{"train_loss": 6.932713508605957, "global_step": 0, "epoch": 0, "lr": 2.0000000000000002e-07}
{"train_loss": 6.9332804679870605, "global_step": 1, "epoch": 0, "lr": 4.0000000000000003e-07}
{"train_loss": 6.933236598968506, "global_step": 2, "epoch": 0, "lr": 6.000000000000001e-07}
{"train_loss": 6.93294620513916, "global_step": 3, "epoch": 0, "lr": 8.000000000000001e-07}
{"train_loss": 6.932689666748047, "global_step": 4, "epoch": 0, "lr": 1.0000000000000002e-06}
{"train_loss": 6.932497024536133, "global_step": 5, "epoch": 0, "lr": 1.2000000000000002e-06}
{"train_loss": 6.933079242706299, "global_step": 6, "epoch": 0, "lr": 1.4000000000000001e-06}
{"train_loss": 6.932825088500977, "global_step": 7, "epoch": 0, "lr": 1.6000000000000001e-06}
{"train_loss": 6.932263374328613, "global_step": 8, "epoch": 0, "lr": 1.8e-06}
{"train_loss": 6.932244777679443, "global_step": 9, "epoch": 0, "lr": 2.0000000000000003e-06}
{"train_loss": 6.932267189025879, "global_step": 10, "epoch": 0, "lr": 2.2e-06}
{"train_loss": 6.931981563568115, "global_step": 11, "epoch": 0, "lr": 2.4000000000000003e-06}
{"train_loss": 6.932559967041016, "global_step": 12, "epoch": 0, "lr": 2.6e-06}
{"train_loss": 6.932189464569092, "global_step": 13, "epoch": 0, "lr": 2.8000000000000003e-06}
{"train_loss": 6.931957244873047, "global_step": 14, "epoch": 0, "lr": 3e-06}
{"train_loss": 6.932117462158203, "global_step": 15, "epoch": 0, "lr": 3.2000000000000003e-06}
{"train_loss": 6.932360649108887, "global_step": 16, "epoch": 0, "lr": 3.4000000000000005e-06}
{"train_loss": 6.931760311126709, "global_step": 17, "epoch": 0, "lr": 3.6e-06}
{"train_loss": 6.931354522705078, "global_step": 18, "epoch": 0, "lr": 3.8e-06}
{"train_loss": 6.931532859802246, "global_step": 19, "epoch": 0, "lr": 4.000000000000001e-06}
{"train_loss": 6.931687355041504, "global_step": 20, "epoch": 0, "lr": 4.2000000000000004e-06}
{"train_loss": 6.931931018829346, "global_step": 21, "epoch": 0, "lr": 4.4e-06}
{"train_loss": 6.931190490722656, "global_step": 22, "epoch": 0, "lr": 4.6e-06}
{"train_loss": 6.9310302734375, "global_step": 23, "epoch": 0, "lr": 4.800000000000001e-06}
{"train_loss": 6.930624008178711, "global_step": 24, "epoch": 0, "lr": 5e-06}
{"train_loss": 6.9302077293396, "global_step": 25, "epoch": 0, "lr": 5.2e-06}
{"train_loss": 6.9304094314575195, "global_step": 26, "epoch": 0, "lr": 5.4e-06}
{"train_loss": 6.930055141448975, "global_step": 27, "epoch": 0, "lr": 5.600000000000001e-06}
{"train_loss": 6.930604934692383, "global_step": 28, "epoch": 0, "lr": 5.8e-06}
{"train_loss": 6.929662704467773, "global_step": 29, "epoch": 0, "lr": 6e-06}
{"train_loss": 6.93073034286499, "global_step": 30, "epoch": 0, "lr": 6.2e-06}
{"train_loss": 6.929998397827148, "global_step": 31, "epoch": 0, "lr": 6.4000000000000006e-06}
{"train_loss": 6.930004119873047, "global_step": 32, "epoch": 0, "lr": 6.6e-06}
{"train_loss": 6.930038928985596, "global_step": 33, "epoch": 0, "lr": 6.800000000000001e-06}
{"train_loss": 6.930491924285889, "global_step": 34, "epoch": 0, "lr": 7.000000000000001e-06}
{"train_loss": 6.929944038391113, "global_step": 35, "epoch": 0, "lr": 7.2e-06}
{"train_loss": 6.930058479309082, "global_step": 36, "epoch": 0, "lr": 7.4e-06}
{"train_loss": 6.929502964019775, "global_step": 37, "epoch": 0, "lr": 7.6e-06}
{"train_loss": 6.929126739501953, "global_step": 38, "epoch": 0, "lr": 7.8e-06}
{"train_loss": 6.92919397354126, "global_step": 39, "epoch": 0, "lr": 8.000000000000001e-06}
{"train_loss": 6.929191589355469, "global_step": 40, "epoch": 0, "lr": 8.200000000000001e-06}
{"train_loss": 6.9298415184021, "global_step": 41, "epoch": 0, "lr": 8.400000000000001e-06}
{"train_loss": 6.927492618560791, "global_step": 42, "epoch": 0, "lr": 8.599999999999999e-06}
{"train_loss": 6.929108142852783, "global_step": 43, "epoch": 0, "lr": 8.8e-06}
{"train_loss": 6.928580284118652, "global_step": 44, "epoch": 0, "lr": 9e-06}
{"train_loss": 6.927523612976074, "global_step": 45, "epoch": 0, "lr": 9.2e-06}
{"train_loss": 6.92759370803833, "global_step": 46, "epoch": 0, "lr": 9.4e-06}
{"train_loss": 6.928363800048828, "global_step": 47, "epoch": 0, "lr": 9.600000000000001e-06}
{"train_loss": 6.926763534545898, "global_step": 48, "epoch": 0, "lr": 9.800000000000001e-06}
{"train_loss": 6.927070140838623, "global_step": 49, "epoch": 0, "lr": 1e-05}
{"train_loss": 6.92661190032959, "global_step": 50, "epoch": 0, "lr": 1.02e-05}
{"train_loss": 6.92678165435791, "global_step": 51, "epoch": 0, "lr": 1.04e-05}
{"train_loss": 6.926735877990723, "global_step": 52, "epoch": 0, "lr": 1.06e-05}
{"train_loss": 6.92642068862915, "global_step": 53, "epoch": 0, "lr": 1.08e-05}
{"train_loss": 6.925541877746582, "global_step": 54, "epoch": 0, "lr": 1.1000000000000001e-05}
{"train_loss": 6.924685478210449, "global_step": 55, "epoch": 0, "lr": 1.1200000000000001e-05}
{"train_loss": 6.925968647003174, "global_step": 56, "epoch": 0, "lr": 1.1400000000000001e-05}
{"train_loss": 6.925690174102783, "global_step": 57, "epoch": 0, "lr": 1.16e-05}
{"train_loss": 6.924768447875977, "global_step": 58, "epoch": 0, "lr": 1.18e-05}
{"train_loss": 6.924370288848877, "global_step": 59, "epoch": 0, "lr": 1.2e-05}
{"train_loss": 6.924038887023926, "global_step": 60, "epoch": 0, "lr": 1.22e-05}
{"train_loss": 6.924880027770996, "global_step": 61, "epoch": 0, "lr": 1.24e-05}
{"train_loss": 6.923833847045898, "global_step": 62, "epoch": 0, "lr": 1.2600000000000001e-05}
{"train_loss": 6.923079967498779, "global_step": 63, "epoch": 0, "lr": 1.2800000000000001e-05}
{"train_loss": 6.923267364501953, "global_step": 64, "epoch": 0, "lr": 1.3000000000000001e-05}
{"train_loss": 6.9227824211120605, "global_step": 65, "epoch": 0, "lr": 1.32e-05}
{"train_loss": 6.921395301818848, "global_step": 66, "epoch": 0, "lr": 1.3400000000000002e-05}
{"train_loss": 6.9211506843566895, "global_step": 67, "epoch": 0, "lr": 1.3600000000000002e-05}
{"train_loss": 6.920882701873779, "global_step": 68, "epoch": 0, "lr": 1.3800000000000002e-05}
{"train_loss": 6.921450614929199, "global_step": 69, "epoch": 0, "lr": 1.4000000000000001e-05}
{"train_loss": 6.918184280395508, "global_step": 70, "epoch": 0, "lr": 1.42e-05}
{"train_loss": 6.918411731719971, "global_step": 71, "epoch": 0, "lr": 1.44e-05}
{"train_loss": 6.918857574462891, "global_step": 72, "epoch": 0, "lr": 1.4599999999999999e-05}
{"train_loss": 6.9177165031433105, "global_step": 73, "epoch": 0, "lr": 1.48e-05}
{"train_loss": 6.917659759521484, "global_step": 74, "epoch": 0, "lr": 1.5e-05}
{"train_loss": 6.91676664352417, "global_step": 75, "epoch": 0, "lr": 1.52e-05}
{"train_loss": 6.915658473968506, "global_step": 76, "epoch": 0, "lr": 1.54e-05}
{"train_loss": 6.912632465362549, "global_step": 77, "epoch": 0, "lr": 1.56e-05}
{"train_loss": 6.914673328399658, "global_step": 78, "epoch": 0, "lr": 1.58e-05}
{"train_loss": 6.912344932556152, "global_step": 79, "epoch": 0, "lr": 1.6000000000000003e-05}
{"train_loss": 6.91166353225708, "global_step": 80, "epoch": 0, "lr": 1.62e-05}
{"train_loss": 6.908838272094727, "global_step": 81, "epoch": 0, "lr": 1.6400000000000002e-05}
{"train_loss": 6.911022186279297, "global_step": 82, "epoch": 0, "lr": 1.66e-05}
{"train_loss": 6.9054274559021, "global_step": 83, "epoch": 0, "lr": 1.6800000000000002e-05}
{"train_loss": 6.90976619720459, "global_step": 84, "epoch": 0, "lr": 1.7000000000000003e-05}
{"train_loss": 6.907440662384033, "global_step": 85, "epoch": 0, "lr": 1.7199999999999998e-05}
{"train_loss": 6.90462589263916, "global_step": 86, "epoch": 0, "lr": 1.74e-05}
{"train_loss": 6.902275085449219, "global_step": 87, "epoch": 0, "lr": 1.76e-05}
{"train_loss": 6.925200773089119, "global_step": 88, "epoch": 0, "lr": 1.78e-05, "train/sim_max_reward_0": 0.19715507366317334, "train/sim_max_reward_1": 0.0006754812217992413, "train/sim_max_reward_2": 0.0, "train/sim_max_reward_3": 0.013004569904596303, "train/sim_max_reward_4": 0.2368928228562067, "train/sim_max_reward_5": 0.25415972548617266, "test/sim_max_reward_4500000": 0.31897924536482425, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.001967323988682413, "test/sim_max_reward_4500004": 0.054580065317177094, "test/sim_max_reward_4500005": 0.14543091411652173, "test/sim_max_reward_4500006": 0.0, "test/sim_max_reward_4500007": 0.0, "test/sim_max_reward_4500008": 0.0, "test/sim_max_reward_4500009": 0.0, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.08528357476338054, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.0, "test/sim_max_reward_4500014": 0.16561182130584123, "test/sim_max_reward_4500015": 0.0, "test/sim_max_reward_4500016": 0.017544158064586216, "test/sim_max_reward_4500017": 0.6157525208066361, "test/sim_max_reward_4500018": 0.004249031229020703, "test/sim_max_reward_4500019": 0.39496762743265906, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.21685804941338724, "test/sim_max_reward_4500022": 0.3915478951530142, "test/sim_max_reward_4500023": 0.431168964091752, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.0, "test/sim_max_reward_4500026": 0.0, "test/sim_max_reward_4500027": 0.0, "test/sim_max_reward_4500028": 0.4324727698166609, "test/sim_max_reward_4500029": 0.06904068944137526, "test/sim_max_reward_4500030": 0.18881693657457926, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.33172781418891445, "test/sim_max_reward_4500033": 0.00012312537725908565, "test/sim_max_reward_4500034": 0.04882485170691779, "test/sim_max_reward_4500035": 0.0779606034003907, "test/sim_max_reward_4500036": 0.3097316393341177, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.07106118112574276, "test/sim_max_reward_4500039": 0.0, "test/sim_max_reward_4500040": 0.15243166474591124, "test/sim_max_reward_4500041": 0.44277847587000685, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.0, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.1175735103924684, "test/sim_max_reward_4500046": 0.45770897074511957, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.0, "test/sim_max_reward_4500049": 0.046842731371186934, "train/mean_score": 0.11698127885532472, "test/mean_score": 0.11182072310276267, "val_loss": 6.896783351898193, "train_action_mse_error": 31718.15625}
{"train_loss": 6.901289939880371, "global_step": 89, "epoch": 1, "lr": 1.8e-05}
{"train_loss": 6.895013809204102, "global_step": 90, "epoch": 1, "lr": 1.8200000000000002e-05}
{"train_loss": 6.896984100341797, "global_step": 91, "epoch": 1, "lr": 1.84e-05}
{"train_loss": 6.893044471740723, "global_step": 92, "epoch": 1, "lr": 1.86e-05}
{"train_loss": 6.894161224365234, "global_step": 93, "epoch": 1, "lr": 1.88e-05}
{"train_loss": 6.893507480621338, "global_step": 94, "epoch": 1, "lr": 1.9e-05}
{"train_loss": 6.889883995056152, "global_step": 95, "epoch": 1, "lr": 1.9200000000000003e-05}
{"train_loss": 6.887935638427734, "global_step": 96, "epoch": 1, "lr": 1.94e-05}
{"train_loss": 6.885251045227051, "global_step": 97, "epoch": 1, "lr": 1.9600000000000002e-05}
{"train_loss": 6.8819475173950195, "global_step": 98, "epoch": 1, "lr": 1.9800000000000004e-05}
{"train_loss": 6.880340099334717, "global_step": 99, "epoch": 1, "lr": 2e-05}
{"train_loss": 6.884303092956543, "global_step": 100, "epoch": 1, "lr": 2.0200000000000003e-05}
{"train_loss": 6.867624759674072, "global_step": 101, "epoch": 1, "lr": 2.04e-05}
{"train_loss": 6.868465423583984, "global_step": 102, "epoch": 1, "lr": 2.06e-05}
{"train_loss": 6.864530563354492, "global_step": 103, "epoch": 1, "lr": 2.08e-05}
{"train_loss": 6.864625930786133, "global_step": 104, "epoch": 1, "lr": 2.1e-05}
{"train_loss": 6.8590087890625, "global_step": 105, "epoch": 1, "lr": 2.12e-05}
{"train_loss": 6.861553192138672, "global_step": 106, "epoch": 1, "lr": 2.1400000000000002e-05}
{"train_loss": 6.859845161437988, "global_step": 107, "epoch": 1, "lr": 2.16e-05}
{"train_loss": 6.851108551025391, "global_step": 108, "epoch": 1, "lr": 2.18e-05}
{"train_loss": 6.844099521636963, "global_step": 109, "epoch": 1, "lr": 2.2000000000000003e-05}
{"train_loss": 6.834292888641357, "global_step": 110, "epoch": 1, "lr": 2.22e-05}
{"train_loss": 6.836869239807129, "global_step": 111, "epoch": 1, "lr": 2.2400000000000002e-05}
{"train_loss": 6.821143627166748, "global_step": 112, "epoch": 1, "lr": 2.26e-05}
{"train_loss": 6.823004722595215, "global_step": 113, "epoch": 1, "lr": 2.2800000000000002e-05}
{"train_loss": 6.8175859451293945, "global_step": 114, "epoch": 1, "lr": 2.3000000000000003e-05}
{"train_loss": 6.811854839324951, "global_step": 115, "epoch": 1, "lr": 2.32e-05}
{"train_loss": 6.798521995544434, "global_step": 116, "epoch": 1, "lr": 2.3400000000000003e-05}
{"train_loss": 6.815557956695557, "global_step": 117, "epoch": 1, "lr": 2.36e-05}
{"train_loss": 6.802323341369629, "global_step": 118, "epoch": 1, "lr": 2.38e-05}
{"train_loss": 6.7985734939575195, "global_step": 119, "epoch": 1, "lr": 2.4e-05}
{"train_loss": 6.768887996673584, "global_step": 120, "epoch": 1, "lr": 2.4200000000000002e-05}
{"train_loss": 6.762290000915527, "global_step": 121, "epoch": 1, "lr": 2.44e-05}
{"train_loss": 6.7596755027771, "global_step": 122, "epoch": 1, "lr": 2.46e-05}
{"train_loss": 6.757098197937012, "global_step": 123, "epoch": 1, "lr": 2.48e-05}
{"train_loss": 6.7499098777771, "global_step": 124, "epoch": 1, "lr": 2.5e-05}
{"train_loss": 6.744334697723389, "global_step": 125, "epoch": 1, "lr": 2.5200000000000003e-05}
{"train_loss": 6.739533424377441, "global_step": 126, "epoch": 1, "lr": 2.54e-05}
{"train_loss": 6.709694862365723, "global_step": 127, "epoch": 1, "lr": 2.5600000000000002e-05}
{"train_loss": 6.7132768630981445, "global_step": 128, "epoch": 1, "lr": 2.58e-05}
{"train_loss": 6.691123008728027, "global_step": 129, "epoch": 1, "lr": 2.6000000000000002e-05}
{"train_loss": 6.687469005584717, "global_step": 130, "epoch": 1, "lr": 2.6200000000000003e-05}
{"train_loss": 6.678691387176514, "global_step": 131, "epoch": 1, "lr": 2.64e-05}
{"train_loss": 6.676074504852295, "global_step": 132, "epoch": 1, "lr": 2.6600000000000003e-05}
{"train_loss": 6.662917613983154, "global_step": 133, "epoch": 1, "lr": 2.6800000000000004e-05}
{"train_loss": 6.628371238708496, "global_step": 134, "epoch": 1, "lr": 2.7000000000000002e-05}
{"train_loss": 6.638820171356201, "global_step": 135, "epoch": 1, "lr": 2.7200000000000004e-05}
{"train_loss": 6.587611675262451, "global_step": 136, "epoch": 1, "lr": 2.7400000000000002e-05}
{"train_loss": 6.561396598815918, "global_step": 137, "epoch": 1, "lr": 2.7600000000000003e-05}
{"train_loss": 6.563365459442139, "global_step": 138, "epoch": 1, "lr": 2.7800000000000005e-05}
{"train_loss": 6.543600559234619, "global_step": 139, "epoch": 1, "lr": 2.8000000000000003e-05}
{"train_loss": 6.545587062835693, "global_step": 140, "epoch": 1, "lr": 2.8199999999999998e-05}
{"train_loss": 6.520416259765625, "global_step": 141, "epoch": 1, "lr": 2.84e-05}
{"train_loss": 6.500746250152588, "global_step": 142, "epoch": 1, "lr": 2.86e-05}
{"train_loss": 6.4718122482299805, "global_step": 143, "epoch": 1, "lr": 2.88e-05}
{"train_loss": 6.441760063171387, "global_step": 144, "epoch": 1, "lr": 2.9e-05}
{"train_loss": 6.404004096984863, "global_step": 145, "epoch": 1, "lr": 2.9199999999999998e-05}
{"train_loss": 6.403056621551514, "global_step": 146, "epoch": 1, "lr": 2.94e-05}
{"train_loss": 6.368744850158691, "global_step": 147, "epoch": 1, "lr": 2.96e-05}
{"train_loss": 6.3305182456970215, "global_step": 148, "epoch": 1, "lr": 2.98e-05}
{"train_loss": 6.289916038513184, "global_step": 149, "epoch": 1, "lr": 3e-05}
{"train_loss": 6.3141703605651855, "global_step": 150, "epoch": 1, "lr": 3.02e-05}
{"train_loss": 6.3133063316345215, "global_step": 151, "epoch": 1, "lr": 3.04e-05}
{"train_loss": 6.215432167053223, "global_step": 152, "epoch": 1, "lr": 3.06e-05}
{"train_loss": 6.230551719665527, "global_step": 153, "epoch": 1, "lr": 3.08e-05}
{"train_loss": 6.15749979019165, "global_step": 154, "epoch": 1, "lr": 3.1e-05}
{"train_loss": 6.157609462738037, "global_step": 155, "epoch": 1, "lr": 3.12e-05}
{"train_loss": 6.088020324707031, "global_step": 156, "epoch": 1, "lr": 3.1400000000000004e-05}
{"train_loss": 6.081495761871338, "global_step": 157, "epoch": 1, "lr": 3.16e-05}
{"train_loss": 6.049703121185303, "global_step": 158, "epoch": 1, "lr": 3.18e-05}
{"train_loss": 6.0071001052856445, "global_step": 159, "epoch": 1, "lr": 3.2000000000000005e-05}
{"train_loss": 5.986988544464111, "global_step": 160, "epoch": 1, "lr": 3.2200000000000003e-05}
{"train_loss": 5.939577579498291, "global_step": 161, "epoch": 1, "lr": 3.24e-05}
{"train_loss": 5.869651794433594, "global_step": 162, "epoch": 1, "lr": 3.26e-05}
{"train_loss": 5.869181156158447, "global_step": 163, "epoch": 1, "lr": 3.2800000000000004e-05}
{"train_loss": 5.771517753601074, "global_step": 164, "epoch": 1, "lr": 3.3e-05}
{"train_loss": 5.760497570037842, "global_step": 165, "epoch": 1, "lr": 3.32e-05}
{"train_loss": 5.682750225067139, "global_step": 166, "epoch": 1, "lr": 3.3400000000000005e-05}
{"train_loss": 5.68614387512207, "global_step": 167, "epoch": 1, "lr": 3.3600000000000004e-05}
{"train_loss": 5.6222639083862305, "global_step": 168, "epoch": 1, "lr": 3.38e-05}
{"train_loss": 5.56826114654541, "global_step": 169, "epoch": 1, "lr": 3.4000000000000007e-05}
{"train_loss": 5.509812831878662, "global_step": 170, "epoch": 1, "lr": 3.4200000000000005e-05}
{"train_loss": 5.443772792816162, "global_step": 171, "epoch": 1, "lr": 3.4399999999999996e-05}
{"train_loss": 5.45237922668457, "global_step": 172, "epoch": 1, "lr": 3.46e-05}
{"train_loss": 5.399904251098633, "global_step": 173, "epoch": 1, "lr": 3.48e-05}
{"train_loss": 5.346742630004883, "global_step": 174, "epoch": 1, "lr": 3.5e-05}
{"train_loss": 5.291194915771484, "global_step": 175, "epoch": 1, "lr": 3.52e-05}
{"train_loss": 5.236894130706787, "global_step": 176, "epoch": 1, "lr": 3.54e-05}
{"train_loss": 6.434324660997712, "global_step": 177, "epoch": 1, "lr": 3.56e-05, "val_loss": 4.985713481903076}
{"train_loss": 5.110969543457031, "global_step": 178, "epoch": 2, "lr": 3.58e-05}
{"train_loss": 5.118423938751221, "global_step": 179, "epoch": 2, "lr": 3.6e-05}
{"train_loss": 5.0419535636901855, "global_step": 180, "epoch": 2, "lr": 3.62e-05}
{"train_loss": 4.962034225463867, "global_step": 181, "epoch": 2, "lr": 3.6400000000000004e-05}
{"train_loss": 4.939216136932373, "global_step": 182, "epoch": 2, "lr": 3.66e-05}
{"train_loss": 4.906742572784424, "global_step": 183, "epoch": 2, "lr": 3.68e-05}
{"train_loss": 4.887089729309082, "global_step": 184, "epoch": 2, "lr": 3.7e-05}
{"train_loss": 4.77922248840332, "global_step": 185, "epoch": 2, "lr": 3.72e-05}
{"train_loss": 4.754168510437012, "global_step": 186, "epoch": 2, "lr": 3.74e-05}
{"train_loss": 4.6817946434021, "global_step": 187, "epoch": 2, "lr": 3.76e-05}
{"train_loss": 4.615882873535156, "global_step": 188, "epoch": 2, "lr": 3.7800000000000004e-05}
{"train_loss": 4.589574813842773, "global_step": 189, "epoch": 2, "lr": 3.8e-05}
{"train_loss": 4.547879695892334, "global_step": 190, "epoch": 2, "lr": 3.82e-05}
{"train_loss": 4.488785266876221, "global_step": 191, "epoch": 2, "lr": 3.8400000000000005e-05}
{"train_loss": 4.448060512542725, "global_step": 192, "epoch": 2, "lr": 3.86e-05}
{"train_loss": 4.419384956359863, "global_step": 193, "epoch": 2, "lr": 3.88e-05}
{"train_loss": 4.466739654541016, "global_step": 194, "epoch": 2, "lr": 3.9000000000000006e-05}
{"train_loss": 4.348387241363525, "global_step": 195, "epoch": 2, "lr": 3.9200000000000004e-05}
{"train_loss": 4.28823184967041, "global_step": 196, "epoch": 2, "lr": 3.94e-05}
{"train_loss": 4.288362503051758, "global_step": 197, "epoch": 2, "lr": 3.960000000000001e-05}
{"train_loss": 4.239752292633057, "global_step": 198, "epoch": 2, "lr": 3.9800000000000005e-05}
{"train_loss": 4.193582057952881, "global_step": 199, "epoch": 2, "lr": 4e-05}
{"train_loss": 4.180926322937012, "global_step": 200, "epoch": 2, "lr": 4.02e-05}
{"train_loss": 4.26678466796875, "global_step": 201, "epoch": 2, "lr": 4.0400000000000006e-05}
{"train_loss": 4.103592395782471, "global_step": 202, "epoch": 2, "lr": 4.0600000000000004e-05}
{"train_loss": 4.11705207824707, "global_step": 203, "epoch": 2, "lr": 4.08e-05}
{"train_loss": 4.01198148727417, "global_step": 204, "epoch": 2, "lr": 4.1e-05}
{"train_loss": 4.072502136230469, "global_step": 205, "epoch": 2, "lr": 4.12e-05}
{"train_loss": 4.0159912109375, "global_step": 206, "epoch": 2, "lr": 4.14e-05}
{"train_loss": 3.947593927383423, "global_step": 207, "epoch": 2, "lr": 4.16e-05}
{"train_loss": 4.055458068847656, "global_step": 208, "epoch": 2, "lr": 4.18e-05}
{"train_loss": 3.8956542015075684, "global_step": 209, "epoch": 2, "lr": 4.2e-05}
{"train_loss": 3.992985248565674, "global_step": 210, "epoch": 2, "lr": 4.22e-05}
{"train_loss": 3.9264326095581055, "global_step": 211, "epoch": 2, "lr": 4.24e-05}
{"train_loss": 3.8947412967681885, "global_step": 212, "epoch": 2, "lr": 4.26e-05}
{"train_loss": 4.064750671386719, "global_step": 213, "epoch": 2, "lr": 4.2800000000000004e-05}
{"train_loss": 4.041012287139893, "global_step": 214, "epoch": 2, "lr": 4.3e-05}
{"train_loss": 3.821971893310547, "global_step": 215, "epoch": 2, "lr": 4.32e-05}
{"train_loss": 3.7720470428466797, "global_step": 216, "epoch": 2, "lr": 4.3400000000000005e-05}
{"train_loss": 3.928957462310791, "global_step": 217, "epoch": 2, "lr": 4.36e-05}
{"train_loss": 3.755373954772949, "global_step": 218, "epoch": 2, "lr": 4.38e-05}
{"train_loss": 3.841949939727783, "global_step": 219, "epoch": 2, "lr": 4.4000000000000006e-05}
{"train_loss": 3.8477039337158203, "global_step": 220, "epoch": 2, "lr": 4.4200000000000004e-05}
{"train_loss": 3.785360813140869, "global_step": 221, "epoch": 2, "lr": 4.44e-05}
{"train_loss": 3.8250465393066406, "global_step": 222, "epoch": 2, "lr": 4.46e-05}
{"train_loss": 3.878143310546875, "global_step": 223, "epoch": 2, "lr": 4.4800000000000005e-05}
{"train_loss": 3.917022228240967, "global_step": 224, "epoch": 2, "lr": 4.5e-05}
{"train_loss": 3.755871534347534, "global_step": 225, "epoch": 2, "lr": 4.52e-05}
{"train_loss": 3.905311107635498, "global_step": 226, "epoch": 2, "lr": 4.5400000000000006e-05}
{"train_loss": 3.9290478229522705, "global_step": 227, "epoch": 2, "lr": 4.5600000000000004e-05}
{"train_loss": 3.719106674194336, "global_step": 228, "epoch": 2, "lr": 4.58e-05}
{"train_loss": 3.714280843734741, "global_step": 229, "epoch": 2, "lr": 4.600000000000001e-05}
{"train_loss": 3.722993850708008, "global_step": 230, "epoch": 2, "lr": 4.6200000000000005e-05}
{"train_loss": 3.854861259460449, "global_step": 231, "epoch": 2, "lr": 4.64e-05}
{"train_loss": 3.7502593994140625, "global_step": 232, "epoch": 2, "lr": 4.660000000000001e-05}
{"train_loss": 3.6783447265625, "global_step": 233, "epoch": 2, "lr": 4.6800000000000006e-05}
{"train_loss": 3.6824591159820557, "global_step": 234, "epoch": 2, "lr": 4.7e-05}
{"train_loss": 3.565603017807007, "global_step": 235, "epoch": 2, "lr": 4.72e-05}
{"train_loss": 3.728410005569458, "global_step": 236, "epoch": 2, "lr": 4.74e-05}
{"train_loss": 3.9171409606933594, "global_step": 237, "epoch": 2, "lr": 4.76e-05}
{"train_loss": 3.714298963546753, "global_step": 238, "epoch": 2, "lr": 4.78e-05}
{"train_loss": 3.599628448486328, "global_step": 239, "epoch": 2, "lr": 4.8e-05}
{"train_loss": 3.886049747467041, "global_step": 240, "epoch": 2, "lr": 4.82e-05}
{"train_loss": 3.591330051422119, "global_step": 241, "epoch": 2, "lr": 4.8400000000000004e-05}
{"train_loss": 3.7040698528289795, "global_step": 242, "epoch": 2, "lr": 4.86e-05}
{"train_loss": 3.6148335933685303, "global_step": 243, "epoch": 2, "lr": 4.88e-05}
{"train_loss": 3.7161314487457275, "global_step": 244, "epoch": 2, "lr": 4.9e-05}
{"train_loss": 3.5673575401306152, "global_step": 245, "epoch": 2, "lr": 4.92e-05}
{"train_loss": 3.5882315635681152, "global_step": 246, "epoch": 2, "lr": 4.94e-05}
{"train_loss": 3.961446762084961, "global_step": 247, "epoch": 2, "lr": 4.96e-05}
{"train_loss": 3.6886701583862305, "global_step": 248, "epoch": 2, "lr": 4.9800000000000004e-05}
{"train_loss": 3.482085704803467, "global_step": 249, "epoch": 2, "lr": 5e-05}
{"train_loss": 3.5940444469451904, "global_step": 250, "epoch": 2, "lr": 5.02e-05}
{"train_loss": 3.58577823638916, "global_step": 251, "epoch": 2, "lr": 5.0400000000000005e-05}
{"train_loss": 3.639737367630005, "global_step": 252, "epoch": 2, "lr": 5.0600000000000003e-05}
{"train_loss": 3.625143051147461, "global_step": 253, "epoch": 2, "lr": 5.08e-05}
{"train_loss": 3.6218271255493164, "global_step": 254, "epoch": 2, "lr": 5.1000000000000006e-05}
{"train_loss": 3.7466280460357666, "global_step": 255, "epoch": 2, "lr": 5.1200000000000004e-05}
{"train_loss": 3.8492939472198486, "global_step": 256, "epoch": 2, "lr": 5.14e-05}
{"train_loss": 3.6616172790527344, "global_step": 257, "epoch": 2, "lr": 5.16e-05}
{"train_loss": 3.5423104763031006, "global_step": 258, "epoch": 2, "lr": 5.1800000000000005e-05}
{"train_loss": 3.775052547454834, "global_step": 259, "epoch": 2, "lr": 5.2000000000000004e-05}
{"train_loss": 3.5655946731567383, "global_step": 260, "epoch": 2, "lr": 5.22e-05}
{"train_loss": 3.5529165267944336, "global_step": 261, "epoch": 2, "lr": 5.2400000000000007e-05}
{"train_loss": 3.7007055282592773, "global_step": 262, "epoch": 2, "lr": 5.2600000000000005e-05}
{"train_loss": 3.8508503437042236, "global_step": 263, "epoch": 2, "lr": 5.28e-05}
{"train_loss": 3.5070834159851074, "global_step": 264, "epoch": 2, "lr": 5.300000000000001e-05}
{"train_loss": 3.6441867351531982, "global_step": 265, "epoch": 2, "lr": 5.3200000000000006e-05}
{"train_loss": 4.00457879666532, "global_step": 266, "epoch": 2, "lr": 5.3400000000000004e-05, "val_loss": 3.497783660888672}
{"train_loss": 3.808495044708252, "global_step": 267, "epoch": 3, "lr": 5.360000000000001e-05}
{"train_loss": 3.7620697021484375, "global_step": 268, "epoch": 3, "lr": 5.380000000000001e-05}
{"train_loss": 3.596808910369873, "global_step": 269, "epoch": 3, "lr": 5.4000000000000005e-05}
{"train_loss": 3.473815679550171, "global_step": 270, "epoch": 3, "lr": 5.420000000000001e-05}
{"train_loss": 3.561655282974243, "global_step": 271, "epoch": 3, "lr": 5.440000000000001e-05}
{"train_loss": 3.333218812942505, "global_step": 272, "epoch": 3, "lr": 5.4600000000000006e-05}
{"train_loss": 3.4456679821014404, "global_step": 273, "epoch": 3, "lr": 5.4800000000000004e-05}
{"train_loss": 3.4894845485687256, "global_step": 274, "epoch": 3, "lr": 5.500000000000001e-05}
{"train_loss": 3.4709830284118652, "global_step": 275, "epoch": 3, "lr": 5.520000000000001e-05}
{"train_loss": 3.5813748836517334, "global_step": 276, "epoch": 3, "lr": 5.5400000000000005e-05}
{"train_loss": 3.5551748275756836, "global_step": 277, "epoch": 3, "lr": 5.560000000000001e-05}
{"train_loss": 3.380336284637451, "global_step": 278, "epoch": 3, "lr": 5.580000000000001e-05}
{"train_loss": 3.432873010635376, "global_step": 279, "epoch": 3, "lr": 5.6000000000000006e-05}
{"train_loss": 3.6559083461761475, "global_step": 280, "epoch": 3, "lr": 5.620000000000001e-05}
{"train_loss": 3.497553825378418, "global_step": 281, "epoch": 3, "lr": 5.6399999999999995e-05}
{"train_loss": 3.7266180515289307, "global_step": 282, "epoch": 3, "lr": 5.66e-05}
{"train_loss": 3.491446018218994, "global_step": 283, "epoch": 3, "lr": 5.68e-05}
{"train_loss": 3.590245485305786, "global_step": 284, "epoch": 3, "lr": 5.6999999999999996e-05}
{"train_loss": 3.7260375022888184, "global_step": 285, "epoch": 3, "lr": 5.72e-05}
{"train_loss": 3.621433973312378, "global_step": 286, "epoch": 3, "lr": 5.74e-05}
{"train_loss": 3.68105411529541, "global_step": 287, "epoch": 3, "lr": 5.76e-05}
{"train_loss": 3.594956636428833, "global_step": 288, "epoch": 3, "lr": 5.7799999999999995e-05}
{"train_loss": 3.3932454586029053, "global_step": 289, "epoch": 3, "lr": 5.8e-05}
{"train_loss": 3.56453800201416, "global_step": 290, "epoch": 3, "lr": 5.82e-05}
{"train_loss": 3.424461841583252, "global_step": 291, "epoch": 3, "lr": 5.8399999999999997e-05}
{"train_loss": 3.5377964973449707, "global_step": 292, "epoch": 3, "lr": 5.86e-05}
{"train_loss": 3.6443607807159424, "global_step": 293, "epoch": 3, "lr": 5.88e-05}
{"train_loss": 3.5812735557556152, "global_step": 294, "epoch": 3, "lr": 5.9e-05}
{"train_loss": 3.5827488899230957, "global_step": 295, "epoch": 3, "lr": 5.92e-05}
{"train_loss": 3.5130882263183594, "global_step": 296, "epoch": 3, "lr": 5.94e-05}
{"train_loss": 3.408066987991333, "global_step": 297, "epoch": 3, "lr": 5.96e-05}
{"train_loss": 3.3960437774658203, "global_step": 298, "epoch": 3, "lr": 5.9800000000000003e-05}
{"train_loss": 3.452132225036621, "global_step": 299, "epoch": 3, "lr": 6e-05}
{"train_loss": 3.5207650661468506, "global_step": 300, "epoch": 3, "lr": 6.02e-05}
{"train_loss": 3.404703378677368, "global_step": 301, "epoch": 3, "lr": 6.04e-05}
{"train_loss": 3.590362548828125, "global_step": 302, "epoch": 3, "lr": 6.06e-05}
{"train_loss": 3.3403196334838867, "global_step": 303, "epoch": 3, "lr": 6.08e-05}
{"train_loss": 3.3296866416931152, "global_step": 304, "epoch": 3, "lr": 6.1e-05}
{"train_loss": 3.564826488494873, "global_step": 305, "epoch": 3, "lr": 6.12e-05}
{"train_loss": 3.5086112022399902, "global_step": 306, "epoch": 3, "lr": 6.14e-05}
{"train_loss": 3.487783432006836, "global_step": 307, "epoch": 3, "lr": 6.16e-05}
{"train_loss": 3.4860925674438477, "global_step": 308, "epoch": 3, "lr": 6.18e-05}
{"train_loss": 3.634716033935547, "global_step": 309, "epoch": 3, "lr": 6.2e-05}
{"train_loss": 3.577261447906494, "global_step": 310, "epoch": 3, "lr": 6.220000000000001e-05}
{"train_loss": 3.4476630687713623, "global_step": 311, "epoch": 3, "lr": 6.24e-05}
{"train_loss": 3.5075106620788574, "global_step": 312, "epoch": 3, "lr": 6.26e-05}
{"train_loss": 3.6588430404663086, "global_step": 313, "epoch": 3, "lr": 6.280000000000001e-05}
{"train_loss": 3.520570755004883, "global_step": 314, "epoch": 3, "lr": 6.3e-05}
{"train_loss": 3.4603323936462402, "global_step": 315, "epoch": 3, "lr": 6.32e-05}
{"train_loss": 3.54449725151062, "global_step": 316, "epoch": 3, "lr": 6.340000000000001e-05}
{"train_loss": 3.3740806579589844, "global_step": 317, "epoch": 3, "lr": 6.36e-05}
{"train_loss": 3.3499350547790527, "global_step": 318, "epoch": 3, "lr": 6.38e-05}
{"train_loss": 3.259657144546509, "global_step": 319, "epoch": 3, "lr": 6.400000000000001e-05}
{"train_loss": 3.3944835662841797, "global_step": 320, "epoch": 3, "lr": 6.42e-05}
{"train_loss": 3.420476198196411, "global_step": 321, "epoch": 3, "lr": 6.440000000000001e-05}
{"train_loss": 3.494746208190918, "global_step": 322, "epoch": 3, "lr": 6.460000000000001e-05}
{"train_loss": 3.4527504444122314, "global_step": 323, "epoch": 3, "lr": 6.48e-05}
{"train_loss": 3.4151995182037354, "global_step": 324, "epoch": 3, "lr": 6.500000000000001e-05}
{"train_loss": 3.620087146759033, "global_step": 325, "epoch": 3, "lr": 6.52e-05}
{"train_loss": 3.5138087272644043, "global_step": 326, "epoch": 3, "lr": 6.54e-05}
{"train_loss": 3.4859752655029297, "global_step": 327, "epoch": 3, "lr": 6.560000000000001e-05}
{"train_loss": 3.10497784614563, "global_step": 328, "epoch": 3, "lr": 6.58e-05}
{"train_loss": 3.4496307373046875, "global_step": 329, "epoch": 3, "lr": 6.6e-05}
{"train_loss": 3.6144654750823975, "global_step": 330, "epoch": 3, "lr": 6.620000000000001e-05}
{"train_loss": 3.4542531967163086, "global_step": 331, "epoch": 3, "lr": 6.64e-05}
{"train_loss": 3.352055788040161, "global_step": 332, "epoch": 3, "lr": 6.66e-05}
{"train_loss": 3.665945529937744, "global_step": 333, "epoch": 3, "lr": 6.680000000000001e-05}
{"train_loss": 3.395611524581909, "global_step": 334, "epoch": 3, "lr": 6.7e-05}
{"train_loss": 3.488248825073242, "global_step": 335, "epoch": 3, "lr": 6.720000000000001e-05}
{"train_loss": 3.3632888793945312, "global_step": 336, "epoch": 3, "lr": 6.740000000000001e-05}
{"train_loss": 3.6207597255706787, "global_step": 337, "epoch": 3, "lr": 6.76e-05}
{"train_loss": 3.54697585105896, "global_step": 338, "epoch": 3, "lr": 6.780000000000001e-05}
{"train_loss": 3.55722975730896, "global_step": 339, "epoch": 3, "lr": 6.800000000000001e-05}
{"train_loss": 3.3327672481536865, "global_step": 340, "epoch": 3, "lr": 6.82e-05}
{"train_loss": 3.3649840354919434, "global_step": 341, "epoch": 3, "lr": 6.840000000000001e-05}
{"train_loss": 3.447554111480713, "global_step": 342, "epoch": 3, "lr": 6.860000000000001e-05}
{"train_loss": 3.3031513690948486, "global_step": 343, "epoch": 3, "lr": 6.879999999999999e-05}
{"train_loss": 3.483933448791504, "global_step": 344, "epoch": 3, "lr": 6.9e-05}
{"train_loss": 3.4282402992248535, "global_step": 345, "epoch": 3, "lr": 6.92e-05}
{"train_loss": 3.380525588989258, "global_step": 346, "epoch": 3, "lr": 6.939999999999999e-05}
{"train_loss": 3.373134136199951, "global_step": 347, "epoch": 3, "lr": 6.96e-05}
{"train_loss": 3.196638822555542, "global_step": 348, "epoch": 3, "lr": 6.98e-05}
{"train_loss": 3.474191427230835, "global_step": 349, "epoch": 3, "lr": 7e-05}
{"train_loss": 3.43233585357666, "global_step": 350, "epoch": 3, "lr": 7.02e-05}
{"train_loss": 3.2362935543060303, "global_step": 351, "epoch": 3, "lr": 7.04e-05}
{"train_loss": 3.69864821434021, "global_step": 352, "epoch": 3, "lr": 7.06e-05}
{"train_loss": 3.4009222984313965, "global_step": 353, "epoch": 3, "lr": 7.08e-05}
{"train_loss": 3.3985631465911865, "global_step": 354, "epoch": 3, "lr": 7.1e-05}
{"train_loss": 3.4842424767740656, "global_step": 355, "epoch": 3, "lr": 7.12e-05, "val_loss": 3.2264442443847656}
{"train_loss": 3.226665735244751, "global_step": 356, "epoch": 4, "lr": 7.14e-05}
{"train_loss": 3.6124956607818604, "global_step": 357, "epoch": 4, "lr": 7.16e-05}
{"train_loss": 3.217254400253296, "global_step": 358, "epoch": 4, "lr": 7.18e-05}
{"train_loss": 3.358743906021118, "global_step": 359, "epoch": 4, "lr": 7.2e-05}
{"train_loss": 3.3305349349975586, "global_step": 360, "epoch": 4, "lr": 7.22e-05}
{"train_loss": 3.429987907409668, "global_step": 361, "epoch": 4, "lr": 7.24e-05}
{"train_loss": 3.53576922416687, "global_step": 362, "epoch": 4, "lr": 7.26e-05}
{"train_loss": 3.3609728813171387, "global_step": 363, "epoch": 4, "lr": 7.280000000000001e-05}
{"train_loss": 3.2203421592712402, "global_step": 364, "epoch": 4, "lr": 7.3e-05}
{"train_loss": 3.425567626953125, "global_step": 365, "epoch": 4, "lr": 7.32e-05}
{"train_loss": 3.27604079246521, "global_step": 366, "epoch": 4, "lr": 7.340000000000001e-05}
{"train_loss": 3.204584836959839, "global_step": 367, "epoch": 4, "lr": 7.36e-05}
{"train_loss": 3.2941625118255615, "global_step": 368, "epoch": 4, "lr": 7.38e-05}
{"train_loss": 3.4362874031066895, "global_step": 369, "epoch": 4, "lr": 7.4e-05}
{"train_loss": 3.253199815750122, "global_step": 370, "epoch": 4, "lr": 7.42e-05}
{"train_loss": 3.4178988933563232, "global_step": 371, "epoch": 4, "lr": 7.44e-05}
{"train_loss": 3.2524313926696777, "global_step": 372, "epoch": 4, "lr": 7.46e-05}
{"train_loss": 3.35612416267395, "global_step": 373, "epoch": 4, "lr": 7.48e-05}
{"train_loss": 3.260645866394043, "global_step": 374, "epoch": 4, "lr": 7.500000000000001e-05}
{"train_loss": 3.202223062515259, "global_step": 375, "epoch": 4, "lr": 7.52e-05}
{"train_loss": 3.1422955989837646, "global_step": 376, "epoch": 4, "lr": 7.54e-05}
{"train_loss": 3.2930586338043213, "global_step": 377, "epoch": 4, "lr": 7.560000000000001e-05}
{"train_loss": 3.2313709259033203, "global_step": 378, "epoch": 4, "lr": 7.58e-05}
{"train_loss": 3.3405768871307373, "global_step": 379, "epoch": 4, "lr": 7.6e-05}
{"train_loss": 3.3523378372192383, "global_step": 380, "epoch": 4, "lr": 7.620000000000001e-05}
{"train_loss": 3.37160325050354, "global_step": 381, "epoch": 4, "lr": 7.64e-05}
{"train_loss": 3.377549648284912, "global_step": 382, "epoch": 4, "lr": 7.66e-05}
{"train_loss": 3.3805410861968994, "global_step": 383, "epoch": 4, "lr": 7.680000000000001e-05}
{"train_loss": 3.2443618774414062, "global_step": 384, "epoch": 4, "lr": 7.7e-05}
{"train_loss": 3.2817094326019287, "global_step": 385, "epoch": 4, "lr": 7.72e-05}
{"train_loss": 3.1946780681610107, "global_step": 386, "epoch": 4, "lr": 7.740000000000001e-05}
{"train_loss": 3.1241095066070557, "global_step": 387, "epoch": 4, "lr": 7.76e-05}
{"train_loss": 3.4697799682617188, "global_step": 388, "epoch": 4, "lr": 7.780000000000001e-05}
{"train_loss": 3.1675448417663574, "global_step": 389, "epoch": 4, "lr": 7.800000000000001e-05}
{"train_loss": 3.2038044929504395, "global_step": 390, "epoch": 4, "lr": 7.82e-05}
{"train_loss": 2.9655065536499023, "global_step": 391, "epoch": 4, "lr": 7.840000000000001e-05}
{"train_loss": 3.188283920288086, "global_step": 392, "epoch": 4, "lr": 7.860000000000001e-05}
{"train_loss": 3.4252195358276367, "global_step": 393, "epoch": 4, "lr": 7.88e-05}
{"train_loss": 3.3656301498413086, "global_step": 394, "epoch": 4, "lr": 7.900000000000001e-05}
{"train_loss": 3.017064094543457, "global_step": 395, "epoch": 4, "lr": 7.920000000000001e-05}
{"train_loss": 3.2177772521972656, "global_step": 396, "epoch": 4, "lr": 7.94e-05}
{"train_loss": 3.3190643787384033, "global_step": 397, "epoch": 4, "lr": 7.960000000000001e-05}
{"train_loss": 3.2152583599090576, "global_step": 398, "epoch": 4, "lr": 7.98e-05}
{"train_loss": 3.3334357738494873, "global_step": 399, "epoch": 4, "lr": 8e-05}
{"train_loss": 3.30047869682312, "global_step": 400, "epoch": 4, "lr": 8.020000000000001e-05}
{"train_loss": 3.243906259536743, "global_step": 401, "epoch": 4, "lr": 8.04e-05}
{"train_loss": 3.3517398834228516, "global_step": 402, "epoch": 4, "lr": 8.060000000000001e-05}
{"train_loss": 3.3653435707092285, "global_step": 403, "epoch": 4, "lr": 8.080000000000001e-05}
{"train_loss": 3.325368881225586, "global_step": 404, "epoch": 4, "lr": 8.1e-05}
{"train_loss": 3.315351963043213, "global_step": 405, "epoch": 4, "lr": 8.120000000000001e-05}
{"train_loss": 3.4262266159057617, "global_step": 406, "epoch": 4, "lr": 8.14e-05}
{"train_loss": 3.216489791870117, "global_step": 407, "epoch": 4, "lr": 8.16e-05}
{"train_loss": 3.294632911682129, "global_step": 408, "epoch": 4, "lr": 8.18e-05}
{"train_loss": 3.2846567630767822, "global_step": 409, "epoch": 4, "lr": 8.2e-05}
{"train_loss": 3.126993179321289, "global_step": 410, "epoch": 4, "lr": 8.22e-05}
{"train_loss": 3.3200321197509766, "global_step": 411, "epoch": 4, "lr": 8.24e-05}
{"train_loss": 3.0471103191375732, "global_step": 412, "epoch": 4, "lr": 8.26e-05}
{"train_loss": 3.2627134323120117, "global_step": 413, "epoch": 4, "lr": 8.28e-05}
{"train_loss": 3.3766934871673584, "global_step": 414, "epoch": 4, "lr": 8.3e-05}
{"train_loss": 3.004138946533203, "global_step": 415, "epoch": 4, "lr": 8.32e-05}
{"train_loss": 3.140504837036133, "global_step": 416, "epoch": 4, "lr": 8.34e-05}
{"train_loss": 3.293484687805176, "global_step": 417, "epoch": 4, "lr": 8.36e-05}
{"train_loss": 3.064620018005371, "global_step": 418, "epoch": 4, "lr": 8.38e-05}
{"train_loss": 3.0510141849517822, "global_step": 419, "epoch": 4, "lr": 8.4e-05}
{"train_loss": 3.1811418533325195, "global_step": 420, "epoch": 4, "lr": 8.42e-05}
{"train_loss": 3.17852783203125, "global_step": 421, "epoch": 4, "lr": 8.44e-05}
{"train_loss": 3.051692485809326, "global_step": 422, "epoch": 4, "lr": 8.46e-05}
{"train_loss": 3.1847686767578125, "global_step": 423, "epoch": 4, "lr": 8.48e-05}
{"train_loss": 3.3424336910247803, "global_step": 424, "epoch": 4, "lr": 8.5e-05}
{"train_loss": 3.174901008605957, "global_step": 425, "epoch": 4, "lr": 8.52e-05}
{"train_loss": 3.32881498336792, "global_step": 426, "epoch": 4, "lr": 8.54e-05}
{"train_loss": 3.2303693294525146, "global_step": 427, "epoch": 4, "lr": 8.560000000000001e-05}
{"train_loss": 3.201098680496216, "global_step": 428, "epoch": 4, "lr": 8.58e-05}
{"train_loss": 2.857927083969116, "global_step": 429, "epoch": 4, "lr": 8.6e-05}
{"train_loss": 3.2062184810638428, "global_step": 430, "epoch": 4, "lr": 8.620000000000001e-05}
{"train_loss": 3.1873867511749268, "global_step": 431, "epoch": 4, "lr": 8.64e-05}
{"train_loss": 3.237363815307617, "global_step": 432, "epoch": 4, "lr": 8.66e-05}
{"train_loss": 3.458624839782715, "global_step": 433, "epoch": 4, "lr": 8.680000000000001e-05}
{"train_loss": 3.162717342376709, "global_step": 434, "epoch": 4, "lr": 8.7e-05}
{"train_loss": 3.1882169246673584, "global_step": 435, "epoch": 4, "lr": 8.72e-05}
{"train_loss": 3.1674859523773193, "global_step": 436, "epoch": 4, "lr": 8.740000000000001e-05}
{"train_loss": 2.966839075088501, "global_step": 437, "epoch": 4, "lr": 8.76e-05}
{"train_loss": 3.0934629440307617, "global_step": 438, "epoch": 4, "lr": 8.78e-05}
{"train_loss": 3.2094016075134277, "global_step": 439, "epoch": 4, "lr": 8.800000000000001e-05}
{"train_loss": 3.116628408432007, "global_step": 440, "epoch": 4, "lr": 8.82e-05}
{"train_loss": 3.1948070526123047, "global_step": 441, "epoch": 4, "lr": 8.840000000000001e-05}
{"train_loss": 3.317453384399414, "global_step": 442, "epoch": 4, "lr": 8.86e-05}
{"train_loss": 3.135988235473633, "global_step": 443, "epoch": 4, "lr": 8.88e-05}
{"train_loss": 3.2505445266037842, "global_step": 444, "epoch": 4, "lr": 8.900000000000001e-05, "val_loss": 3.1402487754821777}
{"train_loss": 3.2681632041931152, "global_step": 445, "epoch": 5, "lr": 8.92e-05}
{"train_loss": 3.0977113246917725, "global_step": 446, "epoch": 5, "lr": 8.94e-05}
{"train_loss": 3.1330244541168213, "global_step": 447, "epoch": 5, "lr": 8.960000000000001e-05}
{"train_loss": 3.1745893955230713, "global_step": 448, "epoch": 5, "lr": 8.98e-05}
{"train_loss": 3.0939931869506836, "global_step": 449, "epoch": 5, "lr": 9e-05}
{"train_loss": 3.1569929122924805, "global_step": 450, "epoch": 5, "lr": 9.020000000000001e-05}
{"train_loss": 3.2364680767059326, "global_step": 451, "epoch": 5, "lr": 9.04e-05}
{"train_loss": 3.0216453075408936, "global_step": 452, "epoch": 5, "lr": 9.06e-05}
{"train_loss": 2.983628749847412, "global_step": 453, "epoch": 5, "lr": 9.080000000000001e-05}
{"train_loss": 2.91839337348938, "global_step": 454, "epoch": 5, "lr": 9.1e-05}
{"train_loss": 3.0857999324798584, "global_step": 455, "epoch": 5, "lr": 9.120000000000001e-05}
{"train_loss": 3.3466362953186035, "global_step": 456, "epoch": 5, "lr": 9.140000000000001e-05}
{"train_loss": 3.0355477333068848, "global_step": 457, "epoch": 5, "lr": 9.16e-05}
{"train_loss": 3.171994686126709, "global_step": 458, "epoch": 5, "lr": 9.180000000000001e-05}
{"train_loss": 2.9173879623413086, "global_step": 459, "epoch": 5, "lr": 9.200000000000001e-05}
{"train_loss": 3.1801650524139404, "global_step": 460, "epoch": 5, "lr": 9.22e-05}
{"train_loss": 3.1212735176086426, "global_step": 461, "epoch": 5, "lr": 9.240000000000001e-05}
{"train_loss": 2.9838736057281494, "global_step": 462, "epoch": 5, "lr": 9.260000000000001e-05}
{"train_loss": 3.130174398422241, "global_step": 463, "epoch": 5, "lr": 9.28e-05}
{"train_loss": 3.1872942447662354, "global_step": 464, "epoch": 5, "lr": 9.300000000000001e-05}
{"train_loss": 3.071723461151123, "global_step": 465, "epoch": 5, "lr": 9.320000000000002e-05}
{"train_loss": 2.9872665405273438, "global_step": 466, "epoch": 5, "lr": 9.340000000000001e-05}
{"train_loss": 3.1209664344787598, "global_step": 467, "epoch": 5, "lr": 9.360000000000001e-05}
{"train_loss": 2.769490957260132, "global_step": 468, "epoch": 5, "lr": 9.38e-05}
{"train_loss": 3.162195920944214, "global_step": 469, "epoch": 5, "lr": 9.4e-05}
{"train_loss": 3.065387487411499, "global_step": 470, "epoch": 5, "lr": 9.42e-05}
{"train_loss": 3.075916290283203, "global_step": 471, "epoch": 5, "lr": 9.44e-05}
{"train_loss": 3.0277554988861084, "global_step": 472, "epoch": 5, "lr": 9.46e-05}
{"train_loss": 3.1168742179870605, "global_step": 473, "epoch": 5, "lr": 9.48e-05}
{"train_loss": 3.2306928634643555, "global_step": 474, "epoch": 5, "lr": 9.5e-05}
{"train_loss": 3.146883010864258, "global_step": 475, "epoch": 5, "lr": 9.52e-05}
{"train_loss": 3.1700246334075928, "global_step": 476, "epoch": 5, "lr": 9.54e-05}
{"train_loss": 3.0419440269470215, "global_step": 477, "epoch": 5, "lr": 9.56e-05}
{"train_loss": 2.8088884353637695, "global_step": 478, "epoch": 5, "lr": 9.58e-05}
{"train_loss": 3.00726318359375, "global_step": 479, "epoch": 5, "lr": 9.6e-05}
{"train_loss": 2.9973292350769043, "global_step": 480, "epoch": 5, "lr": 9.620000000000001e-05}
{"train_loss": 3.0300471782684326, "global_step": 481, "epoch": 5, "lr": 9.64e-05}
{"train_loss": 3.011200189590454, "global_step": 482, "epoch": 5, "lr": 9.66e-05}
{"train_loss": 3.019521713256836, "global_step": 483, "epoch": 5, "lr": 9.680000000000001e-05}
{"train_loss": 3.0192596912384033, "global_step": 484, "epoch": 5, "lr": 9.7e-05}
{"train_loss": 3.017204761505127, "global_step": 485, "epoch": 5, "lr": 9.72e-05}
{"train_loss": 3.212190866470337, "global_step": 486, "epoch": 5, "lr": 9.74e-05}
{"train_loss": 3.0937328338623047, "global_step": 487, "epoch": 5, "lr": 9.76e-05}
{"train_loss": 2.9539546966552734, "global_step": 488, "epoch": 5, "lr": 9.78e-05}
{"train_loss": 3.304901599884033, "global_step": 489, "epoch": 5, "lr": 9.8e-05}
{"train_loss": 3.155863046646118, "global_step": 490, "epoch": 5, "lr": 9.82e-05}
{"train_loss": 2.86562442779541, "global_step": 491, "epoch": 5, "lr": 9.84e-05}
{"train_loss": 3.210834264755249, "global_step": 492, "epoch": 5, "lr": 9.86e-05}
{"train_loss": 2.858262300491333, "global_step": 493, "epoch": 5, "lr": 9.88e-05}
{"train_loss": 3.086055040359497, "global_step": 494, "epoch": 5, "lr": 9.900000000000001e-05}
{"train_loss": 3.210843324661255, "global_step": 495, "epoch": 5, "lr": 9.92e-05}
{"train_loss": 3.1792283058166504, "global_step": 496, "epoch": 5, "lr": 9.94e-05}
{"train_loss": 3.0664374828338623, "global_step": 497, "epoch": 5, "lr": 9.960000000000001e-05}
{"train_loss": 2.745972156524658, "global_step": 498, "epoch": 5, "lr": 9.98e-05}
{"train_loss": 3.161118268966675, "global_step": 499, "epoch": 5, "lr": 0.0001}
{"train_loss": 2.913025140762329, "global_step": 500, "epoch": 5, "lr": 9.999999999663906e-05}
{"train_loss": 2.8590869903564453, "global_step": 501, "epoch": 5, "lr": 9.999999998655622e-05}
{"train_loss": 3.0816197395324707, "global_step": 502, "epoch": 5, "lr": 9.99999999697515e-05}
{"train_loss": 2.982052803039551, "global_step": 503, "epoch": 5, "lr": 9.999999994622489e-05}
{"train_loss": 3.238125801086426, "global_step": 504, "epoch": 5, "lr": 9.999999991597637e-05}
{"train_loss": 3.0657036304473877, "global_step": 505, "epoch": 5, "lr": 9.999999987900597e-05}
{"train_loss": 3.0626721382141113, "global_step": 506, "epoch": 5, "lr": 9.999999983531368e-05}
{"train_loss": 3.2210745811462402, "global_step": 507, "epoch": 5, "lr": 9.999999978489951e-05}
{"train_loss": 3.0067882537841797, "global_step": 508, "epoch": 5, "lr": 9.999999972776344e-05}
{"train_loss": 2.8241026401519775, "global_step": 509, "epoch": 5, "lr": 9.999999966390548e-05}
{"train_loss": 2.9931371212005615, "global_step": 510, "epoch": 5, "lr": 9.999999959332562e-05}
{"train_loss": 3.1418278217315674, "global_step": 511, "epoch": 5, "lr": 9.999999951602388e-05}
{"train_loss": 3.0613577365875244, "global_step": 512, "epoch": 5, "lr": 9.999999943200024e-05}
{"train_loss": 3.207036256790161, "global_step": 513, "epoch": 5, "lr": 9.999999934125473e-05}
{"train_loss": 2.912343740463257, "global_step": 514, "epoch": 5, "lr": 9.999999924378731e-05}
{"train_loss": 3.054011344909668, "global_step": 515, "epoch": 5, "lr": 9.9999999139598e-05}
{"train_loss": 3.0311012268066406, "global_step": 516, "epoch": 5, "lr": 9.999999902868682e-05}
{"train_loss": 3.261530637741089, "global_step": 517, "epoch": 5, "lr": 9.999999891105374e-05}
{"train_loss": 3.0292718410491943, "global_step": 518, "epoch": 5, "lr": 9.999999878669876e-05}
{"train_loss": 3.2782492637634277, "global_step": 519, "epoch": 5, "lr": 9.999999865562189e-05}
{"train_loss": 2.9926631450653076, "global_step": 520, "epoch": 5, "lr": 9.999999851782313e-05}
{"train_loss": 3.0520107746124268, "global_step": 521, "epoch": 5, "lr": 9.999999837330249e-05}
{"train_loss": 2.98087739944458, "global_step": 522, "epoch": 5, "lr": 9.999999822205996e-05}
{"train_loss": 3.06003475189209, "global_step": 523, "epoch": 5, "lr": 9.999999806409552e-05}
{"train_loss": 3.0643608570098877, "global_step": 524, "epoch": 5, "lr": 9.999999789940921e-05}
{"train_loss": 2.852428913116455, "global_step": 525, "epoch": 5, "lr": 9.999999772800101e-05}
{"train_loss": 3.0530428886413574, "global_step": 526, "epoch": 5, "lr": 9.99999975498709e-05}
{"train_loss": 3.0130324363708496, "global_step": 527, "epoch": 5, "lr": 9.999999736501892e-05}
{"train_loss": 2.963200807571411, "global_step": 528, "epoch": 5, "lr": 9.999999717344504e-05}
{"train_loss": 3.0075790882110596, "global_step": 529, "epoch": 5, "lr": 9.999999697514928e-05}
{"train_loss": 3.1028761863708496, "global_step": 530, "epoch": 5, "lr": 9.99999967701316e-05}
{"train_loss": 2.8217811584472656, "global_step": 531, "epoch": 5, "lr": 9.999999655839207e-05}
{"train_loss": 2.9536991119384766, "global_step": 532, "epoch": 5, "lr": 9.999999633993062e-05}
{"train_loss": 3.05931592791268, "global_step": 533, "epoch": 5, "lr": 9.99999961147473e-05, "val_loss": 2.902238607406616, "train_action_mse_error": 345.84197998046875}
{"train_loss": 2.900505542755127, "global_step": 534, "epoch": 6, "lr": 9.999999588284208e-05}
{"train_loss": 2.8863272666931152, "global_step": 535, "epoch": 6, "lr": 9.999999564421496e-05}
{"train_loss": 2.8617677688598633, "global_step": 536, "epoch": 6, "lr": 9.999999539886597e-05}
{"train_loss": 3.1065053939819336, "global_step": 537, "epoch": 6, "lr": 9.999999514679508e-05}
{"train_loss": 2.7791969776153564, "global_step": 538, "epoch": 6, "lr": 9.99999948880023e-05}
{"train_loss": 2.94779109954834, "global_step": 539, "epoch": 6, "lr": 9.999999462248763e-05}
{"train_loss": 3.078566789627075, "global_step": 540, "epoch": 6, "lr": 9.999999435025107e-05}
{"train_loss": 3.040174961090088, "global_step": 541, "epoch": 6, "lr": 9.999999407129262e-05}
{"train_loss": 2.8810508251190186, "global_step": 542, "epoch": 6, "lr": 9.999999378561229e-05}
{"train_loss": 2.858099937438965, "global_step": 543, "epoch": 6, "lr": 9.999999349321005e-05}
{"train_loss": 2.944260597229004, "global_step": 544, "epoch": 6, "lr": 9.999999319408594e-05}
{"train_loss": 2.9548821449279785, "global_step": 545, "epoch": 6, "lr": 9.999999288823993e-05}
{"train_loss": 2.765718936920166, "global_step": 546, "epoch": 6, "lr": 9.999999257567203e-05}
{"train_loss": 2.9751498699188232, "global_step": 547, "epoch": 6, "lr": 9.999999225638226e-05}
{"train_loss": 2.828372001647949, "global_step": 548, "epoch": 6, "lr": 9.999999193037057e-05}
{"train_loss": 2.9077036380767822, "global_step": 549, "epoch": 6, "lr": 9.999999159763701e-05}
{"train_loss": 2.84401798248291, "global_step": 550, "epoch": 6, "lr": 9.999999125818155e-05}
{"train_loss": 3.249375820159912, "global_step": 551, "epoch": 6, "lr": 9.99999909120042e-05}
{"train_loss": 2.8135149478912354, "global_step": 552, "epoch": 6, "lr": 9.999999055910497e-05}
{"train_loss": 3.085820436477661, "global_step": 553, "epoch": 6, "lr": 9.999999019948385e-05}
{"train_loss": 2.8083715438842773, "global_step": 554, "epoch": 6, "lr": 9.999998983314084e-05}
{"train_loss": 2.908252716064453, "global_step": 555, "epoch": 6, "lr": 9.999998946007594e-05}
{"train_loss": 2.8534834384918213, "global_step": 556, "epoch": 6, "lr": 9.999998908028915e-05}
{"train_loss": 2.98336124420166, "global_step": 557, "epoch": 6, "lr": 9.999998869378047e-05}
{"train_loss": 2.719276189804077, "global_step": 558, "epoch": 6, "lr": 9.99999883005499e-05}
{"train_loss": 2.847579002380371, "global_step": 559, "epoch": 6, "lr": 9.999998790059743e-05}
{"train_loss": 3.016294479370117, "global_step": 560, "epoch": 6, "lr": 9.99999874939231e-05}
{"train_loss": 2.805542469024658, "global_step": 561, "epoch": 6, "lr": 9.999998708052686e-05}
{"train_loss": 2.8622357845306396, "global_step": 562, "epoch": 6, "lr": 9.999998666040873e-05}
{"train_loss": 3.1015708446502686, "global_step": 563, "epoch": 6, "lr": 9.999998623356872e-05}
{"train_loss": 2.9065890312194824, "global_step": 564, "epoch": 6, "lr": 9.999998580000683e-05}
{"train_loss": 3.1562376022338867, "global_step": 565, "epoch": 6, "lr": 9.999998535972303e-05}
{"train_loss": 2.7200632095336914, "global_step": 566, "epoch": 6, "lr": 9.999998491271733e-05}
{"train_loss": 2.81278657913208, "global_step": 567, "epoch": 6, "lr": 9.999998445898977e-05}
{"train_loss": 3.145599126815796, "global_step": 568, "epoch": 6, "lr": 9.999998399854031e-05}
{"train_loss": 2.7215383052825928, "global_step": 569, "epoch": 6, "lr": 9.999998353136897e-05}
{"train_loss": 2.894975423812866, "global_step": 570, "epoch": 6, "lr": 9.999998305747574e-05}
{"train_loss": 2.686452865600586, "global_step": 571, "epoch": 6, "lr": 9.999998257686062e-05}
{"train_loss": 3.003713846206665, "global_step": 572, "epoch": 6, "lr": 9.999998208952361e-05}
{"train_loss": 2.912400722503662, "global_step": 573, "epoch": 6, "lr": 9.99999815954647e-05}
{"train_loss": 2.752307176589966, "global_step": 574, "epoch": 6, "lr": 9.999998109468393e-05}
{"train_loss": 2.6942975521087646, "global_step": 575, "epoch": 6, "lr": 9.999998058718125e-05}
{"train_loss": 3.0608115196228027, "global_step": 576, "epoch": 6, "lr": 9.999998007295669e-05}
{"train_loss": 2.896693468093872, "global_step": 577, "epoch": 6, "lr": 9.999997955201024e-05}
{"train_loss": 2.660264015197754, "global_step": 578, "epoch": 6, "lr": 9.99999790243419e-05}
{"train_loss": 2.7296881675720215, "global_step": 579, "epoch": 6, "lr": 9.999997848995168e-05}
{"train_loss": 2.860921859741211, "global_step": 580, "epoch": 6, "lr": 9.999997794883956e-05}
{"train_loss": 2.9137213230133057, "global_step": 581, "epoch": 6, "lr": 9.999997740100555e-05}
{"train_loss": 2.9385221004486084, "global_step": 582, "epoch": 6, "lr": 9.999997684644967e-05}
{"train_loss": 2.887220859527588, "global_step": 583, "epoch": 6, "lr": 9.999997628517189e-05}
{"train_loss": 3.0817019939422607, "global_step": 584, "epoch": 6, "lr": 9.999997571717224e-05}
{"train_loss": 3.0096232891082764, "global_step": 585, "epoch": 6, "lr": 9.999997514245068e-05}
{"train_loss": 2.840980291366577, "global_step": 586, "epoch": 6, "lr": 9.999997456100724e-05}
{"train_loss": 3.0414934158325195, "global_step": 587, "epoch": 6, "lr": 9.999997397284191e-05}
{"train_loss": 2.9266977310180664, "global_step": 588, "epoch": 6, "lr": 9.99999733779547e-05}
{"train_loss": 2.9968302249908447, "global_step": 589, "epoch": 6, "lr": 9.99999727763456e-05}
{"train_loss": 2.9652466773986816, "global_step": 590, "epoch": 6, "lr": 9.999997216801462e-05}
{"train_loss": 2.646266222000122, "global_step": 591, "epoch": 6, "lr": 9.999997155296174e-05}
{"train_loss": 2.8977222442626953, "global_step": 592, "epoch": 6, "lr": 9.999997093118699e-05}
{"train_loss": 2.6979753971099854, "global_step": 593, "epoch": 6, "lr": 9.999997030269034e-05}
{"train_loss": 2.743504524230957, "global_step": 594, "epoch": 6, "lr": 9.999996966747181e-05}
{"train_loss": 2.9392459392547607, "global_step": 595, "epoch": 6, "lr": 9.99999690255314e-05}
{"train_loss": 2.670738697052002, "global_step": 596, "epoch": 6, "lr": 9.999996837686908e-05}
{"train_loss": 3.0327341556549072, "global_step": 597, "epoch": 6, "lr": 9.999996772148488e-05}
{"train_loss": 2.809424877166748, "global_step": 598, "epoch": 6, "lr": 9.999996705937881e-05}
{"train_loss": 2.8103253841400146, "global_step": 599, "epoch": 6, "lr": 9.999996639055085e-05}
{"train_loss": 2.8060977458953857, "global_step": 600, "epoch": 6, "lr": 9.999996571500098e-05}
{"train_loss": 2.983856439590454, "global_step": 601, "epoch": 6, "lr": 9.999996503272925e-05}
{"train_loss": 3.0594868659973145, "global_step": 602, "epoch": 6, "lr": 9.999996434373563e-05}
{"train_loss": 2.7101991176605225, "global_step": 603, "epoch": 6, "lr": 9.999996364802012e-05}
{"train_loss": 2.931814670562744, "global_step": 604, "epoch": 6, "lr": 9.999996294558273e-05}
{"train_loss": 2.868405342102051, "global_step": 605, "epoch": 6, "lr": 9.999996223642345e-05}
{"train_loss": 2.936138153076172, "global_step": 606, "epoch": 6, "lr": 9.999996152054228e-05}
{"train_loss": 2.75601863861084, "global_step": 607, "epoch": 6, "lr": 9.999996079793923e-05}
{"train_loss": 2.8926520347595215, "global_step": 608, "epoch": 6, "lr": 9.999996006861429e-05}
{"train_loss": 2.7464027404785156, "global_step": 609, "epoch": 6, "lr": 9.999995933256748e-05}
{"train_loss": 2.7575759887695312, "global_step": 610, "epoch": 6, "lr": 9.999995858979877e-05}
{"train_loss": 2.913252353668213, "global_step": 611, "epoch": 6, "lr": 9.999995784030818e-05}
{"train_loss": 2.8008546829223633, "global_step": 612, "epoch": 6, "lr": 9.999995708409571e-05}
{"train_loss": 2.817976474761963, "global_step": 613, "epoch": 6, "lr": 9.999995632116134e-05}
{"train_loss": 2.9046730995178223, "global_step": 614, "epoch": 6, "lr": 9.99999555515051e-05}
{"train_loss": 2.9236507415771484, "global_step": 615, "epoch": 6, "lr": 9.999995477512696e-05}
{"train_loss": 2.802497625350952, "global_step": 616, "epoch": 6, "lr": 9.999995399202695e-05}
{"train_loss": 2.9316823482513428, "global_step": 617, "epoch": 6, "lr": 9.999995320220505e-05}
{"train_loss": 2.780601978302002, "global_step": 618, "epoch": 6, "lr": 9.999995240566127e-05}
{"train_loss": 3.0038678646087646, "global_step": 619, "epoch": 6, "lr": 9.999995160239559e-05}
{"train_loss": 2.817890167236328, "global_step": 620, "epoch": 6, "lr": 9.999995079240805e-05}
{"train_loss": 2.5822668075561523, "global_step": 621, "epoch": 6, "lr": 9.999994997569861e-05}
{"train_loss": 2.881960472364104, "global_step": 622, "epoch": 6, "lr": 9.99999491522673e-05, "val_loss": 2.7751927375793457}
{"train_loss": 2.6516008377075195, "global_step": 623, "epoch": 7, "lr": 9.999994832211408e-05}
{"train_loss": 2.693009376525879, "global_step": 624, "epoch": 7, "lr": 9.999994748523899e-05}
{"train_loss": 2.881587028503418, "global_step": 625, "epoch": 7, "lr": 9.999994664164203e-05}
{"train_loss": 2.881605863571167, "global_step": 626, "epoch": 7, "lr": 9.999994579132316e-05}
{"train_loss": 2.7316339015960693, "global_step": 627, "epoch": 7, "lr": 9.999994493428243e-05}
{"train_loss": 2.9320030212402344, "global_step": 628, "epoch": 7, "lr": 9.999994407051981e-05}
{"train_loss": 2.8077542781829834, "global_step": 629, "epoch": 7, "lr": 9.999994320003532e-05}
{"train_loss": 2.762077569961548, "global_step": 630, "epoch": 7, "lr": 9.999994232282891e-05}
{"train_loss": 2.666591167449951, "global_step": 631, "epoch": 7, "lr": 9.999994143890066e-05}
{"train_loss": 2.845883369445801, "global_step": 632, "epoch": 7, "lr": 9.99999405482505e-05}
{"train_loss": 2.6772282123565674, "global_step": 633, "epoch": 7, "lr": 9.999993965087847e-05}
{"train_loss": 2.6744561195373535, "global_step": 634, "epoch": 7, "lr": 9.999993874678455e-05}
{"train_loss": 2.858044385910034, "global_step": 635, "epoch": 7, "lr": 9.999993783596874e-05}
{"train_loss": 2.4325478076934814, "global_step": 636, "epoch": 7, "lr": 9.999993691843107e-05}
{"train_loss": 3.0457656383514404, "global_step": 637, "epoch": 7, "lr": 9.99999359941715e-05}
{"train_loss": 2.6514716148376465, "global_step": 638, "epoch": 7, "lr": 9.999993506319005e-05}
{"train_loss": 2.611392021179199, "global_step": 639, "epoch": 7, "lr": 9.999993412548674e-05}
{"train_loss": 2.6747689247131348, "global_step": 640, "epoch": 7, "lr": 9.999993318106152e-05}
{"train_loss": 2.6658859252929688, "global_step": 641, "epoch": 7, "lr": 9.999993222991444e-05}
{"train_loss": 2.7542896270751953, "global_step": 642, "epoch": 7, "lr": 9.999993127204546e-05}
{"train_loss": 2.717848777770996, "global_step": 643, "epoch": 7, "lr": 9.999993030745461e-05}
{"train_loss": 2.6662678718566895, "global_step": 644, "epoch": 7, "lr": 9.999992933614186e-05}
{"train_loss": 2.7638094425201416, "global_step": 645, "epoch": 7, "lr": 9.999992835810726e-05}
{"train_loss": 2.732208013534546, "global_step": 646, "epoch": 7, "lr": 9.999992737335075e-05}
{"train_loss": 3.080648899078369, "global_step": 647, "epoch": 7, "lr": 9.999992638187238e-05}
{"train_loss": 2.6631739139556885, "global_step": 648, "epoch": 7, "lr": 9.999992538367212e-05}
{"train_loss": 2.822866678237915, "global_step": 649, "epoch": 7, "lr": 9.999992437874999e-05}
{"train_loss": 2.7212750911712646, "global_step": 650, "epoch": 7, "lr": 9.999992336710596e-05}
{"train_loss": 2.703564167022705, "global_step": 651, "epoch": 7, "lr": 9.999992234874006e-05}
{"train_loss": 2.760368585586548, "global_step": 652, "epoch": 7, "lr": 9.999992132365227e-05}
{"train_loss": 2.7581934928894043, "global_step": 653, "epoch": 7, "lr": 9.999992029184261e-05}
{"train_loss": 2.792158603668213, "global_step": 654, "epoch": 7, "lr": 9.999991925331107e-05}
{"train_loss": 2.5063557624816895, "global_step": 655, "epoch": 7, "lr": 9.999991820805766e-05}
{"train_loss": 2.629425048828125, "global_step": 656, "epoch": 7, "lr": 9.999991715608237e-05}
{"train_loss": 2.7029850482940674, "global_step": 657, "epoch": 7, "lr": 9.999991609738518e-05}
{"train_loss": 2.839838743209839, "global_step": 658, "epoch": 7, "lr": 9.999991503196613e-05}
{"train_loss": 2.612196445465088, "global_step": 659, "epoch": 7, "lr": 9.999991395982518e-05}
{"train_loss": 2.7687625885009766, "global_step": 660, "epoch": 7, "lr": 9.999991288096238e-05}
{"train_loss": 2.532315492630005, "global_step": 661, "epoch": 7, "lr": 9.999991179537767e-05}
{"train_loss": 2.786068916320801, "global_step": 662, "epoch": 7, "lr": 9.99999107030711e-05}
{"train_loss": 2.746678113937378, "global_step": 663, "epoch": 7, "lr": 9.999990960404266e-05}
{"train_loss": 2.9308676719665527, "global_step": 664, "epoch": 7, "lr": 9.999990849829233e-05}
{"train_loss": 2.7543158531188965, "global_step": 665, "epoch": 7, "lr": 9.99999073858201e-05}
{"train_loss": 2.946178913116455, "global_step": 666, "epoch": 7, "lr": 9.999990626662602e-05}
{"train_loss": 2.8314390182495117, "global_step": 667, "epoch": 7, "lr": 9.999990514071005e-05}
{"train_loss": 2.708611011505127, "global_step": 668, "epoch": 7, "lr": 9.999990400807222e-05}
{"train_loss": 2.7715389728546143, "global_step": 669, "epoch": 7, "lr": 9.99999028687125e-05}
{"train_loss": 2.545152187347412, "global_step": 670, "epoch": 7, "lr": 9.999990172263089e-05}
{"train_loss": 2.6511311531066895, "global_step": 671, "epoch": 7, "lr": 9.999990056982743e-05}
{"train_loss": 2.8859286308288574, "global_step": 672, "epoch": 7, "lr": 9.999989941030207e-05}
{"train_loss": 2.9256184101104736, "global_step": 673, "epoch": 7, "lr": 9.999989824405484e-05}
{"train_loss": 2.4616611003875732, "global_step": 674, "epoch": 7, "lr": 9.999989707108573e-05}
{"train_loss": 2.917407512664795, "global_step": 675, "epoch": 7, "lr": 9.999989589139475e-05}
{"train_loss": 2.842303514480591, "global_step": 676, "epoch": 7, "lr": 9.999989470498188e-05}
{"train_loss": 2.772686004638672, "global_step": 677, "epoch": 7, "lr": 9.999989351184716e-05}
{"train_loss": 2.821110725402832, "global_step": 678, "epoch": 7, "lr": 9.999989231199053e-05}
{"train_loss": 2.6740012168884277, "global_step": 679, "epoch": 7, "lr": 9.999989110541205e-05}
{"train_loss": 2.578549861907959, "global_step": 680, "epoch": 7, "lr": 9.999988989211169e-05}
{"train_loss": 2.7114696502685547, "global_step": 681, "epoch": 7, "lr": 9.999988867208945e-05}
{"train_loss": 2.6881725788116455, "global_step": 682, "epoch": 7, "lr": 9.999988744534532e-05}
{"train_loss": 2.731436014175415, "global_step": 683, "epoch": 7, "lr": 9.999988621187933e-05}
{"train_loss": 2.9476451873779297, "global_step": 684, "epoch": 7, "lr": 9.999988497169146e-05}
{"train_loss": 2.7318737506866455, "global_step": 685, "epoch": 7, "lr": 9.999988372478172e-05}
{"train_loss": 2.6803276538848877, "global_step": 686, "epoch": 7, "lr": 9.999988247115011e-05}
{"train_loss": 2.5988128185272217, "global_step": 687, "epoch": 7, "lr": 9.999988121079662e-05}
{"train_loss": 2.770670175552368, "global_step": 688, "epoch": 7, "lr": 9.999987994372124e-05}
{"train_loss": 2.694781541824341, "global_step": 689, "epoch": 7, "lr": 9.999987866992401e-05}
{"train_loss": 2.8757057189941406, "global_step": 690, "epoch": 7, "lr": 9.999987738940489e-05}
{"train_loss": 2.769940137863159, "global_step": 691, "epoch": 7, "lr": 9.999987610216391e-05}
{"train_loss": 2.747915029525757, "global_step": 692, "epoch": 7, "lr": 9.999987480820103e-05}
{"train_loss": 2.7567126750946045, "global_step": 693, "epoch": 7, "lr": 9.99998735075163e-05}
{"train_loss": 2.569193124771118, "global_step": 694, "epoch": 7, "lr": 9.999987220010969e-05}
{"train_loss": 2.6484055519104004, "global_step": 695, "epoch": 7, "lr": 9.999987088598121e-05}
{"train_loss": 2.5703485012054443, "global_step": 696, "epoch": 7, "lr": 9.999986956513085e-05}
{"train_loss": 2.4824118614196777, "global_step": 697, "epoch": 7, "lr": 9.999986823755862e-05}
{"train_loss": 2.767491340637207, "global_step": 698, "epoch": 7, "lr": 9.999986690326452e-05}
{"train_loss": 2.643937110900879, "global_step": 699, "epoch": 7, "lr": 9.999986556224855e-05}
{"train_loss": 2.676215887069702, "global_step": 700, "epoch": 7, "lr": 9.999986421451068e-05}
{"train_loss": 2.7639060020446777, "global_step": 701, "epoch": 7, "lr": 9.999986286005098e-05}
{"train_loss": 2.438577890396118, "global_step": 702, "epoch": 7, "lr": 9.999986149886937e-05}
{"train_loss": 2.875030040740967, "global_step": 703, "epoch": 7, "lr": 9.999986013096591e-05}
{"train_loss": 2.5494165420532227, "global_step": 704, "epoch": 7, "lr": 9.999985875634056e-05}
{"train_loss": 2.7894062995910645, "global_step": 705, "epoch": 7, "lr": 9.999985737499338e-05}
{"train_loss": 2.6322245597839355, "global_step": 706, "epoch": 7, "lr": 9.99998559869243e-05}
{"train_loss": 2.4599063396453857, "global_step": 707, "epoch": 7, "lr": 9.999985459213335e-05}
{"train_loss": 2.387714147567749, "global_step": 708, "epoch": 7, "lr": 9.999985319062052e-05}
{"train_loss": 2.828028678894043, "global_step": 709, "epoch": 7, "lr": 9.999985178238582e-05}
{"train_loss": 2.6068897247314453, "global_step": 710, "epoch": 7, "lr": 9.999985036742925e-05}
{"train_loss": 2.7208779233225275, "global_step": 711, "epoch": 7, "lr": 9.999984894575083e-05, "val_loss": 2.6516098976135254}
{"train_loss": 2.5593175888061523, "global_step": 712, "epoch": 8, "lr": 9.999984751735052e-05}
{"train_loss": 2.3716955184936523, "global_step": 713, "epoch": 8, "lr": 9.999984608222834e-05}
{"train_loss": 2.761885166168213, "global_step": 714, "epoch": 8, "lr": 9.99998446403843e-05}
{"train_loss": 2.7288780212402344, "global_step": 715, "epoch": 8, "lr": 9.999984319181839e-05}
{"train_loss": 2.547091007232666, "global_step": 716, "epoch": 8, "lr": 9.99998417365306e-05}
{"train_loss": 2.6157352924346924, "global_step": 717, "epoch": 8, "lr": 9.999984027452096e-05}
{"train_loss": 2.6348047256469727, "global_step": 718, "epoch": 8, "lr": 9.999983880578942e-05}
{"train_loss": 2.4872939586639404, "global_step": 719, "epoch": 8, "lr": 9.999983733033604e-05}
{"train_loss": 2.7764434814453125, "global_step": 720, "epoch": 8, "lr": 9.999983584816079e-05}
{"train_loss": 2.4475796222686768, "global_step": 721, "epoch": 8, "lr": 9.999983435926366e-05}
{"train_loss": 2.5004751682281494, "global_step": 722, "epoch": 8, "lr": 9.999983286364466e-05}
{"train_loss": 2.90203595161438, "global_step": 723, "epoch": 8, "lr": 9.99998313613038e-05}
{"train_loss": 2.6532199382781982, "global_step": 724, "epoch": 8, "lr": 9.999982985224106e-05}
{"train_loss": 2.745696544647217, "global_step": 725, "epoch": 8, "lr": 9.999982833645646e-05}
{"train_loss": 2.586575508117676, "global_step": 726, "epoch": 8, "lr": 9.999982681394999e-05}
{"train_loss": 2.633617639541626, "global_step": 727, "epoch": 8, "lr": 9.999982528472165e-05}
{"train_loss": 2.6974337100982666, "global_step": 728, "epoch": 8, "lr": 9.999982374877145e-05}
{"train_loss": 2.6405529975891113, "global_step": 729, "epoch": 8, "lr": 9.999982220609939e-05}
{"train_loss": 2.3078110218048096, "global_step": 730, "epoch": 8, "lr": 9.999982065670546e-05}
{"train_loss": 2.6080031394958496, "global_step": 731, "epoch": 8, "lr": 9.999981910058966e-05}
{"train_loss": 2.6365387439727783, "global_step": 732, "epoch": 8, "lr": 9.999981753775199e-05}
{"train_loss": 2.5128886699676514, "global_step": 733, "epoch": 8, "lr": 9.999981596819246e-05}
{"train_loss": 2.5257396697998047, "global_step": 734, "epoch": 8, "lr": 9.999981439191105e-05}
{"train_loss": 2.6491169929504395, "global_step": 735, "epoch": 8, "lr": 9.999981280890777e-05}
{"train_loss": 3.0164270401000977, "global_step": 736, "epoch": 8, "lr": 9.999981121918265e-05}
{"train_loss": 2.611612319946289, "global_step": 737, "epoch": 8, "lr": 9.999980962273565e-05}
{"train_loss": 2.687054395675659, "global_step": 738, "epoch": 8, "lr": 9.999980801956679e-05}
{"train_loss": 2.5651214122772217, "global_step": 739, "epoch": 8, "lr": 9.999980640967607e-05}
{"train_loss": 2.5949783325195312, "global_step": 740, "epoch": 8, "lr": 9.999980479306348e-05}
{"train_loss": 2.567204236984253, "global_step": 741, "epoch": 8, "lr": 9.999980316972902e-05}
{"train_loss": 2.5008885860443115, "global_step": 742, "epoch": 8, "lr": 9.99998015396727e-05}
{"train_loss": 2.7352631092071533, "global_step": 743, "epoch": 8, "lr": 9.999979990289452e-05}
{"train_loss": 2.5223371982574463, "global_step": 744, "epoch": 8, "lr": 9.999979825939447e-05}
{"train_loss": 2.599632740020752, "global_step": 745, "epoch": 8, "lr": 9.999979660917256e-05}
{"train_loss": 2.690669298171997, "global_step": 746, "epoch": 8, "lr": 9.99997949522288e-05}
{"train_loss": 2.6778056621551514, "global_step": 747, "epoch": 8, "lr": 9.999979328856315e-05}
{"train_loss": 2.398268461227417, "global_step": 748, "epoch": 8, "lr": 9.999979161817565e-05}
{"train_loss": 2.4846255779266357, "global_step": 749, "epoch": 8, "lr": 9.99997899410663e-05}
{"train_loss": 2.785250663757324, "global_step": 750, "epoch": 8, "lr": 9.999978825723505e-05}
{"train_loss": 2.5830934047698975, "global_step": 751, "epoch": 8, "lr": 9.999978656668199e-05}
{"train_loss": 2.488619089126587, "global_step": 752, "epoch": 8, "lr": 9.999978486940703e-05}
{"train_loss": 2.4658687114715576, "global_step": 753, "epoch": 8, "lr": 9.999978316541024e-05}
{"train_loss": 2.398329973220825, "global_step": 754, "epoch": 8, "lr": 9.999978145469155e-05}
{"train_loss": 2.6703734397888184, "global_step": 755, "epoch": 8, "lr": 9.999977973725102e-05}
{"train_loss": 2.7765328884124756, "global_step": 756, "epoch": 8, "lr": 9.999977801308864e-05}
{"train_loss": 2.379121780395508, "global_step": 757, "epoch": 8, "lr": 9.999977628220437e-05}
{"train_loss": 2.5505173206329346, "global_step": 758, "epoch": 8, "lr": 9.999977454459826e-05}
{"train_loss": 2.594606876373291, "global_step": 759, "epoch": 8, "lr": 9.999977280027029e-05}
{"train_loss": 2.558781385421753, "global_step": 760, "epoch": 8, "lr": 9.999977104922045e-05}
{"train_loss": 2.5716874599456787, "global_step": 761, "epoch": 8, "lr": 9.999976929144876e-05}
{"train_loss": 2.360335111618042, "global_step": 762, "epoch": 8, "lr": 9.99997675269552e-05}
{"train_loss": 2.429851770401001, "global_step": 763, "epoch": 8, "lr": 9.999976575573979e-05}
{"train_loss": 2.5343949794769287, "global_step": 764, "epoch": 8, "lr": 9.99997639778025e-05}
{"train_loss": 2.635303497314453, "global_step": 765, "epoch": 8, "lr": 9.999976219314337e-05}
{"train_loss": 2.5034165382385254, "global_step": 766, "epoch": 8, "lr": 9.999976040176239e-05}
{"train_loss": 2.4796857833862305, "global_step": 767, "epoch": 8, "lr": 9.999975860365955e-05}
{"train_loss": 2.5757033824920654, "global_step": 768, "epoch": 8, "lr": 9.999975679883484e-05}
{"train_loss": 2.288039207458496, "global_step": 769, "epoch": 8, "lr": 9.999975498728827e-05}
{"train_loss": 2.619615077972412, "global_step": 770, "epoch": 8, "lr": 9.999975316901984e-05}
{"train_loss": 2.424415111541748, "global_step": 771, "epoch": 8, "lr": 9.999975134402957e-05}
{"train_loss": 2.4400627613067627, "global_step": 772, "epoch": 8, "lr": 9.999974951231743e-05}
{"train_loss": 2.6107826232910156, "global_step": 773, "epoch": 8, "lr": 9.999974767388343e-05}
{"train_loss": 2.684277057647705, "global_step": 774, "epoch": 8, "lr": 9.999974582872758e-05}
{"train_loss": 2.5002784729003906, "global_step": 775, "epoch": 8, "lr": 9.999974397684988e-05}
{"train_loss": 2.346853733062744, "global_step": 776, "epoch": 8, "lr": 9.999974211825032e-05}
{"train_loss": 2.545642614364624, "global_step": 777, "epoch": 8, "lr": 9.99997402529289e-05}
{"train_loss": 2.545856237411499, "global_step": 778, "epoch": 8, "lr": 9.999973838088563e-05}
{"train_loss": 2.693056106567383, "global_step": 779, "epoch": 8, "lr": 9.999973650212048e-05}
{"train_loss": 2.5831096172332764, "global_step": 780, "epoch": 8, "lr": 9.99997346166335e-05}
{"train_loss": 2.4000167846679688, "global_step": 781, "epoch": 8, "lr": 9.999973272442466e-05}
{"train_loss": 2.4243972301483154, "global_step": 782, "epoch": 8, "lr": 9.999973082549397e-05}
{"train_loss": 2.579725980758667, "global_step": 783, "epoch": 8, "lr": 9.999972891984141e-05}
{"train_loss": 2.5825514793395996, "global_step": 784, "epoch": 8, "lr": 9.999972700746702e-05}
{"train_loss": 2.6334457397460938, "global_step": 785, "epoch": 8, "lr": 9.999972508837076e-05}
{"train_loss": 2.353489875793457, "global_step": 786, "epoch": 8, "lr": 9.999972316255266e-05}
{"train_loss": 2.7052860260009766, "global_step": 787, "epoch": 8, "lr": 9.999972123001269e-05}
{"train_loss": 2.4080801010131836, "global_step": 788, "epoch": 8, "lr": 9.999971929075087e-05}
{"train_loss": 2.285475254058838, "global_step": 789, "epoch": 8, "lr": 9.999971734476721e-05}
{"train_loss": 2.5395705699920654, "global_step": 790, "epoch": 8, "lr": 9.999971539206167e-05}
{"train_loss": 2.628527879714966, "global_step": 791, "epoch": 8, "lr": 9.99997134326343e-05}
{"train_loss": 2.412914752960205, "global_step": 792, "epoch": 8, "lr": 9.999971146648508e-05}
{"train_loss": 2.249959945678711, "global_step": 793, "epoch": 8, "lr": 9.9999709493614e-05}
{"train_loss": 2.2577056884765625, "global_step": 794, "epoch": 8, "lr": 9.999970751402107e-05}
{"train_loss": 2.5182552337646484, "global_step": 795, "epoch": 8, "lr": 9.999970552770627e-05}
{"train_loss": 2.47096586227417, "global_step": 796, "epoch": 8, "lr": 9.999970353466965e-05}
{"train_loss": 2.475132703781128, "global_step": 797, "epoch": 8, "lr": 9.999970153491117e-05}
{"train_loss": 2.2686033248901367, "global_step": 798, "epoch": 8, "lr": 9.999969952843083e-05}
{"train_loss": 2.4685237407684326, "global_step": 799, "epoch": 8, "lr": 9.999969751522866e-05}
{"train_loss": 2.5493014501721674, "global_step": 800, "epoch": 8, "lr": 9.999969549530461e-05, "val_loss": 2.5704410076141357}
{"train_loss": 2.6372456550598145, "global_step": 801, "epoch": 9, "lr": 9.999969346865873e-05}
{"train_loss": 2.7110977172851562, "global_step": 802, "epoch": 9, "lr": 9.9999691435291e-05}
{"train_loss": 2.4381799697875977, "global_step": 803, "epoch": 9, "lr": 9.999968939520141e-05}
{"train_loss": 2.3375229835510254, "global_step": 804, "epoch": 9, "lr": 9.999968734838999e-05}
{"train_loss": 2.293586254119873, "global_step": 805, "epoch": 9, "lr": 9.99996852948567e-05}
{"train_loss": 2.367323875427246, "global_step": 806, "epoch": 9, "lr": 9.999968323460159e-05}
{"train_loss": 2.5230605602264404, "global_step": 807, "epoch": 9, "lr": 9.99996811676246e-05}
{"train_loss": 2.441913366317749, "global_step": 808, "epoch": 9, "lr": 9.999967909392578e-05}
{"train_loss": 2.314042091369629, "global_step": 809, "epoch": 9, "lr": 9.999967701350511e-05}
{"train_loss": 2.399460792541504, "global_step": 810, "epoch": 9, "lr": 9.999967492636259e-05}
{"train_loss": 2.277599811553955, "global_step": 811, "epoch": 9, "lr": 9.999967283249823e-05}
{"train_loss": 2.4192676544189453, "global_step": 812, "epoch": 9, "lr": 9.9999670731912e-05}
{"train_loss": 2.4913036823272705, "global_step": 813, "epoch": 9, "lr": 9.999966862460396e-05}
{"train_loss": 2.4245848655700684, "global_step": 814, "epoch": 9, "lr": 9.999966651057406e-05}
{"train_loss": 2.631784439086914, "global_step": 815, "epoch": 9, "lr": 9.99996643898223e-05}
{"train_loss": 2.396418333053589, "global_step": 816, "epoch": 9, "lr": 9.999966226234871e-05}
{"train_loss": 2.469834566116333, "global_step": 817, "epoch": 9, "lr": 9.999966012815327e-05}
{"train_loss": 2.62135648727417, "global_step": 818, "epoch": 9, "lr": 9.999965798723599e-05}
{"train_loss": 2.62259578704834, "global_step": 819, "epoch": 9, "lr": 9.999965583959685e-05}
{"train_loss": 2.349804639816284, "global_step": 820, "epoch": 9, "lr": 9.999965368523587e-05}
{"train_loss": 2.4760515689849854, "global_step": 821, "epoch": 9, "lr": 9.999965152415306e-05}
{"train_loss": 2.4590630531311035, "global_step": 822, "epoch": 9, "lr": 9.99996493563484e-05}
{"train_loss": 2.554710865020752, "global_step": 823, "epoch": 9, "lr": 9.999964718182189e-05}
{"train_loss": 2.5582640171051025, "global_step": 824, "epoch": 9, "lr": 9.999964500057354e-05}
{"train_loss": 2.411365509033203, "global_step": 825, "epoch": 9, "lr": 9.999964281260336e-05}
{"train_loss": 2.5957489013671875, "global_step": 826, "epoch": 9, "lr": 9.999964061791131e-05}
{"train_loss": 2.292886734008789, "global_step": 827, "epoch": 9, "lr": 9.999963841649744e-05}
{"train_loss": 2.4375855922698975, "global_step": 828, "epoch": 9, "lr": 9.999963620836173e-05}
{"train_loss": 2.251631259918213, "global_step": 829, "epoch": 9, "lr": 9.999963399350416e-05}
{"train_loss": 2.348449945449829, "global_step": 830, "epoch": 9, "lr": 9.999963177192476e-05}
{"train_loss": 2.4199376106262207, "global_step": 831, "epoch": 9, "lr": 9.999962954362352e-05}
{"train_loss": 2.5195038318634033, "global_step": 832, "epoch": 9, "lr": 9.999962730860044e-05}
{"train_loss": 2.3113577365875244, "global_step": 833, "epoch": 9, "lr": 9.999962506685552e-05}
{"train_loss": 2.374293565750122, "global_step": 834, "epoch": 9, "lr": 9.999962281838874e-05}
{"train_loss": 2.668501377105713, "global_step": 835, "epoch": 9, "lr": 9.999962056320015e-05}
{"train_loss": 2.2727301120758057, "global_step": 836, "epoch": 9, "lr": 9.99996183012897e-05}
{"train_loss": 2.413688898086548, "global_step": 837, "epoch": 9, "lr": 9.999961603265741e-05}
{"train_loss": 2.4499449729919434, "global_step": 838, "epoch": 9, "lr": 9.99996137573033e-05}
{"train_loss": 2.4989712238311768, "global_step": 839, "epoch": 9, "lr": 9.999961147522733e-05}
{"train_loss": 2.3532752990722656, "global_step": 840, "epoch": 9, "lr": 9.999960918642955e-05}
{"train_loss": 2.4415719509124756, "global_step": 841, "epoch": 9, "lr": 9.99996068909099e-05}
{"train_loss": 2.4799771308898926, "global_step": 842, "epoch": 9, "lr": 9.999960458866844e-05}
{"train_loss": 2.189952850341797, "global_step": 843, "epoch": 9, "lr": 9.999960227970512e-05}
{"train_loss": 2.4686124324798584, "global_step": 844, "epoch": 9, "lr": 9.999959996401997e-05}
{"train_loss": 2.5385942459106445, "global_step": 845, "epoch": 9, "lr": 9.999959764161299e-05}
{"train_loss": 2.350339412689209, "global_step": 846, "epoch": 9, "lr": 9.999959531248417e-05}
{"train_loss": 2.285384178161621, "global_step": 847, "epoch": 9, "lr": 9.999959297663351e-05}
{"train_loss": 2.610085964202881, "global_step": 848, "epoch": 9, "lr": 9.999959063406102e-05}
{"train_loss": 2.4218533039093018, "global_step": 849, "epoch": 9, "lr": 9.999958828476667e-05}
{"train_loss": 2.675096035003662, "global_step": 850, "epoch": 9, "lr": 9.999958592875052e-05}
{"train_loss": 2.577775239944458, "global_step": 851, "epoch": 9, "lr": 9.99995835660125e-05}
{"train_loss": 2.3325634002685547, "global_step": 852, "epoch": 9, "lr": 9.999958119655268e-05}
{"train_loss": 2.4585564136505127, "global_step": 853, "epoch": 9, "lr": 9.9999578820371e-05}
{"train_loss": 2.3765265941619873, "global_step": 854, "epoch": 9, "lr": 9.999957643746752e-05}
{"train_loss": 2.432067394256592, "global_step": 855, "epoch": 9, "lr": 9.999957404784217e-05}
{"train_loss": 2.252307891845703, "global_step": 856, "epoch": 9, "lr": 9.999957165149501e-05}
{"train_loss": 2.5636863708496094, "global_step": 857, "epoch": 9, "lr": 9.9999569248426e-05}
{"train_loss": 2.4279048442840576, "global_step": 858, "epoch": 9, "lr": 9.999956683863516e-05}
{"train_loss": 2.374906301498413, "global_step": 859, "epoch": 9, "lr": 9.99995644221225e-05}
{"train_loss": 2.3879077434539795, "global_step": 860, "epoch": 9, "lr": 9.999956199888801e-05}
{"train_loss": 2.1864235401153564, "global_step": 861, "epoch": 9, "lr": 9.999955956893167e-05}
{"train_loss": 2.2664122581481934, "global_step": 862, "epoch": 9, "lr": 9.99995571322535e-05}
{"train_loss": 2.456425905227661, "global_step": 863, "epoch": 9, "lr": 9.999955468885351e-05}
{"train_loss": 2.33808970451355, "global_step": 864, "epoch": 9, "lr": 9.999955223873169e-05}
{"train_loss": 2.3768558502197266, "global_step": 865, "epoch": 9, "lr": 9.999954978188804e-05}
{"train_loss": 2.360886812210083, "global_step": 866, "epoch": 9, "lr": 9.999954731832256e-05}
{"train_loss": 2.4578773975372314, "global_step": 867, "epoch": 9, "lr": 9.999954484803524e-05}
{"train_loss": 2.3833770751953125, "global_step": 868, "epoch": 9, "lr": 9.999954237102609e-05}
{"train_loss": 2.2661898136138916, "global_step": 869, "epoch": 9, "lr": 9.999953988729511e-05}
{"train_loss": 2.566246747970581, "global_step": 870, "epoch": 9, "lr": 9.99995373968423e-05}
{"train_loss": 2.422926902770996, "global_step": 871, "epoch": 9, "lr": 9.999953489966768e-05}
{"train_loss": 2.3461577892303467, "global_step": 872, "epoch": 9, "lr": 9.999953239577122e-05}
{"train_loss": 2.470059871673584, "global_step": 873, "epoch": 9, "lr": 9.999952988515293e-05}
{"train_loss": 2.414529800415039, "global_step": 874, "epoch": 9, "lr": 9.999952736781282e-05}
{"train_loss": 2.4482479095458984, "global_step": 875, "epoch": 9, "lr": 9.999952484375089e-05}
{"train_loss": 2.2585537433624268, "global_step": 876, "epoch": 9, "lr": 9.999952231296712e-05}
{"train_loss": 2.4376280307769775, "global_step": 877, "epoch": 9, "lr": 9.999951977546151e-05}
{"train_loss": 2.496455192565918, "global_step": 878, "epoch": 9, "lr": 9.999951723123411e-05}
{"train_loss": 2.3068902492523193, "global_step": 879, "epoch": 9, "lr": 9.999951468028486e-05}
{"train_loss": 2.296280860900879, "global_step": 880, "epoch": 9, "lr": 9.999951212261378e-05}
{"train_loss": 2.2257399559020996, "global_step": 881, "epoch": 9, "lr": 9.999950955822088e-05}
{"train_loss": 2.501323699951172, "global_step": 882, "epoch": 9, "lr": 9.999950698710616e-05}
{"train_loss": 2.396697521209717, "global_step": 883, "epoch": 9, "lr": 9.999950440926962e-05}
{"train_loss": 2.555590867996216, "global_step": 884, "epoch": 9, "lr": 9.999950182471125e-05}
{"train_loss": 2.3374440670013428, "global_step": 885, "epoch": 9, "lr": 9.999949923343106e-05}
{"train_loss": 2.5981850624084473, "global_step": 886, "epoch": 9, "lr": 9.999949663542905e-05}
{"train_loss": 2.351491928100586, "global_step": 887, "epoch": 9, "lr": 9.999949403070521e-05}
{"train_loss": 2.460984945297241, "global_step": 888, "epoch": 9, "lr": 9.999949141925955e-05}
{"train_loss": 2.424846667921945, "global_step": 889, "epoch": 9, "lr": 9.999948880109207e-05, "val_loss": 2.421967029571533}
{"train_loss": 2.24594783782959, "global_step": 890, "epoch": 10, "lr": 9.999948617620276e-05}
{"train_loss": 2.1986312866210938, "global_step": 891, "epoch": 10, "lr": 9.999948354459163e-05}
{"train_loss": 2.34267520904541, "global_step": 892, "epoch": 10, "lr": 9.999948090625868e-05}
{"train_loss": 2.4711363315582275, "global_step": 893, "epoch": 10, "lr": 9.999947826120392e-05}
{"train_loss": 2.2064778804779053, "global_step": 894, "epoch": 10, "lr": 9.999947560942733e-05}
{"train_loss": 2.472062349319458, "global_step": 895, "epoch": 10, "lr": 9.999947295092892e-05}
{"train_loss": 2.294358730316162, "global_step": 896, "epoch": 10, "lr": 9.999947028570868e-05}
{"train_loss": 2.2752628326416016, "global_step": 897, "epoch": 10, "lr": 9.999946761376664e-05}
{"train_loss": 2.5275332927703857, "global_step": 898, "epoch": 10, "lr": 9.999946493510278e-05}
{"train_loss": 2.464272975921631, "global_step": 899, "epoch": 10, "lr": 9.999946224971709e-05}
{"train_loss": 2.1047730445861816, "global_step": 900, "epoch": 10, "lr": 9.99994595576096e-05}
{"train_loss": 2.428760528564453, "global_step": 901, "epoch": 10, "lr": 9.999945685878027e-05}
{"train_loss": 2.3721179962158203, "global_step": 902, "epoch": 10, "lr": 9.999945415322913e-05}
{"train_loss": 2.4388740062713623, "global_step": 903, "epoch": 10, "lr": 9.999945144095617e-05}
{"train_loss": 2.4267401695251465, "global_step": 904, "epoch": 10, "lr": 9.999944872196139e-05}
{"train_loss": 2.296996831893921, "global_step": 905, "epoch": 10, "lr": 9.99994459962448e-05}
{"train_loss": 2.3005547523498535, "global_step": 906, "epoch": 10, "lr": 9.999944326380639e-05}
{"train_loss": 2.4734439849853516, "global_step": 907, "epoch": 10, "lr": 9.999944052464618e-05}
{"train_loss": 2.3688852787017822, "global_step": 908, "epoch": 10, "lr": 9.999943777876414e-05}
{"train_loss": 2.148084878921509, "global_step": 909, "epoch": 10, "lr": 9.99994350261603e-05}
{"train_loss": 2.151923656463623, "global_step": 910, "epoch": 10, "lr": 9.999943226683462e-05}
{"train_loss": 2.270191192626953, "global_step": 911, "epoch": 10, "lr": 9.999942950078714e-05}
{"train_loss": 2.4640066623687744, "global_step": 912, "epoch": 10, "lr": 9.999942672801785e-05}
{"train_loss": 2.329163074493408, "global_step": 913, "epoch": 10, "lr": 9.999942394852673e-05}
{"train_loss": 2.376803159713745, "global_step": 914, "epoch": 10, "lr": 9.999942116231382e-05}
{"train_loss": 2.452582836151123, "global_step": 915, "epoch": 10, "lr": 9.999941836937907e-05}
{"train_loss": 2.3380680084228516, "global_step": 916, "epoch": 10, "lr": 9.999941556972253e-05}
{"train_loss": 2.2264211177825928, "global_step": 917, "epoch": 10, "lr": 9.999941276334416e-05}
{"train_loss": 2.270355224609375, "global_step": 918, "epoch": 10, "lr": 9.9999409950244e-05}
{"train_loss": 2.2670581340789795, "global_step": 919, "epoch": 10, "lr": 9.999940713042202e-05}
{"train_loss": 2.1733500957489014, "global_step": 920, "epoch": 10, "lr": 9.999940430387823e-05}
{"train_loss": 2.3119335174560547, "global_step": 921, "epoch": 10, "lr": 9.999940147061262e-05}
{"train_loss": 2.272779703140259, "global_step": 922, "epoch": 10, "lr": 9.999939863062522e-05}
{"train_loss": 2.350198268890381, "global_step": 923, "epoch": 10, "lr": 9.999939578391599e-05}
{"train_loss": 2.51275897026062, "global_step": 924, "epoch": 10, "lr": 9.999939293048496e-05}
{"train_loss": 2.305576801300049, "global_step": 925, "epoch": 10, "lr": 9.999939007033212e-05}
{"train_loss": 2.1624767780303955, "global_step": 926, "epoch": 10, "lr": 9.999938720345748e-05}
{"train_loss": 2.0409903526306152, "global_step": 927, "epoch": 10, "lr": 9.9999384329861e-05}
{"train_loss": 2.616489887237549, "global_step": 928, "epoch": 10, "lr": 9.999938144954274e-05}
{"train_loss": 2.0768966674804688, "global_step": 929, "epoch": 10, "lr": 9.999937856250266e-05}
{"train_loss": 2.4678828716278076, "global_step": 930, "epoch": 10, "lr": 9.999937566874077e-05}
{"train_loss": 2.4369843006134033, "global_step": 931, "epoch": 10, "lr": 9.99993727682571e-05}
{"train_loss": 2.3076841831207275, "global_step": 932, "epoch": 10, "lr": 9.99993698610516e-05}
{"train_loss": 2.0850017070770264, "global_step": 933, "epoch": 10, "lr": 9.99993669471243e-05}
{"train_loss": 2.31551456451416, "global_step": 934, "epoch": 10, "lr": 9.999936402647521e-05}
{"train_loss": 2.252472162246704, "global_step": 935, "epoch": 10, "lr": 9.999936109910429e-05}
{"train_loss": 2.0734570026397705, "global_step": 936, "epoch": 10, "lr": 9.999935816501158e-05}
{"train_loss": 2.463444709777832, "global_step": 937, "epoch": 10, "lr": 9.999935522419705e-05}
{"train_loss": 2.236025333404541, "global_step": 938, "epoch": 10, "lr": 9.999935227666073e-05}
{"train_loss": 2.385382652282715, "global_step": 939, "epoch": 10, "lr": 9.999934932240262e-05}
{"train_loss": 2.3336057662963867, "global_step": 940, "epoch": 10, "lr": 9.999934636142267e-05}
{"train_loss": 2.224220037460327, "global_step": 941, "epoch": 10, "lr": 9.999934339372095e-05}
{"train_loss": 2.396439552307129, "global_step": 942, "epoch": 10, "lr": 9.999934041929742e-05}
{"train_loss": 2.332850933074951, "global_step": 943, "epoch": 10, "lr": 9.999933743815209e-05}
{"train_loss": 2.3791632652282715, "global_step": 944, "epoch": 10, "lr": 9.999933445028495e-05}
{"train_loss": 2.346384048461914, "global_step": 945, "epoch": 10, "lr": 9.9999331455696e-05}
{"train_loss": 2.1740105152130127, "global_step": 946, "epoch": 10, "lr": 9.999932845438526e-05}
{"train_loss": 2.5942249298095703, "global_step": 947, "epoch": 10, "lr": 9.999932544635272e-05}
{"train_loss": 2.3077759742736816, "global_step": 948, "epoch": 10, "lr": 9.999932243159839e-05}
{"train_loss": 2.2765796184539795, "global_step": 949, "epoch": 10, "lr": 9.999931941012224e-05}
{"train_loss": 2.3655247688293457, "global_step": 950, "epoch": 10, "lr": 9.99993163819243e-05}
{"train_loss": 2.2178375720977783, "global_step": 951, "epoch": 10, "lr": 9.999931334700457e-05}
{"train_loss": 2.11720871925354, "global_step": 952, "epoch": 10, "lr": 9.999931030536303e-05}
{"train_loss": 2.3913187980651855, "global_step": 953, "epoch": 10, "lr": 9.999930725699971e-05}
{"train_loss": 2.3404250144958496, "global_step": 954, "epoch": 10, "lr": 9.999930420191457e-05}
{"train_loss": 2.155703067779541, "global_step": 955, "epoch": 10, "lr": 9.999930114010764e-05}
{"train_loss": 2.2120003700256348, "global_step": 956, "epoch": 10, "lr": 9.999929807157892e-05}
{"train_loss": 2.3345017433166504, "global_step": 957, "epoch": 10, "lr": 9.999929499632839e-05}
{"train_loss": 2.3459255695343018, "global_step": 958, "epoch": 10, "lr": 9.999929191435608e-05}
{"train_loss": 2.2407279014587402, "global_step": 959, "epoch": 10, "lr": 9.999928882566196e-05}
{"train_loss": 2.4157652854919434, "global_step": 960, "epoch": 10, "lr": 9.999928573024607e-05}
{"train_loss": 2.4015707969665527, "global_step": 961, "epoch": 10, "lr": 9.999928262810837e-05}
{"train_loss": 2.3484880924224854, "global_step": 962, "epoch": 10, "lr": 9.999927951924887e-05}
{"train_loss": 2.2943849563598633, "global_step": 963, "epoch": 10, "lr": 9.999927640366758e-05}
{"train_loss": 2.3918399810791016, "global_step": 964, "epoch": 10, "lr": 9.999927328136449e-05}
{"train_loss": 2.4196512699127197, "global_step": 965, "epoch": 10, "lr": 9.999927015233962e-05}
{"train_loss": 2.1970200538635254, "global_step": 966, "epoch": 10, "lr": 9.999926701659295e-05}
{"train_loss": 2.4438424110412598, "global_step": 967, "epoch": 10, "lr": 9.99992638741245e-05}
{"train_loss": 2.3118696212768555, "global_step": 968, "epoch": 10, "lr": 9.999926072493424e-05}
{"train_loss": 2.417314291000366, "global_step": 969, "epoch": 10, "lr": 9.99992575690222e-05}
{"train_loss": 2.351893663406372, "global_step": 970, "epoch": 10, "lr": 9.999925440638836e-05}
{"train_loss": 2.3445746898651123, "global_step": 971, "epoch": 10, "lr": 9.999925123703273e-05}
{"train_loss": 2.1469736099243164, "global_step": 972, "epoch": 10, "lr": 9.999924806095532e-05}
{"train_loss": 2.254249334335327, "global_step": 973, "epoch": 10, "lr": 9.999924487815613e-05}
{"train_loss": 2.218712091445923, "global_step": 974, "epoch": 10, "lr": 9.999924168863513e-05}
{"train_loss": 2.312615156173706, "global_step": 975, "epoch": 10, "lr": 9.999923849239234e-05}
{"train_loss": 2.542372465133667, "global_step": 976, "epoch": 10, "lr": 9.999923528942778e-05}
{"train_loss": 2.1374592781066895, "global_step": 977, "epoch": 10, "lr": 9.999923207974143e-05}
{"train_loss": 2.3184458941556096, "global_step": 978, "epoch": 10, "lr": 9.999922886333329e-05, "val_loss": 2.322009325027466, "train_action_mse_error": 212.97242736816406}
{"train_loss": 2.302503824234009, "global_step": 979, "epoch": 11, "lr": 9.999922564020334e-05}
{"train_loss": 2.2256360054016113, "global_step": 980, "epoch": 11, "lr": 9.999922241035163e-05}
{"train_loss": 2.1399831771850586, "global_step": 981, "epoch": 11, "lr": 9.999921917377813e-05}
{"train_loss": 2.1516706943511963, "global_step": 982, "epoch": 11, "lr": 9.999921593048283e-05}
{"train_loss": 2.3251781463623047, "global_step": 983, "epoch": 11, "lr": 9.999921268046577e-05}
{"train_loss": 2.457951545715332, "global_step": 984, "epoch": 11, "lr": 9.99992094237269e-05}
{"train_loss": 2.4239556789398193, "global_step": 985, "epoch": 11, "lr": 9.999920616026626e-05}
{"train_loss": 2.239097833633423, "global_step": 986, "epoch": 11, "lr": 9.999920289008383e-05}
{"train_loss": 2.173600196838379, "global_step": 987, "epoch": 11, "lr": 9.999919961317963e-05}
{"train_loss": 2.3640875816345215, "global_step": 988, "epoch": 11, "lr": 9.999919632955364e-05}
{"train_loss": 2.116898536682129, "global_step": 989, "epoch": 11, "lr": 9.999919303920585e-05}
{"train_loss": 2.349179744720459, "global_step": 990, "epoch": 11, "lr": 9.999918974213628e-05}
{"train_loss": 2.290705442428589, "global_step": 991, "epoch": 11, "lr": 9.999918643834496e-05}
{"train_loss": 2.072391986846924, "global_step": 992, "epoch": 11, "lr": 9.999918312783183e-05}
{"train_loss": 2.5539896488189697, "global_step": 993, "epoch": 11, "lr": 9.999917981059693e-05}
{"train_loss": 2.1730804443359375, "global_step": 994, "epoch": 11, "lr": 9.999917648664026e-05}
{"train_loss": 2.3985278606414795, "global_step": 995, "epoch": 11, "lr": 9.99991731559618e-05}
{"train_loss": 2.2086546421051025, "global_step": 996, "epoch": 11, "lr": 9.999916981856155e-05}
{"train_loss": 2.1102449893951416, "global_step": 997, "epoch": 11, "lr": 9.999916647443952e-05}
{"train_loss": 2.348322868347168, "global_step": 998, "epoch": 11, "lr": 9.999916312359572e-05}
{"train_loss": 1.955137848854065, "global_step": 999, "epoch": 11, "lr": 9.999915976603015e-05}
{"train_loss": 2.209109306335449, "global_step": 1000, "epoch": 11, "lr": 9.99991564017428e-05}
{"train_loss": 2.1897377967834473, "global_step": 1001, "epoch": 11, "lr": 9.999915303073366e-05}
{"train_loss": 2.345972776412964, "global_step": 1002, "epoch": 11, "lr": 9.999914965300275e-05}
{"train_loss": 2.5351290702819824, "global_step": 1003, "epoch": 11, "lr": 9.999914626855008e-05}
{"train_loss": 1.936091423034668, "global_step": 1004, "epoch": 11, "lr": 9.999914287737561e-05}
{"train_loss": 2.048271894454956, "global_step": 1005, "epoch": 11, "lr": 9.999913947947936e-05}
{"train_loss": 2.1064224243164062, "global_step": 1006, "epoch": 11, "lr": 9.999913607486136e-05}
{"train_loss": 2.1887810230255127, "global_step": 1007, "epoch": 11, "lr": 9.999913266352157e-05}
{"train_loss": 2.199530839920044, "global_step": 1008, "epoch": 11, "lr": 9.999912924546001e-05}
{"train_loss": 2.1733386516571045, "global_step": 1009, "epoch": 11, "lr": 9.999912582067667e-05}
{"train_loss": 2.037095069885254, "global_step": 1010, "epoch": 11, "lr": 9.999912238917157e-05}
{"train_loss": 2.1819262504577637, "global_step": 1011, "epoch": 11, "lr": 9.999911895094469e-05}
{"train_loss": 2.1017262935638428, "global_step": 1012, "epoch": 11, "lr": 9.999911550599605e-05}
{"train_loss": 2.2999722957611084, "global_step": 1013, "epoch": 11, "lr": 9.999911205432562e-05}
{"train_loss": 2.180860757827759, "global_step": 1014, "epoch": 11, "lr": 9.999910859593342e-05}
{"train_loss": 2.240474224090576, "global_step": 1015, "epoch": 11, "lr": 9.999910513081947e-05}
{"train_loss": 2.3284919261932373, "global_step": 1016, "epoch": 11, "lr": 9.999910165898372e-05}
{"train_loss": 2.1790096759796143, "global_step": 1017, "epoch": 11, "lr": 9.999909818042623e-05}
{"train_loss": 2.2504987716674805, "global_step": 1018, "epoch": 11, "lr": 9.999909469514695e-05}
{"train_loss": 2.074288845062256, "global_step": 1019, "epoch": 11, "lr": 9.99990912031459e-05}
{"train_loss": 1.929377555847168, "global_step": 1020, "epoch": 11, "lr": 9.99990877044231e-05}
{"train_loss": 2.1613059043884277, "global_step": 1021, "epoch": 11, "lr": 9.999908419897852e-05}
{"train_loss": 2.3779096603393555, "global_step": 1022, "epoch": 11, "lr": 9.999908068681216e-05}
{"train_loss": 2.146683931350708, "global_step": 1023, "epoch": 11, "lr": 9.999907716792404e-05}
{"train_loss": 1.9433226585388184, "global_step": 1024, "epoch": 11, "lr": 9.999907364231417e-05}
{"train_loss": 2.452488422393799, "global_step": 1025, "epoch": 11, "lr": 9.999907010998253e-05}
{"train_loss": 2.2176766395568848, "global_step": 1026, "epoch": 11, "lr": 9.999906657092912e-05}
{"train_loss": 2.1172173023223877, "global_step": 1027, "epoch": 11, "lr": 9.999906302515393e-05}
{"train_loss": 2.1407976150512695, "global_step": 1028, "epoch": 11, "lr": 9.9999059472657e-05}
{"train_loss": 2.1731910705566406, "global_step": 1029, "epoch": 11, "lr": 9.99990559134383e-05}
{"train_loss": 2.1165411472320557, "global_step": 1030, "epoch": 11, "lr": 9.999905234749783e-05}
{"train_loss": 2.0708537101745605, "global_step": 1031, "epoch": 11, "lr": 9.999904877483559e-05}
{"train_loss": 2.2831077575683594, "global_step": 1032, "epoch": 11, "lr": 9.99990451954516e-05}
{"train_loss": 2.3957879543304443, "global_step": 1033, "epoch": 11, "lr": 9.999904160934583e-05}
{"train_loss": 2.4292407035827637, "global_step": 1034, "epoch": 11, "lr": 9.999903801651832e-05}
{"train_loss": 2.018932580947876, "global_step": 1035, "epoch": 11, "lr": 9.999903441696903e-05}
{"train_loss": 2.272286891937256, "global_step": 1036, "epoch": 11, "lr": 9.999903081069799e-05}
{"train_loss": 2.2251243591308594, "global_step": 1037, "epoch": 11, "lr": 9.999902719770519e-05}
{"train_loss": 2.137948989868164, "global_step": 1038, "epoch": 11, "lr": 9.999902357799064e-05}
{"train_loss": 2.0218703746795654, "global_step": 1039, "epoch": 11, "lr": 9.999901995155431e-05}
{"train_loss": 2.003829002380371, "global_step": 1040, "epoch": 11, "lr": 9.999901631839624e-05}
{"train_loss": 2.3611879348754883, "global_step": 1041, "epoch": 11, "lr": 9.999901267851639e-05}
{"train_loss": 2.145395278930664, "global_step": 1042, "epoch": 11, "lr": 9.99990090319148e-05}
{"train_loss": 2.119994640350342, "global_step": 1043, "epoch": 11, "lr": 9.999900537859144e-05}
{"train_loss": 2.064274549484253, "global_step": 1044, "epoch": 11, "lr": 9.999900171854634e-05}
{"train_loss": 2.124946355819702, "global_step": 1045, "epoch": 11, "lr": 9.999899805177947e-05}
{"train_loss": 2.1354360580444336, "global_step": 1046, "epoch": 11, "lr": 9.999899437829085e-05}
{"train_loss": 2.547764539718628, "global_step": 1047, "epoch": 11, "lr": 9.999899069808047e-05}
{"train_loss": 2.0081639289855957, "global_step": 1048, "epoch": 11, "lr": 9.999898701114834e-05}
{"train_loss": 2.2293498516082764, "global_step": 1049, "epoch": 11, "lr": 9.999898331749446e-05}
{"train_loss": 2.3898794651031494, "global_step": 1050, "epoch": 11, "lr": 9.999897961711882e-05}
{"train_loss": 2.0674190521240234, "global_step": 1051, "epoch": 11, "lr": 9.999897591002141e-05}
{"train_loss": 2.1651265621185303, "global_step": 1052, "epoch": 11, "lr": 9.999897219620227e-05}
{"train_loss": 2.094032049179077, "global_step": 1053, "epoch": 11, "lr": 9.999896847566138e-05}
{"train_loss": 2.172715902328491, "global_step": 1054, "epoch": 11, "lr": 9.999896474839871e-05}
{"train_loss": 2.0840847492218018, "global_step": 1055, "epoch": 11, "lr": 9.999896101441433e-05}
{"train_loss": 2.1237545013427734, "global_step": 1056, "epoch": 11, "lr": 9.999895727370816e-05}
{"train_loss": 2.359673023223877, "global_step": 1057, "epoch": 11, "lr": 9.999895352628026e-05}
{"train_loss": 2.2979729175567627, "global_step": 1058, "epoch": 11, "lr": 9.99989497721306e-05}
{"train_loss": 2.203279972076416, "global_step": 1059, "epoch": 11, "lr": 9.999894601125921e-05}
{"train_loss": 2.0661418437957764, "global_step": 1060, "epoch": 11, "lr": 9.999894224366605e-05}
{"train_loss": 2.1235439777374268, "global_step": 1061, "epoch": 11, "lr": 9.999893846935115e-05}
{"train_loss": 2.2335526943206787, "global_step": 1062, "epoch": 11, "lr": 9.99989346883145e-05}
{"train_loss": 2.2286722660064697, "global_step": 1063, "epoch": 11, "lr": 9.99989309005561e-05}
{"train_loss": 2.018336296081543, "global_step": 1064, "epoch": 11, "lr": 9.999892710607597e-05}
{"train_loss": 1.9778300523757935, "global_step": 1065, "epoch": 11, "lr": 9.999892330487409e-05}
{"train_loss": 2.0980255603790283, "global_step": 1066, "epoch": 11, "lr": 9.999891949695044e-05}
{"train_loss": 2.198113243231613, "global_step": 1067, "epoch": 11, "lr": 9.999891568230506e-05, "val_loss": 2.1325080394744873}
{"train_loss": 2.0691075325012207, "global_step": 1068, "epoch": 12, "lr": 9.999891186093793e-05}
{"train_loss": 1.9608039855957031, "global_step": 1069, "epoch": 12, "lr": 9.999890803284905e-05}
{"train_loss": 1.8448588848114014, "global_step": 1070, "epoch": 12, "lr": 9.999890419803845e-05}
{"train_loss": 1.9709619283676147, "global_step": 1071, "epoch": 12, "lr": 9.99989003565061e-05}
{"train_loss": 2.0359349250793457, "global_step": 1072, "epoch": 12, "lr": 9.999889650825201e-05}
{"train_loss": 2.000732421875, "global_step": 1073, "epoch": 12, "lr": 9.999889265327615e-05}
{"train_loss": 2.2113659381866455, "global_step": 1074, "epoch": 12, "lr": 9.999888879157857e-05}
{"train_loss": 1.9575616121292114, "global_step": 1075, "epoch": 12, "lr": 9.999888492315926e-05}
{"train_loss": 2.041853189468384, "global_step": 1076, "epoch": 12, "lr": 9.999888104801818e-05}
{"train_loss": 2.0457825660705566, "global_step": 1077, "epoch": 12, "lr": 9.999887716615539e-05}
{"train_loss": 2.067751884460449, "global_step": 1078, "epoch": 12, "lr": 9.999887327757082e-05}
{"train_loss": 2.368912935256958, "global_step": 1079, "epoch": 12, "lr": 9.999886938226454e-05}
{"train_loss": 2.2963950634002686, "global_step": 1080, "epoch": 12, "lr": 9.999886548023652e-05}
{"train_loss": 2.1963181495666504, "global_step": 1081, "epoch": 12, "lr": 9.999886157148676e-05}
{"train_loss": 2.0476038455963135, "global_step": 1082, "epoch": 12, "lr": 9.999885765601527e-05}
{"train_loss": 1.8813904523849487, "global_step": 1083, "epoch": 12, "lr": 9.999885373382203e-05}
{"train_loss": 1.980019211769104, "global_step": 1084, "epoch": 12, "lr": 9.999884980490705e-05}
{"train_loss": 2.169231414794922, "global_step": 1085, "epoch": 12, "lr": 9.999884586927035e-05}
{"train_loss": 1.9873294830322266, "global_step": 1086, "epoch": 12, "lr": 9.999884192691191e-05}
{"train_loss": 2.115790843963623, "global_step": 1087, "epoch": 12, "lr": 9.999883797783174e-05}
{"train_loss": 2.248804807662964, "global_step": 1088, "epoch": 12, "lr": 9.999883402202982e-05}
{"train_loss": 2.299102544784546, "global_step": 1089, "epoch": 12, "lr": 9.999883005950618e-05}
{"train_loss": 2.2035083770751953, "global_step": 1090, "epoch": 12, "lr": 9.999882609026079e-05}
{"train_loss": 1.9444111585617065, "global_step": 1091, "epoch": 12, "lr": 9.999882211429367e-05}
{"train_loss": 2.161794900894165, "global_step": 1092, "epoch": 12, "lr": 9.999881813160484e-05}
{"train_loss": 2.056385040283203, "global_step": 1093, "epoch": 12, "lr": 9.999881414219427e-05}
{"train_loss": 1.8836315870285034, "global_step": 1094, "epoch": 12, "lr": 9.999881014606196e-05}
{"train_loss": 2.0025174617767334, "global_step": 1095, "epoch": 12, "lr": 9.99988061432079e-05}
{"train_loss": 2.0423078536987305, "global_step": 1096, "epoch": 12, "lr": 9.999880213363215e-05}
{"train_loss": 2.1171202659606934, "global_step": 1097, "epoch": 12, "lr": 9.999879811733466e-05}
{"train_loss": 2.0046305656433105, "global_step": 1098, "epoch": 12, "lr": 9.999879409431542e-05}
{"train_loss": 2.026261329650879, "global_step": 1099, "epoch": 12, "lr": 9.999879006457446e-05}
{"train_loss": 2.0921669006347656, "global_step": 1100, "epoch": 12, "lr": 9.99987860281118e-05}
{"train_loss": 2.081411361694336, "global_step": 1101, "epoch": 12, "lr": 9.999878198492739e-05}
{"train_loss": 1.950551152229309, "global_step": 1102, "epoch": 12, "lr": 9.999877793502124e-05}
{"train_loss": 2.0760746002197266, "global_step": 1103, "epoch": 12, "lr": 9.999877387839337e-05}
{"train_loss": 2.088998556137085, "global_step": 1104, "epoch": 12, "lr": 9.99987698150438e-05}
{"train_loss": 2.1626458168029785, "global_step": 1105, "epoch": 12, "lr": 9.999876574497246e-05}
{"train_loss": 2.1457266807556152, "global_step": 1106, "epoch": 12, "lr": 9.999876166817943e-05}
{"train_loss": 2.091439723968506, "global_step": 1107, "epoch": 12, "lr": 9.999875758466467e-05}
{"train_loss": 2.022254467010498, "global_step": 1108, "epoch": 12, "lr": 9.999875349442819e-05}
{"train_loss": 1.8884625434875488, "global_step": 1109, "epoch": 12, "lr": 9.999874939746998e-05}
{"train_loss": 2.1700387001037598, "global_step": 1110, "epoch": 12, "lr": 9.999874529379004e-05}
{"train_loss": 2.0821454524993896, "global_step": 1111, "epoch": 12, "lr": 9.999874118338839e-05}
{"train_loss": 2.187255620956421, "global_step": 1112, "epoch": 12, "lr": 9.999873706626501e-05}
{"train_loss": 2.06172513961792, "global_step": 1113, "epoch": 12, "lr": 9.999873294241993e-05}
{"train_loss": 2.0586636066436768, "global_step": 1114, "epoch": 12, "lr": 9.99987288118531e-05}
{"train_loss": 1.9693912267684937, "global_step": 1115, "epoch": 12, "lr": 9.999872467456457e-05}
{"train_loss": 2.0941953659057617, "global_step": 1116, "epoch": 12, "lr": 9.999872053055432e-05}
{"train_loss": 2.046111822128296, "global_step": 1117, "epoch": 12, "lr": 9.999871637982234e-05}
{"train_loss": 1.997897982597351, "global_step": 1118, "epoch": 12, "lr": 9.999871222236865e-05}
{"train_loss": 2.0766942501068115, "global_step": 1119, "epoch": 12, "lr": 9.999870805819324e-05}
{"train_loss": 1.9426729679107666, "global_step": 1120, "epoch": 12, "lr": 9.99987038872961e-05}
{"train_loss": 2.1856133937835693, "global_step": 1121, "epoch": 12, "lr": 9.999869970967727e-05}
{"train_loss": 2.022489547729492, "global_step": 1122, "epoch": 12, "lr": 9.999869552533672e-05}
{"train_loss": 1.8431096076965332, "global_step": 1123, "epoch": 12, "lr": 9.999869133427444e-05}
{"train_loss": 2.0609874725341797, "global_step": 1124, "epoch": 12, "lr": 9.999868713649045e-05}
{"train_loss": 1.9879086017608643, "global_step": 1125, "epoch": 12, "lr": 9.999868293198474e-05}
{"train_loss": 1.910139799118042, "global_step": 1126, "epoch": 12, "lr": 9.999867872075732e-05}
{"train_loss": 2.0156495571136475, "global_step": 1127, "epoch": 12, "lr": 9.99986745028082e-05}
{"train_loss": 2.021584987640381, "global_step": 1128, "epoch": 12, "lr": 9.999867027813737e-05}
{"train_loss": 2.398913621902466, "global_step": 1129, "epoch": 12, "lr": 9.99986660467448e-05}
{"train_loss": 2.1393380165100098, "global_step": 1130, "epoch": 12, "lr": 9.999866180863054e-05}
{"train_loss": 1.997678279876709, "global_step": 1131, "epoch": 12, "lr": 9.999865756379457e-05}
{"train_loss": 2.037213087081909, "global_step": 1132, "epoch": 12, "lr": 9.999865331223688e-05}
{"train_loss": 2.0969202518463135, "global_step": 1133, "epoch": 12, "lr": 9.999864905395749e-05}
{"train_loss": 2.078228235244751, "global_step": 1134, "epoch": 12, "lr": 9.999864478895639e-05}
{"train_loss": 2.2073159217834473, "global_step": 1135, "epoch": 12, "lr": 9.999864051723357e-05}
{"train_loss": 2.1732332706451416, "global_step": 1136, "epoch": 12, "lr": 9.999863623878904e-05}
{"train_loss": 1.9271701574325562, "global_step": 1137, "epoch": 12, "lr": 9.999863195362283e-05}
{"train_loss": 2.2858238220214844, "global_step": 1138, "epoch": 12, "lr": 9.999862766173488e-05}
{"train_loss": 2.0283122062683105, "global_step": 1139, "epoch": 12, "lr": 9.999862336312525e-05}
{"train_loss": 1.9279353618621826, "global_step": 1140, "epoch": 12, "lr": 9.99986190577939e-05}
{"train_loss": 1.9521092176437378, "global_step": 1141, "epoch": 12, "lr": 9.999861474574086e-05}
{"train_loss": 2.134199619293213, "global_step": 1142, "epoch": 12, "lr": 9.99986104269661e-05}
{"train_loss": 2.0410704612731934, "global_step": 1143, "epoch": 12, "lr": 9.999860610146965e-05}
{"train_loss": 2.0912020206451416, "global_step": 1144, "epoch": 12, "lr": 9.999860176925148e-05}
{"train_loss": 2.1513543128967285, "global_step": 1145, "epoch": 12, "lr": 9.999859743031162e-05}
{"train_loss": 2.1121671199798584, "global_step": 1146, "epoch": 12, "lr": 9.999859308465005e-05}
{"train_loss": 2.139505624771118, "global_step": 1147, "epoch": 12, "lr": 9.999858873226678e-05}
{"train_loss": 2.0957882404327393, "global_step": 1148, "epoch": 12, "lr": 9.999858437316181e-05}
{"train_loss": 2.085317611694336, "global_step": 1149, "epoch": 12, "lr": 9.999858000733516e-05}
{"train_loss": 2.081958055496216, "global_step": 1150, "epoch": 12, "lr": 9.999857563478679e-05}
{"train_loss": 2.0702412128448486, "global_step": 1151, "epoch": 12, "lr": 9.99985712555167e-05}
{"train_loss": 2.073474884033203, "global_step": 1152, "epoch": 12, "lr": 9.999856686952494e-05}
{"train_loss": 2.103012800216675, "global_step": 1153, "epoch": 12, "lr": 9.999856247681148e-05}
{"train_loss": 2.0534915924072266, "global_step": 1154, "epoch": 12, "lr": 9.999855807737633e-05}
{"train_loss": 2.17451810836792, "global_step": 1155, "epoch": 12, "lr": 9.999855367121947e-05}
{"train_loss": 2.0739820740196144, "global_step": 1156, "epoch": 12, "lr": 9.999854925834092e-05, "val_loss": 2.080850601196289}
{"train_loss": 2.0128509998321533, "global_step": 1157, "epoch": 13, "lr": 9.999854483874067e-05}
{"train_loss": 2.052471876144409, "global_step": 1158, "epoch": 13, "lr": 9.999854041241871e-05}
{"train_loss": 2.142854690551758, "global_step": 1159, "epoch": 13, "lr": 9.999853597937507e-05}
{"train_loss": 2.0152487754821777, "global_step": 1160, "epoch": 13, "lr": 9.999853153960975e-05}
{"train_loss": 2.0955984592437744, "global_step": 1161, "epoch": 13, "lr": 9.999852709312272e-05}
{"train_loss": 2.0820913314819336, "global_step": 1162, "epoch": 13, "lr": 9.9998522639914e-05}
{"train_loss": 2.180718183517456, "global_step": 1163, "epoch": 13, "lr": 9.999851817998358e-05}
{"train_loss": 1.9333029985427856, "global_step": 1164, "epoch": 13, "lr": 9.999851371333149e-05}
{"train_loss": 2.0593831539154053, "global_step": 1165, "epoch": 13, "lr": 9.99985092399577e-05}
{"train_loss": 1.9498146772384644, "global_step": 1166, "epoch": 13, "lr": 9.99985047598622e-05}
{"train_loss": 2.078366279602051, "global_step": 1167, "epoch": 13, "lr": 9.999850027304504e-05}
{"train_loss": 2.2065958976745605, "global_step": 1168, "epoch": 13, "lr": 9.999849577950617e-05}
{"train_loss": 2.034027576446533, "global_step": 1169, "epoch": 13, "lr": 9.999849127924564e-05}
{"train_loss": 1.7903780937194824, "global_step": 1170, "epoch": 13, "lr": 9.99984867722634e-05}
{"train_loss": 2.006056547164917, "global_step": 1171, "epoch": 13, "lr": 9.999848225855946e-05}
{"train_loss": 2.082601547241211, "global_step": 1172, "epoch": 13, "lr": 9.999847773813386e-05}
{"train_loss": 2.3263182640075684, "global_step": 1173, "epoch": 13, "lr": 9.999847321098655e-05}
{"train_loss": 1.8740421533584595, "global_step": 1174, "epoch": 13, "lr": 9.999846867711757e-05}
{"train_loss": 1.9981803894042969, "global_step": 1175, "epoch": 13, "lr": 9.999846413652691e-05}
{"train_loss": 1.9466410875320435, "global_step": 1176, "epoch": 13, "lr": 9.999845958921456e-05}
{"train_loss": 2.0948870182037354, "global_step": 1177, "epoch": 13, "lr": 9.999845503518053e-05}
{"train_loss": 1.946022868156433, "global_step": 1178, "epoch": 13, "lr": 9.99984504744248e-05}
{"train_loss": 2.0643646717071533, "global_step": 1179, "epoch": 13, "lr": 9.999844590694741e-05}
{"train_loss": 1.9847640991210938, "global_step": 1180, "epoch": 13, "lr": 9.999844133274831e-05}
{"train_loss": 2.0332369804382324, "global_step": 1181, "epoch": 13, "lr": 9.999843675182756e-05}
{"train_loss": 2.041874885559082, "global_step": 1182, "epoch": 13, "lr": 9.999843216418512e-05}
{"train_loss": 2.081791639328003, "global_step": 1183, "epoch": 13, "lr": 9.9998427569821e-05}
{"train_loss": 2.022395133972168, "global_step": 1184, "epoch": 13, "lr": 9.999842296873519e-05}
{"train_loss": 1.8096481561660767, "global_step": 1185, "epoch": 13, "lr": 9.999841836092772e-05}
{"train_loss": 1.9968762397766113, "global_step": 1186, "epoch": 13, "lr": 9.999841374639855e-05}
{"train_loss": 2.030895233154297, "global_step": 1187, "epoch": 13, "lr": 9.999840912514771e-05}
{"train_loss": 2.0006601810455322, "global_step": 1188, "epoch": 13, "lr": 9.99984044971752e-05}
{"train_loss": 2.0207159519195557, "global_step": 1189, "epoch": 13, "lr": 9.999839986248102e-05}
{"train_loss": 1.9186407327651978, "global_step": 1190, "epoch": 13, "lr": 9.999839522106516e-05}
{"train_loss": 1.9711244106292725, "global_step": 1191, "epoch": 13, "lr": 9.999839057292761e-05}
{"train_loss": 1.7794826030731201, "global_step": 1192, "epoch": 13, "lr": 9.99983859180684e-05}
{"train_loss": 1.8979127407073975, "global_step": 1193, "epoch": 13, "lr": 9.999838125648751e-05}
{"train_loss": 2.0216779708862305, "global_step": 1194, "epoch": 13, "lr": 9.999837658818497e-05}
{"train_loss": 2.068880796432495, "global_step": 1195, "epoch": 13, "lr": 9.999837191316072e-05}
{"train_loss": 1.655266284942627, "global_step": 1196, "epoch": 13, "lr": 9.999836723141482e-05}
{"train_loss": 2.063518762588501, "global_step": 1197, "epoch": 13, "lr": 9.999836254294724e-05}
{"train_loss": 2.095677137374878, "global_step": 1198, "epoch": 13, "lr": 9.999835784775799e-05}
{"train_loss": 1.9492771625518799, "global_step": 1199, "epoch": 13, "lr": 9.999835314584709e-05}
{"train_loss": 1.957528829574585, "global_step": 1200, "epoch": 13, "lr": 9.999834843721449e-05}
{"train_loss": 1.756853699684143, "global_step": 1201, "epoch": 13, "lr": 9.999834372186025e-05}
{"train_loss": 2.181386709213257, "global_step": 1202, "epoch": 13, "lr": 9.999833899978433e-05}
{"train_loss": 2.272120952606201, "global_step": 1203, "epoch": 13, "lr": 9.999833427098673e-05}
{"train_loss": 2.0276637077331543, "global_step": 1204, "epoch": 13, "lr": 9.999832953546748e-05}
{"train_loss": 1.944008708000183, "global_step": 1205, "epoch": 13, "lr": 9.999832479322656e-05}
{"train_loss": 2.07686710357666, "global_step": 1206, "epoch": 13, "lr": 9.999832004426397e-05}
{"train_loss": 1.9020518064498901, "global_step": 1207, "epoch": 13, "lr": 9.999831528857972e-05}
{"train_loss": 2.187669277191162, "global_step": 1208, "epoch": 13, "lr": 9.999831052617381e-05}
{"train_loss": 2.1597695350646973, "global_step": 1209, "epoch": 13, "lr": 9.999830575704623e-05}
{"train_loss": 1.8324358463287354, "global_step": 1210, "epoch": 13, "lr": 9.9998300981197e-05}
{"train_loss": 2.1185269355773926, "global_step": 1211, "epoch": 13, "lr": 9.999829619862608e-05}
{"train_loss": 1.7525874376296997, "global_step": 1212, "epoch": 13, "lr": 9.999829140933353e-05}
{"train_loss": 1.9041869640350342, "global_step": 1213, "epoch": 13, "lr": 9.999828661331929e-05}
{"train_loss": 2.0121586322784424, "global_step": 1214, "epoch": 13, "lr": 9.999828181058341e-05}
{"train_loss": 2.014112949371338, "global_step": 1215, "epoch": 13, "lr": 9.999827700112587e-05}
{"train_loss": 2.1022567749023438, "global_step": 1216, "epoch": 13, "lr": 9.999827218494666e-05}
{"train_loss": 1.953566551208496, "global_step": 1217, "epoch": 13, "lr": 9.999826736204579e-05}
{"train_loss": 1.9218477010726929, "global_step": 1218, "epoch": 13, "lr": 9.999826253242328e-05}
{"train_loss": 1.9976518154144287, "global_step": 1219, "epoch": 13, "lr": 9.99982576960791e-05}
{"train_loss": 2.073216438293457, "global_step": 1220, "epoch": 13, "lr": 9.999825285301327e-05}
{"train_loss": 1.930275797843933, "global_step": 1221, "epoch": 13, "lr": 9.999824800322577e-05}
{"train_loss": 1.892163634300232, "global_step": 1222, "epoch": 13, "lr": 9.999824314671663e-05}
{"train_loss": 2.2986199855804443, "global_step": 1223, "epoch": 13, "lr": 9.999823828348583e-05}
{"train_loss": 1.8836029767990112, "global_step": 1224, "epoch": 13, "lr": 9.999823341353339e-05}
{"train_loss": 1.9673634767532349, "global_step": 1225, "epoch": 13, "lr": 9.999822853685928e-05}
{"train_loss": 1.9239261150360107, "global_step": 1226, "epoch": 13, "lr": 9.999822365346352e-05}
{"train_loss": 2.26037859916687, "global_step": 1227, "epoch": 13, "lr": 9.99982187633461e-05}
{"train_loss": 1.968177318572998, "global_step": 1228, "epoch": 13, "lr": 9.999821386650705e-05}
{"train_loss": 1.9569079875946045, "global_step": 1229, "epoch": 13, "lr": 9.999820896294634e-05}
{"train_loss": 1.805978536605835, "global_step": 1230, "epoch": 13, "lr": 9.999820405266397e-05}
{"train_loss": 1.9825780391693115, "global_step": 1231, "epoch": 13, "lr": 9.999819913565995e-05}
{"train_loss": 1.8255829811096191, "global_step": 1232, "epoch": 13, "lr": 9.999819421193431e-05}
{"train_loss": 1.874752402305603, "global_step": 1233, "epoch": 13, "lr": 9.999818928148698e-05}
{"train_loss": 1.7512149810791016, "global_step": 1234, "epoch": 13, "lr": 9.999818434431805e-05}
{"train_loss": 1.9225478172302246, "global_step": 1235, "epoch": 13, "lr": 9.999817940042744e-05}
{"train_loss": 1.9987051486968994, "global_step": 1236, "epoch": 13, "lr": 9.999817444981519e-05}
{"train_loss": 1.6989572048187256, "global_step": 1237, "epoch": 13, "lr": 9.99981694924813e-05}
{"train_loss": 1.7194017171859741, "global_step": 1238, "epoch": 13, "lr": 9.999816452842577e-05}
{"train_loss": 2.0772674083709717, "global_step": 1239, "epoch": 13, "lr": 9.999815955764859e-05}
{"train_loss": 1.7898926734924316, "global_step": 1240, "epoch": 13, "lr": 9.999815458014976e-05}
{"train_loss": 1.955659031867981, "global_step": 1241, "epoch": 13, "lr": 9.99981495959293e-05}
{"train_loss": 1.9859927892684937, "global_step": 1242, "epoch": 13, "lr": 9.999814460498719e-05}
{"train_loss": 1.8572428226470947, "global_step": 1243, "epoch": 13, "lr": 9.999813960732344e-05}
{"train_loss": 1.859869122505188, "global_step": 1244, "epoch": 13, "lr": 9.999813460293805e-05}
{"train_loss": 1.98578301574407, "global_step": 1245, "epoch": 13, "lr": 9.999812959183103e-05, "val_loss": 1.8842054605484009}
{"train_loss": 1.8159030675888062, "global_step": 1246, "epoch": 14, "lr": 9.999812457400235e-05}
{"train_loss": 1.9006515741348267, "global_step": 1247, "epoch": 14, "lr": 9.999811954945204e-05}
{"train_loss": 1.9390839338302612, "global_step": 1248, "epoch": 14, "lr": 9.999811451818011e-05}
{"train_loss": 1.6076475381851196, "global_step": 1249, "epoch": 14, "lr": 9.999810948018653e-05}
{"train_loss": 2.0255117416381836, "global_step": 1250, "epoch": 14, "lr": 9.999810443547131e-05}
{"train_loss": 1.8205808401107788, "global_step": 1251, "epoch": 14, "lr": 9.999809938403445e-05}
{"train_loss": 1.7554348707199097, "global_step": 1252, "epoch": 14, "lr": 9.999809432587597e-05}
{"train_loss": 1.7250922918319702, "global_step": 1253, "epoch": 14, "lr": 9.999808926099584e-05}
{"train_loss": 1.9074137210845947, "global_step": 1254, "epoch": 14, "lr": 9.999808418939409e-05}
{"train_loss": 1.9721770286560059, "global_step": 1255, "epoch": 14, "lr": 9.99980791110707e-05}
{"train_loss": 2.0952820777893066, "global_step": 1256, "epoch": 14, "lr": 9.999807402602568e-05}
{"train_loss": 1.8081635236740112, "global_step": 1257, "epoch": 14, "lr": 9.999806893425903e-05}
{"train_loss": 1.9097239971160889, "global_step": 1258, "epoch": 14, "lr": 9.999806383577072e-05}
{"train_loss": 1.9047855138778687, "global_step": 1259, "epoch": 14, "lr": 9.999805873056082e-05}
{"train_loss": 1.7911497354507446, "global_step": 1260, "epoch": 14, "lr": 9.999805361862927e-05}
{"train_loss": 1.8168195486068726, "global_step": 1261, "epoch": 14, "lr": 9.999804849997609e-05}
{"train_loss": 1.7428745031356812, "global_step": 1262, "epoch": 14, "lr": 9.99980433746013e-05}
{"train_loss": 1.7901183366775513, "global_step": 1263, "epoch": 14, "lr": 9.999803824250487e-05}
{"train_loss": 1.8020743131637573, "global_step": 1264, "epoch": 14, "lr": 9.999803310368681e-05}
{"train_loss": 2.003631114959717, "global_step": 1265, "epoch": 14, "lr": 9.999802795814714e-05}
{"train_loss": 1.9528241157531738, "global_step": 1266, "epoch": 14, "lr": 9.999802280588583e-05}
{"train_loss": 1.9019826650619507, "global_step": 1267, "epoch": 14, "lr": 9.999801764690289e-05}
{"train_loss": 2.034045696258545, "global_step": 1268, "epoch": 14, "lr": 9.999801248119834e-05}
{"train_loss": 2.068699598312378, "global_step": 1269, "epoch": 14, "lr": 9.999800730877216e-05}
{"train_loss": 1.8352797031402588, "global_step": 1270, "epoch": 14, "lr": 9.999800212962436e-05}
{"train_loss": 1.9621555805206299, "global_step": 1271, "epoch": 14, "lr": 9.999799694375494e-05}
{"train_loss": 1.850036382675171, "global_step": 1272, "epoch": 14, "lr": 9.999799175116388e-05}
{"train_loss": 2.0931429862976074, "global_step": 1273, "epoch": 14, "lr": 9.999798655185122e-05}
{"train_loss": 2.015517473220825, "global_step": 1274, "epoch": 14, "lr": 9.999798134581693e-05}
{"train_loss": 1.9467800855636597, "global_step": 1275, "epoch": 14, "lr": 9.999797613306102e-05}
{"train_loss": 2.1399662494659424, "global_step": 1276, "epoch": 14, "lr": 9.99979709135835e-05}
{"train_loss": 1.9352445602416992, "global_step": 1277, "epoch": 14, "lr": 9.999796568738436e-05}
{"train_loss": 1.915698766708374, "global_step": 1278, "epoch": 14, "lr": 9.99979604544636e-05}
{"train_loss": 1.886663794517517, "global_step": 1279, "epoch": 14, "lr": 9.999795521482122e-05}
{"train_loss": 1.8263816833496094, "global_step": 1280, "epoch": 14, "lr": 9.999794996845724e-05}
{"train_loss": 1.7957587242126465, "global_step": 1281, "epoch": 14, "lr": 9.999794471537163e-05}
{"train_loss": 1.8139077425003052, "global_step": 1282, "epoch": 14, "lr": 9.99979394555644e-05}
{"train_loss": 1.7777436971664429, "global_step": 1283, "epoch": 14, "lr": 9.999793418903557e-05}
{"train_loss": 1.7909724712371826, "global_step": 1284, "epoch": 14, "lr": 9.999792891578512e-05}
{"train_loss": 1.7271417379379272, "global_step": 1285, "epoch": 14, "lr": 9.999792363581307e-05}
{"train_loss": 1.9118386507034302, "global_step": 1286, "epoch": 14, "lr": 9.99979183491194e-05}
{"train_loss": 2.127676010131836, "global_step": 1287, "epoch": 14, "lr": 9.999791305570412e-05}
{"train_loss": 1.9551068544387817, "global_step": 1288, "epoch": 14, "lr": 9.999790775556722e-05}
{"train_loss": 1.898418664932251, "global_step": 1289, "epoch": 14, "lr": 9.999790244870872e-05}
{"train_loss": 1.8494114875793457, "global_step": 1290, "epoch": 14, "lr": 9.999789713512862e-05}
{"train_loss": 1.806560754776001, "global_step": 1291, "epoch": 14, "lr": 9.99978918148269e-05}
{"train_loss": 2.088773012161255, "global_step": 1292, "epoch": 14, "lr": 9.999788648780357e-05}
{"train_loss": 1.8463537693023682, "global_step": 1293, "epoch": 14, "lr": 9.999788115405864e-05}
{"train_loss": 1.7932456731796265, "global_step": 1294, "epoch": 14, "lr": 9.99978758135921e-05}
{"train_loss": 1.9444841146469116, "global_step": 1295, "epoch": 14, "lr": 9.999787046640397e-05}
{"train_loss": 1.6664303541183472, "global_step": 1296, "epoch": 14, "lr": 9.999786511249423e-05}
{"train_loss": 1.9040776491165161, "global_step": 1297, "epoch": 14, "lr": 9.999785975186287e-05}
{"train_loss": 1.9292720556259155, "global_step": 1298, "epoch": 14, "lr": 9.999785438450993e-05}
{"train_loss": 1.683323621749878, "global_step": 1299, "epoch": 14, "lr": 9.999784901043536e-05}
{"train_loss": 1.7545462846755981, "global_step": 1300, "epoch": 14, "lr": 9.999784362963921e-05}
{"train_loss": 2.0001220703125, "global_step": 1301, "epoch": 14, "lr": 9.999783824212145e-05}
{"train_loss": 1.8470944166183472, "global_step": 1302, "epoch": 14, "lr": 9.999783284788209e-05}
{"train_loss": 1.8750741481781006, "global_step": 1303, "epoch": 14, "lr": 9.999782744692115e-05}
{"train_loss": 1.7196965217590332, "global_step": 1304, "epoch": 14, "lr": 9.999782203923858e-05}
{"train_loss": 1.8010737895965576, "global_step": 1305, "epoch": 14, "lr": 9.999781662483443e-05}
{"train_loss": 1.8962005376815796, "global_step": 1306, "epoch": 14, "lr": 9.999781120370868e-05}
{"train_loss": 1.8920347690582275, "global_step": 1307, "epoch": 14, "lr": 9.999780577586134e-05}
{"train_loss": 1.8427517414093018, "global_step": 1308, "epoch": 14, "lr": 9.99978003412924e-05}
{"train_loss": 1.6426793336868286, "global_step": 1309, "epoch": 14, "lr": 9.999779490000187e-05}
{"train_loss": 1.6986585855484009, "global_step": 1310, "epoch": 14, "lr": 9.999778945198974e-05}
{"train_loss": 1.9415146112442017, "global_step": 1311, "epoch": 14, "lr": 9.999778399725602e-05}
{"train_loss": 1.8131014108657837, "global_step": 1312, "epoch": 14, "lr": 9.99977785358007e-05}
{"train_loss": 1.7779271602630615, "global_step": 1313, "epoch": 14, "lr": 9.999777306762379e-05}
{"train_loss": 1.7335580587387085, "global_step": 1314, "epoch": 14, "lr": 9.999776759272529e-05}
{"train_loss": 1.6805311441421509, "global_step": 1315, "epoch": 14, "lr": 9.999776211110519e-05}
{"train_loss": 1.8631993532180786, "global_step": 1316, "epoch": 14, "lr": 9.999775662276352e-05}
{"train_loss": 1.8019341230392456, "global_step": 1317, "epoch": 14, "lr": 9.999775112770025e-05}
{"train_loss": 1.7840626239776611, "global_step": 1318, "epoch": 14, "lr": 9.999774562591539e-05}
{"train_loss": 1.7532334327697754, "global_step": 1319, "epoch": 14, "lr": 9.999774011740894e-05}
{"train_loss": 1.5922942161560059, "global_step": 1320, "epoch": 14, "lr": 9.999773460218091e-05}
{"train_loss": 1.8332901000976562, "global_step": 1321, "epoch": 14, "lr": 9.999772908023131e-05}
{"train_loss": 1.7090932130813599, "global_step": 1322, "epoch": 14, "lr": 9.99977235515601e-05}
{"train_loss": 1.8064061403274536, "global_step": 1323, "epoch": 14, "lr": 9.999771801616731e-05}
{"train_loss": 1.952811598777771, "global_step": 1324, "epoch": 14, "lr": 9.999771247405294e-05}
{"train_loss": 1.8945996761322021, "global_step": 1325, "epoch": 14, "lr": 9.9997706925217e-05}
{"train_loss": 1.5938538312911987, "global_step": 1326, "epoch": 14, "lr": 9.999770136965945e-05}
{"train_loss": 1.7287846803665161, "global_step": 1327, "epoch": 14, "lr": 9.999769580738034e-05}
{"train_loss": 1.8886017799377441, "global_step": 1328, "epoch": 14, "lr": 9.999769023837963e-05}
{"train_loss": 1.8987537622451782, "global_step": 1329, "epoch": 14, "lr": 9.999768466265736e-05}
{"train_loss": 1.7246160507202148, "global_step": 1330, "epoch": 14, "lr": 9.999767908021351e-05}
{"train_loss": 1.8584526777267456, "global_step": 1331, "epoch": 14, "lr": 9.999767349104807e-05}
{"train_loss": 1.7261617183685303, "global_step": 1332, "epoch": 14, "lr": 9.999766789516105e-05}
{"train_loss": 1.7876092195510864, "global_step": 1333, "epoch": 14, "lr": 9.999766229255246e-05}
{"train_loss": 1.852149004346869, "global_step": 1334, "epoch": 14, "lr": 9.99976566832223e-05, "val_loss": 1.7990808486938477}
{"train_loss": 1.765041470527649, "global_step": 1335, "epoch": 15, "lr": 9.999765106717057e-05}
{"train_loss": 1.9067542552947998, "global_step": 1336, "epoch": 15, "lr": 9.999764544439724e-05}
{"train_loss": 1.7181849479675293, "global_step": 1337, "epoch": 15, "lr": 9.999763981490234e-05}
{"train_loss": 1.692091941833496, "global_step": 1338, "epoch": 15, "lr": 9.999763417868588e-05}
{"train_loss": 1.7086158990859985, "global_step": 1339, "epoch": 15, "lr": 9.999762853574785e-05}
{"train_loss": 2.015737295150757, "global_step": 1340, "epoch": 15, "lr": 9.999762288608823e-05}
{"train_loss": 2.0912506580352783, "global_step": 1341, "epoch": 15, "lr": 9.999761722970705e-05}
{"train_loss": 1.6945652961730957, "global_step": 1342, "epoch": 15, "lr": 9.99976115666043e-05}
{"train_loss": 1.6001484394073486, "global_step": 1343, "epoch": 15, "lr": 9.999760589677997e-05}
{"train_loss": 1.9088938236236572, "global_step": 1344, "epoch": 15, "lr": 9.99976002202341e-05}
{"train_loss": 1.5580787658691406, "global_step": 1345, "epoch": 15, "lr": 9.999759453696664e-05}
{"train_loss": 1.8226487636566162, "global_step": 1346, "epoch": 15, "lr": 9.99975888469776e-05}
{"train_loss": 1.8765836954116821, "global_step": 1347, "epoch": 15, "lr": 9.999758315026703e-05}
{"train_loss": 1.7821924686431885, "global_step": 1348, "epoch": 15, "lr": 9.999757744683487e-05}
{"train_loss": 1.836251139640808, "global_step": 1349, "epoch": 15, "lr": 9.999757173668115e-05}
{"train_loss": 1.833222508430481, "global_step": 1350, "epoch": 15, "lr": 9.999756601980587e-05}
{"train_loss": 1.735174536705017, "global_step": 1351, "epoch": 15, "lr": 9.999756029620902e-05}
{"train_loss": 1.741690754890442, "global_step": 1352, "epoch": 15, "lr": 9.99975545658906e-05}
{"train_loss": 1.651558518409729, "global_step": 1353, "epoch": 15, "lr": 9.999754882885062e-05}
{"train_loss": 1.6741000413894653, "global_step": 1354, "epoch": 15, "lr": 9.999754308508909e-05}
{"train_loss": 1.8928790092468262, "global_step": 1355, "epoch": 15, "lr": 9.9997537334606e-05}
{"train_loss": 1.8424142599105835, "global_step": 1356, "epoch": 15, "lr": 9.999753157740134e-05}
{"train_loss": 1.676076054573059, "global_step": 1357, "epoch": 15, "lr": 9.999752581347513e-05}
{"train_loss": 1.6080378293991089, "global_step": 1358, "epoch": 15, "lr": 9.999752004282736e-05}
{"train_loss": 1.9079166650772095, "global_step": 1359, "epoch": 15, "lr": 9.999751426545804e-05}
{"train_loss": 1.7763124704360962, "global_step": 1360, "epoch": 15, "lr": 9.999750848136716e-05}
{"train_loss": 1.782739281654358, "global_step": 1361, "epoch": 15, "lr": 9.999750269055472e-05}
{"train_loss": 1.9302239418029785, "global_step": 1362, "epoch": 15, "lr": 9.999749689302072e-05}
{"train_loss": 1.9242522716522217, "global_step": 1363, "epoch": 15, "lr": 9.999749108876518e-05}
{"train_loss": 1.976958990097046, "global_step": 1364, "epoch": 15, "lr": 9.999748527778807e-05}
{"train_loss": 1.6211508512496948, "global_step": 1365, "epoch": 15, "lr": 9.999747946008941e-05}
{"train_loss": 1.6274516582489014, "global_step": 1366, "epoch": 15, "lr": 9.99974736356692e-05}
{"train_loss": 1.8358319997787476, "global_step": 1367, "epoch": 15, "lr": 9.999746780452745e-05}
{"train_loss": 1.7204484939575195, "global_step": 1368, "epoch": 15, "lr": 9.999746196666415e-05}
{"train_loss": 1.8075200319290161, "global_step": 1369, "epoch": 15, "lr": 9.999745612207929e-05}
{"train_loss": 1.7648444175720215, "global_step": 1370, "epoch": 15, "lr": 9.999745027077288e-05}
{"train_loss": 2.019575834274292, "global_step": 1371, "epoch": 15, "lr": 9.999744441274494e-05}
{"train_loss": 1.6119744777679443, "global_step": 1372, "epoch": 15, "lr": 9.999743854799544e-05}
{"train_loss": 1.8169746398925781, "global_step": 1373, "epoch": 15, "lr": 9.99974326765244e-05}
{"train_loss": 1.9672496318817139, "global_step": 1374, "epoch": 15, "lr": 9.99974267983318e-05}
{"train_loss": 1.894054889678955, "global_step": 1375, "epoch": 15, "lr": 9.999742091341766e-05}
{"train_loss": 2.1779513359069824, "global_step": 1376, "epoch": 15, "lr": 9.999741502178198e-05}
{"train_loss": 1.8260139226913452, "global_step": 1377, "epoch": 15, "lr": 9.999740912342475e-05}
{"train_loss": 2.0963542461395264, "global_step": 1378, "epoch": 15, "lr": 9.999740321834599e-05}
{"train_loss": 2.0911037921905518, "global_step": 1379, "epoch": 15, "lr": 9.999739730654568e-05}
{"train_loss": 1.86184561252594, "global_step": 1380, "epoch": 15, "lr": 9.999739138802384e-05}
{"train_loss": 1.4576261043548584, "global_step": 1381, "epoch": 15, "lr": 9.999738546278044e-05}
{"train_loss": 1.7518978118896484, "global_step": 1382, "epoch": 15, "lr": 9.999737953081552e-05}
{"train_loss": 1.9970300197601318, "global_step": 1383, "epoch": 15, "lr": 9.999737359212906e-05}
{"train_loss": 1.7412667274475098, "global_step": 1384, "epoch": 15, "lr": 9.999736764672105e-05}
{"train_loss": 1.7853392362594604, "global_step": 1385, "epoch": 15, "lr": 9.999736169459151e-05}
{"train_loss": 1.8043516874313354, "global_step": 1386, "epoch": 15, "lr": 9.999735573574044e-05}
{"train_loss": 1.8097662925720215, "global_step": 1387, "epoch": 15, "lr": 9.999734977016784e-05}
{"train_loss": 1.5668036937713623, "global_step": 1388, "epoch": 15, "lr": 9.999734379787369e-05}
{"train_loss": 1.934090256690979, "global_step": 1389, "epoch": 15, "lr": 9.999733781885801e-05}
{"train_loss": 1.849361777305603, "global_step": 1390, "epoch": 15, "lr": 9.99973318331208e-05}
{"train_loss": 2.0002615451812744, "global_step": 1391, "epoch": 15, "lr": 9.999732584066205e-05}
{"train_loss": 1.810307264328003, "global_step": 1392, "epoch": 15, "lr": 9.999731984148179e-05}
{"train_loss": 1.738730549812317, "global_step": 1393, "epoch": 15, "lr": 9.999731383557997e-05}
{"train_loss": 1.652894139289856, "global_step": 1394, "epoch": 15, "lr": 9.999730782295664e-05}
{"train_loss": 1.944463849067688, "global_step": 1395, "epoch": 15, "lr": 9.999730180361178e-05}
{"train_loss": 1.6718615293502808, "global_step": 1396, "epoch": 15, "lr": 9.99972957775454e-05}
{"train_loss": 1.7758170366287231, "global_step": 1397, "epoch": 15, "lr": 9.999728974475748e-05}
{"train_loss": 1.5524784326553345, "global_step": 1398, "epoch": 15, "lr": 9.999728370524804e-05}
{"train_loss": 1.6390689611434937, "global_step": 1399, "epoch": 15, "lr": 9.999727765901706e-05}
{"train_loss": 1.7474392652511597, "global_step": 1400, "epoch": 15, "lr": 9.999727160606458e-05}
{"train_loss": 1.6840718984603882, "global_step": 1401, "epoch": 15, "lr": 9.999726554639055e-05}
{"train_loss": 1.729576587677002, "global_step": 1402, "epoch": 15, "lr": 9.999725947999501e-05}
{"train_loss": 1.5984526872634888, "global_step": 1403, "epoch": 15, "lr": 9.999725340687796e-05}
{"train_loss": 1.6786010265350342, "global_step": 1404, "epoch": 15, "lr": 9.999724732703937e-05}
{"train_loss": 1.9473587274551392, "global_step": 1405, "epoch": 15, "lr": 9.999724124047926e-05}
{"train_loss": 1.972167730331421, "global_step": 1406, "epoch": 15, "lr": 9.999723514719764e-05}
{"train_loss": 1.5990569591522217, "global_step": 1407, "epoch": 15, "lr": 9.999722904719451e-05}
{"train_loss": 1.7862590551376343, "global_step": 1408, "epoch": 15, "lr": 9.999722294046985e-05}
{"train_loss": 1.8425610065460205, "global_step": 1409, "epoch": 15, "lr": 9.999721682702366e-05}
{"train_loss": 1.8337953090667725, "global_step": 1410, "epoch": 15, "lr": 9.999721070685597e-05}
{"train_loss": 1.7825151681900024, "global_step": 1411, "epoch": 15, "lr": 9.999720457996675e-05}
{"train_loss": 1.768222451210022, "global_step": 1412, "epoch": 15, "lr": 9.999719844635604e-05}
{"train_loss": 1.8168418407440186, "global_step": 1413, "epoch": 15, "lr": 9.99971923060238e-05}
{"train_loss": 1.8401836156845093, "global_step": 1414, "epoch": 15, "lr": 9.999718615897005e-05}
{"train_loss": 1.6558668613433838, "global_step": 1415, "epoch": 15, "lr": 9.999718000519478e-05}
{"train_loss": 1.7142800092697144, "global_step": 1416, "epoch": 15, "lr": 9.9997173844698e-05}
{"train_loss": 1.6772449016571045, "global_step": 1417, "epoch": 15, "lr": 9.999716767747971e-05}
{"train_loss": 1.689442753791809, "global_step": 1418, "epoch": 15, "lr": 9.999716150353992e-05}
{"train_loss": 1.9482072591781616, "global_step": 1419, "epoch": 15, "lr": 9.999715532287862e-05}
{"train_loss": 1.7088600397109985, "global_step": 1420, "epoch": 15, "lr": 9.999714913549581e-05}
{"train_loss": 1.611402988433838, "global_step": 1421, "epoch": 15, "lr": 9.999714294139148e-05}
{"train_loss": 1.7252765893936157, "global_step": 1422, "epoch": 15, "lr": 9.999713674056566e-05}
{"train_loss": 1.7921871964851122, "global_step": 1423, "epoch": 15, "lr": 9.999713053301834e-05, "val_loss": 1.7255827188491821, "train_action_mse_error": 112.15298461914062}
{"train_loss": 1.820627212524414, "global_step": 1424, "epoch": 16, "lr": 9.99971243187495e-05}
{"train_loss": 1.6737592220306396, "global_step": 1425, "epoch": 16, "lr": 9.999711809775916e-05}
{"train_loss": 1.7390413284301758, "global_step": 1426, "epoch": 16, "lr": 9.99971118700473e-05}
{"train_loss": 1.7104989290237427, "global_step": 1427, "epoch": 16, "lr": 9.999710563561396e-05}
{"train_loss": 1.7493432760238647, "global_step": 1428, "epoch": 16, "lr": 9.999709939445912e-05}
{"train_loss": 1.6907292604446411, "global_step": 1429, "epoch": 16, "lr": 9.999709314658277e-05}
{"train_loss": 1.9297606945037842, "global_step": 1430, "epoch": 16, "lr": 9.999708689198491e-05}
{"train_loss": 1.6130009889602661, "global_step": 1431, "epoch": 16, "lr": 9.999708063066557e-05}
{"train_loss": 1.7070419788360596, "global_step": 1432, "epoch": 16, "lr": 9.999707436262473e-05}
{"train_loss": 1.7195574045181274, "global_step": 1433, "epoch": 16, "lr": 9.999706808786237e-05}
{"train_loss": 1.644057035446167, "global_step": 1434, "epoch": 16, "lr": 9.999706180637854e-05}
{"train_loss": 1.7450920343399048, "global_step": 1435, "epoch": 16, "lr": 9.99970555181732e-05}
{"train_loss": 1.6220604181289673, "global_step": 1436, "epoch": 16, "lr": 9.999704922324638e-05}
{"train_loss": 1.947785496711731, "global_step": 1437, "epoch": 16, "lr": 9.999704292159805e-05}
{"train_loss": 1.8600738048553467, "global_step": 1438, "epoch": 16, "lr": 9.999703661322824e-05}
{"train_loss": 1.7963511943817139, "global_step": 1439, "epoch": 16, "lr": 9.999703029813692e-05}
{"train_loss": 1.994568943977356, "global_step": 1440, "epoch": 16, "lr": 9.999702397632413e-05}
{"train_loss": 1.8889883756637573, "global_step": 1441, "epoch": 16, "lr": 9.999701764778983e-05}
{"train_loss": 1.6008167266845703, "global_step": 1442, "epoch": 16, "lr": 9.999701131253405e-05}
{"train_loss": 1.7304563522338867, "global_step": 1443, "epoch": 16, "lr": 9.999700497055679e-05}
{"train_loss": 1.8540385961532593, "global_step": 1444, "epoch": 16, "lr": 9.999699862185803e-05}
{"train_loss": 1.584116816520691, "global_step": 1445, "epoch": 16, "lr": 9.999699226643778e-05}
{"train_loss": 1.5294451713562012, "global_step": 1446, "epoch": 16, "lr": 9.999698590429605e-05}
{"train_loss": 1.7952369451522827, "global_step": 1447, "epoch": 16, "lr": 9.999697953543284e-05}
{"train_loss": 1.7521239519119263, "global_step": 1448, "epoch": 16, "lr": 9.999697315984814e-05}
{"train_loss": 1.8698161840438843, "global_step": 1449, "epoch": 16, "lr": 9.999696677754196e-05}
{"train_loss": 1.6684556007385254, "global_step": 1450, "epoch": 16, "lr": 9.999696038851429e-05}
{"train_loss": 1.7853267192840576, "global_step": 1451, "epoch": 16, "lr": 9.999695399276514e-05}
{"train_loss": 1.6279854774475098, "global_step": 1452, "epoch": 16, "lr": 9.99969475902945e-05}
{"train_loss": 1.6658233404159546, "global_step": 1453, "epoch": 16, "lr": 9.99969411811024e-05}
{"train_loss": 1.8309389352798462, "global_step": 1454, "epoch": 16, "lr": 9.99969347651888e-05}
{"train_loss": 1.8995555639266968, "global_step": 1455, "epoch": 16, "lr": 9.999692834255375e-05}
{"train_loss": 1.4886596202850342, "global_step": 1456, "epoch": 16, "lr": 9.99969219131972e-05}
{"train_loss": 1.6001121997833252, "global_step": 1457, "epoch": 16, "lr": 9.999691547711918e-05}
{"train_loss": 1.6304786205291748, "global_step": 1458, "epoch": 16, "lr": 9.999690903431967e-05}
{"train_loss": 1.4280755519866943, "global_step": 1459, "epoch": 16, "lr": 9.99969025847987e-05}
{"train_loss": 1.94173264503479, "global_step": 1460, "epoch": 16, "lr": 9.999689612855624e-05}
{"train_loss": 1.6416271924972534, "global_step": 1461, "epoch": 16, "lr": 9.999688966559233e-05}
{"train_loss": 1.5407952070236206, "global_step": 1462, "epoch": 16, "lr": 9.999688319590695e-05}
{"train_loss": 1.7272758483886719, "global_step": 1463, "epoch": 16, "lr": 9.999687671950008e-05}
{"train_loss": 1.7325197458267212, "global_step": 1464, "epoch": 16, "lr": 9.999687023637174e-05}
{"train_loss": 1.5708974599838257, "global_step": 1465, "epoch": 16, "lr": 9.999686374652193e-05}
{"train_loss": 1.5333876609802246, "global_step": 1466, "epoch": 16, "lr": 9.999685724995067e-05}
{"train_loss": 1.7929658889770508, "global_step": 1467, "epoch": 16, "lr": 9.999685074665792e-05}
{"train_loss": 1.6690380573272705, "global_step": 1468, "epoch": 16, "lr": 9.999684423664371e-05}
{"train_loss": 1.6574668884277344, "global_step": 1469, "epoch": 16, "lr": 9.999683771990805e-05}
{"train_loss": 1.8000450134277344, "global_step": 1470, "epoch": 16, "lr": 9.99968311964509e-05}
{"train_loss": 1.5198957920074463, "global_step": 1471, "epoch": 16, "lr": 9.999682466627229e-05}
{"train_loss": 1.6857280731201172, "global_step": 1472, "epoch": 16, "lr": 9.999681812937223e-05}
{"train_loss": 1.8829596042633057, "global_step": 1473, "epoch": 16, "lr": 9.99968115857507e-05}
{"train_loss": 1.6412385702133179, "global_step": 1474, "epoch": 16, "lr": 9.999680503540769e-05}
{"train_loss": 1.7353867292404175, "global_step": 1475, "epoch": 16, "lr": 9.999679847834324e-05}
{"train_loss": 1.6300355195999146, "global_step": 1476, "epoch": 16, "lr": 9.999679191455732e-05}
{"train_loss": 1.6073540449142456, "global_step": 1477, "epoch": 16, "lr": 9.999678534404995e-05}
{"train_loss": 1.6419652700424194, "global_step": 1478, "epoch": 16, "lr": 9.999677876682111e-05}
{"train_loss": 1.4167627096176147, "global_step": 1479, "epoch": 16, "lr": 9.999677218287082e-05}
{"train_loss": 1.5854157209396362, "global_step": 1480, "epoch": 16, "lr": 9.999676559219908e-05}
{"train_loss": 1.9049466848373413, "global_step": 1481, "epoch": 16, "lr": 9.999675899480587e-05}
{"train_loss": 1.4696711301803589, "global_step": 1482, "epoch": 16, "lr": 9.999675239069122e-05}
{"train_loss": 1.3954529762268066, "global_step": 1483, "epoch": 16, "lr": 9.99967457798551e-05}
{"train_loss": 1.4734301567077637, "global_step": 1484, "epoch": 16, "lr": 9.999673916229754e-05}
{"train_loss": 1.8095800876617432, "global_step": 1485, "epoch": 16, "lr": 9.999673253801852e-05}
{"train_loss": 1.4460822343826294, "global_step": 1486, "epoch": 16, "lr": 9.999672590701806e-05}
{"train_loss": 1.7556933164596558, "global_step": 1487, "epoch": 16, "lr": 9.999671926929614e-05}
{"train_loss": 1.3932974338531494, "global_step": 1488, "epoch": 16, "lr": 9.999671262485277e-05}
{"train_loss": 1.694901943206787, "global_step": 1489, "epoch": 16, "lr": 9.999670597368795e-05}
{"train_loss": 1.8372608423233032, "global_step": 1490, "epoch": 16, "lr": 9.999669931580169e-05}
{"train_loss": 1.9214112758636475, "global_step": 1491, "epoch": 16, "lr": 9.999669265119398e-05}
{"train_loss": 1.5895329713821411, "global_step": 1492, "epoch": 16, "lr": 9.999668597986481e-05}
{"train_loss": 1.5627020597457886, "global_step": 1493, "epoch": 16, "lr": 9.999667930181422e-05}
{"train_loss": 1.598970651626587, "global_step": 1494, "epoch": 16, "lr": 9.999667261704216e-05}
{"train_loss": 1.6376802921295166, "global_step": 1495, "epoch": 16, "lr": 9.999666592554867e-05}
{"train_loss": 1.9079804420471191, "global_step": 1496, "epoch": 16, "lr": 9.999665922733374e-05}
{"train_loss": 1.733977198600769, "global_step": 1497, "epoch": 16, "lr": 9.999665252239736e-05}
{"train_loss": 1.6960058212280273, "global_step": 1498, "epoch": 16, "lr": 9.999664581073956e-05}
{"train_loss": 1.562010645866394, "global_step": 1499, "epoch": 16, "lr": 9.99966390923603e-05}
{"train_loss": 1.7722824811935425, "global_step": 1500, "epoch": 16, "lr": 9.99966323672596e-05}
{"train_loss": 1.6463044881820679, "global_step": 1501, "epoch": 16, "lr": 9.999662563543748e-05}
{"train_loss": 1.4945809841156006, "global_step": 1502, "epoch": 16, "lr": 9.999661889689391e-05}
{"train_loss": 1.7889915704727173, "global_step": 1503, "epoch": 16, "lr": 9.99966121516289e-05}
{"train_loss": 1.7692234516143799, "global_step": 1504, "epoch": 16, "lr": 9.999660539964247e-05}
{"train_loss": 1.5865083932876587, "global_step": 1505, "epoch": 16, "lr": 9.999659864093459e-05}
{"train_loss": 1.752570629119873, "global_step": 1506, "epoch": 16, "lr": 9.999659187550528e-05}
{"train_loss": 1.7984930276870728, "global_step": 1507, "epoch": 16, "lr": 9.999658510335454e-05}
{"train_loss": 1.5521267652511597, "global_step": 1508, "epoch": 16, "lr": 9.999657832448237e-05}
{"train_loss": 1.7112889289855957, "global_step": 1509, "epoch": 16, "lr": 9.999657153888878e-05}
{"train_loss": 1.5514163970947266, "global_step": 1510, "epoch": 16, "lr": 9.999656474657375e-05}
{"train_loss": 1.7535957098007202, "global_step": 1511, "epoch": 16, "lr": 9.999655794753729e-05}
{"train_loss": 1.690750033667918, "global_step": 1512, "epoch": 16, "lr": 9.999655114177939e-05, "val_loss": 1.5947511196136475}
{"train_loss": 1.860283374786377, "global_step": 1513, "epoch": 17, "lr": 9.999654432930008e-05}
{"train_loss": 1.4078184366226196, "global_step": 1514, "epoch": 17, "lr": 9.999653751009935e-05}
{"train_loss": 1.6266423463821411, "global_step": 1515, "epoch": 17, "lr": 9.999653068417718e-05}
{"train_loss": 1.5364112854003906, "global_step": 1516, "epoch": 17, "lr": 9.999652385153359e-05}
{"train_loss": 1.7965182065963745, "global_step": 1517, "epoch": 17, "lr": 9.999651701216857e-05}
{"train_loss": 1.8271725177764893, "global_step": 1518, "epoch": 17, "lr": 9.999651016608213e-05}
{"train_loss": 1.8968629837036133, "global_step": 1519, "epoch": 17, "lr": 9.999650331327429e-05}
{"train_loss": 1.5770442485809326, "global_step": 1520, "epoch": 17, "lr": 9.9996496453745e-05}
{"train_loss": 1.6110759973526, "global_step": 1521, "epoch": 17, "lr": 9.999648958749431e-05}
{"train_loss": 1.5847822427749634, "global_step": 1522, "epoch": 17, "lr": 9.99964827145222e-05}
{"train_loss": 1.725602388381958, "global_step": 1523, "epoch": 17, "lr": 9.999647583482866e-05}
{"train_loss": 1.5507636070251465, "global_step": 1524, "epoch": 17, "lr": 9.999646894841371e-05}
{"train_loss": 1.6745877265930176, "global_step": 1525, "epoch": 17, "lr": 9.999646205527735e-05}
{"train_loss": 1.7086591720581055, "global_step": 1526, "epoch": 17, "lr": 9.999645515541956e-05}
{"train_loss": 1.472794771194458, "global_step": 1527, "epoch": 17, "lr": 9.999644824884039e-05}
{"train_loss": 1.5170382261276245, "global_step": 1528, "epoch": 17, "lr": 9.999644133553976e-05}
{"train_loss": 1.5589470863342285, "global_step": 1529, "epoch": 17, "lr": 9.999643441551774e-05}
{"train_loss": 1.6423249244689941, "global_step": 1530, "epoch": 17, "lr": 9.999642748877432e-05}
{"train_loss": 1.8105168342590332, "global_step": 1531, "epoch": 17, "lr": 9.999642055530947e-05}
{"train_loss": 1.6884839534759521, "global_step": 1532, "epoch": 17, "lr": 9.999641361512321e-05}
{"train_loss": 1.715524435043335, "global_step": 1533, "epoch": 17, "lr": 9.999640666821556e-05}
{"train_loss": 1.6601279973983765, "global_step": 1534, "epoch": 17, "lr": 9.99963997145865e-05}
{"train_loss": 1.5756678581237793, "global_step": 1535, "epoch": 17, "lr": 9.999639275423602e-05}
{"train_loss": 1.5613819360733032, "global_step": 1536, "epoch": 17, "lr": 9.999638578716415e-05}
{"train_loss": 1.3650954961776733, "global_step": 1537, "epoch": 17, "lr": 9.999637881337086e-05}
{"train_loss": 1.5945076942443848, "global_step": 1538, "epoch": 17, "lr": 9.999637183285617e-05}
{"train_loss": 1.606336236000061, "global_step": 1539, "epoch": 17, "lr": 9.999636484562009e-05}
{"train_loss": 1.720394492149353, "global_step": 1540, "epoch": 17, "lr": 9.999635785166259e-05}
{"train_loss": 1.7192366123199463, "global_step": 1541, "epoch": 17, "lr": 9.99963508509837e-05}
{"train_loss": 1.5946693420410156, "global_step": 1542, "epoch": 17, "lr": 9.99963438435834e-05}
{"train_loss": 1.5689830780029297, "global_step": 1543, "epoch": 17, "lr": 9.999633682946171e-05}
{"train_loss": 1.528398871421814, "global_step": 1544, "epoch": 17, "lr": 9.999632980861862e-05}
{"train_loss": 1.8131033182144165, "global_step": 1545, "epoch": 17, "lr": 9.999632278105413e-05}
{"train_loss": 1.5192441940307617, "global_step": 1546, "epoch": 17, "lr": 9.999631574676825e-05}
{"train_loss": 1.7013342380523682, "global_step": 1547, "epoch": 17, "lr": 9.999630870576098e-05}
{"train_loss": 1.8535678386688232, "global_step": 1548, "epoch": 17, "lr": 9.99963016580323e-05}
{"train_loss": 1.8279987573623657, "global_step": 1549, "epoch": 17, "lr": 9.999629460358223e-05}
{"train_loss": 1.8175915479660034, "global_step": 1550, "epoch": 17, "lr": 9.999628754241078e-05}
{"train_loss": 1.5212377309799194, "global_step": 1551, "epoch": 17, "lr": 9.999628047451793e-05}
{"train_loss": 1.6624925136566162, "global_step": 1552, "epoch": 17, "lr": 9.999627339990368e-05}
{"train_loss": 1.592318058013916, "global_step": 1553, "epoch": 17, "lr": 9.999626631856804e-05}
{"train_loss": 1.8836537599563599, "global_step": 1554, "epoch": 17, "lr": 9.999625923051104e-05}
{"train_loss": 1.6507936716079712, "global_step": 1555, "epoch": 17, "lr": 9.999625213573263e-05}
{"train_loss": 1.6198573112487793, "global_step": 1556, "epoch": 17, "lr": 9.999624503423284e-05}
{"train_loss": 1.7343924045562744, "global_step": 1557, "epoch": 17, "lr": 9.999623792601165e-05}
{"train_loss": 1.5717921257019043, "global_step": 1558, "epoch": 17, "lr": 9.999623081106908e-05}
{"train_loss": 1.483217477798462, "global_step": 1559, "epoch": 17, "lr": 9.999622368940513e-05}
{"train_loss": 1.7688089609146118, "global_step": 1560, "epoch": 17, "lr": 9.99962165610198e-05}
{"train_loss": 1.6936074495315552, "global_step": 1561, "epoch": 17, "lr": 9.99962094259131e-05}
{"train_loss": 1.9069805145263672, "global_step": 1562, "epoch": 17, "lr": 9.9996202284085e-05}
{"train_loss": 1.5829403400421143, "global_step": 1563, "epoch": 17, "lr": 9.999619513553553e-05}
{"train_loss": 1.7191691398620605, "global_step": 1564, "epoch": 17, "lr": 9.999618798026466e-05}
{"train_loss": 1.485212802886963, "global_step": 1565, "epoch": 17, "lr": 9.999618081827245e-05}
{"train_loss": 1.6295084953308105, "global_step": 1566, "epoch": 17, "lr": 9.999617364955884e-05}
{"train_loss": 1.6285033226013184, "global_step": 1567, "epoch": 17, "lr": 9.999616647412384e-05}
{"train_loss": 1.5217632055282593, "global_step": 1568, "epoch": 17, "lr": 9.999615929196748e-05}
{"train_loss": 1.6114765405654907, "global_step": 1569, "epoch": 17, "lr": 9.999615210308976e-05}
{"train_loss": 1.658778429031372, "global_step": 1570, "epoch": 17, "lr": 9.999614490749065e-05}
{"train_loss": 1.5489158630371094, "global_step": 1571, "epoch": 17, "lr": 9.999613770517017e-05}
{"train_loss": 1.6045622825622559, "global_step": 1572, "epoch": 17, "lr": 9.999613049612833e-05}
{"train_loss": 1.532241702079773, "global_step": 1573, "epoch": 17, "lr": 9.999612328036509e-05}
{"train_loss": 1.5864574909210205, "global_step": 1574, "epoch": 17, "lr": 9.99961160578805e-05}
{"train_loss": 1.7044978141784668, "global_step": 1575, "epoch": 17, "lr": 9.999610882867454e-05}
{"train_loss": 1.7509765625, "global_step": 1576, "epoch": 17, "lr": 9.999610159274722e-05}
{"train_loss": 1.542187213897705, "global_step": 1577, "epoch": 17, "lr": 9.999609435009853e-05}
{"train_loss": 1.7681810855865479, "global_step": 1578, "epoch": 17, "lr": 9.999608710072848e-05}
{"train_loss": 1.727794885635376, "global_step": 1579, "epoch": 17, "lr": 9.999607984463704e-05}
{"train_loss": 1.5507088899612427, "global_step": 1580, "epoch": 17, "lr": 9.999607258182427e-05}
{"train_loss": 1.5765197277069092, "global_step": 1581, "epoch": 17, "lr": 9.99960653122901e-05}
{"train_loss": 1.6196374893188477, "global_step": 1582, "epoch": 17, "lr": 9.99960580360346e-05}
{"train_loss": 1.7254470586776733, "global_step": 1583, "epoch": 17, "lr": 9.999605075305773e-05}
{"train_loss": 1.4579963684082031, "global_step": 1584, "epoch": 17, "lr": 9.99960434633595e-05}
{"train_loss": 1.5780330896377563, "global_step": 1585, "epoch": 17, "lr": 9.999603616693991e-05}
{"train_loss": 1.6162488460540771, "global_step": 1586, "epoch": 17, "lr": 9.999602886379895e-05}
{"train_loss": 1.6506352424621582, "global_step": 1587, "epoch": 17, "lr": 9.999602155393666e-05}
{"train_loss": 1.5754573345184326, "global_step": 1588, "epoch": 17, "lr": 9.999601423735299e-05}
{"train_loss": 1.6210788488388062, "global_step": 1589, "epoch": 17, "lr": 9.999600691404798e-05}
{"train_loss": 1.569011926651001, "global_step": 1590, "epoch": 17, "lr": 9.999599958402161e-05}
{"train_loss": 1.620030403137207, "global_step": 1591, "epoch": 17, "lr": 9.99959922472739e-05}
{"train_loss": 1.6522951126098633, "global_step": 1592, "epoch": 17, "lr": 9.999598490380481e-05}
{"train_loss": 1.5868937969207764, "global_step": 1593, "epoch": 17, "lr": 9.99959775536144e-05}
{"train_loss": 1.5750761032104492, "global_step": 1594, "epoch": 17, "lr": 9.999597019670262e-05}
{"train_loss": 1.296479344367981, "global_step": 1595, "epoch": 17, "lr": 9.999596283306949e-05}
{"train_loss": 1.5865228176116943, "global_step": 1596, "epoch": 17, "lr": 9.999595546271503e-05}
{"train_loss": 1.7414041757583618, "global_step": 1597, "epoch": 17, "lr": 9.999594808563922e-05}
{"train_loss": 1.4503477811813354, "global_step": 1598, "epoch": 17, "lr": 9.999594070184205e-05}
{"train_loss": 1.5308560132980347, "global_step": 1599, "epoch": 17, "lr": 9.999593331132354e-05}
{"train_loss": 1.9484641551971436, "global_step": 1600, "epoch": 17, "lr": 9.99959259140837e-05}
{"train_loss": 1.6367341494292356, "global_step": 1601, "epoch": 17, "lr": 9.999591851012249e-05, "val_loss": 1.5714685916900635}
{"train_loss": 1.541439414024353, "global_step": 1602, "epoch": 18, "lr": 9.999591109943996e-05}
{"train_loss": 1.5551714897155762, "global_step": 1603, "epoch": 18, "lr": 9.999590368203608e-05}
{"train_loss": 1.4887683391571045, "global_step": 1604, "epoch": 18, "lr": 9.999589625791086e-05}
{"train_loss": 1.5398379564285278, "global_step": 1605, "epoch": 18, "lr": 9.99958888270643e-05}
{"train_loss": 1.5948466062545776, "global_step": 1606, "epoch": 18, "lr": 9.99958813894964e-05}
{"train_loss": 1.544936180114746, "global_step": 1607, "epoch": 18, "lr": 9.999587394520717e-05}
{"train_loss": 1.5599013566970825, "global_step": 1608, "epoch": 18, "lr": 9.99958664941966e-05}
{"train_loss": 1.4887263774871826, "global_step": 1609, "epoch": 18, "lr": 9.999585903646472e-05}
{"train_loss": 1.6310503482818604, "global_step": 1610, "epoch": 18, "lr": 9.999585157201148e-05}
{"train_loss": 1.566678524017334, "global_step": 1611, "epoch": 18, "lr": 9.999584410083691e-05}
{"train_loss": 1.718191146850586, "global_step": 1612, "epoch": 18, "lr": 9.9995836622941e-05}
{"train_loss": 1.4749559164047241, "global_step": 1613, "epoch": 18, "lr": 9.999582913832378e-05}
{"train_loss": 1.6592215299606323, "global_step": 1614, "epoch": 18, "lr": 9.999582164698521e-05}
{"train_loss": 1.5153008699417114, "global_step": 1615, "epoch": 18, "lr": 9.999581414892531e-05}
{"train_loss": 1.6208903789520264, "global_step": 1616, "epoch": 18, "lr": 9.999580664414409e-05}
{"train_loss": 1.4924637079238892, "global_step": 1617, "epoch": 18, "lr": 9.999579913264156e-05}
{"train_loss": 1.5331904888153076, "global_step": 1618, "epoch": 18, "lr": 9.999579161441769e-05}
{"train_loss": 1.286362886428833, "global_step": 1619, "epoch": 18, "lr": 9.99957840894725e-05}
{"train_loss": 1.5050249099731445, "global_step": 1620, "epoch": 18, "lr": 9.999577655780598e-05}
{"train_loss": 1.4611307382583618, "global_step": 1621, "epoch": 18, "lr": 9.999576901941814e-05}
{"train_loss": 1.4229987859725952, "global_step": 1622, "epoch": 18, "lr": 9.999576147430896e-05}
{"train_loss": 1.5368849039077759, "global_step": 1623, "epoch": 18, "lr": 9.999575392247849e-05}
{"train_loss": 1.69760262966156, "global_step": 1624, "epoch": 18, "lr": 9.999574636392668e-05}
{"train_loss": 1.4754358530044556, "global_step": 1625, "epoch": 18, "lr": 9.999573879865356e-05}
{"train_loss": 1.5725938081741333, "global_step": 1626, "epoch": 18, "lr": 9.999573122665912e-05}
{"train_loss": 1.4074785709381104, "global_step": 1627, "epoch": 18, "lr": 9.999572364794336e-05}
{"train_loss": 1.71564519405365, "global_step": 1628, "epoch": 18, "lr": 9.999571606250628e-05}
{"train_loss": 1.4078136682510376, "global_step": 1629, "epoch": 18, "lr": 9.999570847034789e-05}
{"train_loss": 1.5724880695343018, "global_step": 1630, "epoch": 18, "lr": 9.99957008714682e-05}
{"train_loss": 1.2199654579162598, "global_step": 1631, "epoch": 18, "lr": 9.999569326586718e-05}
{"train_loss": 1.6709873676300049, "global_step": 1632, "epoch": 18, "lr": 9.999568565354485e-05}
{"train_loss": 1.698024868965149, "global_step": 1633, "epoch": 18, "lr": 9.999567803450123e-05}
{"train_loss": 1.697478175163269, "global_step": 1634, "epoch": 18, "lr": 9.999567040873628e-05}
{"train_loss": 1.6051990985870361, "global_step": 1635, "epoch": 18, "lr": 9.999566277625003e-05}
{"train_loss": 1.507196307182312, "global_step": 1636, "epoch": 18, "lr": 9.999565513704246e-05}
{"train_loss": 1.4600427150726318, "global_step": 1637, "epoch": 18, "lr": 9.999564749111359e-05}
{"train_loss": 1.6518629789352417, "global_step": 1638, "epoch": 18, "lr": 9.999563983846342e-05}
{"train_loss": 1.5308672189712524, "global_step": 1639, "epoch": 18, "lr": 9.999563217909193e-05}
{"train_loss": 1.5470068454742432, "global_step": 1640, "epoch": 18, "lr": 9.999562451299916e-05}
{"train_loss": 1.6426156759262085, "global_step": 1641, "epoch": 18, "lr": 9.999561684018508e-05}
{"train_loss": 1.5260133743286133, "global_step": 1642, "epoch": 18, "lr": 9.999560916064969e-05}
{"train_loss": 1.6930763721466064, "global_step": 1643, "epoch": 18, "lr": 9.999560147439299e-05}
{"train_loss": 1.465458869934082, "global_step": 1644, "epoch": 18, "lr": 9.999559378141502e-05}
{"train_loss": 1.4068936109542847, "global_step": 1645, "epoch": 18, "lr": 9.999558608171573e-05}
{"train_loss": 1.6145495176315308, "global_step": 1646, "epoch": 18, "lr": 9.999557837529516e-05}
{"train_loss": 1.6234984397888184, "global_step": 1647, "epoch": 18, "lr": 9.999557066215328e-05}
{"train_loss": 1.854651689529419, "global_step": 1648, "epoch": 18, "lr": 9.999556294229011e-05}
{"train_loss": 1.6021286249160767, "global_step": 1649, "epoch": 18, "lr": 9.999555521570565e-05}
{"train_loss": 1.4641474485397339, "global_step": 1650, "epoch": 18, "lr": 9.999554748239988e-05}
{"train_loss": 1.5579020977020264, "global_step": 1651, "epoch": 18, "lr": 9.999553974237283e-05}
{"train_loss": 1.3978544473648071, "global_step": 1652, "epoch": 18, "lr": 9.99955319956245e-05}
{"train_loss": 1.4461337327957153, "global_step": 1653, "epoch": 18, "lr": 9.999552424215486e-05}
{"train_loss": 1.712531328201294, "global_step": 1654, "epoch": 18, "lr": 9.999551648196395e-05}
{"train_loss": 1.6256227493286133, "global_step": 1655, "epoch": 18, "lr": 9.999550871505174e-05}
{"train_loss": 1.5399891138076782, "global_step": 1656, "epoch": 18, "lr": 9.999550094141822e-05}
{"train_loss": 1.634557843208313, "global_step": 1657, "epoch": 18, "lr": 9.999549316106345e-05}
{"train_loss": 1.4294782876968384, "global_step": 1658, "epoch": 18, "lr": 9.99954853739874e-05}
{"train_loss": 1.7139240503311157, "global_step": 1659, "epoch": 18, "lr": 9.999547758019003e-05}
{"train_loss": 1.6299155950546265, "global_step": 1660, "epoch": 18, "lr": 9.99954697796714e-05}
{"train_loss": 1.7792035341262817, "global_step": 1661, "epoch": 18, "lr": 9.999546197243151e-05}
{"train_loss": 1.6410835981369019, "global_step": 1662, "epoch": 18, "lr": 9.99954541584703e-05}
{"train_loss": 1.6413367986679077, "global_step": 1663, "epoch": 18, "lr": 9.999544633778784e-05}
{"train_loss": 1.5343420505523682, "global_step": 1664, "epoch": 18, "lr": 9.999543851038407e-05}
{"train_loss": 1.4508039951324463, "global_step": 1665, "epoch": 18, "lr": 9.999543067625905e-05}
{"train_loss": 1.6767414808273315, "global_step": 1666, "epoch": 18, "lr": 9.999542283541274e-05}
{"train_loss": 1.4860312938690186, "global_step": 1667, "epoch": 18, "lr": 9.999541498784517e-05}
{"train_loss": 1.5526976585388184, "global_step": 1668, "epoch": 18, "lr": 9.999540713355631e-05}
{"train_loss": 1.4395290613174438, "global_step": 1669, "epoch": 18, "lr": 9.999539927254617e-05}
{"train_loss": 1.446058750152588, "global_step": 1670, "epoch": 18, "lr": 9.999539140481478e-05}
{"train_loss": 1.5878701210021973, "global_step": 1671, "epoch": 18, "lr": 9.999538353036211e-05}
{"train_loss": 1.5287864208221436, "global_step": 1672, "epoch": 18, "lr": 9.999537564918817e-05}
{"train_loss": 1.5267601013183594, "global_step": 1673, "epoch": 18, "lr": 9.999536776129297e-05}
{"train_loss": 1.5546979904174805, "global_step": 1674, "epoch": 18, "lr": 9.999535986667649e-05}
{"train_loss": 1.4366776943206787, "global_step": 1675, "epoch": 18, "lr": 9.999535196533873e-05}
{"train_loss": 1.4678525924682617, "global_step": 1676, "epoch": 18, "lr": 9.999534405727973e-05}
{"train_loss": 1.7060574293136597, "global_step": 1677, "epoch": 18, "lr": 9.999533614249945e-05}
{"train_loss": 1.4590212106704712, "global_step": 1678, "epoch": 18, "lr": 9.999532822099791e-05}
{"train_loss": 1.4793893098831177, "global_step": 1679, "epoch": 18, "lr": 9.999532029277512e-05}
{"train_loss": 1.5590505599975586, "global_step": 1680, "epoch": 18, "lr": 9.999531235783105e-05}
{"train_loss": 1.8449786901474, "global_step": 1681, "epoch": 18, "lr": 9.999530441616573e-05}
{"train_loss": 1.6342802047729492, "global_step": 1682, "epoch": 18, "lr": 9.999529646777914e-05}
{"train_loss": 1.498436450958252, "global_step": 1683, "epoch": 18, "lr": 9.999528851267131e-05}
{"train_loss": 1.5652985572814941, "global_step": 1684, "epoch": 18, "lr": 9.999528055084222e-05}
{"train_loss": 1.4530006647109985, "global_step": 1685, "epoch": 18, "lr": 9.999527258229187e-05}
{"train_loss": 1.585548758506775, "global_step": 1686, "epoch": 18, "lr": 9.999526460702025e-05}
{"train_loss": 1.1967339515686035, "global_step": 1687, "epoch": 18, "lr": 9.999525662502739e-05}
{"train_loss": 1.4547711610794067, "global_step": 1688, "epoch": 18, "lr": 9.999524863631326e-05}
{"train_loss": 1.464865803718567, "global_step": 1689, "epoch": 18, "lr": 9.99952406408779e-05}
{"train_loss": 1.5493077548702112, "global_step": 1690, "epoch": 18, "lr": 9.999523263872128e-05, "val_loss": 1.531076431274414}
{"train_loss": 1.8196179866790771, "global_step": 1691, "epoch": 19, "lr": 9.999522462984342e-05}
{"train_loss": 1.5136771202087402, "global_step": 1692, "epoch": 19, "lr": 9.99952166142443e-05}
{"train_loss": 1.6289362907409668, "global_step": 1693, "epoch": 19, "lr": 9.999520859192393e-05}
{"train_loss": 1.5500174760818481, "global_step": 1694, "epoch": 19, "lr": 9.999520056288234e-05}
{"train_loss": 1.36724853515625, "global_step": 1695, "epoch": 19, "lr": 9.999519252711947e-05}
{"train_loss": 1.572070598602295, "global_step": 1696, "epoch": 19, "lr": 9.999518448463538e-05}
{"train_loss": 1.5619107484817505, "global_step": 1697, "epoch": 19, "lr": 9.999517643543003e-05}
{"train_loss": 1.5467572212219238, "global_step": 1698, "epoch": 19, "lr": 9.999516837950345e-05}
{"train_loss": 1.4757542610168457, "global_step": 1699, "epoch": 19, "lr": 9.999516031685562e-05}
{"train_loss": 1.4381334781646729, "global_step": 1700, "epoch": 19, "lr": 9.999515224748656e-05}
{"train_loss": 1.3220716714859009, "global_step": 1701, "epoch": 19, "lr": 9.999514417139625e-05}
{"train_loss": 1.5595721006393433, "global_step": 1702, "epoch": 19, "lr": 9.99951360885847e-05}
{"train_loss": 1.3918452262878418, "global_step": 1703, "epoch": 19, "lr": 9.999512799905193e-05}
{"train_loss": 1.5817151069641113, "global_step": 1704, "epoch": 19, "lr": 9.99951199027979e-05}
{"train_loss": 1.5515422821044922, "global_step": 1705, "epoch": 19, "lr": 9.999511179982266e-05}
{"train_loss": 1.3818655014038086, "global_step": 1706, "epoch": 19, "lr": 9.999510369012618e-05}
{"train_loss": 1.377463936805725, "global_step": 1707, "epoch": 19, "lr": 9.999509557370846e-05}
{"train_loss": 1.5419589281082153, "global_step": 1708, "epoch": 19, "lr": 9.999508745056951e-05}
{"train_loss": 1.3827803134918213, "global_step": 1709, "epoch": 19, "lr": 9.999507932070934e-05}
{"train_loss": 1.634404182434082, "global_step": 1710, "epoch": 19, "lr": 9.999507118412794e-05}
{"train_loss": 1.4973430633544922, "global_step": 1711, "epoch": 19, "lr": 9.99950630408253e-05}
{"train_loss": 1.508410930633545, "global_step": 1712, "epoch": 19, "lr": 9.999505489080144e-05}
{"train_loss": 1.6201248168945312, "global_step": 1713, "epoch": 19, "lr": 9.999504673405636e-05}
{"train_loss": 1.377048373222351, "global_step": 1714, "epoch": 19, "lr": 9.999503857059005e-05}
{"train_loss": 1.4536817073822021, "global_step": 1715, "epoch": 19, "lr": 9.99950304004025e-05}
{"train_loss": 1.5942338705062866, "global_step": 1716, "epoch": 19, "lr": 9.999502222349375e-05}
{"train_loss": 1.461302399635315, "global_step": 1717, "epoch": 19, "lr": 9.999501403986377e-05}
{"train_loss": 1.4682213068008423, "global_step": 1718, "epoch": 19, "lr": 9.999500584951259e-05}
{"train_loss": 1.4917075634002686, "global_step": 1719, "epoch": 19, "lr": 9.999499765244016e-05}
{"train_loss": 1.5950452089309692, "global_step": 1720, "epoch": 19, "lr": 9.999498944864652e-05}
{"train_loss": 1.4228299856185913, "global_step": 1721, "epoch": 19, "lr": 9.999498123813167e-05}
{"train_loss": 1.330458641052246, "global_step": 1722, "epoch": 19, "lr": 9.99949730208956e-05}
{"train_loss": 1.4990839958190918, "global_step": 1723, "epoch": 19, "lr": 9.999496479693832e-05}
{"train_loss": 1.530150294303894, "global_step": 1724, "epoch": 19, "lr": 9.999495656625981e-05}
{"train_loss": 1.5308315753936768, "global_step": 1725, "epoch": 19, "lr": 9.99949483288601e-05}
{"train_loss": 1.4456806182861328, "global_step": 1726, "epoch": 19, "lr": 9.999494008473919e-05}
{"train_loss": 1.402086615562439, "global_step": 1727, "epoch": 19, "lr": 9.999493183389705e-05}
{"train_loss": 1.4114965200424194, "global_step": 1728, "epoch": 19, "lr": 9.999492357633372e-05}
{"train_loss": 1.5945814847946167, "global_step": 1729, "epoch": 19, "lr": 9.999491531204917e-05}
{"train_loss": 1.4682962894439697, "global_step": 1730, "epoch": 19, "lr": 9.999490704104341e-05}
{"train_loss": 1.2811493873596191, "global_step": 1731, "epoch": 19, "lr": 9.999489876331645e-05}
{"train_loss": 1.6242114305496216, "global_step": 1732, "epoch": 19, "lr": 9.999489047886827e-05}
{"train_loss": 1.6247013807296753, "global_step": 1733, "epoch": 19, "lr": 9.99948821876989e-05}
{"train_loss": 1.6174832582473755, "global_step": 1734, "epoch": 19, "lr": 9.999487388980834e-05}
{"train_loss": 1.7352484464645386, "global_step": 1735, "epoch": 19, "lr": 9.999486558519657e-05}
{"train_loss": 1.3112398386001587, "global_step": 1736, "epoch": 19, "lr": 9.99948572738636e-05}
{"train_loss": 1.4817636013031006, "global_step": 1737, "epoch": 19, "lr": 9.999484895580941e-05}
{"train_loss": 1.718015193939209, "global_step": 1738, "epoch": 19, "lr": 9.999484063103404e-05}
{"train_loss": 1.4837218523025513, "global_step": 1739, "epoch": 19, "lr": 9.999483229953748e-05}
{"train_loss": 1.7510828971862793, "global_step": 1740, "epoch": 19, "lr": 9.999482396131971e-05}
{"train_loss": 1.3322358131408691, "global_step": 1741, "epoch": 19, "lr": 9.999481561638075e-05}
{"train_loss": 1.788727879524231, "global_step": 1742, "epoch": 19, "lr": 9.99948072647206e-05}
{"train_loss": 1.4626609086990356, "global_step": 1743, "epoch": 19, "lr": 9.999479890633925e-05}
{"train_loss": 1.3969638347625732, "global_step": 1744, "epoch": 19, "lr": 9.99947905412367e-05}
{"train_loss": 1.391923189163208, "global_step": 1745, "epoch": 19, "lr": 9.999478216941298e-05}
{"train_loss": 1.5909554958343506, "global_step": 1746, "epoch": 19, "lr": 9.999477379086807e-05}
{"train_loss": 1.5826094150543213, "global_step": 1747, "epoch": 19, "lr": 9.999476540560197e-05}
{"train_loss": 1.5665031671524048, "global_step": 1748, "epoch": 19, "lr": 9.999475701361467e-05}
{"train_loss": 1.5341793298721313, "global_step": 1749, "epoch": 19, "lr": 9.99947486149062e-05}
{"train_loss": 1.4551684856414795, "global_step": 1750, "epoch": 19, "lr": 9.999474020947654e-05}
{"train_loss": 1.3229644298553467, "global_step": 1751, "epoch": 19, "lr": 9.999473179732569e-05}
{"train_loss": 1.4243361949920654, "global_step": 1752, "epoch": 19, "lr": 9.999472337845366e-05}
{"train_loss": 1.3190162181854248, "global_step": 1753, "epoch": 19, "lr": 9.999471495286046e-05}
{"train_loss": 1.367590308189392, "global_step": 1754, "epoch": 19, "lr": 9.999470652054607e-05}
{"train_loss": 1.5129977464675903, "global_step": 1755, "epoch": 19, "lr": 9.999469808151049e-05}
{"train_loss": 1.2917460203170776, "global_step": 1756, "epoch": 19, "lr": 9.999468963575375e-05}
{"train_loss": 1.4308884143829346, "global_step": 1757, "epoch": 19, "lr": 9.999468118327583e-05}
{"train_loss": 1.4767173528671265, "global_step": 1758, "epoch": 19, "lr": 9.999467272407673e-05}
{"train_loss": 1.4481853246688843, "global_step": 1759, "epoch": 19, "lr": 9.999466425815647e-05}
{"train_loss": 1.5008292198181152, "global_step": 1760, "epoch": 19, "lr": 9.999465578551501e-05}
{"train_loss": 1.3977000713348389, "global_step": 1761, "epoch": 19, "lr": 9.99946473061524e-05}
{"train_loss": 1.332023024559021, "global_step": 1762, "epoch": 19, "lr": 9.999463882006861e-05}
{"train_loss": 1.4857759475708008, "global_step": 1763, "epoch": 19, "lr": 9.999463032726366e-05}
{"train_loss": 1.624930739402771, "global_step": 1764, "epoch": 19, "lr": 9.999462182773753e-05}
{"train_loss": 1.454331874847412, "global_step": 1765, "epoch": 19, "lr": 9.999461332149023e-05}
{"train_loss": 1.5282516479492188, "global_step": 1766, "epoch": 19, "lr": 9.999460480852176e-05}
{"train_loss": 1.5160987377166748, "global_step": 1767, "epoch": 19, "lr": 9.999459628883215e-05}
{"train_loss": 1.4007179737091064, "global_step": 1768, "epoch": 19, "lr": 9.999458776242135e-05}
{"train_loss": 1.3427093029022217, "global_step": 1769, "epoch": 19, "lr": 9.99945792292894e-05}
{"train_loss": 1.643829107284546, "global_step": 1770, "epoch": 19, "lr": 9.999457068943628e-05}
{"train_loss": 1.152889609336853, "global_step": 1771, "epoch": 19, "lr": 9.999456214286201e-05}
{"train_loss": 1.3688437938690186, "global_step": 1772, "epoch": 19, "lr": 9.999455358956658e-05}
{"train_loss": 1.438843846321106, "global_step": 1773, "epoch": 19, "lr": 9.999454502954998e-05}
{"train_loss": 1.7325646877288818, "global_step": 1774, "epoch": 19, "lr": 9.999453646281222e-05}
{"train_loss": 1.6083372831344604, "global_step": 1775, "epoch": 19, "lr": 9.999452788935333e-05}
{"train_loss": 1.3247541189193726, "global_step": 1776, "epoch": 19, "lr": 9.999451930917326e-05}
{"train_loss": 1.5672848224639893, "global_step": 1777, "epoch": 19, "lr": 9.999451072227204e-05}
{"train_loss": 1.5242977142333984, "global_step": 1778, "epoch": 19, "lr": 9.999450212864967e-05}
{"train_loss": 1.4918748113546478, "global_step": 1779, "epoch": 19, "lr": 9.999449352830616e-05, "val_loss": 1.4417228698730469}
{"train_loss": 1.484967827796936, "global_step": 1780, "epoch": 20, "lr": 9.999448492124149e-05}
{"train_loss": 1.5502142906188965, "global_step": 1781, "epoch": 20, "lr": 9.999447630745567e-05}
{"train_loss": 1.7826194763183594, "global_step": 1782, "epoch": 20, "lr": 9.99944676869487e-05}
{"train_loss": 1.5890783071517944, "global_step": 1783, "epoch": 20, "lr": 9.999445905972058e-05}
{"train_loss": 1.2785266637802124, "global_step": 1784, "epoch": 20, "lr": 9.999445042577133e-05}
{"train_loss": 1.2382748126983643, "global_step": 1785, "epoch": 20, "lr": 9.999444178510093e-05}
{"train_loss": 1.4494346380233765, "global_step": 1786, "epoch": 20, "lr": 9.999443313770938e-05}
{"train_loss": 1.440443515777588, "global_step": 1787, "epoch": 20, "lr": 9.999442448359669e-05}
{"train_loss": 1.510379433631897, "global_step": 1788, "epoch": 20, "lr": 9.999441582276285e-05}
{"train_loss": 1.3011372089385986, "global_step": 1789, "epoch": 20, "lr": 9.99944071552079e-05}
{"train_loss": 1.5647770166397095, "global_step": 1790, "epoch": 20, "lr": 9.999439848093177e-05}
{"train_loss": 1.1923432350158691, "global_step": 1791, "epoch": 20, "lr": 9.999438979993452e-05}
{"train_loss": 1.4613474607467651, "global_step": 1792, "epoch": 20, "lr": 9.999438111221615e-05}
{"train_loss": 1.3442838191986084, "global_step": 1793, "epoch": 20, "lr": 9.999437241777662e-05}
{"train_loss": 1.4173797369003296, "global_step": 1794, "epoch": 20, "lr": 9.999436371661597e-05}
{"train_loss": 1.3662948608398438, "global_step": 1795, "epoch": 20, "lr": 9.999435500873418e-05}
{"train_loss": 1.402843713760376, "global_step": 1796, "epoch": 20, "lr": 9.999434629413127e-05}
{"train_loss": 1.5866670608520508, "global_step": 1797, "epoch": 20, "lr": 9.999433757280722e-05}
{"train_loss": 1.2909058332443237, "global_step": 1798, "epoch": 20, "lr": 9.999432884476205e-05}
{"train_loss": 1.3561556339263916, "global_step": 1799, "epoch": 20, "lr": 9.999432010999574e-05}
{"train_loss": 1.4210126399993896, "global_step": 1800, "epoch": 20, "lr": 9.99943113685083e-05}
{"train_loss": 1.576336145401001, "global_step": 1801, "epoch": 20, "lr": 9.999430262029976e-05}
{"train_loss": 1.3301485776901245, "global_step": 1802, "epoch": 20, "lr": 9.999429386537007e-05}
{"train_loss": 1.4771932363510132, "global_step": 1803, "epoch": 20, "lr": 9.999428510371927e-05}
{"train_loss": 1.3322200775146484, "global_step": 1804, "epoch": 20, "lr": 9.999427633534734e-05}
{"train_loss": 1.448026418685913, "global_step": 1805, "epoch": 20, "lr": 9.999426756025429e-05}
{"train_loss": 1.4189929962158203, "global_step": 1806, "epoch": 20, "lr": 9.999425877844012e-05}
{"train_loss": 1.4490641355514526, "global_step": 1807, "epoch": 20, "lr": 9.999424998990483e-05}
{"train_loss": 1.5899962186813354, "global_step": 1808, "epoch": 20, "lr": 9.999424119464842e-05}
{"train_loss": 1.5631259679794312, "global_step": 1809, "epoch": 20, "lr": 9.99942323926709e-05}
{"train_loss": 1.6290513277053833, "global_step": 1810, "epoch": 20, "lr": 9.999422358397225e-05}
{"train_loss": 1.5091229677200317, "global_step": 1811, "epoch": 20, "lr": 9.99942147685525e-05}
{"train_loss": 1.451622724533081, "global_step": 1812, "epoch": 20, "lr": 9.999420594641165e-05}
{"train_loss": 1.6084779500961304, "global_step": 1813, "epoch": 20, "lr": 9.999419711754966e-05}
{"train_loss": 1.353789210319519, "global_step": 1814, "epoch": 20, "lr": 9.999418828196659e-05}
{"train_loss": 1.439016580581665, "global_step": 1815, "epoch": 20, "lr": 9.99941794396624e-05}
{"train_loss": 1.4517637491226196, "global_step": 1816, "epoch": 20, "lr": 9.99941705906371e-05}
{"train_loss": 1.6526862382888794, "global_step": 1817, "epoch": 20, "lr": 9.999416173489068e-05}
{"train_loss": 1.4403687715530396, "global_step": 1818, "epoch": 20, "lr": 9.999415287242317e-05}
{"train_loss": 1.419442057609558, "global_step": 1819, "epoch": 20, "lr": 9.999414400323454e-05}
{"train_loss": 1.2926207780838013, "global_step": 1820, "epoch": 20, "lr": 9.999413512732482e-05}
{"train_loss": 1.3059439659118652, "global_step": 1821, "epoch": 20, "lr": 9.9994126244694e-05}
{"train_loss": 1.1996804475784302, "global_step": 1822, "epoch": 20, "lr": 9.999411735534207e-05}
{"train_loss": 1.5657981634140015, "global_step": 1823, "epoch": 20, "lr": 9.999410845926904e-05}
{"train_loss": 1.6054118871688843, "global_step": 1824, "epoch": 20, "lr": 9.999409955647491e-05}
{"train_loss": 1.7597543001174927, "global_step": 1825, "epoch": 20, "lr": 9.99940906469597e-05}
{"train_loss": 1.6120569705963135, "global_step": 1826, "epoch": 20, "lr": 9.999408173072338e-05}
{"train_loss": 1.5496615171432495, "global_step": 1827, "epoch": 20, "lr": 9.999407280776597e-05}
{"train_loss": 1.614632248878479, "global_step": 1828, "epoch": 20, "lr": 9.999406387808745e-05}
{"train_loss": 1.4887256622314453, "global_step": 1829, "epoch": 20, "lr": 9.999405494168786e-05}
{"train_loss": 1.4167805910110474, "global_step": 1830, "epoch": 20, "lr": 9.999404599856717e-05}
{"train_loss": 1.4816933870315552, "global_step": 1831, "epoch": 20, "lr": 9.999403704872538e-05}
{"train_loss": 1.431419849395752, "global_step": 1832, "epoch": 20, "lr": 9.999402809216251e-05}
{"train_loss": 1.3599419593811035, "global_step": 1833, "epoch": 20, "lr": 9.999401912887855e-05}
{"train_loss": 1.385626196861267, "global_step": 1834, "epoch": 20, "lr": 9.99940101588735e-05}
{"train_loss": 1.6418880224227905, "global_step": 1835, "epoch": 20, "lr": 9.999400118214736e-05}
{"train_loss": 1.5126105546951294, "global_step": 1836, "epoch": 20, "lr": 9.999399219870016e-05}
{"train_loss": 1.4636985063552856, "global_step": 1837, "epoch": 20, "lr": 9.999398320853186e-05}
{"train_loss": 1.7483927011489868, "global_step": 1838, "epoch": 20, "lr": 9.999397421164249e-05}
{"train_loss": 1.5629208087921143, "global_step": 1839, "epoch": 20, "lr": 9.999396520803202e-05}
{"train_loss": 1.5212643146514893, "global_step": 1840, "epoch": 20, "lr": 9.999395619770049e-05}
{"train_loss": 1.606359839439392, "global_step": 1841, "epoch": 20, "lr": 9.999394718064787e-05}
{"train_loss": 1.3530404567718506, "global_step": 1842, "epoch": 20, "lr": 9.999393815687418e-05}
{"train_loss": 1.4418554306030273, "global_step": 1843, "epoch": 20, "lr": 9.999392912637942e-05}
{"train_loss": 1.406943678855896, "global_step": 1844, "epoch": 20, "lr": 9.999392008916357e-05}
{"train_loss": 1.612526535987854, "global_step": 1845, "epoch": 20, "lr": 9.999391104522665e-05}
{"train_loss": 1.4046056270599365, "global_step": 1846, "epoch": 20, "lr": 9.999390199456867e-05}
{"train_loss": 1.2959506511688232, "global_step": 1847, "epoch": 20, "lr": 9.999389293718959e-05}
{"train_loss": 1.4405553340911865, "global_step": 1848, "epoch": 20, "lr": 9.999388387308948e-05}
{"train_loss": 1.4760138988494873, "global_step": 1849, "epoch": 20, "lr": 9.999387480226827e-05}
{"train_loss": 1.488399863243103, "global_step": 1850, "epoch": 20, "lr": 9.999386572472601e-05}
{"train_loss": 1.5045682191848755, "global_step": 1851, "epoch": 20, "lr": 9.999385664046268e-05}
{"train_loss": 1.4540061950683594, "global_step": 1852, "epoch": 20, "lr": 9.999384754947829e-05}
{"train_loss": 1.4940086603164673, "global_step": 1853, "epoch": 20, "lr": 9.999383845177283e-05}
{"train_loss": 1.3945496082305908, "global_step": 1854, "epoch": 20, "lr": 9.999382934734631e-05}
{"train_loss": 1.4404217004776, "global_step": 1855, "epoch": 20, "lr": 9.999382023619872e-05}
{"train_loss": 1.367687463760376, "global_step": 1856, "epoch": 20, "lr": 9.999381111833008e-05}
{"train_loss": 1.3805214166641235, "global_step": 1857, "epoch": 20, "lr": 9.999380199374038e-05}
{"train_loss": 1.4152952432632446, "global_step": 1858, "epoch": 20, "lr": 9.999379286242963e-05}
{"train_loss": 1.4914766550064087, "global_step": 1859, "epoch": 20, "lr": 9.99937837243978e-05}
{"train_loss": 1.409543752670288, "global_step": 1860, "epoch": 20, "lr": 9.999377457964493e-05}
{"train_loss": 1.3855808973312378, "global_step": 1861, "epoch": 20, "lr": 9.999376542817103e-05}
{"train_loss": 1.467983603477478, "global_step": 1862, "epoch": 20, "lr": 9.999375626997605e-05}
{"train_loss": 1.3404295444488525, "global_step": 1863, "epoch": 20, "lr": 9.999374710506002e-05}
{"train_loss": 1.5424160957336426, "global_step": 1864, "epoch": 20, "lr": 9.999373793342295e-05}
{"train_loss": 1.7783048152923584, "global_step": 1865, "epoch": 20, "lr": 9.999372875506481e-05}
{"train_loss": 1.6266887187957764, "global_step": 1866, "epoch": 20, "lr": 9.999371956998564e-05}
{"train_loss": 1.4236729145050049, "global_step": 1867, "epoch": 20, "lr": 9.999371037818544e-05}
{"train_loss": 1.467728780896476, "global_step": 1868, "epoch": 20, "lr": 9.999370117966417e-05, "val_loss": 1.3707828521728516, "train_action_mse_error": 87.06619262695312}
{"train_loss": 1.2585723400115967, "global_step": 1869, "epoch": 21, "lr": 9.999369197442186e-05}
{"train_loss": 1.204729676246643, "global_step": 1870, "epoch": 21, "lr": 9.999368276245851e-05}
{"train_loss": 1.291107416152954, "global_step": 1871, "epoch": 21, "lr": 9.999367354377412e-05}
{"train_loss": 1.3991259336471558, "global_step": 1872, "epoch": 21, "lr": 9.999366431836868e-05}
{"train_loss": 1.3791635036468506, "global_step": 1873, "epoch": 21, "lr": 9.999365508624221e-05}
{"train_loss": 1.058335542678833, "global_step": 1874, "epoch": 21, "lr": 9.999364584739472e-05}
{"train_loss": 1.4654181003570557, "global_step": 1875, "epoch": 21, "lr": 9.999363660182617e-05}
{"train_loss": 1.3590916395187378, "global_step": 1876, "epoch": 21, "lr": 9.99936273495366e-05}
{"train_loss": 1.3528012037277222, "global_step": 1877, "epoch": 21, "lr": 9.999361809052598e-05}
{"train_loss": 1.4304232597351074, "global_step": 1878, "epoch": 21, "lr": 9.999360882479433e-05}
{"train_loss": 1.4164365530014038, "global_step": 1879, "epoch": 21, "lr": 9.999359955234166e-05}
{"train_loss": 1.335654616355896, "global_step": 1880, "epoch": 21, "lr": 9.999359027316795e-05}
{"train_loss": 1.293043851852417, "global_step": 1881, "epoch": 21, "lr": 9.99935809872732e-05}
{"train_loss": 1.345106840133667, "global_step": 1882, "epoch": 21, "lr": 9.999357169465744e-05}
{"train_loss": 1.2621185779571533, "global_step": 1883, "epoch": 21, "lr": 9.999356239532065e-05}
{"train_loss": 1.4065876007080078, "global_step": 1884, "epoch": 21, "lr": 9.999355308926284e-05}
{"train_loss": 1.2346256971359253, "global_step": 1885, "epoch": 21, "lr": 9.9993543776484e-05}
{"train_loss": 1.517448902130127, "global_step": 1886, "epoch": 21, "lr": 9.999353445698414e-05}
{"train_loss": 1.4071465730667114, "global_step": 1887, "epoch": 21, "lr": 9.999352513076326e-05}
{"train_loss": 1.4832733869552612, "global_step": 1888, "epoch": 21, "lr": 9.999351579782135e-05}
{"train_loss": 1.3486285209655762, "global_step": 1889, "epoch": 21, "lr": 9.999350645815844e-05}
{"train_loss": 1.5439757108688354, "global_step": 1890, "epoch": 21, "lr": 9.99934971117745e-05}
{"train_loss": 1.299044132232666, "global_step": 1891, "epoch": 21, "lr": 9.999348775866954e-05}
{"train_loss": 1.4649361371994019, "global_step": 1892, "epoch": 21, "lr": 9.999347839884356e-05}
{"train_loss": 1.4251059293746948, "global_step": 1893, "epoch": 21, "lr": 9.999346903229658e-05}
{"train_loss": 1.356391429901123, "global_step": 1894, "epoch": 21, "lr": 9.999345965902859e-05}
{"train_loss": 1.5128298997879028, "global_step": 1895, "epoch": 21, "lr": 9.999345027903958e-05}
{"train_loss": 1.4059255123138428, "global_step": 1896, "epoch": 21, "lr": 9.999344089232957e-05}
{"train_loss": 1.4188255071640015, "global_step": 1897, "epoch": 21, "lr": 9.999343149889853e-05}
{"train_loss": 1.4336298704147339, "global_step": 1898, "epoch": 21, "lr": 9.999342209874649e-05}
{"train_loss": 1.5149022340774536, "global_step": 1899, "epoch": 21, "lr": 9.999341269187346e-05}
{"train_loss": 1.3510462045669556, "global_step": 1900, "epoch": 21, "lr": 9.999340327827941e-05}
{"train_loss": 1.3443357944488525, "global_step": 1901, "epoch": 21, "lr": 9.999339385796435e-05}
{"train_loss": 1.4268591403961182, "global_step": 1902, "epoch": 21, "lr": 9.999338443092831e-05}
{"train_loss": 1.659347653388977, "global_step": 1903, "epoch": 21, "lr": 9.999337499717126e-05}
{"train_loss": 1.325093150138855, "global_step": 1904, "epoch": 21, "lr": 9.999336555669321e-05}
{"train_loss": 1.641424536705017, "global_step": 1905, "epoch": 21, "lr": 9.999335610949415e-05}
{"train_loss": 1.2541710138320923, "global_step": 1906, "epoch": 21, "lr": 9.999334665557411e-05}
{"train_loss": 1.4957941770553589, "global_step": 1907, "epoch": 21, "lr": 9.999333719493306e-05}
{"train_loss": 1.4072651863098145, "global_step": 1908, "epoch": 21, "lr": 9.999332772757102e-05}
{"train_loss": 1.2936543226242065, "global_step": 1909, "epoch": 21, "lr": 9.999331825348799e-05}
{"train_loss": 1.6633800268173218, "global_step": 1910, "epoch": 21, "lr": 9.999330877268395e-05}
{"train_loss": 1.5781816244125366, "global_step": 1911, "epoch": 21, "lr": 9.999329928515894e-05}
{"train_loss": 1.352980613708496, "global_step": 1912, "epoch": 21, "lr": 9.999328979091294e-05}
{"train_loss": 1.3722387552261353, "global_step": 1913, "epoch": 21, "lr": 9.999328028994593e-05}
{"train_loss": 1.430333137512207, "global_step": 1914, "epoch": 21, "lr": 9.999327078225795e-05}
{"train_loss": 1.4187819957733154, "global_step": 1915, "epoch": 21, "lr": 9.999326126784898e-05}
{"train_loss": 1.4789807796478271, "global_step": 1916, "epoch": 21, "lr": 9.999325174671903e-05}
{"train_loss": 1.4855462312698364, "global_step": 1917, "epoch": 21, "lr": 9.99932422188681e-05}
{"train_loss": 1.4471945762634277, "global_step": 1918, "epoch": 21, "lr": 9.999323268429616e-05}
{"train_loss": 1.125460147857666, "global_step": 1919, "epoch": 21, "lr": 9.999322314300328e-05}
{"train_loss": 1.4594546556472778, "global_step": 1920, "epoch": 21, "lr": 9.99932135949894e-05}
{"train_loss": 1.4576382637023926, "global_step": 1921, "epoch": 21, "lr": 9.999320404025454e-05}
{"train_loss": 1.4460408687591553, "global_step": 1922, "epoch": 21, "lr": 9.999319447879871e-05}
{"train_loss": 1.1784985065460205, "global_step": 1923, "epoch": 21, "lr": 9.999318491062189e-05}
{"train_loss": 1.3977186679840088, "global_step": 1924, "epoch": 21, "lr": 9.99931753357241e-05}
{"train_loss": 1.633049488067627, "global_step": 1925, "epoch": 21, "lr": 9.999316575410535e-05}
{"train_loss": 1.4082399606704712, "global_step": 1926, "epoch": 21, "lr": 9.999315616576562e-05}
{"train_loss": 1.665276288986206, "global_step": 1927, "epoch": 21, "lr": 9.999314657070493e-05}
{"train_loss": 1.5907942056655884, "global_step": 1928, "epoch": 21, "lr": 9.999313696892325e-05}
{"train_loss": 1.3700371980667114, "global_step": 1929, "epoch": 21, "lr": 9.999312736042061e-05}
{"train_loss": 1.412635087966919, "global_step": 1930, "epoch": 21, "lr": 9.999311774519702e-05}
{"train_loss": 1.3165562152862549, "global_step": 1931, "epoch": 21, "lr": 9.999310812325245e-05}
{"train_loss": 1.7266043424606323, "global_step": 1932, "epoch": 21, "lr": 9.99930984945869e-05}
{"train_loss": 1.3858790397644043, "global_step": 1933, "epoch": 21, "lr": 9.999308885920042e-05}
{"train_loss": 1.4693844318389893, "global_step": 1934, "epoch": 21, "lr": 9.999307921709296e-05}
{"train_loss": 1.4703824520111084, "global_step": 1935, "epoch": 21, "lr": 9.999306956826454e-05}
{"train_loss": 1.448373556137085, "global_step": 1936, "epoch": 21, "lr": 9.999305991271517e-05}
{"train_loss": 1.2609277963638306, "global_step": 1937, "epoch": 21, "lr": 9.999305025044483e-05}
{"train_loss": 1.3300182819366455, "global_step": 1938, "epoch": 21, "lr": 9.999304058145354e-05}
{"train_loss": 1.4505977630615234, "global_step": 1939, "epoch": 21, "lr": 9.999303090574131e-05}
{"train_loss": 1.3496767282485962, "global_step": 1940, "epoch": 21, "lr": 9.99930212233081e-05}
{"train_loss": 1.5382874011993408, "global_step": 1941, "epoch": 21, "lr": 9.999301153415395e-05}
{"train_loss": 1.2478528022766113, "global_step": 1942, "epoch": 21, "lr": 9.999300183827884e-05}
{"train_loss": 1.5609467029571533, "global_step": 1943, "epoch": 21, "lr": 9.99929921356828e-05}
{"train_loss": 1.2025585174560547, "global_step": 1944, "epoch": 21, "lr": 9.999298242636579e-05}
{"train_loss": 1.3582375049591064, "global_step": 1945, "epoch": 21, "lr": 9.999297271032784e-05}
{"train_loss": 1.6264632940292358, "global_step": 1946, "epoch": 21, "lr": 9.999296298756895e-05}
{"train_loss": 1.5411678552627563, "global_step": 1947, "epoch": 21, "lr": 9.999295325808911e-05}
{"train_loss": 1.6356967687606812, "global_step": 1948, "epoch": 21, "lr": 9.999294352188833e-05}
{"train_loss": 1.3893413543701172, "global_step": 1949, "epoch": 21, "lr": 9.999293377896661e-05}
{"train_loss": 1.3586515188217163, "global_step": 1950, "epoch": 21, "lr": 9.999292402932394e-05}
{"train_loss": 1.3677635192871094, "global_step": 1951, "epoch": 21, "lr": 9.999291427296033e-05}
{"train_loss": 1.4977473020553589, "global_step": 1952, "epoch": 21, "lr": 9.99929045098758e-05}
{"train_loss": 1.438697099685669, "global_step": 1953, "epoch": 21, "lr": 9.999289474007032e-05}
{"train_loss": 1.5262982845306396, "global_step": 1954, "epoch": 21, "lr": 9.999288496354391e-05}
{"train_loss": 1.2062300443649292, "global_step": 1955, "epoch": 21, "lr": 9.999287518029656e-05}
{"train_loss": 1.3332302570343018, "global_step": 1956, "epoch": 21, "lr": 9.999286539032829e-05}
{"train_loss": 1.4119432267178311, "global_step": 1957, "epoch": 21, "lr": 9.999285559363907e-05, "val_loss": 1.3866544961929321}
{"train_loss": 1.3080676794052124, "global_step": 1958, "epoch": 22, "lr": 9.999284579022892e-05}
{"train_loss": 1.2963923215866089, "global_step": 1959, "epoch": 22, "lr": 9.999283598009785e-05}
{"train_loss": 1.342732310295105, "global_step": 1960, "epoch": 22, "lr": 9.999282616324586e-05}
{"train_loss": 1.5072981119155884, "global_step": 1961, "epoch": 22, "lr": 9.999281633967294e-05}
{"train_loss": 1.445459008216858, "global_step": 1962, "epoch": 22, "lr": 9.999280650937908e-05}
{"train_loss": 1.4253114461898804, "global_step": 1963, "epoch": 22, "lr": 9.999279667236432e-05}
{"train_loss": 1.4951529502868652, "global_step": 1964, "epoch": 22, "lr": 9.999278682862862e-05}
{"train_loss": 1.5338081121444702, "global_step": 1965, "epoch": 22, "lr": 9.999277697817201e-05}
{"train_loss": 1.5976567268371582, "global_step": 1966, "epoch": 22, "lr": 9.999276712099448e-05}
{"train_loss": 1.4531776905059814, "global_step": 1967, "epoch": 22, "lr": 9.999275725709602e-05}
{"train_loss": 1.3823745250701904, "global_step": 1968, "epoch": 22, "lr": 9.999274738647664e-05}
{"train_loss": 1.418783187866211, "global_step": 1969, "epoch": 22, "lr": 9.999273750913636e-05}
{"train_loss": 1.4012998342514038, "global_step": 1970, "epoch": 22, "lr": 9.999272762507517e-05}
{"train_loss": 1.3067127466201782, "global_step": 1971, "epoch": 22, "lr": 9.999271773429305e-05}
{"train_loss": 1.4366004467010498, "global_step": 1972, "epoch": 22, "lr": 9.999270783679002e-05}
{"train_loss": 1.1429935693740845, "global_step": 1973, "epoch": 22, "lr": 9.99926979325661e-05}
{"train_loss": 1.6022250652313232, "global_step": 1974, "epoch": 22, "lr": 9.999268802162124e-05}
{"train_loss": 1.434794545173645, "global_step": 1975, "epoch": 22, "lr": 9.999267810395549e-05}
{"train_loss": 1.29721999168396, "global_step": 1976, "epoch": 22, "lr": 9.999266817956884e-05}
{"train_loss": 1.4214401245117188, "global_step": 1977, "epoch": 22, "lr": 9.999265824846127e-05}
{"train_loss": 1.5509834289550781, "global_step": 1978, "epoch": 22, "lr": 9.999264831063281e-05}
{"train_loss": 1.2977416515350342, "global_step": 1979, "epoch": 22, "lr": 9.999263836608345e-05}
{"train_loss": 1.399235486984253, "global_step": 1980, "epoch": 22, "lr": 9.999262841481319e-05}
{"train_loss": 1.5118366479873657, "global_step": 1981, "epoch": 22, "lr": 9.999261845682202e-05}
{"train_loss": 1.3163185119628906, "global_step": 1982, "epoch": 22, "lr": 9.999260849210994e-05}
{"train_loss": 1.2637666463851929, "global_step": 1983, "epoch": 22, "lr": 9.999259852067698e-05}
{"train_loss": 1.477276086807251, "global_step": 1984, "epoch": 22, "lr": 9.999258854252311e-05}
{"train_loss": 1.3957685232162476, "global_step": 1985, "epoch": 22, "lr": 9.999257855764836e-05}
{"train_loss": 1.6910444498062134, "global_step": 1986, "epoch": 22, "lr": 9.999256856605272e-05}
{"train_loss": 1.5156333446502686, "global_step": 1987, "epoch": 22, "lr": 9.999255856773619e-05}
{"train_loss": 1.303588628768921, "global_step": 1988, "epoch": 22, "lr": 9.999254856269875e-05}
{"train_loss": 1.4942338466644287, "global_step": 1989, "epoch": 22, "lr": 9.999253855094044e-05}
{"train_loss": 1.3536075353622437, "global_step": 1990, "epoch": 22, "lr": 9.999252853246123e-05}
{"train_loss": 1.4884899854660034, "global_step": 1991, "epoch": 22, "lr": 9.999251850726115e-05}
{"train_loss": 1.105116367340088, "global_step": 1992, "epoch": 22, "lr": 9.999250847534016e-05}
{"train_loss": 1.415196180343628, "global_step": 1993, "epoch": 22, "lr": 9.999249843669832e-05}
{"train_loss": 1.3984942436218262, "global_step": 1994, "epoch": 22, "lr": 9.999248839133557e-05}
{"train_loss": 1.3831475973129272, "global_step": 1995, "epoch": 22, "lr": 9.999247833925194e-05}
{"train_loss": 1.4114562273025513, "global_step": 1996, "epoch": 22, "lr": 9.999246828044745e-05}
{"train_loss": 1.4312283992767334, "global_step": 1997, "epoch": 22, "lr": 9.999245821492207e-05}
{"train_loss": 1.3763949871063232, "global_step": 1998, "epoch": 22, "lr": 9.999244814267581e-05}
{"train_loss": 1.3733566999435425, "global_step": 1999, "epoch": 22, "lr": 9.999243806370868e-05}
{"train_loss": 1.4775362014770508, "global_step": 2000, "epoch": 22, "lr": 9.999242797802069e-05}
{"train_loss": 1.390997052192688, "global_step": 2001, "epoch": 22, "lr": 9.99924178856118e-05}
{"train_loss": 1.3189040422439575, "global_step": 2002, "epoch": 22, "lr": 9.999240778648206e-05}
{"train_loss": 1.1933215856552124, "global_step": 2003, "epoch": 22, "lr": 9.999239768063143e-05}
{"train_loss": 1.3387364149093628, "global_step": 2004, "epoch": 22, "lr": 9.999238756805995e-05}
{"train_loss": 1.5017693042755127, "global_step": 2005, "epoch": 22, "lr": 9.99923774487676e-05}
{"train_loss": 1.4220247268676758, "global_step": 2006, "epoch": 22, "lr": 9.999236732275438e-05}
{"train_loss": 1.2549163103103638, "global_step": 2007, "epoch": 22, "lr": 9.999235719002029e-05}
{"train_loss": 1.2981224060058594, "global_step": 2008, "epoch": 22, "lr": 9.999234705056534e-05}
{"train_loss": 1.3029720783233643, "global_step": 2009, "epoch": 22, "lr": 9.999233690438953e-05}
{"train_loss": 1.366994023323059, "global_step": 2010, "epoch": 22, "lr": 9.999232675149286e-05}
{"train_loss": 1.4100134372711182, "global_step": 2011, "epoch": 22, "lr": 9.999231659187533e-05}
{"train_loss": 1.5222375392913818, "global_step": 2012, "epoch": 22, "lr": 9.999230642553694e-05}
{"train_loss": 1.398977518081665, "global_step": 2013, "epoch": 22, "lr": 9.99922962524777e-05}
{"train_loss": 1.3911601305007935, "global_step": 2014, "epoch": 22, "lr": 9.999228607269761e-05}
{"train_loss": 1.4594471454620361, "global_step": 2015, "epoch": 22, "lr": 9.999227588619666e-05}
{"train_loss": 1.252475380897522, "global_step": 2016, "epoch": 22, "lr": 9.999226569297485e-05}
{"train_loss": 1.4269708395004272, "global_step": 2017, "epoch": 22, "lr": 9.99922554930322e-05}
{"train_loss": 1.298756718635559, "global_step": 2018, "epoch": 22, "lr": 9.99922452863687e-05}
{"train_loss": 1.3093339204788208, "global_step": 2019, "epoch": 22, "lr": 9.999223507298434e-05}
{"train_loss": 1.4412713050842285, "global_step": 2020, "epoch": 22, "lr": 9.999222485287914e-05}
{"train_loss": 1.6024683713912964, "global_step": 2021, "epoch": 22, "lr": 9.999221462605311e-05}
{"train_loss": 1.4760313034057617, "global_step": 2022, "epoch": 22, "lr": 9.999220439250622e-05}
{"train_loss": 1.309439778327942, "global_step": 2023, "epoch": 22, "lr": 9.999219415223848e-05}
{"train_loss": 1.3228061199188232, "global_step": 2024, "epoch": 22, "lr": 9.999218390524992e-05}
{"train_loss": 1.399931788444519, "global_step": 2025, "epoch": 22, "lr": 9.99921736515405e-05}
{"train_loss": 1.4976491928100586, "global_step": 2026, "epoch": 22, "lr": 9.999216339111027e-05}
{"train_loss": 1.1693480014801025, "global_step": 2027, "epoch": 22, "lr": 9.999215312395916e-05}
{"train_loss": 1.275851845741272, "global_step": 2028, "epoch": 22, "lr": 9.999214285008724e-05}
{"train_loss": 1.3859516382217407, "global_step": 2029, "epoch": 22, "lr": 9.999213256949449e-05}
{"train_loss": 1.4541221857070923, "global_step": 2030, "epoch": 22, "lr": 9.99921222821809e-05}
{"train_loss": 1.3642569780349731, "global_step": 2031, "epoch": 22, "lr": 9.999211198814647e-05}
{"train_loss": 1.3636479377746582, "global_step": 2032, "epoch": 22, "lr": 9.999210168739123e-05}
{"train_loss": 1.2919135093688965, "global_step": 2033, "epoch": 22, "lr": 9.999209137991514e-05}
{"train_loss": 1.4207484722137451, "global_step": 2034, "epoch": 22, "lr": 9.999208106571824e-05}
{"train_loss": 1.3827078342437744, "global_step": 2035, "epoch": 22, "lr": 9.999207074480051e-05}
{"train_loss": 1.2209526300430298, "global_step": 2036, "epoch": 22, "lr": 9.999206041716194e-05}
{"train_loss": 1.1588407754898071, "global_step": 2037, "epoch": 22, "lr": 9.999205008280257e-05}
{"train_loss": 1.3261889219284058, "global_step": 2038, "epoch": 22, "lr": 9.999203974172236e-05}
{"train_loss": 1.4921818971633911, "global_step": 2039, "epoch": 22, "lr": 9.999202939392134e-05}
{"train_loss": 1.6056461334228516, "global_step": 2040, "epoch": 22, "lr": 9.999201903939949e-05}
{"train_loss": 1.4481450319290161, "global_step": 2041, "epoch": 22, "lr": 9.999200867815684e-05}
{"train_loss": 1.496169090270996, "global_step": 2042, "epoch": 22, "lr": 9.999199831019337e-05}
{"train_loss": 1.3818204402923584, "global_step": 2043, "epoch": 22, "lr": 9.999198793550907e-05}
{"train_loss": 1.4056084156036377, "global_step": 2044, "epoch": 22, "lr": 9.999197755410396e-05}
{"train_loss": 1.2582285404205322, "global_step": 2045, "epoch": 22, "lr": 9.999196716597805e-05}
{"train_loss": 1.3903742198194011, "global_step": 2046, "epoch": 22, "lr": 9.999195677113133e-05, "val_loss": 1.2773492336273193}
{"train_loss": 1.127852201461792, "global_step": 2047, "epoch": 23, "lr": 9.99919463695638e-05}
{"train_loss": 1.311326026916504, "global_step": 2048, "epoch": 23, "lr": 9.999193596127546e-05}
{"train_loss": 1.4200941324234009, "global_step": 2049, "epoch": 23, "lr": 9.999192554626628e-05}
{"train_loss": 1.4556434154510498, "global_step": 2050, "epoch": 23, "lr": 9.999191512453634e-05}
{"train_loss": 1.5193145275115967, "global_step": 2051, "epoch": 23, "lr": 9.999190469608557e-05}
{"train_loss": 1.195569634437561, "global_step": 2052, "epoch": 23, "lr": 9.999189426091402e-05}
{"train_loss": 1.4932119846343994, "global_step": 2053, "epoch": 23, "lr": 9.999188381902164e-05}
{"train_loss": 1.4528326988220215, "global_step": 2054, "epoch": 23, "lr": 9.999187337040849e-05}
{"train_loss": 1.3947525024414062, "global_step": 2055, "epoch": 23, "lr": 9.999186291507453e-05}
{"train_loss": 1.2655211687088013, "global_step": 2056, "epoch": 23, "lr": 9.999185245301978e-05}
{"train_loss": 1.4792512655258179, "global_step": 2057, "epoch": 23, "lr": 9.999184198424423e-05}
{"train_loss": 1.379516363143921, "global_step": 2058, "epoch": 23, "lr": 9.999183150874787e-05}
{"train_loss": 1.3696600198745728, "global_step": 2059, "epoch": 23, "lr": 9.999182102653072e-05}
{"train_loss": 1.1387799978256226, "global_step": 2060, "epoch": 23, "lr": 9.999181053759281e-05}
{"train_loss": 1.4086908102035522, "global_step": 2061, "epoch": 23, "lr": 9.99918000419341e-05}
{"train_loss": 1.2419956922531128, "global_step": 2062, "epoch": 23, "lr": 9.999178953955458e-05}
{"train_loss": 1.2149325609207153, "global_step": 2063, "epoch": 23, "lr": 9.99917790304543e-05}
{"train_loss": 1.238739013671875, "global_step": 2064, "epoch": 23, "lr": 9.999176851463321e-05}
{"train_loss": 1.1709880828857422, "global_step": 2065, "epoch": 23, "lr": 9.999175799209135e-05}
{"train_loss": 1.274094820022583, "global_step": 2066, "epoch": 23, "lr": 9.99917474628287e-05}
{"train_loss": 1.5531294345855713, "global_step": 2067, "epoch": 23, "lr": 9.999173692684527e-05}
{"train_loss": 1.5613064765930176, "global_step": 2068, "epoch": 23, "lr": 9.999172638414107e-05}
{"train_loss": 1.3640886545181274, "global_step": 2069, "epoch": 23, "lr": 9.999171583471609e-05}
{"train_loss": 1.2670923471450806, "global_step": 2070, "epoch": 23, "lr": 9.999170527857031e-05}
{"train_loss": 1.5512888431549072, "global_step": 2071, "epoch": 23, "lr": 9.999169471570377e-05}
{"train_loss": 1.3388915061950684, "global_step": 2072, "epoch": 23, "lr": 9.999168414611647e-05}
{"train_loss": 1.2063385248184204, "global_step": 2073, "epoch": 23, "lr": 9.999167356980838e-05}
{"train_loss": 1.1575207710266113, "global_step": 2074, "epoch": 23, "lr": 9.999166298677953e-05}
{"train_loss": 1.2842869758605957, "global_step": 2075, "epoch": 23, "lr": 9.99916523970299e-05}
{"train_loss": 1.4026384353637695, "global_step": 2076, "epoch": 23, "lr": 9.999164180055951e-05}
{"train_loss": 1.4113352298736572, "global_step": 2077, "epoch": 23, "lr": 9.999163119736835e-05}
{"train_loss": 1.221243143081665, "global_step": 2078, "epoch": 23, "lr": 9.999162058745642e-05}
{"train_loss": 1.089455246925354, "global_step": 2079, "epoch": 23, "lr": 9.999160997082373e-05}
{"train_loss": 1.2849944829940796, "global_step": 2080, "epoch": 23, "lr": 9.99915993474703e-05}
{"train_loss": 1.1138194799423218, "global_step": 2081, "epoch": 23, "lr": 9.999158871739607e-05}
{"train_loss": 1.2925876379013062, "global_step": 2082, "epoch": 23, "lr": 9.999157808060109e-05}
{"train_loss": 1.514908790588379, "global_step": 2083, "epoch": 23, "lr": 9.999156743708537e-05}
{"train_loss": 1.4944589138031006, "global_step": 2084, "epoch": 23, "lr": 9.999155678684888e-05}
{"train_loss": 1.3095124959945679, "global_step": 2085, "epoch": 23, "lr": 9.999154612989163e-05}
{"train_loss": 1.2897510528564453, "global_step": 2086, "epoch": 23, "lr": 9.999153546621363e-05}
{"train_loss": 1.11312735080719, "global_step": 2087, "epoch": 23, "lr": 9.999152479581489e-05}
{"train_loss": 1.4387471675872803, "global_step": 2088, "epoch": 23, "lr": 9.999151411869538e-05}
{"train_loss": 1.2061502933502197, "global_step": 2089, "epoch": 23, "lr": 9.999150343485512e-05}
{"train_loss": 1.3543670177459717, "global_step": 2090, "epoch": 23, "lr": 9.999149274429414e-05}
{"train_loss": 1.5150436162948608, "global_step": 2091, "epoch": 23, "lr": 9.999148204701239e-05}
{"train_loss": 1.29443359375, "global_step": 2092, "epoch": 23, "lr": 9.999147134300988e-05}
{"train_loss": 1.4205366373062134, "global_step": 2093, "epoch": 23, "lr": 9.999146063228665e-05}
{"train_loss": 1.5361642837524414, "global_step": 2094, "epoch": 23, "lr": 9.999144991484267e-05}
{"train_loss": 1.4616832733154297, "global_step": 2095, "epoch": 23, "lr": 9.999143919067795e-05}
{"train_loss": 1.1731126308441162, "global_step": 2096, "epoch": 23, "lr": 9.999142845979249e-05}
{"train_loss": 1.1780024766921997, "global_step": 2097, "epoch": 23, "lr": 9.99914177221863e-05}
{"train_loss": 1.2547236680984497, "global_step": 2098, "epoch": 23, "lr": 9.999140697785936e-05}
{"train_loss": 1.1983364820480347, "global_step": 2099, "epoch": 23, "lr": 9.999139622681168e-05}
{"train_loss": 1.391406774520874, "global_step": 2100, "epoch": 23, "lr": 9.999138546904328e-05}
{"train_loss": 1.4214086532592773, "global_step": 2101, "epoch": 23, "lr": 9.999137470455415e-05}
{"train_loss": 1.1466524600982666, "global_step": 2102, "epoch": 23, "lr": 9.999136393334427e-05}
{"train_loss": 1.3515392541885376, "global_step": 2103, "epoch": 23, "lr": 9.999135315541368e-05}
{"train_loss": 1.3754397630691528, "global_step": 2104, "epoch": 23, "lr": 9.999134237076235e-05}
{"train_loss": 1.3824764490127563, "global_step": 2105, "epoch": 23, "lr": 9.99913315793903e-05}
{"train_loss": 1.288050889968872, "global_step": 2106, "epoch": 23, "lr": 9.999132078129752e-05}
{"train_loss": 1.384938359260559, "global_step": 2107, "epoch": 23, "lr": 9.999130997648403e-05}
{"train_loss": 1.2640646696090698, "global_step": 2108, "epoch": 23, "lr": 9.99912991649498e-05}
{"train_loss": 1.4499839544296265, "global_step": 2109, "epoch": 23, "lr": 9.999128834669485e-05}
{"train_loss": 1.2748926877975464, "global_step": 2110, "epoch": 23, "lr": 9.99912775217192e-05}
{"train_loss": 1.257922887802124, "global_step": 2111, "epoch": 23, "lr": 9.99912666900228e-05}
{"train_loss": 1.2931528091430664, "global_step": 2112, "epoch": 23, "lr": 9.999125585160572e-05}
{"train_loss": 1.3132730722427368, "global_step": 2113, "epoch": 23, "lr": 9.99912450064679e-05}
{"train_loss": 1.3573740720748901, "global_step": 2114, "epoch": 23, "lr": 9.999123415460937e-05}
{"train_loss": 1.3004364967346191, "global_step": 2115, "epoch": 23, "lr": 9.999122329603013e-05}
{"train_loss": 1.360018014907837, "global_step": 2116, "epoch": 23, "lr": 9.999121243073019e-05}
{"train_loss": 1.490903615951538, "global_step": 2117, "epoch": 23, "lr": 9.999120155870953e-05}
{"train_loss": 1.2660753726959229, "global_step": 2118, "epoch": 23, "lr": 9.999119067996817e-05}
{"train_loss": 1.5438578128814697, "global_step": 2119, "epoch": 23, "lr": 9.999117979450609e-05}
{"train_loss": 1.441758155822754, "global_step": 2120, "epoch": 23, "lr": 9.99911689023233e-05}
{"train_loss": 1.379479169845581, "global_step": 2121, "epoch": 23, "lr": 9.999115800341982e-05}
{"train_loss": 1.3502330780029297, "global_step": 2122, "epoch": 23, "lr": 9.999114709779564e-05}
{"train_loss": 1.507718801498413, "global_step": 2123, "epoch": 23, "lr": 9.999113618545076e-05}
{"train_loss": 1.3650290966033936, "global_step": 2124, "epoch": 23, "lr": 9.999112526638518e-05}
{"train_loss": 1.2120988368988037, "global_step": 2125, "epoch": 23, "lr": 9.999111434059889e-05}
{"train_loss": 1.1419578790664673, "global_step": 2126, "epoch": 23, "lr": 9.999110340809193e-05}
{"train_loss": 1.4467613697052002, "global_step": 2127, "epoch": 23, "lr": 9.999109246886424e-05}
{"train_loss": 1.5357199907302856, "global_step": 2128, "epoch": 23, "lr": 9.999108152291589e-05}
{"train_loss": 1.6206285953521729, "global_step": 2129, "epoch": 23, "lr": 9.999107057024684e-05}
{"train_loss": 1.30263090133667, "global_step": 2130, "epoch": 23, "lr": 9.999105961085709e-05}
{"train_loss": 1.2371768951416016, "global_step": 2131, "epoch": 23, "lr": 9.999104864474665e-05}
{"train_loss": 1.3916279077529907, "global_step": 2132, "epoch": 23, "lr": 9.999103767191554e-05}
{"train_loss": 1.417450189590454, "global_step": 2133, "epoch": 23, "lr": 9.999102669236373e-05}
{"train_loss": 1.2704484462738037, "global_step": 2134, "epoch": 23, "lr": 9.999101570609124e-05}
{"train_loss": 1.3398040667008817, "global_step": 2135, "epoch": 23, "lr": 9.999100471309806e-05, "val_loss": 1.2940932512283325}
{"train_loss": 1.2471554279327393, "global_step": 2136, "epoch": 24, "lr": 9.999099371338421e-05}
{"train_loss": 1.2676085233688354, "global_step": 2137, "epoch": 24, "lr": 9.999098270694967e-05}
{"train_loss": 1.253204345703125, "global_step": 2138, "epoch": 24, "lr": 9.999097169379446e-05}
{"train_loss": 1.3979787826538086, "global_step": 2139, "epoch": 24, "lr": 9.999096067391857e-05}
{"train_loss": 1.3767454624176025, "global_step": 2140, "epoch": 24, "lr": 9.999094964732201e-05}
{"train_loss": 1.3965409994125366, "global_step": 2141, "epoch": 24, "lr": 9.999093861400477e-05}
{"train_loss": 1.2209384441375732, "global_step": 2142, "epoch": 24, "lr": 9.999092757396685e-05}
{"train_loss": 1.1623029708862305, "global_step": 2143, "epoch": 24, "lr": 9.999091652720826e-05}
{"train_loss": 1.3387393951416016, "global_step": 2144, "epoch": 24, "lr": 9.999090547372902e-05}
{"train_loss": 1.203157901763916, "global_step": 2145, "epoch": 24, "lr": 9.99908944135291e-05}
{"train_loss": 1.162469506263733, "global_step": 2146, "epoch": 24, "lr": 9.999088334660852e-05}
{"train_loss": 1.211196780204773, "global_step": 2147, "epoch": 24, "lr": 9.999087227296726e-05}
{"train_loss": 1.3586286306381226, "global_step": 2148, "epoch": 24, "lr": 9.999086119260534e-05}
{"train_loss": 1.3563711643218994, "global_step": 2149, "epoch": 24, "lr": 9.999085010552277e-05}
{"train_loss": 1.2813321352005005, "global_step": 2150, "epoch": 24, "lr": 9.999083901171953e-05}
{"train_loss": 1.4429445266723633, "global_step": 2151, "epoch": 24, "lr": 9.999082791119562e-05}
{"train_loss": 1.4102827310562134, "global_step": 2152, "epoch": 24, "lr": 9.999081680395108e-05}
{"train_loss": 1.2629941701889038, "global_step": 2153, "epoch": 24, "lr": 9.999080568998586e-05}
{"train_loss": 1.3841168880462646, "global_step": 2154, "epoch": 24, "lr": 9.99907945693e-05}
{"train_loss": 1.191264033317566, "global_step": 2155, "epoch": 24, "lr": 9.99907834418935e-05}
{"train_loss": 1.353610873222351, "global_step": 2156, "epoch": 24, "lr": 9.999077230776632e-05}
{"train_loss": 1.3906364440917969, "global_step": 2157, "epoch": 24, "lr": 9.99907611669185e-05}
{"train_loss": 1.2544376850128174, "global_step": 2158, "epoch": 24, "lr": 9.999075001935004e-05}
{"train_loss": 1.2238436937332153, "global_step": 2159, "epoch": 24, "lr": 9.999073886506091e-05}
{"train_loss": 1.4397711753845215, "global_step": 2160, "epoch": 24, "lr": 9.999072770405116e-05}
{"train_loss": 1.4781063795089722, "global_step": 2161, "epoch": 24, "lr": 9.999071653632075e-05}
{"train_loss": 1.2896660566329956, "global_step": 2162, "epoch": 24, "lr": 9.999070536186971e-05}
{"train_loss": 1.3574235439300537, "global_step": 2163, "epoch": 24, "lr": 9.999069418069801e-05}
{"train_loss": 1.3374754190444946, "global_step": 2164, "epoch": 24, "lr": 9.99906829928057e-05}
{"train_loss": 1.5214080810546875, "global_step": 2165, "epoch": 24, "lr": 9.999067179819273e-05}
{"train_loss": 1.4408624172210693, "global_step": 2166, "epoch": 24, "lr": 9.999066059685913e-05}
{"train_loss": 1.2033296823501587, "global_step": 2167, "epoch": 24, "lr": 9.999064938880489e-05}
{"train_loss": 1.324853777885437, "global_step": 2168, "epoch": 24, "lr": 9.999063817403003e-05}
{"train_loss": 1.2801710367202759, "global_step": 2169, "epoch": 24, "lr": 9.999062695253452e-05}
{"train_loss": 1.1954777240753174, "global_step": 2170, "epoch": 24, "lr": 9.999061572431838e-05}
{"train_loss": 1.3489048480987549, "global_step": 2171, "epoch": 24, "lr": 9.999060448938163e-05}
{"train_loss": 1.3714221715927124, "global_step": 2172, "epoch": 24, "lr": 9.999059324772424e-05}
{"train_loss": 1.5514979362487793, "global_step": 2173, "epoch": 24, "lr": 9.999058199934623e-05}
{"train_loss": 1.3384536504745483, "global_step": 2174, "epoch": 24, "lr": 9.999057074424758e-05}
{"train_loss": 1.3565987348556519, "global_step": 2175, "epoch": 24, "lr": 9.999055948242832e-05}
{"train_loss": 1.529284119606018, "global_step": 2176, "epoch": 24, "lr": 9.999054821388844e-05}
{"train_loss": 1.5101683139801025, "global_step": 2177, "epoch": 24, "lr": 9.999053693862793e-05}
{"train_loss": 1.4020394086837769, "global_step": 2178, "epoch": 24, "lr": 9.999052565664681e-05}
{"train_loss": 1.139148473739624, "global_step": 2179, "epoch": 24, "lr": 9.999051436794507e-05}
{"train_loss": 1.2866216897964478, "global_step": 2180, "epoch": 24, "lr": 9.999050307252271e-05}
{"train_loss": 1.411219835281372, "global_step": 2181, "epoch": 24, "lr": 9.999049177037974e-05}
{"train_loss": 1.7389066219329834, "global_step": 2182, "epoch": 24, "lr": 9.999048046151616e-05}
{"train_loss": 1.153157114982605, "global_step": 2183, "epoch": 24, "lr": 9.999046914593198e-05}
{"train_loss": 1.1334500312805176, "global_step": 2184, "epoch": 24, "lr": 9.999045782362718e-05}
{"train_loss": 1.5603876113891602, "global_step": 2185, "epoch": 24, "lr": 9.999044649460176e-05}
{"train_loss": 1.3655117750167847, "global_step": 2186, "epoch": 24, "lr": 9.999043515885575e-05}
{"train_loss": 1.2321116924285889, "global_step": 2187, "epoch": 24, "lr": 9.999042381638914e-05}
{"train_loss": 1.3440055847167969, "global_step": 2188, "epoch": 24, "lr": 9.999041246720191e-05}
{"train_loss": 1.1933314800262451, "global_step": 2189, "epoch": 24, "lr": 9.999040111129409e-05}
{"train_loss": 1.393789529800415, "global_step": 2190, "epoch": 24, "lr": 9.999038974866566e-05}
{"train_loss": 1.371337890625, "global_step": 2191, "epoch": 24, "lr": 9.999037837931664e-05}
{"train_loss": 1.346691370010376, "global_step": 2192, "epoch": 24, "lr": 9.999036700324702e-05}
{"train_loss": 1.3288686275482178, "global_step": 2193, "epoch": 24, "lr": 9.99903556204568e-05}
{"train_loss": 1.08066725730896, "global_step": 2194, "epoch": 24, "lr": 9.9990344230946e-05}
{"train_loss": 1.4872113466262817, "global_step": 2195, "epoch": 24, "lr": 9.999033283471459e-05}
{"train_loss": 1.247460961341858, "global_step": 2196, "epoch": 24, "lr": 9.999032143176259e-05}
{"train_loss": 1.325772762298584, "global_step": 2197, "epoch": 24, "lr": 9.999031002209001e-05}
{"train_loss": 1.2103469371795654, "global_step": 2198, "epoch": 24, "lr": 9.999029860569685e-05}
{"train_loss": 1.208290934562683, "global_step": 2199, "epoch": 24, "lr": 9.999028718258309e-05}
{"train_loss": 1.1967241764068604, "global_step": 2200, "epoch": 24, "lr": 9.999027575274876e-05}
{"train_loss": 1.1879479885101318, "global_step": 2201, "epoch": 24, "lr": 9.999026431619383e-05}
{"train_loss": 1.3816263675689697, "global_step": 2202, "epoch": 24, "lr": 9.999025287291832e-05}
{"train_loss": 1.3910810947418213, "global_step": 2203, "epoch": 24, "lr": 9.999024142292223e-05}
{"train_loss": 1.3140809535980225, "global_step": 2204, "epoch": 24, "lr": 9.999022996620557e-05}
{"train_loss": 1.440943717956543, "global_step": 2205, "epoch": 24, "lr": 9.999021850276832e-05}
{"train_loss": 1.3592607975006104, "global_step": 2206, "epoch": 24, "lr": 9.999020703261051e-05}
{"train_loss": 1.435593605041504, "global_step": 2207, "epoch": 24, "lr": 9.999019555573211e-05}
{"train_loss": 1.274780511856079, "global_step": 2208, "epoch": 24, "lr": 9.999018407213314e-05}
{"train_loss": 1.4111968278884888, "global_step": 2209, "epoch": 24, "lr": 9.999017258181361e-05}
{"train_loss": 1.2601584196090698, "global_step": 2210, "epoch": 24, "lr": 9.99901610847735e-05}
{"train_loss": 1.3473063707351685, "global_step": 2211, "epoch": 24, "lr": 9.999014958101283e-05}
{"train_loss": 1.3231316804885864, "global_step": 2212, "epoch": 24, "lr": 9.99901380705316e-05}
{"train_loss": 1.1469075679779053, "global_step": 2213, "epoch": 24, "lr": 9.999012655332978e-05}
{"train_loss": 1.330330491065979, "global_step": 2214, "epoch": 24, "lr": 9.999011502940742e-05}
{"train_loss": 1.2384194135665894, "global_step": 2215, "epoch": 24, "lr": 9.99901034987645e-05}
{"train_loss": 1.2593873739242554, "global_step": 2216, "epoch": 24, "lr": 9.999009196140102e-05}
{"train_loss": 1.276596188545227, "global_step": 2217, "epoch": 24, "lr": 9.999008041731695e-05}
{"train_loss": 1.2996710538864136, "global_step": 2218, "epoch": 24, "lr": 9.999006886651236e-05}
{"train_loss": 1.3492708206176758, "global_step": 2219, "epoch": 24, "lr": 9.99900573089872e-05}
{"train_loss": 1.2118046283721924, "global_step": 2220, "epoch": 24, "lr": 9.999004574474147e-05}
{"train_loss": 1.3917529582977295, "global_step": 2221, "epoch": 24, "lr": 9.999003417377522e-05}
{"train_loss": 1.312816858291626, "global_step": 2222, "epoch": 24, "lr": 9.99900225960884e-05}
{"train_loss": 1.3091062307357788, "global_step": 2223, "epoch": 24, "lr": 9.999001101168102e-05}
{"train_loss": 1.3227968430251218, "global_step": 2224, "epoch": 24, "lr": 9.998999942055312e-05, "val_loss": 1.2741066217422485}
{"train_loss": 1.1019841432571411, "global_step": 2225, "epoch": 25, "lr": 9.998998782270465e-05}
{"train_loss": 1.4623794555664062, "global_step": 2226, "epoch": 25, "lr": 9.998997621813565e-05}
{"train_loss": 1.4278137683868408, "global_step": 2227, "epoch": 25, "lr": 9.998996460684611e-05}
{"train_loss": 1.4666824340820312, "global_step": 2228, "epoch": 25, "lr": 9.9989952988836e-05}
{"train_loss": 1.3410542011260986, "global_step": 2229, "epoch": 25, "lr": 9.998994136410538e-05}
{"train_loss": 1.104791283607483, "global_step": 2230, "epoch": 25, "lr": 9.998992973265422e-05}
{"train_loss": 1.3156001567840576, "global_step": 2231, "epoch": 25, "lr": 9.998991809448251e-05}
{"train_loss": 1.3729729652404785, "global_step": 2232, "epoch": 25, "lr": 9.998990644959028e-05}
{"train_loss": 1.496243953704834, "global_step": 2233, "epoch": 25, "lr": 9.99898947979775e-05}
{"train_loss": 1.2767399549484253, "global_step": 2234, "epoch": 25, "lr": 9.998988313964419e-05}
{"train_loss": 1.0860711336135864, "global_step": 2235, "epoch": 25, "lr": 9.998987147459036e-05}
{"train_loss": 1.0972824096679688, "global_step": 2236, "epoch": 25, "lr": 9.998985980281599e-05}
{"train_loss": 1.1666947603225708, "global_step": 2237, "epoch": 25, "lr": 9.99898481243211e-05}
{"train_loss": 1.2947115898132324, "global_step": 2238, "epoch": 25, "lr": 9.998983643910569e-05}
{"train_loss": 1.4214695692062378, "global_step": 2239, "epoch": 25, "lr": 9.998982474716974e-05}
{"train_loss": 1.443579077720642, "global_step": 2240, "epoch": 25, "lr": 9.998981304851327e-05}
{"train_loss": 1.3497687578201294, "global_step": 2241, "epoch": 25, "lr": 9.998980134313629e-05}
{"train_loss": 1.20378577709198, "global_step": 2242, "epoch": 25, "lr": 9.998978963103879e-05}
{"train_loss": 1.417889952659607, "global_step": 2243, "epoch": 25, "lr": 9.998977791222077e-05}
{"train_loss": 1.2941621541976929, "global_step": 2244, "epoch": 25, "lr": 9.998976618668224e-05}
{"train_loss": 1.3601878881454468, "global_step": 2245, "epoch": 25, "lr": 9.998975445442318e-05}
{"train_loss": 1.443094253540039, "global_step": 2246, "epoch": 25, "lr": 9.998974271544361e-05}
{"train_loss": 1.2773776054382324, "global_step": 2247, "epoch": 25, "lr": 9.998973096974354e-05}
{"train_loss": 1.3685824871063232, "global_step": 2248, "epoch": 25, "lr": 9.998971921732294e-05}
{"train_loss": 1.13401198387146, "global_step": 2249, "epoch": 25, "lr": 9.998970745818183e-05}
{"train_loss": 1.4386850595474243, "global_step": 2250, "epoch": 25, "lr": 9.998969569232023e-05}
{"train_loss": 1.2444233894348145, "global_step": 2251, "epoch": 25, "lr": 9.998968391973812e-05}
{"train_loss": 1.4246044158935547, "global_step": 2252, "epoch": 25, "lr": 9.998967214043551e-05}
{"train_loss": 1.3508667945861816, "global_step": 2253, "epoch": 25, "lr": 9.99896603544124e-05}
{"train_loss": 1.3079596757888794, "global_step": 2254, "epoch": 25, "lr": 9.998964856166877e-05}
{"train_loss": 1.528485655784607, "global_step": 2255, "epoch": 25, "lr": 9.998963676220466e-05}
{"train_loss": 1.3779340982437134, "global_step": 2256, "epoch": 25, "lr": 9.998962495602006e-05}
{"train_loss": 1.3173205852508545, "global_step": 2257, "epoch": 25, "lr": 9.998961314311494e-05}
{"train_loss": 1.072346568107605, "global_step": 2258, "epoch": 25, "lr": 9.998960132348934e-05}
{"train_loss": 1.3863033056259155, "global_step": 2259, "epoch": 25, "lr": 9.998958949714326e-05}
{"train_loss": 1.3634891510009766, "global_step": 2260, "epoch": 25, "lr": 9.998957766407666e-05}
{"train_loss": 1.2929795980453491, "global_step": 2261, "epoch": 25, "lr": 9.998956582428958e-05}
{"train_loss": 1.4362937211990356, "global_step": 2262, "epoch": 25, "lr": 9.998955397778203e-05}
{"train_loss": 1.5353655815124512, "global_step": 2263, "epoch": 25, "lr": 9.998954212455398e-05}
{"train_loss": 1.3302308320999146, "global_step": 2264, "epoch": 25, "lr": 9.998953026460545e-05}
{"train_loss": 1.2651530504226685, "global_step": 2265, "epoch": 25, "lr": 9.998951839793642e-05}
{"train_loss": 1.2772088050842285, "global_step": 2266, "epoch": 25, "lr": 9.998950652454692e-05}
{"train_loss": 1.3959362506866455, "global_step": 2267, "epoch": 25, "lr": 9.998949464443695e-05}
{"train_loss": 0.982821524143219, "global_step": 2268, "epoch": 25, "lr": 9.99894827576065e-05}
{"train_loss": 1.1879637241363525, "global_step": 2269, "epoch": 25, "lr": 9.998947086405556e-05}
{"train_loss": 1.3949496746063232, "global_step": 2270, "epoch": 25, "lr": 9.998945896378415e-05}
{"train_loss": 1.5006580352783203, "global_step": 2271, "epoch": 25, "lr": 9.998944705679227e-05}
{"train_loss": 1.1871938705444336, "global_step": 2272, "epoch": 25, "lr": 9.998943514307992e-05}
{"train_loss": 1.1385775804519653, "global_step": 2273, "epoch": 25, "lr": 9.998942322264708e-05}
{"train_loss": 1.4877609014511108, "global_step": 2274, "epoch": 25, "lr": 9.998941129549379e-05}
{"train_loss": 1.237453818321228, "global_step": 2275, "epoch": 25, "lr": 9.998939936162003e-05}
{"train_loss": 1.3856662511825562, "global_step": 2276, "epoch": 25, "lr": 9.998938742102581e-05}
{"train_loss": 1.2180458307266235, "global_step": 2277, "epoch": 25, "lr": 9.998937547371112e-05}
{"train_loss": 1.343786597251892, "global_step": 2278, "epoch": 25, "lr": 9.998936351967597e-05}
{"train_loss": 1.2169626951217651, "global_step": 2279, "epoch": 25, "lr": 9.998935155892036e-05}
{"train_loss": 1.2042648792266846, "global_step": 2280, "epoch": 25, "lr": 9.998933959144429e-05}
{"train_loss": 1.3831398487091064, "global_step": 2281, "epoch": 25, "lr": 9.998932761724775e-05}
{"train_loss": 1.3134033679962158, "global_step": 2282, "epoch": 25, "lr": 9.998931563633077e-05}
{"train_loss": 1.194537878036499, "global_step": 2283, "epoch": 25, "lr": 9.998930364869333e-05}
{"train_loss": 1.090118646621704, "global_step": 2284, "epoch": 25, "lr": 9.998929165433544e-05}
{"train_loss": 1.5463879108428955, "global_step": 2285, "epoch": 25, "lr": 9.99892796532571e-05}
{"train_loss": 1.3292908668518066, "global_step": 2286, "epoch": 25, "lr": 9.998926764545831e-05}
{"train_loss": 1.2379589080810547, "global_step": 2287, "epoch": 25, "lr": 9.998925563093907e-05}
{"train_loss": 1.0694639682769775, "global_step": 2288, "epoch": 25, "lr": 9.99892436096994e-05}
{"train_loss": 1.1616219282150269, "global_step": 2289, "epoch": 25, "lr": 9.998923158173925e-05}
{"train_loss": 1.1700855493545532, "global_step": 2290, "epoch": 25, "lr": 9.998921954705868e-05}
{"train_loss": 1.399376392364502, "global_step": 2291, "epoch": 25, "lr": 9.998920750565767e-05}
{"train_loss": 1.3187378644943237, "global_step": 2292, "epoch": 25, "lr": 9.998919545753622e-05}
{"train_loss": 1.311542272567749, "global_step": 2293, "epoch": 25, "lr": 9.998918340269433e-05}
{"train_loss": 1.1887543201446533, "global_step": 2294, "epoch": 25, "lr": 9.998917134113199e-05}
{"train_loss": 1.1849219799041748, "global_step": 2295, "epoch": 25, "lr": 9.998915927284923e-05}
{"train_loss": 1.2580726146697998, "global_step": 2296, "epoch": 25, "lr": 9.998914719784603e-05}
{"train_loss": 1.2568094730377197, "global_step": 2297, "epoch": 25, "lr": 9.99891351161224e-05}
{"train_loss": 1.3914308547973633, "global_step": 2298, "epoch": 25, "lr": 9.998912302767836e-05}
{"train_loss": 1.2342654466629028, "global_step": 2299, "epoch": 25, "lr": 9.998911093251386e-05}
{"train_loss": 1.3955316543579102, "global_step": 2300, "epoch": 25, "lr": 9.998909883062895e-05}
{"train_loss": 1.214789628982544, "global_step": 2301, "epoch": 25, "lr": 9.998908672202361e-05}
{"train_loss": 1.422560691833496, "global_step": 2302, "epoch": 25, "lr": 9.998907460669785e-05}
{"train_loss": 1.2262823581695557, "global_step": 2303, "epoch": 25, "lr": 9.998906248465167e-05}
{"train_loss": 1.2976248264312744, "global_step": 2304, "epoch": 25, "lr": 9.998905035588506e-05}
{"train_loss": 1.4070510864257812, "global_step": 2305, "epoch": 25, "lr": 9.998903822039805e-05}
{"train_loss": 1.5185633897781372, "global_step": 2306, "epoch": 25, "lr": 9.998902607819061e-05}
{"train_loss": 1.3422636985778809, "global_step": 2307, "epoch": 25, "lr": 9.998901392926275e-05}
{"train_loss": 1.088071346282959, "global_step": 2308, "epoch": 25, "lr": 9.99890017736145e-05}
{"train_loss": 1.3293379545211792, "global_step": 2309, "epoch": 25, "lr": 9.99889896112458e-05}
{"train_loss": 1.1464959383010864, "global_step": 2310, "epoch": 25, "lr": 9.998897744215672e-05}
{"train_loss": 1.272152066230774, "global_step": 2311, "epoch": 25, "lr": 9.998896526634722e-05}
{"train_loss": 1.277301549911499, "global_step": 2312, "epoch": 25, "lr": 9.99889530838173e-05}
{"train_loss": 1.3050511412406236, "global_step": 2313, "epoch": 25, "lr": 9.9988940894567e-05, "val_loss": 1.3262673616409302, "train_action_mse_error": 73.0625228881836}
{"train_loss": 1.3310071229934692, "global_step": 2314, "epoch": 26, "lr": 9.998892869859629e-05}
{"train_loss": 1.3349765539169312, "global_step": 2315, "epoch": 26, "lr": 9.998891649590517e-05}
{"train_loss": 1.3831229209899902, "global_step": 2316, "epoch": 26, "lr": 9.998890428649366e-05}
{"train_loss": 1.4120434522628784, "global_step": 2317, "epoch": 26, "lr": 9.998889207036173e-05}
{"train_loss": 1.1509767770767212, "global_step": 2318, "epoch": 26, "lr": 9.998887984750942e-05}
{"train_loss": 1.3838406801223755, "global_step": 2319, "epoch": 26, "lr": 9.998886761793672e-05}
{"train_loss": 1.118526816368103, "global_step": 2320, "epoch": 26, "lr": 9.998885538164361e-05}
{"train_loss": 1.4030039310455322, "global_step": 2321, "epoch": 26, "lr": 9.99888431386301e-05}
{"train_loss": 1.2562953233718872, "global_step": 2322, "epoch": 26, "lr": 9.998883088889623e-05}
{"train_loss": 1.5113829374313354, "global_step": 2323, "epoch": 26, "lr": 9.998881863244195e-05}
{"train_loss": 1.475402593612671, "global_step": 2324, "epoch": 26, "lr": 9.998880636926729e-05}
{"train_loss": 1.2617523670196533, "global_step": 2325, "epoch": 26, "lr": 9.998879409937224e-05}
{"train_loss": 1.4784215688705444, "global_step": 2326, "epoch": 26, "lr": 9.998878182275681e-05}
{"train_loss": 1.180230736732483, "global_step": 2327, "epoch": 26, "lr": 9.998876953942099e-05}
{"train_loss": 1.278939962387085, "global_step": 2328, "epoch": 26, "lr": 9.99887572493648e-05}
{"train_loss": 1.2487637996673584, "global_step": 2329, "epoch": 26, "lr": 9.998874495258822e-05}
{"train_loss": 1.255082130432129, "global_step": 2330, "epoch": 26, "lr": 9.998873264909126e-05}
{"train_loss": 1.2611730098724365, "global_step": 2331, "epoch": 26, "lr": 9.998872033887394e-05}
{"train_loss": 1.0795401334762573, "global_step": 2332, "epoch": 26, "lr": 9.998870802193622e-05}
{"train_loss": 1.1617228984832764, "global_step": 2333, "epoch": 26, "lr": 9.998869569827815e-05}
{"train_loss": 1.3411998748779297, "global_step": 2334, "epoch": 26, "lr": 9.998868336789972e-05}
{"train_loss": 1.2589447498321533, "global_step": 2335, "epoch": 26, "lr": 9.998867103080089e-05}
{"train_loss": 1.2736952304840088, "global_step": 2336, "epoch": 26, "lr": 9.998865868698171e-05}
{"train_loss": 1.283342957496643, "global_step": 2337, "epoch": 26, "lr": 9.998864633644218e-05}
{"train_loss": 1.5703197717666626, "global_step": 2338, "epoch": 26, "lr": 9.998863397918225e-05}
{"train_loss": 1.129176378250122, "global_step": 2339, "epoch": 26, "lr": 9.998862161520199e-05}
{"train_loss": 1.2160367965698242, "global_step": 2340, "epoch": 26, "lr": 9.998860924450136e-05}
{"train_loss": 1.1854221820831299, "global_step": 2341, "epoch": 26, "lr": 9.998859686708036e-05}
{"train_loss": 1.2522517442703247, "global_step": 2342, "epoch": 26, "lr": 9.9988584482939e-05}
{"train_loss": 1.3911410570144653, "global_step": 2343, "epoch": 26, "lr": 9.998857209207728e-05}
{"train_loss": 1.2431902885437012, "global_step": 2344, "epoch": 26, "lr": 9.998855969449524e-05}
{"train_loss": 1.3400886058807373, "global_step": 2345, "epoch": 26, "lr": 9.99885472901928e-05}
{"train_loss": 1.3601772785186768, "global_step": 2346, "epoch": 26, "lr": 9.998853487917004e-05}
{"train_loss": 1.140535593032837, "global_step": 2347, "epoch": 26, "lr": 9.998852246142692e-05}
{"train_loss": 1.36344313621521, "global_step": 2348, "epoch": 26, "lr": 9.998851003696346e-05}
{"train_loss": 1.661853313446045, "global_step": 2349, "epoch": 26, "lr": 9.998849760577965e-05}
{"train_loss": 1.2845760583877563, "global_step": 2350, "epoch": 26, "lr": 9.99884851678755e-05}
{"train_loss": 1.238579511642456, "global_step": 2351, "epoch": 26, "lr": 9.9988472723251e-05}
{"train_loss": 1.402659296989441, "global_step": 2352, "epoch": 26, "lr": 9.998846027190617e-05}
{"train_loss": 1.2524433135986328, "global_step": 2353, "epoch": 26, "lr": 9.998844781384099e-05}
{"train_loss": 1.1203253269195557, "global_step": 2354, "epoch": 26, "lr": 9.998843534905548e-05}
{"train_loss": 1.3866890668869019, "global_step": 2355, "epoch": 26, "lr": 9.998842287754962e-05}
{"train_loss": 1.1936779022216797, "global_step": 2356, "epoch": 26, "lr": 9.998841039932344e-05}
{"train_loss": 1.3615939617156982, "global_step": 2357, "epoch": 26, "lr": 9.998839791437694e-05}
{"train_loss": 1.3455570936203003, "global_step": 2358, "epoch": 26, "lr": 9.998838542271007e-05}
{"train_loss": 1.2667709589004517, "global_step": 2359, "epoch": 26, "lr": 9.99883729243229e-05}
{"train_loss": 1.403601884841919, "global_step": 2360, "epoch": 26, "lr": 9.998836041921539e-05}
{"train_loss": 1.3046915531158447, "global_step": 2361, "epoch": 26, "lr": 9.998834790738757e-05}
{"train_loss": 1.1114561557769775, "global_step": 2362, "epoch": 26, "lr": 9.998833538883941e-05}
{"train_loss": 1.299418568611145, "global_step": 2363, "epoch": 26, "lr": 9.998832286357094e-05}
{"train_loss": 1.1445324420928955, "global_step": 2364, "epoch": 26, "lr": 9.998831033158213e-05}
{"train_loss": 1.1744288206100464, "global_step": 2365, "epoch": 26, "lr": 9.998829779287302e-05}
{"train_loss": 1.4349786043167114, "global_step": 2366, "epoch": 26, "lr": 9.998828524744359e-05}
{"train_loss": 1.4002437591552734, "global_step": 2367, "epoch": 26, "lr": 9.998827269529386e-05}
{"train_loss": 1.2608110904693604, "global_step": 2368, "epoch": 26, "lr": 9.998826013642379e-05}
{"train_loss": 1.1613941192626953, "global_step": 2369, "epoch": 26, "lr": 9.99882475708334e-05}
{"train_loss": 1.0627162456512451, "global_step": 2370, "epoch": 26, "lr": 9.998823499852272e-05}
{"train_loss": 1.1666148900985718, "global_step": 2371, "epoch": 26, "lr": 9.998822241949172e-05}
{"train_loss": 1.445141077041626, "global_step": 2372, "epoch": 26, "lr": 9.998820983374042e-05}
{"train_loss": 1.182832956314087, "global_step": 2373, "epoch": 26, "lr": 9.998819724126881e-05}
{"train_loss": 1.3709633350372314, "global_step": 2374, "epoch": 26, "lr": 9.998818464207692e-05}
{"train_loss": 1.3444410562515259, "global_step": 2375, "epoch": 26, "lr": 9.998817203616468e-05}
{"train_loss": 1.1779139041900635, "global_step": 2376, "epoch": 26, "lr": 9.998815942353217e-05}
{"train_loss": 1.146087408065796, "global_step": 2377, "epoch": 26, "lr": 9.998814680417936e-05}
{"train_loss": 1.188723087310791, "global_step": 2378, "epoch": 26, "lr": 9.998813417810625e-05}
{"train_loss": 1.0281758308410645, "global_step": 2379, "epoch": 26, "lr": 9.998812154531286e-05}
{"train_loss": 1.2116223573684692, "global_step": 2380, "epoch": 26, "lr": 9.998810890579915e-05}
{"train_loss": 1.2873930931091309, "global_step": 2381, "epoch": 26, "lr": 9.998809625956516e-05}
{"train_loss": 1.3634262084960938, "global_step": 2382, "epoch": 26, "lr": 9.998808360661087e-05}
{"train_loss": 1.297642469406128, "global_step": 2383, "epoch": 26, "lr": 9.99880709469363e-05}
{"train_loss": 1.3256655931472778, "global_step": 2384, "epoch": 26, "lr": 9.998805828054145e-05}
{"train_loss": 1.233872652053833, "global_step": 2385, "epoch": 26, "lr": 9.998804560742632e-05}
{"train_loss": 1.2836990356445312, "global_step": 2386, "epoch": 26, "lr": 9.998803292759089e-05}
{"train_loss": 1.3887856006622314, "global_step": 2387, "epoch": 26, "lr": 9.998802024103517e-05}
{"train_loss": 1.294116497039795, "global_step": 2388, "epoch": 26, "lr": 9.998800754775919e-05}
{"train_loss": 1.312827467918396, "global_step": 2389, "epoch": 26, "lr": 9.998799484776293e-05}
{"train_loss": 1.1000676155090332, "global_step": 2390, "epoch": 26, "lr": 9.998798214104639e-05}
{"train_loss": 1.203771948814392, "global_step": 2391, "epoch": 26, "lr": 9.998796942760957e-05}
{"train_loss": 1.1979268789291382, "global_step": 2392, "epoch": 26, "lr": 9.998795670745248e-05}
{"train_loss": 1.0327200889587402, "global_step": 2393, "epoch": 26, "lr": 9.998794398057512e-05}
{"train_loss": 1.1017460823059082, "global_step": 2394, "epoch": 26, "lr": 9.998793124697748e-05}
{"train_loss": 1.3802807331085205, "global_step": 2395, "epoch": 26, "lr": 9.998791850665958e-05}
{"train_loss": 1.2159796953201294, "global_step": 2396, "epoch": 26, "lr": 9.998790575962143e-05}
{"train_loss": 1.0286941528320312, "global_step": 2397, "epoch": 26, "lr": 9.998789300586299e-05}
{"train_loss": 1.3296080827713013, "global_step": 2398, "epoch": 26, "lr": 9.99878802453843e-05}
{"train_loss": 1.3101774454116821, "global_step": 2399, "epoch": 26, "lr": 9.998786747818534e-05}
{"train_loss": 1.2716962099075317, "global_step": 2400, "epoch": 26, "lr": 9.998785470426613e-05}
{"train_loss": 1.192191481590271, "global_step": 2401, "epoch": 26, "lr": 9.998784192362664e-05}
{"train_loss": 1.2722230091523588, "global_step": 2402, "epoch": 26, "lr": 9.998782913626692e-05, "val_loss": 1.2392654418945312}
{"train_loss": 1.3972722291946411, "global_step": 2403, "epoch": 27, "lr": 9.998781634218694e-05}
{"train_loss": 1.133418083190918, "global_step": 2404, "epoch": 27, "lr": 9.99878035413867e-05}
{"train_loss": 1.2326987981796265, "global_step": 2405, "epoch": 27, "lr": 9.998779073386622e-05}
{"train_loss": 1.1904921531677246, "global_step": 2406, "epoch": 27, "lr": 9.998777791962548e-05}
{"train_loss": 1.3273667097091675, "global_step": 2407, "epoch": 27, "lr": 9.99877650986645e-05}
{"train_loss": 1.2915372848510742, "global_step": 2408, "epoch": 27, "lr": 9.998775227098328e-05}
{"train_loss": 1.1748074293136597, "global_step": 2409, "epoch": 27, "lr": 9.998773943658179e-05}
{"train_loss": 1.1953610181808472, "global_step": 2410, "epoch": 27, "lr": 9.998772659546009e-05}
{"train_loss": 1.3591333627700806, "global_step": 2411, "epoch": 27, "lr": 9.998771374761813e-05}
{"train_loss": 1.245303988456726, "global_step": 2412, "epoch": 27, "lr": 9.998770089305593e-05}
{"train_loss": 1.175526738166809, "global_step": 2413, "epoch": 27, "lr": 9.998768803177351e-05}
{"train_loss": 1.1031516790390015, "global_step": 2414, "epoch": 27, "lr": 9.998767516377083e-05}
{"train_loss": 1.2770421504974365, "global_step": 2415, "epoch": 27, "lr": 9.998766228904795e-05}
{"train_loss": 1.1958779096603394, "global_step": 2416, "epoch": 27, "lr": 9.998764940760482e-05}
{"train_loss": 1.1949405670166016, "global_step": 2417, "epoch": 27, "lr": 9.998763651944145e-05}
{"train_loss": 1.1706418991088867, "global_step": 2418, "epoch": 27, "lr": 9.998762362455784e-05}
{"train_loss": 1.0553622245788574, "global_step": 2419, "epoch": 27, "lr": 9.998761072295404e-05}
{"train_loss": 1.0190479755401611, "global_step": 2420, "epoch": 27, "lr": 9.998759781462998e-05}
{"train_loss": 1.1721450090408325, "global_step": 2421, "epoch": 27, "lr": 9.998758489958573e-05}
{"train_loss": 1.3278443813323975, "global_step": 2422, "epoch": 27, "lr": 9.998757197782124e-05}
{"train_loss": 1.4221816062927246, "global_step": 2423, "epoch": 27, "lr": 9.998755904933653e-05}
{"train_loss": 1.0813754796981812, "global_step": 2424, "epoch": 27, "lr": 9.998754611413161e-05}
{"train_loss": 1.288170337677002, "global_step": 2425, "epoch": 27, "lr": 9.998753317220646e-05}
{"train_loss": 1.1300863027572632, "global_step": 2426, "epoch": 27, "lr": 9.99875202235611e-05}
{"train_loss": 1.0899876356124878, "global_step": 2427, "epoch": 27, "lr": 9.998750726819555e-05}
{"train_loss": 1.3317217826843262, "global_step": 2428, "epoch": 27, "lr": 9.998749430610976e-05}
{"train_loss": 1.189704179763794, "global_step": 2429, "epoch": 27, "lr": 9.998748133730377e-05}
{"train_loss": 1.3509193658828735, "global_step": 2430, "epoch": 27, "lr": 9.998746836177757e-05}
{"train_loss": 1.0792286396026611, "global_step": 2431, "epoch": 27, "lr": 9.998745537953117e-05}
{"train_loss": 1.2660354375839233, "global_step": 2432, "epoch": 27, "lr": 9.998744239056457e-05}
{"train_loss": 1.3327455520629883, "global_step": 2433, "epoch": 27, "lr": 9.998742939487776e-05}
{"train_loss": 1.4624648094177246, "global_step": 2434, "epoch": 27, "lr": 9.998741639247075e-05}
{"train_loss": 1.2360070943832397, "global_step": 2435, "epoch": 27, "lr": 9.998740338334354e-05}
{"train_loss": 1.2889348268508911, "global_step": 2436, "epoch": 27, "lr": 9.998739036749613e-05}
{"train_loss": 1.0545233488082886, "global_step": 2437, "epoch": 27, "lr": 9.998737734492853e-05}
{"train_loss": 1.2055251598358154, "global_step": 2438, "epoch": 27, "lr": 9.998736431564073e-05}
{"train_loss": 1.1803505420684814, "global_step": 2439, "epoch": 27, "lr": 9.998735127963275e-05}
{"train_loss": 1.1140893697738647, "global_step": 2440, "epoch": 27, "lr": 9.998733823690457e-05}
{"train_loss": 1.2521328926086426, "global_step": 2441, "epoch": 27, "lr": 9.998732518745619e-05}
{"train_loss": 1.0270144939422607, "global_step": 2442, "epoch": 27, "lr": 9.998731213128764e-05}
{"train_loss": 1.2317932844161987, "global_step": 2443, "epoch": 27, "lr": 9.998729906839891e-05}
{"train_loss": 1.2657499313354492, "global_step": 2444, "epoch": 27, "lr": 9.998728599878999e-05}
{"train_loss": 1.3461288213729858, "global_step": 2445, "epoch": 27, "lr": 9.998727292246088e-05}
{"train_loss": 1.3360000848770142, "global_step": 2446, "epoch": 27, "lr": 9.998725983941161e-05}
{"train_loss": 1.3309520483016968, "global_step": 2447, "epoch": 27, "lr": 9.998724674964215e-05}
{"train_loss": 1.1508584022521973, "global_step": 2448, "epoch": 27, "lr": 9.998723365315252e-05}
{"train_loss": 1.4048244953155518, "global_step": 2449, "epoch": 27, "lr": 9.99872205499427e-05}
{"train_loss": 1.2798330783843994, "global_step": 2450, "epoch": 27, "lr": 9.998720744001273e-05}
{"train_loss": 1.1072689294815063, "global_step": 2451, "epoch": 27, "lr": 9.998719432336257e-05}
{"train_loss": 1.3384952545166016, "global_step": 2452, "epoch": 27, "lr": 9.998718119999226e-05}
{"train_loss": 1.2771220207214355, "global_step": 2453, "epoch": 27, "lr": 9.998716806990176e-05}
{"train_loss": 1.300583839416504, "global_step": 2454, "epoch": 27, "lr": 9.998715493309111e-05}
{"train_loss": 1.2147936820983887, "global_step": 2455, "epoch": 27, "lr": 9.99871417895603e-05}
{"train_loss": 1.2766332626342773, "global_step": 2456, "epoch": 27, "lr": 9.998712863930932e-05}
{"train_loss": 1.244065284729004, "global_step": 2457, "epoch": 27, "lr": 9.998711548233817e-05}
{"train_loss": 1.3326082229614258, "global_step": 2458, "epoch": 27, "lr": 9.998710231864687e-05}
{"train_loss": 1.3196176290512085, "global_step": 2459, "epoch": 27, "lr": 9.998708914823543e-05}
{"train_loss": 1.1131055355072021, "global_step": 2460, "epoch": 27, "lr": 9.99870759711038e-05}
{"train_loss": 1.0809142589569092, "global_step": 2461, "epoch": 27, "lr": 9.998706278725204e-05}
{"train_loss": 1.3499248027801514, "global_step": 2462, "epoch": 27, "lr": 9.998704959668013e-05}
{"train_loss": 1.2392486333847046, "global_step": 2463, "epoch": 27, "lr": 9.998703639938806e-05}
{"train_loss": 1.348008632659912, "global_step": 2464, "epoch": 27, "lr": 9.998702319537585e-05}
{"train_loss": 1.4176182746887207, "global_step": 2465, "epoch": 27, "lr": 9.99870099846435e-05}
{"train_loss": 1.2360163927078247, "global_step": 2466, "epoch": 27, "lr": 9.998699676719099e-05}
{"train_loss": 1.256371021270752, "global_step": 2467, "epoch": 27, "lr": 9.998698354301834e-05}
{"train_loss": 1.1780624389648438, "global_step": 2468, "epoch": 27, "lr": 9.998697031212556e-05}
{"train_loss": 1.0229815244674683, "global_step": 2469, "epoch": 27, "lr": 9.998695707451265e-05}
{"train_loss": 1.1788779497146606, "global_step": 2470, "epoch": 27, "lr": 9.998694383017958e-05}
{"train_loss": 1.4732719659805298, "global_step": 2471, "epoch": 27, "lr": 9.998693057912639e-05}
{"train_loss": 1.3257189989089966, "global_step": 2472, "epoch": 27, "lr": 9.998691732135306e-05}
{"train_loss": 1.286749005317688, "global_step": 2473, "epoch": 27, "lr": 9.998690405685958e-05}
{"train_loss": 1.2675634622573853, "global_step": 2474, "epoch": 27, "lr": 9.9986890785646e-05}
{"train_loss": 1.309356927871704, "global_step": 2475, "epoch": 27, "lr": 9.998687750771227e-05}
{"train_loss": 1.0649378299713135, "global_step": 2476, "epoch": 27, "lr": 9.998686422305843e-05}
{"train_loss": 1.024348258972168, "global_step": 2477, "epoch": 27, "lr": 9.998685093168446e-05}
{"train_loss": 1.2803348302841187, "global_step": 2478, "epoch": 27, "lr": 9.998683763359035e-05}
{"train_loss": 1.263253927230835, "global_step": 2479, "epoch": 27, "lr": 9.998682432877615e-05}
{"train_loss": 1.2095617055892944, "global_step": 2480, "epoch": 27, "lr": 9.998681101724181e-05}
{"train_loss": 1.3431603908538818, "global_step": 2481, "epoch": 27, "lr": 9.998679769898736e-05}
{"train_loss": 1.241964340209961, "global_step": 2482, "epoch": 27, "lr": 9.99867843740128e-05}
{"train_loss": 1.2612439393997192, "global_step": 2483, "epoch": 27, "lr": 9.998677104231812e-05}
{"train_loss": 1.262980341911316, "global_step": 2484, "epoch": 27, "lr": 9.998675770390332e-05}
{"train_loss": 1.2862844467163086, "global_step": 2485, "epoch": 27, "lr": 9.998674435876843e-05}
{"train_loss": 1.0589520931243896, "global_step": 2486, "epoch": 27, "lr": 9.998673100691341e-05}
{"train_loss": 1.1748862266540527, "global_step": 2487, "epoch": 27, "lr": 9.998671764833829e-05}
{"train_loss": 1.2631139755249023, "global_step": 2488, "epoch": 27, "lr": 9.998670428304306e-05}
{"train_loss": 1.380066156387329, "global_step": 2489, "epoch": 27, "lr": 9.998669091102773e-05}
{"train_loss": 1.5690202713012695, "global_step": 2490, "epoch": 27, "lr": 9.99866775322923e-05}
{"train_loss": 1.237113155675738, "global_step": 2491, "epoch": 27, "lr": 9.998666414683677e-05, "val_loss": 1.2516324520111084}
{"train_loss": 1.3717013597488403, "global_step": 2492, "epoch": 28, "lr": 9.998665075466116e-05}
{"train_loss": 1.1886850595474243, "global_step": 2493, "epoch": 28, "lr": 9.998663735576543e-05}
{"train_loss": 1.327882170677185, "global_step": 2494, "epoch": 28, "lr": 9.998662395014962e-05}
{"train_loss": 1.3072932958602905, "global_step": 2495, "epoch": 28, "lr": 9.998661053781371e-05}
{"train_loss": 1.3396481275558472, "global_step": 2496, "epoch": 28, "lr": 9.99865971187577e-05}
{"train_loss": 1.0427402257919312, "global_step": 2497, "epoch": 28, "lr": 9.998658369298162e-05}
{"train_loss": 1.1975284814834595, "global_step": 2498, "epoch": 28, "lr": 9.998657026048544e-05}
{"train_loss": 1.3459981679916382, "global_step": 2499, "epoch": 28, "lr": 9.998655682126919e-05}
{"train_loss": 1.345033049583435, "global_step": 2500, "epoch": 28, "lr": 9.998654337533285e-05}
{"train_loss": 1.0875412225723267, "global_step": 2501, "epoch": 28, "lr": 9.998652992267642e-05}
{"train_loss": 1.3772927522659302, "global_step": 2502, "epoch": 28, "lr": 9.99865164632999e-05}
{"train_loss": 1.2949458360671997, "global_step": 2503, "epoch": 28, "lr": 9.998650299720334e-05}
{"train_loss": 1.3373969793319702, "global_step": 2504, "epoch": 28, "lr": 9.998648952438667e-05}
{"train_loss": 1.3716881275177002, "global_step": 2505, "epoch": 28, "lr": 9.998647604484994e-05}
{"train_loss": 1.1535574197769165, "global_step": 2506, "epoch": 28, "lr": 9.998646255859314e-05}
{"train_loss": 1.2980492115020752, "global_step": 2507, "epoch": 28, "lr": 9.998644906561626e-05}
{"train_loss": 1.2153412103652954, "global_step": 2508, "epoch": 28, "lr": 9.998643556591931e-05}
{"train_loss": 1.1409344673156738, "global_step": 2509, "epoch": 28, "lr": 9.99864220595023e-05}
{"train_loss": 1.1901496648788452, "global_step": 2510, "epoch": 28, "lr": 9.998640854636522e-05}
{"train_loss": 1.0832122564315796, "global_step": 2511, "epoch": 28, "lr": 9.998639502650809e-05}
{"train_loss": 1.1461313962936401, "global_step": 2512, "epoch": 28, "lr": 9.998638149993089e-05}
{"train_loss": 1.2016397714614868, "global_step": 2513, "epoch": 28, "lr": 9.998636796663363e-05}
{"train_loss": 1.2634066343307495, "global_step": 2514, "epoch": 28, "lr": 9.99863544266163e-05}
{"train_loss": 1.211421251296997, "global_step": 2515, "epoch": 28, "lr": 9.998634087987893e-05}
{"train_loss": 1.1133294105529785, "global_step": 2516, "epoch": 28, "lr": 9.99863273264215e-05}
{"train_loss": 1.3493517637252808, "global_step": 2517, "epoch": 28, "lr": 9.998631376624403e-05}
{"train_loss": 1.1871498823165894, "global_step": 2518, "epoch": 28, "lr": 9.998630019934649e-05}
{"train_loss": 1.314070463180542, "global_step": 2519, "epoch": 28, "lr": 9.99862866257289e-05}
{"train_loss": 1.2020469903945923, "global_step": 2520, "epoch": 28, "lr": 9.998627304539127e-05}
{"train_loss": 1.1449241638183594, "global_step": 2521, "epoch": 28, "lr": 9.99862594583336e-05}
{"train_loss": 1.304401159286499, "global_step": 2522, "epoch": 28, "lr": 9.998624586455589e-05}
{"train_loss": 1.2481980323791504, "global_step": 2523, "epoch": 28, "lr": 9.998623226405813e-05}
{"train_loss": 1.1717162132263184, "global_step": 2524, "epoch": 28, "lr": 9.998621865684033e-05}
{"train_loss": 1.2317899465560913, "global_step": 2525, "epoch": 28, "lr": 9.99862050429025e-05}
{"train_loss": 1.3195959329605103, "global_step": 2526, "epoch": 28, "lr": 9.998619142224462e-05}
{"train_loss": 1.2685456275939941, "global_step": 2527, "epoch": 28, "lr": 9.998617779486672e-05}
{"train_loss": 1.3011305332183838, "global_step": 2528, "epoch": 28, "lr": 9.998616416076877e-05}
{"train_loss": 1.254562258720398, "global_step": 2529, "epoch": 28, "lr": 9.998615051995082e-05}
{"train_loss": 1.5151662826538086, "global_step": 2530, "epoch": 28, "lr": 9.99861368724128e-05}
{"train_loss": 1.1831432580947876, "global_step": 2531, "epoch": 28, "lr": 9.99861232181548e-05}
{"train_loss": 1.1322251558303833, "global_step": 2532, "epoch": 28, "lr": 9.998610955717673e-05}
{"train_loss": 1.3277573585510254, "global_step": 2533, "epoch": 28, "lr": 9.998609588947866e-05}
{"train_loss": 1.1444456577301025, "global_step": 2534, "epoch": 28, "lr": 9.998608221506057e-05}
{"train_loss": 1.1589491367340088, "global_step": 2535, "epoch": 28, "lr": 9.998606853392245e-05}
{"train_loss": 1.015936017036438, "global_step": 2536, "epoch": 28, "lr": 9.998605484606432e-05}
{"train_loss": 1.2065832614898682, "global_step": 2537, "epoch": 28, "lr": 9.998604115148618e-05}
{"train_loss": 1.34974205493927, "global_step": 2538, "epoch": 28, "lr": 9.998602745018801e-05}
{"train_loss": 1.3172823190689087, "global_step": 2539, "epoch": 28, "lr": 9.998601374216985e-05}
{"train_loss": 1.1822227239608765, "global_step": 2540, "epoch": 28, "lr": 9.998600002743166e-05}
{"train_loss": 1.1488460302352905, "global_step": 2541, "epoch": 28, "lr": 9.998598630597347e-05}
{"train_loss": 1.0754903554916382, "global_step": 2542, "epoch": 28, "lr": 9.998597257779528e-05}
{"train_loss": 1.2594482898712158, "global_step": 2543, "epoch": 28, "lr": 9.998595884289706e-05}
{"train_loss": 1.08074152469635, "global_step": 2544, "epoch": 28, "lr": 9.998594510127885e-05}
{"train_loss": 1.1686320304870605, "global_step": 2545, "epoch": 28, "lr": 9.998593135294065e-05}
{"train_loss": 1.2554075717926025, "global_step": 2546, "epoch": 28, "lr": 9.998591759788246e-05}
{"train_loss": 1.203991174697876, "global_step": 2547, "epoch": 28, "lr": 9.998590383610424e-05}
{"train_loss": 1.4890022277832031, "global_step": 2548, "epoch": 28, "lr": 9.998589006760603e-05}
{"train_loss": 1.1509815454483032, "global_step": 2549, "epoch": 28, "lr": 9.998587629238785e-05}
{"train_loss": 1.163529396057129, "global_step": 2550, "epoch": 28, "lr": 9.998586251044966e-05}
{"train_loss": 1.1174824237823486, "global_step": 2551, "epoch": 28, "lr": 9.998584872179148e-05}
{"train_loss": 1.2669469118118286, "global_step": 2552, "epoch": 28, "lr": 9.998583492641332e-05}
{"train_loss": 0.9443858861923218, "global_step": 2553, "epoch": 28, "lr": 9.998582112431519e-05}
{"train_loss": 1.1853405237197876, "global_step": 2554, "epoch": 28, "lr": 9.998580731549705e-05}
{"train_loss": 1.1899831295013428, "global_step": 2555, "epoch": 28, "lr": 9.998579349995892e-05}
{"train_loss": 1.1750041246414185, "global_step": 2556, "epoch": 28, "lr": 9.998577967770083e-05}
{"train_loss": 1.2277932167053223, "global_step": 2557, "epoch": 28, "lr": 9.998576584872275e-05}
{"train_loss": 1.3161749839782715, "global_step": 2558, "epoch": 28, "lr": 9.99857520130247e-05}
{"train_loss": 1.2572046518325806, "global_step": 2559, "epoch": 28, "lr": 9.998573817060668e-05}
{"train_loss": 1.243848443031311, "global_step": 2560, "epoch": 28, "lr": 9.998572432146866e-05}
{"train_loss": 1.1670690774917603, "global_step": 2561, "epoch": 28, "lr": 9.99857104656107e-05}
{"train_loss": 1.400767207145691, "global_step": 2562, "epoch": 28, "lr": 9.998569660303277e-05}
{"train_loss": 1.066144585609436, "global_step": 2563, "epoch": 28, "lr": 9.998568273373485e-05}
{"train_loss": 1.1316502094268799, "global_step": 2564, "epoch": 28, "lr": 9.998566885771696e-05}
{"train_loss": 1.085188627243042, "global_step": 2565, "epoch": 28, "lr": 9.998565497497913e-05}
{"train_loss": 0.9643771648406982, "global_step": 2566, "epoch": 28, "lr": 9.998564108552132e-05}
{"train_loss": 1.3098849058151245, "global_step": 2567, "epoch": 28, "lr": 9.998562718934356e-05}
{"train_loss": 1.3083263635635376, "global_step": 2568, "epoch": 28, "lr": 9.998561328644585e-05}
{"train_loss": 1.3849910497665405, "global_step": 2569, "epoch": 28, "lr": 9.998559937682817e-05}
{"train_loss": 1.2170419692993164, "global_step": 2570, "epoch": 28, "lr": 9.998558546049054e-05}
{"train_loss": 1.1793712377548218, "global_step": 2571, "epoch": 28, "lr": 9.998557153743295e-05}
{"train_loss": 1.3710317611694336, "global_step": 2572, "epoch": 28, "lr": 9.998555760765543e-05}
{"train_loss": 1.2297754287719727, "global_step": 2573, "epoch": 28, "lr": 9.998554367115794e-05}
{"train_loss": 1.0914771556854248, "global_step": 2574, "epoch": 28, "lr": 9.998552972794051e-05}
{"train_loss": 1.2432887554168701, "global_step": 2575, "epoch": 28, "lr": 9.998551577800312e-05}
{"train_loss": 1.1949052810668945, "global_step": 2576, "epoch": 28, "lr": 9.99855018213458e-05}
{"train_loss": 1.2420077323913574, "global_step": 2577, "epoch": 28, "lr": 9.998548785796855e-05}
{"train_loss": 1.1257530450820923, "global_step": 2578, "epoch": 28, "lr": 9.998547388787135e-05}
{"train_loss": 0.9646136164665222, "global_step": 2579, "epoch": 28, "lr": 9.998545991105421e-05}
{"train_loss": 1.2188405300794023, "global_step": 2580, "epoch": 28, "lr": 9.998544592751713e-05, "val_loss": 1.2345727682113647}
{"train_loss": 1.1701627969741821, "global_step": 2581, "epoch": 29, "lr": 9.998543193726013e-05}
{"train_loss": 1.2007737159729004, "global_step": 2582, "epoch": 29, "lr": 9.998541794028318e-05}
{"train_loss": 1.1438076496124268, "global_step": 2583, "epoch": 29, "lr": 9.998540393658631e-05}
{"train_loss": 1.310431718826294, "global_step": 2584, "epoch": 29, "lr": 9.998538992616953e-05}
{"train_loss": 1.4130733013153076, "global_step": 2585, "epoch": 29, "lr": 9.998537590903278e-05}
{"train_loss": 1.2840536832809448, "global_step": 2586, "epoch": 29, "lr": 9.998536188517615e-05}
{"train_loss": 1.2107765674591064, "global_step": 2587, "epoch": 29, "lr": 9.998534785459956e-05}
{"train_loss": 1.2243627309799194, "global_step": 2588, "epoch": 29, "lr": 9.998533381730309e-05}
{"train_loss": 1.428000569343567, "global_step": 2589, "epoch": 29, "lr": 9.998531977328667e-05}
{"train_loss": 1.2412099838256836, "global_step": 2590, "epoch": 29, "lr": 9.998530572255034e-05}
{"train_loss": 1.2672494649887085, "global_step": 2591, "epoch": 29, "lr": 9.99852916650941e-05}
{"train_loss": 1.2213739156723022, "global_step": 2592, "epoch": 29, "lr": 9.998527760091793e-05}
{"train_loss": 1.190495252609253, "global_step": 2593, "epoch": 29, "lr": 9.998526353002187e-05}
{"train_loss": 1.1405727863311768, "global_step": 2594, "epoch": 29, "lr": 9.998524945240589e-05}
{"train_loss": 1.2065150737762451, "global_step": 2595, "epoch": 29, "lr": 9.998523536807e-05}
{"train_loss": 1.0667375326156616, "global_step": 2596, "epoch": 29, "lr": 9.998522127701422e-05}
{"train_loss": 1.119911551475525, "global_step": 2597, "epoch": 29, "lr": 9.998520717923851e-05}
{"train_loss": 1.1437034606933594, "global_step": 2598, "epoch": 29, "lr": 9.998519307474292e-05}
{"train_loss": 1.3681020736694336, "global_step": 2599, "epoch": 29, "lr": 9.998517896352741e-05}
{"train_loss": 1.4271262884140015, "global_step": 2600, "epoch": 29, "lr": 9.998516484559202e-05}
{"train_loss": 1.0230873823165894, "global_step": 2601, "epoch": 29, "lr": 9.998515072093672e-05}
{"train_loss": 0.9334182739257812, "global_step": 2602, "epoch": 29, "lr": 9.998513658956155e-05}
{"train_loss": 1.182342290878296, "global_step": 2603, "epoch": 29, "lr": 9.998512245146647e-05}
{"train_loss": 1.3595128059387207, "global_step": 2604, "epoch": 29, "lr": 9.998510830665152e-05}
{"train_loss": 1.1374547481536865, "global_step": 2605, "epoch": 29, "lr": 9.998509415511665e-05}
{"train_loss": 1.09683358669281, "global_step": 2606, "epoch": 29, "lr": 9.99850799968619e-05}
{"train_loss": 1.3255136013031006, "global_step": 2607, "epoch": 29, "lr": 9.998506583188728e-05}
{"train_loss": 1.3926976919174194, "global_step": 2608, "epoch": 29, "lr": 9.998505166019276e-05}
{"train_loss": 1.274088978767395, "global_step": 2609, "epoch": 29, "lr": 9.998503748177837e-05}
{"train_loss": 1.4573619365692139, "global_step": 2610, "epoch": 29, "lr": 9.99850232966441e-05}
{"train_loss": 1.2062063217163086, "global_step": 2611, "epoch": 29, "lr": 9.998500910478995e-05}
{"train_loss": 1.1296372413635254, "global_step": 2612, "epoch": 29, "lr": 9.998499490621592e-05}
{"train_loss": 1.192240595817566, "global_step": 2613, "epoch": 29, "lr": 9.998498070092202e-05}
{"train_loss": 1.1193610429763794, "global_step": 2614, "epoch": 29, "lr": 9.998496648890826e-05}
{"train_loss": 0.9847292304039001, "global_step": 2615, "epoch": 29, "lr": 9.998495227017462e-05}
{"train_loss": 1.22435462474823, "global_step": 2616, "epoch": 29, "lr": 9.99849380447211e-05}
{"train_loss": 1.0791375637054443, "global_step": 2617, "epoch": 29, "lr": 9.998492381254774e-05}
{"train_loss": 1.1750770807266235, "global_step": 2618, "epoch": 29, "lr": 9.998490957365452e-05}
{"train_loss": 1.2381656169891357, "global_step": 2619, "epoch": 29, "lr": 9.998489532804141e-05}
{"train_loss": 1.3774343729019165, "global_step": 2620, "epoch": 29, "lr": 9.998488107570844e-05}
{"train_loss": 1.4167729616165161, "global_step": 2621, "epoch": 29, "lr": 9.998486681665564e-05}
{"train_loss": 1.2852632999420166, "global_step": 2622, "epoch": 29, "lr": 9.998485255088297e-05}
{"train_loss": 1.3035953044891357, "global_step": 2623, "epoch": 29, "lr": 9.998483827839045e-05}
{"train_loss": 0.9768487811088562, "global_step": 2624, "epoch": 29, "lr": 9.998482399917806e-05}
{"train_loss": 1.2189764976501465, "global_step": 2625, "epoch": 29, "lr": 9.998480971324584e-05}
{"train_loss": 1.064065933227539, "global_step": 2626, "epoch": 29, "lr": 9.998479542059375e-05}
{"train_loss": 1.17847740650177, "global_step": 2627, "epoch": 29, "lr": 9.998478112122183e-05}
{"train_loss": 1.1017115116119385, "global_step": 2628, "epoch": 29, "lr": 9.998476681513007e-05}
{"train_loss": 1.0566869974136353, "global_step": 2629, "epoch": 29, "lr": 9.998475250231847e-05}
{"train_loss": 1.483403205871582, "global_step": 2630, "epoch": 29, "lr": 9.998473818278702e-05}
{"train_loss": 1.1792842149734497, "global_step": 2631, "epoch": 29, "lr": 9.998472385653572e-05}
{"train_loss": 1.4093838930130005, "global_step": 2632, "epoch": 29, "lr": 9.99847095235646e-05}
{"train_loss": 1.4543030261993408, "global_step": 2633, "epoch": 29, "lr": 9.998469518387364e-05}
{"train_loss": 1.282542109489441, "global_step": 2634, "epoch": 29, "lr": 9.998468083746285e-05}
{"train_loss": 1.3484693765640259, "global_step": 2635, "epoch": 29, "lr": 9.998466648433221e-05}
{"train_loss": 1.085902214050293, "global_step": 2636, "epoch": 29, "lr": 9.998465212448176e-05}
{"train_loss": 1.1128569841384888, "global_step": 2637, "epoch": 29, "lr": 9.998463775791147e-05}
{"train_loss": 1.236299753189087, "global_step": 2638, "epoch": 29, "lr": 9.998462338462137e-05}
{"train_loss": 1.1764495372772217, "global_step": 2639, "epoch": 29, "lr": 9.998460900461144e-05}
{"train_loss": 1.024202585220337, "global_step": 2640, "epoch": 29, "lr": 9.998459461788168e-05}
{"train_loss": 1.1570801734924316, "global_step": 2641, "epoch": 29, "lr": 9.998458022443211e-05}
{"train_loss": 1.115662932395935, "global_step": 2642, "epoch": 29, "lr": 9.998456582426273e-05}
{"train_loss": 1.204606294631958, "global_step": 2643, "epoch": 29, "lr": 9.998455141737352e-05}
{"train_loss": 1.2653801441192627, "global_step": 2644, "epoch": 29, "lr": 9.99845370037645e-05}
{"train_loss": 1.1859341859817505, "global_step": 2645, "epoch": 29, "lr": 9.998452258343567e-05}
{"train_loss": 1.1126646995544434, "global_step": 2646, "epoch": 29, "lr": 9.998450815638703e-05}
{"train_loss": 1.0097813606262207, "global_step": 2647, "epoch": 29, "lr": 9.998449372261857e-05}
{"train_loss": 1.1644341945648193, "global_step": 2648, "epoch": 29, "lr": 9.998447928213032e-05}
{"train_loss": 0.9819552302360535, "global_step": 2649, "epoch": 29, "lr": 9.998446483492227e-05}
{"train_loss": 1.430491328239441, "global_step": 2650, "epoch": 29, "lr": 9.99844503809944e-05}
{"train_loss": 1.3868398666381836, "global_step": 2651, "epoch": 29, "lr": 9.998443592034675e-05}
{"train_loss": 1.058016061782837, "global_step": 2652, "epoch": 29, "lr": 9.998442145297928e-05}
{"train_loss": 1.2286993265151978, "global_step": 2653, "epoch": 29, "lr": 9.998440697889204e-05}
{"train_loss": 1.2421958446502686, "global_step": 2654, "epoch": 29, "lr": 9.998439249808497e-05}
{"train_loss": 1.2432196140289307, "global_step": 2655, "epoch": 29, "lr": 9.998437801055814e-05}
{"train_loss": 1.1992762088775635, "global_step": 2656, "epoch": 29, "lr": 9.99843635163115e-05}
{"train_loss": 0.9264587163925171, "global_step": 2657, "epoch": 29, "lr": 9.998434901534509e-05}
{"train_loss": 1.2483264207839966, "global_step": 2658, "epoch": 29, "lr": 9.998433450765889e-05}
{"train_loss": 1.4195754528045654, "global_step": 2659, "epoch": 29, "lr": 9.99843199932529e-05}
{"train_loss": 1.2473399639129639, "global_step": 2660, "epoch": 29, "lr": 9.998430547212711e-05}
{"train_loss": 1.169457197189331, "global_step": 2661, "epoch": 29, "lr": 9.998429094428156e-05}
{"train_loss": 1.145658254623413, "global_step": 2662, "epoch": 29, "lr": 9.998427640971624e-05}
{"train_loss": 1.190747857093811, "global_step": 2663, "epoch": 29, "lr": 9.998426186843112e-05}
{"train_loss": 1.088127851486206, "global_step": 2664, "epoch": 29, "lr": 9.998424732042623e-05}
{"train_loss": 1.1701616048812866, "global_step": 2665, "epoch": 29, "lr": 9.998423276570158e-05}
{"train_loss": 1.281765341758728, "global_step": 2666, "epoch": 29, "lr": 9.998421820425715e-05}
{"train_loss": 1.1833566427230835, "global_step": 2667, "epoch": 29, "lr": 9.998420363609296e-05}
{"train_loss": 1.356116771697998, "global_step": 2668, "epoch": 29, "lr": 9.998418906120899e-05}
{"train_loss": 1.2110477960511539, "global_step": 2669, "epoch": 29, "lr": 9.998417447960527e-05, "val_loss": 1.1972445249557495}
{"train_loss": 1.1824390888214111, "global_step": 2670, "epoch": 30, "lr": 9.998415989128176e-05}
{"train_loss": 1.1606652736663818, "global_step": 2671, "epoch": 30, "lr": 9.998414529623852e-05}
{"train_loss": 1.2390544414520264, "global_step": 2672, "epoch": 30, "lr": 9.99841306944755e-05}
{"train_loss": 1.216573715209961, "global_step": 2673, "epoch": 30, "lr": 9.998411608599273e-05}
{"train_loss": 1.2932219505310059, "global_step": 2674, "epoch": 30, "lr": 9.998410147079021e-05}
{"train_loss": 1.2554059028625488, "global_step": 2675, "epoch": 30, "lr": 9.998408684886792e-05}
{"train_loss": 1.2723698616027832, "global_step": 2676, "epoch": 30, "lr": 9.99840722202259e-05}
{"train_loss": 1.0841655731201172, "global_step": 2677, "epoch": 30, "lr": 9.998405758486412e-05}
{"train_loss": 1.3217216730117798, "global_step": 2678, "epoch": 30, "lr": 9.998404294278259e-05}
{"train_loss": 1.2474949359893799, "global_step": 2679, "epoch": 30, "lr": 9.998402829398133e-05}
{"train_loss": 1.2869707345962524, "global_step": 2680, "epoch": 30, "lr": 9.99840136384603e-05}
{"train_loss": 1.1547781229019165, "global_step": 2681, "epoch": 30, "lr": 9.998399897621954e-05}
{"train_loss": 1.0681837797164917, "global_step": 2682, "epoch": 30, "lr": 9.998398430725905e-05}
{"train_loss": 1.0688248872756958, "global_step": 2683, "epoch": 30, "lr": 9.998396963157883e-05}
{"train_loss": 1.0995019674301147, "global_step": 2684, "epoch": 30, "lr": 9.998395494917885e-05}
{"train_loss": 1.1281172037124634, "global_step": 2685, "epoch": 30, "lr": 9.998394026005915e-05}
{"train_loss": 1.1644824743270874, "global_step": 2686, "epoch": 30, "lr": 9.998392556421972e-05}
{"train_loss": 1.3544830083847046, "global_step": 2687, "epoch": 30, "lr": 9.998391086166055e-05}
{"train_loss": 1.0716438293457031, "global_step": 2688, "epoch": 30, "lr": 9.998389615238167e-05}
{"train_loss": 1.3230454921722412, "global_step": 2689, "epoch": 30, "lr": 9.998388143638304e-05}
{"train_loss": 1.355358362197876, "global_step": 2690, "epoch": 30, "lr": 9.998386671366471e-05}
{"train_loss": 1.240730881690979, "global_step": 2691, "epoch": 30, "lr": 9.998385198422664e-05}
{"train_loss": 1.11908757686615, "global_step": 2692, "epoch": 30, "lr": 9.998383724806886e-05}
{"train_loss": 0.9933440089225769, "global_step": 2693, "epoch": 30, "lr": 9.998382250519136e-05}
{"train_loss": 1.1427147388458252, "global_step": 2694, "epoch": 30, "lr": 9.998380775559415e-05}
{"train_loss": 1.191773533821106, "global_step": 2695, "epoch": 30, "lr": 9.998379299927723e-05}
{"train_loss": 1.2654845714569092, "global_step": 2696, "epoch": 30, "lr": 9.998377823624058e-05}
{"train_loss": 1.078023910522461, "global_step": 2697, "epoch": 30, "lr": 9.998376346648424e-05}
{"train_loss": 1.164155125617981, "global_step": 2698, "epoch": 30, "lr": 9.998374869000818e-05}
{"train_loss": 1.2794737815856934, "global_step": 2699, "epoch": 30, "lr": 9.998373390681242e-05}
{"train_loss": 1.0430132150650024, "global_step": 2700, "epoch": 30, "lr": 9.998371911689695e-05}
{"train_loss": 1.1737478971481323, "global_step": 2701, "epoch": 30, "lr": 9.998370432026178e-05}
{"train_loss": 1.3745859861373901, "global_step": 2702, "epoch": 30, "lr": 9.99836895169069e-05}
{"train_loss": 1.079440712928772, "global_step": 2703, "epoch": 30, "lr": 9.998367470683234e-05}
{"train_loss": 1.2641096115112305, "global_step": 2704, "epoch": 30, "lr": 9.998365989003807e-05}
{"train_loss": 1.2136951684951782, "global_step": 2705, "epoch": 30, "lr": 9.998364506652412e-05}
{"train_loss": 1.446702480316162, "global_step": 2706, "epoch": 30, "lr": 9.998363023629048e-05}
{"train_loss": 1.0697764158248901, "global_step": 2707, "epoch": 30, "lr": 9.998361539933713e-05}
{"train_loss": 0.9212663173675537, "global_step": 2708, "epoch": 30, "lr": 9.99836005556641e-05}
{"train_loss": 1.107624888420105, "global_step": 2709, "epoch": 30, "lr": 9.998358570527139e-05}
{"train_loss": 1.1680129766464233, "global_step": 2710, "epoch": 30, "lr": 9.9983570848159e-05}
{"train_loss": 1.0841395854949951, "global_step": 2711, "epoch": 30, "lr": 9.998355598432692e-05}
{"train_loss": 1.0686941146850586, "global_step": 2712, "epoch": 30, "lr": 9.998354111377516e-05}
{"train_loss": 1.143911600112915, "global_step": 2713, "epoch": 30, "lr": 9.998352623650372e-05}
{"train_loss": 1.0064818859100342, "global_step": 2714, "epoch": 30, "lr": 9.99835113525126e-05}
{"train_loss": 1.2235478162765503, "global_step": 2715, "epoch": 30, "lr": 9.998349646180183e-05}
{"train_loss": 1.0387532711029053, "global_step": 2716, "epoch": 30, "lr": 9.998348156437137e-05}
{"train_loss": 1.1043587923049927, "global_step": 2717, "epoch": 30, "lr": 9.998346666022124e-05}
{"train_loss": 1.2345178127288818, "global_step": 2718, "epoch": 30, "lr": 9.998345174935144e-05}
{"train_loss": 1.3127251863479614, "global_step": 2719, "epoch": 30, "lr": 9.998343683176199e-05}
{"train_loss": 1.1691545248031616, "global_step": 2720, "epoch": 30, "lr": 9.998342190745286e-05}
{"train_loss": 1.383770227432251, "global_step": 2721, "epoch": 30, "lr": 9.998340697642407e-05}
{"train_loss": 1.1874672174453735, "global_step": 2722, "epoch": 30, "lr": 9.998339203867563e-05}
{"train_loss": 1.1118383407592773, "global_step": 2723, "epoch": 30, "lr": 9.998337709420753e-05}
{"train_loss": 1.077331781387329, "global_step": 2724, "epoch": 30, "lr": 9.998336214301975e-05}
{"train_loss": 1.170987606048584, "global_step": 2725, "epoch": 30, "lr": 9.998334718511235e-05}
{"train_loss": 1.2632670402526855, "global_step": 2726, "epoch": 30, "lr": 9.998333222048529e-05}
{"train_loss": 1.3341124057769775, "global_step": 2727, "epoch": 30, "lr": 9.998331724913856e-05}
{"train_loss": 1.2457146644592285, "global_step": 2728, "epoch": 30, "lr": 9.99833022710722e-05}
{"train_loss": 1.1688696146011353, "global_step": 2729, "epoch": 30, "lr": 9.998328728628619e-05}
{"train_loss": 1.063529133796692, "global_step": 2730, "epoch": 30, "lr": 9.998327229478054e-05}
{"train_loss": 1.2354652881622314, "global_step": 2731, "epoch": 30, "lr": 9.998325729655524e-05}
{"train_loss": 1.056046485900879, "global_step": 2732, "epoch": 30, "lr": 9.998324229161031e-05}
{"train_loss": 1.1483848094940186, "global_step": 2733, "epoch": 30, "lr": 9.998322727994574e-05}
{"train_loss": 1.0209753513336182, "global_step": 2734, "epoch": 30, "lr": 9.998321226156153e-05}
{"train_loss": 1.0196503400802612, "global_step": 2735, "epoch": 30, "lr": 9.998319723645768e-05}
{"train_loss": 1.2805591821670532, "global_step": 2736, "epoch": 30, "lr": 9.99831822046342e-05}
{"train_loss": 1.0108673572540283, "global_step": 2737, "epoch": 30, "lr": 9.99831671660911e-05}
{"train_loss": 1.0296623706817627, "global_step": 2738, "epoch": 30, "lr": 9.998315212082838e-05}
{"train_loss": 0.9813255667686462, "global_step": 2739, "epoch": 30, "lr": 9.998313706884603e-05}
{"train_loss": 1.3298702239990234, "global_step": 2740, "epoch": 30, "lr": 9.998312201014403e-05}
{"train_loss": 1.195948839187622, "global_step": 2741, "epoch": 30, "lr": 9.998310694472244e-05}
{"train_loss": 1.483916163444519, "global_step": 2742, "epoch": 30, "lr": 9.998309187258122e-05}
{"train_loss": 1.269624948501587, "global_step": 2743, "epoch": 30, "lr": 9.998307679372038e-05}
{"train_loss": 1.093935251235962, "global_step": 2744, "epoch": 30, "lr": 9.998306170813994e-05}
{"train_loss": 1.213423490524292, "global_step": 2745, "epoch": 30, "lr": 9.998304661583986e-05}
{"train_loss": 1.1723746061325073, "global_step": 2746, "epoch": 30, "lr": 9.998303151682018e-05}
{"train_loss": 1.1258107423782349, "global_step": 2747, "epoch": 30, "lr": 9.99830164110809e-05}
{"train_loss": 1.3912869691848755, "global_step": 2748, "epoch": 30, "lr": 9.9983001298622e-05}
{"train_loss": 1.2531629800796509, "global_step": 2749, "epoch": 30, "lr": 9.99829861794435e-05}
{"train_loss": 1.184524416923523, "global_step": 2750, "epoch": 30, "lr": 9.99829710535454e-05}
{"train_loss": 1.1611528396606445, "global_step": 2751, "epoch": 30, "lr": 9.99829559209277e-05}
{"train_loss": 1.0857445001602173, "global_step": 2752, "epoch": 30, "lr": 9.998294078159039e-05}
{"train_loss": 1.2120258808135986, "global_step": 2753, "epoch": 30, "lr": 9.998292563553349e-05}
{"train_loss": 1.2322075366973877, "global_step": 2754, "epoch": 30, "lr": 9.998291048275698e-05}
{"train_loss": 1.1401655673980713, "global_step": 2755, "epoch": 30, "lr": 9.998289532326089e-05}
{"train_loss": 1.2142473459243774, "global_step": 2756, "epoch": 30, "lr": 9.998288015704522e-05}
{"train_loss": 1.3392047882080078, "global_step": 2757, "epoch": 30, "lr": 9.998286498410995e-05}
{"train_loss": 1.1804954335930642, "global_step": 2758, "epoch": 30, "lr": 9.99828498044551e-05, "val_loss": 1.215726613998413, "train_action_mse_error": 72.58430480957031}
{"train_loss": 1.3530482053756714, "global_step": 2759, "epoch": 31, "lr": 9.998283461808064e-05}
{"train_loss": 1.048466444015503, "global_step": 2760, "epoch": 31, "lr": 9.998281942498662e-05}
{"train_loss": 1.0907760858535767, "global_step": 2761, "epoch": 31, "lr": 9.998280422517301e-05}
{"train_loss": 1.1795315742492676, "global_step": 2762, "epoch": 31, "lr": 9.998278901863983e-05}
{"train_loss": 1.1344366073608398, "global_step": 2763, "epoch": 31, "lr": 9.998277380538708e-05}
{"train_loss": 1.1147266626358032, "global_step": 2764, "epoch": 31, "lr": 9.998275858541473e-05}
{"train_loss": 1.1278060674667358, "global_step": 2765, "epoch": 31, "lr": 9.998274335872282e-05}
{"train_loss": 1.3699524402618408, "global_step": 2766, "epoch": 31, "lr": 9.998272812531135e-05}
{"train_loss": 1.2649425268173218, "global_step": 2767, "epoch": 31, "lr": 9.99827128851803e-05}
{"train_loss": 0.8628098368644714, "global_step": 2768, "epoch": 31, "lr": 9.998269763832968e-05}
{"train_loss": 1.1520556211471558, "global_step": 2769, "epoch": 31, "lr": 9.998268238475949e-05}
{"train_loss": 1.1264886856079102, "global_step": 2770, "epoch": 31, "lr": 9.998266712446977e-05}
{"train_loss": 1.2378613948822021, "global_step": 2771, "epoch": 31, "lr": 9.998265185746046e-05}
{"train_loss": 1.1308703422546387, "global_step": 2772, "epoch": 31, "lr": 9.998263658373159e-05}
{"train_loss": 1.1841307878494263, "global_step": 2773, "epoch": 31, "lr": 9.998262130328319e-05}
{"train_loss": 1.2186192274093628, "global_step": 2774, "epoch": 31, "lr": 9.998260601611523e-05}
{"train_loss": 1.073408842086792, "global_step": 2775, "epoch": 31, "lr": 9.998259072222769e-05}
{"train_loss": 0.9265751838684082, "global_step": 2776, "epoch": 31, "lr": 9.998257542162062e-05}
{"train_loss": 1.2108336687088013, "global_step": 2777, "epoch": 31, "lr": 9.9982560114294e-05}
{"train_loss": 0.9877594113349915, "global_step": 2778, "epoch": 31, "lr": 9.998254480024782e-05}
{"train_loss": 1.0704187154769897, "global_step": 2779, "epoch": 31, "lr": 9.998252947948213e-05}
{"train_loss": 1.1535332202911377, "global_step": 2780, "epoch": 31, "lr": 9.998251415199686e-05}
{"train_loss": 1.1841387748718262, "global_step": 2781, "epoch": 31, "lr": 9.998249881779206e-05}
{"train_loss": 1.0968133211135864, "global_step": 2782, "epoch": 31, "lr": 9.998248347686775e-05}
{"train_loss": 1.273147463798523, "global_step": 2783, "epoch": 31, "lr": 9.998246812922387e-05}
{"train_loss": 1.3866376876831055, "global_step": 2784, "epoch": 31, "lr": 9.998245277486047e-05}
{"train_loss": 1.1270779371261597, "global_step": 2785, "epoch": 31, "lr": 9.998243741377755e-05}
{"train_loss": 1.1005138158798218, "global_step": 2786, "epoch": 31, "lr": 9.998242204597508e-05}
{"train_loss": 1.2250491380691528, "global_step": 2787, "epoch": 31, "lr": 9.998240667145308e-05}
{"train_loss": 1.1385594606399536, "global_step": 2788, "epoch": 31, "lr": 9.998239129021158e-05}
{"train_loss": 1.1970590353012085, "global_step": 2789, "epoch": 31, "lr": 9.998237590225055e-05}
{"train_loss": 0.9524945020675659, "global_step": 2790, "epoch": 31, "lr": 9.998236050757e-05}
{"train_loss": 1.3253406286239624, "global_step": 2791, "epoch": 31, "lr": 9.998234510616991e-05}
{"train_loss": 1.3351311683654785, "global_step": 2792, "epoch": 31, "lr": 9.998232969805032e-05}
{"train_loss": 1.2144192457199097, "global_step": 2793, "epoch": 31, "lr": 9.998231428321121e-05}
{"train_loss": 1.0321274995803833, "global_step": 2794, "epoch": 31, "lr": 9.998229886165258e-05}
{"train_loss": 1.0108872652053833, "global_step": 2795, "epoch": 31, "lr": 9.998228343337446e-05}
{"train_loss": 1.2613857984542847, "global_step": 2796, "epoch": 31, "lr": 9.99822679983768e-05}
{"train_loss": 1.1297738552093506, "global_step": 2797, "epoch": 31, "lr": 9.998225255665965e-05}
{"train_loss": 1.311944842338562, "global_step": 2798, "epoch": 31, "lr": 9.998223710822301e-05}
{"train_loss": 1.2609236240386963, "global_step": 2799, "epoch": 31, "lr": 9.998222165306685e-05}
{"train_loss": 1.169211745262146, "global_step": 2800, "epoch": 31, "lr": 9.998220619119119e-05}
{"train_loss": 1.2128511667251587, "global_step": 2801, "epoch": 31, "lr": 9.998219072259603e-05}
{"train_loss": 1.0666855573654175, "global_step": 2802, "epoch": 31, "lr": 9.998217524728139e-05}
{"train_loss": 1.3689290285110474, "global_step": 2803, "epoch": 31, "lr": 9.998215976524723e-05}
{"train_loss": 1.0167051553726196, "global_step": 2804, "epoch": 31, "lr": 9.998214427649359e-05}
{"train_loss": 1.0125216245651245, "global_step": 2805, "epoch": 31, "lr": 9.998212878102047e-05}
{"train_loss": 1.0576112270355225, "global_step": 2806, "epoch": 31, "lr": 9.998211327882785e-05}
{"train_loss": 1.1584665775299072, "global_step": 2807, "epoch": 31, "lr": 9.998209776991576e-05}
{"train_loss": 1.3748576641082764, "global_step": 2808, "epoch": 31, "lr": 9.998208225428417e-05}
{"train_loss": 1.2691307067871094, "global_step": 2809, "epoch": 31, "lr": 9.99820667319331e-05}
{"train_loss": 1.21589994430542, "global_step": 2810, "epoch": 31, "lr": 9.998205120286255e-05}
{"train_loss": 1.1108671426773071, "global_step": 2811, "epoch": 31, "lr": 9.998203566707253e-05}
{"train_loss": 1.3320988416671753, "global_step": 2812, "epoch": 31, "lr": 9.998202012456303e-05}
{"train_loss": 1.2626080513000488, "global_step": 2813, "epoch": 31, "lr": 9.998200457533405e-05}
{"train_loss": 1.296201467514038, "global_step": 2814, "epoch": 31, "lr": 9.99819890193856e-05}
{"train_loss": 1.0816271305084229, "global_step": 2815, "epoch": 31, "lr": 9.998197345671769e-05}
{"train_loss": 1.2521839141845703, "global_step": 2816, "epoch": 31, "lr": 9.998195788733032e-05}
{"train_loss": 1.1309325695037842, "global_step": 2817, "epoch": 31, "lr": 9.998194231122347e-05}
{"train_loss": 1.1438767910003662, "global_step": 2818, "epoch": 31, "lr": 9.998192672839716e-05}
{"train_loss": 1.5084283351898193, "global_step": 2819, "epoch": 31, "lr": 9.998191113885138e-05}
{"train_loss": 1.257155418395996, "global_step": 2820, "epoch": 31, "lr": 9.998189554258616e-05}
{"train_loss": 1.0414518117904663, "global_step": 2821, "epoch": 31, "lr": 9.998187993960148e-05}
{"train_loss": 1.0669893026351929, "global_step": 2822, "epoch": 31, "lr": 9.998186432989733e-05}
{"train_loss": 1.1834709644317627, "global_step": 2823, "epoch": 31, "lr": 9.998184871347373e-05}
{"train_loss": 1.0719462633132935, "global_step": 2824, "epoch": 31, "lr": 9.998183309033069e-05}
{"train_loss": 1.029304027557373, "global_step": 2825, "epoch": 31, "lr": 9.99818174604682e-05}
{"train_loss": 0.9348286986351013, "global_step": 2826, "epoch": 31, "lr": 9.998180182388626e-05}
{"train_loss": 0.9792072772979736, "global_step": 2827, "epoch": 31, "lr": 9.99817861805849e-05}
{"train_loss": 1.4128050804138184, "global_step": 2828, "epoch": 31, "lr": 9.998177053056405e-05}
{"train_loss": 1.0940871238708496, "global_step": 2829, "epoch": 31, "lr": 9.99817548738238e-05}
{"train_loss": 1.3071513175964355, "global_step": 2830, "epoch": 31, "lr": 9.99817392103641e-05}
{"train_loss": 1.3570839166641235, "global_step": 2831, "epoch": 31, "lr": 9.998172354018496e-05}
{"train_loss": 1.2247611284255981, "global_step": 2832, "epoch": 31, "lr": 9.998170786328638e-05}
{"train_loss": 1.06786048412323, "global_step": 2833, "epoch": 31, "lr": 9.998169217966839e-05}
{"train_loss": 1.3821848630905151, "global_step": 2834, "epoch": 31, "lr": 9.998167648933096e-05}
{"train_loss": 1.2709505558013916, "global_step": 2835, "epoch": 31, "lr": 9.99816607922741e-05}
{"train_loss": 1.2804034948349, "global_step": 2836, "epoch": 31, "lr": 9.998164508849781e-05}
{"train_loss": 1.3945860862731934, "global_step": 2837, "epoch": 31, "lr": 9.998162937800211e-05}
{"train_loss": 1.3117986917495728, "global_step": 2838, "epoch": 31, "lr": 9.998161366078697e-05}
{"train_loss": 1.059262990951538, "global_step": 2839, "epoch": 31, "lr": 9.998159793685245e-05}
{"train_loss": 0.9786703586578369, "global_step": 2840, "epoch": 31, "lr": 9.998158220619847e-05}
{"train_loss": 1.3160449266433716, "global_step": 2841, "epoch": 31, "lr": 9.998156646882511e-05}
{"train_loss": 1.1549440622329712, "global_step": 2842, "epoch": 31, "lr": 9.998155072473231e-05}
{"train_loss": 1.082883596420288, "global_step": 2843, "epoch": 31, "lr": 9.998153497392011e-05}
{"train_loss": 1.186803936958313, "global_step": 2844, "epoch": 31, "lr": 9.99815192163885e-05}
{"train_loss": 1.0239802598953247, "global_step": 2845, "epoch": 31, "lr": 9.998150345213749e-05}
{"train_loss": 1.367200255393982, "global_step": 2846, "epoch": 31, "lr": 9.998148768116706e-05}
{"train_loss": 1.1786124820119879, "global_step": 2847, "epoch": 31, "lr": 9.998147190347724e-05, "val_loss": 1.2285021543502808}
{"train_loss": 1.1413209438323975, "global_step": 2848, "epoch": 32, "lr": 9.998145611906803e-05}
{"train_loss": 1.307185411453247, "global_step": 2849, "epoch": 32, "lr": 9.998144032793941e-05}
{"train_loss": 1.215878963470459, "global_step": 2850, "epoch": 32, "lr": 9.99814245300914e-05}
{"train_loss": 1.2539095878601074, "global_step": 2851, "epoch": 32, "lr": 9.9981408725524e-05}
{"train_loss": 1.176395058631897, "global_step": 2852, "epoch": 32, "lr": 9.998139291423719e-05}
{"train_loss": 1.3231521844863892, "global_step": 2853, "epoch": 32, "lr": 9.998137709623101e-05}
{"train_loss": 1.1663135290145874, "global_step": 2854, "epoch": 32, "lr": 9.998136127150543e-05}
{"train_loss": 1.3332712650299072, "global_step": 2855, "epoch": 32, "lr": 9.998134544006048e-05}
{"train_loss": 1.1956371068954468, "global_step": 2856, "epoch": 32, "lr": 9.998132960189614e-05}
{"train_loss": 1.3384208679199219, "global_step": 2857, "epoch": 32, "lr": 9.998131375701241e-05}
{"train_loss": 1.2188657522201538, "global_step": 2858, "epoch": 32, "lr": 9.998129790540932e-05}
{"train_loss": 1.0908317565917969, "global_step": 2859, "epoch": 32, "lr": 9.998128204708683e-05}
{"train_loss": 1.0516629219055176, "global_step": 2860, "epoch": 32, "lr": 9.998126618204498e-05}
{"train_loss": 1.0981851816177368, "global_step": 2861, "epoch": 32, "lr": 9.998125031028376e-05}
{"train_loss": 1.1586017608642578, "global_step": 2862, "epoch": 32, "lr": 9.998123443180317e-05}
{"train_loss": 1.2264454364776611, "global_step": 2863, "epoch": 32, "lr": 9.998121854660321e-05}
{"train_loss": 1.1188973188400269, "global_step": 2864, "epoch": 32, "lr": 9.998120265468389e-05}
{"train_loss": 1.1736218929290771, "global_step": 2865, "epoch": 32, "lr": 9.99811867560452e-05}
{"train_loss": 1.1190409660339355, "global_step": 2866, "epoch": 32, "lr": 9.998117085068714e-05}
{"train_loss": 1.1150970458984375, "global_step": 2867, "epoch": 32, "lr": 9.998115493860973e-05}
{"train_loss": 0.9558498859405518, "global_step": 2868, "epoch": 32, "lr": 9.998113901981296e-05}
{"train_loss": 1.2958532571792603, "global_step": 2869, "epoch": 32, "lr": 9.998112309429684e-05}
{"train_loss": 1.1558656692504883, "global_step": 2870, "epoch": 32, "lr": 9.998110716206137e-05}
{"train_loss": 1.3456931114196777, "global_step": 2871, "epoch": 32, "lr": 9.998109122310655e-05}
{"train_loss": 1.035581111907959, "global_step": 2872, "epoch": 32, "lr": 9.998107527743236e-05}
{"train_loss": 1.273133635520935, "global_step": 2873, "epoch": 32, "lr": 9.998105932503885e-05}
{"train_loss": 1.1078662872314453, "global_step": 2874, "epoch": 32, "lr": 9.998104336592598e-05}
{"train_loss": 1.2543765306472778, "global_step": 2875, "epoch": 32, "lr": 9.998102740009378e-05}
{"train_loss": 1.0283619165420532, "global_step": 2876, "epoch": 32, "lr": 9.998101142754224e-05}
{"train_loss": 1.1666138172149658, "global_step": 2877, "epoch": 32, "lr": 9.998099544827135e-05}
{"train_loss": 1.2472684383392334, "global_step": 2878, "epoch": 32, "lr": 9.998097946228113e-05}
{"train_loss": 1.146934151649475, "global_step": 2879, "epoch": 32, "lr": 9.998096346957158e-05}
{"train_loss": 1.1740832328796387, "global_step": 2880, "epoch": 32, "lr": 9.998094747014268e-05}
{"train_loss": 1.2285300493240356, "global_step": 2881, "epoch": 32, "lr": 9.998093146399448e-05}
{"train_loss": 1.1023893356323242, "global_step": 2882, "epoch": 32, "lr": 9.998091545112693e-05}
{"train_loss": 1.035025715827942, "global_step": 2883, "epoch": 32, "lr": 9.998089943154007e-05}
{"train_loss": 1.0567805767059326, "global_step": 2884, "epoch": 32, "lr": 9.998088340523388e-05}
{"train_loss": 1.1430432796478271, "global_step": 2885, "epoch": 32, "lr": 9.998086737220837e-05}
{"train_loss": 1.2236260175704956, "global_step": 2886, "epoch": 32, "lr": 9.998085133246354e-05}
{"train_loss": 1.034119963645935, "global_step": 2887, "epoch": 32, "lr": 9.99808352859994e-05}
{"train_loss": 1.266937494277954, "global_step": 2888, "epoch": 32, "lr": 9.998081923281593e-05}
{"train_loss": 1.1858556270599365, "global_step": 2889, "epoch": 32, "lr": 9.998080317291316e-05}
{"train_loss": 1.054774522781372, "global_step": 2890, "epoch": 32, "lr": 9.998078710629109e-05}
{"train_loss": 1.1840283870697021, "global_step": 2891, "epoch": 32, "lr": 9.998077103294969e-05}
{"train_loss": 1.2666555643081665, "global_step": 2892, "epoch": 32, "lr": 9.998075495288901e-05}
{"train_loss": 1.0762436389923096, "global_step": 2893, "epoch": 32, "lr": 9.998073886610902e-05}
{"train_loss": 1.0598851442337036, "global_step": 2894, "epoch": 32, "lr": 9.998072277260971e-05}
{"train_loss": 1.0572255849838257, "global_step": 2895, "epoch": 32, "lr": 9.998070667239112e-05}
{"train_loss": 0.947826623916626, "global_step": 2896, "epoch": 32, "lr": 9.998069056545322e-05}
{"train_loss": 1.189085602760315, "global_step": 2897, "epoch": 32, "lr": 9.998067445179604e-05}
{"train_loss": 1.0287015438079834, "global_step": 2898, "epoch": 32, "lr": 9.998065833141956e-05}
{"train_loss": 1.1726669073104858, "global_step": 2899, "epoch": 32, "lr": 9.998064220432378e-05}
{"train_loss": 1.0385502576828003, "global_step": 2900, "epoch": 32, "lr": 9.998062607050874e-05}
{"train_loss": 1.0993784666061401, "global_step": 2901, "epoch": 32, "lr": 9.998060992997438e-05}
{"train_loss": 1.0850512981414795, "global_step": 2902, "epoch": 32, "lr": 9.998059378272076e-05}
{"train_loss": 1.2415647506713867, "global_step": 2903, "epoch": 32, "lr": 9.998057762874784e-05}
{"train_loss": 0.9710407257080078, "global_step": 2904, "epoch": 32, "lr": 9.998056146805566e-05}
{"train_loss": 1.1528253555297852, "global_step": 2905, "epoch": 32, "lr": 9.998054530064419e-05}
{"train_loss": 1.2010992765426636, "global_step": 2906, "epoch": 32, "lr": 9.998052912651344e-05}
{"train_loss": 0.9825027585029602, "global_step": 2907, "epoch": 32, "lr": 9.998051294566342e-05}
{"train_loss": 1.1142048835754395, "global_step": 2908, "epoch": 32, "lr": 9.998049675809413e-05}
{"train_loss": 1.0800232887268066, "global_step": 2909, "epoch": 32, "lr": 9.998048056380557e-05}
{"train_loss": 1.3295608758926392, "global_step": 2910, "epoch": 32, "lr": 9.998046436279775e-05}
{"train_loss": 1.1732760667800903, "global_step": 2911, "epoch": 32, "lr": 9.998044815507067e-05}
{"train_loss": 0.9394569396972656, "global_step": 2912, "epoch": 32, "lr": 9.998043194062432e-05}
{"train_loss": 1.3742913007736206, "global_step": 2913, "epoch": 32, "lr": 9.998041571945872e-05}
{"train_loss": 1.0883066654205322, "global_step": 2914, "epoch": 32, "lr": 9.998039949157386e-05}
{"train_loss": 1.3521171808242798, "global_step": 2915, "epoch": 32, "lr": 9.998038325696973e-05}
{"train_loss": 1.5773937702178955, "global_step": 2916, "epoch": 32, "lr": 9.998036701564636e-05}
{"train_loss": 1.3465824127197266, "global_step": 2917, "epoch": 32, "lr": 9.998035076760373e-05}
{"train_loss": 1.1765644550323486, "global_step": 2918, "epoch": 32, "lr": 9.998033451284185e-05}
{"train_loss": 0.9975359439849854, "global_step": 2919, "epoch": 32, "lr": 9.998031825136073e-05}
{"train_loss": 1.2991622686386108, "global_step": 2920, "epoch": 32, "lr": 9.998030198316038e-05}
{"train_loss": 1.2037639617919922, "global_step": 2921, "epoch": 32, "lr": 9.998028570824076e-05}
{"train_loss": 0.9749653339385986, "global_step": 2922, "epoch": 32, "lr": 9.998026942660191e-05}
{"train_loss": 1.0001671314239502, "global_step": 2923, "epoch": 32, "lr": 9.998025313824383e-05}
{"train_loss": 1.0673022270202637, "global_step": 2924, "epoch": 32, "lr": 9.998023684316651e-05}
{"train_loss": 0.9335205554962158, "global_step": 2925, "epoch": 32, "lr": 9.998022054136994e-05}
{"train_loss": 1.2735700607299805, "global_step": 2926, "epoch": 32, "lr": 9.998020423285417e-05}
{"train_loss": 1.2219927310943604, "global_step": 2927, "epoch": 32, "lr": 9.998018791761914e-05}
{"train_loss": 1.2968392372131348, "global_step": 2928, "epoch": 32, "lr": 9.998017159566491e-05}
{"train_loss": 1.0467472076416016, "global_step": 2929, "epoch": 32, "lr": 9.998015526699143e-05}
{"train_loss": 1.09378981590271, "global_step": 2930, "epoch": 32, "lr": 9.998013893159874e-05}
{"train_loss": 1.1547883749008179, "global_step": 2931, "epoch": 32, "lr": 9.998012258948684e-05}
{"train_loss": 1.3024053573608398, "global_step": 2932, "epoch": 32, "lr": 9.99801062406557e-05}
{"train_loss": 1.1685186624526978, "global_step": 2933, "epoch": 32, "lr": 9.998008988510534e-05}
{"train_loss": 1.160072684288025, "global_step": 2934, "epoch": 32, "lr": 9.998007352283578e-05}
{"train_loss": 1.0213356018066406, "global_step": 2935, "epoch": 32, "lr": 9.998005715384702e-05}
{"train_loss": 1.156421946005875, "global_step": 2936, "epoch": 32, "lr": 9.998004077813903e-05, "val_loss": 1.177834391593933}
{"train_loss": 1.1726740598678589, "global_step": 2937, "epoch": 33, "lr": 9.998002439571185e-05}
{"train_loss": 1.1099954843521118, "global_step": 2938, "epoch": 33, "lr": 9.998000800656545e-05}
{"train_loss": 1.0221730470657349, "global_step": 2939, "epoch": 33, "lr": 9.997999161069987e-05}
{"train_loss": 1.126668930053711, "global_step": 2940, "epoch": 33, "lr": 9.997997520811507e-05}
{"train_loss": 1.258488655090332, "global_step": 2941, "epoch": 33, "lr": 9.997995879881106e-05}
{"train_loss": 1.1167998313903809, "global_step": 2942, "epoch": 33, "lr": 9.997994238278787e-05}
{"train_loss": 1.0931661128997803, "global_step": 2943, "epoch": 33, "lr": 9.997992596004548e-05}
{"train_loss": 1.1975762844085693, "global_step": 2944, "epoch": 33, "lr": 9.99799095305839e-05}
{"train_loss": 1.0161329507827759, "global_step": 2945, "epoch": 33, "lr": 9.997989309440315e-05}
{"train_loss": 1.0798102617263794, "global_step": 2946, "epoch": 33, "lr": 9.997987665150318e-05}
{"train_loss": 1.0421297550201416, "global_step": 2947, "epoch": 33, "lr": 9.997986020188403e-05}
{"train_loss": 1.2342504262924194, "global_step": 2948, "epoch": 33, "lr": 9.997984374554572e-05}
{"train_loss": 1.0302543640136719, "global_step": 2949, "epoch": 33, "lr": 9.997982728248821e-05}
{"train_loss": 0.8990676999092102, "global_step": 2950, "epoch": 33, "lr": 9.997981081271153e-05}
{"train_loss": 1.341740608215332, "global_step": 2951, "epoch": 33, "lr": 9.997979433621567e-05}
{"train_loss": 1.0166908502578735, "global_step": 2952, "epoch": 33, "lr": 9.997977785300064e-05}
{"train_loss": 1.179146409034729, "global_step": 2953, "epoch": 33, "lr": 9.997976136306642e-05}
{"train_loss": 1.280375599861145, "global_step": 2954, "epoch": 33, "lr": 9.997974486641306e-05}
{"train_loss": 1.0363715887069702, "global_step": 2955, "epoch": 33, "lr": 9.997972836304051e-05}
{"train_loss": 1.4667710065841675, "global_step": 2956, "epoch": 33, "lr": 9.99797118529488e-05}
{"train_loss": 1.1561038494110107, "global_step": 2957, "epoch": 33, "lr": 9.997969533613794e-05}
{"train_loss": 1.2455793619155884, "global_step": 2958, "epoch": 33, "lr": 9.99796788126079e-05}
{"train_loss": 1.2602444887161255, "global_step": 2959, "epoch": 33, "lr": 9.997966228235871e-05}
{"train_loss": 1.0705804824829102, "global_step": 2960, "epoch": 33, "lr": 9.997964574539036e-05}
{"train_loss": 1.1400630474090576, "global_step": 2961, "epoch": 33, "lr": 9.997962920170285e-05}
{"train_loss": 1.3080297708511353, "global_step": 2962, "epoch": 33, "lr": 9.99796126512962e-05}
{"train_loss": 1.0744441747665405, "global_step": 2963, "epoch": 33, "lr": 9.99795960941704e-05}
{"train_loss": 1.0444905757904053, "global_step": 2964, "epoch": 33, "lr": 9.997957953032545e-05}
{"train_loss": 1.1022027730941772, "global_step": 2965, "epoch": 33, "lr": 9.997956295976135e-05}
{"train_loss": 1.1647553443908691, "global_step": 2966, "epoch": 33, "lr": 9.997954638247812e-05}
{"train_loss": 1.0923576354980469, "global_step": 2967, "epoch": 33, "lr": 9.997952979847573e-05}
{"train_loss": 1.0965739488601685, "global_step": 2968, "epoch": 33, "lr": 9.997951320775422e-05}
{"train_loss": 1.2649171352386475, "global_step": 2969, "epoch": 33, "lr": 9.997949661031357e-05}
{"train_loss": 1.2732003927230835, "global_step": 2970, "epoch": 33, "lr": 9.997948000615377e-05}
{"train_loss": 1.0496466159820557, "global_step": 2971, "epoch": 33, "lr": 9.997946339527486e-05}
{"train_loss": 1.1563889980316162, "global_step": 2972, "epoch": 33, "lr": 9.99794467776768e-05}
{"train_loss": 1.2267911434173584, "global_step": 2973, "epoch": 33, "lr": 9.997943015335963e-05}
{"train_loss": 1.105604648590088, "global_step": 2974, "epoch": 33, "lr": 9.997941352232332e-05}
{"train_loss": 1.1514781713485718, "global_step": 2975, "epoch": 33, "lr": 9.99793968845679e-05}
{"train_loss": 1.1697864532470703, "global_step": 2976, "epoch": 33, "lr": 9.997938024009334e-05}
{"train_loss": 1.0366556644439697, "global_step": 2977, "epoch": 33, "lr": 9.997936358889968e-05}
{"train_loss": 1.2680307626724243, "global_step": 2978, "epoch": 33, "lr": 9.99793469309869e-05}
{"train_loss": 1.0296372175216675, "global_step": 2979, "epoch": 33, "lr": 9.9979330266355e-05}
{"train_loss": 0.8722600936889648, "global_step": 2980, "epoch": 33, "lr": 9.997931359500398e-05}
{"train_loss": 1.1129697561264038, "global_step": 2981, "epoch": 33, "lr": 9.997929691693387e-05}
{"train_loss": 1.3568781614303589, "global_step": 2982, "epoch": 33, "lr": 9.997928023214464e-05}
{"train_loss": 1.1311907768249512, "global_step": 2983, "epoch": 33, "lr": 9.997926354063631e-05}
{"train_loss": 0.9272521138191223, "global_step": 2984, "epoch": 33, "lr": 9.997924684240888e-05}
{"train_loss": 1.1555027961730957, "global_step": 2985, "epoch": 33, "lr": 9.997923013746235e-05}
{"train_loss": 1.1261358261108398, "global_step": 2986, "epoch": 33, "lr": 9.997921342579673e-05}
{"train_loss": 1.1791611909866333, "global_step": 2987, "epoch": 33, "lr": 9.9979196707412e-05}
{"train_loss": 1.0960203409194946, "global_step": 2988, "epoch": 33, "lr": 9.997917998230816e-05}
{"train_loss": 0.9919782280921936, "global_step": 2989, "epoch": 33, "lr": 9.997916325048526e-05}
{"train_loss": 1.095313310623169, "global_step": 2990, "epoch": 33, "lr": 9.997914651194326e-05}
{"train_loss": 1.2683290243148804, "global_step": 2991, "epoch": 33, "lr": 9.997912976668217e-05}
{"train_loss": 1.2294375896453857, "global_step": 2992, "epoch": 33, "lr": 9.9979113014702e-05}
{"train_loss": 1.2489198446273804, "global_step": 2993, "epoch": 33, "lr": 9.997909625600273e-05}
{"train_loss": 1.2250365018844604, "global_step": 2994, "epoch": 33, "lr": 9.99790794905844e-05}
{"train_loss": 1.0356513261795044, "global_step": 2995, "epoch": 33, "lr": 9.997906271844699e-05}
{"train_loss": 1.2066991329193115, "global_step": 2996, "epoch": 33, "lr": 9.997904593959048e-05}
{"train_loss": 1.1878124475479126, "global_step": 2997, "epoch": 33, "lr": 9.997902915401494e-05}
{"train_loss": 1.0474224090576172, "global_step": 2998, "epoch": 33, "lr": 9.997901236172029e-05}
{"train_loss": 1.0981448888778687, "global_step": 2999, "epoch": 33, "lr": 9.997899556270658e-05}
{"train_loss": 1.259401798248291, "global_step": 3000, "epoch": 33, "lr": 9.997897875697382e-05}
{"train_loss": 1.1546608209609985, "global_step": 3001, "epoch": 33, "lr": 9.997896194452199e-05}
{"train_loss": 1.0648884773254395, "global_step": 3002, "epoch": 33, "lr": 9.997894512535108e-05}
{"train_loss": 1.2233108282089233, "global_step": 3003, "epoch": 33, "lr": 9.997892829946111e-05}
{"train_loss": 0.9522812962532043, "global_step": 3004, "epoch": 33, "lr": 9.99789114668521e-05}
{"train_loss": 1.3122330904006958, "global_step": 3005, "epoch": 33, "lr": 9.997889462752404e-05}
{"train_loss": 0.9471096992492676, "global_step": 3006, "epoch": 33, "lr": 9.997887778147691e-05}
{"train_loss": 1.1049648523330688, "global_step": 3007, "epoch": 33, "lr": 9.997886092871074e-05}
{"train_loss": 1.1771186590194702, "global_step": 3008, "epoch": 33, "lr": 9.997884406922551e-05}
{"train_loss": 1.1014554500579834, "global_step": 3009, "epoch": 33, "lr": 9.997882720302123e-05}
{"train_loss": 1.2463699579238892, "global_step": 3010, "epoch": 33, "lr": 9.997881033009793e-05}
{"train_loss": 0.9980184435844421, "global_step": 3011, "epoch": 33, "lr": 9.997879345045556e-05}
{"train_loss": 1.2878071069717407, "global_step": 3012, "epoch": 33, "lr": 9.997877656409417e-05}
{"train_loss": 1.2273597717285156, "global_step": 3013, "epoch": 33, "lr": 9.997875967101373e-05}
{"train_loss": 1.2282732725143433, "global_step": 3014, "epoch": 33, "lr": 9.997874277121426e-05}
{"train_loss": 1.1804592609405518, "global_step": 3015, "epoch": 33, "lr": 9.997872586469577e-05}
{"train_loss": 1.1643569469451904, "global_step": 3016, "epoch": 33, "lr": 9.997870895145822e-05}
{"train_loss": 1.0587488412857056, "global_step": 3017, "epoch": 33, "lr": 9.997869203150168e-05}
{"train_loss": 1.1145776510238647, "global_step": 3018, "epoch": 33, "lr": 9.997867510482608e-05}
{"train_loss": 1.1519901752471924, "global_step": 3019, "epoch": 33, "lr": 9.997865817143147e-05}
{"train_loss": 1.4826791286468506, "global_step": 3020, "epoch": 33, "lr": 9.997864123131784e-05}
{"train_loss": 1.2142614126205444, "global_step": 3021, "epoch": 33, "lr": 9.997862428448519e-05}
{"train_loss": 1.1530447006225586, "global_step": 3022, "epoch": 33, "lr": 9.997860733093352e-05}
{"train_loss": 1.1027343273162842, "global_step": 3023, "epoch": 33, "lr": 9.997859037066284e-05}
{"train_loss": 1.1642073392868042, "global_step": 3024, "epoch": 33, "lr": 9.997857340367313e-05}
{"train_loss": 1.1490876440251812, "global_step": 3025, "epoch": 33, "lr": 9.997855642996443e-05, "val_loss": 1.180017113685608}
{"train_loss": 1.3434913158416748, "global_step": 3026, "epoch": 34, "lr": 9.997853944953673e-05}
{"train_loss": 1.118620753288269, "global_step": 3027, "epoch": 34, "lr": 9.997852246239001e-05}
{"train_loss": 1.0578573942184448, "global_step": 3028, "epoch": 34, "lr": 9.997850546852428e-05}
{"train_loss": 1.1059712171554565, "global_step": 3029, "epoch": 34, "lr": 9.997848846793956e-05}
{"train_loss": 1.0979703664779663, "global_step": 3030, "epoch": 34, "lr": 9.997847146063584e-05}
{"train_loss": 1.1970136165618896, "global_step": 3031, "epoch": 34, "lr": 9.997845444661312e-05}
{"train_loss": 1.1261845827102661, "global_step": 3032, "epoch": 34, "lr": 9.99784374258714e-05}
{"train_loss": 0.9513977766036987, "global_step": 3033, "epoch": 34, "lr": 9.99784203984107e-05}
{"train_loss": 1.2128067016601562, "global_step": 3034, "epoch": 34, "lr": 9.997840336423101e-05}
{"train_loss": 1.391082525253296, "global_step": 3035, "epoch": 34, "lr": 9.997838632333234e-05}
{"train_loss": 1.077562689781189, "global_step": 3036, "epoch": 34, "lr": 9.997836927571467e-05}
{"train_loss": 1.1602391004562378, "global_step": 3037, "epoch": 34, "lr": 9.997835222137802e-05}
{"train_loss": 0.8455329537391663, "global_step": 3038, "epoch": 34, "lr": 9.99783351603224e-05}
{"train_loss": 1.2073012590408325, "global_step": 3039, "epoch": 34, "lr": 9.997831809254778e-05}
{"train_loss": 1.235931634902954, "global_step": 3040, "epoch": 34, "lr": 9.997830101805422e-05}
{"train_loss": 1.3367784023284912, "global_step": 3041, "epoch": 34, "lr": 9.997828393684166e-05}
{"train_loss": 1.2562578916549683, "global_step": 3042, "epoch": 34, "lr": 9.997826684891013e-05}
{"train_loss": 1.1474008560180664, "global_step": 3043, "epoch": 34, "lr": 9.997824975425963e-05}
{"train_loss": 1.276808500289917, "global_step": 3044, "epoch": 34, "lr": 9.997823265289017e-05}
{"train_loss": 1.1466951370239258, "global_step": 3045, "epoch": 34, "lr": 9.997821554480174e-05}
{"train_loss": 1.1751521825790405, "global_step": 3046, "epoch": 34, "lr": 9.997819842999436e-05}
{"train_loss": 1.2331833839416504, "global_step": 3047, "epoch": 34, "lr": 9.9978181308468e-05}
{"train_loss": 1.381030559539795, "global_step": 3048, "epoch": 34, "lr": 9.99781641802227e-05}
{"train_loss": 1.1662629842758179, "global_step": 3049, "epoch": 34, "lr": 9.997814704525845e-05}
{"train_loss": 1.1449296474456787, "global_step": 3050, "epoch": 34, "lr": 9.997812990357524e-05}
{"train_loss": 1.1662293672561646, "global_step": 3051, "epoch": 34, "lr": 9.997811275517307e-05}
{"train_loss": 1.0831201076507568, "global_step": 3052, "epoch": 34, "lr": 9.997809560005196e-05}
{"train_loss": 0.9899998307228088, "global_step": 3053, "epoch": 34, "lr": 9.99780784382119e-05}
{"train_loss": 1.289303183555603, "global_step": 3054, "epoch": 34, "lr": 9.99780612696529e-05}
{"train_loss": 1.1534873247146606, "global_step": 3055, "epoch": 34, "lr": 9.997804409437497e-05}
{"train_loss": 1.1636894941329956, "global_step": 3056, "epoch": 34, "lr": 9.997802691237807e-05}
{"train_loss": 1.1980291604995728, "global_step": 3057, "epoch": 34, "lr": 9.997800972366228e-05}
{"train_loss": 1.223832607269287, "global_step": 3058, "epoch": 34, "lr": 9.997799252822752e-05}
{"train_loss": 1.0338268280029297, "global_step": 3059, "epoch": 34, "lr": 9.997797532607383e-05}
{"train_loss": 0.9923296570777893, "global_step": 3060, "epoch": 34, "lr": 9.997795811720121e-05}
{"train_loss": 1.0030491352081299, "global_step": 3061, "epoch": 34, "lr": 9.997794090160969e-05}
{"train_loss": 1.2772011756896973, "global_step": 3062, "epoch": 34, "lr": 9.997792367929921e-05}
{"train_loss": 0.9439525008201599, "global_step": 3063, "epoch": 34, "lr": 9.997790645026983e-05}
{"train_loss": 1.1507169008255005, "global_step": 3064, "epoch": 34, "lr": 9.997788921452151e-05}
{"train_loss": 1.0466039180755615, "global_step": 3065, "epoch": 34, "lr": 9.997787197205429e-05}
{"train_loss": 1.087886929512024, "global_step": 3066, "epoch": 34, "lr": 9.997785472286816e-05}
{"train_loss": 1.1912918090820312, "global_step": 3067, "epoch": 34, "lr": 9.99778374669631e-05}
{"train_loss": 1.1991939544677734, "global_step": 3068, "epoch": 34, "lr": 9.997782020433913e-05}
{"train_loss": 1.2145566940307617, "global_step": 3069, "epoch": 34, "lr": 9.997780293499625e-05}
{"train_loss": 1.3839364051818848, "global_step": 3070, "epoch": 34, "lr": 9.997778565893447e-05}
{"train_loss": 1.0619701147079468, "global_step": 3071, "epoch": 34, "lr": 9.99777683761538e-05}
{"train_loss": 1.11449134349823, "global_step": 3072, "epoch": 34, "lr": 9.99777510866542e-05}
{"train_loss": 1.0942744016647339, "global_step": 3073, "epoch": 34, "lr": 9.997773379043572e-05}
{"train_loss": 1.1020373106002808, "global_step": 3074, "epoch": 34, "lr": 9.997771648749833e-05}
{"train_loss": 1.2209892272949219, "global_step": 3075, "epoch": 34, "lr": 9.997769917784206e-05}
{"train_loss": 1.0368659496307373, "global_step": 3076, "epoch": 34, "lr": 9.997768186146687e-05}
{"train_loss": 0.9745094776153564, "global_step": 3077, "epoch": 34, "lr": 9.997766453837281e-05}
{"train_loss": 1.2738423347473145, "global_step": 3078, "epoch": 34, "lr": 9.997764720855987e-05}
{"train_loss": 1.1494590044021606, "global_step": 3079, "epoch": 34, "lr": 9.997762987202804e-05}
{"train_loss": 1.3683326244354248, "global_step": 3080, "epoch": 34, "lr": 9.997761252877731e-05}
{"train_loss": 1.0348565578460693, "global_step": 3081, "epoch": 34, "lr": 9.997759517880771e-05}
{"train_loss": 0.9073778986930847, "global_step": 3082, "epoch": 34, "lr": 9.997757782211924e-05}
{"train_loss": 1.1014604568481445, "global_step": 3083, "epoch": 34, "lr": 9.997756045871189e-05}
{"train_loss": 1.3062717914581299, "global_step": 3084, "epoch": 34, "lr": 9.997754308858565e-05}
{"train_loss": 1.2306824922561646, "global_step": 3085, "epoch": 34, "lr": 9.997752571174056e-05}
{"train_loss": 1.200615406036377, "global_step": 3086, "epoch": 34, "lr": 9.997750832817658e-05}
{"train_loss": 1.1510401964187622, "global_step": 3087, "epoch": 34, "lr": 9.997749093789375e-05}
{"train_loss": 1.1152113676071167, "global_step": 3088, "epoch": 34, "lr": 9.997747354089207e-05}
{"train_loss": 1.0105400085449219, "global_step": 3089, "epoch": 34, "lr": 9.997745613717149e-05}
{"train_loss": 1.0544015169143677, "global_step": 3090, "epoch": 34, "lr": 9.997743872673208e-05}
{"train_loss": 0.9887953996658325, "global_step": 3091, "epoch": 34, "lr": 9.99774213095738e-05}
{"train_loss": 1.1242296695709229, "global_step": 3092, "epoch": 34, "lr": 9.997740388569665e-05}
{"train_loss": 1.2125440835952759, "global_step": 3093, "epoch": 34, "lr": 9.997738645510068e-05}
{"train_loss": 1.0220513343811035, "global_step": 3094, "epoch": 34, "lr": 9.997736901778583e-05}
{"train_loss": 1.0257643461227417, "global_step": 3095, "epoch": 34, "lr": 9.997735157375215e-05}
{"train_loss": 1.1457842588424683, "global_step": 3096, "epoch": 34, "lr": 9.997733412299961e-05}
{"train_loss": 1.154701828956604, "global_step": 3097, "epoch": 34, "lr": 9.997731666552824e-05}
{"train_loss": 1.0584334135055542, "global_step": 3098, "epoch": 34, "lr": 9.997729920133803e-05}
{"train_loss": 1.1501551866531372, "global_step": 3099, "epoch": 34, "lr": 9.997728173042898e-05}
{"train_loss": 1.2477275133132935, "global_step": 3100, "epoch": 34, "lr": 9.997726425280107e-05}
{"train_loss": 1.038034439086914, "global_step": 3101, "epoch": 34, "lr": 9.997724676845435e-05}
{"train_loss": 1.1732027530670166, "global_step": 3102, "epoch": 34, "lr": 9.997722927738879e-05}
{"train_loss": 1.1410393714904785, "global_step": 3103, "epoch": 34, "lr": 9.997721177960441e-05}
{"train_loss": 1.0448079109191895, "global_step": 3104, "epoch": 34, "lr": 9.99771942751012e-05}
{"train_loss": 1.1854560375213623, "global_step": 3105, "epoch": 34, "lr": 9.997717676387916e-05}
{"train_loss": 1.1021331548690796, "global_step": 3106, "epoch": 34, "lr": 9.99771592459383e-05}
{"train_loss": 1.2861746549606323, "global_step": 3107, "epoch": 34, "lr": 9.997714172127863e-05}
{"train_loss": 1.1637552976608276, "global_step": 3108, "epoch": 34, "lr": 9.997712418990013e-05}
{"train_loss": 1.1823481321334839, "global_step": 3109, "epoch": 34, "lr": 9.997710665180281e-05}
{"train_loss": 0.9159778356552124, "global_step": 3110, "epoch": 34, "lr": 9.997708910698669e-05}
{"train_loss": 1.1019707918167114, "global_step": 3111, "epoch": 34, "lr": 9.997707155545174e-05}
{"train_loss": 1.1751598119735718, "global_step": 3112, "epoch": 34, "lr": 9.997705399719801e-05}
{"train_loss": 1.1449228525161743, "global_step": 3113, "epoch": 34, "lr": 9.997703643222545e-05}
{"train_loss": 1.1399368730823645, "global_step": 3114, "epoch": 34, "lr": 9.99770188605341e-05, "val_loss": 1.2322968244552612}
{"train_loss": 1.352256178855896, "global_step": 3115, "epoch": 35, "lr": 9.997700128212395e-05}
{"train_loss": 1.249687910079956, "global_step": 3116, "epoch": 35, "lr": 9.9976983696995e-05}
{"train_loss": 1.2001864910125732, "global_step": 3117, "epoch": 35, "lr": 9.997696610514724e-05}
{"train_loss": 1.1642115116119385, "global_step": 3118, "epoch": 35, "lr": 9.99769485065807e-05}
{"train_loss": 1.257495403289795, "global_step": 3119, "epoch": 35, "lr": 9.997693090129536e-05}
{"train_loss": 1.2760168313980103, "global_step": 3120, "epoch": 35, "lr": 9.997691328929125e-05}
{"train_loss": 1.1595035791397095, "global_step": 3121, "epoch": 35, "lr": 9.997689567056833e-05}
{"train_loss": 1.3289541006088257, "global_step": 3122, "epoch": 35, "lr": 9.997687804512663e-05}
{"train_loss": 1.040112018585205, "global_step": 3123, "epoch": 35, "lr": 9.997686041296616e-05}
{"train_loss": 1.026024341583252, "global_step": 3124, "epoch": 35, "lr": 9.997684277408689e-05}
{"train_loss": 1.093095302581787, "global_step": 3125, "epoch": 35, "lr": 9.997682512848886e-05}
{"train_loss": 0.9708200097084045, "global_step": 3126, "epoch": 35, "lr": 9.997680747617205e-05}
{"train_loss": 1.2014737129211426, "global_step": 3127, "epoch": 35, "lr": 9.997678981713648e-05}
{"train_loss": 1.440616488456726, "global_step": 3128, "epoch": 35, "lr": 9.997677215138212e-05}
{"train_loss": 1.1987457275390625, "global_step": 3129, "epoch": 35, "lr": 9.9976754478909e-05}
{"train_loss": 1.274885654449463, "global_step": 3130, "epoch": 35, "lr": 9.997673679971711e-05}
{"train_loss": 1.0350314378738403, "global_step": 3131, "epoch": 35, "lr": 9.997671911380646e-05}
{"train_loss": 1.0703808069229126, "global_step": 3132, "epoch": 35, "lr": 9.997670142117706e-05}
{"train_loss": 1.0698482990264893, "global_step": 3133, "epoch": 35, "lr": 9.997668372182888e-05}
{"train_loss": 1.1664437055587769, "global_step": 3134, "epoch": 35, "lr": 9.997666601576196e-05}
{"train_loss": 1.2115557193756104, "global_step": 3135, "epoch": 35, "lr": 9.997664830297628e-05}
{"train_loss": 1.2222223281860352, "global_step": 3136, "epoch": 35, "lr": 9.997663058347184e-05}
{"train_loss": 1.195482611656189, "global_step": 3137, "epoch": 35, "lr": 9.997661285724867e-05}
{"train_loss": 1.2398009300231934, "global_step": 3138, "epoch": 35, "lr": 9.997659512430675e-05}
{"train_loss": 1.269752025604248, "global_step": 3139, "epoch": 35, "lr": 9.997657738464607e-05}
{"train_loss": 0.8584979176521301, "global_step": 3140, "epoch": 35, "lr": 9.997655963826667e-05}
{"train_loss": 0.9016561508178711, "global_step": 3141, "epoch": 35, "lr": 9.997654188516851e-05}
{"train_loss": 1.233349084854126, "global_step": 3142, "epoch": 35, "lr": 9.997652412535163e-05}
{"train_loss": 1.1311696767807007, "global_step": 3143, "epoch": 35, "lr": 9.997650635881601e-05}
{"train_loss": 1.067312240600586, "global_step": 3144, "epoch": 35, "lr": 9.997648858556165e-05}
{"train_loss": 1.095194935798645, "global_step": 3145, "epoch": 35, "lr": 9.997647080558857e-05}
{"train_loss": 1.2760576009750366, "global_step": 3146, "epoch": 35, "lr": 9.997645301889675e-05}
{"train_loss": 1.189211368560791, "global_step": 3147, "epoch": 35, "lr": 9.997643522548623e-05}
{"train_loss": 1.132069706916809, "global_step": 3148, "epoch": 35, "lr": 9.997641742535697e-05}
{"train_loss": 1.0538952350616455, "global_step": 3149, "epoch": 35, "lr": 9.997639961850899e-05}
{"train_loss": 1.4054546356201172, "global_step": 3150, "epoch": 35, "lr": 9.997638180494228e-05}
{"train_loss": 1.1172857284545898, "global_step": 3151, "epoch": 35, "lr": 9.997636398465689e-05}
{"train_loss": 1.1613389253616333, "global_step": 3152, "epoch": 35, "lr": 9.997634615765276e-05}
{"train_loss": 1.2376362085342407, "global_step": 3153, "epoch": 35, "lr": 9.997632832392993e-05}
{"train_loss": 1.157436728477478, "global_step": 3154, "epoch": 35, "lr": 9.997631048348837e-05}
{"train_loss": 1.1195943355560303, "global_step": 3155, "epoch": 35, "lr": 9.997629263632813e-05}
{"train_loss": 1.259930968284607, "global_step": 3156, "epoch": 35, "lr": 9.997627478244917e-05}
{"train_loss": 1.1279726028442383, "global_step": 3157, "epoch": 35, "lr": 9.997625692185153e-05}
{"train_loss": 1.0344338417053223, "global_step": 3158, "epoch": 35, "lr": 9.997623905453517e-05}
{"train_loss": 0.8586307764053345, "global_step": 3159, "epoch": 35, "lr": 9.997622118050012e-05}
{"train_loss": 1.2554690837860107, "global_step": 3160, "epoch": 35, "lr": 9.997620329974637e-05}
{"train_loss": 1.236566424369812, "global_step": 3161, "epoch": 35, "lr": 9.997618541227395e-05}
{"train_loss": 1.3331396579742432, "global_step": 3162, "epoch": 35, "lr": 9.997616751808283e-05}
{"train_loss": 0.979144811630249, "global_step": 3163, "epoch": 35, "lr": 9.997614961717301e-05}
{"train_loss": 1.0698355436325073, "global_step": 3164, "epoch": 35, "lr": 9.997613170954452e-05}
{"train_loss": 1.2036497592926025, "global_step": 3165, "epoch": 35, "lr": 9.997611379519735e-05}
{"train_loss": 1.1541835069656372, "global_step": 3166, "epoch": 35, "lr": 9.997609587413149e-05}
{"train_loss": 1.084965467453003, "global_step": 3167, "epoch": 35, "lr": 9.997607794634696e-05}
{"train_loss": 1.1002585887908936, "global_step": 3168, "epoch": 35, "lr": 9.997606001184375e-05}
{"train_loss": 1.100353717803955, "global_step": 3169, "epoch": 35, "lr": 9.997604207062188e-05}
{"train_loss": 0.8655222654342651, "global_step": 3170, "epoch": 35, "lr": 9.997602412268132e-05}
{"train_loss": 0.995631992816925, "global_step": 3171, "epoch": 35, "lr": 9.997600616802211e-05}
{"train_loss": 0.9901238679885864, "global_step": 3172, "epoch": 35, "lr": 9.997598820664423e-05}
{"train_loss": 1.2261589765548706, "global_step": 3173, "epoch": 35, "lr": 9.997597023854768e-05}
{"train_loss": 1.2705280780792236, "global_step": 3174, "epoch": 35, "lr": 9.997595226373249e-05}
{"train_loss": 0.9787325859069824, "global_step": 3175, "epoch": 35, "lr": 9.997593428219862e-05}
{"train_loss": 1.3566306829452515, "global_step": 3176, "epoch": 35, "lr": 9.997591629394612e-05}
{"train_loss": 0.9278983473777771, "global_step": 3177, "epoch": 35, "lr": 9.997589829897494e-05}
{"train_loss": 1.1800482273101807, "global_step": 3178, "epoch": 35, "lr": 9.997588029728513e-05}
{"train_loss": 1.1512553691864014, "global_step": 3179, "epoch": 35, "lr": 9.997586228887666e-05}
{"train_loss": 1.1723558902740479, "global_step": 3180, "epoch": 35, "lr": 9.997584427374954e-05}
{"train_loss": 1.2465230226516724, "global_step": 3181, "epoch": 35, "lr": 9.997582625190379e-05}
{"train_loss": 1.1934036016464233, "global_step": 3182, "epoch": 35, "lr": 9.997580822333939e-05}
{"train_loss": 1.0367525815963745, "global_step": 3183, "epoch": 35, "lr": 9.997579018805636e-05}
{"train_loss": 1.2560752630233765, "global_step": 3184, "epoch": 35, "lr": 9.997577214605469e-05}
{"train_loss": 1.0557831525802612, "global_step": 3185, "epoch": 35, "lr": 9.997575409733437e-05}
{"train_loss": 1.125317931175232, "global_step": 3186, "epoch": 35, "lr": 9.997573604189545e-05}
{"train_loss": 1.0216537714004517, "global_step": 3187, "epoch": 35, "lr": 9.997571797973788e-05}
{"train_loss": 0.9665431976318359, "global_step": 3188, "epoch": 35, "lr": 9.997569991086168e-05}
{"train_loss": 1.0943031311035156, "global_step": 3189, "epoch": 35, "lr": 9.997568183526688e-05}
{"train_loss": 1.086076021194458, "global_step": 3190, "epoch": 35, "lr": 9.997566375295344e-05}
{"train_loss": 1.1400034427642822, "global_step": 3191, "epoch": 35, "lr": 9.99756456639214e-05}
{"train_loss": 1.2054147720336914, "global_step": 3192, "epoch": 35, "lr": 9.997562756817072e-05}
{"train_loss": 1.0893913507461548, "global_step": 3193, "epoch": 35, "lr": 9.997560946570144e-05}
{"train_loss": 0.96480792760849, "global_step": 3194, "epoch": 35, "lr": 9.997559135651354e-05}
{"train_loss": 1.3533539772033691, "global_step": 3195, "epoch": 35, "lr": 9.997557324060703e-05}
{"train_loss": 0.982742965221405, "global_step": 3196, "epoch": 35, "lr": 9.997555511798193e-05}
{"train_loss": 1.0393670797348022, "global_step": 3197, "epoch": 35, "lr": 9.997553698863821e-05}
{"train_loss": 1.1706552505493164, "global_step": 3198, "epoch": 35, "lr": 9.997551885257588e-05}
{"train_loss": 1.000221610069275, "global_step": 3199, "epoch": 35, "lr": 9.997550070979497e-05}
{"train_loss": 1.1434911489486694, "global_step": 3200, "epoch": 35, "lr": 9.997548256029545e-05}
{"train_loss": 1.2262505292892456, "global_step": 3201, "epoch": 35, "lr": 9.997546440407735e-05}
{"train_loss": 1.0484957695007324, "global_step": 3202, "epoch": 35, "lr": 9.997544624114066e-05}
{"train_loss": 1.1360685061872675, "global_step": 3203, "epoch": 35, "lr": 9.997542807148537e-05, "val_loss": 1.194636583328247, "train_action_mse_error": 63.01003646850586}
{"train_loss": 1.1484979391098022, "global_step": 3204, "epoch": 36, "lr": 9.997540989511148e-05}
{"train_loss": 1.237894058227539, "global_step": 3205, "epoch": 36, "lr": 9.997539171201901e-05}
{"train_loss": 1.0131115913391113, "global_step": 3206, "epoch": 36, "lr": 9.997537352220799e-05}
{"train_loss": 1.1827466487884521, "global_step": 3207, "epoch": 36, "lr": 9.997535532567835e-05}
{"train_loss": 0.9468509554862976, "global_step": 3208, "epoch": 36, "lr": 9.997533712243014e-05}
{"train_loss": 0.9804393649101257, "global_step": 3209, "epoch": 36, "lr": 9.997531891246336e-05}
{"train_loss": 1.088991403579712, "global_step": 3210, "epoch": 36, "lr": 9.997530069577803e-05}
{"train_loss": 0.9812312126159668, "global_step": 3211, "epoch": 36, "lr": 9.997528247237411e-05}
{"train_loss": 1.2511299848556519, "global_step": 3212, "epoch": 36, "lr": 9.99752642422516e-05}
{"train_loss": 1.1966650485992432, "global_step": 3213, "epoch": 36, "lr": 9.997524600541057e-05}
{"train_loss": 1.0294421911239624, "global_step": 3214, "epoch": 36, "lr": 9.997522776185096e-05}
{"train_loss": 1.1141687631607056, "global_step": 3215, "epoch": 36, "lr": 9.997520951157278e-05}
{"train_loss": 1.1324152946472168, "global_step": 3216, "epoch": 36, "lr": 9.997519125457604e-05}
{"train_loss": 1.215507984161377, "global_step": 3217, "epoch": 36, "lr": 9.997517299086075e-05}
{"train_loss": 1.0821340084075928, "global_step": 3218, "epoch": 36, "lr": 9.997515472042692e-05}
{"train_loss": 1.0542536973953247, "global_step": 3219, "epoch": 36, "lr": 9.997513644327452e-05}
{"train_loss": 0.9741822481155396, "global_step": 3220, "epoch": 36, "lr": 9.997511815940358e-05}
{"train_loss": 1.0770177841186523, "global_step": 3221, "epoch": 36, "lr": 9.997509986881411e-05}
{"train_loss": 1.293721079826355, "global_step": 3222, "epoch": 36, "lr": 9.997508157150606e-05}
{"train_loss": 1.4845261573791504, "global_step": 3223, "epoch": 36, "lr": 9.99750632674795e-05}
{"train_loss": 1.061923623085022, "global_step": 3224, "epoch": 36, "lr": 9.99750449567344e-05}
{"train_loss": 1.0068786144256592, "global_step": 3225, "epoch": 36, "lr": 9.997502663927075e-05}
{"train_loss": 0.9165974259376526, "global_step": 3226, "epoch": 36, "lr": 9.997500831508857e-05}
{"train_loss": 1.261696696281433, "global_step": 3227, "epoch": 36, "lr": 9.997498998418786e-05}
{"train_loss": 1.1362502574920654, "global_step": 3228, "epoch": 36, "lr": 9.997497164656864e-05}
{"train_loss": 1.0385018587112427, "global_step": 3229, "epoch": 36, "lr": 9.997495330223087e-05}
{"train_loss": 0.9573482871055603, "global_step": 3230, "epoch": 36, "lr": 9.997493495117459e-05}
{"train_loss": 1.0584702491760254, "global_step": 3231, "epoch": 36, "lr": 9.997491659339979e-05}
{"train_loss": 1.2547016143798828, "global_step": 3232, "epoch": 36, "lr": 9.997489822890647e-05}
{"train_loss": 1.4450739622116089, "global_step": 3233, "epoch": 36, "lr": 9.997487985769463e-05}
{"train_loss": 1.1902726888656616, "global_step": 3234, "epoch": 36, "lr": 9.997486147976428e-05}
{"train_loss": 1.1878138780593872, "global_step": 3235, "epoch": 36, "lr": 9.99748430951154e-05}
{"train_loss": 1.0428452491760254, "global_step": 3236, "epoch": 36, "lr": 9.997482470374804e-05}
{"train_loss": 1.0992884635925293, "global_step": 3237, "epoch": 36, "lr": 9.997480630566216e-05}
{"train_loss": 1.036622405052185, "global_step": 3238, "epoch": 36, "lr": 9.997478790085777e-05}
{"train_loss": 1.1129298210144043, "global_step": 3239, "epoch": 36, "lr": 9.997476948933488e-05}
{"train_loss": 1.1046019792556763, "global_step": 3240, "epoch": 36, "lr": 9.99747510710935e-05}
{"train_loss": 1.2366176843643188, "global_step": 3241, "epoch": 36, "lr": 9.997473264613362e-05}
{"train_loss": 1.3051737546920776, "global_step": 3242, "epoch": 36, "lr": 9.997471421445525e-05}
{"train_loss": 1.2604973316192627, "global_step": 3243, "epoch": 36, "lr": 9.99746957760584e-05}
{"train_loss": 0.9774469137191772, "global_step": 3244, "epoch": 36, "lr": 9.997467733094304e-05}
{"train_loss": 0.9716585278511047, "global_step": 3245, "epoch": 36, "lr": 9.99746588791092e-05}
{"train_loss": 1.0743197202682495, "global_step": 3246, "epoch": 36, "lr": 9.997464042055689e-05}
{"train_loss": 1.1524847745895386, "global_step": 3247, "epoch": 36, "lr": 9.997462195528609e-05}
{"train_loss": 1.3513946533203125, "global_step": 3248, "epoch": 36, "lr": 9.99746034832968e-05}
{"train_loss": 1.1342344284057617, "global_step": 3249, "epoch": 36, "lr": 9.997458500458904e-05}
{"train_loss": 1.2962819337844849, "global_step": 3250, "epoch": 36, "lr": 9.997456651916281e-05}
{"train_loss": 1.0635004043579102, "global_step": 3251, "epoch": 36, "lr": 9.997454802701811e-05}
{"train_loss": 1.153960108757019, "global_step": 3252, "epoch": 36, "lr": 9.997452952815494e-05}
{"train_loss": 1.259887456893921, "global_step": 3253, "epoch": 36, "lr": 9.997451102257329e-05}
{"train_loss": 1.0095477104187012, "global_step": 3254, "epoch": 36, "lr": 9.99744925102732e-05}
{"train_loss": 1.069923758506775, "global_step": 3255, "epoch": 36, "lr": 9.997447399125464e-05}
{"train_loss": 1.170843482017517, "global_step": 3256, "epoch": 36, "lr": 9.99744554655176e-05}
{"train_loss": 1.155301809310913, "global_step": 3257, "epoch": 36, "lr": 9.997443693306212e-05}
{"train_loss": 1.26748526096344, "global_step": 3258, "epoch": 36, "lr": 9.99744183938882e-05}
{"train_loss": 1.0787732601165771, "global_step": 3259, "epoch": 36, "lr": 9.997439984799582e-05}
{"train_loss": 1.0993455648422241, "global_step": 3260, "epoch": 36, "lr": 9.997438129538497e-05}
{"train_loss": 1.1280585527420044, "global_step": 3261, "epoch": 36, "lr": 9.99743627360557e-05}
{"train_loss": 1.0931265354156494, "global_step": 3262, "epoch": 36, "lr": 9.997434417000797e-05}
{"train_loss": 1.3131719827651978, "global_step": 3263, "epoch": 36, "lr": 9.99743255972418e-05}
{"train_loss": 0.9293196797370911, "global_step": 3264, "epoch": 36, "lr": 9.997430701775721e-05}
{"train_loss": 1.142794132232666, "global_step": 3265, "epoch": 36, "lr": 9.997428843155417e-05}
{"train_loss": 1.1713199615478516, "global_step": 3266, "epoch": 36, "lr": 9.997426983863268e-05}
{"train_loss": 1.0881389379501343, "global_step": 3267, "epoch": 36, "lr": 9.997425123899277e-05}
{"train_loss": 1.016474962234497, "global_step": 3268, "epoch": 36, "lr": 9.997423263263443e-05}
{"train_loss": 1.0783759355545044, "global_step": 3269, "epoch": 36, "lr": 9.99742140195577e-05}
{"train_loss": 1.0958943367004395, "global_step": 3270, "epoch": 36, "lr": 9.99741953997625e-05}
{"train_loss": 1.0875500440597534, "global_step": 3271, "epoch": 36, "lr": 9.99741767732489e-05}
{"train_loss": 1.0597028732299805, "global_step": 3272, "epoch": 36, "lr": 9.997415814001686e-05}
{"train_loss": 1.1533710956573486, "global_step": 3273, "epoch": 36, "lr": 9.997413950006642e-05}
{"train_loss": 1.0013505220413208, "global_step": 3274, "epoch": 36, "lr": 9.997412085339756e-05}
{"train_loss": 0.9299648404121399, "global_step": 3275, "epoch": 36, "lr": 9.997410220001028e-05}
{"train_loss": 1.2581267356872559, "global_step": 3276, "epoch": 36, "lr": 9.997408353990462e-05}
{"train_loss": 1.1005321741104126, "global_step": 3277, "epoch": 36, "lr": 9.997406487308053e-05}
{"train_loss": 1.2077707052230835, "global_step": 3278, "epoch": 36, "lr": 9.997404619953805e-05}
{"train_loss": 1.058868646621704, "global_step": 3279, "epoch": 36, "lr": 9.997402751927716e-05}
{"train_loss": 1.148248553276062, "global_step": 3280, "epoch": 36, "lr": 9.997400883229786e-05}
{"train_loss": 1.2337998151779175, "global_step": 3281, "epoch": 36, "lr": 9.997399013860018e-05}
{"train_loss": 1.2194257974624634, "global_step": 3282, "epoch": 36, "lr": 9.99739714381841e-05}
{"train_loss": 1.301944613456726, "global_step": 3283, "epoch": 36, "lr": 9.997395273104964e-05}
{"train_loss": 1.2118990421295166, "global_step": 3284, "epoch": 36, "lr": 9.997393401719678e-05}
{"train_loss": 1.157127022743225, "global_step": 3285, "epoch": 36, "lr": 9.997391529662555e-05}
{"train_loss": 1.1344186067581177, "global_step": 3286, "epoch": 36, "lr": 9.997389656933592e-05}
{"train_loss": 1.1108733415603638, "global_step": 3287, "epoch": 36, "lr": 9.99738778353279e-05}
{"train_loss": 1.3333549499511719, "global_step": 3288, "epoch": 36, "lr": 9.99738590946015e-05}
{"train_loss": 1.0702883005142212, "global_step": 3289, "epoch": 36, "lr": 9.997384034715674e-05}
{"train_loss": 1.0849004983901978, "global_step": 3290, "epoch": 36, "lr": 9.99738215929936e-05}
{"train_loss": 1.2173821926116943, "global_step": 3291, "epoch": 36, "lr": 9.99738028321121e-05}
{"train_loss": 1.1330563707298107, "global_step": 3292, "epoch": 36, "lr": 9.997378406451221e-05, "val_loss": 1.163321614265442}
{"train_loss": 1.1894021034240723, "global_step": 3293, "epoch": 37, "lr": 9.997376529019398e-05}
{"train_loss": 1.0226267576217651, "global_step": 3294, "epoch": 37, "lr": 9.997374650915736e-05}
{"train_loss": 1.149627923965454, "global_step": 3295, "epoch": 37, "lr": 9.997372772140238e-05}
{"train_loss": 1.0218514204025269, "global_step": 3296, "epoch": 37, "lr": 9.997370892692907e-05}
{"train_loss": 0.9947755932807922, "global_step": 3297, "epoch": 37, "lr": 9.997369012573739e-05}
{"train_loss": 0.9802494645118713, "global_step": 3298, "epoch": 37, "lr": 9.997367131782735e-05}
{"train_loss": 0.910712480545044, "global_step": 3299, "epoch": 37, "lr": 9.997365250319896e-05}
{"train_loss": 1.1682082414627075, "global_step": 3300, "epoch": 37, "lr": 9.997363368185222e-05}
{"train_loss": 1.3246301412582397, "global_step": 3301, "epoch": 37, "lr": 9.997361485378713e-05}
{"train_loss": 1.0340664386749268, "global_step": 3302, "epoch": 37, "lr": 9.997359601900371e-05}
{"train_loss": 1.300539255142212, "global_step": 3303, "epoch": 37, "lr": 9.997357717750195e-05}
{"train_loss": 1.3289823532104492, "global_step": 3304, "epoch": 37, "lr": 9.997355832928184e-05}
{"train_loss": 1.1784014701843262, "global_step": 3305, "epoch": 37, "lr": 9.99735394743434e-05}
{"train_loss": 1.1878718137741089, "global_step": 3306, "epoch": 37, "lr": 9.997352061268663e-05}
{"train_loss": 1.1520030498504639, "global_step": 3307, "epoch": 37, "lr": 9.997350174431152e-05}
{"train_loss": 1.2375184297561646, "global_step": 3308, "epoch": 37, "lr": 9.997348286921808e-05}
{"train_loss": 1.0674799680709839, "global_step": 3309, "epoch": 37, "lr": 9.997346398740633e-05}
{"train_loss": 1.2864052057266235, "global_step": 3310, "epoch": 37, "lr": 9.997344509887624e-05}
{"train_loss": 1.113626480102539, "global_step": 3311, "epoch": 37, "lr": 9.997342620362784e-05}
{"train_loss": 1.1052883863449097, "global_step": 3312, "epoch": 37, "lr": 9.997340730166112e-05}
{"train_loss": 0.9794937968254089, "global_step": 3313, "epoch": 37, "lr": 9.997338839297608e-05}
{"train_loss": 1.0711474418640137, "global_step": 3314, "epoch": 37, "lr": 9.997336947757272e-05}
{"train_loss": 1.012982964515686, "global_step": 3315, "epoch": 37, "lr": 9.997335055545107e-05}
{"train_loss": 1.1576787233352661, "global_step": 3316, "epoch": 37, "lr": 9.99733316266111e-05}
{"train_loss": 1.0764176845550537, "global_step": 3317, "epoch": 37, "lr": 9.997331269105283e-05}
{"train_loss": 0.9927642941474915, "global_step": 3318, "epoch": 37, "lr": 9.997329374877624e-05}
{"train_loss": 1.1141226291656494, "global_step": 3319, "epoch": 37, "lr": 9.997327479978139e-05}
{"train_loss": 1.1705588102340698, "global_step": 3320, "epoch": 37, "lr": 9.997325584406821e-05}
{"train_loss": 1.149510145187378, "global_step": 3321, "epoch": 37, "lr": 9.997323688163673e-05}
{"train_loss": 1.1456407308578491, "global_step": 3322, "epoch": 37, "lr": 9.997321791248698e-05}
{"train_loss": 1.0717592239379883, "global_step": 3323, "epoch": 37, "lr": 9.997319893661891e-05}
{"train_loss": 1.1227459907531738, "global_step": 3324, "epoch": 37, "lr": 9.997317995403256e-05}
{"train_loss": 1.0465617179870605, "global_step": 3325, "epoch": 37, "lr": 9.997316096472796e-05}
{"train_loss": 1.3362489938735962, "global_step": 3326, "epoch": 37, "lr": 9.997314196870505e-05}
{"train_loss": 1.0181739330291748, "global_step": 3327, "epoch": 37, "lr": 9.997312296596387e-05}
{"train_loss": 1.1968858242034912, "global_step": 3328, "epoch": 37, "lr": 9.997310395650439e-05}
{"train_loss": 1.3399096727371216, "global_step": 3329, "epoch": 37, "lr": 9.997308494032665e-05}
{"train_loss": 1.0818721055984497, "global_step": 3330, "epoch": 37, "lr": 9.997306591743065e-05}
{"train_loss": 1.0298463106155396, "global_step": 3331, "epoch": 37, "lr": 9.997304688781637e-05}
{"train_loss": 1.348199486732483, "global_step": 3332, "epoch": 37, "lr": 9.997302785148381e-05}
{"train_loss": 1.1999222040176392, "global_step": 3333, "epoch": 37, "lr": 9.997300880843301e-05}
{"train_loss": 0.9962365031242371, "global_step": 3334, "epoch": 37, "lr": 9.997298975866394e-05}
{"train_loss": 1.0615559816360474, "global_step": 3335, "epoch": 37, "lr": 9.99729707021766e-05}
{"train_loss": 1.0189645290374756, "global_step": 3336, "epoch": 37, "lr": 9.9972951638971e-05}
{"train_loss": 1.0666449069976807, "global_step": 3337, "epoch": 37, "lr": 9.997293256904716e-05}
{"train_loss": 0.9952604174613953, "global_step": 3338, "epoch": 37, "lr": 9.997291349240508e-05}
{"train_loss": 1.2135432958602905, "global_step": 3339, "epoch": 37, "lr": 9.997289440904471e-05}
{"train_loss": 1.1662707328796387, "global_step": 3340, "epoch": 37, "lr": 9.997287531896612e-05}
{"train_loss": 1.0924363136291504, "global_step": 3341, "epoch": 37, "lr": 9.99728562221693e-05}
{"train_loss": 1.105772852897644, "global_step": 3342, "epoch": 37, "lr": 9.997283711865422e-05}
{"train_loss": 0.9763180017471313, "global_step": 3343, "epoch": 37, "lr": 9.99728180084209e-05}
{"train_loss": 1.1834592819213867, "global_step": 3344, "epoch": 37, "lr": 9.997279889146935e-05}
{"train_loss": 0.9771215915679932, "global_step": 3345, "epoch": 37, "lr": 9.997277976779955e-05}
{"train_loss": 1.2367478609085083, "global_step": 3346, "epoch": 37, "lr": 9.997276063741155e-05}
{"train_loss": 1.1662569046020508, "global_step": 3347, "epoch": 37, "lr": 9.997274150030529e-05}
{"train_loss": 1.1655908823013306, "global_step": 3348, "epoch": 37, "lr": 9.997272235648084e-05}
{"train_loss": 1.077568531036377, "global_step": 3349, "epoch": 37, "lr": 9.997270320593813e-05}
{"train_loss": 0.9826284646987915, "global_step": 3350, "epoch": 37, "lr": 9.997268404867721e-05}
{"train_loss": 0.9993242621421814, "global_step": 3351, "epoch": 37, "lr": 9.997266488469808e-05}
{"train_loss": 1.0632891654968262, "global_step": 3352, "epoch": 37, "lr": 9.997264571400074e-05}
{"train_loss": 0.9730017185211182, "global_step": 3353, "epoch": 37, "lr": 9.997262653658517e-05}
{"train_loss": 1.0768815279006958, "global_step": 3354, "epoch": 37, "lr": 9.99726073524514e-05}
{"train_loss": 1.0636354684829712, "global_step": 3355, "epoch": 37, "lr": 9.99725881615994e-05}
{"train_loss": 1.013985514640808, "global_step": 3356, "epoch": 37, "lr": 9.997256896402922e-05}
{"train_loss": 1.1272107362747192, "global_step": 3357, "epoch": 37, "lr": 9.997254975974083e-05}
{"train_loss": 1.0899070501327515, "global_step": 3358, "epoch": 37, "lr": 9.997253054873423e-05}
{"train_loss": 0.9611297249794006, "global_step": 3359, "epoch": 37, "lr": 9.997251133100945e-05}
{"train_loss": 1.0253499746322632, "global_step": 3360, "epoch": 37, "lr": 9.997249210656647e-05}
{"train_loss": 1.1721693277359009, "global_step": 3361, "epoch": 37, "lr": 9.99724728754053e-05}
{"train_loss": 1.1186121702194214, "global_step": 3362, "epoch": 37, "lr": 9.997245363752592e-05}
{"train_loss": 1.1602632999420166, "global_step": 3363, "epoch": 37, "lr": 9.997243439292838e-05}
{"train_loss": 1.1425480842590332, "global_step": 3364, "epoch": 37, "lr": 9.997241514161264e-05}
{"train_loss": 1.147749423980713, "global_step": 3365, "epoch": 37, "lr": 9.997239588357872e-05}
{"train_loss": 1.0956705808639526, "global_step": 3366, "epoch": 37, "lr": 9.997237661882662e-05}
{"train_loss": 1.117663025856018, "global_step": 3367, "epoch": 37, "lr": 9.997235734735633e-05}
{"train_loss": 1.0853919982910156, "global_step": 3368, "epoch": 37, "lr": 9.997233806916789e-05}
{"train_loss": 1.0347373485565186, "global_step": 3369, "epoch": 37, "lr": 9.997231878426126e-05}
{"train_loss": 1.1876660585403442, "global_step": 3370, "epoch": 37, "lr": 9.997229949263647e-05}
{"train_loss": 0.8612529039382935, "global_step": 3371, "epoch": 37, "lr": 9.997228019429352e-05}
{"train_loss": 1.0721948146820068, "global_step": 3372, "epoch": 37, "lr": 9.997226088923239e-05}
{"train_loss": 1.067960500717163, "global_step": 3373, "epoch": 37, "lr": 9.997224157745311e-05}
{"train_loss": 1.069639801979065, "global_step": 3374, "epoch": 37, "lr": 9.997222225895566e-05}
{"train_loss": 0.938767671585083, "global_step": 3375, "epoch": 37, "lr": 9.997220293374008e-05}
{"train_loss": 1.0067691802978516, "global_step": 3376, "epoch": 37, "lr": 9.997218360180631e-05}
{"train_loss": 0.8844733834266663, "global_step": 3377, "epoch": 37, "lr": 9.997216426315441e-05}
{"train_loss": 0.9858359098434448, "global_step": 3378, "epoch": 37, "lr": 9.997214491778437e-05}
{"train_loss": 1.047741174697876, "global_step": 3379, "epoch": 37, "lr": 9.997212556569617e-05}
{"train_loss": 1.0698564052581787, "global_step": 3380, "epoch": 37, "lr": 9.997210620688982e-05}
{"train_loss": 1.0988525252663688, "global_step": 3381, "epoch": 37, "lr": 9.997208684136534e-05, "val_loss": 1.2364628314971924}
{"train_loss": 0.9902428984642029, "global_step": 3382, "epoch": 38, "lr": 9.997206746912273e-05}
{"train_loss": 1.1318542957305908, "global_step": 3383, "epoch": 38, "lr": 9.997204809016196e-05}
{"train_loss": 1.411530613899231, "global_step": 3384, "epoch": 38, "lr": 9.997202870448308e-05}
{"train_loss": 1.2552410364151, "global_step": 3385, "epoch": 38, "lr": 9.997200931208607e-05}
{"train_loss": 1.0599949359893799, "global_step": 3386, "epoch": 38, "lr": 9.997198991297093e-05}
{"train_loss": 1.2680317163467407, "global_step": 3387, "epoch": 38, "lr": 9.997197050713765e-05}
{"train_loss": 1.0385448932647705, "global_step": 3388, "epoch": 38, "lr": 9.997195109458626e-05}
{"train_loss": 1.087373971939087, "global_step": 3389, "epoch": 38, "lr": 9.997193167531674e-05}
{"train_loss": 1.2701056003570557, "global_step": 3390, "epoch": 38, "lr": 9.997191224932912e-05}
{"train_loss": 1.1795767545700073, "global_step": 3391, "epoch": 38, "lr": 9.997189281662338e-05}
{"train_loss": 1.33448326587677, "global_step": 3392, "epoch": 38, "lr": 9.997187337719951e-05}
{"train_loss": 1.1774982213974, "global_step": 3393, "epoch": 38, "lr": 9.997185393105755e-05}
{"train_loss": 1.3005224466323853, "global_step": 3394, "epoch": 38, "lr": 9.997183447819748e-05}
{"train_loss": 1.1996610164642334, "global_step": 3395, "epoch": 38, "lr": 9.99718150186193e-05}
{"train_loss": 1.4052373170852661, "global_step": 3396, "epoch": 38, "lr": 9.997179555232303e-05}
{"train_loss": 1.2683889865875244, "global_step": 3397, "epoch": 38, "lr": 9.997177607930866e-05}
{"train_loss": 1.261094093322754, "global_step": 3398, "epoch": 38, "lr": 9.997175659957619e-05}
{"train_loss": 0.932803213596344, "global_step": 3399, "epoch": 38, "lr": 9.997173711312562e-05}
{"train_loss": 1.3306338787078857, "global_step": 3400, "epoch": 38, "lr": 9.997171761995696e-05}
{"train_loss": 1.1204307079315186, "global_step": 3401, "epoch": 38, "lr": 9.997169812007021e-05}
{"train_loss": 1.3400155305862427, "global_step": 3402, "epoch": 38, "lr": 9.997167861346537e-05}
{"train_loss": 1.178667426109314, "global_step": 3403, "epoch": 38, "lr": 9.997165910014246e-05}
{"train_loss": 1.0744850635528564, "global_step": 3404, "epoch": 38, "lr": 9.997163958010147e-05}
{"train_loss": 1.1054939031600952, "global_step": 3405, "epoch": 38, "lr": 9.997162005334239e-05}
{"train_loss": 1.051318645477295, "global_step": 3406, "epoch": 38, "lr": 9.997160051986525e-05}
{"train_loss": 1.0709282159805298, "global_step": 3407, "epoch": 38, "lr": 9.997158097967004e-05}
{"train_loss": 1.3021008968353271, "global_step": 3408, "epoch": 38, "lr": 9.997156143275674e-05}
{"train_loss": 1.1508009433746338, "global_step": 3409, "epoch": 38, "lr": 9.997154187912539e-05}
{"train_loss": 1.1461787223815918, "global_step": 3410, "epoch": 38, "lr": 9.997152231877595e-05}
{"train_loss": 1.0638443231582642, "global_step": 3411, "epoch": 38, "lr": 9.997150275170848e-05}
{"train_loss": 1.1524016857147217, "global_step": 3412, "epoch": 38, "lr": 9.997148317792293e-05}
{"train_loss": 1.128699779510498, "global_step": 3413, "epoch": 38, "lr": 9.997146359741932e-05}
{"train_loss": 1.345946192741394, "global_step": 3414, "epoch": 38, "lr": 9.997144401019766e-05}
{"train_loss": 1.3005298376083374, "global_step": 3415, "epoch": 38, "lr": 9.997142441625797e-05}
{"train_loss": 1.2650073766708374, "global_step": 3416, "epoch": 38, "lr": 9.997140481560021e-05}
{"train_loss": 1.2291655540466309, "global_step": 3417, "epoch": 38, "lr": 9.99713852082244e-05}
{"train_loss": 1.0747803449630737, "global_step": 3418, "epoch": 38, "lr": 9.997136559413057e-05}
{"train_loss": 1.1235977411270142, "global_step": 3419, "epoch": 38, "lr": 9.997134597331867e-05}
{"train_loss": 0.9943549633026123, "global_step": 3420, "epoch": 38, "lr": 9.997132634578874e-05}
{"train_loss": 1.1052219867706299, "global_step": 3421, "epoch": 38, "lr": 9.997130671154078e-05}
{"train_loss": 1.1530154943466187, "global_step": 3422, "epoch": 38, "lr": 9.99712870705748e-05}
{"train_loss": 1.130112886428833, "global_step": 3423, "epoch": 38, "lr": 9.997126742289077e-05}
{"train_loss": 1.1604533195495605, "global_step": 3424, "epoch": 38, "lr": 9.997124776848872e-05}
{"train_loss": 1.1588765382766724, "global_step": 3425, "epoch": 38, "lr": 9.997122810736863e-05}
{"train_loss": 0.9859888553619385, "global_step": 3426, "epoch": 38, "lr": 9.997120843953054e-05}
{"train_loss": 0.9595518112182617, "global_step": 3427, "epoch": 38, "lr": 9.997118876497442e-05}
{"train_loss": 0.9123858213424683, "global_step": 3428, "epoch": 38, "lr": 9.997116908370028e-05}
{"train_loss": 1.1031715869903564, "global_step": 3429, "epoch": 38, "lr": 9.997114939570813e-05}
{"train_loss": 1.023657202720642, "global_step": 3430, "epoch": 38, "lr": 9.997112970099796e-05}
{"train_loss": 1.276641845703125, "global_step": 3431, "epoch": 38, "lr": 9.997110999956979e-05}
{"train_loss": 1.2431821823120117, "global_step": 3432, "epoch": 38, "lr": 9.997109029142363e-05}
{"train_loss": 1.1791224479675293, "global_step": 3433, "epoch": 38, "lr": 9.997107057655943e-05}
{"train_loss": 0.9576103687286377, "global_step": 3434, "epoch": 38, "lr": 9.997105085497725e-05}
{"train_loss": 1.1534433364868164, "global_step": 3435, "epoch": 38, "lr": 9.997103112667705e-05}
{"train_loss": 1.1645123958587646, "global_step": 3436, "epoch": 38, "lr": 9.997101139165888e-05}
{"train_loss": 0.919157087802887, "global_step": 3437, "epoch": 38, "lr": 9.997099164992269e-05}
{"train_loss": 1.0671032667160034, "global_step": 3438, "epoch": 38, "lr": 9.997097190146853e-05}
{"train_loss": 1.037733554840088, "global_step": 3439, "epoch": 38, "lr": 9.99709521462964e-05}
{"train_loss": 0.9984281063079834, "global_step": 3440, "epoch": 38, "lr": 9.997093238440625e-05}
{"train_loss": 0.8917558789253235, "global_step": 3441, "epoch": 38, "lr": 9.997091261579812e-05}
{"train_loss": 1.0284337997436523, "global_step": 3442, "epoch": 38, "lr": 9.997089284047203e-05}
{"train_loss": 1.1227223873138428, "global_step": 3443, "epoch": 38, "lr": 9.997087305842795e-05}
{"train_loss": 1.2563185691833496, "global_step": 3444, "epoch": 38, "lr": 9.99708532696659e-05}
{"train_loss": 0.9551241397857666, "global_step": 3445, "epoch": 38, "lr": 9.997083347418587e-05}
{"train_loss": 1.1683526039123535, "global_step": 3446, "epoch": 38, "lr": 9.997081367198788e-05}
{"train_loss": 1.017069935798645, "global_step": 3447, "epoch": 38, "lr": 9.997079386307191e-05}
{"train_loss": 1.26876699924469, "global_step": 3448, "epoch": 38, "lr": 9.997077404743798e-05}
{"train_loss": 0.9444539546966553, "global_step": 3449, "epoch": 38, "lr": 9.99707542250861e-05}
{"train_loss": 1.0730427503585815, "global_step": 3450, "epoch": 38, "lr": 9.997073439601626e-05}
{"train_loss": 1.193527102470398, "global_step": 3451, "epoch": 38, "lr": 9.997071456022844e-05}
{"train_loss": 0.9836331605911255, "global_step": 3452, "epoch": 38, "lr": 9.997069471772269e-05}
{"train_loss": 1.1841107606887817, "global_step": 3453, "epoch": 38, "lr": 9.9970674868499e-05}
{"train_loss": 1.0098519325256348, "global_step": 3454, "epoch": 38, "lr": 9.997065501255733e-05}
{"train_loss": 1.3374730348587036, "global_step": 3455, "epoch": 38, "lr": 9.997063514989773e-05}
{"train_loss": 1.2016428709030151, "global_step": 3456, "epoch": 38, "lr": 9.997061528052018e-05}
{"train_loss": 1.096451997756958, "global_step": 3457, "epoch": 38, "lr": 9.99705954044247e-05}
{"train_loss": 1.103076696395874, "global_step": 3458, "epoch": 38, "lr": 9.997057552161129e-05}
{"train_loss": 0.9380964636802673, "global_step": 3459, "epoch": 38, "lr": 9.997055563207993e-05}
{"train_loss": 1.0308812856674194, "global_step": 3460, "epoch": 38, "lr": 9.997053573583064e-05}
{"train_loss": 0.8974012136459351, "global_step": 3461, "epoch": 38, "lr": 9.997051583286344e-05}
{"train_loss": 1.1442137956619263, "global_step": 3462, "epoch": 38, "lr": 9.997049592317829e-05}
{"train_loss": 1.2126671075820923, "global_step": 3463, "epoch": 38, "lr": 9.997047600677522e-05}
{"train_loss": 1.0598876476287842, "global_step": 3464, "epoch": 38, "lr": 9.997045608365423e-05}
{"train_loss": 1.2047145366668701, "global_step": 3465, "epoch": 38, "lr": 9.997043615381531e-05}
{"train_loss": 1.0631276369094849, "global_step": 3466, "epoch": 38, "lr": 9.997041621725849e-05}
{"train_loss": 1.1006230115890503, "global_step": 3467, "epoch": 38, "lr": 9.997039627398374e-05}
{"train_loss": 1.271533489227295, "global_step": 3468, "epoch": 38, "lr": 9.997037632399111e-05}
{"train_loss": 0.9820263385772705, "global_step": 3469, "epoch": 38, "lr": 9.997035636728055e-05}
{"train_loss": 1.1297245742229933, "global_step": 3470, "epoch": 38, "lr": 9.997033640385209e-05, "val_loss": 1.1732614040374756}
{"train_loss": 1.1333200931549072, "global_step": 3471, "epoch": 39, "lr": 9.997031643370572e-05}
{"train_loss": 1.0745067596435547, "global_step": 3472, "epoch": 39, "lr": 9.997029645684146e-05}
{"train_loss": 1.1565262079238892, "global_step": 3473, "epoch": 39, "lr": 9.997027647325929e-05}
{"train_loss": 1.2118011713027954, "global_step": 3474, "epoch": 39, "lr": 9.997025648295924e-05}
{"train_loss": 1.1332803964614868, "global_step": 3475, "epoch": 39, "lr": 9.997023648594128e-05}
{"train_loss": 1.06400465965271, "global_step": 3476, "epoch": 39, "lr": 9.997021648220545e-05}
{"train_loss": 0.9831113219261169, "global_step": 3477, "epoch": 39, "lr": 9.997019647175172e-05}
{"train_loss": 1.077246069908142, "global_step": 3478, "epoch": 39, "lr": 9.997017645458011e-05}
{"train_loss": 0.9981310367584229, "global_step": 3479, "epoch": 39, "lr": 9.997015643069063e-05}
{"train_loss": 0.9942050576210022, "global_step": 3480, "epoch": 39, "lr": 9.997013640008327e-05}
{"train_loss": 1.230687141418457, "global_step": 3481, "epoch": 39, "lr": 9.997011636275802e-05}
{"train_loss": 1.092860221862793, "global_step": 3482, "epoch": 39, "lr": 9.99700963187149e-05}
{"train_loss": 1.1389292478561401, "global_step": 3483, "epoch": 39, "lr": 9.997007626795391e-05}
{"train_loss": 1.3103232383728027, "global_step": 3484, "epoch": 39, "lr": 9.997005621047506e-05}
{"train_loss": 0.9502546191215515, "global_step": 3485, "epoch": 39, "lr": 9.997003614627833e-05}
{"train_loss": 0.9168651700019836, "global_step": 3486, "epoch": 39, "lr": 9.997001607536376e-05}
{"train_loss": 1.2297569513320923, "global_step": 3487, "epoch": 39, "lr": 9.996999599773132e-05}
{"train_loss": 1.0359998941421509, "global_step": 3488, "epoch": 39, "lr": 9.996997591338102e-05}
{"train_loss": 1.1965739727020264, "global_step": 3489, "epoch": 39, "lr": 9.996995582231285e-05}
{"train_loss": 1.3883376121520996, "global_step": 3490, "epoch": 39, "lr": 9.996993572452687e-05}
{"train_loss": 1.0646891593933105, "global_step": 3491, "epoch": 39, "lr": 9.9969915620023e-05}
{"train_loss": 1.2075012922286987, "global_step": 3492, "epoch": 39, "lr": 9.99698955088013e-05}
{"train_loss": 1.2214698791503906, "global_step": 3493, "epoch": 39, "lr": 9.996987539086176e-05}
{"train_loss": 1.0415499210357666, "global_step": 3494, "epoch": 39, "lr": 9.996985526620437e-05}
{"train_loss": 1.031543254852295, "global_step": 3495, "epoch": 39, "lr": 9.996983513482916e-05}
{"train_loss": 1.0729594230651855, "global_step": 3496, "epoch": 39, "lr": 9.99698149967361e-05}
{"train_loss": 1.1185084581375122, "global_step": 3497, "epoch": 39, "lr": 9.996979485192521e-05}
{"train_loss": 1.0362452268600464, "global_step": 3498, "epoch": 39, "lr": 9.996977470039649e-05}
{"train_loss": 1.073422908782959, "global_step": 3499, "epoch": 39, "lr": 9.996975454214994e-05}
{"train_loss": 1.0845016241073608, "global_step": 3500, "epoch": 39, "lr": 9.996973437718557e-05}
{"train_loss": 0.9378294944763184, "global_step": 3501, "epoch": 39, "lr": 9.996971420550339e-05}
{"train_loss": 1.038099765777588, "global_step": 3502, "epoch": 39, "lr": 9.996969402710337e-05}
{"train_loss": 1.069181203842163, "global_step": 3503, "epoch": 39, "lr": 9.996967384198555e-05}
{"train_loss": 1.0399901866912842, "global_step": 3504, "epoch": 39, "lr": 9.99696536501499e-05}
{"train_loss": 1.194210410118103, "global_step": 3505, "epoch": 39, "lr": 9.996963345159644e-05}
{"train_loss": 0.9484338164329529, "global_step": 3506, "epoch": 39, "lr": 9.996961324632519e-05}
{"train_loss": 1.080367088317871, "global_step": 3507, "epoch": 39, "lr": 9.996959303433612e-05}
{"train_loss": 0.9620457887649536, "global_step": 3508, "epoch": 39, "lr": 9.996957281562926e-05}
{"train_loss": 1.0912574529647827, "global_step": 3509, "epoch": 39, "lr": 9.996955259020457e-05}
{"train_loss": 1.0685222148895264, "global_step": 3510, "epoch": 39, "lr": 9.996953235806212e-05}
{"train_loss": 1.0466015338897705, "global_step": 3511, "epoch": 39, "lr": 9.996951211920184e-05}
{"train_loss": 1.1368293762207031, "global_step": 3512, "epoch": 39, "lr": 9.99694918736238e-05}
{"train_loss": 1.108880877494812, "global_step": 3513, "epoch": 39, "lr": 9.996947162132796e-05}
{"train_loss": 1.1818982362747192, "global_step": 3514, "epoch": 39, "lr": 9.996945136231432e-05}
{"train_loss": 1.147336483001709, "global_step": 3515, "epoch": 39, "lr": 9.996943109658292e-05}
{"train_loss": 0.934319257736206, "global_step": 3516, "epoch": 39, "lr": 9.996941082413374e-05}
{"train_loss": 1.1287404298782349, "global_step": 3517, "epoch": 39, "lr": 9.996939054496676e-05}
{"train_loss": 1.0038702487945557, "global_step": 3518, "epoch": 39, "lr": 9.996937025908201e-05}
{"train_loss": 1.0569725036621094, "global_step": 3519, "epoch": 39, "lr": 9.996934996647949e-05}
{"train_loss": 1.0584583282470703, "global_step": 3520, "epoch": 39, "lr": 9.996932966715919e-05}
{"train_loss": 1.1046806573867798, "global_step": 3521, "epoch": 39, "lr": 9.996930936112114e-05}
{"train_loss": 1.0679473876953125, "global_step": 3522, "epoch": 39, "lr": 9.996928904836533e-05}
{"train_loss": 1.2420525550842285, "global_step": 3523, "epoch": 39, "lr": 9.996926872889174e-05}
{"train_loss": 0.8778681755065918, "global_step": 3524, "epoch": 39, "lr": 9.99692484027004e-05}
{"train_loss": 1.0500209331512451, "global_step": 3525, "epoch": 39, "lr": 9.99692280697913e-05}
{"train_loss": 1.1354538202285767, "global_step": 3526, "epoch": 39, "lr": 9.996920773016445e-05}
{"train_loss": 0.98294597864151, "global_step": 3527, "epoch": 39, "lr": 9.996918738381984e-05}
{"train_loss": 1.081898808479309, "global_step": 3528, "epoch": 39, "lr": 9.996916703075749e-05}
{"train_loss": 1.1583276987075806, "global_step": 3529, "epoch": 39, "lr": 9.996914667097738e-05}
{"train_loss": 0.9057188034057617, "global_step": 3530, "epoch": 39, "lr": 9.996912630447955e-05}
{"train_loss": 0.9615591168403625, "global_step": 3531, "epoch": 39, "lr": 9.996910593126397e-05}
{"train_loss": 1.1914058923721313, "global_step": 3532, "epoch": 39, "lr": 9.996908555133066e-05}
{"train_loss": 0.9853142499923706, "global_step": 3533, "epoch": 39, "lr": 9.99690651646796e-05}
{"train_loss": 1.0966154336929321, "global_step": 3534, "epoch": 39, "lr": 9.996904477131082e-05}
{"train_loss": 1.2012038230895996, "global_step": 3535, "epoch": 39, "lr": 9.996902437122431e-05}
{"train_loss": 0.8299131989479065, "global_step": 3536, "epoch": 39, "lr": 9.996900396442008e-05}
{"train_loss": 0.9752130508422852, "global_step": 3537, "epoch": 39, "lr": 9.99689835508981e-05}
{"train_loss": 0.9688066840171814, "global_step": 3538, "epoch": 39, "lr": 9.996896313065843e-05}
{"train_loss": 0.9073764085769653, "global_step": 3539, "epoch": 39, "lr": 9.996894270370103e-05}
{"train_loss": 1.1622610092163086, "global_step": 3540, "epoch": 39, "lr": 9.996892227002591e-05}
{"train_loss": 0.878453254699707, "global_step": 3541, "epoch": 39, "lr": 9.996890182963308e-05}
{"train_loss": 1.0495322942733765, "global_step": 3542, "epoch": 39, "lr": 9.996888138252254e-05}
{"train_loss": 1.1430895328521729, "global_step": 3543, "epoch": 39, "lr": 9.996886092869429e-05}
{"train_loss": 1.2729268074035645, "global_step": 3544, "epoch": 39, "lr": 9.996884046814834e-05}
{"train_loss": 0.9665083885192871, "global_step": 3545, "epoch": 39, "lr": 9.99688200008847e-05}
{"train_loss": 1.1394046545028687, "global_step": 3546, "epoch": 39, "lr": 9.996879952690334e-05}
{"train_loss": 1.2451971769332886, "global_step": 3547, "epoch": 39, "lr": 9.99687790462043e-05}
{"train_loss": 1.0222852230072021, "global_step": 3548, "epoch": 39, "lr": 9.996875855878755e-05}
{"train_loss": 0.9659056663513184, "global_step": 3549, "epoch": 39, "lr": 9.996873806465312e-05}
{"train_loss": 0.9278349876403809, "global_step": 3550, "epoch": 39, "lr": 9.996871756380101e-05}
{"train_loss": 0.9842657446861267, "global_step": 3551, "epoch": 39, "lr": 9.996869705623119e-05}
{"train_loss": 1.0302462577819824, "global_step": 3552, "epoch": 39, "lr": 9.996867654194371e-05}
{"train_loss": 1.1877448558807373, "global_step": 3553, "epoch": 39, "lr": 9.996865602093856e-05}
{"train_loss": 1.074449896812439, "global_step": 3554, "epoch": 39, "lr": 9.996863549321571e-05}
{"train_loss": 1.372443675994873, "global_step": 3555, "epoch": 39, "lr": 9.99686149587752e-05}
{"train_loss": 1.332042932510376, "global_step": 3556, "epoch": 39, "lr": 9.996859441761702e-05}
{"train_loss": 1.006798505783081, "global_step": 3557, "epoch": 39, "lr": 9.996857386974114e-05}
{"train_loss": 1.0224583148956299, "global_step": 3558, "epoch": 39, "lr": 9.996855331514762e-05}
{"train_loss": 1.079247986332754, "global_step": 3559, "epoch": 39, "lr": 9.996853275383644e-05, "val_loss": 1.159305453300476}
{"train_loss": 1.0794618129730225, "global_step": 3560, "epoch": 40, "lr": 9.99685121858076e-05}
{"train_loss": 1.007009744644165, "global_step": 3561, "epoch": 40, "lr": 9.99684916110611e-05}
{"train_loss": 1.111933946609497, "global_step": 3562, "epoch": 40, "lr": 9.996847102959695e-05}
{"train_loss": 1.099793553352356, "global_step": 3563, "epoch": 40, "lr": 9.996845044141514e-05}
{"train_loss": 1.1205743551254272, "global_step": 3564, "epoch": 40, "lr": 9.996842984651567e-05}
{"train_loss": 1.0222212076187134, "global_step": 3565, "epoch": 40, "lr": 9.996840924489858e-05}
{"train_loss": 0.8656809329986572, "global_step": 3566, "epoch": 40, "lr": 9.996838863656382e-05}
{"train_loss": 0.9918510913848877, "global_step": 3567, "epoch": 40, "lr": 9.996836802151143e-05}
{"train_loss": 1.1383397579193115, "global_step": 3568, "epoch": 40, "lr": 9.996834739974141e-05}
{"train_loss": 1.017961025238037, "global_step": 3569, "epoch": 40, "lr": 9.996832677125375e-05}
{"train_loss": 1.158164143562317, "global_step": 3570, "epoch": 40, "lr": 9.996830613604847e-05}
{"train_loss": 1.0972919464111328, "global_step": 3571, "epoch": 40, "lr": 9.996828549412554e-05}
{"train_loss": 1.0690662860870361, "global_step": 3572, "epoch": 40, "lr": 9.996826484548498e-05}
{"train_loss": 1.2184302806854248, "global_step": 3573, "epoch": 40, "lr": 9.996824419012681e-05}
{"train_loss": 1.0602095127105713, "global_step": 3574, "epoch": 40, "lr": 9.996822352805101e-05}
{"train_loss": 1.2712513208389282, "global_step": 3575, "epoch": 40, "lr": 9.996820285925759e-05}
{"train_loss": 1.1520925760269165, "global_step": 3576, "epoch": 40, "lr": 9.996818218374657e-05}
{"train_loss": 1.0990604162216187, "global_step": 3577, "epoch": 40, "lr": 9.996816150151792e-05}
{"train_loss": 0.8481053113937378, "global_step": 3578, "epoch": 40, "lr": 9.996814081257166e-05}
{"train_loss": 0.9193106293678284, "global_step": 3579, "epoch": 40, "lr": 9.996812011690781e-05}
{"train_loss": 1.1166870594024658, "global_step": 3580, "epoch": 40, "lr": 9.996809941452634e-05}
{"train_loss": 0.9162668585777283, "global_step": 3581, "epoch": 40, "lr": 9.996807870542726e-05}
{"train_loss": 1.2798296213150024, "global_step": 3582, "epoch": 40, "lr": 9.996805798961059e-05}
{"train_loss": 1.0272475481033325, "global_step": 3583, "epoch": 40, "lr": 9.996803726707634e-05}
{"train_loss": 1.1526892185211182, "global_step": 3584, "epoch": 40, "lr": 9.996801653782448e-05}
{"train_loss": 1.0511305332183838, "global_step": 3585, "epoch": 40, "lr": 9.996799580185502e-05}
{"train_loss": 1.0972511768341064, "global_step": 3586, "epoch": 40, "lr": 9.9967975059168e-05}
{"train_loss": 1.137374758720398, "global_step": 3587, "epoch": 40, "lr": 9.996795430976337e-05}
{"train_loss": 0.9960055947303772, "global_step": 3588, "epoch": 40, "lr": 9.996793355364117e-05}
{"train_loss": 1.3355849981307983, "global_step": 3589, "epoch": 40, "lr": 9.996791279080137e-05}
{"train_loss": 1.254326343536377, "global_step": 3590, "epoch": 40, "lr": 9.996789202124402e-05}
{"train_loss": 1.167776346206665, "global_step": 3591, "epoch": 40, "lr": 9.996787124496909e-05}
{"train_loss": 1.1086915731430054, "global_step": 3592, "epoch": 40, "lr": 9.996785046197659e-05}
{"train_loss": 1.1160022020339966, "global_step": 3593, "epoch": 40, "lr": 9.99678296722665e-05}
{"train_loss": 1.2800356149673462, "global_step": 3594, "epoch": 40, "lr": 9.996780887583888e-05}
{"train_loss": 1.0002516508102417, "global_step": 3595, "epoch": 40, "lr": 9.996778807269366e-05}
{"train_loss": 1.0433769226074219, "global_step": 3596, "epoch": 40, "lr": 9.99677672628309e-05}
{"train_loss": 0.9256530404090881, "global_step": 3597, "epoch": 40, "lr": 9.996774644625058e-05}
{"train_loss": 1.414772391319275, "global_step": 3598, "epoch": 40, "lr": 9.996772562295272e-05}
{"train_loss": 1.022108554840088, "global_step": 3599, "epoch": 40, "lr": 9.996770479293728e-05}
{"train_loss": 1.0656472444534302, "global_step": 3600, "epoch": 40, "lr": 9.996768395620431e-05}
{"train_loss": 1.1237132549285889, "global_step": 3601, "epoch": 40, "lr": 9.996766311275378e-05}
{"train_loss": 1.1043968200683594, "global_step": 3602, "epoch": 40, "lr": 9.996764226258573e-05}
{"train_loss": 1.0528348684310913, "global_step": 3603, "epoch": 40, "lr": 9.99676214057001e-05}
{"train_loss": 1.1453955173492432, "global_step": 3604, "epoch": 40, "lr": 9.996760054209697e-05}
{"train_loss": 0.9501065611839294, "global_step": 3605, "epoch": 40, "lr": 9.996757967177629e-05}
{"train_loss": 0.9010099768638611, "global_step": 3606, "epoch": 40, "lr": 9.996755879473807e-05}
{"train_loss": 1.0452848672866821, "global_step": 3607, "epoch": 40, "lr": 9.996753791098234e-05}
{"train_loss": 1.2893173694610596, "global_step": 3608, "epoch": 40, "lr": 9.996751702050909e-05}
{"train_loss": 0.838202714920044, "global_step": 3609, "epoch": 40, "lr": 9.996749612331828e-05}
{"train_loss": 0.9525585174560547, "global_step": 3610, "epoch": 40, "lr": 9.996747521940998e-05}
{"train_loss": 1.0385233163833618, "global_step": 3611, "epoch": 40, "lr": 9.996745430878414e-05}
{"train_loss": 1.1690133810043335, "global_step": 3612, "epoch": 40, "lr": 9.996743339144082e-05}
{"train_loss": 0.9798354506492615, "global_step": 3613, "epoch": 40, "lr": 9.996741246737995e-05}
{"train_loss": 1.0998327732086182, "global_step": 3614, "epoch": 40, "lr": 9.996739153660159e-05}
{"train_loss": 0.9124060869216919, "global_step": 3615, "epoch": 40, "lr": 9.996737059910571e-05}
{"train_loss": 0.9784793257713318, "global_step": 3616, "epoch": 40, "lr": 9.996734965489233e-05}
{"train_loss": 1.0338746309280396, "global_step": 3617, "epoch": 40, "lr": 9.996732870396146e-05}
{"train_loss": 1.0484014749526978, "global_step": 3618, "epoch": 40, "lr": 9.996730774631309e-05}
{"train_loss": 0.9326826930046082, "global_step": 3619, "epoch": 40, "lr": 9.996728678194722e-05}
{"train_loss": 1.052855372428894, "global_step": 3620, "epoch": 40, "lr": 9.996726581086385e-05}
{"train_loss": 0.9614386558532715, "global_step": 3621, "epoch": 40, "lr": 9.996724483306299e-05}
{"train_loss": 0.8444228768348694, "global_step": 3622, "epoch": 40, "lr": 9.996722384854465e-05}
{"train_loss": 1.020053744316101, "global_step": 3623, "epoch": 40, "lr": 9.996720285730883e-05}
{"train_loss": 1.064158320426941, "global_step": 3624, "epoch": 40, "lr": 9.996718185935552e-05}
{"train_loss": 0.9169725179672241, "global_step": 3625, "epoch": 40, "lr": 9.996716085468474e-05}
{"train_loss": 1.1702004671096802, "global_step": 3626, "epoch": 40, "lr": 9.996713984329649e-05}
{"train_loss": 0.9556410312652588, "global_step": 3627, "epoch": 40, "lr": 9.996711882519074e-05}
{"train_loss": 1.296345591545105, "global_step": 3628, "epoch": 40, "lr": 9.996709780036755e-05}
{"train_loss": 1.2268582582473755, "global_step": 3629, "epoch": 40, "lr": 9.996707676882688e-05}
{"train_loss": 0.9650378227233887, "global_step": 3630, "epoch": 40, "lr": 9.996705573056873e-05}
{"train_loss": 1.08505117893219, "global_step": 3631, "epoch": 40, "lr": 9.996703468559314e-05}
{"train_loss": 0.8995487093925476, "global_step": 3632, "epoch": 40, "lr": 9.996701363390008e-05}
{"train_loss": 1.1371099948883057, "global_step": 3633, "epoch": 40, "lr": 9.996699257548957e-05}
{"train_loss": 1.0767947435379028, "global_step": 3634, "epoch": 40, "lr": 9.99669715103616e-05}
{"train_loss": 1.0831472873687744, "global_step": 3635, "epoch": 40, "lr": 9.996695043851619e-05}
{"train_loss": 1.2339805364608765, "global_step": 3636, "epoch": 40, "lr": 9.996692935995333e-05}
{"train_loss": 1.1508177518844604, "global_step": 3637, "epoch": 40, "lr": 9.996690827467303e-05}
{"train_loss": 1.0050841569900513, "global_step": 3638, "epoch": 40, "lr": 9.996688718267528e-05}
{"train_loss": 1.2817630767822266, "global_step": 3639, "epoch": 40, "lr": 9.996686608396008e-05}
{"train_loss": 1.2412811517715454, "global_step": 3640, "epoch": 40, "lr": 9.996684497852747e-05}
{"train_loss": 1.0028393268585205, "global_step": 3641, "epoch": 40, "lr": 9.996682386637741e-05}
{"train_loss": 0.862061619758606, "global_step": 3642, "epoch": 40, "lr": 9.996680274750992e-05}
{"train_loss": 1.226253628730774, "global_step": 3643, "epoch": 40, "lr": 9.9966781621925e-05}
{"train_loss": 0.9439100623130798, "global_step": 3644, "epoch": 40, "lr": 9.996676048962266e-05}
{"train_loss": 0.9255433082580566, "global_step": 3645, "epoch": 40, "lr": 9.99667393506029e-05}
{"train_loss": 1.0828739404678345, "global_step": 3646, "epoch": 40, "lr": 9.996671820486572e-05}
{"train_loss": 1.2206863164901733, "global_step": 3647, "epoch": 40, "lr": 9.996669705241112e-05}
{"train_loss": 1.07375040750825, "global_step": 3648, "epoch": 40, "lr": 9.996667589323911e-05, "val_loss": 1.181134581565857, "train_action_mse_error": 57.39595031738281}
{"train_loss": 1.000124454498291, "global_step": 3649, "epoch": 41, "lr": 9.996665472734969e-05}
{"train_loss": 1.3559902906417847, "global_step": 3650, "epoch": 41, "lr": 9.996663355474285e-05}
{"train_loss": 1.115813136100769, "global_step": 3651, "epoch": 41, "lr": 9.996661237541863e-05}
{"train_loss": 1.062555193901062, "global_step": 3652, "epoch": 41, "lr": 9.996659118937699e-05}
{"train_loss": 1.2779433727264404, "global_step": 3653, "epoch": 41, "lr": 9.996656999661795e-05}
{"train_loss": 1.0695587396621704, "global_step": 3654, "epoch": 41, "lr": 9.996654879714152e-05}
{"train_loss": 1.0631630420684814, "global_step": 3655, "epoch": 41, "lr": 9.99665275909477e-05}
{"train_loss": 0.8194300532341003, "global_step": 3656, "epoch": 41, "lr": 9.996650637803648e-05}
{"train_loss": 0.8925865292549133, "global_step": 3657, "epoch": 41, "lr": 9.996648515840787e-05}
{"train_loss": 0.976814329624176, "global_step": 3658, "epoch": 41, "lr": 9.996646393206189e-05}
{"train_loss": 1.0562353134155273, "global_step": 3659, "epoch": 41, "lr": 9.996644269899851e-05}
{"train_loss": 1.1669450998306274, "global_step": 3660, "epoch": 41, "lr": 9.996642145921776e-05}
{"train_loss": 1.174463152885437, "global_step": 3661, "epoch": 41, "lr": 9.996640021271964e-05}
{"train_loss": 0.8677725791931152, "global_step": 3662, "epoch": 41, "lr": 9.996637895950414e-05}
{"train_loss": 1.059537649154663, "global_step": 3663, "epoch": 41, "lr": 9.996635769957127e-05}
{"train_loss": 1.2040003538131714, "global_step": 3664, "epoch": 41, "lr": 9.996633643292103e-05}
{"train_loss": 0.9782583117485046, "global_step": 3665, "epoch": 41, "lr": 9.996631515955343e-05}
{"train_loss": 1.0348773002624512, "global_step": 3666, "epoch": 41, "lr": 9.996629387946845e-05}
{"train_loss": 1.3543946743011475, "global_step": 3667, "epoch": 41, "lr": 9.996627259266613e-05}
{"train_loss": 0.9705684781074524, "global_step": 3668, "epoch": 41, "lr": 9.996625129914645e-05}
{"train_loss": 1.1849910020828247, "global_step": 3669, "epoch": 41, "lr": 9.996622999890942e-05}
{"train_loss": 1.1817891597747803, "global_step": 3670, "epoch": 41, "lr": 9.996620869195503e-05}
{"train_loss": 1.2333712577819824, "global_step": 3671, "epoch": 41, "lr": 9.99661873782833e-05}
{"train_loss": 0.9861034750938416, "global_step": 3672, "epoch": 41, "lr": 9.996616605789421e-05}
{"train_loss": 1.0399069786071777, "global_step": 3673, "epoch": 41, "lr": 9.99661447307878e-05}
{"train_loss": 0.9993600249290466, "global_step": 3674, "epoch": 41, "lr": 9.996612339696405e-05}
{"train_loss": 1.0523419380187988, "global_step": 3675, "epoch": 41, "lr": 9.996610205642295e-05}
{"train_loss": 0.9556635022163391, "global_step": 3676, "epoch": 41, "lr": 9.996608070916453e-05}
{"train_loss": 1.1668061017990112, "global_step": 3677, "epoch": 41, "lr": 9.996605935518877e-05}
{"train_loss": 1.4740917682647705, "global_step": 3678, "epoch": 41, "lr": 9.996603799449568e-05}
{"train_loss": 1.1062289476394653, "global_step": 3679, "epoch": 41, "lr": 9.996601662708527e-05}
{"train_loss": 1.2341952323913574, "global_step": 3680, "epoch": 41, "lr": 9.996599525295753e-05}
{"train_loss": 1.056784749031067, "global_step": 3681, "epoch": 41, "lr": 9.996597387211248e-05}
{"train_loss": 1.0239009857177734, "global_step": 3682, "epoch": 41, "lr": 9.996595248455012e-05}
{"train_loss": 0.9948903918266296, "global_step": 3683, "epoch": 41, "lr": 9.996593109027043e-05}
{"train_loss": 1.149597406387329, "global_step": 3684, "epoch": 41, "lr": 9.996590968927344e-05}
{"train_loss": 0.9904963374137878, "global_step": 3685, "epoch": 41, "lr": 9.996588828155914e-05}
{"train_loss": 0.9051831364631653, "global_step": 3686, "epoch": 41, "lr": 9.996586686712753e-05}
{"train_loss": 1.1881192922592163, "global_step": 3687, "epoch": 41, "lr": 9.996584544597863e-05}
{"train_loss": 1.1444690227508545, "global_step": 3688, "epoch": 41, "lr": 9.996582401811242e-05}
{"train_loss": 1.0957629680633545, "global_step": 3689, "epoch": 41, "lr": 9.996580258352893e-05}
{"train_loss": 1.200813889503479, "global_step": 3690, "epoch": 41, "lr": 9.996578114222813e-05}
{"train_loss": 1.1262506246566772, "global_step": 3691, "epoch": 41, "lr": 9.996575969421004e-05}
{"train_loss": 1.1110897064208984, "global_step": 3692, "epoch": 41, "lr": 9.996573823947466e-05}
{"train_loss": 1.0585088729858398, "global_step": 3693, "epoch": 41, "lr": 9.996571677802201e-05}
{"train_loss": 0.9955161213874817, "global_step": 3694, "epoch": 41, "lr": 9.996569530985208e-05}
{"train_loss": 0.9697607159614563, "global_step": 3695, "epoch": 41, "lr": 9.996567383496486e-05}
{"train_loss": 1.1525367498397827, "global_step": 3696, "epoch": 41, "lr": 9.996565235336036e-05}
{"train_loss": 1.134251594543457, "global_step": 3697, "epoch": 41, "lr": 9.996563086503859e-05}
{"train_loss": 1.1738489866256714, "global_step": 3698, "epoch": 41, "lr": 9.996560936999955e-05}
{"train_loss": 1.09445321559906, "global_step": 3699, "epoch": 41, "lr": 9.996558786824326e-05}
{"train_loss": 1.0176271200180054, "global_step": 3700, "epoch": 41, "lr": 9.996556635976969e-05}
{"train_loss": 1.0165706872940063, "global_step": 3701, "epoch": 41, "lr": 9.996554484457886e-05}
{"train_loss": 1.071258306503296, "global_step": 3702, "epoch": 41, "lr": 9.996552332267077e-05}
{"train_loss": 1.0666325092315674, "global_step": 3703, "epoch": 41, "lr": 9.996550179404543e-05}
{"train_loss": 1.0477123260498047, "global_step": 3704, "epoch": 41, "lr": 9.996548025870284e-05}
{"train_loss": 0.9677273035049438, "global_step": 3705, "epoch": 41, "lr": 9.9965458716643e-05}
{"train_loss": 0.9589359164237976, "global_step": 3706, "epoch": 41, "lr": 9.99654371678659e-05}
{"train_loss": 1.0936155319213867, "global_step": 3707, "epoch": 41, "lr": 9.996541561237155e-05}
{"train_loss": 1.2215807437896729, "global_step": 3708, "epoch": 41, "lr": 9.996539405015998e-05}
{"train_loss": 1.0060536861419678, "global_step": 3709, "epoch": 41, "lr": 9.996537248123117e-05}
{"train_loss": 0.974461019039154, "global_step": 3710, "epoch": 41, "lr": 9.996535090558512e-05}
{"train_loss": 0.9996488094329834, "global_step": 3711, "epoch": 41, "lr": 9.996532932322182e-05}
{"train_loss": 0.9089656472206116, "global_step": 3712, "epoch": 41, "lr": 9.996530773414132e-05}
{"train_loss": 1.1002976894378662, "global_step": 3713, "epoch": 41, "lr": 9.996528613834358e-05}
{"train_loss": 0.9607870578765869, "global_step": 3714, "epoch": 41, "lr": 9.996526453582862e-05}
{"train_loss": 1.189737319946289, "global_step": 3715, "epoch": 41, "lr": 9.996524292659644e-05}
{"train_loss": 1.109605312347412, "global_step": 3716, "epoch": 41, "lr": 9.996522131064702e-05}
{"train_loss": 0.9083390831947327, "global_step": 3717, "epoch": 41, "lr": 9.996519968798042e-05}
{"train_loss": 1.0688259601593018, "global_step": 3718, "epoch": 41, "lr": 9.996517805859658e-05}
{"train_loss": 1.0403680801391602, "global_step": 3719, "epoch": 41, "lr": 9.996515642249554e-05}
{"train_loss": 0.87513667345047, "global_step": 3720, "epoch": 41, "lr": 9.996513477967731e-05}
{"train_loss": 1.211017370223999, "global_step": 3721, "epoch": 41, "lr": 9.996511313014185e-05}
{"train_loss": 1.0113043785095215, "global_step": 3722, "epoch": 41, "lr": 9.99650914738892e-05}
{"train_loss": 0.9317690134048462, "global_step": 3723, "epoch": 41, "lr": 9.996506981091936e-05}
{"train_loss": 1.129927396774292, "global_step": 3724, "epoch": 41, "lr": 9.996504814123232e-05}
{"train_loss": 1.09577214717865, "global_step": 3725, "epoch": 41, "lr": 9.996502646482809e-05}
{"train_loss": 1.0292621850967407, "global_step": 3726, "epoch": 41, "lr": 9.996500478170666e-05}
{"train_loss": 1.2826682329177856, "global_step": 3727, "epoch": 41, "lr": 9.996498309186806e-05}
{"train_loss": 1.0976356267929077, "global_step": 3728, "epoch": 41, "lr": 9.996496139531226e-05}
{"train_loss": 0.9655628204345703, "global_step": 3729, "epoch": 41, "lr": 9.996493969203929e-05}
{"train_loss": 1.183418869972229, "global_step": 3730, "epoch": 41, "lr": 9.996491798204915e-05}
{"train_loss": 1.1847487688064575, "global_step": 3731, "epoch": 41, "lr": 9.996489626534182e-05}
{"train_loss": 1.1085708141326904, "global_step": 3732, "epoch": 41, "lr": 9.996487454191733e-05}
{"train_loss": 1.3165485858917236, "global_step": 3733, "epoch": 41, "lr": 9.996485281177567e-05}
{"train_loss": 1.0902352333068848, "global_step": 3734, "epoch": 41, "lr": 9.996483107491685e-05}
{"train_loss": 1.1829438209533691, "global_step": 3735, "epoch": 41, "lr": 9.996480933134084e-05}
{"train_loss": 1.0118052959442139, "global_step": 3736, "epoch": 41, "lr": 9.996478758104771e-05}
{"train_loss": 1.080854259180219, "global_step": 3737, "epoch": 41, "lr": 9.99647658240374e-05, "val_loss": 1.1743748188018799}
{"train_loss": 0.9208433628082275, "global_step": 3738, "epoch": 42, "lr": 9.996474406030993e-05}
{"train_loss": 1.0223997831344604, "global_step": 3739, "epoch": 42, "lr": 9.996472228986534e-05}
{"train_loss": 0.9690371751785278, "global_step": 3740, "epoch": 42, "lr": 9.996470051270357e-05}
{"train_loss": 1.1164137125015259, "global_step": 3741, "epoch": 42, "lr": 9.996467872882467e-05}
{"train_loss": 1.1492688655853271, "global_step": 3742, "epoch": 42, "lr": 9.996465693822862e-05}
{"train_loss": 0.9553351998329163, "global_step": 3743, "epoch": 42, "lr": 9.996463514091543e-05}
{"train_loss": 1.300195336341858, "global_step": 3744, "epoch": 42, "lr": 9.996461333688511e-05}
{"train_loss": 1.1263036727905273, "global_step": 3745, "epoch": 42, "lr": 9.996459152613766e-05}
{"train_loss": 1.058854579925537, "global_step": 3746, "epoch": 42, "lr": 9.996456970867307e-05}
{"train_loss": 1.1747204065322876, "global_step": 3747, "epoch": 42, "lr": 9.996454788449136e-05}
{"train_loss": 1.098997712135315, "global_step": 3748, "epoch": 42, "lr": 9.996452605359253e-05}
{"train_loss": 1.1207736730575562, "global_step": 3749, "epoch": 42, "lr": 9.996450421597657e-05}
{"train_loss": 1.0000234842300415, "global_step": 3750, "epoch": 42, "lr": 9.996448237164349e-05}
{"train_loss": 1.0591262578964233, "global_step": 3751, "epoch": 42, "lr": 9.996446052059329e-05}
{"train_loss": 0.924574613571167, "global_step": 3752, "epoch": 42, "lr": 9.996443866282599e-05}
{"train_loss": 0.9791214466094971, "global_step": 3753, "epoch": 42, "lr": 9.996441679834158e-05}
{"train_loss": 1.2283767461776733, "global_step": 3754, "epoch": 42, "lr": 9.996439492714006e-05}
{"train_loss": 1.0626994371414185, "global_step": 3755, "epoch": 42, "lr": 9.996437304922144e-05}
{"train_loss": 1.0306532382965088, "global_step": 3756, "epoch": 42, "lr": 9.996435116458571e-05}
{"train_loss": 1.0734668970108032, "global_step": 3757, "epoch": 42, "lr": 9.996432927323289e-05}
{"train_loss": 1.2096483707427979, "global_step": 3758, "epoch": 42, "lr": 9.996430737516295e-05}
{"train_loss": 1.0100183486938477, "global_step": 3759, "epoch": 42, "lr": 9.996428547037596e-05}
{"train_loss": 1.1458576917648315, "global_step": 3760, "epoch": 42, "lr": 9.996426355887185e-05}
{"train_loss": 1.047183632850647, "global_step": 3761, "epoch": 42, "lr": 9.996424164065065e-05}
{"train_loss": 1.105633020401001, "global_step": 3762, "epoch": 42, "lr": 9.996421971571239e-05}
{"train_loss": 1.135857105255127, "global_step": 3763, "epoch": 42, "lr": 9.996419778405703e-05}
{"train_loss": 1.1449921131134033, "global_step": 3764, "epoch": 42, "lr": 9.99641758456846e-05}
{"train_loss": 0.9696033596992493, "global_step": 3765, "epoch": 42, "lr": 9.99641539005951e-05}
{"train_loss": 1.2746707201004028, "global_step": 3766, "epoch": 42, "lr": 9.996413194878853e-05}
{"train_loss": 1.2156790494918823, "global_step": 3767, "epoch": 42, "lr": 9.996410999026488e-05}
{"train_loss": 0.9507851600646973, "global_step": 3768, "epoch": 42, "lr": 9.996408802502418e-05}
{"train_loss": 0.9900832176208496, "global_step": 3769, "epoch": 42, "lr": 9.996406605306641e-05}
{"train_loss": 0.9020207524299622, "global_step": 3770, "epoch": 42, "lr": 9.996404407439158e-05}
{"train_loss": 0.8141438364982605, "global_step": 3771, "epoch": 42, "lr": 9.996402208899969e-05}
{"train_loss": 1.1300580501556396, "global_step": 3772, "epoch": 42, "lr": 9.996400009689075e-05}
{"train_loss": 1.0453754663467407, "global_step": 3773, "epoch": 42, "lr": 9.996397809806476e-05}
{"train_loss": 1.0764856338500977, "global_step": 3774, "epoch": 42, "lr": 9.996395609252171e-05}
{"train_loss": 1.0271282196044922, "global_step": 3775, "epoch": 42, "lr": 9.996393408026163e-05}
{"train_loss": 0.9670277833938599, "global_step": 3776, "epoch": 42, "lr": 9.996391206128451e-05}
{"train_loss": 1.0598180294036865, "global_step": 3777, "epoch": 42, "lr": 9.996389003559035e-05}
{"train_loss": 1.0249223709106445, "global_step": 3778, "epoch": 42, "lr": 9.996386800317914e-05}
{"train_loss": 1.1901369094848633, "global_step": 3779, "epoch": 42, "lr": 9.99638459640509e-05}
{"train_loss": 1.0158202648162842, "global_step": 3780, "epoch": 42, "lr": 9.996382391820564e-05}
{"train_loss": 0.9220395088195801, "global_step": 3781, "epoch": 42, "lr": 9.996380186564335e-05}
{"train_loss": 1.0673683881759644, "global_step": 3782, "epoch": 42, "lr": 9.996377980636402e-05}
{"train_loss": 0.9254785776138306, "global_step": 3783, "epoch": 42, "lr": 9.996375774036769e-05}
{"train_loss": 0.9066141247749329, "global_step": 3784, "epoch": 42, "lr": 9.996373566765434e-05}
{"train_loss": 1.0959731340408325, "global_step": 3785, "epoch": 42, "lr": 9.996371358822396e-05}
{"train_loss": 1.1179568767547607, "global_step": 3786, "epoch": 42, "lr": 9.996369150207658e-05}
{"train_loss": 0.9124541878700256, "global_step": 3787, "epoch": 42, "lr": 9.996366940921218e-05}
{"train_loss": 1.1415358781814575, "global_step": 3788, "epoch": 42, "lr": 9.996364730963079e-05}
{"train_loss": 1.276628017425537, "global_step": 3789, "epoch": 42, "lr": 9.996362520333237e-05}
{"train_loss": 1.1204203367233276, "global_step": 3790, "epoch": 42, "lr": 9.996360309031698e-05}
{"train_loss": 1.053648829460144, "global_step": 3791, "epoch": 42, "lr": 9.996358097058458e-05}
{"train_loss": 1.097998023033142, "global_step": 3792, "epoch": 42, "lr": 9.996355884413518e-05}
{"train_loss": 1.0420056581497192, "global_step": 3793, "epoch": 42, "lr": 9.996353671096879e-05}
{"train_loss": 1.0667383670806885, "global_step": 3794, "epoch": 42, "lr": 9.996351457108541e-05}
{"train_loss": 1.1355656385421753, "global_step": 3795, "epoch": 42, "lr": 9.996349242448506e-05}
{"train_loss": 1.1478462219238281, "global_step": 3796, "epoch": 42, "lr": 9.996347027116773e-05}
{"train_loss": 0.9960207939147949, "global_step": 3797, "epoch": 42, "lr": 9.99634481111334e-05}
{"train_loss": 0.9184514284133911, "global_step": 3798, "epoch": 42, "lr": 9.996342594438209e-05}
{"train_loss": 1.1292002201080322, "global_step": 3799, "epoch": 42, "lr": 9.996340377091382e-05}
{"train_loss": 1.1562423706054688, "global_step": 3800, "epoch": 42, "lr": 9.996338159072858e-05}
{"train_loss": 1.0017470121383667, "global_step": 3801, "epoch": 42, "lr": 9.996335940382638e-05}
{"train_loss": 1.1301170587539673, "global_step": 3802, "epoch": 42, "lr": 9.996333721020719e-05}
{"train_loss": 1.1589319705963135, "global_step": 3803, "epoch": 42, "lr": 9.996331500987104e-05}
{"train_loss": 0.9552607536315918, "global_step": 3804, "epoch": 42, "lr": 9.996329280281795e-05}
{"train_loss": 1.0899349451065063, "global_step": 3805, "epoch": 42, "lr": 9.99632705890479e-05}
{"train_loss": 1.1715189218521118, "global_step": 3806, "epoch": 42, "lr": 9.99632483685609e-05}
{"train_loss": 1.2081069946289062, "global_step": 3807, "epoch": 42, "lr": 9.996322614135693e-05}
{"train_loss": 0.9112981557846069, "global_step": 3808, "epoch": 42, "lr": 9.996320390743604e-05}
{"train_loss": 1.0795973539352417, "global_step": 3809, "epoch": 42, "lr": 9.996318166679819e-05}
{"train_loss": 1.0087512731552124, "global_step": 3810, "epoch": 42, "lr": 9.99631594194434e-05}
{"train_loss": 1.0440140962600708, "global_step": 3811, "epoch": 42, "lr": 9.996313716537169e-05}
{"train_loss": 0.9542542099952698, "global_step": 3812, "epoch": 42, "lr": 9.996311490458303e-05}
{"train_loss": 0.9900184273719788, "global_step": 3813, "epoch": 42, "lr": 9.996309263707744e-05}
{"train_loss": 0.954611599445343, "global_step": 3814, "epoch": 42, "lr": 9.996307036285491e-05}
{"train_loss": 1.0843147039413452, "global_step": 3815, "epoch": 42, "lr": 9.996304808191546e-05}
{"train_loss": 1.0677402019500732, "global_step": 3816, "epoch": 42, "lr": 9.996302579425911e-05}
{"train_loss": 0.9865255355834961, "global_step": 3817, "epoch": 42, "lr": 9.996300349988581e-05}
{"train_loss": 1.0246444940567017, "global_step": 3818, "epoch": 42, "lr": 9.996298119879561e-05}
{"train_loss": 0.9928730130195618, "global_step": 3819, "epoch": 42, "lr": 9.996295889098849e-05}
{"train_loss": 0.996788740158081, "global_step": 3820, "epoch": 42, "lr": 9.996293657646444e-05}
{"train_loss": 1.1783123016357422, "global_step": 3821, "epoch": 42, "lr": 9.996291425522352e-05}
{"train_loss": 1.0011759996414185, "global_step": 3822, "epoch": 42, "lr": 9.996289192726568e-05}
{"train_loss": 1.0245956182479858, "global_step": 3823, "epoch": 42, "lr": 9.996286959259093e-05}
{"train_loss": 0.9438909888267517, "global_step": 3824, "epoch": 42, "lr": 9.996284725119928e-05}
{"train_loss": 0.9585978984832764, "global_step": 3825, "epoch": 42, "lr": 9.996282490309074e-05}
{"train_loss": 1.056734976473819, "global_step": 3826, "epoch": 42, "lr": 9.996280254826531e-05, "val_loss": 1.132793664932251}
{"train_loss": 0.9701558947563171, "global_step": 3827, "epoch": 43, "lr": 9.9962780186723e-05}
{"train_loss": 1.10878586769104, "global_step": 3828, "epoch": 43, "lr": 9.996275781846379e-05}
{"train_loss": 0.9893910884857178, "global_step": 3829, "epoch": 43, "lr": 9.996273544348769e-05}
{"train_loss": 0.9264309406280518, "global_step": 3830, "epoch": 43, "lr": 9.996271306179473e-05}
{"train_loss": 0.9698925614356995, "global_step": 3831, "epoch": 43, "lr": 9.996269067338486e-05}
{"train_loss": 0.9583674073219299, "global_step": 3832, "epoch": 43, "lr": 9.996266827825815e-05}
{"train_loss": 0.8952780365943909, "global_step": 3833, "epoch": 43, "lr": 9.996264587641456e-05}
{"train_loss": 1.1713846921920776, "global_step": 3834, "epoch": 43, "lr": 9.996262346785407e-05}
{"train_loss": 0.8263610005378723, "global_step": 3835, "epoch": 43, "lr": 9.996260105257676e-05}
{"train_loss": 1.0686908960342407, "global_step": 3836, "epoch": 43, "lr": 9.996257863058257e-05}
{"train_loss": 1.0408916473388672, "global_step": 3837, "epoch": 43, "lr": 9.99625562018715e-05}
{"train_loss": 1.057849645614624, "global_step": 3838, "epoch": 43, "lr": 9.996253376644358e-05}
{"train_loss": 1.079970359802246, "global_step": 3839, "epoch": 43, "lr": 9.996251132429881e-05}
{"train_loss": 1.1561733484268188, "global_step": 3840, "epoch": 43, "lr": 9.99624888754372e-05}
{"train_loss": 1.0378910303115845, "global_step": 3841, "epoch": 43, "lr": 9.996246641985874e-05}
{"train_loss": 0.940451979637146, "global_step": 3842, "epoch": 43, "lr": 9.996244395756344e-05}
{"train_loss": 1.0740963220596313, "global_step": 3843, "epoch": 43, "lr": 9.996242148855128e-05}
{"train_loss": 1.0787596702575684, "global_step": 3844, "epoch": 43, "lr": 9.99623990128223e-05}
{"train_loss": 1.1879256963729858, "global_step": 3845, "epoch": 43, "lr": 9.996237653037646e-05}
{"train_loss": 1.390919804573059, "global_step": 3846, "epoch": 43, "lr": 9.996235404121382e-05}
{"train_loss": 1.168829083442688, "global_step": 3847, "epoch": 43, "lr": 9.996233154533432e-05}
{"train_loss": 1.0005300045013428, "global_step": 3848, "epoch": 43, "lr": 9.996230904273802e-05}
{"train_loss": 1.2419064044952393, "global_step": 3849, "epoch": 43, "lr": 9.996228653342488e-05}
{"train_loss": 1.1009531021118164, "global_step": 3850, "epoch": 43, "lr": 9.996226401739492e-05}
{"train_loss": 1.0962376594543457, "global_step": 3851, "epoch": 43, "lr": 9.996224149464815e-05}
{"train_loss": 0.9917469024658203, "global_step": 3852, "epoch": 43, "lr": 9.996221896518456e-05}
{"train_loss": 1.0934545993804932, "global_step": 3853, "epoch": 43, "lr": 9.996219642900416e-05}
{"train_loss": 0.9325832724571228, "global_step": 3854, "epoch": 43, "lr": 9.996217388610695e-05}
{"train_loss": 0.9226099848747253, "global_step": 3855, "epoch": 43, "lr": 9.996215133649294e-05}
{"train_loss": 1.023431420326233, "global_step": 3856, "epoch": 43, "lr": 9.996212878016211e-05}
{"train_loss": 0.6724268198013306, "global_step": 3857, "epoch": 43, "lr": 9.99621062171145e-05}
{"train_loss": 0.9039905667304993, "global_step": 3858, "epoch": 43, "lr": 9.996208364735009e-05}
{"train_loss": 1.240106463432312, "global_step": 3859, "epoch": 43, "lr": 9.996206107086889e-05}
{"train_loss": 0.9908126592636108, "global_step": 3860, "epoch": 43, "lr": 9.99620384876709e-05}
{"train_loss": 0.9864587783813477, "global_step": 3861, "epoch": 43, "lr": 9.996201589775611e-05}
{"train_loss": 1.0373432636260986, "global_step": 3862, "epoch": 43, "lr": 9.996199330112454e-05}
{"train_loss": 1.1195580959320068, "global_step": 3863, "epoch": 43, "lr": 9.99619706977762e-05}
{"train_loss": 0.9834874868392944, "global_step": 3864, "epoch": 43, "lr": 9.996194808771107e-05}
{"train_loss": 0.9443075656890869, "global_step": 3865, "epoch": 43, "lr": 9.996192547092917e-05}
{"train_loss": 1.3165206909179688, "global_step": 3866, "epoch": 43, "lr": 9.99619028474305e-05}
{"train_loss": 0.9051398634910583, "global_step": 3867, "epoch": 43, "lr": 9.996188021721506e-05}
{"train_loss": 1.2100203037261963, "global_step": 3868, "epoch": 43, "lr": 9.996185758028284e-05}
{"train_loss": 0.8049759864807129, "global_step": 3869, "epoch": 43, "lr": 9.99618349366339e-05}
{"train_loss": 1.0939058065414429, "global_step": 3870, "epoch": 43, "lr": 9.996181228626816e-05}
{"train_loss": 1.161492943763733, "global_step": 3871, "epoch": 43, "lr": 9.996178962918567e-05}
{"train_loss": 1.0817146301269531, "global_step": 3872, "epoch": 43, "lr": 9.996176696538642e-05}
{"train_loss": 0.9714536666870117, "global_step": 3873, "epoch": 43, "lr": 9.996174429487045e-05}
{"train_loss": 0.9873449206352234, "global_step": 3874, "epoch": 43, "lr": 9.99617216176377e-05}
{"train_loss": 0.8933320045471191, "global_step": 3875, "epoch": 43, "lr": 9.996169893368822e-05}
{"train_loss": 1.1358027458190918, "global_step": 3876, "epoch": 43, "lr": 9.996167624302199e-05}
{"train_loss": 1.1690888404846191, "global_step": 3877, "epoch": 43, "lr": 9.996165354563903e-05}
{"train_loss": 1.1310193538665771, "global_step": 3878, "epoch": 43, "lr": 9.996163084153934e-05}
{"train_loss": 0.961310863494873, "global_step": 3879, "epoch": 43, "lr": 9.99616081307229e-05}
{"train_loss": 0.9203137755393982, "global_step": 3880, "epoch": 43, "lr": 9.996158541318974e-05}
{"train_loss": 1.0686354637145996, "global_step": 3881, "epoch": 43, "lr": 9.996156268893986e-05}
{"train_loss": 1.126699686050415, "global_step": 3882, "epoch": 43, "lr": 9.996153995797326e-05}
{"train_loss": 1.1263693571090698, "global_step": 3883, "epoch": 43, "lr": 9.996151722028993e-05}
{"train_loss": 0.9261597394943237, "global_step": 3884, "epoch": 43, "lr": 9.996149447588987e-05}
{"train_loss": 1.121082067489624, "global_step": 3885, "epoch": 43, "lr": 9.996147172477312e-05}
{"train_loss": 1.0673513412475586, "global_step": 3886, "epoch": 43, "lr": 9.996144896693964e-05}
{"train_loss": 0.9754729866981506, "global_step": 3887, "epoch": 43, "lr": 9.996142620238946e-05}
{"train_loss": 0.8273153305053711, "global_step": 3888, "epoch": 43, "lr": 9.996140343112259e-05}
{"train_loss": 0.9937941431999207, "global_step": 3889, "epoch": 43, "lr": 9.9961380653139e-05}
{"train_loss": 0.86927729845047, "global_step": 3890, "epoch": 43, "lr": 9.996135786843873e-05}
{"train_loss": 0.9862596392631531, "global_step": 3891, "epoch": 43, "lr": 9.996133507702174e-05}
{"train_loss": 0.9476108551025391, "global_step": 3892, "epoch": 43, "lr": 9.996131227888807e-05}
{"train_loss": 0.8194102644920349, "global_step": 3893, "epoch": 43, "lr": 9.996128947403772e-05}
{"train_loss": 1.0817457437515259, "global_step": 3894, "epoch": 43, "lr": 9.996126666247066e-05}
{"train_loss": 1.0443990230560303, "global_step": 3895, "epoch": 43, "lr": 9.996124384418692e-05}
{"train_loss": 1.1015350818634033, "global_step": 3896, "epoch": 43, "lr": 9.996122101918651e-05}
{"train_loss": 1.208339810371399, "global_step": 3897, "epoch": 43, "lr": 9.996119818746943e-05}
{"train_loss": 0.9777207374572754, "global_step": 3898, "epoch": 43, "lr": 9.996117534903567e-05}
{"train_loss": 0.9676419496536255, "global_step": 3899, "epoch": 43, "lr": 9.996115250388523e-05}
{"train_loss": 1.491166591644287, "global_step": 3900, "epoch": 43, "lr": 9.996112965201813e-05}
{"train_loss": 0.9674870371818542, "global_step": 3901, "epoch": 43, "lr": 9.996110679343435e-05}
{"train_loss": 1.0095648765563965, "global_step": 3902, "epoch": 43, "lr": 9.996108392813392e-05}
{"train_loss": 1.0192950963974, "global_step": 3903, "epoch": 43, "lr": 9.996106105611684e-05}
{"train_loss": 1.0401413440704346, "global_step": 3904, "epoch": 43, "lr": 9.996103817738309e-05}
{"train_loss": 1.2505848407745361, "global_step": 3905, "epoch": 43, "lr": 9.99610152919327e-05}
{"train_loss": 1.0151671171188354, "global_step": 3906, "epoch": 43, "lr": 9.996099239976566e-05}
{"train_loss": 0.9439056515693665, "global_step": 3907, "epoch": 43, "lr": 9.996096950088196e-05}
{"train_loss": 0.9390667676925659, "global_step": 3908, "epoch": 43, "lr": 9.996094659528164e-05}
{"train_loss": 1.1562345027923584, "global_step": 3909, "epoch": 43, "lr": 9.996092368296465e-05}
{"train_loss": 0.933963418006897, "global_step": 3910, "epoch": 43, "lr": 9.996090076393104e-05}
{"train_loss": 1.1668163537979126, "global_step": 3911, "epoch": 43, "lr": 9.996087783818079e-05}
{"train_loss": 1.032723069190979, "global_step": 3912, "epoch": 43, "lr": 9.996085490571391e-05}
{"train_loss": 1.1970093250274658, "global_step": 3913, "epoch": 43, "lr": 9.996083196653042e-05}
{"train_loss": 1.0232428312301636, "global_step": 3914, "epoch": 43, "lr": 9.996080902063027e-05}
{"train_loss": 1.0408179706402039, "global_step": 3915, "epoch": 43, "lr": 9.996078606801354e-05, "val_loss": 1.1383010149002075}
{"train_loss": 1.0291615724563599, "global_step": 3916, "epoch": 44, "lr": 9.996076310868015e-05}
{"train_loss": 0.9110746383666992, "global_step": 3917, "epoch": 44, "lr": 9.996074014263017e-05}
{"train_loss": 1.1452302932739258, "global_step": 3918, "epoch": 44, "lr": 9.996071716986357e-05}
{"train_loss": 0.820729672908783, "global_step": 3919, "epoch": 44, "lr": 9.996069419038035e-05}
{"train_loss": 0.9129824042320251, "global_step": 3920, "epoch": 44, "lr": 9.996067120418053e-05}
{"train_loss": 1.003796100616455, "global_step": 3921, "epoch": 44, "lr": 9.996064821126412e-05}
{"train_loss": 1.1054158210754395, "global_step": 3922, "epoch": 44, "lr": 9.99606252116311e-05}
{"train_loss": 1.0174751281738281, "global_step": 3923, "epoch": 44, "lr": 9.996060220528149e-05}
{"train_loss": 1.02439284324646, "global_step": 3924, "epoch": 44, "lr": 9.996057919221528e-05}
{"train_loss": 1.0302966833114624, "global_step": 3925, "epoch": 44, "lr": 9.996055617243248e-05}
{"train_loss": 1.034039855003357, "global_step": 3926, "epoch": 44, "lr": 9.996053314593308e-05}
{"train_loss": 0.9086899757385254, "global_step": 3927, "epoch": 44, "lr": 9.996051011271711e-05}
{"train_loss": 1.083546757698059, "global_step": 3928, "epoch": 44, "lr": 9.996048707278456e-05}
{"train_loss": 0.9974333643913269, "global_step": 3929, "epoch": 44, "lr": 9.996046402613541e-05}
{"train_loss": 1.148422122001648, "global_step": 3930, "epoch": 44, "lr": 9.99604409727697e-05}
{"train_loss": 1.198378086090088, "global_step": 3931, "epoch": 44, "lr": 9.996041791268743e-05}
{"train_loss": 1.0409307479858398, "global_step": 3932, "epoch": 44, "lr": 9.996039484588857e-05}
{"train_loss": 1.1137964725494385, "global_step": 3933, "epoch": 44, "lr": 9.996037177237317e-05}
{"train_loss": 1.1505932807922363, "global_step": 3934, "epoch": 44, "lr": 9.996034869214119e-05}
{"train_loss": 1.0796663761138916, "global_step": 3935, "epoch": 44, "lr": 9.996032560519264e-05}
{"train_loss": 0.9493108987808228, "global_step": 3936, "epoch": 44, "lr": 9.996030251152755e-05}
{"train_loss": 1.1377735137939453, "global_step": 3937, "epoch": 44, "lr": 9.99602794111459e-05}
{"train_loss": 0.7831444144248962, "global_step": 3938, "epoch": 44, "lr": 9.996025630404769e-05}
{"train_loss": 1.0726375579833984, "global_step": 3939, "epoch": 44, "lr": 9.996023319023294e-05}
{"train_loss": 1.1178247928619385, "global_step": 3940, "epoch": 44, "lr": 9.996021006970165e-05}
{"train_loss": 1.0673474073410034, "global_step": 3941, "epoch": 44, "lr": 9.996018694245381e-05}
{"train_loss": 1.1312915086746216, "global_step": 3942, "epoch": 44, "lr": 9.996016380848944e-05}
{"train_loss": 1.2690709829330444, "global_step": 3943, "epoch": 44, "lr": 9.996014066780852e-05}
{"train_loss": 1.0792958736419678, "global_step": 3944, "epoch": 44, "lr": 9.996011752041108e-05}
{"train_loss": 1.1134544610977173, "global_step": 3945, "epoch": 44, "lr": 9.996009436629712e-05}
{"train_loss": 1.1670702695846558, "global_step": 3946, "epoch": 44, "lr": 9.996007120546662e-05}
{"train_loss": 0.9277099967002869, "global_step": 3947, "epoch": 44, "lr": 9.99600480379196e-05}
{"train_loss": 1.1500910520553589, "global_step": 3948, "epoch": 44, "lr": 9.996002486365608e-05}
{"train_loss": 1.1624985933303833, "global_step": 3949, "epoch": 44, "lr": 9.996000168267602e-05}
{"train_loss": 1.0090395212173462, "global_step": 3950, "epoch": 44, "lr": 9.995997849497944e-05}
{"train_loss": 1.0940269231796265, "global_step": 3951, "epoch": 44, "lr": 9.995995530056636e-05}
{"train_loss": 1.0293891429901123, "global_step": 3952, "epoch": 44, "lr": 9.995993209943679e-05}
{"train_loss": 1.1460084915161133, "global_step": 3953, "epoch": 44, "lr": 9.99599088915907e-05}
{"train_loss": 0.8941785097122192, "global_step": 3954, "epoch": 44, "lr": 9.995988567702812e-05}
{"train_loss": 0.9484989643096924, "global_step": 3955, "epoch": 44, "lr": 9.995986245574903e-05}
{"train_loss": 1.182884693145752, "global_step": 3956, "epoch": 44, "lr": 9.995983922775345e-05}
{"train_loss": 0.9088060855865479, "global_step": 3957, "epoch": 44, "lr": 9.995981599304138e-05}
{"train_loss": 0.984225869178772, "global_step": 3958, "epoch": 44, "lr": 9.995979275161282e-05}
{"train_loss": 0.8839433789253235, "global_step": 3959, "epoch": 44, "lr": 9.995976950346776e-05}
{"train_loss": 1.1657471656799316, "global_step": 3960, "epoch": 44, "lr": 9.995974624860624e-05}
{"train_loss": 1.1170549392700195, "global_step": 3961, "epoch": 44, "lr": 9.995972298702822e-05}
{"train_loss": 1.087349534034729, "global_step": 3962, "epoch": 44, "lr": 9.995969971873374e-05}
{"train_loss": 1.1517399549484253, "global_step": 3963, "epoch": 44, "lr": 9.99596764437228e-05}
{"train_loss": 1.2137391567230225, "global_step": 3964, "epoch": 44, "lr": 9.995965316199537e-05}
{"train_loss": 1.1557458639144897, "global_step": 3965, "epoch": 44, "lr": 9.995962987355148e-05}
{"train_loss": 1.0803874731063843, "global_step": 3966, "epoch": 44, "lr": 9.995960657839112e-05}
{"train_loss": 0.8506565093994141, "global_step": 3967, "epoch": 44, "lr": 9.99595832765143e-05}
{"train_loss": 1.0529735088348389, "global_step": 3968, "epoch": 44, "lr": 9.995955996792104e-05}
{"train_loss": 0.9437519311904907, "global_step": 3969, "epoch": 44, "lr": 9.99595366526113e-05}
{"train_loss": 1.0770773887634277, "global_step": 3970, "epoch": 44, "lr": 9.995951333058513e-05}
{"train_loss": 0.9835348129272461, "global_step": 3971, "epoch": 44, "lr": 9.995949000184251e-05}
{"train_loss": 0.9900027513504028, "global_step": 3972, "epoch": 44, "lr": 9.995946666638343e-05}
{"train_loss": 1.1209136247634888, "global_step": 3973, "epoch": 44, "lr": 9.995944332420792e-05}
{"train_loss": 0.9339480400085449, "global_step": 3974, "epoch": 44, "lr": 9.995941997531597e-05}
{"train_loss": 1.0086332559585571, "global_step": 3975, "epoch": 44, "lr": 9.995939661970759e-05}
{"train_loss": 1.0106593370437622, "global_step": 3976, "epoch": 44, "lr": 9.995937325738279e-05}
{"train_loss": 1.2158445119857788, "global_step": 3977, "epoch": 44, "lr": 9.995934988834152e-05}
{"train_loss": 1.0255316495895386, "global_step": 3978, "epoch": 44, "lr": 9.995932651258387e-05}
{"train_loss": 1.0203512907028198, "global_step": 3979, "epoch": 44, "lr": 9.995930313010977e-05}
{"train_loss": 1.1128380298614502, "global_step": 3980, "epoch": 44, "lr": 9.995927974091925e-05}
{"train_loss": 1.0122486352920532, "global_step": 3981, "epoch": 44, "lr": 9.995925634501232e-05}
{"train_loss": 1.0372674465179443, "global_step": 3982, "epoch": 44, "lr": 9.995923294238898e-05}
{"train_loss": 1.0852105617523193, "global_step": 3983, "epoch": 44, "lr": 9.995920953304923e-05}
{"train_loss": 1.1292215585708618, "global_step": 3984, "epoch": 44, "lr": 9.995918611699307e-05}
{"train_loss": 0.938607394695282, "global_step": 3985, "epoch": 44, "lr": 9.995916269422052e-05}
{"train_loss": 0.9195778965950012, "global_step": 3986, "epoch": 44, "lr": 9.995913926473155e-05}
{"train_loss": 1.0127153396606445, "global_step": 3987, "epoch": 44, "lr": 9.99591158285262e-05}
{"train_loss": 0.9304772615432739, "global_step": 3988, "epoch": 44, "lr": 9.995909238560444e-05}
{"train_loss": 1.1308103799819946, "global_step": 3989, "epoch": 44, "lr": 9.995906893596629e-05}
{"train_loss": 0.6930273175239563, "global_step": 3990, "epoch": 44, "lr": 9.995904547961177e-05}
{"train_loss": 0.8592782020568848, "global_step": 3991, "epoch": 44, "lr": 9.995902201654084e-05}
{"train_loss": 1.065313458442688, "global_step": 3992, "epoch": 44, "lr": 9.995899854675355e-05}
{"train_loss": 0.9615971446037292, "global_step": 3993, "epoch": 44, "lr": 9.995897507024988e-05}
{"train_loss": 1.0809575319290161, "global_step": 3994, "epoch": 44, "lr": 9.995895158702982e-05}
{"train_loss": 0.938115656375885, "global_step": 3995, "epoch": 44, "lr": 9.995892809709338e-05}
{"train_loss": 1.0255886316299438, "global_step": 3996, "epoch": 44, "lr": 9.99589046004406e-05}
{"train_loss": 0.9771687984466553, "global_step": 3997, "epoch": 44, "lr": 9.995888109707144e-05}
{"train_loss": 1.0255858898162842, "global_step": 3998, "epoch": 44, "lr": 9.995885758698592e-05}
{"train_loss": 0.9868127703666687, "global_step": 3999, "epoch": 44, "lr": 9.995883407018402e-05}
{"train_loss": 1.072649359703064, "global_step": 4000, "epoch": 44, "lr": 9.995881054666578e-05}
{"train_loss": 0.9586833119392395, "global_step": 4001, "epoch": 44, "lr": 9.99587870164312e-05}
{"train_loss": 1.107241153717041, "global_step": 4002, "epoch": 44, "lr": 9.995876347948026e-05}
{"train_loss": 1.0165631771087646, "global_step": 4003, "epoch": 44, "lr": 9.995873993581297e-05}
{"train_loss": 1.037957403766975, "global_step": 4004, "epoch": 44, "lr": 9.995871638542932e-05, "val_loss": 1.1202369928359985}
{"train_loss": 1.0125513076782227, "global_step": 4005, "epoch": 45, "lr": 9.995869282832937e-05}
{"train_loss": 1.172757863998413, "global_step": 4006, "epoch": 45, "lr": 9.995866926451305e-05}
{"train_loss": 0.9965099692344666, "global_step": 4007, "epoch": 45, "lr": 9.995864569398041e-05}
{"train_loss": 1.114471673965454, "global_step": 4008, "epoch": 45, "lr": 9.995862211673144e-05}
{"train_loss": 1.0163815021514893, "global_step": 4009, "epoch": 45, "lr": 9.995859853276612e-05}
{"train_loss": 1.0500203371047974, "global_step": 4010, "epoch": 45, "lr": 9.995857494208451e-05}
{"train_loss": 0.9034945964813232, "global_step": 4011, "epoch": 45, "lr": 9.995855134468656e-05}
{"train_loss": 1.1010175943374634, "global_step": 4012, "epoch": 45, "lr": 9.995852774057228e-05}
{"train_loss": 1.0191808938980103, "global_step": 4013, "epoch": 45, "lr": 9.99585041297417e-05}
{"train_loss": 1.069725513458252, "global_step": 4014, "epoch": 45, "lr": 9.99584805121948e-05}
{"train_loss": 1.1914665699005127, "global_step": 4015, "epoch": 45, "lr": 9.99584568879316e-05}
{"train_loss": 1.1007702350616455, "global_step": 4016, "epoch": 45, "lr": 9.99584332569521e-05}
{"train_loss": 1.0982682704925537, "global_step": 4017, "epoch": 45, "lr": 9.995840961925629e-05}
{"train_loss": 0.9962249398231506, "global_step": 4018, "epoch": 45, "lr": 9.995838597484418e-05}
{"train_loss": 1.028691053390503, "global_step": 4019, "epoch": 45, "lr": 9.995836232371577e-05}
{"train_loss": 1.0217872858047485, "global_step": 4020, "epoch": 45, "lr": 9.995833866587107e-05}
{"train_loss": 1.225367546081543, "global_step": 4021, "epoch": 45, "lr": 9.995831500131008e-05}
{"train_loss": 1.1528440713882446, "global_step": 4022, "epoch": 45, "lr": 9.99582913300328e-05}
{"train_loss": 0.9906872510910034, "global_step": 4023, "epoch": 45, "lr": 9.995826765203924e-05}
{"train_loss": 1.3170164823532104, "global_step": 4024, "epoch": 45, "lr": 9.99582439673294e-05}
{"train_loss": 1.110995888710022, "global_step": 4025, "epoch": 45, "lr": 9.995822027590329e-05}
{"train_loss": 1.0061136484146118, "global_step": 4026, "epoch": 45, "lr": 9.99581965777609e-05}
{"train_loss": 0.8557946681976318, "global_step": 4027, "epoch": 45, "lr": 9.995817287290224e-05}
{"train_loss": 0.9866048693656921, "global_step": 4028, "epoch": 45, "lr": 9.995814916132731e-05}
{"train_loss": 1.0209147930145264, "global_step": 4029, "epoch": 45, "lr": 9.99581254430361e-05}
{"train_loss": 1.0298553705215454, "global_step": 4030, "epoch": 45, "lr": 9.995810171802866e-05}
{"train_loss": 1.0083067417144775, "global_step": 4031, "epoch": 45, "lr": 9.995807798630495e-05}
{"train_loss": 0.8754175305366516, "global_step": 4032, "epoch": 45, "lr": 9.995805424786498e-05}
{"train_loss": 0.8853932619094849, "global_step": 4033, "epoch": 45, "lr": 9.995803050270876e-05}
{"train_loss": 1.023752212524414, "global_step": 4034, "epoch": 45, "lr": 9.995800675083631e-05}
{"train_loss": 1.1162844896316528, "global_step": 4035, "epoch": 45, "lr": 9.99579829922476e-05}
{"train_loss": 1.0411570072174072, "global_step": 4036, "epoch": 45, "lr": 9.995795922694265e-05}
{"train_loss": 1.2013620138168335, "global_step": 4037, "epoch": 45, "lr": 9.995793545492144e-05}
{"train_loss": 1.0664687156677246, "global_step": 4038, "epoch": 45, "lr": 9.995791167618403e-05}
{"train_loss": 1.0758615732192993, "global_step": 4039, "epoch": 45, "lr": 9.995788789073036e-05}
{"train_loss": 1.167218565940857, "global_step": 4040, "epoch": 45, "lr": 9.995786409856048e-05}
{"train_loss": 1.0822511911392212, "global_step": 4041, "epoch": 45, "lr": 9.995784029967436e-05}
{"train_loss": 0.866782546043396, "global_step": 4042, "epoch": 45, "lr": 9.995781649407203e-05}
{"train_loss": 1.0955538749694824, "global_step": 4043, "epoch": 45, "lr": 9.995779268175345e-05}
{"train_loss": 1.3166414499282837, "global_step": 4044, "epoch": 45, "lr": 9.995776886271868e-05}
{"train_loss": 1.1696399450302124, "global_step": 4045, "epoch": 45, "lr": 9.995774503696771e-05}
{"train_loss": 0.9562560319900513, "global_step": 4046, "epoch": 45, "lr": 9.99577212045005e-05}
{"train_loss": 1.1413367986679077, "global_step": 4047, "epoch": 45, "lr": 9.995769736531711e-05}
{"train_loss": 1.0196329355239868, "global_step": 4048, "epoch": 45, "lr": 9.995767351941749e-05}
{"train_loss": 1.025996446609497, "global_step": 4049, "epoch": 45, "lr": 9.995764966680168e-05}
{"train_loss": 1.0591318607330322, "global_step": 4050, "epoch": 45, "lr": 9.995762580746968e-05}
{"train_loss": 1.0068458318710327, "global_step": 4051, "epoch": 45, "lr": 9.995760194142149e-05}
{"train_loss": 1.1555858850479126, "global_step": 4052, "epoch": 45, "lr": 9.99575780686571e-05}
{"train_loss": 1.2917885780334473, "global_step": 4053, "epoch": 45, "lr": 9.995755418917652e-05}
{"train_loss": 1.145624041557312, "global_step": 4054, "epoch": 45, "lr": 9.995753030297976e-05}
{"train_loss": 0.9039902687072754, "global_step": 4055, "epoch": 45, "lr": 9.995750641006683e-05}
{"train_loss": 0.9559962153434753, "global_step": 4056, "epoch": 45, "lr": 9.995748251043771e-05}
{"train_loss": 0.9946638941764832, "global_step": 4057, "epoch": 45, "lr": 9.99574586040924e-05}
{"train_loss": 1.0527970790863037, "global_step": 4058, "epoch": 45, "lr": 9.995743469103094e-05}
{"train_loss": 0.964987576007843, "global_step": 4059, "epoch": 45, "lr": 9.995741077125331e-05}
{"train_loss": 1.051234245300293, "global_step": 4060, "epoch": 45, "lr": 9.995738684475952e-05}
{"train_loss": 0.8621966242790222, "global_step": 4061, "epoch": 45, "lr": 9.995736291154956e-05}
{"train_loss": 1.136466145515442, "global_step": 4062, "epoch": 45, "lr": 9.995733897162343e-05}
{"train_loss": 1.0122164487838745, "global_step": 4063, "epoch": 45, "lr": 9.995731502498116e-05}
{"train_loss": 1.1212722063064575, "global_step": 4064, "epoch": 45, "lr": 9.995729107162273e-05}
{"train_loss": 0.9227925539016724, "global_step": 4065, "epoch": 45, "lr": 9.995726711154816e-05}
{"train_loss": 1.0297021865844727, "global_step": 4066, "epoch": 45, "lr": 9.995724314475744e-05}
{"train_loss": 1.0792138576507568, "global_step": 4067, "epoch": 45, "lr": 9.995721917125058e-05}
{"train_loss": 1.1363284587860107, "global_step": 4068, "epoch": 45, "lr": 9.995719519102757e-05}
{"train_loss": 0.9337202906608582, "global_step": 4069, "epoch": 45, "lr": 9.995717120408843e-05}
{"train_loss": 1.1279561519622803, "global_step": 4070, "epoch": 45, "lr": 9.995714721043317e-05}
{"train_loss": 0.8962255120277405, "global_step": 4071, "epoch": 45, "lr": 9.995712321006176e-05}
{"train_loss": 1.05584716796875, "global_step": 4072, "epoch": 45, "lr": 9.995709920297424e-05}
{"train_loss": 1.1720967292785645, "global_step": 4073, "epoch": 45, "lr": 9.99570751891706e-05}
{"train_loss": 0.8728674054145813, "global_step": 4074, "epoch": 45, "lr": 9.995705116865081e-05}
{"train_loss": 1.1125644445419312, "global_step": 4075, "epoch": 45, "lr": 9.995702714141492e-05}
{"train_loss": 1.0623524188995361, "global_step": 4076, "epoch": 45, "lr": 9.995700310746293e-05}
{"train_loss": 1.2561302185058594, "global_step": 4077, "epoch": 45, "lr": 9.995697906679482e-05}
{"train_loss": 0.8577332496643066, "global_step": 4078, "epoch": 45, "lr": 9.99569550194106e-05}
{"train_loss": 1.0507094860076904, "global_step": 4079, "epoch": 45, "lr": 9.995693096531028e-05}
{"train_loss": 0.9149991273880005, "global_step": 4080, "epoch": 45, "lr": 9.995690690449386e-05}
{"train_loss": 1.1311577558517456, "global_step": 4081, "epoch": 45, "lr": 9.995688283696134e-05}
{"train_loss": 0.9721408486366272, "global_step": 4082, "epoch": 45, "lr": 9.995685876271272e-05}
{"train_loss": 1.1756255626678467, "global_step": 4083, "epoch": 45, "lr": 9.995683468174802e-05}
{"train_loss": 1.0196435451507568, "global_step": 4084, "epoch": 45, "lr": 9.995681059406723e-05}
{"train_loss": 1.0484497547149658, "global_step": 4085, "epoch": 45, "lr": 9.995678649967037e-05}
{"train_loss": 0.8496639728546143, "global_step": 4086, "epoch": 45, "lr": 9.99567623985574e-05}
{"train_loss": 1.1589415073394775, "global_step": 4087, "epoch": 45, "lr": 9.995673829072837e-05}
{"train_loss": 1.0654523372650146, "global_step": 4088, "epoch": 45, "lr": 9.995671417618327e-05}
{"train_loss": 1.0077803134918213, "global_step": 4089, "epoch": 45, "lr": 9.995669005492207e-05}
{"train_loss": 1.0589958429336548, "global_step": 4090, "epoch": 45, "lr": 9.995666592694484e-05}
{"train_loss": 1.083430528640747, "global_step": 4091, "epoch": 45, "lr": 9.995664179225152e-05}
{"train_loss": 0.8669159412384033, "global_step": 4092, "epoch": 45, "lr": 9.995661765084215e-05}
{"train_loss": 1.0495309762740404, "global_step": 4093, "epoch": 45, "lr": 9.995659350271672e-05, "val_loss": 1.1596826314926147, "train_action_mse_error": 57.48578643798828}
{"train_loss": 0.9683139324188232, "global_step": 4094, "epoch": 46, "lr": 9.995656934787523e-05}
{"train_loss": 1.0845896005630493, "global_step": 4095, "epoch": 46, "lr": 9.995654518631768e-05}
{"train_loss": 1.0655676126480103, "global_step": 4096, "epoch": 46, "lr": 9.995652101804409e-05}
{"train_loss": 0.9031346440315247, "global_step": 4097, "epoch": 46, "lr": 9.995649684305445e-05}
{"train_loss": 1.0973150730133057, "global_step": 4098, "epoch": 46, "lr": 9.995647266134878e-05}
{"train_loss": 1.0368887186050415, "global_step": 4099, "epoch": 46, "lr": 9.995644847292707e-05}
{"train_loss": 0.917323648929596, "global_step": 4100, "epoch": 46, "lr": 9.995642427778932e-05}
{"train_loss": 1.184830665588379, "global_step": 4101, "epoch": 46, "lr": 9.995640007593555e-05}
{"train_loss": 1.134380578994751, "global_step": 4102, "epoch": 46, "lr": 9.995637586736573e-05}
{"train_loss": 1.0942151546478271, "global_step": 4103, "epoch": 46, "lr": 9.99563516520799e-05}
{"train_loss": 1.1480287313461304, "global_step": 4104, "epoch": 46, "lr": 9.995632743007804e-05}
{"train_loss": 0.9496729373931885, "global_step": 4105, "epoch": 46, "lr": 9.995630320136015e-05}
{"train_loss": 0.9085118174552917, "global_step": 4106, "epoch": 46, "lr": 9.995627896592625e-05}
{"train_loss": 1.1122314929962158, "global_step": 4107, "epoch": 46, "lr": 9.995625472377635e-05}
{"train_loss": 1.1259855031967163, "global_step": 4108, "epoch": 46, "lr": 9.995623047491043e-05}
{"train_loss": 1.0813933610916138, "global_step": 4109, "epoch": 46, "lr": 9.995620621932851e-05}
{"train_loss": 0.9741852879524231, "global_step": 4110, "epoch": 46, "lr": 9.995618195703058e-05}
{"train_loss": 1.2071895599365234, "global_step": 4111, "epoch": 46, "lr": 9.995615768801665e-05}
{"train_loss": 1.0904176235198975, "global_step": 4112, "epoch": 46, "lr": 9.995613341228673e-05}
{"train_loss": 1.0413895845413208, "global_step": 4113, "epoch": 46, "lr": 9.995610912984081e-05}
{"train_loss": 1.2409942150115967, "global_step": 4114, "epoch": 46, "lr": 9.99560848406789e-05}
{"train_loss": 1.2793185710906982, "global_step": 4115, "epoch": 46, "lr": 9.9956060544801e-05}
{"train_loss": 0.9924761056900024, "global_step": 4116, "epoch": 46, "lr": 9.995603624220713e-05}
{"train_loss": 1.0632531642913818, "global_step": 4117, "epoch": 46, "lr": 9.995601193289727e-05}
{"train_loss": 0.9640235304832458, "global_step": 4118, "epoch": 46, "lr": 9.995598761687144e-05}
{"train_loss": 1.1060185432434082, "global_step": 4119, "epoch": 46, "lr": 9.995596329412965e-05}
{"train_loss": 1.2789236307144165, "global_step": 4120, "epoch": 46, "lr": 9.995593896467186e-05}
{"train_loss": 1.1265687942504883, "global_step": 4121, "epoch": 46, "lr": 9.995591462849812e-05}
{"train_loss": 1.0920647382736206, "global_step": 4122, "epoch": 46, "lr": 9.99558902856084e-05}
{"train_loss": 0.9863377809524536, "global_step": 4123, "epoch": 46, "lr": 9.995586593600273e-05}
{"train_loss": 0.9549657106399536, "global_step": 4124, "epoch": 46, "lr": 9.99558415796811e-05}
{"train_loss": 0.923441469669342, "global_step": 4125, "epoch": 46, "lr": 9.995581721664352e-05}
{"train_loss": 1.069490909576416, "global_step": 4126, "epoch": 46, "lr": 9.995579284688999e-05}
{"train_loss": 0.9356847405433655, "global_step": 4127, "epoch": 46, "lr": 9.99557684704205e-05}
{"train_loss": 0.9395031332969666, "global_step": 4128, "epoch": 46, "lr": 9.995574408723509e-05}
{"train_loss": 0.9251219034194946, "global_step": 4129, "epoch": 46, "lr": 9.995571969733373e-05}
{"train_loss": 0.8651533126831055, "global_step": 4130, "epoch": 46, "lr": 9.995569530071642e-05}
{"train_loss": 0.9253534078598022, "global_step": 4131, "epoch": 46, "lr": 9.995567089738318e-05}
{"train_loss": 1.025375485420227, "global_step": 4132, "epoch": 46, "lr": 9.995564648733402e-05}
{"train_loss": 0.9394686222076416, "global_step": 4133, "epoch": 46, "lr": 9.995562207056893e-05}
{"train_loss": 1.1512587070465088, "global_step": 4134, "epoch": 46, "lr": 9.995559764708791e-05}
{"train_loss": 1.008745551109314, "global_step": 4135, "epoch": 46, "lr": 9.995557321689095e-05}
{"train_loss": 0.8870761394500732, "global_step": 4136, "epoch": 46, "lr": 9.99555487799781e-05}
{"train_loss": 1.0658316612243652, "global_step": 4137, "epoch": 46, "lr": 9.995552433634932e-05}
{"train_loss": 0.9693368077278137, "global_step": 4138, "epoch": 46, "lr": 9.995549988600463e-05}
{"train_loss": 0.8367794156074524, "global_step": 4139, "epoch": 46, "lr": 9.995547542894404e-05}
{"train_loss": 0.8867847919464111, "global_step": 4140, "epoch": 46, "lr": 9.995545096516754e-05}
{"train_loss": 1.111721396446228, "global_step": 4141, "epoch": 46, "lr": 9.995542649467515e-05}
{"train_loss": 1.0273066759109497, "global_step": 4142, "epoch": 46, "lr": 9.995540201746684e-05}
{"train_loss": 1.259255290031433, "global_step": 4143, "epoch": 46, "lr": 9.995537753354266e-05}
{"train_loss": 1.0072087049484253, "global_step": 4144, "epoch": 46, "lr": 9.995535304290257e-05}
{"train_loss": 0.9735201001167297, "global_step": 4145, "epoch": 46, "lr": 9.99553285455466e-05}
{"train_loss": 1.0994051694869995, "global_step": 4146, "epoch": 46, "lr": 9.995530404147475e-05}
{"train_loss": 0.8052501082420349, "global_step": 4147, "epoch": 46, "lr": 9.995527953068701e-05}
{"train_loss": 1.1139211654663086, "global_step": 4148, "epoch": 46, "lr": 9.99552550131834e-05}
{"train_loss": 0.855643093585968, "global_step": 4149, "epoch": 46, "lr": 9.99552304889639e-05}
{"train_loss": 1.1032999753952026, "global_step": 4150, "epoch": 46, "lr": 9.995520595802853e-05}
{"train_loss": 0.954077422618866, "global_step": 4151, "epoch": 46, "lr": 9.99551814203773e-05}
{"train_loss": 1.0725154876708984, "global_step": 4152, "epoch": 46, "lr": 9.99551568760102e-05}
{"train_loss": 1.1919275522232056, "global_step": 4153, "epoch": 46, "lr": 9.995513232492723e-05}
{"train_loss": 1.0396792888641357, "global_step": 4154, "epoch": 46, "lr": 9.995510776712842e-05}
{"train_loss": 0.9990805387496948, "global_step": 4155, "epoch": 46, "lr": 9.995508320261375e-05}
{"train_loss": 0.8852612972259521, "global_step": 4156, "epoch": 46, "lr": 9.995505863138321e-05}
{"train_loss": 0.948157012462616, "global_step": 4157, "epoch": 46, "lr": 9.995503405343684e-05}
{"train_loss": 0.9376019835472107, "global_step": 4158, "epoch": 46, "lr": 9.995500946877462e-05}
{"train_loss": 1.0520011186599731, "global_step": 4159, "epoch": 46, "lr": 9.995498487739657e-05}
{"train_loss": 0.9773545265197754, "global_step": 4160, "epoch": 46, "lr": 9.995496027930265e-05}
{"train_loss": 1.0834105014801025, "global_step": 4161, "epoch": 46, "lr": 9.995493567449292e-05}
{"train_loss": 1.0021063089370728, "global_step": 4162, "epoch": 46, "lr": 9.995491106296735e-05}
{"train_loss": 1.0602953433990479, "global_step": 4163, "epoch": 46, "lr": 9.995488644472594e-05}
{"train_loss": 0.9811870455741882, "global_step": 4164, "epoch": 46, "lr": 9.995486181976872e-05}
{"train_loss": 1.0217314958572388, "global_step": 4165, "epoch": 46, "lr": 9.995483718809568e-05}
{"train_loss": 1.0104539394378662, "global_step": 4166, "epoch": 46, "lr": 9.995481254970681e-05}
{"train_loss": 0.9849475622177124, "global_step": 4167, "epoch": 46, "lr": 9.995478790460213e-05}
{"train_loss": 1.1533364057540894, "global_step": 4168, "epoch": 46, "lr": 9.995476325278164e-05}
{"train_loss": 1.1354289054870605, "global_step": 4169, "epoch": 46, "lr": 9.995473859424533e-05}
{"train_loss": 0.9719168543815613, "global_step": 4170, "epoch": 46, "lr": 9.995471392899322e-05}
{"train_loss": 0.9964295029640198, "global_step": 4171, "epoch": 46, "lr": 9.995468925702531e-05}
{"train_loss": 0.8507246375083923, "global_step": 4172, "epoch": 46, "lr": 9.995466457834159e-05}
{"train_loss": 1.0702917575836182, "global_step": 4173, "epoch": 46, "lr": 9.995463989294209e-05}
{"train_loss": 0.920962929725647, "global_step": 4174, "epoch": 46, "lr": 9.995461520082679e-05}
{"train_loss": 1.1858017444610596, "global_step": 4175, "epoch": 46, "lr": 9.995459050199572e-05}
{"train_loss": 1.0626084804534912, "global_step": 4176, "epoch": 46, "lr": 9.995456579644884e-05}
{"train_loss": 0.9993903636932373, "global_step": 4177, "epoch": 46, "lr": 9.995454108418618e-05}
{"train_loss": 1.0430773496627808, "global_step": 4178, "epoch": 46, "lr": 9.995451636520774e-05}
{"train_loss": 1.2853124141693115, "global_step": 4179, "epoch": 46, "lr": 9.995449163951354e-05}
{"train_loss": 0.9469786882400513, "global_step": 4180, "epoch": 46, "lr": 9.995446690710356e-05}
{"train_loss": 1.102676272392273, "global_step": 4181, "epoch": 46, "lr": 9.995444216797781e-05}
{"train_loss": 1.0332224369049072, "global_step": 4182, "epoch": 46, "lr": 9.995441742213627e-05, "val_loss": 1.1533313989639282}
{"train_loss": 1.013501763343811, "global_step": 4183, "epoch": 47, "lr": 9.9954392669579e-05}
{"train_loss": 0.9073970317840576, "global_step": 4184, "epoch": 47, "lr": 9.995436791030595e-05}
{"train_loss": 1.1359572410583496, "global_step": 4185, "epoch": 47, "lr": 9.995434314431716e-05}
{"train_loss": 1.1454561948776245, "global_step": 4186, "epoch": 47, "lr": 9.995431837161261e-05}
{"train_loss": 1.0469225645065308, "global_step": 4187, "epoch": 47, "lr": 9.995429359219232e-05}
{"train_loss": 0.9079652428627014, "global_step": 4188, "epoch": 47, "lr": 9.995426880605627e-05}
{"train_loss": 1.1743483543395996, "global_step": 4189, "epoch": 47, "lr": 9.995424401320448e-05}
{"train_loss": 1.068872332572937, "global_step": 4190, "epoch": 47, "lr": 9.995421921363696e-05}
{"train_loss": 0.9554386138916016, "global_step": 4191, "epoch": 47, "lr": 9.99541944073537e-05}
{"train_loss": 0.950886070728302, "global_step": 4192, "epoch": 47, "lr": 9.995416959435471e-05}
{"train_loss": 1.0016320943832397, "global_step": 4193, "epoch": 47, "lr": 9.995414477463997e-05}
{"train_loss": 0.875919759273529, "global_step": 4194, "epoch": 47, "lr": 9.995411994820954e-05}
{"train_loss": 1.1417748928070068, "global_step": 4195, "epoch": 47, "lr": 9.995409511506335e-05}
{"train_loss": 0.9646599292755127, "global_step": 4196, "epoch": 47, "lr": 9.995407027520147e-05}
{"train_loss": 0.9436089992523193, "global_step": 4197, "epoch": 47, "lr": 9.995404542862386e-05}
{"train_loss": 1.0994936227798462, "global_step": 4198, "epoch": 47, "lr": 9.995402057533054e-05}
{"train_loss": 1.0674794912338257, "global_step": 4199, "epoch": 47, "lr": 9.995399571532151e-05}
{"train_loss": 1.0397956371307373, "global_step": 4200, "epoch": 47, "lr": 9.995397084859677e-05}
{"train_loss": 0.9246767163276672, "global_step": 4201, "epoch": 47, "lr": 9.995394597515633e-05}
{"train_loss": 1.002150058746338, "global_step": 4202, "epoch": 47, "lr": 9.99539210950002e-05}
{"train_loss": 0.9878501296043396, "global_step": 4203, "epoch": 47, "lr": 9.995389620812836e-05}
{"train_loss": 1.0072343349456787, "global_step": 4204, "epoch": 47, "lr": 9.995387131454083e-05}
{"train_loss": 0.9618363976478577, "global_step": 4205, "epoch": 47, "lr": 9.995384641423763e-05}
{"train_loss": 1.064491868019104, "global_step": 4206, "epoch": 47, "lr": 9.995382150721872e-05}
{"train_loss": 0.9760273694992065, "global_step": 4207, "epoch": 47, "lr": 9.995379659348414e-05}
{"train_loss": 1.0241291522979736, "global_step": 4208, "epoch": 47, "lr": 9.995377167303389e-05}
{"train_loss": 0.8486056327819824, "global_step": 4209, "epoch": 47, "lr": 9.995374674586795e-05}
{"train_loss": 1.0661077499389648, "global_step": 4210, "epoch": 47, "lr": 9.995372181198634e-05}
{"train_loss": 0.9123077392578125, "global_step": 4211, "epoch": 47, "lr": 9.995369687138905e-05}
{"train_loss": 1.0585253238677979, "global_step": 4212, "epoch": 47, "lr": 9.995367192407612e-05}
{"train_loss": 1.1305218935012817, "global_step": 4213, "epoch": 47, "lr": 9.99536469700475e-05}
{"train_loss": 1.0571355819702148, "global_step": 4214, "epoch": 47, "lr": 9.995362200930325e-05}
{"train_loss": 0.9200886487960815, "global_step": 4215, "epoch": 47, "lr": 9.995359704184333e-05}
{"train_loss": 1.0105165243148804, "global_step": 4216, "epoch": 47, "lr": 9.995357206766776e-05}
{"train_loss": 1.1700369119644165, "global_step": 4217, "epoch": 47, "lr": 9.995354708677654e-05}
{"train_loss": 1.1243486404418945, "global_step": 4218, "epoch": 47, "lr": 9.995352209916967e-05}
{"train_loss": 1.10566246509552, "global_step": 4219, "epoch": 47, "lr": 9.995349710484715e-05}
{"train_loss": 1.0491794347763062, "global_step": 4220, "epoch": 47, "lr": 9.9953472103809e-05}
{"train_loss": 0.9253031611442566, "global_step": 4221, "epoch": 47, "lr": 9.99534470960552e-05}
{"train_loss": 1.036643624305725, "global_step": 4222, "epoch": 47, "lr": 9.99534220815858e-05}
{"train_loss": 1.1074107885360718, "global_step": 4223, "epoch": 47, "lr": 9.995339706040075e-05}
{"train_loss": 0.9303491711616516, "global_step": 4224, "epoch": 47, "lr": 9.995337203250009e-05}
{"train_loss": 0.9779492616653442, "global_step": 4225, "epoch": 47, "lr": 9.995334699788379e-05}
{"train_loss": 1.0942258834838867, "global_step": 4226, "epoch": 47, "lr": 9.995332195655187e-05}
{"train_loss": 1.0341421365737915, "global_step": 4227, "epoch": 47, "lr": 9.995329690850434e-05}
{"train_loss": 1.0307387113571167, "global_step": 4228, "epoch": 47, "lr": 9.995327185374122e-05}
{"train_loss": 1.169585108757019, "global_step": 4229, "epoch": 47, "lr": 9.995324679226246e-05}
{"train_loss": 1.1087393760681152, "global_step": 4230, "epoch": 47, "lr": 9.995322172406811e-05}
{"train_loss": 0.848942220211029, "global_step": 4231, "epoch": 47, "lr": 9.995319664915816e-05}
{"train_loss": 1.0290971994400024, "global_step": 4232, "epoch": 47, "lr": 9.995317156753261e-05}
{"train_loss": 1.1161381006240845, "global_step": 4233, "epoch": 47, "lr": 9.995314647919146e-05}
{"train_loss": 1.2563058137893677, "global_step": 4234, "epoch": 47, "lr": 9.995312138413471e-05}
{"train_loss": 1.0879876613616943, "global_step": 4235, "epoch": 47, "lr": 9.99530962823624e-05}
{"train_loss": 1.003722071647644, "global_step": 4236, "epoch": 47, "lr": 9.995307117387448e-05}
{"train_loss": 1.1247180700302124, "global_step": 4237, "epoch": 47, "lr": 9.995304605867098e-05}
{"train_loss": 1.0214684009552002, "global_step": 4238, "epoch": 47, "lr": 9.995302093675192e-05}
{"train_loss": 1.3175256252288818, "global_step": 4239, "epoch": 47, "lr": 9.995299580811726e-05}
{"train_loss": 1.1090445518493652, "global_step": 4240, "epoch": 47, "lr": 9.995297067276704e-05}
{"train_loss": 0.9956111311912537, "global_step": 4241, "epoch": 47, "lr": 9.995294553070126e-05}
{"train_loss": 1.144515872001648, "global_step": 4242, "epoch": 47, "lr": 9.99529203819199e-05}
{"train_loss": 1.1226829290390015, "global_step": 4243, "epoch": 47, "lr": 9.995289522642299e-05}
{"train_loss": 1.2036782503128052, "global_step": 4244, "epoch": 47, "lr": 9.995287006421052e-05}
{"train_loss": 1.0833425521850586, "global_step": 4245, "epoch": 47, "lr": 9.995284489528249e-05}
{"train_loss": 1.0651599168777466, "global_step": 4246, "epoch": 47, "lr": 9.995281971963892e-05}
{"train_loss": 1.0000412464141846, "global_step": 4247, "epoch": 47, "lr": 9.99527945372798e-05}
{"train_loss": 1.0690933465957642, "global_step": 4248, "epoch": 47, "lr": 9.995276934820512e-05}
{"train_loss": 1.009645700454712, "global_step": 4249, "epoch": 47, "lr": 9.995274415241491e-05}
{"train_loss": 0.9949353337287903, "global_step": 4250, "epoch": 47, "lr": 9.995271894990916e-05}
{"train_loss": 1.023503303527832, "global_step": 4251, "epoch": 47, "lr": 9.99526937406879e-05}
{"train_loss": 1.0783485174179077, "global_step": 4252, "epoch": 47, "lr": 9.995266852475107e-05}
{"train_loss": 0.9487351775169373, "global_step": 4253, "epoch": 47, "lr": 9.995264330209873e-05}
{"train_loss": 0.900539755821228, "global_step": 4254, "epoch": 47, "lr": 9.995261807273088e-05}
{"train_loss": 0.8720723390579224, "global_step": 4255, "epoch": 47, "lr": 9.995259283664748e-05}
{"train_loss": 0.9495986700057983, "global_step": 4256, "epoch": 47, "lr": 9.995256759384858e-05}
{"train_loss": 1.3646793365478516, "global_step": 4257, "epoch": 47, "lr": 9.995254234433418e-05}
{"train_loss": 1.0532736778259277, "global_step": 4258, "epoch": 47, "lr": 9.995251708810425e-05}
{"train_loss": 1.1900649070739746, "global_step": 4259, "epoch": 47, "lr": 9.995249182515881e-05}
{"train_loss": 1.1503061056137085, "global_step": 4260, "epoch": 47, "lr": 9.995246655549788e-05}
{"train_loss": 1.0802069902420044, "global_step": 4261, "epoch": 47, "lr": 9.995244127912145e-05}
{"train_loss": 0.9919770956039429, "global_step": 4262, "epoch": 47, "lr": 9.99524159960295e-05}
{"train_loss": 1.08251953125, "global_step": 4263, "epoch": 47, "lr": 9.995239070622208e-05}
{"train_loss": 0.9296177625656128, "global_step": 4264, "epoch": 47, "lr": 9.995236540969917e-05}
{"train_loss": 0.9330583214759827, "global_step": 4265, "epoch": 47, "lr": 9.995234010646076e-05}
{"train_loss": 1.0644813776016235, "global_step": 4266, "epoch": 47, "lr": 9.995231479650686e-05}
{"train_loss": 0.9679595232009888, "global_step": 4267, "epoch": 47, "lr": 9.99522894798375e-05}
{"train_loss": 0.9896316528320312, "global_step": 4268, "epoch": 47, "lr": 9.995226415645266e-05}
{"train_loss": 0.7446984052658081, "global_step": 4269, "epoch": 47, "lr": 9.995223882635234e-05}
{"train_loss": 1.0012325048446655, "global_step": 4270, "epoch": 47, "lr": 9.995221348953655e-05}
{"train_loss": 1.0388788808597607, "global_step": 4271, "epoch": 47, "lr": 9.99521881460053e-05, "val_loss": 1.1608412265777588}
{"train_loss": 1.0988030433654785, "global_step": 4272, "epoch": 48, "lr": 9.995216279575858e-05}
{"train_loss": 0.8779818415641785, "global_step": 4273, "epoch": 48, "lr": 9.995213743879641e-05}
{"train_loss": 0.916654646396637, "global_step": 4274, "epoch": 48, "lr": 9.995211207511878e-05}
{"train_loss": 1.0501034259796143, "global_step": 4275, "epoch": 48, "lr": 9.995208670472568e-05}
{"train_loss": 0.9632949829101562, "global_step": 4276, "epoch": 48, "lr": 9.995206132761716e-05}
{"train_loss": 1.0181078910827637, "global_step": 4277, "epoch": 48, "lr": 9.995203594379318e-05}
{"train_loss": 1.184775710105896, "global_step": 4278, "epoch": 48, "lr": 9.995201055325377e-05}
{"train_loss": 0.8109925985336304, "global_step": 4279, "epoch": 48, "lr": 9.995198515599891e-05}
{"train_loss": 1.0553653240203857, "global_step": 4280, "epoch": 48, "lr": 9.99519597520286e-05}
{"train_loss": 0.8221057057380676, "global_step": 4281, "epoch": 48, "lr": 9.995193434134287e-05}
{"train_loss": 1.0888288021087646, "global_step": 4282, "epoch": 48, "lr": 9.995190892394172e-05}
{"train_loss": 1.2184810638427734, "global_step": 4283, "epoch": 48, "lr": 9.995188349982514e-05}
{"train_loss": 0.9789942502975464, "global_step": 4284, "epoch": 48, "lr": 9.995185806899313e-05}
{"train_loss": 1.0535300970077515, "global_step": 4285, "epoch": 48, "lr": 9.99518326314457e-05}
{"train_loss": 0.8674891591072083, "global_step": 4286, "epoch": 48, "lr": 9.995180718718286e-05}
{"train_loss": 0.9118052124977112, "global_step": 4287, "epoch": 48, "lr": 9.995178173620462e-05}
{"train_loss": 1.3302220106124878, "global_step": 4288, "epoch": 48, "lr": 9.995175627851097e-05}
{"train_loss": 1.0190666913986206, "global_step": 4289, "epoch": 48, "lr": 9.99517308141019e-05}
{"train_loss": 1.0307109355926514, "global_step": 4290, "epoch": 48, "lr": 9.995170534297744e-05}
{"train_loss": 1.0472372770309448, "global_step": 4291, "epoch": 48, "lr": 9.995167986513758e-05}
{"train_loss": 1.0184937715530396, "global_step": 4292, "epoch": 48, "lr": 9.995165438058232e-05}
{"train_loss": 1.2678149938583374, "global_step": 4293, "epoch": 48, "lr": 9.995162888931167e-05}
{"train_loss": 1.1640042066574097, "global_step": 4294, "epoch": 48, "lr": 9.995160339132564e-05}
{"train_loss": 0.923799991607666, "global_step": 4295, "epoch": 48, "lr": 9.995157788662422e-05}
{"train_loss": 0.9829433560371399, "global_step": 4296, "epoch": 48, "lr": 9.995155237520742e-05}
{"train_loss": 0.999421238899231, "global_step": 4297, "epoch": 48, "lr": 9.995152685707524e-05}
{"train_loss": 1.091790795326233, "global_step": 4298, "epoch": 48, "lr": 9.995150133222768e-05}
{"train_loss": 1.2951241731643677, "global_step": 4299, "epoch": 48, "lr": 9.995147580066477e-05}
{"train_loss": 0.9039832949638367, "global_step": 4300, "epoch": 48, "lr": 9.995145026238648e-05}
{"train_loss": 1.0049740076065063, "global_step": 4301, "epoch": 48, "lr": 9.995142471739283e-05}
{"train_loss": 1.0516209602355957, "global_step": 4302, "epoch": 48, "lr": 9.995139916568382e-05}
{"train_loss": 1.03130304813385, "global_step": 4303, "epoch": 48, "lr": 9.995137360725943e-05}
{"train_loss": 1.0413672924041748, "global_step": 4304, "epoch": 48, "lr": 9.995134804211972e-05}
{"train_loss": 0.9264423847198486, "global_step": 4305, "epoch": 48, "lr": 9.995132247026464e-05}
{"train_loss": 1.0210859775543213, "global_step": 4306, "epoch": 48, "lr": 9.995129689169422e-05}
{"train_loss": 1.0793254375457764, "global_step": 4307, "epoch": 48, "lr": 9.995127130640847e-05}
{"train_loss": 0.9952173233032227, "global_step": 4308, "epoch": 48, "lr": 9.995124571440736e-05}
{"train_loss": 1.1943156719207764, "global_step": 4309, "epoch": 48, "lr": 9.995122011569093e-05}
{"train_loss": 1.0242379903793335, "global_step": 4310, "epoch": 48, "lr": 9.995119451025916e-05}
{"train_loss": 0.8212441205978394, "global_step": 4311, "epoch": 48, "lr": 9.995116889811204e-05}
{"train_loss": 0.8842837810516357, "global_step": 4312, "epoch": 48, "lr": 9.995114327924962e-05}
{"train_loss": 1.011184573173523, "global_step": 4313, "epoch": 48, "lr": 9.995111765367187e-05}
{"train_loss": 0.9513254761695862, "global_step": 4314, "epoch": 48, "lr": 9.99510920213788e-05}
{"train_loss": 0.9330438375473022, "global_step": 4315, "epoch": 48, "lr": 9.995106638237042e-05}
{"train_loss": 1.0254578590393066, "global_step": 4316, "epoch": 48, "lr": 9.995104073664673e-05}
{"train_loss": 0.9472814202308655, "global_step": 4317, "epoch": 48, "lr": 9.995101508420772e-05}
{"train_loss": 1.1051150560379028, "global_step": 4318, "epoch": 48, "lr": 9.995098942505341e-05}
{"train_loss": 0.7872913479804993, "global_step": 4319, "epoch": 48, "lr": 9.99509637591838e-05}
{"train_loss": 1.0143423080444336, "global_step": 4320, "epoch": 48, "lr": 9.99509380865989e-05}
{"train_loss": 0.9761877655982971, "global_step": 4321, "epoch": 48, "lr": 9.99509124072987e-05}
{"train_loss": 1.026104211807251, "global_step": 4322, "epoch": 48, "lr": 9.99508867212832e-05}
{"train_loss": 0.9684111475944519, "global_step": 4323, "epoch": 48, "lr": 9.995086102855243e-05}
{"train_loss": 1.0050055980682373, "global_step": 4324, "epoch": 48, "lr": 9.995083532910634e-05}
{"train_loss": 0.8813031911849976, "global_step": 4325, "epoch": 48, "lr": 9.995080962294501e-05}
{"train_loss": 1.0511530637741089, "global_step": 4326, "epoch": 48, "lr": 9.995078391006839e-05}
{"train_loss": 0.9062142968177795, "global_step": 4327, "epoch": 48, "lr": 9.995075819047648e-05}
{"train_loss": 1.0104645490646362, "global_step": 4328, "epoch": 48, "lr": 9.99507324641693e-05}
{"train_loss": 1.2427119016647339, "global_step": 4329, "epoch": 48, "lr": 9.995070673114686e-05}
{"train_loss": 0.9969257712364197, "global_step": 4330, "epoch": 48, "lr": 9.995068099140917e-05}
{"train_loss": 1.0820258855819702, "global_step": 4331, "epoch": 48, "lr": 9.99506552449562e-05}
{"train_loss": 0.9428697228431702, "global_step": 4332, "epoch": 48, "lr": 9.995062949178798e-05}
{"train_loss": 1.174411654472351, "global_step": 4333, "epoch": 48, "lr": 9.99506037319045e-05}
{"train_loss": 1.1412616968154907, "global_step": 4334, "epoch": 48, "lr": 9.995057796530579e-05}
{"train_loss": 1.0486595630645752, "global_step": 4335, "epoch": 48, "lr": 9.995055219199183e-05}
{"train_loss": 0.9468510150909424, "global_step": 4336, "epoch": 48, "lr": 9.995052641196261e-05}
{"train_loss": 0.8718605041503906, "global_step": 4337, "epoch": 48, "lr": 9.995050062521816e-05}
{"train_loss": 1.1325308084487915, "global_step": 4338, "epoch": 48, "lr": 9.995047483175848e-05}
{"train_loss": 1.1433926820755005, "global_step": 4339, "epoch": 48, "lr": 9.995044903158355e-05}
{"train_loss": 1.1304006576538086, "global_step": 4340, "epoch": 48, "lr": 9.995042322469341e-05}
{"train_loss": 1.013000249862671, "global_step": 4341, "epoch": 48, "lr": 9.995039741108802e-05}
{"train_loss": 0.9712119698524475, "global_step": 4342, "epoch": 48, "lr": 9.995037159076744e-05}
{"train_loss": 1.0059432983398438, "global_step": 4343, "epoch": 48, "lr": 9.995034576373163e-05}
{"train_loss": 1.0591620206832886, "global_step": 4344, "epoch": 48, "lr": 9.99503199299806e-05}
{"train_loss": 0.8679707646369934, "global_step": 4345, "epoch": 48, "lr": 9.995029408951435e-05}
{"train_loss": 0.8534979820251465, "global_step": 4346, "epoch": 48, "lr": 9.99502682423329e-05}
{"train_loss": 1.0627862215042114, "global_step": 4347, "epoch": 48, "lr": 9.995024238843624e-05}
{"train_loss": 1.0704940557479858, "global_step": 4348, "epoch": 48, "lr": 9.995021652782439e-05}
{"train_loss": 0.9042974710464478, "global_step": 4349, "epoch": 48, "lr": 9.995019066049734e-05}
{"train_loss": 1.192017912864685, "global_step": 4350, "epoch": 48, "lr": 9.995016478645509e-05}
{"train_loss": 1.0684698820114136, "global_step": 4351, "epoch": 48, "lr": 9.995013890569766e-05}
{"train_loss": 0.8679155111312866, "global_step": 4352, "epoch": 48, "lr": 9.995011301822503e-05}
{"train_loss": 0.9376412034034729, "global_step": 4353, "epoch": 48, "lr": 9.995008712403722e-05}
{"train_loss": 1.0429404973983765, "global_step": 4354, "epoch": 48, "lr": 9.995006122313423e-05}
{"train_loss": 0.979685366153717, "global_step": 4355, "epoch": 48, "lr": 9.995003531551606e-05}
{"train_loss": 1.046900749206543, "global_step": 4356, "epoch": 48, "lr": 9.995000940118272e-05}
{"train_loss": 0.9379954934120178, "global_step": 4357, "epoch": 48, "lr": 9.99499834801342e-05}
{"train_loss": 0.7745206952095032, "global_step": 4358, "epoch": 48, "lr": 9.994995755237053e-05}
{"train_loss": 1.068336844444275, "global_step": 4359, "epoch": 48, "lr": 9.994993161789169e-05}
{"train_loss": 1.0161120469650526, "global_step": 4360, "epoch": 48, "lr": 9.994990567669769e-05, "val_loss": 1.1840312480926514}
{"train_loss": 0.994215190410614, "global_step": 4361, "epoch": 49, "lr": 9.994987972878853e-05}
{"train_loss": 1.06046462059021, "global_step": 4362, "epoch": 49, "lr": 9.994985377416423e-05}
{"train_loss": 1.2152950763702393, "global_step": 4363, "epoch": 49, "lr": 9.994982781282477e-05}
{"train_loss": 1.051272988319397, "global_step": 4364, "epoch": 49, "lr": 9.994980184477016e-05}
{"train_loss": 0.8652649521827698, "global_step": 4365, "epoch": 49, "lr": 9.994977587000042e-05}
{"train_loss": 1.0498409271240234, "global_step": 4366, "epoch": 49, "lr": 9.994974988851554e-05}
{"train_loss": 0.8419132232666016, "global_step": 4367, "epoch": 49, "lr": 9.994972390031552e-05}
{"train_loss": 1.1981483697891235, "global_step": 4368, "epoch": 49, "lr": 9.994969790540037e-05}
{"train_loss": 0.855951189994812, "global_step": 4369, "epoch": 49, "lr": 9.994967190377009e-05}
{"train_loss": 0.9826449751853943, "global_step": 4370, "epoch": 49, "lr": 9.99496458954247e-05}
{"train_loss": 1.0501067638397217, "global_step": 4371, "epoch": 49, "lr": 9.994961988036416e-05}
{"train_loss": 1.1018226146697998, "global_step": 4372, "epoch": 49, "lr": 9.994959385858853e-05}
{"train_loss": 0.9493529796600342, "global_step": 4373, "epoch": 49, "lr": 9.994956783009776e-05}
{"train_loss": 1.1081089973449707, "global_step": 4374, "epoch": 49, "lr": 9.994954179489191e-05}
{"train_loss": 1.039689064025879, "global_step": 4375, "epoch": 49, "lr": 9.994951575297093e-05}
{"train_loss": 1.0522137880325317, "global_step": 4376, "epoch": 49, "lr": 9.994948970433485e-05}
{"train_loss": 0.9705895185470581, "global_step": 4377, "epoch": 49, "lr": 9.994946364898368e-05}
{"train_loss": 1.0939918756484985, "global_step": 4378, "epoch": 49, "lr": 9.99494375869174e-05}
{"train_loss": 1.1488902568817139, "global_step": 4379, "epoch": 49, "lr": 9.994941151813604e-05}
{"train_loss": 1.2027537822723389, "global_step": 4380, "epoch": 49, "lr": 9.994938544263959e-05}
{"train_loss": 1.05118727684021, "global_step": 4381, "epoch": 49, "lr": 9.994935936042804e-05}
{"train_loss": 1.0182199478149414, "global_step": 4382, "epoch": 49, "lr": 9.994933327150142e-05}
{"train_loss": 0.8365861177444458, "global_step": 4383, "epoch": 49, "lr": 9.99493071758597e-05}
{"train_loss": 0.893782913684845, "global_step": 4384, "epoch": 49, "lr": 9.994928107350293e-05}
{"train_loss": 0.9852393865585327, "global_step": 4385, "epoch": 49, "lr": 9.994925496443108e-05}
{"train_loss": 1.0493412017822266, "global_step": 4386, "epoch": 49, "lr": 9.994922884864416e-05}
{"train_loss": 1.0208840370178223, "global_step": 4387, "epoch": 49, "lr": 9.994920272614218e-05}
{"train_loss": 0.9628089070320129, "global_step": 4388, "epoch": 49, "lr": 9.994917659692513e-05}
{"train_loss": 1.1186487674713135, "global_step": 4389, "epoch": 49, "lr": 9.994915046099302e-05}
{"train_loss": 1.0133864879608154, "global_step": 4390, "epoch": 49, "lr": 9.994912431834586e-05}
{"train_loss": 0.9672347903251648, "global_step": 4391, "epoch": 49, "lr": 9.994909816898366e-05}
{"train_loss": 1.1189050674438477, "global_step": 4392, "epoch": 49, "lr": 9.99490720129064e-05}
{"train_loss": 1.0327528715133667, "global_step": 4393, "epoch": 49, "lr": 9.99490458501141e-05}
{"train_loss": 1.0386277437210083, "global_step": 4394, "epoch": 49, "lr": 9.994901968060676e-05}
{"train_loss": 1.0933964252471924, "global_step": 4395, "epoch": 49, "lr": 9.994899350438438e-05}
{"train_loss": 1.0966787338256836, "global_step": 4396, "epoch": 49, "lr": 9.994896732144697e-05}
{"train_loss": 1.13645601272583, "global_step": 4397, "epoch": 49, "lr": 9.994894113179453e-05}
{"train_loss": 1.0785716772079468, "global_step": 4398, "epoch": 49, "lr": 9.994891493542706e-05}
{"train_loss": 0.9285169839859009, "global_step": 4399, "epoch": 49, "lr": 9.994888873234455e-05}
{"train_loss": 0.8712635040283203, "global_step": 4400, "epoch": 49, "lr": 9.994886252254705e-05}
{"train_loss": 0.9703087210655212, "global_step": 4401, "epoch": 49, "lr": 9.994883630603453e-05}
{"train_loss": 1.091792345046997, "global_step": 4402, "epoch": 49, "lr": 9.994881008280699e-05}
{"train_loss": 0.9989305734634399, "global_step": 4403, "epoch": 49, "lr": 9.994878385286444e-05}
{"train_loss": 0.96433424949646, "global_step": 4404, "epoch": 49, "lr": 9.994875761620688e-05}
{"train_loss": 0.9425177574157715, "global_step": 4405, "epoch": 49, "lr": 9.994873137283434e-05}
{"train_loss": 1.0324323177337646, "global_step": 4406, "epoch": 49, "lr": 9.994870512274678e-05}
{"train_loss": 0.9355407357215881, "global_step": 4407, "epoch": 49, "lr": 9.994867886594424e-05}
{"train_loss": 1.2267457246780396, "global_step": 4408, "epoch": 49, "lr": 9.99486526024267e-05}
{"train_loss": 1.1918659210205078, "global_step": 4409, "epoch": 49, "lr": 9.994862633219417e-05}
{"train_loss": 1.1310200691223145, "global_step": 4410, "epoch": 49, "lr": 9.994860005524666e-05}
{"train_loss": 0.8558973670005798, "global_step": 4411, "epoch": 49, "lr": 9.994857377158417e-05}
{"train_loss": 0.9956265687942505, "global_step": 4412, "epoch": 49, "lr": 9.994854748120672e-05}
{"train_loss": 1.0676660537719727, "global_step": 4413, "epoch": 49, "lr": 9.994852118411426e-05}
{"train_loss": 1.1372132301330566, "global_step": 4414, "epoch": 49, "lr": 9.994849488030686e-05}
{"train_loss": 1.0639209747314453, "global_step": 4415, "epoch": 49, "lr": 9.994846856978448e-05}
{"train_loss": 0.8591482043266296, "global_step": 4416, "epoch": 49, "lr": 9.994844225254714e-05}
{"train_loss": 1.134877324104309, "global_step": 4417, "epoch": 49, "lr": 9.994841592859486e-05}
{"train_loss": 0.9951401352882385, "global_step": 4418, "epoch": 49, "lr": 9.99483895979276e-05}
{"train_loss": 0.9779409766197205, "global_step": 4419, "epoch": 49, "lr": 9.994836326054538e-05}
{"train_loss": 0.9344500303268433, "global_step": 4420, "epoch": 49, "lr": 9.994833691644823e-05}
{"train_loss": 0.9020078182220459, "global_step": 4421, "epoch": 49, "lr": 9.994831056563613e-05}
{"train_loss": 0.9676207304000854, "global_step": 4422, "epoch": 49, "lr": 9.994828420810909e-05}
{"train_loss": 1.0294615030288696, "global_step": 4423, "epoch": 49, "lr": 9.99482578438671e-05}
{"train_loss": 0.8724108338356018, "global_step": 4424, "epoch": 49, "lr": 9.994823147291019e-05}
{"train_loss": 0.982291579246521, "global_step": 4425, "epoch": 49, "lr": 9.994820509523834e-05}
{"train_loss": 0.9095684885978699, "global_step": 4426, "epoch": 49, "lr": 9.994817871085158e-05}
{"train_loss": 0.8020373582839966, "global_step": 4427, "epoch": 49, "lr": 9.994815231974987e-05}
{"train_loss": 0.9712204933166504, "global_step": 4428, "epoch": 49, "lr": 9.994812592193325e-05}
{"train_loss": 1.0351102352142334, "global_step": 4429, "epoch": 49, "lr": 9.994809951740172e-05}
{"train_loss": 0.9310628175735474, "global_step": 4430, "epoch": 49, "lr": 9.994807310615528e-05}
{"train_loss": 0.9677287340164185, "global_step": 4431, "epoch": 49, "lr": 9.994804668819393e-05}
{"train_loss": 0.9597111344337463, "global_step": 4432, "epoch": 49, "lr": 9.994802026351766e-05}
{"train_loss": 0.9665225148200989, "global_step": 4433, "epoch": 49, "lr": 9.99479938321265e-05}
{"train_loss": 1.2307281494140625, "global_step": 4434, "epoch": 49, "lr": 9.994796739402044e-05}
{"train_loss": 0.9254530668258667, "global_step": 4435, "epoch": 49, "lr": 9.994794094919948e-05}
{"train_loss": 0.8989359140396118, "global_step": 4436, "epoch": 49, "lr": 9.994791449766363e-05}
{"train_loss": 0.8057204484939575, "global_step": 4437, "epoch": 49, "lr": 9.994788803941289e-05}
{"train_loss": 0.9504634141921997, "global_step": 4438, "epoch": 49, "lr": 9.994786157444727e-05}
{"train_loss": 0.9305070042610168, "global_step": 4439, "epoch": 49, "lr": 9.994783510276675e-05}
{"train_loss": 1.2006416320800781, "global_step": 4440, "epoch": 49, "lr": 9.994780862437136e-05}
{"train_loss": 1.0495963096618652, "global_step": 4441, "epoch": 49, "lr": 9.994778213926111e-05}
{"train_loss": 1.0896984338760376, "global_step": 4442, "epoch": 49, "lr": 9.994775564743599e-05}
{"train_loss": 1.0124613046646118, "global_step": 4443, "epoch": 49, "lr": 9.994772914889599e-05}
{"train_loss": 1.1738792657852173, "global_step": 4444, "epoch": 49, "lr": 9.994770264364113e-05}
{"train_loss": 0.9876947999000549, "global_step": 4445, "epoch": 49, "lr": 9.994767613167141e-05}
{"train_loss": 0.8928007483482361, "global_step": 4446, "epoch": 49, "lr": 9.994764961298684e-05}
{"train_loss": 0.894581139087677, "global_step": 4447, "epoch": 49, "lr": 9.994762308758741e-05}
{"train_loss": 0.9392285346984863, "global_step": 4448, "epoch": 49, "lr": 9.994759655547315e-05}
{"train_loss": 1.0125516827186842, "global_step": 4449, "epoch": 49, "lr": 9.9947570016644e-05, "val_loss": 1.0933157205581665}
{"train_loss": 1.1842046976089478, "global_step": 4450, "epoch": 50, "lr": 9.994754347110004e-05}
{"train_loss": 1.0043364763259888, "global_step": 4451, "epoch": 50, "lr": 9.994751691884125e-05}
{"train_loss": 0.8993092775344849, "global_step": 4452, "epoch": 50, "lr": 9.994749035986762e-05}
{"train_loss": 0.8788738250732422, "global_step": 4453, "epoch": 50, "lr": 9.994746379417914e-05}
{"train_loss": 1.0488593578338623, "global_step": 4454, "epoch": 50, "lr": 9.994743722177584e-05}
{"train_loss": 0.8872523903846741, "global_step": 4455, "epoch": 50, "lr": 9.994741064265773e-05}
{"train_loss": 0.7745944857597351, "global_step": 4456, "epoch": 50, "lr": 9.994738405682479e-05}
{"train_loss": 0.9581820964813232, "global_step": 4457, "epoch": 50, "lr": 9.994735746427703e-05}
{"train_loss": 1.1078921556472778, "global_step": 4458, "epoch": 50, "lr": 9.994733086501446e-05}
{"train_loss": 0.9143697619438171, "global_step": 4459, "epoch": 50, "lr": 9.994730425903708e-05}
{"train_loss": 1.1893364191055298, "global_step": 4460, "epoch": 50, "lr": 9.994727764634488e-05}
{"train_loss": 0.8738645315170288, "global_step": 4461, "epoch": 50, "lr": 9.99472510269379e-05}
{"train_loss": 1.0150752067565918, "global_step": 4462, "epoch": 50, "lr": 9.994722440081611e-05}
{"train_loss": 1.1946839094161987, "global_step": 4463, "epoch": 50, "lr": 9.994719776797952e-05}
{"train_loss": 1.2723289728164673, "global_step": 4464, "epoch": 50, "lr": 9.994717112842816e-05}
{"train_loss": 1.057753562927246, "global_step": 4465, "epoch": 50, "lr": 9.994714448216198e-05}
{"train_loss": 1.0588428974151611, "global_step": 4466, "epoch": 50, "lr": 9.994711782918102e-05}
{"train_loss": 1.0396575927734375, "global_step": 4467, "epoch": 50, "lr": 9.99470911694853e-05}
{"train_loss": 1.0194425582885742, "global_step": 4468, "epoch": 50, "lr": 9.994706450307478e-05}
{"train_loss": 0.8183555006980896, "global_step": 4469, "epoch": 50, "lr": 9.994703782994951e-05}
{"train_loss": 0.9071630239486694, "global_step": 4470, "epoch": 50, "lr": 9.994701115010945e-05}
{"train_loss": 0.97499018907547, "global_step": 4471, "epoch": 50, "lr": 9.994698446355464e-05}
{"train_loss": 1.0188788175582886, "global_step": 4472, "epoch": 50, "lr": 9.994695777028505e-05}
{"train_loss": 0.850947916507721, "global_step": 4473, "epoch": 50, "lr": 9.994693107030071e-05}
{"train_loss": 1.0250325202941895, "global_step": 4474, "epoch": 50, "lr": 9.99469043636016e-05}
{"train_loss": 1.019128441810608, "global_step": 4475, "epoch": 50, "lr": 9.994687765018775e-05}
{"train_loss": 1.0570508241653442, "global_step": 4476, "epoch": 50, "lr": 9.994685093005916e-05}
{"train_loss": 1.208038091659546, "global_step": 4477, "epoch": 50, "lr": 9.994682420321582e-05}
{"train_loss": 0.8776758313179016, "global_step": 4478, "epoch": 50, "lr": 9.994679746965771e-05}
{"train_loss": 1.1607277393341064, "global_step": 4479, "epoch": 50, "lr": 9.99467707293849e-05}
{"train_loss": 0.8209713101387024, "global_step": 4480, "epoch": 50, "lr": 9.994674398239734e-05}
{"train_loss": 1.0409082174301147, "global_step": 4481, "epoch": 50, "lr": 9.994671722869504e-05}
{"train_loss": 1.1438158750534058, "global_step": 4482, "epoch": 50, "lr": 9.994669046827802e-05}
{"train_loss": 1.1791058778762817, "global_step": 4483, "epoch": 50, "lr": 9.994666370114628e-05}
{"train_loss": 0.965913712978363, "global_step": 4484, "epoch": 50, "lr": 9.994663692729982e-05}
{"train_loss": 0.8309904932975769, "global_step": 4485, "epoch": 50, "lr": 9.994661014673864e-05}
{"train_loss": 0.9275161623954773, "global_step": 4486, "epoch": 50, "lr": 9.994658335946274e-05}
{"train_loss": 0.9124652147293091, "global_step": 4487, "epoch": 50, "lr": 9.994655656547214e-05}
{"train_loss": 0.911464512348175, "global_step": 4488, "epoch": 50, "lr": 9.994652976476684e-05}
{"train_loss": 0.953420102596283, "global_step": 4489, "epoch": 50, "lr": 9.994650295734683e-05}
{"train_loss": 0.9039812684059143, "global_step": 4490, "epoch": 50, "lr": 9.994647614321212e-05}
{"train_loss": 0.9810242056846619, "global_step": 4491, "epoch": 50, "lr": 9.994644932236272e-05}
{"train_loss": 0.9229355454444885, "global_step": 4492, "epoch": 50, "lr": 9.994642249479862e-05}
{"train_loss": 0.9344497919082642, "global_step": 4493, "epoch": 50, "lr": 9.994639566051985e-05}
{"train_loss": 0.9627548456192017, "global_step": 4494, "epoch": 50, "lr": 9.994636881952637e-05}
{"train_loss": 0.8646302223205566, "global_step": 4495, "epoch": 50, "lr": 9.994634197181822e-05}
{"train_loss": 1.1108015775680542, "global_step": 4496, "epoch": 50, "lr": 9.99463151173954e-05}
{"train_loss": 0.9851706624031067, "global_step": 4497, "epoch": 50, "lr": 9.99462882562579e-05}
{"train_loss": 0.9847578406333923, "global_step": 4498, "epoch": 50, "lr": 9.994626138840573e-05}
{"train_loss": 1.116475224494934, "global_step": 4499, "epoch": 50, "lr": 9.994623451383891e-05}
{"train_loss": 0.8105928897857666, "global_step": 4500, "epoch": 50, "lr": 9.994620763255741e-05}
{"train_loss": 0.9871162176132202, "global_step": 4501, "epoch": 50, "lr": 9.994618074456125e-05}
{"train_loss": 0.9710121750831604, "global_step": 4502, "epoch": 50, "lr": 9.994615384985044e-05}
{"train_loss": 1.1024775505065918, "global_step": 4503, "epoch": 50, "lr": 9.994612694842498e-05}
{"train_loss": 1.068814754486084, "global_step": 4504, "epoch": 50, "lr": 9.994610004028488e-05}
{"train_loss": 0.9205583930015564, "global_step": 4505, "epoch": 50, "lr": 9.994607312543012e-05}
{"train_loss": 0.9180088043212891, "global_step": 4506, "epoch": 50, "lr": 9.994604620386072e-05}
{"train_loss": 1.0283455848693848, "global_step": 4507, "epoch": 50, "lr": 9.99460192755767e-05}
{"train_loss": 0.8244686126708984, "global_step": 4508, "epoch": 50, "lr": 9.994599234057802e-05}
{"train_loss": 1.0472123622894287, "global_step": 4509, "epoch": 50, "lr": 9.994596539886474e-05}
{"train_loss": 0.8066372275352478, "global_step": 4510, "epoch": 50, "lr": 9.99459384504368e-05}
{"train_loss": 0.9612023830413818, "global_step": 4511, "epoch": 50, "lr": 9.994591149529426e-05}
{"train_loss": 0.9076938033103943, "global_step": 4512, "epoch": 50, "lr": 9.99458845334371e-05}
{"train_loss": 1.0223281383514404, "global_step": 4513, "epoch": 50, "lr": 9.994585756486532e-05}
{"train_loss": 1.0222876071929932, "global_step": 4514, "epoch": 50, "lr": 9.994583058957893e-05}
{"train_loss": 1.0144424438476562, "global_step": 4515, "epoch": 50, "lr": 9.994580360757794e-05}
{"train_loss": 1.1218822002410889, "global_step": 4516, "epoch": 50, "lr": 9.994577661886232e-05}
{"train_loss": 0.7955647706985474, "global_step": 4517, "epoch": 50, "lr": 9.994574962343213e-05}
{"train_loss": 0.9647982120513916, "global_step": 4518, "epoch": 50, "lr": 9.994572262128733e-05}
{"train_loss": 1.0558083057403564, "global_step": 4519, "epoch": 50, "lr": 9.994569561242794e-05}
{"train_loss": 1.1138776540756226, "global_step": 4520, "epoch": 50, "lr": 9.994566859685395e-05}
{"train_loss": 0.8152579665184021, "global_step": 4521, "epoch": 50, "lr": 9.994564157456539e-05}
{"train_loss": 0.8190392255783081, "global_step": 4522, "epoch": 50, "lr": 9.994561454556222e-05}
{"train_loss": 0.6672122478485107, "global_step": 4523, "epoch": 50, "lr": 9.994558750984449e-05}
{"train_loss": 1.0036497116088867, "global_step": 4524, "epoch": 50, "lr": 9.994556046741218e-05}
{"train_loss": 1.099168062210083, "global_step": 4525, "epoch": 50, "lr": 9.994553341826531e-05}
{"train_loss": 0.9742959141731262, "global_step": 4526, "epoch": 50, "lr": 9.994550636240385e-05}
{"train_loss": 1.0804523229599, "global_step": 4527, "epoch": 50, "lr": 9.994547929982783e-05}
{"train_loss": 0.9984044432640076, "global_step": 4528, "epoch": 50, "lr": 9.994545223053726e-05}
{"train_loss": 1.1184793710708618, "global_step": 4529, "epoch": 50, "lr": 9.994542515453214e-05}
{"train_loss": 0.9738448262214661, "global_step": 4530, "epoch": 50, "lr": 9.994539807181244e-05}
{"train_loss": 0.9754949808120728, "global_step": 4531, "epoch": 50, "lr": 9.994537098237822e-05}
{"train_loss": 1.184705138206482, "global_step": 4532, "epoch": 50, "lr": 9.994534388622942e-05}
{"train_loss": 0.9229212999343872, "global_step": 4533, "epoch": 50, "lr": 9.99453167833661e-05}
{"train_loss": 0.9843799471855164, "global_step": 4534, "epoch": 50, "lr": 9.994528967378824e-05}
{"train_loss": 0.9897885918617249, "global_step": 4535, "epoch": 50, "lr": 9.994526255749584e-05}
{"train_loss": 1.1562869548797607, "global_step": 4536, "epoch": 50, "lr": 9.994523543448891e-05}
{"train_loss": 0.9276906251907349, "global_step": 4537, "epoch": 50, "lr": 9.994520830476746e-05}
{"train_loss": 0.9902149880870005, "global_step": 4538, "epoch": 50, "lr": 9.994518116833147e-05, "train/sim_max_reward_0": 0.6922912476933741, "train/sim_max_reward_1": 0.3650060324433766, "train/sim_max_reward_2": 0.7525328271120579, "train/sim_max_reward_3": 0.029550621207624545, "train/sim_max_reward_4": 0.5177419950988384, "train/sim_max_reward_5": 0.5178983391953308, "test/sim_max_reward_4500000": 0.4083730431205814, "test/sim_max_reward_4500001": 0.15635468517503293, "test/sim_max_reward_4500002": 0.5443621566907999, "test/sim_max_reward_4500003": 0.6161889452880405, "test/sim_max_reward_4500004": 0.6406405906565715, "test/sim_max_reward_4500005": 0.4483866403017726, "test/sim_max_reward_4500006": 0.4639020207483642, "test/sim_max_reward_4500007": 0.5080990968018543, "test/sim_max_reward_4500008": 0.7126521873443715, "test/sim_max_reward_4500009": 0.05011328265319691, "test/sim_max_reward_4500010": 0.6068579993670355, "test/sim_max_reward_4500011": 0.9636829039597278, "test/sim_max_reward_4500012": 0.5258516649412358, "test/sim_max_reward_4500013": 0.5341688262338811, "test/sim_max_reward_4500014": 0.6301848764392622, "test/sim_max_reward_4500015": 0.3435846654165987, "test/sim_max_reward_4500016": 0.17411439316051203, "test/sim_max_reward_4500017": 0.5333059770909699, "test/sim_max_reward_4500018": 0.555990835432367, "test/sim_max_reward_4500019": 0.7489299055943643, "test/sim_max_reward_4500020": 0.3938340252713986, "test/sim_max_reward_4500021": 0.5659350813204462, "test/sim_max_reward_4500022": 0.421118596307843, "test/sim_max_reward_4500023": 0.5100606937217839, "test/sim_max_reward_4500024": 0.48164055606253736, "test/sim_max_reward_4500025": 0.295226405147396, "test/sim_max_reward_4500026": 0.0, "test/sim_max_reward_4500027": 0.1835367472410496, "test/sim_max_reward_4500028": 0.5054206982207796, "test/sim_max_reward_4500029": 0.2793998998717892, "test/sim_max_reward_4500030": 0.6094073753675298, "test/sim_max_reward_4500031": 0.429107783788196, "test/sim_max_reward_4500032": 0.2898783786221453, "test/sim_max_reward_4500033": 0.34075440958481434, "test/sim_max_reward_4500034": 0.8690860106502468, "test/sim_max_reward_4500035": 0.8547995109123785, "test/sim_max_reward_4500036": 0.573208704075381, "test/sim_max_reward_4500037": 0.14502011455521324, "test/sim_max_reward_4500038": 0.7442826751238394, "test/sim_max_reward_4500039": 0.37786073930660596, "test/sim_max_reward_4500040": 0.8357972913766153, "test/sim_max_reward_4500041": 0.4419290725202051, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.0, "test/sim_max_reward_4500044": 0.443688874950349, "test/sim_max_reward_4500045": 0.5335074712720138, "test/sim_max_reward_4500046": 0.6723769672064015, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9267338739217866, "test/sim_max_reward_4500049": 0.42799151528290236, "train/mean_score": 0.47917017712510046, "test/mean_score": 0.4663469633619637, "val_loss": 1.149796724319458, "train_action_mse_error": 52.61475372314453}
{"train_loss": 0.9648211002349854, "global_step": 4539, "epoch": 51, "lr": 9.994515402518096e-05}
{"train_loss": 1.0776760578155518, "global_step": 4540, "epoch": 51, "lr": 9.994512687531594e-05}
{"train_loss": 1.1998615264892578, "global_step": 4541, "epoch": 51, "lr": 9.994509971873641e-05}
{"train_loss": 1.0438259840011597, "global_step": 4542, "epoch": 51, "lr": 9.994507255544236e-05}
{"train_loss": 0.8681609630584717, "global_step": 4543, "epoch": 51, "lr": 9.994504538543381e-05}
{"train_loss": 0.7066220045089722, "global_step": 4544, "epoch": 51, "lr": 9.994501820871076e-05}
{"train_loss": 1.026766061782837, "global_step": 4545, "epoch": 51, "lr": 9.994499102527321e-05}
{"train_loss": 1.021772027015686, "global_step": 4546, "epoch": 51, "lr": 9.994496383512116e-05}
{"train_loss": 0.9581301808357239, "global_step": 4547, "epoch": 51, "lr": 9.994493663825462e-05}
{"train_loss": 1.0885462760925293, "global_step": 4548, "epoch": 51, "lr": 9.99449094346736e-05}
{"train_loss": 0.9253538250923157, "global_step": 4549, "epoch": 51, "lr": 9.994488222437808e-05}
{"train_loss": 0.9325107336044312, "global_step": 4550, "epoch": 51, "lr": 9.994485500736809e-05}
{"train_loss": 0.9179550409317017, "global_step": 4551, "epoch": 51, "lr": 9.994482778364362e-05}
{"train_loss": 0.9893641471862793, "global_step": 4552, "epoch": 51, "lr": 9.994480055320468e-05}
{"train_loss": 0.8457803130149841, "global_step": 4553, "epoch": 51, "lr": 9.994477331605127e-05}
{"train_loss": 1.0860953330993652, "global_step": 4554, "epoch": 51, "lr": 9.994474607218338e-05}
{"train_loss": 1.0827292203903198, "global_step": 4555, "epoch": 51, "lr": 9.994471882160104e-05}
{"train_loss": 1.0729634761810303, "global_step": 4556, "epoch": 51, "lr": 9.994469156430424e-05}
{"train_loss": 1.0863258838653564, "global_step": 4557, "epoch": 51, "lr": 9.994466430029299e-05}
{"train_loss": 1.020716667175293, "global_step": 4558, "epoch": 51, "lr": 9.994463702956728e-05}
{"train_loss": 0.970898449420929, "global_step": 4559, "epoch": 51, "lr": 9.994460975212713e-05}
{"train_loss": 0.8529950976371765, "global_step": 4560, "epoch": 51, "lr": 9.994458246797254e-05}
{"train_loss": 0.8956959247589111, "global_step": 4561, "epoch": 51, "lr": 9.99445551771035e-05}
{"train_loss": 0.8256528973579407, "global_step": 4562, "epoch": 51, "lr": 9.994452787952002e-05}
{"train_loss": 1.057479977607727, "global_step": 4563, "epoch": 51, "lr": 9.99445005752221e-05}
{"train_loss": 0.9787148833274841, "global_step": 4564, "epoch": 51, "lr": 9.994447326420977e-05}
{"train_loss": 0.9943148493766785, "global_step": 4565, "epoch": 51, "lr": 9.9944445946483e-05}
{"train_loss": 0.8282878398895264, "global_step": 4566, "epoch": 51, "lr": 9.994441862204182e-05}
{"train_loss": 1.0881181955337524, "global_step": 4567, "epoch": 51, "lr": 9.994439129088623e-05}
{"train_loss": 1.1165457963943481, "global_step": 4568, "epoch": 51, "lr": 9.994436395301621e-05}
{"train_loss": 0.9533129930496216, "global_step": 4569, "epoch": 51, "lr": 9.994433660843177e-05}
{"train_loss": 1.185178518295288, "global_step": 4570, "epoch": 51, "lr": 9.994430925713294e-05}
{"train_loss": 1.0459171533584595, "global_step": 4571, "epoch": 51, "lr": 9.994428189911969e-05}
{"train_loss": 0.9463125467300415, "global_step": 4572, "epoch": 51, "lr": 9.994425453439206e-05}
{"train_loss": 0.9660978317260742, "global_step": 4573, "epoch": 51, "lr": 9.994422716295002e-05}
{"train_loss": 0.9882557988166809, "global_step": 4574, "epoch": 51, "lr": 9.994419978479358e-05}
{"train_loss": 1.0734446048736572, "global_step": 4575, "epoch": 51, "lr": 9.994417239992278e-05}
{"train_loss": 0.8115659952163696, "global_step": 4576, "epoch": 51, "lr": 9.994414500833758e-05}
{"train_loss": 1.1290282011032104, "global_step": 4577, "epoch": 51, "lr": 9.994411761003799e-05}
{"train_loss": 0.9598034620285034, "global_step": 4578, "epoch": 51, "lr": 9.994409020502402e-05}
{"train_loss": 0.8817738890647888, "global_step": 4579, "epoch": 51, "lr": 9.994406279329569e-05}
{"train_loss": 1.0051259994506836, "global_step": 4580, "epoch": 51, "lr": 9.994403537485297e-05}
{"train_loss": 0.8720003366470337, "global_step": 4581, "epoch": 51, "lr": 9.99440079496959e-05}
{"train_loss": 1.013114333152771, "global_step": 4582, "epoch": 51, "lr": 9.994398051782446e-05}
{"train_loss": 0.9866513013839722, "global_step": 4583, "epoch": 51, "lr": 9.994395307923867e-05}
{"train_loss": 1.119818091392517, "global_step": 4584, "epoch": 51, "lr": 9.994392563393852e-05}
{"train_loss": 1.0254952907562256, "global_step": 4585, "epoch": 51, "lr": 9.9943898181924e-05}
{"train_loss": 0.7772836685180664, "global_step": 4586, "epoch": 51, "lr": 9.994387072319515e-05}
{"train_loss": 1.012593150138855, "global_step": 4587, "epoch": 51, "lr": 9.994384325775197e-05}
{"train_loss": 1.0026006698608398, "global_step": 4588, "epoch": 51, "lr": 9.994381578559442e-05}
{"train_loss": 0.8991330862045288, "global_step": 4589, "epoch": 51, "lr": 9.994378830672255e-05}
{"train_loss": 0.9706067442893982, "global_step": 4590, "epoch": 51, "lr": 9.994376082113635e-05}
{"train_loss": 1.224913239479065, "global_step": 4591, "epoch": 51, "lr": 9.99437333288358e-05}
{"train_loss": 1.1498041152954102, "global_step": 4592, "epoch": 51, "lr": 9.994370582982093e-05}
{"train_loss": 1.0786845684051514, "global_step": 4593, "epoch": 51, "lr": 9.994367832409175e-05}
{"train_loss": 0.8398547768592834, "global_step": 4594, "epoch": 51, "lr": 9.994365081164824e-05}
{"train_loss": 0.9092586636543274, "global_step": 4595, "epoch": 51, "lr": 9.994362329249042e-05}
{"train_loss": 0.8742000460624695, "global_step": 4596, "epoch": 51, "lr": 9.994359576661828e-05}
{"train_loss": 0.9919139742851257, "global_step": 4597, "epoch": 51, "lr": 9.994356823403184e-05}
{"train_loss": 0.8510578870773315, "global_step": 4598, "epoch": 51, "lr": 9.99435406947311e-05}
{"train_loss": 0.9806202054023743, "global_step": 4599, "epoch": 51, "lr": 9.994351314871605e-05}
{"train_loss": 0.964091956615448, "global_step": 4600, "epoch": 51, "lr": 9.994348559598671e-05}
{"train_loss": 0.9191340208053589, "global_step": 4601, "epoch": 51, "lr": 9.994345803654308e-05}
{"train_loss": 1.049333095550537, "global_step": 4602, "epoch": 51, "lr": 9.994343047038514e-05}
{"train_loss": 0.9123030304908752, "global_step": 4603, "epoch": 51, "lr": 9.994340289751293e-05}
{"train_loss": 1.0468755960464478, "global_step": 4604, "epoch": 51, "lr": 9.994337531792645e-05}
{"train_loss": 0.950236976146698, "global_step": 4605, "epoch": 51, "lr": 9.994334773162568e-05}
{"train_loss": 0.9347396492958069, "global_step": 4606, "epoch": 51, "lr": 9.994332013861062e-05}
{"train_loss": 0.9777485728263855, "global_step": 4607, "epoch": 51, "lr": 9.994329253888132e-05}
{"train_loss": 0.977476954460144, "global_step": 4608, "epoch": 51, "lr": 9.994326493243772e-05}
{"train_loss": 1.1410088539123535, "global_step": 4609, "epoch": 51, "lr": 9.99432373192799e-05}
{"train_loss": 1.0523416996002197, "global_step": 4610, "epoch": 51, "lr": 9.994320969940778e-05}
{"train_loss": 1.1257437467575073, "global_step": 4611, "epoch": 51, "lr": 9.994318207282143e-05}
{"train_loss": 1.2414021492004395, "global_step": 4612, "epoch": 51, "lr": 9.994315443952081e-05}
{"train_loss": 0.847439706325531, "global_step": 4613, "epoch": 51, "lr": 9.994312679950594e-05}
{"train_loss": 1.0651426315307617, "global_step": 4614, "epoch": 51, "lr": 9.994309915277684e-05}
{"train_loss": 1.3301278352737427, "global_step": 4615, "epoch": 51, "lr": 9.99430714993335e-05}
{"train_loss": 1.1577270030975342, "global_step": 4616, "epoch": 51, "lr": 9.994304383917592e-05}
{"train_loss": 1.0912679433822632, "global_step": 4617, "epoch": 51, "lr": 9.99430161723041e-05}
{"train_loss": 0.7128226161003113, "global_step": 4618, "epoch": 51, "lr": 9.994298849871806e-05}
{"train_loss": 0.8597525358200073, "global_step": 4619, "epoch": 51, "lr": 9.994296081841779e-05}
{"train_loss": 1.1359769105911255, "global_step": 4620, "epoch": 51, "lr": 9.994293313140328e-05}
{"train_loss": 1.3438748121261597, "global_step": 4621, "epoch": 51, "lr": 9.994290543767457e-05}
{"train_loss": 1.1554019451141357, "global_step": 4622, "epoch": 51, "lr": 9.994287773723164e-05}
{"train_loss": 0.9586613178253174, "global_step": 4623, "epoch": 51, "lr": 9.994285003007451e-05}
{"train_loss": 1.030733346939087, "global_step": 4624, "epoch": 51, "lr": 9.994282231620315e-05}
{"train_loss": 1.140377402305603, "global_step": 4625, "epoch": 51, "lr": 9.99427945956176e-05}
{"train_loss": 1.0907050371170044, "global_step": 4626, "epoch": 51, "lr": 9.994276686831784e-05}
{"train_loss": 1.0040145556578475, "global_step": 4627, "epoch": 51, "lr": 9.994273913430389e-05, "val_loss": 1.2433416843414307}
{"train_loss": 1.1753852367401123, "global_step": 4628, "epoch": 52, "lr": 9.994271139357576e-05}
{"train_loss": 0.9113931655883789, "global_step": 4629, "epoch": 52, "lr": 9.994268364613344e-05}
{"train_loss": 0.8803572654724121, "global_step": 4630, "epoch": 52, "lr": 9.994265589197692e-05}
{"train_loss": 1.2033498287200928, "global_step": 4631, "epoch": 52, "lr": 9.994262813110622e-05}
{"train_loss": 0.9348551034927368, "global_step": 4632, "epoch": 52, "lr": 9.994260036352135e-05}
{"train_loss": 1.1262942552566528, "global_step": 4633, "epoch": 52, "lr": 9.99425725892223e-05}
{"train_loss": 1.1426723003387451, "global_step": 4634, "epoch": 52, "lr": 9.994254480820908e-05}
{"train_loss": 1.0267302989959717, "global_step": 4635, "epoch": 52, "lr": 9.99425170204817e-05}
{"train_loss": 0.8963574171066284, "global_step": 4636, "epoch": 52, "lr": 9.994248922604016e-05}
{"train_loss": 1.053619146347046, "global_step": 4637, "epoch": 52, "lr": 9.994246142488445e-05}
{"train_loss": 0.8797731995582581, "global_step": 4638, "epoch": 52, "lr": 9.994243361701458e-05}
{"train_loss": 1.2589521408081055, "global_step": 4639, "epoch": 52, "lr": 9.994240580243057e-05}
{"train_loss": 0.9473211765289307, "global_step": 4640, "epoch": 52, "lr": 9.994237798113242e-05}
{"train_loss": 1.2819405794143677, "global_step": 4641, "epoch": 52, "lr": 9.994235015312011e-05}
{"train_loss": 1.0515778064727783, "global_step": 4642, "epoch": 52, "lr": 9.994232231839367e-05}
{"train_loss": 0.993828535079956, "global_step": 4643, "epoch": 52, "lr": 9.994229447695309e-05}
{"train_loss": 1.0633206367492676, "global_step": 4644, "epoch": 52, "lr": 9.994226662879837e-05}
{"train_loss": 0.939592719078064, "global_step": 4645, "epoch": 52, "lr": 9.994223877392954e-05}
{"train_loss": 1.0687885284423828, "global_step": 4646, "epoch": 52, "lr": 9.994221091234657e-05}
{"train_loss": 1.0786287784576416, "global_step": 4647, "epoch": 52, "lr": 9.994218304404948e-05}
{"train_loss": 0.9756709337234497, "global_step": 4648, "epoch": 52, "lr": 9.994215516903826e-05}
{"train_loss": 1.145579218864441, "global_step": 4649, "epoch": 52, "lr": 9.994212728731295e-05}
{"train_loss": 0.9923514723777771, "global_step": 4650, "epoch": 52, "lr": 9.994209939887352e-05}
{"train_loss": 0.9513092041015625, "global_step": 4651, "epoch": 52, "lr": 9.994207150371998e-05}
{"train_loss": 0.9691051840782166, "global_step": 4652, "epoch": 52, "lr": 9.994204360185233e-05}
{"train_loss": 1.0570902824401855, "global_step": 4653, "epoch": 52, "lr": 9.99420156932706e-05}
{"train_loss": 1.1755422353744507, "global_step": 4654, "epoch": 52, "lr": 9.994198777797477e-05}
{"train_loss": 1.269768476486206, "global_step": 4655, "epoch": 52, "lr": 9.994195985596484e-05}
{"train_loss": 0.9226577877998352, "global_step": 4656, "epoch": 52, "lr": 9.994193192724082e-05}
{"train_loss": 1.0704638957977295, "global_step": 4657, "epoch": 52, "lr": 9.994190399180272e-05}
{"train_loss": 1.000097393989563, "global_step": 4658, "epoch": 52, "lr": 9.994187604965056e-05}
{"train_loss": 0.9801414608955383, "global_step": 4659, "epoch": 52, "lr": 9.994184810078429e-05}
{"train_loss": 0.9825245141983032, "global_step": 4660, "epoch": 52, "lr": 9.994182014520397e-05}
{"train_loss": 0.8522620797157288, "global_step": 4661, "epoch": 52, "lr": 9.994179218290956e-05}
{"train_loss": 1.1885687112808228, "global_step": 4662, "epoch": 52, "lr": 9.99417642139011e-05}
{"train_loss": 1.045944333076477, "global_step": 4663, "epoch": 52, "lr": 9.994173623817857e-05}
{"train_loss": 0.9632459282875061, "global_step": 4664, "epoch": 52, "lr": 9.994170825574199e-05}
{"train_loss": 0.9963143467903137, "global_step": 4665, "epoch": 52, "lr": 9.994168026659136e-05}
{"train_loss": 0.8236185312271118, "global_step": 4666, "epoch": 52, "lr": 9.994165227072667e-05}
{"train_loss": 0.9713114500045776, "global_step": 4667, "epoch": 52, "lr": 9.994162426814796e-05}
{"train_loss": 1.1312263011932373, "global_step": 4668, "epoch": 52, "lr": 9.994159625885517e-05}
{"train_loss": 0.9927793741226196, "global_step": 4669, "epoch": 52, "lr": 9.994156824284835e-05}
{"train_loss": 0.9926722049713135, "global_step": 4670, "epoch": 52, "lr": 9.99415402201275e-05}
{"train_loss": 1.2090131044387817, "global_step": 4671, "epoch": 52, "lr": 9.994151219069262e-05}
{"train_loss": 1.063795566558838, "global_step": 4672, "epoch": 52, "lr": 9.994148415454372e-05}
{"train_loss": 1.1302871704101562, "global_step": 4673, "epoch": 52, "lr": 9.994145611168077e-05}
{"train_loss": 0.8593659996986389, "global_step": 4674, "epoch": 52, "lr": 9.994142806210383e-05}
{"train_loss": 0.9585064649581909, "global_step": 4675, "epoch": 52, "lr": 9.994140000581287e-05}
{"train_loss": 0.9615702033042908, "global_step": 4676, "epoch": 52, "lr": 9.994137194280788e-05}
{"train_loss": 0.993573784828186, "global_step": 4677, "epoch": 52, "lr": 9.994134387308888e-05}
{"train_loss": 0.9989722371101379, "global_step": 4678, "epoch": 52, "lr": 9.99413157966559e-05}
{"train_loss": 0.9896681308746338, "global_step": 4679, "epoch": 52, "lr": 9.99412877135089e-05}
{"train_loss": 0.9677727222442627, "global_step": 4680, "epoch": 52, "lr": 9.994125962364791e-05}
{"train_loss": 0.848196268081665, "global_step": 4681, "epoch": 52, "lr": 9.994123152707292e-05}
{"train_loss": 0.8079898357391357, "global_step": 4682, "epoch": 52, "lr": 9.994120342378395e-05}
{"train_loss": 0.9803412556648254, "global_step": 4683, "epoch": 52, "lr": 9.994117531378097e-05}
{"train_loss": 1.0695068836212158, "global_step": 4684, "epoch": 52, "lr": 9.994114719706405e-05}
{"train_loss": 1.1275367736816406, "global_step": 4685, "epoch": 52, "lr": 9.994111907363311e-05}
{"train_loss": 0.9435411095619202, "global_step": 4686, "epoch": 52, "lr": 9.994109094348821e-05}
{"train_loss": 1.0370523929595947, "global_step": 4687, "epoch": 52, "lr": 9.994106280662935e-05}
{"train_loss": 1.1038734912872314, "global_step": 4688, "epoch": 52, "lr": 9.994103466305651e-05}
{"train_loss": 0.8801229000091553, "global_step": 4689, "epoch": 52, "lr": 9.99410065127697e-05}
{"train_loss": 1.0723196268081665, "global_step": 4690, "epoch": 52, "lr": 9.994097835576894e-05}
{"train_loss": 1.2273259162902832, "global_step": 4691, "epoch": 52, "lr": 9.994095019205423e-05}
{"train_loss": 0.9241377115249634, "global_step": 4692, "epoch": 52, "lr": 9.994092202162558e-05}
{"train_loss": 0.7701201438903809, "global_step": 4693, "epoch": 52, "lr": 9.994089384448295e-05}
{"train_loss": 0.9013446569442749, "global_step": 4694, "epoch": 52, "lr": 9.994086566062639e-05}
{"train_loss": 1.281885027885437, "global_step": 4695, "epoch": 52, "lr": 9.994083747005589e-05}
{"train_loss": 0.9172517657279968, "global_step": 4696, "epoch": 52, "lr": 9.994080927277146e-05}
{"train_loss": 0.9160914421081543, "global_step": 4697, "epoch": 52, "lr": 9.994078106877308e-05}
{"train_loss": 0.8863879442214966, "global_step": 4698, "epoch": 52, "lr": 9.994075285806079e-05}
{"train_loss": 1.2496334314346313, "global_step": 4699, "epoch": 52, "lr": 9.994072464063455e-05}
{"train_loss": 0.919349730014801, "global_step": 4700, "epoch": 52, "lr": 9.99406964164944e-05}
{"train_loss": 1.2264924049377441, "global_step": 4701, "epoch": 52, "lr": 9.994066818564034e-05}
{"train_loss": 0.9359703660011292, "global_step": 4702, "epoch": 52, "lr": 9.994063994807237e-05}
{"train_loss": 1.1962300539016724, "global_step": 4703, "epoch": 52, "lr": 9.994061170379048e-05}
{"train_loss": 0.9868506193161011, "global_step": 4704, "epoch": 52, "lr": 9.994058345279467e-05}
{"train_loss": 0.863469123840332, "global_step": 4705, "epoch": 52, "lr": 9.994055519508497e-05}
{"train_loss": 1.0471292734146118, "global_step": 4706, "epoch": 52, "lr": 9.994052693066139e-05}
{"train_loss": 1.2112475633621216, "global_step": 4707, "epoch": 52, "lr": 9.994049865952388e-05}
{"train_loss": 1.1956875324249268, "global_step": 4708, "epoch": 52, "lr": 9.994047038167249e-05}
{"train_loss": 1.0237693786621094, "global_step": 4709, "epoch": 52, "lr": 9.994044209710722e-05}
{"train_loss": 1.013952374458313, "global_step": 4710, "epoch": 52, "lr": 9.994041380582807e-05}
{"train_loss": 1.1257649660110474, "global_step": 4711, "epoch": 52, "lr": 9.994038550783503e-05}
{"train_loss": 1.036384105682373, "global_step": 4712, "epoch": 52, "lr": 9.994035720312811e-05}
{"train_loss": 0.9401353597640991, "global_step": 4713, "epoch": 52, "lr": 9.994032889170733e-05}
{"train_loss": 1.002579927444458, "global_step": 4714, "epoch": 52, "lr": 9.994030057357269e-05}
{"train_loss": 1.0364570617675781, "global_step": 4715, "epoch": 52, "lr": 9.994027224872417e-05}
{"train_loss": 1.0258716585930814, "global_step": 4716, "epoch": 52, "lr": 9.994024391716179e-05, "val_loss": 1.193479299545288}
{"train_loss": 1.0979779958724976, "global_step": 4717, "epoch": 53, "lr": 9.994021557888555e-05}
{"train_loss": 0.9279491901397705, "global_step": 4718, "epoch": 53, "lr": 9.994018723389547e-05}
{"train_loss": 0.9969986081123352, "global_step": 4719, "epoch": 53, "lr": 9.994015888219152e-05}
{"train_loss": 0.7504922747612, "global_step": 4720, "epoch": 53, "lr": 9.994013052377374e-05}
{"train_loss": 1.0289349555969238, "global_step": 4721, "epoch": 53, "lr": 9.994010215864211e-05}
{"train_loss": 0.9720090627670288, "global_step": 4722, "epoch": 53, "lr": 9.994007378679665e-05}
{"train_loss": 1.179732322692871, "global_step": 4723, "epoch": 53, "lr": 9.994004540823735e-05}
{"train_loss": 0.9845531582832336, "global_step": 4724, "epoch": 53, "lr": 9.994001702296422e-05}
{"train_loss": 1.0174471139907837, "global_step": 4725, "epoch": 53, "lr": 9.993998863097726e-05}
{"train_loss": 0.8887773156166077, "global_step": 4726, "epoch": 53, "lr": 9.993996023227648e-05}
{"train_loss": 1.0041800737380981, "global_step": 4727, "epoch": 53, "lr": 9.99399318268619e-05}
{"train_loss": 0.9194318652153015, "global_step": 4728, "epoch": 53, "lr": 9.993990341473348e-05}
{"train_loss": 1.0970604419708252, "global_step": 4729, "epoch": 53, "lr": 9.993987499589126e-05}
{"train_loss": 1.0680441856384277, "global_step": 4730, "epoch": 53, "lr": 9.993984657033522e-05}
{"train_loss": 1.0717869997024536, "global_step": 4731, "epoch": 53, "lr": 9.993981813806539e-05}
{"train_loss": 0.9868263602256775, "global_step": 4732, "epoch": 53, "lr": 9.993978969908175e-05}
{"train_loss": 1.0858774185180664, "global_step": 4733, "epoch": 53, "lr": 9.993976125338432e-05}
{"train_loss": 0.7764616012573242, "global_step": 4734, "epoch": 53, "lr": 9.993973280097311e-05}
{"train_loss": 0.9335663318634033, "global_step": 4735, "epoch": 53, "lr": 9.993970434184809e-05}
{"train_loss": 1.0018004179000854, "global_step": 4736, "epoch": 53, "lr": 9.99396758760093e-05}
{"train_loss": 1.0436272621154785, "global_step": 4737, "epoch": 53, "lr": 9.993964740345671e-05}
{"train_loss": 1.3025645017623901, "global_step": 4738, "epoch": 53, "lr": 9.993961892419036e-05}
{"train_loss": 0.8821789026260376, "global_step": 4739, "epoch": 53, "lr": 9.993959043821024e-05}
{"train_loss": 1.0137572288513184, "global_step": 4740, "epoch": 53, "lr": 9.993956194551634e-05}
{"train_loss": 0.9910047650337219, "global_step": 4741, "epoch": 53, "lr": 9.993953344610868e-05}
{"train_loss": 0.9067850112915039, "global_step": 4742, "epoch": 53, "lr": 9.993950493998725e-05}
{"train_loss": 0.7963988184928894, "global_step": 4743, "epoch": 53, "lr": 9.993947642715206e-05}
{"train_loss": 0.9101585149765015, "global_step": 4744, "epoch": 53, "lr": 9.993944790760314e-05}
{"train_loss": 0.9613872766494751, "global_step": 4745, "epoch": 53, "lr": 9.993941938134047e-05}
{"train_loss": 1.0078201293945312, "global_step": 4746, "epoch": 53, "lr": 9.993939084836403e-05}
{"train_loss": 1.1078639030456543, "global_step": 4747, "epoch": 53, "lr": 9.993936230867385e-05}
{"train_loss": 0.9966875314712524, "global_step": 4748, "epoch": 53, "lr": 9.993933376226994e-05}
{"train_loss": 1.047080636024475, "global_step": 4749, "epoch": 53, "lr": 9.993930520915228e-05}
{"train_loss": 1.1110706329345703, "global_step": 4750, "epoch": 53, "lr": 9.993927664932092e-05}
{"train_loss": 1.0877140760421753, "global_step": 4751, "epoch": 53, "lr": 9.993924808277582e-05}
{"train_loss": 0.9848735332489014, "global_step": 4752, "epoch": 53, "lr": 9.993921950951697e-05}
{"train_loss": 0.9186003804206848, "global_step": 4753, "epoch": 53, "lr": 9.993919092954444e-05}
{"train_loss": 0.7982086539268494, "global_step": 4754, "epoch": 53, "lr": 9.993916234285816e-05}
{"train_loss": 0.9599981904029846, "global_step": 4755, "epoch": 53, "lr": 9.99391337494582e-05}
{"train_loss": 0.9711405634880066, "global_step": 4756, "epoch": 53, "lr": 9.993910514934452e-05}
{"train_loss": 1.0169174671173096, "global_step": 4757, "epoch": 53, "lr": 9.993907654251712e-05}
{"train_loss": 1.0417215824127197, "global_step": 4758, "epoch": 53, "lr": 9.993904792897603e-05}
{"train_loss": 1.0264588594436646, "global_step": 4759, "epoch": 53, "lr": 9.993901930872125e-05}
{"train_loss": 0.890998363494873, "global_step": 4760, "epoch": 53, "lr": 9.993899068175278e-05}
{"train_loss": 1.071370005607605, "global_step": 4761, "epoch": 53, "lr": 9.993896204807062e-05}
{"train_loss": 1.0600671768188477, "global_step": 4762, "epoch": 53, "lr": 9.993893340767478e-05}
{"train_loss": 0.9262747168540955, "global_step": 4763, "epoch": 53, "lr": 9.993890476056525e-05}
{"train_loss": 1.0289369821548462, "global_step": 4764, "epoch": 53, "lr": 9.993887610674204e-05}
{"train_loss": 1.0141799449920654, "global_step": 4765, "epoch": 53, "lr": 9.993884744620515e-05}
{"train_loss": 1.0765069723129272, "global_step": 4766, "epoch": 53, "lr": 9.993881877895461e-05}
{"train_loss": 0.9734048247337341, "global_step": 4767, "epoch": 53, "lr": 9.993879010499039e-05}
{"train_loss": 1.1505603790283203, "global_step": 4768, "epoch": 53, "lr": 9.993876142431251e-05}
{"train_loss": 0.9727284908294678, "global_step": 4769, "epoch": 53, "lr": 9.993873273692097e-05}
{"train_loss": 1.0453543663024902, "global_step": 4770, "epoch": 53, "lr": 9.993870404281578e-05}
{"train_loss": 0.8728432059288025, "global_step": 4771, "epoch": 53, "lr": 9.993867534199694e-05}
{"train_loss": 0.916248083114624, "global_step": 4772, "epoch": 53, "lr": 9.993864663446446e-05}
{"train_loss": 0.8595095276832581, "global_step": 4773, "epoch": 53, "lr": 9.993861792021833e-05}
{"train_loss": 1.05483078956604, "global_step": 4774, "epoch": 53, "lr": 9.993858919925857e-05}
{"train_loss": 1.0538548231124878, "global_step": 4775, "epoch": 53, "lr": 9.993856047158516e-05}
{"train_loss": 0.9016420841217041, "global_step": 4776, "epoch": 53, "lr": 9.993853173719814e-05}
{"train_loss": 0.9843533039093018, "global_step": 4777, "epoch": 53, "lr": 9.993850299609748e-05}
{"train_loss": 0.999157726764679, "global_step": 4778, "epoch": 53, "lr": 9.99384742482832e-05}
{"train_loss": 0.9966967701911926, "global_step": 4779, "epoch": 53, "lr": 9.993844549375531e-05}
{"train_loss": 0.8936054110527039, "global_step": 4780, "epoch": 53, "lr": 9.993841673251379e-05}
{"train_loss": 1.0640475749969482, "global_step": 4781, "epoch": 53, "lr": 9.993838796455865e-05}
{"train_loss": 1.012874722480774, "global_step": 4782, "epoch": 53, "lr": 9.993835918988991e-05}
{"train_loss": 1.0250427722930908, "global_step": 4783, "epoch": 53, "lr": 9.993833040850759e-05}
{"train_loss": 0.8992727994918823, "global_step": 4784, "epoch": 53, "lr": 9.993830162041165e-05}
{"train_loss": 0.8929680585861206, "global_step": 4785, "epoch": 53, "lr": 9.993827282560211e-05}
{"train_loss": 0.9134944081306458, "global_step": 4786, "epoch": 53, "lr": 9.993824402407898e-05}
{"train_loss": 1.1233479976654053, "global_step": 4787, "epoch": 53, "lr": 9.993821521584226e-05}
{"train_loss": 0.9710140824317932, "global_step": 4788, "epoch": 53, "lr": 9.993818640089196e-05}
{"train_loss": 1.0098910331726074, "global_step": 4789, "epoch": 53, "lr": 9.993815757922809e-05}
{"train_loss": 0.8116011619567871, "global_step": 4790, "epoch": 53, "lr": 9.993812875085063e-05}
{"train_loss": 1.0924757719039917, "global_step": 4791, "epoch": 53, "lr": 9.99380999157596e-05}
{"train_loss": 1.0842313766479492, "global_step": 4792, "epoch": 53, "lr": 9.993807107395501e-05}
{"train_loss": 0.8244737386703491, "global_step": 4793, "epoch": 53, "lr": 9.993804222543684e-05}
{"train_loss": 1.1232593059539795, "global_step": 4794, "epoch": 53, "lr": 9.993801337020512e-05}
{"train_loss": 0.9474629759788513, "global_step": 4795, "epoch": 53, "lr": 9.993798450825984e-05}
{"train_loss": 1.074575662612915, "global_step": 4796, "epoch": 53, "lr": 9.9937955639601e-05}
{"train_loss": 1.1116243600845337, "global_step": 4797, "epoch": 53, "lr": 9.993792676422862e-05}
{"train_loss": 1.176248550415039, "global_step": 4798, "epoch": 53, "lr": 9.993789788214269e-05}
{"train_loss": 0.9158663153648376, "global_step": 4799, "epoch": 53, "lr": 9.993786899334321e-05}
{"train_loss": 0.9529393911361694, "global_step": 4800, "epoch": 53, "lr": 9.99378400978302e-05}
{"train_loss": 1.17108952999115, "global_step": 4801, "epoch": 53, "lr": 9.993781119560367e-05}
{"train_loss": 0.9266051054000854, "global_step": 4802, "epoch": 53, "lr": 9.993778228666359e-05}
{"train_loss": 1.0074787139892578, "global_step": 4803, "epoch": 53, "lr": 9.993775337100999e-05}
{"train_loss": 0.9447088241577148, "global_step": 4804, "epoch": 53, "lr": 9.993772444864287e-05}
{"train_loss": 0.9950231199853876, "global_step": 4805, "epoch": 53, "lr": 9.993769551956223e-05, "val_loss": 1.1318994760513306}
{"train_loss": 1.0169726610183716, "global_step": 4806, "epoch": 54, "lr": 9.993766658376808e-05}
{"train_loss": 1.0313990116119385, "global_step": 4807, "epoch": 54, "lr": 9.993763764126042e-05}
{"train_loss": 1.0592474937438965, "global_step": 4808, "epoch": 54, "lr": 9.993760869203924e-05}
{"train_loss": 1.0174293518066406, "global_step": 4809, "epoch": 54, "lr": 9.993757973610457e-05}
{"train_loss": 0.996321976184845, "global_step": 4810, "epoch": 54, "lr": 9.993755077345639e-05}
{"train_loss": 0.9288382530212402, "global_step": 4811, "epoch": 54, "lr": 9.993752180409473e-05}
{"train_loss": 0.9321020245552063, "global_step": 4812, "epoch": 54, "lr": 9.993749282801957e-05}
{"train_loss": 0.9690337777137756, "global_step": 4813, "epoch": 54, "lr": 9.993746384523093e-05}
{"train_loss": 0.8903428912162781, "global_step": 4814, "epoch": 54, "lr": 9.99374348557288e-05}
{"train_loss": 0.9913225769996643, "global_step": 4815, "epoch": 54, "lr": 9.993740585951318e-05}
{"train_loss": 0.9295238256454468, "global_step": 4816, "epoch": 54, "lr": 9.993737685658408e-05}
{"train_loss": 0.7576007843017578, "global_step": 4817, "epoch": 54, "lr": 9.993734784694153e-05}
{"train_loss": 0.8788520097732544, "global_step": 4818, "epoch": 54, "lr": 9.993731883058551e-05}
{"train_loss": 1.0093092918395996, "global_step": 4819, "epoch": 54, "lr": 9.993728980751603e-05}
{"train_loss": 0.8372710943222046, "global_step": 4820, "epoch": 54, "lr": 9.993726077773308e-05}
{"train_loss": 0.9044906497001648, "global_step": 4821, "epoch": 54, "lr": 9.993723174123668e-05}
{"train_loss": 0.9913746118545532, "global_step": 4822, "epoch": 54, "lr": 9.993720269802681e-05}
{"train_loss": 1.0555930137634277, "global_step": 4823, "epoch": 54, "lr": 9.993717364810351e-05}
{"train_loss": 1.006473183631897, "global_step": 4824, "epoch": 54, "lr": 9.993714459146676e-05}
{"train_loss": 1.0233300924301147, "global_step": 4825, "epoch": 54, "lr": 9.993711552811658e-05}
{"train_loss": 1.1784350872039795, "global_step": 4826, "epoch": 54, "lr": 9.993708645805295e-05}
{"train_loss": 0.9113209247589111, "global_step": 4827, "epoch": 54, "lr": 9.993705738127588e-05}
{"train_loss": 0.934407651424408, "global_step": 4828, "epoch": 54, "lr": 9.99370282977854e-05}
{"train_loss": 0.8903661966323853, "global_step": 4829, "epoch": 54, "lr": 9.993699920758149e-05}
{"train_loss": 1.0026088953018188, "global_step": 4830, "epoch": 54, "lr": 9.993697011066414e-05}
{"train_loss": 0.8698214292526245, "global_step": 4831, "epoch": 54, "lr": 9.993694100703339e-05}
{"train_loss": 1.1229217052459717, "global_step": 4832, "epoch": 54, "lr": 9.993691189668924e-05}
{"train_loss": 1.101128339767456, "global_step": 4833, "epoch": 54, "lr": 9.993688277963168e-05}
{"train_loss": 1.160996913909912, "global_step": 4834, "epoch": 54, "lr": 9.993685365586068e-05}
{"train_loss": 0.8635917901992798, "global_step": 4835, "epoch": 54, "lr": 9.99368245253763e-05}
{"train_loss": 1.03865647315979, "global_step": 4836, "epoch": 54, "lr": 9.993679538817852e-05}
{"train_loss": 1.1513879299163818, "global_step": 4837, "epoch": 54, "lr": 9.993676624426735e-05}
{"train_loss": 0.9860291481018066, "global_step": 4838, "epoch": 54, "lr": 9.99367370936428e-05}
{"train_loss": 0.9430426359176636, "global_step": 4839, "epoch": 54, "lr": 9.993670793630485e-05}
{"train_loss": 0.882803201675415, "global_step": 4840, "epoch": 54, "lr": 9.993667877225352e-05}
{"train_loss": 0.9954432845115662, "global_step": 4841, "epoch": 54, "lr": 9.993664960148882e-05}
{"train_loss": 0.935786247253418, "global_step": 4842, "epoch": 54, "lr": 9.993662042401074e-05}
{"train_loss": 0.9088004231452942, "global_step": 4843, "epoch": 54, "lr": 9.993659123981928e-05}
{"train_loss": 1.0524532794952393, "global_step": 4844, "epoch": 54, "lr": 9.993656204891449e-05}
{"train_loss": 0.9540790915489197, "global_step": 4845, "epoch": 54, "lr": 9.993653285129629e-05}
{"train_loss": 0.9366790652275085, "global_step": 4846, "epoch": 54, "lr": 9.993650364696476e-05}
{"train_loss": 0.8894521594047546, "global_step": 4847, "epoch": 54, "lr": 9.993647443591987e-05}
{"train_loss": 0.9167688488960266, "global_step": 4848, "epoch": 54, "lr": 9.993644521816164e-05}
{"train_loss": 0.9285675883293152, "global_step": 4849, "epoch": 54, "lr": 9.993641599369004e-05}
{"train_loss": 0.9431813955307007, "global_step": 4850, "epoch": 54, "lr": 9.993638676250512e-05}
{"train_loss": 0.9845289587974548, "global_step": 4851, "epoch": 54, "lr": 9.993635752460686e-05}
{"train_loss": 0.8556283116340637, "global_step": 4852, "epoch": 54, "lr": 9.993632827999526e-05}
{"train_loss": 0.9986846446990967, "global_step": 4853, "epoch": 54, "lr": 9.993629902867031e-05}
{"train_loss": 1.1192301511764526, "global_step": 4854, "epoch": 54, "lr": 9.993626977063206e-05}
{"train_loss": 0.9597699046134949, "global_step": 4855, "epoch": 54, "lr": 9.993624050588047e-05}
{"train_loss": 0.9956777095794678, "global_step": 4856, "epoch": 54, "lr": 9.993621123441557e-05}
{"train_loss": 0.8786413669586182, "global_step": 4857, "epoch": 54, "lr": 9.993618195623735e-05}
{"train_loss": 1.1128458976745605, "global_step": 4858, "epoch": 54, "lr": 9.993615267134583e-05}
{"train_loss": 1.0554529428482056, "global_step": 4859, "epoch": 54, "lr": 9.9936123379741e-05}
{"train_loss": 1.013417363166809, "global_step": 4860, "epoch": 54, "lr": 9.993609408142286e-05}
{"train_loss": 0.9169602990150452, "global_step": 4861, "epoch": 54, "lr": 9.993606477639143e-05}
{"train_loss": 0.8243192434310913, "global_step": 4862, "epoch": 54, "lr": 9.99360354646467e-05}
{"train_loss": 1.0657655000686646, "global_step": 4863, "epoch": 54, "lr": 9.993600614618866e-05}
{"train_loss": 1.0067532062530518, "global_step": 4864, "epoch": 54, "lr": 9.993597682101737e-05}
{"train_loss": 1.2154432535171509, "global_step": 4865, "epoch": 54, "lr": 9.993594748913276e-05}
{"train_loss": 1.0817160606384277, "global_step": 4866, "epoch": 54, "lr": 9.993591815053489e-05}
{"train_loss": 1.0203279256820679, "global_step": 4867, "epoch": 54, "lr": 9.993588880522374e-05}
{"train_loss": 0.963081419467926, "global_step": 4868, "epoch": 54, "lr": 9.993585945319932e-05}
{"train_loss": 0.9557430148124695, "global_step": 4869, "epoch": 54, "lr": 9.993583009446163e-05}
{"train_loss": 0.7748192548751831, "global_step": 4870, "epoch": 54, "lr": 9.993580072901067e-05}
{"train_loss": 1.0818243026733398, "global_step": 4871, "epoch": 54, "lr": 9.993577135684646e-05}
{"train_loss": 1.2159322500228882, "global_step": 4872, "epoch": 54, "lr": 9.9935741977969e-05}
{"train_loss": 1.0490137338638306, "global_step": 4873, "epoch": 54, "lr": 9.993571259237828e-05}
{"train_loss": 1.0361188650131226, "global_step": 4874, "epoch": 54, "lr": 9.993568320007431e-05}
{"train_loss": 1.069969654083252, "global_step": 4875, "epoch": 54, "lr": 9.99356538010571e-05}
{"train_loss": 1.2184377908706665, "global_step": 4876, "epoch": 54, "lr": 9.993562439532665e-05}
{"train_loss": 0.874496340751648, "global_step": 4877, "epoch": 54, "lr": 9.993559498288296e-05}
{"train_loss": 1.0315124988555908, "global_step": 4878, "epoch": 54, "lr": 9.993556556372604e-05}
{"train_loss": 1.0253883600234985, "global_step": 4879, "epoch": 54, "lr": 9.993553613785588e-05}
{"train_loss": 1.0954456329345703, "global_step": 4880, "epoch": 54, "lr": 9.993550670527252e-05}
{"train_loss": 0.8897145986557007, "global_step": 4881, "epoch": 54, "lr": 9.993547726597592e-05}
{"train_loss": 0.7408289909362793, "global_step": 4882, "epoch": 54, "lr": 9.993544781996613e-05}
{"train_loss": 0.9470099210739136, "global_step": 4883, "epoch": 54, "lr": 9.99354183672431e-05}
{"train_loss": 1.1460561752319336, "global_step": 4884, "epoch": 54, "lr": 9.993538890780688e-05}
{"train_loss": 1.101305365562439, "global_step": 4885, "epoch": 54, "lr": 9.993535944165744e-05}
{"train_loss": 0.74483722448349, "global_step": 4886, "epoch": 54, "lr": 9.99353299687948e-05}
{"train_loss": 0.9633786678314209, "global_step": 4887, "epoch": 54, "lr": 9.993530048921898e-05}
{"train_loss": 0.9221247434616089, "global_step": 4888, "epoch": 54, "lr": 9.993527100292995e-05}
{"train_loss": 0.9066903591156006, "global_step": 4889, "epoch": 54, "lr": 9.993524150992774e-05}
{"train_loss": 1.0108684301376343, "global_step": 4890, "epoch": 54, "lr": 9.993521201021235e-05}
{"train_loss": 0.8818813562393188, "global_step": 4891, "epoch": 54, "lr": 9.993518250378376e-05}
{"train_loss": 0.9655906558036804, "global_step": 4892, "epoch": 54, "lr": 9.993515299064202e-05}
{"train_loss": 1.0725336074829102, "global_step": 4893, "epoch": 54, "lr": 9.993512347078709e-05}
{"train_loss": 0.982787329829141, "global_step": 4894, "epoch": 54, "lr": 9.993509394421898e-05, "val_loss": 1.144362211227417}
{"train_loss": 0.8742351531982422, "global_step": 4895, "epoch": 55, "lr": 9.993506441093772e-05}
{"train_loss": 1.0272066593170166, "global_step": 4896, "epoch": 55, "lr": 9.993503487094332e-05}
{"train_loss": 0.9345059990882874, "global_step": 4897, "epoch": 55, "lr": 9.993500532423573e-05}
{"train_loss": 0.8107420802116394, "global_step": 4898, "epoch": 55, "lr": 9.9934975770815e-05}
{"train_loss": 0.8622451424598694, "global_step": 4899, "epoch": 55, "lr": 9.993494621068112e-05}
{"train_loss": 1.0071582794189453, "global_step": 4900, "epoch": 55, "lr": 9.99349166438341e-05}
{"train_loss": 0.8776319026947021, "global_step": 4901, "epoch": 55, "lr": 9.993488707027393e-05}
{"train_loss": 0.8638260960578918, "global_step": 4902, "epoch": 55, "lr": 9.993485749000062e-05}
{"train_loss": 0.901885986328125, "global_step": 4903, "epoch": 55, "lr": 9.993482790301419e-05}
{"train_loss": 0.9046670198440552, "global_step": 4904, "epoch": 55, "lr": 9.993479830931463e-05}
{"train_loss": 1.148010015487671, "global_step": 4905, "epoch": 55, "lr": 9.993476870890195e-05}
{"train_loss": 0.7628613710403442, "global_step": 4906, "epoch": 55, "lr": 9.993473910177613e-05}
{"train_loss": 0.969700813293457, "global_step": 4907, "epoch": 55, "lr": 9.99347094879372e-05}
{"train_loss": 0.9580878019332886, "global_step": 4908, "epoch": 55, "lr": 9.993467986738516e-05}
{"train_loss": 0.8915271759033203, "global_step": 4909, "epoch": 55, "lr": 9.993465024012e-05}
{"train_loss": 0.888979971408844, "global_step": 4910, "epoch": 55, "lr": 9.993462060614176e-05}
{"train_loss": 0.9801043272018433, "global_step": 4911, "epoch": 55, "lr": 9.99345909654504e-05}
{"train_loss": 0.9269179105758667, "global_step": 4912, "epoch": 55, "lr": 9.993456131804595e-05}
{"train_loss": 1.0162246227264404, "global_step": 4913, "epoch": 55, "lr": 9.993453166392841e-05}
{"train_loss": 1.0698355436325073, "global_step": 4914, "epoch": 55, "lr": 9.993450200309778e-05}
{"train_loss": 1.1566648483276367, "global_step": 4915, "epoch": 55, "lr": 9.993447233555405e-05}
{"train_loss": 0.8768612146377563, "global_step": 4916, "epoch": 55, "lr": 9.993444266129726e-05}
{"train_loss": 0.8695074915885925, "global_step": 4917, "epoch": 55, "lr": 9.993441298032738e-05}
{"train_loss": 0.8922923803329468, "global_step": 4918, "epoch": 55, "lr": 9.993438329264442e-05}
{"train_loss": 1.0344334840774536, "global_step": 4919, "epoch": 55, "lr": 9.99343535982484e-05}
{"train_loss": 1.014397144317627, "global_step": 4920, "epoch": 55, "lr": 9.993432389713932e-05}
{"train_loss": 0.9556273221969604, "global_step": 4921, "epoch": 55, "lr": 9.993429418931716e-05}
{"train_loss": 0.7033050060272217, "global_step": 4922, "epoch": 55, "lr": 9.993426447478195e-05}
{"train_loss": 0.8473742008209229, "global_step": 4923, "epoch": 55, "lr": 9.99342347535337e-05}
{"train_loss": 1.1380010843276978, "global_step": 4924, "epoch": 55, "lr": 9.99342050255724e-05}
{"train_loss": 0.9619215130805969, "global_step": 4925, "epoch": 55, "lr": 9.993417529089804e-05}
{"train_loss": 0.8155103325843811, "global_step": 4926, "epoch": 55, "lr": 9.993414554951064e-05}
{"train_loss": 0.7903115153312683, "global_step": 4927, "epoch": 55, "lr": 9.993411580141022e-05}
{"train_loss": 0.8885788321495056, "global_step": 4928, "epoch": 55, "lr": 9.993408604659675e-05}
{"train_loss": 0.9017910957336426, "global_step": 4929, "epoch": 55, "lr": 9.993405628507027e-05}
{"train_loss": 0.8137122988700867, "global_step": 4930, "epoch": 55, "lr": 9.993402651683075e-05}
{"train_loss": 1.1516963243484497, "global_step": 4931, "epoch": 55, "lr": 9.99339967418782e-05}
{"train_loss": 0.8440536856651306, "global_step": 4932, "epoch": 55, "lr": 9.993396696021265e-05}
{"train_loss": 1.0062307119369507, "global_step": 4933, "epoch": 55, "lr": 9.993393717183408e-05}
{"train_loss": 0.9438084959983826, "global_step": 4934, "epoch": 55, "lr": 9.99339073767425e-05}
{"train_loss": 0.9845543503761292, "global_step": 4935, "epoch": 55, "lr": 9.993387757493791e-05}
{"train_loss": 1.0656768083572388, "global_step": 4936, "epoch": 55, "lr": 9.993384776642032e-05}
{"train_loss": 1.0655275583267212, "global_step": 4937, "epoch": 55, "lr": 9.993381795118975e-05}
{"train_loss": 1.0589178800582886, "global_step": 4938, "epoch": 55, "lr": 9.993378812924616e-05}
{"train_loss": 1.033798098564148, "global_step": 4939, "epoch": 55, "lr": 9.993375830058959e-05}
{"train_loss": 0.950226366519928, "global_step": 4940, "epoch": 55, "lr": 9.993372846522005e-05}
{"train_loss": 0.9532289505004883, "global_step": 4941, "epoch": 55, "lr": 9.993369862313751e-05}
{"train_loss": 0.9086039662361145, "global_step": 4942, "epoch": 55, "lr": 9.993366877434201e-05}
{"train_loss": 0.8654269576072693, "global_step": 4943, "epoch": 55, "lr": 9.993363891883352e-05}
{"train_loss": 0.8518527150154114, "global_step": 4944, "epoch": 55, "lr": 9.993360905661208e-05}
{"train_loss": 0.8479731678962708, "global_step": 4945, "epoch": 55, "lr": 9.993357918767766e-05}
{"train_loss": 0.9509256482124329, "global_step": 4946, "epoch": 55, "lr": 9.993354931203028e-05}
{"train_loss": 0.7479411363601685, "global_step": 4947, "epoch": 55, "lr": 9.993351942966995e-05}
{"train_loss": 1.166169285774231, "global_step": 4948, "epoch": 55, "lr": 9.993348954059667e-05}
{"train_loss": 1.062153935432434, "global_step": 4949, "epoch": 55, "lr": 9.993345964481042e-05}
{"train_loss": 0.9670084118843079, "global_step": 4950, "epoch": 55, "lr": 9.993342974231123e-05}
{"train_loss": 1.0177676677703857, "global_step": 4951, "epoch": 55, "lr": 9.993339983309912e-05}
{"train_loss": 1.0595834255218506, "global_step": 4952, "epoch": 55, "lr": 9.993336991717406e-05}
{"train_loss": 0.7369418740272522, "global_step": 4953, "epoch": 55, "lr": 9.993333999453607e-05}
{"train_loss": 0.8928627967834473, "global_step": 4954, "epoch": 55, "lr": 9.993331006518514e-05}
{"train_loss": 0.927478551864624, "global_step": 4955, "epoch": 55, "lr": 9.993328012912128e-05}
{"train_loss": 0.9346118569374084, "global_step": 4956, "epoch": 55, "lr": 9.993325018634452e-05}
{"train_loss": 0.8804879188537598, "global_step": 4957, "epoch": 55, "lr": 9.993322023685484e-05}
{"train_loss": 0.861831545829773, "global_step": 4958, "epoch": 55, "lr": 9.993319028065224e-05}
{"train_loss": 1.0788062810897827, "global_step": 4959, "epoch": 55, "lr": 9.993316031773674e-05}
{"train_loss": 0.968078076839447, "global_step": 4960, "epoch": 55, "lr": 9.993313034810832e-05}
{"train_loss": 0.7406824827194214, "global_step": 4961, "epoch": 55, "lr": 9.993310037176702e-05}
{"train_loss": 0.9789761304855347, "global_step": 4962, "epoch": 55, "lr": 9.99330703887128e-05}
{"train_loss": 0.9487571716308594, "global_step": 4963, "epoch": 55, "lr": 9.993304039894571e-05}
{"train_loss": 1.011366367340088, "global_step": 4964, "epoch": 55, "lr": 9.99330104024657e-05}
{"train_loss": 0.9653134942054749, "global_step": 4965, "epoch": 55, "lr": 9.993298039927284e-05}
{"train_loss": 0.9446437358856201, "global_step": 4966, "epoch": 55, "lr": 9.993295038936708e-05}
{"train_loss": 0.9125608801841736, "global_step": 4967, "epoch": 55, "lr": 9.993292037274845e-05}
{"train_loss": 0.9686986804008484, "global_step": 4968, "epoch": 55, "lr": 9.993289034941695e-05}
{"train_loss": 1.152726173400879, "global_step": 4969, "epoch": 55, "lr": 9.993286031937256e-05}
{"train_loss": 1.0674412250518799, "global_step": 4970, "epoch": 55, "lr": 9.993283028261533e-05}
{"train_loss": 0.8604623675346375, "global_step": 4971, "epoch": 55, "lr": 9.993280023914524e-05}
{"train_loss": 1.0695325136184692, "global_step": 4972, "epoch": 55, "lr": 9.993277018896228e-05}
{"train_loss": 0.8789753913879395, "global_step": 4973, "epoch": 55, "lr": 9.993274013206648e-05}
{"train_loss": 1.0556604862213135, "global_step": 4974, "epoch": 55, "lr": 9.993271006845781e-05}
{"train_loss": 0.9505290389060974, "global_step": 4975, "epoch": 55, "lr": 9.993267999813631e-05}
{"train_loss": 0.9186868071556091, "global_step": 4976, "epoch": 55, "lr": 9.993264992110198e-05}
{"train_loss": 1.1517572402954102, "global_step": 4977, "epoch": 55, "lr": 9.99326198373548e-05}
{"train_loss": 1.0762983560562134, "global_step": 4978, "epoch": 55, "lr": 9.993258974689478e-05}
{"train_loss": 1.0092862844467163, "global_step": 4979, "epoch": 55, "lr": 9.993255964972197e-05}
{"train_loss": 0.9911304116249084, "global_step": 4980, "epoch": 55, "lr": 9.993252954583629e-05}
{"train_loss": 0.9402837753295898, "global_step": 4981, "epoch": 55, "lr": 9.993249943523781e-05}
{"train_loss": 0.9580274224281311, "global_step": 4982, "epoch": 55, "lr": 9.993246931792653e-05}
{"train_loss": 0.9512132727697994, "global_step": 4983, "epoch": 55, "lr": 9.993243919390242e-05, "val_loss": 1.1211004257202148, "train_action_mse_error": 51.02386474609375}
{"train_loss": 0.8529114723205566, "global_step": 4984, "epoch": 56, "lr": 9.99324090631655e-05}
{"train_loss": 0.856928288936615, "global_step": 4985, "epoch": 56, "lr": 9.993237892571579e-05}
{"train_loss": 0.9496259093284607, "global_step": 4986, "epoch": 56, "lr": 9.993234878155326e-05}
{"train_loss": 0.8692200183868408, "global_step": 4987, "epoch": 56, "lr": 9.993231863067795e-05}
{"train_loss": 1.074691653251648, "global_step": 4988, "epoch": 56, "lr": 9.993228847308983e-05}
{"train_loss": 0.8681009411811829, "global_step": 4989, "epoch": 56, "lr": 9.993225830878894e-05}
{"train_loss": 0.9842247366905212, "global_step": 4990, "epoch": 56, "lr": 9.993222813777528e-05}
{"train_loss": 0.9640782475471497, "global_step": 4991, "epoch": 56, "lr": 9.993219796004882e-05}
{"train_loss": 0.8958231806755066, "global_step": 4992, "epoch": 56, "lr": 9.993216777560957e-05}
{"train_loss": 0.9906089305877686, "global_step": 4993, "epoch": 56, "lr": 9.993213758445757e-05}
{"train_loss": 1.132179856300354, "global_step": 4994, "epoch": 56, "lr": 9.99321073865928e-05}
{"train_loss": 1.0075584650039673, "global_step": 4995, "epoch": 56, "lr": 9.993207718201527e-05}
{"train_loss": 0.9062501192092896, "global_step": 4996, "epoch": 56, "lr": 9.993204697072498e-05}
{"train_loss": 0.8440624475479126, "global_step": 4997, "epoch": 56, "lr": 9.993201675272192e-05}
{"train_loss": 0.7921074032783508, "global_step": 4998, "epoch": 56, "lr": 9.993198652800613e-05}
{"train_loss": 0.826891303062439, "global_step": 4999, "epoch": 56, "lr": 9.993195629657757e-05}
{"train_loss": 1.0352400541305542, "global_step": 5000, "epoch": 56, "lr": 9.993192605843629e-05}
{"train_loss": 0.9747372269630432, "global_step": 5001, "epoch": 56, "lr": 9.993189581358225e-05}
{"train_loss": 0.8134956359863281, "global_step": 5002, "epoch": 56, "lr": 9.99318655620155e-05}
{"train_loss": 0.7950087189674377, "global_step": 5003, "epoch": 56, "lr": 9.9931835303736e-05}
{"train_loss": 0.9631023406982422, "global_step": 5004, "epoch": 56, "lr": 9.993180503874377e-05}
{"train_loss": 0.9507957696914673, "global_step": 5005, "epoch": 56, "lr": 9.993177476703883e-05}
{"train_loss": 0.9003918170928955, "global_step": 5006, "epoch": 56, "lr": 9.993174448862117e-05}
{"train_loss": 1.0414564609527588, "global_step": 5007, "epoch": 56, "lr": 9.993171420349078e-05}
{"train_loss": 0.7678854465484619, "global_step": 5008, "epoch": 56, "lr": 9.993168391164771e-05}
{"train_loss": 0.9566177725791931, "global_step": 5009, "epoch": 56, "lr": 9.993165361309192e-05}
{"train_loss": 0.772320568561554, "global_step": 5010, "epoch": 56, "lr": 9.993162330782342e-05}
{"train_loss": 0.9857920408248901, "global_step": 5011, "epoch": 56, "lr": 9.993159299584222e-05}
{"train_loss": 0.9919988512992859, "global_step": 5012, "epoch": 56, "lr": 9.993156267714834e-05}
{"train_loss": 0.7712094783782959, "global_step": 5013, "epoch": 56, "lr": 9.993153235174175e-05}
{"train_loss": 0.9524126648902893, "global_step": 5014, "epoch": 56, "lr": 9.993150201962249e-05}
{"train_loss": 0.8764636516571045, "global_step": 5015, "epoch": 56, "lr": 9.993147168079055e-05}
{"train_loss": 1.1066818237304688, "global_step": 5016, "epoch": 56, "lr": 9.993144133524593e-05}
{"train_loss": 0.8871915936470032, "global_step": 5017, "epoch": 56, "lr": 9.993141098298862e-05}
{"train_loss": 0.9209460020065308, "global_step": 5018, "epoch": 56, "lr": 9.993138062401866e-05}
{"train_loss": 1.0921523571014404, "global_step": 5019, "epoch": 56, "lr": 9.993135025833602e-05}
{"train_loss": 0.9435528516769409, "global_step": 5020, "epoch": 56, "lr": 9.993131988594074e-05}
{"train_loss": 0.8346403241157532, "global_step": 5021, "epoch": 56, "lr": 9.993128950683278e-05}
{"train_loss": 0.9297210574150085, "global_step": 5022, "epoch": 56, "lr": 9.993125912101218e-05}
{"train_loss": 1.0859529972076416, "global_step": 5023, "epoch": 56, "lr": 9.993122872847894e-05}
{"train_loss": 1.0029637813568115, "global_step": 5024, "epoch": 56, "lr": 9.993119832923303e-05}
{"train_loss": 0.9826633334159851, "global_step": 5025, "epoch": 56, "lr": 9.993116792327448e-05}
{"train_loss": 1.1027474403381348, "global_step": 5026, "epoch": 56, "lr": 9.99311375106033e-05}
{"train_loss": 0.9462237358093262, "global_step": 5027, "epoch": 56, "lr": 9.99311070912195e-05}
{"train_loss": 0.8495662808418274, "global_step": 5028, "epoch": 56, "lr": 9.993107666512306e-05}
{"train_loss": 0.9711830615997314, "global_step": 5029, "epoch": 56, "lr": 9.9931046232314e-05}
{"train_loss": 0.9855025410652161, "global_step": 5030, "epoch": 56, "lr": 9.993101579279232e-05}
{"train_loss": 1.1319345235824585, "global_step": 5031, "epoch": 56, "lr": 9.993098534655801e-05}
{"train_loss": 0.9688283801078796, "global_step": 5032, "epoch": 56, "lr": 9.993095489361111e-05}
{"train_loss": 0.94273442029953, "global_step": 5033, "epoch": 56, "lr": 9.993092443395158e-05}
{"train_loss": 1.0629069805145264, "global_step": 5034, "epoch": 56, "lr": 9.993089396757945e-05}
{"train_loss": 0.8702434301376343, "global_step": 5035, "epoch": 56, "lr": 9.993086349449473e-05}
{"train_loss": 1.14336097240448, "global_step": 5036, "epoch": 56, "lr": 9.993083301469741e-05}
{"train_loss": 1.0254249572753906, "global_step": 5037, "epoch": 56, "lr": 9.99308025281875e-05}
{"train_loss": 0.8076215982437134, "global_step": 5038, "epoch": 56, "lr": 9.9930772034965e-05}
{"train_loss": 1.07716703414917, "global_step": 5039, "epoch": 56, "lr": 9.99307415350299e-05}
{"train_loss": 1.0781426429748535, "global_step": 5040, "epoch": 56, "lr": 9.993071102838225e-05}
{"train_loss": 1.1632188558578491, "global_step": 5041, "epoch": 56, "lr": 9.9930680515022e-05}
{"train_loss": 0.8846249580383301, "global_step": 5042, "epoch": 56, "lr": 9.993064999494918e-05}
{"train_loss": 0.9053745269775391, "global_step": 5043, "epoch": 56, "lr": 9.99306194681638e-05}
{"train_loss": 0.8661502003669739, "global_step": 5044, "epoch": 56, "lr": 9.993058893466587e-05}
{"train_loss": 1.0885692834854126, "global_step": 5045, "epoch": 56, "lr": 9.993055839445538e-05}
{"train_loss": 0.9127156138420105, "global_step": 5046, "epoch": 56, "lr": 9.993052784753231e-05}
{"train_loss": 1.0335583686828613, "global_step": 5047, "epoch": 56, "lr": 9.993049729389671e-05}
{"train_loss": 0.8185780048370361, "global_step": 5048, "epoch": 56, "lr": 9.993046673354855e-05}
{"train_loss": 0.7951616644859314, "global_step": 5049, "epoch": 56, "lr": 9.993043616648786e-05}
{"train_loss": 1.1241456270217896, "global_step": 5050, "epoch": 56, "lr": 9.993040559271462e-05}
{"train_loss": 0.865038275718689, "global_step": 5051, "epoch": 56, "lr": 9.993037501222884e-05}
{"train_loss": 0.8024862408638, "global_step": 5052, "epoch": 56, "lr": 9.993034442503056e-05}
{"train_loss": 1.0324969291687012, "global_step": 5053, "epoch": 56, "lr": 9.993031383111973e-05}
{"train_loss": 1.137681484222412, "global_step": 5054, "epoch": 56, "lr": 9.993028323049638e-05}
{"train_loss": 1.1443047523498535, "global_step": 5055, "epoch": 56, "lr": 9.993025262316051e-05}
{"train_loss": 0.7877836227416992, "global_step": 5056, "epoch": 56, "lr": 9.993022200911213e-05}
{"train_loss": 0.9920486211776733, "global_step": 5057, "epoch": 56, "lr": 9.993019138835124e-05}
{"train_loss": 0.9053454399108887, "global_step": 5058, "epoch": 56, "lr": 9.993016076087785e-05}
{"train_loss": 1.098577857017517, "global_step": 5059, "epoch": 56, "lr": 9.993013012669195e-05}
{"train_loss": 0.9452729225158691, "global_step": 5060, "epoch": 56, "lr": 9.993009948579355e-05}
{"train_loss": 0.8673340678215027, "global_step": 5061, "epoch": 56, "lr": 9.993006883818267e-05}
{"train_loss": 0.994098961353302, "global_step": 5062, "epoch": 56, "lr": 9.993003818385929e-05}
{"train_loss": 1.0803935527801514, "global_step": 5063, "epoch": 56, "lr": 9.993000752282342e-05}
{"train_loss": 0.890178918838501, "global_step": 5064, "epoch": 56, "lr": 9.992997685507509e-05}
{"train_loss": 0.8824174404144287, "global_step": 5065, "epoch": 56, "lr": 9.992994618061427e-05}
{"train_loss": 0.9061271548271179, "global_step": 5066, "epoch": 56, "lr": 9.992991549944097e-05}
{"train_loss": 1.1223019361495972, "global_step": 5067, "epoch": 56, "lr": 9.992988481155521e-05}
{"train_loss": 1.0769100189208984, "global_step": 5068, "epoch": 56, "lr": 9.9929854116957e-05}
{"train_loss": 0.9653593301773071, "global_step": 5069, "epoch": 56, "lr": 9.99298234156463e-05}
{"train_loss": 1.0178855657577515, "global_step": 5070, "epoch": 56, "lr": 9.992979270762315e-05}
{"train_loss": 0.9447652697563171, "global_step": 5071, "epoch": 56, "lr": 9.992976199288756e-05}
{"train_loss": 0.9574468142530891, "global_step": 5072, "epoch": 56, "lr": 9.992973127143952e-05, "val_loss": 1.1252354383468628}
{"train_loss": 0.882460355758667, "global_step": 5073, "epoch": 57, "lr": 9.992970054327904e-05}
{"train_loss": 0.9094142317771912, "global_step": 5074, "epoch": 57, "lr": 9.99296698084061e-05}
{"train_loss": 1.0103840827941895, "global_step": 5075, "epoch": 57, "lr": 9.992963906682074e-05}
{"train_loss": 0.853218138217926, "global_step": 5076, "epoch": 57, "lr": 9.992960831852295e-05}
{"train_loss": 1.0819447040557861, "global_step": 5077, "epoch": 57, "lr": 9.992957756351273e-05}
{"train_loss": 1.043458104133606, "global_step": 5078, "epoch": 57, "lr": 9.992954680179008e-05}
{"train_loss": 0.6562656760215759, "global_step": 5079, "epoch": 57, "lr": 9.992951603335503e-05}
{"train_loss": 0.8337490558624268, "global_step": 5080, "epoch": 57, "lr": 9.992948525820756e-05}
{"train_loss": 0.9130958318710327, "global_step": 5081, "epoch": 57, "lr": 9.992945447634767e-05}
{"train_loss": 0.9898018836975098, "global_step": 5082, "epoch": 57, "lr": 9.992942368777536e-05}
{"train_loss": 1.063452124595642, "global_step": 5083, "epoch": 57, "lr": 9.992939289249068e-05}
{"train_loss": 0.90401291847229, "global_step": 5084, "epoch": 57, "lr": 9.992936209049357e-05}
{"train_loss": 1.031347632408142, "global_step": 5085, "epoch": 57, "lr": 9.992933128178408e-05}
{"train_loss": 1.017549991607666, "global_step": 5086, "epoch": 57, "lr": 9.99293004663622e-05}
{"train_loss": 1.0902457237243652, "global_step": 5087, "epoch": 57, "lr": 9.992926964422794e-05}
{"train_loss": 0.9751477241516113, "global_step": 5088, "epoch": 57, "lr": 9.99292388153813e-05}
{"train_loss": 0.9835498332977295, "global_step": 5089, "epoch": 57, "lr": 9.992920797982227e-05}
{"train_loss": 0.8857231736183167, "global_step": 5090, "epoch": 57, "lr": 9.992917713755087e-05}
{"train_loss": 0.8362894654273987, "global_step": 5091, "epoch": 57, "lr": 9.99291462885671e-05}
{"train_loss": 1.008563756942749, "global_step": 5092, "epoch": 57, "lr": 9.992911543287097e-05}
{"train_loss": 0.9245576858520508, "global_step": 5093, "epoch": 57, "lr": 9.992908457046249e-05}
{"train_loss": 0.8598975539207458, "global_step": 5094, "epoch": 57, "lr": 9.992905370134163e-05}
{"train_loss": 0.8141947984695435, "global_step": 5095, "epoch": 57, "lr": 9.992902282550842e-05}
{"train_loss": 1.1475918292999268, "global_step": 5096, "epoch": 57, "lr": 9.992899194296286e-05}
{"train_loss": 0.9176505208015442, "global_step": 5097, "epoch": 57, "lr": 9.992896105370497e-05}
{"train_loss": 0.8205633759498596, "global_step": 5098, "epoch": 57, "lr": 9.992893015773474e-05}
{"train_loss": 1.0343677997589111, "global_step": 5099, "epoch": 57, "lr": 9.992889925505217e-05}
{"train_loss": 0.9563487768173218, "global_step": 5100, "epoch": 57, "lr": 9.992886834565727e-05}
{"train_loss": 1.0162954330444336, "global_step": 5101, "epoch": 57, "lr": 9.992883742955003e-05}
{"train_loss": 0.9608868956565857, "global_step": 5102, "epoch": 57, "lr": 9.992880650673047e-05}
{"train_loss": 1.010725975036621, "global_step": 5103, "epoch": 57, "lr": 9.99287755771986e-05}
{"train_loss": 0.9497447609901428, "global_step": 5104, "epoch": 57, "lr": 9.99287446409544e-05}
{"train_loss": 0.8238652944564819, "global_step": 5105, "epoch": 57, "lr": 9.99287136979979e-05}
{"train_loss": 0.9811946749687195, "global_step": 5106, "epoch": 57, "lr": 9.99286827483291e-05}
{"train_loss": 0.9870485663414001, "global_step": 5107, "epoch": 57, "lr": 9.992865179194798e-05}
{"train_loss": 0.8019059300422668, "global_step": 5108, "epoch": 57, "lr": 9.992862082885458e-05}
{"train_loss": 0.9902544021606445, "global_step": 5109, "epoch": 57, "lr": 9.992858985904888e-05}
{"train_loss": 0.7493945360183716, "global_step": 5110, "epoch": 57, "lr": 9.992855888253088e-05}
{"train_loss": 0.876846969127655, "global_step": 5111, "epoch": 57, "lr": 9.992852789930059e-05}
{"train_loss": 0.9233763217926025, "global_step": 5112, "epoch": 57, "lr": 9.992849690935803e-05}
{"train_loss": 0.8254044055938721, "global_step": 5113, "epoch": 57, "lr": 9.992846591270319e-05}
{"train_loss": 1.0312893390655518, "global_step": 5114, "epoch": 57, "lr": 9.992843490933608e-05}
{"train_loss": 1.187248706817627, "global_step": 5115, "epoch": 57, "lr": 9.992840389925669e-05}
{"train_loss": 0.8623554706573486, "global_step": 5116, "epoch": 57, "lr": 9.992837288246504e-05}
{"train_loss": 1.0000026226043701, "global_step": 5117, "epoch": 57, "lr": 9.992834185896111e-05}
{"train_loss": 0.886738121509552, "global_step": 5118, "epoch": 57, "lr": 9.992831082874495e-05}
{"train_loss": 0.8191471099853516, "global_step": 5119, "epoch": 57, "lr": 9.992827979181654e-05}
{"train_loss": 0.83668053150177, "global_step": 5120, "epoch": 57, "lr": 9.992824874817586e-05}
{"train_loss": 0.9290317296981812, "global_step": 5121, "epoch": 57, "lr": 9.992821769782294e-05}
{"train_loss": 0.8933439254760742, "global_step": 5122, "epoch": 57, "lr": 9.992818664075779e-05}
{"train_loss": 0.9545549154281616, "global_step": 5123, "epoch": 57, "lr": 9.992815557698041e-05}
{"train_loss": 0.9495030045509338, "global_step": 5124, "epoch": 57, "lr": 9.992812450649078e-05}
{"train_loss": 1.0570218563079834, "global_step": 5125, "epoch": 57, "lr": 9.992809342928893e-05}
{"train_loss": 0.9290884137153625, "global_step": 5126, "epoch": 57, "lr": 9.992806234537486e-05}
{"train_loss": 0.7485989332199097, "global_step": 5127, "epoch": 57, "lr": 9.992803125474857e-05}
{"train_loss": 0.8350794315338135, "global_step": 5128, "epoch": 57, "lr": 9.992800015741006e-05}
{"train_loss": 1.0469343662261963, "global_step": 5129, "epoch": 57, "lr": 9.992796905335933e-05}
{"train_loss": 0.9256532192230225, "global_step": 5130, "epoch": 57, "lr": 9.99279379425964e-05}
{"train_loss": 0.9001988768577576, "global_step": 5131, "epoch": 57, "lr": 9.992790682512128e-05}
{"train_loss": 0.8102205991744995, "global_step": 5132, "epoch": 57, "lr": 9.992787570093395e-05}
{"train_loss": 0.9037083387374878, "global_step": 5133, "epoch": 57, "lr": 9.992784457003444e-05}
{"train_loss": 1.0442699193954468, "global_step": 5134, "epoch": 57, "lr": 9.992781343242272e-05}
{"train_loss": 0.8419303894042969, "global_step": 5135, "epoch": 57, "lr": 9.992778228809881e-05}
{"train_loss": 0.8466886281967163, "global_step": 5136, "epoch": 57, "lr": 9.992775113706274e-05}
{"train_loss": 1.0646754503250122, "global_step": 5137, "epoch": 57, "lr": 9.992771997931447e-05}
{"train_loss": 0.8383426666259766, "global_step": 5138, "epoch": 57, "lr": 9.992768881485405e-05}
{"train_loss": 0.8158254027366638, "global_step": 5139, "epoch": 57, "lr": 9.992765764368144e-05}
{"train_loss": 0.9888799786567688, "global_step": 5140, "epoch": 57, "lr": 9.992762646579667e-05}
{"train_loss": 1.0210884809494019, "global_step": 5141, "epoch": 57, "lr": 9.992759528119975e-05}
{"train_loss": 0.9112282395362854, "global_step": 5142, "epoch": 57, "lr": 9.992756408989066e-05}
{"train_loss": 0.9050679206848145, "global_step": 5143, "epoch": 57, "lr": 9.992753289186943e-05}
{"train_loss": 1.1397130489349365, "global_step": 5144, "epoch": 57, "lr": 9.992750168713606e-05}
{"train_loss": 1.0623575448989868, "global_step": 5145, "epoch": 57, "lr": 9.992747047569052e-05}
{"train_loss": 0.7763786315917969, "global_step": 5146, "epoch": 57, "lr": 9.992743925753285e-05}
{"train_loss": 0.8823336362838745, "global_step": 5147, "epoch": 57, "lr": 9.992740803266304e-05}
{"train_loss": 0.9318125247955322, "global_step": 5148, "epoch": 57, "lr": 9.992737680108111e-05}
{"train_loss": 0.985079824924469, "global_step": 5149, "epoch": 57, "lr": 9.992734556278705e-05}
{"train_loss": 0.9426295161247253, "global_step": 5150, "epoch": 57, "lr": 9.992731431778086e-05}
{"train_loss": 0.814802885055542, "global_step": 5151, "epoch": 57, "lr": 9.992728306606255e-05}
{"train_loss": 1.172146201133728, "global_step": 5152, "epoch": 57, "lr": 9.992725180763214e-05}
{"train_loss": 0.8204575777053833, "global_step": 5153, "epoch": 57, "lr": 9.992722054248962e-05}
{"train_loss": 0.971288800239563, "global_step": 5154, "epoch": 57, "lr": 9.992718927063498e-05}
{"train_loss": 1.015455722808838, "global_step": 5155, "epoch": 57, "lr": 9.992715799206825e-05}
{"train_loss": 1.048620343208313, "global_step": 5156, "epoch": 57, "lr": 9.992712670678941e-05}
{"train_loss": 1.0075404644012451, "global_step": 5157, "epoch": 57, "lr": 9.992709541479848e-05}
{"train_loss": 1.036779522895813, "global_step": 5158, "epoch": 57, "lr": 9.992706411609546e-05}
{"train_loss": 0.8192185759544373, "global_step": 5159, "epoch": 57, "lr": 9.992703281068035e-05}
{"train_loss": 0.9104135036468506, "global_step": 5160, "epoch": 57, "lr": 9.992700149855316e-05}
{"train_loss": 0.9386468750707219, "global_step": 5161, "epoch": 57, "lr": 9.99269701797139e-05, "val_loss": 1.1321465969085693}
{"train_loss": 0.9658505916595459, "global_step": 5162, "epoch": 58, "lr": 9.992693885416258e-05}
{"train_loss": 0.9419940710067749, "global_step": 5163, "epoch": 58, "lr": 9.992690752189917e-05}
{"train_loss": 0.7659512758255005, "global_step": 5164, "epoch": 58, "lr": 9.99268761829237e-05}
{"train_loss": 0.8288857936859131, "global_step": 5165, "epoch": 58, "lr": 9.992684483723617e-05}
{"train_loss": 1.1435153484344482, "global_step": 5166, "epoch": 58, "lr": 9.992681348483659e-05}
{"train_loss": 0.885076105594635, "global_step": 5167, "epoch": 58, "lr": 9.992678212572495e-05}
{"train_loss": 1.0204695463180542, "global_step": 5168, "epoch": 58, "lr": 9.992675075990127e-05}
{"train_loss": 0.9229691624641418, "global_step": 5169, "epoch": 58, "lr": 9.992671938736555e-05}
{"train_loss": 0.9192073941230774, "global_step": 5170, "epoch": 58, "lr": 9.992668800811778e-05}
{"train_loss": 0.8774765729904175, "global_step": 5171, "epoch": 58, "lr": 9.992665662215799e-05}
{"train_loss": 0.7814776301383972, "global_step": 5172, "epoch": 58, "lr": 9.992662522948616e-05}
{"train_loss": 0.9633482098579407, "global_step": 5173, "epoch": 58, "lr": 9.99265938301023e-05}
{"train_loss": 0.9161373972892761, "global_step": 5174, "epoch": 58, "lr": 9.992656242400642e-05}
{"train_loss": 0.9798516035079956, "global_step": 5175, "epoch": 58, "lr": 9.992653101119853e-05}
{"train_loss": 0.8120191693305969, "global_step": 5176, "epoch": 58, "lr": 9.992649959167862e-05}
{"train_loss": 0.8353706002235413, "global_step": 5177, "epoch": 58, "lr": 9.992646816544671e-05}
{"train_loss": 0.8485450148582458, "global_step": 5178, "epoch": 58, "lr": 9.992643673250279e-05}
{"train_loss": 0.8353564739227295, "global_step": 5179, "epoch": 58, "lr": 9.992640529284686e-05}
{"train_loss": 1.018783450126648, "global_step": 5180, "epoch": 58, "lr": 9.992637384647894e-05}
{"train_loss": 0.8756048679351807, "global_step": 5181, "epoch": 58, "lr": 9.992634239339903e-05}
{"train_loss": 0.9056791663169861, "global_step": 5182, "epoch": 58, "lr": 9.992631093360711e-05}
{"train_loss": 1.0660693645477295, "global_step": 5183, "epoch": 58, "lr": 9.992627946710323e-05}
{"train_loss": 0.9652994275093079, "global_step": 5184, "epoch": 58, "lr": 9.992624799388737e-05}
{"train_loss": 0.9051972031593323, "global_step": 5185, "epoch": 58, "lr": 9.992621651395953e-05}
{"train_loss": 0.9584802985191345, "global_step": 5186, "epoch": 58, "lr": 9.992618502731972e-05}
{"train_loss": 0.8734596967697144, "global_step": 5187, "epoch": 58, "lr": 9.992615353396795e-05}
{"train_loss": 1.0035825967788696, "global_step": 5188, "epoch": 58, "lr": 9.99261220339042e-05}
{"train_loss": 0.9551369547843933, "global_step": 5189, "epoch": 58, "lr": 9.99260905271285e-05}
{"train_loss": 0.9815607070922852, "global_step": 5190, "epoch": 58, "lr": 9.992605901364085e-05}
{"train_loss": 0.9583044648170471, "global_step": 5191, "epoch": 58, "lr": 9.992602749344124e-05}
{"train_loss": 0.9997251033782959, "global_step": 5192, "epoch": 58, "lr": 9.99259959665297e-05}
{"train_loss": 0.8340536952018738, "global_step": 5193, "epoch": 58, "lr": 9.992596443290619e-05}
{"train_loss": 1.1036282777786255, "global_step": 5194, "epoch": 58, "lr": 9.992593289257077e-05}
{"train_loss": 0.8250203132629395, "global_step": 5195, "epoch": 58, "lr": 9.99259013455234e-05}
{"train_loss": 0.8881521224975586, "global_step": 5196, "epoch": 58, "lr": 9.99258697917641e-05}
{"train_loss": 0.8531764149665833, "global_step": 5197, "epoch": 58, "lr": 9.992583823129289e-05}
{"train_loss": 0.7773436307907104, "global_step": 5198, "epoch": 58, "lr": 9.992580666410975e-05}
{"train_loss": 0.9460753798484802, "global_step": 5199, "epoch": 58, "lr": 9.99257750902147e-05}
{"train_loss": 1.043151617050171, "global_step": 5200, "epoch": 58, "lr": 9.992574350960774e-05}
{"train_loss": 0.8955914974212646, "global_step": 5201, "epoch": 58, "lr": 9.992571192228886e-05}
{"train_loss": 1.0453130006790161, "global_step": 5202, "epoch": 58, "lr": 9.992568032825809e-05}
{"train_loss": 0.9518733620643616, "global_step": 5203, "epoch": 58, "lr": 9.992564872751542e-05}
{"train_loss": 0.9120595455169678, "global_step": 5204, "epoch": 58, "lr": 9.992561712006084e-05}
{"train_loss": 1.0573335886001587, "global_step": 5205, "epoch": 58, "lr": 9.992558550589438e-05}
{"train_loss": 0.7742291688919067, "global_step": 5206, "epoch": 58, "lr": 9.992555388501603e-05}
{"train_loss": 0.9615147709846497, "global_step": 5207, "epoch": 58, "lr": 9.992552225742581e-05}
{"train_loss": 1.00043785572052, "global_step": 5208, "epoch": 58, "lr": 9.99254906231237e-05}
{"train_loss": 0.9332196712493896, "global_step": 5209, "epoch": 58, "lr": 9.992545898210973e-05}
{"train_loss": 0.9830412864685059, "global_step": 5210, "epoch": 58, "lr": 9.992542733438387e-05}
{"train_loss": 1.19679594039917, "global_step": 5211, "epoch": 58, "lr": 9.992539567994614e-05}
{"train_loss": 1.1270381212234497, "global_step": 5212, "epoch": 58, "lr": 9.992536401879657e-05}
{"train_loss": 1.0689347982406616, "global_step": 5213, "epoch": 58, "lr": 9.992533235093513e-05}
{"train_loss": 1.0585098266601562, "global_step": 5214, "epoch": 58, "lr": 9.992530067636185e-05}
{"train_loss": 0.9120830297470093, "global_step": 5215, "epoch": 58, "lr": 9.992526899507672e-05}
{"train_loss": 0.76844322681427, "global_step": 5216, "epoch": 58, "lr": 9.992523730707974e-05}
{"train_loss": 0.8240532279014587, "global_step": 5217, "epoch": 58, "lr": 9.992520561237092e-05}
{"train_loss": 0.988508403301239, "global_step": 5218, "epoch": 58, "lr": 9.992517391095026e-05}
{"train_loss": 0.9662478566169739, "global_step": 5219, "epoch": 58, "lr": 9.992514220281779e-05}
{"train_loss": 0.9184591770172119, "global_step": 5220, "epoch": 58, "lr": 9.992511048797347e-05}
{"train_loss": 0.85748291015625, "global_step": 5221, "epoch": 58, "lr": 9.992507876641734e-05}
{"train_loss": 1.1360127925872803, "global_step": 5222, "epoch": 58, "lr": 9.992504703814939e-05}
{"train_loss": 1.0227395296096802, "global_step": 5223, "epoch": 58, "lr": 9.992501530316962e-05}
{"train_loss": 0.9127802848815918, "global_step": 5224, "epoch": 58, "lr": 9.992498356147803e-05}
{"train_loss": 0.8473502993583679, "global_step": 5225, "epoch": 58, "lr": 9.992495181307466e-05}
{"train_loss": 1.1599342823028564, "global_step": 5226, "epoch": 58, "lr": 9.992492005795948e-05}
{"train_loss": 0.9567104578018188, "global_step": 5227, "epoch": 58, "lr": 9.99248882961325e-05}
{"train_loss": 0.7995880246162415, "global_step": 5228, "epoch": 58, "lr": 9.992485652759372e-05}
{"train_loss": 1.0801734924316406, "global_step": 5229, "epoch": 58, "lr": 9.992482475234318e-05}
{"train_loss": 1.067257285118103, "global_step": 5230, "epoch": 58, "lr": 9.992479297038082e-05}
{"train_loss": 0.9344308972358704, "global_step": 5231, "epoch": 58, "lr": 9.99247611817067e-05}
{"train_loss": 0.8404463529586792, "global_step": 5232, "epoch": 58, "lr": 9.992472938632079e-05}
{"train_loss": 1.0095468759536743, "global_step": 5233, "epoch": 58, "lr": 9.992469758422312e-05}
{"train_loss": 1.0417046546936035, "global_step": 5234, "epoch": 58, "lr": 9.992466577541368e-05}
{"train_loss": 0.9722194075584412, "global_step": 5235, "epoch": 58, "lr": 9.992463395989248e-05}
{"train_loss": 1.0083281993865967, "global_step": 5236, "epoch": 58, "lr": 9.992460213765952e-05}
{"train_loss": 0.9138780832290649, "global_step": 5237, "epoch": 58, "lr": 9.99245703087148e-05}
{"train_loss": 0.9218873977661133, "global_step": 5238, "epoch": 58, "lr": 9.992453847305834e-05}
{"train_loss": 1.025010585784912, "global_step": 5239, "epoch": 58, "lr": 9.992450663069013e-05}
{"train_loss": 0.9188688397407532, "global_step": 5240, "epoch": 58, "lr": 9.992447478161017e-05}
{"train_loss": 0.8456118106842041, "global_step": 5241, "epoch": 58, "lr": 9.992444292581849e-05}
{"train_loss": 0.9521478414535522, "global_step": 5242, "epoch": 58, "lr": 9.992441106331507e-05}
{"train_loss": 0.9388490319252014, "global_step": 5243, "epoch": 58, "lr": 9.992437919409992e-05}
{"train_loss": 1.0034122467041016, "global_step": 5244, "epoch": 58, "lr": 9.992434731817304e-05}
{"train_loss": 0.9669091701507568, "global_step": 5245, "epoch": 58, "lr": 9.992431543553444e-05}
{"train_loss": 0.7574057579040527, "global_step": 5246, "epoch": 58, "lr": 9.992428354618413e-05}
{"train_loss": 1.0052731037139893, "global_step": 5247, "epoch": 58, "lr": 9.992425165012211e-05}
{"train_loss": 0.9814535975456238, "global_step": 5248, "epoch": 58, "lr": 9.992421974734838e-05}
{"train_loss": 1.1145817041397095, "global_step": 5249, "epoch": 58, "lr": 9.992418783786295e-05}
{"train_loss": 0.946201180474142, "global_step": 5250, "epoch": 58, "lr": 9.992415592166582e-05, "val_loss": 1.1431686878204346}
{"train_loss": 0.9067461490631104, "global_step": 5251, "epoch": 59, "lr": 9.9924123998757e-05}
{"train_loss": 0.9447338581085205, "global_step": 5252, "epoch": 59, "lr": 9.992409206913648e-05}
{"train_loss": 0.8429723978042603, "global_step": 5253, "epoch": 59, "lr": 9.992406013280428e-05}
{"train_loss": 1.091352105140686, "global_step": 5254, "epoch": 59, "lr": 9.99240281897604e-05}
{"train_loss": 0.8756594657897949, "global_step": 5255, "epoch": 59, "lr": 9.992399624000484e-05}
{"train_loss": 0.9717439413070679, "global_step": 5256, "epoch": 59, "lr": 9.992396428353761e-05}
{"train_loss": 1.0556143522262573, "global_step": 5257, "epoch": 59, "lr": 9.99239323203587e-05}
{"train_loss": 0.9914954900741577, "global_step": 5258, "epoch": 59, "lr": 9.992390035046814e-05}
{"train_loss": 1.0153234004974365, "global_step": 5259, "epoch": 59, "lr": 9.992386837386591e-05}
{"train_loss": 0.8307526707649231, "global_step": 5260, "epoch": 59, "lr": 9.992383639055203e-05}
{"train_loss": 0.9214857816696167, "global_step": 5261, "epoch": 59, "lr": 9.99238044005265e-05}
{"train_loss": 1.2005234956741333, "global_step": 5262, "epoch": 59, "lr": 9.992377240378932e-05}
{"train_loss": 0.9958641529083252, "global_step": 5263, "epoch": 59, "lr": 9.992374040034051e-05}
{"train_loss": 0.9217125773429871, "global_step": 5264, "epoch": 59, "lr": 9.992370839018004e-05}
{"train_loss": 0.9161455631256104, "global_step": 5265, "epoch": 59, "lr": 9.992367637330795e-05}
{"train_loss": 0.9348385334014893, "global_step": 5266, "epoch": 59, "lr": 9.992364434972424e-05}
{"train_loss": 0.795154869556427, "global_step": 5267, "epoch": 59, "lr": 9.992361231942888e-05}
{"train_loss": 0.8178099393844604, "global_step": 5268, "epoch": 59, "lr": 9.99235802824219e-05}
{"train_loss": 0.8572114706039429, "global_step": 5269, "epoch": 59, "lr": 9.992354823870331e-05}
{"train_loss": 1.1924595832824707, "global_step": 5270, "epoch": 59, "lr": 9.992351618827311e-05}
{"train_loss": 0.7979462146759033, "global_step": 5271, "epoch": 59, "lr": 9.99234841311313e-05}
{"train_loss": 0.9595159888267517, "global_step": 5272, "epoch": 59, "lr": 9.99234520672779e-05}
{"train_loss": 1.1310839653015137, "global_step": 5273, "epoch": 59, "lr": 9.992341999671289e-05}
{"train_loss": 1.118378758430481, "global_step": 5274, "epoch": 59, "lr": 9.992338791943629e-05}
{"train_loss": 0.8482104539871216, "global_step": 5275, "epoch": 59, "lr": 9.992335583544809e-05}
{"train_loss": 1.020155429840088, "global_step": 5276, "epoch": 59, "lr": 9.99233237447483e-05}
{"train_loss": 1.0581635236740112, "global_step": 5277, "epoch": 59, "lr": 9.992329164733694e-05}
{"train_loss": 0.8990045785903931, "global_step": 5278, "epoch": 59, "lr": 9.992325954321399e-05}
{"train_loss": 1.0026826858520508, "global_step": 5279, "epoch": 59, "lr": 9.992322743237947e-05}
{"train_loss": 0.890438973903656, "global_step": 5280, "epoch": 59, "lr": 9.992319531483338e-05}
{"train_loss": 0.9277499318122864, "global_step": 5281, "epoch": 59, "lr": 9.992316319057573e-05}
{"train_loss": 0.9876388907432556, "global_step": 5282, "epoch": 59, "lr": 9.992313105960651e-05}
{"train_loss": 0.9080013036727905, "global_step": 5283, "epoch": 59, "lr": 9.992309892192574e-05}
{"train_loss": 1.0995289087295532, "global_step": 5284, "epoch": 59, "lr": 9.992306677753343e-05}
{"train_loss": 0.9616784453392029, "global_step": 5285, "epoch": 59, "lr": 9.992303462642954e-05}
{"train_loss": 0.9239723086357117, "global_step": 5286, "epoch": 59, "lr": 9.992300246861414e-05}
{"train_loss": 0.9391236901283264, "global_step": 5287, "epoch": 59, "lr": 9.992297030408718e-05}
{"train_loss": 0.9437897205352783, "global_step": 5288, "epoch": 59, "lr": 9.99229381328487e-05}
{"train_loss": 0.9484415650367737, "global_step": 5289, "epoch": 59, "lr": 9.992290595489868e-05}
{"train_loss": 0.9380249977111816, "global_step": 5290, "epoch": 59, "lr": 9.992287377023712e-05}
{"train_loss": 0.9837081432342529, "global_step": 5291, "epoch": 59, "lr": 9.992284157886407e-05}
{"train_loss": 0.8563231229782104, "global_step": 5292, "epoch": 59, "lr": 9.992280938077946e-05}
{"train_loss": 1.0493546724319458, "global_step": 5293, "epoch": 59, "lr": 9.992277717598337e-05}
{"train_loss": 1.0533634424209595, "global_step": 5294, "epoch": 59, "lr": 9.992274496447578e-05}
{"train_loss": 1.3332496881484985, "global_step": 5295, "epoch": 59, "lr": 9.992271274625666e-05}
{"train_loss": 0.978516697883606, "global_step": 5296, "epoch": 59, "lr": 9.992268052132605e-05}
{"train_loss": 0.8188771605491638, "global_step": 5297, "epoch": 59, "lr": 9.992264828968394e-05}
{"train_loss": 0.7577415704727173, "global_step": 5298, "epoch": 59, "lr": 9.992261605133035e-05}
{"train_loss": 1.0044515132904053, "global_step": 5299, "epoch": 59, "lr": 9.992258380626525e-05}
{"train_loss": 1.0167573690414429, "global_step": 5300, "epoch": 59, "lr": 9.992255155448868e-05}
{"train_loss": 0.902112603187561, "global_step": 5301, "epoch": 59, "lr": 9.992251929600063e-05}
{"train_loss": 0.8981761932373047, "global_step": 5302, "epoch": 59, "lr": 9.992248703080112e-05}
{"train_loss": 0.918352484703064, "global_step": 5303, "epoch": 59, "lr": 9.992245475889013e-05}
{"train_loss": 0.9055800437927246, "global_step": 5304, "epoch": 59, "lr": 9.992242248026766e-05}
{"train_loss": 0.9243117570877075, "global_step": 5305, "epoch": 59, "lr": 9.992239019493375e-05}
{"train_loss": 0.7798999547958374, "global_step": 5306, "epoch": 59, "lr": 9.992235790288838e-05}
{"train_loss": 0.7504459619522095, "global_step": 5307, "epoch": 59, "lr": 9.992232560413154e-05}
{"train_loss": 0.8084732890129089, "global_step": 5308, "epoch": 59, "lr": 9.992229329866327e-05}
{"train_loss": 1.1494792699813843, "global_step": 5309, "epoch": 59, "lr": 9.992226098648355e-05}
{"train_loss": 0.8972359299659729, "global_step": 5310, "epoch": 59, "lr": 9.992222866759237e-05}
{"train_loss": 0.933500349521637, "global_step": 5311, "epoch": 59, "lr": 9.992219634198978e-05}
{"train_loss": 1.0947586297988892, "global_step": 5312, "epoch": 59, "lr": 9.992216400967575e-05}
{"train_loss": 1.0299469232559204, "global_step": 5313, "epoch": 59, "lr": 9.99221316706503e-05}
{"train_loss": 0.9733970761299133, "global_step": 5314, "epoch": 59, "lr": 9.992209932491343e-05}
{"train_loss": 0.9369558691978455, "global_step": 5315, "epoch": 59, "lr": 9.992206697246514e-05}
{"train_loss": 0.8566110134124756, "global_step": 5316, "epoch": 59, "lr": 9.992203461330544e-05}
{"train_loss": 1.1148995161056519, "global_step": 5317, "epoch": 59, "lr": 9.992200224743432e-05}
{"train_loss": 0.9550459384918213, "global_step": 5318, "epoch": 59, "lr": 9.992196987485181e-05}
{"train_loss": 0.7466291785240173, "global_step": 5319, "epoch": 59, "lr": 9.992193749555788e-05}
{"train_loss": 0.93852299451828, "global_step": 5320, "epoch": 59, "lr": 9.992190510955257e-05}
{"train_loss": 1.0784932374954224, "global_step": 5321, "epoch": 59, "lr": 9.992187271683586e-05}
{"train_loss": 1.020185947418213, "global_step": 5322, "epoch": 59, "lr": 9.992184031740776e-05}
{"train_loss": 0.9607868790626526, "global_step": 5323, "epoch": 59, "lr": 9.992180791126829e-05}
{"train_loss": 0.9715234637260437, "global_step": 5324, "epoch": 59, "lr": 9.992177549841743e-05}
{"train_loss": 0.9029183387756348, "global_step": 5325, "epoch": 59, "lr": 9.99217430788552e-05}
{"train_loss": 0.821118175983429, "global_step": 5326, "epoch": 59, "lr": 9.99217106525816e-05}
{"train_loss": 1.2175594568252563, "global_step": 5327, "epoch": 59, "lr": 9.992167821959663e-05}
{"train_loss": 0.9205413460731506, "global_step": 5328, "epoch": 59, "lr": 9.99216457799003e-05}
{"train_loss": 1.0378525257110596, "global_step": 5329, "epoch": 59, "lr": 9.992161333349262e-05}
{"train_loss": 1.043701410293579, "global_step": 5330, "epoch": 59, "lr": 9.992158088037358e-05}
{"train_loss": 0.9745793342590332, "global_step": 5331, "epoch": 59, "lr": 9.99215484205432e-05}
{"train_loss": 0.8360520005226135, "global_step": 5332, "epoch": 59, "lr": 9.992151595400147e-05}
{"train_loss": 0.8794470429420471, "global_step": 5333, "epoch": 59, "lr": 9.99214834807484e-05}
{"train_loss": 1.0647395849227905, "global_step": 5334, "epoch": 59, "lr": 9.992145100078401e-05}
{"train_loss": 0.9228423833847046, "global_step": 5335, "epoch": 59, "lr": 9.992141851410825e-05}
{"train_loss": 0.8600831627845764, "global_step": 5336, "epoch": 59, "lr": 9.992138602072121e-05}
{"train_loss": 0.9369843602180481, "global_step": 5337, "epoch": 59, "lr": 9.992135352062282e-05}
{"train_loss": 0.9099968671798706, "global_step": 5338, "epoch": 59, "lr": 9.992132101381313e-05}
{"train_loss": 0.9563710495327296, "global_step": 5339, "epoch": 59, "lr": 9.99212885002921e-05, "val_loss": 1.1298010349273682}
{"train_loss": 0.9198883771896362, "global_step": 5340, "epoch": 60, "lr": 9.992125598005979e-05}
{"train_loss": 0.9925978183746338, "global_step": 5341, "epoch": 60, "lr": 9.992122345311617e-05}
{"train_loss": 1.0177500247955322, "global_step": 5342, "epoch": 60, "lr": 9.992119091946124e-05}
{"train_loss": 0.912299633026123, "global_step": 5343, "epoch": 60, "lr": 9.992115837909502e-05}
{"train_loss": 0.9999732375144958, "global_step": 5344, "epoch": 60, "lr": 9.99211258320175e-05}
{"train_loss": 0.886235237121582, "global_step": 5345, "epoch": 60, "lr": 9.992109327822871e-05}
{"train_loss": 1.0554167032241821, "global_step": 5346, "epoch": 60, "lr": 9.992106071772862e-05}
{"train_loss": 0.9554275274276733, "global_step": 5347, "epoch": 60, "lr": 9.992102815051726e-05}
{"train_loss": 0.9413899183273315, "global_step": 5348, "epoch": 60, "lr": 9.992099557659463e-05}
{"train_loss": 0.8465031385421753, "global_step": 5349, "epoch": 60, "lr": 9.992096299596072e-05}
{"train_loss": 0.7752246260643005, "global_step": 5350, "epoch": 60, "lr": 9.992093040861555e-05}
{"train_loss": 0.8355070352554321, "global_step": 5351, "epoch": 60, "lr": 9.992089781455914e-05}
{"train_loss": 0.8308242559432983, "global_step": 5352, "epoch": 60, "lr": 9.992086521379145e-05}
{"train_loss": 0.8797454833984375, "global_step": 5353, "epoch": 60, "lr": 9.992083260631251e-05}
{"train_loss": 0.9502130746841431, "global_step": 5354, "epoch": 60, "lr": 9.992079999212232e-05}
{"train_loss": 0.8748713135719299, "global_step": 5355, "epoch": 60, "lr": 9.99207673712209e-05}
{"train_loss": 1.2001296281814575, "global_step": 5356, "epoch": 60, "lr": 9.992073474360822e-05}
{"train_loss": 0.8352527618408203, "global_step": 5357, "epoch": 60, "lr": 9.992070210928432e-05}
{"train_loss": 0.9130308032035828, "global_step": 5358, "epoch": 60, "lr": 9.99206694682492e-05}
{"train_loss": 0.8683953881263733, "global_step": 5359, "epoch": 60, "lr": 9.992063682050283e-05}
{"train_loss": 0.8492010235786438, "global_step": 5360, "epoch": 60, "lr": 9.992060416604526e-05}
{"train_loss": 1.073359489440918, "global_step": 5361, "epoch": 60, "lr": 9.992057150487646e-05}
{"train_loss": 0.9226412773132324, "global_step": 5362, "epoch": 60, "lr": 9.992053883699646e-05}
{"train_loss": 0.8704718351364136, "global_step": 5363, "epoch": 60, "lr": 9.992050616240525e-05}
{"train_loss": 0.854487419128418, "global_step": 5364, "epoch": 60, "lr": 9.992047348110283e-05}
{"train_loss": 0.8875486254692078, "global_step": 5365, "epoch": 60, "lr": 9.99204407930892e-05}
{"train_loss": 0.7706787586212158, "global_step": 5366, "epoch": 60, "lr": 9.992040809836439e-05}
{"train_loss": 0.8857861757278442, "global_step": 5367, "epoch": 60, "lr": 9.992037539692839e-05}
{"train_loss": 0.9386267066001892, "global_step": 5368, "epoch": 60, "lr": 9.99203426887812e-05}
{"train_loss": 0.9462835192680359, "global_step": 5369, "epoch": 60, "lr": 9.992030997392284e-05}
{"train_loss": 0.9105038046836853, "global_step": 5370, "epoch": 60, "lr": 9.992027725235328e-05}
{"train_loss": 0.9282714128494263, "global_step": 5371, "epoch": 60, "lr": 9.992024452407256e-05}
{"train_loss": 0.7962160706520081, "global_step": 5372, "epoch": 60, "lr": 9.992021178908068e-05}
{"train_loss": 1.0349661111831665, "global_step": 5373, "epoch": 60, "lr": 9.992017904737762e-05}
{"train_loss": 0.8448358774185181, "global_step": 5374, "epoch": 60, "lr": 9.99201462989634e-05}
{"train_loss": 0.9204681515693665, "global_step": 5375, "epoch": 60, "lr": 9.992011354383803e-05}
{"train_loss": 0.860519289970398, "global_step": 5376, "epoch": 60, "lr": 9.992008078200152e-05}
{"train_loss": 1.0084850788116455, "global_step": 5377, "epoch": 60, "lr": 9.992004801345385e-05}
{"train_loss": 0.9361751079559326, "global_step": 5378, "epoch": 60, "lr": 9.992001523819506e-05}
{"train_loss": 1.0685515403747559, "global_step": 5379, "epoch": 60, "lr": 9.99199824562251e-05}
{"train_loss": 0.9643735885620117, "global_step": 5380, "epoch": 60, "lr": 9.991994966754403e-05}
{"train_loss": 1.1196035146713257, "global_step": 5381, "epoch": 60, "lr": 9.991991687215181e-05}
{"train_loss": 0.9827725887298584, "global_step": 5382, "epoch": 60, "lr": 9.991988407004848e-05}
{"train_loss": 0.7977972030639648, "global_step": 5383, "epoch": 60, "lr": 9.991985126123405e-05}
{"train_loss": 0.9399306178092957, "global_step": 5384, "epoch": 60, "lr": 9.991981844570847e-05}
{"train_loss": 0.7959873080253601, "global_step": 5385, "epoch": 60, "lr": 9.99197856234718e-05}
{"train_loss": 0.9429464340209961, "global_step": 5386, "epoch": 60, "lr": 9.991975279452401e-05}
{"train_loss": 0.8989397287368774, "global_step": 5387, "epoch": 60, "lr": 9.991971995886513e-05}
{"train_loss": 0.91229248046875, "global_step": 5388, "epoch": 60, "lr": 9.991968711649513e-05}
{"train_loss": 1.0917434692382812, "global_step": 5389, "epoch": 60, "lr": 9.991965426741405e-05}
{"train_loss": 0.8600994348526001, "global_step": 5390, "epoch": 60, "lr": 9.991962141162188e-05}
{"train_loss": 0.8736233115196228, "global_step": 5391, "epoch": 60, "lr": 9.991958854911864e-05}
{"train_loss": 1.0058479309082031, "global_step": 5392, "epoch": 60, "lr": 9.99195556799043e-05}
{"train_loss": 0.8109712600708008, "global_step": 5393, "epoch": 60, "lr": 9.991952280397889e-05}
{"train_loss": 0.8856815099716187, "global_step": 5394, "epoch": 60, "lr": 9.991948992134242e-05}
{"train_loss": 0.9649862051010132, "global_step": 5395, "epoch": 60, "lr": 9.991945703199488e-05}
{"train_loss": 0.8659586906433105, "global_step": 5396, "epoch": 60, "lr": 9.991942413593627e-05}
{"train_loss": 0.7475911974906921, "global_step": 5397, "epoch": 60, "lr": 9.991939123316661e-05}
{"train_loss": 0.7876631021499634, "global_step": 5398, "epoch": 60, "lr": 9.991935832368587e-05}
{"train_loss": 0.9285609126091003, "global_step": 5399, "epoch": 60, "lr": 9.99193254074941e-05}
{"train_loss": 0.9418671131134033, "global_step": 5400, "epoch": 60, "lr": 9.991929248459129e-05}
{"train_loss": 0.7889868021011353, "global_step": 5401, "epoch": 60, "lr": 9.991925955497744e-05}
{"train_loss": 1.0506850481033325, "global_step": 5402, "epoch": 60, "lr": 9.991922661865254e-05}
{"train_loss": 0.8995295763015747, "global_step": 5403, "epoch": 60, "lr": 9.991919367561662e-05}
{"train_loss": 1.02272629737854, "global_step": 5404, "epoch": 60, "lr": 9.991916072586967e-05}
{"train_loss": 0.7741696834564209, "global_step": 5405, "epoch": 60, "lr": 9.99191277694117e-05}
{"train_loss": 1.064058780670166, "global_step": 5406, "epoch": 60, "lr": 9.991909480624271e-05}
{"train_loss": 0.9414371848106384, "global_step": 5407, "epoch": 60, "lr": 9.991906183636271e-05}
{"train_loss": 0.9479284882545471, "global_step": 5408, "epoch": 60, "lr": 9.991902885977169e-05}
{"train_loss": 1.006591796875, "global_step": 5409, "epoch": 60, "lr": 9.991899587646967e-05}
{"train_loss": 0.9804331660270691, "global_step": 5410, "epoch": 60, "lr": 9.991896288645665e-05}
{"train_loss": 0.8174203634262085, "global_step": 5411, "epoch": 60, "lr": 9.991892988973263e-05}
{"train_loss": 0.9900323748588562, "global_step": 5412, "epoch": 60, "lr": 9.991889688629763e-05}
{"train_loss": 0.8187214732170105, "global_step": 5413, "epoch": 60, "lr": 9.991886387615163e-05}
{"train_loss": 0.8538878560066223, "global_step": 5414, "epoch": 60, "lr": 9.991883085929466e-05}
{"train_loss": 0.8105813264846802, "global_step": 5415, "epoch": 60, "lr": 9.991879783572669e-05}
{"train_loss": 0.9472454190254211, "global_step": 5416, "epoch": 60, "lr": 9.991876480544776e-05}
{"train_loss": 1.0956008434295654, "global_step": 5417, "epoch": 60, "lr": 9.991873176845786e-05}
{"train_loss": 0.9237757325172424, "global_step": 5418, "epoch": 60, "lr": 9.991869872475698e-05}
{"train_loss": 1.0077654123306274, "global_step": 5419, "epoch": 60, "lr": 9.991866567434517e-05}
{"train_loss": 0.961426854133606, "global_step": 5420, "epoch": 60, "lr": 9.991863261722239e-05}
{"train_loss": 0.7491742968559265, "global_step": 5421, "epoch": 60, "lr": 9.991859955338865e-05}
{"train_loss": 0.839516818523407, "global_step": 5422, "epoch": 60, "lr": 9.991856648284397e-05}
{"train_loss": 0.9511027336120605, "global_step": 5423, "epoch": 60, "lr": 9.991853340558835e-05}
{"train_loss": 0.9059675335884094, "global_step": 5424, "epoch": 60, "lr": 9.99185003216218e-05}
{"train_loss": 0.8333523869514465, "global_step": 5425, "epoch": 60, "lr": 9.991846723094428e-05}
{"train_loss": 0.8192685842514038, "global_step": 5426, "epoch": 60, "lr": 9.991843413355586e-05}
{"train_loss": 0.9052069783210754, "global_step": 5427, "epoch": 60, "lr": 9.991840102945651e-05}
{"train_loss": 0.9184303283691406, "global_step": 5428, "epoch": 60, "lr": 9.991836791864623e-05, "val_loss": 1.1266767978668213, "train_action_mse_error": 52.470787048339844}
{"train_loss": 1.0335707664489746, "global_step": 5429, "epoch": 61, "lr": 9.991833480112506e-05}
{"train_loss": 0.868887186050415, "global_step": 5430, "epoch": 61, "lr": 9.991830167689295e-05}
{"train_loss": 0.9061545133590698, "global_step": 5431, "epoch": 61, "lr": 9.991826854594995e-05}
{"train_loss": 1.0862940549850464, "global_step": 5432, "epoch": 61, "lr": 9.991823540829603e-05}
{"train_loss": 1.0739012956619263, "global_step": 5433, "epoch": 61, "lr": 9.991820226393123e-05}
{"train_loss": 0.8947715163230896, "global_step": 5434, "epoch": 61, "lr": 9.991816911285553e-05}
{"train_loss": 0.9620575308799744, "global_step": 5435, "epoch": 61, "lr": 9.991813595506894e-05}
{"train_loss": 0.941693902015686, "global_step": 5436, "epoch": 61, "lr": 9.991810279057147e-05}
{"train_loss": 1.0149096250534058, "global_step": 5437, "epoch": 61, "lr": 9.99180696193631e-05}
{"train_loss": 0.9344783425331116, "global_step": 5438, "epoch": 61, "lr": 9.991803644144388e-05}
{"train_loss": 0.9214489459991455, "global_step": 5439, "epoch": 61, "lr": 9.991800325681378e-05}
{"train_loss": 0.934303343296051, "global_step": 5440, "epoch": 61, "lr": 9.991797006547281e-05}
{"train_loss": 0.9397234320640564, "global_step": 5441, "epoch": 61, "lr": 9.991793686742098e-05}
{"train_loss": 0.9159035086631775, "global_step": 5442, "epoch": 61, "lr": 9.991790366265828e-05}
{"train_loss": 0.9575902819633484, "global_step": 5443, "epoch": 61, "lr": 9.991787045118472e-05}
{"train_loss": 1.1170802116394043, "global_step": 5444, "epoch": 61, "lr": 9.991783723300035e-05}
{"train_loss": 0.8534413576126099, "global_step": 5445, "epoch": 61, "lr": 9.99178040081051e-05}
{"train_loss": 1.1542026996612549, "global_step": 5446, "epoch": 61, "lr": 9.991777077649902e-05}
{"train_loss": 1.1542587280273438, "global_step": 5447, "epoch": 61, "lr": 9.99177375381821e-05}
{"train_loss": 0.7710779309272766, "global_step": 5448, "epoch": 61, "lr": 9.991770429315435e-05}
{"train_loss": 0.9453026652336121, "global_step": 5449, "epoch": 61, "lr": 9.991767104141578e-05}
{"train_loss": 1.086334466934204, "global_step": 5450, "epoch": 61, "lr": 9.991763778296639e-05}
{"train_loss": 0.9337490200996399, "global_step": 5451, "epoch": 61, "lr": 9.991760451780616e-05}
{"train_loss": 0.9835394024848938, "global_step": 5452, "epoch": 61, "lr": 9.991757124593514e-05}
{"train_loss": 1.0286774635314941, "global_step": 5453, "epoch": 61, "lr": 9.991753796735331e-05}
{"train_loss": 0.8405270576477051, "global_step": 5454, "epoch": 61, "lr": 9.991750468206067e-05}
{"train_loss": 1.0339423418045044, "global_step": 5455, "epoch": 61, "lr": 9.991747139005722e-05}
{"train_loss": 0.9423249363899231, "global_step": 5456, "epoch": 61, "lr": 9.991743809134299e-05}
{"train_loss": 0.8472382426261902, "global_step": 5457, "epoch": 61, "lr": 9.991740478591796e-05}
{"train_loss": 0.8732763528823853, "global_step": 5458, "epoch": 61, "lr": 9.991737147378214e-05}
{"train_loss": 0.8126412034034729, "global_step": 5459, "epoch": 61, "lr": 9.991733815493553e-05}
{"train_loss": 0.9944522976875305, "global_step": 5460, "epoch": 61, "lr": 9.991730482937817e-05}
{"train_loss": 0.7314030528068542, "global_step": 5461, "epoch": 61, "lr": 9.991727149711002e-05}
{"train_loss": 0.8189442753791809, "global_step": 5462, "epoch": 61, "lr": 9.99172381581311e-05}
{"train_loss": 0.9403370022773743, "global_step": 5463, "epoch": 61, "lr": 9.991720481244142e-05}
{"train_loss": 0.8332380056381226, "global_step": 5464, "epoch": 61, "lr": 9.991717146004097e-05}
{"train_loss": 0.8610755205154419, "global_step": 5465, "epoch": 61, "lr": 9.991713810092978e-05}
{"train_loss": 0.7026180624961853, "global_step": 5466, "epoch": 61, "lr": 9.991710473510783e-05}
{"train_loss": 0.8285611867904663, "global_step": 5467, "epoch": 61, "lr": 9.991707136257514e-05}
{"train_loss": 0.9047630429267883, "global_step": 5468, "epoch": 61, "lr": 9.99170379833317e-05}
{"train_loss": 0.955043375492096, "global_step": 5469, "epoch": 61, "lr": 9.991700459737754e-05}
{"train_loss": 0.8034711480140686, "global_step": 5470, "epoch": 61, "lr": 9.991697120471262e-05}
{"train_loss": 0.8834334015846252, "global_step": 5471, "epoch": 61, "lr": 9.991693780533699e-05}
{"train_loss": 0.8557295203208923, "global_step": 5472, "epoch": 61, "lr": 9.991690439925064e-05}
{"train_loss": 0.8591289520263672, "global_step": 5473, "epoch": 61, "lr": 9.991687098645356e-05}
{"train_loss": 0.6702103614807129, "global_step": 5474, "epoch": 61, "lr": 9.991683756694577e-05}
{"train_loss": 0.9360312819480896, "global_step": 5475, "epoch": 61, "lr": 9.991680414072726e-05}
{"train_loss": 0.9134267568588257, "global_step": 5476, "epoch": 61, "lr": 9.991677070779806e-05}
{"train_loss": 1.1123586893081665, "global_step": 5477, "epoch": 61, "lr": 9.991673726815814e-05}
{"train_loss": 0.9713447690010071, "global_step": 5478, "epoch": 61, "lr": 9.991670382180754e-05}
{"train_loss": 0.9649066925048828, "global_step": 5479, "epoch": 61, "lr": 9.991667036874623e-05}
{"train_loss": 0.8984798192977905, "global_step": 5480, "epoch": 61, "lr": 9.991663690897426e-05}
{"train_loss": 0.8322321772575378, "global_step": 5481, "epoch": 61, "lr": 9.991660344249158e-05}
{"train_loss": 0.786022961139679, "global_step": 5482, "epoch": 61, "lr": 9.991656996929824e-05}
{"train_loss": 0.8078717589378357, "global_step": 5483, "epoch": 61, "lr": 9.99165364893942e-05}
{"train_loss": 0.8492798209190369, "global_step": 5484, "epoch": 61, "lr": 9.991650300277952e-05}
{"train_loss": 0.6361051201820374, "global_step": 5485, "epoch": 61, "lr": 9.991646950945415e-05}
{"train_loss": 0.9377132058143616, "global_step": 5486, "epoch": 61, "lr": 9.991643600941813e-05}
{"train_loss": 0.9958285093307495, "global_step": 5487, "epoch": 61, "lr": 9.991640250267146e-05}
{"train_loss": 0.6856544017791748, "global_step": 5488, "epoch": 61, "lr": 9.991636898921413e-05}
{"train_loss": 0.8858183026313782, "global_step": 5489, "epoch": 61, "lr": 9.991633546904615e-05}
{"train_loss": 0.966167151927948, "global_step": 5490, "epoch": 61, "lr": 9.991630194216753e-05}
{"train_loss": 1.1155734062194824, "global_step": 5491, "epoch": 61, "lr": 9.991626840857828e-05}
{"train_loss": 1.0535199642181396, "global_step": 5492, "epoch": 61, "lr": 9.991623486827839e-05}
{"train_loss": 0.8978985548019409, "global_step": 5493, "epoch": 61, "lr": 9.991620132126787e-05}
{"train_loss": 0.877164363861084, "global_step": 5494, "epoch": 61, "lr": 9.991616776754672e-05}
{"train_loss": 1.0425564050674438, "global_step": 5495, "epoch": 61, "lr": 9.991613420711495e-05}
{"train_loss": 0.7895277142524719, "global_step": 5496, "epoch": 61, "lr": 9.991610063997256e-05}
{"train_loss": 0.9438009262084961, "global_step": 5497, "epoch": 61, "lr": 9.991606706611958e-05}
{"train_loss": 0.7015343308448792, "global_step": 5498, "epoch": 61, "lr": 9.991603348555599e-05}
{"train_loss": 1.0355602502822876, "global_step": 5499, "epoch": 61, "lr": 9.991599989828178e-05}
{"train_loss": 1.000171184539795, "global_step": 5500, "epoch": 61, "lr": 9.991596630429698e-05}
{"train_loss": 1.0024099349975586, "global_step": 5501, "epoch": 61, "lr": 9.991593270360158e-05}
{"train_loss": 0.858273983001709, "global_step": 5502, "epoch": 61, "lr": 9.991589909619561e-05}
{"train_loss": 0.7937560081481934, "global_step": 5503, "epoch": 61, "lr": 9.991586548207902e-05}
{"train_loss": 0.772648274898529, "global_step": 5504, "epoch": 61, "lr": 9.991583186125189e-05}
{"train_loss": 0.9423937797546387, "global_step": 5505, "epoch": 61, "lr": 9.991579823371418e-05}
{"train_loss": 1.0025231838226318, "global_step": 5506, "epoch": 61, "lr": 9.991576459946588e-05}
{"train_loss": 0.9895521998405457, "global_step": 5507, "epoch": 61, "lr": 9.991573095850701e-05}
{"train_loss": 0.8596980571746826, "global_step": 5508, "epoch": 61, "lr": 9.991569731083759e-05}
{"train_loss": 1.0035895109176636, "global_step": 5509, "epoch": 61, "lr": 9.991566365645761e-05}
{"train_loss": 1.2404130697250366, "global_step": 5510, "epoch": 61, "lr": 9.99156299953671e-05}
{"train_loss": 0.9364936351776123, "global_step": 5511, "epoch": 61, "lr": 9.991559632756602e-05}
{"train_loss": 0.9012889266014099, "global_step": 5512, "epoch": 61, "lr": 9.991556265305439e-05}
{"train_loss": 0.9045734405517578, "global_step": 5513, "epoch": 61, "lr": 9.991552897183223e-05}
{"train_loss": 0.9984771013259888, "global_step": 5514, "epoch": 61, "lr": 9.991549528389954e-05}
{"train_loss": 1.1109594106674194, "global_step": 5515, "epoch": 61, "lr": 9.991546158925631e-05}
{"train_loss": 0.9398471117019653, "global_step": 5516, "epoch": 61, "lr": 9.991542788790255e-05}
{"train_loss": 0.928175626845842, "global_step": 5517, "epoch": 61, "lr": 9.99153941798383e-05, "val_loss": 1.143863558769226}
{"train_loss": 0.7962874174118042, "global_step": 5518, "epoch": 62, "lr": 9.991536046506351e-05}
{"train_loss": 1.095901608467102, "global_step": 5519, "epoch": 62, "lr": 9.99153267435782e-05}
{"train_loss": 1.0202040672302246, "global_step": 5520, "epoch": 62, "lr": 9.99152930153824e-05}
{"train_loss": 0.9413306713104248, "global_step": 5521, "epoch": 62, "lr": 9.99152592804761e-05}
{"train_loss": 0.9672098159790039, "global_step": 5522, "epoch": 62, "lr": 9.991522553885928e-05}
{"train_loss": 0.9365646839141846, "global_step": 5523, "epoch": 62, "lr": 9.991519179053198e-05}
{"train_loss": 0.90590900182724, "global_step": 5524, "epoch": 62, "lr": 9.991515803549419e-05}
{"train_loss": 0.9357409477233887, "global_step": 5525, "epoch": 62, "lr": 9.991512427374592e-05}
{"train_loss": 1.064680814743042, "global_step": 5526, "epoch": 62, "lr": 9.991509050528715e-05}
{"train_loss": 1.130272388458252, "global_step": 5527, "epoch": 62, "lr": 9.991505673011794e-05}
{"train_loss": 0.8429902791976929, "global_step": 5528, "epoch": 62, "lr": 9.991502294823822e-05}
{"train_loss": 0.9235112071037292, "global_step": 5529, "epoch": 62, "lr": 9.991498915964806e-05}
{"train_loss": 1.0623811483383179, "global_step": 5530, "epoch": 62, "lr": 9.991495536434743e-05}
{"train_loss": 0.9042566418647766, "global_step": 5531, "epoch": 62, "lr": 9.991492156233635e-05}
{"train_loss": 1.0032047033309937, "global_step": 5532, "epoch": 62, "lr": 9.991488775361481e-05}
{"train_loss": 0.9376116394996643, "global_step": 5533, "epoch": 62, "lr": 9.991485393818281e-05}
{"train_loss": 1.0470441579818726, "global_step": 5534, "epoch": 62, "lr": 9.991482011604039e-05}
{"train_loss": 0.9944790601730347, "global_step": 5535, "epoch": 62, "lr": 9.99147862871875e-05}
{"train_loss": 0.9862580895423889, "global_step": 5536, "epoch": 62, "lr": 9.991475245162421e-05}
{"train_loss": 1.0594303607940674, "global_step": 5537, "epoch": 62, "lr": 9.991471860935047e-05}
{"train_loss": 0.8448453545570374, "global_step": 5538, "epoch": 62, "lr": 9.991468476036632e-05}
{"train_loss": 0.9079052209854126, "global_step": 5539, "epoch": 62, "lr": 9.991465090467173e-05}
{"train_loss": 0.9201442003250122, "global_step": 5540, "epoch": 62, "lr": 9.991461704226672e-05}
{"train_loss": 0.9328094124794006, "global_step": 5541, "epoch": 62, "lr": 9.991458317315132e-05}
{"train_loss": 0.8721755743026733, "global_step": 5542, "epoch": 62, "lr": 9.991454929732551e-05}
{"train_loss": 0.9715529680252075, "global_step": 5543, "epoch": 62, "lr": 9.991451541478929e-05}
{"train_loss": 1.0113904476165771, "global_step": 5544, "epoch": 62, "lr": 9.991448152554266e-05}
{"train_loss": 1.0665568113327026, "global_step": 5545, "epoch": 62, "lr": 9.991444762958565e-05}
{"train_loss": 0.9832808375358582, "global_step": 5546, "epoch": 62, "lr": 9.991441372691825e-05}
{"train_loss": 0.8440412878990173, "global_step": 5547, "epoch": 62, "lr": 9.991437981754046e-05}
{"train_loss": 0.9368341565132141, "global_step": 5548, "epoch": 62, "lr": 9.99143459014523e-05}
{"train_loss": 1.0042392015457153, "global_step": 5549, "epoch": 62, "lr": 9.991431197865375e-05}
{"train_loss": 1.0026969909667969, "global_step": 5550, "epoch": 62, "lr": 9.991427804914484e-05}
{"train_loss": 0.9369581937789917, "global_step": 5551, "epoch": 62, "lr": 9.991424411292556e-05}
{"train_loss": 0.8939768671989441, "global_step": 5552, "epoch": 62, "lr": 9.991421016999591e-05}
{"train_loss": 0.9669549465179443, "global_step": 5553, "epoch": 62, "lr": 9.991417622035593e-05}
{"train_loss": 0.9559095501899719, "global_step": 5554, "epoch": 62, "lr": 9.991414226400557e-05}
{"train_loss": 0.7824354767799377, "global_step": 5555, "epoch": 62, "lr": 9.991410830094488e-05}
{"train_loss": 0.9236605167388916, "global_step": 5556, "epoch": 62, "lr": 9.991407433117383e-05}
{"train_loss": 0.9097274541854858, "global_step": 5557, "epoch": 62, "lr": 9.991404035469245e-05}
{"train_loss": 0.878229558467865, "global_step": 5558, "epoch": 62, "lr": 9.991400637150073e-05}
{"train_loss": 0.9667889475822449, "global_step": 5559, "epoch": 62, "lr": 9.991397238159869e-05}
{"train_loss": 0.921843409538269, "global_step": 5560, "epoch": 62, "lr": 9.991393838498631e-05}
{"train_loss": 1.1264221668243408, "global_step": 5561, "epoch": 62, "lr": 9.991390438166362e-05}
{"train_loss": 0.8255112767219543, "global_step": 5562, "epoch": 62, "lr": 9.991387037163062e-05}
{"train_loss": 1.0613893270492554, "global_step": 5563, "epoch": 62, "lr": 9.99138363548873e-05}
{"train_loss": 1.0231602191925049, "global_step": 5564, "epoch": 62, "lr": 9.991380233143368e-05}
{"train_loss": 0.866561770439148, "global_step": 5565, "epoch": 62, "lr": 9.991376830126975e-05}
{"train_loss": 0.981059193611145, "global_step": 5566, "epoch": 62, "lr": 9.991373426439552e-05}
{"train_loss": 0.9297716021537781, "global_step": 5567, "epoch": 62, "lr": 9.9913700220811e-05}
{"train_loss": 1.0288560390472412, "global_step": 5568, "epoch": 62, "lr": 9.99136661705162e-05}
{"train_loss": 1.0656390190124512, "global_step": 5569, "epoch": 62, "lr": 9.99136321135111e-05}
{"train_loss": 0.8780877590179443, "global_step": 5570, "epoch": 62, "lr": 9.991359804979572e-05}
{"train_loss": 0.8790019750595093, "global_step": 5571, "epoch": 62, "lr": 9.991356397937009e-05}
{"train_loss": 0.9127714037895203, "global_step": 5572, "epoch": 62, "lr": 9.991352990223417e-05}
{"train_loss": 0.9999886751174927, "global_step": 5573, "epoch": 62, "lr": 9.991349581838798e-05}
{"train_loss": 0.9269627928733826, "global_step": 5574, "epoch": 62, "lr": 9.991346172783155e-05}
{"train_loss": 0.85386061668396, "global_step": 5575, "epoch": 62, "lr": 9.991342763056483e-05}
{"train_loss": 0.9380218982696533, "global_step": 5576, "epoch": 62, "lr": 9.99133935265879e-05}
{"train_loss": 1.1214780807495117, "global_step": 5577, "epoch": 62, "lr": 9.99133594159007e-05}
{"train_loss": 0.8548387289047241, "global_step": 5578, "epoch": 62, "lr": 9.991332529850325e-05}
{"train_loss": 0.929573655128479, "global_step": 5579, "epoch": 62, "lr": 9.991329117439558e-05}
{"train_loss": 0.9226369857788086, "global_step": 5580, "epoch": 62, "lr": 9.991325704357766e-05}
{"train_loss": 0.965566873550415, "global_step": 5581, "epoch": 62, "lr": 9.991322290604953e-05}
{"train_loss": 0.9115360975265503, "global_step": 5582, "epoch": 62, "lr": 9.991318876181114e-05}
{"train_loss": 0.8854731917381287, "global_step": 5583, "epoch": 62, "lr": 9.991315461086256e-05}
{"train_loss": 0.9908081293106079, "global_step": 5584, "epoch": 62, "lr": 9.991312045320376e-05}
{"train_loss": 0.9653436541557312, "global_step": 5585, "epoch": 62, "lr": 9.991308628883475e-05}
{"train_loss": 0.9667289853096008, "global_step": 5586, "epoch": 62, "lr": 9.991305211775552e-05}
{"train_loss": 0.90239018201828, "global_step": 5587, "epoch": 62, "lr": 9.99130179399661e-05}
{"train_loss": 1.0412558317184448, "global_step": 5588, "epoch": 62, "lr": 9.99129837554665e-05}
{"train_loss": 0.9011960029602051, "global_step": 5589, "epoch": 62, "lr": 9.991294956425668e-05}
{"train_loss": 0.8143976330757141, "global_step": 5590, "epoch": 62, "lr": 9.991291536633669e-05}
{"train_loss": 0.7278074026107788, "global_step": 5591, "epoch": 62, "lr": 9.99128811617065e-05}
{"train_loss": 1.0228092670440674, "global_step": 5592, "epoch": 62, "lr": 9.991284695036614e-05}
{"train_loss": 0.8987935781478882, "global_step": 5593, "epoch": 62, "lr": 9.991281273231561e-05}
{"train_loss": 1.0314576625823975, "global_step": 5594, "epoch": 62, "lr": 9.99127785075549e-05}
{"train_loss": 0.8547204732894897, "global_step": 5595, "epoch": 62, "lr": 9.991274427608403e-05}
{"train_loss": 0.8690955638885498, "global_step": 5596, "epoch": 62, "lr": 9.991271003790301e-05}
{"train_loss": 0.8763302564620972, "global_step": 5597, "epoch": 62, "lr": 9.991267579301182e-05}
{"train_loss": 0.7180402278900146, "global_step": 5598, "epoch": 62, "lr": 9.991264154141049e-05}
{"train_loss": 0.8958487510681152, "global_step": 5599, "epoch": 62, "lr": 9.991260728309901e-05}
{"train_loss": 0.9349756240844727, "global_step": 5600, "epoch": 62, "lr": 9.991257301807739e-05}
{"train_loss": 1.014369249343872, "global_step": 5601, "epoch": 62, "lr": 9.991253874634563e-05}
{"train_loss": 0.7504782676696777, "global_step": 5602, "epoch": 62, "lr": 9.991250446790374e-05}
{"train_loss": 0.7727108001708984, "global_step": 5603, "epoch": 62, "lr": 9.991247018275172e-05}
{"train_loss": 1.169071078300476, "global_step": 5604, "epoch": 62, "lr": 9.991243589088957e-05}
{"train_loss": 0.8997420072555542, "global_step": 5605, "epoch": 62, "lr": 9.991240159231732e-05}
{"train_loss": 0.9414713114834903, "global_step": 5606, "epoch": 62, "lr": 9.991236728703495e-05, "val_loss": 1.1694847345352173}
{"train_loss": 0.8840000033378601, "global_step": 5607, "epoch": 63, "lr": 9.991233297504245e-05}
{"train_loss": 0.9411228895187378, "global_step": 5608, "epoch": 63, "lr": 9.991229865633987e-05}
{"train_loss": 0.8292482495307922, "global_step": 5609, "epoch": 63, "lr": 9.991226433092717e-05}
{"train_loss": 0.9625231027603149, "global_step": 5610, "epoch": 63, "lr": 9.991222999880439e-05}
{"train_loss": 0.8603245615959167, "global_step": 5611, "epoch": 63, "lr": 9.991219565997152e-05}
{"train_loss": 1.1136584281921387, "global_step": 5612, "epoch": 63, "lr": 9.991216131442856e-05}
{"train_loss": 1.0510097742080688, "global_step": 5613, "epoch": 63, "lr": 9.991212696217552e-05}
{"train_loss": 0.8011013865470886, "global_step": 5614, "epoch": 63, "lr": 9.99120926032124e-05}
{"train_loss": 0.7818597555160522, "global_step": 5615, "epoch": 63, "lr": 9.99120582375392e-05}
{"train_loss": 0.8542069792747498, "global_step": 5616, "epoch": 63, "lr": 9.991202386515592e-05}
{"train_loss": 1.163375973701477, "global_step": 5617, "epoch": 63, "lr": 9.99119894860626e-05}
{"train_loss": 0.9447181820869446, "global_step": 5618, "epoch": 63, "lr": 9.991195510025922e-05}
{"train_loss": 0.9463710784912109, "global_step": 5619, "epoch": 63, "lr": 9.991192070774577e-05}
{"train_loss": 0.9811927676200867, "global_step": 5620, "epoch": 63, "lr": 9.99118863085223e-05}
{"train_loss": 0.7982914447784424, "global_step": 5621, "epoch": 63, "lr": 9.991185190258875e-05}
{"train_loss": 0.8776370882987976, "global_step": 5622, "epoch": 63, "lr": 9.991181748994519e-05}
{"train_loss": 0.8502070307731628, "global_step": 5623, "epoch": 63, "lr": 9.991178307059158e-05}
{"train_loss": 0.9172927141189575, "global_step": 5624, "epoch": 63, "lr": 9.991174864452795e-05}
{"train_loss": 0.9414440989494324, "global_step": 5625, "epoch": 63, "lr": 9.991171421175427e-05}
{"train_loss": 0.9265923500061035, "global_step": 5626, "epoch": 63, "lr": 9.991167977227058e-05}
{"train_loss": 1.0550693273544312, "global_step": 5627, "epoch": 63, "lr": 9.991164532607689e-05}
{"train_loss": 0.9787680506706238, "global_step": 5628, "epoch": 63, "lr": 9.991161087317317e-05}
{"train_loss": 0.9901410341262817, "global_step": 5629, "epoch": 63, "lr": 9.991157641355945e-05}
{"train_loss": 0.962670624256134, "global_step": 5630, "epoch": 63, "lr": 9.991154194723572e-05}
{"train_loss": 0.8008667826652527, "global_step": 5631, "epoch": 63, "lr": 9.9911507474202e-05}
{"train_loss": 0.8130255341529846, "global_step": 5632, "epoch": 63, "lr": 9.991147299445828e-05}
{"train_loss": 0.9949965476989746, "global_step": 5633, "epoch": 63, "lr": 9.991143850800458e-05}
{"train_loss": 0.8048624992370605, "global_step": 5634, "epoch": 63, "lr": 9.991140401484087e-05}
{"train_loss": 0.8133090734481812, "global_step": 5635, "epoch": 63, "lr": 9.991136951496721e-05}
{"train_loss": 0.819668173789978, "global_step": 5636, "epoch": 63, "lr": 9.991133500838356e-05}
{"train_loss": 1.0520718097686768, "global_step": 5637, "epoch": 63, "lr": 9.991130049508994e-05}
{"train_loss": 1.0903785228729248, "global_step": 5638, "epoch": 63, "lr": 9.991126597508636e-05}
{"train_loss": 0.8103678822517395, "global_step": 5639, "epoch": 63, "lr": 9.99112314483728e-05}
{"train_loss": 0.8745905756950378, "global_step": 5640, "epoch": 63, "lr": 9.99111969149493e-05}
{"train_loss": 0.8800439238548279, "global_step": 5641, "epoch": 63, "lr": 9.991116237481585e-05}
{"train_loss": 0.872191309928894, "global_step": 5642, "epoch": 63, "lr": 9.991112782797245e-05}
{"train_loss": 1.0544166564941406, "global_step": 5643, "epoch": 63, "lr": 9.99110932744191e-05}
{"train_loss": 0.9374493956565857, "global_step": 5644, "epoch": 63, "lr": 9.991105871415581e-05}
{"train_loss": 0.8616104125976562, "global_step": 5645, "epoch": 63, "lr": 9.99110241471826e-05}
{"train_loss": 0.994739830493927, "global_step": 5646, "epoch": 63, "lr": 9.991098957349945e-05}
{"train_loss": 1.0567680597305298, "global_step": 5647, "epoch": 63, "lr": 9.991095499310639e-05}
{"train_loss": 1.0069868564605713, "global_step": 5648, "epoch": 63, "lr": 9.991092040600339e-05}
{"train_loss": 0.893937349319458, "global_step": 5649, "epoch": 63, "lr": 9.99108858121905e-05}
{"train_loss": 0.8155187964439392, "global_step": 5650, "epoch": 63, "lr": 9.991085121166768e-05}
{"train_loss": 0.8871200084686279, "global_step": 5651, "epoch": 63, "lr": 9.991081660443495e-05}
{"train_loss": 0.7826009392738342, "global_step": 5652, "epoch": 63, "lr": 9.991078199049234e-05}
{"train_loss": 0.6773930191993713, "global_step": 5653, "epoch": 63, "lr": 9.991074736983981e-05}
{"train_loss": 1.0734678506851196, "global_step": 5654, "epoch": 63, "lr": 9.991071274247741e-05}
{"train_loss": 1.0057222843170166, "global_step": 5655, "epoch": 63, "lr": 9.99106781084051e-05}
{"train_loss": 1.081792950630188, "global_step": 5656, "epoch": 63, "lr": 9.991064346762293e-05}
{"train_loss": 0.8818538188934326, "global_step": 5657, "epoch": 63, "lr": 9.991060882013088e-05}
{"train_loss": 0.9295097589492798, "global_step": 5658, "epoch": 63, "lr": 9.991057416592895e-05}
{"train_loss": 0.8548481464385986, "global_step": 5659, "epoch": 63, "lr": 9.991053950501714e-05}
{"train_loss": 0.8672760128974915, "global_step": 5660, "epoch": 63, "lr": 9.991050483739548e-05}
{"train_loss": 0.7662531733512878, "global_step": 5661, "epoch": 63, "lr": 9.991047016306397e-05}
{"train_loss": 0.7942222952842712, "global_step": 5662, "epoch": 63, "lr": 9.991043548202259e-05}
{"train_loss": 0.8309588432312012, "global_step": 5663, "epoch": 63, "lr": 9.991040079427136e-05}
{"train_loss": 0.8499025702476501, "global_step": 5664, "epoch": 63, "lr": 9.99103660998103e-05}
{"train_loss": 0.9869378805160522, "global_step": 5665, "epoch": 63, "lr": 9.991033139863939e-05}
{"train_loss": 1.024165391921997, "global_step": 5666, "epoch": 63, "lr": 9.991029669075864e-05}
{"train_loss": 0.7814705967903137, "global_step": 5667, "epoch": 63, "lr": 9.991026197616806e-05}
{"train_loss": 1.0556164979934692, "global_step": 5668, "epoch": 63, "lr": 9.991022725486766e-05}
{"train_loss": 0.9759507179260254, "global_step": 5669, "epoch": 63, "lr": 9.991019252685742e-05}
{"train_loss": 0.8293493986129761, "global_step": 5670, "epoch": 63, "lr": 9.99101577921374e-05}
{"train_loss": 1.0704163312911987, "global_step": 5671, "epoch": 63, "lr": 9.991012305070753e-05}
{"train_loss": 0.633924126625061, "global_step": 5672, "epoch": 63, "lr": 9.991008830256788e-05}
{"train_loss": 1.054884672164917, "global_step": 5673, "epoch": 63, "lr": 9.991005354771841e-05}
{"train_loss": 0.9124670028686523, "global_step": 5674, "epoch": 63, "lr": 9.991001878615915e-05}
{"train_loss": 0.8303136825561523, "global_step": 5675, "epoch": 63, "lr": 9.990998401789009e-05}
{"train_loss": 0.8218688368797302, "global_step": 5676, "epoch": 63, "lr": 9.990994924291124e-05}
{"train_loss": 0.8215303421020508, "global_step": 5677, "epoch": 63, "lr": 9.990991446122261e-05}
{"train_loss": 0.7963318228721619, "global_step": 5678, "epoch": 63, "lr": 9.99098796728242e-05}
{"train_loss": 1.152191162109375, "global_step": 5679, "epoch": 63, "lr": 9.990984487771601e-05}
{"train_loss": 0.7969862222671509, "global_step": 5680, "epoch": 63, "lr": 9.990981007589805e-05}
{"train_loss": 0.7068932056427002, "global_step": 5681, "epoch": 63, "lr": 9.990977526737033e-05}
{"train_loss": 1.03085196018219, "global_step": 5682, "epoch": 63, "lr": 9.990974045213286e-05}
{"train_loss": 0.9454854726791382, "global_step": 5683, "epoch": 63, "lr": 9.990970563018562e-05}
{"train_loss": 1.17274010181427, "global_step": 5684, "epoch": 63, "lr": 9.990967080152862e-05}
{"train_loss": 0.8480765223503113, "global_step": 5685, "epoch": 63, "lr": 9.99096359661619e-05}
{"train_loss": 0.8709149360656738, "global_step": 5686, "epoch": 63, "lr": 9.99096011240854e-05}
{"train_loss": 0.8070991039276123, "global_step": 5687, "epoch": 63, "lr": 9.990956627529919e-05}
{"train_loss": 0.8632020354270935, "global_step": 5688, "epoch": 63, "lr": 9.990953141980323e-05}
{"train_loss": 1.0288488864898682, "global_step": 5689, "epoch": 63, "lr": 9.990949655759757e-05}
{"train_loss": 1.1935150623321533, "global_step": 5690, "epoch": 63, "lr": 9.990946168868217e-05}
{"train_loss": 0.8964329957962036, "global_step": 5691, "epoch": 63, "lr": 9.990942681305703e-05}
{"train_loss": 0.9136334657669067, "global_step": 5692, "epoch": 63, "lr": 9.99093919307222e-05}
{"train_loss": 1.0072071552276611, "global_step": 5693, "epoch": 63, "lr": 9.990935704167766e-05}
{"train_loss": 0.8257330060005188, "global_step": 5694, "epoch": 63, "lr": 9.990932214592342e-05}
{"train_loss": 0.9173299342059018, "global_step": 5695, "epoch": 63, "lr": 9.990928724345947e-05, "val_loss": 1.1717078685760498}
{"train_loss": 1.0780285596847534, "global_step": 5696, "epoch": 64, "lr": 9.990925233428581e-05}
{"train_loss": 0.8894839882850647, "global_step": 5697, "epoch": 64, "lr": 9.990921741840249e-05}
{"train_loss": 0.9245235323905945, "global_step": 5698, "epoch": 64, "lr": 9.990918249580947e-05}
{"train_loss": 0.846570611000061, "global_step": 5699, "epoch": 64, "lr": 9.990914756650677e-05}
{"train_loss": 0.9777336716651917, "global_step": 5700, "epoch": 64, "lr": 9.990911263049439e-05}
{"train_loss": 1.052626132965088, "global_step": 5701, "epoch": 64, "lr": 9.990907768777233e-05}
{"train_loss": 0.8433568477630615, "global_step": 5702, "epoch": 64, "lr": 9.990904273834063e-05}
{"train_loss": 0.941668689250946, "global_step": 5703, "epoch": 64, "lr": 9.990900778219924e-05}
{"train_loss": 0.9493224024772644, "global_step": 5704, "epoch": 64, "lr": 9.99089728193482e-05}
{"train_loss": 0.8485304117202759, "global_step": 5705, "epoch": 64, "lr": 9.990893784978751e-05}
{"train_loss": 0.9904322624206543, "global_step": 5706, "epoch": 64, "lr": 9.990890287351717e-05}
{"train_loss": 0.9676728248596191, "global_step": 5707, "epoch": 64, "lr": 9.990886789053718e-05}
{"train_loss": 0.794358491897583, "global_step": 5708, "epoch": 64, "lr": 9.990883290084755e-05}
{"train_loss": 1.0283656120300293, "global_step": 5709, "epoch": 64, "lr": 9.99087979044483e-05}
{"train_loss": 0.8612363338470459, "global_step": 5710, "epoch": 64, "lr": 9.990876290133941e-05}
{"train_loss": 1.0150543451309204, "global_step": 5711, "epoch": 64, "lr": 9.990872789152091e-05}
{"train_loss": 0.9313101172447205, "global_step": 5712, "epoch": 64, "lr": 9.990869287499278e-05}
{"train_loss": 1.0058859586715698, "global_step": 5713, "epoch": 64, "lr": 9.990865785175502e-05}
{"train_loss": 0.8724527955055237, "global_step": 5714, "epoch": 64, "lr": 9.990862282180767e-05}
{"train_loss": 0.8761149048805237, "global_step": 5715, "epoch": 64, "lr": 9.99085877851507e-05}
{"train_loss": 0.817361056804657, "global_step": 5716, "epoch": 64, "lr": 9.990855274178414e-05}
{"train_loss": 0.8652884364128113, "global_step": 5717, "epoch": 64, "lr": 9.990851769170799e-05}
{"train_loss": 1.2130622863769531, "global_step": 5718, "epoch": 64, "lr": 9.990848263492223e-05}
{"train_loss": 0.9124072790145874, "global_step": 5719, "epoch": 64, "lr": 9.99084475714269e-05}
{"train_loss": 0.8992653489112854, "global_step": 5720, "epoch": 64, "lr": 9.990841250122198e-05}
{"train_loss": 0.9554989337921143, "global_step": 5721, "epoch": 64, "lr": 9.990837742430747e-05}
{"train_loss": 0.8579180836677551, "global_step": 5722, "epoch": 64, "lr": 9.990834234068341e-05}
{"train_loss": 0.9894178509712219, "global_step": 5723, "epoch": 64, "lr": 9.990830725034975e-05}
{"train_loss": 0.929240882396698, "global_step": 5724, "epoch": 64, "lr": 9.990827215330654e-05}
{"train_loss": 0.9320625066757202, "global_step": 5725, "epoch": 64, "lr": 9.99082370495538e-05}
{"train_loss": 0.8958920836448669, "global_step": 5726, "epoch": 64, "lr": 9.990820193909147e-05}
{"train_loss": 0.84576016664505, "global_step": 5727, "epoch": 64, "lr": 9.990816682191959e-05}
{"train_loss": 0.9463984370231628, "global_step": 5728, "epoch": 64, "lr": 9.990813169803818e-05}
{"train_loss": 1.0601086616516113, "global_step": 5729, "epoch": 64, "lr": 9.990809656744723e-05}
{"train_loss": 0.7872706651687622, "global_step": 5730, "epoch": 64, "lr": 9.990806143014674e-05}
{"train_loss": 1.118826985359192, "global_step": 5731, "epoch": 64, "lr": 9.990802628613673e-05}
{"train_loss": 0.9646947383880615, "global_step": 5732, "epoch": 64, "lr": 9.990799113541718e-05}
{"train_loss": 0.8029849529266357, "global_step": 5733, "epoch": 64, "lr": 9.990795597798813e-05}
{"train_loss": 0.6596531867980957, "global_step": 5734, "epoch": 64, "lr": 9.990792081384953e-05}
{"train_loss": 0.9614436030387878, "global_step": 5735, "epoch": 64, "lr": 9.990788564300144e-05}
{"train_loss": 0.9576351642608643, "global_step": 5736, "epoch": 64, "lr": 9.990785046544384e-05}
{"train_loss": 1.0174661874771118, "global_step": 5737, "epoch": 64, "lr": 9.990781528117675e-05}
{"train_loss": 0.9464371800422668, "global_step": 5738, "epoch": 64, "lr": 9.990778009020014e-05}
{"train_loss": 0.9458176493644714, "global_step": 5739, "epoch": 64, "lr": 9.990774489251405e-05}
{"train_loss": 0.931083619594574, "global_step": 5740, "epoch": 64, "lr": 9.990770968811847e-05}
{"train_loss": 0.8425281643867493, "global_step": 5741, "epoch": 64, "lr": 9.990767447701339e-05}
{"train_loss": 0.8131626844406128, "global_step": 5742, "epoch": 64, "lr": 9.990763925919886e-05}
{"train_loss": 0.843565046787262, "global_step": 5743, "epoch": 64, "lr": 9.990760403467483e-05}
{"train_loss": 0.9208219647407532, "global_step": 5744, "epoch": 64, "lr": 9.990756880344134e-05}
{"train_loss": 0.9569706916809082, "global_step": 5745, "epoch": 64, "lr": 9.99075335654984e-05}
{"train_loss": 0.8445051908493042, "global_step": 5746, "epoch": 64, "lr": 9.990749832084598e-05}
{"train_loss": 0.9167490005493164, "global_step": 5747, "epoch": 64, "lr": 9.990746306948413e-05}
{"train_loss": 1.1505659818649292, "global_step": 5748, "epoch": 64, "lr": 9.990742781141282e-05}
{"train_loss": 0.9713038206100464, "global_step": 5749, "epoch": 64, "lr": 9.990739254663204e-05}
{"train_loss": 0.7862295508384705, "global_step": 5750, "epoch": 64, "lr": 9.990735727514185e-05}
{"train_loss": 1.0235706567764282, "global_step": 5751, "epoch": 64, "lr": 9.990732199694222e-05}
{"train_loss": 1.0736706256866455, "global_step": 5752, "epoch": 64, "lr": 9.990728671203315e-05}
{"train_loss": 0.6876062154769897, "global_step": 5753, "epoch": 64, "lr": 9.990725142041466e-05}
{"train_loss": 0.8058964610099792, "global_step": 5754, "epoch": 64, "lr": 9.990721612208673e-05}
{"train_loss": 0.982819139957428, "global_step": 5755, "epoch": 64, "lr": 9.990718081704941e-05}
{"train_loss": 0.8424030542373657, "global_step": 5756, "epoch": 64, "lr": 9.990714550530266e-05}
{"train_loss": 0.9165213704109192, "global_step": 5757, "epoch": 64, "lr": 9.990711018684651e-05}
{"train_loss": 0.7709046602249146, "global_step": 5758, "epoch": 64, "lr": 9.990707486168095e-05}
{"train_loss": 0.935431718826294, "global_step": 5759, "epoch": 64, "lr": 9.9907039529806e-05}
{"train_loss": 0.93635094165802, "global_step": 5760, "epoch": 64, "lr": 9.990700419122165e-05}
{"train_loss": 0.8925034403800964, "global_step": 5761, "epoch": 64, "lr": 9.990696884592791e-05}
{"train_loss": 0.8694901466369629, "global_step": 5762, "epoch": 64, "lr": 9.99069334939248e-05}
{"train_loss": 0.9105298519134521, "global_step": 5763, "epoch": 64, "lr": 9.99068981352123e-05}
{"train_loss": 0.8331506252288818, "global_step": 5764, "epoch": 64, "lr": 9.990686276979043e-05}
{"train_loss": 0.9557176232337952, "global_step": 5765, "epoch": 64, "lr": 9.990682739765919e-05}
{"train_loss": 0.9715514183044434, "global_step": 5766, "epoch": 64, "lr": 9.990679201881859e-05}
{"train_loss": 0.8226223587989807, "global_step": 5767, "epoch": 64, "lr": 9.990675663326861e-05}
{"train_loss": 0.857653021812439, "global_step": 5768, "epoch": 64, "lr": 9.99067212410093e-05}
{"train_loss": 0.7673471570014954, "global_step": 5769, "epoch": 64, "lr": 9.990668584204063e-05}
{"train_loss": 0.8567669987678528, "global_step": 5770, "epoch": 64, "lr": 9.990665043636261e-05}
{"train_loss": 0.8972370028495789, "global_step": 5771, "epoch": 64, "lr": 9.990661502397524e-05}
{"train_loss": 0.7575851678848267, "global_step": 5772, "epoch": 64, "lr": 9.990657960487855e-05}
{"train_loss": 1.0295089483261108, "global_step": 5773, "epoch": 64, "lr": 9.990654417907253e-05}
{"train_loss": 0.9036986231803894, "global_step": 5774, "epoch": 64, "lr": 9.990650874655719e-05}
{"train_loss": 0.9816014766693115, "global_step": 5775, "epoch": 64, "lr": 9.990647330733251e-05}
{"train_loss": 0.8609914183616638, "global_step": 5776, "epoch": 64, "lr": 9.990643786139852e-05}
{"train_loss": 0.9994107484817505, "global_step": 5777, "epoch": 64, "lr": 9.990640240875522e-05}
{"train_loss": 0.9625595808029175, "global_step": 5778, "epoch": 64, "lr": 9.99063669494026e-05}
{"train_loss": 0.864497184753418, "global_step": 5779, "epoch": 64, "lr": 9.99063314833407e-05}
{"train_loss": 0.891934871673584, "global_step": 5780, "epoch": 64, "lr": 9.990629601056949e-05}
{"train_loss": 1.0993857383728027, "global_step": 5781, "epoch": 64, "lr": 9.990626053108897e-05}
{"train_loss": 0.8079078197479248, "global_step": 5782, "epoch": 64, "lr": 9.99062250448992e-05}
{"train_loss": 0.8987646698951721, "global_step": 5783, "epoch": 64, "lr": 9.99061895520001e-05}
{"train_loss": 0.9170808363496588, "global_step": 5784, "epoch": 64, "lr": 9.990615405239175e-05, "val_loss": 1.1634591817855835}
{"train_loss": 1.1744725704193115, "global_step": 5785, "epoch": 65, "lr": 9.990611854607412e-05}
{"train_loss": 1.0236443281173706, "global_step": 5786, "epoch": 65, "lr": 9.990608303304722e-05}
{"train_loss": 0.9014344215393066, "global_step": 5787, "epoch": 65, "lr": 9.990604751331106e-05}
{"train_loss": 0.8965836763381958, "global_step": 5788, "epoch": 65, "lr": 9.990601198686564e-05}
{"train_loss": 0.7759298086166382, "global_step": 5789, "epoch": 65, "lr": 9.990597645371097e-05}
{"train_loss": 1.1262104511260986, "global_step": 5790, "epoch": 65, "lr": 9.990594091384704e-05}
{"train_loss": 0.974989116191864, "global_step": 5791, "epoch": 65, "lr": 9.990590536727387e-05}
{"train_loss": 0.8655648231506348, "global_step": 5792, "epoch": 65, "lr": 9.990586981399144e-05}
{"train_loss": 0.8448460102081299, "global_step": 5793, "epoch": 65, "lr": 9.99058342539998e-05}
{"train_loss": 0.9025834798812866, "global_step": 5794, "epoch": 65, "lr": 9.990579868729891e-05}
{"train_loss": 0.8557243943214417, "global_step": 5795, "epoch": 65, "lr": 9.990576311388881e-05}
{"train_loss": 0.9147699475288391, "global_step": 5796, "epoch": 65, "lr": 9.990572753376948e-05}
{"train_loss": 0.9488211274147034, "global_step": 5797, "epoch": 65, "lr": 9.990569194694092e-05}
{"train_loss": 0.8710737228393555, "global_step": 5798, "epoch": 65, "lr": 9.990565635340316e-05}
{"train_loss": 0.782295286655426, "global_step": 5799, "epoch": 65, "lr": 9.990562075315621e-05}
{"train_loss": 0.9189416170120239, "global_step": 5800, "epoch": 65, "lr": 9.990558514620003e-05}
{"train_loss": 1.0032497644424438, "global_step": 5801, "epoch": 65, "lr": 9.990554953253467e-05}
{"train_loss": 0.8796460628509521, "global_step": 5802, "epoch": 65, "lr": 9.990551391216011e-05}
{"train_loss": 0.9260320067405701, "global_step": 5803, "epoch": 65, "lr": 9.990547828507637e-05}
{"train_loss": 0.8720948696136475, "global_step": 5804, "epoch": 65, "lr": 9.990544265128343e-05}
{"train_loss": 0.8734285235404968, "global_step": 5805, "epoch": 65, "lr": 9.990540701078133e-05}
{"train_loss": 0.8991550207138062, "global_step": 5806, "epoch": 65, "lr": 9.990537136357005e-05}
{"train_loss": 0.8881055116653442, "global_step": 5807, "epoch": 65, "lr": 9.99053357096496e-05}
{"train_loss": 0.8596859574317932, "global_step": 5808, "epoch": 65, "lr": 9.990530004901997e-05}
{"train_loss": 0.9766149520874023, "global_step": 5809, "epoch": 65, "lr": 9.990526438168119e-05}
{"train_loss": 1.0662450790405273, "global_step": 5810, "epoch": 65, "lr": 9.990522870763328e-05}
{"train_loss": 0.9049931168556213, "global_step": 5811, "epoch": 65, "lr": 9.99051930268762e-05}
{"train_loss": 0.7940581440925598, "global_step": 5812, "epoch": 65, "lr": 9.990515733940997e-05}
{"train_loss": 0.7734123468399048, "global_step": 5813, "epoch": 65, "lr": 9.99051216452346e-05}
{"train_loss": 1.0203937292099, "global_step": 5814, "epoch": 65, "lr": 9.99050859443501e-05}
{"train_loss": 0.9214996099472046, "global_step": 5815, "epoch": 65, "lr": 9.990505023675647e-05}
{"train_loss": 0.9869050979614258, "global_step": 5816, "epoch": 65, "lr": 9.990501452245373e-05}
{"train_loss": 0.8322970271110535, "global_step": 5817, "epoch": 65, "lr": 9.990497880144185e-05}
{"train_loss": 0.9956617951393127, "global_step": 5818, "epoch": 65, "lr": 9.990494307372085e-05}
{"train_loss": 0.7910579442977905, "global_step": 5819, "epoch": 65, "lr": 9.990490733929075e-05}
{"train_loss": 0.9459569454193115, "global_step": 5820, "epoch": 65, "lr": 9.990487159815153e-05}
{"train_loss": 0.7664191126823425, "global_step": 5821, "epoch": 65, "lr": 9.990483585030322e-05}
{"train_loss": 1.1026109457015991, "global_step": 5822, "epoch": 65, "lr": 9.990480009574581e-05}
{"train_loss": 0.982875645160675, "global_step": 5823, "epoch": 65, "lr": 9.99047643344793e-05}
{"train_loss": 0.9016082882881165, "global_step": 5824, "epoch": 65, "lr": 9.990472856650373e-05}
{"train_loss": 1.159176230430603, "global_step": 5825, "epoch": 65, "lr": 9.990469279181905e-05}
{"train_loss": 1.0652780532836914, "global_step": 5826, "epoch": 65, "lr": 9.990465701042531e-05}
{"train_loss": 0.9905335307121277, "global_step": 5827, "epoch": 65, "lr": 9.990462122232248e-05}
{"train_loss": 0.8769218921661377, "global_step": 5828, "epoch": 65, "lr": 9.99045854275106e-05}
{"train_loss": 0.9466241598129272, "global_step": 5829, "epoch": 65, "lr": 9.990454962598964e-05}
{"train_loss": 0.9652619361877441, "global_step": 5830, "epoch": 65, "lr": 9.990451381775963e-05}
{"train_loss": 0.7955721616744995, "global_step": 5831, "epoch": 65, "lr": 9.990447800282056e-05}
{"train_loss": 0.9213863015174866, "global_step": 5832, "epoch": 65, "lr": 9.990444218117245e-05}
{"train_loss": 0.8277378082275391, "global_step": 5833, "epoch": 65, "lr": 9.990440635281529e-05}
{"train_loss": 0.8473173975944519, "global_step": 5834, "epoch": 65, "lr": 9.990437051774909e-05}
{"train_loss": 0.7904688715934753, "global_step": 5835, "epoch": 65, "lr": 9.990433467597386e-05}
{"train_loss": 0.8538753390312195, "global_step": 5836, "epoch": 65, "lr": 9.990429882748959e-05}
{"train_loss": 1.012410044670105, "global_step": 5837, "epoch": 65, "lr": 9.990426297229632e-05}
{"train_loss": 0.674048125743866, "global_step": 5838, "epoch": 65, "lr": 9.9904227110394e-05}
{"train_loss": 1.1317074298858643, "global_step": 5839, "epoch": 65, "lr": 9.990419124178269e-05}
{"train_loss": 0.9177418947219849, "global_step": 5840, "epoch": 65, "lr": 9.990415536646236e-05}
{"train_loss": 1.0739843845367432, "global_step": 5841, "epoch": 65, "lr": 9.9904119484433e-05}
{"train_loss": 1.1473925113677979, "global_step": 5842, "epoch": 65, "lr": 9.990408359569468e-05}
{"train_loss": 1.0320321321487427, "global_step": 5843, "epoch": 65, "lr": 9.990404770024735e-05}
{"train_loss": 0.8138082027435303, "global_step": 5844, "epoch": 65, "lr": 9.990401179809103e-05}
{"train_loss": 1.0535480976104736, "global_step": 5845, "epoch": 65, "lr": 9.990397588922571e-05}
{"train_loss": 1.0253984928131104, "global_step": 5846, "epoch": 65, "lr": 9.990393997365141e-05}
{"train_loss": 0.8846163749694824, "global_step": 5847, "epoch": 65, "lr": 9.990390405136816e-05}
{"train_loss": 0.9999052286148071, "global_step": 5848, "epoch": 65, "lr": 9.990386812237592e-05}
{"train_loss": 0.9927129745483398, "global_step": 5849, "epoch": 65, "lr": 9.990383218667472e-05}
{"train_loss": 1.0770617723464966, "global_step": 5850, "epoch": 65, "lr": 9.990379624426454e-05}
{"train_loss": 0.964994490146637, "global_step": 5851, "epoch": 65, "lr": 9.990376029514541e-05}
{"train_loss": 0.7244574427604675, "global_step": 5852, "epoch": 65, "lr": 9.990372433931734e-05}
{"train_loss": 0.8165199160575867, "global_step": 5853, "epoch": 65, "lr": 9.990368837678031e-05}
{"train_loss": 0.7731630802154541, "global_step": 5854, "epoch": 65, "lr": 9.990365240753434e-05}
{"train_loss": 0.8814224004745483, "global_step": 5855, "epoch": 65, "lr": 9.990361643157944e-05}
{"train_loss": 1.0265686511993408, "global_step": 5856, "epoch": 65, "lr": 9.99035804489156e-05}
{"train_loss": 0.8546185493469238, "global_step": 5857, "epoch": 65, "lr": 9.990354445954283e-05}
{"train_loss": 0.8336616158485413, "global_step": 5858, "epoch": 65, "lr": 9.990350846346115e-05}
{"train_loss": 0.9235672354698181, "global_step": 5859, "epoch": 65, "lr": 9.990347246067053e-05}
{"train_loss": 1.0796267986297607, "global_step": 5860, "epoch": 65, "lr": 9.990343645117102e-05}
{"train_loss": 1.023099422454834, "global_step": 5861, "epoch": 65, "lr": 9.990340043496258e-05}
{"train_loss": 1.0669145584106445, "global_step": 5862, "epoch": 65, "lr": 9.990336441204524e-05}
{"train_loss": 1.0194401741027832, "global_step": 5863, "epoch": 65, "lr": 9.9903328382419e-05}
{"train_loss": 0.7974987626075745, "global_step": 5864, "epoch": 65, "lr": 9.990329234608389e-05}
{"train_loss": 0.9684257507324219, "global_step": 5865, "epoch": 65, "lr": 9.990325630303988e-05}
{"train_loss": 0.809431254863739, "global_step": 5866, "epoch": 65, "lr": 9.990322025328696e-05}
{"train_loss": 0.9135273098945618, "global_step": 5867, "epoch": 65, "lr": 9.990318419682519e-05}
{"train_loss": 0.9368858933448792, "global_step": 5868, "epoch": 65, "lr": 9.990314813365452e-05}
{"train_loss": 1.0355494022369385, "global_step": 5869, "epoch": 65, "lr": 9.990311206377499e-05}
{"train_loss": 0.9837870597839355, "global_step": 5870, "epoch": 65, "lr": 9.99030759871866e-05}
{"train_loss": 1.0745830535888672, "global_step": 5871, "epoch": 65, "lr": 9.990303990388935e-05}
{"train_loss": 0.7054789066314697, "global_step": 5872, "epoch": 65, "lr": 9.990300381388325e-05}
{"train_loss": 0.9286989007103309, "global_step": 5873, "epoch": 65, "lr": 9.990296771716829e-05, "val_loss": 1.126146912574768, "train_action_mse_error": 52.28858184814453}
{"train_loss": 0.8619420528411865, "global_step": 5874, "epoch": 66, "lr": 9.990293161374447e-05}
{"train_loss": 0.981840968132019, "global_step": 5875, "epoch": 66, "lr": 9.990289550361182e-05}
{"train_loss": 1.0277942419052124, "global_step": 5876, "epoch": 66, "lr": 9.990285938677036e-05}
{"train_loss": 1.0996675491333008, "global_step": 5877, "epoch": 66, "lr": 9.990282326322005e-05}
{"train_loss": 0.9953346252441406, "global_step": 5878, "epoch": 66, "lr": 9.990278713296091e-05}
{"train_loss": 0.9247003197669983, "global_step": 5879, "epoch": 66, "lr": 9.990275099599294e-05}
{"train_loss": 0.914275050163269, "global_step": 5880, "epoch": 66, "lr": 9.990271485231616e-05}
{"train_loss": 1.094181776046753, "global_step": 5881, "epoch": 66, "lr": 9.990267870193057e-05}
{"train_loss": 0.7882929444313049, "global_step": 5882, "epoch": 66, "lr": 9.990264254483618e-05}
{"train_loss": 0.9866750836372375, "global_step": 5883, "epoch": 66, "lr": 9.990260638103297e-05}
{"train_loss": 0.9617648720741272, "global_step": 5884, "epoch": 66, "lr": 9.990257021052098e-05}
{"train_loss": 1.0507688522338867, "global_step": 5885, "epoch": 66, "lr": 9.990253403330019e-05}
{"train_loss": 0.8706479668617249, "global_step": 5886, "epoch": 66, "lr": 9.990249784937062e-05}
{"train_loss": 0.8087130188941956, "global_step": 5887, "epoch": 66, "lr": 9.990246165873226e-05}
{"train_loss": 0.8691080808639526, "global_step": 5888, "epoch": 66, "lr": 9.990242546138512e-05}
{"train_loss": 0.7961480617523193, "global_step": 5889, "epoch": 66, "lr": 9.990238925732922e-05}
{"train_loss": 0.7410144805908203, "global_step": 5890, "epoch": 66, "lr": 9.990235304656455e-05}
{"train_loss": 0.8250006437301636, "global_step": 5891, "epoch": 66, "lr": 9.99023168290911e-05}
{"train_loss": 0.9554980993270874, "global_step": 5892, "epoch": 66, "lr": 9.99022806049089e-05}
{"train_loss": 0.8052696585655212, "global_step": 5893, "epoch": 66, "lr": 9.990224437401796e-05}
{"train_loss": 1.026552438735962, "global_step": 5894, "epoch": 66, "lr": 9.990220813641826e-05}
{"train_loss": 0.8031671047210693, "global_step": 5895, "epoch": 66, "lr": 9.990217189210982e-05}
{"train_loss": 1.0831913948059082, "global_step": 5896, "epoch": 66, "lr": 9.990213564109263e-05}
{"train_loss": 0.948356032371521, "global_step": 5897, "epoch": 66, "lr": 9.990209938336672e-05}
{"train_loss": 0.845478892326355, "global_step": 5898, "epoch": 66, "lr": 9.990206311893207e-05}
{"train_loss": 0.9168906807899475, "global_step": 5899, "epoch": 66, "lr": 9.99020268477887e-05}
{"train_loss": 0.8936522006988525, "global_step": 5900, "epoch": 66, "lr": 9.990199056993661e-05}
{"train_loss": 0.9310618042945862, "global_step": 5901, "epoch": 66, "lr": 9.990195428537581e-05}
{"train_loss": 1.000824213027954, "global_step": 5902, "epoch": 66, "lr": 9.99019179941063e-05}
{"train_loss": 1.0273737907409668, "global_step": 5903, "epoch": 66, "lr": 9.990188169612809e-05}
{"train_loss": 0.8684477806091309, "global_step": 5904, "epoch": 66, "lr": 9.990184539144117e-05}
{"train_loss": 0.9517494440078735, "global_step": 5905, "epoch": 66, "lr": 9.990180908004555e-05}
{"train_loss": 0.721096396446228, "global_step": 5906, "epoch": 66, "lr": 9.990177276194125e-05}
{"train_loss": 0.8707351088523865, "global_step": 5907, "epoch": 66, "lr": 9.990173643712826e-05}
{"train_loss": 0.8788396716117859, "global_step": 5908, "epoch": 66, "lr": 9.99017001056066e-05}
{"train_loss": 1.03574538230896, "global_step": 5909, "epoch": 66, "lr": 9.990166376737625e-05}
{"train_loss": 0.8777688145637512, "global_step": 5910, "epoch": 66, "lr": 9.990162742243724e-05}
{"train_loss": 0.8362078666687012, "global_step": 5911, "epoch": 66, "lr": 9.990159107078955e-05}
{"train_loss": 0.8717749118804932, "global_step": 5912, "epoch": 66, "lr": 9.990155471243322e-05}
{"train_loss": 0.9871407151222229, "global_step": 5913, "epoch": 66, "lr": 9.990151834736823e-05}
{"train_loss": 0.7392423748970032, "global_step": 5914, "epoch": 66, "lr": 9.990148197559459e-05}
{"train_loss": 0.7842033505439758, "global_step": 5915, "epoch": 66, "lr": 9.990144559711229e-05}
{"train_loss": 0.7907598614692688, "global_step": 5916, "epoch": 66, "lr": 9.990140921192135e-05}
{"train_loss": 0.9518304467201233, "global_step": 5917, "epoch": 66, "lr": 9.99013728200218e-05}
{"train_loss": 0.773927628993988, "global_step": 5918, "epoch": 66, "lr": 9.990133642141359e-05}
{"train_loss": 1.0358221530914307, "global_step": 5919, "epoch": 66, "lr": 9.990130001609676e-05}
{"train_loss": 0.8157024383544922, "global_step": 5920, "epoch": 66, "lr": 9.99012636040713e-05}
{"train_loss": 0.7913328409194946, "global_step": 5921, "epoch": 66, "lr": 9.990122718533724e-05}
{"train_loss": 1.0981541872024536, "global_step": 5922, "epoch": 66, "lr": 9.990119075989457e-05}
{"train_loss": 0.9245736002922058, "global_step": 5923, "epoch": 66, "lr": 9.990115432774327e-05}
{"train_loss": 0.8380159139633179, "global_step": 5924, "epoch": 66, "lr": 9.990111788888339e-05}
{"train_loss": 0.9044829607009888, "global_step": 5925, "epoch": 66, "lr": 9.990108144331492e-05}
{"train_loss": 0.8533251881599426, "global_step": 5926, "epoch": 66, "lr": 9.990104499103782e-05}
{"train_loss": 1.0512295961380005, "global_step": 5927, "epoch": 66, "lr": 9.990100853205217e-05}
{"train_loss": 0.7254168391227722, "global_step": 5928, "epoch": 66, "lr": 9.990097206635794e-05}
{"train_loss": 0.8902287483215332, "global_step": 5929, "epoch": 66, "lr": 9.990093559395511e-05}
{"train_loss": 0.8094177842140198, "global_step": 5930, "epoch": 66, "lr": 9.990089911484371e-05}
{"train_loss": 1.0612868070602417, "global_step": 5931, "epoch": 66, "lr": 9.990086262902374e-05}
{"train_loss": 0.7754570841789246, "global_step": 5932, "epoch": 66, "lr": 9.990082613649522e-05}
{"train_loss": 0.8612217903137207, "global_step": 5933, "epoch": 66, "lr": 9.990078963725813e-05}
{"train_loss": 0.8904284238815308, "global_step": 5934, "epoch": 66, "lr": 9.99007531313125e-05}
{"train_loss": 1.1178210973739624, "global_step": 5935, "epoch": 66, "lr": 9.990071661865832e-05}
{"train_loss": 0.9222795367240906, "global_step": 5936, "epoch": 66, "lr": 9.990068009929558e-05}
{"train_loss": 0.8062581419944763, "global_step": 5937, "epoch": 66, "lr": 9.990064357322432e-05}
{"train_loss": 0.838310718536377, "global_step": 5938, "epoch": 66, "lr": 9.990060704044453e-05}
{"train_loss": 0.9688560366630554, "global_step": 5939, "epoch": 66, "lr": 9.99005705009562e-05}
{"train_loss": 0.800035297870636, "global_step": 5940, "epoch": 66, "lr": 9.990053395475933e-05}
{"train_loss": 0.9080997109413147, "global_step": 5941, "epoch": 66, "lr": 9.990049740185396e-05}
{"train_loss": 0.7481852173805237, "global_step": 5942, "epoch": 66, "lr": 9.990046084224009e-05}
{"train_loss": 0.8853167295455933, "global_step": 5943, "epoch": 66, "lr": 9.99004242759177e-05}
{"train_loss": 0.930420458316803, "global_step": 5944, "epoch": 66, "lr": 9.99003877028868e-05}
{"train_loss": 0.771996796131134, "global_step": 5945, "epoch": 66, "lr": 9.99003511231474e-05}
{"train_loss": 0.9941007494926453, "global_step": 5946, "epoch": 66, "lr": 9.990031453669952e-05}
{"train_loss": 0.9646777510643005, "global_step": 5947, "epoch": 66, "lr": 9.990027794354314e-05}
{"train_loss": 0.922344446182251, "global_step": 5948, "epoch": 66, "lr": 9.990024134367828e-05}
{"train_loss": 0.8146218657493591, "global_step": 5949, "epoch": 66, "lr": 9.990020473710493e-05}
{"train_loss": 0.8367421627044678, "global_step": 5950, "epoch": 66, "lr": 9.990016812382312e-05}
{"train_loss": 0.9813642501831055, "global_step": 5951, "epoch": 66, "lr": 9.990013150383284e-05}
{"train_loss": 0.9610848426818848, "global_step": 5952, "epoch": 66, "lr": 9.990009487713408e-05}
{"train_loss": 0.7783365249633789, "global_step": 5953, "epoch": 66, "lr": 9.990005824372687e-05}
{"train_loss": 0.9259039163589478, "global_step": 5954, "epoch": 66, "lr": 9.990002160361121e-05}
{"train_loss": 1.0791112184524536, "global_step": 5955, "epoch": 66, "lr": 9.98999849567871e-05}
{"train_loss": 1.2362114191055298, "global_step": 5956, "epoch": 66, "lr": 9.989994830325453e-05}
{"train_loss": 0.8747972249984741, "global_step": 5957, "epoch": 66, "lr": 9.989991164301353e-05}
{"train_loss": 0.9748231172561646, "global_step": 5958, "epoch": 66, "lr": 9.98998749760641e-05}
{"train_loss": 0.9559529423713684, "global_step": 5959, "epoch": 66, "lr": 9.989983830240625e-05}
{"train_loss": 1.0422738790512085, "global_step": 5960, "epoch": 66, "lr": 9.989980162203996e-05}
{"train_loss": 0.6928827166557312, "global_step": 5961, "epoch": 66, "lr": 9.989976493496525e-05}
{"train_loss": 0.9090945318843542, "global_step": 5962, "epoch": 66, "lr": 9.989972824118212e-05, "val_loss": 1.1493953466415405}
{"train_loss": 1.0092177391052246, "global_step": 5963, "epoch": 67, "lr": 9.989969154069058e-05}
{"train_loss": 1.08542799949646, "global_step": 5964, "epoch": 67, "lr": 9.989965483349065e-05}
{"train_loss": 1.157819390296936, "global_step": 5965, "epoch": 67, "lr": 9.98996181195823e-05}
{"train_loss": 0.9366103410720825, "global_step": 5966, "epoch": 67, "lr": 9.989958139896558e-05}
{"train_loss": 0.8131084442138672, "global_step": 5967, "epoch": 67, "lr": 9.989954467164045e-05}
{"train_loss": 1.0732988119125366, "global_step": 5968, "epoch": 67, "lr": 9.989950793760695e-05}
{"train_loss": 1.023796558380127, "global_step": 5969, "epoch": 67, "lr": 9.989947119686504e-05}
{"train_loss": 1.050489902496338, "global_step": 5970, "epoch": 67, "lr": 9.989943444941478e-05}
{"train_loss": 1.052528738975525, "global_step": 5971, "epoch": 67, "lr": 9.989939769525614e-05}
{"train_loss": 0.8467664122581482, "global_step": 5972, "epoch": 67, "lr": 9.989936093438915e-05}
{"train_loss": 0.8722323179244995, "global_step": 5973, "epoch": 67, "lr": 9.989932416681378e-05}
{"train_loss": 0.8974708914756775, "global_step": 5974, "epoch": 67, "lr": 9.989928739253005e-05}
{"train_loss": 1.0595446825027466, "global_step": 5975, "epoch": 67, "lr": 9.989925061153798e-05}
{"train_loss": 0.899424135684967, "global_step": 5976, "epoch": 67, "lr": 9.989921382383756e-05}
{"train_loss": 0.8855178356170654, "global_step": 5977, "epoch": 67, "lr": 9.989917702942881e-05}
{"train_loss": 1.015262246131897, "global_step": 5978, "epoch": 67, "lr": 9.989914022831171e-05}
{"train_loss": 0.9041748046875, "global_step": 5979, "epoch": 67, "lr": 9.989910342048627e-05}
{"train_loss": 0.8586102724075317, "global_step": 5980, "epoch": 67, "lr": 9.989906660595252e-05}
{"train_loss": 0.8397986888885498, "global_step": 5981, "epoch": 67, "lr": 9.989902978471045e-05}
{"train_loss": 0.8956365585327148, "global_step": 5982, "epoch": 67, "lr": 9.989899295676007e-05}
{"train_loss": 0.9943309426307678, "global_step": 5983, "epoch": 67, "lr": 9.989895612210136e-05}
{"train_loss": 0.775588870048523, "global_step": 5984, "epoch": 67, "lr": 9.989891928073435e-05}
{"train_loss": 0.8222982883453369, "global_step": 5985, "epoch": 67, "lr": 9.989888243265903e-05}
{"train_loss": 0.9120109677314758, "global_step": 5986, "epoch": 67, "lr": 9.989884557787543e-05}
{"train_loss": 1.1191214323043823, "global_step": 5987, "epoch": 67, "lr": 9.989880871638353e-05}
{"train_loss": 0.9587887525558472, "global_step": 5988, "epoch": 67, "lr": 9.989877184818335e-05}
{"train_loss": 1.0234038829803467, "global_step": 5989, "epoch": 67, "lr": 9.989873497327489e-05}
{"train_loss": 0.8232249617576599, "global_step": 5990, "epoch": 67, "lr": 9.989869809165815e-05}
{"train_loss": 1.0664290189743042, "global_step": 5991, "epoch": 67, "lr": 9.989866120333313e-05}
{"train_loss": 0.8147774934768677, "global_step": 5992, "epoch": 67, "lr": 9.989862430829984e-05}
{"train_loss": 0.9977076053619385, "global_step": 5993, "epoch": 67, "lr": 9.98985874065583e-05}
{"train_loss": 0.8491250872612, "global_step": 5994, "epoch": 67, "lr": 9.98985504981085e-05}
{"train_loss": 0.9172706007957458, "global_step": 5995, "epoch": 67, "lr": 9.989851358295044e-05}
{"train_loss": 0.8917747735977173, "global_step": 5996, "epoch": 67, "lr": 9.989847666108415e-05}
{"train_loss": 1.0168263912200928, "global_step": 5997, "epoch": 67, "lr": 9.989843973250961e-05}
{"train_loss": 0.945090651512146, "global_step": 5998, "epoch": 67, "lr": 9.989840279722683e-05}
{"train_loss": 0.982826292514801, "global_step": 5999, "epoch": 67, "lr": 9.989836585523582e-05}
{"train_loss": 0.8131015300750732, "global_step": 6000, "epoch": 67, "lr": 9.989832890653658e-05}
{"train_loss": 0.9768080711364746, "global_step": 6001, "epoch": 67, "lr": 9.989829195112912e-05}
{"train_loss": 0.9046886563301086, "global_step": 6002, "epoch": 67, "lr": 9.989825498901344e-05}
{"train_loss": 1.0266544818878174, "global_step": 6003, "epoch": 67, "lr": 9.989821802018955e-05}
{"train_loss": 0.9631789922714233, "global_step": 6004, "epoch": 67, "lr": 9.989818104465745e-05}
{"train_loss": 0.9546740055084229, "global_step": 6005, "epoch": 67, "lr": 9.989814406241716e-05}
{"train_loss": 0.9342157244682312, "global_step": 6006, "epoch": 67, "lr": 9.989810707346866e-05}
{"train_loss": 0.9181064367294312, "global_step": 6007, "epoch": 67, "lr": 9.989807007781197e-05}
{"train_loss": 0.9649137258529663, "global_step": 6008, "epoch": 67, "lr": 9.98980330754471e-05}
{"train_loss": 0.8867170810699463, "global_step": 6009, "epoch": 67, "lr": 9.989799606637404e-05}
{"train_loss": 0.9309790134429932, "global_step": 6010, "epoch": 67, "lr": 9.98979590505928e-05}
{"train_loss": 0.979600727558136, "global_step": 6011, "epoch": 67, "lr": 9.98979220281034e-05}
{"train_loss": 1.0521420240402222, "global_step": 6012, "epoch": 67, "lr": 9.989788499890582e-05}
{"train_loss": 0.8043138384819031, "global_step": 6013, "epoch": 67, "lr": 9.989784796300008e-05}
{"train_loss": 0.8069005608558655, "global_step": 6014, "epoch": 67, "lr": 9.989781092038619e-05}
{"train_loss": 0.7602548003196716, "global_step": 6015, "epoch": 67, "lr": 9.989777387106414e-05}
{"train_loss": 0.863867461681366, "global_step": 6016, "epoch": 67, "lr": 9.989773681503394e-05}
{"train_loss": 0.8789007067680359, "global_step": 6017, "epoch": 67, "lr": 9.989769975229561e-05}
{"train_loss": 0.8334807753562927, "global_step": 6018, "epoch": 67, "lr": 9.989766268284913e-05}
{"train_loss": 0.6982209086418152, "global_step": 6019, "epoch": 67, "lr": 9.989762560669452e-05}
{"train_loss": 0.9366495013237, "global_step": 6020, "epoch": 67, "lr": 9.989758852383178e-05}
{"train_loss": 0.8119973540306091, "global_step": 6021, "epoch": 67, "lr": 9.989755143426093e-05}
{"train_loss": 0.9515250325202942, "global_step": 6022, "epoch": 67, "lr": 9.989751433798195e-05}
{"train_loss": 0.9163720011711121, "global_step": 6023, "epoch": 67, "lr": 9.989747723499487e-05}
{"train_loss": 0.9885221719741821, "global_step": 6024, "epoch": 67, "lr": 9.989744012529967e-05}
{"train_loss": 0.9401116371154785, "global_step": 6025, "epoch": 67, "lr": 9.989740300889638e-05}
{"train_loss": 0.7806265950202942, "global_step": 6026, "epoch": 67, "lr": 9.989736588578498e-05}
{"train_loss": 0.8238251209259033, "global_step": 6027, "epoch": 67, "lr": 9.98973287559655e-05}
{"train_loss": 0.882592499256134, "global_step": 6028, "epoch": 67, "lr": 9.98972916194379e-05}
{"train_loss": 0.9931057691574097, "global_step": 6029, "epoch": 67, "lr": 9.989725447620225e-05}
{"train_loss": 0.9574440717697144, "global_step": 6030, "epoch": 67, "lr": 9.989721732625852e-05}
{"train_loss": 0.8970860838890076, "global_step": 6031, "epoch": 67, "lr": 9.98971801696067e-05}
{"train_loss": 1.1049649715423584, "global_step": 6032, "epoch": 67, "lr": 9.989714300624681e-05}
{"train_loss": 0.984560489654541, "global_step": 6033, "epoch": 67, "lr": 9.989710583617889e-05}
{"train_loss": 0.6947473883628845, "global_step": 6034, "epoch": 67, "lr": 9.989706865940289e-05}
{"train_loss": 0.8246420621871948, "global_step": 6035, "epoch": 67, "lr": 9.989703147591883e-05}
{"train_loss": 0.7364146709442139, "global_step": 6036, "epoch": 67, "lr": 9.989699428572674e-05}
{"train_loss": 0.8646835088729858, "global_step": 6037, "epoch": 67, "lr": 9.989695708882659e-05}
{"train_loss": 1.0610040426254272, "global_step": 6038, "epoch": 67, "lr": 9.98969198852184e-05}
{"train_loss": 0.8342528939247131, "global_step": 6039, "epoch": 67, "lr": 9.98968826749022e-05}
{"train_loss": 0.9947043061256409, "global_step": 6040, "epoch": 67, "lr": 9.989684545787796e-05}
{"train_loss": 0.7962448596954346, "global_step": 6041, "epoch": 67, "lr": 9.98968082341457e-05}
{"train_loss": 0.7426167726516724, "global_step": 6042, "epoch": 67, "lr": 9.989677100370543e-05}
{"train_loss": 0.9847164154052734, "global_step": 6043, "epoch": 67, "lr": 9.989673376655712e-05}
{"train_loss": 0.8380054235458374, "global_step": 6044, "epoch": 67, "lr": 9.989669652270082e-05}
{"train_loss": 0.9728332757949829, "global_step": 6045, "epoch": 67, "lr": 9.989665927213652e-05}
{"train_loss": 0.8092319369316101, "global_step": 6046, "epoch": 67, "lr": 9.989662201486422e-05}
{"train_loss": 0.7448315024375916, "global_step": 6047, "epoch": 67, "lr": 9.989658475088392e-05}
{"train_loss": 0.9530272483825684, "global_step": 6048, "epoch": 67, "lr": 9.989654748019565e-05}
{"train_loss": 0.7038834691047668, "global_step": 6049, "epoch": 67, "lr": 9.989651020279938e-05}
{"train_loss": 0.9679021835327148, "global_step": 6050, "epoch": 67, "lr": 9.989647291869515e-05}
{"train_loss": 0.9148105141821872, "global_step": 6051, "epoch": 67, "lr": 9.989643562788294e-05, "val_loss": 1.1460285186767578}
{"train_loss": 0.7749548554420471, "global_step": 6052, "epoch": 68, "lr": 9.989639833036275e-05}
{"train_loss": 0.8921828269958496, "global_step": 6053, "epoch": 68, "lr": 9.98963610261346e-05}
{"train_loss": 0.9755691289901733, "global_step": 6054, "epoch": 68, "lr": 9.989632371519851e-05}
{"train_loss": 0.941318154335022, "global_step": 6055, "epoch": 68, "lr": 9.989628639755446e-05}
{"train_loss": 0.8871976733207703, "global_step": 6056, "epoch": 68, "lr": 9.989624907320246e-05}
{"train_loss": 1.052476167678833, "global_step": 6057, "epoch": 68, "lr": 9.98962117421425e-05}
{"train_loss": 0.7825770378112793, "global_step": 6058, "epoch": 68, "lr": 9.989617440437462e-05}
{"train_loss": 0.7655972838401794, "global_step": 6059, "epoch": 68, "lr": 9.989613705989881e-05}
{"train_loss": 0.8650884628295898, "global_step": 6060, "epoch": 68, "lr": 9.989609970871507e-05}
{"train_loss": 0.900388240814209, "global_step": 6061, "epoch": 68, "lr": 9.98960623508234e-05}
{"train_loss": 0.828628420829773, "global_step": 6062, "epoch": 68, "lr": 9.989602498622382e-05}
{"train_loss": 0.9566602110862732, "global_step": 6063, "epoch": 68, "lr": 9.989598761491633e-05}
{"train_loss": 0.8857666850090027, "global_step": 6064, "epoch": 68, "lr": 9.989595023690093e-05}
{"train_loss": 0.9233683943748474, "global_step": 6065, "epoch": 68, "lr": 9.989591285217761e-05}
{"train_loss": 0.8254428505897522, "global_step": 6066, "epoch": 68, "lr": 9.989587546074641e-05}
{"train_loss": 0.7588016986846924, "global_step": 6067, "epoch": 68, "lr": 9.989583806260733e-05}
{"train_loss": 0.7390506863594055, "global_step": 6068, "epoch": 68, "lr": 9.989580065776034e-05}
{"train_loss": 0.7064099907875061, "global_step": 6069, "epoch": 68, "lr": 9.989576324620549e-05}
{"train_loss": 0.8384553790092468, "global_step": 6070, "epoch": 68, "lr": 9.989572582794274e-05}
{"train_loss": 0.91835618019104, "global_step": 6071, "epoch": 68, "lr": 9.989568840297212e-05}
{"train_loss": 0.9318015575408936, "global_step": 6072, "epoch": 68, "lr": 9.989565097129365e-05}
{"train_loss": 0.8618423938751221, "global_step": 6073, "epoch": 68, "lr": 9.98956135329073e-05}
{"train_loss": 0.8093127012252808, "global_step": 6074, "epoch": 68, "lr": 9.98955760878131e-05}
{"train_loss": 0.8136756420135498, "global_step": 6075, "epoch": 68, "lr": 9.989553863601107e-05}
{"train_loss": 0.9895513653755188, "global_step": 6076, "epoch": 68, "lr": 9.989550117750116e-05}
{"train_loss": 1.096873164176941, "global_step": 6077, "epoch": 68, "lr": 9.989546371228342e-05}
{"train_loss": 0.8935088515281677, "global_step": 6078, "epoch": 68, "lr": 9.989542624035783e-05}
{"train_loss": 0.9615557193756104, "global_step": 6079, "epoch": 68, "lr": 9.989538876172444e-05}
{"train_loss": 0.8441743850708008, "global_step": 6080, "epoch": 68, "lr": 9.98953512763832e-05}
{"train_loss": 0.8507243990898132, "global_step": 6081, "epoch": 68, "lr": 9.989531378433415e-05}
{"train_loss": 0.7703317999839783, "global_step": 6082, "epoch": 68, "lr": 9.989527628557727e-05}
{"train_loss": 0.8656421303749084, "global_step": 6083, "epoch": 68, "lr": 9.989523878011258e-05}
{"train_loss": 0.7360067367553711, "global_step": 6084, "epoch": 68, "lr": 9.989520126794009e-05}
{"train_loss": 0.8833261728286743, "global_step": 6085, "epoch": 68, "lr": 9.98951637490598e-05}
{"train_loss": 0.7914342880249023, "global_step": 6086, "epoch": 68, "lr": 9.98951262234717e-05}
{"train_loss": 1.100249171257019, "global_step": 6087, "epoch": 68, "lr": 9.989508869117581e-05}
{"train_loss": 0.8569915890693665, "global_step": 6088, "epoch": 68, "lr": 9.989505115217214e-05}
{"train_loss": 0.8941678404808044, "global_step": 6089, "epoch": 68, "lr": 9.98950136064607e-05}
{"train_loss": 0.8206294775009155, "global_step": 6090, "epoch": 68, "lr": 9.989497605404148e-05}
{"train_loss": 0.9430313110351562, "global_step": 6091, "epoch": 68, "lr": 9.989493849491448e-05}
{"train_loss": 0.819511353969574, "global_step": 6092, "epoch": 68, "lr": 9.98949009290797e-05}
{"train_loss": 0.9551209807395935, "global_step": 6093, "epoch": 68, "lr": 9.989486335653718e-05}
{"train_loss": 1.0313127040863037, "global_step": 6094, "epoch": 68, "lr": 9.98948257772869e-05}
{"train_loss": 0.8585673570632935, "global_step": 6095, "epoch": 68, "lr": 9.989478819132887e-05}
{"train_loss": 0.8081154823303223, "global_step": 6096, "epoch": 68, "lr": 9.989475059866309e-05}
{"train_loss": 0.9824594855308533, "global_step": 6097, "epoch": 68, "lr": 9.989471299928956e-05}
{"train_loss": 0.8787629008293152, "global_step": 6098, "epoch": 68, "lr": 9.989467539320831e-05}
{"train_loss": 0.7578644156455994, "global_step": 6099, "epoch": 68, "lr": 9.989463778041931e-05}
{"train_loss": 0.8687976598739624, "global_step": 6100, "epoch": 68, "lr": 9.98946001609226e-05}
{"train_loss": 0.858536958694458, "global_step": 6101, "epoch": 68, "lr": 9.989456253471818e-05}
{"train_loss": 0.903788149356842, "global_step": 6102, "epoch": 68, "lr": 9.989452490180602e-05}
{"train_loss": 0.8742011785507202, "global_step": 6103, "epoch": 68, "lr": 9.989448726218616e-05}
{"train_loss": 0.9890038967132568, "global_step": 6104, "epoch": 68, "lr": 9.989444961585859e-05}
{"train_loss": 0.9100788831710815, "global_step": 6105, "epoch": 68, "lr": 9.989441196282331e-05}
{"train_loss": 0.9409480094909668, "global_step": 6106, "epoch": 68, "lr": 9.989437430308037e-05}
{"train_loss": 0.9056568145751953, "global_step": 6107, "epoch": 68, "lr": 9.989433663662971e-05}
{"train_loss": 1.0554850101470947, "global_step": 6108, "epoch": 68, "lr": 9.989429896347137e-05}
{"train_loss": 0.7428034543991089, "global_step": 6109, "epoch": 68, "lr": 9.989426128360535e-05}
{"train_loss": 0.9914458394050598, "global_step": 6110, "epoch": 68, "lr": 9.989422359703165e-05}
{"train_loss": 0.9493997097015381, "global_step": 6111, "epoch": 68, "lr": 9.989418590375028e-05}
{"train_loss": 0.9497458934783936, "global_step": 6112, "epoch": 68, "lr": 9.989414820376126e-05}
{"train_loss": 0.91493821144104, "global_step": 6113, "epoch": 68, "lr": 9.989411049706456e-05}
{"train_loss": 1.0216950178146362, "global_step": 6114, "epoch": 68, "lr": 9.989407278366023e-05}
{"train_loss": 0.8384274244308472, "global_step": 6115, "epoch": 68, "lr": 9.989403506354823e-05}
{"train_loss": 0.9323578476905823, "global_step": 6116, "epoch": 68, "lr": 9.98939973367286e-05}
{"train_loss": 1.0662376880645752, "global_step": 6117, "epoch": 68, "lr": 9.989395960320132e-05}
{"train_loss": 0.9075536131858826, "global_step": 6118, "epoch": 68, "lr": 9.98939218629664e-05}
{"train_loss": 0.9311122298240662, "global_step": 6119, "epoch": 68, "lr": 9.989388411602385e-05}
{"train_loss": 0.9346094131469727, "global_step": 6120, "epoch": 68, "lr": 9.989384636237367e-05}
{"train_loss": 0.8572848439216614, "global_step": 6121, "epoch": 68, "lr": 9.98938086020159e-05}
{"train_loss": 0.8587979078292847, "global_step": 6122, "epoch": 68, "lr": 9.989377083495049e-05}
{"train_loss": 0.9357317090034485, "global_step": 6123, "epoch": 68, "lr": 9.989373306117748e-05}
{"train_loss": 0.8352857828140259, "global_step": 6124, "epoch": 68, "lr": 9.989369528069685e-05}
{"train_loss": 1.0873956680297852, "global_step": 6125, "epoch": 68, "lr": 9.989365749350864e-05}
{"train_loss": 0.7586596608161926, "global_step": 6126, "epoch": 68, "lr": 9.989361969961283e-05}
{"train_loss": 0.9932533502578735, "global_step": 6127, "epoch": 68, "lr": 9.989358189900943e-05}
{"train_loss": 0.8710086345672607, "global_step": 6128, "epoch": 68, "lr": 9.989354409169844e-05}
{"train_loss": 0.8550341725349426, "global_step": 6129, "epoch": 68, "lr": 9.989350627767988e-05}
{"train_loss": 0.9651010632514954, "global_step": 6130, "epoch": 68, "lr": 9.989346845695376e-05}
{"train_loss": 0.9334589838981628, "global_step": 6131, "epoch": 68, "lr": 9.989343062952005e-05}
{"train_loss": 1.057746410369873, "global_step": 6132, "epoch": 68, "lr": 9.989339279537878e-05}
{"train_loss": 0.8971943259239197, "global_step": 6133, "epoch": 68, "lr": 9.989335495452996e-05}
{"train_loss": 0.8661336302757263, "global_step": 6134, "epoch": 68, "lr": 9.989331710697357e-05}
{"train_loss": 0.8205500841140747, "global_step": 6135, "epoch": 68, "lr": 9.989327925270965e-05}
{"train_loss": 1.0721285343170166, "global_step": 6136, "epoch": 68, "lr": 9.989324139173817e-05}
{"train_loss": 1.0151951313018799, "global_step": 6137, "epoch": 68, "lr": 9.989320352405917e-05}
{"train_loss": 1.0700700283050537, "global_step": 6138, "epoch": 68, "lr": 9.989316564967261e-05}
{"train_loss": 0.9532920718193054, "global_step": 6139, "epoch": 68, "lr": 9.989312776857855e-05}
{"train_loss": 0.900998153713312, "global_step": 6140, "epoch": 68, "lr": 9.989308988077696e-05, "val_loss": 1.1654632091522217}
{"train_loss": 0.9797104597091675, "global_step": 6141, "epoch": 69, "lr": 9.989305198626784e-05}
{"train_loss": 0.869321346282959, "global_step": 6142, "epoch": 69, "lr": 9.989301408505123e-05}
{"train_loss": 0.6656492948532104, "global_step": 6143, "epoch": 69, "lr": 9.98929761771271e-05}
{"train_loss": 0.7827280759811401, "global_step": 6144, "epoch": 69, "lr": 9.989293826249546e-05}
{"train_loss": 0.7910097241401672, "global_step": 6145, "epoch": 69, "lr": 9.989290034115632e-05}
{"train_loss": 0.8712226152420044, "global_step": 6146, "epoch": 69, "lr": 9.989286241310969e-05}
{"train_loss": 0.814586877822876, "global_step": 6147, "epoch": 69, "lr": 9.989282447835558e-05}
{"train_loss": 0.8209415078163147, "global_step": 6148, "epoch": 69, "lr": 9.989278653689399e-05}
{"train_loss": 0.9122288823127747, "global_step": 6149, "epoch": 69, "lr": 9.989274858872493e-05}
{"train_loss": 0.772636353969574, "global_step": 6150, "epoch": 69, "lr": 9.989271063384837e-05}
{"train_loss": 0.9311283230781555, "global_step": 6151, "epoch": 69, "lr": 9.989267267226436e-05}
{"train_loss": 0.8882286548614502, "global_step": 6152, "epoch": 69, "lr": 9.98926347039729e-05}
{"train_loss": 0.8129733800888062, "global_step": 6153, "epoch": 69, "lr": 9.989259672897397e-05}
{"train_loss": 0.8780701160430908, "global_step": 6154, "epoch": 69, "lr": 9.98925587472676e-05}
{"train_loss": 0.7534120678901672, "global_step": 6155, "epoch": 69, "lr": 9.989252075885376e-05}
{"train_loss": 0.7839162349700928, "global_step": 6156, "epoch": 69, "lr": 9.98924827637325e-05}
{"train_loss": 0.8967124819755554, "global_step": 6157, "epoch": 69, "lr": 9.989244476190379e-05}
{"train_loss": 0.9537086486816406, "global_step": 6158, "epoch": 69, "lr": 9.989240675336767e-05}
{"train_loss": 0.9807204604148865, "global_step": 6159, "epoch": 69, "lr": 9.98923687381241e-05}
{"train_loss": 0.891876220703125, "global_step": 6160, "epoch": 69, "lr": 9.989233071617313e-05}
{"train_loss": 0.9125291705131531, "global_step": 6161, "epoch": 69, "lr": 9.989229268751473e-05}
{"train_loss": 0.9856542348861694, "global_step": 6162, "epoch": 69, "lr": 9.989225465214891e-05}
{"train_loss": 0.7830561995506287, "global_step": 6163, "epoch": 69, "lr": 9.98922166100757e-05}
{"train_loss": 0.8259265422821045, "global_step": 6164, "epoch": 69, "lr": 9.989217856129509e-05}
{"train_loss": 0.9263498783111572, "global_step": 6165, "epoch": 69, "lr": 9.989214050580707e-05}
{"train_loss": 0.8112730979919434, "global_step": 6166, "epoch": 69, "lr": 9.989210244361168e-05}
{"train_loss": 0.8956640958786011, "global_step": 6167, "epoch": 69, "lr": 9.989206437470889e-05}
{"train_loss": 0.8761183023452759, "global_step": 6168, "epoch": 69, "lr": 9.989202629909873e-05}
{"train_loss": 0.9429589509963989, "global_step": 6169, "epoch": 69, "lr": 9.989198821678119e-05}
{"train_loss": 0.8696335554122925, "global_step": 6170, "epoch": 69, "lr": 9.989195012775627e-05}
{"train_loss": 0.9661449193954468, "global_step": 6171, "epoch": 69, "lr": 9.989191203202401e-05}
{"train_loss": 0.7628965973854065, "global_step": 6172, "epoch": 69, "lr": 9.989187392958436e-05}
{"train_loss": 0.8855330348014832, "global_step": 6173, "epoch": 69, "lr": 9.989183582043738e-05}
{"train_loss": 0.8716073632240295, "global_step": 6174, "epoch": 69, "lr": 9.989179770458306e-05}
{"train_loss": 0.8193399310112, "global_step": 6175, "epoch": 69, "lr": 9.989175958202137e-05}
{"train_loss": 0.9762812256813049, "global_step": 6176, "epoch": 69, "lr": 9.989172145275235e-05}
{"train_loss": 1.1580091714859009, "global_step": 6177, "epoch": 69, "lr": 9.9891683316776e-05}
{"train_loss": 0.8191961050033569, "global_step": 6178, "epoch": 69, "lr": 9.98916451740923e-05}
{"train_loss": 0.9832373857498169, "global_step": 6179, "epoch": 69, "lr": 9.989160702470131e-05}
{"train_loss": 0.9991332292556763, "global_step": 6180, "epoch": 69, "lr": 9.989156886860298e-05}
{"train_loss": 0.8213399648666382, "global_step": 6181, "epoch": 69, "lr": 9.989153070579735e-05}
{"train_loss": 0.9795539379119873, "global_step": 6182, "epoch": 69, "lr": 9.98914925362844e-05}
{"train_loss": 1.0884883403778076, "global_step": 6183, "epoch": 69, "lr": 9.989145436006415e-05}
{"train_loss": 1.0802584886550903, "global_step": 6184, "epoch": 69, "lr": 9.98914161771366e-05}
{"train_loss": 0.8356616497039795, "global_step": 6185, "epoch": 69, "lr": 9.989137798750177e-05}
{"train_loss": 0.8018447756767273, "global_step": 6186, "epoch": 69, "lr": 9.989133979115963e-05}
{"train_loss": 0.8913401961326599, "global_step": 6187, "epoch": 69, "lr": 9.989130158811024e-05}
{"train_loss": 0.825710654258728, "global_step": 6188, "epoch": 69, "lr": 9.989126337835353e-05}
{"train_loss": 0.8395166993141174, "global_step": 6189, "epoch": 69, "lr": 9.989122516188958e-05}
{"train_loss": 0.9616745710372925, "global_step": 6190, "epoch": 69, "lr": 9.989118693871835e-05}
{"train_loss": 0.870341420173645, "global_step": 6191, "epoch": 69, "lr": 9.989114870883986e-05}
{"train_loss": 0.8867558240890503, "global_step": 6192, "epoch": 69, "lr": 9.989111047225413e-05}
{"train_loss": 0.8785552382469177, "global_step": 6193, "epoch": 69, "lr": 9.989107222896113e-05}
{"train_loss": 0.8953714370727539, "global_step": 6194, "epoch": 69, "lr": 9.989103397896088e-05}
{"train_loss": 0.8528634309768677, "global_step": 6195, "epoch": 69, "lr": 9.98909957222534e-05}
{"train_loss": 0.9506973028182983, "global_step": 6196, "epoch": 69, "lr": 9.989095745883867e-05}
{"train_loss": 0.9291663765907288, "global_step": 6197, "epoch": 69, "lr": 9.989091918871671e-05}
{"train_loss": 1.0742894411087036, "global_step": 6198, "epoch": 69, "lr": 9.989088091188754e-05}
{"train_loss": 0.9010456800460815, "global_step": 6199, "epoch": 69, "lr": 9.989084262835115e-05}
{"train_loss": 1.0145081281661987, "global_step": 6200, "epoch": 69, "lr": 9.989080433810752e-05}
{"train_loss": 0.7830901145935059, "global_step": 6201, "epoch": 69, "lr": 9.98907660411567e-05}
{"train_loss": 0.9727070927619934, "global_step": 6202, "epoch": 69, "lr": 9.989072773749867e-05}
{"train_loss": 1.0366276502609253, "global_step": 6203, "epoch": 69, "lr": 9.989068942713345e-05}
{"train_loss": 0.8178105354309082, "global_step": 6204, "epoch": 69, "lr": 9.989065111006102e-05}
{"train_loss": 1.0035556554794312, "global_step": 6205, "epoch": 69, "lr": 9.98906127862814e-05}
{"train_loss": 1.0048725605010986, "global_step": 6206, "epoch": 69, "lr": 9.98905744557946e-05}
{"train_loss": 0.9448642134666443, "global_step": 6207, "epoch": 69, "lr": 9.989053611860063e-05}
{"train_loss": 0.8852352499961853, "global_step": 6208, "epoch": 69, "lr": 9.989049777469945e-05}
{"train_loss": 0.8979355692863464, "global_step": 6209, "epoch": 69, "lr": 9.989045942409113e-05}
{"train_loss": 0.9447073936462402, "global_step": 6210, "epoch": 69, "lr": 9.989042106677564e-05}
{"train_loss": 0.7680354714393616, "global_step": 6211, "epoch": 69, "lr": 9.9890382702753e-05}
{"train_loss": 0.9282404780387878, "global_step": 6212, "epoch": 69, "lr": 9.98903443320232e-05}
{"train_loss": 0.7408635020256042, "global_step": 6213, "epoch": 69, "lr": 9.989030595458624e-05}
{"train_loss": 0.9077907204627991, "global_step": 6214, "epoch": 69, "lr": 9.989026757044216e-05}
{"train_loss": 0.9670488834381104, "global_step": 6215, "epoch": 69, "lr": 9.989022917959092e-05}
{"train_loss": 0.7575825452804565, "global_step": 6216, "epoch": 69, "lr": 9.989019078203254e-05}
{"train_loss": 0.7984259128570557, "global_step": 6217, "epoch": 69, "lr": 9.989015237776706e-05}
{"train_loss": 0.9213050007820129, "global_step": 6218, "epoch": 69, "lr": 9.989011396679444e-05}
{"train_loss": 0.8837453722953796, "global_step": 6219, "epoch": 69, "lr": 9.98900755491147e-05}
{"train_loss": 1.02971351146698, "global_step": 6220, "epoch": 69, "lr": 9.989003712472786e-05}
{"train_loss": 1.0289056301116943, "global_step": 6221, "epoch": 69, "lr": 9.98899986936339e-05}
{"train_loss": 1.0735290050506592, "global_step": 6222, "epoch": 69, "lr": 9.988996025583285e-05}
{"train_loss": 0.989142119884491, "global_step": 6223, "epoch": 69, "lr": 9.988992181132469e-05}
{"train_loss": 0.8273674249649048, "global_step": 6224, "epoch": 69, "lr": 9.988988336010944e-05}
{"train_loss": 1.0346133708953857, "global_step": 6225, "epoch": 69, "lr": 9.988984490218712e-05}
{"train_loss": 1.0427355766296387, "global_step": 6226, "epoch": 69, "lr": 9.98898064375577e-05}
{"train_loss": 0.7924733757972717, "global_step": 6227, "epoch": 69, "lr": 9.98897679662212e-05}
{"train_loss": 0.8889622092247009, "global_step": 6228, "epoch": 69, "lr": 9.988972948817765e-05}
{"train_loss": 0.8998099538717377, "global_step": 6229, "epoch": 69, "lr": 9.988969100342702e-05, "val_loss": 1.1432812213897705}
{"train_loss": 0.8241618871688843, "global_step": 6230, "epoch": 70, "lr": 9.988965251196933e-05}
{"train_loss": 1.0780411958694458, "global_step": 6231, "epoch": 70, "lr": 9.98896140138046e-05}
{"train_loss": 0.9091938138008118, "global_step": 6232, "epoch": 70, "lr": 9.988957550893279e-05}
{"train_loss": 0.8451358079910278, "global_step": 6233, "epoch": 70, "lr": 9.988953699735395e-05}
{"train_loss": 0.9937177300453186, "global_step": 6234, "epoch": 70, "lr": 9.988949847906807e-05}
{"train_loss": 0.9437673091888428, "global_step": 6235, "epoch": 70, "lr": 9.988945995407514e-05}
{"train_loss": 0.9434447884559631, "global_step": 6236, "epoch": 70, "lr": 9.98894214223752e-05}
{"train_loss": 0.7129753828048706, "global_step": 6237, "epoch": 70, "lr": 9.988938288396824e-05}
{"train_loss": 1.050199031829834, "global_step": 6238, "epoch": 70, "lr": 9.988934433885426e-05}
{"train_loss": 0.9080849885940552, "global_step": 6239, "epoch": 70, "lr": 9.988930578703323e-05}
{"train_loss": 0.8298395872116089, "global_step": 6240, "epoch": 70, "lr": 9.988926722850524e-05}
{"train_loss": 0.8726608157157898, "global_step": 6241, "epoch": 70, "lr": 9.988922866327022e-05}
{"train_loss": 0.6984422206878662, "global_step": 6242, "epoch": 70, "lr": 9.98891900913282e-05}
{"train_loss": 0.6976169347763062, "global_step": 6243, "epoch": 70, "lr": 9.988915151267919e-05}
{"train_loss": 0.6994881629943848, "global_step": 6244, "epoch": 70, "lr": 9.988911292732319e-05}
{"train_loss": 0.7491648197174072, "global_step": 6245, "epoch": 70, "lr": 9.988907433526022e-05}
{"train_loss": 0.8585352897644043, "global_step": 6246, "epoch": 70, "lr": 9.988903573649026e-05}
{"train_loss": 0.8418838381767273, "global_step": 6247, "epoch": 70, "lr": 9.988899713101331e-05}
{"train_loss": 0.9282233119010925, "global_step": 6248, "epoch": 70, "lr": 9.988895851882943e-05}
{"train_loss": 0.95301753282547, "global_step": 6249, "epoch": 70, "lr": 9.988891989993856e-05}
{"train_loss": 1.0333348512649536, "global_step": 6250, "epoch": 70, "lr": 9.988888127434075e-05}
{"train_loss": 0.9361765384674072, "global_step": 6251, "epoch": 70, "lr": 9.988884264203597e-05}
{"train_loss": 0.8363676071166992, "global_step": 6252, "epoch": 70, "lr": 9.988880400302425e-05}
{"train_loss": 0.9934666156768799, "global_step": 6253, "epoch": 70, "lr": 9.98887653573056e-05}
{"train_loss": 1.133789300918579, "global_step": 6254, "epoch": 70, "lr": 9.988872670488e-05}
{"train_loss": 1.0796359777450562, "global_step": 6255, "epoch": 70, "lr": 9.988868804574747e-05}
{"train_loss": 1.0409269332885742, "global_step": 6256, "epoch": 70, "lr": 9.988864937990803e-05}
{"train_loss": 0.7981762290000916, "global_step": 6257, "epoch": 70, "lr": 9.988861070736165e-05}
{"train_loss": 0.8417089581489563, "global_step": 6258, "epoch": 70, "lr": 9.988857202810836e-05}
{"train_loss": 0.9117388725280762, "global_step": 6259, "epoch": 70, "lr": 9.988853334214816e-05}
{"train_loss": 0.9072946906089783, "global_step": 6260, "epoch": 70, "lr": 9.988849464948106e-05}
{"train_loss": 0.6477347612380981, "global_step": 6261, "epoch": 70, "lr": 9.988845595010705e-05}
{"train_loss": 0.9161893725395203, "global_step": 6262, "epoch": 70, "lr": 9.988841724402616e-05}
{"train_loss": 0.8373231291770935, "global_step": 6263, "epoch": 70, "lr": 9.988837853123835e-05}
{"train_loss": 0.9830475449562073, "global_step": 6264, "epoch": 70, "lr": 9.988833981174367e-05}
{"train_loss": 0.9113353490829468, "global_step": 6265, "epoch": 70, "lr": 9.988830108554213e-05}
{"train_loss": 0.6393099427223206, "global_step": 6266, "epoch": 70, "lr": 9.988826235263371e-05}
{"train_loss": 0.8746069073677063, "global_step": 6267, "epoch": 70, "lr": 9.988822361301842e-05}
{"train_loss": 0.847579836845398, "global_step": 6268, "epoch": 70, "lr": 9.988818486669624e-05}
{"train_loss": 0.8525196313858032, "global_step": 6269, "epoch": 70, "lr": 9.988814611366723e-05}
{"train_loss": 0.9601415991783142, "global_step": 6270, "epoch": 70, "lr": 9.988810735393136e-05}
{"train_loss": 0.7428556680679321, "global_step": 6271, "epoch": 70, "lr": 9.988806858748865e-05}
{"train_loss": 0.987202525138855, "global_step": 6272, "epoch": 70, "lr": 9.988802981433908e-05}
{"train_loss": 1.0516587495803833, "global_step": 6273, "epoch": 70, "lr": 9.988799103448267e-05}
{"train_loss": 1.218506097793579, "global_step": 6274, "epoch": 70, "lr": 9.988795224791945e-05}
{"train_loss": 0.7261818051338196, "global_step": 6275, "epoch": 70, "lr": 9.988791345464938e-05}
{"train_loss": 0.8679404258728027, "global_step": 6276, "epoch": 70, "lr": 9.98878746546725e-05}
{"train_loss": 1.019653081893921, "global_step": 6277, "epoch": 70, "lr": 9.988783584798882e-05}
{"train_loss": 0.9286871552467346, "global_step": 6278, "epoch": 70, "lr": 9.98877970345983e-05}
{"train_loss": 0.8405539393424988, "global_step": 6279, "epoch": 70, "lr": 9.988775821450098e-05}
{"train_loss": 0.9418195486068726, "global_step": 6280, "epoch": 70, "lr": 9.988771938769687e-05}
{"train_loss": 0.8525205850601196, "global_step": 6281, "epoch": 70, "lr": 9.988768055418596e-05}
{"train_loss": 1.1321114301681519, "global_step": 6282, "epoch": 70, "lr": 9.988764171396825e-05}
{"train_loss": 0.9117560982704163, "global_step": 6283, "epoch": 70, "lr": 9.988760286704378e-05}
{"train_loss": 0.8001595735549927, "global_step": 6284, "epoch": 70, "lr": 9.98875640134125e-05}
{"train_loss": 0.8609203696250916, "global_step": 6285, "epoch": 70, "lr": 9.988752515307446e-05}
{"train_loss": 0.8015241622924805, "global_step": 6286, "epoch": 70, "lr": 9.988748628602965e-05}
{"train_loss": 0.8739272356033325, "global_step": 6287, "epoch": 70, "lr": 9.988744741227807e-05}
{"train_loss": 0.8188349008560181, "global_step": 6288, "epoch": 70, "lr": 9.988740853181973e-05}
{"train_loss": 0.9826956391334534, "global_step": 6289, "epoch": 70, "lr": 9.988736964465464e-05}
{"train_loss": 1.023046851158142, "global_step": 6290, "epoch": 70, "lr": 9.98873307507828e-05}
{"train_loss": 0.8130412697792053, "global_step": 6291, "epoch": 70, "lr": 9.988729185020422e-05}
{"train_loss": 1.0985702276229858, "global_step": 6292, "epoch": 70, "lr": 9.988725294291889e-05}
{"train_loss": 0.7340555191040039, "global_step": 6293, "epoch": 70, "lr": 9.988721402892684e-05}
{"train_loss": 1.066859483718872, "global_step": 6294, "epoch": 70, "lr": 9.988717510822805e-05}
{"train_loss": 0.9699029922485352, "global_step": 6295, "epoch": 70, "lr": 9.988713618082255e-05}
{"train_loss": 0.9037035703659058, "global_step": 6296, "epoch": 70, "lr": 9.988709724671033e-05}
{"train_loss": 0.6707354187965393, "global_step": 6297, "epoch": 70, "lr": 9.988705830589139e-05}
{"train_loss": 0.731360137462616, "global_step": 6298, "epoch": 70, "lr": 9.988701935836575e-05}
{"train_loss": 0.9373492002487183, "global_step": 6299, "epoch": 70, "lr": 9.98869804041334e-05}
{"train_loss": 0.9838222861289978, "global_step": 6300, "epoch": 70, "lr": 9.988694144319437e-05}
{"train_loss": 0.8835368752479553, "global_step": 6301, "epoch": 70, "lr": 9.988690247554864e-05}
{"train_loss": 1.0038758516311646, "global_step": 6302, "epoch": 70, "lr": 9.988686350119623e-05}
{"train_loss": 1.0655449628829956, "global_step": 6303, "epoch": 70, "lr": 9.988682452013711e-05}
{"train_loss": 0.9092773199081421, "global_step": 6304, "epoch": 70, "lr": 9.988678553237134e-05}
{"train_loss": 0.9969372749328613, "global_step": 6305, "epoch": 70, "lr": 9.98867465378989e-05}
{"train_loss": 0.7883780002593994, "global_step": 6306, "epoch": 70, "lr": 9.988670753671979e-05}
{"train_loss": 0.8737557530403137, "global_step": 6307, "epoch": 70, "lr": 9.988666852883403e-05}
{"train_loss": 0.7722896933555603, "global_step": 6308, "epoch": 70, "lr": 9.988662951424161e-05}
{"train_loss": 1.0061551332473755, "global_step": 6309, "epoch": 70, "lr": 9.988659049294253e-05}
{"train_loss": 0.8524890542030334, "global_step": 6310, "epoch": 70, "lr": 9.988655146493681e-05}
{"train_loss": 0.7803945541381836, "global_step": 6311, "epoch": 70, "lr": 9.988651243022447e-05}
{"train_loss": 0.8746421337127686, "global_step": 6312, "epoch": 70, "lr": 9.988647338880546e-05}
{"train_loss": 0.9560147523880005, "global_step": 6313, "epoch": 70, "lr": 9.988643434067986e-05}
{"train_loss": 0.764380156993866, "global_step": 6314, "epoch": 70, "lr": 9.98863952858476e-05}
{"train_loss": 0.9333484768867493, "global_step": 6315, "epoch": 70, "lr": 9.988635622430876e-05}
{"train_loss": 0.8576467037200928, "global_step": 6316, "epoch": 70, "lr": 9.988631715606328e-05}
{"train_loss": 0.9073975682258606, "global_step": 6317, "epoch": 70, "lr": 9.988627808111121e-05}
{"train_loss": 0.8967737850178493, "global_step": 6318, "epoch": 70, "lr": 9.988623899945252e-05, "val_loss": 1.1273705959320068, "train_action_mse_error": 53.33353805541992}
{"train_loss": 0.8678111433982849, "global_step": 6319, "epoch": 71, "lr": 9.988619991108725e-05}
{"train_loss": 1.000321865081787, "global_step": 6320, "epoch": 71, "lr": 9.988616081601538e-05}
{"train_loss": 0.7545925974845886, "global_step": 6321, "epoch": 71, "lr": 9.988612171423693e-05}
{"train_loss": 0.8853915929794312, "global_step": 6322, "epoch": 71, "lr": 9.98860826057519e-05}
{"train_loss": 0.9609965682029724, "global_step": 6323, "epoch": 71, "lr": 9.988604349056027e-05}
{"train_loss": 0.8412249088287354, "global_step": 6324, "epoch": 71, "lr": 9.988600436866209e-05}
{"train_loss": 0.7699323892593384, "global_step": 6325, "epoch": 71, "lr": 9.988596524005735e-05}
{"train_loss": 0.8987612724304199, "global_step": 6326, "epoch": 71, "lr": 9.988592610474604e-05}
{"train_loss": 0.9694361686706543, "global_step": 6327, "epoch": 71, "lr": 9.988588696272818e-05}
{"train_loss": 0.861855685710907, "global_step": 6328, "epoch": 71, "lr": 9.988584781400376e-05}
{"train_loss": 0.8912855982780457, "global_step": 6329, "epoch": 71, "lr": 9.988580865857281e-05}
{"train_loss": 0.8061895966529846, "global_step": 6330, "epoch": 71, "lr": 9.98857694964353e-05}
{"train_loss": 0.8253721594810486, "global_step": 6331, "epoch": 71, "lr": 9.988573032759129e-05}
{"train_loss": 0.8423486948013306, "global_step": 6332, "epoch": 71, "lr": 9.988569115204073e-05}
{"train_loss": 0.8478639721870422, "global_step": 6333, "epoch": 71, "lr": 9.988565196978366e-05}
{"train_loss": 0.8440540432929993, "global_step": 6334, "epoch": 71, "lr": 9.988561278082005e-05}
{"train_loss": 0.9614157676696777, "global_step": 6335, "epoch": 71, "lr": 9.988557358514994e-05}
{"train_loss": 0.9611304402351379, "global_step": 6336, "epoch": 71, "lr": 9.988553438277333e-05}
{"train_loss": 0.8301619291305542, "global_step": 6337, "epoch": 71, "lr": 9.98854951736902e-05}
{"train_loss": 0.8451194167137146, "global_step": 6338, "epoch": 71, "lr": 9.988545595790058e-05}
{"train_loss": 0.9869953989982605, "global_step": 6339, "epoch": 71, "lr": 9.988541673540448e-05}
{"train_loss": 0.7514348030090332, "global_step": 6340, "epoch": 71, "lr": 9.988537750620189e-05}
{"train_loss": 0.7440100908279419, "global_step": 6341, "epoch": 71, "lr": 9.988533827029281e-05}
{"train_loss": 0.7963033318519592, "global_step": 6342, "epoch": 71, "lr": 9.988529902767726e-05}
{"train_loss": 0.6544637084007263, "global_step": 6343, "epoch": 71, "lr": 9.988525977835524e-05}
{"train_loss": 0.9141150116920471, "global_step": 6344, "epoch": 71, "lr": 9.988522052232675e-05}
{"train_loss": 0.8494911789894104, "global_step": 6345, "epoch": 71, "lr": 9.98851812595918e-05}
{"train_loss": 0.6926157474517822, "global_step": 6346, "epoch": 71, "lr": 9.98851419901504e-05}
{"train_loss": 1.0159941911697388, "global_step": 6347, "epoch": 71, "lr": 9.988510271400256e-05}
{"train_loss": 0.7775008082389832, "global_step": 6348, "epoch": 71, "lr": 9.988506343114826e-05}
{"train_loss": 0.8050540089607239, "global_step": 6349, "epoch": 71, "lr": 9.988502414158753e-05}
{"train_loss": 0.8357870578765869, "global_step": 6350, "epoch": 71, "lr": 9.988498484532036e-05}
{"train_loss": 1.0381784439086914, "global_step": 6351, "epoch": 71, "lr": 9.988494554234677e-05}
{"train_loss": 0.9830635190010071, "global_step": 6352, "epoch": 71, "lr": 9.988490623266674e-05}
{"train_loss": 0.9395163655281067, "global_step": 6353, "epoch": 71, "lr": 9.98848669162803e-05}
{"train_loss": 0.7907049655914307, "global_step": 6354, "epoch": 71, "lr": 9.988482759318746e-05}
{"train_loss": 0.9759818315505981, "global_step": 6355, "epoch": 71, "lr": 9.98847882633882e-05}
{"train_loss": 0.953993558883667, "global_step": 6356, "epoch": 71, "lr": 9.988474892688255e-05}
{"train_loss": 0.842027485370636, "global_step": 6357, "epoch": 71, "lr": 9.988470958367049e-05}
{"train_loss": 0.7836853265762329, "global_step": 6358, "epoch": 71, "lr": 9.988467023375205e-05}
{"train_loss": 1.1347377300262451, "global_step": 6359, "epoch": 71, "lr": 9.988463087712721e-05}
{"train_loss": 0.9558992385864258, "global_step": 6360, "epoch": 71, "lr": 9.988459151379601e-05}
{"train_loss": 0.8991426825523376, "global_step": 6361, "epoch": 71, "lr": 9.988455214375843e-05}
{"train_loss": 1.090502142906189, "global_step": 6362, "epoch": 71, "lr": 9.988451276701447e-05}
{"train_loss": 1.1633522510528564, "global_step": 6363, "epoch": 71, "lr": 9.988447338356415e-05}
{"train_loss": 0.9131587743759155, "global_step": 6364, "epoch": 71, "lr": 9.988443399340745e-05}
{"train_loss": 0.97054123878479, "global_step": 6365, "epoch": 71, "lr": 9.988439459654443e-05}
{"train_loss": 0.9368064403533936, "global_step": 6366, "epoch": 71, "lr": 9.988435519297504e-05}
{"train_loss": 0.9685330390930176, "global_step": 6367, "epoch": 71, "lr": 9.988431578269933e-05}
{"train_loss": 1.0536305904388428, "global_step": 6368, "epoch": 71, "lr": 9.988427636571725e-05}
{"train_loss": 0.9385541081428528, "global_step": 6369, "epoch": 71, "lr": 9.988423694202886e-05}
{"train_loss": 0.9929795861244202, "global_step": 6370, "epoch": 71, "lr": 9.988419751163413e-05}
{"train_loss": 0.9488635063171387, "global_step": 6371, "epoch": 71, "lr": 9.988415807453309e-05}
{"train_loss": 0.7944455742835999, "global_step": 6372, "epoch": 71, "lr": 9.988411863072571e-05}
{"train_loss": 0.9470886588096619, "global_step": 6373, "epoch": 71, "lr": 9.988407918021204e-05}
{"train_loss": 0.8611928820610046, "global_step": 6374, "epoch": 71, "lr": 9.988403972299205e-05}
{"train_loss": 0.648033082485199, "global_step": 6375, "epoch": 71, "lr": 9.988400025906577e-05}
{"train_loss": 0.6931177377700806, "global_step": 6376, "epoch": 71, "lr": 9.988396078843318e-05}
{"train_loss": 0.9751578569412231, "global_step": 6377, "epoch": 71, "lr": 9.988392131109432e-05}
{"train_loss": 0.7260838150978088, "global_step": 6378, "epoch": 71, "lr": 9.988388182704916e-05}
{"train_loss": 0.9713271856307983, "global_step": 6379, "epoch": 71, "lr": 9.988384233629772e-05}
{"train_loss": 0.9241218566894531, "global_step": 6380, "epoch": 71, "lr": 9.988380283884001e-05}
{"train_loss": 0.7696892619132996, "global_step": 6381, "epoch": 71, "lr": 9.988376333467605e-05}
{"train_loss": 0.9230680465698242, "global_step": 6382, "epoch": 71, "lr": 9.98837238238058e-05}
{"train_loss": 0.7956393361091614, "global_step": 6383, "epoch": 71, "lr": 9.988368430622928e-05}
{"train_loss": 1.0582189559936523, "global_step": 6384, "epoch": 71, "lr": 9.988364478194654e-05}
{"train_loss": 0.8524611592292786, "global_step": 6385, "epoch": 71, "lr": 9.988360525095753e-05}
{"train_loss": 1.0517067909240723, "global_step": 6386, "epoch": 71, "lr": 9.988356571326228e-05}
{"train_loss": 0.8960906863212585, "global_step": 6387, "epoch": 71, "lr": 9.988352616886079e-05}
{"train_loss": 0.8814181089401245, "global_step": 6388, "epoch": 71, "lr": 9.988348661775308e-05}
{"train_loss": 0.8749333024024963, "global_step": 6389, "epoch": 71, "lr": 9.988344705993914e-05}
{"train_loss": 0.8132604360580444, "global_step": 6390, "epoch": 71, "lr": 9.988340749541898e-05}
{"train_loss": 0.9449111819267273, "global_step": 6391, "epoch": 71, "lr": 9.98833679241926e-05}
{"train_loss": 0.7678837180137634, "global_step": 6392, "epoch": 71, "lr": 9.988332834625999e-05}
{"train_loss": 1.016945481300354, "global_step": 6393, "epoch": 71, "lr": 9.98832887616212e-05}
{"train_loss": 0.7927986979484558, "global_step": 6394, "epoch": 71, "lr": 9.98832491702762e-05}
{"train_loss": 0.9835577011108398, "global_step": 6395, "epoch": 71, "lr": 9.988320957222502e-05}
{"train_loss": 0.901757001876831, "global_step": 6396, "epoch": 71, "lr": 9.988316996746762e-05}
{"train_loss": 0.7877467274665833, "global_step": 6397, "epoch": 71, "lr": 9.988313035600407e-05}
{"train_loss": 0.855225145816803, "global_step": 6398, "epoch": 71, "lr": 9.988309073783433e-05}
{"train_loss": 0.9285871982574463, "global_step": 6399, "epoch": 71, "lr": 9.988305111295841e-05}
{"train_loss": 0.9014797210693359, "global_step": 6400, "epoch": 71, "lr": 9.988301148137631e-05}
{"train_loss": 0.8400150537490845, "global_step": 6401, "epoch": 71, "lr": 9.988297184308806e-05}
{"train_loss": 0.9417965412139893, "global_step": 6402, "epoch": 71, "lr": 9.988293219809366e-05}
{"train_loss": 0.9363864660263062, "global_step": 6403, "epoch": 71, "lr": 9.98828925463931e-05}
{"train_loss": 0.8336329460144043, "global_step": 6404, "epoch": 71, "lr": 9.98828528879864e-05}
{"train_loss": 0.7091304659843445, "global_step": 6405, "epoch": 71, "lr": 9.988281322287354e-05}
{"train_loss": 0.769516110420227, "global_step": 6406, "epoch": 71, "lr": 9.988277355105457e-05}
{"train_loss": 0.8842320087250699, "global_step": 6407, "epoch": 71, "lr": 9.988273387252944e-05, "val_loss": 1.1279293298721313}
{"train_loss": 0.8633929491043091, "global_step": 6408, "epoch": 72, "lr": 9.988269418729821e-05}
{"train_loss": 0.864663302898407, "global_step": 6409, "epoch": 72, "lr": 9.988265449536085e-05}
{"train_loss": 0.6938633918762207, "global_step": 6410, "epoch": 72, "lr": 9.988261479671736e-05}
{"train_loss": 0.9062690138816833, "global_step": 6411, "epoch": 72, "lr": 9.988257509136779e-05}
{"train_loss": 0.8813474774360657, "global_step": 6412, "epoch": 72, "lr": 9.98825353793121e-05}
{"train_loss": 0.882982611656189, "global_step": 6413, "epoch": 72, "lr": 9.98824956605503e-05}
{"train_loss": 0.8090344071388245, "global_step": 6414, "epoch": 72, "lr": 9.988245593508243e-05}
{"train_loss": 0.9105563163757324, "global_step": 6415, "epoch": 72, "lr": 9.988241620290846e-05}
{"train_loss": 0.9253464937210083, "global_step": 6416, "epoch": 72, "lr": 9.98823764640284e-05}
{"train_loss": 0.8572989106178284, "global_step": 6417, "epoch": 72, "lr": 9.988233671844228e-05}
{"train_loss": 1.0005971193313599, "global_step": 6418, "epoch": 72, "lr": 9.988229696615007e-05}
{"train_loss": 0.9175390601158142, "global_step": 6419, "epoch": 72, "lr": 9.98822572071518e-05}
{"train_loss": 1.0130258798599243, "global_step": 6420, "epoch": 72, "lr": 9.988221744144747e-05}
{"train_loss": 1.0636439323425293, "global_step": 6421, "epoch": 72, "lr": 9.988217766903708e-05}
{"train_loss": 0.9160333871841431, "global_step": 6422, "epoch": 72, "lr": 9.988213788992065e-05}
{"train_loss": 0.9511304497718811, "global_step": 6423, "epoch": 72, "lr": 9.988209810409816e-05}
{"train_loss": 0.8487663269042969, "global_step": 6424, "epoch": 72, "lr": 9.988205831156963e-05}
{"train_loss": 0.9862140417098999, "global_step": 6425, "epoch": 72, "lr": 9.988201851233507e-05}
{"train_loss": 0.7956894040107727, "global_step": 6426, "epoch": 72, "lr": 9.988197870639448e-05}
{"train_loss": 0.7844589948654175, "global_step": 6427, "epoch": 72, "lr": 9.988193889374788e-05}
{"train_loss": 0.839440643787384, "global_step": 6428, "epoch": 72, "lr": 9.988189907439524e-05}
{"train_loss": 0.8424229025840759, "global_step": 6429, "epoch": 72, "lr": 9.98818592483366e-05}
{"train_loss": 0.7949581742286682, "global_step": 6430, "epoch": 72, "lr": 9.988181941557195e-05}
{"train_loss": 0.8697145581245422, "global_step": 6431, "epoch": 72, "lr": 9.988177957610129e-05}
{"train_loss": 0.7389545440673828, "global_step": 6432, "epoch": 72, "lr": 9.988173972992465e-05}
{"train_loss": 0.7824337482452393, "global_step": 6433, "epoch": 72, "lr": 9.9881699877042e-05}
{"train_loss": 1.0787010192871094, "global_step": 6434, "epoch": 72, "lr": 9.988166001745337e-05}
{"train_loss": 0.8827853202819824, "global_step": 6435, "epoch": 72, "lr": 9.988162015115877e-05}
{"train_loss": 0.75535649061203, "global_step": 6436, "epoch": 72, "lr": 9.988158027815817e-05}
{"train_loss": 0.8780558705329895, "global_step": 6437, "epoch": 72, "lr": 9.988154039845162e-05}
{"train_loss": 0.7407213449478149, "global_step": 6438, "epoch": 72, "lr": 9.988150051203908e-05}
{"train_loss": 0.891198992729187, "global_step": 6439, "epoch": 72, "lr": 9.988146061892061e-05}
{"train_loss": 0.8642611503601074, "global_step": 6440, "epoch": 72, "lr": 9.988142071909616e-05}
{"train_loss": 0.8491884469985962, "global_step": 6441, "epoch": 72, "lr": 9.988138081256578e-05}
{"train_loss": 0.8837167620658875, "global_step": 6442, "epoch": 72, "lr": 9.988134089932945e-05}
{"train_loss": 0.960037112236023, "global_step": 6443, "epoch": 72, "lr": 9.988130097938718e-05}
{"train_loss": 0.895966112613678, "global_step": 6444, "epoch": 72, "lr": 9.988126105273898e-05}
{"train_loss": 0.8326586484909058, "global_step": 6445, "epoch": 72, "lr": 9.988122111938485e-05}
{"train_loss": 0.9150816202163696, "global_step": 6446, "epoch": 72, "lr": 9.98811811793248e-05}
{"train_loss": 0.7073432207107544, "global_step": 6447, "epoch": 72, "lr": 9.988114123255883e-05}
{"train_loss": 1.0098692178726196, "global_step": 6448, "epoch": 72, "lr": 9.988110127908695e-05}
{"train_loss": 0.9065458178520203, "global_step": 6449, "epoch": 72, "lr": 9.988106131890918e-05}
{"train_loss": 0.7535626888275146, "global_step": 6450, "epoch": 72, "lr": 9.988102135202548e-05}
{"train_loss": 0.8820063471794128, "global_step": 6451, "epoch": 72, "lr": 9.98809813784359e-05}
{"train_loss": 0.9338144659996033, "global_step": 6452, "epoch": 72, "lr": 9.988094139814044e-05}
{"train_loss": 0.8001223206520081, "global_step": 6453, "epoch": 72, "lr": 9.988090141113908e-05}
{"train_loss": 0.8807053565979004, "global_step": 6454, "epoch": 72, "lr": 9.988086141743184e-05}
{"train_loss": 0.7531542778015137, "global_step": 6455, "epoch": 72, "lr": 9.988082141701873e-05}
{"train_loss": 1.0254101753234863, "global_step": 6456, "epoch": 72, "lr": 9.988078140989974e-05}
{"train_loss": 0.893593966960907, "global_step": 6457, "epoch": 72, "lr": 9.98807413960749e-05}
{"train_loss": 1.053025484085083, "global_step": 6458, "epoch": 72, "lr": 9.988070137554422e-05}
{"train_loss": 0.9860943555831909, "global_step": 6459, "epoch": 72, "lr": 9.988066134830766e-05}
{"train_loss": 0.9117891788482666, "global_step": 6460, "epoch": 72, "lr": 9.988062131436526e-05}
{"train_loss": 0.7973742485046387, "global_step": 6461, "epoch": 72, "lr": 9.988058127371703e-05}
{"train_loss": 0.9532383680343628, "global_step": 6462, "epoch": 72, "lr": 9.988054122636295e-05}
{"train_loss": 0.7869473099708557, "global_step": 6463, "epoch": 72, "lr": 9.988050117230304e-05}
{"train_loss": 0.9034150838851929, "global_step": 6464, "epoch": 72, "lr": 9.98804611115373e-05}
{"train_loss": 0.834303081035614, "global_step": 6465, "epoch": 72, "lr": 9.988042104406577e-05}
{"train_loss": 0.9517372846603394, "global_step": 6466, "epoch": 72, "lr": 9.98803809698884e-05}
{"train_loss": 0.9701094031333923, "global_step": 6467, "epoch": 72, "lr": 9.988034088900521e-05}
{"train_loss": 1.0151699781417847, "global_step": 6468, "epoch": 72, "lr": 9.988030080141623e-05}
{"train_loss": 1.1014306545257568, "global_step": 6469, "epoch": 72, "lr": 9.988026070712146e-05}
{"train_loss": 0.6837961077690125, "global_step": 6470, "epoch": 72, "lr": 9.988022060612088e-05}
{"train_loss": 0.7307130098342896, "global_step": 6471, "epoch": 72, "lr": 9.988018049841453e-05}
{"train_loss": 0.768799364566803, "global_step": 6472, "epoch": 72, "lr": 9.988014038400238e-05}
{"train_loss": 1.0367618799209595, "global_step": 6473, "epoch": 72, "lr": 9.988010026288447e-05}
{"train_loss": 0.8894898891448975, "global_step": 6474, "epoch": 72, "lr": 9.988006013506078e-05}
{"train_loss": 0.8265717029571533, "global_step": 6475, "epoch": 72, "lr": 9.988002000053131e-05}
{"train_loss": 0.729588508605957, "global_step": 6476, "epoch": 72, "lr": 9.98799798592961e-05}
{"train_loss": 0.8411767482757568, "global_step": 6477, "epoch": 72, "lr": 9.987993971135513e-05}
{"train_loss": 0.8885276913642883, "global_step": 6478, "epoch": 72, "lr": 9.987989955670842e-05}
{"train_loss": 0.809119462966919, "global_step": 6479, "epoch": 72, "lr": 9.987985939535594e-05}
{"train_loss": 0.7215483784675598, "global_step": 6480, "epoch": 72, "lr": 9.987981922729774e-05}
{"train_loss": 0.880144476890564, "global_step": 6481, "epoch": 72, "lr": 9.987977905253381e-05}
{"train_loss": 1.0539127588272095, "global_step": 6482, "epoch": 72, "lr": 9.987973887106413e-05}
{"train_loss": 0.7966163158416748, "global_step": 6483, "epoch": 72, "lr": 9.987969868288874e-05}
{"train_loss": 0.7699175477027893, "global_step": 6484, "epoch": 72, "lr": 9.987965848800763e-05}
{"train_loss": 0.9043349623680115, "global_step": 6485, "epoch": 72, "lr": 9.987961828642081e-05}
{"train_loss": 0.9173336029052734, "global_step": 6486, "epoch": 72, "lr": 9.987957807812829e-05}
{"train_loss": 0.8225838541984558, "global_step": 6487, "epoch": 72, "lr": 9.987953786313006e-05}
{"train_loss": 1.015391230583191, "global_step": 6488, "epoch": 72, "lr": 9.987949764142614e-05}
{"train_loss": 0.8771086931228638, "global_step": 6489, "epoch": 72, "lr": 9.987945741301652e-05}
{"train_loss": 0.9855729341506958, "global_step": 6490, "epoch": 72, "lr": 9.987941717790122e-05}
{"train_loss": 0.9152030944824219, "global_step": 6491, "epoch": 72, "lr": 9.987937693608023e-05}
{"train_loss": 0.9900377988815308, "global_step": 6492, "epoch": 72, "lr": 9.987933668755358e-05}
{"train_loss": 0.9628476500511169, "global_step": 6493, "epoch": 72, "lr": 9.987929643232126e-05}
{"train_loss": 0.9554135203361511, "global_step": 6494, "epoch": 72, "lr": 9.987925617038327e-05}
{"train_loss": 0.7463179230690002, "global_step": 6495, "epoch": 72, "lr": 9.987921590173963e-05}
{"train_loss": 0.8800802893852919, "global_step": 6496, "epoch": 72, "lr": 9.987917562639033e-05, "val_loss": 1.1445449590682983}
{"train_loss": 0.8078761100769043, "global_step": 6497, "epoch": 73, "lr": 9.98791353443354e-05}
{"train_loss": 0.8510653972625732, "global_step": 6498, "epoch": 73, "lr": 9.98790950555748e-05}
{"train_loss": 0.717476487159729, "global_step": 6499, "epoch": 73, "lr": 9.987905476010858e-05}
{"train_loss": 0.8274014592170715, "global_step": 6500, "epoch": 73, "lr": 9.987901445793673e-05}
{"train_loss": 0.7438237071037292, "global_step": 6501, "epoch": 73, "lr": 9.987897414905925e-05}
{"train_loss": 0.7833496332168579, "global_step": 6502, "epoch": 73, "lr": 9.987893383347616e-05}
{"train_loss": 0.9390261173248291, "global_step": 6503, "epoch": 73, "lr": 9.987889351118744e-05}
{"train_loss": 0.8784629106521606, "global_step": 6504, "epoch": 73, "lr": 9.987885318219312e-05}
{"train_loss": 0.8589338064193726, "global_step": 6505, "epoch": 73, "lr": 9.987881284649318e-05}
{"train_loss": 0.7052813768386841, "global_step": 6506, "epoch": 73, "lr": 9.987877250408766e-05}
{"train_loss": 0.7272897958755493, "global_step": 6507, "epoch": 73, "lr": 9.987873215497653e-05}
{"train_loss": 0.8857512474060059, "global_step": 6508, "epoch": 73, "lr": 9.987869179915982e-05}
{"train_loss": 0.919903576374054, "global_step": 6509, "epoch": 73, "lr": 9.987865143663755e-05}
{"train_loss": 0.7728745341300964, "global_step": 6510, "epoch": 73, "lr": 9.987861106740968e-05}
{"train_loss": 0.8649973273277283, "global_step": 6511, "epoch": 73, "lr": 9.987857069147625e-05}
{"train_loss": 0.8552285432815552, "global_step": 6512, "epoch": 73, "lr": 9.987853030883724e-05}
{"train_loss": 0.8272978663444519, "global_step": 6513, "epoch": 73, "lr": 9.987848991949269e-05}
{"train_loss": 1.02475905418396, "global_step": 6514, "epoch": 73, "lr": 9.987844952344257e-05}
{"train_loss": 0.9418939352035522, "global_step": 6515, "epoch": 73, "lr": 9.98784091206869e-05}
{"train_loss": 0.9242525696754456, "global_step": 6516, "epoch": 73, "lr": 9.987836871122569e-05}
{"train_loss": 0.7194306254386902, "global_step": 6517, "epoch": 73, "lr": 9.987832829505894e-05}
{"train_loss": 0.8555998802185059, "global_step": 6518, "epoch": 73, "lr": 9.987828787218665e-05}
{"train_loss": 0.7386587262153625, "global_step": 6519, "epoch": 73, "lr": 9.987824744260884e-05}
{"train_loss": 0.7985837459564209, "global_step": 6520, "epoch": 73, "lr": 9.987820700632551e-05}
{"train_loss": 0.9439701437950134, "global_step": 6521, "epoch": 73, "lr": 9.987816656333666e-05}
{"train_loss": 0.9781163334846497, "global_step": 6522, "epoch": 73, "lr": 9.987812611364228e-05}
{"train_loss": 0.6429740786552429, "global_step": 6523, "epoch": 73, "lr": 9.987808565724243e-05}
{"train_loss": 0.8082904815673828, "global_step": 6524, "epoch": 73, "lr": 9.987804519413706e-05}
{"train_loss": 0.9363555908203125, "global_step": 6525, "epoch": 73, "lr": 9.98780047243262e-05}
{"train_loss": 0.7612469792366028, "global_step": 6526, "epoch": 73, "lr": 9.987796424780985e-05}
{"train_loss": 0.9588568210601807, "global_step": 6527, "epoch": 73, "lr": 9.9877923764588e-05}
{"train_loss": 0.8216816186904907, "global_step": 6528, "epoch": 73, "lr": 9.987788327466068e-05}
{"train_loss": 0.8291678428649902, "global_step": 6529, "epoch": 73, "lr": 9.98778427780279e-05}
{"train_loss": 0.8163183331489563, "global_step": 6530, "epoch": 73, "lr": 9.987780227468964e-05}
{"train_loss": 1.0302897691726685, "global_step": 6531, "epoch": 73, "lr": 9.987776176464592e-05}
{"train_loss": 0.7967913746833801, "global_step": 6532, "epoch": 73, "lr": 9.987772124789674e-05}
{"train_loss": 0.7388988733291626, "global_step": 6533, "epoch": 73, "lr": 9.98776807244421e-05}
{"train_loss": 1.0294290781021118, "global_step": 6534, "epoch": 73, "lr": 9.987764019428203e-05}
{"train_loss": 0.8997802138328552, "global_step": 6535, "epoch": 73, "lr": 9.987759965741652e-05}
{"train_loss": 0.8639981150627136, "global_step": 6536, "epoch": 73, "lr": 9.987755911384556e-05}
{"train_loss": 0.9281430244445801, "global_step": 6537, "epoch": 73, "lr": 9.987751856356917e-05}
{"train_loss": 0.8797643184661865, "global_step": 6538, "epoch": 73, "lr": 9.987747800658736e-05}
{"train_loss": 0.8269416689872742, "global_step": 6539, "epoch": 73, "lr": 9.987743744290015e-05}
{"train_loss": 0.8835994601249695, "global_step": 6540, "epoch": 73, "lr": 9.98773968725075e-05}
{"train_loss": 0.9626930356025696, "global_step": 6541, "epoch": 73, "lr": 9.987735629540946e-05}
{"train_loss": 0.7202099561691284, "global_step": 6542, "epoch": 73, "lr": 9.987731571160601e-05}
{"train_loss": 0.8478485941886902, "global_step": 6543, "epoch": 73, "lr": 9.987727512109715e-05}
{"train_loss": 0.8448067903518677, "global_step": 6544, "epoch": 73, "lr": 9.987723452388292e-05}
{"train_loss": 0.9022891521453857, "global_step": 6545, "epoch": 73, "lr": 9.987719391996329e-05}
{"train_loss": 0.8971742391586304, "global_step": 6546, "epoch": 73, "lr": 9.98771533093383e-05}
{"train_loss": 0.8673861026763916, "global_step": 6547, "epoch": 73, "lr": 9.98771126920079e-05}
{"train_loss": 1.1174391508102417, "global_step": 6548, "epoch": 73, "lr": 9.987707206797215e-05}
{"train_loss": 0.8312426209449768, "global_step": 6549, "epoch": 73, "lr": 9.987703143723104e-05}
{"train_loss": 0.8385183811187744, "global_step": 6550, "epoch": 73, "lr": 9.987699079978456e-05}
{"train_loss": 0.9205741882324219, "global_step": 6551, "epoch": 73, "lr": 9.987695015563273e-05}
{"train_loss": 0.8367257118225098, "global_step": 6552, "epoch": 73, "lr": 9.987690950477555e-05}
{"train_loss": 0.8779271245002747, "global_step": 6553, "epoch": 73, "lr": 9.987686884721304e-05}
{"train_loss": 0.947721004486084, "global_step": 6554, "epoch": 73, "lr": 9.987682818294517e-05}
{"train_loss": 0.946343183517456, "global_step": 6555, "epoch": 73, "lr": 9.987678751197199e-05}
{"train_loss": 0.9424429535865784, "global_step": 6556, "epoch": 73, "lr": 9.987674683429348e-05}
{"train_loss": 0.8887487053871155, "global_step": 6557, "epoch": 73, "lr": 9.987670614990965e-05}
{"train_loss": 0.8573111891746521, "global_step": 6558, "epoch": 73, "lr": 9.987666545882048e-05}
{"train_loss": 0.8763937950134277, "global_step": 6559, "epoch": 73, "lr": 9.987662476102604e-05}
{"train_loss": 0.9986320734024048, "global_step": 6560, "epoch": 73, "lr": 9.987658405652626e-05}
{"train_loss": 0.9540514349937439, "global_step": 6561, "epoch": 73, "lr": 9.98765433453212e-05}
{"train_loss": 0.9954536557197571, "global_step": 6562, "epoch": 73, "lr": 9.987650262741085e-05}
{"train_loss": 0.8702967166900635, "global_step": 6563, "epoch": 73, "lr": 9.98764619027952e-05}
{"train_loss": 0.9158576130867004, "global_step": 6564, "epoch": 73, "lr": 9.987642117147427e-05}
{"train_loss": 0.9541705846786499, "global_step": 6565, "epoch": 73, "lr": 9.987638043344807e-05}
{"train_loss": 0.8875552415847778, "global_step": 6566, "epoch": 73, "lr": 9.98763396887166e-05}
{"train_loss": 0.8611724972724915, "global_step": 6567, "epoch": 73, "lr": 9.987629893727985e-05}
{"train_loss": 0.9040041565895081, "global_step": 6568, "epoch": 73, "lr": 9.987625817913786e-05}
{"train_loss": 1.0988391637802124, "global_step": 6569, "epoch": 73, "lr": 9.987621741429061e-05}
{"train_loss": 1.058067798614502, "global_step": 6570, "epoch": 73, "lr": 9.987617664273809e-05}
{"train_loss": 0.8711813688278198, "global_step": 6571, "epoch": 73, "lr": 9.987613586448034e-05}
{"train_loss": 0.9627838134765625, "global_step": 6572, "epoch": 73, "lr": 9.987609507951736e-05}
{"train_loss": 0.7944258451461792, "global_step": 6573, "epoch": 73, "lr": 9.987605428784914e-05}
{"train_loss": 0.9238524436950684, "global_step": 6574, "epoch": 73, "lr": 9.987601348947568e-05}
{"train_loss": 0.8083928227424622, "global_step": 6575, "epoch": 73, "lr": 9.987597268439702e-05}
{"train_loss": 0.8842551112174988, "global_step": 6576, "epoch": 73, "lr": 9.987593187261312e-05}
{"train_loss": 1.2105634212493896, "global_step": 6577, "epoch": 73, "lr": 9.987589105412403e-05}
{"train_loss": 0.9746271967887878, "global_step": 6578, "epoch": 73, "lr": 9.987585022892973e-05}
{"train_loss": 0.9407700300216675, "global_step": 6579, "epoch": 73, "lr": 9.987580939703022e-05}
{"train_loss": 0.814906656742096, "global_step": 6580, "epoch": 73, "lr": 9.987576855842552e-05}
{"train_loss": 0.7277799248695374, "global_step": 6581, "epoch": 73, "lr": 9.987572771311563e-05}
{"train_loss": 0.7150475978851318, "global_step": 6582, "epoch": 73, "lr": 9.987568686110056e-05}
{"train_loss": 1.1335728168487549, "global_step": 6583, "epoch": 73, "lr": 9.987564600238031e-05}
{"train_loss": 0.7449232339859009, "global_step": 6584, "epoch": 73, "lr": 9.987560513695488e-05}
{"train_loss": 0.8771733581350091, "global_step": 6585, "epoch": 73, "lr": 9.987556426482429e-05, "val_loss": 1.1438524723052979}
{"train_loss": 0.8692476153373718, "global_step": 6586, "epoch": 74, "lr": 9.987552338598855e-05}
{"train_loss": 0.7949204444885254, "global_step": 6587, "epoch": 74, "lr": 9.987548250044764e-05}
{"train_loss": 0.6665617227554321, "global_step": 6588, "epoch": 74, "lr": 9.987544160820157e-05}
{"train_loss": 0.8185617923736572, "global_step": 6589, "epoch": 74, "lr": 9.987540070925037e-05}
{"train_loss": 0.7126040458679199, "global_step": 6590, "epoch": 74, "lr": 9.987535980359403e-05}
{"train_loss": 0.9443985223770142, "global_step": 6591, "epoch": 74, "lr": 9.987531889123255e-05}
{"train_loss": 0.743590772151947, "global_step": 6592, "epoch": 74, "lr": 9.987527797216593e-05}
{"train_loss": 0.8617615699768066, "global_step": 6593, "epoch": 74, "lr": 9.987523704639421e-05}
{"train_loss": 0.6887993812561035, "global_step": 6594, "epoch": 74, "lr": 9.987519611391736e-05}
{"train_loss": 0.840792715549469, "global_step": 6595, "epoch": 74, "lr": 9.987515517473541e-05}
{"train_loss": 0.7830618023872375, "global_step": 6596, "epoch": 74, "lr": 9.987511422884835e-05}
{"train_loss": 0.7731695175170898, "global_step": 6597, "epoch": 74, "lr": 9.987507327625617e-05}
{"train_loss": 0.7862399816513062, "global_step": 6598, "epoch": 74, "lr": 9.987503231695891e-05}
{"train_loss": 1.0651121139526367, "global_step": 6599, "epoch": 74, "lr": 9.987499135095655e-05}
{"train_loss": 0.6178427934646606, "global_step": 6600, "epoch": 74, "lr": 9.987495037824912e-05}
{"train_loss": 0.8869197964668274, "global_step": 6601, "epoch": 74, "lr": 9.987490939883659e-05}
{"train_loss": 0.861059308052063, "global_step": 6602, "epoch": 74, "lr": 9.9874868412719e-05}
{"train_loss": 0.8729219436645508, "global_step": 6603, "epoch": 74, "lr": 9.987482741989634e-05}
{"train_loss": 0.8990753889083862, "global_step": 6604, "epoch": 74, "lr": 9.987478642036863e-05}
{"train_loss": 0.8615512847900391, "global_step": 6605, "epoch": 74, "lr": 9.987474541413584e-05}
{"train_loss": 0.771928608417511, "global_step": 6606, "epoch": 74, "lr": 9.987470440119801e-05}
{"train_loss": 0.7009963989257812, "global_step": 6607, "epoch": 74, "lr": 9.987466338155513e-05}
{"train_loss": 0.7949011325836182, "global_step": 6608, "epoch": 74, "lr": 9.987462235520722e-05}
{"train_loss": 0.937838613986969, "global_step": 6609, "epoch": 74, "lr": 9.987458132215426e-05}
{"train_loss": 0.8420583605766296, "global_step": 6610, "epoch": 74, "lr": 9.987454028239628e-05}
{"train_loss": 1.0426141023635864, "global_step": 6611, "epoch": 74, "lr": 9.987449923593327e-05}
{"train_loss": 0.7852890491485596, "global_step": 6612, "epoch": 74, "lr": 9.987445818276525e-05}
{"train_loss": 0.7225556969642639, "global_step": 6613, "epoch": 74, "lr": 9.98744171228922e-05}
{"train_loss": 0.7656297087669373, "global_step": 6614, "epoch": 74, "lr": 9.987437605631417e-05}
{"train_loss": 0.8595090508460999, "global_step": 6615, "epoch": 74, "lr": 9.987433498303111e-05}
{"train_loss": 0.8988150954246521, "global_step": 6616, "epoch": 74, "lr": 9.987429390304308e-05}
{"train_loss": 0.8910877108573914, "global_step": 6617, "epoch": 74, "lr": 9.987425281635003e-05}
{"train_loss": 0.8112084865570068, "global_step": 6618, "epoch": 74, "lr": 9.987421172295201e-05}
{"train_loss": 0.8687602281570435, "global_step": 6619, "epoch": 74, "lr": 9.987417062284902e-05}
{"train_loss": 0.8950516581535339, "global_step": 6620, "epoch": 74, "lr": 9.987412951604105e-05}
{"train_loss": 0.9993335008621216, "global_step": 6621, "epoch": 74, "lr": 9.98740884025281e-05}
{"train_loss": 0.7927255630493164, "global_step": 6622, "epoch": 74, "lr": 9.987404728231019e-05}
{"train_loss": 0.8531852960586548, "global_step": 6623, "epoch": 74, "lr": 9.987400615538733e-05}
{"train_loss": 0.7315691709518433, "global_step": 6624, "epoch": 74, "lr": 9.987396502175952e-05}
{"train_loss": 0.8885097503662109, "global_step": 6625, "epoch": 74, "lr": 9.987392388142675e-05}
{"train_loss": 0.8792581558227539, "global_step": 6626, "epoch": 74, "lr": 9.987388273438904e-05}
{"train_loss": 0.7510766983032227, "global_step": 6627, "epoch": 74, "lr": 9.987384158064641e-05}
{"train_loss": 0.8595476746559143, "global_step": 6628, "epoch": 74, "lr": 9.987380042019883e-05}
{"train_loss": 0.7882503271102905, "global_step": 6629, "epoch": 74, "lr": 9.987375925304634e-05}
{"train_loss": 0.770631730556488, "global_step": 6630, "epoch": 74, "lr": 9.987371807918892e-05}
{"train_loss": 0.9601801633834839, "global_step": 6631, "epoch": 74, "lr": 9.987367689862658e-05}
{"train_loss": 0.8047276735305786, "global_step": 6632, "epoch": 74, "lr": 9.987363571135936e-05}
{"train_loss": 0.9273611903190613, "global_step": 6633, "epoch": 74, "lr": 9.987359451738722e-05}
{"train_loss": 0.6799459457397461, "global_step": 6634, "epoch": 74, "lr": 9.987355331671018e-05}
{"train_loss": 0.7061767578125, "global_step": 6635, "epoch": 74, "lr": 9.987351210932826e-05}
{"train_loss": 0.8775537610054016, "global_step": 6636, "epoch": 74, "lr": 9.987347089524144e-05}
{"train_loss": 1.0095373392105103, "global_step": 6637, "epoch": 74, "lr": 9.987342967444974e-05}
{"train_loss": 0.7152063846588135, "global_step": 6638, "epoch": 74, "lr": 9.987338844695319e-05}
{"train_loss": 0.5881341695785522, "global_step": 6639, "epoch": 74, "lr": 9.987334721275174e-05}
{"train_loss": 0.7789835929870605, "global_step": 6640, "epoch": 74, "lr": 9.987330597184546e-05}
{"train_loss": 0.793509304523468, "global_step": 6641, "epoch": 74, "lr": 9.98732647242343e-05}
{"train_loss": 0.9536011219024658, "global_step": 6642, "epoch": 74, "lr": 9.987322346991828e-05}
{"train_loss": 0.9445191621780396, "global_step": 6643, "epoch": 74, "lr": 9.987318220889744e-05}
{"train_loss": 0.8836462497711182, "global_step": 6644, "epoch": 74, "lr": 9.987314094117172e-05}
{"train_loss": 0.7570223212242126, "global_step": 6645, "epoch": 74, "lr": 9.987309966674119e-05}
{"train_loss": 0.8311970829963684, "global_step": 6646, "epoch": 74, "lr": 9.987305838560582e-05}
{"train_loss": 0.704104483127594, "global_step": 6647, "epoch": 74, "lr": 9.987301709776564e-05}
{"train_loss": 0.8641665577888489, "global_step": 6648, "epoch": 74, "lr": 9.987297580322062e-05}
{"train_loss": 0.9225216507911682, "global_step": 6649, "epoch": 74, "lr": 9.98729345019708e-05}
{"train_loss": 0.7733824849128723, "global_step": 6650, "epoch": 74, "lr": 9.987289319401617e-05}
{"train_loss": 0.9046499133110046, "global_step": 6651, "epoch": 74, "lr": 9.987285187935674e-05}
{"train_loss": 0.8912345767021179, "global_step": 6652, "epoch": 74, "lr": 9.98728105579925e-05}
{"train_loss": 1.0429927110671997, "global_step": 6653, "epoch": 74, "lr": 9.987276922992347e-05}
{"train_loss": 0.9171886444091797, "global_step": 6654, "epoch": 74, "lr": 9.987272789514968e-05}
{"train_loss": 0.9152972102165222, "global_step": 6655, "epoch": 74, "lr": 9.987268655367107e-05}
{"train_loss": 0.9179843068122864, "global_step": 6656, "epoch": 74, "lr": 9.987264520548772e-05}
{"train_loss": 0.8218955993652344, "global_step": 6657, "epoch": 74, "lr": 9.987260385059958e-05}
{"train_loss": 0.808523952960968, "global_step": 6658, "epoch": 74, "lr": 9.987256248900669e-05}
{"train_loss": 0.9076150059700012, "global_step": 6659, "epoch": 74, "lr": 9.987252112070904e-05}
{"train_loss": 0.9077426195144653, "global_step": 6660, "epoch": 74, "lr": 9.987247974570663e-05}
{"train_loss": 0.9357773065567017, "global_step": 6661, "epoch": 74, "lr": 9.987243836399946e-05}
{"train_loss": 0.9625969529151917, "global_step": 6662, "epoch": 74, "lr": 9.987239697558757e-05}
{"train_loss": 0.7537226676940918, "global_step": 6663, "epoch": 74, "lr": 9.987235558047095e-05}
{"train_loss": 0.8677856922149658, "global_step": 6664, "epoch": 74, "lr": 9.987231417864958e-05}
{"train_loss": 0.8987640738487244, "global_step": 6665, "epoch": 74, "lr": 9.98722727701235e-05}
{"train_loss": 1.1391681432724, "global_step": 6666, "epoch": 74, "lr": 9.987223135489269e-05}
{"train_loss": 0.7217270731925964, "global_step": 6667, "epoch": 74, "lr": 9.987218993295717e-05}
{"train_loss": 0.9946938157081604, "global_step": 6668, "epoch": 74, "lr": 9.987214850431694e-05}
{"train_loss": 0.900235652923584, "global_step": 6669, "epoch": 74, "lr": 9.987210706897201e-05}
{"train_loss": 0.7780330181121826, "global_step": 6670, "epoch": 74, "lr": 9.987206562692238e-05}
{"train_loss": 0.9217991232872009, "global_step": 6671, "epoch": 74, "lr": 9.987202417816806e-05}
{"train_loss": 0.8301611542701721, "global_step": 6672, "epoch": 74, "lr": 9.987198272270906e-05}
{"train_loss": 0.8987314701080322, "global_step": 6673, "epoch": 74, "lr": 9.987194126054537e-05}
{"train_loss": 0.8452729589483711, "global_step": 6674, "epoch": 74, "lr": 9.987189979167702e-05, "val_loss": 1.1664783954620361}
{"train_loss": 0.7900480031967163, "global_step": 6675, "epoch": 75, "lr": 9.987185831610398e-05}
{"train_loss": 0.665996789932251, "global_step": 6676, "epoch": 75, "lr": 9.987181683382629e-05}
{"train_loss": 0.772019624710083, "global_step": 6677, "epoch": 75, "lr": 9.987177534484393e-05}
{"train_loss": 0.8915731906890869, "global_step": 6678, "epoch": 75, "lr": 9.987173384915693e-05}
{"train_loss": 0.7808650135993958, "global_step": 6679, "epoch": 75, "lr": 9.987169234676528e-05}
{"train_loss": 0.9772620797157288, "global_step": 6680, "epoch": 75, "lr": 9.987165083766899e-05}
{"train_loss": 1.0294231176376343, "global_step": 6681, "epoch": 75, "lr": 9.987160932186806e-05}
{"train_loss": 0.9537104964256287, "global_step": 6682, "epoch": 75, "lr": 9.98715677993625e-05}
{"train_loss": 0.8787747621536255, "global_step": 6683, "epoch": 75, "lr": 9.987152627015233e-05}
{"train_loss": 1.0028525590896606, "global_step": 6684, "epoch": 75, "lr": 9.987148473423752e-05}
{"train_loss": 0.912562370300293, "global_step": 6685, "epoch": 75, "lr": 9.987144319161812e-05}
{"train_loss": 0.9053401350975037, "global_step": 6686, "epoch": 75, "lr": 9.98714016422941e-05}
{"train_loss": 0.9986454844474792, "global_step": 6687, "epoch": 75, "lr": 9.987136008626547e-05}
{"train_loss": 0.8270463347434998, "global_step": 6688, "epoch": 75, "lr": 9.987131852353225e-05}
{"train_loss": 0.9380794763565063, "global_step": 6689, "epoch": 75, "lr": 9.987127695409444e-05}
{"train_loss": 0.9950875639915466, "global_step": 6690, "epoch": 75, "lr": 9.987123537795204e-05}
{"train_loss": 1.0712409019470215, "global_step": 6691, "epoch": 75, "lr": 9.987119379510508e-05}
{"train_loss": 0.8768075108528137, "global_step": 6692, "epoch": 75, "lr": 9.987115220555352e-05}
{"train_loss": 0.7804470658302307, "global_step": 6693, "epoch": 75, "lr": 9.987111060929742e-05}
{"train_loss": 0.8697435259819031, "global_step": 6694, "epoch": 75, "lr": 9.987106900633673e-05}
{"train_loss": 0.8395028710365295, "global_step": 6695, "epoch": 75, "lr": 9.987102739667149e-05}
{"train_loss": 0.8709625601768494, "global_step": 6696, "epoch": 75, "lr": 9.98709857803017e-05}
{"train_loss": 0.7416501045227051, "global_step": 6697, "epoch": 75, "lr": 9.987094415722737e-05}
{"train_loss": 0.7981268763542175, "global_step": 6698, "epoch": 75, "lr": 9.987090252744848e-05}
{"train_loss": 0.6362591981887817, "global_step": 6699, "epoch": 75, "lr": 9.987086089096507e-05}
{"train_loss": 0.6870520114898682, "global_step": 6700, "epoch": 75, "lr": 9.987081924777712e-05}
{"train_loss": 0.7523798942565918, "global_step": 6701, "epoch": 75, "lr": 9.987077759788467e-05}
{"train_loss": 0.8934059143066406, "global_step": 6702, "epoch": 75, "lr": 9.987073594128769e-05}
{"train_loss": 1.1889369487762451, "global_step": 6703, "epoch": 75, "lr": 9.98706942779862e-05}
{"train_loss": 0.6219350695610046, "global_step": 6704, "epoch": 75, "lr": 9.987065260798019e-05}
{"train_loss": 0.84588223695755, "global_step": 6705, "epoch": 75, "lr": 9.987061093126968e-05}
{"train_loss": 0.7983488440513611, "global_step": 6706, "epoch": 75, "lr": 9.987056924785467e-05}
{"train_loss": 0.9677690863609314, "global_step": 6707, "epoch": 75, "lr": 9.987052755773519e-05}
{"train_loss": 1.0743427276611328, "global_step": 6708, "epoch": 75, "lr": 9.987048586091121e-05}
{"train_loss": 0.724658191204071, "global_step": 6709, "epoch": 75, "lr": 9.987044415738275e-05}
{"train_loss": 0.9666256308555603, "global_step": 6710, "epoch": 75, "lr": 9.987040244714984e-05}
{"train_loss": 0.8129763603210449, "global_step": 6711, "epoch": 75, "lr": 9.987036073021243e-05}
{"train_loss": 0.9094375967979431, "global_step": 6712, "epoch": 75, "lr": 9.987031900657058e-05}
{"train_loss": 0.8711227178573608, "global_step": 6713, "epoch": 75, "lr": 9.987027727622427e-05}
{"train_loss": 0.9938928484916687, "global_step": 6714, "epoch": 75, "lr": 9.98702355391735e-05}
{"train_loss": 0.9984726905822754, "global_step": 6715, "epoch": 75, "lr": 9.987019379541829e-05}
{"train_loss": 0.8064799308776855, "global_step": 6716, "epoch": 75, "lr": 9.987015204495865e-05}
{"train_loss": 0.7563477158546448, "global_step": 6717, "epoch": 75, "lr": 9.987011028779457e-05}
{"train_loss": 1.064665675163269, "global_step": 6718, "epoch": 75, "lr": 9.987006852392604e-05}
{"train_loss": 0.6986417770385742, "global_step": 6719, "epoch": 75, "lr": 9.98700267533531e-05}
{"train_loss": 0.9193845987319946, "global_step": 6720, "epoch": 75, "lr": 9.986998497607576e-05}
{"train_loss": 0.7688649892807007, "global_step": 6721, "epoch": 75, "lr": 9.9869943192094e-05}
{"train_loss": 0.9267801642417908, "global_step": 6722, "epoch": 75, "lr": 9.986990140140782e-05}
{"train_loss": 1.0504302978515625, "global_step": 6723, "epoch": 75, "lr": 9.986985960401724e-05}
{"train_loss": 0.7737342715263367, "global_step": 6724, "epoch": 75, "lr": 9.986981779992228e-05}
{"train_loss": 0.9090399742126465, "global_step": 6725, "epoch": 75, "lr": 9.986977598912292e-05}
{"train_loss": 0.9629056453704834, "global_step": 6726, "epoch": 75, "lr": 9.98697341716192e-05}
{"train_loss": 0.8123929500579834, "global_step": 6727, "epoch": 75, "lr": 9.986969234741108e-05}
{"train_loss": 0.7576718330383301, "global_step": 6728, "epoch": 75, "lr": 9.98696505164986e-05}
{"train_loss": 0.9066884517669678, "global_step": 6729, "epoch": 75, "lr": 9.986960867888173e-05}
{"train_loss": 0.8979414105415344, "global_step": 6730, "epoch": 75, "lr": 9.986956683456052e-05}
{"train_loss": 0.7990538477897644, "global_step": 6731, "epoch": 75, "lr": 9.986952498353494e-05}
{"train_loss": 0.9481915235519409, "global_step": 6732, "epoch": 75, "lr": 9.986948312580503e-05}
{"train_loss": 0.9321796298027039, "global_step": 6733, "epoch": 75, "lr": 9.986944126137076e-05}
{"train_loss": 0.9715943932533264, "global_step": 6734, "epoch": 75, "lr": 9.986939939023217e-05}
{"train_loss": 0.9461362957954407, "global_step": 6735, "epoch": 75, "lr": 9.986935751238922e-05}
{"train_loss": 0.728386640548706, "global_step": 6736, "epoch": 75, "lr": 9.986931562784197e-05}
{"train_loss": 0.8663063049316406, "global_step": 6737, "epoch": 75, "lr": 9.986927373659036e-05}
{"train_loss": 1.0001473426818848, "global_step": 6738, "epoch": 75, "lr": 9.986923183863447e-05}
{"train_loss": 0.864482581615448, "global_step": 6739, "epoch": 75, "lr": 9.986918993397424e-05}
{"train_loss": 1.1259534358978271, "global_step": 6740, "epoch": 75, "lr": 9.986914802260974e-05}
{"train_loss": 0.9043283462524414, "global_step": 6741, "epoch": 75, "lr": 9.986910610454093e-05}
{"train_loss": 0.975906252861023, "global_step": 6742, "epoch": 75, "lr": 9.986906417976781e-05}
{"train_loss": 0.8817220330238342, "global_step": 6743, "epoch": 75, "lr": 9.986902224829042e-05}
{"train_loss": 0.728570818901062, "global_step": 6744, "epoch": 75, "lr": 9.986898031010874e-05}
{"train_loss": 0.9501751661300659, "global_step": 6745, "epoch": 75, "lr": 9.986893836522278e-05}
{"train_loss": 0.7389265298843384, "global_step": 6746, "epoch": 75, "lr": 9.986889641363255e-05}
{"train_loss": 0.8505359888076782, "global_step": 6747, "epoch": 75, "lr": 9.986885445533805e-05}
{"train_loss": 0.7592633962631226, "global_step": 6748, "epoch": 75, "lr": 9.98688124903393e-05}
{"train_loss": 1.0471502542495728, "global_step": 6749, "epoch": 75, "lr": 9.98687705186363e-05}
{"train_loss": 0.7608973979949951, "global_step": 6750, "epoch": 75, "lr": 9.986872854022904e-05}
{"train_loss": 0.9961790442466736, "global_step": 6751, "epoch": 75, "lr": 9.986868655511754e-05}
{"train_loss": 0.8551845550537109, "global_step": 6752, "epoch": 75, "lr": 9.986864456330181e-05}
{"train_loss": 0.8540840744972229, "global_step": 6753, "epoch": 75, "lr": 9.986860256478186e-05}
{"train_loss": 0.8229506015777588, "global_step": 6754, "epoch": 75, "lr": 9.986856055955766e-05}
{"train_loss": 0.9721965789794922, "global_step": 6755, "epoch": 75, "lr": 9.986851854762925e-05}
{"train_loss": 0.856338381767273, "global_step": 6756, "epoch": 75, "lr": 9.986847652899663e-05}
{"train_loss": 0.7751129865646362, "global_step": 6757, "epoch": 75, "lr": 9.986843450365978e-05}
{"train_loss": 0.9563723206520081, "global_step": 6758, "epoch": 75, "lr": 9.986839247161875e-05}
{"train_loss": 0.9726765155792236, "global_step": 6759, "epoch": 75, "lr": 9.98683504328735e-05}
{"train_loss": 0.8149641752243042, "global_step": 6760, "epoch": 75, "lr": 9.986830838742406e-05}
{"train_loss": 1.043459415435791, "global_step": 6761, "epoch": 75, "lr": 9.986826633527046e-05}
{"train_loss": 0.8449544310569763, "global_step": 6762, "epoch": 75, "lr": 9.986822427641267e-05}
{"train_loss": 0.8789950790030233, "global_step": 6763, "epoch": 75, "lr": 9.98681822108507e-05, "val_loss": 1.1872520446777344, "train_action_mse_error": 54.001922607421875}
{"train_loss": 0.9687129259109497, "global_step": 6764, "epoch": 76, "lr": 9.986814013858456e-05}
{"train_loss": 0.8749562501907349, "global_step": 6765, "epoch": 76, "lr": 9.986809805961426e-05}
{"train_loss": 1.0324712991714478, "global_step": 6766, "epoch": 76, "lr": 9.98680559739398e-05}
{"train_loss": 0.7243067026138306, "global_step": 6767, "epoch": 76, "lr": 9.986801388156118e-05}
{"train_loss": 0.7982078194618225, "global_step": 6768, "epoch": 76, "lr": 9.986797178247842e-05}
{"train_loss": 0.7799775004386902, "global_step": 6769, "epoch": 76, "lr": 9.986792967669152e-05}
{"train_loss": 0.8989860415458679, "global_step": 6770, "epoch": 76, "lr": 9.986788756420049e-05}
{"train_loss": 0.7925360798835754, "global_step": 6771, "epoch": 76, "lr": 9.986784544500532e-05}
{"train_loss": 0.8984234929084778, "global_step": 6772, "epoch": 76, "lr": 9.986780331910603e-05}
{"train_loss": 0.7234264612197876, "global_step": 6773, "epoch": 76, "lr": 9.986776118650263e-05}
{"train_loss": 0.6570737361907959, "global_step": 6774, "epoch": 76, "lr": 9.98677190471951e-05}
{"train_loss": 0.8673938512802124, "global_step": 6775, "epoch": 76, "lr": 9.986767690118347e-05}
{"train_loss": 0.7870639562606812, "global_step": 6776, "epoch": 76, "lr": 9.986763474846775e-05}
{"train_loss": 0.8557959794998169, "global_step": 6777, "epoch": 76, "lr": 9.986759258904793e-05}
{"train_loss": 0.846164882183075, "global_step": 6778, "epoch": 76, "lr": 9.986755042292401e-05}
{"train_loss": 0.8351641297340393, "global_step": 6779, "epoch": 76, "lr": 9.9867508250096e-05}
{"train_loss": 0.8685659766197205, "global_step": 6780, "epoch": 76, "lr": 9.986746607056392e-05}
{"train_loss": 0.8396275639533997, "global_step": 6781, "epoch": 76, "lr": 9.986742388432777e-05}
{"train_loss": 0.8606820702552795, "global_step": 6782, "epoch": 76, "lr": 9.986738169138756e-05}
{"train_loss": 0.7824132442474365, "global_step": 6783, "epoch": 76, "lr": 9.986733949174327e-05}
{"train_loss": 0.8907238245010376, "global_step": 6784, "epoch": 76, "lr": 9.986729728539495e-05}
{"train_loss": 0.7810006737709045, "global_step": 6785, "epoch": 76, "lr": 9.986725507234256e-05}
{"train_loss": 1.1076529026031494, "global_step": 6786, "epoch": 76, "lr": 9.986721285258613e-05}
{"train_loss": 0.8217848539352417, "global_step": 6787, "epoch": 76, "lr": 9.986717062612565e-05}
{"train_loss": 0.7737144827842712, "global_step": 6788, "epoch": 76, "lr": 9.986712839296115e-05}
{"train_loss": 0.9574798941612244, "global_step": 6789, "epoch": 76, "lr": 9.986708615309262e-05}
{"train_loss": 0.8212668299674988, "global_step": 6790, "epoch": 76, "lr": 9.986704390652006e-05}
{"train_loss": 0.8236429691314697, "global_step": 6791, "epoch": 76, "lr": 9.98670016532435e-05}
{"train_loss": 0.7599582672119141, "global_step": 6792, "epoch": 76, "lr": 9.986695939326292e-05}
{"train_loss": 0.8220968842506409, "global_step": 6793, "epoch": 76, "lr": 9.986691712657834e-05}
{"train_loss": 0.7936357855796814, "global_step": 6794, "epoch": 76, "lr": 9.986687485318975e-05}
{"train_loss": 0.8020332455635071, "global_step": 6795, "epoch": 76, "lr": 9.986683257309717e-05}
{"train_loss": 0.7781192064285278, "global_step": 6796, "epoch": 76, "lr": 9.986679028630062e-05}
{"train_loss": 0.8233498334884644, "global_step": 6797, "epoch": 76, "lr": 9.986674799280007e-05}
{"train_loss": 0.7384615540504456, "global_step": 6798, "epoch": 76, "lr": 9.986670569259553e-05}
{"train_loss": 0.800545871257782, "global_step": 6799, "epoch": 76, "lr": 9.986666338568704e-05}
{"train_loss": 0.9295869469642639, "global_step": 6800, "epoch": 76, "lr": 9.986662107207458e-05}
{"train_loss": 0.8160092234611511, "global_step": 6801, "epoch": 76, "lr": 9.986657875175817e-05}
{"train_loss": 0.8344365954399109, "global_step": 6802, "epoch": 76, "lr": 9.98665364247378e-05}
{"train_loss": 0.9196028113365173, "global_step": 6803, "epoch": 76, "lr": 9.986649409101347e-05}
{"train_loss": 0.7615624070167542, "global_step": 6804, "epoch": 76, "lr": 9.98664517505852e-05}
{"train_loss": 0.8722966909408569, "global_step": 6805, "epoch": 76, "lr": 9.986640940345302e-05}
{"train_loss": 0.751649796962738, "global_step": 6806, "epoch": 76, "lr": 9.986636704961687e-05}
{"train_loss": 0.9345253705978394, "global_step": 6807, "epoch": 76, "lr": 9.986632468907682e-05}
{"train_loss": 0.8438615202903748, "global_step": 6808, "epoch": 76, "lr": 9.986628232183283e-05}
{"train_loss": 0.8204555511474609, "global_step": 6809, "epoch": 76, "lr": 9.986623994788495e-05}
{"train_loss": 0.7924255728721619, "global_step": 6810, "epoch": 76, "lr": 9.986619756723317e-05}
{"train_loss": 0.7560397982597351, "global_step": 6811, "epoch": 76, "lr": 9.986615517987745e-05}
{"train_loss": 0.9341257214546204, "global_step": 6812, "epoch": 76, "lr": 9.986611278581786e-05}
{"train_loss": 0.8357941508293152, "global_step": 6813, "epoch": 76, "lr": 9.986607038505437e-05}
{"train_loss": 0.8771914839744568, "global_step": 6814, "epoch": 76, "lr": 9.9866027977587e-05}
{"train_loss": 0.9145227074623108, "global_step": 6815, "epoch": 76, "lr": 9.986598556341576e-05}
{"train_loss": 0.8123321533203125, "global_step": 6816, "epoch": 76, "lr": 9.986594314254062e-05}
{"train_loss": 0.9783599376678467, "global_step": 6817, "epoch": 76, "lr": 9.986590071496162e-05}
{"train_loss": 1.040557861328125, "global_step": 6818, "epoch": 76, "lr": 9.986585828067875e-05}
{"train_loss": 0.9031768441200256, "global_step": 6819, "epoch": 76, "lr": 9.986581583969206e-05}
{"train_loss": 0.7948583960533142, "global_step": 6820, "epoch": 76, "lr": 9.986577339200149e-05}
{"train_loss": 0.9296715259552002, "global_step": 6821, "epoch": 76, "lr": 9.986573093760707e-05}
{"train_loss": 0.852328896522522, "global_step": 6822, "epoch": 76, "lr": 9.986568847650883e-05}
{"train_loss": 0.8700077533721924, "global_step": 6823, "epoch": 76, "lr": 9.986564600870674e-05}
{"train_loss": 0.866173505783081, "global_step": 6824, "epoch": 76, "lr": 9.986560353420083e-05}
{"train_loss": 1.004172682762146, "global_step": 6825, "epoch": 76, "lr": 9.986556105299109e-05}
{"train_loss": 0.892587423324585, "global_step": 6826, "epoch": 76, "lr": 9.986551856507754e-05}
{"train_loss": 0.8590527772903442, "global_step": 6827, "epoch": 76, "lr": 9.986547607046017e-05}
{"train_loss": 0.913169264793396, "global_step": 6828, "epoch": 76, "lr": 9.9865433569139e-05}
{"train_loss": 0.8145495653152466, "global_step": 6829, "epoch": 76, "lr": 9.986539106111403e-05}
{"train_loss": 0.7715758681297302, "global_step": 6830, "epoch": 76, "lr": 9.986534854638527e-05}
{"train_loss": 0.8380360007286072, "global_step": 6831, "epoch": 76, "lr": 9.986530602495273e-05}
{"train_loss": 0.7969479560852051, "global_step": 6832, "epoch": 76, "lr": 9.986526349681637e-05}
{"train_loss": 1.0213404893875122, "global_step": 6833, "epoch": 76, "lr": 9.986522096197628e-05}
{"train_loss": 0.9479964971542358, "global_step": 6834, "epoch": 76, "lr": 9.986517842043239e-05}
{"train_loss": 0.6763466596603394, "global_step": 6835, "epoch": 76, "lr": 9.986513587218474e-05}
{"train_loss": 0.8113281726837158, "global_step": 6836, "epoch": 76, "lr": 9.986509331723333e-05}
{"train_loss": 0.8857157826423645, "global_step": 6837, "epoch": 76, "lr": 9.986505075557817e-05}
{"train_loss": 0.8836036920547485, "global_step": 6838, "epoch": 76, "lr": 9.986500818721926e-05}
{"train_loss": 0.8773580193519592, "global_step": 6839, "epoch": 76, "lr": 9.986496561215662e-05}
{"train_loss": 0.9518656134605408, "global_step": 6840, "epoch": 76, "lr": 9.986492303039022e-05}
{"train_loss": 0.8016445636749268, "global_step": 6841, "epoch": 76, "lr": 9.98648804419201e-05}
{"train_loss": 0.7845463752746582, "global_step": 6842, "epoch": 76, "lr": 9.986483784674625e-05}
{"train_loss": 0.879047691822052, "global_step": 6843, "epoch": 76, "lr": 9.986479524486869e-05}
{"train_loss": 0.8832156658172607, "global_step": 6844, "epoch": 76, "lr": 9.98647526362874e-05}
{"train_loss": 0.891482412815094, "global_step": 6845, "epoch": 76, "lr": 9.986471002100242e-05}
{"train_loss": 0.8626123070716858, "global_step": 6846, "epoch": 76, "lr": 9.986466739901372e-05}
{"train_loss": 0.8082754611968994, "global_step": 6847, "epoch": 76, "lr": 9.986462477032135e-05}
{"train_loss": 1.0756815671920776, "global_step": 6848, "epoch": 76, "lr": 9.986458213492527e-05}
{"train_loss": 0.800209641456604, "global_step": 6849, "epoch": 76, "lr": 9.98645394928255e-05}
{"train_loss": 0.9419731497764587, "global_step": 6850, "epoch": 76, "lr": 9.986449684402207e-05}
{"train_loss": 0.8320786356925964, "global_step": 6851, "epoch": 76, "lr": 9.986445418851493e-05}
{"train_loss": 0.8541348214899556, "global_step": 6852, "epoch": 76, "lr": 9.986441152630414e-05, "val_loss": 1.1597537994384766}
{"train_loss": 0.7977253198623657, "global_step": 6853, "epoch": 77, "lr": 9.98643688573897e-05}
{"train_loss": 0.7196584939956665, "global_step": 6854, "epoch": 77, "lr": 9.986432618177159e-05}
{"train_loss": 1.0147706270217896, "global_step": 6855, "epoch": 77, "lr": 9.986428349944985e-05}
{"train_loss": 0.9877288937568665, "global_step": 6856, "epoch": 77, "lr": 9.986424081042445e-05}
{"train_loss": 1.0101808309555054, "global_step": 6857, "epoch": 77, "lr": 9.98641981146954e-05}
{"train_loss": 0.7418436408042908, "global_step": 6858, "epoch": 77, "lr": 9.986415541226272e-05}
{"train_loss": 0.9343043565750122, "global_step": 6859, "epoch": 77, "lr": 9.986411270312643e-05}
{"train_loss": 0.7035237550735474, "global_step": 6860, "epoch": 77, "lr": 9.98640699872865e-05}
{"train_loss": 0.7377656698226929, "global_step": 6861, "epoch": 77, "lr": 9.986402726474297e-05}
{"train_loss": 0.9170051217079163, "global_step": 6862, "epoch": 77, "lr": 9.986398453549582e-05}
{"train_loss": 0.8165328502655029, "global_step": 6863, "epoch": 77, "lr": 9.986394179954505e-05}
{"train_loss": 0.962602436542511, "global_step": 6864, "epoch": 77, "lr": 9.98638990568907e-05}
{"train_loss": 0.8138008713722229, "global_step": 6865, "epoch": 77, "lr": 9.986385630753276e-05}
{"train_loss": 0.9433835744857788, "global_step": 6866, "epoch": 77, "lr": 9.986381355147123e-05}
{"train_loss": 0.8163196444511414, "global_step": 6867, "epoch": 77, "lr": 9.98637707887061e-05}
{"train_loss": 0.8857336044311523, "global_step": 6868, "epoch": 77, "lr": 9.986372801923741e-05}
{"train_loss": 0.8070326447486877, "global_step": 6869, "epoch": 77, "lr": 9.986368524306515e-05}
{"train_loss": 0.8110877275466919, "global_step": 6870, "epoch": 77, "lr": 9.986364246018933e-05}
{"train_loss": 0.7668579816818237, "global_step": 6871, "epoch": 77, "lr": 9.986359967060992e-05}
{"train_loss": 0.756485104560852, "global_step": 6872, "epoch": 77, "lr": 9.9863556874327e-05}
{"train_loss": 0.8885707855224609, "global_step": 6873, "epoch": 77, "lr": 9.98635140713405e-05}
{"train_loss": 1.0060983896255493, "global_step": 6874, "epoch": 77, "lr": 9.986347126165047e-05}
{"train_loss": 0.8133776783943176, "global_step": 6875, "epoch": 77, "lr": 9.98634284452569e-05}
{"train_loss": 0.8765407204627991, "global_step": 6876, "epoch": 77, "lr": 9.98633856221598e-05}
{"train_loss": 0.9202756285667419, "global_step": 6877, "epoch": 77, "lr": 9.986334279235918e-05}
{"train_loss": 0.8140937089920044, "global_step": 6878, "epoch": 77, "lr": 9.986329995585504e-05}
{"train_loss": 1.0070500373840332, "global_step": 6879, "epoch": 77, "lr": 9.986325711264739e-05}
{"train_loss": 0.8722591996192932, "global_step": 6880, "epoch": 77, "lr": 9.986321426273623e-05}
{"train_loss": 0.8667341470718384, "global_step": 6881, "epoch": 77, "lr": 9.986317140612157e-05}
{"train_loss": 0.7109910249710083, "global_step": 6882, "epoch": 77, "lr": 9.986312854280342e-05}
{"train_loss": 0.8760413527488708, "global_step": 6883, "epoch": 77, "lr": 9.986308567278176e-05}
{"train_loss": 0.779749870300293, "global_step": 6884, "epoch": 77, "lr": 9.986304279605663e-05}
{"train_loss": 0.7921046018600464, "global_step": 6885, "epoch": 77, "lr": 9.986299991262803e-05}
{"train_loss": 0.9137060642242432, "global_step": 6886, "epoch": 77, "lr": 9.986295702249595e-05}
{"train_loss": 0.7362877130508423, "global_step": 6887, "epoch": 77, "lr": 9.98629141256604e-05}
{"train_loss": 0.7279683947563171, "global_step": 6888, "epoch": 77, "lr": 9.98628712221214e-05}
{"train_loss": 0.9716112613677979, "global_step": 6889, "epoch": 77, "lr": 9.986282831187893e-05}
{"train_loss": 1.03678560256958, "global_step": 6890, "epoch": 77, "lr": 9.986278539493301e-05}
{"train_loss": 0.8012166023254395, "global_step": 6891, "epoch": 77, "lr": 9.986274247128365e-05}
{"train_loss": 0.8382524251937866, "global_step": 6892, "epoch": 77, "lr": 9.986269954093085e-05}
{"train_loss": 0.9123410582542419, "global_step": 6893, "epoch": 77, "lr": 9.986265660387461e-05}
{"train_loss": 0.8834488391876221, "global_step": 6894, "epoch": 77, "lr": 9.986261366011497e-05}
{"train_loss": 0.8816007375717163, "global_step": 6895, "epoch": 77, "lr": 9.986257070965189e-05}
{"train_loss": 0.8847036957740784, "global_step": 6896, "epoch": 77, "lr": 9.98625277524854e-05}
{"train_loss": 0.877457857131958, "global_step": 6897, "epoch": 77, "lr": 9.98624847886155e-05}
{"train_loss": 0.8379901647567749, "global_step": 6898, "epoch": 77, "lr": 9.98624418180422e-05}
{"train_loss": 0.7479461431503296, "global_step": 6899, "epoch": 77, "lr": 9.98623988407655e-05}
{"train_loss": 0.8264995217323303, "global_step": 6900, "epoch": 77, "lr": 9.98623558567854e-05}
{"train_loss": 0.8797199130058289, "global_step": 6901, "epoch": 77, "lr": 9.986231286610192e-05}
{"train_loss": 0.9020456075668335, "global_step": 6902, "epoch": 77, "lr": 9.986226986871506e-05}
{"train_loss": 0.8821452260017395, "global_step": 6903, "epoch": 77, "lr": 9.986222686462482e-05}
{"train_loss": 0.7378910183906555, "global_step": 6904, "epoch": 77, "lr": 9.986218385383123e-05}
{"train_loss": 0.7764769792556763, "global_step": 6905, "epoch": 77, "lr": 9.986214083633427e-05}
{"train_loss": 1.0013890266418457, "global_step": 6906, "epoch": 77, "lr": 9.986209781213394e-05}
{"train_loss": 0.8452242016792297, "global_step": 6907, "epoch": 77, "lr": 9.986205478123028e-05}
{"train_loss": 0.9110585451126099, "global_step": 6908, "epoch": 77, "lr": 9.986201174362325e-05}
{"train_loss": 0.8095069527626038, "global_step": 6909, "epoch": 77, "lr": 9.986196869931289e-05}
{"train_loss": 0.7939844727516174, "global_step": 6910, "epoch": 77, "lr": 9.986192564829919e-05}
{"train_loss": 0.8392086625099182, "global_step": 6911, "epoch": 77, "lr": 9.986188259058217e-05}
{"train_loss": 0.8514727354049683, "global_step": 6912, "epoch": 77, "lr": 9.986183952616184e-05}
{"train_loss": 0.9576929211616516, "global_step": 6913, "epoch": 77, "lr": 9.986179645503818e-05}
{"train_loss": 0.8385964035987854, "global_step": 6914, "epoch": 77, "lr": 9.986175337721121e-05}
{"train_loss": 1.0356653928756714, "global_step": 6915, "epoch": 77, "lr": 9.986171029268093e-05}
{"train_loss": 0.781649649143219, "global_step": 6916, "epoch": 77, "lr": 9.986166720144737e-05}
{"train_loss": 1.0267276763916016, "global_step": 6917, "epoch": 77, "lr": 9.986162410351049e-05}
{"train_loss": 0.9396976232528687, "global_step": 6918, "epoch": 77, "lr": 9.986158099887032e-05}
{"train_loss": 0.9358652234077454, "global_step": 6919, "epoch": 77, "lr": 9.98615378875269e-05}
{"train_loss": 0.8353981971740723, "global_step": 6920, "epoch": 77, "lr": 9.986149476948019e-05}
{"train_loss": 0.7645416855812073, "global_step": 6921, "epoch": 77, "lr": 9.98614516447302e-05}
{"train_loss": 0.772447943687439, "global_step": 6922, "epoch": 77, "lr": 9.986140851327694e-05}
{"train_loss": 0.9321328401565552, "global_step": 6923, "epoch": 77, "lr": 9.986136537512044e-05}
{"train_loss": 0.8988890051841736, "global_step": 6924, "epoch": 77, "lr": 9.986132223026067e-05}
{"train_loss": 0.8946566581726074, "global_step": 6925, "epoch": 77, "lr": 9.986127907869768e-05}
{"train_loss": 0.9246546626091003, "global_step": 6926, "epoch": 77, "lr": 9.986123592043142e-05}
{"train_loss": 0.9525552988052368, "global_step": 6927, "epoch": 77, "lr": 9.986119275546195e-05}
{"train_loss": 0.9410330653190613, "global_step": 6928, "epoch": 77, "lr": 9.986114958378923e-05}
{"train_loss": 0.716320276260376, "global_step": 6929, "epoch": 77, "lr": 9.98611064054133e-05}
{"train_loss": 0.7548965811729431, "global_step": 6930, "epoch": 77, "lr": 9.986106322033414e-05}
{"train_loss": 0.9257205128669739, "global_step": 6931, "epoch": 77, "lr": 9.986102002855177e-05}
{"train_loss": 0.9173975586891174, "global_step": 6932, "epoch": 77, "lr": 9.98609768300662e-05}
{"train_loss": 0.8720635771751404, "global_step": 6933, "epoch": 77, "lr": 9.986093362487742e-05}
{"train_loss": 0.838806688785553, "global_step": 6934, "epoch": 77, "lr": 9.986089041298546e-05}
{"train_loss": 0.8821056485176086, "global_step": 6935, "epoch": 77, "lr": 9.986084719439031e-05}
{"train_loss": 0.9809914827346802, "global_step": 6936, "epoch": 77, "lr": 9.986080396909195e-05}
{"train_loss": 1.0479830503463745, "global_step": 6937, "epoch": 77, "lr": 9.986076073709044e-05}
{"train_loss": 0.888001561164856, "global_step": 6938, "epoch": 77, "lr": 9.986071749838575e-05}
{"train_loss": 0.7278541326522827, "global_step": 6939, "epoch": 77, "lr": 9.98606742529779e-05}
{"train_loss": 0.7361311912536621, "global_step": 6940, "epoch": 77, "lr": 9.986063100086688e-05}
{"train_loss": 0.8613465381472298, "global_step": 6941, "epoch": 77, "lr": 9.986058774205271e-05, "val_loss": 1.117944359779358}
{"train_loss": 0.9549052119255066, "global_step": 6942, "epoch": 78, "lr": 9.986054447653539e-05}
{"train_loss": 0.9247298836708069, "global_step": 6943, "epoch": 78, "lr": 9.986050120431493e-05}
{"train_loss": 0.8199458122253418, "global_step": 6944, "epoch": 78, "lr": 9.986045792539133e-05}
{"train_loss": 0.7925000190734863, "global_step": 6945, "epoch": 78, "lr": 9.98604146397646e-05}
{"train_loss": 0.9966006875038147, "global_step": 6946, "epoch": 78, "lr": 9.986037134743475e-05}
{"train_loss": 0.8488203883171082, "global_step": 6947, "epoch": 78, "lr": 9.986032804840178e-05}
{"train_loss": 0.6978840827941895, "global_step": 6948, "epoch": 78, "lr": 9.98602847426657e-05}
{"train_loss": 0.7216681838035583, "global_step": 6949, "epoch": 78, "lr": 9.98602414302265e-05}
{"train_loss": 0.9531465768814087, "global_step": 6950, "epoch": 78, "lr": 9.98601981110842e-05}
{"train_loss": 0.7538909316062927, "global_step": 6951, "epoch": 78, "lr": 9.986015478523881e-05}
{"train_loss": 0.8986053466796875, "global_step": 6952, "epoch": 78, "lr": 9.986011145269032e-05}
{"train_loss": 0.7631548643112183, "global_step": 6953, "epoch": 78, "lr": 9.986006811343875e-05}
{"train_loss": 0.956932544708252, "global_step": 6954, "epoch": 78, "lr": 9.986002476748412e-05}
{"train_loss": 0.8364214301109314, "global_step": 6955, "epoch": 78, "lr": 9.98599814148264e-05}
{"train_loss": 0.8894475102424622, "global_step": 6956, "epoch": 78, "lr": 9.985993805546561e-05}
{"train_loss": 0.8462522029876709, "global_step": 6957, "epoch": 78, "lr": 9.985989468940178e-05}
{"train_loss": 0.9620730876922607, "global_step": 6958, "epoch": 78, "lr": 9.985985131663487e-05}
{"train_loss": 0.7233362197875977, "global_step": 6959, "epoch": 78, "lr": 9.985980793716492e-05}
{"train_loss": 0.7438431978225708, "global_step": 6960, "epoch": 78, "lr": 9.985976455099193e-05}
{"train_loss": 0.9558027386665344, "global_step": 6961, "epoch": 78, "lr": 9.985972115811591e-05}
{"train_loss": 0.9425017833709717, "global_step": 6962, "epoch": 78, "lr": 9.985967775853684e-05}
{"train_loss": 0.9192394614219666, "global_step": 6963, "epoch": 78, "lr": 9.985963435225476e-05}
{"train_loss": 0.7953306436538696, "global_step": 6964, "epoch": 78, "lr": 9.985959093926964e-05}
{"train_loss": 0.6831415295600891, "global_step": 6965, "epoch": 78, "lr": 9.985954751958152e-05}
{"train_loss": 0.8564939498901367, "global_step": 6966, "epoch": 78, "lr": 9.985950409319039e-05}
{"train_loss": 0.906046986579895, "global_step": 6967, "epoch": 78, "lr": 9.985946066009626e-05}
{"train_loss": 0.936616063117981, "global_step": 6968, "epoch": 78, "lr": 9.985941722029913e-05}
{"train_loss": 0.9125984907150269, "global_step": 6969, "epoch": 78, "lr": 9.985937377379901e-05}
{"train_loss": 0.9514999389648438, "global_step": 6970, "epoch": 78, "lr": 9.98593303205959e-05}
{"train_loss": 0.8511234521865845, "global_step": 6971, "epoch": 78, "lr": 9.985928686068982e-05}
{"train_loss": 0.7675771713256836, "global_step": 6972, "epoch": 78, "lr": 9.985924339408076e-05}
{"train_loss": 0.9367438554763794, "global_step": 6973, "epoch": 78, "lr": 9.985919992076874e-05}
{"train_loss": 0.8919933438301086, "global_step": 6974, "epoch": 78, "lr": 9.985915644075374e-05}
{"train_loss": 0.666903018951416, "global_step": 6975, "epoch": 78, "lr": 9.985911295403581e-05}
{"train_loss": 0.8612822890281677, "global_step": 6976, "epoch": 78, "lr": 9.98590694606149e-05}
{"train_loss": 0.9763432145118713, "global_step": 6977, "epoch": 78, "lr": 9.985902596049106e-05}
{"train_loss": 0.8988359570503235, "global_step": 6978, "epoch": 78, "lr": 9.98589824536643e-05}
{"train_loss": 0.8639095425605774, "global_step": 6979, "epoch": 78, "lr": 9.985893894013458e-05}
{"train_loss": 0.750158965587616, "global_step": 6980, "epoch": 78, "lr": 9.985889541990194e-05}
{"train_loss": 1.1138609647750854, "global_step": 6981, "epoch": 78, "lr": 9.985885189296638e-05}
{"train_loss": 0.7273762822151184, "global_step": 6982, "epoch": 78, "lr": 9.985880835932791e-05}
{"train_loss": 0.8034144639968872, "global_step": 6983, "epoch": 78, "lr": 9.985876481898654e-05}
{"train_loss": 0.9774419665336609, "global_step": 6984, "epoch": 78, "lr": 9.985872127194224e-05}
{"train_loss": 0.9722395539283752, "global_step": 6985, "epoch": 78, "lr": 9.985867771819507e-05}
{"train_loss": 0.7335770130157471, "global_step": 6986, "epoch": 78, "lr": 9.9858634157745e-05}
{"train_loss": 0.8234907984733582, "global_step": 6987, "epoch": 78, "lr": 9.985859059059203e-05}
{"train_loss": 0.8419094085693359, "global_step": 6988, "epoch": 78, "lr": 9.985854701673621e-05}
{"train_loss": 0.8058127760887146, "global_step": 6989, "epoch": 78, "lr": 9.985850343617749e-05}
{"train_loss": 0.9931784272193909, "global_step": 6990, "epoch": 78, "lr": 9.985845984891591e-05}
{"train_loss": 0.9864360094070435, "global_step": 6991, "epoch": 78, "lr": 9.985841625495147e-05}
{"train_loss": 0.9150152802467346, "global_step": 6992, "epoch": 78, "lr": 9.985837265428417e-05}
{"train_loss": 0.8760724067687988, "global_step": 6993, "epoch": 78, "lr": 9.985832904691402e-05}
{"train_loss": 0.7763466238975525, "global_step": 6994, "epoch": 78, "lr": 9.985828543284102e-05}
{"train_loss": 0.7090178728103638, "global_step": 6995, "epoch": 78, "lr": 9.98582418120652e-05}
{"train_loss": 0.8126470446586609, "global_step": 6996, "epoch": 78, "lr": 9.985819818458653e-05}
{"train_loss": 1.0399378538131714, "global_step": 6997, "epoch": 78, "lr": 9.985815455040503e-05}
{"train_loss": 0.7478119730949402, "global_step": 6998, "epoch": 78, "lr": 9.985811090952072e-05}
{"train_loss": 0.7518066167831421, "global_step": 6999, "epoch": 78, "lr": 9.985806726193359e-05}
{"train_loss": 0.7364529967308044, "global_step": 7000, "epoch": 78, "lr": 9.985802360764365e-05}
{"train_loss": 0.8325701951980591, "global_step": 7001, "epoch": 78, "lr": 9.985797994665091e-05}
{"train_loss": 0.966116189956665, "global_step": 7002, "epoch": 78, "lr": 9.985793627895537e-05}
{"train_loss": 1.1094270944595337, "global_step": 7003, "epoch": 78, "lr": 9.985789260455704e-05}
{"train_loss": 0.7377744913101196, "global_step": 7004, "epoch": 78, "lr": 9.985784892345592e-05}
{"train_loss": 0.8253152966499329, "global_step": 7005, "epoch": 78, "lr": 9.985780523565202e-05}
{"train_loss": 0.8017984628677368, "global_step": 7006, "epoch": 78, "lr": 9.985776154114535e-05}
{"train_loss": 0.9033370614051819, "global_step": 7007, "epoch": 78, "lr": 9.985771783993592e-05}
{"train_loss": 1.0303242206573486, "global_step": 7008, "epoch": 78, "lr": 9.98576741320237e-05}
{"train_loss": 0.7181363701820374, "global_step": 7009, "epoch": 78, "lr": 9.985763041740875e-05}
{"train_loss": 0.9665076732635498, "global_step": 7010, "epoch": 78, "lr": 9.985758669609105e-05}
{"train_loss": 0.8885114192962646, "global_step": 7011, "epoch": 78, "lr": 9.985754296807059e-05}
{"train_loss": 0.9077447056770325, "global_step": 7012, "epoch": 78, "lr": 9.985749923334739e-05}
{"train_loss": 0.9839645028114319, "global_step": 7013, "epoch": 78, "lr": 9.985745549192146e-05}
{"train_loss": 0.7954614162445068, "global_step": 7014, "epoch": 78, "lr": 9.98574117437928e-05}
{"train_loss": 0.8571561574935913, "global_step": 7015, "epoch": 78, "lr": 9.985736798896144e-05}
{"train_loss": 0.7513183951377869, "global_step": 7016, "epoch": 78, "lr": 9.985732422742734e-05}
{"train_loss": 1.0408838987350464, "global_step": 7017, "epoch": 78, "lr": 9.985728045919054e-05}
{"train_loss": 0.852239727973938, "global_step": 7018, "epoch": 78, "lr": 9.985723668425103e-05}
{"train_loss": 0.7235120534896851, "global_step": 7019, "epoch": 78, "lr": 9.985719290260883e-05}
{"train_loss": 0.7477914094924927, "global_step": 7020, "epoch": 78, "lr": 9.985714911426394e-05}
{"train_loss": 0.8526081442832947, "global_step": 7021, "epoch": 78, "lr": 9.985710531921635e-05}
{"train_loss": 0.8368329405784607, "global_step": 7022, "epoch": 78, "lr": 9.98570615174661e-05}
{"train_loss": 0.8928723335266113, "global_step": 7023, "epoch": 78, "lr": 9.985701770901317e-05}
{"train_loss": 0.7037533521652222, "global_step": 7024, "epoch": 78, "lr": 9.985697389385755e-05}
{"train_loss": 0.7700788378715515, "global_step": 7025, "epoch": 78, "lr": 9.98569300719993e-05}
{"train_loss": 0.9933471083641052, "global_step": 7026, "epoch": 78, "lr": 9.985688624343836e-05}
{"train_loss": 0.6180863976478577, "global_step": 7027, "epoch": 78, "lr": 9.98568424081748e-05}
{"train_loss": 0.768871009349823, "global_step": 7028, "epoch": 78, "lr": 9.985679856620858e-05}
{"train_loss": 0.9000409841537476, "global_step": 7029, "epoch": 78, "lr": 9.985675471753973e-05}
{"train_loss": 0.8577548359217269, "global_step": 7030, "epoch": 78, "lr": 9.985671086216823e-05, "val_loss": 1.2014576196670532}
{"train_loss": 0.9030807018280029, "global_step": 7031, "epoch": 79, "lr": 9.985666700009411e-05}
{"train_loss": 0.8354310393333435, "global_step": 7032, "epoch": 79, "lr": 9.985662313131738e-05}
{"train_loss": 0.8177936673164368, "global_step": 7033, "epoch": 79, "lr": 9.985657925583803e-05}
{"train_loss": 1.1031129360198975, "global_step": 7034, "epoch": 79, "lr": 9.985653537365607e-05}
{"train_loss": 0.8943422436714172, "global_step": 7035, "epoch": 79, "lr": 9.98564914847715e-05}
{"train_loss": 0.8047983050346375, "global_step": 7036, "epoch": 79, "lr": 9.985644758918433e-05}
{"train_loss": 1.1295734643936157, "global_step": 7037, "epoch": 79, "lr": 9.985640368689457e-05}
{"train_loss": 0.8706558346748352, "global_step": 7038, "epoch": 79, "lr": 9.985635977790224e-05}
{"train_loss": 0.8844597339630127, "global_step": 7039, "epoch": 79, "lr": 9.985631586220732e-05}
{"train_loss": 0.9391955733299255, "global_step": 7040, "epoch": 79, "lr": 9.985627193980982e-05}
{"train_loss": 0.910632312297821, "global_step": 7041, "epoch": 79, "lr": 9.985622801070975e-05}
{"train_loss": 0.8156608939170837, "global_step": 7042, "epoch": 79, "lr": 9.985618407490713e-05}
{"train_loss": 0.8850210905075073, "global_step": 7043, "epoch": 79, "lr": 9.985614013240194e-05}
{"train_loss": 0.8702682852745056, "global_step": 7044, "epoch": 79, "lr": 9.985609618319422e-05}
{"train_loss": 0.8634074926376343, "global_step": 7045, "epoch": 79, "lr": 9.985605222728394e-05}
{"train_loss": 0.9039945006370544, "global_step": 7046, "epoch": 79, "lr": 9.985600826467112e-05}
{"train_loss": 0.8547679781913757, "global_step": 7047, "epoch": 79, "lr": 9.985596429535578e-05}
{"train_loss": 0.72539883852005, "global_step": 7048, "epoch": 79, "lr": 9.985592031933791e-05}
{"train_loss": 0.9288108348846436, "global_step": 7049, "epoch": 79, "lr": 9.98558763366175e-05}
{"train_loss": 0.7112772464752197, "global_step": 7050, "epoch": 79, "lr": 9.98558323471946e-05}
{"train_loss": 0.908276379108429, "global_step": 7051, "epoch": 79, "lr": 9.985578835106918e-05}
{"train_loss": 0.7694587111473083, "global_step": 7052, "epoch": 79, "lr": 9.985574434824126e-05}
{"train_loss": 0.9190186262130737, "global_step": 7053, "epoch": 79, "lr": 9.985570033871084e-05}
{"train_loss": 0.9660409688949585, "global_step": 7054, "epoch": 79, "lr": 9.985565632247793e-05}
{"train_loss": 0.7769328355789185, "global_step": 7055, "epoch": 79, "lr": 9.985561229954253e-05}
{"train_loss": 0.7123624086380005, "global_step": 7056, "epoch": 79, "lr": 9.985556826990466e-05}
{"train_loss": 0.755730390548706, "global_step": 7057, "epoch": 79, "lr": 9.98555242335643e-05}
{"train_loss": 0.8790960311889648, "global_step": 7058, "epoch": 79, "lr": 9.985548019052149e-05}
{"train_loss": 1.060265302658081, "global_step": 7059, "epoch": 79, "lr": 9.985543614077622e-05}
{"train_loss": 0.9997918605804443, "global_step": 7060, "epoch": 79, "lr": 9.985539208432848e-05}
{"train_loss": 1.0072612762451172, "global_step": 7061, "epoch": 79, "lr": 9.98553480211783e-05}
{"train_loss": 1.034123420715332, "global_step": 7062, "epoch": 79, "lr": 9.985530395132567e-05}
{"train_loss": 0.7513335943222046, "global_step": 7063, "epoch": 79, "lr": 9.98552598747706e-05}
{"train_loss": 0.9432384967803955, "global_step": 7064, "epoch": 79, "lr": 9.98552157915131e-05}
{"train_loss": 0.9048657417297363, "global_step": 7065, "epoch": 79, "lr": 9.985517170155319e-05}
{"train_loss": 0.8700640797615051, "global_step": 7066, "epoch": 79, "lr": 9.985512760489085e-05}
{"train_loss": 1.0254155397415161, "global_step": 7067, "epoch": 79, "lr": 9.985508350152609e-05}
{"train_loss": 0.8493322134017944, "global_step": 7068, "epoch": 79, "lr": 9.985503939145891e-05}
{"train_loss": 0.84068363904953, "global_step": 7069, "epoch": 79, "lr": 9.985499527468935e-05}
{"train_loss": 0.9483858346939087, "global_step": 7070, "epoch": 79, "lr": 9.985495115121737e-05}
{"train_loss": 0.7608922123908997, "global_step": 7071, "epoch": 79, "lr": 9.985490702104303e-05}
{"train_loss": 0.9153768420219421, "global_step": 7072, "epoch": 79, "lr": 9.985486288416628e-05}
{"train_loss": 1.1179176568984985, "global_step": 7073, "epoch": 79, "lr": 9.985481874058716e-05}
{"train_loss": 0.9236013889312744, "global_step": 7074, "epoch": 79, "lr": 9.985477459030567e-05}
{"train_loss": 0.7756431698799133, "global_step": 7075, "epoch": 79, "lr": 9.98547304333218e-05}
{"train_loss": 0.8242431282997131, "global_step": 7076, "epoch": 79, "lr": 9.985468626963558e-05}
{"train_loss": 0.9252806901931763, "global_step": 7077, "epoch": 79, "lr": 9.985464209924702e-05}
{"train_loss": 0.7342908978462219, "global_step": 7078, "epoch": 79, "lr": 9.985459792215609e-05}
{"train_loss": 0.9052254557609558, "global_step": 7079, "epoch": 79, "lr": 9.985455373836283e-05}
{"train_loss": 0.9481141567230225, "global_step": 7080, "epoch": 79, "lr": 9.985450954786721e-05}
{"train_loss": 0.802457869052887, "global_step": 7081, "epoch": 79, "lr": 9.985446535066929e-05}
{"train_loss": 0.9946433305740356, "global_step": 7082, "epoch": 79, "lr": 9.985442114676903e-05}
{"train_loss": 0.767525315284729, "global_step": 7083, "epoch": 79, "lr": 9.985437693616645e-05}
{"train_loss": 1.0054599046707153, "global_step": 7084, "epoch": 79, "lr": 9.985433271886156e-05}
{"train_loss": 0.8428481817245483, "global_step": 7085, "epoch": 79, "lr": 9.985428849485436e-05}
{"train_loss": 0.7695767283439636, "global_step": 7086, "epoch": 79, "lr": 9.985424426414485e-05}
{"train_loss": 0.8596237301826477, "global_step": 7087, "epoch": 79, "lr": 9.985420002673306e-05}
{"train_loss": 1.039831519126892, "global_step": 7088, "epoch": 79, "lr": 9.985415578261897e-05}
{"train_loss": 0.9516395926475525, "global_step": 7089, "epoch": 79, "lr": 9.98541115318026e-05}
{"train_loss": 0.8463948965072632, "global_step": 7090, "epoch": 79, "lr": 9.985406727428395e-05}
{"train_loss": 0.9551003575325012, "global_step": 7091, "epoch": 79, "lr": 9.985402301006304e-05}
{"train_loss": 0.6271958947181702, "global_step": 7092, "epoch": 79, "lr": 9.985397873913984e-05}
{"train_loss": 0.9900555610656738, "global_step": 7093, "epoch": 79, "lr": 9.985393446151441e-05}
{"train_loss": 0.8311202526092529, "global_step": 7094, "epoch": 79, "lr": 9.98538901771867e-05}
{"train_loss": 0.9502323269844055, "global_step": 7095, "epoch": 79, "lr": 9.985384588615675e-05}
{"train_loss": 1.000852346420288, "global_step": 7096, "epoch": 79, "lr": 9.985380158842456e-05}
{"train_loss": 0.7805631160736084, "global_step": 7097, "epoch": 79, "lr": 9.985375728399014e-05}
{"train_loss": 0.6019081473350525, "global_step": 7098, "epoch": 79, "lr": 9.985371297285348e-05}
{"train_loss": 0.8594130873680115, "global_step": 7099, "epoch": 79, "lr": 9.985366865501461e-05}
{"train_loss": 0.7634321451187134, "global_step": 7100, "epoch": 79, "lr": 9.985362433047352e-05}
{"train_loss": 0.6634422540664673, "global_step": 7101, "epoch": 79, "lr": 9.98535799992302e-05}
{"train_loss": 0.7375353574752808, "global_step": 7102, "epoch": 79, "lr": 9.985353566128468e-05}
{"train_loss": 0.7911378145217896, "global_step": 7103, "epoch": 79, "lr": 9.985349131663697e-05}
{"train_loss": 0.7044805884361267, "global_step": 7104, "epoch": 79, "lr": 9.985344696528707e-05}
{"train_loss": 0.7985957264900208, "global_step": 7105, "epoch": 79, "lr": 9.985340260723497e-05}
{"train_loss": 0.8505155444145203, "global_step": 7106, "epoch": 79, "lr": 9.985335824248068e-05}
{"train_loss": 0.946215033531189, "global_step": 7107, "epoch": 79, "lr": 9.985331387102422e-05}
{"train_loss": 0.8493297100067139, "global_step": 7108, "epoch": 79, "lr": 9.98532694928656e-05}
{"train_loss": 0.946980893611908, "global_step": 7109, "epoch": 79, "lr": 9.985322510800481e-05}
{"train_loss": 0.9714123606681824, "global_step": 7110, "epoch": 79, "lr": 9.985318071644185e-05}
{"train_loss": 0.7131974101066589, "global_step": 7111, "epoch": 79, "lr": 9.985313631817675e-05}
{"train_loss": 0.8436791896820068, "global_step": 7112, "epoch": 79, "lr": 9.98530919132095e-05}
{"train_loss": 0.8238177299499512, "global_step": 7113, "epoch": 79, "lr": 9.985304750154012e-05}
{"train_loss": 0.744875967502594, "global_step": 7114, "epoch": 79, "lr": 9.985300308316859e-05}
{"train_loss": 0.7637490034103394, "global_step": 7115, "epoch": 79, "lr": 9.985295865809494e-05}
{"train_loss": 0.8078071475028992, "global_step": 7116, "epoch": 79, "lr": 9.985291422631917e-05}
{"train_loss": 0.7382941842079163, "global_step": 7117, "epoch": 79, "lr": 9.985286978784127e-05}
{"train_loss": 0.804696798324585, "global_step": 7118, "epoch": 79, "lr": 9.985282534266127e-05}
{"train_loss": 0.8655228648292884, "global_step": 7119, "epoch": 79, "lr": 9.985278089077917e-05, "val_loss": 1.1591055393218994}
{"train_loss": 0.8040913939476013, "global_step": 7120, "epoch": 80, "lr": 9.985273643219495e-05}
{"train_loss": 0.9264664649963379, "global_step": 7121, "epoch": 80, "lr": 9.985269196690865e-05}
{"train_loss": 0.6611582636833191, "global_step": 7122, "epoch": 80, "lr": 9.985264749492026e-05}
{"train_loss": 0.7797499895095825, "global_step": 7123, "epoch": 80, "lr": 9.98526030162298e-05}
{"train_loss": 0.8401495218276978, "global_step": 7124, "epoch": 80, "lr": 9.985255853083725e-05}
{"train_loss": 0.7936885952949524, "global_step": 7125, "epoch": 80, "lr": 9.985251403874263e-05}
{"train_loss": 0.7525783181190491, "global_step": 7126, "epoch": 80, "lr": 9.985246953994597e-05}
{"train_loss": 0.7875915765762329, "global_step": 7127, "epoch": 80, "lr": 9.985242503444725e-05}
{"train_loss": 0.7911261916160583, "global_step": 7128, "epoch": 80, "lr": 9.985238052224646e-05}
{"train_loss": 0.920828640460968, "global_step": 7129, "epoch": 80, "lr": 9.985233600334363e-05}
{"train_loss": 0.8587402105331421, "global_step": 7130, "epoch": 80, "lr": 9.985229147773876e-05}
{"train_loss": 0.7206117510795593, "global_step": 7131, "epoch": 80, "lr": 9.985224694543187e-05}
{"train_loss": 0.6833920478820801, "global_step": 7132, "epoch": 80, "lr": 9.985220240642294e-05}
{"train_loss": 0.893451452255249, "global_step": 7133, "epoch": 80, "lr": 9.985215786071199e-05}
{"train_loss": 0.7283288240432739, "global_step": 7134, "epoch": 80, "lr": 9.985211330829903e-05}
{"train_loss": 0.9228775501251221, "global_step": 7135, "epoch": 80, "lr": 9.985206874918406e-05}
{"train_loss": 0.8412091135978699, "global_step": 7136, "epoch": 80, "lr": 9.985202418336709e-05}
{"train_loss": 0.8477609157562256, "global_step": 7137, "epoch": 80, "lr": 9.985197961084812e-05}
{"train_loss": 1.0149962902069092, "global_step": 7138, "epoch": 80, "lr": 9.985193503162715e-05}
{"train_loss": 0.7894918918609619, "global_step": 7139, "epoch": 80, "lr": 9.98518904457042e-05}
{"train_loss": 0.7909353971481323, "global_step": 7140, "epoch": 80, "lr": 9.985184585307927e-05}
{"train_loss": 0.980277955532074, "global_step": 7141, "epoch": 80, "lr": 9.985180125375237e-05}
{"train_loss": 0.9192052483558655, "global_step": 7142, "epoch": 80, "lr": 9.985175664772351e-05}
{"train_loss": 0.8168514966964722, "global_step": 7143, "epoch": 80, "lr": 9.985171203499268e-05}
{"train_loss": 0.8182063102722168, "global_step": 7144, "epoch": 80, "lr": 9.985166741555991e-05}
{"train_loss": 0.8677042722702026, "global_step": 7145, "epoch": 80, "lr": 9.985162278942516e-05}
{"train_loss": 0.8512029051780701, "global_step": 7146, "epoch": 80, "lr": 9.98515781565885e-05}
{"train_loss": 0.9239487051963806, "global_step": 7147, "epoch": 80, "lr": 9.985153351704989e-05}
{"train_loss": 0.8274593949317932, "global_step": 7148, "epoch": 80, "lr": 9.985148887080936e-05}
{"train_loss": 0.8796589970588684, "global_step": 7149, "epoch": 80, "lr": 9.985144421786688e-05}
{"train_loss": 0.8870373964309692, "global_step": 7150, "epoch": 80, "lr": 9.985139955822249e-05}
{"train_loss": 0.7688384652137756, "global_step": 7151, "epoch": 80, "lr": 9.98513548918762e-05}
{"train_loss": 0.7532083988189697, "global_step": 7152, "epoch": 80, "lr": 9.985131021882799e-05}
{"train_loss": 0.8782116174697876, "global_step": 7153, "epoch": 80, "lr": 9.985126553907787e-05}
{"train_loss": 0.9567752480506897, "global_step": 7154, "epoch": 80, "lr": 9.985122085262588e-05}
{"train_loss": 0.789588987827301, "global_step": 7155, "epoch": 80, "lr": 9.985117615947198e-05}
{"train_loss": 0.8124204874038696, "global_step": 7156, "epoch": 80, "lr": 9.985113145961621e-05}
{"train_loss": 0.9104528427124023, "global_step": 7157, "epoch": 80, "lr": 9.985108675305856e-05}
{"train_loss": 0.9566254019737244, "global_step": 7158, "epoch": 80, "lr": 9.985104203979903e-05}
{"train_loss": 0.7283206582069397, "global_step": 7159, "epoch": 80, "lr": 9.985099731983765e-05}
{"train_loss": 0.8494809865951538, "global_step": 7160, "epoch": 80, "lr": 9.98509525931744e-05}
{"train_loss": 0.8772462010383606, "global_step": 7161, "epoch": 80, "lr": 9.985090785980929e-05}
{"train_loss": 0.7122504711151123, "global_step": 7162, "epoch": 80, "lr": 9.985086311974235e-05}
{"train_loss": 0.8617292642593384, "global_step": 7163, "epoch": 80, "lr": 9.985081837297357e-05}
{"train_loss": 0.8448414206504822, "global_step": 7164, "epoch": 80, "lr": 9.985077361950295e-05}
{"train_loss": 0.8062815070152283, "global_step": 7165, "epoch": 80, "lr": 9.985072885933049e-05}
{"train_loss": 0.6738584637641907, "global_step": 7166, "epoch": 80, "lr": 9.985068409245621e-05}
{"train_loss": 0.8247370719909668, "global_step": 7167, "epoch": 80, "lr": 9.985063931888012e-05}
{"train_loss": 0.9598609805107117, "global_step": 7168, "epoch": 80, "lr": 9.985059453860222e-05}
{"train_loss": 0.8203906416893005, "global_step": 7169, "epoch": 80, "lr": 9.985054975162251e-05}
{"train_loss": 0.8176485896110535, "global_step": 7170, "epoch": 80, "lr": 9.985050495794101e-05}
{"train_loss": 0.8969258666038513, "global_step": 7171, "epoch": 80, "lr": 9.98504601575577e-05}
{"train_loss": 1.0018266439437866, "global_step": 7172, "epoch": 80, "lr": 9.985041535047262e-05}
{"train_loss": 0.8081527352333069, "global_step": 7173, "epoch": 80, "lr": 9.985037053668575e-05}
{"train_loss": 0.8175028562545776, "global_step": 7174, "epoch": 80, "lr": 9.985032571619711e-05}
{"train_loss": 0.8889048099517822, "global_step": 7175, "epoch": 80, "lr": 9.98502808890067e-05}
{"train_loss": 0.8643577694892883, "global_step": 7176, "epoch": 80, "lr": 9.985023605511453e-05}
{"train_loss": 0.959022581577301, "global_step": 7177, "epoch": 80, "lr": 9.985019121452059e-05}
{"train_loss": 0.8447101712226868, "global_step": 7178, "epoch": 80, "lr": 9.985014636722491e-05}
{"train_loss": 0.7249916791915894, "global_step": 7179, "epoch": 80, "lr": 9.985010151322749e-05}
{"train_loss": 0.9250370860099792, "global_step": 7180, "epoch": 80, "lr": 9.985005665252833e-05}
{"train_loss": 0.7839956283569336, "global_step": 7181, "epoch": 80, "lr": 9.985001178512744e-05}
{"train_loss": 0.9764109253883362, "global_step": 7182, "epoch": 80, "lr": 9.984996691102481e-05}
{"train_loss": 0.9865370392799377, "global_step": 7183, "epoch": 80, "lr": 9.984992203022048e-05}
{"train_loss": 0.9963880181312561, "global_step": 7184, "epoch": 80, "lr": 9.98498771427144e-05}
{"train_loss": 0.825512170791626, "global_step": 7185, "epoch": 80, "lr": 9.984983224850665e-05}
{"train_loss": 0.9184051752090454, "global_step": 7186, "epoch": 80, "lr": 9.984978734759718e-05}
{"train_loss": 0.8146550059318542, "global_step": 7187, "epoch": 80, "lr": 9.984974243998602e-05}
{"train_loss": 0.9213662147521973, "global_step": 7188, "epoch": 80, "lr": 9.984969752567317e-05}
{"train_loss": 0.9232085943222046, "global_step": 7189, "epoch": 80, "lr": 9.984965260465863e-05}
{"train_loss": 0.9119598269462585, "global_step": 7190, "epoch": 80, "lr": 9.984960767694242e-05}
{"train_loss": 0.7974396347999573, "global_step": 7191, "epoch": 80, "lr": 9.984956274252452e-05}
{"train_loss": 0.799068808555603, "global_step": 7192, "epoch": 80, "lr": 9.984951780140497e-05}
{"train_loss": 0.9354878067970276, "global_step": 7193, "epoch": 80, "lr": 9.984947285358375e-05}
{"train_loss": 0.8258464932441711, "global_step": 7194, "epoch": 80, "lr": 9.984942789906088e-05}
{"train_loss": 1.0009963512420654, "global_step": 7195, "epoch": 80, "lr": 9.984938293783636e-05}
{"train_loss": 0.8728641867637634, "global_step": 7196, "epoch": 80, "lr": 9.984933796991021e-05}
{"train_loss": 0.9467521905899048, "global_step": 7197, "epoch": 80, "lr": 9.98492929952824e-05}
{"train_loss": 0.9064067006111145, "global_step": 7198, "epoch": 80, "lr": 9.984924801395299e-05}
{"train_loss": 0.8924536108970642, "global_step": 7199, "epoch": 80, "lr": 9.984920302592193e-05}
{"train_loss": 0.7710352540016174, "global_step": 7200, "epoch": 80, "lr": 9.984915803118928e-05}
{"train_loss": 0.9509854316711426, "global_step": 7201, "epoch": 80, "lr": 9.984911302975499e-05}
{"train_loss": 0.8206298351287842, "global_step": 7202, "epoch": 80, "lr": 9.984906802161911e-05}
{"train_loss": 0.9599852561950684, "global_step": 7203, "epoch": 80, "lr": 9.984902300678163e-05}
{"train_loss": 0.9496808052062988, "global_step": 7204, "epoch": 80, "lr": 9.984897798524255e-05}
{"train_loss": 0.6542004942893982, "global_step": 7205, "epoch": 80, "lr": 9.984893295700188e-05}
{"train_loss": 0.8824823498725891, "global_step": 7206, "epoch": 80, "lr": 9.984888792205965e-05}
{"train_loss": 1.0214442014694214, "global_step": 7207, "epoch": 80, "lr": 9.984884288041581e-05}
{"train_loss": 0.8555845102567351, "global_step": 7208, "epoch": 80, "lr": 9.984879783207043e-05, "val_loss": 1.1315854787826538, "train_action_mse_error": 44.522064208984375}
{"train_loss": 0.8401248455047607, "global_step": 7209, "epoch": 81, "lr": 9.984875277702348e-05}
{"train_loss": 0.7511652708053589, "global_step": 7210, "epoch": 81, "lr": 9.984870771527497e-05}
{"train_loss": 0.897462785243988, "global_step": 7211, "epoch": 81, "lr": 9.98486626468249e-05}
{"train_loss": 1.0184377431869507, "global_step": 7212, "epoch": 81, "lr": 9.98486175716733e-05}
{"train_loss": 0.8322059512138367, "global_step": 7213, "epoch": 81, "lr": 9.984857248982014e-05}
{"train_loss": 0.9362156987190247, "global_step": 7214, "epoch": 81, "lr": 9.984852740126547e-05}
{"train_loss": 0.9734911918640137, "global_step": 7215, "epoch": 81, "lr": 9.984848230600926e-05}
{"train_loss": 0.8401070833206177, "global_step": 7216, "epoch": 81, "lr": 9.984843720405152e-05}
{"train_loss": 0.7809852361679077, "global_step": 7217, "epoch": 81, "lr": 9.98483920953923e-05}
{"train_loss": 0.7428888082504272, "global_step": 7218, "epoch": 81, "lr": 9.984834698003154e-05}
{"train_loss": 0.8640958070755005, "global_step": 7219, "epoch": 81, "lr": 9.984830185796928e-05}
{"train_loss": 0.8104153871536255, "global_step": 7220, "epoch": 81, "lr": 9.984825672920553e-05}
{"train_loss": 0.853622317314148, "global_step": 7221, "epoch": 81, "lr": 9.984821159374029e-05}
{"train_loss": 0.706821858882904, "global_step": 7222, "epoch": 81, "lr": 9.984816645157356e-05}
{"train_loss": 0.8694520592689514, "global_step": 7223, "epoch": 81, "lr": 9.984812130270536e-05}
{"train_loss": 0.820107638835907, "global_step": 7224, "epoch": 81, "lr": 9.98480761471357e-05}
{"train_loss": 1.1050355434417725, "global_step": 7225, "epoch": 81, "lr": 9.984803098486455e-05}
{"train_loss": 0.8397214412689209, "global_step": 7226, "epoch": 81, "lr": 9.984798581589194e-05}
{"train_loss": 0.8962135314941406, "global_step": 7227, "epoch": 81, "lr": 9.984794064021788e-05}
{"train_loss": 0.6755605340003967, "global_step": 7228, "epoch": 81, "lr": 9.984789545784239e-05}
{"train_loss": 0.896166980266571, "global_step": 7229, "epoch": 81, "lr": 9.984785026876544e-05}
{"train_loss": 0.7837719917297363, "global_step": 7230, "epoch": 81, "lr": 9.984780507298706e-05}
{"train_loss": 0.9908391833305359, "global_step": 7231, "epoch": 81, "lr": 9.984775987050724e-05}
{"train_loss": 0.7512180805206299, "global_step": 7232, "epoch": 81, "lr": 9.9847714661326e-05}
{"train_loss": 0.9749654531478882, "global_step": 7233, "epoch": 81, "lr": 9.984766944544334e-05}
{"train_loss": 0.9235095977783203, "global_step": 7234, "epoch": 81, "lr": 9.984762422285928e-05}
{"train_loss": 0.8816815614700317, "global_step": 7235, "epoch": 81, "lr": 9.984757899357381e-05}
{"train_loss": 0.9027841687202454, "global_step": 7236, "epoch": 81, "lr": 9.984753375758694e-05}
{"train_loss": 0.7933850884437561, "global_step": 7237, "epoch": 81, "lr": 9.984748851489868e-05}
{"train_loss": 0.7108834385871887, "global_step": 7238, "epoch": 81, "lr": 9.984744326550904e-05}
{"train_loss": 0.9500476121902466, "global_step": 7239, "epoch": 81, "lr": 9.9847398009418e-05}
{"train_loss": 0.7354970574378967, "global_step": 7240, "epoch": 81, "lr": 9.984735274662558e-05}
{"train_loss": 0.7647703886032104, "global_step": 7241, "epoch": 81, "lr": 9.98473074771318e-05}
{"train_loss": 0.7883858680725098, "global_step": 7242, "epoch": 81, "lr": 9.984726220093667e-05}
{"train_loss": 0.7614237666130066, "global_step": 7243, "epoch": 81, "lr": 9.984721691804018e-05}
{"train_loss": 0.852240264415741, "global_step": 7244, "epoch": 81, "lr": 9.984717162844233e-05}
{"train_loss": 0.8210409879684448, "global_step": 7245, "epoch": 81, "lr": 9.984712633214313e-05}
{"train_loss": 0.8643428087234497, "global_step": 7246, "epoch": 81, "lr": 9.984708102914259e-05}
{"train_loss": 0.9153215289115906, "global_step": 7247, "epoch": 81, "lr": 9.984703571944075e-05}
{"train_loss": 0.6184055805206299, "global_step": 7248, "epoch": 81, "lr": 9.984699040303756e-05}
{"train_loss": 0.8869096636772156, "global_step": 7249, "epoch": 81, "lr": 9.984694507993305e-05}
{"train_loss": 0.9292749762535095, "global_step": 7250, "epoch": 81, "lr": 9.984689975012722e-05}
{"train_loss": 0.918980598449707, "global_step": 7251, "epoch": 81, "lr": 9.984685441362009e-05}
{"train_loss": 0.8212229013442993, "global_step": 7252, "epoch": 81, "lr": 9.984680907041165e-05}
{"train_loss": 0.7455560564994812, "global_step": 7253, "epoch": 81, "lr": 9.984676372050193e-05}
{"train_loss": 0.813307523727417, "global_step": 7254, "epoch": 81, "lr": 9.984671836389091e-05}
{"train_loss": 0.6742735505104065, "global_step": 7255, "epoch": 81, "lr": 9.98466730005786e-05}
{"train_loss": 0.7049031257629395, "global_step": 7256, "epoch": 81, "lr": 9.984662763056503e-05}
{"train_loss": 0.7314985990524292, "global_step": 7257, "epoch": 81, "lr": 9.984658225385017e-05}
{"train_loss": 0.7592703104019165, "global_step": 7258, "epoch": 81, "lr": 9.984653687043405e-05}
{"train_loss": 0.5835261344909668, "global_step": 7259, "epoch": 81, "lr": 9.984649148031667e-05}
{"train_loss": 0.7679811120033264, "global_step": 7260, "epoch": 81, "lr": 9.984644608349804e-05}
{"train_loss": 0.8544332981109619, "global_step": 7261, "epoch": 81, "lr": 9.984640067997818e-05}
{"train_loss": 0.872391939163208, "global_step": 7262, "epoch": 81, "lr": 9.984635526975705e-05}
{"train_loss": 0.6523246169090271, "global_step": 7263, "epoch": 81, "lr": 9.98463098528347e-05}
{"train_loss": 0.9764171838760376, "global_step": 7264, "epoch": 81, "lr": 9.98462644292111e-05}
{"train_loss": 0.9035733342170715, "global_step": 7265, "epoch": 81, "lr": 9.98462189988863e-05}
{"train_loss": 0.7100447416305542, "global_step": 7266, "epoch": 81, "lr": 9.984617356186028e-05}
{"train_loss": 0.9169553518295288, "global_step": 7267, "epoch": 81, "lr": 9.984612811813305e-05}
{"train_loss": 0.9099051356315613, "global_step": 7268, "epoch": 81, "lr": 9.984608266770461e-05}
{"train_loss": 0.7557271122932434, "global_step": 7269, "epoch": 81, "lr": 9.984603721057497e-05}
{"train_loss": 0.8150414824485779, "global_step": 7270, "epoch": 81, "lr": 9.984599174674415e-05}
{"train_loss": 0.8718082308769226, "global_step": 7271, "epoch": 81, "lr": 9.984594627621214e-05}
{"train_loss": 0.8452930450439453, "global_step": 7272, "epoch": 81, "lr": 9.984590079897894e-05}
{"train_loss": 1.0264431238174438, "global_step": 7273, "epoch": 81, "lr": 9.984585531504458e-05}
{"train_loss": 0.8718303442001343, "global_step": 7274, "epoch": 81, "lr": 9.984580982440905e-05}
{"train_loss": 0.8322239518165588, "global_step": 7275, "epoch": 81, "lr": 9.984576432707234e-05}
{"train_loss": 0.795256495475769, "global_step": 7276, "epoch": 81, "lr": 9.984571882303449e-05}
{"train_loss": 0.9365063905715942, "global_step": 7277, "epoch": 81, "lr": 9.984567331229549e-05}
{"train_loss": 0.8338385224342346, "global_step": 7278, "epoch": 81, "lr": 9.984562779485536e-05}
{"train_loss": 1.0565077066421509, "global_step": 7279, "epoch": 81, "lr": 9.984558227071407e-05}
{"train_loss": 0.7260589599609375, "global_step": 7280, "epoch": 81, "lr": 9.984553673987165e-05}
{"train_loss": 0.8211279511451721, "global_step": 7281, "epoch": 81, "lr": 9.984549120232811e-05}
{"train_loss": 0.903293251991272, "global_step": 7282, "epoch": 81, "lr": 9.984544565808347e-05}
{"train_loss": 0.7752217650413513, "global_step": 7283, "epoch": 81, "lr": 9.984540010713769e-05}
{"train_loss": 0.6556566953659058, "global_step": 7284, "epoch": 81, "lr": 9.984535454949082e-05}
{"train_loss": 0.9038421511650085, "global_step": 7285, "epoch": 81, "lr": 9.984530898514284e-05}
{"train_loss": 0.879271388053894, "global_step": 7286, "epoch": 81, "lr": 9.984526341409376e-05}
{"train_loss": 0.8269804120063782, "global_step": 7287, "epoch": 81, "lr": 9.984521783634361e-05}
{"train_loss": 0.7425568103790283, "global_step": 7288, "epoch": 81, "lr": 9.984517225189237e-05}
{"train_loss": 0.7560113668441772, "global_step": 7289, "epoch": 81, "lr": 9.984512666074006e-05}
{"train_loss": 0.9000733494758606, "global_step": 7290, "epoch": 81, "lr": 9.984508106288669e-05}
{"train_loss": 0.8618781566619873, "global_step": 7291, "epoch": 81, "lr": 9.984503545833222e-05}
{"train_loss": 0.8179338574409485, "global_step": 7292, "epoch": 81, "lr": 9.984498984707672e-05}
{"train_loss": 1.0349247455596924, "global_step": 7293, "epoch": 81, "lr": 9.984494422912016e-05}
{"train_loss": 0.8872507810592651, "global_step": 7294, "epoch": 81, "lr": 9.984489860446256e-05}
{"train_loss": 0.7660768628120422, "global_step": 7295, "epoch": 81, "lr": 9.984485297310391e-05}
{"train_loss": 0.6295580267906189, "global_step": 7296, "epoch": 81, "lr": 9.984480733504424e-05}
{"train_loss": 0.8337815136052249, "global_step": 7297, "epoch": 81, "lr": 9.984476169028353e-05, "val_loss": 1.1749836206436157}
{"train_loss": 0.817330002784729, "global_step": 7298, "epoch": 82, "lr": 9.984471603882181e-05}
{"train_loss": 0.7037861943244934, "global_step": 7299, "epoch": 82, "lr": 9.984467038065908e-05}
{"train_loss": 0.8860369324684143, "global_step": 7300, "epoch": 82, "lr": 9.984462471579532e-05}
{"train_loss": 0.8252578973770142, "global_step": 7301, "epoch": 82, "lr": 9.984457904423058e-05}
{"train_loss": 0.8918721675872803, "global_step": 7302, "epoch": 82, "lr": 9.984453336596484e-05}
{"train_loss": 0.8267045021057129, "global_step": 7303, "epoch": 82, "lr": 9.98444876809981e-05}
{"train_loss": 0.839755654335022, "global_step": 7304, "epoch": 82, "lr": 9.984444198933038e-05}
{"train_loss": 0.9272934198379517, "global_step": 7305, "epoch": 82, "lr": 9.984439629096169e-05}
{"train_loss": 0.894119918346405, "global_step": 7306, "epoch": 82, "lr": 9.984435058589202e-05}
{"train_loss": 0.8204274773597717, "global_step": 7307, "epoch": 82, "lr": 9.984430487412139e-05}
{"train_loss": 0.7551412582397461, "global_step": 7308, "epoch": 82, "lr": 9.98442591556498e-05}
{"train_loss": 0.7365522384643555, "global_step": 7309, "epoch": 82, "lr": 9.984421343047725e-05}
{"train_loss": 0.8543692827224731, "global_step": 7310, "epoch": 82, "lr": 9.984416769860376e-05}
{"train_loss": 0.8486990928649902, "global_step": 7311, "epoch": 82, "lr": 9.984412196002933e-05}
{"train_loss": 0.7693814635276794, "global_step": 7312, "epoch": 82, "lr": 9.984407621475395e-05}
{"train_loss": 0.9100195169448853, "global_step": 7313, "epoch": 82, "lr": 9.984403046277766e-05}
{"train_loss": 0.7355255484580994, "global_step": 7314, "epoch": 82, "lr": 9.984398470410043e-05}
{"train_loss": 0.7285891175270081, "global_step": 7315, "epoch": 82, "lr": 9.98439389387223e-05}
{"train_loss": 0.8056449294090271, "global_step": 7316, "epoch": 82, "lr": 9.984389316664327e-05}
{"train_loss": 0.8005349040031433, "global_step": 7317, "epoch": 82, "lr": 9.984384738786331e-05}
{"train_loss": 0.8153524398803711, "global_step": 7318, "epoch": 82, "lr": 9.984380160238247e-05}
{"train_loss": 0.7970550060272217, "global_step": 7319, "epoch": 82, "lr": 9.984375581020073e-05}
{"train_loss": 0.7344991564750671, "global_step": 7320, "epoch": 82, "lr": 9.98437100113181e-05}
{"train_loss": 0.8171444535255432, "global_step": 7321, "epoch": 82, "lr": 9.98436642057346e-05}
{"train_loss": 0.8439289927482605, "global_step": 7322, "epoch": 82, "lr": 9.984361839345022e-05}
{"train_loss": 0.7755857110023499, "global_step": 7323, "epoch": 82, "lr": 9.984357257446499e-05}
{"train_loss": 0.8645749092102051, "global_step": 7324, "epoch": 82, "lr": 9.984352674877889e-05}
{"train_loss": 0.6821959614753723, "global_step": 7325, "epoch": 82, "lr": 9.984348091639191e-05}
{"train_loss": 0.7835236191749573, "global_step": 7326, "epoch": 82, "lr": 9.984343507730411e-05}
{"train_loss": 0.8289721012115479, "global_step": 7327, "epoch": 82, "lr": 9.984338923151547e-05}
{"train_loss": 0.9395610690116882, "global_step": 7328, "epoch": 82, "lr": 9.984334337902598e-05}
{"train_loss": 0.8993100523948669, "global_step": 7329, "epoch": 82, "lr": 9.984329751983566e-05}
{"train_loss": 0.7810448408126831, "global_step": 7330, "epoch": 82, "lr": 9.984325165394452e-05}
{"train_loss": 0.8730878233909607, "global_step": 7331, "epoch": 82, "lr": 9.984320578135257e-05}
{"train_loss": 0.9292298555374146, "global_step": 7332, "epoch": 82, "lr": 9.984315990205979e-05}
{"train_loss": 0.9665617942810059, "global_step": 7333, "epoch": 82, "lr": 9.984311401606621e-05}
{"train_loss": 0.7417938709259033, "global_step": 7334, "epoch": 82, "lr": 9.984306812337185e-05}
{"train_loss": 0.9239699840545654, "global_step": 7335, "epoch": 82, "lr": 9.984302222397667e-05}
{"train_loss": 0.7260707020759583, "global_step": 7336, "epoch": 82, "lr": 9.984297631788073e-05}
{"train_loss": 1.0802273750305176, "global_step": 7337, "epoch": 82, "lr": 9.984293040508399e-05}
{"train_loss": 0.8290265798568726, "global_step": 7338, "epoch": 82, "lr": 9.984288448558648e-05}
{"train_loss": 0.7775889039039612, "global_step": 7339, "epoch": 82, "lr": 9.984283855938821e-05}
{"train_loss": 0.9219366908073425, "global_step": 7340, "epoch": 82, "lr": 9.984279262648917e-05}
{"train_loss": 0.9023103713989258, "global_step": 7341, "epoch": 82, "lr": 9.984274668688937e-05}
{"train_loss": 0.9787980318069458, "global_step": 7342, "epoch": 82, "lr": 9.984270074058883e-05}
{"train_loss": 0.9600109457969666, "global_step": 7343, "epoch": 82, "lr": 9.984265478758754e-05}
{"train_loss": 0.7690045237541199, "global_step": 7344, "epoch": 82, "lr": 9.984260882788552e-05}
{"train_loss": 1.0005860328674316, "global_step": 7345, "epoch": 82, "lr": 9.984256286148277e-05}
{"train_loss": 0.9345364570617676, "global_step": 7346, "epoch": 82, "lr": 9.984251688837928e-05}
{"train_loss": 0.8898884654045105, "global_step": 7347, "epoch": 82, "lr": 9.984247090857508e-05}
{"train_loss": 0.828087568283081, "global_step": 7348, "epoch": 82, "lr": 9.984242492207017e-05}
{"train_loss": 0.9055986404418945, "global_step": 7349, "epoch": 82, "lr": 9.984237892886455e-05}
{"train_loss": 1.0040786266326904, "global_step": 7350, "epoch": 82, "lr": 9.984233292895823e-05}
{"train_loss": 0.8491101861000061, "global_step": 7351, "epoch": 82, "lr": 9.984228692235122e-05}
{"train_loss": 0.8600109815597534, "global_step": 7352, "epoch": 82, "lr": 9.984224090904351e-05}
{"train_loss": 0.8917251229286194, "global_step": 7353, "epoch": 82, "lr": 9.984219488903512e-05}
{"train_loss": 0.8639256954193115, "global_step": 7354, "epoch": 82, "lr": 9.984214886232605e-05}
{"train_loss": 0.8292826414108276, "global_step": 7355, "epoch": 82, "lr": 9.984210282891633e-05}
{"train_loss": 0.749923586845398, "global_step": 7356, "epoch": 82, "lr": 9.984205678880594e-05}
{"train_loss": 0.7489400506019592, "global_step": 7357, "epoch": 82, "lr": 9.984201074199488e-05}
{"train_loss": 0.7810251712799072, "global_step": 7358, "epoch": 82, "lr": 9.98419646884832e-05}
{"train_loss": 0.8458455204963684, "global_step": 7359, "epoch": 82, "lr": 9.984191862827085e-05}
{"train_loss": 1.01292085647583, "global_step": 7360, "epoch": 82, "lr": 9.984187256135787e-05}
{"train_loss": 0.8776562809944153, "global_step": 7361, "epoch": 82, "lr": 9.984182648774425e-05}
{"train_loss": 0.9755827784538269, "global_step": 7362, "epoch": 82, "lr": 9.984178040743001e-05}
{"train_loss": 0.7970577478408813, "global_step": 7363, "epoch": 82, "lr": 9.984173432041515e-05}
{"train_loss": 0.6785028576850891, "global_step": 7364, "epoch": 82, "lr": 9.984168822669966e-05}
{"train_loss": 0.9372377395629883, "global_step": 7365, "epoch": 82, "lr": 9.984164212628359e-05}
{"train_loss": 0.6271610260009766, "global_step": 7366, "epoch": 82, "lr": 9.984159601916689e-05}
{"train_loss": 0.8331213593482971, "global_step": 7367, "epoch": 82, "lr": 9.984154990534962e-05}
{"train_loss": 0.7569146752357483, "global_step": 7368, "epoch": 82, "lr": 9.984150378483174e-05}
{"train_loss": 0.846793532371521, "global_step": 7369, "epoch": 82, "lr": 9.98414576576133e-05}
{"train_loss": 0.8377870917320251, "global_step": 7370, "epoch": 82, "lr": 9.984141152369427e-05}
{"train_loss": 0.7743874192237854, "global_step": 7371, "epoch": 82, "lr": 9.984136538307468e-05}
{"train_loss": 0.8308184742927551, "global_step": 7372, "epoch": 82, "lr": 9.984131923575451e-05}
{"train_loss": 0.7476427555084229, "global_step": 7373, "epoch": 82, "lr": 9.984127308173379e-05}
{"train_loss": 0.857187032699585, "global_step": 7374, "epoch": 82, "lr": 9.984122692101252e-05}
{"train_loss": 0.9498640298843384, "global_step": 7375, "epoch": 82, "lr": 9.98411807535907e-05}
{"train_loss": 0.8126217722892761, "global_step": 7376, "epoch": 82, "lr": 9.984113457946836e-05}
{"train_loss": 0.8213884234428406, "global_step": 7377, "epoch": 82, "lr": 9.984108839864545e-05}
{"train_loss": 0.7601917386054993, "global_step": 7378, "epoch": 82, "lr": 9.984104221112204e-05}
{"train_loss": 0.901287317276001, "global_step": 7379, "epoch": 82, "lr": 9.984099601689811e-05}
{"train_loss": 0.8061966300010681, "global_step": 7380, "epoch": 82, "lr": 9.984094981597365e-05}
{"train_loss": 0.5972080826759338, "global_step": 7381, "epoch": 82, "lr": 9.98409036083487e-05}
{"train_loss": 0.9593115448951721, "global_step": 7382, "epoch": 82, "lr": 9.984085739402323e-05}
{"train_loss": 0.7933865785598755, "global_step": 7383, "epoch": 82, "lr": 9.984081117299728e-05}
{"train_loss": 0.803536593914032, "global_step": 7384, "epoch": 82, "lr": 9.984076494527084e-05}
{"train_loss": 0.947186291217804, "global_step": 7385, "epoch": 82, "lr": 9.984071871084391e-05}
{"train_loss": 0.8385135102807806, "global_step": 7386, "epoch": 82, "lr": 9.984067246971649e-05, "val_loss": 1.1661434173583984}
{"train_loss": 0.7002614736557007, "global_step": 7387, "epoch": 83, "lr": 9.984062622188861e-05}
{"train_loss": 0.8699848651885986, "global_step": 7388, "epoch": 83, "lr": 9.984057996736028e-05}
{"train_loss": 0.7735087871551514, "global_step": 7389, "epoch": 83, "lr": 9.984053370613147e-05}
{"train_loss": 1.0277491807937622, "global_step": 7390, "epoch": 83, "lr": 9.984048743820222e-05}
{"train_loss": 0.6102144122123718, "global_step": 7391, "epoch": 83, "lr": 9.984044116357252e-05}
{"train_loss": 0.6853057146072388, "global_step": 7392, "epoch": 83, "lr": 9.984039488224238e-05}
{"train_loss": 0.7471307516098022, "global_step": 7393, "epoch": 83, "lr": 9.984034859421181e-05}
{"train_loss": 0.785828173160553, "global_step": 7394, "epoch": 83, "lr": 9.984030229948081e-05}
{"train_loss": 0.8480753898620605, "global_step": 7395, "epoch": 83, "lr": 9.984025599804938e-05}
{"train_loss": 0.7516818046569824, "global_step": 7396, "epoch": 83, "lr": 9.984020968991754e-05}
{"train_loss": 0.8879584670066833, "global_step": 7397, "epoch": 83, "lr": 9.98401633750853e-05}
{"train_loss": 0.7376201748847961, "global_step": 7398, "epoch": 83, "lr": 9.984011705355265e-05}
{"train_loss": 0.8347966074943542, "global_step": 7399, "epoch": 83, "lr": 9.98400707253196e-05}
{"train_loss": 0.8795129656791687, "global_step": 7400, "epoch": 83, "lr": 9.984002439038617e-05}
{"train_loss": 0.9760664701461792, "global_step": 7401, "epoch": 83, "lr": 9.983997804875236e-05}
{"train_loss": 0.7156140804290771, "global_step": 7402, "epoch": 83, "lr": 9.983993170041816e-05}
{"train_loss": 0.8091950416564941, "global_step": 7403, "epoch": 83, "lr": 9.983988534538358e-05}
{"train_loss": 0.9117172360420227, "global_step": 7404, "epoch": 83, "lr": 9.983983898364866e-05}
{"train_loss": 0.6570488810539246, "global_step": 7405, "epoch": 83, "lr": 9.983979261521336e-05}
{"train_loss": 0.7340532541275024, "global_step": 7406, "epoch": 83, "lr": 9.983974624007772e-05}
{"train_loss": 0.8534266948699951, "global_step": 7407, "epoch": 83, "lr": 9.983969985824173e-05}
{"train_loss": 0.9602601528167725, "global_step": 7408, "epoch": 83, "lr": 9.98396534697054e-05}
{"train_loss": 0.6975866556167603, "global_step": 7409, "epoch": 83, "lr": 9.983960707446873e-05}
{"train_loss": 0.7917492389678955, "global_step": 7410, "epoch": 83, "lr": 9.983956067253174e-05}
{"train_loss": 0.9170687198638916, "global_step": 7411, "epoch": 83, "lr": 9.983951426389443e-05}
{"train_loss": 0.8717588782310486, "global_step": 7412, "epoch": 83, "lr": 9.98394678485568e-05}
{"train_loss": 0.8318065404891968, "global_step": 7413, "epoch": 83, "lr": 9.983942142651886e-05}
{"train_loss": 0.8184401988983154, "global_step": 7414, "epoch": 83, "lr": 9.983937499778062e-05}
{"train_loss": 0.6956321001052856, "global_step": 7415, "epoch": 83, "lr": 9.983932856234208e-05}
{"train_loss": 0.7675722241401672, "global_step": 7416, "epoch": 83, "lr": 9.983928212020326e-05}
{"train_loss": 0.9981459379196167, "global_step": 7417, "epoch": 83, "lr": 9.983923567136416e-05}
{"train_loss": 0.7879859209060669, "global_step": 7418, "epoch": 83, "lr": 9.983918921582476e-05}
{"train_loss": 0.8682850003242493, "global_step": 7419, "epoch": 83, "lr": 9.983914275358511e-05}
{"train_loss": 0.9240392446517944, "global_step": 7420, "epoch": 83, "lr": 9.983909628464516e-05}
{"train_loss": 1.1467578411102295, "global_step": 7421, "epoch": 83, "lr": 9.983904980900497e-05}
{"train_loss": 0.7128252983093262, "global_step": 7422, "epoch": 83, "lr": 9.983900332666455e-05}
{"train_loss": 0.8624565601348877, "global_step": 7423, "epoch": 83, "lr": 9.983895683762385e-05}
{"train_loss": 0.8044307231903076, "global_step": 7424, "epoch": 83, "lr": 9.983891034188293e-05}
{"train_loss": 0.9294440150260925, "global_step": 7425, "epoch": 83, "lr": 9.983886383944178e-05}
{"train_loss": 0.9543338418006897, "global_step": 7426, "epoch": 83, "lr": 9.983881733030039e-05}
{"train_loss": 0.7069839239120483, "global_step": 7427, "epoch": 83, "lr": 9.983877081445879e-05}
{"train_loss": 0.8490622043609619, "global_step": 7428, "epoch": 83, "lr": 9.983872429191696e-05}
{"train_loss": 0.7667572498321533, "global_step": 7429, "epoch": 83, "lr": 9.983867776267492e-05}
{"train_loss": 0.8590908646583557, "global_step": 7430, "epoch": 83, "lr": 9.98386312267327e-05}
{"train_loss": 0.7918540835380554, "global_step": 7431, "epoch": 83, "lr": 9.983858468409026e-05}
{"train_loss": 0.8070821762084961, "global_step": 7432, "epoch": 83, "lr": 9.983853813474763e-05}
{"train_loss": 0.7469583749771118, "global_step": 7433, "epoch": 83, "lr": 9.983849157870484e-05}
{"train_loss": 0.8744195699691772, "global_step": 7434, "epoch": 83, "lr": 9.983844501596184e-05}
{"train_loss": 1.0744645595550537, "global_step": 7435, "epoch": 83, "lr": 9.98383984465187e-05}
{"train_loss": 0.5981997847557068, "global_step": 7436, "epoch": 83, "lr": 9.983835187037538e-05}
{"train_loss": 0.7557613253593445, "global_step": 7437, "epoch": 83, "lr": 9.983830528753188e-05}
{"train_loss": 0.8075911402702332, "global_step": 7438, "epoch": 83, "lr": 9.983825869798826e-05}
{"train_loss": 0.8326741456985474, "global_step": 7439, "epoch": 83, "lr": 9.98382121017445e-05}
{"train_loss": 0.728475034236908, "global_step": 7440, "epoch": 83, "lr": 9.983816549880057e-05}
{"train_loss": 0.7586539387702942, "global_step": 7441, "epoch": 83, "lr": 9.983811888915651e-05}
{"train_loss": 0.7882200479507446, "global_step": 7442, "epoch": 83, "lr": 9.983807227281233e-05}
{"train_loss": 0.8305181264877319, "global_step": 7443, "epoch": 83, "lr": 9.983802564976802e-05}
{"train_loss": 0.7617727518081665, "global_step": 7444, "epoch": 83, "lr": 9.98379790200236e-05}
{"train_loss": 0.8679507374763489, "global_step": 7445, "epoch": 83, "lr": 9.983793238357908e-05}
{"train_loss": 0.7612456679344177, "global_step": 7446, "epoch": 83, "lr": 9.983788574043446e-05}
{"train_loss": 0.7930348515510559, "global_step": 7447, "epoch": 83, "lr": 9.983783909058973e-05}
{"train_loss": 0.9621853232383728, "global_step": 7448, "epoch": 83, "lr": 9.983779243404492e-05}
{"train_loss": 0.8355666995048523, "global_step": 7449, "epoch": 83, "lr": 9.983774577080003e-05}
{"train_loss": 0.8929045796394348, "global_step": 7450, "epoch": 83, "lr": 9.983769910085506e-05}
{"train_loss": 0.8946245312690735, "global_step": 7451, "epoch": 83, "lr": 9.983765242421e-05}
{"train_loss": 0.8950543403625488, "global_step": 7452, "epoch": 83, "lr": 9.98376057408649e-05}
{"train_loss": 0.7492496967315674, "global_step": 7453, "epoch": 83, "lr": 9.983755905081972e-05}
{"train_loss": 0.6258291006088257, "global_step": 7454, "epoch": 83, "lr": 9.983751235407449e-05}
{"train_loss": 0.9292277693748474, "global_step": 7455, "epoch": 83, "lr": 9.983746565062922e-05}
{"train_loss": 0.6990393996238708, "global_step": 7456, "epoch": 83, "lr": 9.983741894048392e-05}
{"train_loss": 0.8323927521705627, "global_step": 7457, "epoch": 83, "lr": 9.983737222363858e-05}
{"train_loss": 0.9057406783103943, "global_step": 7458, "epoch": 83, "lr": 9.98373255000932e-05}
{"train_loss": 0.7795258164405823, "global_step": 7459, "epoch": 83, "lr": 9.983727876984781e-05}
{"train_loss": 0.7246068716049194, "global_step": 7460, "epoch": 83, "lr": 9.98372320329024e-05}
{"train_loss": 0.9009511470794678, "global_step": 7461, "epoch": 83, "lr": 9.9837185289257e-05}
{"train_loss": 0.885676920413971, "global_step": 7462, "epoch": 83, "lr": 9.983713853891158e-05}
{"train_loss": 0.6260535717010498, "global_step": 7463, "epoch": 83, "lr": 9.983709178186616e-05}
{"train_loss": 0.8060698509216309, "global_step": 7464, "epoch": 83, "lr": 9.983704501812077e-05}
{"train_loss": 0.672266960144043, "global_step": 7465, "epoch": 83, "lr": 9.983699824767538e-05}
{"train_loss": 0.6956620216369629, "global_step": 7466, "epoch": 83, "lr": 9.983695147053002e-05}
{"train_loss": 1.0025683641433716, "global_step": 7467, "epoch": 83, "lr": 9.983690468668468e-05}
{"train_loss": 0.8979981541633606, "global_step": 7468, "epoch": 83, "lr": 9.983685789613939e-05}
{"train_loss": 0.9213176369667053, "global_step": 7469, "epoch": 83, "lr": 9.983681109889413e-05}
{"train_loss": 0.7344027757644653, "global_step": 7470, "epoch": 83, "lr": 9.983676429494893e-05}
{"train_loss": 0.9465835690498352, "global_step": 7471, "epoch": 83, "lr": 9.983671748430378e-05}
{"train_loss": 0.7110381126403809, "global_step": 7472, "epoch": 83, "lr": 9.983667066695869e-05}
{"train_loss": 0.820110023021698, "global_step": 7473, "epoch": 83, "lr": 9.983662384291366e-05}
{"train_loss": 0.9220449924468994, "global_step": 7474, "epoch": 83, "lr": 9.98365770121687e-05}
{"train_loss": 0.8204290732908784, "global_step": 7475, "epoch": 83, "lr": 9.983653017472384e-05, "val_loss": 1.1224154233932495}
{"train_loss": 0.859699010848999, "global_step": 7476, "epoch": 84, "lr": 9.983648333057905e-05}
{"train_loss": 0.8562131524085999, "global_step": 7477, "epoch": 84, "lr": 9.983643647973435e-05}
{"train_loss": 0.794599711894989, "global_step": 7478, "epoch": 84, "lr": 9.983638962218976e-05}
{"train_loss": 0.7850585579872131, "global_step": 7479, "epoch": 84, "lr": 9.983634275794526e-05}
{"train_loss": 0.8864302635192871, "global_step": 7480, "epoch": 84, "lr": 9.983629588700089e-05}
{"train_loss": 0.7119666934013367, "global_step": 7481, "epoch": 84, "lr": 9.983624900935664e-05}
{"train_loss": 0.9816979169845581, "global_step": 7482, "epoch": 84, "lr": 9.98362021250125e-05}
{"train_loss": 0.804229736328125, "global_step": 7483, "epoch": 84, "lr": 9.98361552339685e-05}
{"train_loss": 0.7211275100708008, "global_step": 7484, "epoch": 84, "lr": 9.983610833622462e-05}
{"train_loss": 0.856755256652832, "global_step": 7485, "epoch": 84, "lr": 9.98360614317809e-05}
{"train_loss": 0.7336291670799255, "global_step": 7486, "epoch": 84, "lr": 9.983601452063731e-05}
{"train_loss": 0.714046835899353, "global_step": 7487, "epoch": 84, "lr": 9.983596760279388e-05}
{"train_loss": 0.9237768650054932, "global_step": 7488, "epoch": 84, "lr": 9.983592067825063e-05}
{"train_loss": 0.8592056632041931, "global_step": 7489, "epoch": 84, "lr": 9.983587374700753e-05}
{"train_loss": 0.8823338747024536, "global_step": 7490, "epoch": 84, "lr": 9.983582680906461e-05}
{"train_loss": 0.7571423053741455, "global_step": 7491, "epoch": 84, "lr": 9.983577986442187e-05}
{"train_loss": 0.8438618779182434, "global_step": 7492, "epoch": 84, "lr": 9.983573291307932e-05}
{"train_loss": 0.7316008806228638, "global_step": 7493, "epoch": 84, "lr": 9.983568595503696e-05}
{"train_loss": 0.7136486172676086, "global_step": 7494, "epoch": 84, "lr": 9.98356389902948e-05}
{"train_loss": 1.0172796249389648, "global_step": 7495, "epoch": 84, "lr": 9.983559201885284e-05}
{"train_loss": 0.8824041485786438, "global_step": 7496, "epoch": 84, "lr": 9.98355450407111e-05}
{"train_loss": 0.7978125214576721, "global_step": 7497, "epoch": 84, "lr": 9.983549805586957e-05}
{"train_loss": 0.8359060883522034, "global_step": 7498, "epoch": 84, "lr": 9.983545106432827e-05}
{"train_loss": 0.7341592311859131, "global_step": 7499, "epoch": 84, "lr": 9.98354040660872e-05}
{"train_loss": 0.683779239654541, "global_step": 7500, "epoch": 84, "lr": 9.983535706114638e-05}
{"train_loss": 0.826192319393158, "global_step": 7501, "epoch": 84, "lr": 9.983531004950579e-05}
{"train_loss": 0.8439760804176331, "global_step": 7502, "epoch": 84, "lr": 9.983526303116543e-05}
{"train_loss": 0.6960098147392273, "global_step": 7503, "epoch": 84, "lr": 9.983521600612536e-05}
{"train_loss": 0.8741660118103027, "global_step": 7504, "epoch": 84, "lr": 9.983516897438555e-05}
{"train_loss": 0.8412965536117554, "global_step": 7505, "epoch": 84, "lr": 9.9835121935946e-05}
{"train_loss": 0.7688615918159485, "global_step": 7506, "epoch": 84, "lr": 9.983507489080673e-05}
{"train_loss": 0.6521241664886475, "global_step": 7507, "epoch": 84, "lr": 9.983502783896773e-05}
{"train_loss": 0.7646154761314392, "global_step": 7508, "epoch": 84, "lr": 9.983498078042902e-05}
{"train_loss": 0.8294076323509216, "global_step": 7509, "epoch": 84, "lr": 9.983493371519062e-05}
{"train_loss": 0.6655862927436829, "global_step": 7510, "epoch": 84, "lr": 9.98348866432525e-05}
{"train_loss": 0.8385624289512634, "global_step": 7511, "epoch": 84, "lr": 9.98348395646147e-05}
{"train_loss": 0.8563355207443237, "global_step": 7512, "epoch": 84, "lr": 9.983479247927721e-05}
{"train_loss": 0.9342052936553955, "global_step": 7513, "epoch": 84, "lr": 9.983474538724004e-05}
{"train_loss": 0.6785591244697571, "global_step": 7514, "epoch": 84, "lr": 9.98346982885032e-05}
{"train_loss": 0.838607668876648, "global_step": 7515, "epoch": 84, "lr": 9.983465118306668e-05}
{"train_loss": 0.9531251192092896, "global_step": 7516, "epoch": 84, "lr": 9.98346040709305e-05}
{"train_loss": 0.9529354572296143, "global_step": 7517, "epoch": 84, "lr": 9.983455695209467e-05}
{"train_loss": 0.6774746775627136, "global_step": 7518, "epoch": 84, "lr": 9.983450982655919e-05}
{"train_loss": 0.6655271649360657, "global_step": 7519, "epoch": 84, "lr": 9.983446269432408e-05}
{"train_loss": 0.6870740652084351, "global_step": 7520, "epoch": 84, "lr": 9.983441555538933e-05}
{"train_loss": 0.8385646939277649, "global_step": 7521, "epoch": 84, "lr": 9.983436840975493e-05}
{"train_loss": 0.7752588391304016, "global_step": 7522, "epoch": 84, "lr": 9.983432125742091e-05}
{"train_loss": 0.847382664680481, "global_step": 7523, "epoch": 84, "lr": 9.98342740983873e-05}
{"train_loss": 0.9052512049674988, "global_step": 7524, "epoch": 84, "lr": 9.983422693265406e-05}
{"train_loss": 0.8545238971710205, "global_step": 7525, "epoch": 84, "lr": 9.983417976022123e-05}
{"train_loss": 0.7953757047653198, "global_step": 7526, "epoch": 84, "lr": 9.983413258108877e-05}
{"train_loss": 0.7101662158966064, "global_step": 7527, "epoch": 84, "lr": 9.983408539525674e-05}
{"train_loss": 0.8151490092277527, "global_step": 7528, "epoch": 84, "lr": 9.983403820272512e-05}
{"train_loss": 0.9722530841827393, "global_step": 7529, "epoch": 84, "lr": 9.983399100349394e-05}
{"train_loss": 0.7407398819923401, "global_step": 7530, "epoch": 84, "lr": 9.983394379756316e-05}
{"train_loss": 0.843910276889801, "global_step": 7531, "epoch": 84, "lr": 9.983389658493283e-05}
{"train_loss": 0.9883744120597839, "global_step": 7532, "epoch": 84, "lr": 9.983384936560293e-05}
{"train_loss": 0.7232099175453186, "global_step": 7533, "epoch": 84, "lr": 9.983380213957349e-05}
{"train_loss": 0.9296718835830688, "global_step": 7534, "epoch": 84, "lr": 9.983375490684448e-05}
{"train_loss": 0.980246365070343, "global_step": 7535, "epoch": 84, "lr": 9.983370766741595e-05}
{"train_loss": 0.7788397669792175, "global_step": 7536, "epoch": 84, "lr": 9.983366042128788e-05}
{"train_loss": 0.9393568634986877, "global_step": 7537, "epoch": 84, "lr": 9.983361316846027e-05}
{"train_loss": 0.6473028659820557, "global_step": 7538, "epoch": 84, "lr": 9.983356590893316e-05}
{"train_loss": 0.8751771450042725, "global_step": 7539, "epoch": 84, "lr": 9.983351864270652e-05}
{"train_loss": 0.8715142607688904, "global_step": 7540, "epoch": 84, "lr": 9.983347136978036e-05}
{"train_loss": 0.823157787322998, "global_step": 7541, "epoch": 84, "lr": 9.983342409015472e-05}
{"train_loss": 0.8061599731445312, "global_step": 7542, "epoch": 84, "lr": 9.983337680382958e-05}
{"train_loss": 0.9194076061248779, "global_step": 7543, "epoch": 84, "lr": 9.983332951080493e-05}
{"train_loss": 0.7592502236366272, "global_step": 7544, "epoch": 84, "lr": 9.983328221108082e-05}
{"train_loss": 0.7263949513435364, "global_step": 7545, "epoch": 84, "lr": 9.983323490465721e-05}
{"train_loss": 0.7017850875854492, "global_step": 7546, "epoch": 84, "lr": 9.983318759153415e-05}
{"train_loss": 0.7723050713539124, "global_step": 7547, "epoch": 84, "lr": 9.983314027171162e-05}
{"train_loss": 0.8163854479789734, "global_step": 7548, "epoch": 84, "lr": 9.983309294518964e-05}
{"train_loss": 0.7804840803146362, "global_step": 7549, "epoch": 84, "lr": 9.983304561196817e-05}
{"train_loss": 0.8699384927749634, "global_step": 7550, "epoch": 84, "lr": 9.983299827204729e-05}
{"train_loss": 0.7800405025482178, "global_step": 7551, "epoch": 84, "lr": 9.983295092542697e-05}
{"train_loss": 0.8433417677879333, "global_step": 7552, "epoch": 84, "lr": 9.983290357210721e-05}
{"train_loss": 0.8083569407463074, "global_step": 7553, "epoch": 84, "lr": 9.983285621208802e-05}
{"train_loss": 0.7605129480361938, "global_step": 7554, "epoch": 84, "lr": 9.983280884536941e-05}
{"train_loss": 0.6825379729270935, "global_step": 7555, "epoch": 84, "lr": 9.983276147195139e-05}
{"train_loss": 0.9246062636375427, "global_step": 7556, "epoch": 84, "lr": 9.983271409183395e-05}
{"train_loss": 0.8271906971931458, "global_step": 7557, "epoch": 84, "lr": 9.983266670501712e-05}
{"train_loss": 0.7508569359779358, "global_step": 7558, "epoch": 84, "lr": 9.98326193115009e-05}
{"train_loss": 0.7041118144989014, "global_step": 7559, "epoch": 84, "lr": 9.983257191128529e-05}
{"train_loss": 0.9563618898391724, "global_step": 7560, "epoch": 84, "lr": 9.983252450437029e-05}
{"train_loss": 0.8712388873100281, "global_step": 7561, "epoch": 84, "lr": 9.983247709075593e-05}
{"train_loss": 0.7786345481872559, "global_step": 7562, "epoch": 84, "lr": 9.983242967044219e-05}
{"train_loss": 0.8131638765335083, "global_step": 7563, "epoch": 84, "lr": 9.983238224342908e-05}
{"train_loss": 0.8126074986511402, "global_step": 7564, "epoch": 84, "lr": 9.983233480971663e-05, "val_loss": 1.179524540901184}
{"train_loss": 0.739729642868042, "global_step": 7565, "epoch": 85, "lr": 9.983228736930481e-05}
{"train_loss": 0.9907019138336182, "global_step": 7566, "epoch": 85, "lr": 9.983223992219366e-05}
{"train_loss": 0.8776815533638, "global_step": 7567, "epoch": 85, "lr": 9.983219246838318e-05}
{"train_loss": 0.8970618844032288, "global_step": 7568, "epoch": 85, "lr": 9.983214500787335e-05}
{"train_loss": 0.9094586372375488, "global_step": 7569, "epoch": 85, "lr": 9.98320975406642e-05}
{"train_loss": 0.772609293460846, "global_step": 7570, "epoch": 85, "lr": 9.983205006675574e-05}
{"train_loss": 0.7408576607704163, "global_step": 7571, "epoch": 85, "lr": 9.983200258614796e-05}
{"train_loss": 0.798768937587738, "global_step": 7572, "epoch": 85, "lr": 9.983195509884089e-05}
{"train_loss": 0.8506155014038086, "global_step": 7573, "epoch": 85, "lr": 9.983190760483451e-05}
{"train_loss": 0.6810228824615479, "global_step": 7574, "epoch": 85, "lr": 9.983186010412884e-05}
{"train_loss": 1.0546900033950806, "global_step": 7575, "epoch": 85, "lr": 9.983181259672388e-05}
{"train_loss": 0.7607847452163696, "global_step": 7576, "epoch": 85, "lr": 9.983176508261964e-05}
{"train_loss": 0.8163476586341858, "global_step": 7577, "epoch": 85, "lr": 9.983171756181613e-05}
{"train_loss": 0.7257250547409058, "global_step": 7578, "epoch": 85, "lr": 9.983167003431335e-05}
{"train_loss": 0.9241107702255249, "global_step": 7579, "epoch": 85, "lr": 9.983162250011131e-05}
{"train_loss": 0.8237968683242798, "global_step": 7580, "epoch": 85, "lr": 9.983157495921002e-05}
{"train_loss": 0.8629779815673828, "global_step": 7581, "epoch": 85, "lr": 9.983152741160948e-05}
{"train_loss": 0.7249557375907898, "global_step": 7582, "epoch": 85, "lr": 9.983147985730969e-05}
{"train_loss": 0.9125886559486389, "global_step": 7583, "epoch": 85, "lr": 9.983143229631067e-05}
{"train_loss": 0.7721124887466431, "global_step": 7584, "epoch": 85, "lr": 9.983138472861243e-05}
{"train_loss": 0.7365097403526306, "global_step": 7585, "epoch": 85, "lr": 9.983133715421496e-05}
{"train_loss": 0.9250093102455139, "global_step": 7586, "epoch": 85, "lr": 9.983128957311826e-05}
{"train_loss": 0.9846113324165344, "global_step": 7587, "epoch": 85, "lr": 9.983124198532238e-05}
{"train_loss": 1.116779088973999, "global_step": 7588, "epoch": 85, "lr": 9.983119439082727e-05}
{"train_loss": 0.924286425113678, "global_step": 7589, "epoch": 85, "lr": 9.9831146789633e-05}
{"train_loss": 0.9139095544815063, "global_step": 7590, "epoch": 85, "lr": 9.98310991817395e-05}
{"train_loss": 0.8433215022087097, "global_step": 7591, "epoch": 85, "lr": 9.983105156714683e-05}
{"train_loss": 0.8805252909660339, "global_step": 7592, "epoch": 85, "lr": 9.983100394585498e-05}
{"train_loss": 1.0027689933776855, "global_step": 7593, "epoch": 85, "lr": 9.983095631786397e-05}
{"train_loss": 0.7573796510696411, "global_step": 7594, "epoch": 85, "lr": 9.983090868317379e-05}
{"train_loss": 0.7228171229362488, "global_step": 7595, "epoch": 85, "lr": 9.983086104178445e-05}
{"train_loss": 0.9831416010856628, "global_step": 7596, "epoch": 85, "lr": 9.983081339369596e-05}
{"train_loss": 0.8725509643554688, "global_step": 7597, "epoch": 85, "lr": 9.98307657389083e-05}
{"train_loss": 1.030842900276184, "global_step": 7598, "epoch": 85, "lr": 9.983071807742154e-05}
{"train_loss": 0.9301097393035889, "global_step": 7599, "epoch": 85, "lr": 9.983067040923562e-05}
{"train_loss": 0.8069992661476135, "global_step": 7600, "epoch": 85, "lr": 9.983062273435059e-05}
{"train_loss": 0.8743607997894287, "global_step": 7601, "epoch": 85, "lr": 9.983057505276643e-05}
{"train_loss": 0.7403789162635803, "global_step": 7602, "epoch": 85, "lr": 9.983052736448316e-05}
{"train_loss": 0.8682162165641785, "global_step": 7603, "epoch": 85, "lr": 9.983047966950079e-05}
{"train_loss": 0.9130479097366333, "global_step": 7604, "epoch": 85, "lr": 9.983043196781931e-05}
{"train_loss": 0.9372973442077637, "global_step": 7605, "epoch": 85, "lr": 9.983038425943873e-05}
{"train_loss": 0.9353922009468079, "global_step": 7606, "epoch": 85, "lr": 9.983033654435908e-05}
{"train_loss": 0.9176079034805298, "global_step": 7607, "epoch": 85, "lr": 9.983028882258034e-05}
{"train_loss": 0.8871256113052368, "global_step": 7608, "epoch": 85, "lr": 9.983024109410253e-05}
{"train_loss": 0.9252849817276001, "global_step": 7609, "epoch": 85, "lr": 9.983019335892564e-05}
{"train_loss": 0.7767186164855957, "global_step": 7610, "epoch": 85, "lr": 9.983014561704968e-05}
{"train_loss": 0.9332942962646484, "global_step": 7611, "epoch": 85, "lr": 9.983009786847469e-05}
{"train_loss": 0.9574569463729858, "global_step": 7612, "epoch": 85, "lr": 9.983005011320063e-05}
{"train_loss": 0.8212029933929443, "global_step": 7613, "epoch": 85, "lr": 9.983000235122754e-05}
{"train_loss": 0.7222200036048889, "global_step": 7614, "epoch": 85, "lr": 9.982995458255541e-05}
{"train_loss": 0.8714430332183838, "global_step": 7615, "epoch": 85, "lr": 9.982990680718424e-05}
{"train_loss": 0.6797018051147461, "global_step": 7616, "epoch": 85, "lr": 9.982985902511407e-05}
{"train_loss": 0.9822096824645996, "global_step": 7617, "epoch": 85, "lr": 9.982981123634487e-05}
{"train_loss": 0.9524819254875183, "global_step": 7618, "epoch": 85, "lr": 9.982976344087666e-05}
{"train_loss": 0.9178968071937561, "global_step": 7619, "epoch": 85, "lr": 9.982971563870945e-05}
{"train_loss": 0.8490970730781555, "global_step": 7620, "epoch": 85, "lr": 9.982966782984322e-05}
{"train_loss": 0.9149788618087769, "global_step": 7621, "epoch": 85, "lr": 9.982962001427802e-05}
{"train_loss": 0.8775259256362915, "global_step": 7622, "epoch": 85, "lr": 9.982957219201383e-05}
{"train_loss": 0.8566962480545044, "global_step": 7623, "epoch": 85, "lr": 9.982952436305067e-05}
{"train_loss": 0.715799868106842, "global_step": 7624, "epoch": 85, "lr": 9.982947652738852e-05}
{"train_loss": 0.7114686369895935, "global_step": 7625, "epoch": 85, "lr": 9.98294286850274e-05}
{"train_loss": 0.7702934741973877, "global_step": 7626, "epoch": 85, "lr": 9.982938083596733e-05}
{"train_loss": 0.7900329232215881, "global_step": 7627, "epoch": 85, "lr": 9.982933298020833e-05}
{"train_loss": 0.7824870347976685, "global_step": 7628, "epoch": 85, "lr": 9.982928511775036e-05}
{"train_loss": 0.7752004861831665, "global_step": 7629, "epoch": 85, "lr": 9.982923724859346e-05}
{"train_loss": 1.0172078609466553, "global_step": 7630, "epoch": 85, "lr": 9.982918937273761e-05}
{"train_loss": 0.8414437770843506, "global_step": 7631, "epoch": 85, "lr": 9.982914149018285e-05}
{"train_loss": 0.8577507138252258, "global_step": 7632, "epoch": 85, "lr": 9.982909360092915e-05}
{"train_loss": 0.7080729603767395, "global_step": 7633, "epoch": 85, "lr": 9.982904570497656e-05}
{"train_loss": 1.0897619724273682, "global_step": 7634, "epoch": 85, "lr": 9.982899780232504e-05}
{"train_loss": 0.8074938058853149, "global_step": 7635, "epoch": 85, "lr": 9.982894989297461e-05}
{"train_loss": 0.9180117249488831, "global_step": 7636, "epoch": 85, "lr": 9.982890197692531e-05}
{"train_loss": 0.977177083492279, "global_step": 7637, "epoch": 85, "lr": 9.982885405417712e-05}
{"train_loss": 0.7740034461021423, "global_step": 7638, "epoch": 85, "lr": 9.982880612473005e-05}
{"train_loss": 0.9448198080062866, "global_step": 7639, "epoch": 85, "lr": 9.98287581885841e-05}
{"train_loss": 0.8791670799255371, "global_step": 7640, "epoch": 85, "lr": 9.982871024573928e-05}
{"train_loss": 0.7411289215087891, "global_step": 7641, "epoch": 85, "lr": 9.982866229619558e-05}
{"train_loss": 0.8659113049507141, "global_step": 7642, "epoch": 85, "lr": 9.982861433995305e-05}
{"train_loss": 0.7113000154495239, "global_step": 7643, "epoch": 85, "lr": 9.982856637701166e-05}
{"train_loss": 0.8829087018966675, "global_step": 7644, "epoch": 85, "lr": 9.982851840737141e-05}
{"train_loss": 0.9815734028816223, "global_step": 7645, "epoch": 85, "lr": 9.982847043103235e-05}
{"train_loss": 0.7473993301391602, "global_step": 7646, "epoch": 85, "lr": 9.982842244799444e-05}
{"train_loss": 0.6034890413284302, "global_step": 7647, "epoch": 85, "lr": 9.982837445825772e-05}
{"train_loss": 0.6493812799453735, "global_step": 7648, "epoch": 85, "lr": 9.982832646182217e-05}
{"train_loss": 0.7399065494537354, "global_step": 7649, "epoch": 85, "lr": 9.982827845868782e-05}
{"train_loss": 0.8390631675720215, "global_step": 7650, "epoch": 85, "lr": 9.982823044885467e-05}
{"train_loss": 0.8115333914756775, "global_step": 7651, "epoch": 85, "lr": 9.98281824323227e-05}
{"train_loss": 0.955440104007721, "global_step": 7652, "epoch": 85, "lr": 9.982813440909195e-05}
{"train_loss": 0.8524015137318814, "global_step": 7653, "epoch": 85, "lr": 9.982808637916242e-05, "val_loss": 1.1708228588104248, "train_action_mse_error": 45.16130828857422}
{"train_loss": 0.6655012965202332, "global_step": 7654, "epoch": 86, "lr": 9.982803834253411e-05}
{"train_loss": 0.8281822800636292, "global_step": 7655, "epoch": 86, "lr": 9.982799029920702e-05}
{"train_loss": 0.9608399868011475, "global_step": 7656, "epoch": 86, "lr": 9.982794224918118e-05}
{"train_loss": 0.9889946579933167, "global_step": 7657, "epoch": 86, "lr": 9.982789419245655e-05}
{"train_loss": 0.7962426543235779, "global_step": 7658, "epoch": 86, "lr": 9.98278461290332e-05}
{"train_loss": 0.9296656847000122, "global_step": 7659, "epoch": 86, "lr": 9.982779805891107e-05}
{"train_loss": 0.909433126449585, "global_step": 7660, "epoch": 86, "lr": 9.982774998209023e-05}
{"train_loss": 0.682828962802887, "global_step": 7661, "epoch": 86, "lr": 9.982770189857064e-05}
{"train_loss": 0.7702414393424988, "global_step": 7662, "epoch": 86, "lr": 9.982765380835232e-05}
{"train_loss": 1.1619317531585693, "global_step": 7663, "epoch": 86, "lr": 9.98276057114353e-05}
{"train_loss": 0.9411080479621887, "global_step": 7664, "epoch": 86, "lr": 9.982755760781954e-05}
{"train_loss": 0.713139533996582, "global_step": 7665, "epoch": 86, "lr": 9.982750949750509e-05}
{"train_loss": 0.6546438336372375, "global_step": 7666, "epoch": 86, "lr": 9.982746138049194e-05}
{"train_loss": 0.8604669570922852, "global_step": 7667, "epoch": 86, "lr": 9.982741325678009e-05}
{"train_loss": 0.8262885212898254, "global_step": 7668, "epoch": 86, "lr": 9.982736512636955e-05}
{"train_loss": 0.8818663358688354, "global_step": 7669, "epoch": 86, "lr": 9.982731698926033e-05}
{"train_loss": 0.7566497325897217, "global_step": 7670, "epoch": 86, "lr": 9.982726884545243e-05}
{"train_loss": 0.738413393497467, "global_step": 7671, "epoch": 86, "lr": 9.982722069494586e-05}
{"train_loss": 0.7292259931564331, "global_step": 7672, "epoch": 86, "lr": 9.982717253774064e-05}
{"train_loss": 0.8210461735725403, "global_step": 7673, "epoch": 86, "lr": 9.982712437383674e-05}
{"train_loss": 0.8221731185913086, "global_step": 7674, "epoch": 86, "lr": 9.982707620323421e-05}
{"train_loss": 0.7019726037979126, "global_step": 7675, "epoch": 86, "lr": 9.982702802593303e-05}
{"train_loss": 0.6614358425140381, "global_step": 7676, "epoch": 86, "lr": 9.98269798419332e-05}
{"train_loss": 0.6653698682785034, "global_step": 7677, "epoch": 86, "lr": 9.982693165123477e-05}
{"train_loss": 0.7460749745368958, "global_step": 7678, "epoch": 86, "lr": 9.98268834538377e-05}
{"train_loss": 0.7566519975662231, "global_step": 7679, "epoch": 86, "lr": 9.982683524974201e-05}
{"train_loss": 0.805471658706665, "global_step": 7680, "epoch": 86, "lr": 9.982678703894772e-05}
{"train_loss": 0.6118717789649963, "global_step": 7681, "epoch": 86, "lr": 9.982673882145482e-05}
{"train_loss": 0.740778386592865, "global_step": 7682, "epoch": 86, "lr": 9.982669059726332e-05}
{"train_loss": 0.6394913792610168, "global_step": 7683, "epoch": 86, "lr": 9.982664236637323e-05}
{"train_loss": 0.768182635307312, "global_step": 7684, "epoch": 86, "lr": 9.982659412878457e-05}
{"train_loss": 0.7542363405227661, "global_step": 7685, "epoch": 86, "lr": 9.982654588449732e-05}
{"train_loss": 0.8797910809516907, "global_step": 7686, "epoch": 86, "lr": 9.982649763351148e-05}
{"train_loss": 0.915763795375824, "global_step": 7687, "epoch": 86, "lr": 9.982644937582711e-05}
{"train_loss": 0.6869929432868958, "global_step": 7688, "epoch": 86, "lr": 9.982640111144415e-05}
{"train_loss": 1.0326056480407715, "global_step": 7689, "epoch": 86, "lr": 9.982635284036264e-05}
{"train_loss": 0.7544054985046387, "global_step": 7690, "epoch": 86, "lr": 9.98263045625826e-05}
{"train_loss": 0.8271637558937073, "global_step": 7691, "epoch": 86, "lr": 9.982625627810402e-05}
{"train_loss": 0.9700901508331299, "global_step": 7692, "epoch": 86, "lr": 9.98262079869269e-05}
{"train_loss": 0.801629364490509, "global_step": 7693, "epoch": 86, "lr": 9.982615968905125e-05}
{"train_loss": 0.7198033332824707, "global_step": 7694, "epoch": 86, "lr": 9.982611138447708e-05}
{"train_loss": 0.8679370284080505, "global_step": 7695, "epoch": 86, "lr": 9.982606307320442e-05}
{"train_loss": 0.7216382622718811, "global_step": 7696, "epoch": 86, "lr": 9.982601475523324e-05}
{"train_loss": 0.6853815317153931, "global_step": 7697, "epoch": 86, "lr": 9.982596643056355e-05}
{"train_loss": 0.6658552885055542, "global_step": 7698, "epoch": 86, "lr": 9.982591809919536e-05}
{"train_loss": 0.6567437648773193, "global_step": 7699, "epoch": 86, "lr": 9.982586976112869e-05}
{"train_loss": 0.839259147644043, "global_step": 7700, "epoch": 86, "lr": 9.982582141636354e-05}
{"train_loss": 0.8303455710411072, "global_step": 7701, "epoch": 86, "lr": 9.982577306489992e-05}
{"train_loss": 0.7031318545341492, "global_step": 7702, "epoch": 86, "lr": 9.982572470673783e-05}
{"train_loss": 0.97238689661026, "global_step": 7703, "epoch": 86, "lr": 9.982567634187728e-05}
{"train_loss": 0.9049987196922302, "global_step": 7704, "epoch": 86, "lr": 9.982562797031828e-05}
{"train_loss": 0.9516421556472778, "global_step": 7705, "epoch": 86, "lr": 9.982557959206083e-05}
{"train_loss": 0.9172408580780029, "global_step": 7706, "epoch": 86, "lr": 9.982553120710493e-05}
{"train_loss": 0.9426204562187195, "global_step": 7707, "epoch": 86, "lr": 9.982548281545058e-05}
{"train_loss": 0.9137337803840637, "global_step": 7708, "epoch": 86, "lr": 9.982543441709784e-05}
{"train_loss": 0.878669798374176, "global_step": 7709, "epoch": 86, "lr": 9.982538601204665e-05}
{"train_loss": 1.0353741645812988, "global_step": 7710, "epoch": 86, "lr": 9.982533760029706e-05}
{"train_loss": 0.8677443861961365, "global_step": 7711, "epoch": 86, "lr": 9.982528918184904e-05}
{"train_loss": 0.6738950610160828, "global_step": 7712, "epoch": 86, "lr": 9.982524075670263e-05}
{"train_loss": 0.5617595314979553, "global_step": 7713, "epoch": 86, "lr": 9.982519232485784e-05}
{"train_loss": 0.8937345147132874, "global_step": 7714, "epoch": 86, "lr": 9.982514388631463e-05}
{"train_loss": 0.8706645369529724, "global_step": 7715, "epoch": 86, "lr": 9.982509544107304e-05}
{"train_loss": 0.8244590163230896, "global_step": 7716, "epoch": 86, "lr": 9.982504698913308e-05}
{"train_loss": 0.7190069556236267, "global_step": 7717, "epoch": 86, "lr": 9.982499853049476e-05}
{"train_loss": 0.7294512391090393, "global_step": 7718, "epoch": 86, "lr": 9.982495006515807e-05}
{"train_loss": 0.7605270743370056, "global_step": 7719, "epoch": 86, "lr": 9.982490159312302e-05}
{"train_loss": 0.7798676490783691, "global_step": 7720, "epoch": 86, "lr": 9.982485311438962e-05}
{"train_loss": 0.7969698309898376, "global_step": 7721, "epoch": 86, "lr": 9.982480462895788e-05}
{"train_loss": 0.8744070529937744, "global_step": 7722, "epoch": 86, "lr": 9.98247561368278e-05}
{"train_loss": 0.8823091983795166, "global_step": 7723, "epoch": 86, "lr": 9.982470763799938e-05}
{"train_loss": 0.713468074798584, "global_step": 7724, "epoch": 86, "lr": 9.982465913247265e-05}
{"train_loss": 0.7067268490791321, "global_step": 7725, "epoch": 86, "lr": 9.982461062024759e-05}
{"train_loss": 0.8916864991188049, "global_step": 7726, "epoch": 86, "lr": 9.982456210132422e-05}
{"train_loss": 0.7693930864334106, "global_step": 7727, "epoch": 86, "lr": 9.982451357570255e-05}
{"train_loss": 0.9230623841285706, "global_step": 7728, "epoch": 86, "lr": 9.982446504338259e-05}
{"train_loss": 1.0915465354919434, "global_step": 7729, "epoch": 86, "lr": 9.982441650436432e-05}
{"train_loss": 0.7172494530677795, "global_step": 7730, "epoch": 86, "lr": 9.982436795864778e-05}
{"train_loss": 0.9643523097038269, "global_step": 7731, "epoch": 86, "lr": 9.982431940623295e-05}
{"train_loss": 0.7775587439537048, "global_step": 7732, "epoch": 86, "lr": 9.982427084711985e-05}
{"train_loss": 0.9070517420768738, "global_step": 7733, "epoch": 86, "lr": 9.982422228130848e-05}
{"train_loss": 0.8251752257347107, "global_step": 7734, "epoch": 86, "lr": 9.982417370879886e-05}
{"train_loss": 0.8929160833358765, "global_step": 7735, "epoch": 86, "lr": 9.982412512959099e-05}
{"train_loss": 0.8459241986274719, "global_step": 7736, "epoch": 86, "lr": 9.982407654368485e-05}
{"train_loss": 0.6871963143348694, "global_step": 7737, "epoch": 86, "lr": 9.98240279510805e-05}
{"train_loss": 0.7352074980735779, "global_step": 7738, "epoch": 86, "lr": 9.98239793517779e-05}
{"train_loss": 0.8444620966911316, "global_step": 7739, "epoch": 86, "lr": 9.982393074577708e-05}
{"train_loss": 0.8040695190429688, "global_step": 7740, "epoch": 86, "lr": 9.982388213307803e-05}
{"train_loss": 0.8826329708099365, "global_step": 7741, "epoch": 86, "lr": 9.982383351368076e-05}
{"train_loss": 0.8149267026547635, "global_step": 7742, "epoch": 86, "lr": 9.98237848875853e-05, "val_loss": 1.1813478469848633}
{"train_loss": 0.8267481327056885, "global_step": 7743, "epoch": 87, "lr": 9.982373625479163e-05}
{"train_loss": 0.8191977739334106, "global_step": 7744, "epoch": 87, "lr": 9.982368761529978e-05}
{"train_loss": 0.7247130870819092, "global_step": 7745, "epoch": 87, "lr": 9.982363896910973e-05}
{"train_loss": 0.7738933563232422, "global_step": 7746, "epoch": 87, "lr": 9.98235903162215e-05}
{"train_loss": 0.8024804592132568, "global_step": 7747, "epoch": 87, "lr": 9.98235416566351e-05}
{"train_loss": 0.8410915732383728, "global_step": 7748, "epoch": 87, "lr": 9.982349299035052e-05}
{"train_loss": 0.6217560768127441, "global_step": 7749, "epoch": 87, "lr": 9.982344431736779e-05}
{"train_loss": 0.7542096972465515, "global_step": 7750, "epoch": 87, "lr": 9.98233956376869e-05}
{"train_loss": 0.8317292332649231, "global_step": 7751, "epoch": 87, "lr": 9.982334695130788e-05}
{"train_loss": 0.6542405486106873, "global_step": 7752, "epoch": 87, "lr": 9.982329825823069e-05}
{"train_loss": 0.773287296295166, "global_step": 7753, "epoch": 87, "lr": 9.982324955845539e-05}
{"train_loss": 0.8725114464759827, "global_step": 7754, "epoch": 87, "lr": 9.982320085198193e-05}
{"train_loss": 1.012331247329712, "global_step": 7755, "epoch": 87, "lr": 9.982315213881038e-05}
{"train_loss": 0.7763122320175171, "global_step": 7756, "epoch": 87, "lr": 9.98231034189407e-05}
{"train_loss": 0.7218093872070312, "global_step": 7757, "epoch": 87, "lr": 9.982305469237291e-05}
{"train_loss": 0.7052512764930725, "global_step": 7758, "epoch": 87, "lr": 9.982300595910702e-05}
{"train_loss": 0.7586876153945923, "global_step": 7759, "epoch": 87, "lr": 9.982295721914303e-05}
{"train_loss": 0.7030310034751892, "global_step": 7760, "epoch": 87, "lr": 9.982290847248097e-05}
{"train_loss": 0.8094870448112488, "global_step": 7761, "epoch": 87, "lr": 9.98228597191208e-05}
{"train_loss": 0.8198590874671936, "global_step": 7762, "epoch": 87, "lr": 9.982281095906256e-05}
{"train_loss": 0.8188589215278625, "global_step": 7763, "epoch": 87, "lr": 9.982276219230627e-05}
{"train_loss": 0.8271779417991638, "global_step": 7764, "epoch": 87, "lr": 9.98227134188519e-05}
{"train_loss": 0.7315723896026611, "global_step": 7765, "epoch": 87, "lr": 9.982266463869947e-05}
{"train_loss": 0.8355909585952759, "global_step": 7766, "epoch": 87, "lr": 9.9822615851849e-05}
{"train_loss": 0.7680222392082214, "global_step": 7767, "epoch": 87, "lr": 9.982256705830048e-05}
{"train_loss": 0.8471863865852356, "global_step": 7768, "epoch": 87, "lr": 9.982251825805394e-05}
{"train_loss": 0.7940911650657654, "global_step": 7769, "epoch": 87, "lr": 9.982246945110934e-05}
{"train_loss": 0.6761916279792786, "global_step": 7770, "epoch": 87, "lr": 9.982242063746675e-05}
{"train_loss": 0.6033804416656494, "global_step": 7771, "epoch": 87, "lr": 9.982237181712611e-05}
{"train_loss": 0.7291902899742126, "global_step": 7772, "epoch": 87, "lr": 9.982232299008748e-05}
{"train_loss": 0.8793884515762329, "global_step": 7773, "epoch": 87, "lr": 9.982227415635084e-05}
{"train_loss": 0.6966752409934998, "global_step": 7774, "epoch": 87, "lr": 9.98222253159162e-05}
{"train_loss": 0.8174448609352112, "global_step": 7775, "epoch": 87, "lr": 9.982217646878358e-05}
{"train_loss": 0.7531685829162598, "global_step": 7776, "epoch": 87, "lr": 9.982212761495295e-05}
{"train_loss": 0.7683005928993225, "global_step": 7777, "epoch": 87, "lr": 9.982207875442438e-05}
{"train_loss": 0.9175137877464294, "global_step": 7778, "epoch": 87, "lr": 9.982202988719781e-05}
{"train_loss": 0.7652254700660706, "global_step": 7779, "epoch": 87, "lr": 9.982198101327328e-05}
{"train_loss": 0.7265115976333618, "global_step": 7780, "epoch": 87, "lr": 9.98219321326508e-05}
{"train_loss": 0.8821309208869934, "global_step": 7781, "epoch": 87, "lr": 9.982188324533035e-05}
{"train_loss": 0.7135112881660461, "global_step": 7782, "epoch": 87, "lr": 9.982183435131198e-05}
{"train_loss": 0.7297850847244263, "global_step": 7783, "epoch": 87, "lr": 9.982178545059566e-05}
{"train_loss": 0.5742493271827698, "global_step": 7784, "epoch": 87, "lr": 9.98217365431814e-05}
{"train_loss": 0.7247637510299683, "global_step": 7785, "epoch": 87, "lr": 9.982168762906922e-05}
{"train_loss": 0.935340166091919, "global_step": 7786, "epoch": 87, "lr": 9.982163870825914e-05}
{"train_loss": 0.8039900064468384, "global_step": 7787, "epoch": 87, "lr": 9.982158978075114e-05}
{"train_loss": 0.7531477808952332, "global_step": 7788, "epoch": 87, "lr": 9.982154084654522e-05}
{"train_loss": 0.8494173884391785, "global_step": 7789, "epoch": 87, "lr": 9.98214919056414e-05}
{"train_loss": 1.0452635288238525, "global_step": 7790, "epoch": 87, "lr": 9.98214429580397e-05}
{"train_loss": 1.0987281799316406, "global_step": 7791, "epoch": 87, "lr": 9.982139400374011e-05}
{"train_loss": 0.8392171859741211, "global_step": 7792, "epoch": 87, "lr": 9.982134504274265e-05}
{"train_loss": 0.9930572509765625, "global_step": 7793, "epoch": 87, "lr": 9.982129607504731e-05}
{"train_loss": 0.7490428686141968, "global_step": 7794, "epoch": 87, "lr": 9.98212471006541e-05}
{"train_loss": 0.8067514300346375, "global_step": 7795, "epoch": 87, "lr": 9.982119811956304e-05}
{"train_loss": 0.8559790253639221, "global_step": 7796, "epoch": 87, "lr": 9.982114913177411e-05}
{"train_loss": 0.8273502588272095, "global_step": 7797, "epoch": 87, "lr": 9.982110013728734e-05}
{"train_loss": 0.9047678709030151, "global_step": 7798, "epoch": 87, "lr": 9.982105113610274e-05}
{"train_loss": 0.9212570786476135, "global_step": 7799, "epoch": 87, "lr": 9.98210021282203e-05}
{"train_loss": 0.8198376297950745, "global_step": 7800, "epoch": 87, "lr": 9.982095311364003e-05}
{"train_loss": 0.7483489513397217, "global_step": 7801, "epoch": 87, "lr": 9.982090409236195e-05}
{"train_loss": 0.8181741833686829, "global_step": 7802, "epoch": 87, "lr": 9.982085506438605e-05}
{"train_loss": 1.0822190046310425, "global_step": 7803, "epoch": 87, "lr": 9.982080602971234e-05}
{"train_loss": 0.9357236623764038, "global_step": 7804, "epoch": 87, "lr": 9.982075698834083e-05}
{"train_loss": 0.7615970373153687, "global_step": 7805, "epoch": 87, "lr": 9.982070794027154e-05}
{"train_loss": 0.910116970539093, "global_step": 7806, "epoch": 87, "lr": 9.982065888550444e-05}
{"train_loss": 0.8811026215553284, "global_step": 7807, "epoch": 87, "lr": 9.982060982403958e-05}
{"train_loss": 0.8847033977508545, "global_step": 7808, "epoch": 87, "lr": 9.982056075587694e-05}
{"train_loss": 1.0412583351135254, "global_step": 7809, "epoch": 87, "lr": 9.982051168101654e-05}
{"train_loss": 0.6879758834838867, "global_step": 7810, "epoch": 87, "lr": 9.982046259945836e-05}
{"train_loss": 0.9340869784355164, "global_step": 7811, "epoch": 87, "lr": 9.982041351120244e-05}
{"train_loss": 0.8525230884552002, "global_step": 7812, "epoch": 87, "lr": 9.982036441624878e-05}
{"train_loss": 1.0821202993392944, "global_step": 7813, "epoch": 87, "lr": 9.982031531459737e-05}
{"train_loss": 0.8132696747779846, "global_step": 7814, "epoch": 87, "lr": 9.982026620624822e-05}
{"train_loss": 0.705411970615387, "global_step": 7815, "epoch": 87, "lr": 9.982021709120136e-05}
{"train_loss": 0.7858709096908569, "global_step": 7816, "epoch": 87, "lr": 9.982016796945677e-05}
{"train_loss": 0.7053382992744446, "global_step": 7817, "epoch": 87, "lr": 9.982011884101445e-05}
{"train_loss": 0.9131304025650024, "global_step": 7818, "epoch": 87, "lr": 9.982006970587444e-05}
{"train_loss": 0.8594695925712585, "global_step": 7819, "epoch": 87, "lr": 9.982002056403672e-05}
{"train_loss": 0.7197516560554504, "global_step": 7820, "epoch": 87, "lr": 9.981997141550132e-05}
{"train_loss": 0.714981198310852, "global_step": 7821, "epoch": 87, "lr": 9.981992226026822e-05}
{"train_loss": 0.8371981978416443, "global_step": 7822, "epoch": 87, "lr": 9.981987309833743e-05}
{"train_loss": 0.7687550783157349, "global_step": 7823, "epoch": 87, "lr": 9.981982392970899e-05}
{"train_loss": 0.9503353834152222, "global_step": 7824, "epoch": 87, "lr": 9.981977475438286e-05}
{"train_loss": 1.0149728059768677, "global_step": 7825, "epoch": 87, "lr": 9.981972557235909e-05}
{"train_loss": 0.7444850206375122, "global_step": 7826, "epoch": 87, "lr": 9.981967638363763e-05}
{"train_loss": 0.7857086062431335, "global_step": 7827, "epoch": 87, "lr": 9.981962718821854e-05}
{"train_loss": 0.6505417227745056, "global_step": 7828, "epoch": 87, "lr": 9.981957798610182e-05}
{"train_loss": 0.7839930653572083, "global_step": 7829, "epoch": 87, "lr": 9.981952877728746e-05}
{"train_loss": 0.931757390499115, "global_step": 7830, "epoch": 87, "lr": 9.981947956177546e-05}
{"train_loss": 0.814863267909275, "global_step": 7831, "epoch": 87, "lr": 9.981943033956585e-05, "val_loss": 1.1630054712295532}
{"train_loss": 0.9743803143501282, "global_step": 7832, "epoch": 88, "lr": 9.981938111065863e-05}
{"train_loss": 0.8042067885398865, "global_step": 7833, "epoch": 88, "lr": 9.981933187505378e-05}
{"train_loss": 0.6373368501663208, "global_step": 7834, "epoch": 88, "lr": 9.981928263275134e-05}
{"train_loss": 0.6605414748191833, "global_step": 7835, "epoch": 88, "lr": 9.98192333837513e-05}
{"train_loss": 0.7458016872406006, "global_step": 7836, "epoch": 88, "lr": 9.981918412805369e-05}
{"train_loss": 0.568565309047699, "global_step": 7837, "epoch": 88, "lr": 9.981913486565848e-05}
{"train_loss": 0.8556534647941589, "global_step": 7838, "epoch": 88, "lr": 9.98190855965657e-05}
{"train_loss": 0.8902560472488403, "global_step": 7839, "epoch": 88, "lr": 9.981903632077534e-05}
{"train_loss": 0.8564611077308655, "global_step": 7840, "epoch": 88, "lr": 9.981898703828742e-05}
{"train_loss": 0.5566238760948181, "global_step": 7841, "epoch": 88, "lr": 9.981893774910196e-05}
{"train_loss": 1.0710909366607666, "global_step": 7842, "epoch": 88, "lr": 9.981888845321894e-05}
{"train_loss": 0.7652671337127686, "global_step": 7843, "epoch": 88, "lr": 9.981883915063839e-05}
{"train_loss": 0.7163160443305969, "global_step": 7844, "epoch": 88, "lr": 9.98187898413603e-05}
{"train_loss": 0.6717967391014099, "global_step": 7845, "epoch": 88, "lr": 9.981874052538466e-05}
{"train_loss": 0.731467068195343, "global_step": 7846, "epoch": 88, "lr": 9.981869120271152e-05}
{"train_loss": 0.686080276966095, "global_step": 7847, "epoch": 88, "lr": 9.981864187334085e-05}
{"train_loss": 0.7930864095687866, "global_step": 7848, "epoch": 88, "lr": 9.981859253727268e-05}
{"train_loss": 0.7799668908119202, "global_step": 7849, "epoch": 88, "lr": 9.9818543194507e-05}
{"train_loss": 0.9814366698265076, "global_step": 7850, "epoch": 88, "lr": 9.981849384504385e-05}
{"train_loss": 0.7698723077774048, "global_step": 7851, "epoch": 88, "lr": 9.981844448888317e-05}
{"train_loss": 0.7329216003417969, "global_step": 7852, "epoch": 88, "lr": 9.981839512602504e-05}
{"train_loss": 0.9001962542533875, "global_step": 7853, "epoch": 88, "lr": 9.981834575646942e-05}
{"train_loss": 0.8669820427894592, "global_step": 7854, "epoch": 88, "lr": 9.981829638021633e-05}
{"train_loss": 0.6382008194923401, "global_step": 7855, "epoch": 88, "lr": 9.981824699726578e-05}
{"train_loss": 0.8392277956008911, "global_step": 7856, "epoch": 88, "lr": 9.981819760761777e-05}
{"train_loss": 0.9156419634819031, "global_step": 7857, "epoch": 88, "lr": 9.981814821127233e-05}
{"train_loss": 0.6910722255706787, "global_step": 7858, "epoch": 88, "lr": 9.981809880822942e-05}
{"train_loss": 0.7148623466491699, "global_step": 7859, "epoch": 88, "lr": 9.981804939848908e-05}
{"train_loss": 0.8127738237380981, "global_step": 7860, "epoch": 88, "lr": 9.981799998205133e-05}
{"train_loss": 0.7807738780975342, "global_step": 7861, "epoch": 88, "lr": 9.981795055891614e-05}
{"train_loss": 0.8961454629898071, "global_step": 7862, "epoch": 88, "lr": 9.981790112908354e-05}
{"train_loss": 0.682466447353363, "global_step": 7863, "epoch": 88, "lr": 9.981785169255352e-05}
{"train_loss": 0.7845680117607117, "global_step": 7864, "epoch": 88, "lr": 9.98178022493261e-05}
{"train_loss": 0.8207827806472778, "global_step": 7865, "epoch": 88, "lr": 9.981775279940129e-05}
{"train_loss": 0.9315436482429504, "global_step": 7866, "epoch": 88, "lr": 9.981770334277909e-05}
{"train_loss": 0.8943181037902832, "global_step": 7867, "epoch": 88, "lr": 9.98176538794595e-05}
{"train_loss": 0.8716289401054382, "global_step": 7868, "epoch": 88, "lr": 9.981760440944255e-05}
{"train_loss": 0.9784746170043945, "global_step": 7869, "epoch": 88, "lr": 9.981755493272821e-05}
{"train_loss": 0.9009296894073486, "global_step": 7870, "epoch": 88, "lr": 9.981750544931651e-05}
{"train_loss": 0.8212974667549133, "global_step": 7871, "epoch": 88, "lr": 9.981745595920747e-05}
{"train_loss": 0.9813870787620544, "global_step": 7872, "epoch": 88, "lr": 9.981740646240106e-05}
{"train_loss": 0.9101340174674988, "global_step": 7873, "epoch": 88, "lr": 9.98173569588973e-05}
{"train_loss": 0.7545891404151917, "global_step": 7874, "epoch": 88, "lr": 9.981730744869625e-05}
{"train_loss": 0.8313840627670288, "global_step": 7875, "epoch": 88, "lr": 9.981725793179784e-05}
{"train_loss": 0.7784273028373718, "global_step": 7876, "epoch": 88, "lr": 9.98172084082021e-05}
{"train_loss": 0.82407146692276, "global_step": 7877, "epoch": 88, "lr": 9.981715887790905e-05}
{"train_loss": 0.8254690766334534, "global_step": 7878, "epoch": 88, "lr": 9.981710934091868e-05}
{"train_loss": 0.7513955235481262, "global_step": 7879, "epoch": 88, "lr": 9.981705979723102e-05}
{"train_loss": 0.9035683274269104, "global_step": 7880, "epoch": 88, "lr": 9.981701024684607e-05}
{"train_loss": 0.8417689800262451, "global_step": 7881, "epoch": 88, "lr": 9.981696068976382e-05}
{"train_loss": 0.9537119269371033, "global_step": 7882, "epoch": 88, "lr": 9.981691112598428e-05}
{"train_loss": 0.7332812547683716, "global_step": 7883, "epoch": 88, "lr": 9.981686155550747e-05}
{"train_loss": 0.7158843278884888, "global_step": 7884, "epoch": 88, "lr": 9.981681197833339e-05}
{"train_loss": 0.7512606978416443, "global_step": 7885, "epoch": 88, "lr": 9.981676239446205e-05}
{"train_loss": 0.6636641025543213, "global_step": 7886, "epoch": 88, "lr": 9.981671280389346e-05}
{"train_loss": 0.745568573474884, "global_step": 7887, "epoch": 88, "lr": 9.981666320662762e-05}
{"train_loss": 0.6635674834251404, "global_step": 7888, "epoch": 88, "lr": 9.981661360266452e-05}
{"train_loss": 0.7051668763160706, "global_step": 7889, "epoch": 88, "lr": 9.981656399200419e-05}
{"train_loss": 0.747860848903656, "global_step": 7890, "epoch": 88, "lr": 9.981651437464663e-05}
{"train_loss": 0.8513124585151672, "global_step": 7891, "epoch": 88, "lr": 9.981646475059186e-05}
{"train_loss": 0.8438597917556763, "global_step": 7892, "epoch": 88, "lr": 9.981641511983985e-05}
{"train_loss": 0.7347460985183716, "global_step": 7893, "epoch": 88, "lr": 9.981636548239064e-05}
{"train_loss": 0.8591618537902832, "global_step": 7894, "epoch": 88, "lr": 9.981631583824424e-05}
{"train_loss": 0.7116096615791321, "global_step": 7895, "epoch": 88, "lr": 9.981626618740063e-05}
{"train_loss": 0.8622449636459351, "global_step": 7896, "epoch": 88, "lr": 9.981621652985985e-05}
{"train_loss": 0.8214017748832703, "global_step": 7897, "epoch": 88, "lr": 9.981616686562185e-05}
{"train_loss": 0.8992867469787598, "global_step": 7898, "epoch": 88, "lr": 9.98161171946867e-05}
{"train_loss": 0.7950795888900757, "global_step": 7899, "epoch": 88, "lr": 9.981606751705437e-05}
{"train_loss": 0.9203956127166748, "global_step": 7900, "epoch": 88, "lr": 9.981601783272489e-05}
{"train_loss": 0.83668452501297, "global_step": 7901, "epoch": 88, "lr": 9.981596814169823e-05}
{"train_loss": 0.6617698669433594, "global_step": 7902, "epoch": 88, "lr": 9.981591844397444e-05}
{"train_loss": 0.7313536405563354, "global_step": 7903, "epoch": 88, "lr": 9.981586873955351e-05}
{"train_loss": 0.9015471339225769, "global_step": 7904, "epoch": 88, "lr": 9.981581902843542e-05}
{"train_loss": 0.8071721196174622, "global_step": 7905, "epoch": 88, "lr": 9.981576931062022e-05}
{"train_loss": 0.797808051109314, "global_step": 7906, "epoch": 88, "lr": 9.98157195861079e-05}
{"train_loss": 0.9154617190361023, "global_step": 7907, "epoch": 88, "lr": 9.981566985489846e-05}
{"train_loss": 0.8803917169570923, "global_step": 7908, "epoch": 88, "lr": 9.98156201169919e-05}
{"train_loss": 0.6786844730377197, "global_step": 7909, "epoch": 88, "lr": 9.981557037238824e-05}
{"train_loss": 0.7710003852844238, "global_step": 7910, "epoch": 88, "lr": 9.981552062108749e-05}
{"train_loss": 0.815223217010498, "global_step": 7911, "epoch": 88, "lr": 9.981547086308964e-05}
{"train_loss": 0.7954567670822144, "global_step": 7912, "epoch": 88, "lr": 9.981542109839473e-05}
{"train_loss": 0.7760032415390015, "global_step": 7913, "epoch": 88, "lr": 9.981537132700272e-05}
{"train_loss": 0.8060779571533203, "global_step": 7914, "epoch": 88, "lr": 9.981532154891364e-05}
{"train_loss": 0.8258060216903687, "global_step": 7915, "epoch": 88, "lr": 9.981527176412751e-05}
{"train_loss": 0.7814933061599731, "global_step": 7916, "epoch": 88, "lr": 9.981522197264432e-05}
{"train_loss": 0.9408295750617981, "global_step": 7917, "epoch": 88, "lr": 9.981517217446409e-05}
{"train_loss": 0.7414311170578003, "global_step": 7918, "epoch": 88, "lr": 9.981512236958681e-05}
{"train_loss": 0.7837637066841125, "global_step": 7919, "epoch": 88, "lr": 9.981507255801249e-05}
{"train_loss": 0.8054338713710227, "global_step": 7920, "epoch": 88, "lr": 9.981502273974115e-05, "val_loss": 1.165209412574768}
{"train_loss": 0.7542263269424438, "global_step": 7921, "epoch": 89, "lr": 9.981497291477277e-05}
{"train_loss": 0.8972839713096619, "global_step": 7922, "epoch": 89, "lr": 9.981492308310738e-05}
{"train_loss": 0.7807501554489136, "global_step": 7923, "epoch": 89, "lr": 9.981487324474499e-05}
{"train_loss": 0.690605878829956, "global_step": 7924, "epoch": 89, "lr": 9.98148233996856e-05}
{"train_loss": 0.7347785830497742, "global_step": 7925, "epoch": 89, "lr": 9.981477354792919e-05}
{"train_loss": 0.7992680072784424, "global_step": 7926, "epoch": 89, "lr": 9.981472368947582e-05}
{"train_loss": 0.8570514917373657, "global_step": 7927, "epoch": 89, "lr": 9.981467382432546e-05}
{"train_loss": 0.7053315043449402, "global_step": 7928, "epoch": 89, "lr": 9.981462395247811e-05}
{"train_loss": 0.7080768346786499, "global_step": 7929, "epoch": 89, "lr": 9.981457407393379e-05}
{"train_loss": 0.8283252716064453, "global_step": 7930, "epoch": 89, "lr": 9.981452418869252e-05}
{"train_loss": 0.7382888197898865, "global_step": 7931, "epoch": 89, "lr": 9.98144742967543e-05}
{"train_loss": 0.7528226971626282, "global_step": 7932, "epoch": 89, "lr": 9.981442439811912e-05}
{"train_loss": 0.823256254196167, "global_step": 7933, "epoch": 89, "lr": 9.9814374492787e-05}
{"train_loss": 0.9015394449234009, "global_step": 7934, "epoch": 89, "lr": 9.981432458075795e-05}
{"train_loss": 0.6222951412200928, "global_step": 7935, "epoch": 89, "lr": 9.981427466203197e-05}
{"train_loss": 0.7500094175338745, "global_step": 7936, "epoch": 89, "lr": 9.981422473660906e-05}
{"train_loss": 0.8001436591148376, "global_step": 7937, "epoch": 89, "lr": 9.981417480448925e-05}
{"train_loss": 0.9747790098190308, "global_step": 7938, "epoch": 89, "lr": 9.98141248656725e-05}
{"train_loss": 0.8104113340377808, "global_step": 7939, "epoch": 89, "lr": 9.981407492015888e-05}
{"train_loss": 0.6343920826911926, "global_step": 7940, "epoch": 89, "lr": 9.981402496794837e-05}
{"train_loss": 0.8613625764846802, "global_step": 7941, "epoch": 89, "lr": 9.981397500904094e-05}
{"train_loss": 0.8258923292160034, "global_step": 7942, "epoch": 89, "lr": 9.981392504343664e-05}
{"train_loss": 0.7828124761581421, "global_step": 7943, "epoch": 89, "lr": 9.981387507113547e-05}
{"train_loss": 0.7771603465080261, "global_step": 7944, "epoch": 89, "lr": 9.981382509213744e-05}
{"train_loss": 0.7219088077545166, "global_step": 7945, "epoch": 89, "lr": 9.981377510644252e-05}
{"train_loss": 1.0064170360565186, "global_step": 7946, "epoch": 89, "lr": 9.981372511405077e-05}
{"train_loss": 0.6949755549430847, "global_step": 7947, "epoch": 89, "lr": 9.981367511496216e-05}
{"train_loss": 0.6689296960830688, "global_step": 7948, "epoch": 89, "lr": 9.981362510917672e-05}
{"train_loss": 0.7946392893791199, "global_step": 7949, "epoch": 89, "lr": 9.981357509669444e-05}
{"train_loss": 0.825337827205658, "global_step": 7950, "epoch": 89, "lr": 9.981352507751534e-05}
{"train_loss": 0.6115767359733582, "global_step": 7951, "epoch": 89, "lr": 9.98134750516394e-05}
{"train_loss": 0.8540710210800171, "global_step": 7952, "epoch": 89, "lr": 9.981342501906665e-05}
{"train_loss": 0.8531208038330078, "global_step": 7953, "epoch": 89, "lr": 9.98133749797971e-05}
{"train_loss": 0.9228644967079163, "global_step": 7954, "epoch": 89, "lr": 9.981332493383074e-05}
{"train_loss": 0.8916088342666626, "global_step": 7955, "epoch": 89, "lr": 9.981327488116759e-05}
{"train_loss": 0.9005251526832581, "global_step": 7956, "epoch": 89, "lr": 9.981322482180766e-05}
{"train_loss": 1.025072693824768, "global_step": 7957, "epoch": 89, "lr": 9.981317475575094e-05}
{"train_loss": 0.8424792885780334, "global_step": 7958, "epoch": 89, "lr": 9.981312468299745e-05}
{"train_loss": 0.6670644879341125, "global_step": 7959, "epoch": 89, "lr": 9.981307460354719e-05}
{"train_loss": 0.7231840491294861, "global_step": 7960, "epoch": 89, "lr": 9.981302451740018e-05}
{"train_loss": 0.9038050174713135, "global_step": 7961, "epoch": 89, "lr": 9.98129744245564e-05}
{"train_loss": 0.7245393395423889, "global_step": 7962, "epoch": 89, "lr": 9.981292432501587e-05}
{"train_loss": 0.8002977967262268, "global_step": 7963, "epoch": 89, "lr": 9.981287421877861e-05}
{"train_loss": 0.7834375500679016, "global_step": 7964, "epoch": 89, "lr": 9.981282410584463e-05}
{"train_loss": 0.6919422149658203, "global_step": 7965, "epoch": 89, "lr": 9.981277398621389e-05}
{"train_loss": 0.7522857785224915, "global_step": 7966, "epoch": 89, "lr": 9.981272385988646e-05}
{"train_loss": 0.6668252348899841, "global_step": 7967, "epoch": 89, "lr": 9.981267372686229e-05}
{"train_loss": 0.8625714182853699, "global_step": 7968, "epoch": 89, "lr": 9.981262358714142e-05}
{"train_loss": 0.751591145992279, "global_step": 7969, "epoch": 89, "lr": 9.981257344072387e-05}
{"train_loss": 0.8319692611694336, "global_step": 7970, "epoch": 89, "lr": 9.981252328760961e-05}
{"train_loss": 0.824863612651825, "global_step": 7971, "epoch": 89, "lr": 9.981247312779867e-05}
{"train_loss": 0.8074325323104858, "global_step": 7972, "epoch": 89, "lr": 9.981242296129106e-05}
{"train_loss": 1.0155149698257446, "global_step": 7973, "epoch": 89, "lr": 9.981237278808675e-05}
{"train_loss": 0.6882585287094116, "global_step": 7974, "epoch": 89, "lr": 9.98123226081858e-05}
{"train_loss": 0.8589363694190979, "global_step": 7975, "epoch": 89, "lr": 9.981227242158818e-05}
{"train_loss": 0.857930064201355, "global_step": 7976, "epoch": 89, "lr": 9.98122222282939e-05}
{"train_loss": 0.8137770891189575, "global_step": 7977, "epoch": 89, "lr": 9.981217202830298e-05}
{"train_loss": 0.7085938453674316, "global_step": 7978, "epoch": 89, "lr": 9.981212182161542e-05}
{"train_loss": 0.8993733525276184, "global_step": 7979, "epoch": 89, "lr": 9.981207160823123e-05}
{"train_loss": 0.8000938296318054, "global_step": 7980, "epoch": 89, "lr": 9.98120213881504e-05}
{"train_loss": 0.8254703879356384, "global_step": 7981, "epoch": 89, "lr": 9.981197116137297e-05}
{"train_loss": 0.7621408700942993, "global_step": 7982, "epoch": 89, "lr": 9.98119209278989e-05}
{"train_loss": 0.78486168384552, "global_step": 7983, "epoch": 89, "lr": 9.981187068772827e-05}
{"train_loss": 0.8472845554351807, "global_step": 7984, "epoch": 89, "lr": 9.9811820440861e-05}
{"train_loss": 0.8406935334205627, "global_step": 7985, "epoch": 89, "lr": 9.981177018729716e-05}
{"train_loss": 0.7661677002906799, "global_step": 7986, "epoch": 89, "lr": 9.981171992703672e-05}
{"train_loss": 0.8001183867454529, "global_step": 7987, "epoch": 89, "lr": 9.98116696600797e-05}
{"train_loss": 0.7506200075149536, "global_step": 7988, "epoch": 89, "lr": 9.981161938642612e-05}
{"train_loss": 0.7437456250190735, "global_step": 7989, "epoch": 89, "lr": 9.981156910607597e-05}
{"train_loss": 0.8372426629066467, "global_step": 7990, "epoch": 89, "lr": 9.981151881902927e-05}
{"train_loss": 0.819122850894928, "global_step": 7991, "epoch": 89, "lr": 9.981146852528602e-05}
{"train_loss": 0.8911570310592651, "global_step": 7992, "epoch": 89, "lr": 9.98114182248462e-05}
{"train_loss": 0.6707143783569336, "global_step": 7993, "epoch": 89, "lr": 9.981136791770986e-05}
{"train_loss": 0.8241004347801208, "global_step": 7994, "epoch": 89, "lr": 9.981131760387699e-05}
{"train_loss": 0.8993817567825317, "global_step": 7995, "epoch": 89, "lr": 9.981126728334759e-05}
{"train_loss": 0.7221147418022156, "global_step": 7996, "epoch": 89, "lr": 9.981121695612168e-05}
{"train_loss": 0.8809799551963806, "global_step": 7997, "epoch": 89, "lr": 9.981116662219924e-05}
{"train_loss": 0.796735942363739, "global_step": 7998, "epoch": 89, "lr": 9.981111628158034e-05}
{"train_loss": 0.7547369599342346, "global_step": 7999, "epoch": 89, "lr": 9.981106593426491e-05}
{"train_loss": 0.8995321393013, "global_step": 8000, "epoch": 89, "lr": 9.981101558025299e-05}
{"train_loss": 1.0876781940460205, "global_step": 8001, "epoch": 89, "lr": 9.981096521954459e-05}
{"train_loss": 0.7438201308250427, "global_step": 8002, "epoch": 89, "lr": 9.981091485213969e-05}
{"train_loss": 0.7837316393852234, "global_step": 8003, "epoch": 89, "lr": 9.981086447803834e-05}
{"train_loss": 0.7841659188270569, "global_step": 8004, "epoch": 89, "lr": 9.981081409724053e-05}
{"train_loss": 0.9012555480003357, "global_step": 8005, "epoch": 89, "lr": 9.981076370974626e-05}
{"train_loss": 0.6544104218482971, "global_step": 8006, "epoch": 89, "lr": 9.981071331555554e-05}
{"train_loss": 0.7815731763839722, "global_step": 8007, "epoch": 89, "lr": 9.981066291466838e-05}
{"train_loss": 0.7993669509887695, "global_step": 8008, "epoch": 89, "lr": 9.981061250708479e-05}
{"train_loss": 0.8011230250422874, "global_step": 8009, "epoch": 89, "lr": 9.981056209280475e-05, "val_loss": 1.1782563924789429}
{"train_loss": 0.6271584033966064, "global_step": 8010, "epoch": 90, "lr": 9.98105116718283e-05}
{"train_loss": 0.9050347208976746, "global_step": 8011, "epoch": 90, "lr": 9.981046124415543e-05}
{"train_loss": 0.6390990614891052, "global_step": 8012, "epoch": 90, "lr": 9.981041080978615e-05}
{"train_loss": 0.7690272331237793, "global_step": 8013, "epoch": 90, "lr": 9.981036036872047e-05}
{"train_loss": 0.8481297492980957, "global_step": 8014, "epoch": 90, "lr": 9.981030992095838e-05}
{"train_loss": 0.8704258799552917, "global_step": 8015, "epoch": 90, "lr": 9.981025946649993e-05}
{"train_loss": 0.7044236063957214, "global_step": 8016, "epoch": 90, "lr": 9.981020900534508e-05}
{"train_loss": 0.9488925337791443, "global_step": 8017, "epoch": 90, "lr": 9.981015853749385e-05}
{"train_loss": 0.6593292951583862, "global_step": 8018, "epoch": 90, "lr": 9.981010806294627e-05}
{"train_loss": 0.7184491157531738, "global_step": 8019, "epoch": 90, "lr": 9.981005758170231e-05}
{"train_loss": 1.0511720180511475, "global_step": 8020, "epoch": 90, "lr": 9.9810007093762e-05}
{"train_loss": 0.8729124069213867, "global_step": 8021, "epoch": 90, "lr": 9.980995659912533e-05}
{"train_loss": 0.7725273966789246, "global_step": 8022, "epoch": 90, "lr": 9.980990609779234e-05}
{"train_loss": 0.6985906958580017, "global_step": 8023, "epoch": 90, "lr": 9.980985558976301e-05}
{"train_loss": 0.8393936157226562, "global_step": 8024, "epoch": 90, "lr": 9.980980507503735e-05}
{"train_loss": 0.7561938166618347, "global_step": 8025, "epoch": 90, "lr": 9.980975455361537e-05}
{"train_loss": 0.7550637722015381, "global_step": 8026, "epoch": 90, "lr": 9.980970402549706e-05}
{"train_loss": 0.775258481502533, "global_step": 8027, "epoch": 90, "lr": 9.980965349068246e-05}
{"train_loss": 0.7970773577690125, "global_step": 8028, "epoch": 90, "lr": 9.980960294917154e-05}
{"train_loss": 0.980959415435791, "global_step": 8029, "epoch": 90, "lr": 9.980955240096435e-05}
{"train_loss": 0.6011056303977966, "global_step": 8030, "epoch": 90, "lr": 9.980950184606086e-05}
{"train_loss": 0.8236013054847717, "global_step": 8031, "epoch": 90, "lr": 9.98094512844611e-05}
{"train_loss": 0.7363394498825073, "global_step": 8032, "epoch": 90, "lr": 9.980940071616505e-05}
{"train_loss": 0.9073167443275452, "global_step": 8033, "epoch": 90, "lr": 9.980935014117274e-05}
{"train_loss": 0.5804241299629211, "global_step": 8034, "epoch": 90, "lr": 9.980929955948417e-05}
{"train_loss": 0.8344191908836365, "global_step": 8035, "epoch": 90, "lr": 9.980924897109936e-05}
{"train_loss": 0.6792535185813904, "global_step": 8036, "epoch": 90, "lr": 9.980919837601828e-05}
{"train_loss": 0.844684362411499, "global_step": 8037, "epoch": 90, "lr": 9.980914777424097e-05}
{"train_loss": 0.8103975057601929, "global_step": 8038, "epoch": 90, "lr": 9.980909716576744e-05}
{"train_loss": 0.9190483689308167, "global_step": 8039, "epoch": 90, "lr": 9.980904655059768e-05}
{"train_loss": 0.8103534579277039, "global_step": 8040, "epoch": 90, "lr": 9.980899592873169e-05}
{"train_loss": 0.8509876728057861, "global_step": 8041, "epoch": 90, "lr": 9.980894530016948e-05}
{"train_loss": 0.6784314513206482, "global_step": 8042, "epoch": 90, "lr": 9.980889466491107e-05}
{"train_loss": 0.7048563361167908, "global_step": 8043, "epoch": 90, "lr": 9.980884402295648e-05}
{"train_loss": 0.8418447375297546, "global_step": 8044, "epoch": 90, "lr": 9.980879337430568e-05}
{"train_loss": 0.7045154571533203, "global_step": 8045, "epoch": 90, "lr": 9.98087427189587e-05}
{"train_loss": 0.6242653727531433, "global_step": 8046, "epoch": 90, "lr": 9.980869205691555e-05}
{"train_loss": 1.039873719215393, "global_step": 8047, "epoch": 90, "lr": 9.980864138817622e-05}
{"train_loss": 0.6708481311798096, "global_step": 8048, "epoch": 90, "lr": 9.980859071274073e-05}
{"train_loss": 0.6660230159759521, "global_step": 8049, "epoch": 90, "lr": 9.980854003060906e-05}
{"train_loss": 0.8364130854606628, "global_step": 8050, "epoch": 90, "lr": 9.980848934178127e-05}
{"train_loss": 1.0185822248458862, "global_step": 8051, "epoch": 90, "lr": 9.980843864625731e-05}
{"train_loss": 0.8987457752227783, "global_step": 8052, "epoch": 90, "lr": 9.980838794403724e-05}
{"train_loss": 0.66987544298172, "global_step": 8053, "epoch": 90, "lr": 9.980833723512102e-05}
{"train_loss": 0.6522923707962036, "global_step": 8054, "epoch": 90, "lr": 9.980828651950867e-05}
{"train_loss": 0.7723525762557983, "global_step": 8055, "epoch": 90, "lr": 9.980823579720022e-05}
{"train_loss": 0.7264057993888855, "global_step": 8056, "epoch": 90, "lr": 9.980818506819565e-05}
{"train_loss": 1.0302101373672485, "global_step": 8057, "epoch": 90, "lr": 9.980813433249499e-05}
{"train_loss": 0.8929126262664795, "global_step": 8058, "epoch": 90, "lr": 9.980808359009821e-05}
{"train_loss": 0.9233853220939636, "global_step": 8059, "epoch": 90, "lr": 9.980803284100536e-05}
{"train_loss": 0.9577949047088623, "global_step": 8060, "epoch": 90, "lr": 9.980798208521643e-05}
{"train_loss": 0.7239052057266235, "global_step": 8061, "epoch": 90, "lr": 9.98079313227314e-05}
{"train_loss": 0.7437210083007812, "global_step": 8062, "epoch": 90, "lr": 9.980788055355031e-05}
{"train_loss": 0.7275197505950928, "global_step": 8063, "epoch": 90, "lr": 9.980782977767317e-05}
{"train_loss": 0.8270546793937683, "global_step": 8064, "epoch": 90, "lr": 9.980777899509997e-05}
{"train_loss": 0.8427704572677612, "global_step": 8065, "epoch": 90, "lr": 9.980772820583071e-05}
{"train_loss": 0.7446476221084595, "global_step": 8066, "epoch": 90, "lr": 9.980767740986541e-05}
{"train_loss": 0.7839623093605042, "global_step": 8067, "epoch": 90, "lr": 9.980762660720409e-05}
{"train_loss": 0.719379186630249, "global_step": 8068, "epoch": 90, "lr": 9.980757579784673e-05}
{"train_loss": 0.9032524824142456, "global_step": 8069, "epoch": 90, "lr": 9.980752498179335e-05}
{"train_loss": 0.8197026252746582, "global_step": 8070, "epoch": 90, "lr": 9.980747415904395e-05}
{"train_loss": 0.8347669243812561, "global_step": 8071, "epoch": 90, "lr": 9.980742332959856e-05}
{"train_loss": 0.9147974848747253, "global_step": 8072, "epoch": 90, "lr": 9.980737249345716e-05}
{"train_loss": 0.9724319577217102, "global_step": 8073, "epoch": 90, "lr": 9.980732165061976e-05}
{"train_loss": 0.8429194688796997, "global_step": 8074, "epoch": 90, "lr": 9.980727080108638e-05}
{"train_loss": 0.8388327956199646, "global_step": 8075, "epoch": 90, "lr": 9.9807219944857e-05}
{"train_loss": 0.9392060041427612, "global_step": 8076, "epoch": 90, "lr": 9.980716908193167e-05}
{"train_loss": 0.778739869594574, "global_step": 8077, "epoch": 90, "lr": 9.980711821231037e-05}
{"train_loss": 0.8628261685371399, "global_step": 8078, "epoch": 90, "lr": 9.980706733599309e-05}
{"train_loss": 0.7057111859321594, "global_step": 8079, "epoch": 90, "lr": 9.980701645297989e-05}
{"train_loss": 0.7336705923080444, "global_step": 8080, "epoch": 90, "lr": 9.980696556327071e-05}
{"train_loss": 0.8266778588294983, "global_step": 8081, "epoch": 90, "lr": 9.980691466686562e-05}
{"train_loss": 0.7331193089485168, "global_step": 8082, "epoch": 90, "lr": 9.980686376376457e-05}
{"train_loss": 0.9120512008666992, "global_step": 8083, "epoch": 90, "lr": 9.98068128539676e-05}
{"train_loss": 0.9889382719993591, "global_step": 8084, "epoch": 90, "lr": 9.980676193747473e-05}
{"train_loss": 0.7705177664756775, "global_step": 8085, "epoch": 90, "lr": 9.980671101428594e-05}
{"train_loss": 0.8924275636672974, "global_step": 8086, "epoch": 90, "lr": 9.980666008440124e-05}
{"train_loss": 0.9470136761665344, "global_step": 8087, "epoch": 90, "lr": 9.980660914782064e-05}
{"train_loss": 0.7306666374206543, "global_step": 8088, "epoch": 90, "lr": 9.980655820454415e-05}
{"train_loss": 0.8278405070304871, "global_step": 8089, "epoch": 90, "lr": 9.980650725457177e-05}
{"train_loss": 0.8670640587806702, "global_step": 8090, "epoch": 90, "lr": 9.980645629790353e-05}
{"train_loss": 0.8526373505592346, "global_step": 8091, "epoch": 90, "lr": 9.980640533453941e-05}
{"train_loss": 1.0262356996536255, "global_step": 8092, "epoch": 90, "lr": 9.980635436447942e-05}
{"train_loss": 0.8060641288757324, "global_step": 8093, "epoch": 90, "lr": 9.980630338772358e-05}
{"train_loss": 0.7561876177787781, "global_step": 8094, "epoch": 90, "lr": 9.980625240427188e-05}
{"train_loss": 0.878017246723175, "global_step": 8095, "epoch": 90, "lr": 9.980620141412435e-05}
{"train_loss": 0.8814993500709534, "global_step": 8096, "epoch": 90, "lr": 9.980615041728098e-05}
{"train_loss": 0.7916630506515503, "global_step": 8097, "epoch": 90, "lr": 9.980609941374178e-05}
{"train_loss": 0.8100811879286606, "global_step": 8098, "epoch": 90, "lr": 9.980604840350673e-05, "val_loss": 1.1599836349487305, "train_action_mse_error": 43.54259490966797}
{"train_loss": 0.8496010303497314, "global_step": 8099, "epoch": 91, "lr": 9.98059973865759e-05}
{"train_loss": 0.7907184958457947, "global_step": 8100, "epoch": 91, "lr": 9.980594636294924e-05}
{"train_loss": 0.8429309725761414, "global_step": 8101, "epoch": 91, "lr": 9.980589533262679e-05}
{"train_loss": 0.8893170356750488, "global_step": 8102, "epoch": 91, "lr": 9.980584429560853e-05}
{"train_loss": 0.8180502653121948, "global_step": 8103, "epoch": 91, "lr": 9.980579325189449e-05}
{"train_loss": 0.8281522989273071, "global_step": 8104, "epoch": 91, "lr": 9.980574220148468e-05}
{"train_loss": 0.9255717396736145, "global_step": 8105, "epoch": 91, "lr": 9.980569114437907e-05}
{"train_loss": 0.6887381076812744, "global_step": 8106, "epoch": 91, "lr": 9.980564008057772e-05}
{"train_loss": 0.780579149723053, "global_step": 8107, "epoch": 91, "lr": 9.980558901008059e-05}
{"train_loss": 0.8160529136657715, "global_step": 8108, "epoch": 91, "lr": 9.980553793288771e-05}
{"train_loss": 0.7745670080184937, "global_step": 8109, "epoch": 91, "lr": 9.980548684899909e-05}
{"train_loss": 0.9034273624420166, "global_step": 8110, "epoch": 91, "lr": 9.980543575841473e-05}
{"train_loss": 0.9038015604019165, "global_step": 8111, "epoch": 91, "lr": 9.980538466113462e-05}
{"train_loss": 0.9303697943687439, "global_step": 8112, "epoch": 91, "lr": 9.980533355715879e-05}
{"train_loss": 0.8489187359809875, "global_step": 8113, "epoch": 91, "lr": 9.980528244648723e-05}
{"train_loss": 0.6171237826347351, "global_step": 8114, "epoch": 91, "lr": 9.980523132911997e-05}
{"train_loss": 0.8632510900497437, "global_step": 8115, "epoch": 91, "lr": 9.9805180205057e-05}
{"train_loss": 0.9981014728546143, "global_step": 8116, "epoch": 91, "lr": 9.980512907429833e-05}
{"train_loss": 0.8138057589530945, "global_step": 8117, "epoch": 91, "lr": 9.980507793684397e-05}
{"train_loss": 0.8253262639045715, "global_step": 8118, "epoch": 91, "lr": 9.980502679269393e-05}
{"train_loss": 0.9216344356536865, "global_step": 8119, "epoch": 91, "lr": 9.98049756418482e-05}
{"train_loss": 0.7841808795928955, "global_step": 8120, "epoch": 91, "lr": 9.980492448430681e-05}
{"train_loss": 0.9449209570884705, "global_step": 8121, "epoch": 91, "lr": 9.980487332006975e-05}
{"train_loss": 0.8973387479782104, "global_step": 8122, "epoch": 91, "lr": 9.980482214913703e-05}
{"train_loss": 0.8611984252929688, "global_step": 8123, "epoch": 91, "lr": 9.980477097150865e-05}
{"train_loss": 0.8459149599075317, "global_step": 8124, "epoch": 91, "lr": 9.980471978718464e-05}
{"train_loss": 0.9853321313858032, "global_step": 8125, "epoch": 91, "lr": 9.980466859616499e-05}
{"train_loss": 0.6725267767906189, "global_step": 8126, "epoch": 91, "lr": 9.98046173984497e-05}
{"train_loss": 0.8111954927444458, "global_step": 8127, "epoch": 91, "lr": 9.980456619403879e-05}
{"train_loss": 0.9040709733963013, "global_step": 8128, "epoch": 91, "lr": 9.980451498293226e-05}
{"train_loss": 0.7583333849906921, "global_step": 8129, "epoch": 91, "lr": 9.980446376513013e-05}
{"train_loss": 0.8535141944885254, "global_step": 8130, "epoch": 91, "lr": 9.98044125406324e-05}
{"train_loss": 0.8065453171730042, "global_step": 8131, "epoch": 91, "lr": 9.980436130943907e-05}
{"train_loss": 0.9281505346298218, "global_step": 8132, "epoch": 91, "lr": 9.980431007155014e-05}
{"train_loss": 0.6860039234161377, "global_step": 8133, "epoch": 91, "lr": 9.980425882696562e-05}
{"train_loss": 0.762597382068634, "global_step": 8134, "epoch": 91, "lr": 9.980420757568554e-05}
{"train_loss": 0.5912930965423584, "global_step": 8135, "epoch": 91, "lr": 9.98041563177099e-05}
{"train_loss": 0.8711615204811096, "global_step": 8136, "epoch": 91, "lr": 9.980410505303869e-05}
{"train_loss": 0.8509847521781921, "global_step": 8137, "epoch": 91, "lr": 9.980405378167193e-05}
{"train_loss": 0.7455510497093201, "global_step": 8138, "epoch": 91, "lr": 9.98040025036096e-05}
{"train_loss": 0.649031400680542, "global_step": 8139, "epoch": 91, "lr": 9.980395121885175e-05}
{"train_loss": 0.6800976395606995, "global_step": 8140, "epoch": 91, "lr": 9.980389992739835e-05}
{"train_loss": 0.8628857135772705, "global_step": 8141, "epoch": 91, "lr": 9.980384862924944e-05}
{"train_loss": 0.799602746963501, "global_step": 8142, "epoch": 91, "lr": 9.9803797324405e-05}
{"train_loss": 1.0087288618087769, "global_step": 8143, "epoch": 91, "lr": 9.980374601286506e-05}
{"train_loss": 0.8047497272491455, "global_step": 8144, "epoch": 91, "lr": 9.98036946946296e-05}
{"train_loss": 0.9226661920547485, "global_step": 8145, "epoch": 91, "lr": 9.980364336969863e-05}
{"train_loss": 0.9776785373687744, "global_step": 8146, "epoch": 91, "lr": 9.980359203807219e-05}
{"train_loss": 0.6961633563041687, "global_step": 8147, "epoch": 91, "lr": 9.980354069975024e-05}
{"train_loss": 0.8000637292861938, "global_step": 8148, "epoch": 91, "lr": 9.980348935473282e-05}
{"train_loss": 0.8660122156143188, "global_step": 8149, "epoch": 91, "lr": 9.980343800301994e-05}
{"train_loss": 0.7955598831176758, "global_step": 8150, "epoch": 91, "lr": 9.98033866446116e-05}
{"train_loss": 0.8191738724708557, "global_step": 8151, "epoch": 91, "lr": 9.980333527950778e-05}
{"train_loss": 0.7592095136642456, "global_step": 8152, "epoch": 91, "lr": 9.980328390770852e-05}
{"train_loss": 0.883427083492279, "global_step": 8153, "epoch": 91, "lr": 9.980323252921381e-05}
{"train_loss": 0.791460394859314, "global_step": 8154, "epoch": 91, "lr": 9.980318114402367e-05}
{"train_loss": 0.9199643135070801, "global_step": 8155, "epoch": 91, "lr": 9.980312975213809e-05}
{"train_loss": 0.7469359636306763, "global_step": 8156, "epoch": 91, "lr": 9.980307835355709e-05}
{"train_loss": 0.9060002565383911, "global_step": 8157, "epoch": 91, "lr": 9.980302694828068e-05}
{"train_loss": 0.648877739906311, "global_step": 8158, "epoch": 91, "lr": 9.980297553630886e-05}
{"train_loss": 0.8722283244132996, "global_step": 8159, "epoch": 91, "lr": 9.980292411764163e-05}
{"train_loss": 0.7069768309593201, "global_step": 8160, "epoch": 91, "lr": 9.980287269227899e-05}
{"train_loss": 0.7127745151519775, "global_step": 8161, "epoch": 91, "lr": 9.980282126022098e-05}
{"train_loss": 0.9547091126441956, "global_step": 8162, "epoch": 91, "lr": 9.980276982146759e-05}
{"train_loss": 0.7971930503845215, "global_step": 8163, "epoch": 91, "lr": 9.98027183760188e-05}
{"train_loss": 0.7935135364532471, "global_step": 8164, "epoch": 91, "lr": 9.980266692387468e-05}
{"train_loss": 0.6531926989555359, "global_step": 8165, "epoch": 91, "lr": 9.980261546503518e-05}
{"train_loss": 0.7511513829231262, "global_step": 8166, "epoch": 91, "lr": 9.98025639995003e-05}
{"train_loss": 0.764744758605957, "global_step": 8167, "epoch": 91, "lr": 9.980251252727011e-05}
{"train_loss": 0.7343282103538513, "global_step": 8168, "epoch": 91, "lr": 9.980246104834457e-05}
{"train_loss": 0.8185703754425049, "global_step": 8169, "epoch": 91, "lr": 9.980240956272369e-05}
{"train_loss": 0.9400420784950256, "global_step": 8170, "epoch": 91, "lr": 9.980235807040748e-05}
{"train_loss": 0.824840784072876, "global_step": 8171, "epoch": 91, "lr": 9.980230657139596e-05}
{"train_loss": 0.8134093880653381, "global_step": 8172, "epoch": 91, "lr": 9.980225506568911e-05}
{"train_loss": 0.7435077428817749, "global_step": 8173, "epoch": 91, "lr": 9.980220355328697e-05}
{"train_loss": 0.7443244457244873, "global_step": 8174, "epoch": 91, "lr": 9.980215203418952e-05}
{"train_loss": 0.827186107635498, "global_step": 8175, "epoch": 91, "lr": 9.980210050839678e-05}
{"train_loss": 0.8644202947616577, "global_step": 8176, "epoch": 91, "lr": 9.980204897590875e-05}
{"train_loss": 0.8995903730392456, "global_step": 8177, "epoch": 91, "lr": 9.980199743672546e-05}
{"train_loss": 0.9166197180747986, "global_step": 8178, "epoch": 91, "lr": 9.980194589084688e-05}
{"train_loss": 0.8335589170455933, "global_step": 8179, "epoch": 91, "lr": 9.980189433827305e-05}
{"train_loss": 0.8086193203926086, "global_step": 8180, "epoch": 91, "lr": 9.980184277900395e-05}
{"train_loss": 0.8034170269966125, "global_step": 8181, "epoch": 91, "lr": 9.980179121303961e-05}
{"train_loss": 0.7595080137252808, "global_step": 8182, "epoch": 91, "lr": 9.980173964038002e-05}
{"train_loss": 0.7368016839027405, "global_step": 8183, "epoch": 91, "lr": 9.980168806102519e-05}
{"train_loss": 0.8483098745346069, "global_step": 8184, "epoch": 91, "lr": 9.980163647497513e-05}
{"train_loss": 0.7485690712928772, "global_step": 8185, "epoch": 91, "lr": 9.980158488222986e-05}
{"train_loss": 0.8397650718688965, "global_step": 8186, "epoch": 91, "lr": 9.980153328278936e-05}
{"train_loss": 0.8171926629677247, "global_step": 8187, "epoch": 91, "lr": 9.980148167665366e-05, "val_loss": 1.1870965957641602}
{"train_loss": 0.8725036382675171, "global_step": 8188, "epoch": 92, "lr": 9.980143006382275e-05}
{"train_loss": 0.6808392405509949, "global_step": 8189, "epoch": 92, "lr": 9.980137844429664e-05}
{"train_loss": 0.7257812023162842, "global_step": 8190, "epoch": 92, "lr": 9.980132681807536e-05}
{"train_loss": 0.6985610127449036, "global_step": 8191, "epoch": 92, "lr": 9.980127518515888e-05}
{"train_loss": 0.7131773829460144, "global_step": 8192, "epoch": 92, "lr": 9.980122354554724e-05}
{"train_loss": 0.6925360560417175, "global_step": 8193, "epoch": 92, "lr": 9.980117189924044e-05}
{"train_loss": 0.6569046378135681, "global_step": 8194, "epoch": 92, "lr": 9.980112024623844e-05}
{"train_loss": 0.5990517139434814, "global_step": 8195, "epoch": 92, "lr": 9.980106858654133e-05}
{"train_loss": 0.6910496354103088, "global_step": 8196, "epoch": 92, "lr": 9.980101692014905e-05}
{"train_loss": 0.9001405835151672, "global_step": 8197, "epoch": 92, "lr": 9.980096524706164e-05}
{"train_loss": 0.6812007427215576, "global_step": 8198, "epoch": 92, "lr": 9.980091356727909e-05}
{"train_loss": 0.7693759799003601, "global_step": 8199, "epoch": 92, "lr": 9.980086188080142e-05}
{"train_loss": 0.6927358508110046, "global_step": 8200, "epoch": 92, "lr": 9.980081018762862e-05}
{"train_loss": 0.6731145977973938, "global_step": 8201, "epoch": 92, "lr": 9.980075848776072e-05}
{"train_loss": 0.7993079423904419, "global_step": 8202, "epoch": 92, "lr": 9.980070678119771e-05}
{"train_loss": 0.7573940753936768, "global_step": 8203, "epoch": 92, "lr": 9.98006550679396e-05}
{"train_loss": 0.8211553692817688, "global_step": 8204, "epoch": 92, "lr": 9.98006033479864e-05}
{"train_loss": 0.881565272808075, "global_step": 8205, "epoch": 92, "lr": 9.980055162133813e-05}
{"train_loss": 0.8237568140029907, "global_step": 8206, "epoch": 92, "lr": 9.980049988799475e-05}
{"train_loss": 0.8812499046325684, "global_step": 8207, "epoch": 92, "lr": 9.980044814795633e-05}
{"train_loss": 0.7881844639778137, "global_step": 8208, "epoch": 92, "lr": 9.980039640122283e-05}
{"train_loss": 0.7551583051681519, "global_step": 8209, "epoch": 92, "lr": 9.980034464779429e-05}
{"train_loss": 0.67939293384552, "global_step": 8210, "epoch": 92, "lr": 9.980029288767069e-05}
{"train_loss": 0.6431591510772705, "global_step": 8211, "epoch": 92, "lr": 9.980024112085204e-05}
{"train_loss": 0.6071943044662476, "global_step": 8212, "epoch": 92, "lr": 9.980018934733837e-05}
{"train_loss": 0.6184603571891785, "global_step": 8213, "epoch": 92, "lr": 9.980013756712967e-05}
{"train_loss": 0.7481415271759033, "global_step": 8214, "epoch": 92, "lr": 9.980008578022594e-05}
{"train_loss": 0.8359574675559998, "global_step": 8215, "epoch": 92, "lr": 9.98000339866272e-05}
{"train_loss": 0.6863541007041931, "global_step": 8216, "epoch": 92, "lr": 9.979998218633346e-05}
{"train_loss": 0.9209370613098145, "global_step": 8217, "epoch": 92, "lr": 9.97999303793447e-05}
{"train_loss": 0.8976618647575378, "global_step": 8218, "epoch": 92, "lr": 9.979987856566095e-05}
{"train_loss": 0.8662542104721069, "global_step": 8219, "epoch": 92, "lr": 9.979982674528223e-05}
{"train_loss": 0.8406163454055786, "global_step": 8220, "epoch": 92, "lr": 9.979977491820852e-05}
{"train_loss": 0.8075281977653503, "global_step": 8221, "epoch": 92, "lr": 9.979972308443983e-05}
{"train_loss": 0.8178790807723999, "global_step": 8222, "epoch": 92, "lr": 9.979967124397619e-05}
{"train_loss": 0.6921223998069763, "global_step": 8223, "epoch": 92, "lr": 9.979961939681759e-05}
{"train_loss": 0.816771924495697, "global_step": 8224, "epoch": 92, "lr": 9.979956754296403e-05}
{"train_loss": 0.8681220412254333, "global_step": 8225, "epoch": 92, "lr": 9.979951568241552e-05}
{"train_loss": 0.7772988677024841, "global_step": 8226, "epoch": 92, "lr": 9.979946381517209e-05}
{"train_loss": 0.612672746181488, "global_step": 8227, "epoch": 92, "lr": 9.97994119412337e-05}
{"train_loss": 0.7437572479248047, "global_step": 8228, "epoch": 92, "lr": 9.97993600606004e-05}
{"train_loss": 0.8918020725250244, "global_step": 8229, "epoch": 92, "lr": 9.97993081732722e-05}
{"train_loss": 0.7418915033340454, "global_step": 8230, "epoch": 92, "lr": 9.979925627924908e-05}
{"train_loss": 0.6785991787910461, "global_step": 8231, "epoch": 92, "lr": 9.979920437853105e-05}
{"train_loss": 0.8408634066581726, "global_step": 8232, "epoch": 92, "lr": 9.979915247111813e-05}
{"train_loss": 0.7525417804718018, "global_step": 8233, "epoch": 92, "lr": 9.979910055701032e-05}
{"train_loss": 0.7996309995651245, "global_step": 8234, "epoch": 92, "lr": 9.979904863620762e-05}
{"train_loss": 0.7534505724906921, "global_step": 8235, "epoch": 92, "lr": 9.979899670871004e-05}
{"train_loss": 0.8412761688232422, "global_step": 8236, "epoch": 92, "lr": 9.979894477451761e-05}
{"train_loss": 0.9301196932792664, "global_step": 8237, "epoch": 92, "lr": 9.979889283363032e-05}
{"train_loss": 0.8819345235824585, "global_step": 8238, "epoch": 92, "lr": 9.979884088604817e-05}
{"train_loss": 0.7101106643676758, "global_step": 8239, "epoch": 92, "lr": 9.979878893177117e-05}
{"train_loss": 0.7545490860939026, "global_step": 8240, "epoch": 92, "lr": 9.979873697079934e-05}
{"train_loss": 0.9672129154205322, "global_step": 8241, "epoch": 92, "lr": 9.979868500313265e-05}
{"train_loss": 0.7888627648353577, "global_step": 8242, "epoch": 92, "lr": 9.979863302877115e-05}
{"train_loss": 0.7642174959182739, "global_step": 8243, "epoch": 92, "lr": 9.979858104771483e-05}
{"train_loss": 0.83687424659729, "global_step": 8244, "epoch": 92, "lr": 9.979852905996371e-05}
{"train_loss": 0.6626490950584412, "global_step": 8245, "epoch": 92, "lr": 9.979847706551777e-05}
{"train_loss": 0.8341071605682373, "global_step": 8246, "epoch": 92, "lr": 9.979842506437704e-05}
{"train_loss": 0.6924703121185303, "global_step": 8247, "epoch": 92, "lr": 9.979837305654152e-05}
{"train_loss": 0.8934595584869385, "global_step": 8248, "epoch": 92, "lr": 9.97983210420112e-05}
{"train_loss": 0.7470805048942566, "global_step": 8249, "epoch": 92, "lr": 9.979826902078611e-05}
{"train_loss": 0.7567249536514282, "global_step": 8250, "epoch": 92, "lr": 9.979821699286626e-05}
{"train_loss": 0.5923670530319214, "global_step": 8251, "epoch": 92, "lr": 9.979816495825164e-05}
{"train_loss": 0.9437426328659058, "global_step": 8252, "epoch": 92, "lr": 9.979811291694226e-05}
{"train_loss": 0.7725280523300171, "global_step": 8253, "epoch": 92, "lr": 9.979806086893813e-05}
{"train_loss": 0.8959506154060364, "global_step": 8254, "epoch": 92, "lr": 9.979800881423926e-05}
{"train_loss": 0.860469400882721, "global_step": 8255, "epoch": 92, "lr": 9.979795675284567e-05}
{"train_loss": 0.684508204460144, "global_step": 8256, "epoch": 92, "lr": 9.979790468475733e-05}
{"train_loss": 0.7652120590209961, "global_step": 8257, "epoch": 92, "lr": 9.979785260997427e-05}
{"train_loss": 0.7451527714729309, "global_step": 8258, "epoch": 92, "lr": 9.97978005284965e-05}
{"train_loss": 0.9562948346138, "global_step": 8259, "epoch": 92, "lr": 9.979774844032403e-05}
{"train_loss": 0.7658461928367615, "global_step": 8260, "epoch": 92, "lr": 9.979769634545686e-05}
{"train_loss": 1.0588043928146362, "global_step": 8261, "epoch": 92, "lr": 9.9797644243895e-05}
{"train_loss": 1.096638560295105, "global_step": 8262, "epoch": 92, "lr": 9.979759213563843e-05}
{"train_loss": 0.862071692943573, "global_step": 8263, "epoch": 92, "lr": 9.979754002068719e-05}
{"train_loss": 0.8522348403930664, "global_step": 8264, "epoch": 92, "lr": 9.979748789904128e-05}
{"train_loss": 0.8125739097595215, "global_step": 8265, "epoch": 92, "lr": 9.979743577070073e-05}
{"train_loss": 0.7138146162033081, "global_step": 8266, "epoch": 92, "lr": 9.979738363566549e-05}
{"train_loss": 0.8643168210983276, "global_step": 8267, "epoch": 92, "lr": 9.979733149393561e-05}
{"train_loss": 0.9341084361076355, "global_step": 8268, "epoch": 92, "lr": 9.979727934551109e-05}
{"train_loss": 0.9278578758239746, "global_step": 8269, "epoch": 92, "lr": 9.979722719039191e-05}
{"train_loss": 0.7951192855834961, "global_step": 8270, "epoch": 92, "lr": 9.979717502857813e-05}
{"train_loss": 0.724105954170227, "global_step": 8271, "epoch": 92, "lr": 9.97971228600697e-05}
{"train_loss": 0.8233077526092529, "global_step": 8272, "epoch": 92, "lr": 9.979707068486667e-05}
{"train_loss": 0.7309945225715637, "global_step": 8273, "epoch": 92, "lr": 9.979701850296903e-05}
{"train_loss": 0.7020045518875122, "global_step": 8274, "epoch": 92, "lr": 9.979696631437679e-05}
{"train_loss": 0.8682874441146851, "global_step": 8275, "epoch": 92, "lr": 9.979691411908994e-05}
{"train_loss": 0.7857689201162102, "global_step": 8276, "epoch": 92, "lr": 9.979686191710852e-05, "val_loss": 1.1327191591262817}
{"train_loss": 0.7954620718955994, "global_step": 8277, "epoch": 93, "lr": 9.979680970843251e-05}
{"train_loss": 0.8420974016189575, "global_step": 8278, "epoch": 93, "lr": 9.979675749306192e-05}
{"train_loss": 0.8556203246116638, "global_step": 8279, "epoch": 93, "lr": 9.979670527099679e-05}
{"train_loss": 0.7001852989196777, "global_step": 8280, "epoch": 93, "lr": 9.979665304223706e-05}
{"train_loss": 0.7668429017066956, "global_step": 8281, "epoch": 93, "lr": 9.97966008067828e-05}
{"train_loss": 0.6865836381912231, "global_step": 8282, "epoch": 93, "lr": 9.979654856463399e-05}
{"train_loss": 0.7629743218421936, "global_step": 8283, "epoch": 93, "lr": 9.979649631579065e-05}
{"train_loss": 0.9086929559707642, "global_step": 8284, "epoch": 93, "lr": 9.979644406025277e-05}
{"train_loss": 0.6406999230384827, "global_step": 8285, "epoch": 93, "lr": 9.979639179802036e-05}
{"train_loss": 0.6415713429450989, "global_step": 8286, "epoch": 93, "lr": 9.979633952909344e-05}
{"train_loss": 0.7300817966461182, "global_step": 8287, "epoch": 93, "lr": 9.9796287253472e-05}
{"train_loss": 0.7361805438995361, "global_step": 8288, "epoch": 93, "lr": 9.979623497115606e-05}
{"train_loss": 0.6387262344360352, "global_step": 8289, "epoch": 93, "lr": 9.979618268214563e-05}
{"train_loss": 0.8330976963043213, "global_step": 8290, "epoch": 93, "lr": 9.979613038644072e-05}
{"train_loss": 0.6225096583366394, "global_step": 8291, "epoch": 93, "lr": 9.97960780840413e-05}
{"train_loss": 0.8303976655006409, "global_step": 8292, "epoch": 93, "lr": 9.979602577494742e-05}
{"train_loss": 0.7477295398712158, "global_step": 8293, "epoch": 93, "lr": 9.979597345915907e-05}
{"train_loss": 0.6421586275100708, "global_step": 8294, "epoch": 93, "lr": 9.979592113667626e-05}
{"train_loss": 0.6185994744300842, "global_step": 8295, "epoch": 93, "lr": 9.979586880749898e-05}
{"train_loss": 0.7502694725990295, "global_step": 8296, "epoch": 93, "lr": 9.979581647162726e-05}
{"train_loss": 0.5672257542610168, "global_step": 8297, "epoch": 93, "lr": 9.979576412906111e-05}
{"train_loss": 0.7415366172790527, "global_step": 8298, "epoch": 93, "lr": 9.979571177980052e-05}
{"train_loss": 0.8138231635093689, "global_step": 8299, "epoch": 93, "lr": 9.97956594238455e-05}
{"train_loss": 0.8292357325553894, "global_step": 8300, "epoch": 93, "lr": 9.979560706119606e-05}
{"train_loss": 0.7367210388183594, "global_step": 8301, "epoch": 93, "lr": 9.979555469185223e-05}
{"train_loss": 0.7385110259056091, "global_step": 8302, "epoch": 93, "lr": 9.979550231581396e-05}
{"train_loss": 0.8744626641273499, "global_step": 8303, "epoch": 93, "lr": 9.979544993308131e-05}
{"train_loss": 0.7449798583984375, "global_step": 8304, "epoch": 93, "lr": 9.979539754365426e-05}
{"train_loss": 0.79375159740448, "global_step": 8305, "epoch": 93, "lr": 9.979534514753283e-05}
{"train_loss": 0.753373384475708, "global_step": 8306, "epoch": 93, "lr": 9.979529274471703e-05}
{"train_loss": 0.7095175385475159, "global_step": 8307, "epoch": 93, "lr": 9.979524033520686e-05}
{"train_loss": 0.8326064348220825, "global_step": 8308, "epoch": 93, "lr": 9.979518791900232e-05}
{"train_loss": 0.6035667657852173, "global_step": 8309, "epoch": 93, "lr": 9.979513549610342e-05}
{"train_loss": 0.7763910293579102, "global_step": 8310, "epoch": 93, "lr": 9.979508306651018e-05}
{"train_loss": 0.7650840282440186, "global_step": 8311, "epoch": 93, "lr": 9.979503063022259e-05}
{"train_loss": 0.866858959197998, "global_step": 8312, "epoch": 93, "lr": 9.979497818724067e-05}
{"train_loss": 0.681603193283081, "global_step": 8313, "epoch": 93, "lr": 9.979492573756443e-05}
{"train_loss": 0.7953366041183472, "global_step": 8314, "epoch": 93, "lr": 9.979487328119386e-05}
{"train_loss": 0.7917101383209229, "global_step": 8315, "epoch": 93, "lr": 9.979482081812896e-05}
{"train_loss": 0.8121211528778076, "global_step": 8316, "epoch": 93, "lr": 9.979476834836978e-05}
{"train_loss": 0.645164966583252, "global_step": 8317, "epoch": 93, "lr": 9.979471587191628e-05}
{"train_loss": 0.7718911170959473, "global_step": 8318, "epoch": 93, "lr": 9.97946633887685e-05}
{"train_loss": 0.6956930160522461, "global_step": 8319, "epoch": 93, "lr": 9.979461089892644e-05}
{"train_loss": 0.8367875218391418, "global_step": 8320, "epoch": 93, "lr": 9.97945584023901e-05}
{"train_loss": 0.6317926049232483, "global_step": 8321, "epoch": 93, "lr": 9.979450589915948e-05}
{"train_loss": 0.745841383934021, "global_step": 8322, "epoch": 93, "lr": 9.979445338923459e-05}
{"train_loss": 0.6064347624778748, "global_step": 8323, "epoch": 93, "lr": 9.979440087261546e-05}
{"train_loss": 0.7306782007217407, "global_step": 8324, "epoch": 93, "lr": 9.979434834930205e-05}
{"train_loss": 0.6827985048294067, "global_step": 8325, "epoch": 93, "lr": 9.979429581929442e-05}
{"train_loss": 0.8020116090774536, "global_step": 8326, "epoch": 93, "lr": 9.979424328259257e-05}
{"train_loss": 0.7855794429779053, "global_step": 8327, "epoch": 93, "lr": 9.979419073919646e-05}
{"train_loss": 0.581213116645813, "global_step": 8328, "epoch": 93, "lr": 9.979413818910614e-05}
{"train_loss": 0.8364757895469666, "global_step": 8329, "epoch": 93, "lr": 9.979408563232161e-05}
{"train_loss": 0.7918075323104858, "global_step": 8330, "epoch": 93, "lr": 9.979403306884287e-05}
{"train_loss": 0.8430024981498718, "global_step": 8331, "epoch": 93, "lr": 9.979398049866992e-05}
{"train_loss": 0.9298632144927979, "global_step": 8332, "epoch": 93, "lr": 9.979392792180278e-05}
{"train_loss": 0.7368942499160767, "global_step": 8333, "epoch": 93, "lr": 9.979387533824146e-05}
{"train_loss": 0.8306047320365906, "global_step": 8334, "epoch": 93, "lr": 9.979382274798595e-05}
{"train_loss": 0.7822332978248596, "global_step": 8335, "epoch": 93, "lr": 9.979377015103629e-05}
{"train_loss": 0.825823962688446, "global_step": 8336, "epoch": 93, "lr": 9.979371754739245e-05}
{"train_loss": 0.6753309369087219, "global_step": 8337, "epoch": 93, "lr": 9.979366493705443e-05}
{"train_loss": 0.8200159072875977, "global_step": 8338, "epoch": 93, "lr": 9.97936123200223e-05}
{"train_loss": 0.8935742974281311, "global_step": 8339, "epoch": 93, "lr": 9.979355969629598e-05}
{"train_loss": 0.5077943801879883, "global_step": 8340, "epoch": 93, "lr": 9.979350706587556e-05}
{"train_loss": 0.7282374501228333, "global_step": 8341, "epoch": 93, "lr": 9.979345442876099e-05}
{"train_loss": 0.723136305809021, "global_step": 8342, "epoch": 93, "lr": 9.97934017849523e-05}
{"train_loss": 0.6485216617584229, "global_step": 8343, "epoch": 93, "lr": 9.97933491344495e-05}
{"train_loss": 0.7648124098777771, "global_step": 8344, "epoch": 93, "lr": 9.979329647725259e-05}
{"train_loss": 0.7021975517272949, "global_step": 8345, "epoch": 93, "lr": 9.979324381336157e-05}
{"train_loss": 0.992219865322113, "global_step": 8346, "epoch": 93, "lr": 9.979319114277646e-05}
{"train_loss": 0.9383692145347595, "global_step": 8347, "epoch": 93, "lr": 9.979313846549727e-05}
{"train_loss": 0.8153548240661621, "global_step": 8348, "epoch": 93, "lr": 9.9793085781524e-05}
{"train_loss": 0.9024320840835571, "global_step": 8349, "epoch": 93, "lr": 9.979303309085663e-05}
{"train_loss": 0.5868372321128845, "global_step": 8350, "epoch": 93, "lr": 9.979298039349522e-05}
{"train_loss": 0.7204017043113708, "global_step": 8351, "epoch": 93, "lr": 9.979292768943975e-05}
{"train_loss": 0.710630476474762, "global_step": 8352, "epoch": 93, "lr": 9.979287497869022e-05}
{"train_loss": 0.9769545793533325, "global_step": 8353, "epoch": 93, "lr": 9.979282226124665e-05}
{"train_loss": 0.835656464099884, "global_step": 8354, "epoch": 93, "lr": 9.979276953710903e-05}
{"train_loss": 0.832073986530304, "global_step": 8355, "epoch": 93, "lr": 9.97927168062774e-05}
{"train_loss": 0.9189532399177551, "global_step": 8356, "epoch": 93, "lr": 9.979266406875173e-05}
{"train_loss": 0.7126824259757996, "global_step": 8357, "epoch": 93, "lr": 9.979261132453205e-05}
{"train_loss": 0.7388675212860107, "global_step": 8358, "epoch": 93, "lr": 9.979255857361836e-05}
{"train_loss": 0.79286789894104, "global_step": 8359, "epoch": 93, "lr": 9.979250581601067e-05}
{"train_loss": 0.6884436011314392, "global_step": 8360, "epoch": 93, "lr": 9.979245305170896e-05}
{"train_loss": 0.8768851161003113, "global_step": 8361, "epoch": 93, "lr": 9.979240028071329e-05}
{"train_loss": 0.7573992609977722, "global_step": 8362, "epoch": 93, "lr": 9.979234750302363e-05}
{"train_loss": 0.7576608657836914, "global_step": 8363, "epoch": 93, "lr": 9.979229471864e-05}
{"train_loss": 0.8758298754692078, "global_step": 8364, "epoch": 93, "lr": 9.97922419275624e-05}
{"train_loss": 0.7612553901886672, "global_step": 8365, "epoch": 93, "lr": 9.979218912979083e-05, "val_loss": 1.1945620775222778}
{"train_loss": 0.8442081809043884, "global_step": 8366, "epoch": 94, "lr": 9.979213632532532e-05}
{"train_loss": 0.7745071053504944, "global_step": 8367, "epoch": 94, "lr": 9.979208351416586e-05}
{"train_loss": 0.6121504902839661, "global_step": 8368, "epoch": 94, "lr": 9.979203069631246e-05}
{"train_loss": 0.8240628242492676, "global_step": 8369, "epoch": 94, "lr": 9.979197787176512e-05}
{"train_loss": 0.7766987085342407, "global_step": 8370, "epoch": 94, "lr": 9.979192504052387e-05}
{"train_loss": 0.7618434429168701, "global_step": 8371, "epoch": 94, "lr": 9.97918722025887e-05}
{"train_loss": 0.7328535318374634, "global_step": 8372, "epoch": 94, "lr": 9.979181935795962e-05}
{"train_loss": 0.796635627746582, "global_step": 8373, "epoch": 94, "lr": 9.979176650663663e-05}
{"train_loss": 0.8417417407035828, "global_step": 8374, "epoch": 94, "lr": 9.979171364861974e-05}
{"train_loss": 0.7323423027992249, "global_step": 8375, "epoch": 94, "lr": 9.979166078390897e-05}
{"train_loss": 0.5828220248222351, "global_step": 8376, "epoch": 94, "lr": 9.979160791250432e-05}
{"train_loss": 0.7164316773414612, "global_step": 8377, "epoch": 94, "lr": 9.979155503440579e-05}
{"train_loss": 0.7795605659484863, "global_step": 8378, "epoch": 94, "lr": 9.979150214961339e-05}
{"train_loss": 0.8400302529335022, "global_step": 8379, "epoch": 94, "lr": 9.979144925812714e-05}
{"train_loss": 0.7562146782875061, "global_step": 8380, "epoch": 94, "lr": 9.979139635994703e-05}
{"train_loss": 0.8346478343009949, "global_step": 8381, "epoch": 94, "lr": 9.979134345507308e-05}
{"train_loss": 0.7719404101371765, "global_step": 8382, "epoch": 94, "lr": 9.979129054350526e-05}
{"train_loss": 0.7294859290122986, "global_step": 8383, "epoch": 94, "lr": 9.979123762524365e-05}
{"train_loss": 0.7652856111526489, "global_step": 8384, "epoch": 94, "lr": 9.979118470028819e-05}
{"train_loss": 0.7290254235267639, "global_step": 8385, "epoch": 94, "lr": 9.979113176863892e-05}
{"train_loss": 0.7138252258300781, "global_step": 8386, "epoch": 94, "lr": 9.979107883029584e-05}
{"train_loss": 0.8640180230140686, "global_step": 8387, "epoch": 94, "lr": 9.979102588525895e-05}
{"train_loss": 0.843478798866272, "global_step": 8388, "epoch": 94, "lr": 9.979097293352828e-05}
{"train_loss": 0.9724225401878357, "global_step": 8389, "epoch": 94, "lr": 9.97909199751038e-05}
{"train_loss": 0.686402440071106, "global_step": 8390, "epoch": 94, "lr": 9.979086700998554e-05}
{"train_loss": 0.8549256324768066, "global_step": 8391, "epoch": 94, "lr": 9.979081403817352e-05}
{"train_loss": 0.7821019887924194, "global_step": 8392, "epoch": 94, "lr": 9.979076105966772e-05}
{"train_loss": 0.8850438594818115, "global_step": 8393, "epoch": 94, "lr": 9.979070807446816e-05}
{"train_loss": 0.7086888551712036, "global_step": 8394, "epoch": 94, "lr": 9.979065508257486e-05}
{"train_loss": 0.8234057426452637, "global_step": 8395, "epoch": 94, "lr": 9.979060208398779e-05}
{"train_loss": 0.8041685223579407, "global_step": 8396, "epoch": 94, "lr": 9.9790549078707e-05}
{"train_loss": 0.7419112920761108, "global_step": 8397, "epoch": 94, "lr": 9.979049606673247e-05}
{"train_loss": 0.7969338893890381, "global_step": 8398, "epoch": 94, "lr": 9.979044304806421e-05}
{"train_loss": 0.8461214900016785, "global_step": 8399, "epoch": 94, "lr": 9.979039002270222e-05}
{"train_loss": 0.6910665035247803, "global_step": 8400, "epoch": 94, "lr": 9.979033699064655e-05}
{"train_loss": 0.8719929456710815, "global_step": 8401, "epoch": 94, "lr": 9.979028395189716e-05}
{"train_loss": 0.7988590002059937, "global_step": 8402, "epoch": 94, "lr": 9.979023090645408e-05}
{"train_loss": 0.8310191631317139, "global_step": 8403, "epoch": 94, "lr": 9.97901778543173e-05}
{"train_loss": 0.7329894304275513, "global_step": 8404, "epoch": 94, "lr": 9.979012479548684e-05}
{"train_loss": 0.7265656590461731, "global_step": 8405, "epoch": 94, "lr": 9.979007172996271e-05}
{"train_loss": 0.8997911810874939, "global_step": 8406, "epoch": 94, "lr": 9.979001865774491e-05}
{"train_loss": 0.7410974502563477, "global_step": 8407, "epoch": 94, "lr": 9.978996557883345e-05}
{"train_loss": 0.9730795621871948, "global_step": 8408, "epoch": 94, "lr": 9.978991249322833e-05}
{"train_loss": 0.8788453340530396, "global_step": 8409, "epoch": 94, "lr": 9.978985940092956e-05}
{"train_loss": 0.9070054292678833, "global_step": 8410, "epoch": 94, "lr": 9.978980630193715e-05}
{"train_loss": 0.6913271546363831, "global_step": 8411, "epoch": 94, "lr": 9.978975319625113e-05}
{"train_loss": 0.7430557012557983, "global_step": 8412, "epoch": 94, "lr": 9.978970008387147e-05}
{"train_loss": 0.595876157283783, "global_step": 8413, "epoch": 94, "lr": 9.978964696479818e-05}
{"train_loss": 0.9550126791000366, "global_step": 8414, "epoch": 94, "lr": 9.978959383903128e-05}
{"train_loss": 0.8663338422775269, "global_step": 8415, "epoch": 94, "lr": 9.978954070657079e-05}
{"train_loss": 1.026853322982788, "global_step": 8416, "epoch": 94, "lr": 9.97894875674167e-05}
{"train_loss": 0.7658981084823608, "global_step": 8417, "epoch": 94, "lr": 9.978943442156902e-05}
{"train_loss": 0.6917034983634949, "global_step": 8418, "epoch": 94, "lr": 9.978938126902775e-05}
{"train_loss": 0.8190150260925293, "global_step": 8419, "epoch": 94, "lr": 9.978932810979292e-05}
{"train_loss": 0.6882118582725525, "global_step": 8420, "epoch": 94, "lr": 9.978927494386451e-05}
{"train_loss": 0.852736234664917, "global_step": 8421, "epoch": 94, "lr": 9.978922177124254e-05}
{"train_loss": 0.7996138334274292, "global_step": 8422, "epoch": 94, "lr": 9.978916859192702e-05}
{"train_loss": 0.7448757290840149, "global_step": 8423, "epoch": 94, "lr": 9.978911540591794e-05}
{"train_loss": 0.9003373384475708, "global_step": 8424, "epoch": 94, "lr": 9.978906221321534e-05}
{"train_loss": 0.8185597062110901, "global_step": 8425, "epoch": 94, "lr": 9.97890090138192e-05}
{"train_loss": 0.7830020189285278, "global_step": 8426, "epoch": 94, "lr": 9.978895580772953e-05}
{"train_loss": 0.8542847633361816, "global_step": 8427, "epoch": 94, "lr": 9.978890259494633e-05}
{"train_loss": 0.8328981995582581, "global_step": 8428, "epoch": 94, "lr": 9.978884937546964e-05}
{"train_loss": 0.7938899397850037, "global_step": 8429, "epoch": 94, "lr": 9.978879614929944e-05}
{"train_loss": 0.7747907042503357, "global_step": 8430, "epoch": 94, "lr": 9.978874291643574e-05}
{"train_loss": 0.7353834509849548, "global_step": 8431, "epoch": 94, "lr": 9.978868967687856e-05}
{"train_loss": 0.8462588787078857, "global_step": 8432, "epoch": 94, "lr": 9.978863643062788e-05}
{"train_loss": 0.744213342666626, "global_step": 8433, "epoch": 94, "lr": 9.978858317768373e-05}
{"train_loss": 0.8979262113571167, "global_step": 8434, "epoch": 94, "lr": 9.978852991804613e-05}
{"train_loss": 0.8075942397117615, "global_step": 8435, "epoch": 94, "lr": 9.978847665171505e-05}
{"train_loss": 0.6887273192405701, "global_step": 8436, "epoch": 94, "lr": 9.978842337869052e-05}
{"train_loss": 0.8262150883674622, "global_step": 8437, "epoch": 94, "lr": 9.978837009897253e-05}
{"train_loss": 0.8428755402565002, "global_step": 8438, "epoch": 94, "lr": 9.978831681256112e-05}
{"train_loss": 0.7903532981872559, "global_step": 8439, "epoch": 94, "lr": 9.978826351945627e-05}
{"train_loss": 0.860233724117279, "global_step": 8440, "epoch": 94, "lr": 9.9788210219658e-05}
{"train_loss": 0.7361947894096375, "global_step": 8441, "epoch": 94, "lr": 9.978815691316633e-05}
{"train_loss": 0.690898597240448, "global_step": 8442, "epoch": 94, "lr": 9.978810359998121e-05}
{"train_loss": 0.8521376252174377, "global_step": 8443, "epoch": 94, "lr": 9.978805028010271e-05}
{"train_loss": 0.7750234603881836, "global_step": 8444, "epoch": 94, "lr": 9.978799695353081e-05}
{"train_loss": 0.7867771983146667, "global_step": 8445, "epoch": 94, "lr": 9.978794362026552e-05}
{"train_loss": 0.6886612772941589, "global_step": 8446, "epoch": 94, "lr": 9.978789028030684e-05}
{"train_loss": 0.7448982000350952, "global_step": 8447, "epoch": 94, "lr": 9.978783693365478e-05}
{"train_loss": 0.8642228245735168, "global_step": 8448, "epoch": 94, "lr": 9.978778358030937e-05}
{"train_loss": 0.677769124507904, "global_step": 8449, "epoch": 94, "lr": 9.978773022027059e-05}
{"train_loss": 1.085556983947754, "global_step": 8450, "epoch": 94, "lr": 9.978767685353847e-05}
{"train_loss": 0.6810141205787659, "global_step": 8451, "epoch": 94, "lr": 9.9787623480113e-05}
{"train_loss": 0.7414829134941101, "global_step": 8452, "epoch": 94, "lr": 9.978757009999418e-05}
{"train_loss": 0.769277811050415, "global_step": 8453, "epoch": 94, "lr": 9.978751671318203e-05}
{"train_loss": 0.7927579009131099, "global_step": 8454, "epoch": 94, "lr": 9.978746331967655e-05, "val_loss": 1.2080522775650024}
{"train_loss": 0.7859975099563599, "global_step": 8455, "epoch": 95, "lr": 9.978740991947777e-05}
{"train_loss": 0.8533181548118591, "global_step": 8456, "epoch": 95, "lr": 9.978735651258566e-05}
{"train_loss": 0.9262166023254395, "global_step": 8457, "epoch": 95, "lr": 9.978730309900026e-05}
{"train_loss": 0.7743924260139465, "global_step": 8458, "epoch": 95, "lr": 9.978724967872156e-05}
{"train_loss": 0.8712804913520813, "global_step": 8459, "epoch": 95, "lr": 9.978719625174956e-05}
{"train_loss": 0.8284898996353149, "global_step": 8460, "epoch": 95, "lr": 9.97871428180843e-05}
{"train_loss": 0.6800650358200073, "global_step": 8461, "epoch": 95, "lr": 9.978708937772576e-05}
{"train_loss": 0.7358981966972351, "global_step": 8462, "epoch": 95, "lr": 9.978703593067395e-05}
{"train_loss": 0.7349987626075745, "global_step": 8463, "epoch": 95, "lr": 9.978698247692888e-05}
{"train_loss": 0.7108363509178162, "global_step": 8464, "epoch": 95, "lr": 9.978692901649055e-05}
{"train_loss": 0.7786718606948853, "global_step": 8465, "epoch": 95, "lr": 9.978687554935898e-05}
{"train_loss": 0.8535104393959045, "global_step": 8466, "epoch": 95, "lr": 9.978682207553416e-05}
{"train_loss": 0.7764208316802979, "global_step": 8467, "epoch": 95, "lr": 9.978676859501612e-05}
{"train_loss": 0.8215177655220032, "global_step": 8468, "epoch": 95, "lr": 9.978671510780486e-05}
{"train_loss": 0.7577047348022461, "global_step": 8469, "epoch": 95, "lr": 9.978666161390038e-05}
{"train_loss": 0.6485748291015625, "global_step": 8470, "epoch": 95, "lr": 9.978660811330268e-05}
{"train_loss": 0.9843014478683472, "global_step": 8471, "epoch": 95, "lr": 9.978655460601179e-05}
{"train_loss": 0.6991350650787354, "global_step": 8472, "epoch": 95, "lr": 9.978650109202769e-05}
{"train_loss": 0.7024404406547546, "global_step": 8473, "epoch": 95, "lr": 9.978644757135043e-05}
{"train_loss": 0.841294527053833, "global_step": 8474, "epoch": 95, "lr": 9.978639404397996e-05}
{"train_loss": 0.9762281179428101, "global_step": 8475, "epoch": 95, "lr": 9.978634050991633e-05}
{"train_loss": 0.9204092621803284, "global_step": 8476, "epoch": 95, "lr": 9.978628696915953e-05}
{"train_loss": 0.6756746768951416, "global_step": 8477, "epoch": 95, "lr": 9.978623342170956e-05}
{"train_loss": 0.6774223446846008, "global_step": 8478, "epoch": 95, "lr": 9.978617986756645e-05}
{"train_loss": 0.7180858254432678, "global_step": 8479, "epoch": 95, "lr": 9.97861263067302e-05}
{"train_loss": 0.7691786885261536, "global_step": 8480, "epoch": 95, "lr": 9.97860727392008e-05}
{"train_loss": 0.7596203684806824, "global_step": 8481, "epoch": 95, "lr": 9.978601916497828e-05}
{"train_loss": 0.6817194819450378, "global_step": 8482, "epoch": 95, "lr": 9.978596558406262e-05}
{"train_loss": 0.5724242329597473, "global_step": 8483, "epoch": 95, "lr": 9.978591199645387e-05}
{"train_loss": 0.8934047818183899, "global_step": 8484, "epoch": 95, "lr": 9.9785858402152e-05}
{"train_loss": 0.6576066017150879, "global_step": 8485, "epoch": 95, "lr": 9.978580480115701e-05}
{"train_loss": 0.7834053635597229, "global_step": 8486, "epoch": 95, "lr": 9.978575119346894e-05}
{"train_loss": 0.7979162335395813, "global_step": 8487, "epoch": 95, "lr": 9.978569757908778e-05}
{"train_loss": 0.8314942121505737, "global_step": 8488, "epoch": 95, "lr": 9.978564395801353e-05}
{"train_loss": 0.721422016620636, "global_step": 8489, "epoch": 95, "lr": 9.978559033024624e-05}
{"train_loss": 0.8371545672416687, "global_step": 8490, "epoch": 95, "lr": 9.978553669578586e-05}
{"train_loss": 0.7234060168266296, "global_step": 8491, "epoch": 95, "lr": 9.978548305463242e-05}
{"train_loss": 0.8924517631530762, "global_step": 8492, "epoch": 95, "lr": 9.978542940678592e-05}
{"train_loss": 0.8344504833221436, "global_step": 8493, "epoch": 95, "lr": 9.97853757522464e-05}
{"train_loss": 0.6830770969390869, "global_step": 8494, "epoch": 95, "lr": 9.978532209101383e-05}
{"train_loss": 0.9292321801185608, "global_step": 8495, "epoch": 95, "lr": 9.978526842308823e-05}
{"train_loss": 0.9059011936187744, "global_step": 8496, "epoch": 95, "lr": 9.97852147484696e-05}
{"train_loss": 0.8207619786262512, "global_step": 8497, "epoch": 95, "lr": 9.978516106715798e-05}
{"train_loss": 0.8041496276855469, "global_step": 8498, "epoch": 95, "lr": 9.978510737915332e-05}
{"train_loss": 0.7105590105056763, "global_step": 8499, "epoch": 95, "lr": 9.978505368445568e-05}
{"train_loss": 0.7935572266578674, "global_step": 8500, "epoch": 95, "lr": 9.978499998306504e-05}
{"train_loss": 0.7940112352371216, "global_step": 8501, "epoch": 95, "lr": 9.978494627498142e-05}
{"train_loss": 0.7251770496368408, "global_step": 8502, "epoch": 95, "lr": 9.978489256020482e-05}
{"train_loss": 0.8865075707435608, "global_step": 8503, "epoch": 95, "lr": 9.978483883873524e-05}
{"train_loss": 0.9002922773361206, "global_step": 8504, "epoch": 95, "lr": 9.97847851105727e-05}
{"train_loss": 0.841442346572876, "global_step": 8505, "epoch": 95, "lr": 9.97847313757172e-05}
{"train_loss": 0.7615883946418762, "global_step": 8506, "epoch": 95, "lr": 9.978467763416874e-05}
{"train_loss": 0.7441169619560242, "global_step": 8507, "epoch": 95, "lr": 9.978462388592735e-05}
{"train_loss": 0.6974867582321167, "global_step": 8508, "epoch": 95, "lr": 9.978457013099303e-05}
{"train_loss": 0.6713824272155762, "global_step": 8509, "epoch": 95, "lr": 9.978451636936577e-05}
{"train_loss": 0.8244633674621582, "global_step": 8510, "epoch": 95, "lr": 9.97844626010456e-05}
{"train_loss": 0.764287531375885, "global_step": 8511, "epoch": 95, "lr": 9.97844088260325e-05}
{"train_loss": 0.8243023157119751, "global_step": 8512, "epoch": 95, "lr": 9.978435504432649e-05}
{"train_loss": 0.9251465201377869, "global_step": 8513, "epoch": 95, "lr": 9.97843012559276e-05}
{"train_loss": 0.9043389558792114, "global_step": 8514, "epoch": 95, "lr": 9.978424746083582e-05}
{"train_loss": 0.763713002204895, "global_step": 8515, "epoch": 95, "lr": 9.978419365905112e-05}
{"train_loss": 0.8197884559631348, "global_step": 8516, "epoch": 95, "lr": 9.978413985057358e-05}
{"train_loss": 0.7127846479415894, "global_step": 8517, "epoch": 95, "lr": 9.978408603540315e-05}
{"train_loss": 0.8648472428321838, "global_step": 8518, "epoch": 95, "lr": 9.978403221353985e-05}
{"train_loss": 0.6602016687393188, "global_step": 8519, "epoch": 95, "lr": 9.978397838498371e-05}
{"train_loss": 0.8923751711845398, "global_step": 8520, "epoch": 95, "lr": 9.978392454973471e-05}
{"train_loss": 0.8130757808685303, "global_step": 8521, "epoch": 95, "lr": 9.978387070779287e-05}
{"train_loss": 0.8781808018684387, "global_step": 8522, "epoch": 95, "lr": 9.978381685915821e-05}
{"train_loss": 0.7758925557136536, "global_step": 8523, "epoch": 95, "lr": 9.97837630038307e-05}
{"train_loss": 0.8445359468460083, "global_step": 8524, "epoch": 95, "lr": 9.97837091418104e-05}
{"train_loss": 0.8947001099586487, "global_step": 8525, "epoch": 95, "lr": 9.978365527309725e-05}
{"train_loss": 0.7009665966033936, "global_step": 8526, "epoch": 95, "lr": 9.978360139769132e-05}
{"train_loss": 0.8029105067253113, "global_step": 8527, "epoch": 95, "lr": 9.978354751559257e-05}
{"train_loss": 0.6966548562049866, "global_step": 8528, "epoch": 95, "lr": 9.978349362680105e-05}
{"train_loss": 0.7787498235702515, "global_step": 8529, "epoch": 95, "lr": 9.978343973131674e-05}
{"train_loss": 0.8573631048202515, "global_step": 8530, "epoch": 95, "lr": 9.978338582913965e-05}
{"train_loss": 0.7335900068283081, "global_step": 8531, "epoch": 95, "lr": 9.97833319202698e-05}
{"train_loss": 0.7812778353691101, "global_step": 8532, "epoch": 95, "lr": 9.978327800470718e-05}
{"train_loss": 0.9591115117073059, "global_step": 8533, "epoch": 95, "lr": 9.97832240824518e-05}
{"train_loss": 0.7491592764854431, "global_step": 8534, "epoch": 95, "lr": 9.978317015350368e-05}
{"train_loss": 0.8926500082015991, "global_step": 8535, "epoch": 95, "lr": 9.978311621786281e-05}
{"train_loss": 0.623982310295105, "global_step": 8536, "epoch": 95, "lr": 9.978306227552923e-05}
{"train_loss": 0.9294180274009705, "global_step": 8537, "epoch": 95, "lr": 9.978300832650289e-05}
{"train_loss": 0.6944081783294678, "global_step": 8538, "epoch": 95, "lr": 9.978295437078386e-05}
{"train_loss": 0.718804121017456, "global_step": 8539, "epoch": 95, "lr": 9.97829004083721e-05}
{"train_loss": 0.8753752708435059, "global_step": 8540, "epoch": 95, "lr": 9.978284643926765e-05}
{"train_loss": 0.7873954772949219, "global_step": 8541, "epoch": 95, "lr": 9.978279246347048e-05}
{"train_loss": 0.8149884939193726, "global_step": 8542, "epoch": 95, "lr": 9.978273848098066e-05}
{"train_loss": 0.7934287130162957, "global_step": 8543, "epoch": 95, "lr": 9.978268449179812e-05, "val_loss": 1.2615505456924438, "train_action_mse_error": 52.891387939453125}
{"train_loss": 0.7221267223358154, "global_step": 8544, "epoch": 96, "lr": 9.978263049592293e-05}
{"train_loss": 0.8378860950469971, "global_step": 8545, "epoch": 96, "lr": 9.978257649335505e-05}
{"train_loss": 0.8977034687995911, "global_step": 8546, "epoch": 96, "lr": 9.978252248409452e-05}
{"train_loss": 0.8557755947113037, "global_step": 8547, "epoch": 96, "lr": 9.978246846814134e-05}
{"train_loss": 0.6964870095252991, "global_step": 8548, "epoch": 96, "lr": 9.978241444549552e-05}
{"train_loss": 0.8031612038612366, "global_step": 8549, "epoch": 96, "lr": 9.978236041615704e-05}
{"train_loss": 0.7348578572273254, "global_step": 8550, "epoch": 96, "lr": 9.978230638012596e-05}
{"train_loss": 0.8124184608459473, "global_step": 8551, "epoch": 96, "lr": 9.978225233740222e-05}
{"train_loss": 0.7004318833351135, "global_step": 8552, "epoch": 96, "lr": 9.978219828798589e-05}
{"train_loss": 0.7118501663208008, "global_step": 8553, "epoch": 96, "lr": 9.978214423187693e-05}
{"train_loss": 0.7952906489372253, "global_step": 8554, "epoch": 96, "lr": 9.978209016907539e-05}
{"train_loss": 0.7710760831832886, "global_step": 8555, "epoch": 96, "lr": 9.978203609958124e-05}
{"train_loss": 0.6945604085922241, "global_step": 8556, "epoch": 96, "lr": 9.97819820233945e-05}
{"train_loss": 0.8406169414520264, "global_step": 8557, "epoch": 96, "lr": 9.978192794051517e-05}
{"train_loss": 0.7596166133880615, "global_step": 8558, "epoch": 96, "lr": 9.978187385094329e-05}
{"train_loss": 0.7064940929412842, "global_step": 8559, "epoch": 96, "lr": 9.978181975467884e-05}
{"train_loss": 0.8955240845680237, "global_step": 8560, "epoch": 96, "lr": 9.978176565172182e-05}
{"train_loss": 0.9690709114074707, "global_step": 8561, "epoch": 96, "lr": 9.978171154207225e-05}
{"train_loss": 0.636063814163208, "global_step": 8562, "epoch": 96, "lr": 9.978165742573013e-05}
{"train_loss": 0.9327862858772278, "global_step": 8563, "epoch": 96, "lr": 9.978160330269549e-05}
{"train_loss": 0.667365312576294, "global_step": 8564, "epoch": 96, "lr": 9.978154917296832e-05}
{"train_loss": 0.6397141218185425, "global_step": 8565, "epoch": 96, "lr": 9.978149503654862e-05}
{"train_loss": 0.8740135431289673, "global_step": 8566, "epoch": 96, "lr": 9.97814408934364e-05}
{"train_loss": 1.0320967435836792, "global_step": 8567, "epoch": 96, "lr": 9.978138674363167e-05}
{"train_loss": 0.7844820022583008, "global_step": 8568, "epoch": 96, "lr": 9.978133258713444e-05}
{"train_loss": 0.8611359596252441, "global_step": 8569, "epoch": 96, "lr": 9.978127842394473e-05}
{"train_loss": 0.8130977153778076, "global_step": 8570, "epoch": 96, "lr": 9.978122425406253e-05}
{"train_loss": 0.7414076328277588, "global_step": 8571, "epoch": 96, "lr": 9.978117007748785e-05}
{"train_loss": 0.8240110874176025, "global_step": 8572, "epoch": 96, "lr": 9.978111589422069e-05}
{"train_loss": 0.7993623614311218, "global_step": 8573, "epoch": 96, "lr": 9.978106170426107e-05}
{"train_loss": 0.8551745414733887, "global_step": 8574, "epoch": 96, "lr": 9.9781007507609e-05}
{"train_loss": 0.7353345155715942, "global_step": 8575, "epoch": 96, "lr": 9.978095330426446e-05}
{"train_loss": 0.7217816710472107, "global_step": 8576, "epoch": 96, "lr": 9.978089909422749e-05}
{"train_loss": 0.7847328186035156, "global_step": 8577, "epoch": 96, "lr": 9.97808448774981e-05}
{"train_loss": 0.7352715730667114, "global_step": 8578, "epoch": 96, "lr": 9.978079065407626e-05}
{"train_loss": 0.6410485506057739, "global_step": 8579, "epoch": 96, "lr": 9.978073642396202e-05}
{"train_loss": 0.5668780207633972, "global_step": 8580, "epoch": 96, "lr": 9.978068218715535e-05}
{"train_loss": 0.776803731918335, "global_step": 8581, "epoch": 96, "lr": 9.978062794365628e-05}
{"train_loss": 0.8260754942893982, "global_step": 8582, "epoch": 96, "lr": 9.97805736934648e-05}
{"train_loss": 0.9392761588096619, "global_step": 8583, "epoch": 96, "lr": 9.978051943658094e-05}
{"train_loss": 0.9077810645103455, "global_step": 8584, "epoch": 96, "lr": 9.97804651730047e-05}
{"train_loss": 0.8117863535881042, "global_step": 8585, "epoch": 96, "lr": 9.978041090273608e-05}
{"train_loss": 0.7825118899345398, "global_step": 8586, "epoch": 96, "lr": 9.978035662577509e-05}
{"train_loss": 0.8351285457611084, "global_step": 8587, "epoch": 96, "lr": 9.978030234212174e-05}
{"train_loss": 0.7454061508178711, "global_step": 8588, "epoch": 96, "lr": 9.978024805177603e-05}
{"train_loss": 0.7855648994445801, "global_step": 8589, "epoch": 96, "lr": 9.978019375473797e-05}
{"train_loss": 0.8029695749282837, "global_step": 8590, "epoch": 96, "lr": 9.978013945100757e-05}
{"train_loss": 0.8695129752159119, "global_step": 8591, "epoch": 96, "lr": 9.978008514058485e-05}
{"train_loss": 0.8423686623573303, "global_step": 8592, "epoch": 96, "lr": 9.978003082346978e-05}
{"train_loss": 0.7883796095848083, "global_step": 8593, "epoch": 96, "lr": 9.977997649966242e-05}
{"train_loss": 0.9103388786315918, "global_step": 8594, "epoch": 96, "lr": 9.977992216916274e-05}
{"train_loss": 0.8555490970611572, "global_step": 8595, "epoch": 96, "lr": 9.977986783197074e-05}
{"train_loss": 0.7105028033256531, "global_step": 8596, "epoch": 96, "lr": 9.977981348808644e-05}
{"train_loss": 0.8119586706161499, "global_step": 8597, "epoch": 96, "lr": 9.977975913750987e-05}
{"train_loss": 0.9555377960205078, "global_step": 8598, "epoch": 96, "lr": 9.977970478024103e-05}
{"train_loss": 0.8433958292007446, "global_step": 8599, "epoch": 96, "lr": 9.97796504162799e-05}
{"train_loss": 0.6759627461433411, "global_step": 8600, "epoch": 96, "lr": 9.97795960456265e-05}
{"train_loss": 0.7990078926086426, "global_step": 8601, "epoch": 96, "lr": 9.977954166828083e-05}
{"train_loss": 0.7882950901985168, "global_step": 8602, "epoch": 96, "lr": 9.977948728424294e-05}
{"train_loss": 0.7768349051475525, "global_step": 8603, "epoch": 96, "lr": 9.977943289351277e-05}
{"train_loss": 0.9575605392456055, "global_step": 8604, "epoch": 96, "lr": 9.977937849609038e-05}
{"train_loss": 0.8979211449623108, "global_step": 8605, "epoch": 96, "lr": 9.977932409197574e-05}
{"train_loss": 0.8933554887771606, "global_step": 8606, "epoch": 96, "lr": 9.977926968116891e-05}
{"train_loss": 0.7312891483306885, "global_step": 8607, "epoch": 96, "lr": 9.977921526366984e-05}
{"train_loss": 0.8363136649131775, "global_step": 8608, "epoch": 96, "lr": 9.977916083947857e-05}
{"train_loss": 0.8408914804458618, "global_step": 8609, "epoch": 96, "lr": 9.977910640859509e-05}
{"train_loss": 0.7499768137931824, "global_step": 8610, "epoch": 96, "lr": 9.977905197101942e-05}
{"train_loss": 0.6422860026359558, "global_step": 8611, "epoch": 96, "lr": 9.977899752675157e-05}
{"train_loss": 0.8940649032592773, "global_step": 8612, "epoch": 96, "lr": 9.977894307579154e-05}
{"train_loss": 0.8038235902786255, "global_step": 8613, "epoch": 96, "lr": 9.977888861813934e-05}
{"train_loss": 0.7784143686294556, "global_step": 8614, "epoch": 96, "lr": 9.977883415379497e-05}
{"train_loss": 0.6776595711708069, "global_step": 8615, "epoch": 96, "lr": 9.977877968275843e-05}
{"train_loss": 0.8276844024658203, "global_step": 8616, "epoch": 96, "lr": 9.977872520502976e-05}
{"train_loss": 0.6515728235244751, "global_step": 8617, "epoch": 96, "lr": 9.977867072060893e-05}
{"train_loss": 0.8205485343933105, "global_step": 8618, "epoch": 96, "lr": 9.977861622949597e-05}
{"train_loss": 0.7982869744300842, "global_step": 8619, "epoch": 96, "lr": 9.977856173169087e-05}
{"train_loss": 0.6688215136528015, "global_step": 8620, "epoch": 96, "lr": 9.977850722719367e-05}
{"train_loss": 0.7164986729621887, "global_step": 8621, "epoch": 96, "lr": 9.977845271600434e-05}
{"train_loss": 0.777081310749054, "global_step": 8622, "epoch": 96, "lr": 9.977839819812292e-05}
{"train_loss": 0.7726375460624695, "global_step": 8623, "epoch": 96, "lr": 9.97783436735494e-05}
{"train_loss": 0.8218100666999817, "global_step": 8624, "epoch": 96, "lr": 9.977828914228377e-05}
{"train_loss": 0.711982250213623, "global_step": 8625, "epoch": 96, "lr": 9.977823460432608e-05}
{"train_loss": 0.6468033790588379, "global_step": 8626, "epoch": 96, "lr": 9.97781800596763e-05}
{"train_loss": 0.9627941250801086, "global_step": 8627, "epoch": 96, "lr": 9.977812550833446e-05}
{"train_loss": 0.9667997360229492, "global_step": 8628, "epoch": 96, "lr": 9.977807095030055e-05}
{"train_loss": 0.7585073709487915, "global_step": 8629, "epoch": 96, "lr": 9.977801638557458e-05}
{"train_loss": 0.9149384498596191, "global_step": 8630, "epoch": 96, "lr": 9.977796181415656e-05}
{"train_loss": 0.8086815476417542, "global_step": 8631, "epoch": 96, "lr": 9.97779072360465e-05}
{"train_loss": 0.7957808020409574, "global_step": 8632, "epoch": 96, "lr": 9.977785265124441e-05, "val_loss": 1.1681034564971924}
{"train_loss": 0.7266683578491211, "global_step": 8633, "epoch": 97, "lr": 9.977779805975031e-05}
{"train_loss": 0.6688275337219238, "global_step": 8634, "epoch": 97, "lr": 9.97777434615642e-05}
{"train_loss": 0.711407482624054, "global_step": 8635, "epoch": 97, "lr": 9.977768885668604e-05}
{"train_loss": 0.8411921858787537, "global_step": 8636, "epoch": 97, "lr": 9.97776342451159e-05}
{"train_loss": 0.7430510520935059, "global_step": 8637, "epoch": 97, "lr": 9.977757962685376e-05}
{"train_loss": 0.6991307139396667, "global_step": 8638, "epoch": 97, "lr": 9.977752500189962e-05}
{"train_loss": 0.6357169151306152, "global_step": 8639, "epoch": 97, "lr": 9.97774703702535e-05}
{"train_loss": 0.6594617366790771, "global_step": 8640, "epoch": 97, "lr": 9.977741573191544e-05}
{"train_loss": 0.6691086292266846, "global_step": 8641, "epoch": 97, "lr": 9.977736108688538e-05}
{"train_loss": 0.7213123440742493, "global_step": 8642, "epoch": 97, "lr": 9.977730643516337e-05}
{"train_loss": 0.79330974817276, "global_step": 8643, "epoch": 97, "lr": 9.977725177674941e-05}
{"train_loss": 0.6788223385810852, "global_step": 8644, "epoch": 97, "lr": 9.97771971116435e-05}
{"train_loss": 0.6219082474708557, "global_step": 8645, "epoch": 97, "lr": 9.977714243984563e-05}
{"train_loss": 0.583418071269989, "global_step": 8646, "epoch": 97, "lr": 9.977708776135587e-05}
{"train_loss": 0.5822478532791138, "global_step": 8647, "epoch": 97, "lr": 9.977703307617417e-05}
{"train_loss": 0.6179762482643127, "global_step": 8648, "epoch": 97, "lr": 9.977697838430055e-05}
{"train_loss": 0.6894114017486572, "global_step": 8649, "epoch": 97, "lr": 9.977692368573504e-05}
{"train_loss": 0.6571255326271057, "global_step": 8650, "epoch": 97, "lr": 9.977686898047761e-05}
{"train_loss": 0.7830655574798584, "global_step": 8651, "epoch": 97, "lr": 9.97768142685283e-05}
{"train_loss": 0.7275069952011108, "global_step": 8652, "epoch": 97, "lr": 9.97767595498871e-05}
{"train_loss": 0.7818127274513245, "global_step": 8653, "epoch": 97, "lr": 9.977670482455401e-05}
{"train_loss": 0.6942868232727051, "global_step": 8654, "epoch": 97, "lr": 9.977665009252906e-05}
{"train_loss": 0.7969990968704224, "global_step": 8655, "epoch": 97, "lr": 9.977659535381224e-05}
{"train_loss": 0.7429584264755249, "global_step": 8656, "epoch": 97, "lr": 9.977654060840357e-05}
{"train_loss": 1.0646981000900269, "global_step": 8657, "epoch": 97, "lr": 9.977648585630305e-05}
{"train_loss": 0.6765240430831909, "global_step": 8658, "epoch": 97, "lr": 9.977643109751069e-05}
{"train_loss": 0.7445291876792908, "global_step": 8659, "epoch": 97, "lr": 9.977637633202648e-05}
{"train_loss": 0.8607591390609741, "global_step": 8660, "epoch": 97, "lr": 9.977632155985046e-05}
{"train_loss": 0.8561003804206848, "global_step": 8661, "epoch": 97, "lr": 9.977626678098261e-05}
{"train_loss": 0.7446703910827637, "global_step": 8662, "epoch": 97, "lr": 9.977621199542295e-05}
{"train_loss": 0.7720598578453064, "global_step": 8663, "epoch": 97, "lr": 9.977615720317149e-05}
{"train_loss": 0.7108860015869141, "global_step": 8664, "epoch": 97, "lr": 9.977610240422823e-05}
{"train_loss": 0.7767255902290344, "global_step": 8665, "epoch": 97, "lr": 9.977604759859319e-05}
{"train_loss": 0.7570674419403076, "global_step": 8666, "epoch": 97, "lr": 9.977599278626635e-05}
{"train_loss": 0.7804084420204163, "global_step": 8667, "epoch": 97, "lr": 9.977593796724773e-05}
{"train_loss": 0.9181824922561646, "global_step": 8668, "epoch": 97, "lr": 9.977588314153736e-05}
{"train_loss": 0.7154543399810791, "global_step": 8669, "epoch": 97, "lr": 9.977582830913523e-05}
{"train_loss": 0.6225854158401489, "global_step": 8670, "epoch": 97, "lr": 9.977577347004134e-05}
{"train_loss": 0.657006561756134, "global_step": 8671, "epoch": 97, "lr": 9.97757186242557e-05}
{"train_loss": 0.8154115676879883, "global_step": 8672, "epoch": 97, "lr": 9.977566377177832e-05}
{"train_loss": 0.6304677128791809, "global_step": 8673, "epoch": 97, "lr": 9.977560891260922e-05}
{"train_loss": 0.8513617515563965, "global_step": 8674, "epoch": 97, "lr": 9.97755540467484e-05}
{"train_loss": 0.8489418029785156, "global_step": 8675, "epoch": 97, "lr": 9.977549917419585e-05}
{"train_loss": 0.9170209765434265, "global_step": 8676, "epoch": 97, "lr": 9.977544429495158e-05}
{"train_loss": 0.7728583812713623, "global_step": 8677, "epoch": 97, "lr": 9.977538940901561e-05}
{"train_loss": 1.0447782278060913, "global_step": 8678, "epoch": 97, "lr": 9.977533451638796e-05}
{"train_loss": 0.8794534802436829, "global_step": 8679, "epoch": 97, "lr": 9.977527961706863e-05}
{"train_loss": 0.9841854572296143, "global_step": 8680, "epoch": 97, "lr": 9.977522471105761e-05}
{"train_loss": 0.7437533736228943, "global_step": 8681, "epoch": 97, "lr": 9.97751697983549e-05}
{"train_loss": 0.8084304928779602, "global_step": 8682, "epoch": 97, "lr": 9.977511487896056e-05}
{"train_loss": 0.908836305141449, "global_step": 8683, "epoch": 97, "lr": 9.977505995287453e-05}
{"train_loss": 0.8665463924407959, "global_step": 8684, "epoch": 97, "lr": 9.977500502009687e-05}
{"train_loss": 0.8152462840080261, "global_step": 8685, "epoch": 97, "lr": 9.977495008062756e-05}
{"train_loss": 0.9672102928161621, "global_step": 8686, "epoch": 97, "lr": 9.977489513446663e-05}
{"train_loss": 0.8544512391090393, "global_step": 8687, "epoch": 97, "lr": 9.977484018161404e-05}
{"train_loss": 0.87056565284729, "global_step": 8688, "epoch": 97, "lr": 9.977478522206987e-05}
{"train_loss": 0.7407504916191101, "global_step": 8689, "epoch": 97, "lr": 9.977473025583406e-05}
{"train_loss": 0.7697665691375732, "global_step": 8690, "epoch": 97, "lr": 9.977467528290664e-05}
{"train_loss": 0.7738407254219055, "global_step": 8691, "epoch": 97, "lr": 9.977462030328764e-05}
{"train_loss": 0.8648524880409241, "global_step": 8692, "epoch": 97, "lr": 9.977456531697704e-05}
{"train_loss": 0.8202885389328003, "global_step": 8693, "epoch": 97, "lr": 9.977451032397485e-05}
{"train_loss": 0.8650767803192139, "global_step": 8694, "epoch": 97, "lr": 9.97744553242811e-05}
{"train_loss": 0.8639866709709167, "global_step": 8695, "epoch": 97, "lr": 9.977440031789577e-05}
{"train_loss": 0.7637166380882263, "global_step": 8696, "epoch": 97, "lr": 9.977434530481887e-05}
{"train_loss": 0.9153739809989929, "global_step": 8697, "epoch": 97, "lr": 9.977429028505043e-05}
{"train_loss": 0.8859567642211914, "global_step": 8698, "epoch": 97, "lr": 9.977423525859044e-05}
{"train_loss": 0.8344243764877319, "global_step": 8699, "epoch": 97, "lr": 9.977418022543891e-05}
{"train_loss": 0.7515411376953125, "global_step": 8700, "epoch": 97, "lr": 9.977412518559584e-05}
{"train_loss": 0.7074496746063232, "global_step": 8701, "epoch": 97, "lr": 9.977407013906126e-05}
{"train_loss": 0.8325802683830261, "global_step": 8702, "epoch": 97, "lr": 9.977401508583515e-05}
{"train_loss": 0.9348907470703125, "global_step": 8703, "epoch": 97, "lr": 9.977396002591755e-05}
{"train_loss": 0.8408079147338867, "global_step": 8704, "epoch": 97, "lr": 9.977390495930843e-05}
{"train_loss": 0.9280840754508972, "global_step": 8705, "epoch": 97, "lr": 9.977384988600782e-05}
{"train_loss": 0.81468266248703, "global_step": 8706, "epoch": 97, "lr": 9.977379480601572e-05}
{"train_loss": 0.6811913847923279, "global_step": 8707, "epoch": 97, "lr": 9.977373971933214e-05}
{"train_loss": 0.710517942905426, "global_step": 8708, "epoch": 97, "lr": 9.977368462595709e-05}
{"train_loss": 0.6815350651741028, "global_step": 8709, "epoch": 97, "lr": 9.977362952589057e-05}
{"train_loss": 0.7493533492088318, "global_step": 8710, "epoch": 97, "lr": 9.97735744191326e-05}
{"train_loss": 0.8464347124099731, "global_step": 8711, "epoch": 97, "lr": 9.977351930568317e-05}
{"train_loss": 0.7740727066993713, "global_step": 8712, "epoch": 97, "lr": 9.977346418554231e-05}
{"train_loss": 0.6886322498321533, "global_step": 8713, "epoch": 97, "lr": 9.977340905871e-05}
{"train_loss": 0.8569791913032532, "global_step": 8714, "epoch": 97, "lr": 9.977335392518627e-05}
{"train_loss": 0.8311997056007385, "global_step": 8715, "epoch": 97, "lr": 9.977329878497112e-05}
{"train_loss": 0.799839973449707, "global_step": 8716, "epoch": 97, "lr": 9.977324363806455e-05}
{"train_loss": 0.8159125447273254, "global_step": 8717, "epoch": 97, "lr": 9.97731884844666e-05}
{"train_loss": 1.076282262802124, "global_step": 8718, "epoch": 97, "lr": 9.977313332417721e-05}
{"train_loss": 0.6591734886169434, "global_step": 8719, "epoch": 97, "lr": 9.977307815719645e-05}
{"train_loss": 0.657049298286438, "global_step": 8720, "epoch": 97, "lr": 9.977302298352431e-05}
{"train_loss": 0.7803472762697199, "global_step": 8721, "epoch": 97, "lr": 9.977296780316079e-05, "val_loss": 1.2011958360671997}
{"train_loss": 0.8235501050949097, "global_step": 8722, "epoch": 98, "lr": 9.97729126161059e-05}
{"train_loss": 0.8545911312103271, "global_step": 8723, "epoch": 98, "lr": 9.977285742235964e-05}
{"train_loss": 0.8890230655670166, "global_step": 8724, "epoch": 98, "lr": 9.977280222192205e-05}
{"train_loss": 0.6778677105903625, "global_step": 8725, "epoch": 98, "lr": 9.977274701479309e-05}
{"train_loss": 0.625103235244751, "global_step": 8726, "epoch": 98, "lr": 9.97726918009728e-05}
{"train_loss": 0.841330349445343, "global_step": 8727, "epoch": 98, "lr": 9.977263658046119e-05}
{"train_loss": 0.7959259748458862, "global_step": 8728, "epoch": 98, "lr": 9.977258135325824e-05}
{"train_loss": 0.7752255201339722, "global_step": 8729, "epoch": 98, "lr": 9.977252611936397e-05}
{"train_loss": 0.8072275519371033, "global_step": 8730, "epoch": 98, "lr": 9.977247087877839e-05}
{"train_loss": 0.7423624396324158, "global_step": 8731, "epoch": 98, "lr": 9.977241563150153e-05}
{"train_loss": 0.7799875140190125, "global_step": 8732, "epoch": 98, "lr": 9.977236037753335e-05}
{"train_loss": 0.5831719636917114, "global_step": 8733, "epoch": 98, "lr": 9.97723051168739e-05}
{"train_loss": 0.6824504137039185, "global_step": 8734, "epoch": 98, "lr": 9.977224984952316e-05}
{"train_loss": 0.8097568154335022, "global_step": 8735, "epoch": 98, "lr": 9.977219457548115e-05}
{"train_loss": 0.8121993541717529, "global_step": 8736, "epoch": 98, "lr": 9.977213929474786e-05}
{"train_loss": 0.7889310121536255, "global_step": 8737, "epoch": 98, "lr": 9.977208400732333e-05}
{"train_loss": 0.7150989770889282, "global_step": 8738, "epoch": 98, "lr": 9.977202871320757e-05}
{"train_loss": 0.9174951314926147, "global_step": 8739, "epoch": 98, "lr": 9.977197341240053e-05}
{"train_loss": 0.8227608799934387, "global_step": 8740, "epoch": 98, "lr": 9.977191810490227e-05}
{"train_loss": 0.6188644766807556, "global_step": 8741, "epoch": 98, "lr": 9.977186279071278e-05}
{"train_loss": 0.6650702357292175, "global_step": 8742, "epoch": 98, "lr": 9.977180746983208e-05}
{"train_loss": 0.6752580404281616, "global_step": 8743, "epoch": 98, "lr": 9.977175214226015e-05}
{"train_loss": 0.7436378002166748, "global_step": 8744, "epoch": 98, "lr": 9.977169680799702e-05}
{"train_loss": 0.8944998383522034, "global_step": 8745, "epoch": 98, "lr": 9.977164146704271e-05}
{"train_loss": 0.779908299446106, "global_step": 8746, "epoch": 98, "lr": 9.977158611939719e-05}
{"train_loss": 0.7982754111289978, "global_step": 8747, "epoch": 98, "lr": 9.97715307650605e-05}
{"train_loss": 0.676161527633667, "global_step": 8748, "epoch": 98, "lr": 9.977147540403262e-05}
{"train_loss": 0.7701174020767212, "global_step": 8749, "epoch": 98, "lr": 9.977142003631358e-05}
{"train_loss": 0.7443163394927979, "global_step": 8750, "epoch": 98, "lr": 9.977136466190338e-05}
{"train_loss": 0.6866481900215149, "global_step": 8751, "epoch": 98, "lr": 9.977130928080202e-05}
{"train_loss": 0.7674988508224487, "global_step": 8752, "epoch": 98, "lr": 9.977125389300953e-05}
{"train_loss": 0.7336856722831726, "global_step": 8753, "epoch": 98, "lr": 9.977119849852588e-05}
{"train_loss": 0.7280396223068237, "global_step": 8754, "epoch": 98, "lr": 9.977114309735111e-05}
{"train_loss": 0.8911775946617126, "global_step": 8755, "epoch": 98, "lr": 9.977108768948521e-05}
{"train_loss": 0.8011754155158997, "global_step": 8756, "epoch": 98, "lr": 9.97710322749282e-05}
{"train_loss": 0.7946605086326599, "global_step": 8757, "epoch": 98, "lr": 9.977097685368009e-05}
{"train_loss": 0.7430784702301025, "global_step": 8758, "epoch": 98, "lr": 9.977092142574086e-05}
{"train_loss": 0.6935263276100159, "global_step": 8759, "epoch": 98, "lr": 9.977086599111054e-05}
{"train_loss": 0.8526368141174316, "global_step": 8760, "epoch": 98, "lr": 9.977081054978914e-05}
{"train_loss": 0.677291989326477, "global_step": 8761, "epoch": 98, "lr": 9.977075510177667e-05}
{"train_loss": 0.7087507247924805, "global_step": 8762, "epoch": 98, "lr": 9.977069964707311e-05}
{"train_loss": 0.7887417078018188, "global_step": 8763, "epoch": 98, "lr": 9.97706441856785e-05}
{"train_loss": 0.8767923712730408, "global_step": 8764, "epoch": 98, "lr": 9.977058871759283e-05}
{"train_loss": 0.7584044933319092, "global_step": 8765, "epoch": 98, "lr": 9.977053324281612e-05}
{"train_loss": 0.7282062768936157, "global_step": 8766, "epoch": 98, "lr": 9.977047776134835e-05}
{"train_loss": 0.7344235777854919, "global_step": 8767, "epoch": 98, "lr": 9.977042227318955e-05}
{"train_loss": 0.7863165736198425, "global_step": 8768, "epoch": 98, "lr": 9.977036677833973e-05}
{"train_loss": 0.7039987444877625, "global_step": 8769, "epoch": 98, "lr": 9.977031127679888e-05}
{"train_loss": 0.7929853796958923, "global_step": 8770, "epoch": 98, "lr": 9.977025576856703e-05}
{"train_loss": 0.750114917755127, "global_step": 8771, "epoch": 98, "lr": 9.977020025364417e-05}
{"train_loss": 0.6678932309150696, "global_step": 8772, "epoch": 98, "lr": 9.977014473203032e-05}
{"train_loss": 0.907856285572052, "global_step": 8773, "epoch": 98, "lr": 9.977008920372548e-05}
{"train_loss": 0.536158561706543, "global_step": 8774, "epoch": 98, "lr": 9.977003366872964e-05}
{"train_loss": 0.6557071805000305, "global_step": 8775, "epoch": 98, "lr": 9.976997812704285e-05}
{"train_loss": 0.6808086037635803, "global_step": 8776, "epoch": 98, "lr": 9.976992257866507e-05}
{"train_loss": 0.7740784883499146, "global_step": 8777, "epoch": 98, "lr": 9.976986702359634e-05}
{"train_loss": 0.7599469423294067, "global_step": 8778, "epoch": 98, "lr": 9.976981146183668e-05}
{"train_loss": 0.8712291717529297, "global_step": 8779, "epoch": 98, "lr": 9.976975589338604e-05}
{"train_loss": 0.6606669425964355, "global_step": 8780, "epoch": 98, "lr": 9.976970031824449e-05}
{"train_loss": 0.7690168023109436, "global_step": 8781, "epoch": 98, "lr": 9.976964473641199e-05}
{"train_loss": 0.7004274725914001, "global_step": 8782, "epoch": 98, "lr": 9.976958914788858e-05}
{"train_loss": 0.8197091221809387, "global_step": 8783, "epoch": 98, "lr": 9.976953355267425e-05}
{"train_loss": 0.7792962193489075, "global_step": 8784, "epoch": 98, "lr": 9.976947795076902e-05}
{"train_loss": 0.7887302041053772, "global_step": 8785, "epoch": 98, "lr": 9.976942234217289e-05}
{"train_loss": 0.6303508877754211, "global_step": 8786, "epoch": 98, "lr": 9.976936672688586e-05}
{"train_loss": 0.7804949283599854, "global_step": 8787, "epoch": 98, "lr": 9.976931110490795e-05}
{"train_loss": 0.7885687351226807, "global_step": 8788, "epoch": 98, "lr": 9.976925547623916e-05}
{"train_loss": 0.6263269782066345, "global_step": 8789, "epoch": 98, "lr": 9.976919984087951e-05}
{"train_loss": 0.8356251120567322, "global_step": 8790, "epoch": 98, "lr": 9.976914419882898e-05}
{"train_loss": 0.7022355794906616, "global_step": 8791, "epoch": 98, "lr": 9.97690885500876e-05}
{"train_loss": 0.9782346487045288, "global_step": 8792, "epoch": 98, "lr": 9.976903289465538e-05}
{"train_loss": 0.9190928339958191, "global_step": 8793, "epoch": 98, "lr": 9.976897723253232e-05}
{"train_loss": 0.6566773653030396, "global_step": 8794, "epoch": 98, "lr": 9.976892156371841e-05}
{"train_loss": 0.9452311992645264, "global_step": 8795, "epoch": 98, "lr": 9.976886588821371e-05}
{"train_loss": 0.8513807654380798, "global_step": 8796, "epoch": 98, "lr": 9.976881020601817e-05}
{"train_loss": 0.7337857484817505, "global_step": 8797, "epoch": 98, "lr": 9.976875451713182e-05}
{"train_loss": 0.7736929655075073, "global_step": 8798, "epoch": 98, "lr": 9.976869882155467e-05}
{"train_loss": 0.6890521049499512, "global_step": 8799, "epoch": 98, "lr": 9.976864311928673e-05}
{"train_loss": 0.7978430986404419, "global_step": 8800, "epoch": 98, "lr": 9.976858741032799e-05}
{"train_loss": 0.8201274275779724, "global_step": 8801, "epoch": 98, "lr": 9.976853169467848e-05}
{"train_loss": 0.8905335068702698, "global_step": 8802, "epoch": 98, "lr": 9.97684759723382e-05}
{"train_loss": 0.8343664407730103, "global_step": 8803, "epoch": 98, "lr": 9.976842024330715e-05}
{"train_loss": 0.8460679650306702, "global_step": 8804, "epoch": 98, "lr": 9.976836450758534e-05}
{"train_loss": 0.6530269384384155, "global_step": 8805, "epoch": 98, "lr": 9.976830876517278e-05}
{"train_loss": 0.6686140298843384, "global_step": 8806, "epoch": 98, "lr": 9.976825301606949e-05}
{"train_loss": 0.8640682697296143, "global_step": 8807, "epoch": 98, "lr": 9.976819726027545e-05}
{"train_loss": 0.7366685271263123, "global_step": 8808, "epoch": 98, "lr": 9.97681414977907e-05}
{"train_loss": 0.9819600582122803, "global_step": 8809, "epoch": 98, "lr": 9.97680857286152e-05}
{"train_loss": 0.7673747733737646, "global_step": 8810, "epoch": 98, "lr": 9.9768029952749e-05, "val_loss": 1.2090033292770386}
{"train_loss": 0.7207021117210388, "global_step": 8811, "epoch": 99, "lr": 9.976797417019211e-05}
{"train_loss": 0.7955110669136047, "global_step": 8812, "epoch": 99, "lr": 9.976791838094453e-05}
{"train_loss": 0.6814398765563965, "global_step": 8813, "epoch": 99, "lr": 9.976786258500624e-05}
{"train_loss": 0.7369686365127563, "global_step": 8814, "epoch": 99, "lr": 9.976780678237726e-05}
{"train_loss": 0.844720721244812, "global_step": 8815, "epoch": 99, "lr": 9.976775097305762e-05}
{"train_loss": 0.6805549263954163, "global_step": 8816, "epoch": 99, "lr": 9.97676951570473e-05}
{"train_loss": 0.7705270051956177, "global_step": 8817, "epoch": 99, "lr": 9.976763933434632e-05}
{"train_loss": 0.6903289556503296, "global_step": 8818, "epoch": 99, "lr": 9.97675835049547e-05}
{"train_loss": 0.7414244413375854, "global_step": 8819, "epoch": 99, "lr": 9.976752766887243e-05}
{"train_loss": 0.7391154766082764, "global_step": 8820, "epoch": 99, "lr": 9.976747182609953e-05}
{"train_loss": 0.7588574886322021, "global_step": 8821, "epoch": 99, "lr": 9.976741597663599e-05}
{"train_loss": 0.5620066523551941, "global_step": 8822, "epoch": 99, "lr": 9.976736012048182e-05}
{"train_loss": 0.8477253913879395, "global_step": 8823, "epoch": 99, "lr": 9.976730425763705e-05}
{"train_loss": 0.8299459218978882, "global_step": 8824, "epoch": 99, "lr": 9.976724838810168e-05}
{"train_loss": 0.7743079662322998, "global_step": 8825, "epoch": 99, "lr": 9.97671925118757e-05}
{"train_loss": 0.7202740907669067, "global_step": 8826, "epoch": 99, "lr": 9.976713662895912e-05}
{"train_loss": 0.7774364948272705, "global_step": 8827, "epoch": 99, "lr": 9.976708073935195e-05}
{"train_loss": 0.6310833096504211, "global_step": 8828, "epoch": 99, "lr": 9.976702484305422e-05}
{"train_loss": 0.740972101688385, "global_step": 8829, "epoch": 99, "lr": 9.976696894006592e-05}
{"train_loss": 0.688551127910614, "global_step": 8830, "epoch": 99, "lr": 9.976691303038705e-05}
{"train_loss": 0.7881482243537903, "global_step": 8831, "epoch": 99, "lr": 9.976685711401763e-05}
{"train_loss": 0.8630125522613525, "global_step": 8832, "epoch": 99, "lr": 9.976680119095765e-05}
{"train_loss": 0.9128261208534241, "global_step": 8833, "epoch": 99, "lr": 9.976674526120712e-05}
{"train_loss": 0.7566171288490295, "global_step": 8834, "epoch": 99, "lr": 9.97666893247661e-05}
{"train_loss": 0.5857567191123962, "global_step": 8835, "epoch": 99, "lr": 9.976663338163453e-05}
{"train_loss": 0.5699783563613892, "global_step": 8836, "epoch": 99, "lr": 9.976657743181244e-05}
{"train_loss": 0.8119814991950989, "global_step": 8837, "epoch": 99, "lr": 9.976652147529984e-05}
{"train_loss": 0.8144401907920837, "global_step": 8838, "epoch": 99, "lr": 9.976646551209675e-05}
{"train_loss": 0.6820385456085205, "global_step": 8839, "epoch": 99, "lr": 9.976640954220315e-05}
{"train_loss": 0.7697437405586243, "global_step": 8840, "epoch": 99, "lr": 9.976635356561908e-05}
{"train_loss": 0.9627568125724792, "global_step": 8841, "epoch": 99, "lr": 9.97662975823445e-05}
{"train_loss": 0.7825354933738708, "global_step": 8842, "epoch": 99, "lr": 9.976624159237948e-05}
{"train_loss": 0.7477921843528748, "global_step": 8843, "epoch": 99, "lr": 9.976618559572398e-05}
{"train_loss": 0.8318647146224976, "global_step": 8844, "epoch": 99, "lr": 9.976612959237803e-05}
{"train_loss": 0.7433019876480103, "global_step": 8845, "epoch": 99, "lr": 9.976607358234163e-05}
{"train_loss": 0.8943148851394653, "global_step": 8846, "epoch": 99, "lr": 9.976601756561478e-05}
{"train_loss": 0.8175702691078186, "global_step": 8847, "epoch": 99, "lr": 9.976596154219751e-05}
{"train_loss": 0.5450450778007507, "global_step": 8848, "epoch": 99, "lr": 9.97659055120898e-05}
{"train_loss": 0.7528993487358093, "global_step": 8849, "epoch": 99, "lr": 9.976584947529167e-05}
{"train_loss": 0.7267480492591858, "global_step": 8850, "epoch": 99, "lr": 9.976579343180312e-05}
{"train_loss": 1.027362585067749, "global_step": 8851, "epoch": 99, "lr": 9.976573738162418e-05}
{"train_loss": 0.6434029340744019, "global_step": 8852, "epoch": 99, "lr": 9.976568132475484e-05}
{"train_loss": 0.8828200101852417, "global_step": 8853, "epoch": 99, "lr": 9.976562526119512e-05}
{"train_loss": 0.8467764258384705, "global_step": 8854, "epoch": 99, "lr": 9.9765569190945e-05}
{"train_loss": 0.7215726375579834, "global_step": 8855, "epoch": 99, "lr": 9.976551311400452e-05}
{"train_loss": 0.6170885562896729, "global_step": 8856, "epoch": 99, "lr": 9.976545703037367e-05}
{"train_loss": 0.9747240543365479, "global_step": 8857, "epoch": 99, "lr": 9.976540094005247e-05}
{"train_loss": 0.668714702129364, "global_step": 8858, "epoch": 99, "lr": 9.97653448430409e-05}
{"train_loss": 0.8379462957382202, "global_step": 8859, "epoch": 99, "lr": 9.976528873933899e-05}
{"train_loss": 0.7693982720375061, "global_step": 8860, "epoch": 99, "lr": 9.976523262894675e-05}
{"train_loss": 0.9081922769546509, "global_step": 8861, "epoch": 99, "lr": 9.976517651186418e-05}
{"train_loss": 0.8024987578392029, "global_step": 8862, "epoch": 99, "lr": 9.976512038809129e-05}
{"train_loss": 0.7048368453979492, "global_step": 8863, "epoch": 99, "lr": 9.976506425762809e-05}
{"train_loss": 0.7072703242301941, "global_step": 8864, "epoch": 99, "lr": 9.976500812047457e-05}
{"train_loss": 0.8873635530471802, "global_step": 8865, "epoch": 99, "lr": 9.976495197663076e-05}
{"train_loss": 0.6631900668144226, "global_step": 8866, "epoch": 99, "lr": 9.976489582609665e-05}
{"train_loss": 0.8052118420600891, "global_step": 8867, "epoch": 99, "lr": 9.976483966887226e-05}
{"train_loss": 0.7381241917610168, "global_step": 8868, "epoch": 99, "lr": 9.97647835049576e-05}
{"train_loss": 0.7184420228004456, "global_step": 8869, "epoch": 99, "lr": 9.976472733435267e-05}
{"train_loss": 0.7874367833137512, "global_step": 8870, "epoch": 99, "lr": 9.976467115705748e-05}
{"train_loss": 0.7095975279808044, "global_step": 8871, "epoch": 99, "lr": 9.976461497307203e-05}
{"train_loss": 0.9006631374359131, "global_step": 8872, "epoch": 99, "lr": 9.976455878239635e-05}
{"train_loss": 0.8235030174255371, "global_step": 8873, "epoch": 99, "lr": 9.976450258503041e-05}
{"train_loss": 0.7620030641555786, "global_step": 8874, "epoch": 99, "lr": 9.976444638097425e-05}
{"train_loss": 0.7230198979377747, "global_step": 8875, "epoch": 99, "lr": 9.976439017022788e-05}
{"train_loss": 0.5554796457290649, "global_step": 8876, "epoch": 99, "lr": 9.976433395279127e-05}
{"train_loss": 0.597934365272522, "global_step": 8877, "epoch": 99, "lr": 9.976427772866446e-05}
{"train_loss": 0.7176162600517273, "global_step": 8878, "epoch": 99, "lr": 9.976422149784745e-05}
{"train_loss": 0.8460789322853088, "global_step": 8879, "epoch": 99, "lr": 9.976416526034025e-05}
{"train_loss": 0.8657874464988708, "global_step": 8880, "epoch": 99, "lr": 9.976410901614286e-05}
{"train_loss": 0.7061180472373962, "global_step": 8881, "epoch": 99, "lr": 9.976405276525529e-05}
{"train_loss": 0.7782805562019348, "global_step": 8882, "epoch": 99, "lr": 9.976399650767756e-05}
{"train_loss": 0.7506242990493774, "global_step": 8883, "epoch": 99, "lr": 9.976394024340966e-05}
{"train_loss": 0.7140517830848694, "global_step": 8884, "epoch": 99, "lr": 9.97638839724516e-05}
{"train_loss": 0.8175699710845947, "global_step": 8885, "epoch": 99, "lr": 9.97638276948034e-05}
{"train_loss": 0.7380543947219849, "global_step": 8886, "epoch": 99, "lr": 9.976377141046507e-05}
{"train_loss": 0.7599952816963196, "global_step": 8887, "epoch": 99, "lr": 9.976371511943658e-05}
{"train_loss": 0.7569363713264465, "global_step": 8888, "epoch": 99, "lr": 9.976365882171799e-05}
{"train_loss": 0.7199434638023376, "global_step": 8889, "epoch": 99, "lr": 9.976360251730928e-05}
{"train_loss": 0.7224142551422119, "global_step": 8890, "epoch": 99, "lr": 9.976354620621044e-05}
{"train_loss": 0.7923921346664429, "global_step": 8891, "epoch": 99, "lr": 9.976348988842152e-05}
{"train_loss": 0.770142138004303, "global_step": 8892, "epoch": 99, "lr": 9.976343356394249e-05}
{"train_loss": 0.7685322165489197, "global_step": 8893, "epoch": 99, "lr": 9.976337723277338e-05}
{"train_loss": 0.7617586255073547, "global_step": 8894, "epoch": 99, "lr": 9.97633208949142e-05}
{"train_loss": 0.8740368485450745, "global_step": 8895, "epoch": 99, "lr": 9.976326455036493e-05}
{"train_loss": 0.8036838173866272, "global_step": 8896, "epoch": 99, "lr": 9.976320819912561e-05}
{"train_loss": 0.7700157165527344, "global_step": 8897, "epoch": 99, "lr": 9.976315184119622e-05}
{"train_loss": 0.5704443454742432, "global_step": 8898, "epoch": 99, "lr": 9.976309547657679e-05}
{"train_loss": 0.7582676517829466, "global_step": 8899, "epoch": 99, "lr": 9.976303910526732e-05, "val_loss": 1.2214523553848267}
{"train_loss": 0.6061617136001587, "global_step": 8900, "epoch": 100, "lr": 9.97629827272678e-05}
{"train_loss": 0.9327266216278076, "global_step": 8901, "epoch": 100, "lr": 9.976292634257827e-05}
{"train_loss": 0.6185746192932129, "global_step": 8902, "epoch": 100, "lr": 9.976286995119872e-05}
{"train_loss": 0.7103487849235535, "global_step": 8903, "epoch": 100, "lr": 9.976281355312916e-05}
{"train_loss": 0.675951361656189, "global_step": 8904, "epoch": 100, "lr": 9.976275714836959e-05}
{"train_loss": 0.7797152996063232, "global_step": 8905, "epoch": 100, "lr": 9.976270073692003e-05}
{"train_loss": 0.6923990249633789, "global_step": 8906, "epoch": 100, "lr": 9.976264431878047e-05}
{"train_loss": 0.8459954857826233, "global_step": 8907, "epoch": 100, "lr": 9.976258789395094e-05}
{"train_loss": 0.8289494514465332, "global_step": 8908, "epoch": 100, "lr": 9.976253146243143e-05}
{"train_loss": 0.7362664341926575, "global_step": 8909, "epoch": 100, "lr": 9.976247502422195e-05}
{"train_loss": 0.7016642689704895, "global_step": 8910, "epoch": 100, "lr": 9.976241857932253e-05}
{"train_loss": 0.6729850172996521, "global_step": 8911, "epoch": 100, "lr": 9.976236212773314e-05}
{"train_loss": 0.7547470927238464, "global_step": 8912, "epoch": 100, "lr": 9.976230566945382e-05}
{"train_loss": 0.8191685676574707, "global_step": 8913, "epoch": 100, "lr": 9.976224920448456e-05}
{"train_loss": 0.6613373160362244, "global_step": 8914, "epoch": 100, "lr": 9.976219273282537e-05}
{"train_loss": 0.9011943936347961, "global_step": 8915, "epoch": 100, "lr": 9.976213625447626e-05}
{"train_loss": 0.6930715441703796, "global_step": 8916, "epoch": 100, "lr": 9.976207976943723e-05}
{"train_loss": 0.7817218899726868, "global_step": 8917, "epoch": 100, "lr": 9.976202327770832e-05}
{"train_loss": 0.6812463998794556, "global_step": 8918, "epoch": 100, "lr": 9.976196677928949e-05}
{"train_loss": 0.7862012982368469, "global_step": 8919, "epoch": 100, "lr": 9.976191027418078e-05}
{"train_loss": 0.857231855392456, "global_step": 8920, "epoch": 100, "lr": 9.976185376238217e-05}
{"train_loss": 0.5798540711402893, "global_step": 8921, "epoch": 100, "lr": 9.97617972438937e-05}
{"train_loss": 0.6913115382194519, "global_step": 8922, "epoch": 100, "lr": 9.976174071871536e-05}
{"train_loss": 0.7405151724815369, "global_step": 8923, "epoch": 100, "lr": 9.976168418684717e-05}
{"train_loss": 0.8306033611297607, "global_step": 8924, "epoch": 100, "lr": 9.976162764828912e-05}
{"train_loss": 0.9350431561470032, "global_step": 8925, "epoch": 100, "lr": 9.976157110304122e-05}
{"train_loss": 0.8110305666923523, "global_step": 8926, "epoch": 100, "lr": 9.976151455110349e-05}
{"train_loss": 0.7645607590675354, "global_step": 8927, "epoch": 100, "lr": 9.976145799247594e-05}
{"train_loss": 0.6984178423881531, "global_step": 8928, "epoch": 100, "lr": 9.976140142715855e-05}
{"train_loss": 0.8105949759483337, "global_step": 8929, "epoch": 100, "lr": 9.976134485515135e-05}
{"train_loss": 0.6903896331787109, "global_step": 8930, "epoch": 100, "lr": 9.976128827645434e-05}
{"train_loss": 0.6716237664222717, "global_step": 8931, "epoch": 100, "lr": 9.976123169106755e-05}
{"train_loss": 0.7542504668235779, "global_step": 8932, "epoch": 100, "lr": 9.976117509899095e-05}
{"train_loss": 0.7084846496582031, "global_step": 8933, "epoch": 100, "lr": 9.976111850022458e-05}
{"train_loss": 0.6304293274879456, "global_step": 8934, "epoch": 100, "lr": 9.976106189476844e-05}
{"train_loss": 0.6450044512748718, "global_step": 8935, "epoch": 100, "lr": 9.976100528262251e-05}
{"train_loss": 0.8591590523719788, "global_step": 8936, "epoch": 100, "lr": 9.976094866378683e-05}
{"train_loss": 0.8942747116088867, "global_step": 8937, "epoch": 100, "lr": 9.97608920382614e-05}
{"train_loss": 0.8836408853530884, "global_step": 8938, "epoch": 100, "lr": 9.976083540604621e-05}
{"train_loss": 0.828832745552063, "global_step": 8939, "epoch": 100, "lr": 9.97607787671413e-05}
{"train_loss": 0.7955162525177002, "global_step": 8940, "epoch": 100, "lr": 9.976072212154666e-05}
{"train_loss": 0.6669787168502808, "global_step": 8941, "epoch": 100, "lr": 9.976066546926228e-05}
{"train_loss": 0.8346713185310364, "global_step": 8942, "epoch": 100, "lr": 9.97606088102882e-05}
{"train_loss": 0.7104343771934509, "global_step": 8943, "epoch": 100, "lr": 9.976055214462441e-05}
{"train_loss": 0.7548291683197021, "global_step": 8944, "epoch": 100, "lr": 9.976049547227093e-05}
{"train_loss": 0.8771553635597229, "global_step": 8945, "epoch": 100, "lr": 9.976043879322774e-05}
{"train_loss": 0.9060064554214478, "global_step": 8946, "epoch": 100, "lr": 9.976038210749488e-05}
{"train_loss": 0.659052312374115, "global_step": 8947, "epoch": 100, "lr": 9.976032541507232e-05}
{"train_loss": 0.8533231616020203, "global_step": 8948, "epoch": 100, "lr": 9.976026871596011e-05}
{"train_loss": 0.8134006261825562, "global_step": 8949, "epoch": 100, "lr": 9.976021201015824e-05}
{"train_loss": 1.0040394067764282, "global_step": 8950, "epoch": 100, "lr": 9.976015529766671e-05}
{"train_loss": 0.866891086101532, "global_step": 8951, "epoch": 100, "lr": 9.976009857848552e-05}
{"train_loss": 0.9017369151115417, "global_step": 8952, "epoch": 100, "lr": 9.976004185261471e-05}
{"train_loss": 0.8040116429328918, "global_step": 8953, "epoch": 100, "lr": 9.975998512005426e-05}
{"train_loss": 0.6990986466407776, "global_step": 8954, "epoch": 100, "lr": 9.97599283808042e-05}
{"train_loss": 0.8794538974761963, "global_step": 8955, "epoch": 100, "lr": 9.975987163486451e-05}
{"train_loss": 0.7714559435844421, "global_step": 8956, "epoch": 100, "lr": 9.975981488223522e-05}
{"train_loss": 0.8532336354255676, "global_step": 8957, "epoch": 100, "lr": 9.975975812291632e-05}
{"train_loss": 0.7538276314735413, "global_step": 8958, "epoch": 100, "lr": 9.975970135690784e-05}
{"train_loss": 0.7953503727912903, "global_step": 8959, "epoch": 100, "lr": 9.975964458420977e-05}
{"train_loss": 0.7196053862571716, "global_step": 8960, "epoch": 100, "lr": 9.975958780482211e-05}
{"train_loss": 0.8721734881401062, "global_step": 8961, "epoch": 100, "lr": 9.975953101874489e-05}
{"train_loss": 0.7757407426834106, "global_step": 8962, "epoch": 100, "lr": 9.975947422597811e-05}
{"train_loss": 0.6847500801086426, "global_step": 8963, "epoch": 100, "lr": 9.975941742652178e-05}
{"train_loss": 0.7445845007896423, "global_step": 8964, "epoch": 100, "lr": 9.975936062037589e-05}
{"train_loss": 0.7543805837631226, "global_step": 8965, "epoch": 100, "lr": 9.975930380754046e-05}
{"train_loss": 0.8235307335853577, "global_step": 8966, "epoch": 100, "lr": 9.975924698801551e-05}
{"train_loss": 0.7522087693214417, "global_step": 8967, "epoch": 100, "lr": 9.975919016180102e-05}
{"train_loss": 0.7097089886665344, "global_step": 8968, "epoch": 100, "lr": 9.975913332889704e-05}
{"train_loss": 0.8655797243118286, "global_step": 8969, "epoch": 100, "lr": 9.975907648930353e-05}
{"train_loss": 0.755668580532074, "global_step": 8970, "epoch": 100, "lr": 9.975901964302052e-05}
{"train_loss": 0.9123867750167847, "global_step": 8971, "epoch": 100, "lr": 9.975896279004802e-05}
{"train_loss": 0.7319115400314331, "global_step": 8972, "epoch": 100, "lr": 9.975890593038604e-05}
{"train_loss": 0.7636774182319641, "global_step": 8973, "epoch": 100, "lr": 9.975884906403456e-05}
{"train_loss": 0.728227972984314, "global_step": 8974, "epoch": 100, "lr": 9.975879219099362e-05}
{"train_loss": 1.108519434928894, "global_step": 8975, "epoch": 100, "lr": 9.975873531126323e-05}
{"train_loss": 0.8170779943466187, "global_step": 8976, "epoch": 100, "lr": 9.975867842484336e-05}
{"train_loss": 0.6954312920570374, "global_step": 8977, "epoch": 100, "lr": 9.975862153173408e-05}
{"train_loss": 0.6754558682441711, "global_step": 8978, "epoch": 100, "lr": 9.975856463193533e-05}
{"train_loss": 0.7502942085266113, "global_step": 8979, "epoch": 100, "lr": 9.975850772544715e-05}
{"train_loss": 0.760344386100769, "global_step": 8980, "epoch": 100, "lr": 9.975845081226956e-05}
{"train_loss": 0.8329525589942932, "global_step": 8981, "epoch": 100, "lr": 9.975839389240253e-05}
{"train_loss": 0.5377368927001953, "global_step": 8982, "epoch": 100, "lr": 9.975833696584611e-05}
{"train_loss": 0.6739488244056702, "global_step": 8983, "epoch": 100, "lr": 9.975828003260029e-05}
{"train_loss": 0.777940034866333, "global_step": 8984, "epoch": 100, "lr": 9.975822309266504e-05}
{"train_loss": 0.6703535318374634, "global_step": 8985, "epoch": 100, "lr": 9.975816614604045e-05}
{"train_loss": 0.7628877758979797, "global_step": 8986, "epoch": 100, "lr": 9.975810919272645e-05}
{"train_loss": 0.8041269779205322, "global_step": 8987, "epoch": 100, "lr": 9.97580522327231e-05}
{"train_loss": 0.7719695661844832, "global_step": 8988, "epoch": 100, "lr": 9.975799526603037e-05, "train/sim_max_reward_0": 0.4903115798636688, "train/sim_max_reward_1": 0.802740951252063, "train/sim_max_reward_2": 0.44868132113264936, "train/sim_max_reward_3": 0.4352935975198635, "train/sim_max_reward_4": 0.49078869104090916, "train/sim_max_reward_5": 0.7014437335828497, "test/sim_max_reward_4500000": 0.4586817394622463, "test/sim_max_reward_4500001": 0.8692317449718084, "test/sim_max_reward_4500002": 0.7246158399798339, "test/sim_max_reward_4500003": 0.2261365710961097, "test/sim_max_reward_4500004": 0.4539928100661603, "test/sim_max_reward_4500005": 0.9250314458762344, "test/sim_max_reward_4500006": 0.8825421959073855, "test/sim_max_reward_4500007": 0.4323499293984466, "test/sim_max_reward_4500008": 0.9343039902830311, "test/sim_max_reward_4500009": 0.916057074566355, "test/sim_max_reward_4500010": 0.33768558821859457, "test/sim_max_reward_4500011": 0.9693622349762344, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.340167066099364, "test/sim_max_reward_4500014": 0.9862625494503466, "test/sim_max_reward_4500015": 0.3784721400461217, "test/sim_max_reward_4500016": 0.986051645907781, "test/sim_max_reward_4500017": 0.2009746237473517, "test/sim_max_reward_4500018": 0.7371249595742685, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.23340621216291393, "test/sim_max_reward_4500021": 0.5055687237119776, "test/sim_max_reward_4500022": 0.7330548248857722, "test/sim_max_reward_4500023": 0.7135258138110286, "test/sim_max_reward_4500024": 0.603604788625639, "test/sim_max_reward_4500025": 0.44427102679814623, "test/sim_max_reward_4500026": 0.0, "test/sim_max_reward_4500027": 0.7096919766417823, "test/sim_max_reward_4500028": 0.7269340996046744, "test/sim_max_reward_4500029": 0.48129162238255185, "test/sim_max_reward_4500030": 0.9982342479253548, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.3694697623313258, "test/sim_max_reward_4500033": 0.6229876582844103, "test/sim_max_reward_4500034": 0.467426651912268, "test/sim_max_reward_4500035": 0.6888379823407272, "test/sim_max_reward_4500036": 0.577863202384648, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.5590362161211506, "test/sim_max_reward_4500039": 0.5798070401891212, "test/sim_max_reward_4500040": 0.9703636399163532, "test/sim_max_reward_4500041": 0.46701347391848785, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.6389943200974347, "test/sim_max_reward_4500044": 0.48600301046443883, "test/sim_max_reward_4500045": 0.5205253469944753, "test/sim_max_reward_4500046": 0.4524360662101321, "test/sim_max_reward_4500047": 0.5098946370120968, "test/sim_max_reward_4500048": 0.26959223737431576, "test/sim_max_reward_4500049": 0.886718406269888, "train/mean_score": 0.5615433123986673, "test/mean_score": 0.5595119427599758, "val_loss": 1.212754487991333, "train_action_mse_error": 40.39751434326172}
{"train_loss": 0.8711550831794739, "global_step": 8989, "epoch": 101, "lr": 9.975793829264829e-05}
{"train_loss": 0.7926644682884216, "global_step": 8990, "epoch": 101, "lr": 9.975788131257686e-05}
{"train_loss": 0.9526244401931763, "global_step": 8991, "epoch": 101, "lr": 9.975782432581609e-05}
{"train_loss": 0.8823350667953491, "global_step": 8992, "epoch": 101, "lr": 9.9757767332366e-05}
{"train_loss": 0.713455855846405, "global_step": 8993, "epoch": 101, "lr": 9.975771033222657e-05}
{"train_loss": 0.6716702580451965, "global_step": 8994, "epoch": 101, "lr": 9.975765332539781e-05}
{"train_loss": 0.7177920937538147, "global_step": 8995, "epoch": 101, "lr": 9.975759631187976e-05}
{"train_loss": 0.8070473670959473, "global_step": 8996, "epoch": 101, "lr": 9.97575392916724e-05}
{"train_loss": 0.7980887293815613, "global_step": 8997, "epoch": 101, "lr": 9.975748226477575e-05}
{"train_loss": 0.8143099546432495, "global_step": 8998, "epoch": 101, "lr": 9.97574252311898e-05}
{"train_loss": 0.6351383328437805, "global_step": 8999, "epoch": 101, "lr": 9.975736819091459e-05}
{"train_loss": 0.8545238971710205, "global_step": 9000, "epoch": 101, "lr": 9.975731114395011e-05}
{"train_loss": 0.8233932256698608, "global_step": 9001, "epoch": 101, "lr": 9.975725409029634e-05}
{"train_loss": 0.6482101082801819, "global_step": 9002, "epoch": 101, "lr": 9.975719702995335e-05}
{"train_loss": 0.7851378917694092, "global_step": 9003, "epoch": 101, "lr": 9.975713996292108e-05}
{"train_loss": 0.7274391651153564, "global_step": 9004, "epoch": 101, "lr": 9.975708288919957e-05}
{"train_loss": 0.6586435437202454, "global_step": 9005, "epoch": 101, "lr": 9.975702580878884e-05}
{"train_loss": 0.6023716926574707, "global_step": 9006, "epoch": 101, "lr": 9.975696872168888e-05}
{"train_loss": 0.6376179456710815, "global_step": 9007, "epoch": 101, "lr": 9.975691162789971e-05}
{"train_loss": 0.5847235321998596, "global_step": 9008, "epoch": 101, "lr": 9.975685452742132e-05}
{"train_loss": 0.8304945826530457, "global_step": 9009, "epoch": 101, "lr": 9.975679742025373e-05}
{"train_loss": 0.6929196119308472, "global_step": 9010, "epoch": 101, "lr": 9.975674030639694e-05}
{"train_loss": 0.6417720913887024, "global_step": 9011, "epoch": 101, "lr": 9.975668318585096e-05}
{"train_loss": 0.6839131712913513, "global_step": 9012, "epoch": 101, "lr": 9.975662605861582e-05}
{"train_loss": 0.7538391947746277, "global_step": 9013, "epoch": 101, "lr": 9.97565689246915e-05}
{"train_loss": 0.7466036081314087, "global_step": 9014, "epoch": 101, "lr": 9.9756511784078e-05}
{"train_loss": 0.8628036379814148, "global_step": 9015, "epoch": 101, "lr": 9.975645463677535e-05}
{"train_loss": 0.7228642702102661, "global_step": 9016, "epoch": 101, "lr": 9.975639748278356e-05}
{"train_loss": 0.7521630525588989, "global_step": 9017, "epoch": 101, "lr": 9.975634032210262e-05}
{"train_loss": 0.8303748369216919, "global_step": 9018, "epoch": 101, "lr": 9.975628315473256e-05}
{"train_loss": 0.682832658290863, "global_step": 9019, "epoch": 101, "lr": 9.975622598067335e-05}
{"train_loss": 0.9987453818321228, "global_step": 9020, "epoch": 101, "lr": 9.975616879992505e-05}
{"train_loss": 0.7641030550003052, "global_step": 9021, "epoch": 101, "lr": 9.975611161248763e-05}
{"train_loss": 0.645969569683075, "global_step": 9022, "epoch": 101, "lr": 9.97560544183611e-05}
{"train_loss": 0.8226207494735718, "global_step": 9023, "epoch": 101, "lr": 9.975599721754547e-05}
{"train_loss": 0.7606338858604431, "global_step": 9024, "epoch": 101, "lr": 9.975594001004078e-05}
{"train_loss": 0.712914228439331, "global_step": 9025, "epoch": 101, "lr": 9.975588279584697e-05}
{"train_loss": 0.6561746001243591, "global_step": 9026, "epoch": 101, "lr": 9.975582557496413e-05}
{"train_loss": 0.7892663478851318, "global_step": 9027, "epoch": 101, "lr": 9.97557683473922e-05}
{"train_loss": 0.8380330801010132, "global_step": 9028, "epoch": 101, "lr": 9.975571111313123e-05}
{"train_loss": 0.6799769997596741, "global_step": 9029, "epoch": 101, "lr": 9.975565387218118e-05}
{"train_loss": 0.7165668606758118, "global_step": 9030, "epoch": 101, "lr": 9.975559662454211e-05}
{"train_loss": 0.6497495770454407, "global_step": 9031, "epoch": 101, "lr": 9.975553937021402e-05}
{"train_loss": 0.7239420413970947, "global_step": 9032, "epoch": 101, "lr": 9.975548210919688e-05}
{"train_loss": 0.7295576930046082, "global_step": 9033, "epoch": 101, "lr": 9.975542484149073e-05}
{"train_loss": 0.9325239658355713, "global_step": 9034, "epoch": 101, "lr": 9.975536756709559e-05}
{"train_loss": 0.7356744408607483, "global_step": 9035, "epoch": 101, "lr": 9.975531028601141e-05}
{"train_loss": 0.7347716689109802, "global_step": 9036, "epoch": 101, "lr": 9.975525299823824e-05}
{"train_loss": 0.840044379234314, "global_step": 9037, "epoch": 101, "lr": 9.975519570377611e-05}
{"train_loss": 0.735075831413269, "global_step": 9038, "epoch": 101, "lr": 9.975513840262498e-05}
{"train_loss": 0.7696337699890137, "global_step": 9039, "epoch": 101, "lr": 9.975508109478488e-05}
{"train_loss": 0.8190920352935791, "global_step": 9040, "epoch": 101, "lr": 9.975502378025582e-05}
{"train_loss": 0.8702782392501831, "global_step": 9041, "epoch": 101, "lr": 9.97549664590378e-05}
{"train_loss": 0.5980448722839355, "global_step": 9042, "epoch": 101, "lr": 9.975490913113084e-05}
{"train_loss": 0.6997002959251404, "global_step": 9043, "epoch": 101, "lr": 9.975485179653494e-05}
{"train_loss": 0.6730693578720093, "global_step": 9044, "epoch": 101, "lr": 9.97547944552501e-05}
{"train_loss": 0.8190255761146545, "global_step": 9045, "epoch": 101, "lr": 9.975473710727635e-05}
{"train_loss": 0.5987818241119385, "global_step": 9046, "epoch": 101, "lr": 9.975467975261366e-05}
{"train_loss": 0.7962602972984314, "global_step": 9047, "epoch": 101, "lr": 9.975462239126206e-05}
{"train_loss": 0.7113010883331299, "global_step": 9048, "epoch": 101, "lr": 9.975456502322156e-05}
{"train_loss": 0.8702127933502197, "global_step": 9049, "epoch": 101, "lr": 9.975450764849217e-05}
{"train_loss": 0.8360751867294312, "global_step": 9050, "epoch": 101, "lr": 9.975445026707389e-05}
{"train_loss": 0.8572524189949036, "global_step": 9051, "epoch": 101, "lr": 9.975439287896673e-05}
{"train_loss": 0.6601847410202026, "global_step": 9052, "epoch": 101, "lr": 9.97543354841707e-05}
{"train_loss": 0.7991739511489868, "global_step": 9053, "epoch": 101, "lr": 9.975427808268582e-05}
{"train_loss": 0.6409960389137268, "global_step": 9054, "epoch": 101, "lr": 9.975422067451206e-05}
{"train_loss": 0.8609992265701294, "global_step": 9055, "epoch": 101, "lr": 9.975416325964946e-05}
{"train_loss": 0.7101247906684875, "global_step": 9056, "epoch": 101, "lr": 9.975410583809803e-05}
{"train_loss": 0.6806632876396179, "global_step": 9057, "epoch": 101, "lr": 9.975404840985775e-05}
{"train_loss": 0.6973774433135986, "global_step": 9058, "epoch": 101, "lr": 9.975399097492865e-05}
{"train_loss": 0.7678913474082947, "global_step": 9059, "epoch": 101, "lr": 9.975393353331075e-05}
{"train_loss": 0.9065580368041992, "global_step": 9060, "epoch": 101, "lr": 9.975387608500401e-05}
{"train_loss": 0.690301239490509, "global_step": 9061, "epoch": 101, "lr": 9.975381863000849e-05}
{"train_loss": 0.6627174615859985, "global_step": 9062, "epoch": 101, "lr": 9.975376116832416e-05}
{"train_loss": 0.7197381258010864, "global_step": 9063, "epoch": 101, "lr": 9.975370369995104e-05}
{"train_loss": 0.6681830883026123, "global_step": 9064, "epoch": 101, "lr": 9.975364622488916e-05}
{"train_loss": 0.6574938297271729, "global_step": 9065, "epoch": 101, "lr": 9.975358874313852e-05}
{"train_loss": 0.7258249521255493, "global_step": 9066, "epoch": 101, "lr": 9.975353125469908e-05}
{"train_loss": 0.6075229644775391, "global_step": 9067, "epoch": 101, "lr": 9.975347375957091e-05}
{"train_loss": 0.871225118637085, "global_step": 9068, "epoch": 101, "lr": 9.975341625775398e-05}
{"train_loss": 0.7799164056777954, "global_step": 9069, "epoch": 101, "lr": 9.975335874924832e-05}
{"train_loss": 0.8991451263427734, "global_step": 9070, "epoch": 101, "lr": 9.975330123405391e-05}
{"train_loss": 0.8509357571601868, "global_step": 9071, "epoch": 101, "lr": 9.975324371217078e-05}
{"train_loss": 0.5693328380584717, "global_step": 9072, "epoch": 101, "lr": 9.975318618359894e-05}
{"train_loss": 0.8942676186561584, "global_step": 9073, "epoch": 101, "lr": 9.97531286483384e-05}
{"train_loss": 0.9332727789878845, "global_step": 9074, "epoch": 101, "lr": 9.975307110638913e-05}
{"train_loss": 0.6778814792633057, "global_step": 9075, "epoch": 101, "lr": 9.97530135577512e-05}
{"train_loss": 0.6227734684944153, "global_step": 9076, "epoch": 101, "lr": 9.975295600242455e-05}
{"train_loss": 0.7500341983323686, "global_step": 9077, "epoch": 101, "lr": 9.975289844040924e-05, "val_loss": 1.1906242370605469}
{"train_loss": 0.8448644280433655, "global_step": 9078, "epoch": 102, "lr": 9.975284087170526e-05}
{"train_loss": 0.7489856481552124, "global_step": 9079, "epoch": 102, "lr": 9.975278329631262e-05}
{"train_loss": 0.7372860312461853, "global_step": 9080, "epoch": 102, "lr": 9.975272571423133e-05}
{"train_loss": 0.8446584939956665, "global_step": 9081, "epoch": 102, "lr": 9.975266812546139e-05}
{"train_loss": 0.6001437902450562, "global_step": 9082, "epoch": 102, "lr": 9.975261053000279e-05}
{"train_loss": 0.7814887166023254, "global_step": 9083, "epoch": 102, "lr": 9.975255292785556e-05}
{"train_loss": 0.7241829633712769, "global_step": 9084, "epoch": 102, "lr": 9.975249531901972e-05}
{"train_loss": 0.668940007686615, "global_step": 9085, "epoch": 102, "lr": 9.975243770349526e-05}
{"train_loss": 0.7518081068992615, "global_step": 9086, "epoch": 102, "lr": 9.975238008128218e-05}
{"train_loss": 0.7550885081291199, "global_step": 9087, "epoch": 102, "lr": 9.975232245238051e-05}
{"train_loss": 0.6546159982681274, "global_step": 9088, "epoch": 102, "lr": 9.975226481679026e-05}
{"train_loss": 0.7588263154029846, "global_step": 9089, "epoch": 102, "lr": 9.975220717451141e-05}
{"train_loss": 0.7342635989189148, "global_step": 9090, "epoch": 102, "lr": 9.975214952554397e-05}
{"train_loss": 0.6350533962249756, "global_step": 9091, "epoch": 102, "lr": 9.975209186988798e-05}
{"train_loss": 0.7867825627326965, "global_step": 9092, "epoch": 102, "lr": 9.975203420754342e-05}
{"train_loss": 0.5686153173446655, "global_step": 9093, "epoch": 102, "lr": 9.97519765385103e-05}
{"train_loss": 0.867091953754425, "global_step": 9094, "epoch": 102, "lr": 9.975191886278865e-05}
{"train_loss": 0.8637822270393372, "global_step": 9095, "epoch": 102, "lr": 9.975186118037843e-05}
{"train_loss": 0.5102554559707642, "global_step": 9096, "epoch": 102, "lr": 9.975180349127971e-05}
{"train_loss": 0.6613248586654663, "global_step": 9097, "epoch": 102, "lr": 9.975174579549245e-05}
{"train_loss": 0.6404699683189392, "global_step": 9098, "epoch": 102, "lr": 9.975168809301668e-05}
{"train_loss": 0.7213165163993835, "global_step": 9099, "epoch": 102, "lr": 9.97516303838524e-05}
{"train_loss": 0.7926260828971863, "global_step": 9100, "epoch": 102, "lr": 9.975157266799961e-05}
{"train_loss": 0.7895996570587158, "global_step": 9101, "epoch": 102, "lr": 9.975151494545833e-05}
{"train_loss": 0.6676880121231079, "global_step": 9102, "epoch": 102, "lr": 9.975145721622858e-05}
{"train_loss": 0.6702986359596252, "global_step": 9103, "epoch": 102, "lr": 9.975139948031035e-05}
{"train_loss": 0.7923635840415955, "global_step": 9104, "epoch": 102, "lr": 9.975134173770365e-05}
{"train_loss": 0.696332573890686, "global_step": 9105, "epoch": 102, "lr": 9.975128398840847e-05}
{"train_loss": 0.8349077105522156, "global_step": 9106, "epoch": 102, "lr": 9.975122623242486e-05}
{"train_loss": 0.7094921469688416, "global_step": 9107, "epoch": 102, "lr": 9.975116846975278e-05}
{"train_loss": 0.7707061767578125, "global_step": 9108, "epoch": 102, "lr": 9.97511107003923e-05}
{"train_loss": 0.7944194674491882, "global_step": 9109, "epoch": 102, "lr": 9.975105292434336e-05}
{"train_loss": 0.8506485223770142, "global_step": 9110, "epoch": 102, "lr": 9.9750995141606e-05}
{"train_loss": 0.7505378723144531, "global_step": 9111, "epoch": 102, "lr": 9.975093735218023e-05}
{"train_loss": 0.8013775944709778, "global_step": 9112, "epoch": 102, "lr": 9.975087955606605e-05}
{"train_loss": 0.7247887253761292, "global_step": 9113, "epoch": 102, "lr": 9.975082175326347e-05}
{"train_loss": 0.8647251725196838, "global_step": 9114, "epoch": 102, "lr": 9.97507639437725e-05}
{"train_loss": 0.858383297920227, "global_step": 9115, "epoch": 102, "lr": 9.975070612759314e-05}
{"train_loss": 0.7898619771003723, "global_step": 9116, "epoch": 102, "lr": 9.975064830472543e-05}
{"train_loss": 0.874096155166626, "global_step": 9117, "epoch": 102, "lr": 9.975059047516933e-05}
{"train_loss": 0.7412236928939819, "global_step": 9118, "epoch": 102, "lr": 9.975053263892487e-05}
{"train_loss": 0.7265252470970154, "global_step": 9119, "epoch": 102, "lr": 9.975047479599204e-05}
{"train_loss": 0.7968029975891113, "global_step": 9120, "epoch": 102, "lr": 9.97504169463709e-05}
{"train_loss": 0.8199188709259033, "global_step": 9121, "epoch": 102, "lr": 9.975035909006141e-05}
{"train_loss": 0.8229728937149048, "global_step": 9122, "epoch": 102, "lr": 9.975030122706358e-05}
{"train_loss": 0.6570053696632385, "global_step": 9123, "epoch": 102, "lr": 9.975024335737743e-05}
{"train_loss": 0.8971903920173645, "global_step": 9124, "epoch": 102, "lr": 9.975018548100298e-05}
{"train_loss": 0.9300413131713867, "global_step": 9125, "epoch": 102, "lr": 9.975012759794021e-05}
{"train_loss": 0.8689076900482178, "global_step": 9126, "epoch": 102, "lr": 9.975006970818914e-05}
{"train_loss": 0.6022595763206482, "global_step": 9127, "epoch": 102, "lr": 9.975001181174979e-05}
{"train_loss": 0.8308444023132324, "global_step": 9128, "epoch": 102, "lr": 9.974995390862216e-05}
{"train_loss": 0.7894186973571777, "global_step": 9129, "epoch": 102, "lr": 9.974989599880625e-05}
{"train_loss": 0.7868560552597046, "global_step": 9130, "epoch": 102, "lr": 9.974983808230207e-05}
{"train_loss": 0.8301494717597961, "global_step": 9131, "epoch": 102, "lr": 9.974978015910963e-05}
{"train_loss": 0.7219789624214172, "global_step": 9132, "epoch": 102, "lr": 9.974972222922895e-05}
{"train_loss": 0.857079803943634, "global_step": 9133, "epoch": 102, "lr": 9.974966429266002e-05}
{"train_loss": 0.6670399308204651, "global_step": 9134, "epoch": 102, "lr": 9.974960634940284e-05}
{"train_loss": 0.8413628339767456, "global_step": 9135, "epoch": 102, "lr": 9.974954839945745e-05}
{"train_loss": 0.7759177088737488, "global_step": 9136, "epoch": 102, "lr": 9.974949044282383e-05}
{"train_loss": 0.6511059999465942, "global_step": 9137, "epoch": 102, "lr": 9.9749432479502e-05}
{"train_loss": 0.7303876280784607, "global_step": 9138, "epoch": 102, "lr": 9.974937450949196e-05}
{"train_loss": 0.7016096711158752, "global_step": 9139, "epoch": 102, "lr": 9.974931653279373e-05}
{"train_loss": 0.7313594222068787, "global_step": 9140, "epoch": 102, "lr": 9.97492585494073e-05}
{"train_loss": 0.8414132595062256, "global_step": 9141, "epoch": 102, "lr": 9.974920055933271e-05}
{"train_loss": 0.9796369671821594, "global_step": 9142, "epoch": 102, "lr": 9.974914256256993e-05}
{"train_loss": 0.688096821308136, "global_step": 9143, "epoch": 102, "lr": 9.9749084559119e-05}
{"train_loss": 0.6617946624755859, "global_step": 9144, "epoch": 102, "lr": 9.974902654897989e-05}
{"train_loss": 0.8559627532958984, "global_step": 9145, "epoch": 102, "lr": 9.974896853215264e-05}
{"train_loss": 0.7129533886909485, "global_step": 9146, "epoch": 102, "lr": 9.974891050863724e-05}
{"train_loss": 0.70489102602005, "global_step": 9147, "epoch": 102, "lr": 9.974885247843372e-05}
{"train_loss": 0.8903313279151917, "global_step": 9148, "epoch": 102, "lr": 9.974879444154206e-05}
{"train_loss": 0.5903881192207336, "global_step": 9149, "epoch": 102, "lr": 9.97487363979623e-05}
{"train_loss": 0.7073323726654053, "global_step": 9150, "epoch": 102, "lr": 9.974867834769442e-05}
{"train_loss": 0.8550249338150024, "global_step": 9151, "epoch": 102, "lr": 9.974862029073842e-05}
{"train_loss": 0.7081283330917358, "global_step": 9152, "epoch": 102, "lr": 9.974856222709434e-05}
{"train_loss": 0.7248548269271851, "global_step": 9153, "epoch": 102, "lr": 9.974850415676216e-05}
{"train_loss": 0.8398474454879761, "global_step": 9154, "epoch": 102, "lr": 9.974844607974192e-05}
{"train_loss": 0.7335706949234009, "global_step": 9155, "epoch": 102, "lr": 9.974838799603358e-05}
{"train_loss": 0.7536629438400269, "global_step": 9156, "epoch": 102, "lr": 9.97483299056372e-05}
{"train_loss": 0.6526011824607849, "global_step": 9157, "epoch": 102, "lr": 9.974827180855275e-05}
{"train_loss": 0.728635311126709, "global_step": 9158, "epoch": 102, "lr": 9.974821370478027e-05}
{"train_loss": 0.7885127067565918, "global_step": 9159, "epoch": 102, "lr": 9.974815559431973e-05}
{"train_loss": 0.7237558364868164, "global_step": 9160, "epoch": 102, "lr": 9.974809747717116e-05}
{"train_loss": 0.708317220211029, "global_step": 9161, "epoch": 102, "lr": 9.974803935333456e-05}
{"train_loss": 0.8510299324989319, "global_step": 9162, "epoch": 102, "lr": 9.974798122280995e-05}
{"train_loss": 0.8029384016990662, "global_step": 9163, "epoch": 102, "lr": 9.974792308559733e-05}
{"train_loss": 0.8322094678878784, "global_step": 9164, "epoch": 102, "lr": 9.974786494169671e-05}
{"train_loss": 0.8472521305084229, "global_step": 9165, "epoch": 102, "lr": 9.974780679110809e-05}
{"train_loss": 0.7588488406009888, "global_step": 9166, "epoch": 102, "lr": 9.97477486338315e-05, "val_loss": 1.2412415742874146}
{"train_loss": 0.7928966879844666, "global_step": 9167, "epoch": 103, "lr": 9.974769046986692e-05}
{"train_loss": 0.7352400422096252, "global_step": 9168, "epoch": 103, "lr": 9.974763229921435e-05}
{"train_loss": 0.8099196553230286, "global_step": 9169, "epoch": 103, "lr": 9.974757412187385e-05}
{"train_loss": 0.8377107977867126, "global_step": 9170, "epoch": 103, "lr": 9.974751593784538e-05}
{"train_loss": 0.7358755469322205, "global_step": 9171, "epoch": 103, "lr": 9.974745774712896e-05}
{"train_loss": 0.9575309157371521, "global_step": 9172, "epoch": 103, "lr": 9.974739954972461e-05}
{"train_loss": 0.7382897138595581, "global_step": 9173, "epoch": 103, "lr": 9.974734134563232e-05}
{"train_loss": 0.7596762776374817, "global_step": 9174, "epoch": 103, "lr": 9.97472831348521e-05}
{"train_loss": 0.5486905574798584, "global_step": 9175, "epoch": 103, "lr": 9.974722491738397e-05}
{"train_loss": 0.7891123294830322, "global_step": 9176, "epoch": 103, "lr": 9.974716669322794e-05}
{"train_loss": 0.700141191482544, "global_step": 9177, "epoch": 103, "lr": 9.9747108462384e-05}
{"train_loss": 0.9121816754341125, "global_step": 9178, "epoch": 103, "lr": 9.974705022485218e-05}
{"train_loss": 0.7131421566009521, "global_step": 9179, "epoch": 103, "lr": 9.974699198063246e-05}
{"train_loss": 0.8377890586853027, "global_step": 9180, "epoch": 103, "lr": 9.974693372972488e-05}
{"train_loss": 0.7542441487312317, "global_step": 9181, "epoch": 103, "lr": 9.974687547212942e-05}
{"train_loss": 0.7814072966575623, "global_step": 9182, "epoch": 103, "lr": 9.97468172078461e-05}
{"train_loss": 0.6340256929397583, "global_step": 9183, "epoch": 103, "lr": 9.974675893687492e-05}
{"train_loss": 0.5981475710868835, "global_step": 9184, "epoch": 103, "lr": 9.974670065921591e-05}
{"train_loss": 0.7101290822029114, "global_step": 9185, "epoch": 103, "lr": 9.974664237486905e-05}
{"train_loss": 0.7303776741027832, "global_step": 9186, "epoch": 103, "lr": 9.974658408383438e-05}
{"train_loss": 0.5986788868904114, "global_step": 9187, "epoch": 103, "lr": 9.974652578611185e-05}
{"train_loss": 0.6369693279266357, "global_step": 9188, "epoch": 103, "lr": 9.974646748170153e-05}
{"train_loss": 1.0192643404006958, "global_step": 9189, "epoch": 103, "lr": 9.974640917060341e-05}
{"train_loss": 0.7934049367904663, "global_step": 9190, "epoch": 103, "lr": 9.974635085281748e-05}
{"train_loss": 0.7086640000343323, "global_step": 9191, "epoch": 103, "lr": 9.974629252834377e-05}
{"train_loss": 0.6694658994674683, "global_step": 9192, "epoch": 103, "lr": 9.974623419718228e-05}
{"train_loss": 0.7176520228385925, "global_step": 9193, "epoch": 103, "lr": 9.974617585933299e-05}
{"train_loss": 0.7055462598800659, "global_step": 9194, "epoch": 103, "lr": 9.974611751479595e-05}
{"train_loss": 0.6926688551902771, "global_step": 9195, "epoch": 103, "lr": 9.974605916357115e-05}
{"train_loss": 0.7704421281814575, "global_step": 9196, "epoch": 103, "lr": 9.97460008056586e-05}
{"train_loss": 0.6495282053947449, "global_step": 9197, "epoch": 103, "lr": 9.97459424410583e-05}
{"train_loss": 0.7988186478614807, "global_step": 9198, "epoch": 103, "lr": 9.974588406977027e-05}
{"train_loss": 0.8608171939849854, "global_step": 9199, "epoch": 103, "lr": 9.974582569179452e-05}
{"train_loss": 0.9134226441383362, "global_step": 9200, "epoch": 103, "lr": 9.974576730713103e-05}
{"train_loss": 0.7962114214897156, "global_step": 9201, "epoch": 103, "lr": 9.974570891577985e-05}
{"train_loss": 0.9122638702392578, "global_step": 9202, "epoch": 103, "lr": 9.974565051774095e-05}
{"train_loss": 0.7450912594795227, "global_step": 9203, "epoch": 103, "lr": 9.974559211301437e-05}
{"train_loss": 0.8487562537193298, "global_step": 9204, "epoch": 103, "lr": 9.974553370160009e-05}
{"train_loss": 0.6576268672943115, "global_step": 9205, "epoch": 103, "lr": 9.974547528349812e-05}
{"train_loss": 0.6889363527297974, "global_step": 9206, "epoch": 103, "lr": 9.974541685870848e-05}
{"train_loss": 0.8388654589653015, "global_step": 9207, "epoch": 103, "lr": 9.974535842723119e-05}
{"train_loss": 0.7058367133140564, "global_step": 9208, "epoch": 103, "lr": 9.974529998906624e-05}
{"train_loss": 0.6559940576553345, "global_step": 9209, "epoch": 103, "lr": 9.974524154421363e-05}
{"train_loss": 0.7277181148529053, "global_step": 9210, "epoch": 103, "lr": 9.974518309267338e-05}
{"train_loss": 0.6513205766677856, "global_step": 9211, "epoch": 103, "lr": 9.974512463444549e-05}
{"train_loss": 0.6819593906402588, "global_step": 9212, "epoch": 103, "lr": 9.974506616953e-05}
{"train_loss": 0.7333036661148071, "global_step": 9213, "epoch": 103, "lr": 9.974500769792688e-05}
{"train_loss": 0.7378968000411987, "global_step": 9214, "epoch": 103, "lr": 9.974494921963615e-05}
{"train_loss": 0.7272614240646362, "global_step": 9215, "epoch": 103, "lr": 9.97448907346578e-05}
{"train_loss": 0.6111462116241455, "global_step": 9216, "epoch": 103, "lr": 9.974483224299187e-05}
{"train_loss": 0.8244284391403198, "global_step": 9217, "epoch": 103, "lr": 9.974477374463836e-05}
{"train_loss": 0.7651469111442566, "global_step": 9218, "epoch": 103, "lr": 9.974471523959727e-05}
{"train_loss": 0.7518308162689209, "global_step": 9219, "epoch": 103, "lr": 9.97446567278686e-05}
{"train_loss": 0.7524930238723755, "global_step": 9220, "epoch": 103, "lr": 9.974459820945236e-05}
{"train_loss": 0.9114086627960205, "global_step": 9221, "epoch": 103, "lr": 9.974453968434859e-05}
{"train_loss": 0.8520517349243164, "global_step": 9222, "epoch": 103, "lr": 9.974448115255726e-05}
{"train_loss": 0.7876025438308716, "global_step": 9223, "epoch": 103, "lr": 9.974442261407839e-05}
{"train_loss": 0.7932214140892029, "global_step": 9224, "epoch": 103, "lr": 9.9744364068912e-05}
{"train_loss": 0.7350847125053406, "global_step": 9225, "epoch": 103, "lr": 9.974430551705806e-05}
{"train_loss": 0.7533000707626343, "global_step": 9226, "epoch": 103, "lr": 9.974424695851662e-05}
{"train_loss": 0.7605016231536865, "global_step": 9227, "epoch": 103, "lr": 9.974418839328768e-05}
{"train_loss": 0.7074521780014038, "global_step": 9228, "epoch": 103, "lr": 9.974412982137124e-05}
{"train_loss": 0.773689866065979, "global_step": 9229, "epoch": 103, "lr": 9.97440712427673e-05}
{"train_loss": 0.7594484090805054, "global_step": 9230, "epoch": 103, "lr": 9.974401265747587e-05}
{"train_loss": 0.87935471534729, "global_step": 9231, "epoch": 103, "lr": 9.974395406549697e-05}
{"train_loss": 0.709503173828125, "global_step": 9232, "epoch": 103, "lr": 9.97438954668306e-05}
{"train_loss": 0.665783166885376, "global_step": 9233, "epoch": 103, "lr": 9.974383686147677e-05}
{"train_loss": 0.7300326824188232, "global_step": 9234, "epoch": 103, "lr": 9.97437782494355e-05}
{"train_loss": 0.8453303575515747, "global_step": 9235, "epoch": 103, "lr": 9.974371963070676e-05}
{"train_loss": 0.796151876449585, "global_step": 9236, "epoch": 103, "lr": 9.974366100529061e-05}
{"train_loss": 0.7332680225372314, "global_step": 9237, "epoch": 103, "lr": 9.974360237318701e-05}
{"train_loss": 1.0305169820785522, "global_step": 9238, "epoch": 103, "lr": 9.9743543734396e-05}
{"train_loss": 0.8176382780075073, "global_step": 9239, "epoch": 103, "lr": 9.974348508891757e-05}
{"train_loss": 0.6443009972572327, "global_step": 9240, "epoch": 103, "lr": 9.974342643675175e-05}
{"train_loss": 0.7296972274780273, "global_step": 9241, "epoch": 103, "lr": 9.974336777789852e-05}
{"train_loss": 0.8034533858299255, "global_step": 9242, "epoch": 103, "lr": 9.974330911235791e-05}
{"train_loss": 0.6443054676055908, "global_step": 9243, "epoch": 103, "lr": 9.974325044012991e-05}
{"train_loss": 0.7724744081497192, "global_step": 9244, "epoch": 103, "lr": 9.974319176121454e-05}
{"train_loss": 0.7349589467048645, "global_step": 9245, "epoch": 103, "lr": 9.974313307561179e-05}
{"train_loss": 0.6813514232635498, "global_step": 9246, "epoch": 103, "lr": 9.97430743833217e-05}
{"train_loss": 0.7500202655792236, "global_step": 9247, "epoch": 103, "lr": 9.974301568434426e-05}
{"train_loss": 0.6892732381820679, "global_step": 9248, "epoch": 103, "lr": 9.974295697867945e-05}
{"train_loss": 0.7193347215652466, "global_step": 9249, "epoch": 103, "lr": 9.974289826632735e-05}
{"train_loss": 0.5835453271865845, "global_step": 9250, "epoch": 103, "lr": 9.974283954728789e-05}
{"train_loss": 0.6574066877365112, "global_step": 9251, "epoch": 103, "lr": 9.974278082156112e-05}
{"train_loss": 0.6884546875953674, "global_step": 9252, "epoch": 103, "lr": 9.974272208914704e-05}
{"train_loss": 0.7941546440124512, "global_step": 9253, "epoch": 103, "lr": 9.974266335004566e-05}
{"train_loss": 0.8498679399490356, "global_step": 9254, "epoch": 103, "lr": 9.9742604604257e-05}
{"train_loss": 0.7515166080399845, "global_step": 9255, "epoch": 103, "lr": 9.974254585178102e-05, "val_loss": 1.2366273403167725}
{"train_loss": 0.801681637763977, "global_step": 9256, "epoch": 104, "lr": 9.974248709261778e-05}
{"train_loss": 0.672317624092102, "global_step": 9257, "epoch": 104, "lr": 9.974242832676726e-05}
{"train_loss": 0.7373701930046082, "global_step": 9258, "epoch": 104, "lr": 9.974236955422948e-05}
{"train_loss": 0.7883782982826233, "global_step": 9259, "epoch": 104, "lr": 9.974231077500447e-05}
{"train_loss": 1.0076717138290405, "global_step": 9260, "epoch": 104, "lr": 9.974225198909218e-05}
{"train_loss": 0.7476345300674438, "global_step": 9261, "epoch": 104, "lr": 9.974219319649267e-05}
{"train_loss": 0.7338821291923523, "global_step": 9262, "epoch": 104, "lr": 9.974213439720591e-05}
{"train_loss": 0.8033618927001953, "global_step": 9263, "epoch": 104, "lr": 9.974207559123194e-05}
{"train_loss": 0.7194282412528992, "global_step": 9264, "epoch": 104, "lr": 9.974201677857076e-05}
{"train_loss": 0.5673213601112366, "global_step": 9265, "epoch": 104, "lr": 9.974195795922236e-05}
{"train_loss": 0.7242827415466309, "global_step": 9266, "epoch": 104, "lr": 9.974189913318676e-05}
{"train_loss": 0.7222110033035278, "global_step": 9267, "epoch": 104, "lr": 9.974184030046396e-05}
{"train_loss": 0.8662652969360352, "global_step": 9268, "epoch": 104, "lr": 9.9741781461054e-05}
{"train_loss": 0.6934369802474976, "global_step": 9269, "epoch": 104, "lr": 9.974172261495685e-05}
{"train_loss": 0.6443739533424377, "global_step": 9270, "epoch": 104, "lr": 9.974166376217252e-05}
{"train_loss": 0.7129392623901367, "global_step": 9271, "epoch": 104, "lr": 9.974160490270104e-05}
{"train_loss": 0.6976574659347534, "global_step": 9272, "epoch": 104, "lr": 9.97415460365424e-05}
{"train_loss": 0.6353421211242676, "global_step": 9273, "epoch": 104, "lr": 9.974148716369664e-05}
{"train_loss": 0.793385922908783, "global_step": 9274, "epoch": 104, "lr": 9.974142828416373e-05}
{"train_loss": 0.9096145629882812, "global_step": 9275, "epoch": 104, "lr": 9.974136939794369e-05}
{"train_loss": 0.6096661686897278, "global_step": 9276, "epoch": 104, "lr": 9.974131050503652e-05}
{"train_loss": 0.7255517244338989, "global_step": 9277, "epoch": 104, "lr": 9.974125160544224e-05}
{"train_loss": 0.8622398376464844, "global_step": 9278, "epoch": 104, "lr": 9.974119269916087e-05}
{"train_loss": 0.7958278656005859, "global_step": 9279, "epoch": 104, "lr": 9.974113378619239e-05}
{"train_loss": 0.5237760543823242, "global_step": 9280, "epoch": 104, "lr": 9.974107486653682e-05}
{"train_loss": 0.6286508440971375, "global_step": 9281, "epoch": 104, "lr": 9.974101594019417e-05}
{"train_loss": 0.8040742874145508, "global_step": 9282, "epoch": 104, "lr": 9.974095700716446e-05}
{"train_loss": 0.9205202460289001, "global_step": 9283, "epoch": 104, "lr": 9.974089806744767e-05}
{"train_loss": 0.6516852378845215, "global_step": 9284, "epoch": 104, "lr": 9.974083912104381e-05}
{"train_loss": 0.7071889042854309, "global_step": 9285, "epoch": 104, "lr": 9.974078016795294e-05}
{"train_loss": 0.7481916546821594, "global_step": 9286, "epoch": 104, "lr": 9.974072120817499e-05}
{"train_loss": 0.72602379322052, "global_step": 9287, "epoch": 104, "lr": 9.974066224171003e-05}
{"train_loss": 0.7352555394172668, "global_step": 9288, "epoch": 104, "lr": 9.974060326855804e-05}
{"train_loss": 0.9564009308815002, "global_step": 9289, "epoch": 104, "lr": 9.974054428871901e-05}
{"train_loss": 0.8879691362380981, "global_step": 9290, "epoch": 104, "lr": 9.9740485302193e-05}
{"train_loss": 0.628980815410614, "global_step": 9291, "epoch": 104, "lr": 9.974042630897997e-05}
{"train_loss": 0.8699715733528137, "global_step": 9292, "epoch": 104, "lr": 9.974036730907995e-05}
{"train_loss": 0.8125690221786499, "global_step": 9293, "epoch": 104, "lr": 9.974030830249294e-05}
{"train_loss": 0.7284557819366455, "global_step": 9294, "epoch": 104, "lr": 9.974024928921896e-05}
{"train_loss": 0.7395970821380615, "global_step": 9295, "epoch": 104, "lr": 9.974019026925801e-05}
{"train_loss": 0.8023192882537842, "global_step": 9296, "epoch": 104, "lr": 9.97401312426101e-05}
{"train_loss": 0.7456974983215332, "global_step": 9297, "epoch": 104, "lr": 9.974007220927522e-05}
{"train_loss": 0.5903061628341675, "global_step": 9298, "epoch": 104, "lr": 9.974001316925341e-05}
{"train_loss": 0.6387320756912231, "global_step": 9299, "epoch": 104, "lr": 9.973995412254466e-05}
{"train_loss": 0.6867406964302063, "global_step": 9300, "epoch": 104, "lr": 9.973989506914897e-05}
{"train_loss": 0.6940183043479919, "global_step": 9301, "epoch": 104, "lr": 9.973983600906635e-05}
{"train_loss": 0.7740662693977356, "global_step": 9302, "epoch": 104, "lr": 9.973977694229684e-05}
{"train_loss": 0.7606303691864014, "global_step": 9303, "epoch": 104, "lr": 9.973971786884041e-05}
{"train_loss": 0.6299434304237366, "global_step": 9304, "epoch": 104, "lr": 9.973965878869709e-05}
{"train_loss": 0.8758705854415894, "global_step": 9305, "epoch": 104, "lr": 9.973959970186686e-05}
{"train_loss": 0.7672106623649597, "global_step": 9306, "epoch": 104, "lr": 9.973954060834975e-05}
{"train_loss": 0.7275136709213257, "global_step": 9307, "epoch": 104, "lr": 9.973948150814578e-05}
{"train_loss": 0.9016916155815125, "global_step": 9308, "epoch": 104, "lr": 9.973942240125494e-05}
{"train_loss": 0.7843550443649292, "global_step": 9309, "epoch": 104, "lr": 9.973936328767723e-05}
{"train_loss": 0.7814509272575378, "global_step": 9310, "epoch": 104, "lr": 9.973930416741268e-05}
{"train_loss": 0.6914064884185791, "global_step": 9311, "epoch": 104, "lr": 9.973924504046129e-05}
{"train_loss": 0.6022588014602661, "global_step": 9312, "epoch": 104, "lr": 9.973918590682305e-05}
{"train_loss": 0.8648740649223328, "global_step": 9313, "epoch": 104, "lr": 9.973912676649799e-05}
{"train_loss": 0.8790109157562256, "global_step": 9314, "epoch": 104, "lr": 9.97390676194861e-05}
{"train_loss": 0.7778787016868591, "global_step": 9315, "epoch": 104, "lr": 9.973900846578743e-05}
{"train_loss": 0.6529215574264526, "global_step": 9316, "epoch": 104, "lr": 9.973894930540193e-05}
{"train_loss": 0.8703782558441162, "global_step": 9317, "epoch": 104, "lr": 9.973889013832964e-05}
{"train_loss": 0.7389696836471558, "global_step": 9318, "epoch": 104, "lr": 9.973883096457055e-05}
{"train_loss": 0.6971227526664734, "global_step": 9319, "epoch": 104, "lr": 9.97387717841247e-05}
{"train_loss": 0.7846915125846863, "global_step": 9320, "epoch": 104, "lr": 9.973871259699207e-05}
{"train_loss": 0.7132034301757812, "global_step": 9321, "epoch": 104, "lr": 9.973865340317267e-05}
{"train_loss": 0.648582935333252, "global_step": 9322, "epoch": 104, "lr": 9.973859420266654e-05}
{"train_loss": 0.6350568532943726, "global_step": 9323, "epoch": 104, "lr": 9.973853499547364e-05}
{"train_loss": 0.7645718455314636, "global_step": 9324, "epoch": 104, "lr": 9.9738475781594e-05}
{"train_loss": 0.7024036645889282, "global_step": 9325, "epoch": 104, "lr": 9.973841656102763e-05}
{"train_loss": 0.6903475522994995, "global_step": 9326, "epoch": 104, "lr": 9.973835733377455e-05}
{"train_loss": 0.6656449437141418, "global_step": 9327, "epoch": 104, "lr": 9.973829809983472e-05}
{"train_loss": 0.697833240032196, "global_step": 9328, "epoch": 104, "lr": 9.973823885920821e-05}
{"train_loss": 0.6337294578552246, "global_step": 9329, "epoch": 104, "lr": 9.9738179611895e-05}
{"train_loss": 0.7589622735977173, "global_step": 9330, "epoch": 104, "lr": 9.973812035789509e-05}
{"train_loss": 1.0214011669158936, "global_step": 9331, "epoch": 104, "lr": 9.97380610972085e-05}
{"train_loss": 0.720374584197998, "global_step": 9332, "epoch": 104, "lr": 9.973800182983523e-05}
{"train_loss": 0.7834810614585876, "global_step": 9333, "epoch": 104, "lr": 9.97379425557753e-05}
{"train_loss": 0.9254392385482788, "global_step": 9334, "epoch": 104, "lr": 9.97378832750287e-05}
{"train_loss": 0.8699032664299011, "global_step": 9335, "epoch": 104, "lr": 9.973782398759545e-05}
{"train_loss": 0.677656888961792, "global_step": 9336, "epoch": 104, "lr": 9.973776469347557e-05}
{"train_loss": 0.7110028266906738, "global_step": 9337, "epoch": 104, "lr": 9.973770539266904e-05}
{"train_loss": 0.7431122660636902, "global_step": 9338, "epoch": 104, "lr": 9.973764608517589e-05}
{"train_loss": 0.820058286190033, "global_step": 9339, "epoch": 104, "lr": 9.97375867709961e-05}
{"train_loss": 0.5215296745300293, "global_step": 9340, "epoch": 104, "lr": 9.973752745012973e-05}
{"train_loss": 0.7102034687995911, "global_step": 9341, "epoch": 104, "lr": 9.973746812257671e-05}
{"train_loss": 0.6373986005783081, "global_step": 9342, "epoch": 104, "lr": 9.973740878833712e-05}
{"train_loss": 0.9374294877052307, "global_step": 9343, "epoch": 104, "lr": 9.973734944741096e-05}
{"train_loss": 0.7463159708494551, "global_step": 9344, "epoch": 104, "lr": 9.97372900997982e-05, "val_loss": 1.1887222528457642}
{"train_loss": 0.6557208895683289, "global_step": 9345, "epoch": 105, "lr": 9.973723074549885e-05}
{"train_loss": 0.7368171215057373, "global_step": 9346, "epoch": 105, "lr": 9.973717138451297e-05}
{"train_loss": 0.7831871509552002, "global_step": 9347, "epoch": 105, "lr": 9.97371120168405e-05}
{"train_loss": 0.7403658628463745, "global_step": 9348, "epoch": 105, "lr": 9.97370526424815e-05}
{"train_loss": 0.5885341167449951, "global_step": 9349, "epoch": 105, "lr": 9.973699326143597e-05}
{"train_loss": 0.7710174322128296, "global_step": 9350, "epoch": 105, "lr": 9.97369338737039e-05}
{"train_loss": 0.6603415012359619, "global_step": 9351, "epoch": 105, "lr": 9.97368744792853e-05}
{"train_loss": 0.7733030915260315, "global_step": 9352, "epoch": 105, "lr": 9.973681507818018e-05}
{"train_loss": 0.6514176726341248, "global_step": 9353, "epoch": 105, "lr": 9.973675567038856e-05}
{"train_loss": 0.8562873005867004, "global_step": 9354, "epoch": 105, "lr": 9.973669625591044e-05}
{"train_loss": 0.8247776627540588, "global_step": 9355, "epoch": 105, "lr": 9.973663683474581e-05}
{"train_loss": 0.7073912620544434, "global_step": 9356, "epoch": 105, "lr": 9.973657740689471e-05}
{"train_loss": 0.6938632726669312, "global_step": 9357, "epoch": 105, "lr": 9.973651797235714e-05}
{"train_loss": 0.7639181017875671, "global_step": 9358, "epoch": 105, "lr": 9.973645853113309e-05}
{"train_loss": 0.7173362374305725, "global_step": 9359, "epoch": 105, "lr": 9.973639908322259e-05}
{"train_loss": 0.6232401728630066, "global_step": 9360, "epoch": 105, "lr": 9.973633962862563e-05}
{"train_loss": 0.5828306078910828, "global_step": 9361, "epoch": 105, "lr": 9.973628016734222e-05}
{"train_loss": 0.624463677406311, "global_step": 9362, "epoch": 105, "lr": 9.973622069937239e-05}
{"train_loss": 0.7913185358047485, "global_step": 9363, "epoch": 105, "lr": 9.973616122471612e-05}
{"train_loss": 0.6960379481315613, "global_step": 9364, "epoch": 105, "lr": 9.973610174337342e-05}
{"train_loss": 0.6170938611030579, "global_step": 9365, "epoch": 105, "lr": 9.973604225534432e-05}
{"train_loss": 0.7046037912368774, "global_step": 9366, "epoch": 105, "lr": 9.973598276062881e-05}
{"train_loss": 0.7236781120300293, "global_step": 9367, "epoch": 105, "lr": 9.97359232592269e-05}
{"train_loss": 0.6719685196876526, "global_step": 9368, "epoch": 105, "lr": 9.973586375113861e-05}
{"train_loss": 0.7861953973770142, "global_step": 9369, "epoch": 105, "lr": 9.973580423636394e-05}
{"train_loss": 0.7135353088378906, "global_step": 9370, "epoch": 105, "lr": 9.973574471490288e-05}
{"train_loss": 0.7512753009796143, "global_step": 9371, "epoch": 105, "lr": 9.973568518675548e-05}
{"train_loss": 0.7098290324211121, "global_step": 9372, "epoch": 105, "lr": 9.973562565192172e-05}
{"train_loss": 0.8667950630187988, "global_step": 9373, "epoch": 105, "lr": 9.97355661104016e-05}
{"train_loss": 0.7974018454551697, "global_step": 9374, "epoch": 105, "lr": 9.973550656219514e-05}
{"train_loss": 0.931029200553894, "global_step": 9375, "epoch": 105, "lr": 9.973544700730235e-05}
{"train_loss": 0.6840302348136902, "global_step": 9376, "epoch": 105, "lr": 9.973538744572324e-05}
{"train_loss": 0.5754404067993164, "global_step": 9377, "epoch": 105, "lr": 9.97353278774578e-05}
{"train_loss": 0.7751233577728271, "global_step": 9378, "epoch": 105, "lr": 9.973526830250606e-05}
{"train_loss": 0.7329387664794922, "global_step": 9379, "epoch": 105, "lr": 9.973520872086803e-05}
{"train_loss": 0.8329548239707947, "global_step": 9380, "epoch": 105, "lr": 9.97351491325437e-05}
{"train_loss": 0.6948775053024292, "global_step": 9381, "epoch": 105, "lr": 9.973508953753307e-05}
{"train_loss": 0.6814916729927063, "global_step": 9382, "epoch": 105, "lr": 9.973502993583618e-05}
{"train_loss": 0.7588556408882141, "global_step": 9383, "epoch": 105, "lr": 9.973497032745301e-05}
{"train_loss": 0.6914125084877014, "global_step": 9384, "epoch": 105, "lr": 9.973491071238359e-05}
{"train_loss": 0.8173668384552002, "global_step": 9385, "epoch": 105, "lr": 9.973485109062791e-05}
{"train_loss": 0.8855841159820557, "global_step": 9386, "epoch": 105, "lr": 9.973479146218598e-05}
{"train_loss": 0.7676986455917358, "global_step": 9387, "epoch": 105, "lr": 9.973473182705783e-05}
{"train_loss": 0.7080878615379333, "global_step": 9388, "epoch": 105, "lr": 9.973467218524345e-05}
{"train_loss": 0.8172000646591187, "global_step": 9389, "epoch": 105, "lr": 9.973461253674285e-05}
{"train_loss": 0.9358647465705872, "global_step": 9390, "epoch": 105, "lr": 9.973455288155603e-05}
{"train_loss": 0.7627962827682495, "global_step": 9391, "epoch": 105, "lr": 9.9734493219683e-05}
{"train_loss": 0.6697858572006226, "global_step": 9392, "epoch": 105, "lr": 9.973443355112377e-05}
{"train_loss": 0.8988366723060608, "global_step": 9393, "epoch": 105, "lr": 9.973437387587837e-05}
{"train_loss": 0.6190803050994873, "global_step": 9394, "epoch": 105, "lr": 9.973431419394678e-05}
{"train_loss": 0.737575888633728, "global_step": 9395, "epoch": 105, "lr": 9.973425450532903e-05}
{"train_loss": 0.7495759725570679, "global_step": 9396, "epoch": 105, "lr": 9.97341948100251e-05}
{"train_loss": 0.8731861710548401, "global_step": 9397, "epoch": 105, "lr": 9.973413510803502e-05}
{"train_loss": 0.6904801726341248, "global_step": 9398, "epoch": 105, "lr": 9.97340753993588e-05}
{"train_loss": 0.5771431922912598, "global_step": 9399, "epoch": 105, "lr": 9.973401568399642e-05}
{"train_loss": 0.7914473414421082, "global_step": 9400, "epoch": 105, "lr": 9.973395596194792e-05}
{"train_loss": 0.8096588850021362, "global_step": 9401, "epoch": 105, "lr": 9.973389623321328e-05}
{"train_loss": 0.8072212934494019, "global_step": 9402, "epoch": 105, "lr": 9.973383649779254e-05}
{"train_loss": 0.857840895652771, "global_step": 9403, "epoch": 105, "lr": 9.97337767556857e-05}
{"train_loss": 0.9136724472045898, "global_step": 9404, "epoch": 105, "lr": 9.973371700689274e-05}
{"train_loss": 0.8329289555549622, "global_step": 9405, "epoch": 105, "lr": 9.973365725141371e-05}
{"train_loss": 0.7267922759056091, "global_step": 9406, "epoch": 105, "lr": 9.973359748924858e-05}
{"train_loss": 0.6925008296966553, "global_step": 9407, "epoch": 105, "lr": 9.973353772039737e-05}
{"train_loss": 0.7917126417160034, "global_step": 9408, "epoch": 105, "lr": 9.97334779448601e-05}
{"train_loss": 0.6304593682289124, "global_step": 9409, "epoch": 105, "lr": 9.973341816263678e-05}
{"train_loss": 0.8249409794807434, "global_step": 9410, "epoch": 105, "lr": 9.97333583737274e-05}
{"train_loss": 0.7876794338226318, "global_step": 9411, "epoch": 105, "lr": 9.973329857813196e-05}
{"train_loss": 0.7240881323814392, "global_step": 9412, "epoch": 105, "lr": 9.973323877585051e-05}
{"train_loss": 0.9528392553329468, "global_step": 9413, "epoch": 105, "lr": 9.973317896688301e-05}
{"train_loss": 0.833486795425415, "global_step": 9414, "epoch": 105, "lr": 9.97331191512295e-05}
{"train_loss": 0.8780480027198792, "global_step": 9415, "epoch": 105, "lr": 9.973305932888996e-05}
{"train_loss": 0.7307153940200806, "global_step": 9416, "epoch": 105, "lr": 9.973299949986445e-05}
{"train_loss": 0.6587140560150146, "global_step": 9417, "epoch": 105, "lr": 9.973293966415294e-05}
{"train_loss": 0.6231592893600464, "global_step": 9418, "epoch": 105, "lr": 9.973287982175542e-05}
{"train_loss": 0.7674822807312012, "global_step": 9419, "epoch": 105, "lr": 9.973281997267193e-05}
{"train_loss": 0.7657127380371094, "global_step": 9420, "epoch": 105, "lr": 9.973276011690247e-05}
{"train_loss": 0.7636241316795349, "global_step": 9421, "epoch": 105, "lr": 9.973270025444705e-05}
{"train_loss": 0.5799490213394165, "global_step": 9422, "epoch": 105, "lr": 9.973264038530567e-05}
{"train_loss": 0.9334799647331238, "global_step": 9423, "epoch": 105, "lr": 9.973258050947835e-05}
{"train_loss": 0.671737551689148, "global_step": 9424, "epoch": 105, "lr": 9.973252062696508e-05}
{"train_loss": 0.6536065936088562, "global_step": 9425, "epoch": 105, "lr": 9.973246073776588e-05}
{"train_loss": 0.8427087664604187, "global_step": 9426, "epoch": 105, "lr": 9.973240084188075e-05}
{"train_loss": 0.7203818559646606, "global_step": 9427, "epoch": 105, "lr": 9.973234093930972e-05}
{"train_loss": 0.6264499425888062, "global_step": 9428, "epoch": 105, "lr": 9.973228103005279e-05}
{"train_loss": 0.746941089630127, "global_step": 9429, "epoch": 105, "lr": 9.973222111410995e-05}
{"train_loss": 0.7569198608398438, "global_step": 9430, "epoch": 105, "lr": 9.973216119148121e-05}
{"train_loss": 0.7306835055351257, "global_step": 9431, "epoch": 105, "lr": 9.97321012621666e-05}
{"train_loss": 0.6893320083618164, "global_step": 9432, "epoch": 105, "lr": 9.973204132616611e-05}
{"train_loss": 0.7461165015617114, "global_step": 9433, "epoch": 105, "lr": 9.973198138347974e-05, "val_loss": 1.1773940324783325, "train_action_mse_error": 56.11376190185547}
{"train_loss": 0.7380841970443726, "global_step": 9434, "epoch": 106, "lr": 9.973192143410755e-05}
{"train_loss": 0.7218285202980042, "global_step": 9435, "epoch": 106, "lr": 9.973186147804947e-05}
{"train_loss": 0.5904785990715027, "global_step": 9436, "epoch": 106, "lr": 9.973180151530557e-05}
{"train_loss": 0.7553192377090454, "global_step": 9437, "epoch": 106, "lr": 9.973174154587581e-05}
{"train_loss": 0.7433048486709595, "global_step": 9438, "epoch": 106, "lr": 9.973168156976025e-05}
{"train_loss": 0.6365424394607544, "global_step": 9439, "epoch": 106, "lr": 9.973162158695886e-05}
{"train_loss": 0.7358654737472534, "global_step": 9440, "epoch": 106, "lr": 9.973156159747166e-05}
{"train_loss": 0.8869142532348633, "global_step": 9441, "epoch": 106, "lr": 9.973150160129866e-05}
{"train_loss": 0.794107973575592, "global_step": 9442, "epoch": 106, "lr": 9.973144159843987e-05}
{"train_loss": 0.6335565447807312, "global_step": 9443, "epoch": 106, "lr": 9.973138158889527e-05}
{"train_loss": 0.7305074334144592, "global_step": 9444, "epoch": 106, "lr": 9.973132157266492e-05}
{"train_loss": 0.5784733295440674, "global_step": 9445, "epoch": 106, "lr": 9.97312615497488e-05}
{"train_loss": 0.7968448996543884, "global_step": 9446, "epoch": 106, "lr": 9.97312015201469e-05}
{"train_loss": 0.9360175132751465, "global_step": 9447, "epoch": 106, "lr": 9.973114148385925e-05}
{"train_loss": 0.7404326796531677, "global_step": 9448, "epoch": 106, "lr": 9.973108144088588e-05}
{"train_loss": 0.783967137336731, "global_step": 9449, "epoch": 106, "lr": 9.973102139122675e-05}
{"train_loss": 0.6418036222457886, "global_step": 9450, "epoch": 106, "lr": 9.97309613348819e-05}
{"train_loss": 0.6792634725570679, "global_step": 9451, "epoch": 106, "lr": 9.97309012718513e-05}
{"train_loss": 0.689069390296936, "global_step": 9452, "epoch": 106, "lr": 9.9730841202135e-05}
{"train_loss": 0.7197975516319275, "global_step": 9453, "epoch": 106, "lr": 9.973078112573301e-05}
{"train_loss": 0.6721243858337402, "global_step": 9454, "epoch": 106, "lr": 9.973072104264532e-05}
{"train_loss": 0.6057291030883789, "global_step": 9455, "epoch": 106, "lr": 9.973066095287194e-05}
{"train_loss": 0.8534965515136719, "global_step": 9456, "epoch": 106, "lr": 9.973060085641286e-05}
{"train_loss": 0.7567197680473328, "global_step": 9457, "epoch": 106, "lr": 9.973054075326811e-05}
{"train_loss": 0.6489207148551941, "global_step": 9458, "epoch": 106, "lr": 9.973048064343771e-05}
{"train_loss": 0.6340941190719604, "global_step": 9459, "epoch": 106, "lr": 9.973042052692166e-05}
{"train_loss": 0.7033209204673767, "global_step": 9460, "epoch": 106, "lr": 9.973036040371995e-05}
{"train_loss": 0.6208481788635254, "global_step": 9461, "epoch": 106, "lr": 9.97303002738326e-05}
{"train_loss": 0.8583243489265442, "global_step": 9462, "epoch": 106, "lr": 9.97302401372596e-05}
{"train_loss": 0.5887590646743774, "global_step": 9463, "epoch": 106, "lr": 9.9730179994001e-05}
{"train_loss": 0.8208743333816528, "global_step": 9464, "epoch": 106, "lr": 9.973011984405677e-05}
{"train_loss": 0.7929783463478088, "global_step": 9465, "epoch": 106, "lr": 9.973005968742694e-05}
{"train_loss": 0.7488071322441101, "global_step": 9466, "epoch": 106, "lr": 9.97299995241115e-05}
{"train_loss": 0.7991713881492615, "global_step": 9467, "epoch": 106, "lr": 9.972993935411048e-05}
{"train_loss": 0.8214295506477356, "global_step": 9468, "epoch": 106, "lr": 9.972987917742386e-05}
{"train_loss": 0.7397105097770691, "global_step": 9469, "epoch": 106, "lr": 9.972981899405167e-05}
{"train_loss": 0.7166056632995605, "global_step": 9470, "epoch": 106, "lr": 9.972975880399392e-05}
{"train_loss": 0.636178195476532, "global_step": 9471, "epoch": 106, "lr": 9.972969860725061e-05}
{"train_loss": 0.7384015321731567, "global_step": 9472, "epoch": 106, "lr": 9.972963840382173e-05}
{"train_loss": 0.7668180465698242, "global_step": 9473, "epoch": 106, "lr": 9.972957819370732e-05}
{"train_loss": 0.7691279053688049, "global_step": 9474, "epoch": 106, "lr": 9.972951797690737e-05}
{"train_loss": 0.5964038968086243, "global_step": 9475, "epoch": 106, "lr": 9.972945775342191e-05}
{"train_loss": 0.6399396061897278, "global_step": 9476, "epoch": 106, "lr": 9.972939752325091e-05}
{"train_loss": 0.8034833669662476, "global_step": 9477, "epoch": 106, "lr": 9.972933728639439e-05}
{"train_loss": 0.5815971493721008, "global_step": 9478, "epoch": 106, "lr": 9.97292770428524e-05}
{"train_loss": 0.7509936094284058, "global_step": 9479, "epoch": 106, "lr": 9.972921679262489e-05}
{"train_loss": 0.6924718618392944, "global_step": 9480, "epoch": 106, "lr": 9.97291565357119e-05}
{"train_loss": 0.8639678359031677, "global_step": 9481, "epoch": 106, "lr": 9.972909627211343e-05}
{"train_loss": 0.7511577010154724, "global_step": 9482, "epoch": 106, "lr": 9.97290360018295e-05}
{"train_loss": 0.7213531732559204, "global_step": 9483, "epoch": 106, "lr": 9.972897572486008e-05}
{"train_loss": 0.7957433462142944, "global_step": 9484, "epoch": 106, "lr": 9.972891544120523e-05}
{"train_loss": 0.6728695631027222, "global_step": 9485, "epoch": 106, "lr": 9.972885515086492e-05}
{"train_loss": 0.7277263402938843, "global_step": 9486, "epoch": 106, "lr": 9.972879485383918e-05}
{"train_loss": 0.7702794075012207, "global_step": 9487, "epoch": 106, "lr": 9.972873455012802e-05}
{"train_loss": 0.7815440893173218, "global_step": 9488, "epoch": 106, "lr": 9.97286742397314e-05}
{"train_loss": 0.7800386548042297, "global_step": 9489, "epoch": 106, "lr": 9.97286139226494e-05}
{"train_loss": 0.8707214593887329, "global_step": 9490, "epoch": 106, "lr": 9.972855359888198e-05}
{"train_loss": 0.7636964917182922, "global_step": 9491, "epoch": 106, "lr": 9.972849326842917e-05}
{"train_loss": 0.6252830624580383, "global_step": 9492, "epoch": 106, "lr": 9.972843293129097e-05}
{"train_loss": 0.6284875869750977, "global_step": 9493, "epoch": 106, "lr": 9.97283725874674e-05}
{"train_loss": 0.9578273296356201, "global_step": 9494, "epoch": 106, "lr": 9.972831223695842e-05}
{"train_loss": 0.7503279447555542, "global_step": 9495, "epoch": 106, "lr": 9.972825187976411e-05}
{"train_loss": 0.6072105169296265, "global_step": 9496, "epoch": 106, "lr": 9.972819151588441e-05}
{"train_loss": 0.7611355781555176, "global_step": 9497, "epoch": 106, "lr": 9.972813114531938e-05}
{"train_loss": 0.9050959944725037, "global_step": 9498, "epoch": 106, "lr": 9.972807076806902e-05}
{"train_loss": 0.7484039664268494, "global_step": 9499, "epoch": 106, "lr": 9.972801038413331e-05}
{"train_loss": 0.6531344056129456, "global_step": 9500, "epoch": 106, "lr": 9.972794999351227e-05}
{"train_loss": 0.8401080965995789, "global_step": 9501, "epoch": 106, "lr": 9.972788959620593e-05}
{"train_loss": 0.8235284090042114, "global_step": 9502, "epoch": 106, "lr": 9.972782919221428e-05}
{"train_loss": 0.9409921169281006, "global_step": 9503, "epoch": 106, "lr": 9.972776878153732e-05}
{"train_loss": 0.9399257898330688, "global_step": 9504, "epoch": 106, "lr": 9.972770836417506e-05}
{"train_loss": 0.7899118661880493, "global_step": 9505, "epoch": 106, "lr": 9.972764794012754e-05}
{"train_loss": 0.8684893250465393, "global_step": 9506, "epoch": 106, "lr": 9.972758750939472e-05}
{"train_loss": 0.7976962924003601, "global_step": 9507, "epoch": 106, "lr": 9.972752707197666e-05}
{"train_loss": 0.7692926526069641, "global_step": 9508, "epoch": 106, "lr": 9.97274666278733e-05}
{"train_loss": 0.74997478723526, "global_step": 9509, "epoch": 106, "lr": 9.972740617708473e-05}
{"train_loss": 0.7614314556121826, "global_step": 9510, "epoch": 106, "lr": 9.972734571961089e-05}
{"train_loss": 0.9485805034637451, "global_step": 9511, "epoch": 106, "lr": 9.972728525545183e-05}
{"train_loss": 0.8096770644187927, "global_step": 9512, "epoch": 106, "lr": 9.972722478460753e-05}
{"train_loss": 0.7387449145317078, "global_step": 9513, "epoch": 106, "lr": 9.972716430707802e-05}
{"train_loss": 0.7921072840690613, "global_step": 9514, "epoch": 106, "lr": 9.97271038228633e-05}
{"train_loss": 0.677899956703186, "global_step": 9515, "epoch": 106, "lr": 9.972704333196336e-05}
{"train_loss": 0.8941766023635864, "global_step": 9516, "epoch": 106, "lr": 9.972698283437824e-05}
{"train_loss": 0.7778234481811523, "global_step": 9517, "epoch": 106, "lr": 9.972692233010793e-05}
{"train_loss": 0.7519793510437012, "global_step": 9518, "epoch": 106, "lr": 9.972686181915243e-05}
{"train_loss": 0.6569065451622009, "global_step": 9519, "epoch": 106, "lr": 9.972680130151177e-05}
{"train_loss": 0.6297833323478699, "global_step": 9520, "epoch": 106, "lr": 9.972674077718596e-05}
{"train_loss": 0.7973285913467407, "global_step": 9521, "epoch": 106, "lr": 9.972668024617498e-05}
{"train_loss": 0.7467143589191223, "global_step": 9522, "epoch": 106, "lr": 9.972661970847886e-05, "val_loss": 1.24604332447052}
{"train_loss": 0.7736082077026367, "global_step": 9523, "epoch": 107, "lr": 9.972655916409759e-05}
{"train_loss": 0.7483425140380859, "global_step": 9524, "epoch": 107, "lr": 9.972649861303121e-05}
{"train_loss": 0.8354153037071228, "global_step": 9525, "epoch": 107, "lr": 9.97264380552797e-05}
{"train_loss": 0.6942264437675476, "global_step": 9526, "epoch": 107, "lr": 9.972637749084307e-05}
{"train_loss": 0.8652708530426025, "global_step": 9527, "epoch": 107, "lr": 9.972631691972134e-05}
{"train_loss": 0.5531935095787048, "global_step": 9528, "epoch": 107, "lr": 9.972625634191452e-05}
{"train_loss": 0.6517207622528076, "global_step": 9529, "epoch": 107, "lr": 9.972619575742259e-05}
{"train_loss": 0.7628195285797119, "global_step": 9530, "epoch": 107, "lr": 9.97261351662456e-05}
{"train_loss": 0.7738266587257385, "global_step": 9531, "epoch": 107, "lr": 9.972607456838353e-05}
{"train_loss": 0.8849381804466248, "global_step": 9532, "epoch": 107, "lr": 9.97260139638364e-05}
{"train_loss": 0.6646697521209717, "global_step": 9533, "epoch": 107, "lr": 9.97259533526042e-05}
{"train_loss": 0.6734510064125061, "global_step": 9534, "epoch": 107, "lr": 9.972589273468695e-05}
{"train_loss": 0.7205794453620911, "global_step": 9535, "epoch": 107, "lr": 9.972583211008468e-05}
{"train_loss": 0.8902096748352051, "global_step": 9536, "epoch": 107, "lr": 9.972577147879737e-05}
{"train_loss": 0.8916418552398682, "global_step": 9537, "epoch": 107, "lr": 9.972571084082503e-05}
{"train_loss": 0.5578166246414185, "global_step": 9538, "epoch": 107, "lr": 9.972565019616768e-05}
{"train_loss": 0.9229581356048584, "global_step": 9539, "epoch": 107, "lr": 9.972558954482531e-05}
{"train_loss": 0.6463354229927063, "global_step": 9540, "epoch": 107, "lr": 9.972552888679796e-05}
{"train_loss": 0.7791656255722046, "global_step": 9541, "epoch": 107, "lr": 9.972546822208561e-05}
{"train_loss": 0.8337510228157043, "global_step": 9542, "epoch": 107, "lr": 9.972540755068829e-05}
{"train_loss": 0.7188507914543152, "global_step": 9543, "epoch": 107, "lr": 9.972534687260597e-05}
{"train_loss": 0.6312288045883179, "global_step": 9544, "epoch": 107, "lr": 9.97252861878387e-05}
{"train_loss": 0.7829737067222595, "global_step": 9545, "epoch": 107, "lr": 9.972522549638645e-05}
{"train_loss": 0.7738051414489746, "global_step": 9546, "epoch": 107, "lr": 9.972516479824928e-05}
{"train_loss": 0.7130953073501587, "global_step": 9547, "epoch": 107, "lr": 9.972510409342714e-05}
{"train_loss": 0.9242753386497498, "global_step": 9548, "epoch": 107, "lr": 9.97250433819201e-05}
{"train_loss": 0.8539167046546936, "global_step": 9549, "epoch": 107, "lr": 9.972498266372811e-05}
{"train_loss": 0.577028214931488, "global_step": 9550, "epoch": 107, "lr": 9.972492193885119e-05}
{"train_loss": 0.7015630006790161, "global_step": 9551, "epoch": 107, "lr": 9.972486120728938e-05}
{"train_loss": 0.7330722212791443, "global_step": 9552, "epoch": 107, "lr": 9.972480046904266e-05}
{"train_loss": 0.8394820690155029, "global_step": 9553, "epoch": 107, "lr": 9.972473972411106e-05}
{"train_loss": 0.648115873336792, "global_step": 9554, "epoch": 107, "lr": 9.972467897249456e-05}
{"train_loss": 0.6435327529907227, "global_step": 9555, "epoch": 107, "lr": 9.972461821419319e-05}
{"train_loss": 0.5803501605987549, "global_step": 9556, "epoch": 107, "lr": 9.972455744920695e-05}
{"train_loss": 0.7605927586555481, "global_step": 9557, "epoch": 107, "lr": 9.972449667753585e-05}
{"train_loss": 0.9176130890846252, "global_step": 9558, "epoch": 107, "lr": 9.97244358991799e-05}
{"train_loss": 0.7645454406738281, "global_step": 9559, "epoch": 107, "lr": 9.97243751141391e-05}
{"train_loss": 0.7183783054351807, "global_step": 9560, "epoch": 107, "lr": 9.972431432241347e-05}
{"train_loss": 0.7273819446563721, "global_step": 9561, "epoch": 107, "lr": 9.9724253524003e-05}
{"train_loss": 0.525634765625, "global_step": 9562, "epoch": 107, "lr": 9.972419271890772e-05}
{"train_loss": 0.8331316709518433, "global_step": 9563, "epoch": 107, "lr": 9.972413190712763e-05}
{"train_loss": 0.6383592486381531, "global_step": 9564, "epoch": 107, "lr": 9.972407108866273e-05}
{"train_loss": 0.6748700141906738, "global_step": 9565, "epoch": 107, "lr": 9.972401026351303e-05}
{"train_loss": 0.6536434888839722, "global_step": 9566, "epoch": 107, "lr": 9.972394943167855e-05}
{"train_loss": 0.8003826141357422, "global_step": 9567, "epoch": 107, "lr": 9.97238885931593e-05}
{"train_loss": 0.6157581806182861, "global_step": 9568, "epoch": 107, "lr": 9.972382774795527e-05}
{"train_loss": 0.7168949842453003, "global_step": 9569, "epoch": 107, "lr": 9.972376689606649e-05}
{"train_loss": 0.8843610286712646, "global_step": 9570, "epoch": 107, "lr": 9.972370603749292e-05}
{"train_loss": 0.7108604907989502, "global_step": 9571, "epoch": 107, "lr": 9.972364517223464e-05}
{"train_loss": 0.6645169258117676, "global_step": 9572, "epoch": 107, "lr": 9.97235843002916e-05}
{"train_loss": 0.8755069375038147, "global_step": 9573, "epoch": 107, "lr": 9.972352342166385e-05}
{"train_loss": 0.7602512240409851, "global_step": 9574, "epoch": 107, "lr": 9.972346253635136e-05}
{"train_loss": 0.7852765917778015, "global_step": 9575, "epoch": 107, "lr": 9.972340164435416e-05}
{"train_loss": 0.9361627101898193, "global_step": 9576, "epoch": 107, "lr": 9.972334074567227e-05}
{"train_loss": 0.6247238516807556, "global_step": 9577, "epoch": 107, "lr": 9.972327984030567e-05}
{"train_loss": 0.6376395225524902, "global_step": 9578, "epoch": 107, "lr": 9.972321892825438e-05}
{"train_loss": 0.8204399347305298, "global_step": 9579, "epoch": 107, "lr": 9.972315800951841e-05}
{"train_loss": 0.6954355239868164, "global_step": 9580, "epoch": 107, "lr": 9.972309708409777e-05}
{"train_loss": 0.589169442653656, "global_step": 9581, "epoch": 107, "lr": 9.972303615199247e-05}
{"train_loss": 0.6103290915489197, "global_step": 9582, "epoch": 107, "lr": 9.97229752132025e-05}
{"train_loss": 0.7525038123130798, "global_step": 9583, "epoch": 107, "lr": 9.972291426772791e-05}
{"train_loss": 0.7176302075386047, "global_step": 9584, "epoch": 107, "lr": 9.972285331556866e-05}
{"train_loss": 0.7243504524230957, "global_step": 9585, "epoch": 107, "lr": 9.972279235672477e-05}
{"train_loss": 0.7704947590827942, "global_step": 9586, "epoch": 107, "lr": 9.972273139119627e-05}
{"train_loss": 0.7269992828369141, "global_step": 9587, "epoch": 107, "lr": 9.972267041898315e-05}
{"train_loss": 0.6461289525032043, "global_step": 9588, "epoch": 107, "lr": 9.972260944008542e-05}
{"train_loss": 0.6495304107666016, "global_step": 9589, "epoch": 107, "lr": 9.97225484545031e-05}
{"train_loss": 0.7778523564338684, "global_step": 9590, "epoch": 107, "lr": 9.972248746223618e-05}
{"train_loss": 0.7693417072296143, "global_step": 9591, "epoch": 107, "lr": 9.97224264632847e-05}
{"train_loss": 0.6550241708755493, "global_step": 9592, "epoch": 107, "lr": 9.972236545764862e-05}
{"train_loss": 0.7532181739807129, "global_step": 9593, "epoch": 107, "lr": 9.972230444532798e-05}
{"train_loss": 0.7537167072296143, "global_step": 9594, "epoch": 107, "lr": 9.972224342632278e-05}
{"train_loss": 0.5736099481582642, "global_step": 9595, "epoch": 107, "lr": 9.972218240063304e-05}
{"train_loss": 0.6624915599822998, "global_step": 9596, "epoch": 107, "lr": 9.972212136825876e-05}
{"train_loss": 0.6610613465309143, "global_step": 9597, "epoch": 107, "lr": 9.972206032919993e-05}
{"train_loss": 0.8486303091049194, "global_step": 9598, "epoch": 107, "lr": 9.972199928345658e-05}
{"train_loss": 0.8462666869163513, "global_step": 9599, "epoch": 107, "lr": 9.972193823102872e-05}
{"train_loss": 0.9304100871086121, "global_step": 9600, "epoch": 107, "lr": 9.972187717191635e-05}
{"train_loss": 0.682779848575592, "global_step": 9601, "epoch": 107, "lr": 9.972181610611948e-05}
{"train_loss": 0.7227542400360107, "global_step": 9602, "epoch": 107, "lr": 9.972175503363813e-05}
{"train_loss": 0.8165243268013, "global_step": 9603, "epoch": 107, "lr": 9.972169395447227e-05}
{"train_loss": 1.0297871828079224, "global_step": 9604, "epoch": 107, "lr": 9.972163286862195e-05}
{"train_loss": 0.5717042684555054, "global_step": 9605, "epoch": 107, "lr": 9.972157177608716e-05}
{"train_loss": 0.6646637320518494, "global_step": 9606, "epoch": 107, "lr": 9.972151067686792e-05}
{"train_loss": 0.6193200945854187, "global_step": 9607, "epoch": 107, "lr": 9.972144957096421e-05}
{"train_loss": 0.9375966191291809, "global_step": 9608, "epoch": 107, "lr": 9.972138845837606e-05}
{"train_loss": 0.7643832564353943, "global_step": 9609, "epoch": 107, "lr": 9.972132733910348e-05}
{"train_loss": 0.702275276184082, "global_step": 9610, "epoch": 107, "lr": 9.972126621314649e-05}
{"train_loss": 0.7397360406564862, "global_step": 9611, "epoch": 107, "lr": 9.972120508050507e-05, "val_loss": 1.174152135848999}
{"train_loss": 0.7275655269622803, "global_step": 9612, "epoch": 108, "lr": 9.972114394117924e-05}
{"train_loss": 0.7366265058517456, "global_step": 9613, "epoch": 108, "lr": 9.9721082795169e-05}
{"train_loss": 0.8351495862007141, "global_step": 9614, "epoch": 108, "lr": 9.972102164247437e-05}
{"train_loss": 0.8506720066070557, "global_step": 9615, "epoch": 108, "lr": 9.972096048309536e-05}
{"train_loss": 0.6834989786148071, "global_step": 9616, "epoch": 108, "lr": 9.972089931703196e-05}
{"train_loss": 0.5995776057243347, "global_step": 9617, "epoch": 108, "lr": 9.972083814428422e-05}
{"train_loss": 0.7048618197441101, "global_step": 9618, "epoch": 108, "lr": 9.972077696485209e-05}
{"train_loss": 0.8085030913352966, "global_step": 9619, "epoch": 108, "lr": 9.972071577873561e-05}
{"train_loss": 0.8120272755622864, "global_step": 9620, "epoch": 108, "lr": 9.97206545859348e-05}
{"train_loss": 0.6568664312362671, "global_step": 9621, "epoch": 108, "lr": 9.972059338644965e-05}
{"train_loss": 0.7925043106079102, "global_step": 9622, "epoch": 108, "lr": 9.972053218028016e-05}
{"train_loss": 0.6267378926277161, "global_step": 9623, "epoch": 108, "lr": 9.972047096742635e-05}
{"train_loss": 0.7921587824821472, "global_step": 9624, "epoch": 108, "lr": 9.972040974788825e-05}
{"train_loss": 0.6494389772415161, "global_step": 9625, "epoch": 108, "lr": 9.972034852166585e-05}
{"train_loss": 0.5663082599639893, "global_step": 9626, "epoch": 108, "lr": 9.972028728875913e-05}
{"train_loss": 0.6571913957595825, "global_step": 9627, "epoch": 108, "lr": 9.972022604916813e-05}
{"train_loss": 0.7420685291290283, "global_step": 9628, "epoch": 108, "lr": 9.972016480289286e-05}
{"train_loss": 0.8641957640647888, "global_step": 9629, "epoch": 108, "lr": 9.972010354993332e-05}
{"train_loss": 0.6738594770431519, "global_step": 9630, "epoch": 108, "lr": 9.97200422902895e-05}
{"train_loss": 0.6429161429405212, "global_step": 9631, "epoch": 108, "lr": 9.971998102396145e-05}
{"train_loss": 0.6381344199180603, "global_step": 9632, "epoch": 108, "lr": 9.971991975094915e-05}
{"train_loss": 0.735814094543457, "global_step": 9633, "epoch": 108, "lr": 9.97198584712526e-05}
{"train_loss": 0.849271833896637, "global_step": 9634, "epoch": 108, "lr": 9.971979718487184e-05}
{"train_loss": 0.7295944094657898, "global_step": 9635, "epoch": 108, "lr": 9.971973589180684e-05}
{"train_loss": 0.747563898563385, "global_step": 9636, "epoch": 108, "lr": 9.971967459205764e-05}
{"train_loss": 0.6779494285583496, "global_step": 9637, "epoch": 108, "lr": 9.971961328562423e-05}
{"train_loss": 0.663909375667572, "global_step": 9638, "epoch": 108, "lr": 9.971955197250661e-05}
{"train_loss": 0.7555633187294006, "global_step": 9639, "epoch": 108, "lr": 9.971949065270482e-05}
{"train_loss": 0.7700103521347046, "global_step": 9640, "epoch": 108, "lr": 9.971942932621885e-05}
{"train_loss": 0.5704967379570007, "global_step": 9641, "epoch": 108, "lr": 9.971936799304871e-05}
{"train_loss": 0.8015424013137817, "global_step": 9642, "epoch": 108, "lr": 9.97193066531944e-05}
{"train_loss": 0.7509651184082031, "global_step": 9643, "epoch": 108, "lr": 9.971924530665595e-05}
{"train_loss": 0.8013366460800171, "global_step": 9644, "epoch": 108, "lr": 9.971918395343334e-05}
{"train_loss": 0.7262418866157532, "global_step": 9645, "epoch": 108, "lr": 9.97191225935266e-05}
{"train_loss": 0.5166780948638916, "global_step": 9646, "epoch": 108, "lr": 9.971906122693571e-05}
{"train_loss": 0.733622133731842, "global_step": 9647, "epoch": 108, "lr": 9.97189998536607e-05}
{"train_loss": 0.8107359409332275, "global_step": 9648, "epoch": 108, "lr": 9.97189384737016e-05}
{"train_loss": 0.7885767221450806, "global_step": 9649, "epoch": 108, "lr": 9.971887708705839e-05}
{"train_loss": 0.7553473114967346, "global_step": 9650, "epoch": 108, "lr": 9.971881569373108e-05}
{"train_loss": 0.8743934035301208, "global_step": 9651, "epoch": 108, "lr": 9.971875429371967e-05}
{"train_loss": 0.915687084197998, "global_step": 9652, "epoch": 108, "lr": 9.971869288702419e-05}
{"train_loss": 0.845584511756897, "global_step": 9653, "epoch": 108, "lr": 9.971863147364464e-05}
{"train_loss": 0.7762044668197632, "global_step": 9654, "epoch": 108, "lr": 9.971857005358101e-05}
{"train_loss": 0.6437169313430786, "global_step": 9655, "epoch": 108, "lr": 9.971850862683334e-05}
{"train_loss": 0.8238765001296997, "global_step": 9656, "epoch": 108, "lr": 9.971844719340162e-05}
{"train_loss": 0.9797486066818237, "global_step": 9657, "epoch": 108, "lr": 9.971838575328585e-05}
{"train_loss": 1.0686522722244263, "global_step": 9658, "epoch": 108, "lr": 9.971832430648607e-05}
{"train_loss": 0.7656707763671875, "global_step": 9659, "epoch": 108, "lr": 9.971826285300226e-05}
{"train_loss": 0.8949050903320312, "global_step": 9660, "epoch": 108, "lr": 9.971820139283441e-05}
{"train_loss": 0.7644751667976379, "global_step": 9661, "epoch": 108, "lr": 9.971813992598258e-05}
{"train_loss": 0.750027060508728, "global_step": 9662, "epoch": 108, "lr": 9.971807845244674e-05}
{"train_loss": 0.6446666121482849, "global_step": 9663, "epoch": 108, "lr": 9.971801697222691e-05}
{"train_loss": 0.6988064050674438, "global_step": 9664, "epoch": 108, "lr": 9.971795548532313e-05}
{"train_loss": 0.8742787837982178, "global_step": 9665, "epoch": 108, "lr": 9.971789399173534e-05}
{"train_loss": 0.9324460029602051, "global_step": 9666, "epoch": 108, "lr": 9.971783249146359e-05}
{"train_loss": 0.7648458480834961, "global_step": 9667, "epoch": 108, "lr": 9.971777098450789e-05}
{"train_loss": 0.783760130405426, "global_step": 9668, "epoch": 108, "lr": 9.971770947086825e-05}
{"train_loss": 0.8314602971076965, "global_step": 9669, "epoch": 108, "lr": 9.971764795054465e-05}
{"train_loss": 0.8422521352767944, "global_step": 9670, "epoch": 108, "lr": 9.971758642353714e-05}
{"train_loss": 0.7080157399177551, "global_step": 9671, "epoch": 108, "lr": 9.971752488984569e-05}
{"train_loss": 0.9106300473213196, "global_step": 9672, "epoch": 108, "lr": 9.971746334947034e-05}
{"train_loss": 0.7641927003860474, "global_step": 9673, "epoch": 108, "lr": 9.971740180241107e-05}
{"train_loss": 0.7983298301696777, "global_step": 9674, "epoch": 108, "lr": 9.97173402486679e-05}
{"train_loss": 0.8082835078239441, "global_step": 9675, "epoch": 108, "lr": 9.971727868824085e-05}
{"train_loss": 0.7699902653694153, "global_step": 9676, "epoch": 108, "lr": 9.971721712112991e-05}
{"train_loss": 0.7028000950813293, "global_step": 9677, "epoch": 108, "lr": 9.97171555473351e-05}
{"train_loss": 0.7828409075737, "global_step": 9678, "epoch": 108, "lr": 9.971709396685643e-05}
{"train_loss": 0.794722318649292, "global_step": 9679, "epoch": 108, "lr": 9.971703237969389e-05}
{"train_loss": 0.7605007886886597, "global_step": 9680, "epoch": 108, "lr": 9.971697078584751e-05}
{"train_loss": 0.7394949793815613, "global_step": 9681, "epoch": 108, "lr": 9.971690918531729e-05}
{"train_loss": 0.6149945855140686, "global_step": 9682, "epoch": 108, "lr": 9.971684757810322e-05}
{"train_loss": 0.7510973811149597, "global_step": 9683, "epoch": 108, "lr": 9.971678596420535e-05}
{"train_loss": 0.8123263120651245, "global_step": 9684, "epoch": 108, "lr": 9.971672434362365e-05}
{"train_loss": 0.7917677760124207, "global_step": 9685, "epoch": 108, "lr": 9.971666271635814e-05}
{"train_loss": 0.9046758413314819, "global_step": 9686, "epoch": 108, "lr": 9.971660108240884e-05}
{"train_loss": 0.5766065716743469, "global_step": 9687, "epoch": 108, "lr": 9.971653944177575e-05}
{"train_loss": 0.6704925298690796, "global_step": 9688, "epoch": 108, "lr": 9.971647779445887e-05}
{"train_loss": 0.7413515448570251, "global_step": 9689, "epoch": 108, "lr": 9.971641614045821e-05}
{"train_loss": 0.6821544170379639, "global_step": 9690, "epoch": 108, "lr": 9.97163544797738e-05}
{"train_loss": 0.5902771353721619, "global_step": 9691, "epoch": 108, "lr": 9.971629281240562e-05}
{"train_loss": 0.8059251308441162, "global_step": 9692, "epoch": 108, "lr": 9.97162311383537e-05}
{"train_loss": 0.6726080179214478, "global_step": 9693, "epoch": 108, "lr": 9.971616945761804e-05}
{"train_loss": 0.6734948754310608, "global_step": 9694, "epoch": 108, "lr": 9.971610777019864e-05}
{"train_loss": 0.6162630319595337, "global_step": 9695, "epoch": 108, "lr": 9.97160460760955e-05}
{"train_loss": 0.634579598903656, "global_step": 9696, "epoch": 108, "lr": 9.971598437530867e-05}
{"train_loss": 0.6981747150421143, "global_step": 9697, "epoch": 108, "lr": 9.971592266783812e-05}
{"train_loss": 0.8107942342758179, "global_step": 9698, "epoch": 108, "lr": 9.971586095368386e-05}
{"train_loss": 0.7646475434303284, "global_step": 9699, "epoch": 108, "lr": 9.971579923284592e-05}
{"train_loss": 0.7504781518089637, "global_step": 9700, "epoch": 108, "lr": 9.97157375053243e-05, "val_loss": 1.230624794960022}
{"train_loss": 0.6978731751441956, "global_step": 9701, "epoch": 109, "lr": 9.971567577111901e-05}
{"train_loss": 0.6667569875717163, "global_step": 9702, "epoch": 109, "lr": 9.971561403023004e-05}
{"train_loss": 0.6478211879730225, "global_step": 9703, "epoch": 109, "lr": 9.971555228265742e-05}
{"train_loss": 0.7533718347549438, "global_step": 9704, "epoch": 109, "lr": 9.971549052840115e-05}
{"train_loss": 0.7621483206748962, "global_step": 9705, "epoch": 109, "lr": 9.971542876746122e-05}
{"train_loss": 0.7155983448028564, "global_step": 9706, "epoch": 109, "lr": 9.971536699983768e-05}
{"train_loss": 0.7763921022415161, "global_step": 9707, "epoch": 109, "lr": 9.971530522553051e-05}
{"train_loss": 0.7198296785354614, "global_step": 9708, "epoch": 109, "lr": 9.971524344453972e-05}
{"train_loss": 0.71396404504776, "global_step": 9709, "epoch": 109, "lr": 9.971518165686532e-05}
{"train_loss": 0.9415889382362366, "global_step": 9710, "epoch": 109, "lr": 9.971511986250732e-05}
{"train_loss": 0.6039004325866699, "global_step": 9711, "epoch": 109, "lr": 9.971505806146573e-05}
{"train_loss": 0.5941845178604126, "global_step": 9712, "epoch": 109, "lr": 9.971499625374054e-05}
{"train_loss": 0.6008557081222534, "global_step": 9713, "epoch": 109, "lr": 9.97149344393318e-05}
{"train_loss": 0.8600823283195496, "global_step": 9714, "epoch": 109, "lr": 9.971487261823948e-05}
{"train_loss": 0.6440341472625732, "global_step": 9715, "epoch": 109, "lr": 9.97148107904636e-05}
{"train_loss": 0.8495532274246216, "global_step": 9716, "epoch": 109, "lr": 9.971474895600418e-05}
{"train_loss": 0.766274094581604, "global_step": 9717, "epoch": 109, "lr": 9.97146871148612e-05}
{"train_loss": 0.6690800786018372, "global_step": 9718, "epoch": 109, "lr": 9.971462526703471e-05}
{"train_loss": 0.7152740955352783, "global_step": 9719, "epoch": 109, "lr": 9.971456341252467e-05}
{"train_loss": 0.7082452178001404, "global_step": 9720, "epoch": 109, "lr": 9.971450155133113e-05}
{"train_loss": 0.6850701570510864, "global_step": 9721, "epoch": 109, "lr": 9.971443968345409e-05}
{"train_loss": 0.7126362919807434, "global_step": 9722, "epoch": 109, "lr": 9.971437780889352e-05}
{"train_loss": 0.6307697892189026, "global_step": 9723, "epoch": 109, "lr": 9.971431592764948e-05}
{"train_loss": 0.7560903429985046, "global_step": 9724, "epoch": 109, "lr": 9.971425403972196e-05}
{"train_loss": 0.7861103415489197, "global_step": 9725, "epoch": 109, "lr": 9.971419214511095e-05}
{"train_loss": 0.736424446105957, "global_step": 9726, "epoch": 109, "lr": 9.971413024381648e-05}
{"train_loss": 0.628858208656311, "global_step": 9727, "epoch": 109, "lr": 9.971406833583854e-05}
{"train_loss": 0.7078566551208496, "global_step": 9728, "epoch": 109, "lr": 9.971400642117716e-05}
{"train_loss": 0.7453471422195435, "global_step": 9729, "epoch": 109, "lr": 9.971394449983233e-05}
{"train_loss": 0.7655170559883118, "global_step": 9730, "epoch": 109, "lr": 9.971388257180408e-05}
{"train_loss": 0.6596859693527222, "global_step": 9731, "epoch": 109, "lr": 9.971382063709239e-05}
{"train_loss": 0.7453368306159973, "global_step": 9732, "epoch": 109, "lr": 9.971375869569729e-05}
{"train_loss": 0.693162202835083, "global_step": 9733, "epoch": 109, "lr": 9.971369674761877e-05}
{"train_loss": 0.9355465173721313, "global_step": 9734, "epoch": 109, "lr": 9.971363479285687e-05}
{"train_loss": 0.6941007375717163, "global_step": 9735, "epoch": 109, "lr": 9.971357283141156e-05}
{"train_loss": 0.7244340181350708, "global_step": 9736, "epoch": 109, "lr": 9.971351086328288e-05}
{"train_loss": 0.6305810809135437, "global_step": 9737, "epoch": 109, "lr": 9.971344888847081e-05}
{"train_loss": 0.9546991586685181, "global_step": 9738, "epoch": 109, "lr": 9.971338690697538e-05}
{"train_loss": 0.5821853280067444, "global_step": 9739, "epoch": 109, "lr": 9.971332491879659e-05}
{"train_loss": 0.7892987132072449, "global_step": 9740, "epoch": 109, "lr": 9.971326292393445e-05}
{"train_loss": 0.6934971213340759, "global_step": 9741, "epoch": 109, "lr": 9.971320092238898e-05}
{"train_loss": 0.6642425060272217, "global_step": 9742, "epoch": 109, "lr": 9.971313891416016e-05}
{"train_loss": 0.8367621898651123, "global_step": 9743, "epoch": 109, "lr": 9.971307689924802e-05}
{"train_loss": 0.7325013279914856, "global_step": 9744, "epoch": 109, "lr": 9.971301487765256e-05}
{"train_loss": 0.6437545418739319, "global_step": 9745, "epoch": 109, "lr": 9.971295284937378e-05}
{"train_loss": 0.8100913166999817, "global_step": 9746, "epoch": 109, "lr": 9.971289081441173e-05}
{"train_loss": 0.6917974948883057, "global_step": 9747, "epoch": 109, "lr": 9.971282877276637e-05}
{"train_loss": 0.6006253957748413, "global_step": 9748, "epoch": 109, "lr": 9.971276672443772e-05}
{"train_loss": 0.5500056147575378, "global_step": 9749, "epoch": 109, "lr": 9.97127046694258e-05}
{"train_loss": 0.6505011916160583, "global_step": 9750, "epoch": 109, "lr": 9.971264260773061e-05}
{"train_loss": 0.8204852342605591, "global_step": 9751, "epoch": 109, "lr": 9.971258053935217e-05}
{"train_loss": 0.7740062475204468, "global_step": 9752, "epoch": 109, "lr": 9.971251846429047e-05}
{"train_loss": 0.7278736233711243, "global_step": 9753, "epoch": 109, "lr": 9.971245638254553e-05}
{"train_loss": 0.7906237840652466, "global_step": 9754, "epoch": 109, "lr": 9.971239429411736e-05}
{"train_loss": 0.7191835045814514, "global_step": 9755, "epoch": 109, "lr": 9.971233219900596e-05}
{"train_loss": 0.7234088778495789, "global_step": 9756, "epoch": 109, "lr": 9.971227009721134e-05}
{"train_loss": 0.8125308156013489, "global_step": 9757, "epoch": 109, "lr": 9.971220798873353e-05}
{"train_loss": 0.7405878901481628, "global_step": 9758, "epoch": 109, "lr": 9.97121458735725e-05}
{"train_loss": 0.7843435406684875, "global_step": 9759, "epoch": 109, "lr": 9.971208375172829e-05}
{"train_loss": 0.7215471863746643, "global_step": 9760, "epoch": 109, "lr": 9.971202162320088e-05}
{"train_loss": 0.7039779424667358, "global_step": 9761, "epoch": 109, "lr": 9.97119594879903e-05}
{"train_loss": 0.7333189845085144, "global_step": 9762, "epoch": 109, "lr": 9.971189734609657e-05}
{"train_loss": 0.7612161040306091, "global_step": 9763, "epoch": 109, "lr": 9.971183519751966e-05}
{"train_loss": 0.6062595248222351, "global_step": 9764, "epoch": 109, "lr": 9.971177304225961e-05}
{"train_loss": 0.7312155365943909, "global_step": 9765, "epoch": 109, "lr": 9.971171088031641e-05}
{"train_loss": 0.8326273560523987, "global_step": 9766, "epoch": 109, "lr": 9.971164871169009e-05}
{"train_loss": 0.8484187722206116, "global_step": 9767, "epoch": 109, "lr": 9.971158653638064e-05}
{"train_loss": 0.6800068616867065, "global_step": 9768, "epoch": 109, "lr": 9.971152435438807e-05}
{"train_loss": 0.7861682772636414, "global_step": 9769, "epoch": 109, "lr": 9.971146216571239e-05}
{"train_loss": 0.8023374080657959, "global_step": 9770, "epoch": 109, "lr": 9.971139997035361e-05}
{"train_loss": 0.7684608101844788, "global_step": 9771, "epoch": 109, "lr": 9.971133776831174e-05}
{"train_loss": 0.8471623659133911, "global_step": 9772, "epoch": 109, "lr": 9.97112755595868e-05}
{"train_loss": 0.8583094477653503, "global_step": 9773, "epoch": 109, "lr": 9.971121334417876e-05}
{"train_loss": 0.7188681364059448, "global_step": 9774, "epoch": 109, "lr": 9.971115112208767e-05}
{"train_loss": 0.7973803877830505, "global_step": 9775, "epoch": 109, "lr": 9.971108889331351e-05}
{"train_loss": 0.7040740251541138, "global_step": 9776, "epoch": 109, "lr": 9.97110266578563e-05}
{"train_loss": 0.6676574349403381, "global_step": 9777, "epoch": 109, "lr": 9.971096441571606e-05}
{"train_loss": 0.6749887466430664, "global_step": 9778, "epoch": 109, "lr": 9.971090216689278e-05}
{"train_loss": 0.863296627998352, "global_step": 9779, "epoch": 109, "lr": 9.971083991138647e-05}
{"train_loss": 0.7986517548561096, "global_step": 9780, "epoch": 109, "lr": 9.971077764919714e-05}
{"train_loss": 0.7259838581085205, "global_step": 9781, "epoch": 109, "lr": 9.971071538032482e-05}
{"train_loss": 0.6576364636421204, "global_step": 9782, "epoch": 109, "lr": 9.97106531047695e-05}
{"train_loss": 0.6485544443130493, "global_step": 9783, "epoch": 109, "lr": 9.971059082253118e-05}
{"train_loss": 0.7802407145500183, "global_step": 9784, "epoch": 109, "lr": 9.971052853360987e-05}
{"train_loss": 0.7395759224891663, "global_step": 9785, "epoch": 109, "lr": 9.971046623800559e-05}
{"train_loss": 0.616933286190033, "global_step": 9786, "epoch": 109, "lr": 9.971040393571835e-05}
{"train_loss": 0.7750723958015442, "global_step": 9787, "epoch": 109, "lr": 9.971034162674815e-05}
{"train_loss": 0.7242876887321472, "global_step": 9788, "epoch": 109, "lr": 9.9710279311095e-05}
{"train_loss": 0.7318035190025073, "global_step": 9789, "epoch": 109, "lr": 9.97102169887589e-05, "val_loss": 1.3110474348068237}
{"train_loss": 0.9267386794090271, "global_step": 9790, "epoch": 110, "lr": 9.971015465973988e-05}
{"train_loss": 0.8519977331161499, "global_step": 9791, "epoch": 110, "lr": 9.971009232403795e-05}
{"train_loss": 0.847225546836853, "global_step": 9792, "epoch": 110, "lr": 9.971002998165306e-05}
{"train_loss": 0.7664556503295898, "global_step": 9793, "epoch": 110, "lr": 9.97099676325853e-05}
{"train_loss": 0.7697708606719971, "global_step": 9794, "epoch": 110, "lr": 9.970990527683462e-05}
{"train_loss": 0.743864893913269, "global_step": 9795, "epoch": 110, "lr": 9.970984291440107e-05}
{"train_loss": 0.7425926327705383, "global_step": 9796, "epoch": 110, "lr": 9.970978054528462e-05}
{"train_loss": 0.9186096787452698, "global_step": 9797, "epoch": 110, "lr": 9.970971816948528e-05}
{"train_loss": 0.7596790790557861, "global_step": 9798, "epoch": 110, "lr": 9.970965578700311e-05}
{"train_loss": 0.6509431600570679, "global_step": 9799, "epoch": 110, "lr": 9.970959339783806e-05}
{"train_loss": 0.8304794430732727, "global_step": 9800, "epoch": 110, "lr": 9.970953100199016e-05}
{"train_loss": 0.7679327130317688, "global_step": 9801, "epoch": 110, "lr": 9.970946859945944e-05}
{"train_loss": 0.7586851716041565, "global_step": 9802, "epoch": 110, "lr": 9.970940619024587e-05}
{"train_loss": 0.7589507699012756, "global_step": 9803, "epoch": 110, "lr": 9.970934377434949e-05}
{"train_loss": 0.7720659375190735, "global_step": 9804, "epoch": 110, "lr": 9.97092813517703e-05}
{"train_loss": 0.7225226759910583, "global_step": 9805, "epoch": 110, "lr": 9.970921892250826e-05}
{"train_loss": 0.8296070098876953, "global_step": 9806, "epoch": 110, "lr": 9.970915648656346e-05}
{"train_loss": 0.7561488747596741, "global_step": 9807, "epoch": 110, "lr": 9.970909404393586e-05}
{"train_loss": 0.8351537585258484, "global_step": 9808, "epoch": 110, "lr": 9.970903159462548e-05}
{"train_loss": 0.691827654838562, "global_step": 9809, "epoch": 110, "lr": 9.970896913863233e-05}
{"train_loss": 0.7521531581878662, "global_step": 9810, "epoch": 110, "lr": 9.97089066759564e-05}
{"train_loss": 0.7979857325553894, "global_step": 9811, "epoch": 110, "lr": 9.970884420659772e-05}
{"train_loss": 0.6215683817863464, "global_step": 9812, "epoch": 110, "lr": 9.97087817305563e-05}
{"train_loss": 0.8583499193191528, "global_step": 9813, "epoch": 110, "lr": 9.970871924783214e-05}
{"train_loss": 0.6807006001472473, "global_step": 9814, "epoch": 110, "lr": 9.970865675842525e-05}
{"train_loss": 0.7319308519363403, "global_step": 9815, "epoch": 110, "lr": 9.970859426233562e-05}
{"train_loss": 0.7112829089164734, "global_step": 9816, "epoch": 110, "lr": 9.970853175956329e-05}
{"train_loss": 0.5694389343261719, "global_step": 9817, "epoch": 110, "lr": 9.970846925010825e-05}
{"train_loss": 0.7547621130943298, "global_step": 9818, "epoch": 110, "lr": 9.970840673397051e-05}
{"train_loss": 0.7697789669036865, "global_step": 9819, "epoch": 110, "lr": 9.970834421115009e-05}
{"train_loss": 0.6576651334762573, "global_step": 9820, "epoch": 110, "lr": 9.970828168164698e-05}
{"train_loss": 0.8627172708511353, "global_step": 9821, "epoch": 110, "lr": 9.97082191454612e-05}
{"train_loss": 0.7075251340866089, "global_step": 9822, "epoch": 110, "lr": 9.970815660259275e-05}
{"train_loss": 0.7902253866195679, "global_step": 9823, "epoch": 110, "lr": 9.970809405304164e-05}
{"train_loss": 0.7331746816635132, "global_step": 9824, "epoch": 110, "lr": 9.97080314968079e-05}
{"train_loss": 0.8070210814476013, "global_step": 9825, "epoch": 110, "lr": 9.970796893389152e-05}
{"train_loss": 0.8153671622276306, "global_step": 9826, "epoch": 110, "lr": 9.97079063642925e-05}
{"train_loss": 0.928377091884613, "global_step": 9827, "epoch": 110, "lr": 9.970784378801086e-05}
{"train_loss": 0.7619631290435791, "global_step": 9828, "epoch": 110, "lr": 9.97077812050466e-05}
{"train_loss": 0.707908570766449, "global_step": 9829, "epoch": 110, "lr": 9.970771861539974e-05}
{"train_loss": 0.6775995492935181, "global_step": 9830, "epoch": 110, "lr": 9.970765601907028e-05}
{"train_loss": 0.6000155210494995, "global_step": 9831, "epoch": 110, "lr": 9.970759341605825e-05}
{"train_loss": 0.7118297219276428, "global_step": 9832, "epoch": 110, "lr": 9.970753080636361e-05}
{"train_loss": 0.78954017162323, "global_step": 9833, "epoch": 110, "lr": 9.970746818998641e-05}
{"train_loss": 0.7694699764251709, "global_step": 9834, "epoch": 110, "lr": 9.970740556692665e-05}
{"train_loss": 0.7185452580451965, "global_step": 9835, "epoch": 110, "lr": 9.970734293718435e-05}
{"train_loss": 0.7343701720237732, "global_step": 9836, "epoch": 110, "lr": 9.970728030075949e-05}
{"train_loss": 0.9777649641036987, "global_step": 9837, "epoch": 110, "lr": 9.970721765765209e-05}
{"train_loss": 0.6818578839302063, "global_step": 9838, "epoch": 110, "lr": 9.970715500786216e-05}
{"train_loss": 0.6996440887451172, "global_step": 9839, "epoch": 110, "lr": 9.97070923513897e-05}
{"train_loss": 0.6439694762229919, "global_step": 9840, "epoch": 110, "lr": 9.970702968823474e-05}
{"train_loss": 0.74622642993927, "global_step": 9841, "epoch": 110, "lr": 9.970696701839728e-05}
{"train_loss": 0.7517881989479065, "global_step": 9842, "epoch": 110, "lr": 9.970690434187731e-05}
{"train_loss": 0.7688054442405701, "global_step": 9843, "epoch": 110, "lr": 9.970684165867487e-05}
{"train_loss": 0.6825171113014221, "global_step": 9844, "epoch": 110, "lr": 9.970677896878994e-05}
{"train_loss": 0.8956683278083801, "global_step": 9845, "epoch": 110, "lr": 9.970671627222255e-05}
{"train_loss": 0.6882826685905457, "global_step": 9846, "epoch": 110, "lr": 9.970665356897267e-05}
{"train_loss": 0.8761643171310425, "global_step": 9847, "epoch": 110, "lr": 9.970659085904036e-05}
{"train_loss": 0.8198990225791931, "global_step": 9848, "epoch": 110, "lr": 9.970652814242561e-05}
{"train_loss": 0.7637659907341003, "global_step": 9849, "epoch": 110, "lr": 9.97064654191284e-05}
{"train_loss": 0.6495034098625183, "global_step": 9850, "epoch": 110, "lr": 9.970640268914877e-05}
{"train_loss": 0.5759755373001099, "global_step": 9851, "epoch": 110, "lr": 9.970633995248673e-05}
{"train_loss": 0.5837852954864502, "global_step": 9852, "epoch": 110, "lr": 9.970627720914228e-05}
{"train_loss": 0.6764265298843384, "global_step": 9853, "epoch": 110, "lr": 9.970621445911541e-05}
{"train_loss": 0.7183302044868469, "global_step": 9854, "epoch": 110, "lr": 9.970615170240615e-05}
{"train_loss": 0.862365186214447, "global_step": 9855, "epoch": 110, "lr": 9.970608893901452e-05}
{"train_loss": 0.7998949885368347, "global_step": 9856, "epoch": 110, "lr": 9.970602616894049e-05}
{"train_loss": 0.5373063087463379, "global_step": 9857, "epoch": 110, "lr": 9.97059633921841e-05}
{"train_loss": 0.6942299604415894, "global_step": 9858, "epoch": 110, "lr": 9.970590060874536e-05}
{"train_loss": 0.7087174654006958, "global_step": 9859, "epoch": 110, "lr": 9.970583781862426e-05}
{"train_loss": 0.6478757262229919, "global_step": 9860, "epoch": 110, "lr": 9.970577502182082e-05}
{"train_loss": 0.7106464505195618, "global_step": 9861, "epoch": 110, "lr": 9.970571221833502e-05}
{"train_loss": 0.5967711806297302, "global_step": 9862, "epoch": 110, "lr": 9.970564940816691e-05}
{"train_loss": 0.5945055484771729, "global_step": 9863, "epoch": 110, "lr": 9.970558659131648e-05}
{"train_loss": 0.7069590091705322, "global_step": 9864, "epoch": 110, "lr": 9.970552376778375e-05}
{"train_loss": 0.6233261823654175, "global_step": 9865, "epoch": 110, "lr": 9.970546093756871e-05}
{"train_loss": 0.7946815490722656, "global_step": 9866, "epoch": 110, "lr": 9.970539810067137e-05}
{"train_loss": 0.5908987522125244, "global_step": 9867, "epoch": 110, "lr": 9.970533525709176e-05}
{"train_loss": 0.8665294051170349, "global_step": 9868, "epoch": 110, "lr": 9.970527240682986e-05}
{"train_loss": 0.7406064867973328, "global_step": 9869, "epoch": 110, "lr": 9.97052095498857e-05}
{"train_loss": 0.8121672868728638, "global_step": 9870, "epoch": 110, "lr": 9.970514668625927e-05}
{"train_loss": 0.8208171725273132, "global_step": 9871, "epoch": 110, "lr": 9.970508381595061e-05}
{"train_loss": 0.7010173797607422, "global_step": 9872, "epoch": 110, "lr": 9.970502093895968e-05}
{"train_loss": 0.5875230431556702, "global_step": 9873, "epoch": 110, "lr": 9.970495805528653e-05}
{"train_loss": 0.5160632729530334, "global_step": 9874, "epoch": 110, "lr": 9.970489516493116e-05}
{"train_loss": 0.758901834487915, "global_step": 9875, "epoch": 110, "lr": 9.970483226789357e-05}
{"train_loss": 0.7828373312950134, "global_step": 9876, "epoch": 110, "lr": 9.970476936417376e-05}
{"train_loss": 0.6060820817947388, "global_step": 9877, "epoch": 110, "lr": 9.970470645377176e-05}
{"train_loss": 0.7398671360498064, "global_step": 9878, "epoch": 110, "lr": 9.970464353668755e-05, "val_loss": 1.2118875980377197, "train_action_mse_error": 44.46106719970703}
{"train_loss": 0.6876353621482849, "global_step": 9879, "epoch": 111, "lr": 9.970458061292118e-05}
{"train_loss": 0.8110758066177368, "global_step": 9880, "epoch": 111, "lr": 9.970451768247263e-05}
{"train_loss": 0.6150354743003845, "global_step": 9881, "epoch": 111, "lr": 9.97044547453419e-05}
{"train_loss": 0.7436053156852722, "global_step": 9882, "epoch": 111, "lr": 9.970439180152904e-05}
{"train_loss": 0.7412320971488953, "global_step": 9883, "epoch": 111, "lr": 9.9704328851034e-05}
{"train_loss": 0.7762879133224487, "global_step": 9884, "epoch": 111, "lr": 9.970426589385684e-05}
{"train_loss": 0.6434614062309265, "global_step": 9885, "epoch": 111, "lr": 9.970420292999753e-05}
{"train_loss": 0.6178128123283386, "global_step": 9886, "epoch": 111, "lr": 9.970413995945611e-05}
{"train_loss": 0.7066665291786194, "global_step": 9887, "epoch": 111, "lr": 9.970407698223257e-05}
{"train_loss": 0.7394373416900635, "global_step": 9888, "epoch": 111, "lr": 9.970401399832692e-05}
{"train_loss": 0.7271844744682312, "global_step": 9889, "epoch": 111, "lr": 9.970395100773917e-05}
{"train_loss": 0.6108648777008057, "global_step": 9890, "epoch": 111, "lr": 9.970388801046932e-05}
{"train_loss": 0.669491171836853, "global_step": 9891, "epoch": 111, "lr": 9.97038250065174e-05}
{"train_loss": 0.7258151769638062, "global_step": 9892, "epoch": 111, "lr": 9.970376199588341e-05}
{"train_loss": 0.5752382278442383, "global_step": 9893, "epoch": 111, "lr": 9.970369897856735e-05}
{"train_loss": 0.8956703543663025, "global_step": 9894, "epoch": 111, "lr": 9.970363595456924e-05}
{"train_loss": 0.6851691603660583, "global_step": 9895, "epoch": 111, "lr": 9.970357292388908e-05}
{"train_loss": 0.7219550013542175, "global_step": 9896, "epoch": 111, "lr": 9.970350988652688e-05}
{"train_loss": 0.7557924389839172, "global_step": 9897, "epoch": 111, "lr": 9.970344684248264e-05}
{"train_loss": 0.6477043628692627, "global_step": 9898, "epoch": 111, "lr": 9.970338379175638e-05}
{"train_loss": 0.5586749315261841, "global_step": 9899, "epoch": 111, "lr": 9.970332073434811e-05}
{"train_loss": 0.7449600696563721, "global_step": 9900, "epoch": 111, "lr": 9.970325767025783e-05}
{"train_loss": 0.7174956798553467, "global_step": 9901, "epoch": 111, "lr": 9.970319459948556e-05}
{"train_loss": 0.5782923698425293, "global_step": 9902, "epoch": 111, "lr": 9.970313152203131e-05}
{"train_loss": 0.7695528864860535, "global_step": 9903, "epoch": 111, "lr": 9.970306843789506e-05}
{"train_loss": 0.6718147397041321, "global_step": 9904, "epoch": 111, "lr": 9.970300534707684e-05}
{"train_loss": 0.8110307455062866, "global_step": 9905, "epoch": 111, "lr": 9.970294224957666e-05}
{"train_loss": 0.7928022742271423, "global_step": 9906, "epoch": 111, "lr": 9.970287914539453e-05}
{"train_loss": 0.7331805229187012, "global_step": 9907, "epoch": 111, "lr": 9.970281603453045e-05}
{"train_loss": 0.8764877319335938, "global_step": 9908, "epoch": 111, "lr": 9.970275291698443e-05}
{"train_loss": 0.6861328482627869, "global_step": 9909, "epoch": 111, "lr": 9.970268979275649e-05}
{"train_loss": 0.6257928013801575, "global_step": 9910, "epoch": 111, "lr": 9.970262666184662e-05}
{"train_loss": 0.6382391452789307, "global_step": 9911, "epoch": 111, "lr": 9.970256352425485e-05}
{"train_loss": 0.8128636479377747, "global_step": 9912, "epoch": 111, "lr": 9.970250037998116e-05}
{"train_loss": 0.7275102138519287, "global_step": 9913, "epoch": 111, "lr": 9.970243722902557e-05}
{"train_loss": 0.7028968930244446, "global_step": 9914, "epoch": 111, "lr": 9.970237407138811e-05}
{"train_loss": 0.7358735203742981, "global_step": 9915, "epoch": 111, "lr": 9.970231090706877e-05}
{"train_loss": 0.7338761687278748, "global_step": 9916, "epoch": 111, "lr": 9.970224773606756e-05}
{"train_loss": 0.7246038913726807, "global_step": 9917, "epoch": 111, "lr": 9.970218455838447e-05}
{"train_loss": 0.7995770573616028, "global_step": 9918, "epoch": 111, "lr": 9.970212137401955e-05}
{"train_loss": 0.6216630339622498, "global_step": 9919, "epoch": 111, "lr": 9.970205818297278e-05}
{"train_loss": 0.750394344329834, "global_step": 9920, "epoch": 111, "lr": 9.970199498524418e-05}
{"train_loss": 0.6727249622344971, "global_step": 9921, "epoch": 111, "lr": 9.970193178083374e-05}
{"train_loss": 0.8088811635971069, "global_step": 9922, "epoch": 111, "lr": 9.970186856974148e-05}
{"train_loss": 0.7592953443527222, "global_step": 9923, "epoch": 111, "lr": 9.97018053519674e-05}
{"train_loss": 0.8259020447731018, "global_step": 9924, "epoch": 111, "lr": 9.970174212751154e-05}
{"train_loss": 0.8202099204063416, "global_step": 9925, "epoch": 111, "lr": 9.970167889637388e-05}
{"train_loss": 0.7120248675346375, "global_step": 9926, "epoch": 111, "lr": 9.970161565855444e-05}
{"train_loss": 0.7967715263366699, "global_step": 9927, "epoch": 111, "lr": 9.970155241405321e-05}
{"train_loss": 0.7170580625534058, "global_step": 9928, "epoch": 111, "lr": 9.970148916287023e-05}
{"train_loss": 0.8285805583000183, "global_step": 9929, "epoch": 111, "lr": 9.970142590500548e-05}
{"train_loss": 0.82843416929245, "global_step": 9930, "epoch": 111, "lr": 9.970136264045898e-05}
{"train_loss": 0.8467368483543396, "global_step": 9931, "epoch": 111, "lr": 9.970129936923074e-05}
{"train_loss": 0.9275835156440735, "global_step": 9932, "epoch": 111, "lr": 9.970123609132076e-05}
{"train_loss": 0.8299888372421265, "global_step": 9933, "epoch": 111, "lr": 9.970117280672906e-05}
{"train_loss": 0.864219069480896, "global_step": 9934, "epoch": 111, "lr": 9.970110951545564e-05}
{"train_loss": 0.7095632553100586, "global_step": 9935, "epoch": 111, "lr": 9.97010462175005e-05}
{"train_loss": 0.7602008581161499, "global_step": 9936, "epoch": 111, "lr": 9.970098291286367e-05}
{"train_loss": 0.7132493853569031, "global_step": 9937, "epoch": 111, "lr": 9.970091960154515e-05}
{"train_loss": 0.8073113560676575, "global_step": 9938, "epoch": 111, "lr": 9.970085628354495e-05}
{"train_loss": 0.829262912273407, "global_step": 9939, "epoch": 111, "lr": 9.970079295886308e-05}
{"train_loss": 0.7395824193954468, "global_step": 9940, "epoch": 111, "lr": 9.970072962749954e-05}
{"train_loss": 0.792195737361908, "global_step": 9941, "epoch": 111, "lr": 9.970066628945434e-05}
{"train_loss": 0.803480327129364, "global_step": 9942, "epoch": 111, "lr": 9.97006029447275e-05}
{"train_loss": 0.6892542839050293, "global_step": 9943, "epoch": 111, "lr": 9.9700539593319e-05}
{"train_loss": 0.834223747253418, "global_step": 9944, "epoch": 111, "lr": 9.970047623522888e-05}
{"train_loss": 0.6945521831512451, "global_step": 9945, "epoch": 111, "lr": 9.970041287045714e-05}
{"train_loss": 0.7191009521484375, "global_step": 9946, "epoch": 111, "lr": 9.970034949900378e-05}
{"train_loss": 0.7846716046333313, "global_step": 9947, "epoch": 111, "lr": 9.970028612086882e-05}
{"train_loss": 0.7546666860580444, "global_step": 9948, "epoch": 111, "lr": 9.970022273605226e-05}
{"train_loss": 0.7808260321617126, "global_step": 9949, "epoch": 111, "lr": 9.97001593445541e-05}
{"train_loss": 0.8111061453819275, "global_step": 9950, "epoch": 111, "lr": 9.970009594637438e-05}
{"train_loss": 0.7202187776565552, "global_step": 9951, "epoch": 111, "lr": 9.970003254151306e-05}
{"train_loss": 0.7173555493354797, "global_step": 9952, "epoch": 111, "lr": 9.96999691299702e-05}
{"train_loss": 0.7215673923492432, "global_step": 9953, "epoch": 111, "lr": 9.969990571174576e-05}
{"train_loss": 0.7114041447639465, "global_step": 9954, "epoch": 111, "lr": 9.969984228683981e-05}
{"train_loss": 0.80156010389328, "global_step": 9955, "epoch": 111, "lr": 9.96997788552523e-05}
{"train_loss": 0.7765634655952454, "global_step": 9956, "epoch": 111, "lr": 9.969971541698326e-05}
{"train_loss": 0.8493298888206482, "global_step": 9957, "epoch": 111, "lr": 9.96996519720327e-05}
{"train_loss": 0.8890095353126526, "global_step": 9958, "epoch": 111, "lr": 9.969958852040063e-05}
{"train_loss": 0.8244510889053345, "global_step": 9959, "epoch": 111, "lr": 9.969952506208707e-05}
{"train_loss": 0.8883054256439209, "global_step": 9960, "epoch": 111, "lr": 9.969946159709198e-05}
{"train_loss": 0.5962270498275757, "global_step": 9961, "epoch": 111, "lr": 9.969939812541542e-05}
{"train_loss": 0.8099656701087952, "global_step": 9962, "epoch": 111, "lr": 9.96993346470574e-05}
{"train_loss": 0.9090883731842041, "global_step": 9963, "epoch": 111, "lr": 9.96992711620179e-05}
{"train_loss": 0.5868390798568726, "global_step": 9964, "epoch": 111, "lr": 9.969920767029693e-05}
{"train_loss": 0.6614071130752563, "global_step": 9965, "epoch": 111, "lr": 9.96991441718945e-05}
{"train_loss": 0.7589658498764038, "global_step": 9966, "epoch": 111, "lr": 9.969908066681064e-05}
{"train_loss": 0.7444246277380525, "global_step": 9967, "epoch": 111, "lr": 9.969901715504535e-05, "val_loss": 1.2490062713623047}
{"train_loss": 0.6758073568344116, "global_step": 9968, "epoch": 112, "lr": 9.969895363659862e-05}
{"train_loss": 0.8673143982887268, "global_step": 9969, "epoch": 112, "lr": 9.969889011147047e-05}
{"train_loss": 0.7538383603096008, "global_step": 9970, "epoch": 112, "lr": 9.969882657966093e-05}
{"train_loss": 0.7597266435623169, "global_step": 9971, "epoch": 112, "lr": 9.969876304116997e-05}
{"train_loss": 0.7683071494102478, "global_step": 9972, "epoch": 112, "lr": 9.969869949599762e-05}
{"train_loss": 0.6213163137435913, "global_step": 9973, "epoch": 112, "lr": 9.969863594414388e-05}
{"train_loss": 0.6233254075050354, "global_step": 9974, "epoch": 112, "lr": 9.969857238560878e-05}
{"train_loss": 0.7917551398277283, "global_step": 9975, "epoch": 112, "lr": 9.96985088203923e-05}
{"train_loss": 0.750779390335083, "global_step": 9976, "epoch": 112, "lr": 9.969844524849445e-05}
{"train_loss": 0.5780170559883118, "global_step": 9977, "epoch": 112, "lr": 9.969838166991527e-05}
{"train_loss": 0.6988417506217957, "global_step": 9978, "epoch": 112, "lr": 9.969831808465475e-05}
{"train_loss": 0.7535241842269897, "global_step": 9979, "epoch": 112, "lr": 9.969825449271288e-05}
{"train_loss": 0.7598545551300049, "global_step": 9980, "epoch": 112, "lr": 9.96981908940897e-05}
{"train_loss": 0.7789637446403503, "global_step": 9981, "epoch": 112, "lr": 9.96981272887852e-05}
{"train_loss": 0.7650101184844971, "global_step": 9982, "epoch": 112, "lr": 9.96980636767994e-05}
{"train_loss": 0.7192338109016418, "global_step": 9983, "epoch": 112, "lr": 9.969800005813228e-05}
{"train_loss": 0.6761312484741211, "global_step": 9984, "epoch": 112, "lr": 9.969793643278389e-05}
{"train_loss": 0.6640591025352478, "global_step": 9985, "epoch": 112, "lr": 9.969787280075421e-05}
{"train_loss": 0.5444709658622742, "global_step": 9986, "epoch": 112, "lr": 9.969780916204325e-05}
{"train_loss": 0.7640498280525208, "global_step": 9987, "epoch": 112, "lr": 9.969774551665105e-05}
{"train_loss": 0.6866337060928345, "global_step": 9988, "epoch": 112, "lr": 9.969768186457757e-05}
{"train_loss": 0.7792418003082275, "global_step": 9989, "epoch": 112, "lr": 9.969761820582284e-05}
{"train_loss": 0.68989497423172, "global_step": 9990, "epoch": 112, "lr": 9.969755454038688e-05}
{"train_loss": 0.6768046021461487, "global_step": 9991, "epoch": 112, "lr": 9.969749086826969e-05}
{"train_loss": 0.783619225025177, "global_step": 9992, "epoch": 112, "lr": 9.969742718947129e-05}
{"train_loss": 0.7033579349517822, "global_step": 9993, "epoch": 112, "lr": 9.969736350399166e-05}
{"train_loss": 0.6293490529060364, "global_step": 9994, "epoch": 112, "lr": 9.969729981183083e-05}
{"train_loss": 0.7001439332962036, "global_step": 9995, "epoch": 112, "lr": 9.96972361129888e-05}
{"train_loss": 0.6520376801490784, "global_step": 9996, "epoch": 112, "lr": 9.969717240746559e-05}
{"train_loss": 0.6201117038726807, "global_step": 9997, "epoch": 112, "lr": 9.969710869526118e-05}
{"train_loss": 0.751556932926178, "global_step": 9998, "epoch": 112, "lr": 9.969704497637561e-05}
{"train_loss": 0.9013813138008118, "global_step": 9999, "epoch": 112, "lr": 9.969698125080891e-05}
{"train_loss": 0.585612416267395, "global_step": 10000, "epoch": 112, "lr": 9.969691751856102e-05}
{"train_loss": 0.5962985754013062, "global_step": 10001, "epoch": 112, "lr": 9.9696853779632e-05}
{"train_loss": 0.7517135143280029, "global_step": 10002, "epoch": 112, "lr": 9.969679003402183e-05}
{"train_loss": 0.6799125075340271, "global_step": 10003, "epoch": 112, "lr": 9.969672628173055e-05}
{"train_loss": 0.6412997245788574, "global_step": 10004, "epoch": 112, "lr": 9.969666252275813e-05}
{"train_loss": 0.7761090397834778, "global_step": 10005, "epoch": 112, "lr": 9.969659875710462e-05}
{"train_loss": 0.6869014501571655, "global_step": 10006, "epoch": 112, "lr": 9.969653498477e-05}
{"train_loss": 0.6331490278244019, "global_step": 10007, "epoch": 112, "lr": 9.96964712057543e-05}
{"train_loss": 0.7715197801589966, "global_step": 10008, "epoch": 112, "lr": 9.969640742005749e-05}
{"train_loss": 0.8645185232162476, "global_step": 10009, "epoch": 112, "lr": 9.969634362767962e-05}
{"train_loss": 0.7380543351173401, "global_step": 10010, "epoch": 112, "lr": 9.969627982862066e-05}
{"train_loss": 0.4507911801338196, "global_step": 10011, "epoch": 112, "lr": 9.969621602288067e-05}
{"train_loss": 0.7093605995178223, "global_step": 10012, "epoch": 112, "lr": 9.969615221045962e-05}
{"train_loss": 0.8984928727149963, "global_step": 10013, "epoch": 112, "lr": 9.969608839135752e-05}
{"train_loss": 0.6231626272201538, "global_step": 10014, "epoch": 112, "lr": 9.96960245655744e-05}
{"train_loss": 0.8514918088912964, "global_step": 10015, "epoch": 112, "lr": 9.969596073311025e-05}
{"train_loss": 0.6573140621185303, "global_step": 10016, "epoch": 112, "lr": 9.969589689396508e-05}
{"train_loss": 0.7235147356987, "global_step": 10017, "epoch": 112, "lr": 9.96958330481389e-05}
{"train_loss": 0.7276772260665894, "global_step": 10018, "epoch": 112, "lr": 9.969576919563172e-05}
{"train_loss": 0.740925669670105, "global_step": 10019, "epoch": 112, "lr": 9.969570533644356e-05}
{"train_loss": 0.7117025852203369, "global_step": 10020, "epoch": 112, "lr": 9.969564147057441e-05}
{"train_loss": 0.5864128470420837, "global_step": 10021, "epoch": 112, "lr": 9.969557759802429e-05}
{"train_loss": 0.6772317886352539, "global_step": 10022, "epoch": 112, "lr": 9.969551371879321e-05}
{"train_loss": 0.8767911195755005, "global_step": 10023, "epoch": 112, "lr": 9.969544983288118e-05}
{"train_loss": 0.7373109459877014, "global_step": 10024, "epoch": 112, "lr": 9.969538594028818e-05}
{"train_loss": 0.661971390247345, "global_step": 10025, "epoch": 112, "lr": 9.969532204101424e-05}
{"train_loss": 0.7868400812149048, "global_step": 10026, "epoch": 112, "lr": 9.96952581350594e-05}
{"train_loss": 0.9061257243156433, "global_step": 10027, "epoch": 112, "lr": 9.969519422242361e-05}
{"train_loss": 0.8417103886604309, "global_step": 10028, "epoch": 112, "lr": 9.969513030310691e-05}
{"train_loss": 0.7769848108291626, "global_step": 10029, "epoch": 112, "lr": 9.969506637710932e-05}
{"train_loss": 0.7302696108818054, "global_step": 10030, "epoch": 112, "lr": 9.969500244443082e-05}
{"train_loss": 0.8541253209114075, "global_step": 10031, "epoch": 112, "lr": 9.969493850507144e-05}
{"train_loss": 0.6373787522315979, "global_step": 10032, "epoch": 112, "lr": 9.969487455903119e-05}
{"train_loss": 0.8173891305923462, "global_step": 10033, "epoch": 112, "lr": 9.969481060631005e-05}
{"train_loss": 0.8426195383071899, "global_step": 10034, "epoch": 112, "lr": 9.969474664690806e-05}
{"train_loss": 0.6834420561790466, "global_step": 10035, "epoch": 112, "lr": 9.969468268082523e-05}
{"train_loss": 0.9473649263381958, "global_step": 10036, "epoch": 112, "lr": 9.969461870806153e-05}
{"train_loss": 0.7418897151947021, "global_step": 10037, "epoch": 112, "lr": 9.9694554728617e-05}
{"train_loss": 0.857533872127533, "global_step": 10038, "epoch": 112, "lr": 9.969449074249166e-05}
{"train_loss": 0.8175023794174194, "global_step": 10039, "epoch": 112, "lr": 9.969442674968547e-05}
{"train_loss": 0.7239437103271484, "global_step": 10040, "epoch": 112, "lr": 9.96943627501985e-05}
{"train_loss": 0.8519722819328308, "global_step": 10041, "epoch": 112, "lr": 9.969429874403072e-05}
{"train_loss": 0.8068616390228271, "global_step": 10042, "epoch": 112, "lr": 9.969423473118215e-05}
{"train_loss": 0.7366840243339539, "global_step": 10043, "epoch": 112, "lr": 9.969417071165279e-05}
{"train_loss": 0.7667338252067566, "global_step": 10044, "epoch": 112, "lr": 9.969410668544265e-05}
{"train_loss": 0.7280383110046387, "global_step": 10045, "epoch": 112, "lr": 9.969404265255175e-05}
{"train_loss": 0.7208170294761658, "global_step": 10046, "epoch": 112, "lr": 9.969397861298009e-05}
{"train_loss": 0.8512334823608398, "global_step": 10047, "epoch": 112, "lr": 9.969391456672767e-05}
{"train_loss": 0.7429695725440979, "global_step": 10048, "epoch": 112, "lr": 9.969385051379453e-05}
{"train_loss": 0.9230538010597229, "global_step": 10049, "epoch": 112, "lr": 9.969378645418065e-05}
{"train_loss": 0.6331647634506226, "global_step": 10050, "epoch": 112, "lr": 9.969372238788604e-05}
{"train_loss": 0.7530226111412048, "global_step": 10051, "epoch": 112, "lr": 9.969365831491072e-05}
{"train_loss": 0.791161298751831, "global_step": 10052, "epoch": 112, "lr": 9.969359423525468e-05}
{"train_loss": 0.85273677110672, "global_step": 10053, "epoch": 112, "lr": 9.969353014891795e-05}
{"train_loss": 0.7182180881500244, "global_step": 10054, "epoch": 112, "lr": 9.969346605590055e-05}
{"train_loss": 0.792201817035675, "global_step": 10055, "epoch": 112, "lr": 9.969340195620245e-05}
{"train_loss": 0.7362030885192785, "global_step": 10056, "epoch": 112, "lr": 9.969333784982367e-05, "val_loss": 1.229552149772644}
{"train_loss": 0.8063379526138306, "global_step": 10057, "epoch": 113, "lr": 9.969327373676425e-05}
{"train_loss": 0.7915701270103455, "global_step": 10058, "epoch": 113, "lr": 9.969320961702415e-05}
{"train_loss": 0.6229519248008728, "global_step": 10059, "epoch": 113, "lr": 9.969314549060342e-05}
{"train_loss": 0.8613809943199158, "global_step": 10060, "epoch": 113, "lr": 9.969308135750204e-05}
{"train_loss": 0.7520294785499573, "global_step": 10061, "epoch": 113, "lr": 9.969301721772004e-05}
{"train_loss": 0.7180374264717102, "global_step": 10062, "epoch": 113, "lr": 9.969295307125743e-05}
{"train_loss": 0.5968579053878784, "global_step": 10063, "epoch": 113, "lr": 9.969288891811419e-05}
{"train_loss": 0.5591955184936523, "global_step": 10064, "epoch": 113, "lr": 9.969282475829036e-05}
{"train_loss": 0.6887980103492737, "global_step": 10065, "epoch": 113, "lr": 9.969276059178592e-05}
{"train_loss": 0.8276998400688171, "global_step": 10066, "epoch": 113, "lr": 9.969269641860091e-05}
{"train_loss": 0.6534034609794617, "global_step": 10067, "epoch": 113, "lr": 9.96926322387353e-05}
{"train_loss": 0.7834175825119019, "global_step": 10068, "epoch": 113, "lr": 9.969256805218916e-05}
{"train_loss": 0.6561436057090759, "global_step": 10069, "epoch": 113, "lr": 9.969250385896243e-05}
{"train_loss": 0.7024354934692383, "global_step": 10070, "epoch": 113, "lr": 9.969243965905514e-05}
{"train_loss": 0.7881000638008118, "global_step": 10071, "epoch": 113, "lr": 9.969237545246733e-05}
{"train_loss": 0.6890828013420105, "global_step": 10072, "epoch": 113, "lr": 9.969231123919897e-05}
{"train_loss": 0.7235709428787231, "global_step": 10073, "epoch": 113, "lr": 9.96922470192501e-05}
{"train_loss": 0.5885531902313232, "global_step": 10074, "epoch": 113, "lr": 9.96921827926207e-05}
{"train_loss": 0.816752016544342, "global_step": 10075, "epoch": 113, "lr": 9.96921185593108e-05}
{"train_loss": 0.7028562426567078, "global_step": 10076, "epoch": 113, "lr": 9.969205431932041e-05}
{"train_loss": 0.5926857590675354, "global_step": 10077, "epoch": 113, "lr": 9.969199007264951e-05}
{"train_loss": 0.6512623429298401, "global_step": 10078, "epoch": 113, "lr": 9.969192581929814e-05}
{"train_loss": 0.6017941236495972, "global_step": 10079, "epoch": 113, "lr": 9.969186155926629e-05}
{"train_loss": 0.6420103311538696, "global_step": 10080, "epoch": 113, "lr": 9.969179729255395e-05}
{"train_loss": 0.7864463329315186, "global_step": 10081, "epoch": 113, "lr": 9.969173301916119e-05}
{"train_loss": 0.7515823841094971, "global_step": 10082, "epoch": 113, "lr": 9.969166873908797e-05}
{"train_loss": 0.7276681065559387, "global_step": 10083, "epoch": 113, "lr": 9.969160445233431e-05}
{"train_loss": 0.694840133190155, "global_step": 10084, "epoch": 113, "lr": 9.969154015890023e-05}
{"train_loss": 0.7898191213607788, "global_step": 10085, "epoch": 113, "lr": 9.969147585878572e-05}
{"train_loss": 0.6502209901809692, "global_step": 10086, "epoch": 113, "lr": 9.969141155199079e-05}
{"train_loss": 0.8323952555656433, "global_step": 10087, "epoch": 113, "lr": 9.969134723851545e-05}
{"train_loss": 0.8966583013534546, "global_step": 10088, "epoch": 113, "lr": 9.969128291835974e-05}
{"train_loss": 0.7486666440963745, "global_step": 10089, "epoch": 113, "lr": 9.969121859152362e-05}
{"train_loss": 0.8694889545440674, "global_step": 10090, "epoch": 113, "lr": 9.969115425800713e-05}
{"train_loss": 0.627994954586029, "global_step": 10091, "epoch": 113, "lr": 9.969108991781026e-05}
{"train_loss": 0.9022055864334106, "global_step": 10092, "epoch": 113, "lr": 9.969102557093304e-05}
{"train_loss": 0.8766916990280151, "global_step": 10093, "epoch": 113, "lr": 9.969096121737547e-05}
{"train_loss": 0.7558797597885132, "global_step": 10094, "epoch": 113, "lr": 9.969089685713755e-05}
{"train_loss": 0.7192903161048889, "global_step": 10095, "epoch": 113, "lr": 9.96908324902193e-05}
{"train_loss": 0.7066924571990967, "global_step": 10096, "epoch": 113, "lr": 9.969076811662071e-05}
{"train_loss": 0.7496705055236816, "global_step": 10097, "epoch": 113, "lr": 9.969070373634181e-05}
{"train_loss": 0.7466626763343811, "global_step": 10098, "epoch": 113, "lr": 9.96906393493826e-05}
{"train_loss": 0.8875977993011475, "global_step": 10099, "epoch": 113, "lr": 9.969057495574309e-05}
{"train_loss": 0.5117167234420776, "global_step": 10100, "epoch": 113, "lr": 9.969051055542329e-05}
{"train_loss": 0.7866480350494385, "global_step": 10101, "epoch": 113, "lr": 9.969044614842321e-05}
{"train_loss": 0.9137414693832397, "global_step": 10102, "epoch": 113, "lr": 9.969038173474286e-05}
{"train_loss": 0.8495749235153198, "global_step": 10103, "epoch": 113, "lr": 9.969031731438222e-05}
{"train_loss": 0.5982570052146912, "global_step": 10104, "epoch": 113, "lr": 9.969025288734135e-05}
{"train_loss": 0.5807356834411621, "global_step": 10105, "epoch": 113, "lr": 9.96901884536202e-05}
{"train_loss": 0.709185004234314, "global_step": 10106, "epoch": 113, "lr": 9.969012401321883e-05}
{"train_loss": 0.7606241703033447, "global_step": 10107, "epoch": 113, "lr": 9.969005956613722e-05}
{"train_loss": 0.6860095262527466, "global_step": 10108, "epoch": 113, "lr": 9.96899951123754e-05}
{"train_loss": 0.8589830994606018, "global_step": 10109, "epoch": 113, "lr": 9.968993065193335e-05}
{"train_loss": 0.7401304841041565, "global_step": 10110, "epoch": 113, "lr": 9.96898661848111e-05}
{"train_loss": 0.7372068166732788, "global_step": 10111, "epoch": 113, "lr": 9.968980171100866e-05}
{"train_loss": 0.6206789016723633, "global_step": 10112, "epoch": 113, "lr": 9.968973723052603e-05}
{"train_loss": 0.7494885921478271, "global_step": 10113, "epoch": 113, "lr": 9.968967274336323e-05}
{"train_loss": 0.7513284087181091, "global_step": 10114, "epoch": 113, "lr": 9.968960824952024e-05}
{"train_loss": 0.6977399587631226, "global_step": 10115, "epoch": 113, "lr": 9.96895437489971e-05}
{"train_loss": 0.6554515361785889, "global_step": 10116, "epoch": 113, "lr": 9.968947924179379e-05}
{"train_loss": 0.6794072985649109, "global_step": 10117, "epoch": 113, "lr": 9.968941472791035e-05}
{"train_loss": 0.785371720790863, "global_step": 10118, "epoch": 113, "lr": 9.968935020734678e-05}
{"train_loss": 0.9185138940811157, "global_step": 10119, "epoch": 113, "lr": 9.968928568010306e-05}
{"train_loss": 0.8940489888191223, "global_step": 10120, "epoch": 113, "lr": 9.968922114617924e-05}
{"train_loss": 0.6188893914222717, "global_step": 10121, "epoch": 113, "lr": 9.96891566055753e-05}
{"train_loss": 0.6855889558792114, "global_step": 10122, "epoch": 113, "lr": 9.968909205829126e-05}
{"train_loss": 0.6052998900413513, "global_step": 10123, "epoch": 113, "lr": 9.968902750432713e-05}
{"train_loss": 0.7349192500114441, "global_step": 10124, "epoch": 113, "lr": 9.968896294368292e-05}
{"train_loss": 0.7191399931907654, "global_step": 10125, "epoch": 113, "lr": 9.968889837635861e-05}
{"train_loss": 0.6937891840934753, "global_step": 10126, "epoch": 113, "lr": 9.968883380235428e-05}
{"train_loss": 0.723522424697876, "global_step": 10127, "epoch": 113, "lr": 9.968876922166985e-05}
{"train_loss": 0.8284651637077332, "global_step": 10128, "epoch": 113, "lr": 9.96887046343054e-05}
{"train_loss": 0.8822077512741089, "global_step": 10129, "epoch": 113, "lr": 9.968864004026089e-05}
{"train_loss": 0.6726117134094238, "global_step": 10130, "epoch": 113, "lr": 9.968857543953635e-05}
{"train_loss": 0.7188501954078674, "global_step": 10131, "epoch": 113, "lr": 9.968851083213179e-05}
{"train_loss": 0.7808998227119446, "global_step": 10132, "epoch": 113, "lr": 9.968844621804721e-05}
{"train_loss": 0.728818416595459, "global_step": 10133, "epoch": 113, "lr": 9.968838159728263e-05}
{"train_loss": 0.6928443908691406, "global_step": 10134, "epoch": 113, "lr": 9.968831696983804e-05}
{"train_loss": 0.7559016942977905, "global_step": 10135, "epoch": 113, "lr": 9.968825233571348e-05}
{"train_loss": 0.6524977087974548, "global_step": 10136, "epoch": 113, "lr": 9.968818769490894e-05}
{"train_loss": 0.7908751368522644, "global_step": 10137, "epoch": 113, "lr": 9.968812304742442e-05}
{"train_loss": 0.7164231538772583, "global_step": 10138, "epoch": 113, "lr": 9.968805839325994e-05}
{"train_loss": 0.653927206993103, "global_step": 10139, "epoch": 113, "lr": 9.968799373241551e-05}
{"train_loss": 0.6290378570556641, "global_step": 10140, "epoch": 113, "lr": 9.968792906489112e-05}
{"train_loss": 0.7042171359062195, "global_step": 10141, "epoch": 113, "lr": 9.968786439068682e-05}
{"train_loss": 0.8893944621086121, "global_step": 10142, "epoch": 113, "lr": 9.968779970980258e-05}
{"train_loss": 0.8614139556884766, "global_step": 10143, "epoch": 113, "lr": 9.968773502223841e-05}
{"train_loss": 0.7511259317398071, "global_step": 10144, "epoch": 113, "lr": 9.968767032799433e-05}
{"train_loss": 0.733772892630502, "global_step": 10145, "epoch": 113, "lr": 9.968760562707035e-05, "val_loss": 1.2287057638168335}
{"train_loss": 0.7010819911956787, "global_step": 10146, "epoch": 114, "lr": 9.968754091946648e-05}
{"train_loss": 0.7815977334976196, "global_step": 10147, "epoch": 114, "lr": 9.968747620518275e-05}
{"train_loss": 0.689197838306427, "global_step": 10148, "epoch": 114, "lr": 9.968741148421912e-05}
{"train_loss": 0.7907516360282898, "global_step": 10149, "epoch": 114, "lr": 9.968734675657561e-05}
{"train_loss": 0.9593685865402222, "global_step": 10150, "epoch": 114, "lr": 9.968728202225225e-05}
{"train_loss": 0.754097044467926, "global_step": 10151, "epoch": 114, "lr": 9.968721728124906e-05}
{"train_loss": 0.9093886613845825, "global_step": 10152, "epoch": 114, "lr": 9.968715253356601e-05}
{"train_loss": 0.682981550693512, "global_step": 10153, "epoch": 114, "lr": 9.968708777920315e-05}
{"train_loss": 0.5465991497039795, "global_step": 10154, "epoch": 114, "lr": 9.968702301816044e-05}
{"train_loss": 0.7874460816383362, "global_step": 10155, "epoch": 114, "lr": 9.968695825043792e-05}
{"train_loss": 0.8449360728263855, "global_step": 10156, "epoch": 114, "lr": 9.968689347603561e-05}
{"train_loss": 0.6578748822212219, "global_step": 10157, "epoch": 114, "lr": 9.96868286949535e-05}
{"train_loss": 0.4864758551120758, "global_step": 10158, "epoch": 114, "lr": 9.968676390719158e-05}
{"train_loss": 0.7894043922424316, "global_step": 10159, "epoch": 114, "lr": 9.96866991127499e-05}
{"train_loss": 0.5436787009239197, "global_step": 10160, "epoch": 114, "lr": 9.968663431162846e-05}
{"train_loss": 0.6188755035400391, "global_step": 10161, "epoch": 114, "lr": 9.968656950382723e-05}
{"train_loss": 0.771342396736145, "global_step": 10162, "epoch": 114, "lr": 9.968650468934626e-05}
{"train_loss": 0.807084858417511, "global_step": 10163, "epoch": 114, "lr": 9.968643986818554e-05}
{"train_loss": 0.8148261904716492, "global_step": 10164, "epoch": 114, "lr": 9.968637504034509e-05}
{"train_loss": 0.6062374114990234, "global_step": 10165, "epoch": 114, "lr": 9.96863102058249e-05}
{"train_loss": 0.8903797268867493, "global_step": 10166, "epoch": 114, "lr": 9.968624536462501e-05}
{"train_loss": 0.7242568135261536, "global_step": 10167, "epoch": 114, "lr": 9.968618051674541e-05}
{"train_loss": 0.6745153069496155, "global_step": 10168, "epoch": 114, "lr": 9.968611566218608e-05}
{"train_loss": 0.690110981464386, "global_step": 10169, "epoch": 114, "lr": 9.968605080094709e-05}
{"train_loss": 0.8670569062232971, "global_step": 10170, "epoch": 114, "lr": 9.968598593302841e-05}
{"train_loss": 0.7605764865875244, "global_step": 10171, "epoch": 114, "lr": 9.968592105843004e-05}
{"train_loss": 0.6737551093101501, "global_step": 10172, "epoch": 114, "lr": 9.968585617715201e-05}
{"train_loss": 0.8987032175064087, "global_step": 10173, "epoch": 114, "lr": 9.968579128919433e-05}
{"train_loss": 0.8402668833732605, "global_step": 10174, "epoch": 114, "lr": 9.968572639455699e-05}
{"train_loss": 0.6714826822280884, "global_step": 10175, "epoch": 114, "lr": 9.968566149324002e-05}
{"train_loss": 0.6089668273925781, "global_step": 10176, "epoch": 114, "lr": 9.96855965852434e-05}
{"train_loss": 0.6665382385253906, "global_step": 10177, "epoch": 114, "lr": 9.968553167056718e-05}
{"train_loss": 0.6821150779724121, "global_step": 10178, "epoch": 114, "lr": 9.968546674921133e-05}
{"train_loss": 0.8521028757095337, "global_step": 10179, "epoch": 114, "lr": 9.968540182117587e-05}
{"train_loss": 0.8758375644683838, "global_step": 10180, "epoch": 114, "lr": 9.968533688646082e-05}
{"train_loss": 0.8514801263809204, "global_step": 10181, "epoch": 114, "lr": 9.968527194506619e-05}
{"train_loss": 0.7464115023612976, "global_step": 10182, "epoch": 114, "lr": 9.968520699699198e-05}
{"train_loss": 0.8429321646690369, "global_step": 10183, "epoch": 114, "lr": 9.96851420422382e-05}
{"train_loss": 0.5735061764717102, "global_step": 10184, "epoch": 114, "lr": 9.968507708080485e-05}
{"train_loss": 0.6055887341499329, "global_step": 10185, "epoch": 114, "lr": 9.968501211269194e-05}
{"train_loss": 0.8242554068565369, "global_step": 10186, "epoch": 114, "lr": 9.968494713789951e-05}
{"train_loss": 0.6690628528594971, "global_step": 10187, "epoch": 114, "lr": 9.968488215642753e-05}
{"train_loss": 0.9182679653167725, "global_step": 10188, "epoch": 114, "lr": 9.968481716827603e-05}
{"train_loss": 0.8125731945037842, "global_step": 10189, "epoch": 114, "lr": 9.968475217344499e-05}
{"train_loss": 0.6830130815505981, "global_step": 10190, "epoch": 114, "lr": 9.968468717193447e-05}
{"train_loss": 0.6960938572883606, "global_step": 10191, "epoch": 114, "lr": 9.968462216374444e-05}
{"train_loss": 0.6987090706825256, "global_step": 10192, "epoch": 114, "lr": 9.96845571488749e-05}
{"train_loss": 0.7560893297195435, "global_step": 10193, "epoch": 114, "lr": 9.968449212732589e-05}
{"train_loss": 0.7428144812583923, "global_step": 10194, "epoch": 114, "lr": 9.968442709909741e-05}
{"train_loss": 0.7291017174720764, "global_step": 10195, "epoch": 114, "lr": 9.968436206418946e-05}
{"train_loss": 0.6171239614486694, "global_step": 10196, "epoch": 114, "lr": 9.968429702260205e-05}
{"train_loss": 0.7250840067863464, "global_step": 10197, "epoch": 114, "lr": 9.968423197433521e-05}
{"train_loss": 0.8860894441604614, "global_step": 10198, "epoch": 114, "lr": 9.96841669193889e-05}
{"train_loss": 0.716722846031189, "global_step": 10199, "epoch": 114, "lr": 9.968410185776318e-05}
{"train_loss": 0.632379949092865, "global_step": 10200, "epoch": 114, "lr": 9.968403678945804e-05}
{"train_loss": 0.747955858707428, "global_step": 10201, "epoch": 114, "lr": 9.968397171447348e-05}
{"train_loss": 0.6017817854881287, "global_step": 10202, "epoch": 114, "lr": 9.96839066328095e-05}
{"train_loss": 0.630672037601471, "global_step": 10203, "epoch": 114, "lr": 9.968384154446615e-05}
{"train_loss": 0.7360362410545349, "global_step": 10204, "epoch": 114, "lr": 9.96837764494434e-05}
{"train_loss": 0.8216797113418579, "global_step": 10205, "epoch": 114, "lr": 9.968371134774128e-05}
{"train_loss": 0.6964191794395447, "global_step": 10206, "epoch": 114, "lr": 9.968364623935978e-05}
{"train_loss": 0.6511409878730774, "global_step": 10207, "epoch": 114, "lr": 9.968358112429892e-05}
{"train_loss": 0.7520655393600464, "global_step": 10208, "epoch": 114, "lr": 9.968351600255872e-05}
{"train_loss": 0.780185878276825, "global_step": 10209, "epoch": 114, "lr": 9.968345087413917e-05}
{"train_loss": 0.6629152297973633, "global_step": 10210, "epoch": 114, "lr": 9.968338573904029e-05}
{"train_loss": 0.7780300974845886, "global_step": 10211, "epoch": 114, "lr": 9.968332059726209e-05}
{"train_loss": 0.6375604867935181, "global_step": 10212, "epoch": 114, "lr": 9.968325544880455e-05}
{"train_loss": 0.6507554054260254, "global_step": 10213, "epoch": 114, "lr": 9.968319029366772e-05}
{"train_loss": 0.6068796515464783, "global_step": 10214, "epoch": 114, "lr": 9.968312513185159e-05}
{"train_loss": 0.578986406326294, "global_step": 10215, "epoch": 114, "lr": 9.968305996335617e-05}
{"train_loss": 0.6497270464897156, "global_step": 10216, "epoch": 114, "lr": 9.968299478818145e-05}
{"train_loss": 0.6363374590873718, "global_step": 10217, "epoch": 114, "lr": 9.968292960632747e-05}
{"train_loss": 0.7283784747123718, "global_step": 10218, "epoch": 114, "lr": 9.968286441779422e-05}
{"train_loss": 0.757276177406311, "global_step": 10219, "epoch": 114, "lr": 9.968279922258172e-05}
{"train_loss": 0.7101175785064697, "global_step": 10220, "epoch": 114, "lr": 9.968273402068997e-05}
{"train_loss": 0.7735783457756042, "global_step": 10221, "epoch": 114, "lr": 9.9682668812119e-05}
{"train_loss": 0.6733939051628113, "global_step": 10222, "epoch": 114, "lr": 9.968260359686879e-05}
{"train_loss": 0.6327643394470215, "global_step": 10223, "epoch": 114, "lr": 9.968253837493935e-05}
{"train_loss": 0.744056761264801, "global_step": 10224, "epoch": 114, "lr": 9.968247314633071e-05}
{"train_loss": 0.6360334157943726, "global_step": 10225, "epoch": 114, "lr": 9.968240791104285e-05}
{"train_loss": 0.6664165258407593, "global_step": 10226, "epoch": 114, "lr": 9.968234266907581e-05}
{"train_loss": 0.844792902469635, "global_step": 10227, "epoch": 114, "lr": 9.968227742042958e-05}
{"train_loss": 0.6275539994239807, "global_step": 10228, "epoch": 114, "lr": 9.968221216510417e-05}
{"train_loss": 0.5757925510406494, "global_step": 10229, "epoch": 114, "lr": 9.96821469030996e-05}
{"train_loss": 0.7667779326438904, "global_step": 10230, "epoch": 114, "lr": 9.968208163441586e-05}
{"train_loss": 0.8902847766876221, "global_step": 10231, "epoch": 114, "lr": 9.968201635905299e-05}
{"train_loss": 0.6472666263580322, "global_step": 10232, "epoch": 114, "lr": 9.968195107701097e-05}
{"train_loss": 0.7594119906425476, "global_step": 10233, "epoch": 114, "lr": 9.968188578828981e-05}
{"train_loss": 0.7213083187515816, "global_step": 10234, "epoch": 114, "lr": 9.968182049288952e-05, "val_loss": 1.2368038892745972}
{"train_loss": 0.5672426223754883, "global_step": 10235, "epoch": 115, "lr": 9.968175519081012e-05}
{"train_loss": 0.7194070219993591, "global_step": 10236, "epoch": 115, "lr": 9.968168988205161e-05}
{"train_loss": 0.6349398493766785, "global_step": 10237, "epoch": 115, "lr": 9.968162456661401e-05}
{"train_loss": 0.5607240796089172, "global_step": 10238, "epoch": 115, "lr": 9.968155924449733e-05}
{"train_loss": 0.4862019121646881, "global_step": 10239, "epoch": 115, "lr": 9.968149391570156e-05}
{"train_loss": 0.5933699011802673, "global_step": 10240, "epoch": 115, "lr": 9.968142858022672e-05}
{"train_loss": 0.6367262005805969, "global_step": 10241, "epoch": 115, "lr": 9.968136323807282e-05}
{"train_loss": 0.8263415098190308, "global_step": 10242, "epoch": 115, "lr": 9.968129788923986e-05}
{"train_loss": 0.4571853578090668, "global_step": 10243, "epoch": 115, "lr": 9.968123253372785e-05}
{"train_loss": 0.7698739767074585, "global_step": 10244, "epoch": 115, "lr": 9.968116717153681e-05}
{"train_loss": 0.6726282835006714, "global_step": 10245, "epoch": 115, "lr": 9.968110180266674e-05}
{"train_loss": 0.5606042742729187, "global_step": 10246, "epoch": 115, "lr": 9.968103642711767e-05}
{"train_loss": 0.5692668557167053, "global_step": 10247, "epoch": 115, "lr": 9.968097104488957e-05}
{"train_loss": 0.6830235123634338, "global_step": 10248, "epoch": 115, "lr": 9.968090565598247e-05}
{"train_loss": 0.7936905026435852, "global_step": 10249, "epoch": 115, "lr": 9.968084026039639e-05}
{"train_loss": 0.6617494225502014, "global_step": 10250, "epoch": 115, "lr": 9.968077485813132e-05}
{"train_loss": 0.6178172826766968, "global_step": 10251, "epoch": 115, "lr": 9.968070944918727e-05}
{"train_loss": 0.8572419881820679, "global_step": 10252, "epoch": 115, "lr": 9.968064403356427e-05}
{"train_loss": 0.812365710735321, "global_step": 10253, "epoch": 115, "lr": 9.96805786112623e-05}
{"train_loss": 0.7569643259048462, "global_step": 10254, "epoch": 115, "lr": 9.968051318228137e-05}
{"train_loss": 0.5299941897392273, "global_step": 10255, "epoch": 115, "lr": 9.968044774662153e-05}
{"train_loss": 0.7832511067390442, "global_step": 10256, "epoch": 115, "lr": 9.968038230428273e-05}
{"train_loss": 0.7217364311218262, "global_step": 10257, "epoch": 115, "lr": 9.968031685526502e-05}
{"train_loss": 0.691902756690979, "global_step": 10258, "epoch": 115, "lr": 9.968025139956841e-05}
{"train_loss": 0.7724570035934448, "global_step": 10259, "epoch": 115, "lr": 9.968018593719288e-05}
{"train_loss": 0.8523171544075012, "global_step": 10260, "epoch": 115, "lr": 9.968012046813846e-05}
{"train_loss": 0.7101613283157349, "global_step": 10261, "epoch": 115, "lr": 9.968005499240514e-05}
{"train_loss": 0.8108977675437927, "global_step": 10262, "epoch": 115, "lr": 9.967998950999295e-05}
{"train_loss": 0.6433990001678467, "global_step": 10263, "epoch": 115, "lr": 9.96799240209019e-05}
{"train_loss": 0.7671511173248291, "global_step": 10264, "epoch": 115, "lr": 9.967985852513199e-05}
{"train_loss": 0.735119640827179, "global_step": 10265, "epoch": 115, "lr": 9.967979302268323e-05}
{"train_loss": 0.6520087122917175, "global_step": 10266, "epoch": 115, "lr": 9.967972751355562e-05}
{"train_loss": 0.9452274441719055, "global_step": 10267, "epoch": 115, "lr": 9.967966199774916e-05}
{"train_loss": 0.7147446870803833, "global_step": 10268, "epoch": 115, "lr": 9.967959647526391e-05}
{"train_loss": 0.6892527341842651, "global_step": 10269, "epoch": 115, "lr": 9.967953094609981e-05}
{"train_loss": 0.7331096529960632, "global_step": 10270, "epoch": 115, "lr": 9.967946541025692e-05}
{"train_loss": 0.7809279561042786, "global_step": 10271, "epoch": 115, "lr": 9.967939986773523e-05}
{"train_loss": 0.6929738521575928, "global_step": 10272, "epoch": 115, "lr": 9.967933431853476e-05}
{"train_loss": 0.6771634221076965, "global_step": 10273, "epoch": 115, "lr": 9.967926876265549e-05}
{"train_loss": 0.8771394491195679, "global_step": 10274, "epoch": 115, "lr": 9.967920320009745e-05}
{"train_loss": 0.7029722929000854, "global_step": 10275, "epoch": 115, "lr": 9.967913763086065e-05}
{"train_loss": 0.7888554930686951, "global_step": 10276, "epoch": 115, "lr": 9.96790720549451e-05}
{"train_loss": 0.6544870138168335, "global_step": 10277, "epoch": 115, "lr": 9.96790064723508e-05}
{"train_loss": 0.7802371382713318, "global_step": 10278, "epoch": 115, "lr": 9.967894088307776e-05}
{"train_loss": 0.7780984044075012, "global_step": 10279, "epoch": 115, "lr": 9.9678875287126e-05}
{"train_loss": 0.8144220113754272, "global_step": 10280, "epoch": 115, "lr": 9.967880968449552e-05}
{"train_loss": 0.5952587723731995, "global_step": 10281, "epoch": 115, "lr": 9.967874407518634e-05}
{"train_loss": 0.7311421036720276, "global_step": 10282, "epoch": 115, "lr": 9.967867845919845e-05}
{"train_loss": 0.7347143888473511, "global_step": 10283, "epoch": 115, "lr": 9.967861283653186e-05}
{"train_loss": 0.6273451447486877, "global_step": 10284, "epoch": 115, "lr": 9.967854720718659e-05}
{"train_loss": 0.6538683176040649, "global_step": 10285, "epoch": 115, "lr": 9.967848157116264e-05}
{"train_loss": 0.6034331321716309, "global_step": 10286, "epoch": 115, "lr": 9.967841592846004e-05}
{"train_loss": 0.7512038946151733, "global_step": 10287, "epoch": 115, "lr": 9.967835027907876e-05}
{"train_loss": 0.9945827126502991, "global_step": 10288, "epoch": 115, "lr": 9.967828462301884e-05}
{"train_loss": 0.6187010407447815, "global_step": 10289, "epoch": 115, "lr": 9.967821896028028e-05}
{"train_loss": 0.7384750247001648, "global_step": 10290, "epoch": 115, "lr": 9.967815329086309e-05}
{"train_loss": 0.6494090557098389, "global_step": 10291, "epoch": 115, "lr": 9.967808761476728e-05}
{"train_loss": 0.6746240854263306, "global_step": 10292, "epoch": 115, "lr": 9.967802193199285e-05}
{"train_loss": 0.7515748739242554, "global_step": 10293, "epoch": 115, "lr": 9.967795624253981e-05}
{"train_loss": 0.8464312553405762, "global_step": 10294, "epoch": 115, "lr": 9.96778905464082e-05}
{"train_loss": 0.658515453338623, "global_step": 10295, "epoch": 115, "lr": 9.967782484359798e-05}
{"train_loss": 0.8495922088623047, "global_step": 10296, "epoch": 115, "lr": 9.96777591341092e-05}
{"train_loss": 0.6873142719268799, "global_step": 10297, "epoch": 115, "lr": 9.967769341794184e-05}
{"train_loss": 0.8571226596832275, "global_step": 10298, "epoch": 115, "lr": 9.967762769509592e-05}
{"train_loss": 0.6740825772285461, "global_step": 10299, "epoch": 115, "lr": 9.967756196557144e-05}
{"train_loss": 0.7954219579696655, "global_step": 10300, "epoch": 115, "lr": 9.967749622936843e-05}
{"train_loss": 0.7002845406532288, "global_step": 10301, "epoch": 115, "lr": 9.967743048648687e-05}
{"train_loss": 0.7058035135269165, "global_step": 10302, "epoch": 115, "lr": 9.967736473692682e-05}
{"train_loss": 0.8689380884170532, "global_step": 10303, "epoch": 115, "lr": 9.967729898068823e-05}
{"train_loss": 0.5814180970191956, "global_step": 10304, "epoch": 115, "lr": 9.967723321777112e-05}
{"train_loss": 0.7177166938781738, "global_step": 10305, "epoch": 115, "lr": 9.967716744817553e-05}
{"train_loss": 0.7195068597793579, "global_step": 10306, "epoch": 115, "lr": 9.967710167190144e-05}
{"train_loss": 0.7382405400276184, "global_step": 10307, "epoch": 115, "lr": 9.967703588894888e-05}
{"train_loss": 0.6989039182662964, "global_step": 10308, "epoch": 115, "lr": 9.967697009931785e-05}
{"train_loss": 0.6651595830917358, "global_step": 10309, "epoch": 115, "lr": 9.967690430300833e-05}
{"train_loss": 0.6438028216362, "global_step": 10310, "epoch": 115, "lr": 9.967683850002039e-05}
{"train_loss": 0.6783665418624878, "global_step": 10311, "epoch": 115, "lr": 9.967677269035399e-05}
{"train_loss": 0.6463119387626648, "global_step": 10312, "epoch": 115, "lr": 9.967670687400914e-05}
{"train_loss": 0.8692563772201538, "global_step": 10313, "epoch": 115, "lr": 9.967664105098588e-05}
{"train_loss": 0.658109724521637, "global_step": 10314, "epoch": 115, "lr": 9.967657522128419e-05}
{"train_loss": 0.7740780115127563, "global_step": 10315, "epoch": 115, "lr": 9.96765093849041e-05}
{"train_loss": 0.7087652683258057, "global_step": 10316, "epoch": 115, "lr": 9.96764435418456e-05}
{"train_loss": 0.8350591659545898, "global_step": 10317, "epoch": 115, "lr": 9.967637769210871e-05}
{"train_loss": 0.58698570728302, "global_step": 10318, "epoch": 115, "lr": 9.967631183569344e-05}
{"train_loss": 0.570426344871521, "global_step": 10319, "epoch": 115, "lr": 9.96762459725998e-05}
{"train_loss": 0.7022984027862549, "global_step": 10320, "epoch": 115, "lr": 9.967618010282779e-05}
{"train_loss": 0.6219606995582581, "global_step": 10321, "epoch": 115, "lr": 9.967611422637742e-05}
{"train_loss": 0.6210700273513794, "global_step": 10322, "epoch": 115, "lr": 9.96760483432487e-05}
{"train_loss": 0.7079355502396487, "global_step": 10323, "epoch": 115, "lr": 9.967598245344165e-05, "val_loss": 1.2530547380447388, "train_action_mse_error": 55.73484420776367}
{"train_loss": 0.7567503452301025, "global_step": 10324, "epoch": 116, "lr": 9.967591655695626e-05}
{"train_loss": 0.7848234176635742, "global_step": 10325, "epoch": 116, "lr": 9.967585065379256e-05}
{"train_loss": 0.6024011373519897, "global_step": 10326, "epoch": 116, "lr": 9.967578474395054e-05}
{"train_loss": 0.7074146866798401, "global_step": 10327, "epoch": 116, "lr": 9.967571882743021e-05}
{"train_loss": 0.8529732823371887, "global_step": 10328, "epoch": 116, "lr": 9.96756529042316e-05}
{"train_loss": 0.6016783714294434, "global_step": 10329, "epoch": 116, "lr": 9.967558697435468e-05}
{"train_loss": 0.6809849739074707, "global_step": 10330, "epoch": 116, "lr": 9.96755210377995e-05}
{"train_loss": 0.7753415107727051, "global_step": 10331, "epoch": 116, "lr": 9.967545509456605e-05}
{"train_loss": 0.880984365940094, "global_step": 10332, "epoch": 116, "lr": 9.967538914465434e-05}
{"train_loss": 0.7238366603851318, "global_step": 10333, "epoch": 116, "lr": 9.967532318806439e-05}
{"train_loss": 0.7612248063087463, "global_step": 10334, "epoch": 116, "lr": 9.967525722479619e-05}
{"train_loss": 0.6794004440307617, "global_step": 10335, "epoch": 116, "lr": 9.967519125484974e-05}
{"train_loss": 0.625262975692749, "global_step": 10336, "epoch": 116, "lr": 9.967512527822509e-05}
{"train_loss": 0.749538779258728, "global_step": 10337, "epoch": 116, "lr": 9.96750592949222e-05}
{"train_loss": 0.6733709573745728, "global_step": 10338, "epoch": 116, "lr": 9.967499330494113e-05}
{"train_loss": 0.7206177711486816, "global_step": 10339, "epoch": 116, "lr": 9.967492730828185e-05}
{"train_loss": 0.7422090172767639, "global_step": 10340, "epoch": 116, "lr": 9.96748613049444e-05}
{"train_loss": 0.6871971487998962, "global_step": 10341, "epoch": 116, "lr": 9.967479529492873e-05}
{"train_loss": 0.6980071663856506, "global_step": 10342, "epoch": 116, "lr": 9.967472927823491e-05}
{"train_loss": 0.9277507662773132, "global_step": 10343, "epoch": 116, "lr": 9.967466325486295e-05}
{"train_loss": 0.6509236693382263, "global_step": 10344, "epoch": 116, "lr": 9.96745972248128e-05}
{"train_loss": 0.6244098544120789, "global_step": 10345, "epoch": 116, "lr": 9.967453118808453e-05}
{"train_loss": 0.7709517478942871, "global_step": 10346, "epoch": 116, "lr": 9.967446514467811e-05}
{"train_loss": 0.726913332939148, "global_step": 10347, "epoch": 116, "lr": 9.967439909459359e-05}
{"train_loss": 0.659548282623291, "global_step": 10348, "epoch": 116, "lr": 9.967433303783092e-05}
{"train_loss": 0.8649525046348572, "global_step": 10349, "epoch": 116, "lr": 9.967426697439017e-05}
{"train_loss": 0.7824016809463501, "global_step": 10350, "epoch": 116, "lr": 9.96742009042713e-05}
{"train_loss": 0.637847363948822, "global_step": 10351, "epoch": 116, "lr": 9.967413482747435e-05}
{"train_loss": 0.7237149477005005, "global_step": 10352, "epoch": 116, "lr": 9.967406874399931e-05}
{"train_loss": 0.6860603094100952, "global_step": 10353, "epoch": 116, "lr": 9.96740026538462e-05}
{"train_loss": 0.7775774002075195, "global_step": 10354, "epoch": 116, "lr": 9.967393655701503e-05}
{"train_loss": 0.5148607492446899, "global_step": 10355, "epoch": 116, "lr": 9.96738704535058e-05}
{"train_loss": 0.6226892471313477, "global_step": 10356, "epoch": 116, "lr": 9.967380434331852e-05}
{"train_loss": 0.790837824344635, "global_step": 10357, "epoch": 116, "lr": 9.96737382264532e-05}
{"train_loss": 0.4823794364929199, "global_step": 10358, "epoch": 116, "lr": 9.967367210290986e-05}
{"train_loss": 0.6640353798866272, "global_step": 10359, "epoch": 116, "lr": 9.96736059726885e-05}
{"train_loss": 0.6524999737739563, "global_step": 10360, "epoch": 116, "lr": 9.967353983578913e-05}
{"train_loss": 0.8120346069335938, "global_step": 10361, "epoch": 116, "lr": 9.967347369221176e-05}
{"train_loss": 0.8527954816818237, "global_step": 10362, "epoch": 116, "lr": 9.967340754195639e-05}
{"train_loss": 0.5404254198074341, "global_step": 10363, "epoch": 116, "lr": 9.967334138502304e-05}
{"train_loss": 0.6598309278488159, "global_step": 10364, "epoch": 116, "lr": 9.96732752214117e-05}
{"train_loss": 0.7286332249641418, "global_step": 10365, "epoch": 116, "lr": 9.967320905112241e-05}
{"train_loss": 0.8261949419975281, "global_step": 10366, "epoch": 116, "lr": 9.967314287415517e-05}
{"train_loss": 0.6899946928024292, "global_step": 10367, "epoch": 116, "lr": 9.967307669050996e-05}
{"train_loss": 0.750494122505188, "global_step": 10368, "epoch": 116, "lr": 9.967301050018683e-05}
{"train_loss": 0.7203846573829651, "global_step": 10369, "epoch": 116, "lr": 9.967294430318574e-05}
{"train_loss": 0.6640335321426392, "global_step": 10370, "epoch": 116, "lr": 9.967287809950677e-05}
{"train_loss": 0.7723835110664368, "global_step": 10371, "epoch": 116, "lr": 9.967281188914985e-05}
{"train_loss": 0.8419303894042969, "global_step": 10372, "epoch": 116, "lr": 9.967274567211504e-05}
{"train_loss": 0.8952073454856873, "global_step": 10373, "epoch": 116, "lr": 9.967267944840234e-05}
{"train_loss": 0.7087064385414124, "global_step": 10374, "epoch": 116, "lr": 9.967261321801175e-05}
{"train_loss": 0.7278035879135132, "global_step": 10375, "epoch": 116, "lr": 9.967254698094328e-05}
{"train_loss": 0.8766670823097229, "global_step": 10376, "epoch": 116, "lr": 9.967248073719695e-05}
{"train_loss": 0.6237556338310242, "global_step": 10377, "epoch": 116, "lr": 9.967241448677275e-05}
{"train_loss": 0.6933063268661499, "global_step": 10378, "epoch": 116, "lr": 9.96723482296707e-05}
{"train_loss": 0.7103260159492493, "global_step": 10379, "epoch": 116, "lr": 9.96722819658908e-05}
{"train_loss": 0.8625974059104919, "global_step": 10380, "epoch": 116, "lr": 9.967221569543308e-05}
{"train_loss": 0.6753756999969482, "global_step": 10381, "epoch": 116, "lr": 9.967214941829755e-05}
{"train_loss": 0.7362915277481079, "global_step": 10382, "epoch": 116, "lr": 9.967208313448418e-05}
{"train_loss": 0.7987344861030579, "global_step": 10383, "epoch": 116, "lr": 9.967201684399301e-05}
{"train_loss": 0.6969186067581177, "global_step": 10384, "epoch": 116, "lr": 9.967195054682406e-05}
{"train_loss": 0.7806355953216553, "global_step": 10385, "epoch": 116, "lr": 9.967188424297729e-05}
{"train_loss": 0.8638113141059875, "global_step": 10386, "epoch": 116, "lr": 9.967181793245275e-05}
{"train_loss": 0.6501636505126953, "global_step": 10387, "epoch": 116, "lr": 9.967175161525047e-05}
{"train_loss": 0.732631266117096, "global_step": 10388, "epoch": 116, "lr": 9.967168529137039e-05}
{"train_loss": 0.7529077529907227, "global_step": 10389, "epoch": 116, "lr": 9.967161896081259e-05}
{"train_loss": 0.6686424612998962, "global_step": 10390, "epoch": 116, "lr": 9.967155262357701e-05}
{"train_loss": 0.8226125240325928, "global_step": 10391, "epoch": 116, "lr": 9.967148627966371e-05}
{"train_loss": 0.8635135889053345, "global_step": 10392, "epoch": 116, "lr": 9.967141992907269e-05}
{"train_loss": 0.8496658802032471, "global_step": 10393, "epoch": 116, "lr": 9.967135357180396e-05}
{"train_loss": 0.8371473550796509, "global_step": 10394, "epoch": 116, "lr": 9.967128720785753e-05}
{"train_loss": 0.7038859724998474, "global_step": 10395, "epoch": 116, "lr": 9.967122083723338e-05}
{"train_loss": 0.6978832483291626, "global_step": 10396, "epoch": 116, "lr": 9.967115445993154e-05}
{"train_loss": 0.7326395511627197, "global_step": 10397, "epoch": 116, "lr": 9.967108807595202e-05}
{"train_loss": 0.813675045967102, "global_step": 10398, "epoch": 116, "lr": 9.967102168529482e-05}
{"train_loss": 0.6729230284690857, "global_step": 10399, "epoch": 116, "lr": 9.967095528795998e-05}
{"train_loss": 0.6530191898345947, "global_step": 10400, "epoch": 116, "lr": 9.967088888394747e-05}
{"train_loss": 0.6223430633544922, "global_step": 10401, "epoch": 116, "lr": 9.967082247325732e-05}
{"train_loss": 0.5592088103294373, "global_step": 10402, "epoch": 116, "lr": 9.967075605588951e-05}
{"train_loss": 0.5242121815681458, "global_step": 10403, "epoch": 116, "lr": 9.96706896318441e-05}
{"train_loss": 0.5988136529922485, "global_step": 10404, "epoch": 116, "lr": 9.967062320112107e-05}
{"train_loss": 0.7721471190452576, "global_step": 10405, "epoch": 116, "lr": 9.967055676372041e-05}
{"train_loss": 0.7019745707511902, "global_step": 10406, "epoch": 116, "lr": 9.967049031964217e-05}
{"train_loss": 0.7451671361923218, "global_step": 10407, "epoch": 116, "lr": 9.967042386888635e-05}
{"train_loss": 0.6789497137069702, "global_step": 10408, "epoch": 116, "lr": 9.967035741145292e-05}
{"train_loss": 0.6229636073112488, "global_step": 10409, "epoch": 116, "lr": 9.967029094734192e-05}
{"train_loss": 0.7094508409500122, "global_step": 10410, "epoch": 116, "lr": 9.967022447655335e-05}
{"train_loss": 0.8128871321678162, "global_step": 10411, "epoch": 116, "lr": 9.967015799908723e-05}
{"train_loss": 0.7214814839738138, "global_step": 10412, "epoch": 116, "lr": 9.967009151494357e-05, "val_loss": 1.2364171743392944}
{"train_loss": 0.6345239877700806, "global_step": 10413, "epoch": 117, "lr": 9.967002502412237e-05}
{"train_loss": 0.5898803472518921, "global_step": 10414, "epoch": 117, "lr": 9.966995852662363e-05}
{"train_loss": 0.6941174864768982, "global_step": 10415, "epoch": 117, "lr": 9.966989202244737e-05}
{"train_loss": 0.755990743637085, "global_step": 10416, "epoch": 117, "lr": 9.966982551159361e-05}
{"train_loss": 0.7231202125549316, "global_step": 10417, "epoch": 117, "lr": 9.966975899406234e-05}
{"train_loss": 0.67794269323349, "global_step": 10418, "epoch": 117, "lr": 9.966969246985358e-05}
{"train_loss": 0.7356398701667786, "global_step": 10419, "epoch": 117, "lr": 9.966962593896734e-05}
{"train_loss": 0.7225577235221863, "global_step": 10420, "epoch": 117, "lr": 9.966955940140361e-05}
{"train_loss": 0.6395723223686218, "global_step": 10421, "epoch": 117, "lr": 9.966949285716243e-05}
{"train_loss": 0.6469089388847351, "global_step": 10422, "epoch": 117, "lr": 9.966942630624378e-05}
{"train_loss": 0.6897522807121277, "global_step": 10423, "epoch": 117, "lr": 9.966935974864768e-05}
{"train_loss": 0.6126164197921753, "global_step": 10424, "epoch": 117, "lr": 9.966929318437414e-05}
{"train_loss": 0.606768786907196, "global_step": 10425, "epoch": 117, "lr": 9.966922661342317e-05}
{"train_loss": 0.5891452431678772, "global_step": 10426, "epoch": 117, "lr": 9.966916003579478e-05}
{"train_loss": 0.6334090232849121, "global_step": 10427, "epoch": 117, "lr": 9.966909345148898e-05}
{"train_loss": 0.973590612411499, "global_step": 10428, "epoch": 117, "lr": 9.966902686050577e-05}
{"train_loss": 0.5962886810302734, "global_step": 10429, "epoch": 117, "lr": 9.966896026284517e-05}
{"train_loss": 0.7113385796546936, "global_step": 10430, "epoch": 117, "lr": 9.966889365850718e-05}
{"train_loss": 0.7075716853141785, "global_step": 10431, "epoch": 117, "lr": 9.966882704749182e-05}
{"train_loss": 0.8019259572029114, "global_step": 10432, "epoch": 117, "lr": 9.96687604297991e-05}
{"train_loss": 0.7122957706451416, "global_step": 10433, "epoch": 117, "lr": 9.9668693805429e-05}
{"train_loss": 0.6169717907905579, "global_step": 10434, "epoch": 117, "lr": 9.966862717438154e-05}
{"train_loss": 0.7855846881866455, "global_step": 10435, "epoch": 117, "lr": 9.966856053665676e-05}
{"train_loss": 0.6814266443252563, "global_step": 10436, "epoch": 117, "lr": 9.966849389225464e-05}
{"train_loss": 0.6711934208869934, "global_step": 10437, "epoch": 117, "lr": 9.96684272411752e-05}
{"train_loss": 0.7151811718940735, "global_step": 10438, "epoch": 117, "lr": 9.966836058341844e-05}
{"train_loss": 0.6575589179992676, "global_step": 10439, "epoch": 117, "lr": 9.966829391898437e-05}
{"train_loss": 0.6499303579330444, "global_step": 10440, "epoch": 117, "lr": 9.966822724787302e-05}
{"train_loss": 0.6990736722946167, "global_step": 10441, "epoch": 117, "lr": 9.966816057008436e-05}
{"train_loss": 0.7147476673126221, "global_step": 10442, "epoch": 117, "lr": 9.966809388561846e-05}
{"train_loss": 0.714306652545929, "global_step": 10443, "epoch": 117, "lr": 9.966802719447525e-05}
{"train_loss": 0.6030515432357788, "global_step": 10444, "epoch": 117, "lr": 9.96679604966548e-05}
{"train_loss": 0.7353707551956177, "global_step": 10445, "epoch": 117, "lr": 9.966789379215709e-05}
{"train_loss": 0.6214278340339661, "global_step": 10446, "epoch": 117, "lr": 9.966782708098213e-05}
{"train_loss": 0.6129940152168274, "global_step": 10447, "epoch": 117, "lr": 9.966776036312994e-05}
{"train_loss": 0.6765061616897583, "global_step": 10448, "epoch": 117, "lr": 9.966769363860053e-05}
{"train_loss": 0.6592080593109131, "global_step": 10449, "epoch": 117, "lr": 9.966762690739391e-05}
{"train_loss": 0.6476876735687256, "global_step": 10450, "epoch": 117, "lr": 9.966756016951008e-05}
{"train_loss": 0.7074217200279236, "global_step": 10451, "epoch": 117, "lr": 9.966749342494906e-05}
{"train_loss": 0.6836482882499695, "global_step": 10452, "epoch": 117, "lr": 9.966742667371083e-05}
{"train_loss": 0.7017316222190857, "global_step": 10453, "epoch": 117, "lr": 9.966735991579543e-05}
{"train_loss": 0.7528056502342224, "global_step": 10454, "epoch": 117, "lr": 9.966729315120285e-05}
{"train_loss": 0.8074998259544373, "global_step": 10455, "epoch": 117, "lr": 9.966722637993312e-05}
{"train_loss": 0.6745356917381287, "global_step": 10456, "epoch": 117, "lr": 9.966715960198623e-05}
{"train_loss": 0.6066531538963318, "global_step": 10457, "epoch": 117, "lr": 9.96670928173622e-05}
{"train_loss": 0.6882245540618896, "global_step": 10458, "epoch": 117, "lr": 9.966702602606103e-05}
{"train_loss": 0.5843701362609863, "global_step": 10459, "epoch": 117, "lr": 9.966695922808274e-05}
{"train_loss": 0.6340639591217041, "global_step": 10460, "epoch": 117, "lr": 9.966689242342733e-05}
{"train_loss": 0.7300571203231812, "global_step": 10461, "epoch": 117, "lr": 9.966682561209482e-05}
{"train_loss": 0.7198708057403564, "global_step": 10462, "epoch": 117, "lr": 9.966675879408519e-05}
{"train_loss": 0.7852954268455505, "global_step": 10463, "epoch": 117, "lr": 9.966669196939848e-05}
{"train_loss": 0.6908817887306213, "global_step": 10464, "epoch": 117, "lr": 9.96666251380347e-05}
{"train_loss": 0.6657198071479797, "global_step": 10465, "epoch": 117, "lr": 9.966655829999383e-05}
{"train_loss": 0.6935572624206543, "global_step": 10466, "epoch": 117, "lr": 9.966649145527591e-05}
{"train_loss": 0.8033002614974976, "global_step": 10467, "epoch": 117, "lr": 9.966642460388092e-05}
{"train_loss": 0.6917663812637329, "global_step": 10468, "epoch": 117, "lr": 9.96663577458089e-05}
{"train_loss": 0.600100040435791, "global_step": 10469, "epoch": 117, "lr": 9.966629088105983e-05}
{"train_loss": 0.9159426689147949, "global_step": 10470, "epoch": 117, "lr": 9.966622400963375e-05}
{"train_loss": 0.590706467628479, "global_step": 10471, "epoch": 117, "lr": 9.966615713153065e-05}
{"train_loss": 0.66895592212677, "global_step": 10472, "epoch": 117, "lr": 9.966609024675052e-05}
{"train_loss": 0.6826595664024353, "global_step": 10473, "epoch": 117, "lr": 9.96660233552934e-05}
{"train_loss": 0.6475170850753784, "global_step": 10474, "epoch": 117, "lr": 9.96659564571593e-05}
{"train_loss": 0.804153323173523, "global_step": 10475, "epoch": 117, "lr": 9.966588955234821e-05}
{"train_loss": 0.6000372767448425, "global_step": 10476, "epoch": 117, "lr": 9.966582264086012e-05}
{"train_loss": 0.6770104169845581, "global_step": 10477, "epoch": 117, "lr": 9.96657557226951e-05}
{"train_loss": 0.5996201038360596, "global_step": 10478, "epoch": 117, "lr": 9.966568879785312e-05}
{"train_loss": 0.769611120223999, "global_step": 10479, "epoch": 117, "lr": 9.966562186633419e-05}
{"train_loss": 0.8298050761222839, "global_step": 10480, "epoch": 117, "lr": 9.966555492813831e-05}
{"train_loss": 0.7532809376716614, "global_step": 10481, "epoch": 117, "lr": 9.966548798326552e-05}
{"train_loss": 0.6397038102149963, "global_step": 10482, "epoch": 117, "lr": 9.96654210317158e-05}
{"train_loss": 0.6656087040901184, "global_step": 10483, "epoch": 117, "lr": 9.966535407348919e-05}
{"train_loss": 0.733197033405304, "global_step": 10484, "epoch": 117, "lr": 9.966528710858565e-05}
{"train_loss": 0.724769115447998, "global_step": 10485, "epoch": 117, "lr": 9.966522013700522e-05}
{"train_loss": 0.59751957654953, "global_step": 10486, "epoch": 117, "lr": 9.966515315874792e-05}
{"train_loss": 0.6949049830436707, "global_step": 10487, "epoch": 117, "lr": 9.966508617381373e-05}
{"train_loss": 0.6848173141479492, "global_step": 10488, "epoch": 117, "lr": 9.966501918220269e-05}
{"train_loss": 0.6125626564025879, "global_step": 10489, "epoch": 117, "lr": 9.966495218391478e-05}
{"train_loss": 0.6726135015487671, "global_step": 10490, "epoch": 117, "lr": 9.966488517895004e-05}
{"train_loss": 0.7355667948722839, "global_step": 10491, "epoch": 117, "lr": 9.966481816730845e-05}
{"train_loss": 0.6634159088134766, "global_step": 10492, "epoch": 117, "lr": 9.966475114899003e-05}
{"train_loss": 0.7890878915786743, "global_step": 10493, "epoch": 117, "lr": 9.966468412399479e-05}
{"train_loss": 0.7748862504959106, "global_step": 10494, "epoch": 117, "lr": 9.966461709232274e-05}
{"train_loss": 0.6544560194015503, "global_step": 10495, "epoch": 117, "lr": 9.96645500539739e-05}
{"train_loss": 0.6969650983810425, "global_step": 10496, "epoch": 117, "lr": 9.966448300894824e-05}
{"train_loss": 0.5671762228012085, "global_step": 10497, "epoch": 117, "lr": 9.966441595724582e-05}
{"train_loss": 0.6742649078369141, "global_step": 10498, "epoch": 117, "lr": 9.966434889886661e-05}
{"train_loss": 0.7956988215446472, "global_step": 10499, "epoch": 117, "lr": 9.966428183381064e-05}
{"train_loss": 0.6970884799957275, "global_step": 10500, "epoch": 117, "lr": 9.966421476207791e-05}
{"train_loss": 0.6909390982617153, "global_step": 10501, "epoch": 117, "lr": 9.966414768366844e-05, "val_loss": 1.2297569513320923}
{"train_loss": 0.7615801692008972, "global_step": 10502, "epoch": 118, "lr": 9.966408059858221e-05}
{"train_loss": 0.5382869243621826, "global_step": 10503, "epoch": 118, "lr": 9.966401350681927e-05}
{"train_loss": 0.6763972640037537, "global_step": 10504, "epoch": 118, "lr": 9.966394640837959e-05}
{"train_loss": 0.7670232653617859, "global_step": 10505, "epoch": 118, "lr": 9.966387930326322e-05}
{"train_loss": 0.7130705118179321, "global_step": 10506, "epoch": 118, "lr": 9.966381219147014e-05}
{"train_loss": 0.7647624015808105, "global_step": 10507, "epoch": 118, "lr": 9.966374507300036e-05}
{"train_loss": 0.7393162250518799, "global_step": 10508, "epoch": 118, "lr": 9.96636779478539e-05}
{"train_loss": 0.6040283441543579, "global_step": 10509, "epoch": 118, "lr": 9.966361081603076e-05}
{"train_loss": 0.6063037514686584, "global_step": 10510, "epoch": 118, "lr": 9.966354367753094e-05}
{"train_loss": 0.6975605487823486, "global_step": 10511, "epoch": 118, "lr": 9.966347653235448e-05}
{"train_loss": 0.7224410176277161, "global_step": 10512, "epoch": 118, "lr": 9.966340938050137e-05}
{"train_loss": 0.6833896040916443, "global_step": 10513, "epoch": 118, "lr": 9.966334222197161e-05}
{"train_loss": 0.6930029392242432, "global_step": 10514, "epoch": 118, "lr": 9.966327505676523e-05}
{"train_loss": 0.6036747694015503, "global_step": 10515, "epoch": 118, "lr": 9.966320788488221e-05}
{"train_loss": 0.6353482604026794, "global_step": 10516, "epoch": 118, "lr": 9.96631407063226e-05}
{"train_loss": 0.7471210360527039, "global_step": 10517, "epoch": 118, "lr": 9.966307352108637e-05}
{"train_loss": 0.6434706449508667, "global_step": 10518, "epoch": 118, "lr": 9.966300632917355e-05}
{"train_loss": 0.5919591188430786, "global_step": 10519, "epoch": 118, "lr": 9.966293913058415e-05}
{"train_loss": 0.6014561057090759, "global_step": 10520, "epoch": 118, "lr": 9.966287192531817e-05}
{"train_loss": 0.7613860964775085, "global_step": 10521, "epoch": 118, "lr": 9.966280471337563e-05}
{"train_loss": 0.5865923762321472, "global_step": 10522, "epoch": 118, "lr": 9.966273749475652e-05}
{"train_loss": 0.6424081325531006, "global_step": 10523, "epoch": 118, "lr": 9.966267026946086e-05}
{"train_loss": 0.7429876923561096, "global_step": 10524, "epoch": 118, "lr": 9.966260303748866e-05}
{"train_loss": 0.6837292313575745, "global_step": 10525, "epoch": 118, "lr": 9.966253579883993e-05}
{"train_loss": 0.6416588425636292, "global_step": 10526, "epoch": 118, "lr": 9.966246855351467e-05}
{"train_loss": 0.5570803284645081, "global_step": 10527, "epoch": 118, "lr": 9.96624013015129e-05}
{"train_loss": 0.8125823736190796, "global_step": 10528, "epoch": 118, "lr": 9.966233404283462e-05}
{"train_loss": 0.6356114745140076, "global_step": 10529, "epoch": 118, "lr": 9.966226677747988e-05}
{"train_loss": 0.800389289855957, "global_step": 10530, "epoch": 118, "lr": 9.966219950544862e-05}
{"train_loss": 0.6281664371490479, "global_step": 10531, "epoch": 118, "lr": 9.966213222674088e-05}
{"train_loss": 0.7204373478889465, "global_step": 10532, "epoch": 118, "lr": 9.966206494135668e-05}
{"train_loss": 0.6272600293159485, "global_step": 10533, "epoch": 118, "lr": 9.966199764929603e-05}
{"train_loss": 0.6606033444404602, "global_step": 10534, "epoch": 118, "lr": 9.966193035055891e-05}
{"train_loss": 0.6568798422813416, "global_step": 10535, "epoch": 118, "lr": 9.966186304514537e-05}
{"train_loss": 0.7303027510643005, "global_step": 10536, "epoch": 118, "lr": 9.966179573305538e-05}
{"train_loss": 0.7057770490646362, "global_step": 10537, "epoch": 118, "lr": 9.966172841428898e-05}
{"train_loss": 0.7133859395980835, "global_step": 10538, "epoch": 118, "lr": 9.966166108884616e-05}
{"train_loss": 0.6875573396682739, "global_step": 10539, "epoch": 118, "lr": 9.966159375672694e-05}
{"train_loss": 0.6582042574882507, "global_step": 10540, "epoch": 118, "lr": 9.966152641793132e-05}
{"train_loss": 0.6821103096008301, "global_step": 10541, "epoch": 118, "lr": 9.966145907245931e-05}
{"train_loss": 0.8530251979827881, "global_step": 10542, "epoch": 118, "lr": 9.966139172031093e-05}
{"train_loss": 0.6426806449890137, "global_step": 10543, "epoch": 118, "lr": 9.966132436148616e-05}
{"train_loss": 0.6661422848701477, "global_step": 10544, "epoch": 118, "lr": 9.966125699598505e-05}
{"train_loss": 0.7912588715553284, "global_step": 10545, "epoch": 118, "lr": 9.966118962380759e-05}
{"train_loss": 0.6556537747383118, "global_step": 10546, "epoch": 118, "lr": 9.966112224495379e-05}
{"train_loss": 0.5935795307159424, "global_step": 10547, "epoch": 118, "lr": 9.966105485942365e-05}
{"train_loss": 0.8417779207229614, "global_step": 10548, "epoch": 118, "lr": 9.966098746721718e-05}
{"train_loss": 0.7814615964889526, "global_step": 10549, "epoch": 118, "lr": 9.96609200683344e-05}
{"train_loss": 0.7515341639518738, "global_step": 10550, "epoch": 118, "lr": 9.966085266277531e-05}
{"train_loss": 0.546273946762085, "global_step": 10551, "epoch": 118, "lr": 9.966078525053994e-05}
{"train_loss": 0.8125999569892883, "global_step": 10552, "epoch": 118, "lr": 9.966071783162828e-05}
{"train_loss": 0.7509937882423401, "global_step": 10553, "epoch": 118, "lr": 9.966065040604034e-05}
{"train_loss": 0.690823495388031, "global_step": 10554, "epoch": 118, "lr": 9.966058297377612e-05}
{"train_loss": 0.6423608064651489, "global_step": 10555, "epoch": 118, "lr": 9.966051553483564e-05}
{"train_loss": 0.6713733673095703, "global_step": 10556, "epoch": 118, "lr": 9.966044808921893e-05}
{"train_loss": 0.6836985349655151, "global_step": 10557, "epoch": 118, "lr": 9.966038063692595e-05}
{"train_loss": 0.594963550567627, "global_step": 10558, "epoch": 118, "lr": 9.966031317795676e-05}
{"train_loss": 0.69467693567276, "global_step": 10559, "epoch": 118, "lr": 9.966024571231134e-05}
{"train_loss": 0.7625814080238342, "global_step": 10560, "epoch": 118, "lr": 9.96601782399897e-05}
{"train_loss": 0.7346522212028503, "global_step": 10561, "epoch": 118, "lr": 9.966011076099185e-05}
{"train_loss": 0.6850311756134033, "global_step": 10562, "epoch": 118, "lr": 9.966004327531781e-05}
{"train_loss": 0.7848143577575684, "global_step": 10563, "epoch": 118, "lr": 9.965997578296759e-05}
{"train_loss": 0.5955646634101868, "global_step": 10564, "epoch": 118, "lr": 9.965990828394118e-05}
{"train_loss": 0.7234103679656982, "global_step": 10565, "epoch": 118, "lr": 9.96598407782386e-05}
{"train_loss": 0.794288694858551, "global_step": 10566, "epoch": 118, "lr": 9.965977326585987e-05}
{"train_loss": 0.7800513505935669, "global_step": 10567, "epoch": 118, "lr": 9.965970574680498e-05}
{"train_loss": 0.6488687992095947, "global_step": 10568, "epoch": 118, "lr": 9.965963822107395e-05}
{"train_loss": 0.7944341897964478, "global_step": 10569, "epoch": 118, "lr": 9.965957068866678e-05}
{"train_loss": 0.7256301045417786, "global_step": 10570, "epoch": 118, "lr": 9.96595031495835e-05}
{"train_loss": 0.6856595873832703, "global_step": 10571, "epoch": 118, "lr": 9.965943560382409e-05}
{"train_loss": 0.5655572414398193, "global_step": 10572, "epoch": 118, "lr": 9.965936805138859e-05}
{"train_loss": 0.6382663249969482, "global_step": 10573, "epoch": 118, "lr": 9.965930049227699e-05}
{"train_loss": 0.809097409248352, "global_step": 10574, "epoch": 118, "lr": 9.965923292648928e-05}
{"train_loss": 0.5951929688453674, "global_step": 10575, "epoch": 118, "lr": 9.965916535402553e-05}
{"train_loss": 0.6246986389160156, "global_step": 10576, "epoch": 118, "lr": 9.965909777488568e-05}
{"train_loss": 0.6642277836799622, "global_step": 10577, "epoch": 118, "lr": 9.965903018906979e-05}
{"train_loss": 0.665176272392273, "global_step": 10578, "epoch": 118, "lr": 9.965896259657783e-05}
{"train_loss": 0.6570878624916077, "global_step": 10579, "epoch": 118, "lr": 9.965889499740983e-05}
{"train_loss": 0.7639142274856567, "global_step": 10580, "epoch": 118, "lr": 9.965882739156581e-05}
{"train_loss": 0.6889165639877319, "global_step": 10581, "epoch": 118, "lr": 9.965875977904576e-05}
{"train_loss": 0.6293798089027405, "global_step": 10582, "epoch": 118, "lr": 9.965869215984969e-05}
{"train_loss": 0.6808079481124878, "global_step": 10583, "epoch": 118, "lr": 9.965862453397763e-05}
{"train_loss": 0.6425641775131226, "global_step": 10584, "epoch": 118, "lr": 9.965855690142955e-05}
{"train_loss": 0.5079109072685242, "global_step": 10585, "epoch": 118, "lr": 9.965848926220548e-05}
{"train_loss": 0.7583482265472412, "global_step": 10586, "epoch": 118, "lr": 9.965842161630547e-05}
{"train_loss": 0.7327605485916138, "global_step": 10587, "epoch": 118, "lr": 9.965835396372946e-05}
{"train_loss": 0.594227135181427, "global_step": 10588, "epoch": 118, "lr": 9.96582863044775e-05}
{"train_loss": 0.7649718523025513, "global_step": 10589, "epoch": 118, "lr": 9.965821863854957e-05}
{"train_loss": 0.6885181799363554, "global_step": 10590, "epoch": 118, "lr": 9.965815096594572e-05, "val_loss": 1.2950736284255981}
{"train_loss": 0.7950255870819092, "global_step": 10591, "epoch": 119, "lr": 9.965808328666592e-05}
{"train_loss": 0.6357056498527527, "global_step": 10592, "epoch": 119, "lr": 9.96580156007102e-05}
{"train_loss": 0.5102607011795044, "global_step": 10593, "epoch": 119, "lr": 9.965794790807858e-05}
{"train_loss": 0.7612488865852356, "global_step": 10594, "epoch": 119, "lr": 9.965788020877103e-05}
{"train_loss": 0.7774280309677124, "global_step": 10595, "epoch": 119, "lr": 9.965781250278761e-05}
{"train_loss": 0.6553167700767517, "global_step": 10596, "epoch": 119, "lr": 9.965774479012828e-05}
{"train_loss": 0.5046221017837524, "global_step": 10597, "epoch": 119, "lr": 9.965767707079309e-05}
{"train_loss": 0.6859986782073975, "global_step": 10598, "epoch": 119, "lr": 9.965760934478201e-05}
{"train_loss": 0.7336232662200928, "global_step": 10599, "epoch": 119, "lr": 9.965754161209508e-05}
{"train_loss": 0.731138288974762, "global_step": 10600, "epoch": 119, "lr": 9.96574738727323e-05}
{"train_loss": 0.7055082321166992, "global_step": 10601, "epoch": 119, "lr": 9.965740612669367e-05}
{"train_loss": 0.7635181546211243, "global_step": 10602, "epoch": 119, "lr": 9.965733837397922e-05}
{"train_loss": 0.6921131014823914, "global_step": 10603, "epoch": 119, "lr": 9.965727061458895e-05}
{"train_loss": 0.6803684234619141, "global_step": 10604, "epoch": 119, "lr": 9.965720284852285e-05}
{"train_loss": 0.6935881972312927, "global_step": 10605, "epoch": 119, "lr": 9.965713507578095e-05}
{"train_loss": 0.4813753664493561, "global_step": 10606, "epoch": 119, "lr": 9.965706729636326e-05}
{"train_loss": 0.6894109845161438, "global_step": 10607, "epoch": 119, "lr": 9.965699951026977e-05}
{"train_loss": 0.6978951096534729, "global_step": 10608, "epoch": 119, "lr": 9.96569317175005e-05}
{"train_loss": 0.6726779937744141, "global_step": 10609, "epoch": 119, "lr": 9.965686391805547e-05}
{"train_loss": 0.7407062649726868, "global_step": 10610, "epoch": 119, "lr": 9.965679611193469e-05}
{"train_loss": 0.8475915789604187, "global_step": 10611, "epoch": 119, "lr": 9.965672829913814e-05}
{"train_loss": 0.773711085319519, "global_step": 10612, "epoch": 119, "lr": 9.965666047966586e-05}
{"train_loss": 0.6781826019287109, "global_step": 10613, "epoch": 119, "lr": 9.965659265351784e-05}
{"train_loss": 0.7139742970466614, "global_step": 10614, "epoch": 119, "lr": 9.96565248206941e-05}
{"train_loss": 0.6984202265739441, "global_step": 10615, "epoch": 119, "lr": 9.965645698119463e-05}
{"train_loss": 0.8095195293426514, "global_step": 10616, "epoch": 119, "lr": 9.965638913501948e-05}
{"train_loss": 0.7297875881195068, "global_step": 10617, "epoch": 119, "lr": 9.965632128216862e-05}
{"train_loss": 0.6943225860595703, "global_step": 10618, "epoch": 119, "lr": 9.965625342264208e-05}
{"train_loss": 0.8273226618766785, "global_step": 10619, "epoch": 119, "lr": 9.965618555643985e-05}
{"train_loss": 0.6804970502853394, "global_step": 10620, "epoch": 119, "lr": 9.965611768356196e-05}
{"train_loss": 0.7095393538475037, "global_step": 10621, "epoch": 119, "lr": 9.96560498040084e-05}
{"train_loss": 0.6094583868980408, "global_step": 10622, "epoch": 119, "lr": 9.96559819177792e-05}
{"train_loss": 0.6315032243728638, "global_step": 10623, "epoch": 119, "lr": 9.965591402487436e-05}
{"train_loss": 0.6978809237480164, "global_step": 10624, "epoch": 119, "lr": 9.965584612529389e-05}
{"train_loss": 0.6029742360115051, "global_step": 10625, "epoch": 119, "lr": 9.965577821903779e-05}
{"train_loss": 0.7669957280158997, "global_step": 10626, "epoch": 119, "lr": 9.965571030610608e-05}
{"train_loss": 0.6604978442192078, "global_step": 10627, "epoch": 119, "lr": 9.965564238649877e-05}
{"train_loss": 0.5753737688064575, "global_step": 10628, "epoch": 119, "lr": 9.965557446021584e-05}
{"train_loss": 0.7275506854057312, "global_step": 10629, "epoch": 119, "lr": 9.965550652725735e-05}
{"train_loss": 0.7759514451026917, "global_step": 10630, "epoch": 119, "lr": 9.965543858762326e-05}
{"train_loss": 0.6298348307609558, "global_step": 10631, "epoch": 119, "lr": 9.96553706413136e-05}
{"train_loss": 0.6959466934204102, "global_step": 10632, "epoch": 119, "lr": 9.965530268832842e-05}
{"train_loss": 0.7754159569740295, "global_step": 10633, "epoch": 119, "lr": 9.965523472866765e-05}
{"train_loss": 0.7368924021720886, "global_step": 10634, "epoch": 119, "lr": 9.965516676233135e-05}
{"train_loss": 0.7617198824882507, "global_step": 10635, "epoch": 119, "lr": 9.965509878931952e-05}
{"train_loss": 0.7805148959159851, "global_step": 10636, "epoch": 119, "lr": 9.965503080963217e-05}
{"train_loss": 0.6172661781311035, "global_step": 10637, "epoch": 119, "lr": 9.965496282326929e-05}
{"train_loss": 0.6299429535865784, "global_step": 10638, "epoch": 119, "lr": 9.965489483023094e-05}
{"train_loss": 0.7143797278404236, "global_step": 10639, "epoch": 119, "lr": 9.965482683051706e-05}
{"train_loss": 0.7210698127746582, "global_step": 10640, "epoch": 119, "lr": 9.965475882412771e-05}
{"train_loss": 0.7752113342285156, "global_step": 10641, "epoch": 119, "lr": 9.965469081106287e-05}
{"train_loss": 0.6136815547943115, "global_step": 10642, "epoch": 119, "lr": 9.965462279132259e-05}
{"train_loss": 0.53244948387146, "global_step": 10643, "epoch": 119, "lr": 9.965455476490683e-05}
{"train_loss": 0.6859714984893799, "global_step": 10644, "epoch": 119, "lr": 9.965448673181562e-05}
{"train_loss": 0.7407220602035522, "global_step": 10645, "epoch": 119, "lr": 9.965441869204897e-05}
{"train_loss": 0.6924945712089539, "global_step": 10646, "epoch": 119, "lr": 9.965435064560689e-05}
{"train_loss": 0.7364758253097534, "global_step": 10647, "epoch": 119, "lr": 9.965428259248939e-05}
{"train_loss": 0.6697295904159546, "global_step": 10648, "epoch": 119, "lr": 9.965421453269648e-05}
{"train_loss": 0.7063920497894287, "global_step": 10649, "epoch": 119, "lr": 9.965414646622816e-05}
{"train_loss": 0.6698770523071289, "global_step": 10650, "epoch": 119, "lr": 9.965407839308446e-05}
{"train_loss": 0.7632158398628235, "global_step": 10651, "epoch": 119, "lr": 9.965401031326536e-05}
{"train_loss": 0.6561623215675354, "global_step": 10652, "epoch": 119, "lr": 9.965394222677087e-05}
{"train_loss": 0.6643021106719971, "global_step": 10653, "epoch": 119, "lr": 9.965387413360104e-05}
{"train_loss": 0.7776936292648315, "global_step": 10654, "epoch": 119, "lr": 9.965380603375584e-05}
{"train_loss": 0.619967520236969, "global_step": 10655, "epoch": 119, "lr": 9.96537379272353e-05}
{"train_loss": 0.7827395796775818, "global_step": 10656, "epoch": 119, "lr": 9.96536698140394e-05}
{"train_loss": 0.6264614462852478, "global_step": 10657, "epoch": 119, "lr": 9.965360169416818e-05}
{"train_loss": 0.7331748604774475, "global_step": 10658, "epoch": 119, "lr": 9.965353356762163e-05}
{"train_loss": 0.6976079344749451, "global_step": 10659, "epoch": 119, "lr": 9.965346543439978e-05}
{"train_loss": 0.7431480884552002, "global_step": 10660, "epoch": 119, "lr": 9.965339729450262e-05}
{"train_loss": 0.7336548566818237, "global_step": 10661, "epoch": 119, "lr": 9.965332914793017e-05}
{"train_loss": 0.6553416848182678, "global_step": 10662, "epoch": 119, "lr": 9.965326099468244e-05}
{"train_loss": 0.8419260382652283, "global_step": 10663, "epoch": 119, "lr": 9.965319283475942e-05}
{"train_loss": 0.664448618888855, "global_step": 10664, "epoch": 119, "lr": 9.965312466816115e-05}
{"train_loss": 0.7377392649650574, "global_step": 10665, "epoch": 119, "lr": 9.965305649488761e-05}
{"train_loss": 0.8142870664596558, "global_step": 10666, "epoch": 119, "lr": 9.965298831493883e-05}
{"train_loss": 0.9327604174613953, "global_step": 10667, "epoch": 119, "lr": 9.96529201283148e-05}
{"train_loss": 0.637121856212616, "global_step": 10668, "epoch": 119, "lr": 9.965285193501555e-05}
{"train_loss": 0.5507504343986511, "global_step": 10669, "epoch": 119, "lr": 9.965278373504107e-05}
{"train_loss": 0.6804465055465698, "global_step": 10670, "epoch": 119, "lr": 9.96527155283914e-05}
{"train_loss": 0.670632004737854, "global_step": 10671, "epoch": 119, "lr": 9.96526473150665e-05}
{"train_loss": 0.546380877494812, "global_step": 10672, "epoch": 119, "lr": 9.965257909506643e-05}
{"train_loss": 0.7550057768821716, "global_step": 10673, "epoch": 119, "lr": 9.965251086839116e-05}
{"train_loss": 0.6766795516014099, "global_step": 10674, "epoch": 119, "lr": 9.965244263504072e-05}
{"train_loss": 0.6593667268753052, "global_step": 10675, "epoch": 119, "lr": 9.965237439501511e-05}
{"train_loss": 0.7126147747039795, "global_step": 10676, "epoch": 119, "lr": 9.965230614831436e-05}
{"train_loss": 0.6476479172706604, "global_step": 10677, "epoch": 119, "lr": 9.965223789493844e-05}
{"train_loss": 0.5794130563735962, "global_step": 10678, "epoch": 119, "lr": 9.96521696348874e-05}
{"train_loss": 0.6967002424631226, "global_step": 10679, "epoch": 119, "lr": 9.965210136816121e-05, "val_loss": 1.2719147205352783}
{"train_loss": 0.7196394205093384, "global_step": 10680, "epoch": 120, "lr": 9.965203309475992e-05}
{"train_loss": 0.6092426776885986, "global_step": 10681, "epoch": 120, "lr": 9.965196481468351e-05}
{"train_loss": 0.7138580679893494, "global_step": 10682, "epoch": 120, "lr": 9.965189652793201e-05}
{"train_loss": 0.5915718078613281, "global_step": 10683, "epoch": 120, "lr": 9.96518282345054e-05}
{"train_loss": 0.779567539691925, "global_step": 10684, "epoch": 120, "lr": 9.965175993440372e-05}
{"train_loss": 0.60845947265625, "global_step": 10685, "epoch": 120, "lr": 9.965169162762695e-05}
{"train_loss": 0.7249768972396851, "global_step": 10686, "epoch": 120, "lr": 9.965162331417513e-05}
{"train_loss": 0.45261120796203613, "global_step": 10687, "epoch": 120, "lr": 9.965155499404826e-05}
{"train_loss": 0.6563916206359863, "global_step": 10688, "epoch": 120, "lr": 9.965148666724634e-05}
{"train_loss": 0.5889315009117126, "global_step": 10689, "epoch": 120, "lr": 9.965141833376937e-05}
{"train_loss": 0.7088413238525391, "global_step": 10690, "epoch": 120, "lr": 9.965134999361738e-05}
{"train_loss": 0.61029452085495, "global_step": 10691, "epoch": 120, "lr": 9.965128164679036e-05}
{"train_loss": 0.6003895998001099, "global_step": 10692, "epoch": 120, "lr": 9.965121329328834e-05}
{"train_loss": 0.6629101037979126, "global_step": 10693, "epoch": 120, "lr": 9.965114493311134e-05}
{"train_loss": 0.7722342014312744, "global_step": 10694, "epoch": 120, "lr": 9.965107656625931e-05}
{"train_loss": 0.6206198334693909, "global_step": 10695, "epoch": 120, "lr": 9.965100819273233e-05}
{"train_loss": 0.6874853372573853, "global_step": 10696, "epoch": 120, "lr": 9.965093981253037e-05}
{"train_loss": 0.5254780054092407, "global_step": 10697, "epoch": 120, "lr": 9.965087142565344e-05}
{"train_loss": 0.7536416053771973, "global_step": 10698, "epoch": 120, "lr": 9.965080303210155e-05}
{"train_loss": 0.8264949321746826, "global_step": 10699, "epoch": 120, "lr": 9.965073463187472e-05}
{"train_loss": 0.686294674873352, "global_step": 10700, "epoch": 120, "lr": 9.965066622497296e-05}
{"train_loss": 0.6603274345397949, "global_step": 10701, "epoch": 120, "lr": 9.965059781139626e-05}
{"train_loss": 0.6539252996444702, "global_step": 10702, "epoch": 120, "lr": 9.965052939114466e-05}
{"train_loss": 0.734810471534729, "global_step": 10703, "epoch": 120, "lr": 9.965046096421813e-05}
{"train_loss": 0.8119586110115051, "global_step": 10704, "epoch": 120, "lr": 9.965039253061672e-05}
{"train_loss": 0.6566022038459778, "global_step": 10705, "epoch": 120, "lr": 9.96503240903404e-05}
{"train_loss": 0.5955581068992615, "global_step": 10706, "epoch": 120, "lr": 9.965025564338923e-05}
{"train_loss": 0.6568793058395386, "global_step": 10707, "epoch": 120, "lr": 9.965018718976316e-05}
{"train_loss": 0.8105635643005371, "global_step": 10708, "epoch": 120, "lr": 9.965011872946224e-05}
{"train_loss": 0.6654013991355896, "global_step": 10709, "epoch": 120, "lr": 9.965005026248646e-05}
{"train_loss": 0.5687276721000671, "global_step": 10710, "epoch": 120, "lr": 9.964998178883584e-05}
{"train_loss": 0.6442670822143555, "global_step": 10711, "epoch": 120, "lr": 9.964991330851038e-05}
{"train_loss": 0.8021827340126038, "global_step": 10712, "epoch": 120, "lr": 9.96498448215101e-05}
{"train_loss": 0.6004486083984375, "global_step": 10713, "epoch": 120, "lr": 9.9649776327835e-05}
{"train_loss": 0.716454267501831, "global_step": 10714, "epoch": 120, "lr": 9.964970782748511e-05}
{"train_loss": 0.7779486775398254, "global_step": 10715, "epoch": 120, "lr": 9.96496393204604e-05}
{"train_loss": 0.7035472989082336, "global_step": 10716, "epoch": 120, "lr": 9.964957080676091e-05}
{"train_loss": 0.8089354038238525, "global_step": 10717, "epoch": 120, "lr": 9.964950228638663e-05}
{"train_loss": 0.6701349020004272, "global_step": 10718, "epoch": 120, "lr": 9.964943375933759e-05}
{"train_loss": 0.6881375908851624, "global_step": 10719, "epoch": 120, "lr": 9.964936522561378e-05}
{"train_loss": 0.7616503834724426, "global_step": 10720, "epoch": 120, "lr": 9.964929668521523e-05}
{"train_loss": 0.7528814077377319, "global_step": 10721, "epoch": 120, "lr": 9.964922813814194e-05}
{"train_loss": 0.6254639625549316, "global_step": 10722, "epoch": 120, "lr": 9.964915958439391e-05}
{"train_loss": 0.6110134720802307, "global_step": 10723, "epoch": 120, "lr": 9.964909102397115e-05}
{"train_loss": 0.7760170698165894, "global_step": 10724, "epoch": 120, "lr": 9.964902245687369e-05}
{"train_loss": 0.6091102361679077, "global_step": 10725, "epoch": 120, "lr": 9.96489538831015e-05}
{"train_loss": 0.6912473440170288, "global_step": 10726, "epoch": 120, "lr": 9.964888530265464e-05}
{"train_loss": 0.6545567512512207, "global_step": 10727, "epoch": 120, "lr": 9.964881671553307e-05}
{"train_loss": 0.5790978670120239, "global_step": 10728, "epoch": 120, "lr": 9.964874812173683e-05}
{"train_loss": 0.7185426354408264, "global_step": 10729, "epoch": 120, "lr": 9.964867952126594e-05}
{"train_loss": 0.642167866230011, "global_step": 10730, "epoch": 120, "lr": 9.964861091412035e-05}
{"train_loss": 0.6706821322441101, "global_step": 10731, "epoch": 120, "lr": 9.964854230030014e-05}
{"train_loss": 0.8617066740989685, "global_step": 10732, "epoch": 120, "lr": 9.964847367980528e-05}
{"train_loss": 0.696387767791748, "global_step": 10733, "epoch": 120, "lr": 9.964840505263578e-05}
{"train_loss": 0.828372597694397, "global_step": 10734, "epoch": 120, "lr": 9.964833641879166e-05}
{"train_loss": 0.7914270162582397, "global_step": 10735, "epoch": 120, "lr": 9.964826777827294e-05}
{"train_loss": 0.9286838173866272, "global_step": 10736, "epoch": 120, "lr": 9.96481991310796e-05}
{"train_loss": 0.632889986038208, "global_step": 10737, "epoch": 120, "lr": 9.964813047721168e-05}
{"train_loss": 0.8157212734222412, "global_step": 10738, "epoch": 120, "lr": 9.964806181666917e-05}
{"train_loss": 0.6976355314254761, "global_step": 10739, "epoch": 120, "lr": 9.964799314945209e-05}
{"train_loss": 0.7193888425827026, "global_step": 10740, "epoch": 120, "lr": 9.964792447556041e-05}
{"train_loss": 0.7402093410491943, "global_step": 10741, "epoch": 120, "lr": 9.96478557949942e-05}
{"train_loss": 0.5769538283348083, "global_step": 10742, "epoch": 120, "lr": 9.964778710775342e-05}
{"train_loss": 0.6035509705543518, "global_step": 10743, "epoch": 120, "lr": 9.964771841383812e-05}
{"train_loss": 0.7050262689590454, "global_step": 10744, "epoch": 120, "lr": 9.964764971324829e-05}
{"train_loss": 0.7096670866012573, "global_step": 10745, "epoch": 120, "lr": 9.964758100598393e-05}
{"train_loss": 0.7740334272384644, "global_step": 10746, "epoch": 120, "lr": 9.964751229204505e-05}
{"train_loss": 0.5615909099578857, "global_step": 10747, "epoch": 120, "lr": 9.964744357143168e-05}
{"train_loss": 0.7192221879959106, "global_step": 10748, "epoch": 120, "lr": 9.964737484414382e-05}
{"train_loss": 0.6491637825965881, "global_step": 10749, "epoch": 120, "lr": 9.964730611018148e-05}
{"train_loss": 0.6588269472122192, "global_step": 10750, "epoch": 120, "lr": 9.964723736954465e-05}
{"train_loss": 0.6407310366630554, "global_step": 10751, "epoch": 120, "lr": 9.964716862223334e-05}
{"train_loss": 0.6958118677139282, "global_step": 10752, "epoch": 120, "lr": 9.96470998682476e-05}
{"train_loss": 0.7554238438606262, "global_step": 10753, "epoch": 120, "lr": 9.96470311075874e-05}
{"train_loss": 0.6604204773902893, "global_step": 10754, "epoch": 120, "lr": 9.964696234025278e-05}
{"train_loss": 0.7297891974449158, "global_step": 10755, "epoch": 120, "lr": 9.96468935662437e-05}
{"train_loss": 0.6614680290222168, "global_step": 10756, "epoch": 120, "lr": 9.964682478556022e-05}
{"train_loss": 0.8996937274932861, "global_step": 10757, "epoch": 120, "lr": 9.964675599820233e-05}
{"train_loss": 0.9630016684532166, "global_step": 10758, "epoch": 120, "lr": 9.964668720417003e-05}
{"train_loss": 0.6730930209159851, "global_step": 10759, "epoch": 120, "lr": 9.964661840346334e-05}
{"train_loss": 0.6762052178382874, "global_step": 10760, "epoch": 120, "lr": 9.964654959608225e-05}
{"train_loss": 0.6599820256233215, "global_step": 10761, "epoch": 120, "lr": 9.964648078202684e-05}
{"train_loss": 0.6890706419944763, "global_step": 10762, "epoch": 120, "lr": 9.964641196129703e-05}
{"train_loss": 0.7008089423179626, "global_step": 10763, "epoch": 120, "lr": 9.964634313389286e-05}
{"train_loss": 0.8190984129905701, "global_step": 10764, "epoch": 120, "lr": 9.964627429981435e-05}
{"train_loss": 0.770740807056427, "global_step": 10765, "epoch": 120, "lr": 9.964620545906151e-05}
{"train_loss": 0.7748386263847351, "global_step": 10766, "epoch": 120, "lr": 9.964613661163433e-05}
{"train_loss": 0.6622231006622314, "global_step": 10767, "epoch": 120, "lr": 9.964606775753285e-05}
{"train_loss": 0.6988404975848251, "global_step": 10768, "epoch": 120, "lr": 9.964599889675704e-05, "val_loss": 1.2790998220443726, "train_action_mse_error": 45.245445251464844}
{"train_loss": 0.7144033312797546, "global_step": 10769, "epoch": 121, "lr": 9.964593002930693e-05}
{"train_loss": 0.5792255401611328, "global_step": 10770, "epoch": 121, "lr": 9.964586115518255e-05}
{"train_loss": 0.653240442276001, "global_step": 10771, "epoch": 121, "lr": 9.964579227438389e-05}
{"train_loss": 0.5587618350982666, "global_step": 10772, "epoch": 121, "lr": 9.964572338691095e-05}
{"train_loss": 0.6067857146263123, "global_step": 10773, "epoch": 121, "lr": 9.964565449276374e-05}
{"train_loss": 0.7086561322212219, "global_step": 10774, "epoch": 121, "lr": 9.964558559194229e-05}
{"train_loss": 0.8868255615234375, "global_step": 10775, "epoch": 121, "lr": 9.964551668444658e-05}
{"train_loss": 0.7670465111732483, "global_step": 10776, "epoch": 121, "lr": 9.964544777027664e-05}
{"train_loss": 0.8317384719848633, "global_step": 10777, "epoch": 121, "lr": 9.964537884943248e-05}
{"train_loss": 0.7968501448631287, "global_step": 10778, "epoch": 121, "lr": 9.964530992191411e-05}
{"train_loss": 0.5467832088470459, "global_step": 10779, "epoch": 121, "lr": 9.964524098772152e-05}
{"train_loss": 0.8469383716583252, "global_step": 10780, "epoch": 121, "lr": 9.964517204685475e-05}
{"train_loss": 0.844353199005127, "global_step": 10781, "epoch": 121, "lr": 9.964510309931378e-05}
{"train_loss": 0.7260878086090088, "global_step": 10782, "epoch": 121, "lr": 9.964503414509863e-05}
{"train_loss": 0.7765481472015381, "global_step": 10783, "epoch": 121, "lr": 9.964496518420931e-05}
{"train_loss": 0.7484267354011536, "global_step": 10784, "epoch": 121, "lr": 9.964489621664583e-05}
{"train_loss": 0.9037143588066101, "global_step": 10785, "epoch": 121, "lr": 9.96448272424082e-05}
{"train_loss": 0.5733880400657654, "global_step": 10786, "epoch": 121, "lr": 9.964475826149644e-05}
{"train_loss": 0.7367600202560425, "global_step": 10787, "epoch": 121, "lr": 9.964468927391053e-05}
{"train_loss": 0.7483320236206055, "global_step": 10788, "epoch": 121, "lr": 9.964462027965049e-05}
{"train_loss": 0.644825279712677, "global_step": 10789, "epoch": 121, "lr": 9.964455127871634e-05}
{"train_loss": 0.7600263357162476, "global_step": 10790, "epoch": 121, "lr": 9.964448227110811e-05}
{"train_loss": 0.5677518844604492, "global_step": 10791, "epoch": 121, "lr": 9.964441325682577e-05}
{"train_loss": 0.6981019973754883, "global_step": 10792, "epoch": 121, "lr": 9.964434423586934e-05}
{"train_loss": 0.7462384104728699, "global_step": 10793, "epoch": 121, "lr": 9.964427520823883e-05}
{"train_loss": 0.7483291029930115, "global_step": 10794, "epoch": 121, "lr": 9.964420617393426e-05}
{"train_loss": 0.7048729062080383, "global_step": 10795, "epoch": 121, "lr": 9.964413713295562e-05}
{"train_loss": 0.7187053561210632, "global_step": 10796, "epoch": 121, "lr": 9.964406808530295e-05}
{"train_loss": 0.7598299384117126, "global_step": 10797, "epoch": 121, "lr": 9.964399903097623e-05}
{"train_loss": 0.6615089178085327, "global_step": 10798, "epoch": 121, "lr": 9.964392996997547e-05}
{"train_loss": 0.7089442014694214, "global_step": 10799, "epoch": 121, "lr": 9.96438609023007e-05}
{"train_loss": 0.6584461331367493, "global_step": 10800, "epoch": 121, "lr": 9.964379182795192e-05}
{"train_loss": 0.7935910224914551, "global_step": 10801, "epoch": 121, "lr": 9.964372274692913e-05}
{"train_loss": 0.6795843243598938, "global_step": 10802, "epoch": 121, "lr": 9.964365365923234e-05}
{"train_loss": 0.7085644006729126, "global_step": 10803, "epoch": 121, "lr": 9.964358456486159e-05}
{"train_loss": 0.7165262699127197, "global_step": 10804, "epoch": 121, "lr": 9.964351546381683e-05}
{"train_loss": 0.8983291983604431, "global_step": 10805, "epoch": 121, "lr": 9.964344635609814e-05}
{"train_loss": 1.0363303422927856, "global_step": 10806, "epoch": 121, "lr": 9.964337724170547e-05}
{"train_loss": 0.6536831855773926, "global_step": 10807, "epoch": 121, "lr": 9.964330812063886e-05}
{"train_loss": 0.748382031917572, "global_step": 10808, "epoch": 121, "lr": 9.964323899289831e-05}
{"train_loss": 0.7106801867485046, "global_step": 10809, "epoch": 121, "lr": 9.964316985848384e-05}
{"train_loss": 0.7452638745307922, "global_step": 10810, "epoch": 121, "lr": 9.964310071739544e-05}
{"train_loss": 0.7928896546363831, "global_step": 10811, "epoch": 121, "lr": 9.964303156963315e-05}
{"train_loss": 0.6617032885551453, "global_step": 10812, "epoch": 121, "lr": 9.964296241519694e-05}
{"train_loss": 0.5671296119689941, "global_step": 10813, "epoch": 121, "lr": 9.964289325408685e-05}
{"train_loss": 0.8158063292503357, "global_step": 10814, "epoch": 121, "lr": 9.964282408630287e-05}
{"train_loss": 0.6877791881561279, "global_step": 10815, "epoch": 121, "lr": 9.964275491184501e-05}
{"train_loss": 0.599387526512146, "global_step": 10816, "epoch": 121, "lr": 9.964268573071331e-05}
{"train_loss": 0.7495360970497131, "global_step": 10817, "epoch": 121, "lr": 9.964261654290774e-05}
{"train_loss": 0.6967697739601135, "global_step": 10818, "epoch": 121, "lr": 9.964254734842833e-05}
{"train_loss": 0.6911360025405884, "global_step": 10819, "epoch": 121, "lr": 9.964247814727508e-05}
{"train_loss": 0.8387555480003357, "global_step": 10820, "epoch": 121, "lr": 9.964240893944801e-05}
{"train_loss": 0.6781688928604126, "global_step": 10821, "epoch": 121, "lr": 9.964233972494711e-05}
{"train_loss": 0.7319958806037903, "global_step": 10822, "epoch": 121, "lr": 9.964227050377242e-05}
{"train_loss": 0.6661384105682373, "global_step": 10823, "epoch": 121, "lr": 9.964220127592391e-05}
{"train_loss": 0.4989258944988251, "global_step": 10824, "epoch": 121, "lr": 9.964213204140162e-05}
{"train_loss": 0.6706059575080872, "global_step": 10825, "epoch": 121, "lr": 9.964206280020557e-05}
{"train_loss": 0.6737560033798218, "global_step": 10826, "epoch": 121, "lr": 9.964199355233572e-05}
{"train_loss": 0.6335698962211609, "global_step": 10827, "epoch": 121, "lr": 9.964192429779213e-05}
{"train_loss": 0.952585756778717, "global_step": 10828, "epoch": 121, "lr": 9.964185503657478e-05}
{"train_loss": 0.6872064471244812, "global_step": 10829, "epoch": 121, "lr": 9.96417857686837e-05}
{"train_loss": 0.7197384238243103, "global_step": 10830, "epoch": 121, "lr": 9.964171649411887e-05}
{"train_loss": 0.8513546586036682, "global_step": 10831, "epoch": 121, "lr": 9.964164721288031e-05}
{"train_loss": 0.6951566338539124, "global_step": 10832, "epoch": 121, "lr": 9.964157792496807e-05}
{"train_loss": 0.6945584416389465, "global_step": 10833, "epoch": 121, "lr": 9.964150863038209e-05}
{"train_loss": 0.6220636367797852, "global_step": 10834, "epoch": 121, "lr": 9.964143932912243e-05}
{"train_loss": 0.6284047365188599, "global_step": 10835, "epoch": 121, "lr": 9.964137002118908e-05}
{"train_loss": 0.5874768495559692, "global_step": 10836, "epoch": 121, "lr": 9.964130070658206e-05}
{"train_loss": 0.6343162655830383, "global_step": 10837, "epoch": 121, "lr": 9.964123138530137e-05}
{"train_loss": 0.8020725250244141, "global_step": 10838, "epoch": 121, "lr": 9.964116205734701e-05}
{"train_loss": 0.7391791939735413, "global_step": 10839, "epoch": 121, "lr": 9.9641092722719e-05}
{"train_loss": 0.7393908500671387, "global_step": 10840, "epoch": 121, "lr": 9.964102338141737e-05}
{"train_loss": 0.6147078275680542, "global_step": 10841, "epoch": 121, "lr": 9.964095403344209e-05}
{"train_loss": 0.5833323001861572, "global_step": 10842, "epoch": 121, "lr": 9.96408846787932e-05}
{"train_loss": 0.7502529621124268, "global_step": 10843, "epoch": 121, "lr": 9.96408153174707e-05}
{"train_loss": 0.7225858569145203, "global_step": 10844, "epoch": 121, "lr": 9.96407459494746e-05}
{"train_loss": 0.5553072690963745, "global_step": 10845, "epoch": 121, "lr": 9.964067657480489e-05}
{"train_loss": 0.7236201167106628, "global_step": 10846, "epoch": 121, "lr": 9.96406071934616e-05}
{"train_loss": 0.7364532947540283, "global_step": 10847, "epoch": 121, "lr": 9.964053780544475e-05}
{"train_loss": 0.5815926194190979, "global_step": 10848, "epoch": 121, "lr": 9.964046841075432e-05}
{"train_loss": 0.6139230132102966, "global_step": 10849, "epoch": 121, "lr": 9.964039900939034e-05}
{"train_loss": 0.7350277900695801, "global_step": 10850, "epoch": 121, "lr": 9.96403296013528e-05}
{"train_loss": 0.5705912113189697, "global_step": 10851, "epoch": 121, "lr": 9.964026018664174e-05}
{"train_loss": 0.5928810834884644, "global_step": 10852, "epoch": 121, "lr": 9.964019076525715e-05}
{"train_loss": 0.7589853405952454, "global_step": 10853, "epoch": 121, "lr": 9.964012133719904e-05}
{"train_loss": 0.7732674479484558, "global_step": 10854, "epoch": 121, "lr": 9.964005190246742e-05}
{"train_loss": 0.7483879923820496, "global_step": 10855, "epoch": 121, "lr": 9.963998246106229e-05}
{"train_loss": 0.6072065234184265, "global_step": 10856, "epoch": 121, "lr": 9.963991301298368e-05}
{"train_loss": 0.7064292347163297, "global_step": 10857, "epoch": 121, "lr": 9.96398435582316e-05, "val_loss": 1.2578670978546143}
{"train_loss": 0.6399749517440796, "global_step": 10858, "epoch": 122, "lr": 9.963977409680602e-05}
{"train_loss": 0.7420493960380554, "global_step": 10859, "epoch": 122, "lr": 9.9639704628707e-05}
{"train_loss": 0.5874045491218567, "global_step": 10860, "epoch": 122, "lr": 9.963963515393453e-05}
{"train_loss": 0.7220643162727356, "global_step": 10861, "epoch": 122, "lr": 9.96395656724886e-05}
{"train_loss": 0.7167004346847534, "global_step": 10862, "epoch": 122, "lr": 9.963949618436924e-05}
{"train_loss": 0.6659907102584839, "global_step": 10863, "epoch": 122, "lr": 9.963942668957647e-05}
{"train_loss": 0.5950055122375488, "global_step": 10864, "epoch": 122, "lr": 9.963935718811027e-05}
{"train_loss": 0.6648338437080383, "global_step": 10865, "epoch": 122, "lr": 9.963928767997065e-05}
{"train_loss": 0.7805017828941345, "global_step": 10866, "epoch": 122, "lr": 9.963921816515764e-05}
{"train_loss": 0.6950528621673584, "global_step": 10867, "epoch": 122, "lr": 9.963914864367125e-05}
{"train_loss": 0.6531437039375305, "global_step": 10868, "epoch": 122, "lr": 9.963907911551149e-05}
{"train_loss": 0.7562102675437927, "global_step": 10869, "epoch": 122, "lr": 9.963900958067836e-05}
{"train_loss": 0.7094093561172485, "global_step": 10870, "epoch": 122, "lr": 9.963894003917184e-05}
{"train_loss": 0.7108129262924194, "global_step": 10871, "epoch": 122, "lr": 9.9638870490992e-05}
{"train_loss": 0.5348573327064514, "global_step": 10872, "epoch": 122, "lr": 9.96388009361388e-05}
{"train_loss": 0.63230299949646, "global_step": 10873, "epoch": 122, "lr": 9.963873137461228e-05}
{"train_loss": 0.6073692440986633, "global_step": 10874, "epoch": 122, "lr": 9.963866180641243e-05}
{"train_loss": 0.7000369429588318, "global_step": 10875, "epoch": 122, "lr": 9.963859223153928e-05}
{"train_loss": 0.6900927424430847, "global_step": 10876, "epoch": 122, "lr": 9.963852264999282e-05}
{"train_loss": 0.7899279594421387, "global_step": 10877, "epoch": 122, "lr": 9.963845306177306e-05}
{"train_loss": 0.6589581370353699, "global_step": 10878, "epoch": 122, "lr": 9.963838346688003e-05}
{"train_loss": 0.7491774559020996, "global_step": 10879, "epoch": 122, "lr": 9.96383138653137e-05}
{"train_loss": 0.6778151988983154, "global_step": 10880, "epoch": 122, "lr": 9.963824425707413e-05}
{"train_loss": 0.7481112480163574, "global_step": 10881, "epoch": 122, "lr": 9.963817464216128e-05}
{"train_loss": 0.7789332866668701, "global_step": 10882, "epoch": 122, "lr": 9.96381050205752e-05}
{"train_loss": 0.6638073325157166, "global_step": 10883, "epoch": 122, "lr": 9.963803539231587e-05}
{"train_loss": 0.7086010575294495, "global_step": 10884, "epoch": 122, "lr": 9.963796575738331e-05}
{"train_loss": 0.5597688555717468, "global_step": 10885, "epoch": 122, "lr": 9.963789611577754e-05}
{"train_loss": 0.7236362099647522, "global_step": 10886, "epoch": 122, "lr": 9.963782646749856e-05}
{"train_loss": 0.5810811519622803, "global_step": 10887, "epoch": 122, "lr": 9.963775681254637e-05}
{"train_loss": 0.7740174531936646, "global_step": 10888, "epoch": 122, "lr": 9.9637687150921e-05}
{"train_loss": 0.8993092179298401, "global_step": 10889, "epoch": 122, "lr": 9.963761748262244e-05}
{"train_loss": 0.7954711318016052, "global_step": 10890, "epoch": 122, "lr": 9.963754780765072e-05}
{"train_loss": 0.7002267837524414, "global_step": 10891, "epoch": 122, "lr": 9.963747812600581e-05}
{"train_loss": 0.6920379400253296, "global_step": 10892, "epoch": 122, "lr": 9.963740843768776e-05}
{"train_loss": 0.7444460988044739, "global_step": 10893, "epoch": 122, "lr": 9.963733874269657e-05}
{"train_loss": 0.6960577368736267, "global_step": 10894, "epoch": 122, "lr": 9.963726904103226e-05}
{"train_loss": 0.581951379776001, "global_step": 10895, "epoch": 122, "lr": 9.96371993326948e-05}
{"train_loss": 0.7236616611480713, "global_step": 10896, "epoch": 122, "lr": 9.963712961768422e-05}
{"train_loss": 0.6612210273742676, "global_step": 10897, "epoch": 122, "lr": 9.963705989600053e-05}
{"train_loss": 0.7248309850692749, "global_step": 10898, "epoch": 122, "lr": 9.963699016764377e-05}
{"train_loss": 0.6531001925468445, "global_step": 10899, "epoch": 122, "lr": 9.96369204326139e-05}
{"train_loss": 0.8449355363845825, "global_step": 10900, "epoch": 122, "lr": 9.963685069091096e-05}
{"train_loss": 0.7049079537391663, "global_step": 10901, "epoch": 122, "lr": 9.963678094253494e-05}
{"train_loss": 0.6697424650192261, "global_step": 10902, "epoch": 122, "lr": 9.963671118748588e-05}
{"train_loss": 0.7339779138565063, "global_step": 10903, "epoch": 122, "lr": 9.963664142576376e-05}
{"train_loss": 0.6330237984657288, "global_step": 10904, "epoch": 122, "lr": 9.96365716573686e-05}
{"train_loss": 0.6651681065559387, "global_step": 10905, "epoch": 122, "lr": 9.96365018823004e-05}
{"train_loss": 0.6381306052207947, "global_step": 10906, "epoch": 122, "lr": 9.963643210055917e-05}
{"train_loss": 0.7148260474205017, "global_step": 10907, "epoch": 122, "lr": 9.963636231214495e-05}
{"train_loss": 0.6655226349830627, "global_step": 10908, "epoch": 122, "lr": 9.96362925170577e-05}
{"train_loss": 0.6326221227645874, "global_step": 10909, "epoch": 122, "lr": 9.963622271529748e-05}
{"train_loss": 0.6446784734725952, "global_step": 10910, "epoch": 122, "lr": 9.963615290686426e-05}
{"train_loss": 0.6534717082977295, "global_step": 10911, "epoch": 122, "lr": 9.963608309175808e-05}
{"train_loss": 0.6303758025169373, "global_step": 10912, "epoch": 122, "lr": 9.963601326997891e-05}
{"train_loss": 0.7826995849609375, "global_step": 10913, "epoch": 122, "lr": 9.96359434415268e-05}
{"train_loss": 0.816180944442749, "global_step": 10914, "epoch": 122, "lr": 9.963587360640173e-05}
{"train_loss": 0.7712059617042542, "global_step": 10915, "epoch": 122, "lr": 9.963580376460373e-05}
{"train_loss": 0.749627947807312, "global_step": 10916, "epoch": 122, "lr": 9.96357339161328e-05}
{"train_loss": 0.9335759282112122, "global_step": 10917, "epoch": 122, "lr": 9.963566406098895e-05}
{"train_loss": 0.9156066179275513, "global_step": 10918, "epoch": 122, "lr": 9.963559419917218e-05}
{"train_loss": 0.6418026685714722, "global_step": 10919, "epoch": 122, "lr": 9.963552433068253e-05}
{"train_loss": 0.6876758337020874, "global_step": 10920, "epoch": 122, "lr": 9.963545445551999e-05}
{"train_loss": 0.6302149891853333, "global_step": 10921, "epoch": 122, "lr": 9.963538457368455e-05}
{"train_loss": 0.7777786254882812, "global_step": 10922, "epoch": 122, "lr": 9.963531468517624e-05}
{"train_loss": 0.7793931365013123, "global_step": 10923, "epoch": 122, "lr": 9.963524478999507e-05}
{"train_loss": 0.5927060842514038, "global_step": 10924, "epoch": 122, "lr": 9.963517488814104e-05}
{"train_loss": 0.81200110912323, "global_step": 10925, "epoch": 122, "lr": 9.963510497961418e-05}
{"train_loss": 0.9245278835296631, "global_step": 10926, "epoch": 122, "lr": 9.963503506441447e-05}
{"train_loss": 0.6270153522491455, "global_step": 10927, "epoch": 122, "lr": 9.963496514254195e-05}
{"train_loss": 0.5635934472084045, "global_step": 10928, "epoch": 122, "lr": 9.96348952139966e-05}
{"train_loss": 0.6491223573684692, "global_step": 10929, "epoch": 122, "lr": 9.963482527877845e-05}
{"train_loss": 0.625463604927063, "global_step": 10930, "epoch": 122, "lr": 9.963475533688751e-05}
{"train_loss": 0.8952318429946899, "global_step": 10931, "epoch": 122, "lr": 9.963468538832377e-05}
{"train_loss": 0.7027029991149902, "global_step": 10932, "epoch": 122, "lr": 9.963461543308727e-05}
{"train_loss": 0.6913081407546997, "global_step": 10933, "epoch": 122, "lr": 9.963454547117798e-05}
{"train_loss": 0.8322732448577881, "global_step": 10934, "epoch": 122, "lr": 9.963447550259593e-05}
{"train_loss": 0.7167691588401794, "global_step": 10935, "epoch": 122, "lr": 9.963440552734115e-05}
{"train_loss": 0.7545656561851501, "global_step": 10936, "epoch": 122, "lr": 9.963433554541361e-05}
{"train_loss": 0.739043653011322, "global_step": 10937, "epoch": 122, "lr": 9.963426555681334e-05}
{"train_loss": 0.6499146819114685, "global_step": 10938, "epoch": 122, "lr": 9.963419556154037e-05}
{"train_loss": 0.8098230957984924, "global_step": 10939, "epoch": 122, "lr": 9.963412555959467e-05}
{"train_loss": 0.818916380405426, "global_step": 10940, "epoch": 122, "lr": 9.963405555097626e-05}
{"train_loss": 0.763982892036438, "global_step": 10941, "epoch": 122, "lr": 9.963398553568516e-05}
{"train_loss": 0.7195292115211487, "global_step": 10942, "epoch": 122, "lr": 9.963391551372138e-05}
{"train_loss": 0.7686837911605835, "global_step": 10943, "epoch": 122, "lr": 9.963384548508492e-05}
{"train_loss": 0.706192135810852, "global_step": 10944, "epoch": 122, "lr": 9.96337754497758e-05}
{"train_loss": 0.6974495053291321, "global_step": 10945, "epoch": 122, "lr": 9.963370540779403e-05}
{"train_loss": 0.7116442818320199, "global_step": 10946, "epoch": 122, "lr": 9.96336353591396e-05, "val_loss": 1.3066085577011108}
{"train_loss": 0.6928240060806274, "global_step": 10947, "epoch": 123, "lr": 9.963356530381254e-05}
{"train_loss": 0.7133723497390747, "global_step": 10948, "epoch": 123, "lr": 9.963349524181285e-05}
{"train_loss": 0.6718509793281555, "global_step": 10949, "epoch": 123, "lr": 9.963342517314054e-05}
{"train_loss": 0.7613497972488403, "global_step": 10950, "epoch": 123, "lr": 9.963335509779561e-05}
{"train_loss": 0.6617720723152161, "global_step": 10951, "epoch": 123, "lr": 9.96332850157781e-05}
{"train_loss": 0.8720375299453735, "global_step": 10952, "epoch": 123, "lr": 9.9633214927088e-05}
{"train_loss": 0.7832810878753662, "global_step": 10953, "epoch": 123, "lr": 9.96331448317253e-05}
{"train_loss": 0.730277419090271, "global_step": 10954, "epoch": 123, "lr": 9.963307472969005e-05}
{"train_loss": 0.6049278378486633, "global_step": 10955, "epoch": 123, "lr": 9.963300462098221e-05}
{"train_loss": 0.7859392166137695, "global_step": 10956, "epoch": 123, "lr": 9.963293450560185e-05}
{"train_loss": 0.5315394401550293, "global_step": 10957, "epoch": 123, "lr": 9.963286438354893e-05}
{"train_loss": 0.7369971871376038, "global_step": 10958, "epoch": 123, "lr": 9.963279425482348e-05}
{"train_loss": 0.6634398102760315, "global_step": 10959, "epoch": 123, "lr": 9.96327241194255e-05}
{"train_loss": 0.7673859000205994, "global_step": 10960, "epoch": 123, "lr": 9.963265397735501e-05}
{"train_loss": 0.7475491762161255, "global_step": 10961, "epoch": 123, "lr": 9.963258382861203e-05}
{"train_loss": 0.6444118022918701, "global_step": 10962, "epoch": 123, "lr": 9.963251367319653e-05}
{"train_loss": 0.4861815273761749, "global_step": 10963, "epoch": 123, "lr": 9.963244351110855e-05}
{"train_loss": 0.6276763677597046, "global_step": 10964, "epoch": 123, "lr": 9.96323733423481e-05}
{"train_loss": 0.6969466209411621, "global_step": 10965, "epoch": 123, "lr": 9.963230316691518e-05}
{"train_loss": 0.6461484432220459, "global_step": 10966, "epoch": 123, "lr": 9.963223298480979e-05}
{"train_loss": 0.6962034702301025, "global_step": 10967, "epoch": 123, "lr": 9.963216279603196e-05}
{"train_loss": 0.6298066973686218, "global_step": 10968, "epoch": 123, "lr": 9.96320926005817e-05}
{"train_loss": 0.6744992733001709, "global_step": 10969, "epoch": 123, "lr": 9.9632022398459e-05}
{"train_loss": 0.5683026909828186, "global_step": 10970, "epoch": 123, "lr": 9.963195218966388e-05}
{"train_loss": 0.6336821913719177, "global_step": 10971, "epoch": 123, "lr": 9.963188197419635e-05}
{"train_loss": 0.6998960971832275, "global_step": 10972, "epoch": 123, "lr": 9.963181175205641e-05}
{"train_loss": 0.625036895275116, "global_step": 10973, "epoch": 123, "lr": 9.963174152324409e-05}
{"train_loss": 0.6969856023788452, "global_step": 10974, "epoch": 123, "lr": 9.963167128775938e-05}
{"train_loss": 0.7686595916748047, "global_step": 10975, "epoch": 123, "lr": 9.96316010456023e-05}
{"train_loss": 0.517732560634613, "global_step": 10976, "epoch": 123, "lr": 9.963153079677285e-05}
{"train_loss": 0.6916956901550293, "global_step": 10977, "epoch": 123, "lr": 9.963146054127106e-05}
{"train_loss": 0.7090805768966675, "global_step": 10978, "epoch": 123, "lr": 9.963139027909692e-05}
{"train_loss": 0.8447306156158447, "global_step": 10979, "epoch": 123, "lr": 9.963132001025044e-05}
{"train_loss": 0.6806747317314148, "global_step": 10980, "epoch": 123, "lr": 9.963124973473164e-05}
{"train_loss": 0.6554515361785889, "global_step": 10981, "epoch": 123, "lr": 9.963117945254051e-05}
{"train_loss": 0.665107786655426, "global_step": 10982, "epoch": 123, "lr": 9.963110916367709e-05}
{"train_loss": 0.6704712510108948, "global_step": 10983, "epoch": 123, "lr": 9.963103886814135e-05}
{"train_loss": 0.6510007977485657, "global_step": 10984, "epoch": 123, "lr": 9.963096856593334e-05}
{"train_loss": 0.6513577103614807, "global_step": 10985, "epoch": 123, "lr": 9.963089825705306e-05}
{"train_loss": 0.5808380842208862, "global_step": 10986, "epoch": 123, "lr": 9.96308279415005e-05}
{"train_loss": 0.7636215686798096, "global_step": 10987, "epoch": 123, "lr": 9.963075761927567e-05}
{"train_loss": 0.5536667704582214, "global_step": 10988, "epoch": 123, "lr": 9.963068729037859e-05}
{"train_loss": 0.6526530385017395, "global_step": 10989, "epoch": 123, "lr": 9.963061695480928e-05}
{"train_loss": 0.9038166403770447, "global_step": 10990, "epoch": 123, "lr": 9.963054661256773e-05}
{"train_loss": 0.5686591267585754, "global_step": 10991, "epoch": 123, "lr": 9.963047626365396e-05}
{"train_loss": 0.7090910077095032, "global_step": 10992, "epoch": 123, "lr": 9.963040590806797e-05}
{"train_loss": 0.670077383518219, "global_step": 10993, "epoch": 123, "lr": 9.96303355458098e-05}
{"train_loss": 0.7373506426811218, "global_step": 10994, "epoch": 123, "lr": 9.963026517687942e-05}
{"train_loss": 0.642353355884552, "global_step": 10995, "epoch": 123, "lr": 9.963019480127687e-05}
{"train_loss": 0.6183490753173828, "global_step": 10996, "epoch": 123, "lr": 9.963012441900213e-05}
{"train_loss": 0.5042476654052734, "global_step": 10997, "epoch": 123, "lr": 9.963005403005523e-05}
{"train_loss": 0.6012014150619507, "global_step": 10998, "epoch": 123, "lr": 9.962998363443617e-05}
{"train_loss": 0.542235255241394, "global_step": 10999, "epoch": 123, "lr": 9.962991323214497e-05}
{"train_loss": 0.7944233417510986, "global_step": 11000, "epoch": 123, "lr": 9.962984282318161e-05}
{"train_loss": 0.6182600855827332, "global_step": 11001, "epoch": 123, "lr": 9.962977240754614e-05}
{"train_loss": 0.7343431115150452, "global_step": 11002, "epoch": 123, "lr": 9.962970198523857e-05}
{"train_loss": 0.6707721948623657, "global_step": 11003, "epoch": 123, "lr": 9.962963155625887e-05}
{"train_loss": 0.6730260848999023, "global_step": 11004, "epoch": 123, "lr": 9.962956112060708e-05}
{"train_loss": 0.5964882373809814, "global_step": 11005, "epoch": 123, "lr": 9.962949067828319e-05}
{"train_loss": 0.5736587047576904, "global_step": 11006, "epoch": 123, "lr": 9.962942022928723e-05}
{"train_loss": 0.6544561982154846, "global_step": 11007, "epoch": 123, "lr": 9.96293497736192e-05}
{"train_loss": 0.5111743211746216, "global_step": 11008, "epoch": 123, "lr": 9.96292793112791e-05}
{"train_loss": 0.7564544081687927, "global_step": 11009, "epoch": 123, "lr": 9.962920884226696e-05}
{"train_loss": 0.736147403717041, "global_step": 11010, "epoch": 123, "lr": 9.962913836658276e-05}
{"train_loss": 0.6597643494606018, "global_step": 11011, "epoch": 123, "lr": 9.962906788422655e-05}
{"train_loss": 0.7061986923217773, "global_step": 11012, "epoch": 123, "lr": 9.96289973951983e-05}
{"train_loss": 0.7720705270767212, "global_step": 11013, "epoch": 123, "lr": 9.962892689949803e-05}
{"train_loss": 0.7117814421653748, "global_step": 11014, "epoch": 123, "lr": 9.962885639712578e-05}
{"train_loss": 0.5711736679077148, "global_step": 11015, "epoch": 123, "lr": 9.962878588808152e-05}
{"train_loss": 0.7522684335708618, "global_step": 11016, "epoch": 123, "lr": 9.962871537236527e-05}
{"train_loss": 0.6828585267066956, "global_step": 11017, "epoch": 123, "lr": 9.962864484997705e-05}
{"train_loss": 0.6615928411483765, "global_step": 11018, "epoch": 123, "lr": 9.962857432091686e-05}
{"train_loss": 0.6609901785850525, "global_step": 11019, "epoch": 123, "lr": 9.962850378518473e-05}
{"train_loss": 0.6831713318824768, "global_step": 11020, "epoch": 123, "lr": 9.962843324278064e-05}
{"train_loss": 0.7099946737289429, "global_step": 11021, "epoch": 123, "lr": 9.962836269370461e-05}
{"train_loss": 0.5291900634765625, "global_step": 11022, "epoch": 123, "lr": 9.962829213795666e-05}
{"train_loss": 0.6168712377548218, "global_step": 11023, "epoch": 123, "lr": 9.962822157553677e-05}
{"train_loss": 0.5010631084442139, "global_step": 11024, "epoch": 123, "lr": 9.962815100644498e-05}
{"train_loss": 0.5196120142936707, "global_step": 11025, "epoch": 123, "lr": 9.962808043068131e-05}
{"train_loss": 0.4830780625343323, "global_step": 11026, "epoch": 123, "lr": 9.962800984824573e-05}
{"train_loss": 0.7934436202049255, "global_step": 11027, "epoch": 123, "lr": 9.962793925913827e-05}
{"train_loss": 0.6827934980392456, "global_step": 11028, "epoch": 123, "lr": 9.962786866335894e-05}
{"train_loss": 0.7260450124740601, "global_step": 11029, "epoch": 123, "lr": 9.962779806090775e-05}
{"train_loss": 0.7266557216644287, "global_step": 11030, "epoch": 123, "lr": 9.96277274517847e-05}
{"train_loss": 0.646056592464447, "global_step": 11031, "epoch": 123, "lr": 9.962765683598983e-05}
{"train_loss": 0.6422924399375916, "global_step": 11032, "epoch": 123, "lr": 9.96275862135231e-05}
{"train_loss": 0.6722623109817505, "global_step": 11033, "epoch": 123, "lr": 9.962751558438455e-05}
{"train_loss": 0.8187159299850464, "global_step": 11034, "epoch": 123, "lr": 9.96274449485742e-05}
{"train_loss": 0.6703047923157724, "global_step": 11035, "epoch": 123, "lr": 9.962737430609204e-05, "val_loss": 1.2763044834136963}
{"train_loss": 0.737511396408081, "global_step": 11036, "epoch": 124, "lr": 9.962730365693808e-05}
{"train_loss": 0.6714621186256409, "global_step": 11037, "epoch": 124, "lr": 9.962723300111233e-05}
{"train_loss": 0.7314049601554871, "global_step": 11038, "epoch": 124, "lr": 9.962716233861481e-05}
{"train_loss": 0.6101608872413635, "global_step": 11039, "epoch": 124, "lr": 9.962709166944553e-05}
{"train_loss": 0.5350173711776733, "global_step": 11040, "epoch": 124, "lr": 9.962702099360448e-05}
{"train_loss": 0.6087044477462769, "global_step": 11041, "epoch": 124, "lr": 9.962695031109168e-05}
{"train_loss": 0.5024263858795166, "global_step": 11042, "epoch": 124, "lr": 9.962687962190716e-05}
{"train_loss": 0.6051491498947144, "global_step": 11043, "epoch": 124, "lr": 9.962680892605088e-05}
{"train_loss": 0.7112996578216553, "global_step": 11044, "epoch": 124, "lr": 9.96267382235229e-05}
{"train_loss": 0.7150163650512695, "global_step": 11045, "epoch": 124, "lr": 9.962666751432322e-05}
{"train_loss": 0.596527636051178, "global_step": 11046, "epoch": 124, "lr": 9.962659679845181e-05}
{"train_loss": 0.6107279658317566, "global_step": 11047, "epoch": 124, "lr": 9.962652607590873e-05}
{"train_loss": 0.8285511136054993, "global_step": 11048, "epoch": 124, "lr": 9.962645534669397e-05}
{"train_loss": 0.6062945127487183, "global_step": 11049, "epoch": 124, "lr": 9.962638461080753e-05}
{"train_loss": 0.7296836972236633, "global_step": 11050, "epoch": 124, "lr": 9.962631386824944e-05}
{"train_loss": 0.6827060580253601, "global_step": 11051, "epoch": 124, "lr": 9.962624311901968e-05}
{"train_loss": 0.69866943359375, "global_step": 11052, "epoch": 124, "lr": 9.962617236311828e-05}
{"train_loss": 0.7491365671157837, "global_step": 11053, "epoch": 124, "lr": 9.962610160054525e-05}
{"train_loss": 0.6757272481918335, "global_step": 11054, "epoch": 124, "lr": 9.96260308313006e-05}
{"train_loss": 0.5494416356086731, "global_step": 11055, "epoch": 124, "lr": 9.962596005538432e-05}
{"train_loss": 0.7804360389709473, "global_step": 11056, "epoch": 124, "lr": 9.962588927279646e-05}
{"train_loss": 0.8308063745498657, "global_step": 11057, "epoch": 124, "lr": 9.962581848353698e-05}
{"train_loss": 0.8079765439033508, "global_step": 11058, "epoch": 124, "lr": 9.962574768760592e-05}
{"train_loss": 0.7574421167373657, "global_step": 11059, "epoch": 124, "lr": 9.962567688500328e-05}
{"train_loss": 0.6911932229995728, "global_step": 11060, "epoch": 124, "lr": 9.962560607572908e-05}
{"train_loss": 0.6699183583259583, "global_step": 11061, "epoch": 124, "lr": 9.962553525978334e-05}
{"train_loss": 0.8274316191673279, "global_step": 11062, "epoch": 124, "lr": 9.962546443716602e-05}
{"train_loss": 0.6016443371772766, "global_step": 11063, "epoch": 124, "lr": 9.962539360787718e-05}
{"train_loss": 0.6202796697616577, "global_step": 11064, "epoch": 124, "lr": 9.962532277191679e-05}
{"train_loss": 0.6667063236236572, "global_step": 11065, "epoch": 124, "lr": 9.96252519292849e-05}
{"train_loss": 0.6570960283279419, "global_step": 11066, "epoch": 124, "lr": 9.962518107998149e-05}
{"train_loss": 0.6233285665512085, "global_step": 11067, "epoch": 124, "lr": 9.962511022400658e-05}
{"train_loss": 0.5501686334609985, "global_step": 11068, "epoch": 124, "lr": 9.962503936136018e-05}
{"train_loss": 0.7692462801933289, "global_step": 11069, "epoch": 124, "lr": 9.962496849204231e-05}
{"train_loss": 0.7090751528739929, "global_step": 11070, "epoch": 124, "lr": 9.962489761605295e-05}
{"train_loss": 0.5851629376411438, "global_step": 11071, "epoch": 124, "lr": 9.962482673339213e-05}
{"train_loss": 0.6167529225349426, "global_step": 11072, "epoch": 124, "lr": 9.962475584405986e-05}
{"train_loss": 0.6558990478515625, "global_step": 11073, "epoch": 124, "lr": 9.962468494805615e-05}
{"train_loss": 0.8225511908531189, "global_step": 11074, "epoch": 124, "lr": 9.962461404538101e-05}
{"train_loss": 0.7804784774780273, "global_step": 11075, "epoch": 124, "lr": 9.962454313603445e-05}
{"train_loss": 0.8389953970909119, "global_step": 11076, "epoch": 124, "lr": 9.962447222001645e-05}
{"train_loss": 0.818710446357727, "global_step": 11077, "epoch": 124, "lr": 9.962440129732707e-05}
{"train_loss": 0.7527832984924316, "global_step": 11078, "epoch": 124, "lr": 9.962433036796628e-05}
{"train_loss": 0.6873107552528381, "global_step": 11079, "epoch": 124, "lr": 9.96242594319341e-05}
{"train_loss": 0.6807917952537537, "global_step": 11080, "epoch": 124, "lr": 9.962418848923055e-05}
{"train_loss": 0.7225779294967651, "global_step": 11081, "epoch": 124, "lr": 9.962411753985564e-05}
{"train_loss": 0.6198950409889221, "global_step": 11082, "epoch": 124, "lr": 9.962404658380935e-05}
{"train_loss": 0.77419513463974, "global_step": 11083, "epoch": 124, "lr": 9.962397562109173e-05}
{"train_loss": 0.6924577355384827, "global_step": 11084, "epoch": 124, "lr": 9.962390465170277e-05}
{"train_loss": 0.6887761354446411, "global_step": 11085, "epoch": 124, "lr": 9.962383367564248e-05}
{"train_loss": 0.7291385531425476, "global_step": 11086, "epoch": 124, "lr": 9.962376269291087e-05}
{"train_loss": 0.6574097275733948, "global_step": 11087, "epoch": 124, "lr": 9.962369170350797e-05}
{"train_loss": 0.5618170499801636, "global_step": 11088, "epoch": 124, "lr": 9.962362070743373e-05}
{"train_loss": 0.659699559211731, "global_step": 11089, "epoch": 124, "lr": 9.962354970468821e-05}
{"train_loss": 0.6915246248245239, "global_step": 11090, "epoch": 124, "lr": 9.962347869527142e-05}
{"train_loss": 0.741010844707489, "global_step": 11091, "epoch": 124, "lr": 9.962340767918336e-05}
{"train_loss": 0.5393725037574768, "global_step": 11092, "epoch": 124, "lr": 9.962333665642403e-05}
{"train_loss": 0.6444884538650513, "global_step": 11093, "epoch": 124, "lr": 9.962326562699344e-05}
{"train_loss": 0.7250990867614746, "global_step": 11094, "epoch": 124, "lr": 9.962319459089162e-05}
{"train_loss": 0.6543580889701843, "global_step": 11095, "epoch": 124, "lr": 9.962312354811857e-05}
{"train_loss": 0.7277386784553528, "global_step": 11096, "epoch": 124, "lr": 9.962305249867429e-05}
{"train_loss": 0.6000505685806274, "global_step": 11097, "epoch": 124, "lr": 9.962298144255879e-05}
{"train_loss": 0.8395247459411621, "global_step": 11098, "epoch": 124, "lr": 9.96229103797721e-05}
{"train_loss": 0.7297934889793396, "global_step": 11099, "epoch": 124, "lr": 9.96228393103142e-05}
{"train_loss": 0.5979375839233398, "global_step": 11100, "epoch": 124, "lr": 9.96227682341851e-05}
{"train_loss": 0.8350043296813965, "global_step": 11101, "epoch": 124, "lr": 9.962269715138484e-05}
{"train_loss": 0.7512445449829102, "global_step": 11102, "epoch": 124, "lr": 9.962262606191341e-05}
{"train_loss": 0.5685055255889893, "global_step": 11103, "epoch": 124, "lr": 9.962255496577084e-05}
{"train_loss": 0.970589816570282, "global_step": 11104, "epoch": 124, "lr": 9.962248386295712e-05}
{"train_loss": 0.6313331127166748, "global_step": 11105, "epoch": 124, "lr": 9.962241275347223e-05}
{"train_loss": 0.7523715496063232, "global_step": 11106, "epoch": 124, "lr": 9.962234163731624e-05}
{"train_loss": 0.7687230706214905, "global_step": 11107, "epoch": 124, "lr": 9.962227051448912e-05}
{"train_loss": 0.6669522523880005, "global_step": 11108, "epoch": 124, "lr": 9.96221993849909e-05}
{"train_loss": 0.8265303373336792, "global_step": 11109, "epoch": 124, "lr": 9.962212824882157e-05}
{"train_loss": 0.7927207350730896, "global_step": 11110, "epoch": 124, "lr": 9.962205710598116e-05}
{"train_loss": 0.6669595241546631, "global_step": 11111, "epoch": 124, "lr": 9.962198595646965e-05}
{"train_loss": 0.6973037719726562, "global_step": 11112, "epoch": 124, "lr": 9.962191480028709e-05}
{"train_loss": 0.669806182384491, "global_step": 11113, "epoch": 124, "lr": 9.962184363743346e-05}
{"train_loss": 0.8484318256378174, "global_step": 11114, "epoch": 124, "lr": 9.962177246790878e-05}
{"train_loss": 0.6552200317382812, "global_step": 11115, "epoch": 124, "lr": 9.962170129171306e-05}
{"train_loss": 0.7386370301246643, "global_step": 11116, "epoch": 124, "lr": 9.96216301088463e-05}
{"train_loss": 0.6120880246162415, "global_step": 11117, "epoch": 124, "lr": 9.962155891930851e-05}
{"train_loss": 0.5952625870704651, "global_step": 11118, "epoch": 124, "lr": 9.962148772309971e-05}
{"train_loss": 0.6299805045127869, "global_step": 11119, "epoch": 124, "lr": 9.962141652021993e-05}
{"train_loss": 0.6797371506690979, "global_step": 11120, "epoch": 124, "lr": 9.962134531066914e-05}
{"train_loss": 0.6823788285255432, "global_step": 11121, "epoch": 124, "lr": 9.962127409444735e-05}
{"train_loss": 0.680729866027832, "global_step": 11122, "epoch": 124, "lr": 9.96212028715546e-05}
{"train_loss": 0.7310552000999451, "global_step": 11123, "epoch": 124, "lr": 9.962113164199087e-05}
{"train_loss": 0.6930296836274393, "global_step": 11124, "epoch": 124, "lr": 9.96210604057562e-05, "val_loss": 1.2892287969589233}
{"train_loss": 0.8192130923271179, "global_step": 11125, "epoch": 125, "lr": 9.962098916285057e-05}
{"train_loss": 0.8069227933883667, "global_step": 11126, "epoch": 125, "lr": 9.962091791327402e-05}
{"train_loss": 0.7626476287841797, "global_step": 11127, "epoch": 125, "lr": 9.962084665702654e-05}
{"train_loss": 0.6446974277496338, "global_step": 11128, "epoch": 125, "lr": 9.962077539410812e-05}
{"train_loss": 0.633330225944519, "global_step": 11129, "epoch": 125, "lr": 9.962070412451881e-05}
{"train_loss": 0.78310227394104, "global_step": 11130, "epoch": 125, "lr": 9.96206328482586e-05}
{"train_loss": 0.6513454914093018, "global_step": 11131, "epoch": 125, "lr": 9.962056156532748e-05}
{"train_loss": 0.7748439311981201, "global_step": 11132, "epoch": 125, "lr": 9.96204902757255e-05}
{"train_loss": 0.6622875928878784, "global_step": 11133, "epoch": 125, "lr": 9.962041897945265e-05}
{"train_loss": 0.8193628787994385, "global_step": 11134, "epoch": 125, "lr": 9.962034767650895e-05}
{"train_loss": 0.6992252469062805, "global_step": 11135, "epoch": 125, "lr": 9.962027636689437e-05}
{"train_loss": 0.6323433518409729, "global_step": 11136, "epoch": 125, "lr": 9.962020505060898e-05}
{"train_loss": 0.6138083934783936, "global_step": 11137, "epoch": 125, "lr": 9.962013372765273e-05}
{"train_loss": 0.6924679279327393, "global_step": 11138, "epoch": 125, "lr": 9.962006239802568e-05}
{"train_loss": 0.6182309985160828, "global_step": 11139, "epoch": 125, "lr": 9.96199910617278e-05}
{"train_loss": 0.6721562743186951, "global_step": 11140, "epoch": 125, "lr": 9.961991971875913e-05}
{"train_loss": 0.6234468221664429, "global_step": 11141, "epoch": 125, "lr": 9.961984836911966e-05}
{"train_loss": 0.6927157044410706, "global_step": 11142, "epoch": 125, "lr": 9.96197770128094e-05}
{"train_loss": 0.8281630873680115, "global_step": 11143, "epoch": 125, "lr": 9.961970564982839e-05}
{"train_loss": 0.6247780323028564, "global_step": 11144, "epoch": 125, "lr": 9.961963428017658e-05}
{"train_loss": 0.735085129737854, "global_step": 11145, "epoch": 125, "lr": 9.961956290385403e-05}
{"train_loss": 0.7112263441085815, "global_step": 11146, "epoch": 125, "lr": 9.961949152086075e-05}
{"train_loss": 0.7204106450080872, "global_step": 11147, "epoch": 125, "lr": 9.961942013119671e-05}
{"train_loss": 0.6519149541854858, "global_step": 11148, "epoch": 125, "lr": 9.961934873486197e-05}
{"train_loss": 0.5911785960197449, "global_step": 11149, "epoch": 125, "lr": 9.961927733185649e-05}
{"train_loss": 0.5749941468238831, "global_step": 11150, "epoch": 125, "lr": 9.961920592218033e-05}
{"train_loss": 0.7194374799728394, "global_step": 11151, "epoch": 125, "lr": 9.961913450583344e-05}
{"train_loss": 0.5515114068984985, "global_step": 11152, "epoch": 125, "lr": 9.961906308281588e-05}
{"train_loss": 0.7109246850013733, "global_step": 11153, "epoch": 125, "lr": 9.961899165312765e-05}
{"train_loss": 0.6429852247238159, "global_step": 11154, "epoch": 125, "lr": 9.961892021676873e-05}
{"train_loss": 0.7092090249061584, "global_step": 11155, "epoch": 125, "lr": 9.961884877373917e-05}
{"train_loss": 0.6348835229873657, "global_step": 11156, "epoch": 125, "lr": 9.961877732403895e-05}
{"train_loss": 0.5428459644317627, "global_step": 11157, "epoch": 125, "lr": 9.961870586766809e-05}
{"train_loss": 0.732738733291626, "global_step": 11158, "epoch": 125, "lr": 9.96186344046266e-05}
{"train_loss": 0.6021873950958252, "global_step": 11159, "epoch": 125, "lr": 9.96185629349145e-05}
{"train_loss": 0.5890455842018127, "global_step": 11160, "epoch": 125, "lr": 9.961849145853177e-05}
{"train_loss": 0.7292495965957642, "global_step": 11161, "epoch": 125, "lr": 9.961841997547845e-05}
{"train_loss": 0.7027657628059387, "global_step": 11162, "epoch": 125, "lr": 9.961834848575453e-05}
{"train_loss": 0.6238407492637634, "global_step": 11163, "epoch": 125, "lr": 9.961827698936003e-05}
{"train_loss": 0.6966118216514587, "global_step": 11164, "epoch": 125, "lr": 9.961820548629497e-05}
{"train_loss": 0.7107962369918823, "global_step": 11165, "epoch": 125, "lr": 9.961813397655935e-05}
{"train_loss": 0.6915010809898376, "global_step": 11166, "epoch": 125, "lr": 9.961806246015315e-05}
{"train_loss": 0.6984745860099792, "global_step": 11167, "epoch": 125, "lr": 9.961799093707642e-05}
{"train_loss": 0.6788356900215149, "global_step": 11168, "epoch": 125, "lr": 9.961791940732916e-05}
{"train_loss": 0.5947117209434509, "global_step": 11169, "epoch": 125, "lr": 9.961784787091137e-05}
{"train_loss": 0.585521399974823, "global_step": 11170, "epoch": 125, "lr": 9.961777632782308e-05}
{"train_loss": 0.6729229092597961, "global_step": 11171, "epoch": 125, "lr": 9.961770477806427e-05}
{"train_loss": 0.7087494134902954, "global_step": 11172, "epoch": 125, "lr": 9.961763322163498e-05}
{"train_loss": 0.8039348125457764, "global_step": 11173, "epoch": 125, "lr": 9.961756165853517e-05}
{"train_loss": 0.8052460551261902, "global_step": 11174, "epoch": 125, "lr": 9.96174900887649e-05}
{"train_loss": 0.7875112891197205, "global_step": 11175, "epoch": 125, "lr": 9.961741851232418e-05}
{"train_loss": 0.5172123312950134, "global_step": 11176, "epoch": 125, "lr": 9.9617346929213e-05}
{"train_loss": 0.7592636942863464, "global_step": 11177, "epoch": 125, "lr": 9.961727533943135e-05}
{"train_loss": 0.6558557152748108, "global_step": 11178, "epoch": 125, "lr": 9.961720374297928e-05}
{"train_loss": 0.648494303226471, "global_step": 11179, "epoch": 125, "lr": 9.961713213985679e-05}
{"train_loss": 0.5728545188903809, "global_step": 11180, "epoch": 125, "lr": 9.961706053006385e-05}
{"train_loss": 0.6880948543548584, "global_step": 11181, "epoch": 125, "lr": 9.961698891360053e-05}
{"train_loss": 0.6945654153823853, "global_step": 11182, "epoch": 125, "lr": 9.96169172904668e-05}
{"train_loss": 0.6639981269836426, "global_step": 11183, "epoch": 125, "lr": 9.961684566066267e-05}
{"train_loss": 0.765673041343689, "global_step": 11184, "epoch": 125, "lr": 9.961677402418818e-05}
{"train_loss": 0.5512333512306213, "global_step": 11185, "epoch": 125, "lr": 9.96167023810433e-05}
{"train_loss": 0.7215093374252319, "global_step": 11186, "epoch": 125, "lr": 9.961663073122807e-05}
{"train_loss": 0.6122958660125732, "global_step": 11187, "epoch": 125, "lr": 9.96165590747425e-05}
{"train_loss": 0.581849217414856, "global_step": 11188, "epoch": 125, "lr": 9.961648741158657e-05}
{"train_loss": 0.6365427374839783, "global_step": 11189, "epoch": 125, "lr": 9.961641574176032e-05}
{"train_loss": 0.6628352403640747, "global_step": 11190, "epoch": 125, "lr": 9.961634406526375e-05}
{"train_loss": 0.565260112285614, "global_step": 11191, "epoch": 125, "lr": 9.961627238209685e-05}
{"train_loss": 0.7240361571311951, "global_step": 11192, "epoch": 125, "lr": 9.961620069225965e-05}
{"train_loss": 0.9120633006095886, "global_step": 11193, "epoch": 125, "lr": 9.961612899575217e-05}
{"train_loss": 0.6516337394714355, "global_step": 11194, "epoch": 125, "lr": 9.961605729257439e-05}
{"train_loss": 0.6405734419822693, "global_step": 11195, "epoch": 125, "lr": 9.961598558272635e-05}
{"train_loss": 0.8382270336151123, "global_step": 11196, "epoch": 125, "lr": 9.961591386620804e-05}
{"train_loss": 0.6608473658561707, "global_step": 11197, "epoch": 125, "lr": 9.961584214301946e-05}
{"train_loss": 0.5527809262275696, "global_step": 11198, "epoch": 125, "lr": 9.961577041316066e-05}
{"train_loss": 0.6018677949905396, "global_step": 11199, "epoch": 125, "lr": 9.961569867663162e-05}
{"train_loss": 0.7241019010543823, "global_step": 11200, "epoch": 125, "lr": 9.961562693343233e-05}
{"train_loss": 0.6401833891868591, "global_step": 11201, "epoch": 125, "lr": 9.961555518356286e-05}
{"train_loss": 0.7958148717880249, "global_step": 11202, "epoch": 125, "lr": 9.961548342702316e-05}
{"train_loss": 0.7477015852928162, "global_step": 11203, "epoch": 125, "lr": 9.961541166381327e-05}
{"train_loss": 0.6690119504928589, "global_step": 11204, "epoch": 125, "lr": 9.961533989393319e-05}
{"train_loss": 0.8009775280952454, "global_step": 11205, "epoch": 125, "lr": 9.961526811738294e-05}
{"train_loss": 0.662865400314331, "global_step": 11206, "epoch": 125, "lr": 9.961519633416251e-05}
{"train_loss": 0.45769360661506653, "global_step": 11207, "epoch": 125, "lr": 9.961512454427194e-05}
{"train_loss": 0.7662789821624756, "global_step": 11208, "epoch": 125, "lr": 9.961505274771119e-05}
{"train_loss": 0.6045805811882019, "global_step": 11209, "epoch": 125, "lr": 9.961498094448032e-05}
{"train_loss": 0.750453531742096, "global_step": 11210, "epoch": 125, "lr": 9.961490913457932e-05}
{"train_loss": 0.5689883828163147, "global_step": 11211, "epoch": 125, "lr": 9.961483731800821e-05}
{"train_loss": 0.7191711664199829, "global_step": 11212, "epoch": 125, "lr": 9.961476549476697e-05}
{"train_loss": 0.679915443230211, "global_step": 11213, "epoch": 125, "lr": 9.961469366485564e-05, "val_loss": 1.2611621618270874, "train_action_mse_error": 43.69166564941406}
{"train_loss": 0.627280592918396, "global_step": 11214, "epoch": 126, "lr": 9.961462182827421e-05}
{"train_loss": 0.6718589067459106, "global_step": 11215, "epoch": 126, "lr": 9.96145499850227e-05}
{"train_loss": 0.6642789840698242, "global_step": 11216, "epoch": 126, "lr": 9.961447813510114e-05}
{"train_loss": 0.530430018901825, "global_step": 11217, "epoch": 126, "lr": 9.96144062785095e-05}
{"train_loss": 0.6174347996711731, "global_step": 11218, "epoch": 126, "lr": 9.961433441524781e-05}
{"train_loss": 0.6881856322288513, "global_step": 11219, "epoch": 126, "lr": 9.961426254531608e-05}
{"train_loss": 0.6880478262901306, "global_step": 11220, "epoch": 126, "lr": 9.961419066871431e-05}
{"train_loss": 0.751596987247467, "global_step": 11221, "epoch": 126, "lr": 9.961411878544254e-05}
{"train_loss": 0.6071394085884094, "global_step": 11222, "epoch": 126, "lr": 9.961404689550073e-05}
{"train_loss": 0.6158795356750488, "global_step": 11223, "epoch": 126, "lr": 9.961397499888893e-05}
{"train_loss": 0.7042301297187805, "global_step": 11224, "epoch": 126, "lr": 9.961390309560712e-05}
{"train_loss": 0.6402711272239685, "global_step": 11225, "epoch": 126, "lr": 9.961383118565535e-05}
{"train_loss": 0.7137368321418762, "global_step": 11226, "epoch": 126, "lr": 9.961375926903357e-05}
{"train_loss": 0.6839947700500488, "global_step": 11227, "epoch": 126, "lr": 9.961368734574186e-05}
{"train_loss": 0.6337273716926575, "global_step": 11228, "epoch": 126, "lr": 9.961361541578018e-05}
{"train_loss": 0.6743754744529724, "global_step": 11229, "epoch": 126, "lr": 9.961354347914855e-05}
{"train_loss": 0.578658938407898, "global_step": 11230, "epoch": 126, "lr": 9.961347153584699e-05}
{"train_loss": 0.7035866975784302, "global_step": 11231, "epoch": 126, "lr": 9.96133995858755e-05}
{"train_loss": 0.6634438633918762, "global_step": 11232, "epoch": 126, "lr": 9.96133276292341e-05}
{"train_loss": 0.7277803421020508, "global_step": 11233, "epoch": 126, "lr": 9.961325566592278e-05}
{"train_loss": 0.5907307863235474, "global_step": 11234, "epoch": 126, "lr": 9.961318369594157e-05}
{"train_loss": 0.5921521186828613, "global_step": 11235, "epoch": 126, "lr": 9.961311171929049e-05}
{"train_loss": 0.7103444337844849, "global_step": 11236, "epoch": 126, "lr": 9.96130397359695e-05}
{"train_loss": 0.7566064596176147, "global_step": 11237, "epoch": 126, "lr": 9.961296774597866e-05}
{"train_loss": 0.7267807722091675, "global_step": 11238, "epoch": 126, "lr": 9.961289574931796e-05}
{"train_loss": 0.7692031264305115, "global_step": 11239, "epoch": 126, "lr": 9.96128237459874e-05}
{"train_loss": 0.5882588028907776, "global_step": 11240, "epoch": 126, "lr": 9.961275173598701e-05}
{"train_loss": 0.5079594254493713, "global_step": 11241, "epoch": 126, "lr": 9.96126797193168e-05}
{"train_loss": 0.638360857963562, "global_step": 11242, "epoch": 126, "lr": 9.961260769597676e-05}
{"train_loss": 0.6503227353096008, "global_step": 11243, "epoch": 126, "lr": 9.96125356659669e-05}
{"train_loss": 0.6519153118133545, "global_step": 11244, "epoch": 126, "lr": 9.961246362928725e-05}
{"train_loss": 0.8145882487297058, "global_step": 11245, "epoch": 126, "lr": 9.96123915859378e-05}
{"train_loss": 0.5919293761253357, "global_step": 11246, "epoch": 126, "lr": 9.961231953591858e-05}
{"train_loss": 0.7229515910148621, "global_step": 11247, "epoch": 126, "lr": 9.961224747922958e-05}
{"train_loss": 0.6198371052742004, "global_step": 11248, "epoch": 126, "lr": 9.961217541587083e-05}
{"train_loss": 0.7445393204689026, "global_step": 11249, "epoch": 126, "lr": 9.961210334584231e-05}
{"train_loss": 0.775000810623169, "global_step": 11250, "epoch": 126, "lr": 9.961203126914406e-05}
{"train_loss": 0.6599081158638, "global_step": 11251, "epoch": 126, "lr": 9.961195918577607e-05}
{"train_loss": 0.6075637340545654, "global_step": 11252, "epoch": 126, "lr": 9.961188709573837e-05}
{"train_loss": 0.7479574680328369, "global_step": 11253, "epoch": 126, "lr": 9.961181499903094e-05}
{"train_loss": 0.8012795448303223, "global_step": 11254, "epoch": 126, "lr": 9.96117428956538e-05}
{"train_loss": 0.5825303792953491, "global_step": 11255, "epoch": 126, "lr": 9.961167078560698e-05}
{"train_loss": 0.6718643307685852, "global_step": 11256, "epoch": 126, "lr": 9.961159866889048e-05}
{"train_loss": 0.7000419497489929, "global_step": 11257, "epoch": 126, "lr": 9.961152654550429e-05}
{"train_loss": 0.5320853590965271, "global_step": 11258, "epoch": 126, "lr": 9.961145441544846e-05}
{"train_loss": 0.500537097454071, "global_step": 11259, "epoch": 126, "lr": 9.961138227872295e-05}
{"train_loss": 0.5871526002883911, "global_step": 11260, "epoch": 126, "lr": 9.961131013532781e-05}
{"train_loss": 0.6919012665748596, "global_step": 11261, "epoch": 126, "lr": 9.961123798526303e-05}
{"train_loss": 0.5878444910049438, "global_step": 11262, "epoch": 126, "lr": 9.961116582852862e-05}
{"train_loss": 0.5745635032653809, "global_step": 11263, "epoch": 126, "lr": 9.96110936651246e-05}
{"train_loss": 0.6937554478645325, "global_step": 11264, "epoch": 126, "lr": 9.961102149505096e-05}
{"train_loss": 0.5580479502677917, "global_step": 11265, "epoch": 126, "lr": 9.961094931830774e-05}
{"train_loss": 0.6753796935081482, "global_step": 11266, "epoch": 126, "lr": 9.961087713489493e-05}
{"train_loss": 0.7253520488739014, "global_step": 11267, "epoch": 126, "lr": 9.961080494481252e-05}
{"train_loss": 0.7423015236854553, "global_step": 11268, "epoch": 126, "lr": 9.961073274806057e-05}
{"train_loss": 0.6411784887313843, "global_step": 11269, "epoch": 126, "lr": 9.961066054463904e-05}
{"train_loss": 0.6980724334716797, "global_step": 11270, "epoch": 126, "lr": 9.961058833454797e-05}
{"train_loss": 0.709639310836792, "global_step": 11271, "epoch": 126, "lr": 9.961051611778736e-05}
{"train_loss": 0.6605759859085083, "global_step": 11272, "epoch": 126, "lr": 9.961044389435722e-05}
{"train_loss": 0.5574108362197876, "global_step": 11273, "epoch": 126, "lr": 9.961037166425756e-05}
{"train_loss": 0.7811942100524902, "global_step": 11274, "epoch": 126, "lr": 9.96102994274884e-05}
{"train_loss": 0.6423265337944031, "global_step": 11275, "epoch": 126, "lr": 9.961022718404973e-05}
{"train_loss": 0.5924597382545471, "global_step": 11276, "epoch": 126, "lr": 9.961015493394157e-05}
{"train_loss": 0.700560450553894, "global_step": 11277, "epoch": 126, "lr": 9.961008267716393e-05}
{"train_loss": 0.5455877780914307, "global_step": 11278, "epoch": 126, "lr": 9.961001041371682e-05}
{"train_loss": 0.7412771582603455, "global_step": 11279, "epoch": 126, "lr": 9.960993814360024e-05}
{"train_loss": 0.6858903169631958, "global_step": 11280, "epoch": 126, "lr": 9.960986586681422e-05}
{"train_loss": 0.6494943499565125, "global_step": 11281, "epoch": 126, "lr": 9.960979358335876e-05}
{"train_loss": 0.740077555179596, "global_step": 11282, "epoch": 126, "lr": 9.960972129323387e-05}
{"train_loss": 0.6643081307411194, "global_step": 11283, "epoch": 126, "lr": 9.960964899643956e-05}
{"train_loss": 0.8093439340591431, "global_step": 11284, "epoch": 126, "lr": 9.960957669297582e-05}
{"train_loss": 0.6749096512794495, "global_step": 11285, "epoch": 126, "lr": 9.960950438284269e-05}
{"train_loss": 0.8005168437957764, "global_step": 11286, "epoch": 126, "lr": 9.960943206604016e-05}
{"train_loss": 0.6789183020591736, "global_step": 11287, "epoch": 126, "lr": 9.960935974256824e-05}
{"train_loss": 0.7750937938690186, "global_step": 11288, "epoch": 126, "lr": 9.960928741242696e-05}
{"train_loss": 0.7493390440940857, "global_step": 11289, "epoch": 126, "lr": 9.960921507561631e-05}
{"train_loss": 0.6705923676490784, "global_step": 11290, "epoch": 126, "lr": 9.96091427321363e-05}
{"train_loss": 0.801568329334259, "global_step": 11291, "epoch": 126, "lr": 9.960907038198697e-05}
{"train_loss": 0.6019287705421448, "global_step": 11292, "epoch": 126, "lr": 9.960899802516829e-05}
{"train_loss": 0.5993735790252686, "global_step": 11293, "epoch": 126, "lr": 9.960892566168028e-05}
{"train_loss": 0.663486123085022, "global_step": 11294, "epoch": 126, "lr": 9.960885329152296e-05}
{"train_loss": 0.7031898498535156, "global_step": 11295, "epoch": 126, "lr": 9.960878091469632e-05}
{"train_loss": 0.7936603426933289, "global_step": 11296, "epoch": 126, "lr": 9.96087085312004e-05}
{"train_loss": 0.7399190664291382, "global_step": 11297, "epoch": 126, "lr": 9.960863614103519e-05}
{"train_loss": 0.5516523718833923, "global_step": 11298, "epoch": 126, "lr": 9.960856374420071e-05}
{"train_loss": 0.6242085695266724, "global_step": 11299, "epoch": 126, "lr": 9.960849134069696e-05}
{"train_loss": 0.7441235780715942, "global_step": 11300, "epoch": 126, "lr": 9.960841893052394e-05}
{"train_loss": 0.7415958046913147, "global_step": 11301, "epoch": 126, "lr": 9.960834651368168e-05}
{"train_loss": 0.6702060518639811, "global_step": 11302, "epoch": 126, "lr": 9.960827409017018e-05, "val_loss": 1.2766484022140503}
{"train_loss": 0.5667803883552551, "global_step": 11303, "epoch": 127, "lr": 9.960820165998946e-05}
{"train_loss": 0.5146195292472839, "global_step": 11304, "epoch": 127, "lr": 9.960812922313953e-05}
{"train_loss": 0.6129120588302612, "global_step": 11305, "epoch": 127, "lr": 9.960805677962038e-05}
{"train_loss": 0.6283130049705505, "global_step": 11306, "epoch": 127, "lr": 9.960798432943202e-05}
{"train_loss": 0.7618481516838074, "global_step": 11307, "epoch": 127, "lr": 9.960791187257449e-05}
{"train_loss": 0.6938239336013794, "global_step": 11308, "epoch": 127, "lr": 9.960783940904776e-05}
{"train_loss": 0.6090891361236572, "global_step": 11309, "epoch": 127, "lr": 9.960776693885187e-05}
{"train_loss": 0.6896416544914246, "global_step": 11310, "epoch": 127, "lr": 9.960769446198683e-05}
{"train_loss": 0.7762297987937927, "global_step": 11311, "epoch": 127, "lr": 9.960762197845264e-05}
{"train_loss": 0.6212562918663025, "global_step": 11312, "epoch": 127, "lr": 9.96075494882493e-05}
{"train_loss": 0.6795937418937683, "global_step": 11313, "epoch": 127, "lr": 9.960747699137684e-05}
{"train_loss": 0.7605067491531372, "global_step": 11314, "epoch": 127, "lr": 9.960740448783526e-05}
{"train_loss": 0.7180187106132507, "global_step": 11315, "epoch": 127, "lr": 9.960733197762455e-05}
{"train_loss": 0.7422239780426025, "global_step": 11316, "epoch": 127, "lr": 9.960725946074476e-05}
{"train_loss": 0.6838876008987427, "global_step": 11317, "epoch": 127, "lr": 9.960718693719586e-05}
{"train_loss": 0.5327832698822021, "global_step": 11318, "epoch": 127, "lr": 9.960711440697789e-05}
{"train_loss": 0.5202198624610901, "global_step": 11319, "epoch": 127, "lr": 9.960704187009086e-05}
{"train_loss": 0.6609723567962646, "global_step": 11320, "epoch": 127, "lr": 9.960696932653475e-05}
{"train_loss": 0.6390698552131653, "global_step": 11321, "epoch": 127, "lr": 9.960689677630959e-05}
{"train_loss": 0.6583890914916992, "global_step": 11322, "epoch": 127, "lr": 9.960682421941538e-05}
{"train_loss": 0.637706995010376, "global_step": 11323, "epoch": 127, "lr": 9.960675165585217e-05}
{"train_loss": 0.8531739115715027, "global_step": 11324, "epoch": 127, "lr": 9.96066790856199e-05}
{"train_loss": 0.6884754300117493, "global_step": 11325, "epoch": 127, "lr": 9.960660650871862e-05}
{"train_loss": 0.623960554599762, "global_step": 11326, "epoch": 127, "lr": 9.960653392514836e-05}
{"train_loss": 0.5536837577819824, "global_step": 11327, "epoch": 127, "lr": 9.96064613349091e-05}
{"train_loss": 0.7308419942855835, "global_step": 11328, "epoch": 127, "lr": 9.960638873800084e-05}
{"train_loss": 0.7416317462921143, "global_step": 11329, "epoch": 127, "lr": 9.960631613442362e-05}
{"train_loss": 0.7378758788108826, "global_step": 11330, "epoch": 127, "lr": 9.960624352417742e-05}
{"train_loss": 0.7076719999313354, "global_step": 11331, "epoch": 127, "lr": 9.960617090726227e-05}
{"train_loss": 0.6858686208724976, "global_step": 11332, "epoch": 127, "lr": 9.960609828367819e-05}
{"train_loss": 0.7224746346473694, "global_step": 11333, "epoch": 127, "lr": 9.960602565342517e-05}
{"train_loss": 0.7972293496131897, "global_step": 11334, "epoch": 127, "lr": 9.96059530165032e-05}
{"train_loss": 0.6435448527336121, "global_step": 11335, "epoch": 127, "lr": 9.960588037291234e-05}
{"train_loss": 0.6421675086021423, "global_step": 11336, "epoch": 127, "lr": 9.960580772265256e-05}
{"train_loss": 0.6804965734481812, "global_step": 11337, "epoch": 127, "lr": 9.96057350657239e-05}
{"train_loss": 0.6233886480331421, "global_step": 11338, "epoch": 127, "lr": 9.960566240212635e-05}
{"train_loss": 0.6309704780578613, "global_step": 11339, "epoch": 127, "lr": 9.960558973185991e-05}
{"train_loss": 0.5408706665039062, "global_step": 11340, "epoch": 127, "lr": 9.96055170549246e-05}
{"train_loss": 0.6173666715621948, "global_step": 11341, "epoch": 127, "lr": 9.960544437132044e-05}
{"train_loss": 0.6519336700439453, "global_step": 11342, "epoch": 127, "lr": 9.960537168104745e-05}
{"train_loss": 0.5942649841308594, "global_step": 11343, "epoch": 127, "lr": 9.96052989841056e-05}
{"train_loss": 0.621399998664856, "global_step": 11344, "epoch": 127, "lr": 9.960522628049493e-05}
{"train_loss": 0.6112152338027954, "global_step": 11345, "epoch": 127, "lr": 9.960515357021544e-05}
{"train_loss": 0.5757311582565308, "global_step": 11346, "epoch": 127, "lr": 9.960508085326715e-05}
{"train_loss": 0.6449123620986938, "global_step": 11347, "epoch": 127, "lr": 9.960500812965006e-05}
{"train_loss": 0.697008490562439, "global_step": 11348, "epoch": 127, "lr": 9.960493539936417e-05}
{"train_loss": 0.6436604857444763, "global_step": 11349, "epoch": 127, "lr": 9.960486266240951e-05}
{"train_loss": 0.6466944217681885, "global_step": 11350, "epoch": 127, "lr": 9.960478991878608e-05}
{"train_loss": 0.6474688649177551, "global_step": 11351, "epoch": 127, "lr": 9.960471716849389e-05}
{"train_loss": 0.7104652523994446, "global_step": 11352, "epoch": 127, "lr": 9.960464441153294e-05}
{"train_loss": 0.7678822875022888, "global_step": 11353, "epoch": 127, "lr": 9.960457164790327e-05}
{"train_loss": 0.7102740406990051, "global_step": 11354, "epoch": 127, "lr": 9.960449887760485e-05}
{"train_loss": 0.7374716401100159, "global_step": 11355, "epoch": 127, "lr": 9.960442610063774e-05}
{"train_loss": 0.6771941184997559, "global_step": 11356, "epoch": 127, "lr": 9.960435331700188e-05}
{"train_loss": 0.7815203070640564, "global_step": 11357, "epoch": 127, "lr": 9.960428052669735e-05}
{"train_loss": 0.6668558716773987, "global_step": 11358, "epoch": 127, "lr": 9.960420772972412e-05}
{"train_loss": 0.7143730521202087, "global_step": 11359, "epoch": 127, "lr": 9.96041349260822e-05}
{"train_loss": 0.7224210500717163, "global_step": 11360, "epoch": 127, "lr": 9.960406211577162e-05}
{"train_loss": 0.6178258657455444, "global_step": 11361, "epoch": 127, "lr": 9.960398929879237e-05}
{"train_loss": 0.5440270900726318, "global_step": 11362, "epoch": 127, "lr": 9.960391647514446e-05}
{"train_loss": 0.5668356418609619, "global_step": 11363, "epoch": 127, "lr": 9.960384364482794e-05}
{"train_loss": 0.6243000030517578, "global_step": 11364, "epoch": 127, "lr": 9.960377080784275e-05}
{"train_loss": 0.7042188048362732, "global_step": 11365, "epoch": 127, "lr": 9.960369796418896e-05}
{"train_loss": 0.6810243129730225, "global_step": 11366, "epoch": 127, "lr": 9.960362511386656e-05}
{"train_loss": 0.5952445268630981, "global_step": 11367, "epoch": 127, "lr": 9.960355225687554e-05}
{"train_loss": 0.6138179898262024, "global_step": 11368, "epoch": 127, "lr": 9.960347939321595e-05}
{"train_loss": 0.7454757690429688, "global_step": 11369, "epoch": 127, "lr": 9.960340652288776e-05}
{"train_loss": 0.5927509069442749, "global_step": 11370, "epoch": 127, "lr": 9.9603333645891e-05}
{"train_loss": 0.6132475137710571, "global_step": 11371, "epoch": 127, "lr": 9.960326076222567e-05}
{"train_loss": 0.9020955562591553, "global_step": 11372, "epoch": 127, "lr": 9.960318787189179e-05}
{"train_loss": 0.8216232657432556, "global_step": 11373, "epoch": 127, "lr": 9.960311497488937e-05}
{"train_loss": 0.7105273604393005, "global_step": 11374, "epoch": 127, "lr": 9.960304207121842e-05}
{"train_loss": 0.7512478232383728, "global_step": 11375, "epoch": 127, "lr": 9.960296916087894e-05}
{"train_loss": 0.6891874074935913, "global_step": 11376, "epoch": 127, "lr": 9.960289624387093e-05}
{"train_loss": 0.6872377991676331, "global_step": 11377, "epoch": 127, "lr": 9.960282332019444e-05}
{"train_loss": 0.91738361120224, "global_step": 11378, "epoch": 127, "lr": 9.960275038984946e-05}
{"train_loss": 0.7263413667678833, "global_step": 11379, "epoch": 127, "lr": 9.960267745283596e-05}
{"train_loss": 0.5908275842666626, "global_step": 11380, "epoch": 127, "lr": 9.960260450915401e-05}
{"train_loss": 0.6069777011871338, "global_step": 11381, "epoch": 127, "lr": 9.960253155880359e-05}
{"train_loss": 0.6040723323822021, "global_step": 11382, "epoch": 127, "lr": 9.960245860178471e-05}
{"train_loss": 0.7079957723617554, "global_step": 11383, "epoch": 127, "lr": 9.960238563809738e-05}
{"train_loss": 0.6912240386009216, "global_step": 11384, "epoch": 127, "lr": 9.960231266774163e-05}
{"train_loss": 0.6847153306007385, "global_step": 11385, "epoch": 127, "lr": 9.960223969071744e-05}
{"train_loss": 0.7635406255722046, "global_step": 11386, "epoch": 127, "lr": 9.960216670702485e-05}
{"train_loss": 0.6290208101272583, "global_step": 11387, "epoch": 127, "lr": 9.960209371666384e-05}
{"train_loss": 0.6890300512313843, "global_step": 11388, "epoch": 127, "lr": 9.960202071963442e-05}
{"train_loss": 0.7779288291931152, "global_step": 11389, "epoch": 127, "lr": 9.960194771593662e-05}
{"train_loss": 0.7107213139533997, "global_step": 11390, "epoch": 127, "lr": 9.960187470557046e-05}
{"train_loss": 0.6759207308962104, "global_step": 11391, "epoch": 127, "lr": 9.960180168853592e-05, "val_loss": 1.3094464540481567}
{"train_loss": 0.7152146100997925, "global_step": 11392, "epoch": 128, "lr": 9.960172866483303e-05}
{"train_loss": 0.6532211303710938, "global_step": 11393, "epoch": 128, "lr": 9.960165563446178e-05}
{"train_loss": 0.5916862487792969, "global_step": 11394, "epoch": 128, "lr": 9.96015825974222e-05}
{"train_loss": 0.6892299652099609, "global_step": 11395, "epoch": 128, "lr": 9.96015095537143e-05}
{"train_loss": 0.6023492217063904, "global_step": 11396, "epoch": 128, "lr": 9.960143650333807e-05}
{"train_loss": 0.708161473274231, "global_step": 11397, "epoch": 128, "lr": 9.960136344629354e-05}
{"train_loss": 0.5416432023048401, "global_step": 11398, "epoch": 128, "lr": 9.96012903825807e-05}
{"train_loss": 0.5983203649520874, "global_step": 11399, "epoch": 128, "lr": 9.960121731219958e-05}
{"train_loss": 0.6991464495658875, "global_step": 11400, "epoch": 128, "lr": 9.960114423515017e-05}
{"train_loss": 0.7465851306915283, "global_step": 11401, "epoch": 128, "lr": 9.960107115143251e-05}
{"train_loss": 0.7448861598968506, "global_step": 11402, "epoch": 128, "lr": 9.960099806104656e-05}
{"train_loss": 0.8079639673233032, "global_step": 11403, "epoch": 128, "lr": 9.960092496399238e-05}
{"train_loss": 0.7622173428535461, "global_step": 11404, "epoch": 128, "lr": 9.960085186026997e-05}
{"train_loss": 0.6093611121177673, "global_step": 11405, "epoch": 128, "lr": 9.960077874987932e-05}
{"train_loss": 0.6377443671226501, "global_step": 11406, "epoch": 128, "lr": 9.960070563282044e-05}
{"train_loss": 0.6951560974121094, "global_step": 11407, "epoch": 128, "lr": 9.960063250909335e-05}
{"train_loss": 0.8074764013290405, "global_step": 11408, "epoch": 128, "lr": 9.960055937869807e-05}
{"train_loss": 0.6595034599304199, "global_step": 11409, "epoch": 128, "lr": 9.960048624163461e-05}
{"train_loss": 0.6209426522254944, "global_step": 11410, "epoch": 128, "lr": 9.960041309790295e-05}
{"train_loss": 0.7101078629493713, "global_step": 11411, "epoch": 128, "lr": 9.960033994750313e-05}
{"train_loss": 0.6162475943565369, "global_step": 11412, "epoch": 128, "lr": 9.960026679043514e-05}
{"train_loss": 0.809986412525177, "global_step": 11413, "epoch": 128, "lr": 9.9600193626699e-05}
{"train_loss": 0.7949115037918091, "global_step": 11414, "epoch": 128, "lr": 9.960012045629473e-05}
{"train_loss": 0.7456519603729248, "global_step": 11415, "epoch": 128, "lr": 9.960004727922232e-05}
{"train_loss": 0.7831009030342102, "global_step": 11416, "epoch": 128, "lr": 9.959997409548179e-05}
{"train_loss": 0.7220712900161743, "global_step": 11417, "epoch": 128, "lr": 9.959990090507314e-05}
{"train_loss": 0.7263681292533875, "global_step": 11418, "epoch": 128, "lr": 9.959982770799639e-05}
{"train_loss": 0.7048364877700806, "global_step": 11419, "epoch": 128, "lr": 9.959975450425155e-05}
{"train_loss": 0.7642523050308228, "global_step": 11420, "epoch": 128, "lr": 9.959968129383863e-05}
{"train_loss": 0.7941723465919495, "global_step": 11421, "epoch": 128, "lr": 9.959960807675763e-05}
{"train_loss": 0.5996387600898743, "global_step": 11422, "epoch": 128, "lr": 9.95995348530086e-05}
{"train_loss": 0.7253806591033936, "global_step": 11423, "epoch": 128, "lr": 9.959946162259148e-05}
{"train_loss": 0.8081867098808289, "global_step": 11424, "epoch": 128, "lr": 9.959938838550633e-05}
{"train_loss": 0.7274817824363708, "global_step": 11425, "epoch": 128, "lr": 9.959931514175314e-05}
{"train_loss": 0.8469408750534058, "global_step": 11426, "epoch": 128, "lr": 9.959924189133194e-05}
{"train_loss": 0.738075852394104, "global_step": 11427, "epoch": 128, "lr": 9.95991686342427e-05}
{"train_loss": 0.8392181396484375, "global_step": 11428, "epoch": 128, "lr": 9.959909537048548e-05}
{"train_loss": 0.8501203656196594, "global_step": 11429, "epoch": 128, "lr": 9.959902210006026e-05}
{"train_loss": 0.8052924871444702, "global_step": 11430, "epoch": 128, "lr": 9.959894882296706e-05}
{"train_loss": 0.6220139861106873, "global_step": 11431, "epoch": 128, "lr": 9.959887553920588e-05}
{"train_loss": 0.8034501075744629, "global_step": 11432, "epoch": 128, "lr": 9.959880224877673e-05}
{"train_loss": 0.541018009185791, "global_step": 11433, "epoch": 128, "lr": 9.959872895167964e-05}
{"train_loss": 0.6979337334632874, "global_step": 11434, "epoch": 128, "lr": 9.95986556479146e-05}
{"train_loss": 0.5664918422698975, "global_step": 11435, "epoch": 128, "lr": 9.959858233748161e-05}
{"train_loss": 0.6978353261947632, "global_step": 11436, "epoch": 128, "lr": 9.959850902038071e-05}
{"train_loss": 1.062203288078308, "global_step": 11437, "epoch": 128, "lr": 9.959843569661189e-05}
{"train_loss": 0.691440999507904, "global_step": 11438, "epoch": 128, "lr": 9.959836236617518e-05}
{"train_loss": 0.8708195090293884, "global_step": 11439, "epoch": 128, "lr": 9.959828902907055e-05}
{"train_loss": 0.7215721607208252, "global_step": 11440, "epoch": 128, "lr": 9.959821568529806e-05}
{"train_loss": 0.8876327276229858, "global_step": 11441, "epoch": 128, "lr": 9.959814233485768e-05}
{"train_loss": 0.6347182393074036, "global_step": 11442, "epoch": 128, "lr": 9.959806897774943e-05}
{"train_loss": 0.8247647881507874, "global_step": 11443, "epoch": 128, "lr": 9.959799561397333e-05}
{"train_loss": 0.6686401963233948, "global_step": 11444, "epoch": 128, "lr": 9.959792224352939e-05}
{"train_loss": 0.5570375919342041, "global_step": 11445, "epoch": 128, "lr": 9.95978488664176e-05}
{"train_loss": 0.4961538016796112, "global_step": 11446, "epoch": 128, "lr": 9.9597775482638e-05}
{"train_loss": 0.629980206489563, "global_step": 11447, "epoch": 128, "lr": 9.959770209219058e-05}
{"train_loss": 0.6792713403701782, "global_step": 11448, "epoch": 128, "lr": 9.959762869507534e-05}
{"train_loss": 0.6528865694999695, "global_step": 11449, "epoch": 128, "lr": 9.959755529129232e-05}
{"train_loss": 0.6327557563781738, "global_step": 11450, "epoch": 128, "lr": 9.95974818808415e-05}
{"train_loss": 0.7479545474052429, "global_step": 11451, "epoch": 128, "lr": 9.959740846372292e-05}
{"train_loss": 0.6395286917686462, "global_step": 11452, "epoch": 128, "lr": 9.959733503993655e-05}
{"train_loss": 0.598494291305542, "global_step": 11453, "epoch": 128, "lr": 9.959726160948244e-05}
{"train_loss": 0.5868256688117981, "global_step": 11454, "epoch": 128, "lr": 9.959718817236059e-05}
{"train_loss": 0.6175201535224915, "global_step": 11455, "epoch": 128, "lr": 9.959711472857098e-05}
{"train_loss": 0.6129782199859619, "global_step": 11456, "epoch": 128, "lr": 9.959704127811366e-05}
{"train_loss": 0.6726096272468567, "global_step": 11457, "epoch": 128, "lr": 9.959696782098861e-05}
{"train_loss": 0.680099368095398, "global_step": 11458, "epoch": 128, "lr": 9.959689435719586e-05}
{"train_loss": 0.7436427474021912, "global_step": 11459, "epoch": 128, "lr": 9.95968208867354e-05}
{"train_loss": 0.7930916547775269, "global_step": 11460, "epoch": 128, "lr": 9.959674740960727e-05}
{"train_loss": 0.5702832937240601, "global_step": 11461, "epoch": 128, "lr": 9.959667392581145e-05}
{"train_loss": 0.7539655566215515, "global_step": 11462, "epoch": 128, "lr": 9.959660043534797e-05}
{"train_loss": 0.8594440817832947, "global_step": 11463, "epoch": 128, "lr": 9.959652693821682e-05}
{"train_loss": 0.5712141990661621, "global_step": 11464, "epoch": 128, "lr": 9.959645343441802e-05}
{"train_loss": 0.6137006878852844, "global_step": 11465, "epoch": 128, "lr": 9.959637992395159e-05}
{"train_loss": 0.689333438873291, "global_step": 11466, "epoch": 128, "lr": 9.959630640681753e-05}
{"train_loss": 0.6570460796356201, "global_step": 11467, "epoch": 128, "lr": 9.959623288301585e-05}
{"train_loss": 0.7613323330879211, "global_step": 11468, "epoch": 128, "lr": 9.959615935254655e-05}
{"train_loss": 0.6318531036376953, "global_step": 11469, "epoch": 128, "lr": 9.959608581540968e-05}
{"train_loss": 0.7339450120925903, "global_step": 11470, "epoch": 128, "lr": 9.95960122716052e-05}
{"train_loss": 0.6323111653327942, "global_step": 11471, "epoch": 128, "lr": 9.959593872113316e-05}
{"train_loss": 0.7440202236175537, "global_step": 11472, "epoch": 128, "lr": 9.959586516399353e-05}
{"train_loss": 0.6636159420013428, "global_step": 11473, "epoch": 128, "lr": 9.959579160018634e-05}
{"train_loss": 0.6300141215324402, "global_step": 11474, "epoch": 128, "lr": 9.95957180297116e-05}
{"train_loss": 0.5802423357963562, "global_step": 11475, "epoch": 128, "lr": 9.959564445256932e-05}
{"train_loss": 0.6988267302513123, "global_step": 11476, "epoch": 128, "lr": 9.959557086875953e-05}
{"train_loss": 0.6613062620162964, "global_step": 11477, "epoch": 128, "lr": 9.95954972782822e-05}
{"train_loss": 0.59076988697052, "global_step": 11478, "epoch": 128, "lr": 9.959542368113737e-05}
{"train_loss": 0.7070817351341248, "global_step": 11479, "epoch": 128, "lr": 9.959535007732504e-05}
{"train_loss": 0.6963134511802973, "global_step": 11480, "epoch": 128, "lr": 9.959527646684521e-05, "val_loss": 1.3292280435562134}
{"train_loss": 0.7414928674697876, "global_step": 11481, "epoch": 129, "lr": 9.959520284969789e-05}
{"train_loss": 0.6343297958374023, "global_step": 11482, "epoch": 129, "lr": 9.959512922588313e-05}
{"train_loss": 0.7362020015716553, "global_step": 11483, "epoch": 129, "lr": 9.95950555954009e-05}
{"train_loss": 0.6822114586830139, "global_step": 11484, "epoch": 129, "lr": 9.95949819582512e-05}
{"train_loss": 0.6214189529418945, "global_step": 11485, "epoch": 129, "lr": 9.959490831443408e-05}
{"train_loss": 0.5999869108200073, "global_step": 11486, "epoch": 129, "lr": 9.959483466394952e-05}
{"train_loss": 0.7001491785049438, "global_step": 11487, "epoch": 129, "lr": 9.959476100679753e-05}
{"train_loss": 0.6808584928512573, "global_step": 11488, "epoch": 129, "lr": 9.959468734297814e-05}
{"train_loss": 0.5829554796218872, "global_step": 11489, "epoch": 129, "lr": 9.959461367249134e-05}
{"train_loss": 0.6399524211883545, "global_step": 11490, "epoch": 129, "lr": 9.959453999533716e-05}
{"train_loss": 0.7533211708068848, "global_step": 11491, "epoch": 129, "lr": 9.95944663115156e-05}
{"train_loss": 0.7301114201545715, "global_step": 11492, "epoch": 129, "lr": 9.959439262102667e-05}
{"train_loss": 0.665583610534668, "global_step": 11493, "epoch": 129, "lr": 9.959431892387037e-05}
{"train_loss": 0.5813511610031128, "global_step": 11494, "epoch": 129, "lr": 9.959424522004672e-05}
{"train_loss": 0.576958417892456, "global_step": 11495, "epoch": 129, "lr": 9.959417150955573e-05}
{"train_loss": 0.8228389024734497, "global_step": 11496, "epoch": 129, "lr": 9.95940977923974e-05}
{"train_loss": 0.5993748903274536, "global_step": 11497, "epoch": 129, "lr": 9.959402406857176e-05}
{"train_loss": 0.8459054827690125, "global_step": 11498, "epoch": 129, "lr": 9.959395033807879e-05}
{"train_loss": 0.6905331611633301, "global_step": 11499, "epoch": 129, "lr": 9.959387660091853e-05}
{"train_loss": 0.6154729723930359, "global_step": 11500, "epoch": 129, "lr": 9.9593802857091e-05}
{"train_loss": 0.7166763544082642, "global_step": 11501, "epoch": 129, "lr": 9.959372910659615e-05}
{"train_loss": 0.7349959015846252, "global_step": 11502, "epoch": 129, "lr": 9.959365534943404e-05}
{"train_loss": 0.6844747066497803, "global_step": 11503, "epoch": 129, "lr": 9.959358158560467e-05}
{"train_loss": 0.7966539263725281, "global_step": 11504, "epoch": 129, "lr": 9.959350781510806e-05}
{"train_loss": 0.6083391904830933, "global_step": 11505, "epoch": 129, "lr": 9.959343403794418e-05}
{"train_loss": 0.638582706451416, "global_step": 11506, "epoch": 129, "lr": 9.959336025411309e-05}
{"train_loss": 0.6483765244483948, "global_step": 11507, "epoch": 129, "lr": 9.959328646361477e-05}
{"train_loss": 0.8079931735992432, "global_step": 11508, "epoch": 129, "lr": 9.959321266644924e-05}
{"train_loss": 0.6940687894821167, "global_step": 11509, "epoch": 129, "lr": 9.95931388626165e-05}
{"train_loss": 0.8266767859458923, "global_step": 11510, "epoch": 129, "lr": 9.959306505211656e-05}
{"train_loss": 0.7844815850257874, "global_step": 11511, "epoch": 129, "lr": 9.959299123494945e-05}
{"train_loss": 0.6403475999832153, "global_step": 11512, "epoch": 129, "lr": 9.959291741111517e-05}
{"train_loss": 0.6354390382766724, "global_step": 11513, "epoch": 129, "lr": 9.959284358061372e-05}
{"train_loss": 0.6013005375862122, "global_step": 11514, "epoch": 129, "lr": 9.959276974344511e-05}
{"train_loss": 0.6382686495780945, "global_step": 11515, "epoch": 129, "lr": 9.959269589960937e-05}
{"train_loss": 0.6913471817970276, "global_step": 11516, "epoch": 129, "lr": 9.959262204910648e-05}
{"train_loss": 0.7241684198379517, "global_step": 11517, "epoch": 129, "lr": 9.959254819193648e-05}
{"train_loss": 0.6729288101196289, "global_step": 11518, "epoch": 129, "lr": 9.959247432809937e-05}
{"train_loss": 0.5958738327026367, "global_step": 11519, "epoch": 129, "lr": 9.959240045759516e-05}
{"train_loss": 0.5906603336334229, "global_step": 11520, "epoch": 129, "lr": 9.959232658042384e-05}
{"train_loss": 0.5618190169334412, "global_step": 11521, "epoch": 129, "lr": 9.959225269658544e-05}
{"train_loss": 0.6773396134376526, "global_step": 11522, "epoch": 129, "lr": 9.959217880607997e-05}
{"train_loss": 0.7021745443344116, "global_step": 11523, "epoch": 129, "lr": 9.959210490890744e-05}
{"train_loss": 0.7744986414909363, "global_step": 11524, "epoch": 129, "lr": 9.959203100506785e-05}
{"train_loss": 0.7542715072631836, "global_step": 11525, "epoch": 129, "lr": 9.959195709456121e-05}
{"train_loss": 0.6434530019760132, "global_step": 11526, "epoch": 129, "lr": 9.959188317738755e-05}
{"train_loss": 0.6433913707733154, "global_step": 11527, "epoch": 129, "lr": 9.959180925354685e-05}
{"train_loss": 0.5668174624443054, "global_step": 11528, "epoch": 129, "lr": 9.959173532303914e-05}
{"train_loss": 0.628947913646698, "global_step": 11529, "epoch": 129, "lr": 9.959166138586443e-05}
{"train_loss": 0.6941638588905334, "global_step": 11530, "epoch": 129, "lr": 9.959158744202272e-05}
{"train_loss": 0.6468572616577148, "global_step": 11531, "epoch": 129, "lr": 9.959151349151403e-05}
{"train_loss": 0.6174784898757935, "global_step": 11532, "epoch": 129, "lr": 9.959143953433838e-05}
{"train_loss": 0.7136391997337341, "global_step": 11533, "epoch": 129, "lr": 9.959136557049574e-05}
{"train_loss": 0.522165834903717, "global_step": 11534, "epoch": 129, "lr": 9.959129159998616e-05}
{"train_loss": 0.5832744836807251, "global_step": 11535, "epoch": 129, "lr": 9.959121762280963e-05}
{"train_loss": 0.6141568422317505, "global_step": 11536, "epoch": 129, "lr": 9.959114363896616e-05}
{"train_loss": 0.6116526126861572, "global_step": 11537, "epoch": 129, "lr": 9.959106964845579e-05}
{"train_loss": 0.5351780652999878, "global_step": 11538, "epoch": 129, "lr": 9.959099565127848e-05}
{"train_loss": 0.7152122259140015, "global_step": 11539, "epoch": 129, "lr": 9.959092164743426e-05}
{"train_loss": 0.7045205235481262, "global_step": 11540, "epoch": 129, "lr": 9.959084763692316e-05}
{"train_loss": 0.7354809045791626, "global_step": 11541, "epoch": 129, "lr": 9.959077361974518e-05}
{"train_loss": 0.6343475580215454, "global_step": 11542, "epoch": 129, "lr": 9.959069959590032e-05}
{"train_loss": 0.5893087983131409, "global_step": 11543, "epoch": 129, "lr": 9.959062556538858e-05}
{"train_loss": 0.6320110559463501, "global_step": 11544, "epoch": 129, "lr": 9.959055152821e-05}
{"train_loss": 0.7425867915153503, "global_step": 11545, "epoch": 129, "lr": 9.959047748436457e-05}
{"train_loss": 0.4993457794189453, "global_step": 11546, "epoch": 129, "lr": 9.959040343385231e-05}
{"train_loss": 0.682890772819519, "global_step": 11547, "epoch": 129, "lr": 9.959032937667322e-05}
{"train_loss": 0.7013697624206543, "global_step": 11548, "epoch": 129, "lr": 9.959025531282732e-05}
{"train_loss": 0.6801955699920654, "global_step": 11549, "epoch": 129, "lr": 9.95901812423146e-05}
{"train_loss": 0.6477770209312439, "global_step": 11550, "epoch": 129, "lr": 9.959010716513511e-05}
{"train_loss": 0.5164970755577087, "global_step": 11551, "epoch": 129, "lr": 9.959003308128881e-05}
{"train_loss": 0.8276128172874451, "global_step": 11552, "epoch": 129, "lr": 9.958995899077574e-05}
{"train_loss": 0.5586009621620178, "global_step": 11553, "epoch": 129, "lr": 9.958988489359591e-05}
{"train_loss": 0.6506814360618591, "global_step": 11554, "epoch": 129, "lr": 9.958981078974932e-05}
{"train_loss": 0.716643750667572, "global_step": 11555, "epoch": 129, "lr": 9.958973667923599e-05}
{"train_loss": 0.6576640009880066, "global_step": 11556, "epoch": 129, "lr": 9.958966256205593e-05}
{"train_loss": 0.6002255082130432, "global_step": 11557, "epoch": 129, "lr": 9.958958843820913e-05}
{"train_loss": 0.627842366695404, "global_step": 11558, "epoch": 129, "lr": 9.958951430769562e-05}
{"train_loss": 0.7113860845565796, "global_step": 11559, "epoch": 129, "lr": 9.95894401705154e-05}
{"train_loss": 0.8622328042984009, "global_step": 11560, "epoch": 129, "lr": 9.958936602666849e-05}
{"train_loss": 0.6114657521247864, "global_step": 11561, "epoch": 129, "lr": 9.958929187615489e-05}
{"train_loss": 0.7235924601554871, "global_step": 11562, "epoch": 129, "lr": 9.958921771897462e-05}
{"train_loss": 0.5123782157897949, "global_step": 11563, "epoch": 129, "lr": 9.958914355512768e-05}
{"train_loss": 0.6287001967430115, "global_step": 11564, "epoch": 129, "lr": 9.958906938461407e-05}
{"train_loss": 0.7671688795089722, "global_step": 11565, "epoch": 129, "lr": 9.958899520743383e-05}
{"train_loss": 0.7338366508483887, "global_step": 11566, "epoch": 129, "lr": 9.958892102358697e-05}
{"train_loss": 0.8306894898414612, "global_step": 11567, "epoch": 129, "lr": 9.958884683307345e-05}
{"train_loss": 0.6818119287490845, "global_step": 11568, "epoch": 129, "lr": 9.958877263589334e-05}
{"train_loss": 0.6700406811210546, "global_step": 11569, "epoch": 129, "lr": 9.958869843204661e-05, "val_loss": 1.345611572265625}
{"train_loss": 0.6599117517471313, "global_step": 11570, "epoch": 130, "lr": 9.958862422153329e-05}
{"train_loss": 0.5404897928237915, "global_step": 11571, "epoch": 130, "lr": 9.958855000435339e-05}
{"train_loss": 0.7837885618209839, "global_step": 11572, "epoch": 130, "lr": 9.95884757805069e-05}
{"train_loss": 0.7662044763565063, "global_step": 11573, "epoch": 130, "lr": 9.958840154999385e-05}
{"train_loss": 0.6468397974967957, "global_step": 11574, "epoch": 130, "lr": 9.958832731281425e-05}
{"train_loss": 0.5719356536865234, "global_step": 11575, "epoch": 130, "lr": 9.958825306896808e-05}
{"train_loss": 0.6968955397605896, "global_step": 11576, "epoch": 130, "lr": 9.95881788184554e-05}
{"train_loss": 0.6084499955177307, "global_step": 11577, "epoch": 130, "lr": 9.958810456127619e-05}
{"train_loss": 0.619611382484436, "global_step": 11578, "epoch": 130, "lr": 9.958803029743044e-05}
{"train_loss": 0.634631872177124, "global_step": 11579, "epoch": 130, "lr": 9.958795602691821e-05}
{"train_loss": 0.7302225232124329, "global_step": 11580, "epoch": 130, "lr": 9.958788174973947e-05}
{"train_loss": 0.6760582327842712, "global_step": 11581, "epoch": 130, "lr": 9.958780746589427e-05}
{"train_loss": 0.7659623622894287, "global_step": 11582, "epoch": 130, "lr": 9.958773317538257e-05}
{"train_loss": 0.5901491045951843, "global_step": 11583, "epoch": 130, "lr": 9.95876588782044e-05}
{"train_loss": 0.663550078868866, "global_step": 11584, "epoch": 130, "lr": 9.958758457435978e-05}
{"train_loss": 0.6470488905906677, "global_step": 11585, "epoch": 130, "lr": 9.958751026384872e-05}
{"train_loss": 0.736655056476593, "global_step": 11586, "epoch": 130, "lr": 9.958743594667121e-05}
{"train_loss": 0.6606142520904541, "global_step": 11587, "epoch": 130, "lr": 9.958736162282728e-05}
{"train_loss": 0.7102662920951843, "global_step": 11588, "epoch": 130, "lr": 9.958728729231695e-05}
{"train_loss": 0.6856556534767151, "global_step": 11589, "epoch": 130, "lr": 9.95872129551402e-05}
{"train_loss": 0.6400895714759827, "global_step": 11590, "epoch": 130, "lr": 9.958713861129705e-05}
{"train_loss": 0.6185804009437561, "global_step": 11591, "epoch": 130, "lr": 9.95870642607875e-05}
{"train_loss": 0.5167633295059204, "global_step": 11592, "epoch": 130, "lr": 9.958698990361159e-05}
{"train_loss": 0.6934897303581238, "global_step": 11593, "epoch": 130, "lr": 9.958691553976931e-05}
{"train_loss": 0.6337194442749023, "global_step": 11594, "epoch": 130, "lr": 9.958684116926069e-05}
{"train_loss": 0.6196886301040649, "global_step": 11595, "epoch": 130, "lr": 9.958676679208571e-05}
{"train_loss": 0.5917817950248718, "global_step": 11596, "epoch": 130, "lr": 9.958669240824439e-05}
{"train_loss": 0.6770910024642944, "global_step": 11597, "epoch": 130, "lr": 9.958661801773675e-05}
{"train_loss": 0.5982076525688171, "global_step": 11598, "epoch": 130, "lr": 9.95865436205628e-05}
{"train_loss": 0.7946435213088989, "global_step": 11599, "epoch": 130, "lr": 9.958646921672253e-05}
{"train_loss": 0.5702728033065796, "global_step": 11600, "epoch": 130, "lr": 9.958639480621596e-05}
{"train_loss": 0.5836575031280518, "global_step": 11601, "epoch": 130, "lr": 9.958632038904311e-05}
{"train_loss": 0.5765814781188965, "global_step": 11602, "epoch": 130, "lr": 9.958624596520398e-05}
{"train_loss": 0.6214801669120789, "global_step": 11603, "epoch": 130, "lr": 9.95861715346986e-05}
{"train_loss": 0.6683798432350159, "global_step": 11604, "epoch": 130, "lr": 9.958609709752695e-05}
{"train_loss": 0.6333618760108948, "global_step": 11605, "epoch": 130, "lr": 9.958602265368906e-05}
{"train_loss": 0.673943042755127, "global_step": 11606, "epoch": 130, "lr": 9.958594820318492e-05}
{"train_loss": 0.5925513505935669, "global_step": 11607, "epoch": 130, "lr": 9.958587374601456e-05}
{"train_loss": 0.6290651559829712, "global_step": 11608, "epoch": 130, "lr": 9.958579928217799e-05}
{"train_loss": 0.5860056281089783, "global_step": 11609, "epoch": 130, "lr": 9.958572481167521e-05}
{"train_loss": 0.9564297795295715, "global_step": 11610, "epoch": 130, "lr": 9.958565033450625e-05}
{"train_loss": 0.6135431528091431, "global_step": 11611, "epoch": 130, "lr": 9.958557585067108e-05}
{"train_loss": 0.6583072543144226, "global_step": 11612, "epoch": 130, "lr": 9.958550136016975e-05}
{"train_loss": 0.6485157012939453, "global_step": 11613, "epoch": 130, "lr": 9.958542686300222e-05}
{"train_loss": 0.6290035843849182, "global_step": 11614, "epoch": 130, "lr": 9.958535235916857e-05}
{"train_loss": 0.7293112277984619, "global_step": 11615, "epoch": 130, "lr": 9.958527784866877e-05}
{"train_loss": 0.7436935901641846, "global_step": 11616, "epoch": 130, "lr": 9.958520333150283e-05}
{"train_loss": 0.5796096920967102, "global_step": 11617, "epoch": 130, "lr": 9.958512880767076e-05}
{"train_loss": 0.6634909510612488, "global_step": 11618, "epoch": 130, "lr": 9.958505427717258e-05}
{"train_loss": 0.8216122984886169, "global_step": 11619, "epoch": 130, "lr": 9.958497974000828e-05}
{"train_loss": 0.6464424729347229, "global_step": 11620, "epoch": 130, "lr": 9.958490519617789e-05}
{"train_loss": 0.6205710768699646, "global_step": 11621, "epoch": 130, "lr": 9.958483064568142e-05}
{"train_loss": 0.7315514087677002, "global_step": 11622, "epoch": 130, "lr": 9.958475608851888e-05}
{"train_loss": 0.7078006267547607, "global_step": 11623, "epoch": 130, "lr": 9.958468152469025e-05}
{"train_loss": 0.7234606146812439, "global_step": 11624, "epoch": 130, "lr": 9.958460695419558e-05}
{"train_loss": 0.7193102836608887, "global_step": 11625, "epoch": 130, "lr": 9.958453237703486e-05}
{"train_loss": 0.8068596720695496, "global_step": 11626, "epoch": 130, "lr": 9.958445779320811e-05}
{"train_loss": 0.6241634488105774, "global_step": 11627, "epoch": 130, "lr": 9.958438320271533e-05}
{"train_loss": 0.6169148683547974, "global_step": 11628, "epoch": 130, "lr": 9.958430860555652e-05}
{"train_loss": 0.6290795207023621, "global_step": 11629, "epoch": 130, "lr": 9.958423400173173e-05}
{"train_loss": 0.6800872683525085, "global_step": 11630, "epoch": 130, "lr": 9.958415939124093e-05}
{"train_loss": 0.6955720782279968, "global_step": 11631, "epoch": 130, "lr": 9.958408477408414e-05}
{"train_loss": 0.5846821069717407, "global_step": 11632, "epoch": 130, "lr": 9.958401015026137e-05}
{"train_loss": 0.9335048794746399, "global_step": 11633, "epoch": 130, "lr": 9.958393551977265e-05}
{"train_loss": 0.6197296380996704, "global_step": 11634, "epoch": 130, "lr": 9.958386088261797e-05}
{"train_loss": 0.7095643281936646, "global_step": 11635, "epoch": 130, "lr": 9.958378623879733e-05}
{"train_loss": 0.6662729978561401, "global_step": 11636, "epoch": 130, "lr": 9.958371158831078e-05}
{"train_loss": 0.690066933631897, "global_step": 11637, "epoch": 130, "lr": 9.958363693115829e-05}
{"train_loss": 0.6502859592437744, "global_step": 11638, "epoch": 130, "lr": 9.958356226733989e-05}
{"train_loss": 0.6529519557952881, "global_step": 11639, "epoch": 130, "lr": 9.958348759685559e-05}
{"train_loss": 0.7672111988067627, "global_step": 11640, "epoch": 130, "lr": 9.958341291970537e-05}
{"train_loss": 0.7311532497406006, "global_step": 11641, "epoch": 130, "lr": 9.958333823588927e-05}
{"train_loss": 0.7695040702819824, "global_step": 11642, "epoch": 130, "lr": 9.958326354540731e-05}
{"train_loss": 0.6233802437782288, "global_step": 11643, "epoch": 130, "lr": 9.958318884825948e-05}
{"train_loss": 0.7830106616020203, "global_step": 11644, "epoch": 130, "lr": 9.958311414444579e-05}
{"train_loss": 0.6407781839370728, "global_step": 11645, "epoch": 130, "lr": 9.958303943396626e-05}
{"train_loss": 0.5860528349876404, "global_step": 11646, "epoch": 130, "lr": 9.95829647168209e-05}
{"train_loss": 0.7632555961608887, "global_step": 11647, "epoch": 130, "lr": 9.958288999300969e-05}
{"train_loss": 0.6154902577400208, "global_step": 11648, "epoch": 130, "lr": 9.958281526253269e-05}
{"train_loss": 0.6684454083442688, "global_step": 11649, "epoch": 130, "lr": 9.958274052538988e-05}
{"train_loss": 0.7728271484375, "global_step": 11650, "epoch": 130, "lr": 9.958266578158127e-05}
{"train_loss": 0.6836830377578735, "global_step": 11651, "epoch": 130, "lr": 9.958259103110688e-05}
{"train_loss": 0.8900863528251648, "global_step": 11652, "epoch": 130, "lr": 9.95825162739667e-05}
{"train_loss": 0.8308590650558472, "global_step": 11653, "epoch": 130, "lr": 9.958244151016076e-05}
{"train_loss": 0.6935245394706726, "global_step": 11654, "epoch": 130, "lr": 9.958236673968908e-05}
{"train_loss": 0.5634670257568359, "global_step": 11655, "epoch": 130, "lr": 9.958229196255164e-05}
{"train_loss": 0.6955649256706238, "global_step": 11656, "epoch": 130, "lr": 9.958221717874846e-05}
{"train_loss": 0.6376926302909851, "global_step": 11657, "epoch": 130, "lr": 9.958214238827959e-05}
{"train_loss": 0.6767210739382198, "global_step": 11658, "epoch": 130, "lr": 9.958206759114497e-05, "val_loss": 1.3335589170455933, "train_action_mse_error": 47.17362976074219}
{"train_loss": 0.6543332934379578, "global_step": 11659, "epoch": 131, "lr": 9.958199278734466e-05}
{"train_loss": 0.7914316058158875, "global_step": 11660, "epoch": 131, "lr": 9.958191797687866e-05}
{"train_loss": 0.7038342356681824, "global_step": 11661, "epoch": 131, "lr": 9.958184315974697e-05}
{"train_loss": 0.5989619493484497, "global_step": 11662, "epoch": 131, "lr": 9.958176833594959e-05}
{"train_loss": 0.6883972883224487, "global_step": 11663, "epoch": 131, "lr": 9.958169350548657e-05}
{"train_loss": 0.7207601070404053, "global_step": 11664, "epoch": 131, "lr": 9.958161866835788e-05}
{"train_loss": 0.6233925223350525, "global_step": 11665, "epoch": 131, "lr": 9.958154382456355e-05}
{"train_loss": 0.7605701684951782, "global_step": 11666, "epoch": 131, "lr": 9.958146897410359e-05}
{"train_loss": 0.6860671043395996, "global_step": 11667, "epoch": 131, "lr": 9.958139411697799e-05}
{"train_loss": 0.5497453212738037, "global_step": 11668, "epoch": 131, "lr": 9.958131925318679e-05}
{"train_loss": 0.7918917536735535, "global_step": 11669, "epoch": 131, "lr": 9.958124438272999e-05}
{"train_loss": 0.5677180290222168, "global_step": 11670, "epoch": 131, "lr": 9.958116950560758e-05}
{"train_loss": 0.485943466424942, "global_step": 11671, "epoch": 131, "lr": 9.958109462181961e-05}
{"train_loss": 0.7903256416320801, "global_step": 11672, "epoch": 131, "lr": 9.958101973136604e-05}
{"train_loss": 0.7612888216972351, "global_step": 11673, "epoch": 131, "lr": 9.958094483424691e-05}
{"train_loss": 0.5909523367881775, "global_step": 11674, "epoch": 131, "lr": 9.958086993046225e-05}
{"train_loss": 0.7417091727256775, "global_step": 11675, "epoch": 131, "lr": 9.958079502001203e-05}
{"train_loss": 0.6527801752090454, "global_step": 11676, "epoch": 131, "lr": 9.958072010289628e-05}
{"train_loss": 0.6794105768203735, "global_step": 11677, "epoch": 131, "lr": 9.9580645179115e-05}
{"train_loss": 0.693655788898468, "global_step": 11678, "epoch": 131, "lr": 9.95805702486682e-05}
{"train_loss": 0.7844676971435547, "global_step": 11679, "epoch": 131, "lr": 9.958049531155591e-05}
{"train_loss": 0.7221094369888306, "global_step": 11680, "epoch": 131, "lr": 9.958042036777812e-05}
{"train_loss": 0.6015941500663757, "global_step": 11681, "epoch": 131, "lr": 9.958034541733485e-05}
{"train_loss": 0.6417996883392334, "global_step": 11682, "epoch": 131, "lr": 9.95802704602261e-05}
{"train_loss": 0.6185952425003052, "global_step": 11683, "epoch": 131, "lr": 9.95801954964519e-05}
{"train_loss": 0.5994492769241333, "global_step": 11684, "epoch": 131, "lr": 9.958012052601224e-05}
{"train_loss": 0.777525007724762, "global_step": 11685, "epoch": 131, "lr": 9.958004554890714e-05}
{"train_loss": 0.7120614051818848, "global_step": 11686, "epoch": 131, "lr": 9.957997056513662e-05}
{"train_loss": 0.7461564540863037, "global_step": 11687, "epoch": 131, "lr": 9.957989557470065e-05}
{"train_loss": 0.6222529411315918, "global_step": 11688, "epoch": 131, "lr": 9.95798205775993e-05}
{"train_loss": 0.665390133857727, "global_step": 11689, "epoch": 131, "lr": 9.957974557383252e-05}
{"train_loss": 0.5637562870979309, "global_step": 11690, "epoch": 131, "lr": 9.957967056340034e-05}
{"train_loss": 0.861281156539917, "global_step": 11691, "epoch": 131, "lr": 9.95795955463028e-05}
{"train_loss": 0.8780655860900879, "global_step": 11692, "epoch": 131, "lr": 9.957952052253987e-05}
{"train_loss": 0.6936920285224915, "global_step": 11693, "epoch": 131, "lr": 9.95794454921116e-05}
{"train_loss": 0.6987213492393494, "global_step": 11694, "epoch": 131, "lr": 9.957937045501796e-05}
{"train_loss": 0.5133016109466553, "global_step": 11695, "epoch": 131, "lr": 9.957929541125899e-05}
{"train_loss": 0.7107584476470947, "global_step": 11696, "epoch": 131, "lr": 9.957922036083467e-05}
{"train_loss": 0.761898398399353, "global_step": 11697, "epoch": 131, "lr": 9.957914530374505e-05}
{"train_loss": 0.7248618602752686, "global_step": 11698, "epoch": 131, "lr": 9.95790702399901e-05}
{"train_loss": 0.6841859817504883, "global_step": 11699, "epoch": 131, "lr": 9.957899516956985e-05}
{"train_loss": 0.6941262483596802, "global_step": 11700, "epoch": 131, "lr": 9.957892009248433e-05}
{"train_loss": 0.710508406162262, "global_step": 11701, "epoch": 131, "lr": 9.957884500873352e-05}
{"train_loss": 0.6494197845458984, "global_step": 11702, "epoch": 131, "lr": 9.957876991831742e-05}
{"train_loss": 0.6587401628494263, "global_step": 11703, "epoch": 131, "lr": 9.957869482123607e-05}
{"train_loss": 0.7217851877212524, "global_step": 11704, "epoch": 131, "lr": 9.957861971748947e-05}
{"train_loss": 0.7543371319770813, "global_step": 11705, "epoch": 131, "lr": 9.957854460707763e-05}
{"train_loss": 0.7594351768493652, "global_step": 11706, "epoch": 131, "lr": 9.957846949000056e-05}
{"train_loss": 0.6104241609573364, "global_step": 11707, "epoch": 131, "lr": 9.957839436625826e-05}
{"train_loss": 0.7043632864952087, "global_step": 11708, "epoch": 131, "lr": 9.957831923585075e-05}
{"train_loss": 0.6107959151268005, "global_step": 11709, "epoch": 131, "lr": 9.957824409877806e-05}
{"train_loss": 0.76157546043396, "global_step": 11710, "epoch": 131, "lr": 9.957816895504016e-05}
{"train_loss": 0.7526168823242188, "global_step": 11711, "epoch": 131, "lr": 9.957809380463707e-05}
{"train_loss": 0.6740256547927856, "global_step": 11712, "epoch": 131, "lr": 9.957801864756883e-05}
{"train_loss": 0.5287610292434692, "global_step": 11713, "epoch": 131, "lr": 9.957794348383542e-05}
{"train_loss": 0.8067739009857178, "global_step": 11714, "epoch": 131, "lr": 9.957786831343686e-05}
{"train_loss": 0.6184844970703125, "global_step": 11715, "epoch": 131, "lr": 9.957779313637316e-05}
{"train_loss": 0.587656557559967, "global_step": 11716, "epoch": 131, "lr": 9.957771795264433e-05}
{"train_loss": 0.6305427551269531, "global_step": 11717, "epoch": 131, "lr": 9.957764276225039e-05}
{"train_loss": 0.5545933842658997, "global_step": 11718, "epoch": 131, "lr": 9.957756756519132e-05}
{"train_loss": 0.7232531905174255, "global_step": 11719, "epoch": 131, "lr": 9.957749236146717e-05}
{"train_loss": 0.4791886806488037, "global_step": 11720, "epoch": 131, "lr": 9.957741715107792e-05}
{"train_loss": 0.7044406533241272, "global_step": 11721, "epoch": 131, "lr": 9.957734193402359e-05}
{"train_loss": 0.5709385275840759, "global_step": 11722, "epoch": 131, "lr": 9.95772667103042e-05}
{"train_loss": 0.7425514459609985, "global_step": 11723, "epoch": 131, "lr": 9.957719147991976e-05}
{"train_loss": 0.5391027331352234, "global_step": 11724, "epoch": 131, "lr": 9.957711624287024e-05}
{"train_loss": 0.6133477687835693, "global_step": 11725, "epoch": 131, "lr": 9.95770409991557e-05}
{"train_loss": 0.6212627291679382, "global_step": 11726, "epoch": 131, "lr": 9.957696574877613e-05}
{"train_loss": 0.5558147430419922, "global_step": 11727, "epoch": 131, "lr": 9.957689049173154e-05}
{"train_loss": 0.659508466720581, "global_step": 11728, "epoch": 131, "lr": 9.957681522802194e-05}
{"train_loss": 0.716680645942688, "global_step": 11729, "epoch": 131, "lr": 9.957673995764733e-05}
{"train_loss": 0.7444804906845093, "global_step": 11730, "epoch": 131, "lr": 9.957666468060775e-05}
{"train_loss": 0.6877719759941101, "global_step": 11731, "epoch": 131, "lr": 9.957658939690319e-05}
{"train_loss": 0.6812518239021301, "global_step": 11732, "epoch": 131, "lr": 9.957651410653365e-05}
{"train_loss": 0.720029890537262, "global_step": 11733, "epoch": 131, "lr": 9.957643880949916e-05}
{"train_loss": 0.7109862565994263, "global_step": 11734, "epoch": 131, "lr": 9.957636350579972e-05}
{"train_loss": 0.510021984577179, "global_step": 11735, "epoch": 131, "lr": 9.957628819543534e-05}
{"train_loss": 0.610675036907196, "global_step": 11736, "epoch": 131, "lr": 9.957621287840603e-05}
{"train_loss": 0.6789072751998901, "global_step": 11737, "epoch": 131, "lr": 9.957613755471183e-05}
{"train_loss": 0.6281488537788391, "global_step": 11738, "epoch": 131, "lr": 9.95760622243527e-05}
{"train_loss": 0.6071531176567078, "global_step": 11739, "epoch": 131, "lr": 9.957598688732865e-05}
{"train_loss": 0.6039722561836243, "global_step": 11740, "epoch": 131, "lr": 9.957591154363975e-05}
{"train_loss": 0.7616117596626282, "global_step": 11741, "epoch": 131, "lr": 9.957583619328595e-05}
{"train_loss": 0.7070837616920471, "global_step": 11742, "epoch": 131, "lr": 9.95757608362673e-05}
{"train_loss": 0.5981135368347168, "global_step": 11743, "epoch": 131, "lr": 9.957568547258379e-05}
{"train_loss": 0.65012526512146, "global_step": 11744, "epoch": 131, "lr": 9.957561010223542e-05}
{"train_loss": 0.7352736592292786, "global_step": 11745, "epoch": 131, "lr": 9.957553472522223e-05}
{"train_loss": 0.6146177649497986, "global_step": 11746, "epoch": 131, "lr": 9.957545934154419e-05}
{"train_loss": 0.6730560683802272, "global_step": 11747, "epoch": 131, "lr": 9.957538395120137e-05, "val_loss": 1.3368492126464844}
{"train_loss": 0.5032333135604858, "global_step": 11748, "epoch": 132, "lr": 9.95753085541937e-05}
{"train_loss": 0.7715491056442261, "global_step": 11749, "epoch": 132, "lr": 9.957523315052126e-05}
{"train_loss": 0.5388381481170654, "global_step": 11750, "epoch": 132, "lr": 9.957515774018404e-05}
{"train_loss": 0.7195245027542114, "global_step": 11751, "epoch": 132, "lr": 9.957508232318204e-05}
{"train_loss": 0.6826804280281067, "global_step": 11752, "epoch": 132, "lr": 9.957500689951527e-05}
{"train_loss": 0.5441259741783142, "global_step": 11753, "epoch": 132, "lr": 9.957493146918373e-05}
{"train_loss": 0.6520112156867981, "global_step": 11754, "epoch": 132, "lr": 9.957485603218746e-05}
{"train_loss": 0.6381503343582153, "global_step": 11755, "epoch": 132, "lr": 9.957478058852647e-05}
{"train_loss": 0.7138903141021729, "global_step": 11756, "epoch": 132, "lr": 9.957470513820072e-05}
{"train_loss": 0.5987322926521301, "global_step": 11757, "epoch": 132, "lr": 9.957462968121027e-05}
{"train_loss": 0.5774551630020142, "global_step": 11758, "epoch": 132, "lr": 9.957455421755513e-05}
{"train_loss": 0.5961225628852844, "global_step": 11759, "epoch": 132, "lr": 9.957447874723529e-05}
{"train_loss": 0.6760304570198059, "global_step": 11760, "epoch": 132, "lr": 9.957440327025076e-05}
{"train_loss": 0.7341333627700806, "global_step": 11761, "epoch": 132, "lr": 9.957432778660155e-05}
{"train_loss": 0.718092143535614, "global_step": 11762, "epoch": 132, "lr": 9.957425229628768e-05}
{"train_loss": 0.7064560651779175, "global_step": 11763, "epoch": 132, "lr": 9.957417679930917e-05}
{"train_loss": 0.6207115054130554, "global_step": 11764, "epoch": 132, "lr": 9.9574101295666e-05}
{"train_loss": 0.47389963269233704, "global_step": 11765, "epoch": 132, "lr": 9.957402578535819e-05}
{"train_loss": 0.6446688175201416, "global_step": 11766, "epoch": 132, "lr": 9.957395026838578e-05}
{"train_loss": 0.7132959961891174, "global_step": 11767, "epoch": 132, "lr": 9.957387474474873e-05}
{"train_loss": 0.6752616167068481, "global_step": 11768, "epoch": 132, "lr": 9.95737992144471e-05}
{"train_loss": 0.5821841955184937, "global_step": 11769, "epoch": 132, "lr": 9.957372367748086e-05}
{"train_loss": 0.5891709923744202, "global_step": 11770, "epoch": 132, "lr": 9.957364813385003e-05}
{"train_loss": 0.6140188574790955, "global_step": 11771, "epoch": 132, "lr": 9.957357258355467e-05}
{"train_loss": 0.608488917350769, "global_step": 11772, "epoch": 132, "lr": 9.95734970265947e-05}
{"train_loss": 0.9473140835762024, "global_step": 11773, "epoch": 132, "lr": 9.95734214629702e-05}
{"train_loss": 0.5352514982223511, "global_step": 11774, "epoch": 132, "lr": 9.957334589268116e-05}
{"train_loss": 0.7140355706214905, "global_step": 11775, "epoch": 132, "lr": 9.957327031572758e-05}
{"train_loss": 0.6954392790794373, "global_step": 11776, "epoch": 132, "lr": 9.957319473210947e-05}
{"train_loss": 0.677638053894043, "global_step": 11777, "epoch": 132, "lr": 9.957311914182686e-05}
{"train_loss": 0.645939290523529, "global_step": 11778, "epoch": 132, "lr": 9.957304354487976e-05}
{"train_loss": 0.562715470790863, "global_step": 11779, "epoch": 132, "lr": 9.957296794126814e-05}
{"train_loss": 0.6064853072166443, "global_step": 11780, "epoch": 132, "lr": 9.957289233099206e-05}
{"train_loss": 0.6492031216621399, "global_step": 11781, "epoch": 132, "lr": 9.957281671405149e-05}
{"train_loss": 0.6143439412117004, "global_step": 11782, "epoch": 132, "lr": 9.957274109044648e-05}
{"train_loss": 0.6278437972068787, "global_step": 11783, "epoch": 132, "lr": 9.957266546017702e-05}
{"train_loss": 0.6823374629020691, "global_step": 11784, "epoch": 132, "lr": 9.957258982324312e-05}
{"train_loss": 0.8965142369270325, "global_step": 11785, "epoch": 132, "lr": 9.957251417964477e-05}
{"train_loss": 0.7483406662940979, "global_step": 11786, "epoch": 132, "lr": 9.957243852938201e-05}
{"train_loss": 0.5949289202690125, "global_step": 11787, "epoch": 132, "lr": 9.957236287245484e-05}
{"train_loss": 0.5849474668502808, "global_step": 11788, "epoch": 132, "lr": 9.957228720886327e-05}
{"train_loss": 0.6002243161201477, "global_step": 11789, "epoch": 132, "lr": 9.95722115386073e-05}
{"train_loss": 0.5675191283226013, "global_step": 11790, "epoch": 132, "lr": 9.957213586168696e-05}
{"train_loss": 0.7012118697166443, "global_step": 11791, "epoch": 132, "lr": 9.957206017810225e-05}
{"train_loss": 0.6717889308929443, "global_step": 11792, "epoch": 132, "lr": 9.957198448785319e-05}
{"train_loss": 0.632096529006958, "global_step": 11793, "epoch": 132, "lr": 9.957190879093976e-05}
{"train_loss": 0.7034221291542053, "global_step": 11794, "epoch": 132, "lr": 9.9571833087362e-05}
{"train_loss": 0.6547534465789795, "global_step": 11795, "epoch": 132, "lr": 9.957175737711992e-05}
{"train_loss": 0.5496945381164551, "global_step": 11796, "epoch": 132, "lr": 9.957168166021352e-05}
{"train_loss": 0.7889392375946045, "global_step": 11797, "epoch": 132, "lr": 9.957160593664279e-05}
{"train_loss": 0.6637322902679443, "global_step": 11798, "epoch": 132, "lr": 9.957153020640779e-05}
{"train_loss": 0.6273313760757446, "global_step": 11799, "epoch": 132, "lr": 9.957145446950848e-05}
{"train_loss": 0.6405442357063293, "global_step": 11800, "epoch": 132, "lr": 9.95713787259449e-05}
{"train_loss": 0.5698176026344299, "global_step": 11801, "epoch": 132, "lr": 9.957130297571706e-05}
{"train_loss": 0.7139251828193665, "global_step": 11802, "epoch": 132, "lr": 9.957122721882497e-05}
{"train_loss": 0.6250352263450623, "global_step": 11803, "epoch": 132, "lr": 9.957115145526861e-05}
{"train_loss": 0.587668240070343, "global_step": 11804, "epoch": 132, "lr": 9.957107568504802e-05}
{"train_loss": 0.6451086401939392, "global_step": 11805, "epoch": 132, "lr": 9.957099990816321e-05}
{"train_loss": 0.7602537274360657, "global_step": 11806, "epoch": 132, "lr": 9.957092412461419e-05}
{"train_loss": 0.5713313221931458, "global_step": 11807, "epoch": 132, "lr": 9.957084833440095e-05}
{"train_loss": 0.6753067374229431, "global_step": 11808, "epoch": 132, "lr": 9.957077253752352e-05}
{"train_loss": 0.8899299502372742, "global_step": 11809, "epoch": 132, "lr": 9.95706967339819e-05}
{"train_loss": 0.7116257548332214, "global_step": 11810, "epoch": 132, "lr": 9.95706209237761e-05}
{"train_loss": 0.8924622535705566, "global_step": 11811, "epoch": 132, "lr": 9.957054510690613e-05}
{"train_loss": 0.8931626677513123, "global_step": 11812, "epoch": 132, "lr": 9.957046928337202e-05}
{"train_loss": 0.7521647810935974, "global_step": 11813, "epoch": 132, "lr": 9.957039345317377e-05}
{"train_loss": 0.8324381113052368, "global_step": 11814, "epoch": 132, "lr": 9.957031761631135e-05}
{"train_loss": 0.8350103497505188, "global_step": 11815, "epoch": 132, "lr": 9.957024177278483e-05}
{"train_loss": 0.7823563814163208, "global_step": 11816, "epoch": 132, "lr": 9.957016592259421e-05}
{"train_loss": 0.6446359157562256, "global_step": 11817, "epoch": 132, "lr": 9.957009006573945e-05}
{"train_loss": 0.6430136561393738, "global_step": 11818, "epoch": 132, "lr": 9.957001420222061e-05}
{"train_loss": 0.5800549983978271, "global_step": 11819, "epoch": 132, "lr": 9.956993833203769e-05}
{"train_loss": 0.7616021037101746, "global_step": 11820, "epoch": 132, "lr": 9.956986245519071e-05}
{"train_loss": 0.5963525772094727, "global_step": 11821, "epoch": 132, "lr": 9.956978657167965e-05}
{"train_loss": 0.7671399712562561, "global_step": 11822, "epoch": 132, "lr": 9.956971068150454e-05}
{"train_loss": 0.5919296741485596, "global_step": 11823, "epoch": 132, "lr": 9.956963478466538e-05}
{"train_loss": 0.6439411640167236, "global_step": 11824, "epoch": 132, "lr": 9.95695588811622e-05}
{"train_loss": 0.6297461986541748, "global_step": 11825, "epoch": 132, "lr": 9.956948297099498e-05}
{"train_loss": 0.7341857552528381, "global_step": 11826, "epoch": 132, "lr": 9.956940705416375e-05}
{"train_loss": 0.686242938041687, "global_step": 11827, "epoch": 132, "lr": 9.956933113066853e-05}
{"train_loss": 0.7486723065376282, "global_step": 11828, "epoch": 132, "lr": 9.95692552005093e-05}
{"train_loss": 0.6239246726036072, "global_step": 11829, "epoch": 132, "lr": 9.95691792636861e-05}
{"train_loss": 0.7134976983070374, "global_step": 11830, "epoch": 132, "lr": 9.956910332019894e-05}
{"train_loss": 0.5714614391326904, "global_step": 11831, "epoch": 132, "lr": 9.956902737004778e-05}
{"train_loss": 0.6062073707580566, "global_step": 11832, "epoch": 132, "lr": 9.956895141323271e-05}
{"train_loss": 0.7403280735015869, "global_step": 11833, "epoch": 132, "lr": 9.956887544975367e-05}
{"train_loss": 0.6615310907363892, "global_step": 11834, "epoch": 132, "lr": 9.956879947961071e-05}
{"train_loss": 0.6108497977256775, "global_step": 11835, "epoch": 132, "lr": 9.956872350280384e-05}
{"train_loss": 0.6683917725354098, "global_step": 11836, "epoch": 132, "lr": 9.956864751933304e-05, "val_loss": 1.3582979440689087}
{"train_loss": 0.6762804985046387, "global_step": 11837, "epoch": 133, "lr": 9.956857152919835e-05}
{"train_loss": 0.7897597551345825, "global_step": 11838, "epoch": 133, "lr": 9.956849553239977e-05}
{"train_loss": 0.6349182724952698, "global_step": 11839, "epoch": 133, "lr": 9.956841952893731e-05}
{"train_loss": 0.6093578934669495, "global_step": 11840, "epoch": 133, "lr": 9.956834351881096e-05}
{"train_loss": 0.6609378457069397, "global_step": 11841, "epoch": 133, "lr": 9.956826750202077e-05}
{"train_loss": 0.5502915978431702, "global_step": 11842, "epoch": 133, "lr": 9.956819147856675e-05}
{"train_loss": 0.8213369250297546, "global_step": 11843, "epoch": 133, "lr": 9.956811544844886e-05}
{"train_loss": 0.646734893321991, "global_step": 11844, "epoch": 133, "lr": 9.956803941166715e-05}
{"train_loss": 0.7002170085906982, "global_step": 11845, "epoch": 133, "lr": 9.956796336822163e-05}
{"train_loss": 0.6420860290527344, "global_step": 11846, "epoch": 133, "lr": 9.956788731811229e-05}
{"train_loss": 0.6078991293907166, "global_step": 11847, "epoch": 133, "lr": 9.956781126133914e-05}
{"train_loss": 0.7545247077941895, "global_step": 11848, "epoch": 133, "lr": 9.956773519790224e-05}
{"train_loss": 0.6950100064277649, "global_step": 11849, "epoch": 133, "lr": 9.956765912780152e-05}
{"train_loss": 0.6648992300033569, "global_step": 11850, "epoch": 133, "lr": 9.956758305103706e-05}
{"train_loss": 0.6128508448600769, "global_step": 11851, "epoch": 133, "lr": 9.956750696760883e-05}
{"train_loss": 0.7083048224449158, "global_step": 11852, "epoch": 133, "lr": 9.956743087751686e-05}
{"train_loss": 0.6450568437576294, "global_step": 11853, "epoch": 133, "lr": 9.956735478076114e-05}
{"train_loss": 0.8985484838485718, "global_step": 11854, "epoch": 133, "lr": 9.95672786773417e-05}
{"train_loss": 0.7202780246734619, "global_step": 11855, "epoch": 133, "lr": 9.956720256725856e-05}
{"train_loss": 0.5927359461784363, "global_step": 11856, "epoch": 133, "lr": 9.95671264505117e-05}
{"train_loss": 0.8377389311790466, "global_step": 11857, "epoch": 133, "lr": 9.956705032710115e-05}
{"train_loss": 0.7449249625205994, "global_step": 11858, "epoch": 133, "lr": 9.95669741970269e-05}
{"train_loss": 0.6813368797302246, "global_step": 11859, "epoch": 133, "lr": 9.956689806028898e-05}
{"train_loss": 0.6544651389122009, "global_step": 11860, "epoch": 133, "lr": 9.95668219168874e-05}
{"train_loss": 0.7161751389503479, "global_step": 11861, "epoch": 133, "lr": 9.956674576682217e-05}
{"train_loss": 0.5923486948013306, "global_step": 11862, "epoch": 133, "lr": 9.956666961009329e-05}
{"train_loss": 0.7584161162376404, "global_step": 11863, "epoch": 133, "lr": 9.956659344670076e-05}
{"train_loss": 0.5865360498428345, "global_step": 11864, "epoch": 133, "lr": 9.956651727664462e-05}
{"train_loss": 0.6059768795967102, "global_step": 11865, "epoch": 133, "lr": 9.956644109992487e-05}
{"train_loss": 0.7125833034515381, "global_step": 11866, "epoch": 133, "lr": 9.956636491654151e-05}
{"train_loss": 0.710087776184082, "global_step": 11867, "epoch": 133, "lr": 9.956628872649456e-05}
{"train_loss": 0.6895710229873657, "global_step": 11868, "epoch": 133, "lr": 9.956621252978402e-05}
{"train_loss": 0.6105068922042847, "global_step": 11869, "epoch": 133, "lr": 9.95661363264099e-05}
{"train_loss": 0.5747794508934021, "global_step": 11870, "epoch": 133, "lr": 9.956606011637224e-05}
{"train_loss": 0.7097153663635254, "global_step": 11871, "epoch": 133, "lr": 9.9565983899671e-05}
{"train_loss": 0.6762726306915283, "global_step": 11872, "epoch": 133, "lr": 9.956590767630624e-05}
{"train_loss": 0.835422158241272, "global_step": 11873, "epoch": 133, "lr": 9.956583144627794e-05}
{"train_loss": 0.7505779266357422, "global_step": 11874, "epoch": 133, "lr": 9.956575520958614e-05}
{"train_loss": 0.5668190717697144, "global_step": 11875, "epoch": 133, "lr": 9.956567896623079e-05}
{"train_loss": 0.6820934414863586, "global_step": 11876, "epoch": 133, "lr": 9.956560271621195e-05}
{"train_loss": 0.5862957835197449, "global_step": 11877, "epoch": 133, "lr": 9.956552645952963e-05}
{"train_loss": 0.6955602765083313, "global_step": 11878, "epoch": 133, "lr": 9.956545019618383e-05}
{"train_loss": 0.614992082118988, "global_step": 11879, "epoch": 133, "lr": 9.956537392617456e-05}
{"train_loss": 0.6234866380691528, "global_step": 11880, "epoch": 133, "lr": 9.95652976495018e-05}
{"train_loss": 0.6555088758468628, "global_step": 11881, "epoch": 133, "lr": 9.956522136616562e-05}
{"train_loss": 0.7004555463790894, "global_step": 11882, "epoch": 133, "lr": 9.956514507616599e-05}
{"train_loss": 0.698115348815918, "global_step": 11883, "epoch": 133, "lr": 9.956506877950294e-05}
{"train_loss": 0.6129071116447449, "global_step": 11884, "epoch": 133, "lr": 9.956499247617645e-05}
{"train_loss": 0.6509292721748352, "global_step": 11885, "epoch": 133, "lr": 9.956491616618657e-05}
{"train_loss": 0.6478434801101685, "global_step": 11886, "epoch": 133, "lr": 9.956483984953329e-05}
{"train_loss": 0.674298107624054, "global_step": 11887, "epoch": 133, "lr": 9.956476352621661e-05}
{"train_loss": 0.6547451019287109, "global_step": 11888, "epoch": 133, "lr": 9.956468719623657e-05}
{"train_loss": 0.5211257338523865, "global_step": 11889, "epoch": 133, "lr": 9.956461085959314e-05}
{"train_loss": 0.6045829057693481, "global_step": 11890, "epoch": 133, "lr": 9.956453451628637e-05}
{"train_loss": 0.8348771333694458, "global_step": 11891, "epoch": 133, "lr": 9.956445816631624e-05}
{"train_loss": 0.5841630101203918, "global_step": 11892, "epoch": 133, "lr": 9.956438180968279e-05}
{"train_loss": 0.7070978879928589, "global_step": 11893, "epoch": 133, "lr": 9.9564305446386e-05}
{"train_loss": 0.7494210600852966, "global_step": 11894, "epoch": 133, "lr": 9.956422907642589e-05}
{"train_loss": 0.5371062159538269, "global_step": 11895, "epoch": 133, "lr": 9.956415269980248e-05}
{"train_loss": 0.5682080984115601, "global_step": 11896, "epoch": 133, "lr": 9.956407631651577e-05}
{"train_loss": 0.8077735900878906, "global_step": 11897, "epoch": 133, "lr": 9.956399992656578e-05}
{"train_loss": 0.6000970602035522, "global_step": 11898, "epoch": 133, "lr": 9.95639235299525e-05}
{"train_loss": 0.5077828764915466, "global_step": 11899, "epoch": 133, "lr": 9.956384712667597e-05}
{"train_loss": 0.5811618566513062, "global_step": 11900, "epoch": 133, "lr": 9.956377071673617e-05}
{"train_loss": 0.7388662099838257, "global_step": 11901, "epoch": 133, "lr": 9.956369430013314e-05}
{"train_loss": 0.5931100249290466, "global_step": 11902, "epoch": 133, "lr": 9.956361787686687e-05}
{"train_loss": 0.8642531037330627, "global_step": 11903, "epoch": 133, "lr": 9.956354144693738e-05}
{"train_loss": 0.5921850204467773, "global_step": 11904, "epoch": 133, "lr": 9.956346501034466e-05}
{"train_loss": 0.5879826545715332, "global_step": 11905, "epoch": 133, "lr": 9.956338856708875e-05}
{"train_loss": 0.5023402571678162, "global_step": 11906, "epoch": 133, "lr": 9.956331211716964e-05}
{"train_loss": 0.7391022443771362, "global_step": 11907, "epoch": 133, "lr": 9.956323566058737e-05}
{"train_loss": 0.5356842875480652, "global_step": 11908, "epoch": 133, "lr": 9.95631591973419e-05}
{"train_loss": 0.7804352641105652, "global_step": 11909, "epoch": 133, "lr": 9.956308272743327e-05}
{"train_loss": 0.4501383900642395, "global_step": 11910, "epoch": 133, "lr": 9.956300625086149e-05}
{"train_loss": 0.6779234409332275, "global_step": 11911, "epoch": 133, "lr": 9.956292976762658e-05}
{"train_loss": 0.7797707319259644, "global_step": 11912, "epoch": 133, "lr": 9.956285327772852e-05}
{"train_loss": 0.8685199022293091, "global_step": 11913, "epoch": 133, "lr": 9.956277678116735e-05}
{"train_loss": 0.7756075859069824, "global_step": 11914, "epoch": 133, "lr": 9.956270027794306e-05}
{"train_loss": 0.7262066602706909, "global_step": 11915, "epoch": 133, "lr": 9.956262376805567e-05}
{"train_loss": 0.7507062554359436, "global_step": 11916, "epoch": 133, "lr": 9.95625472515052e-05}
{"train_loss": 0.8258780837059021, "global_step": 11917, "epoch": 133, "lr": 9.956247072829164e-05}
{"train_loss": 0.6981651186943054, "global_step": 11918, "epoch": 133, "lr": 9.956239419841502e-05}
{"train_loss": 0.5757662653923035, "global_step": 11919, "epoch": 133, "lr": 9.956231766187534e-05}
{"train_loss": 0.6875181198120117, "global_step": 11920, "epoch": 133, "lr": 9.95622411186726e-05}
{"train_loss": 0.6880844831466675, "global_step": 11921, "epoch": 133, "lr": 9.956216456880683e-05}
{"train_loss": 0.6735588908195496, "global_step": 11922, "epoch": 133, "lr": 9.956208801227801e-05}
{"train_loss": 0.6331885457038879, "global_step": 11923, "epoch": 133, "lr": 9.956201144908619e-05}
{"train_loss": 0.5984945893287659, "global_step": 11924, "epoch": 133, "lr": 9.956193487923137e-05}
{"train_loss": 0.6726906909031815, "global_step": 11925, "epoch": 133, "lr": 9.956185830271353e-05, "val_loss": 1.3012676239013672}
{"train_loss": 0.621810257434845, "global_step": 11926, "epoch": 134, "lr": 9.956178171953272e-05}
{"train_loss": 0.739828884601593, "global_step": 11927, "epoch": 134, "lr": 9.956170512968892e-05}
{"train_loss": 0.6496577262878418, "global_step": 11928, "epoch": 134, "lr": 9.956162853318216e-05}
{"train_loss": 0.6087931394577026, "global_step": 11929, "epoch": 134, "lr": 9.956155193001244e-05}
{"train_loss": 0.6541584730148315, "global_step": 11930, "epoch": 134, "lr": 9.956147532017979e-05}
{"train_loss": 0.7336170673370361, "global_step": 11931, "epoch": 134, "lr": 9.956139870368418e-05}
{"train_loss": 0.6202789545059204, "global_step": 11932, "epoch": 134, "lr": 9.956132208052565e-05}
{"train_loss": 0.5064846873283386, "global_step": 11933, "epoch": 134, "lr": 9.956124545070421e-05}
{"train_loss": 0.6803290843963623, "global_step": 11934, "epoch": 134, "lr": 9.956116881421986e-05}
{"train_loss": 0.6214150786399841, "global_step": 11935, "epoch": 134, "lr": 9.956109217107262e-05}
{"train_loss": 0.5550363063812256, "global_step": 11936, "epoch": 134, "lr": 9.956101552126249e-05}
{"train_loss": 0.5953380465507507, "global_step": 11937, "epoch": 134, "lr": 9.956093886478949e-05}
{"train_loss": 0.8061856031417847, "global_step": 11938, "epoch": 134, "lr": 9.956086220165362e-05}
{"train_loss": 0.6954447031021118, "global_step": 11939, "epoch": 134, "lr": 9.956078553185491e-05}
{"train_loss": 0.6560498476028442, "global_step": 11940, "epoch": 134, "lr": 9.956070885539333e-05}
{"train_loss": 0.6095932126045227, "global_step": 11941, "epoch": 134, "lr": 9.956063217226894e-05}
{"train_loss": 0.6721521019935608, "global_step": 11942, "epoch": 134, "lr": 9.956055548248172e-05}
{"train_loss": 0.6968657374382019, "global_step": 11943, "epoch": 134, "lr": 9.956047878603168e-05}
{"train_loss": 0.6389337778091431, "global_step": 11944, "epoch": 134, "lr": 9.956040208291886e-05}
{"train_loss": 0.6161435842514038, "global_step": 11945, "epoch": 134, "lr": 9.956032537314323e-05}
{"train_loss": 0.6344027519226074, "global_step": 11946, "epoch": 134, "lr": 9.956024865670481e-05}
{"train_loss": 0.7483097314834595, "global_step": 11947, "epoch": 134, "lr": 9.956017193360363e-05}
{"train_loss": 0.8255931735038757, "global_step": 11948, "epoch": 134, "lr": 9.956009520383969e-05}
{"train_loss": 0.478748619556427, "global_step": 11949, "epoch": 134, "lr": 9.956001846741301e-05}
{"train_loss": 0.7647837996482849, "global_step": 11950, "epoch": 134, "lr": 9.955994172432356e-05}
{"train_loss": 0.6713608503341675, "global_step": 11951, "epoch": 134, "lr": 9.955986497457142e-05}
{"train_loss": 0.6104335188865662, "global_step": 11952, "epoch": 134, "lr": 9.955978821815653e-05}
{"train_loss": 0.6013503670692444, "global_step": 11953, "epoch": 134, "lr": 9.955971145507895e-05}
{"train_loss": 0.7159115076065063, "global_step": 11954, "epoch": 134, "lr": 9.955963468533864e-05}
{"train_loss": 0.7059443593025208, "global_step": 11955, "epoch": 134, "lr": 9.955955790893566e-05}
{"train_loss": 0.8184232115745544, "global_step": 11956, "epoch": 134, "lr": 9.955948112587e-05}
{"train_loss": 0.6950037479400635, "global_step": 11957, "epoch": 134, "lr": 9.955940433614168e-05}
{"train_loss": 0.5645437240600586, "global_step": 11958, "epoch": 134, "lr": 9.955932753975069e-05}
{"train_loss": 0.5864362716674805, "global_step": 11959, "epoch": 134, "lr": 9.955925073669706e-05}
{"train_loss": 0.6486430168151855, "global_step": 11960, "epoch": 134, "lr": 9.95591739269808e-05}
{"train_loss": 0.6027101278305054, "global_step": 11961, "epoch": 134, "lr": 9.95590971106019e-05}
{"train_loss": 0.6322267055511475, "global_step": 11962, "epoch": 134, "lr": 9.95590202875604e-05}
{"train_loss": 0.5161576271057129, "global_step": 11963, "epoch": 134, "lr": 9.955894345785626e-05}
{"train_loss": 0.6741062998771667, "global_step": 11964, "epoch": 134, "lr": 9.955886662148955e-05}
{"train_loss": 0.7782245874404907, "global_step": 11965, "epoch": 134, "lr": 9.955878977846025e-05}
{"train_loss": 0.6191926002502441, "global_step": 11966, "epoch": 134, "lr": 9.955871292876837e-05}
{"train_loss": 0.6104698181152344, "global_step": 11967, "epoch": 134, "lr": 9.955863607241394e-05}
{"train_loss": 0.6397754549980164, "global_step": 11968, "epoch": 134, "lr": 9.955855920939694e-05}
{"train_loss": 0.7341460585594177, "global_step": 11969, "epoch": 134, "lr": 9.95584823397174e-05}
{"train_loss": 0.7666925191879272, "global_step": 11970, "epoch": 134, "lr": 9.955840546337533e-05}
{"train_loss": 0.7004366517066956, "global_step": 11971, "epoch": 134, "lr": 9.955832858037073e-05}
{"train_loss": 0.5186423063278198, "global_step": 11972, "epoch": 134, "lr": 9.955825169070362e-05}
{"train_loss": 0.6578927636146545, "global_step": 11973, "epoch": 134, "lr": 9.9558174794374e-05}
{"train_loss": 0.6872533559799194, "global_step": 11974, "epoch": 134, "lr": 9.95580978913819e-05}
{"train_loss": 0.6482285261154175, "global_step": 11975, "epoch": 134, "lr": 9.955802098172731e-05}
{"train_loss": 0.6738901138305664, "global_step": 11976, "epoch": 134, "lr": 9.955794406541026e-05}
{"train_loss": 0.8140633702278137, "global_step": 11977, "epoch": 134, "lr": 9.955786714243073e-05}
{"train_loss": 0.6516433954238892, "global_step": 11978, "epoch": 134, "lr": 9.955779021278875e-05}
{"train_loss": 0.5752561092376709, "global_step": 11979, "epoch": 134, "lr": 9.955771327648434e-05}
{"train_loss": 0.5637218952178955, "global_step": 11980, "epoch": 134, "lr": 9.95576363335175e-05}
{"train_loss": 0.6747571229934692, "global_step": 11981, "epoch": 134, "lr": 9.955755938388825e-05}
{"train_loss": 0.5146796107292175, "global_step": 11982, "epoch": 134, "lr": 9.955748242759656e-05}
{"train_loss": 0.6454095244407654, "global_step": 11983, "epoch": 134, "lr": 9.955740546464249e-05}
{"train_loss": 0.7269099354743958, "global_step": 11984, "epoch": 134, "lr": 9.955732849502603e-05}
{"train_loss": 0.5153955221176147, "global_step": 11985, "epoch": 134, "lr": 9.955725151874719e-05}
{"train_loss": 0.7826133966445923, "global_step": 11986, "epoch": 134, "lr": 9.955717453580597e-05}
{"train_loss": 0.7016278505325317, "global_step": 11987, "epoch": 134, "lr": 9.955709754620242e-05}
{"train_loss": 0.6340471506118774, "global_step": 11988, "epoch": 134, "lr": 9.955702054993649e-05}
{"train_loss": 0.5429200530052185, "global_step": 11989, "epoch": 134, "lr": 9.955694354700825e-05}
{"train_loss": 0.7044264078140259, "global_step": 11990, "epoch": 134, "lr": 9.955686653741766e-05}
{"train_loss": 0.7286217212677002, "global_step": 11991, "epoch": 134, "lr": 9.955678952116476e-05}
{"train_loss": 0.7997599840164185, "global_step": 11992, "epoch": 134, "lr": 9.955671249824956e-05}
{"train_loss": 0.7369704246520996, "global_step": 11993, "epoch": 134, "lr": 9.955663546867206e-05}
{"train_loss": 0.6428449153900146, "global_step": 11994, "epoch": 134, "lr": 9.955655843243227e-05}
{"train_loss": 0.6325925588607788, "global_step": 11995, "epoch": 134, "lr": 9.955648138953021e-05}
{"train_loss": 0.6619852781295776, "global_step": 11996, "epoch": 134, "lr": 9.95564043399659e-05}
{"train_loss": 0.6157594323158264, "global_step": 11997, "epoch": 134, "lr": 9.955632728373931e-05}
{"train_loss": 0.825795590877533, "global_step": 11998, "epoch": 134, "lr": 9.955625022085049e-05}
{"train_loss": 0.47690466046333313, "global_step": 11999, "epoch": 134, "lr": 9.955617315129943e-05}
{"train_loss": 0.74952232837677, "global_step": 12000, "epoch": 134, "lr": 9.955609607508615e-05}
{"train_loss": 0.6804655194282532, "global_step": 12001, "epoch": 134, "lr": 9.955601899221066e-05}
{"train_loss": 0.8206884860992432, "global_step": 12002, "epoch": 134, "lr": 9.955594190267295e-05}
{"train_loss": 0.6051502823829651, "global_step": 12003, "epoch": 134, "lr": 9.955586480647307e-05}
{"train_loss": 0.5127388834953308, "global_step": 12004, "epoch": 134, "lr": 9.9555787703611e-05}
{"train_loss": 0.6818060874938965, "global_step": 12005, "epoch": 134, "lr": 9.955571059408675e-05}
{"train_loss": 0.7104089856147766, "global_step": 12006, "epoch": 134, "lr": 9.955563347790036e-05}
{"train_loss": 0.6338596343994141, "global_step": 12007, "epoch": 134, "lr": 9.95555563550518e-05}
{"train_loss": 0.6997459530830383, "global_step": 12008, "epoch": 134, "lr": 9.955547922554109e-05}
{"train_loss": 0.6704832911491394, "global_step": 12009, "epoch": 134, "lr": 9.955540208936826e-05}
{"train_loss": 0.5103359818458557, "global_step": 12010, "epoch": 134, "lr": 9.955532494653332e-05}
{"train_loss": 0.6927666664123535, "global_step": 12011, "epoch": 134, "lr": 9.955524779703627e-05}
{"train_loss": 0.6508651971817017, "global_step": 12012, "epoch": 134, "lr": 9.955517064087711e-05}
{"train_loss": 0.6969491839408875, "global_step": 12013, "epoch": 134, "lr": 9.955509347805587e-05}
{"train_loss": 0.659290791562434, "global_step": 12014, "epoch": 134, "lr": 9.955501630857255e-05, "val_loss": 1.300289273262024}
{"train_loss": 0.6745287179946899, "global_step": 12015, "epoch": 135, "lr": 9.955493913242717e-05}
{"train_loss": 0.5796835422515869, "global_step": 12016, "epoch": 135, "lr": 9.955486194961971e-05}
{"train_loss": 0.6035687923431396, "global_step": 12017, "epoch": 135, "lr": 9.955478476015023e-05}
{"train_loss": 0.7110512852668762, "global_step": 12018, "epoch": 135, "lr": 9.95547075640187e-05}
{"train_loss": 0.7549989819526672, "global_step": 12019, "epoch": 135, "lr": 9.955463036122515e-05}
{"train_loss": 0.7625183463096619, "global_step": 12020, "epoch": 135, "lr": 9.955455315176956e-05}
{"train_loss": 0.6058092713356018, "global_step": 12021, "epoch": 135, "lr": 9.955447593565199e-05}
{"train_loss": 0.6723002195358276, "global_step": 12022, "epoch": 135, "lr": 9.955439871287241e-05}
{"train_loss": 0.4951845407485962, "global_step": 12023, "epoch": 135, "lr": 9.955432148343085e-05}
{"train_loss": 0.7797048687934875, "global_step": 12024, "epoch": 135, "lr": 9.955424424732731e-05}
{"train_loss": 0.4495452344417572, "global_step": 12025, "epoch": 135, "lr": 9.955416700456182e-05}
{"train_loss": 0.8004472255706787, "global_step": 12026, "epoch": 135, "lr": 9.955408975513436e-05}
{"train_loss": 0.562267541885376, "global_step": 12027, "epoch": 135, "lr": 9.955401249904497e-05}
{"train_loss": 0.6016787886619568, "global_step": 12028, "epoch": 135, "lr": 9.955393523629364e-05}
{"train_loss": 0.6507542133331299, "global_step": 12029, "epoch": 135, "lr": 9.955385796688039e-05}
{"train_loss": 0.6724780201911926, "global_step": 12030, "epoch": 135, "lr": 9.955378069080523e-05}
{"train_loss": 0.6692442297935486, "global_step": 12031, "epoch": 135, "lr": 9.955370340806816e-05}
{"train_loss": 0.6360096335411072, "global_step": 12032, "epoch": 135, "lr": 9.955362611866921e-05}
{"train_loss": 0.6425766348838806, "global_step": 12033, "epoch": 135, "lr": 9.955354882260837e-05}
{"train_loss": 0.703567385673523, "global_step": 12034, "epoch": 135, "lr": 9.955347151988566e-05}
{"train_loss": 0.5456239581108093, "global_step": 12035, "epoch": 135, "lr": 9.955339421050109e-05}
{"train_loss": 0.6053228974342346, "global_step": 12036, "epoch": 135, "lr": 9.955331689445469e-05}
{"train_loss": 0.691687285900116, "global_step": 12037, "epoch": 135, "lr": 9.955323957174642e-05}
{"train_loss": 0.608042299747467, "global_step": 12038, "epoch": 135, "lr": 9.955316224237634e-05}
{"train_loss": 0.8058798313140869, "global_step": 12039, "epoch": 135, "lr": 9.955308490634443e-05}
{"train_loss": 0.6575841903686523, "global_step": 12040, "epoch": 135, "lr": 9.955300756365072e-05}
{"train_loss": 0.6120778918266296, "global_step": 12041, "epoch": 135, "lr": 9.95529302142952e-05}
{"train_loss": 0.46173062920570374, "global_step": 12042, "epoch": 135, "lr": 9.955285285827791e-05}
{"train_loss": 0.6345698833465576, "global_step": 12043, "epoch": 135, "lr": 9.955277549559883e-05}
{"train_loss": 0.641575276851654, "global_step": 12044, "epoch": 135, "lr": 9.955269812625799e-05}
{"train_loss": 0.7209475636482239, "global_step": 12045, "epoch": 135, "lr": 9.955262075025538e-05}
{"train_loss": 0.6426020860671997, "global_step": 12046, "epoch": 135, "lr": 9.955254336759105e-05}
{"train_loss": 0.617305338382721, "global_step": 12047, "epoch": 135, "lr": 9.955246597826496e-05}
{"train_loss": 0.5452362298965454, "global_step": 12048, "epoch": 135, "lr": 9.955238858227715e-05}
{"train_loss": 0.5989332795143127, "global_step": 12049, "epoch": 135, "lr": 9.955231117962764e-05}
{"train_loss": 0.734652042388916, "global_step": 12050, "epoch": 135, "lr": 9.955223377031642e-05}
{"train_loss": 0.6416409611701965, "global_step": 12051, "epoch": 135, "lr": 9.955215635434349e-05}
{"train_loss": 0.7624205946922302, "global_step": 12052, "epoch": 135, "lr": 9.955207893170889e-05}
{"train_loss": 0.5966317653656006, "global_step": 12053, "epoch": 135, "lr": 9.955200150241261e-05}
{"train_loss": 0.5854695439338684, "global_step": 12054, "epoch": 135, "lr": 9.955192406645469e-05}
{"train_loss": 0.5513858795166016, "global_step": 12055, "epoch": 135, "lr": 9.95518466238351e-05}
{"train_loss": 0.5293893218040466, "global_step": 12056, "epoch": 135, "lr": 9.955176917455385e-05}
{"train_loss": 0.6914063692092896, "global_step": 12057, "epoch": 135, "lr": 9.9551691718611e-05}
{"train_loss": 0.6158900260925293, "global_step": 12058, "epoch": 135, "lr": 9.955161425600652e-05}
{"train_loss": 0.6422158479690552, "global_step": 12059, "epoch": 135, "lr": 9.95515367867404e-05}
{"train_loss": 0.655332624912262, "global_step": 12060, "epoch": 135, "lr": 9.955145931081271e-05}
{"train_loss": 0.5910798907279968, "global_step": 12061, "epoch": 135, "lr": 9.955138182822343e-05}
{"train_loss": 0.6571511626243591, "global_step": 12062, "epoch": 135, "lr": 9.955130433897257e-05}
{"train_loss": 0.614179253578186, "global_step": 12063, "epoch": 135, "lr": 9.955122684306013e-05}
{"train_loss": 0.810319185256958, "global_step": 12064, "epoch": 135, "lr": 9.955114934048615e-05}
{"train_loss": 0.6232438087463379, "global_step": 12065, "epoch": 135, "lr": 9.95510718312506e-05}
{"train_loss": 0.7326731085777283, "global_step": 12066, "epoch": 135, "lr": 9.955099431535353e-05}
{"train_loss": 0.6198579668998718, "global_step": 12067, "epoch": 135, "lr": 9.955091679279493e-05}
{"train_loss": 0.6936282515525818, "global_step": 12068, "epoch": 135, "lr": 9.955083926357481e-05}
{"train_loss": 0.6002877354621887, "global_step": 12069, "epoch": 135, "lr": 9.955076172769318e-05}
{"train_loss": 0.5574120879173279, "global_step": 12070, "epoch": 135, "lr": 9.955068418515006e-05}
{"train_loss": 0.7928508520126343, "global_step": 12071, "epoch": 135, "lr": 9.955060663594544e-05}
{"train_loss": 0.5934098362922668, "global_step": 12072, "epoch": 135, "lr": 9.955052908007936e-05}
{"train_loss": 0.5312239527702332, "global_step": 12073, "epoch": 135, "lr": 9.955045151755182e-05}
{"train_loss": 0.7544671297073364, "global_step": 12074, "epoch": 135, "lr": 9.95503739483628e-05}
{"train_loss": 0.769335925579071, "global_step": 12075, "epoch": 135, "lr": 9.955029637251237e-05}
{"train_loss": 0.8237999677658081, "global_step": 12076, "epoch": 135, "lr": 9.955021879000049e-05}
{"train_loss": 0.6341713666915894, "global_step": 12077, "epoch": 135, "lr": 9.955014120082717e-05}
{"train_loss": 0.786461353302002, "global_step": 12078, "epoch": 135, "lr": 9.955006360499246e-05}
{"train_loss": 0.7913349866867065, "global_step": 12079, "epoch": 135, "lr": 9.954998600249634e-05}
{"train_loss": 0.6539499759674072, "global_step": 12080, "epoch": 135, "lr": 9.954990839333883e-05}
{"train_loss": 0.650837779045105, "global_step": 12081, "epoch": 135, "lr": 9.954983077751993e-05}
{"train_loss": 0.7751984596252441, "global_step": 12082, "epoch": 135, "lr": 9.954975315503968e-05}
{"train_loss": 0.8510658740997314, "global_step": 12083, "epoch": 135, "lr": 9.954967552589806e-05}
{"train_loss": 0.6613234281539917, "global_step": 12084, "epoch": 135, "lr": 9.954959789009508e-05}
{"train_loss": 0.5527881979942322, "global_step": 12085, "epoch": 135, "lr": 9.954952024763077e-05}
{"train_loss": 0.6390022039413452, "global_step": 12086, "epoch": 135, "lr": 9.954944259850513e-05}
{"train_loss": 0.7736368775367737, "global_step": 12087, "epoch": 135, "lr": 9.954936494271817e-05}
{"train_loss": 0.6425788402557373, "global_step": 12088, "epoch": 135, "lr": 9.95492872802699e-05}
{"train_loss": 0.6088969707489014, "global_step": 12089, "epoch": 135, "lr": 9.954920961116034e-05}
{"train_loss": 0.6601042747497559, "global_step": 12090, "epoch": 135, "lr": 9.954913193538949e-05}
{"train_loss": 0.6963365077972412, "global_step": 12091, "epoch": 135, "lr": 9.954905425295736e-05}
{"train_loss": 0.7207853198051453, "global_step": 12092, "epoch": 135, "lr": 9.954897656386396e-05}
{"train_loss": 0.8100748658180237, "global_step": 12093, "epoch": 135, "lr": 9.95488988681093e-05}
{"train_loss": 0.6317017674446106, "global_step": 12094, "epoch": 135, "lr": 9.954882116569342e-05}
{"train_loss": 0.6894111633300781, "global_step": 12095, "epoch": 135, "lr": 9.954874345661627e-05}
{"train_loss": 0.774545431137085, "global_step": 12096, "epoch": 135, "lr": 9.954866574087793e-05}
{"train_loss": 0.7096180319786072, "global_step": 12097, "epoch": 135, "lr": 9.954858801847835e-05}
{"train_loss": 0.7161685228347778, "global_step": 12098, "epoch": 135, "lr": 9.954851028941758e-05}
{"train_loss": 0.5940479040145874, "global_step": 12099, "epoch": 135, "lr": 9.954843255369561e-05}
{"train_loss": 0.6046478748321533, "global_step": 12100, "epoch": 135, "lr": 9.954835481131245e-05}
{"train_loss": 0.57344651222229, "global_step": 12101, "epoch": 135, "lr": 9.954827706226812e-05}
{"train_loss": 0.7421302199363708, "global_step": 12102, "epoch": 135, "lr": 9.954819930656264e-05}
{"train_loss": 0.659852335962017, "global_step": 12103, "epoch": 135, "lr": 9.954812154419602e-05, "val_loss": 1.3669474124908447, "train_action_mse_error": 47.68579864501953}
{"train_loss": 0.7686589956283569, "global_step": 12104, "epoch": 136, "lr": 9.954804377516823e-05}
{"train_loss": 0.7318158149719238, "global_step": 12105, "epoch": 136, "lr": 9.954796599947932e-05}
{"train_loss": 0.6615099906921387, "global_step": 12106, "epoch": 136, "lr": 9.95478882171293e-05}
{"train_loss": 0.6344664096832275, "global_step": 12107, "epoch": 136, "lr": 9.954781042811816e-05}
{"train_loss": 0.5866609215736389, "global_step": 12108, "epoch": 136, "lr": 9.954773263244592e-05}
{"train_loss": 0.5916683673858643, "global_step": 12109, "epoch": 136, "lr": 9.954765483011259e-05}
{"train_loss": 0.6802370548248291, "global_step": 12110, "epoch": 136, "lr": 9.954757702111818e-05}
{"train_loss": 0.6495840549468994, "global_step": 12111, "epoch": 136, "lr": 9.954749920546271e-05}
{"train_loss": 0.793960690498352, "global_step": 12112, "epoch": 136, "lr": 9.954742138314619e-05}
{"train_loss": 0.6418115496635437, "global_step": 12113, "epoch": 136, "lr": 9.954734355416861e-05}
{"train_loss": 0.6427954435348511, "global_step": 12114, "epoch": 136, "lr": 9.954726571853e-05}
{"train_loss": 0.629814863204956, "global_step": 12115, "epoch": 136, "lr": 9.954718787623035e-05}
{"train_loss": 0.5894858241081238, "global_step": 12116, "epoch": 136, "lr": 9.95471100272697e-05}
{"train_loss": 0.6735143065452576, "global_step": 12117, "epoch": 136, "lr": 9.954703217164804e-05}
{"train_loss": 0.8161693811416626, "global_step": 12118, "epoch": 136, "lr": 9.954695430936538e-05}
{"train_loss": 0.6754271388053894, "global_step": 12119, "epoch": 136, "lr": 9.954687644042174e-05}
{"train_loss": 0.7324875593185425, "global_step": 12120, "epoch": 136, "lr": 9.954679856481714e-05}
{"train_loss": 0.5311332941055298, "global_step": 12121, "epoch": 136, "lr": 9.954672068255156e-05}
{"train_loss": 0.6483708620071411, "global_step": 12122, "epoch": 136, "lr": 9.954664279362502e-05}
{"train_loss": 0.8245694637298584, "global_step": 12123, "epoch": 136, "lr": 9.954656489803755e-05}
{"train_loss": 0.8592720031738281, "global_step": 12124, "epoch": 136, "lr": 9.954648699578915e-05}
{"train_loss": 0.6838918328285217, "global_step": 12125, "epoch": 136, "lr": 9.954640908687983e-05}
{"train_loss": 0.5747650265693665, "global_step": 12126, "epoch": 136, "lr": 9.95463311713096e-05}
{"train_loss": 0.6078936457633972, "global_step": 12127, "epoch": 136, "lr": 9.954625324907847e-05}
{"train_loss": 0.5992913842201233, "global_step": 12128, "epoch": 136, "lr": 9.954617532018644e-05}
{"train_loss": 0.608262300491333, "global_step": 12129, "epoch": 136, "lr": 9.954609738463355e-05}
{"train_loss": 0.7084910273551941, "global_step": 12130, "epoch": 136, "lr": 9.954601944241977e-05}
{"train_loss": 0.5609398484230042, "global_step": 12131, "epoch": 136, "lr": 9.954594149354512e-05}
{"train_loss": 0.6296117901802063, "global_step": 12132, "epoch": 136, "lr": 9.954586353800965e-05}
{"train_loss": 0.6755414605140686, "global_step": 12133, "epoch": 136, "lr": 9.954578557581333e-05}
{"train_loss": 0.6759487390518188, "global_step": 12134, "epoch": 136, "lr": 9.95457076069562e-05}
{"train_loss": 0.6626687049865723, "global_step": 12135, "epoch": 136, "lr": 9.954562963143822e-05}
{"train_loss": 0.8048158884048462, "global_step": 12136, "epoch": 136, "lr": 9.954555164925947e-05}
{"train_loss": 0.6505844593048096, "global_step": 12137, "epoch": 136, "lr": 9.954547366041991e-05}
{"train_loss": 0.6259674429893494, "global_step": 12138, "epoch": 136, "lr": 9.954539566491957e-05}
{"train_loss": 0.545325517654419, "global_step": 12139, "epoch": 136, "lr": 9.954531766275843e-05}
{"train_loss": 0.6281394362449646, "global_step": 12140, "epoch": 136, "lr": 9.954523965393654e-05}
{"train_loss": 0.47422438859939575, "global_step": 12141, "epoch": 136, "lr": 9.95451616384539e-05}
{"train_loss": 0.6911374926567078, "global_step": 12142, "epoch": 136, "lr": 9.954508361631053e-05}
{"train_loss": 0.7091367244720459, "global_step": 12143, "epoch": 136, "lr": 9.95450055875064e-05}
{"train_loss": 0.6339889764785767, "global_step": 12144, "epoch": 136, "lr": 9.954492755204158e-05}
{"train_loss": 0.5782467126846313, "global_step": 12145, "epoch": 136, "lr": 9.954484950991603e-05}
{"train_loss": 0.5281387567520142, "global_step": 12146, "epoch": 136, "lr": 9.954477146112978e-05}
{"train_loss": 0.6922104358673096, "global_step": 12147, "epoch": 136, "lr": 9.954469340568284e-05}
{"train_loss": 0.6180745363235474, "global_step": 12148, "epoch": 136, "lr": 9.954461534357522e-05}
{"train_loss": 0.6952047944068909, "global_step": 12149, "epoch": 136, "lr": 9.954453727480693e-05}
{"train_loss": 0.7445676326751709, "global_step": 12150, "epoch": 136, "lr": 9.954445919937798e-05}
{"train_loss": 0.5189350843429565, "global_step": 12151, "epoch": 136, "lr": 9.954438111728838e-05}
{"train_loss": 0.52585369348526, "global_step": 12152, "epoch": 136, "lr": 9.954430302853814e-05}
{"train_loss": 0.6166999340057373, "global_step": 12153, "epoch": 136, "lr": 9.954422493312728e-05}
{"train_loss": 0.688955545425415, "global_step": 12154, "epoch": 136, "lr": 9.95441468310558e-05}
{"train_loss": 0.6389061212539673, "global_step": 12155, "epoch": 136, "lr": 9.954406872232371e-05}
{"train_loss": 0.652815043926239, "global_step": 12156, "epoch": 136, "lr": 9.954399060693103e-05}
{"train_loss": 0.7465694546699524, "global_step": 12157, "epoch": 136, "lr": 9.954391248487776e-05}
{"train_loss": 0.6118700504302979, "global_step": 12158, "epoch": 136, "lr": 9.954383435616392e-05}
{"train_loss": 0.6372926235198975, "global_step": 12159, "epoch": 136, "lr": 9.954375622078951e-05}
{"train_loss": 0.8062852621078491, "global_step": 12160, "epoch": 136, "lr": 9.954367807875454e-05}
{"train_loss": 0.5669126510620117, "global_step": 12161, "epoch": 136, "lr": 9.954359993005903e-05}
{"train_loss": 0.7782608866691589, "global_step": 12162, "epoch": 136, "lr": 9.9543521774703e-05}
{"train_loss": 0.5573168396949768, "global_step": 12163, "epoch": 136, "lr": 9.954344361268644e-05}
{"train_loss": 0.6635924577713013, "global_step": 12164, "epoch": 136, "lr": 9.954336544400936e-05}
{"train_loss": 0.734765350818634, "global_step": 12165, "epoch": 136, "lr": 9.954328726867179e-05}
{"train_loss": 0.6490098834037781, "global_step": 12166, "epoch": 136, "lr": 9.954320908667372e-05}
{"train_loss": 0.8174356818199158, "global_step": 12167, "epoch": 136, "lr": 9.954313089801517e-05}
{"train_loss": 0.8337280750274658, "global_step": 12168, "epoch": 136, "lr": 9.954305270269615e-05}
{"train_loss": 0.7612570524215698, "global_step": 12169, "epoch": 136, "lr": 9.954297450071667e-05}
{"train_loss": 0.7176048755645752, "global_step": 12170, "epoch": 136, "lr": 9.954289629207673e-05}
{"train_loss": 0.6378470063209534, "global_step": 12171, "epoch": 136, "lr": 9.954281807677637e-05}
{"train_loss": 0.6023414731025696, "global_step": 12172, "epoch": 136, "lr": 9.954273985481559e-05}
{"train_loss": 0.8331955075263977, "global_step": 12173, "epoch": 136, "lr": 9.954266162619437e-05}
{"train_loss": 0.7275745868682861, "global_step": 12174, "epoch": 136, "lr": 9.954258339091274e-05}
{"train_loss": 0.7423756718635559, "global_step": 12175, "epoch": 136, "lr": 9.954250514897074e-05}
{"train_loss": 0.5293347835540771, "global_step": 12176, "epoch": 136, "lr": 9.954242690036833e-05}
{"train_loss": 0.6078541874885559, "global_step": 12177, "epoch": 136, "lr": 9.954234864510556e-05}
{"train_loss": 0.5801414847373962, "global_step": 12178, "epoch": 136, "lr": 9.954227038318243e-05}
{"train_loss": 0.8820920586585999, "global_step": 12179, "epoch": 136, "lr": 9.954219211459892e-05}
{"train_loss": 0.5556574463844299, "global_step": 12180, "epoch": 136, "lr": 9.954211383935507e-05}
{"train_loss": 0.5811880230903625, "global_step": 12181, "epoch": 136, "lr": 9.954203555745091e-05}
{"train_loss": 0.6598969101905823, "global_step": 12182, "epoch": 136, "lr": 9.954195726888641e-05}
{"train_loss": 0.8949925899505615, "global_step": 12183, "epoch": 136, "lr": 9.954187897366161e-05}
{"train_loss": 0.7857298851013184, "global_step": 12184, "epoch": 136, "lr": 9.95418006717765e-05}
{"train_loss": 0.7611315250396729, "global_step": 12185, "epoch": 136, "lr": 9.95417223632311e-05}
{"train_loss": 0.5670389533042908, "global_step": 12186, "epoch": 136, "lr": 9.95416440480254e-05}
{"train_loss": 0.6253071427345276, "global_step": 12187, "epoch": 136, "lr": 9.954156572615946e-05}
{"train_loss": 0.7829475402832031, "global_step": 12188, "epoch": 136, "lr": 9.954148739763325e-05}
{"train_loss": 0.5800694823265076, "global_step": 12189, "epoch": 136, "lr": 9.954140906244677e-05}
{"train_loss": 0.6875024437904358, "global_step": 12190, "epoch": 136, "lr": 9.954133072060008e-05}
{"train_loss": 0.6553657650947571, "global_step": 12191, "epoch": 136, "lr": 9.954125237209315e-05}
{"train_loss": 0.6672876865676279, "global_step": 12192, "epoch": 136, "lr": 9.9541174016926e-05, "val_loss": 1.3808199167251587}
{"train_loss": 0.672959566116333, "global_step": 12193, "epoch": 137, "lr": 9.954109565509866e-05}
{"train_loss": 0.6326478719711304, "global_step": 12194, "epoch": 137, "lr": 9.954101728661112e-05}
{"train_loss": 0.6747826337814331, "global_step": 12195, "epoch": 137, "lr": 9.954093891146336e-05}
{"train_loss": 0.622180163860321, "global_step": 12196, "epoch": 137, "lr": 9.954086052965547e-05}
{"train_loss": 0.7868309617042542, "global_step": 12197, "epoch": 137, "lr": 9.954078214118739e-05}
{"train_loss": 0.7008556127548218, "global_step": 12198, "epoch": 137, "lr": 9.954070374605916e-05}
{"train_loss": 0.6457375884056091, "global_step": 12199, "epoch": 137, "lr": 9.954062534427079e-05}
{"train_loss": 0.7443093061447144, "global_step": 12200, "epoch": 137, "lr": 9.954054693582229e-05}
{"train_loss": 0.7619324922561646, "global_step": 12201, "epoch": 137, "lr": 9.954046852071365e-05}
{"train_loss": 0.6784902215003967, "global_step": 12202, "epoch": 137, "lr": 9.954039009894491e-05}
{"train_loss": 0.7031581997871399, "global_step": 12203, "epoch": 137, "lr": 9.954031167051607e-05}
{"train_loss": 0.8113654255867004, "global_step": 12204, "epoch": 137, "lr": 9.954023323542713e-05}
{"train_loss": 0.7064971923828125, "global_step": 12205, "epoch": 137, "lr": 9.95401547936781e-05}
{"train_loss": 0.6320571899414062, "global_step": 12206, "epoch": 137, "lr": 9.954007634526903e-05}
{"train_loss": 0.7173015475273132, "global_step": 12207, "epoch": 137, "lr": 9.953999789019989e-05}
{"train_loss": 0.6432335376739502, "global_step": 12208, "epoch": 137, "lr": 9.953991942847069e-05}
{"train_loss": 0.7061081528663635, "global_step": 12209, "epoch": 137, "lr": 9.953984096008146e-05}
{"train_loss": 0.8200691342353821, "global_step": 12210, "epoch": 137, "lr": 9.953976248503221e-05}
{"train_loss": 0.7190188765525818, "global_step": 12211, "epoch": 137, "lr": 9.953968400332292e-05}
{"train_loss": 0.5148130655288696, "global_step": 12212, "epoch": 137, "lr": 9.953960551495364e-05}
{"train_loss": 0.7824246287345886, "global_step": 12213, "epoch": 137, "lr": 9.953952701992438e-05}
{"train_loss": 0.6387430429458618, "global_step": 12214, "epoch": 137, "lr": 9.95394485182351e-05}
{"train_loss": 0.501426637172699, "global_step": 12215, "epoch": 137, "lr": 9.953937000988586e-05}
{"train_loss": 0.6852730512619019, "global_step": 12216, "epoch": 137, "lr": 9.953929149487667e-05}
{"train_loss": 0.673068106174469, "global_step": 12217, "epoch": 137, "lr": 9.95392129732075e-05}
{"train_loss": 0.7339427471160889, "global_step": 12218, "epoch": 137, "lr": 9.953913444487839e-05}
{"train_loss": 0.5923411846160889, "global_step": 12219, "epoch": 137, "lr": 9.953905590988937e-05}
{"train_loss": 0.5814274549484253, "global_step": 12220, "epoch": 137, "lr": 9.95389773682404e-05}
{"train_loss": 0.5874804854393005, "global_step": 12221, "epoch": 137, "lr": 9.953889881993154e-05}
{"train_loss": 0.6461207270622253, "global_step": 12222, "epoch": 137, "lr": 9.953882026496276e-05}
{"train_loss": 0.706365168094635, "global_step": 12223, "epoch": 137, "lr": 9.953874170333411e-05}
{"train_loss": 0.5930928587913513, "global_step": 12224, "epoch": 137, "lr": 9.953866313504557e-05}
{"train_loss": 0.6531505584716797, "global_step": 12225, "epoch": 137, "lr": 9.953858456009715e-05}
{"train_loss": 0.5144643187522888, "global_step": 12226, "epoch": 137, "lr": 9.953850597848888e-05}
{"train_loss": 0.5542773008346558, "global_step": 12227, "epoch": 137, "lr": 9.953842739022077e-05}
{"train_loss": 0.5706985592842102, "global_step": 12228, "epoch": 137, "lr": 9.953834879529281e-05}
{"train_loss": 0.8123918771743774, "global_step": 12229, "epoch": 137, "lr": 9.953827019370502e-05}
{"train_loss": 0.7309770584106445, "global_step": 12230, "epoch": 137, "lr": 9.953819158545743e-05}
{"train_loss": 0.8981388211250305, "global_step": 12231, "epoch": 137, "lr": 9.953811297055002e-05}
{"train_loss": 0.7778119444847107, "global_step": 12232, "epoch": 137, "lr": 9.953803434898284e-05}
{"train_loss": 0.5466780066490173, "global_step": 12233, "epoch": 137, "lr": 9.953795572075584e-05}
{"train_loss": 0.6896065473556519, "global_step": 12234, "epoch": 137, "lr": 9.953787708586908e-05}
{"train_loss": 0.9026975035667419, "global_step": 12235, "epoch": 137, "lr": 9.953779844432256e-05}
{"train_loss": 0.6768383979797363, "global_step": 12236, "epoch": 137, "lr": 9.953771979611629e-05}
{"train_loss": 0.5770865082740784, "global_step": 12237, "epoch": 137, "lr": 9.953764114125026e-05}
{"train_loss": 0.6565168499946594, "global_step": 12238, "epoch": 137, "lr": 9.95375624797245e-05}
{"train_loss": 0.6334168314933777, "global_step": 12239, "epoch": 137, "lr": 9.953748381153904e-05}
{"train_loss": 0.6047396063804626, "global_step": 12240, "epoch": 137, "lr": 9.953740513669385e-05}
{"train_loss": 0.5622694492340088, "global_step": 12241, "epoch": 137, "lr": 9.953732645518896e-05}
{"train_loss": 0.6131417751312256, "global_step": 12242, "epoch": 137, "lr": 9.95372477670244e-05}
{"train_loss": 0.6763048768043518, "global_step": 12243, "epoch": 137, "lr": 9.953716907220015e-05}
{"train_loss": 0.5523882508277893, "global_step": 12244, "epoch": 137, "lr": 9.95370903707162e-05}
{"train_loss": 0.462777704000473, "global_step": 12245, "epoch": 137, "lr": 9.953701166257265e-05}
{"train_loss": 0.7378501296043396, "global_step": 12246, "epoch": 137, "lr": 9.953693294776942e-05}
{"train_loss": 0.5480207800865173, "global_step": 12247, "epoch": 137, "lr": 9.953685422630654e-05}
{"train_loss": 0.7179563641548157, "global_step": 12248, "epoch": 137, "lr": 9.953677549818405e-05}
{"train_loss": 0.5769345760345459, "global_step": 12249, "epoch": 137, "lr": 9.953669676340196e-05}
{"train_loss": 0.7364102602005005, "global_step": 12250, "epoch": 137, "lr": 9.953661802196023e-05}
{"train_loss": 0.6524614095687866, "global_step": 12251, "epoch": 137, "lr": 9.953653927385893e-05}
{"train_loss": 0.730250358581543, "global_step": 12252, "epoch": 137, "lr": 9.953646051909804e-05}
{"train_loss": 0.6106017231941223, "global_step": 12253, "epoch": 137, "lr": 9.953638175767758e-05}
{"train_loss": 0.5375465750694275, "global_step": 12254, "epoch": 137, "lr": 9.953630298959755e-05}
{"train_loss": 0.7855671048164368, "global_step": 12255, "epoch": 137, "lr": 9.953622421485797e-05}
{"train_loss": 0.6215505003929138, "global_step": 12256, "epoch": 137, "lr": 9.953614543345886e-05}
{"train_loss": 0.6937087178230286, "global_step": 12257, "epoch": 137, "lr": 9.953606664540021e-05}
{"train_loss": 0.5913522839546204, "global_step": 12258, "epoch": 137, "lr": 9.953598785068203e-05}
{"train_loss": 0.7052698731422424, "global_step": 12259, "epoch": 137, "lr": 9.953590904930437e-05}
{"train_loss": 0.7921631336212158, "global_step": 12260, "epoch": 137, "lr": 9.953583024126718e-05}
{"train_loss": 0.6158683896064758, "global_step": 12261, "epoch": 137, "lr": 9.953575142657052e-05}
{"train_loss": 0.8738695383071899, "global_step": 12262, "epoch": 137, "lr": 9.953567260521436e-05}
{"train_loss": 0.7455348372459412, "global_step": 12263, "epoch": 137, "lr": 9.953559377719874e-05}
{"train_loss": 0.6705885529518127, "global_step": 12264, "epoch": 137, "lr": 9.953551494252368e-05}
{"train_loss": 0.620496928691864, "global_step": 12265, "epoch": 137, "lr": 9.953543610118917e-05}
{"train_loss": 0.7270707488059998, "global_step": 12266, "epoch": 137, "lr": 9.953535725319521e-05}
{"train_loss": 0.6873804330825806, "global_step": 12267, "epoch": 137, "lr": 9.953527839854184e-05}
{"train_loss": 0.6945475935935974, "global_step": 12268, "epoch": 137, "lr": 9.953519953722904e-05}
{"train_loss": 0.7560247778892517, "global_step": 12269, "epoch": 137, "lr": 9.953512066925684e-05}
{"train_loss": 0.550930917263031, "global_step": 12270, "epoch": 137, "lr": 9.953504179462524e-05}
{"train_loss": 0.6298056840896606, "global_step": 12271, "epoch": 137, "lr": 9.953496291333429e-05}
{"train_loss": 0.7220227122306824, "global_step": 12272, "epoch": 137, "lr": 9.953488402538393e-05}
{"train_loss": 0.6925746202468872, "global_step": 12273, "epoch": 137, "lr": 9.953480513077424e-05}
{"train_loss": 0.5611464381217957, "global_step": 12274, "epoch": 137, "lr": 9.953472622950517e-05}
{"train_loss": 0.8573110699653625, "global_step": 12275, "epoch": 137, "lr": 9.953464732157678e-05}
{"train_loss": 0.7388224601745605, "global_step": 12276, "epoch": 137, "lr": 9.953456840698905e-05}
{"train_loss": 0.7031459212303162, "global_step": 12277, "epoch": 137, "lr": 9.9534489485742e-05}
{"train_loss": 0.7594132423400879, "global_step": 12278, "epoch": 137, "lr": 9.953441055783565e-05}
{"train_loss": 0.7859203219413757, "global_step": 12279, "epoch": 137, "lr": 9.953433162327001e-05}
{"train_loss": 0.5541264414787292, "global_step": 12280, "epoch": 137, "lr": 9.953425268204507e-05}
{"train_loss": 0.672989678181959, "global_step": 12281, "epoch": 137, "lr": 9.953417373416085e-05, "val_loss": 1.298633337020874}
{"train_loss": 0.5533518195152283, "global_step": 12282, "epoch": 138, "lr": 9.953409477961737e-05}
{"train_loss": 0.6867587566375732, "global_step": 12283, "epoch": 138, "lr": 9.953401581841464e-05}
{"train_loss": 0.748631477355957, "global_step": 12284, "epoch": 138, "lr": 9.953393685055266e-05}
{"train_loss": 0.603253960609436, "global_step": 12285, "epoch": 138, "lr": 9.953385787603144e-05}
{"train_loss": 0.7045930027961731, "global_step": 12286, "epoch": 138, "lr": 9.9533778894851e-05}
{"train_loss": 0.5981569886207581, "global_step": 12287, "epoch": 138, "lr": 9.953369990701136e-05}
{"train_loss": 0.7481899857521057, "global_step": 12288, "epoch": 138, "lr": 9.953362091251251e-05}
{"train_loss": 0.49643903970718384, "global_step": 12289, "epoch": 138, "lr": 9.953354191135445e-05}
{"train_loss": 0.501904308795929, "global_step": 12290, "epoch": 138, "lr": 9.953346290353722e-05}
{"train_loss": 0.7072497010231018, "global_step": 12291, "epoch": 138, "lr": 9.953338388906084e-05}
{"train_loss": 0.7039124965667725, "global_step": 12292, "epoch": 138, "lr": 9.953330486792527e-05}
{"train_loss": 0.6619575023651123, "global_step": 12293, "epoch": 138, "lr": 9.953322584013057e-05}
{"train_loss": 0.5981982350349426, "global_step": 12294, "epoch": 138, "lr": 9.953314680567673e-05}
{"train_loss": 0.5292513370513916, "global_step": 12295, "epoch": 138, "lr": 9.953306776456376e-05}
{"train_loss": 0.6491765975952148, "global_step": 12296, "epoch": 138, "lr": 9.953298871679166e-05}
{"train_loss": 0.5621391534805298, "global_step": 12297, "epoch": 138, "lr": 9.953290966236047e-05}
{"train_loss": 0.6218071579933167, "global_step": 12298, "epoch": 138, "lr": 9.953283060127019e-05}
{"train_loss": 0.5243081450462341, "global_step": 12299, "epoch": 138, "lr": 9.953275153352081e-05}
{"train_loss": 0.7151947021484375, "global_step": 12300, "epoch": 138, "lr": 9.953267245911236e-05}
{"train_loss": 0.6700742244720459, "global_step": 12301, "epoch": 138, "lr": 9.953259337804484e-05}
{"train_loss": 0.5605593919754028, "global_step": 12302, "epoch": 138, "lr": 9.953251429031827e-05}
{"train_loss": 0.6176478266716003, "global_step": 12303, "epoch": 138, "lr": 9.953243519593266e-05}
{"train_loss": 0.6232742667198181, "global_step": 12304, "epoch": 138, "lr": 9.953235609488803e-05}
{"train_loss": 0.6221913695335388, "global_step": 12305, "epoch": 138, "lr": 9.953227698718435e-05}
{"train_loss": 0.7628750205039978, "global_step": 12306, "epoch": 138, "lr": 9.953219787282168e-05}
{"train_loss": 0.7258380055427551, "global_step": 12307, "epoch": 138, "lr": 9.95321187518e-05}
{"train_loss": 0.8238042593002319, "global_step": 12308, "epoch": 138, "lr": 9.953203962411933e-05}
{"train_loss": 0.6109225749969482, "global_step": 12309, "epoch": 138, "lr": 9.953196048977969e-05}
{"train_loss": 0.6988196969032288, "global_step": 12310, "epoch": 138, "lr": 9.953188134878107e-05}
{"train_loss": 0.6915238499641418, "global_step": 12311, "epoch": 138, "lr": 9.95318022011235e-05}
{"train_loss": 0.8323647379875183, "global_step": 12312, "epoch": 138, "lr": 9.953172304680698e-05}
{"train_loss": 0.8329852223396301, "global_step": 12313, "epoch": 138, "lr": 9.953164388583153e-05}
{"train_loss": 0.5301550030708313, "global_step": 12314, "epoch": 138, "lr": 9.953156471819715e-05}
{"train_loss": 0.6901735067367554, "global_step": 12315, "epoch": 138, "lr": 9.953148554390386e-05}
{"train_loss": 0.7026307582855225, "global_step": 12316, "epoch": 138, "lr": 9.953140636295165e-05}
{"train_loss": 0.6587411165237427, "global_step": 12317, "epoch": 138, "lr": 9.953132717534055e-05}
{"train_loss": 0.6201616525650024, "global_step": 12318, "epoch": 138, "lr": 9.953124798107058e-05}
{"train_loss": 0.6215418577194214, "global_step": 12319, "epoch": 138, "lr": 9.953116878014172e-05}
{"train_loss": 0.6574450731277466, "global_step": 12320, "epoch": 138, "lr": 9.953108957255401e-05}
{"train_loss": 0.6575677394866943, "global_step": 12321, "epoch": 138, "lr": 9.953101035830745e-05}
{"train_loss": 0.6512770652770996, "global_step": 12322, "epoch": 138, "lr": 9.953093113740205e-05}
{"train_loss": 0.5845723152160645, "global_step": 12323, "epoch": 138, "lr": 9.953085190983782e-05}
{"train_loss": 0.6024545431137085, "global_step": 12324, "epoch": 138, "lr": 9.953077267561477e-05}
{"train_loss": 0.5704382658004761, "global_step": 12325, "epoch": 138, "lr": 9.95306934347329e-05}
{"train_loss": 0.5980488061904907, "global_step": 12326, "epoch": 138, "lr": 9.953061418719224e-05}
{"train_loss": 0.6227613091468811, "global_step": 12327, "epoch": 138, "lr": 9.953053493299279e-05}
{"train_loss": 0.5693174600601196, "global_step": 12328, "epoch": 138, "lr": 9.953045567213456e-05}
{"train_loss": 0.654214084148407, "global_step": 12329, "epoch": 138, "lr": 9.953037640461758e-05}
{"train_loss": 0.5965940952301025, "global_step": 12330, "epoch": 138, "lr": 9.953029713044184e-05}
{"train_loss": 0.6636528968811035, "global_step": 12331, "epoch": 138, "lr": 9.953021784960735e-05}
{"train_loss": 0.7110806107521057, "global_step": 12332, "epoch": 138, "lr": 9.953013856211411e-05}
{"train_loss": 0.5086458921432495, "global_step": 12333, "epoch": 138, "lr": 9.953005926796218e-05}
{"train_loss": 0.5984501838684082, "global_step": 12334, "epoch": 138, "lr": 9.952997996715151e-05}
{"train_loss": 0.5335564613342285, "global_step": 12335, "epoch": 138, "lr": 9.952990065968214e-05}
{"train_loss": 0.6927761435508728, "global_step": 12336, "epoch": 138, "lr": 9.952982134555409e-05}
{"train_loss": 0.7631779909133911, "global_step": 12337, "epoch": 138, "lr": 9.952974202476736e-05}
{"train_loss": 0.5636845827102661, "global_step": 12338, "epoch": 138, "lr": 9.952966269732196e-05}
{"train_loss": 0.597165584564209, "global_step": 12339, "epoch": 138, "lr": 9.95295833632179e-05}
{"train_loss": 0.5701271891593933, "global_step": 12340, "epoch": 138, "lr": 9.952950402245518e-05}
{"train_loss": 0.6017717719078064, "global_step": 12341, "epoch": 138, "lr": 9.952942467503382e-05}
{"train_loss": 0.7132281064987183, "global_step": 12342, "epoch": 138, "lr": 9.952934532095384e-05}
{"train_loss": 0.6196277141571045, "global_step": 12343, "epoch": 138, "lr": 9.952926596021525e-05}
{"train_loss": 0.44842880964279175, "global_step": 12344, "epoch": 138, "lr": 9.952918659281805e-05}
{"train_loss": 0.6961823105812073, "global_step": 12345, "epoch": 138, "lr": 9.952910721876225e-05}
{"train_loss": 0.7306109070777893, "global_step": 12346, "epoch": 138, "lr": 9.952902783804787e-05}
{"train_loss": 0.7907782793045044, "global_step": 12347, "epoch": 138, "lr": 9.95289484506749e-05}
{"train_loss": 0.5596542954444885, "global_step": 12348, "epoch": 138, "lr": 9.952886905664338e-05}
{"train_loss": 0.6359184384346008, "global_step": 12349, "epoch": 138, "lr": 9.952878965595331e-05}
{"train_loss": 0.7104926705360413, "global_step": 12350, "epoch": 138, "lr": 9.952871024860471e-05}
{"train_loss": 0.589961051940918, "global_step": 12351, "epoch": 138, "lr": 9.952863083459756e-05}
{"train_loss": 0.6278480291366577, "global_step": 12352, "epoch": 138, "lr": 9.952855141393189e-05}
{"train_loss": 0.7234201431274414, "global_step": 12353, "epoch": 138, "lr": 9.95284719866077e-05}
{"train_loss": 0.5341653227806091, "global_step": 12354, "epoch": 138, "lr": 9.952839255262502e-05}
{"train_loss": 0.7415170669555664, "global_step": 12355, "epoch": 138, "lr": 9.952831311198388e-05}
{"train_loss": 0.6243258118629456, "global_step": 12356, "epoch": 138, "lr": 9.952823366468423e-05}
{"train_loss": 0.6067906618118286, "global_step": 12357, "epoch": 138, "lr": 9.952815421072612e-05}
{"train_loss": 0.6976311206817627, "global_step": 12358, "epoch": 138, "lr": 9.952807475010954e-05}
{"train_loss": 0.681347668170929, "global_step": 12359, "epoch": 138, "lr": 9.952799528283454e-05}
{"train_loss": 0.6466715931892395, "global_step": 12360, "epoch": 138, "lr": 9.952791580890108e-05}
{"train_loss": 0.8045307397842407, "global_step": 12361, "epoch": 138, "lr": 9.952783632830922e-05}
{"train_loss": 0.6186514496803284, "global_step": 12362, "epoch": 138, "lr": 9.952775684105892e-05}
{"train_loss": 0.679899275302887, "global_step": 12363, "epoch": 138, "lr": 9.952767734715024e-05}
{"train_loss": 0.7476149201393127, "global_step": 12364, "epoch": 138, "lr": 9.952759784658315e-05}
{"train_loss": 0.7079416513442993, "global_step": 12365, "epoch": 138, "lr": 9.95275183393577e-05}
{"train_loss": 0.5721057057380676, "global_step": 12366, "epoch": 138, "lr": 9.952743882547387e-05}
{"train_loss": 0.5549818873405457, "global_step": 12367, "epoch": 138, "lr": 9.952735930493166e-05}
{"train_loss": 0.4881477952003479, "global_step": 12368, "epoch": 138, "lr": 9.95272797777311e-05}
{"train_loss": 0.830595076084137, "global_step": 12369, "epoch": 138, "lr": 9.952720024387224e-05}
{"train_loss": 0.643890551636728, "global_step": 12370, "epoch": 138, "lr": 9.952712070335501e-05, "val_loss": 1.3686012029647827}
{"train_loss": 0.5388387441635132, "global_step": 12371, "epoch": 139, "lr": 9.952704115617948e-05}
{"train_loss": 0.6831316947937012, "global_step": 12372, "epoch": 139, "lr": 9.952696160234565e-05}
{"train_loss": 0.6815296411514282, "global_step": 12373, "epoch": 139, "lr": 9.95268820418535e-05}
{"train_loss": 0.7924622893333435, "global_step": 12374, "epoch": 139, "lr": 9.95268024747031e-05}
{"train_loss": 0.851893961429596, "global_step": 12375, "epoch": 139, "lr": 9.95267229008944e-05}
{"train_loss": 0.5673679709434509, "global_step": 12376, "epoch": 139, "lr": 9.952664332042743e-05}
{"train_loss": 0.6335466504096985, "global_step": 12377, "epoch": 139, "lr": 9.952656373330222e-05}
{"train_loss": 0.6551878452301025, "global_step": 12378, "epoch": 139, "lr": 9.952648413951877e-05}
{"train_loss": 0.7070318460464478, "global_step": 12379, "epoch": 139, "lr": 9.952640453907706e-05}
{"train_loss": 0.7220333814620972, "global_step": 12380, "epoch": 139, "lr": 9.952632493197717e-05}
{"train_loss": 0.6103547215461731, "global_step": 12381, "epoch": 139, "lr": 9.952624531821905e-05}
{"train_loss": 0.45960721373558044, "global_step": 12382, "epoch": 139, "lr": 9.952616569780274e-05}
{"train_loss": 0.7437505722045898, "global_step": 12383, "epoch": 139, "lr": 9.952608607072822e-05}
{"train_loss": 0.5852245092391968, "global_step": 12384, "epoch": 139, "lr": 9.952600643699552e-05}
{"train_loss": 0.7738822102546692, "global_step": 12385, "epoch": 139, "lr": 9.952592679660466e-05}
{"train_loss": 0.7717282772064209, "global_step": 12386, "epoch": 139, "lr": 9.952584714955566e-05}
{"train_loss": 0.6298678517341614, "global_step": 12387, "epoch": 139, "lr": 9.95257674958485e-05}
{"train_loss": 0.529671847820282, "global_step": 12388, "epoch": 139, "lr": 9.95256878354832e-05}
{"train_loss": 0.49955594539642334, "global_step": 12389, "epoch": 139, "lr": 9.952560816845977e-05}
{"train_loss": 0.5577688217163086, "global_step": 12390, "epoch": 139, "lr": 9.952552849477825e-05}
{"train_loss": 0.46516650915145874, "global_step": 12391, "epoch": 139, "lr": 9.952544881443861e-05}
{"train_loss": 0.7418863773345947, "global_step": 12392, "epoch": 139, "lr": 9.952536912744089e-05}
{"train_loss": 0.7492933869361877, "global_step": 12393, "epoch": 139, "lr": 9.952528943378505e-05}
{"train_loss": 0.6058919429779053, "global_step": 12394, "epoch": 139, "lr": 9.952520973347117e-05}
{"train_loss": 0.5379857420921326, "global_step": 12395, "epoch": 139, "lr": 9.952513002649922e-05}
{"train_loss": 0.5582795739173889, "global_step": 12396, "epoch": 139, "lr": 9.952505031286923e-05}
{"train_loss": 0.7007277011871338, "global_step": 12397, "epoch": 139, "lr": 9.952497059258119e-05}
{"train_loss": 0.5559036135673523, "global_step": 12398, "epoch": 139, "lr": 9.952489086563512e-05}
{"train_loss": 0.5499362349510193, "global_step": 12399, "epoch": 139, "lr": 9.952481113203105e-05}
{"train_loss": 0.6258617639541626, "global_step": 12400, "epoch": 139, "lr": 9.952473139176896e-05}
{"train_loss": 0.6908699870109558, "global_step": 12401, "epoch": 139, "lr": 9.952465164484887e-05}
{"train_loss": 0.5752479434013367, "global_step": 12402, "epoch": 139, "lr": 9.952457189127081e-05}
{"train_loss": 0.6866391897201538, "global_step": 12403, "epoch": 139, "lr": 9.952449213103475e-05}
{"train_loss": 0.7466406226158142, "global_step": 12404, "epoch": 139, "lr": 9.952441236414074e-05}
{"train_loss": 0.6787713766098022, "global_step": 12405, "epoch": 139, "lr": 9.952433259058878e-05}
{"train_loss": 0.7450379729270935, "global_step": 12406, "epoch": 139, "lr": 9.952425281037886e-05}
{"train_loss": 0.5555907487869263, "global_step": 12407, "epoch": 139, "lr": 9.952417302351102e-05}
{"train_loss": 0.5932378768920898, "global_step": 12408, "epoch": 139, "lr": 9.952409322998526e-05}
{"train_loss": 0.6562387347221375, "global_step": 12409, "epoch": 139, "lr": 9.95240134298016e-05}
{"train_loss": 0.8018048405647278, "global_step": 12410, "epoch": 139, "lr": 9.952393362296002e-05}
{"train_loss": 0.7835132479667664, "global_step": 12411, "epoch": 139, "lr": 9.952385380946057e-05}
{"train_loss": 0.7455257773399353, "global_step": 12412, "epoch": 139, "lr": 9.952377398930322e-05}
{"train_loss": 0.5655220746994019, "global_step": 12413, "epoch": 139, "lr": 9.952369416248802e-05}
{"train_loss": 0.5309361219406128, "global_step": 12414, "epoch": 139, "lr": 9.952361432901493e-05}
{"train_loss": 0.8306147456169128, "global_step": 12415, "epoch": 139, "lr": 9.952353448888404e-05}
{"train_loss": 0.567974865436554, "global_step": 12416, "epoch": 139, "lr": 9.95234546420953e-05}
{"train_loss": 0.5880061984062195, "global_step": 12417, "epoch": 139, "lr": 9.952337478864871e-05}
{"train_loss": 0.7375210523605347, "global_step": 12418, "epoch": 139, "lr": 9.952329492854433e-05}
{"train_loss": 0.6587267518043518, "global_step": 12419, "epoch": 139, "lr": 9.952321506178213e-05}
{"train_loss": 0.5188674330711365, "global_step": 12420, "epoch": 139, "lr": 9.952313518836216e-05}
{"train_loss": 0.5976359844207764, "global_step": 12421, "epoch": 139, "lr": 9.952305530828439e-05}
{"train_loss": 0.7652783989906311, "global_step": 12422, "epoch": 139, "lr": 9.952297542154884e-05}
{"train_loss": 0.6742771863937378, "global_step": 12423, "epoch": 139, "lr": 9.952289552815555e-05}
{"train_loss": 0.6291220188140869, "global_step": 12424, "epoch": 139, "lr": 9.952281562810451e-05}
{"train_loss": 0.7094114422798157, "global_step": 12425, "epoch": 139, "lr": 9.952273572139573e-05}
{"train_loss": 0.7533615231513977, "global_step": 12426, "epoch": 139, "lr": 9.952265580802922e-05}
{"train_loss": 0.667472243309021, "global_step": 12427, "epoch": 139, "lr": 9.952257588800498e-05}
{"train_loss": 0.622099757194519, "global_step": 12428, "epoch": 139, "lr": 9.952249596132305e-05}
{"train_loss": 0.6797145009040833, "global_step": 12429, "epoch": 139, "lr": 9.952241602798343e-05}
{"train_loss": 0.605869472026825, "global_step": 12430, "epoch": 139, "lr": 9.95223360879861e-05}
{"train_loss": 0.5581291317939758, "global_step": 12431, "epoch": 139, "lr": 9.95222561413311e-05}
{"train_loss": 0.8789857625961304, "global_step": 12432, "epoch": 139, "lr": 9.952217618801844e-05}
{"train_loss": 0.7157503366470337, "global_step": 12433, "epoch": 139, "lr": 9.952209622804814e-05}
{"train_loss": 0.7096903920173645, "global_step": 12434, "epoch": 139, "lr": 9.952201626142019e-05}
{"train_loss": 0.658128023147583, "global_step": 12435, "epoch": 139, "lr": 9.952193628813461e-05}
{"train_loss": 0.8359336853027344, "global_step": 12436, "epoch": 139, "lr": 9.95218563081914e-05}
{"train_loss": 0.6429790258407593, "global_step": 12437, "epoch": 139, "lr": 9.952177632159059e-05}
{"train_loss": 0.8136521577835083, "global_step": 12438, "epoch": 139, "lr": 9.952169632833216e-05}
{"train_loss": 0.6929916143417358, "global_step": 12439, "epoch": 139, "lr": 9.952161632841617e-05}
{"train_loss": 0.5480148792266846, "global_step": 12440, "epoch": 139, "lr": 9.952153632184261e-05}
{"train_loss": 0.5891122817993164, "global_step": 12441, "epoch": 139, "lr": 9.952145630861145e-05}
{"train_loss": 0.6868746876716614, "global_step": 12442, "epoch": 139, "lr": 9.952137628872274e-05}
{"train_loss": 0.8908663392066956, "global_step": 12443, "epoch": 139, "lr": 9.95212962621765e-05}
{"train_loss": 0.8927891254425049, "global_step": 12444, "epoch": 139, "lr": 9.952121622897273e-05}
{"train_loss": 0.8644938468933105, "global_step": 12445, "epoch": 139, "lr": 9.95211361891114e-05}
{"train_loss": 0.7500642538070679, "global_step": 12446, "epoch": 139, "lr": 9.95210561425926e-05}
{"train_loss": 0.7157493829727173, "global_step": 12447, "epoch": 139, "lr": 9.952097608941627e-05}
{"train_loss": 0.6877813935279846, "global_step": 12448, "epoch": 139, "lr": 9.952089602958245e-05}
{"train_loss": 0.8067818880081177, "global_step": 12449, "epoch": 139, "lr": 9.952081596309116e-05}
{"train_loss": 0.7587683796882629, "global_step": 12450, "epoch": 139, "lr": 9.952073588994239e-05}
{"train_loss": 0.7663825750350952, "global_step": 12451, "epoch": 139, "lr": 9.952065581013616e-05}
{"train_loss": 0.6012738943099976, "global_step": 12452, "epoch": 139, "lr": 9.952057572367251e-05}
{"train_loss": 0.7197347283363342, "global_step": 12453, "epoch": 139, "lr": 9.95204956305514e-05}
{"train_loss": 0.7387790679931641, "global_step": 12454, "epoch": 139, "lr": 9.952041553077285e-05}
{"train_loss": 0.741506814956665, "global_step": 12455, "epoch": 139, "lr": 9.95203354243369e-05}
{"train_loss": 0.6325180530548096, "global_step": 12456, "epoch": 139, "lr": 9.952025531124354e-05}
{"train_loss": 0.6257269382476807, "global_step": 12457, "epoch": 139, "lr": 9.952017519149278e-05}
{"train_loss": 0.7294867038726807, "global_step": 12458, "epoch": 139, "lr": 9.952009506508464e-05}
{"train_loss": 0.6732994361539905, "global_step": 12459, "epoch": 139, "lr": 9.952001493201912e-05, "val_loss": 1.3959556818008423}
{"train_loss": 0.755574643611908, "global_step": 12460, "epoch": 140, "lr": 9.951993479229627e-05}
{"train_loss": 0.7140271663665771, "global_step": 12461, "epoch": 140, "lr": 9.951985464591603e-05}
{"train_loss": 0.5850675106048584, "global_step": 12462, "epoch": 140, "lr": 9.951977449287846e-05}
{"train_loss": 0.5684167742729187, "global_step": 12463, "epoch": 140, "lr": 9.951969433318357e-05}
{"train_loss": 0.7922593355178833, "global_step": 12464, "epoch": 140, "lr": 9.951961416683136e-05}
{"train_loss": 0.7098615765571594, "global_step": 12465, "epoch": 140, "lr": 9.951953399382184e-05}
{"train_loss": 0.7412127256393433, "global_step": 12466, "epoch": 140, "lr": 9.9519453814155e-05}
{"train_loss": 0.6715278029441833, "global_step": 12467, "epoch": 140, "lr": 9.95193736278309e-05}
{"train_loss": 0.7163323760032654, "global_step": 12468, "epoch": 140, "lr": 9.951929343484952e-05}
{"train_loss": 0.5601377487182617, "global_step": 12469, "epoch": 140, "lr": 9.951921323521087e-05}
{"train_loss": 0.7672267556190491, "global_step": 12470, "epoch": 140, "lr": 9.951913302891496e-05}
{"train_loss": 0.548908531665802, "global_step": 12471, "epoch": 140, "lr": 9.951905281596182e-05}
{"train_loss": 0.7256297469139099, "global_step": 12472, "epoch": 140, "lr": 9.951897259635143e-05}
{"train_loss": 0.5932292342185974, "global_step": 12473, "epoch": 140, "lr": 9.951889237008381e-05}
{"train_loss": 0.5764659643173218, "global_step": 12474, "epoch": 140, "lr": 9.9518812137159e-05}
{"train_loss": 0.6851611137390137, "global_step": 12475, "epoch": 140, "lr": 9.951873189757699e-05}
{"train_loss": 0.6135022640228271, "global_step": 12476, "epoch": 140, "lr": 9.951865165133778e-05}
{"train_loss": 0.5668686032295227, "global_step": 12477, "epoch": 140, "lr": 9.951857139844138e-05}
{"train_loss": 0.6347078680992126, "global_step": 12478, "epoch": 140, "lr": 9.951849113888784e-05}
{"train_loss": 0.9319524765014648, "global_step": 12479, "epoch": 140, "lr": 9.951841087267713e-05}
{"train_loss": 0.6220084428787231, "global_step": 12480, "epoch": 140, "lr": 9.951833059980927e-05}
{"train_loss": 0.8183575868606567, "global_step": 12481, "epoch": 140, "lr": 9.951825032028427e-05}
{"train_loss": 0.6268239617347717, "global_step": 12482, "epoch": 140, "lr": 9.951817003410216e-05}
{"train_loss": 0.663429856300354, "global_step": 12483, "epoch": 140, "lr": 9.951808974126291e-05}
{"train_loss": 0.7830395102500916, "global_step": 12484, "epoch": 140, "lr": 9.951800944176657e-05}
{"train_loss": 0.7753720283508301, "global_step": 12485, "epoch": 140, "lr": 9.951792913561314e-05}
{"train_loss": 0.6958064436912537, "global_step": 12486, "epoch": 140, "lr": 9.951784882280263e-05}
{"train_loss": 0.6910427808761597, "global_step": 12487, "epoch": 140, "lr": 9.951776850333505e-05}
{"train_loss": 0.5114397406578064, "global_step": 12488, "epoch": 140, "lr": 9.95176881772104e-05}
{"train_loss": 0.8318847417831421, "global_step": 12489, "epoch": 140, "lr": 9.951760784442871e-05}
{"train_loss": 0.6596815586090088, "global_step": 12490, "epoch": 140, "lr": 9.951752750498997e-05}
{"train_loss": 0.4560212194919586, "global_step": 12491, "epoch": 140, "lr": 9.951744715889421e-05}
{"train_loss": 0.6720994710922241, "global_step": 12492, "epoch": 140, "lr": 9.951736680614143e-05}
{"train_loss": 0.8428467512130737, "global_step": 12493, "epoch": 140, "lr": 9.951728644673165e-05}
{"train_loss": 0.6098912954330444, "global_step": 12494, "epoch": 140, "lr": 9.951720608066486e-05}
{"train_loss": 0.6160497665405273, "global_step": 12495, "epoch": 140, "lr": 9.951712570794109e-05}
{"train_loss": 0.6336653828620911, "global_step": 12496, "epoch": 140, "lr": 9.951704532856037e-05}
{"train_loss": 0.640763521194458, "global_step": 12497, "epoch": 140, "lr": 9.951696494252266e-05}
{"train_loss": 0.658194363117218, "global_step": 12498, "epoch": 140, "lr": 9.951688454982801e-05}
{"train_loss": 0.6739687323570251, "global_step": 12499, "epoch": 140, "lr": 9.95168041504764e-05}
{"train_loss": 0.5995520353317261, "global_step": 12500, "epoch": 140, "lr": 9.951672374446788e-05}
{"train_loss": 0.5750218629837036, "global_step": 12501, "epoch": 140, "lr": 9.951664333180244e-05}
{"train_loss": 0.6307175159454346, "global_step": 12502, "epoch": 140, "lr": 9.951656291248008e-05}
{"train_loss": 0.6026344895362854, "global_step": 12503, "epoch": 140, "lr": 9.951648248650083e-05}
{"train_loss": 0.6755916476249695, "global_step": 12504, "epoch": 140, "lr": 9.951640205386468e-05}
{"train_loss": 0.5649275779724121, "global_step": 12505, "epoch": 140, "lr": 9.951632161457165e-05}
{"train_loss": 0.6630227565765381, "global_step": 12506, "epoch": 140, "lr": 9.951624116862178e-05}
{"train_loss": 0.6184177398681641, "global_step": 12507, "epoch": 140, "lr": 9.951616071601504e-05}
{"train_loss": 0.6484180688858032, "global_step": 12508, "epoch": 140, "lr": 9.951608025675145e-05}
{"train_loss": 0.7877164483070374, "global_step": 12509, "epoch": 140, "lr": 9.951599979083103e-05}
{"train_loss": 0.5360901355743408, "global_step": 12510, "epoch": 140, "lr": 9.951591931825379e-05}
{"train_loss": 0.6523769497871399, "global_step": 12511, "epoch": 140, "lr": 9.951583883901974e-05}
{"train_loss": 0.666155993938446, "global_step": 12512, "epoch": 140, "lr": 9.951575835312888e-05}
{"train_loss": 0.7009528875350952, "global_step": 12513, "epoch": 140, "lr": 9.951567786058125e-05}
{"train_loss": 0.5129361152648926, "global_step": 12514, "epoch": 140, "lr": 9.95155973613768e-05}
{"train_loss": 0.6362907290458679, "global_step": 12515, "epoch": 140, "lr": 9.951551685551562e-05}
{"train_loss": 0.6630328893661499, "global_step": 12516, "epoch": 140, "lr": 9.951543634299768e-05}
{"train_loss": 0.6927772760391235, "global_step": 12517, "epoch": 140, "lr": 9.951535582382297e-05}
{"train_loss": 0.5489524006843567, "global_step": 12518, "epoch": 140, "lr": 9.951527529799154e-05}
{"train_loss": 0.5238115787506104, "global_step": 12519, "epoch": 140, "lr": 9.951519476550339e-05}
{"train_loss": 0.62534499168396, "global_step": 12520, "epoch": 140, "lr": 9.951511422635851e-05}
{"train_loss": 0.6977957487106323, "global_step": 12521, "epoch": 140, "lr": 9.951503368055694e-05}
{"train_loss": 0.6161545515060425, "global_step": 12522, "epoch": 140, "lr": 9.951495312809866e-05}
{"train_loss": 0.6294627785682678, "global_step": 12523, "epoch": 140, "lr": 9.951487256898372e-05}
{"train_loss": 0.5236740708351135, "global_step": 12524, "epoch": 140, "lr": 9.95147920032121e-05}
{"train_loss": 0.5723906755447388, "global_step": 12525, "epoch": 140, "lr": 9.951471143078381e-05}
{"train_loss": 0.6323874592781067, "global_step": 12526, "epoch": 140, "lr": 9.951463085169889e-05}
{"train_loss": 0.5816264748573303, "global_step": 12527, "epoch": 140, "lr": 9.951455026595731e-05}
{"train_loss": 0.6750193238258362, "global_step": 12528, "epoch": 140, "lr": 9.951446967355911e-05}
{"train_loss": 0.5316312313079834, "global_step": 12529, "epoch": 140, "lr": 9.95143890745043e-05}
{"train_loss": 0.6196960210800171, "global_step": 12530, "epoch": 140, "lr": 9.951430846879288e-05}
{"train_loss": 0.7167640924453735, "global_step": 12531, "epoch": 140, "lr": 9.951422785642486e-05}
{"train_loss": 0.733260452747345, "global_step": 12532, "epoch": 140, "lr": 9.951414723740027e-05}
{"train_loss": 0.6561108231544495, "global_step": 12533, "epoch": 140, "lr": 9.95140666117191e-05}
{"train_loss": 0.6792651414871216, "global_step": 12534, "epoch": 140, "lr": 9.951398597938135e-05}
{"train_loss": 0.5867024064064026, "global_step": 12535, "epoch": 140, "lr": 9.951390534038707e-05}
{"train_loss": 0.6433738470077515, "global_step": 12536, "epoch": 140, "lr": 9.951382469473624e-05}
{"train_loss": 0.7171367406845093, "global_step": 12537, "epoch": 140, "lr": 9.951374404242889e-05}
{"train_loss": 0.865922212600708, "global_step": 12538, "epoch": 140, "lr": 9.951366338346499e-05}
{"train_loss": 0.8147159814834595, "global_step": 12539, "epoch": 140, "lr": 9.95135827178446e-05}
{"train_loss": 0.6675153374671936, "global_step": 12540, "epoch": 140, "lr": 9.951350204556772e-05}
{"train_loss": 0.6319497227668762, "global_step": 12541, "epoch": 140, "lr": 9.951342136663435e-05}
{"train_loss": 0.6637360453605652, "global_step": 12542, "epoch": 140, "lr": 9.95133406810445e-05}
{"train_loss": 0.5787373781204224, "global_step": 12543, "epoch": 140, "lr": 9.951325998879818e-05}
{"train_loss": 0.834738552570343, "global_step": 12544, "epoch": 140, "lr": 9.951317928989541e-05}
{"train_loss": 0.7446596622467041, "global_step": 12545, "epoch": 140, "lr": 9.95130985843362e-05}
{"train_loss": 0.62408047914505, "global_step": 12546, "epoch": 140, "lr": 9.951301787212057e-05}
{"train_loss": 0.6617012023925781, "global_step": 12547, "epoch": 140, "lr": 9.95129371532485e-05}
{"train_loss": 0.6596931590792838, "global_step": 12548, "epoch": 140, "lr": 9.951285642772001e-05, "val_loss": 1.3858565092086792, "train_action_mse_error": 38.286170959472656}
{"train_loss": 0.6011843681335449, "global_step": 12549, "epoch": 141, "lr": 9.951277569553514e-05}
{"train_loss": 0.5223647356033325, "global_step": 12550, "epoch": 141, "lr": 9.951269495669387e-05}
{"train_loss": 0.6844015717506409, "global_step": 12551, "epoch": 141, "lr": 9.951261421119623e-05}
{"train_loss": 0.6676894426345825, "global_step": 12552, "epoch": 141, "lr": 9.951253345904222e-05}
{"train_loss": 0.7177342176437378, "global_step": 12553, "epoch": 141, "lr": 9.951245270023184e-05}
{"train_loss": 0.8021947145462036, "global_step": 12554, "epoch": 141, "lr": 9.951237193476512e-05}
{"train_loss": 0.9155232906341553, "global_step": 12555, "epoch": 141, "lr": 9.951229116264208e-05}
{"train_loss": 0.6227220296859741, "global_step": 12556, "epoch": 141, "lr": 9.951221038386271e-05}
{"train_loss": 0.5574136972427368, "global_step": 12557, "epoch": 141, "lr": 9.951212959842702e-05}
{"train_loss": 0.561481773853302, "global_step": 12558, "epoch": 141, "lr": 9.951204880633504e-05}
{"train_loss": 0.5054140686988831, "global_step": 12559, "epoch": 141, "lr": 9.951196800758676e-05}
{"train_loss": 0.6610986590385437, "global_step": 12560, "epoch": 141, "lr": 9.951188720218218e-05}
{"train_loss": 0.7085759043693542, "global_step": 12561, "epoch": 141, "lr": 9.951180639012136e-05}
{"train_loss": 0.6338909864425659, "global_step": 12562, "epoch": 141, "lr": 9.951172557140427e-05}
{"train_loss": 0.7779282331466675, "global_step": 12563, "epoch": 141, "lr": 9.951164474603093e-05}
{"train_loss": 0.632098913192749, "global_step": 12564, "epoch": 141, "lr": 9.951156391400135e-05}
{"train_loss": 0.5965800285339355, "global_step": 12565, "epoch": 141, "lr": 9.951148307531555e-05}
{"train_loss": 0.7600598931312561, "global_step": 12566, "epoch": 141, "lr": 9.951140222997353e-05}
{"train_loss": 0.5328289270401001, "global_step": 12567, "epoch": 141, "lr": 9.951132137797531e-05}
{"train_loss": 0.5704677700996399, "global_step": 12568, "epoch": 141, "lr": 9.951124051932091e-05}
{"train_loss": 0.57660311460495, "global_step": 12569, "epoch": 141, "lr": 9.95111596540103e-05}
{"train_loss": 0.6565041542053223, "global_step": 12570, "epoch": 141, "lr": 9.951107878204353e-05}
{"train_loss": 0.622775673866272, "global_step": 12571, "epoch": 141, "lr": 9.95109979034206e-05}
{"train_loss": 0.6182776093482971, "global_step": 12572, "epoch": 141, "lr": 9.951091701814153e-05}
{"train_loss": 0.6148000955581665, "global_step": 12573, "epoch": 141, "lr": 9.95108361262063e-05}
{"train_loss": 0.5227128267288208, "global_step": 12574, "epoch": 141, "lr": 9.951075522761494e-05}
{"train_loss": 0.4822986125946045, "global_step": 12575, "epoch": 141, "lr": 9.951067432236749e-05}
{"train_loss": 0.5499526858329773, "global_step": 12576, "epoch": 141, "lr": 9.951059341046392e-05}
{"train_loss": 0.5860302448272705, "global_step": 12577, "epoch": 141, "lr": 9.951051249190423e-05}
{"train_loss": 0.6346331834793091, "global_step": 12578, "epoch": 141, "lr": 9.951043156668849e-05}
{"train_loss": 0.8076801896095276, "global_step": 12579, "epoch": 141, "lr": 9.951035063481666e-05}
{"train_loss": 0.7026807069778442, "global_step": 12580, "epoch": 141, "lr": 9.951026969628875e-05}
{"train_loss": 0.6248310804367065, "global_step": 12581, "epoch": 141, "lr": 9.951018875110482e-05}
{"train_loss": 0.6883867979049683, "global_step": 12582, "epoch": 141, "lr": 9.951010779926484e-05}
{"train_loss": 0.5464637279510498, "global_step": 12583, "epoch": 141, "lr": 9.951002684076883e-05}
{"train_loss": 0.6076652407646179, "global_step": 12584, "epoch": 141, "lr": 9.950994587561679e-05}
{"train_loss": 0.7011100649833679, "global_step": 12585, "epoch": 141, "lr": 9.950986490380874e-05}
{"train_loss": 0.6864115595817566, "global_step": 12586, "epoch": 141, "lr": 9.950978392534471e-05}
{"train_loss": 0.637004017829895, "global_step": 12587, "epoch": 141, "lr": 9.950970294022466e-05}
{"train_loss": 0.7371801733970642, "global_step": 12588, "epoch": 141, "lr": 9.950962194844866e-05}
{"train_loss": 0.5796301364898682, "global_step": 12589, "epoch": 141, "lr": 9.95095409500167e-05}
{"train_loss": 0.5851312875747681, "global_step": 12590, "epoch": 141, "lr": 9.950945994492877e-05}
{"train_loss": 0.6223255395889282, "global_step": 12591, "epoch": 141, "lr": 9.95093789331849e-05}
{"train_loss": 0.6407808065414429, "global_step": 12592, "epoch": 141, "lr": 9.95092979147851e-05}
{"train_loss": 0.5877284407615662, "global_step": 12593, "epoch": 141, "lr": 9.950921688972939e-05}
{"train_loss": 0.7292992472648621, "global_step": 12594, "epoch": 141, "lr": 9.950913585801775e-05}
{"train_loss": 0.6888989210128784, "global_step": 12595, "epoch": 141, "lr": 9.95090548196502e-05}
{"train_loss": 0.6937853097915649, "global_step": 12596, "epoch": 141, "lr": 9.95089737746268e-05}
{"train_loss": 0.7729827165603638, "global_step": 12597, "epoch": 141, "lr": 9.950889272294748e-05}
{"train_loss": 0.571953296661377, "global_step": 12598, "epoch": 141, "lr": 9.950881166461232e-05}
{"train_loss": 0.782966673374176, "global_step": 12599, "epoch": 141, "lr": 9.95087305996213e-05}
{"train_loss": 0.7059147953987122, "global_step": 12600, "epoch": 141, "lr": 9.950864952797443e-05}
{"train_loss": 0.5672569870948792, "global_step": 12601, "epoch": 141, "lr": 9.950856844967173e-05}
{"train_loss": 0.693725049495697, "global_step": 12602, "epoch": 141, "lr": 9.95084873647132e-05}
{"train_loss": 0.5051496028900146, "global_step": 12603, "epoch": 141, "lr": 9.950840627309886e-05}
{"train_loss": 0.667142927646637, "global_step": 12604, "epoch": 141, "lr": 9.950832517482871e-05}
{"train_loss": 0.48332133889198303, "global_step": 12605, "epoch": 141, "lr": 9.950824406990278e-05}
{"train_loss": 0.544085681438446, "global_step": 12606, "epoch": 141, "lr": 9.950816295832107e-05}
{"train_loss": 0.6470267176628113, "global_step": 12607, "epoch": 141, "lr": 9.950808184008359e-05}
{"train_loss": 0.6393389105796814, "global_step": 12608, "epoch": 141, "lr": 9.950800071519035e-05}
{"train_loss": 0.8244380354881287, "global_step": 12609, "epoch": 141, "lr": 9.950791958364137e-05}
{"train_loss": 0.5633734464645386, "global_step": 12610, "epoch": 141, "lr": 9.950783844543664e-05}
{"train_loss": 0.6944010257720947, "global_step": 12611, "epoch": 141, "lr": 9.950775730057619e-05}
{"train_loss": 0.7744108438491821, "global_step": 12612, "epoch": 141, "lr": 9.950767614906002e-05}
{"train_loss": 0.7496306300163269, "global_step": 12613, "epoch": 141, "lr": 9.950759499088815e-05}
{"train_loss": 0.6063894629478455, "global_step": 12614, "epoch": 141, "lr": 9.95075138260606e-05}
{"train_loss": 0.5149906277656555, "global_step": 12615, "epoch": 141, "lr": 9.950743265457736e-05}
{"train_loss": 0.4932444989681244, "global_step": 12616, "epoch": 141, "lr": 9.950735147643844e-05}
{"train_loss": 0.4880240559577942, "global_step": 12617, "epoch": 141, "lr": 9.950727029164388e-05}
{"train_loss": 0.5185213088989258, "global_step": 12618, "epoch": 141, "lr": 9.950718910019366e-05}
{"train_loss": 0.6192867755889893, "global_step": 12619, "epoch": 141, "lr": 9.95071079020878e-05}
{"train_loss": 0.6296758651733398, "global_step": 12620, "epoch": 141, "lr": 9.95070266973263e-05}
{"train_loss": 0.7085895538330078, "global_step": 12621, "epoch": 141, "lr": 9.95069454859092e-05}
{"train_loss": 0.5764596462249756, "global_step": 12622, "epoch": 141, "lr": 9.950686426783649e-05}
{"train_loss": 0.7335304617881775, "global_step": 12623, "epoch": 141, "lr": 9.950678304310818e-05}
{"train_loss": 0.6642608642578125, "global_step": 12624, "epoch": 141, "lr": 9.95067018117243e-05}
{"train_loss": 0.7763587832450867, "global_step": 12625, "epoch": 141, "lr": 9.950662057368484e-05}
{"train_loss": 0.7498331665992737, "global_step": 12626, "epoch": 141, "lr": 9.950653932898981e-05}
{"train_loss": 0.628263533115387, "global_step": 12627, "epoch": 141, "lr": 9.950645807763925e-05}
{"train_loss": 0.6267302632331848, "global_step": 12628, "epoch": 141, "lr": 9.950637681963313e-05}
{"train_loss": 0.8367030024528503, "global_step": 12629, "epoch": 141, "lr": 9.950629555497148e-05}
{"train_loss": 0.7078871130943298, "global_step": 12630, "epoch": 141, "lr": 9.950621428365433e-05}
{"train_loss": 0.6688228249549866, "global_step": 12631, "epoch": 141, "lr": 9.950613300568167e-05}
{"train_loss": 0.6046947240829468, "global_step": 12632, "epoch": 141, "lr": 9.95060517210535e-05}
{"train_loss": 0.667410671710968, "global_step": 12633, "epoch": 141, "lr": 9.950597042976986e-05}
{"train_loss": 0.6629236340522766, "global_step": 12634, "epoch": 141, "lr": 9.950588913183073e-05}
{"train_loss": 0.6577371954917908, "global_step": 12635, "epoch": 141, "lr": 9.950580782723615e-05}
{"train_loss": 0.6514137387275696, "global_step": 12636, "epoch": 141, "lr": 9.950572651598611e-05}
{"train_loss": 0.6442791692326578, "global_step": 12637, "epoch": 141, "lr": 9.950564519808063e-05, "val_loss": 1.3616517782211304}
{"train_loss": 0.5897413492202759, "global_step": 12638, "epoch": 142, "lr": 9.950556387351973e-05}
{"train_loss": 0.691778838634491, "global_step": 12639, "epoch": 142, "lr": 9.950548254230339e-05}
{"train_loss": 0.7159301042556763, "global_step": 12640, "epoch": 142, "lr": 9.950540120443166e-05}
{"train_loss": 0.6050564646720886, "global_step": 12641, "epoch": 142, "lr": 9.950531985990452e-05}
{"train_loss": 0.6873800754547119, "global_step": 12642, "epoch": 142, "lr": 9.9505238508722e-05}
{"train_loss": 0.6398219466209412, "global_step": 12643, "epoch": 142, "lr": 9.950515715088409e-05}
{"train_loss": 0.5705334544181824, "global_step": 12644, "epoch": 142, "lr": 9.950507578639084e-05}
{"train_loss": 0.7781898379325867, "global_step": 12645, "epoch": 142, "lr": 9.950499441524221e-05}
{"train_loss": 0.6354207396507263, "global_step": 12646, "epoch": 142, "lr": 9.950491303743826e-05}
{"train_loss": 0.569381833076477, "global_step": 12647, "epoch": 142, "lr": 9.950483165297896e-05}
{"train_loss": 0.5855075120925903, "global_step": 12648, "epoch": 142, "lr": 9.950475026186434e-05}
{"train_loss": 0.5206345915794373, "global_step": 12649, "epoch": 142, "lr": 9.950466886409442e-05}
{"train_loss": 0.723365068435669, "global_step": 12650, "epoch": 142, "lr": 9.95045874596692e-05}
{"train_loss": 0.646856427192688, "global_step": 12651, "epoch": 142, "lr": 9.950450604858867e-05}
{"train_loss": 0.6828724145889282, "global_step": 12652, "epoch": 142, "lr": 9.950442463085289e-05}
{"train_loss": 0.5290052890777588, "global_step": 12653, "epoch": 142, "lr": 9.950434320646183e-05}
{"train_loss": 0.48585328459739685, "global_step": 12654, "epoch": 142, "lr": 9.950426177541551e-05}
{"train_loss": 0.6811996102333069, "global_step": 12655, "epoch": 142, "lr": 9.950418033771396e-05}
{"train_loss": 0.5544863939285278, "global_step": 12656, "epoch": 142, "lr": 9.950409889335717e-05}
{"train_loss": 0.5467591881752014, "global_step": 12657, "epoch": 142, "lr": 9.950401744234516e-05}
{"train_loss": 0.7139455080032349, "global_step": 12658, "epoch": 142, "lr": 9.950393598467793e-05}
{"train_loss": 0.6149659156799316, "global_step": 12659, "epoch": 142, "lr": 9.950385452035552e-05}
{"train_loss": 0.6125513911247253, "global_step": 12660, "epoch": 142, "lr": 9.950377304937789e-05}
{"train_loss": 0.6195299625396729, "global_step": 12661, "epoch": 142, "lr": 9.95036915717451e-05}
{"train_loss": 0.6672603487968445, "global_step": 12662, "epoch": 142, "lr": 9.950361008745713e-05}
{"train_loss": 0.6595848202705383, "global_step": 12663, "epoch": 142, "lr": 9.950352859651401e-05}
{"train_loss": 0.4843912422657013, "global_step": 12664, "epoch": 142, "lr": 9.950344709891575e-05}
{"train_loss": 0.6874168515205383, "global_step": 12665, "epoch": 142, "lr": 9.950336559466236e-05}
{"train_loss": 0.6514317989349365, "global_step": 12666, "epoch": 142, "lr": 9.950328408375382e-05}
{"train_loss": 0.7500765919685364, "global_step": 12667, "epoch": 142, "lr": 9.95032025661902e-05}
{"train_loss": 0.8046001195907593, "global_step": 12668, "epoch": 142, "lr": 9.950312104197145e-05}
{"train_loss": 0.6161616444587708, "global_step": 12669, "epoch": 142, "lr": 9.950303951109761e-05}
{"train_loss": 0.5948436856269836, "global_step": 12670, "epoch": 142, "lr": 9.95029579735687e-05}
{"train_loss": 0.5190929174423218, "global_step": 12671, "epoch": 142, "lr": 9.950287642938471e-05}
{"train_loss": 0.6004632711410522, "global_step": 12672, "epoch": 142, "lr": 9.950279487854569e-05}
{"train_loss": 0.6396889090538025, "global_step": 12673, "epoch": 142, "lr": 9.950271332105158e-05}
{"train_loss": 0.7222310304641724, "global_step": 12674, "epoch": 142, "lr": 9.950263175690246e-05}
{"train_loss": 0.5774308443069458, "global_step": 12675, "epoch": 142, "lr": 9.950255018609832e-05}
{"train_loss": 0.6160773634910583, "global_step": 12676, "epoch": 142, "lr": 9.950246860863915e-05}
{"train_loss": 0.6446242332458496, "global_step": 12677, "epoch": 142, "lr": 9.950238702452499e-05}
{"train_loss": 0.58303302526474, "global_step": 12678, "epoch": 142, "lr": 9.950230543375583e-05}
{"train_loss": 0.6984253525733948, "global_step": 12679, "epoch": 142, "lr": 9.950222383633168e-05}
{"train_loss": 0.5806843638420105, "global_step": 12680, "epoch": 142, "lr": 9.950214223225258e-05}
{"train_loss": 0.700380265712738, "global_step": 12681, "epoch": 142, "lr": 9.95020606215185e-05}
{"train_loss": 0.5916507840156555, "global_step": 12682, "epoch": 142, "lr": 9.950197900412948e-05}
{"train_loss": 0.5516816973686218, "global_step": 12683, "epoch": 142, "lr": 9.950189738008553e-05}
{"train_loss": 0.595439076423645, "global_step": 12684, "epoch": 142, "lr": 9.950181574938664e-05}
{"train_loss": 0.632266104221344, "global_step": 12685, "epoch": 142, "lr": 9.950173411203284e-05}
{"train_loss": 0.7025235891342163, "global_step": 12686, "epoch": 142, "lr": 9.950165246802414e-05}
{"train_loss": 0.5964787006378174, "global_step": 12687, "epoch": 142, "lr": 9.950157081736054e-05}
{"train_loss": 0.5530819296836853, "global_step": 12688, "epoch": 142, "lr": 9.950148916004204e-05}
{"train_loss": 0.6493136286735535, "global_step": 12689, "epoch": 142, "lr": 9.95014074960687e-05}
{"train_loss": 0.732197105884552, "global_step": 12690, "epoch": 142, "lr": 9.950132582544049e-05}
{"train_loss": 0.6626930832862854, "global_step": 12691, "epoch": 142, "lr": 9.950124414815743e-05}
{"train_loss": 0.6642778515815735, "global_step": 12692, "epoch": 142, "lr": 9.950116246421952e-05}
{"train_loss": 0.6176385283470154, "global_step": 12693, "epoch": 142, "lr": 9.950108077362678e-05}
{"train_loss": 0.6359469890594482, "global_step": 12694, "epoch": 142, "lr": 9.950099907637924e-05}
{"train_loss": 0.6892025470733643, "global_step": 12695, "epoch": 142, "lr": 9.950091737247689e-05}
{"train_loss": 0.5936347842216492, "global_step": 12696, "epoch": 142, "lr": 9.950083566191973e-05}
{"train_loss": 0.6125098466873169, "global_step": 12697, "epoch": 142, "lr": 9.950075394470779e-05}
{"train_loss": 0.6362583041191101, "global_step": 12698, "epoch": 142, "lr": 9.950067222084109e-05}
{"train_loss": 0.7380545139312744, "global_step": 12699, "epoch": 142, "lr": 9.950059049031962e-05}
{"train_loss": 0.533839225769043, "global_step": 12700, "epoch": 142, "lr": 9.95005087531434e-05}
{"train_loss": 0.7383280396461487, "global_step": 12701, "epoch": 142, "lr": 9.950042700931244e-05}
{"train_loss": 0.6949118375778198, "global_step": 12702, "epoch": 142, "lr": 9.950034525882676e-05}
{"train_loss": 0.7116661071777344, "global_step": 12703, "epoch": 142, "lr": 9.950026350168635e-05}
{"train_loss": 0.6150843501091003, "global_step": 12704, "epoch": 142, "lr": 9.950018173789124e-05}
{"train_loss": 0.7681684494018555, "global_step": 12705, "epoch": 142, "lr": 9.950009996744144e-05}
{"train_loss": 0.6351913809776306, "global_step": 12706, "epoch": 142, "lr": 9.950001819033692e-05}
{"train_loss": 0.6622097492218018, "global_step": 12707, "epoch": 142, "lr": 9.949993640657776e-05}
{"train_loss": 0.630438506603241, "global_step": 12708, "epoch": 142, "lr": 9.949985461616392e-05}
{"train_loss": 0.6313309669494629, "global_step": 12709, "epoch": 142, "lr": 9.949977281909544e-05}
{"train_loss": 0.6231445074081421, "global_step": 12710, "epoch": 142, "lr": 9.949969101537231e-05}
{"train_loss": 0.6247469782829285, "global_step": 12711, "epoch": 142, "lr": 9.949960920499456e-05}
{"train_loss": 0.6717600226402283, "global_step": 12712, "epoch": 142, "lr": 9.949952738796218e-05}
{"train_loss": 0.5070889592170715, "global_step": 12713, "epoch": 142, "lr": 9.949944556427519e-05}
{"train_loss": 0.7088282108306885, "global_step": 12714, "epoch": 142, "lr": 9.949936373393362e-05}
{"train_loss": 0.6499577760696411, "global_step": 12715, "epoch": 142, "lr": 9.949928189693745e-05}
{"train_loss": 0.8842155933380127, "global_step": 12716, "epoch": 142, "lr": 9.94992000532867e-05}
{"train_loss": 0.5785222053527832, "global_step": 12717, "epoch": 142, "lr": 9.94991182029814e-05}
{"train_loss": 0.5938007831573486, "global_step": 12718, "epoch": 142, "lr": 9.949903634602155e-05}
{"train_loss": 0.5951123237609863, "global_step": 12719, "epoch": 142, "lr": 9.949895448240714e-05}
{"train_loss": 0.6979858875274658, "global_step": 12720, "epoch": 142, "lr": 9.949887261213822e-05}
{"train_loss": 0.886642575263977, "global_step": 12721, "epoch": 142, "lr": 9.949879073521475e-05}
{"train_loss": 0.6313840746879578, "global_step": 12722, "epoch": 142, "lr": 9.949870885163679e-05}
{"train_loss": 0.6370669007301331, "global_step": 12723, "epoch": 142, "lr": 9.949862696140433e-05}
{"train_loss": 0.640192449092865, "global_step": 12724, "epoch": 142, "lr": 9.94985450645174e-05}
{"train_loss": 0.7493148446083069, "global_step": 12725, "epoch": 142, "lr": 9.949846316097598e-05}
{"train_loss": 0.6429471540986822, "global_step": 12726, "epoch": 142, "lr": 9.949838125078008e-05, "val_loss": 1.358892560005188}
{"train_loss": 0.5349488854408264, "global_step": 12727, "epoch": 143, "lr": 9.949829933392973e-05}
{"train_loss": 0.6864455342292786, "global_step": 12728, "epoch": 143, "lr": 9.949821741042495e-05}
{"train_loss": 0.737907350063324, "global_step": 12729, "epoch": 143, "lr": 9.949813548026575e-05}
{"train_loss": 0.5753670930862427, "global_step": 12730, "epoch": 143, "lr": 9.949805354345209e-05}
{"train_loss": 0.638073205947876, "global_step": 12731, "epoch": 143, "lr": 9.949797159998405e-05}
{"train_loss": 0.5891972780227661, "global_step": 12732, "epoch": 143, "lr": 9.949788964986159e-05}
{"train_loss": 0.6654981374740601, "global_step": 12733, "epoch": 143, "lr": 9.949780769308476e-05}
{"train_loss": 0.5860669612884521, "global_step": 12734, "epoch": 143, "lr": 9.949772572965355e-05}
{"train_loss": 0.7195513844490051, "global_step": 12735, "epoch": 143, "lr": 9.949764375956798e-05}
{"train_loss": 0.654975175857544, "global_step": 12736, "epoch": 143, "lr": 9.949756178282805e-05}
{"train_loss": 0.694082498550415, "global_step": 12737, "epoch": 143, "lr": 9.949747979943377e-05}
{"train_loss": 0.6879401803016663, "global_step": 12738, "epoch": 143, "lr": 9.949739780938516e-05}
{"train_loss": 0.6180154085159302, "global_step": 12739, "epoch": 143, "lr": 9.949731581268223e-05}
{"train_loss": 0.46338820457458496, "global_step": 12740, "epoch": 143, "lr": 9.949723380932498e-05}
{"train_loss": 0.677318811416626, "global_step": 12741, "epoch": 143, "lr": 9.949715179931345e-05}
{"train_loss": 0.767669677734375, "global_step": 12742, "epoch": 143, "lr": 9.949706978264761e-05}
{"train_loss": 0.5950366258621216, "global_step": 12743, "epoch": 143, "lr": 9.94969877593275e-05}
{"train_loss": 0.6182193160057068, "global_step": 12744, "epoch": 143, "lr": 9.949690572935313e-05}
{"train_loss": 0.590600848197937, "global_step": 12745, "epoch": 143, "lr": 9.94968236927245e-05}
{"train_loss": 0.6001153588294983, "global_step": 12746, "epoch": 143, "lr": 9.949674164944163e-05}
{"train_loss": 0.7087106108665466, "global_step": 12747, "epoch": 143, "lr": 9.949665959950452e-05}
{"train_loss": 0.7219114303588867, "global_step": 12748, "epoch": 143, "lr": 9.949657754291318e-05}
{"train_loss": 0.6072721481323242, "global_step": 12749, "epoch": 143, "lr": 9.949649547966764e-05}
{"train_loss": 0.5312598347663879, "global_step": 12750, "epoch": 143, "lr": 9.94964134097679e-05}
{"train_loss": 0.44000473618507385, "global_step": 12751, "epoch": 143, "lr": 9.949633133321396e-05}
{"train_loss": 0.6316246390342712, "global_step": 12752, "epoch": 143, "lr": 9.949624925000585e-05}
{"train_loss": 0.8254662752151489, "global_step": 12753, "epoch": 143, "lr": 9.949616716014357e-05}
{"train_loss": 0.8523034453392029, "global_step": 12754, "epoch": 143, "lr": 9.949608506362713e-05}
{"train_loss": 0.5478686094284058, "global_step": 12755, "epoch": 143, "lr": 9.949600296045654e-05}
{"train_loss": 0.7104557156562805, "global_step": 12756, "epoch": 143, "lr": 9.949592085063184e-05}
{"train_loss": 0.5822162628173828, "global_step": 12757, "epoch": 143, "lr": 9.949583873415301e-05}
{"train_loss": 0.707973062992096, "global_step": 12758, "epoch": 143, "lr": 9.949575661102005e-05}
{"train_loss": 0.6784312725067139, "global_step": 12759, "epoch": 143, "lr": 9.949567448123299e-05}
{"train_loss": 0.6974411606788635, "global_step": 12760, "epoch": 143, "lr": 9.949559234479184e-05}
{"train_loss": 0.7296672463417053, "global_step": 12761, "epoch": 143, "lr": 9.949551020169663e-05}
{"train_loss": 0.7129127383232117, "global_step": 12762, "epoch": 143, "lr": 9.949542805194733e-05}
{"train_loss": 0.5633788108825684, "global_step": 12763, "epoch": 143, "lr": 9.949534589554399e-05}
{"train_loss": 0.7397693991661072, "global_step": 12764, "epoch": 143, "lr": 9.94952637324866e-05}
{"train_loss": 0.5948442816734314, "global_step": 12765, "epoch": 143, "lr": 9.949518156277517e-05}
{"train_loss": 0.7440499067306519, "global_step": 12766, "epoch": 143, "lr": 9.949509938640971e-05}
{"train_loss": 0.5775581002235413, "global_step": 12767, "epoch": 143, "lr": 9.949501720339025e-05}
{"train_loss": 0.4370371401309967, "global_step": 12768, "epoch": 143, "lr": 9.949493501371678e-05}
{"train_loss": 0.7771657109260559, "global_step": 12769, "epoch": 143, "lr": 9.949485281738933e-05}
{"train_loss": 0.6992735266685486, "global_step": 12770, "epoch": 143, "lr": 9.949477061440789e-05}
{"train_loss": 0.7087843418121338, "global_step": 12771, "epoch": 143, "lr": 9.94946884047725e-05}
{"train_loss": 0.6492199897766113, "global_step": 12772, "epoch": 143, "lr": 9.949460618848312e-05}
{"train_loss": 0.5399919748306274, "global_step": 12773, "epoch": 143, "lr": 9.949452396553983e-05}
{"train_loss": 0.6840208172798157, "global_step": 12774, "epoch": 143, "lr": 9.949444173594258e-05}
{"train_loss": 0.6276538968086243, "global_step": 12775, "epoch": 143, "lr": 9.94943594996914e-05}
{"train_loss": 0.6467334628105164, "global_step": 12776, "epoch": 143, "lr": 9.949427725678633e-05}
{"train_loss": 0.8110052943229675, "global_step": 12777, "epoch": 143, "lr": 9.949419500722735e-05}
{"train_loss": 0.5556238293647766, "global_step": 12778, "epoch": 143, "lr": 9.949411275101445e-05}
{"train_loss": 0.5968286991119385, "global_step": 12779, "epoch": 143, "lr": 9.94940304881477e-05}
{"train_loss": 0.7083306312561035, "global_step": 12780, "epoch": 143, "lr": 9.949394821862708e-05}
{"train_loss": 0.5535435676574707, "global_step": 12781, "epoch": 143, "lr": 9.94938659424526e-05}
{"train_loss": 0.6655341982841492, "global_step": 12782, "epoch": 143, "lr": 9.949378365962426e-05}
{"train_loss": 0.6899769306182861, "global_step": 12783, "epoch": 143, "lr": 9.949370137014209e-05}
{"train_loss": 0.6810460090637207, "global_step": 12784, "epoch": 143, "lr": 9.949361907400609e-05}
{"train_loss": 0.5149136781692505, "global_step": 12785, "epoch": 143, "lr": 9.949353677121629e-05}
{"train_loss": 0.630486786365509, "global_step": 12786, "epoch": 143, "lr": 9.949345446177269e-05}
{"train_loss": 0.6386667490005493, "global_step": 12787, "epoch": 143, "lr": 9.94933721456753e-05}
{"train_loss": 0.5136063694953918, "global_step": 12788, "epoch": 143, "lr": 9.94932898229241e-05}
{"train_loss": 0.6140707731246948, "global_step": 12789, "epoch": 143, "lr": 9.949320749351916e-05}
{"train_loss": 0.5794637203216553, "global_step": 12790, "epoch": 143, "lr": 9.949312515746045e-05}
{"train_loss": 0.6881669759750366, "global_step": 12791, "epoch": 143, "lr": 9.9493042814748e-05}
{"train_loss": 0.6272179484367371, "global_step": 12792, "epoch": 143, "lr": 9.94929604653818e-05}
{"train_loss": 0.6115548610687256, "global_step": 12793, "epoch": 143, "lr": 9.949287810936188e-05}
{"train_loss": 0.5976154208183289, "global_step": 12794, "epoch": 143, "lr": 9.949279574668824e-05}
{"train_loss": 0.6630954742431641, "global_step": 12795, "epoch": 143, "lr": 9.949271337736092e-05}
{"train_loss": 0.623236894607544, "global_step": 12796, "epoch": 143, "lr": 9.949263100137989e-05}
{"train_loss": 0.6547337174415588, "global_step": 12797, "epoch": 143, "lr": 9.949254861874518e-05}
{"train_loss": 0.6941572427749634, "global_step": 12798, "epoch": 143, "lr": 9.94924662294568e-05}
{"train_loss": 0.6517394781112671, "global_step": 12799, "epoch": 143, "lr": 9.949238383351475e-05}
{"train_loss": 0.673568844795227, "global_step": 12800, "epoch": 143, "lr": 9.949230143091907e-05}
{"train_loss": 0.6060370802879333, "global_step": 12801, "epoch": 143, "lr": 9.949221902166977e-05}
{"train_loss": 0.6868337988853455, "global_step": 12802, "epoch": 143, "lr": 9.949213660576682e-05}
{"train_loss": 0.7271833419799805, "global_step": 12803, "epoch": 143, "lr": 9.949205418321025e-05}
{"train_loss": 0.633560836315155, "global_step": 12804, "epoch": 143, "lr": 9.949197175400009e-05}
{"train_loss": 0.6221681833267212, "global_step": 12805, "epoch": 143, "lr": 9.949188931813633e-05}
{"train_loss": 0.6114100217819214, "global_step": 12806, "epoch": 143, "lr": 9.949180687561899e-05}
{"train_loss": 0.5991217494010925, "global_step": 12807, "epoch": 143, "lr": 9.949172442644807e-05}
{"train_loss": 0.63313227891922, "global_step": 12808, "epoch": 143, "lr": 9.949164197062362e-05}
{"train_loss": 0.6967945098876953, "global_step": 12809, "epoch": 143, "lr": 9.94915595081456e-05}
{"train_loss": 0.6744586229324341, "global_step": 12810, "epoch": 143, "lr": 9.949147703901404e-05}
{"train_loss": 0.6896023154258728, "global_step": 12811, "epoch": 143, "lr": 9.949139456322896e-05}
{"train_loss": 0.6010475158691406, "global_step": 12812, "epoch": 143, "lr": 9.949131208079037e-05}
{"train_loss": 0.5673818588256836, "global_step": 12813, "epoch": 143, "lr": 9.949122959169827e-05}
{"train_loss": 0.6507987380027771, "global_step": 12814, "epoch": 143, "lr": 9.949114709595269e-05}
{"train_loss": 0.6457174393568146, "global_step": 12815, "epoch": 143, "lr": 9.949106459355361e-05, "val_loss": 1.336251974105835}
{"train_loss": 0.676473081111908, "global_step": 12816, "epoch": 144, "lr": 9.949098208450107e-05}
{"train_loss": 0.5921695232391357, "global_step": 12817, "epoch": 144, "lr": 9.949089956879508e-05}
{"train_loss": 0.6361656188964844, "global_step": 12818, "epoch": 144, "lr": 9.949081704643562e-05}
{"train_loss": 0.5262787938117981, "global_step": 12819, "epoch": 144, "lr": 9.949073451742273e-05}
{"train_loss": 0.5565121173858643, "global_step": 12820, "epoch": 144, "lr": 9.949065198175642e-05}
{"train_loss": 0.6439196467399597, "global_step": 12821, "epoch": 144, "lr": 9.949056943943668e-05}
{"train_loss": 0.5850801467895508, "global_step": 12822, "epoch": 144, "lr": 9.949048689046355e-05}
{"train_loss": 0.6632471084594727, "global_step": 12823, "epoch": 144, "lr": 9.949040433483703e-05}
{"train_loss": 0.6083901524543762, "global_step": 12824, "epoch": 144, "lr": 9.949032177255712e-05}
{"train_loss": 0.686332643032074, "global_step": 12825, "epoch": 144, "lr": 9.949023920362385e-05}
{"train_loss": 0.6219235062599182, "global_step": 12826, "epoch": 144, "lr": 9.949015662803721e-05}
{"train_loss": 0.5234235525131226, "global_step": 12827, "epoch": 144, "lr": 9.949007404579722e-05}
{"train_loss": 0.560924768447876, "global_step": 12828, "epoch": 144, "lr": 9.94899914569039e-05}
{"train_loss": 0.7629714608192444, "global_step": 12829, "epoch": 144, "lr": 9.948990886135725e-05}
{"train_loss": 0.5802326798439026, "global_step": 12830, "epoch": 144, "lr": 9.94898262591573e-05}
{"train_loss": 0.5877551436424255, "global_step": 12831, "epoch": 144, "lr": 9.948974365030404e-05}
{"train_loss": 0.6849954128265381, "global_step": 12832, "epoch": 144, "lr": 9.948966103479747e-05}
{"train_loss": 0.8239476680755615, "global_step": 12833, "epoch": 144, "lr": 9.948957841263762e-05}
{"train_loss": 0.6923670172691345, "global_step": 12834, "epoch": 144, "lr": 9.948949578382451e-05}
{"train_loss": 0.546208381652832, "global_step": 12835, "epoch": 144, "lr": 9.948941314835813e-05}
{"train_loss": 0.5879140496253967, "global_step": 12836, "epoch": 144, "lr": 9.948933050623852e-05}
{"train_loss": 0.6043299436569214, "global_step": 12837, "epoch": 144, "lr": 9.948924785746567e-05}
{"train_loss": 0.5997029542922974, "global_step": 12838, "epoch": 144, "lr": 9.948916520203958e-05}
{"train_loss": 0.6185077428817749, "global_step": 12839, "epoch": 144, "lr": 9.948908253996028e-05}
{"train_loss": 0.6174246072769165, "global_step": 12840, "epoch": 144, "lr": 9.948899987122777e-05}
{"train_loss": 0.5113171339035034, "global_step": 12841, "epoch": 144, "lr": 9.948891719584208e-05}
{"train_loss": 0.5904762744903564, "global_step": 12842, "epoch": 144, "lr": 9.948883451380321e-05}
{"train_loss": 0.6771553754806519, "global_step": 12843, "epoch": 144, "lr": 9.948875182511115e-05}
{"train_loss": 0.5792032480239868, "global_step": 12844, "epoch": 144, "lr": 9.948866912976595e-05}
{"train_loss": 0.6886342763900757, "global_step": 12845, "epoch": 144, "lr": 9.94885864277676e-05}
{"train_loss": 0.5900266170501709, "global_step": 12846, "epoch": 144, "lr": 9.948850371911611e-05}
{"train_loss": 0.6026328802108765, "global_step": 12847, "epoch": 144, "lr": 9.948842100381148e-05}
{"train_loss": 0.4802376627922058, "global_step": 12848, "epoch": 144, "lr": 9.948833828185375e-05}
{"train_loss": 0.6191185712814331, "global_step": 12849, "epoch": 144, "lr": 9.948825555324291e-05}
{"train_loss": 0.6401012539863586, "global_step": 12850, "epoch": 144, "lr": 9.948817281797897e-05}
{"train_loss": 0.5898599028587341, "global_step": 12851, "epoch": 144, "lr": 9.948809007606197e-05}
{"train_loss": 0.5504857301712036, "global_step": 12852, "epoch": 144, "lr": 9.948800732749188e-05}
{"train_loss": 0.6551238298416138, "global_step": 12853, "epoch": 144, "lr": 9.948792457226874e-05}
{"train_loss": 0.5464611649513245, "global_step": 12854, "epoch": 144, "lr": 9.948784181039254e-05}
{"train_loss": 0.6334837079048157, "global_step": 12855, "epoch": 144, "lr": 9.948775904186333e-05}
{"train_loss": 0.6191224455833435, "global_step": 12856, "epoch": 144, "lr": 9.948767626668107e-05}
{"train_loss": 0.7689604759216309, "global_step": 12857, "epoch": 144, "lr": 9.94875934848458e-05}
{"train_loss": 0.6157079935073853, "global_step": 12858, "epoch": 144, "lr": 9.948751069635754e-05}
{"train_loss": 0.5702077150344849, "global_step": 12859, "epoch": 144, "lr": 9.948742790121627e-05}
{"train_loss": 0.588683009147644, "global_step": 12860, "epoch": 144, "lr": 9.948734509942201e-05}
{"train_loss": 0.6581928730010986, "global_step": 12861, "epoch": 144, "lr": 9.94872622909748e-05}
{"train_loss": 0.6004683375358582, "global_step": 12862, "epoch": 144, "lr": 9.948717947587462e-05}
{"train_loss": 0.6927036643028259, "global_step": 12863, "epoch": 144, "lr": 9.94870966541215e-05}
{"train_loss": 0.8398756384849548, "global_step": 12864, "epoch": 144, "lr": 9.948701382571544e-05}
{"train_loss": 0.5041907429695129, "global_step": 12865, "epoch": 144, "lr": 9.948693099065645e-05}
{"train_loss": 0.619810938835144, "global_step": 12866, "epoch": 144, "lr": 9.948684814894455e-05}
{"train_loss": 0.5726231336593628, "global_step": 12867, "epoch": 144, "lr": 9.948676530057975e-05}
{"train_loss": 0.7522897720336914, "global_step": 12868, "epoch": 144, "lr": 9.948668244556205e-05}
{"train_loss": 0.6135801076889038, "global_step": 12869, "epoch": 144, "lr": 9.948659958389146e-05}
{"train_loss": 0.6963039040565491, "global_step": 12870, "epoch": 144, "lr": 9.948651671556801e-05}
{"train_loss": 0.5567148327827454, "global_step": 12871, "epoch": 144, "lr": 9.948643384059171e-05}
{"train_loss": 0.6587860584259033, "global_step": 12872, "epoch": 144, "lr": 9.948635095896255e-05}
{"train_loss": 0.5898150205612183, "global_step": 12873, "epoch": 144, "lr": 9.948626807068055e-05}
{"train_loss": 0.5772050619125366, "global_step": 12874, "epoch": 144, "lr": 9.948618517574575e-05}
{"train_loss": 0.7698293924331665, "global_step": 12875, "epoch": 144, "lr": 9.948610227415811e-05}
{"train_loss": 0.5285756587982178, "global_step": 12876, "epoch": 144, "lr": 9.948601936591767e-05}
{"train_loss": 0.667557418346405, "global_step": 12877, "epoch": 144, "lr": 9.948593645102445e-05}
{"train_loss": 0.6836113929748535, "global_step": 12878, "epoch": 144, "lr": 9.948585352947843e-05}
{"train_loss": 0.585935115814209, "global_step": 12879, "epoch": 144, "lr": 9.948577060127966e-05}
{"train_loss": 0.5927794575691223, "global_step": 12880, "epoch": 144, "lr": 9.948568766642811e-05}
{"train_loss": 0.8091728687286377, "global_step": 12881, "epoch": 144, "lr": 9.948560472492384e-05}
{"train_loss": 0.4494788944721222, "global_step": 12882, "epoch": 144, "lr": 9.948552177676681e-05}
{"train_loss": 0.6646167039871216, "global_step": 12883, "epoch": 144, "lr": 9.948543882195707e-05}
{"train_loss": 0.6239836812019348, "global_step": 12884, "epoch": 144, "lr": 9.94853558604946e-05}
{"train_loss": 0.6858746409416199, "global_step": 12885, "epoch": 144, "lr": 9.948527289237943e-05}
{"train_loss": 0.6118399500846863, "global_step": 12886, "epoch": 144, "lr": 9.948518991761157e-05}
{"train_loss": 0.5692200660705566, "global_step": 12887, "epoch": 144, "lr": 9.948510693619104e-05}
{"train_loss": 0.5789821147918701, "global_step": 12888, "epoch": 144, "lr": 9.948502394811782e-05}
{"train_loss": 0.6344483494758606, "global_step": 12889, "epoch": 144, "lr": 9.948494095339197e-05}
{"train_loss": 0.621101438999176, "global_step": 12890, "epoch": 144, "lr": 9.948485795201345e-05}
{"train_loss": 0.7363383769989014, "global_step": 12891, "epoch": 144, "lr": 9.948477494398231e-05}
{"train_loss": 0.6217063665390015, "global_step": 12892, "epoch": 144, "lr": 9.948469192929855e-05}
{"train_loss": 0.5908678770065308, "global_step": 12893, "epoch": 144, "lr": 9.948460890796215e-05}
{"train_loss": 0.7457140684127808, "global_step": 12894, "epoch": 144, "lr": 9.948452587997317e-05}
{"train_loss": 0.6683446764945984, "global_step": 12895, "epoch": 144, "lr": 9.948444284533158e-05}
{"train_loss": 0.648017942905426, "global_step": 12896, "epoch": 144, "lr": 9.948435980403743e-05}
{"train_loss": 0.6414914727210999, "global_step": 12897, "epoch": 144, "lr": 9.948427675609069e-05}
{"train_loss": 0.6172192096710205, "global_step": 12898, "epoch": 144, "lr": 9.94841937014914e-05}
{"train_loss": 0.7157933115959167, "global_step": 12899, "epoch": 144, "lr": 9.948411064023958e-05}
{"train_loss": 0.6221583485603333, "global_step": 12900, "epoch": 144, "lr": 9.94840275723352e-05}
{"train_loss": 0.7604883909225464, "global_step": 12901, "epoch": 144, "lr": 9.948394449777831e-05}
{"train_loss": 0.6620118021965027, "global_step": 12902, "epoch": 144, "lr": 9.94838614165689e-05}
{"train_loss": 0.6398022174835205, "global_step": 12903, "epoch": 144, "lr": 9.948377832870698e-05}
{"train_loss": 0.6326157400447331, "global_step": 12904, "epoch": 144, "lr": 9.948369523419259e-05, "val_loss": 1.3452270030975342}
{"train_loss": 0.5319994688034058, "global_step": 12905, "epoch": 145, "lr": 9.94836121330257e-05}
{"train_loss": 0.6451418399810791, "global_step": 12906, "epoch": 145, "lr": 9.948352902520637e-05}
{"train_loss": 0.6679040789604187, "global_step": 12907, "epoch": 145, "lr": 9.948344591073455e-05}
{"train_loss": 0.7059164047241211, "global_step": 12908, "epoch": 145, "lr": 9.94833627896103e-05}
{"train_loss": 0.8629361391067505, "global_step": 12909, "epoch": 145, "lr": 9.948327966183362e-05}
{"train_loss": 0.7574516534805298, "global_step": 12910, "epoch": 145, "lr": 9.948319652740451e-05}
{"train_loss": 0.6794174909591675, "global_step": 12911, "epoch": 145, "lr": 9.948311338632299e-05}
{"train_loss": 0.6135783195495605, "global_step": 12912, "epoch": 145, "lr": 9.948303023858905e-05}
{"train_loss": 0.5952582359313965, "global_step": 12913, "epoch": 145, "lr": 9.948294708420274e-05}
{"train_loss": 0.6995181441307068, "global_step": 12914, "epoch": 145, "lr": 9.948286392316405e-05}
{"train_loss": 0.6957255005836487, "global_step": 12915, "epoch": 145, "lr": 9.948278075547299e-05}
{"train_loss": 0.6922041177749634, "global_step": 12916, "epoch": 145, "lr": 9.948269758112957e-05}
{"train_loss": 0.8627713918685913, "global_step": 12917, "epoch": 145, "lr": 9.94826144001338e-05}
{"train_loss": 0.5794641375541687, "global_step": 12918, "epoch": 145, "lr": 9.94825312124857e-05}
{"train_loss": 0.572870671749115, "global_step": 12919, "epoch": 145, "lr": 9.948244801818528e-05}
{"train_loss": 0.6235318779945374, "global_step": 12920, "epoch": 145, "lr": 9.948236481723254e-05}
{"train_loss": 0.4636107087135315, "global_step": 12921, "epoch": 145, "lr": 9.948228160962751e-05}
{"train_loss": 0.4935421347618103, "global_step": 12922, "epoch": 145, "lr": 9.948219839537018e-05}
{"train_loss": 0.5714438557624817, "global_step": 12923, "epoch": 145, "lr": 9.948211517446058e-05}
{"train_loss": 0.5903449654579163, "global_step": 12924, "epoch": 145, "lr": 9.94820319468987e-05}
{"train_loss": 0.6192126274108887, "global_step": 12925, "epoch": 145, "lr": 9.948194871268459e-05}
{"train_loss": 0.6321715116500854, "global_step": 12926, "epoch": 145, "lr": 9.948186547181821e-05}
{"train_loss": 0.4597977101802826, "global_step": 12927, "epoch": 145, "lr": 9.948178222429961e-05}
{"train_loss": 0.7133461236953735, "global_step": 12928, "epoch": 145, "lr": 9.948169897012878e-05}
{"train_loss": 0.5681877732276917, "global_step": 12929, "epoch": 145, "lr": 9.948161570930574e-05}
{"train_loss": 0.5274984836578369, "global_step": 12930, "epoch": 145, "lr": 9.94815324418305e-05}
{"train_loss": 0.5956817865371704, "global_step": 12931, "epoch": 145, "lr": 9.948144916770308e-05}
{"train_loss": 0.6054186820983887, "global_step": 12932, "epoch": 145, "lr": 9.948136588692347e-05}
{"train_loss": 0.7107371687889099, "global_step": 12933, "epoch": 145, "lr": 9.94812825994917e-05}
{"train_loss": 0.6509471535682678, "global_step": 12934, "epoch": 145, "lr": 9.948119930540777e-05}
{"train_loss": 0.6529589891433716, "global_step": 12935, "epoch": 145, "lr": 9.94811160046717e-05}
{"train_loss": 0.618135392665863, "global_step": 12936, "epoch": 145, "lr": 9.94810326972835e-05}
{"train_loss": 0.737253725528717, "global_step": 12937, "epoch": 145, "lr": 9.948094938324317e-05}
{"train_loss": 0.7248401045799255, "global_step": 12938, "epoch": 145, "lr": 9.948086606255073e-05}
{"train_loss": 0.5853322744369507, "global_step": 12939, "epoch": 145, "lr": 9.948078273520619e-05}
{"train_loss": 0.661704957485199, "global_step": 12940, "epoch": 145, "lr": 9.948069940120957e-05}
{"train_loss": 0.6831752061843872, "global_step": 12941, "epoch": 145, "lr": 9.948061606056087e-05}
{"train_loss": 0.6253482103347778, "global_step": 12942, "epoch": 145, "lr": 9.94805327132601e-05}
{"train_loss": 0.5531901717185974, "global_step": 12943, "epoch": 145, "lr": 9.948044935930727e-05}
{"train_loss": 0.4125218391418457, "global_step": 12944, "epoch": 145, "lr": 9.94803659987024e-05}
{"train_loss": 0.4471350312232971, "global_step": 12945, "epoch": 145, "lr": 9.948028263144551e-05}
{"train_loss": 0.5400400161743164, "global_step": 12946, "epoch": 145, "lr": 9.94801992575366e-05}
{"train_loss": 0.6028112173080444, "global_step": 12947, "epoch": 145, "lr": 9.948011587697566e-05}
{"train_loss": 0.666767418384552, "global_step": 12948, "epoch": 145, "lr": 9.948003248976274e-05}
{"train_loss": 0.6270608901977539, "global_step": 12949, "epoch": 145, "lr": 9.947994909589782e-05}
{"train_loss": 0.6472585201263428, "global_step": 12950, "epoch": 145, "lr": 9.947986569538094e-05}
{"train_loss": 0.6864035725593567, "global_step": 12951, "epoch": 145, "lr": 9.947978228821209e-05}
{"train_loss": 0.571249783039093, "global_step": 12952, "epoch": 145, "lr": 9.947969887439127e-05}
{"train_loss": 0.8628298044204712, "global_step": 12953, "epoch": 145, "lr": 9.947961545391853e-05}
{"train_loss": 0.6009184718132019, "global_step": 12954, "epoch": 145, "lr": 9.947953202679384e-05}
{"train_loss": 0.6424084305763245, "global_step": 12955, "epoch": 145, "lr": 9.947944859301724e-05}
{"train_loss": 0.8061471581459045, "global_step": 12956, "epoch": 145, "lr": 9.947936515258874e-05}
{"train_loss": 0.5895606875419617, "global_step": 12957, "epoch": 145, "lr": 9.947928170550831e-05}
{"train_loss": 0.6409890651702881, "global_step": 12958, "epoch": 145, "lr": 9.947919825177603e-05}
{"train_loss": 0.6530881524085999, "global_step": 12959, "epoch": 145, "lr": 9.947911479139185e-05}
{"train_loss": 0.560026228427887, "global_step": 12960, "epoch": 145, "lr": 9.947903132435582e-05}
{"train_loss": 0.5952744483947754, "global_step": 12961, "epoch": 145, "lr": 9.947894785066793e-05}
{"train_loss": 0.593607485294342, "global_step": 12962, "epoch": 145, "lr": 9.94788643703282e-05}
{"train_loss": 0.7925727963447571, "global_step": 12963, "epoch": 145, "lr": 9.947878088333664e-05}
{"train_loss": 0.7294470071792603, "global_step": 12964, "epoch": 145, "lr": 9.947869738969326e-05}
{"train_loss": 0.5645830631256104, "global_step": 12965, "epoch": 145, "lr": 9.947861388939808e-05}
{"train_loss": 0.7038406729698181, "global_step": 12966, "epoch": 145, "lr": 9.947853038245109e-05}
{"train_loss": 0.6781401038169861, "global_step": 12967, "epoch": 145, "lr": 9.947844686885233e-05}
{"train_loss": 0.6480726599693298, "global_step": 12968, "epoch": 145, "lr": 9.947836334860179e-05}
{"train_loss": 0.7479643821716309, "global_step": 12969, "epoch": 145, "lr": 9.947827982169949e-05}
{"train_loss": 0.6138043403625488, "global_step": 12970, "epoch": 145, "lr": 9.947819628814543e-05}
{"train_loss": 0.5320587754249573, "global_step": 12971, "epoch": 145, "lr": 9.947811274793963e-05}
{"train_loss": 0.6463592052459717, "global_step": 12972, "epoch": 145, "lr": 9.947802920108212e-05}
{"train_loss": 0.7102035880088806, "global_step": 12973, "epoch": 145, "lr": 9.947794564757287e-05}
{"train_loss": 0.7961710691452026, "global_step": 12974, "epoch": 145, "lr": 9.947786208741192e-05}
{"train_loss": 0.6017090678215027, "global_step": 12975, "epoch": 145, "lr": 9.947777852059927e-05}
{"train_loss": 0.7405199408531189, "global_step": 12976, "epoch": 145, "lr": 9.947769494713495e-05}
{"train_loss": 0.8125032186508179, "global_step": 12977, "epoch": 145, "lr": 9.947761136701894e-05}
{"train_loss": 0.6979333758354187, "global_step": 12978, "epoch": 145, "lr": 9.947752778025128e-05}
{"train_loss": 0.7892232537269592, "global_step": 12979, "epoch": 145, "lr": 9.947744418683195e-05}
{"train_loss": 0.6183167099952698, "global_step": 12980, "epoch": 145, "lr": 9.947736058676101e-05}
{"train_loss": 0.6746008992195129, "global_step": 12981, "epoch": 145, "lr": 9.947727698003842e-05}
{"train_loss": 0.6974073052406311, "global_step": 12982, "epoch": 145, "lr": 9.947719336666422e-05}
{"train_loss": 0.5486109256744385, "global_step": 12983, "epoch": 145, "lr": 9.947710974663842e-05}
{"train_loss": 0.575896143913269, "global_step": 12984, "epoch": 145, "lr": 9.947702611996101e-05}
{"train_loss": 0.5126115679740906, "global_step": 12985, "epoch": 145, "lr": 9.947694248663203e-05}
{"train_loss": 0.607309877872467, "global_step": 12986, "epoch": 145, "lr": 9.947685884665148e-05}
{"train_loss": 0.6033935546875, "global_step": 12987, "epoch": 145, "lr": 9.947677520001936e-05}
{"train_loss": 0.5729854702949524, "global_step": 12988, "epoch": 145, "lr": 9.94766915467357e-05}
{"train_loss": 0.5676178932189941, "global_step": 12989, "epoch": 145, "lr": 9.947660788680049e-05}
{"train_loss": 0.5508449077606201, "global_step": 12990, "epoch": 145, "lr": 9.947652422021377e-05}
{"train_loss": 0.6585198044776917, "global_step": 12991, "epoch": 145, "lr": 9.947644054697551e-05}
{"train_loss": 0.5450633764266968, "global_step": 12992, "epoch": 145, "lr": 9.947635686708576e-05}
{"train_loss": 0.6387178432405665, "global_step": 12993, "epoch": 145, "lr": 9.947627318054453e-05, "val_loss": 1.3329925537109375, "train_action_mse_error": 49.47108459472656}
{"train_loss": 0.7219056487083435, "global_step": 12994, "epoch": 146, "lr": 9.94761894873518e-05}
{"train_loss": 0.5562770962715149, "global_step": 12995, "epoch": 146, "lr": 9.947610578750761e-05}
{"train_loss": 0.5652269721031189, "global_step": 12996, "epoch": 146, "lr": 9.947602208101195e-05}
{"train_loss": 0.6144227981567383, "global_step": 12997, "epoch": 146, "lr": 9.947593836786485e-05}
{"train_loss": 0.5388457775115967, "global_step": 12998, "epoch": 146, "lr": 9.94758546480663e-05}
{"train_loss": 0.5318213105201721, "global_step": 12999, "epoch": 146, "lr": 9.947577092161634e-05}
{"train_loss": 0.7170907258987427, "global_step": 13000, "epoch": 146, "lr": 9.947568718851496e-05}
{"train_loss": 0.4724047780036926, "global_step": 13001, "epoch": 146, "lr": 9.947560344876218e-05}
{"train_loss": 0.5539278984069824, "global_step": 13002, "epoch": 146, "lr": 9.9475519702358e-05}
{"train_loss": 0.723147451877594, "global_step": 13003, "epoch": 146, "lr": 9.947543594930245e-05}
{"train_loss": 0.5483359694480896, "global_step": 13004, "epoch": 146, "lr": 9.947535218959553e-05}
{"train_loss": 0.7349134683609009, "global_step": 13005, "epoch": 146, "lr": 9.947526842323724e-05}
{"train_loss": 0.5633434653282166, "global_step": 13006, "epoch": 146, "lr": 9.947518465022761e-05}
{"train_loss": 0.6920839548110962, "global_step": 13007, "epoch": 146, "lr": 9.947510087056665e-05}
{"train_loss": 0.5565674304962158, "global_step": 13008, "epoch": 146, "lr": 9.947501708425435e-05}
{"train_loss": 0.5767024755477905, "global_step": 13009, "epoch": 146, "lr": 9.947493329129076e-05}
{"train_loss": 0.7713331580162048, "global_step": 13010, "epoch": 146, "lr": 9.947484949167585e-05}
{"train_loss": 0.6173815727233887, "global_step": 13011, "epoch": 146, "lr": 9.947476568540965e-05}
{"train_loss": 0.6867012977600098, "global_step": 13012, "epoch": 146, "lr": 9.947468187249218e-05}
{"train_loss": 0.6646680235862732, "global_step": 13013, "epoch": 146, "lr": 9.947459805292344e-05}
{"train_loss": 0.5464097857475281, "global_step": 13014, "epoch": 146, "lr": 9.947451422670344e-05}
{"train_loss": 0.7131130695343018, "global_step": 13015, "epoch": 146, "lr": 9.947443039383219e-05}
{"train_loss": 0.7464014291763306, "global_step": 13016, "epoch": 146, "lr": 9.947434655430972e-05}
{"train_loss": 0.5545447468757629, "global_step": 13017, "epoch": 146, "lr": 9.947426270813602e-05}
{"train_loss": 0.5759364366531372, "global_step": 13018, "epoch": 146, "lr": 9.947417885531111e-05}
{"train_loss": 0.5383105874061584, "global_step": 13019, "epoch": 146, "lr": 9.9474094995835e-05}
{"train_loss": 0.5584015250205994, "global_step": 13020, "epoch": 146, "lr": 9.947401112970769e-05}
{"train_loss": 0.7513843178749084, "global_step": 13021, "epoch": 146, "lr": 9.947392725692921e-05}
{"train_loss": 0.5875943899154663, "global_step": 13022, "epoch": 146, "lr": 9.947384337749957e-05}
{"train_loss": 0.7045848369598389, "global_step": 13023, "epoch": 146, "lr": 9.947375949141876e-05}
{"train_loss": 0.6055442690849304, "global_step": 13024, "epoch": 146, "lr": 9.94736755986868e-05}
{"train_loss": 0.6311829090118408, "global_step": 13025, "epoch": 146, "lr": 9.947359169930373e-05}
{"train_loss": 0.6090036630630493, "global_step": 13026, "epoch": 146, "lr": 9.947350779326954e-05}
{"train_loss": 0.732876718044281, "global_step": 13027, "epoch": 146, "lr": 9.947342388058423e-05}
{"train_loss": 0.7225828170776367, "global_step": 13028, "epoch": 146, "lr": 9.947333996124781e-05}
{"train_loss": 0.7240601181983948, "global_step": 13029, "epoch": 146, "lr": 9.947325603526032e-05}
{"train_loss": 0.7429832220077515, "global_step": 13030, "epoch": 146, "lr": 9.947317210262175e-05}
{"train_loss": 0.5749009847640991, "global_step": 13031, "epoch": 146, "lr": 9.947308816333212e-05}
{"train_loss": 0.6305185556411743, "global_step": 13032, "epoch": 146, "lr": 9.947300421739142e-05}
{"train_loss": 0.5890486836433411, "global_step": 13033, "epoch": 146, "lr": 9.947292026479968e-05}
{"train_loss": 0.5784953236579895, "global_step": 13034, "epoch": 146, "lr": 9.947283630555692e-05}
{"train_loss": 0.6290817260742188, "global_step": 13035, "epoch": 146, "lr": 9.947275233966313e-05}
{"train_loss": 0.7117701768875122, "global_step": 13036, "epoch": 146, "lr": 9.947266836711835e-05}
{"train_loss": 0.6106934547424316, "global_step": 13037, "epoch": 146, "lr": 9.947258438792255e-05}
{"train_loss": 0.7081447243690491, "global_step": 13038, "epoch": 146, "lr": 9.947250040207579e-05}
{"train_loss": 0.6153451204299927, "global_step": 13039, "epoch": 146, "lr": 9.947241640957803e-05}
{"train_loss": 0.765583872795105, "global_step": 13040, "epoch": 146, "lr": 9.947233241042931e-05}
{"train_loss": 0.5499860644340515, "global_step": 13041, "epoch": 146, "lr": 9.947224840462966e-05}
{"train_loss": 0.6372724175453186, "global_step": 13042, "epoch": 146, "lr": 9.947216439217904e-05}
{"train_loss": 0.668474555015564, "global_step": 13043, "epoch": 146, "lr": 9.947208037307751e-05}
{"train_loss": 0.5636364817619324, "global_step": 13044, "epoch": 146, "lr": 9.947199634732505e-05}
{"train_loss": 0.6546167135238647, "global_step": 13045, "epoch": 146, "lr": 9.947191231492168e-05}
{"train_loss": 0.6353974342346191, "global_step": 13046, "epoch": 146, "lr": 9.947182827586743e-05}
{"train_loss": 0.5834136605262756, "global_step": 13047, "epoch": 146, "lr": 9.947174423016228e-05}
{"train_loss": 0.6764233708381653, "global_step": 13048, "epoch": 146, "lr": 9.947166017780627e-05}
{"train_loss": 0.6901851296424866, "global_step": 13049, "epoch": 146, "lr": 9.947157611879939e-05}
{"train_loss": 0.6634584069252014, "global_step": 13050, "epoch": 146, "lr": 9.947149205314165e-05}
{"train_loss": 0.748495876789093, "global_step": 13051, "epoch": 146, "lr": 9.94714079808331e-05}
{"train_loss": 0.6684442162513733, "global_step": 13052, "epoch": 146, "lr": 9.94713239018737e-05}
{"train_loss": 0.5805294513702393, "global_step": 13053, "epoch": 146, "lr": 9.94712398162635e-05}
{"train_loss": 0.516216516494751, "global_step": 13054, "epoch": 146, "lr": 9.947115572400247e-05}
{"train_loss": 0.7077777981758118, "global_step": 13055, "epoch": 146, "lr": 9.947107162509066e-05}
{"train_loss": 0.65266352891922, "global_step": 13056, "epoch": 146, "lr": 9.947098751952809e-05}
{"train_loss": 0.5688406229019165, "global_step": 13057, "epoch": 146, "lr": 9.94709034073147e-05}
{"train_loss": 0.7486572265625, "global_step": 13058, "epoch": 146, "lr": 9.94708192884506e-05}
{"train_loss": 0.5826408863067627, "global_step": 13059, "epoch": 146, "lr": 9.947073516293572e-05}
{"train_loss": 0.6366381049156189, "global_step": 13060, "epoch": 146, "lr": 9.94706510307701e-05}
{"train_loss": 0.637461245059967, "global_step": 13061, "epoch": 146, "lr": 9.947056689195377e-05}
{"train_loss": 0.8018258213996887, "global_step": 13062, "epoch": 146, "lr": 9.947048274648671e-05}
{"train_loss": 0.6326237320899963, "global_step": 13063, "epoch": 146, "lr": 9.947039859436896e-05}
{"train_loss": 0.5718181729316711, "global_step": 13064, "epoch": 146, "lr": 9.94703144356005e-05}
{"train_loss": 0.5742523074150085, "global_step": 13065, "epoch": 146, "lr": 9.947023027018139e-05}
{"train_loss": 0.5974891185760498, "global_step": 13066, "epoch": 146, "lr": 9.947014609811158e-05}
{"train_loss": 0.6558814644813538, "global_step": 13067, "epoch": 146, "lr": 9.947006191939113e-05}
{"train_loss": 0.6745507717132568, "global_step": 13068, "epoch": 146, "lr": 9.946997773402003e-05}
{"train_loss": 0.544856071472168, "global_step": 13069, "epoch": 146, "lr": 9.946989354199829e-05}
{"train_loss": 0.5558390617370605, "global_step": 13070, "epoch": 146, "lr": 9.946980934332594e-05}
{"train_loss": 0.6418986320495605, "global_step": 13071, "epoch": 146, "lr": 9.946972513800297e-05}
{"train_loss": 0.6782547235488892, "global_step": 13072, "epoch": 146, "lr": 9.946964092602938e-05}
{"train_loss": 0.7322146892547607, "global_step": 13073, "epoch": 146, "lr": 9.946955670740521e-05}
{"train_loss": 0.7114693522453308, "global_step": 13074, "epoch": 146, "lr": 9.946947248213048e-05}
{"train_loss": 0.5955392718315125, "global_step": 13075, "epoch": 146, "lr": 9.946938825020516e-05}
{"train_loss": 0.6791165471076965, "global_step": 13076, "epoch": 146, "lr": 9.946930401162929e-05}
{"train_loss": 0.6706863045692444, "global_step": 13077, "epoch": 146, "lr": 9.946921976640288e-05}
{"train_loss": 0.5689401626586914, "global_step": 13078, "epoch": 146, "lr": 9.946913551452593e-05}
{"train_loss": 0.5628397464752197, "global_step": 13079, "epoch": 146, "lr": 9.946905125599846e-05}
{"train_loss": 0.6210423111915588, "global_step": 13080, "epoch": 146, "lr": 9.946896699082048e-05}
{"train_loss": 0.7151392102241516, "global_step": 13081, "epoch": 146, "lr": 9.9468882718992e-05}
{"train_loss": 0.6371478049942617, "global_step": 13082, "epoch": 146, "lr": 9.946879844051302e-05, "val_loss": 1.366942286491394}
{"train_loss": 0.7779008746147156, "global_step": 13083, "epoch": 147, "lr": 9.946871415538358e-05}
{"train_loss": 0.6116295456886292, "global_step": 13084, "epoch": 147, "lr": 9.946862986360367e-05}
{"train_loss": 0.47643882036209106, "global_step": 13085, "epoch": 147, "lr": 9.94685455651733e-05}
{"train_loss": 0.585589587688446, "global_step": 13086, "epoch": 147, "lr": 9.946846126009249e-05}
{"train_loss": 0.5656241178512573, "global_step": 13087, "epoch": 147, "lr": 9.946837694836125e-05}
{"train_loss": 0.6232112050056458, "global_step": 13088, "epoch": 147, "lr": 9.946829262997959e-05}
{"train_loss": 0.641427218914032, "global_step": 13089, "epoch": 147, "lr": 9.946820830494751e-05}
{"train_loss": 0.5274971127510071, "global_step": 13090, "epoch": 147, "lr": 9.946812397326506e-05}
{"train_loss": 0.567027747631073, "global_step": 13091, "epoch": 147, "lr": 9.94680396349322e-05}
{"train_loss": 0.5332188010215759, "global_step": 13092, "epoch": 147, "lr": 9.946795528994897e-05}
{"train_loss": 0.5985777378082275, "global_step": 13093, "epoch": 147, "lr": 9.946787093831538e-05}
{"train_loss": 0.5655373930931091, "global_step": 13094, "epoch": 147, "lr": 9.946778658003143e-05}
{"train_loss": 0.5517198443412781, "global_step": 13095, "epoch": 147, "lr": 9.946770221509713e-05}
{"train_loss": 0.670049250125885, "global_step": 13096, "epoch": 147, "lr": 9.946761784351253e-05}
{"train_loss": 0.5257116556167603, "global_step": 13097, "epoch": 147, "lr": 9.94675334652776e-05}
{"train_loss": 0.6375052332878113, "global_step": 13098, "epoch": 147, "lr": 9.946744908039235e-05}
{"train_loss": 0.5787070989608765, "global_step": 13099, "epoch": 147, "lr": 9.946736468885682e-05}
{"train_loss": 0.603784441947937, "global_step": 13100, "epoch": 147, "lr": 9.946728029067099e-05}
{"train_loss": 0.5266756415367126, "global_step": 13101, "epoch": 147, "lr": 9.946719588583491e-05}
{"train_loss": 0.6861090064048767, "global_step": 13102, "epoch": 147, "lr": 9.946711147434855e-05}
{"train_loss": 0.4603813588619232, "global_step": 13103, "epoch": 147, "lr": 9.946702705621194e-05}
{"train_loss": 0.5578195452690125, "global_step": 13104, "epoch": 147, "lr": 9.946694263142509e-05}
{"train_loss": 0.5837342143058777, "global_step": 13105, "epoch": 147, "lr": 9.946685819998802e-05}
{"train_loss": 0.5511061549186707, "global_step": 13106, "epoch": 147, "lr": 9.946677376190073e-05}
{"train_loss": 0.566538393497467, "global_step": 13107, "epoch": 147, "lr": 9.946668931716324e-05}
{"train_loss": 0.7163053750991821, "global_step": 13108, "epoch": 147, "lr": 9.946660486577555e-05}
{"train_loss": 0.6609064340591431, "global_step": 13109, "epoch": 147, "lr": 9.946652040773768e-05}
{"train_loss": 0.6210976839065552, "global_step": 13110, "epoch": 147, "lr": 9.946643594304966e-05}
{"train_loss": 0.6172164082527161, "global_step": 13111, "epoch": 147, "lr": 9.946635147171145e-05}
{"train_loss": 0.5396847128868103, "global_step": 13112, "epoch": 147, "lr": 9.94662669937231e-05}
{"train_loss": 0.5368600487709045, "global_step": 13113, "epoch": 147, "lr": 9.946618250908461e-05}
{"train_loss": 0.5680360198020935, "global_step": 13114, "epoch": 147, "lr": 9.9466098017796e-05}
{"train_loss": 0.5250663161277771, "global_step": 13115, "epoch": 147, "lr": 9.94660135198573e-05}
{"train_loss": 0.5796948671340942, "global_step": 13116, "epoch": 147, "lr": 9.946592901526845e-05}
{"train_loss": 0.7372811436653137, "global_step": 13117, "epoch": 147, "lr": 9.946584450402954e-05}
{"train_loss": 0.5033460855484009, "global_step": 13118, "epoch": 147, "lr": 9.946575998614054e-05}
{"train_loss": 0.5262705087661743, "global_step": 13119, "epoch": 147, "lr": 9.946567546160148e-05}
{"train_loss": 0.6670421361923218, "global_step": 13120, "epoch": 147, "lr": 9.946559093041236e-05}
{"train_loss": 0.6583635210990906, "global_step": 13121, "epoch": 147, "lr": 9.946550639257319e-05}
{"train_loss": 0.4080900251865387, "global_step": 13122, "epoch": 147, "lr": 9.946542184808398e-05}
{"train_loss": 0.6227856278419495, "global_step": 13123, "epoch": 147, "lr": 9.946533729694477e-05}
{"train_loss": 0.6926320195198059, "global_step": 13124, "epoch": 147, "lr": 9.946525273915553e-05}
{"train_loss": 0.6103390455245972, "global_step": 13125, "epoch": 147, "lr": 9.94651681747163e-05}
{"train_loss": 0.5214725732803345, "global_step": 13126, "epoch": 147, "lr": 9.946508360362706e-05}
{"train_loss": 0.5316328406333923, "global_step": 13127, "epoch": 147, "lr": 9.946499902588786e-05}
{"train_loss": 0.746528148651123, "global_step": 13128, "epoch": 147, "lr": 9.946491444149867e-05}
{"train_loss": 0.5868924856185913, "global_step": 13129, "epoch": 147, "lr": 9.946482985045956e-05}
{"train_loss": 0.7665222883224487, "global_step": 13130, "epoch": 147, "lr": 9.94647452527705e-05}
{"train_loss": 0.6424902677536011, "global_step": 13131, "epoch": 147, "lr": 9.946466064843149e-05}
{"train_loss": 0.6103703379631042, "global_step": 13132, "epoch": 147, "lr": 9.946457603744257e-05}
{"train_loss": 0.7179251909255981, "global_step": 13133, "epoch": 147, "lr": 9.946449141980374e-05}
{"train_loss": 0.5713961720466614, "global_step": 13134, "epoch": 147, "lr": 9.9464406795515e-05}
{"train_loss": 0.45348456501960754, "global_step": 13135, "epoch": 147, "lr": 9.94643221645764e-05}
{"train_loss": 0.5389006733894348, "global_step": 13136, "epoch": 147, "lr": 9.94642375269879e-05}
{"train_loss": 0.6599216461181641, "global_step": 13137, "epoch": 147, "lr": 9.946415288274955e-05}
{"train_loss": 0.6442069411277771, "global_step": 13138, "epoch": 147, "lr": 9.946406823186135e-05}
{"train_loss": 0.6313357353210449, "global_step": 13139, "epoch": 147, "lr": 9.94639835743233e-05}
{"train_loss": 0.7746142745018005, "global_step": 13140, "epoch": 147, "lr": 9.946389891013541e-05}
{"train_loss": 0.6307266354560852, "global_step": 13141, "epoch": 147, "lr": 9.946381423929771e-05}
{"train_loss": 0.5863968133926392, "global_step": 13142, "epoch": 147, "lr": 9.946372956181021e-05}
{"train_loss": 0.5444058775901794, "global_step": 13143, "epoch": 147, "lr": 9.946364487767293e-05}
{"train_loss": 0.5744606852531433, "global_step": 13144, "epoch": 147, "lr": 9.946356018688583e-05}
{"train_loss": 0.82956862449646, "global_step": 13145, "epoch": 147, "lr": 9.946347548944898e-05}
{"train_loss": 0.861051082611084, "global_step": 13146, "epoch": 147, "lr": 9.946339078536236e-05}
{"train_loss": 0.5494748950004578, "global_step": 13147, "epoch": 147, "lr": 9.946330607462599e-05}
{"train_loss": 0.5725194215774536, "global_step": 13148, "epoch": 147, "lr": 9.946322135723989e-05}
{"train_loss": 0.7098671197891235, "global_step": 13149, "epoch": 147, "lr": 9.946313663320405e-05}
{"train_loss": 0.5056816935539246, "global_step": 13150, "epoch": 147, "lr": 9.94630519025185e-05}
{"train_loss": 0.68104088306427, "global_step": 13151, "epoch": 147, "lr": 9.946296716518325e-05}
{"train_loss": 0.7899556756019592, "global_step": 13152, "epoch": 147, "lr": 9.94628824211983e-05}
{"train_loss": 0.6916982531547546, "global_step": 13153, "epoch": 147, "lr": 9.946279767056368e-05}
{"train_loss": 0.6980852484703064, "global_step": 13154, "epoch": 147, "lr": 9.946271291327937e-05}
{"train_loss": 0.6125071048736572, "global_step": 13155, "epoch": 147, "lr": 9.946262814934542e-05}
{"train_loss": 0.5835861563682556, "global_step": 13156, "epoch": 147, "lr": 9.946254337876182e-05}
{"train_loss": 0.5823786854743958, "global_step": 13157, "epoch": 147, "lr": 9.946245860152858e-05}
{"train_loss": 0.6022133827209473, "global_step": 13158, "epoch": 147, "lr": 9.946237381764572e-05}
{"train_loss": 0.6819300651550293, "global_step": 13159, "epoch": 147, "lr": 9.946228902711323e-05}
{"train_loss": 0.5624187588691711, "global_step": 13160, "epoch": 147, "lr": 9.946220422993116e-05}
{"train_loss": 0.6149298548698425, "global_step": 13161, "epoch": 147, "lr": 9.94621194260995e-05}
{"train_loss": 0.6089344620704651, "global_step": 13162, "epoch": 147, "lr": 9.946203461561822e-05}
{"train_loss": 0.6556541323661804, "global_step": 13163, "epoch": 147, "lr": 9.946194979848742e-05}
{"train_loss": 0.6124721169471741, "global_step": 13164, "epoch": 147, "lr": 9.946186497470704e-05}
{"train_loss": 0.5818643569946289, "global_step": 13165, "epoch": 147, "lr": 9.946178014427712e-05}
{"train_loss": 0.6101416945457458, "global_step": 13166, "epoch": 147, "lr": 9.946169530719767e-05}
{"train_loss": 0.7753721475601196, "global_step": 13167, "epoch": 147, "lr": 9.946161046346869e-05}
{"train_loss": 0.6402275562286377, "global_step": 13168, "epoch": 147, "lr": 9.946152561309022e-05}
{"train_loss": 0.6454421877861023, "global_step": 13169, "epoch": 147, "lr": 9.946144075606222e-05}
{"train_loss": 0.6332497596740723, "global_step": 13170, "epoch": 147, "lr": 9.946135589238475e-05}
{"train_loss": 0.6119863394271122, "global_step": 13171, "epoch": 147, "lr": 9.94612710220578e-05, "val_loss": 1.332400918006897}
{"train_loss": 0.536209225654602, "global_step": 13172, "epoch": 148, "lr": 9.946118614508139e-05}
{"train_loss": 0.675997257232666, "global_step": 13173, "epoch": 148, "lr": 9.946110126145551e-05}
{"train_loss": 0.6444540619850159, "global_step": 13174, "epoch": 148, "lr": 9.946101637118021e-05}
{"train_loss": 0.7497382164001465, "global_step": 13175, "epoch": 148, "lr": 9.946093147425547e-05}
{"train_loss": 0.6792869567871094, "global_step": 13176, "epoch": 148, "lr": 9.946084657068131e-05}
{"train_loss": 0.5759533047676086, "global_step": 13177, "epoch": 148, "lr": 9.946076166045774e-05}
{"train_loss": 0.5419052839279175, "global_step": 13178, "epoch": 148, "lr": 9.946067674358477e-05}
{"train_loss": 0.4828122854232788, "global_step": 13179, "epoch": 148, "lr": 9.946059182006243e-05}
{"train_loss": 0.6434975862503052, "global_step": 13180, "epoch": 148, "lr": 9.94605068898907e-05}
{"train_loss": 0.49446991086006165, "global_step": 13181, "epoch": 148, "lr": 9.946042195306961e-05}
{"train_loss": 0.5762881636619568, "global_step": 13182, "epoch": 148, "lr": 9.946033700959918e-05}
{"train_loss": 0.5818175673484802, "global_step": 13183, "epoch": 148, "lr": 9.94602520594794e-05}
{"train_loss": 0.5215503573417664, "global_step": 13184, "epoch": 148, "lr": 9.94601671027103e-05}
{"train_loss": 0.5693061351776123, "global_step": 13185, "epoch": 148, "lr": 9.946008213929188e-05}
{"train_loss": 0.7249100804328918, "global_step": 13186, "epoch": 148, "lr": 9.945999716922416e-05}
{"train_loss": 0.648003876209259, "global_step": 13187, "epoch": 148, "lr": 9.945991219250713e-05}
{"train_loss": 0.7360053658485413, "global_step": 13188, "epoch": 148, "lr": 9.945982720914084e-05}
{"train_loss": 0.4818333685398102, "global_step": 13189, "epoch": 148, "lr": 9.945974221912527e-05}
{"train_loss": 0.636716902256012, "global_step": 13190, "epoch": 148, "lr": 9.945965722246043e-05}
{"train_loss": 0.5061717629432678, "global_step": 13191, "epoch": 148, "lr": 9.945957221914635e-05}
{"train_loss": 0.60585618019104, "global_step": 13192, "epoch": 148, "lr": 9.945948720918303e-05}
{"train_loss": 0.4632189869880676, "global_step": 13193, "epoch": 148, "lr": 9.945940219257048e-05}
{"train_loss": 0.5912695527076721, "global_step": 13194, "epoch": 148, "lr": 9.945931716930873e-05}
{"train_loss": 0.6065365076065063, "global_step": 13195, "epoch": 148, "lr": 9.945923213939776e-05}
{"train_loss": 0.5949881672859192, "global_step": 13196, "epoch": 148, "lr": 9.945914710283761e-05}
{"train_loss": 0.655154287815094, "global_step": 13197, "epoch": 148, "lr": 9.94590620596283e-05}
{"train_loss": 0.6859937906265259, "global_step": 13198, "epoch": 148, "lr": 9.94589770097698e-05}
{"train_loss": 0.5761651992797852, "global_step": 13199, "epoch": 148, "lr": 9.945889195326214e-05}
{"train_loss": 0.6003852486610413, "global_step": 13200, "epoch": 148, "lr": 9.945880689010533e-05}
{"train_loss": 0.6768444180488586, "global_step": 13201, "epoch": 148, "lr": 9.945872182029941e-05}
{"train_loss": 0.7460603713989258, "global_step": 13202, "epoch": 148, "lr": 9.945863674384435e-05}
{"train_loss": 0.5582396388053894, "global_step": 13203, "epoch": 148, "lr": 9.945855166074018e-05}
{"train_loss": 0.6346869468688965, "global_step": 13204, "epoch": 148, "lr": 9.945846657098692e-05}
{"train_loss": 0.6277603507041931, "global_step": 13205, "epoch": 148, "lr": 9.945838147458457e-05}
{"train_loss": 0.6709250807762146, "global_step": 13206, "epoch": 148, "lr": 9.945829637153315e-05}
{"train_loss": 0.6819180250167847, "global_step": 13207, "epoch": 148, "lr": 9.945821126183264e-05}
{"train_loss": 0.6253591179847717, "global_step": 13208, "epoch": 148, "lr": 9.945812614548309e-05}
{"train_loss": 0.6956679224967957, "global_step": 13209, "epoch": 148, "lr": 9.94580410224845e-05}
{"train_loss": 0.6856418251991272, "global_step": 13210, "epoch": 148, "lr": 9.945795589283688e-05}
{"train_loss": 0.6602559089660645, "global_step": 13211, "epoch": 148, "lr": 9.945787075654024e-05}
{"train_loss": 0.6290538907051086, "global_step": 13212, "epoch": 148, "lr": 9.945778561359458e-05}
{"train_loss": 0.5909107327461243, "global_step": 13213, "epoch": 148, "lr": 9.945770046399993e-05}
{"train_loss": 0.5680353045463562, "global_step": 13214, "epoch": 148, "lr": 9.945761530775632e-05}
{"train_loss": 0.5893781185150146, "global_step": 13215, "epoch": 148, "lr": 9.945753014486371e-05}
{"train_loss": 0.5054283142089844, "global_step": 13216, "epoch": 148, "lr": 9.945744497532214e-05}
{"train_loss": 0.685219407081604, "global_step": 13217, "epoch": 148, "lr": 9.945735979913162e-05}
{"train_loss": 0.6715171933174133, "global_step": 13218, "epoch": 148, "lr": 9.945727461629217e-05}
{"train_loss": 0.5787055492401123, "global_step": 13219, "epoch": 148, "lr": 9.945718942680377e-05}
{"train_loss": 0.6632497906684875, "global_step": 13220, "epoch": 148, "lr": 9.945710423066648e-05}
{"train_loss": 0.5879389643669128, "global_step": 13221, "epoch": 148, "lr": 9.945701902788027e-05}
{"train_loss": 0.6179420351982117, "global_step": 13222, "epoch": 148, "lr": 9.945693381844517e-05}
{"train_loss": 0.48288604617118835, "global_step": 13223, "epoch": 148, "lr": 9.94568486023612e-05}
{"train_loss": 0.6082018613815308, "global_step": 13224, "epoch": 148, "lr": 9.945676337962835e-05}
{"train_loss": 0.6842362880706787, "global_step": 13225, "epoch": 148, "lr": 9.945667815024664e-05}
{"train_loss": 0.7482233643531799, "global_step": 13226, "epoch": 148, "lr": 9.945659291421607e-05}
{"train_loss": 0.7964159846305847, "global_step": 13227, "epoch": 148, "lr": 9.945650767153668e-05}
{"train_loss": 0.7259268760681152, "global_step": 13228, "epoch": 148, "lr": 9.945642242220846e-05}
{"train_loss": 0.5568918585777283, "global_step": 13229, "epoch": 148, "lr": 9.945633716623144e-05}
{"train_loss": 0.6354913711547852, "global_step": 13230, "epoch": 148, "lr": 9.945625190360559e-05}
{"train_loss": 0.7596715092658997, "global_step": 13231, "epoch": 148, "lr": 9.945616663433097e-05}
{"train_loss": 0.6199750900268555, "global_step": 13232, "epoch": 148, "lr": 9.945608135840758e-05}
{"train_loss": 0.8242278695106506, "global_step": 13233, "epoch": 148, "lr": 9.94559960758354e-05}
{"train_loss": 0.6264048218727112, "global_step": 13234, "epoch": 148, "lr": 9.945591078661446e-05}
{"train_loss": 0.5665550231933594, "global_step": 13235, "epoch": 148, "lr": 9.94558254907448e-05}
{"train_loss": 0.8544211387634277, "global_step": 13236, "epoch": 148, "lr": 9.945574018822641e-05}
{"train_loss": 0.6671265363693237, "global_step": 13237, "epoch": 148, "lr": 9.945565487905928e-05}
{"train_loss": 0.745402455329895, "global_step": 13238, "epoch": 148, "lr": 9.945556956324344e-05}
{"train_loss": 0.6597316861152649, "global_step": 13239, "epoch": 148, "lr": 9.94554842407789e-05}
{"train_loss": 0.5873099565505981, "global_step": 13240, "epoch": 148, "lr": 9.945539891166569e-05}
{"train_loss": 0.6786719560623169, "global_step": 13241, "epoch": 148, "lr": 9.945531357590378e-05}
{"train_loss": 0.6364244222640991, "global_step": 13242, "epoch": 148, "lr": 9.945522823349322e-05}
{"train_loss": 0.4160098731517792, "global_step": 13243, "epoch": 148, "lr": 9.945514288443402e-05}
{"train_loss": 0.6412267088890076, "global_step": 13244, "epoch": 148, "lr": 9.945505752872617e-05}
{"train_loss": 0.6911514401435852, "global_step": 13245, "epoch": 148, "lr": 9.945497216636968e-05}
{"train_loss": 0.726248025894165, "global_step": 13246, "epoch": 148, "lr": 9.945488679736457e-05}
{"train_loss": 0.5468806028366089, "global_step": 13247, "epoch": 148, "lr": 9.945480142171087e-05}
{"train_loss": 0.5530329942703247, "global_step": 13248, "epoch": 148, "lr": 9.945471603940857e-05}
{"train_loss": 0.5805245041847229, "global_step": 13249, "epoch": 148, "lr": 9.945463065045768e-05}
{"train_loss": 0.4990035891532898, "global_step": 13250, "epoch": 148, "lr": 9.945454525485822e-05}
{"train_loss": 0.5700050592422485, "global_step": 13251, "epoch": 148, "lr": 9.945445985261019e-05}
{"train_loss": 0.5953606963157654, "global_step": 13252, "epoch": 148, "lr": 9.945437444371363e-05}
{"train_loss": 0.6567718386650085, "global_step": 13253, "epoch": 148, "lr": 9.945428902816852e-05}
{"train_loss": 0.8128389716148376, "global_step": 13254, "epoch": 148, "lr": 9.945420360597488e-05}
{"train_loss": 0.6439734101295471, "global_step": 13255, "epoch": 148, "lr": 9.945411817713273e-05}
{"train_loss": 0.5264853835105896, "global_step": 13256, "epoch": 148, "lr": 9.94540327416421e-05}
{"train_loss": 0.48107632994651794, "global_step": 13257, "epoch": 148, "lr": 9.945394729950293e-05}
{"train_loss": 0.6058496832847595, "global_step": 13258, "epoch": 148, "lr": 9.945386185071532e-05}
{"train_loss": 0.6702772378921509, "global_step": 13259, "epoch": 148, "lr": 9.945377639527921e-05}
{"train_loss": 0.6273547059364533, "global_step": 13260, "epoch": 148, "lr": 9.945369093319466e-05, "val_loss": 1.4267159700393677}
{"train_loss": 0.758746325969696, "global_step": 13261, "epoch": 149, "lr": 9.945360546446167e-05}
{"train_loss": 0.6413170695304871, "global_step": 13262, "epoch": 149, "lr": 9.945351998908023e-05}
{"train_loss": 0.7816325426101685, "global_step": 13263, "epoch": 149, "lr": 9.945343450705039e-05}
{"train_loss": 0.6662384271621704, "global_step": 13264, "epoch": 149, "lr": 9.945334901837212e-05}
{"train_loss": 0.7369120717048645, "global_step": 13265, "epoch": 149, "lr": 9.945326352304545e-05}
{"train_loss": 0.5727229714393616, "global_step": 13266, "epoch": 149, "lr": 9.945317802107039e-05}
{"train_loss": 0.628808319568634, "global_step": 13267, "epoch": 149, "lr": 9.945309251244696e-05}
{"train_loss": 0.5229048132896423, "global_step": 13268, "epoch": 149, "lr": 9.945300699717516e-05}
{"train_loss": 0.5489594340324402, "global_step": 13269, "epoch": 149, "lr": 9.9452921475255e-05}
{"train_loss": 0.4934721291065216, "global_step": 13270, "epoch": 149, "lr": 9.94528359466865e-05}
{"train_loss": 0.652439534664154, "global_step": 13271, "epoch": 149, "lr": 9.945275041146968e-05}
{"train_loss": 0.6406729221343994, "global_step": 13272, "epoch": 149, "lr": 9.945266486960454e-05}
{"train_loss": 0.5527070760726929, "global_step": 13273, "epoch": 149, "lr": 9.94525793210911e-05}
{"train_loss": 0.6756724715232849, "global_step": 13274, "epoch": 149, "lr": 9.945249376592933e-05}
{"train_loss": 0.4983268082141876, "global_step": 13275, "epoch": 149, "lr": 9.945240820411929e-05}
{"train_loss": 0.6371197700500488, "global_step": 13276, "epoch": 149, "lr": 9.9452322635661e-05}
{"train_loss": 0.4321907162666321, "global_step": 13277, "epoch": 149, "lr": 9.945223706055442e-05}
{"train_loss": 0.5259004235267639, "global_step": 13278, "epoch": 149, "lr": 9.94521514787996e-05}
{"train_loss": 0.5162467360496521, "global_step": 13279, "epoch": 149, "lr": 9.945206589039656e-05}
{"train_loss": 0.7591577172279358, "global_step": 13280, "epoch": 149, "lr": 9.945198029534526e-05}
{"train_loss": 0.6054126620292664, "global_step": 13281, "epoch": 149, "lr": 9.945189469364576e-05}
{"train_loss": 0.559349000453949, "global_step": 13282, "epoch": 149, "lr": 9.945180908529805e-05}
{"train_loss": 0.6421664953231812, "global_step": 13283, "epoch": 149, "lr": 9.945172347030213e-05}
{"train_loss": 0.6288846731185913, "global_step": 13284, "epoch": 149, "lr": 9.945163784865807e-05}
{"train_loss": 0.5325071215629578, "global_step": 13285, "epoch": 149, "lr": 9.945155222036581e-05}
{"train_loss": 0.5524913668632507, "global_step": 13286, "epoch": 149, "lr": 9.945146658542539e-05}
{"train_loss": 0.5741129517555237, "global_step": 13287, "epoch": 149, "lr": 9.945138094383684e-05}
{"train_loss": 0.5691221356391907, "global_step": 13288, "epoch": 149, "lr": 9.945129529560014e-05}
{"train_loss": 0.5164728760719299, "global_step": 13289, "epoch": 149, "lr": 9.945120964071533e-05}
{"train_loss": 0.5725642442703247, "global_step": 13290, "epoch": 149, "lr": 9.94511239791824e-05}
{"train_loss": 0.6047871112823486, "global_step": 13291, "epoch": 149, "lr": 9.945103831100138e-05}
{"train_loss": 0.5598078370094299, "global_step": 13292, "epoch": 149, "lr": 9.945095263617226e-05}
{"train_loss": 0.7680211067199707, "global_step": 13293, "epoch": 149, "lr": 9.945086695469505e-05}
{"train_loss": 0.5351917147636414, "global_step": 13294, "epoch": 149, "lr": 9.94507812665698e-05}
{"train_loss": 0.6243652105331421, "global_step": 13295, "epoch": 149, "lr": 9.945069557179647e-05}
{"train_loss": 0.4942764937877655, "global_step": 13296, "epoch": 149, "lr": 9.945060987037511e-05}
{"train_loss": 0.6953253149986267, "global_step": 13297, "epoch": 149, "lr": 9.945052416230571e-05}
{"train_loss": 0.6717345714569092, "global_step": 13298, "epoch": 149, "lr": 9.94504384475883e-05}
{"train_loss": 0.6293669939041138, "global_step": 13299, "epoch": 149, "lr": 9.945035272622287e-05}
{"train_loss": 0.7182625532150269, "global_step": 13300, "epoch": 149, "lr": 9.945026699820945e-05}
{"train_loss": 0.49093160033226013, "global_step": 13301, "epoch": 149, "lr": 9.945018126354805e-05}
{"train_loss": 0.7314170002937317, "global_step": 13302, "epoch": 149, "lr": 9.945009552223867e-05}
{"train_loss": 0.6318746209144592, "global_step": 13303, "epoch": 149, "lr": 9.945000977428132e-05}
{"train_loss": 0.7464224100112915, "global_step": 13304, "epoch": 149, "lr": 9.944992401967603e-05}
{"train_loss": 0.6207979917526245, "global_step": 13305, "epoch": 149, "lr": 9.94498382584228e-05}
{"train_loss": 0.5723632574081421, "global_step": 13306, "epoch": 149, "lr": 9.944975249052164e-05}
{"train_loss": 0.586668848991394, "global_step": 13307, "epoch": 149, "lr": 9.944966671597255e-05}
{"train_loss": 0.4924752116203308, "global_step": 13308, "epoch": 149, "lr": 9.944958093477558e-05}
{"train_loss": 0.6764415502548218, "global_step": 13309, "epoch": 149, "lr": 9.944949514693069e-05}
{"train_loss": 0.6134811043739319, "global_step": 13310, "epoch": 149, "lr": 9.944940935243794e-05}
{"train_loss": 0.6936588883399963, "global_step": 13311, "epoch": 149, "lr": 9.944932355129732e-05}
{"train_loss": 0.6659095883369446, "global_step": 13312, "epoch": 149, "lr": 9.944923774350883e-05}
{"train_loss": 0.6780279278755188, "global_step": 13313, "epoch": 149, "lr": 9.944915192907249e-05}
{"train_loss": 0.5411038398742676, "global_step": 13314, "epoch": 149, "lr": 9.944906610798832e-05}
{"train_loss": 0.5765491127967834, "global_step": 13315, "epoch": 149, "lr": 9.944898028025633e-05}
{"train_loss": 0.7112546563148499, "global_step": 13316, "epoch": 149, "lr": 9.944889444587652e-05}
{"train_loss": 0.7552956938743591, "global_step": 13317, "epoch": 149, "lr": 9.94488086048489e-05}
{"train_loss": 0.7276448607444763, "global_step": 13318, "epoch": 149, "lr": 9.944872275717351e-05}
{"train_loss": 0.5746614933013916, "global_step": 13319, "epoch": 149, "lr": 9.944863690285034e-05}
{"train_loss": 0.758307695388794, "global_step": 13320, "epoch": 149, "lr": 9.944855104187939e-05}
{"train_loss": 0.6838876008987427, "global_step": 13321, "epoch": 149, "lr": 9.94484651742607e-05}
{"train_loss": 0.528553307056427, "global_step": 13322, "epoch": 149, "lr": 9.944837929999426e-05}
{"train_loss": 0.671539843082428, "global_step": 13323, "epoch": 149, "lr": 9.944829341908009e-05}
{"train_loss": 0.7078783512115479, "global_step": 13324, "epoch": 149, "lr": 9.944820753151818e-05}
{"train_loss": 0.6123692393302917, "global_step": 13325, "epoch": 149, "lr": 9.944812163730858e-05}
{"train_loss": 0.6989870071411133, "global_step": 13326, "epoch": 149, "lr": 9.944803573645129e-05}
{"train_loss": 0.656758725643158, "global_step": 13327, "epoch": 149, "lr": 9.94479498289463e-05}
{"train_loss": 0.6467708349227905, "global_step": 13328, "epoch": 149, "lr": 9.944786391479364e-05}
{"train_loss": 0.8860360980033875, "global_step": 13329, "epoch": 149, "lr": 9.944777799399332e-05}
{"train_loss": 0.5804629921913147, "global_step": 13330, "epoch": 149, "lr": 9.944769206654535e-05}
{"train_loss": 0.715320885181427, "global_step": 13331, "epoch": 149, "lr": 9.944760613244972e-05}
{"train_loss": 0.6692081689834595, "global_step": 13332, "epoch": 149, "lr": 9.94475201917065e-05}
{"train_loss": 0.5571197867393494, "global_step": 13333, "epoch": 149, "lr": 9.944743424431563e-05}
{"train_loss": 0.6607595682144165, "global_step": 13334, "epoch": 149, "lr": 9.944734829027718e-05}
{"train_loss": 0.6538468599319458, "global_step": 13335, "epoch": 149, "lr": 9.944726232959111e-05}
{"train_loss": 0.5822270512580872, "global_step": 13336, "epoch": 149, "lr": 9.944717636225748e-05}
{"train_loss": 0.7265605926513672, "global_step": 13337, "epoch": 149, "lr": 9.944709038827626e-05}
{"train_loss": 0.6380773782730103, "global_step": 13338, "epoch": 149, "lr": 9.94470044076475e-05}
{"train_loss": 0.5793582201004028, "global_step": 13339, "epoch": 149, "lr": 9.944691842037119e-05}
{"train_loss": 0.6039603352546692, "global_step": 13340, "epoch": 149, "lr": 9.944683242644735e-05}
{"train_loss": 0.71009361743927, "global_step": 13341, "epoch": 149, "lr": 9.944674642587597e-05}
{"train_loss": 0.5873267650604248, "global_step": 13342, "epoch": 149, "lr": 9.94466604186571e-05}
{"train_loss": 0.49557727575302124, "global_step": 13343, "epoch": 149, "lr": 9.944657440479071e-05}
{"train_loss": 0.5747870206832886, "global_step": 13344, "epoch": 149, "lr": 9.944648838427682e-05}
{"train_loss": 0.4589342474937439, "global_step": 13345, "epoch": 149, "lr": 9.944640235711546e-05}
{"train_loss": 0.6902425289154053, "global_step": 13346, "epoch": 149, "lr": 9.944631632330666e-05}
{"train_loss": 0.43360647559165955, "global_step": 13347, "epoch": 149, "lr": 9.944623028285039e-05}
{"train_loss": 0.6677414774894714, "global_step": 13348, "epoch": 149, "lr": 9.944614423574666e-05}
{"train_loss": 0.6221271837025546, "global_step": 13349, "epoch": 149, "lr": 9.944605818199551e-05, "val_loss": 1.3746963739395142}
{"train_loss": 0.6235353350639343, "global_step": 13350, "epoch": 150, "lr": 9.944597212159695e-05}
{"train_loss": 0.5552242994308472, "global_step": 13351, "epoch": 150, "lr": 9.944588605455097e-05}
{"train_loss": 0.6273617148399353, "global_step": 13352, "epoch": 150, "lr": 9.944579998085761e-05}
{"train_loss": 0.6197223663330078, "global_step": 13353, "epoch": 150, "lr": 9.944571390051685e-05}
{"train_loss": 0.5823349952697754, "global_step": 13354, "epoch": 150, "lr": 9.944562781352871e-05}
{"train_loss": 0.5794854164123535, "global_step": 13355, "epoch": 150, "lr": 9.944554171989322e-05}
{"train_loss": 0.5018883943557739, "global_step": 13356, "epoch": 150, "lr": 9.944545561961037e-05}
{"train_loss": 0.5799700617790222, "global_step": 13357, "epoch": 150, "lr": 9.94453695126802e-05}
{"train_loss": 0.5337043404579163, "global_step": 13358, "epoch": 150, "lr": 9.944528339910268e-05}
{"train_loss": 0.5353020429611206, "global_step": 13359, "epoch": 150, "lr": 9.944519727887785e-05}
{"train_loss": 0.6488305926322937, "global_step": 13360, "epoch": 150, "lr": 9.944511115200573e-05}
{"train_loss": 0.5538573265075684, "global_step": 13361, "epoch": 150, "lr": 9.94450250184863e-05}
{"train_loss": 0.5517802834510803, "global_step": 13362, "epoch": 150, "lr": 9.94449388783196e-05}
{"train_loss": 0.6173096299171448, "global_step": 13363, "epoch": 150, "lr": 9.944485273150564e-05}
{"train_loss": 0.6016848087310791, "global_step": 13364, "epoch": 150, "lr": 9.94447665780444e-05}
{"train_loss": 0.5697751045227051, "global_step": 13365, "epoch": 150, "lr": 9.944468041793592e-05}
{"train_loss": 0.6437404751777649, "global_step": 13366, "epoch": 150, "lr": 9.94445942511802e-05}
{"train_loss": 0.5937679409980774, "global_step": 13367, "epoch": 150, "lr": 9.944450807777727e-05}
{"train_loss": 0.5035809874534607, "global_step": 13368, "epoch": 150, "lr": 9.944442189772714e-05}
{"train_loss": 0.6969749331474304, "global_step": 13369, "epoch": 150, "lr": 9.944433571102978e-05}
{"train_loss": 0.6889508962631226, "global_step": 13370, "epoch": 150, "lr": 9.944424951768525e-05}
{"train_loss": 0.4710225760936737, "global_step": 13371, "epoch": 150, "lr": 9.944416331769354e-05}
{"train_loss": 0.6570690274238586, "global_step": 13372, "epoch": 150, "lr": 9.944407711105466e-05}
{"train_loss": 0.49003270268440247, "global_step": 13373, "epoch": 150, "lr": 9.944399089776864e-05}
{"train_loss": 0.5706071257591248, "global_step": 13374, "epoch": 150, "lr": 9.944390467783547e-05}
{"train_loss": 0.6575804352760315, "global_step": 13375, "epoch": 150, "lr": 9.944381845125516e-05}
{"train_loss": 0.6470891237258911, "global_step": 13376, "epoch": 150, "lr": 9.944373221802774e-05}
{"train_loss": 0.6024981141090393, "global_step": 13377, "epoch": 150, "lr": 9.944364597815321e-05}
{"train_loss": 0.5543871521949768, "global_step": 13378, "epoch": 150, "lr": 9.94435597316316e-05}
{"train_loss": 0.5849394202232361, "global_step": 13379, "epoch": 150, "lr": 9.944347347846287e-05}
{"train_loss": 0.6949851512908936, "global_step": 13380, "epoch": 150, "lr": 9.94433872186471e-05}
{"train_loss": 0.5636304020881653, "global_step": 13381, "epoch": 150, "lr": 9.944330095218426e-05}
{"train_loss": 0.6261752843856812, "global_step": 13382, "epoch": 150, "lr": 9.944321467907438e-05}
{"train_loss": 0.6206246018409729, "global_step": 13383, "epoch": 150, "lr": 9.944312839931744e-05}
{"train_loss": 0.6218752861022949, "global_step": 13384, "epoch": 150, "lr": 9.944304211291349e-05}
{"train_loss": 0.6597738862037659, "global_step": 13385, "epoch": 150, "lr": 9.944295581986252e-05}
{"train_loss": 0.5931461453437805, "global_step": 13386, "epoch": 150, "lr": 9.944286952016454e-05}
{"train_loss": 0.6830443739891052, "global_step": 13387, "epoch": 150, "lr": 9.944278321381959e-05}
{"train_loss": 0.6235597729682922, "global_step": 13388, "epoch": 150, "lr": 9.944269690082766e-05}
{"train_loss": 0.46195971965789795, "global_step": 13389, "epoch": 150, "lr": 9.944261058118875e-05}
{"train_loss": 0.6051897406578064, "global_step": 13390, "epoch": 150, "lr": 9.944252425490288e-05}
{"train_loss": 0.5621774196624756, "global_step": 13391, "epoch": 150, "lr": 9.944243792197008e-05}
{"train_loss": 0.5156463384628296, "global_step": 13392, "epoch": 150, "lr": 9.944235158239033e-05}
{"train_loss": 0.5967755317687988, "global_step": 13393, "epoch": 150, "lr": 9.944226523616367e-05}
{"train_loss": 0.5661098957061768, "global_step": 13394, "epoch": 150, "lr": 9.944217888329009e-05}
{"train_loss": 0.5716504454612732, "global_step": 13395, "epoch": 150, "lr": 9.944209252376961e-05}
{"train_loss": 0.6992055773735046, "global_step": 13396, "epoch": 150, "lr": 9.944200615760225e-05}
{"train_loss": 0.5279116630554199, "global_step": 13397, "epoch": 150, "lr": 9.944191978478802e-05}
{"train_loss": 0.48312920331954956, "global_step": 13398, "epoch": 150, "lr": 9.944183340532693e-05}
{"train_loss": 0.46731114387512207, "global_step": 13399, "epoch": 150, "lr": 9.944174701921898e-05}
{"train_loss": 0.7697712182998657, "global_step": 13400, "epoch": 150, "lr": 9.944166062646419e-05}
{"train_loss": 0.737373411655426, "global_step": 13401, "epoch": 150, "lr": 9.944157422706256e-05}
{"train_loss": 0.642987072467804, "global_step": 13402, "epoch": 150, "lr": 9.944148782101412e-05}
{"train_loss": 0.49371209740638733, "global_step": 13403, "epoch": 150, "lr": 9.944140140831888e-05}
{"train_loss": 0.7223894596099854, "global_step": 13404, "epoch": 150, "lr": 9.944131498897684e-05}
{"train_loss": 0.4167911112308502, "global_step": 13405, "epoch": 150, "lr": 9.944122856298803e-05}
{"train_loss": 0.6634913682937622, "global_step": 13406, "epoch": 150, "lr": 9.944114213035245e-05}
{"train_loss": 0.5995818376541138, "global_step": 13407, "epoch": 150, "lr": 9.94410556910701e-05}
{"train_loss": 0.6011888384819031, "global_step": 13408, "epoch": 150, "lr": 9.9440969245141e-05}
{"train_loss": 0.7244937419891357, "global_step": 13409, "epoch": 150, "lr": 9.944088279256517e-05}
{"train_loss": 0.6298221349716187, "global_step": 13410, "epoch": 150, "lr": 9.944079633334262e-05}
{"train_loss": 0.5902296900749207, "global_step": 13411, "epoch": 150, "lr": 9.944070986747335e-05}
{"train_loss": 0.5117897391319275, "global_step": 13412, "epoch": 150, "lr": 9.944062339495738e-05}
{"train_loss": 0.7227288484573364, "global_step": 13413, "epoch": 150, "lr": 9.944053691579473e-05}
{"train_loss": 0.5754127502441406, "global_step": 13414, "epoch": 150, "lr": 9.94404504299854e-05}
{"train_loss": 0.7013435959815979, "global_step": 13415, "epoch": 150, "lr": 9.944036393752939e-05}
{"train_loss": 0.5319175720214844, "global_step": 13416, "epoch": 150, "lr": 9.944027743842675e-05}
{"train_loss": 0.6333286762237549, "global_step": 13417, "epoch": 150, "lr": 9.944019093267745e-05}
{"train_loss": 0.6776342988014221, "global_step": 13418, "epoch": 150, "lr": 9.944010442028151e-05}
{"train_loss": 0.6832094192504883, "global_step": 13419, "epoch": 150, "lr": 9.944001790123896e-05}
{"train_loss": 0.5504572987556458, "global_step": 13420, "epoch": 150, "lr": 9.943993137554981e-05}
{"train_loss": 0.6090949177742004, "global_step": 13421, "epoch": 150, "lr": 9.943984484321404e-05}
{"train_loss": 0.6999688744544983, "global_step": 13422, "epoch": 150, "lr": 9.943975830423171e-05}
{"train_loss": 0.5989996790885925, "global_step": 13423, "epoch": 150, "lr": 9.94396717586028e-05}
{"train_loss": 0.7484349012374878, "global_step": 13424, "epoch": 150, "lr": 9.943958520632733e-05}
{"train_loss": 0.6935003399848938, "global_step": 13425, "epoch": 150, "lr": 9.943949864740532e-05}
{"train_loss": 0.7794697880744934, "global_step": 13426, "epoch": 150, "lr": 9.943941208183676e-05}
{"train_loss": 0.5609589219093323, "global_step": 13427, "epoch": 150, "lr": 9.943932550962168e-05}
{"train_loss": 0.8594051003456116, "global_step": 13428, "epoch": 150, "lr": 9.943923893076009e-05}
{"train_loss": 0.648162305355072, "global_step": 13429, "epoch": 150, "lr": 9.943915234525197e-05}
{"train_loss": 0.5511879920959473, "global_step": 13430, "epoch": 150, "lr": 9.943906575309738e-05}
{"train_loss": 0.5615320801734924, "global_step": 13431, "epoch": 150, "lr": 9.943897915429629e-05}
{"train_loss": 0.5504583120346069, "global_step": 13432, "epoch": 150, "lr": 9.943889254884876e-05}
{"train_loss": 0.5758275985717773, "global_step": 13433, "epoch": 150, "lr": 9.943880593675476e-05}
{"train_loss": 0.5933067202568054, "global_step": 13434, "epoch": 150, "lr": 9.943871931801431e-05}
{"train_loss": 0.7217241525650024, "global_step": 13435, "epoch": 150, "lr": 9.943863269262745e-05}
{"train_loss": 0.66562819480896, "global_step": 13436, "epoch": 150, "lr": 9.943854606059415e-05}
{"train_loss": 0.6011562347412109, "global_step": 13437, "epoch": 150, "lr": 9.943845942191443e-05}
{"train_loss": 0.6108269550827112, "global_step": 13438, "epoch": 150, "lr": 9.943837277658833e-05, "train/sim_max_reward_0": 0.6407557387561871, "train/sim_max_reward_1": 0.9697831585098865, "train/sim_max_reward_2": 0.13546330546024807, "train/sim_max_reward_3": 0.3481243462004995, "train/sim_max_reward_4": 0.009503307570729362, "train/sim_max_reward_5": 0.6102708960546799, "test/sim_max_reward_4500000": 0.8583013407861027, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9539889293807122, "test/sim_max_reward_4500003": 0.49581848602063855, "test/sim_max_reward_4500004": 0.49652625063763484, "test/sim_max_reward_4500005": 0.9029425328647372, "test/sim_max_reward_4500006": 0.4321664743552902, "test/sim_max_reward_4500007": 0.5827332239331455, "test/sim_max_reward_4500008": 0.0, "test/sim_max_reward_4500009": 0.5315181508497085, "test/sim_max_reward_4500010": 0.9480175108302007, "test/sim_max_reward_4500011": 0.9514920445544924, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.540438123742596, "test/sim_max_reward_4500014": 0.8486398561207056, "test/sim_max_reward_4500015": 0.8549336791400105, "test/sim_max_reward_4500016": 0.5112336833689249, "test/sim_max_reward_4500017": 0.4156421068063586, "test/sim_max_reward_4500018": 0.6281752882925755, "test/sim_max_reward_4500019": 0.39496762743265906, "test/sim_max_reward_4500020": 0.3867153745522534, "test/sim_max_reward_4500021": 0.931469575712853, "test/sim_max_reward_4500022": 0.8128949854152201, "test/sim_max_reward_4500023": 0.6739527257290941, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.4047687818763128, "test/sim_max_reward_4500026": 0.7723263219545247, "test/sim_max_reward_4500027": 0.7515733554100659, "test/sim_max_reward_4500028": 0.5952314168542491, "test/sim_max_reward_4500029": 0.07516474435023965, "test/sim_max_reward_4500030": 0.46671096540283397, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.49074139031520225, "test/sim_max_reward_4500033": 0.542169455781329, "test/sim_max_reward_4500034": 0.9438146779853512, "test/sim_max_reward_4500035": 0.6764746424858614, "test/sim_max_reward_4500036": 0.9473225816479262, "test/sim_max_reward_4500037": 0.19826174665814478, "test/sim_max_reward_4500038": 0.5136833743849414, "test/sim_max_reward_4500039": 0.029767022369779732, "test/sim_max_reward_4500040": 0.17536757291094343, "test/sim_max_reward_4500041": 0.5474519679846411, "test/sim_max_reward_4500042": 0.6349575496754543, "test/sim_max_reward_4500043": 0.5936571026087455, "test/sim_max_reward_4500044": 0.4932884695350994, "test/sim_max_reward_4500045": 0.9064731194678297, "test/sim_max_reward_4500046": 0.5697257401423718, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9523004774064109, "test/sim_max_reward_4500049": 0.5833637170784158, "train/mean_score": 0.45231679209203834, "test/mean_score": 0.5403432832962518, "val_loss": 1.4198447465896606, "train_action_mse_error": 40.46923065185547}
{"train_loss": 0.741027295589447, "global_step": 13439, "epoch": 151, "lr": 9.943828612461583e-05}
{"train_loss": 0.7684245705604553, "global_step": 13440, "epoch": 151, "lr": 9.943819946599698e-05}
{"train_loss": 0.5996198058128357, "global_step": 13441, "epoch": 151, "lr": 9.943811280073175e-05}
{"train_loss": 0.4106834828853607, "global_step": 13442, "epoch": 151, "lr": 9.943802612882016e-05}
{"train_loss": 0.6624206304550171, "global_step": 13443, "epoch": 151, "lr": 9.943793945026223e-05}
{"train_loss": 0.5404545664787292, "global_step": 13444, "epoch": 151, "lr": 9.943785276505797e-05}
{"train_loss": 0.5320919156074524, "global_step": 13445, "epoch": 151, "lr": 9.943776607320741e-05}
{"train_loss": 0.7337386012077332, "global_step": 13446, "epoch": 151, "lr": 9.943767937471053e-05}
{"train_loss": 0.6554486155509949, "global_step": 13447, "epoch": 151, "lr": 9.943759266956737e-05}
{"train_loss": 0.45199158787727356, "global_step": 13448, "epoch": 151, "lr": 9.943750595777792e-05}
{"train_loss": 0.5203402638435364, "global_step": 13449, "epoch": 151, "lr": 9.94374192393422e-05}
{"train_loss": 0.7565509080886841, "global_step": 13450, "epoch": 151, "lr": 9.943733251426022e-05}
{"train_loss": 0.6009183526039124, "global_step": 13451, "epoch": 151, "lr": 9.9437245782532e-05}
{"train_loss": 0.5291526913642883, "global_step": 13452, "epoch": 151, "lr": 9.943715904415754e-05}
{"train_loss": 0.45767447352409363, "global_step": 13453, "epoch": 151, "lr": 9.943707229913686e-05}
{"train_loss": 0.6364010572433472, "global_step": 13454, "epoch": 151, "lr": 9.943698554746996e-05}
{"train_loss": 0.573951780796051, "global_step": 13455, "epoch": 151, "lr": 9.943689878915687e-05}
{"train_loss": 0.46047645807266235, "global_step": 13456, "epoch": 151, "lr": 9.943681202419759e-05}
{"train_loss": 0.6100448369979858, "global_step": 13457, "epoch": 151, "lr": 9.943672525259213e-05}
{"train_loss": 0.5318132638931274, "global_step": 13458, "epoch": 151, "lr": 9.943663847434051e-05}
{"train_loss": 0.5117050409317017, "global_step": 13459, "epoch": 151, "lr": 9.943655168944272e-05}
{"train_loss": 0.7227185964584351, "global_step": 13460, "epoch": 151, "lr": 9.94364648978988e-05}
{"train_loss": 0.4833080768585205, "global_step": 13461, "epoch": 151, "lr": 9.943637809970874e-05}
{"train_loss": 0.6939346194267273, "global_step": 13462, "epoch": 151, "lr": 9.943629129487257e-05}
{"train_loss": 0.5700170993804932, "global_step": 13463, "epoch": 151, "lr": 9.94362044833903e-05}
{"train_loss": 0.49732068181037903, "global_step": 13464, "epoch": 151, "lr": 9.943611766526193e-05}
{"train_loss": 0.5444098711013794, "global_step": 13465, "epoch": 151, "lr": 9.943603084048749e-05}
{"train_loss": 0.6926623582839966, "global_step": 13466, "epoch": 151, "lr": 9.943594400906695e-05}
{"train_loss": 0.7371048927307129, "global_step": 13467, "epoch": 151, "lr": 9.943585717100037e-05}
{"train_loss": 0.6344559192657471, "global_step": 13468, "epoch": 151, "lr": 9.943577032628773e-05}
{"train_loss": 0.5892270803451538, "global_step": 13469, "epoch": 151, "lr": 9.943568347492906e-05}
{"train_loss": 0.6266423463821411, "global_step": 13470, "epoch": 151, "lr": 9.943559661692436e-05}
{"train_loss": 0.5361081957817078, "global_step": 13471, "epoch": 151, "lr": 9.943550975227365e-05}
{"train_loss": 0.6150137186050415, "global_step": 13472, "epoch": 151, "lr": 9.943542288097694e-05}
{"train_loss": 0.5675219297409058, "global_step": 13473, "epoch": 151, "lr": 9.943533600303425e-05}
{"train_loss": 0.5448352694511414, "global_step": 13474, "epoch": 151, "lr": 9.943524911844556e-05}
{"train_loss": 0.5338324904441833, "global_step": 13475, "epoch": 151, "lr": 9.943516222721093e-05}
{"train_loss": 0.5895906686782837, "global_step": 13476, "epoch": 151, "lr": 9.943507532933032e-05}
{"train_loss": 0.5748664736747742, "global_step": 13477, "epoch": 151, "lr": 9.943498842480377e-05}
{"train_loss": 0.6147558689117432, "global_step": 13478, "epoch": 151, "lr": 9.943490151363128e-05}
{"train_loss": 0.6188797354698181, "global_step": 13479, "epoch": 151, "lr": 9.943481459581289e-05}
{"train_loss": 0.7008470296859741, "global_step": 13480, "epoch": 151, "lr": 9.943472767134858e-05}
{"train_loss": 0.4388048052787781, "global_step": 13481, "epoch": 151, "lr": 9.943464074023838e-05}
{"train_loss": 0.7711570858955383, "global_step": 13482, "epoch": 151, "lr": 9.94345538024823e-05}
{"train_loss": 0.5466665029525757, "global_step": 13483, "epoch": 151, "lr": 9.943446685808034e-05}
{"train_loss": 0.6701698303222656, "global_step": 13484, "epoch": 151, "lr": 9.943437990703252e-05}
{"train_loss": 0.6081233620643616, "global_step": 13485, "epoch": 151, "lr": 9.943429294933884e-05}
{"train_loss": 0.636410653591156, "global_step": 13486, "epoch": 151, "lr": 9.943420598499933e-05}
{"train_loss": 0.7074499726295471, "global_step": 13487, "epoch": 151, "lr": 9.9434119014014e-05}
{"train_loss": 0.46695560216903687, "global_step": 13488, "epoch": 151, "lr": 9.943403203638283e-05}
{"train_loss": 0.5618218183517456, "global_step": 13489, "epoch": 151, "lr": 9.943394505210589e-05}
{"train_loss": 0.691860556602478, "global_step": 13490, "epoch": 151, "lr": 9.943385806118315e-05}
{"train_loss": 0.7242707014083862, "global_step": 13491, "epoch": 151, "lr": 9.943377106361462e-05}
{"train_loss": 0.5765247941017151, "global_step": 13492, "epoch": 151, "lr": 9.943368405940032e-05}
{"train_loss": 0.583651065826416, "global_step": 13493, "epoch": 151, "lr": 9.943359704854027e-05}
{"train_loss": 0.5498822927474976, "global_step": 13494, "epoch": 151, "lr": 9.943351003103448e-05}
{"train_loss": 0.5577873587608337, "global_step": 13495, "epoch": 151, "lr": 9.943342300688295e-05}
{"train_loss": 0.6111466288566589, "global_step": 13496, "epoch": 151, "lr": 9.94333359760857e-05}
{"train_loss": 0.4572807252407074, "global_step": 13497, "epoch": 151, "lr": 9.943324893864274e-05}
{"train_loss": 0.6424863934516907, "global_step": 13498, "epoch": 151, "lr": 9.94331618945541e-05}
{"train_loss": 0.6205775141716003, "global_step": 13499, "epoch": 151, "lr": 9.943307484381975e-05}
{"train_loss": 0.6318320035934448, "global_step": 13500, "epoch": 151, "lr": 9.943298778643974e-05}
{"train_loss": 0.452007532119751, "global_step": 13501, "epoch": 151, "lr": 9.943290072241406e-05}
{"train_loss": 0.5956289172172546, "global_step": 13502, "epoch": 151, "lr": 9.943281365174274e-05}
{"train_loss": 0.5782818794250488, "global_step": 13503, "epoch": 151, "lr": 9.943272657442575e-05}
{"train_loss": 0.710232675075531, "global_step": 13504, "epoch": 151, "lr": 9.943263949046315e-05}
{"train_loss": 0.6624635457992554, "global_step": 13505, "epoch": 151, "lr": 9.943255239985494e-05}
{"train_loss": 0.6078173518180847, "global_step": 13506, "epoch": 151, "lr": 9.943246530260113e-05}
{"train_loss": 0.7286421656608582, "global_step": 13507, "epoch": 151, "lr": 9.943237819870173e-05}
{"train_loss": 0.6057434678077698, "global_step": 13508, "epoch": 151, "lr": 9.943229108815673e-05}
{"train_loss": 0.7088345885276794, "global_step": 13509, "epoch": 151, "lr": 9.943220397096618e-05}
{"train_loss": 0.657616138458252, "global_step": 13510, "epoch": 151, "lr": 9.943211684713005e-05}
{"train_loss": 0.5860111117362976, "global_step": 13511, "epoch": 151, "lr": 9.94320297166484e-05}
{"train_loss": 0.7212076187133789, "global_step": 13512, "epoch": 151, "lr": 9.94319425795212e-05}
{"train_loss": 0.720662534236908, "global_step": 13513, "epoch": 151, "lr": 9.943185543574847e-05}
{"train_loss": 0.6244498491287231, "global_step": 13514, "epoch": 151, "lr": 9.943176828533026e-05}
{"train_loss": 0.7073508501052856, "global_step": 13515, "epoch": 151, "lr": 9.943168112826653e-05}
{"train_loss": 0.5521281361579895, "global_step": 13516, "epoch": 151, "lr": 9.94315939645573e-05}
{"train_loss": 0.6587174534797668, "global_step": 13517, "epoch": 151, "lr": 9.943150679420261e-05}
{"train_loss": 0.6780005693435669, "global_step": 13518, "epoch": 151, "lr": 9.943141961720245e-05}
{"train_loss": 0.5954350829124451, "global_step": 13519, "epoch": 151, "lr": 9.943133243355687e-05}
{"train_loss": 0.7548280358314514, "global_step": 13520, "epoch": 151, "lr": 9.943124524326582e-05}
{"train_loss": 0.5631695985794067, "global_step": 13521, "epoch": 151, "lr": 9.943115804632934e-05}
{"train_loss": 0.5494651198387146, "global_step": 13522, "epoch": 151, "lr": 9.943107084274746e-05}
{"train_loss": 0.6416320204734802, "global_step": 13523, "epoch": 151, "lr": 9.943098363252015e-05}
{"train_loss": 0.5681214928627014, "global_step": 13524, "epoch": 151, "lr": 9.943089641564747e-05}
{"train_loss": 0.5514723062515259, "global_step": 13525, "epoch": 151, "lr": 9.943080919212939e-05}
{"train_loss": 0.6348685622215271, "global_step": 13526, "epoch": 151, "lr": 9.943072196196596e-05}
{"train_loss": 0.6045234066047026, "global_step": 13527, "epoch": 151, "lr": 9.943063472515716e-05, "val_loss": 1.427273154258728}
{"train_loss": 0.5853363871574402, "global_step": 13528, "epoch": 152, "lr": 9.943054748170301e-05}
{"train_loss": 0.5733900666236877, "global_step": 13529, "epoch": 152, "lr": 9.943046023160353e-05}
{"train_loss": 0.6354489922523499, "global_step": 13530, "epoch": 152, "lr": 9.943037297485874e-05}
{"train_loss": 0.7668415904045105, "global_step": 13531, "epoch": 152, "lr": 9.943028571146862e-05}
{"train_loss": 0.599053680896759, "global_step": 13532, "epoch": 152, "lr": 9.94301984414332e-05}
{"train_loss": 0.6272480487823486, "global_step": 13533, "epoch": 152, "lr": 9.943011116475251e-05}
{"train_loss": 0.5939714312553406, "global_step": 13534, "epoch": 152, "lr": 9.943002388142653e-05}
{"train_loss": 0.6180961728096008, "global_step": 13535, "epoch": 152, "lr": 9.942993659145528e-05}
{"train_loss": 0.5697643756866455, "global_step": 13536, "epoch": 152, "lr": 9.94298492948388e-05}
{"train_loss": 0.5581956505775452, "global_step": 13537, "epoch": 152, "lr": 9.942976199157705e-05}
{"train_loss": 0.8216449618339539, "global_step": 13538, "epoch": 152, "lr": 9.942967468167011e-05}
{"train_loss": 0.5855487585067749, "global_step": 13539, "epoch": 152, "lr": 9.942958736511791e-05}
{"train_loss": 0.6045846939086914, "global_step": 13540, "epoch": 152, "lr": 9.942950004192054e-05}
{"train_loss": 0.5518319010734558, "global_step": 13541, "epoch": 152, "lr": 9.942941271207795e-05}
{"train_loss": 0.4970480799674988, "global_step": 13542, "epoch": 152, "lr": 9.942932537559019e-05}
{"train_loss": 0.7058980464935303, "global_step": 13543, "epoch": 152, "lr": 9.942923803245726e-05}
{"train_loss": 0.6870671510696411, "global_step": 13544, "epoch": 152, "lr": 9.942915068267917e-05}
{"train_loss": 0.6509461402893066, "global_step": 13545, "epoch": 152, "lr": 9.942906332625593e-05}
{"train_loss": 0.6652186512947083, "global_step": 13546, "epoch": 152, "lr": 9.942897596318756e-05}
{"train_loss": 0.6692116260528564, "global_step": 13547, "epoch": 152, "lr": 9.942888859347405e-05}
{"train_loss": 0.5429312586784363, "global_step": 13548, "epoch": 152, "lr": 9.942880121711545e-05}
{"train_loss": 0.5624464154243469, "global_step": 13549, "epoch": 152, "lr": 9.942871383411175e-05}
{"train_loss": 0.5219783186912537, "global_step": 13550, "epoch": 152, "lr": 9.942862644446295e-05}
{"train_loss": 0.6770830750465393, "global_step": 13551, "epoch": 152, "lr": 9.942853904816908e-05}
{"train_loss": 0.7063049077987671, "global_step": 13552, "epoch": 152, "lr": 9.942845164523014e-05}
{"train_loss": 0.8062846064567566, "global_step": 13553, "epoch": 152, "lr": 9.942836423564614e-05}
{"train_loss": 0.5814589858055115, "global_step": 13554, "epoch": 152, "lr": 9.942827681941712e-05}
{"train_loss": 0.4615274965763092, "global_step": 13555, "epoch": 152, "lr": 9.942818939654306e-05}
{"train_loss": 0.6716704368591309, "global_step": 13556, "epoch": 152, "lr": 9.942810196702399e-05}
{"train_loss": 0.5878735780715942, "global_step": 13557, "epoch": 152, "lr": 9.94280145308599e-05}
{"train_loss": 0.6335716843605042, "global_step": 13558, "epoch": 152, "lr": 9.942792708805081e-05}
{"train_loss": 0.6186692118644714, "global_step": 13559, "epoch": 152, "lr": 9.942783963859676e-05}
{"train_loss": 0.5049484968185425, "global_step": 13560, "epoch": 152, "lr": 9.942775218249773e-05}
{"train_loss": 0.46105077862739563, "global_step": 13561, "epoch": 152, "lr": 9.942766471975375e-05}
{"train_loss": 0.6420060396194458, "global_step": 13562, "epoch": 152, "lr": 9.94275772503648e-05}
{"train_loss": 0.5654969811439514, "global_step": 13563, "epoch": 152, "lr": 9.942748977433094e-05}
{"train_loss": 0.5852469205856323, "global_step": 13564, "epoch": 152, "lr": 9.942740229165213e-05}
{"train_loss": 0.7365542650222778, "global_step": 13565, "epoch": 152, "lr": 9.942731480232843e-05}
{"train_loss": 0.5981782078742981, "global_step": 13566, "epoch": 152, "lr": 9.942722730635983e-05}
{"train_loss": 0.7219750285148621, "global_step": 13567, "epoch": 152, "lr": 9.942713980374633e-05}
{"train_loss": 0.671049952507019, "global_step": 13568, "epoch": 152, "lr": 9.942705229448795e-05}
{"train_loss": 0.596796452999115, "global_step": 13569, "epoch": 152, "lr": 9.942696477858472e-05}
{"train_loss": 0.6459166407585144, "global_step": 13570, "epoch": 152, "lr": 9.942687725603663e-05}
{"train_loss": 0.7351880669593811, "global_step": 13571, "epoch": 152, "lr": 9.94267897268437e-05}
{"train_loss": 0.6946818232536316, "global_step": 13572, "epoch": 152, "lr": 9.942670219100594e-05}
{"train_loss": 0.5458057522773743, "global_step": 13573, "epoch": 152, "lr": 9.942661464852337e-05}
{"train_loss": 0.6809813976287842, "global_step": 13574, "epoch": 152, "lr": 9.942652709939597e-05}
{"train_loss": 0.6442312598228455, "global_step": 13575, "epoch": 152, "lr": 9.94264395436238e-05}
{"train_loss": 0.6870443224906921, "global_step": 13576, "epoch": 152, "lr": 9.942635198120684e-05}
{"train_loss": 0.6846495866775513, "global_step": 13577, "epoch": 152, "lr": 9.942626441214511e-05}
{"train_loss": 0.9688299298286438, "global_step": 13578, "epoch": 152, "lr": 9.942617683643862e-05}
{"train_loss": 0.7114081978797913, "global_step": 13579, "epoch": 152, "lr": 9.942608925408739e-05}
{"train_loss": 0.5733686685562134, "global_step": 13580, "epoch": 152, "lr": 9.942600166509141e-05}
{"train_loss": 0.6377924680709839, "global_step": 13581, "epoch": 152, "lr": 9.942591406945071e-05}
{"train_loss": 0.7539606690406799, "global_step": 13582, "epoch": 152, "lr": 9.94258264671653e-05}
{"train_loss": 0.6074037551879883, "global_step": 13583, "epoch": 152, "lr": 9.942573885823519e-05}
{"train_loss": 0.5720851421356201, "global_step": 13584, "epoch": 152, "lr": 9.94256512426604e-05}
{"train_loss": 0.5739984512329102, "global_step": 13585, "epoch": 152, "lr": 9.942556362044093e-05}
{"train_loss": 0.586815595626831, "global_step": 13586, "epoch": 152, "lr": 9.942547599157679e-05}
{"train_loss": 0.5413727164268494, "global_step": 13587, "epoch": 152, "lr": 9.942538835606801e-05}
{"train_loss": 0.7431423664093018, "global_step": 13588, "epoch": 152, "lr": 9.942530071391458e-05}
{"train_loss": 0.8508545160293579, "global_step": 13589, "epoch": 152, "lr": 9.942521306511651e-05}
{"train_loss": 0.5913718342781067, "global_step": 13590, "epoch": 152, "lr": 9.942512540967383e-05}
{"train_loss": 0.5900917649269104, "global_step": 13591, "epoch": 152, "lr": 9.942503774758656e-05}
{"train_loss": 0.8271145820617676, "global_step": 13592, "epoch": 152, "lr": 9.942495007885468e-05}
{"train_loss": 0.7713762521743774, "global_step": 13593, "epoch": 152, "lr": 9.942486240347821e-05}
{"train_loss": 0.6807543039321899, "global_step": 13594, "epoch": 152, "lr": 9.94247747214572e-05}
{"train_loss": 0.7683120965957642, "global_step": 13595, "epoch": 152, "lr": 9.94246870327916e-05}
{"train_loss": 0.6240817904472351, "global_step": 13596, "epoch": 152, "lr": 9.942459933748146e-05}
{"train_loss": 0.48396825790405273, "global_step": 13597, "epoch": 152, "lr": 9.94245116355268e-05}
{"train_loss": 0.6467258334159851, "global_step": 13598, "epoch": 152, "lr": 9.94244239269276e-05}
{"train_loss": 0.6201865077018738, "global_step": 13599, "epoch": 152, "lr": 9.94243362116839e-05}
{"train_loss": 0.5609994530677795, "global_step": 13600, "epoch": 152, "lr": 9.942424848979569e-05}
{"train_loss": 0.5372664928436279, "global_step": 13601, "epoch": 152, "lr": 9.9424160761263e-05}
{"train_loss": 0.6129701137542725, "global_step": 13602, "epoch": 152, "lr": 9.942407302608583e-05}
{"train_loss": 0.582782506942749, "global_step": 13603, "epoch": 152, "lr": 9.94239852842642e-05}
{"train_loss": 0.7353234887123108, "global_step": 13604, "epoch": 152, "lr": 9.942389753579811e-05}
{"train_loss": 0.6656762957572937, "global_step": 13605, "epoch": 152, "lr": 9.942380978068759e-05}
{"train_loss": 0.872902512550354, "global_step": 13606, "epoch": 152, "lr": 9.942372201893264e-05}
{"train_loss": 0.7398218512535095, "global_step": 13607, "epoch": 152, "lr": 9.942363425053326e-05}
{"train_loss": 0.6166315674781799, "global_step": 13608, "epoch": 152, "lr": 9.942354647548948e-05}
{"train_loss": 0.7285659909248352, "global_step": 13609, "epoch": 152, "lr": 9.942345869380133e-05}
{"train_loss": 0.473470538854599, "global_step": 13610, "epoch": 152, "lr": 9.942337090546877e-05}
{"train_loss": 0.6424199342727661, "global_step": 13611, "epoch": 152, "lr": 9.942328311049185e-05}
{"train_loss": 0.7850369215011597, "global_step": 13612, "epoch": 152, "lr": 9.942319530887057e-05}
{"train_loss": 0.6336405873298645, "global_step": 13613, "epoch": 152, "lr": 9.942310750060495e-05}
{"train_loss": 0.6219096183776855, "global_step": 13614, "epoch": 152, "lr": 9.942301968569498e-05}
{"train_loss": 0.7731919884681702, "global_step": 13615, "epoch": 152, "lr": 9.94229318641407e-05}
{"train_loss": 0.6429560428924774, "global_step": 13616, "epoch": 152, "lr": 9.942284403594212e-05, "val_loss": 1.4257566928863525}
{"train_loss": 0.6299245953559875, "global_step": 13617, "epoch": 153, "lr": 9.942275620109923e-05}
{"train_loss": 0.540023148059845, "global_step": 13618, "epoch": 153, "lr": 9.942266835961204e-05}
{"train_loss": 0.5180900692939758, "global_step": 13619, "epoch": 153, "lr": 9.942258051148059e-05}
{"train_loss": 0.5322583913803101, "global_step": 13620, "epoch": 153, "lr": 9.942249265670487e-05}
{"train_loss": 0.5631978511810303, "global_step": 13621, "epoch": 153, "lr": 9.94224047952849e-05}
{"train_loss": 0.7347480654716492, "global_step": 13622, "epoch": 153, "lr": 9.942231692722068e-05}
{"train_loss": 0.6385297775268555, "global_step": 13623, "epoch": 153, "lr": 9.942222905251224e-05}
{"train_loss": 0.5176061391830444, "global_step": 13624, "epoch": 153, "lr": 9.942214117115959e-05}
{"train_loss": 0.6521024703979492, "global_step": 13625, "epoch": 153, "lr": 9.942205328316272e-05}
{"train_loss": 0.6914035677909851, "global_step": 13626, "epoch": 153, "lr": 9.942196538852168e-05}
{"train_loss": 0.6582610011100769, "global_step": 13627, "epoch": 153, "lr": 9.942187748723645e-05}
{"train_loss": 0.8273605108261108, "global_step": 13628, "epoch": 153, "lr": 9.942178957930704e-05}
{"train_loss": 0.7191635370254517, "global_step": 13629, "epoch": 153, "lr": 9.942170166473349e-05}
{"train_loss": 0.5541680455207825, "global_step": 13630, "epoch": 153, "lr": 9.942161374351579e-05}
{"train_loss": 0.6775942444801331, "global_step": 13631, "epoch": 153, "lr": 9.942152581565394e-05}
{"train_loss": 0.7739472985267639, "global_step": 13632, "epoch": 153, "lr": 9.942143788114799e-05}
{"train_loss": 0.9358692765235901, "global_step": 13633, "epoch": 153, "lr": 9.942134993999793e-05}
{"train_loss": 0.8998631834983826, "global_step": 13634, "epoch": 153, "lr": 9.942126199220374e-05}
{"train_loss": 0.6997575163841248, "global_step": 13635, "epoch": 153, "lr": 9.942117403776549e-05}
{"train_loss": 0.7425593733787537, "global_step": 13636, "epoch": 153, "lr": 9.942108607668317e-05}
{"train_loss": 0.7243074178695679, "global_step": 13637, "epoch": 153, "lr": 9.942099810895678e-05}
{"train_loss": 0.6844400763511658, "global_step": 13638, "epoch": 153, "lr": 9.942091013458633e-05}
{"train_loss": 0.6292934417724609, "global_step": 13639, "epoch": 153, "lr": 9.942082215357185e-05}
{"train_loss": 0.6018838286399841, "global_step": 13640, "epoch": 153, "lr": 9.942073416591335e-05}
{"train_loss": 0.7723897695541382, "global_step": 13641, "epoch": 153, "lr": 9.942064617161083e-05}
{"train_loss": 0.520986020565033, "global_step": 13642, "epoch": 153, "lr": 9.942055817066429e-05}
{"train_loss": 0.6611068844795227, "global_step": 13643, "epoch": 153, "lr": 9.942047016307378e-05}
{"train_loss": 0.5630170106887817, "global_step": 13644, "epoch": 153, "lr": 9.942038214883927e-05}
{"train_loss": 0.5296300053596497, "global_step": 13645, "epoch": 153, "lr": 9.942029412796082e-05}
{"train_loss": 0.5918265581130981, "global_step": 13646, "epoch": 153, "lr": 9.942020610043839e-05}
{"train_loss": 0.7180578708648682, "global_step": 13647, "epoch": 153, "lr": 9.942011806627202e-05}
{"train_loss": 0.570609986782074, "global_step": 13648, "epoch": 153, "lr": 9.942003002546172e-05}
{"train_loss": 0.5312365889549255, "global_step": 13649, "epoch": 153, "lr": 9.94199419780075e-05}
{"train_loss": 0.6163232922554016, "global_step": 13650, "epoch": 153, "lr": 9.941985392390937e-05}
{"train_loss": 0.5595827102661133, "global_step": 13651, "epoch": 153, "lr": 9.941976586316735e-05}
{"train_loss": 0.768671989440918, "global_step": 13652, "epoch": 153, "lr": 9.941967779578143e-05}
{"train_loss": 0.6287625432014465, "global_step": 13653, "epoch": 153, "lr": 9.941958972175165e-05}
{"train_loss": 0.634726345539093, "global_step": 13654, "epoch": 153, "lr": 9.9419501641078e-05}
{"train_loss": 0.6309022903442383, "global_step": 13655, "epoch": 153, "lr": 9.94194135537605e-05}
{"train_loss": 0.5029613375663757, "global_step": 13656, "epoch": 153, "lr": 9.941932545979917e-05}
{"train_loss": 0.4777044653892517, "global_step": 13657, "epoch": 153, "lr": 9.9419237359194e-05}
{"train_loss": 0.6064292192459106, "global_step": 13658, "epoch": 153, "lr": 9.941914925194504e-05}
{"train_loss": 0.6810335516929626, "global_step": 13659, "epoch": 153, "lr": 9.941906113805227e-05}
{"train_loss": 0.5540729761123657, "global_step": 13660, "epoch": 153, "lr": 9.94189730175157e-05}
{"train_loss": 0.6661255955696106, "global_step": 13661, "epoch": 153, "lr": 9.941888489033535e-05}
{"train_loss": 0.5864666104316711, "global_step": 13662, "epoch": 153, "lr": 9.941879675651123e-05}
{"train_loss": 0.7992863655090332, "global_step": 13663, "epoch": 153, "lr": 9.941870861604337e-05}
{"train_loss": 0.5070905685424805, "global_step": 13664, "epoch": 153, "lr": 9.941862046893176e-05}
{"train_loss": 0.6293894052505493, "global_step": 13665, "epoch": 153, "lr": 9.941853231517641e-05}
{"train_loss": 0.477945476770401, "global_step": 13666, "epoch": 153, "lr": 9.941844415477736e-05}
{"train_loss": 0.5824242234230042, "global_step": 13667, "epoch": 153, "lr": 9.941835598773459e-05}
{"train_loss": 0.618902325630188, "global_step": 13668, "epoch": 153, "lr": 9.941826781404812e-05}
{"train_loss": 0.5891287326812744, "global_step": 13669, "epoch": 153, "lr": 9.941817963371798e-05}
{"train_loss": 0.8037896156311035, "global_step": 13670, "epoch": 153, "lr": 9.941809144674417e-05}
{"train_loss": 0.6305240988731384, "global_step": 13671, "epoch": 153, "lr": 9.941800325312668e-05}
{"train_loss": 0.5182113647460938, "global_step": 13672, "epoch": 153, "lr": 9.941791505286555e-05}
{"train_loss": 0.47156500816345215, "global_step": 13673, "epoch": 153, "lr": 9.941782684596079e-05}
{"train_loss": 0.6178796291351318, "global_step": 13674, "epoch": 153, "lr": 9.94177386324124e-05}
{"train_loss": 0.8240022659301758, "global_step": 13675, "epoch": 153, "lr": 9.94176504122204e-05}
{"train_loss": 0.6494477987289429, "global_step": 13676, "epoch": 153, "lr": 9.941756218538479e-05}
{"train_loss": 0.587966799736023, "global_step": 13677, "epoch": 153, "lr": 9.941747395190561e-05}
{"train_loss": 0.5652452707290649, "global_step": 13678, "epoch": 153, "lr": 9.941738571178284e-05}
{"train_loss": 0.6672075390815735, "global_step": 13679, "epoch": 153, "lr": 9.941729746501651e-05}
{"train_loss": 0.6195806264877319, "global_step": 13680, "epoch": 153, "lr": 9.941720921160662e-05}
{"train_loss": 0.688874363899231, "global_step": 13681, "epoch": 153, "lr": 9.941712095155319e-05}
{"train_loss": 0.5538798570632935, "global_step": 13682, "epoch": 153, "lr": 9.941703268485623e-05}
{"train_loss": 0.6370699405670166, "global_step": 13683, "epoch": 153, "lr": 9.941694441151576e-05}
{"train_loss": 0.7300171256065369, "global_step": 13684, "epoch": 153, "lr": 9.941685613153178e-05}
{"train_loss": 0.6685728430747986, "global_step": 13685, "epoch": 153, "lr": 9.94167678449043e-05}
{"train_loss": 0.6395223140716553, "global_step": 13686, "epoch": 153, "lr": 9.941667955163335e-05}
{"train_loss": 0.6241533160209656, "global_step": 13687, "epoch": 153, "lr": 9.941659125171892e-05}
{"train_loss": 0.5374966263771057, "global_step": 13688, "epoch": 153, "lr": 9.941650294516102e-05}
{"train_loss": 0.6819844245910645, "global_step": 13689, "epoch": 153, "lr": 9.941641463195969e-05}
{"train_loss": 0.626453697681427, "global_step": 13690, "epoch": 153, "lr": 9.941632631211493e-05}
{"train_loss": 0.6655619144439697, "global_step": 13691, "epoch": 153, "lr": 9.941623798562674e-05}
{"train_loss": 0.5615624785423279, "global_step": 13692, "epoch": 153, "lr": 9.941614965249515e-05}
{"train_loss": 0.7118780612945557, "global_step": 13693, "epoch": 153, "lr": 9.941606131272014e-05}
{"train_loss": 0.6285533308982849, "global_step": 13694, "epoch": 153, "lr": 9.941597296630176e-05}
{"train_loss": 0.49463173747062683, "global_step": 13695, "epoch": 153, "lr": 9.941588461323998e-05}
{"train_loss": 0.5499227643013, "global_step": 13696, "epoch": 153, "lr": 9.941579625353486e-05}
{"train_loss": 0.6628327965736389, "global_step": 13697, "epoch": 153, "lr": 9.941570788718639e-05}
{"train_loss": 0.5933467745780945, "global_step": 13698, "epoch": 153, "lr": 9.941561951419458e-05}
{"train_loss": 0.6748454570770264, "global_step": 13699, "epoch": 153, "lr": 9.941553113455944e-05}
{"train_loss": 0.48839524388313293, "global_step": 13700, "epoch": 153, "lr": 9.941544274828096e-05}
{"train_loss": 0.6468921303749084, "global_step": 13701, "epoch": 153, "lr": 9.94153543553592e-05}
{"train_loss": 0.6563791036605835, "global_step": 13702, "epoch": 153, "lr": 9.941526595579415e-05}
{"train_loss": 0.6035645604133606, "global_step": 13703, "epoch": 153, "lr": 9.941517754958582e-05}
{"train_loss": 0.6132163405418396, "global_step": 13704, "epoch": 153, "lr": 9.94150891367342e-05}
{"train_loss": 0.6314385382646925, "global_step": 13705, "epoch": 153, "lr": 9.941500071723935e-05, "val_loss": 1.4022982120513916}
{"train_loss": 0.5603166222572327, "global_step": 13706, "epoch": 154, "lr": 9.941491229110124e-05}
{"train_loss": 0.37138959765434265, "global_step": 13707, "epoch": 154, "lr": 9.941482385831991e-05}
{"train_loss": 0.5603221654891968, "global_step": 13708, "epoch": 154, "lr": 9.941473541889534e-05}
{"train_loss": 0.6473448276519775, "global_step": 13709, "epoch": 154, "lr": 9.941464697282757e-05}
{"train_loss": 0.5738959908485413, "global_step": 13710, "epoch": 154, "lr": 9.941455852011661e-05}
{"train_loss": 0.6009849309921265, "global_step": 13711, "epoch": 154, "lr": 9.941447006076246e-05}
{"train_loss": 0.6030243039131165, "global_step": 13712, "epoch": 154, "lr": 9.941438159476513e-05}
{"train_loss": 0.6605542302131653, "global_step": 13713, "epoch": 154, "lr": 9.941429312212464e-05}
{"train_loss": 0.5648161172866821, "global_step": 13714, "epoch": 154, "lr": 9.941420464284101e-05}
{"train_loss": 0.7717193961143494, "global_step": 13715, "epoch": 154, "lr": 9.941411615691425e-05}
{"train_loss": 0.6785665154457092, "global_step": 13716, "epoch": 154, "lr": 9.941402766434435e-05}
{"train_loss": 0.6027352809906006, "global_step": 13717, "epoch": 154, "lr": 9.941393916513133e-05}
{"train_loss": 0.5384091138839722, "global_step": 13718, "epoch": 154, "lr": 9.941385065927522e-05}
{"train_loss": 0.4838680028915405, "global_step": 13719, "epoch": 154, "lr": 9.941376214677602e-05}
{"train_loss": 0.7587876915931702, "global_step": 13720, "epoch": 154, "lr": 9.941367362763374e-05}
{"train_loss": 0.6087764501571655, "global_step": 13721, "epoch": 154, "lr": 9.941358510184839e-05}
{"train_loss": 0.5560932755470276, "global_step": 13722, "epoch": 154, "lr": 9.941349656941999e-05}
{"train_loss": 0.5497446656227112, "global_step": 13723, "epoch": 154, "lr": 9.941340803034854e-05}
{"train_loss": 0.6081218123435974, "global_step": 13724, "epoch": 154, "lr": 9.941331948463407e-05}
{"train_loss": 0.5489251017570496, "global_step": 13725, "epoch": 154, "lr": 9.941323093227658e-05}
{"train_loss": 0.5799352526664734, "global_step": 13726, "epoch": 154, "lr": 9.941314237327608e-05}
{"train_loss": 0.6295547485351562, "global_step": 13727, "epoch": 154, "lr": 9.941305380763259e-05}
{"train_loss": 0.5827124714851379, "global_step": 13728, "epoch": 154, "lr": 9.941296523534611e-05}
{"train_loss": 0.6020818948745728, "global_step": 13729, "epoch": 154, "lr": 9.941287665641665e-05}
{"train_loss": 0.618073582649231, "global_step": 13730, "epoch": 154, "lr": 9.941278807084425e-05}
{"train_loss": 0.6641348004341125, "global_step": 13731, "epoch": 154, "lr": 9.94126994786289e-05}
{"train_loss": 0.7084605693817139, "global_step": 13732, "epoch": 154, "lr": 9.941261087977061e-05}
{"train_loss": 0.6765753030776978, "global_step": 13733, "epoch": 154, "lr": 9.94125222742694e-05}
{"train_loss": 0.7008340954780579, "global_step": 13734, "epoch": 154, "lr": 9.941243366212527e-05}
{"train_loss": 0.6013873219490051, "global_step": 13735, "epoch": 154, "lr": 9.941234504333825e-05}
{"train_loss": 0.45562365651130676, "global_step": 13736, "epoch": 154, "lr": 9.941225641790834e-05}
{"train_loss": 0.5182456374168396, "global_step": 13737, "epoch": 154, "lr": 9.941216778583555e-05}
{"train_loss": 0.6056289076805115, "global_step": 13738, "epoch": 154, "lr": 9.941207914711989e-05}
{"train_loss": 0.7522456645965576, "global_step": 13739, "epoch": 154, "lr": 9.94119905017614e-05}
{"train_loss": 0.5858533978462219, "global_step": 13740, "epoch": 154, "lr": 9.941190184976006e-05}
{"train_loss": 0.6146001219749451, "global_step": 13741, "epoch": 154, "lr": 9.941181319111589e-05}
{"train_loss": 0.662559449672699, "global_step": 13742, "epoch": 154, "lr": 9.94117245258289e-05}
{"train_loss": 0.5558279156684875, "global_step": 13743, "epoch": 154, "lr": 9.941163585389912e-05}
{"train_loss": 0.5961831212043762, "global_step": 13744, "epoch": 154, "lr": 9.941154717532652e-05}
{"train_loss": 0.6541336178779602, "global_step": 13745, "epoch": 154, "lr": 9.941145849011117e-05}
{"train_loss": 0.6323493123054504, "global_step": 13746, "epoch": 154, "lr": 9.941136979825303e-05}
{"train_loss": 0.6720966696739197, "global_step": 13747, "epoch": 154, "lr": 9.941128109975215e-05}
{"train_loss": 0.6189587116241455, "global_step": 13748, "epoch": 154, "lr": 9.941119239460852e-05}
{"train_loss": 0.5907944440841675, "global_step": 13749, "epoch": 154, "lr": 9.941110368282215e-05}
{"train_loss": 0.629816472530365, "global_step": 13750, "epoch": 154, "lr": 9.941101496439306e-05}
{"train_loss": 0.5511384606361389, "global_step": 13751, "epoch": 154, "lr": 9.941092623932127e-05}
{"train_loss": 0.5616399645805359, "global_step": 13752, "epoch": 154, "lr": 9.941083750760678e-05}
{"train_loss": 0.7186911106109619, "global_step": 13753, "epoch": 154, "lr": 9.94107487692496e-05}
{"train_loss": 0.5488228797912598, "global_step": 13754, "epoch": 154, "lr": 9.941066002424976e-05}
{"train_loss": 0.5775450468063354, "global_step": 13755, "epoch": 154, "lr": 9.941057127260725e-05}
{"train_loss": 0.6523171663284302, "global_step": 13756, "epoch": 154, "lr": 9.941048251432209e-05}
{"train_loss": 0.5614718198776245, "global_step": 13757, "epoch": 154, "lr": 9.94103937493943e-05}
{"train_loss": 0.7147828936576843, "global_step": 13758, "epoch": 154, "lr": 9.941030497782388e-05}
{"train_loss": 0.68596351146698, "global_step": 13759, "epoch": 154, "lr": 9.941021619961084e-05}
{"train_loss": 0.6814268231391907, "global_step": 13760, "epoch": 154, "lr": 9.94101274147552e-05}
{"train_loss": 0.5763244032859802, "global_step": 13761, "epoch": 154, "lr": 9.941003862325698e-05}
{"train_loss": 0.5866795182228088, "global_step": 13762, "epoch": 154, "lr": 9.940994982511618e-05}
{"train_loss": 0.49009376764297485, "global_step": 13763, "epoch": 154, "lr": 9.94098610203328e-05}
{"train_loss": 0.48039910197257996, "global_step": 13764, "epoch": 154, "lr": 9.940977220890688e-05}
{"train_loss": 0.49757784605026245, "global_step": 13765, "epoch": 154, "lr": 9.940968339083842e-05}
{"train_loss": 0.562997043132782, "global_step": 13766, "epoch": 154, "lr": 9.940959456612743e-05}
{"train_loss": 0.780828058719635, "global_step": 13767, "epoch": 154, "lr": 9.940950573477392e-05}
{"train_loss": 0.5870494842529297, "global_step": 13768, "epoch": 154, "lr": 9.94094168967779e-05}
{"train_loss": 0.7658599019050598, "global_step": 13769, "epoch": 154, "lr": 9.940932805213939e-05}
{"train_loss": 0.6056524515151978, "global_step": 13770, "epoch": 154, "lr": 9.94092392008584e-05}
{"train_loss": 0.6304953098297119, "global_step": 13771, "epoch": 154, "lr": 9.940915034293494e-05}
{"train_loss": 0.5650491714477539, "global_step": 13772, "epoch": 154, "lr": 9.940906147836901e-05}
{"train_loss": 0.5649498701095581, "global_step": 13773, "epoch": 154, "lr": 9.940897260716066e-05}
{"train_loss": 0.6115105748176575, "global_step": 13774, "epoch": 154, "lr": 9.940888372930986e-05}
{"train_loss": 0.5994685888290405, "global_step": 13775, "epoch": 154, "lr": 9.940879484481663e-05}
{"train_loss": 0.7824141979217529, "global_step": 13776, "epoch": 154, "lr": 9.940870595368099e-05}
{"train_loss": 0.6426942348480225, "global_step": 13777, "epoch": 154, "lr": 9.940861705590296e-05}
{"train_loss": 0.6398224234580994, "global_step": 13778, "epoch": 154, "lr": 9.940852815148254e-05}
{"train_loss": 0.534000039100647, "global_step": 13779, "epoch": 154, "lr": 9.940843924041976e-05}
{"train_loss": 0.7076337337493896, "global_step": 13780, "epoch": 154, "lr": 9.940835032271461e-05}
{"train_loss": 0.6502097249031067, "global_step": 13781, "epoch": 154, "lr": 9.94082613983671e-05}
{"train_loss": 0.608130693435669, "global_step": 13782, "epoch": 154, "lr": 9.940817246737726e-05}
{"train_loss": 0.5437761545181274, "global_step": 13783, "epoch": 154, "lr": 9.940808352974509e-05}
{"train_loss": 0.5778810977935791, "global_step": 13784, "epoch": 154, "lr": 9.94079945854706e-05}
{"train_loss": 0.6511337161064148, "global_step": 13785, "epoch": 154, "lr": 9.94079056345538e-05}
{"train_loss": 0.5498049259185791, "global_step": 13786, "epoch": 154, "lr": 9.940781667699473e-05}
{"train_loss": 0.6839998960494995, "global_step": 13787, "epoch": 154, "lr": 9.940772771279338e-05}
{"train_loss": 0.5710998773574829, "global_step": 13788, "epoch": 154, "lr": 9.940763874194976e-05}
{"train_loss": 0.6112064719200134, "global_step": 13789, "epoch": 154, "lr": 9.940754976446387e-05}
{"train_loss": 0.6664139628410339, "global_step": 13790, "epoch": 154, "lr": 9.940746078033575e-05}
{"train_loss": 0.689147412776947, "global_step": 13791, "epoch": 154, "lr": 9.940737178956541e-05}
{"train_loss": 0.6135662794113159, "global_step": 13792, "epoch": 154, "lr": 9.940728279215283e-05}
{"train_loss": 0.7433915138244629, "global_step": 13793, "epoch": 154, "lr": 9.940719378809806e-05}
{"train_loss": 0.6134892718845539, "global_step": 13794, "epoch": 154, "lr": 9.940710477740108e-05, "val_loss": 1.3914778232574463}
{"train_loss": 0.610875129699707, "global_step": 13795, "epoch": 155, "lr": 9.940701576006192e-05}
{"train_loss": 0.6898494362831116, "global_step": 13796, "epoch": 155, "lr": 9.940692673608058e-05}
{"train_loss": 0.5350165963172913, "global_step": 13797, "epoch": 155, "lr": 9.940683770545712e-05}
{"train_loss": 0.5147709846496582, "global_step": 13798, "epoch": 155, "lr": 9.940674866819148e-05}
{"train_loss": 0.5852880477905273, "global_step": 13799, "epoch": 155, "lr": 9.94066596242837e-05}
{"train_loss": 0.6164711713790894, "global_step": 13800, "epoch": 155, "lr": 9.940657057373381e-05}
{"train_loss": 0.7472186088562012, "global_step": 13801, "epoch": 155, "lr": 9.940648151654181e-05}
{"train_loss": 0.5913788080215454, "global_step": 13802, "epoch": 155, "lr": 9.94063924527077e-05}
{"train_loss": 0.6305068731307983, "global_step": 13803, "epoch": 155, "lr": 9.940630338223152e-05}
{"train_loss": 0.6826329827308655, "global_step": 13804, "epoch": 155, "lr": 9.940621430511325e-05}
{"train_loss": 0.5150843262672424, "global_step": 13805, "epoch": 155, "lr": 9.940612522135293e-05}
{"train_loss": 0.46184778213500977, "global_step": 13806, "epoch": 155, "lr": 9.940603613095054e-05}
{"train_loss": 0.588674783706665, "global_step": 13807, "epoch": 155, "lr": 9.940594703390613e-05}
{"train_loss": 0.6831550002098083, "global_step": 13808, "epoch": 155, "lr": 9.940585793021968e-05}
{"train_loss": 0.6323458552360535, "global_step": 13809, "epoch": 155, "lr": 9.940576881989122e-05}
{"train_loss": 0.4419075548648834, "global_step": 13810, "epoch": 155, "lr": 9.940567970292075e-05}
{"train_loss": 0.6109268069267273, "global_step": 13811, "epoch": 155, "lr": 9.940559057930829e-05}
{"train_loss": 0.6232966780662537, "global_step": 13812, "epoch": 155, "lr": 9.940550144905384e-05}
{"train_loss": 0.5652402639389038, "global_step": 13813, "epoch": 155, "lr": 9.940541231215745e-05}
{"train_loss": 0.6908808350563049, "global_step": 13814, "epoch": 155, "lr": 9.940532316861909e-05}
{"train_loss": 0.541479766368866, "global_step": 13815, "epoch": 155, "lr": 9.94052340184388e-05}
{"train_loss": 0.5414465665817261, "global_step": 13816, "epoch": 155, "lr": 9.940514486161656e-05}
{"train_loss": 0.5776099562644958, "global_step": 13817, "epoch": 155, "lr": 9.940505569815239e-05}
{"train_loss": 0.6002456545829773, "global_step": 13818, "epoch": 155, "lr": 9.940496652804634e-05}
{"train_loss": 0.5544978976249695, "global_step": 13819, "epoch": 155, "lr": 9.940487735129838e-05}
{"train_loss": 0.4212068021297455, "global_step": 13820, "epoch": 155, "lr": 9.940478816790854e-05}
{"train_loss": 0.6919429302215576, "global_step": 13821, "epoch": 155, "lr": 9.940469897787683e-05}
{"train_loss": 0.8285929560661316, "global_step": 13822, "epoch": 155, "lr": 9.940460978120325e-05}
{"train_loss": 0.4761735498905182, "global_step": 13823, "epoch": 155, "lr": 9.940452057788784e-05}
{"train_loss": 0.47785684466362, "global_step": 13824, "epoch": 155, "lr": 9.940443136793058e-05}
{"train_loss": 0.5941246151924133, "global_step": 13825, "epoch": 155, "lr": 9.94043421513315e-05}
{"train_loss": 0.5923591256141663, "global_step": 13826, "epoch": 155, "lr": 9.940425292809061e-05}
{"train_loss": 0.5594936013221741, "global_step": 13827, "epoch": 155, "lr": 9.940416369820793e-05}
{"train_loss": 0.6579028367996216, "global_step": 13828, "epoch": 155, "lr": 9.940407446168345e-05}
{"train_loss": 0.7718878388404846, "global_step": 13829, "epoch": 155, "lr": 9.94039852185172e-05}
{"train_loss": 0.5854312181472778, "global_step": 13830, "epoch": 155, "lr": 9.940389596870919e-05}
{"train_loss": 0.9132089018821716, "global_step": 13831, "epoch": 155, "lr": 9.940380671225941e-05}
{"train_loss": 0.7004541158676147, "global_step": 13832, "epoch": 155, "lr": 9.94037174491679e-05}
{"train_loss": 0.7040921449661255, "global_step": 13833, "epoch": 155, "lr": 9.940362817943465e-05}
{"train_loss": 0.6595114469528198, "global_step": 13834, "epoch": 155, "lr": 9.94035389030597e-05}
{"train_loss": 0.6048663854598999, "global_step": 13835, "epoch": 155, "lr": 9.940344962004306e-05}
{"train_loss": 0.5756694078445435, "global_step": 13836, "epoch": 155, "lr": 9.940336033038471e-05}
{"train_loss": 0.6938537359237671, "global_step": 13837, "epoch": 155, "lr": 9.940327103408467e-05}
{"train_loss": 0.7140610814094543, "global_step": 13838, "epoch": 155, "lr": 9.940318173114298e-05}
{"train_loss": 0.6917710900306702, "global_step": 13839, "epoch": 155, "lr": 9.940309242155963e-05}
{"train_loss": 0.6937719583511353, "global_step": 13840, "epoch": 155, "lr": 9.940300310533463e-05}
{"train_loss": 0.5444929003715515, "global_step": 13841, "epoch": 155, "lr": 9.940291378246802e-05}
{"train_loss": 0.5149694681167603, "global_step": 13842, "epoch": 155, "lr": 9.940282445295977e-05}
{"train_loss": 0.6054535508155823, "global_step": 13843, "epoch": 155, "lr": 9.940273511680992e-05}
{"train_loss": 0.6710370182991028, "global_step": 13844, "epoch": 155, "lr": 9.940264577401848e-05}
{"train_loss": 0.5588517189025879, "global_step": 13845, "epoch": 155, "lr": 9.940255642458544e-05}
{"train_loss": 0.555570662021637, "global_step": 13846, "epoch": 155, "lr": 9.940246706851084e-05}
{"train_loss": 0.5913767218589783, "global_step": 13847, "epoch": 155, "lr": 9.940237770579467e-05}
{"train_loss": 0.6176983714103699, "global_step": 13848, "epoch": 155, "lr": 9.940228833643697e-05}
{"train_loss": 0.6774240136146545, "global_step": 13849, "epoch": 155, "lr": 9.940219896043772e-05}
{"train_loss": 0.5706067085266113, "global_step": 13850, "epoch": 155, "lr": 9.940210957779694e-05}
{"train_loss": 0.5984339714050293, "global_step": 13851, "epoch": 155, "lr": 9.940202018851466e-05}
{"train_loss": 0.5086416006088257, "global_step": 13852, "epoch": 155, "lr": 9.940193079259088e-05}
{"train_loss": 0.5115144848823547, "global_step": 13853, "epoch": 155, "lr": 9.940184139002562e-05}
{"train_loss": 0.6522428393363953, "global_step": 13854, "epoch": 155, "lr": 9.940175198081887e-05}
{"train_loss": 0.6156498193740845, "global_step": 13855, "epoch": 155, "lr": 9.940166256497067e-05}
{"train_loss": 0.585472822189331, "global_step": 13856, "epoch": 155, "lr": 9.9401573142481e-05}
{"train_loss": 0.7242781519889832, "global_step": 13857, "epoch": 155, "lr": 9.940148371334991e-05}
{"train_loss": 0.5997254848480225, "global_step": 13858, "epoch": 155, "lr": 9.940139427757738e-05}
{"train_loss": 0.604038417339325, "global_step": 13859, "epoch": 155, "lr": 9.940130483516344e-05}
{"train_loss": 0.5635854005813599, "global_step": 13860, "epoch": 155, "lr": 9.94012153861081e-05}
{"train_loss": 0.7633767127990723, "global_step": 13861, "epoch": 155, "lr": 9.940112593041137e-05}
{"train_loss": 0.6706255078315735, "global_step": 13862, "epoch": 155, "lr": 9.940103646807325e-05}
{"train_loss": 0.5314085483551025, "global_step": 13863, "epoch": 155, "lr": 9.940094699909377e-05}
{"train_loss": 0.629707396030426, "global_step": 13864, "epoch": 155, "lr": 9.940085752347293e-05}
{"train_loss": 0.4893994927406311, "global_step": 13865, "epoch": 155, "lr": 9.940076804121074e-05}
{"train_loss": 0.5620995163917542, "global_step": 13866, "epoch": 155, "lr": 9.940067855230724e-05}
{"train_loss": 0.5989011526107788, "global_step": 13867, "epoch": 155, "lr": 9.940058905676241e-05}
{"train_loss": 0.5775817632675171, "global_step": 13868, "epoch": 155, "lr": 9.940049955457627e-05}
{"train_loss": 0.48894578218460083, "global_step": 13869, "epoch": 155, "lr": 9.940041004574884e-05}
{"train_loss": 0.6360301971435547, "global_step": 13870, "epoch": 155, "lr": 9.940032053028012e-05}
{"train_loss": 0.5464376211166382, "global_step": 13871, "epoch": 155, "lr": 9.940023100817013e-05}
{"train_loss": 0.6065423488616943, "global_step": 13872, "epoch": 155, "lr": 9.94001414794189e-05}
{"train_loss": 0.49124276638031006, "global_step": 13873, "epoch": 155, "lr": 9.94000519440264e-05}
{"train_loss": 0.5961359143257141, "global_step": 13874, "epoch": 155, "lr": 9.939996240199267e-05}
{"train_loss": 0.8136416673660278, "global_step": 13875, "epoch": 155, "lr": 9.939987285331772e-05}
{"train_loss": 0.7606073021888733, "global_step": 13876, "epoch": 155, "lr": 9.939978329800158e-05}
{"train_loss": 0.5691072940826416, "global_step": 13877, "epoch": 155, "lr": 9.939969373604421e-05}
{"train_loss": 0.5409156084060669, "global_step": 13878, "epoch": 155, "lr": 9.939960416744566e-05}
{"train_loss": 0.6477263569831848, "global_step": 13879, "epoch": 155, "lr": 9.939951459220594e-05}
{"train_loss": 0.6611222624778748, "global_step": 13880, "epoch": 155, "lr": 9.939942501032506e-05}
{"train_loss": 0.5828068852424622, "global_step": 13881, "epoch": 155, "lr": 9.939933542180302e-05}
{"train_loss": 0.5626973509788513, "global_step": 13882, "epoch": 155, "lr": 9.939924582663985e-05}
{"train_loss": 0.6098211615273122, "global_step": 13883, "epoch": 155, "lr": 9.939915622483555e-05, "val_loss": 1.4470720291137695, "train_action_mse_error": 51.44746398925781}
{"train_loss": 0.6301636099815369, "global_step": 13884, "epoch": 156, "lr": 9.939906661639014e-05}
{"train_loss": 0.7270892858505249, "global_step": 13885, "epoch": 156, "lr": 9.939897700130362e-05}
{"train_loss": 0.558404803276062, "global_step": 13886, "epoch": 156, "lr": 9.939888737957602e-05}
{"train_loss": 0.621956467628479, "global_step": 13887, "epoch": 156, "lr": 9.939879775120732e-05}
{"train_loss": 0.5868151783943176, "global_step": 13888, "epoch": 156, "lr": 9.939870811619758e-05}
{"train_loss": 0.5159444808959961, "global_step": 13889, "epoch": 156, "lr": 9.939861847454676e-05}
{"train_loss": 0.5571382641792297, "global_step": 13890, "epoch": 156, "lr": 9.939852882625492e-05}
{"train_loss": 0.6641137003898621, "global_step": 13891, "epoch": 156, "lr": 9.939843917132204e-05}
{"train_loss": 0.5664724707603455, "global_step": 13892, "epoch": 156, "lr": 9.939834950974814e-05}
{"train_loss": 0.5414688587188721, "global_step": 13893, "epoch": 156, "lr": 9.939825984153324e-05}
{"train_loss": 0.638041615486145, "global_step": 13894, "epoch": 156, "lr": 9.939817016667735e-05}
{"train_loss": 0.5774992108345032, "global_step": 13895, "epoch": 156, "lr": 9.939808048518046e-05}
{"train_loss": 0.6569443941116333, "global_step": 13896, "epoch": 156, "lr": 9.939799079704262e-05}
{"train_loss": 0.6285626292228699, "global_step": 13897, "epoch": 156, "lr": 9.939790110226381e-05}
{"train_loss": 0.6860140562057495, "global_step": 13898, "epoch": 156, "lr": 9.939781140084408e-05}
{"train_loss": 0.6616710424423218, "global_step": 13899, "epoch": 156, "lr": 9.939772169278339e-05}
{"train_loss": 0.6356042623519897, "global_step": 13900, "epoch": 156, "lr": 9.939763197808179e-05}
{"train_loss": 0.6181282997131348, "global_step": 13901, "epoch": 156, "lr": 9.939754225673926e-05}
{"train_loss": 0.5342998504638672, "global_step": 13902, "epoch": 156, "lr": 9.939745252875585e-05}
{"train_loss": 0.5939338803291321, "global_step": 13903, "epoch": 156, "lr": 9.939736279413154e-05}
{"train_loss": 0.5038196444511414, "global_step": 13904, "epoch": 156, "lr": 9.939727305286638e-05}
{"train_loss": 0.6188250184059143, "global_step": 13905, "epoch": 156, "lr": 9.939718330496035e-05}
{"train_loss": 0.699500560760498, "global_step": 13906, "epoch": 156, "lr": 9.939709355041347e-05}
{"train_loss": 0.4471927583217621, "global_step": 13907, "epoch": 156, "lr": 9.939700378922574e-05}
{"train_loss": 0.5875861048698425, "global_step": 13908, "epoch": 156, "lr": 9.939691402139719e-05}
{"train_loss": 0.5558033585548401, "global_step": 13909, "epoch": 156, "lr": 9.939682424692784e-05}
{"train_loss": 0.5093676447868347, "global_step": 13910, "epoch": 156, "lr": 9.939673446581768e-05}
{"train_loss": 0.5323992967605591, "global_step": 13911, "epoch": 156, "lr": 9.939664467806673e-05}
{"train_loss": 0.6084859371185303, "global_step": 13912, "epoch": 156, "lr": 9.939655488367501e-05}
{"train_loss": 0.5885949730873108, "global_step": 13913, "epoch": 156, "lr": 9.939646508264252e-05}
{"train_loss": 0.4780510663986206, "global_step": 13914, "epoch": 156, "lr": 9.939637527496928e-05}
{"train_loss": 0.6839767098426819, "global_step": 13915, "epoch": 156, "lr": 9.939628546065529e-05}
{"train_loss": 0.6106817722320557, "global_step": 13916, "epoch": 156, "lr": 9.939619563970059e-05}
{"train_loss": 0.6786927580833435, "global_step": 13917, "epoch": 156, "lr": 9.939610581210515e-05}
{"train_loss": 0.764523983001709, "global_step": 13918, "epoch": 156, "lr": 9.939601597786903e-05}
{"train_loss": 0.6288171410560608, "global_step": 13919, "epoch": 156, "lr": 9.93959261369922e-05}
{"train_loss": 0.5095231533050537, "global_step": 13920, "epoch": 156, "lr": 9.93958362894747e-05}
{"train_loss": 0.5396434664726257, "global_step": 13921, "epoch": 156, "lr": 9.939574643531653e-05}
{"train_loss": 0.519917368888855, "global_step": 13922, "epoch": 156, "lr": 9.93956565745177e-05}
{"train_loss": 0.5899004936218262, "global_step": 13923, "epoch": 156, "lr": 9.939556670707823e-05}
{"train_loss": 0.5149461627006531, "global_step": 13924, "epoch": 156, "lr": 9.939547683299813e-05}
{"train_loss": 0.6685614585876465, "global_step": 13925, "epoch": 156, "lr": 9.93953869522774e-05}
{"train_loss": 0.6578384637832642, "global_step": 13926, "epoch": 156, "lr": 9.939529706491606e-05}
{"train_loss": 0.4658474326133728, "global_step": 13927, "epoch": 156, "lr": 9.939520717091412e-05}
{"train_loss": 0.7353401184082031, "global_step": 13928, "epoch": 156, "lr": 9.939511727027162e-05}
{"train_loss": 0.5878605246543884, "global_step": 13929, "epoch": 156, "lr": 9.939502736298854e-05}
{"train_loss": 0.7337626218795776, "global_step": 13930, "epoch": 156, "lr": 9.93949374490649e-05}
{"train_loss": 0.6076982021331787, "global_step": 13931, "epoch": 156, "lr": 9.939484752850071e-05}
{"train_loss": 0.6122353672981262, "global_step": 13932, "epoch": 156, "lr": 9.939475760129598e-05}
{"train_loss": 0.5497878193855286, "global_step": 13933, "epoch": 156, "lr": 9.939466766745074e-05}
{"train_loss": 0.6335257291793823, "global_step": 13934, "epoch": 156, "lr": 9.939457772696497e-05}
{"train_loss": 0.641126811504364, "global_step": 13935, "epoch": 156, "lr": 9.939448777983873e-05}
{"train_loss": 0.6707192659378052, "global_step": 13936, "epoch": 156, "lr": 9.939439782607199e-05}
{"train_loss": 0.5492909550666809, "global_step": 13937, "epoch": 156, "lr": 9.939430786566476e-05}
{"train_loss": 0.6251650452613831, "global_step": 13938, "epoch": 156, "lr": 9.939421789861707e-05}
{"train_loss": 0.5909632444381714, "global_step": 13939, "epoch": 156, "lr": 9.939412792492895e-05}
{"train_loss": 0.6735832095146179, "global_step": 13940, "epoch": 156, "lr": 9.939403794460037e-05}
{"train_loss": 0.4980826675891876, "global_step": 13941, "epoch": 156, "lr": 9.939394795763137e-05}
{"train_loss": 0.6045451760292053, "global_step": 13942, "epoch": 156, "lr": 9.939385796402196e-05}
{"train_loss": 0.6779576539993286, "global_step": 13943, "epoch": 156, "lr": 9.939376796377215e-05}
{"train_loss": 0.7611432075500488, "global_step": 13944, "epoch": 156, "lr": 9.939367795688195e-05}
{"train_loss": 0.5989648103713989, "global_step": 13945, "epoch": 156, "lr": 9.939358794335137e-05}
{"train_loss": 0.666532576084137, "global_step": 13946, "epoch": 156, "lr": 9.939349792318042e-05}
{"train_loss": 0.5375235080718994, "global_step": 13947, "epoch": 156, "lr": 9.939340789636912e-05}
{"train_loss": 0.6760353446006775, "global_step": 13948, "epoch": 156, "lr": 9.939331786291748e-05}
{"train_loss": 0.6192837953567505, "global_step": 13949, "epoch": 156, "lr": 9.93932278228255e-05}
{"train_loss": 0.5409179925918579, "global_step": 13950, "epoch": 156, "lr": 9.93931377760932e-05}
{"train_loss": 0.6288111805915833, "global_step": 13951, "epoch": 156, "lr": 9.939304772272061e-05}
{"train_loss": 0.6639225482940674, "global_step": 13952, "epoch": 156, "lr": 9.939295766270773e-05}
{"train_loss": 0.6796581745147705, "global_step": 13953, "epoch": 156, "lr": 9.939286759605456e-05}
{"train_loss": 0.500443696975708, "global_step": 13954, "epoch": 156, "lr": 9.939277752276111e-05}
{"train_loss": 0.48169368505477905, "global_step": 13955, "epoch": 156, "lr": 9.939268744282742e-05}
{"train_loss": 0.5131389498710632, "global_step": 13956, "epoch": 156, "lr": 9.939259735625348e-05}
{"train_loss": 0.6886138319969177, "global_step": 13957, "epoch": 156, "lr": 9.93925072630393e-05}
{"train_loss": 0.6629136800765991, "global_step": 13958, "epoch": 156, "lr": 9.939241716318491e-05}
{"train_loss": 0.6133702397346497, "global_step": 13959, "epoch": 156, "lr": 9.939232705669031e-05}
{"train_loss": 0.6966721415519714, "global_step": 13960, "epoch": 156, "lr": 9.93922369435555e-05}
{"train_loss": 0.76922208070755, "global_step": 13961, "epoch": 156, "lr": 9.939214682378052e-05}
{"train_loss": 0.6724633574485779, "global_step": 13962, "epoch": 156, "lr": 9.939205669736538e-05}
{"train_loss": 0.6290187239646912, "global_step": 13963, "epoch": 156, "lr": 9.939196656431006e-05}
{"train_loss": 0.6959632039070129, "global_step": 13964, "epoch": 156, "lr": 9.939187642461459e-05}
{"train_loss": 0.5179697275161743, "global_step": 13965, "epoch": 156, "lr": 9.9391786278279e-05}
{"train_loss": 0.4060657322406769, "global_step": 13966, "epoch": 156, "lr": 9.939169612530329e-05}
{"train_loss": 0.4450031816959381, "global_step": 13967, "epoch": 156, "lr": 9.939160596568744e-05}
{"train_loss": 0.6122654676437378, "global_step": 13968, "epoch": 156, "lr": 9.939151579943151e-05}
{"train_loss": 0.4777025282382965, "global_step": 13969, "epoch": 156, "lr": 9.939142562653548e-05}
{"train_loss": 0.5900607109069824, "global_step": 13970, "epoch": 156, "lr": 9.939133544699939e-05}
{"train_loss": 0.784123957157135, "global_step": 13971, "epoch": 156, "lr": 9.939124526082323e-05}
{"train_loss": 0.6053490742538752, "global_step": 13972, "epoch": 156, "lr": 9.939115506800703e-05, "val_loss": 1.4406028985977173}
{"train_loss": 0.5913744568824768, "global_step": 13973, "epoch": 157, "lr": 9.939106486855077e-05}
{"train_loss": 0.7023093104362488, "global_step": 13974, "epoch": 157, "lr": 9.93909746624545e-05}
{"train_loss": 0.6310164332389832, "global_step": 13975, "epoch": 157, "lr": 9.93908844497182e-05}
{"train_loss": 0.6594023108482361, "global_step": 13976, "epoch": 157, "lr": 9.939079423034192e-05}
{"train_loss": 0.5487425327301025, "global_step": 13977, "epoch": 157, "lr": 9.939070400432563e-05}
{"train_loss": 0.5353177785873413, "global_step": 13978, "epoch": 157, "lr": 9.939061377166938e-05}
{"train_loss": 0.713647186756134, "global_step": 13979, "epoch": 157, "lr": 9.939052353237314e-05}
{"train_loss": 0.6687133312225342, "global_step": 13980, "epoch": 157, "lr": 9.939043328643696e-05}
{"train_loss": 0.5852940082550049, "global_step": 13981, "epoch": 157, "lr": 9.939034303386083e-05}
{"train_loss": 0.5922607183456421, "global_step": 13982, "epoch": 157, "lr": 9.939025277464478e-05}
{"train_loss": 0.6233440637588501, "global_step": 13983, "epoch": 157, "lr": 9.939016250878881e-05}
{"train_loss": 0.6990106105804443, "global_step": 13984, "epoch": 157, "lr": 9.939007223629294e-05}
{"train_loss": 0.456863671541214, "global_step": 13985, "epoch": 157, "lr": 9.938998195715717e-05}
{"train_loss": 0.5647461414337158, "global_step": 13986, "epoch": 157, "lr": 9.938989167138153e-05}
{"train_loss": 0.5746487975120544, "global_step": 13987, "epoch": 157, "lr": 9.9389801378966e-05}
{"train_loss": 0.47831496596336365, "global_step": 13988, "epoch": 157, "lr": 9.938971107991063e-05}
{"train_loss": 0.5774128437042236, "global_step": 13989, "epoch": 157, "lr": 9.93896207742154e-05}
{"train_loss": 0.530885636806488, "global_step": 13990, "epoch": 157, "lr": 9.938953046188035e-05}
{"train_loss": 0.5614715814590454, "global_step": 13991, "epoch": 157, "lr": 9.938944014290547e-05}
{"train_loss": 0.6101704239845276, "global_step": 13992, "epoch": 157, "lr": 9.93893498172908e-05}
{"train_loss": 0.5840727686882019, "global_step": 13993, "epoch": 157, "lr": 9.938925948503633e-05}
{"train_loss": 0.676639974117279, "global_step": 13994, "epoch": 157, "lr": 9.938916914614207e-05}
{"train_loss": 0.5801864266395569, "global_step": 13995, "epoch": 157, "lr": 9.938907880060804e-05}
{"train_loss": 0.6384446620941162, "global_step": 13996, "epoch": 157, "lr": 9.938898844843425e-05}
{"train_loss": 0.5544585585594177, "global_step": 13997, "epoch": 157, "lr": 9.93888980896207e-05}
{"train_loss": 0.6034426093101501, "global_step": 13998, "epoch": 157, "lr": 9.938880772416744e-05}
{"train_loss": 0.717305600643158, "global_step": 13999, "epoch": 157, "lr": 9.938871735207442e-05}
{"train_loss": 0.482846736907959, "global_step": 14000, "epoch": 157, "lr": 9.938862697334171e-05}
{"train_loss": 0.5290064215660095, "global_step": 14001, "epoch": 157, "lr": 9.938853658796931e-05}
{"train_loss": 0.6214055418968201, "global_step": 14002, "epoch": 157, "lr": 9.938844619595722e-05}
{"train_loss": 0.6022653579711914, "global_step": 14003, "epoch": 157, "lr": 9.938835579730544e-05}
{"train_loss": 0.7328585386276245, "global_step": 14004, "epoch": 157, "lr": 9.9388265392014e-05}
{"train_loss": 0.5292187333106995, "global_step": 14005, "epoch": 157, "lr": 9.938817498008293e-05}
{"train_loss": 0.7112640738487244, "global_step": 14006, "epoch": 157, "lr": 9.93880845615122e-05}
{"train_loss": 0.5555951595306396, "global_step": 14007, "epoch": 157, "lr": 9.938799413630186e-05}
{"train_loss": 0.5598871111869812, "global_step": 14008, "epoch": 157, "lr": 9.93879037044519e-05}
{"train_loss": 0.5916383862495422, "global_step": 14009, "epoch": 157, "lr": 9.938781326596233e-05}
{"train_loss": 0.6637231707572937, "global_step": 14010, "epoch": 157, "lr": 9.938772282083318e-05}
{"train_loss": 0.610313892364502, "global_step": 14011, "epoch": 157, "lr": 9.938763236906445e-05}
{"train_loss": 0.5279505252838135, "global_step": 14012, "epoch": 157, "lr": 9.938754191065616e-05}
{"train_loss": 0.6588994264602661, "global_step": 14013, "epoch": 157, "lr": 9.93874514456083e-05}
{"train_loss": 0.6068048477172852, "global_step": 14014, "epoch": 157, "lr": 9.938736097392092e-05}
{"train_loss": 0.4864599108695984, "global_step": 14015, "epoch": 157, "lr": 9.9387270495594e-05}
{"train_loss": 0.69289630651474, "global_step": 14016, "epoch": 157, "lr": 9.938718001062757e-05}
{"train_loss": 0.5397076606750488, "global_step": 14017, "epoch": 157, "lr": 9.938708951902162e-05}
{"train_loss": 0.5300590991973877, "global_step": 14018, "epoch": 157, "lr": 9.938699902077619e-05}
{"train_loss": 0.7095736861228943, "global_step": 14019, "epoch": 157, "lr": 9.938690851589129e-05}
{"train_loss": 0.6327667236328125, "global_step": 14020, "epoch": 157, "lr": 9.93868180043669e-05}
{"train_loss": 0.5390402674674988, "global_step": 14021, "epoch": 157, "lr": 9.938672748620307e-05}
{"train_loss": 0.5655079483985901, "global_step": 14022, "epoch": 157, "lr": 9.938663696139979e-05}
{"train_loss": 0.3801633417606354, "global_step": 14023, "epoch": 157, "lr": 9.938654642995708e-05}
{"train_loss": 0.5500330924987793, "global_step": 14024, "epoch": 157, "lr": 9.938645589187496e-05}
{"train_loss": 0.6312652826309204, "global_step": 14025, "epoch": 157, "lr": 9.938636534715342e-05}
{"train_loss": 0.5485674142837524, "global_step": 14026, "epoch": 157, "lr": 9.938627479579249e-05}
{"train_loss": 0.4264795184135437, "global_step": 14027, "epoch": 157, "lr": 9.938618423779218e-05}
{"train_loss": 0.598297119140625, "global_step": 14028, "epoch": 157, "lr": 9.938609367315249e-05}
{"train_loss": 0.6412381529808044, "global_step": 14029, "epoch": 157, "lr": 9.938600310187345e-05}
{"train_loss": 0.5824486613273621, "global_step": 14030, "epoch": 157, "lr": 9.938591252395506e-05}
{"train_loss": 0.4809783697128296, "global_step": 14031, "epoch": 157, "lr": 9.938582193939733e-05}
{"train_loss": 0.6287691593170166, "global_step": 14032, "epoch": 157, "lr": 9.93857313482003e-05}
{"train_loss": 0.5644878149032593, "global_step": 14033, "epoch": 157, "lr": 9.938564075036394e-05}
{"train_loss": 0.5967822670936584, "global_step": 14034, "epoch": 157, "lr": 9.938555014588829e-05}
{"train_loss": 0.5791810750961304, "global_step": 14035, "epoch": 157, "lr": 9.938545953477335e-05}
{"train_loss": 0.5970211029052734, "global_step": 14036, "epoch": 157, "lr": 9.938536891701915e-05}
{"train_loss": 0.6092044711112976, "global_step": 14037, "epoch": 157, "lr": 9.938527829262568e-05}
{"train_loss": 0.6575665473937988, "global_step": 14038, "epoch": 157, "lr": 9.938518766159296e-05}
{"train_loss": 0.6259093284606934, "global_step": 14039, "epoch": 157, "lr": 9.9385097023921e-05}
{"train_loss": 0.7700321078300476, "global_step": 14040, "epoch": 157, "lr": 9.938500637960981e-05}
{"train_loss": 0.6251862645149231, "global_step": 14041, "epoch": 157, "lr": 9.938491572865943e-05}
{"train_loss": 0.5661665797233582, "global_step": 14042, "epoch": 157, "lr": 9.938482507106984e-05}
{"train_loss": 0.6038304567337036, "global_step": 14043, "epoch": 157, "lr": 9.938473440684105e-05}
{"train_loss": 0.4013100862503052, "global_step": 14044, "epoch": 157, "lr": 9.938464373597309e-05}
{"train_loss": 0.629037082195282, "global_step": 14045, "epoch": 157, "lr": 9.938455305846596e-05}
{"train_loss": 0.5064476728439331, "global_step": 14046, "epoch": 157, "lr": 9.938446237431971e-05}
{"train_loss": 0.5625007152557373, "global_step": 14047, "epoch": 157, "lr": 9.938437168353429e-05}
{"train_loss": 0.6442228555679321, "global_step": 14048, "epoch": 157, "lr": 9.938428098610976e-05}
{"train_loss": 0.6087629199028015, "global_step": 14049, "epoch": 157, "lr": 9.938419028204611e-05}
{"train_loss": 0.6886948347091675, "global_step": 14050, "epoch": 157, "lr": 9.938409957134335e-05}
{"train_loss": 0.5977051258087158, "global_step": 14051, "epoch": 157, "lr": 9.938400885400151e-05}
{"train_loss": 0.6662784218788147, "global_step": 14052, "epoch": 157, "lr": 9.93839181300206e-05}
{"train_loss": 0.6185666918754578, "global_step": 14053, "epoch": 157, "lr": 9.93838273994006e-05}
{"train_loss": 0.5914919376373291, "global_step": 14054, "epoch": 157, "lr": 9.938373666214156e-05}
{"train_loss": 0.5869030356407166, "global_step": 14055, "epoch": 157, "lr": 9.938364591824347e-05}
{"train_loss": 0.603740930557251, "global_step": 14056, "epoch": 157, "lr": 9.938355516770637e-05}
{"train_loss": 0.5870246887207031, "global_step": 14057, "epoch": 157, "lr": 9.938346441053024e-05}
{"train_loss": 0.5896726250648499, "global_step": 14058, "epoch": 157, "lr": 9.93833736467151e-05}
{"train_loss": 0.6811708807945251, "global_step": 14059, "epoch": 157, "lr": 9.938328287626097e-05}
{"train_loss": 0.6128105521202087, "global_step": 14060, "epoch": 157, "lr": 9.938319209916787e-05}
{"train_loss": 0.5966935402221893, "global_step": 14061, "epoch": 157, "lr": 9.938310131543578e-05, "val_loss": 1.456577181816101}
{"train_loss": 0.672558069229126, "global_step": 14062, "epoch": 158, "lr": 9.938301052506475e-05}
{"train_loss": 0.5949400663375854, "global_step": 14063, "epoch": 158, "lr": 9.938291972805477e-05}
{"train_loss": 0.4518662989139557, "global_step": 14064, "epoch": 158, "lr": 9.938282892440587e-05}
{"train_loss": 0.5593377947807312, "global_step": 14065, "epoch": 158, "lr": 9.938273811411805e-05}
{"train_loss": 0.5479601621627808, "global_step": 14066, "epoch": 158, "lr": 9.938264729719132e-05}
{"train_loss": 0.6390787363052368, "global_step": 14067, "epoch": 158, "lr": 9.938255647362567e-05}
{"train_loss": 0.5922024250030518, "global_step": 14068, "epoch": 158, "lr": 9.938246564342118e-05}
{"train_loss": 0.6151255965232849, "global_step": 14069, "epoch": 158, "lr": 9.938237480657779e-05}
{"train_loss": 0.6413921117782593, "global_step": 14070, "epoch": 158, "lr": 9.938228396309554e-05}
{"train_loss": 0.5289045572280884, "global_step": 14071, "epoch": 158, "lr": 9.938219311297446e-05}
{"train_loss": 0.7014487981796265, "global_step": 14072, "epoch": 158, "lr": 9.938210225621454e-05}
{"train_loss": 0.5393211245536804, "global_step": 14073, "epoch": 158, "lr": 9.93820113928158e-05}
{"train_loss": 0.6045606136322021, "global_step": 14074, "epoch": 158, "lr": 9.938192052277823e-05}
{"train_loss": 0.40068110823631287, "global_step": 14075, "epoch": 158, "lr": 9.938182964610189e-05}
{"train_loss": 0.43689867854118347, "global_step": 14076, "epoch": 158, "lr": 9.938173876278676e-05}
{"train_loss": 0.6658633351325989, "global_step": 14077, "epoch": 158, "lr": 9.938164787283284e-05}
{"train_loss": 0.5200220942497253, "global_step": 14078, "epoch": 158, "lr": 9.938155697624017e-05}
{"train_loss": 0.613550066947937, "global_step": 14079, "epoch": 158, "lr": 9.938146607300875e-05}
{"train_loss": 0.6637162566184998, "global_step": 14080, "epoch": 158, "lr": 9.938137516313861e-05}
{"train_loss": 0.6466944813728333, "global_step": 14081, "epoch": 158, "lr": 9.938128424662974e-05}
{"train_loss": 0.7549323439598083, "global_step": 14082, "epoch": 158, "lr": 9.938119332348214e-05}
{"train_loss": 0.545354962348938, "global_step": 14083, "epoch": 158, "lr": 9.938110239369585e-05}
{"train_loss": 0.6283380389213562, "global_step": 14084, "epoch": 158, "lr": 9.938101145727089e-05}
{"train_loss": 0.5152105689048767, "global_step": 14085, "epoch": 158, "lr": 9.938092051420722e-05}
{"train_loss": 0.7926607131958008, "global_step": 14086, "epoch": 158, "lr": 9.938082956450491e-05}
{"train_loss": 0.8084466457366943, "global_step": 14087, "epoch": 158, "lr": 9.938073860816394e-05}
{"train_loss": 0.5368066430091858, "global_step": 14088, "epoch": 158, "lr": 9.938064764518434e-05}
{"train_loss": 0.6779292225837708, "global_step": 14089, "epoch": 158, "lr": 9.938055667556612e-05}
{"train_loss": 0.594940185546875, "global_step": 14090, "epoch": 158, "lr": 9.938046569930926e-05}
{"train_loss": 0.4811820387840271, "global_step": 14091, "epoch": 158, "lr": 9.938037471641381e-05}
{"train_loss": 0.4582734704017639, "global_step": 14092, "epoch": 158, "lr": 9.938028372687979e-05}
{"train_loss": 0.533394455909729, "global_step": 14093, "epoch": 158, "lr": 9.938019273070718e-05}
{"train_loss": 0.7718048691749573, "global_step": 14094, "epoch": 158, "lr": 9.9380101727896e-05}
{"train_loss": 0.6211872100830078, "global_step": 14095, "epoch": 158, "lr": 9.938001071844627e-05}
{"train_loss": 0.48081517219543457, "global_step": 14096, "epoch": 158, "lr": 9.9379919702358e-05}
{"train_loss": 0.7802170515060425, "global_step": 14097, "epoch": 158, "lr": 9.937982867963121e-05}
{"train_loss": 0.6073265075683594, "global_step": 14098, "epoch": 158, "lr": 9.937973765026589e-05}
{"train_loss": 0.7095242738723755, "global_step": 14099, "epoch": 158, "lr": 9.937964661426206e-05}
{"train_loss": 0.5763956904411316, "global_step": 14100, "epoch": 158, "lr": 9.937955557161975e-05}
{"train_loss": 0.4501815736293793, "global_step": 14101, "epoch": 158, "lr": 9.937946452233897e-05}
{"train_loss": 0.5442057847976685, "global_step": 14102, "epoch": 158, "lr": 9.937937346641971e-05}
{"train_loss": 0.6043491363525391, "global_step": 14103, "epoch": 158, "lr": 9.9379282403862e-05}
{"train_loss": 0.6247149705886841, "global_step": 14104, "epoch": 158, "lr": 9.937919133466584e-05}
{"train_loss": 0.5918993949890137, "global_step": 14105, "epoch": 158, "lr": 9.937910025883126e-05}
{"train_loss": 0.7815589904785156, "global_step": 14106, "epoch": 158, "lr": 9.937900917635825e-05}
{"train_loss": 0.5543591380119324, "global_step": 14107, "epoch": 158, "lr": 9.937891808724685e-05}
{"train_loss": 0.5764198899269104, "global_step": 14108, "epoch": 158, "lr": 9.937882699149705e-05}
{"train_loss": 0.7343312501907349, "global_step": 14109, "epoch": 158, "lr": 9.937873588910887e-05}
{"train_loss": 0.6119746565818787, "global_step": 14110, "epoch": 158, "lr": 9.93786447800823e-05}
{"train_loss": 0.5330929160118103, "global_step": 14111, "epoch": 158, "lr": 9.93785536644174e-05}
{"train_loss": 0.576059103012085, "global_step": 14112, "epoch": 158, "lr": 9.937846254211413e-05}
{"train_loss": 0.6900972723960876, "global_step": 14113, "epoch": 158, "lr": 9.937837141317254e-05}
{"train_loss": 0.5107117891311646, "global_step": 14114, "epoch": 158, "lr": 9.937828027759263e-05}
{"train_loss": 0.5327540040016174, "global_step": 14115, "epoch": 158, "lr": 9.937818913537442e-05}
{"train_loss": 0.5864860415458679, "global_step": 14116, "epoch": 158, "lr": 9.937809798651792e-05}
{"train_loss": 0.7336795926094055, "global_step": 14117, "epoch": 158, "lr": 9.937800683102312e-05}
{"train_loss": 0.46896612644195557, "global_step": 14118, "epoch": 158, "lr": 9.937791566889006e-05}
{"train_loss": 0.5875836610794067, "global_step": 14119, "epoch": 158, "lr": 9.937782450011873e-05}
{"train_loss": 0.6008466482162476, "global_step": 14120, "epoch": 158, "lr": 9.937773332470916e-05}
{"train_loss": 0.6374146938323975, "global_step": 14121, "epoch": 158, "lr": 9.937764214266136e-05}
{"train_loss": 0.5398629307746887, "global_step": 14122, "epoch": 158, "lr": 9.937755095397532e-05}
{"train_loss": 0.47328606247901917, "global_step": 14123, "epoch": 158, "lr": 9.937745975865109e-05}
{"train_loss": 0.5419283509254456, "global_step": 14124, "epoch": 158, "lr": 9.937736855668866e-05}
{"train_loss": 0.6603232026100159, "global_step": 14125, "epoch": 158, "lr": 9.937727734808804e-05}
{"train_loss": 0.5996787548065186, "global_step": 14126, "epoch": 158, "lr": 9.937718613284923e-05}
{"train_loss": 0.5812305808067322, "global_step": 14127, "epoch": 158, "lr": 9.937709491097228e-05}
{"train_loss": 0.5828545689582825, "global_step": 14128, "epoch": 158, "lr": 9.937700368245717e-05}
{"train_loss": 0.6886187791824341, "global_step": 14129, "epoch": 158, "lr": 9.937691244730394e-05}
{"train_loss": 0.6391465067863464, "global_step": 14130, "epoch": 158, "lr": 9.937682120551257e-05}
{"train_loss": 0.609910786151886, "global_step": 14131, "epoch": 158, "lr": 9.937672995708309e-05}
{"train_loss": 0.6658017039299011, "global_step": 14132, "epoch": 158, "lr": 9.937663870201552e-05}
{"train_loss": 0.661194384098053, "global_step": 14133, "epoch": 158, "lr": 9.937654744030986e-05}
{"train_loss": 0.5962040424346924, "global_step": 14134, "epoch": 158, "lr": 9.937645617196612e-05}
{"train_loss": 0.6841174960136414, "global_step": 14135, "epoch": 158, "lr": 9.937636489698432e-05}
{"train_loss": 0.5621616840362549, "global_step": 14136, "epoch": 158, "lr": 9.937627361536447e-05}
{"train_loss": 0.5216812491416931, "global_step": 14137, "epoch": 158, "lr": 9.937618232710658e-05}
{"train_loss": 0.6118465662002563, "global_step": 14138, "epoch": 158, "lr": 9.937609103221066e-05}
{"train_loss": 0.4668639004230499, "global_step": 14139, "epoch": 158, "lr": 9.937599973067675e-05}
{"train_loss": 0.5108214616775513, "global_step": 14140, "epoch": 158, "lr": 9.937590842250481e-05}
{"train_loss": 0.6523219347000122, "global_step": 14141, "epoch": 158, "lr": 9.937581710769489e-05}
{"train_loss": 0.4219478368759155, "global_step": 14142, "epoch": 158, "lr": 9.937572578624698e-05}
{"train_loss": 0.5073312520980835, "global_step": 14143, "epoch": 158, "lr": 9.937563445816114e-05}
{"train_loss": 0.6982940435409546, "global_step": 14144, "epoch": 158, "lr": 9.937554312343731e-05}
{"train_loss": 0.5989002585411072, "global_step": 14145, "epoch": 158, "lr": 9.937545178207557e-05}
{"train_loss": 0.5469323992729187, "global_step": 14146, "epoch": 158, "lr": 9.93753604340759e-05}
{"train_loss": 0.5215954184532166, "global_step": 14147, "epoch": 158, "lr": 9.93752690794383e-05}
{"train_loss": 0.7014317512512207, "global_step": 14148, "epoch": 158, "lr": 9.937517771816281e-05}
{"train_loss": 0.7505047917366028, "global_step": 14149, "epoch": 158, "lr": 9.937508635024941e-05}
{"train_loss": 0.5967153886730752, "global_step": 14150, "epoch": 158, "lr": 9.937499497569813e-05, "val_loss": 1.397762417793274}
{"train_loss": 0.5768498182296753, "global_step": 14151, "epoch": 159, "lr": 9.9374903594509e-05}
{"train_loss": 0.5489141345024109, "global_step": 14152, "epoch": 159, "lr": 9.937481220668201e-05}
{"train_loss": 0.5280110239982605, "global_step": 14153, "epoch": 159, "lr": 9.93747208122172e-05}
{"train_loss": 0.675082802772522, "global_step": 14154, "epoch": 159, "lr": 9.937462941111452e-05}
{"train_loss": 0.47516342997550964, "global_step": 14155, "epoch": 159, "lr": 9.937453800337406e-05}
{"train_loss": 0.6041637659072876, "global_step": 14156, "epoch": 159, "lr": 9.937444658899578e-05}
{"train_loss": 0.7115217447280884, "global_step": 14157, "epoch": 159, "lr": 9.937435516797969e-05}
{"train_loss": 0.6523849368095398, "global_step": 14158, "epoch": 159, "lr": 9.937426374032583e-05}
{"train_loss": 0.6096764206886292, "global_step": 14159, "epoch": 159, "lr": 9.937417230603421e-05}
{"train_loss": 0.54676753282547, "global_step": 14160, "epoch": 159, "lr": 9.937408086510484e-05}
{"train_loss": 0.48008012771606445, "global_step": 14161, "epoch": 159, "lr": 9.937398941753771e-05}
{"train_loss": 0.7238571047782898, "global_step": 14162, "epoch": 159, "lr": 9.937389796333287e-05}
{"train_loss": 0.7084765434265137, "global_step": 14163, "epoch": 159, "lr": 9.937380650249029e-05}
{"train_loss": 0.5420097708702087, "global_step": 14164, "epoch": 159, "lr": 9.937371503501001e-05}
{"train_loss": 0.5706527233123779, "global_step": 14165, "epoch": 159, "lr": 9.937362356089203e-05}
{"train_loss": 0.5400089621543884, "global_step": 14166, "epoch": 159, "lr": 9.93735320801364e-05}
{"train_loss": 0.537330150604248, "global_step": 14167, "epoch": 159, "lr": 9.937344059274307e-05}
{"train_loss": 0.5820583701133728, "global_step": 14168, "epoch": 159, "lr": 9.937334909871208e-05}
{"train_loss": 0.6402890086174011, "global_step": 14169, "epoch": 159, "lr": 9.937325759804345e-05}
{"train_loss": 0.47999832034111023, "global_step": 14170, "epoch": 159, "lr": 9.93731660907372e-05}
{"train_loss": 0.8161419630050659, "global_step": 14171, "epoch": 159, "lr": 9.937307457679332e-05}
{"train_loss": 0.7213931679725647, "global_step": 14172, "epoch": 159, "lr": 9.937298305621183e-05}
{"train_loss": 0.7731932401657104, "global_step": 14173, "epoch": 159, "lr": 9.937289152899276e-05}
{"train_loss": 0.5892171263694763, "global_step": 14174, "epoch": 159, "lr": 9.937279999513609e-05}
{"train_loss": 0.5495493412017822, "global_step": 14175, "epoch": 159, "lr": 9.937270845464184e-05}
{"train_loss": 0.562147855758667, "global_step": 14176, "epoch": 159, "lr": 9.937261690751006e-05}
{"train_loss": 0.7642074823379517, "global_step": 14177, "epoch": 159, "lr": 9.93725253537407e-05}
{"train_loss": 0.6641227602958679, "global_step": 14178, "epoch": 159, "lr": 9.937243379333382e-05}
{"train_loss": 0.5834305882453918, "global_step": 14179, "epoch": 159, "lr": 9.937234222628944e-05}
{"train_loss": 0.6726164221763611, "global_step": 14180, "epoch": 159, "lr": 9.937225065260754e-05}
{"train_loss": 0.5970200300216675, "global_step": 14181, "epoch": 159, "lr": 9.937215907228812e-05}
{"train_loss": 0.5235162377357483, "global_step": 14182, "epoch": 159, "lr": 9.937206748533123e-05}
{"train_loss": 0.6823979616165161, "global_step": 14183, "epoch": 159, "lr": 9.937197589173687e-05}
{"train_loss": 0.6950114965438843, "global_step": 14184, "epoch": 159, "lr": 9.937188429150503e-05}
{"train_loss": 0.6134876012802124, "global_step": 14185, "epoch": 159, "lr": 9.937179268463577e-05}
{"train_loss": 0.5662485361099243, "global_step": 14186, "epoch": 159, "lr": 9.937170107112907e-05}
{"train_loss": 0.5278702974319458, "global_step": 14187, "epoch": 159, "lr": 9.937160945098494e-05}
{"train_loss": 0.6388880014419556, "global_step": 14188, "epoch": 159, "lr": 9.937151782420339e-05}
{"train_loss": 0.7436071038246155, "global_step": 14189, "epoch": 159, "lr": 9.937142619078444e-05}
{"train_loss": 0.682447075843811, "global_step": 14190, "epoch": 159, "lr": 9.937133455072811e-05}
{"train_loss": 0.6101232767105103, "global_step": 14191, "epoch": 159, "lr": 9.93712429040344e-05}
{"train_loss": 0.5119556188583374, "global_step": 14192, "epoch": 159, "lr": 9.937115125070335e-05}
{"train_loss": 0.3981128931045532, "global_step": 14193, "epoch": 159, "lr": 9.937105959073494e-05}
{"train_loss": 0.5152413845062256, "global_step": 14194, "epoch": 159, "lr": 9.937096792412919e-05}
{"train_loss": 0.6196869015693665, "global_step": 14195, "epoch": 159, "lr": 9.937087625088612e-05}
{"train_loss": 0.6893309354782104, "global_step": 14196, "epoch": 159, "lr": 9.937078457100573e-05}
{"train_loss": 0.5424098372459412, "global_step": 14197, "epoch": 159, "lr": 9.937069288448803e-05}
{"train_loss": 0.5813618898391724, "global_step": 14198, "epoch": 159, "lr": 9.937060119133306e-05}
{"train_loss": 0.48664093017578125, "global_step": 14199, "epoch": 159, "lr": 9.937050949154081e-05}
{"train_loss": 0.6225792169570923, "global_step": 14200, "epoch": 159, "lr": 9.93704177851113e-05}
{"train_loss": 0.5992847084999084, "global_step": 14201, "epoch": 159, "lr": 9.937032607204453e-05}
{"train_loss": 0.5730977654457092, "global_step": 14202, "epoch": 159, "lr": 9.937023435234052e-05}
{"train_loss": 0.7674463987350464, "global_step": 14203, "epoch": 159, "lr": 9.937014262599929e-05}
{"train_loss": 0.5688272714614868, "global_step": 14204, "epoch": 159, "lr": 9.937005089302086e-05}
{"train_loss": 0.5716733336448669, "global_step": 14205, "epoch": 159, "lr": 9.93699591534052e-05}
{"train_loss": 0.5793398022651672, "global_step": 14206, "epoch": 159, "lr": 9.936986740715236e-05}
{"train_loss": 0.5943402051925659, "global_step": 14207, "epoch": 159, "lr": 9.936977565426235e-05}
{"train_loss": 0.5880402326583862, "global_step": 14208, "epoch": 159, "lr": 9.936968389473517e-05}
{"train_loss": 0.6819918155670166, "global_step": 14209, "epoch": 159, "lr": 9.936959212857084e-05}
{"train_loss": 0.5362615585327148, "global_step": 14210, "epoch": 159, "lr": 9.936950035576938e-05}
{"train_loss": 0.6437050700187683, "global_step": 14211, "epoch": 159, "lr": 9.936940857633077e-05}
{"train_loss": 0.6285474300384521, "global_step": 14212, "epoch": 159, "lr": 9.936931679025506e-05}
{"train_loss": 0.5649166703224182, "global_step": 14213, "epoch": 159, "lr": 9.936922499754225e-05}
{"train_loss": 0.7357137799263, "global_step": 14214, "epoch": 159, "lr": 9.936913319819235e-05}
{"train_loss": 0.6013575196266174, "global_step": 14215, "epoch": 159, "lr": 9.936904139220535e-05}
{"train_loss": 0.6945441365242004, "global_step": 14216, "epoch": 159, "lr": 9.936894957958132e-05}
{"train_loss": 0.6823160648345947, "global_step": 14217, "epoch": 159, "lr": 9.936885776032021e-05}
{"train_loss": 0.6435233950614929, "global_step": 14218, "epoch": 159, "lr": 9.936876593442206e-05}
{"train_loss": 0.5848053693771362, "global_step": 14219, "epoch": 159, "lr": 9.936867410188689e-05}
{"train_loss": 0.6303389668464661, "global_step": 14220, "epoch": 159, "lr": 9.936858226271469e-05}
{"train_loss": 0.5048301219940186, "global_step": 14221, "epoch": 159, "lr": 9.93684904169055e-05}
{"train_loss": 0.5965405106544495, "global_step": 14222, "epoch": 159, "lr": 9.936839856445932e-05}
{"train_loss": 0.5878252983093262, "global_step": 14223, "epoch": 159, "lr": 9.936830670537616e-05}
{"train_loss": 0.5094167590141296, "global_step": 14224, "epoch": 159, "lr": 9.936821483965601e-05}
{"train_loss": 0.649507999420166, "global_step": 14225, "epoch": 159, "lr": 9.936812296729894e-05}
{"train_loss": 0.6267645955085754, "global_step": 14226, "epoch": 159, "lr": 9.93680310883049e-05}
{"train_loss": 0.6462603211402893, "global_step": 14227, "epoch": 159, "lr": 9.936793920267394e-05}
{"train_loss": 0.6113473773002625, "global_step": 14228, "epoch": 159, "lr": 9.936784731040605e-05}
{"train_loss": 0.8276629447937012, "global_step": 14229, "epoch": 159, "lr": 9.936775541150127e-05}
{"train_loss": 0.5488909482955933, "global_step": 14230, "epoch": 159, "lr": 9.93676635059596e-05}
{"train_loss": 0.48516830801963806, "global_step": 14231, "epoch": 159, "lr": 9.936757159378105e-05}
{"train_loss": 0.5895791053771973, "global_step": 14232, "epoch": 159, "lr": 9.936747967496563e-05}
{"train_loss": 0.6845569610595703, "global_step": 14233, "epoch": 159, "lr": 9.936738774951334e-05}
{"train_loss": 0.6635458469390869, "global_step": 14234, "epoch": 159, "lr": 9.936729581742421e-05}
{"train_loss": 0.7042028903961182, "global_step": 14235, "epoch": 159, "lr": 9.936720387869827e-05}
{"train_loss": 0.5716853737831116, "global_step": 14236, "epoch": 159, "lr": 9.936711193333548e-05}
{"train_loss": 0.62336665391922, "global_step": 14237, "epoch": 159, "lr": 9.936701998133591e-05}
{"train_loss": 0.6956490874290466, "global_step": 14238, "epoch": 159, "lr": 9.936692802269954e-05}
{"train_loss": 0.6148902817388598, "global_step": 14239, "epoch": 159, "lr": 9.936683605742637e-05, "val_loss": 1.448954463005066}
{"train_loss": 0.5569921731948853, "global_step": 14240, "epoch": 160, "lr": 9.936674408551645e-05}
{"train_loss": 0.5181177854537964, "global_step": 14241, "epoch": 160, "lr": 9.936665210696977e-05}
{"train_loss": 0.5837744474411011, "global_step": 14242, "epoch": 160, "lr": 9.936656012178634e-05}
{"train_loss": 0.4780583083629608, "global_step": 14243, "epoch": 160, "lr": 9.936646812996617e-05}
{"train_loss": 0.5556450486183167, "global_step": 14244, "epoch": 160, "lr": 9.93663761315093e-05}
{"train_loss": 0.7044286131858826, "global_step": 14245, "epoch": 160, "lr": 9.936628412641571e-05}
{"train_loss": 0.5414856672286987, "global_step": 14246, "epoch": 160, "lr": 9.936619211468542e-05}
{"train_loss": 0.5760058164596558, "global_step": 14247, "epoch": 160, "lr": 9.936610009631846e-05}
{"train_loss": 0.5075935125350952, "global_step": 14248, "epoch": 160, "lr": 9.936600807131483e-05}
{"train_loss": 0.6226336359977722, "global_step": 14249, "epoch": 160, "lr": 9.936591603967453e-05}
{"train_loss": 0.41579562425613403, "global_step": 14250, "epoch": 160, "lr": 9.936582400139757e-05}
{"train_loss": 0.5247495770454407, "global_step": 14251, "epoch": 160, "lr": 9.936573195648401e-05}
{"train_loss": 0.49979540705680847, "global_step": 14252, "epoch": 160, "lr": 9.936563990493382e-05}
{"train_loss": 0.6275367736816406, "global_step": 14253, "epoch": 160, "lr": 9.936554784674701e-05}
{"train_loss": 0.46793627738952637, "global_step": 14254, "epoch": 160, "lr": 9.93654557819236e-05}
{"train_loss": 0.4819105565547943, "global_step": 14255, "epoch": 160, "lr": 9.936536371046364e-05}
{"train_loss": 0.5151529312133789, "global_step": 14256, "epoch": 160, "lr": 9.936527163236708e-05}
{"train_loss": 0.6956825256347656, "global_step": 14257, "epoch": 160, "lr": 9.936517954763396e-05}
{"train_loss": 0.5128851532936096, "global_step": 14258, "epoch": 160, "lr": 9.93650874562643e-05}
{"train_loss": 0.5912039875984192, "global_step": 14259, "epoch": 160, "lr": 9.93649953582581e-05}
{"train_loss": 0.6242245435714722, "global_step": 14260, "epoch": 160, "lr": 9.936490325361538e-05}
{"train_loss": 0.6293252110481262, "global_step": 14261, "epoch": 160, "lr": 9.936481114233615e-05}
{"train_loss": 0.6623892188072205, "global_step": 14262, "epoch": 160, "lr": 9.936471902442044e-05}
{"train_loss": 0.5289056301116943, "global_step": 14263, "epoch": 160, "lr": 9.936462689986823e-05}
{"train_loss": 0.6002652645111084, "global_step": 14264, "epoch": 160, "lr": 9.936453476867955e-05}
{"train_loss": 0.40552660822868347, "global_step": 14265, "epoch": 160, "lr": 9.93644426308544e-05}
{"train_loss": 0.4411642849445343, "global_step": 14266, "epoch": 160, "lr": 9.936435048639282e-05}
{"train_loss": 0.6875830292701721, "global_step": 14267, "epoch": 160, "lr": 9.936425833529479e-05}
{"train_loss": 0.5144150853157043, "global_step": 14268, "epoch": 160, "lr": 9.936416617756035e-05}
{"train_loss": 0.5845528841018677, "global_step": 14269, "epoch": 160, "lr": 9.936407401318949e-05}
{"train_loss": 0.5516965985298157, "global_step": 14270, "epoch": 160, "lr": 9.936398184218223e-05}
{"train_loss": 0.5663111209869385, "global_step": 14271, "epoch": 160, "lr": 9.93638896645386e-05}
{"train_loss": 0.46230605244636536, "global_step": 14272, "epoch": 160, "lr": 9.936379748025858e-05}
{"train_loss": 0.5812599062919617, "global_step": 14273, "epoch": 160, "lr": 9.93637052893422e-05}
{"train_loss": 0.5334903597831726, "global_step": 14274, "epoch": 160, "lr": 9.936361309178947e-05}
{"train_loss": 0.591840386390686, "global_step": 14275, "epoch": 160, "lr": 9.936352088760043e-05}
{"train_loss": 0.6043238639831543, "global_step": 14276, "epoch": 160, "lr": 9.936342867677504e-05}
{"train_loss": 0.5703452825546265, "global_step": 14277, "epoch": 160, "lr": 9.936333645931335e-05}
{"train_loss": 0.46858543157577515, "global_step": 14278, "epoch": 160, "lr": 9.936324423521535e-05}
{"train_loss": 0.47999563813209534, "global_step": 14279, "epoch": 160, "lr": 9.936315200448107e-05}
{"train_loss": 0.4967329800128937, "global_step": 14280, "epoch": 160, "lr": 9.936305976711051e-05}
{"train_loss": 0.6493597030639648, "global_step": 14281, "epoch": 160, "lr": 9.93629675231037e-05}
{"train_loss": 0.621425211429596, "global_step": 14282, "epoch": 160, "lr": 9.936287527246063e-05}
{"train_loss": 0.5833621621131897, "global_step": 14283, "epoch": 160, "lr": 9.936278301518134e-05}
{"train_loss": 0.5404159426689148, "global_step": 14284, "epoch": 160, "lr": 9.936269075126581e-05}
{"train_loss": 0.567840576171875, "global_step": 14285, "epoch": 160, "lr": 9.936259848071406e-05}
{"train_loss": 0.7001748085021973, "global_step": 14286, "epoch": 160, "lr": 9.936250620352613e-05}
{"train_loss": 0.530988872051239, "global_step": 14287, "epoch": 160, "lr": 9.9362413919702e-05}
{"train_loss": 0.6473212838172913, "global_step": 14288, "epoch": 160, "lr": 9.93623216292417e-05}
{"train_loss": 0.5276795625686646, "global_step": 14289, "epoch": 160, "lr": 9.936222933214524e-05}
{"train_loss": 0.5589232444763184, "global_step": 14290, "epoch": 160, "lr": 9.936213702841261e-05}
{"train_loss": 0.6046386957168579, "global_step": 14291, "epoch": 160, "lr": 9.936204471804387e-05}
{"train_loss": 0.659805417060852, "global_step": 14292, "epoch": 160, "lr": 9.936195240103899e-05}
{"train_loss": 0.6204354166984558, "global_step": 14293, "epoch": 160, "lr": 9.9361860077398e-05}
{"train_loss": 0.6112133264541626, "global_step": 14294, "epoch": 160, "lr": 9.936176774712091e-05}
{"train_loss": 0.551899254322052, "global_step": 14295, "epoch": 160, "lr": 9.936167541020773e-05}
{"train_loss": 0.7430596947669983, "global_step": 14296, "epoch": 160, "lr": 9.936158306665849e-05}
{"train_loss": 0.6308807730674744, "global_step": 14297, "epoch": 160, "lr": 9.936149071647317e-05}
{"train_loss": 0.6945599317550659, "global_step": 14298, "epoch": 160, "lr": 9.93613983596518e-05}
{"train_loss": 0.6366793513298035, "global_step": 14299, "epoch": 160, "lr": 9.936130599619439e-05}
{"train_loss": 0.7518055438995361, "global_step": 14300, "epoch": 160, "lr": 9.936121362610095e-05}
{"train_loss": 0.6759716272354126, "global_step": 14301, "epoch": 160, "lr": 9.936112124937152e-05}
{"train_loss": 0.5714329481124878, "global_step": 14302, "epoch": 160, "lr": 9.936102886600606e-05}
{"train_loss": 0.6836879849433899, "global_step": 14303, "epoch": 160, "lr": 9.936093647600464e-05}
{"train_loss": 0.6140333414077759, "global_step": 14304, "epoch": 160, "lr": 9.936084407936722e-05}
{"train_loss": 0.5566749572753906, "global_step": 14305, "epoch": 160, "lr": 9.936075167609384e-05}
{"train_loss": 0.8441765308380127, "global_step": 14306, "epoch": 160, "lr": 9.936065926618452e-05}
{"train_loss": 0.6113367676734924, "global_step": 14307, "epoch": 160, "lr": 9.936056684963926e-05}
{"train_loss": 0.526167631149292, "global_step": 14308, "epoch": 160, "lr": 9.936047442645806e-05}
{"train_loss": 0.6179800629615784, "global_step": 14309, "epoch": 160, "lr": 9.936038199664095e-05}
{"train_loss": 0.5606482028961182, "global_step": 14310, "epoch": 160, "lr": 9.936028956018794e-05}
{"train_loss": 0.5944267511367798, "global_step": 14311, "epoch": 160, "lr": 9.936019711709905e-05}
{"train_loss": 0.7126815319061279, "global_step": 14312, "epoch": 160, "lr": 9.936010466737428e-05}
{"train_loss": 0.6641234159469604, "global_step": 14313, "epoch": 160, "lr": 9.936001221101364e-05}
{"train_loss": 0.6343422532081604, "global_step": 14314, "epoch": 160, "lr": 9.935991974801715e-05}
{"train_loss": 0.5567868947982788, "global_step": 14315, "epoch": 160, "lr": 9.935982727838482e-05}
{"train_loss": 0.7988041043281555, "global_step": 14316, "epoch": 160, "lr": 9.935973480211665e-05}
{"train_loss": 0.5160958170890808, "global_step": 14317, "epoch": 160, "lr": 9.935964231921269e-05}
{"train_loss": 0.5383098125457764, "global_step": 14318, "epoch": 160, "lr": 9.93595498296729e-05}
{"train_loss": 0.5350624918937683, "global_step": 14319, "epoch": 160, "lr": 9.935945733349733e-05}
{"train_loss": 0.7007862329483032, "global_step": 14320, "epoch": 160, "lr": 9.9359364830686e-05}
{"train_loss": 0.7078108191490173, "global_step": 14321, "epoch": 160, "lr": 9.935927232123889e-05}
{"train_loss": 0.48527827858924866, "global_step": 14322, "epoch": 160, "lr": 9.935917980515603e-05}
{"train_loss": 0.5066618323326111, "global_step": 14323, "epoch": 160, "lr": 9.935908728243743e-05}
{"train_loss": 0.5872461795806885, "global_step": 14324, "epoch": 160, "lr": 9.935899475308311e-05}
{"train_loss": 0.7776978611946106, "global_step": 14325, "epoch": 160, "lr": 9.935890221709306e-05}
{"train_loss": 0.62644362449646, "global_step": 14326, "epoch": 160, "lr": 9.935880967446732e-05}
{"train_loss": 0.5302848219871521, "global_step": 14327, "epoch": 160, "lr": 9.935871712520588e-05}
{"train_loss": 0.5861912580688348, "global_step": 14328, "epoch": 160, "lr": 9.935862456930877e-05, "val_loss": 1.423182725906372, "train_action_mse_error": 42.22608947753906}
{"train_loss": 0.565441906452179, "global_step": 14329, "epoch": 161, "lr": 9.935853200677599e-05}
{"train_loss": 0.6439380049705505, "global_step": 14330, "epoch": 161, "lr": 9.935843943760757e-05}
{"train_loss": 0.5227763652801514, "global_step": 14331, "epoch": 161, "lr": 9.935834686180349e-05}
{"train_loss": 0.5832249522209167, "global_step": 14332, "epoch": 161, "lr": 9.935825427936378e-05}
{"train_loss": 0.6945018768310547, "global_step": 14333, "epoch": 161, "lr": 9.935816169028847e-05}
{"train_loss": 0.6426575183868408, "global_step": 14334, "epoch": 161, "lr": 9.935806909457755e-05}
{"train_loss": 0.5294561982154846, "global_step": 14335, "epoch": 161, "lr": 9.935797649223105e-05}
{"train_loss": 0.5938769578933716, "global_step": 14336, "epoch": 161, "lr": 9.935788388324895e-05}
{"train_loss": 0.6295236945152283, "global_step": 14337, "epoch": 161, "lr": 9.93577912676313e-05}
{"train_loss": 0.6033057570457458, "global_step": 14338, "epoch": 161, "lr": 9.935769864537809e-05}
{"train_loss": 0.6200969219207764, "global_step": 14339, "epoch": 161, "lr": 9.935760601648934e-05}
{"train_loss": 0.6301195621490479, "global_step": 14340, "epoch": 161, "lr": 9.935751338096506e-05}
{"train_loss": 0.5402114391326904, "global_step": 14341, "epoch": 161, "lr": 9.935742073880526e-05}
{"train_loss": 0.7862408757209778, "global_step": 14342, "epoch": 161, "lr": 9.935732809000997e-05}
{"train_loss": 0.6030019521713257, "global_step": 14343, "epoch": 161, "lr": 9.935723543457917e-05}
{"train_loss": 0.5701795816421509, "global_step": 14344, "epoch": 161, "lr": 9.93571427725129e-05}
{"train_loss": 0.6287615299224854, "global_step": 14345, "epoch": 161, "lr": 9.935705010381117e-05}
{"train_loss": 0.578221321105957, "global_step": 14346, "epoch": 161, "lr": 9.935695742847399e-05}
{"train_loss": 0.7826948165893555, "global_step": 14347, "epoch": 161, "lr": 9.935686474650135e-05}
{"train_loss": 0.5900052785873413, "global_step": 14348, "epoch": 161, "lr": 9.93567720578933e-05}
{"train_loss": 0.6028132438659668, "global_step": 14349, "epoch": 161, "lr": 9.935667936264982e-05}
{"train_loss": 0.6330615878105164, "global_step": 14350, "epoch": 161, "lr": 9.935658666077095e-05}
{"train_loss": 0.5507050156593323, "global_step": 14351, "epoch": 161, "lr": 9.935649395225668e-05}
{"train_loss": 0.6033594608306885, "global_step": 14352, "epoch": 161, "lr": 9.935640123710703e-05}
{"train_loss": 0.5735364556312561, "global_step": 14353, "epoch": 161, "lr": 9.9356308515322e-05}
{"train_loss": 0.611395537853241, "global_step": 14354, "epoch": 161, "lr": 9.935621578690164e-05}
{"train_loss": 0.5588133931159973, "global_step": 14355, "epoch": 161, "lr": 9.935612305184592e-05}
{"train_loss": 0.6904354691505432, "global_step": 14356, "epoch": 161, "lr": 9.935603031015489e-05}
{"train_loss": 0.6393323540687561, "global_step": 14357, "epoch": 161, "lr": 9.935593756182852e-05}
{"train_loss": 0.6213706135749817, "global_step": 14358, "epoch": 161, "lr": 9.935584480686688e-05}
{"train_loss": 0.6438062787055969, "global_step": 14359, "epoch": 161, "lr": 9.935575204526992e-05}
{"train_loss": 0.5757226943969727, "global_step": 14360, "epoch": 161, "lr": 9.935565927703769e-05}
{"train_loss": 0.6168684959411621, "global_step": 14361, "epoch": 161, "lr": 9.935556650217019e-05}
{"train_loss": 0.599252462387085, "global_step": 14362, "epoch": 161, "lr": 9.935547372066743e-05}
{"train_loss": 0.6617015600204468, "global_step": 14363, "epoch": 161, "lr": 9.935538093252943e-05}
{"train_loss": 0.6076201796531677, "global_step": 14364, "epoch": 161, "lr": 9.935528813775621e-05}
{"train_loss": 0.5875284075737, "global_step": 14365, "epoch": 161, "lr": 9.935519533634778e-05}
{"train_loss": 0.6457989811897278, "global_step": 14366, "epoch": 161, "lr": 9.935510252830412e-05}
{"train_loss": 0.4552757143974304, "global_step": 14367, "epoch": 161, "lr": 9.935500971362528e-05}
{"train_loss": 0.6825980544090271, "global_step": 14368, "epoch": 161, "lr": 9.935491689231127e-05}
{"train_loss": 0.5696765184402466, "global_step": 14369, "epoch": 161, "lr": 9.935482406436209e-05}
{"train_loss": 0.6495017409324646, "global_step": 14370, "epoch": 161, "lr": 9.935473122977775e-05}
{"train_loss": 0.6286050081253052, "global_step": 14371, "epoch": 161, "lr": 9.935463838855827e-05}
{"train_loss": 0.5887554287910461, "global_step": 14372, "epoch": 161, "lr": 9.935454554070365e-05}
{"train_loss": 0.9242187142372131, "global_step": 14373, "epoch": 161, "lr": 9.935445268621392e-05}
{"train_loss": 0.6701859831809998, "global_step": 14374, "epoch": 161, "lr": 9.935435982508909e-05}
{"train_loss": 0.699305534362793, "global_step": 14375, "epoch": 161, "lr": 9.935426695732917e-05}
{"train_loss": 0.5708119869232178, "global_step": 14376, "epoch": 161, "lr": 9.935417408293416e-05}
{"train_loss": 0.5295029878616333, "global_step": 14377, "epoch": 161, "lr": 9.935408120190409e-05}
{"train_loss": 0.5846819877624512, "global_step": 14378, "epoch": 161, "lr": 9.935398831423897e-05}
{"train_loss": 0.5823625326156616, "global_step": 14379, "epoch": 161, "lr": 9.93538954199388e-05}
{"train_loss": 0.6380287408828735, "global_step": 14380, "epoch": 161, "lr": 9.935380251900358e-05}
{"train_loss": 0.5578368306159973, "global_step": 14381, "epoch": 161, "lr": 9.935370961143337e-05}
{"train_loss": 0.48477011919021606, "global_step": 14382, "epoch": 161, "lr": 9.935361669722816e-05}
{"train_loss": 0.6031948328018188, "global_step": 14383, "epoch": 161, "lr": 9.935352377638795e-05}
{"train_loss": 0.6768428683280945, "global_step": 14384, "epoch": 161, "lr": 9.935343084891275e-05}
{"train_loss": 0.45512470602989197, "global_step": 14385, "epoch": 161, "lr": 9.93533379148026e-05}
{"train_loss": 0.5875521898269653, "global_step": 14386, "epoch": 161, "lr": 9.935324497405747e-05}
{"train_loss": 0.5627089142799377, "global_step": 14387, "epoch": 161, "lr": 9.935315202667743e-05}
{"train_loss": 0.6189880967140198, "global_step": 14388, "epoch": 161, "lr": 9.935305907266243e-05}
{"train_loss": 0.6238571405410767, "global_step": 14389, "epoch": 161, "lr": 9.935296611201254e-05}
{"train_loss": 0.8385093212127686, "global_step": 14390, "epoch": 161, "lr": 9.935287314472772e-05}
{"train_loss": 0.6052748560905457, "global_step": 14391, "epoch": 161, "lr": 9.935278017080802e-05}
{"train_loss": 0.6303911805152893, "global_step": 14392, "epoch": 161, "lr": 9.935268719025345e-05}
{"train_loss": 0.5729661583900452, "global_step": 14393, "epoch": 161, "lr": 9.9352594203064e-05}
{"train_loss": 0.5202829241752625, "global_step": 14394, "epoch": 161, "lr": 9.935250120923969e-05}
{"train_loss": 0.7519990801811218, "global_step": 14395, "epoch": 161, "lr": 9.935240820878056e-05}
{"train_loss": 0.5539450645446777, "global_step": 14396, "epoch": 161, "lr": 9.935231520168658e-05}
{"train_loss": 0.7545862793922424, "global_step": 14397, "epoch": 161, "lr": 9.935222218795779e-05}
{"train_loss": 0.608407199382782, "global_step": 14398, "epoch": 161, "lr": 9.935212916759419e-05}
{"train_loss": 0.8036100268363953, "global_step": 14399, "epoch": 161, "lr": 9.93520361405958e-05}
{"train_loss": 0.533786416053772, "global_step": 14400, "epoch": 161, "lr": 9.935194310696263e-05}
{"train_loss": 0.5509605407714844, "global_step": 14401, "epoch": 161, "lr": 9.935185006669469e-05}
{"train_loss": 0.6073228716850281, "global_step": 14402, "epoch": 161, "lr": 9.9351757019792e-05}
{"train_loss": 0.6879540681838989, "global_step": 14403, "epoch": 161, "lr": 9.935166396625457e-05}
{"train_loss": 0.6105840802192688, "global_step": 14404, "epoch": 161, "lr": 9.935157090608241e-05}
{"train_loss": 0.5909011960029602, "global_step": 14405, "epoch": 161, "lr": 9.935147783927554e-05}
{"train_loss": 0.6171692609786987, "global_step": 14406, "epoch": 161, "lr": 9.935138476583394e-05}
{"train_loss": 0.5347089171409607, "global_step": 14407, "epoch": 161, "lr": 9.935129168575768e-05}
{"train_loss": 0.5150366425514221, "global_step": 14408, "epoch": 161, "lr": 9.935119859904672e-05}
{"train_loss": 0.581693172454834, "global_step": 14409, "epoch": 161, "lr": 9.93511055057011e-05}
{"train_loss": 0.5967093706130981, "global_step": 14410, "epoch": 161, "lr": 9.935101240572082e-05}
{"train_loss": 0.6587197780609131, "global_step": 14411, "epoch": 161, "lr": 9.935091929910589e-05}
{"train_loss": 0.5812968015670776, "global_step": 14412, "epoch": 161, "lr": 9.935082618585634e-05}
{"train_loss": 0.5638525485992432, "global_step": 14413, "epoch": 161, "lr": 9.935073306597219e-05}
{"train_loss": 0.6029254794120789, "global_step": 14414, "epoch": 161, "lr": 9.935063993945341e-05}
{"train_loss": 0.6185382604598999, "global_step": 14415, "epoch": 161, "lr": 9.935054680630004e-05}
{"train_loss": 0.4973277747631073, "global_step": 14416, "epoch": 161, "lr": 9.93504536665121e-05}
{"train_loss": 0.6118149456013454, "global_step": 14417, "epoch": 161, "lr": 9.935036052008959e-05, "val_loss": 1.4194023609161377}
{"train_loss": 0.5136466026306152, "global_step": 14418, "epoch": 162, "lr": 9.935026736703253e-05}
{"train_loss": 0.4938209056854248, "global_step": 14419, "epoch": 162, "lr": 9.935017420734092e-05}
{"train_loss": 0.5719578862190247, "global_step": 14420, "epoch": 162, "lr": 9.93500810410148e-05}
{"train_loss": 0.6569515466690063, "global_step": 14421, "epoch": 162, "lr": 9.934998786805414e-05}
{"train_loss": 0.5516979694366455, "global_step": 14422, "epoch": 162, "lr": 9.9349894688459e-05}
{"train_loss": 0.45058193802833557, "global_step": 14423, "epoch": 162, "lr": 9.934980150222933e-05}
{"train_loss": 0.6324252486228943, "global_step": 14424, "epoch": 162, "lr": 9.934970830936521e-05}
{"train_loss": 0.6896029710769653, "global_step": 14425, "epoch": 162, "lr": 9.93496151098666e-05}
{"train_loss": 0.5337012410163879, "global_step": 14426, "epoch": 162, "lr": 9.934952190373357e-05}
{"train_loss": 0.6249796152114868, "global_step": 14427, "epoch": 162, "lr": 9.934942869096607e-05}
{"train_loss": 0.5020102262496948, "global_step": 14428, "epoch": 162, "lr": 9.934933547156416e-05}
{"train_loss": 0.4633846879005432, "global_step": 14429, "epoch": 162, "lr": 9.934924224552782e-05}
{"train_loss": 0.6603888273239136, "global_step": 14430, "epoch": 162, "lr": 9.934914901285708e-05}
{"train_loss": 0.6407515406608582, "global_step": 14431, "epoch": 162, "lr": 9.934905577355194e-05}
{"train_loss": 0.5407363772392273, "global_step": 14432, "epoch": 162, "lr": 9.934896252761243e-05}
{"train_loss": 0.5527727603912354, "global_step": 14433, "epoch": 162, "lr": 9.934886927503855e-05}
{"train_loss": 0.6375913619995117, "global_step": 14434, "epoch": 162, "lr": 9.934877601583033e-05}
{"train_loss": 0.5410363078117371, "global_step": 14435, "epoch": 162, "lr": 9.934868274998774e-05}
{"train_loss": 0.49847927689552307, "global_step": 14436, "epoch": 162, "lr": 9.934858947751084e-05}
{"train_loss": 0.6869262456893921, "global_step": 14437, "epoch": 162, "lr": 9.934849619839963e-05}
{"train_loss": 0.742489755153656, "global_step": 14438, "epoch": 162, "lr": 9.93484029126541e-05}
{"train_loss": 0.43521416187286377, "global_step": 14439, "epoch": 162, "lr": 9.934830962027428e-05}
{"train_loss": 0.7024666666984558, "global_step": 14440, "epoch": 162, "lr": 9.934821632126019e-05}
{"train_loss": 0.5048736929893494, "global_step": 14441, "epoch": 162, "lr": 9.934812301561184e-05}
{"train_loss": 0.440372496843338, "global_step": 14442, "epoch": 162, "lr": 9.934802970332923e-05}
{"train_loss": 0.5998551249504089, "global_step": 14443, "epoch": 162, "lr": 9.934793638441237e-05}
{"train_loss": 0.5948939919471741, "global_step": 14444, "epoch": 162, "lr": 9.934784305886128e-05}
{"train_loss": 0.640629231929779, "global_step": 14445, "epoch": 162, "lr": 9.9347749726676e-05}
{"train_loss": 0.6064806580543518, "global_step": 14446, "epoch": 162, "lr": 9.934765638785649e-05}
{"train_loss": 0.7594422698020935, "global_step": 14447, "epoch": 162, "lr": 9.93475630424028e-05}
{"train_loss": 0.820743203163147, "global_step": 14448, "epoch": 162, "lr": 9.934746969031492e-05}
{"train_loss": 0.5119377970695496, "global_step": 14449, "epoch": 162, "lr": 9.934737633159289e-05}
{"train_loss": 0.7603442072868347, "global_step": 14450, "epoch": 162, "lr": 9.93472829662367e-05}
{"train_loss": 0.5550469160079956, "global_step": 14451, "epoch": 162, "lr": 9.934718959424637e-05}
{"train_loss": 0.44736766815185547, "global_step": 14452, "epoch": 162, "lr": 9.93470962156219e-05}
{"train_loss": 0.6498633623123169, "global_step": 14453, "epoch": 162, "lr": 9.934700283036334e-05}
{"train_loss": 0.507810115814209, "global_step": 14454, "epoch": 162, "lr": 9.934690943847066e-05}
{"train_loss": 0.7549521923065186, "global_step": 14455, "epoch": 162, "lr": 9.934681603994389e-05}
{"train_loss": 0.5134849548339844, "global_step": 14456, "epoch": 162, "lr": 9.934672263478303e-05}
{"train_loss": 0.6815897822380066, "global_step": 14457, "epoch": 162, "lr": 9.934662922298813e-05}
{"train_loss": 0.6003222465515137, "global_step": 14458, "epoch": 162, "lr": 9.934653580455918e-05}
{"train_loss": 0.6318681240081787, "global_step": 14459, "epoch": 162, "lr": 9.934644237949616e-05}
{"train_loss": 0.5931785106658936, "global_step": 14460, "epoch": 162, "lr": 9.934634894779913e-05}
{"train_loss": 0.6107596158981323, "global_step": 14461, "epoch": 162, "lr": 9.934625550946808e-05}
{"train_loss": 0.598267138004303, "global_step": 14462, "epoch": 162, "lr": 9.934616206450303e-05}
{"train_loss": 0.5415075421333313, "global_step": 14463, "epoch": 162, "lr": 9.934606861290398e-05}
{"train_loss": 0.6124213337898254, "global_step": 14464, "epoch": 162, "lr": 9.934597515467097e-05}
{"train_loss": 0.5198707580566406, "global_step": 14465, "epoch": 162, "lr": 9.934588168980399e-05}
{"train_loss": 0.6489596366882324, "global_step": 14466, "epoch": 162, "lr": 9.934578821830305e-05}
{"train_loss": 0.5778582096099854, "global_step": 14467, "epoch": 162, "lr": 9.934569474016818e-05}
{"train_loss": 0.48582521080970764, "global_step": 14468, "epoch": 162, "lr": 9.934560125539938e-05}
{"train_loss": 0.5355929136276245, "global_step": 14469, "epoch": 162, "lr": 9.934550776399667e-05}
{"train_loss": 0.594267725944519, "global_step": 14470, "epoch": 162, "lr": 9.934541426596004e-05}
{"train_loss": 0.5993705987930298, "global_step": 14471, "epoch": 162, "lr": 9.934532076128952e-05}
{"train_loss": 0.5014576315879822, "global_step": 14472, "epoch": 162, "lr": 9.934522724998515e-05}
{"train_loss": 0.5814771056175232, "global_step": 14473, "epoch": 162, "lr": 9.93451337320469e-05}
{"train_loss": 0.542158842086792, "global_step": 14474, "epoch": 162, "lr": 9.934504020747479e-05}
{"train_loss": 0.5501937866210938, "global_step": 14475, "epoch": 162, "lr": 9.934494667626886e-05}
{"train_loss": 0.6684937477111816, "global_step": 14476, "epoch": 162, "lr": 9.93448531384291e-05}
{"train_loss": 0.661342442035675, "global_step": 14477, "epoch": 162, "lr": 9.93447595939555e-05}
{"train_loss": 0.598179817199707, "global_step": 14478, "epoch": 162, "lr": 9.934466604284813e-05}
{"train_loss": 0.4855043292045593, "global_step": 14479, "epoch": 162, "lr": 9.934457248510696e-05}
{"train_loss": 0.4842836260795593, "global_step": 14480, "epoch": 162, "lr": 9.934447892073201e-05}
{"train_loss": 0.5662109851837158, "global_step": 14481, "epoch": 162, "lr": 9.93443853497233e-05}
{"train_loss": 0.6511932015419006, "global_step": 14482, "epoch": 162, "lr": 9.934429177208084e-05}
{"train_loss": 0.5102638602256775, "global_step": 14483, "epoch": 162, "lr": 9.934419818780463e-05}
{"train_loss": 0.7356672286987305, "global_step": 14484, "epoch": 162, "lr": 9.93441045968947e-05}
{"train_loss": 0.6398118734359741, "global_step": 14485, "epoch": 162, "lr": 9.934401099935106e-05}
{"train_loss": 0.41448527574539185, "global_step": 14486, "epoch": 162, "lr": 9.934391739517373e-05}
{"train_loss": 0.4720206558704376, "global_step": 14487, "epoch": 162, "lr": 9.934382378436269e-05}
{"train_loss": 0.7125245928764343, "global_step": 14488, "epoch": 162, "lr": 9.934373016691798e-05}
{"train_loss": 0.49052879214286804, "global_step": 14489, "epoch": 162, "lr": 9.934363654283961e-05}
{"train_loss": 0.5178202390670776, "global_step": 14490, "epoch": 162, "lr": 9.93435429121276e-05}
{"train_loss": 0.6675047278404236, "global_step": 14491, "epoch": 162, "lr": 9.934344927478194e-05}
{"train_loss": 0.629903256893158, "global_step": 14492, "epoch": 162, "lr": 9.934335563080265e-05}
{"train_loss": 0.6690450310707092, "global_step": 14493, "epoch": 162, "lr": 9.934326198018977e-05}
{"train_loss": 0.5872629880905151, "global_step": 14494, "epoch": 162, "lr": 9.934316832294326e-05}
{"train_loss": 0.6111056804656982, "global_step": 14495, "epoch": 162, "lr": 9.934307465906318e-05}
{"train_loss": 0.5053967833518982, "global_step": 14496, "epoch": 162, "lr": 9.934298098854953e-05}
{"train_loss": 0.6881181001663208, "global_step": 14497, "epoch": 162, "lr": 9.93428873114023e-05}
{"train_loss": 0.6905758380889893, "global_step": 14498, "epoch": 162, "lr": 9.934279362762153e-05}
{"train_loss": 0.595196008682251, "global_step": 14499, "epoch": 162, "lr": 9.934269993720722e-05}
{"train_loss": 0.5976842045783997, "global_step": 14500, "epoch": 162, "lr": 9.934260624015939e-05}
{"train_loss": 0.7450058460235596, "global_step": 14501, "epoch": 162, "lr": 9.934251253647803e-05}
{"train_loss": 0.6366068720817566, "global_step": 14502, "epoch": 162, "lr": 9.93424188261632e-05}
{"train_loss": 0.66200852394104, "global_step": 14503, "epoch": 162, "lr": 9.934232510921487e-05}
{"train_loss": 0.6105808615684509, "global_step": 14504, "epoch": 162, "lr": 9.934223138563306e-05}
{"train_loss": 0.5636234879493713, "global_step": 14505, "epoch": 162, "lr": 9.93421376554178e-05}
{"train_loss": 0.5930065700177396, "global_step": 14506, "epoch": 162, "lr": 9.934204391856908e-05, "val_loss": 1.490073561668396}
{"train_loss": 0.5530834197998047, "global_step": 14507, "epoch": 163, "lr": 9.934195017508694e-05}
{"train_loss": 0.49175554513931274, "global_step": 14508, "epoch": 163, "lr": 9.934185642497136e-05}
{"train_loss": 0.5552443861961365, "global_step": 14509, "epoch": 163, "lr": 9.934176266822237e-05}
{"train_loss": 0.6257532835006714, "global_step": 14510, "epoch": 163, "lr": 9.934166890483997e-05}
{"train_loss": 0.6089444756507874, "global_step": 14511, "epoch": 163, "lr": 9.934157513482422e-05}
{"train_loss": 0.5886108875274658, "global_step": 14512, "epoch": 163, "lr": 9.934148135817508e-05}
{"train_loss": 0.562217652797699, "global_step": 14513, "epoch": 163, "lr": 9.934138757489256e-05}
{"train_loss": 0.7086755037307739, "global_step": 14514, "epoch": 163, "lr": 9.934129378497672e-05}
{"train_loss": 0.6338227391242981, "global_step": 14515, "epoch": 163, "lr": 9.934119998842752e-05}
{"train_loss": 0.5131396055221558, "global_step": 14516, "epoch": 163, "lr": 9.934110618524501e-05}
{"train_loss": 0.578058123588562, "global_step": 14517, "epoch": 163, "lr": 9.934101237542918e-05}
{"train_loss": 0.5300442576408386, "global_step": 14518, "epoch": 163, "lr": 9.934091855898007e-05}
{"train_loss": 0.6145420670509338, "global_step": 14519, "epoch": 163, "lr": 9.934082473589765e-05}
{"train_loss": 0.6748289465904236, "global_step": 14520, "epoch": 163, "lr": 9.9340730906182e-05}
{"train_loss": 0.5221357941627502, "global_step": 14521, "epoch": 163, "lr": 9.934063706983305e-05}
{"train_loss": 0.7120501399040222, "global_step": 14522, "epoch": 163, "lr": 9.934054322685085e-05}
{"train_loss": 0.5113293528556824, "global_step": 14523, "epoch": 163, "lr": 9.934044937723544e-05}
{"train_loss": 0.5748364329338074, "global_step": 14524, "epoch": 163, "lr": 9.934035552098679e-05}
{"train_loss": 0.7068690061569214, "global_step": 14525, "epoch": 163, "lr": 9.934026165810495e-05}
{"train_loss": 0.5819352865219116, "global_step": 14526, "epoch": 163, "lr": 9.93401677885899e-05}
{"train_loss": 0.5360000729560852, "global_step": 14527, "epoch": 163, "lr": 9.934007391244167e-05}
{"train_loss": 0.6544800400733948, "global_step": 14528, "epoch": 163, "lr": 9.933998002966027e-05}
{"train_loss": 0.6541268825531006, "global_step": 14529, "epoch": 163, "lr": 9.93398861402457e-05}
{"train_loss": 0.6186385154724121, "global_step": 14530, "epoch": 163, "lr": 9.9339792244198e-05}
{"train_loss": 0.5862392783164978, "global_step": 14531, "epoch": 163, "lr": 9.933969834151715e-05}
{"train_loss": 0.612450122833252, "global_step": 14532, "epoch": 163, "lr": 9.933960443220319e-05}
{"train_loss": 0.597969651222229, "global_step": 14533, "epoch": 163, "lr": 9.933951051625612e-05}
{"train_loss": 0.5603490471839905, "global_step": 14534, "epoch": 163, "lr": 9.933941659367594e-05}
{"train_loss": 0.46771103143692017, "global_step": 14535, "epoch": 163, "lr": 9.933932266446269e-05}
{"train_loss": 0.4841069281101227, "global_step": 14536, "epoch": 163, "lr": 9.933922872861637e-05}
{"train_loss": 0.6131680607795715, "global_step": 14537, "epoch": 163, "lr": 9.933913478613698e-05}
{"train_loss": 0.7564041018486023, "global_step": 14538, "epoch": 163, "lr": 9.933904083702456e-05}
{"train_loss": 0.7022756934165955, "global_step": 14539, "epoch": 163, "lr": 9.933894688127912e-05}
{"train_loss": 0.6698248386383057, "global_step": 14540, "epoch": 163, "lr": 9.933885291890064e-05}
{"train_loss": 0.5515847206115723, "global_step": 14541, "epoch": 163, "lr": 9.933875894988914e-05}
{"train_loss": 0.6208322644233704, "global_step": 14542, "epoch": 163, "lr": 9.933866497424467e-05}
{"train_loss": 0.5663110613822937, "global_step": 14543, "epoch": 163, "lr": 9.933857099196721e-05}
{"train_loss": 0.5468145608901978, "global_step": 14544, "epoch": 163, "lr": 9.933847700305677e-05}
{"train_loss": 0.5709540247917175, "global_step": 14545, "epoch": 163, "lr": 9.933838300751338e-05}
{"train_loss": 0.6527804136276245, "global_step": 14546, "epoch": 163, "lr": 9.933828900533705e-05}
{"train_loss": 0.5063411593437195, "global_step": 14547, "epoch": 163, "lr": 9.933819499652778e-05}
{"train_loss": 0.6049660444259644, "global_step": 14548, "epoch": 163, "lr": 9.93381009810856e-05}
{"train_loss": 0.5643860101699829, "global_step": 14549, "epoch": 163, "lr": 9.93380069590105e-05}
{"train_loss": 0.606860876083374, "global_step": 14550, "epoch": 163, "lr": 9.933791293030252e-05}
{"train_loss": 0.581429123878479, "global_step": 14551, "epoch": 163, "lr": 9.933781889496166e-05}
{"train_loss": 0.6043768525123596, "global_step": 14552, "epoch": 163, "lr": 9.933772485298794e-05}
{"train_loss": 0.6280593872070312, "global_step": 14553, "epoch": 163, "lr": 9.933763080438135e-05}
{"train_loss": 0.5618174076080322, "global_step": 14554, "epoch": 163, "lr": 9.933753674914192e-05}
{"train_loss": 0.4986894726753235, "global_step": 14555, "epoch": 163, "lr": 9.933744268726966e-05}
{"train_loss": 0.6959380507469177, "global_step": 14556, "epoch": 163, "lr": 9.933734861876457e-05}
{"train_loss": 0.5028665065765381, "global_step": 14557, "epoch": 163, "lr": 9.933725454362668e-05}
{"train_loss": 0.550906777381897, "global_step": 14558, "epoch": 163, "lr": 9.9337160461856e-05}
{"train_loss": 0.40338802337646484, "global_step": 14559, "epoch": 163, "lr": 9.933706637345255e-05}
{"train_loss": 0.6166831851005554, "global_step": 14560, "epoch": 163, "lr": 9.933697227841632e-05}
{"train_loss": 0.7800492644309998, "global_step": 14561, "epoch": 163, "lr": 9.933687817674734e-05}
{"train_loss": 0.6606752276420593, "global_step": 14562, "epoch": 163, "lr": 9.933678406844563e-05}
{"train_loss": 0.5958535671234131, "global_step": 14563, "epoch": 163, "lr": 9.933668995351118e-05}
{"train_loss": 0.5160720348358154, "global_step": 14564, "epoch": 163, "lr": 9.9336595831944e-05}
{"train_loss": 0.5329193472862244, "global_step": 14565, "epoch": 163, "lr": 9.933650170374414e-05}
{"train_loss": 0.6614697575569153, "global_step": 14566, "epoch": 163, "lr": 9.933640756891158e-05}
{"train_loss": 0.4386151134967804, "global_step": 14567, "epoch": 163, "lr": 9.933631342744632e-05}
{"train_loss": 0.4985990822315216, "global_step": 14568, "epoch": 163, "lr": 9.933621927934842e-05}
{"train_loss": 0.5673736929893494, "global_step": 14569, "epoch": 163, "lr": 9.933612512461786e-05}
{"train_loss": 0.6252302527427673, "global_step": 14570, "epoch": 163, "lr": 9.933603096325465e-05}
{"train_loss": 0.6265109777450562, "global_step": 14571, "epoch": 163, "lr": 9.933593679525883e-05}
{"train_loss": 0.6873307228088379, "global_step": 14572, "epoch": 163, "lr": 9.933584262063039e-05}
{"train_loss": 0.5264468789100647, "global_step": 14573, "epoch": 163, "lr": 9.933574843936934e-05}
{"train_loss": 0.5862289667129517, "global_step": 14574, "epoch": 163, "lr": 9.93356542514757e-05}
{"train_loss": 0.5043281316757202, "global_step": 14575, "epoch": 163, "lr": 9.933556005694949e-05}
{"train_loss": 0.6264834403991699, "global_step": 14576, "epoch": 163, "lr": 9.93354658557907e-05}
{"train_loss": 0.5847455859184265, "global_step": 14577, "epoch": 163, "lr": 9.933537164799937e-05}
{"train_loss": 0.7713796496391296, "global_step": 14578, "epoch": 163, "lr": 9.93352774335755e-05}
{"train_loss": 0.5523746013641357, "global_step": 14579, "epoch": 163, "lr": 9.93351832125191e-05}
{"train_loss": 0.4630403220653534, "global_step": 14580, "epoch": 163, "lr": 9.933508898483019e-05}
{"train_loss": 0.5188698172569275, "global_step": 14581, "epoch": 163, "lr": 9.933499475050878e-05}
{"train_loss": 0.7733935117721558, "global_step": 14582, "epoch": 163, "lr": 9.933490050955487e-05}
{"train_loss": 0.5483993887901306, "global_step": 14583, "epoch": 163, "lr": 9.933480626196849e-05}
{"train_loss": 0.6371415257453918, "global_step": 14584, "epoch": 163, "lr": 9.933471200774966e-05}
{"train_loss": 0.5836954116821289, "global_step": 14585, "epoch": 163, "lr": 9.933461774689836e-05}
{"train_loss": 0.44381386041641235, "global_step": 14586, "epoch": 163, "lr": 9.933452347941464e-05}
{"train_loss": 0.5308372378349304, "global_step": 14587, "epoch": 163, "lr": 9.933442920529848e-05}
{"train_loss": 0.6060454845428467, "global_step": 14588, "epoch": 163, "lr": 9.933433492454991e-05}
{"train_loss": 0.5947468280792236, "global_step": 14589, "epoch": 163, "lr": 9.933424063716894e-05}
{"train_loss": 0.7263567447662354, "global_step": 14590, "epoch": 163, "lr": 9.933414634315559e-05}
{"train_loss": 0.6130236983299255, "global_step": 14591, "epoch": 163, "lr": 9.933405204250985e-05}
{"train_loss": 0.5247085690498352, "global_step": 14592, "epoch": 163, "lr": 9.933395773523176e-05}
{"train_loss": 0.53275066614151, "global_step": 14593, "epoch": 163, "lr": 9.933386342132132e-05}
{"train_loss": 0.6744090914726257, "global_step": 14594, "epoch": 163, "lr": 9.933376910077854e-05}
{"train_loss": 0.5892880183927128, "global_step": 14595, "epoch": 163, "lr": 9.933367477360344e-05, "val_loss": 1.520076870918274}
{"train_loss": 0.6806752681732178, "global_step": 14596, "epoch": 164, "lr": 9.933358043979602e-05}
{"train_loss": 0.6005325317382812, "global_step": 14597, "epoch": 164, "lr": 9.933348609935632e-05}
{"train_loss": 0.5460187196731567, "global_step": 14598, "epoch": 164, "lr": 9.93333917522843e-05}
{"train_loss": 0.5069721937179565, "global_step": 14599, "epoch": 164, "lr": 9.933329739858004e-05}
{"train_loss": 0.7487534284591675, "global_step": 14600, "epoch": 164, "lr": 9.933320303824352e-05}
{"train_loss": 0.7137700319290161, "global_step": 14601, "epoch": 164, "lr": 9.933310867127472e-05}
{"train_loss": 0.6709799766540527, "global_step": 14602, "epoch": 164, "lr": 9.933301429767371e-05}
{"train_loss": 0.5097874999046326, "global_step": 14603, "epoch": 164, "lr": 9.933291991744047e-05}
{"train_loss": 0.5300686955451965, "global_step": 14604, "epoch": 164, "lr": 9.933282553057504e-05}
{"train_loss": 0.550701379776001, "global_step": 14605, "epoch": 164, "lr": 9.933273113707738e-05}
{"train_loss": 0.533120334148407, "global_step": 14606, "epoch": 164, "lr": 9.933263673694755e-05}
{"train_loss": 0.5816096663475037, "global_step": 14607, "epoch": 164, "lr": 9.933254233018557e-05}
{"train_loss": 0.44576624035835266, "global_step": 14608, "epoch": 164, "lr": 9.933244791679139e-05}
{"train_loss": 0.5518883466720581, "global_step": 14609, "epoch": 164, "lr": 9.933235349676508e-05}
{"train_loss": 0.4319337010383606, "global_step": 14610, "epoch": 164, "lr": 9.933225907010665e-05}
{"train_loss": 0.588917076587677, "global_step": 14611, "epoch": 164, "lr": 9.933216463681609e-05}
{"train_loss": 0.44694048166275024, "global_step": 14612, "epoch": 164, "lr": 9.933207019689343e-05}
{"train_loss": 0.5872092843055725, "global_step": 14613, "epoch": 164, "lr": 9.933197575033866e-05}
{"train_loss": 0.5536901950836182, "global_step": 14614, "epoch": 164, "lr": 9.933188129715181e-05}
{"train_loss": 0.6953352689743042, "global_step": 14615, "epoch": 164, "lr": 9.933178683733288e-05}
{"train_loss": 0.5881175994873047, "global_step": 14616, "epoch": 164, "lr": 9.933169237088192e-05}
{"train_loss": 0.5401513576507568, "global_step": 14617, "epoch": 164, "lr": 9.933159789779889e-05}
{"train_loss": 0.6080410480499268, "global_step": 14618, "epoch": 164, "lr": 9.933150341808383e-05}
{"train_loss": 0.5467125773429871, "global_step": 14619, "epoch": 164, "lr": 9.933140893173678e-05}
{"train_loss": 0.5414236783981323, "global_step": 14620, "epoch": 164, "lr": 9.93313144387577e-05}
{"train_loss": 0.7005405426025391, "global_step": 14621, "epoch": 164, "lr": 9.933121993914662e-05}
{"train_loss": 0.5803654789924622, "global_step": 14622, "epoch": 164, "lr": 9.933112543290357e-05}
{"train_loss": 0.5499739646911621, "global_step": 14623, "epoch": 164, "lr": 9.933103092002856e-05}
{"train_loss": 0.5292850732803345, "global_step": 14624, "epoch": 164, "lr": 9.933093640052158e-05}
{"train_loss": 0.43836021423339844, "global_step": 14625, "epoch": 164, "lr": 9.933084187438265e-05}
{"train_loss": 0.4155896306037903, "global_step": 14626, "epoch": 164, "lr": 9.933074734161181e-05}
{"train_loss": 0.6119685769081116, "global_step": 14627, "epoch": 164, "lr": 9.933065280220903e-05}
{"train_loss": 0.6570819616317749, "global_step": 14628, "epoch": 164, "lr": 9.933055825617436e-05}
{"train_loss": 0.5062772631645203, "global_step": 14629, "epoch": 164, "lr": 9.933046370350779e-05}
{"train_loss": 0.42950794100761414, "global_step": 14630, "epoch": 164, "lr": 9.933036914420934e-05}
{"train_loss": 0.5851600766181946, "global_step": 14631, "epoch": 164, "lr": 9.933027457827903e-05}
{"train_loss": 0.45180147886276245, "global_step": 14632, "epoch": 164, "lr": 9.933018000571686e-05}
{"train_loss": 0.5063091516494751, "global_step": 14633, "epoch": 164, "lr": 9.933008542652286e-05}
{"train_loss": 0.6150704026222229, "global_step": 14634, "epoch": 164, "lr": 9.932999084069703e-05}
{"train_loss": 0.49042391777038574, "global_step": 14635, "epoch": 164, "lr": 9.932989624823936e-05}
{"train_loss": 0.6017520427703857, "global_step": 14636, "epoch": 164, "lr": 9.932980164914992e-05}
{"train_loss": 0.6519514918327332, "global_step": 14637, "epoch": 164, "lr": 9.932970704342867e-05}
{"train_loss": 0.474247008562088, "global_step": 14638, "epoch": 164, "lr": 9.932961243107563e-05}
{"train_loss": 0.5586119890213013, "global_step": 14639, "epoch": 164, "lr": 9.932951781209085e-05}
{"train_loss": 0.5889889001846313, "global_step": 14640, "epoch": 164, "lr": 9.932942318647432e-05}
{"train_loss": 0.5411620140075684, "global_step": 14641, "epoch": 164, "lr": 9.932932855422603e-05}
{"train_loss": 0.7005869150161743, "global_step": 14642, "epoch": 164, "lr": 9.932923391534602e-05}
{"train_loss": 0.6097623705863953, "global_step": 14643, "epoch": 164, "lr": 9.93291392698343e-05}
{"train_loss": 0.5519989132881165, "global_step": 14644, "epoch": 164, "lr": 9.932904461769087e-05}
{"train_loss": 0.4844556450843811, "global_step": 14645, "epoch": 164, "lr": 9.932894995891575e-05}
{"train_loss": 0.5322414040565491, "global_step": 14646, "epoch": 164, "lr": 9.932885529350897e-05}
{"train_loss": 0.5740675330162048, "global_step": 14647, "epoch": 164, "lr": 9.932876062147052e-05}
{"train_loss": 0.5619201064109802, "global_step": 14648, "epoch": 164, "lr": 9.932866594280041e-05}
{"train_loss": 0.6545993089675903, "global_step": 14649, "epoch": 164, "lr": 9.932857125749867e-05}
{"train_loss": 0.5279046297073364, "global_step": 14650, "epoch": 164, "lr": 9.93284765655653e-05}
{"train_loss": 0.7276427149772644, "global_step": 14651, "epoch": 164, "lr": 9.932838186700034e-05}
{"train_loss": 0.6084688901901245, "global_step": 14652, "epoch": 164, "lr": 9.932828716180374e-05}
{"train_loss": 0.5514941215515137, "global_step": 14653, "epoch": 164, "lr": 9.932819244997557e-05}
{"train_loss": 0.6095291972160339, "global_step": 14654, "epoch": 164, "lr": 9.932809773151584e-05}
{"train_loss": 0.5758880376815796, "global_step": 14655, "epoch": 164, "lr": 9.932800300642453e-05}
{"train_loss": 0.5058582425117493, "global_step": 14656, "epoch": 164, "lr": 9.932790827470169e-05}
{"train_loss": 0.556857705116272, "global_step": 14657, "epoch": 164, "lr": 9.93278135363473e-05}
{"train_loss": 0.6319604516029358, "global_step": 14658, "epoch": 164, "lr": 9.932771879136139e-05}
{"train_loss": 0.5125384330749512, "global_step": 14659, "epoch": 164, "lr": 9.932762403974397e-05}
{"train_loss": 0.580696702003479, "global_step": 14660, "epoch": 164, "lr": 9.932752928149505e-05}
{"train_loss": 0.5904749035835266, "global_step": 14661, "epoch": 164, "lr": 9.932743451661464e-05}
{"train_loss": 0.6805998682975769, "global_step": 14662, "epoch": 164, "lr": 9.932733974510277e-05}
{"train_loss": 0.6001399159431458, "global_step": 14663, "epoch": 164, "lr": 9.932724496695944e-05}
{"train_loss": 0.5105435848236084, "global_step": 14664, "epoch": 164, "lr": 9.932715018218466e-05}
{"train_loss": 0.60280841588974, "global_step": 14665, "epoch": 164, "lr": 9.932705539077845e-05}
{"train_loss": 0.7467639446258545, "global_step": 14666, "epoch": 164, "lr": 9.93269605927408e-05}
{"train_loss": 0.5191172361373901, "global_step": 14667, "epoch": 164, "lr": 9.932686578807176e-05}
{"train_loss": 0.565041720867157, "global_step": 14668, "epoch": 164, "lr": 9.932677097677132e-05}
{"train_loss": 0.6044852137565613, "global_step": 14669, "epoch": 164, "lr": 9.932667615883948e-05}
{"train_loss": 0.7085718512535095, "global_step": 14670, "epoch": 164, "lr": 9.93265813342763e-05}
{"train_loss": 0.47605305910110474, "global_step": 14671, "epoch": 164, "lr": 9.932648650308173e-05}
{"train_loss": 0.6118934154510498, "global_step": 14672, "epoch": 164, "lr": 9.932639166525584e-05}
{"train_loss": 0.604545533657074, "global_step": 14673, "epoch": 164, "lr": 9.932629682079862e-05}
{"train_loss": 0.6513945460319519, "global_step": 14674, "epoch": 164, "lr": 9.932620196971006e-05}
{"train_loss": 0.5884395241737366, "global_step": 14675, "epoch": 164, "lr": 9.932610711199021e-05}
{"train_loss": 0.7032256126403809, "global_step": 14676, "epoch": 164, "lr": 9.932601224763905e-05}
{"train_loss": 0.5909284353256226, "global_step": 14677, "epoch": 164, "lr": 9.932591737665662e-05}
{"train_loss": 0.6447497010231018, "global_step": 14678, "epoch": 164, "lr": 9.932582249904292e-05}
{"train_loss": 0.657915472984314, "global_step": 14679, "epoch": 164, "lr": 9.932572761479796e-05}
{"train_loss": 0.6335562467575073, "global_step": 14680, "epoch": 164, "lr": 9.932563272392177e-05}
{"train_loss": 0.6105373501777649, "global_step": 14681, "epoch": 164, "lr": 9.932553782641433e-05}
{"train_loss": 0.675708532333374, "global_step": 14682, "epoch": 164, "lr": 9.932544292227571e-05}
{"train_loss": 0.6076328158378601, "global_step": 14683, "epoch": 164, "lr": 9.932534801150583e-05}
{"train_loss": 0.579182618454601, "global_step": 14684, "epoch": 164, "lr": 9.932525309410481e-05, "val_loss": 1.543041467666626}
{"train_loss": 0.7108409404754639, "global_step": 14685, "epoch": 165, "lr": 9.932515817007257e-05}
{"train_loss": 0.6619461178779602, "global_step": 14686, "epoch": 165, "lr": 9.93250632394092e-05}
{"train_loss": 0.5597589015960693, "global_step": 14687, "epoch": 165, "lr": 9.932496830211465e-05}
{"train_loss": 0.6796613931655884, "global_step": 14688, "epoch": 165, "lr": 9.932487335818896e-05}
{"train_loss": 0.5423716902732849, "global_step": 14689, "epoch": 165, "lr": 9.932477840763216e-05}
{"train_loss": 0.4850464165210724, "global_step": 14690, "epoch": 165, "lr": 9.932468345044423e-05}
{"train_loss": 0.5678598284721375, "global_step": 14691, "epoch": 165, "lr": 9.932458848662521e-05}
{"train_loss": 0.5229485034942627, "global_step": 14692, "epoch": 165, "lr": 9.932449351617508e-05}
{"train_loss": 0.4661375880241394, "global_step": 14693, "epoch": 165, "lr": 9.93243985390939e-05}
{"train_loss": 0.6493561863899231, "global_step": 14694, "epoch": 165, "lr": 9.932430355538164e-05}
{"train_loss": 0.6004564166069031, "global_step": 14695, "epoch": 165, "lr": 9.932420856503833e-05}
{"train_loss": 0.593498945236206, "global_step": 14696, "epoch": 165, "lr": 9.932411356806397e-05}
{"train_loss": 0.5428162813186646, "global_step": 14697, "epoch": 165, "lr": 9.93240185644586e-05}
{"train_loss": 0.4573926329612732, "global_step": 14698, "epoch": 165, "lr": 9.932392355422222e-05}
{"train_loss": 0.5921272039413452, "global_step": 14699, "epoch": 165, "lr": 9.932382853735482e-05}
{"train_loss": 0.501282811164856, "global_step": 14700, "epoch": 165, "lr": 9.932373351385644e-05}
{"train_loss": 0.47832903265953064, "global_step": 14701, "epoch": 165, "lr": 9.93236384837271e-05}
{"train_loss": 0.576802670955658, "global_step": 14702, "epoch": 165, "lr": 9.932354344696679e-05}
{"train_loss": 0.6837756633758545, "global_step": 14703, "epoch": 165, "lr": 9.932344840357553e-05}
{"train_loss": 0.6698995232582092, "global_step": 14704, "epoch": 165, "lr": 9.932335335355331e-05}
{"train_loss": 0.5948545336723328, "global_step": 14705, "epoch": 165, "lr": 9.93232582969002e-05}
{"train_loss": 0.45204058289527893, "global_step": 14706, "epoch": 165, "lr": 9.932316323361617e-05}
{"train_loss": 0.5375763773918152, "global_step": 14707, "epoch": 165, "lr": 9.932306816370124e-05}
{"train_loss": 0.5606665015220642, "global_step": 14708, "epoch": 165, "lr": 9.932297308715542e-05}
{"train_loss": 0.6119230389595032, "global_step": 14709, "epoch": 165, "lr": 9.932287800397872e-05}
{"train_loss": 0.48289239406585693, "global_step": 14710, "epoch": 165, "lr": 9.932278291417119e-05}
{"train_loss": 0.5709458589553833, "global_step": 14711, "epoch": 165, "lr": 9.932268781773279e-05}
{"train_loss": 0.6014664173126221, "global_step": 14712, "epoch": 165, "lr": 9.932259271466356e-05}
{"train_loss": 0.639193594455719, "global_step": 14713, "epoch": 165, "lr": 9.932249760496351e-05}
{"train_loss": 0.5996144413948059, "global_step": 14714, "epoch": 165, "lr": 9.932240248863266e-05}
{"train_loss": 0.5835738778114319, "global_step": 14715, "epoch": 165, "lr": 9.932230736567098e-05}
{"train_loss": 0.5271846652030945, "global_step": 14716, "epoch": 165, "lr": 9.932221223607855e-05}
{"train_loss": 0.6275262832641602, "global_step": 14717, "epoch": 165, "lr": 9.932211709985535e-05}
{"train_loss": 0.6026502847671509, "global_step": 14718, "epoch": 165, "lr": 9.932202195700137e-05}
{"train_loss": 0.7315854430198669, "global_step": 14719, "epoch": 165, "lr": 9.932192680751666e-05}
{"train_loss": 0.7177290916442871, "global_step": 14720, "epoch": 165, "lr": 9.932183165140122e-05}
{"train_loss": 0.5141576528549194, "global_step": 14721, "epoch": 165, "lr": 9.932173648865506e-05}
{"train_loss": 0.6648055911064148, "global_step": 14722, "epoch": 165, "lr": 9.93216413192782e-05}
{"train_loss": 0.6292104721069336, "global_step": 14723, "epoch": 165, "lr": 9.932154614327063e-05}
{"train_loss": 0.7463061809539795, "global_step": 14724, "epoch": 165, "lr": 9.93214509606324e-05}
{"train_loss": 0.5133131742477417, "global_step": 14725, "epoch": 165, "lr": 9.932135577136348e-05}
{"train_loss": 0.5344591736793518, "global_step": 14726, "epoch": 165, "lr": 9.932126057546391e-05}
{"train_loss": 0.5600481629371643, "global_step": 14727, "epoch": 165, "lr": 9.93211653729337e-05}
{"train_loss": 0.6293319463729858, "global_step": 14728, "epoch": 165, "lr": 9.932107016377286e-05}
{"train_loss": 0.6332987546920776, "global_step": 14729, "epoch": 165, "lr": 9.93209749479814e-05}
{"train_loss": 0.7255675196647644, "global_step": 14730, "epoch": 165, "lr": 9.932087972555934e-05}
{"train_loss": 0.4334680736064911, "global_step": 14731, "epoch": 165, "lr": 9.93207844965067e-05}
{"train_loss": 0.5578600168228149, "global_step": 14732, "epoch": 165, "lr": 9.932068926082347e-05}
{"train_loss": 0.5782204270362854, "global_step": 14733, "epoch": 165, "lr": 9.932059401850968e-05}
{"train_loss": 0.5484742522239685, "global_step": 14734, "epoch": 165, "lr": 9.932049876956533e-05}
{"train_loss": 0.6936945915222168, "global_step": 14735, "epoch": 165, "lr": 9.932040351399043e-05}
{"train_loss": 0.46963852643966675, "global_step": 14736, "epoch": 165, "lr": 9.932030825178502e-05}
{"train_loss": 0.6284641623497009, "global_step": 14737, "epoch": 165, "lr": 9.93202129829491e-05}
{"train_loss": 0.5599621534347534, "global_step": 14738, "epoch": 165, "lr": 9.932011770748267e-05}
{"train_loss": 0.5253949761390686, "global_step": 14739, "epoch": 165, "lr": 9.932002242538574e-05}
{"train_loss": 0.6236461400985718, "global_step": 14740, "epoch": 165, "lr": 9.931992713665837e-05}
{"train_loss": 0.605843186378479, "global_step": 14741, "epoch": 165, "lr": 9.931983184130049e-05}
{"train_loss": 0.5590335130691528, "global_step": 14742, "epoch": 165, "lr": 9.93197365393122e-05}
{"train_loss": 0.5011555552482605, "global_step": 14743, "epoch": 165, "lr": 9.931964123069345e-05}
{"train_loss": 0.5906597971916199, "global_step": 14744, "epoch": 165, "lr": 9.931954591544428e-05}
{"train_loss": 0.6590722799301147, "global_step": 14745, "epoch": 165, "lr": 9.93194505935647e-05}
{"train_loss": 0.5692490935325623, "global_step": 14746, "epoch": 165, "lr": 9.931935526505471e-05}
{"train_loss": 0.6599260568618774, "global_step": 14747, "epoch": 165, "lr": 9.931925992991436e-05}
{"train_loss": 0.42248567938804626, "global_step": 14748, "epoch": 165, "lr": 9.931916458814361e-05}
{"train_loss": 0.6066554188728333, "global_step": 14749, "epoch": 165, "lr": 9.931906923974251e-05}
{"train_loss": 0.5460789799690247, "global_step": 14750, "epoch": 165, "lr": 9.931897388471106e-05}
{"train_loss": 0.5706334710121155, "global_step": 14751, "epoch": 165, "lr": 9.931887852304928e-05}
{"train_loss": 0.5877978801727295, "global_step": 14752, "epoch": 165, "lr": 9.931878315475718e-05}
{"train_loss": 0.5424565076828003, "global_step": 14753, "epoch": 165, "lr": 9.931868777983475e-05}
{"train_loss": 0.6104937195777893, "global_step": 14754, "epoch": 165, "lr": 9.931859239828205e-05}
{"train_loss": 0.5954674482345581, "global_step": 14755, "epoch": 165, "lr": 9.931849701009905e-05}
{"train_loss": 0.5964445471763611, "global_step": 14756, "epoch": 165, "lr": 9.931840161528579e-05}
{"train_loss": 0.6595931053161621, "global_step": 14757, "epoch": 165, "lr": 9.931830621384226e-05}
{"train_loss": 0.5145174860954285, "global_step": 14758, "epoch": 165, "lr": 9.93182108057685e-05}
{"train_loss": 0.6455801725387573, "global_step": 14759, "epoch": 165, "lr": 9.93181153910645e-05}
{"train_loss": 0.4942404627799988, "global_step": 14760, "epoch": 165, "lr": 9.931801996973027e-05}
{"train_loss": 0.5533029437065125, "global_step": 14761, "epoch": 165, "lr": 9.931792454176586e-05}
{"train_loss": 0.5701584815979004, "global_step": 14762, "epoch": 165, "lr": 9.931782910717122e-05}
{"train_loss": 0.6848986148834229, "global_step": 14763, "epoch": 165, "lr": 9.931773366594643e-05}
{"train_loss": 0.633801281452179, "global_step": 14764, "epoch": 165, "lr": 9.931763821809145e-05}
{"train_loss": 0.6795745491981506, "global_step": 14765, "epoch": 165, "lr": 9.931754276360633e-05}
{"train_loss": 0.5695688724517822, "global_step": 14766, "epoch": 165, "lr": 9.931744730249107e-05}
{"train_loss": 0.5010613799095154, "global_step": 14767, "epoch": 165, "lr": 9.931735183474566e-05}
{"train_loss": 0.4970446825027466, "global_step": 14768, "epoch": 165, "lr": 9.931725636037015e-05}
{"train_loss": 0.6366216540336609, "global_step": 14769, "epoch": 165, "lr": 9.931716087936453e-05}
{"train_loss": 0.4974431097507477, "global_step": 14770, "epoch": 165, "lr": 9.931706539172883e-05}
{"train_loss": 0.6867523789405823, "global_step": 14771, "epoch": 165, "lr": 9.931696989746303e-05}
{"train_loss": 0.5384587049484253, "global_step": 14772, "epoch": 165, "lr": 9.931687439656718e-05}
{"train_loss": 0.583473344197434, "global_step": 14773, "epoch": 165, "lr": 9.931677888904127e-05, "val_loss": 1.4291603565216064, "train_action_mse_error": 37.476070404052734}
{"train_loss": 0.4810726046562195, "global_step": 14774, "epoch": 166, "lr": 9.931668337488533e-05}
{"train_loss": 0.4701584577560425, "global_step": 14775, "epoch": 166, "lr": 9.931658785409935e-05}
{"train_loss": 0.5320632457733154, "global_step": 14776, "epoch": 166, "lr": 9.931649232668336e-05}
{"train_loss": 0.4494275152683258, "global_step": 14777, "epoch": 166, "lr": 9.931639679263738e-05}
{"train_loss": 0.576023280620575, "global_step": 14778, "epoch": 166, "lr": 9.93163012519614e-05}
{"train_loss": 0.625613808631897, "global_step": 14779, "epoch": 166, "lr": 9.931620570465545e-05}
{"train_loss": 0.6420559883117676, "global_step": 14780, "epoch": 166, "lr": 9.931611015071953e-05}
{"train_loss": 0.5800631642341614, "global_step": 14781, "epoch": 166, "lr": 9.931601459015367e-05}
{"train_loss": 0.5437807440757751, "global_step": 14782, "epoch": 166, "lr": 9.931591902295787e-05}
{"train_loss": 0.6351301074028015, "global_step": 14783, "epoch": 166, "lr": 9.931582344913214e-05}
{"train_loss": 0.49750664830207825, "global_step": 14784, "epoch": 166, "lr": 9.93157278686765e-05}
{"train_loss": 0.44632452726364136, "global_step": 14785, "epoch": 166, "lr": 9.931563228159097e-05}
{"train_loss": 0.549516499042511, "global_step": 14786, "epoch": 166, "lr": 9.931553668787554e-05}
{"train_loss": 0.4441530704498291, "global_step": 14787, "epoch": 166, "lr": 9.931544108753024e-05}
{"train_loss": 0.5622656345367432, "global_step": 14788, "epoch": 166, "lr": 9.931534548055509e-05}
{"train_loss": 0.46077466011047363, "global_step": 14789, "epoch": 166, "lr": 9.93152498669501e-05}
{"train_loss": 0.49445152282714844, "global_step": 14790, "epoch": 166, "lr": 9.931515424671525e-05}
{"train_loss": 0.4982879161834717, "global_step": 14791, "epoch": 166, "lr": 9.93150586198506e-05}
{"train_loss": 0.5485636591911316, "global_step": 14792, "epoch": 166, "lr": 9.931496298635612e-05}
{"train_loss": 0.6007758378982544, "global_step": 14793, "epoch": 166, "lr": 9.931486734623187e-05}
{"train_loss": 0.5018124580383301, "global_step": 14794, "epoch": 166, "lr": 9.931477169947783e-05}
{"train_loss": 0.5315340757369995, "global_step": 14795, "epoch": 166, "lr": 9.9314676046094e-05}
{"train_loss": 0.5537141561508179, "global_step": 14796, "epoch": 166, "lr": 9.931458038608044e-05}
{"train_loss": 0.6767634749412537, "global_step": 14797, "epoch": 166, "lr": 9.931448471943712e-05}
{"train_loss": 0.531303882598877, "global_step": 14798, "epoch": 166, "lr": 9.931438904616406e-05}
{"train_loss": 0.5732958316802979, "global_step": 14799, "epoch": 166, "lr": 9.93142933662613e-05}
{"train_loss": 0.4625512659549713, "global_step": 14800, "epoch": 166, "lr": 9.931419767972883e-05}
{"train_loss": 0.5231189131736755, "global_step": 14801, "epoch": 166, "lr": 9.931410198656666e-05}
{"train_loss": 0.5917080640792847, "global_step": 14802, "epoch": 166, "lr": 9.931400628677482e-05}
{"train_loss": 0.4798279106616974, "global_step": 14803, "epoch": 166, "lr": 9.931391058035331e-05}
{"train_loss": 0.5211565494537354, "global_step": 14804, "epoch": 166, "lr": 9.931381486730214e-05}
{"train_loss": 0.5746932029724121, "global_step": 14805, "epoch": 166, "lr": 9.931371914762133e-05}
{"train_loss": 0.6344767808914185, "global_step": 14806, "epoch": 166, "lr": 9.93136234213109e-05}
{"train_loss": 0.5456686615943909, "global_step": 14807, "epoch": 166, "lr": 9.931352768837084e-05}
{"train_loss": 0.48599153757095337, "global_step": 14808, "epoch": 166, "lr": 9.931343194880118e-05}
{"train_loss": 0.6246039867401123, "global_step": 14809, "epoch": 166, "lr": 9.931333620260195e-05}
{"train_loss": 0.6036439538002014, "global_step": 14810, "epoch": 166, "lr": 9.931324044977313e-05}
{"train_loss": 0.5112437009811401, "global_step": 14811, "epoch": 166, "lr": 9.931314469031475e-05}
{"train_loss": 0.44811007380485535, "global_step": 14812, "epoch": 166, "lr": 9.93130489242268e-05}
{"train_loss": 0.6391032338142395, "global_step": 14813, "epoch": 166, "lr": 9.931295315150933e-05}
{"train_loss": 0.5649231672286987, "global_step": 14814, "epoch": 166, "lr": 9.931285737216232e-05}
{"train_loss": 0.5349335074424744, "global_step": 14815, "epoch": 166, "lr": 9.931276158618581e-05}
{"train_loss": 0.6535549759864807, "global_step": 14816, "epoch": 166, "lr": 9.93126657935798e-05}
{"train_loss": 0.5542948842048645, "global_step": 14817, "epoch": 166, "lr": 9.931256999434431e-05}
{"train_loss": 0.6344330310821533, "global_step": 14818, "epoch": 166, "lr": 9.931247418847934e-05}
{"train_loss": 0.4895196557044983, "global_step": 14819, "epoch": 166, "lr": 9.931237837598488e-05}
{"train_loss": 0.5271275043487549, "global_step": 14820, "epoch": 166, "lr": 9.9312282556861e-05}
{"train_loss": 0.5111405849456787, "global_step": 14821, "epoch": 166, "lr": 9.93121867311077e-05}
{"train_loss": 0.5268901586532593, "global_step": 14822, "epoch": 166, "lr": 9.931209089872495e-05}
{"train_loss": 0.5629583597183228, "global_step": 14823, "epoch": 166, "lr": 9.931199505971281e-05}
{"train_loss": 0.6403532028198242, "global_step": 14824, "epoch": 166, "lr": 9.931189921407124e-05}
{"train_loss": 0.4821324944496155, "global_step": 14825, "epoch": 166, "lr": 9.931180336180031e-05}
{"train_loss": 0.580089807510376, "global_step": 14826, "epoch": 166, "lr": 9.931170750290001e-05}
{"train_loss": 0.720382571220398, "global_step": 14827, "epoch": 166, "lr": 9.931161163737035e-05}
{"train_loss": 0.6198159456253052, "global_step": 14828, "epoch": 166, "lr": 9.931151576521135e-05}
{"train_loss": 0.6649536490440369, "global_step": 14829, "epoch": 166, "lr": 9.9311419886423e-05}
{"train_loss": 0.6456806659698486, "global_step": 14830, "epoch": 166, "lr": 9.931132400100536e-05}
{"train_loss": 0.5619558691978455, "global_step": 14831, "epoch": 166, "lr": 9.931122810895838e-05}
{"train_loss": 0.4960813820362091, "global_step": 14832, "epoch": 166, "lr": 9.931113221028214e-05}
{"train_loss": 0.49265527725219727, "global_step": 14833, "epoch": 166, "lr": 9.93110363049766e-05}
{"train_loss": 0.598562479019165, "global_step": 14834, "epoch": 166, "lr": 9.931094039304178e-05}
{"train_loss": 0.9289146065711975, "global_step": 14835, "epoch": 166, "lr": 9.931084447447774e-05}
{"train_loss": 0.5414323210716248, "global_step": 14836, "epoch": 166, "lr": 9.931074854928443e-05}
{"train_loss": 0.8013622760772705, "global_step": 14837, "epoch": 166, "lr": 9.93106526174619e-05}
{"train_loss": 0.5545066595077515, "global_step": 14838, "epoch": 166, "lr": 9.931055667901015e-05}
{"train_loss": 0.5436901450157166, "global_step": 14839, "epoch": 166, "lr": 9.93104607339292e-05}
{"train_loss": 0.6351765394210815, "global_step": 14840, "epoch": 166, "lr": 9.931036478221906e-05}
{"train_loss": 0.7233222126960754, "global_step": 14841, "epoch": 166, "lr": 9.931026882387974e-05}
{"train_loss": 0.6406972408294678, "global_step": 14842, "epoch": 166, "lr": 9.931017285891125e-05}
{"train_loss": 0.6493272185325623, "global_step": 14843, "epoch": 166, "lr": 9.931007688731362e-05}
{"train_loss": 0.6912887692451477, "global_step": 14844, "epoch": 166, "lr": 9.930998090908685e-05}
{"train_loss": 0.586776614189148, "global_step": 14845, "epoch": 166, "lr": 9.930988492423095e-05}
{"train_loss": 0.7279803156852722, "global_step": 14846, "epoch": 166, "lr": 9.930978893274592e-05}
{"train_loss": 0.544410228729248, "global_step": 14847, "epoch": 166, "lr": 9.930969293463181e-05}
{"train_loss": 0.547448992729187, "global_step": 14848, "epoch": 166, "lr": 9.930959692988861e-05}
{"train_loss": 0.5127905607223511, "global_step": 14849, "epoch": 166, "lr": 9.930950091851634e-05}
{"train_loss": 0.5453413128852844, "global_step": 14850, "epoch": 166, "lr": 9.9309404900515e-05}
{"train_loss": 0.6794370412826538, "global_step": 14851, "epoch": 166, "lr": 9.930930887588461e-05}
{"train_loss": 0.6396112442016602, "global_step": 14852, "epoch": 166, "lr": 9.930921284462519e-05}
{"train_loss": 0.486503541469574, "global_step": 14853, "epoch": 166, "lr": 9.930911680673676e-05}
{"train_loss": 0.5412616729736328, "global_step": 14854, "epoch": 166, "lr": 9.93090207622193e-05}
{"train_loss": 0.6489898562431335, "global_step": 14855, "epoch": 166, "lr": 9.930892471107284e-05}
{"train_loss": 0.6836255192756653, "global_step": 14856, "epoch": 166, "lr": 9.930882865329742e-05}
{"train_loss": 0.6346325874328613, "global_step": 14857, "epoch": 166, "lr": 9.930873258889301e-05}
{"train_loss": 0.6000891327857971, "global_step": 14858, "epoch": 166, "lr": 9.930863651785964e-05}
{"train_loss": 0.5896270871162415, "global_step": 14859, "epoch": 166, "lr": 9.930854044019733e-05}
{"train_loss": 0.5408502817153931, "global_step": 14860, "epoch": 166, "lr": 9.93084443559061e-05}
{"train_loss": 0.5365665555000305, "global_step": 14861, "epoch": 166, "lr": 9.930834826498593e-05}
{"train_loss": 0.573977493168263, "global_step": 14862, "epoch": 166, "lr": 9.930825216743688e-05, "val_loss": 1.4046858549118042}
{"train_loss": 0.526712954044342, "global_step": 14863, "epoch": 167, "lr": 9.930815606325892e-05}
{"train_loss": 0.5322062969207764, "global_step": 14864, "epoch": 167, "lr": 9.930805995245207e-05}
{"train_loss": 0.629514217376709, "global_step": 14865, "epoch": 167, "lr": 9.930796383501637e-05}
{"train_loss": 0.5476025938987732, "global_step": 14866, "epoch": 167, "lr": 9.93078677109518e-05}
{"train_loss": 0.607285737991333, "global_step": 14867, "epoch": 167, "lr": 9.93077715802584e-05}
{"train_loss": 0.6263710260391235, "global_step": 14868, "epoch": 167, "lr": 9.930767544293616e-05}
{"train_loss": 0.6066243052482605, "global_step": 14869, "epoch": 167, "lr": 9.930757929898511e-05}
{"train_loss": 0.7934208512306213, "global_step": 14870, "epoch": 167, "lr": 9.930748314840525e-05}
{"train_loss": 0.5762987732887268, "global_step": 14871, "epoch": 167, "lr": 9.930738699119662e-05}
{"train_loss": 0.6083313226699829, "global_step": 14872, "epoch": 167, "lr": 9.93072908273592e-05}
{"train_loss": 0.5289016962051392, "global_step": 14873, "epoch": 167, "lr": 9.930719465689302e-05}
{"train_loss": 0.621965229511261, "global_step": 14874, "epoch": 167, "lr": 9.930709847979809e-05}
{"train_loss": 0.5805101990699768, "global_step": 14875, "epoch": 167, "lr": 9.930700229607441e-05}
{"train_loss": 0.4815561771392822, "global_step": 14876, "epoch": 167, "lr": 9.9306906105722e-05}
{"train_loss": 0.4829659163951874, "global_step": 14877, "epoch": 167, "lr": 9.930680990874089e-05}
{"train_loss": 0.5353783369064331, "global_step": 14878, "epoch": 167, "lr": 9.930671370513109e-05}
{"train_loss": 0.5370696187019348, "global_step": 14879, "epoch": 167, "lr": 9.930661749489258e-05}
{"train_loss": 0.4407719373703003, "global_step": 14880, "epoch": 167, "lr": 9.930652127802543e-05}
{"train_loss": 0.4772050082683563, "global_step": 14881, "epoch": 167, "lr": 9.930642505452958e-05}
{"train_loss": 0.5439882874488831, "global_step": 14882, "epoch": 167, "lr": 9.93063288244051e-05}
{"train_loss": 0.6463662385940552, "global_step": 14883, "epoch": 167, "lr": 9.930623258765199e-05}
{"train_loss": 0.5713883638381958, "global_step": 14884, "epoch": 167, "lr": 9.930613634427024e-05}
{"train_loss": 0.7056092619895935, "global_step": 14885, "epoch": 167, "lr": 9.93060400942599e-05}
{"train_loss": 0.39177951216697693, "global_step": 14886, "epoch": 167, "lr": 9.930594383762095e-05}
{"train_loss": 0.5187594890594482, "global_step": 14887, "epoch": 167, "lr": 9.930584757435343e-05}
{"train_loss": 0.5704083442687988, "global_step": 14888, "epoch": 167, "lr": 9.930575130445734e-05}
{"train_loss": 0.4360445439815521, "global_step": 14889, "epoch": 167, "lr": 9.930565502793268e-05}
{"train_loss": 0.6700303554534912, "global_step": 14890, "epoch": 167, "lr": 9.930555874477949e-05}
{"train_loss": 0.5775814056396484, "global_step": 14891, "epoch": 167, "lr": 9.930546245499776e-05}
{"train_loss": 0.5920810699462891, "global_step": 14892, "epoch": 167, "lr": 9.93053661585875e-05}
{"train_loss": 0.632902204990387, "global_step": 14893, "epoch": 167, "lr": 9.930526985554875e-05}
{"train_loss": 0.5434425473213196, "global_step": 14894, "epoch": 167, "lr": 9.930517354588151e-05}
{"train_loss": 0.5118818879127502, "global_step": 14895, "epoch": 167, "lr": 9.930507722958578e-05}
{"train_loss": 0.47993457317352295, "global_step": 14896, "epoch": 167, "lr": 9.930498090666158e-05}
{"train_loss": 0.6037983894348145, "global_step": 14897, "epoch": 167, "lr": 9.930488457710893e-05}
{"train_loss": 0.5613080263137817, "global_step": 14898, "epoch": 167, "lr": 9.930478824092786e-05}
{"train_loss": 0.6569882035255432, "global_step": 14899, "epoch": 167, "lr": 9.930469189811834e-05}
{"train_loss": 0.5896961092948914, "global_step": 14900, "epoch": 167, "lr": 9.930459554868041e-05}
{"train_loss": 0.5500169396400452, "global_step": 14901, "epoch": 167, "lr": 9.930449919261408e-05}
{"train_loss": 0.6466397047042847, "global_step": 14902, "epoch": 167, "lr": 9.930440282991935e-05}
{"train_loss": 0.5140436887741089, "global_step": 14903, "epoch": 167, "lr": 9.930430646059626e-05}
{"train_loss": 0.6342443227767944, "global_step": 14904, "epoch": 167, "lr": 9.930421008464481e-05}
{"train_loss": 0.6487564444541931, "global_step": 14905, "epoch": 167, "lr": 9.9304113702065e-05}
{"train_loss": 0.6805239319801331, "global_step": 14906, "epoch": 167, "lr": 9.930401731285685e-05}
{"train_loss": 0.6495711803436279, "global_step": 14907, "epoch": 167, "lr": 9.930392091702037e-05}
{"train_loss": 0.6083467602729797, "global_step": 14908, "epoch": 167, "lr": 9.93038245145556e-05}
{"train_loss": 0.5465307235717773, "global_step": 14909, "epoch": 167, "lr": 9.930372810546251e-05}
{"train_loss": 0.6180103421211243, "global_step": 14910, "epoch": 167, "lr": 9.930363168974115e-05}
{"train_loss": 0.5438506007194519, "global_step": 14911, "epoch": 167, "lr": 9.930353526739151e-05}
{"train_loss": 0.49117541313171387, "global_step": 14912, "epoch": 167, "lr": 9.930343883841361e-05}
{"train_loss": 0.5104491710662842, "global_step": 14913, "epoch": 167, "lr": 9.930334240280746e-05}
{"train_loss": 0.510017454624176, "global_step": 14914, "epoch": 167, "lr": 9.930324596057308e-05}
{"train_loss": 0.6579702496528625, "global_step": 14915, "epoch": 167, "lr": 9.930314951171049e-05}
{"train_loss": 0.46019455790519714, "global_step": 14916, "epoch": 167, "lr": 9.930305305621969e-05}
{"train_loss": 0.6056845784187317, "global_step": 14917, "epoch": 167, "lr": 9.930295659410069e-05}
{"train_loss": 0.6909500360488892, "global_step": 14918, "epoch": 167, "lr": 9.93028601253535e-05}
{"train_loss": 0.5612332820892334, "global_step": 14919, "epoch": 167, "lr": 9.930276364997815e-05}
{"train_loss": 0.6024767756462097, "global_step": 14920, "epoch": 167, "lr": 9.930266716797465e-05}
{"train_loss": 0.5795149207115173, "global_step": 14921, "epoch": 167, "lr": 9.9302570679343e-05}
{"train_loss": 0.68123459815979, "global_step": 14922, "epoch": 167, "lr": 9.930247418408323e-05}
{"train_loss": 0.5146646499633789, "global_step": 14923, "epoch": 167, "lr": 9.930237768219532e-05}
{"train_loss": 0.4301696717739105, "global_step": 14924, "epoch": 167, "lr": 9.930228117367933e-05}
{"train_loss": 0.6186493635177612, "global_step": 14925, "epoch": 167, "lr": 9.930218465853524e-05}
{"train_loss": 0.5205894112586975, "global_step": 14926, "epoch": 167, "lr": 9.930208813676307e-05}
{"train_loss": 0.5144416689872742, "global_step": 14927, "epoch": 167, "lr": 9.930199160836285e-05}
{"train_loss": 0.6369736194610596, "global_step": 14928, "epoch": 167, "lr": 9.930189507333455e-05}
{"train_loss": 0.7332953810691833, "global_step": 14929, "epoch": 167, "lr": 9.930179853167823e-05}
{"train_loss": 0.52069491147995, "global_step": 14930, "epoch": 167, "lr": 9.930170198339388e-05}
{"train_loss": 0.5526986122131348, "global_step": 14931, "epoch": 167, "lr": 9.930160542848152e-05}
{"train_loss": 0.5677222013473511, "global_step": 14932, "epoch": 167, "lr": 9.930150886694116e-05}
{"train_loss": 0.5460138320922852, "global_step": 14933, "epoch": 167, "lr": 9.930141229877282e-05}
{"train_loss": 0.6490752696990967, "global_step": 14934, "epoch": 167, "lr": 9.930131572397648e-05}
{"train_loss": 0.6609206199645996, "global_step": 14935, "epoch": 167, "lr": 9.93012191425522e-05}
{"train_loss": 0.6402597427368164, "global_step": 14936, "epoch": 167, "lr": 9.930112255449997e-05}
{"train_loss": 0.5358487367630005, "global_step": 14937, "epoch": 167, "lr": 9.93010259598198e-05}
{"train_loss": 0.5390526056289673, "global_step": 14938, "epoch": 167, "lr": 9.930092935851171e-05}
{"train_loss": 0.6222397685050964, "global_step": 14939, "epoch": 167, "lr": 9.93008327505757e-05}
{"train_loss": 0.6497440338134766, "global_step": 14940, "epoch": 167, "lr": 9.930073613601182e-05}
{"train_loss": 0.6714950799942017, "global_step": 14941, "epoch": 167, "lr": 9.930063951482005e-05}
{"train_loss": 0.5986497402191162, "global_step": 14942, "epoch": 167, "lr": 9.93005428870004e-05}
{"train_loss": 0.6921144127845764, "global_step": 14943, "epoch": 167, "lr": 9.93004462525529e-05}
{"train_loss": 0.6117801070213318, "global_step": 14944, "epoch": 167, "lr": 9.930034961147753e-05}
{"train_loss": 0.6770007610321045, "global_step": 14945, "epoch": 167, "lr": 9.930025296377436e-05}
{"train_loss": 0.6193823218345642, "global_step": 14946, "epoch": 167, "lr": 9.930015630944336e-05}
{"train_loss": 0.6063852906227112, "global_step": 14947, "epoch": 167, "lr": 9.930005964848457e-05}
{"train_loss": 0.5903319120407104, "global_step": 14948, "epoch": 167, "lr": 9.929996298089797e-05}
{"train_loss": 0.5310608744621277, "global_step": 14949, "epoch": 167, "lr": 9.92998663066836e-05}
{"train_loss": 0.5690465569496155, "global_step": 14950, "epoch": 167, "lr": 9.929976962584145e-05}
{"train_loss": 0.5808324284767836, "global_step": 14951, "epoch": 167, "lr": 9.929967293837156e-05, "val_loss": 1.5022573471069336}
{"train_loss": 0.5659045577049255, "global_step": 14952, "epoch": 168, "lr": 9.929957624427392e-05}
{"train_loss": 0.6773965358734131, "global_step": 14953, "epoch": 168, "lr": 9.929947954354857e-05}
{"train_loss": 0.490876168012619, "global_step": 14954, "epoch": 168, "lr": 9.929938283619548e-05}
{"train_loss": 0.46451589465141296, "global_step": 14955, "epoch": 168, "lr": 9.92992861222147e-05}
{"train_loss": 0.47712138295173645, "global_step": 14956, "epoch": 168, "lr": 9.929918940160625e-05}
{"train_loss": 0.5731747150421143, "global_step": 14957, "epoch": 168, "lr": 9.92990926743701e-05}
{"train_loss": 0.6278491020202637, "global_step": 14958, "epoch": 168, "lr": 9.929899594050632e-05}
{"train_loss": 0.6136631965637207, "global_step": 14959, "epoch": 168, "lr": 9.929889920001486e-05}
{"train_loss": 0.567535400390625, "global_step": 14960, "epoch": 168, "lr": 9.929880245289578e-05}
{"train_loss": 0.6418101787567139, "global_step": 14961, "epoch": 168, "lr": 9.929870569914906e-05}
{"train_loss": 0.4864230453968048, "global_step": 14962, "epoch": 168, "lr": 9.929860893877474e-05}
{"train_loss": 0.5450190901756287, "global_step": 14963, "epoch": 168, "lr": 9.929851217177283e-05}
{"train_loss": 0.6897743940353394, "global_step": 14964, "epoch": 168, "lr": 9.929841539814333e-05}
{"train_loss": 0.5099943280220032, "global_step": 14965, "epoch": 168, "lr": 9.929831861788625e-05}
{"train_loss": 0.550764799118042, "global_step": 14966, "epoch": 168, "lr": 9.929822183100162e-05}
{"train_loss": 0.6871380805969238, "global_step": 14967, "epoch": 168, "lr": 9.929812503748945e-05}
{"train_loss": 0.679020881652832, "global_step": 14968, "epoch": 168, "lr": 9.929802823734974e-05}
{"train_loss": 0.552507758140564, "global_step": 14969, "epoch": 168, "lr": 9.929793143058252e-05}
{"train_loss": 0.6530889868736267, "global_step": 14970, "epoch": 168, "lr": 9.929783461718777e-05}
{"train_loss": 0.48640158772468567, "global_step": 14971, "epoch": 168, "lr": 9.929773779716556e-05}
{"train_loss": 0.40645354986190796, "global_step": 14972, "epoch": 168, "lr": 9.929764097051585e-05}
{"train_loss": 0.5872016549110413, "global_step": 14973, "epoch": 168, "lr": 9.929754413723868e-05}
{"train_loss": 0.597387433052063, "global_step": 14974, "epoch": 168, "lr": 9.929744729733405e-05}
{"train_loss": 0.5742533802986145, "global_step": 14975, "epoch": 168, "lr": 9.929735045080199e-05}
{"train_loss": 0.6321770548820496, "global_step": 14976, "epoch": 168, "lr": 9.929725359764249e-05}
{"train_loss": 0.7133710980415344, "global_step": 14977, "epoch": 168, "lr": 9.929715673785557e-05}
{"train_loss": 0.5151124596595764, "global_step": 14978, "epoch": 168, "lr": 9.929705987144127e-05}
{"train_loss": 0.496504008769989, "global_step": 14979, "epoch": 168, "lr": 9.929696299839956e-05}
{"train_loss": 0.6174044013023376, "global_step": 14980, "epoch": 168, "lr": 9.929686611873048e-05}
{"train_loss": 0.5725860595703125, "global_step": 14981, "epoch": 168, "lr": 9.929676923243405e-05}
{"train_loss": 0.7483992576599121, "global_step": 14982, "epoch": 168, "lr": 9.929667233951026e-05}
{"train_loss": 0.5332472920417786, "global_step": 14983, "epoch": 168, "lr": 9.929657543995914e-05}
{"train_loss": 0.6358857750892639, "global_step": 14984, "epoch": 168, "lr": 9.929647853378068e-05}
{"train_loss": 0.6012200117111206, "global_step": 14985, "epoch": 168, "lr": 9.929638162097493e-05}
{"train_loss": 0.5713412165641785, "global_step": 14986, "epoch": 168, "lr": 9.929628470154186e-05}
{"train_loss": 0.591994047164917, "global_step": 14987, "epoch": 168, "lr": 9.929618777548153e-05}
{"train_loss": 0.5494917631149292, "global_step": 14988, "epoch": 168, "lr": 9.92960908427939e-05}
{"train_loss": 0.5734666585922241, "global_step": 14989, "epoch": 168, "lr": 9.929599390347904e-05}
{"train_loss": 0.5484699010848999, "global_step": 14990, "epoch": 168, "lr": 9.929589695753692e-05}
{"train_loss": 0.5962521433830261, "global_step": 14991, "epoch": 168, "lr": 9.929580000496756e-05}
{"train_loss": 0.4923703968524933, "global_step": 14992, "epoch": 168, "lr": 9.929570304577099e-05}
{"train_loss": 0.5632309317588806, "global_step": 14993, "epoch": 168, "lr": 9.92956060799472e-05}
{"train_loss": 0.6971207857131958, "global_step": 14994, "epoch": 168, "lr": 9.929550910749625e-05}
{"train_loss": 0.7226316332817078, "global_step": 14995, "epoch": 168, "lr": 9.92954121284181e-05}
{"train_loss": 0.46118679642677307, "global_step": 14996, "epoch": 168, "lr": 9.929531514271278e-05}
{"train_loss": 0.6519832015037537, "global_step": 14997, "epoch": 168, "lr": 9.929521815038029e-05}
{"train_loss": 0.6028866767883301, "global_step": 14998, "epoch": 168, "lr": 9.92951211514207e-05}
{"train_loss": 0.5773476362228394, "global_step": 14999, "epoch": 168, "lr": 9.929502414583395e-05}
{"train_loss": 0.5143557786941528, "global_step": 15000, "epoch": 168, "lr": 9.92949271336201e-05}
{"train_loss": 0.5788694620132446, "global_step": 15001, "epoch": 168, "lr": 9.929483011477914e-05}
{"train_loss": 0.5465384125709534, "global_step": 15002, "epoch": 168, "lr": 9.929473308931109e-05}
{"train_loss": 0.745105504989624, "global_step": 15003, "epoch": 168, "lr": 9.929463605721596e-05}
{"train_loss": 0.634604275226593, "global_step": 15004, "epoch": 168, "lr": 9.929453901849377e-05}
{"train_loss": 0.6166127324104309, "global_step": 15005, "epoch": 168, "lr": 9.929444197314454e-05}
{"train_loss": 0.5916830897331238, "global_step": 15006, "epoch": 168, "lr": 9.929434492116828e-05}
{"train_loss": 0.5669229030609131, "global_step": 15007, "epoch": 168, "lr": 9.929424786256496e-05}
{"train_loss": 0.6215711832046509, "global_step": 15008, "epoch": 168, "lr": 9.929415079733466e-05}
{"train_loss": 0.5774850845336914, "global_step": 15009, "epoch": 168, "lr": 9.929405372547736e-05}
{"train_loss": 0.6089316606521606, "global_step": 15010, "epoch": 168, "lr": 9.929395664699305e-05}
{"train_loss": 0.6491150856018066, "global_step": 15011, "epoch": 168, "lr": 9.92938595618818e-05}
{"train_loss": 0.6451601982116699, "global_step": 15012, "epoch": 168, "lr": 9.929376247014357e-05}
{"train_loss": 0.5092152953147888, "global_step": 15013, "epoch": 168, "lr": 9.92936653717784e-05}
{"train_loss": 0.6554286479949951, "global_step": 15014, "epoch": 168, "lr": 9.92935682667863e-05}
{"train_loss": 0.4617208242416382, "global_step": 15015, "epoch": 168, "lr": 9.929347115516729e-05}
{"train_loss": 0.645553708076477, "global_step": 15016, "epoch": 168, "lr": 9.929337403692136e-05}
{"train_loss": 0.5949555039405823, "global_step": 15017, "epoch": 168, "lr": 9.929327691204855e-05}
{"train_loss": 0.619475781917572, "global_step": 15018, "epoch": 168, "lr": 9.929317978054885e-05}
{"train_loss": 0.6358141899108887, "global_step": 15019, "epoch": 168, "lr": 9.929308264242227e-05}
{"train_loss": 0.6844473481178284, "global_step": 15020, "epoch": 168, "lr": 9.929298549766886e-05}
{"train_loss": 0.5124427080154419, "global_step": 15021, "epoch": 168, "lr": 9.92928883462886e-05}
{"train_loss": 0.5511667132377625, "global_step": 15022, "epoch": 168, "lr": 9.929279118828151e-05}
{"train_loss": 0.49703899025917053, "global_step": 15023, "epoch": 168, "lr": 9.92926940236476e-05}
{"train_loss": 0.5539370775222778, "global_step": 15024, "epoch": 168, "lr": 9.92925968523869e-05}
{"train_loss": 0.5367745757102966, "global_step": 15025, "epoch": 168, "lr": 9.92924996744994e-05}
{"train_loss": 0.47906169295310974, "global_step": 15026, "epoch": 168, "lr": 9.929240248998513e-05}
{"train_loss": 0.7491693496704102, "global_step": 15027, "epoch": 168, "lr": 9.92923052988441e-05}
{"train_loss": 0.5229566693305969, "global_step": 15028, "epoch": 168, "lr": 9.929220810107631e-05}
{"train_loss": 0.5878578424453735, "global_step": 15029, "epoch": 168, "lr": 9.92921108966818e-05}
{"train_loss": 0.7058000564575195, "global_step": 15030, "epoch": 168, "lr": 9.929201368566054e-05}
{"train_loss": 0.48019149899482727, "global_step": 15031, "epoch": 168, "lr": 9.929191646801259e-05}
{"train_loss": 0.6016881465911865, "global_step": 15032, "epoch": 168, "lr": 9.929181924373793e-05}
{"train_loss": 0.5983667373657227, "global_step": 15033, "epoch": 168, "lr": 9.929172201283661e-05}
{"train_loss": 0.5318408012390137, "global_step": 15034, "epoch": 168, "lr": 9.92916247753086e-05}
{"train_loss": 0.6344627141952515, "global_step": 15035, "epoch": 168, "lr": 9.929152753115394e-05}
{"train_loss": 0.7051547765731812, "global_step": 15036, "epoch": 168, "lr": 9.929143028037262e-05}
{"train_loss": 0.8615303039550781, "global_step": 15037, "epoch": 168, "lr": 9.929133302296468e-05}
{"train_loss": 0.7257339358329773, "global_step": 15038, "epoch": 168, "lr": 9.929123575893011e-05}
{"train_loss": 0.5371123552322388, "global_step": 15039, "epoch": 168, "lr": 9.929113848826896e-05}
{"train_loss": 0.5908239269524478, "global_step": 15040, "epoch": 168, "lr": 9.92910412109812e-05, "val_loss": 1.5293737649917603}
{"train_loss": 0.7429355978965759, "global_step": 15041, "epoch": 169, "lr": 9.929094392706686e-05}
{"train_loss": 0.647271454334259, "global_step": 15042, "epoch": 169, "lr": 9.929084663652595e-05}
{"train_loss": 0.512273907661438, "global_step": 15043, "epoch": 169, "lr": 9.929074933935849e-05}
{"train_loss": 0.4969068169593811, "global_step": 15044, "epoch": 169, "lr": 9.929065203556449e-05}
{"train_loss": 0.6086966395378113, "global_step": 15045, "epoch": 169, "lr": 9.929055472514396e-05}
{"train_loss": 0.611963152885437, "global_step": 15046, "epoch": 169, "lr": 9.929045740809692e-05}
{"train_loss": 0.542273759841919, "global_step": 15047, "epoch": 169, "lr": 9.929036008442337e-05}
{"train_loss": 0.6328689455986023, "global_step": 15048, "epoch": 169, "lr": 9.929026275412334e-05}
{"train_loss": 0.6599496603012085, "global_step": 15049, "epoch": 169, "lr": 9.929016541719683e-05}
{"train_loss": 0.6325989365577698, "global_step": 15050, "epoch": 169, "lr": 9.929006807364387e-05}
{"train_loss": 0.6710233092308044, "global_step": 15051, "epoch": 169, "lr": 9.928997072346444e-05}
{"train_loss": 0.5959513783454895, "global_step": 15052, "epoch": 169, "lr": 9.92898733666586e-05}
{"train_loss": 0.5805402398109436, "global_step": 15053, "epoch": 169, "lr": 9.928977600322632e-05}
{"train_loss": 0.6015451550483704, "global_step": 15054, "epoch": 169, "lr": 9.928967863316764e-05}
{"train_loss": 0.54698246717453, "global_step": 15055, "epoch": 169, "lr": 9.928958125648255e-05}
{"train_loss": 0.7028962969779968, "global_step": 15056, "epoch": 169, "lr": 9.92894838731711e-05}
{"train_loss": 0.573916494846344, "global_step": 15057, "epoch": 169, "lr": 9.928938648323325e-05}
{"train_loss": 0.4818791449069977, "global_step": 15058, "epoch": 169, "lr": 9.928928908666906e-05}
{"train_loss": 0.7812778353691101, "global_step": 15059, "epoch": 169, "lr": 9.928919168347852e-05}
{"train_loss": 0.5979717373847961, "global_step": 15060, "epoch": 169, "lr": 9.928909427366165e-05}
{"train_loss": 0.6353704333305359, "global_step": 15061, "epoch": 169, "lr": 9.928899685721846e-05}
{"train_loss": 0.48252710700035095, "global_step": 15062, "epoch": 169, "lr": 9.928889943414897e-05}
{"train_loss": 0.4857500493526459, "global_step": 15063, "epoch": 169, "lr": 9.928880200445318e-05}
{"train_loss": 0.4462156593799591, "global_step": 15064, "epoch": 169, "lr": 9.928870456813113e-05}
{"train_loss": 0.53214430809021, "global_step": 15065, "epoch": 169, "lr": 9.92886071251828e-05}
{"train_loss": 0.5077048540115356, "global_step": 15066, "epoch": 169, "lr": 9.928850967560824e-05}
{"train_loss": 0.5728509426116943, "global_step": 15067, "epoch": 169, "lr": 9.928841221940741e-05}
{"train_loss": 0.521086573600769, "global_step": 15068, "epoch": 169, "lr": 9.928831475658035e-05}
{"train_loss": 0.6789336800575256, "global_step": 15069, "epoch": 169, "lr": 9.92882172871271e-05}
{"train_loss": 0.6056435704231262, "global_step": 15070, "epoch": 169, "lr": 9.928811981104764e-05}
{"train_loss": 0.535888135433197, "global_step": 15071, "epoch": 169, "lr": 9.9288022328342e-05}
{"train_loss": 0.5516397953033447, "global_step": 15072, "epoch": 169, "lr": 9.928792483901017e-05}
{"train_loss": 0.5763632655143738, "global_step": 15073, "epoch": 169, "lr": 9.92878273430522e-05}
{"train_loss": 0.5660645961761475, "global_step": 15074, "epoch": 169, "lr": 9.928772984046806e-05}
{"train_loss": 0.5552953481674194, "global_step": 15075, "epoch": 169, "lr": 9.92876323312578e-05}
{"train_loss": 0.49150753021240234, "global_step": 15076, "epoch": 169, "lr": 9.928753481542143e-05}
{"train_loss": 0.5588939189910889, "global_step": 15077, "epoch": 169, "lr": 9.928743729295892e-05}
{"train_loss": 0.5750238299369812, "global_step": 15078, "epoch": 169, "lr": 9.928733976387035e-05}
{"train_loss": 0.6261788010597229, "global_step": 15079, "epoch": 169, "lr": 9.928724222815568e-05}
{"train_loss": 0.5787011981010437, "global_step": 15080, "epoch": 169, "lr": 9.928714468581494e-05}
{"train_loss": 0.6912199854850769, "global_step": 15081, "epoch": 169, "lr": 9.928704713684815e-05}
{"train_loss": 0.5539499521255493, "global_step": 15082, "epoch": 169, "lr": 9.92869495812553e-05}
{"train_loss": 0.5860357284545898, "global_step": 15083, "epoch": 169, "lr": 9.928685201903643e-05}
{"train_loss": 0.5513251423835754, "global_step": 15084, "epoch": 169, "lr": 9.928675445019155e-05}
{"train_loss": 0.5796512365341187, "global_step": 15085, "epoch": 169, "lr": 9.928665687472067e-05}
{"train_loss": 0.6470077037811279, "global_step": 15086, "epoch": 169, "lr": 9.92865592926238e-05}
{"train_loss": 0.5188599228858948, "global_step": 15087, "epoch": 169, "lr": 9.928646170390094e-05}
{"train_loss": 0.5269956588745117, "global_step": 15088, "epoch": 169, "lr": 9.928636410855212e-05}
{"train_loss": 0.5506910085678101, "global_step": 15089, "epoch": 169, "lr": 9.928626650657736e-05}
{"train_loss": 0.4864564538002014, "global_step": 15090, "epoch": 169, "lr": 9.928616889797664e-05}
{"train_loss": 0.5407264232635498, "global_step": 15091, "epoch": 169, "lr": 9.928607128275e-05}
{"train_loss": 0.3922941982746124, "global_step": 15092, "epoch": 169, "lr": 9.928597366089747e-05}
{"train_loss": 0.6189773678779602, "global_step": 15093, "epoch": 169, "lr": 9.928587603241902e-05}
{"train_loss": 0.6297256946563721, "global_step": 15094, "epoch": 169, "lr": 9.92857783973147e-05}
{"train_loss": 0.61458420753479, "global_step": 15095, "epoch": 169, "lr": 9.92856807555845e-05}
{"train_loss": 0.5599919557571411, "global_step": 15096, "epoch": 169, "lr": 9.928558310722845e-05}
{"train_loss": 0.6732750535011292, "global_step": 15097, "epoch": 169, "lr": 9.928548545224654e-05}
{"train_loss": 0.5345983505249023, "global_step": 15098, "epoch": 169, "lr": 9.92853877906388e-05}
{"train_loss": 0.615412712097168, "global_step": 15099, "epoch": 169, "lr": 9.928529012240523e-05}
{"train_loss": 0.6476737856864929, "global_step": 15100, "epoch": 169, "lr": 9.928519244754588e-05}
{"train_loss": 0.6897861957550049, "global_step": 15101, "epoch": 169, "lr": 9.928509476606072e-05}
{"train_loss": 0.6751450896263123, "global_step": 15102, "epoch": 169, "lr": 9.928499707794978e-05}
{"train_loss": 0.4867989420890808, "global_step": 15103, "epoch": 169, "lr": 9.928489938321308e-05}
{"train_loss": 0.6671613454818726, "global_step": 15104, "epoch": 169, "lr": 9.928480168185062e-05}
{"train_loss": 0.5502452850341797, "global_step": 15105, "epoch": 169, "lr": 9.928470397386242e-05}
{"train_loss": 0.6952216625213623, "global_step": 15106, "epoch": 169, "lr": 9.928460625924849e-05}
{"train_loss": 0.556688129901886, "global_step": 15107, "epoch": 169, "lr": 9.928450853800886e-05}
{"train_loss": 0.7255715131759644, "global_step": 15108, "epoch": 169, "lr": 9.928441081014352e-05}
{"train_loss": 0.6543251276016235, "global_step": 15109, "epoch": 169, "lr": 9.928431307565249e-05}
{"train_loss": 0.6061971187591553, "global_step": 15110, "epoch": 169, "lr": 9.928421533453579e-05}
{"train_loss": 0.5259054899215698, "global_step": 15111, "epoch": 169, "lr": 9.928411758679342e-05}
{"train_loss": 0.5613830089569092, "global_step": 15112, "epoch": 169, "lr": 9.92840198324254e-05}
{"train_loss": 0.6890428066253662, "global_step": 15113, "epoch": 169, "lr": 9.928392207143173e-05}
{"train_loss": 0.526922345161438, "global_step": 15114, "epoch": 169, "lr": 9.928382430381247e-05}
{"train_loss": 0.6877683401107788, "global_step": 15115, "epoch": 169, "lr": 9.928372652956756e-05}
{"train_loss": 0.6970378756523132, "global_step": 15116, "epoch": 169, "lr": 9.92836287486971e-05}
{"train_loss": 0.662442147731781, "global_step": 15117, "epoch": 169, "lr": 9.928353096120102e-05}
{"train_loss": 0.5066357254981995, "global_step": 15118, "epoch": 169, "lr": 9.92834331670794e-05}
{"train_loss": 0.5922122597694397, "global_step": 15119, "epoch": 169, "lr": 9.92833353663322e-05}
{"train_loss": 0.6802043318748474, "global_step": 15120, "epoch": 169, "lr": 9.928323755895946e-05}
{"train_loss": 0.604177713394165, "global_step": 15121, "epoch": 169, "lr": 9.92831397449612e-05}
{"train_loss": 0.6048739552497864, "global_step": 15122, "epoch": 169, "lr": 9.928304192433742e-05}
{"train_loss": 0.6221850514411926, "global_step": 15123, "epoch": 169, "lr": 9.928294409708812e-05}
{"train_loss": 0.6290959715843201, "global_step": 15124, "epoch": 169, "lr": 9.928284626321334e-05}
{"train_loss": 0.5917274355888367, "global_step": 15125, "epoch": 169, "lr": 9.928274842271309e-05}
{"train_loss": 0.6283272504806519, "global_step": 15126, "epoch": 169, "lr": 9.928265057558737e-05}
{"train_loss": 0.6597972512245178, "global_step": 15127, "epoch": 169, "lr": 9.92825527218362e-05}
{"train_loss": 0.5489823222160339, "global_step": 15128, "epoch": 169, "lr": 9.928245486145959e-05}
{"train_loss": 0.5919393398118823, "global_step": 15129, "epoch": 169, "lr": 9.928235699445753e-05, "val_loss": 1.434165120124817}
{"train_loss": 0.5001755952835083, "global_step": 15130, "epoch": 170, "lr": 9.928225912083009e-05}
{"train_loss": 0.5862318277359009, "global_step": 15131, "epoch": 170, "lr": 9.928216124057725e-05}
{"train_loss": 0.61570805311203, "global_step": 15132, "epoch": 170, "lr": 9.928206335369901e-05}
{"train_loss": 0.510005533695221, "global_step": 15133, "epoch": 170, "lr": 9.92819654601954e-05}
{"train_loss": 0.49847105145454407, "global_step": 15134, "epoch": 170, "lr": 9.928186756006643e-05}
{"train_loss": 0.5692077279090881, "global_step": 15135, "epoch": 170, "lr": 9.928176965331213e-05}
{"train_loss": 0.6865370869636536, "global_step": 15136, "epoch": 170, "lr": 9.928167173993248e-05}
{"train_loss": 0.5149112343788147, "global_step": 15137, "epoch": 170, "lr": 9.928157381992751e-05}
{"train_loss": 0.5262855291366577, "global_step": 15138, "epoch": 170, "lr": 9.928147589329724e-05}
{"train_loss": 0.795363187789917, "global_step": 15139, "epoch": 170, "lr": 9.928137796004167e-05}
{"train_loss": 0.7030541896820068, "global_step": 15140, "epoch": 170, "lr": 9.928128002016081e-05}
{"train_loss": 0.6884533762931824, "global_step": 15141, "epoch": 170, "lr": 9.928118207365471e-05}
{"train_loss": 0.6631834506988525, "global_step": 15142, "epoch": 170, "lr": 9.928108412052333e-05}
{"train_loss": 0.6736880540847778, "global_step": 15143, "epoch": 170, "lr": 9.928098616076672e-05}
{"train_loss": 0.6074440479278564, "global_step": 15144, "epoch": 170, "lr": 9.928088819438487e-05}
{"train_loss": 0.4937501847743988, "global_step": 15145, "epoch": 170, "lr": 9.928079022137783e-05}
{"train_loss": 0.5865898728370667, "global_step": 15146, "epoch": 170, "lr": 9.928069224174556e-05}
{"train_loss": 0.5072746872901917, "global_step": 15147, "epoch": 170, "lr": 9.928059425548812e-05}
{"train_loss": 0.5768246650695801, "global_step": 15148, "epoch": 170, "lr": 9.928049626260551e-05}
{"train_loss": 0.5783888697624207, "global_step": 15149, "epoch": 170, "lr": 9.928039826309772e-05}
{"train_loss": 0.643648624420166, "global_step": 15150, "epoch": 170, "lr": 9.92803002569648e-05}
{"train_loss": 0.6651535630226135, "global_step": 15151, "epoch": 170, "lr": 9.928020224420673e-05}
{"train_loss": 0.6125861406326294, "global_step": 15152, "epoch": 170, "lr": 9.928010422482354e-05}
{"train_loss": 0.6573430895805359, "global_step": 15153, "epoch": 170, "lr": 9.928000619881525e-05}
{"train_loss": 0.640462338924408, "global_step": 15154, "epoch": 170, "lr": 9.927990816618185e-05}
{"train_loss": 0.7165203094482422, "global_step": 15155, "epoch": 170, "lr": 9.927981012692337e-05}
{"train_loss": 0.594382107257843, "global_step": 15156, "epoch": 170, "lr": 9.927971208103983e-05}
{"train_loss": 0.5008316040039062, "global_step": 15157, "epoch": 170, "lr": 9.927961402853124e-05}
{"train_loss": 0.5603163242340088, "global_step": 15158, "epoch": 170, "lr": 9.927951596939757e-05}
{"train_loss": 0.588549017906189, "global_step": 15159, "epoch": 170, "lr": 9.927941790363891e-05}
{"train_loss": 0.67779141664505, "global_step": 15160, "epoch": 170, "lr": 9.92793198312552e-05}
{"train_loss": 0.6738063097000122, "global_step": 15161, "epoch": 170, "lr": 9.927922175224651e-05}
{"train_loss": 0.5601404309272766, "global_step": 15162, "epoch": 170, "lr": 9.927912366661282e-05}
{"train_loss": 0.5386675596237183, "global_step": 15163, "epoch": 170, "lr": 9.927902557435415e-05}
{"train_loss": 0.3832114338874817, "global_step": 15164, "epoch": 170, "lr": 9.927892747547052e-05}
{"train_loss": 0.6679965257644653, "global_step": 15165, "epoch": 170, "lr": 9.927882936996193e-05}
{"train_loss": 0.5223448276519775, "global_step": 15166, "epoch": 170, "lr": 9.927873125782841e-05}
{"train_loss": 0.6981101632118225, "global_step": 15167, "epoch": 170, "lr": 9.927863313906997e-05}
{"train_loss": 0.6576523780822754, "global_step": 15168, "epoch": 170, "lr": 9.927853501368663e-05}
{"train_loss": 0.5403045415878296, "global_step": 15169, "epoch": 170, "lr": 9.927843688167836e-05}
{"train_loss": 0.557367742061615, "global_step": 15170, "epoch": 170, "lr": 9.927833874304522e-05}
{"train_loss": 0.5793939232826233, "global_step": 15171, "epoch": 170, "lr": 9.927824059778721e-05}
{"train_loss": 0.5967021584510803, "global_step": 15172, "epoch": 170, "lr": 9.927814244590435e-05}
{"train_loss": 0.6562501788139343, "global_step": 15173, "epoch": 170, "lr": 9.927804428739662e-05}
{"train_loss": 0.738536536693573, "global_step": 15174, "epoch": 170, "lr": 9.927794612226408e-05}
{"train_loss": 0.6530978083610535, "global_step": 15175, "epoch": 170, "lr": 9.927784795050669e-05}
{"train_loss": 0.5573851466178894, "global_step": 15176, "epoch": 170, "lr": 9.927774977212453e-05}
{"train_loss": 0.6529092788696289, "global_step": 15177, "epoch": 170, "lr": 9.927765158711757e-05}
{"train_loss": 0.6280550360679626, "global_step": 15178, "epoch": 170, "lr": 9.927755339548581e-05}
{"train_loss": 0.5415347814559937, "global_step": 15179, "epoch": 170, "lr": 9.927745519722931e-05}
{"train_loss": 0.6447288393974304, "global_step": 15180, "epoch": 170, "lr": 9.927735699234804e-05}
{"train_loss": 0.5574048161506653, "global_step": 15181, "epoch": 170, "lr": 9.927725878084203e-05}
{"train_loss": 0.5097194314002991, "global_step": 15182, "epoch": 170, "lr": 9.92771605627113e-05}
{"train_loss": 0.5984818339347839, "global_step": 15183, "epoch": 170, "lr": 9.927706233795585e-05}
{"train_loss": 0.554824948310852, "global_step": 15184, "epoch": 170, "lr": 9.92769641065757e-05}
{"train_loss": 0.5416259169578552, "global_step": 15185, "epoch": 170, "lr": 9.927686586857087e-05}
{"train_loss": 0.46846234798431396, "global_step": 15186, "epoch": 170, "lr": 9.927676762394136e-05}
{"train_loss": 0.4664890170097351, "global_step": 15187, "epoch": 170, "lr": 9.927666937268719e-05}
{"train_loss": 0.6942718029022217, "global_step": 15188, "epoch": 170, "lr": 9.927657111480838e-05}
{"train_loss": 0.6260614395141602, "global_step": 15189, "epoch": 170, "lr": 9.927647285030492e-05}
{"train_loss": 0.5639322400093079, "global_step": 15190, "epoch": 170, "lr": 9.927637457917684e-05}
{"train_loss": 0.5590907335281372, "global_step": 15191, "epoch": 170, "lr": 9.927627630142418e-05}
{"train_loss": 0.5852599740028381, "global_step": 15192, "epoch": 170, "lr": 9.92761780170469e-05}
{"train_loss": 0.5062641501426697, "global_step": 15193, "epoch": 170, "lr": 9.927607972604504e-05}
{"train_loss": 0.5916682481765747, "global_step": 15194, "epoch": 170, "lr": 9.927598142841861e-05}
{"train_loss": 0.5464128255844116, "global_step": 15195, "epoch": 170, "lr": 9.927588312416763e-05}
{"train_loss": 0.48663660883903503, "global_step": 15196, "epoch": 170, "lr": 9.92757848132921e-05}
{"train_loss": 0.6140210032463074, "global_step": 15197, "epoch": 170, "lr": 9.927568649579206e-05}
{"train_loss": 0.5497487783432007, "global_step": 15198, "epoch": 170, "lr": 9.927558817166749e-05}
{"train_loss": 0.46083864569664, "global_step": 15199, "epoch": 170, "lr": 9.927548984091842e-05}
{"train_loss": 0.6807816028594971, "global_step": 15200, "epoch": 170, "lr": 9.927539150354486e-05}
{"train_loss": 0.6010056138038635, "global_step": 15201, "epoch": 170, "lr": 9.927529315954682e-05}
{"train_loss": 0.721548318862915, "global_step": 15202, "epoch": 170, "lr": 9.927519480892434e-05}
{"train_loss": 0.5565977096557617, "global_step": 15203, "epoch": 170, "lr": 9.927509645167739e-05}
{"train_loss": 0.4688047468662262, "global_step": 15204, "epoch": 170, "lr": 9.9274998087806e-05}
{"train_loss": 0.7109732031822205, "global_step": 15205, "epoch": 170, "lr": 9.927489971731022e-05}
{"train_loss": 0.6006402969360352, "global_step": 15206, "epoch": 170, "lr": 9.927480134018998e-05}
{"train_loss": 0.6470199227333069, "global_step": 15207, "epoch": 170, "lr": 9.927470295644537e-05}
{"train_loss": 0.5742716193199158, "global_step": 15208, "epoch": 170, "lr": 9.927460456607639e-05}
{"train_loss": 0.6834936141967773, "global_step": 15209, "epoch": 170, "lr": 9.927450616908303e-05}
{"train_loss": 0.5457908511161804, "global_step": 15210, "epoch": 170, "lr": 9.92744077654653e-05}
{"train_loss": 0.7211412191390991, "global_step": 15211, "epoch": 170, "lr": 9.927430935522324e-05}
{"train_loss": 0.5786306858062744, "global_step": 15212, "epoch": 170, "lr": 9.927421093835685e-05}
{"train_loss": 0.6194082498550415, "global_step": 15213, "epoch": 170, "lr": 9.927411251486613e-05}
{"train_loss": 0.45079994201660156, "global_step": 15214, "epoch": 170, "lr": 9.927401408475112e-05}
{"train_loss": 0.49767163395881653, "global_step": 15215, "epoch": 170, "lr": 9.927391564801181e-05}
{"train_loss": 0.6414108872413635, "global_step": 15216, "epoch": 170, "lr": 9.927381720464823e-05}
{"train_loss": 0.6026417016983032, "global_step": 15217, "epoch": 170, "lr": 9.927371875466039e-05}
{"train_loss": 0.5909683543644594, "global_step": 15218, "epoch": 170, "lr": 9.927362029804827e-05, "val_loss": 1.5432101488113403, "train_action_mse_error": 37.90612030029297}
{"train_loss": 0.5400490164756775, "global_step": 15219, "epoch": 171, "lr": 9.927352183481193e-05}
{"train_loss": 0.7852362990379333, "global_step": 15220, "epoch": 171, "lr": 9.927342336495138e-05}
{"train_loss": 0.6477695107460022, "global_step": 15221, "epoch": 171, "lr": 9.927332488846662e-05}
{"train_loss": 0.5685753226280212, "global_step": 15222, "epoch": 171, "lr": 9.927322640535763e-05}
{"train_loss": 0.5303955674171448, "global_step": 15223, "epoch": 171, "lr": 9.927312791562449e-05}
{"train_loss": 0.6218953728675842, "global_step": 15224, "epoch": 171, "lr": 9.927302941926715e-05}
{"train_loss": 0.5835320949554443, "global_step": 15225, "epoch": 171, "lr": 9.927293091628566e-05}
{"train_loss": 0.6719571948051453, "global_step": 15226, "epoch": 171, "lr": 9.927283240668004e-05}
{"train_loss": 0.46909841895103455, "global_step": 15227, "epoch": 171, "lr": 9.927273389045029e-05}
{"train_loss": 0.6939631700515747, "global_step": 15228, "epoch": 171, "lr": 9.92726353675964e-05}
{"train_loss": 0.6423390507698059, "global_step": 15229, "epoch": 171, "lr": 9.927253683811841e-05}
{"train_loss": 0.6714325547218323, "global_step": 15230, "epoch": 171, "lr": 9.927243830201634e-05}
{"train_loss": 0.5862335562705994, "global_step": 15231, "epoch": 171, "lr": 9.927233975929018e-05}
{"train_loss": 0.5281774401664734, "global_step": 15232, "epoch": 171, "lr": 9.927224120993997e-05}
{"train_loss": 0.6605531573295593, "global_step": 15233, "epoch": 171, "lr": 9.927214265396569e-05}
{"train_loss": 0.6123499274253845, "global_step": 15234, "epoch": 171, "lr": 9.927204409136738e-05}
{"train_loss": 0.5302808284759521, "global_step": 15235, "epoch": 171, "lr": 9.927194552214504e-05}
{"train_loss": 0.5988384485244751, "global_step": 15236, "epoch": 171, "lr": 9.927184694629869e-05}
{"train_loss": 0.572295606136322, "global_step": 15237, "epoch": 171, "lr": 9.927174836382834e-05}
{"train_loss": 0.7071625590324402, "global_step": 15238, "epoch": 171, "lr": 9.9271649774734e-05}
{"train_loss": 0.591866135597229, "global_step": 15239, "epoch": 171, "lr": 9.92715511790157e-05}
{"train_loss": 0.5542284250259399, "global_step": 15240, "epoch": 171, "lr": 9.927145257667344e-05}
{"train_loss": 0.4828896224498749, "global_step": 15241, "epoch": 171, "lr": 9.927135396770721e-05}
{"train_loss": 0.5526266694068909, "global_step": 15242, "epoch": 171, "lr": 9.927125535211705e-05}
{"train_loss": 0.5356143712997437, "global_step": 15243, "epoch": 171, "lr": 9.927115672990299e-05}
{"train_loss": 0.5640773773193359, "global_step": 15244, "epoch": 171, "lr": 9.927105810106502e-05}
{"train_loss": 0.5216689705848694, "global_step": 15245, "epoch": 171, "lr": 9.927095946560315e-05}
{"train_loss": 0.48473525047302246, "global_step": 15246, "epoch": 171, "lr": 9.92708608235174e-05}
{"train_loss": 0.5310811400413513, "global_step": 15247, "epoch": 171, "lr": 9.92707621748078e-05}
{"train_loss": 0.5757229924201965, "global_step": 15248, "epoch": 171, "lr": 9.927066351947431e-05}
{"train_loss": 0.5411919355392456, "global_step": 15249, "epoch": 171, "lr": 9.927056485751702e-05}
{"train_loss": 0.5862025618553162, "global_step": 15250, "epoch": 171, "lr": 9.927046618893587e-05}
{"train_loss": 0.5147805213928223, "global_step": 15251, "epoch": 171, "lr": 9.927036751373093e-05}
{"train_loss": 0.5016749501228333, "global_step": 15252, "epoch": 171, "lr": 9.927026883190218e-05}
{"train_loss": 0.6460278034210205, "global_step": 15253, "epoch": 171, "lr": 9.927017014344965e-05}
{"train_loss": 0.7040598392486572, "global_step": 15254, "epoch": 171, "lr": 9.927007144837334e-05}
{"train_loss": 0.775993824005127, "global_step": 15255, "epoch": 171, "lr": 9.926997274667326e-05}
{"train_loss": 0.6988188624382019, "global_step": 15256, "epoch": 171, "lr": 9.926987403834946e-05}
{"train_loss": 0.6266003847122192, "global_step": 15257, "epoch": 171, "lr": 9.926977532340189e-05}
{"train_loss": 0.5583553314208984, "global_step": 15258, "epoch": 171, "lr": 9.926967660183063e-05}
{"train_loss": 0.5855540633201599, "global_step": 15259, "epoch": 171, "lr": 9.926957787363565e-05}
{"train_loss": 0.5994338393211365, "global_step": 15260, "epoch": 171, "lr": 9.926947913881696e-05}
{"train_loss": 0.5997636318206787, "global_step": 15261, "epoch": 171, "lr": 9.926938039737461e-05}
{"train_loss": 0.6833153963088989, "global_step": 15262, "epoch": 171, "lr": 9.92692816493086e-05}
{"train_loss": 0.6093550324440002, "global_step": 15263, "epoch": 171, "lr": 9.926918289461891e-05}
{"train_loss": 0.5000194907188416, "global_step": 15264, "epoch": 171, "lr": 9.92690841333056e-05}
{"train_loss": 0.5476474165916443, "global_step": 15265, "epoch": 171, "lr": 9.926898536536866e-05}
{"train_loss": 0.7171708941459656, "global_step": 15266, "epoch": 171, "lr": 9.926888659080809e-05}
{"train_loss": 0.5355204939842224, "global_step": 15267, "epoch": 171, "lr": 9.926878780962393e-05}
{"train_loss": 0.566541314125061, "global_step": 15268, "epoch": 171, "lr": 9.926868902181619e-05}
{"train_loss": 0.613664448261261, "global_step": 15269, "epoch": 171, "lr": 9.926859022738486e-05}
{"train_loss": 0.4918077886104584, "global_step": 15270, "epoch": 171, "lr": 9.926849142632997e-05}
{"train_loss": 0.5107991695404053, "global_step": 15271, "epoch": 171, "lr": 9.926839261865152e-05}
{"train_loss": 0.5549920797348022, "global_step": 15272, "epoch": 171, "lr": 9.926829380434957e-05}
{"train_loss": 0.5173166990280151, "global_step": 15273, "epoch": 171, "lr": 9.926819498342407e-05}
{"train_loss": 0.6436013579368591, "global_step": 15274, "epoch": 171, "lr": 9.926809615587508e-05}
{"train_loss": 0.5238804221153259, "global_step": 15275, "epoch": 171, "lr": 9.926799732170258e-05}
{"train_loss": 0.5628387928009033, "global_step": 15276, "epoch": 171, "lr": 9.92678984809066e-05}
{"train_loss": 0.5085955262184143, "global_step": 15277, "epoch": 171, "lr": 9.926779963348716e-05}
{"train_loss": 0.4554601311683655, "global_step": 15278, "epoch": 171, "lr": 9.926770077944426e-05}
{"train_loss": 0.568791389465332, "global_step": 15279, "epoch": 171, "lr": 9.926760191877792e-05}
{"train_loss": 0.5833801031112671, "global_step": 15280, "epoch": 171, "lr": 9.926750305148816e-05}
{"train_loss": 0.48995620012283325, "global_step": 15281, "epoch": 171, "lr": 9.926740417757496e-05}
{"train_loss": 0.4836403727531433, "global_step": 15282, "epoch": 171, "lr": 9.926730529703838e-05}
{"train_loss": 0.5698487162590027, "global_step": 15283, "epoch": 171, "lr": 9.926720640987842e-05}
{"train_loss": 0.6321160197257996, "global_step": 15284, "epoch": 171, "lr": 9.926710751609505e-05}
{"train_loss": 0.5286611318588257, "global_step": 15285, "epoch": 171, "lr": 9.926700861568835e-05}
{"train_loss": 0.6738226413726807, "global_step": 15286, "epoch": 171, "lr": 9.926690970865828e-05}
{"train_loss": 0.7017500996589661, "global_step": 15287, "epoch": 171, "lr": 9.926681079500489e-05}
{"train_loss": 0.6261020302772522, "global_step": 15288, "epoch": 171, "lr": 9.926671187472816e-05}
{"train_loss": 0.621472954750061, "global_step": 15289, "epoch": 171, "lr": 9.926661294782814e-05}
{"train_loss": 0.663724958896637, "global_step": 15290, "epoch": 171, "lr": 9.92665140143048e-05}
{"train_loss": 0.583766520023346, "global_step": 15291, "epoch": 171, "lr": 9.92664150741582e-05}
{"train_loss": 0.5365809202194214, "global_step": 15292, "epoch": 171, "lr": 9.926631612738833e-05}
{"train_loss": 0.49862897396087646, "global_step": 15293, "epoch": 171, "lr": 9.926621717399519e-05}
{"train_loss": 0.7275263071060181, "global_step": 15294, "epoch": 171, "lr": 9.926611821397882e-05}
{"train_loss": 0.4678927958011627, "global_step": 15295, "epoch": 171, "lr": 9.92660192473392e-05}
{"train_loss": 0.5379873514175415, "global_step": 15296, "epoch": 171, "lr": 9.926592027407639e-05}
{"train_loss": 0.6435526013374329, "global_step": 15297, "epoch": 171, "lr": 9.926582129419038e-05}
{"train_loss": 0.5663418173789978, "global_step": 15298, "epoch": 171, "lr": 9.926572230768116e-05}
{"train_loss": 0.6368018984794617, "global_step": 15299, "epoch": 171, "lr": 9.926562331454877e-05}
{"train_loss": 0.6345049142837524, "global_step": 15300, "epoch": 171, "lr": 9.926552431479322e-05}
{"train_loss": 0.7966001629829407, "global_step": 15301, "epoch": 171, "lr": 9.926542530841452e-05}
{"train_loss": 0.6495275497436523, "global_step": 15302, "epoch": 171, "lr": 9.926532629541269e-05}
{"train_loss": 0.49782267212867737, "global_step": 15303, "epoch": 171, "lr": 9.926522727578773e-05}
{"train_loss": 0.5280643105506897, "global_step": 15304, "epoch": 171, "lr": 9.926512824953966e-05}
{"train_loss": 0.5271835923194885, "global_step": 15305, "epoch": 171, "lr": 9.92650292166685e-05}
{"train_loss": 0.614594042301178, "global_step": 15306, "epoch": 171, "lr": 9.926493017717426e-05}
{"train_loss": 0.58890522028623, "global_step": 15307, "epoch": 171, "lr": 9.926483113105694e-05, "val_loss": 1.5159690380096436}
{"train_loss": 0.5234811902046204, "global_step": 15308, "epoch": 172, "lr": 9.926473207831656e-05}
{"train_loss": 0.5525522828102112, "global_step": 15309, "epoch": 172, "lr": 9.926463301895316e-05}
{"train_loss": 0.5644550323486328, "global_step": 15310, "epoch": 172, "lr": 9.926453395296671e-05}
{"train_loss": 0.5500514507293701, "global_step": 15311, "epoch": 172, "lr": 9.926443488035726e-05}
{"train_loss": 0.5417112112045288, "global_step": 15312, "epoch": 172, "lr": 9.92643358011248e-05}
{"train_loss": 0.6370027661323547, "global_step": 15313, "epoch": 172, "lr": 9.926423671526934e-05}
{"train_loss": 0.5255244374275208, "global_step": 15314, "epoch": 172, "lr": 9.926413762279092e-05}
{"train_loss": 0.5680364966392517, "global_step": 15315, "epoch": 172, "lr": 9.926403852368953e-05}
{"train_loss": 0.6263378858566284, "global_step": 15316, "epoch": 172, "lr": 9.926393941796519e-05}
{"train_loss": 0.40860220789909363, "global_step": 15317, "epoch": 172, "lr": 9.926384030561791e-05}
{"train_loss": 0.5604180097579956, "global_step": 15318, "epoch": 172, "lr": 9.92637411866477e-05}
{"train_loss": 0.46780505776405334, "global_step": 15319, "epoch": 172, "lr": 9.92636420610546e-05}
{"train_loss": 0.6303849220275879, "global_step": 15320, "epoch": 172, "lr": 9.92635429288386e-05}
{"train_loss": 0.4463147222995758, "global_step": 15321, "epoch": 172, "lr": 9.92634437899997e-05}
{"train_loss": 0.8143453598022461, "global_step": 15322, "epoch": 172, "lr": 9.926334464453795e-05}
{"train_loss": 0.6263057589530945, "global_step": 15323, "epoch": 172, "lr": 9.926324549245334e-05}
{"train_loss": 0.5864272713661194, "global_step": 15324, "epoch": 172, "lr": 9.926314633374588e-05}
{"train_loss": 0.6782922148704529, "global_step": 15325, "epoch": 172, "lr": 9.926304716841559e-05}
{"train_loss": 0.6193758249282837, "global_step": 15326, "epoch": 172, "lr": 9.926294799646248e-05}
{"train_loss": 0.5924996137619019, "global_step": 15327, "epoch": 172, "lr": 9.926284881788658e-05}
{"train_loss": 0.5414201617240906, "global_step": 15328, "epoch": 172, "lr": 9.926274963268789e-05}
{"train_loss": 0.5568049550056458, "global_step": 15329, "epoch": 172, "lr": 9.926265044086641e-05}
{"train_loss": 0.4407070577144623, "global_step": 15330, "epoch": 172, "lr": 9.926255124242219e-05}
{"train_loss": 0.6936023831367493, "global_step": 15331, "epoch": 172, "lr": 9.92624520373552e-05}
{"train_loss": 0.5598504543304443, "global_step": 15332, "epoch": 172, "lr": 9.926235282566547e-05}
{"train_loss": 0.6173515319824219, "global_step": 15333, "epoch": 172, "lr": 9.926225360735304e-05}
{"train_loss": 0.5395041108131409, "global_step": 15334, "epoch": 172, "lr": 9.926215438241789e-05}
{"train_loss": 0.5819454789161682, "global_step": 15335, "epoch": 172, "lr": 9.926205515086003e-05}
{"train_loss": 0.634611964225769, "global_step": 15336, "epoch": 172, "lr": 9.926195591267951e-05}
{"train_loss": 0.571985125541687, "global_step": 15337, "epoch": 172, "lr": 9.92618566678763e-05}
{"train_loss": 0.5902906656265259, "global_step": 15338, "epoch": 172, "lr": 9.926175741645044e-05}
{"train_loss": 0.553431510925293, "global_step": 15339, "epoch": 172, "lr": 9.926165815840193e-05}
{"train_loss": 0.6047270894050598, "global_step": 15340, "epoch": 172, "lr": 9.926155889373081e-05}
{"train_loss": 0.610042929649353, "global_step": 15341, "epoch": 172, "lr": 9.926145962243707e-05}
{"train_loss": 0.4756350815296173, "global_step": 15342, "epoch": 172, "lr": 9.926136034452071e-05}
{"train_loss": 0.5628853440284729, "global_step": 15343, "epoch": 172, "lr": 9.926126105998178e-05}
{"train_loss": 0.6033858060836792, "global_step": 15344, "epoch": 172, "lr": 9.926116176882026e-05}
{"train_loss": 0.6609665751457214, "global_step": 15345, "epoch": 172, "lr": 9.92610624710362e-05}
{"train_loss": 0.6171908378601074, "global_step": 15346, "epoch": 172, "lr": 9.926096316662956e-05}
{"train_loss": 0.6110200881958008, "global_step": 15347, "epoch": 172, "lr": 9.92608638556004e-05}
{"train_loss": 0.7769514322280884, "global_step": 15348, "epoch": 172, "lr": 9.92607645379487e-05}
{"train_loss": 0.6496003270149231, "global_step": 15349, "epoch": 172, "lr": 9.926066521367452e-05}
{"train_loss": 0.5451350212097168, "global_step": 15350, "epoch": 172, "lr": 9.926056588277782e-05}
{"train_loss": 0.5897576808929443, "global_step": 15351, "epoch": 172, "lr": 9.926046654525865e-05}
{"train_loss": 0.6681650876998901, "global_step": 15352, "epoch": 172, "lr": 9.9260367201117e-05}
{"train_loss": 0.6309755444526672, "global_step": 15353, "epoch": 172, "lr": 9.926026785035291e-05}
{"train_loss": 0.6737914085388184, "global_step": 15354, "epoch": 172, "lr": 9.926016849296636e-05}
{"train_loss": 0.6218109130859375, "global_step": 15355, "epoch": 172, "lr": 9.92600691289574e-05}
{"train_loss": 0.5482624769210815, "global_step": 15356, "epoch": 172, "lr": 9.925996975832602e-05}
{"train_loss": 0.49758902192115784, "global_step": 15357, "epoch": 172, "lr": 9.925987038107222e-05}
{"train_loss": 0.5848771333694458, "global_step": 15358, "epoch": 172, "lr": 9.925977099719604e-05}
{"train_loss": 0.6168347597122192, "global_step": 15359, "epoch": 172, "lr": 9.92596716066975e-05}
{"train_loss": 0.5253525376319885, "global_step": 15360, "epoch": 172, "lr": 9.925957220957656e-05}
{"train_loss": 0.641331136226654, "global_step": 15361, "epoch": 172, "lr": 9.92594728058333e-05}
{"train_loss": 0.6312214732170105, "global_step": 15362, "epoch": 172, "lr": 9.925937339546771e-05}
{"train_loss": 0.5017701983451843, "global_step": 15363, "epoch": 172, "lr": 9.925927397847979e-05}
{"train_loss": 0.7567822337150574, "global_step": 15364, "epoch": 172, "lr": 9.925917455486955e-05}
{"train_loss": 0.6066009402275085, "global_step": 15365, "epoch": 172, "lr": 9.925907512463701e-05}
{"train_loss": 0.6253533959388733, "global_step": 15366, "epoch": 172, "lr": 9.92589756877822e-05}
{"train_loss": 0.6179628372192383, "global_step": 15367, "epoch": 172, "lr": 9.925887624430512e-05}
{"train_loss": 0.6158545017242432, "global_step": 15368, "epoch": 172, "lr": 9.925877679420579e-05}
{"train_loss": 0.6513182520866394, "global_step": 15369, "epoch": 172, "lr": 9.925867733748421e-05}
{"train_loss": 0.6143207550048828, "global_step": 15370, "epoch": 172, "lr": 9.925857787414042e-05}
{"train_loss": 0.6468253135681152, "global_step": 15371, "epoch": 172, "lr": 9.925847840417438e-05}
{"train_loss": 0.6623906493186951, "global_step": 15372, "epoch": 172, "lr": 9.925837892758616e-05}
{"train_loss": 0.591639518737793, "global_step": 15373, "epoch": 172, "lr": 9.925827944437575e-05}
{"train_loss": 0.4523278474807739, "global_step": 15374, "epoch": 172, "lr": 9.925817995454316e-05}
{"train_loss": 0.571412980556488, "global_step": 15375, "epoch": 172, "lr": 9.925808045808841e-05}
{"train_loss": 0.6049979329109192, "global_step": 15376, "epoch": 172, "lr": 9.92579809550115e-05}
{"train_loss": 0.5186010599136353, "global_step": 15377, "epoch": 172, "lr": 9.925788144531247e-05}
{"train_loss": 0.5900593400001526, "global_step": 15378, "epoch": 172, "lr": 9.925778192899132e-05}
{"train_loss": 0.5501739978790283, "global_step": 15379, "epoch": 172, "lr": 9.925768240604805e-05}
{"train_loss": 0.4556572437286377, "global_step": 15380, "epoch": 172, "lr": 9.92575828764827e-05}
{"train_loss": 0.5481556057929993, "global_step": 15381, "epoch": 172, "lr": 9.925748334029525e-05}
{"train_loss": 0.7437615394592285, "global_step": 15382, "epoch": 172, "lr": 9.925738379748575e-05}
{"train_loss": 0.5287338495254517, "global_step": 15383, "epoch": 172, "lr": 9.925728424805419e-05}
{"train_loss": 0.7157998085021973, "global_step": 15384, "epoch": 172, "lr": 9.925718469200059e-05}
{"train_loss": 0.5660741329193115, "global_step": 15385, "epoch": 172, "lr": 9.925708512932495e-05}
{"train_loss": 0.44801098108291626, "global_step": 15386, "epoch": 172, "lr": 9.92569855600273e-05}
{"train_loss": 0.5922898054122925, "global_step": 15387, "epoch": 172, "lr": 9.925688598410765e-05}
{"train_loss": 0.4583495855331421, "global_step": 15388, "epoch": 172, "lr": 9.925678640156602e-05}
{"train_loss": 0.5629489421844482, "global_step": 15389, "epoch": 172, "lr": 9.925668681240241e-05}
{"train_loss": 0.6156770586967468, "global_step": 15390, "epoch": 172, "lr": 9.925658721661683e-05}
{"train_loss": 0.5692947506904602, "global_step": 15391, "epoch": 172, "lr": 9.925648761420932e-05}
{"train_loss": 0.6197158098220825, "global_step": 15392, "epoch": 172, "lr": 9.925638800517985e-05}
{"train_loss": 0.590913712978363, "global_step": 15393, "epoch": 172, "lr": 9.925628838952847e-05}
{"train_loss": 0.6060278415679932, "global_step": 15394, "epoch": 172, "lr": 9.92561887672552e-05}
{"train_loss": 0.575752854347229, "global_step": 15395, "epoch": 172, "lr": 9.925608913836002e-05}
{"train_loss": 0.5894686412275507, "global_step": 15396, "epoch": 172, "lr": 9.925598950284296e-05, "val_loss": 1.5138576030731201}
{"train_loss": 0.5635371208190918, "global_step": 15397, "epoch": 173, "lr": 9.925588986070403e-05}
{"train_loss": 0.5095366835594177, "global_step": 15398, "epoch": 173, "lr": 9.925579021194325e-05}
{"train_loss": 0.6050034165382385, "global_step": 15399, "epoch": 173, "lr": 9.925569055656063e-05}
{"train_loss": 0.5228252410888672, "global_step": 15400, "epoch": 173, "lr": 9.925559089455618e-05}
{"train_loss": 0.55962735414505, "global_step": 15401, "epoch": 173, "lr": 9.925549122592993e-05}
{"train_loss": 0.7749663591384888, "global_step": 15402, "epoch": 173, "lr": 9.925539155068186e-05}
{"train_loss": 0.5358474254608154, "global_step": 15403, "epoch": 173, "lr": 9.925529186881202e-05}
{"train_loss": 0.525871992111206, "global_step": 15404, "epoch": 173, "lr": 9.925519218032039e-05}
{"train_loss": 0.6761478185653687, "global_step": 15405, "epoch": 173, "lr": 9.9255092485207e-05}
{"train_loss": 0.5058765411376953, "global_step": 15406, "epoch": 173, "lr": 9.925499278347188e-05}
{"train_loss": 0.5125937461853027, "global_step": 15407, "epoch": 173, "lr": 9.925489307511502e-05}
{"train_loss": 0.5364280343055725, "global_step": 15408, "epoch": 173, "lr": 9.925479336013644e-05}
{"train_loss": 0.4712941348552704, "global_step": 15409, "epoch": 173, "lr": 9.925469363853614e-05}
{"train_loss": 0.5306248664855957, "global_step": 15410, "epoch": 173, "lr": 9.925459391031415e-05}
{"train_loss": 0.700061559677124, "global_step": 15411, "epoch": 173, "lr": 9.925449417547049e-05}
{"train_loss": 0.49347996711730957, "global_step": 15412, "epoch": 173, "lr": 9.925439443400516e-05}
{"train_loss": 0.4928676187992096, "global_step": 15413, "epoch": 173, "lr": 9.925429468591819e-05}
{"train_loss": 0.5933331847190857, "global_step": 15414, "epoch": 173, "lr": 9.925419493120956e-05}
{"train_loss": 0.653389573097229, "global_step": 15415, "epoch": 173, "lr": 9.925409516987931e-05}
{"train_loss": 0.5367303490638733, "global_step": 15416, "epoch": 173, "lr": 9.925399540192745e-05}
{"train_loss": 0.6984988451004028, "global_step": 15417, "epoch": 173, "lr": 9.925389562735398e-05}
{"train_loss": 0.5621064901351929, "global_step": 15418, "epoch": 173, "lr": 9.925379584615893e-05}
{"train_loss": 0.5829114317893982, "global_step": 15419, "epoch": 173, "lr": 9.925369605834232e-05}
{"train_loss": 0.5326926708221436, "global_step": 15420, "epoch": 173, "lr": 9.925359626390413e-05}
{"train_loss": 0.5450802445411682, "global_step": 15421, "epoch": 173, "lr": 9.92534964628444e-05}
{"train_loss": 0.5618918538093567, "global_step": 15422, "epoch": 173, "lr": 9.925339665516315e-05}
{"train_loss": 0.5217786431312561, "global_step": 15423, "epoch": 173, "lr": 9.925329684086037e-05}
{"train_loss": 0.5125442147254944, "global_step": 15424, "epoch": 173, "lr": 9.92531970199361e-05}
{"train_loss": 0.6186185479164124, "global_step": 15425, "epoch": 173, "lr": 9.925309719239033e-05}
{"train_loss": 0.45952388644218445, "global_step": 15426, "epoch": 173, "lr": 9.925299735822307e-05}
{"train_loss": 0.5260197520256042, "global_step": 15427, "epoch": 173, "lr": 9.925289751743436e-05}
{"train_loss": 0.5681381225585938, "global_step": 15428, "epoch": 173, "lr": 9.92527976700242e-05}
{"train_loss": 0.5932521224021912, "global_step": 15429, "epoch": 173, "lr": 9.925269781599259e-05}
{"train_loss": 0.512917697429657, "global_step": 15430, "epoch": 173, "lr": 9.925259795533956e-05}
{"train_loss": 0.573566734790802, "global_step": 15431, "epoch": 173, "lr": 9.925249808806511e-05}
{"train_loss": 0.5584629774093628, "global_step": 15432, "epoch": 173, "lr": 9.925239821416927e-05}
{"train_loss": 0.5813089609146118, "global_step": 15433, "epoch": 173, "lr": 9.925229833365204e-05}
{"train_loss": 0.73621666431427, "global_step": 15434, "epoch": 173, "lr": 9.925219844651346e-05}
{"train_loss": 0.5856223702430725, "global_step": 15435, "epoch": 173, "lr": 9.92520985527535e-05}
{"train_loss": 0.503535807132721, "global_step": 15436, "epoch": 173, "lr": 9.92519986523722e-05}
{"train_loss": 0.5886921286582947, "global_step": 15437, "epoch": 173, "lr": 9.925189874536956e-05}
{"train_loss": 0.5645654797554016, "global_step": 15438, "epoch": 173, "lr": 9.925179883174563e-05}
{"train_loss": 0.5724778771400452, "global_step": 15439, "epoch": 173, "lr": 9.925169891150037e-05}
{"train_loss": 0.6939693689346313, "global_step": 15440, "epoch": 173, "lr": 9.925159898463383e-05}
{"train_loss": 0.5698034763336182, "global_step": 15441, "epoch": 173, "lr": 9.925149905114602e-05}
{"train_loss": 0.6717641353607178, "global_step": 15442, "epoch": 173, "lr": 9.925139911103692e-05}
{"train_loss": 0.6232260465621948, "global_step": 15443, "epoch": 173, "lr": 9.925129916430658e-05}
{"train_loss": 0.545601487159729, "global_step": 15444, "epoch": 173, "lr": 9.925119921095503e-05}
{"train_loss": 0.5305532813072205, "global_step": 15445, "epoch": 173, "lr": 9.925109925098223e-05}
{"train_loss": 0.5439915060997009, "global_step": 15446, "epoch": 173, "lr": 9.925099928438823e-05}
{"train_loss": 0.5974184274673462, "global_step": 15447, "epoch": 173, "lr": 9.925089931117302e-05}
{"train_loss": 0.5011506676673889, "global_step": 15448, "epoch": 173, "lr": 9.925079933133664e-05}
{"train_loss": 0.6502542495727539, "global_step": 15449, "epoch": 173, "lr": 9.925069934487909e-05}
{"train_loss": 0.5691526532173157, "global_step": 15450, "epoch": 173, "lr": 9.925059935180038e-05}
{"train_loss": 0.5617209672927856, "global_step": 15451, "epoch": 173, "lr": 9.925049935210052e-05}
{"train_loss": 0.5141218900680542, "global_step": 15452, "epoch": 173, "lr": 9.925039934577955e-05}
{"train_loss": 0.5937833786010742, "global_step": 15453, "epoch": 173, "lr": 9.925029933283745e-05}
{"train_loss": 0.48624953627586365, "global_step": 15454, "epoch": 173, "lr": 9.925019931327425e-05}
{"train_loss": 0.5979968905448914, "global_step": 15455, "epoch": 173, "lr": 9.925009928708994e-05}
{"train_loss": 0.7256680130958557, "global_step": 15456, "epoch": 173, "lr": 9.92499992542846e-05}
{"train_loss": 0.5745529532432556, "global_step": 15457, "epoch": 173, "lr": 9.924989921485817e-05}
{"train_loss": 0.543906033039093, "global_step": 15458, "epoch": 173, "lr": 9.924979916881069e-05}
{"train_loss": 0.6643412709236145, "global_step": 15459, "epoch": 173, "lr": 9.924969911614219e-05}
{"train_loss": 0.5161609649658203, "global_step": 15460, "epoch": 173, "lr": 9.924959905685265e-05}
{"train_loss": 0.5169146656990051, "global_step": 15461, "epoch": 173, "lr": 9.924949899094211e-05}
{"train_loss": 0.5333999395370483, "global_step": 15462, "epoch": 173, "lr": 9.924939891841056e-05}
{"train_loss": 0.4742361009120941, "global_step": 15463, "epoch": 173, "lr": 9.924929883925805e-05}
{"train_loss": 0.5440314412117004, "global_step": 15464, "epoch": 173, "lr": 9.924919875348457e-05}
{"train_loss": 0.5833613276481628, "global_step": 15465, "epoch": 173, "lr": 9.924909866109013e-05}
{"train_loss": 0.5437172055244446, "global_step": 15466, "epoch": 173, "lr": 9.924899856207475e-05}
{"train_loss": 0.6218138933181763, "global_step": 15467, "epoch": 173, "lr": 9.924889845643843e-05}
{"train_loss": 0.68935626745224, "global_step": 15468, "epoch": 173, "lr": 9.924879834418122e-05}
{"train_loss": 0.6468127965927124, "global_step": 15469, "epoch": 173, "lr": 9.924869822530309e-05}
{"train_loss": 0.5435341000556946, "global_step": 15470, "epoch": 173, "lr": 9.92485980998041e-05}
{"train_loss": 0.6523236632347107, "global_step": 15471, "epoch": 173, "lr": 9.92484979676842e-05}
{"train_loss": 0.49883759021759033, "global_step": 15472, "epoch": 173, "lr": 9.924839782894346e-05}
{"train_loss": 0.4868021011352539, "global_step": 15473, "epoch": 173, "lr": 9.924829768358185e-05}
{"train_loss": 0.6905856728553772, "global_step": 15474, "epoch": 173, "lr": 9.924819753159944e-05}
{"train_loss": 0.6823591589927673, "global_step": 15475, "epoch": 173, "lr": 9.924809737299619e-05}
{"train_loss": 0.5745031237602234, "global_step": 15476, "epoch": 173, "lr": 9.924799720777215e-05}
{"train_loss": 0.44905486702919006, "global_step": 15477, "epoch": 173, "lr": 9.924789703592729e-05}
{"train_loss": 0.5687023401260376, "global_step": 15478, "epoch": 173, "lr": 9.924779685746168e-05}
{"train_loss": 0.6076761484146118, "global_step": 15479, "epoch": 173, "lr": 9.924769667237529e-05}
{"train_loss": 0.5497303605079651, "global_step": 15480, "epoch": 173, "lr": 9.924759648066814e-05}
{"train_loss": 0.4840206205844879, "global_step": 15481, "epoch": 173, "lr": 9.924749628234026e-05}
{"train_loss": 0.5764833688735962, "global_step": 15482, "epoch": 173, "lr": 9.924739607739166e-05}
{"train_loss": 0.5210863947868347, "global_step": 15483, "epoch": 173, "lr": 9.924729586582234e-05}
{"train_loss": 0.5810065269470215, "global_step": 15484, "epoch": 173, "lr": 9.924719564763231e-05}
{"train_loss": 0.5713977127262716, "global_step": 15485, "epoch": 173, "lr": 9.924709542282163e-05, "val_loss": 1.48517644405365}
{"train_loss": 0.6269242763519287, "global_step": 15486, "epoch": 174, "lr": 9.924699519139025e-05}
{"train_loss": 0.6856602430343628, "global_step": 15487, "epoch": 174, "lr": 9.924689495333821e-05}
{"train_loss": 0.4758640229701996, "global_step": 15488, "epoch": 174, "lr": 9.924679470866554e-05}
{"train_loss": 0.6015510559082031, "global_step": 15489, "epoch": 174, "lr": 9.924669445737223e-05}
{"train_loss": 0.5714476704597473, "global_step": 15490, "epoch": 174, "lr": 9.924659419945829e-05}
{"train_loss": 0.6652958989143372, "global_step": 15491, "epoch": 174, "lr": 9.924649393492378e-05}
{"train_loss": 0.425608366727829, "global_step": 15492, "epoch": 174, "lr": 9.924639366376865e-05}
{"train_loss": 0.5775097012519836, "global_step": 15493, "epoch": 174, "lr": 9.924629338599295e-05}
{"train_loss": 0.5535390973091125, "global_step": 15494, "epoch": 174, "lr": 9.924619310159669e-05}
{"train_loss": 0.5782929062843323, "global_step": 15495, "epoch": 174, "lr": 9.924609281057988e-05}
{"train_loss": 0.6870416402816772, "global_step": 15496, "epoch": 174, "lr": 9.924599251294254e-05}
{"train_loss": 0.42133477330207825, "global_step": 15497, "epoch": 174, "lr": 9.924589220868465e-05}
{"train_loss": 0.5569572448730469, "global_step": 15498, "epoch": 174, "lr": 9.924579189780627e-05}
{"train_loss": 0.5788953304290771, "global_step": 15499, "epoch": 174, "lr": 9.924569158030739e-05}
{"train_loss": 0.5271471738815308, "global_step": 15500, "epoch": 174, "lr": 9.924559125618801e-05}
{"train_loss": 0.5402272939682007, "global_step": 15501, "epoch": 174, "lr": 9.924549092544819e-05}
{"train_loss": 0.5399746894836426, "global_step": 15502, "epoch": 174, "lr": 9.92453905880879e-05}
{"train_loss": 0.4655667245388031, "global_step": 15503, "epoch": 174, "lr": 9.924529024410717e-05}
{"train_loss": 0.6364113688468933, "global_step": 15504, "epoch": 174, "lr": 9.9245189893506e-05}
{"train_loss": 0.615612268447876, "global_step": 15505, "epoch": 174, "lr": 9.924508953628444e-05}
{"train_loss": 0.39493969082832336, "global_step": 15506, "epoch": 174, "lr": 9.924498917244244e-05}
{"train_loss": 0.587264358997345, "global_step": 15507, "epoch": 174, "lr": 9.924488880198008e-05}
{"train_loss": 0.589888870716095, "global_step": 15508, "epoch": 174, "lr": 9.924478842489735e-05}
{"train_loss": 0.5781821608543396, "global_step": 15509, "epoch": 174, "lr": 9.924468804119425e-05}
{"train_loss": 0.6398454308509827, "global_step": 15510, "epoch": 174, "lr": 9.92445876508708e-05}
{"train_loss": 0.49872323870658875, "global_step": 15511, "epoch": 174, "lr": 9.924448725392701e-05}
{"train_loss": 0.6471875309944153, "global_step": 15512, "epoch": 174, "lr": 9.92443868503629e-05}
{"train_loss": 0.5989569425582886, "global_step": 15513, "epoch": 174, "lr": 9.924428644017849e-05}
{"train_loss": 0.5760747194290161, "global_step": 15514, "epoch": 174, "lr": 9.924418602337378e-05}
{"train_loss": 0.6031354665756226, "global_step": 15515, "epoch": 174, "lr": 9.924408559994877e-05}
{"train_loss": 0.5094043612480164, "global_step": 15516, "epoch": 174, "lr": 9.924398516990353e-05}
{"train_loss": 0.6474856734275818, "global_step": 15517, "epoch": 174, "lr": 9.924388473323802e-05}
{"train_loss": 0.5740023851394653, "global_step": 15518, "epoch": 174, "lr": 9.924378428995226e-05}
{"train_loss": 0.5416964292526245, "global_step": 15519, "epoch": 174, "lr": 9.92436838400463e-05}
{"train_loss": 0.5192230939865112, "global_step": 15520, "epoch": 174, "lr": 9.92435833835201e-05}
{"train_loss": 0.5361069440841675, "global_step": 15521, "epoch": 174, "lr": 9.92434829203737e-05}
{"train_loss": 0.5632489323616028, "global_step": 15522, "epoch": 174, "lr": 9.924338245060714e-05}
{"train_loss": 0.592577338218689, "global_step": 15523, "epoch": 174, "lr": 9.924328197422039e-05}
{"train_loss": 0.4742090702056885, "global_step": 15524, "epoch": 174, "lr": 9.924318149121348e-05}
{"train_loss": 0.5282686352729797, "global_step": 15525, "epoch": 174, "lr": 9.924308100158642e-05}
{"train_loss": 0.5559908747673035, "global_step": 15526, "epoch": 174, "lr": 9.924298050533924e-05}
{"train_loss": 0.4952113926410675, "global_step": 15527, "epoch": 174, "lr": 9.924288000247194e-05}
{"train_loss": 0.5747694969177246, "global_step": 15528, "epoch": 174, "lr": 9.924277949298453e-05}
{"train_loss": 0.556939959526062, "global_step": 15529, "epoch": 174, "lr": 9.924267897687702e-05}
{"train_loss": 0.5915870666503906, "global_step": 15530, "epoch": 174, "lr": 9.924257845414944e-05}
{"train_loss": 0.45825207233428955, "global_step": 15531, "epoch": 174, "lr": 9.924247792480181e-05}
{"train_loss": 0.5250964164733887, "global_step": 15532, "epoch": 174, "lr": 9.92423773888341e-05}
{"train_loss": 0.6181715130805969, "global_step": 15533, "epoch": 174, "lr": 9.924227684624639e-05}
{"train_loss": 0.5643712282180786, "global_step": 15534, "epoch": 174, "lr": 9.924217629703862e-05}
{"train_loss": 0.6440358757972717, "global_step": 15535, "epoch": 174, "lr": 9.924207574121087e-05}
{"train_loss": 0.48738861083984375, "global_step": 15536, "epoch": 174, "lr": 9.92419751787631e-05}
{"train_loss": 0.6107085347175598, "global_step": 15537, "epoch": 174, "lr": 9.924187460969536e-05}
{"train_loss": 0.5692792534828186, "global_step": 15538, "epoch": 174, "lr": 9.924177403400765e-05}
{"train_loss": 0.6932194828987122, "global_step": 15539, "epoch": 174, "lr": 9.924167345169997e-05}
{"train_loss": 0.6575562953948975, "global_step": 15540, "epoch": 174, "lr": 9.924157286277236e-05}
{"train_loss": 0.634738564491272, "global_step": 15541, "epoch": 174, "lr": 9.92414722672248e-05}
{"train_loss": 0.5211879014968872, "global_step": 15542, "epoch": 174, "lr": 9.924137166505735e-05}
{"train_loss": 0.44580063223838806, "global_step": 15543, "epoch": 174, "lr": 9.924127105626999e-05}
{"train_loss": 0.5231829285621643, "global_step": 15544, "epoch": 174, "lr": 9.924117044086274e-05}
{"train_loss": 0.5355345010757446, "global_step": 15545, "epoch": 174, "lr": 9.92410698188356e-05}
{"train_loss": 0.4909684956073761, "global_step": 15546, "epoch": 174, "lr": 9.924096919018861e-05}
{"train_loss": 0.5420578718185425, "global_step": 15547, "epoch": 174, "lr": 9.924086855492178e-05}
{"train_loss": 0.4631352722644806, "global_step": 15548, "epoch": 174, "lr": 9.924076791303511e-05}
{"train_loss": 0.38253045082092285, "global_step": 15549, "epoch": 174, "lr": 9.924066726452863e-05}
{"train_loss": 0.5759181380271912, "global_step": 15550, "epoch": 174, "lr": 9.924056660940233e-05}
{"train_loss": 0.6224387884140015, "global_step": 15551, "epoch": 174, "lr": 9.924046594765624e-05}
{"train_loss": 0.683067262172699, "global_step": 15552, "epoch": 174, "lr": 9.924036527929035e-05}
{"train_loss": 0.6473100185394287, "global_step": 15553, "epoch": 174, "lr": 9.924026460430471e-05}
{"train_loss": 0.475650429725647, "global_step": 15554, "epoch": 174, "lr": 9.924016392269933e-05}
{"train_loss": 0.41757333278656006, "global_step": 15555, "epoch": 174, "lr": 9.92400632344742e-05}
{"train_loss": 0.5358502864837646, "global_step": 15556, "epoch": 174, "lr": 9.923996253962933e-05}
{"train_loss": 0.5129234790802002, "global_step": 15557, "epoch": 174, "lr": 9.923986183816476e-05}
{"train_loss": 0.6365046501159668, "global_step": 15558, "epoch": 174, "lr": 9.923976113008049e-05}
{"train_loss": 0.6365386247634888, "global_step": 15559, "epoch": 174, "lr": 9.923966041537652e-05}
{"train_loss": 0.5288593173027039, "global_step": 15560, "epoch": 174, "lr": 9.92395596940529e-05}
{"train_loss": 0.5788229703903198, "global_step": 15561, "epoch": 174, "lr": 9.92394589661096e-05}
{"train_loss": 0.6033716797828674, "global_step": 15562, "epoch": 174, "lr": 9.923935823154667e-05}
{"train_loss": 0.6950227618217468, "global_step": 15563, "epoch": 174, "lr": 9.923925749036412e-05}
{"train_loss": 0.6339870691299438, "global_step": 15564, "epoch": 174, "lr": 9.923915674256193e-05}
{"train_loss": 0.49606454372406006, "global_step": 15565, "epoch": 174, "lr": 9.923905598814014e-05}
{"train_loss": 0.6218550205230713, "global_step": 15566, "epoch": 174, "lr": 9.923895522709876e-05}
{"train_loss": 0.6905985474586487, "global_step": 15567, "epoch": 174, "lr": 9.923885445943779e-05}
{"train_loss": 0.5609175562858582, "global_step": 15568, "epoch": 174, "lr": 9.923875368515727e-05}
{"train_loss": 0.586319625377655, "global_step": 15569, "epoch": 174, "lr": 9.92386529042572e-05}
{"train_loss": 0.6563370823860168, "global_step": 15570, "epoch": 174, "lr": 9.923855211673761e-05}
{"train_loss": 0.5044952630996704, "global_step": 15571, "epoch": 174, "lr": 9.923845132259846e-05}
{"train_loss": 0.586781919002533, "global_step": 15572, "epoch": 174, "lr": 9.923835052183983e-05}
{"train_loss": 0.7002384066581726, "global_step": 15573, "epoch": 174, "lr": 9.92382497144617e-05}
{"train_loss": 0.5663675853375638, "global_step": 15574, "epoch": 174, "lr": 9.923814890046408e-05, "val_loss": 1.469100832939148}
{"train_loss": 0.4844966232776642, "global_step": 15575, "epoch": 175, "lr": 9.923804807984698e-05}
{"train_loss": 0.6354271173477173, "global_step": 15576, "epoch": 175, "lr": 9.923794725261044e-05}
{"train_loss": 0.6245622038841248, "global_step": 15577, "epoch": 175, "lr": 9.923784641875446e-05}
{"train_loss": 0.49365612864494324, "global_step": 15578, "epoch": 175, "lr": 9.923774557827905e-05}
{"train_loss": 0.604728639125824, "global_step": 15579, "epoch": 175, "lr": 9.923764473118422e-05}
{"train_loss": 0.6316333413124084, "global_step": 15580, "epoch": 175, "lr": 9.923754387746998e-05}
{"train_loss": 0.6049121618270874, "global_step": 15581, "epoch": 175, "lr": 9.923744301713637e-05}
{"train_loss": 0.4942290186882019, "global_step": 15582, "epoch": 175, "lr": 9.923734215018337e-05}
{"train_loss": 0.6478083729743958, "global_step": 15583, "epoch": 175, "lr": 9.923724127661102e-05}
{"train_loss": 0.5148116946220398, "global_step": 15584, "epoch": 175, "lr": 9.923714039641934e-05}
{"train_loss": 0.4880509376525879, "global_step": 15585, "epoch": 175, "lr": 9.92370395096083e-05}
{"train_loss": 0.47099068760871887, "global_step": 15586, "epoch": 175, "lr": 9.923693861617794e-05}
{"train_loss": 0.6591569185256958, "global_step": 15587, "epoch": 175, "lr": 9.923683771612829e-05}
{"train_loss": 0.5702967643737793, "global_step": 15588, "epoch": 175, "lr": 9.923673680945933e-05}
{"train_loss": 0.4452891945838928, "global_step": 15589, "epoch": 175, "lr": 9.92366358961711e-05}
{"train_loss": 0.5907231569290161, "global_step": 15590, "epoch": 175, "lr": 9.923653497626362e-05}
{"train_loss": 0.5760616064071655, "global_step": 15591, "epoch": 175, "lr": 9.923643404973686e-05}
{"train_loss": 0.4919651746749878, "global_step": 15592, "epoch": 175, "lr": 9.923633311659087e-05}
{"train_loss": 0.6116592884063721, "global_step": 15593, "epoch": 175, "lr": 9.923623217682567e-05}
{"train_loss": 0.49066483974456787, "global_step": 15594, "epoch": 175, "lr": 9.923613123044124e-05}
{"train_loss": 0.3862185478210449, "global_step": 15595, "epoch": 175, "lr": 9.923603027743762e-05}
{"train_loss": 0.4253002405166626, "global_step": 15596, "epoch": 175, "lr": 9.923592931781482e-05}
{"train_loss": 0.4856443405151367, "global_step": 15597, "epoch": 175, "lr": 9.923582835157285e-05}
{"train_loss": 0.40686383843421936, "global_step": 15598, "epoch": 175, "lr": 9.923572737871171e-05}
{"train_loss": 0.4993654191493988, "global_step": 15599, "epoch": 175, "lr": 9.923562639923143e-05}
{"train_loss": 0.434552401304245, "global_step": 15600, "epoch": 175, "lr": 9.923552541313204e-05}
{"train_loss": 0.5200563669204712, "global_step": 15601, "epoch": 175, "lr": 9.923542442041351e-05}
{"train_loss": 0.5553712248802185, "global_step": 15602, "epoch": 175, "lr": 9.923532342107589e-05}
{"train_loss": 0.7592927813529968, "global_step": 15603, "epoch": 175, "lr": 9.923522241511918e-05}
{"train_loss": 0.5485101342201233, "global_step": 15604, "epoch": 175, "lr": 9.923512140254338e-05}
{"train_loss": 0.5870601534843445, "global_step": 15605, "epoch": 175, "lr": 9.923502038334853e-05}
{"train_loss": 0.49225887656211853, "global_step": 15606, "epoch": 175, "lr": 9.923491935753463e-05}
{"train_loss": 0.504398763179779, "global_step": 15607, "epoch": 175, "lr": 9.92348183251017e-05}
{"train_loss": 0.7143095135688782, "global_step": 15608, "epoch": 175, "lr": 9.923471728604974e-05}
{"train_loss": 0.7505017518997192, "global_step": 15609, "epoch": 175, "lr": 9.923461624037877e-05}
{"train_loss": 0.4738660156726837, "global_step": 15610, "epoch": 175, "lr": 9.923451518808881e-05}
{"train_loss": 0.6947346925735474, "global_step": 15611, "epoch": 175, "lr": 9.923441412917989e-05}
{"train_loss": 0.5259158611297607, "global_step": 15612, "epoch": 175, "lr": 9.923431306365198e-05}
{"train_loss": 0.6027812957763672, "global_step": 15613, "epoch": 175, "lr": 9.923421199150513e-05}
{"train_loss": 0.6806087493896484, "global_step": 15614, "epoch": 175, "lr": 9.923411091273933e-05}
{"train_loss": 0.5300629138946533, "global_step": 15615, "epoch": 175, "lr": 9.923400982735461e-05}
{"train_loss": 0.48279303312301636, "global_step": 15616, "epoch": 175, "lr": 9.923390873535097e-05}
{"train_loss": 0.5627851486206055, "global_step": 15617, "epoch": 175, "lr": 9.923380763672843e-05}
{"train_loss": 0.5587612390518188, "global_step": 15618, "epoch": 175, "lr": 9.923370653148702e-05}
{"train_loss": 0.5868991613388062, "global_step": 15619, "epoch": 175, "lr": 9.923360541962672e-05}
{"train_loss": 0.5984212160110474, "global_step": 15620, "epoch": 175, "lr": 9.923350430114757e-05}
{"train_loss": 0.42089322209358215, "global_step": 15621, "epoch": 175, "lr": 9.923340317604957e-05}
{"train_loss": 0.6424228549003601, "global_step": 15622, "epoch": 175, "lr": 9.923330204433275e-05}
{"train_loss": 0.5705470442771912, "global_step": 15623, "epoch": 175, "lr": 9.923320090599711e-05}
{"train_loss": 0.564386248588562, "global_step": 15624, "epoch": 175, "lr": 9.923309976104266e-05}
{"train_loss": 0.6174517869949341, "global_step": 15625, "epoch": 175, "lr": 9.923299860946942e-05}
{"train_loss": 0.5158693790435791, "global_step": 15626, "epoch": 175, "lr": 9.923289745127741e-05}
{"train_loss": 0.5808700323104858, "global_step": 15627, "epoch": 175, "lr": 9.923279628646662e-05}
{"train_loss": 0.4975857138633728, "global_step": 15628, "epoch": 175, "lr": 9.92326951150371e-05}
{"train_loss": 0.5998090505599976, "global_step": 15629, "epoch": 175, "lr": 9.923259393698885e-05}
{"train_loss": 0.6176977753639221, "global_step": 15630, "epoch": 175, "lr": 9.923249275232186e-05}
{"train_loss": 0.5851231217384338, "global_step": 15631, "epoch": 175, "lr": 9.923239156103616e-05}
{"train_loss": 0.5617258548736572, "global_step": 15632, "epoch": 175, "lr": 9.923229036313177e-05}
{"train_loss": 0.4544251561164856, "global_step": 15633, "epoch": 175, "lr": 9.92321891586087e-05}
{"train_loss": 0.6738313436508179, "global_step": 15634, "epoch": 175, "lr": 9.923208794746696e-05}
{"train_loss": 0.5643976926803589, "global_step": 15635, "epoch": 175, "lr": 9.923198672970657e-05}
{"train_loss": 0.4565800428390503, "global_step": 15636, "epoch": 175, "lr": 9.923188550532755e-05}
{"train_loss": 0.6564496159553528, "global_step": 15637, "epoch": 175, "lr": 9.923178427432989e-05}
{"train_loss": 0.5076309442520142, "global_step": 15638, "epoch": 175, "lr": 9.923168303671361e-05}
{"train_loss": 0.4678039848804474, "global_step": 15639, "epoch": 175, "lr": 9.923158179247875e-05}
{"train_loss": 0.6321457624435425, "global_step": 15640, "epoch": 175, "lr": 9.923148054162529e-05}
{"train_loss": 0.40711086988449097, "global_step": 15641, "epoch": 175, "lr": 9.923137928415325e-05}
{"train_loss": 0.6490917801856995, "global_step": 15642, "epoch": 175, "lr": 9.923127802006267e-05}
{"train_loss": 0.4818103015422821, "global_step": 15643, "epoch": 175, "lr": 9.923117674935354e-05}
{"train_loss": 0.654463529586792, "global_step": 15644, "epoch": 175, "lr": 9.923107547202586e-05}
{"train_loss": 0.6094300746917725, "global_step": 15645, "epoch": 175, "lr": 9.923097418807969e-05}
{"train_loss": 0.5364588499069214, "global_step": 15646, "epoch": 175, "lr": 9.9230872897515e-05}
{"train_loss": 0.6743397116661072, "global_step": 15647, "epoch": 175, "lr": 9.92307716003318e-05}
{"train_loss": 0.6023955345153809, "global_step": 15648, "epoch": 175, "lr": 9.923067029653015e-05}
{"train_loss": 0.505452573299408, "global_step": 15649, "epoch": 175, "lr": 9.923056898611003e-05}
{"train_loss": 0.43821069598197937, "global_step": 15650, "epoch": 175, "lr": 9.923046766907148e-05}
{"train_loss": 0.5355625152587891, "global_step": 15651, "epoch": 175, "lr": 9.923036634541446e-05}
{"train_loss": 0.5808057188987732, "global_step": 15652, "epoch": 175, "lr": 9.923026501513904e-05}
{"train_loss": 0.6706161499023438, "global_step": 15653, "epoch": 175, "lr": 9.923016367824519e-05}
{"train_loss": 0.5328606963157654, "global_step": 15654, "epoch": 175, "lr": 9.923006233473295e-05}
{"train_loss": 0.6138836145401001, "global_step": 15655, "epoch": 175, "lr": 9.922996098460234e-05}
{"train_loss": 0.5225616097450256, "global_step": 15656, "epoch": 175, "lr": 9.922985962785336e-05}
{"train_loss": 0.5432495474815369, "global_step": 15657, "epoch": 175, "lr": 9.922975826448601e-05}
{"train_loss": 0.5210647583007812, "global_step": 15658, "epoch": 175, "lr": 9.922965689450035e-05}
{"train_loss": 0.44268232583999634, "global_step": 15659, "epoch": 175, "lr": 9.922955551789633e-05}
{"train_loss": 0.4925343692302704, "global_step": 15660, "epoch": 175, "lr": 9.922945413467402e-05}
{"train_loss": 0.6006270051002502, "global_step": 15661, "epoch": 175, "lr": 9.922935274483339e-05}
{"train_loss": 0.4724433422088623, "global_step": 15662, "epoch": 175, "lr": 9.922925134837448e-05}
{"train_loss": 0.554125088989065, "global_step": 15663, "epoch": 175, "lr": 9.92291499452973e-05, "val_loss": 1.5194332599639893, "train_action_mse_error": 46.19257736206055}
{"train_loss": 0.5790390372276306, "global_step": 15664, "epoch": 176, "lr": 9.922904853560187e-05}
{"train_loss": 0.4195522367954254, "global_step": 15665, "epoch": 176, "lr": 9.922894711928818e-05}
{"train_loss": 0.5977419018745422, "global_step": 15666, "epoch": 176, "lr": 9.922884569635627e-05}
{"train_loss": 0.46933674812316895, "global_step": 15667, "epoch": 176, "lr": 9.922874426680612e-05}
{"train_loss": 0.6023281812667847, "global_step": 15668, "epoch": 176, "lr": 9.922864283063778e-05}
{"train_loss": 0.4750076234340668, "global_step": 15669, "epoch": 176, "lr": 9.922854138785126e-05}
{"train_loss": 0.6654673218727112, "global_step": 15670, "epoch": 176, "lr": 9.922843993844655e-05}
{"train_loss": 0.6484581232070923, "global_step": 15671, "epoch": 176, "lr": 9.922833848242368e-05}
{"train_loss": 0.6552569270133972, "global_step": 15672, "epoch": 176, "lr": 9.922823701978266e-05}
{"train_loss": 0.4187779128551483, "global_step": 15673, "epoch": 176, "lr": 9.92281355505235e-05}
{"train_loss": 0.5523714423179626, "global_step": 15674, "epoch": 176, "lr": 9.922803407464621e-05}
{"train_loss": 0.5925662517547607, "global_step": 15675, "epoch": 176, "lr": 9.922793259215082e-05}
{"train_loss": 0.6594178676605225, "global_step": 15676, "epoch": 176, "lr": 9.922783110303733e-05}
{"train_loss": 0.6057918667793274, "global_step": 15677, "epoch": 176, "lr": 9.922772960730577e-05}
{"train_loss": 0.523590624332428, "global_step": 15678, "epoch": 176, "lr": 9.922762810495614e-05}
{"train_loss": 0.5723376274108887, "global_step": 15679, "epoch": 176, "lr": 9.922752659598844e-05}
{"train_loss": 0.5194879770278931, "global_step": 15680, "epoch": 176, "lr": 9.922742508040271e-05}
{"train_loss": 0.4798338711261749, "global_step": 15681, "epoch": 176, "lr": 9.922732355819894e-05}
{"train_loss": 0.5477540493011475, "global_step": 15682, "epoch": 176, "lr": 9.922722202937717e-05}
{"train_loss": 0.5048268437385559, "global_step": 15683, "epoch": 176, "lr": 9.922712049393741e-05}
{"train_loss": 0.59291672706604, "global_step": 15684, "epoch": 176, "lr": 9.922701895187965e-05}
{"train_loss": 0.5651442408561707, "global_step": 15685, "epoch": 176, "lr": 9.922691740320392e-05}
{"train_loss": 0.5145103931427002, "global_step": 15686, "epoch": 176, "lr": 9.922681584791023e-05}
{"train_loss": 0.5525085926055908, "global_step": 15687, "epoch": 176, "lr": 9.92267142859986e-05}
{"train_loss": 0.47586318850517273, "global_step": 15688, "epoch": 176, "lr": 9.922661271746902e-05}
{"train_loss": 0.5275125503540039, "global_step": 15689, "epoch": 176, "lr": 9.922651114232153e-05}
{"train_loss": 0.4196230173110962, "global_step": 15690, "epoch": 176, "lr": 9.922640956055615e-05}
{"train_loss": 0.5653209090232849, "global_step": 15691, "epoch": 176, "lr": 9.922630797217287e-05}
{"train_loss": 0.39999672770500183, "global_step": 15692, "epoch": 176, "lr": 9.92262063771717e-05}
{"train_loss": 0.6468686461448669, "global_step": 15693, "epoch": 176, "lr": 9.922610477555268e-05}
{"train_loss": 0.5419623851776123, "global_step": 15694, "epoch": 176, "lr": 9.922600316731582e-05}
{"train_loss": 0.54490727186203, "global_step": 15695, "epoch": 176, "lr": 9.92259015524611e-05}
{"train_loss": 0.49041467905044556, "global_step": 15696, "epoch": 176, "lr": 9.922579993098857e-05}
{"train_loss": 0.5080437660217285, "global_step": 15697, "epoch": 176, "lr": 9.922569830289824e-05}
{"train_loss": 0.615403413772583, "global_step": 15698, "epoch": 176, "lr": 9.922559666819012e-05}
{"train_loss": 0.45629438757896423, "global_step": 15699, "epoch": 176, "lr": 9.922549502686419e-05}
{"train_loss": 0.5689248442649841, "global_step": 15700, "epoch": 176, "lr": 9.92253933789205e-05}
{"train_loss": 0.5649952292442322, "global_step": 15701, "epoch": 176, "lr": 9.922529172435907e-05}
{"train_loss": 0.6056410074234009, "global_step": 15702, "epoch": 176, "lr": 9.92251900631799e-05}
{"train_loss": 0.5742934346199036, "global_step": 15703, "epoch": 176, "lr": 9.922508839538299e-05}
{"train_loss": 0.5258437991142273, "global_step": 15704, "epoch": 176, "lr": 9.922498672096836e-05}
{"train_loss": 0.627667248249054, "global_step": 15705, "epoch": 176, "lr": 9.922488503993606e-05}
{"train_loss": 0.5993496775627136, "global_step": 15706, "epoch": 176, "lr": 9.922478335228604e-05}
{"train_loss": 0.6926959156990051, "global_step": 15707, "epoch": 176, "lr": 9.922468165801837e-05}
{"train_loss": 0.546090841293335, "global_step": 15708, "epoch": 176, "lr": 9.922457995713303e-05}
{"train_loss": 0.5264450907707214, "global_step": 15709, "epoch": 176, "lr": 9.922447824963005e-05}
{"train_loss": 0.5628834366798401, "global_step": 15710, "epoch": 176, "lr": 9.922437653550945e-05}
{"train_loss": 0.5613402724266052, "global_step": 15711, "epoch": 176, "lr": 9.922427481477121e-05}
{"train_loss": 0.5236112475395203, "global_step": 15712, "epoch": 176, "lr": 9.922417308741538e-05}
{"train_loss": 0.47414854168891907, "global_step": 15713, "epoch": 176, "lr": 9.922407135344196e-05}
{"train_loss": 0.6099494099617004, "global_step": 15714, "epoch": 176, "lr": 9.922396961285097e-05}
{"train_loss": 0.5736123323440552, "global_step": 15715, "epoch": 176, "lr": 9.922386786564241e-05}
{"train_loss": 0.5195820331573486, "global_step": 15716, "epoch": 176, "lr": 9.92237661118163e-05}
{"train_loss": 0.6693443059921265, "global_step": 15717, "epoch": 176, "lr": 9.922366435137264e-05}
{"train_loss": 0.4478640854358673, "global_step": 15718, "epoch": 176, "lr": 9.922356258431149e-05}
{"train_loss": 0.578536331653595, "global_step": 15719, "epoch": 176, "lr": 9.922346081063281e-05}
{"train_loss": 0.5706779956817627, "global_step": 15720, "epoch": 176, "lr": 9.922335903033664e-05}
{"train_loss": 0.5214585661888123, "global_step": 15721, "epoch": 176, "lr": 9.9223257243423e-05}
{"train_loss": 0.5155318975448608, "global_step": 15722, "epoch": 176, "lr": 9.922315544989187e-05}
{"train_loss": 0.6409028172492981, "global_step": 15723, "epoch": 176, "lr": 9.922305364974331e-05}
{"train_loss": 0.3776873052120209, "global_step": 15724, "epoch": 176, "lr": 9.92229518429773e-05}
{"train_loss": 0.45814162492752075, "global_step": 15725, "epoch": 176, "lr": 9.922285002959387e-05}
{"train_loss": 0.5369517803192139, "global_step": 15726, "epoch": 176, "lr": 9.9222748209593e-05}
{"train_loss": 0.5028687715530396, "global_step": 15727, "epoch": 176, "lr": 9.922264638297477e-05}
{"train_loss": 0.5809040069580078, "global_step": 15728, "epoch": 176, "lr": 9.922254454973914e-05}
{"train_loss": 0.6147572994232178, "global_step": 15729, "epoch": 176, "lr": 9.922244270988613e-05}
{"train_loss": 0.5826645493507385, "global_step": 15730, "epoch": 176, "lr": 9.922234086341577e-05}
{"train_loss": 0.529590368270874, "global_step": 15731, "epoch": 176, "lr": 9.922223901032808e-05}
{"train_loss": 0.5632523894309998, "global_step": 15732, "epoch": 176, "lr": 9.922213715062303e-05}
{"train_loss": 0.5597289204597473, "global_step": 15733, "epoch": 176, "lr": 9.922203528430069e-05}
{"train_loss": 0.45694929361343384, "global_step": 15734, "epoch": 176, "lr": 9.922193341136104e-05}
{"train_loss": 0.5354551672935486, "global_step": 15735, "epoch": 176, "lr": 9.922183153180408e-05}
{"train_loss": 0.5872616171836853, "global_step": 15736, "epoch": 176, "lr": 9.922172964562988e-05}
{"train_loss": 0.6683269143104553, "global_step": 15737, "epoch": 176, "lr": 9.922162775283839e-05}
{"train_loss": 0.6127772927284241, "global_step": 15738, "epoch": 176, "lr": 9.922152585342968e-05}
{"train_loss": 0.6465738415718079, "global_step": 15739, "epoch": 176, "lr": 9.92214239474037e-05}
{"train_loss": 0.6714243292808533, "global_step": 15740, "epoch": 176, "lr": 9.922132203476052e-05}
{"train_loss": 0.6748329997062683, "global_step": 15741, "epoch": 176, "lr": 9.922122011550013e-05}
{"train_loss": 0.5641102194786072, "global_step": 15742, "epoch": 176, "lr": 9.922111818962255e-05}
{"train_loss": 0.6141329407691956, "global_step": 15743, "epoch": 176, "lr": 9.922101625712778e-05}
{"train_loss": 0.5679723620414734, "global_step": 15744, "epoch": 176, "lr": 9.922091431801587e-05}
{"train_loss": 0.66511470079422, "global_step": 15745, "epoch": 176, "lr": 9.922081237228679e-05}
{"train_loss": 0.5984377264976501, "global_step": 15746, "epoch": 176, "lr": 9.922071041994057e-05}
{"train_loss": 0.5867093801498413, "global_step": 15747, "epoch": 176, "lr": 9.922060846097723e-05}
{"train_loss": 0.6170060634613037, "global_step": 15748, "epoch": 176, "lr": 9.922050649539679e-05}
{"train_loss": 0.5473307371139526, "global_step": 15749, "epoch": 176, "lr": 9.922040452319923e-05}
{"train_loss": 0.5896525979042053, "global_step": 15750, "epoch": 176, "lr": 9.92203025443846e-05}
{"train_loss": 0.6206096410751343, "global_step": 15751, "epoch": 176, "lr": 9.92202005589529e-05}
{"train_loss": 0.5598628326748194, "global_step": 15752, "epoch": 176, "lr": 9.922009856690413e-05, "val_loss": 1.5170446634292603}
{"train_loss": 0.623688817024231, "global_step": 15753, "epoch": 177, "lr": 9.921999656823833e-05}
{"train_loss": 0.423788458108902, "global_step": 15754, "epoch": 177, "lr": 9.92198945629555e-05}
{"train_loss": 0.5542161464691162, "global_step": 15755, "epoch": 177, "lr": 9.921979255105565e-05}
{"train_loss": 0.6036803126335144, "global_step": 15756, "epoch": 177, "lr": 9.921969053253881e-05}
{"train_loss": 0.5155165791511536, "global_step": 15757, "epoch": 177, "lr": 9.921958850740499e-05}
{"train_loss": 0.5787382125854492, "global_step": 15758, "epoch": 177, "lr": 9.921948647565417e-05}
{"train_loss": 0.6146416068077087, "global_step": 15759, "epoch": 177, "lr": 9.92193844372864e-05}
{"train_loss": 0.4504558742046356, "global_step": 15760, "epoch": 177, "lr": 9.92192823923017e-05}
{"train_loss": 0.4671405553817749, "global_step": 15761, "epoch": 177, "lr": 9.921918034070004e-05}
{"train_loss": 0.43752771615982056, "global_step": 15762, "epoch": 177, "lr": 9.921907828248148e-05}
{"train_loss": 0.498843252658844, "global_step": 15763, "epoch": 177, "lr": 9.9218976217646e-05}
{"train_loss": 0.4776669144630432, "global_step": 15764, "epoch": 177, "lr": 9.921887414619365e-05}
{"train_loss": 0.47760844230651855, "global_step": 15765, "epoch": 177, "lr": 9.92187720681244e-05}
{"train_loss": 0.5789225101470947, "global_step": 15766, "epoch": 177, "lr": 9.92186699834383e-05}
{"train_loss": 0.6132887601852417, "global_step": 15767, "epoch": 177, "lr": 9.921856789213535e-05}
{"train_loss": 0.5332257151603699, "global_step": 15768, "epoch": 177, "lr": 9.921846579421556e-05}
{"train_loss": 0.5626077651977539, "global_step": 15769, "epoch": 177, "lr": 9.921836368967894e-05}
{"train_loss": 0.4798119068145752, "global_step": 15770, "epoch": 177, "lr": 9.921826157852552e-05}
{"train_loss": 0.5431539416313171, "global_step": 15771, "epoch": 177, "lr": 9.92181594607553e-05}
{"train_loss": 0.6005797386169434, "global_step": 15772, "epoch": 177, "lr": 9.92180573363683e-05}
{"train_loss": 0.5869227647781372, "global_step": 15773, "epoch": 177, "lr": 9.921795520536455e-05}
{"train_loss": 0.5952232480049133, "global_step": 15774, "epoch": 177, "lr": 9.921785306774402e-05}
{"train_loss": 0.5550984740257263, "global_step": 15775, "epoch": 177, "lr": 9.921775092350676e-05}
{"train_loss": 0.523871660232544, "global_step": 15776, "epoch": 177, "lr": 9.921764877265276e-05}
{"train_loss": 0.5909692049026489, "global_step": 15777, "epoch": 177, "lr": 9.921754661518205e-05}
{"train_loss": 0.4065893888473511, "global_step": 15778, "epoch": 177, "lr": 9.921744445109467e-05}
{"train_loss": 0.5125262141227722, "global_step": 15779, "epoch": 177, "lr": 9.921734228039056e-05}
{"train_loss": 0.4991305470466614, "global_step": 15780, "epoch": 177, "lr": 9.921724010306981e-05}
{"train_loss": 0.6124521493911743, "global_step": 15781, "epoch": 177, "lr": 9.92171379191324e-05}
{"train_loss": 0.4250441789627075, "global_step": 15782, "epoch": 177, "lr": 9.921703572857835e-05}
{"train_loss": 0.6671077609062195, "global_step": 15783, "epoch": 177, "lr": 9.921693353140765e-05}
{"train_loss": 0.5327414870262146, "global_step": 15784, "epoch": 177, "lr": 9.921683132762034e-05}
{"train_loss": 0.4206118583679199, "global_step": 15785, "epoch": 177, "lr": 9.921672911721642e-05}
{"train_loss": 0.4720190465450287, "global_step": 15786, "epoch": 177, "lr": 9.921662690019593e-05}
{"train_loss": 0.6524394154548645, "global_step": 15787, "epoch": 177, "lr": 9.921652467655884e-05}
{"train_loss": 0.5954048037528992, "global_step": 15788, "epoch": 177, "lr": 9.921642244630521e-05}
{"train_loss": 0.6174585819244385, "global_step": 15789, "epoch": 177, "lr": 9.921632020943504e-05}
{"train_loss": 0.5374556183815002, "global_step": 15790, "epoch": 177, "lr": 9.921621796594832e-05}
{"train_loss": 0.6382318139076233, "global_step": 15791, "epoch": 177, "lr": 9.921611571584506e-05}
{"train_loss": 0.4694454073905945, "global_step": 15792, "epoch": 177, "lr": 9.921601345912533e-05}
{"train_loss": 0.47792014479637146, "global_step": 15793, "epoch": 177, "lr": 9.921591119578907e-05}
{"train_loss": 0.5150439739227295, "global_step": 15794, "epoch": 177, "lr": 9.921580892583635e-05}
{"train_loss": 0.5605857968330383, "global_step": 15795, "epoch": 177, "lr": 9.921570664926716e-05}
{"train_loss": 0.5304999947547913, "global_step": 15796, "epoch": 177, "lr": 9.921560436608153e-05}
{"train_loss": 0.5967251658439636, "global_step": 15797, "epoch": 177, "lr": 9.921550207627945e-05}
{"train_loss": 0.5401828289031982, "global_step": 15798, "epoch": 177, "lr": 9.921539977986095e-05}
{"train_loss": 0.3902193009853363, "global_step": 15799, "epoch": 177, "lr": 9.921529747682604e-05}
{"train_loss": 0.5008454918861389, "global_step": 15800, "epoch": 177, "lr": 9.921519516717473e-05}
{"train_loss": 0.508714497089386, "global_step": 15801, "epoch": 177, "lr": 9.921509285090705e-05}
{"train_loss": 0.49784255027770996, "global_step": 15802, "epoch": 177, "lr": 9.921499052802298e-05}
{"train_loss": 0.5566805601119995, "global_step": 15803, "epoch": 177, "lr": 9.921488819852256e-05}
{"train_loss": 0.4996318519115448, "global_step": 15804, "epoch": 177, "lr": 9.92147858624058e-05}
{"train_loss": 0.6596060395240784, "global_step": 15805, "epoch": 177, "lr": 9.921468351967271e-05}
{"train_loss": 0.5936669707298279, "global_step": 15806, "epoch": 177, "lr": 9.921458117032332e-05}
{"train_loss": 0.485159695148468, "global_step": 15807, "epoch": 177, "lr": 9.921447881435762e-05}
{"train_loss": 0.4836161732673645, "global_step": 15808, "epoch": 177, "lr": 9.921437645177562e-05}
{"train_loss": 0.5973857045173645, "global_step": 15809, "epoch": 177, "lr": 9.921427408257737e-05}
{"train_loss": 0.6177825331687927, "global_step": 15810, "epoch": 177, "lr": 9.921417170676284e-05}
{"train_loss": 0.6032609939575195, "global_step": 15811, "epoch": 177, "lr": 9.921406932433208e-05}
{"train_loss": 0.4549504220485687, "global_step": 15812, "epoch": 177, "lr": 9.921396693528508e-05}
{"train_loss": 0.4701816141605377, "global_step": 15813, "epoch": 177, "lr": 9.921386453962187e-05}
{"train_loss": 0.4648107588291168, "global_step": 15814, "epoch": 177, "lr": 9.921376213734244e-05}
{"train_loss": 0.4337189197540283, "global_step": 15815, "epoch": 177, "lr": 9.921365972844683e-05}
{"train_loss": 0.6016985774040222, "global_step": 15816, "epoch": 177, "lr": 9.921355731293505e-05}
{"train_loss": 0.489539235830307, "global_step": 15817, "epoch": 177, "lr": 9.921345489080708e-05}
{"train_loss": 0.6101465821266174, "global_step": 15818, "epoch": 177, "lr": 9.9213352462063e-05}
{"train_loss": 0.4847307503223419, "global_step": 15819, "epoch": 177, "lr": 9.921325002670277e-05}
{"train_loss": 0.4899875819683075, "global_step": 15820, "epoch": 177, "lr": 9.92131475847264e-05}
{"train_loss": 0.5626234412193298, "global_step": 15821, "epoch": 177, "lr": 9.921304513613394e-05}
{"train_loss": 0.573466420173645, "global_step": 15822, "epoch": 177, "lr": 9.921294268092538e-05}
{"train_loss": 0.5258606672286987, "global_step": 15823, "epoch": 177, "lr": 9.921284021910073e-05}
{"train_loss": 0.6581690907478333, "global_step": 15824, "epoch": 177, "lr": 9.921273775066004e-05}
{"train_loss": 0.5525586605072021, "global_step": 15825, "epoch": 177, "lr": 9.921263527560327e-05}
{"train_loss": 0.6547021269798279, "global_step": 15826, "epoch": 177, "lr": 9.921253279393047e-05}
{"train_loss": 0.5133452415466309, "global_step": 15827, "epoch": 177, "lr": 9.921243030564166e-05}
{"train_loss": 0.5275865197181702, "global_step": 15828, "epoch": 177, "lr": 9.921232781073682e-05}
{"train_loss": 0.6894667744636536, "global_step": 15829, "epoch": 177, "lr": 9.921222530921598e-05}
{"train_loss": 0.5662410855293274, "global_step": 15830, "epoch": 177, "lr": 9.921212280107917e-05}
{"train_loss": 0.6657629609107971, "global_step": 15831, "epoch": 177, "lr": 9.92120202863264e-05}
{"train_loss": 0.493575781583786, "global_step": 15832, "epoch": 177, "lr": 9.921191776495764e-05}
{"train_loss": 0.5303789377212524, "global_step": 15833, "epoch": 177, "lr": 9.921181523697296e-05}
{"train_loss": 0.4780561625957489, "global_step": 15834, "epoch": 177, "lr": 9.921171270237235e-05}
{"train_loss": 0.44685226678848267, "global_step": 15835, "epoch": 177, "lr": 9.921161016115582e-05}
{"train_loss": 0.4618220925331116, "global_step": 15836, "epoch": 177, "lr": 9.921150761332339e-05}
{"train_loss": 0.5223181247711182, "global_step": 15837, "epoch": 177, "lr": 9.921140505887507e-05}
{"train_loss": 0.605237603187561, "global_step": 15838, "epoch": 177, "lr": 9.921130249781088e-05}
{"train_loss": 0.6885679364204407, "global_step": 15839, "epoch": 177, "lr": 9.921119993013084e-05}
{"train_loss": 0.6090235114097595, "global_step": 15840, "epoch": 177, "lr": 9.921109735583494e-05}
{"train_loss": 0.5410399624470914, "global_step": 15841, "epoch": 177, "lr": 9.921099477492322e-05, "val_loss": 1.5635355710983276}
{"train_loss": 0.4664524495601654, "global_step": 15842, "epoch": 178, "lr": 9.921089218739567e-05}
{"train_loss": 0.6066916584968567, "global_step": 15843, "epoch": 178, "lr": 9.921078959325232e-05}
{"train_loss": 0.634511411190033, "global_step": 15844, "epoch": 178, "lr": 9.921068699249316e-05}
{"train_loss": 0.6163731813430786, "global_step": 15845, "epoch": 178, "lr": 9.921058438511825e-05}
{"train_loss": 0.4486359655857086, "global_step": 15846, "epoch": 178, "lr": 9.921048177112757e-05}
{"train_loss": 0.49452584981918335, "global_step": 15847, "epoch": 178, "lr": 9.921037915052114e-05}
{"train_loss": 0.574248731136322, "global_step": 15848, "epoch": 178, "lr": 9.921027652329898e-05}
{"train_loss": 0.5686168670654297, "global_step": 15849, "epoch": 178, "lr": 9.921017388946108e-05}
{"train_loss": 0.591412365436554, "global_step": 15850, "epoch": 178, "lr": 9.921007124900747e-05}
{"train_loss": 0.566825270652771, "global_step": 15851, "epoch": 178, "lr": 9.920996860193819e-05}
{"train_loss": 0.5881571769714355, "global_step": 15852, "epoch": 178, "lr": 9.920986594825321e-05}
{"train_loss": 0.43915948271751404, "global_step": 15853, "epoch": 178, "lr": 9.920976328795257e-05}
{"train_loss": 0.562533974647522, "global_step": 15854, "epoch": 178, "lr": 9.920966062103629e-05}
{"train_loss": 0.4678809940814972, "global_step": 15855, "epoch": 178, "lr": 9.920955794750435e-05}
{"train_loss": 0.5290079116821289, "global_step": 15856, "epoch": 178, "lr": 9.920945526735679e-05}
{"train_loss": 0.7344956994056702, "global_step": 15857, "epoch": 178, "lr": 9.920935258059363e-05}
{"train_loss": 0.5587818622589111, "global_step": 15858, "epoch": 178, "lr": 9.920924988721486e-05}
{"train_loss": 0.42465874552726746, "global_step": 15859, "epoch": 178, "lr": 9.92091471872205e-05}
{"train_loss": 0.49766290187835693, "global_step": 15860, "epoch": 178, "lr": 9.920904448061059e-05}
{"train_loss": 0.5209106206893921, "global_step": 15861, "epoch": 178, "lr": 9.920894176738511e-05}
{"train_loss": 0.5732533931732178, "global_step": 15862, "epoch": 178, "lr": 9.92088390475441e-05}
{"train_loss": 0.606910228729248, "global_step": 15863, "epoch": 178, "lr": 9.920873632108755e-05}
{"train_loss": 0.6631431579589844, "global_step": 15864, "epoch": 178, "lr": 9.920863358801548e-05}
{"train_loss": 0.5900863409042358, "global_step": 15865, "epoch": 178, "lr": 9.920853084832793e-05}
{"train_loss": 0.6016205549240112, "global_step": 15866, "epoch": 178, "lr": 9.920842810202487e-05}
{"train_loss": 0.5943273305892944, "global_step": 15867, "epoch": 178, "lr": 9.920832534910636e-05}
{"train_loss": 0.6498556733131409, "global_step": 15868, "epoch": 178, "lr": 9.920822258957236e-05}
{"train_loss": 0.5782580375671387, "global_step": 15869, "epoch": 178, "lr": 9.920811982342294e-05}
{"train_loss": 0.550462543964386, "global_step": 15870, "epoch": 178, "lr": 9.920801705065808e-05}
{"train_loss": 0.54658043384552, "global_step": 15871, "epoch": 178, "lr": 9.92079142712778e-05}
{"train_loss": 0.5068010687828064, "global_step": 15872, "epoch": 178, "lr": 9.920781148528212e-05}
{"train_loss": 0.6457497477531433, "global_step": 15873, "epoch": 178, "lr": 9.920770869267104e-05}
{"train_loss": 0.6935418248176575, "global_step": 15874, "epoch": 178, "lr": 9.920760589344459e-05}
{"train_loss": 0.7782782316207886, "global_step": 15875, "epoch": 178, "lr": 9.920750308760279e-05}
{"train_loss": 0.5310196876525879, "global_step": 15876, "epoch": 178, "lr": 9.920740027514562e-05}
{"train_loss": 0.6947662830352783, "global_step": 15877, "epoch": 178, "lr": 9.92072974560731e-05}
{"train_loss": 0.556908130645752, "global_step": 15878, "epoch": 178, "lr": 9.920719463038531e-05}
{"train_loss": 0.5554551482200623, "global_step": 15879, "epoch": 178, "lr": 9.920709179808216e-05}
{"train_loss": 0.47080880403518677, "global_step": 15880, "epoch": 178, "lr": 9.920698895916375e-05}
{"train_loss": 0.6728761196136475, "global_step": 15881, "epoch": 178, "lr": 9.920688611363005e-05}
{"train_loss": 0.6299989819526672, "global_step": 15882, "epoch": 178, "lr": 9.920678326148109e-05}
{"train_loss": 0.5390722751617432, "global_step": 15883, "epoch": 178, "lr": 9.920668040271687e-05}
{"train_loss": 0.48891696333885193, "global_step": 15884, "epoch": 178, "lr": 9.920657753733741e-05}
{"train_loss": 0.5027657747268677, "global_step": 15885, "epoch": 178, "lr": 9.920647466534273e-05}
{"train_loss": 0.5040868520736694, "global_step": 15886, "epoch": 178, "lr": 9.920637178673283e-05}
{"train_loss": 0.49930456280708313, "global_step": 15887, "epoch": 178, "lr": 9.920626890150775e-05}
{"train_loss": 0.582828938961029, "global_step": 15888, "epoch": 178, "lr": 9.920616600966748e-05}
{"train_loss": 0.521719753742218, "global_step": 15889, "epoch": 178, "lr": 9.920606311121203e-05}
{"train_loss": 0.5506452322006226, "global_step": 15890, "epoch": 178, "lr": 9.920596020614143e-05}
{"train_loss": 0.5440068244934082, "global_step": 15891, "epoch": 178, "lr": 9.920585729445569e-05}
{"train_loss": 0.5809101462364197, "global_step": 15892, "epoch": 178, "lr": 9.920575437615483e-05}
{"train_loss": 0.7044652700424194, "global_step": 15893, "epoch": 178, "lr": 9.920565145123886e-05}
{"train_loss": 0.6876468658447266, "global_step": 15894, "epoch": 178, "lr": 9.920554851970777e-05}
{"train_loss": 0.5445404648780823, "global_step": 15895, "epoch": 178, "lr": 9.92054455815616e-05}
{"train_loss": 0.4958922564983368, "global_step": 15896, "epoch": 178, "lr": 9.920534263680037e-05}
{"train_loss": 0.5304182171821594, "global_step": 15897, "epoch": 178, "lr": 9.920523968542405e-05}
{"train_loss": 0.6232428550720215, "global_step": 15898, "epoch": 178, "lr": 9.920513672743273e-05}
{"train_loss": 0.4126265048980713, "global_step": 15899, "epoch": 178, "lr": 9.920503376282635e-05}
{"train_loss": 0.6860281229019165, "global_step": 15900, "epoch": 178, "lr": 9.920493079160496e-05}
{"train_loss": 0.6210417151451111, "global_step": 15901, "epoch": 178, "lr": 9.920482781376856e-05}
{"train_loss": 0.557013213634491, "global_step": 15902, "epoch": 178, "lr": 9.920472482931718e-05}
{"train_loss": 0.5364747643470764, "global_step": 15903, "epoch": 178, "lr": 9.920462183825082e-05}
{"train_loss": 0.7042021751403809, "global_step": 15904, "epoch": 178, "lr": 9.920451884056949e-05}
{"train_loss": 0.5289642810821533, "global_step": 15905, "epoch": 178, "lr": 9.920441583627323e-05}
{"train_loss": 0.7082321047782898, "global_step": 15906, "epoch": 178, "lr": 9.920431282536203e-05}
{"train_loss": 0.4975048899650574, "global_step": 15907, "epoch": 178, "lr": 9.920420980783591e-05}
{"train_loss": 0.5167078971862793, "global_step": 15908, "epoch": 178, "lr": 9.920410678369488e-05}
{"train_loss": 0.6241847276687622, "global_step": 15909, "epoch": 178, "lr": 9.920400375293895e-05}
{"train_loss": 0.489618718624115, "global_step": 15910, "epoch": 178, "lr": 9.920390071556815e-05}
{"train_loss": 0.6308681964874268, "global_step": 15911, "epoch": 178, "lr": 9.92037976715825e-05}
{"train_loss": 0.597268283367157, "global_step": 15912, "epoch": 178, "lr": 9.920369462098198e-05}
{"train_loss": 0.5231187343597412, "global_step": 15913, "epoch": 178, "lr": 9.920359156376663e-05}
{"train_loss": 0.525788426399231, "global_step": 15914, "epoch": 178, "lr": 9.920348849993646e-05}
{"train_loss": 0.6220371127128601, "global_step": 15915, "epoch": 178, "lr": 9.920338542949147e-05}
{"train_loss": 0.47054538130760193, "global_step": 15916, "epoch": 178, "lr": 9.920328235243168e-05}
{"train_loss": 0.5531330108642578, "global_step": 15917, "epoch": 178, "lr": 9.920317926875713e-05}
{"train_loss": 0.6743239760398865, "global_step": 15918, "epoch": 178, "lr": 9.92030761784678e-05}
{"train_loss": 0.5910989046096802, "global_step": 15919, "epoch": 178, "lr": 9.920297308156371e-05}
{"train_loss": 0.7325815558433533, "global_step": 15920, "epoch": 178, "lr": 9.920286997804489e-05}
{"train_loss": 0.5290694832801819, "global_step": 15921, "epoch": 178, "lr": 9.920276686791134e-05}
{"train_loss": 0.5050516128540039, "global_step": 15922, "epoch": 178, "lr": 9.920266375116307e-05}
{"train_loss": 0.5406637191772461, "global_step": 15923, "epoch": 178, "lr": 9.920256062780011e-05}
{"train_loss": 0.5220862030982971, "global_step": 15924, "epoch": 178, "lr": 9.920245749782247e-05}
{"train_loss": 0.6531010866165161, "global_step": 15925, "epoch": 178, "lr": 9.920235436123015e-05}
{"train_loss": 0.5261353254318237, "global_step": 15926, "epoch": 178, "lr": 9.920225121802318e-05}
{"train_loss": 0.6194034218788147, "global_step": 15927, "epoch": 178, "lr": 9.920214806820156e-05}
{"train_loss": 0.4898245632648468, "global_step": 15928, "epoch": 178, "lr": 9.920204491176532e-05}
{"train_loss": 0.48672449588775635, "global_step": 15929, "epoch": 178, "lr": 9.920194174871447e-05}
{"train_loss": 0.5691184404860722, "global_step": 15930, "epoch": 178, "lr": 9.9201838579049e-05, "val_loss": 1.505696177482605}
{"train_loss": 0.5895214080810547, "global_step": 15931, "epoch": 179, "lr": 9.920173540276896e-05}
{"train_loss": 0.6246035099029541, "global_step": 15932, "epoch": 179, "lr": 9.920163221987433e-05}
{"train_loss": 0.6100810170173645, "global_step": 15933, "epoch": 179, "lr": 9.920152903036514e-05}
{"train_loss": 0.44308537244796753, "global_step": 15934, "epoch": 179, "lr": 9.920142583424142e-05}
{"train_loss": 0.5732381939888, "global_step": 15935, "epoch": 179, "lr": 9.920132263150315e-05}
{"train_loss": 0.5428574085235596, "global_step": 15936, "epoch": 179, "lr": 9.920121942215037e-05}
{"train_loss": 0.6306276321411133, "global_step": 15937, "epoch": 179, "lr": 9.920111620618309e-05}
{"train_loss": 0.46398717164993286, "global_step": 15938, "epoch": 179, "lr": 9.920101298360131e-05}
{"train_loss": 0.5519464015960693, "global_step": 15939, "epoch": 179, "lr": 9.920090975440508e-05}
{"train_loss": 0.4324739873409271, "global_step": 15940, "epoch": 179, "lr": 9.920080651859435e-05}
{"train_loss": 0.643687903881073, "global_step": 15941, "epoch": 179, "lr": 9.920070327616919e-05}
{"train_loss": 0.5483108758926392, "global_step": 15942, "epoch": 179, "lr": 9.920060002712961e-05}
{"train_loss": 0.47984158992767334, "global_step": 15943, "epoch": 179, "lr": 9.920049677147558e-05}
{"train_loss": 0.42353928089141846, "global_step": 15944, "epoch": 179, "lr": 9.920039350920715e-05}
{"train_loss": 0.5108556151390076, "global_step": 15945, "epoch": 179, "lr": 9.920029024032434e-05}
{"train_loss": 0.46721750497817993, "global_step": 15946, "epoch": 179, "lr": 9.920018696482716e-05}
{"train_loss": 0.6109470725059509, "global_step": 15947, "epoch": 179, "lr": 9.920008368271558e-05}
{"train_loss": 0.6894475221633911, "global_step": 15948, "epoch": 179, "lr": 9.919998039398968e-05}
{"train_loss": 0.547630250453949, "global_step": 15949, "epoch": 179, "lr": 9.919987709864942e-05}
{"train_loss": 0.5160742998123169, "global_step": 15950, "epoch": 179, "lr": 9.919977379669485e-05}
{"train_loss": 0.6128791570663452, "global_step": 15951, "epoch": 179, "lr": 9.919967048812597e-05}
{"train_loss": 0.572124719619751, "global_step": 15952, "epoch": 179, "lr": 9.919956717294279e-05}
{"train_loss": 0.7288643717765808, "global_step": 15953, "epoch": 179, "lr": 9.919946385114532e-05}
{"train_loss": 0.5639593601226807, "global_step": 15954, "epoch": 179, "lr": 9.919936052273358e-05}
{"train_loss": 0.6580163240432739, "global_step": 15955, "epoch": 179, "lr": 9.91992571877076e-05}
{"train_loss": 0.6705100536346436, "global_step": 15956, "epoch": 179, "lr": 9.919915384606737e-05}
{"train_loss": 0.5692268013954163, "global_step": 15957, "epoch": 179, "lr": 9.919905049781293e-05}
{"train_loss": 0.6297466158866882, "global_step": 15958, "epoch": 179, "lr": 9.919894714294426e-05}
{"train_loss": 0.6238571405410767, "global_step": 15959, "epoch": 179, "lr": 9.919884378146139e-05}
{"train_loss": 0.5579864382743835, "global_step": 15960, "epoch": 179, "lr": 9.919874041336435e-05}
{"train_loss": 0.5993395447731018, "global_step": 15961, "epoch": 179, "lr": 9.919863703865312e-05}
{"train_loss": 0.5983396172523499, "global_step": 15962, "epoch": 179, "lr": 9.919853365732774e-05}
{"train_loss": 0.5634676218032837, "global_step": 15963, "epoch": 179, "lr": 9.919843026938821e-05}
{"train_loss": 0.5350192189216614, "global_step": 15964, "epoch": 179, "lr": 9.919832687483457e-05}
{"train_loss": 0.6592976450920105, "global_step": 15965, "epoch": 179, "lr": 9.919822347366681e-05}
{"train_loss": 0.620357871055603, "global_step": 15966, "epoch": 179, "lr": 9.919812006588494e-05}
{"train_loss": 0.554126501083374, "global_step": 15967, "epoch": 179, "lr": 9.919801665148898e-05}
{"train_loss": 0.6294028759002686, "global_step": 15968, "epoch": 179, "lr": 9.919791323047896e-05}
{"train_loss": 0.7090029716491699, "global_step": 15969, "epoch": 179, "lr": 9.919780980285487e-05}
{"train_loss": 0.5484104156494141, "global_step": 15970, "epoch": 179, "lr": 9.919770636861673e-05}
{"train_loss": 0.7202194929122925, "global_step": 15971, "epoch": 179, "lr": 9.919760292776456e-05}
{"train_loss": 0.591849684715271, "global_step": 15972, "epoch": 179, "lr": 9.91974994802984e-05}
{"train_loss": 0.4681968688964844, "global_step": 15973, "epoch": 179, "lr": 9.91973960262182e-05}
{"train_loss": 0.6198867559432983, "global_step": 15974, "epoch": 179, "lr": 9.919729256552402e-05}
{"train_loss": 0.5244626402854919, "global_step": 15975, "epoch": 179, "lr": 9.919718909821587e-05}
{"train_loss": 0.4901995360851288, "global_step": 15976, "epoch": 179, "lr": 9.919708562429375e-05}
{"train_loss": 0.6221531629562378, "global_step": 15977, "epoch": 179, "lr": 9.919698214375768e-05}
{"train_loss": 0.6313053965568542, "global_step": 15978, "epoch": 179, "lr": 9.919687865660768e-05}
{"train_loss": 0.5279940962791443, "global_step": 15979, "epoch": 179, "lr": 9.919677516284376e-05}
{"train_loss": 0.5185753107070923, "global_step": 15980, "epoch": 179, "lr": 9.919667166246594e-05}
{"train_loss": 0.6034421324729919, "global_step": 15981, "epoch": 179, "lr": 9.919656815547421e-05}
{"train_loss": 0.52134108543396, "global_step": 15982, "epoch": 179, "lr": 9.919646464186862e-05}
{"train_loss": 0.5426951050758362, "global_step": 15983, "epoch": 179, "lr": 9.919636112164915e-05}
{"train_loss": 0.6232742071151733, "global_step": 15984, "epoch": 179, "lr": 9.919625759481583e-05}
{"train_loss": 0.5408296585083008, "global_step": 15985, "epoch": 179, "lr": 9.919615406136869e-05}
{"train_loss": 0.5728289484977722, "global_step": 15986, "epoch": 179, "lr": 9.919605052130771e-05}
{"train_loss": 0.6251972317695618, "global_step": 15987, "epoch": 179, "lr": 9.919594697463292e-05}
{"train_loss": 0.6322888731956482, "global_step": 15988, "epoch": 179, "lr": 9.919584342134434e-05}
{"train_loss": 0.5661309957504272, "global_step": 15989, "epoch": 179, "lr": 9.919573986144197e-05}
{"train_loss": 0.6483638882637024, "global_step": 15990, "epoch": 179, "lr": 9.919563629492585e-05}
{"train_loss": 0.46254709362983704, "global_step": 15991, "epoch": 179, "lr": 9.919553272179596e-05}
{"train_loss": 0.6403952836990356, "global_step": 15992, "epoch": 179, "lr": 9.919542914205233e-05}
{"train_loss": 0.4944913983345032, "global_step": 15993, "epoch": 179, "lr": 9.919532555569498e-05}
{"train_loss": 0.6852238774299622, "global_step": 15994, "epoch": 179, "lr": 9.919522196272393e-05}
{"train_loss": 0.6525987386703491, "global_step": 15995, "epoch": 179, "lr": 9.919511836313917e-05}
{"train_loss": 0.6555993556976318, "global_step": 15996, "epoch": 179, "lr": 9.919501475694073e-05}
{"train_loss": 0.5487356185913086, "global_step": 15997, "epoch": 179, "lr": 9.919491114412861e-05}
{"train_loss": 0.5523557662963867, "global_step": 15998, "epoch": 179, "lr": 9.919480752470284e-05}
{"train_loss": 0.4666091799736023, "global_step": 15999, "epoch": 179, "lr": 9.919470389866344e-05}
{"train_loss": 0.586902916431427, "global_step": 16000, "epoch": 179, "lr": 9.919460026601039e-05}
{"train_loss": 0.5416094064712524, "global_step": 16001, "epoch": 179, "lr": 9.919449662674374e-05}
{"train_loss": 0.4752400815486908, "global_step": 16002, "epoch": 179, "lr": 9.919439298086348e-05}
{"train_loss": 0.6266205906867981, "global_step": 16003, "epoch": 179, "lr": 9.919428932836963e-05}
{"train_loss": 0.4938068985939026, "global_step": 16004, "epoch": 179, "lr": 9.919418566926222e-05}
{"train_loss": 0.5801555514335632, "global_step": 16005, "epoch": 179, "lr": 9.919408200354125e-05}
{"train_loss": 0.562885582447052, "global_step": 16006, "epoch": 179, "lr": 9.919397833120672e-05}
{"train_loss": 0.697158694267273, "global_step": 16007, "epoch": 179, "lr": 9.919387465225868e-05}
{"train_loss": 0.5256945490837097, "global_step": 16008, "epoch": 179, "lr": 9.919377096669711e-05}
{"train_loss": 0.7145045399665833, "global_step": 16009, "epoch": 179, "lr": 9.919366727452204e-05}
{"train_loss": 0.4509197473526001, "global_step": 16010, "epoch": 179, "lr": 9.919356357573348e-05}
{"train_loss": 0.514075756072998, "global_step": 16011, "epoch": 179, "lr": 9.919345987033147e-05}
{"train_loss": 0.6645223498344421, "global_step": 16012, "epoch": 179, "lr": 9.919335615831596e-05}
{"train_loss": 0.4865073561668396, "global_step": 16013, "epoch": 179, "lr": 9.919325243968702e-05}
{"train_loss": 0.6015615463256836, "global_step": 16014, "epoch": 179, "lr": 9.919314871444465e-05}
{"train_loss": 0.48416295647621155, "global_step": 16015, "epoch": 179, "lr": 9.919304498258886e-05}
{"train_loss": 0.6205171942710876, "global_step": 16016, "epoch": 179, "lr": 9.919294124411965e-05}
{"train_loss": 0.6349789500236511, "global_step": 16017, "epoch": 179, "lr": 9.919283749903708e-05}
{"train_loss": 0.455835223197937, "global_step": 16018, "epoch": 179, "lr": 9.91927337473411e-05}
{"train_loss": 0.5753886381561837, "global_step": 16019, "epoch": 179, "lr": 9.919262998903176e-05, "val_loss": 1.5575066804885864}
{"train_loss": 0.6040391325950623, "global_step": 16020, "epoch": 180, "lr": 9.919252622410908e-05}
{"train_loss": 0.6051222085952759, "global_step": 16021, "epoch": 180, "lr": 9.919242245257308e-05}
{"train_loss": 0.4396769106388092, "global_step": 16022, "epoch": 180, "lr": 9.919231867442373e-05}
{"train_loss": 0.4884192943572998, "global_step": 16023, "epoch": 180, "lr": 9.919221488966109e-05}
{"train_loss": 0.5042476058006287, "global_step": 16024, "epoch": 180, "lr": 9.919211109828515e-05}
{"train_loss": 0.5695528388023376, "global_step": 16025, "epoch": 180, "lr": 9.919200730029593e-05}
{"train_loss": 0.513115644454956, "global_step": 16026, "epoch": 180, "lr": 9.919190349569346e-05}
{"train_loss": 0.5673151612281799, "global_step": 16027, "epoch": 180, "lr": 9.919179968447771e-05}
{"train_loss": 0.5249766707420349, "global_step": 16028, "epoch": 180, "lr": 9.919169586664873e-05}
{"train_loss": 0.5412840843200684, "global_step": 16029, "epoch": 180, "lr": 9.919159204220654e-05}
{"train_loss": 0.674365758895874, "global_step": 16030, "epoch": 180, "lr": 9.919148821115113e-05}
{"train_loss": 0.4377230107784271, "global_step": 16031, "epoch": 180, "lr": 9.919138437348253e-05}
{"train_loss": 0.6095312833786011, "global_step": 16032, "epoch": 180, "lr": 9.919128052920073e-05}
{"train_loss": 0.4262930154800415, "global_step": 16033, "epoch": 180, "lr": 9.919117667830578e-05}
{"train_loss": 0.5620034337043762, "global_step": 16034, "epoch": 180, "lr": 9.919107282079767e-05}
{"train_loss": 0.49841850996017456, "global_step": 16035, "epoch": 180, "lr": 9.919096895667643e-05}
{"train_loss": 0.695581316947937, "global_step": 16036, "epoch": 180, "lr": 9.919086508594205e-05}
{"train_loss": 0.5921757221221924, "global_step": 16037, "epoch": 180, "lr": 9.919076120859456e-05}
{"train_loss": 0.4929443299770355, "global_step": 16038, "epoch": 180, "lr": 9.919065732463398e-05}
{"train_loss": 0.5252399444580078, "global_step": 16039, "epoch": 180, "lr": 9.91905534340603e-05}
{"train_loss": 0.4755744934082031, "global_step": 16040, "epoch": 180, "lr": 9.919044953687358e-05}
{"train_loss": 0.519916296005249, "global_step": 16041, "epoch": 180, "lr": 9.919034563307376e-05}
{"train_loss": 0.6193960309028625, "global_step": 16042, "epoch": 180, "lr": 9.919024172266093e-05}
{"train_loss": 0.5651295185089111, "global_step": 16043, "epoch": 180, "lr": 9.919013780563506e-05}
{"train_loss": 0.5154057145118713, "global_step": 16044, "epoch": 180, "lr": 9.919003388199618e-05}
{"train_loss": 0.41191282868385315, "global_step": 16045, "epoch": 180, "lr": 9.91899299517443e-05}
{"train_loss": 0.5018607378005981, "global_step": 16046, "epoch": 180, "lr": 9.918982601487942e-05}
{"train_loss": 0.5522505640983582, "global_step": 16047, "epoch": 180, "lr": 9.918972207140159e-05}
{"train_loss": 0.6337476968765259, "global_step": 16048, "epoch": 180, "lr": 9.918961812131078e-05}
{"train_loss": 0.5390918254852295, "global_step": 16049, "epoch": 180, "lr": 9.918951416460705e-05}
{"train_loss": 0.6351012587547302, "global_step": 16050, "epoch": 180, "lr": 9.918941020129036e-05}
{"train_loss": 0.5588244795799255, "global_step": 16051, "epoch": 180, "lr": 9.918930623136077e-05}
{"train_loss": 0.4695737659931183, "global_step": 16052, "epoch": 180, "lr": 9.918920225481828e-05}
{"train_loss": 0.45660725235939026, "global_step": 16053, "epoch": 180, "lr": 9.918909827166289e-05}
{"train_loss": 0.5956300497055054, "global_step": 16054, "epoch": 180, "lr": 9.918899428189464e-05}
{"train_loss": 0.5086150765419006, "global_step": 16055, "epoch": 180, "lr": 9.918889028551352e-05}
{"train_loss": 0.6366432905197144, "global_step": 16056, "epoch": 180, "lr": 9.918878628251955e-05}
{"train_loss": 0.39747509360313416, "global_step": 16057, "epoch": 180, "lr": 9.918868227291275e-05}
{"train_loss": 0.5900737643241882, "global_step": 16058, "epoch": 180, "lr": 9.918857825669314e-05}
{"train_loss": 0.5542635917663574, "global_step": 16059, "epoch": 180, "lr": 9.918847423386072e-05}
{"train_loss": 0.7034434676170349, "global_step": 16060, "epoch": 180, "lr": 9.91883702044155e-05}
{"train_loss": 0.6765549778938293, "global_step": 16061, "epoch": 180, "lr": 9.918826616835751e-05}
{"train_loss": 0.5157946944236755, "global_step": 16062, "epoch": 180, "lr": 9.918816212568676e-05}
{"train_loss": 0.7174127101898193, "global_step": 16063, "epoch": 180, "lr": 9.918805807640325e-05}
{"train_loss": 0.6087074875831604, "global_step": 16064, "epoch": 180, "lr": 9.918795402050703e-05}
{"train_loss": 0.5426366925239563, "global_step": 16065, "epoch": 180, "lr": 9.918784995799805e-05}
{"train_loss": 0.6333098411560059, "global_step": 16066, "epoch": 180, "lr": 9.91877458888764e-05}
{"train_loss": 0.49638769030570984, "global_step": 16067, "epoch": 180, "lr": 9.918764181314205e-05}
{"train_loss": 0.684619128704071, "global_step": 16068, "epoch": 180, "lr": 9.918753773079502e-05}
{"train_loss": 0.5942927002906799, "global_step": 16069, "epoch": 180, "lr": 9.918743364183531e-05}
{"train_loss": 0.514266848564148, "global_step": 16070, "epoch": 180, "lr": 9.918732954626296e-05}
{"train_loss": 0.6069968342781067, "global_step": 16071, "epoch": 180, "lr": 9.918722544407797e-05}
{"train_loss": 0.6550986170768738, "global_step": 16072, "epoch": 180, "lr": 9.918712133528035e-05}
{"train_loss": 0.7380385398864746, "global_step": 16073, "epoch": 180, "lr": 9.918701721987014e-05}
{"train_loss": 0.47676730155944824, "global_step": 16074, "epoch": 180, "lr": 9.918691309784731e-05}
{"train_loss": 0.4741594195365906, "global_step": 16075, "epoch": 180, "lr": 9.918680896921193e-05}
{"train_loss": 0.7539699673652649, "global_step": 16076, "epoch": 180, "lr": 9.918670483396395e-05}
{"train_loss": 0.5456880927085876, "global_step": 16077, "epoch": 180, "lr": 9.918660069210343e-05}
{"train_loss": 0.5909301042556763, "global_step": 16078, "epoch": 180, "lr": 9.918649654363039e-05}
{"train_loss": 0.5522521734237671, "global_step": 16079, "epoch": 180, "lr": 9.91863923885448e-05}
{"train_loss": 0.49497416615486145, "global_step": 16080, "epoch": 180, "lr": 9.918628822684671e-05}
{"train_loss": 0.5737452507019043, "global_step": 16081, "epoch": 180, "lr": 9.918618405853612e-05}
{"train_loss": 0.5628778338432312, "global_step": 16082, "epoch": 180, "lr": 9.918607988361305e-05}
{"train_loss": 0.4317498803138733, "global_step": 16083, "epoch": 180, "lr": 9.91859757020775e-05}
{"train_loss": 0.47429224848747253, "global_step": 16084, "epoch": 180, "lr": 9.918587151392951e-05}
{"train_loss": 0.6140862703323364, "global_step": 16085, "epoch": 180, "lr": 9.918576731916908e-05}
{"train_loss": 0.67220139503479, "global_step": 16086, "epoch": 180, "lr": 9.918566311779621e-05}
{"train_loss": 0.7420623302459717, "global_step": 16087, "epoch": 180, "lr": 9.918555890981094e-05}
{"train_loss": 0.573980450630188, "global_step": 16088, "epoch": 180, "lr": 9.918545469521326e-05}
{"train_loss": 0.626153290271759, "global_step": 16089, "epoch": 180, "lr": 9.91853504740032e-05}
{"train_loss": 0.5600448846817017, "global_step": 16090, "epoch": 180, "lr": 9.918524624618077e-05}
{"train_loss": 0.49224480986595154, "global_step": 16091, "epoch": 180, "lr": 9.918514201174598e-05}
{"train_loss": 0.5584104061126709, "global_step": 16092, "epoch": 180, "lr": 9.918503777069885e-05}
{"train_loss": 0.5343378186225891, "global_step": 16093, "epoch": 180, "lr": 9.918493352303939e-05}
{"train_loss": 0.5486141443252563, "global_step": 16094, "epoch": 180, "lr": 9.918482926876762e-05}
{"train_loss": 0.5285249352455139, "global_step": 16095, "epoch": 180, "lr": 9.918472500788354e-05}
{"train_loss": 0.551098108291626, "global_step": 16096, "epoch": 180, "lr": 9.91846207403872e-05}
{"train_loss": 0.5511684417724609, "global_step": 16097, "epoch": 180, "lr": 9.918451646627855e-05}
{"train_loss": 0.7071555852890015, "global_step": 16098, "epoch": 180, "lr": 9.918441218555765e-05}
{"train_loss": 0.4038240611553192, "global_step": 16099, "epoch": 180, "lr": 9.918430789822451e-05}
{"train_loss": 0.6758260726928711, "global_step": 16100, "epoch": 180, "lr": 9.918420360427915e-05}
{"train_loss": 0.5759020447731018, "global_step": 16101, "epoch": 180, "lr": 9.918409930372157e-05}
{"train_loss": 0.5913053750991821, "global_step": 16102, "epoch": 180, "lr": 9.918399499655177e-05}
{"train_loss": 0.5157527923583984, "global_step": 16103, "epoch": 180, "lr": 9.91838906827698e-05}
{"train_loss": 0.41598761081695557, "global_step": 16104, "epoch": 180, "lr": 9.918378636237564e-05}
{"train_loss": 0.4364282190799713, "global_step": 16105, "epoch": 180, "lr": 9.918368203536933e-05}
{"train_loss": 0.4059954285621643, "global_step": 16106, "epoch": 180, "lr": 9.918357770175087e-05}
{"train_loss": 0.5624553561210632, "global_step": 16107, "epoch": 180, "lr": 9.918347336152029e-05}
{"train_loss": 0.5571025365524078, "global_step": 16108, "epoch": 180, "lr": 9.918336901467757e-05, "val_loss": 1.5332211256027222, "train_action_mse_error": 27.968341827392578}
{"train_loss": 0.5748775601387024, "global_step": 16109, "epoch": 181, "lr": 9.918326466122276e-05}
{"train_loss": 0.49558961391448975, "global_step": 16110, "epoch": 181, "lr": 9.918316030115586e-05}
{"train_loss": 0.5381594300270081, "global_step": 16111, "epoch": 181, "lr": 9.918305593447687e-05}
{"train_loss": 0.4617520570755005, "global_step": 16112, "epoch": 181, "lr": 9.918295156118582e-05}
{"train_loss": 0.47179532051086426, "global_step": 16113, "epoch": 181, "lr": 9.918284718128274e-05}
{"train_loss": 0.461588054895401, "global_step": 16114, "epoch": 181, "lr": 9.91827427947676e-05}
{"train_loss": 0.46227195858955383, "global_step": 16115, "epoch": 181, "lr": 9.918263840164046e-05}
{"train_loss": 0.48814448714256287, "global_step": 16116, "epoch": 181, "lr": 9.918253400190132e-05}
{"train_loss": 0.5670105814933777, "global_step": 16117, "epoch": 181, "lr": 9.918242959555016e-05}
{"train_loss": 0.3660687208175659, "global_step": 16118, "epoch": 181, "lr": 9.918232518258703e-05}
{"train_loss": 0.5258603096008301, "global_step": 16119, "epoch": 181, "lr": 9.918222076301196e-05}
{"train_loss": 0.6261602640151978, "global_step": 16120, "epoch": 181, "lr": 9.918211633682491e-05}
{"train_loss": 0.5020970702171326, "global_step": 16121, "epoch": 181, "lr": 9.918201190402594e-05}
{"train_loss": 0.46872615814208984, "global_step": 16122, "epoch": 181, "lr": 9.918190746461505e-05}
{"train_loss": 0.5184126496315002, "global_step": 16123, "epoch": 181, "lr": 9.918180301859224e-05}
{"train_loss": 0.5040486454963684, "global_step": 16124, "epoch": 181, "lr": 9.918169856595753e-05}
{"train_loss": 0.4596681594848633, "global_step": 16125, "epoch": 181, "lr": 9.918159410671097e-05}
{"train_loss": 0.6691863536834717, "global_step": 16126, "epoch": 181, "lr": 9.918148964085252e-05}
{"train_loss": 0.5924015045166016, "global_step": 16127, "epoch": 181, "lr": 9.918138516838223e-05}
{"train_loss": 0.5391280055046082, "global_step": 16128, "epoch": 181, "lr": 9.918128068930009e-05}
{"train_loss": 0.4569205641746521, "global_step": 16129, "epoch": 181, "lr": 9.918117620360615e-05}
{"train_loss": 0.5771155953407288, "global_step": 16130, "epoch": 181, "lr": 9.918107171130036e-05}
{"train_loss": 0.5260472297668457, "global_step": 16131, "epoch": 181, "lr": 9.91809672123828e-05}
{"train_loss": 0.45003941655158997, "global_step": 16132, "epoch": 181, "lr": 9.918086270685346e-05}
{"train_loss": 0.5213055610656738, "global_step": 16133, "epoch": 181, "lr": 9.918075819471236e-05}
{"train_loss": 0.6575829386711121, "global_step": 16134, "epoch": 181, "lr": 9.918065367595948e-05}
{"train_loss": 0.5441171526908875, "global_step": 16135, "epoch": 181, "lr": 9.918054915059488e-05}
{"train_loss": 0.5174047946929932, "global_step": 16136, "epoch": 181, "lr": 9.918044461861856e-05}
{"train_loss": 0.48066025972366333, "global_step": 16137, "epoch": 181, "lr": 9.918034008003051e-05}
{"train_loss": 0.6357917189598083, "global_step": 16138, "epoch": 181, "lr": 9.918023553483077e-05}
{"train_loss": 0.5787153244018555, "global_step": 16139, "epoch": 181, "lr": 9.918013098301935e-05}
{"train_loss": 0.5635796189308167, "global_step": 16140, "epoch": 181, "lr": 9.918002642459624e-05}
{"train_loss": 0.5902780890464783, "global_step": 16141, "epoch": 181, "lr": 9.917992185956151e-05}
{"train_loss": 0.6162311434745789, "global_step": 16142, "epoch": 181, "lr": 9.917981728791513e-05}
{"train_loss": 0.5690605640411377, "global_step": 16143, "epoch": 181, "lr": 9.917971270965711e-05}
{"train_loss": 0.6092207431793213, "global_step": 16144, "epoch": 181, "lr": 9.917960812478749e-05}
{"train_loss": 0.5747694373130798, "global_step": 16145, "epoch": 181, "lr": 9.917950353330624e-05}
{"train_loss": 0.4178251028060913, "global_step": 16146, "epoch": 181, "lr": 9.917939893521344e-05}
{"train_loss": 0.5671291947364807, "global_step": 16147, "epoch": 181, "lr": 9.917929433050906e-05}
{"train_loss": 0.530945360660553, "global_step": 16148, "epoch": 181, "lr": 9.917918971919312e-05}
{"train_loss": 0.531663715839386, "global_step": 16149, "epoch": 181, "lr": 9.917908510126563e-05}
{"train_loss": 0.5660861730575562, "global_step": 16150, "epoch": 181, "lr": 9.917898047672663e-05}
{"train_loss": 0.5673851370811462, "global_step": 16151, "epoch": 181, "lr": 9.91788758455761e-05}
{"train_loss": 0.48013627529144287, "global_step": 16152, "epoch": 181, "lr": 9.917877120781409e-05}
{"train_loss": 0.6097201108932495, "global_step": 16153, "epoch": 181, "lr": 9.917866656344057e-05}
{"train_loss": 0.6428717374801636, "global_step": 16154, "epoch": 181, "lr": 9.917856191245558e-05}
{"train_loss": 0.38616305589675903, "global_step": 16155, "epoch": 181, "lr": 9.917845725485914e-05}
{"train_loss": 0.5587692260742188, "global_step": 16156, "epoch": 181, "lr": 9.917835259065126e-05}
{"train_loss": 0.44503307342529297, "global_step": 16157, "epoch": 181, "lr": 9.917824791983194e-05}
{"train_loss": 0.5906316041946411, "global_step": 16158, "epoch": 181, "lr": 9.917814324240121e-05}
{"train_loss": 0.5081772804260254, "global_step": 16159, "epoch": 181, "lr": 9.917803855835908e-05}
{"train_loss": 0.48232102394104004, "global_step": 16160, "epoch": 181, "lr": 9.917793386770555e-05}
{"train_loss": 0.4292603135108948, "global_step": 16161, "epoch": 181, "lr": 9.917782917044066e-05}
{"train_loss": 0.5497353672981262, "global_step": 16162, "epoch": 181, "lr": 9.91777244665644e-05}
{"train_loss": 0.5309972763061523, "global_step": 16163, "epoch": 181, "lr": 9.91776197560768e-05}
{"train_loss": 0.5450469255447388, "global_step": 16164, "epoch": 181, "lr": 9.917751503897787e-05}
{"train_loss": 0.524261474609375, "global_step": 16165, "epoch": 181, "lr": 9.917741031526762e-05}
{"train_loss": 0.5304957628250122, "global_step": 16166, "epoch": 181, "lr": 9.917730558494606e-05}
{"train_loss": 0.4941175878047943, "global_step": 16167, "epoch": 181, "lr": 9.917720084801323e-05}
{"train_loss": 0.6586169004440308, "global_step": 16168, "epoch": 181, "lr": 9.91770961044691e-05}
{"train_loss": 0.5840137600898743, "global_step": 16169, "epoch": 181, "lr": 9.917699135431373e-05}
{"train_loss": 0.6205511689186096, "global_step": 16170, "epoch": 181, "lr": 9.917688659754709e-05}
{"train_loss": 0.6030838489532471, "global_step": 16171, "epoch": 181, "lr": 9.917678183416924e-05}
{"train_loss": 0.593252956867218, "global_step": 16172, "epoch": 181, "lr": 9.917667706418017e-05}
{"train_loss": 0.5034654140472412, "global_step": 16173, "epoch": 181, "lr": 9.917657228757989e-05}
{"train_loss": 0.46135157346725464, "global_step": 16174, "epoch": 181, "lr": 9.91764675043684e-05}
{"train_loss": 0.5557864308357239, "global_step": 16175, "epoch": 181, "lr": 9.917636271454575e-05}
{"train_loss": 0.43484944105148315, "global_step": 16176, "epoch": 181, "lr": 9.917625791811193e-05}
{"train_loss": 0.4850856065750122, "global_step": 16177, "epoch": 181, "lr": 9.917615311506699e-05}
{"train_loss": 0.5616180300712585, "global_step": 16178, "epoch": 181, "lr": 9.917604830541087e-05}
{"train_loss": 0.5601361393928528, "global_step": 16179, "epoch": 181, "lr": 9.917594348914366e-05}
{"train_loss": 0.5351784825325012, "global_step": 16180, "epoch": 181, "lr": 9.917583866626534e-05}
{"train_loss": 0.5434998273849487, "global_step": 16181, "epoch": 181, "lr": 9.917573383677592e-05}
{"train_loss": 0.46643567085266113, "global_step": 16182, "epoch": 181, "lr": 9.917562900067542e-05}
{"train_loss": 0.5417433381080627, "global_step": 16183, "epoch": 181, "lr": 9.917552415796387e-05}
{"train_loss": 0.5138298869132996, "global_step": 16184, "epoch": 181, "lr": 9.917541930864127e-05}
{"train_loss": 0.5247946381568909, "global_step": 16185, "epoch": 181, "lr": 9.917531445270762e-05}
{"train_loss": 0.5047879219055176, "global_step": 16186, "epoch": 181, "lr": 9.917520959016296e-05}
{"train_loss": 0.5387148261070251, "global_step": 16187, "epoch": 181, "lr": 9.917510472100728e-05}
{"train_loss": 0.5858438014984131, "global_step": 16188, "epoch": 181, "lr": 9.917499984524063e-05}
{"train_loss": 0.6042913794517517, "global_step": 16189, "epoch": 181, "lr": 9.917489496286298e-05}
{"train_loss": 0.4942331910133362, "global_step": 16190, "epoch": 181, "lr": 9.917479007387438e-05}
{"train_loss": 0.4369615316390991, "global_step": 16191, "epoch": 181, "lr": 9.917468517827482e-05}
{"train_loss": 0.43882957100868225, "global_step": 16192, "epoch": 181, "lr": 9.917458027606431e-05}
{"train_loss": 0.6583252549171448, "global_step": 16193, "epoch": 181, "lr": 9.91744753672429e-05}
{"train_loss": 0.47907930612564087, "global_step": 16194, "epoch": 181, "lr": 9.917437045181057e-05}
{"train_loss": 0.528458297252655, "global_step": 16195, "epoch": 181, "lr": 9.917426552976735e-05}
{"train_loss": 0.5138513445854187, "global_step": 16196, "epoch": 181, "lr": 9.917416060111323e-05}
{"train_loss": 0.5308243434750632, "global_step": 16197, "epoch": 181, "lr": 9.917405566584826e-05, "val_loss": 1.5558860301971436}
{"train_loss": 0.48906177282333374, "global_step": 16198, "epoch": 182, "lr": 9.917395072397244e-05}
{"train_loss": 0.46834903955459595, "global_step": 16199, "epoch": 182, "lr": 9.917384577548576e-05}
{"train_loss": 0.5003593564033508, "global_step": 16200, "epoch": 182, "lr": 9.917374082038828e-05}
{"train_loss": 0.567602813243866, "global_step": 16201, "epoch": 182, "lr": 9.917363585867998e-05}
{"train_loss": 0.5699372291564941, "global_step": 16202, "epoch": 182, "lr": 9.91735308903609e-05}
{"train_loss": 0.4957371950149536, "global_step": 16203, "epoch": 182, "lr": 9.917342591543102e-05}
{"train_loss": 0.5288130044937134, "global_step": 16204, "epoch": 182, "lr": 9.917332093389038e-05}
{"train_loss": 0.5694405436515808, "global_step": 16205, "epoch": 182, "lr": 9.917321594573899e-05}
{"train_loss": 0.5367287397384644, "global_step": 16206, "epoch": 182, "lr": 9.917311095097685e-05}
{"train_loss": 0.6194043159484863, "global_step": 16207, "epoch": 182, "lr": 9.917300594960398e-05}
{"train_loss": 0.4222279191017151, "global_step": 16208, "epoch": 182, "lr": 9.917290094162041e-05}
{"train_loss": 0.5910696387290955, "global_step": 16209, "epoch": 182, "lr": 9.917279592702614e-05}
{"train_loss": 0.6743753552436829, "global_step": 16210, "epoch": 182, "lr": 9.917269090582118e-05}
{"train_loss": 0.6573561429977417, "global_step": 16211, "epoch": 182, "lr": 9.917258587800557e-05}
{"train_loss": 0.6738399863243103, "global_step": 16212, "epoch": 182, "lr": 9.917248084357929e-05}
{"train_loss": 0.4133182168006897, "global_step": 16213, "epoch": 182, "lr": 9.917237580254237e-05}
{"train_loss": 0.6537603139877319, "global_step": 16214, "epoch": 182, "lr": 9.917227075489483e-05}
{"train_loss": 0.5198017358779907, "global_step": 16215, "epoch": 182, "lr": 9.917216570063668e-05}
{"train_loss": 0.47509050369262695, "global_step": 16216, "epoch": 182, "lr": 9.917206063976791e-05}
{"train_loss": 0.6526645421981812, "global_step": 16217, "epoch": 182, "lr": 9.917195557228858e-05}
{"train_loss": 0.5872878432273865, "global_step": 16218, "epoch": 182, "lr": 9.917185049819866e-05}
{"train_loss": 0.4874056577682495, "global_step": 16219, "epoch": 182, "lr": 9.917174541749821e-05}
{"train_loss": 0.5871658325195312, "global_step": 16220, "epoch": 182, "lr": 9.917164033018719e-05}
{"train_loss": 0.689419686794281, "global_step": 16221, "epoch": 182, "lr": 9.917153523626566e-05}
{"train_loss": 0.4249730110168457, "global_step": 16222, "epoch": 182, "lr": 9.917143013573361e-05}
{"train_loss": 0.5300166606903076, "global_step": 16223, "epoch": 182, "lr": 9.917132502859107e-05}
{"train_loss": 0.5066111087799072, "global_step": 16224, "epoch": 182, "lr": 9.917121991483802e-05}
{"train_loss": 0.6885378360748291, "global_step": 16225, "epoch": 182, "lr": 9.917111479447453e-05}
{"train_loss": 0.4770019054412842, "global_step": 16226, "epoch": 182, "lr": 9.917100966750057e-05}
{"train_loss": 0.51995849609375, "global_step": 16227, "epoch": 182, "lr": 9.917090453391615e-05}
{"train_loss": 0.49871501326560974, "global_step": 16228, "epoch": 182, "lr": 9.917079939372133e-05}
{"train_loss": 0.4710148870944977, "global_step": 16229, "epoch": 182, "lr": 9.917069424691607e-05}
{"train_loss": 0.4748108386993408, "global_step": 16230, "epoch": 182, "lr": 9.917058909350044e-05}
{"train_loss": 0.49525928497314453, "global_step": 16231, "epoch": 182, "lr": 9.917048393347439e-05}
{"train_loss": 0.5793464183807373, "global_step": 16232, "epoch": 182, "lr": 9.917037876683799e-05}
{"train_loss": 0.38877010345458984, "global_step": 16233, "epoch": 182, "lr": 9.917027359359122e-05}
{"train_loss": 0.5174649357795715, "global_step": 16234, "epoch": 182, "lr": 9.917016841373412e-05}
{"train_loss": 0.6509184837341309, "global_step": 16235, "epoch": 182, "lr": 9.917006322726667e-05}
{"train_loss": 0.5830702781677246, "global_step": 16236, "epoch": 182, "lr": 9.916995803418891e-05}
{"train_loss": 0.5876510143280029, "global_step": 16237, "epoch": 182, "lr": 9.916985283450087e-05}
{"train_loss": 0.6329719424247742, "global_step": 16238, "epoch": 182, "lr": 9.916974762820252e-05}
{"train_loss": 0.5919535756111145, "global_step": 16239, "epoch": 182, "lr": 9.91696424152939e-05}
{"train_loss": 0.599949836730957, "global_step": 16240, "epoch": 182, "lr": 9.916953719577503e-05}
{"train_loss": 0.5206313729286194, "global_step": 16241, "epoch": 182, "lr": 9.916943196964591e-05}
{"train_loss": 0.6474355459213257, "global_step": 16242, "epoch": 182, "lr": 9.916932673690656e-05}
{"train_loss": 0.3994271159172058, "global_step": 16243, "epoch": 182, "lr": 9.916922149755699e-05}
{"train_loss": 0.5337234735488892, "global_step": 16244, "epoch": 182, "lr": 9.916911625159721e-05}
{"train_loss": 0.535471498966217, "global_step": 16245, "epoch": 182, "lr": 9.916901099902726e-05}
{"train_loss": 0.6020991802215576, "global_step": 16246, "epoch": 182, "lr": 9.916890573984712e-05}
{"train_loss": 0.5901258587837219, "global_step": 16247, "epoch": 182, "lr": 9.916880047405683e-05}
{"train_loss": 0.5563406944274902, "global_step": 16248, "epoch": 182, "lr": 9.91686952016564e-05}
{"train_loss": 0.6215415596961975, "global_step": 16249, "epoch": 182, "lr": 9.916858992264582e-05}
{"train_loss": 0.5758566856384277, "global_step": 16250, "epoch": 182, "lr": 9.916848463702513e-05}
{"train_loss": 0.5406675934791565, "global_step": 16251, "epoch": 182, "lr": 9.916837934479433e-05}
{"train_loss": 0.5450992584228516, "global_step": 16252, "epoch": 182, "lr": 9.916827404595346e-05}
{"train_loss": 0.6334212422370911, "global_step": 16253, "epoch": 182, "lr": 9.91681687405025e-05}
{"train_loss": 0.6074547171592712, "global_step": 16254, "epoch": 182, "lr": 9.916806342844148e-05}
{"train_loss": 0.6092609167098999, "global_step": 16255, "epoch": 182, "lr": 9.916795810977042e-05}
{"train_loss": 0.5181991457939148, "global_step": 16256, "epoch": 182, "lr": 9.916785278448931e-05}
{"train_loss": 0.4845461845397949, "global_step": 16257, "epoch": 182, "lr": 9.916774745259821e-05}
{"train_loss": 0.43591222167015076, "global_step": 16258, "epoch": 182, "lr": 9.916764211409708e-05}
{"train_loss": 0.5174711346626282, "global_step": 16259, "epoch": 182, "lr": 9.916753676898596e-05}
{"train_loss": 0.5428006649017334, "global_step": 16260, "epoch": 182, "lr": 9.916743141726488e-05}
{"train_loss": 0.6245650053024292, "global_step": 16261, "epoch": 182, "lr": 9.916732605893383e-05}
{"train_loss": 0.47630488872528076, "global_step": 16262, "epoch": 182, "lr": 9.916722069399283e-05}
{"train_loss": 0.5860247611999512, "global_step": 16263, "epoch": 182, "lr": 9.916711532244191e-05}
{"train_loss": 0.5064249038696289, "global_step": 16264, "epoch": 182, "lr": 9.916700994428106e-05}
{"train_loss": 0.6024202108383179, "global_step": 16265, "epoch": 182, "lr": 9.91669045595103e-05}
{"train_loss": 0.5850566625595093, "global_step": 16266, "epoch": 182, "lr": 9.916679916812965e-05}
{"train_loss": 0.5398007035255432, "global_step": 16267, "epoch": 182, "lr": 9.916669377013913e-05}
{"train_loss": 0.5086346864700317, "global_step": 16268, "epoch": 182, "lr": 9.916658836553875e-05}
{"train_loss": 0.5018706321716309, "global_step": 16269, "epoch": 182, "lr": 9.916648295432851e-05}
{"train_loss": 0.5236402750015259, "global_step": 16270, "epoch": 182, "lr": 9.916637753650844e-05}
{"train_loss": 0.641094982624054, "global_step": 16271, "epoch": 182, "lr": 9.916627211207856e-05}
{"train_loss": 0.6474669575691223, "global_step": 16272, "epoch": 182, "lr": 9.916616668103886e-05}
{"train_loss": 0.6317413449287415, "global_step": 16273, "epoch": 182, "lr": 9.916606124338938e-05}
{"train_loss": 0.567054808139801, "global_step": 16274, "epoch": 182, "lr": 9.91659557991301e-05}
{"train_loss": 0.5912969708442688, "global_step": 16275, "epoch": 182, "lr": 9.916585034826107e-05}
{"train_loss": 0.5202062129974365, "global_step": 16276, "epoch": 182, "lr": 9.916574489078229e-05}
{"train_loss": 0.48386332392692566, "global_step": 16277, "epoch": 182, "lr": 9.91656394266938e-05}
{"train_loss": 0.6041066646575928, "global_step": 16278, "epoch": 182, "lr": 9.916553395599557e-05}
{"train_loss": 0.6120688319206238, "global_step": 16279, "epoch": 182, "lr": 9.916542847868762e-05}
{"train_loss": 0.6145538687705994, "global_step": 16280, "epoch": 182, "lr": 9.916532299477e-05}
{"train_loss": 0.5881778001785278, "global_step": 16281, "epoch": 182, "lr": 9.916521750424267e-05}
{"train_loss": 0.4416848421096802, "global_step": 16282, "epoch": 182, "lr": 9.91651120071057e-05}
{"train_loss": 0.45394039154052734, "global_step": 16283, "epoch": 182, "lr": 9.916500650335907e-05}
{"train_loss": 0.5147208571434021, "global_step": 16284, "epoch": 182, "lr": 9.916490099300282e-05}
{"train_loss": 0.4949897229671478, "global_step": 16285, "epoch": 182, "lr": 9.916479547603693e-05}
{"train_loss": 0.5508377428135175, "global_step": 16286, "epoch": 182, "lr": 9.916468995246144e-05, "val_loss": 1.5185415744781494}
{"train_loss": 0.6226785182952881, "global_step": 16287, "epoch": 183, "lr": 9.916458442227636e-05}
{"train_loss": 0.4380896985530853, "global_step": 16288, "epoch": 183, "lr": 9.91644788854817e-05}
{"train_loss": 0.6408929228782654, "global_step": 16289, "epoch": 183, "lr": 9.916437334207747e-05}
{"train_loss": 0.5742500424385071, "global_step": 16290, "epoch": 183, "lr": 9.916426779206369e-05}
{"train_loss": 0.5799062848091125, "global_step": 16291, "epoch": 183, "lr": 9.916416223544037e-05}
{"train_loss": 0.638244092464447, "global_step": 16292, "epoch": 183, "lr": 9.916405667220754e-05}
{"train_loss": 0.6072977185249329, "global_step": 16293, "epoch": 183, "lr": 9.91639511023652e-05}
{"train_loss": 0.5525241494178772, "global_step": 16294, "epoch": 183, "lr": 9.916384552591336e-05}
{"train_loss": 0.5452638864517212, "global_step": 16295, "epoch": 183, "lr": 9.916373994285203e-05}
{"train_loss": 0.4971766173839569, "global_step": 16296, "epoch": 183, "lr": 9.916363435318124e-05}
{"train_loss": 0.5014631152153015, "global_step": 16297, "epoch": 183, "lr": 9.9163528756901e-05}
{"train_loss": 0.5641226172447205, "global_step": 16298, "epoch": 183, "lr": 9.916342315401134e-05}
{"train_loss": 0.4713011384010315, "global_step": 16299, "epoch": 183, "lr": 9.916331754451225e-05}
{"train_loss": 0.6061652898788452, "global_step": 16300, "epoch": 183, "lr": 9.916321192840373e-05}
{"train_loss": 0.4020980894565582, "global_step": 16301, "epoch": 183, "lr": 9.916310630568584e-05}
{"train_loss": 0.580795168876648, "global_step": 16302, "epoch": 183, "lr": 9.916300067635856e-05}
{"train_loss": 0.5145671367645264, "global_step": 16303, "epoch": 183, "lr": 9.916289504042192e-05}
{"train_loss": 0.7254584431648254, "global_step": 16304, "epoch": 183, "lr": 9.916278939787591e-05}
{"train_loss": 0.4253831207752228, "global_step": 16305, "epoch": 183, "lr": 9.916268374872058e-05}
{"train_loss": 0.5007226467132568, "global_step": 16306, "epoch": 183, "lr": 9.916257809295591e-05}
{"train_loss": 0.5319648385047913, "global_step": 16307, "epoch": 183, "lr": 9.916247243058194e-05}
{"train_loss": 0.5589354038238525, "global_step": 16308, "epoch": 183, "lr": 9.916236676159867e-05}
{"train_loss": 0.3964517414569855, "global_step": 16309, "epoch": 183, "lr": 9.916226108600612e-05}
{"train_loss": 0.5401431322097778, "global_step": 16310, "epoch": 183, "lr": 9.916215540380431e-05}
{"train_loss": 0.6265108585357666, "global_step": 16311, "epoch": 183, "lr": 9.916204971499325e-05}
{"train_loss": 0.6922071576118469, "global_step": 16312, "epoch": 183, "lr": 9.916194401957296e-05}
{"train_loss": 0.45662981271743774, "global_step": 16313, "epoch": 183, "lr": 9.916183831754342e-05}
{"train_loss": 0.5790047645568848, "global_step": 16314, "epoch": 183, "lr": 9.916173260890469e-05}
{"train_loss": 0.43568405508995056, "global_step": 16315, "epoch": 183, "lr": 9.916162689365674e-05}
{"train_loss": 0.6433103084564209, "global_step": 16316, "epoch": 183, "lr": 9.916152117179963e-05}
{"train_loss": 0.7695820927619934, "global_step": 16317, "epoch": 183, "lr": 9.916141544333335e-05}
{"train_loss": 0.6242115497589111, "global_step": 16318, "epoch": 183, "lr": 9.916130970825792e-05}
{"train_loss": 0.7181941866874695, "global_step": 16319, "epoch": 183, "lr": 9.916120396657333e-05}
{"train_loss": 0.4745800495147705, "global_step": 16320, "epoch": 183, "lr": 9.916109821827964e-05}
{"train_loss": 0.4938075542449951, "global_step": 16321, "epoch": 183, "lr": 9.916099246337683e-05}
{"train_loss": 0.6675373911857605, "global_step": 16322, "epoch": 183, "lr": 9.916088670186492e-05}
{"train_loss": 0.6312230229377747, "global_step": 16323, "epoch": 183, "lr": 9.916078093374393e-05}
{"train_loss": 0.585319459438324, "global_step": 16324, "epoch": 183, "lr": 9.916067515901388e-05}
{"train_loss": 0.4939902126789093, "global_step": 16325, "epoch": 183, "lr": 9.916056937767478e-05}
{"train_loss": 0.4914376139640808, "global_step": 16326, "epoch": 183, "lr": 9.916046358972662e-05}
{"train_loss": 0.5313229560852051, "global_step": 16327, "epoch": 183, "lr": 9.916035779516946e-05}
{"train_loss": 0.6064794063568115, "global_step": 16328, "epoch": 183, "lr": 9.916025199400327e-05}
{"train_loss": 0.6038673520088196, "global_step": 16329, "epoch": 183, "lr": 9.916014618622809e-05}
{"train_loss": 0.607059895992279, "global_step": 16330, "epoch": 183, "lr": 9.916004037184392e-05}
{"train_loss": 0.5030100345611572, "global_step": 16331, "epoch": 183, "lr": 9.915993455085079e-05}
{"train_loss": 0.6395187973976135, "global_step": 16332, "epoch": 183, "lr": 9.915982872324869e-05}
{"train_loss": 0.5516878366470337, "global_step": 16333, "epoch": 183, "lr": 9.915972288903768e-05}
{"train_loss": 0.6058622598648071, "global_step": 16334, "epoch": 183, "lr": 9.915961704821773e-05}
{"train_loss": 0.6388791799545288, "global_step": 16335, "epoch": 183, "lr": 9.915951120078886e-05}
{"train_loss": 0.4701084494590759, "global_step": 16336, "epoch": 183, "lr": 9.91594053467511e-05}
{"train_loss": 0.7136580348014832, "global_step": 16337, "epoch": 183, "lr": 9.915929948610444e-05}
{"train_loss": 0.5086371898651123, "global_step": 16338, "epoch": 183, "lr": 9.915919361884894e-05}
{"train_loss": 0.533515453338623, "global_step": 16339, "epoch": 183, "lr": 9.915908774498458e-05}
{"train_loss": 0.5555229187011719, "global_step": 16340, "epoch": 183, "lr": 9.915898186451138e-05}
{"train_loss": 0.46791088581085205, "global_step": 16341, "epoch": 183, "lr": 9.915887597742935e-05}
{"train_loss": 0.5879664421081543, "global_step": 16342, "epoch": 183, "lr": 9.915877008373849e-05}
{"train_loss": 0.7110442519187927, "global_step": 16343, "epoch": 183, "lr": 9.915866418343885e-05}
{"train_loss": 0.5209519267082214, "global_step": 16344, "epoch": 183, "lr": 9.915855827653043e-05}
{"train_loss": 0.4529200494289398, "global_step": 16345, "epoch": 183, "lr": 9.915845236301325e-05}
{"train_loss": 0.6680462956428528, "global_step": 16346, "epoch": 183, "lr": 9.915834644288728e-05}
{"train_loss": 0.4621047377586365, "global_step": 16347, "epoch": 183, "lr": 9.91582405161526e-05}
{"train_loss": 0.5422906279563904, "global_step": 16348, "epoch": 183, "lr": 9.915813458280919e-05}
{"train_loss": 0.5094611644744873, "global_step": 16349, "epoch": 183, "lr": 9.915802864285706e-05}
{"train_loss": 0.624785840511322, "global_step": 16350, "epoch": 183, "lr": 9.915792269629624e-05}
{"train_loss": 0.5113033056259155, "global_step": 16351, "epoch": 183, "lr": 9.915781674312672e-05}
{"train_loss": 0.5450217127799988, "global_step": 16352, "epoch": 183, "lr": 9.915771078334855e-05}
{"train_loss": 0.5649674534797668, "global_step": 16353, "epoch": 183, "lr": 9.915760481696172e-05}
{"train_loss": 0.5736464262008667, "global_step": 16354, "epoch": 183, "lr": 9.915749884396623e-05}
{"train_loss": 0.4919908940792084, "global_step": 16355, "epoch": 183, "lr": 9.915739286436214e-05}
{"train_loss": 0.5815252065658569, "global_step": 16356, "epoch": 183, "lr": 9.915728687814942e-05}
{"train_loss": 0.655889630317688, "global_step": 16357, "epoch": 183, "lr": 9.915718088532812e-05}
{"train_loss": 0.6194736957550049, "global_step": 16358, "epoch": 183, "lr": 9.915707488589822e-05}
{"train_loss": 0.6533769965171814, "global_step": 16359, "epoch": 183, "lr": 9.915696887985976e-05}
{"train_loss": 0.5070821642875671, "global_step": 16360, "epoch": 183, "lr": 9.915686286721272e-05}
{"train_loss": 0.6149863004684448, "global_step": 16361, "epoch": 183, "lr": 9.915675684795717e-05}
{"train_loss": 0.5547455549240112, "global_step": 16362, "epoch": 183, "lr": 9.915665082209309e-05}
{"train_loss": 0.5579614639282227, "global_step": 16363, "epoch": 183, "lr": 9.915654478962049e-05}
{"train_loss": 0.5946976542472839, "global_step": 16364, "epoch": 183, "lr": 9.91564387505394e-05}
{"train_loss": 0.6202347874641418, "global_step": 16365, "epoch": 183, "lr": 9.91563327048498e-05}
{"train_loss": 0.5257017016410828, "global_step": 16366, "epoch": 183, "lr": 9.915622665255176e-05}
{"train_loss": 0.6239715218544006, "global_step": 16367, "epoch": 183, "lr": 9.915612059364526e-05}
{"train_loss": 0.5253444910049438, "global_step": 16368, "epoch": 183, "lr": 9.91560145281303e-05}
{"train_loss": 0.6253070831298828, "global_step": 16369, "epoch": 183, "lr": 9.915590845600694e-05}
{"train_loss": 0.5764479041099548, "global_step": 16370, "epoch": 183, "lr": 9.915580237727513e-05}
{"train_loss": 0.6364646553993225, "global_step": 16371, "epoch": 183, "lr": 9.915569629193495e-05}
{"train_loss": 0.6463049650192261, "global_step": 16372, "epoch": 183, "lr": 9.915559019998637e-05}
{"train_loss": 0.573711633682251, "global_step": 16373, "epoch": 183, "lr": 9.915548410142943e-05}
{"train_loss": 0.4833540618419647, "global_step": 16374, "epoch": 183, "lr": 9.915537799626413e-05}
{"train_loss": 0.5673662611607755, "global_step": 16375, "epoch": 183, "lr": 9.91552718844905e-05, "val_loss": 1.526196837425232}
{"train_loss": 0.6091825366020203, "global_step": 16376, "epoch": 184, "lr": 9.915516576610853e-05}
{"train_loss": 0.47738704085350037, "global_step": 16377, "epoch": 184, "lr": 9.915505964111825e-05}
{"train_loss": 0.5439181923866272, "global_step": 16378, "epoch": 184, "lr": 9.915495350951968e-05}
{"train_loss": 0.49552372097969055, "global_step": 16379, "epoch": 184, "lr": 9.915484737131283e-05}
{"train_loss": 0.3827170729637146, "global_step": 16380, "epoch": 184, "lr": 9.915474122649768e-05}
{"train_loss": 0.46235156059265137, "global_step": 16381, "epoch": 184, "lr": 9.91546350750743e-05}
{"train_loss": 0.6672530770301819, "global_step": 16382, "epoch": 184, "lr": 9.915452891704267e-05}
{"train_loss": 0.647872805595398, "global_step": 16383, "epoch": 184, "lr": 9.91544227524028e-05}
{"train_loss": 0.48516711592674255, "global_step": 16384, "epoch": 184, "lr": 9.915431658115474e-05}
{"train_loss": 0.5319758057594299, "global_step": 16385, "epoch": 184, "lr": 9.915421040329847e-05}
{"train_loss": 0.6105411052703857, "global_step": 16386, "epoch": 184, "lr": 9.915410421883402e-05}
{"train_loss": 0.5231225490570068, "global_step": 16387, "epoch": 184, "lr": 9.915399802776139e-05}
{"train_loss": 0.5341213345527649, "global_step": 16388, "epoch": 184, "lr": 9.915389183008061e-05}
{"train_loss": 0.5977510213851929, "global_step": 16389, "epoch": 184, "lr": 9.91537856257917e-05}
{"train_loss": 0.5172476172447205, "global_step": 16390, "epoch": 184, "lr": 9.915367941489466e-05}
{"train_loss": 0.5124762654304504, "global_step": 16391, "epoch": 184, "lr": 9.915357319738949e-05}
{"train_loss": 0.46725285053253174, "global_step": 16392, "epoch": 184, "lr": 9.915346697327623e-05}
{"train_loss": 0.4740760624408722, "global_step": 16393, "epoch": 184, "lr": 9.915336074255488e-05}
{"train_loss": 0.5191280841827393, "global_step": 16394, "epoch": 184, "lr": 9.915325450522547e-05}
{"train_loss": 0.3948208689689636, "global_step": 16395, "epoch": 184, "lr": 9.9153148261288e-05}
{"train_loss": 0.5204864740371704, "global_step": 16396, "epoch": 184, "lr": 9.915304201074249e-05}
{"train_loss": 0.5088668465614319, "global_step": 16397, "epoch": 184, "lr": 9.915293575358894e-05}
{"train_loss": 0.5789438486099243, "global_step": 16398, "epoch": 184, "lr": 9.915282948982739e-05}
{"train_loss": 0.5601853132247925, "global_step": 16399, "epoch": 184, "lr": 9.915272321945785e-05}
{"train_loss": 0.5302047729492188, "global_step": 16400, "epoch": 184, "lr": 9.91526169424803e-05}
{"train_loss": 0.5533323884010315, "global_step": 16401, "epoch": 184, "lr": 9.91525106588948e-05}
{"train_loss": 0.5554924011230469, "global_step": 16402, "epoch": 184, "lr": 9.915240436870134e-05}
{"train_loss": 0.6346105933189392, "global_step": 16403, "epoch": 184, "lr": 9.915229807189993e-05}
{"train_loss": 0.525483250617981, "global_step": 16404, "epoch": 184, "lr": 9.91521917684906e-05}
{"train_loss": 0.6695957183837891, "global_step": 16405, "epoch": 184, "lr": 9.915208545847335e-05}
{"train_loss": 0.5728117227554321, "global_step": 16406, "epoch": 184, "lr": 9.915197914184822e-05}
{"train_loss": 0.5758285522460938, "global_step": 16407, "epoch": 184, "lr": 9.915187281861518e-05}
{"train_loss": 0.6242064237594604, "global_step": 16408, "epoch": 184, "lr": 9.915176648877429e-05}
{"train_loss": 0.6748145222663879, "global_step": 16409, "epoch": 184, "lr": 9.915166015232553e-05}
{"train_loss": 0.5388320684432983, "global_step": 16410, "epoch": 184, "lr": 9.915155380926894e-05}
{"train_loss": 0.5656262636184692, "global_step": 16411, "epoch": 184, "lr": 9.915144745960453e-05}
{"train_loss": 0.5655336380004883, "global_step": 16412, "epoch": 184, "lr": 9.915134110333228e-05}
{"train_loss": 0.5988363027572632, "global_step": 16413, "epoch": 184, "lr": 9.915123474045224e-05}
{"train_loss": 0.6493432521820068, "global_step": 16414, "epoch": 184, "lr": 9.915112837096443e-05}
{"train_loss": 0.7370391488075256, "global_step": 16415, "epoch": 184, "lr": 9.915102199486883e-05}
{"train_loss": 0.54863440990448, "global_step": 16416, "epoch": 184, "lr": 9.915091561216548e-05}
{"train_loss": 0.42176952958106995, "global_step": 16417, "epoch": 184, "lr": 9.91508092228544e-05}
{"train_loss": 0.5661841630935669, "global_step": 16418, "epoch": 184, "lr": 9.915070282693559e-05}
{"train_loss": 0.5362208485603333, "global_step": 16419, "epoch": 184, "lr": 9.915059642440906e-05}
{"train_loss": 0.5613316893577576, "global_step": 16420, "epoch": 184, "lr": 9.915049001527483e-05}
{"train_loss": 0.5739464163780212, "global_step": 16421, "epoch": 184, "lr": 9.915038359953293e-05}
{"train_loss": 0.6282962560653687, "global_step": 16422, "epoch": 184, "lr": 9.915027717718334e-05}
{"train_loss": 0.5366597175598145, "global_step": 16423, "epoch": 184, "lr": 9.91501707482261e-05}
{"train_loss": 0.4148087501525879, "global_step": 16424, "epoch": 184, "lr": 9.915006431266123e-05}
{"train_loss": 0.5347558259963989, "global_step": 16425, "epoch": 184, "lr": 9.914995787048873e-05}
{"train_loss": 0.6885495185852051, "global_step": 16426, "epoch": 184, "lr": 9.91498514217086e-05}
{"train_loss": 0.8403397798538208, "global_step": 16427, "epoch": 184, "lr": 9.91497449663209e-05}
{"train_loss": 0.5170484185218811, "global_step": 16428, "epoch": 184, "lr": 9.914963850432559e-05}
{"train_loss": 0.5928767919540405, "global_step": 16429, "epoch": 184, "lr": 9.914953203572272e-05}
{"train_loss": 0.5054680109024048, "global_step": 16430, "epoch": 184, "lr": 9.91494255605123e-05}
{"train_loss": 0.550908088684082, "global_step": 16431, "epoch": 184, "lr": 9.914931907869433e-05}
{"train_loss": 0.5821505784988403, "global_step": 16432, "epoch": 184, "lr": 9.914921259026883e-05}
{"train_loss": 0.6134353280067444, "global_step": 16433, "epoch": 184, "lr": 9.914910609523584e-05}
{"train_loss": 0.5043485760688782, "global_step": 16434, "epoch": 184, "lr": 9.914899959359534e-05}
{"train_loss": 0.49938222765922546, "global_step": 16435, "epoch": 184, "lr": 9.914889308534734e-05}
{"train_loss": 0.5234252214431763, "global_step": 16436, "epoch": 184, "lr": 9.914878657049189e-05}
{"train_loss": 0.8587976694107056, "global_step": 16437, "epoch": 184, "lr": 9.914868004902898e-05}
{"train_loss": 0.5191353559494019, "global_step": 16438, "epoch": 184, "lr": 9.914857352095861e-05}
{"train_loss": 0.5324995517730713, "global_step": 16439, "epoch": 184, "lr": 9.914846698628084e-05}
{"train_loss": 0.5973871946334839, "global_step": 16440, "epoch": 184, "lr": 9.914836044499565e-05}
{"train_loss": 0.4149097502231598, "global_step": 16441, "epoch": 184, "lr": 9.914825389710306e-05}
{"train_loss": 0.5643795728683472, "global_step": 16442, "epoch": 184, "lr": 9.914814734260306e-05}
{"train_loss": 0.627103328704834, "global_step": 16443, "epoch": 184, "lr": 9.914804078149573e-05}
{"train_loss": 0.5425291061401367, "global_step": 16444, "epoch": 184, "lr": 9.914793421378103e-05}
{"train_loss": 0.4787159562110901, "global_step": 16445, "epoch": 184, "lr": 9.9147827639459e-05}
{"train_loss": 0.5611652135848999, "global_step": 16446, "epoch": 184, "lr": 9.914772105852963e-05}
{"train_loss": 0.5376793742179871, "global_step": 16447, "epoch": 184, "lr": 9.914761447099295e-05}
{"train_loss": 0.4348362982273102, "global_step": 16448, "epoch": 184, "lr": 9.914750787684898e-05}
{"train_loss": 0.570018470287323, "global_step": 16449, "epoch": 184, "lr": 9.914740127609772e-05}
{"train_loss": 0.6396206021308899, "global_step": 16450, "epoch": 184, "lr": 9.914729466873919e-05}
{"train_loss": 0.49597054719924927, "global_step": 16451, "epoch": 184, "lr": 9.91471880547734e-05}
{"train_loss": 0.5185011625289917, "global_step": 16452, "epoch": 184, "lr": 9.914708143420037e-05}
{"train_loss": 0.4637957811355591, "global_step": 16453, "epoch": 184, "lr": 9.914697480702013e-05}
{"train_loss": 0.5250985622406006, "global_step": 16454, "epoch": 184, "lr": 9.914686817323266e-05}
{"train_loss": 0.6299629211425781, "global_step": 16455, "epoch": 184, "lr": 9.914676153283801e-05}
{"train_loss": 0.5441014766693115, "global_step": 16456, "epoch": 184, "lr": 9.914665488583617e-05}
{"train_loss": 0.6541637778282166, "global_step": 16457, "epoch": 184, "lr": 9.914654823222716e-05}
{"train_loss": 0.6741125583648682, "global_step": 16458, "epoch": 184, "lr": 9.914644157201099e-05}
{"train_loss": 0.6427817344665527, "global_step": 16459, "epoch": 184, "lr": 9.914633490518769e-05}
{"train_loss": 0.591861367225647, "global_step": 16460, "epoch": 184, "lr": 9.914622823175725e-05}
{"train_loss": 0.6320415735244751, "global_step": 16461, "epoch": 184, "lr": 9.914612155171971e-05}
{"train_loss": 0.34950119256973267, "global_step": 16462, "epoch": 184, "lr": 9.914601486507508e-05}
{"train_loss": 0.5532549619674683, "global_step": 16463, "epoch": 184, "lr": 9.914590817182335e-05}
{"train_loss": 0.5574286700634474, "global_step": 16464, "epoch": 184, "lr": 9.914580147196456e-05, "val_loss": 1.6054192781448364}
{"train_loss": 0.687031626701355, "global_step": 16465, "epoch": 185, "lr": 9.914569476549873e-05}
{"train_loss": 0.5837520956993103, "global_step": 16466, "epoch": 185, "lr": 9.914558805242585e-05}
{"train_loss": 0.5589689016342163, "global_step": 16467, "epoch": 185, "lr": 9.914548133274593e-05}
{"train_loss": 0.6193657517433167, "global_step": 16468, "epoch": 185, "lr": 9.914537460645901e-05}
{"train_loss": 0.4813688099384308, "global_step": 16469, "epoch": 185, "lr": 9.914526787356509e-05}
{"train_loss": 0.58036208152771, "global_step": 16470, "epoch": 185, "lr": 9.91451611340642e-05}
{"train_loss": 0.6157547831535339, "global_step": 16471, "epoch": 185, "lr": 9.914505438795631e-05}
{"train_loss": 0.4956490397453308, "global_step": 16472, "epoch": 185, "lr": 9.91449476352415e-05}
{"train_loss": 0.49684083461761475, "global_step": 16473, "epoch": 185, "lr": 9.914484087591974e-05}
{"train_loss": 0.5749475955963135, "global_step": 16474, "epoch": 185, "lr": 9.914473410999106e-05}
{"train_loss": 0.5640712976455688, "global_step": 16475, "epoch": 185, "lr": 9.914462733745546e-05}
{"train_loss": 0.589424192905426, "global_step": 16476, "epoch": 185, "lr": 9.914452055831297e-05}
{"train_loss": 0.45014286041259766, "global_step": 16477, "epoch": 185, "lr": 9.914441377256359e-05}
{"train_loss": 0.5186521410942078, "global_step": 16478, "epoch": 185, "lr": 9.914430698020735e-05}
{"train_loss": 0.5921561121940613, "global_step": 16479, "epoch": 185, "lr": 9.914420018124426e-05}
{"train_loss": 0.5067845582962036, "global_step": 16480, "epoch": 185, "lr": 9.914409337567432e-05}
{"train_loss": 0.44978946447372437, "global_step": 16481, "epoch": 185, "lr": 9.914398656349757e-05}
{"train_loss": 0.5260578393936157, "global_step": 16482, "epoch": 185, "lr": 9.9143879744714e-05}
{"train_loss": 0.48972639441490173, "global_step": 16483, "epoch": 185, "lr": 9.914377291932365e-05}
{"train_loss": 0.6013967990875244, "global_step": 16484, "epoch": 185, "lr": 9.91436660873265e-05}
{"train_loss": 0.6456392407417297, "global_step": 16485, "epoch": 185, "lr": 9.914355924872259e-05}
{"train_loss": 0.5128630995750427, "global_step": 16486, "epoch": 185, "lr": 9.914345240351193e-05}
{"train_loss": 0.5079673528671265, "global_step": 16487, "epoch": 185, "lr": 9.914334555169453e-05}
{"train_loss": 0.4951730966567993, "global_step": 16488, "epoch": 185, "lr": 9.91432386932704e-05}
{"train_loss": 0.40563759207725525, "global_step": 16489, "epoch": 185, "lr": 9.914313182823956e-05}
{"train_loss": 0.46673789620399475, "global_step": 16490, "epoch": 185, "lr": 9.914302495660203e-05}
{"train_loss": 0.5293768644332886, "global_step": 16491, "epoch": 185, "lr": 9.914291807835782e-05}
{"train_loss": 0.5031573176383972, "global_step": 16492, "epoch": 185, "lr": 9.914281119350695e-05}
{"train_loss": 0.4698769450187683, "global_step": 16493, "epoch": 185, "lr": 9.914270430204941e-05}
{"train_loss": 0.5917614698410034, "global_step": 16494, "epoch": 185, "lr": 9.914259740398524e-05}
{"train_loss": 0.4895058870315552, "global_step": 16495, "epoch": 185, "lr": 9.914249049931446e-05}
{"train_loss": 0.4936508536338806, "global_step": 16496, "epoch": 185, "lr": 9.914238358803707e-05}
{"train_loss": 0.655062735080719, "global_step": 16497, "epoch": 185, "lr": 9.914227667015306e-05}
{"train_loss": 0.597322940826416, "global_step": 16498, "epoch": 185, "lr": 9.914216974566249e-05}
{"train_loss": 0.6673976182937622, "global_step": 16499, "epoch": 185, "lr": 9.914206281456535e-05}
{"train_loss": 0.6785555481910706, "global_step": 16500, "epoch": 185, "lr": 9.914195587686166e-05}
{"train_loss": 0.5717745423316956, "global_step": 16501, "epoch": 185, "lr": 9.914184893255144e-05}
{"train_loss": 0.4520310163497925, "global_step": 16502, "epoch": 185, "lr": 9.914174198163468e-05}
{"train_loss": 0.40755510330200195, "global_step": 16503, "epoch": 185, "lr": 9.914163502411143e-05}
{"train_loss": 0.6271902918815613, "global_step": 16504, "epoch": 185, "lr": 9.914152805998167e-05}
{"train_loss": 0.40996596217155457, "global_step": 16505, "epoch": 185, "lr": 9.914142108924544e-05}
{"train_loss": 0.5923281311988831, "global_step": 16506, "epoch": 185, "lr": 9.914131411190275e-05}
{"train_loss": 0.562311589717865, "global_step": 16507, "epoch": 185, "lr": 9.91412071279536e-05}
{"train_loss": 0.5030668377876282, "global_step": 16508, "epoch": 185, "lr": 9.914110013739801e-05}
{"train_loss": 0.4618208706378937, "global_step": 16509, "epoch": 185, "lr": 9.914099314023601e-05}
{"train_loss": 0.5180280804634094, "global_step": 16510, "epoch": 185, "lr": 9.914088613646759e-05}
{"train_loss": 0.6775319576263428, "global_step": 16511, "epoch": 185, "lr": 9.914077912609279e-05}
{"train_loss": 0.6695238351821899, "global_step": 16512, "epoch": 185, "lr": 9.91406721091116e-05}
{"train_loss": 0.6270750164985657, "global_step": 16513, "epoch": 185, "lr": 9.914056508552404e-05}
{"train_loss": 0.5429978370666504, "global_step": 16514, "epoch": 185, "lr": 9.914045805533016e-05}
{"train_loss": 0.5588639974594116, "global_step": 16515, "epoch": 185, "lr": 9.914035101852992e-05}
{"train_loss": 0.6069560050964355, "global_step": 16516, "epoch": 185, "lr": 9.914024397512336e-05}
{"train_loss": 0.6557984352111816, "global_step": 16517, "epoch": 185, "lr": 9.914013692511051e-05}
{"train_loss": 0.4734434187412262, "global_step": 16518, "epoch": 185, "lr": 9.914002986849134e-05}
{"train_loss": 0.6084535717964172, "global_step": 16519, "epoch": 185, "lr": 9.91399228052659e-05}
{"train_loss": 0.5636453628540039, "global_step": 16520, "epoch": 185, "lr": 9.91398157354342e-05}
{"train_loss": 0.5642542839050293, "global_step": 16521, "epoch": 185, "lr": 9.913970865899627e-05}
{"train_loss": 0.544622540473938, "global_step": 16522, "epoch": 185, "lr": 9.913960157595209e-05}
{"train_loss": 0.5189927816390991, "global_step": 16523, "epoch": 185, "lr": 9.913949448630168e-05}
{"train_loss": 0.5449737310409546, "global_step": 16524, "epoch": 185, "lr": 9.913938739004506e-05}
{"train_loss": 0.582863986492157, "global_step": 16525, "epoch": 185, "lr": 9.913928028718227e-05}
{"train_loss": 0.6840507984161377, "global_step": 16526, "epoch": 185, "lr": 9.913917317771329e-05}
{"train_loss": 0.7053083181381226, "global_step": 16527, "epoch": 185, "lr": 9.913906606163815e-05}
{"train_loss": 0.6787857413291931, "global_step": 16528, "epoch": 185, "lr": 9.913895893895686e-05}
{"train_loss": 0.5311333537101746, "global_step": 16529, "epoch": 185, "lr": 9.913885180966944e-05}
{"train_loss": 0.6938034892082214, "global_step": 16530, "epoch": 185, "lr": 9.913874467377589e-05}
{"train_loss": 0.6236861944198608, "global_step": 16531, "epoch": 185, "lr": 9.913863753127625e-05}
{"train_loss": 0.4823158085346222, "global_step": 16532, "epoch": 185, "lr": 9.91385303821705e-05}
{"train_loss": 0.6457874178886414, "global_step": 16533, "epoch": 185, "lr": 9.913842322645869e-05}
{"train_loss": 0.43187394738197327, "global_step": 16534, "epoch": 185, "lr": 9.913831606414081e-05}
{"train_loss": 0.4539337158203125, "global_step": 16535, "epoch": 185, "lr": 9.913820889521688e-05}
{"train_loss": 0.47726500034332275, "global_step": 16536, "epoch": 185, "lr": 9.913810171968693e-05}
{"train_loss": 0.6237714886665344, "global_step": 16537, "epoch": 185, "lr": 9.913799453755094e-05}
{"train_loss": 0.5978920459747314, "global_step": 16538, "epoch": 185, "lr": 9.913788734880897e-05}
{"train_loss": 0.6348493099212646, "global_step": 16539, "epoch": 185, "lr": 9.9137780153461e-05}
{"train_loss": 0.5960662961006165, "global_step": 16540, "epoch": 185, "lr": 9.913767295150704e-05}
{"train_loss": 0.651379406452179, "global_step": 16541, "epoch": 185, "lr": 9.913756574294714e-05}
{"train_loss": 0.4945470988750458, "global_step": 16542, "epoch": 185, "lr": 9.913745852778128e-05}
{"train_loss": 0.6062326431274414, "global_step": 16543, "epoch": 185, "lr": 9.91373513060095e-05}
{"train_loss": 0.603421151638031, "global_step": 16544, "epoch": 185, "lr": 9.913724407763179e-05}
{"train_loss": 0.5227228403091431, "global_step": 16545, "epoch": 185, "lr": 9.913713684264818e-05}
{"train_loss": 0.7431159615516663, "global_step": 16546, "epoch": 185, "lr": 9.913702960105869e-05}
{"train_loss": 0.41932934522628784, "global_step": 16547, "epoch": 185, "lr": 9.913692235286331e-05}
{"train_loss": 0.6046128869056702, "global_step": 16548, "epoch": 185, "lr": 9.913681509806208e-05}
{"train_loss": 0.48447149991989136, "global_step": 16549, "epoch": 185, "lr": 9.913670783665498e-05}
{"train_loss": 0.5262684226036072, "global_step": 16550, "epoch": 185, "lr": 9.913660056864208e-05}
{"train_loss": 0.487964004278183, "global_step": 16551, "epoch": 185, "lr": 9.913649329402335e-05}
{"train_loss": 0.475675106048584, "global_step": 16552, "epoch": 185, "lr": 9.913638601279883e-05}
{"train_loss": 0.5567984604433681, "global_step": 16553, "epoch": 185, "lr": 9.91362787249685e-05, "val_loss": 1.5169237852096558, "train_action_mse_error": 42.6696662902832}
{"train_loss": 0.3950270116329193, "global_step": 16554, "epoch": 186, "lr": 9.91361714305324e-05}
{"train_loss": 0.5748347043991089, "global_step": 16555, "epoch": 186, "lr": 9.913606412949054e-05}
{"train_loss": 0.4394482970237732, "global_step": 16556, "epoch": 186, "lr": 9.913595682184295e-05}
{"train_loss": 0.543880045413971, "global_step": 16557, "epoch": 186, "lr": 9.913584950758963e-05}
{"train_loss": 0.4575098156929016, "global_step": 16558, "epoch": 186, "lr": 9.913574218673059e-05}
{"train_loss": 0.5547470450401306, "global_step": 16559, "epoch": 186, "lr": 9.913563485926583e-05}
{"train_loss": 0.6111772060394287, "global_step": 16560, "epoch": 186, "lr": 9.91355275251954e-05}
{"train_loss": 0.5836177468299866, "global_step": 16561, "epoch": 186, "lr": 9.91354201845193e-05}
{"train_loss": 0.6138486862182617, "global_step": 16562, "epoch": 186, "lr": 9.913531283723754e-05}
{"train_loss": 0.48418155312538147, "global_step": 16563, "epoch": 186, "lr": 9.913520548335014e-05}
{"train_loss": 0.4958699941635132, "global_step": 16564, "epoch": 186, "lr": 9.913509812285708e-05}
{"train_loss": 0.6456190347671509, "global_step": 16565, "epoch": 186, "lr": 9.913499075575844e-05}
{"train_loss": 0.42980092763900757, "global_step": 16566, "epoch": 186, "lr": 9.913488338205418e-05}
{"train_loss": 0.5257319808006287, "global_step": 16567, "epoch": 186, "lr": 9.913477600174434e-05}
{"train_loss": 0.5714749693870544, "global_step": 16568, "epoch": 186, "lr": 9.913466861482894e-05}
{"train_loss": 0.5131412148475647, "global_step": 16569, "epoch": 186, "lr": 9.913456122130797e-05}
{"train_loss": 0.43098020553588867, "global_step": 16570, "epoch": 186, "lr": 9.913445382118146e-05}
{"train_loss": 0.6621609926223755, "global_step": 16571, "epoch": 186, "lr": 9.91343464144494e-05}
{"train_loss": 0.5691457986831665, "global_step": 16572, "epoch": 186, "lr": 9.913423900111186e-05}
{"train_loss": 0.5485406517982483, "global_step": 16573, "epoch": 186, "lr": 9.91341315811688e-05}
{"train_loss": 0.5364052653312683, "global_step": 16574, "epoch": 186, "lr": 9.913402415462027e-05}
{"train_loss": 0.5337505340576172, "global_step": 16575, "epoch": 186, "lr": 9.913391672146626e-05}
{"train_loss": 0.5897766947746277, "global_step": 16576, "epoch": 186, "lr": 9.91338092817068e-05}
{"train_loss": 0.5517336130142212, "global_step": 16577, "epoch": 186, "lr": 9.913370183534189e-05}
{"train_loss": 0.5902807116508484, "global_step": 16578, "epoch": 186, "lr": 9.913359438237156e-05}
{"train_loss": 0.5944497585296631, "global_step": 16579, "epoch": 186, "lr": 9.91334869227958e-05}
{"train_loss": 0.4641244411468506, "global_step": 16580, "epoch": 186, "lr": 9.913337945661465e-05}
{"train_loss": 0.42153412103652954, "global_step": 16581, "epoch": 186, "lr": 9.913327198382812e-05}
{"train_loss": 0.4145205318927765, "global_step": 16582, "epoch": 186, "lr": 9.913316450443622e-05}
{"train_loss": 0.6480122208595276, "global_step": 16583, "epoch": 186, "lr": 9.913305701843898e-05}
{"train_loss": 0.6018795371055603, "global_step": 16584, "epoch": 186, "lr": 9.913294952583639e-05}
{"train_loss": 0.5638120174407959, "global_step": 16585, "epoch": 186, "lr": 9.913284202662845e-05}
{"train_loss": 0.42466533184051514, "global_step": 16586, "epoch": 186, "lr": 9.913273452081522e-05}
{"train_loss": 0.6049762964248657, "global_step": 16587, "epoch": 186, "lr": 9.91326270083967e-05}
{"train_loss": 0.6980233788490295, "global_step": 16588, "epoch": 186, "lr": 9.913251948937288e-05}
{"train_loss": 0.5875078439712524, "global_step": 16589, "epoch": 186, "lr": 9.913241196374378e-05}
{"train_loss": 0.5217457413673401, "global_step": 16590, "epoch": 186, "lr": 9.913230443150946e-05}
{"train_loss": 0.484681099653244, "global_step": 16591, "epoch": 186, "lr": 9.913219689266988e-05}
{"train_loss": 0.6227757334709167, "global_step": 16592, "epoch": 186, "lr": 9.913208934722509e-05}
{"train_loss": 0.5487477779388428, "global_step": 16593, "epoch": 186, "lr": 9.913198179517506e-05}
{"train_loss": 0.6455298066139221, "global_step": 16594, "epoch": 186, "lr": 9.913187423651986e-05}
{"train_loss": 0.45435571670532227, "global_step": 16595, "epoch": 186, "lr": 9.913176667125947e-05}
{"train_loss": 0.4734160900115967, "global_step": 16596, "epoch": 186, "lr": 9.913165909939393e-05}
{"train_loss": 0.532538115978241, "global_step": 16597, "epoch": 186, "lr": 9.913155152092322e-05}
{"train_loss": 0.5044457316398621, "global_step": 16598, "epoch": 186, "lr": 9.913144393584736e-05}
{"train_loss": 0.6266614198684692, "global_step": 16599, "epoch": 186, "lr": 9.91313363441664e-05}
{"train_loss": 0.6386963725090027, "global_step": 16600, "epoch": 186, "lr": 9.913122874588032e-05}
{"train_loss": 0.5541043877601624, "global_step": 16601, "epoch": 186, "lr": 9.913112114098915e-05}
{"train_loss": 0.4871186316013336, "global_step": 16602, "epoch": 186, "lr": 9.913101352949287e-05}
{"train_loss": 0.5458984375, "global_step": 16603, "epoch": 186, "lr": 9.913090591139156e-05}
{"train_loss": 0.5591580271720886, "global_step": 16604, "epoch": 186, "lr": 9.913079828668519e-05}
{"train_loss": 0.4845120310783386, "global_step": 16605, "epoch": 186, "lr": 9.913069065537379e-05}
{"train_loss": 0.6169013977050781, "global_step": 16606, "epoch": 186, "lr": 9.913058301745735e-05}
{"train_loss": 0.5946227312088013, "global_step": 16607, "epoch": 186, "lr": 9.91304753729359e-05}
{"train_loss": 0.5504562854766846, "global_step": 16608, "epoch": 186, "lr": 9.913036772180949e-05}
{"train_loss": 0.4649612605571747, "global_step": 16609, "epoch": 186, "lr": 9.913026006407807e-05}
{"train_loss": 0.4748963713645935, "global_step": 16610, "epoch": 186, "lr": 9.913015239974168e-05}
{"train_loss": 0.5099135041236877, "global_step": 16611, "epoch": 186, "lr": 9.913004472880036e-05}
{"train_loss": 0.4812324643135071, "global_step": 16612, "epoch": 186, "lr": 9.912993705125409e-05}
{"train_loss": 0.4636131823062897, "global_step": 16613, "epoch": 186, "lr": 9.912982936710291e-05}
{"train_loss": 0.5891144871711731, "global_step": 16614, "epoch": 186, "lr": 9.912972167634682e-05}
{"train_loss": 0.48342493176460266, "global_step": 16615, "epoch": 186, "lr": 9.912961397898584e-05}
{"train_loss": 0.4230011999607086, "global_step": 16616, "epoch": 186, "lr": 9.912950627501997e-05}
{"train_loss": 0.5420095324516296, "global_step": 16617, "epoch": 186, "lr": 9.912939856444925e-05}
{"train_loss": 0.48473024368286133, "global_step": 16618, "epoch": 186, "lr": 9.912929084727367e-05}
{"train_loss": 0.42664384841918945, "global_step": 16619, "epoch": 186, "lr": 9.912918312349326e-05}
{"train_loss": 0.6868945956230164, "global_step": 16620, "epoch": 186, "lr": 9.912907539310802e-05}
{"train_loss": 0.5675005316734314, "global_step": 16621, "epoch": 186, "lr": 9.9128967656118e-05}
{"train_loss": 0.4158972203731537, "global_step": 16622, "epoch": 186, "lr": 9.912885991252318e-05}
{"train_loss": 0.6152086853981018, "global_step": 16623, "epoch": 186, "lr": 9.912875216232359e-05}
{"train_loss": 0.519943118095398, "global_step": 16624, "epoch": 186, "lr": 9.912864440551921e-05}
{"train_loss": 0.497683048248291, "global_step": 16625, "epoch": 186, "lr": 9.912853664211011e-05}
{"train_loss": 0.4622196555137634, "global_step": 16626, "epoch": 186, "lr": 9.912842887209626e-05}
{"train_loss": 0.607530951499939, "global_step": 16627, "epoch": 186, "lr": 9.91283210954777e-05}
{"train_loss": 0.5759210586547852, "global_step": 16628, "epoch": 186, "lr": 9.912821331225443e-05}
{"train_loss": 0.7020334005355835, "global_step": 16629, "epoch": 186, "lr": 9.912810552242647e-05}
{"train_loss": 0.5568513870239258, "global_step": 16630, "epoch": 186, "lr": 9.912799772599384e-05}
{"train_loss": 0.6244146823883057, "global_step": 16631, "epoch": 186, "lr": 9.912788992295655e-05}
{"train_loss": 0.5315364003181458, "global_step": 16632, "epoch": 186, "lr": 9.912778211331461e-05}
{"train_loss": 0.6270087361335754, "global_step": 16633, "epoch": 186, "lr": 9.912767429706805e-05}
{"train_loss": 0.48375898599624634, "global_step": 16634, "epoch": 186, "lr": 9.912756647421686e-05}
{"train_loss": 0.5204483866691589, "global_step": 16635, "epoch": 186, "lr": 9.912745864476107e-05}
{"train_loss": 0.491436243057251, "global_step": 16636, "epoch": 186, "lr": 9.912735080870071e-05}
{"train_loss": 0.544562041759491, "global_step": 16637, "epoch": 186, "lr": 9.912724296603575e-05}
{"train_loss": 0.57200026512146, "global_step": 16638, "epoch": 186, "lr": 9.912713511676625e-05}
{"train_loss": 0.41148000955581665, "global_step": 16639, "epoch": 186, "lr": 9.91270272608922e-05}
{"train_loss": 0.4930332899093628, "global_step": 16640, "epoch": 186, "lr": 9.912691939841362e-05}
{"train_loss": 0.5469136834144592, "global_step": 16641, "epoch": 186, "lr": 9.912681152933051e-05}
{"train_loss": 0.5383820908792903, "global_step": 16642, "epoch": 186, "lr": 9.912670365364292e-05, "val_loss": 1.6317087411880493}
{"train_loss": 0.49048879742622375, "global_step": 16643, "epoch": 187, "lr": 9.912659577135082e-05}
{"train_loss": 0.5447677969932556, "global_step": 16644, "epoch": 187, "lr": 9.912648788245427e-05}
{"train_loss": 0.5050054788589478, "global_step": 16645, "epoch": 187, "lr": 9.912637998695324e-05}
{"train_loss": 0.5684221982955933, "global_step": 16646, "epoch": 187, "lr": 9.912627208484779e-05}
{"train_loss": 0.6144625544548035, "global_step": 16647, "epoch": 187, "lr": 9.912616417613791e-05}
{"train_loss": 0.751038670539856, "global_step": 16648, "epoch": 187, "lr": 9.912605626082361e-05}
{"train_loss": 0.6490055322647095, "global_step": 16649, "epoch": 187, "lr": 9.912594833890489e-05}
{"train_loss": 0.4953543543815613, "global_step": 16650, "epoch": 187, "lr": 9.912584041038183e-05}
{"train_loss": 0.4971044063568115, "global_step": 16651, "epoch": 187, "lr": 9.912573247525436e-05}
{"train_loss": 0.43038409948349, "global_step": 16652, "epoch": 187, "lr": 9.912562453352254e-05}
{"train_loss": 0.6100549697875977, "global_step": 16653, "epoch": 187, "lr": 9.91255165851864e-05}
{"train_loss": 0.6267295479774475, "global_step": 16654, "epoch": 187, "lr": 9.91254086302459e-05}
{"train_loss": 0.6133771538734436, "global_step": 16655, "epoch": 187, "lr": 9.912530066870112e-05}
{"train_loss": 0.42545071244239807, "global_step": 16656, "epoch": 187, "lr": 9.912519270055203e-05}
{"train_loss": 0.5597221255302429, "global_step": 16657, "epoch": 187, "lr": 9.912508472579865e-05}
{"train_loss": 0.5076887011528015, "global_step": 16658, "epoch": 187, "lr": 9.9124976744441e-05}
{"train_loss": 0.6449404954910278, "global_step": 16659, "epoch": 187, "lr": 9.912486875647912e-05}
{"train_loss": 0.5581603646278381, "global_step": 16660, "epoch": 187, "lr": 9.912476076191297e-05}
{"train_loss": 0.484224408864975, "global_step": 16661, "epoch": 187, "lr": 9.91246527607426e-05}
{"train_loss": 0.4581734538078308, "global_step": 16662, "epoch": 187, "lr": 9.912454475296804e-05}
{"train_loss": 0.5409985184669495, "global_step": 16663, "epoch": 187, "lr": 9.912443673858927e-05}
{"train_loss": 0.5087597370147705, "global_step": 16664, "epoch": 187, "lr": 9.912432871760631e-05}
{"train_loss": 0.4273620545864105, "global_step": 16665, "epoch": 187, "lr": 9.912422069001919e-05}
{"train_loss": 0.43444204330444336, "global_step": 16666, "epoch": 187, "lr": 9.912411265582793e-05}
{"train_loss": 0.6072840094566345, "global_step": 16667, "epoch": 187, "lr": 9.912400461503251e-05}
{"train_loss": 0.5249180197715759, "global_step": 16668, "epoch": 187, "lr": 9.9123896567633e-05}
{"train_loss": 0.5389912724494934, "global_step": 16669, "epoch": 187, "lr": 9.912378851362934e-05}
{"train_loss": 0.4469527304172516, "global_step": 16670, "epoch": 187, "lr": 9.912368045302161e-05}
{"train_loss": 0.48611530661582947, "global_step": 16671, "epoch": 187, "lr": 9.912357238580978e-05}
{"train_loss": 0.4981490969657898, "global_step": 16672, "epoch": 187, "lr": 9.91234643119939e-05}
{"train_loss": 0.5499030947685242, "global_step": 16673, "epoch": 187, "lr": 9.912335623157398e-05}
{"train_loss": 0.48294633626937866, "global_step": 16674, "epoch": 187, "lr": 9.912324814455001e-05}
{"train_loss": 0.6118806600570679, "global_step": 16675, "epoch": 187, "lr": 9.9123140050922e-05}
{"train_loss": 0.5900986790657043, "global_step": 16676, "epoch": 187, "lr": 9.912303195069001e-05}
{"train_loss": 0.4992089569568634, "global_step": 16677, "epoch": 187, "lr": 9.912292384385401e-05}
{"train_loss": 0.5308717489242554, "global_step": 16678, "epoch": 187, "lr": 9.912281573041405e-05}
{"train_loss": 0.5596828460693359, "global_step": 16679, "epoch": 187, "lr": 9.91227076103701e-05}
{"train_loss": 0.5714498162269592, "global_step": 16680, "epoch": 187, "lr": 9.912259948372222e-05}
{"train_loss": 0.573032557964325, "global_step": 16681, "epoch": 187, "lr": 9.91224913504704e-05}
{"train_loss": 0.3398125171661377, "global_step": 16682, "epoch": 187, "lr": 9.912238321061467e-05}
{"train_loss": 0.40148699283599854, "global_step": 16683, "epoch": 187, "lr": 9.912227506415502e-05}
{"train_loss": 0.41205722093582153, "global_step": 16684, "epoch": 187, "lr": 9.912216691109148e-05}
{"train_loss": 0.536044716835022, "global_step": 16685, "epoch": 187, "lr": 9.912205875142407e-05}
{"train_loss": 0.5381521582603455, "global_step": 16686, "epoch": 187, "lr": 9.912195058515279e-05}
{"train_loss": 0.5881481766700745, "global_step": 16687, "epoch": 187, "lr": 9.912184241227767e-05}
{"train_loss": 0.6302926540374756, "global_step": 16688, "epoch": 187, "lr": 9.912173423279871e-05}
{"train_loss": 0.620665967464447, "global_step": 16689, "epoch": 187, "lr": 9.912162604671595e-05}
{"train_loss": 0.6973266005516052, "global_step": 16690, "epoch": 187, "lr": 9.912151785402937e-05}
{"train_loss": 0.49603286385536194, "global_step": 16691, "epoch": 187, "lr": 9.912140965473899e-05}
{"train_loss": 0.49275487661361694, "global_step": 16692, "epoch": 187, "lr": 9.912130144884486e-05}
{"train_loss": 0.581291675567627, "global_step": 16693, "epoch": 187, "lr": 9.912119323634696e-05}
{"train_loss": 0.5154137015342712, "global_step": 16694, "epoch": 187, "lr": 9.912108501724531e-05}
{"train_loss": 0.5801067352294922, "global_step": 16695, "epoch": 187, "lr": 9.912097679153993e-05}
{"train_loss": 0.5158640146255493, "global_step": 16696, "epoch": 187, "lr": 9.912086855923084e-05}
{"train_loss": 0.5975123643875122, "global_step": 16697, "epoch": 187, "lr": 9.912076032031803e-05}
{"train_loss": 0.5188114047050476, "global_step": 16698, "epoch": 187, "lr": 9.912065207480154e-05}
{"train_loss": 0.5086332559585571, "global_step": 16699, "epoch": 187, "lr": 9.912054382268139e-05}
{"train_loss": 0.494034081697464, "global_step": 16700, "epoch": 187, "lr": 9.912043556395757e-05}
{"train_loss": 0.6023886203765869, "global_step": 16701, "epoch": 187, "lr": 9.91203272986301e-05}
{"train_loss": 0.44944947957992554, "global_step": 16702, "epoch": 187, "lr": 9.912021902669901e-05}
{"train_loss": 0.5309696197509766, "global_step": 16703, "epoch": 187, "lr": 9.912011074816431e-05}
{"train_loss": 0.5767738223075867, "global_step": 16704, "epoch": 187, "lr": 9.912000246302601e-05}
{"train_loss": 0.6478518843650818, "global_step": 16705, "epoch": 187, "lr": 9.911989417128412e-05}
{"train_loss": 0.5063682198524475, "global_step": 16706, "epoch": 187, "lr": 9.911978587293866e-05}
{"train_loss": 0.6102999448776245, "global_step": 16707, "epoch": 187, "lr": 9.911967756798963e-05}
{"train_loss": 0.5099976062774658, "global_step": 16708, "epoch": 187, "lr": 9.911956925643708e-05}
{"train_loss": 0.6951366662979126, "global_step": 16709, "epoch": 187, "lr": 9.911946093828099e-05}
{"train_loss": 0.6385625004768372, "global_step": 16710, "epoch": 187, "lr": 9.911935261352139e-05}
{"train_loss": 0.47837120294570923, "global_step": 16711, "epoch": 187, "lr": 9.911924428215829e-05}
{"train_loss": 0.46932005882263184, "global_step": 16712, "epoch": 187, "lr": 9.91191359441917e-05}
{"train_loss": 0.655350923538208, "global_step": 16713, "epoch": 187, "lr": 9.911902759962166e-05}
{"train_loss": 0.5495098829269409, "global_step": 16714, "epoch": 187, "lr": 9.911891924844815e-05}
{"train_loss": 0.6197577714920044, "global_step": 16715, "epoch": 187, "lr": 9.911881089067122e-05}
{"train_loss": 0.665444016456604, "global_step": 16716, "epoch": 187, "lr": 9.911870252629083e-05}
{"train_loss": 0.5789186954498291, "global_step": 16717, "epoch": 187, "lr": 9.911859415530705e-05}
{"train_loss": 0.5080567598342896, "global_step": 16718, "epoch": 187, "lr": 9.911848577771987e-05}
{"train_loss": 0.5487644076347351, "global_step": 16719, "epoch": 187, "lr": 9.911837739352932e-05}
{"train_loss": 0.595669150352478, "global_step": 16720, "epoch": 187, "lr": 9.911826900273539e-05}
{"train_loss": 0.6256648898124695, "global_step": 16721, "epoch": 187, "lr": 9.91181606053381e-05}
{"train_loss": 0.5653306841850281, "global_step": 16722, "epoch": 187, "lr": 9.911805220133749e-05}
{"train_loss": 0.4771610498428345, "global_step": 16723, "epoch": 187, "lr": 9.911794379073353e-05}
{"train_loss": 0.48885464668273926, "global_step": 16724, "epoch": 187, "lr": 9.911783537352629e-05}
{"train_loss": 0.46871817111968994, "global_step": 16725, "epoch": 187, "lr": 9.911772694971575e-05}
{"train_loss": 0.48894137144088745, "global_step": 16726, "epoch": 187, "lr": 9.911761851930193e-05}
{"train_loss": 0.5973504185676575, "global_step": 16727, "epoch": 187, "lr": 9.911751008228482e-05}
{"train_loss": 0.5157483816146851, "global_step": 16728, "epoch": 187, "lr": 9.91174016386645e-05}
{"train_loss": 0.6652856469154358, "global_step": 16729, "epoch": 187, "lr": 9.911729318844091e-05}
{"train_loss": 0.5065492391586304, "global_step": 16730, "epoch": 187, "lr": 9.911718473161411e-05}
{"train_loss": 0.5448161695110664, "global_step": 16731, "epoch": 187, "lr": 9.911707626818412e-05, "val_loss": 1.5843983888626099}
{"train_loss": 0.6024315357208252, "global_step": 16732, "epoch": 188, "lr": 9.911696779815092e-05}
{"train_loss": 0.5748433470726013, "global_step": 16733, "epoch": 188, "lr": 9.911685932151454e-05}
{"train_loss": 0.42034420371055603, "global_step": 16734, "epoch": 188, "lr": 9.911675083827501e-05}
{"train_loss": 0.5749640464782715, "global_step": 16735, "epoch": 188, "lr": 9.911664234843232e-05}
{"train_loss": 0.5508679747581482, "global_step": 16736, "epoch": 188, "lr": 9.91165338519865e-05}
{"train_loss": 0.575851321220398, "global_step": 16737, "epoch": 188, "lr": 9.911642534893755e-05}
{"train_loss": 0.5072596669197083, "global_step": 16738, "epoch": 188, "lr": 9.911631683928551e-05}
{"train_loss": 0.5124324560165405, "global_step": 16739, "epoch": 188, "lr": 9.911620832303038e-05}
{"train_loss": 0.49060115218162537, "global_step": 16740, "epoch": 188, "lr": 9.911609980017217e-05}
{"train_loss": 0.484988272190094, "global_step": 16741, "epoch": 188, "lr": 9.911599127071089e-05}
{"train_loss": 0.5781112909317017, "global_step": 16742, "epoch": 188, "lr": 9.911588273464658e-05}
{"train_loss": 0.6178194284439087, "global_step": 16743, "epoch": 188, "lr": 9.911577419197924e-05}
{"train_loss": 0.46529537439346313, "global_step": 16744, "epoch": 188, "lr": 9.911566564270887e-05}
{"train_loss": 0.5098633170127869, "global_step": 16745, "epoch": 188, "lr": 9.91155570868355e-05}
{"train_loss": 0.5195959806442261, "global_step": 16746, "epoch": 188, "lr": 9.911544852435915e-05}
{"train_loss": 0.6361761689186096, "global_step": 16747, "epoch": 188, "lr": 9.91153399552798e-05}
{"train_loss": 0.4412339925765991, "global_step": 16748, "epoch": 188, "lr": 9.911523137959752e-05}
{"train_loss": 0.4953743815422058, "global_step": 16749, "epoch": 188, "lr": 9.911512279731229e-05}
{"train_loss": 0.4488908648490906, "global_step": 16750, "epoch": 188, "lr": 9.911501420842413e-05}
{"train_loss": 0.5103324055671692, "global_step": 16751, "epoch": 188, "lr": 9.911490561293305e-05}
{"train_loss": 0.5647563934326172, "global_step": 16752, "epoch": 188, "lr": 9.911479701083907e-05}
{"train_loss": 0.41134941577911377, "global_step": 16753, "epoch": 188, "lr": 9.91146884021422e-05}
{"train_loss": 0.5044037699699402, "global_step": 16754, "epoch": 188, "lr": 9.911457978684246e-05}
{"train_loss": 0.5177037119865417, "global_step": 16755, "epoch": 188, "lr": 9.911447116493987e-05}
{"train_loss": 0.5916780829429626, "global_step": 16756, "epoch": 188, "lr": 9.911436253643445e-05}
{"train_loss": 0.6364012956619263, "global_step": 16757, "epoch": 188, "lr": 9.911425390132618e-05}
{"train_loss": 0.5684853792190552, "global_step": 16758, "epoch": 188, "lr": 9.911414525961511e-05}
{"train_loss": 0.607746958732605, "global_step": 16759, "epoch": 188, "lr": 9.911403661130123e-05}
{"train_loss": 0.5437983870506287, "global_step": 16760, "epoch": 188, "lr": 9.911392795638457e-05}
{"train_loss": 0.41626667976379395, "global_step": 16761, "epoch": 188, "lr": 9.911381929486514e-05}
{"train_loss": 0.5912492871284485, "global_step": 16762, "epoch": 188, "lr": 9.911371062674297e-05}
{"train_loss": 0.5050262808799744, "global_step": 16763, "epoch": 188, "lr": 9.911360195201804e-05}
{"train_loss": 0.474336177110672, "global_step": 16764, "epoch": 188, "lr": 9.91134932706904e-05}
{"train_loss": 0.6084172129631042, "global_step": 16765, "epoch": 188, "lr": 9.911338458276004e-05}
{"train_loss": 0.5917602777481079, "global_step": 16766, "epoch": 188, "lr": 9.9113275888227e-05}
{"train_loss": 0.5060541033744812, "global_step": 16767, "epoch": 188, "lr": 9.911316718709128e-05}
{"train_loss": 0.5371991991996765, "global_step": 16768, "epoch": 188, "lr": 9.911305847935285e-05}
{"train_loss": 0.5089276432991028, "global_step": 16769, "epoch": 188, "lr": 9.911294976501182e-05}
{"train_loss": 0.5372123122215271, "global_step": 16770, "epoch": 188, "lr": 9.911284104406813e-05}
{"train_loss": 0.48401498794555664, "global_step": 16771, "epoch": 188, "lr": 9.91127323165218e-05}
{"train_loss": 0.5789461731910706, "global_step": 16772, "epoch": 188, "lr": 9.91126235823729e-05}
{"train_loss": 0.422028124332428, "global_step": 16773, "epoch": 188, "lr": 9.911251484162138e-05}
{"train_loss": 0.6112425327301025, "global_step": 16774, "epoch": 188, "lr": 9.911240609426729e-05}
{"train_loss": 0.5416345000267029, "global_step": 16775, "epoch": 188, "lr": 9.911229734031063e-05}
{"train_loss": 0.6721004843711853, "global_step": 16776, "epoch": 188, "lr": 9.911218857975142e-05}
{"train_loss": 0.5972038507461548, "global_step": 16777, "epoch": 188, "lr": 9.911207981258968e-05}
{"train_loss": 0.5602062940597534, "global_step": 16778, "epoch": 188, "lr": 9.911197103882542e-05}
{"train_loss": 0.49703550338745117, "global_step": 16779, "epoch": 188, "lr": 9.911186225845865e-05}
{"train_loss": 0.6690802574157715, "global_step": 16780, "epoch": 188, "lr": 9.91117534714894e-05}
{"train_loss": 0.5074166059494019, "global_step": 16781, "epoch": 188, "lr": 9.911164467791765e-05}
{"train_loss": 0.4962255656719208, "global_step": 16782, "epoch": 188, "lr": 9.911153587774346e-05}
{"train_loss": 0.5805147290229797, "global_step": 16783, "epoch": 188, "lr": 9.911142707096681e-05}
{"train_loss": 0.6776887774467468, "global_step": 16784, "epoch": 188, "lr": 9.911131825758773e-05}
{"train_loss": 0.4373248219490051, "global_step": 16785, "epoch": 188, "lr": 9.911120943760624e-05}
{"train_loss": 0.5400671362876892, "global_step": 16786, "epoch": 188, "lr": 9.911110061102233e-05}
{"train_loss": 0.4999884068965912, "global_step": 16787, "epoch": 188, "lr": 9.911099177783605e-05}
{"train_loss": 0.4926435053348541, "global_step": 16788, "epoch": 188, "lr": 9.911088293804737e-05}
{"train_loss": 0.5537300705909729, "global_step": 16789, "epoch": 188, "lr": 9.911077409165637e-05}
{"train_loss": 0.6415360569953918, "global_step": 16790, "epoch": 188, "lr": 9.911066523866299e-05}
{"train_loss": 0.553990364074707, "global_step": 16791, "epoch": 188, "lr": 9.91105563790673e-05}
{"train_loss": 0.6086640954017639, "global_step": 16792, "epoch": 188, "lr": 9.911044751286928e-05}
{"train_loss": 0.41163337230682373, "global_step": 16793, "epoch": 188, "lr": 9.911033864006895e-05}
{"train_loss": 0.5855939388275146, "global_step": 16794, "epoch": 188, "lr": 9.911022976066636e-05}
{"train_loss": 0.6353757977485657, "global_step": 16795, "epoch": 188, "lr": 9.91101208746615e-05}
{"train_loss": 0.48965731263160706, "global_step": 16796, "epoch": 188, "lr": 9.911001198205436e-05}
{"train_loss": 0.4589993953704834, "global_step": 16797, "epoch": 188, "lr": 9.910990308284499e-05}
{"train_loss": 0.5335888862609863, "global_step": 16798, "epoch": 188, "lr": 9.910979417703338e-05}
{"train_loss": 0.49419766664505005, "global_step": 16799, "epoch": 188, "lr": 9.910968526461957e-05}
{"train_loss": 0.5556492805480957, "global_step": 16800, "epoch": 188, "lr": 9.910957634560356e-05}
{"train_loss": 0.4600520730018616, "global_step": 16801, "epoch": 188, "lr": 9.910946741998536e-05}
{"train_loss": 0.4624297022819519, "global_step": 16802, "epoch": 188, "lr": 9.910935848776499e-05}
{"train_loss": 0.48806631565093994, "global_step": 16803, "epoch": 188, "lr": 9.910924954894247e-05}
{"train_loss": 0.6129182577133179, "global_step": 16804, "epoch": 188, "lr": 9.910914060351782e-05}
{"train_loss": 0.5533280372619629, "global_step": 16805, "epoch": 188, "lr": 9.910903165149101e-05}
{"train_loss": 0.4232735335826874, "global_step": 16806, "epoch": 188, "lr": 9.910892269286211e-05}
{"train_loss": 0.5331794023513794, "global_step": 16807, "epoch": 188, "lr": 9.910881372763114e-05}
{"train_loss": 0.614711344242096, "global_step": 16808, "epoch": 188, "lr": 9.910870475579806e-05}
{"train_loss": 0.5293853282928467, "global_step": 16809, "epoch": 188, "lr": 9.910859577736291e-05}
{"train_loss": 0.682902991771698, "global_step": 16810, "epoch": 188, "lr": 9.910848679232571e-05}
{"train_loss": 0.5278995037078857, "global_step": 16811, "epoch": 188, "lr": 9.910837780068649e-05}
{"train_loss": 0.5717877149581909, "global_step": 16812, "epoch": 188, "lr": 9.910826880244524e-05}
{"train_loss": 0.5060905814170837, "global_step": 16813, "epoch": 188, "lr": 9.910815979760198e-05}
{"train_loss": 0.5826398730278015, "global_step": 16814, "epoch": 188, "lr": 9.910805078615672e-05}
{"train_loss": 0.49912935495376587, "global_step": 16815, "epoch": 188, "lr": 9.910794176810949e-05}
{"train_loss": 0.5367861986160278, "global_step": 16816, "epoch": 188, "lr": 9.910783274346029e-05}
{"train_loss": 0.6545011401176453, "global_step": 16817, "epoch": 188, "lr": 9.910772371220914e-05}
{"train_loss": 0.4350687563419342, "global_step": 16818, "epoch": 188, "lr": 9.910761467435605e-05}
{"train_loss": 0.6335115432739258, "global_step": 16819, "epoch": 188, "lr": 9.910750562990105e-05}
{"train_loss": 0.5395110982857393, "global_step": 16820, "epoch": 188, "lr": 9.910739657884413e-05, "val_loss": 1.558735966682434}
{"train_loss": 0.5101407170295715, "global_step": 16821, "epoch": 189, "lr": 9.910728752118534e-05}
{"train_loss": 0.4736783802509308, "global_step": 16822, "epoch": 189, "lr": 9.910717845692465e-05}
{"train_loss": 0.5754631161689758, "global_step": 16823, "epoch": 189, "lr": 9.910706938606212e-05}
{"train_loss": 0.5132720470428467, "global_step": 16824, "epoch": 189, "lr": 9.910696030859773e-05}
{"train_loss": 0.5113887786865234, "global_step": 16825, "epoch": 189, "lr": 9.910685122453152e-05}
{"train_loss": 0.5760178565979004, "global_step": 16826, "epoch": 189, "lr": 9.910674213386349e-05}
{"train_loss": 0.577177882194519, "global_step": 16827, "epoch": 189, "lr": 9.910663303659364e-05}
{"train_loss": 0.5681195259094238, "global_step": 16828, "epoch": 189, "lr": 9.910652393272202e-05}
{"train_loss": 0.5935956239700317, "global_step": 16829, "epoch": 189, "lr": 9.910641482224862e-05}
{"train_loss": 0.4927976429462433, "global_step": 16830, "epoch": 189, "lr": 9.910630570517347e-05}
{"train_loss": 0.5554372072219849, "global_step": 16831, "epoch": 189, "lr": 9.910619658149656e-05}
{"train_loss": 0.6533198356628418, "global_step": 16832, "epoch": 189, "lr": 9.910608745121792e-05}
{"train_loss": 0.516389787197113, "global_step": 16833, "epoch": 189, "lr": 9.910597831433758e-05}
{"train_loss": 0.4677835702896118, "global_step": 16834, "epoch": 189, "lr": 9.910586917085554e-05}
{"train_loss": 0.4750964939594269, "global_step": 16835, "epoch": 189, "lr": 9.91057600207718e-05}
{"train_loss": 0.4855009913444519, "global_step": 16836, "epoch": 189, "lr": 9.910565086408639e-05}
{"train_loss": 0.5304441452026367, "global_step": 16837, "epoch": 189, "lr": 9.910554170079934e-05}
{"train_loss": 0.642109215259552, "global_step": 16838, "epoch": 189, "lr": 9.910543253091062e-05}
{"train_loss": 0.7040805816650391, "global_step": 16839, "epoch": 189, "lr": 9.91053233544203e-05}
{"train_loss": 0.44150471687316895, "global_step": 16840, "epoch": 189, "lr": 9.910521417132837e-05}
{"train_loss": 0.6440367102622986, "global_step": 16841, "epoch": 189, "lr": 9.910510498163482e-05}
{"train_loss": 0.48906564712524414, "global_step": 16842, "epoch": 189, "lr": 9.910499578533969e-05}
{"train_loss": 0.4853999614715576, "global_step": 16843, "epoch": 189, "lr": 9.910488658244302e-05}
{"train_loss": 0.6052201986312866, "global_step": 16844, "epoch": 189, "lr": 9.910477737294476e-05}
{"train_loss": 0.6021333336830139, "global_step": 16845, "epoch": 189, "lr": 9.910466815684499e-05}
{"train_loss": 0.6415135860443115, "global_step": 16846, "epoch": 189, "lr": 9.910455893414367e-05}
{"train_loss": 0.5075647234916687, "global_step": 16847, "epoch": 189, "lr": 9.910444970484086e-05}
{"train_loss": 0.5166981816291809, "global_step": 16848, "epoch": 189, "lr": 9.910434046893655e-05}
{"train_loss": 0.6337443590164185, "global_step": 16849, "epoch": 189, "lr": 9.910423122643075e-05}
{"train_loss": 0.4874121844768524, "global_step": 16850, "epoch": 189, "lr": 9.910412197732351e-05}
{"train_loss": 0.5981345772743225, "global_step": 16851, "epoch": 189, "lr": 9.910401272161481e-05}
{"train_loss": 0.5602086186408997, "global_step": 16852, "epoch": 189, "lr": 9.910390345930465e-05}
{"train_loss": 0.49648210406303406, "global_step": 16853, "epoch": 189, "lr": 9.910379419039309e-05}
{"train_loss": 0.6588970422744751, "global_step": 16854, "epoch": 189, "lr": 9.910368491488012e-05}
{"train_loss": 0.5603528022766113, "global_step": 16855, "epoch": 189, "lr": 9.910357563276575e-05}
{"train_loss": 0.6493147015571594, "global_step": 16856, "epoch": 189, "lr": 9.910346634405001e-05}
{"train_loss": 0.6212872266769409, "global_step": 16857, "epoch": 189, "lr": 9.910335704873291e-05}
{"train_loss": 0.6579044461250305, "global_step": 16858, "epoch": 189, "lr": 9.910324774681447e-05}
{"train_loss": 0.6221659779548645, "global_step": 16859, "epoch": 189, "lr": 9.910313843829469e-05}
{"train_loss": 0.5711535811424255, "global_step": 16860, "epoch": 189, "lr": 9.910302912317358e-05}
{"train_loss": 0.4449135959148407, "global_step": 16861, "epoch": 189, "lr": 9.910291980145116e-05}
{"train_loss": 0.4524213969707489, "global_step": 16862, "epoch": 189, "lr": 9.910281047312748e-05}
{"train_loss": 0.6260696053504944, "global_step": 16863, "epoch": 189, "lr": 9.91027011382025e-05}
{"train_loss": 0.5257117748260498, "global_step": 16864, "epoch": 189, "lr": 9.910259179667628e-05}
{"train_loss": 0.48656371235847473, "global_step": 16865, "epoch": 189, "lr": 9.91024824485488e-05}
{"train_loss": 0.6270322203636169, "global_step": 16866, "epoch": 189, "lr": 9.91023730938201e-05}
{"train_loss": 0.49413323402404785, "global_step": 16867, "epoch": 189, "lr": 9.910226373249018e-05}
{"train_loss": 0.49654436111450195, "global_step": 16868, "epoch": 189, "lr": 9.910215436455905e-05}
{"train_loss": 0.43342819809913635, "global_step": 16869, "epoch": 189, "lr": 9.910204499002676e-05}
{"train_loss": 0.525015115737915, "global_step": 16870, "epoch": 189, "lr": 9.910193560889327e-05}
{"train_loss": 0.5465594530105591, "global_step": 16871, "epoch": 189, "lr": 9.910182622115864e-05}
{"train_loss": 0.6561093330383301, "global_step": 16872, "epoch": 189, "lr": 9.910171682682286e-05}
{"train_loss": 0.450127512216568, "global_step": 16873, "epoch": 189, "lr": 9.910160742588596e-05}
{"train_loss": 0.5001787543296814, "global_step": 16874, "epoch": 189, "lr": 9.910149801834794e-05}
{"train_loss": 0.5341189503669739, "global_step": 16875, "epoch": 189, "lr": 9.910138860420883e-05}
{"train_loss": 0.5340998768806458, "global_step": 16876, "epoch": 189, "lr": 9.910127918346863e-05}
{"train_loss": 0.6429688930511475, "global_step": 16877, "epoch": 189, "lr": 9.910116975612735e-05}
{"train_loss": 0.4902566969394684, "global_step": 16878, "epoch": 189, "lr": 9.910106032218504e-05}
{"train_loss": 0.6994878649711609, "global_step": 16879, "epoch": 189, "lr": 9.910095088164168e-05}
{"train_loss": 0.6088975667953491, "global_step": 16880, "epoch": 189, "lr": 9.91008414344973e-05}
{"train_loss": 0.6318647861480713, "global_step": 16881, "epoch": 189, "lr": 9.910073198075189e-05}
{"train_loss": 0.5970086455345154, "global_step": 16882, "epoch": 189, "lr": 9.91006225204055e-05}
{"train_loss": 0.5632423758506775, "global_step": 16883, "epoch": 189, "lr": 9.910051305345814e-05}
{"train_loss": 0.5853381752967834, "global_step": 16884, "epoch": 189, "lr": 9.91004035799098e-05}
{"train_loss": 0.6311600804328918, "global_step": 16885, "epoch": 189, "lr": 9.910029409976052e-05}
{"train_loss": 0.5572584867477417, "global_step": 16886, "epoch": 189, "lr": 9.91001846130103e-05}
{"train_loss": 0.5355799794197083, "global_step": 16887, "epoch": 189, "lr": 9.910007511965915e-05}
{"train_loss": 0.6083815693855286, "global_step": 16888, "epoch": 189, "lr": 9.909996561970711e-05}
{"train_loss": 0.670117199420929, "global_step": 16889, "epoch": 189, "lr": 9.909985611315417e-05}
{"train_loss": 0.4937458336353302, "global_step": 16890, "epoch": 189, "lr": 9.909974660000034e-05}
{"train_loss": 0.5911816358566284, "global_step": 16891, "epoch": 189, "lr": 9.909963708024566e-05}
{"train_loss": 0.39570561051368713, "global_step": 16892, "epoch": 189, "lr": 9.909952755389013e-05}
{"train_loss": 0.6809604167938232, "global_step": 16893, "epoch": 189, "lr": 9.909941802093378e-05}
{"train_loss": 0.624130368232727, "global_step": 16894, "epoch": 189, "lr": 9.90993084813766e-05}
{"train_loss": 0.5895678400993347, "global_step": 16895, "epoch": 189, "lr": 9.909919893521862e-05}
{"train_loss": 0.5535709857940674, "global_step": 16896, "epoch": 189, "lr": 9.909908938245984e-05}
{"train_loss": 0.5867677927017212, "global_step": 16897, "epoch": 189, "lr": 9.909897982310029e-05}
{"train_loss": 0.5449984073638916, "global_step": 16898, "epoch": 189, "lr": 9.909887025713998e-05}
{"train_loss": 0.4225781261920929, "global_step": 16899, "epoch": 189, "lr": 9.909876068457893e-05}
{"train_loss": 0.5502290725708008, "global_step": 16900, "epoch": 189, "lr": 9.909865110541716e-05}
{"train_loss": 0.7598778009414673, "global_step": 16901, "epoch": 189, "lr": 9.909854151965466e-05}
{"train_loss": 0.5438322424888611, "global_step": 16902, "epoch": 189, "lr": 9.909843192729146e-05}
{"train_loss": 0.5574772953987122, "global_step": 16903, "epoch": 189, "lr": 9.909832232832758e-05}
{"train_loss": 0.48506349325180054, "global_step": 16904, "epoch": 189, "lr": 9.909821272276303e-05}
{"train_loss": 0.6700698137283325, "global_step": 16905, "epoch": 189, "lr": 9.909810311059781e-05}
{"train_loss": 0.4958166480064392, "global_step": 16906, "epoch": 189, "lr": 9.909799349183196e-05}
{"train_loss": 0.57874995470047, "global_step": 16907, "epoch": 189, "lr": 9.909788386646548e-05}
{"train_loss": 0.5086167454719543, "global_step": 16908, "epoch": 189, "lr": 9.90977742344984e-05}
{"train_loss": 0.5577600966678576, "global_step": 16909, "epoch": 189, "lr": 9.90976645959307e-05, "val_loss": 1.5196888446807861}
{"train_loss": 0.5655657649040222, "global_step": 16910, "epoch": 190, "lr": 9.909755495076244e-05}
{"train_loss": 0.5620750188827515, "global_step": 16911, "epoch": 190, "lr": 9.90974452989936e-05}
{"train_loss": 0.5932239294052124, "global_step": 16912, "epoch": 190, "lr": 9.909733564062421e-05}
{"train_loss": 0.5192148685455322, "global_step": 16913, "epoch": 190, "lr": 9.909722597565428e-05}
{"train_loss": 0.4202146530151367, "global_step": 16914, "epoch": 190, "lr": 9.909711630408383e-05}
{"train_loss": 0.555416464805603, "global_step": 16915, "epoch": 190, "lr": 9.909700662591286e-05}
{"train_loss": 0.4887184202671051, "global_step": 16916, "epoch": 190, "lr": 9.909689694114141e-05}
{"train_loss": 0.5450004935264587, "global_step": 16917, "epoch": 190, "lr": 9.909678724976947e-05}
{"train_loss": 0.6973153352737427, "global_step": 16918, "epoch": 190, "lr": 9.909667755179709e-05}
{"train_loss": 0.5930948853492737, "global_step": 16919, "epoch": 190, "lr": 9.909656784722423e-05}
{"train_loss": 0.6008686423301697, "global_step": 16920, "epoch": 190, "lr": 9.909645813605093e-05}
{"train_loss": 0.4166093170642853, "global_step": 16921, "epoch": 190, "lr": 9.909634841827724e-05}
{"train_loss": 0.6626102924346924, "global_step": 16922, "epoch": 190, "lr": 9.909623869390312e-05}
{"train_loss": 0.6633884310722351, "global_step": 16923, "epoch": 190, "lr": 9.909612896292861e-05}
{"train_loss": 0.4546176493167877, "global_step": 16924, "epoch": 190, "lr": 9.909601922535375e-05}
{"train_loss": 0.5801955461502075, "global_step": 16925, "epoch": 190, "lr": 9.90959094811785e-05}
{"train_loss": 0.6372760534286499, "global_step": 16926, "epoch": 190, "lr": 9.909579973040291e-05}
{"train_loss": 0.5242863893508911, "global_step": 16927, "epoch": 190, "lr": 9.909568997302701e-05}
{"train_loss": 0.5988994240760803, "global_step": 16928, "epoch": 190, "lr": 9.909558020905077e-05}
{"train_loss": 0.5724053382873535, "global_step": 16929, "epoch": 190, "lr": 9.909547043847423e-05}
{"train_loss": 0.3891269266605377, "global_step": 16930, "epoch": 190, "lr": 9.90953606612974e-05}
{"train_loss": 0.5765759348869324, "global_step": 16931, "epoch": 190, "lr": 9.90952508775203e-05}
{"train_loss": 0.44334477186203003, "global_step": 16932, "epoch": 190, "lr": 9.909514108714294e-05}
{"train_loss": 0.5172224044799805, "global_step": 16933, "epoch": 190, "lr": 9.909503129016535e-05}
{"train_loss": 0.49784815311431885, "global_step": 16934, "epoch": 190, "lr": 9.909492148658753e-05}
{"train_loss": 0.5006381869316101, "global_step": 16935, "epoch": 190, "lr": 9.909481167640949e-05}
{"train_loss": 0.5814192295074463, "global_step": 16936, "epoch": 190, "lr": 9.909470185963123e-05}
{"train_loss": 0.5244380235671997, "global_step": 16937, "epoch": 190, "lr": 9.909459203625282e-05}
{"train_loss": 0.5268813967704773, "global_step": 16938, "epoch": 190, "lr": 9.909448220627422e-05}
{"train_loss": 0.6046419143676758, "global_step": 16939, "epoch": 190, "lr": 9.909437236969547e-05}
{"train_loss": 0.6058939099311829, "global_step": 16940, "epoch": 190, "lr": 9.909426252651658e-05}
{"train_loss": 0.6320809721946716, "global_step": 16941, "epoch": 190, "lr": 9.909415267673756e-05}
{"train_loss": 0.37927210330963135, "global_step": 16942, "epoch": 190, "lr": 9.909404282035843e-05}
{"train_loss": 0.5440018177032471, "global_step": 16943, "epoch": 190, "lr": 9.909393295737922e-05}
{"train_loss": 0.6507548093795776, "global_step": 16944, "epoch": 190, "lr": 9.909382308779992e-05}
{"train_loss": 0.4958849549293518, "global_step": 16945, "epoch": 190, "lr": 9.909371321162055e-05}
{"train_loss": 0.7690656781196594, "global_step": 16946, "epoch": 190, "lr": 9.909360332884112e-05}
{"train_loss": 0.48323896527290344, "global_step": 16947, "epoch": 190, "lr": 9.909349343946168e-05}
{"train_loss": 0.5237332582473755, "global_step": 16948, "epoch": 190, "lr": 9.909338354348219e-05}
{"train_loss": 0.5345124006271362, "global_step": 16949, "epoch": 190, "lr": 9.909327364090271e-05}
{"train_loss": 0.44419294595718384, "global_step": 16950, "epoch": 190, "lr": 9.909316373172324e-05}
{"train_loss": 0.5572865605354309, "global_step": 16951, "epoch": 190, "lr": 9.909305381594377e-05}
{"train_loss": 0.5050484538078308, "global_step": 16952, "epoch": 190, "lr": 9.909294389356436e-05}
{"train_loss": 0.526025116443634, "global_step": 16953, "epoch": 190, "lr": 9.909283396458499e-05}
{"train_loss": 0.46507033705711365, "global_step": 16954, "epoch": 190, "lr": 9.90927240290057e-05}
{"train_loss": 0.47609469294548035, "global_step": 16955, "epoch": 190, "lr": 9.909261408682648e-05}
{"train_loss": 0.47699248790740967, "global_step": 16956, "epoch": 190, "lr": 9.909250413804735e-05}
{"train_loss": 0.5605495572090149, "global_step": 16957, "epoch": 190, "lr": 9.909239418266836e-05}
{"train_loss": 0.5156600475311279, "global_step": 16958, "epoch": 190, "lr": 9.909228422068948e-05}
{"train_loss": 0.5057798624038696, "global_step": 16959, "epoch": 190, "lr": 9.909217425211073e-05}
{"train_loss": 0.4409230053424835, "global_step": 16960, "epoch": 190, "lr": 9.909206427693214e-05}
{"train_loss": 0.49246734380722046, "global_step": 16961, "epoch": 190, "lr": 9.909195429515373e-05}
{"train_loss": 0.3825903534889221, "global_step": 16962, "epoch": 190, "lr": 9.90918443067755e-05}
{"train_loss": 0.5553119778633118, "global_step": 16963, "epoch": 190, "lr": 9.909173431179746e-05}
{"train_loss": 0.48687729239463806, "global_step": 16964, "epoch": 190, "lr": 9.909162431021964e-05}
{"train_loss": 0.5266129970550537, "global_step": 16965, "epoch": 190, "lr": 9.909151430204207e-05}
{"train_loss": 0.5781050324440002, "global_step": 16966, "epoch": 190, "lr": 9.909140428726471e-05}
{"train_loss": 0.6134617328643799, "global_step": 16967, "epoch": 190, "lr": 9.909129426588763e-05}
{"train_loss": 0.591723620891571, "global_step": 16968, "epoch": 190, "lr": 9.909118423791082e-05}
{"train_loss": 0.49511808156967163, "global_step": 16969, "epoch": 190, "lr": 9.90910742033343e-05}
{"train_loss": 0.6058859825134277, "global_step": 16970, "epoch": 190, "lr": 9.909096416215809e-05}
{"train_loss": 0.6659473776817322, "global_step": 16971, "epoch": 190, "lr": 9.909085411438218e-05}
{"train_loss": 0.47449231147766113, "global_step": 16972, "epoch": 190, "lr": 9.909074406000662e-05}
{"train_loss": 0.5601263642311096, "global_step": 16973, "epoch": 190, "lr": 9.90906339990314e-05}
{"train_loss": 0.5887064337730408, "global_step": 16974, "epoch": 190, "lr": 9.909052393145653e-05}
{"train_loss": 0.4498492479324341, "global_step": 16975, "epoch": 190, "lr": 9.909041385728206e-05}
{"train_loss": 0.5796857476234436, "global_step": 16976, "epoch": 190, "lr": 9.909030377650796e-05}
{"train_loss": 0.5285567045211792, "global_step": 16977, "epoch": 190, "lr": 9.909019368913429e-05}
{"train_loss": 0.4994281828403473, "global_step": 16978, "epoch": 190, "lr": 9.909008359516103e-05}
{"train_loss": 0.5737244486808777, "global_step": 16979, "epoch": 190, "lr": 9.908997349458822e-05}
{"train_loss": 0.5593197345733643, "global_step": 16980, "epoch": 190, "lr": 9.908986338741584e-05}
{"train_loss": 0.5292761325836182, "global_step": 16981, "epoch": 190, "lr": 9.908975327364394e-05}
{"train_loss": 0.4998491108417511, "global_step": 16982, "epoch": 190, "lr": 9.908964315327251e-05}
{"train_loss": 0.5487451553344727, "global_step": 16983, "epoch": 190, "lr": 9.908953302630159e-05}
{"train_loss": 0.503497302532196, "global_step": 16984, "epoch": 190, "lr": 9.908942289273117e-05}
{"train_loss": 0.555405855178833, "global_step": 16985, "epoch": 190, "lr": 9.908931275256128e-05}
{"train_loss": 0.46386295557022095, "global_step": 16986, "epoch": 190, "lr": 9.908920260579192e-05}
{"train_loss": 0.4093755781650543, "global_step": 16987, "epoch": 190, "lr": 9.908909245242313e-05}
{"train_loss": 0.5921075344085693, "global_step": 16988, "epoch": 190, "lr": 9.90889822924549e-05}
{"train_loss": 0.598917543888092, "global_step": 16989, "epoch": 190, "lr": 9.908887212588727e-05}
{"train_loss": 0.5415130853652954, "global_step": 16990, "epoch": 190, "lr": 9.908876195272021e-05}
{"train_loss": 0.5145018100738525, "global_step": 16991, "epoch": 190, "lr": 9.908865177295379e-05}
{"train_loss": 0.6217142939567566, "global_step": 16992, "epoch": 190, "lr": 9.908854158658799e-05}
{"train_loss": 0.5248691439628601, "global_step": 16993, "epoch": 190, "lr": 9.908843139362283e-05}
{"train_loss": 0.537360668182373, "global_step": 16994, "epoch": 190, "lr": 9.908832119405834e-05}
{"train_loss": 0.4236496090888977, "global_step": 16995, "epoch": 190, "lr": 9.908821098789451e-05}
{"train_loss": 0.48388487100601196, "global_step": 16996, "epoch": 190, "lr": 9.908810077513137e-05}
{"train_loss": 0.5982055068016052, "global_step": 16997, "epoch": 190, "lr": 9.908799055576896e-05}
{"train_loss": 0.5368571937753913, "global_step": 16998, "epoch": 190, "lr": 9.908788032980724e-05, "val_loss": 1.5707141160964966, "train_action_mse_error": 41.730857849121094}
{"train_loss": 0.48448213934898376, "global_step": 16999, "epoch": 191, "lr": 9.908777009724626e-05}
{"train_loss": 0.5682946443557739, "global_step": 17000, "epoch": 191, "lr": 9.908765985808603e-05}
{"train_loss": 0.5144713521003723, "global_step": 17001, "epoch": 191, "lr": 9.908754961232655e-05}
{"train_loss": 0.4359704852104187, "global_step": 17002, "epoch": 191, "lr": 9.908743935996786e-05}
{"train_loss": 0.44056904315948486, "global_step": 17003, "epoch": 191, "lr": 9.908732910100995e-05}
{"train_loss": 0.45871633291244507, "global_step": 17004, "epoch": 191, "lr": 9.908721883545286e-05}
{"train_loss": 0.4529013931751251, "global_step": 17005, "epoch": 191, "lr": 9.90871085632966e-05}
{"train_loss": 0.553204357624054, "global_step": 17006, "epoch": 191, "lr": 9.908699828454115e-05}
{"train_loss": 0.5432571172714233, "global_step": 17007, "epoch": 191, "lr": 9.908688799918657e-05}
{"train_loss": 0.4734838604927063, "global_step": 17008, "epoch": 191, "lr": 9.908677770723285e-05}
{"train_loss": 0.5670195817947388, "global_step": 17009, "epoch": 191, "lr": 9.908666740868002e-05}
{"train_loss": 0.5462068319320679, "global_step": 17010, "epoch": 191, "lr": 9.908655710352808e-05}
{"train_loss": 0.4249286651611328, "global_step": 17011, "epoch": 191, "lr": 9.908644679177704e-05}
{"train_loss": 0.5273430943489075, "global_step": 17012, "epoch": 191, "lr": 9.908633647342693e-05}
{"train_loss": 0.5836544036865234, "global_step": 17013, "epoch": 191, "lr": 9.908622614847777e-05}
{"train_loss": 0.5453876852989197, "global_step": 17014, "epoch": 191, "lr": 9.908611581692957e-05}
{"train_loss": 0.6310462951660156, "global_step": 17015, "epoch": 191, "lr": 9.908600547878232e-05}
{"train_loss": 0.4586087465286255, "global_step": 17016, "epoch": 191, "lr": 9.908589513403607e-05}
{"train_loss": 0.4522210359573364, "global_step": 17017, "epoch": 191, "lr": 9.908578478269082e-05}
{"train_loss": 0.5546828508377075, "global_step": 17018, "epoch": 191, "lr": 9.908567442474656e-05}
{"train_loss": 0.5407291054725647, "global_step": 17019, "epoch": 191, "lr": 9.908556406020335e-05}
{"train_loss": 0.48382148146629333, "global_step": 17020, "epoch": 191, "lr": 9.908545368906119e-05}
{"train_loss": 0.5636516809463501, "global_step": 17021, "epoch": 191, "lr": 9.908534331132009e-05}
{"train_loss": 0.6561555862426758, "global_step": 17022, "epoch": 191, "lr": 9.908523292698005e-05}
{"train_loss": 0.6047514081001282, "global_step": 17023, "epoch": 191, "lr": 9.908512253604109e-05}
{"train_loss": 0.4709392189979553, "global_step": 17024, "epoch": 191, "lr": 9.908501213850325e-05}
{"train_loss": 0.559269905090332, "global_step": 17025, "epoch": 191, "lr": 9.908490173436653e-05}
{"train_loss": 0.5334560871124268, "global_step": 17026, "epoch": 191, "lr": 9.908479132363095e-05}
{"train_loss": 0.5912832021713257, "global_step": 17027, "epoch": 191, "lr": 9.90846809062965e-05}
{"train_loss": 0.6107959151268005, "global_step": 17028, "epoch": 191, "lr": 9.908457048236323e-05}
{"train_loss": 0.7309102416038513, "global_step": 17029, "epoch": 191, "lr": 9.908446005183112e-05}
{"train_loss": 0.40539461374282837, "global_step": 17030, "epoch": 191, "lr": 9.908434961470022e-05}
{"train_loss": 0.4346839487552643, "global_step": 17031, "epoch": 191, "lr": 9.908423917097052e-05}
{"train_loss": 0.6232505440711975, "global_step": 17032, "epoch": 191, "lr": 9.908412872064203e-05}
{"train_loss": 0.5411867499351501, "global_step": 17033, "epoch": 191, "lr": 9.908401826371479e-05}
{"train_loss": 0.5615555047988892, "global_step": 17034, "epoch": 191, "lr": 9.90839078001888e-05}
{"train_loss": 0.4962153434753418, "global_step": 17035, "epoch": 191, "lr": 9.908379733006407e-05}
{"train_loss": 0.7750495076179504, "global_step": 17036, "epoch": 191, "lr": 9.908368685334064e-05}
{"train_loss": 0.4905420243740082, "global_step": 17037, "epoch": 191, "lr": 9.908357637001848e-05}
{"train_loss": 0.5062502026557922, "global_step": 17038, "epoch": 191, "lr": 9.908346588009764e-05}
{"train_loss": 0.5059081315994263, "global_step": 17039, "epoch": 191, "lr": 9.908335538357814e-05}
{"train_loss": 0.5400317907333374, "global_step": 17040, "epoch": 191, "lr": 9.908324488045999e-05}
{"train_loss": 0.4417874813079834, "global_step": 17041, "epoch": 191, "lr": 9.908313437074317e-05}
{"train_loss": 0.5318945050239563, "global_step": 17042, "epoch": 191, "lr": 9.908302385442772e-05}
{"train_loss": 0.5669422149658203, "global_step": 17043, "epoch": 191, "lr": 9.908291333151367e-05}
{"train_loss": 0.6110528707504272, "global_step": 17044, "epoch": 191, "lr": 9.908280280200102e-05}
{"train_loss": 0.4964887499809265, "global_step": 17045, "epoch": 191, "lr": 9.908269226588978e-05}
{"train_loss": 0.5221104025840759, "global_step": 17046, "epoch": 191, "lr": 9.908258172317997e-05}
{"train_loss": 0.5886012315750122, "global_step": 17047, "epoch": 191, "lr": 9.908247117387162e-05}
{"train_loss": 0.623414158821106, "global_step": 17048, "epoch": 191, "lr": 9.908236061796471e-05}
{"train_loss": 0.5146237015724182, "global_step": 17049, "epoch": 191, "lr": 9.908225005545927e-05}
{"train_loss": 0.5421485304832458, "global_step": 17050, "epoch": 191, "lr": 9.908213948635534e-05}
{"train_loss": 0.37333452701568604, "global_step": 17051, "epoch": 191, "lr": 9.908202891065291e-05}
{"train_loss": 0.5757131576538086, "global_step": 17052, "epoch": 191, "lr": 9.908191832835198e-05}
{"train_loss": 0.5341511368751526, "global_step": 17053, "epoch": 191, "lr": 9.908180773945262e-05}
{"train_loss": 0.6429282426834106, "global_step": 17054, "epoch": 191, "lr": 9.908169714395478e-05}
{"train_loss": 0.46121886372566223, "global_step": 17055, "epoch": 191, "lr": 9.908158654185852e-05}
{"train_loss": 0.5151455998420715, "global_step": 17056, "epoch": 191, "lr": 9.908147593316383e-05}
{"train_loss": 0.5950570702552795, "global_step": 17057, "epoch": 191, "lr": 9.908136531787073e-05}
{"train_loss": 0.520977258682251, "global_step": 17058, "epoch": 191, "lr": 9.908125469597924e-05}
{"train_loss": 0.42288684844970703, "global_step": 17059, "epoch": 191, "lr": 9.908114406748938e-05}
{"train_loss": 0.5315874218940735, "global_step": 17060, "epoch": 191, "lr": 9.908103343240116e-05}
{"train_loss": 0.5643649697303772, "global_step": 17061, "epoch": 191, "lr": 9.908092279071459e-05}
{"train_loss": 0.5194269418716431, "global_step": 17062, "epoch": 191, "lr": 9.908081214242969e-05}
{"train_loss": 0.432794451713562, "global_step": 17063, "epoch": 191, "lr": 9.908070148754647e-05}
{"train_loss": 0.4835500717163086, "global_step": 17064, "epoch": 191, "lr": 9.908059082606496e-05}
{"train_loss": 0.5634145140647888, "global_step": 17065, "epoch": 191, "lr": 9.908048015798514e-05}
{"train_loss": 0.6051186323165894, "global_step": 17066, "epoch": 191, "lr": 9.908036948330706e-05}
{"train_loss": 0.5524632930755615, "global_step": 17067, "epoch": 191, "lr": 9.908025880203073e-05}
{"train_loss": 0.6429609060287476, "global_step": 17068, "epoch": 191, "lr": 9.908014811415613e-05}
{"train_loss": 0.4437233507633209, "global_step": 17069, "epoch": 191, "lr": 9.908003741968334e-05}
{"train_loss": 0.6302540898323059, "global_step": 17070, "epoch": 191, "lr": 9.907992671861232e-05}
{"train_loss": 0.6478418707847595, "global_step": 17071, "epoch": 191, "lr": 9.907981601094309e-05}
{"train_loss": 0.7900149822235107, "global_step": 17072, "epoch": 191, "lr": 9.90797052966757e-05}
{"train_loss": 0.5465930700302124, "global_step": 17073, "epoch": 191, "lr": 9.907959457581011e-05}
{"train_loss": 0.44760292768478394, "global_step": 17074, "epoch": 191, "lr": 9.90794838483464e-05}
{"train_loss": 0.5586248636245728, "global_step": 17075, "epoch": 191, "lr": 9.907937311428454e-05}
{"train_loss": 0.6425785422325134, "global_step": 17076, "epoch": 191, "lr": 9.907926237362455e-05}
{"train_loss": 0.7245164513587952, "global_step": 17077, "epoch": 191, "lr": 9.907915162636646e-05}
{"train_loss": 0.5464828014373779, "global_step": 17078, "epoch": 191, "lr": 9.907904087251027e-05}
{"train_loss": 0.6517571806907654, "global_step": 17079, "epoch": 191, "lr": 9.907893011205601e-05}
{"train_loss": 0.612144947052002, "global_step": 17080, "epoch": 191, "lr": 9.907881934500367e-05}
{"train_loss": 0.6534825563430786, "global_step": 17081, "epoch": 191, "lr": 9.907870857135329e-05}
{"train_loss": 0.4805721044540405, "global_step": 17082, "epoch": 191, "lr": 9.907859779110488e-05}
{"train_loss": 0.532882809638977, "global_step": 17083, "epoch": 191, "lr": 9.907848700425843e-05}
{"train_loss": 0.5141623020172119, "global_step": 17084, "epoch": 191, "lr": 9.907837621081401e-05}
{"train_loss": 0.5319401025772095, "global_step": 17085, "epoch": 191, "lr": 9.907826541077157e-05}
{"train_loss": 0.6073514819145203, "global_step": 17086, "epoch": 191, "lr": 9.907815460413116e-05}
{"train_loss": 0.5447803417618355, "global_step": 17087, "epoch": 191, "lr": 9.90780437908928e-05, "val_loss": 1.5827789306640625}
{"train_loss": 0.5388420224189758, "global_step": 17088, "epoch": 192, "lr": 9.90779329710565e-05}
{"train_loss": 0.6730645298957825, "global_step": 17089, "epoch": 192, "lr": 9.907782214462227e-05}
{"train_loss": 0.5007791519165039, "global_step": 17090, "epoch": 192, "lr": 9.907771131159011e-05}
{"train_loss": 0.5019861459732056, "global_step": 17091, "epoch": 192, "lr": 9.907760047196006e-05}
{"train_loss": 0.59494948387146, "global_step": 17092, "epoch": 192, "lr": 9.907748962573213e-05}
{"train_loss": 0.4811301827430725, "global_step": 17093, "epoch": 192, "lr": 9.907737877290632e-05}
{"train_loss": 0.643263578414917, "global_step": 17094, "epoch": 192, "lr": 9.907726791348266e-05}
{"train_loss": 0.5282413959503174, "global_step": 17095, "epoch": 192, "lr": 9.907715704746115e-05}
{"train_loss": 0.5788732171058655, "global_step": 17096, "epoch": 192, "lr": 9.907704617484182e-05}
{"train_loss": 0.4646105468273163, "global_step": 17097, "epoch": 192, "lr": 9.907693529562469e-05}
{"train_loss": 0.5144326686859131, "global_step": 17098, "epoch": 192, "lr": 9.907682440980976e-05}
{"train_loss": 0.5899024605751038, "global_step": 17099, "epoch": 192, "lr": 9.907671351739703e-05}
{"train_loss": 0.604350745677948, "global_step": 17100, "epoch": 192, "lr": 9.907660261838656e-05}
{"train_loss": 0.44200584292411804, "global_step": 17101, "epoch": 192, "lr": 9.907649171277833e-05}
{"train_loss": 0.44946569204330444, "global_step": 17102, "epoch": 192, "lr": 9.907638080057236e-05}
{"train_loss": 0.49963366985321045, "global_step": 17103, "epoch": 192, "lr": 9.907626988176865e-05}
{"train_loss": 0.5317786335945129, "global_step": 17104, "epoch": 192, "lr": 9.907615895636726e-05}
{"train_loss": 0.5318055152893066, "global_step": 17105, "epoch": 192, "lr": 9.907604802436819e-05}
{"train_loss": 0.4150253236293793, "global_step": 17106, "epoch": 192, "lr": 9.90759370857714e-05}
{"train_loss": 0.48023542761802673, "global_step": 17107, "epoch": 192, "lr": 9.907582614057699e-05}
{"train_loss": 0.5455834865570068, "global_step": 17108, "epoch": 192, "lr": 9.907571518878491e-05}
{"train_loss": 0.6008921265602112, "global_step": 17109, "epoch": 192, "lr": 9.907560423039522e-05}
{"train_loss": 0.5028266310691833, "global_step": 17110, "epoch": 192, "lr": 9.907549326540789e-05}
{"train_loss": 0.4593910872936249, "global_step": 17111, "epoch": 192, "lr": 9.907538229382296e-05}
{"train_loss": 0.5311393141746521, "global_step": 17112, "epoch": 192, "lr": 9.907527131564046e-05}
{"train_loss": 0.48293337225914, "global_step": 17113, "epoch": 192, "lr": 9.907516033086037e-05}
{"train_loss": 0.4662041962146759, "global_step": 17114, "epoch": 192, "lr": 9.907504933948273e-05}
{"train_loss": 0.48353397846221924, "global_step": 17115, "epoch": 192, "lr": 9.907493834150756e-05}
{"train_loss": 0.5437222719192505, "global_step": 17116, "epoch": 192, "lr": 9.907482733693484e-05}
{"train_loss": 0.47790658473968506, "global_step": 17117, "epoch": 192, "lr": 9.907471632576462e-05}
{"train_loss": 0.5691218972206116, "global_step": 17118, "epoch": 192, "lr": 9.90746053079969e-05}
{"train_loss": 0.49749046564102173, "global_step": 17119, "epoch": 192, "lr": 9.907449428363169e-05}
{"train_loss": 0.42100489139556885, "global_step": 17120, "epoch": 192, "lr": 9.907438325266903e-05}
{"train_loss": 0.7309430241584778, "global_step": 17121, "epoch": 192, "lr": 9.907427221510891e-05}
{"train_loss": 0.5634856224060059, "global_step": 17122, "epoch": 192, "lr": 9.907416117095133e-05}
{"train_loss": 0.45064470171928406, "global_step": 17123, "epoch": 192, "lr": 9.907405012019637e-05}
{"train_loss": 0.5395762324333191, "global_step": 17124, "epoch": 192, "lr": 9.907393906284397e-05}
{"train_loss": 0.41602590680122375, "global_step": 17125, "epoch": 192, "lr": 9.90738279988942e-05}
{"train_loss": 0.38818731904029846, "global_step": 17126, "epoch": 192, "lr": 9.907371692834704e-05}
{"train_loss": 0.6570452451705933, "global_step": 17127, "epoch": 192, "lr": 9.907360585120252e-05}
{"train_loss": 0.5779799222946167, "global_step": 17128, "epoch": 192, "lr": 9.907349476746064e-05}
{"train_loss": 0.5470030903816223, "global_step": 17129, "epoch": 192, "lr": 9.907338367712142e-05}
{"train_loss": 0.5826273560523987, "global_step": 17130, "epoch": 192, "lr": 9.907327258018491e-05}
{"train_loss": 0.4786528944969177, "global_step": 17131, "epoch": 192, "lr": 9.907316147665107e-05}
{"train_loss": 0.5155077576637268, "global_step": 17132, "epoch": 192, "lr": 9.907305036651996e-05}
{"train_loss": 0.4936717450618744, "global_step": 17133, "epoch": 192, "lr": 9.907293924979158e-05}
{"train_loss": 0.7899484634399414, "global_step": 17134, "epoch": 192, "lr": 9.907282812646592e-05}
{"train_loss": 0.6101655960083008, "global_step": 17135, "epoch": 192, "lr": 9.907271699654303e-05}
{"train_loss": 0.6228137612342834, "global_step": 17136, "epoch": 192, "lr": 9.907260586002292e-05}
{"train_loss": 0.615618884563446, "global_step": 17137, "epoch": 192, "lr": 9.907249471690557e-05}
{"train_loss": 0.37868398427963257, "global_step": 17138, "epoch": 192, "lr": 9.907238356719105e-05}
{"train_loss": 0.6398932933807373, "global_step": 17139, "epoch": 192, "lr": 9.907227241087932e-05}
{"train_loss": 0.7665865421295166, "global_step": 17140, "epoch": 192, "lr": 9.907216124797042e-05}
{"train_loss": 0.5075905919075012, "global_step": 17141, "epoch": 192, "lr": 9.90720500784644e-05}
{"train_loss": 0.5174134969711304, "global_step": 17142, "epoch": 192, "lr": 9.907193890236121e-05}
{"train_loss": 0.5916734933853149, "global_step": 17143, "epoch": 192, "lr": 9.907182771966092e-05}
{"train_loss": 0.6065647602081299, "global_step": 17144, "epoch": 192, "lr": 9.907171653036351e-05}
{"train_loss": 0.5464265942573547, "global_step": 17145, "epoch": 192, "lr": 9.907160533446899e-05}
{"train_loss": 0.4679224491119385, "global_step": 17146, "epoch": 192, "lr": 9.90714941319774e-05}
{"train_loss": 0.5721035003662109, "global_step": 17147, "epoch": 192, "lr": 9.907138292288875e-05}
{"train_loss": 0.6158692836761475, "global_step": 17148, "epoch": 192, "lr": 9.907127170720306e-05}
{"train_loss": 0.623270571231842, "global_step": 17149, "epoch": 192, "lr": 9.907116048492033e-05}
{"train_loss": 0.6082109212875366, "global_step": 17150, "epoch": 192, "lr": 9.907104925604059e-05}
{"train_loss": 0.5873966217041016, "global_step": 17151, "epoch": 192, "lr": 9.907093802056382e-05}
{"train_loss": 0.6862093806266785, "global_step": 17152, "epoch": 192, "lr": 9.907082677849007e-05}
{"train_loss": 0.5900182127952576, "global_step": 17153, "epoch": 192, "lr": 9.907071552981935e-05}
{"train_loss": 0.5861361622810364, "global_step": 17154, "epoch": 192, "lr": 9.907060427455167e-05}
{"train_loss": 0.48435118794441223, "global_step": 17155, "epoch": 192, "lr": 9.907049301268704e-05}
{"train_loss": 0.5354010462760925, "global_step": 17156, "epoch": 192, "lr": 9.907038174422549e-05}
{"train_loss": 0.6025262475013733, "global_step": 17157, "epoch": 192, "lr": 9.907027046916701e-05}
{"train_loss": 0.6725830435752869, "global_step": 17158, "epoch": 192, "lr": 9.907015918751164e-05}
{"train_loss": 0.551218569278717, "global_step": 17159, "epoch": 192, "lr": 9.90700478992594e-05}
{"train_loss": 0.5603733062744141, "global_step": 17160, "epoch": 192, "lr": 9.906993660441026e-05}
{"train_loss": 0.6459998488426208, "global_step": 17161, "epoch": 192, "lr": 9.90698253029643e-05}
{"train_loss": 0.6083710193634033, "global_step": 17162, "epoch": 192, "lr": 9.906971399492146e-05}
{"train_loss": 0.5400636196136475, "global_step": 17163, "epoch": 192, "lr": 9.906960268028183e-05}
{"train_loss": 0.48868328332901, "global_step": 17164, "epoch": 192, "lr": 9.906949135904538e-05}
{"train_loss": 0.6009055376052856, "global_step": 17165, "epoch": 192, "lr": 9.906938003121212e-05}
{"train_loss": 0.6135115027427673, "global_step": 17166, "epoch": 192, "lr": 9.906926869678209e-05}
{"train_loss": 0.5601174235343933, "global_step": 17167, "epoch": 192, "lr": 9.90691573557553e-05}
{"train_loss": 0.5245142579078674, "global_step": 17168, "epoch": 192, "lr": 9.906904600813175e-05}
{"train_loss": 0.5408065319061279, "global_step": 17169, "epoch": 192, "lr": 9.906893465391147e-05}
{"train_loss": 0.7224525213241577, "global_step": 17170, "epoch": 192, "lr": 9.906882329309447e-05}
{"train_loss": 0.4358300268650055, "global_step": 17171, "epoch": 192, "lr": 9.906871192568077e-05}
{"train_loss": 0.4876248240470886, "global_step": 17172, "epoch": 192, "lr": 9.906860055167037e-05}
{"train_loss": 0.4745836555957794, "global_step": 17173, "epoch": 192, "lr": 9.906848917106331e-05}
{"train_loss": 0.49363863468170166, "global_step": 17174, "epoch": 192, "lr": 9.906837778385957e-05}
{"train_loss": 0.542344868183136, "global_step": 17175, "epoch": 192, "lr": 9.906826639005919e-05}
{"train_loss": 0.5494535096575704, "global_step": 17176, "epoch": 192, "lr": 9.906815498966217e-05, "val_loss": 1.5325028896331787}
{"train_loss": 0.41531598567962646, "global_step": 17177, "epoch": 193, "lr": 9.906804358266856e-05}
{"train_loss": 0.5912060737609863, "global_step": 17178, "epoch": 193, "lr": 9.906793216907834e-05}
{"train_loss": 0.5114294290542603, "global_step": 17179, "epoch": 193, "lr": 9.906782074889153e-05}
{"train_loss": 0.5112801194190979, "global_step": 17180, "epoch": 193, "lr": 9.906770932210815e-05}
{"train_loss": 0.5313252210617065, "global_step": 17181, "epoch": 193, "lr": 9.906759788872821e-05}
{"train_loss": 0.42992138862609863, "global_step": 17182, "epoch": 193, "lr": 9.906748644875174e-05}
{"train_loss": 0.651466429233551, "global_step": 17183, "epoch": 193, "lr": 9.906737500217874e-05}
{"train_loss": 0.610781729221344, "global_step": 17184, "epoch": 193, "lr": 9.906726354900924e-05}
{"train_loss": 0.5668094158172607, "global_step": 17185, "epoch": 193, "lr": 9.906715208924322e-05}
{"train_loss": 0.6574775576591492, "global_step": 17186, "epoch": 193, "lr": 9.906704062288073e-05}
{"train_loss": 0.5782230496406555, "global_step": 17187, "epoch": 193, "lr": 9.906692914992179e-05}
{"train_loss": 0.5290751457214355, "global_step": 17188, "epoch": 193, "lr": 9.906681767036638e-05}
{"train_loss": 0.5563099980354309, "global_step": 17189, "epoch": 193, "lr": 9.906670618421454e-05}
{"train_loss": 0.4726778566837311, "global_step": 17190, "epoch": 193, "lr": 9.906659469146628e-05}
{"train_loss": 0.559190571308136, "global_step": 17191, "epoch": 193, "lr": 9.906648319212161e-05}
{"train_loss": 0.6577181220054626, "global_step": 17192, "epoch": 193, "lr": 9.906637168618056e-05}
{"train_loss": 0.5813271403312683, "global_step": 17193, "epoch": 193, "lr": 9.906626017364313e-05}
{"train_loss": 0.5040982365608215, "global_step": 17194, "epoch": 193, "lr": 9.906614865450933e-05}
{"train_loss": 0.4273894429206848, "global_step": 17195, "epoch": 193, "lr": 9.90660371287792e-05}
{"train_loss": 0.5301014184951782, "global_step": 17196, "epoch": 193, "lr": 9.906592559645274e-05}
{"train_loss": 0.43447843194007874, "global_step": 17197, "epoch": 193, "lr": 9.906581405752995e-05}
{"train_loss": 0.4752933382987976, "global_step": 17198, "epoch": 193, "lr": 9.906570251201086e-05}
{"train_loss": 0.4933534264564514, "global_step": 17199, "epoch": 193, "lr": 9.90655909598955e-05}
{"train_loss": 0.4382242262363434, "global_step": 17200, "epoch": 193, "lr": 9.906547940118386e-05}
{"train_loss": 0.5350853204727173, "global_step": 17201, "epoch": 193, "lr": 9.906536783587596e-05}
{"train_loss": 0.45159947872161865, "global_step": 17202, "epoch": 193, "lr": 9.906525626397182e-05}
{"train_loss": 0.5406452417373657, "global_step": 17203, "epoch": 193, "lr": 9.906514468547145e-05}
{"train_loss": 0.5888566970825195, "global_step": 17204, "epoch": 193, "lr": 9.90650331003749e-05}
{"train_loss": 0.4928644895553589, "global_step": 17205, "epoch": 193, "lr": 9.906492150868212e-05}
{"train_loss": 0.4790617823600769, "global_step": 17206, "epoch": 193, "lr": 9.906480991039317e-05}
{"train_loss": 0.4791962504386902, "global_step": 17207, "epoch": 193, "lr": 9.906469830550805e-05}
{"train_loss": 0.39142435789108276, "global_step": 17208, "epoch": 193, "lr": 9.906458669402679e-05}
{"train_loss": 0.6472444534301758, "global_step": 17209, "epoch": 193, "lr": 9.906447507594938e-05}
{"train_loss": 0.5133399963378906, "global_step": 17210, "epoch": 193, "lr": 9.906436345127587e-05}
{"train_loss": 0.5163095593452454, "global_step": 17211, "epoch": 193, "lr": 9.906425182000624e-05}
{"train_loss": 0.5209180116653442, "global_step": 17212, "epoch": 193, "lr": 9.906414018214053e-05}
{"train_loss": 0.5105398297309875, "global_step": 17213, "epoch": 193, "lr": 9.906402853767872e-05}
{"train_loss": 0.4720558822154999, "global_step": 17214, "epoch": 193, "lr": 9.906391688662086e-05}
{"train_loss": 0.5134009718894958, "global_step": 17215, "epoch": 193, "lr": 9.906380522896697e-05}
{"train_loss": 0.42509955167770386, "global_step": 17216, "epoch": 193, "lr": 9.906369356471703e-05}
{"train_loss": 0.43491092324256897, "global_step": 17217, "epoch": 193, "lr": 9.906358189387108e-05}
{"train_loss": 0.5322160720825195, "global_step": 17218, "epoch": 193, "lr": 9.906347021642914e-05}
{"train_loss": 0.41999000310897827, "global_step": 17219, "epoch": 193, "lr": 9.906335853239122e-05}
{"train_loss": 0.48765885829925537, "global_step": 17220, "epoch": 193, "lr": 9.90632468417573e-05}
{"train_loss": 0.6091039776802063, "global_step": 17221, "epoch": 193, "lr": 9.906313514452746e-05}
{"train_loss": 0.5222017765045166, "global_step": 17222, "epoch": 193, "lr": 9.906302344070165e-05}
{"train_loss": 0.5543556809425354, "global_step": 17223, "epoch": 193, "lr": 9.906291173027993e-05}
{"train_loss": 0.5663133263587952, "global_step": 17224, "epoch": 193, "lr": 9.90628000132623e-05}
{"train_loss": 0.5869651436805725, "global_step": 17225, "epoch": 193, "lr": 9.906268828964875e-05}
{"train_loss": 0.5821529626846313, "global_step": 17226, "epoch": 193, "lr": 9.906257655943933e-05}
{"train_loss": 0.6906951069831848, "global_step": 17227, "epoch": 193, "lr": 9.906246482263406e-05}
{"train_loss": 0.5986918807029724, "global_step": 17228, "epoch": 193, "lr": 9.906235307923293e-05}
{"train_loss": 0.5125892162322998, "global_step": 17229, "epoch": 193, "lr": 9.906224132923598e-05}
{"train_loss": 0.4641624689102173, "global_step": 17230, "epoch": 193, "lr": 9.906212957264319e-05}
{"train_loss": 0.5102845430374146, "global_step": 17231, "epoch": 193, "lr": 9.906201780945462e-05}
{"train_loss": 0.5385406017303467, "global_step": 17232, "epoch": 193, "lr": 9.906190603967023e-05}
{"train_loss": 0.665337860584259, "global_step": 17233, "epoch": 193, "lr": 9.906179426329007e-05}
{"train_loss": 0.3681243658065796, "global_step": 17234, "epoch": 193, "lr": 9.906168248031416e-05}
{"train_loss": 0.50168377161026, "global_step": 17235, "epoch": 193, "lr": 9.90615706907425e-05}
{"train_loss": 0.583644688129425, "global_step": 17236, "epoch": 193, "lr": 9.906145889457511e-05}
{"train_loss": 0.5589647889137268, "global_step": 17237, "epoch": 193, "lr": 9.9061347091812e-05}
{"train_loss": 0.40880489349365234, "global_step": 17238, "epoch": 193, "lr": 9.90612352824532e-05}
{"train_loss": 0.5688357353210449, "global_step": 17239, "epoch": 193, "lr": 9.906112346649871e-05}
{"train_loss": 0.6656976938247681, "global_step": 17240, "epoch": 193, "lr": 9.906101164394855e-05}
{"train_loss": 0.4542660117149353, "global_step": 17241, "epoch": 193, "lr": 9.906089981480274e-05}
{"train_loss": 0.6220356822013855, "global_step": 17242, "epoch": 193, "lr": 9.906078797906128e-05}
{"train_loss": 0.557902455329895, "global_step": 17243, "epoch": 193, "lr": 9.90606761367242e-05}
{"train_loss": 0.686833918094635, "global_step": 17244, "epoch": 193, "lr": 9.906056428779152e-05}
{"train_loss": 0.500389814376831, "global_step": 17245, "epoch": 193, "lr": 9.906045243226323e-05}
{"train_loss": 0.47002866864204407, "global_step": 17246, "epoch": 193, "lr": 9.906034057013935e-05}
{"train_loss": 0.6263472437858582, "global_step": 17247, "epoch": 193, "lr": 9.906022870141993e-05}
{"train_loss": 0.5327329635620117, "global_step": 17248, "epoch": 193, "lr": 9.906011682610495e-05}
{"train_loss": 0.5299315452575684, "global_step": 17249, "epoch": 193, "lr": 9.906000494419443e-05}
{"train_loss": 0.4847157597541809, "global_step": 17250, "epoch": 193, "lr": 9.905989305568839e-05}
{"train_loss": 0.5632674098014832, "global_step": 17251, "epoch": 193, "lr": 9.905978116058686e-05}
{"train_loss": 0.4405718445777893, "global_step": 17252, "epoch": 193, "lr": 9.905966925888983e-05}
{"train_loss": 0.5883259773254395, "global_step": 17253, "epoch": 193, "lr": 9.905955735059733e-05}
{"train_loss": 0.6054739356040955, "global_step": 17254, "epoch": 193, "lr": 9.905944543570937e-05}
{"train_loss": 0.5747178196907043, "global_step": 17255, "epoch": 193, "lr": 9.905933351422596e-05}
{"train_loss": 0.4391458332538605, "global_step": 17256, "epoch": 193, "lr": 9.905922158614712e-05}
{"train_loss": 0.5361530780792236, "global_step": 17257, "epoch": 193, "lr": 9.905910965147287e-05}
{"train_loss": 0.5800943970680237, "global_step": 17258, "epoch": 193, "lr": 9.905899771020322e-05}
{"train_loss": 0.6137162446975708, "global_step": 17259, "epoch": 193, "lr": 9.905888576233819e-05}
{"train_loss": 0.4408330023288727, "global_step": 17260, "epoch": 193, "lr": 9.905877380787779e-05}
{"train_loss": 0.48159173130989075, "global_step": 17261, "epoch": 193, "lr": 9.905866184682202e-05}
{"train_loss": 0.5124735236167908, "global_step": 17262, "epoch": 193, "lr": 9.905854987917094e-05}
{"train_loss": 0.7403749823570251, "global_step": 17263, "epoch": 193, "lr": 9.90584379049245e-05}
{"train_loss": 0.5452678203582764, "global_step": 17264, "epoch": 193, "lr": 9.905832592408278e-05}
{"train_loss": 0.5336033042227284, "global_step": 17265, "epoch": 193, "lr": 9.905821393664576e-05, "val_loss": 1.5721849203109741}
{"train_loss": 0.47488895058631897, "global_step": 17266, "epoch": 194, "lr": 9.905810194261345e-05}
{"train_loss": 0.42799651622772217, "global_step": 17267, "epoch": 194, "lr": 9.905798994198589e-05}
{"train_loss": 0.5475627779960632, "global_step": 17268, "epoch": 194, "lr": 9.905787793476307e-05}
{"train_loss": 0.44761958718299866, "global_step": 17269, "epoch": 194, "lr": 9.905776592094502e-05}
{"train_loss": 0.592137336730957, "global_step": 17270, "epoch": 194, "lr": 9.905765390053174e-05}
{"train_loss": 0.4193700850009918, "global_step": 17271, "epoch": 194, "lr": 9.905754187352328e-05}
{"train_loss": 0.47803449630737305, "global_step": 17272, "epoch": 194, "lr": 9.905742983991961e-05}
{"train_loss": 0.4196360111236572, "global_step": 17273, "epoch": 194, "lr": 9.905731779972078e-05}
{"train_loss": 0.6099669933319092, "global_step": 17274, "epoch": 194, "lr": 9.905720575292678e-05}
{"train_loss": 0.46806496381759644, "global_step": 17275, "epoch": 194, "lr": 9.905709369953767e-05}
{"train_loss": 0.5990917682647705, "global_step": 17276, "epoch": 194, "lr": 9.90569816395534e-05}
{"train_loss": 0.45182308554649353, "global_step": 17277, "epoch": 194, "lr": 9.905686957297402e-05}
{"train_loss": 0.4535951316356659, "global_step": 17278, "epoch": 194, "lr": 9.905675749979955e-05}
{"train_loss": 0.3581124246120453, "global_step": 17279, "epoch": 194, "lr": 9.905664542002998e-05}
{"train_loss": 0.5398120284080505, "global_step": 17280, "epoch": 194, "lr": 9.905653333366535e-05}
{"train_loss": 0.42949217557907104, "global_step": 17281, "epoch": 194, "lr": 9.905642124070567e-05}
{"train_loss": 0.6016910076141357, "global_step": 17282, "epoch": 194, "lr": 9.905630914115096e-05}
{"train_loss": 0.4904367923736572, "global_step": 17283, "epoch": 194, "lr": 9.905619703500122e-05}
{"train_loss": 0.547530472278595, "global_step": 17284, "epoch": 194, "lr": 9.905608492225646e-05}
{"train_loss": 0.4669816493988037, "global_step": 17285, "epoch": 194, "lr": 9.905597280291672e-05}
{"train_loss": 0.5439727902412415, "global_step": 17286, "epoch": 194, "lr": 9.9055860676982e-05}
{"train_loss": 0.5358145236968994, "global_step": 17287, "epoch": 194, "lr": 9.905574854445233e-05}
{"train_loss": 0.4370054006576538, "global_step": 17288, "epoch": 194, "lr": 9.90556364053277e-05}
{"train_loss": 0.5992537140846252, "global_step": 17289, "epoch": 194, "lr": 9.905552425960813e-05}
{"train_loss": 0.4054523706436157, "global_step": 17290, "epoch": 194, "lr": 9.905541210729365e-05}
{"train_loss": 0.5507725477218628, "global_step": 17291, "epoch": 194, "lr": 9.905529994838428e-05}
{"train_loss": 0.48979151248931885, "global_step": 17292, "epoch": 194, "lr": 9.905518778288e-05}
{"train_loss": 0.48274552822113037, "global_step": 17293, "epoch": 194, "lr": 9.905507561078087e-05}
{"train_loss": 0.6211414337158203, "global_step": 17294, "epoch": 194, "lr": 9.905496343208687e-05}
{"train_loss": 0.5163307785987854, "global_step": 17295, "epoch": 194, "lr": 9.905485124679803e-05}
{"train_loss": 0.4908815920352936, "global_step": 17296, "epoch": 194, "lr": 9.905473905491437e-05}
{"train_loss": 0.4056634306907654, "global_step": 17297, "epoch": 194, "lr": 9.905462685643589e-05}
{"train_loss": 0.5731518864631653, "global_step": 17298, "epoch": 194, "lr": 9.905451465136261e-05}
{"train_loss": 0.45904114842414856, "global_step": 17299, "epoch": 194, "lr": 9.905440243969456e-05}
{"train_loss": 0.5205901861190796, "global_step": 17300, "epoch": 194, "lr": 9.905429022143173e-05}
{"train_loss": 0.4994639456272125, "global_step": 17301, "epoch": 194, "lr": 9.905417799657418e-05}
{"train_loss": 0.4223954677581787, "global_step": 17302, "epoch": 194, "lr": 9.905406576512187e-05}
{"train_loss": 0.5673298239707947, "global_step": 17303, "epoch": 194, "lr": 9.905395352707483e-05}
{"train_loss": 0.4387286305427551, "global_step": 17304, "epoch": 194, "lr": 9.90538412824331e-05}
{"train_loss": 0.4700840711593628, "global_step": 17305, "epoch": 194, "lr": 9.905372903119665e-05}
{"train_loss": 0.5664048790931702, "global_step": 17306, "epoch": 194, "lr": 9.905361677336555e-05}
{"train_loss": 0.461660236120224, "global_step": 17307, "epoch": 194, "lr": 9.905350450893978e-05}
{"train_loss": 0.5384587645530701, "global_step": 17308, "epoch": 194, "lr": 9.905339223791938e-05}
{"train_loss": 0.5749794840812683, "global_step": 17309, "epoch": 194, "lr": 9.905327996030433e-05}
{"train_loss": 0.5077715516090393, "global_step": 17310, "epoch": 194, "lr": 9.905316767609468e-05}
{"train_loss": 0.5660550594329834, "global_step": 17311, "epoch": 194, "lr": 9.905305538529042e-05}
{"train_loss": 0.5735135674476624, "global_step": 17312, "epoch": 194, "lr": 9.905294308789157e-05}
{"train_loss": 0.4824107587337494, "global_step": 17313, "epoch": 194, "lr": 9.905283078389815e-05}
{"train_loss": 0.5949064493179321, "global_step": 17314, "epoch": 194, "lr": 9.90527184733102e-05}
{"train_loss": 0.4263631999492645, "global_step": 17315, "epoch": 194, "lr": 9.905260615612768e-05}
{"train_loss": 0.4161883294582367, "global_step": 17316, "epoch": 194, "lr": 9.905249383235065e-05}
{"train_loss": 0.4693104922771454, "global_step": 17317, "epoch": 194, "lr": 9.905238150197911e-05}
{"train_loss": 0.4786399006843567, "global_step": 17318, "epoch": 194, "lr": 9.905226916501307e-05}
{"train_loss": 0.5913368463516235, "global_step": 17319, "epoch": 194, "lr": 9.905215682145255e-05}
{"train_loss": 0.529977560043335, "global_step": 17320, "epoch": 194, "lr": 9.905204447129756e-05}
{"train_loss": 0.5105536580085754, "global_step": 17321, "epoch": 194, "lr": 9.905193211454813e-05}
{"train_loss": 0.5701209306716919, "global_step": 17322, "epoch": 194, "lr": 9.905181975120425e-05}
{"train_loss": 0.46808961033821106, "global_step": 17323, "epoch": 194, "lr": 9.905170738126597e-05}
{"train_loss": 0.6290097832679749, "global_step": 17324, "epoch": 194, "lr": 9.905159500473329e-05}
{"train_loss": 0.4321373701095581, "global_step": 17325, "epoch": 194, "lr": 9.905148262160621e-05}
{"train_loss": 0.5578024387359619, "global_step": 17326, "epoch": 194, "lr": 9.905137023188475e-05}
{"train_loss": 0.6730821132659912, "global_step": 17327, "epoch": 194, "lr": 9.905125783556895e-05}
{"train_loss": 0.5342648029327393, "global_step": 17328, "epoch": 194, "lr": 9.905114543265879e-05}
{"train_loss": 0.5271729230880737, "global_step": 17329, "epoch": 194, "lr": 9.905103302315429e-05}
{"train_loss": 0.5603984594345093, "global_step": 17330, "epoch": 194, "lr": 9.90509206070555e-05}
{"train_loss": 0.5350520610809326, "global_step": 17331, "epoch": 194, "lr": 9.905080818436239e-05}
{"train_loss": 0.5117179751396179, "global_step": 17332, "epoch": 194, "lr": 9.905069575507502e-05}
{"train_loss": 0.3568585216999054, "global_step": 17333, "epoch": 194, "lr": 9.905058331919337e-05}
{"train_loss": 0.5416345000267029, "global_step": 17334, "epoch": 194, "lr": 9.905047087671747e-05}
{"train_loss": 0.6805640459060669, "global_step": 17335, "epoch": 194, "lr": 9.905035842764733e-05}
{"train_loss": 0.6845167279243469, "global_step": 17336, "epoch": 194, "lr": 9.905024597198297e-05}
{"train_loss": 0.6055546402931213, "global_step": 17337, "epoch": 194, "lr": 9.90501335097244e-05}
{"train_loss": 0.5035676956176758, "global_step": 17338, "epoch": 194, "lr": 9.905002104087164e-05}
{"train_loss": 0.5094157457351685, "global_step": 17339, "epoch": 194, "lr": 9.904990856542468e-05}
{"train_loss": 0.4991779327392578, "global_step": 17340, "epoch": 194, "lr": 9.904979608338359e-05}
{"train_loss": 0.5751311779022217, "global_step": 17341, "epoch": 194, "lr": 9.904968359474833e-05}
{"train_loss": 0.627945065498352, "global_step": 17342, "epoch": 194, "lr": 9.904957109951896e-05}
{"train_loss": 0.6807822585105896, "global_step": 17343, "epoch": 194, "lr": 9.904945859769546e-05}
{"train_loss": 0.4869599938392639, "global_step": 17344, "epoch": 194, "lr": 9.904934608927785e-05}
{"train_loss": 0.49765321612358093, "global_step": 17345, "epoch": 194, "lr": 9.904923357426616e-05}
{"train_loss": 0.5949164628982544, "global_step": 17346, "epoch": 194, "lr": 9.904912105266042e-05}
{"train_loss": 0.5246734023094177, "global_step": 17347, "epoch": 194, "lr": 9.904900852446059e-05}
{"train_loss": 0.6418352127075195, "global_step": 17348, "epoch": 194, "lr": 9.904889598966673e-05}
{"train_loss": 0.5521516799926758, "global_step": 17349, "epoch": 194, "lr": 9.904878344827886e-05}
{"train_loss": 0.562171220779419, "global_step": 17350, "epoch": 194, "lr": 9.904867090029696e-05}
{"train_loss": 0.49604105949401855, "global_step": 17351, "epoch": 194, "lr": 9.904855834572106e-05}
{"train_loss": 0.5699653029441833, "global_step": 17352, "epoch": 194, "lr": 9.904844578455119e-05}
{"train_loss": 0.4445742964744568, "global_step": 17353, "epoch": 194, "lr": 9.904833321678736e-05}
{"train_loss": 0.5201519295740663, "global_step": 17354, "epoch": 194, "lr": 9.904822064242958e-05, "val_loss": 1.5688139200210571}
{"train_loss": 0.5327185988426208, "global_step": 17355, "epoch": 195, "lr": 9.904810806147785e-05}
{"train_loss": 0.5835020542144775, "global_step": 17356, "epoch": 195, "lr": 9.90479954739322e-05}
{"train_loss": 0.4753987491130829, "global_step": 17357, "epoch": 195, "lr": 9.904788287979265e-05}
{"train_loss": 0.5326455235481262, "global_step": 17358, "epoch": 195, "lr": 9.904777027905922e-05}
{"train_loss": 0.4590122699737549, "global_step": 17359, "epoch": 195, "lr": 9.90476576717319e-05}
{"train_loss": 0.6847620606422424, "global_step": 17360, "epoch": 195, "lr": 9.904754505781074e-05}
{"train_loss": 0.5127994418144226, "global_step": 17361, "epoch": 195, "lr": 9.90474324372957e-05}
{"train_loss": 0.4897248148918152, "global_step": 17362, "epoch": 195, "lr": 9.904731981018687e-05}
{"train_loss": 0.4456428587436676, "global_step": 17363, "epoch": 195, "lr": 9.904720717648421e-05}
{"train_loss": 0.47134867310523987, "global_step": 17364, "epoch": 195, "lr": 9.904709453618775e-05}
{"train_loss": 0.37168824672698975, "global_step": 17365, "epoch": 195, "lr": 9.90469818892975e-05}
{"train_loss": 0.5583891272544861, "global_step": 17366, "epoch": 195, "lr": 9.904686923581348e-05}
{"train_loss": 0.6815919876098633, "global_step": 17367, "epoch": 195, "lr": 9.904675657573572e-05}
{"train_loss": 0.6368008255958557, "global_step": 17368, "epoch": 195, "lr": 9.904664390906421e-05}
{"train_loss": 0.5624457001686096, "global_step": 17369, "epoch": 195, "lr": 9.904653123579898e-05}
{"train_loss": 0.46917203068733215, "global_step": 17370, "epoch": 195, "lr": 9.904641855594003e-05}
{"train_loss": 0.6294905543327332, "global_step": 17371, "epoch": 195, "lr": 9.904630586948741e-05}
{"train_loss": 0.5211636424064636, "global_step": 17372, "epoch": 195, "lr": 9.90461931764411e-05}
{"train_loss": 0.4728451669216156, "global_step": 17373, "epoch": 195, "lr": 9.904608047680113e-05}
{"train_loss": 0.6490836143493652, "global_step": 17374, "epoch": 195, "lr": 9.90459677705675e-05}
{"train_loss": 0.452938437461853, "global_step": 17375, "epoch": 195, "lr": 9.904585505774025e-05}
{"train_loss": 0.527155876159668, "global_step": 17376, "epoch": 195, "lr": 9.904574233831938e-05}
{"train_loss": 0.45176568627357483, "global_step": 17377, "epoch": 195, "lr": 9.90456296123049e-05}
{"train_loss": 0.6729281544685364, "global_step": 17378, "epoch": 195, "lr": 9.904551687969685e-05}
{"train_loss": 0.4446214437484741, "global_step": 17379, "epoch": 195, "lr": 9.904540414049521e-05}
{"train_loss": 0.5561097860336304, "global_step": 17380, "epoch": 195, "lr": 9.904529139470002e-05}
{"train_loss": 0.6120607852935791, "global_step": 17381, "epoch": 195, "lr": 9.904517864231131e-05}
{"train_loss": 0.5063823461532593, "global_step": 17382, "epoch": 195, "lr": 9.904506588332905e-05}
{"train_loss": 0.5488886833190918, "global_step": 17383, "epoch": 195, "lr": 9.904495311775328e-05}
{"train_loss": 0.5206928253173828, "global_step": 17384, "epoch": 195, "lr": 9.904484034558401e-05}
{"train_loss": 0.4269000291824341, "global_step": 17385, "epoch": 195, "lr": 9.904472756682127e-05}
{"train_loss": 0.5710852742195129, "global_step": 17386, "epoch": 195, "lr": 9.904461478146506e-05}
{"train_loss": 0.5115116238594055, "global_step": 17387, "epoch": 195, "lr": 9.904450198951541e-05}
{"train_loss": 0.6314177513122559, "global_step": 17388, "epoch": 195, "lr": 9.904438919097229e-05}
{"train_loss": 0.6219469308853149, "global_step": 17389, "epoch": 195, "lr": 9.904427638583578e-05}
{"train_loss": 0.6054091453552246, "global_step": 17390, "epoch": 195, "lr": 9.904416357410586e-05}
{"train_loss": 0.6846376657485962, "global_step": 17391, "epoch": 195, "lr": 9.904405075578255e-05}
{"train_loss": 0.6018498539924622, "global_step": 17392, "epoch": 195, "lr": 9.904393793086586e-05}
{"train_loss": 0.4759628474712372, "global_step": 17393, "epoch": 195, "lr": 9.904382509935582e-05}
{"train_loss": 0.6822670102119446, "global_step": 17394, "epoch": 195, "lr": 9.904371226125241e-05}
{"train_loss": 0.5390594005584717, "global_step": 17395, "epoch": 195, "lr": 9.90435994165557e-05}
{"train_loss": 0.4971255362033844, "global_step": 17396, "epoch": 195, "lr": 9.904348656526567e-05}
{"train_loss": 0.29067033529281616, "global_step": 17397, "epoch": 195, "lr": 9.904337370738232e-05}
{"train_loss": 0.6275150179862976, "global_step": 17398, "epoch": 195, "lr": 9.904326084290571e-05}
{"train_loss": 0.546050488948822, "global_step": 17399, "epoch": 195, "lr": 9.904314797183583e-05}
{"train_loss": 0.5553160905838013, "global_step": 17400, "epoch": 195, "lr": 9.904303509417268e-05}
{"train_loss": 0.6549386382102966, "global_step": 17401, "epoch": 195, "lr": 9.90429222099163e-05}
{"train_loss": 0.5851394534111023, "global_step": 17402, "epoch": 195, "lr": 9.904280931906671e-05}
{"train_loss": 0.6399015784263611, "global_step": 17403, "epoch": 195, "lr": 9.904269642162389e-05}
{"train_loss": 0.630330502986908, "global_step": 17404, "epoch": 195, "lr": 9.904258351758788e-05}
{"train_loss": 0.41832658648490906, "global_step": 17405, "epoch": 195, "lr": 9.904247060695871e-05}
{"train_loss": 0.5884848833084106, "global_step": 17406, "epoch": 195, "lr": 9.904235768973638e-05}
{"train_loss": 0.5753712058067322, "global_step": 17407, "epoch": 195, "lr": 9.904224476592087e-05}
{"train_loss": 0.49855220317840576, "global_step": 17408, "epoch": 195, "lr": 9.904213183551225e-05}
{"train_loss": 0.42871540784835815, "global_step": 17409, "epoch": 195, "lr": 9.904201889851053e-05}
{"train_loss": 0.5299964547157288, "global_step": 17410, "epoch": 195, "lr": 9.904190595491568e-05}
{"train_loss": 0.43590039014816284, "global_step": 17411, "epoch": 195, "lr": 9.904179300472775e-05}
{"train_loss": 0.6032446026802063, "global_step": 17412, "epoch": 195, "lr": 9.904168004794676e-05}
{"train_loss": 0.6021128296852112, "global_step": 17413, "epoch": 195, "lr": 9.90415670845727e-05}
{"train_loss": 0.4726570248603821, "global_step": 17414, "epoch": 195, "lr": 9.904145411460561e-05}
{"train_loss": 0.4251863360404968, "global_step": 17415, "epoch": 195, "lr": 9.904134113804548e-05}
{"train_loss": 0.4419889450073242, "global_step": 17416, "epoch": 195, "lr": 9.904122815489235e-05}
{"train_loss": 0.6605293154716492, "global_step": 17417, "epoch": 195, "lr": 9.904111516514622e-05}
{"train_loss": 0.6864753365516663, "global_step": 17418, "epoch": 195, "lr": 9.904100216880712e-05}
{"train_loss": 0.7233479022979736, "global_step": 17419, "epoch": 195, "lr": 9.904088916587504e-05}
{"train_loss": 0.5551213026046753, "global_step": 17420, "epoch": 195, "lr": 9.904077615635002e-05}
{"train_loss": 0.6008148789405823, "global_step": 17421, "epoch": 195, "lr": 9.904066314023206e-05}
{"train_loss": 0.5820413827896118, "global_step": 17422, "epoch": 195, "lr": 9.904055011752118e-05}
{"train_loss": 0.5779935121536255, "global_step": 17423, "epoch": 195, "lr": 9.904043708821741e-05}
{"train_loss": 0.4978578984737396, "global_step": 17424, "epoch": 195, "lr": 9.904032405232073e-05}
{"train_loss": 0.46392622590065, "global_step": 17425, "epoch": 195, "lr": 9.904021100983119e-05}
{"train_loss": 0.554874062538147, "global_step": 17426, "epoch": 195, "lr": 9.904009796074881e-05}
{"train_loss": 0.5412346720695496, "global_step": 17427, "epoch": 195, "lr": 9.903998490507355e-05}
{"train_loss": 0.6840429902076721, "global_step": 17428, "epoch": 195, "lr": 9.903987184280548e-05}
{"train_loss": 0.6147992014884949, "global_step": 17429, "epoch": 195, "lr": 9.903975877394458e-05}
{"train_loss": 0.5243865847587585, "global_step": 17430, "epoch": 195, "lr": 9.90396456984909e-05}
{"train_loss": 0.6060870289802551, "global_step": 17431, "epoch": 195, "lr": 9.903953261644444e-05}
{"train_loss": 0.5019024610519409, "global_step": 17432, "epoch": 195, "lr": 9.90394195278052e-05}
{"train_loss": 0.7127335071563721, "global_step": 17433, "epoch": 195, "lr": 9.903930643257322e-05}
{"train_loss": 0.5714548230171204, "global_step": 17434, "epoch": 195, "lr": 9.903919333074849e-05}
{"train_loss": 0.4537309408187866, "global_step": 17435, "epoch": 195, "lr": 9.903908022233104e-05}
{"train_loss": 0.5199938416481018, "global_step": 17436, "epoch": 195, "lr": 9.903896710732089e-05}
{"train_loss": 0.553882896900177, "global_step": 17437, "epoch": 195, "lr": 9.903885398571803e-05}
{"train_loss": 0.4132309854030609, "global_step": 17438, "epoch": 195, "lr": 9.903874085752253e-05}
{"train_loss": 0.5581762790679932, "global_step": 17439, "epoch": 195, "lr": 9.903862772273434e-05}
{"train_loss": 0.5837329626083374, "global_step": 17440, "epoch": 195, "lr": 9.903851458135352e-05}
{"train_loss": 0.6611783504486084, "global_step": 17441, "epoch": 195, "lr": 9.903840143338006e-05}
{"train_loss": 0.5825151205062866, "global_step": 17442, "epoch": 195, "lr": 9.903828827881398e-05}
{"train_loss": 0.5506630546591255, "global_step": 17443, "epoch": 195, "lr": 9.903817511765532e-05, "val_loss": 1.586652159690857, "train_action_mse_error": 29.014404296875}
{"train_loss": 0.48480328917503357, "global_step": 17444, "epoch": 196, "lr": 9.903806194990406e-05}
{"train_loss": 0.5358630418777466, "global_step": 17445, "epoch": 196, "lr": 9.903794877556023e-05}
{"train_loss": 0.529682457447052, "global_step": 17446, "epoch": 196, "lr": 9.903783559462386e-05}
{"train_loss": 0.6146320700645447, "global_step": 17447, "epoch": 196, "lr": 9.903772240709493e-05}
{"train_loss": 0.5323240756988525, "global_step": 17448, "epoch": 196, "lr": 9.903760921297348e-05}
{"train_loss": 0.45977169275283813, "global_step": 17449, "epoch": 196, "lr": 9.903749601225953e-05}
{"train_loss": 0.4482570290565491, "global_step": 17450, "epoch": 196, "lr": 9.903738280495307e-05}
{"train_loss": 0.596781849861145, "global_step": 17451, "epoch": 196, "lr": 9.903726959105415e-05}
{"train_loss": 0.5700519680976868, "global_step": 17452, "epoch": 196, "lr": 9.903715637056277e-05}
{"train_loss": 0.5115058422088623, "global_step": 17453, "epoch": 196, "lr": 9.903704314347892e-05}
{"train_loss": 0.4075479805469513, "global_step": 17454, "epoch": 196, "lr": 9.903692990980265e-05}
{"train_loss": 0.580227792263031, "global_step": 17455, "epoch": 196, "lr": 9.903681666953396e-05}
{"train_loss": 0.5576369166374207, "global_step": 17456, "epoch": 196, "lr": 9.903670342267287e-05}
{"train_loss": 0.5054607391357422, "global_step": 17457, "epoch": 196, "lr": 9.90365901692194e-05}
{"train_loss": 0.4968411326408386, "global_step": 17458, "epoch": 196, "lr": 9.903647690917355e-05}
{"train_loss": 0.5534980893135071, "global_step": 17459, "epoch": 196, "lr": 9.903636364253535e-05}
{"train_loss": 0.6001936793327332, "global_step": 17460, "epoch": 196, "lr": 9.903625036930479e-05}
{"train_loss": 0.48061856627464294, "global_step": 17461, "epoch": 196, "lr": 9.903613708948192e-05}
{"train_loss": 0.5660800337791443, "global_step": 17462, "epoch": 196, "lr": 9.903602380306674e-05}
{"train_loss": 0.4764834940433502, "global_step": 17463, "epoch": 196, "lr": 9.903591051005925e-05}
{"train_loss": 0.5857500433921814, "global_step": 17464, "epoch": 196, "lr": 9.903579721045949e-05}
{"train_loss": 0.5748405456542969, "global_step": 17465, "epoch": 196, "lr": 9.903568390426748e-05}
{"train_loss": 0.5863093137741089, "global_step": 17466, "epoch": 196, "lr": 9.903557059148319e-05}
{"train_loss": 0.5093198418617249, "global_step": 17467, "epoch": 196, "lr": 9.903545727210668e-05}
{"train_loss": 0.6224003434181213, "global_step": 17468, "epoch": 196, "lr": 9.903534394613794e-05}
{"train_loss": 0.515988826751709, "global_step": 17469, "epoch": 196, "lr": 9.903523061357701e-05}
{"train_loss": 0.4717283248901367, "global_step": 17470, "epoch": 196, "lr": 9.90351172744239e-05}
{"train_loss": 0.6825544834136963, "global_step": 17471, "epoch": 196, "lr": 9.903500392867859e-05}
{"train_loss": 0.6072800755500793, "global_step": 17472, "epoch": 196, "lr": 9.903489057634113e-05}
{"train_loss": 0.4717782735824585, "global_step": 17473, "epoch": 196, "lr": 9.903477721741154e-05}
{"train_loss": 0.6774589419364929, "global_step": 17474, "epoch": 196, "lr": 9.90346638518898e-05}
{"train_loss": 0.5398446917533875, "global_step": 17475, "epoch": 196, "lr": 9.903455047977597e-05}
{"train_loss": 0.5585560202598572, "global_step": 17476, "epoch": 196, "lr": 9.903443710107003e-05}
{"train_loss": 0.5564878582954407, "global_step": 17477, "epoch": 196, "lr": 9.9034323715772e-05}
{"train_loss": 0.5609922409057617, "global_step": 17478, "epoch": 196, "lr": 9.903421032388191e-05}
{"train_loss": 0.460903525352478, "global_step": 17479, "epoch": 196, "lr": 9.903409692539978e-05}
{"train_loss": 0.4561806619167328, "global_step": 17480, "epoch": 196, "lr": 9.90339835203256e-05}
{"train_loss": 0.5318055152893066, "global_step": 17481, "epoch": 196, "lr": 9.90338701086594e-05}
{"train_loss": 0.4791363477706909, "global_step": 17482, "epoch": 196, "lr": 9.90337566904012e-05}
{"train_loss": 0.4955397844314575, "global_step": 17483, "epoch": 196, "lr": 9.9033643265551e-05}
{"train_loss": 0.6057568788528442, "global_step": 17484, "epoch": 196, "lr": 9.903352983410882e-05}
{"train_loss": 0.4620189964771271, "global_step": 17485, "epoch": 196, "lr": 9.90334163960747e-05}
{"train_loss": 0.5749330520629883, "global_step": 17486, "epoch": 196, "lr": 9.903330295144861e-05}
{"train_loss": 0.6022035479545593, "global_step": 17487, "epoch": 196, "lr": 9.903318950023061e-05}
{"train_loss": 0.4719100892543793, "global_step": 17488, "epoch": 196, "lr": 9.903307604242069e-05}
{"train_loss": 0.4890306890010834, "global_step": 17489, "epoch": 196, "lr": 9.903296257801888e-05}
{"train_loss": 0.44782114028930664, "global_step": 17490, "epoch": 196, "lr": 9.903284910702517e-05}
{"train_loss": 0.4920969605445862, "global_step": 17491, "epoch": 196, "lr": 9.90327356294396e-05}
{"train_loss": 0.4709329307079315, "global_step": 17492, "epoch": 196, "lr": 9.903262214526217e-05}
{"train_loss": 0.4494059979915619, "global_step": 17493, "epoch": 196, "lr": 9.90325086544929e-05}
{"train_loss": 0.4745301306247711, "global_step": 17494, "epoch": 196, "lr": 9.903239515713181e-05}
{"train_loss": 0.6232548952102661, "global_step": 17495, "epoch": 196, "lr": 9.90322816531789e-05}
{"train_loss": 0.48293349146842957, "global_step": 17496, "epoch": 196, "lr": 9.903216814263421e-05}
{"train_loss": 0.46556442975997925, "global_step": 17497, "epoch": 196, "lr": 9.903205462549777e-05}
{"train_loss": 0.5216269493103027, "global_step": 17498, "epoch": 196, "lr": 9.903194110176952e-05}
{"train_loss": 0.6774516105651855, "global_step": 17499, "epoch": 196, "lr": 9.903182757144955e-05}
{"train_loss": 0.6725949645042419, "global_step": 17500, "epoch": 196, "lr": 9.903171403453783e-05}
{"train_loss": 0.6363348960876465, "global_step": 17501, "epoch": 196, "lr": 9.903160049103441e-05}
{"train_loss": 0.4647418260574341, "global_step": 17502, "epoch": 196, "lr": 9.903148694093929e-05}
{"train_loss": 0.44896671175956726, "global_step": 17503, "epoch": 196, "lr": 9.903137338425248e-05}
{"train_loss": 0.5797044038772583, "global_step": 17504, "epoch": 196, "lr": 9.9031259820974e-05}
{"train_loss": 0.5040582418441772, "global_step": 17505, "epoch": 196, "lr": 9.903114625110386e-05}
{"train_loss": 0.6535524129867554, "global_step": 17506, "epoch": 196, "lr": 9.90310326746421e-05}
{"train_loss": 0.6561172604560852, "global_step": 17507, "epoch": 196, "lr": 9.903091909158868e-05}
{"train_loss": 0.4172302782535553, "global_step": 17508, "epoch": 196, "lr": 9.903080550194368e-05}
{"train_loss": 0.38726088404655457, "global_step": 17509, "epoch": 196, "lr": 9.903069190570706e-05}
{"train_loss": 0.5252259373664856, "global_step": 17510, "epoch": 196, "lr": 9.903057830287887e-05}
{"train_loss": 0.4616098999977112, "global_step": 17511, "epoch": 196, "lr": 9.903046469345913e-05}
{"train_loss": 0.5725390911102295, "global_step": 17512, "epoch": 196, "lr": 9.903035107744784e-05}
{"train_loss": 0.5852625370025635, "global_step": 17513, "epoch": 196, "lr": 9.9030237454845e-05}
{"train_loss": 0.6287840008735657, "global_step": 17514, "epoch": 196, "lr": 9.903012382565067e-05}
{"train_loss": 0.5016829967498779, "global_step": 17515, "epoch": 196, "lr": 9.903001018986481e-05}
{"train_loss": 0.6608710289001465, "global_step": 17516, "epoch": 196, "lr": 9.902989654748747e-05}
{"train_loss": 0.5172135829925537, "global_step": 17517, "epoch": 196, "lr": 9.902978289851867e-05}
{"train_loss": 0.5368889570236206, "global_step": 17518, "epoch": 196, "lr": 9.90296692429584e-05}
{"train_loss": 0.6185243129730225, "global_step": 17519, "epoch": 196, "lr": 9.902955558080669e-05}
{"train_loss": 0.43334895372390747, "global_step": 17520, "epoch": 196, "lr": 9.902944191206356e-05}
{"train_loss": 0.41705283522605896, "global_step": 17521, "epoch": 196, "lr": 9.902932823672902e-05}
{"train_loss": 0.5110480785369873, "global_step": 17522, "epoch": 196, "lr": 9.902921455480307e-05}
{"train_loss": 0.5038455128669739, "global_step": 17523, "epoch": 196, "lr": 9.902910086628575e-05}
{"train_loss": 0.5511771440505981, "global_step": 17524, "epoch": 196, "lr": 9.902898717117707e-05}
{"train_loss": 0.5868229866027832, "global_step": 17525, "epoch": 196, "lr": 9.902887346947703e-05}
{"train_loss": 0.5255258679389954, "global_step": 17526, "epoch": 196, "lr": 9.902875976118566e-05}
{"train_loss": 0.5373720526695251, "global_step": 17527, "epoch": 196, "lr": 9.902864604630298e-05}
{"train_loss": 0.6198727488517761, "global_step": 17528, "epoch": 196, "lr": 9.902853232482898e-05}
{"train_loss": 0.4901582598686218, "global_step": 17529, "epoch": 196, "lr": 9.902841859676371e-05}
{"train_loss": 0.35898086428642273, "global_step": 17530, "epoch": 196, "lr": 9.902830486210714e-05}
{"train_loss": 0.5406162142753601, "global_step": 17531, "epoch": 196, "lr": 9.902819112085933e-05}
{"train_loss": 0.5331685459345914, "global_step": 17532, "epoch": 196, "lr": 9.902807737302028e-05, "val_loss": 1.6382497549057007}
{"train_loss": 0.40261897444725037, "global_step": 17533, "epoch": 197, "lr": 9.902796361858999e-05}
{"train_loss": 0.41092538833618164, "global_step": 17534, "epoch": 197, "lr": 9.90278498575685e-05}
{"train_loss": 0.4705188572406769, "global_step": 17535, "epoch": 197, "lr": 9.902773608995582e-05}
{"train_loss": 0.43431317806243896, "global_step": 17536, "epoch": 197, "lr": 9.902762231575193e-05}
{"train_loss": 0.4609273374080658, "global_step": 17537, "epoch": 197, "lr": 9.90275085349569e-05}
{"train_loss": 0.6420366764068604, "global_step": 17538, "epoch": 197, "lr": 9.90273947475707e-05}
{"train_loss": 0.45794105529785156, "global_step": 17539, "epoch": 197, "lr": 9.902728095359338e-05}
{"train_loss": 0.6843502521514893, "global_step": 17540, "epoch": 197, "lr": 9.902716715302494e-05}
{"train_loss": 0.4824204444885254, "global_step": 17541, "epoch": 197, "lr": 9.902705334586539e-05}
{"train_loss": 0.4292956292629242, "global_step": 17542, "epoch": 197, "lr": 9.902693953211474e-05}
{"train_loss": 0.4299752414226532, "global_step": 17543, "epoch": 197, "lr": 9.902682571177301e-05}
{"train_loss": 0.4978521764278412, "global_step": 17544, "epoch": 197, "lr": 9.902671188484024e-05}
{"train_loss": 0.6129038333892822, "global_step": 17545, "epoch": 197, "lr": 9.902659805131643e-05}
{"train_loss": 0.5121846795082092, "global_step": 17546, "epoch": 197, "lr": 9.902648421120158e-05}
{"train_loss": 0.5610542297363281, "global_step": 17547, "epoch": 197, "lr": 9.902637036449573e-05}
{"train_loss": 0.45691484212875366, "global_step": 17548, "epoch": 197, "lr": 9.902625651119886e-05}
{"train_loss": 0.6009756922721863, "global_step": 17549, "epoch": 197, "lr": 9.902614265131102e-05}
{"train_loss": 0.501154363155365, "global_step": 17550, "epoch": 197, "lr": 9.902602878483222e-05}
{"train_loss": 0.5258479118347168, "global_step": 17551, "epoch": 197, "lr": 9.902591491176245e-05}
{"train_loss": 0.6181614995002747, "global_step": 17552, "epoch": 197, "lr": 9.902580103210176e-05}
{"train_loss": 0.42330312728881836, "global_step": 17553, "epoch": 197, "lr": 9.902568714585014e-05}
{"train_loss": 0.5605647563934326, "global_step": 17554, "epoch": 197, "lr": 9.902557325300761e-05}
{"train_loss": 0.4171139597892761, "global_step": 17555, "epoch": 197, "lr": 9.90254593535742e-05}
{"train_loss": 0.6251179575920105, "global_step": 17556, "epoch": 197, "lr": 9.902534544754991e-05}
{"train_loss": 0.5497469305992126, "global_step": 17557, "epoch": 197, "lr": 9.902523153493476e-05}
{"train_loss": 0.4511505961418152, "global_step": 17558, "epoch": 197, "lr": 9.902511761572877e-05}
{"train_loss": 0.6111629009246826, "global_step": 17559, "epoch": 197, "lr": 9.902500368993194e-05}
{"train_loss": 0.6012492775917053, "global_step": 17560, "epoch": 197, "lr": 9.902488975754429e-05}
{"train_loss": 0.5395290851593018, "global_step": 17561, "epoch": 197, "lr": 9.902477581856586e-05}
{"train_loss": 0.5995609164237976, "global_step": 17562, "epoch": 197, "lr": 9.902466187299664e-05}
{"train_loss": 0.4918915629386902, "global_step": 17563, "epoch": 197, "lr": 9.902454792083664e-05}
{"train_loss": 0.4875822961330414, "global_step": 17564, "epoch": 197, "lr": 9.90244339620859e-05}
{"train_loss": 0.5880445837974548, "global_step": 17565, "epoch": 197, "lr": 9.902431999674441e-05}
{"train_loss": 0.6724685430526733, "global_step": 17566, "epoch": 197, "lr": 9.902420602481222e-05}
{"train_loss": 0.5316758751869202, "global_step": 17567, "epoch": 197, "lr": 9.902409204628931e-05}
{"train_loss": 0.5484670996665955, "global_step": 17568, "epoch": 197, "lr": 9.90239780611757e-05}
{"train_loss": 0.4584420323371887, "global_step": 17569, "epoch": 197, "lr": 9.902386406947143e-05}
{"train_loss": 0.6406378149986267, "global_step": 17570, "epoch": 197, "lr": 9.902375007117649e-05}
{"train_loss": 0.47923627495765686, "global_step": 17571, "epoch": 197, "lr": 9.902363606629091e-05}
{"train_loss": 0.4928589165210724, "global_step": 17572, "epoch": 197, "lr": 9.902352205481469e-05}
{"train_loss": 0.675390362739563, "global_step": 17573, "epoch": 197, "lr": 9.902340803674785e-05}
{"train_loss": 0.5792750716209412, "global_step": 17574, "epoch": 197, "lr": 9.902329401209043e-05}
{"train_loss": 0.4265139698982239, "global_step": 17575, "epoch": 197, "lr": 9.902317998084241e-05}
{"train_loss": 0.6728875637054443, "global_step": 17576, "epoch": 197, "lr": 9.902306594300383e-05}
{"train_loss": 0.48522984981536865, "global_step": 17577, "epoch": 197, "lr": 9.90229518985747e-05}
{"train_loss": 0.5587437152862549, "global_step": 17578, "epoch": 197, "lr": 9.902283784755502e-05}
{"train_loss": 0.4924905598163605, "global_step": 17579, "epoch": 197, "lr": 9.902272378994481e-05}
{"train_loss": 0.6155756115913391, "global_step": 17580, "epoch": 197, "lr": 9.902260972574411e-05}
{"train_loss": 0.4918873906135559, "global_step": 17581, "epoch": 197, "lr": 9.902249565495291e-05}
{"train_loss": 0.4958352744579315, "global_step": 17582, "epoch": 197, "lr": 9.902238157757125e-05}
{"train_loss": 0.5358467102050781, "global_step": 17583, "epoch": 197, "lr": 9.90222674935991e-05}
{"train_loss": 0.4610753059387207, "global_step": 17584, "epoch": 197, "lr": 9.90221534030365e-05}
{"train_loss": 0.49396824836730957, "global_step": 17585, "epoch": 197, "lr": 9.90220393058835e-05}
{"train_loss": 0.5583405494689941, "global_step": 17586, "epoch": 197, "lr": 9.902192520214007e-05}
{"train_loss": 0.49089452624320984, "global_step": 17587, "epoch": 197, "lr": 9.902181109180624e-05}
{"train_loss": 0.5918183922767639, "global_step": 17588, "epoch": 197, "lr": 9.902169697488202e-05}
{"train_loss": 0.5828747153282166, "global_step": 17589, "epoch": 197, "lr": 9.902158285136744e-05}
{"train_loss": 0.6130722165107727, "global_step": 17590, "epoch": 197, "lr": 9.90214687212625e-05}
{"train_loss": 0.8145819306373596, "global_step": 17591, "epoch": 197, "lr": 9.902135458456723e-05}
{"train_loss": 0.48734793066978455, "global_step": 17592, "epoch": 197, "lr": 9.902124044128163e-05}
{"train_loss": 0.4848824739456177, "global_step": 17593, "epoch": 197, "lr": 9.90211262914057e-05}
{"train_loss": 0.5838358998298645, "global_step": 17594, "epoch": 197, "lr": 9.902101213493951e-05}
{"train_loss": 0.49945780634880066, "global_step": 17595, "epoch": 197, "lr": 9.902089797188303e-05}
{"train_loss": 0.5720186829566956, "global_step": 17596, "epoch": 197, "lr": 9.902078380223627e-05}
{"train_loss": 0.6479436159133911, "global_step": 17597, "epoch": 197, "lr": 9.90206696259993e-05}
{"train_loss": 0.4731399416923523, "global_step": 17598, "epoch": 197, "lr": 9.902055544317207e-05}
{"train_loss": 0.513297438621521, "global_step": 17599, "epoch": 197, "lr": 9.902044125375463e-05}
{"train_loss": 0.6170609593391418, "global_step": 17600, "epoch": 197, "lr": 9.9020327057747e-05}
{"train_loss": 0.5992285013198853, "global_step": 17601, "epoch": 197, "lr": 9.902021285514917e-05}
{"train_loss": 0.44241222739219666, "global_step": 17602, "epoch": 197, "lr": 9.902009864596116e-05}
{"train_loss": 0.4526187479496002, "global_step": 17603, "epoch": 197, "lr": 9.901998443018302e-05}
{"train_loss": 0.6038748621940613, "global_step": 17604, "epoch": 197, "lr": 9.901987020781474e-05}
{"train_loss": 0.547386884689331, "global_step": 17605, "epoch": 197, "lr": 9.901975597885632e-05}
{"train_loss": 0.6385114789009094, "global_step": 17606, "epoch": 197, "lr": 9.90196417433078e-05}
{"train_loss": 0.5037975907325745, "global_step": 17607, "epoch": 197, "lr": 9.901952750116919e-05}
{"train_loss": 0.43943271040916443, "global_step": 17608, "epoch": 197, "lr": 9.901941325244048e-05}
{"train_loss": 0.6342827081680298, "global_step": 17609, "epoch": 197, "lr": 9.901929899712173e-05}
{"train_loss": 0.45488959550857544, "global_step": 17610, "epoch": 197, "lr": 9.901918473521293e-05}
{"train_loss": 0.47026991844177246, "global_step": 17611, "epoch": 197, "lr": 9.90190704667141e-05}
{"train_loss": 0.5743458271026611, "global_step": 17612, "epoch": 197, "lr": 9.901895619162525e-05}
{"train_loss": 0.5327643752098083, "global_step": 17613, "epoch": 197, "lr": 9.901884190994639e-05}
{"train_loss": 0.562352180480957, "global_step": 17614, "epoch": 197, "lr": 9.901872762167756e-05}
{"train_loss": 0.758604109287262, "global_step": 17615, "epoch": 197, "lr": 9.901861332681874e-05}
{"train_loss": 0.5477054119110107, "global_step": 17616, "epoch": 197, "lr": 9.901849902536998e-05}
{"train_loss": 0.5488705635070801, "global_step": 17617, "epoch": 197, "lr": 9.901838471733127e-05}
{"train_loss": 0.4540286362171173, "global_step": 17618, "epoch": 197, "lr": 9.901827040270264e-05}
{"train_loss": 0.5429121851921082, "global_step": 17619, "epoch": 197, "lr": 9.90181560814841e-05}
{"train_loss": 0.5522986054420471, "global_step": 17620, "epoch": 197, "lr": 9.901804175367568e-05}
{"train_loss": 0.5370524910728584, "global_step": 17621, "epoch": 197, "lr": 9.901792741927737e-05, "val_loss": 1.5799771547317505}
{"train_loss": 0.5106703042984009, "global_step": 17622, "epoch": 198, "lr": 9.901781307828919e-05}
{"train_loss": 0.5738945007324219, "global_step": 17623, "epoch": 198, "lr": 9.901769873071118e-05}
{"train_loss": 0.5428510904312134, "global_step": 17624, "epoch": 198, "lr": 9.901758437654332e-05}
{"train_loss": 0.7061429619789124, "global_step": 17625, "epoch": 198, "lr": 9.901747001578565e-05}
{"train_loss": 0.5772840976715088, "global_step": 17626, "epoch": 198, "lr": 9.90173556484382e-05}
{"train_loss": 0.5706998109817505, "global_step": 17627, "epoch": 198, "lr": 9.901724127450094e-05}
{"train_loss": 0.5873245000839233, "global_step": 17628, "epoch": 198, "lr": 9.901712689397392e-05}
{"train_loss": 0.443012535572052, "global_step": 17629, "epoch": 198, "lr": 9.901701250685713e-05}
{"train_loss": 0.5664275884628296, "global_step": 17630, "epoch": 198, "lr": 9.90168981131506e-05}
{"train_loss": 0.7037991285324097, "global_step": 17631, "epoch": 198, "lr": 9.901678371285435e-05}
{"train_loss": 0.4670812487602234, "global_step": 17632, "epoch": 198, "lr": 9.90166693059684e-05}
{"train_loss": 0.5640875697135925, "global_step": 17633, "epoch": 198, "lr": 9.901655489249276e-05}
{"train_loss": 0.44431212544441223, "global_step": 17634, "epoch": 198, "lr": 9.901644047242741e-05}
{"train_loss": 0.5712186098098755, "global_step": 17635, "epoch": 198, "lr": 9.901632604577243e-05}
{"train_loss": 0.649276852607727, "global_step": 17636, "epoch": 198, "lr": 9.901621161252778e-05}
{"train_loss": 0.6320140361785889, "global_step": 17637, "epoch": 198, "lr": 9.901609717269352e-05}
{"train_loss": 0.5004863739013672, "global_step": 17638, "epoch": 198, "lr": 9.901598272626964e-05}
{"train_loss": 0.5383108258247375, "global_step": 17639, "epoch": 198, "lr": 9.901586827325615e-05}
{"train_loss": 0.46687665581703186, "global_step": 17640, "epoch": 198, "lr": 9.901575381365307e-05}
{"train_loss": 0.7364722490310669, "global_step": 17641, "epoch": 198, "lr": 9.901563934746043e-05}
{"train_loss": 0.6372257471084595, "global_step": 17642, "epoch": 198, "lr": 9.901552487467823e-05}
{"train_loss": 0.49122941493988037, "global_step": 17643, "epoch": 198, "lr": 9.901541039530648e-05}
{"train_loss": 0.4428265392780304, "global_step": 17644, "epoch": 198, "lr": 9.901529590934522e-05}
{"train_loss": 0.637392520904541, "global_step": 17645, "epoch": 198, "lr": 9.901518141679444e-05}
{"train_loss": 0.44591522216796875, "global_step": 17646, "epoch": 198, "lr": 9.901506691765417e-05}
{"train_loss": 0.546471357345581, "global_step": 17647, "epoch": 198, "lr": 9.901495241192443e-05}
{"train_loss": 0.3868511915206909, "global_step": 17648, "epoch": 198, "lr": 9.901483789960522e-05}
{"train_loss": 0.6644569039344788, "global_step": 17649, "epoch": 198, "lr": 9.901472338069657e-05}
{"train_loss": 0.5632416009902954, "global_step": 17650, "epoch": 198, "lr": 9.901460885519847e-05}
{"train_loss": 0.5159209966659546, "global_step": 17651, "epoch": 198, "lr": 9.901449432311098e-05}
{"train_loss": 0.483662873506546, "global_step": 17652, "epoch": 198, "lr": 9.901437978443407e-05}
{"train_loss": 0.5217933654785156, "global_step": 17653, "epoch": 198, "lr": 9.901426523916777e-05}
{"train_loss": 0.548204779624939, "global_step": 17654, "epoch": 198, "lr": 9.90141506873121e-05}
{"train_loss": 0.4586327373981476, "global_step": 17655, "epoch": 198, "lr": 9.901403612886708e-05}
{"train_loss": 0.5444180369377136, "global_step": 17656, "epoch": 198, "lr": 9.901392156383273e-05}
{"train_loss": 0.6400616765022278, "global_step": 17657, "epoch": 198, "lr": 9.901380699220904e-05}
{"train_loss": 0.5511917471885681, "global_step": 17658, "epoch": 198, "lr": 9.901369241399605e-05}
{"train_loss": 0.591328501701355, "global_step": 17659, "epoch": 198, "lr": 9.901357782919378e-05}
{"train_loss": 0.4762001037597656, "global_step": 17660, "epoch": 198, "lr": 9.901346323780221e-05}
{"train_loss": 0.5074408650398254, "global_step": 17661, "epoch": 198, "lr": 9.901334863982138e-05}
{"train_loss": 0.49171629548072815, "global_step": 17662, "epoch": 198, "lr": 9.901323403525131e-05}
{"train_loss": 0.48863449692726135, "global_step": 17663, "epoch": 198, "lr": 9.901311942409199e-05}
{"train_loss": 0.41392630338668823, "global_step": 17664, "epoch": 198, "lr": 9.901300480634348e-05}
{"train_loss": 0.5364377498626709, "global_step": 17665, "epoch": 198, "lr": 9.901289018200575e-05}
{"train_loss": 0.5449848771095276, "global_step": 17666, "epoch": 198, "lr": 9.901277555107884e-05}
{"train_loss": 0.5111801624298096, "global_step": 17667, "epoch": 198, "lr": 9.901266091356276e-05}
{"train_loss": 0.6004058718681335, "global_step": 17668, "epoch": 198, "lr": 9.901254626945753e-05}
{"train_loss": 0.4790680706501007, "global_step": 17669, "epoch": 198, "lr": 9.901243161876316e-05}
{"train_loss": 0.6273696422576904, "global_step": 17670, "epoch": 198, "lr": 9.901231696147965e-05}
{"train_loss": 0.38415613770484924, "global_step": 17671, "epoch": 198, "lr": 9.901220229760706e-05}
{"train_loss": 0.3408966362476349, "global_step": 17672, "epoch": 198, "lr": 9.901208762714536e-05}
{"train_loss": 0.5786349773406982, "global_step": 17673, "epoch": 198, "lr": 9.901197295009457e-05}
{"train_loss": 0.5380875468254089, "global_step": 17674, "epoch": 198, "lr": 9.901185826645474e-05}
{"train_loss": 0.5518823266029358, "global_step": 17675, "epoch": 198, "lr": 9.901174357622585e-05}
{"train_loss": 0.48962023854255676, "global_step": 17676, "epoch": 198, "lr": 9.901162887940793e-05}
{"train_loss": 0.432443767786026, "global_step": 17677, "epoch": 198, "lr": 9.9011514176001e-05}
{"train_loss": 0.6329834461212158, "global_step": 17678, "epoch": 198, "lr": 9.901139946600507e-05}
{"train_loss": 0.612442672252655, "global_step": 17679, "epoch": 198, "lr": 9.901128474942015e-05}
{"train_loss": 0.5301222801208496, "global_step": 17680, "epoch": 198, "lr": 9.901117002624625e-05}
{"train_loss": 0.4873173236846924, "global_step": 17681, "epoch": 198, "lr": 9.901105529648341e-05}
{"train_loss": 0.5742190480232239, "global_step": 17682, "epoch": 198, "lr": 9.901094056013163e-05}
{"train_loss": 0.5566743016242981, "global_step": 17683, "epoch": 198, "lr": 9.901082581719092e-05}
{"train_loss": 0.6362702250480652, "global_step": 17684, "epoch": 198, "lr": 9.90107110676613e-05}
{"train_loss": 0.684619128704071, "global_step": 17685, "epoch": 198, "lr": 9.901059631154279e-05}
{"train_loss": 0.46482887864112854, "global_step": 17686, "epoch": 198, "lr": 9.901048154883541e-05}
{"train_loss": 0.6205245852470398, "global_step": 17687, "epoch": 198, "lr": 9.901036677953917e-05}
{"train_loss": 0.5157616138458252, "global_step": 17688, "epoch": 198, "lr": 9.901025200365407e-05}
{"train_loss": 0.5826955437660217, "global_step": 17689, "epoch": 198, "lr": 9.901013722118014e-05}
{"train_loss": 0.4412378668785095, "global_step": 17690, "epoch": 198, "lr": 9.90100224321174e-05}
{"train_loss": 0.3774757981300354, "global_step": 17691, "epoch": 198, "lr": 9.900990763646586e-05}
{"train_loss": 0.42004117369651794, "global_step": 17692, "epoch": 198, "lr": 9.900979283422553e-05}
{"train_loss": 0.6277917623519897, "global_step": 17693, "epoch": 198, "lr": 9.900967802539645e-05}
{"train_loss": 0.42017313838005066, "global_step": 17694, "epoch": 198, "lr": 9.900956320997859e-05}
{"train_loss": 0.5938811898231506, "global_step": 17695, "epoch": 198, "lr": 9.9009448387972e-05}
{"train_loss": 0.4779777526855469, "global_step": 17696, "epoch": 198, "lr": 9.90093335593767e-05}
{"train_loss": 0.5454645156860352, "global_step": 17697, "epoch": 198, "lr": 9.900921872419267e-05}
{"train_loss": 0.5601443648338318, "global_step": 17698, "epoch": 198, "lr": 9.900910388241997e-05}
{"train_loss": 0.5045454502105713, "global_step": 17699, "epoch": 198, "lr": 9.900898903405857e-05}
{"train_loss": 0.5756931304931641, "global_step": 17700, "epoch": 198, "lr": 9.900887417910853e-05}
{"train_loss": 0.614378809928894, "global_step": 17701, "epoch": 198, "lr": 9.900875931756984e-05}
{"train_loss": 0.677047610282898, "global_step": 17702, "epoch": 198, "lr": 9.900864444944251e-05}
{"train_loss": 0.6406832933425903, "global_step": 17703, "epoch": 198, "lr": 9.900852957472657e-05}
{"train_loss": 0.6639951467514038, "global_step": 17704, "epoch": 198, "lr": 9.900841469342204e-05}
{"train_loss": 0.7130091190338135, "global_step": 17705, "epoch": 198, "lr": 9.900829980552893e-05}
{"train_loss": 0.5514339804649353, "global_step": 17706, "epoch": 198, "lr": 9.900818491104722e-05}
{"train_loss": 0.44167739152908325, "global_step": 17707, "epoch": 198, "lr": 9.900807000997698e-05}
{"train_loss": 0.6132674217224121, "global_step": 17708, "epoch": 198, "lr": 9.90079551023182e-05}
{"train_loss": 0.5356237888336182, "global_step": 17709, "epoch": 198, "lr": 9.90078401880709e-05}
{"train_loss": 0.5441293103641338, "global_step": 17710, "epoch": 198, "lr": 9.900772526723509e-05, "val_loss": 1.5732342004776}
{"train_loss": 0.5173202157020569, "global_step": 17711, "epoch": 199, "lr": 9.900761033981079e-05}
{"train_loss": 0.494378000497818, "global_step": 17712, "epoch": 199, "lr": 9.900749540579802e-05}
{"train_loss": 0.549505352973938, "global_step": 17713, "epoch": 199, "lr": 9.900738046519678e-05}
{"train_loss": 0.4799617826938629, "global_step": 17714, "epoch": 199, "lr": 9.900726551800709e-05}
{"train_loss": 0.6172738671302795, "global_step": 17715, "epoch": 199, "lr": 9.900715056422898e-05}
{"train_loss": 0.5103576183319092, "global_step": 17716, "epoch": 199, "lr": 9.900703560386244e-05}
{"train_loss": 0.7181068658828735, "global_step": 17717, "epoch": 199, "lr": 9.900692063690752e-05}
{"train_loss": 0.5314800143241882, "global_step": 17718, "epoch": 199, "lr": 9.900680566336421e-05}
{"train_loss": 0.5735695362091064, "global_step": 17719, "epoch": 199, "lr": 9.900669068323254e-05}
{"train_loss": 0.4562109708786011, "global_step": 17720, "epoch": 199, "lr": 9.900657569651251e-05}
{"train_loss": 0.4463176727294922, "global_step": 17721, "epoch": 199, "lr": 9.900646070320414e-05}
{"train_loss": 0.4952002167701721, "global_step": 17722, "epoch": 199, "lr": 9.900634570330746e-05}
{"train_loss": 0.4820386469364166, "global_step": 17723, "epoch": 199, "lr": 9.900623069682247e-05}
{"train_loss": 0.6116971969604492, "global_step": 17724, "epoch": 199, "lr": 9.900611568374918e-05}
{"train_loss": 0.6230917572975159, "global_step": 17725, "epoch": 199, "lr": 9.900600066408763e-05}
{"train_loss": 0.6008411049842834, "global_step": 17726, "epoch": 199, "lr": 9.90058856378378e-05}
{"train_loss": 0.3806414008140564, "global_step": 17727, "epoch": 199, "lr": 9.900577060499974e-05}
{"train_loss": 0.4862576425075531, "global_step": 17728, "epoch": 199, "lr": 9.900565556557346e-05}
{"train_loss": 0.429057240486145, "global_step": 17729, "epoch": 199, "lr": 9.900554051955895e-05}
{"train_loss": 0.6439245343208313, "global_step": 17730, "epoch": 199, "lr": 9.900542546695625e-05}
{"train_loss": 0.5326464176177979, "global_step": 17731, "epoch": 199, "lr": 9.900531040776537e-05}
{"train_loss": 0.527466893196106, "global_step": 17732, "epoch": 199, "lr": 9.900519534198631e-05}
{"train_loss": 0.49021005630493164, "global_step": 17733, "epoch": 199, "lr": 9.900508026961912e-05}
{"train_loss": 0.5901888012886047, "global_step": 17734, "epoch": 199, "lr": 9.900496519066378e-05}
{"train_loss": 0.548143744468689, "global_step": 17735, "epoch": 199, "lr": 9.900485010512033e-05}
{"train_loss": 0.5518430471420288, "global_step": 17736, "epoch": 199, "lr": 9.900473501298876e-05}
{"train_loss": 0.38834553956985474, "global_step": 17737, "epoch": 199, "lr": 9.90046199142691e-05}
{"train_loss": 0.5849071741104126, "global_step": 17738, "epoch": 199, "lr": 9.900450480896139e-05}
{"train_loss": 0.627481997013092, "global_step": 17739, "epoch": 199, "lr": 9.900438969706561e-05}
{"train_loss": 0.5161364674568176, "global_step": 17740, "epoch": 199, "lr": 9.900427457858177e-05}
{"train_loss": 0.46716317534446716, "global_step": 17741, "epoch": 199, "lr": 9.900415945350991e-05}
{"train_loss": 0.6611935496330261, "global_step": 17742, "epoch": 199, "lr": 9.900404432185005e-05}
{"train_loss": 0.6163327693939209, "global_step": 17743, "epoch": 199, "lr": 9.90039291836022e-05}
{"train_loss": 0.45382267236709595, "global_step": 17744, "epoch": 199, "lr": 9.900381403876635e-05}
{"train_loss": 0.5772174596786499, "global_step": 17745, "epoch": 199, "lr": 9.900369888734253e-05}
{"train_loss": 0.5847108960151672, "global_step": 17746, "epoch": 199, "lr": 9.900358372933078e-05}
{"train_loss": 0.6196601390838623, "global_step": 17747, "epoch": 199, "lr": 9.900346856473108e-05}
{"train_loss": 0.5657233595848083, "global_step": 17748, "epoch": 199, "lr": 9.900335339354347e-05}
{"train_loss": 0.4611091613769531, "global_step": 17749, "epoch": 199, "lr": 9.900323821576795e-05}
{"train_loss": 0.6819542646408081, "global_step": 17750, "epoch": 199, "lr": 9.900312303140455e-05}
{"train_loss": 0.4394761621952057, "global_step": 17751, "epoch": 199, "lr": 9.900300784045328e-05}
{"train_loss": 0.5226973295211792, "global_step": 17752, "epoch": 199, "lr": 9.900289264291415e-05}
{"train_loss": 0.4784756302833557, "global_step": 17753, "epoch": 199, "lr": 9.900277743878716e-05}
{"train_loss": 0.42777276039123535, "global_step": 17754, "epoch": 199, "lr": 9.900266222807237e-05}
{"train_loss": 0.6201249361038208, "global_step": 17755, "epoch": 199, "lr": 9.900254701076975e-05}
{"train_loss": 0.5330637097358704, "global_step": 17756, "epoch": 199, "lr": 9.900243178687935e-05}
{"train_loss": 0.494266539812088, "global_step": 17757, "epoch": 199, "lr": 9.900231655640115e-05}
{"train_loss": 0.5310198664665222, "global_step": 17758, "epoch": 199, "lr": 9.900220131933521e-05}
{"train_loss": 0.5278639793395996, "global_step": 17759, "epoch": 199, "lr": 9.90020860756815e-05}
{"train_loss": 0.5835946202278137, "global_step": 17760, "epoch": 199, "lr": 9.900197082544007e-05}
{"train_loss": 0.6051886081695557, "global_step": 17761, "epoch": 199, "lr": 9.900185556861092e-05}
{"train_loss": 0.4996950626373291, "global_step": 17762, "epoch": 199, "lr": 9.900174030519407e-05}
{"train_loss": 0.7015081644058228, "global_step": 17763, "epoch": 199, "lr": 9.900162503518952e-05}
{"train_loss": 0.5466325283050537, "global_step": 17764, "epoch": 199, "lr": 9.900150975859732e-05}
{"train_loss": 0.5806360244750977, "global_step": 17765, "epoch": 199, "lr": 9.900139447541746e-05}
{"train_loss": 0.5939778089523315, "global_step": 17766, "epoch": 199, "lr": 9.900127918564994e-05}
{"train_loss": 0.718705415725708, "global_step": 17767, "epoch": 199, "lr": 9.900116388929482e-05}
{"train_loss": 0.5207443237304688, "global_step": 17768, "epoch": 199, "lr": 9.900104858635207e-05}
{"train_loss": 0.5070375204086304, "global_step": 17769, "epoch": 199, "lr": 9.900093327682174e-05}
{"train_loss": 0.6177070140838623, "global_step": 17770, "epoch": 199, "lr": 9.900081796070382e-05}
{"train_loss": 0.49375659227371216, "global_step": 17771, "epoch": 199, "lr": 9.900070263799834e-05}
{"train_loss": 0.5312968492507935, "global_step": 17772, "epoch": 199, "lr": 9.900058730870531e-05}
{"train_loss": 0.6563858985900879, "global_step": 17773, "epoch": 199, "lr": 9.900047197282478e-05}
{"train_loss": 0.6444336771965027, "global_step": 17774, "epoch": 199, "lr": 9.90003566303567e-05}
{"train_loss": 0.6109439730644226, "global_step": 17775, "epoch": 199, "lr": 9.900024128130113e-05}
{"train_loss": 0.5725752711296082, "global_step": 17776, "epoch": 199, "lr": 9.900012592565807e-05}
{"train_loss": 0.5924479961395264, "global_step": 17777, "epoch": 199, "lr": 9.900001056342754e-05}
{"train_loss": 0.3962804079055786, "global_step": 17778, "epoch": 199, "lr": 9.899989519460955e-05}
{"train_loss": 0.4052931070327759, "global_step": 17779, "epoch": 199, "lr": 9.899977981920414e-05}
{"train_loss": 0.6031665205955505, "global_step": 17780, "epoch": 199, "lr": 9.89996644372113e-05}
{"train_loss": 0.6088383197784424, "global_step": 17781, "epoch": 199, "lr": 9.899954904863104e-05}
{"train_loss": 0.6676353812217712, "global_step": 17782, "epoch": 199, "lr": 9.89994336534634e-05}
{"train_loss": 0.6238986253738403, "global_step": 17783, "epoch": 199, "lr": 9.899931825170839e-05}
{"train_loss": 0.6440244317054749, "global_step": 17784, "epoch": 199, "lr": 9.899920284336601e-05}
{"train_loss": 0.4781838059425354, "global_step": 17785, "epoch": 199, "lr": 9.899908742843628e-05}
{"train_loss": 0.47523000836372375, "global_step": 17786, "epoch": 199, "lr": 9.899897200691922e-05}
{"train_loss": 0.700819730758667, "global_step": 17787, "epoch": 199, "lr": 9.899885657881485e-05}
{"train_loss": 0.4595887064933777, "global_step": 17788, "epoch": 199, "lr": 9.899874114412318e-05}
{"train_loss": 0.6132667660713196, "global_step": 17789, "epoch": 199, "lr": 9.899862570284422e-05}
{"train_loss": 0.5252124667167664, "global_step": 17790, "epoch": 199, "lr": 9.899851025497802e-05}
{"train_loss": 0.6242145299911499, "global_step": 17791, "epoch": 199, "lr": 9.899839480052452e-05}
{"train_loss": 0.44889476895332336, "global_step": 17792, "epoch": 199, "lr": 9.899827933948383e-05}
{"train_loss": 0.5986223220825195, "global_step": 17793, "epoch": 199, "lr": 9.89981638718559e-05}
{"train_loss": 0.539242148399353, "global_step": 17794, "epoch": 199, "lr": 9.899804839764076e-05}
{"train_loss": 0.6674084067344666, "global_step": 17795, "epoch": 199, "lr": 9.899793291683843e-05}
{"train_loss": 0.4908114969730377, "global_step": 17796, "epoch": 199, "lr": 9.899781742944892e-05}
{"train_loss": 0.5205778479576111, "global_step": 17797, "epoch": 199, "lr": 9.899770193547226e-05}
{"train_loss": 0.48286962509155273, "global_step": 17798, "epoch": 199, "lr": 9.899758643490845e-05}
{"train_loss": 0.5493485529101296, "global_step": 17799, "epoch": 199, "lr": 9.899747092775752e-05, "val_loss": 1.5842770338058472}
{"train_loss": 0.41908058524131775, "global_step": 17800, "epoch": 200, "lr": 9.899735541401946e-05}
{"train_loss": 0.46619918942451477, "global_step": 17801, "epoch": 200, "lr": 9.899723989369433e-05}
{"train_loss": 0.3694590926170349, "global_step": 17802, "epoch": 200, "lr": 9.89971243667821e-05}
{"train_loss": 0.5638557076454163, "global_step": 17803, "epoch": 200, "lr": 9.899700883328281e-05}
{"train_loss": 0.49494025111198425, "global_step": 17804, "epoch": 200, "lr": 9.899689329319645e-05}
{"train_loss": 0.5135818719863892, "global_step": 17805, "epoch": 200, "lr": 9.899677774652308e-05}
{"train_loss": 0.5460548400878906, "global_step": 17806, "epoch": 200, "lr": 9.899666219326269e-05}
{"train_loss": 0.6820642948150635, "global_step": 17807, "epoch": 200, "lr": 9.899654663341529e-05}
{"train_loss": 0.5133375525474548, "global_step": 17808, "epoch": 200, "lr": 9.899643106698088e-05}
{"train_loss": 0.5496483445167542, "global_step": 17809, "epoch": 200, "lr": 9.899631549395953e-05}
{"train_loss": 0.5299686789512634, "global_step": 17810, "epoch": 200, "lr": 9.899619991435121e-05}
{"train_loss": 0.4286860227584839, "global_step": 17811, "epoch": 200, "lr": 9.899608432815595e-05}
{"train_loss": 0.4538620710372925, "global_step": 17812, "epoch": 200, "lr": 9.899596873537376e-05}
{"train_loss": 0.571271538734436, "global_step": 17813, "epoch": 200, "lr": 9.899585313600465e-05}
{"train_loss": 0.507016122341156, "global_step": 17814, "epoch": 200, "lr": 9.899573753004865e-05}
{"train_loss": 0.4672144055366516, "global_step": 17815, "epoch": 200, "lr": 9.899562191750577e-05}
{"train_loss": 0.3894868791103363, "global_step": 17816, "epoch": 200, "lr": 9.899550629837604e-05}
{"train_loss": 0.3939383029937744, "global_step": 17817, "epoch": 200, "lr": 9.899539067265946e-05}
{"train_loss": 0.42641782760620117, "global_step": 17818, "epoch": 200, "lr": 9.899527504035603e-05}
{"train_loss": 0.6113649010658264, "global_step": 17819, "epoch": 200, "lr": 9.899515940146579e-05}
{"train_loss": 0.4938451051712036, "global_step": 17820, "epoch": 200, "lr": 9.899504375598876e-05}
{"train_loss": 0.5516937375068665, "global_step": 17821, "epoch": 200, "lr": 9.899492810392493e-05}
{"train_loss": 0.4027312994003296, "global_step": 17822, "epoch": 200, "lr": 9.899481244527433e-05}
{"train_loss": 0.4797111451625824, "global_step": 17823, "epoch": 200, "lr": 9.899469678003698e-05}
{"train_loss": 0.4812542200088501, "global_step": 17824, "epoch": 200, "lr": 9.89945811082129e-05}
{"train_loss": 0.4940124452114105, "global_step": 17825, "epoch": 200, "lr": 9.899446542980206e-05}
{"train_loss": 0.4573667347431183, "global_step": 17826, "epoch": 200, "lr": 9.899434974480454e-05}
{"train_loss": 0.5584507584571838, "global_step": 17827, "epoch": 200, "lr": 9.899423405322032e-05}
{"train_loss": 0.4485037624835968, "global_step": 17828, "epoch": 200, "lr": 9.899411835504942e-05}
{"train_loss": 0.4876173138618469, "global_step": 17829, "epoch": 200, "lr": 9.899400265029186e-05}
{"train_loss": 0.6426166296005249, "global_step": 17830, "epoch": 200, "lr": 9.899388693894765e-05}
{"train_loss": 0.5497469305992126, "global_step": 17831, "epoch": 200, "lr": 9.899377122101682e-05}
{"train_loss": 0.5849816203117371, "global_step": 17832, "epoch": 200, "lr": 9.899365549649937e-05}
{"train_loss": 0.5954008102416992, "global_step": 17833, "epoch": 200, "lr": 9.899353976539531e-05}
{"train_loss": 0.5466097593307495, "global_step": 17834, "epoch": 200, "lr": 9.899342402770468e-05}
{"train_loss": 0.5605771541595459, "global_step": 17835, "epoch": 200, "lr": 9.899330828342748e-05}
{"train_loss": 0.4561672806739807, "global_step": 17836, "epoch": 200, "lr": 9.899319253256372e-05}
{"train_loss": 0.6794795989990234, "global_step": 17837, "epoch": 200, "lr": 9.899307677511343e-05}
{"train_loss": 0.45377489924430847, "global_step": 17838, "epoch": 200, "lr": 9.89929610110766e-05}
{"train_loss": 0.4820198714733124, "global_step": 17839, "epoch": 200, "lr": 9.899284524045328e-05}
{"train_loss": 0.5338397026062012, "global_step": 17840, "epoch": 200, "lr": 9.899272946324348e-05}
{"train_loss": 0.486467570066452, "global_step": 17841, "epoch": 200, "lr": 9.899261367944717e-05}
{"train_loss": 0.4710010588169098, "global_step": 17842, "epoch": 200, "lr": 9.899249788906443e-05}
{"train_loss": 0.6798450350761414, "global_step": 17843, "epoch": 200, "lr": 9.899238209209523e-05}
{"train_loss": 0.38374343514442444, "global_step": 17844, "epoch": 200, "lr": 9.899226628853962e-05}
{"train_loss": 0.5499025583267212, "global_step": 17845, "epoch": 200, "lr": 9.899215047839757e-05}
{"train_loss": 0.4711996018886566, "global_step": 17846, "epoch": 200, "lr": 9.899203466166914e-05}
{"train_loss": 0.5441694259643555, "global_step": 17847, "epoch": 200, "lr": 9.899191883835432e-05}
{"train_loss": 0.6605182886123657, "global_step": 17848, "epoch": 200, "lr": 9.899180300845315e-05}
{"train_loss": 0.5753055214881897, "global_step": 17849, "epoch": 200, "lr": 9.899168717196561e-05}
{"train_loss": 0.5052942633628845, "global_step": 17850, "epoch": 200, "lr": 9.899157132889174e-05}
{"train_loss": 0.5974730849266052, "global_step": 17851, "epoch": 200, "lr": 9.899145547923155e-05}
{"train_loss": 0.49696141481399536, "global_step": 17852, "epoch": 200, "lr": 9.899133962298506e-05}
{"train_loss": 0.3581252694129944, "global_step": 17853, "epoch": 200, "lr": 9.899122376015228e-05}
{"train_loss": 0.523126482963562, "global_step": 17854, "epoch": 200, "lr": 9.899110789073323e-05}
{"train_loss": 0.4471728205680847, "global_step": 17855, "epoch": 200, "lr": 9.899099201472792e-05}
{"train_loss": 0.5213804244995117, "global_step": 17856, "epoch": 200, "lr": 9.899087613213636e-05}
{"train_loss": 0.5763491988182068, "global_step": 17857, "epoch": 200, "lr": 9.899076024295859e-05}
{"train_loss": 0.5521573424339294, "global_step": 17858, "epoch": 200, "lr": 9.89906443471946e-05}
{"train_loss": 0.5474055409431458, "global_step": 17859, "epoch": 200, "lr": 9.899052844484443e-05}
{"train_loss": 0.4755403399467468, "global_step": 17860, "epoch": 200, "lr": 9.899041253590806e-05}
{"train_loss": 0.5222630500793457, "global_step": 17861, "epoch": 200, "lr": 9.899029662038552e-05}
{"train_loss": 0.5013318061828613, "global_step": 17862, "epoch": 200, "lr": 9.899018069827686e-05}
{"train_loss": 0.397988498210907, "global_step": 17863, "epoch": 200, "lr": 9.899006476958203e-05}
{"train_loss": 0.42476722598075867, "global_step": 17864, "epoch": 200, "lr": 9.898994883430111e-05}
{"train_loss": 0.5818073749542236, "global_step": 17865, "epoch": 200, "lr": 9.898983289243408e-05}
{"train_loss": 0.49239280819892883, "global_step": 17866, "epoch": 200, "lr": 9.898971694398099e-05}
{"train_loss": 0.4569251239299774, "global_step": 17867, "epoch": 200, "lr": 9.898960098894178e-05}
{"train_loss": 0.46814343333244324, "global_step": 17868, "epoch": 200, "lr": 9.898948502731655e-05}
{"train_loss": 0.4007537364959717, "global_step": 17869, "epoch": 200, "lr": 9.898936905910528e-05}
{"train_loss": 0.4793630540370941, "global_step": 17870, "epoch": 200, "lr": 9.898925308430797e-05}
{"train_loss": 0.513378381729126, "global_step": 17871, "epoch": 200, "lr": 9.898913710292467e-05}
{"train_loss": 0.5089373588562012, "global_step": 17872, "epoch": 200, "lr": 9.898902111495538e-05}
{"train_loss": 0.5419041514396667, "global_step": 17873, "epoch": 200, "lr": 9.898890512040009e-05}
{"train_loss": 0.3878708481788635, "global_step": 17874, "epoch": 200, "lr": 9.898878911925886e-05}
{"train_loss": 0.4583466053009033, "global_step": 17875, "epoch": 200, "lr": 9.898867311153168e-05}
{"train_loss": 0.45154502987861633, "global_step": 17876, "epoch": 200, "lr": 9.898855709721854e-05}
{"train_loss": 0.405021071434021, "global_step": 17877, "epoch": 200, "lr": 9.898844107631953e-05}
{"train_loss": 0.4812614917755127, "global_step": 17878, "epoch": 200, "lr": 9.898832504883459e-05}
{"train_loss": 0.44085440039634705, "global_step": 17879, "epoch": 200, "lr": 9.898820901476378e-05}
{"train_loss": 0.5520361065864563, "global_step": 17880, "epoch": 200, "lr": 9.898809297410711e-05}
{"train_loss": 0.5131688714027405, "global_step": 17881, "epoch": 200, "lr": 9.898797692686458e-05}
{"train_loss": 0.48580440878868103, "global_step": 17882, "epoch": 200, "lr": 9.898786087303622e-05}
{"train_loss": 0.45386454463005066, "global_step": 17883, "epoch": 200, "lr": 9.898774481262203e-05}
{"train_loss": 0.5428450703620911, "global_step": 17884, "epoch": 200, "lr": 9.898762874562203e-05}
{"train_loss": 0.511223554611206, "global_step": 17885, "epoch": 200, "lr": 9.898751267203626e-05}
{"train_loss": 0.6216807961463928, "global_step": 17886, "epoch": 200, "lr": 9.898739659186469e-05}
{"train_loss": 0.5748472809791565, "global_step": 17887, "epoch": 200, "lr": 9.898728050510737e-05}
{"train_loss": 0.505356146043606, "global_step": 17888, "epoch": 200, "lr": 9.898716441176432e-05, "train/sim_max_reward_0": 0.5743707367432906, "train/sim_max_reward_1": 0.9544628002905422, "train/sim_max_reward_2": 0.9971057508833743, "train/sim_max_reward_3": 0.6960796025379328, "train/sim_max_reward_4": 0.9223010663017028, "train/sim_max_reward_5": 0.703096531292743, "test/sim_max_reward_4500000": 0.5273260988441731, "test/sim_max_reward_4500001": 0.43403651255312525, "test/sim_max_reward_4500002": 0.978137863999676, "test/sim_max_reward_4500003": 0.5573374647274806, "test/sim_max_reward_4500004": 0.8771602038942835, "test/sim_max_reward_4500005": 0.9899246623481394, "test/sim_max_reward_4500006": 0.482252968180937, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.11734025541131937, "test/sim_max_reward_4500009": 0.336838324207544, "test/sim_max_reward_4500010": 0.6910414076479774, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.6992271406487158, "test/sim_max_reward_4500014": 0.44933129900665286, "test/sim_max_reward_4500015": 0.9488539098830784, "test/sim_max_reward_4500016": 0.5698280808303233, "test/sim_max_reward_4500017": 0.8551946946614148, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.9858955361997486, "test/sim_max_reward_4500021": 0.8752506129968709, "test/sim_max_reward_4500022": 0.9633062348611072, "test/sim_max_reward_4500023": 0.9620136815747411, "test/sim_max_reward_4500024": 0.9920710039219504, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 0.5519471993832458, "test/sim_max_reward_4500028": 0.9862749806983595, "test/sim_max_reward_4500029": 0.7176418186424042, "test/sim_max_reward_4500030": 0.9473024493115442, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.549532864015158, "test/sim_max_reward_4500033": 0.5182699676933543, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.6675336684096032, "test/sim_max_reward_4500036": 0.7977153116712058, "test/sim_max_reward_4500037": 0.9766156834662009, "test/sim_max_reward_4500038": 0.652080198439761, "test/sim_max_reward_4500039": 0.8297885586450626, "test/sim_max_reward_4500040": 0.9962976455121472, "test/sim_max_reward_4500041": 0.449759694690503, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.7077353610532222, "test/sim_max_reward_4500044": 0.5678239060938479, "test/sim_max_reward_4500045": 0.5335739491985643, "test/sim_max_reward_4500046": 0.6961329073441352, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.5689232723968478, "train/mean_score": 0.8079027480082642, "test/mean_score": 0.7001463478612885, "val_loss": 1.6494888067245483, "train_action_mse_error": 43.451934814453125}
{"train_loss": 0.4694828987121582, "global_step": 17889, "epoch": 201, "lr": 9.898704831183554e-05}
{"train_loss": 0.4914616346359253, "global_step": 17890, "epoch": 201, "lr": 9.898693220532104e-05}
{"train_loss": 0.5042930245399475, "global_step": 17891, "epoch": 201, "lr": 9.898681609222085e-05}
{"train_loss": 0.5007025003433228, "global_step": 17892, "epoch": 201, "lr": 9.898669997253498e-05}
{"train_loss": 0.5664247274398804, "global_step": 17893, "epoch": 201, "lr": 9.898658384626344e-05}
{"train_loss": 0.472174733877182, "global_step": 17894, "epoch": 201, "lr": 9.898646771340624e-05}
{"train_loss": 0.5804944038391113, "global_step": 17895, "epoch": 201, "lr": 9.898635157396343e-05}
{"train_loss": 0.5807514190673828, "global_step": 17896, "epoch": 201, "lr": 9.898623542793499e-05}
{"train_loss": 0.4477253258228302, "global_step": 17897, "epoch": 201, "lr": 9.898611927532096e-05}
{"train_loss": 0.42892035841941833, "global_step": 17898, "epoch": 201, "lr": 9.898600311612131e-05}
{"train_loss": 0.5958470702171326, "global_step": 17899, "epoch": 201, "lr": 9.898588695033612e-05}
{"train_loss": 0.5668250322341919, "global_step": 17900, "epoch": 201, "lr": 9.898577077796537e-05}
{"train_loss": 0.5718969106674194, "global_step": 17901, "epoch": 201, "lr": 9.898565459900909e-05}
{"train_loss": 0.4542296230792999, "global_step": 17902, "epoch": 201, "lr": 9.898553841346725e-05}
{"train_loss": 0.3611430823802948, "global_step": 17903, "epoch": 201, "lr": 9.898542222133992e-05}
{"train_loss": 0.37209001183509827, "global_step": 17904, "epoch": 201, "lr": 9.898530602262712e-05}
{"train_loss": 0.663133978843689, "global_step": 17905, "epoch": 201, "lr": 9.898518981732882e-05}
{"train_loss": 0.4678431451320648, "global_step": 17906, "epoch": 201, "lr": 9.898507360544506e-05}
{"train_loss": 0.4880909323692322, "global_step": 17907, "epoch": 201, "lr": 9.898495738697586e-05}
{"train_loss": 0.5873228311538696, "global_step": 17908, "epoch": 201, "lr": 9.898484116192122e-05}
{"train_loss": 0.5495360493659973, "global_step": 17909, "epoch": 201, "lr": 9.898472493028117e-05}
{"train_loss": 0.5571111440658569, "global_step": 17910, "epoch": 201, "lr": 9.898460869205573e-05}
{"train_loss": 0.5765801072120667, "global_step": 17911, "epoch": 201, "lr": 9.898449244724489e-05}
{"train_loss": 0.41454261541366577, "global_step": 17912, "epoch": 201, "lr": 9.898437619584868e-05}
{"train_loss": 0.6935712695121765, "global_step": 17913, "epoch": 201, "lr": 9.898425993786713e-05}
{"train_loss": 0.5051498413085938, "global_step": 17914, "epoch": 201, "lr": 9.898414367330024e-05}
{"train_loss": 0.512168288230896, "global_step": 17915, "epoch": 201, "lr": 9.898402740214803e-05}
{"train_loss": 0.5615286231040955, "global_step": 17916, "epoch": 201, "lr": 9.898391112441052e-05}
{"train_loss": 0.5102195143699646, "global_step": 17917, "epoch": 201, "lr": 9.898379484008771e-05}
{"train_loss": 0.5128303170204163, "global_step": 17918, "epoch": 201, "lr": 9.898367854917962e-05}
{"train_loss": 0.5265393257141113, "global_step": 17919, "epoch": 201, "lr": 9.89835622516863e-05}
{"train_loss": 0.5116654634475708, "global_step": 17920, "epoch": 201, "lr": 9.898344594760771e-05}
{"train_loss": 0.5687412023544312, "global_step": 17921, "epoch": 201, "lr": 9.89833296369439e-05}
{"train_loss": 0.38875052332878113, "global_step": 17922, "epoch": 201, "lr": 9.898321331969489e-05}
{"train_loss": 0.4390900731086731, "global_step": 17923, "epoch": 201, "lr": 9.898309699586067e-05}
{"train_loss": 0.4381670355796814, "global_step": 17924, "epoch": 201, "lr": 9.898298066544128e-05}
{"train_loss": 0.39154118299484253, "global_step": 17925, "epoch": 201, "lr": 9.898286432843671e-05}
{"train_loss": 0.5094125270843506, "global_step": 17926, "epoch": 201, "lr": 9.898274798484699e-05}
{"train_loss": 0.5609841346740723, "global_step": 17927, "epoch": 201, "lr": 9.898263163467215e-05}
{"train_loss": 0.5325044393539429, "global_step": 17928, "epoch": 201, "lr": 9.898251527791218e-05}
{"train_loss": 0.4654659032821655, "global_step": 17929, "epoch": 201, "lr": 9.898239891456713e-05}
{"train_loss": 0.5803557634353638, "global_step": 17930, "epoch": 201, "lr": 9.898228254463697e-05}
{"train_loss": 0.4250667095184326, "global_step": 17931, "epoch": 201, "lr": 9.898216616812175e-05}
{"train_loss": 0.42274218797683716, "global_step": 17932, "epoch": 201, "lr": 9.898204978502148e-05}
{"train_loss": 0.4123176634311676, "global_step": 17933, "epoch": 201, "lr": 9.898193339533617e-05}
{"train_loss": 0.5331343412399292, "global_step": 17934, "epoch": 201, "lr": 9.898181699906582e-05}
{"train_loss": 0.5482169985771179, "global_step": 17935, "epoch": 201, "lr": 9.898170059621047e-05}
{"train_loss": 0.5264756083488464, "global_step": 17936, "epoch": 201, "lr": 9.898158418677013e-05}
{"train_loss": 0.5725215077400208, "global_step": 17937, "epoch": 201, "lr": 9.898146777074482e-05}
{"train_loss": 0.6801695823669434, "global_step": 17938, "epoch": 201, "lr": 9.898135134813453e-05}
{"train_loss": 0.47061169147491455, "global_step": 17939, "epoch": 201, "lr": 9.898123491893932e-05}
{"train_loss": 0.5504940748214722, "global_step": 17940, "epoch": 201, "lr": 9.898111848315915e-05}
{"train_loss": 0.4568234980106354, "global_step": 17941, "epoch": 201, "lr": 9.898100204079408e-05}
{"train_loss": 0.518535852432251, "global_step": 17942, "epoch": 201, "lr": 9.898088559184411e-05}
{"train_loss": 0.519412100315094, "global_step": 17943, "epoch": 201, "lr": 9.898076913630927e-05}
{"train_loss": 0.43497222661972046, "global_step": 17944, "epoch": 201, "lr": 9.898065267418955e-05}
{"train_loss": 0.5027483701705933, "global_step": 17945, "epoch": 201, "lr": 9.898053620548498e-05}
{"train_loss": 0.5419031977653503, "global_step": 17946, "epoch": 201, "lr": 9.898041973019556e-05}
{"train_loss": 0.6907667517662048, "global_step": 17947, "epoch": 201, "lr": 9.898030324832134e-05}
{"train_loss": 0.5622530579566956, "global_step": 17948, "epoch": 201, "lr": 9.898018675986231e-05}
{"train_loss": 0.6477056741714478, "global_step": 17949, "epoch": 201, "lr": 9.898007026481848e-05}
{"train_loss": 0.5696895122528076, "global_step": 17950, "epoch": 201, "lr": 9.897995376318989e-05}
{"train_loss": 0.44954314827919006, "global_step": 17951, "epoch": 201, "lr": 9.897983725497653e-05}
{"train_loss": 0.42430630326271057, "global_step": 17952, "epoch": 201, "lr": 9.897972074017845e-05}
{"train_loss": 0.6775587797164917, "global_step": 17953, "epoch": 201, "lr": 9.897960421879562e-05}
{"train_loss": 0.5240462422370911, "global_step": 17954, "epoch": 201, "lr": 9.89794876908281e-05}
{"train_loss": 0.46810609102249146, "global_step": 17955, "epoch": 201, "lr": 9.897937115627587e-05}
{"train_loss": 0.427068293094635, "global_step": 17956, "epoch": 201, "lr": 9.897925461513895e-05}
{"train_loss": 0.4676808714866638, "global_step": 17957, "epoch": 201, "lr": 9.897913806741739e-05}
{"train_loss": 0.48256349563598633, "global_step": 17958, "epoch": 201, "lr": 9.897902151311117e-05}
{"train_loss": 0.6557642221450806, "global_step": 17959, "epoch": 201, "lr": 9.897890495222032e-05}
{"train_loss": 0.5786838531494141, "global_step": 17960, "epoch": 201, "lr": 9.897878838474486e-05}
{"train_loss": 0.5762002468109131, "global_step": 17961, "epoch": 201, "lr": 9.897867181068479e-05}
{"train_loss": 0.4865834414958954, "global_step": 17962, "epoch": 201, "lr": 9.897855523004013e-05}
{"train_loss": 0.4220970869064331, "global_step": 17963, "epoch": 201, "lr": 9.897843864281092e-05}
{"train_loss": 0.5492440462112427, "global_step": 17964, "epoch": 201, "lr": 9.897832204899713e-05}
{"train_loss": 0.5681156516075134, "global_step": 17965, "epoch": 201, "lr": 9.897820544859882e-05}
{"train_loss": 0.4779505729675293, "global_step": 17966, "epoch": 201, "lr": 9.897808884161596e-05}
{"train_loss": 0.4537217319011688, "global_step": 17967, "epoch": 201, "lr": 9.897797222804862e-05}
{"train_loss": 0.5540529489517212, "global_step": 17968, "epoch": 201, "lr": 9.897785560789679e-05}
{"train_loss": 0.5622264742851257, "global_step": 17969, "epoch": 201, "lr": 9.897773898116047e-05}
{"train_loss": 0.4813493490219116, "global_step": 17970, "epoch": 201, "lr": 9.89776223478397e-05}
{"train_loss": 0.6109752655029297, "global_step": 17971, "epoch": 201, "lr": 9.897750570793448e-05}
{"train_loss": 0.4642826020717621, "global_step": 17972, "epoch": 201, "lr": 9.897738906144483e-05}
{"train_loss": 0.5956824421882629, "global_step": 17973, "epoch": 201, "lr": 9.897727240837077e-05}
{"train_loss": 0.44690102338790894, "global_step": 17974, "epoch": 201, "lr": 9.897715574871232e-05}
{"train_loss": 0.48780345916748047, "global_step": 17975, "epoch": 201, "lr": 9.897703908246946e-05}
{"train_loss": 0.5271656513214111, "global_step": 17976, "epoch": 201, "lr": 9.897692240964228e-05}
{"train_loss": 0.5160113301839722, "global_step": 17977, "epoch": 201, "lr": 9.897680573023071e-05, "val_loss": 1.560893177986145}
{"train_loss": 0.578976035118103, "global_step": 17978, "epoch": 202, "lr": 9.897668904423483e-05}
{"train_loss": 0.5694078207015991, "global_step": 17979, "epoch": 202, "lr": 9.897657235165462e-05}
{"train_loss": 0.5351352691650391, "global_step": 17980, "epoch": 202, "lr": 9.897645565249011e-05}
{"train_loss": 0.472025603055954, "global_step": 17981, "epoch": 202, "lr": 9.897633894674132e-05}
{"train_loss": 0.6117215752601624, "global_step": 17982, "epoch": 202, "lr": 9.897622223440824e-05}
{"train_loss": 0.42992791533470154, "global_step": 17983, "epoch": 202, "lr": 9.897610551549091e-05}
{"train_loss": 0.5298219323158264, "global_step": 17984, "epoch": 202, "lr": 9.897598878998935e-05}
{"train_loss": 0.515448272228241, "global_step": 17985, "epoch": 202, "lr": 9.897587205790356e-05}
{"train_loss": 0.4984525740146637, "global_step": 17986, "epoch": 202, "lr": 9.897575531923356e-05}
{"train_loss": 0.5205084681510925, "global_step": 17987, "epoch": 202, "lr": 9.897563857397936e-05}
{"train_loss": 0.5473253130912781, "global_step": 17988, "epoch": 202, "lr": 9.897552182214098e-05}
{"train_loss": 0.4658087491989136, "global_step": 17989, "epoch": 202, "lr": 9.897540506371845e-05}
{"train_loss": 0.4925561845302582, "global_step": 17990, "epoch": 202, "lr": 9.897528829871176e-05}
{"train_loss": 0.40643540024757385, "global_step": 17991, "epoch": 202, "lr": 9.897517152712096e-05}
{"train_loss": 0.5931956768035889, "global_step": 17992, "epoch": 202, "lr": 9.897505474894603e-05}
{"train_loss": 0.5075211524963379, "global_step": 17993, "epoch": 202, "lr": 9.8974937964187e-05}
{"train_loss": 0.6166195869445801, "global_step": 17994, "epoch": 202, "lr": 9.89748211728439e-05}
{"train_loss": 0.5181263089179993, "global_step": 17995, "epoch": 202, "lr": 9.897470437491672e-05}
{"train_loss": 0.5315786004066467, "global_step": 17996, "epoch": 202, "lr": 9.89745875704055e-05}
{"train_loss": 0.4280478358268738, "global_step": 17997, "epoch": 202, "lr": 9.897447075931023e-05}
{"train_loss": 0.5755340456962585, "global_step": 17998, "epoch": 202, "lr": 9.897435394163093e-05}
{"train_loss": 0.4693305790424347, "global_step": 17999, "epoch": 202, "lr": 9.897423711736764e-05}
{"train_loss": 0.572206437587738, "global_step": 18000, "epoch": 202, "lr": 9.897412028652037e-05}
{"train_loss": 0.48327046632766724, "global_step": 18001, "epoch": 202, "lr": 9.897400344908911e-05}
{"train_loss": 0.43227213621139526, "global_step": 18002, "epoch": 202, "lr": 9.897388660507389e-05}
{"train_loss": 0.52325439453125, "global_step": 18003, "epoch": 202, "lr": 9.897376975447475e-05}
{"train_loss": 0.5388638973236084, "global_step": 18004, "epoch": 202, "lr": 9.897365289729167e-05}
{"train_loss": 0.5195724964141846, "global_step": 18005, "epoch": 202, "lr": 9.897353603352469e-05}
{"train_loss": 0.6027061343193054, "global_step": 18006, "epoch": 202, "lr": 9.897341916317379e-05}
{"train_loss": 0.5590915679931641, "global_step": 18007, "epoch": 202, "lr": 9.897330228623902e-05}
{"train_loss": 0.5565807223320007, "global_step": 18008, "epoch": 202, "lr": 9.89731854027204e-05}
{"train_loss": 0.45905011892318726, "global_step": 18009, "epoch": 202, "lr": 9.897306851261792e-05}
{"train_loss": 0.4620683789253235, "global_step": 18010, "epoch": 202, "lr": 9.89729516159316e-05}
{"train_loss": 0.37943148612976074, "global_step": 18011, "epoch": 202, "lr": 9.897283471266149e-05}
{"train_loss": 0.40907421708106995, "global_step": 18012, "epoch": 202, "lr": 9.897271780280756e-05}
{"train_loss": 0.4562495946884155, "global_step": 18013, "epoch": 202, "lr": 9.897260088636986e-05}
{"train_loss": 0.427337110042572, "global_step": 18014, "epoch": 202, "lr": 9.897248396334837e-05}
{"train_loss": 0.5582394599914551, "global_step": 18015, "epoch": 202, "lr": 9.897236703374315e-05}
{"train_loss": 0.6513543128967285, "global_step": 18016, "epoch": 202, "lr": 9.897225009755416e-05}
{"train_loss": 0.6108279228210449, "global_step": 18017, "epoch": 202, "lr": 9.897213315478147e-05}
{"train_loss": 0.5476480722427368, "global_step": 18018, "epoch": 202, "lr": 9.897201620542506e-05}
{"train_loss": 0.4903072118759155, "global_step": 18019, "epoch": 202, "lr": 9.897189924948497e-05}
{"train_loss": 0.5609748363494873, "global_step": 18020, "epoch": 202, "lr": 9.897178228696123e-05}
{"train_loss": 0.5768672227859497, "global_step": 18021, "epoch": 202, "lr": 9.897166531785379e-05}
{"train_loss": 0.45982152223587036, "global_step": 18022, "epoch": 202, "lr": 9.897154834216272e-05}
{"train_loss": 0.4105885922908783, "global_step": 18023, "epoch": 202, "lr": 9.897143135988802e-05}
{"train_loss": 0.43056994676589966, "global_step": 18024, "epoch": 202, "lr": 9.897131437102971e-05}
{"train_loss": 0.5662993788719177, "global_step": 18025, "epoch": 202, "lr": 9.897119737558781e-05}
{"train_loss": 0.6993537545204163, "global_step": 18026, "epoch": 202, "lr": 9.897108037356234e-05}
{"train_loss": 0.657402515411377, "global_step": 18027, "epoch": 202, "lr": 9.897096336495329e-05}
{"train_loss": 0.5503460764884949, "global_step": 18028, "epoch": 202, "lr": 9.897084634976068e-05}
{"train_loss": 0.5228665471076965, "global_step": 18029, "epoch": 202, "lr": 9.897072932798454e-05}
{"train_loss": 0.48248058557510376, "global_step": 18030, "epoch": 202, "lr": 9.89706122996249e-05}
{"train_loss": 0.5905749797821045, "global_step": 18031, "epoch": 202, "lr": 9.897049526468173e-05}
{"train_loss": 0.5312736630439758, "global_step": 18032, "epoch": 202, "lr": 9.89703782231551e-05}
{"train_loss": 0.8099867701530457, "global_step": 18033, "epoch": 202, "lr": 9.897026117504499e-05}
{"train_loss": 0.5214651823043823, "global_step": 18034, "epoch": 202, "lr": 9.897014412035143e-05}
{"train_loss": 0.5912439227104187, "global_step": 18035, "epoch": 202, "lr": 9.897002705907443e-05}
{"train_loss": 0.47917667031288147, "global_step": 18036, "epoch": 202, "lr": 9.896990999121401e-05}
{"train_loss": 0.6013056039810181, "global_step": 18037, "epoch": 202, "lr": 9.896979291677016e-05}
{"train_loss": 0.4985131025314331, "global_step": 18038, "epoch": 202, "lr": 9.896967583574295e-05}
{"train_loss": 0.4705958664417267, "global_step": 18039, "epoch": 202, "lr": 9.896955874813234e-05}
{"train_loss": 0.5362544655799866, "global_step": 18040, "epoch": 202, "lr": 9.896944165393838e-05}
{"train_loss": 0.5623676776885986, "global_step": 18041, "epoch": 202, "lr": 9.896932455316107e-05}
{"train_loss": 0.6218671798706055, "global_step": 18042, "epoch": 202, "lr": 9.896920744580043e-05}
{"train_loss": 0.3512440621852875, "global_step": 18043, "epoch": 202, "lr": 9.896909033185649e-05}
{"train_loss": 0.5780280232429504, "global_step": 18044, "epoch": 202, "lr": 9.896897321132924e-05}
{"train_loss": 0.4480734169483185, "global_step": 18045, "epoch": 202, "lr": 9.896885608421871e-05}
{"train_loss": 0.678875744342804, "global_step": 18046, "epoch": 202, "lr": 9.896873895052492e-05}
{"train_loss": 0.4818146228790283, "global_step": 18047, "epoch": 202, "lr": 9.896862181024787e-05}
{"train_loss": 0.5805957317352295, "global_step": 18048, "epoch": 202, "lr": 9.896850466338761e-05}
{"train_loss": 0.5586180686950684, "global_step": 18049, "epoch": 202, "lr": 9.89683875099441e-05}
{"train_loss": 0.4823834002017975, "global_step": 18050, "epoch": 202, "lr": 9.89682703499174e-05}
{"train_loss": 0.41899392008781433, "global_step": 18051, "epoch": 202, "lr": 9.896815318330752e-05}
{"train_loss": 0.38573941588401794, "global_step": 18052, "epoch": 202, "lr": 9.896803601011446e-05}
{"train_loss": 0.3680572807788849, "global_step": 18053, "epoch": 202, "lr": 9.896791883033825e-05}
{"train_loss": 0.5913969278335571, "global_step": 18054, "epoch": 202, "lr": 9.89678016439789e-05}
{"train_loss": 0.5034363269805908, "global_step": 18055, "epoch": 202, "lr": 9.896768445103642e-05}
{"train_loss": 0.5628088116645813, "global_step": 18056, "epoch": 202, "lr": 9.896756725151084e-05}
{"train_loss": 0.5296204090118408, "global_step": 18057, "epoch": 202, "lr": 9.896745004540215e-05}
{"train_loss": 0.4606420695781708, "global_step": 18058, "epoch": 202, "lr": 9.89673328327104e-05}
{"train_loss": 0.632719874382019, "global_step": 18059, "epoch": 202, "lr": 9.896721561343558e-05}
{"train_loss": 0.4351865351200104, "global_step": 18060, "epoch": 202, "lr": 9.896709838757771e-05}
{"train_loss": 0.5756840705871582, "global_step": 18061, "epoch": 202, "lr": 9.896698115513683e-05}
{"train_loss": 0.45240750908851624, "global_step": 18062, "epoch": 202, "lr": 9.896686391611292e-05}
{"train_loss": 0.438750684261322, "global_step": 18063, "epoch": 202, "lr": 9.896674667050602e-05}
{"train_loss": 0.44320330023765564, "global_step": 18064, "epoch": 202, "lr": 9.896662941831613e-05}
{"train_loss": 0.51813143491745, "global_step": 18065, "epoch": 202, "lr": 9.896651215954329e-05}
{"train_loss": 0.5220918086137665, "global_step": 18066, "epoch": 202, "lr": 9.896639489418749e-05, "val_loss": 1.6374887228012085}
{"train_loss": 0.40651243925094604, "global_step": 18067, "epoch": 203, "lr": 9.896627762224874e-05}
{"train_loss": 0.5265958905220032, "global_step": 18068, "epoch": 203, "lr": 9.89661603437271e-05}
{"train_loss": 0.47189831733703613, "global_step": 18069, "epoch": 203, "lr": 9.896604305862253e-05}
{"train_loss": 0.5737826228141785, "global_step": 18070, "epoch": 203, "lr": 9.896592576693509e-05}
{"train_loss": 0.6164327263832092, "global_step": 18071, "epoch": 203, "lr": 9.896580846866476e-05}
{"train_loss": 0.5874887704849243, "global_step": 18072, "epoch": 203, "lr": 9.896569116381158e-05}
{"train_loss": 0.46225860714912415, "global_step": 18073, "epoch": 203, "lr": 9.896557385237557e-05}
{"train_loss": 0.37841618061065674, "global_step": 18074, "epoch": 203, "lr": 9.896545653435673e-05}
{"train_loss": 0.5075131058692932, "global_step": 18075, "epoch": 203, "lr": 9.896533920975508e-05}
{"train_loss": 0.5788484215736389, "global_step": 18076, "epoch": 203, "lr": 9.896522187857064e-05}
{"train_loss": 0.5287452936172485, "global_step": 18077, "epoch": 203, "lr": 9.896510454080341e-05}
{"train_loss": 0.5566083788871765, "global_step": 18078, "epoch": 203, "lr": 9.896498719645345e-05}
{"train_loss": 0.4599471092224121, "global_step": 18079, "epoch": 203, "lr": 9.896486984552071e-05}
{"train_loss": 0.41420477628707886, "global_step": 18080, "epoch": 203, "lr": 9.896475248800525e-05}
{"train_loss": 0.5108173489570618, "global_step": 18081, "epoch": 203, "lr": 9.896463512390708e-05}
{"train_loss": 0.5982972979545593, "global_step": 18082, "epoch": 203, "lr": 9.896451775322622e-05}
{"train_loss": 0.5060570240020752, "global_step": 18083, "epoch": 203, "lr": 9.896440037596266e-05}
{"train_loss": 0.47843924164772034, "global_step": 18084, "epoch": 203, "lr": 9.896428299211644e-05}
{"train_loss": 0.5060919523239136, "global_step": 18085, "epoch": 203, "lr": 9.896416560168758e-05}
{"train_loss": 0.3949527144432068, "global_step": 18086, "epoch": 203, "lr": 9.896404820467606e-05}
{"train_loss": 0.5094970464706421, "global_step": 18087, "epoch": 203, "lr": 9.896393080108194e-05}
{"train_loss": 0.48478466272354126, "global_step": 18088, "epoch": 203, "lr": 9.896381339090522e-05}
{"train_loss": 0.5281239748001099, "global_step": 18089, "epoch": 203, "lr": 9.89636959741459e-05}
{"train_loss": 0.47008949518203735, "global_step": 18090, "epoch": 203, "lr": 9.8963578550804e-05}
{"train_loss": 0.3871403932571411, "global_step": 18091, "epoch": 203, "lr": 9.896346112087955e-05}
{"train_loss": 0.5180941820144653, "global_step": 18092, "epoch": 203, "lr": 9.896334368437257e-05}
{"train_loss": 0.4570189416408539, "global_step": 18093, "epoch": 203, "lr": 9.896322624128307e-05}
{"train_loss": 0.5963412523269653, "global_step": 18094, "epoch": 203, "lr": 9.896310879161105e-05}
{"train_loss": 0.3637702465057373, "global_step": 18095, "epoch": 203, "lr": 9.896299133535654e-05}
{"train_loss": 0.6566953659057617, "global_step": 18096, "epoch": 203, "lr": 9.896287387251955e-05}
{"train_loss": 0.47639763355255127, "global_step": 18097, "epoch": 203, "lr": 9.896275640310009e-05}
{"train_loss": 0.7224264740943909, "global_step": 18098, "epoch": 203, "lr": 9.896263892709819e-05}
{"train_loss": 0.4435572624206543, "global_step": 18099, "epoch": 203, "lr": 9.896252144451386e-05}
{"train_loss": 0.4700549244880676, "global_step": 18100, "epoch": 203, "lr": 9.896240395534712e-05}
{"train_loss": 0.48337405920028687, "global_step": 18101, "epoch": 203, "lr": 9.896228645959799e-05}
{"train_loss": 0.444781631231308, "global_step": 18102, "epoch": 203, "lr": 9.896216895726646e-05}
{"train_loss": 0.4663311243057251, "global_step": 18103, "epoch": 203, "lr": 9.896205144835258e-05}
{"train_loss": 0.472987562417984, "global_step": 18104, "epoch": 203, "lr": 9.896193393285633e-05}
{"train_loss": 0.4777544438838959, "global_step": 18105, "epoch": 203, "lr": 9.896181641077776e-05}
{"train_loss": 0.4423059821128845, "global_step": 18106, "epoch": 203, "lr": 9.896169888211685e-05}
{"train_loss": 0.5132961869239807, "global_step": 18107, "epoch": 203, "lr": 9.896158134687365e-05}
{"train_loss": 0.3813888728618622, "global_step": 18108, "epoch": 203, "lr": 9.896146380504818e-05}
{"train_loss": 0.6153813600540161, "global_step": 18109, "epoch": 203, "lr": 9.896134625664042e-05}
{"train_loss": 0.4263988137245178, "global_step": 18110, "epoch": 203, "lr": 9.89612287016504e-05}
{"train_loss": 0.5848429203033447, "global_step": 18111, "epoch": 203, "lr": 9.896111114007814e-05}
{"train_loss": 0.4218459129333496, "global_step": 18112, "epoch": 203, "lr": 9.896099357192367e-05}
{"train_loss": 0.4513435363769531, "global_step": 18113, "epoch": 203, "lr": 9.896087599718698e-05}
{"train_loss": 0.37071627378463745, "global_step": 18114, "epoch": 203, "lr": 9.89607584158681e-05}
{"train_loss": 0.5227869749069214, "global_step": 18115, "epoch": 203, "lr": 9.896064082796703e-05}
{"train_loss": 0.5098399519920349, "global_step": 18116, "epoch": 203, "lr": 9.896052323348382e-05}
{"train_loss": 0.6320927143096924, "global_step": 18117, "epoch": 203, "lr": 9.896040563241844e-05}
{"train_loss": 0.5275173187255859, "global_step": 18118, "epoch": 203, "lr": 9.896028802477096e-05}
{"train_loss": 0.5857065320014954, "global_step": 18119, "epoch": 203, "lr": 9.896017041054134e-05}
{"train_loss": 0.5268096327781677, "global_step": 18120, "epoch": 203, "lr": 9.896005278972963e-05}
{"train_loss": 0.42076411843299866, "global_step": 18121, "epoch": 203, "lr": 9.895993516233583e-05}
{"train_loss": 0.5601465106010437, "global_step": 18122, "epoch": 203, "lr": 9.895981752835999e-05}
{"train_loss": 0.4814598262310028, "global_step": 18123, "epoch": 203, "lr": 9.895969988780208e-05}
{"train_loss": 0.7222285270690918, "global_step": 18124, "epoch": 203, "lr": 9.895958224066214e-05}
{"train_loss": 0.5176852941513062, "global_step": 18125, "epoch": 203, "lr": 9.895946458694018e-05}
{"train_loss": 0.4944029450416565, "global_step": 18126, "epoch": 203, "lr": 9.895934692663621e-05}
{"train_loss": 0.5417807698249817, "global_step": 18127, "epoch": 203, "lr": 9.895922925975025e-05}
{"train_loss": 0.5379708409309387, "global_step": 18128, "epoch": 203, "lr": 9.895911158628233e-05}
{"train_loss": 0.573788046836853, "global_step": 18129, "epoch": 203, "lr": 9.895899390623245e-05}
{"train_loss": 0.6029395461082458, "global_step": 18130, "epoch": 203, "lr": 9.895887621960063e-05}
{"train_loss": 0.5746637582778931, "global_step": 18131, "epoch": 203, "lr": 9.895875852638688e-05}
{"train_loss": 0.35565170645713806, "global_step": 18132, "epoch": 203, "lr": 9.895864082659124e-05}
{"train_loss": 0.41987597942352295, "global_step": 18133, "epoch": 203, "lr": 9.895852312021369e-05}
{"train_loss": 0.4230133593082428, "global_step": 18134, "epoch": 203, "lr": 9.895840540725427e-05}
{"train_loss": 0.5067269206047058, "global_step": 18135, "epoch": 203, "lr": 9.895828768771298e-05}
{"train_loss": 0.4424310326576233, "global_step": 18136, "epoch": 203, "lr": 9.895816996158987e-05}
{"train_loss": 0.48438867926597595, "global_step": 18137, "epoch": 203, "lr": 9.89580522288849e-05}
{"train_loss": 0.5267758369445801, "global_step": 18138, "epoch": 203, "lr": 9.895793448959814e-05}
{"train_loss": 0.5320392847061157, "global_step": 18139, "epoch": 203, "lr": 9.895781674372955e-05}
{"train_loss": 0.5146428942680359, "global_step": 18140, "epoch": 203, "lr": 9.895769899127921e-05}
{"train_loss": 0.5257314443588257, "global_step": 18141, "epoch": 203, "lr": 9.89575812322471e-05}
{"train_loss": 0.46869194507598877, "global_step": 18142, "epoch": 203, "lr": 9.895746346663323e-05}
{"train_loss": 0.5135576725006104, "global_step": 18143, "epoch": 203, "lr": 9.895734569443763e-05}
{"train_loss": 0.47773972153663635, "global_step": 18144, "epoch": 203, "lr": 9.895722791566031e-05}
{"train_loss": 0.47069230675697327, "global_step": 18145, "epoch": 203, "lr": 9.895711013030129e-05}
{"train_loss": 0.3999931812286377, "global_step": 18146, "epoch": 203, "lr": 9.895699233836058e-05}
{"train_loss": 0.49087777733802795, "global_step": 18147, "epoch": 203, "lr": 9.895687453983819e-05}
{"train_loss": 0.6198834776878357, "global_step": 18148, "epoch": 203, "lr": 9.895675673473417e-05}
{"train_loss": 0.43515563011169434, "global_step": 18149, "epoch": 203, "lr": 9.89566389230485e-05}
{"train_loss": 0.5316919088363647, "global_step": 18150, "epoch": 203, "lr": 9.89565211047812e-05}
{"train_loss": 0.4807824194431305, "global_step": 18151, "epoch": 203, "lr": 9.89564032799323e-05}
{"train_loss": 0.6527684926986694, "global_step": 18152, "epoch": 203, "lr": 9.89562854485018e-05}
{"train_loss": 0.5703186392784119, "global_step": 18153, "epoch": 203, "lr": 9.895616761048973e-05}
{"train_loss": 0.5255142450332642, "global_step": 18154, "epoch": 203, "lr": 9.89560497658961e-05}
{"train_loss": 0.5040977905305584, "global_step": 18155, "epoch": 203, "lr": 9.895593191472092e-05, "val_loss": 1.6014735698699951}
{"train_loss": 0.5543614029884338, "global_step": 18156, "epoch": 204, "lr": 9.895581405696422e-05}
{"train_loss": 0.541570782661438, "global_step": 18157, "epoch": 204, "lr": 9.895569619262599e-05}
{"train_loss": 0.6265919804573059, "global_step": 18158, "epoch": 204, "lr": 9.895557832170628e-05}
{"train_loss": 0.5586900115013123, "global_step": 18159, "epoch": 204, "lr": 9.895546044420509e-05}
{"train_loss": 0.48186999559402466, "global_step": 18160, "epoch": 204, "lr": 9.895534256012243e-05}
{"train_loss": 0.4937496781349182, "global_step": 18161, "epoch": 204, "lr": 9.895522466945832e-05}
{"train_loss": 0.40490612387657166, "global_step": 18162, "epoch": 204, "lr": 9.895510677221278e-05}
{"train_loss": 0.43233969807624817, "global_step": 18163, "epoch": 204, "lr": 9.895498886838582e-05}
{"train_loss": 0.4735867381095886, "global_step": 18164, "epoch": 204, "lr": 9.895487095797747e-05}
{"train_loss": 0.4566955268383026, "global_step": 18165, "epoch": 204, "lr": 9.895475304098771e-05}
{"train_loss": 0.4758986532688141, "global_step": 18166, "epoch": 204, "lr": 9.89546351174166e-05}
{"train_loss": 0.4702335596084595, "global_step": 18167, "epoch": 204, "lr": 9.895451718726413e-05}
{"train_loss": 0.4777330756187439, "global_step": 18168, "epoch": 204, "lr": 9.895439925053032e-05}
{"train_loss": 0.45274537801742554, "global_step": 18169, "epoch": 204, "lr": 9.89542813072152e-05}
{"train_loss": 0.47905686497688293, "global_step": 18170, "epoch": 204, "lr": 9.895416335731875e-05}
{"train_loss": 0.3534586429595947, "global_step": 18171, "epoch": 204, "lr": 9.895404540084103e-05}
{"train_loss": 0.5693840980529785, "global_step": 18172, "epoch": 204, "lr": 9.895392743778202e-05}
{"train_loss": 0.38322943449020386, "global_step": 18173, "epoch": 204, "lr": 9.895380946814177e-05}
{"train_loss": 0.5143702030181885, "global_step": 18174, "epoch": 204, "lr": 9.895369149192027e-05}
{"train_loss": 0.5202602744102478, "global_step": 18175, "epoch": 204, "lr": 9.895357350911753e-05}
{"train_loss": 0.371995747089386, "global_step": 18176, "epoch": 204, "lr": 9.895345551973359e-05}
{"train_loss": 0.42307940125465393, "global_step": 18177, "epoch": 204, "lr": 9.895333752376846e-05}
{"train_loss": 0.6688041090965271, "global_step": 18178, "epoch": 204, "lr": 9.895321952122212e-05}
{"train_loss": 0.6457141041755676, "global_step": 18179, "epoch": 204, "lr": 9.895310151209464e-05}
{"train_loss": 0.48535940051078796, "global_step": 18180, "epoch": 204, "lr": 9.895298349638603e-05}
{"train_loss": 0.49672773480415344, "global_step": 18181, "epoch": 204, "lr": 9.895286547409626e-05}
{"train_loss": 0.4720895290374756, "global_step": 18182, "epoch": 204, "lr": 9.895274744522538e-05}
{"train_loss": 0.47042930126190186, "global_step": 18183, "epoch": 204, "lr": 9.89526294097734e-05}
{"train_loss": 0.4997641146183014, "global_step": 18184, "epoch": 204, "lr": 9.895251136774033e-05}
{"train_loss": 0.4584590494632721, "global_step": 18185, "epoch": 204, "lr": 9.895239331912621e-05}
{"train_loss": 0.5157959461212158, "global_step": 18186, "epoch": 204, "lr": 9.895227526393103e-05}
{"train_loss": 0.6168650984764099, "global_step": 18187, "epoch": 204, "lr": 9.89521572021548e-05}
{"train_loss": 0.46996092796325684, "global_step": 18188, "epoch": 204, "lr": 9.895203913379756e-05}
{"train_loss": 0.5260759592056274, "global_step": 18189, "epoch": 204, "lr": 9.89519210588593e-05}
{"train_loss": 0.6534126996994019, "global_step": 18190, "epoch": 204, "lr": 9.895180297734008e-05}
{"train_loss": 0.6942296624183655, "global_step": 18191, "epoch": 204, "lr": 9.895168488923987e-05}
{"train_loss": 0.4643198549747467, "global_step": 18192, "epoch": 204, "lr": 9.89515667945587e-05}
{"train_loss": 0.41428306698799133, "global_step": 18193, "epoch": 204, "lr": 9.89514486932966e-05}
{"train_loss": 0.4139902591705322, "global_step": 18194, "epoch": 204, "lr": 9.895133058545357e-05}
{"train_loss": 0.5511854290962219, "global_step": 18195, "epoch": 204, "lr": 9.895121247102963e-05}
{"train_loss": 0.6271740198135376, "global_step": 18196, "epoch": 204, "lr": 9.895109435002479e-05}
{"train_loss": 0.5564354658126831, "global_step": 18197, "epoch": 204, "lr": 9.895097622243907e-05}
{"train_loss": 0.5476639270782471, "global_step": 18198, "epoch": 204, "lr": 9.89508580882725e-05}
{"train_loss": 0.5949447751045227, "global_step": 18199, "epoch": 204, "lr": 9.895073994752508e-05}
{"train_loss": 0.5522594451904297, "global_step": 18200, "epoch": 204, "lr": 9.895062180019681e-05}
{"train_loss": 0.5802319645881653, "global_step": 18201, "epoch": 204, "lr": 9.895050364628775e-05}
{"train_loss": 0.5027185678482056, "global_step": 18202, "epoch": 204, "lr": 9.89503854857979e-05}
{"train_loss": 0.5787427425384521, "global_step": 18203, "epoch": 204, "lr": 9.895026731872723e-05}
{"train_loss": 0.6753453016281128, "global_step": 18204, "epoch": 204, "lr": 9.895014914507583e-05}
{"train_loss": 0.4412481486797333, "global_step": 18205, "epoch": 204, "lr": 9.895003096484365e-05}
{"train_loss": 0.5104258060455322, "global_step": 18206, "epoch": 204, "lr": 9.894991277803075e-05}
{"train_loss": 0.5961015820503235, "global_step": 18207, "epoch": 204, "lr": 9.894979458463713e-05}
{"train_loss": 0.5199325680732727, "global_step": 18208, "epoch": 204, "lr": 9.89496763846628e-05}
{"train_loss": 0.46259140968322754, "global_step": 18209, "epoch": 204, "lr": 9.89495581781078e-05}
{"train_loss": 0.6092934608459473, "global_step": 18210, "epoch": 204, "lr": 9.894943996497212e-05}
{"train_loss": 0.6432284116744995, "global_step": 18211, "epoch": 204, "lr": 9.894932174525577e-05}
{"train_loss": 0.5204823017120361, "global_step": 18212, "epoch": 204, "lr": 9.89492035189588e-05}
{"train_loss": 0.4856642484664917, "global_step": 18213, "epoch": 204, "lr": 9.89490852860812e-05}
{"train_loss": 0.6009112596511841, "global_step": 18214, "epoch": 204, "lr": 9.894896704662297e-05}
{"train_loss": 0.5482044816017151, "global_step": 18215, "epoch": 204, "lr": 9.894884880058417e-05}
{"train_loss": 0.6782625317573547, "global_step": 18216, "epoch": 204, "lr": 9.89487305479648e-05}
{"train_loss": 0.5411588549613953, "global_step": 18217, "epoch": 204, "lr": 9.894861228876486e-05}
{"train_loss": 0.5931774973869324, "global_step": 18218, "epoch": 204, "lr": 9.894849402298439e-05}
{"train_loss": 0.6088665127754211, "global_step": 18219, "epoch": 204, "lr": 9.894837575062337e-05}
{"train_loss": 0.57254558801651, "global_step": 18220, "epoch": 204, "lr": 9.894825747168183e-05}
{"train_loss": 0.5656441450119019, "global_step": 18221, "epoch": 204, "lr": 9.894813918615982e-05}
{"train_loss": 0.5210486650466919, "global_step": 18222, "epoch": 204, "lr": 9.894802089405732e-05}
{"train_loss": 0.6318944692611694, "global_step": 18223, "epoch": 204, "lr": 9.894790259537435e-05}
{"train_loss": 0.4865924119949341, "global_step": 18224, "epoch": 204, "lr": 9.894778429011094e-05}
{"train_loss": 0.6305079460144043, "global_step": 18225, "epoch": 204, "lr": 9.89476659782671e-05}
{"train_loss": 0.5002849102020264, "global_step": 18226, "epoch": 204, "lr": 9.894754765984282e-05}
{"train_loss": 0.5321633219718933, "global_step": 18227, "epoch": 204, "lr": 9.894742933483816e-05}
{"train_loss": 0.4294886589050293, "global_step": 18228, "epoch": 204, "lr": 9.894731100325311e-05}
{"train_loss": 0.49591416120529175, "global_step": 18229, "epoch": 204, "lr": 9.89471926650877e-05}
{"train_loss": 0.5580494403839111, "global_step": 18230, "epoch": 204, "lr": 9.894707432034191e-05}
{"train_loss": 0.5526062250137329, "global_step": 18231, "epoch": 204, "lr": 9.894695596901579e-05}
{"train_loss": 0.5787901282310486, "global_step": 18232, "epoch": 204, "lr": 9.894683761110936e-05}
{"train_loss": 0.4956552982330322, "global_step": 18233, "epoch": 204, "lr": 9.894671924662263e-05}
{"train_loss": 0.5543642044067383, "global_step": 18234, "epoch": 204, "lr": 9.894660087555561e-05}
{"train_loss": 0.6672477722167969, "global_step": 18235, "epoch": 204, "lr": 9.89464824979083e-05}
{"train_loss": 0.49369820952415466, "global_step": 18236, "epoch": 204, "lr": 9.894636411368075e-05}
{"train_loss": 0.6568030118942261, "global_step": 18237, "epoch": 204, "lr": 9.894624572287293e-05}
{"train_loss": 0.5720024108886719, "global_step": 18238, "epoch": 204, "lr": 9.894612732548491e-05}
{"train_loss": 0.4822630286216736, "global_step": 18239, "epoch": 204, "lr": 9.894600892151667e-05}
{"train_loss": 0.46336567401885986, "global_step": 18240, "epoch": 204, "lr": 9.894589051096822e-05}
{"train_loss": 0.5646235346794128, "global_step": 18241, "epoch": 204, "lr": 9.894577209383962e-05}
{"train_loss": 0.5858155488967896, "global_step": 18242, "epoch": 204, "lr": 9.894565367013085e-05}
{"train_loss": 0.5700498819351196, "global_step": 18243, "epoch": 204, "lr": 9.894553523984192e-05}
{"train_loss": 0.5313450995455967, "global_step": 18244, "epoch": 204, "lr": 9.894541680297287e-05, "val_loss": 1.6952745914459229}
{"train_loss": 0.6524190902709961, "global_step": 18245, "epoch": 205, "lr": 9.894529835952371e-05}
{"train_loss": 0.8246707320213318, "global_step": 18246, "epoch": 205, "lr": 9.894517990949443e-05}
{"train_loss": 0.527182936668396, "global_step": 18247, "epoch": 205, "lr": 9.89450614528851e-05}
{"train_loss": 0.43711042404174805, "global_step": 18248, "epoch": 205, "lr": 9.894494298969568e-05}
{"train_loss": 0.46142423152923584, "global_step": 18249, "epoch": 205, "lr": 9.894482451992622e-05}
{"train_loss": 0.49056223034858704, "global_step": 18250, "epoch": 205, "lr": 9.89447060435767e-05}
{"train_loss": 0.4534294307231903, "global_step": 18251, "epoch": 205, "lr": 9.894458756064718e-05}
{"train_loss": 0.48561185598373413, "global_step": 18252, "epoch": 205, "lr": 9.894446907113765e-05}
{"train_loss": 0.5180192589759827, "global_step": 18253, "epoch": 205, "lr": 9.894435057504815e-05}
{"train_loss": 0.5639466643333435, "global_step": 18254, "epoch": 205, "lr": 9.894423207237867e-05}
{"train_loss": 0.4888858497142792, "global_step": 18255, "epoch": 205, "lr": 9.894411356312923e-05}
{"train_loss": 0.5998176336288452, "global_step": 18256, "epoch": 205, "lr": 9.894399504729985e-05}
{"train_loss": 0.45578959584236145, "global_step": 18257, "epoch": 205, "lr": 9.894387652489054e-05}
{"train_loss": 0.5149705410003662, "global_step": 18258, "epoch": 205, "lr": 9.894375799590135e-05}
{"train_loss": 0.5231234431266785, "global_step": 18259, "epoch": 205, "lr": 9.894363946033225e-05}
{"train_loss": 0.45784103870391846, "global_step": 18260, "epoch": 205, "lr": 9.894352091818327e-05}
{"train_loss": 0.48207783699035645, "global_step": 18261, "epoch": 205, "lr": 9.894340236945442e-05}
{"train_loss": 0.5729416608810425, "global_step": 18262, "epoch": 205, "lr": 9.894328381414575e-05}
{"train_loss": 0.5487019419670105, "global_step": 18263, "epoch": 205, "lr": 9.894316525225724e-05}
{"train_loss": 0.6865523457527161, "global_step": 18264, "epoch": 205, "lr": 9.894304668378892e-05}
{"train_loss": 0.6374765634536743, "global_step": 18265, "epoch": 205, "lr": 9.894292810874081e-05}
{"train_loss": 0.7328586578369141, "global_step": 18266, "epoch": 205, "lr": 9.894280952711291e-05}
{"train_loss": 0.4625650644302368, "global_step": 18267, "epoch": 205, "lr": 9.894269093890524e-05}
{"train_loss": 0.5167264342308044, "global_step": 18268, "epoch": 205, "lr": 9.894257234411783e-05}
{"train_loss": 0.5061506032943726, "global_step": 18269, "epoch": 205, "lr": 9.89424537427507e-05}
{"train_loss": 0.566227376461029, "global_step": 18270, "epoch": 205, "lr": 9.894233513480385e-05}
{"train_loss": 0.47800952196121216, "global_step": 18271, "epoch": 205, "lr": 9.894221652027728e-05}
{"train_loss": 0.5123750567436218, "global_step": 18272, "epoch": 205, "lr": 9.894209789917103e-05}
{"train_loss": 0.5175610780715942, "global_step": 18273, "epoch": 205, "lr": 9.894197927148512e-05}
{"train_loss": 0.5630441904067993, "global_step": 18274, "epoch": 205, "lr": 9.894186063721957e-05}
{"train_loss": 0.5018184185028076, "global_step": 18275, "epoch": 205, "lr": 9.894174199637436e-05}
{"train_loss": 0.525544285774231, "global_step": 18276, "epoch": 205, "lr": 9.894162334894954e-05}
{"train_loss": 0.5058727264404297, "global_step": 18277, "epoch": 205, "lr": 9.894150469494512e-05}
{"train_loss": 0.5129144787788391, "global_step": 18278, "epoch": 205, "lr": 9.894138603436111e-05}
{"train_loss": 0.538305401802063, "global_step": 18279, "epoch": 205, "lr": 9.894126736719751e-05}
{"train_loss": 0.5831406116485596, "global_step": 18280, "epoch": 205, "lr": 9.894114869345437e-05}
{"train_loss": 0.6693920493125916, "global_step": 18281, "epoch": 205, "lr": 9.894103001313168e-05}
{"train_loss": 0.5350417494773865, "global_step": 18282, "epoch": 205, "lr": 9.894091132622948e-05}
{"train_loss": 0.5083348751068115, "global_step": 18283, "epoch": 205, "lr": 9.894079263274777e-05}
{"train_loss": 0.5227716565132141, "global_step": 18284, "epoch": 205, "lr": 9.894067393268655e-05}
{"train_loss": 0.5967951416969299, "global_step": 18285, "epoch": 205, "lr": 9.894055522604586e-05}
{"train_loss": 0.6068868637084961, "global_step": 18286, "epoch": 205, "lr": 9.894043651282571e-05}
{"train_loss": 0.5042140483856201, "global_step": 18287, "epoch": 205, "lr": 9.89403177930261e-05}
{"train_loss": 0.5513107776641846, "global_step": 18288, "epoch": 205, "lr": 9.894019906664709e-05}
{"train_loss": 0.5411863327026367, "global_step": 18289, "epoch": 205, "lr": 9.894008033368865e-05}
{"train_loss": 0.5982837677001953, "global_step": 18290, "epoch": 205, "lr": 9.893996159415082e-05}
{"train_loss": 0.7185521721839905, "global_step": 18291, "epoch": 205, "lr": 9.893984284803359e-05}
{"train_loss": 0.5092198848724365, "global_step": 18292, "epoch": 205, "lr": 9.893972409533702e-05}
{"train_loss": 0.5130094885826111, "global_step": 18293, "epoch": 205, "lr": 9.893960533606109e-05}
{"train_loss": 0.4341648817062378, "global_step": 18294, "epoch": 205, "lr": 9.893948657020582e-05}
{"train_loss": 0.4605811536312103, "global_step": 18295, "epoch": 205, "lr": 9.893936779777124e-05}
{"train_loss": 0.5054387450218201, "global_step": 18296, "epoch": 205, "lr": 9.893924901875735e-05}
{"train_loss": 0.5965712070465088, "global_step": 18297, "epoch": 205, "lr": 9.893913023316417e-05}
{"train_loss": 0.5234302878379822, "global_step": 18298, "epoch": 205, "lr": 9.893901144099173e-05}
{"train_loss": 0.5633831024169922, "global_step": 18299, "epoch": 205, "lr": 9.893889264224005e-05}
{"train_loss": 0.5674399733543396, "global_step": 18300, "epoch": 205, "lr": 9.893877383690912e-05}
{"train_loss": 0.46743378043174744, "global_step": 18301, "epoch": 205, "lr": 9.893865502499897e-05}
{"train_loss": 0.5378807187080383, "global_step": 18302, "epoch": 205, "lr": 9.893853620650961e-05}
{"train_loss": 0.5915405750274658, "global_step": 18303, "epoch": 205, "lr": 9.893841738144106e-05}
{"train_loss": 0.45893386006355286, "global_step": 18304, "epoch": 205, "lr": 9.893829854979333e-05}
{"train_loss": 0.5439854860305786, "global_step": 18305, "epoch": 205, "lr": 9.893817971156646e-05}
{"train_loss": 0.535747230052948, "global_step": 18306, "epoch": 205, "lr": 9.893806086676044e-05}
{"train_loss": 0.5699551105499268, "global_step": 18307, "epoch": 205, "lr": 9.89379420153753e-05}
{"train_loss": 0.6327800750732422, "global_step": 18308, "epoch": 205, "lr": 9.893782315741104e-05}
{"train_loss": 0.6650417447090149, "global_step": 18309, "epoch": 205, "lr": 9.893770429286768e-05}
{"train_loss": 0.5708765983581543, "global_step": 18310, "epoch": 205, "lr": 9.893758542174527e-05}
{"train_loss": 0.6945244073867798, "global_step": 18311, "epoch": 205, "lr": 9.893746654404377e-05}
{"train_loss": 0.4705944061279297, "global_step": 18312, "epoch": 205, "lr": 9.893734765976323e-05}
{"train_loss": 0.6741054058074951, "global_step": 18313, "epoch": 205, "lr": 9.893722876890367e-05}
{"train_loss": 0.4755766987800598, "global_step": 18314, "epoch": 205, "lr": 9.89371098714651e-05}
{"train_loss": 0.5154401659965515, "global_step": 18315, "epoch": 205, "lr": 9.893699096744751e-05}
{"train_loss": 0.5223348736763, "global_step": 18316, "epoch": 205, "lr": 9.893687205685096e-05}
{"train_loss": 0.5975688695907593, "global_step": 18317, "epoch": 205, "lr": 9.893675313967544e-05}
{"train_loss": 0.5445824265480042, "global_step": 18318, "epoch": 205, "lr": 9.893663421592094e-05}
{"train_loss": 0.40275058150291443, "global_step": 18319, "epoch": 205, "lr": 9.893651528558754e-05}
{"train_loss": 0.42856502532958984, "global_step": 18320, "epoch": 205, "lr": 9.893639634867521e-05}
{"train_loss": 0.49855926632881165, "global_step": 18321, "epoch": 205, "lr": 9.893627740518399e-05}
{"train_loss": 0.5260113477706909, "global_step": 18322, "epoch": 205, "lr": 9.893615845511386e-05}
{"train_loss": 0.6573712825775146, "global_step": 18323, "epoch": 205, "lr": 9.893603949846489e-05}
{"train_loss": 0.5154221057891846, "global_step": 18324, "epoch": 205, "lr": 9.893592053523704e-05}
{"train_loss": 0.6033546924591064, "global_step": 18325, "epoch": 205, "lr": 9.893580156543036e-05}
{"train_loss": 0.625464916229248, "global_step": 18326, "epoch": 205, "lr": 9.893568258904487e-05}
{"train_loss": 0.5189156532287598, "global_step": 18327, "epoch": 205, "lr": 9.893556360608056e-05}
{"train_loss": 0.624374508857727, "global_step": 18328, "epoch": 205, "lr": 9.893544461653745e-05}
{"train_loss": 0.5607346892356873, "global_step": 18329, "epoch": 205, "lr": 9.893532562041559e-05}
{"train_loss": 0.5423727035522461, "global_step": 18330, "epoch": 205, "lr": 9.893520661771496e-05}
{"train_loss": 0.4899917244911194, "global_step": 18331, "epoch": 205, "lr": 9.893508760843559e-05}
{"train_loss": 0.4475410580635071, "global_step": 18332, "epoch": 205, "lr": 9.89349685925775e-05}
{"train_loss": 0.546364368682497, "global_step": 18333, "epoch": 205, "lr": 9.893484957014069e-05, "val_loss": 1.5530080795288086, "train_action_mse_error": 24.84980010986328}
{"train_loss": 0.4658004343509674, "global_step": 18334, "epoch": 206, "lr": 9.893473054112519e-05}
{"train_loss": 0.47083282470703125, "global_step": 18335, "epoch": 206, "lr": 9.893461150553101e-05}
{"train_loss": 0.4576972424983978, "global_step": 18336, "epoch": 206, "lr": 9.893449246335816e-05}
{"train_loss": 0.5499479174613953, "global_step": 18337, "epoch": 206, "lr": 9.893437341460668e-05}
{"train_loss": 0.5607939958572388, "global_step": 18338, "epoch": 206, "lr": 9.893425435927657e-05}
{"train_loss": 0.6342435479164124, "global_step": 18339, "epoch": 206, "lr": 9.893413529736783e-05}
{"train_loss": 0.5882626175880432, "global_step": 18340, "epoch": 206, "lr": 9.89340162288805e-05}
{"train_loss": 0.3385283648967743, "global_step": 18341, "epoch": 206, "lr": 9.893389715381459e-05}
{"train_loss": 0.5182865858078003, "global_step": 18342, "epoch": 206, "lr": 9.893377807217011e-05}
{"train_loss": 0.5132253766059875, "global_step": 18343, "epoch": 206, "lr": 9.893365898394709e-05}
{"train_loss": 0.6354851126670837, "global_step": 18344, "epoch": 206, "lr": 9.893353988914553e-05}
{"train_loss": 0.6572097539901733, "global_step": 18345, "epoch": 206, "lr": 9.893342078776544e-05}
{"train_loss": 0.4297604262828827, "global_step": 18346, "epoch": 206, "lr": 9.893330167980685e-05}
{"train_loss": 0.4841347336769104, "global_step": 18347, "epoch": 206, "lr": 9.89331825652698e-05}
{"train_loss": 0.5242341756820679, "global_step": 18348, "epoch": 206, "lr": 9.893306344415425e-05}
{"train_loss": 0.5055710673332214, "global_step": 18349, "epoch": 206, "lr": 9.893294431646027e-05}
{"train_loss": 0.43221741914749146, "global_step": 18350, "epoch": 206, "lr": 9.893282518218783e-05}
{"train_loss": 0.450687438249588, "global_step": 18351, "epoch": 206, "lr": 9.893270604133698e-05}
{"train_loss": 0.4372796416282654, "global_step": 18352, "epoch": 206, "lr": 9.893258689390772e-05}
{"train_loss": 0.4177795350551605, "global_step": 18353, "epoch": 206, "lr": 9.893246773990008e-05}
{"train_loss": 0.4089271128177643, "global_step": 18354, "epoch": 206, "lr": 9.893234857931406e-05}
{"train_loss": 0.4502512514591217, "global_step": 18355, "epoch": 206, "lr": 9.893222941214969e-05}
{"train_loss": 0.4290809631347656, "global_step": 18356, "epoch": 206, "lr": 9.893211023840697e-05}
{"train_loss": 0.6642243266105652, "global_step": 18357, "epoch": 206, "lr": 9.89319910580859e-05}
{"train_loss": 0.4575071334838867, "global_step": 18358, "epoch": 206, "lr": 9.893187187118655e-05}
{"train_loss": 0.3570978343486786, "global_step": 18359, "epoch": 206, "lr": 9.893175267770892e-05}
{"train_loss": 0.4736987352371216, "global_step": 18360, "epoch": 206, "lr": 9.893163347765298e-05}
{"train_loss": 0.46217966079711914, "global_step": 18361, "epoch": 206, "lr": 9.893151427101879e-05}
{"train_loss": 0.5618847608566284, "global_step": 18362, "epoch": 206, "lr": 9.893139505780635e-05}
{"train_loss": 0.6176000833511353, "global_step": 18363, "epoch": 206, "lr": 9.89312758380157e-05}
{"train_loss": 0.4221997857093811, "global_step": 18364, "epoch": 206, "lr": 9.89311566116468e-05}
{"train_loss": 0.4755246639251709, "global_step": 18365, "epoch": 206, "lr": 9.893103737869974e-05}
{"train_loss": 0.48918619751930237, "global_step": 18366, "epoch": 206, "lr": 9.893091813917447e-05}
{"train_loss": 0.4249621331691742, "global_step": 18367, "epoch": 206, "lr": 9.893079889307106e-05}
{"train_loss": 0.6174417734146118, "global_step": 18368, "epoch": 206, "lr": 9.893067964038948e-05}
{"train_loss": 0.45200812816619873, "global_step": 18369, "epoch": 206, "lr": 9.893056038112978e-05}
{"train_loss": 0.4299742579460144, "global_step": 18370, "epoch": 206, "lr": 9.893044111529195e-05}
{"train_loss": 0.4360649883747101, "global_step": 18371, "epoch": 206, "lr": 9.893032184287603e-05}
{"train_loss": 0.4560959041118622, "global_step": 18372, "epoch": 206, "lr": 9.893020256388201e-05}
{"train_loss": 0.5402775406837463, "global_step": 18373, "epoch": 206, "lr": 9.893008327830992e-05}
{"train_loss": 0.5559995174407959, "global_step": 18374, "epoch": 206, "lr": 9.892996398615979e-05}
{"train_loss": 0.40521037578582764, "global_step": 18375, "epoch": 206, "lr": 9.892984468743163e-05}
{"train_loss": 0.5327944755554199, "global_step": 18376, "epoch": 206, "lr": 9.892972538212543e-05}
{"train_loss": 0.5581530928611755, "global_step": 18377, "epoch": 206, "lr": 9.892960607024125e-05}
{"train_loss": 0.47321587800979614, "global_step": 18378, "epoch": 206, "lr": 9.892948675177905e-05}
{"train_loss": 0.36755478382110596, "global_step": 18379, "epoch": 206, "lr": 9.892936742673888e-05}
{"train_loss": 0.545346200466156, "global_step": 18380, "epoch": 206, "lr": 9.892924809512077e-05}
{"train_loss": 0.45851364731788635, "global_step": 18381, "epoch": 206, "lr": 9.892912875692472e-05}
{"train_loss": 0.52818763256073, "global_step": 18382, "epoch": 206, "lr": 9.892900941215073e-05}
{"train_loss": 0.5118914842605591, "global_step": 18383, "epoch": 206, "lr": 9.892889006079884e-05}
{"train_loss": 0.32172930240631104, "global_step": 18384, "epoch": 206, "lr": 9.892877070286905e-05}
{"train_loss": 0.5004320740699768, "global_step": 18385, "epoch": 206, "lr": 9.89286513383614e-05}
{"train_loss": 0.5819198489189148, "global_step": 18386, "epoch": 206, "lr": 9.892853196727588e-05}
{"train_loss": 0.6633036732673645, "global_step": 18387, "epoch": 206, "lr": 9.892841258961251e-05}
{"train_loss": 0.5184414982795715, "global_step": 18388, "epoch": 206, "lr": 9.892829320537132e-05}
{"train_loss": 0.48009631037712097, "global_step": 18389, "epoch": 206, "lr": 9.892817381455231e-05}
{"train_loss": 0.6482672095298767, "global_step": 18390, "epoch": 206, "lr": 9.892805441715549e-05}
{"train_loss": 0.47107937932014465, "global_step": 18391, "epoch": 206, "lr": 9.892793501318092e-05}
{"train_loss": 0.5621641874313354, "global_step": 18392, "epoch": 206, "lr": 9.892781560262856e-05}
{"train_loss": 0.4579753279685974, "global_step": 18393, "epoch": 206, "lr": 9.892769618549846e-05}
{"train_loss": 0.64830082654953, "global_step": 18394, "epoch": 206, "lr": 9.892757676179063e-05}
{"train_loss": 0.494658887386322, "global_step": 18395, "epoch": 206, "lr": 9.892745733150507e-05}
{"train_loss": 0.5657730102539062, "global_step": 18396, "epoch": 206, "lr": 9.892733789464184e-05}
{"train_loss": 0.49336346983909607, "global_step": 18397, "epoch": 206, "lr": 9.89272184512009e-05}
{"train_loss": 0.6595832109451294, "global_step": 18398, "epoch": 206, "lr": 9.892709900118232e-05}
{"train_loss": 0.48559466004371643, "global_step": 18399, "epoch": 206, "lr": 9.892697954458605e-05}
{"train_loss": 0.6818844079971313, "global_step": 18400, "epoch": 206, "lr": 9.892686008141216e-05}
{"train_loss": 0.4395126700401306, "global_step": 18401, "epoch": 206, "lr": 9.892674061166066e-05}
{"train_loss": 0.5876380205154419, "global_step": 18402, "epoch": 206, "lr": 9.892662113533156e-05}
{"train_loss": 0.622748851776123, "global_step": 18403, "epoch": 206, "lr": 9.892650165242485e-05}
{"train_loss": 0.5828826427459717, "global_step": 18404, "epoch": 206, "lr": 9.892638216294057e-05}
{"train_loss": 0.5093468427658081, "global_step": 18405, "epoch": 206, "lr": 9.892626266687875e-05}
{"train_loss": 0.552990734577179, "global_step": 18406, "epoch": 206, "lr": 9.892614316423938e-05}
{"train_loss": 0.6281861066818237, "global_step": 18407, "epoch": 206, "lr": 9.892602365502249e-05}
{"train_loss": 0.5805116891860962, "global_step": 18408, "epoch": 206, "lr": 9.892590413922809e-05}
{"train_loss": 0.522786021232605, "global_step": 18409, "epoch": 206, "lr": 9.892578461685619e-05}
{"train_loss": 0.649391770362854, "global_step": 18410, "epoch": 206, "lr": 9.892566508790683e-05}
{"train_loss": 0.6172003149986267, "global_step": 18411, "epoch": 206, "lr": 9.892554555238001e-05}
{"train_loss": 0.5154975652694702, "global_step": 18412, "epoch": 206, "lr": 9.892542601027574e-05}
{"train_loss": 0.6034518480300903, "global_step": 18413, "epoch": 206, "lr": 9.892530646159405e-05}
{"train_loss": 0.48955345153808594, "global_step": 18414, "epoch": 206, "lr": 9.892518690633494e-05}
{"train_loss": 0.4241829514503479, "global_step": 18415, "epoch": 206, "lr": 9.892506734449845e-05}
{"train_loss": 0.5639103055000305, "global_step": 18416, "epoch": 206, "lr": 9.892494777608456e-05}
{"train_loss": 0.5441721677780151, "global_step": 18417, "epoch": 206, "lr": 9.892482820109332e-05}
{"train_loss": 0.6143871545791626, "global_step": 18418, "epoch": 206, "lr": 9.892470861952472e-05}
{"train_loss": 0.6713395118713379, "global_step": 18419, "epoch": 206, "lr": 9.89245890313788e-05}
{"train_loss": 0.340518981218338, "global_step": 18420, "epoch": 206, "lr": 9.892446943665557e-05}
{"train_loss": 0.648585319519043, "global_step": 18421, "epoch": 206, "lr": 9.892434983535504e-05}
{"train_loss": 0.5190466551968221, "global_step": 18422, "epoch": 206, "lr": 9.892423022747721e-05, "val_loss": 1.625430703163147}
{"train_loss": 0.46311038732528687, "global_step": 18423, "epoch": 207, "lr": 9.892411061302214e-05}
{"train_loss": 0.4965539276599884, "global_step": 18424, "epoch": 207, "lr": 9.89239909919898e-05}
{"train_loss": 0.4749309718608856, "global_step": 18425, "epoch": 207, "lr": 9.892387136438024e-05}
{"train_loss": 0.5550833344459534, "global_step": 18426, "epoch": 207, "lr": 9.892375173019345e-05}
{"train_loss": 0.43488556146621704, "global_step": 18427, "epoch": 207, "lr": 9.892363208942945e-05}
{"train_loss": 0.47846660017967224, "global_step": 18428, "epoch": 207, "lr": 9.892351244208829e-05}
{"train_loss": 0.6308929920196533, "global_step": 18429, "epoch": 207, "lr": 9.892339278816994e-05}
{"train_loss": 0.4453914761543274, "global_step": 18430, "epoch": 207, "lr": 9.892327312767444e-05}
{"train_loss": 0.4215337336063385, "global_step": 18431, "epoch": 207, "lr": 9.892315346060181e-05}
{"train_loss": 0.5730953812599182, "global_step": 18432, "epoch": 207, "lr": 9.892303378695205e-05}
{"train_loss": 0.6094706058502197, "global_step": 18433, "epoch": 207, "lr": 9.892291410672518e-05}
{"train_loss": 0.5597772002220154, "global_step": 18434, "epoch": 207, "lr": 9.892279441992123e-05}
{"train_loss": 0.554178774356842, "global_step": 18435, "epoch": 207, "lr": 9.892267472654021e-05}
{"train_loss": 0.5286346673965454, "global_step": 18436, "epoch": 207, "lr": 9.892255502658212e-05}
{"train_loss": 0.3403093218803406, "global_step": 18437, "epoch": 207, "lr": 9.8922435320047e-05}
{"train_loss": 0.3487034738063812, "global_step": 18438, "epoch": 207, "lr": 9.892231560693486e-05}
{"train_loss": 0.4988844394683838, "global_step": 18439, "epoch": 207, "lr": 9.89221958872457e-05}
{"train_loss": 0.5478603839874268, "global_step": 18440, "epoch": 207, "lr": 9.892207616097954e-05}
{"train_loss": 0.4685002863407135, "global_step": 18441, "epoch": 207, "lr": 9.892195642813641e-05}
{"train_loss": 0.6251396536827087, "global_step": 18442, "epoch": 207, "lr": 9.892183668871633e-05}
{"train_loss": 0.5855286121368408, "global_step": 18443, "epoch": 207, "lr": 9.892171694271929e-05}
{"train_loss": 0.512831449508667, "global_step": 18444, "epoch": 207, "lr": 9.892159719014533e-05}
{"train_loss": 0.48430293798446655, "global_step": 18445, "epoch": 207, "lr": 9.892147743099445e-05}
{"train_loss": 0.45499205589294434, "global_step": 18446, "epoch": 207, "lr": 9.892135766526668e-05}
{"train_loss": 0.46482619643211365, "global_step": 18447, "epoch": 207, "lr": 9.892123789296203e-05}
{"train_loss": 0.49741050601005554, "global_step": 18448, "epoch": 207, "lr": 9.892111811408051e-05}
{"train_loss": 0.4531344175338745, "global_step": 18449, "epoch": 207, "lr": 9.892099832862214e-05}
{"train_loss": 0.4152635931968689, "global_step": 18450, "epoch": 207, "lr": 9.892087853658695e-05}
{"train_loss": 0.5109716057777405, "global_step": 18451, "epoch": 207, "lr": 9.892075873797494e-05}
{"train_loss": 0.5784662365913391, "global_step": 18452, "epoch": 207, "lr": 9.892063893278612e-05}
{"train_loss": 0.6903110146522522, "global_step": 18453, "epoch": 207, "lr": 9.892051912102054e-05}
{"train_loss": 0.4914093017578125, "global_step": 18454, "epoch": 207, "lr": 9.892039930267817e-05}
{"train_loss": 0.5080869793891907, "global_step": 18455, "epoch": 207, "lr": 9.892027947775906e-05}
{"train_loss": 0.4407030940055847, "global_step": 18456, "epoch": 207, "lr": 9.89201596462632e-05}
{"train_loss": 0.49958866834640503, "global_step": 18457, "epoch": 207, "lr": 9.892003980819063e-05}
{"train_loss": 0.5213589072227478, "global_step": 18458, "epoch": 207, "lr": 9.891991996354136e-05}
{"train_loss": 0.43943020701408386, "global_step": 18459, "epoch": 207, "lr": 9.891980011231541e-05}
{"train_loss": 0.6723822355270386, "global_step": 18460, "epoch": 207, "lr": 9.891968025451277e-05}
{"train_loss": 0.42023128271102905, "global_step": 18461, "epoch": 207, "lr": 9.891956039013349e-05}
{"train_loss": 0.5818731784820557, "global_step": 18462, "epoch": 207, "lr": 9.891944051917757e-05}
{"train_loss": 0.47279617190361023, "global_step": 18463, "epoch": 207, "lr": 9.891932064164502e-05}
{"train_loss": 0.5422210097312927, "global_step": 18464, "epoch": 207, "lr": 9.891920075753587e-05}
{"train_loss": 0.5655673146247864, "global_step": 18465, "epoch": 207, "lr": 9.891908086685013e-05}
{"train_loss": 0.5370948910713196, "global_step": 18466, "epoch": 207, "lr": 9.891896096958782e-05}
{"train_loss": 0.6123924851417542, "global_step": 18467, "epoch": 207, "lr": 9.891884106574894e-05}
{"train_loss": 0.37877172231674194, "global_step": 18468, "epoch": 207, "lr": 9.891872115533352e-05}
{"train_loss": 0.5272990465164185, "global_step": 18469, "epoch": 207, "lr": 9.891860123834157e-05}
{"train_loss": 0.5190337896347046, "global_step": 18470, "epoch": 207, "lr": 9.891848131477313e-05}
{"train_loss": 0.48758623003959656, "global_step": 18471, "epoch": 207, "lr": 9.891836138462817e-05}
{"train_loss": 0.6170290112495422, "global_step": 18472, "epoch": 207, "lr": 9.891824144790676e-05}
{"train_loss": 0.5774950385093689, "global_step": 18473, "epoch": 207, "lr": 9.891812150460887e-05}
{"train_loss": 0.6062858700752258, "global_step": 18474, "epoch": 207, "lr": 9.891800155473454e-05}
{"train_loss": 0.49776342511177063, "global_step": 18475, "epoch": 207, "lr": 9.891788159828378e-05}
{"train_loss": 0.4931270480155945, "global_step": 18476, "epoch": 207, "lr": 9.891776163525661e-05}
{"train_loss": 0.5732199549674988, "global_step": 18477, "epoch": 207, "lr": 9.891764166565305e-05}
{"train_loss": 0.5394890308380127, "global_step": 18478, "epoch": 207, "lr": 9.89175216894731e-05}
{"train_loss": 0.4096852242946625, "global_step": 18479, "epoch": 207, "lr": 9.891740170671679e-05}
{"train_loss": 0.5132482051849365, "global_step": 18480, "epoch": 207, "lr": 9.891728171738412e-05}
{"train_loss": 0.644067108631134, "global_step": 18481, "epoch": 207, "lr": 9.891716172147512e-05}
{"train_loss": 0.6496900916099548, "global_step": 18482, "epoch": 207, "lr": 9.891704171898981e-05}
{"train_loss": 0.532194972038269, "global_step": 18483, "epoch": 207, "lr": 9.891692170992822e-05}
{"train_loss": 0.44188886880874634, "global_step": 18484, "epoch": 207, "lr": 9.891680169429031e-05}
{"train_loss": 0.4595224857330322, "global_step": 18485, "epoch": 207, "lr": 9.891668167207615e-05}
{"train_loss": 0.5389541983604431, "global_step": 18486, "epoch": 207, "lr": 9.891656164328574e-05}
{"train_loss": 0.663578450679779, "global_step": 18487, "epoch": 207, "lr": 9.891644160791908e-05}
{"train_loss": 0.5129103064537048, "global_step": 18488, "epoch": 207, "lr": 9.891632156597621e-05}
{"train_loss": 0.5457158088684082, "global_step": 18489, "epoch": 207, "lr": 9.891620151745714e-05}
{"train_loss": 0.5131325125694275, "global_step": 18490, "epoch": 207, "lr": 9.89160814623619e-05}
{"train_loss": 0.6006705164909363, "global_step": 18491, "epoch": 207, "lr": 9.891596140069046e-05}
{"train_loss": 0.6377744674682617, "global_step": 18492, "epoch": 207, "lr": 9.891584133244287e-05}
{"train_loss": 0.5188423991203308, "global_step": 18493, "epoch": 207, "lr": 9.891572125761914e-05}
{"train_loss": 0.6517567038536072, "global_step": 18494, "epoch": 207, "lr": 9.891560117621929e-05}
{"train_loss": 0.6192001104354858, "global_step": 18495, "epoch": 207, "lr": 9.891548108824335e-05}
{"train_loss": 0.6172346472740173, "global_step": 18496, "epoch": 207, "lr": 9.89153609936913e-05}
{"train_loss": 0.4901408553123474, "global_step": 18497, "epoch": 207, "lr": 9.891524089256319e-05}
{"train_loss": 0.41443711519241333, "global_step": 18498, "epoch": 207, "lr": 9.8915120784859e-05}
{"train_loss": 0.5564078092575073, "global_step": 18499, "epoch": 207, "lr": 9.89150006705788e-05}
{"train_loss": 0.49906155467033386, "global_step": 18500, "epoch": 207, "lr": 9.891488054972256e-05}
{"train_loss": 0.7743242979049683, "global_step": 18501, "epoch": 207, "lr": 9.89147604222903e-05}
{"train_loss": 0.5220709443092346, "global_step": 18502, "epoch": 207, "lr": 9.891464028828205e-05}
{"train_loss": 0.6846905946731567, "global_step": 18503, "epoch": 207, "lr": 9.891452014769783e-05}
{"train_loss": 0.570019543170929, "global_step": 18504, "epoch": 207, "lr": 9.891440000053764e-05}
{"train_loss": 0.621825635433197, "global_step": 18505, "epoch": 207, "lr": 9.891427984680153e-05}
{"train_loss": 0.7005214095115662, "global_step": 18506, "epoch": 207, "lr": 9.891415968648948e-05}
{"train_loss": 0.5562099814414978, "global_step": 18507, "epoch": 207, "lr": 9.89140395196015e-05}
{"train_loss": 0.6809149980545044, "global_step": 18508, "epoch": 207, "lr": 9.891391934613762e-05}
{"train_loss": 0.4420941472053528, "global_step": 18509, "epoch": 207, "lr": 9.891379916609789e-05}
{"train_loss": 0.5231808423995972, "global_step": 18510, "epoch": 207, "lr": 9.891367897948228e-05}
{"train_loss": 0.5330617347460115, "global_step": 18511, "epoch": 207, "lr": 9.89135587862908e-05, "val_loss": 1.6749565601348877}
{"train_loss": 0.58002108335495, "global_step": 18512, "epoch": 208, "lr": 9.891343858652352e-05}
{"train_loss": 0.5427403450012207, "global_step": 18513, "epoch": 208, "lr": 9.891331838018041e-05}
{"train_loss": 0.5156542062759399, "global_step": 18514, "epoch": 208, "lr": 9.891319816726151e-05}
{"train_loss": 0.6040388941764832, "global_step": 18515, "epoch": 208, "lr": 9.891307794776683e-05}
{"train_loss": 0.3597865700721741, "global_step": 18516, "epoch": 208, "lr": 9.891295772169636e-05}
{"train_loss": 0.5171151757240295, "global_step": 18517, "epoch": 208, "lr": 9.891283748905016e-05}
{"train_loss": 0.5542169809341431, "global_step": 18518, "epoch": 208, "lr": 9.891271724982822e-05}
{"train_loss": 0.5930845737457275, "global_step": 18519, "epoch": 208, "lr": 9.891259700403055e-05}
{"train_loss": 0.5950967669487, "global_step": 18520, "epoch": 208, "lr": 9.89124767516572e-05}
{"train_loss": 0.4961244761943817, "global_step": 18521, "epoch": 208, "lr": 9.891235649270815e-05}
{"train_loss": 0.4769158959388733, "global_step": 18522, "epoch": 208, "lr": 9.891223622718342e-05}
{"train_loss": 0.45656150579452515, "global_step": 18523, "epoch": 208, "lr": 9.891211595508305e-05}
{"train_loss": 0.5322186350822449, "global_step": 18524, "epoch": 208, "lr": 9.891199567640705e-05}
{"train_loss": 0.5079527497291565, "global_step": 18525, "epoch": 208, "lr": 9.891187539115541e-05}
{"train_loss": 0.6487843990325928, "global_step": 18526, "epoch": 208, "lr": 9.891175509932817e-05}
{"train_loss": 0.6011942028999329, "global_step": 18527, "epoch": 208, "lr": 9.891163480092533e-05}
{"train_loss": 0.6499142646789551, "global_step": 18528, "epoch": 208, "lr": 9.891151449594695e-05}
{"train_loss": 0.459208607673645, "global_step": 18529, "epoch": 208, "lr": 9.891139418439299e-05}
{"train_loss": 0.6142414808273315, "global_step": 18530, "epoch": 208, "lr": 9.891127386626349e-05}
{"train_loss": 0.48920440673828125, "global_step": 18531, "epoch": 208, "lr": 9.891115354155847e-05}
{"train_loss": 0.6531127691268921, "global_step": 18532, "epoch": 208, "lr": 9.891103321027793e-05}
{"train_loss": 0.5294780135154724, "global_step": 18533, "epoch": 208, "lr": 9.891091287242191e-05}
{"train_loss": 0.5595875978469849, "global_step": 18534, "epoch": 208, "lr": 9.89107925279904e-05}
{"train_loss": 0.5747648477554321, "global_step": 18535, "epoch": 208, "lr": 9.891067217698344e-05}
{"train_loss": 0.4653429090976715, "global_step": 18536, "epoch": 208, "lr": 9.891055181940105e-05}
{"train_loss": 0.519855260848999, "global_step": 18537, "epoch": 208, "lr": 9.891043145524321e-05}
{"train_loss": 0.5542253255844116, "global_step": 18538, "epoch": 208, "lr": 9.891031108450997e-05}
{"train_loss": 0.5020818710327148, "global_step": 18539, "epoch": 208, "lr": 9.891019070720133e-05}
{"train_loss": 0.354295551776886, "global_step": 18540, "epoch": 208, "lr": 9.891007032331732e-05}
{"train_loss": 0.5104039311408997, "global_step": 18541, "epoch": 208, "lr": 9.890994993285793e-05}
{"train_loss": 0.46981868147850037, "global_step": 18542, "epoch": 208, "lr": 9.890982953582321e-05}
{"train_loss": 0.6012566089630127, "global_step": 18543, "epoch": 208, "lr": 9.890970913221315e-05}
{"train_loss": 0.42514321208000183, "global_step": 18544, "epoch": 208, "lr": 9.890958872202778e-05}
{"train_loss": 0.5416014790534973, "global_step": 18545, "epoch": 208, "lr": 9.890946830526711e-05}
{"train_loss": 0.5433884263038635, "global_step": 18546, "epoch": 208, "lr": 9.890934788193116e-05}
{"train_loss": 0.4504791498184204, "global_step": 18547, "epoch": 208, "lr": 9.890922745201993e-05}
{"train_loss": 0.49930307269096375, "global_step": 18548, "epoch": 208, "lr": 9.890910701553348e-05}
{"train_loss": 0.5951938629150391, "global_step": 18549, "epoch": 208, "lr": 9.890898657247178e-05}
{"train_loss": 0.4557492136955261, "global_step": 18550, "epoch": 208, "lr": 9.890886612283485e-05}
{"train_loss": 0.5081411600112915, "global_step": 18551, "epoch": 208, "lr": 9.890874566662274e-05}
{"train_loss": 0.4638538956642151, "global_step": 18552, "epoch": 208, "lr": 9.890862520383543e-05}
{"train_loss": 0.5642778873443604, "global_step": 18553, "epoch": 208, "lr": 9.890850473447298e-05}
{"train_loss": 0.4373669922351837, "global_step": 18554, "epoch": 208, "lr": 9.890838425853535e-05}
{"train_loss": 0.621051013469696, "global_step": 18555, "epoch": 208, "lr": 9.890826377602259e-05}
{"train_loss": 0.38155049085617065, "global_step": 18556, "epoch": 208, "lr": 9.890814328693471e-05}
{"train_loss": 0.4744367003440857, "global_step": 18557, "epoch": 208, "lr": 9.890802279127173e-05}
{"train_loss": 0.6793617010116577, "global_step": 18558, "epoch": 208, "lr": 9.890790228903366e-05}
{"train_loss": 0.518366813659668, "global_step": 18559, "epoch": 208, "lr": 9.890778178022051e-05}
{"train_loss": 0.5040444135665894, "global_step": 18560, "epoch": 208, "lr": 9.890766126483231e-05}
{"train_loss": 0.5173807740211487, "global_step": 18561, "epoch": 208, "lr": 9.890754074286908e-05}
{"train_loss": 0.593681275844574, "global_step": 18562, "epoch": 208, "lr": 9.890742021433081e-05}
{"train_loss": 0.4286215603351593, "global_step": 18563, "epoch": 208, "lr": 9.890729967921755e-05}
{"train_loss": 0.48838841915130615, "global_step": 18564, "epoch": 208, "lr": 9.89071791375293e-05}
{"train_loss": 0.5758805274963379, "global_step": 18565, "epoch": 208, "lr": 9.890705858926607e-05}
{"train_loss": 0.5598655939102173, "global_step": 18566, "epoch": 208, "lr": 9.890693803442787e-05}
{"train_loss": 0.5600429773330688, "global_step": 18567, "epoch": 208, "lr": 9.890681747301475e-05}
{"train_loss": 0.5545181632041931, "global_step": 18568, "epoch": 208, "lr": 9.890669690502669e-05}
{"train_loss": 0.47597578167915344, "global_step": 18569, "epoch": 208, "lr": 9.890657633046373e-05}
{"train_loss": 0.4588916003704071, "global_step": 18570, "epoch": 208, "lr": 9.890645574932587e-05}
{"train_loss": 0.46869584918022156, "global_step": 18571, "epoch": 208, "lr": 9.890633516161314e-05}
{"train_loss": 0.4497232735157013, "global_step": 18572, "epoch": 208, "lr": 9.890621456732554e-05}
{"train_loss": 0.4971078038215637, "global_step": 18573, "epoch": 208, "lr": 9.890609396646311e-05}
{"train_loss": 0.4401301443576813, "global_step": 18574, "epoch": 208, "lr": 9.890597335902584e-05}
{"train_loss": 0.5385003089904785, "global_step": 18575, "epoch": 208, "lr": 9.890585274501376e-05}
{"train_loss": 0.5590474009513855, "global_step": 18576, "epoch": 208, "lr": 9.890573212442688e-05}
{"train_loss": 0.648030698299408, "global_step": 18577, "epoch": 208, "lr": 9.890561149726522e-05}
{"train_loss": 0.44855281710624695, "global_step": 18578, "epoch": 208, "lr": 9.890549086352881e-05}
{"train_loss": 0.6438572406768799, "global_step": 18579, "epoch": 208, "lr": 9.890537022321765e-05}
{"train_loss": 0.5138624906539917, "global_step": 18580, "epoch": 208, "lr": 9.890524957633173e-05}
{"train_loss": 0.5529423952102661, "global_step": 18581, "epoch": 208, "lr": 9.890512892287113e-05}
{"train_loss": 0.5554802417755127, "global_step": 18582, "epoch": 208, "lr": 9.890500826283581e-05}
{"train_loss": 0.49554529786109924, "global_step": 18583, "epoch": 208, "lr": 9.890488759622582e-05}
{"train_loss": 0.5405638217926025, "global_step": 18584, "epoch": 208, "lr": 9.890476692304117e-05}
{"train_loss": 0.454105019569397, "global_step": 18585, "epoch": 208, "lr": 9.890464624328186e-05}
{"train_loss": 0.40335792303085327, "global_step": 18586, "epoch": 208, "lr": 9.890452555694792e-05}
{"train_loss": 0.5920273065567017, "global_step": 18587, "epoch": 208, "lr": 9.890440486403936e-05}
{"train_loss": 0.4758688807487488, "global_step": 18588, "epoch": 208, "lr": 9.89042841645562e-05}
{"train_loss": 0.6088394522666931, "global_step": 18589, "epoch": 208, "lr": 9.890416345849846e-05}
{"train_loss": 0.44708457589149475, "global_step": 18590, "epoch": 208, "lr": 9.890404274586615e-05}
{"train_loss": 0.5338348746299744, "global_step": 18591, "epoch": 208, "lr": 9.890392202665928e-05}
{"train_loss": 0.5375767946243286, "global_step": 18592, "epoch": 208, "lr": 9.890380130087788e-05}
{"train_loss": 0.5536959171295166, "global_step": 18593, "epoch": 208, "lr": 9.890368056852195e-05}
{"train_loss": 0.614361584186554, "global_step": 18594, "epoch": 208, "lr": 9.890355982959152e-05}
{"train_loss": 0.47075119614601135, "global_step": 18595, "epoch": 208, "lr": 9.890343908408662e-05}
{"train_loss": 0.4396387040615082, "global_step": 18596, "epoch": 208, "lr": 9.890331833200722e-05}
{"train_loss": 0.4544481039047241, "global_step": 18597, "epoch": 208, "lr": 9.890319757335337e-05}
{"train_loss": 0.581262469291687, "global_step": 18598, "epoch": 208, "lr": 9.890307680812511e-05}
{"train_loss": 0.4853236675262451, "global_step": 18599, "epoch": 208, "lr": 9.89029560363224e-05}
{"train_loss": 0.523895040991601, "global_step": 18600, "epoch": 208, "lr": 9.89028352579453e-05, "val_loss": 1.7262214422225952}
{"train_loss": 0.5383787751197815, "global_step": 18601, "epoch": 209, "lr": 9.89027144729938e-05}
{"train_loss": 0.6576065421104431, "global_step": 18602, "epoch": 209, "lr": 9.890259368146793e-05}
{"train_loss": 0.47307199239730835, "global_step": 18603, "epoch": 209, "lr": 9.890247288336769e-05}
{"train_loss": 0.5166983008384705, "global_step": 18604, "epoch": 209, "lr": 9.890235207869311e-05}
{"train_loss": 0.4664592146873474, "global_step": 18605, "epoch": 209, "lr": 9.890223126744422e-05}
{"train_loss": 0.4530816972255707, "global_step": 18606, "epoch": 209, "lr": 9.890211044962102e-05}
{"train_loss": 0.46952691674232483, "global_step": 18607, "epoch": 209, "lr": 9.890198962522352e-05}
{"train_loss": 0.41829484701156616, "global_step": 18608, "epoch": 209, "lr": 9.890186879425174e-05}
{"train_loss": 0.5030539631843567, "global_step": 18609, "epoch": 209, "lr": 9.89017479567057e-05}
{"train_loss": 0.4289383292198181, "global_step": 18610, "epoch": 209, "lr": 9.890162711258543e-05}
{"train_loss": 0.49640747904777527, "global_step": 18611, "epoch": 209, "lr": 9.890150626189092e-05}
{"train_loss": 0.5235729813575745, "global_step": 18612, "epoch": 209, "lr": 9.89013854046222e-05}
{"train_loss": 0.3983388841152191, "global_step": 18613, "epoch": 209, "lr": 9.890126454077928e-05}
{"train_loss": 0.4519684910774231, "global_step": 18614, "epoch": 209, "lr": 9.89011436703622e-05}
{"train_loss": 0.3427252173423767, "global_step": 18615, "epoch": 209, "lr": 9.890102279337094e-05}
{"train_loss": 0.5761446952819824, "global_step": 18616, "epoch": 209, "lr": 9.890090190980554e-05}
{"train_loss": 0.5371894836425781, "global_step": 18617, "epoch": 209, "lr": 9.8900781019666e-05}
{"train_loss": 0.5828444361686707, "global_step": 18618, "epoch": 209, "lr": 9.890066012295236e-05}
{"train_loss": 0.49481549859046936, "global_step": 18619, "epoch": 209, "lr": 9.890053921966462e-05}
{"train_loss": 0.5183411836624146, "global_step": 18620, "epoch": 209, "lr": 9.890041830980279e-05}
{"train_loss": 0.3598085045814514, "global_step": 18621, "epoch": 209, "lr": 9.890029739336691e-05}
{"train_loss": 0.5392396450042725, "global_step": 18622, "epoch": 209, "lr": 9.890017647035697e-05}
{"train_loss": 0.4195168912410736, "global_step": 18623, "epoch": 209, "lr": 9.890005554077298e-05}
{"train_loss": 0.4934636652469635, "global_step": 18624, "epoch": 209, "lr": 9.8899934604615e-05}
{"train_loss": 0.5057125091552734, "global_step": 18625, "epoch": 209, "lr": 9.889981366188302e-05}
{"train_loss": 0.4350114166736603, "global_step": 18626, "epoch": 209, "lr": 9.889969271257704e-05}
{"train_loss": 0.4785134196281433, "global_step": 18627, "epoch": 209, "lr": 9.88995717566971e-05}
{"train_loss": 0.5440578460693359, "global_step": 18628, "epoch": 209, "lr": 9.88994507942432e-05}
{"train_loss": 0.5377660393714905, "global_step": 18629, "epoch": 209, "lr": 9.889932982521537e-05}
{"train_loss": 0.5143615007400513, "global_step": 18630, "epoch": 209, "lr": 9.889920884961363e-05}
{"train_loss": 0.4743151366710663, "global_step": 18631, "epoch": 209, "lr": 9.889908786743798e-05}
{"train_loss": 0.4200250804424286, "global_step": 18632, "epoch": 209, "lr": 9.889896687868844e-05}
{"train_loss": 0.571922242641449, "global_step": 18633, "epoch": 209, "lr": 9.889884588336505e-05}
{"train_loss": 0.4295331537723541, "global_step": 18634, "epoch": 209, "lr": 9.889872488146779e-05}
{"train_loss": 0.4364323616027832, "global_step": 18635, "epoch": 209, "lr": 9.889860387299669e-05}
{"train_loss": 0.48604097962379456, "global_step": 18636, "epoch": 209, "lr": 9.889848285795177e-05}
{"train_loss": 0.5512160658836365, "global_step": 18637, "epoch": 209, "lr": 9.889836183633305e-05}
{"train_loss": 0.5019062757492065, "global_step": 18638, "epoch": 209, "lr": 9.889824080814053e-05}
{"train_loss": 0.4462108910083771, "global_step": 18639, "epoch": 209, "lr": 9.889811977337424e-05}
{"train_loss": 0.539998471736908, "global_step": 18640, "epoch": 209, "lr": 9.88979987320342e-05}
{"train_loss": 0.4939645528793335, "global_step": 18641, "epoch": 209, "lr": 9.889787768412042e-05}
{"train_loss": 0.3902438282966614, "global_step": 18642, "epoch": 209, "lr": 9.889775662963292e-05}
{"train_loss": 0.4973209798336029, "global_step": 18643, "epoch": 209, "lr": 9.889763556857172e-05}
{"train_loss": 0.4675041437149048, "global_step": 18644, "epoch": 209, "lr": 9.889751450093681e-05}
{"train_loss": 0.43502822518348694, "global_step": 18645, "epoch": 209, "lr": 9.889739342672823e-05}
{"train_loss": 0.5818923115730286, "global_step": 18646, "epoch": 209, "lr": 9.889727234594599e-05}
{"train_loss": 0.5340906381607056, "global_step": 18647, "epoch": 209, "lr": 9.889715125859011e-05}
{"train_loss": 0.6408193707466125, "global_step": 18648, "epoch": 209, "lr": 9.88970301646606e-05}
{"train_loss": 0.5759638547897339, "global_step": 18649, "epoch": 209, "lr": 9.88969090641575e-05}
{"train_loss": 0.4915733337402344, "global_step": 18650, "epoch": 209, "lr": 9.889678795708079e-05}
{"train_loss": 0.5561618208885193, "global_step": 18651, "epoch": 209, "lr": 9.889666684343051e-05}
{"train_loss": 0.5421088933944702, "global_step": 18652, "epoch": 209, "lr": 9.889654572320665e-05}
{"train_loss": 0.5085239410400391, "global_step": 18653, "epoch": 209, "lr": 9.889642459640927e-05}
{"train_loss": 0.5125220417976379, "global_step": 18654, "epoch": 209, "lr": 9.889630346303836e-05}
{"train_loss": 0.521281898021698, "global_step": 18655, "epoch": 209, "lr": 9.889618232309393e-05}
{"train_loss": 0.5039969086647034, "global_step": 18656, "epoch": 209, "lr": 9.8896061176576e-05}
{"train_loss": 0.462604820728302, "global_step": 18657, "epoch": 209, "lr": 9.889594002348458e-05}
{"train_loss": 0.5454331040382385, "global_step": 18658, "epoch": 209, "lr": 9.889581886381971e-05}
{"train_loss": 0.47203487157821655, "global_step": 18659, "epoch": 209, "lr": 9.88956976975814e-05}
{"train_loss": 0.419119268655777, "global_step": 18660, "epoch": 209, "lr": 9.889557652476966e-05}
{"train_loss": 0.6046348810195923, "global_step": 18661, "epoch": 209, "lr": 9.88954553453845e-05}
{"train_loss": 0.4743375778198242, "global_step": 18662, "epoch": 209, "lr": 9.889533415942594e-05}
{"train_loss": 0.545647382736206, "global_step": 18663, "epoch": 209, "lr": 9.889521296689399e-05}
{"train_loss": 0.48397985100746155, "global_step": 18664, "epoch": 209, "lr": 9.889509176778869e-05}
{"train_loss": 0.530382513999939, "global_step": 18665, "epoch": 209, "lr": 9.889497056211003e-05}
{"train_loss": 0.446837842464447, "global_step": 18666, "epoch": 209, "lr": 9.889484934985804e-05}
{"train_loss": 0.5480064153671265, "global_step": 18667, "epoch": 209, "lr": 9.889472813103274e-05}
{"train_loss": 0.5337235331535339, "global_step": 18668, "epoch": 209, "lr": 9.889460690563415e-05}
{"train_loss": 0.46767228841781616, "global_step": 18669, "epoch": 209, "lr": 9.889448567366225e-05}
{"train_loss": 0.506483793258667, "global_step": 18670, "epoch": 209, "lr": 9.88943644351171e-05}
{"train_loss": 0.4333469271659851, "global_step": 18671, "epoch": 209, "lr": 9.889424318999869e-05}
{"train_loss": 0.5100962519645691, "global_step": 18672, "epoch": 209, "lr": 9.889412193830705e-05}
{"train_loss": 0.4326004385948181, "global_step": 18673, "epoch": 209, "lr": 9.889400068004218e-05}
{"train_loss": 0.5808857679367065, "global_step": 18674, "epoch": 209, "lr": 9.889387941520411e-05}
{"train_loss": 0.3594612181186676, "global_step": 18675, "epoch": 209, "lr": 9.889375814379287e-05}
{"train_loss": 0.5059764385223389, "global_step": 18676, "epoch": 209, "lr": 9.889363686580845e-05}
{"train_loss": 0.46727532148361206, "global_step": 18677, "epoch": 209, "lr": 9.889351558125087e-05}
{"train_loss": 0.4093455672264099, "global_step": 18678, "epoch": 209, "lr": 9.889339429012017e-05}
{"train_loss": 0.353607714176178, "global_step": 18679, "epoch": 209, "lr": 9.889327299241633e-05}
{"train_loss": 0.4729253351688385, "global_step": 18680, "epoch": 209, "lr": 9.88931516881394e-05}
{"train_loss": 0.4280737340450287, "global_step": 18681, "epoch": 209, "lr": 9.889303037728936e-05}
{"train_loss": 0.5359145998954773, "global_step": 18682, "epoch": 209, "lr": 9.889290905986628e-05}
{"train_loss": 0.39931520819664, "global_step": 18683, "epoch": 209, "lr": 9.889278773587012e-05}
{"train_loss": 0.6045169830322266, "global_step": 18684, "epoch": 209, "lr": 9.889266640530092e-05}
{"train_loss": 0.4200616478919983, "global_step": 18685, "epoch": 209, "lr": 9.889254506815871e-05}
{"train_loss": 0.5645098686218262, "global_step": 18686, "epoch": 209, "lr": 9.88924237244435e-05}
{"train_loss": 0.44551581144332886, "global_step": 18687, "epoch": 209, "lr": 9.889230237415527e-05}
{"train_loss": 0.5551764965057373, "global_step": 18688, "epoch": 209, "lr": 9.889218101729409e-05}
{"train_loss": 0.4925072404106012, "global_step": 18689, "epoch": 209, "lr": 9.889205965385995e-05, "val_loss": 1.6644113063812256}
{"train_loss": 0.4133443832397461, "global_step": 18690, "epoch": 210, "lr": 9.889193828385286e-05}
{"train_loss": 0.5366146564483643, "global_step": 18691, "epoch": 210, "lr": 9.889181690727285e-05}
{"train_loss": 0.509820818901062, "global_step": 18692, "epoch": 210, "lr": 9.889169552411993e-05}
{"train_loss": 0.5368265509605408, "global_step": 18693, "epoch": 210, "lr": 9.889157413439412e-05}
{"train_loss": 0.46801286935806274, "global_step": 18694, "epoch": 210, "lr": 9.889145273809544e-05}
{"train_loss": 0.5032414793968201, "global_step": 18695, "epoch": 210, "lr": 9.889133133522388e-05}
{"train_loss": 0.4469819664955139, "global_step": 18696, "epoch": 210, "lr": 9.889120992577949e-05}
{"train_loss": 0.562811553478241, "global_step": 18697, "epoch": 210, "lr": 9.889108850976228e-05}
{"train_loss": 0.474920392036438, "global_step": 18698, "epoch": 210, "lr": 9.889096708717225e-05}
{"train_loss": 0.5163040161132812, "global_step": 18699, "epoch": 210, "lr": 9.889084565800941e-05}
{"train_loss": 0.3159217834472656, "global_step": 18700, "epoch": 210, "lr": 9.889072422227381e-05}
{"train_loss": 0.5066689848899841, "global_step": 18701, "epoch": 210, "lr": 9.889060277996545e-05}
{"train_loss": 0.5285754203796387, "global_step": 18702, "epoch": 210, "lr": 9.889048133108433e-05}
{"train_loss": 0.4841287136077881, "global_step": 18703, "epoch": 210, "lr": 9.88903598756305e-05}
{"train_loss": 0.5297326445579529, "global_step": 18704, "epoch": 210, "lr": 9.889023841360395e-05}
{"train_loss": 0.49577850103378296, "global_step": 18705, "epoch": 210, "lr": 9.88901169450047e-05}
{"train_loss": 0.46466681361198425, "global_step": 18706, "epoch": 210, "lr": 9.888999546983278e-05}
{"train_loss": 0.4497702121734619, "global_step": 18707, "epoch": 210, "lr": 9.888987398808818e-05}
{"train_loss": 0.4971172511577606, "global_step": 18708, "epoch": 210, "lr": 9.888975249977093e-05}
{"train_loss": 0.3945627212524414, "global_step": 18709, "epoch": 210, "lr": 9.888963100488106e-05}
{"train_loss": 0.44157013297080994, "global_step": 18710, "epoch": 210, "lr": 9.888950950341858e-05}
{"train_loss": 0.4368876516819, "global_step": 18711, "epoch": 210, "lr": 9.888938799538347e-05}
{"train_loss": 0.5587694644927979, "global_step": 18712, "epoch": 210, "lr": 9.888926648077581e-05}
{"train_loss": 0.5915970802307129, "global_step": 18713, "epoch": 210, "lr": 9.888914495959558e-05}
{"train_loss": 0.5041902661323547, "global_step": 18714, "epoch": 210, "lr": 9.888902343184279e-05}
{"train_loss": 0.46897420287132263, "global_step": 18715, "epoch": 210, "lr": 9.888890189751748e-05}
{"train_loss": 0.6461099982261658, "global_step": 18716, "epoch": 210, "lr": 9.888878035661963e-05}
{"train_loss": 0.49984800815582275, "global_step": 18717, "epoch": 210, "lr": 9.88886588091493e-05}
{"train_loss": 0.584503173828125, "global_step": 18718, "epoch": 210, "lr": 9.888853725510648e-05}
{"train_loss": 0.5389961004257202, "global_step": 18719, "epoch": 210, "lr": 9.88884156944912e-05}
{"train_loss": 0.4494887888431549, "global_step": 18720, "epoch": 210, "lr": 9.888829412730347e-05}
{"train_loss": 0.45214715600013733, "global_step": 18721, "epoch": 210, "lr": 9.888817255354328e-05}
{"train_loss": 0.4802815020084381, "global_step": 18722, "epoch": 210, "lr": 9.888805097321068e-05}
{"train_loss": 0.4138498306274414, "global_step": 18723, "epoch": 210, "lr": 9.888792938630568e-05}
{"train_loss": 0.5017278790473938, "global_step": 18724, "epoch": 210, "lr": 9.888780779282832e-05}
{"train_loss": 0.3246650993824005, "global_step": 18725, "epoch": 210, "lr": 9.888768619277855e-05}
{"train_loss": 0.531633734703064, "global_step": 18726, "epoch": 210, "lr": 9.888756458615645e-05}
{"train_loss": 0.6309030652046204, "global_step": 18727, "epoch": 210, "lr": 9.888744297296201e-05}
{"train_loss": 0.5363488793373108, "global_step": 18728, "epoch": 210, "lr": 9.888732135319525e-05}
{"train_loss": 0.5824627876281738, "global_step": 18729, "epoch": 210, "lr": 9.888719972685618e-05}
{"train_loss": 0.5269361734390259, "global_step": 18730, "epoch": 210, "lr": 9.888707809394482e-05}
{"train_loss": 0.5170376896858215, "global_step": 18731, "epoch": 210, "lr": 9.888695645446119e-05}
{"train_loss": 0.38661208748817444, "global_step": 18732, "epoch": 210, "lr": 9.888683480840531e-05}
{"train_loss": 0.6668745279312134, "global_step": 18733, "epoch": 210, "lr": 9.888671315577718e-05}
{"train_loss": 0.470039963722229, "global_step": 18734, "epoch": 210, "lr": 9.888659149657684e-05}
{"train_loss": 0.4358022212982178, "global_step": 18735, "epoch": 210, "lr": 9.888646983080428e-05}
{"train_loss": 0.5824577808380127, "global_step": 18736, "epoch": 210, "lr": 9.888634815845954e-05}
{"train_loss": 0.6110964417457581, "global_step": 18737, "epoch": 210, "lr": 9.888622647954263e-05}
{"train_loss": 0.4988861083984375, "global_step": 18738, "epoch": 210, "lr": 9.888610479405356e-05}
{"train_loss": 0.5455679297447205, "global_step": 18739, "epoch": 210, "lr": 9.888598310199234e-05}
{"train_loss": 0.4895562529563904, "global_step": 18740, "epoch": 210, "lr": 9.8885861403359e-05}
{"train_loss": 0.55499267578125, "global_step": 18741, "epoch": 210, "lr": 9.888573969815356e-05}
{"train_loss": 0.48849189281463623, "global_step": 18742, "epoch": 210, "lr": 9.888561798637601e-05}
{"train_loss": 0.3902173340320587, "global_step": 18743, "epoch": 210, "lr": 9.88854962680264e-05}
{"train_loss": 0.5098541378974915, "global_step": 18744, "epoch": 210, "lr": 9.888537454310473e-05}
{"train_loss": 0.47197410464286804, "global_step": 18745, "epoch": 210, "lr": 9.8885252811611e-05}
{"train_loss": 0.5351119041442871, "global_step": 18746, "epoch": 210, "lr": 9.888513107354527e-05}
{"train_loss": 0.4346228837966919, "global_step": 18747, "epoch": 210, "lr": 9.888500932890752e-05}
{"train_loss": 0.5774102807044983, "global_step": 18748, "epoch": 210, "lr": 9.888488757769776e-05}
{"train_loss": 0.5002592206001282, "global_step": 18749, "epoch": 210, "lr": 9.888476581991604e-05}
{"train_loss": 0.45806872844696045, "global_step": 18750, "epoch": 210, "lr": 9.888464405556237e-05}
{"train_loss": 0.5177584290504456, "global_step": 18751, "epoch": 210, "lr": 9.888452228463673e-05}
{"train_loss": 0.5128658413887024, "global_step": 18752, "epoch": 210, "lr": 9.888440050713917e-05}
{"train_loss": 0.47790831327438354, "global_step": 18753, "epoch": 210, "lr": 9.88842787230697e-05}
{"train_loss": 0.4837600290775299, "global_step": 18754, "epoch": 210, "lr": 9.888415693242834e-05}
{"train_loss": 0.5729033946990967, "global_step": 18755, "epoch": 210, "lr": 9.888403513521509e-05}
{"train_loss": 0.4681304395198822, "global_step": 18756, "epoch": 210, "lr": 9.888391333142998e-05}
{"train_loss": 0.7623690962791443, "global_step": 18757, "epoch": 210, "lr": 9.888379152107303e-05}
{"train_loss": 0.4850539565086365, "global_step": 18758, "epoch": 210, "lr": 9.888366970414424e-05}
{"train_loss": 0.583097517490387, "global_step": 18759, "epoch": 210, "lr": 9.888354788064365e-05}
{"train_loss": 0.7058060765266418, "global_step": 18760, "epoch": 210, "lr": 9.888342605057126e-05}
{"train_loss": 0.5965548753738403, "global_step": 18761, "epoch": 210, "lr": 9.888330421392709e-05}
{"train_loss": 0.5809821486473083, "global_step": 18762, "epoch": 210, "lr": 9.888318237071115e-05}
{"train_loss": 0.5430048108100891, "global_step": 18763, "epoch": 210, "lr": 9.888306052092346e-05}
{"train_loss": 0.6272671222686768, "global_step": 18764, "epoch": 210, "lr": 9.888293866456404e-05}
{"train_loss": 0.5737541317939758, "global_step": 18765, "epoch": 210, "lr": 9.888281680163291e-05}
{"train_loss": 0.4964430332183838, "global_step": 18766, "epoch": 210, "lr": 9.888269493213008e-05}
{"train_loss": 0.56376051902771, "global_step": 18767, "epoch": 210, "lr": 9.888257305605556e-05}
{"train_loss": 0.560106635093689, "global_step": 18768, "epoch": 210, "lr": 9.888245117340938e-05}
{"train_loss": 0.4283045530319214, "global_step": 18769, "epoch": 210, "lr": 9.888232928419154e-05}
{"train_loss": 0.48677995800971985, "global_step": 18770, "epoch": 210, "lr": 9.888220738840208e-05}
{"train_loss": 0.5871544480323792, "global_step": 18771, "epoch": 210, "lr": 9.8882085486041e-05}
{"train_loss": 0.5323495268821716, "global_step": 18772, "epoch": 210, "lr": 9.888196357710831e-05}
{"train_loss": 0.45551908016204834, "global_step": 18773, "epoch": 210, "lr": 9.888184166160406e-05}
{"train_loss": 0.5890539288520813, "global_step": 18774, "epoch": 210, "lr": 9.888171973952822e-05}
{"train_loss": 0.6108703017234802, "global_step": 18775, "epoch": 210, "lr": 9.888159781088083e-05}
{"train_loss": 0.5054956674575806, "global_step": 18776, "epoch": 210, "lr": 9.88814758756619e-05}
{"train_loss": 0.5311876535415649, "global_step": 18777, "epoch": 210, "lr": 9.888135393387147e-05}
{"train_loss": 0.5148163427797596, "global_step": 18778, "epoch": 210, "lr": 9.888123198550953e-05, "val_loss": 1.6832200288772583, "train_action_mse_error": 41.42491149902344}
{"train_loss": 0.4742034375667572, "global_step": 18779, "epoch": 211, "lr": 9.88811100305761e-05}
{"train_loss": 0.4912004768848419, "global_step": 18780, "epoch": 211, "lr": 9.88809880690712e-05}
{"train_loss": 0.5653447508811951, "global_step": 18781, "epoch": 211, "lr": 9.888086610099485e-05}
{"train_loss": 0.49733057618141174, "global_step": 18782, "epoch": 211, "lr": 9.888074412634706e-05}
{"train_loss": 0.5200132727622986, "global_step": 18783, "epoch": 211, "lr": 9.888062214512786e-05}
{"train_loss": 0.6119612455368042, "global_step": 18784, "epoch": 211, "lr": 9.888050015733725e-05}
{"train_loss": 0.5787375569343567, "global_step": 18785, "epoch": 211, "lr": 9.888037816297525e-05}
{"train_loss": 0.6189377903938293, "global_step": 18786, "epoch": 211, "lr": 9.888025616204188e-05}
{"train_loss": 0.546949028968811, "global_step": 18787, "epoch": 211, "lr": 9.888013415453716e-05}
{"train_loss": 0.4536589980125427, "global_step": 18788, "epoch": 211, "lr": 9.88800121404611e-05}
{"train_loss": 0.5026996731758118, "global_step": 18789, "epoch": 211, "lr": 9.887989011981371e-05}
{"train_loss": 0.6244291663169861, "global_step": 18790, "epoch": 211, "lr": 9.887976809259502e-05}
{"train_loss": 0.44088661670684814, "global_step": 18791, "epoch": 211, "lr": 9.887964605880505e-05}
{"train_loss": 0.49640953540802, "global_step": 18792, "epoch": 211, "lr": 9.887952401844378e-05}
{"train_loss": 0.44947969913482666, "global_step": 18793, "epoch": 211, "lr": 9.887940197151129e-05}
{"train_loss": 0.46320459246635437, "global_step": 18794, "epoch": 211, "lr": 9.887927991800754e-05}
{"train_loss": 0.5891416668891907, "global_step": 18795, "epoch": 211, "lr": 9.887915785793256e-05}
{"train_loss": 0.4529348909854889, "global_step": 18796, "epoch": 211, "lr": 9.887903579128638e-05}
{"train_loss": 0.5570861101150513, "global_step": 18797, "epoch": 211, "lr": 9.8878913718069e-05}
{"train_loss": 0.5171907544136047, "global_step": 18798, "epoch": 211, "lr": 9.887879163828047e-05}
{"train_loss": 0.47731584310531616, "global_step": 18799, "epoch": 211, "lr": 9.887866955192078e-05}
{"train_loss": 0.4274095892906189, "global_step": 18800, "epoch": 211, "lr": 9.887854745898992e-05}
{"train_loss": 0.513833224773407, "global_step": 18801, "epoch": 211, "lr": 9.887842535948797e-05}
{"train_loss": 0.5220499634742737, "global_step": 18802, "epoch": 211, "lr": 9.887830325341489e-05}
{"train_loss": 0.3917222321033478, "global_step": 18803, "epoch": 211, "lr": 9.887818114077072e-05}
{"train_loss": 0.5316406488418579, "global_step": 18804, "epoch": 211, "lr": 9.887805902155546e-05}
{"train_loss": 0.4703138768672943, "global_step": 18805, "epoch": 211, "lr": 9.887793689576917e-05}
{"train_loss": 0.47865983843803406, "global_step": 18806, "epoch": 211, "lr": 9.887781476341183e-05}
{"train_loss": 0.5914037823677063, "global_step": 18807, "epoch": 211, "lr": 9.887769262448345e-05}
{"train_loss": 0.4158400893211365, "global_step": 18808, "epoch": 211, "lr": 9.887757047898408e-05}
{"train_loss": 0.46741652488708496, "global_step": 18809, "epoch": 211, "lr": 9.88774483269137e-05}
{"train_loss": 0.48352953791618347, "global_step": 18810, "epoch": 211, "lr": 9.887732616827235e-05}
{"train_loss": 0.4018379747867584, "global_step": 18811, "epoch": 211, "lr": 9.887720400306003e-05}
{"train_loss": 0.49148502945899963, "global_step": 18812, "epoch": 211, "lr": 9.887708183127678e-05}
{"train_loss": 0.46240198612213135, "global_step": 18813, "epoch": 211, "lr": 9.887695965292259e-05}
{"train_loss": 0.44939514994621277, "global_step": 18814, "epoch": 211, "lr": 9.887683746799749e-05}
{"train_loss": 0.46970319747924805, "global_step": 18815, "epoch": 211, "lr": 9.88767152765015e-05}
{"train_loss": 0.4069964587688446, "global_step": 18816, "epoch": 211, "lr": 9.887659307843462e-05}
{"train_loss": 0.5635581016540527, "global_step": 18817, "epoch": 211, "lr": 9.88764708737969e-05}
{"train_loss": 0.4260529577732086, "global_step": 18818, "epoch": 211, "lr": 9.887634866258831e-05}
{"train_loss": 0.47424018383026123, "global_step": 18819, "epoch": 211, "lr": 9.88762264448089e-05}
{"train_loss": 0.5157960653305054, "global_step": 18820, "epoch": 211, "lr": 9.887610422045867e-05}
{"train_loss": 0.5367588996887207, "global_step": 18821, "epoch": 211, "lr": 9.887598198953766e-05}
{"train_loss": 0.4077669084072113, "global_step": 18822, "epoch": 211, "lr": 9.887585975204587e-05}
{"train_loss": 0.5210138559341431, "global_step": 18823, "epoch": 211, "lr": 9.887573750798332e-05}
{"train_loss": 0.40486788749694824, "global_step": 18824, "epoch": 211, "lr": 9.887561525735e-05}
{"train_loss": 0.5443163514137268, "global_step": 18825, "epoch": 211, "lr": 9.887549300014596e-05}
{"train_loss": 0.5717929601669312, "global_step": 18826, "epoch": 211, "lr": 9.887537073637121e-05}
{"train_loss": 0.5872540473937988, "global_step": 18827, "epoch": 211, "lr": 9.887524846602575e-05}
{"train_loss": 0.43308091163635254, "global_step": 18828, "epoch": 211, "lr": 9.887512618910961e-05}
{"train_loss": 0.5924220085144043, "global_step": 18829, "epoch": 211, "lr": 9.887500390562282e-05}
{"train_loss": 0.41464245319366455, "global_step": 18830, "epoch": 211, "lr": 9.887488161556537e-05}
{"train_loss": 0.4858628809452057, "global_step": 18831, "epoch": 211, "lr": 9.887475931893729e-05}
{"train_loss": 0.5248599052429199, "global_step": 18832, "epoch": 211, "lr": 9.887463701573861e-05}
{"train_loss": 0.47510480880737305, "global_step": 18833, "epoch": 211, "lr": 9.887451470596931e-05}
{"train_loss": 0.5491283535957336, "global_step": 18834, "epoch": 211, "lr": 9.887439238962943e-05}
{"train_loss": 0.37840381264686584, "global_step": 18835, "epoch": 211, "lr": 9.887427006671899e-05}
{"train_loss": 0.4550246000289917, "global_step": 18836, "epoch": 211, "lr": 9.8874147737238e-05}
{"train_loss": 0.5169740319252014, "global_step": 18837, "epoch": 211, "lr": 9.887402540118646e-05}
{"train_loss": 0.40154165029525757, "global_step": 18838, "epoch": 211, "lr": 9.887390305856442e-05}
{"train_loss": 0.4682992100715637, "global_step": 18839, "epoch": 211, "lr": 9.887378070937188e-05}
{"train_loss": 0.38330256938934326, "global_step": 18840, "epoch": 211, "lr": 9.887365835360884e-05}
{"train_loss": 0.5028014779090881, "global_step": 18841, "epoch": 211, "lr": 9.887353599127536e-05}
{"train_loss": 0.4519668221473694, "global_step": 18842, "epoch": 211, "lr": 9.887341362237141e-05}
{"train_loss": 0.44048064947128296, "global_step": 18843, "epoch": 211, "lr": 9.887329124689703e-05}
{"train_loss": 0.4460933208465576, "global_step": 18844, "epoch": 211, "lr": 9.887316886485222e-05}
{"train_loss": 0.5013192892074585, "global_step": 18845, "epoch": 211, "lr": 9.887304647623703e-05}
{"train_loss": 0.41787415742874146, "global_step": 18846, "epoch": 211, "lr": 9.887292408105145e-05}
{"train_loss": 0.5684369206428528, "global_step": 18847, "epoch": 211, "lr": 9.887280167929548e-05}
{"train_loss": 0.5609769821166992, "global_step": 18848, "epoch": 211, "lr": 9.887267927096917e-05}
{"train_loss": 0.7087523341178894, "global_step": 18849, "epoch": 211, "lr": 9.887255685607252e-05}
{"train_loss": 0.5361514687538147, "global_step": 18850, "epoch": 211, "lr": 9.887243443460557e-05}
{"train_loss": 0.421924889087677, "global_step": 18851, "epoch": 211, "lr": 9.887231200656828e-05}
{"train_loss": 0.5062335133552551, "global_step": 18852, "epoch": 211, "lr": 9.887218957196073e-05}
{"train_loss": 0.4250186085700989, "global_step": 18853, "epoch": 211, "lr": 9.887206713078292e-05}
{"train_loss": 0.569888710975647, "global_step": 18854, "epoch": 211, "lr": 9.887194468303484e-05}
{"train_loss": 0.6645662188529968, "global_step": 18855, "epoch": 211, "lr": 9.887182222871652e-05}
{"train_loss": 0.5975355505943298, "global_step": 18856, "epoch": 211, "lr": 9.887169976782799e-05}
{"train_loss": 0.4808587431907654, "global_step": 18857, "epoch": 211, "lr": 9.887157730036925e-05}
{"train_loss": 0.43898096680641174, "global_step": 18858, "epoch": 211, "lr": 9.887145482634031e-05}
{"train_loss": 0.5712432861328125, "global_step": 18859, "epoch": 211, "lr": 9.887133234574121e-05}
{"train_loss": 0.6262734532356262, "global_step": 18860, "epoch": 211, "lr": 9.887120985857195e-05}
{"train_loss": 0.4571079611778259, "global_step": 18861, "epoch": 211, "lr": 9.887108736483257e-05}
{"train_loss": 0.5962260365486145, "global_step": 18862, "epoch": 211, "lr": 9.887096486452306e-05}
{"train_loss": 0.6631026864051819, "global_step": 18863, "epoch": 211, "lr": 9.887084235764343e-05}
{"train_loss": 0.6186914443969727, "global_step": 18864, "epoch": 211, "lr": 9.887071984419371e-05}
{"train_loss": 0.49743300676345825, "global_step": 18865, "epoch": 211, "lr": 9.887059732417394e-05}
{"train_loss": 0.34104791283607483, "global_step": 18866, "epoch": 211, "lr": 9.88704747975841e-05}
{"train_loss": 0.5007385762220018, "global_step": 18867, "epoch": 211, "lr": 9.887035226442422e-05, "val_loss": 1.6314986944198608}
{"train_loss": 0.46664664149284363, "global_step": 18868, "epoch": 212, "lr": 9.887022972469431e-05}
{"train_loss": 0.4855853021144867, "global_step": 18869, "epoch": 212, "lr": 9.88701071783944e-05}
{"train_loss": 0.5060490369796753, "global_step": 18870, "epoch": 212, "lr": 9.88699846255245e-05}
{"train_loss": 0.45362159609794617, "global_step": 18871, "epoch": 212, "lr": 9.886986206608465e-05}
{"train_loss": 0.38133659958839417, "global_step": 18872, "epoch": 212, "lr": 9.88697395000748e-05}
{"train_loss": 0.41795614361763, "global_step": 18873, "epoch": 212, "lr": 9.886961692749503e-05}
{"train_loss": 0.453429251909256, "global_step": 18874, "epoch": 212, "lr": 9.886949434834533e-05}
{"train_loss": 0.409938782453537, "global_step": 18875, "epoch": 212, "lr": 9.886937176262573e-05}
{"train_loss": 0.5535781383514404, "global_step": 18876, "epoch": 212, "lr": 9.886924917033624e-05}
{"train_loss": 0.44069817662239075, "global_step": 18877, "epoch": 212, "lr": 9.886912657147687e-05}
{"train_loss": 0.41082099080085754, "global_step": 18878, "epoch": 212, "lr": 9.886900396604763e-05}
{"train_loss": 0.5972121953964233, "global_step": 18879, "epoch": 212, "lr": 9.886888135404857e-05}
{"train_loss": 0.5013921856880188, "global_step": 18880, "epoch": 212, "lr": 9.886875873547967e-05}
{"train_loss": 0.5096109509468079, "global_step": 18881, "epoch": 212, "lr": 9.886863611034096e-05}
{"train_loss": 0.42910999059677124, "global_step": 18882, "epoch": 212, "lr": 9.886851347863247e-05}
{"train_loss": 0.5195658802986145, "global_step": 18883, "epoch": 212, "lr": 9.88683908403542e-05}
{"train_loss": 0.49753305315971375, "global_step": 18884, "epoch": 212, "lr": 9.886826819550617e-05}
{"train_loss": 0.4824358820915222, "global_step": 18885, "epoch": 212, "lr": 9.886814554408838e-05}
{"train_loss": 0.6051856875419617, "global_step": 18886, "epoch": 212, "lr": 9.886802288610088e-05}
{"train_loss": 0.42351770401000977, "global_step": 18887, "epoch": 212, "lr": 9.886790022154367e-05}
{"train_loss": 0.5377508997917175, "global_step": 18888, "epoch": 212, "lr": 9.886777755041677e-05}
{"train_loss": 0.5586089491844177, "global_step": 18889, "epoch": 212, "lr": 9.886765487272018e-05}
{"train_loss": 0.4693247079849243, "global_step": 18890, "epoch": 212, "lr": 9.886753218845395e-05}
{"train_loss": 0.5428915619850159, "global_step": 18891, "epoch": 212, "lr": 9.886740949761805e-05}
{"train_loss": 0.5137635469436646, "global_step": 18892, "epoch": 212, "lr": 9.886728680021253e-05}
{"train_loss": 0.44785648584365845, "global_step": 18893, "epoch": 212, "lr": 9.886716409623741e-05}
{"train_loss": 0.4545157849788666, "global_step": 18894, "epoch": 212, "lr": 9.88670413856927e-05}
{"train_loss": 0.40096524357795715, "global_step": 18895, "epoch": 212, "lr": 9.88669186685784e-05}
{"train_loss": 0.4916168749332428, "global_step": 18896, "epoch": 212, "lr": 9.886679594489455e-05}
{"train_loss": 0.5937907099723816, "global_step": 18897, "epoch": 212, "lr": 9.886667321464114e-05}
{"train_loss": 0.543113648891449, "global_step": 18898, "epoch": 212, "lr": 9.88665504778182e-05}
{"train_loss": 0.42041072249412537, "global_step": 18899, "epoch": 212, "lr": 9.886642773442574e-05}
{"train_loss": 0.49748218059539795, "global_step": 18900, "epoch": 212, "lr": 9.88663049844638e-05}
{"train_loss": 0.4841329753398895, "global_step": 18901, "epoch": 212, "lr": 9.886618222793238e-05}
{"train_loss": 0.47263336181640625, "global_step": 18902, "epoch": 212, "lr": 9.88660594648315e-05}
{"train_loss": 0.5813339352607727, "global_step": 18903, "epoch": 212, "lr": 9.886593669516118e-05}
{"train_loss": 0.43521079421043396, "global_step": 18904, "epoch": 212, "lr": 9.88658139189214e-05}
{"train_loss": 0.5047063827514648, "global_step": 18905, "epoch": 212, "lr": 9.886569113611223e-05}
{"train_loss": 0.48223546147346497, "global_step": 18906, "epoch": 212, "lr": 9.886556834673367e-05}
{"train_loss": 0.5069548487663269, "global_step": 18907, "epoch": 212, "lr": 9.886544555078572e-05}
{"train_loss": 0.4978834092617035, "global_step": 18908, "epoch": 212, "lr": 9.88653227482684e-05}
{"train_loss": 0.5514190793037415, "global_step": 18909, "epoch": 212, "lr": 9.886519993918176e-05}
{"train_loss": 0.5456328988075256, "global_step": 18910, "epoch": 212, "lr": 9.886507712352576e-05}
{"train_loss": 0.49628371000289917, "global_step": 18911, "epoch": 212, "lr": 9.886495430130045e-05}
{"train_loss": 0.41355884075164795, "global_step": 18912, "epoch": 212, "lr": 9.886483147250584e-05}
{"train_loss": 0.5251708626747131, "global_step": 18913, "epoch": 212, "lr": 9.886470863714196e-05}
{"train_loss": 0.5762600302696228, "global_step": 18914, "epoch": 212, "lr": 9.886458579520882e-05}
{"train_loss": 0.5432071685791016, "global_step": 18915, "epoch": 212, "lr": 9.886446294670642e-05}
{"train_loss": 0.4537213444709778, "global_step": 18916, "epoch": 212, "lr": 9.886434009163479e-05}
{"train_loss": 0.5221006870269775, "global_step": 18917, "epoch": 212, "lr": 9.886421722999396e-05}
{"train_loss": 0.4967355728149414, "global_step": 18918, "epoch": 212, "lr": 9.886409436178392e-05}
{"train_loss": 0.6619259119033813, "global_step": 18919, "epoch": 212, "lr": 9.88639714870047e-05}
{"train_loss": 0.39850130677223206, "global_step": 18920, "epoch": 212, "lr": 9.88638486056563e-05}
{"train_loss": 0.37946486473083496, "global_step": 18921, "epoch": 212, "lr": 9.886372571773877e-05}
{"train_loss": 0.3569146692752838, "global_step": 18922, "epoch": 212, "lr": 9.88636028232521e-05}
{"train_loss": 0.4360353648662567, "global_step": 18923, "epoch": 212, "lr": 9.886347992219631e-05}
{"train_loss": 0.5593526363372803, "global_step": 18924, "epoch": 212, "lr": 9.886335701457144e-05}
{"train_loss": 0.5621750950813293, "global_step": 18925, "epoch": 212, "lr": 9.886323410037746e-05}
{"train_loss": 0.36705276370048523, "global_step": 18926, "epoch": 212, "lr": 9.886311117961443e-05}
{"train_loss": 0.47891420125961304, "global_step": 18927, "epoch": 212, "lr": 9.886298825228235e-05}
{"train_loss": 0.6564435958862305, "global_step": 18928, "epoch": 212, "lr": 9.886286531838124e-05}
{"train_loss": 0.4433302879333496, "global_step": 18929, "epoch": 212, "lr": 9.88627423779111e-05}
{"train_loss": 0.6617329716682434, "global_step": 18930, "epoch": 212, "lr": 9.886261943087197e-05}
{"train_loss": 0.5888204574584961, "global_step": 18931, "epoch": 212, "lr": 9.886249647726386e-05}
{"train_loss": 0.48754212260246277, "global_step": 18932, "epoch": 212, "lr": 9.886237351708677e-05}
{"train_loss": 0.49347177147865295, "global_step": 18933, "epoch": 212, "lr": 9.886225055034073e-05}
{"train_loss": 0.6087949275970459, "global_step": 18934, "epoch": 212, "lr": 9.886212757702577e-05}
{"train_loss": 0.382327139377594, "global_step": 18935, "epoch": 212, "lr": 9.886200459714187e-05}
{"train_loss": 0.4627366364002228, "global_step": 18936, "epoch": 212, "lr": 9.88618816106891e-05}
{"train_loss": 0.38349130749702454, "global_step": 18937, "epoch": 212, "lr": 9.886175861766742e-05}
{"train_loss": 0.6439483165740967, "global_step": 18938, "epoch": 212, "lr": 9.886163561807689e-05}
{"train_loss": 0.6148673295974731, "global_step": 18939, "epoch": 212, "lr": 9.88615126119175e-05}
{"train_loss": 0.5427671670913696, "global_step": 18940, "epoch": 212, "lr": 9.886138959918927e-05}
{"train_loss": 0.49862250685691833, "global_step": 18941, "epoch": 212, "lr": 9.886126657989224e-05}
{"train_loss": 0.44314953684806824, "global_step": 18942, "epoch": 212, "lr": 9.886114355402639e-05}
{"train_loss": 0.4673765003681183, "global_step": 18943, "epoch": 212, "lr": 9.886102052159175e-05}
{"train_loss": 0.5609129071235657, "global_step": 18944, "epoch": 212, "lr": 9.886089748258835e-05}
{"train_loss": 0.451753169298172, "global_step": 18945, "epoch": 212, "lr": 9.88607744370162e-05}
{"train_loss": 0.5590030550956726, "global_step": 18946, "epoch": 212, "lr": 9.886065138487532e-05}
{"train_loss": 0.45129725337028503, "global_step": 18947, "epoch": 212, "lr": 9.88605283261657e-05}
{"train_loss": 0.433583527803421, "global_step": 18948, "epoch": 212, "lr": 9.88604052608874e-05}
{"train_loss": 0.33748194575309753, "global_step": 18949, "epoch": 212, "lr": 9.88602821890404e-05}
{"train_loss": 0.6439139246940613, "global_step": 18950, "epoch": 212, "lr": 9.886015911062474e-05}
{"train_loss": 0.505244791507721, "global_step": 18951, "epoch": 212, "lr": 9.886003602564042e-05}
{"train_loss": 0.4815036356449127, "global_step": 18952, "epoch": 212, "lr": 9.885991293408747e-05}
{"train_loss": 0.46645689010620117, "global_step": 18953, "epoch": 212, "lr": 9.88597898359659e-05}
{"train_loss": 0.4820328652858734, "global_step": 18954, "epoch": 212, "lr": 9.885966673127572e-05}
{"train_loss": 0.5047227740287781, "global_step": 18955, "epoch": 212, "lr": 9.885954362001696e-05}
{"train_loss": 0.4958145896370491, "global_step": 18956, "epoch": 212, "lr": 9.885942050218962e-05, "val_loss": 1.6347389221191406}
{"train_loss": 0.5220640301704407, "global_step": 18957, "epoch": 213, "lr": 9.885929737779373e-05}
{"train_loss": 0.4890904128551483, "global_step": 18958, "epoch": 213, "lr": 9.885917424682931e-05}
{"train_loss": 0.4368394911289215, "global_step": 18959, "epoch": 213, "lr": 9.885905110929637e-05}
{"train_loss": 0.5021862983703613, "global_step": 18960, "epoch": 213, "lr": 9.885892796519493e-05}
{"train_loss": 0.4651363492012024, "global_step": 18961, "epoch": 213, "lr": 9.885880481452498e-05}
{"train_loss": 0.5292035937309265, "global_step": 18962, "epoch": 213, "lr": 9.885868165728657e-05}
{"train_loss": 0.49672648310661316, "global_step": 18963, "epoch": 213, "lr": 9.885855849347973e-05}
{"train_loss": 0.5081895589828491, "global_step": 18964, "epoch": 213, "lr": 9.885843532310442e-05}
{"train_loss": 0.5459495782852173, "global_step": 18965, "epoch": 213, "lr": 9.88583121461607e-05}
{"train_loss": 0.5472926497459412, "global_step": 18966, "epoch": 213, "lr": 9.885818896264858e-05}
{"train_loss": 0.46912217140197754, "global_step": 18967, "epoch": 213, "lr": 9.885806577256806e-05}
{"train_loss": 0.47027549147605896, "global_step": 18968, "epoch": 213, "lr": 9.885794257591918e-05}
{"train_loss": 0.5228516459465027, "global_step": 18969, "epoch": 213, "lr": 9.885781937270195e-05}
{"train_loss": 0.4856416583061218, "global_step": 18970, "epoch": 213, "lr": 9.885769616291638e-05}
{"train_loss": 0.4337645173072815, "global_step": 18971, "epoch": 213, "lr": 9.885757294656249e-05}
{"train_loss": 0.4896025061607361, "global_step": 18972, "epoch": 213, "lr": 9.885744972364028e-05}
{"train_loss": 0.5124996304512024, "global_step": 18973, "epoch": 213, "lr": 9.885732649414978e-05}
{"train_loss": 0.49161797761917114, "global_step": 18974, "epoch": 213, "lr": 9.885720325809103e-05}
{"train_loss": 0.4660128951072693, "global_step": 18975, "epoch": 213, "lr": 9.8857080015464e-05}
{"train_loss": 0.42911630868911743, "global_step": 18976, "epoch": 213, "lr": 9.885695676626874e-05}
{"train_loss": 0.5343496203422546, "global_step": 18977, "epoch": 213, "lr": 9.885683351050525e-05}
{"train_loss": 0.4417390525341034, "global_step": 18978, "epoch": 213, "lr": 9.885671024817358e-05}
{"train_loss": 0.39050671458244324, "global_step": 18979, "epoch": 213, "lr": 9.885658697927369e-05}
{"train_loss": 0.4495448172092438, "global_step": 18980, "epoch": 213, "lr": 9.885646370380566e-05}
{"train_loss": 0.49906450510025024, "global_step": 18981, "epoch": 213, "lr": 9.885634042176944e-05}
{"train_loss": 0.551547110080719, "global_step": 18982, "epoch": 213, "lr": 9.88562171331651e-05}
{"train_loss": 0.48192623257637024, "global_step": 18983, "epoch": 213, "lr": 9.885609383799264e-05}
{"train_loss": 0.3961191177368164, "global_step": 18984, "epoch": 213, "lr": 9.885597053625206e-05}
{"train_loss": 0.44715505838394165, "global_step": 18985, "epoch": 213, "lr": 9.88558472279434e-05}
{"train_loss": 0.4504924416542053, "global_step": 18986, "epoch": 213, "lr": 9.885572391306666e-05}
{"train_loss": 0.48299601674079895, "global_step": 18987, "epoch": 213, "lr": 9.885560059162186e-05}
{"train_loss": 0.5084882974624634, "global_step": 18988, "epoch": 213, "lr": 9.885547726360905e-05}
{"train_loss": 0.5288134813308716, "global_step": 18989, "epoch": 213, "lr": 9.885535392902818e-05}
{"train_loss": 0.5746011734008789, "global_step": 18990, "epoch": 213, "lr": 9.88552305878793e-05}
{"train_loss": 0.4185977280139923, "global_step": 18991, "epoch": 213, "lr": 9.885510724016247e-05}
{"train_loss": 0.5666118264198303, "global_step": 18992, "epoch": 213, "lr": 9.885498388587764e-05}
{"train_loss": 0.519218921661377, "global_step": 18993, "epoch": 213, "lr": 9.885486052502487e-05}
{"train_loss": 0.530164361000061, "global_step": 18994, "epoch": 213, "lr": 9.885473715760412e-05}
{"train_loss": 0.5128273963928223, "global_step": 18995, "epoch": 213, "lr": 9.885461378361548e-05}
{"train_loss": 0.4877470135688782, "global_step": 18996, "epoch": 213, "lr": 9.885449040305891e-05}
{"train_loss": 0.5155458450317383, "global_step": 18997, "epoch": 213, "lr": 9.885436701593448e-05}
{"train_loss": 0.4552208483219147, "global_step": 18998, "epoch": 213, "lr": 9.885424362224216e-05}
{"train_loss": 0.5335263013839722, "global_step": 18999, "epoch": 213, "lr": 9.885412022198197e-05}
{"train_loss": 0.3943784832954407, "global_step": 19000, "epoch": 213, "lr": 9.885399681515396e-05}
{"train_loss": 0.4092514216899872, "global_step": 19001, "epoch": 213, "lr": 9.885387340175811e-05}
{"train_loss": 0.5299307107925415, "global_step": 19002, "epoch": 213, "lr": 9.885374998179447e-05}
{"train_loss": 0.5389443635940552, "global_step": 19003, "epoch": 213, "lr": 9.885362655526302e-05}
{"train_loss": 0.4889027774333954, "global_step": 19004, "epoch": 213, "lr": 9.885350312216379e-05}
{"train_loss": 0.5021654367446899, "global_step": 19005, "epoch": 213, "lr": 9.885337968249682e-05}
{"train_loss": 0.40611326694488525, "global_step": 19006, "epoch": 213, "lr": 9.885325623626212e-05}
{"train_loss": 0.40348654985427856, "global_step": 19007, "epoch": 213, "lr": 9.885313278345967e-05}
{"train_loss": 0.4709654450416565, "global_step": 19008, "epoch": 213, "lr": 9.885300932408951e-05}
{"train_loss": 0.5395087599754333, "global_step": 19009, "epoch": 213, "lr": 9.885288585815167e-05}
{"train_loss": 0.41403743624687195, "global_step": 19010, "epoch": 213, "lr": 9.885276238564616e-05}
{"train_loss": 0.4980739951133728, "global_step": 19011, "epoch": 213, "lr": 9.885263890657299e-05}
{"train_loss": 0.49537384510040283, "global_step": 19012, "epoch": 213, "lr": 9.885251542093217e-05}
{"train_loss": 0.6083083152770996, "global_step": 19013, "epoch": 213, "lr": 9.885239192872372e-05}
{"train_loss": 0.46042194962501526, "global_step": 19014, "epoch": 213, "lr": 9.885226842994767e-05}
{"train_loss": 0.5358710289001465, "global_step": 19015, "epoch": 213, "lr": 9.885214492460402e-05}
{"train_loss": 0.5765487551689148, "global_step": 19016, "epoch": 213, "lr": 9.885202141269281e-05}
{"train_loss": 0.5424878597259521, "global_step": 19017, "epoch": 213, "lr": 9.885189789421402e-05}
{"train_loss": 0.5474737882614136, "global_step": 19018, "epoch": 213, "lr": 9.88517743691677e-05}
{"train_loss": 0.5105627179145813, "global_step": 19019, "epoch": 213, "lr": 9.885165083755386e-05}
{"train_loss": 0.4167533814907074, "global_step": 19020, "epoch": 213, "lr": 9.885152729937251e-05}
{"train_loss": 0.6038507223129272, "global_step": 19021, "epoch": 213, "lr": 9.885140375462367e-05}
{"train_loss": 0.5853436589241028, "global_step": 19022, "epoch": 213, "lr": 9.885128020330734e-05}
{"train_loss": 0.4642918109893799, "global_step": 19023, "epoch": 213, "lr": 9.885115664542355e-05}
{"train_loss": 0.5340668559074402, "global_step": 19024, "epoch": 213, "lr": 9.885103308097233e-05}
{"train_loss": 0.3809466063976288, "global_step": 19025, "epoch": 213, "lr": 9.885090950995368e-05}
{"train_loss": 0.4275193214416504, "global_step": 19026, "epoch": 213, "lr": 9.885078593236761e-05}
{"train_loss": 0.6165040731430054, "global_step": 19027, "epoch": 213, "lr": 9.885066234821416e-05}
{"train_loss": 0.6122249960899353, "global_step": 19028, "epoch": 213, "lr": 9.885053875749333e-05}
{"train_loss": 0.4766693413257599, "global_step": 19029, "epoch": 213, "lr": 9.885041516020514e-05}
{"train_loss": 0.4498472809791565, "global_step": 19030, "epoch": 213, "lr": 9.885029155634961e-05}
{"train_loss": 0.6157548427581787, "global_step": 19031, "epoch": 213, "lr": 9.885016794592677e-05}
{"train_loss": 0.43253636360168457, "global_step": 19032, "epoch": 213, "lr": 9.885004432893659e-05}
{"train_loss": 0.5743905305862427, "global_step": 19033, "epoch": 213, "lr": 9.884992070537913e-05}
{"train_loss": 0.44389885663986206, "global_step": 19034, "epoch": 213, "lr": 9.88497970752544e-05}
{"train_loss": 0.42452892661094666, "global_step": 19035, "epoch": 213, "lr": 9.884967343856239e-05}
{"train_loss": 0.4264131784439087, "global_step": 19036, "epoch": 213, "lr": 9.884954979530316e-05}
{"train_loss": 0.4475521445274353, "global_step": 19037, "epoch": 213, "lr": 9.88494261454767e-05}
{"train_loss": 0.42735251784324646, "global_step": 19038, "epoch": 213, "lr": 9.884930248908302e-05}
{"train_loss": 0.6503748297691345, "global_step": 19039, "epoch": 213, "lr": 9.884917882612215e-05}
{"train_loss": 0.5000866055488586, "global_step": 19040, "epoch": 213, "lr": 9.884905515659411e-05}
{"train_loss": 0.5233974456787109, "global_step": 19041, "epoch": 213, "lr": 9.88489314804989e-05}
{"train_loss": 0.5374150276184082, "global_step": 19042, "epoch": 213, "lr": 9.884880779783654e-05}
{"train_loss": 0.4382367432117462, "global_step": 19043, "epoch": 213, "lr": 9.884868410860706e-05}
{"train_loss": 0.5608121156692505, "global_step": 19044, "epoch": 213, "lr": 9.884856041281048e-05}
{"train_loss": 0.4961300736732697, "global_step": 19045, "epoch": 213, "lr": 9.884843671044679e-05, "val_loss": 1.7136306762695312}
{"train_loss": 0.5324828624725342, "global_step": 19046, "epoch": 214, "lr": 9.884831300151605e-05}
{"train_loss": 0.5530866980552673, "global_step": 19047, "epoch": 214, "lr": 9.884818928601823e-05}
{"train_loss": 0.5854306817054749, "global_step": 19048, "epoch": 214, "lr": 9.884806556395337e-05}
{"train_loss": 0.5228439569473267, "global_step": 19049, "epoch": 214, "lr": 9.884794183532147e-05}
{"train_loss": 0.5973352193832397, "global_step": 19050, "epoch": 214, "lr": 9.884781810012258e-05}
{"train_loss": 0.4595317244529724, "global_step": 19051, "epoch": 214, "lr": 9.884769435835668e-05}
{"train_loss": 0.4937977194786072, "global_step": 19052, "epoch": 214, "lr": 9.884757061002381e-05}
{"train_loss": 0.5040768384933472, "global_step": 19053, "epoch": 214, "lr": 9.884744685512399e-05}
{"train_loss": 0.5163238644599915, "global_step": 19054, "epoch": 214, "lr": 9.884732309365721e-05}
{"train_loss": 0.4902755320072174, "global_step": 19055, "epoch": 214, "lr": 9.88471993256235e-05}
{"train_loss": 0.44993332028388977, "global_step": 19056, "epoch": 214, "lr": 9.88470755510229e-05}
{"train_loss": 0.5357946753501892, "global_step": 19057, "epoch": 214, "lr": 9.884695176985539e-05}
{"train_loss": 0.4067956805229187, "global_step": 19058, "epoch": 214, "lr": 9.8846827982121e-05}
{"train_loss": 0.41251617670059204, "global_step": 19059, "epoch": 214, "lr": 9.884670418781976e-05}
{"train_loss": 0.5254103541374207, "global_step": 19060, "epoch": 214, "lr": 9.884658038695167e-05}
{"train_loss": 0.4742009937763214, "global_step": 19061, "epoch": 214, "lr": 9.884645657951676e-05}
{"train_loss": 0.5539687871932983, "global_step": 19062, "epoch": 214, "lr": 9.884633276551502e-05}
{"train_loss": 0.4989359676837921, "global_step": 19063, "epoch": 214, "lr": 9.884620894494652e-05}
{"train_loss": 0.5530381798744202, "global_step": 19064, "epoch": 214, "lr": 9.884608511781122e-05}
{"train_loss": 0.3749881982803345, "global_step": 19065, "epoch": 214, "lr": 9.884596128410915e-05}
{"train_loss": 0.48466330766677856, "global_step": 19066, "epoch": 214, "lr": 9.884583744384035e-05}
{"train_loss": 0.4453239440917969, "global_step": 19067, "epoch": 214, "lr": 9.884571359700482e-05}
{"train_loss": 0.6704848408699036, "global_step": 19068, "epoch": 214, "lr": 9.884558974360258e-05}
{"train_loss": 0.4349853992462158, "global_step": 19069, "epoch": 214, "lr": 9.884546588363364e-05}
{"train_loss": 0.5843092799186707, "global_step": 19070, "epoch": 214, "lr": 9.884534201709803e-05}
{"train_loss": 0.525236189365387, "global_step": 19071, "epoch": 214, "lr": 9.884521814399575e-05}
{"train_loss": 0.6239610910415649, "global_step": 19072, "epoch": 214, "lr": 9.884509426432685e-05}
{"train_loss": 0.5245577692985535, "global_step": 19073, "epoch": 214, "lr": 9.88449703780913e-05}
{"train_loss": 0.49455705285072327, "global_step": 19074, "epoch": 214, "lr": 9.884484648528913e-05}
{"train_loss": 0.45847296714782715, "global_step": 19075, "epoch": 214, "lr": 9.88447225859204e-05}
{"train_loss": 0.5899257063865662, "global_step": 19076, "epoch": 214, "lr": 9.884459867998505e-05}
{"train_loss": 0.6882256269454956, "global_step": 19077, "epoch": 214, "lr": 9.884447476748317e-05}
{"train_loss": 0.45925647020339966, "global_step": 19078, "epoch": 214, "lr": 9.884435084841474e-05}
{"train_loss": 0.5268174409866333, "global_step": 19079, "epoch": 214, "lr": 9.884422692277977e-05}
{"train_loss": 0.6020194888114929, "global_step": 19080, "epoch": 214, "lr": 9.88441029905783e-05}
{"train_loss": 0.5247066020965576, "global_step": 19081, "epoch": 214, "lr": 9.884397905181033e-05}
{"train_loss": 0.36934301257133484, "global_step": 19082, "epoch": 214, "lr": 9.884385510647589e-05}
{"train_loss": 0.5099815130233765, "global_step": 19083, "epoch": 214, "lr": 9.8843731154575e-05}
{"train_loss": 0.6006413102149963, "global_step": 19084, "epoch": 214, "lr": 9.884360719610763e-05}
{"train_loss": 0.5500958561897278, "global_step": 19085, "epoch": 214, "lr": 9.884348323107386e-05}
{"train_loss": 0.5449200868606567, "global_step": 19086, "epoch": 214, "lr": 9.884335925947368e-05}
{"train_loss": 0.3858078122138977, "global_step": 19087, "epoch": 214, "lr": 9.88432352813071e-05}
{"train_loss": 0.524563193321228, "global_step": 19088, "epoch": 214, "lr": 9.884311129657413e-05}
{"train_loss": 0.6212143898010254, "global_step": 19089, "epoch": 214, "lr": 9.884298730527482e-05}
{"train_loss": 0.5633643865585327, "global_step": 19090, "epoch": 214, "lr": 9.884286330740915e-05}
{"train_loss": 0.5201217532157898, "global_step": 19091, "epoch": 214, "lr": 9.884273930297715e-05}
{"train_loss": 0.47522756457328796, "global_step": 19092, "epoch": 214, "lr": 9.884261529197885e-05}
{"train_loss": 0.5083433985710144, "global_step": 19093, "epoch": 214, "lr": 9.884249127441426e-05}
{"train_loss": 0.47955888509750366, "global_step": 19094, "epoch": 214, "lr": 9.884236725028338e-05}
{"train_loss": 0.497256875038147, "global_step": 19095, "epoch": 214, "lr": 9.884224321958624e-05}
{"train_loss": 0.5364016890525818, "global_step": 19096, "epoch": 214, "lr": 9.884211918232285e-05}
{"train_loss": 0.47697657346725464, "global_step": 19097, "epoch": 214, "lr": 9.884199513849324e-05}
{"train_loss": 0.5360081791877747, "global_step": 19098, "epoch": 214, "lr": 9.884187108809743e-05}
{"train_loss": 0.5125644207000732, "global_step": 19099, "epoch": 214, "lr": 9.884174703113541e-05}
{"train_loss": 0.45735517144203186, "global_step": 19100, "epoch": 214, "lr": 9.884162296760721e-05}
{"train_loss": 0.44006282091140747, "global_step": 19101, "epoch": 214, "lr": 9.884149889751286e-05}
{"train_loss": 0.549378514289856, "global_step": 19102, "epoch": 214, "lr": 9.884137482085236e-05}
{"train_loss": 0.4608491361141205, "global_step": 19103, "epoch": 214, "lr": 9.884125073762574e-05}
{"train_loss": 0.5585877895355225, "global_step": 19104, "epoch": 214, "lr": 9.8841126647833e-05}
{"train_loss": 0.6527582406997681, "global_step": 19105, "epoch": 214, "lr": 9.884100255147417e-05}
{"train_loss": 0.4646737575531006, "global_step": 19106, "epoch": 214, "lr": 9.884087844854928e-05}
{"train_loss": 0.5898897051811218, "global_step": 19107, "epoch": 214, "lr": 9.884075433905831e-05}
{"train_loss": 0.5051794648170471, "global_step": 19108, "epoch": 214, "lr": 9.884063022300129e-05}
{"train_loss": 0.41764384508132935, "global_step": 19109, "epoch": 214, "lr": 9.884050610037825e-05}
{"train_loss": 0.5056436657905579, "global_step": 19110, "epoch": 214, "lr": 9.88403819711892e-05}
{"train_loss": 0.5897306203842163, "global_step": 19111, "epoch": 214, "lr": 9.884025783543416e-05}
{"train_loss": 0.5446829795837402, "global_step": 19112, "epoch": 214, "lr": 9.884013369311313e-05}
{"train_loss": 0.5858438014984131, "global_step": 19113, "epoch": 214, "lr": 9.884000954422615e-05}
{"train_loss": 0.43520665168762207, "global_step": 19114, "epoch": 214, "lr": 9.883988538877322e-05}
{"train_loss": 0.5547416806221008, "global_step": 19115, "epoch": 214, "lr": 9.883976122675438e-05}
{"train_loss": 0.5189797878265381, "global_step": 19116, "epoch": 214, "lr": 9.883963705816961e-05}
{"train_loss": 0.34147781133651733, "global_step": 19117, "epoch": 214, "lr": 9.883951288301895e-05}
{"train_loss": 0.45899122953414917, "global_step": 19118, "epoch": 214, "lr": 9.883938870130242e-05}
{"train_loss": 0.4825773537158966, "global_step": 19119, "epoch": 214, "lr": 9.883926451302001e-05}
{"train_loss": 0.5469180345535278, "global_step": 19120, "epoch": 214, "lr": 9.883914031817179e-05}
{"train_loss": 0.508692741394043, "global_step": 19121, "epoch": 214, "lr": 9.883901611675772e-05}
{"train_loss": 0.4058535099029541, "global_step": 19122, "epoch": 214, "lr": 9.883889190877784e-05}
{"train_loss": 0.41478925943374634, "global_step": 19123, "epoch": 214, "lr": 9.883876769423218e-05}
{"train_loss": 0.5198098421096802, "global_step": 19124, "epoch": 214, "lr": 9.883864347312071e-05}
{"train_loss": 0.44754934310913086, "global_step": 19125, "epoch": 214, "lr": 9.883851924544351e-05}
{"train_loss": 0.5777133703231812, "global_step": 19126, "epoch": 214, "lr": 9.883839501120055e-05}
{"train_loss": 0.5900118350982666, "global_step": 19127, "epoch": 214, "lr": 9.883827077039187e-05}
{"train_loss": 0.6641868352890015, "global_step": 19128, "epoch": 214, "lr": 9.883814652301748e-05}
{"train_loss": 0.4625847637653351, "global_step": 19129, "epoch": 214, "lr": 9.88380222690774e-05}
{"train_loss": 0.49513497948646545, "global_step": 19130, "epoch": 214, "lr": 9.883789800857164e-05}
{"train_loss": 0.5853133201599121, "global_step": 19131, "epoch": 214, "lr": 9.883777374150022e-05}
{"train_loss": 0.4755170941352844, "global_step": 19132, "epoch": 214, "lr": 9.883764946786316e-05}
{"train_loss": 0.6937639713287354, "global_step": 19133, "epoch": 214, "lr": 9.883752518766048e-05}
{"train_loss": 0.5154250451688016, "global_step": 19134, "epoch": 214, "lr": 9.883740090089217e-05, "val_loss": 1.7424554824829102}
{"train_loss": 0.5533272624015808, "global_step": 19135, "epoch": 215, "lr": 9.883727660755829e-05}
{"train_loss": 0.41578707098960876, "global_step": 19136, "epoch": 215, "lr": 9.883715230765881e-05}
{"train_loss": 0.5253921151161194, "global_step": 19137, "epoch": 215, "lr": 9.883702800119377e-05}
{"train_loss": 0.44989797472953796, "global_step": 19138, "epoch": 215, "lr": 9.88369036881632e-05}
{"train_loss": 0.5048931241035461, "global_step": 19139, "epoch": 215, "lr": 9.88367793685671e-05}
{"train_loss": 0.4438732862472534, "global_step": 19140, "epoch": 215, "lr": 9.883665504240549e-05}
{"train_loss": 0.3833559453487396, "global_step": 19141, "epoch": 215, "lr": 9.883653070967839e-05}
{"train_loss": 0.42239293456077576, "global_step": 19142, "epoch": 215, "lr": 9.883640637038581e-05}
{"train_loss": 0.41841113567352295, "global_step": 19143, "epoch": 215, "lr": 9.883628202452777e-05}
{"train_loss": 0.45509642362594604, "global_step": 19144, "epoch": 215, "lr": 9.88361576721043e-05}
{"train_loss": 0.5735270380973816, "global_step": 19145, "epoch": 215, "lr": 9.883603331311538e-05}
{"train_loss": 0.5370186567306519, "global_step": 19146, "epoch": 215, "lr": 9.883590894756107e-05}
{"train_loss": 0.5620964765548706, "global_step": 19147, "epoch": 215, "lr": 9.883578457544136e-05}
{"train_loss": 0.3590240478515625, "global_step": 19148, "epoch": 215, "lr": 9.883566019675627e-05}
{"train_loss": 0.5058233141899109, "global_step": 19149, "epoch": 215, "lr": 9.883553581150583e-05}
{"train_loss": 0.497358113527298, "global_step": 19150, "epoch": 215, "lr": 9.883541141969004e-05}
{"train_loss": 0.5436109304428101, "global_step": 19151, "epoch": 215, "lr": 9.883528702130893e-05}
{"train_loss": 0.5379741787910461, "global_step": 19152, "epoch": 215, "lr": 9.883516261636251e-05}
{"train_loss": 0.37073326110839844, "global_step": 19153, "epoch": 215, "lr": 9.88350382048508e-05}
{"train_loss": 0.5693626403808594, "global_step": 19154, "epoch": 215, "lr": 9.883491378677381e-05}
{"train_loss": 0.4433046281337738, "global_step": 19155, "epoch": 215, "lr": 9.883478936213155e-05}
{"train_loss": 0.5318249464035034, "global_step": 19156, "epoch": 215, "lr": 9.883466493092407e-05}
{"train_loss": 0.5153228044509888, "global_step": 19157, "epoch": 215, "lr": 9.883454049315136e-05}
{"train_loss": 0.5305691361427307, "global_step": 19158, "epoch": 215, "lr": 9.883441604881341e-05}
{"train_loss": 0.3890843093395233, "global_step": 19159, "epoch": 215, "lr": 9.88342915979103e-05}
{"train_loss": 0.37918350100517273, "global_step": 19160, "epoch": 215, "lr": 9.8834167140442e-05}
{"train_loss": 0.5565334558486938, "global_step": 19161, "epoch": 215, "lr": 9.883404267640856e-05}
{"train_loss": 0.5006760358810425, "global_step": 19162, "epoch": 215, "lr": 9.883391820580996e-05}
{"train_loss": 0.6231165528297424, "global_step": 19163, "epoch": 215, "lr": 9.883379372864624e-05}
{"train_loss": 0.5626837015151978, "global_step": 19164, "epoch": 215, "lr": 9.883366924491741e-05}
{"train_loss": 0.4999651312828064, "global_step": 19165, "epoch": 215, "lr": 9.88335447546235e-05}
{"train_loss": 0.4632725417613983, "global_step": 19166, "epoch": 215, "lr": 9.88334202577645e-05}
{"train_loss": 0.5219724774360657, "global_step": 19167, "epoch": 215, "lr": 9.883329575434045e-05}
{"train_loss": 0.4852137863636017, "global_step": 19168, "epoch": 215, "lr": 9.883317124435136e-05}
{"train_loss": 0.6086073517799377, "global_step": 19169, "epoch": 215, "lr": 9.883304672779724e-05}
{"train_loss": 0.4816056191921234, "global_step": 19170, "epoch": 215, "lr": 9.883292220467812e-05}
{"train_loss": 0.5411438941955566, "global_step": 19171, "epoch": 215, "lr": 9.8832797674994e-05}
{"train_loss": 0.4735187292098999, "global_step": 19172, "epoch": 215, "lr": 9.883267313874492e-05}
{"train_loss": 0.39240995049476624, "global_step": 19173, "epoch": 215, "lr": 9.883254859593085e-05}
{"train_loss": 0.5791078209877014, "global_step": 19174, "epoch": 215, "lr": 9.883242404655187e-05}
{"train_loss": 0.5461740493774414, "global_step": 19175, "epoch": 215, "lr": 9.883229949060795e-05}
{"train_loss": 0.3985828459262848, "global_step": 19176, "epoch": 215, "lr": 9.883217492809912e-05}
{"train_loss": 0.5019339919090271, "global_step": 19177, "epoch": 215, "lr": 9.883205035902541e-05}
{"train_loss": 0.3497200012207031, "global_step": 19178, "epoch": 215, "lr": 9.883192578338683e-05}
{"train_loss": 0.3853248953819275, "global_step": 19179, "epoch": 215, "lr": 9.883180120118338e-05}
{"train_loss": 0.4438774585723877, "global_step": 19180, "epoch": 215, "lr": 9.88316766124151e-05}
{"train_loss": 0.6110907196998596, "global_step": 19181, "epoch": 215, "lr": 9.883155201708199e-05}
{"train_loss": 0.5072787404060364, "global_step": 19182, "epoch": 215, "lr": 9.883142741518407e-05}
{"train_loss": 0.505109965801239, "global_step": 19183, "epoch": 215, "lr": 9.883130280672136e-05}
{"train_loss": 0.40827250480651855, "global_step": 19184, "epoch": 215, "lr": 9.883117819169389e-05}
{"train_loss": 0.5487241148948669, "global_step": 19185, "epoch": 215, "lr": 9.883105357010166e-05}
{"train_loss": 0.5971925854682922, "global_step": 19186, "epoch": 215, "lr": 9.883092894194468e-05}
{"train_loss": 0.41034117341041565, "global_step": 19187, "epoch": 215, "lr": 9.883080430722298e-05}
{"train_loss": 0.5034419894218445, "global_step": 19188, "epoch": 215, "lr": 9.883067966593658e-05}
{"train_loss": 0.3551122844219208, "global_step": 19189, "epoch": 215, "lr": 9.883055501808549e-05}
{"train_loss": 0.565648078918457, "global_step": 19190, "epoch": 215, "lr": 9.883043036366972e-05}
{"train_loss": 0.4607542157173157, "global_step": 19191, "epoch": 215, "lr": 9.883030570268929e-05}
{"train_loss": 0.49586135149002075, "global_step": 19192, "epoch": 215, "lr": 9.883018103514423e-05}
{"train_loss": 0.5466275811195374, "global_step": 19193, "epoch": 215, "lr": 9.883005636103455e-05}
{"train_loss": 0.5294793844223022, "global_step": 19194, "epoch": 215, "lr": 9.882993168036026e-05}
{"train_loss": 0.48652181029319763, "global_step": 19195, "epoch": 215, "lr": 9.882980699312136e-05}
{"train_loss": 0.41280511021614075, "global_step": 19196, "epoch": 215, "lr": 9.882968229931791e-05}
{"train_loss": 0.5465158224105835, "global_step": 19197, "epoch": 215, "lr": 9.882955759894991e-05}
{"train_loss": 0.5154792070388794, "global_step": 19198, "epoch": 215, "lr": 9.882943289201736e-05}
{"train_loss": 0.47827497124671936, "global_step": 19199, "epoch": 215, "lr": 9.88293081785203e-05}
{"train_loss": 0.38195309042930603, "global_step": 19200, "epoch": 215, "lr": 9.882918345845871e-05}
{"train_loss": 0.4714454412460327, "global_step": 19201, "epoch": 215, "lr": 9.882905873183264e-05}
{"train_loss": 0.5333503484725952, "global_step": 19202, "epoch": 215, "lr": 9.882893399864211e-05}
{"train_loss": 0.6227251291275024, "global_step": 19203, "epoch": 215, "lr": 9.882880925888713e-05}
{"train_loss": 0.48790305852890015, "global_step": 19204, "epoch": 215, "lr": 9.882868451256769e-05}
{"train_loss": 0.545841634273529, "global_step": 19205, "epoch": 215, "lr": 9.882855975968383e-05}
{"train_loss": 0.5390818119049072, "global_step": 19206, "epoch": 215, "lr": 9.882843500023559e-05}
{"train_loss": 0.48055386543273926, "global_step": 19207, "epoch": 215, "lr": 9.882831023422294e-05}
{"train_loss": 0.5568937659263611, "global_step": 19208, "epoch": 215, "lr": 9.882818546164593e-05}
{"train_loss": 0.4382908344268799, "global_step": 19209, "epoch": 215, "lr": 9.882806068250456e-05}
{"train_loss": 0.42925482988357544, "global_step": 19210, "epoch": 215, "lr": 9.882793589679886e-05}
{"train_loss": 0.44615694880485535, "global_step": 19211, "epoch": 215, "lr": 9.882781110452882e-05}
{"train_loss": 0.5814353227615356, "global_step": 19212, "epoch": 215, "lr": 9.88276863056945e-05}
{"train_loss": 0.5885550379753113, "global_step": 19213, "epoch": 215, "lr": 9.882756150029588e-05}
{"train_loss": 0.5329488515853882, "global_step": 19214, "epoch": 215, "lr": 9.882743668833299e-05}
{"train_loss": 0.42066141963005066, "global_step": 19215, "epoch": 215, "lr": 9.882731186980584e-05}
{"train_loss": 0.502264142036438, "global_step": 19216, "epoch": 215, "lr": 9.882718704471446e-05}
{"train_loss": 0.44215095043182373, "global_step": 19217, "epoch": 215, "lr": 9.882706221305886e-05}
{"train_loss": 0.5069763660430908, "global_step": 19218, "epoch": 215, "lr": 9.882693737483905e-05}
{"train_loss": 0.622389018535614, "global_step": 19219, "epoch": 215, "lr": 9.882681253005507e-05}
{"train_loss": 0.5678666830062866, "global_step": 19220, "epoch": 215, "lr": 9.882668767870691e-05}
{"train_loss": 0.4492802917957306, "global_step": 19221, "epoch": 215, "lr": 9.882656282079459e-05}
{"train_loss": 0.5785192847251892, "global_step": 19222, "epoch": 215, "lr": 9.882643795631814e-05}
{"train_loss": 0.49320493186457776, "global_step": 19223, "epoch": 215, "lr": 9.882631308527757e-05, "val_loss": 1.7561817169189453, "train_action_mse_error": 30.195728302001953}
{"train_loss": 0.4417753219604492, "global_step": 19224, "epoch": 216, "lr": 9.882618820767291e-05}
{"train_loss": 0.484645813703537, "global_step": 19225, "epoch": 216, "lr": 9.882606332350415e-05}
{"train_loss": 0.5367273688316345, "global_step": 19226, "epoch": 216, "lr": 9.882593843277133e-05}
{"train_loss": 0.42847946286201477, "global_step": 19227, "epoch": 216, "lr": 9.882581353547444e-05}
{"train_loss": 0.5568886995315552, "global_step": 19228, "epoch": 216, "lr": 9.882568863161353e-05}
{"train_loss": 0.41898852586746216, "global_step": 19229, "epoch": 216, "lr": 9.882556372118859e-05}
{"train_loss": 0.6185368895530701, "global_step": 19230, "epoch": 216, "lr": 9.882543880419966e-05}
{"train_loss": 0.4229998290538788, "global_step": 19231, "epoch": 216, "lr": 9.882531388064672e-05}
{"train_loss": 0.5739156007766724, "global_step": 19232, "epoch": 216, "lr": 9.882518895052984e-05}
{"train_loss": 0.5248212814331055, "global_step": 19233, "epoch": 216, "lr": 9.882506401384901e-05}
{"train_loss": 0.49299877882003784, "global_step": 19234, "epoch": 216, "lr": 9.882493907060424e-05}
{"train_loss": 0.432815283536911, "global_step": 19235, "epoch": 216, "lr": 9.882481412079553e-05}
{"train_loss": 0.46580377221107483, "global_step": 19236, "epoch": 216, "lr": 9.882468916442294e-05}
{"train_loss": 0.6042544841766357, "global_step": 19237, "epoch": 216, "lr": 9.882456420148648e-05}
{"train_loss": 0.40993908047676086, "global_step": 19238, "epoch": 216, "lr": 9.882443923198612e-05}
{"train_loss": 0.3895935118198395, "global_step": 19239, "epoch": 216, "lr": 9.882431425592192e-05}
{"train_loss": 0.4670393168926239, "global_step": 19240, "epoch": 216, "lr": 9.882418927329388e-05}
{"train_loss": 0.5463780164718628, "global_step": 19241, "epoch": 216, "lr": 9.882406428410205e-05}
{"train_loss": 0.49476152658462524, "global_step": 19242, "epoch": 216, "lr": 9.882393928834639e-05}
{"train_loss": 0.5055187940597534, "global_step": 19243, "epoch": 216, "lr": 9.882381428602697e-05}
{"train_loss": 0.39956510066986084, "global_step": 19244, "epoch": 216, "lr": 9.882368927714376e-05}
{"train_loss": 0.48648497462272644, "global_step": 19245, "epoch": 216, "lr": 9.882356426169681e-05}
{"train_loss": 0.39124491810798645, "global_step": 19246, "epoch": 216, "lr": 9.882343923968614e-05}
{"train_loss": 0.5078787803649902, "global_step": 19247, "epoch": 216, "lr": 9.882331421111175e-05}
{"train_loss": 0.5713083744049072, "global_step": 19248, "epoch": 216, "lr": 9.882318917597365e-05}
{"train_loss": 0.5673257112503052, "global_step": 19249, "epoch": 216, "lr": 9.882306413427185e-05}
{"train_loss": 0.3815020024776459, "global_step": 19250, "epoch": 216, "lr": 9.882293908600641e-05}
{"train_loss": 0.3144318163394928, "global_step": 19251, "epoch": 216, "lr": 9.882281403117733e-05}
{"train_loss": 0.602063775062561, "global_step": 19252, "epoch": 216, "lr": 9.88226889697846e-05}
{"train_loss": 0.41501593589782715, "global_step": 19253, "epoch": 216, "lr": 9.882256390182825e-05}
{"train_loss": 0.4181891977787018, "global_step": 19254, "epoch": 216, "lr": 9.882243882730831e-05}
{"train_loss": 0.6188582181930542, "global_step": 19255, "epoch": 216, "lr": 9.882231374622478e-05}
{"train_loss": 0.4506780207157135, "global_step": 19256, "epoch": 216, "lr": 9.88221886585777e-05}
{"train_loss": 0.5053314566612244, "global_step": 19257, "epoch": 216, "lr": 9.882206356436705e-05}
{"train_loss": 0.48193562030792236, "global_step": 19258, "epoch": 216, "lr": 9.882193846359288e-05}
{"train_loss": 0.40612128376960754, "global_step": 19259, "epoch": 216, "lr": 9.88218133562552e-05}
{"train_loss": 0.546161413192749, "global_step": 19260, "epoch": 216, "lr": 9.882168824235403e-05}
{"train_loss": 0.45039477944374084, "global_step": 19261, "epoch": 216, "lr": 9.882156312188938e-05}
{"train_loss": 0.5148381590843201, "global_step": 19262, "epoch": 216, "lr": 9.882143799486123e-05}
{"train_loss": 0.5499929785728455, "global_step": 19263, "epoch": 216, "lr": 9.882131286126967e-05}
{"train_loss": 0.5227515697479248, "global_step": 19264, "epoch": 216, "lr": 9.882118772111466e-05}
{"train_loss": 0.4512152373790741, "global_step": 19265, "epoch": 216, "lr": 9.882106257439625e-05}
{"train_loss": 0.4191102385520935, "global_step": 19266, "epoch": 216, "lr": 9.882093742111444e-05}
{"train_loss": 0.43295082449913025, "global_step": 19267, "epoch": 216, "lr": 9.882081226126925e-05}
{"train_loss": 0.6352576017379761, "global_step": 19268, "epoch": 216, "lr": 9.882068709486068e-05}
{"train_loss": 0.38089463114738464, "global_step": 19269, "epoch": 216, "lr": 9.882056192188877e-05}
{"train_loss": 0.48633936047554016, "global_step": 19270, "epoch": 216, "lr": 9.882043674235355e-05}
{"train_loss": 0.4814654588699341, "global_step": 19271, "epoch": 216, "lr": 9.882031155625499e-05}
{"train_loss": 0.49862563610076904, "global_step": 19272, "epoch": 216, "lr": 9.882018636359316e-05}
{"train_loss": 0.46927374601364136, "global_step": 19273, "epoch": 216, "lr": 9.882006116436804e-05}
{"train_loss": 0.49393418431282043, "global_step": 19274, "epoch": 216, "lr": 9.881993595857966e-05}
{"train_loss": 0.3693801462650299, "global_step": 19275, "epoch": 216, "lr": 9.881981074622804e-05}
{"train_loss": 0.43545758724212646, "global_step": 19276, "epoch": 216, "lr": 9.881968552731318e-05}
{"train_loss": 0.4452567398548126, "global_step": 19277, "epoch": 216, "lr": 9.88195603018351e-05}
{"train_loss": 0.48240506649017334, "global_step": 19278, "epoch": 216, "lr": 9.881943506979385e-05}
{"train_loss": 0.5313507318496704, "global_step": 19279, "epoch": 216, "lr": 9.88193098311894e-05}
{"train_loss": 0.484164834022522, "global_step": 19280, "epoch": 216, "lr": 9.881918458602179e-05}
{"train_loss": 0.4631076455116272, "global_step": 19281, "epoch": 216, "lr": 9.881905933429105e-05}
{"train_loss": 0.4491369128227234, "global_step": 19282, "epoch": 216, "lr": 9.881893407599719e-05}
{"train_loss": 0.5568636059761047, "global_step": 19283, "epoch": 216, "lr": 9.881880881114019e-05}
{"train_loss": 0.4399920105934143, "global_step": 19284, "epoch": 216, "lr": 9.881868353972012e-05}
{"train_loss": 0.45663028955459595, "global_step": 19285, "epoch": 216, "lr": 9.881855826173696e-05}
{"train_loss": 0.3963383138179779, "global_step": 19286, "epoch": 216, "lr": 9.881843297719075e-05}
{"train_loss": 0.37968629598617554, "global_step": 19287, "epoch": 216, "lr": 9.881830768608149e-05}
{"train_loss": 0.4905564486980438, "global_step": 19288, "epoch": 216, "lr": 9.881818238840921e-05}
{"train_loss": 0.4094059467315674, "global_step": 19289, "epoch": 216, "lr": 9.881805708417391e-05}
{"train_loss": 0.643227219581604, "global_step": 19290, "epoch": 216, "lr": 9.881793177337562e-05}
{"train_loss": 0.43759962916374207, "global_step": 19291, "epoch": 216, "lr": 9.881780645601435e-05}
{"train_loss": 0.5700735449790955, "global_step": 19292, "epoch": 216, "lr": 9.881768113209013e-05}
{"train_loss": 0.5886347889900208, "global_step": 19293, "epoch": 216, "lr": 9.881755580160297e-05}
{"train_loss": 0.5000618696212769, "global_step": 19294, "epoch": 216, "lr": 9.881743046455288e-05}
{"train_loss": 0.6079249382019043, "global_step": 19295, "epoch": 216, "lr": 9.881730512093988e-05}
{"train_loss": 0.43300408124923706, "global_step": 19296, "epoch": 216, "lr": 9.881717977076399e-05}
{"train_loss": 0.4218548834323883, "global_step": 19297, "epoch": 216, "lr": 9.881705441402522e-05}
{"train_loss": 0.4994605481624603, "global_step": 19298, "epoch": 216, "lr": 9.88169290507236e-05}
{"train_loss": 0.44487106800079346, "global_step": 19299, "epoch": 216, "lr": 9.881680368085912e-05}
{"train_loss": 0.5022110342979431, "global_step": 19300, "epoch": 216, "lr": 9.881667830443184e-05}
{"train_loss": 0.4743660092353821, "global_step": 19301, "epoch": 216, "lr": 9.881655292144176e-05}
{"train_loss": 0.5806043148040771, "global_step": 19302, "epoch": 216, "lr": 9.881642753188886e-05}
{"train_loss": 0.49905654788017273, "global_step": 19303, "epoch": 216, "lr": 9.88163021357732e-05}
{"train_loss": 0.46904969215393066, "global_step": 19304, "epoch": 216, "lr": 9.881617673309478e-05}
{"train_loss": 0.47886332869529724, "global_step": 19305, "epoch": 216, "lr": 9.881605132385363e-05}
{"train_loss": 0.35747867822647095, "global_step": 19306, "epoch": 216, "lr": 9.881592590804976e-05}
{"train_loss": 0.49861082434654236, "global_step": 19307, "epoch": 216, "lr": 9.881580048568316e-05}
{"train_loss": 0.4220154285430908, "global_step": 19308, "epoch": 216, "lr": 9.881567505675389e-05}
{"train_loss": 0.4391041696071625, "global_step": 19309, "epoch": 216, "lr": 9.881554962126194e-05}
{"train_loss": 0.5268460512161255, "global_step": 19310, "epoch": 216, "lr": 9.881542417920732e-05}
{"train_loss": 0.42580971121788025, "global_step": 19311, "epoch": 216, "lr": 9.881529873059009e-05}
{"train_loss": 0.48049467935990753, "global_step": 19312, "epoch": 216, "lr": 9.881517327541024e-05, "val_loss": 1.7539442777633667}
{"train_loss": 0.5286246538162231, "global_step": 19313, "epoch": 217, "lr": 9.881504781366775e-05}
{"train_loss": 0.5232729315757751, "global_step": 19314, "epoch": 217, "lr": 9.881492234536269e-05}
{"train_loss": 0.4523521661758423, "global_step": 19315, "epoch": 217, "lr": 9.881479687049507e-05}
{"train_loss": 0.3710636496543884, "global_step": 19316, "epoch": 217, "lr": 9.881467138906489e-05}
{"train_loss": 0.44963330030441284, "global_step": 19317, "epoch": 217, "lr": 9.881454590107217e-05}
{"train_loss": 0.5900759100914001, "global_step": 19318, "epoch": 217, "lr": 9.881442040651691e-05}
{"train_loss": 0.4632314443588257, "global_step": 19319, "epoch": 217, "lr": 9.881429490539918e-05}
{"train_loss": 0.5460813045501709, "global_step": 19320, "epoch": 217, "lr": 9.881416939771895e-05}
{"train_loss": 0.5263229608535767, "global_step": 19321, "epoch": 217, "lr": 9.881404388347623e-05}
{"train_loss": 0.47886794805526733, "global_step": 19322, "epoch": 217, "lr": 9.881391836267108e-05}
{"train_loss": 0.43550586700439453, "global_step": 19323, "epoch": 217, "lr": 9.881379283530349e-05}
{"train_loss": 0.5189175605773926, "global_step": 19324, "epoch": 217, "lr": 9.881366730137346e-05}
{"train_loss": 0.43571969866752625, "global_step": 19325, "epoch": 217, "lr": 9.881354176088107e-05}
{"train_loss": 0.4532725214958191, "global_step": 19326, "epoch": 217, "lr": 9.881341621382627e-05}
{"train_loss": 0.545263946056366, "global_step": 19327, "epoch": 217, "lr": 9.88132906602091e-05}
{"train_loss": 0.49853429198265076, "global_step": 19328, "epoch": 217, "lr": 9.881316510002957e-05}
{"train_loss": 0.457835853099823, "global_step": 19329, "epoch": 217, "lr": 9.881303953328771e-05}
{"train_loss": 0.4948235750198364, "global_step": 19330, "epoch": 217, "lr": 9.881291395998353e-05}
{"train_loss": 0.46341267228126526, "global_step": 19331, "epoch": 217, "lr": 9.881278838011706e-05}
{"train_loss": 0.5341497659683228, "global_step": 19332, "epoch": 217, "lr": 9.88126627936883e-05}
{"train_loss": 0.47407615184783936, "global_step": 19333, "epoch": 217, "lr": 9.881253720069727e-05}
{"train_loss": 0.5352805852890015, "global_step": 19334, "epoch": 217, "lr": 9.8812411601144e-05}
{"train_loss": 0.6197686791419983, "global_step": 19335, "epoch": 217, "lr": 9.881228599502847e-05}
{"train_loss": 0.5571336150169373, "global_step": 19336, "epoch": 217, "lr": 9.881216038235075e-05}
{"train_loss": 0.5212767124176025, "global_step": 19337, "epoch": 217, "lr": 9.881203476311082e-05}
{"train_loss": 0.5297200679779053, "global_step": 19338, "epoch": 217, "lr": 9.881190913730872e-05}
{"train_loss": 0.4687921106815338, "global_step": 19339, "epoch": 217, "lr": 9.881178350494444e-05}
{"train_loss": 0.5677163600921631, "global_step": 19340, "epoch": 217, "lr": 9.881165786601801e-05}
{"train_loss": 0.33993515372276306, "global_step": 19341, "epoch": 217, "lr": 9.881153222052943e-05}
{"train_loss": 0.49036213755607605, "global_step": 19342, "epoch": 217, "lr": 9.881140656847875e-05}
{"train_loss": 0.4701908528804779, "global_step": 19343, "epoch": 217, "lr": 9.881128090986599e-05}
{"train_loss": 0.5228069424629211, "global_step": 19344, "epoch": 217, "lr": 9.881115524469114e-05}
{"train_loss": 0.43274983763694763, "global_step": 19345, "epoch": 217, "lr": 9.88110295729542e-05}
{"train_loss": 0.5700498223304749, "global_step": 19346, "epoch": 217, "lr": 9.881090389465523e-05}
{"train_loss": 0.5592699646949768, "global_step": 19347, "epoch": 217, "lr": 9.881077820979424e-05}
{"train_loss": 0.3728942573070526, "global_step": 19348, "epoch": 217, "lr": 9.881065251837123e-05}
{"train_loss": 0.5313200354576111, "global_step": 19349, "epoch": 217, "lr": 9.881052682038621e-05}
{"train_loss": 0.5082868933677673, "global_step": 19350, "epoch": 217, "lr": 9.881040111583922e-05}
{"train_loss": 0.44191911816596985, "global_step": 19351, "epoch": 217, "lr": 9.881027540473027e-05}
{"train_loss": 0.5518587827682495, "global_step": 19352, "epoch": 217, "lr": 9.881014968705936e-05}
{"train_loss": 0.5366565585136414, "global_step": 19353, "epoch": 217, "lr": 9.881002396282653e-05}
{"train_loss": 0.42745140194892883, "global_step": 19354, "epoch": 217, "lr": 9.880989823203179e-05}
{"train_loss": 0.6610754728317261, "global_step": 19355, "epoch": 217, "lr": 9.880977249467514e-05}
{"train_loss": 0.49467021226882935, "global_step": 19356, "epoch": 217, "lr": 9.880964675075662e-05}
{"train_loss": 0.5207429528236389, "global_step": 19357, "epoch": 217, "lr": 9.880952100027624e-05}
{"train_loss": 0.4877641201019287, "global_step": 19358, "epoch": 217, "lr": 9.880939524323401e-05}
{"train_loss": 0.46604105830192566, "global_step": 19359, "epoch": 217, "lr": 9.880926947962996e-05}
{"train_loss": 0.5414797067642212, "global_step": 19360, "epoch": 217, "lr": 9.88091437094641e-05}
{"train_loss": 0.44124704599380493, "global_step": 19361, "epoch": 217, "lr": 9.880901793273643e-05}
{"train_loss": 0.49811995029449463, "global_step": 19362, "epoch": 217, "lr": 9.880889214944699e-05}
{"train_loss": 0.4421863555908203, "global_step": 19363, "epoch": 217, "lr": 9.88087663595958e-05}
{"train_loss": 0.4737972319126129, "global_step": 19364, "epoch": 217, "lr": 9.880864056318284e-05}
{"train_loss": 0.4175318479537964, "global_step": 19365, "epoch": 217, "lr": 9.880851476020818e-05}
{"train_loss": 0.7094686031341553, "global_step": 19366, "epoch": 217, "lr": 9.88083889506718e-05}
{"train_loss": 0.5030919313430786, "global_step": 19367, "epoch": 217, "lr": 9.880826313457372e-05}
{"train_loss": 0.5232300758361816, "global_step": 19368, "epoch": 217, "lr": 9.880813731191396e-05}
{"train_loss": 0.49475589394569397, "global_step": 19369, "epoch": 217, "lr": 9.880801148269257e-05}
{"train_loss": 0.4989278018474579, "global_step": 19370, "epoch": 217, "lr": 9.88078856469095e-05}
{"train_loss": 0.46855610609054565, "global_step": 19371, "epoch": 217, "lr": 9.880775980456482e-05}
{"train_loss": 0.4537005126476288, "global_step": 19372, "epoch": 217, "lr": 9.880763395565855e-05}
{"train_loss": 0.5950300097465515, "global_step": 19373, "epoch": 217, "lr": 9.880750810019067e-05}
{"train_loss": 0.4424275755882263, "global_step": 19374, "epoch": 217, "lr": 9.880738223816121e-05}
{"train_loss": 0.5223798155784607, "global_step": 19375, "epoch": 217, "lr": 9.880725636957022e-05}
{"train_loss": 0.5848999619483948, "global_step": 19376, "epoch": 217, "lr": 9.880713049441767e-05}
{"train_loss": 0.4248773455619812, "global_step": 19377, "epoch": 217, "lr": 9.880700461270359e-05}
{"train_loss": 0.4592617452144623, "global_step": 19378, "epoch": 217, "lr": 9.880687872442801e-05}
{"train_loss": 0.44451022148132324, "global_step": 19379, "epoch": 217, "lr": 9.880675282959095e-05}
{"train_loss": 0.443662166595459, "global_step": 19380, "epoch": 217, "lr": 9.88066269281924e-05}
{"train_loss": 0.47040775418281555, "global_step": 19381, "epoch": 217, "lr": 9.88065010202324e-05}
{"train_loss": 0.4891921579837799, "global_step": 19382, "epoch": 217, "lr": 9.880637510571095e-05}
{"train_loss": 0.5626436471939087, "global_step": 19383, "epoch": 217, "lr": 9.880624918462811e-05}
{"train_loss": 0.5816616415977478, "global_step": 19384, "epoch": 217, "lr": 9.880612325698385e-05}
{"train_loss": 0.4458371102809906, "global_step": 19385, "epoch": 217, "lr": 9.880599732277818e-05}
{"train_loss": 0.4177832007408142, "global_step": 19386, "epoch": 217, "lr": 9.880587138201116e-05}
{"train_loss": 0.6566320657730103, "global_step": 19387, "epoch": 217, "lr": 9.880574543468277e-05}
{"train_loss": 0.3726748526096344, "global_step": 19388, "epoch": 217, "lr": 9.880561948079305e-05}
{"train_loss": 0.4458162188529968, "global_step": 19389, "epoch": 217, "lr": 9.880549352034202e-05}
{"train_loss": 0.45465636253356934, "global_step": 19390, "epoch": 217, "lr": 9.880536755332968e-05}
{"train_loss": 0.542218029499054, "global_step": 19391, "epoch": 217, "lr": 9.880524157975605e-05}
{"train_loss": 0.42309024930000305, "global_step": 19392, "epoch": 217, "lr": 9.880511559962114e-05}
{"train_loss": 0.5165566205978394, "global_step": 19393, "epoch": 217, "lr": 9.8804989612925e-05}
{"train_loss": 0.5616395473480225, "global_step": 19394, "epoch": 217, "lr": 9.880486361966761e-05}
{"train_loss": 0.5470930933952332, "global_step": 19395, "epoch": 217, "lr": 9.8804737619849e-05}
{"train_loss": 0.6010405421257019, "global_step": 19396, "epoch": 217, "lr": 9.880461161346919e-05}
{"train_loss": 0.5223311185836792, "global_step": 19397, "epoch": 217, "lr": 9.88044856005282e-05}
{"train_loss": 0.5175036787986755, "global_step": 19398, "epoch": 217, "lr": 9.880435958102603e-05}
{"train_loss": 0.4541530907154083, "global_step": 19399, "epoch": 217, "lr": 9.880423355496271e-05}
{"train_loss": 0.5551894903182983, "global_step": 19400, "epoch": 217, "lr": 9.880410752233827e-05}
{"train_loss": 0.5001160177621948, "global_step": 19401, "epoch": 217, "lr": 9.880398148315271e-05, "val_loss": 1.722238302230835}
{"train_loss": 0.5448192358016968, "global_step": 19402, "epoch": 218, "lr": 9.880385543740605e-05}
{"train_loss": 0.5271670818328857, "global_step": 19403, "epoch": 218, "lr": 9.880372938509828e-05}
{"train_loss": 0.3820738196372986, "global_step": 19404, "epoch": 218, "lr": 9.880360332622947e-05}
{"train_loss": 0.5252376198768616, "global_step": 19405, "epoch": 218, "lr": 9.88034772607996e-05}
{"train_loss": 0.5341445207595825, "global_step": 19406, "epoch": 218, "lr": 9.88033511888087e-05}
{"train_loss": 0.378332257270813, "global_step": 19407, "epoch": 218, "lr": 9.880322511025679e-05}
{"train_loss": 0.5057255625724792, "global_step": 19408, "epoch": 218, "lr": 9.880309902514389e-05}
{"train_loss": 0.47909823060035706, "global_step": 19409, "epoch": 218, "lr": 9.880297293346999e-05}
{"train_loss": 0.4983994960784912, "global_step": 19410, "epoch": 218, "lr": 9.880284683523513e-05}
{"train_loss": 0.3826189339160919, "global_step": 19411, "epoch": 218, "lr": 9.880272073043933e-05}
{"train_loss": 0.5197274088859558, "global_step": 19412, "epoch": 218, "lr": 9.880259461908258e-05}
{"train_loss": 0.4365730285644531, "global_step": 19413, "epoch": 218, "lr": 9.880246850116493e-05}
{"train_loss": 0.5372740626335144, "global_step": 19414, "epoch": 218, "lr": 9.880234237668638e-05}
{"train_loss": 0.5474699139595032, "global_step": 19415, "epoch": 218, "lr": 9.880221624564694e-05}
{"train_loss": 0.5302473306655884, "global_step": 19416, "epoch": 218, "lr": 9.880209010804666e-05}
{"train_loss": 0.6069725751876831, "global_step": 19417, "epoch": 218, "lr": 9.880196396388551e-05}
{"train_loss": 0.34841692447662354, "global_step": 19418, "epoch": 218, "lr": 9.880183781316355e-05}
{"train_loss": 0.5289742350578308, "global_step": 19419, "epoch": 218, "lr": 9.880171165588077e-05}
{"train_loss": 0.4553129971027374, "global_step": 19420, "epoch": 218, "lr": 9.880158549203718e-05}
{"train_loss": 0.47388288378715515, "global_step": 19421, "epoch": 218, "lr": 9.880145932163283e-05}
{"train_loss": 0.38883161544799805, "global_step": 19422, "epoch": 218, "lr": 9.880133314466771e-05}
{"train_loss": 0.4169662892818451, "global_step": 19423, "epoch": 218, "lr": 9.880120696114186e-05}
{"train_loss": 0.3857251703739166, "global_step": 19424, "epoch": 218, "lr": 9.880108077105527e-05}
{"train_loss": 0.3741481900215149, "global_step": 19425, "epoch": 218, "lr": 9.880095457440797e-05}
{"train_loss": 0.4283219277858734, "global_step": 19426, "epoch": 218, "lr": 9.880082837119997e-05}
{"train_loss": 0.4453577399253845, "global_step": 19427, "epoch": 218, "lr": 9.880070216143131e-05}
{"train_loss": 0.5614146590232849, "global_step": 19428, "epoch": 218, "lr": 9.880057594510198e-05}
{"train_loss": 0.4429560601711273, "global_step": 19429, "epoch": 218, "lr": 9.8800449722212e-05}
{"train_loss": 0.49958163499832153, "global_step": 19430, "epoch": 218, "lr": 9.880032349276142e-05}
{"train_loss": 0.40439680218696594, "global_step": 19431, "epoch": 218, "lr": 9.88001972567502e-05}
{"train_loss": 0.34760046005249023, "global_step": 19432, "epoch": 218, "lr": 9.880007101417841e-05}
{"train_loss": 0.47462666034698486, "global_step": 19433, "epoch": 218, "lr": 9.879994476504604e-05}
{"train_loss": 0.525141716003418, "global_step": 19434, "epoch": 218, "lr": 9.879981850935311e-05}
{"train_loss": 0.6651102304458618, "global_step": 19435, "epoch": 218, "lr": 9.879969224709964e-05}
{"train_loss": 0.5528351068496704, "global_step": 19436, "epoch": 218, "lr": 9.879956597828564e-05}
{"train_loss": 0.50263512134552, "global_step": 19437, "epoch": 218, "lr": 9.879943970291114e-05}
{"train_loss": 0.5120323896408081, "global_step": 19438, "epoch": 218, "lr": 9.879931342097614e-05}
{"train_loss": 0.35896173119544983, "global_step": 19439, "epoch": 218, "lr": 9.879918713248068e-05}
{"train_loss": 0.506292998790741, "global_step": 19440, "epoch": 218, "lr": 9.879906083742478e-05}
{"train_loss": 0.6841822266578674, "global_step": 19441, "epoch": 218, "lr": 9.879893453580842e-05}
{"train_loss": 0.5313448905944824, "global_step": 19442, "epoch": 218, "lr": 9.879880822763163e-05}
{"train_loss": 0.5408251881599426, "global_step": 19443, "epoch": 218, "lr": 9.879868191289445e-05}
{"train_loss": 0.39300620555877686, "global_step": 19444, "epoch": 218, "lr": 9.879855559159687e-05}
{"train_loss": 0.4992128014564514, "global_step": 19445, "epoch": 218, "lr": 9.879842926373892e-05}
{"train_loss": 0.48617079854011536, "global_step": 19446, "epoch": 218, "lr": 9.879830292932062e-05}
{"train_loss": 0.5839062333106995, "global_step": 19447, "epoch": 218, "lr": 9.8798176588342e-05}
{"train_loss": 0.5094900727272034, "global_step": 19448, "epoch": 218, "lr": 9.879805024080304e-05}
{"train_loss": 0.5089648962020874, "global_step": 19449, "epoch": 218, "lr": 9.879792388670378e-05}
{"train_loss": 0.41633689403533936, "global_step": 19450, "epoch": 218, "lr": 9.879779752604424e-05}
{"train_loss": 0.38980358839035034, "global_step": 19451, "epoch": 218, "lr": 9.879767115882441e-05}
{"train_loss": 0.512116014957428, "global_step": 19452, "epoch": 218, "lr": 9.879754478504434e-05}
{"train_loss": 0.5388610363006592, "global_step": 19453, "epoch": 218, "lr": 9.879741840470404e-05}
{"train_loss": 0.5539351105690002, "global_step": 19454, "epoch": 218, "lr": 9.879729201780353e-05}
{"train_loss": 0.3839164078235626, "global_step": 19455, "epoch": 218, "lr": 9.879716562434281e-05}
{"train_loss": 0.44175195693969727, "global_step": 19456, "epoch": 218, "lr": 9.87970392243219e-05}
{"train_loss": 0.4306719899177551, "global_step": 19457, "epoch": 218, "lr": 9.879691281774084e-05}
{"train_loss": 0.5178357362747192, "global_step": 19458, "epoch": 218, "lr": 9.879678640459962e-05}
{"train_loss": 0.4554447531700134, "global_step": 19459, "epoch": 218, "lr": 9.879665998489827e-05}
{"train_loss": 0.6764414310455322, "global_step": 19460, "epoch": 218, "lr": 9.879653355863679e-05}
{"train_loss": 0.5453056693077087, "global_step": 19461, "epoch": 218, "lr": 9.879640712581522e-05}
{"train_loss": 0.45095986127853394, "global_step": 19462, "epoch": 218, "lr": 9.879628068643356e-05}
{"train_loss": 0.3625163733959198, "global_step": 19463, "epoch": 218, "lr": 9.879615424049185e-05}
{"train_loss": 0.47925928235054016, "global_step": 19464, "epoch": 218, "lr": 9.879602778799007e-05}
{"train_loss": 0.4429702162742615, "global_step": 19465, "epoch": 218, "lr": 9.879590132892829e-05}
{"train_loss": 0.6144293546676636, "global_step": 19466, "epoch": 218, "lr": 9.879577486330647e-05}
{"train_loss": 0.426941454410553, "global_step": 19467, "epoch": 218, "lr": 9.879564839112465e-05}
{"train_loss": 0.5522396564483643, "global_step": 19468, "epoch": 218, "lr": 9.879552191238288e-05}
{"train_loss": 0.5026516914367676, "global_step": 19469, "epoch": 218, "lr": 9.879539542708111e-05}
{"train_loss": 0.41946902871131897, "global_step": 19470, "epoch": 218, "lr": 9.879526893521942e-05}
{"train_loss": 0.42207419872283936, "global_step": 19471, "epoch": 218, "lr": 9.87951424367978e-05}
{"train_loss": 0.5065224170684814, "global_step": 19472, "epoch": 218, "lr": 9.879501593181626e-05}
{"train_loss": 0.5180399417877197, "global_step": 19473, "epoch": 218, "lr": 9.879488942027482e-05}
{"train_loss": 0.6445948481559753, "global_step": 19474, "epoch": 218, "lr": 9.879476290217351e-05}
{"train_loss": 0.5998615622520447, "global_step": 19475, "epoch": 218, "lr": 9.879463637751233e-05}
{"train_loss": 0.601753294467926, "global_step": 19476, "epoch": 218, "lr": 9.87945098462913e-05}
{"train_loss": 0.581709086894989, "global_step": 19477, "epoch": 218, "lr": 9.879438330851046e-05}
{"train_loss": 0.5154122114181519, "global_step": 19478, "epoch": 218, "lr": 9.879425676416981e-05}
{"train_loss": 0.4493964910507202, "global_step": 19479, "epoch": 218, "lr": 9.879413021326936e-05}
{"train_loss": 0.5719438195228577, "global_step": 19480, "epoch": 218, "lr": 9.879400365580912e-05}
{"train_loss": 0.5548073053359985, "global_step": 19481, "epoch": 218, "lr": 9.879387709178916e-05}
{"train_loss": 0.47832268476486206, "global_step": 19482, "epoch": 218, "lr": 9.879375052120943e-05}
{"train_loss": 0.4542626142501831, "global_step": 19483, "epoch": 218, "lr": 9.879362394406997e-05}
{"train_loss": 0.4215843975543976, "global_step": 19484, "epoch": 218, "lr": 9.879349736037083e-05}
{"train_loss": 0.5549721717834473, "global_step": 19485, "epoch": 218, "lr": 9.879337077011197e-05}
{"train_loss": 0.5315153002738953, "global_step": 19486, "epoch": 218, "lr": 9.879324417329344e-05}
{"train_loss": 0.5993072986602783, "global_step": 19487, "epoch": 218, "lr": 9.879311756991526e-05}
{"train_loss": 0.5361190438270569, "global_step": 19488, "epoch": 218, "lr": 9.879299095997743e-05}
{"train_loss": 0.6596472263336182, "global_step": 19489, "epoch": 218, "lr": 9.879286434348e-05}
{"train_loss": 0.49544734834285264, "global_step": 19490, "epoch": 218, "lr": 9.879273772042294e-05, "val_loss": 1.6928805112838745}
{"train_loss": 0.5039322972297668, "global_step": 19491, "epoch": 219, "lr": 9.87926110908063e-05}
{"train_loss": 0.4605315327644348, "global_step": 19492, "epoch": 219, "lr": 9.87924844546301e-05}
{"train_loss": 0.6832141280174255, "global_step": 19493, "epoch": 219, "lr": 9.879235781189433e-05}
{"train_loss": 0.48783376812934875, "global_step": 19494, "epoch": 219, "lr": 9.879223116259903e-05}
{"train_loss": 0.4534752368927002, "global_step": 19495, "epoch": 219, "lr": 9.879210450674421e-05}
{"train_loss": 0.513043999671936, "global_step": 19496, "epoch": 219, "lr": 9.879197784432988e-05}
{"train_loss": 0.5272064805030823, "global_step": 19497, "epoch": 219, "lr": 9.879185117535605e-05}
{"train_loss": 0.4170372188091278, "global_step": 19498, "epoch": 219, "lr": 9.879172449982278e-05}
{"train_loss": 0.6449456810951233, "global_step": 19499, "epoch": 219, "lr": 9.879159781773004e-05}
{"train_loss": 0.41101816296577454, "global_step": 19500, "epoch": 219, "lr": 9.879147112907787e-05}
{"train_loss": 0.6094803214073181, "global_step": 19501, "epoch": 219, "lr": 9.879134443386629e-05}
{"train_loss": 0.45065054297447205, "global_step": 19502, "epoch": 219, "lr": 9.879121773209528e-05}
{"train_loss": 0.4770878851413727, "global_step": 19503, "epoch": 219, "lr": 9.879109102376491e-05}
{"train_loss": 0.632573127746582, "global_step": 19504, "epoch": 219, "lr": 9.879096430887515e-05}
{"train_loss": 0.5438051819801331, "global_step": 19505, "epoch": 219, "lr": 9.879083758742606e-05}
{"train_loss": 0.4975290894508362, "global_step": 19506, "epoch": 219, "lr": 9.879071085941764e-05}
{"train_loss": 0.49082571268081665, "global_step": 19507, "epoch": 219, "lr": 9.87905841248499e-05}
{"train_loss": 0.3968077301979065, "global_step": 19508, "epoch": 219, "lr": 9.879045738372285e-05}
{"train_loss": 0.49434882402420044, "global_step": 19509, "epoch": 219, "lr": 9.879033063603652e-05}
{"train_loss": 0.41802147030830383, "global_step": 19510, "epoch": 219, "lr": 9.879020388179092e-05}
{"train_loss": 0.4573849141597748, "global_step": 19511, "epoch": 219, "lr": 9.879007712098609e-05}
{"train_loss": 0.5333998799324036, "global_step": 19512, "epoch": 219, "lr": 9.878995035362202e-05}
{"train_loss": 0.44138047099113464, "global_step": 19513, "epoch": 219, "lr": 9.878982357969872e-05}
{"train_loss": 0.412221223115921, "global_step": 19514, "epoch": 219, "lr": 9.878969679921625e-05}
{"train_loss": 0.5990422368049622, "global_step": 19515, "epoch": 219, "lr": 9.878957001217459e-05}
{"train_loss": 0.6173174977302551, "global_step": 19516, "epoch": 219, "lr": 9.878944321857376e-05}
{"train_loss": 0.508316159248352, "global_step": 19517, "epoch": 219, "lr": 9.878931641841379e-05}
{"train_loss": 0.6747702360153198, "global_step": 19518, "epoch": 219, "lr": 9.878918961169469e-05}
{"train_loss": 0.5363447070121765, "global_step": 19519, "epoch": 219, "lr": 9.878906279841647e-05}
{"train_loss": 0.41668304800987244, "global_step": 19520, "epoch": 219, "lr": 9.878893597857916e-05}
{"train_loss": 0.4922577440738678, "global_step": 19521, "epoch": 219, "lr": 9.878880915218279e-05}
{"train_loss": 0.4641569256782532, "global_step": 19522, "epoch": 219, "lr": 9.878868231922733e-05}
{"train_loss": 0.4427441358566284, "global_step": 19523, "epoch": 219, "lr": 9.878855547971284e-05}
{"train_loss": 0.5034439563751221, "global_step": 19524, "epoch": 219, "lr": 9.878842863363933e-05}
{"train_loss": 0.7236895561218262, "global_step": 19525, "epoch": 219, "lr": 9.87883017810068e-05}
{"train_loss": 0.48701247572898865, "global_step": 19526, "epoch": 219, "lr": 9.878817492181529e-05}
{"train_loss": 0.517808198928833, "global_step": 19527, "epoch": 219, "lr": 9.878804805606479e-05}
{"train_loss": 0.46020564436912537, "global_step": 19528, "epoch": 219, "lr": 9.878792118375534e-05}
{"train_loss": 0.5787619948387146, "global_step": 19529, "epoch": 219, "lr": 9.878779430488693e-05}
{"train_loss": 0.5972350239753723, "global_step": 19530, "epoch": 219, "lr": 9.878766741945963e-05}
{"train_loss": 0.5615797638893127, "global_step": 19531, "epoch": 219, "lr": 9.878754052747341e-05}
{"train_loss": 0.5597745180130005, "global_step": 19532, "epoch": 219, "lr": 9.87874136289283e-05}
{"train_loss": 0.5299405455589294, "global_step": 19533, "epoch": 219, "lr": 9.878728672382429e-05}
{"train_loss": 0.5044028759002686, "global_step": 19534, "epoch": 219, "lr": 9.878715981216146e-05}
{"train_loss": 0.608898937702179, "global_step": 19535, "epoch": 219, "lr": 9.878703289393977e-05}
{"train_loss": 0.6004335880279541, "global_step": 19536, "epoch": 219, "lr": 9.878690596915927e-05}
{"train_loss": 0.5291220545768738, "global_step": 19537, "epoch": 219, "lr": 9.878677903781996e-05}
{"train_loss": 0.6183551549911499, "global_step": 19538, "epoch": 219, "lr": 9.878665209992186e-05}
{"train_loss": 0.5580527782440186, "global_step": 19539, "epoch": 219, "lr": 9.878652515546498e-05}
{"train_loss": 0.38198789954185486, "global_step": 19540, "epoch": 219, "lr": 9.878639820444937e-05}
{"train_loss": 0.5074551105499268, "global_step": 19541, "epoch": 219, "lr": 9.8786271246875e-05}
{"train_loss": 0.43196815252304077, "global_step": 19542, "epoch": 219, "lr": 9.878614428274194e-05}
{"train_loss": 0.5937491059303284, "global_step": 19543, "epoch": 219, "lr": 9.878601731205015e-05}
{"train_loss": 0.5087488889694214, "global_step": 19544, "epoch": 219, "lr": 9.878589033479968e-05}
{"train_loss": 0.6541635990142822, "global_step": 19545, "epoch": 219, "lr": 9.878576335099054e-05}
{"train_loss": 0.6327688097953796, "global_step": 19546, "epoch": 219, "lr": 9.878563636062276e-05}
{"train_loss": 0.43243715167045593, "global_step": 19547, "epoch": 219, "lr": 9.878550936369633e-05}
{"train_loss": 0.47457319498062134, "global_step": 19548, "epoch": 219, "lr": 9.878538236021129e-05}
{"train_loss": 0.4351613521575928, "global_step": 19549, "epoch": 219, "lr": 9.878525535016766e-05}
{"train_loss": 0.46717751026153564, "global_step": 19550, "epoch": 219, "lr": 9.878512833356544e-05}
{"train_loss": 0.5655624866485596, "global_step": 19551, "epoch": 219, "lr": 9.878500131040465e-05}
{"train_loss": 0.479437917470932, "global_step": 19552, "epoch": 219, "lr": 9.878487428068532e-05}
{"train_loss": 0.46517643332481384, "global_step": 19553, "epoch": 219, "lr": 9.878474724440745e-05}
{"train_loss": 0.45700252056121826, "global_step": 19554, "epoch": 219, "lr": 9.878462020157107e-05}
{"train_loss": 0.5726135969161987, "global_step": 19555, "epoch": 219, "lr": 9.87844931521762e-05}
{"train_loss": 0.6651178002357483, "global_step": 19556, "epoch": 219, "lr": 9.878436609622283e-05}
{"train_loss": 0.5214372277259827, "global_step": 19557, "epoch": 219, "lr": 9.878423903371102e-05}
{"train_loss": 0.48458391427993774, "global_step": 19558, "epoch": 219, "lr": 9.878411196464073e-05}
{"train_loss": 0.5490069389343262, "global_step": 19559, "epoch": 219, "lr": 9.878398488901204e-05}
{"train_loss": 0.43960118293762207, "global_step": 19560, "epoch": 219, "lr": 9.878385780682493e-05}
{"train_loss": 0.4703183174133301, "global_step": 19561, "epoch": 219, "lr": 9.878373071807942e-05}
{"train_loss": 0.5958114266395569, "global_step": 19562, "epoch": 219, "lr": 9.878360362277554e-05}
{"train_loss": 0.5545229911804199, "global_step": 19563, "epoch": 219, "lr": 9.878347652091328e-05}
{"train_loss": 0.5073171257972717, "global_step": 19564, "epoch": 219, "lr": 9.878334941249269e-05}
{"train_loss": 0.5710790157318115, "global_step": 19565, "epoch": 219, "lr": 9.878322229751378e-05}
{"train_loss": 0.5588968992233276, "global_step": 19566, "epoch": 219, "lr": 9.878309517597656e-05}
{"train_loss": 0.4788622558116913, "global_step": 19567, "epoch": 219, "lr": 9.878296804788104e-05}
{"train_loss": 0.43700098991394043, "global_step": 19568, "epoch": 219, "lr": 9.878284091322725e-05}
{"train_loss": 0.4839867651462555, "global_step": 19569, "epoch": 219, "lr": 9.87827137720152e-05}
{"train_loss": 0.5030554533004761, "global_step": 19570, "epoch": 219, "lr": 9.878258662424491e-05}
{"train_loss": 0.4590562582015991, "global_step": 19571, "epoch": 219, "lr": 9.878245946991638e-05}
{"train_loss": 0.5525420904159546, "global_step": 19572, "epoch": 219, "lr": 9.878233230902966e-05}
{"train_loss": 0.437968909740448, "global_step": 19573, "epoch": 219, "lr": 9.878220514158475e-05}
{"train_loss": 0.4467198848724365, "global_step": 19574, "epoch": 219, "lr": 9.878207796758165e-05}
{"train_loss": 0.44272905588150024, "global_step": 19575, "epoch": 219, "lr": 9.878195078702042e-05}
{"train_loss": 0.5439922213554382, "global_step": 19576, "epoch": 219, "lr": 9.878182359990104e-05}
{"train_loss": 0.41114041209220886, "global_step": 19577, "epoch": 219, "lr": 9.878169640622353e-05}
{"train_loss": 0.39956143498420715, "global_step": 19578, "epoch": 219, "lr": 9.878156920598793e-05}
{"train_loss": 0.5124413397874725, "global_step": 19579, "epoch": 219, "lr": 9.878144199919423e-05, "val_loss": 1.683954119682312}
{"train_loss": 0.3904876410961151, "global_step": 19580, "epoch": 220, "lr": 9.878131478584245e-05}
{"train_loss": 0.517055869102478, "global_step": 19581, "epoch": 220, "lr": 9.878118756593264e-05}
{"train_loss": 0.4442434012889862, "global_step": 19582, "epoch": 220, "lr": 9.878106033946479e-05}
{"train_loss": 0.34400683641433716, "global_step": 19583, "epoch": 220, "lr": 9.878093310643892e-05}
{"train_loss": 0.6064602136611938, "global_step": 19584, "epoch": 220, "lr": 9.878080586685504e-05}
{"train_loss": 0.5376672744750977, "global_step": 19585, "epoch": 220, "lr": 9.878067862071318e-05}
{"train_loss": 0.3754221498966217, "global_step": 19586, "epoch": 220, "lr": 9.878055136801335e-05}
{"train_loss": 0.33207494020462036, "global_step": 19587, "epoch": 220, "lr": 9.878042410875556e-05}
{"train_loss": 0.5185296535491943, "global_step": 19588, "epoch": 220, "lr": 9.878029684293986e-05}
{"train_loss": 0.5100234746932983, "global_step": 19589, "epoch": 220, "lr": 9.878016957056623e-05}
{"train_loss": 0.48688170313835144, "global_step": 19590, "epoch": 220, "lr": 9.87800422916347e-05}
{"train_loss": 0.5421482920646667, "global_step": 19591, "epoch": 220, "lr": 9.87799150061453e-05}
{"train_loss": 0.4742431938648224, "global_step": 19592, "epoch": 220, "lr": 9.877978771409802e-05}
{"train_loss": 0.43517547845840454, "global_step": 19593, "epoch": 220, "lr": 9.877966041549289e-05}
{"train_loss": 0.5156422853469849, "global_step": 19594, "epoch": 220, "lr": 9.877953311032995e-05}
{"train_loss": 0.4681236147880554, "global_step": 19595, "epoch": 220, "lr": 9.877940579860918e-05}
{"train_loss": 0.6062140464782715, "global_step": 19596, "epoch": 220, "lr": 9.87792784803306e-05}
{"train_loss": 0.6406651139259338, "global_step": 19597, "epoch": 220, "lr": 9.877915115549426e-05}
{"train_loss": 0.459892600774765, "global_step": 19598, "epoch": 220, "lr": 9.877902382410016e-05}
{"train_loss": 0.4607512652873993, "global_step": 19599, "epoch": 220, "lr": 9.87788964861483e-05}
{"train_loss": 0.4459618628025055, "global_step": 19600, "epoch": 220, "lr": 9.877876914163873e-05}
{"train_loss": 0.5398545265197754, "global_step": 19601, "epoch": 220, "lr": 9.877864179057143e-05}
{"train_loss": 0.6372497081756592, "global_step": 19602, "epoch": 220, "lr": 9.877851443294646e-05}
{"train_loss": 0.36626917123794556, "global_step": 19603, "epoch": 220, "lr": 9.87783870687638e-05}
{"train_loss": 0.4002143144607544, "global_step": 19604, "epoch": 220, "lr": 9.877825969802348e-05}
{"train_loss": 0.49898913502693176, "global_step": 19605, "epoch": 220, "lr": 9.87781323207255e-05}
{"train_loss": 0.5519041419029236, "global_step": 19606, "epoch": 220, "lr": 9.877800493686993e-05}
{"train_loss": 0.4967264235019684, "global_step": 19607, "epoch": 220, "lr": 9.877787754645673e-05}
{"train_loss": 0.5222592353820801, "global_step": 19608, "epoch": 220, "lr": 9.877775014948594e-05}
{"train_loss": 0.3525262773036957, "global_step": 19609, "epoch": 220, "lr": 9.877762274595758e-05}
{"train_loss": 0.42472341656684875, "global_step": 19610, "epoch": 220, "lr": 9.877749533587166e-05}
{"train_loss": 0.4700554609298706, "global_step": 19611, "epoch": 220, "lr": 9.87773679192282e-05}
{"train_loss": 0.42708346247673035, "global_step": 19612, "epoch": 220, "lr": 9.87772404960272e-05}
{"train_loss": 0.5286583304405212, "global_step": 19613, "epoch": 220, "lr": 9.877711306626873e-05}
{"train_loss": 0.48900753259658813, "global_step": 19614, "epoch": 220, "lr": 9.877698562995275e-05}
{"train_loss": 0.38082778453826904, "global_step": 19615, "epoch": 220, "lr": 9.87768581870793e-05}
{"train_loss": 0.4854796230792999, "global_step": 19616, "epoch": 220, "lr": 9.877673073764841e-05}
{"train_loss": 0.5345562696456909, "global_step": 19617, "epoch": 220, "lr": 9.877660328166008e-05}
{"train_loss": 0.4366483688354492, "global_step": 19618, "epoch": 220, "lr": 9.877647581911431e-05}
{"train_loss": 0.5594757795333862, "global_step": 19619, "epoch": 220, "lr": 9.877634835001115e-05}
{"train_loss": 0.49548205733299255, "global_step": 19620, "epoch": 220, "lr": 9.87762208743506e-05}
{"train_loss": 0.5310742855072021, "global_step": 19621, "epoch": 220, "lr": 9.87760933921327e-05}
{"train_loss": 0.6093881130218506, "global_step": 19622, "epoch": 220, "lr": 9.877596590335742e-05}
{"train_loss": 0.4565978944301605, "global_step": 19623, "epoch": 220, "lr": 9.877583840802484e-05}
{"train_loss": 0.3939669728279114, "global_step": 19624, "epoch": 220, "lr": 9.87757109061349e-05}
{"train_loss": 0.4979900121688843, "global_step": 19625, "epoch": 220, "lr": 9.87755833976877e-05}
{"train_loss": 0.5092115998268127, "global_step": 19626, "epoch": 220, "lr": 9.87754558826832e-05}
{"train_loss": 0.4794938564300537, "global_step": 19627, "epoch": 220, "lr": 9.877532836112144e-05}
{"train_loss": 0.5273558497428894, "global_step": 19628, "epoch": 220, "lr": 9.877520083300243e-05}
{"train_loss": 0.40749749541282654, "global_step": 19629, "epoch": 220, "lr": 9.877507329832618e-05}
{"train_loss": 0.5724973678588867, "global_step": 19630, "epoch": 220, "lr": 9.877494575709273e-05}
{"train_loss": 0.3889677822589874, "global_step": 19631, "epoch": 220, "lr": 9.877481820930208e-05}
{"train_loss": 0.49354735016822815, "global_step": 19632, "epoch": 220, "lr": 9.877469065495423e-05}
{"train_loss": 0.6640541553497314, "global_step": 19633, "epoch": 220, "lr": 9.877456309404924e-05}
{"train_loss": 0.46006888151168823, "global_step": 19634, "epoch": 220, "lr": 9.87744355265871e-05}
{"train_loss": 0.5319274663925171, "global_step": 19635, "epoch": 220, "lr": 9.877430795256784e-05}
{"train_loss": 0.5029287338256836, "global_step": 19636, "epoch": 220, "lr": 9.877418037199144e-05}
{"train_loss": 0.5620469450950623, "global_step": 19637, "epoch": 220, "lr": 9.877405278485796e-05}
{"train_loss": 0.4762789309024811, "global_step": 19638, "epoch": 220, "lr": 9.877392519116741e-05}
{"train_loss": 0.5468734502792358, "global_step": 19639, "epoch": 220, "lr": 9.877379759091981e-05}
{"train_loss": 0.5385717153549194, "global_step": 19640, "epoch": 220, "lr": 9.877366998411514e-05}
{"train_loss": 0.4665387272834778, "global_step": 19641, "epoch": 220, "lr": 9.877354237075346e-05}
{"train_loss": 0.4639270603656769, "global_step": 19642, "epoch": 220, "lr": 9.877341475083478e-05}
{"train_loss": 0.5868407487869263, "global_step": 19643, "epoch": 220, "lr": 9.877328712435909e-05}
{"train_loss": 0.6249722838401794, "global_step": 19644, "epoch": 220, "lr": 9.877315949132643e-05}
{"train_loss": 0.6246707439422607, "global_step": 19645, "epoch": 220, "lr": 9.877303185173684e-05}
{"train_loss": 0.6332666873931885, "global_step": 19646, "epoch": 220, "lr": 9.877290420559028e-05}
{"train_loss": 0.4218067228794098, "global_step": 19647, "epoch": 220, "lr": 9.877277655288681e-05}
{"train_loss": 0.5805999040603638, "global_step": 19648, "epoch": 220, "lr": 9.877264889362644e-05}
{"train_loss": 0.5133066773414612, "global_step": 19649, "epoch": 220, "lr": 9.877252122780916e-05}
{"train_loss": 0.46445155143737793, "global_step": 19650, "epoch": 220, "lr": 9.877239355543502e-05}
{"train_loss": 0.42592960596084595, "global_step": 19651, "epoch": 220, "lr": 9.877226587650404e-05}
{"train_loss": 0.3751457631587982, "global_step": 19652, "epoch": 220, "lr": 9.87721381910162e-05}
{"train_loss": 0.4435189366340637, "global_step": 19653, "epoch": 220, "lr": 9.877201049897156e-05}
{"train_loss": 0.5854030847549438, "global_step": 19654, "epoch": 220, "lr": 9.877188280037011e-05}
{"train_loss": 0.4815802276134491, "global_step": 19655, "epoch": 220, "lr": 9.877175509521186e-05}
{"train_loss": 0.49801522493362427, "global_step": 19656, "epoch": 220, "lr": 9.877162738349686e-05}
{"train_loss": 0.5270520448684692, "global_step": 19657, "epoch": 220, "lr": 9.877149966522511e-05}
{"train_loss": 0.49427908658981323, "global_step": 19658, "epoch": 220, "lr": 9.877137194039662e-05}
{"train_loss": 0.4676385819911957, "global_step": 19659, "epoch": 220, "lr": 9.877124420901142e-05}
{"train_loss": 0.5313074588775635, "global_step": 19660, "epoch": 220, "lr": 9.877111647106952e-05}
{"train_loss": 0.4714532494544983, "global_step": 19661, "epoch": 220, "lr": 9.877098872657092e-05}
{"train_loss": 0.4964665472507477, "global_step": 19662, "epoch": 220, "lr": 9.877086097551567e-05}
{"train_loss": 0.44780775904655457, "global_step": 19663, "epoch": 220, "lr": 9.877073321790378e-05}
{"train_loss": 0.5244890451431274, "global_step": 19664, "epoch": 220, "lr": 9.877060545373525e-05}
{"train_loss": 0.641808032989502, "global_step": 19665, "epoch": 220, "lr": 9.87704776830101e-05}
{"train_loss": 0.6533588767051697, "global_step": 19666, "epoch": 220, "lr": 9.877034990572835e-05}
{"train_loss": 0.4031884968280792, "global_step": 19667, "epoch": 220, "lr": 9.877022212189005e-05}
{"train_loss": 0.4967722316806236, "global_step": 19668, "epoch": 220, "lr": 9.877009433149516e-05, "val_loss": 1.6972357034683228, "train_action_mse_error": 37.760223388671875}
{"train_loss": 0.4789927303791046, "global_step": 19669, "epoch": 221, "lr": 9.876996653454373e-05}
{"train_loss": 0.49297136068344116, "global_step": 19670, "epoch": 221, "lr": 9.876983873103578e-05}
{"train_loss": 0.6018224954605103, "global_step": 19671, "epoch": 221, "lr": 9.876971092097132e-05}
{"train_loss": 0.491742342710495, "global_step": 19672, "epoch": 221, "lr": 9.876958310435036e-05}
{"train_loss": 0.398351788520813, "global_step": 19673, "epoch": 221, "lr": 9.876945528117293e-05}
{"train_loss": 0.4904314875602722, "global_step": 19674, "epoch": 221, "lr": 9.876932745143903e-05}
{"train_loss": 0.5629231929779053, "global_step": 19675, "epoch": 221, "lr": 9.87691996151487e-05}
{"train_loss": 0.4352596700191498, "global_step": 19676, "epoch": 221, "lr": 9.876907177230194e-05}
{"train_loss": 0.4268282651901245, "global_step": 19677, "epoch": 221, "lr": 9.876894392289879e-05}
{"train_loss": 0.5282172560691833, "global_step": 19678, "epoch": 221, "lr": 9.876881606693923e-05}
{"train_loss": 0.4402674436569214, "global_step": 19679, "epoch": 221, "lr": 9.87686882044233e-05}
{"train_loss": 0.4989941716194153, "global_step": 19680, "epoch": 221, "lr": 9.876856033535102e-05}
{"train_loss": 0.49888378381729126, "global_step": 19681, "epoch": 221, "lr": 9.87684324597224e-05}
{"train_loss": 0.429010808467865, "global_step": 19682, "epoch": 221, "lr": 9.876830457753745e-05}
{"train_loss": 0.5232564806938171, "global_step": 19683, "epoch": 221, "lr": 9.87681766887962e-05}
{"train_loss": 0.4686781167984009, "global_step": 19684, "epoch": 221, "lr": 9.876804879349867e-05}
{"train_loss": 0.44190657138824463, "global_step": 19685, "epoch": 221, "lr": 9.876792089164487e-05}
{"train_loss": 0.5872613191604614, "global_step": 19686, "epoch": 221, "lr": 9.87677929832348e-05}
{"train_loss": 0.5689915418624878, "global_step": 19687, "epoch": 221, "lr": 9.876766506826851e-05}
{"train_loss": 0.37599024176597595, "global_step": 19688, "epoch": 221, "lr": 9.8767537146746e-05}
{"train_loss": 0.4403964579105377, "global_step": 19689, "epoch": 221, "lr": 9.876740921866729e-05}
{"train_loss": 0.4142107665538788, "global_step": 19690, "epoch": 221, "lr": 9.87672812840324e-05}
{"train_loss": 0.42706960439682007, "global_step": 19691, "epoch": 221, "lr": 9.876715334284134e-05}
{"train_loss": 0.4512518346309662, "global_step": 19692, "epoch": 221, "lr": 9.876702539509412e-05}
{"train_loss": 0.5075950622558594, "global_step": 19693, "epoch": 221, "lr": 9.876689744079077e-05}
{"train_loss": 0.4045727252960205, "global_step": 19694, "epoch": 221, "lr": 9.876676947993132e-05}
{"train_loss": 0.44261500239372253, "global_step": 19695, "epoch": 221, "lr": 9.876664151251577e-05}
{"train_loss": 0.4000338613986969, "global_step": 19696, "epoch": 221, "lr": 9.876651353854414e-05}
{"train_loss": 0.4312340319156647, "global_step": 19697, "epoch": 221, "lr": 9.876638555801643e-05}
{"train_loss": 0.4932266175746918, "global_step": 19698, "epoch": 221, "lr": 9.876625757093269e-05}
{"train_loss": 0.6622203588485718, "global_step": 19699, "epoch": 221, "lr": 9.876612957729292e-05}
{"train_loss": 0.5458659529685974, "global_step": 19700, "epoch": 221, "lr": 9.876600157709712e-05}
{"train_loss": 0.49949535727500916, "global_step": 19701, "epoch": 221, "lr": 9.876587357034534e-05}
{"train_loss": 0.42517080903053284, "global_step": 19702, "epoch": 221, "lr": 9.876574555703758e-05}
{"train_loss": 0.4493745267391205, "global_step": 19703, "epoch": 221, "lr": 9.876561753717386e-05}
{"train_loss": 0.4720302224159241, "global_step": 19704, "epoch": 221, "lr": 9.876548951075419e-05}
{"train_loss": 0.4020838737487793, "global_step": 19705, "epoch": 221, "lr": 9.876536147777861e-05}
{"train_loss": 0.45992612838745117, "global_step": 19706, "epoch": 221, "lr": 9.876523343824712e-05}
{"train_loss": 0.4067436754703522, "global_step": 19707, "epoch": 221, "lr": 9.876510539215973e-05}
{"train_loss": 0.5830870866775513, "global_step": 19708, "epoch": 221, "lr": 9.876497733951646e-05}
{"train_loss": 0.4532422721385956, "global_step": 19709, "epoch": 221, "lr": 9.876484928031736e-05}
{"train_loss": 0.3982655704021454, "global_step": 19710, "epoch": 221, "lr": 9.87647212145624e-05}
{"train_loss": 0.5371093153953552, "global_step": 19711, "epoch": 221, "lr": 9.876459314225161e-05}
{"train_loss": 0.36476612091064453, "global_step": 19712, "epoch": 221, "lr": 9.876446506338502e-05}
{"train_loss": 0.5340380668640137, "global_step": 19713, "epoch": 221, "lr": 9.876433697796266e-05}
{"train_loss": 0.5900655388832092, "global_step": 19714, "epoch": 221, "lr": 9.876420888598451e-05}
{"train_loss": 0.2676776647567749, "global_step": 19715, "epoch": 221, "lr": 9.876408078745061e-05}
{"train_loss": 0.5363948941230774, "global_step": 19716, "epoch": 221, "lr": 9.876395268236098e-05}
{"train_loss": 0.46570083498954773, "global_step": 19717, "epoch": 221, "lr": 9.876382457071562e-05}
{"train_loss": 0.4329836070537567, "global_step": 19718, "epoch": 221, "lr": 9.876369645251457e-05}
{"train_loss": 0.5368896722793579, "global_step": 19719, "epoch": 221, "lr": 9.876356832775784e-05}
{"train_loss": 0.5807352066040039, "global_step": 19720, "epoch": 221, "lr": 9.876344019644542e-05}
{"train_loss": 0.47355976700782776, "global_step": 19721, "epoch": 221, "lr": 9.876331205857736e-05}
{"train_loss": 0.445028692483902, "global_step": 19722, "epoch": 221, "lr": 9.876318391415368e-05}
{"train_loss": 0.5617959499359131, "global_step": 19723, "epoch": 221, "lr": 9.876305576317437e-05}
{"train_loss": 0.4433978796005249, "global_step": 19724, "epoch": 221, "lr": 9.876292760563946e-05}
{"train_loss": 0.5187235474586487, "global_step": 19725, "epoch": 221, "lr": 9.876279944154898e-05}
{"train_loss": 0.3828009366989136, "global_step": 19726, "epoch": 221, "lr": 9.876267127090293e-05}
{"train_loss": 0.5998623967170715, "global_step": 19727, "epoch": 221, "lr": 9.876254309370133e-05}
{"train_loss": 0.6109833121299744, "global_step": 19728, "epoch": 221, "lr": 9.87624149099442e-05}
{"train_loss": 0.5523045659065247, "global_step": 19729, "epoch": 221, "lr": 9.876228671963157e-05}
{"train_loss": 0.5115666389465332, "global_step": 19730, "epoch": 221, "lr": 9.876215852276344e-05}
{"train_loss": 0.37456032633781433, "global_step": 19731, "epoch": 221, "lr": 9.876203031933983e-05}
{"train_loss": 0.6751375198364258, "global_step": 19732, "epoch": 221, "lr": 9.876190210936076e-05}
{"train_loss": 0.5568135976791382, "global_step": 19733, "epoch": 221, "lr": 9.876177389282623e-05}
{"train_loss": 0.5044461488723755, "global_step": 19734, "epoch": 221, "lr": 9.876164566973629e-05}
{"train_loss": 0.6017800569534302, "global_step": 19735, "epoch": 221, "lr": 9.876151744009095e-05}
{"train_loss": 0.5760905146598816, "global_step": 19736, "epoch": 221, "lr": 9.87613892038902e-05}
{"train_loss": 0.3528406322002411, "global_step": 19737, "epoch": 221, "lr": 9.876126096113409e-05}
{"train_loss": 0.38284388184547424, "global_step": 19738, "epoch": 221, "lr": 9.876113271182261e-05}
{"train_loss": 0.5010525584220886, "global_step": 19739, "epoch": 221, "lr": 9.87610044559558e-05}
{"train_loss": 0.4995449483394623, "global_step": 19740, "epoch": 221, "lr": 9.876087619353366e-05}
{"train_loss": 0.5036671757698059, "global_step": 19741, "epoch": 221, "lr": 9.876074792455621e-05}
{"train_loss": 0.5636005997657776, "global_step": 19742, "epoch": 221, "lr": 9.87606196490235e-05}
{"train_loss": 0.5140154957771301, "global_step": 19743, "epoch": 221, "lr": 9.876049136693549e-05}
{"train_loss": 0.4296281039714813, "global_step": 19744, "epoch": 221, "lr": 9.876036307829223e-05}
{"train_loss": 0.5506492853164673, "global_step": 19745, "epoch": 221, "lr": 9.876023478309375e-05}
{"train_loss": 0.46858981251716614, "global_step": 19746, "epoch": 221, "lr": 9.876010648134002e-05}
{"train_loss": 0.5637481212615967, "global_step": 19747, "epoch": 221, "lr": 9.875997817303111e-05}
{"train_loss": 0.42771318554878235, "global_step": 19748, "epoch": 221, "lr": 9.875984985816702e-05}
{"train_loss": 0.5345640182495117, "global_step": 19749, "epoch": 221, "lr": 9.875972153674776e-05}
{"train_loss": 0.47469961643218994, "global_step": 19750, "epoch": 221, "lr": 9.875959320877334e-05}
{"train_loss": 0.5561995506286621, "global_step": 19751, "epoch": 221, "lr": 9.875946487424379e-05}
{"train_loss": 0.43553557991981506, "global_step": 19752, "epoch": 221, "lr": 9.875933653315912e-05}
{"train_loss": 0.5814081430435181, "global_step": 19753, "epoch": 221, "lr": 9.875920818551935e-05}
{"train_loss": 0.486670583486557, "global_step": 19754, "epoch": 221, "lr": 9.87590798313245e-05}
{"train_loss": 0.5424275398254395, "global_step": 19755, "epoch": 221, "lr": 9.875895147057462e-05}
{"train_loss": 0.4955809712409973, "global_step": 19756, "epoch": 221, "lr": 9.875882310326964e-05}
{"train_loss": 0.48867269647255374, "global_step": 19757, "epoch": 221, "lr": 9.875869472940967e-05, "val_loss": 1.7098332643508911}
{"train_loss": 0.511265218257904, "global_step": 19758, "epoch": 222, "lr": 9.875856634899467e-05}
{"train_loss": 0.49200138449668884, "global_step": 19759, "epoch": 222, "lr": 9.875843796202468e-05}
{"train_loss": 0.5880283117294312, "global_step": 19760, "epoch": 222, "lr": 9.875830956849972e-05}
{"train_loss": 0.4187070429325104, "global_step": 19761, "epoch": 222, "lr": 9.875818116841979e-05}
{"train_loss": 0.5240421295166016, "global_step": 19762, "epoch": 222, "lr": 9.875805276178491e-05}
{"train_loss": 0.6062715649604797, "global_step": 19763, "epoch": 222, "lr": 9.87579243485951e-05}
{"train_loss": 0.5323588848114014, "global_step": 19764, "epoch": 222, "lr": 9.87577959288504e-05}
{"train_loss": 0.4890361428260803, "global_step": 19765, "epoch": 222, "lr": 9.87576675025508e-05}
{"train_loss": 0.5304439067840576, "global_step": 19766, "epoch": 222, "lr": 9.875753906969634e-05}
{"train_loss": 0.4718327820301056, "global_step": 19767, "epoch": 222, "lr": 9.8757410630287e-05}
{"train_loss": 0.4100726842880249, "global_step": 19768, "epoch": 222, "lr": 9.875728218432283e-05}
{"train_loss": 0.5201716423034668, "global_step": 19769, "epoch": 222, "lr": 9.875715373180385e-05}
{"train_loss": 0.3720003068447113, "global_step": 19770, "epoch": 222, "lr": 9.875702527273004e-05}
{"train_loss": 0.5380493998527527, "global_step": 19771, "epoch": 222, "lr": 9.875689680710145e-05}
{"train_loss": 0.4494950771331787, "global_step": 19772, "epoch": 222, "lr": 9.875676833491811e-05}
{"train_loss": 0.49092748761177063, "global_step": 19773, "epoch": 222, "lr": 9.875663985618e-05}
{"train_loss": 0.6241605281829834, "global_step": 19774, "epoch": 222, "lr": 9.875651137088715e-05}
{"train_loss": 0.5471493005752563, "global_step": 19775, "epoch": 222, "lr": 9.875638287903959e-05}
{"train_loss": 0.5202223062515259, "global_step": 19776, "epoch": 222, "lr": 9.875625438063732e-05}
{"train_loss": 0.4818810820579529, "global_step": 19777, "epoch": 222, "lr": 9.875612587568038e-05}
{"train_loss": 0.418174684047699, "global_step": 19778, "epoch": 222, "lr": 9.875599736416876e-05}
{"train_loss": 0.4950198531150818, "global_step": 19779, "epoch": 222, "lr": 9.87558688461025e-05}
{"train_loss": 0.3334156274795532, "global_step": 19780, "epoch": 222, "lr": 9.875574032148161e-05}
{"train_loss": 0.3762055039405823, "global_step": 19781, "epoch": 222, "lr": 9.875561179030608e-05}
{"train_loss": 0.4395851194858551, "global_step": 19782, "epoch": 222, "lr": 9.875548325257598e-05}
{"train_loss": 0.4703027009963989, "global_step": 19783, "epoch": 222, "lr": 9.875535470829129e-05}
{"train_loss": 0.47285735607147217, "global_step": 19784, "epoch": 222, "lr": 9.875522615745203e-05}
{"train_loss": 0.525253176689148, "global_step": 19785, "epoch": 222, "lr": 9.875509760005823e-05}
{"train_loss": 0.4855934679508209, "global_step": 19786, "epoch": 222, "lr": 9.87549690361099e-05}
{"train_loss": 0.39427438378334045, "global_step": 19787, "epoch": 222, "lr": 9.875484046560706e-05}
{"train_loss": 0.5951743721961975, "global_step": 19788, "epoch": 222, "lr": 9.875471188854972e-05}
{"train_loss": 0.5534096956253052, "global_step": 19789, "epoch": 222, "lr": 9.875458330493792e-05}
{"train_loss": 0.5230322480201721, "global_step": 19790, "epoch": 222, "lr": 9.875445471477165e-05}
{"train_loss": 0.48373863101005554, "global_step": 19791, "epoch": 222, "lr": 9.875432611805094e-05}
{"train_loss": 0.5429577231407166, "global_step": 19792, "epoch": 222, "lr": 9.87541975147758e-05}
{"train_loss": 0.49421069025993347, "global_step": 19793, "epoch": 222, "lr": 9.875406890494625e-05}
{"train_loss": 0.4957423210144043, "global_step": 19794, "epoch": 222, "lr": 9.875394028856233e-05}
{"train_loss": 0.5257534980773926, "global_step": 19795, "epoch": 222, "lr": 9.8753811665624e-05}
{"train_loss": 0.43544042110443115, "global_step": 19796, "epoch": 222, "lr": 9.875368303613135e-05}
{"train_loss": 0.6403202414512634, "global_step": 19797, "epoch": 222, "lr": 9.875355440008435e-05}
{"train_loss": 0.6291334629058838, "global_step": 19798, "epoch": 222, "lr": 9.875342575748303e-05}
{"train_loss": 0.486179381608963, "global_step": 19799, "epoch": 222, "lr": 9.87532971083274e-05}
{"train_loss": 0.37696248292922974, "global_step": 19800, "epoch": 222, "lr": 9.875316845261748e-05}
{"train_loss": 0.41911959648132324, "global_step": 19801, "epoch": 222, "lr": 9.875303979035331e-05}
{"train_loss": 0.4278266429901123, "global_step": 19802, "epoch": 222, "lr": 9.875291112153486e-05}
{"train_loss": 0.557339072227478, "global_step": 19803, "epoch": 222, "lr": 9.875278244616219e-05}
{"train_loss": 0.4552774429321289, "global_step": 19804, "epoch": 222, "lr": 9.87526537642353e-05}
{"train_loss": 0.43354320526123047, "global_step": 19805, "epoch": 222, "lr": 9.875252507575422e-05}
{"train_loss": 0.5570238828659058, "global_step": 19806, "epoch": 222, "lr": 9.875239638071894e-05}
{"train_loss": 0.4991047978401184, "global_step": 19807, "epoch": 222, "lr": 9.87522676791295e-05}
{"train_loss": 0.46478110551834106, "global_step": 19808, "epoch": 222, "lr": 9.875213897098593e-05}
{"train_loss": 0.3904917240142822, "global_step": 19809, "epoch": 222, "lr": 9.875201025628822e-05}
{"train_loss": 0.6120216846466064, "global_step": 19810, "epoch": 222, "lr": 9.875188153503638e-05}
{"train_loss": 0.5868392586708069, "global_step": 19811, "epoch": 222, "lr": 9.875175280723046e-05}
{"train_loss": 0.434231162071228, "global_step": 19812, "epoch": 222, "lr": 9.875162407287047e-05}
{"train_loss": 0.5842318534851074, "global_step": 19813, "epoch": 222, "lr": 9.87514953319564e-05}
{"train_loss": 0.4573105275630951, "global_step": 19814, "epoch": 222, "lr": 9.875136658448829e-05}
{"train_loss": 0.5514513850212097, "global_step": 19815, "epoch": 222, "lr": 9.875123783046615e-05}
{"train_loss": 0.40188470482826233, "global_step": 19816, "epoch": 222, "lr": 9.875110906989e-05}
{"train_loss": 0.5642615556716919, "global_step": 19817, "epoch": 222, "lr": 9.875098030275987e-05}
{"train_loss": 0.5794126987457275, "global_step": 19818, "epoch": 222, "lr": 9.875085152907576e-05}
{"train_loss": 0.30671170353889465, "global_step": 19819, "epoch": 222, "lr": 9.875072274883768e-05}
{"train_loss": 0.45128336548805237, "global_step": 19820, "epoch": 222, "lr": 9.875059396204567e-05}
{"train_loss": 0.4698619246482849, "global_step": 19821, "epoch": 222, "lr": 9.875046516869973e-05}
{"train_loss": 0.4726202189922333, "global_step": 19822, "epoch": 222, "lr": 9.87503363687999e-05}
{"train_loss": 0.5143350958824158, "global_step": 19823, "epoch": 222, "lr": 9.875020756234616e-05}
{"train_loss": 0.45859795808792114, "global_step": 19824, "epoch": 222, "lr": 9.875007874933855e-05}
{"train_loss": 0.6820100545883179, "global_step": 19825, "epoch": 222, "lr": 9.87499499297771e-05}
{"train_loss": 0.5682215094566345, "global_step": 19826, "epoch": 222, "lr": 9.874982110366182e-05}
{"train_loss": 0.444378137588501, "global_step": 19827, "epoch": 222, "lr": 9.87496922709927e-05}
{"train_loss": 0.4664633572101593, "global_step": 19828, "epoch": 222, "lr": 9.874956343176979e-05}
{"train_loss": 0.37533190846443176, "global_step": 19829, "epoch": 222, "lr": 9.87494345859931e-05}
{"train_loss": 0.5714083313941956, "global_step": 19830, "epoch": 222, "lr": 9.874930573366263e-05}
{"train_loss": 0.3958507776260376, "global_step": 19831, "epoch": 222, "lr": 9.874917687477841e-05}
{"train_loss": 0.4743344485759735, "global_step": 19832, "epoch": 222, "lr": 9.874904800934047e-05}
{"train_loss": 0.537126362323761, "global_step": 19833, "epoch": 222, "lr": 9.87489191373488e-05}
{"train_loss": 0.46487823128700256, "global_step": 19834, "epoch": 222, "lr": 9.874879025880343e-05}
{"train_loss": 0.5088841915130615, "global_step": 19835, "epoch": 222, "lr": 9.87486613737044e-05}
{"train_loss": 0.5634355545043945, "global_step": 19836, "epoch": 222, "lr": 9.87485324820517e-05}
{"train_loss": 0.49019938707351685, "global_step": 19837, "epoch": 222, "lr": 9.874840358384534e-05}
{"train_loss": 0.4929467737674713, "global_step": 19838, "epoch": 222, "lr": 9.874827467908536e-05}
{"train_loss": 0.386339008808136, "global_step": 19839, "epoch": 222, "lr": 9.874814576777178e-05}
{"train_loss": 0.46469128131866455, "global_step": 19840, "epoch": 222, "lr": 9.874801684990459e-05}
{"train_loss": 0.606378972530365, "global_step": 19841, "epoch": 222, "lr": 9.874788792548383e-05}
{"train_loss": 0.38855087757110596, "global_step": 19842, "epoch": 222, "lr": 9.87477589945095e-05}
{"train_loss": 0.5097218751907349, "global_step": 19843, "epoch": 222, "lr": 9.874763005698164e-05}
{"train_loss": 0.5328987836837769, "global_step": 19844, "epoch": 222, "lr": 9.874750111290025e-05}
{"train_loss": 0.644234836101532, "global_step": 19845, "epoch": 222, "lr": 9.874737216226537e-05}
{"train_loss": 0.4956274792719423, "global_step": 19846, "epoch": 222, "lr": 9.874724320507698e-05, "val_loss": 1.742110252380371}
{"train_loss": 0.40601035952568054, "global_step": 19847, "epoch": 223, "lr": 9.874711424133511e-05}
{"train_loss": 0.4286349415779114, "global_step": 19848, "epoch": 223, "lr": 9.874698527103981e-05}
{"train_loss": 0.5173624157905579, "global_step": 19849, "epoch": 223, "lr": 9.874685629419106e-05}
{"train_loss": 0.42797529697418213, "global_step": 19850, "epoch": 223, "lr": 9.874672731078889e-05}
{"train_loss": 0.4423852264881134, "global_step": 19851, "epoch": 223, "lr": 9.87465983208333e-05}
{"train_loss": 0.454551637172699, "global_step": 19852, "epoch": 223, "lr": 9.874646932432434e-05}
{"train_loss": 0.5185878276824951, "global_step": 19853, "epoch": 223, "lr": 9.874634032126202e-05}
{"train_loss": 0.4187028408050537, "global_step": 19854, "epoch": 223, "lr": 9.874621131164634e-05}
{"train_loss": 0.46001094579696655, "global_step": 19855, "epoch": 223, "lr": 9.874608229547732e-05}
{"train_loss": 0.5070307850837708, "global_step": 19856, "epoch": 223, "lr": 9.8745953272755e-05}
{"train_loss": 0.5421208143234253, "global_step": 19857, "epoch": 223, "lr": 9.874582424347935e-05}
{"train_loss": 0.40165990591049194, "global_step": 19858, "epoch": 223, "lr": 9.874569520765044e-05}
{"train_loss": 0.4226413369178772, "global_step": 19859, "epoch": 223, "lr": 9.874556616526826e-05}
{"train_loss": 0.30968177318573, "global_step": 19860, "epoch": 223, "lr": 9.874543711633284e-05}
{"train_loss": 0.4426087439060211, "global_step": 19861, "epoch": 223, "lr": 9.874530806084419e-05}
{"train_loss": 0.48926183581352234, "global_step": 19862, "epoch": 223, "lr": 9.874517899880233e-05}
{"train_loss": 0.44206154346466064, "global_step": 19863, "epoch": 223, "lr": 9.874504993020726e-05}
{"train_loss": 0.5185302495956421, "global_step": 19864, "epoch": 223, "lr": 9.874492085505902e-05}
{"train_loss": 0.4782475233078003, "global_step": 19865, "epoch": 223, "lr": 9.874479177335763e-05}
{"train_loss": 0.32235288619995117, "global_step": 19866, "epoch": 223, "lr": 9.874466268510307e-05}
{"train_loss": 0.469760537147522, "global_step": 19867, "epoch": 223, "lr": 9.87445335902954e-05}
{"train_loss": 0.3549395203590393, "global_step": 19868, "epoch": 223, "lr": 9.874440448893463e-05}
{"train_loss": 0.5119878053665161, "global_step": 19869, "epoch": 223, "lr": 9.874427538102075e-05}
{"train_loss": 0.6094882488250732, "global_step": 19870, "epoch": 223, "lr": 9.874414626655382e-05}
{"train_loss": 0.5537765026092529, "global_step": 19871, "epoch": 223, "lr": 9.874401714553381e-05}
{"train_loss": 0.41840851306915283, "global_step": 19872, "epoch": 223, "lr": 9.874388801796077e-05}
{"train_loss": 0.5678132176399231, "global_step": 19873, "epoch": 223, "lr": 9.874375888383472e-05}
{"train_loss": 0.4386991858482361, "global_step": 19874, "epoch": 223, "lr": 9.874362974315565e-05}
{"train_loss": 0.46779176592826843, "global_step": 19875, "epoch": 223, "lr": 9.874350059592361e-05}
{"train_loss": 0.4667779803276062, "global_step": 19876, "epoch": 223, "lr": 9.874337144213859e-05}
{"train_loss": 0.5444782972335815, "global_step": 19877, "epoch": 223, "lr": 9.87432422818006e-05}
{"train_loss": 0.4545588493347168, "global_step": 19878, "epoch": 223, "lr": 9.87431131149097e-05}
{"train_loss": 0.5828166007995605, "global_step": 19879, "epoch": 223, "lr": 9.874298394146589e-05}
{"train_loss": 0.5219612717628479, "global_step": 19880, "epoch": 223, "lr": 9.874285476146916e-05}
{"train_loss": 0.47580486536026, "global_step": 19881, "epoch": 223, "lr": 9.874272557491955e-05}
{"train_loss": 0.3669213652610779, "global_step": 19882, "epoch": 223, "lr": 9.874259638181707e-05}
{"train_loss": 0.6143050193786621, "global_step": 19883, "epoch": 223, "lr": 9.874246718216176e-05}
{"train_loss": 0.4468693137168884, "global_step": 19884, "epoch": 223, "lr": 9.87423379759536e-05}
{"train_loss": 0.4662521481513977, "global_step": 19885, "epoch": 223, "lr": 9.874220876319263e-05}
{"train_loss": 0.4297597110271454, "global_step": 19886, "epoch": 223, "lr": 9.874207954387887e-05}
{"train_loss": 0.4473956823348999, "global_step": 19887, "epoch": 223, "lr": 9.874195031801233e-05}
{"train_loss": 0.5052487254142761, "global_step": 19888, "epoch": 223, "lr": 9.874182108559304e-05}
{"train_loss": 0.39993172883987427, "global_step": 19889, "epoch": 223, "lr": 9.874169184662099e-05}
{"train_loss": 0.5687602162361145, "global_step": 19890, "epoch": 223, "lr": 9.874156260109621e-05}
{"train_loss": 0.5754780173301697, "global_step": 19891, "epoch": 223, "lr": 9.874143334901873e-05}
{"train_loss": 0.46015310287475586, "global_step": 19892, "epoch": 223, "lr": 9.874130409038857e-05}
{"train_loss": 0.44601598381996155, "global_step": 19893, "epoch": 223, "lr": 9.874117482520571e-05}
{"train_loss": 0.41693177819252014, "global_step": 19894, "epoch": 223, "lr": 9.874104555347021e-05}
{"train_loss": 0.5156631469726562, "global_step": 19895, "epoch": 223, "lr": 9.874091627518208e-05}
{"train_loss": 0.3975261449813843, "global_step": 19896, "epoch": 223, "lr": 9.874078699034131e-05}
{"train_loss": 0.518869161605835, "global_step": 19897, "epoch": 223, "lr": 9.874065769894794e-05}
{"train_loss": 0.40742701292037964, "global_step": 19898, "epoch": 223, "lr": 9.874052840100198e-05}
{"train_loss": 0.44770127534866333, "global_step": 19899, "epoch": 223, "lr": 9.874039909650345e-05}
{"train_loss": 0.5679799318313599, "global_step": 19900, "epoch": 223, "lr": 9.874026978545238e-05}
{"train_loss": 0.6434934735298157, "global_step": 19901, "epoch": 223, "lr": 9.874014046784876e-05}
{"train_loss": 0.5232614278793335, "global_step": 19902, "epoch": 223, "lr": 9.874001114369264e-05}
{"train_loss": 0.5801048874855042, "global_step": 19903, "epoch": 223, "lr": 9.8739881812984e-05}
{"train_loss": 0.5713868141174316, "global_step": 19904, "epoch": 223, "lr": 9.873975247572289e-05}
{"train_loss": 0.44577986001968384, "global_step": 19905, "epoch": 223, "lr": 9.87396231319093e-05}
{"train_loss": 0.4480282962322235, "global_step": 19906, "epoch": 223, "lr": 9.873949378154327e-05}
{"train_loss": 0.46969321370124817, "global_step": 19907, "epoch": 223, "lr": 9.873936442462481e-05}
{"train_loss": 0.4162743389606476, "global_step": 19908, "epoch": 223, "lr": 9.873923506115395e-05}
{"train_loss": 0.41042494773864746, "global_step": 19909, "epoch": 223, "lr": 9.873910569113067e-05}
{"train_loss": 0.5162540674209595, "global_step": 19910, "epoch": 223, "lr": 9.873897631455501e-05}
{"train_loss": 0.6505374312400818, "global_step": 19911, "epoch": 223, "lr": 9.873884693142701e-05}
{"train_loss": 0.41416969895362854, "global_step": 19912, "epoch": 223, "lr": 9.873871754174666e-05}
{"train_loss": 0.754718005657196, "global_step": 19913, "epoch": 223, "lr": 9.873858814551398e-05}
{"train_loss": 0.550862193107605, "global_step": 19914, "epoch": 223, "lr": 9.873845874272899e-05}
{"train_loss": 0.5203540325164795, "global_step": 19915, "epoch": 223, "lr": 9.873832933339171e-05}
{"train_loss": 0.5801441669464111, "global_step": 19916, "epoch": 223, "lr": 9.873819991750215e-05}
{"train_loss": 0.5371077060699463, "global_step": 19917, "epoch": 223, "lr": 9.873807049506034e-05}
{"train_loss": 0.5719254612922668, "global_step": 19918, "epoch": 223, "lr": 9.873794106606629e-05}
{"train_loss": 0.49496448040008545, "global_step": 19919, "epoch": 223, "lr": 9.873781163052001e-05}
{"train_loss": 0.5862752199172974, "global_step": 19920, "epoch": 223, "lr": 9.873768218842153e-05}
{"train_loss": 0.4656381607055664, "global_step": 19921, "epoch": 223, "lr": 9.873755273977086e-05}
{"train_loss": 0.4345824122428894, "global_step": 19922, "epoch": 223, "lr": 9.873742328456803e-05}
{"train_loss": 0.545698881149292, "global_step": 19923, "epoch": 223, "lr": 9.873729382281304e-05}
{"train_loss": 0.40769338607788086, "global_step": 19924, "epoch": 223, "lr": 9.873716435450592e-05}
{"train_loss": 0.500231146812439, "global_step": 19925, "epoch": 223, "lr": 9.873703487964667e-05}
{"train_loss": 0.49645060300827026, "global_step": 19926, "epoch": 223, "lr": 9.873690539823533e-05}
{"train_loss": 0.5357835292816162, "global_step": 19927, "epoch": 223, "lr": 9.87367759102719e-05}
{"train_loss": 0.41131553053855896, "global_step": 19928, "epoch": 223, "lr": 9.873664641575642e-05}
{"train_loss": 0.5742594599723816, "global_step": 19929, "epoch": 223, "lr": 9.873651691468888e-05}
{"train_loss": 0.4545111358165741, "global_step": 19930, "epoch": 223, "lr": 9.873638740706931e-05}
{"train_loss": 0.4767545759677887, "global_step": 19931, "epoch": 223, "lr": 9.873625789289774e-05}
{"train_loss": 0.5191309452056885, "global_step": 19932, "epoch": 223, "lr": 9.873612837217415e-05}
{"train_loss": 0.4992521405220032, "global_step": 19933, "epoch": 223, "lr": 9.873599884489859e-05}
{"train_loss": 0.41427451372146606, "global_step": 19934, "epoch": 223, "lr": 9.873586931107108e-05}
{"train_loss": 0.4873519302084205, "global_step": 19935, "epoch": 223, "lr": 9.873573977069161e-05, "val_loss": 1.7151451110839844}
{"train_loss": 0.5162302255630493, "global_step": 19936, "epoch": 224, "lr": 9.873561022376023e-05}
{"train_loss": 0.5133700966835022, "global_step": 19937, "epoch": 224, "lr": 9.873548067027693e-05}
{"train_loss": 0.4416384994983673, "global_step": 19938, "epoch": 224, "lr": 9.873535111024175e-05}
{"train_loss": 0.5960650444030762, "global_step": 19939, "epoch": 224, "lr": 9.873522154365469e-05}
{"train_loss": 0.40779998898506165, "global_step": 19940, "epoch": 224, "lr": 9.873509197051575e-05}
{"train_loss": 0.48637720942497253, "global_step": 19941, "epoch": 224, "lr": 9.8734962390825e-05}
{"train_loss": 0.5310970544815063, "global_step": 19942, "epoch": 224, "lr": 9.873483280458242e-05}
{"train_loss": 0.40568894147872925, "global_step": 19943, "epoch": 224, "lr": 9.873470321178804e-05}
{"train_loss": 0.5230458378791809, "global_step": 19944, "epoch": 224, "lr": 9.873457361244187e-05}
{"train_loss": 0.4266497790813446, "global_step": 19945, "epoch": 224, "lr": 9.873444400654393e-05}
{"train_loss": 0.46724802255630493, "global_step": 19946, "epoch": 224, "lr": 9.873431439409423e-05}
{"train_loss": 0.34870395064353943, "global_step": 19947, "epoch": 224, "lr": 9.873418477509282e-05}
{"train_loss": 0.3956197500228882, "global_step": 19948, "epoch": 224, "lr": 9.873405514953967e-05}
{"train_loss": 0.6436182260513306, "global_step": 19949, "epoch": 224, "lr": 9.873392551743483e-05}
{"train_loss": 0.5502139329910278, "global_step": 19950, "epoch": 224, "lr": 9.87337958787783e-05}
{"train_loss": 0.3503255844116211, "global_step": 19951, "epoch": 224, "lr": 9.873366623357011e-05}
{"train_loss": 0.37497109174728394, "global_step": 19952, "epoch": 224, "lr": 9.873353658181028e-05}
{"train_loss": 0.5978243350982666, "global_step": 19953, "epoch": 224, "lr": 9.873340692349881e-05}
{"train_loss": 0.4640169143676758, "global_step": 19954, "epoch": 224, "lr": 9.873327725863572e-05}
{"train_loss": 0.5556476712226868, "global_step": 19955, "epoch": 224, "lr": 9.873314758722105e-05}
{"train_loss": 0.48538699746131897, "global_step": 19956, "epoch": 224, "lr": 9.87330179092548e-05}
{"train_loss": 0.5043453574180603, "global_step": 19957, "epoch": 224, "lr": 9.873288822473698e-05}
{"train_loss": 0.5502727031707764, "global_step": 19958, "epoch": 224, "lr": 9.873275853366763e-05}
{"train_loss": 0.5643609166145325, "global_step": 19959, "epoch": 224, "lr": 9.873262883604675e-05}
{"train_loss": 0.4803398847579956, "global_step": 19960, "epoch": 224, "lr": 9.873249913187436e-05}
{"train_loss": 0.5245814323425293, "global_step": 19961, "epoch": 224, "lr": 9.873236942115048e-05}
{"train_loss": 0.41490137577056885, "global_step": 19962, "epoch": 224, "lr": 9.873223970387513e-05}
{"train_loss": 0.4408375918865204, "global_step": 19963, "epoch": 224, "lr": 9.873210998004832e-05}
{"train_loss": 0.4933561682701111, "global_step": 19964, "epoch": 224, "lr": 9.873198024967006e-05}
{"train_loss": 0.5063861608505249, "global_step": 19965, "epoch": 224, "lr": 9.87318505127404e-05}
{"train_loss": 0.3602146804332733, "global_step": 19966, "epoch": 224, "lr": 9.873172076925933e-05}
{"train_loss": 0.5019658803939819, "global_step": 19967, "epoch": 224, "lr": 9.873159101922686e-05}
{"train_loss": 0.48509982228279114, "global_step": 19968, "epoch": 224, "lr": 9.873146126264304e-05}
{"train_loss": 0.6068251132965088, "global_step": 19969, "epoch": 224, "lr": 9.873133149950786e-05}
{"train_loss": 0.48210811614990234, "global_step": 19970, "epoch": 224, "lr": 9.873120172982135e-05}
{"train_loss": 0.5317097306251526, "global_step": 19971, "epoch": 224, "lr": 9.873107195358353e-05}
{"train_loss": 0.4908679127693176, "global_step": 19972, "epoch": 224, "lr": 9.873094217079441e-05}
{"train_loss": 0.48860472440719604, "global_step": 19973, "epoch": 224, "lr": 9.8730812381454e-05}
{"train_loss": 0.5263110399246216, "global_step": 19974, "epoch": 224, "lr": 9.873068258556233e-05}
{"train_loss": 0.5087798237800598, "global_step": 19975, "epoch": 224, "lr": 9.873055278311941e-05}
{"train_loss": 0.4863625466823578, "global_step": 19976, "epoch": 224, "lr": 9.873042297412527e-05}
{"train_loss": 0.4248420298099518, "global_step": 19977, "epoch": 224, "lr": 9.873029315857992e-05}
{"train_loss": 0.4596918225288391, "global_step": 19978, "epoch": 224, "lr": 9.873016333648335e-05}
{"train_loss": 0.503075361251831, "global_step": 19979, "epoch": 224, "lr": 9.873003350783562e-05}
{"train_loss": 0.45172756910324097, "global_step": 19980, "epoch": 224, "lr": 9.872990367263674e-05}
{"train_loss": 0.39427295327186584, "global_step": 19981, "epoch": 224, "lr": 9.872977383088672e-05}
{"train_loss": 0.4275071620941162, "global_step": 19982, "epoch": 224, "lr": 9.872964398258556e-05}
{"train_loss": 0.4415746033191681, "global_step": 19983, "epoch": 224, "lr": 9.87295141277333e-05}
{"train_loss": 0.6032674312591553, "global_step": 19984, "epoch": 224, "lr": 9.872938426632996e-05}
{"train_loss": 0.6288576722145081, "global_step": 19985, "epoch": 224, "lr": 9.872925439837553e-05}
{"train_loss": 0.4215973913669586, "global_step": 19986, "epoch": 224, "lr": 9.872912452387007e-05}
{"train_loss": 0.3738948404788971, "global_step": 19987, "epoch": 224, "lr": 9.872899464281355e-05}
{"train_loss": 0.593490719795227, "global_step": 19988, "epoch": 224, "lr": 9.872886475520602e-05}
{"train_loss": 0.4307464063167572, "global_step": 19989, "epoch": 224, "lr": 9.87287348610475e-05}
{"train_loss": 0.4273728132247925, "global_step": 19990, "epoch": 224, "lr": 9.872860496033798e-05}
{"train_loss": 0.5671178698539734, "global_step": 19991, "epoch": 224, "lr": 9.872847505307749e-05}
{"train_loss": 0.6223348379135132, "global_step": 19992, "epoch": 224, "lr": 9.872834513926605e-05}
{"train_loss": 0.4465324580669403, "global_step": 19993, "epoch": 224, "lr": 9.872821521890369e-05}
{"train_loss": 0.342349648475647, "global_step": 19994, "epoch": 224, "lr": 9.87280852919904e-05}
{"train_loss": 0.4604039788246155, "global_step": 19995, "epoch": 224, "lr": 9.872795535852624e-05}
{"train_loss": 0.42584583163261414, "global_step": 19996, "epoch": 224, "lr": 9.872782541851118e-05}
{"train_loss": 0.45998555421829224, "global_step": 19997, "epoch": 224, "lr": 9.872769547194524e-05}
{"train_loss": 0.3945331871509552, "global_step": 19998, "epoch": 224, "lr": 9.87275655188285e-05}
{"train_loss": 0.3989538848400116, "global_step": 19999, "epoch": 224, "lr": 9.87274355591609e-05}
{"train_loss": 0.41979458928108215, "global_step": 20000, "epoch": 224, "lr": 9.87273055929425e-05}
{"train_loss": 0.49122828245162964, "global_step": 20001, "epoch": 224, "lr": 9.87271756201733e-05}
{"train_loss": 0.48605602979660034, "global_step": 20002, "epoch": 224, "lr": 9.872704564085334e-05}
{"train_loss": 0.5168904662132263, "global_step": 20003, "epoch": 224, "lr": 9.872691565498261e-05}
{"train_loss": 0.41895434260368347, "global_step": 20004, "epoch": 224, "lr": 9.872678566256115e-05}
{"train_loss": 0.40233197808265686, "global_step": 20005, "epoch": 224, "lr": 9.872665566358896e-05}
{"train_loss": 0.41022467613220215, "global_step": 20006, "epoch": 224, "lr": 9.872652565806606e-05}
{"train_loss": 0.4803611636161804, "global_step": 20007, "epoch": 224, "lr": 9.872639564599249e-05}
{"train_loss": 0.4209953248500824, "global_step": 20008, "epoch": 224, "lr": 9.872626562736824e-05}
{"train_loss": 0.5526441335678101, "global_step": 20009, "epoch": 224, "lr": 9.872613560219332e-05}
{"train_loss": 0.4871763288974762, "global_step": 20010, "epoch": 224, "lr": 9.872600557046778e-05}
{"train_loss": 0.4835764765739441, "global_step": 20011, "epoch": 224, "lr": 9.872587553219164e-05}
{"train_loss": 0.6449037194252014, "global_step": 20012, "epoch": 224, "lr": 9.872574548736489e-05}
{"train_loss": 0.4948265552520752, "global_step": 20013, "epoch": 224, "lr": 9.872561543598755e-05}
{"train_loss": 0.4725097715854645, "global_step": 20014, "epoch": 224, "lr": 9.872548537805963e-05}
{"train_loss": 0.4134071171283722, "global_step": 20015, "epoch": 224, "lr": 9.872535531358118e-05}
{"train_loss": 0.41243892908096313, "global_step": 20016, "epoch": 224, "lr": 9.87252252425522e-05}
{"train_loss": 0.41691455245018005, "global_step": 20017, "epoch": 224, "lr": 9.872509516497272e-05}
{"train_loss": 0.6147474050521851, "global_step": 20018, "epoch": 224, "lr": 9.872496508084273e-05}
{"train_loss": 0.6690495610237122, "global_step": 20019, "epoch": 224, "lr": 9.872483499016226e-05}
{"train_loss": 0.4807947278022766, "global_step": 20020, "epoch": 224, "lr": 9.872470489293134e-05}
{"train_loss": 0.440555214881897, "global_step": 20021, "epoch": 224, "lr": 9.872457478914996e-05}
{"train_loss": 0.36955273151397705, "global_step": 20022, "epoch": 224, "lr": 9.872444467881817e-05}
{"train_loss": 0.5684674382209778, "global_step": 20023, "epoch": 224, "lr": 9.872431456193596e-05}
{"train_loss": 0.48085998584715167, "global_step": 20024, "epoch": 224, "lr": 9.872418443850338e-05, "val_loss": 1.817948341369629}
{"train_loss": 0.489524245262146, "global_step": 20025, "epoch": 225, "lr": 9.872405430852041e-05}
{"train_loss": 0.46496468782424927, "global_step": 20026, "epoch": 225, "lr": 9.872392417198708e-05}
{"train_loss": 0.4518010914325714, "global_step": 20027, "epoch": 225, "lr": 9.872379402890343e-05}
{"train_loss": 0.503831148147583, "global_step": 20028, "epoch": 225, "lr": 9.872366387926944e-05}
{"train_loss": 0.5435704588890076, "global_step": 20029, "epoch": 225, "lr": 9.872353372308516e-05}
{"train_loss": 0.5073056221008301, "global_step": 20030, "epoch": 225, "lr": 9.872340356035059e-05}
{"train_loss": 0.5200734734535217, "global_step": 20031, "epoch": 225, "lr": 9.872327339106576e-05}
{"train_loss": 0.5496748089790344, "global_step": 20032, "epoch": 225, "lr": 9.872314321523069e-05}
{"train_loss": 0.4693238139152527, "global_step": 20033, "epoch": 225, "lr": 9.872301303284538e-05}
{"train_loss": 0.5546835064888, "global_step": 20034, "epoch": 225, "lr": 9.872288284390983e-05}
{"train_loss": 0.6014317274093628, "global_step": 20035, "epoch": 225, "lr": 9.87227526484241e-05}
{"train_loss": 0.39844781160354614, "global_step": 20036, "epoch": 225, "lr": 9.87226224463882e-05}
{"train_loss": 0.5694397687911987, "global_step": 20037, "epoch": 225, "lr": 9.872249223780212e-05}
{"train_loss": 0.567549467086792, "global_step": 20038, "epoch": 225, "lr": 9.87223620226659e-05}
{"train_loss": 0.5091188549995422, "global_step": 20039, "epoch": 225, "lr": 9.872223180097954e-05}
{"train_loss": 0.5196570754051208, "global_step": 20040, "epoch": 225, "lr": 9.872210157274309e-05}
{"train_loss": 0.5256643295288086, "global_step": 20041, "epoch": 225, "lr": 9.872197133795654e-05}
{"train_loss": 0.5648475289344788, "global_step": 20042, "epoch": 225, "lr": 9.872184109661992e-05}
{"train_loss": 0.4289097189903259, "global_step": 20043, "epoch": 225, "lr": 9.872171084873323e-05}
{"train_loss": 0.4425431191921234, "global_step": 20044, "epoch": 225, "lr": 9.872158059429651e-05}
{"train_loss": 0.450362890958786, "global_step": 20045, "epoch": 225, "lr": 9.872145033330978e-05}
{"train_loss": 0.5868389010429382, "global_step": 20046, "epoch": 225, "lr": 9.872132006577302e-05}
{"train_loss": 0.3983725607395172, "global_step": 20047, "epoch": 225, "lr": 9.872118979168628e-05}
{"train_loss": 0.3879266083240509, "global_step": 20048, "epoch": 225, "lr": 9.872105951104958e-05}
{"train_loss": 0.45283377170562744, "global_step": 20049, "epoch": 225, "lr": 9.872092922386291e-05}
{"train_loss": 0.4573386609554291, "global_step": 20050, "epoch": 225, "lr": 9.872079893012631e-05}
{"train_loss": 0.4243144989013672, "global_step": 20051, "epoch": 225, "lr": 9.87206686298398e-05}
{"train_loss": 0.6264063715934753, "global_step": 20052, "epoch": 225, "lr": 9.872053832300339e-05}
{"train_loss": 0.5281924605369568, "global_step": 20053, "epoch": 225, "lr": 9.872040800961709e-05}
{"train_loss": 0.5996018648147583, "global_step": 20054, "epoch": 225, "lr": 9.872027768968093e-05}
{"train_loss": 0.46362337470054626, "global_step": 20055, "epoch": 225, "lr": 9.872014736319492e-05}
{"train_loss": 0.5924561619758606, "global_step": 20056, "epoch": 225, "lr": 9.872001703015908e-05}
{"train_loss": 0.5311591625213623, "global_step": 20057, "epoch": 225, "lr": 9.871988669057344e-05}
{"train_loss": 0.46292296051979065, "global_step": 20058, "epoch": 225, "lr": 9.871975634443798e-05}
{"train_loss": 0.405216246843338, "global_step": 20059, "epoch": 225, "lr": 9.871962599175277e-05}
{"train_loss": 0.5219371318817139, "global_step": 20060, "epoch": 225, "lr": 9.871949563251778e-05}
{"train_loss": 0.41402941942214966, "global_step": 20061, "epoch": 225, "lr": 9.871936526673305e-05}
{"train_loss": 0.4574577808380127, "global_step": 20062, "epoch": 225, "lr": 9.871923489439861e-05}
{"train_loss": 0.4676375985145569, "global_step": 20063, "epoch": 225, "lr": 9.871910451551445e-05}
{"train_loss": 0.4520597457885742, "global_step": 20064, "epoch": 225, "lr": 9.87189741300806e-05}
{"train_loss": 0.41118723154067993, "global_step": 20065, "epoch": 225, "lr": 9.871884373809708e-05}
{"train_loss": 0.3670780658721924, "global_step": 20066, "epoch": 225, "lr": 9.87187133395639e-05}
{"train_loss": 0.514214277267456, "global_step": 20067, "epoch": 225, "lr": 9.871858293448108e-05}
{"train_loss": 0.5123903751373291, "global_step": 20068, "epoch": 225, "lr": 9.871845252284866e-05}
{"train_loss": 0.6105401515960693, "global_step": 20069, "epoch": 225, "lr": 9.871832210466663e-05}
{"train_loss": 0.48168087005615234, "global_step": 20070, "epoch": 225, "lr": 9.871819167993501e-05}
{"train_loss": 0.5111387372016907, "global_step": 20071, "epoch": 225, "lr": 9.871806124865381e-05}
{"train_loss": 0.4901568293571472, "global_step": 20072, "epoch": 225, "lr": 9.87179308108231e-05}
{"train_loss": 0.47527462244033813, "global_step": 20073, "epoch": 225, "lr": 9.871780036644282e-05}
{"train_loss": 0.5669739842414856, "global_step": 20074, "epoch": 225, "lr": 9.871766991551303e-05}
{"train_loss": 0.568731963634491, "global_step": 20075, "epoch": 225, "lr": 9.871753945803376e-05}
{"train_loss": 0.5426156520843506, "global_step": 20076, "epoch": 225, "lr": 9.8717408994005e-05}
{"train_loss": 0.5957564115524292, "global_step": 20077, "epoch": 225, "lr": 9.871727852342679e-05}
{"train_loss": 0.47610390186309814, "global_step": 20078, "epoch": 225, "lr": 9.871714804629912e-05}
{"train_loss": 0.4242763817310333, "global_step": 20079, "epoch": 225, "lr": 9.871701756262202e-05}
{"train_loss": 0.475722074508667, "global_step": 20080, "epoch": 225, "lr": 9.871688707239553e-05}
{"train_loss": 0.3994704782962799, "global_step": 20081, "epoch": 225, "lr": 9.871675657561963e-05}
{"train_loss": 0.48944181203842163, "global_step": 20082, "epoch": 225, "lr": 9.871662607229437e-05}
{"train_loss": 0.3974256217479706, "global_step": 20083, "epoch": 225, "lr": 9.871649556241975e-05}
{"train_loss": 0.5155247449874878, "global_step": 20084, "epoch": 225, "lr": 9.871636504599579e-05}
{"train_loss": 0.3592435121536255, "global_step": 20085, "epoch": 225, "lr": 9.87162345230225e-05}
{"train_loss": 0.4696086049079895, "global_step": 20086, "epoch": 225, "lr": 9.871610399349991e-05}
{"train_loss": 0.34653568267822266, "global_step": 20087, "epoch": 225, "lr": 9.871597345742805e-05}
{"train_loss": 0.42391547560691833, "global_step": 20088, "epoch": 225, "lr": 9.87158429148069e-05}
{"train_loss": 0.46478471159935, "global_step": 20089, "epoch": 225, "lr": 9.87157123656365e-05}
{"train_loss": 0.5484103560447693, "global_step": 20090, "epoch": 225, "lr": 9.871558180991687e-05}
{"train_loss": 0.5158185362815857, "global_step": 20091, "epoch": 225, "lr": 9.871545124764803e-05}
{"train_loss": 0.266245037317276, "global_step": 20092, "epoch": 225, "lr": 9.871532067882999e-05}
{"train_loss": 0.613864004611969, "global_step": 20093, "epoch": 225, "lr": 9.871519010346278e-05}
{"train_loss": 0.4295453131198883, "global_step": 20094, "epoch": 225, "lr": 9.871505952154639e-05}
{"train_loss": 0.5305505394935608, "global_step": 20095, "epoch": 225, "lr": 9.871492893308085e-05}
{"train_loss": 0.6081308722496033, "global_step": 20096, "epoch": 225, "lr": 9.871479833806618e-05}
{"train_loss": 0.461002379655838, "global_step": 20097, "epoch": 225, "lr": 9.871466773650242e-05}
{"train_loss": 0.4853856563568115, "global_step": 20098, "epoch": 225, "lr": 9.871453712838955e-05}
{"train_loss": 0.5381097197532654, "global_step": 20099, "epoch": 225, "lr": 9.87144065137276e-05}
{"train_loss": 0.46399664878845215, "global_step": 20100, "epoch": 225, "lr": 9.87142758925166e-05}
{"train_loss": 0.5843484401702881, "global_step": 20101, "epoch": 225, "lr": 9.871414526475658e-05}
{"train_loss": 0.6160269975662231, "global_step": 20102, "epoch": 225, "lr": 9.87140146304475e-05}
{"train_loss": 0.4128549098968506, "global_step": 20103, "epoch": 225, "lr": 9.871388398958944e-05}
{"train_loss": 0.38801464438438416, "global_step": 20104, "epoch": 225, "lr": 9.871375334218238e-05}
{"train_loss": 0.4732803404331207, "global_step": 20105, "epoch": 225, "lr": 9.871362268822635e-05}
{"train_loss": 0.40230995416641235, "global_step": 20106, "epoch": 225, "lr": 9.871349202772138e-05}
{"train_loss": 0.6231855154037476, "global_step": 20107, "epoch": 225, "lr": 9.871336136066746e-05}
{"train_loss": 0.519530713558197, "global_step": 20108, "epoch": 225, "lr": 9.871323068706463e-05}
{"train_loss": 0.497489333152771, "global_step": 20109, "epoch": 225, "lr": 9.87131000069129e-05}
{"train_loss": 0.5369514226913452, "global_step": 20110, "epoch": 225, "lr": 9.871296932021228e-05}
{"train_loss": 0.4835064709186554, "global_step": 20111, "epoch": 225, "lr": 9.87128386269628e-05}
{"train_loss": 0.538968563079834, "global_step": 20112, "epoch": 225, "lr": 9.871270792716447e-05}
{"train_loss": 0.4932571816980169, "global_step": 20113, "epoch": 225, "lr": 9.87125772208173e-05, "val_loss": 1.832870364189148, "train_action_mse_error": 51.23461151123047}
{"train_loss": 0.4546695351600647, "global_step": 20114, "epoch": 226, "lr": 9.871244650792133e-05}
{"train_loss": 0.49561989307403564, "global_step": 20115, "epoch": 226, "lr": 9.871231578847657e-05}
{"train_loss": 0.6419820785522461, "global_step": 20116, "epoch": 226, "lr": 9.871218506248304e-05}
{"train_loss": 0.37142449617385864, "global_step": 20117, "epoch": 226, "lr": 9.871205432994073e-05}
{"train_loss": 0.5810583829879761, "global_step": 20118, "epoch": 226, "lr": 9.871192359084968e-05}
{"train_loss": 0.42911797761917114, "global_step": 20119, "epoch": 226, "lr": 9.871179284520993e-05}
{"train_loss": 0.5491024255752563, "global_step": 20120, "epoch": 226, "lr": 9.871166209302145e-05}
{"train_loss": 0.498215913772583, "global_step": 20121, "epoch": 226, "lr": 9.871153133428428e-05}
{"train_loss": 0.6685230731964111, "global_step": 20122, "epoch": 226, "lr": 9.871140056899845e-05}
{"train_loss": 0.5261282324790955, "global_step": 20123, "epoch": 226, "lr": 9.871126979716396e-05}
{"train_loss": 0.5195260047912598, "global_step": 20124, "epoch": 226, "lr": 9.871113901878083e-05}
{"train_loss": 0.48992079496383667, "global_step": 20125, "epoch": 226, "lr": 9.871100823384909e-05}
{"train_loss": 0.4996146261692047, "global_step": 20126, "epoch": 226, "lr": 9.871087744236873e-05}
{"train_loss": 0.3970808982849121, "global_step": 20127, "epoch": 226, "lr": 9.871074664433982e-05}
{"train_loss": 0.4676547050476074, "global_step": 20128, "epoch": 226, "lr": 9.87106158397623e-05}
{"train_loss": 0.4888211786746979, "global_step": 20129, "epoch": 226, "lr": 9.871048502863627e-05}
{"train_loss": 0.3850622773170471, "global_step": 20130, "epoch": 226, "lr": 9.871035421096169e-05}
{"train_loss": 0.42579835653305054, "global_step": 20131, "epoch": 226, "lr": 9.87102233867386e-05}
{"train_loss": 0.3760469853878021, "global_step": 20132, "epoch": 226, "lr": 9.871009255596701e-05}
{"train_loss": 0.4294326603412628, "global_step": 20133, "epoch": 226, "lr": 9.870996171864695e-05}
{"train_loss": 0.592937171459198, "global_step": 20134, "epoch": 226, "lr": 9.870983087477844e-05}
{"train_loss": 0.4393054246902466, "global_step": 20135, "epoch": 226, "lr": 9.870970002436147e-05}
{"train_loss": 0.44800427556037903, "global_step": 20136, "epoch": 226, "lr": 9.870956916739607e-05}
{"train_loss": 0.422475129365921, "global_step": 20137, "epoch": 226, "lr": 9.870943830388226e-05}
{"train_loss": 0.5146400332450867, "global_step": 20138, "epoch": 226, "lr": 9.870930743382008e-05}
{"train_loss": 0.3801741302013397, "global_step": 20139, "epoch": 226, "lr": 9.87091765572095e-05}
{"train_loss": 0.448000967502594, "global_step": 20140, "epoch": 226, "lr": 9.87090456740506e-05}
{"train_loss": 0.370223730802536, "global_step": 20141, "epoch": 226, "lr": 9.870891478434333e-05}
{"train_loss": 0.5035418272018433, "global_step": 20142, "epoch": 226, "lr": 9.870878388808775e-05}
{"train_loss": 0.47925108671188354, "global_step": 20143, "epoch": 226, "lr": 9.870865298528388e-05}
{"train_loss": 0.5217800140380859, "global_step": 20144, "epoch": 226, "lr": 9.870852207593171e-05}
{"train_loss": 0.5078274607658386, "global_step": 20145, "epoch": 226, "lr": 9.870839116003128e-05}
{"train_loss": 0.4936184287071228, "global_step": 20146, "epoch": 226, "lr": 9.87082602375826e-05}
{"train_loss": 0.5002930760383606, "global_step": 20147, "epoch": 226, "lr": 9.870812930858568e-05}
{"train_loss": 0.4573063850402832, "global_step": 20148, "epoch": 226, "lr": 9.870799837304057e-05}
{"train_loss": 0.5623923540115356, "global_step": 20149, "epoch": 226, "lr": 9.870786743094724e-05}
{"train_loss": 0.43401873111724854, "global_step": 20150, "epoch": 226, "lr": 9.870773648230574e-05}
{"train_loss": 0.3125656545162201, "global_step": 20151, "epoch": 226, "lr": 9.870760552711608e-05}
{"train_loss": 0.4025561213493347, "global_step": 20152, "epoch": 226, "lr": 9.870747456537826e-05}
{"train_loss": 0.4332484304904938, "global_step": 20153, "epoch": 226, "lr": 9.870734359709232e-05}
{"train_loss": 0.3821270763874054, "global_step": 20154, "epoch": 226, "lr": 9.870721262225829e-05}
{"train_loss": 0.437890887260437, "global_step": 20155, "epoch": 226, "lr": 9.870708164087616e-05}
{"train_loss": 0.5217097401618958, "global_step": 20156, "epoch": 226, "lr": 9.870695065294596e-05}
{"train_loss": 0.3902941644191742, "global_step": 20157, "epoch": 226, "lr": 9.870681965846769e-05}
{"train_loss": 0.4266199469566345, "global_step": 20158, "epoch": 226, "lr": 9.870668865744139e-05}
{"train_loss": 0.4549265205860138, "global_step": 20159, "epoch": 226, "lr": 9.870655764986707e-05}
{"train_loss": 0.47978872060775757, "global_step": 20160, "epoch": 226, "lr": 9.870642663574474e-05}
{"train_loss": 0.6574527025222778, "global_step": 20161, "epoch": 226, "lr": 9.870629561507442e-05}
{"train_loss": 0.4145268201828003, "global_step": 20162, "epoch": 226, "lr": 9.870616458785615e-05}
{"train_loss": 0.6469260454177856, "global_step": 20163, "epoch": 226, "lr": 9.870603355408991e-05}
{"train_loss": 0.5350078344345093, "global_step": 20164, "epoch": 226, "lr": 9.870590251377576e-05}
{"train_loss": 0.47735488414764404, "global_step": 20165, "epoch": 226, "lr": 9.870577146691369e-05}
{"train_loss": 0.48352450132369995, "global_step": 20166, "epoch": 226, "lr": 9.870564041350372e-05}
{"train_loss": 0.6539204716682434, "global_step": 20167, "epoch": 226, "lr": 9.870550935354586e-05}
{"train_loss": 0.5791460275650024, "global_step": 20168, "epoch": 226, "lr": 9.870537828704015e-05}
{"train_loss": 0.5392712354660034, "global_step": 20169, "epoch": 226, "lr": 9.870524721398658e-05}
{"train_loss": 0.5483472943305969, "global_step": 20170, "epoch": 226, "lr": 9.87051161343852e-05}
{"train_loss": 0.37862905859947205, "global_step": 20171, "epoch": 226, "lr": 9.8704985048236e-05}
{"train_loss": 0.47320666909217834, "global_step": 20172, "epoch": 226, "lr": 9.870485395553901e-05}
{"train_loss": 0.5005576610565186, "global_step": 20173, "epoch": 226, "lr": 9.870472285629427e-05}
{"train_loss": 0.5082008838653564, "global_step": 20174, "epoch": 226, "lr": 9.870459175050174e-05}
{"train_loss": 0.6622036099433899, "global_step": 20175, "epoch": 226, "lr": 9.870446063816148e-05}
{"train_loss": 0.5504989624023438, "global_step": 20176, "epoch": 226, "lr": 9.87043295192735e-05}
{"train_loss": 0.5361252427101135, "global_step": 20177, "epoch": 226, "lr": 9.870419839383783e-05}
{"train_loss": 0.542275071144104, "global_step": 20178, "epoch": 226, "lr": 9.870406726185446e-05}
{"train_loss": 0.4302482008934021, "global_step": 20179, "epoch": 226, "lr": 9.870393612332342e-05}
{"train_loss": 0.47654736042022705, "global_step": 20180, "epoch": 226, "lr": 9.870380497824474e-05}
{"train_loss": 0.49703550338745117, "global_step": 20181, "epoch": 226, "lr": 9.870367382661842e-05}
{"train_loss": 0.5093148350715637, "global_step": 20182, "epoch": 226, "lr": 9.870354266844448e-05}
{"train_loss": 0.5704474449157715, "global_step": 20183, "epoch": 226, "lr": 9.870341150372295e-05}
{"train_loss": 0.4245510995388031, "global_step": 20184, "epoch": 226, "lr": 9.870328033245384e-05}
{"train_loss": 0.4436993896961212, "global_step": 20185, "epoch": 226, "lr": 9.870314915463716e-05}
{"train_loss": 0.5691361427307129, "global_step": 20186, "epoch": 226, "lr": 9.870301797027295e-05}
{"train_loss": 0.4850419759750366, "global_step": 20187, "epoch": 226, "lr": 9.870288677936121e-05}
{"train_loss": 0.384230375289917, "global_step": 20188, "epoch": 226, "lr": 9.870275558190194e-05}
{"train_loss": 0.433057963848114, "global_step": 20189, "epoch": 226, "lr": 9.87026243778952e-05}
{"train_loss": 0.41996127367019653, "global_step": 20190, "epoch": 226, "lr": 9.870249316734098e-05}
{"train_loss": 0.49299222230911255, "global_step": 20191, "epoch": 226, "lr": 9.87023619502393e-05}
{"train_loss": 0.392619252204895, "global_step": 20192, "epoch": 226, "lr": 9.870223072659018e-05}
{"train_loss": 0.4697699546813965, "global_step": 20193, "epoch": 226, "lr": 9.870209949639365e-05}
{"train_loss": 0.494176983833313, "global_step": 20194, "epoch": 226, "lr": 9.870196825964971e-05}
{"train_loss": 0.6797299981117249, "global_step": 20195, "epoch": 226, "lr": 9.870183701635838e-05}
{"train_loss": 0.4970630407333374, "global_step": 20196, "epoch": 226, "lr": 9.870170576651968e-05}
{"train_loss": 0.44157344102859497, "global_step": 20197, "epoch": 226, "lr": 9.870157451013363e-05}
{"train_loss": 0.48554831743240356, "global_step": 20198, "epoch": 226, "lr": 9.870144324720026e-05}
{"train_loss": 0.5262380838394165, "global_step": 20199, "epoch": 226, "lr": 9.870131197771955e-05}
{"train_loss": 0.5374502539634705, "global_step": 20200, "epoch": 226, "lr": 9.870118070169157e-05}
{"train_loss": 0.37969592213630676, "global_step": 20201, "epoch": 226, "lr": 9.870104941911629e-05}
{"train_loss": 0.4870413441336557, "global_step": 20202, "epoch": 226, "lr": 9.870091812999376e-05, "val_loss": 1.781206488609314}
{"train_loss": 0.44958463311195374, "global_step": 20203, "epoch": 227, "lr": 9.870078683432397e-05}
{"train_loss": 0.5162051320075989, "global_step": 20204, "epoch": 227, "lr": 9.870065553210696e-05}
{"train_loss": 0.5130227208137512, "global_step": 20205, "epoch": 227, "lr": 9.870052422334275e-05}
{"train_loss": 0.4048923850059509, "global_step": 20206, "epoch": 227, "lr": 9.870039290803133e-05}
{"train_loss": 0.3978213667869568, "global_step": 20207, "epoch": 227, "lr": 9.870026158617275e-05}
{"train_loss": 0.6044125556945801, "global_step": 20208, "epoch": 227, "lr": 9.870013025776702e-05}
{"train_loss": 0.44766151905059814, "global_step": 20209, "epoch": 227, "lr": 9.869999892281413e-05}
{"train_loss": 0.48639944195747375, "global_step": 20210, "epoch": 227, "lr": 9.869986758131413e-05}
{"train_loss": 0.35992148518562317, "global_step": 20211, "epoch": 227, "lr": 9.869973623326703e-05}
{"train_loss": 0.4876294434070587, "global_step": 20212, "epoch": 227, "lr": 9.869960487867282e-05}
{"train_loss": 0.5326433777809143, "global_step": 20213, "epoch": 227, "lr": 9.869947351753157e-05}
{"train_loss": 0.4462987780570984, "global_step": 20214, "epoch": 227, "lr": 9.869934214984326e-05}
{"train_loss": 0.4189110994338989, "global_step": 20215, "epoch": 227, "lr": 9.869921077560792e-05}
{"train_loss": 0.5823596119880676, "global_step": 20216, "epoch": 227, "lr": 9.869907939482556e-05}
{"train_loss": 0.47800084948539734, "global_step": 20217, "epoch": 227, "lr": 9.86989480074962e-05}
{"train_loss": 0.7002483606338501, "global_step": 20218, "epoch": 227, "lr": 9.869881661361987e-05}
{"train_loss": 0.46881580352783203, "global_step": 20219, "epoch": 227, "lr": 9.869868521319658e-05}
{"train_loss": 0.5807132124900818, "global_step": 20220, "epoch": 227, "lr": 9.869855380622634e-05}
{"train_loss": 0.5132538676261902, "global_step": 20221, "epoch": 227, "lr": 9.869842239270917e-05}
{"train_loss": 0.5286388993263245, "global_step": 20222, "epoch": 227, "lr": 9.869829097264509e-05}
{"train_loss": 0.3901495039463043, "global_step": 20223, "epoch": 227, "lr": 9.869815954603412e-05}
{"train_loss": 0.5798202753067017, "global_step": 20224, "epoch": 227, "lr": 9.869802811287627e-05}
{"train_loss": 0.5727328062057495, "global_step": 20225, "epoch": 227, "lr": 9.869789667317158e-05}
{"train_loss": 0.56084144115448, "global_step": 20226, "epoch": 227, "lr": 9.869776522692003e-05}
{"train_loss": 0.5978137850761414, "global_step": 20227, "epoch": 227, "lr": 9.869763377412166e-05}
{"train_loss": 0.4285319149494171, "global_step": 20228, "epoch": 227, "lr": 9.86975023147765e-05}
{"train_loss": 0.3990342915058136, "global_step": 20229, "epoch": 227, "lr": 9.869737084888455e-05}
{"train_loss": 0.5140188932418823, "global_step": 20230, "epoch": 227, "lr": 9.869723937644584e-05}
{"train_loss": 0.5084564685821533, "global_step": 20231, "epoch": 227, "lr": 9.869710789746036e-05}
{"train_loss": 0.7511641383171082, "global_step": 20232, "epoch": 227, "lr": 9.869697641192816e-05}
{"train_loss": 0.571576714515686, "global_step": 20233, "epoch": 227, "lr": 9.869684491984925e-05}
{"train_loss": 0.4791911244392395, "global_step": 20234, "epoch": 227, "lr": 9.869671342122365e-05}
{"train_loss": 0.5124141573905945, "global_step": 20235, "epoch": 227, "lr": 9.869658191605136e-05}
{"train_loss": 0.37925100326538086, "global_step": 20236, "epoch": 227, "lr": 9.869645040433239e-05}
{"train_loss": 0.46627292037010193, "global_step": 20237, "epoch": 227, "lr": 9.86963188860668e-05}
{"train_loss": 0.34629374742507935, "global_step": 20238, "epoch": 227, "lr": 9.869618736125457e-05}
{"train_loss": 0.4031575620174408, "global_step": 20239, "epoch": 227, "lr": 9.869605582989574e-05}
{"train_loss": 0.4036043584346771, "global_step": 20240, "epoch": 227, "lr": 9.86959242919903e-05}
{"train_loss": 0.4678754508495331, "global_step": 20241, "epoch": 227, "lr": 9.869579274753831e-05}
{"train_loss": 0.47956961393356323, "global_step": 20242, "epoch": 227, "lr": 9.869566119653976e-05}
{"train_loss": 0.4457717537879944, "global_step": 20243, "epoch": 227, "lr": 9.869552963899467e-05}
{"train_loss": 0.4102269113063812, "global_step": 20244, "epoch": 227, "lr": 9.869539807490305e-05}
{"train_loss": 0.6220272183418274, "global_step": 20245, "epoch": 227, "lr": 9.869526650426493e-05}
{"train_loss": 0.4418870508670807, "global_step": 20246, "epoch": 227, "lr": 9.869513492708034e-05}
{"train_loss": 0.43469393253326416, "global_step": 20247, "epoch": 227, "lr": 9.869500334334927e-05}
{"train_loss": 0.5938551425933838, "global_step": 20248, "epoch": 227, "lr": 9.869487175307175e-05}
{"train_loss": 0.5732583999633789, "global_step": 20249, "epoch": 227, "lr": 9.86947401562478e-05}
{"train_loss": 0.4200580418109894, "global_step": 20250, "epoch": 227, "lr": 9.869460855287744e-05}
{"train_loss": 0.5131829380989075, "global_step": 20251, "epoch": 227, "lr": 9.869447694296068e-05}
{"train_loss": 0.5946054458618164, "global_step": 20252, "epoch": 227, "lr": 9.869434532649754e-05}
{"train_loss": 0.4699726104736328, "global_step": 20253, "epoch": 227, "lr": 9.869421370348804e-05}
{"train_loss": 0.48517993092536926, "global_step": 20254, "epoch": 227, "lr": 9.869408207393219e-05}
{"train_loss": 0.5012494325637817, "global_step": 20255, "epoch": 227, "lr": 9.869395043783003e-05}
{"train_loss": 0.4986684024333954, "global_step": 20256, "epoch": 227, "lr": 9.869381879518154e-05}
{"train_loss": 0.5082124471664429, "global_step": 20257, "epoch": 227, "lr": 9.869368714598677e-05}
{"train_loss": 0.5034216046333313, "global_step": 20258, "epoch": 227, "lr": 9.869355549024573e-05}
{"train_loss": 0.402831494808197, "global_step": 20259, "epoch": 227, "lr": 9.869342382795843e-05}
{"train_loss": 0.5935989022254944, "global_step": 20260, "epoch": 227, "lr": 9.869329215912491e-05}
{"train_loss": 0.4185040295124054, "global_step": 20261, "epoch": 227, "lr": 9.869316048374515e-05}
{"train_loss": 0.4688965976238251, "global_step": 20262, "epoch": 227, "lr": 9.869302880181919e-05}
{"train_loss": 0.4508536159992218, "global_step": 20263, "epoch": 227, "lr": 9.869289711334707e-05}
{"train_loss": 0.4887326955795288, "global_step": 20264, "epoch": 227, "lr": 9.869276541832874e-05}
{"train_loss": 0.5148365497589111, "global_step": 20265, "epoch": 227, "lr": 9.86926337167643e-05}
{"train_loss": 0.3838941752910614, "global_step": 20266, "epoch": 227, "lr": 9.86925020086537e-05}
{"train_loss": 0.5125459432601929, "global_step": 20267, "epoch": 227, "lr": 9.8692370293997e-05}
{"train_loss": 0.5498707890510559, "global_step": 20268, "epoch": 227, "lr": 9.869223857279422e-05}
{"train_loss": 0.41484561562538147, "global_step": 20269, "epoch": 227, "lr": 9.869210684504534e-05}
{"train_loss": 0.4744851887226105, "global_step": 20270, "epoch": 227, "lr": 9.869197511075041e-05}
{"train_loss": 0.5116358399391174, "global_step": 20271, "epoch": 227, "lr": 9.869184336990943e-05}
{"train_loss": 0.4178098738193512, "global_step": 20272, "epoch": 227, "lr": 9.869171162252242e-05}
{"train_loss": 0.5334237813949585, "global_step": 20273, "epoch": 227, "lr": 9.869157986858942e-05}
{"train_loss": 0.4540671706199646, "global_step": 20274, "epoch": 227, "lr": 9.869144810811041e-05}
{"train_loss": 0.4541562497615814, "global_step": 20275, "epoch": 227, "lr": 9.869131634108546e-05}
{"train_loss": 0.4119037985801697, "global_step": 20276, "epoch": 227, "lr": 9.869118456751453e-05}
{"train_loss": 0.5060186386108398, "global_step": 20277, "epoch": 227, "lr": 9.869105278739767e-05}
{"train_loss": 0.5018047094345093, "global_step": 20278, "epoch": 227, "lr": 9.86909210007349e-05}
{"train_loss": 0.5315251350402832, "global_step": 20279, "epoch": 227, "lr": 9.869078920752621e-05}
{"train_loss": 0.5644059181213379, "global_step": 20280, "epoch": 227, "lr": 9.869065740777166e-05}
{"train_loss": 0.5083187222480774, "global_step": 20281, "epoch": 227, "lr": 9.869052560147123e-05}
{"train_loss": 0.44375988841056824, "global_step": 20282, "epoch": 227, "lr": 9.869039378862495e-05}
{"train_loss": 0.48712772130966187, "global_step": 20283, "epoch": 227, "lr": 9.869026196923284e-05}
{"train_loss": 0.5597748160362244, "global_step": 20284, "epoch": 227, "lr": 9.869013014329492e-05}
{"train_loss": 0.5481719970703125, "global_step": 20285, "epoch": 227, "lr": 9.86899983108112e-05}
{"train_loss": 0.5351605415344238, "global_step": 20286, "epoch": 227, "lr": 9.868986647178172e-05}
{"train_loss": 0.49321848154067993, "global_step": 20287, "epoch": 227, "lr": 9.868973462620648e-05}
{"train_loss": 0.5290067195892334, "global_step": 20288, "epoch": 227, "lr": 9.868960277408548e-05}
{"train_loss": 0.45868992805480957, "global_step": 20289, "epoch": 227, "lr": 9.868947091541876e-05}
{"train_loss": 0.5152249336242676, "global_step": 20290, "epoch": 227, "lr": 9.868933905020636e-05}
{"train_loss": 0.4927133284927754, "global_step": 20291, "epoch": 227, "lr": 9.868920717844826e-05, "val_loss": 1.7597062587738037}
{"train_loss": 0.41901567578315735, "global_step": 20292, "epoch": 228, "lr": 9.868907530014446e-05}
{"train_loss": 0.4004802703857422, "global_step": 20293, "epoch": 228, "lr": 9.868894341529504e-05}
{"train_loss": 0.39729398488998413, "global_step": 20294, "epoch": 228, "lr": 9.868881152389997e-05}
{"train_loss": 0.4181166887283325, "global_step": 20295, "epoch": 228, "lr": 9.868867962595928e-05}
{"train_loss": 0.5192749500274658, "global_step": 20296, "epoch": 228, "lr": 9.8688547721473e-05}
{"train_loss": 0.4188576340675354, "global_step": 20297, "epoch": 228, "lr": 9.868841581044115e-05}
{"train_loss": 0.49549105763435364, "global_step": 20298, "epoch": 228, "lr": 9.868828389286371e-05}
{"train_loss": 0.4769861400127411, "global_step": 20299, "epoch": 228, "lr": 9.868815196874073e-05}
{"train_loss": 0.4626086354255676, "global_step": 20300, "epoch": 228, "lr": 9.868802003807222e-05}
{"train_loss": 0.5079526901245117, "global_step": 20301, "epoch": 228, "lr": 9.868788810085823e-05}
{"train_loss": 0.44327476620674133, "global_step": 20302, "epoch": 228, "lr": 9.868775615709871e-05}
{"train_loss": 0.458164781332016, "global_step": 20303, "epoch": 228, "lr": 9.868762420679374e-05}
{"train_loss": 0.5597923398017883, "global_step": 20304, "epoch": 228, "lr": 9.86874922499433e-05}
{"train_loss": 0.38499879837036133, "global_step": 20305, "epoch": 228, "lr": 9.868736028654742e-05}
{"train_loss": 0.46581971645355225, "global_step": 20306, "epoch": 228, "lr": 9.868722831660611e-05}
{"train_loss": 0.5881206393241882, "global_step": 20307, "epoch": 228, "lr": 9.868709634011941e-05}
{"train_loss": 0.442468523979187, "global_step": 20308, "epoch": 228, "lr": 9.86869643570873e-05}
{"train_loss": 0.42421844601631165, "global_step": 20309, "epoch": 228, "lr": 9.868683236750983e-05}
{"train_loss": 0.5112951993942261, "global_step": 20310, "epoch": 228, "lr": 9.868670037138704e-05}
{"train_loss": 0.5027275085449219, "global_step": 20311, "epoch": 228, "lr": 9.868656836871888e-05}
{"train_loss": 0.425497829914093, "global_step": 20312, "epoch": 228, "lr": 9.868643635950542e-05}
{"train_loss": 0.5190853476524353, "global_step": 20313, "epoch": 228, "lr": 9.868630434374667e-05}
{"train_loss": 0.3912736475467682, "global_step": 20314, "epoch": 228, "lr": 9.868617232144262e-05}
{"train_loss": 0.468570739030838, "global_step": 20315, "epoch": 228, "lr": 9.868604029259331e-05}
{"train_loss": 0.4672943949699402, "global_step": 20316, "epoch": 228, "lr": 9.868590825719876e-05}
{"train_loss": 0.4854569435119629, "global_step": 20317, "epoch": 228, "lr": 9.8685776215259e-05}
{"train_loss": 0.4247323274612427, "global_step": 20318, "epoch": 228, "lr": 9.868564416677402e-05}
{"train_loss": 0.32510408759117126, "global_step": 20319, "epoch": 228, "lr": 9.868551211174383e-05}
{"train_loss": 0.5608863234519958, "global_step": 20320, "epoch": 228, "lr": 9.868538005016848e-05}
{"train_loss": 0.7032299041748047, "global_step": 20321, "epoch": 228, "lr": 9.868524798204798e-05}
{"train_loss": 0.5067432522773743, "global_step": 20322, "epoch": 228, "lr": 9.868511590738235e-05}
{"train_loss": 0.6098276972770691, "global_step": 20323, "epoch": 228, "lr": 9.868498382617158e-05}
{"train_loss": 0.4824185073375702, "global_step": 20324, "epoch": 228, "lr": 9.868485173841571e-05}
{"train_loss": 0.3910180330276489, "global_step": 20325, "epoch": 228, "lr": 9.868471964411476e-05}
{"train_loss": 0.350513219833374, "global_step": 20326, "epoch": 228, "lr": 9.868458754326872e-05}
{"train_loss": 0.42107710242271423, "global_step": 20327, "epoch": 228, "lr": 9.868445543587767e-05}
{"train_loss": 0.5686856508255005, "global_step": 20328, "epoch": 228, "lr": 9.868432332194156e-05}
{"train_loss": 0.5054715871810913, "global_step": 20329, "epoch": 228, "lr": 9.868419120146045e-05}
{"train_loss": 0.4903150796890259, "global_step": 20330, "epoch": 228, "lr": 9.868405907443434e-05}
{"train_loss": 0.5842210650444031, "global_step": 20331, "epoch": 228, "lr": 9.868392694086326e-05}
{"train_loss": 0.5519580245018005, "global_step": 20332, "epoch": 228, "lr": 9.868379480074721e-05}
{"train_loss": 0.6461262106895447, "global_step": 20333, "epoch": 228, "lr": 9.868366265408622e-05}
{"train_loss": 0.5063866376876831, "global_step": 20334, "epoch": 228, "lr": 9.868353050088029e-05}
{"train_loss": 0.518006443977356, "global_step": 20335, "epoch": 228, "lr": 9.868339834112947e-05}
{"train_loss": 0.5803470611572266, "global_step": 20336, "epoch": 228, "lr": 9.868326617483375e-05}
{"train_loss": 0.4559995234012604, "global_step": 20337, "epoch": 228, "lr": 9.868313400199317e-05}
{"train_loss": 0.5527001023292542, "global_step": 20338, "epoch": 228, "lr": 9.868300182260773e-05}
{"train_loss": 0.4804947078227997, "global_step": 20339, "epoch": 228, "lr": 9.868286963667745e-05}
{"train_loss": 0.482738196849823, "global_step": 20340, "epoch": 228, "lr": 9.868273744420236e-05}
{"train_loss": 0.42831894755363464, "global_step": 20341, "epoch": 228, "lr": 9.868260524518246e-05}
{"train_loss": 0.4655253291130066, "global_step": 20342, "epoch": 228, "lr": 9.86824730396178e-05}
{"train_loss": 0.4472085237503052, "global_step": 20343, "epoch": 228, "lr": 9.868234082750835e-05}
{"train_loss": 0.4305035173892975, "global_step": 20344, "epoch": 228, "lr": 9.868220860885415e-05}
{"train_loss": 0.534649133682251, "global_step": 20345, "epoch": 228, "lr": 9.868207638365523e-05}
{"train_loss": 0.4299602806568146, "global_step": 20346, "epoch": 228, "lr": 9.868194415191161e-05}
{"train_loss": 0.66749107837677, "global_step": 20347, "epoch": 228, "lr": 9.868181191362327e-05}
{"train_loss": 0.4271298050880432, "global_step": 20348, "epoch": 228, "lr": 9.868167966879027e-05}
{"train_loss": 0.4740225374698639, "global_step": 20349, "epoch": 228, "lr": 9.868154741741263e-05}
{"train_loss": 0.48049068450927734, "global_step": 20350, "epoch": 228, "lr": 9.868141515949032e-05}
{"train_loss": 0.4660845398902893, "global_step": 20351, "epoch": 228, "lr": 9.868128289502339e-05}
{"train_loss": 0.6254188418388367, "global_step": 20352, "epoch": 228, "lr": 9.868115062401187e-05}
{"train_loss": 0.4697229266166687, "global_step": 20353, "epoch": 228, "lr": 9.868101834645574e-05}
{"train_loss": 0.6021992564201355, "global_step": 20354, "epoch": 228, "lr": 9.868088606235506e-05}
{"train_loss": 0.3851405382156372, "global_step": 20355, "epoch": 228, "lr": 9.868075377170983e-05}
{"train_loss": 0.4788356125354767, "global_step": 20356, "epoch": 228, "lr": 9.868062147452004e-05}
{"train_loss": 0.4565669596195221, "global_step": 20357, "epoch": 228, "lr": 9.868048917078576e-05}
{"train_loss": 0.5702366232872009, "global_step": 20358, "epoch": 228, "lr": 9.868035686050698e-05}
{"train_loss": 0.5121116638183594, "global_step": 20359, "epoch": 228, "lr": 9.86802245436837e-05}
{"train_loss": 0.5407958626747131, "global_step": 20360, "epoch": 228, "lr": 9.868009222031598e-05}
{"train_loss": 0.5287549495697021, "global_step": 20361, "epoch": 228, "lr": 9.86799598904038e-05}
{"train_loss": 0.4143381714820862, "global_step": 20362, "epoch": 228, "lr": 9.86798275539472e-05}
{"train_loss": 0.5647145509719849, "global_step": 20363, "epoch": 228, "lr": 9.867969521094618e-05}
{"train_loss": 0.45527341961860657, "global_step": 20364, "epoch": 228, "lr": 9.86795628614008e-05}
{"train_loss": 0.6155439019203186, "global_step": 20365, "epoch": 228, "lr": 9.867943050531101e-05}
{"train_loss": 0.552077054977417, "global_step": 20366, "epoch": 228, "lr": 9.867929814267687e-05}
{"train_loss": 0.5672891736030579, "global_step": 20367, "epoch": 228, "lr": 9.867916577349841e-05}
{"train_loss": 0.5769658088684082, "global_step": 20368, "epoch": 228, "lr": 9.86790333977756e-05}
{"train_loss": 0.5285000801086426, "global_step": 20369, "epoch": 228, "lr": 9.86789010155085e-05}
{"train_loss": 0.40095967054367065, "global_step": 20370, "epoch": 228, "lr": 9.867876862669713e-05}
{"train_loss": 0.6058114767074585, "global_step": 20371, "epoch": 228, "lr": 9.867863623134148e-05}
{"train_loss": 0.4568661153316498, "global_step": 20372, "epoch": 228, "lr": 9.867850382944159e-05}
{"train_loss": 0.551671028137207, "global_step": 20373, "epoch": 228, "lr": 9.867837142099745e-05}
{"train_loss": 0.5245651006698608, "global_step": 20374, "epoch": 228, "lr": 9.867823900600912e-05}
{"train_loss": 0.5052506327629089, "global_step": 20375, "epoch": 228, "lr": 9.867810658447658e-05}
{"train_loss": 0.4145396649837494, "global_step": 20376, "epoch": 228, "lr": 9.867797415639985e-05}
{"train_loss": 0.45525923371315, "global_step": 20377, "epoch": 228, "lr": 9.867784172177899e-05}
{"train_loss": 0.45556411147117615, "global_step": 20378, "epoch": 228, "lr": 9.867770928061397e-05}
{"train_loss": 0.4588668942451477, "global_step": 20379, "epoch": 228, "lr": 9.867757683290481e-05}
{"train_loss": 0.49239757891451374, "global_step": 20380, "epoch": 228, "lr": 9.867744437865157e-05, "val_loss": 1.8033924102783203}
{"train_loss": 0.4072745442390442, "global_step": 20381, "epoch": 229, "lr": 9.867731191785423e-05}
{"train_loss": 0.590644121170044, "global_step": 20382, "epoch": 229, "lr": 9.867717945051283e-05}
{"train_loss": 0.47406360507011414, "global_step": 20383, "epoch": 229, "lr": 9.867704697662736e-05}
{"train_loss": 0.7278048992156982, "global_step": 20384, "epoch": 229, "lr": 9.867691449619786e-05}
{"train_loss": 0.46311521530151367, "global_step": 20385, "epoch": 229, "lr": 9.867678200922435e-05}
{"train_loss": 0.4868050217628479, "global_step": 20386, "epoch": 229, "lr": 9.867664951570682e-05}
{"train_loss": 0.34853219985961914, "global_step": 20387, "epoch": 229, "lr": 9.867651701564532e-05}
{"train_loss": 0.45515623688697815, "global_step": 20388, "epoch": 229, "lr": 9.867638450903986e-05}
{"train_loss": 0.515290379524231, "global_step": 20389, "epoch": 229, "lr": 9.867625199589045e-05}
{"train_loss": 0.49689823389053345, "global_step": 20390, "epoch": 229, "lr": 9.86761194761971e-05}
{"train_loss": 0.5972838997840881, "global_step": 20391, "epoch": 229, "lr": 9.867598694995985e-05}
{"train_loss": 0.42950424551963806, "global_step": 20392, "epoch": 229, "lr": 9.86758544171787e-05}
{"train_loss": 0.42142295837402344, "global_step": 20393, "epoch": 229, "lr": 9.867572187785369e-05}
{"train_loss": 0.5275737047195435, "global_step": 20394, "epoch": 229, "lr": 9.867558933198482e-05}
{"train_loss": 0.4634643793106079, "global_step": 20395, "epoch": 229, "lr": 9.867545677957209e-05}
{"train_loss": 0.5057325959205627, "global_step": 20396, "epoch": 229, "lr": 9.867532422061556e-05}
{"train_loss": 0.57632976770401, "global_step": 20397, "epoch": 229, "lr": 9.867519165511521e-05}
{"train_loss": 0.4390597343444824, "global_step": 20398, "epoch": 229, "lr": 9.867505908307109e-05}
{"train_loss": 0.42433643341064453, "global_step": 20399, "epoch": 229, "lr": 9.867492650448318e-05}
{"train_loss": 0.5959935784339905, "global_step": 20400, "epoch": 229, "lr": 9.867479391935154e-05}
{"train_loss": 0.4325045347213745, "global_step": 20401, "epoch": 229, "lr": 9.867466132767616e-05}
{"train_loss": 0.4315336048603058, "global_step": 20402, "epoch": 229, "lr": 9.867452872945706e-05}
{"train_loss": 0.40942493081092834, "global_step": 20403, "epoch": 229, "lr": 9.867439612469427e-05}
{"train_loss": 0.5217415690422058, "global_step": 20404, "epoch": 229, "lr": 9.86742635133878e-05}
{"train_loss": 0.44635510444641113, "global_step": 20405, "epoch": 229, "lr": 9.867413089553767e-05}
{"train_loss": 0.5571894645690918, "global_step": 20406, "epoch": 229, "lr": 9.867399827114388e-05}
{"train_loss": 0.7237645983695984, "global_step": 20407, "epoch": 229, "lr": 9.867386564020648e-05}
{"train_loss": 0.4381677508354187, "global_step": 20408, "epoch": 229, "lr": 9.867373300272546e-05}
{"train_loss": 0.37672263383865356, "global_step": 20409, "epoch": 229, "lr": 9.867360035870087e-05}
{"train_loss": 0.6018570065498352, "global_step": 20410, "epoch": 229, "lr": 9.867346770813269e-05}
{"train_loss": 0.4574209153652191, "global_step": 20411, "epoch": 229, "lr": 9.867333505102097e-05}
{"train_loss": 0.5130252242088318, "global_step": 20412, "epoch": 229, "lr": 9.86732023873657e-05}
{"train_loss": 0.39717862010002136, "global_step": 20413, "epoch": 229, "lr": 9.867306971716692e-05}
{"train_loss": 0.4199872612953186, "global_step": 20414, "epoch": 229, "lr": 9.867293704042464e-05}
{"train_loss": 0.4833172559738159, "global_step": 20415, "epoch": 229, "lr": 9.867280435713886e-05}
{"train_loss": 0.4967345595359802, "global_step": 20416, "epoch": 229, "lr": 9.867267166730964e-05}
{"train_loss": 0.6329192519187927, "global_step": 20417, "epoch": 229, "lr": 9.867253897093695e-05}
{"train_loss": 0.5099329948425293, "global_step": 20418, "epoch": 229, "lr": 9.867240626802085e-05}
{"train_loss": 0.5131315588951111, "global_step": 20419, "epoch": 229, "lr": 9.867227355856132e-05}
{"train_loss": 0.5767934322357178, "global_step": 20420, "epoch": 229, "lr": 9.867214084255841e-05}
{"train_loss": 0.44064754247665405, "global_step": 20421, "epoch": 229, "lr": 9.867200812001212e-05}
{"train_loss": 0.6077392101287842, "global_step": 20422, "epoch": 229, "lr": 9.867187539092246e-05}
{"train_loss": 0.4758479595184326, "global_step": 20423, "epoch": 229, "lr": 9.867174265528947e-05}
{"train_loss": 0.3731106221675873, "global_step": 20424, "epoch": 229, "lr": 9.867160991311316e-05}
{"train_loss": 0.43214350938796997, "global_step": 20425, "epoch": 229, "lr": 9.867147716439355e-05}
{"train_loss": 0.4922834038734436, "global_step": 20426, "epoch": 229, "lr": 9.867134440913064e-05}
{"train_loss": 0.43922170996665955, "global_step": 20427, "epoch": 229, "lr": 9.867121164732447e-05}
{"train_loss": 0.6350740790367126, "global_step": 20428, "epoch": 229, "lr": 9.867107887897504e-05}
{"train_loss": 0.4805869162082672, "global_step": 20429, "epoch": 229, "lr": 9.867094610408238e-05}
{"train_loss": 0.4310246706008911, "global_step": 20430, "epoch": 229, "lr": 9.867081332264651e-05}
{"train_loss": 0.4723711311817169, "global_step": 20431, "epoch": 229, "lr": 9.867068053466744e-05}
{"train_loss": 0.5836920738220215, "global_step": 20432, "epoch": 229, "lr": 9.867054774014518e-05}
{"train_loss": 0.4245910942554474, "global_step": 20433, "epoch": 229, "lr": 9.867041493907978e-05}
{"train_loss": 0.47426164150238037, "global_step": 20434, "epoch": 229, "lr": 9.867028213147121e-05}
{"train_loss": 0.5993735194206238, "global_step": 20435, "epoch": 229, "lr": 9.867014931731952e-05}
{"train_loss": 0.6489893198013306, "global_step": 20436, "epoch": 229, "lr": 9.867001649662473e-05}
{"train_loss": 0.47895878553390503, "global_step": 20437, "epoch": 229, "lr": 9.866988366938685e-05}
{"train_loss": 0.4312751293182373, "global_step": 20438, "epoch": 229, "lr": 9.86697508356059e-05}
{"train_loss": 0.5105032324790955, "global_step": 20439, "epoch": 229, "lr": 9.866961799528189e-05}
{"train_loss": 0.47567614912986755, "global_step": 20440, "epoch": 229, "lr": 9.866948514841484e-05}
{"train_loss": 0.4593263864517212, "global_step": 20441, "epoch": 229, "lr": 9.866935229500477e-05}
{"train_loss": 0.5288017392158508, "global_step": 20442, "epoch": 229, "lr": 9.866921943505171e-05}
{"train_loss": 0.5862511992454529, "global_step": 20443, "epoch": 229, "lr": 9.866908656855566e-05}
{"train_loss": 0.5886138677597046, "global_step": 20444, "epoch": 229, "lr": 9.866895369551664e-05}
{"train_loss": 0.4204823970794678, "global_step": 20445, "epoch": 229, "lr": 9.866882081593468e-05}
{"train_loss": 0.5074509978294373, "global_step": 20446, "epoch": 229, "lr": 9.86686879298098e-05}
{"train_loss": 0.5114918947219849, "global_step": 20447, "epoch": 229, "lr": 9.866855503714199e-05}
{"train_loss": 0.4543432891368866, "global_step": 20448, "epoch": 229, "lr": 9.866842213793128e-05}
{"train_loss": 0.6731139421463013, "global_step": 20449, "epoch": 229, "lr": 9.86682892321777e-05}
{"train_loss": 0.4440099895000458, "global_step": 20450, "epoch": 229, "lr": 9.866815631988127e-05}
{"train_loss": 0.4320942163467407, "global_step": 20451, "epoch": 229, "lr": 9.8668023401042e-05}
{"train_loss": 0.4375276267528534, "global_step": 20452, "epoch": 229, "lr": 9.86678904756599e-05}
{"train_loss": 0.4297321140766144, "global_step": 20453, "epoch": 229, "lr": 9.866775754373501e-05}
{"train_loss": 0.5693213939666748, "global_step": 20454, "epoch": 229, "lr": 9.866762460526731e-05}
{"train_loss": 0.4048857092857361, "global_step": 20455, "epoch": 229, "lr": 9.866749166025686e-05}
{"train_loss": 0.4135180115699768, "global_step": 20456, "epoch": 229, "lr": 9.866735870870365e-05}
{"train_loss": 0.41185158491134644, "global_step": 20457, "epoch": 229, "lr": 9.866722575060771e-05}
{"train_loss": 0.4967838525772095, "global_step": 20458, "epoch": 229, "lr": 9.866709278596904e-05}
{"train_loss": 0.5894072651863098, "global_step": 20459, "epoch": 229, "lr": 9.866695981478769e-05}
{"train_loss": 0.4069828689098358, "global_step": 20460, "epoch": 229, "lr": 9.866682683706365e-05}
{"train_loss": 0.552583634853363, "global_step": 20461, "epoch": 229, "lr": 9.866669385279696e-05}
{"train_loss": 0.5602662563323975, "global_step": 20462, "epoch": 229, "lr": 9.866656086198762e-05}
{"train_loss": 0.4923236668109894, "global_step": 20463, "epoch": 229, "lr": 9.866642786463566e-05}
{"train_loss": 0.47072872519493103, "global_step": 20464, "epoch": 229, "lr": 9.866629486074108e-05}
{"train_loss": 0.4473157227039337, "global_step": 20465, "epoch": 229, "lr": 9.866616185030393e-05}
{"train_loss": 0.36143991351127625, "global_step": 20466, "epoch": 229, "lr": 9.866602883332419e-05}
{"train_loss": 0.4634697139263153, "global_step": 20467, "epoch": 229, "lr": 9.86658958098019e-05}
{"train_loss": 0.6122235059738159, "global_step": 20468, "epoch": 229, "lr": 9.866576277973707e-05}
{"train_loss": 0.4948806076237325, "global_step": 20469, "epoch": 229, "lr": 9.866562974312972e-05, "val_loss": 1.7834908962249756}
{"train_loss": 0.48602232336997986, "global_step": 20470, "epoch": 230, "lr": 9.866549669997987e-05}
{"train_loss": 0.6183291673660278, "global_step": 20471, "epoch": 230, "lr": 9.866536365028754e-05}
{"train_loss": 0.4874882102012634, "global_step": 20472, "epoch": 230, "lr": 9.866523059405275e-05}
{"train_loss": 0.5278005003929138, "global_step": 20473, "epoch": 230, "lr": 9.866509753127551e-05}
{"train_loss": 0.48742496967315674, "global_step": 20474, "epoch": 230, "lr": 9.866496446195583e-05}
{"train_loss": 0.48059603571891785, "global_step": 20475, "epoch": 230, "lr": 9.866483138609376e-05}
{"train_loss": 0.40081948041915894, "global_step": 20476, "epoch": 230, "lr": 9.866469830368929e-05}
{"train_loss": 0.47338616847991943, "global_step": 20477, "epoch": 230, "lr": 9.866456521474243e-05}
{"train_loss": 0.47953474521636963, "global_step": 20478, "epoch": 230, "lr": 9.866443211925322e-05}
{"train_loss": 0.40600189566612244, "global_step": 20479, "epoch": 230, "lr": 9.866429901722167e-05}
{"train_loss": 0.3908108174800873, "global_step": 20480, "epoch": 230, "lr": 9.86641659086478e-05}
{"train_loss": 0.5183945894241333, "global_step": 20481, "epoch": 230, "lr": 9.866403279353162e-05}
{"train_loss": 0.37936854362487793, "global_step": 20482, "epoch": 230, "lr": 9.866389967187317e-05}
{"train_loss": 0.3832365870475769, "global_step": 20483, "epoch": 230, "lr": 9.866376654367244e-05}
{"train_loss": 0.5297816395759583, "global_step": 20484, "epoch": 230, "lr": 9.866363340892946e-05}
{"train_loss": 0.497348427772522, "global_step": 20485, "epoch": 230, "lr": 9.866350026764426e-05}
{"train_loss": 0.4759584069252014, "global_step": 20486, "epoch": 230, "lr": 9.866336711981683e-05}
{"train_loss": 0.39478799700737, "global_step": 20487, "epoch": 230, "lr": 9.866323396544721e-05}
{"train_loss": 0.4854535162448883, "global_step": 20488, "epoch": 230, "lr": 9.86631008045354e-05}
{"train_loss": 0.39942467212677, "global_step": 20489, "epoch": 230, "lr": 9.866296763708145e-05}
{"train_loss": 0.49405738711357117, "global_step": 20490, "epoch": 230, "lr": 9.866283446308534e-05}
{"train_loss": 0.4897415339946747, "global_step": 20491, "epoch": 230, "lr": 9.866270128254711e-05}
{"train_loss": 0.4411153793334961, "global_step": 20492, "epoch": 230, "lr": 9.866256809546677e-05}
{"train_loss": 0.5740490555763245, "global_step": 20493, "epoch": 230, "lr": 9.866243490184435e-05}
{"train_loss": 0.40710002183914185, "global_step": 20494, "epoch": 230, "lr": 9.866230170167985e-05}
{"train_loss": 0.43520477414131165, "global_step": 20495, "epoch": 230, "lr": 9.866216849497329e-05}
{"train_loss": 0.49149471521377563, "global_step": 20496, "epoch": 230, "lr": 9.866203528172471e-05}
{"train_loss": 0.4808969795703888, "global_step": 20497, "epoch": 230, "lr": 9.866190206193411e-05}
{"train_loss": 0.43531471490859985, "global_step": 20498, "epoch": 230, "lr": 9.86617688356015e-05}
{"train_loss": 0.4353364109992981, "global_step": 20499, "epoch": 230, "lr": 9.866163560272693e-05}
{"train_loss": 0.39170360565185547, "global_step": 20500, "epoch": 230, "lr": 9.866150236331037e-05}
{"train_loss": 0.5597987174987793, "global_step": 20501, "epoch": 230, "lr": 9.866136911735189e-05}
{"train_loss": 0.3423200249671936, "global_step": 20502, "epoch": 230, "lr": 9.866123586485145e-05}
{"train_loss": 0.5725284218788147, "global_step": 20503, "epoch": 230, "lr": 9.866110260580911e-05}
{"train_loss": 0.3432299792766571, "global_step": 20504, "epoch": 230, "lr": 9.86609693402249e-05}
{"train_loss": 0.4589283764362335, "global_step": 20505, "epoch": 230, "lr": 9.86608360680988e-05}
{"train_loss": 0.36310234665870667, "global_step": 20506, "epoch": 230, "lr": 9.866070278943085e-05}
{"train_loss": 0.5072576999664307, "global_step": 20507, "epoch": 230, "lr": 9.866056950422105e-05}
{"train_loss": 0.6323789954185486, "global_step": 20508, "epoch": 230, "lr": 9.866043621246943e-05}
{"train_loss": 0.3908478617668152, "global_step": 20509, "epoch": 230, "lr": 9.866030291417602e-05}
{"train_loss": 0.4753117561340332, "global_step": 20510, "epoch": 230, "lr": 9.866016960934081e-05}
{"train_loss": 0.4650577902793884, "global_step": 20511, "epoch": 230, "lr": 9.866003629796386e-05}
{"train_loss": 0.4477122128009796, "global_step": 20512, "epoch": 230, "lr": 9.865990298004514e-05}
{"train_loss": 0.632615327835083, "global_step": 20513, "epoch": 230, "lr": 9.865976965558469e-05}
{"train_loss": 0.42222362756729126, "global_step": 20514, "epoch": 230, "lr": 9.865963632458252e-05}
{"train_loss": 0.503731369972229, "global_step": 20515, "epoch": 230, "lr": 9.865950298703868e-05}
{"train_loss": 0.4632643163204193, "global_step": 20516, "epoch": 230, "lr": 9.865936964295315e-05}
{"train_loss": 0.4886893928050995, "global_step": 20517, "epoch": 230, "lr": 9.865923629232595e-05}
{"train_loss": 0.3255242705345154, "global_step": 20518, "epoch": 230, "lr": 9.865910293515712e-05}
{"train_loss": 0.3872414529323578, "global_step": 20519, "epoch": 230, "lr": 9.865896957144665e-05}
{"train_loss": 0.5494673252105713, "global_step": 20520, "epoch": 230, "lr": 9.865883620119461e-05}
{"train_loss": 0.5230841636657715, "global_step": 20521, "epoch": 230, "lr": 9.865870282440095e-05}
{"train_loss": 0.32730841636657715, "global_step": 20522, "epoch": 230, "lr": 9.865856944106573e-05}
{"train_loss": 0.4435478448867798, "global_step": 20523, "epoch": 230, "lr": 9.865843605118896e-05}
{"train_loss": 0.4911409616470337, "global_step": 20524, "epoch": 230, "lr": 9.865830265477065e-05}
{"train_loss": 0.5103680491447449, "global_step": 20525, "epoch": 230, "lr": 9.865816925181084e-05}
{"train_loss": 0.45164960622787476, "global_step": 20526, "epoch": 230, "lr": 9.865803584230952e-05}
{"train_loss": 0.47613489627838135, "global_step": 20527, "epoch": 230, "lr": 9.865790242626672e-05}
{"train_loss": 0.42208555340766907, "global_step": 20528, "epoch": 230, "lr": 9.865776900368247e-05}
{"train_loss": 0.5504302382469177, "global_step": 20529, "epoch": 230, "lr": 9.865763557455676e-05}
{"train_loss": 0.5347416996955872, "global_step": 20530, "epoch": 230, "lr": 9.865750213888962e-05}
{"train_loss": 0.49943819642066956, "global_step": 20531, "epoch": 230, "lr": 9.86573686966811e-05}
{"train_loss": 0.47408196330070496, "global_step": 20532, "epoch": 230, "lr": 9.865723524793116e-05}
{"train_loss": 0.5291595458984375, "global_step": 20533, "epoch": 230, "lr": 9.865710179263987e-05}
{"train_loss": 0.44826579093933105, "global_step": 20534, "epoch": 230, "lr": 9.86569683308072e-05}
{"train_loss": 0.4531151056289673, "global_step": 20535, "epoch": 230, "lr": 9.865683486243323e-05}
{"train_loss": 0.588045060634613, "global_step": 20536, "epoch": 230, "lr": 9.86567013875179e-05}
{"train_loss": 0.48093828558921814, "global_step": 20537, "epoch": 230, "lr": 9.86565679060613e-05}
{"train_loss": 0.30627214908599854, "global_step": 20538, "epoch": 230, "lr": 9.865643441806341e-05}
{"train_loss": 0.44757673144340515, "global_step": 20539, "epoch": 230, "lr": 9.865630092352424e-05}
{"train_loss": 0.37756770849227905, "global_step": 20540, "epoch": 230, "lr": 9.865616742244385e-05}
{"train_loss": 0.6345704793930054, "global_step": 20541, "epoch": 230, "lr": 9.865603391482222e-05}
{"train_loss": 0.35188454389572144, "global_step": 20542, "epoch": 230, "lr": 9.865590040065937e-05}
{"train_loss": 0.48061805963516235, "global_step": 20543, "epoch": 230, "lr": 9.865576687995534e-05}
{"train_loss": 0.5037950873374939, "global_step": 20544, "epoch": 230, "lr": 9.865563335271013e-05}
{"train_loss": 0.5169405341148376, "global_step": 20545, "epoch": 230, "lr": 9.865549981892377e-05}
{"train_loss": 0.4670340418815613, "global_step": 20546, "epoch": 230, "lr": 9.865536627859626e-05}
{"train_loss": 0.511733889579773, "global_step": 20547, "epoch": 230, "lr": 9.865523273172763e-05}
{"train_loss": 0.4332999885082245, "global_step": 20548, "epoch": 230, "lr": 9.86550991783179e-05}
{"train_loss": 0.5172500610351562, "global_step": 20549, "epoch": 230, "lr": 9.86549656183671e-05}
{"train_loss": 0.5713980793952942, "global_step": 20550, "epoch": 230, "lr": 9.86548320518752e-05}
{"train_loss": 0.5362122058868408, "global_step": 20551, "epoch": 230, "lr": 9.865469847884228e-05}
{"train_loss": 0.34885725378990173, "global_step": 20552, "epoch": 230, "lr": 9.865456489926831e-05}
{"train_loss": 0.5360457301139832, "global_step": 20553, "epoch": 230, "lr": 9.865443131315333e-05}
{"train_loss": 0.3680828809738159, "global_step": 20554, "epoch": 230, "lr": 9.865429772049737e-05}
{"train_loss": 0.48831191658973694, "global_step": 20555, "epoch": 230, "lr": 9.865416412130042e-05}
{"train_loss": 0.44444340467453003, "global_step": 20556, "epoch": 230, "lr": 9.86540305155625e-05}
{"train_loss": 0.5008394122123718, "global_step": 20557, "epoch": 230, "lr": 9.865389690328366e-05}
{"train_loss": 0.46779406874367363, "global_step": 20558, "epoch": 230, "lr": 9.865376328446388e-05, "val_loss": 1.7538563013076782, "train_action_mse_error": 48.91645050048828}
{"train_loss": 0.49859800934791565, "global_step": 20559, "epoch": 231, "lr": 9.865362965910321e-05}
{"train_loss": 0.5788424611091614, "global_step": 20560, "epoch": 231, "lr": 9.865349602720165e-05}
{"train_loss": 0.5862295627593994, "global_step": 20561, "epoch": 231, "lr": 9.865336238875921e-05}
{"train_loss": 0.537953794002533, "global_step": 20562, "epoch": 231, "lr": 9.865322874377592e-05}
{"train_loss": 0.4317273497581482, "global_step": 20563, "epoch": 231, "lr": 9.86530950922518e-05}
{"train_loss": 0.47876444458961487, "global_step": 20564, "epoch": 231, "lr": 9.865296143418688e-05}
{"train_loss": 0.477976530790329, "global_step": 20565, "epoch": 231, "lr": 9.865282776958114e-05}
{"train_loss": 0.44908198714256287, "global_step": 20566, "epoch": 231, "lr": 9.865269409843463e-05}
{"train_loss": 0.4501528739929199, "global_step": 20567, "epoch": 231, "lr": 9.865256042074736e-05}
{"train_loss": 0.4713260531425476, "global_step": 20568, "epoch": 231, "lr": 9.865242673651933e-05}
{"train_loss": 0.591133713722229, "global_step": 20569, "epoch": 231, "lr": 9.865229304575059e-05}
{"train_loss": 0.4459240138530731, "global_step": 20570, "epoch": 231, "lr": 9.865215934844113e-05}
{"train_loss": 0.5264386534690857, "global_step": 20571, "epoch": 231, "lr": 9.8652025644591e-05}
{"train_loss": 0.4261462092399597, "global_step": 20572, "epoch": 231, "lr": 9.865189193420016e-05}
{"train_loss": 0.3824923038482666, "global_step": 20573, "epoch": 231, "lr": 9.86517582172687e-05}
{"train_loss": 0.46748656034469604, "global_step": 20574, "epoch": 231, "lr": 9.86516244937966e-05}
{"train_loss": 0.41800457239151, "global_step": 20575, "epoch": 231, "lr": 9.865149076378389e-05}
{"train_loss": 0.4909561276435852, "global_step": 20576, "epoch": 231, "lr": 9.865135702723056e-05}
{"train_loss": 0.49612414836883545, "global_step": 20577, "epoch": 231, "lr": 9.865122328413665e-05}
{"train_loss": 0.5298046469688416, "global_step": 20578, "epoch": 231, "lr": 9.86510895345022e-05}
{"train_loss": 0.5041099786758423, "global_step": 20579, "epoch": 231, "lr": 9.865095577832717e-05}
{"train_loss": 0.38603729009628296, "global_step": 20580, "epoch": 231, "lr": 9.865082201561164e-05}
{"train_loss": 0.5398699641227722, "global_step": 20581, "epoch": 231, "lr": 9.865068824635558e-05}
{"train_loss": 0.48828059434890747, "global_step": 20582, "epoch": 231, "lr": 9.865055447055904e-05}
{"train_loss": 0.4618011713027954, "global_step": 20583, "epoch": 231, "lr": 9.865042068822202e-05}
{"train_loss": 0.4379361867904663, "global_step": 20584, "epoch": 231, "lr": 9.865028689934454e-05}
{"train_loss": 0.3776519298553467, "global_step": 20585, "epoch": 231, "lr": 9.865015310392663e-05}
{"train_loss": 0.4308215081691742, "global_step": 20586, "epoch": 231, "lr": 9.865001930196831e-05}
{"train_loss": 0.4321461021900177, "global_step": 20587, "epoch": 231, "lr": 9.864988549346956e-05}
{"train_loss": 0.5470897555351257, "global_step": 20588, "epoch": 231, "lr": 9.864975167843045e-05}
{"train_loss": 0.38736408948898315, "global_step": 20589, "epoch": 231, "lr": 9.864961785685097e-05}
{"train_loss": 0.48373934626579285, "global_step": 20590, "epoch": 231, "lr": 9.864948402873113e-05}
{"train_loss": 0.4230009615421295, "global_step": 20591, "epoch": 231, "lr": 9.864935019407098e-05}
{"train_loss": 0.5445947051048279, "global_step": 20592, "epoch": 231, "lr": 9.86492163528705e-05}
{"train_loss": 0.3506046533584595, "global_step": 20593, "epoch": 231, "lr": 9.864908250512972e-05}
{"train_loss": 0.39878684282302856, "global_step": 20594, "epoch": 231, "lr": 9.864894865084868e-05}
{"train_loss": 0.40679019689559937, "global_step": 20595, "epoch": 231, "lr": 9.864881479002738e-05}
{"train_loss": 0.5414539575576782, "global_step": 20596, "epoch": 231, "lr": 9.864868092266584e-05}
{"train_loss": 0.567995011806488, "global_step": 20597, "epoch": 231, "lr": 9.864854704876407e-05}
{"train_loss": 0.3738429546356201, "global_step": 20598, "epoch": 231, "lr": 9.86484131683221e-05}
{"train_loss": 0.4442652761936188, "global_step": 20599, "epoch": 231, "lr": 9.864827928133994e-05}
{"train_loss": 0.4494774043560028, "global_step": 20600, "epoch": 231, "lr": 9.864814538781761e-05}
{"train_loss": 0.4861578941345215, "global_step": 20601, "epoch": 231, "lr": 9.864801148775515e-05}
{"train_loss": 0.5044817328453064, "global_step": 20602, "epoch": 231, "lr": 9.864787758115254e-05}
{"train_loss": 0.530444860458374, "global_step": 20603, "epoch": 231, "lr": 9.864774366800981e-05}
{"train_loss": 0.5682203769683838, "global_step": 20604, "epoch": 231, "lr": 9.864760974832699e-05}
{"train_loss": 0.8248893022537231, "global_step": 20605, "epoch": 231, "lr": 9.86474758221041e-05}
{"train_loss": 0.559295654296875, "global_step": 20606, "epoch": 231, "lr": 9.864734188934114e-05}
{"train_loss": 0.4922935962677002, "global_step": 20607, "epoch": 231, "lr": 9.864720795003815e-05}
{"train_loss": 0.49709975719451904, "global_step": 20608, "epoch": 231, "lr": 9.864707400419513e-05}
{"train_loss": 0.49307212233543396, "global_step": 20609, "epoch": 231, "lr": 9.86469400518121e-05}
{"train_loss": 0.5030961632728577, "global_step": 20610, "epoch": 231, "lr": 9.864680609288907e-05}
{"train_loss": 0.5022367238998413, "global_step": 20611, "epoch": 231, "lr": 9.86466721274261e-05}
{"train_loss": 0.38319408893585205, "global_step": 20612, "epoch": 231, "lr": 9.864653815542316e-05}
{"train_loss": 0.47834527492523193, "global_step": 20613, "epoch": 231, "lr": 9.864640417688028e-05}
{"train_loss": 0.43874236941337585, "global_step": 20614, "epoch": 231, "lr": 9.864627019179749e-05}
{"train_loss": 0.4647913873195648, "global_step": 20615, "epoch": 231, "lr": 9.86461362001748e-05}
{"train_loss": 0.512671172618866, "global_step": 20616, "epoch": 231, "lr": 9.864600220201223e-05}
{"train_loss": 0.48535841703414917, "global_step": 20617, "epoch": 231, "lr": 9.864586819730981e-05}
{"train_loss": 0.4296239912509918, "global_step": 20618, "epoch": 231, "lr": 9.864573418606753e-05}
{"train_loss": 0.43264150619506836, "global_step": 20619, "epoch": 231, "lr": 9.864560016828543e-05}
{"train_loss": 0.38192859292030334, "global_step": 20620, "epoch": 231, "lr": 9.864546614396353e-05}
{"train_loss": 0.42997124791145325, "global_step": 20621, "epoch": 231, "lr": 9.864533211310183e-05}
{"train_loss": 0.5021104216575623, "global_step": 20622, "epoch": 231, "lr": 9.864519807570036e-05}
{"train_loss": 0.42548999190330505, "global_step": 20623, "epoch": 231, "lr": 9.864506403175914e-05}
{"train_loss": 0.4386364817619324, "global_step": 20624, "epoch": 231, "lr": 9.864492998127818e-05}
{"train_loss": 0.476595014333725, "global_step": 20625, "epoch": 231, "lr": 9.86447959242575e-05}
{"train_loss": 0.396285742521286, "global_step": 20626, "epoch": 231, "lr": 9.864466186069712e-05}
{"train_loss": 0.6067640781402588, "global_step": 20627, "epoch": 231, "lr": 9.864452779059706e-05}
{"train_loss": 0.41111627221107483, "global_step": 20628, "epoch": 231, "lr": 9.864439371395734e-05}
{"train_loss": 0.5216930508613586, "global_step": 20629, "epoch": 231, "lr": 9.864425963077796e-05}
{"train_loss": 0.6325012445449829, "global_step": 20630, "epoch": 231, "lr": 9.864412554105898e-05}
{"train_loss": 0.5027413368225098, "global_step": 20631, "epoch": 231, "lr": 9.864399144480036e-05}
{"train_loss": 0.3859271705150604, "global_step": 20632, "epoch": 231, "lr": 9.864385734200217e-05}
{"train_loss": 0.44013693928718567, "global_step": 20633, "epoch": 231, "lr": 9.86437232326644e-05}
{"train_loss": 0.49531739950180054, "global_step": 20634, "epoch": 231, "lr": 9.864358911678708e-05}
{"train_loss": 0.6107314825057983, "global_step": 20635, "epoch": 231, "lr": 9.86434549943702e-05}
{"train_loss": 0.48651790618896484, "global_step": 20636, "epoch": 231, "lr": 9.864332086541381e-05}
{"train_loss": 0.6190617680549622, "global_step": 20637, "epoch": 231, "lr": 9.864318672991793e-05}
{"train_loss": 0.6616221070289612, "global_step": 20638, "epoch": 231, "lr": 9.864305258788256e-05}
{"train_loss": 0.33117321133613586, "global_step": 20639, "epoch": 231, "lr": 9.864291843930773e-05}
{"train_loss": 0.4650709331035614, "global_step": 20640, "epoch": 231, "lr": 9.864278428419345e-05}
{"train_loss": 0.5750690698623657, "global_step": 20641, "epoch": 231, "lr": 9.864265012253973e-05}
{"train_loss": 0.5062063336372375, "global_step": 20642, "epoch": 231, "lr": 9.86425159543466e-05}
{"train_loss": 0.46749147772789, "global_step": 20643, "epoch": 231, "lr": 9.864238177961409e-05}
{"train_loss": 0.5017065405845642, "global_step": 20644, "epoch": 231, "lr": 9.86422475983422e-05}
{"train_loss": 0.585562527179718, "global_step": 20645, "epoch": 231, "lr": 9.864211341053095e-05}
{"train_loss": 0.5300526022911072, "global_step": 20646, "epoch": 231, "lr": 9.864197921618034e-05}
{"train_loss": 0.485311350125945, "global_step": 20647, "epoch": 231, "lr": 9.864184501529044e-05, "val_loss": 1.802390694618225}
{"train_loss": 0.41302475333213806, "global_step": 20648, "epoch": 232, "lr": 9.864171080786123e-05}
{"train_loss": 0.45791810750961304, "global_step": 20649, "epoch": 232, "lr": 9.864157659389273e-05}
{"train_loss": 0.47630196809768677, "global_step": 20650, "epoch": 232, "lr": 9.864144237338496e-05}
{"train_loss": 0.5411251187324524, "global_step": 20651, "epoch": 232, "lr": 9.864130814633793e-05}
{"train_loss": 0.4344751238822937, "global_step": 20652, "epoch": 232, "lr": 9.864117391275169e-05}
{"train_loss": 0.4404350817203522, "global_step": 20653, "epoch": 232, "lr": 9.864103967262625e-05}
{"train_loss": 0.5078030228614807, "global_step": 20654, "epoch": 232, "lr": 9.864090542596159e-05}
{"train_loss": 0.4674874246120453, "global_step": 20655, "epoch": 232, "lr": 9.864077117275776e-05}
{"train_loss": 0.4419369697570801, "global_step": 20656, "epoch": 232, "lr": 9.864063691301478e-05}
{"train_loss": 0.5924346446990967, "global_step": 20657, "epoch": 232, "lr": 9.864050264673263e-05}
{"train_loss": 0.5389179587364197, "global_step": 20658, "epoch": 232, "lr": 9.864036837391139e-05}
{"train_loss": 0.6249755024909973, "global_step": 20659, "epoch": 232, "lr": 9.864023409455103e-05}
{"train_loss": 0.4964281916618347, "global_step": 20660, "epoch": 232, "lr": 9.864009980865158e-05}
{"train_loss": 0.5414552688598633, "global_step": 20661, "epoch": 232, "lr": 9.863996551621306e-05}
{"train_loss": 0.5450726747512817, "global_step": 20662, "epoch": 232, "lr": 9.86398312172355e-05}
{"train_loss": 0.45606228709220886, "global_step": 20663, "epoch": 232, "lr": 9.86396969117189e-05}
{"train_loss": 0.5736134648323059, "global_step": 20664, "epoch": 232, "lr": 9.86395625996633e-05}
{"train_loss": 0.4669291377067566, "global_step": 20665, "epoch": 232, "lr": 9.863942828106869e-05}
{"train_loss": 0.47660839557647705, "global_step": 20666, "epoch": 232, "lr": 9.86392939559351e-05}
{"train_loss": 0.46590858697891235, "global_step": 20667, "epoch": 232, "lr": 9.863915962426255e-05}
{"train_loss": 0.5461713671684265, "global_step": 20668, "epoch": 232, "lr": 9.863902528605107e-05}
{"train_loss": 0.5455697774887085, "global_step": 20669, "epoch": 232, "lr": 9.863889094130065e-05}
{"train_loss": 0.6012297868728638, "global_step": 20670, "epoch": 232, "lr": 9.863875659001134e-05}
{"train_loss": 0.6275156736373901, "global_step": 20671, "epoch": 232, "lr": 9.863862223218314e-05}
{"train_loss": 0.49278026819229126, "global_step": 20672, "epoch": 232, "lr": 9.863848786781605e-05}
{"train_loss": 0.4652979373931885, "global_step": 20673, "epoch": 232, "lr": 9.863835349691013e-05}
{"train_loss": 0.3781808316707611, "global_step": 20674, "epoch": 232, "lr": 9.863821911946536e-05}
{"train_loss": 0.5535250306129456, "global_step": 20675, "epoch": 232, "lr": 9.86380847354818e-05}
{"train_loss": 0.5050995349884033, "global_step": 20676, "epoch": 232, "lr": 9.863795034495942e-05}
{"train_loss": 0.45042943954467773, "global_step": 20677, "epoch": 232, "lr": 9.863781594789825e-05}
{"train_loss": 0.5479978919029236, "global_step": 20678, "epoch": 232, "lr": 9.863768154429833e-05}
{"train_loss": 0.45195093750953674, "global_step": 20679, "epoch": 232, "lr": 9.863754713415969e-05}
{"train_loss": 0.4876559376716614, "global_step": 20680, "epoch": 232, "lr": 9.86374127174823e-05}
{"train_loss": 0.46262848377227783, "global_step": 20681, "epoch": 232, "lr": 9.863727829426621e-05}
{"train_loss": 0.49821457266807556, "global_step": 20682, "epoch": 232, "lr": 9.863714386451142e-05}
{"train_loss": 0.49459144473075867, "global_step": 20683, "epoch": 232, "lr": 9.863700942821797e-05}
{"train_loss": 0.5321813821792603, "global_step": 20684, "epoch": 232, "lr": 9.863687498538586e-05}
{"train_loss": 0.6570876240730286, "global_step": 20685, "epoch": 232, "lr": 9.863674053601513e-05}
{"train_loss": 0.4162024259567261, "global_step": 20686, "epoch": 232, "lr": 9.863660608010576e-05}
{"train_loss": 0.5071530342102051, "global_step": 20687, "epoch": 232, "lr": 9.86364716176578e-05}
{"train_loss": 0.4676361680030823, "global_step": 20688, "epoch": 232, "lr": 9.863633714867127e-05}
{"train_loss": 0.5666135549545288, "global_step": 20689, "epoch": 232, "lr": 9.863620267314618e-05}
{"train_loss": 0.572405993938446, "global_step": 20690, "epoch": 232, "lr": 9.863606819108252e-05}
{"train_loss": 0.5226708650588989, "global_step": 20691, "epoch": 232, "lr": 9.863593370248034e-05}
{"train_loss": 0.46028727293014526, "global_step": 20692, "epoch": 232, "lr": 9.863579920733968e-05}
{"train_loss": 0.593917965888977, "global_step": 20693, "epoch": 232, "lr": 9.86356647056605e-05}
{"train_loss": 0.4426390826702118, "global_step": 20694, "epoch": 232, "lr": 9.863553019744285e-05}
{"train_loss": 0.4221000373363495, "global_step": 20695, "epoch": 232, "lr": 9.863539568268676e-05}
{"train_loss": 0.4779999256134033, "global_step": 20696, "epoch": 232, "lr": 9.863526116139223e-05}
{"train_loss": 0.42650628089904785, "global_step": 20697, "epoch": 232, "lr": 9.863512663355928e-05}
{"train_loss": 0.46057212352752686, "global_step": 20698, "epoch": 232, "lr": 9.863499209918792e-05}
{"train_loss": 0.4266429543495178, "global_step": 20699, "epoch": 232, "lr": 9.863485755827819e-05}
{"train_loss": 0.5399222373962402, "global_step": 20700, "epoch": 232, "lr": 9.863472301083008e-05}
{"train_loss": 0.5051246881484985, "global_step": 20701, "epoch": 232, "lr": 9.863458845684364e-05}
{"train_loss": 0.41781651973724365, "global_step": 20702, "epoch": 232, "lr": 9.863445389631888e-05}
{"train_loss": 0.46405336260795593, "global_step": 20703, "epoch": 232, "lr": 9.863431932925579e-05}
{"train_loss": 0.4064541161060333, "global_step": 20704, "epoch": 232, "lr": 9.863418475565442e-05}
{"train_loss": 0.45915982127189636, "global_step": 20705, "epoch": 232, "lr": 9.863405017551478e-05}
{"train_loss": 0.5245798230171204, "global_step": 20706, "epoch": 232, "lr": 9.863391558883687e-05}
{"train_loss": 0.4725288450717926, "global_step": 20707, "epoch": 232, "lr": 9.863378099562074e-05}
{"train_loss": 0.38574883341789246, "global_step": 20708, "epoch": 232, "lr": 9.863364639586639e-05}
{"train_loss": 0.582674503326416, "global_step": 20709, "epoch": 232, "lr": 9.863351178957383e-05}
{"train_loss": 0.4454183578491211, "global_step": 20710, "epoch": 232, "lr": 9.863337717674307e-05}
{"train_loss": 0.5698041915893555, "global_step": 20711, "epoch": 232, "lr": 9.863324255737419e-05}
{"train_loss": 0.4154313802719116, "global_step": 20712, "epoch": 232, "lr": 9.863310793146712e-05}
{"train_loss": 0.6326541900634766, "global_step": 20713, "epoch": 232, "lr": 9.863297329902194e-05}
{"train_loss": 0.49585655331611633, "global_step": 20714, "epoch": 232, "lr": 9.863283866003865e-05}
{"train_loss": 0.44747936725616455, "global_step": 20715, "epoch": 232, "lr": 9.863270401451727e-05}
{"train_loss": 0.41790449619293213, "global_step": 20716, "epoch": 232, "lr": 9.863256936245782e-05}
{"train_loss": 0.6003569960594177, "global_step": 20717, "epoch": 232, "lr": 9.86324347038603e-05}
{"train_loss": 0.4671255052089691, "global_step": 20718, "epoch": 232, "lr": 9.863230003872474e-05}
{"train_loss": 0.45252951979637146, "global_step": 20719, "epoch": 232, "lr": 9.863216536705118e-05}
{"train_loss": 0.4140474200248718, "global_step": 20720, "epoch": 232, "lr": 9.86320306888396e-05}
{"train_loss": 0.5555983781814575, "global_step": 20721, "epoch": 232, "lr": 9.863189600409004e-05}
{"train_loss": 0.4578230679035187, "global_step": 20722, "epoch": 232, "lr": 9.863176131280251e-05}
{"train_loss": 0.48770520091056824, "global_step": 20723, "epoch": 232, "lr": 9.863162661497704e-05}
{"train_loss": 0.5607157945632935, "global_step": 20724, "epoch": 232, "lr": 9.863149191061365e-05}
{"train_loss": 0.5006262063980103, "global_step": 20725, "epoch": 232, "lr": 9.863135719971234e-05}
{"train_loss": 0.5479859113693237, "global_step": 20726, "epoch": 232, "lr": 9.863122248227314e-05}
{"train_loss": 0.3584256172180176, "global_step": 20727, "epoch": 232, "lr": 9.863108775829606e-05}
{"train_loss": 0.5795674920082092, "global_step": 20728, "epoch": 232, "lr": 9.863095302778112e-05}
{"train_loss": 0.46471381187438965, "global_step": 20729, "epoch": 232, "lr": 9.863081829072833e-05}
{"train_loss": 0.4898546636104584, "global_step": 20730, "epoch": 232, "lr": 9.863068354713774e-05}
{"train_loss": 0.5401341915130615, "global_step": 20731, "epoch": 232, "lr": 9.863054879700935e-05}
{"train_loss": 0.43689799308776855, "global_step": 20732, "epoch": 232, "lr": 9.863041404034317e-05}
{"train_loss": 0.6330711841583252, "global_step": 20733, "epoch": 232, "lr": 9.863027927713921e-05}
{"train_loss": 0.473017156124115, "global_step": 20734, "epoch": 232, "lr": 9.863014450739751e-05}
{"train_loss": 0.39614418148994446, "global_step": 20735, "epoch": 232, "lr": 9.863000973111807e-05}
{"train_loss": 0.495968984753898, "global_step": 20736, "epoch": 232, "lr": 9.862987494830094e-05, "val_loss": 1.707641839981079}
{"train_loss": 0.524899959564209, "global_step": 20737, "epoch": 233, "lr": 9.86297401589461e-05}
{"train_loss": 0.4613722860813141, "global_step": 20738, "epoch": 233, "lr": 9.862960536305358e-05}
{"train_loss": 0.4020998179912567, "global_step": 20739, "epoch": 233, "lr": 9.862947056062342e-05}
{"train_loss": 0.39300358295440674, "global_step": 20740, "epoch": 233, "lr": 9.862933575165561e-05}
{"train_loss": 0.43575286865234375, "global_step": 20741, "epoch": 233, "lr": 9.862920093615017e-05}
{"train_loss": 0.5180118083953857, "global_step": 20742, "epoch": 233, "lr": 9.862906611410714e-05}
{"train_loss": 0.48776960372924805, "global_step": 20743, "epoch": 233, "lr": 9.862893128552652e-05}
{"train_loss": 0.5103721618652344, "global_step": 20744, "epoch": 233, "lr": 9.862879645040834e-05}
{"train_loss": 0.5200549364089966, "global_step": 20745, "epoch": 233, "lr": 9.862866160875261e-05}
{"train_loss": 0.44712740182876587, "global_step": 20746, "epoch": 233, "lr": 9.862852676055934e-05}
{"train_loss": 0.3639621138572693, "global_step": 20747, "epoch": 233, "lr": 9.862839190582858e-05}
{"train_loss": 0.42084184288978577, "global_step": 20748, "epoch": 233, "lr": 9.862825704456028e-05}
{"train_loss": 0.5326085686683655, "global_step": 20749, "epoch": 233, "lr": 9.862812217675454e-05}
{"train_loss": 0.4799865782260895, "global_step": 20750, "epoch": 233, "lr": 9.862798730241134e-05}
{"train_loss": 0.4853504002094269, "global_step": 20751, "epoch": 233, "lr": 9.86278524215307e-05}
{"train_loss": 0.4624314606189728, "global_step": 20752, "epoch": 233, "lr": 9.862771753411261e-05}
{"train_loss": 0.3931509256362915, "global_step": 20753, "epoch": 233, "lr": 9.862758264015715e-05}
{"train_loss": 0.5070723295211792, "global_step": 20754, "epoch": 233, "lr": 9.862744773966428e-05}
{"train_loss": 0.3896244466304779, "global_step": 20755, "epoch": 233, "lr": 9.862731283263407e-05}
{"train_loss": 0.48320916295051575, "global_step": 20756, "epoch": 233, "lr": 9.862717791906649e-05}
{"train_loss": 0.4360322058200836, "global_step": 20757, "epoch": 233, "lr": 9.862704299896157e-05}
{"train_loss": 0.3322696089744568, "global_step": 20758, "epoch": 233, "lr": 9.862690807231935e-05}
{"train_loss": 0.5158222317695618, "global_step": 20759, "epoch": 233, "lr": 9.862677313913984e-05}
{"train_loss": 0.5632111430168152, "global_step": 20760, "epoch": 233, "lr": 9.862663819942304e-05}
{"train_loss": 0.4772813618183136, "global_step": 20761, "epoch": 233, "lr": 9.862650325316898e-05}
{"train_loss": 0.387876033782959, "global_step": 20762, "epoch": 233, "lr": 9.86263683003777e-05}
{"train_loss": 0.48100465536117554, "global_step": 20763, "epoch": 233, "lr": 9.862623334104919e-05}
{"train_loss": 0.4994138181209564, "global_step": 20764, "epoch": 233, "lr": 9.862609837518346e-05}
{"train_loss": 0.38767731189727783, "global_step": 20765, "epoch": 233, "lr": 9.862596340278056e-05}
{"train_loss": 0.4218124449253082, "global_step": 20766, "epoch": 233, "lr": 9.862582842384048e-05}
{"train_loss": 0.40963393449783325, "global_step": 20767, "epoch": 233, "lr": 9.862569343836325e-05}
{"train_loss": 0.5738194584846497, "global_step": 20768, "epoch": 233, "lr": 9.862555844634891e-05}
{"train_loss": 0.4781726896762848, "global_step": 20769, "epoch": 233, "lr": 9.862542344779744e-05}
{"train_loss": 0.34753331542015076, "global_step": 20770, "epoch": 233, "lr": 9.862528844270886e-05}
{"train_loss": 0.45286688208580017, "global_step": 20771, "epoch": 233, "lr": 9.862515343108323e-05}
{"train_loss": 0.46457579731941223, "global_step": 20772, "epoch": 233, "lr": 9.862501841292052e-05}
{"train_loss": 0.448136568069458, "global_step": 20773, "epoch": 233, "lr": 9.862488338822078e-05}
{"train_loss": 0.3842821419239044, "global_step": 20774, "epoch": 233, "lr": 9.8624748356984e-05}
{"train_loss": 0.4241393208503723, "global_step": 20775, "epoch": 233, "lr": 9.862461331921025e-05}
{"train_loss": 0.5097370147705078, "global_step": 20776, "epoch": 233, "lr": 9.862447827489949e-05}
{"train_loss": 0.43390390276908875, "global_step": 20777, "epoch": 233, "lr": 9.862434322405177e-05}
{"train_loss": 0.41726619005203247, "global_step": 20778, "epoch": 233, "lr": 9.862420816666709e-05}
{"train_loss": 0.3624444305896759, "global_step": 20779, "epoch": 233, "lr": 9.862407310274548e-05}
{"train_loss": 0.646215558052063, "global_step": 20780, "epoch": 233, "lr": 9.862393803228696e-05}
{"train_loss": 0.5226865410804749, "global_step": 20781, "epoch": 233, "lr": 9.862380295529154e-05}
{"train_loss": 0.5121898651123047, "global_step": 20782, "epoch": 233, "lr": 9.862366787175923e-05}
{"train_loss": 0.4420150816440582, "global_step": 20783, "epoch": 233, "lr": 9.862353278169008e-05}
{"train_loss": 0.46636080741882324, "global_step": 20784, "epoch": 233, "lr": 9.862339768508409e-05}
{"train_loss": 0.45595505833625793, "global_step": 20785, "epoch": 233, "lr": 9.862326258194126e-05}
{"train_loss": 0.5536866188049316, "global_step": 20786, "epoch": 233, "lr": 9.862312747226165e-05}
{"train_loss": 0.371827632188797, "global_step": 20787, "epoch": 233, "lr": 9.862299235604523e-05}
{"train_loss": 0.3990134596824646, "global_step": 20788, "epoch": 233, "lr": 9.862285723329206e-05}
{"train_loss": 0.45429226756095886, "global_step": 20789, "epoch": 233, "lr": 9.862272210400211e-05}
{"train_loss": 0.5385999083518982, "global_step": 20790, "epoch": 233, "lr": 9.862258696817546e-05}
{"train_loss": 0.46094173192977905, "global_step": 20791, "epoch": 233, "lr": 9.862245182581208e-05}
{"train_loss": 0.3961396813392639, "global_step": 20792, "epoch": 233, "lr": 9.862231667691201e-05}
{"train_loss": 0.45520469546318054, "global_step": 20793, "epoch": 233, "lr": 9.862218152147525e-05}
{"train_loss": 0.5947185158729553, "global_step": 20794, "epoch": 233, "lr": 9.862204635950185e-05}
{"train_loss": 0.4134151041507721, "global_step": 20795, "epoch": 233, "lr": 9.862191119099179e-05}
{"train_loss": 0.5105668902397156, "global_step": 20796, "epoch": 233, "lr": 9.862177601594511e-05}
{"train_loss": 0.46115148067474365, "global_step": 20797, "epoch": 233, "lr": 9.862164083436183e-05}
{"train_loss": 0.5727958679199219, "global_step": 20798, "epoch": 233, "lr": 9.862150564624198e-05}
{"train_loss": 0.6492250561714172, "global_step": 20799, "epoch": 233, "lr": 9.862137045158552e-05}
{"train_loss": 0.5596426725387573, "global_step": 20800, "epoch": 233, "lr": 9.862123525039255e-05}
{"train_loss": 0.5878491997718811, "global_step": 20801, "epoch": 233, "lr": 9.862110004266303e-05}
{"train_loss": 0.43865031003952026, "global_step": 20802, "epoch": 233, "lr": 9.862096482839699e-05}
{"train_loss": 0.4981432259082794, "global_step": 20803, "epoch": 233, "lr": 9.862082960759447e-05}
{"train_loss": 0.5659390091896057, "global_step": 20804, "epoch": 233, "lr": 9.862069438025546e-05}
{"train_loss": 0.5406351685523987, "global_step": 20805, "epoch": 233, "lr": 9.862055914637998e-05}
{"train_loss": 0.4994410276412964, "global_step": 20806, "epoch": 233, "lr": 9.862042390596807e-05}
{"train_loss": 0.4607663154602051, "global_step": 20807, "epoch": 233, "lr": 9.862028865901973e-05}
{"train_loss": 0.5870939493179321, "global_step": 20808, "epoch": 233, "lr": 9.862015340553501e-05}
{"train_loss": 0.5155713558197021, "global_step": 20809, "epoch": 233, "lr": 9.862001814551387e-05}
{"train_loss": 0.506126880645752, "global_step": 20810, "epoch": 233, "lr": 9.861988287895638e-05}
{"train_loss": 0.4896481931209564, "global_step": 20811, "epoch": 233, "lr": 9.861974760586255e-05}
{"train_loss": 0.5195196866989136, "global_step": 20812, "epoch": 233, "lr": 9.861961232623236e-05}
{"train_loss": 0.533754289150238, "global_step": 20813, "epoch": 233, "lr": 9.861947704006587e-05}
{"train_loss": 0.4905041754245758, "global_step": 20814, "epoch": 233, "lr": 9.861934174736309e-05}
{"train_loss": 0.38890284299850464, "global_step": 20815, "epoch": 233, "lr": 9.861920644812402e-05}
{"train_loss": 0.477204293012619, "global_step": 20816, "epoch": 233, "lr": 9.86190711423487e-05}
{"train_loss": 0.4948646128177643, "global_step": 20817, "epoch": 233, "lr": 9.861893583003712e-05}
{"train_loss": 0.47388672828674316, "global_step": 20818, "epoch": 233, "lr": 9.861880051118933e-05}
{"train_loss": 0.502763569355011, "global_step": 20819, "epoch": 233, "lr": 9.861866518580533e-05}
{"train_loss": 0.5018560886383057, "global_step": 20820, "epoch": 233, "lr": 9.861852985388516e-05}
{"train_loss": 0.476409912109375, "global_step": 20821, "epoch": 233, "lr": 9.86183945154288e-05}
{"train_loss": 0.4036199450492859, "global_step": 20822, "epoch": 233, "lr": 9.86182591704363e-05}
{"train_loss": 0.48236778378486633, "global_step": 20823, "epoch": 233, "lr": 9.861812381890767e-05}
{"train_loss": 0.5441393852233887, "global_step": 20824, "epoch": 233, "lr": 9.861798846084293e-05}
{"train_loss": 0.47533219316032493, "global_step": 20825, "epoch": 233, "lr": 9.861785309624208e-05, "val_loss": 1.8035310506820679}
{"train_loss": 0.6024522185325623, "global_step": 20826, "epoch": 234, "lr": 9.861771772510515e-05}
{"train_loss": 0.41081610321998596, "global_step": 20827, "epoch": 234, "lr": 9.861758234743217e-05}
{"train_loss": 0.5915700197219849, "global_step": 20828, "epoch": 234, "lr": 9.861744696322314e-05}
{"train_loss": 0.4456373155117035, "global_step": 20829, "epoch": 234, "lr": 9.861731157247811e-05}
{"train_loss": 0.47095945477485657, "global_step": 20830, "epoch": 234, "lr": 9.861717617519705e-05}
{"train_loss": 0.48226019740104675, "global_step": 20831, "epoch": 234, "lr": 9.861704077138001e-05}
{"train_loss": 0.45584654808044434, "global_step": 20832, "epoch": 234, "lr": 9.861690536102701e-05}
{"train_loss": 0.3196188509464264, "global_step": 20833, "epoch": 234, "lr": 9.861676994413806e-05}
{"train_loss": 0.5188782215118408, "global_step": 20834, "epoch": 234, "lr": 9.861663452071317e-05}
{"train_loss": 0.47593772411346436, "global_step": 20835, "epoch": 234, "lr": 9.861649909075237e-05}
{"train_loss": 0.44550374150276184, "global_step": 20836, "epoch": 234, "lr": 9.861636365425567e-05}
{"train_loss": 0.4481131434440613, "global_step": 20837, "epoch": 234, "lr": 9.861622821122309e-05}
{"train_loss": 0.5584231019020081, "global_step": 20838, "epoch": 234, "lr": 9.861609276165466e-05}
{"train_loss": 0.4781743884086609, "global_step": 20839, "epoch": 234, "lr": 9.861595730555037e-05}
{"train_loss": 0.3520974814891815, "global_step": 20840, "epoch": 234, "lr": 9.861582184291027e-05}
{"train_loss": 0.5332025289535522, "global_step": 20841, "epoch": 234, "lr": 9.861568637373437e-05}
{"train_loss": 0.5888747572898865, "global_step": 20842, "epoch": 234, "lr": 9.861555089802268e-05}
{"train_loss": 0.4727051258087158, "global_step": 20843, "epoch": 234, "lr": 9.861541541577523e-05}
{"train_loss": 0.4078478217124939, "global_step": 20844, "epoch": 234, "lr": 9.861527992699202e-05}
{"train_loss": 0.4875212013721466, "global_step": 20845, "epoch": 234, "lr": 9.861514443167307e-05}
{"train_loss": 0.5259779691696167, "global_step": 20846, "epoch": 234, "lr": 9.861500892981842e-05}
{"train_loss": 0.4099219739437103, "global_step": 20847, "epoch": 234, "lr": 9.861487342142808e-05}
{"train_loss": 0.40228164196014404, "global_step": 20848, "epoch": 234, "lr": 9.861473790650204e-05}
{"train_loss": 0.4366370439529419, "global_step": 20849, "epoch": 234, "lr": 9.861460238504036e-05}
{"train_loss": 0.4685913920402527, "global_step": 20850, "epoch": 234, "lr": 9.861446685704303e-05}
{"train_loss": 0.43336021900177, "global_step": 20851, "epoch": 234, "lr": 9.861433132251008e-05}
{"train_loss": 0.40446025133132935, "global_step": 20852, "epoch": 234, "lr": 9.861419578144153e-05}
{"train_loss": 0.4179016053676605, "global_step": 20853, "epoch": 234, "lr": 9.86140602338374e-05}
{"train_loss": 0.5661736726760864, "global_step": 20854, "epoch": 234, "lr": 9.861392467969768e-05}
{"train_loss": 0.5879873037338257, "global_step": 20855, "epoch": 234, "lr": 9.861378911902243e-05}
{"train_loss": 0.43821123242378235, "global_step": 20856, "epoch": 234, "lr": 9.861365355181164e-05}
{"train_loss": 0.6296001672744751, "global_step": 20857, "epoch": 234, "lr": 9.861351797806533e-05}
{"train_loss": 0.4915600121021271, "global_step": 20858, "epoch": 234, "lr": 9.861338239778354e-05}
{"train_loss": 0.7098267078399658, "global_step": 20859, "epoch": 234, "lr": 9.861324681096626e-05}
{"train_loss": 0.42294278740882874, "global_step": 20860, "epoch": 234, "lr": 9.861311121761353e-05}
{"train_loss": 0.5067803263664246, "global_step": 20861, "epoch": 234, "lr": 9.861297561772536e-05}
{"train_loss": 0.45924925804138184, "global_step": 20862, "epoch": 234, "lr": 9.861284001130177e-05}
{"train_loss": 0.6117763519287109, "global_step": 20863, "epoch": 234, "lr": 9.861270439834275e-05}
{"train_loss": 0.5581867694854736, "global_step": 20864, "epoch": 234, "lr": 9.861256877884838e-05}
{"train_loss": 0.4555315673351288, "global_step": 20865, "epoch": 234, "lr": 9.861243315281862e-05}
{"train_loss": 0.457690954208374, "global_step": 20866, "epoch": 234, "lr": 9.861229752025352e-05}
{"train_loss": 0.43903836607933044, "global_step": 20867, "epoch": 234, "lr": 9.861216188115309e-05}
{"train_loss": 0.4356745481491089, "global_step": 20868, "epoch": 234, "lr": 9.861202623551735e-05}
{"train_loss": 0.43424174189567566, "global_step": 20869, "epoch": 234, "lr": 9.861189058334632e-05}
{"train_loss": 0.41914498805999756, "global_step": 20870, "epoch": 234, "lr": 9.861175492464e-05}
{"train_loss": 0.48803648352622986, "global_step": 20871, "epoch": 234, "lr": 9.861161925939843e-05}
{"train_loss": 0.43952611088752747, "global_step": 20872, "epoch": 234, "lr": 9.861148358762163e-05}
{"train_loss": 0.4260990619659424, "global_step": 20873, "epoch": 234, "lr": 9.861134790930958e-05}
{"train_loss": 0.5761438012123108, "global_step": 20874, "epoch": 234, "lr": 9.861121222446235e-05}
{"train_loss": 0.4197501838207245, "global_step": 20875, "epoch": 234, "lr": 9.861107653307993e-05}
{"train_loss": 0.4660278856754303, "global_step": 20876, "epoch": 234, "lr": 9.861094083516233e-05}
{"train_loss": 0.558502733707428, "global_step": 20877, "epoch": 234, "lr": 9.86108051307096e-05}
{"train_loss": 0.43713170289993286, "global_step": 20878, "epoch": 234, "lr": 9.861066941972173e-05}
{"train_loss": 0.5075375437736511, "global_step": 20879, "epoch": 234, "lr": 9.861053370219875e-05}
{"train_loss": 0.5099895000457764, "global_step": 20880, "epoch": 234, "lr": 9.861039797814068e-05}
{"train_loss": 0.37952885031700134, "global_step": 20881, "epoch": 234, "lr": 9.861026224754752e-05}
{"train_loss": 0.41404709219932556, "global_step": 20882, "epoch": 234, "lr": 9.861012651041932e-05}
{"train_loss": 0.5564456582069397, "global_step": 20883, "epoch": 234, "lr": 9.860999076675608e-05}
{"train_loss": 0.3806570768356323, "global_step": 20884, "epoch": 234, "lr": 9.860985501655782e-05}
{"train_loss": 0.5071234703063965, "global_step": 20885, "epoch": 234, "lr": 9.860971925982454e-05}
{"train_loss": 0.5603211522102356, "global_step": 20886, "epoch": 234, "lr": 9.860958349655629e-05}
{"train_loss": 0.4733790159225464, "global_step": 20887, "epoch": 234, "lr": 9.860944772675307e-05}
{"train_loss": 0.5192702412605286, "global_step": 20888, "epoch": 234, "lr": 9.86093119504149e-05}
{"train_loss": 0.5062623620033264, "global_step": 20889, "epoch": 234, "lr": 9.860917616754181e-05}
{"train_loss": 0.48926177620887756, "global_step": 20890, "epoch": 234, "lr": 9.86090403781338e-05}
{"train_loss": 0.44354674220085144, "global_step": 20891, "epoch": 234, "lr": 9.860890458219089e-05}
{"train_loss": 0.5030651092529297, "global_step": 20892, "epoch": 234, "lr": 9.860876877971312e-05}
{"train_loss": 0.4912492036819458, "global_step": 20893, "epoch": 234, "lr": 9.860863297070049e-05}
{"train_loss": 0.3596223294734955, "global_step": 20894, "epoch": 234, "lr": 9.860849715515303e-05}
{"train_loss": 0.41449159383773804, "global_step": 20895, "epoch": 234, "lr": 9.860836133307073e-05}
{"train_loss": 0.5805346965789795, "global_step": 20896, "epoch": 234, "lr": 9.860822550445363e-05}
{"train_loss": 0.3583531677722931, "global_step": 20897, "epoch": 234, "lr": 9.860808966930177e-05}
{"train_loss": 0.427041232585907, "global_step": 20898, "epoch": 234, "lr": 9.860795382761512e-05}
{"train_loss": 0.42079100012779236, "global_step": 20899, "epoch": 234, "lr": 9.860781797939373e-05}
{"train_loss": 0.540909469127655, "global_step": 20900, "epoch": 234, "lr": 9.860768212463762e-05}
{"train_loss": 0.48927491903305054, "global_step": 20901, "epoch": 234, "lr": 9.86075462633468e-05}
{"train_loss": 0.4064445495605469, "global_step": 20902, "epoch": 234, "lr": 9.860741039552128e-05}
{"train_loss": 0.4609470069408417, "global_step": 20903, "epoch": 234, "lr": 9.860727452116109e-05}
{"train_loss": 0.4967704713344574, "global_step": 20904, "epoch": 234, "lr": 9.860713864026625e-05}
{"train_loss": 0.5417798161506653, "global_step": 20905, "epoch": 234, "lr": 9.860700275283675e-05}
{"train_loss": 0.35934311151504517, "global_step": 20906, "epoch": 234, "lr": 9.860686685887264e-05}
{"train_loss": 0.4331221580505371, "global_step": 20907, "epoch": 234, "lr": 9.860673095837393e-05}
{"train_loss": 0.4983278810977936, "global_step": 20908, "epoch": 234, "lr": 9.860659505134065e-05}
{"train_loss": 0.4737798571586609, "global_step": 20909, "epoch": 234, "lr": 9.86064591377728e-05}
{"train_loss": 0.5405967235565186, "global_step": 20910, "epoch": 234, "lr": 9.86063232176704e-05}
{"train_loss": 0.4183484613895416, "global_step": 20911, "epoch": 234, "lr": 9.860618729103348e-05}
{"train_loss": 0.4733284115791321, "global_step": 20912, "epoch": 234, "lr": 9.860605135786205e-05}
{"train_loss": 0.5053123831748962, "global_step": 20913, "epoch": 234, "lr": 9.860591541815611e-05}
{"train_loss": 0.4754278760277823, "global_step": 20914, "epoch": 234, "lr": 9.860577947191572e-05, "val_loss": 1.821799874305725}
{"train_loss": 0.35968607664108276, "global_step": 20915, "epoch": 235, "lr": 9.860564351914086e-05}
{"train_loss": 0.420813649892807, "global_step": 20916, "epoch": 235, "lr": 9.860550755983158e-05}
{"train_loss": 0.547295093536377, "global_step": 20917, "epoch": 235, "lr": 9.860537159398786e-05}
{"train_loss": 0.5166695713996887, "global_step": 20918, "epoch": 235, "lr": 9.860523562160974e-05}
{"train_loss": 0.5738700032234192, "global_step": 20919, "epoch": 235, "lr": 9.860509964269726e-05}
{"train_loss": 0.4405776858329773, "global_step": 20920, "epoch": 235, "lr": 9.860496365725041e-05}
{"train_loss": 0.5052315592765808, "global_step": 20921, "epoch": 235, "lr": 9.860482766526921e-05}
{"train_loss": 0.6412498354911804, "global_step": 20922, "epoch": 235, "lr": 9.860469166675369e-05}
{"train_loss": 0.4807327687740326, "global_step": 20923, "epoch": 235, "lr": 9.860455566170386e-05}
{"train_loss": 0.4691850543022156, "global_step": 20924, "epoch": 235, "lr": 9.860441965011973e-05}
{"train_loss": 0.611406683921814, "global_step": 20925, "epoch": 235, "lr": 9.860428363200134e-05}
{"train_loss": 0.48488813638687134, "global_step": 20926, "epoch": 235, "lr": 9.860414760734869e-05}
{"train_loss": 0.45346492528915405, "global_step": 20927, "epoch": 235, "lr": 9.860401157616182e-05}
{"train_loss": 0.37330251932144165, "global_step": 20928, "epoch": 235, "lr": 9.86038755384407e-05}
{"train_loss": 0.39741700887680054, "global_step": 20929, "epoch": 235, "lr": 9.860373949418541e-05}
{"train_loss": 0.4567798674106598, "global_step": 20930, "epoch": 235, "lr": 9.860360344339592e-05}
{"train_loss": 0.4797297418117523, "global_step": 20931, "epoch": 235, "lr": 9.860346738607229e-05}
{"train_loss": 0.43865394592285156, "global_step": 20932, "epoch": 235, "lr": 9.86033313222145e-05}
{"train_loss": 0.44186365604400635, "global_step": 20933, "epoch": 235, "lr": 9.86031952518226e-05}
{"train_loss": 0.467849999666214, "global_step": 20934, "epoch": 235, "lr": 9.860305917489657e-05}
{"train_loss": 0.5356617569923401, "global_step": 20935, "epoch": 235, "lr": 9.860292309143646e-05}
{"train_loss": 0.4901571273803711, "global_step": 20936, "epoch": 235, "lr": 9.860278700144229e-05}
{"train_loss": 0.5841140747070312, "global_step": 20937, "epoch": 235, "lr": 9.860265090491405e-05}
{"train_loss": 0.513446569442749, "global_step": 20938, "epoch": 235, "lr": 9.860251480185178e-05}
{"train_loss": 0.4280005097389221, "global_step": 20939, "epoch": 235, "lr": 9.860237869225552e-05}
{"train_loss": 0.44138121604919434, "global_step": 20940, "epoch": 235, "lr": 9.860224257612523e-05}
{"train_loss": 0.3417510688304901, "global_step": 20941, "epoch": 235, "lr": 9.860210645346097e-05}
{"train_loss": 0.5409491062164307, "global_step": 20942, "epoch": 235, "lr": 9.860197032426277e-05}
{"train_loss": 0.5648622512817383, "global_step": 20943, "epoch": 235, "lr": 9.860183418853059e-05}
{"train_loss": 0.3996546268463135, "global_step": 20944, "epoch": 235, "lr": 9.860169804626451e-05}
{"train_loss": 0.5272432565689087, "global_step": 20945, "epoch": 235, "lr": 9.860156189746452e-05}
{"train_loss": 0.5412908792495728, "global_step": 20946, "epoch": 235, "lr": 9.860142574213065e-05}
{"train_loss": 0.4008767306804657, "global_step": 20947, "epoch": 235, "lr": 9.86012895802629e-05}
{"train_loss": 0.41546911001205444, "global_step": 20948, "epoch": 235, "lr": 9.86011534118613e-05}
{"train_loss": 0.5455480813980103, "global_step": 20949, "epoch": 235, "lr": 9.860101723692588e-05}
{"train_loss": 0.4425894021987915, "global_step": 20950, "epoch": 235, "lr": 9.860088105545663e-05}
{"train_loss": 0.47051090002059937, "global_step": 20951, "epoch": 235, "lr": 9.860074486745359e-05}
{"train_loss": 0.5859041810035706, "global_step": 20952, "epoch": 235, "lr": 9.860060867291677e-05}
{"train_loss": 0.46482983231544495, "global_step": 20953, "epoch": 235, "lr": 9.860047247184618e-05}
{"train_loss": 0.552592933177948, "global_step": 20954, "epoch": 235, "lr": 9.860033626424187e-05}
{"train_loss": 0.4158006012439728, "global_step": 20955, "epoch": 235, "lr": 9.860020005010383e-05}
{"train_loss": 0.46207910776138306, "global_step": 20956, "epoch": 235, "lr": 9.860006382943208e-05}
{"train_loss": 0.41398635506629944, "global_step": 20957, "epoch": 235, "lr": 9.859992760222665e-05}
{"train_loss": 0.5382870435714722, "global_step": 20958, "epoch": 235, "lr": 9.859979136848755e-05}
{"train_loss": 0.590637743473053, "global_step": 20959, "epoch": 235, "lr": 9.85996551282148e-05}
{"train_loss": 0.4752580225467682, "global_step": 20960, "epoch": 235, "lr": 9.859951888140842e-05}
{"train_loss": 0.44301679730415344, "global_step": 20961, "epoch": 235, "lr": 9.859938262806843e-05}
{"train_loss": 0.5137559175491333, "global_step": 20962, "epoch": 235, "lr": 9.859924636819483e-05}
{"train_loss": 0.5022606253623962, "global_step": 20963, "epoch": 235, "lr": 9.859911010178767e-05}
{"train_loss": 0.4121645987033844, "global_step": 20964, "epoch": 235, "lr": 9.859897382884694e-05}
{"train_loss": 0.510159432888031, "global_step": 20965, "epoch": 235, "lr": 9.859883754937266e-05}
{"train_loss": 0.49126726388931274, "global_step": 20966, "epoch": 235, "lr": 9.859870126336489e-05}
{"train_loss": 0.5128966569900513, "global_step": 20967, "epoch": 235, "lr": 9.85985649708236e-05}
{"train_loss": 0.49813854694366455, "global_step": 20968, "epoch": 235, "lr": 9.859842867174882e-05}
{"train_loss": 0.41799038648605347, "global_step": 20969, "epoch": 235, "lr": 9.859829236614058e-05}
{"train_loss": 0.4520898461341858, "global_step": 20970, "epoch": 235, "lr": 9.85981560539989e-05}
{"train_loss": 0.39776915311813354, "global_step": 20971, "epoch": 235, "lr": 9.859801973532379e-05}
{"train_loss": 0.5219694375991821, "global_step": 20972, "epoch": 235, "lr": 9.859788341011526e-05}
{"train_loss": 0.48266249895095825, "global_step": 20973, "epoch": 235, "lr": 9.859774707837333e-05}
{"train_loss": 0.5499498248100281, "global_step": 20974, "epoch": 235, "lr": 9.859761074009804e-05}
{"train_loss": 0.6129719614982605, "global_step": 20975, "epoch": 235, "lr": 9.859747439528939e-05}
{"train_loss": 0.43691661953926086, "global_step": 20976, "epoch": 235, "lr": 9.85973380439474e-05}
{"train_loss": 0.5198696255683899, "global_step": 20977, "epoch": 235, "lr": 9.859720168607208e-05}
{"train_loss": 0.477605938911438, "global_step": 20978, "epoch": 235, "lr": 9.859706532166347e-05}
{"train_loss": 0.48634934425354004, "global_step": 20979, "epoch": 235, "lr": 9.859692895072158e-05}
{"train_loss": 0.5416329503059387, "global_step": 20980, "epoch": 235, "lr": 9.859679257324642e-05}
{"train_loss": 0.432827889919281, "global_step": 20981, "epoch": 235, "lr": 9.859665618923801e-05}
{"train_loss": 0.4172687232494354, "global_step": 20982, "epoch": 235, "lr": 9.859651979869637e-05}
{"train_loss": 0.41719764471054077, "global_step": 20983, "epoch": 235, "lr": 9.859638340162154e-05}
{"train_loss": 0.5092281699180603, "global_step": 20984, "epoch": 235, "lr": 9.85962469980135e-05}
{"train_loss": 0.5979042053222656, "global_step": 20985, "epoch": 235, "lr": 9.859611058787228e-05}
{"train_loss": 0.5420458912849426, "global_step": 20986, "epoch": 235, "lr": 9.859597417119792e-05}
{"train_loss": 0.5865387320518494, "global_step": 20987, "epoch": 235, "lr": 9.859583774799042e-05}
{"train_loss": 0.4503328502178192, "global_step": 20988, "epoch": 235, "lr": 9.85957013182498e-05}
{"train_loss": 0.4212726056575775, "global_step": 20989, "epoch": 235, "lr": 9.859556488197609e-05}
{"train_loss": 0.5357292294502258, "global_step": 20990, "epoch": 235, "lr": 9.859542843916929e-05}
{"train_loss": 0.5103383660316467, "global_step": 20991, "epoch": 235, "lr": 9.859529198982942e-05}
{"train_loss": 0.5555542707443237, "global_step": 20992, "epoch": 235, "lr": 9.859515553395652e-05}
{"train_loss": 0.3514430522918701, "global_step": 20993, "epoch": 235, "lr": 9.859501907155057e-05}
{"train_loss": 0.4669511318206787, "global_step": 20994, "epoch": 235, "lr": 9.859488260261165e-05}
{"train_loss": 0.5118553638458252, "global_step": 20995, "epoch": 235, "lr": 9.85947461271397e-05}
{"train_loss": 0.5583721995353699, "global_step": 20996, "epoch": 235, "lr": 9.85946096451348e-05}
{"train_loss": 0.42541593313217163, "global_step": 20997, "epoch": 235, "lr": 9.859447315659695e-05}
{"train_loss": 0.4540860950946808, "global_step": 20998, "epoch": 235, "lr": 9.859433666152617e-05}
{"train_loss": 0.43989458680152893, "global_step": 20999, "epoch": 235, "lr": 9.859420015992246e-05}
{"train_loss": 0.6138201355934143, "global_step": 21000, "epoch": 235, "lr": 9.859406365178585e-05}
{"train_loss": 0.5465754866600037, "global_step": 21001, "epoch": 235, "lr": 9.859392713711637e-05}
{"train_loss": 0.5039023160934448, "global_step": 21002, "epoch": 235, "lr": 9.859379061591403e-05}
{"train_loss": 0.48947433574815813, "global_step": 21003, "epoch": 235, "lr": 9.859365408817884e-05, "val_loss": 1.750205159187317, "train_action_mse_error": 24.62740707397461}
{"train_loss": 0.45766642689704895, "global_step": 21004, "epoch": 236, "lr": 9.859351755391082e-05}
{"train_loss": 0.431598961353302, "global_step": 21005, "epoch": 236, "lr": 9.859338101311e-05}
{"train_loss": 0.43575459718704224, "global_step": 21006, "epoch": 236, "lr": 9.859324446577641e-05}
{"train_loss": 0.4795904755592346, "global_step": 21007, "epoch": 236, "lr": 9.859310791191002e-05}
{"train_loss": 0.5163092613220215, "global_step": 21008, "epoch": 236, "lr": 9.859297135151088e-05}
{"train_loss": 0.37925100326538086, "global_step": 21009, "epoch": 236, "lr": 9.859283478457903e-05}
{"train_loss": 0.4943695664405823, "global_step": 21010, "epoch": 236, "lr": 9.859269821111445e-05}
{"train_loss": 0.520887553691864, "global_step": 21011, "epoch": 236, "lr": 9.859256163111719e-05}
{"train_loss": 0.3787825405597687, "global_step": 21012, "epoch": 236, "lr": 9.859242504458725e-05}
{"train_loss": 0.38118961453437805, "global_step": 21013, "epoch": 236, "lr": 9.859228845152464e-05}
{"train_loss": 0.5267886519432068, "global_step": 21014, "epoch": 236, "lr": 9.859215185192938e-05}
{"train_loss": 0.5057231187820435, "global_step": 21015, "epoch": 236, "lr": 9.859201524580152e-05}
{"train_loss": 0.5063309073448181, "global_step": 21016, "epoch": 236, "lr": 9.859187863314103e-05}
{"train_loss": 0.4843440353870392, "global_step": 21017, "epoch": 236, "lr": 9.859174201394798e-05}
{"train_loss": 0.444336473941803, "global_step": 21018, "epoch": 236, "lr": 9.859160538822235e-05}
{"train_loss": 0.45442768931388855, "global_step": 21019, "epoch": 236, "lr": 9.859146875596417e-05}
{"train_loss": 0.5611236691474915, "global_step": 21020, "epoch": 236, "lr": 9.859133211717346e-05}
{"train_loss": 0.5423557162284851, "global_step": 21021, "epoch": 236, "lr": 9.859119547185025e-05}
{"train_loss": 0.5304020643234253, "global_step": 21022, "epoch": 236, "lr": 9.859105881999453e-05}
{"train_loss": 0.3423192501068115, "global_step": 21023, "epoch": 236, "lr": 9.859092216160634e-05}
{"train_loss": 0.418610543012619, "global_step": 21024, "epoch": 236, "lr": 9.85907854966857e-05}
{"train_loss": 0.4627833366394043, "global_step": 21025, "epoch": 236, "lr": 9.859064882523261e-05}
{"train_loss": 0.5311067700386047, "global_step": 21026, "epoch": 236, "lr": 9.85905121472471e-05}
{"train_loss": 0.42507994174957275, "global_step": 21027, "epoch": 236, "lr": 9.85903754627292e-05}
{"train_loss": 0.3456735908985138, "global_step": 21028, "epoch": 236, "lr": 9.85902387716789e-05}
{"train_loss": 0.5284998416900635, "global_step": 21029, "epoch": 236, "lr": 9.859010207409625e-05}
{"train_loss": 0.4190538227558136, "global_step": 21030, "epoch": 236, "lr": 9.858996536998125e-05}
{"train_loss": 0.44478389620780945, "global_step": 21031, "epoch": 236, "lr": 9.85898286593339e-05}
{"train_loss": 0.32740518450737, "global_step": 21032, "epoch": 236, "lr": 9.858969194215426e-05}
{"train_loss": 0.4753856956958771, "global_step": 21033, "epoch": 236, "lr": 9.858955521844234e-05}
{"train_loss": 0.4992735981941223, "global_step": 21034, "epoch": 236, "lr": 9.858941848819812e-05}
{"train_loss": 0.44331109523773193, "global_step": 21035, "epoch": 236, "lr": 9.858928175142166e-05}
{"train_loss": 0.4312494695186615, "global_step": 21036, "epoch": 236, "lr": 9.858914500811295e-05}
{"train_loss": 0.3664271831512451, "global_step": 21037, "epoch": 236, "lr": 9.858900825827204e-05}
{"train_loss": 0.4802416265010834, "global_step": 21038, "epoch": 236, "lr": 9.858887150189891e-05}
{"train_loss": 0.4241558611392975, "global_step": 21039, "epoch": 236, "lr": 9.858873473899362e-05}
{"train_loss": 0.44344010949134827, "global_step": 21040, "epoch": 236, "lr": 9.858859796955615e-05}
{"train_loss": 0.4599977135658264, "global_step": 21041, "epoch": 236, "lr": 9.858846119358653e-05}
{"train_loss": 0.45844054222106934, "global_step": 21042, "epoch": 236, "lr": 9.858832441108481e-05}
{"train_loss": 0.430446058511734, "global_step": 21043, "epoch": 236, "lr": 9.858818762205096e-05}
{"train_loss": 0.4139504134654999, "global_step": 21044, "epoch": 236, "lr": 9.858805082648503e-05}
{"train_loss": 0.5020503997802734, "global_step": 21045, "epoch": 236, "lr": 9.858791402438703e-05}
{"train_loss": 0.46662747859954834, "global_step": 21046, "epoch": 236, "lr": 9.858777721575697e-05}
{"train_loss": 0.4587245285511017, "global_step": 21047, "epoch": 236, "lr": 9.858764040059487e-05}
{"train_loss": 0.39032721519470215, "global_step": 21048, "epoch": 236, "lr": 9.858750357890077e-05}
{"train_loss": 0.4800020754337311, "global_step": 21049, "epoch": 236, "lr": 9.858736675067467e-05}
{"train_loss": 0.5287841558456421, "global_step": 21050, "epoch": 236, "lr": 9.858722991591658e-05}
{"train_loss": 0.4499407112598419, "global_step": 21051, "epoch": 236, "lr": 9.858709307462655e-05}
{"train_loss": 0.49620679020881653, "global_step": 21052, "epoch": 236, "lr": 9.858695622680455e-05}
{"train_loss": 0.32363662123680115, "global_step": 21053, "epoch": 236, "lr": 9.858681937245065e-05}
{"train_loss": 0.29769349098205566, "global_step": 21054, "epoch": 236, "lr": 9.858668251156482e-05}
{"train_loss": 0.44619280099868774, "global_step": 21055, "epoch": 236, "lr": 9.858654564414711e-05}
{"train_loss": 0.5078783631324768, "global_step": 21056, "epoch": 236, "lr": 9.858640877019755e-05}
{"train_loss": 0.40522924065589905, "global_step": 21057, "epoch": 236, "lr": 9.858627188971613e-05}
{"train_loss": 0.5024712085723877, "global_step": 21058, "epoch": 236, "lr": 9.858613500270286e-05}
{"train_loss": 0.5631653666496277, "global_step": 21059, "epoch": 236, "lr": 9.85859981091578e-05}
{"train_loss": 0.4398120939731598, "global_step": 21060, "epoch": 236, "lr": 9.858586120908093e-05}
{"train_loss": 0.5096559524536133, "global_step": 21061, "epoch": 236, "lr": 9.858572430247228e-05}
{"train_loss": 0.46543601155281067, "global_step": 21062, "epoch": 236, "lr": 9.85855873893319e-05}
{"train_loss": 0.5754953622817993, "global_step": 21063, "epoch": 236, "lr": 9.858545046965975e-05}
{"train_loss": 0.4395819306373596, "global_step": 21064, "epoch": 236, "lr": 9.858531354345588e-05}
{"train_loss": 0.44849860668182373, "global_step": 21065, "epoch": 236, "lr": 9.858517661072033e-05}
{"train_loss": 0.39741989970207214, "global_step": 21066, "epoch": 236, "lr": 9.858503967145307e-05}
{"train_loss": 0.4256106913089752, "global_step": 21067, "epoch": 236, "lr": 9.858490272565415e-05}
{"train_loss": 0.3594224154949188, "global_step": 21068, "epoch": 236, "lr": 9.85847657733236e-05}
{"train_loss": 0.5267862677574158, "global_step": 21069, "epoch": 236, "lr": 9.85846288144614e-05}
{"train_loss": 0.40483975410461426, "global_step": 21070, "epoch": 236, "lr": 9.85844918490676e-05}
{"train_loss": 0.4379653334617615, "global_step": 21071, "epoch": 236, "lr": 9.858435487714219e-05}
{"train_loss": 0.336091548204422, "global_step": 21072, "epoch": 236, "lr": 9.858421789868521e-05}
{"train_loss": 0.4506847858428955, "global_step": 21073, "epoch": 236, "lr": 9.85840809136967e-05}
{"train_loss": 0.39657774567604065, "global_step": 21074, "epoch": 236, "lr": 9.858394392217663e-05}
{"train_loss": 0.5287694931030273, "global_step": 21075, "epoch": 236, "lr": 9.858380692412504e-05}
{"train_loss": 0.37508493661880493, "global_step": 21076, "epoch": 236, "lr": 9.858366991954196e-05}
{"train_loss": 0.44680508971214294, "global_step": 21077, "epoch": 236, "lr": 9.858353290842739e-05}
{"train_loss": 0.5307943820953369, "global_step": 21078, "epoch": 236, "lr": 9.858339589078135e-05}
{"train_loss": 0.4682963788509369, "global_step": 21079, "epoch": 236, "lr": 9.85832588666039e-05}
{"train_loss": 0.4095364511013031, "global_step": 21080, "epoch": 236, "lr": 9.858312183589497e-05}
{"train_loss": 0.40678390860557556, "global_step": 21081, "epoch": 236, "lr": 9.858298479865466e-05}
{"train_loss": 0.5075438022613525, "global_step": 21082, "epoch": 236, "lr": 9.858284775488296e-05}
{"train_loss": 0.33269745111465454, "global_step": 21083, "epoch": 236, "lr": 9.858271070457989e-05}
{"train_loss": 0.3361210823059082, "global_step": 21084, "epoch": 236, "lr": 9.858257364774546e-05}
{"train_loss": 0.5604134798049927, "global_step": 21085, "epoch": 236, "lr": 9.85824365843797e-05}
{"train_loss": 0.46868816018104553, "global_step": 21086, "epoch": 236, "lr": 9.858229951448262e-05}
{"train_loss": 0.3884565532207489, "global_step": 21087, "epoch": 236, "lr": 9.858216243805424e-05}
{"train_loss": 0.32336336374282837, "global_step": 21088, "epoch": 236, "lr": 9.858202535509458e-05}
{"train_loss": 0.6648846864700317, "global_step": 21089, "epoch": 236, "lr": 9.858188826560367e-05}
{"train_loss": 0.48297712206840515, "global_step": 21090, "epoch": 236, "lr": 9.85817511695815e-05}
{"train_loss": 0.420543372631073, "global_step": 21091, "epoch": 236, "lr": 9.858161406702813e-05}
{"train_loss": 0.4519409466995282, "global_step": 21092, "epoch": 236, "lr": 9.858147695794353e-05, "val_loss": 1.8661035299301147}
{"train_loss": 0.436520516872406, "global_step": 21093, "epoch": 237, "lr": 9.858133984232775e-05}
{"train_loss": 0.41410380601882935, "global_step": 21094, "epoch": 237, "lr": 9.858120272018081e-05}
{"train_loss": 0.47575291991233826, "global_step": 21095, "epoch": 237, "lr": 9.858106559150271e-05}
{"train_loss": 0.5635500550270081, "global_step": 21096, "epoch": 237, "lr": 9.858092845629348e-05}
{"train_loss": 0.4893256425857544, "global_step": 21097, "epoch": 237, "lr": 9.858079131455313e-05}
{"train_loss": 0.5139442682266235, "global_step": 21098, "epoch": 237, "lr": 9.85806541662817e-05}
{"train_loss": 0.3784361183643341, "global_step": 21099, "epoch": 237, "lr": 9.858051701147918e-05}
{"train_loss": 0.49055778980255127, "global_step": 21100, "epoch": 237, "lr": 9.858037985014561e-05}
{"train_loss": 0.40179088711738586, "global_step": 21101, "epoch": 237, "lr": 9.858024268228099e-05}
{"train_loss": 0.50933837890625, "global_step": 21102, "epoch": 237, "lr": 9.858010550788535e-05}
{"train_loss": 0.470363974571228, "global_step": 21103, "epoch": 237, "lr": 9.857996832695872e-05}
{"train_loss": 0.5033351182937622, "global_step": 21104, "epoch": 237, "lr": 9.857983113950109e-05}
{"train_loss": 0.3396875560283661, "global_step": 21105, "epoch": 237, "lr": 9.85796939455125e-05}
{"train_loss": 0.500263512134552, "global_step": 21106, "epoch": 237, "lr": 9.857955674499297e-05}
{"train_loss": 0.5774328708648682, "global_step": 21107, "epoch": 237, "lr": 9.857941953794249e-05}
{"train_loss": 0.33844009041786194, "global_step": 21108, "epoch": 237, "lr": 9.857928232436112e-05}
{"train_loss": 0.45938488841056824, "global_step": 21109, "epoch": 237, "lr": 9.857914510424885e-05}
{"train_loss": 0.3916109502315521, "global_step": 21110, "epoch": 237, "lr": 9.85790078776057e-05}
{"train_loss": 0.4143648147583008, "global_step": 21111, "epoch": 237, "lr": 9.85788706444317e-05}
{"train_loss": 0.5158373117446899, "global_step": 21112, "epoch": 237, "lr": 9.857873340472687e-05}
{"train_loss": 0.41787415742874146, "global_step": 21113, "epoch": 237, "lr": 9.857859615849122e-05}
{"train_loss": 0.5410251617431641, "global_step": 21114, "epoch": 237, "lr": 9.857845890572477e-05}
{"train_loss": 0.4784105718135834, "global_step": 21115, "epoch": 237, "lr": 9.857832164642753e-05}
{"train_loss": 0.4088301658630371, "global_step": 21116, "epoch": 237, "lr": 9.857818438059952e-05}
{"train_loss": 0.4149773120880127, "global_step": 21117, "epoch": 237, "lr": 9.857804710824077e-05}
{"train_loss": 0.4361850917339325, "global_step": 21118, "epoch": 237, "lr": 9.85779098293513e-05}
{"train_loss": 0.47506749629974365, "global_step": 21119, "epoch": 237, "lr": 9.857777254393112e-05}
{"train_loss": 0.5245126485824585, "global_step": 21120, "epoch": 237, "lr": 9.857763525198027e-05}
{"train_loss": 0.5099791288375854, "global_step": 21121, "epoch": 237, "lr": 9.857749795349872e-05}
{"train_loss": 0.4699556529521942, "global_step": 21122, "epoch": 237, "lr": 9.857736064848653e-05}
{"train_loss": 0.5349360108375549, "global_step": 21123, "epoch": 237, "lr": 9.857722333694371e-05}
{"train_loss": 0.4388382136821747, "global_step": 21124, "epoch": 237, "lr": 9.857708601887025e-05}
{"train_loss": 0.3851555585861206, "global_step": 21125, "epoch": 237, "lr": 9.857694869426621e-05}
{"train_loss": 0.5763771533966064, "global_step": 21126, "epoch": 237, "lr": 9.85768113631316e-05}
{"train_loss": 0.44978708028793335, "global_step": 21127, "epoch": 237, "lr": 9.857667402546642e-05}
{"train_loss": 0.37887006998062134, "global_step": 21128, "epoch": 237, "lr": 9.85765366812707e-05}
{"train_loss": 0.49790844321250916, "global_step": 21129, "epoch": 237, "lr": 9.857639933054446e-05}
{"train_loss": 0.6758817434310913, "global_step": 21130, "epoch": 237, "lr": 9.857626197328771e-05}
{"train_loss": 0.3769974708557129, "global_step": 21131, "epoch": 237, "lr": 9.857612460950048e-05}
{"train_loss": 0.5056717395782471, "global_step": 21132, "epoch": 237, "lr": 9.857598723918278e-05}
{"train_loss": 0.5634588003158569, "global_step": 21133, "epoch": 237, "lr": 9.857584986233464e-05}
{"train_loss": 0.5847700834274292, "global_step": 21134, "epoch": 237, "lr": 9.857571247895605e-05}
{"train_loss": 0.4657190144062042, "global_step": 21135, "epoch": 237, "lr": 9.857557508904705e-05}
{"train_loss": 0.3815787136554718, "global_step": 21136, "epoch": 237, "lr": 9.857543769260766e-05}
{"train_loss": 0.4090169370174408, "global_step": 21137, "epoch": 237, "lr": 9.857530028963789e-05}
{"train_loss": 0.5348604917526245, "global_step": 21138, "epoch": 237, "lr": 9.857516288013777e-05}
{"train_loss": 0.5425059795379639, "global_step": 21139, "epoch": 237, "lr": 9.85750254641073e-05}
{"train_loss": 0.5351333618164062, "global_step": 21140, "epoch": 237, "lr": 9.857488804154653e-05}
{"train_loss": 0.49043986201286316, "global_step": 21141, "epoch": 237, "lr": 9.857475061245545e-05}
{"train_loss": 0.3881276249885559, "global_step": 21142, "epoch": 237, "lr": 9.857461317683409e-05}
{"train_loss": 0.49122875928878784, "global_step": 21143, "epoch": 237, "lr": 9.857447573468246e-05}
{"train_loss": 0.5353244543075562, "global_step": 21144, "epoch": 237, "lr": 9.857433828600058e-05}
{"train_loss": 0.49350619316101074, "global_step": 21145, "epoch": 237, "lr": 9.857420083078848e-05}
{"train_loss": 0.3921416401863098, "global_step": 21146, "epoch": 237, "lr": 9.857406336904617e-05}
{"train_loss": 0.5097950100898743, "global_step": 21147, "epoch": 237, "lr": 9.857392590077366e-05}
{"train_loss": 0.4362453818321228, "global_step": 21148, "epoch": 237, "lr": 9.857378842597099e-05}
{"train_loss": 0.38177844882011414, "global_step": 21149, "epoch": 237, "lr": 9.857365094463815e-05}
{"train_loss": 0.46125712990760803, "global_step": 21150, "epoch": 237, "lr": 9.857351345677519e-05}
{"train_loss": 0.4986763000488281, "global_step": 21151, "epoch": 237, "lr": 9.857337596238211e-05}
{"train_loss": 0.40230119228363037, "global_step": 21152, "epoch": 237, "lr": 9.857323846145893e-05}
{"train_loss": 0.43651750683784485, "global_step": 21153, "epoch": 237, "lr": 9.857310095400567e-05}
{"train_loss": 0.4167393743991852, "global_step": 21154, "epoch": 237, "lr": 9.857296344002233e-05}
{"train_loss": 0.531913161277771, "global_step": 21155, "epoch": 237, "lr": 9.857282591950898e-05}
{"train_loss": 0.4144972562789917, "global_step": 21156, "epoch": 237, "lr": 9.857268839246558e-05}
{"train_loss": 0.45380720496177673, "global_step": 21157, "epoch": 237, "lr": 9.857255085889218e-05}
{"train_loss": 0.5612990856170654, "global_step": 21158, "epoch": 237, "lr": 9.857241331878881e-05}
{"train_loss": 0.4894619584083557, "global_step": 21159, "epoch": 237, "lr": 9.857227577215547e-05}
{"train_loss": 0.38047343492507935, "global_step": 21160, "epoch": 237, "lr": 9.857213821899216e-05}
{"train_loss": 0.4290865957736969, "global_step": 21161, "epoch": 237, "lr": 9.857200065929893e-05}
{"train_loss": 0.4061315655708313, "global_step": 21162, "epoch": 237, "lr": 9.857186309307578e-05}
{"train_loss": 0.4133365750312805, "global_step": 21163, "epoch": 237, "lr": 9.857172552032274e-05}
{"train_loss": 0.5763081312179565, "global_step": 21164, "epoch": 237, "lr": 9.857158794103982e-05}
{"train_loss": 0.5998148918151855, "global_step": 21165, "epoch": 237, "lr": 9.857145035522703e-05}
{"train_loss": 0.4992435872554779, "global_step": 21166, "epoch": 237, "lr": 9.857131276288443e-05}
{"train_loss": 0.4424899220466614, "global_step": 21167, "epoch": 237, "lr": 9.857117516401199e-05}
{"train_loss": 0.5143349766731262, "global_step": 21168, "epoch": 237, "lr": 9.857103755860975e-05}
{"train_loss": 0.4555760324001312, "global_step": 21169, "epoch": 237, "lr": 9.857089994667773e-05}
{"train_loss": 0.3941600024700165, "global_step": 21170, "epoch": 237, "lr": 9.857076232821593e-05}
{"train_loss": 0.5267200469970703, "global_step": 21171, "epoch": 237, "lr": 9.85706247032244e-05}
{"train_loss": 0.3905222713947296, "global_step": 21172, "epoch": 237, "lr": 9.857048707170313e-05}
{"train_loss": 0.348080575466156, "global_step": 21173, "epoch": 237, "lr": 9.857034943365215e-05}
{"train_loss": 0.3942452371120453, "global_step": 21174, "epoch": 237, "lr": 9.857021178907148e-05}
{"train_loss": 0.47031939029693604, "global_step": 21175, "epoch": 237, "lr": 9.857007413796115e-05}
{"train_loss": 0.5922688841819763, "global_step": 21176, "epoch": 237, "lr": 9.856993648032116e-05}
{"train_loss": 0.4475667178630829, "global_step": 21177, "epoch": 237, "lr": 9.856979881615153e-05}
{"train_loss": 0.5646944046020508, "global_step": 21178, "epoch": 237, "lr": 9.856966114545228e-05}
{"train_loss": 0.505795955657959, "global_step": 21179, "epoch": 237, "lr": 9.856952346822344e-05}
{"train_loss": 0.5079128742218018, "global_step": 21180, "epoch": 237, "lr": 9.856938578446501e-05}
{"train_loss": 0.46815847747781303, "global_step": 21181, "epoch": 237, "lr": 9.856924809417702e-05, "val_loss": 1.7678232192993164}
{"train_loss": 0.5455342531204224, "global_step": 21182, "epoch": 238, "lr": 9.856911039735948e-05}
{"train_loss": 0.4731687307357788, "global_step": 21183, "epoch": 238, "lr": 9.856897269401243e-05}
{"train_loss": 0.5030152797698975, "global_step": 21184, "epoch": 238, "lr": 9.856883498413587e-05}
{"train_loss": 0.4297405481338501, "global_step": 21185, "epoch": 238, "lr": 9.856869726772981e-05}
{"train_loss": 0.44858744740486145, "global_step": 21186, "epoch": 238, "lr": 9.85685595447943e-05}
{"train_loss": 0.5474644899368286, "global_step": 21187, "epoch": 238, "lr": 9.856842181532932e-05}
{"train_loss": 0.5141133666038513, "global_step": 21188, "epoch": 238, "lr": 9.856828407933492e-05}
{"train_loss": 0.425672322511673, "global_step": 21189, "epoch": 238, "lr": 9.856814633681111e-05}
{"train_loss": 0.38708701729774475, "global_step": 21190, "epoch": 238, "lr": 9.856800858775789e-05}
{"train_loss": 0.5489230751991272, "global_step": 21191, "epoch": 238, "lr": 9.856787083217528e-05}
{"train_loss": 0.44020047783851624, "global_step": 21192, "epoch": 238, "lr": 9.856773307006335e-05}
{"train_loss": 0.5114960670471191, "global_step": 21193, "epoch": 238, "lr": 9.856759530142206e-05}
{"train_loss": 0.5078650116920471, "global_step": 21194, "epoch": 238, "lr": 9.856745752625144e-05}
{"train_loss": 0.503084123134613, "global_step": 21195, "epoch": 238, "lr": 9.856731974455153e-05}
{"train_loss": 0.47730308771133423, "global_step": 21196, "epoch": 238, "lr": 9.856718195632234e-05}
{"train_loss": 0.5636986494064331, "global_step": 21197, "epoch": 238, "lr": 9.856704416156388e-05}
{"train_loss": 0.3518837094306946, "global_step": 21198, "epoch": 238, "lr": 9.856690636027617e-05}
{"train_loss": 0.48107266426086426, "global_step": 21199, "epoch": 238, "lr": 9.856676855245923e-05}
{"train_loss": 0.423534095287323, "global_step": 21200, "epoch": 238, "lr": 9.856663073811308e-05}
{"train_loss": 0.47255223989486694, "global_step": 21201, "epoch": 238, "lr": 9.856649291723774e-05}
{"train_loss": 0.3915902376174927, "global_step": 21202, "epoch": 238, "lr": 9.856635508983323e-05}
{"train_loss": 0.4705178141593933, "global_step": 21203, "epoch": 238, "lr": 9.856621725589957e-05}
{"train_loss": 0.5054382085800171, "global_step": 21204, "epoch": 238, "lr": 9.856607941543677e-05}
{"train_loss": 0.32229456305503845, "global_step": 21205, "epoch": 238, "lr": 9.856594156844485e-05}
{"train_loss": 0.44660186767578125, "global_step": 21206, "epoch": 238, "lr": 9.856580371492383e-05}
{"train_loss": 0.37699684500694275, "global_step": 21207, "epoch": 238, "lr": 9.856566585487372e-05}
{"train_loss": 0.515279233455658, "global_step": 21208, "epoch": 238, "lr": 9.856552798829457e-05}
{"train_loss": 0.4976719617843628, "global_step": 21209, "epoch": 238, "lr": 9.856539011518636e-05}
{"train_loss": 0.39991459250450134, "global_step": 21210, "epoch": 238, "lr": 9.856525223554913e-05}
{"train_loss": 0.36728435754776, "global_step": 21211, "epoch": 238, "lr": 9.85651143493829e-05}
{"train_loss": 0.4161054790019989, "global_step": 21212, "epoch": 238, "lr": 9.856497645668767e-05}
{"train_loss": 0.43691280484199524, "global_step": 21213, "epoch": 238, "lr": 9.856483855746348e-05}
{"train_loss": 0.4205251932144165, "global_step": 21214, "epoch": 238, "lr": 9.856470065171034e-05}
{"train_loss": 0.36589163541793823, "global_step": 21215, "epoch": 238, "lr": 9.856456273942827e-05}
{"train_loss": 0.43563762307167053, "global_step": 21216, "epoch": 238, "lr": 9.856442482061727e-05}
{"train_loss": 0.4775394797325134, "global_step": 21217, "epoch": 238, "lr": 9.856428689527739e-05}
{"train_loss": 0.4578179121017456, "global_step": 21218, "epoch": 238, "lr": 9.856414896340863e-05}
{"train_loss": 0.4204370975494385, "global_step": 21219, "epoch": 238, "lr": 9.856401102501101e-05}
{"train_loss": 0.36689886450767517, "global_step": 21220, "epoch": 238, "lr": 9.856387308008456e-05}
{"train_loss": 0.444938063621521, "global_step": 21221, "epoch": 238, "lr": 9.856373512862929e-05}
{"train_loss": 0.4950254261493683, "global_step": 21222, "epoch": 238, "lr": 9.856359717064521e-05}
{"train_loss": 0.4652479588985443, "global_step": 21223, "epoch": 238, "lr": 9.856345920613233e-05}
{"train_loss": 0.585060715675354, "global_step": 21224, "epoch": 238, "lr": 9.856332123509072e-05}
{"train_loss": 0.5497215986251831, "global_step": 21225, "epoch": 238, "lr": 9.856318325752034e-05}
{"train_loss": 0.5275903940200806, "global_step": 21226, "epoch": 238, "lr": 9.856304527342123e-05}
{"train_loss": 0.48266664147377014, "global_step": 21227, "epoch": 238, "lr": 9.856290728279342e-05}
{"train_loss": 0.4447489082813263, "global_step": 21228, "epoch": 238, "lr": 9.85627692856369e-05}
{"train_loss": 0.6599552631378174, "global_step": 21229, "epoch": 238, "lr": 9.856263128195173e-05}
{"train_loss": 0.5459478497505188, "global_step": 21230, "epoch": 238, "lr": 9.85624932717379e-05}
{"train_loss": 0.49734577536582947, "global_step": 21231, "epoch": 238, "lr": 9.856235525499545e-05}
{"train_loss": 0.40508899092674255, "global_step": 21232, "epoch": 238, "lr": 9.856221723172436e-05}
{"train_loss": 0.3831745386123657, "global_step": 21233, "epoch": 238, "lr": 9.856207920192466e-05}
{"train_loss": 0.4199984073638916, "global_step": 21234, "epoch": 238, "lr": 9.85619411655964e-05}
{"train_loss": 0.4998769164085388, "global_step": 21235, "epoch": 238, "lr": 9.85618031227396e-05}
{"train_loss": 0.4144833981990814, "global_step": 21236, "epoch": 238, "lr": 9.856166507335423e-05}
{"train_loss": 0.3295479118824005, "global_step": 21237, "epoch": 238, "lr": 9.856152701744034e-05}
{"train_loss": 0.4597574770450592, "global_step": 21238, "epoch": 238, "lr": 9.856138895499794e-05}
{"train_loss": 0.3464754521846771, "global_step": 21239, "epoch": 238, "lr": 9.856125088602706e-05}
{"train_loss": 0.4445168673992157, "global_step": 21240, "epoch": 238, "lr": 9.856111281052772e-05}
{"train_loss": 0.40962132811546326, "global_step": 21241, "epoch": 238, "lr": 9.85609747284999e-05}
{"train_loss": 0.3973951041698456, "global_step": 21242, "epoch": 238, "lr": 9.856083663994369e-05}
{"train_loss": 0.36861133575439453, "global_step": 21243, "epoch": 238, "lr": 9.856069854485903e-05}
{"train_loss": 0.44294771552085876, "global_step": 21244, "epoch": 238, "lr": 9.8560560443246e-05}
{"train_loss": 0.5032371878623962, "global_step": 21245, "epoch": 238, "lr": 9.856042233510458e-05}
{"train_loss": 0.4418245255947113, "global_step": 21246, "epoch": 238, "lr": 9.856028422043482e-05}
{"train_loss": 0.534421443939209, "global_step": 21247, "epoch": 238, "lr": 9.856014609923672e-05}
{"train_loss": 0.4460025429725647, "global_step": 21248, "epoch": 238, "lr": 9.856000797151028e-05}
{"train_loss": 0.39137089252471924, "global_step": 21249, "epoch": 238, "lr": 9.855986983725556e-05}
{"train_loss": 0.6152991056442261, "global_step": 21250, "epoch": 238, "lr": 9.855973169647255e-05}
{"train_loss": 0.421957403421402, "global_step": 21251, "epoch": 238, "lr": 9.855959354916127e-05}
{"train_loss": 0.4014037251472473, "global_step": 21252, "epoch": 238, "lr": 9.855945539532176e-05}
{"train_loss": 0.4190754294395447, "global_step": 21253, "epoch": 238, "lr": 9.8559317234954e-05}
{"train_loss": 0.4943278133869171, "global_step": 21254, "epoch": 238, "lr": 9.855917906805805e-05}
{"train_loss": 0.37503373622894287, "global_step": 21255, "epoch": 238, "lr": 9.85590408946339e-05}
{"train_loss": 0.4612007439136505, "global_step": 21256, "epoch": 238, "lr": 9.855890271468158e-05}
{"train_loss": 0.5466690063476562, "global_step": 21257, "epoch": 238, "lr": 9.855876452820112e-05}
{"train_loss": 0.5568209886550903, "global_step": 21258, "epoch": 238, "lr": 9.85586263351925e-05}
{"train_loss": 0.4173198342323303, "global_step": 21259, "epoch": 238, "lr": 9.85584881356558e-05}
{"train_loss": 0.4724627137184143, "global_step": 21260, "epoch": 238, "lr": 9.855834992959097e-05}
{"train_loss": 0.4601794183254242, "global_step": 21261, "epoch": 238, "lr": 9.855821171699809e-05}
{"train_loss": 0.37836727499961853, "global_step": 21262, "epoch": 238, "lr": 9.855807349787714e-05}
{"train_loss": 0.5784996151924133, "global_step": 21263, "epoch": 238, "lr": 9.855793527222814e-05}
{"train_loss": 0.44073545932769775, "global_step": 21264, "epoch": 238, "lr": 9.855779704005113e-05}
{"train_loss": 0.40079036355018616, "global_step": 21265, "epoch": 238, "lr": 9.85576588013461e-05}
{"train_loss": 0.44191619753837585, "global_step": 21266, "epoch": 238, "lr": 9.855752055611309e-05}
{"train_loss": 0.3542364537715912, "global_step": 21267, "epoch": 238, "lr": 9.855738230435211e-05}
{"train_loss": 0.32386091351509094, "global_step": 21268, "epoch": 238, "lr": 9.85572440460632e-05}
{"train_loss": 0.4597480297088623, "global_step": 21269, "epoch": 238, "lr": 9.855710578124635e-05}
{"train_loss": 0.4539306853594405, "global_step": 21270, "epoch": 238, "lr": 9.855696750990158e-05, "val_loss": 1.8348395824432373}
{"train_loss": 0.46467623114585876, "global_step": 21271, "epoch": 239, "lr": 9.855682923202892e-05}
{"train_loss": 0.6115965843200684, "global_step": 21272, "epoch": 239, "lr": 9.85566909476284e-05}
{"train_loss": 0.4224071204662323, "global_step": 21273, "epoch": 239, "lr": 9.855655265670001e-05}
{"train_loss": 0.2941695749759674, "global_step": 21274, "epoch": 239, "lr": 9.855641435924379e-05}
{"train_loss": 0.4336647093296051, "global_step": 21275, "epoch": 239, "lr": 9.855627605525974e-05}
{"train_loss": 0.4209979176521301, "global_step": 21276, "epoch": 239, "lr": 9.855613774474791e-05}
{"train_loss": 0.47050875425338745, "global_step": 21277, "epoch": 239, "lr": 9.85559994277083e-05}
{"train_loss": 0.39255356788635254, "global_step": 21278, "epoch": 239, "lr": 9.855586110414091e-05}
{"train_loss": 0.39166057109832764, "global_step": 21279, "epoch": 239, "lr": 9.85557227740458e-05}
{"train_loss": 0.5005677938461304, "global_step": 21280, "epoch": 239, "lr": 9.855558443742294e-05}
{"train_loss": 0.48087063431739807, "global_step": 21281, "epoch": 239, "lr": 9.855544609427238e-05}
{"train_loss": 0.4578230381011963, "global_step": 21282, "epoch": 239, "lr": 9.855530774459414e-05}
{"train_loss": 0.5219627022743225, "global_step": 21283, "epoch": 239, "lr": 9.855516938838822e-05}
{"train_loss": 0.34469032287597656, "global_step": 21284, "epoch": 239, "lr": 9.855503102565466e-05}
{"train_loss": 0.46924176812171936, "global_step": 21285, "epoch": 239, "lr": 9.855489265639346e-05}
{"train_loss": 0.39591184258461, "global_step": 21286, "epoch": 239, "lr": 9.855475428060466e-05}
{"train_loss": 0.49391886591911316, "global_step": 21287, "epoch": 239, "lr": 9.855461589828825e-05}
{"train_loss": 0.3581191599369049, "global_step": 21288, "epoch": 239, "lr": 9.855447750944426e-05}
{"train_loss": 0.4333759546279907, "global_step": 21289, "epoch": 239, "lr": 9.855433911407272e-05}
{"train_loss": 0.49663636088371277, "global_step": 21290, "epoch": 239, "lr": 9.855420071217365e-05}
{"train_loss": 0.5168482065200806, "global_step": 21291, "epoch": 239, "lr": 9.855406230374705e-05}
{"train_loss": 0.45334774255752563, "global_step": 21292, "epoch": 239, "lr": 9.855392388879295e-05}
{"train_loss": 0.517163872718811, "global_step": 21293, "epoch": 239, "lr": 9.855378546731135e-05}
{"train_loss": 0.44481751322746277, "global_step": 21294, "epoch": 239, "lr": 9.855364703930231e-05}
{"train_loss": 0.6428800225257874, "global_step": 21295, "epoch": 239, "lr": 9.855350860476582e-05}
{"train_loss": 0.4029504954814911, "global_step": 21296, "epoch": 239, "lr": 9.85533701637019e-05}
{"train_loss": 0.392531156539917, "global_step": 21297, "epoch": 239, "lr": 9.855323171611056e-05}
{"train_loss": 0.45057857036590576, "global_step": 21298, "epoch": 239, "lr": 9.855309326199184e-05}
{"train_loss": 0.48733535408973694, "global_step": 21299, "epoch": 239, "lr": 9.855295480134575e-05}
{"train_loss": 0.5001218914985657, "global_step": 21300, "epoch": 239, "lr": 9.85528163341723e-05}
{"train_loss": 0.3809352517127991, "global_step": 21301, "epoch": 239, "lr": 9.855267786047152e-05}
{"train_loss": 0.37231582403182983, "global_step": 21302, "epoch": 239, "lr": 9.855253938024342e-05}
{"train_loss": 0.37173518538475037, "global_step": 21303, "epoch": 239, "lr": 9.855240089348803e-05}
{"train_loss": 0.5843655467033386, "global_step": 21304, "epoch": 239, "lr": 9.855226240020537e-05}
{"train_loss": 0.5091614723205566, "global_step": 21305, "epoch": 239, "lr": 9.855212390039544e-05}
{"train_loss": 0.4311694800853729, "global_step": 21306, "epoch": 239, "lr": 9.855198539405826e-05}
{"train_loss": 0.44362369179725647, "global_step": 21307, "epoch": 239, "lr": 9.855184688119386e-05}
{"train_loss": 0.4804922342300415, "global_step": 21308, "epoch": 239, "lr": 9.855170836180225e-05}
{"train_loss": 0.47926783561706543, "global_step": 21309, "epoch": 239, "lr": 9.855156983588347e-05}
{"train_loss": 0.5122899413108826, "global_step": 21310, "epoch": 239, "lr": 9.855143130343752e-05}
{"train_loss": 0.5276782512664795, "global_step": 21311, "epoch": 239, "lr": 9.855129276446442e-05}
{"train_loss": 0.46031004190444946, "global_step": 21312, "epoch": 239, "lr": 9.855115421896419e-05}
{"train_loss": 0.5416358113288879, "global_step": 21313, "epoch": 239, "lr": 9.855101566693685e-05}
{"train_loss": 0.43725118041038513, "global_step": 21314, "epoch": 239, "lr": 9.855087710838241e-05}
{"train_loss": 0.5301319360733032, "global_step": 21315, "epoch": 239, "lr": 9.855073854330091e-05}
{"train_loss": 0.46501824259757996, "global_step": 21316, "epoch": 239, "lr": 9.855059997169235e-05}
{"train_loss": 0.627694308757782, "global_step": 21317, "epoch": 239, "lr": 9.855046139355675e-05}
{"train_loss": 0.49562713503837585, "global_step": 21318, "epoch": 239, "lr": 9.855032280889414e-05}
{"train_loss": 0.46888405084609985, "global_step": 21319, "epoch": 239, "lr": 9.855018421770452e-05}
{"train_loss": 0.5094795227050781, "global_step": 21320, "epoch": 239, "lr": 9.855004561998791e-05}
{"train_loss": 0.5011715888977051, "global_step": 21321, "epoch": 239, "lr": 9.854990701574435e-05}
{"train_loss": 0.39597418904304504, "global_step": 21322, "epoch": 239, "lr": 9.854976840497387e-05}
{"train_loss": 0.45229029655456543, "global_step": 21323, "epoch": 239, "lr": 9.854962978767643e-05}
{"train_loss": 0.49287885427474976, "global_step": 21324, "epoch": 239, "lr": 9.85494911638521e-05}
{"train_loss": 0.6172149777412415, "global_step": 21325, "epoch": 239, "lr": 9.854935253350088e-05}
{"train_loss": 0.48655468225479126, "global_step": 21326, "epoch": 239, "lr": 9.85492138966228e-05}
{"train_loss": 0.38230207562446594, "global_step": 21327, "epoch": 239, "lr": 9.854907525321786e-05}
{"train_loss": 0.444454163312912, "global_step": 21328, "epoch": 239, "lr": 9.854893660328609e-05}
{"train_loss": 0.4899556338787079, "global_step": 21329, "epoch": 239, "lr": 9.854879794682752e-05}
{"train_loss": 0.5448383092880249, "global_step": 21330, "epoch": 239, "lr": 9.854865928384214e-05}
{"train_loss": 0.4875051975250244, "global_step": 21331, "epoch": 239, "lr": 9.854852061432997e-05}
{"train_loss": 0.5064741373062134, "global_step": 21332, "epoch": 239, "lr": 9.854838193829109e-05}
{"train_loss": 0.4167172908782959, "global_step": 21333, "epoch": 239, "lr": 9.854824325572542e-05}
{"train_loss": 0.4434378743171692, "global_step": 21334, "epoch": 239, "lr": 9.854810456663305e-05}
{"train_loss": 0.4579262435436249, "global_step": 21335, "epoch": 239, "lr": 9.854796587101399e-05}
{"train_loss": 0.3340489864349365, "global_step": 21336, "epoch": 239, "lr": 9.854782716886825e-05}
{"train_loss": 0.4798223376274109, "global_step": 21337, "epoch": 239, "lr": 9.854768846019582e-05}
{"train_loss": 0.3587951958179474, "global_step": 21338, "epoch": 239, "lr": 9.854754974499678e-05}
{"train_loss": 0.5497005581855774, "global_step": 21339, "epoch": 239, "lr": 9.85474110232711e-05}
{"train_loss": 0.5106509923934937, "global_step": 21340, "epoch": 239, "lr": 9.854727229501878e-05}
{"train_loss": 0.47890788316726685, "global_step": 21341, "epoch": 239, "lr": 9.854713356023992e-05}
{"train_loss": 0.5262896418571472, "global_step": 21342, "epoch": 239, "lr": 9.854699481893445e-05}
{"train_loss": 0.3995867073535919, "global_step": 21343, "epoch": 239, "lr": 9.854685607110245e-05}
{"train_loss": 0.5740458369255066, "global_step": 21344, "epoch": 239, "lr": 9.854671731674391e-05}
{"train_loss": 0.5844185948371887, "global_step": 21345, "epoch": 239, "lr": 9.854657855585888e-05}
{"train_loss": 0.4375515878200531, "global_step": 21346, "epoch": 239, "lr": 9.854643978844733e-05}
{"train_loss": 0.47712260484695435, "global_step": 21347, "epoch": 239, "lr": 9.85463010145093e-05}
{"train_loss": 0.4862500727176666, "global_step": 21348, "epoch": 239, "lr": 9.854616223404483e-05}
{"train_loss": 0.4671255350112915, "global_step": 21349, "epoch": 239, "lr": 9.85460234470539e-05}
{"train_loss": 0.43893906474113464, "global_step": 21350, "epoch": 239, "lr": 9.854588465353656e-05}
{"train_loss": 0.5348222255706787, "global_step": 21351, "epoch": 239, "lr": 9.85457458534928e-05}
{"train_loss": 0.5539348721504211, "global_step": 21352, "epoch": 239, "lr": 9.854560704692268e-05}
{"train_loss": 0.48526039719581604, "global_step": 21353, "epoch": 239, "lr": 9.854546823382618e-05}
{"train_loss": 0.47037366032600403, "global_step": 21354, "epoch": 239, "lr": 9.854532941420335e-05}
{"train_loss": 0.5607752799987793, "global_step": 21355, "epoch": 239, "lr": 9.854519058805418e-05}
{"train_loss": 0.6216987371444702, "global_step": 21356, "epoch": 239, "lr": 9.85450517553787e-05}
{"train_loss": 0.5099572539329529, "global_step": 21357, "epoch": 239, "lr": 9.854491291617694e-05}
{"train_loss": 0.5978541970252991, "global_step": 21358, "epoch": 239, "lr": 9.854477407044889e-05}
{"train_loss": 0.4757433195462388, "global_step": 21359, "epoch": 239, "lr": 9.85446352181946e-05, "val_loss": 1.8506096601486206}
{"train_loss": 0.46578362584114075, "global_step": 21360, "epoch": 240, "lr": 9.854449635941407e-05}
{"train_loss": 0.5750604271888733, "global_step": 21361, "epoch": 240, "lr": 9.854435749410733e-05}
{"train_loss": 0.564423680305481, "global_step": 21362, "epoch": 240, "lr": 9.854421862227439e-05}
{"train_loss": 0.5476251244544983, "global_step": 21363, "epoch": 240, "lr": 9.854407974391526e-05}
{"train_loss": 0.5063236355781555, "global_step": 21364, "epoch": 240, "lr": 9.854394085902998e-05}
{"train_loss": 0.5394768118858337, "global_step": 21365, "epoch": 240, "lr": 9.854380196761857e-05}
{"train_loss": 0.5077259540557861, "global_step": 21366, "epoch": 240, "lr": 9.854366306968101e-05}
{"train_loss": 0.452960342168808, "global_step": 21367, "epoch": 240, "lr": 9.854352416521737e-05}
{"train_loss": 0.5165056586265564, "global_step": 21368, "epoch": 240, "lr": 9.854338525422764e-05}
{"train_loss": 0.48244088888168335, "global_step": 21369, "epoch": 240, "lr": 9.854324633671183e-05}
{"train_loss": 0.49120020866394043, "global_step": 21370, "epoch": 240, "lr": 9.854310741267e-05}
{"train_loss": 0.5994303822517395, "global_step": 21371, "epoch": 240, "lr": 9.854296848210211e-05}
{"train_loss": 0.7276936173439026, "global_step": 21372, "epoch": 240, "lr": 9.854282954500823e-05}
{"train_loss": 0.6431517601013184, "global_step": 21373, "epoch": 240, "lr": 9.854269060138834e-05}
{"train_loss": 0.3984810411930084, "global_step": 21374, "epoch": 240, "lr": 9.85425516512425e-05}
{"train_loss": 0.5537097454071045, "global_step": 21375, "epoch": 240, "lr": 9.854241269457069e-05}
{"train_loss": 0.69136643409729, "global_step": 21376, "epoch": 240, "lr": 9.854227373137294e-05}
{"train_loss": 0.5564960241317749, "global_step": 21377, "epoch": 240, "lr": 9.854213476164929e-05}
{"train_loss": 0.5302221179008484, "global_step": 21378, "epoch": 240, "lr": 9.854199578539973e-05}
{"train_loss": 0.5599872469902039, "global_step": 21379, "epoch": 240, "lr": 9.85418568026243e-05}
{"train_loss": 0.5614699721336365, "global_step": 21380, "epoch": 240, "lr": 9.8541717813323e-05}
{"train_loss": 0.569861888885498, "global_step": 21381, "epoch": 240, "lr": 9.854157881749586e-05}
{"train_loss": 0.4622303247451782, "global_step": 21382, "epoch": 240, "lr": 9.85414398151429e-05}
{"train_loss": 0.4039435088634491, "global_step": 21383, "epoch": 240, "lr": 9.854130080626414e-05}
{"train_loss": 0.5274902582168579, "global_step": 21384, "epoch": 240, "lr": 9.854116179085958e-05}
{"train_loss": 0.5579413771629333, "global_step": 21385, "epoch": 240, "lr": 9.854102276892924e-05}
{"train_loss": 0.42467114329338074, "global_step": 21386, "epoch": 240, "lr": 9.854088374047318e-05}
{"train_loss": 0.3919140696525574, "global_step": 21387, "epoch": 240, "lr": 9.854074470549137e-05}
{"train_loss": 0.5110835433006287, "global_step": 21388, "epoch": 240, "lr": 9.854060566398387e-05}
{"train_loss": 0.48737117648124695, "global_step": 21389, "epoch": 240, "lr": 9.854046661595066e-05}
{"train_loss": 0.5153675675392151, "global_step": 21390, "epoch": 240, "lr": 9.854032756139178e-05}
{"train_loss": 0.5911996960639954, "global_step": 21391, "epoch": 240, "lr": 9.854018850030724e-05}
{"train_loss": 0.43893638253211975, "global_step": 21392, "epoch": 240, "lr": 9.854004943269707e-05}
{"train_loss": 0.4424671530723572, "global_step": 21393, "epoch": 240, "lr": 9.853991035856128e-05}
{"train_loss": 0.5628073811531067, "global_step": 21394, "epoch": 240, "lr": 9.853977127789989e-05}
{"train_loss": 0.5804052352905273, "global_step": 21395, "epoch": 240, "lr": 9.853963219071292e-05}
{"train_loss": 0.576423168182373, "global_step": 21396, "epoch": 240, "lr": 9.853949309700039e-05}
{"train_loss": 0.36655059456825256, "global_step": 21397, "epoch": 240, "lr": 9.853935399676232e-05}
{"train_loss": 0.40208935737609863, "global_step": 21398, "epoch": 240, "lr": 9.853921488999871e-05}
{"train_loss": 0.4792078733444214, "global_step": 21399, "epoch": 240, "lr": 9.853907577670961e-05}
{"train_loss": 0.4121512770652771, "global_step": 21400, "epoch": 240, "lr": 9.853893665689501e-05}
{"train_loss": 0.3899568021297455, "global_step": 21401, "epoch": 240, "lr": 9.853879753055495e-05}
{"train_loss": 0.31160885095596313, "global_step": 21402, "epoch": 240, "lr": 9.853865839768946e-05}
{"train_loss": 0.41232016682624817, "global_step": 21403, "epoch": 240, "lr": 9.85385192582985e-05}
{"train_loss": 0.4639648199081421, "global_step": 21404, "epoch": 240, "lr": 9.853838011238215e-05}
{"train_loss": 0.47871169447898865, "global_step": 21405, "epoch": 240, "lr": 9.85382409599404e-05}
{"train_loss": 0.41662946343421936, "global_step": 21406, "epoch": 240, "lr": 9.853810180097329e-05}
{"train_loss": 0.6014182567596436, "global_step": 21407, "epoch": 240, "lr": 9.85379626354808e-05}
{"train_loss": 0.4190593957901001, "global_step": 21408, "epoch": 240, "lr": 9.853782346346298e-05}
{"train_loss": 0.49587032198905945, "global_step": 21409, "epoch": 240, "lr": 9.853768428491985e-05}
{"train_loss": 0.5629478096961975, "global_step": 21410, "epoch": 240, "lr": 9.853754509985143e-05}
{"train_loss": 0.43969082832336426, "global_step": 21411, "epoch": 240, "lr": 9.853740590825771e-05}
{"train_loss": 0.5594363808631897, "global_step": 21412, "epoch": 240, "lr": 9.853726671013872e-05}
{"train_loss": 0.4617105722427368, "global_step": 21413, "epoch": 240, "lr": 9.85371275054945e-05}
{"train_loss": 0.429423451423645, "global_step": 21414, "epoch": 240, "lr": 9.853698829432505e-05}
{"train_loss": 0.32773643732070923, "global_step": 21415, "epoch": 240, "lr": 9.85368490766304e-05}
{"train_loss": 0.46343380212783813, "global_step": 21416, "epoch": 240, "lr": 9.853670985241054e-05}
{"train_loss": 0.6244874000549316, "global_step": 21417, "epoch": 240, "lr": 9.853657062166555e-05}
{"train_loss": 0.3998248279094696, "global_step": 21418, "epoch": 240, "lr": 9.853643138439537e-05}
{"train_loss": 0.4482095241546631, "global_step": 21419, "epoch": 240, "lr": 9.853629214060009e-05}
{"train_loss": 0.4840007424354553, "global_step": 21420, "epoch": 240, "lr": 9.853615289027968e-05}
{"train_loss": 0.5697550177574158, "global_step": 21421, "epoch": 240, "lr": 9.853601363343418e-05}
{"train_loss": 0.4175322949886322, "global_step": 21422, "epoch": 240, "lr": 9.85358743700636e-05}
{"train_loss": 0.43248969316482544, "global_step": 21423, "epoch": 240, "lr": 9.853573510016797e-05}
{"train_loss": 0.4640309810638428, "global_step": 21424, "epoch": 240, "lr": 9.853559582374729e-05}
{"train_loss": 0.48227575421333313, "global_step": 21425, "epoch": 240, "lr": 9.85354565408016e-05}
{"train_loss": 0.4754699766635895, "global_step": 21426, "epoch": 240, "lr": 9.853531725133092e-05}
{"train_loss": 0.4026646912097931, "global_step": 21427, "epoch": 240, "lr": 9.853517795533525e-05}
{"train_loss": 0.3695673644542694, "global_step": 21428, "epoch": 240, "lr": 9.853503865281461e-05}
{"train_loss": 0.446828156709671, "global_step": 21429, "epoch": 240, "lr": 9.853489934376903e-05}
{"train_loss": 0.3610493242740631, "global_step": 21430, "epoch": 240, "lr": 9.853476002819853e-05}
{"train_loss": 0.5310708284378052, "global_step": 21431, "epoch": 240, "lr": 9.853462070610311e-05}
{"train_loss": 0.45207613706588745, "global_step": 21432, "epoch": 240, "lr": 9.853448137748281e-05}
{"train_loss": 0.5132732391357422, "global_step": 21433, "epoch": 240, "lr": 9.853434204233765e-05}
{"train_loss": 0.5620871186256409, "global_step": 21434, "epoch": 240, "lr": 9.853420270066763e-05}
{"train_loss": 0.577518105506897, "global_step": 21435, "epoch": 240, "lr": 9.853406335247277e-05}
{"train_loss": 0.44307661056518555, "global_step": 21436, "epoch": 240, "lr": 9.85339239977531e-05}
{"train_loss": 0.41720956563949585, "global_step": 21437, "epoch": 240, "lr": 9.853378463650864e-05}
{"train_loss": 0.5189520716667175, "global_step": 21438, "epoch": 240, "lr": 9.85336452687394e-05}
{"train_loss": 0.4898427426815033, "global_step": 21439, "epoch": 240, "lr": 9.853350589444541e-05}
{"train_loss": 0.45135292410850525, "global_step": 21440, "epoch": 240, "lr": 9.853336651362669e-05}
{"train_loss": 0.4220080077648163, "global_step": 21441, "epoch": 240, "lr": 9.853322712628324e-05}
{"train_loss": 0.4550043046474457, "global_step": 21442, "epoch": 240, "lr": 9.853308773241507e-05}
{"train_loss": 0.4944329559803009, "global_step": 21443, "epoch": 240, "lr": 9.853294833202224e-05}
{"train_loss": 0.4010022282600403, "global_step": 21444, "epoch": 240, "lr": 9.853280892510474e-05}
{"train_loss": 0.4812908470630646, "global_step": 21445, "epoch": 240, "lr": 9.85326695116626e-05}
{"train_loss": 0.524368405342102, "global_step": 21446, "epoch": 240, "lr": 9.853253009169582e-05}
{"train_loss": 0.4686931371688843, "global_step": 21447, "epoch": 240, "lr": 9.853239066520445e-05}
{"train_loss": 0.48656078068058145, "global_step": 21448, "epoch": 240, "lr": 9.853225123218849e-05, "val_loss": 1.8548119068145752, "train_action_mse_error": 21.455013275146484}
{"train_loss": 0.4247502088546753, "global_step": 21449, "epoch": 241, "lr": 9.853211179264796e-05}
{"train_loss": 0.43256086111068726, "global_step": 21450, "epoch": 241, "lr": 9.853197234658287e-05}
{"train_loss": 0.5551238656044006, "global_step": 21451, "epoch": 241, "lr": 9.853183289399326e-05}
{"train_loss": 0.4859703481197357, "global_step": 21452, "epoch": 241, "lr": 9.853169343487912e-05}
{"train_loss": 0.465285062789917, "global_step": 21453, "epoch": 241, "lr": 9.85315539692405e-05}
{"train_loss": 0.5087088346481323, "global_step": 21454, "epoch": 241, "lr": 9.85314144970774e-05}
{"train_loss": 0.5266141891479492, "global_step": 21455, "epoch": 241, "lr": 9.853127501838985e-05}
{"train_loss": 0.5114946961402893, "global_step": 21456, "epoch": 241, "lr": 9.853113553317786e-05}
{"train_loss": 0.40223708748817444, "global_step": 21457, "epoch": 241, "lr": 9.853099604144143e-05}
{"train_loss": 0.3795662522315979, "global_step": 21458, "epoch": 241, "lr": 9.853085654318061e-05}
{"train_loss": 0.4416220188140869, "global_step": 21459, "epoch": 241, "lr": 9.853071703839541e-05}
{"train_loss": 0.44467654824256897, "global_step": 21460, "epoch": 241, "lr": 9.853057752708585e-05}
{"train_loss": 0.40299177169799805, "global_step": 21461, "epoch": 241, "lr": 9.853043800925195e-05}
{"train_loss": 0.3765997290611267, "global_step": 21462, "epoch": 241, "lr": 9.853029848489371e-05}
{"train_loss": 0.4296284019947052, "global_step": 21463, "epoch": 241, "lr": 9.853015895401116e-05}
{"train_loss": 0.5263928174972534, "global_step": 21464, "epoch": 241, "lr": 9.853001941660434e-05}
{"train_loss": 0.4322718679904938, "global_step": 21465, "epoch": 241, "lr": 9.852987987267325e-05}
{"train_loss": 0.38744574785232544, "global_step": 21466, "epoch": 241, "lr": 9.85297403222179e-05}
{"train_loss": 0.5232235193252563, "global_step": 21467, "epoch": 241, "lr": 9.852960076523832e-05}
{"train_loss": 0.39416617155075073, "global_step": 21468, "epoch": 241, "lr": 9.852946120173452e-05}
{"train_loss": 0.4671705961227417, "global_step": 21469, "epoch": 241, "lr": 9.852932163170654e-05}
{"train_loss": 0.5016157627105713, "global_step": 21470, "epoch": 241, "lr": 9.852918205515436e-05}
{"train_loss": 0.5029165744781494, "global_step": 21471, "epoch": 241, "lr": 9.852904247207805e-05}
{"train_loss": 0.4199048578739166, "global_step": 21472, "epoch": 241, "lr": 9.852890288247758e-05}
{"train_loss": 0.4531303942203522, "global_step": 21473, "epoch": 241, "lr": 9.852876328635302e-05}
{"train_loss": 0.43720388412475586, "global_step": 21474, "epoch": 241, "lr": 9.852862368370432e-05}
{"train_loss": 0.4522288143634796, "global_step": 21475, "epoch": 241, "lr": 9.852848407453157e-05}
{"train_loss": 0.47235071659088135, "global_step": 21476, "epoch": 241, "lr": 9.852834445883475e-05}
{"train_loss": 0.44720494747161865, "global_step": 21477, "epoch": 241, "lr": 9.852820483661387e-05}
{"train_loss": 0.47931748628616333, "global_step": 21478, "epoch": 241, "lr": 9.852806520786897e-05}
{"train_loss": 0.61588054895401, "global_step": 21479, "epoch": 241, "lr": 9.852792557260006e-05}
{"train_loss": 0.4776810109615326, "global_step": 21480, "epoch": 241, "lr": 9.852778593080718e-05}
{"train_loss": 0.399522066116333, "global_step": 21481, "epoch": 241, "lr": 9.852764628249032e-05}
{"train_loss": 0.5336103439331055, "global_step": 21482, "epoch": 241, "lr": 9.852750662764951e-05}
{"train_loss": 0.5934033989906311, "global_step": 21483, "epoch": 241, "lr": 9.852736696628478e-05}
{"train_loss": 0.48255592584609985, "global_step": 21484, "epoch": 241, "lr": 9.852722729839612e-05}
{"train_loss": 0.3624320328235626, "global_step": 21485, "epoch": 241, "lr": 9.852708762398358e-05}
{"train_loss": 0.4619031846523285, "global_step": 21486, "epoch": 241, "lr": 9.852694794304714e-05}
{"train_loss": 0.40895190834999084, "global_step": 21487, "epoch": 241, "lr": 9.852680825558686e-05}
{"train_loss": 0.46861928701400757, "global_step": 21488, "epoch": 241, "lr": 9.852666856160272e-05}
{"train_loss": 0.4314226508140564, "global_step": 21489, "epoch": 241, "lr": 9.85265288610948e-05}
{"train_loss": 0.4210560619831085, "global_step": 21490, "epoch": 241, "lr": 9.852638915406306e-05}
{"train_loss": 0.511613667011261, "global_step": 21491, "epoch": 241, "lr": 9.852624944050754e-05}
{"train_loss": 0.4858582317829132, "global_step": 21492, "epoch": 241, "lr": 9.852610972042826e-05}
{"train_loss": 0.49732136726379395, "global_step": 21493, "epoch": 241, "lr": 9.852596999382523e-05}
{"train_loss": 0.37604016065597534, "global_step": 21494, "epoch": 241, "lr": 9.852583026069849e-05}
{"train_loss": 0.5026818513870239, "global_step": 21495, "epoch": 241, "lr": 9.852569052104803e-05}
{"train_loss": 0.432524174451828, "global_step": 21496, "epoch": 241, "lr": 9.852555077487387e-05}
{"train_loss": 0.4427187442779541, "global_step": 21497, "epoch": 241, "lr": 9.852541102217605e-05}
{"train_loss": 0.5008674263954163, "global_step": 21498, "epoch": 241, "lr": 9.852527126295459e-05}
{"train_loss": 0.39827993512153625, "global_step": 21499, "epoch": 241, "lr": 9.852513149720949e-05}
{"train_loss": 0.5698681473731995, "global_step": 21500, "epoch": 241, "lr": 9.852499172494078e-05}
{"train_loss": 0.4960222542285919, "global_step": 21501, "epoch": 241, "lr": 9.852485194614848e-05}
{"train_loss": 0.38662800192832947, "global_step": 21502, "epoch": 241, "lr": 9.85247121608326e-05}
{"train_loss": 0.39496931433677673, "global_step": 21503, "epoch": 241, "lr": 9.852457236899318e-05}
{"train_loss": 0.4121023118495941, "global_step": 21504, "epoch": 241, "lr": 9.85244325706302e-05}
{"train_loss": 0.4371209442615509, "global_step": 21505, "epoch": 241, "lr": 9.85242927657437e-05}
{"train_loss": 0.304485946893692, "global_step": 21506, "epoch": 241, "lr": 9.852415295433371e-05}
{"train_loss": 0.4373437166213989, "global_step": 21507, "epoch": 241, "lr": 9.852401313640025e-05}
{"train_loss": 0.4897680878639221, "global_step": 21508, "epoch": 241, "lr": 9.852387331194332e-05}
{"train_loss": 0.6233353614807129, "global_step": 21509, "epoch": 241, "lr": 9.852373348096293e-05}
{"train_loss": 0.4970623850822449, "global_step": 21510, "epoch": 241, "lr": 9.852359364345913e-05}
{"train_loss": 0.39881160855293274, "global_step": 21511, "epoch": 241, "lr": 9.852345379943193e-05}
{"train_loss": 0.5466668009757996, "global_step": 21512, "epoch": 241, "lr": 9.852331394888133e-05}
{"train_loss": 0.4970252811908722, "global_step": 21513, "epoch": 241, "lr": 9.852317409180738e-05}
{"train_loss": 0.4489627778530121, "global_step": 21514, "epoch": 241, "lr": 9.852303422821006e-05}
{"train_loss": 0.30257517099380493, "global_step": 21515, "epoch": 241, "lr": 9.852289435808942e-05}
{"train_loss": 0.5048493146896362, "global_step": 21516, "epoch": 241, "lr": 9.852275448144546e-05}
{"train_loss": 0.34565499424934387, "global_step": 21517, "epoch": 241, "lr": 9.852261459827822e-05}
{"train_loss": 0.4467160701751709, "global_step": 21518, "epoch": 241, "lr": 9.852247470858771e-05}
{"train_loss": 0.3012796640396118, "global_step": 21519, "epoch": 241, "lr": 9.852233481237393e-05}
{"train_loss": 0.3997840881347656, "global_step": 21520, "epoch": 241, "lr": 9.852219490963692e-05}
{"train_loss": 0.46976238489151, "global_step": 21521, "epoch": 241, "lr": 9.852205500037668e-05}
{"train_loss": 0.43247053027153015, "global_step": 21522, "epoch": 241, "lr": 9.852191508459325e-05}
{"train_loss": 0.4814620614051819, "global_step": 21523, "epoch": 241, "lr": 9.852177516228664e-05}
{"train_loss": 0.5192779898643494, "global_step": 21524, "epoch": 241, "lr": 9.852163523345687e-05}
{"train_loss": 0.4454750418663025, "global_step": 21525, "epoch": 241, "lr": 9.852149529810398e-05}
{"train_loss": 0.35730814933776855, "global_step": 21526, "epoch": 241, "lr": 9.852135535622792e-05}
{"train_loss": 0.538888692855835, "global_step": 21527, "epoch": 241, "lr": 9.85212154078288e-05}
{"train_loss": 0.4882301688194275, "global_step": 21528, "epoch": 241, "lr": 9.852107545290656e-05}
{"train_loss": 0.4353848695755005, "global_step": 21529, "epoch": 241, "lr": 9.852093549146126e-05}
{"train_loss": 0.5155789256095886, "global_step": 21530, "epoch": 241, "lr": 9.852079552349293e-05}
{"train_loss": 0.5056722164154053, "global_step": 21531, "epoch": 241, "lr": 9.852065554900156e-05}
{"train_loss": 0.3854302763938904, "global_step": 21532, "epoch": 241, "lr": 9.852051556798717e-05}
{"train_loss": 0.48023590445518494, "global_step": 21533, "epoch": 241, "lr": 9.85203755804498e-05}
{"train_loss": 0.46427857875823975, "global_step": 21534, "epoch": 241, "lr": 9.852023558638945e-05}
{"train_loss": 0.47218185663223267, "global_step": 21535, "epoch": 241, "lr": 9.852009558580616e-05}
{"train_loss": 0.45497047901153564, "global_step": 21536, "epoch": 241, "lr": 9.85199555786999e-05}
{"train_loss": 0.4575193959675478, "global_step": 21537, "epoch": 241, "lr": 9.851981556507076e-05, "val_loss": 1.8523095846176147}
{"train_loss": 0.4521777033805847, "global_step": 21538, "epoch": 242, "lr": 9.85196755449187e-05}
{"train_loss": 0.46518319845199585, "global_step": 21539, "epoch": 242, "lr": 9.851953551824377e-05}
{"train_loss": 0.4300132989883423, "global_step": 21540, "epoch": 242, "lr": 9.851939548504598e-05}
{"train_loss": 0.48082634806632996, "global_step": 21541, "epoch": 242, "lr": 9.851925544532534e-05}
{"train_loss": 0.48514237999916077, "global_step": 21542, "epoch": 242, "lr": 9.85191153990819e-05}
{"train_loss": 0.46855276823043823, "global_step": 21543, "epoch": 242, "lr": 9.851897534631563e-05}
{"train_loss": 0.462795227766037, "global_step": 21544, "epoch": 242, "lr": 9.851883528702659e-05}
{"train_loss": 0.40324267745018005, "global_step": 21545, "epoch": 242, "lr": 9.851869522121478e-05}
{"train_loss": 0.36280325055122375, "global_step": 21546, "epoch": 242, "lr": 9.851855514888022e-05}
{"train_loss": 0.5486783981323242, "global_step": 21547, "epoch": 242, "lr": 9.851841507002293e-05}
{"train_loss": 0.4396442174911499, "global_step": 21548, "epoch": 242, "lr": 9.851827498464295e-05}
{"train_loss": 0.5843617916107178, "global_step": 21549, "epoch": 242, "lr": 9.851813489274026e-05}
{"train_loss": 0.4726497530937195, "global_step": 21550, "epoch": 242, "lr": 9.85179947943149e-05}
{"train_loss": 0.47864001989364624, "global_step": 21551, "epoch": 242, "lr": 9.85178546893669e-05}
{"train_loss": 0.42274901270866394, "global_step": 21552, "epoch": 242, "lr": 9.851771457789625e-05}
{"train_loss": 0.42380839586257935, "global_step": 21553, "epoch": 242, "lr": 9.851757445990298e-05}
{"train_loss": 0.557390034198761, "global_step": 21554, "epoch": 242, "lr": 9.851743433538713e-05}
{"train_loss": 0.5057161450386047, "global_step": 21555, "epoch": 242, "lr": 9.85172942043487e-05}
{"train_loss": 0.4377322793006897, "global_step": 21556, "epoch": 242, "lr": 9.85171540667877e-05}
{"train_loss": 0.3914770483970642, "global_step": 21557, "epoch": 242, "lr": 9.851701392270415e-05}
{"train_loss": 0.5392355918884277, "global_step": 21558, "epoch": 242, "lr": 9.851687377209811e-05}
{"train_loss": 0.5056887269020081, "global_step": 21559, "epoch": 242, "lr": 9.851673361496956e-05}
{"train_loss": 0.44530928134918213, "global_step": 21560, "epoch": 242, "lr": 9.851659345131851e-05}
{"train_loss": 0.3321089744567871, "global_step": 21561, "epoch": 242, "lr": 9.8516453281145e-05}
{"train_loss": 0.40299129486083984, "global_step": 21562, "epoch": 242, "lr": 9.851631310444906e-05}
{"train_loss": 0.4253942370414734, "global_step": 21563, "epoch": 242, "lr": 9.851617292123068e-05}
{"train_loss": 0.4478500187397003, "global_step": 21564, "epoch": 242, "lr": 9.85160327314899e-05}
{"train_loss": 0.4246782064437866, "global_step": 21565, "epoch": 242, "lr": 9.851589253522672e-05}
{"train_loss": 0.4480401277542114, "global_step": 21566, "epoch": 242, "lr": 9.851575233244117e-05}
{"train_loss": 0.47339940071105957, "global_step": 21567, "epoch": 242, "lr": 9.851561212313327e-05}
{"train_loss": 0.4592372179031372, "global_step": 21568, "epoch": 242, "lr": 9.851547190730304e-05}
{"train_loss": 0.41517993807792664, "global_step": 21569, "epoch": 242, "lr": 9.85153316849505e-05}
{"train_loss": 0.38843053579330444, "global_step": 21570, "epoch": 242, "lr": 9.851519145607566e-05}
{"train_loss": 0.43613380193710327, "global_step": 21571, "epoch": 242, "lr": 9.851505122067854e-05}
{"train_loss": 0.499394029378891, "global_step": 21572, "epoch": 242, "lr": 9.851491097875918e-05}
{"train_loss": 0.442239910364151, "global_step": 21573, "epoch": 242, "lr": 9.851477073031757e-05}
{"train_loss": 0.5304065346717834, "global_step": 21574, "epoch": 242, "lr": 9.851463047535374e-05}
{"train_loss": 0.4130397439002991, "global_step": 21575, "epoch": 242, "lr": 9.851449021386771e-05}
{"train_loss": 0.417931467294693, "global_step": 21576, "epoch": 242, "lr": 9.851434994585948e-05}
{"train_loss": 0.4966046214103699, "global_step": 21577, "epoch": 242, "lr": 9.851420967132911e-05}
{"train_loss": 0.42470303177833557, "global_step": 21578, "epoch": 242, "lr": 9.851406939027659e-05}
{"train_loss": 0.38933461904525757, "global_step": 21579, "epoch": 242, "lr": 9.851392910270194e-05}
{"train_loss": 0.3858814537525177, "global_step": 21580, "epoch": 242, "lr": 9.85137888086052e-05}
{"train_loss": 0.34874048829078674, "global_step": 21581, "epoch": 242, "lr": 9.851364850798635e-05}
{"train_loss": 0.3513241410255432, "global_step": 21582, "epoch": 242, "lr": 9.851350820084544e-05}
{"train_loss": 0.4888944923877716, "global_step": 21583, "epoch": 242, "lr": 9.851336788718248e-05}
{"train_loss": 0.4560421109199524, "global_step": 21584, "epoch": 242, "lr": 9.85132275669975e-05}
{"train_loss": 0.34380730986595154, "global_step": 21585, "epoch": 242, "lr": 9.85130872402905e-05}
{"train_loss": 0.34062811732292175, "global_step": 21586, "epoch": 242, "lr": 9.85129469070615e-05}
{"train_loss": 0.4543325901031494, "global_step": 21587, "epoch": 242, "lr": 9.851280656731053e-05}
{"train_loss": 0.4242672622203827, "global_step": 21588, "epoch": 242, "lr": 9.851266622103761e-05}
{"train_loss": 0.5293561220169067, "global_step": 21589, "epoch": 242, "lr": 9.851252586824276e-05}
{"train_loss": 0.45106086134910583, "global_step": 21590, "epoch": 242, "lr": 9.851238550892596e-05}
{"train_loss": 0.5640692114830017, "global_step": 21591, "epoch": 242, "lr": 9.851224514308729e-05}
{"train_loss": 0.5422860383987427, "global_step": 21592, "epoch": 242, "lr": 9.851210477072673e-05}
{"train_loss": 0.4029020667076111, "global_step": 21593, "epoch": 242, "lr": 9.851196439184432e-05}
{"train_loss": 0.4032742977142334, "global_step": 21594, "epoch": 242, "lr": 9.851182400644006e-05}
{"train_loss": 0.5389129519462585, "global_step": 21595, "epoch": 242, "lr": 9.851168361451397e-05}
{"train_loss": 0.46281981468200684, "global_step": 21596, "epoch": 242, "lr": 9.851154321606609e-05}
{"train_loss": 0.41174477338790894, "global_step": 21597, "epoch": 242, "lr": 9.85114028110964e-05}
{"train_loss": 0.44386982917785645, "global_step": 21598, "epoch": 242, "lr": 9.851126239960495e-05}
{"train_loss": 0.4619044065475464, "global_step": 21599, "epoch": 242, "lr": 9.851112198159177e-05}
{"train_loss": 0.513218879699707, "global_step": 21600, "epoch": 242, "lr": 9.851098155705684e-05}
{"train_loss": 0.48520204424858093, "global_step": 21601, "epoch": 242, "lr": 9.851084112600022e-05}
{"train_loss": 0.4446142315864563, "global_step": 21602, "epoch": 242, "lr": 9.851070068842189e-05}
{"train_loss": 0.3754671812057495, "global_step": 21603, "epoch": 242, "lr": 9.85105602443219e-05}
{"train_loss": 0.4838847517967224, "global_step": 21604, "epoch": 242, "lr": 9.851041979370025e-05}
{"train_loss": 0.38963547348976135, "global_step": 21605, "epoch": 242, "lr": 9.851027933655698e-05}
{"train_loss": 0.39756956696510315, "global_step": 21606, "epoch": 242, "lr": 9.851013887289207e-05}
{"train_loss": 0.44882380962371826, "global_step": 21607, "epoch": 242, "lr": 9.850999840270558e-05}
{"train_loss": 0.43051770329475403, "global_step": 21608, "epoch": 242, "lr": 9.850985792599751e-05}
{"train_loss": 0.4967459440231323, "global_step": 21609, "epoch": 242, "lr": 9.850971744276788e-05}
{"train_loss": 0.4305230975151062, "global_step": 21610, "epoch": 242, "lr": 9.850957695301671e-05}
{"train_loss": 0.40603798627853394, "global_step": 21611, "epoch": 242, "lr": 9.850943645674403e-05}
{"train_loss": 0.5931784510612488, "global_step": 21612, "epoch": 242, "lr": 9.850929595394983e-05}
{"train_loss": 0.4675447940826416, "global_step": 21613, "epoch": 242, "lr": 9.850915544463414e-05}
{"train_loss": 0.3353622257709503, "global_step": 21614, "epoch": 242, "lr": 9.8509014928797e-05}
{"train_loss": 0.4409092366695404, "global_step": 21615, "epoch": 242, "lr": 9.850887440643841e-05}
{"train_loss": 0.49044981598854065, "global_step": 21616, "epoch": 242, "lr": 9.850873387755839e-05}
{"train_loss": 0.4126836061477661, "global_step": 21617, "epoch": 242, "lr": 9.850859334215697e-05}
{"train_loss": 0.44686073064804077, "global_step": 21618, "epoch": 242, "lr": 9.850845280023416e-05}
{"train_loss": 0.5419166684150696, "global_step": 21619, "epoch": 242, "lr": 9.850831225178996e-05}
{"train_loss": 0.43453866243362427, "global_step": 21620, "epoch": 242, "lr": 9.850817169682443e-05}
{"train_loss": 0.4417683184146881, "global_step": 21621, "epoch": 242, "lr": 9.850803113533757e-05}
{"train_loss": 0.4314461648464203, "global_step": 21622, "epoch": 242, "lr": 9.850789056732939e-05}
{"train_loss": 0.432233989238739, "global_step": 21623, "epoch": 242, "lr": 9.85077499927999e-05}
{"train_loss": 0.4496786594390869, "global_step": 21624, "epoch": 242, "lr": 9.850760941174914e-05}
{"train_loss": 0.36002683639526367, "global_step": 21625, "epoch": 242, "lr": 9.850746882417714e-05}
{"train_loss": 0.44912931758366276, "global_step": 21626, "epoch": 242, "lr": 9.85073282300839e-05, "val_loss": 1.8632885217666626}
{"train_loss": 0.5425843000411987, "global_step": 21627, "epoch": 243, "lr": 9.850718762946943e-05}
{"train_loss": 0.45615482330322266, "global_step": 21628, "epoch": 243, "lr": 9.850704702233376e-05}
{"train_loss": 0.4281851053237915, "global_step": 21629, "epoch": 243, "lr": 9.850690640867692e-05}
{"train_loss": 0.6480196118354797, "global_step": 21630, "epoch": 243, "lr": 9.85067657884989e-05}
{"train_loss": 0.34971433877944946, "global_step": 21631, "epoch": 243, "lr": 9.850662516179976e-05}
{"train_loss": 0.5086736679077148, "global_step": 21632, "epoch": 243, "lr": 9.850648452857949e-05}
{"train_loss": 0.5017552375793457, "global_step": 21633, "epoch": 243, "lr": 9.85063438888381e-05}
{"train_loss": 0.458627849817276, "global_step": 21634, "epoch": 243, "lr": 9.850620324257562e-05}
{"train_loss": 0.4537820816040039, "global_step": 21635, "epoch": 243, "lr": 9.85060625897921e-05}
{"train_loss": 0.5178163647651672, "global_step": 21636, "epoch": 243, "lr": 9.85059219304875e-05}
{"train_loss": 0.5588253140449524, "global_step": 21637, "epoch": 243, "lr": 9.850578126466188e-05}
{"train_loss": 0.4264131784439087, "global_step": 21638, "epoch": 243, "lr": 9.850564059231526e-05}
{"train_loss": 0.6026649475097656, "global_step": 21639, "epoch": 243, "lr": 9.850549991344764e-05}
{"train_loss": 0.3592279851436615, "global_step": 21640, "epoch": 243, "lr": 9.850535922805904e-05}
{"train_loss": 0.44932791590690613, "global_step": 21641, "epoch": 243, "lr": 9.85052185361495e-05}
{"train_loss": 0.48296046257019043, "global_step": 21642, "epoch": 243, "lr": 9.850507783771901e-05}
{"train_loss": 0.5149974226951599, "global_step": 21643, "epoch": 243, "lr": 9.850493713276762e-05}
{"train_loss": 0.501979410648346, "global_step": 21644, "epoch": 243, "lr": 9.85047964212953e-05}
{"train_loss": 0.4875830411911011, "global_step": 21645, "epoch": 243, "lr": 9.850465570330214e-05}
{"train_loss": 0.458079069852829, "global_step": 21646, "epoch": 243, "lr": 9.850451497878809e-05}
{"train_loss": 0.3655313551425934, "global_step": 21647, "epoch": 243, "lr": 9.850437424775323e-05}
{"train_loss": 0.4053577482700348, "global_step": 21648, "epoch": 243, "lr": 9.850423351019752e-05}
{"train_loss": 0.4653725326061249, "global_step": 21649, "epoch": 243, "lr": 9.850409276612103e-05}
{"train_loss": 0.4078047275543213, "global_step": 21650, "epoch": 243, "lr": 9.850395201552373e-05}
{"train_loss": 0.5040733814239502, "global_step": 21651, "epoch": 243, "lr": 9.850381125840568e-05}
{"train_loss": 0.4724017083644867, "global_step": 21652, "epoch": 243, "lr": 9.85036704947669e-05}
{"train_loss": 0.3923739790916443, "global_step": 21653, "epoch": 243, "lr": 9.850352972460737e-05}
{"train_loss": 0.6045095920562744, "global_step": 21654, "epoch": 243, "lr": 9.850338894792714e-05}
{"train_loss": 0.445421427488327, "global_step": 21655, "epoch": 243, "lr": 9.850324816472622e-05}
{"train_loss": 0.44899335503578186, "global_step": 21656, "epoch": 243, "lr": 9.850310737500463e-05}
{"train_loss": 0.4968516230583191, "global_step": 21657, "epoch": 243, "lr": 9.850296657876238e-05}
{"train_loss": 0.5215468406677246, "global_step": 21658, "epoch": 243, "lr": 9.85028257759995e-05}
{"train_loss": 0.43745699524879456, "global_step": 21659, "epoch": 243, "lr": 9.850268496671602e-05}
{"train_loss": 0.44776445627212524, "global_step": 21660, "epoch": 243, "lr": 9.850254415091194e-05}
{"train_loss": 0.4783726930618286, "global_step": 21661, "epoch": 243, "lr": 9.850240332858727e-05}
{"train_loss": 0.5850878953933716, "global_step": 21662, "epoch": 243, "lr": 9.850226249974205e-05}
{"train_loss": 0.3674972355365753, "global_step": 21663, "epoch": 243, "lr": 9.850212166437629e-05}
{"train_loss": 0.37161269783973694, "global_step": 21664, "epoch": 243, "lr": 9.850198082249001e-05}
{"train_loss": 0.5222165584564209, "global_step": 21665, "epoch": 243, "lr": 9.850183997408325e-05}
{"train_loss": 0.5241820812225342, "global_step": 21666, "epoch": 243, "lr": 9.850169911915598e-05}
{"train_loss": 0.5261117219924927, "global_step": 21667, "epoch": 243, "lr": 9.850155825770826e-05}
{"train_loss": 0.49325716495513916, "global_step": 21668, "epoch": 243, "lr": 9.85014173897401e-05}
{"train_loss": 0.5248301029205322, "global_step": 21669, "epoch": 243, "lr": 9.850127651525152e-05}
{"train_loss": 0.48391133546829224, "global_step": 21670, "epoch": 243, "lr": 9.850113563424252e-05}
{"train_loss": 0.4367387294769287, "global_step": 21671, "epoch": 243, "lr": 9.850099474671315e-05}
{"train_loss": 0.44750526547431946, "global_step": 21672, "epoch": 243, "lr": 9.85008538526634e-05}
{"train_loss": 0.39262905716896057, "global_step": 21673, "epoch": 243, "lr": 9.850071295209331e-05}
{"train_loss": 0.5520452857017517, "global_step": 21674, "epoch": 243, "lr": 9.850057204500287e-05}
{"train_loss": 0.45635858178138733, "global_step": 21675, "epoch": 243, "lr": 9.850043113139214e-05}
{"train_loss": 0.4013182520866394, "global_step": 21676, "epoch": 243, "lr": 9.850029021126112e-05}
{"train_loss": 0.5425727367401123, "global_step": 21677, "epoch": 243, "lr": 9.850014928460982e-05}
{"train_loss": 0.4259933531284332, "global_step": 21678, "epoch": 243, "lr": 9.850000835143827e-05}
{"train_loss": 0.4707368314266205, "global_step": 21679, "epoch": 243, "lr": 9.849986741174647e-05}
{"train_loss": 0.553256630897522, "global_step": 21680, "epoch": 243, "lr": 9.849972646553446e-05}
{"train_loss": 0.46526896953582764, "global_step": 21681, "epoch": 243, "lr": 9.849958551280226e-05}
{"train_loss": 0.511432945728302, "global_step": 21682, "epoch": 243, "lr": 9.849944455354988e-05}
{"train_loss": 0.3824126720428467, "global_step": 21683, "epoch": 243, "lr": 9.849930358777735e-05}
{"train_loss": 0.5649794936180115, "global_step": 21684, "epoch": 243, "lr": 9.849916261548467e-05}
{"train_loss": 0.4931206703186035, "global_step": 21685, "epoch": 243, "lr": 9.849902163667186e-05}
{"train_loss": 0.4166414141654968, "global_step": 21686, "epoch": 243, "lr": 9.849888065133897e-05}
{"train_loss": 0.45544594526290894, "global_step": 21687, "epoch": 243, "lr": 9.849873965948597e-05}
{"train_loss": 0.39571547508239746, "global_step": 21688, "epoch": 243, "lr": 9.849859866111292e-05}
{"train_loss": 0.6348935961723328, "global_step": 21689, "epoch": 243, "lr": 9.849845765621982e-05}
{"train_loss": 0.48999661207199097, "global_step": 21690, "epoch": 243, "lr": 9.849831664480669e-05}
{"train_loss": 0.46922847628593445, "global_step": 21691, "epoch": 243, "lr": 9.849817562687356e-05}
{"train_loss": 0.49878185987472534, "global_step": 21692, "epoch": 243, "lr": 9.849803460242045e-05}
{"train_loss": 0.35139548778533936, "global_step": 21693, "epoch": 243, "lr": 9.849789357144736e-05}
{"train_loss": 0.4394502341747284, "global_step": 21694, "epoch": 243, "lr": 9.849775253395433e-05}
{"train_loss": 0.4061204493045807, "global_step": 21695, "epoch": 243, "lr": 9.849761148994134e-05}
{"train_loss": 0.48275843262672424, "global_step": 21696, "epoch": 243, "lr": 9.849747043940847e-05}
{"train_loss": 0.4816601872444153, "global_step": 21697, "epoch": 243, "lr": 9.849732938235569e-05}
{"train_loss": 0.39587798714637756, "global_step": 21698, "epoch": 243, "lr": 9.849718831878303e-05}
{"train_loss": 0.3375551104545593, "global_step": 21699, "epoch": 243, "lr": 9.849704724869052e-05}
{"train_loss": 0.42835038900375366, "global_step": 21700, "epoch": 243, "lr": 9.849690617207817e-05}
{"train_loss": 0.5266409516334534, "global_step": 21701, "epoch": 243, "lr": 9.849676508894601e-05}
{"train_loss": 0.516556441783905, "global_step": 21702, "epoch": 243, "lr": 9.849662399929405e-05}
{"train_loss": 0.46109291911125183, "global_step": 21703, "epoch": 243, "lr": 9.84964829031223e-05}
{"train_loss": 0.5029551982879639, "global_step": 21704, "epoch": 243, "lr": 9.84963418004308e-05}
{"train_loss": 0.3369186222553253, "global_step": 21705, "epoch": 243, "lr": 9.849620069121957e-05}
{"train_loss": 0.42047369480133057, "global_step": 21706, "epoch": 243, "lr": 9.849605957548859e-05}
{"train_loss": 0.5278739929199219, "global_step": 21707, "epoch": 243, "lr": 9.849591845323791e-05}
{"train_loss": 0.4286932051181793, "global_step": 21708, "epoch": 243, "lr": 9.849577732446755e-05}
{"train_loss": 0.462176650762558, "global_step": 21709, "epoch": 243, "lr": 9.849563618917754e-05}
{"train_loss": 0.3580227792263031, "global_step": 21710, "epoch": 243, "lr": 9.849549504736786e-05}
{"train_loss": 0.4000962972640991, "global_step": 21711, "epoch": 243, "lr": 9.849535389903857e-05}
{"train_loss": 0.43119996786117554, "global_step": 21712, "epoch": 243, "lr": 9.849521274418964e-05}
{"train_loss": 0.4143531024456024, "global_step": 21713, "epoch": 243, "lr": 9.849507158282116e-05}
{"train_loss": 0.372634619474411, "global_step": 21714, "epoch": 243, "lr": 9.849493041493307e-05}
{"train_loss": 0.46700404601150686, "global_step": 21715, "epoch": 243, "lr": 9.849478924052547e-05, "val_loss": 1.8430894613265991}
{"train_loss": 0.5341002345085144, "global_step": 21716, "epoch": 244, "lr": 9.849464805959829e-05}
{"train_loss": 0.41363605856895447, "global_step": 21717, "epoch": 244, "lr": 9.849450687215163e-05}
{"train_loss": 0.4555646479129791, "global_step": 21718, "epoch": 244, "lr": 9.849436567818547e-05}
{"train_loss": 0.4679212272167206, "global_step": 21719, "epoch": 244, "lr": 9.849422447769982e-05}
{"train_loss": 0.4909471273422241, "global_step": 21720, "epoch": 244, "lr": 9.849408327069471e-05}
{"train_loss": 0.4371453523635864, "global_step": 21721, "epoch": 244, "lr": 9.849394205717018e-05}
{"train_loss": 0.40089142322540283, "global_step": 21722, "epoch": 244, "lr": 9.849380083712622e-05}
{"train_loss": 0.4213275611400604, "global_step": 21723, "epoch": 244, "lr": 9.849365961056287e-05}
{"train_loss": 0.4008484482765198, "global_step": 21724, "epoch": 244, "lr": 9.849351837748014e-05}
{"train_loss": 0.46322330832481384, "global_step": 21725, "epoch": 244, "lr": 9.849337713787802e-05}
{"train_loss": 0.5455341339111328, "global_step": 21726, "epoch": 244, "lr": 9.849323589175659e-05}
{"train_loss": 0.4195510447025299, "global_step": 21727, "epoch": 244, "lr": 9.84930946391158e-05}
{"train_loss": 0.44880035519599915, "global_step": 21728, "epoch": 244, "lr": 9.849295337995573e-05}
{"train_loss": 0.43328073620796204, "global_step": 21729, "epoch": 244, "lr": 9.849281211427637e-05}
{"train_loss": 0.4667147397994995, "global_step": 21730, "epoch": 244, "lr": 9.849267084207773e-05}
{"train_loss": 0.6059727072715759, "global_step": 21731, "epoch": 244, "lr": 9.849252956335986e-05}
{"train_loss": 0.435447096824646, "global_step": 21732, "epoch": 244, "lr": 9.849238827812275e-05}
{"train_loss": 0.5109308362007141, "global_step": 21733, "epoch": 244, "lr": 9.849224698636644e-05}
{"train_loss": 0.6376617550849915, "global_step": 21734, "epoch": 244, "lr": 9.849210568809093e-05}
{"train_loss": 0.49360138177871704, "global_step": 21735, "epoch": 244, "lr": 9.849196438329625e-05}
{"train_loss": 0.4097028374671936, "global_step": 21736, "epoch": 244, "lr": 9.849182307198241e-05}
{"train_loss": 0.3366515636444092, "global_step": 21737, "epoch": 244, "lr": 9.849168175414943e-05}
{"train_loss": 0.5196734070777893, "global_step": 21738, "epoch": 244, "lr": 9.849154042979735e-05}
{"train_loss": 0.4323408901691437, "global_step": 21739, "epoch": 244, "lr": 9.849139909892618e-05}
{"train_loss": 0.5199880599975586, "global_step": 21740, "epoch": 244, "lr": 9.84912577615359e-05}
{"train_loss": 0.46843045949935913, "global_step": 21741, "epoch": 244, "lr": 9.849111641762659e-05}
{"train_loss": 0.5168280005455017, "global_step": 21742, "epoch": 244, "lr": 9.849097506719822e-05}
{"train_loss": 0.4738035500049591, "global_step": 21743, "epoch": 244, "lr": 9.849083371025085e-05}
{"train_loss": 0.41759583353996277, "global_step": 21744, "epoch": 244, "lr": 9.849069234678448e-05}
{"train_loss": 0.3667060136795044, "global_step": 21745, "epoch": 244, "lr": 9.849055097679911e-05}
{"train_loss": 0.6027833819389343, "global_step": 21746, "epoch": 244, "lr": 9.849040960029478e-05}
{"train_loss": 0.6191737055778503, "global_step": 21747, "epoch": 244, "lr": 9.84902682172715e-05}
{"train_loss": 0.4158843457698822, "global_step": 21748, "epoch": 244, "lr": 9.849012682772931e-05}
{"train_loss": 0.4268032908439636, "global_step": 21749, "epoch": 244, "lr": 9.84899854316682e-05}
{"train_loss": 0.4247940182685852, "global_step": 21750, "epoch": 244, "lr": 9.848984402908821e-05}
{"train_loss": 0.4356202185153961, "global_step": 21751, "epoch": 244, "lr": 9.848970261998936e-05}
{"train_loss": 0.5517059564590454, "global_step": 21752, "epoch": 244, "lr": 9.848956120437166e-05}
{"train_loss": 0.33537253737449646, "global_step": 21753, "epoch": 244, "lr": 9.848941978223511e-05}
{"train_loss": 0.4160124361515045, "global_step": 21754, "epoch": 244, "lr": 9.848927835357975e-05}
{"train_loss": 0.44093021750450134, "global_step": 21755, "epoch": 244, "lr": 9.848913691840561e-05}
{"train_loss": 0.45994651317596436, "global_step": 21756, "epoch": 244, "lr": 9.848899547671268e-05}
{"train_loss": 0.5463661551475525, "global_step": 21757, "epoch": 244, "lr": 9.848885402850102e-05}
{"train_loss": 0.3924423158168793, "global_step": 21758, "epoch": 244, "lr": 9.848871257377062e-05}
{"train_loss": 0.47021862864494324, "global_step": 21759, "epoch": 244, "lr": 9.848857111252149e-05}
{"train_loss": 0.49683335423469543, "global_step": 21760, "epoch": 244, "lr": 9.848842964475366e-05}
{"train_loss": 0.4674990475177765, "global_step": 21761, "epoch": 244, "lr": 9.848828817046717e-05}
{"train_loss": 0.5762109756469727, "global_step": 21762, "epoch": 244, "lr": 9.848814668966201e-05}
{"train_loss": 0.5397441983222961, "global_step": 21763, "epoch": 244, "lr": 9.84880052023382e-05}
{"train_loss": 0.4330796003341675, "global_step": 21764, "epoch": 244, "lr": 9.848786370849579e-05}
{"train_loss": 0.3465416431427002, "global_step": 21765, "epoch": 244, "lr": 9.848772220813477e-05}
{"train_loss": 0.47737833857536316, "global_step": 21766, "epoch": 244, "lr": 9.848758070125517e-05}
{"train_loss": 0.45389366149902344, "global_step": 21767, "epoch": 244, "lr": 9.848743918785699e-05}
{"train_loss": 0.5386161804199219, "global_step": 21768, "epoch": 244, "lr": 9.848729766794029e-05}
{"train_loss": 0.5610511898994446, "global_step": 21769, "epoch": 244, "lr": 9.848715614150503e-05}
{"train_loss": 0.5380260348320007, "global_step": 21770, "epoch": 244, "lr": 9.848701460855129e-05}
{"train_loss": 0.5207768082618713, "global_step": 21771, "epoch": 244, "lr": 9.848687306907905e-05}
{"train_loss": 0.502834677696228, "global_step": 21772, "epoch": 244, "lr": 9.848673152308834e-05}
{"train_loss": 0.43675994873046875, "global_step": 21773, "epoch": 244, "lr": 9.848658997057919e-05}
{"train_loss": 0.6172584295272827, "global_step": 21774, "epoch": 244, "lr": 9.84864484115516e-05}
{"train_loss": 0.4483705461025238, "global_step": 21775, "epoch": 244, "lr": 9.84863068460056e-05}
{"train_loss": 0.4288974404335022, "global_step": 21776, "epoch": 244, "lr": 9.848616527394121e-05}
{"train_loss": 0.41352176666259766, "global_step": 21777, "epoch": 244, "lr": 9.848602369535844e-05}
{"train_loss": 0.46453702449798584, "global_step": 21778, "epoch": 244, "lr": 9.848588211025733e-05}
{"train_loss": 0.4174479842185974, "global_step": 21779, "epoch": 244, "lr": 9.848574051863787e-05}
{"train_loss": 0.3712828755378723, "global_step": 21780, "epoch": 244, "lr": 9.848559892050008e-05}
{"train_loss": 0.47193634510040283, "global_step": 21781, "epoch": 244, "lr": 9.848545731584402e-05}
{"train_loss": 0.37923428416252136, "global_step": 21782, "epoch": 244, "lr": 9.848531570466967e-05}
{"train_loss": 0.521369218826294, "global_step": 21783, "epoch": 244, "lr": 9.848517408697706e-05}
{"train_loss": 0.5633236169815063, "global_step": 21784, "epoch": 244, "lr": 9.848503246276622e-05}
{"train_loss": 0.4733256995677948, "global_step": 21785, "epoch": 244, "lr": 9.848489083203714e-05}
{"train_loss": 0.41985711455345154, "global_step": 21786, "epoch": 244, "lr": 9.848474919478987e-05}
{"train_loss": 0.5762830376625061, "global_step": 21787, "epoch": 244, "lr": 9.84846075510244e-05}
{"train_loss": 0.4750446081161499, "global_step": 21788, "epoch": 244, "lr": 9.848446590074079e-05}
{"train_loss": 0.46022242307662964, "global_step": 21789, "epoch": 244, "lr": 9.8484324243939e-05}
{"train_loss": 0.4606803059577942, "global_step": 21790, "epoch": 244, "lr": 9.848418258061911e-05}
{"train_loss": 0.5414460897445679, "global_step": 21791, "epoch": 244, "lr": 9.848404091078111e-05}
{"train_loss": 0.40400105714797974, "global_step": 21792, "epoch": 244, "lr": 9.848389923442503e-05}
{"train_loss": 0.43925797939300537, "global_step": 21793, "epoch": 244, "lr": 9.848375755155085e-05}
{"train_loss": 0.3859999179840088, "global_step": 21794, "epoch": 244, "lr": 9.848361586215866e-05}
{"train_loss": 0.35917189717292786, "global_step": 21795, "epoch": 244, "lr": 9.848347416624842e-05}
{"train_loss": 0.5897770524024963, "global_step": 21796, "epoch": 244, "lr": 9.848333246382016e-05}
{"train_loss": 0.48554423451423645, "global_step": 21797, "epoch": 244, "lr": 9.84831907548739e-05}
{"train_loss": 0.39448317885398865, "global_step": 21798, "epoch": 244, "lr": 9.84830490394097e-05}
{"train_loss": 0.48997777700424194, "global_step": 21799, "epoch": 244, "lr": 9.848290731742751e-05}
{"train_loss": 0.5792343020439148, "global_step": 21800, "epoch": 244, "lr": 9.84827655889274e-05}
{"train_loss": 0.4235542118549347, "global_step": 21801, "epoch": 244, "lr": 9.848262385390938e-05}
{"train_loss": 0.46350306272506714, "global_step": 21802, "epoch": 244, "lr": 9.848248211237346e-05}
{"train_loss": 0.38448184728622437, "global_step": 21803, "epoch": 244, "lr": 9.848234036431965e-05}
{"train_loss": 0.4680307551046436, "global_step": 21804, "epoch": 244, "lr": 9.848219860974797e-05, "val_loss": 1.7899001836776733}
{"train_loss": 0.4809597134590149, "global_step": 21805, "epoch": 245, "lr": 9.848205684865848e-05}
{"train_loss": 0.37216904759407043, "global_step": 21806, "epoch": 245, "lr": 9.848191508105114e-05}
{"train_loss": 0.4503242075443268, "global_step": 21807, "epoch": 245, "lr": 9.8481773306926e-05}
{"train_loss": 0.4402066469192505, "global_step": 21808, "epoch": 245, "lr": 9.84816315262831e-05}
{"train_loss": 0.4444303810596466, "global_step": 21809, "epoch": 245, "lr": 9.848148973912242e-05}
{"train_loss": 0.37504449486732483, "global_step": 21810, "epoch": 245, "lr": 9.848134794544399e-05}
{"train_loss": 0.43610531091690063, "global_step": 21811, "epoch": 245, "lr": 9.848120614524785e-05}
{"train_loss": 0.44079458713531494, "global_step": 21812, "epoch": 245, "lr": 9.848106433853399e-05}
{"train_loss": 0.3943932354450226, "global_step": 21813, "epoch": 245, "lr": 9.848092252530244e-05}
{"train_loss": 0.4897150695323944, "global_step": 21814, "epoch": 245, "lr": 9.848078070555323e-05}
{"train_loss": 0.45050036907196045, "global_step": 21815, "epoch": 245, "lr": 9.848063887928637e-05}
{"train_loss": 0.3138662576675415, "global_step": 21816, "epoch": 245, "lr": 9.848049704650186e-05}
{"train_loss": 0.4258324205875397, "global_step": 21817, "epoch": 245, "lr": 9.848035520719976e-05}
{"train_loss": 0.4290355443954468, "global_step": 21818, "epoch": 245, "lr": 9.848021336138007e-05}
{"train_loss": 0.5757468938827515, "global_step": 21819, "epoch": 245, "lr": 9.848007150904279e-05}
{"train_loss": 0.40671175718307495, "global_step": 21820, "epoch": 245, "lr": 9.847992965018796e-05}
{"train_loss": 0.4464215338230133, "global_step": 21821, "epoch": 245, "lr": 9.847978778481559e-05}
{"train_loss": 0.5710560083389282, "global_step": 21822, "epoch": 245, "lr": 9.847964591292572e-05}
{"train_loss": 0.36412617564201355, "global_step": 21823, "epoch": 245, "lr": 9.847950403451834e-05}
{"train_loss": 0.46917590498924255, "global_step": 21824, "epoch": 245, "lr": 9.847936214959348e-05}
{"train_loss": 0.517397403717041, "global_step": 21825, "epoch": 245, "lr": 9.847922025815116e-05}
{"train_loss": 0.402397483587265, "global_step": 21826, "epoch": 245, "lr": 9.847907836019141e-05}
{"train_loss": 0.4393329918384552, "global_step": 21827, "epoch": 245, "lr": 9.847893645571422e-05}
{"train_loss": 0.3480623662471771, "global_step": 21828, "epoch": 245, "lr": 9.847879454471964e-05}
{"train_loss": 0.44508177042007446, "global_step": 21829, "epoch": 245, "lr": 9.847865262720768e-05}
{"train_loss": 0.578158974647522, "global_step": 21830, "epoch": 245, "lr": 9.847851070317836e-05}
{"train_loss": 0.6322851777076721, "global_step": 21831, "epoch": 245, "lr": 9.847836877263168e-05}
{"train_loss": 0.530253529548645, "global_step": 21832, "epoch": 245, "lr": 9.847822683556769e-05}
{"train_loss": 0.5091686248779297, "global_step": 21833, "epoch": 245, "lr": 9.847808489198637e-05}
{"train_loss": 0.6638785600662231, "global_step": 21834, "epoch": 245, "lr": 9.847794294188779e-05}
{"train_loss": 0.3171752989292145, "global_step": 21835, "epoch": 245, "lr": 9.847780098527192e-05}
{"train_loss": 0.51059889793396, "global_step": 21836, "epoch": 245, "lr": 9.847765902213881e-05}
{"train_loss": 0.41925743222236633, "global_step": 21837, "epoch": 245, "lr": 9.847751705248848e-05}
{"train_loss": 0.5136743187904358, "global_step": 21838, "epoch": 245, "lr": 9.847737507632093e-05}
{"train_loss": 0.4548409581184387, "global_step": 21839, "epoch": 245, "lr": 9.847723309363619e-05}
{"train_loss": 0.4751679599285126, "global_step": 21840, "epoch": 245, "lr": 9.847709110443426e-05}
{"train_loss": 0.41756483912467957, "global_step": 21841, "epoch": 245, "lr": 9.847694910871519e-05}
{"train_loss": 0.36855289340019226, "global_step": 21842, "epoch": 245, "lr": 9.847680710647899e-05}
{"train_loss": 0.5133354663848877, "global_step": 21843, "epoch": 245, "lr": 9.847666509772566e-05}
{"train_loss": 0.4512622356414795, "global_step": 21844, "epoch": 245, "lr": 9.847652308245525e-05}
{"train_loss": 0.6379222273826599, "global_step": 21845, "epoch": 245, "lr": 9.847638106066774e-05}
{"train_loss": 0.4875573217868805, "global_step": 21846, "epoch": 245, "lr": 9.847623903236321e-05}
{"train_loss": 0.4512539207935333, "global_step": 21847, "epoch": 245, "lr": 9.84760969975416e-05}
{"train_loss": 0.42782390117645264, "global_step": 21848, "epoch": 245, "lr": 9.847595495620299e-05}
{"train_loss": 0.5848809480667114, "global_step": 21849, "epoch": 245, "lr": 9.847581290834737e-05}
{"train_loss": 0.5520015358924866, "global_step": 21850, "epoch": 245, "lr": 9.847567085397478e-05}
{"train_loss": 0.519497811794281, "global_step": 21851, "epoch": 245, "lr": 9.847552879308521e-05}
{"train_loss": 0.47508543729782104, "global_step": 21852, "epoch": 245, "lr": 9.847538672567871e-05}
{"train_loss": 0.44041359424591064, "global_step": 21853, "epoch": 245, "lr": 9.847524465175527e-05}
{"train_loss": 0.5662786364555359, "global_step": 21854, "epoch": 245, "lr": 9.847510257131493e-05}
{"train_loss": 0.6268122792243958, "global_step": 21855, "epoch": 245, "lr": 9.847496048435772e-05}
{"train_loss": 0.5016210079193115, "global_step": 21856, "epoch": 245, "lr": 9.847481839088362e-05}
{"train_loss": 0.43586140871047974, "global_step": 21857, "epoch": 245, "lr": 9.847467629089269e-05}
{"train_loss": 0.42996472120285034, "global_step": 21858, "epoch": 245, "lr": 9.847453418438491e-05}
{"train_loss": 0.4743300676345825, "global_step": 21859, "epoch": 245, "lr": 9.847439207136033e-05}
{"train_loss": 0.37622928619384766, "global_step": 21860, "epoch": 245, "lr": 9.847424995181897e-05}
{"train_loss": 0.4621299207210541, "global_step": 21861, "epoch": 245, "lr": 9.847410782576083e-05}
{"train_loss": 0.4113122522830963, "global_step": 21862, "epoch": 245, "lr": 9.847396569318593e-05}
{"train_loss": 0.5866504907608032, "global_step": 21863, "epoch": 245, "lr": 9.84738235540943e-05}
{"train_loss": 0.3794918358325958, "global_step": 21864, "epoch": 245, "lr": 9.847368140848597e-05}
{"train_loss": 0.3490197956562042, "global_step": 21865, "epoch": 245, "lr": 9.847353925636091e-05}
{"train_loss": 0.43668508529663086, "global_step": 21866, "epoch": 245, "lr": 9.84733970977192e-05}
{"train_loss": 0.4823416471481323, "global_step": 21867, "epoch": 245, "lr": 9.847325493256084e-05}
{"train_loss": 0.3882717490196228, "global_step": 21868, "epoch": 245, "lr": 9.847311276088582e-05}
{"train_loss": 0.5062682628631592, "global_step": 21869, "epoch": 245, "lr": 9.847297058269417e-05}
{"train_loss": 0.39926090836524963, "global_step": 21870, "epoch": 245, "lr": 9.847282839798595e-05}
{"train_loss": 0.5003412961959839, "global_step": 21871, "epoch": 245, "lr": 9.847268620676115e-05}
{"train_loss": 0.47562652826309204, "global_step": 21872, "epoch": 245, "lr": 9.847254400901977e-05}
{"train_loss": 0.5671038031578064, "global_step": 21873, "epoch": 245, "lr": 9.847240180476185e-05}
{"train_loss": 0.40068554878234863, "global_step": 21874, "epoch": 245, "lr": 9.847225959398741e-05}
{"train_loss": 0.5536773800849915, "global_step": 21875, "epoch": 245, "lr": 9.847211737669646e-05}
{"train_loss": 0.4172791838645935, "global_step": 21876, "epoch": 245, "lr": 9.847197515288903e-05}
{"train_loss": 0.5852146148681641, "global_step": 21877, "epoch": 245, "lr": 9.847183292256515e-05}
{"train_loss": 0.4842846989631653, "global_step": 21878, "epoch": 245, "lr": 9.84716906857248e-05}
{"train_loss": 0.557235062122345, "global_step": 21879, "epoch": 245, "lr": 9.847154844236802e-05}
{"train_loss": 0.41147544980049133, "global_step": 21880, "epoch": 245, "lr": 9.847140619249484e-05}
{"train_loss": 0.5139867067337036, "global_step": 21881, "epoch": 245, "lr": 9.847126393610527e-05}
{"train_loss": 0.5418726801872253, "global_step": 21882, "epoch": 245, "lr": 9.847112167319933e-05}
{"train_loss": 0.5145382285118103, "global_step": 21883, "epoch": 245, "lr": 9.847097940377704e-05}
{"train_loss": 0.5365824699401855, "global_step": 21884, "epoch": 245, "lr": 9.847083712783843e-05}
{"train_loss": 0.38392823934555054, "global_step": 21885, "epoch": 245, "lr": 9.847069484538347e-05}
{"train_loss": 0.6310253143310547, "global_step": 21886, "epoch": 245, "lr": 9.847055255641224e-05}
{"train_loss": 0.472073495388031, "global_step": 21887, "epoch": 245, "lr": 9.847041026092474e-05}
{"train_loss": 0.38807374238967896, "global_step": 21888, "epoch": 245, "lr": 9.847026795892099e-05}
{"train_loss": 0.49464935064315796, "global_step": 21889, "epoch": 245, "lr": 9.847012565040097e-05}
{"train_loss": 0.5500431060791016, "global_step": 21890, "epoch": 245, "lr": 9.846998333536476e-05}
{"train_loss": 0.44500672817230225, "global_step": 21891, "epoch": 245, "lr": 9.846984101381234e-05}
{"train_loss": 0.5565302968025208, "global_step": 21892, "epoch": 245, "lr": 9.846969868574374e-05}
{"train_loss": 0.47259843349456787, "global_step": 21893, "epoch": 245, "lr": 9.846955635115899e-05, "val_loss": 1.841950535774231, "train_action_mse_error": 37.07166290283203}
{"train_loss": 0.49467960000038147, "global_step": 21894, "epoch": 246, "lr": 9.84694140100581e-05}
{"train_loss": 0.4338325560092926, "global_step": 21895, "epoch": 246, "lr": 9.846927166244107e-05}
{"train_loss": 0.4332740902900696, "global_step": 21896, "epoch": 246, "lr": 9.846912930830795e-05}
{"train_loss": 0.4846997559070587, "global_step": 21897, "epoch": 246, "lr": 9.846898694765875e-05}
{"train_loss": 0.5700044631958008, "global_step": 21898, "epoch": 246, "lr": 9.846884458049348e-05}
{"train_loss": 0.43598654866218567, "global_step": 21899, "epoch": 246, "lr": 9.846870220681217e-05}
{"train_loss": 0.4362892806529999, "global_step": 21900, "epoch": 246, "lr": 9.846855982661483e-05}
{"train_loss": 0.36690640449523926, "global_step": 21901, "epoch": 246, "lr": 9.846841743990147e-05}
{"train_loss": 0.40918469429016113, "global_step": 21902, "epoch": 246, "lr": 9.846827504667213e-05}
{"train_loss": 0.5745588541030884, "global_step": 21903, "epoch": 246, "lr": 9.846813264692683e-05}
{"train_loss": 0.5373189449310303, "global_step": 21904, "epoch": 246, "lr": 9.846799024066558e-05}
{"train_loss": 0.4144307076931, "global_step": 21905, "epoch": 246, "lr": 9.846784782788838e-05}
{"train_loss": 0.3853142559528351, "global_step": 21906, "epoch": 246, "lr": 9.84677054085953e-05}
{"train_loss": 0.4296960234642029, "global_step": 21907, "epoch": 246, "lr": 9.84675629827863e-05}
{"train_loss": 0.44825392961502075, "global_step": 21908, "epoch": 246, "lr": 9.846742055046144e-05}
{"train_loss": 0.5434017777442932, "global_step": 21909, "epoch": 246, "lr": 9.846727811162073e-05}
{"train_loss": 0.4928285479545593, "global_step": 21910, "epoch": 246, "lr": 9.846713566626417e-05}
{"train_loss": 0.4553123116493225, "global_step": 21911, "epoch": 246, "lr": 9.846699321439181e-05}
{"train_loss": 0.4249241054058075, "global_step": 21912, "epoch": 246, "lr": 9.846685075600364e-05}
{"train_loss": 0.44858917593955994, "global_step": 21913, "epoch": 246, "lr": 9.84667082910997e-05}
{"train_loss": 0.45409420132637024, "global_step": 21914, "epoch": 246, "lr": 9.846656581968e-05}
{"train_loss": 0.419683039188385, "global_step": 21915, "epoch": 246, "lr": 9.846642334174456e-05}
{"train_loss": 0.47370731830596924, "global_step": 21916, "epoch": 246, "lr": 9.846628085729341e-05}
{"train_loss": 0.34930679202079773, "global_step": 21917, "epoch": 246, "lr": 9.846613836632655e-05}
{"train_loss": 0.4224544167518616, "global_step": 21918, "epoch": 246, "lr": 9.846599586884401e-05}
{"train_loss": 0.5282138586044312, "global_step": 21919, "epoch": 246, "lr": 9.84658533648458e-05}
{"train_loss": 0.4768042266368866, "global_step": 21920, "epoch": 246, "lr": 9.846571085433195e-05}
{"train_loss": 0.4163690209388733, "global_step": 21921, "epoch": 246, "lr": 9.84655683373025e-05}
{"train_loss": 0.47991451621055603, "global_step": 21922, "epoch": 246, "lr": 9.846542581375742e-05}
{"train_loss": 0.3928891718387604, "global_step": 21923, "epoch": 246, "lr": 9.846528328369675e-05}
{"train_loss": 0.4169870913028717, "global_step": 21924, "epoch": 246, "lr": 9.846514074712051e-05}
{"train_loss": 0.6051672697067261, "global_step": 21925, "epoch": 246, "lr": 9.846499820402875e-05}
{"train_loss": 0.40650948882102966, "global_step": 21926, "epoch": 246, "lr": 9.846485565442145e-05}
{"train_loss": 0.4271179437637329, "global_step": 21927, "epoch": 246, "lr": 9.846471309829863e-05}
{"train_loss": 0.3679632544517517, "global_step": 21928, "epoch": 246, "lr": 9.846457053566034e-05}
{"train_loss": 0.47333329916000366, "global_step": 21929, "epoch": 246, "lr": 9.846442796650656e-05}
{"train_loss": 0.429074227809906, "global_step": 21930, "epoch": 246, "lr": 9.846428539083734e-05}
{"train_loss": 0.34745651483535767, "global_step": 21931, "epoch": 246, "lr": 9.846414280865268e-05}
{"train_loss": 0.4108109474182129, "global_step": 21932, "epoch": 246, "lr": 9.846400021995261e-05}
{"train_loss": 0.42233648896217346, "global_step": 21933, "epoch": 246, "lr": 9.846385762473715e-05}
{"train_loss": 0.42524755001068115, "global_step": 21934, "epoch": 246, "lr": 9.846371502300631e-05}
{"train_loss": 0.41755202412605286, "global_step": 21935, "epoch": 246, "lr": 9.846357241476012e-05}
{"train_loss": 0.4913686215877533, "global_step": 21936, "epoch": 246, "lr": 9.846342979999858e-05}
{"train_loss": 0.4709455966949463, "global_step": 21937, "epoch": 246, "lr": 9.846328717872173e-05}
{"train_loss": 0.44763436913490295, "global_step": 21938, "epoch": 246, "lr": 9.846314455092957e-05}
{"train_loss": 0.44628605246543884, "global_step": 21939, "epoch": 246, "lr": 9.846300191662216e-05}
{"train_loss": 0.4963928461074829, "global_step": 21940, "epoch": 246, "lr": 9.846285927579947e-05}
{"train_loss": 0.4552140235900879, "global_step": 21941, "epoch": 246, "lr": 9.846271662846154e-05}
{"train_loss": 0.5103992223739624, "global_step": 21942, "epoch": 246, "lr": 9.846257397460838e-05}
{"train_loss": 0.4368925988674164, "global_step": 21943, "epoch": 246, "lr": 9.846243131424004e-05}
{"train_loss": 0.5015061497688293, "global_step": 21944, "epoch": 246, "lr": 9.84622886473565e-05}
{"train_loss": 0.42274707555770874, "global_step": 21945, "epoch": 246, "lr": 9.846214597395781e-05}
{"train_loss": 0.4373930096626282, "global_step": 21946, "epoch": 246, "lr": 9.846200329404396e-05}
{"train_loss": 0.4176253080368042, "global_step": 21947, "epoch": 246, "lr": 9.8461860607615e-05}
{"train_loss": 0.3585606515407562, "global_step": 21948, "epoch": 246, "lr": 9.846171791467091e-05}
{"train_loss": 0.3691626191139221, "global_step": 21949, "epoch": 246, "lr": 9.846157521521176e-05}
{"train_loss": 0.47646814584732056, "global_step": 21950, "epoch": 246, "lr": 9.846143250923752e-05}
{"train_loss": 0.5813164710998535, "global_step": 21951, "epoch": 246, "lr": 9.846128979674825e-05}
{"train_loss": 0.4180322587490082, "global_step": 21952, "epoch": 246, "lr": 9.846114707774394e-05}
{"train_loss": 0.4992527961730957, "global_step": 21953, "epoch": 246, "lr": 9.846100435222461e-05}
{"train_loss": 0.5068190693855286, "global_step": 21954, "epoch": 246, "lr": 9.84608616201903e-05}
{"train_loss": 0.37138161063194275, "global_step": 21955, "epoch": 246, "lr": 9.846071888164101e-05}
{"train_loss": 0.37734004855155945, "global_step": 21956, "epoch": 246, "lr": 9.846057613657678e-05}
{"train_loss": 0.4474344253540039, "global_step": 21957, "epoch": 246, "lr": 9.846043338499762e-05}
{"train_loss": 0.37542712688446045, "global_step": 21958, "epoch": 246, "lr": 9.846029062690353e-05}
{"train_loss": 0.4178042411804199, "global_step": 21959, "epoch": 246, "lr": 9.846014786229454e-05}
{"train_loss": 0.5309365391731262, "global_step": 21960, "epoch": 246, "lr": 9.846000509117068e-05}
{"train_loss": 0.5545058250427246, "global_step": 21961, "epoch": 246, "lr": 9.845986231353197e-05}
{"train_loss": 0.46141713857650757, "global_step": 21962, "epoch": 246, "lr": 9.845971952937842e-05}
{"train_loss": 0.31020253896713257, "global_step": 21963, "epoch": 246, "lr": 9.845957673871004e-05}
{"train_loss": 0.4498496651649475, "global_step": 21964, "epoch": 246, "lr": 9.845943394152688e-05}
{"train_loss": 0.5866410136222839, "global_step": 21965, "epoch": 246, "lr": 9.845929113782892e-05}
{"train_loss": 0.4185366630554199, "global_step": 21966, "epoch": 246, "lr": 9.845914832761622e-05}
{"train_loss": 0.3161511719226837, "global_step": 21967, "epoch": 246, "lr": 9.845900551088877e-05}
{"train_loss": 0.4975232779979706, "global_step": 21968, "epoch": 246, "lr": 9.845886268764659e-05}
{"train_loss": 0.42559653520584106, "global_step": 21969, "epoch": 246, "lr": 9.845871985788971e-05}
{"train_loss": 0.3900728225708008, "global_step": 21970, "epoch": 246, "lr": 9.845857702161815e-05}
{"train_loss": 0.5523003935813904, "global_step": 21971, "epoch": 246, "lr": 9.845843417883191e-05}
{"train_loss": 0.5013397336006165, "global_step": 21972, "epoch": 246, "lr": 9.845829132953104e-05}
{"train_loss": 0.48356112837791443, "global_step": 21973, "epoch": 246, "lr": 9.845814847371555e-05}
{"train_loss": 0.43415719270706177, "global_step": 21974, "epoch": 246, "lr": 9.845800561138543e-05}
{"train_loss": 0.4325094521045685, "global_step": 21975, "epoch": 246, "lr": 9.845786274254073e-05}
{"train_loss": 0.5113615393638611, "global_step": 21976, "epoch": 246, "lr": 9.845771986718147e-05}
{"train_loss": 0.35919976234436035, "global_step": 21977, "epoch": 246, "lr": 9.845757698530766e-05}
{"train_loss": 0.5456662774085999, "global_step": 21978, "epoch": 246, "lr": 9.845743409691931e-05}
{"train_loss": 0.42047953605651855, "global_step": 21979, "epoch": 246, "lr": 9.845729120201645e-05}
{"train_loss": 0.49937278032302856, "global_step": 21980, "epoch": 246, "lr": 9.84571483005991e-05}
{"train_loss": 0.491009920835495, "global_step": 21981, "epoch": 246, "lr": 9.845700539266728e-05}
{"train_loss": 0.45133964499730744, "global_step": 21982, "epoch": 246, "lr": 9.845686247822101e-05, "val_loss": 1.8575026988983154}
{"train_loss": 0.3542463481426239, "global_step": 21983, "epoch": 247, "lr": 9.845671955726029e-05}
{"train_loss": 0.4465540647506714, "global_step": 21984, "epoch": 247, "lr": 9.845657662978517e-05}
{"train_loss": 0.457226037979126, "global_step": 21985, "epoch": 247, "lr": 9.845643369579565e-05}
{"train_loss": 0.422758549451828, "global_step": 21986, "epoch": 247, "lr": 9.845629075529174e-05}
{"train_loss": 0.49600622057914734, "global_step": 21987, "epoch": 247, "lr": 9.845614780827348e-05}
{"train_loss": 0.4313971698284149, "global_step": 21988, "epoch": 247, "lr": 9.845600485474088e-05}
{"train_loss": 0.5398158431053162, "global_step": 21989, "epoch": 247, "lr": 9.845586189469398e-05}
{"train_loss": 0.41267669200897217, "global_step": 21990, "epoch": 247, "lr": 9.845571892813277e-05}
{"train_loss": 0.4747193455696106, "global_step": 21991, "epoch": 247, "lr": 9.845557595505727e-05}
{"train_loss": 0.38547343015670776, "global_step": 21992, "epoch": 247, "lr": 9.845543297546751e-05}
{"train_loss": 0.40852487087249756, "global_step": 21993, "epoch": 247, "lr": 9.845528998936351e-05}
{"train_loss": 0.501491904258728, "global_step": 21994, "epoch": 247, "lr": 9.845514699674528e-05}
{"train_loss": 0.3983454406261444, "global_step": 21995, "epoch": 247, "lr": 9.845500399761286e-05}
{"train_loss": 0.4505714774131775, "global_step": 21996, "epoch": 247, "lr": 9.845486099196624e-05}
{"train_loss": 0.4016790986061096, "global_step": 21997, "epoch": 247, "lr": 9.845471797980548e-05}
{"train_loss": 0.5075446367263794, "global_step": 21998, "epoch": 247, "lr": 9.845457496113055e-05}
{"train_loss": 0.45668700337409973, "global_step": 21999, "epoch": 247, "lr": 9.84544319359415e-05}
{"train_loss": 0.42317888140678406, "global_step": 22000, "epoch": 247, "lr": 9.845428890423833e-05}
{"train_loss": 0.41642752289772034, "global_step": 22001, "epoch": 247, "lr": 9.845414586602109e-05}
{"train_loss": 0.4863368570804596, "global_step": 22002, "epoch": 247, "lr": 9.845400282128978e-05}
{"train_loss": 0.39599743485450745, "global_step": 22003, "epoch": 247, "lr": 9.845385977004441e-05}
{"train_loss": 0.4114510416984558, "global_step": 22004, "epoch": 247, "lr": 9.845371671228502e-05}
{"train_loss": 0.5023835897445679, "global_step": 22005, "epoch": 247, "lr": 9.84535736480116e-05}
{"train_loss": 0.4279036521911621, "global_step": 22006, "epoch": 247, "lr": 9.84534305772242e-05}
{"train_loss": 0.46093225479125977, "global_step": 22007, "epoch": 247, "lr": 9.845328749992284e-05}
{"train_loss": 0.40593940019607544, "global_step": 22008, "epoch": 247, "lr": 9.84531444161075e-05}
{"train_loss": 0.5219407081604004, "global_step": 22009, "epoch": 247, "lr": 9.845300132577823e-05}
{"train_loss": 0.5705931186676025, "global_step": 22010, "epoch": 247, "lr": 9.845285822893506e-05}
{"train_loss": 0.5498611330986023, "global_step": 22011, "epoch": 247, "lr": 9.845271512557799e-05}
{"train_loss": 0.367382675409317, "global_step": 22012, "epoch": 247, "lr": 9.845257201570703e-05}
{"train_loss": 0.32808324694633484, "global_step": 22013, "epoch": 247, "lr": 9.845242889932223e-05}
{"train_loss": 0.5713815689086914, "global_step": 22014, "epoch": 247, "lr": 9.845228577642357e-05}
{"train_loss": 0.45813900232315063, "global_step": 22015, "epoch": 247, "lr": 9.84521426470111e-05}
{"train_loss": 0.41491079330444336, "global_step": 22016, "epoch": 247, "lr": 9.845199951108483e-05}
{"train_loss": 0.46598708629608154, "global_step": 22017, "epoch": 247, "lr": 9.84518563686448e-05}
{"train_loss": 0.44290024042129517, "global_step": 22018, "epoch": 247, "lr": 9.845171321969099e-05}
{"train_loss": 0.4203375279903412, "global_step": 22019, "epoch": 247, "lr": 9.845157006422343e-05}
{"train_loss": 0.41673552989959717, "global_step": 22020, "epoch": 247, "lr": 9.845142690224216e-05}
{"train_loss": 0.5760924220085144, "global_step": 22021, "epoch": 247, "lr": 9.845128373374717e-05}
{"train_loss": 0.5642444491386414, "global_step": 22022, "epoch": 247, "lr": 9.845114055873851e-05}
{"train_loss": 0.5078346729278564, "global_step": 22023, "epoch": 247, "lr": 9.84509973772162e-05}
{"train_loss": 0.3556117117404938, "global_step": 22024, "epoch": 247, "lr": 9.845085418918021e-05}
{"train_loss": 0.4119598865509033, "global_step": 22025, "epoch": 247, "lr": 9.845071099463061e-05}
{"train_loss": 0.4064010977745056, "global_step": 22026, "epoch": 247, "lr": 9.84505677935674e-05}
{"train_loss": 0.5086712837219238, "global_step": 22027, "epoch": 247, "lr": 9.845042458599061e-05}
{"train_loss": 0.5277835130691528, "global_step": 22028, "epoch": 247, "lr": 9.845028137190025e-05}
{"train_loss": 0.5110280513763428, "global_step": 22029, "epoch": 247, "lr": 9.845013815129632e-05}
{"train_loss": 0.47268447279930115, "global_step": 22030, "epoch": 247, "lr": 9.844999492417888e-05}
{"train_loss": 0.49754270911216736, "global_step": 22031, "epoch": 247, "lr": 9.844985169054793e-05}
{"train_loss": 0.4487634599208832, "global_step": 22032, "epoch": 247, "lr": 9.844970845040346e-05}
{"train_loss": 0.3414378762245178, "global_step": 22033, "epoch": 247, "lr": 9.844956520374555e-05}
{"train_loss": 0.30829229950904846, "global_step": 22034, "epoch": 247, "lr": 9.844942195057418e-05}
{"train_loss": 0.36039724946022034, "global_step": 22035, "epoch": 247, "lr": 9.844927869088935e-05}
{"train_loss": 0.403137743473053, "global_step": 22036, "epoch": 247, "lr": 9.844913542469115e-05}
{"train_loss": 0.38428983092308044, "global_step": 22037, "epoch": 247, "lr": 9.844899215197951e-05}
{"train_loss": 0.4384446442127228, "global_step": 22038, "epoch": 247, "lr": 9.844884887275452e-05}
{"train_loss": 0.4527455270290375, "global_step": 22039, "epoch": 247, "lr": 9.844870558701616e-05}
{"train_loss": 0.5350558757781982, "global_step": 22040, "epoch": 247, "lr": 9.844856229476446e-05}
{"train_loss": 0.42968544363975525, "global_step": 22041, "epoch": 247, "lr": 9.844841899599945e-05}
{"train_loss": 0.4133088290691376, "global_step": 22042, "epoch": 247, "lr": 9.844827569072115e-05}
{"train_loss": 0.5211719274520874, "global_step": 22043, "epoch": 247, "lr": 9.844813237892956e-05}
{"train_loss": 0.5051280856132507, "global_step": 22044, "epoch": 247, "lr": 9.84479890606247e-05}
{"train_loss": 0.4326923191547394, "global_step": 22045, "epoch": 247, "lr": 9.844784573580661e-05}
{"train_loss": 0.5505993366241455, "global_step": 22046, "epoch": 247, "lr": 9.84477024044753e-05}
{"train_loss": 0.6349713206291199, "global_step": 22047, "epoch": 247, "lr": 9.844755906663077e-05}
{"train_loss": 0.7299526333808899, "global_step": 22048, "epoch": 247, "lr": 9.844741572227307e-05}
{"train_loss": 0.5608746409416199, "global_step": 22049, "epoch": 247, "lr": 9.84472723714022e-05}
{"train_loss": 0.4903462529182434, "global_step": 22050, "epoch": 247, "lr": 9.844712901401819e-05}
{"train_loss": 0.5337952375411987, "global_step": 22051, "epoch": 247, "lr": 9.844698565012105e-05}
{"train_loss": 0.4331548810005188, "global_step": 22052, "epoch": 247, "lr": 9.844684227971081e-05}
{"train_loss": 0.5368432402610779, "global_step": 22053, "epoch": 247, "lr": 9.844669890278748e-05}
{"train_loss": 0.5773308277130127, "global_step": 22054, "epoch": 247, "lr": 9.844655551935107e-05}
{"train_loss": 0.366178423166275, "global_step": 22055, "epoch": 247, "lr": 9.844641212940162e-05}
{"train_loss": 0.4734474718570709, "global_step": 22056, "epoch": 247, "lr": 9.844626873293914e-05}
{"train_loss": 0.4647637605667114, "global_step": 22057, "epoch": 247, "lr": 9.844612532996365e-05}
{"train_loss": 0.39378461241722107, "global_step": 22058, "epoch": 247, "lr": 9.844598192047516e-05}
{"train_loss": 0.5122912526130676, "global_step": 22059, "epoch": 247, "lr": 9.844583850447371e-05}
{"train_loss": 0.5178280472755432, "global_step": 22060, "epoch": 247, "lr": 9.844569508195931e-05}
{"train_loss": 0.5459257960319519, "global_step": 22061, "epoch": 247, "lr": 9.844555165293196e-05}
{"train_loss": 0.5555474162101746, "global_step": 22062, "epoch": 247, "lr": 9.844540821739172e-05}
{"train_loss": 0.6177839040756226, "global_step": 22063, "epoch": 247, "lr": 9.844526477533857e-05}
{"train_loss": 0.5531601905822754, "global_step": 22064, "epoch": 247, "lr": 9.844512132677255e-05}
{"train_loss": 0.41038089990615845, "global_step": 22065, "epoch": 247, "lr": 9.844497787169366e-05}
{"train_loss": 0.6185060143470764, "global_step": 22066, "epoch": 247, "lr": 9.844483441010195e-05}
{"train_loss": 0.5598776340484619, "global_step": 22067, "epoch": 247, "lr": 9.844469094199742e-05}
{"train_loss": 0.5156338810920715, "global_step": 22068, "epoch": 247, "lr": 9.844454746738009e-05}
{"train_loss": 0.3804187476634979, "global_step": 22069, "epoch": 247, "lr": 9.844440398624998e-05}
{"train_loss": 0.47424352169036865, "global_step": 22070, "epoch": 247, "lr": 9.844426049860711e-05}
{"train_loss": 0.46926393401756716, "global_step": 22071, "epoch": 247, "lr": 9.844411700445149e-05, "val_loss": 1.9565321207046509}
{"train_loss": 0.5400745272636414, "global_step": 22072, "epoch": 248, "lr": 9.844397350378317e-05}
{"train_loss": 0.48699989914894104, "global_step": 22073, "epoch": 248, "lr": 9.844382999660213e-05}
{"train_loss": 0.5655003786087036, "global_step": 22074, "epoch": 248, "lr": 9.844368648290842e-05}
{"train_loss": 0.5374079346656799, "global_step": 22075, "epoch": 248, "lr": 9.844354296270204e-05}
{"train_loss": 0.5216435790061951, "global_step": 22076, "epoch": 248, "lr": 9.844339943598302e-05}
{"train_loss": 0.6006066799163818, "global_step": 22077, "epoch": 248, "lr": 9.844325590275136e-05}
{"train_loss": 0.42771559953689575, "global_step": 22078, "epoch": 248, "lr": 9.844311236300712e-05}
{"train_loss": 0.49601438641548157, "global_step": 22079, "epoch": 248, "lr": 9.844296881675028e-05}
{"train_loss": 0.3687460422515869, "global_step": 22080, "epoch": 248, "lr": 9.844282526398088e-05}
{"train_loss": 0.5663396716117859, "global_step": 22081, "epoch": 248, "lr": 9.844268170469892e-05}
{"train_loss": 0.4788186848163605, "global_step": 22082, "epoch": 248, "lr": 9.844253813890444e-05}
{"train_loss": 0.5048766136169434, "global_step": 22083, "epoch": 248, "lr": 9.844239456659746e-05}
{"train_loss": 0.5869572758674622, "global_step": 22084, "epoch": 248, "lr": 9.844225098777797e-05}
{"train_loss": 0.4130646288394928, "global_step": 22085, "epoch": 248, "lr": 9.844210740244602e-05}
{"train_loss": 0.48142343759536743, "global_step": 22086, "epoch": 248, "lr": 9.844196381060163e-05}
{"train_loss": 0.49425870180130005, "global_step": 22087, "epoch": 248, "lr": 9.844182021224479e-05}
{"train_loss": 0.5021440386772156, "global_step": 22088, "epoch": 248, "lr": 9.844167660737553e-05}
{"train_loss": 0.6131075024604797, "global_step": 22089, "epoch": 248, "lr": 9.84415329959939e-05}
{"train_loss": 0.5368223786354065, "global_step": 22090, "epoch": 248, "lr": 9.844138937809989e-05}
{"train_loss": 0.43983903527259827, "global_step": 22091, "epoch": 248, "lr": 9.844124575369353e-05}
{"train_loss": 0.4753507971763611, "global_step": 22092, "epoch": 248, "lr": 9.844110212277482e-05}
{"train_loss": 0.5519100427627563, "global_step": 22093, "epoch": 248, "lr": 9.84409584853438e-05}
{"train_loss": 0.4507460594177246, "global_step": 22094, "epoch": 248, "lr": 9.84408148414005e-05}
{"train_loss": 0.37174421548843384, "global_step": 22095, "epoch": 248, "lr": 9.84406711909449e-05}
{"train_loss": 0.49764126539230347, "global_step": 22096, "epoch": 248, "lr": 9.844052753397706e-05}
{"train_loss": 0.43954896926879883, "global_step": 22097, "epoch": 248, "lr": 9.844038387049697e-05}
{"train_loss": 0.42456528544425964, "global_step": 22098, "epoch": 248, "lr": 9.844024020050467e-05}
{"train_loss": 0.48829302191734314, "global_step": 22099, "epoch": 248, "lr": 9.844009652400016e-05}
{"train_loss": 0.563188910484314, "global_step": 22100, "epoch": 248, "lr": 9.843995284098348e-05}
{"train_loss": 0.4950617551803589, "global_step": 22101, "epoch": 248, "lr": 9.843980915145463e-05}
{"train_loss": 0.4090133011341095, "global_step": 22102, "epoch": 248, "lr": 9.843966545541363e-05}
{"train_loss": 0.5117688775062561, "global_step": 22103, "epoch": 248, "lr": 9.843952175286053e-05}
{"train_loss": 0.3810584247112274, "global_step": 22104, "epoch": 248, "lr": 9.843937804379531e-05}
{"train_loss": 0.4561823904514313, "global_step": 22105, "epoch": 248, "lr": 9.843923432821801e-05}
{"train_loss": 0.4541850686073303, "global_step": 22106, "epoch": 248, "lr": 9.843909060612865e-05}
{"train_loss": 0.4987116754055023, "global_step": 22107, "epoch": 248, "lr": 9.843894687752724e-05}
{"train_loss": 0.5299003720283508, "global_step": 22108, "epoch": 248, "lr": 9.84388031424138e-05}
{"train_loss": 0.3420451581478119, "global_step": 22109, "epoch": 248, "lr": 9.843865940078836e-05}
{"train_loss": 0.5227700471878052, "global_step": 22110, "epoch": 248, "lr": 9.843851565265093e-05}
{"train_loss": 0.45775511860847473, "global_step": 22111, "epoch": 248, "lr": 9.843837189800154e-05}
{"train_loss": 0.4869861900806427, "global_step": 22112, "epoch": 248, "lr": 9.843822813684018e-05}
{"train_loss": 0.513708233833313, "global_step": 22113, "epoch": 248, "lr": 9.843808436916691e-05}
{"train_loss": 0.5437915921211243, "global_step": 22114, "epoch": 248, "lr": 9.843794059498172e-05}
{"train_loss": 0.5749517679214478, "global_step": 22115, "epoch": 248, "lr": 9.843779681428464e-05}
{"train_loss": 0.45240169763565063, "global_step": 22116, "epoch": 248, "lr": 9.84376530270757e-05}
{"train_loss": 0.28960251808166504, "global_step": 22117, "epoch": 248, "lr": 9.843750923335491e-05}
{"train_loss": 0.45510387420654297, "global_step": 22118, "epoch": 248, "lr": 9.843736543312227e-05}
{"train_loss": 0.4262063205242157, "global_step": 22119, "epoch": 248, "lr": 9.843722162637783e-05}
{"train_loss": 0.4175635874271393, "global_step": 22120, "epoch": 248, "lr": 9.84370778131216e-05}
{"train_loss": 0.3733420670032501, "global_step": 22121, "epoch": 248, "lr": 9.843693399335357e-05}
{"train_loss": 0.4854740798473358, "global_step": 22122, "epoch": 248, "lr": 9.84367901670738e-05}
{"train_loss": 0.4383995532989502, "global_step": 22123, "epoch": 248, "lr": 9.84366463342823e-05}
{"train_loss": 0.4546158015727997, "global_step": 22124, "epoch": 248, "lr": 9.843650249497906e-05}
{"train_loss": 0.4595440626144409, "global_step": 22125, "epoch": 248, "lr": 9.843635864916415e-05}
{"train_loss": 0.5191709995269775, "global_step": 22126, "epoch": 248, "lr": 9.843621479683756e-05}
{"train_loss": 0.3714207410812378, "global_step": 22127, "epoch": 248, "lr": 9.84360709379993e-05}
{"train_loss": 0.46860846877098083, "global_step": 22128, "epoch": 248, "lr": 9.843592707264939e-05}
{"train_loss": 0.43689098954200745, "global_step": 22129, "epoch": 248, "lr": 9.843578320078788e-05}
{"train_loss": 0.3384692966938019, "global_step": 22130, "epoch": 248, "lr": 9.843563932241477e-05}
{"train_loss": 0.33468005061149597, "global_step": 22131, "epoch": 248, "lr": 9.843549543753007e-05}
{"train_loss": 0.4276915192604065, "global_step": 22132, "epoch": 248, "lr": 9.843535154613381e-05}
{"train_loss": 0.5336712002754211, "global_step": 22133, "epoch": 248, "lr": 9.843520764822601e-05}
{"train_loss": 0.35996150970458984, "global_step": 22134, "epoch": 248, "lr": 9.843506374380668e-05}
{"train_loss": 0.3800339102745056, "global_step": 22135, "epoch": 248, "lr": 9.843491983287586e-05}
{"train_loss": 0.441639244556427, "global_step": 22136, "epoch": 248, "lr": 9.843477591543353e-05}
{"train_loss": 0.44260919094085693, "global_step": 22137, "epoch": 248, "lr": 9.843463199147975e-05}
{"train_loss": 0.47607916593551636, "global_step": 22138, "epoch": 248, "lr": 9.843448806101452e-05}
{"train_loss": 0.32449039816856384, "global_step": 22139, "epoch": 248, "lr": 9.843434412403787e-05}
{"train_loss": 0.42356258630752563, "global_step": 22140, "epoch": 248, "lr": 9.84342001805498e-05}
{"train_loss": 0.32312121987342834, "global_step": 22141, "epoch": 248, "lr": 9.843405623055037e-05}
{"train_loss": 0.3330046534538269, "global_step": 22142, "epoch": 248, "lr": 9.843391227403955e-05}
{"train_loss": 0.4183390140533447, "global_step": 22143, "epoch": 248, "lr": 9.843376831101738e-05}
{"train_loss": 0.4059930741786957, "global_step": 22144, "epoch": 248, "lr": 9.843362434148389e-05}
{"train_loss": 0.41112858057022095, "global_step": 22145, "epoch": 248, "lr": 9.843348036543907e-05}
{"train_loss": 0.5293852686882019, "global_step": 22146, "epoch": 248, "lr": 9.843333638288299e-05}
{"train_loss": 0.44013798236846924, "global_step": 22147, "epoch": 248, "lr": 9.843319239381561e-05}
{"train_loss": 0.37967756390571594, "global_step": 22148, "epoch": 248, "lr": 9.8433048398237e-05}
{"train_loss": 0.44430387020111084, "global_step": 22149, "epoch": 248, "lr": 9.843290439614715e-05}
{"train_loss": 0.5488336086273193, "global_step": 22150, "epoch": 248, "lr": 9.843276038754607e-05}
{"train_loss": 0.3912917673587799, "global_step": 22151, "epoch": 248, "lr": 9.84326163724338e-05}
{"train_loss": 0.5439366698265076, "global_step": 22152, "epoch": 248, "lr": 9.843247235081037e-05}
{"train_loss": 0.41369807720184326, "global_step": 22153, "epoch": 248, "lr": 9.843232832267578e-05}
{"train_loss": 0.5062775611877441, "global_step": 22154, "epoch": 248, "lr": 9.843218428803005e-05}
{"train_loss": 0.347055584192276, "global_step": 22155, "epoch": 248, "lr": 9.84320402468732e-05}
{"train_loss": 0.4767381548881531, "global_step": 22156, "epoch": 248, "lr": 9.843189619920525e-05}
{"train_loss": 0.5718494653701782, "global_step": 22157, "epoch": 248, "lr": 9.843175214502624e-05}
{"train_loss": 0.6146952509880066, "global_step": 22158, "epoch": 248, "lr": 9.843160808433616e-05}
{"train_loss": 0.43866196274757385, "global_step": 22159, "epoch": 248, "lr": 9.843146401713504e-05}
{"train_loss": 0.4628858117575056, "global_step": 22160, "epoch": 248, "lr": 9.843131994342289e-05, "val_loss": 1.8857858180999756}
{"train_loss": 0.40766629576683044, "global_step": 22161, "epoch": 249, "lr": 9.843117586319976e-05}
{"train_loss": 0.420642226934433, "global_step": 22162, "epoch": 249, "lr": 9.843103177646563e-05}
{"train_loss": 0.38984763622283936, "global_step": 22163, "epoch": 249, "lr": 9.843088768322055e-05}
{"train_loss": 0.45405614376068115, "global_step": 22164, "epoch": 249, "lr": 9.843074358346451e-05}
{"train_loss": 0.41980695724487305, "global_step": 22165, "epoch": 249, "lr": 9.843059947719757e-05}
{"train_loss": 0.3028858006000519, "global_step": 22166, "epoch": 249, "lr": 9.843045536441972e-05}
{"train_loss": 0.38299524784088135, "global_step": 22167, "epoch": 249, "lr": 9.843031124513098e-05}
{"train_loss": 0.5061350464820862, "global_step": 22168, "epoch": 249, "lr": 9.843016711933137e-05}
{"train_loss": 0.495461642742157, "global_step": 22169, "epoch": 249, "lr": 9.843002298702093e-05}
{"train_loss": 0.4337088465690613, "global_step": 22170, "epoch": 249, "lr": 9.842987884819965e-05}
{"train_loss": 0.441449910402298, "global_step": 22171, "epoch": 249, "lr": 9.842973470286757e-05}
{"train_loss": 0.4749031066894531, "global_step": 22172, "epoch": 249, "lr": 9.84295905510247e-05}
{"train_loss": 0.3529841899871826, "global_step": 22173, "epoch": 249, "lr": 9.842944639267107e-05}
{"train_loss": 0.32319176197052, "global_step": 22174, "epoch": 249, "lr": 9.842930222780668e-05}
{"train_loss": 0.416157603263855, "global_step": 22175, "epoch": 249, "lr": 9.842915805643155e-05}
{"train_loss": 0.36942237615585327, "global_step": 22176, "epoch": 249, "lr": 9.842901387854572e-05}
{"train_loss": 0.38137203454971313, "global_step": 22177, "epoch": 249, "lr": 9.842886969414922e-05}
{"train_loss": 0.4503009021282196, "global_step": 22178, "epoch": 249, "lr": 9.842872550324202e-05}
{"train_loss": 0.4215763509273529, "global_step": 22179, "epoch": 249, "lr": 9.84285813058242e-05}
{"train_loss": 0.445823073387146, "global_step": 22180, "epoch": 249, "lr": 9.842843710189572e-05}
{"train_loss": 0.3364867866039276, "global_step": 22181, "epoch": 249, "lr": 9.842829289145663e-05}
{"train_loss": 0.393381267786026, "global_step": 22182, "epoch": 249, "lr": 9.842814867450696e-05}
{"train_loss": 0.4590618908405304, "global_step": 22183, "epoch": 249, "lr": 9.84280044510467e-05}
{"train_loss": 0.4177890717983246, "global_step": 22184, "epoch": 249, "lr": 9.842786022107589e-05}
{"train_loss": 0.46029746532440186, "global_step": 22185, "epoch": 249, "lr": 9.842771598459455e-05}
{"train_loss": 0.4458034336566925, "global_step": 22186, "epoch": 249, "lr": 9.842757174160268e-05}
{"train_loss": 0.45017126202583313, "global_step": 22187, "epoch": 249, "lr": 9.842742749210034e-05}
{"train_loss": 0.5179531574249268, "global_step": 22188, "epoch": 249, "lr": 9.84272832360875e-05}
{"train_loss": 0.43804931640625, "global_step": 22189, "epoch": 249, "lr": 9.842713897356421e-05}
{"train_loss": 0.5053710341453552, "global_step": 22190, "epoch": 249, "lr": 9.842699470453048e-05}
{"train_loss": 0.4319745898246765, "global_step": 22191, "epoch": 249, "lr": 9.842685042898633e-05}
{"train_loss": 0.44987353682518005, "global_step": 22192, "epoch": 249, "lr": 9.842670614693176e-05}
{"train_loss": 0.46048739552497864, "global_step": 22193, "epoch": 249, "lr": 9.842656185836685e-05}
{"train_loss": 0.38626253604888916, "global_step": 22194, "epoch": 249, "lr": 9.842641756329156e-05}
{"train_loss": 0.5423039793968201, "global_step": 22195, "epoch": 249, "lr": 9.842627326170593e-05}
{"train_loss": 0.481953501701355, "global_step": 22196, "epoch": 249, "lr": 9.842612895360997e-05}
{"train_loss": 0.4052482843399048, "global_step": 22197, "epoch": 249, "lr": 9.842598463900371e-05}
{"train_loss": 0.42791664600372314, "global_step": 22198, "epoch": 249, "lr": 9.842584031788718e-05}
{"train_loss": 0.46838417649269104, "global_step": 22199, "epoch": 249, "lr": 9.842569599026037e-05}
{"train_loss": 0.5242753624916077, "global_step": 22200, "epoch": 249, "lr": 9.842555165612331e-05}
{"train_loss": 0.4471384286880493, "global_step": 22201, "epoch": 249, "lr": 9.842540731547603e-05}
{"train_loss": 0.4846108555793762, "global_step": 22202, "epoch": 249, "lr": 9.842526296831856e-05}
{"train_loss": 0.4756684899330139, "global_step": 22203, "epoch": 249, "lr": 9.842511861465089e-05}
{"train_loss": 0.4483816623687744, "global_step": 22204, "epoch": 249, "lr": 9.842497425447305e-05}
{"train_loss": 0.46122556924819946, "global_step": 22205, "epoch": 249, "lr": 9.842482988778507e-05}
{"train_loss": 0.41832107305526733, "global_step": 22206, "epoch": 249, "lr": 9.842468551458697e-05}
{"train_loss": 0.5244141817092896, "global_step": 22207, "epoch": 249, "lr": 9.842454113487875e-05}
{"train_loss": 0.5573908090591431, "global_step": 22208, "epoch": 249, "lr": 9.842439674866045e-05}
{"train_loss": 0.3000132739543915, "global_step": 22209, "epoch": 249, "lr": 9.842425235593207e-05}
{"train_loss": 0.3677893280982971, "global_step": 22210, "epoch": 249, "lr": 9.842410795669364e-05}
{"train_loss": 0.39445924758911133, "global_step": 22211, "epoch": 249, "lr": 9.842396355094516e-05}
{"train_loss": 0.4228009283542633, "global_step": 22212, "epoch": 249, "lr": 9.84238191386867e-05}
{"train_loss": 0.4515516757965088, "global_step": 22213, "epoch": 249, "lr": 9.842367471991823e-05}
{"train_loss": 0.3372724652290344, "global_step": 22214, "epoch": 249, "lr": 9.84235302946398e-05}
{"train_loss": 0.3207095265388489, "global_step": 22215, "epoch": 249, "lr": 9.842338586285141e-05}
{"train_loss": 0.4129471182823181, "global_step": 22216, "epoch": 249, "lr": 9.842324142455308e-05}
{"train_loss": 0.46595507860183716, "global_step": 22217, "epoch": 249, "lr": 9.842309697974486e-05}
{"train_loss": 0.3576880395412445, "global_step": 22218, "epoch": 249, "lr": 9.842295252842671e-05}
{"train_loss": 0.5338165163993835, "global_step": 22219, "epoch": 249, "lr": 9.84228080705987e-05}
{"train_loss": 0.4827268421649933, "global_step": 22220, "epoch": 249, "lr": 9.842266360626084e-05}
{"train_loss": 0.3811681866645813, "global_step": 22221, "epoch": 249, "lr": 9.842251913541314e-05}
{"train_loss": 0.3695722818374634, "global_step": 22222, "epoch": 249, "lr": 9.842237465805561e-05}
{"train_loss": 0.5006733536720276, "global_step": 22223, "epoch": 249, "lr": 9.84222301741883e-05}
{"train_loss": 0.4922977387905121, "global_step": 22224, "epoch": 249, "lr": 9.84220856838112e-05}
{"train_loss": 0.42032089829444885, "global_step": 22225, "epoch": 249, "lr": 9.842194118692433e-05}
{"train_loss": 0.43873241543769836, "global_step": 22226, "epoch": 249, "lr": 9.842179668352775e-05}
{"train_loss": 0.25158172845840454, "global_step": 22227, "epoch": 249, "lr": 9.842165217362142e-05}
{"train_loss": 0.5117462873458862, "global_step": 22228, "epoch": 249, "lr": 9.842150765720542e-05}
{"train_loss": 0.3736547827720642, "global_step": 22229, "epoch": 249, "lr": 9.842136313427972e-05}
{"train_loss": 0.4904671609401703, "global_step": 22230, "epoch": 249, "lr": 9.842121860484436e-05}
{"train_loss": 0.47715845704078674, "global_step": 22231, "epoch": 249, "lr": 9.842107406889935e-05}
{"train_loss": 0.2986048460006714, "global_step": 22232, "epoch": 249, "lr": 9.842092952644473e-05}
{"train_loss": 0.4138251841068268, "global_step": 22233, "epoch": 249, "lr": 9.84207849774805e-05}
{"train_loss": 0.4682866036891937, "global_step": 22234, "epoch": 249, "lr": 9.842064042200668e-05}
{"train_loss": 0.4194665253162384, "global_step": 22235, "epoch": 249, "lr": 9.84204958600233e-05}
{"train_loss": 0.5454282164573669, "global_step": 22236, "epoch": 249, "lr": 9.842035129153038e-05}
{"train_loss": 0.38042402267456055, "global_step": 22237, "epoch": 249, "lr": 9.842020671652792e-05}
{"train_loss": 0.5170563459396362, "global_step": 22238, "epoch": 249, "lr": 9.842006213501597e-05}
{"train_loss": 0.4679003059864044, "global_step": 22239, "epoch": 249, "lr": 9.841991754699453e-05}
{"train_loss": 0.4538651406764984, "global_step": 22240, "epoch": 249, "lr": 9.841977295246361e-05}
{"train_loss": 0.49933043122291565, "global_step": 22241, "epoch": 249, "lr": 9.841962835142324e-05}
{"train_loss": 0.39089182019233704, "global_step": 22242, "epoch": 249, "lr": 9.841948374387346e-05}
{"train_loss": 0.48271578550338745, "global_step": 22243, "epoch": 249, "lr": 9.841933912981424e-05}
{"train_loss": 0.5134799480438232, "global_step": 22244, "epoch": 249, "lr": 9.841919450924566e-05}
{"train_loss": 0.3843594491481781, "global_step": 22245, "epoch": 249, "lr": 9.84190498821677e-05}
{"train_loss": 0.5041834115982056, "global_step": 22246, "epoch": 249, "lr": 9.841890524858039e-05}
{"train_loss": 0.4097684919834137, "global_step": 22247, "epoch": 249, "lr": 9.841876060848375e-05}
{"train_loss": 0.3822002708911896, "global_step": 22248, "epoch": 249, "lr": 9.841861596187781e-05}
{"train_loss": 0.43432237961319053, "global_step": 22249, "epoch": 249, "lr": 9.841847130876254e-05, "val_loss": 1.9017813205718994}
{"train_loss": 0.5296180844306946, "global_step": 22250, "epoch": 250, "lr": 9.841832664913803e-05}
{"train_loss": 0.4143437445163727, "global_step": 22251, "epoch": 250, "lr": 9.841818198300426e-05}
{"train_loss": 0.4926711916923523, "global_step": 22252, "epoch": 250, "lr": 9.841803731036123e-05}
{"train_loss": 0.37762048840522766, "global_step": 22253, "epoch": 250, "lr": 9.841789263120902e-05}
{"train_loss": 0.37519457936286926, "global_step": 22254, "epoch": 250, "lr": 9.84177479455476e-05}
{"train_loss": 0.4292335510253906, "global_step": 22255, "epoch": 250, "lr": 9.8417603253377e-05}
{"train_loss": 0.47300365567207336, "global_step": 22256, "epoch": 250, "lr": 9.841745855469724e-05}
{"train_loss": 0.356256902217865, "global_step": 22257, "epoch": 250, "lr": 9.841731384950837e-05}
{"train_loss": 0.5029295086860657, "global_step": 22258, "epoch": 250, "lr": 9.841716913781036e-05}
{"train_loss": 0.5351755619049072, "global_step": 22259, "epoch": 250, "lr": 9.841702441960324e-05}
{"train_loss": 0.39133358001708984, "global_step": 22260, "epoch": 250, "lr": 9.841687969488706e-05}
{"train_loss": 0.48736634850502014, "global_step": 22261, "epoch": 250, "lr": 9.841673496366182e-05}
{"train_loss": 0.5021609663963318, "global_step": 22262, "epoch": 250, "lr": 9.841659022592753e-05}
{"train_loss": 0.3878667652606964, "global_step": 22263, "epoch": 250, "lr": 9.841644548168423e-05}
{"train_loss": 0.5161229968070984, "global_step": 22264, "epoch": 250, "lr": 9.841630073093191e-05}
{"train_loss": 0.4479747712612152, "global_step": 22265, "epoch": 250, "lr": 9.841615597367063e-05}
{"train_loss": 0.5141919255256653, "global_step": 22266, "epoch": 250, "lr": 9.841601120990039e-05}
{"train_loss": 0.5759507417678833, "global_step": 22267, "epoch": 250, "lr": 9.84158664396212e-05}
{"train_loss": 0.3659243881702423, "global_step": 22268, "epoch": 250, "lr": 9.841572166283309e-05}
{"train_loss": 0.3545018434524536, "global_step": 22269, "epoch": 250, "lr": 9.841557687953606e-05}
{"train_loss": 0.39218005537986755, "global_step": 22270, "epoch": 250, "lr": 9.841543208973015e-05}
{"train_loss": 0.40559402108192444, "global_step": 22271, "epoch": 250, "lr": 9.841528729341537e-05}
{"train_loss": 0.27602171897888184, "global_step": 22272, "epoch": 250, "lr": 9.841514249059177e-05}
{"train_loss": 0.47427430748939514, "global_step": 22273, "epoch": 250, "lr": 9.841499768125934e-05}
{"train_loss": 0.4663877785205841, "global_step": 22274, "epoch": 250, "lr": 9.84148528654181e-05}
{"train_loss": 0.49312224984169006, "global_step": 22275, "epoch": 250, "lr": 9.841470804306805e-05}
{"train_loss": 0.42361852526664734, "global_step": 22276, "epoch": 250, "lr": 9.841456321420925e-05}
{"train_loss": 0.3961314558982849, "global_step": 22277, "epoch": 250, "lr": 9.841441837884171e-05}
{"train_loss": 0.3020861744880676, "global_step": 22278, "epoch": 250, "lr": 9.841427353696544e-05}
{"train_loss": 0.31093886494636536, "global_step": 22279, "epoch": 250, "lr": 9.841412868858045e-05}
{"train_loss": 0.4157300591468811, "global_step": 22280, "epoch": 250, "lr": 9.841398383368677e-05}
{"train_loss": 0.3908035159111023, "global_step": 22281, "epoch": 250, "lr": 9.841383897228443e-05}
{"train_loss": 0.49565058946609497, "global_step": 22282, "epoch": 250, "lr": 9.841369410437345e-05}
{"train_loss": 0.39732345938682556, "global_step": 22283, "epoch": 250, "lr": 9.84135492299538e-05}
{"train_loss": 0.4149066209793091, "global_step": 22284, "epoch": 250, "lr": 9.841340434902557e-05}
{"train_loss": 0.5654969811439514, "global_step": 22285, "epoch": 250, "lr": 9.841325946158875e-05}
{"train_loss": 0.4319501221179962, "global_step": 22286, "epoch": 250, "lr": 9.841311456764334e-05}
{"train_loss": 0.4980052709579468, "global_step": 22287, "epoch": 250, "lr": 9.841296966718938e-05}
{"train_loss": 0.4201425313949585, "global_step": 22288, "epoch": 250, "lr": 9.841282476022689e-05}
{"train_loss": 0.3809369206428528, "global_step": 22289, "epoch": 250, "lr": 9.84126798467559e-05}
{"train_loss": 0.5721828937530518, "global_step": 22290, "epoch": 250, "lr": 9.84125349267764e-05}
{"train_loss": 0.4812890291213989, "global_step": 22291, "epoch": 250, "lr": 9.841239000028842e-05}
{"train_loss": 0.4786446988582611, "global_step": 22292, "epoch": 250, "lr": 9.841224506729199e-05}
{"train_loss": 0.35594847798347473, "global_step": 22293, "epoch": 250, "lr": 9.841210012778712e-05}
{"train_loss": 0.44751840829849243, "global_step": 22294, "epoch": 250, "lr": 9.841195518177384e-05}
{"train_loss": 0.3450948894023895, "global_step": 22295, "epoch": 250, "lr": 9.841181022925217e-05}
{"train_loss": 0.3644096553325653, "global_step": 22296, "epoch": 250, "lr": 9.841166527022211e-05}
{"train_loss": 0.43402746319770813, "global_step": 22297, "epoch": 250, "lr": 9.841152030468369e-05}
{"train_loss": 0.503605842590332, "global_step": 22298, "epoch": 250, "lr": 9.841137533263694e-05}
{"train_loss": 0.5719815492630005, "global_step": 22299, "epoch": 250, "lr": 9.841123035408187e-05}
{"train_loss": 0.4758640229701996, "global_step": 22300, "epoch": 250, "lr": 9.841108536901848e-05}
{"train_loss": 0.3955692648887634, "global_step": 22301, "epoch": 250, "lr": 9.841094037744684e-05}
{"train_loss": 0.34435614943504333, "global_step": 22302, "epoch": 250, "lr": 9.841079537936693e-05}
{"train_loss": 0.43490079045295715, "global_step": 22303, "epoch": 250, "lr": 9.841065037477877e-05}
{"train_loss": 0.44469934701919556, "global_step": 22304, "epoch": 250, "lr": 9.84105053636824e-05}
{"train_loss": 0.4065261483192444, "global_step": 22305, "epoch": 250, "lr": 9.841036034607781e-05}
{"train_loss": 0.3156077563762665, "global_step": 22306, "epoch": 250, "lr": 9.841021532196504e-05}
{"train_loss": 0.46403804421424866, "global_step": 22307, "epoch": 250, "lr": 9.841007029134412e-05}
{"train_loss": 0.5667324066162109, "global_step": 22308, "epoch": 250, "lr": 9.840992525421504e-05}
{"train_loss": 0.45955604314804077, "global_step": 22309, "epoch": 250, "lr": 9.840978021057785e-05}
{"train_loss": 0.481920063495636, "global_step": 22310, "epoch": 250, "lr": 9.840963516043255e-05}
{"train_loss": 0.51522296667099, "global_step": 22311, "epoch": 250, "lr": 9.840949010377916e-05}
{"train_loss": 0.36134880781173706, "global_step": 22312, "epoch": 250, "lr": 9.840934504061771e-05}
{"train_loss": 0.4480692446231842, "global_step": 22313, "epoch": 250, "lr": 9.840919997094822e-05}
{"train_loss": 0.4894965887069702, "global_step": 22314, "epoch": 250, "lr": 9.840905489477068e-05}
{"train_loss": 0.49251464009284973, "global_step": 22315, "epoch": 250, "lr": 9.840890981208515e-05}
{"train_loss": 0.38348227739334106, "global_step": 22316, "epoch": 250, "lr": 9.840876472289163e-05}
{"train_loss": 0.5847442746162415, "global_step": 22317, "epoch": 250, "lr": 9.840861962719014e-05}
{"train_loss": 0.4625604748725891, "global_step": 22318, "epoch": 250, "lr": 9.840847452498071e-05}
{"train_loss": 0.44533878564834595, "global_step": 22319, "epoch": 250, "lr": 9.840832941626335e-05}
{"train_loss": 0.5127644538879395, "global_step": 22320, "epoch": 250, "lr": 9.840818430103806e-05}
{"train_loss": 0.3600377142429352, "global_step": 22321, "epoch": 250, "lr": 9.84080391793049e-05}
{"train_loss": 0.28070640563964844, "global_step": 22322, "epoch": 250, "lr": 9.840789405106385e-05}
{"train_loss": 0.5058813691139221, "global_step": 22323, "epoch": 250, "lr": 9.840774891631497e-05}
{"train_loss": 0.4974011778831482, "global_step": 22324, "epoch": 250, "lr": 9.840760377505824e-05}
{"train_loss": 0.3936963379383087, "global_step": 22325, "epoch": 250, "lr": 9.840745862729372e-05}
{"train_loss": 0.4799850583076477, "global_step": 22326, "epoch": 250, "lr": 9.84073134730214e-05}
{"train_loss": 0.5219365954399109, "global_step": 22327, "epoch": 250, "lr": 9.840716831224128e-05}
{"train_loss": 0.49537885189056396, "global_step": 22328, "epoch": 250, "lr": 9.840702314495344e-05}
{"train_loss": 0.409111350774765, "global_step": 22329, "epoch": 250, "lr": 9.840687797115784e-05}
{"train_loss": 0.4781181812286377, "global_step": 22330, "epoch": 250, "lr": 9.840673279085455e-05}
{"train_loss": 0.5418078303337097, "global_step": 22331, "epoch": 250, "lr": 9.840658760404355e-05}
{"train_loss": 0.38072526454925537, "global_step": 22332, "epoch": 250, "lr": 9.840644241072487e-05}
{"train_loss": 0.47222670912742615, "global_step": 22333, "epoch": 250, "lr": 9.840629721089854e-05}
{"train_loss": 0.5228585600852966, "global_step": 22334, "epoch": 250, "lr": 9.840615200456458e-05}
{"train_loss": 0.46581029891967773, "global_step": 22335, "epoch": 250, "lr": 9.840600679172299e-05}
{"train_loss": 0.4435073137283325, "global_step": 22336, "epoch": 250, "lr": 9.840586157237382e-05}
{"train_loss": 0.42707788944244385, "global_step": 22337, "epoch": 250, "lr": 9.840571634651705e-05}
{"train_loss": 0.44513373194115885, "global_step": 22338, "epoch": 250, "lr": 9.840557111415275e-05, "train/sim_max_reward_0": 0.7064739742925265, "train/sim_max_reward_1": 0.6150822923513918, "train/sim_max_reward_2": 0.4882681316448172, "train/sim_max_reward_3": 0.6090450073919859, "train/sim_max_reward_4": 0.9685763064151995, "train/sim_max_reward_5": 0.7177995919998239, "test/sim_max_reward_4500000": 0.9213978536388142, "test/sim_max_reward_4500001": 0.9496798085965353, "test/sim_max_reward_4500002": 0.9610140144134053, "test/sim_max_reward_4500003": 0.5318604763092941, "test/sim_max_reward_4500004": 0.45674776006379886, "test/sim_max_reward_4500005": 0.2694213576936069, "test/sim_max_reward_4500006": 0.0, "test/sim_max_reward_4500007": 0.49546420276535397, "test/sim_max_reward_4500008": 0.4591427387937918, "test/sim_max_reward_4500009": 0.8134103544831013, "test/sim_max_reward_4500010": 0.214658992595451, "test/sim_max_reward_4500011": 0.962650142664688, "test/sim_max_reward_4500012": 0.9829420347326645, "test/sim_max_reward_4500013": 0.5666165048639547, "test/sim_max_reward_4500014": 0.9570241844422666, "test/sim_max_reward_4500015": 0.6909173161687295, "test/sim_max_reward_4500016": 0.19654604370487055, "test/sim_max_reward_4500017": 0.7019085539010774, "test/sim_max_reward_4500018": 0.46832522333274873, "test/sim_max_reward_4500019": 0.925043944162693, "test/sim_max_reward_4500020": 0.9890191795563802, "test/sim_max_reward_4500021": 0.8556639256297245, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 1.0, "test/sim_max_reward_4500024": 0.6493684508480756, "test/sim_max_reward_4500025": 0.5954762547308884, "test/sim_max_reward_4500026": 0.5389187920095899, "test/sim_max_reward_4500027": 0.9992840326528739, "test/sim_max_reward_4500028": 0.8986299924686605, "test/sim_max_reward_4500029": 0.12774321795275415, "test/sim_max_reward_4500030": 0.9956686814913818, "test/sim_max_reward_4500031": 0.12678218884763903, "test/sim_max_reward_4500032": 0.4474995696422705, "test/sim_max_reward_4500033": 0.6776416611770503, "test/sim_max_reward_4500034": 0.40477630304517714, "test/sim_max_reward_4500035": 0.6893640756473846, "test/sim_max_reward_4500036": 0.9059274744008157, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.6869375540136146, "test/sim_max_reward_4500039": 0.9071007972887645, "test/sim_max_reward_4500040": 0.9521048440596923, "test/sim_max_reward_4500041": 0.8952525262886255, "test/sim_max_reward_4500042": 0.0018570030276895484, "test/sim_max_reward_4500043": 0.6465150936057836, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.9450775447113916, "test/sim_max_reward_4500046": 0.649767164521032, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9857904727429895, "test/sim_max_reward_4500049": 0.10165681944744036, "train/mean_score": 0.6842075506826242, "test/mean_score": 0.6239719025426907, "val_loss": 1.9683315753936768, "train_action_mse_error": 35.38774871826172}
{"train_loss": 0.44634827971458435, "global_step": 22339, "epoch": 251, "lr": 9.840542587528087e-05}
{"train_loss": 0.3927403390407562, "global_step": 22340, "epoch": 251, "lr": 9.840528062990151e-05}
{"train_loss": 0.42401552200317383, "global_step": 22341, "epoch": 251, "lr": 9.840513537801463e-05}
{"train_loss": 0.5572267174720764, "global_step": 22342, "epoch": 251, "lr": 9.840499011962027e-05}
{"train_loss": 0.42938727140426636, "global_step": 22343, "epoch": 251, "lr": 9.840484485471845e-05}
{"train_loss": 0.5517801642417908, "global_step": 22344, "epoch": 251, "lr": 9.84046995833092e-05}
{"train_loss": 0.48052525520324707, "global_step": 22345, "epoch": 251, "lr": 9.840455430539251e-05}
{"train_loss": 0.4194619655609131, "global_step": 22346, "epoch": 251, "lr": 9.840440902096844e-05}
{"train_loss": 0.49170830845832825, "global_step": 22347, "epoch": 251, "lr": 9.840426373003696e-05}
{"train_loss": 0.5182525515556335, "global_step": 22348, "epoch": 251, "lr": 9.840411843259814e-05}
{"train_loss": 0.5192556381225586, "global_step": 22349, "epoch": 251, "lr": 9.840397312865196e-05}
{"train_loss": 0.4646989405155182, "global_step": 22350, "epoch": 251, "lr": 9.840382781819846e-05}
{"train_loss": 0.41213396191596985, "global_step": 22351, "epoch": 251, "lr": 9.840368250123765e-05}
{"train_loss": 0.4494560956954956, "global_step": 22352, "epoch": 251, "lr": 9.840353717776958e-05}
{"train_loss": 0.4651307165622711, "global_step": 22353, "epoch": 251, "lr": 9.840339184779422e-05}
{"train_loss": 0.42746061086654663, "global_step": 22354, "epoch": 251, "lr": 9.840324651131162e-05}
{"train_loss": 0.5015367865562439, "global_step": 22355, "epoch": 251, "lr": 9.84031011683218e-05}
{"train_loss": 0.4374672770500183, "global_step": 22356, "epoch": 251, "lr": 9.840295581882476e-05}
{"train_loss": 0.5510852932929993, "global_step": 22357, "epoch": 251, "lr": 9.840281046282054e-05}
{"train_loss": 0.4668339788913727, "global_step": 22358, "epoch": 251, "lr": 9.840266510030915e-05}
{"train_loss": 0.477862149477005, "global_step": 22359, "epoch": 251, "lr": 9.840251973129062e-05}
{"train_loss": 0.35292112827301025, "global_step": 22360, "epoch": 251, "lr": 9.840237435576496e-05}
{"train_loss": 0.3877037763595581, "global_step": 22361, "epoch": 251, "lr": 9.840222897373218e-05}
{"train_loss": 0.41282564401626587, "global_step": 22362, "epoch": 251, "lr": 9.840208358519231e-05}
{"train_loss": 0.502720057964325, "global_step": 22363, "epoch": 251, "lr": 9.840193819014539e-05}
{"train_loss": 0.47327229380607605, "global_step": 22364, "epoch": 251, "lr": 9.84017927885914e-05}
{"train_loss": 0.3936188519001007, "global_step": 22365, "epoch": 251, "lr": 9.840164738053038e-05}
{"train_loss": 0.47310203313827515, "global_step": 22366, "epoch": 251, "lr": 9.840150196596235e-05}
{"train_loss": 0.3822614848613739, "global_step": 22367, "epoch": 251, "lr": 9.840135654488733e-05}
{"train_loss": 0.4812850058078766, "global_step": 22368, "epoch": 251, "lr": 9.840121111730534e-05}
{"train_loss": 0.35204821825027466, "global_step": 22369, "epoch": 251, "lr": 9.840106568321639e-05}
{"train_loss": 0.38015875220298767, "global_step": 22370, "epoch": 251, "lr": 9.840092024262052e-05}
{"train_loss": 0.48077476024627686, "global_step": 22371, "epoch": 251, "lr": 9.840077479551772e-05}
{"train_loss": 0.5729056000709534, "global_step": 22372, "epoch": 251, "lr": 9.840062934190803e-05}
{"train_loss": 0.3249993622303009, "global_step": 22373, "epoch": 251, "lr": 9.840048388179148e-05}
{"train_loss": 0.42401206493377686, "global_step": 22374, "epoch": 251, "lr": 9.840033841516806e-05}
{"train_loss": 0.43399158120155334, "global_step": 22375, "epoch": 251, "lr": 9.84001929420378e-05}
{"train_loss": 0.40494823455810547, "global_step": 22376, "epoch": 251, "lr": 9.840004746240074e-05}
{"train_loss": 0.36265480518341064, "global_step": 22377, "epoch": 251, "lr": 9.839990197625689e-05}
{"train_loss": 0.43568259477615356, "global_step": 22378, "epoch": 251, "lr": 9.839975648360623e-05}
{"train_loss": 0.4706885814666748, "global_step": 22379, "epoch": 251, "lr": 9.839961098444883e-05}
{"train_loss": 0.4109352231025696, "global_step": 22380, "epoch": 251, "lr": 9.839946547878469e-05}
{"train_loss": 0.5586298108100891, "global_step": 22381, "epoch": 251, "lr": 9.839931996661385e-05}
{"train_loss": 0.3615531921386719, "global_step": 22382, "epoch": 251, "lr": 9.83991744479363e-05}
{"train_loss": 0.4524686932563782, "global_step": 22383, "epoch": 251, "lr": 9.839902892275206e-05}
{"train_loss": 0.44003137946128845, "global_step": 22384, "epoch": 251, "lr": 9.839888339106116e-05}
{"train_loss": 0.46099579334259033, "global_step": 22385, "epoch": 251, "lr": 9.839873785286364e-05}
{"train_loss": 0.6029316186904907, "global_step": 22386, "epoch": 251, "lr": 9.839859230815949e-05}
{"train_loss": 0.4077986478805542, "global_step": 22387, "epoch": 251, "lr": 9.839844675694874e-05}
{"train_loss": 0.46550536155700684, "global_step": 22388, "epoch": 251, "lr": 9.839830119923138e-05}
{"train_loss": 0.42131519317626953, "global_step": 22389, "epoch": 251, "lr": 9.839815563500749e-05}
{"train_loss": 0.4196797013282776, "global_step": 22390, "epoch": 251, "lr": 9.839801006427706e-05}
{"train_loss": 0.47701117396354675, "global_step": 22391, "epoch": 251, "lr": 9.839786448704009e-05}
{"train_loss": 0.5178043246269226, "global_step": 22392, "epoch": 251, "lr": 9.839771890329663e-05}
{"train_loss": 0.38221845030784607, "global_step": 22393, "epoch": 251, "lr": 9.839757331304668e-05}
{"train_loss": 0.40517285466194153, "global_step": 22394, "epoch": 251, "lr": 9.839742771629028e-05}
{"train_loss": 0.34885740280151367, "global_step": 22395, "epoch": 251, "lr": 9.839728211302742e-05}
{"train_loss": 0.5092005133628845, "global_step": 22396, "epoch": 251, "lr": 9.839713650325815e-05}
{"train_loss": 0.4547062814235687, "global_step": 22397, "epoch": 251, "lr": 9.839699088698245e-05}
{"train_loss": 0.32956498861312866, "global_step": 22398, "epoch": 251, "lr": 9.839684526420039e-05}
{"train_loss": 0.542121946811676, "global_step": 22399, "epoch": 251, "lr": 9.839669963491195e-05}
{"train_loss": 0.4475118815898895, "global_step": 22400, "epoch": 251, "lr": 9.839655399911715e-05}
{"train_loss": 0.5530659556388855, "global_step": 22401, "epoch": 251, "lr": 9.839640835681605e-05}
{"train_loss": 0.4627757966518402, "global_step": 22402, "epoch": 251, "lr": 9.839626270800864e-05}
{"train_loss": 0.48024341464042664, "global_step": 22403, "epoch": 251, "lr": 9.839611705269494e-05}
{"train_loss": 0.3753165006637573, "global_step": 22404, "epoch": 251, "lr": 9.839597139087495e-05}
{"train_loss": 0.41246935725212097, "global_step": 22405, "epoch": 251, "lr": 9.839582572254874e-05}
{"train_loss": 0.5287114381790161, "global_step": 22406, "epoch": 251, "lr": 9.839568004771628e-05}
{"train_loss": 0.3962969183921814, "global_step": 22407, "epoch": 251, "lr": 9.839553436637763e-05}
{"train_loss": 0.517361581325531, "global_step": 22408, "epoch": 251, "lr": 9.839538867853277e-05}
{"train_loss": 0.4889437258243561, "global_step": 22409, "epoch": 251, "lr": 9.839524298418176e-05}
{"train_loss": 0.37317153811454773, "global_step": 22410, "epoch": 251, "lr": 9.83950972833246e-05}
{"train_loss": 0.5242384672164917, "global_step": 22411, "epoch": 251, "lr": 9.839495157596128e-05}
{"train_loss": 0.5133577585220337, "global_step": 22412, "epoch": 251, "lr": 9.839480586209188e-05}
{"train_loss": 0.4343753457069397, "global_step": 22413, "epoch": 251, "lr": 9.839466014171636e-05}
{"train_loss": 0.4970935583114624, "global_step": 22414, "epoch": 251, "lr": 9.839451441483479e-05}
{"train_loss": 0.44660159945487976, "global_step": 22415, "epoch": 251, "lr": 9.839436868144717e-05}
{"train_loss": 0.48525509238243103, "global_step": 22416, "epoch": 251, "lr": 9.83942229415535e-05}
{"train_loss": 0.4130658805370331, "global_step": 22417, "epoch": 251, "lr": 9.839407719515383e-05}
{"train_loss": 0.5946754217147827, "global_step": 22418, "epoch": 251, "lr": 9.839393144224815e-05}
{"train_loss": 0.46253839135169983, "global_step": 22419, "epoch": 251, "lr": 9.839378568283652e-05}
{"train_loss": 0.39329349994659424, "global_step": 22420, "epoch": 251, "lr": 9.839363991691892e-05}
{"train_loss": 0.5497420430183411, "global_step": 22421, "epoch": 251, "lr": 9.839349414449539e-05}
{"train_loss": 0.47342318296432495, "global_step": 22422, "epoch": 251, "lr": 9.839334836556593e-05}
{"train_loss": 0.4520665407180786, "global_step": 22423, "epoch": 251, "lr": 9.839320258013059e-05}
{"train_loss": 0.4694848954677582, "global_step": 22424, "epoch": 251, "lr": 9.839305678818938e-05}
{"train_loss": 0.5350807309150696, "global_step": 22425, "epoch": 251, "lr": 9.839291098974229e-05}
{"train_loss": 0.43770021200180054, "global_step": 22426, "epoch": 251, "lr": 9.839276518478937e-05}
{"train_loss": 0.45553197016876734, "global_step": 22427, "epoch": 251, "lr": 9.839261937333063e-05, "val_loss": 1.887873888015747}
{"train_loss": 0.3357868790626526, "global_step": 22428, "epoch": 252, "lr": 9.839247355536612e-05}
{"train_loss": 0.4402327835559845, "global_step": 22429, "epoch": 252, "lr": 9.839232773089581e-05}
{"train_loss": 0.49778231978416443, "global_step": 22430, "epoch": 252, "lr": 9.839218189991974e-05}
{"train_loss": 0.5005930662155151, "global_step": 22431, "epoch": 252, "lr": 9.839203606243794e-05}
{"train_loss": 0.57687908411026, "global_step": 22432, "epoch": 252, "lr": 9.839189021845042e-05}
{"train_loss": 0.5075188279151917, "global_step": 22433, "epoch": 252, "lr": 9.839174436795718e-05}
{"train_loss": 0.4698615074157715, "global_step": 22434, "epoch": 252, "lr": 9.839159851095827e-05}
{"train_loss": 0.3812309801578522, "global_step": 22435, "epoch": 252, "lr": 9.839145264745371e-05}
{"train_loss": 0.44625693559646606, "global_step": 22436, "epoch": 252, "lr": 9.839130677744351e-05}
{"train_loss": 0.4803514778614044, "global_step": 22437, "epoch": 252, "lr": 9.839116090092769e-05}
{"train_loss": 0.4151879847049713, "global_step": 22438, "epoch": 252, "lr": 9.839101501790624e-05}
{"train_loss": 0.4389922022819519, "global_step": 22439, "epoch": 252, "lr": 9.839086912837924e-05}
{"train_loss": 0.5095128417015076, "global_step": 22440, "epoch": 252, "lr": 9.839072323234666e-05}
{"train_loss": 0.41046470403671265, "global_step": 22441, "epoch": 252, "lr": 9.839057732980854e-05}
{"train_loss": 0.39637240767478943, "global_step": 22442, "epoch": 252, "lr": 9.83904314207649e-05}
{"train_loss": 0.4572697579860687, "global_step": 22443, "epoch": 252, "lr": 9.839028550521577e-05}
{"train_loss": 0.5428071022033691, "global_step": 22444, "epoch": 252, "lr": 9.839013958316112e-05}
{"train_loss": 0.459473192691803, "global_step": 22445, "epoch": 252, "lr": 9.838999365460105e-05}
{"train_loss": 0.4634619951248169, "global_step": 22446, "epoch": 252, "lr": 9.83898477195355e-05}
{"train_loss": 0.3220215439796448, "global_step": 22447, "epoch": 252, "lr": 9.838970177796454e-05}
{"train_loss": 0.4555554986000061, "global_step": 22448, "epoch": 252, "lr": 9.838955582988816e-05}
{"train_loss": 0.32770925760269165, "global_step": 22449, "epoch": 252, "lr": 9.838940987530641e-05}
{"train_loss": 0.26886531710624695, "global_step": 22450, "epoch": 252, "lr": 9.838926391421929e-05}
{"train_loss": 0.455835223197937, "global_step": 22451, "epoch": 252, "lr": 9.838911794662681e-05}
{"train_loss": 0.39273762702941895, "global_step": 22452, "epoch": 252, "lr": 9.838897197252903e-05}
{"train_loss": 0.4464501142501831, "global_step": 22453, "epoch": 252, "lr": 9.838882599192592e-05}
{"train_loss": 0.40969952940940857, "global_step": 22454, "epoch": 252, "lr": 9.838868000481753e-05}
{"train_loss": 0.36376819014549255, "global_step": 22455, "epoch": 252, "lr": 9.838853401120388e-05}
{"train_loss": 0.4566963016986847, "global_step": 22456, "epoch": 252, "lr": 9.838838801108497e-05}
{"train_loss": 0.4895038604736328, "global_step": 22457, "epoch": 252, "lr": 9.838824200446083e-05}
{"train_loss": 0.32322371006011963, "global_step": 22458, "epoch": 252, "lr": 9.838809599133147e-05}
{"train_loss": 0.48965248465538025, "global_step": 22459, "epoch": 252, "lr": 9.838794997169695e-05}
{"train_loss": 0.3511095941066742, "global_step": 22460, "epoch": 252, "lr": 9.838780394555724e-05}
{"train_loss": 0.4537588953971863, "global_step": 22461, "epoch": 252, "lr": 9.838765791291237e-05}
{"train_loss": 0.4238303601741791, "global_step": 22462, "epoch": 252, "lr": 9.838751187376239e-05}
{"train_loss": 0.4120231568813324, "global_step": 22463, "epoch": 252, "lr": 9.83873658281073e-05}
{"train_loss": 0.507559061050415, "global_step": 22464, "epoch": 252, "lr": 9.838721977594712e-05}
{"train_loss": 0.384293794631958, "global_step": 22465, "epoch": 252, "lr": 9.838707371728184e-05}
{"train_loss": 0.4761318862438202, "global_step": 22466, "epoch": 252, "lr": 9.838692765211153e-05}
{"train_loss": 0.5233504772186279, "global_step": 22467, "epoch": 252, "lr": 9.83867815804362e-05}
{"train_loss": 0.5193896889686584, "global_step": 22468, "epoch": 252, "lr": 9.838663550225582e-05}
{"train_loss": 0.4612843096256256, "global_step": 22469, "epoch": 252, "lr": 9.838648941757047e-05}
{"train_loss": 0.4192592203617096, "global_step": 22470, "epoch": 252, "lr": 9.838634332638015e-05}
{"train_loss": 0.4331497848033905, "global_step": 22471, "epoch": 252, "lr": 9.838619722868487e-05}
{"train_loss": 0.5005909204483032, "global_step": 22472, "epoch": 252, "lr": 9.838605112448464e-05}
{"train_loss": 0.45257994532585144, "global_step": 22473, "epoch": 252, "lr": 9.838590501377952e-05}
{"train_loss": 0.4157085418701172, "global_step": 22474, "epoch": 252, "lr": 9.838575889656949e-05}
{"train_loss": 0.5229392647743225, "global_step": 22475, "epoch": 252, "lr": 9.838561277285458e-05}
{"train_loss": 0.45800089836120605, "global_step": 22476, "epoch": 252, "lr": 9.838546664263484e-05}
{"train_loss": 0.5021874904632568, "global_step": 22477, "epoch": 252, "lr": 9.838532050591025e-05}
{"train_loss": 0.5173099040985107, "global_step": 22478, "epoch": 252, "lr": 9.838517436268083e-05}
{"train_loss": 0.4570489823818207, "global_step": 22479, "epoch": 252, "lr": 9.838502821294663e-05}
{"train_loss": 0.3973601460456848, "global_step": 22480, "epoch": 252, "lr": 9.838488205670763e-05}
{"train_loss": 0.5481327772140503, "global_step": 22481, "epoch": 252, "lr": 9.838473589396391e-05}
{"train_loss": 0.5448083281517029, "global_step": 22482, "epoch": 252, "lr": 9.838458972471543e-05}
{"train_loss": 0.46062368154525757, "global_step": 22483, "epoch": 252, "lr": 9.838444354896222e-05}
{"train_loss": 0.4085268974304199, "global_step": 22484, "epoch": 252, "lr": 9.838429736670432e-05}
{"train_loss": 0.6148179173469543, "global_step": 22485, "epoch": 252, "lr": 9.838415117794175e-05}
{"train_loss": 0.5135096907615662, "global_step": 22486, "epoch": 252, "lr": 9.838400498267451e-05}
{"train_loss": 0.4739680588245392, "global_step": 22487, "epoch": 252, "lr": 9.838385878090264e-05}
{"train_loss": 0.4589685797691345, "global_step": 22488, "epoch": 252, "lr": 9.838371257262616e-05}
{"train_loss": 0.46075645089149475, "global_step": 22489, "epoch": 252, "lr": 9.838356635784505e-05}
{"train_loss": 0.5065102577209473, "global_step": 22490, "epoch": 252, "lr": 9.838342013655937e-05}
{"train_loss": 0.5631449818611145, "global_step": 22491, "epoch": 252, "lr": 9.838327390876913e-05}
{"train_loss": 0.41911283135414124, "global_step": 22492, "epoch": 252, "lr": 9.838312767447436e-05}
{"train_loss": 0.31203699111938477, "global_step": 22493, "epoch": 252, "lr": 9.838298143367506e-05}
{"train_loss": 0.44081076979637146, "global_step": 22494, "epoch": 252, "lr": 9.838283518637126e-05}
{"train_loss": 0.3975408673286438, "global_step": 22495, "epoch": 252, "lr": 9.838268893256297e-05}
{"train_loss": 0.42148691415786743, "global_step": 22496, "epoch": 252, "lr": 9.838254267225022e-05}
{"train_loss": 0.5063238143920898, "global_step": 22497, "epoch": 252, "lr": 9.838239640543304e-05}
{"train_loss": 0.5035456418991089, "global_step": 22498, "epoch": 252, "lr": 9.838225013211142e-05}
{"train_loss": 0.4523613452911377, "global_step": 22499, "epoch": 252, "lr": 9.83821038522854e-05}
{"train_loss": 0.45668312907218933, "global_step": 22500, "epoch": 252, "lr": 9.838195756595499e-05}
{"train_loss": 0.3799041211605072, "global_step": 22501, "epoch": 252, "lr": 9.838181127312022e-05}
{"train_loss": 0.4002498686313629, "global_step": 22502, "epoch": 252, "lr": 9.838166497378112e-05}
{"train_loss": 0.41445645689964294, "global_step": 22503, "epoch": 252, "lr": 9.838151866793768e-05}
{"train_loss": 0.43972229957580566, "global_step": 22504, "epoch": 252, "lr": 9.838137235558993e-05}
{"train_loss": 0.5105878710746765, "global_step": 22505, "epoch": 252, "lr": 9.838122603673789e-05}
{"train_loss": 0.47341659665107727, "global_step": 22506, "epoch": 252, "lr": 9.838107971138161e-05}
{"train_loss": 0.503671407699585, "global_step": 22507, "epoch": 252, "lr": 9.838093337952107e-05}
{"train_loss": 0.3929787278175354, "global_step": 22508, "epoch": 252, "lr": 9.83807870411563e-05}
{"train_loss": 0.4783841669559479, "global_step": 22509, "epoch": 252, "lr": 9.838064069628731e-05}
{"train_loss": 0.5179870128631592, "global_step": 22510, "epoch": 252, "lr": 9.838049434491417e-05}
{"train_loss": 0.5188072919845581, "global_step": 22511, "epoch": 252, "lr": 9.838034798703683e-05}
{"train_loss": 0.3664289116859436, "global_step": 22512, "epoch": 252, "lr": 9.838020162265536e-05}
{"train_loss": 0.5975490808486938, "global_step": 22513, "epoch": 252, "lr": 9.838005525176974e-05}
{"train_loss": 0.47308221459388733, "global_step": 22514, "epoch": 252, "lr": 9.837990887438003e-05}
{"train_loss": 0.44179320335388184, "global_step": 22515, "epoch": 252, "lr": 9.837976249048621e-05}
{"train_loss": 0.4529834399732311, "global_step": 22516, "epoch": 252, "lr": 9.837961610008835e-05, "val_loss": 2.0185914039611816}
{"train_loss": 0.39949384331703186, "global_step": 22517, "epoch": 253, "lr": 9.837946970318642e-05}
{"train_loss": 0.6057344079017639, "global_step": 22518, "epoch": 253, "lr": 9.837932329978046e-05}
{"train_loss": 0.5234909057617188, "global_step": 22519, "epoch": 253, "lr": 9.837917688987051e-05}
{"train_loss": 0.43578028678894043, "global_step": 22520, "epoch": 253, "lr": 9.837903047345655e-05}
{"train_loss": 0.4862997233867645, "global_step": 22521, "epoch": 253, "lr": 9.837888405053862e-05}
{"train_loss": 0.4211280643939972, "global_step": 22522, "epoch": 253, "lr": 9.837873762111675e-05}
{"train_loss": 0.4422558546066284, "global_step": 22523, "epoch": 253, "lr": 9.837859118519093e-05}
{"train_loss": 0.600311815738678, "global_step": 22524, "epoch": 253, "lr": 9.837844474276121e-05}
{"train_loss": 0.6057024598121643, "global_step": 22525, "epoch": 253, "lr": 9.837829829382761e-05}
{"train_loss": 0.48475950956344604, "global_step": 22526, "epoch": 253, "lr": 9.837815183839013e-05}
{"train_loss": 0.42680126428604126, "global_step": 22527, "epoch": 253, "lr": 9.837800537644878e-05}
{"train_loss": 0.4929974377155304, "global_step": 22528, "epoch": 253, "lr": 9.837785890800361e-05}
{"train_loss": 0.5028375387191772, "global_step": 22529, "epoch": 253, "lr": 9.837771243305463e-05}
{"train_loss": 0.4520360231399536, "global_step": 22530, "epoch": 253, "lr": 9.837756595160184e-05}
{"train_loss": 0.4757023751735687, "global_step": 22531, "epoch": 253, "lr": 9.837741946364529e-05}
{"train_loss": 0.3638792932033539, "global_step": 22532, "epoch": 253, "lr": 9.837727296918497e-05}
{"train_loss": 0.5252124667167664, "global_step": 22533, "epoch": 253, "lr": 9.837712646822094e-05}
{"train_loss": 0.47502848505973816, "global_step": 22534, "epoch": 253, "lr": 9.837697996075318e-05}
{"train_loss": 0.4060886800289154, "global_step": 22535, "epoch": 253, "lr": 9.837683344678172e-05}
{"train_loss": 0.43672436475753784, "global_step": 22536, "epoch": 253, "lr": 9.837668692630659e-05}
{"train_loss": 0.4776836633682251, "global_step": 22537, "epoch": 253, "lr": 9.837654039932782e-05}
{"train_loss": 0.4234938323497772, "global_step": 22538, "epoch": 253, "lr": 9.83763938658454e-05}
{"train_loss": 0.4307485818862915, "global_step": 22539, "epoch": 253, "lr": 9.837624732585935e-05}
{"train_loss": 0.550054669380188, "global_step": 22540, "epoch": 253, "lr": 9.837610077936972e-05}
{"train_loss": 0.43973591923713684, "global_step": 22541, "epoch": 253, "lr": 9.837595422637651e-05}
{"train_loss": 0.512951135635376, "global_step": 22542, "epoch": 253, "lr": 9.837580766687975e-05}
{"train_loss": 0.478723406791687, "global_step": 22543, "epoch": 253, "lr": 9.837566110087944e-05}
{"train_loss": 0.46003708243370056, "global_step": 22544, "epoch": 253, "lr": 9.837551452837561e-05}
{"train_loss": 0.3500358462333679, "global_step": 22545, "epoch": 253, "lr": 9.837536794936829e-05}
{"train_loss": 0.4663212299346924, "global_step": 22546, "epoch": 253, "lr": 9.837522136385748e-05}
{"train_loss": 0.49053460359573364, "global_step": 22547, "epoch": 253, "lr": 9.837507477184322e-05}
{"train_loss": 0.3823077082633972, "global_step": 22548, "epoch": 253, "lr": 9.837492817332553e-05}
{"train_loss": 0.4710133373737335, "global_step": 22549, "epoch": 253, "lr": 9.837478156830442e-05}
{"train_loss": 0.4061983823776245, "global_step": 22550, "epoch": 253, "lr": 9.83746349567799e-05}
{"train_loss": 0.42505118250846863, "global_step": 22551, "epoch": 253, "lr": 9.837448833875201e-05}
{"train_loss": 0.3942890763282776, "global_step": 22552, "epoch": 253, "lr": 9.837434171422073e-05}
{"train_loss": 0.43856775760650635, "global_step": 22553, "epoch": 253, "lr": 9.837419508318614e-05}
{"train_loss": 0.6293720602989197, "global_step": 22554, "epoch": 253, "lr": 9.837404844564821e-05}
{"train_loss": 0.5061028003692627, "global_step": 22555, "epoch": 253, "lr": 9.8373901801607e-05}
{"train_loss": 0.42525714635849, "global_step": 22556, "epoch": 253, "lr": 9.837375515106249e-05}
{"train_loss": 0.5823367238044739, "global_step": 22557, "epoch": 253, "lr": 9.837360849401473e-05}
{"train_loss": 0.45230215787887573, "global_step": 22558, "epoch": 253, "lr": 9.837346183046373e-05}
{"train_loss": 0.600446343421936, "global_step": 22559, "epoch": 253, "lr": 9.837331516040949e-05}
{"train_loss": 0.4350847601890564, "global_step": 22560, "epoch": 253, "lr": 9.837316848385205e-05}
{"train_loss": 0.41817817091941833, "global_step": 22561, "epoch": 253, "lr": 9.837302180079144e-05}
{"train_loss": 0.42042380571365356, "global_step": 22562, "epoch": 253, "lr": 9.837287511122766e-05}
{"train_loss": 0.3779642581939697, "global_step": 22563, "epoch": 253, "lr": 9.837272841516075e-05}
{"train_loss": 0.5434631109237671, "global_step": 22564, "epoch": 253, "lr": 9.83725817125907e-05}
{"train_loss": 0.400646835565567, "global_step": 22565, "epoch": 253, "lr": 9.837243500351755e-05}
{"train_loss": 0.4796631336212158, "global_step": 22566, "epoch": 253, "lr": 9.837228828794132e-05}
{"train_loss": 0.4761536419391632, "global_step": 22567, "epoch": 253, "lr": 9.837214156586202e-05}
{"train_loss": 0.5082855224609375, "global_step": 22568, "epoch": 253, "lr": 9.837199483727968e-05}
{"train_loss": 0.5952504277229309, "global_step": 22569, "epoch": 253, "lr": 9.83718481021943e-05}
{"train_loss": 0.5374388694763184, "global_step": 22570, "epoch": 253, "lr": 9.837170136060594e-05}
{"train_loss": 0.4796864986419678, "global_step": 22571, "epoch": 253, "lr": 9.837155461251458e-05}
{"train_loss": 0.4427851140499115, "global_step": 22572, "epoch": 253, "lr": 9.837140785792025e-05}
{"train_loss": 0.47988009452819824, "global_step": 22573, "epoch": 253, "lr": 9.837126109682298e-05}
{"train_loss": 0.556220531463623, "global_step": 22574, "epoch": 253, "lr": 9.837111432922278e-05}
{"train_loss": 0.4713146686553955, "global_step": 22575, "epoch": 253, "lr": 9.837096755511968e-05}
{"train_loss": 0.44463106989860535, "global_step": 22576, "epoch": 253, "lr": 9.837082077451368e-05}
{"train_loss": 0.5591455698013306, "global_step": 22577, "epoch": 253, "lr": 9.837067398740482e-05}
{"train_loss": 0.48979413509368896, "global_step": 22578, "epoch": 253, "lr": 9.837052719379312e-05}
{"train_loss": 0.3909973204135895, "global_step": 22579, "epoch": 253, "lr": 9.837038039367859e-05}
{"train_loss": 0.39017024636268616, "global_step": 22580, "epoch": 253, "lr": 9.837023358706123e-05}
{"train_loss": 0.5996326208114624, "global_step": 22581, "epoch": 253, "lr": 9.837008677394111e-05}
{"train_loss": 0.5133088827133179, "global_step": 22582, "epoch": 253, "lr": 9.83699399543182e-05}
{"train_loss": 0.4169124364852905, "global_step": 22583, "epoch": 253, "lr": 9.836979312819256e-05}
{"train_loss": 0.4050971567630768, "global_step": 22584, "epoch": 253, "lr": 9.836964629556418e-05}
{"train_loss": 0.5567288994789124, "global_step": 22585, "epoch": 253, "lr": 9.83694994564331e-05}
{"train_loss": 0.6081306338310242, "global_step": 22586, "epoch": 253, "lr": 9.836935261079932e-05}
{"train_loss": 0.3548305630683899, "global_step": 22587, "epoch": 253, "lr": 9.836920575866287e-05}
{"train_loss": 0.5120673775672913, "global_step": 22588, "epoch": 253, "lr": 9.836905890002377e-05}
{"train_loss": 0.5947065949440002, "global_step": 22589, "epoch": 253, "lr": 9.836891203488206e-05}
{"train_loss": 0.46117648482322693, "global_step": 22590, "epoch": 253, "lr": 9.83687651632377e-05}
{"train_loss": 0.41963109374046326, "global_step": 22591, "epoch": 253, "lr": 9.836861828509079e-05}
{"train_loss": 0.45285525918006897, "global_step": 22592, "epoch": 253, "lr": 9.836847140044127e-05}
{"train_loss": 0.5316454172134399, "global_step": 22593, "epoch": 253, "lr": 9.836832450928924e-05}
{"train_loss": 0.46849748492240906, "global_step": 22594, "epoch": 253, "lr": 9.836817761163465e-05}
{"train_loss": 0.43830373883247375, "global_step": 22595, "epoch": 253, "lr": 9.836803070747755e-05}
{"train_loss": 0.4627682864665985, "global_step": 22596, "epoch": 253, "lr": 9.836788379681796e-05}
{"train_loss": 0.5051929950714111, "global_step": 22597, "epoch": 253, "lr": 9.83677368796559e-05}
{"train_loss": 0.44966769218444824, "global_step": 22598, "epoch": 253, "lr": 9.83675899559914e-05}
{"train_loss": 0.5423222780227661, "global_step": 22599, "epoch": 253, "lr": 9.836744302582444e-05}
{"train_loss": 0.39536887407302856, "global_step": 22600, "epoch": 253, "lr": 9.83672960891551e-05}
{"train_loss": 0.34451013803482056, "global_step": 22601, "epoch": 253, "lr": 9.836714914598333e-05}
{"train_loss": 0.518056333065033, "global_step": 22602, "epoch": 253, "lr": 9.83670021963092e-05}
{"train_loss": 0.4146430790424347, "global_step": 22603, "epoch": 253, "lr": 9.836685524013274e-05}
{"train_loss": 0.4399651288986206, "global_step": 22604, "epoch": 253, "lr": 9.836670827745391e-05}
{"train_loss": 0.47414381584424653, "global_step": 22605, "epoch": 253, "lr": 9.836656130827278e-05, "val_loss": 1.8902418613433838}
{"train_loss": 0.37313181161880493, "global_step": 22606, "epoch": 254, "lr": 9.836641433258937e-05}
{"train_loss": 0.5673508644104004, "global_step": 22607, "epoch": 254, "lr": 9.836626735040366e-05}
{"train_loss": 0.42083898186683655, "global_step": 22608, "epoch": 254, "lr": 9.836612036171571e-05}
{"train_loss": 0.38831886649131775, "global_step": 22609, "epoch": 254, "lr": 9.836597336652553e-05}
{"train_loss": 0.49549898505210876, "global_step": 22610, "epoch": 254, "lr": 9.836582636483313e-05}
{"train_loss": 0.538658857345581, "global_step": 22611, "epoch": 254, "lr": 9.836567935663851e-05}
{"train_loss": 0.5345737934112549, "global_step": 22612, "epoch": 254, "lr": 9.836553234194175e-05}
{"train_loss": 0.39800721406936646, "global_step": 22613, "epoch": 254, "lr": 9.836538532074283e-05}
{"train_loss": 0.39247193932533264, "global_step": 22614, "epoch": 254, "lr": 9.836523829304174e-05}
{"train_loss": 0.4463416337966919, "global_step": 22615, "epoch": 254, "lr": 9.836509125883856e-05}
{"train_loss": 0.30979329347610474, "global_step": 22616, "epoch": 254, "lr": 9.836494421813327e-05}
{"train_loss": 0.4152844548225403, "global_step": 22617, "epoch": 254, "lr": 9.836479717092592e-05}
{"train_loss": 0.4571209251880646, "global_step": 22618, "epoch": 254, "lr": 9.83646501172165e-05}
{"train_loss": 0.471513569355011, "global_step": 22619, "epoch": 254, "lr": 9.836450305700504e-05}
{"train_loss": 0.5244361758232117, "global_step": 22620, "epoch": 254, "lr": 9.836435599029156e-05}
{"train_loss": 0.41337403655052185, "global_step": 22621, "epoch": 254, "lr": 9.83642089170761e-05}
{"train_loss": 0.3995608389377594, "global_step": 22622, "epoch": 254, "lr": 9.836406183735865e-05}
{"train_loss": 0.46083253622055054, "global_step": 22623, "epoch": 254, "lr": 9.836391475113924e-05}
{"train_loss": 0.4184612035751343, "global_step": 22624, "epoch": 254, "lr": 9.836376765841789e-05}
{"train_loss": 0.46745574474334717, "global_step": 22625, "epoch": 254, "lr": 9.836362055919464e-05}
{"train_loss": 0.4699617922306061, "global_step": 22626, "epoch": 254, "lr": 9.836347345346946e-05}
{"train_loss": 0.45346003770828247, "global_step": 22627, "epoch": 254, "lr": 9.836332634124241e-05}
{"train_loss": 0.4731540381908417, "global_step": 22628, "epoch": 254, "lr": 9.836317922251352e-05}
{"train_loss": 0.37830308079719543, "global_step": 22629, "epoch": 254, "lr": 9.836303209728276e-05}
{"train_loss": 0.3486357629299164, "global_step": 22630, "epoch": 254, "lr": 9.83628849655502e-05}
{"train_loss": 0.38403958082199097, "global_step": 22631, "epoch": 254, "lr": 9.836273782731584e-05}
{"train_loss": 0.55202716588974, "global_step": 22632, "epoch": 254, "lr": 9.836259068257969e-05}
{"train_loss": 0.2761392295360565, "global_step": 22633, "epoch": 254, "lr": 9.836244353134179e-05}
{"train_loss": 0.3854098916053772, "global_step": 22634, "epoch": 254, "lr": 9.836229637360214e-05}
{"train_loss": 0.452903687953949, "global_step": 22635, "epoch": 254, "lr": 9.836214920936076e-05}
{"train_loss": 0.37526029348373413, "global_step": 22636, "epoch": 254, "lr": 9.836200203861769e-05}
{"train_loss": 0.43878173828125, "global_step": 22637, "epoch": 254, "lr": 9.836185486137295e-05}
{"train_loss": 0.39608269929885864, "global_step": 22638, "epoch": 254, "lr": 9.836170767762653e-05}
{"train_loss": 0.4703840911388397, "global_step": 22639, "epoch": 254, "lr": 9.836156048737848e-05}
{"train_loss": 0.48946690559387207, "global_step": 22640, "epoch": 254, "lr": 9.83614132906288e-05}
{"train_loss": 0.3839362561702728, "global_step": 22641, "epoch": 254, "lr": 9.836126608737752e-05}
{"train_loss": 0.5428640246391296, "global_step": 22642, "epoch": 254, "lr": 9.836111887762465e-05}
{"train_loss": 0.3805857002735138, "global_step": 22643, "epoch": 254, "lr": 9.836097166137022e-05}
{"train_loss": 0.3739098608493805, "global_step": 22644, "epoch": 254, "lr": 9.836082443861425e-05}
{"train_loss": 0.5211188197135925, "global_step": 22645, "epoch": 254, "lr": 9.836067720935676e-05}
{"train_loss": 0.5150198936462402, "global_step": 22646, "epoch": 254, "lr": 9.836052997359776e-05}
{"train_loss": 0.3803180456161499, "global_step": 22647, "epoch": 254, "lr": 9.836038273133728e-05}
{"train_loss": 0.5513964295387268, "global_step": 22648, "epoch": 254, "lr": 9.836023548257533e-05}
{"train_loss": 0.38008230924606323, "global_step": 22649, "epoch": 254, "lr": 9.836008822731194e-05}
{"train_loss": 0.32934388518333435, "global_step": 22650, "epoch": 254, "lr": 9.835994096554712e-05}
{"train_loss": 0.3684414029121399, "global_step": 22651, "epoch": 254, "lr": 9.83597936972809e-05}
{"train_loss": 0.40236085653305054, "global_step": 22652, "epoch": 254, "lr": 9.83596464225133e-05}
{"train_loss": 0.4853123426437378, "global_step": 22653, "epoch": 254, "lr": 9.835949914124434e-05}
{"train_loss": 0.5121527910232544, "global_step": 22654, "epoch": 254, "lr": 9.835935185347402e-05}
{"train_loss": 0.45430177450180054, "global_step": 22655, "epoch": 254, "lr": 9.835920455920238e-05}
{"train_loss": 0.4376745820045471, "global_step": 22656, "epoch": 254, "lr": 9.835905725842944e-05}
{"train_loss": 0.3431406319141388, "global_step": 22657, "epoch": 254, "lr": 9.835890995115519e-05}
{"train_loss": 0.33606699109077454, "global_step": 22658, "epoch": 254, "lr": 9.83587626373797e-05}
{"train_loss": 0.46040910482406616, "global_step": 22659, "epoch": 254, "lr": 9.835861531710297e-05}
{"train_loss": 0.5043888688087463, "global_step": 22660, "epoch": 254, "lr": 9.8358467990325e-05}
{"train_loss": 0.5394449830055237, "global_step": 22661, "epoch": 254, "lr": 9.835832065704582e-05}
{"train_loss": 0.45959004759788513, "global_step": 22662, "epoch": 254, "lr": 9.835817331726546e-05}
{"train_loss": 0.40309348702430725, "global_step": 22663, "epoch": 254, "lr": 9.835802597098393e-05}
{"train_loss": 0.4453568160533905, "global_step": 22664, "epoch": 254, "lr": 9.835787861820125e-05}
{"train_loss": 0.4663272798061371, "global_step": 22665, "epoch": 254, "lr": 9.835773125891745e-05}
{"train_loss": 0.41376277804374695, "global_step": 22666, "epoch": 254, "lr": 9.835758389313253e-05}
{"train_loss": 0.47287920117378235, "global_step": 22667, "epoch": 254, "lr": 9.835743652084654e-05}
{"train_loss": 0.46172598004341125, "global_step": 22668, "epoch": 254, "lr": 9.835728914205946e-05}
{"train_loss": 0.3681102395057678, "global_step": 22669, "epoch": 254, "lr": 9.835714175677135e-05}
{"train_loss": 0.4466075897216797, "global_step": 22670, "epoch": 254, "lr": 9.83569943649822e-05}
{"train_loss": 0.441272497177124, "global_step": 22671, "epoch": 254, "lr": 9.835684696669206e-05}
{"train_loss": 0.46774598956108093, "global_step": 22672, "epoch": 254, "lr": 9.835669956190091e-05}
{"train_loss": 0.5163321495056152, "global_step": 22673, "epoch": 254, "lr": 9.835655215060881e-05}
{"train_loss": 0.46611297130584717, "global_step": 22674, "epoch": 254, "lr": 9.835640473281575e-05}
{"train_loss": 0.44805827736854553, "global_step": 22675, "epoch": 254, "lr": 9.835625730852176e-05}
{"train_loss": 0.4313078224658966, "global_step": 22676, "epoch": 254, "lr": 9.835610987772685e-05}
{"train_loss": 0.474778950214386, "global_step": 22677, "epoch": 254, "lr": 9.835596244043107e-05}
{"train_loss": 0.4848814010620117, "global_step": 22678, "epoch": 254, "lr": 9.83558149966344e-05}
{"train_loss": 0.411834716796875, "global_step": 22679, "epoch": 254, "lr": 9.835566754633691e-05}
{"train_loss": 0.4694980978965759, "global_step": 22680, "epoch": 254, "lr": 9.835552008953856e-05}
{"train_loss": 0.538098931312561, "global_step": 22681, "epoch": 254, "lr": 9.835537262623941e-05}
{"train_loss": 0.5079185962677002, "global_step": 22682, "epoch": 254, "lr": 9.835522515643948e-05}
{"train_loss": 0.3706122934818268, "global_step": 22683, "epoch": 254, "lr": 9.835507768013877e-05}
{"train_loss": 0.5710031986236572, "global_step": 22684, "epoch": 254, "lr": 9.835493019733732e-05}
{"train_loss": 0.39971426129341125, "global_step": 22685, "epoch": 254, "lr": 9.835478270803512e-05}
{"train_loss": 0.5116368532180786, "global_step": 22686, "epoch": 254, "lr": 9.835463521223222e-05}
{"train_loss": 0.32927682995796204, "global_step": 22687, "epoch": 254, "lr": 9.835448770992864e-05}
{"train_loss": 0.4283770024776459, "global_step": 22688, "epoch": 254, "lr": 9.835434020112437e-05}
{"train_loss": 0.3625156879425049, "global_step": 22689, "epoch": 254, "lr": 9.835419268581945e-05}
{"train_loss": 0.3145856261253357, "global_step": 22690, "epoch": 254, "lr": 9.83540451640139e-05}
{"train_loss": 0.40442216396331787, "global_step": 22691, "epoch": 254, "lr": 9.835389763570773e-05}
{"train_loss": 0.3891471028327942, "global_step": 22692, "epoch": 254, "lr": 9.835375010090097e-05}
{"train_loss": 0.5020782947540283, "global_step": 22693, "epoch": 254, "lr": 9.835360255959366e-05}
{"train_loss": 0.43955863627155173, "global_step": 22694, "epoch": 254, "lr": 9.835345501178579e-05, "val_loss": 1.8901937007904053}
{"train_loss": 0.5364148020744324, "global_step": 22695, "epoch": 255, "lr": 9.835330745747738e-05}
{"train_loss": 0.46908023953437805, "global_step": 22696, "epoch": 255, "lr": 9.835315989666844e-05}
{"train_loss": 0.4166899025440216, "global_step": 22697, "epoch": 255, "lr": 9.835301232935904e-05}
{"train_loss": 0.2929726243019104, "global_step": 22698, "epoch": 255, "lr": 9.835286475554915e-05}
{"train_loss": 0.454393208026886, "global_step": 22699, "epoch": 255, "lr": 9.83527171752388e-05}
{"train_loss": 0.5950760841369629, "global_step": 22700, "epoch": 255, "lr": 9.835256958842803e-05}
{"train_loss": 0.43655335903167725, "global_step": 22701, "epoch": 255, "lr": 9.835242199511684e-05}
{"train_loss": 0.5571898221969604, "global_step": 22702, "epoch": 255, "lr": 9.835227439530526e-05}
{"train_loss": 0.4959348142147064, "global_step": 22703, "epoch": 255, "lr": 9.835212678899331e-05}
{"train_loss": 0.49870145320892334, "global_step": 22704, "epoch": 255, "lr": 9.835197917618098e-05}
{"train_loss": 0.5371310710906982, "global_step": 22705, "epoch": 255, "lr": 9.835183155686833e-05}
{"train_loss": 0.6017740368843079, "global_step": 22706, "epoch": 255, "lr": 9.835168393105538e-05}
{"train_loss": 0.6023643612861633, "global_step": 22707, "epoch": 255, "lr": 9.835153629874212e-05}
{"train_loss": 0.3736252784729004, "global_step": 22708, "epoch": 255, "lr": 9.835138865992859e-05}
{"train_loss": 0.5191742777824402, "global_step": 22709, "epoch": 255, "lr": 9.835124101461481e-05}
{"train_loss": 0.4632815718650818, "global_step": 22710, "epoch": 255, "lr": 9.835109336280079e-05}
{"train_loss": 0.5634900331497192, "global_step": 22711, "epoch": 255, "lr": 9.835094570448656e-05}
{"train_loss": 0.435474157333374, "global_step": 22712, "epoch": 255, "lr": 9.835079803967214e-05}
{"train_loss": 0.6118661165237427, "global_step": 22713, "epoch": 255, "lr": 9.835065036835752e-05}
{"train_loss": 0.5382195115089417, "global_step": 22714, "epoch": 255, "lr": 9.835050269054276e-05}
{"train_loss": 0.41221141815185547, "global_step": 22715, "epoch": 255, "lr": 9.835035500622787e-05}
{"train_loss": 0.45458224415779114, "global_step": 22716, "epoch": 255, "lr": 9.835020731541285e-05}
{"train_loss": 0.419432133436203, "global_step": 22717, "epoch": 255, "lr": 9.835005961809775e-05}
{"train_loss": 0.4486524164676666, "global_step": 22718, "epoch": 255, "lr": 9.834991191428256e-05}
{"train_loss": 0.3666937053203583, "global_step": 22719, "epoch": 255, "lr": 9.834976420396731e-05}
{"train_loss": 0.4262489080429077, "global_step": 22720, "epoch": 255, "lr": 9.834961648715203e-05}
{"train_loss": 0.43469104170799255, "global_step": 22721, "epoch": 255, "lr": 9.834946876383675e-05}
{"train_loss": 0.4894685447216034, "global_step": 22722, "epoch": 255, "lr": 9.834932103402146e-05}
{"train_loss": 0.4747805595397949, "global_step": 22723, "epoch": 255, "lr": 9.834917329770619e-05}
{"train_loss": 0.40056538581848145, "global_step": 22724, "epoch": 255, "lr": 9.834902555489096e-05}
{"train_loss": 0.37107381224632263, "global_step": 22725, "epoch": 255, "lr": 9.83488778055758e-05}
{"train_loss": 0.36099711060523987, "global_step": 22726, "epoch": 255, "lr": 9.834873004976072e-05}
{"train_loss": 0.4187975823879242, "global_step": 22727, "epoch": 255, "lr": 9.834858228744575e-05}
{"train_loss": 0.4390357434749603, "global_step": 22728, "epoch": 255, "lr": 9.834843451863089e-05}
{"train_loss": 0.47764572501182556, "global_step": 22729, "epoch": 255, "lr": 9.834828674331617e-05}
{"train_loss": 0.5384153723716736, "global_step": 22730, "epoch": 255, "lr": 9.834813896150162e-05}
{"train_loss": 0.4061836004257202, "global_step": 22731, "epoch": 255, "lr": 9.834799117318725e-05}
{"train_loss": 0.4318884313106537, "global_step": 22732, "epoch": 255, "lr": 9.834784337837308e-05}
{"train_loss": 0.4597277045249939, "global_step": 22733, "epoch": 255, "lr": 9.834769557705914e-05}
{"train_loss": 0.3774701654911041, "global_step": 22734, "epoch": 255, "lr": 9.834754776924543e-05}
{"train_loss": 0.3081611394882202, "global_step": 22735, "epoch": 255, "lr": 9.8347399954932e-05}
{"train_loss": 0.41117385029792786, "global_step": 22736, "epoch": 255, "lr": 9.834725213411884e-05}
{"train_loss": 0.5330711603164673, "global_step": 22737, "epoch": 255, "lr": 9.834710430680598e-05}
{"train_loss": 0.49159523844718933, "global_step": 22738, "epoch": 255, "lr": 9.834695647299343e-05}
{"train_loss": 0.3940800428390503, "global_step": 22739, "epoch": 255, "lr": 9.834680863268124e-05}
{"train_loss": 0.4592837691307068, "global_step": 22740, "epoch": 255, "lr": 9.83466607858694e-05}
{"train_loss": 0.4053034484386444, "global_step": 22741, "epoch": 255, "lr": 9.834651293255796e-05}
{"train_loss": 0.4874679744243622, "global_step": 22742, "epoch": 255, "lr": 9.83463650727469e-05}
{"train_loss": 0.3891991674900055, "global_step": 22743, "epoch": 255, "lr": 9.834621720643626e-05}
{"train_loss": 0.5466480255126953, "global_step": 22744, "epoch": 255, "lr": 9.834606933362606e-05}
{"train_loss": 0.4248417019844055, "global_step": 22745, "epoch": 255, "lr": 9.834592145431634e-05}
{"train_loss": 0.4463990032672882, "global_step": 22746, "epoch": 255, "lr": 9.834577356850708e-05}
{"train_loss": 0.48932889103889465, "global_step": 22747, "epoch": 255, "lr": 9.834562567619833e-05}
{"train_loss": 0.5648863911628723, "global_step": 22748, "epoch": 255, "lr": 9.834547777739009e-05}
{"train_loss": 0.4445318877696991, "global_step": 22749, "epoch": 255, "lr": 9.83453298720824e-05}
{"train_loss": 0.35169774293899536, "global_step": 22750, "epoch": 255, "lr": 9.834518196027527e-05}
{"train_loss": 0.3324066698551178, "global_step": 22751, "epoch": 255, "lr": 9.834503404196871e-05}
{"train_loss": 0.5186277031898499, "global_step": 22752, "epoch": 255, "lr": 9.834488611716275e-05}
{"train_loss": 0.46825718879699707, "global_step": 22753, "epoch": 255, "lr": 9.834473818585742e-05}
{"train_loss": 0.41086503863334656, "global_step": 22754, "epoch": 255, "lr": 9.834459024805271e-05}
{"train_loss": 0.4653848707675934, "global_step": 22755, "epoch": 255, "lr": 9.834444230374869e-05}
{"train_loss": 0.3996773362159729, "global_step": 22756, "epoch": 255, "lr": 9.834429435294532e-05}
{"train_loss": 0.5170989036560059, "global_step": 22757, "epoch": 255, "lr": 9.834414639564267e-05}
{"train_loss": 0.49635055661201477, "global_step": 22758, "epoch": 255, "lr": 9.834399843184072e-05}
{"train_loss": 0.7252362966537476, "global_step": 22759, "epoch": 255, "lr": 9.834385046153952e-05}
{"train_loss": 0.5044853687286377, "global_step": 22760, "epoch": 255, "lr": 9.834370248473907e-05}
{"train_loss": 0.40059539675712585, "global_step": 22761, "epoch": 255, "lr": 9.834355450143941e-05}
{"train_loss": 0.47252845764160156, "global_step": 22762, "epoch": 255, "lr": 9.834340651164054e-05}
{"train_loss": 0.467021107673645, "global_step": 22763, "epoch": 255, "lr": 9.834325851534249e-05}
{"train_loss": 0.40557631850242615, "global_step": 22764, "epoch": 255, "lr": 9.834311051254527e-05}
{"train_loss": 0.4550887644290924, "global_step": 22765, "epoch": 255, "lr": 9.834296250324892e-05}
{"train_loss": 0.5608912110328674, "global_step": 22766, "epoch": 255, "lr": 9.834281448745346e-05}
{"train_loss": 0.39025774598121643, "global_step": 22767, "epoch": 255, "lr": 9.834266646515889e-05}
{"train_loss": 0.42172345519065857, "global_step": 22768, "epoch": 255, "lr": 9.834251843636521e-05}
{"train_loss": 0.3839606046676636, "global_step": 22769, "epoch": 255, "lr": 9.83423704010725e-05}
{"train_loss": 0.4121851325035095, "global_step": 22770, "epoch": 255, "lr": 9.834222235928073e-05}
{"train_loss": 0.45829010009765625, "global_step": 22771, "epoch": 255, "lr": 9.834207431098994e-05}
{"train_loss": 0.4897555112838745, "global_step": 22772, "epoch": 255, "lr": 9.834192625620016e-05}
{"train_loss": 0.4479966163635254, "global_step": 22773, "epoch": 255, "lr": 9.834177819491138e-05}
{"train_loss": 0.5475872159004211, "global_step": 22774, "epoch": 255, "lr": 9.834163012712366e-05}
{"train_loss": 0.511927604675293, "global_step": 22775, "epoch": 255, "lr": 9.834148205283698e-05}
{"train_loss": 0.3745833933353424, "global_step": 22776, "epoch": 255, "lr": 9.834133397205136e-05}
{"train_loss": 0.5695490837097168, "global_step": 22777, "epoch": 255, "lr": 9.834118588476686e-05}
{"train_loss": 0.4861888885498047, "global_step": 22778, "epoch": 255, "lr": 9.834103779098348e-05}
{"train_loss": 0.517646849155426, "global_step": 22779, "epoch": 255, "lr": 9.834088969070124e-05}
{"train_loss": 0.4480064809322357, "global_step": 22780, "epoch": 255, "lr": 9.834074158392012e-05}
{"train_loss": 0.4888840615749359, "global_step": 22781, "epoch": 255, "lr": 9.834059347064022e-05}
{"train_loss": 0.4503742754459381, "global_step": 22782, "epoch": 255, "lr": 9.834044535086149e-05}
{"train_loss": 0.4623449890131361, "global_step": 22783, "epoch": 255, "lr": 9.834029722458397e-05, "val_loss": 1.8781658411026, "train_action_mse_error": 19.638505935668945}
{"train_loss": 0.5224568843841553, "global_step": 22784, "epoch": 256, "lr": 9.83401490918077e-05}
{"train_loss": 0.5663625001907349, "global_step": 22785, "epoch": 256, "lr": 9.83400009525327e-05}
{"train_loss": 0.5828092098236084, "global_step": 22786, "epoch": 256, "lr": 9.833985280675898e-05}
{"train_loss": 0.39226073026657104, "global_step": 22787, "epoch": 256, "lr": 9.833970465448652e-05}
{"train_loss": 0.4154091477394104, "global_step": 22788, "epoch": 256, "lr": 9.83395564957154e-05}
{"train_loss": 0.42061182856559753, "global_step": 22789, "epoch": 256, "lr": 9.833940833044561e-05}
{"train_loss": 0.48962610960006714, "global_step": 22790, "epoch": 256, "lr": 9.833926015867719e-05}
{"train_loss": 0.33817559480667114, "global_step": 22791, "epoch": 256, "lr": 9.833911198041013e-05}
{"train_loss": 0.45197534561157227, "global_step": 22792, "epoch": 256, "lr": 9.833896379564445e-05}
{"train_loss": 0.5479084849357605, "global_step": 22793, "epoch": 256, "lr": 9.833881560438021e-05}
{"train_loss": 0.48015379905700684, "global_step": 22794, "epoch": 256, "lr": 9.833866740661739e-05}
{"train_loss": 0.45417019724845886, "global_step": 22795, "epoch": 256, "lr": 9.833851920235604e-05}
{"train_loss": 0.31135278940200806, "global_step": 22796, "epoch": 256, "lr": 9.833837099159616e-05}
{"train_loss": 0.38343021273612976, "global_step": 22797, "epoch": 256, "lr": 9.833822277433777e-05}
{"train_loss": 0.4735986292362213, "global_step": 22798, "epoch": 256, "lr": 9.833807455058091e-05}
{"train_loss": 0.3476327359676361, "global_step": 22799, "epoch": 256, "lr": 9.833792632032557e-05}
{"train_loss": 0.49932077527046204, "global_step": 22800, "epoch": 256, "lr": 9.833777808357178e-05}
{"train_loss": 0.46113213896751404, "global_step": 22801, "epoch": 256, "lr": 9.833762984031958e-05}
{"train_loss": 0.4619915783405304, "global_step": 22802, "epoch": 256, "lr": 9.833748159056897e-05}
{"train_loss": 0.5762641429901123, "global_step": 22803, "epoch": 256, "lr": 9.833733333431998e-05}
{"train_loss": 0.4794655740261078, "global_step": 22804, "epoch": 256, "lr": 9.833718507157261e-05}
{"train_loss": 0.4980289041996002, "global_step": 22805, "epoch": 256, "lr": 9.833703680232691e-05}
{"train_loss": 0.5251208543777466, "global_step": 22806, "epoch": 256, "lr": 9.833688852658288e-05}
{"train_loss": 0.416832834482193, "global_step": 22807, "epoch": 256, "lr": 9.833674024434054e-05}
{"train_loss": 0.4985519051551819, "global_step": 22808, "epoch": 256, "lr": 9.833659195559991e-05}
{"train_loss": 0.48286232352256775, "global_step": 22809, "epoch": 256, "lr": 9.833644366036103e-05}
{"train_loss": 0.49268925189971924, "global_step": 22810, "epoch": 256, "lr": 9.833629535862389e-05}
{"train_loss": 0.5015090703964233, "global_step": 22811, "epoch": 256, "lr": 9.833614705038852e-05}
{"train_loss": 0.5480841398239136, "global_step": 22812, "epoch": 256, "lr": 9.833599873565495e-05}
{"train_loss": 0.5404282808303833, "global_step": 22813, "epoch": 256, "lr": 9.833585041442322e-05}
{"train_loss": 0.4873385429382324, "global_step": 22814, "epoch": 256, "lr": 9.83357020866933e-05}
{"train_loss": 0.48996710777282715, "global_step": 22815, "epoch": 256, "lr": 9.833555375246523e-05}
{"train_loss": 0.5164318084716797, "global_step": 22816, "epoch": 256, "lr": 9.833540541173904e-05}
{"train_loss": 0.3760809898376465, "global_step": 22817, "epoch": 256, "lr": 9.833525706451473e-05}
{"train_loss": 0.4606909155845642, "global_step": 22818, "epoch": 256, "lr": 9.833510871079236e-05}
{"train_loss": 0.39844128489494324, "global_step": 22819, "epoch": 256, "lr": 9.83349603505719e-05}
{"train_loss": 0.4314829707145691, "global_step": 22820, "epoch": 256, "lr": 9.833481198385342e-05}
{"train_loss": 0.3937831521034241, "global_step": 22821, "epoch": 256, "lr": 9.833466361063688e-05}
{"train_loss": 0.5046706795692444, "global_step": 22822, "epoch": 256, "lr": 9.833451523092236e-05}
{"train_loss": 0.5796850323677063, "global_step": 22823, "epoch": 256, "lr": 9.833436684470985e-05}
{"train_loss": 0.5127521753311157, "global_step": 22824, "epoch": 256, "lr": 9.833421845199937e-05}
{"train_loss": 0.4441043436527252, "global_step": 22825, "epoch": 256, "lr": 9.833407005279094e-05}
{"train_loss": 0.46308597922325134, "global_step": 22826, "epoch": 256, "lr": 9.833392164708459e-05}
{"train_loss": 0.4722220003604889, "global_step": 22827, "epoch": 256, "lr": 9.833377323488034e-05}
{"train_loss": 0.4953349530696869, "global_step": 22828, "epoch": 256, "lr": 9.833362481617819e-05}
{"train_loss": 0.5351858735084534, "global_step": 22829, "epoch": 256, "lr": 9.833347639097816e-05}
{"train_loss": 0.4960431754589081, "global_step": 22830, "epoch": 256, "lr": 9.833332795928029e-05}
{"train_loss": 0.56395024061203, "global_step": 22831, "epoch": 256, "lr": 9.833317952108461e-05}
{"train_loss": 0.5061433911323547, "global_step": 22832, "epoch": 256, "lr": 9.833303107639112e-05}
{"train_loss": 0.4789735674858093, "global_step": 22833, "epoch": 256, "lr": 9.833288262519984e-05}
{"train_loss": 0.5146350860595703, "global_step": 22834, "epoch": 256, "lr": 9.833273416751078e-05}
{"train_loss": 0.44155627489089966, "global_step": 22835, "epoch": 256, "lr": 9.8332585703324e-05}
{"train_loss": 0.4084246754646301, "global_step": 22836, "epoch": 256, "lr": 9.833243723263948e-05}
{"train_loss": 0.4879043698310852, "global_step": 22837, "epoch": 256, "lr": 9.833228875545724e-05}
{"train_loss": 0.41416504979133606, "global_step": 22838, "epoch": 256, "lr": 9.833214027177733e-05}
{"train_loss": 0.5674594640731812, "global_step": 22839, "epoch": 256, "lr": 9.833199178159974e-05}
{"train_loss": 0.39657071232795715, "global_step": 22840, "epoch": 256, "lr": 9.83318432849245e-05}
{"train_loss": 0.5446131229400635, "global_step": 22841, "epoch": 256, "lr": 9.833169478175165e-05}
{"train_loss": 0.6173681616783142, "global_step": 22842, "epoch": 256, "lr": 9.833154627208119e-05}
{"train_loss": 0.3907614052295685, "global_step": 22843, "epoch": 256, "lr": 9.833139775591313e-05}
{"train_loss": 0.39688077569007874, "global_step": 22844, "epoch": 256, "lr": 9.833124923324752e-05}
{"train_loss": 0.4898321330547333, "global_step": 22845, "epoch": 256, "lr": 9.833110070408434e-05}
{"train_loss": 0.44461917877197266, "global_step": 22846, "epoch": 256, "lr": 9.833095216842364e-05}
{"train_loss": 0.39192521572113037, "global_step": 22847, "epoch": 256, "lr": 9.833080362626544e-05}
{"train_loss": 0.5003514885902405, "global_step": 22848, "epoch": 256, "lr": 9.833065507760975e-05}
{"train_loss": 0.5260449647903442, "global_step": 22849, "epoch": 256, "lr": 9.83305065224566e-05}
{"train_loss": 0.4088863432407379, "global_step": 22850, "epoch": 256, "lr": 9.833035796080599e-05}
{"train_loss": 0.4087238609790802, "global_step": 22851, "epoch": 256, "lr": 9.833020939265796e-05}
{"train_loss": 0.5081437230110168, "global_step": 22852, "epoch": 256, "lr": 9.833006081801252e-05}
{"train_loss": 0.5375248193740845, "global_step": 22853, "epoch": 256, "lr": 9.832991223686969e-05}
{"train_loss": 0.4486660063266754, "global_step": 22854, "epoch": 256, "lr": 9.83297636492295e-05}
{"train_loss": 0.3694033622741699, "global_step": 22855, "epoch": 256, "lr": 9.832961505509195e-05}
{"train_loss": 0.4587010443210602, "global_step": 22856, "epoch": 256, "lr": 9.832946645445709e-05}
{"train_loss": 0.4143974781036377, "global_step": 22857, "epoch": 256, "lr": 9.83293178473249e-05}
{"train_loss": 0.5410902500152588, "global_step": 22858, "epoch": 256, "lr": 9.832916923369544e-05}
{"train_loss": 0.4346248209476471, "global_step": 22859, "epoch": 256, "lr": 9.83290206135687e-05}
{"train_loss": 0.4284818172454834, "global_step": 22860, "epoch": 256, "lr": 9.832887198694471e-05}
{"train_loss": 0.3997592031955719, "global_step": 22861, "epoch": 256, "lr": 9.83287233538235e-05}
{"train_loss": 0.501579999923706, "global_step": 22862, "epoch": 256, "lr": 9.832857471420508e-05}
{"train_loss": 0.6025936007499695, "global_step": 22863, "epoch": 256, "lr": 9.832842606808949e-05}
{"train_loss": 0.5381511449813843, "global_step": 22864, "epoch": 256, "lr": 9.832827741547672e-05}
{"train_loss": 0.3769006133079529, "global_step": 22865, "epoch": 256, "lr": 9.832812875636679e-05}
{"train_loss": 0.3976033329963684, "global_step": 22866, "epoch": 256, "lr": 9.832798009075974e-05}
{"train_loss": 0.36612045764923096, "global_step": 22867, "epoch": 256, "lr": 9.832783141865558e-05}
{"train_loss": 0.5123982429504395, "global_step": 22868, "epoch": 256, "lr": 9.832768274005435e-05}
{"train_loss": 0.35996702313423157, "global_step": 22869, "epoch": 256, "lr": 9.832753405495603e-05}
{"train_loss": 0.4178891181945801, "global_step": 22870, "epoch": 256, "lr": 9.832738536336067e-05}
{"train_loss": 0.47680234909057617, "global_step": 22871, "epoch": 256, "lr": 9.832723666526828e-05}
{"train_loss": 0.4677221289511477, "global_step": 22872, "epoch": 256, "lr": 9.832708796067888e-05, "val_loss": 1.9756797552108765}
{"train_loss": 0.3803127706050873, "global_step": 22873, "epoch": 257, "lr": 9.83269392495925e-05}
{"train_loss": 0.38728153705596924, "global_step": 22874, "epoch": 257, "lr": 9.832679053200915e-05}
{"train_loss": 0.3741748631000519, "global_step": 22875, "epoch": 257, "lr": 9.832664180792884e-05}
{"train_loss": 0.5693846344947815, "global_step": 22876, "epoch": 257, "lr": 9.832649307735162e-05}
{"train_loss": 0.5468283295631409, "global_step": 22877, "epoch": 257, "lr": 9.832634434027748e-05}
{"train_loss": 0.5392076969146729, "global_step": 22878, "epoch": 257, "lr": 9.832619559670645e-05}
{"train_loss": 0.41770139336586, "global_step": 22879, "epoch": 257, "lr": 9.832604684663857e-05}
{"train_loss": 0.5366285443305969, "global_step": 22880, "epoch": 257, "lr": 9.832589809007382e-05}
{"train_loss": 0.38068458437919617, "global_step": 22881, "epoch": 257, "lr": 9.832574932701227e-05}
{"train_loss": 0.4367401897907257, "global_step": 22882, "epoch": 257, "lr": 9.832560055745388e-05}
{"train_loss": 0.4911099374294281, "global_step": 22883, "epoch": 257, "lr": 9.832545178139871e-05}
{"train_loss": 0.5573390126228333, "global_step": 22884, "epoch": 257, "lr": 9.832530299884679e-05}
{"train_loss": 0.38230472803115845, "global_step": 22885, "epoch": 257, "lr": 9.832515420979811e-05}
{"train_loss": 0.5679091215133667, "global_step": 22886, "epoch": 257, "lr": 9.83250054142527e-05}
{"train_loss": 0.27781832218170166, "global_step": 22887, "epoch": 257, "lr": 9.832485661221058e-05}
{"train_loss": 0.5931881070137024, "global_step": 22888, "epoch": 257, "lr": 9.832470780367178e-05}
{"train_loss": 0.4462335407733917, "global_step": 22889, "epoch": 257, "lr": 9.83245589886363e-05}
{"train_loss": 0.4024784564971924, "global_step": 22890, "epoch": 257, "lr": 9.832441016710418e-05}
{"train_loss": 0.4905846118927002, "global_step": 22891, "epoch": 257, "lr": 9.832426133907544e-05}
{"train_loss": 0.39339983463287354, "global_step": 22892, "epoch": 257, "lr": 9.832411250455007e-05}
{"train_loss": 0.43987393379211426, "global_step": 22893, "epoch": 257, "lr": 9.832396366352814e-05}
{"train_loss": 0.5158146619796753, "global_step": 22894, "epoch": 257, "lr": 9.832381481600963e-05}
{"train_loss": 0.5460712909698486, "global_step": 22895, "epoch": 257, "lr": 9.832366596199456e-05}
{"train_loss": 0.5961414575576782, "global_step": 22896, "epoch": 257, "lr": 9.832351710148297e-05}
{"train_loss": 0.4313211143016815, "global_step": 22897, "epoch": 257, "lr": 9.832336823447488e-05}
{"train_loss": 0.5548425912857056, "global_step": 22898, "epoch": 257, "lr": 9.832321936097029e-05}
{"train_loss": 0.5927795767784119, "global_step": 22899, "epoch": 257, "lr": 9.832307048096925e-05}
{"train_loss": 0.441351979970932, "global_step": 22900, "epoch": 257, "lr": 9.832292159447174e-05}
{"train_loss": 0.5449903011322021, "global_step": 22901, "epoch": 257, "lr": 9.832277270147782e-05}
{"train_loss": 0.46832385659217834, "global_step": 22902, "epoch": 257, "lr": 9.83226238019875e-05}
{"train_loss": 0.47317925095558167, "global_step": 22903, "epoch": 257, "lr": 9.832247489600076e-05}
{"train_loss": 0.533439576625824, "global_step": 22904, "epoch": 257, "lr": 9.832232598351767e-05}
{"train_loss": 0.4357982277870178, "global_step": 22905, "epoch": 257, "lr": 9.832217706453824e-05}
{"train_loss": 0.5217534303665161, "global_step": 22906, "epoch": 257, "lr": 9.832202813906246e-05}
{"train_loss": 0.5270869731903076, "global_step": 22907, "epoch": 257, "lr": 9.83218792070904e-05}
{"train_loss": 0.4574216604232788, "global_step": 22908, "epoch": 257, "lr": 9.832173026862203e-05}
{"train_loss": 0.5025003552436829, "global_step": 22909, "epoch": 257, "lr": 9.83215813236574e-05}
{"train_loss": 0.38318878412246704, "global_step": 22910, "epoch": 257, "lr": 9.832143237219653e-05}
{"train_loss": 0.3973899185657501, "global_step": 22911, "epoch": 257, "lr": 9.832128341423943e-05}
{"train_loss": 0.39641737937927246, "global_step": 22912, "epoch": 257, "lr": 9.83211344497861e-05}
{"train_loss": 0.4307812452316284, "global_step": 22913, "epoch": 257, "lr": 9.832098547883662e-05}
{"train_loss": 0.3057624101638794, "global_step": 22914, "epoch": 257, "lr": 9.832083650139095e-05}
{"train_loss": 0.6398383975028992, "global_step": 22915, "epoch": 257, "lr": 9.832068751744913e-05}
{"train_loss": 0.37152940034866333, "global_step": 22916, "epoch": 257, "lr": 9.832053852701119e-05}
{"train_loss": 0.46262499690055847, "global_step": 22917, "epoch": 257, "lr": 9.832038953007713e-05}
{"train_loss": 0.3957033157348633, "global_step": 22918, "epoch": 257, "lr": 9.8320240526647e-05}
{"train_loss": 0.4473634362220764, "global_step": 22919, "epoch": 257, "lr": 9.832009151672079e-05}
{"train_loss": 0.5135518908500671, "global_step": 22920, "epoch": 257, "lr": 9.831994250029854e-05}
{"train_loss": 0.3745400607585907, "global_step": 22921, "epoch": 257, "lr": 9.831979347738027e-05}
{"train_loss": 0.4078678786754608, "global_step": 22922, "epoch": 257, "lr": 9.831964444796597e-05}
{"train_loss": 0.5108158588409424, "global_step": 22923, "epoch": 257, "lr": 9.83194954120557e-05}
{"train_loss": 0.5252242684364319, "global_step": 22924, "epoch": 257, "lr": 9.831934636964945e-05}
{"train_loss": 0.44586655497550964, "global_step": 22925, "epoch": 257, "lr": 9.831919732074728e-05}
{"train_loss": 0.5862188935279846, "global_step": 22926, "epoch": 257, "lr": 9.831904826534916e-05}
{"train_loss": 0.42345309257507324, "global_step": 22927, "epoch": 257, "lr": 9.831889920345513e-05}
{"train_loss": 0.4521642327308655, "global_step": 22928, "epoch": 257, "lr": 9.831875013506523e-05}
{"train_loss": 0.4568229913711548, "global_step": 22929, "epoch": 257, "lr": 9.831860106017945e-05}
{"train_loss": 0.499837189912796, "global_step": 22930, "epoch": 257, "lr": 9.831845197879783e-05}
{"train_loss": 0.4647301435470581, "global_step": 22931, "epoch": 257, "lr": 9.831830289092038e-05}
{"train_loss": 0.4727417826652527, "global_step": 22932, "epoch": 257, "lr": 9.831815379654712e-05}
{"train_loss": 0.32719123363494873, "global_step": 22933, "epoch": 257, "lr": 9.831800469567808e-05}
{"train_loss": 0.4509010314941406, "global_step": 22934, "epoch": 257, "lr": 9.831785558831327e-05}
{"train_loss": 0.348381370306015, "global_step": 22935, "epoch": 257, "lr": 9.831770647445271e-05}
{"train_loss": 0.4102070927619934, "global_step": 22936, "epoch": 257, "lr": 9.831755735409643e-05}
{"train_loss": 0.4727643132209778, "global_step": 22937, "epoch": 257, "lr": 9.831740822724444e-05}
{"train_loss": 0.4409363269805908, "global_step": 22938, "epoch": 257, "lr": 9.831725909389676e-05}
{"train_loss": 0.4463011622428894, "global_step": 22939, "epoch": 257, "lr": 9.83171099540534e-05}
{"train_loss": 0.47034963965415955, "global_step": 22940, "epoch": 257, "lr": 9.831696080771442e-05}
{"train_loss": 0.46910956501960754, "global_step": 22941, "epoch": 257, "lr": 9.831681165487981e-05}
{"train_loss": 0.4469868540763855, "global_step": 22942, "epoch": 257, "lr": 9.831666249554959e-05}
{"train_loss": 0.47371071577072144, "global_step": 22943, "epoch": 257, "lr": 9.831651332972377e-05}
{"train_loss": 0.43687885999679565, "global_step": 22944, "epoch": 257, "lr": 9.831636415740239e-05}
{"train_loss": 0.42179083824157715, "global_step": 22945, "epoch": 257, "lr": 9.831621497858547e-05}
{"train_loss": 0.39601555466651917, "global_step": 22946, "epoch": 257, "lr": 9.831606579327301e-05}
{"train_loss": 0.3551788330078125, "global_step": 22947, "epoch": 257, "lr": 9.831591660146507e-05}
{"train_loss": 0.4599345326423645, "global_step": 22948, "epoch": 257, "lr": 9.831576740316162e-05}
{"train_loss": 0.5065984725952148, "global_step": 22949, "epoch": 257, "lr": 9.831561819836272e-05}
{"train_loss": 0.45417866110801697, "global_step": 22950, "epoch": 257, "lr": 9.831546898706836e-05}
{"train_loss": 0.5105587244033813, "global_step": 22951, "epoch": 257, "lr": 9.831531976927858e-05}
{"train_loss": 0.4496954381465912, "global_step": 22952, "epoch": 257, "lr": 9.83151705449934e-05}
{"train_loss": 0.5841878056526184, "global_step": 22953, "epoch": 257, "lr": 9.831502131421283e-05}
{"train_loss": 0.31223064661026, "global_step": 22954, "epoch": 257, "lr": 9.831487207693689e-05}
{"train_loss": 0.3775812089443207, "global_step": 22955, "epoch": 257, "lr": 9.83147228331656e-05}
{"train_loss": 0.6726704835891724, "global_step": 22956, "epoch": 257, "lr": 9.8314573582899e-05}
{"train_loss": 0.48073506355285645, "global_step": 22957, "epoch": 257, "lr": 9.831442432613709e-05}
{"train_loss": 0.43550872802734375, "global_step": 22958, "epoch": 257, "lr": 9.831427506287988e-05}
{"train_loss": 0.5179161429405212, "global_step": 22959, "epoch": 257, "lr": 9.831412579312741e-05}
{"train_loss": 0.5469541549682617, "global_step": 22960, "epoch": 257, "lr": 9.831397651687972e-05}
{"train_loss": 0.46350715535410336, "global_step": 22961, "epoch": 257, "lr": 9.831382723413677e-05, "val_loss": 1.8868721723556519}
{"train_loss": 0.36509352922439575, "global_step": 22962, "epoch": 258, "lr": 9.831367794489864e-05}
{"train_loss": 0.38786017894744873, "global_step": 22963, "epoch": 258, "lr": 9.831352864916532e-05}
{"train_loss": 0.4631223976612091, "global_step": 22964, "epoch": 258, "lr": 9.831337934693683e-05}
{"train_loss": 0.5202651619911194, "global_step": 22965, "epoch": 258, "lr": 9.83132300382132e-05}
{"train_loss": 0.4439281225204468, "global_step": 22966, "epoch": 258, "lr": 9.831308072299443e-05}
{"train_loss": 0.5679296255111694, "global_step": 22967, "epoch": 258, "lr": 9.831293140128057e-05}
{"train_loss": 0.38447487354278564, "global_step": 22968, "epoch": 258, "lr": 9.831278207307162e-05}
{"train_loss": 0.5067393183708191, "global_step": 22969, "epoch": 258, "lr": 9.83126327383676e-05}
{"train_loss": 0.531931459903717, "global_step": 22970, "epoch": 258, "lr": 9.831248339716854e-05}
{"train_loss": 0.37685441970825195, "global_step": 22971, "epoch": 258, "lr": 9.831233404947446e-05}
{"train_loss": 0.45174479484558105, "global_step": 22972, "epoch": 258, "lr": 9.831218469528537e-05}
{"train_loss": 0.44064784049987793, "global_step": 22973, "epoch": 258, "lr": 9.83120353346013e-05}
{"train_loss": 0.5491303205490112, "global_step": 22974, "epoch": 258, "lr": 9.831188596742227e-05}
{"train_loss": 0.5079524517059326, "global_step": 22975, "epoch": 258, "lr": 9.831173659374828e-05}
{"train_loss": 0.48905298113822937, "global_step": 22976, "epoch": 258, "lr": 9.831158721357938e-05}
{"train_loss": 0.35676631331443787, "global_step": 22977, "epoch": 258, "lr": 9.831143782691557e-05}
{"train_loss": 0.4553270637989044, "global_step": 22978, "epoch": 258, "lr": 9.831128843375687e-05}
{"train_loss": 0.4393683671951294, "global_step": 22979, "epoch": 258, "lr": 9.831113903410332e-05}
{"train_loss": 0.38592439889907837, "global_step": 22980, "epoch": 258, "lr": 9.831098962795491e-05}
{"train_loss": 0.48568055033683777, "global_step": 22981, "epoch": 258, "lr": 9.83108402153117e-05}
{"train_loss": 0.40801528096199036, "global_step": 22982, "epoch": 258, "lr": 9.831069079617367e-05}
{"train_loss": 0.5886555314064026, "global_step": 22983, "epoch": 258, "lr": 9.831054137054086e-05}
{"train_loss": 0.48201385140419006, "global_step": 22984, "epoch": 258, "lr": 9.831039193841327e-05}
{"train_loss": 0.45640525221824646, "global_step": 22985, "epoch": 258, "lr": 9.831024249979094e-05}
{"train_loss": 0.5099917054176331, "global_step": 22986, "epoch": 258, "lr": 9.831009305467391e-05}
{"train_loss": 0.5428341627120972, "global_step": 22987, "epoch": 258, "lr": 9.830994360306217e-05}
{"train_loss": 0.36703208088874817, "global_step": 22988, "epoch": 258, "lr": 9.830979414495574e-05}
{"train_loss": 0.3886001706123352, "global_step": 22989, "epoch": 258, "lr": 9.830964468035466e-05}
{"train_loss": 0.4622576832771301, "global_step": 22990, "epoch": 258, "lr": 9.830949520925892e-05}
{"train_loss": 0.40755197405815125, "global_step": 22991, "epoch": 258, "lr": 9.830934573166855e-05}
{"train_loss": 0.47901707887649536, "global_step": 22992, "epoch": 258, "lr": 9.83091962475836e-05}
{"train_loss": 0.5179657936096191, "global_step": 22993, "epoch": 258, "lr": 9.830904675700405e-05}
{"train_loss": 0.4809480905532837, "global_step": 22994, "epoch": 258, "lr": 9.830889725992995e-05}
{"train_loss": 0.4171447455883026, "global_step": 22995, "epoch": 258, "lr": 9.83087477563613e-05}
{"train_loss": 0.4600655138492584, "global_step": 22996, "epoch": 258, "lr": 9.830859824629813e-05}
{"train_loss": 0.4040357172489166, "global_step": 22997, "epoch": 258, "lr": 9.830844872974047e-05}
{"train_loss": 0.49712416529655457, "global_step": 22998, "epoch": 258, "lr": 9.830829920668831e-05}
{"train_loss": 0.49125394225120544, "global_step": 22999, "epoch": 258, "lr": 9.830814967714169e-05}
{"train_loss": 0.4789254665374756, "global_step": 23000, "epoch": 258, "lr": 9.830800014110064e-05}
{"train_loss": 0.35377442836761475, "global_step": 23001, "epoch": 258, "lr": 9.830785059856515e-05}
{"train_loss": 0.5593341588973999, "global_step": 23002, "epoch": 258, "lr": 9.830770104953527e-05}
{"train_loss": 0.49146604537963867, "global_step": 23003, "epoch": 258, "lr": 9.830755149401101e-05}
{"train_loss": 0.48012205958366394, "global_step": 23004, "epoch": 258, "lr": 9.830740193199239e-05}
{"train_loss": 0.5209227204322815, "global_step": 23005, "epoch": 258, "lr": 9.830725236347942e-05}
{"train_loss": 0.47821879386901855, "global_step": 23006, "epoch": 258, "lr": 9.830710278847213e-05}
{"train_loss": 0.3987581133842468, "global_step": 23007, "epoch": 258, "lr": 9.830695320697054e-05}
{"train_loss": 0.4396932125091553, "global_step": 23008, "epoch": 258, "lr": 9.830680361897467e-05}
{"train_loss": 0.5025968551635742, "global_step": 23009, "epoch": 258, "lr": 9.830665402448453e-05}
{"train_loss": 0.46597820520401, "global_step": 23010, "epoch": 258, "lr": 9.830650442350018e-05}
{"train_loss": 0.39587870240211487, "global_step": 23011, "epoch": 258, "lr": 9.830635481602158e-05}
{"train_loss": 0.4870249330997467, "global_step": 23012, "epoch": 258, "lr": 9.830620520204878e-05}
{"train_loss": 0.46321624517440796, "global_step": 23013, "epoch": 258, "lr": 9.830605558158181e-05}
{"train_loss": 0.4259952902793884, "global_step": 23014, "epoch": 258, "lr": 9.830590595462068e-05}
{"train_loss": 0.5157124996185303, "global_step": 23015, "epoch": 258, "lr": 9.83057563211654e-05}
{"train_loss": 0.43645066022872925, "global_step": 23016, "epoch": 258, "lr": 9.830560668121601e-05}
{"train_loss": 0.43899402022361755, "global_step": 23017, "epoch": 258, "lr": 9.830545703477252e-05}
{"train_loss": 0.40380898118019104, "global_step": 23018, "epoch": 258, "lr": 9.830530738183494e-05}
{"train_loss": 0.4220982789993286, "global_step": 23019, "epoch": 258, "lr": 9.830515772240331e-05}
{"train_loss": 0.43450847268104553, "global_step": 23020, "epoch": 258, "lr": 9.830500805647763e-05}
{"train_loss": 0.557973325252533, "global_step": 23021, "epoch": 258, "lr": 9.830485838405794e-05}
{"train_loss": 0.4577191472053528, "global_step": 23022, "epoch": 258, "lr": 9.830470870514425e-05}
{"train_loss": 0.4606371819972992, "global_step": 23023, "epoch": 258, "lr": 9.830455901973657e-05}
{"train_loss": 0.4242435097694397, "global_step": 23024, "epoch": 258, "lr": 9.830440932783495e-05}
{"train_loss": 0.5061191320419312, "global_step": 23025, "epoch": 258, "lr": 9.830425962943937e-05}
{"train_loss": 0.5324950218200684, "global_step": 23026, "epoch": 258, "lr": 9.830410992454988e-05}
{"train_loss": 0.5378938317298889, "global_step": 23027, "epoch": 258, "lr": 9.83039602131665e-05}
{"train_loss": 0.4081781804561615, "global_step": 23028, "epoch": 258, "lr": 9.830381049528923e-05}
{"train_loss": 0.3814220726490021, "global_step": 23029, "epoch": 258, "lr": 9.830366077091811e-05}
{"train_loss": 0.4641053378582001, "global_step": 23030, "epoch": 258, "lr": 9.830351104005314e-05}
{"train_loss": 0.38068634271621704, "global_step": 23031, "epoch": 258, "lr": 9.830336130269436e-05}
{"train_loss": 0.460013210773468, "global_step": 23032, "epoch": 258, "lr": 9.830321155884179e-05}
{"train_loss": 0.4915294051170349, "global_step": 23033, "epoch": 258, "lr": 9.830306180849543e-05}
{"train_loss": 0.3715125620365143, "global_step": 23034, "epoch": 258, "lr": 9.83029120516553e-05}
{"train_loss": 0.48935040831565857, "global_step": 23035, "epoch": 258, "lr": 9.830276228832145e-05}
{"train_loss": 0.4912737011909485, "global_step": 23036, "epoch": 258, "lr": 9.830261251849387e-05}
{"train_loss": 0.4733373522758484, "global_step": 23037, "epoch": 258, "lr": 9.83024627421726e-05}
{"train_loss": 0.4029231667518616, "global_step": 23038, "epoch": 258, "lr": 9.830231295935766e-05}
{"train_loss": 0.516566812992096, "global_step": 23039, "epoch": 258, "lr": 9.830216317004905e-05}
{"train_loss": 0.3965567946434021, "global_step": 23040, "epoch": 258, "lr": 9.830201337424681e-05}
{"train_loss": 0.49044328927993774, "global_step": 23041, "epoch": 258, "lr": 9.830186357195096e-05}
{"train_loss": 0.4793563485145569, "global_step": 23042, "epoch": 258, "lr": 9.830171376316151e-05}
{"train_loss": 0.5468053817749023, "global_step": 23043, "epoch": 258, "lr": 9.830156394787846e-05}
{"train_loss": 0.3627098798751831, "global_step": 23044, "epoch": 258, "lr": 9.830141412610188e-05}
{"train_loss": 0.4979855418205261, "global_step": 23045, "epoch": 258, "lr": 9.830126429783175e-05}
{"train_loss": 0.46116340160369873, "global_step": 23046, "epoch": 258, "lr": 9.830111446306811e-05}
{"train_loss": 0.5304813385009766, "global_step": 23047, "epoch": 258, "lr": 9.830096462181098e-05}
{"train_loss": 0.5401392579078674, "global_step": 23048, "epoch": 258, "lr": 9.830081477406037e-05}
{"train_loss": 0.40675073862075806, "global_step": 23049, "epoch": 258, "lr": 9.83006649198163e-05}
{"train_loss": 0.46221507633669995, "global_step": 23050, "epoch": 258, "lr": 9.830051505907879e-05, "val_loss": 1.8290876150131226}
{"train_loss": 0.4123653173446655, "global_step": 23051, "epoch": 259, "lr": 9.830036519184786e-05}
{"train_loss": 0.3875938057899475, "global_step": 23052, "epoch": 259, "lr": 9.830021531812355e-05}
{"train_loss": 0.3811053931713104, "global_step": 23053, "epoch": 259, "lr": 9.830006543790586e-05}
{"train_loss": 0.37948736548423767, "global_step": 23054, "epoch": 259, "lr": 9.829991555119482e-05}
{"train_loss": 0.4407598078250885, "global_step": 23055, "epoch": 259, "lr": 9.829976565799043e-05}
{"train_loss": 0.37984031438827515, "global_step": 23056, "epoch": 259, "lr": 9.829961575829275e-05}
{"train_loss": 0.554370105266571, "global_step": 23057, "epoch": 259, "lr": 9.829946585210174e-05}
{"train_loss": 0.4469297528266907, "global_step": 23058, "epoch": 259, "lr": 9.829931593941748e-05}
{"train_loss": 0.3802727460861206, "global_step": 23059, "epoch": 259, "lr": 9.829916602023996e-05}
{"train_loss": 0.4250463843345642, "global_step": 23060, "epoch": 259, "lr": 9.829901609456921e-05}
{"train_loss": 0.44506990909576416, "global_step": 23061, "epoch": 259, "lr": 9.829886616240524e-05}
{"train_loss": 0.44993123412132263, "global_step": 23062, "epoch": 259, "lr": 9.829871622374809e-05}
{"train_loss": 0.442496657371521, "global_step": 23063, "epoch": 259, "lr": 9.829856627859774e-05}
{"train_loss": 0.4621475040912628, "global_step": 23064, "epoch": 259, "lr": 9.829841632695426e-05}
{"train_loss": 0.455895334482193, "global_step": 23065, "epoch": 259, "lr": 9.829826636881765e-05}
{"train_loss": 0.34321966767311096, "global_step": 23066, "epoch": 259, "lr": 9.829811640418791e-05}
{"train_loss": 0.47166112065315247, "global_step": 23067, "epoch": 259, "lr": 9.829796643306508e-05}
{"train_loss": 0.43002527952194214, "global_step": 23068, "epoch": 259, "lr": 9.829781645544917e-05}
{"train_loss": 0.6063809990882874, "global_step": 23069, "epoch": 259, "lr": 9.829766647134022e-05}
{"train_loss": 0.4223015010356903, "global_step": 23070, "epoch": 259, "lr": 9.829751648073823e-05}
{"train_loss": 0.4708464443683624, "global_step": 23071, "epoch": 259, "lr": 9.829736648364323e-05}
{"train_loss": 0.4925103783607483, "global_step": 23072, "epoch": 259, "lr": 9.829721648005525e-05}
{"train_loss": 0.4984557628631592, "global_step": 23073, "epoch": 259, "lr": 9.829706646997427e-05}
{"train_loss": 0.3891916275024414, "global_step": 23074, "epoch": 259, "lr": 9.829691645340037e-05}
{"train_loss": 0.47912532091140747, "global_step": 23075, "epoch": 259, "lr": 9.829676643033351e-05}
{"train_loss": 0.4729374945163727, "global_step": 23076, "epoch": 259, "lr": 9.829661640077376e-05}
{"train_loss": 0.3987620174884796, "global_step": 23077, "epoch": 259, "lr": 9.829646636472111e-05}
{"train_loss": 0.48451709747314453, "global_step": 23078, "epoch": 259, "lr": 9.82963163221756e-05}
{"train_loss": 0.44862601161003113, "global_step": 23079, "epoch": 259, "lr": 9.829616627313722e-05}
{"train_loss": 0.43137800693511963, "global_step": 23080, "epoch": 259, "lr": 9.829601621760602e-05}
{"train_loss": 0.3148943781852722, "global_step": 23081, "epoch": 259, "lr": 9.8295866155582e-05}
{"train_loss": 0.46297070384025574, "global_step": 23082, "epoch": 259, "lr": 9.82957160870652e-05}
{"train_loss": 0.43616512417793274, "global_step": 23083, "epoch": 259, "lr": 9.829556601205563e-05}
{"train_loss": 0.5048461556434631, "global_step": 23084, "epoch": 259, "lr": 9.82954159305533e-05}
{"train_loss": 0.48109549283981323, "global_step": 23085, "epoch": 259, "lr": 9.829526584255823e-05}
{"train_loss": 0.5045942068099976, "global_step": 23086, "epoch": 259, "lr": 9.829511574807048e-05}
{"train_loss": 0.37039241194725037, "global_step": 23087, "epoch": 259, "lr": 9.829496564709002e-05}
{"train_loss": 0.5105205178260803, "global_step": 23088, "epoch": 259, "lr": 9.829481553961691e-05}
{"train_loss": 0.4460434913635254, "global_step": 23089, "epoch": 259, "lr": 9.829466542565112e-05}
{"train_loss": 0.4372004568576813, "global_step": 23090, "epoch": 259, "lr": 9.829451530519271e-05}
{"train_loss": 0.5052387714385986, "global_step": 23091, "epoch": 259, "lr": 9.829436517824171e-05}
{"train_loss": 0.48499295115470886, "global_step": 23092, "epoch": 259, "lr": 9.829421504479811e-05}
{"train_loss": 0.44235461950302124, "global_step": 23093, "epoch": 259, "lr": 9.829406490486194e-05}
{"train_loss": 0.37638047337532043, "global_step": 23094, "epoch": 259, "lr": 9.829391475843322e-05}
{"train_loss": 0.36483141779899597, "global_step": 23095, "epoch": 259, "lr": 9.829376460551198e-05}
{"train_loss": 0.46021154522895813, "global_step": 23096, "epoch": 259, "lr": 9.829361444609821e-05}
{"train_loss": 0.4308200776576996, "global_step": 23097, "epoch": 259, "lr": 9.829346428019199e-05}
{"train_loss": 0.4760236442089081, "global_step": 23098, "epoch": 259, "lr": 9.829331410779327e-05}
{"train_loss": 0.4302966892719269, "global_step": 23099, "epoch": 259, "lr": 9.829316392890211e-05}
{"train_loss": 0.4995185434818268, "global_step": 23100, "epoch": 259, "lr": 9.829301374351853e-05}
{"train_loss": 0.4486689865589142, "global_step": 23101, "epoch": 259, "lr": 9.829286355164253e-05}
{"train_loss": 0.4778532087802887, "global_step": 23102, "epoch": 259, "lr": 9.829271335327417e-05}
{"train_loss": 0.47183462977409363, "global_step": 23103, "epoch": 259, "lr": 9.829256314841341e-05}
{"train_loss": 0.45105478167533875, "global_step": 23104, "epoch": 259, "lr": 9.829241293706033e-05}
{"train_loss": 0.3452620506286621, "global_step": 23105, "epoch": 259, "lr": 9.82922627192149e-05}
{"train_loss": 0.5050724744796753, "global_step": 23106, "epoch": 259, "lr": 9.829211249487718e-05}
{"train_loss": 0.3929772973060608, "global_step": 23107, "epoch": 259, "lr": 9.829196226404718e-05}
{"train_loss": 0.43124157190322876, "global_step": 23108, "epoch": 259, "lr": 9.829181202672491e-05}
{"train_loss": 0.47809016704559326, "global_step": 23109, "epoch": 259, "lr": 9.829166178291038e-05}
{"train_loss": 0.4221125543117523, "global_step": 23110, "epoch": 259, "lr": 9.829151153260363e-05}
{"train_loss": 0.32410919666290283, "global_step": 23111, "epoch": 259, "lr": 9.829136127580468e-05}
{"train_loss": 0.42669445276260376, "global_step": 23112, "epoch": 259, "lr": 9.829121101251354e-05}
{"train_loss": 0.44090843200683594, "global_step": 23113, "epoch": 259, "lr": 9.829106074273025e-05}
{"train_loss": 0.4833969175815582, "global_step": 23114, "epoch": 259, "lr": 9.82909104664548e-05}
{"train_loss": 0.4366115927696228, "global_step": 23115, "epoch": 259, "lr": 9.829076018368723e-05}
{"train_loss": 0.34839922189712524, "global_step": 23116, "epoch": 259, "lr": 9.829060989442756e-05}
{"train_loss": 0.4582025706768036, "global_step": 23117, "epoch": 259, "lr": 9.82904595986758e-05}
{"train_loss": 0.3885861039161682, "global_step": 23118, "epoch": 259, "lr": 9.829030929643197e-05}
{"train_loss": 0.5107443332672119, "global_step": 23119, "epoch": 259, "lr": 9.829015898769612e-05}
{"train_loss": 0.45174774527549744, "global_step": 23120, "epoch": 259, "lr": 9.829000867246823e-05}
{"train_loss": 0.36417773365974426, "global_step": 23121, "epoch": 259, "lr": 9.828985835074833e-05}
{"train_loss": 0.44621542096138, "global_step": 23122, "epoch": 259, "lr": 9.828970802253645e-05}
{"train_loss": 0.45141854882240295, "global_step": 23123, "epoch": 259, "lr": 9.828955768783262e-05}
{"train_loss": 0.3912389278411865, "global_step": 23124, "epoch": 259, "lr": 9.828940734663684e-05}
{"train_loss": 0.5703880190849304, "global_step": 23125, "epoch": 259, "lr": 9.828925699894915e-05}
{"train_loss": 0.5426069498062134, "global_step": 23126, "epoch": 259, "lr": 9.828910664476953e-05}
{"train_loss": 0.39738672971725464, "global_step": 23127, "epoch": 259, "lr": 9.828895628409804e-05}
{"train_loss": 0.41467687487602234, "global_step": 23128, "epoch": 259, "lr": 9.82888059169347e-05}
{"train_loss": 0.3425735831260681, "global_step": 23129, "epoch": 259, "lr": 9.828865554327952e-05}
{"train_loss": 0.43966609239578247, "global_step": 23130, "epoch": 259, "lr": 9.828850516313251e-05}
{"train_loss": 0.385466068983078, "global_step": 23131, "epoch": 259, "lr": 9.82883547764937e-05}
{"train_loss": 0.3055298924446106, "global_step": 23132, "epoch": 259, "lr": 9.82882043833631e-05}
{"train_loss": 0.4740767478942871, "global_step": 23133, "epoch": 259, "lr": 9.828805398374074e-05}
{"train_loss": 0.42837846279144287, "global_step": 23134, "epoch": 259, "lr": 9.828790357762666e-05}
{"train_loss": 0.3559894561767578, "global_step": 23135, "epoch": 259, "lr": 9.828775316502085e-05}
{"train_loss": 0.44069352746009827, "global_step": 23136, "epoch": 259, "lr": 9.828760274592334e-05}
{"train_loss": 0.332004576921463, "global_step": 23137, "epoch": 259, "lr": 9.828745232033414e-05}
{"train_loss": 0.401487797498703, "global_step": 23138, "epoch": 259, "lr": 9.828730188825329e-05}
{"train_loss": 0.4366651006628958, "global_step": 23139, "epoch": 259, "lr": 9.82871514496808e-05, "val_loss": 1.9018144607543945}
{"train_loss": 0.4641992449760437, "global_step": 23140, "epoch": 260, "lr": 9.82870010046167e-05}
{"train_loss": 0.4494459629058838, "global_step": 23141, "epoch": 260, "lr": 9.828685055306098e-05}
{"train_loss": 0.42538973689079285, "global_step": 23142, "epoch": 260, "lr": 9.82867000950137e-05}
{"train_loss": 0.5224913954734802, "global_step": 23143, "epoch": 260, "lr": 9.828654963047485e-05}
{"train_loss": 0.3758334815502167, "global_step": 23144, "epoch": 260, "lr": 9.828639915944447e-05}
{"train_loss": 0.4410572052001953, "global_step": 23145, "epoch": 260, "lr": 9.828624868192256e-05}
{"train_loss": 0.36104902625083923, "global_step": 23146, "epoch": 260, "lr": 9.828609819790917e-05}
{"train_loss": 0.4318455457687378, "global_step": 23147, "epoch": 260, "lr": 9.828594770740429e-05}
{"train_loss": 0.3745113015174866, "global_step": 23148, "epoch": 260, "lr": 9.828579721040796e-05}
{"train_loss": 0.42233753204345703, "global_step": 23149, "epoch": 260, "lr": 9.828564670692019e-05}
{"train_loss": 0.3955278992652893, "global_step": 23150, "epoch": 260, "lr": 9.828549619694101e-05}
{"train_loss": 0.36194169521331787, "global_step": 23151, "epoch": 260, "lr": 9.828534568047043e-05}
{"train_loss": 0.23598843812942505, "global_step": 23152, "epoch": 260, "lr": 9.828519515750847e-05}
{"train_loss": 0.4453117251396179, "global_step": 23153, "epoch": 260, "lr": 9.828504462805514e-05}
{"train_loss": 0.5755831599235535, "global_step": 23154, "epoch": 260, "lr": 9.82848940921105e-05}
{"train_loss": 0.44917792081832886, "global_step": 23155, "epoch": 260, "lr": 9.828474354967453e-05}
{"train_loss": 0.4476326107978821, "global_step": 23156, "epoch": 260, "lr": 9.828459300074726e-05}
{"train_loss": 0.4762950539588928, "global_step": 23157, "epoch": 260, "lr": 9.828444244532873e-05}
{"train_loss": 0.42098724842071533, "global_step": 23158, "epoch": 260, "lr": 9.828429188341895e-05}
{"train_loss": 0.47322505712509155, "global_step": 23159, "epoch": 260, "lr": 9.828414131501791e-05}
{"train_loss": 0.33142992854118347, "global_step": 23160, "epoch": 260, "lr": 9.828399074012566e-05}
{"train_loss": 0.3447728455066681, "global_step": 23161, "epoch": 260, "lr": 9.828384015874224e-05}
{"train_loss": 0.5169296860694885, "global_step": 23162, "epoch": 260, "lr": 9.828368957086762e-05}
{"train_loss": 0.43327364325523376, "global_step": 23163, "epoch": 260, "lr": 9.828353897650186e-05}
{"train_loss": 0.5358280539512634, "global_step": 23164, "epoch": 260, "lr": 9.828338837564497e-05}
{"train_loss": 0.4283190965652466, "global_step": 23165, "epoch": 260, "lr": 9.828323776829696e-05}
{"train_loss": 0.39552173018455505, "global_step": 23166, "epoch": 260, "lr": 9.828308715445787e-05}
{"train_loss": 0.37471768260002136, "global_step": 23167, "epoch": 260, "lr": 9.828293653412769e-05}
{"train_loss": 0.46260136365890503, "global_step": 23168, "epoch": 260, "lr": 9.828278590730647e-05}
{"train_loss": 0.49069347977638245, "global_step": 23169, "epoch": 260, "lr": 9.828263527399421e-05}
{"train_loss": 0.36179494857788086, "global_step": 23170, "epoch": 260, "lr": 9.828248463419094e-05}
{"train_loss": 0.42158234119415283, "global_step": 23171, "epoch": 260, "lr": 9.828233398789668e-05}
{"train_loss": 0.38349995017051697, "global_step": 23172, "epoch": 260, "lr": 9.828218333511143e-05}
{"train_loss": 0.46661779284477234, "global_step": 23173, "epoch": 260, "lr": 9.828203267583526e-05}
{"train_loss": 0.4023842513561249, "global_step": 23174, "epoch": 260, "lr": 9.828188201006813e-05}
{"train_loss": 0.3426726758480072, "global_step": 23175, "epoch": 260, "lr": 9.828173133781011e-05}
{"train_loss": 0.4630166292190552, "global_step": 23176, "epoch": 260, "lr": 9.82815806590612e-05}
{"train_loss": 0.4152921438217163, "global_step": 23177, "epoch": 260, "lr": 9.828142997382142e-05}
{"train_loss": 0.41678375005722046, "global_step": 23178, "epoch": 260, "lr": 9.828127928209077e-05}
{"train_loss": 0.4063403308391571, "global_step": 23179, "epoch": 260, "lr": 9.82811285838693e-05}
{"train_loss": 0.43672606348991394, "global_step": 23180, "epoch": 260, "lr": 9.828097787915704e-05}
{"train_loss": 0.3984808325767517, "global_step": 23181, "epoch": 260, "lr": 9.828082716795397e-05}
{"train_loss": 0.32950037717819214, "global_step": 23182, "epoch": 260, "lr": 9.828067645026015e-05}
{"train_loss": 0.4062276780605316, "global_step": 23183, "epoch": 260, "lr": 9.828052572607556e-05}
{"train_loss": 0.49488365650177, "global_step": 23184, "epoch": 260, "lr": 9.828037499540027e-05}
{"train_loss": 0.31755533814430237, "global_step": 23185, "epoch": 260, "lr": 9.828022425823424e-05}
{"train_loss": 0.4295690953731537, "global_step": 23186, "epoch": 260, "lr": 9.828007351457754e-05}
{"train_loss": 0.39989155530929565, "global_step": 23187, "epoch": 260, "lr": 9.827992276443016e-05}
{"train_loss": 0.5219054818153381, "global_step": 23188, "epoch": 260, "lr": 9.827977200779215e-05}
{"train_loss": 0.5511387586593628, "global_step": 23189, "epoch": 260, "lr": 9.82796212446635e-05}
{"train_loss": 0.40207546949386597, "global_step": 23190, "epoch": 260, "lr": 9.827947047504424e-05}
{"train_loss": 0.381317138671875, "global_step": 23191, "epoch": 260, "lr": 9.82793196989344e-05}
{"train_loss": 0.4169102609157562, "global_step": 23192, "epoch": 260, "lr": 9.8279168916334e-05}
{"train_loss": 0.4134308099746704, "global_step": 23193, "epoch": 260, "lr": 9.827901812724305e-05}
{"train_loss": 0.5752322673797607, "global_step": 23194, "epoch": 260, "lr": 9.827886733166157e-05}
{"train_loss": 0.47506260871887207, "global_step": 23195, "epoch": 260, "lr": 9.82787165295896e-05}
{"train_loss": 0.525359034538269, "global_step": 23196, "epoch": 260, "lr": 9.827856572102712e-05}
{"train_loss": 0.3447601795196533, "global_step": 23197, "epoch": 260, "lr": 9.82784149059742e-05}
{"train_loss": 0.4776262044906616, "global_step": 23198, "epoch": 260, "lr": 9.827826408443082e-05}
{"train_loss": 0.40779170393943787, "global_step": 23199, "epoch": 260, "lr": 9.827811325639701e-05}
{"train_loss": 0.33039501309394836, "global_step": 23200, "epoch": 260, "lr": 9.827796242187281e-05}
{"train_loss": 0.3984895944595337, "global_step": 23201, "epoch": 260, "lr": 9.827781158085823e-05}
{"train_loss": 0.515839159488678, "global_step": 23202, "epoch": 260, "lr": 9.827766073335327e-05}
{"train_loss": 0.441731721162796, "global_step": 23203, "epoch": 260, "lr": 9.827750987935798e-05}
{"train_loss": 0.4105224609375, "global_step": 23204, "epoch": 260, "lr": 9.827735901887236e-05}
{"train_loss": 0.4839845895767212, "global_step": 23205, "epoch": 260, "lr": 9.827720815189644e-05}
{"train_loss": 0.6126887798309326, "global_step": 23206, "epoch": 260, "lr": 9.827705727843024e-05}
{"train_loss": 0.4479621350765228, "global_step": 23207, "epoch": 260, "lr": 9.827690639847378e-05}
{"train_loss": 0.4558006525039673, "global_step": 23208, "epoch": 260, "lr": 9.827675551202709e-05}
{"train_loss": 0.45845386385917664, "global_step": 23209, "epoch": 260, "lr": 9.827660461909015e-05}
{"train_loss": 0.4781816303730011, "global_step": 23210, "epoch": 260, "lr": 9.827645371966303e-05}
{"train_loss": 0.4715043902397156, "global_step": 23211, "epoch": 260, "lr": 9.827630281374571e-05}
{"train_loss": 0.39481812715530396, "global_step": 23212, "epoch": 260, "lr": 9.827615190133825e-05}
{"train_loss": 0.44278204441070557, "global_step": 23213, "epoch": 260, "lr": 9.827600098244063e-05}
{"train_loss": 0.5216806530952454, "global_step": 23214, "epoch": 260, "lr": 9.827585005705291e-05}
{"train_loss": 0.41366249322891235, "global_step": 23215, "epoch": 260, "lr": 9.827569912517509e-05}
{"train_loss": 0.5389125347137451, "global_step": 23216, "epoch": 260, "lr": 9.827554818680718e-05}
{"train_loss": 0.48405593633651733, "global_step": 23217, "epoch": 260, "lr": 9.82753972419492e-05}
{"train_loss": 0.43290412425994873, "global_step": 23218, "epoch": 260, "lr": 9.82752462906012e-05}
{"train_loss": 0.35342246294021606, "global_step": 23219, "epoch": 260, "lr": 9.827509533276319e-05}
{"train_loss": 0.43065953254699707, "global_step": 23220, "epoch": 260, "lr": 9.827494436843515e-05}
{"train_loss": 0.4196385145187378, "global_step": 23221, "epoch": 260, "lr": 9.827479339761716e-05}
{"train_loss": 0.46527063846588135, "global_step": 23222, "epoch": 260, "lr": 9.82746424203092e-05}
{"train_loss": 0.4532248377799988, "global_step": 23223, "epoch": 260, "lr": 9.82744914365113e-05}
{"train_loss": 0.3984996974468231, "global_step": 23224, "epoch": 260, "lr": 9.827434044622349e-05}
{"train_loss": 0.49116095900535583, "global_step": 23225, "epoch": 260, "lr": 9.827418944944577e-05}
{"train_loss": 0.4847394824028015, "global_step": 23226, "epoch": 260, "lr": 9.82740384461782e-05}
{"train_loss": 0.39459148049354553, "global_step": 23227, "epoch": 260, "lr": 9.827388743642075e-05}
{"train_loss": 0.43454810675610317, "global_step": 23228, "epoch": 260, "lr": 9.827373642017348e-05, "val_loss": 1.909659743309021, "train_action_mse_error": 47.98323059082031}
{"train_loss": 0.48293280601501465, "global_step": 23229, "epoch": 261, "lr": 9.827358539743639e-05}
{"train_loss": 0.5395842790603638, "global_step": 23230, "epoch": 261, "lr": 9.827343436820949e-05}
{"train_loss": 0.6038019061088562, "global_step": 23231, "epoch": 261, "lr": 9.827328333249284e-05}
{"train_loss": 0.443705290555954, "global_step": 23232, "epoch": 261, "lr": 9.827313229028642e-05}
{"train_loss": 0.41374483704566956, "global_step": 23233, "epoch": 261, "lr": 9.827298124159027e-05}
{"train_loss": 0.4932900071144104, "global_step": 23234, "epoch": 261, "lr": 9.82728301864044e-05}
{"train_loss": 0.5238640308380127, "global_step": 23235, "epoch": 261, "lr": 9.827267912472883e-05}
{"train_loss": 0.33394655585289, "global_step": 23236, "epoch": 261, "lr": 9.82725280565636e-05}
{"train_loss": 0.5288476943969727, "global_step": 23237, "epoch": 261, "lr": 9.827237698190873e-05}
{"train_loss": 0.41450604796409607, "global_step": 23238, "epoch": 261, "lr": 9.82722259007642e-05}
{"train_loss": 0.4114416837692261, "global_step": 23239, "epoch": 261, "lr": 9.827207481313008e-05}
{"train_loss": 0.5270419716835022, "global_step": 23240, "epoch": 261, "lr": 9.827192371900635e-05}
{"train_loss": 0.5308526158332825, "global_step": 23241, "epoch": 261, "lr": 9.827177261839306e-05}
{"train_loss": 0.4477618634700775, "global_step": 23242, "epoch": 261, "lr": 9.827162151129021e-05}
{"train_loss": 0.4074479043483734, "global_step": 23243, "epoch": 261, "lr": 9.827147039769784e-05}
{"train_loss": 0.3762989342212677, "global_step": 23244, "epoch": 261, "lr": 9.827131927761595e-05}
{"train_loss": 0.4161822497844696, "global_step": 23245, "epoch": 261, "lr": 9.827116815104457e-05}
{"train_loss": 0.44644635915756226, "global_step": 23246, "epoch": 261, "lr": 9.827101701798371e-05}
{"train_loss": 0.4042627215385437, "global_step": 23247, "epoch": 261, "lr": 9.827086587843341e-05}
{"train_loss": 0.40441685914993286, "global_step": 23248, "epoch": 261, "lr": 9.827071473239369e-05}
{"train_loss": 0.34038791060447693, "global_step": 23249, "epoch": 261, "lr": 9.827056357986455e-05}
{"train_loss": 0.39900919795036316, "global_step": 23250, "epoch": 261, "lr": 9.827041242084603e-05}
{"train_loss": 0.4099424183368683, "global_step": 23251, "epoch": 261, "lr": 9.827026125533814e-05}
{"train_loss": 0.4985131025314331, "global_step": 23252, "epoch": 261, "lr": 9.827011008334089e-05}
{"train_loss": 0.4250822067260742, "global_step": 23253, "epoch": 261, "lr": 9.826995890485431e-05}
{"train_loss": 0.39308032393455505, "global_step": 23254, "epoch": 261, "lr": 9.826980771987844e-05}
{"train_loss": 0.5036585330963135, "global_step": 23255, "epoch": 261, "lr": 9.826965652841326e-05}
{"train_loss": 0.45526865124702454, "global_step": 23256, "epoch": 261, "lr": 9.826950533045885e-05}
{"train_loss": 0.4076319932937622, "global_step": 23257, "epoch": 261, "lr": 9.826935412601516e-05}
{"train_loss": 0.4536776542663574, "global_step": 23258, "epoch": 261, "lr": 9.826920291508226e-05}
{"train_loss": 0.5178444385528564, "global_step": 23259, "epoch": 261, "lr": 9.826905169766014e-05}
{"train_loss": 0.4508935213088989, "global_step": 23260, "epoch": 261, "lr": 9.826890047374883e-05}
{"train_loss": 0.3729448914527893, "global_step": 23261, "epoch": 261, "lr": 9.826874924334838e-05}
{"train_loss": 0.5538983345031738, "global_step": 23262, "epoch": 261, "lr": 9.826859800645878e-05}
{"train_loss": 0.5200185179710388, "global_step": 23263, "epoch": 261, "lr": 9.826844676308004e-05}
{"train_loss": 0.5124718546867371, "global_step": 23264, "epoch": 261, "lr": 9.82682955132122e-05}
{"train_loss": 0.596921980381012, "global_step": 23265, "epoch": 261, "lr": 9.826814425685528e-05}
{"train_loss": 0.41239961981773376, "global_step": 23266, "epoch": 261, "lr": 9.82679929940093e-05}
{"train_loss": 0.5565190315246582, "global_step": 23267, "epoch": 261, "lr": 9.826784172467427e-05}
{"train_loss": 0.4780340790748596, "global_step": 23268, "epoch": 261, "lr": 9.826769044885022e-05}
{"train_loss": 0.5679822564125061, "global_step": 23269, "epoch": 261, "lr": 9.826753916653715e-05}
{"train_loss": 0.5078356862068176, "global_step": 23270, "epoch": 261, "lr": 9.826738787773513e-05}
{"train_loss": 0.4479779303073883, "global_step": 23271, "epoch": 261, "lr": 9.826723658244412e-05}
{"train_loss": 0.4411115050315857, "global_step": 23272, "epoch": 261, "lr": 9.826708528066419e-05}
{"train_loss": 0.3689330518245697, "global_step": 23273, "epoch": 261, "lr": 9.826693397239532e-05}
{"train_loss": 0.4506053924560547, "global_step": 23274, "epoch": 261, "lr": 9.826678265763756e-05}
{"train_loss": 0.5130060315132141, "global_step": 23275, "epoch": 261, "lr": 9.826663133639092e-05}
{"train_loss": 0.46513715386390686, "global_step": 23276, "epoch": 261, "lr": 9.826648000865542e-05}
{"train_loss": 0.4886322021484375, "global_step": 23277, "epoch": 261, "lr": 9.826632867443108e-05}
{"train_loss": 0.5982626676559448, "global_step": 23278, "epoch": 261, "lr": 9.82661773337179e-05}
{"train_loss": 0.3372938930988312, "global_step": 23279, "epoch": 261, "lr": 9.826602598651596e-05}
{"train_loss": 0.520759642124176, "global_step": 23280, "epoch": 261, "lr": 9.82658746328252e-05}
{"train_loss": 0.46042749285697937, "global_step": 23281, "epoch": 261, "lr": 9.826572327264571e-05}
{"train_loss": 0.4826578199863434, "global_step": 23282, "epoch": 261, "lr": 9.826557190597747e-05}
{"train_loss": 0.4728047549724579, "global_step": 23283, "epoch": 261, "lr": 9.82654205328205e-05}
{"train_loss": 0.5489510297775269, "global_step": 23284, "epoch": 261, "lr": 9.826526915317487e-05}
{"train_loss": 0.37948131561279297, "global_step": 23285, "epoch": 261, "lr": 9.826511776704053e-05}
{"train_loss": 0.5340086817741394, "global_step": 23286, "epoch": 261, "lr": 9.826496637441753e-05}
{"train_loss": 0.43147966265678406, "global_step": 23287, "epoch": 261, "lr": 9.826481497530592e-05}
{"train_loss": 0.45963218808174133, "global_step": 23288, "epoch": 261, "lr": 9.826466356970567e-05}
{"train_loss": 0.470359206199646, "global_step": 23289, "epoch": 261, "lr": 9.826451215761685e-05}
{"train_loss": 0.3396379351615906, "global_step": 23290, "epoch": 261, "lr": 9.826436073903944e-05}
{"train_loss": 0.38066133856773376, "global_step": 23291, "epoch": 261, "lr": 9.826420931397347e-05}
{"train_loss": 0.45403963327407837, "global_step": 23292, "epoch": 261, "lr": 9.826405788241897e-05}
{"train_loss": 0.5443994998931885, "global_step": 23293, "epoch": 261, "lr": 9.826390644437596e-05}
{"train_loss": 0.4457065761089325, "global_step": 23294, "epoch": 261, "lr": 9.826375499984444e-05}
{"train_loss": 0.3586789071559906, "global_step": 23295, "epoch": 261, "lr": 9.826360354882446e-05}
{"train_loss": 0.5100854635238647, "global_step": 23296, "epoch": 261, "lr": 9.826345209131602e-05}
{"train_loss": 0.5803560018539429, "global_step": 23297, "epoch": 261, "lr": 9.826330062731916e-05}
{"train_loss": 0.5938557982444763, "global_step": 23298, "epoch": 261, "lr": 9.826314915683387e-05}
{"train_loss": 0.4473860263824463, "global_step": 23299, "epoch": 261, "lr": 9.82629976798602e-05}
{"train_loss": 0.44693782925605774, "global_step": 23300, "epoch": 261, "lr": 9.826284619639816e-05}
{"train_loss": 0.4209619462490082, "global_step": 23301, "epoch": 261, "lr": 9.826269470644775e-05}
{"train_loss": 0.3392893671989441, "global_step": 23302, "epoch": 261, "lr": 9.826254321000903e-05}
{"train_loss": 0.5094774961471558, "global_step": 23303, "epoch": 261, "lr": 9.826239170708199e-05}
{"train_loss": 0.4344449043273926, "global_step": 23304, "epoch": 261, "lr": 9.826224019766665e-05}
{"train_loss": 0.41837430000305176, "global_step": 23305, "epoch": 261, "lr": 9.826208868176307e-05}
{"train_loss": 0.3595784902572632, "global_step": 23306, "epoch": 261, "lr": 9.826193715937122e-05}
{"train_loss": 0.3994978964328766, "global_step": 23307, "epoch": 261, "lr": 9.826178563049114e-05}
{"train_loss": 0.43048232793807983, "global_step": 23308, "epoch": 261, "lr": 9.826163409512286e-05}
{"train_loss": 0.3933282494544983, "global_step": 23309, "epoch": 261, "lr": 9.826148255326637e-05}
{"train_loss": 0.47692573070526123, "global_step": 23310, "epoch": 261, "lr": 9.826133100492174e-05}
{"train_loss": 0.4568730294704437, "global_step": 23311, "epoch": 261, "lr": 9.826117945008894e-05}
{"train_loss": 0.46896040439605713, "global_step": 23312, "epoch": 261, "lr": 9.826102788876804e-05}
{"train_loss": 0.38895249366760254, "global_step": 23313, "epoch": 261, "lr": 9.826087632095899e-05}
{"train_loss": 0.465599000453949, "global_step": 23314, "epoch": 261, "lr": 9.826072474666188e-05}
{"train_loss": 0.3614727556705475, "global_step": 23315, "epoch": 261, "lr": 9.826057316587671e-05}
{"train_loss": 0.4539583921432495, "global_step": 23316, "epoch": 261, "lr": 9.826042157860349e-05}
{"train_loss": 0.4568727605798271, "global_step": 23317, "epoch": 261, "lr": 9.826026998484223e-05, "val_loss": 1.9027918577194214}
{"train_loss": 0.4545699656009674, "global_step": 23318, "epoch": 262, "lr": 9.826011838459299e-05}
{"train_loss": 0.4725300967693329, "global_step": 23319, "epoch": 262, "lr": 9.825996677785574e-05}
{"train_loss": 0.33975836634635925, "global_step": 23320, "epoch": 262, "lr": 9.825981516463055e-05}
{"train_loss": 0.47858113050460815, "global_step": 23321, "epoch": 262, "lr": 9.82596635449174e-05}
{"train_loss": 0.3719722628593445, "global_step": 23322, "epoch": 262, "lr": 9.825951191871632e-05}
{"train_loss": 0.5060914754867554, "global_step": 23323, "epoch": 262, "lr": 9.825936028602735e-05}
{"train_loss": 0.34409084916114807, "global_step": 23324, "epoch": 262, "lr": 9.82592086468505e-05}
{"train_loss": 0.3580215573310852, "global_step": 23325, "epoch": 262, "lr": 9.825905700118578e-05}
{"train_loss": 0.34791702032089233, "global_step": 23326, "epoch": 262, "lr": 9.825890534903322e-05}
{"train_loss": 0.3833136558532715, "global_step": 23327, "epoch": 262, "lr": 9.825875369039283e-05}
{"train_loss": 0.4720972180366516, "global_step": 23328, "epoch": 262, "lr": 9.825860202526465e-05}
{"train_loss": 0.4881610572338104, "global_step": 23329, "epoch": 262, "lr": 9.825845035364868e-05}
{"train_loss": 0.36252111196517944, "global_step": 23330, "epoch": 262, "lr": 9.825829867554496e-05}
{"train_loss": 0.506491482257843, "global_step": 23331, "epoch": 262, "lr": 9.825814699095349e-05}
{"train_loss": 0.36142122745513916, "global_step": 23332, "epoch": 262, "lr": 9.82579952998743e-05}
{"train_loss": 0.480913907289505, "global_step": 23333, "epoch": 262, "lr": 9.825784360230743e-05}
{"train_loss": 0.553240180015564, "global_step": 23334, "epoch": 262, "lr": 9.825769189825286e-05}
{"train_loss": 0.4811680316925049, "global_step": 23335, "epoch": 262, "lr": 9.825754018771063e-05}
{"train_loss": 0.3691137731075287, "global_step": 23336, "epoch": 262, "lr": 9.825738847068077e-05}
{"train_loss": 0.42889559268951416, "global_step": 23337, "epoch": 262, "lr": 9.82572367471633e-05}
{"train_loss": 0.3664534389972687, "global_step": 23338, "epoch": 262, "lr": 9.825708501715822e-05}
{"train_loss": 0.47290971875190735, "global_step": 23339, "epoch": 262, "lr": 9.825693328066556e-05}
{"train_loss": 0.4428616166114807, "global_step": 23340, "epoch": 262, "lr": 9.825678153768536e-05}
{"train_loss": 0.4912784695625305, "global_step": 23341, "epoch": 262, "lr": 9.825662978821762e-05}
{"train_loss": 0.3349660634994507, "global_step": 23342, "epoch": 262, "lr": 9.825647803226236e-05}
{"train_loss": 0.48954543471336365, "global_step": 23343, "epoch": 262, "lr": 9.82563262698196e-05}
{"train_loss": 0.32643499970436096, "global_step": 23344, "epoch": 262, "lr": 9.825617450088938e-05}
{"train_loss": 0.3933066129684448, "global_step": 23345, "epoch": 262, "lr": 9.825602272547168e-05}
{"train_loss": 0.3843505382537842, "global_step": 23346, "epoch": 262, "lr": 9.825587094356656e-05}
{"train_loss": 0.4638567268848419, "global_step": 23347, "epoch": 262, "lr": 9.825571915517404e-05}
{"train_loss": 0.448819100856781, "global_step": 23348, "epoch": 262, "lr": 9.825556736029411e-05}
{"train_loss": 0.4905610680580139, "global_step": 23349, "epoch": 262, "lr": 9.825541555892682e-05}
{"train_loss": 0.5378899574279785, "global_step": 23350, "epoch": 262, "lr": 9.825526375107217e-05}
{"train_loss": 0.4671138823032379, "global_step": 23351, "epoch": 262, "lr": 9.825511193673018e-05}
{"train_loss": 0.43015629053115845, "global_step": 23352, "epoch": 262, "lr": 9.825496011590088e-05}
{"train_loss": 0.4279906451702118, "global_step": 23353, "epoch": 262, "lr": 9.825480828858428e-05}
{"train_loss": 0.385333776473999, "global_step": 23354, "epoch": 262, "lr": 9.825465645478045e-05}
{"train_loss": 0.5087010264396667, "global_step": 23355, "epoch": 262, "lr": 9.825450461448933e-05}
{"train_loss": 0.4583773612976074, "global_step": 23356, "epoch": 262, "lr": 9.8254352767711e-05}
{"train_loss": 0.4961283504962921, "global_step": 23357, "epoch": 262, "lr": 9.825420091444545e-05}
{"train_loss": 0.5051714777946472, "global_step": 23358, "epoch": 262, "lr": 9.82540490546927e-05}
{"train_loss": 0.4447476267814636, "global_step": 23359, "epoch": 262, "lr": 9.82538971884528e-05}
{"train_loss": 0.43467751145362854, "global_step": 23360, "epoch": 262, "lr": 9.825374531572576e-05}
{"train_loss": 0.3183808922767639, "global_step": 23361, "epoch": 262, "lr": 9.825359343651157e-05}
{"train_loss": 0.2911999523639679, "global_step": 23362, "epoch": 262, "lr": 9.825344155081029e-05}
{"train_loss": 0.5108692049980164, "global_step": 23363, "epoch": 262, "lr": 9.82532896586219e-05}
{"train_loss": 0.4122943580150604, "global_step": 23364, "epoch": 262, "lr": 9.825313775994647e-05}
{"train_loss": 0.5174790620803833, "global_step": 23365, "epoch": 262, "lr": 9.825298585478398e-05}
{"train_loss": 0.3976357579231262, "global_step": 23366, "epoch": 262, "lr": 9.825283394313446e-05}
{"train_loss": 0.5020068883895874, "global_step": 23367, "epoch": 262, "lr": 9.825268202499793e-05}
{"train_loss": 0.49498096108436584, "global_step": 23368, "epoch": 262, "lr": 9.825253010037442e-05}
{"train_loss": 0.4722476303577423, "global_step": 23369, "epoch": 262, "lr": 9.825237816926396e-05}
{"train_loss": 0.4728015065193176, "global_step": 23370, "epoch": 262, "lr": 9.825222623166656e-05}
{"train_loss": 0.359127014875412, "global_step": 23371, "epoch": 262, "lr": 9.825207428758221e-05}
{"train_loss": 0.4363533556461334, "global_step": 23372, "epoch": 262, "lr": 9.825192233701097e-05}
{"train_loss": 0.4184262752532959, "global_step": 23373, "epoch": 262, "lr": 9.825177037995286e-05}
{"train_loss": 0.3999081254005432, "global_step": 23374, "epoch": 262, "lr": 9.825161841640786e-05}
{"train_loss": 0.441597580909729, "global_step": 23375, "epoch": 262, "lr": 9.825146644637604e-05}
{"train_loss": 0.3898528218269348, "global_step": 23376, "epoch": 262, "lr": 9.825131446985739e-05}
{"train_loss": 0.431455135345459, "global_step": 23377, "epoch": 262, "lr": 9.825116248685193e-05}
{"train_loss": 0.5272714495658875, "global_step": 23378, "epoch": 262, "lr": 9.825101049735971e-05}
{"train_loss": 0.3887260854244232, "global_step": 23379, "epoch": 262, "lr": 9.825085850138072e-05}
{"train_loss": 0.44303667545318604, "global_step": 23380, "epoch": 262, "lr": 9.8250706498915e-05}
{"train_loss": 0.4399689733982086, "global_step": 23381, "epoch": 262, "lr": 9.825055448996253e-05}
{"train_loss": 0.4890762269496918, "global_step": 23382, "epoch": 262, "lr": 9.825040247452338e-05}
{"train_loss": 0.43979164958000183, "global_step": 23383, "epoch": 262, "lr": 9.825025045259755e-05}
{"train_loss": 0.3907744586467743, "global_step": 23384, "epoch": 262, "lr": 9.825009842418506e-05}
{"train_loss": 0.4192291498184204, "global_step": 23385, "epoch": 262, "lr": 9.824994638928594e-05}
{"train_loss": 0.45811203122138977, "global_step": 23386, "epoch": 262, "lr": 9.82497943479002e-05}
{"train_loss": 0.45410001277923584, "global_step": 23387, "epoch": 262, "lr": 9.824964230002786e-05}
{"train_loss": 0.41595086455345154, "global_step": 23388, "epoch": 262, "lr": 9.824949024566896e-05}
{"train_loss": 0.6673415899276733, "global_step": 23389, "epoch": 262, "lr": 9.824933818482348e-05}
{"train_loss": 0.5061886310577393, "global_step": 23390, "epoch": 262, "lr": 9.824918611749148e-05}
{"train_loss": 0.4614388346672058, "global_step": 23391, "epoch": 262, "lr": 9.824903404367296e-05}
{"train_loss": 0.4029291868209839, "global_step": 23392, "epoch": 262, "lr": 9.824888196336795e-05}
{"train_loss": 0.40862107276916504, "global_step": 23393, "epoch": 262, "lr": 9.824872987657645e-05}
{"train_loss": 0.3568788468837738, "global_step": 23394, "epoch": 262, "lr": 9.82485777832985e-05}
{"train_loss": 0.3567574918270111, "global_step": 23395, "epoch": 262, "lr": 9.824842568353413e-05}
{"train_loss": 0.434927761554718, "global_step": 23396, "epoch": 262, "lr": 9.824827357728334e-05}
{"train_loss": 0.3636237382888794, "global_step": 23397, "epoch": 262, "lr": 9.824812146454616e-05}
{"train_loss": 0.4388924241065979, "global_step": 23398, "epoch": 262, "lr": 9.824796934532261e-05}
{"train_loss": 0.36405324935913086, "global_step": 23399, "epoch": 262, "lr": 9.82478172196127e-05}
{"train_loss": 0.36766132712364197, "global_step": 23400, "epoch": 262, "lr": 9.824766508741648e-05}
{"train_loss": 0.5524275898933411, "global_step": 23401, "epoch": 262, "lr": 9.824751294873393e-05}
{"train_loss": 0.3758348226547241, "global_step": 23402, "epoch": 262, "lr": 9.82473608035651e-05}
{"train_loss": 0.4260171949863434, "global_step": 23403, "epoch": 262, "lr": 9.824720865190997e-05}
{"train_loss": 0.5535833239555359, "global_step": 23404, "epoch": 262, "lr": 9.824705649376862e-05}
{"train_loss": 0.4908880591392517, "global_step": 23405, "epoch": 262, "lr": 9.824690432914103e-05}
{"train_loss": 0.4369674623012543, "global_step": 23406, "epoch": 262, "lr": 9.824675215802725e-05, "val_loss": 1.8960262537002563}
{"train_loss": 0.3997139632701874, "global_step": 23407, "epoch": 263, "lr": 9.824659998042727e-05}
{"train_loss": 0.38958069682121277, "global_step": 23408, "epoch": 263, "lr": 9.824644779634113e-05}
{"train_loss": 0.5005565881729126, "global_step": 23409, "epoch": 263, "lr": 9.824629560576883e-05}
{"train_loss": 0.36773380637168884, "global_step": 23410, "epoch": 263, "lr": 9.824614340871041e-05}
{"train_loss": 0.48672744631767273, "global_step": 23411, "epoch": 263, "lr": 9.824599120516588e-05}
{"train_loss": 0.4165082275867462, "global_step": 23412, "epoch": 263, "lr": 9.824583899513527e-05}
{"train_loss": 0.3550773561000824, "global_step": 23413, "epoch": 263, "lr": 9.82456867786186e-05}
{"train_loss": 0.47987738251686096, "global_step": 23414, "epoch": 263, "lr": 9.824553455561587e-05}
{"train_loss": 0.38999149203300476, "global_step": 23415, "epoch": 263, "lr": 9.824538232612712e-05}
{"train_loss": 0.480067640542984, "global_step": 23416, "epoch": 263, "lr": 9.824523009015238e-05}
{"train_loss": 0.4030717611312866, "global_step": 23417, "epoch": 263, "lr": 9.824507784769165e-05}
{"train_loss": 0.4497370719909668, "global_step": 23418, "epoch": 263, "lr": 9.824492559874495e-05}
{"train_loss": 0.5120230317115784, "global_step": 23419, "epoch": 263, "lr": 9.824477334331231e-05}
{"train_loss": 0.5122016668319702, "global_step": 23420, "epoch": 263, "lr": 9.824462108139376e-05}
{"train_loss": 0.5235857963562012, "global_step": 23421, "epoch": 263, "lr": 9.82444688129893e-05}
{"train_loss": 0.4443644881248474, "global_step": 23422, "epoch": 263, "lr": 9.824431653809896e-05}
{"train_loss": 0.4669656753540039, "global_step": 23423, "epoch": 263, "lr": 9.824416425672276e-05}
{"train_loss": 0.3332687318325043, "global_step": 23424, "epoch": 263, "lr": 9.82440119688607e-05}
{"train_loss": 0.43531855940818787, "global_step": 23425, "epoch": 263, "lr": 9.824385967451285e-05}
{"train_loss": 0.4115789234638214, "global_step": 23426, "epoch": 263, "lr": 9.824370737367919e-05}
{"train_loss": 0.40772491693496704, "global_step": 23427, "epoch": 263, "lr": 9.824355506635977e-05}
{"train_loss": 0.48632070422172546, "global_step": 23428, "epoch": 263, "lr": 9.824340275255456e-05}
{"train_loss": 0.6132387518882751, "global_step": 23429, "epoch": 263, "lr": 9.824325043226363e-05}
{"train_loss": 0.490750789642334, "global_step": 23430, "epoch": 263, "lr": 9.824309810548698e-05}
{"train_loss": 0.2906084954738617, "global_step": 23431, "epoch": 263, "lr": 9.824294577222464e-05}
{"train_loss": 0.49713024497032166, "global_step": 23432, "epoch": 263, "lr": 9.824279343247659e-05}
{"train_loss": 0.4765050411224365, "global_step": 23433, "epoch": 263, "lr": 9.824264108624293e-05}
{"train_loss": 0.43746402859687805, "global_step": 23434, "epoch": 263, "lr": 9.824248873352362e-05}
{"train_loss": 0.39420852065086365, "global_step": 23435, "epoch": 263, "lr": 9.824233637431868e-05}
{"train_loss": 0.4707332253456116, "global_step": 23436, "epoch": 263, "lr": 9.824218400862817e-05}
{"train_loss": 0.44000786542892456, "global_step": 23437, "epoch": 263, "lr": 9.824203163645207e-05}
{"train_loss": 0.42841801047325134, "global_step": 23438, "epoch": 263, "lr": 9.824187925779042e-05}
{"train_loss": 0.44795048236846924, "global_step": 23439, "epoch": 263, "lr": 9.824172687264324e-05}
{"train_loss": 0.5624925494194031, "global_step": 23440, "epoch": 263, "lr": 9.824157448101054e-05}
{"train_loss": 0.38164180517196655, "global_step": 23441, "epoch": 263, "lr": 9.824142208289236e-05}
{"train_loss": 0.3712836503982544, "global_step": 23442, "epoch": 263, "lr": 9.82412696782887e-05}
{"train_loss": 0.4322372078895569, "global_step": 23443, "epoch": 263, "lr": 9.82411172671996e-05}
{"train_loss": 0.44576358795166016, "global_step": 23444, "epoch": 263, "lr": 9.824096484962506e-05}
{"train_loss": 0.4785243272781372, "global_step": 23445, "epoch": 263, "lr": 9.82408124255651e-05}
{"train_loss": 0.4365357756614685, "global_step": 23446, "epoch": 263, "lr": 9.824065999501977e-05}
{"train_loss": 0.418951153755188, "global_step": 23447, "epoch": 263, "lr": 9.824050755798905e-05}
{"train_loss": 0.39369019865989685, "global_step": 23448, "epoch": 263, "lr": 9.8240355114473e-05}
{"train_loss": 0.42245614528656006, "global_step": 23449, "epoch": 263, "lr": 9.824020266447162e-05}
{"train_loss": 0.47982314229011536, "global_step": 23450, "epoch": 263, "lr": 9.824005020798494e-05}
{"train_loss": 0.37068405747413635, "global_step": 23451, "epoch": 263, "lr": 9.823989774501296e-05}
{"train_loss": 0.5368217825889587, "global_step": 23452, "epoch": 263, "lr": 9.823974527555573e-05}
{"train_loss": 0.36921027302742004, "global_step": 23453, "epoch": 263, "lr": 9.823959279961324e-05}
{"train_loss": 0.4481663107872009, "global_step": 23454, "epoch": 263, "lr": 9.823944031718552e-05}
{"train_loss": 0.6028271913528442, "global_step": 23455, "epoch": 263, "lr": 9.82392878282726e-05}
{"train_loss": 0.5293265581130981, "global_step": 23456, "epoch": 263, "lr": 9.82391353328745e-05}
{"train_loss": 0.3218120336532593, "global_step": 23457, "epoch": 263, "lr": 9.823898283099124e-05}
{"train_loss": 0.39505434036254883, "global_step": 23458, "epoch": 263, "lr": 9.823883032262282e-05}
{"train_loss": 0.5252878665924072, "global_step": 23459, "epoch": 263, "lr": 9.823867780776929e-05}
{"train_loss": 0.4989900588989258, "global_step": 23460, "epoch": 263, "lr": 9.823852528643066e-05}
{"train_loss": 0.46790358424186707, "global_step": 23461, "epoch": 263, "lr": 9.823837275860694e-05}
{"train_loss": 0.3887206017971039, "global_step": 23462, "epoch": 263, "lr": 9.823822022429817e-05}
{"train_loss": 0.4931149184703827, "global_step": 23463, "epoch": 263, "lr": 9.823806768350434e-05}
{"train_loss": 0.3610144257545471, "global_step": 23464, "epoch": 263, "lr": 9.823791513622551e-05}
{"train_loss": 0.4017947316169739, "global_step": 23465, "epoch": 263, "lr": 9.823776258246167e-05}
{"train_loss": 0.44682103395462036, "global_step": 23466, "epoch": 263, "lr": 9.823761002221285e-05}
{"train_loss": 0.4831434190273285, "global_step": 23467, "epoch": 263, "lr": 9.823745745547907e-05}
{"train_loss": 0.43330541253089905, "global_step": 23468, "epoch": 263, "lr": 9.823730488226036e-05}
{"train_loss": 0.4426223933696747, "global_step": 23469, "epoch": 263, "lr": 9.823715230255672e-05}
{"train_loss": 0.41315722465515137, "global_step": 23470, "epoch": 263, "lr": 9.82369997163682e-05}
{"train_loss": 0.39818310737609863, "global_step": 23471, "epoch": 263, "lr": 9.82368471236948e-05}
{"train_loss": 0.38968971371650696, "global_step": 23472, "epoch": 263, "lr": 9.823669452453653e-05}
{"train_loss": 0.5059900879859924, "global_step": 23473, "epoch": 263, "lr": 9.823654191889344e-05}
{"train_loss": 0.40875744819641113, "global_step": 23474, "epoch": 263, "lr": 9.823638930676553e-05}
{"train_loss": 0.4494192600250244, "global_step": 23475, "epoch": 263, "lr": 9.823623668815281e-05}
{"train_loss": 0.41090407967567444, "global_step": 23476, "epoch": 263, "lr": 9.823608406305533e-05}
{"train_loss": 0.4771426022052765, "global_step": 23477, "epoch": 263, "lr": 9.82359314314731e-05}
{"train_loss": 0.47561055421829224, "global_step": 23478, "epoch": 263, "lr": 9.823577879340613e-05}
{"train_loss": 0.4483950138092041, "global_step": 23479, "epoch": 263, "lr": 9.823562614885444e-05}
{"train_loss": 0.5123395919799805, "global_step": 23480, "epoch": 263, "lr": 9.823547349781807e-05}
{"train_loss": 0.4388749301433563, "global_step": 23481, "epoch": 263, "lr": 9.823532084029704e-05}
{"train_loss": 0.443435937166214, "global_step": 23482, "epoch": 263, "lr": 9.823516817629133e-05}
{"train_loss": 0.42858272790908813, "global_step": 23483, "epoch": 263, "lr": 9.823501550580101e-05}
{"train_loss": 0.4748910069465637, "global_step": 23484, "epoch": 263, "lr": 9.823486282882608e-05}
{"train_loss": 0.3860653042793274, "global_step": 23485, "epoch": 263, "lr": 9.823471014536655e-05}
{"train_loss": 0.48471641540527344, "global_step": 23486, "epoch": 263, "lr": 9.823455745542245e-05}
{"train_loss": 0.44131162762641907, "global_step": 23487, "epoch": 263, "lr": 9.823440475899381e-05}
{"train_loss": 0.40735191106796265, "global_step": 23488, "epoch": 263, "lr": 9.823425205608065e-05}
{"train_loss": 0.4295715391635895, "global_step": 23489, "epoch": 263, "lr": 9.823409934668296e-05}
{"train_loss": 0.4050965905189514, "global_step": 23490, "epoch": 263, "lr": 9.82339466308008e-05}
{"train_loss": 0.4333013892173767, "global_step": 23491, "epoch": 263, "lr": 9.823379390843417e-05}
{"train_loss": 0.4197635352611542, "global_step": 23492, "epoch": 263, "lr": 9.82336411795831e-05}
{"train_loss": 0.3993389904499054, "global_step": 23493, "epoch": 263, "lr": 9.823348844424759e-05}
{"train_loss": 0.4470958709716797, "global_step": 23494, "epoch": 263, "lr": 9.823333570242768e-05}
{"train_loss": 0.44268895802872904, "global_step": 23495, "epoch": 263, "lr": 9.823318295412339e-05, "val_loss": 1.9515904188156128}
{"train_loss": 0.3950028717517853, "global_step": 23496, "epoch": 264, "lr": 9.823303019933474e-05}
{"train_loss": 0.45194384455680847, "global_step": 23497, "epoch": 264, "lr": 9.823287743806175e-05}
{"train_loss": 0.5011149644851685, "global_step": 23498, "epoch": 264, "lr": 9.823272467030443e-05}
{"train_loss": 0.3080155849456787, "global_step": 23499, "epoch": 264, "lr": 9.82325718960628e-05}
{"train_loss": 0.5020220875740051, "global_step": 23500, "epoch": 264, "lr": 9.82324191153369e-05}
{"train_loss": 0.4729248881340027, "global_step": 23501, "epoch": 264, "lr": 9.823226632812674e-05}
{"train_loss": 0.41728755831718445, "global_step": 23502, "epoch": 264, "lr": 9.823211353443233e-05}
{"train_loss": 0.4348473846912384, "global_step": 23503, "epoch": 264, "lr": 9.823196073425371e-05}
{"train_loss": 0.3743877410888672, "global_step": 23504, "epoch": 264, "lr": 9.823180792759088e-05}
{"train_loss": 0.21922822296619415, "global_step": 23505, "epoch": 264, "lr": 9.823165511444388e-05}
{"train_loss": 0.5132283568382263, "global_step": 23506, "epoch": 264, "lr": 9.823150229481273e-05}
{"train_loss": 0.4693724811077118, "global_step": 23507, "epoch": 264, "lr": 9.823134946869743e-05}
{"train_loss": 0.3485609292984009, "global_step": 23508, "epoch": 264, "lr": 9.8231196636098e-05}
{"train_loss": 0.4767885208129883, "global_step": 23509, "epoch": 264, "lr": 9.82310437970145e-05}
{"train_loss": 0.30637502670288086, "global_step": 23510, "epoch": 264, "lr": 9.823089095144692e-05}
{"train_loss": 0.3605353832244873, "global_step": 23511, "epoch": 264, "lr": 9.823073809939527e-05}
{"train_loss": 0.4849279522895813, "global_step": 23512, "epoch": 264, "lr": 9.82305852408596e-05}
{"train_loss": 0.5057716369628906, "global_step": 23513, "epoch": 264, "lr": 9.823043237583991e-05}
{"train_loss": 0.4628366231918335, "global_step": 23514, "epoch": 264, "lr": 9.823027950433622e-05}
{"train_loss": 0.3575914204120636, "global_step": 23515, "epoch": 264, "lr": 9.823012662634857e-05}
{"train_loss": 0.3916829526424408, "global_step": 23516, "epoch": 264, "lr": 9.822997374187695e-05}
{"train_loss": 0.5279508829116821, "global_step": 23517, "epoch": 264, "lr": 9.822982085092141e-05}
{"train_loss": 0.4436313509941101, "global_step": 23518, "epoch": 264, "lr": 9.822966795348195e-05}
{"train_loss": 0.47648242115974426, "global_step": 23519, "epoch": 264, "lr": 9.822951504955861e-05}
{"train_loss": 0.48277029395103455, "global_step": 23520, "epoch": 264, "lr": 9.822936213915139e-05}
{"train_loss": 0.4100855588912964, "global_step": 23521, "epoch": 264, "lr": 9.822920922226032e-05}
{"train_loss": 0.47856855392456055, "global_step": 23522, "epoch": 264, "lr": 9.822905629888542e-05}
{"train_loss": 0.3880615532398224, "global_step": 23523, "epoch": 264, "lr": 9.822890336902673e-05}
{"train_loss": 0.41620680689811707, "global_step": 23524, "epoch": 264, "lr": 9.822875043268422e-05}
{"train_loss": 0.38077133893966675, "global_step": 23525, "epoch": 264, "lr": 9.822859748985796e-05}
{"train_loss": 0.39692968130111694, "global_step": 23526, "epoch": 264, "lr": 9.822844454054796e-05}
{"train_loss": 0.5237569808959961, "global_step": 23527, "epoch": 264, "lr": 9.822829158475423e-05}
{"train_loss": 0.4860037565231323, "global_step": 23528, "epoch": 264, "lr": 9.822813862247679e-05}
{"train_loss": 0.44456741213798523, "global_step": 23529, "epoch": 264, "lr": 9.822798565371566e-05}
{"train_loss": 0.37851428985595703, "global_step": 23530, "epoch": 264, "lr": 9.822783267847087e-05}
{"train_loss": 0.37806373834609985, "global_step": 23531, "epoch": 264, "lr": 9.822767969674243e-05}
{"train_loss": 0.3766830563545227, "global_step": 23532, "epoch": 264, "lr": 9.822752670853037e-05}
{"train_loss": 0.49662676453590393, "global_step": 23533, "epoch": 264, "lr": 9.822737371383472e-05}
{"train_loss": 0.5099381804466248, "global_step": 23534, "epoch": 264, "lr": 9.822722071265546e-05}
{"train_loss": 0.39915013313293457, "global_step": 23535, "epoch": 264, "lr": 9.822706770499266e-05}
{"train_loss": 0.34177935123443604, "global_step": 23536, "epoch": 264, "lr": 9.822691469084632e-05}
{"train_loss": 0.2681020498275757, "global_step": 23537, "epoch": 264, "lr": 9.822676167021645e-05}
{"train_loss": 0.3542918562889099, "global_step": 23538, "epoch": 264, "lr": 9.822660864310308e-05}
{"train_loss": 0.43473273515701294, "global_step": 23539, "epoch": 264, "lr": 9.822645560950624e-05}
{"train_loss": 0.39033007621765137, "global_step": 23540, "epoch": 264, "lr": 9.822630256942593e-05}
{"train_loss": 0.48444807529449463, "global_step": 23541, "epoch": 264, "lr": 9.822614952286218e-05}
{"train_loss": 0.5563127994537354, "global_step": 23542, "epoch": 264, "lr": 9.8225996469815e-05}
{"train_loss": 0.5874088406562805, "global_step": 23543, "epoch": 264, "lr": 9.822584341028445e-05}
{"train_loss": 0.3990756571292877, "global_step": 23544, "epoch": 264, "lr": 9.822569034427052e-05}
{"train_loss": 0.4193364083766937, "global_step": 23545, "epoch": 264, "lr": 9.822553727177322e-05}
{"train_loss": 0.4845609664916992, "global_step": 23546, "epoch": 264, "lr": 9.822538419279259e-05}
{"train_loss": 0.4990851581096649, "global_step": 23547, "epoch": 264, "lr": 9.822523110732864e-05}
{"train_loss": 0.42668667435646057, "global_step": 23548, "epoch": 264, "lr": 9.82250780153814e-05}
{"train_loss": 0.6138159036636353, "global_step": 23549, "epoch": 264, "lr": 9.822492491695088e-05}
{"train_loss": 0.34237876534461975, "global_step": 23550, "epoch": 264, "lr": 9.822477181203712e-05}
{"train_loss": 0.5144226551055908, "global_step": 23551, "epoch": 264, "lr": 9.822461870064013e-05}
{"train_loss": 0.4296184480190277, "global_step": 23552, "epoch": 264, "lr": 9.82244655827599e-05}
{"train_loss": 0.38043922185897827, "global_step": 23553, "epoch": 264, "lr": 9.822431245839649e-05}
{"train_loss": 0.46989554166793823, "global_step": 23554, "epoch": 264, "lr": 9.822415932754992e-05}
{"train_loss": 0.49093183875083923, "global_step": 23555, "epoch": 264, "lr": 9.822400619022021e-05}
{"train_loss": 0.3941129446029663, "global_step": 23556, "epoch": 264, "lr": 9.822385304640734e-05}
{"train_loss": 0.3279729187488556, "global_step": 23557, "epoch": 264, "lr": 9.822369989611137e-05}
{"train_loss": 0.43587881326675415, "global_step": 23558, "epoch": 264, "lr": 9.822354673933233e-05}
{"train_loss": 0.5495644807815552, "global_step": 23559, "epoch": 264, "lr": 9.82233935760702e-05}
{"train_loss": 0.44886231422424316, "global_step": 23560, "epoch": 264, "lr": 9.822324040632504e-05}
{"train_loss": 0.4590323269367218, "global_step": 23561, "epoch": 264, "lr": 9.822308723009683e-05}
{"train_loss": 0.35767602920532227, "global_step": 23562, "epoch": 264, "lr": 9.822293404738564e-05}
{"train_loss": 0.4961116909980774, "global_step": 23563, "epoch": 264, "lr": 9.822278085819146e-05}
{"train_loss": 0.459829181432724, "global_step": 23564, "epoch": 264, "lr": 9.82226276625143e-05}
{"train_loss": 0.6050701141357422, "global_step": 23565, "epoch": 264, "lr": 9.822247446035419e-05}
{"train_loss": 0.616068959236145, "global_step": 23566, "epoch": 264, "lr": 9.822232125171118e-05}
{"train_loss": 0.42295727133750916, "global_step": 23567, "epoch": 264, "lr": 9.822216803658525e-05}
{"train_loss": 0.5532089471817017, "global_step": 23568, "epoch": 264, "lr": 9.822201481497644e-05}
{"train_loss": 0.3525620102882385, "global_step": 23569, "epoch": 264, "lr": 9.822186158688478e-05}
{"train_loss": 0.38698285818099976, "global_step": 23570, "epoch": 264, "lr": 9.822170835231026e-05}
{"train_loss": 0.5635777711868286, "global_step": 23571, "epoch": 264, "lr": 9.822155511125292e-05}
{"train_loss": 0.542382538318634, "global_step": 23572, "epoch": 264, "lr": 9.82214018637128e-05}
{"train_loss": 0.4495016038417816, "global_step": 23573, "epoch": 264, "lr": 9.822124860968989e-05}
{"train_loss": 0.376882404088974, "global_step": 23574, "epoch": 264, "lr": 9.822109534918422e-05}
{"train_loss": 0.562577486038208, "global_step": 23575, "epoch": 264, "lr": 9.822094208219581e-05}
{"train_loss": 0.34879571199417114, "global_step": 23576, "epoch": 264, "lr": 9.822078880872467e-05}
{"train_loss": 0.425393283367157, "global_step": 23577, "epoch": 264, "lr": 9.822063552877085e-05}
{"train_loss": 0.4577825963497162, "global_step": 23578, "epoch": 264, "lr": 9.822048224233435e-05}
{"train_loss": 0.34640735387802124, "global_step": 23579, "epoch": 264, "lr": 9.82203289494152e-05}
{"train_loss": 0.3613417148590088, "global_step": 23580, "epoch": 264, "lr": 9.82201756500134e-05}
{"train_loss": 0.5168047547340393, "global_step": 23581, "epoch": 264, "lr": 9.8220022344129e-05}
{"train_loss": 0.32752323150634766, "global_step": 23582, "epoch": 264, "lr": 9.821986903176199e-05}
{"train_loss": 0.3171590566635132, "global_step": 23583, "epoch": 264, "lr": 9.821971571291241e-05}
{"train_loss": 0.4367837358391687, "global_step": 23584, "epoch": 264, "lr": 9.821956238758027e-05, "val_loss": 1.8736521005630493}
{"train_loss": 0.459885835647583, "global_step": 23585, "epoch": 265, "lr": 9.821940905576561e-05}
{"train_loss": 0.41451308131217957, "global_step": 23586, "epoch": 265, "lr": 9.821925571746844e-05}
{"train_loss": 0.4734571874141693, "global_step": 23587, "epoch": 265, "lr": 9.821910237268878e-05}
{"train_loss": 0.35502520203590393, "global_step": 23588, "epoch": 265, "lr": 9.821894902142664e-05}
{"train_loss": 0.4371636211872101, "global_step": 23589, "epoch": 265, "lr": 9.821879566368206e-05}
{"train_loss": 0.413240522146225, "global_step": 23590, "epoch": 265, "lr": 9.821864229945504e-05}
{"train_loss": 0.362195760011673, "global_step": 23591, "epoch": 265, "lr": 9.821848892874561e-05}
{"train_loss": 0.4703909754753113, "global_step": 23592, "epoch": 265, "lr": 9.82183355515538e-05}
{"train_loss": 0.5493308305740356, "global_step": 23593, "epoch": 265, "lr": 9.821818216787962e-05}
{"train_loss": 0.49943214654922485, "global_step": 23594, "epoch": 265, "lr": 9.82180287777231e-05}
{"train_loss": 0.47950440645217896, "global_step": 23595, "epoch": 265, "lr": 9.821787538108424e-05}
{"train_loss": 0.4269663691520691, "global_step": 23596, "epoch": 265, "lr": 9.821772197796309e-05}
{"train_loss": 0.4181637465953827, "global_step": 23597, "epoch": 265, "lr": 9.821756856835964e-05}
{"train_loss": 0.4073333442211151, "global_step": 23598, "epoch": 265, "lr": 9.821741515227394e-05}
{"train_loss": 0.47100287675857544, "global_step": 23599, "epoch": 265, "lr": 9.821726172970598e-05}
{"train_loss": 0.5067697167396545, "global_step": 23600, "epoch": 265, "lr": 9.821710830065581e-05}
{"train_loss": 0.6104212999343872, "global_step": 23601, "epoch": 265, "lr": 9.821695486512344e-05}
{"train_loss": 0.46256083250045776, "global_step": 23602, "epoch": 265, "lr": 9.821680142310888e-05}
{"train_loss": 0.36633867025375366, "global_step": 23603, "epoch": 265, "lr": 9.821664797461215e-05}
{"train_loss": 0.4316997826099396, "global_step": 23604, "epoch": 265, "lr": 9.821649451963329e-05}
{"train_loss": 0.4665982127189636, "global_step": 23605, "epoch": 265, "lr": 9.821634105817231e-05}
{"train_loss": 0.6703269481658936, "global_step": 23606, "epoch": 265, "lr": 9.821618759022924e-05}
{"train_loss": 0.44522154331207275, "global_step": 23607, "epoch": 265, "lr": 9.821603411580406e-05}
{"train_loss": 0.6523057222366333, "global_step": 23608, "epoch": 265, "lr": 9.821588063489685e-05}
{"train_loss": 0.3917821943759918, "global_step": 23609, "epoch": 265, "lr": 9.821572714750759e-05}
{"train_loss": 0.39081230759620667, "global_step": 23610, "epoch": 265, "lr": 9.821557365363633e-05}
{"train_loss": 0.33681339025497437, "global_step": 23611, "epoch": 265, "lr": 9.821542015328305e-05}
{"train_loss": 0.4677759110927582, "global_step": 23612, "epoch": 265, "lr": 9.821526664644781e-05}
{"train_loss": 0.5310341119766235, "global_step": 23613, "epoch": 265, "lr": 9.821511313313061e-05}
{"train_loss": 0.47749167680740356, "global_step": 23614, "epoch": 265, "lr": 9.821495961333148e-05}
{"train_loss": 0.5018017292022705, "global_step": 23615, "epoch": 265, "lr": 9.821480608705044e-05}
{"train_loss": 0.4875550866127014, "global_step": 23616, "epoch": 265, "lr": 9.821465255428751e-05}
{"train_loss": 0.48306819796562195, "global_step": 23617, "epoch": 265, "lr": 9.82144990150427e-05}
{"train_loss": 0.36462366580963135, "global_step": 23618, "epoch": 265, "lr": 9.821434546931603e-05}
{"train_loss": 0.44970443844795227, "global_step": 23619, "epoch": 265, "lr": 9.821419191710752e-05}
{"train_loss": 0.45248815417289734, "global_step": 23620, "epoch": 265, "lr": 9.821403835841722e-05}
{"train_loss": 0.39147740602493286, "global_step": 23621, "epoch": 265, "lr": 9.821388479324513e-05}
{"train_loss": 0.29611560702323914, "global_step": 23622, "epoch": 265, "lr": 9.821373122159127e-05}
{"train_loss": 0.48929455876350403, "global_step": 23623, "epoch": 265, "lr": 9.821357764345564e-05}
{"train_loss": 0.4085328280925751, "global_step": 23624, "epoch": 265, "lr": 9.821342405883831e-05}
{"train_loss": 0.550830066204071, "global_step": 23625, "epoch": 265, "lr": 9.821327046773926e-05}
{"train_loss": 0.5172907114028931, "global_step": 23626, "epoch": 265, "lr": 9.821311687015853e-05}
{"train_loss": 0.48723092675209045, "global_step": 23627, "epoch": 265, "lr": 9.821296326609612e-05}
{"train_loss": 0.4848943054676056, "global_step": 23628, "epoch": 265, "lr": 9.821280965555209e-05}
{"train_loss": 0.41776174306869507, "global_step": 23629, "epoch": 265, "lr": 9.82126560385264e-05}
{"train_loss": 0.5284155011177063, "global_step": 23630, "epoch": 265, "lr": 9.821250241501913e-05}
{"train_loss": 0.46039679646492004, "global_step": 23631, "epoch": 265, "lr": 9.821234878503028e-05}
{"train_loss": 0.5698235034942627, "global_step": 23632, "epoch": 265, "lr": 9.821219514855986e-05}
{"train_loss": 0.3872714340686798, "global_step": 23633, "epoch": 265, "lr": 9.82120415056079e-05}
{"train_loss": 0.4093235731124878, "global_step": 23634, "epoch": 265, "lr": 9.821188785617442e-05}
{"train_loss": 0.4277324974536896, "global_step": 23635, "epoch": 265, "lr": 9.821173420025943e-05}
{"train_loss": 0.3520696461200714, "global_step": 23636, "epoch": 265, "lr": 9.821158053786297e-05}
{"train_loss": 0.41716060042381287, "global_step": 23637, "epoch": 265, "lr": 9.821142686898505e-05}
{"train_loss": 0.4451562762260437, "global_step": 23638, "epoch": 265, "lr": 9.821127319362568e-05}
{"train_loss": 0.5045390725135803, "global_step": 23639, "epoch": 265, "lr": 9.82111195117849e-05}
{"train_loss": 0.37308284640312195, "global_step": 23640, "epoch": 265, "lr": 9.821096582346272e-05}
{"train_loss": 0.3761650025844574, "global_step": 23641, "epoch": 265, "lr": 9.821081212865917e-05}
{"train_loss": 0.40801241993904114, "global_step": 23642, "epoch": 265, "lr": 9.821065842737426e-05}
{"train_loss": 0.5111264586448669, "global_step": 23643, "epoch": 265, "lr": 9.821050471960802e-05}
{"train_loss": 0.5242008566856384, "global_step": 23644, "epoch": 265, "lr": 9.821035100536045e-05}
{"train_loss": 0.590285062789917, "global_step": 23645, "epoch": 265, "lr": 9.82101972846316e-05}
{"train_loss": 0.3523132801055908, "global_step": 23646, "epoch": 265, "lr": 9.821004355742147e-05}
{"train_loss": 0.42746585607528687, "global_step": 23647, "epoch": 265, "lr": 9.820988982373009e-05}
{"train_loss": 0.47806689143180847, "global_step": 23648, "epoch": 265, "lr": 9.820973608355748e-05}
{"train_loss": 0.5549290180206299, "global_step": 23649, "epoch": 265, "lr": 9.820958233690365e-05}
{"train_loss": 0.45352745056152344, "global_step": 23650, "epoch": 265, "lr": 9.820942858376863e-05}
{"train_loss": 0.45470738410949707, "global_step": 23651, "epoch": 265, "lr": 9.820927482415246e-05}
{"train_loss": 0.46063458919525146, "global_step": 23652, "epoch": 265, "lr": 9.820912105805511e-05}
{"train_loss": 0.40296947956085205, "global_step": 23653, "epoch": 265, "lr": 9.820896728547664e-05}
{"train_loss": 0.4633125960826874, "global_step": 23654, "epoch": 265, "lr": 9.820881350641709e-05}
{"train_loss": 0.4378940165042877, "global_step": 23655, "epoch": 265, "lr": 9.820865972087643e-05}
{"train_loss": 0.4181096851825714, "global_step": 23656, "epoch": 265, "lr": 9.820850592885468e-05}
{"train_loss": 0.3868604600429535, "global_step": 23657, "epoch": 265, "lr": 9.820835213035193e-05}
{"train_loss": 0.4448707699775696, "global_step": 23658, "epoch": 265, "lr": 9.820819832536812e-05}
{"train_loss": 0.5035484433174133, "global_step": 23659, "epoch": 265, "lr": 9.820804451390332e-05}
{"train_loss": 0.3378795087337494, "global_step": 23660, "epoch": 265, "lr": 9.820789069595754e-05}
{"train_loss": 0.421362966299057, "global_step": 23661, "epoch": 265, "lr": 9.82077368715308e-05}
{"train_loss": 0.31707271933555603, "global_step": 23662, "epoch": 265, "lr": 9.82075830406231e-05}
{"train_loss": 0.4597436785697937, "global_step": 23663, "epoch": 265, "lr": 9.82074292032345e-05}
{"train_loss": 0.46592018008232117, "global_step": 23664, "epoch": 265, "lr": 9.820727535936497e-05}
{"train_loss": 0.41012465953826904, "global_step": 23665, "epoch": 265, "lr": 9.820712150901456e-05}
{"train_loss": 0.41720014810562134, "global_step": 23666, "epoch": 265, "lr": 9.820696765218332e-05}
{"train_loss": 0.4779968559741974, "global_step": 23667, "epoch": 265, "lr": 9.820681378887123e-05}
{"train_loss": 0.470061719417572, "global_step": 23668, "epoch": 265, "lr": 9.820665991907831e-05}
{"train_loss": 0.3848447799682617, "global_step": 23669, "epoch": 265, "lr": 9.820650604280459e-05}
{"train_loss": 0.42441630363464355, "global_step": 23670, "epoch": 265, "lr": 9.82063521600501e-05}
{"train_loss": 0.41794657707214355, "global_step": 23671, "epoch": 265, "lr": 9.820619827081486e-05}
{"train_loss": 0.4968101978302002, "global_step": 23672, "epoch": 265, "lr": 9.820604437509888e-05}
{"train_loss": 0.45071111368329336, "global_step": 23673, "epoch": 265, "lr": 9.820589047290219e-05, "val_loss": 1.9941703081130981, "train_action_mse_error": 27.36053466796875}
{"train_loss": 0.289673775434494, "global_step": 23674, "epoch": 266, "lr": 9.82057365642248e-05}
{"train_loss": 0.40220701694488525, "global_step": 23675, "epoch": 266, "lr": 9.820558264906674e-05}
{"train_loss": 0.3761150538921356, "global_step": 23676, "epoch": 266, "lr": 9.820542872742802e-05}
{"train_loss": 0.37999796867370605, "global_step": 23677, "epoch": 266, "lr": 9.820527479930866e-05}
{"train_loss": 0.301472932100296, "global_step": 23678, "epoch": 266, "lr": 9.82051208647087e-05}
{"train_loss": 0.31674355268478394, "global_step": 23679, "epoch": 266, "lr": 9.820496692362814e-05}
{"train_loss": 0.5096818804740906, "global_step": 23680, "epoch": 266, "lr": 9.820481297606702e-05}
{"train_loss": 0.4098163843154907, "global_step": 23681, "epoch": 266, "lr": 9.820465902202536e-05}
{"train_loss": 0.5085587501525879, "global_step": 23682, "epoch": 266, "lr": 9.820450506150316e-05}
{"train_loss": 0.31260260939598083, "global_step": 23683, "epoch": 266, "lr": 9.820435109450046e-05}
{"train_loss": 0.4339298605918884, "global_step": 23684, "epoch": 266, "lr": 9.820419712101725e-05}
{"train_loss": 0.3916001319885254, "global_step": 23685, "epoch": 266, "lr": 9.820404314105359e-05}
{"train_loss": 0.35663890838623047, "global_step": 23686, "epoch": 266, "lr": 9.820388915460947e-05}
{"train_loss": 0.3072599470615387, "global_step": 23687, "epoch": 266, "lr": 9.820373516168495e-05}
{"train_loss": 0.3967629671096802, "global_step": 23688, "epoch": 266, "lr": 9.820358116228001e-05}
{"train_loss": 0.5021649599075317, "global_step": 23689, "epoch": 266, "lr": 9.820342715639468e-05}
{"train_loss": 0.47734683752059937, "global_step": 23690, "epoch": 266, "lr": 9.8203273144029e-05}
{"train_loss": 0.49826881289482117, "global_step": 23691, "epoch": 266, "lr": 9.820311912518297e-05}
{"train_loss": 0.37992051243782043, "global_step": 23692, "epoch": 266, "lr": 9.820296509985663e-05}
{"train_loss": 0.340096116065979, "global_step": 23693, "epoch": 266, "lr": 9.820281106804995e-05}
{"train_loss": 0.36647146940231323, "global_step": 23694, "epoch": 266, "lr": 9.820265702976304e-05}
{"train_loss": 0.5198537111282349, "global_step": 23695, "epoch": 266, "lr": 9.820250298499585e-05}
{"train_loss": 0.5159555673599243, "global_step": 23696, "epoch": 266, "lr": 9.820234893374841e-05}
{"train_loss": 0.40621691942214966, "global_step": 23697, "epoch": 266, "lr": 9.820219487602076e-05}
{"train_loss": 0.4641173779964447, "global_step": 23698, "epoch": 266, "lr": 9.82020408118129e-05}
{"train_loss": 0.3500845432281494, "global_step": 23699, "epoch": 266, "lr": 9.82018867411249e-05}
{"train_loss": 0.5235127806663513, "global_step": 23700, "epoch": 266, "lr": 9.820173266395672e-05}
{"train_loss": 0.4778626561164856, "global_step": 23701, "epoch": 266, "lr": 9.82015785803084e-05}
{"train_loss": 0.3849024176597595, "global_step": 23702, "epoch": 266, "lr": 9.820142449017996e-05}
{"train_loss": 0.4520433843135834, "global_step": 23703, "epoch": 266, "lr": 9.820127039357143e-05}
{"train_loss": 0.38160744309425354, "global_step": 23704, "epoch": 266, "lr": 9.820111629048284e-05}
{"train_loss": 0.4489381015300751, "global_step": 23705, "epoch": 266, "lr": 9.82009621809142e-05}
{"train_loss": 0.40286508202552795, "global_step": 23706, "epoch": 266, "lr": 9.820080806486551e-05}
{"train_loss": 0.3627261519432068, "global_step": 23707, "epoch": 266, "lr": 9.820065394233682e-05}
{"train_loss": 0.3895738124847412, "global_step": 23708, "epoch": 266, "lr": 9.820049981332812e-05}
{"train_loss": 0.4579320549964905, "global_step": 23709, "epoch": 266, "lr": 9.820034567783948e-05}
{"train_loss": 0.45265570282936096, "global_step": 23710, "epoch": 266, "lr": 9.820019153587088e-05}
{"train_loss": 0.36202794313430786, "global_step": 23711, "epoch": 266, "lr": 9.820003738742234e-05}
{"train_loss": 0.44881996512413025, "global_step": 23712, "epoch": 266, "lr": 9.819988323249391e-05}
{"train_loss": 0.5747509598731995, "global_step": 23713, "epoch": 266, "lr": 9.819972907108558e-05}
{"train_loss": 0.44079136848449707, "global_step": 23714, "epoch": 266, "lr": 9.81995749031974e-05}
{"train_loss": 0.5194373726844788, "global_step": 23715, "epoch": 266, "lr": 9.819942072882935e-05}
{"train_loss": 0.3586651086807251, "global_step": 23716, "epoch": 266, "lr": 9.819926654798148e-05}
{"train_loss": 0.4020358920097351, "global_step": 23717, "epoch": 266, "lr": 9.819911236065382e-05}
{"train_loss": 0.4208149313926697, "global_step": 23718, "epoch": 266, "lr": 9.819895816684637e-05}
{"train_loss": 0.4759827256202698, "global_step": 23719, "epoch": 266, "lr": 9.819880396655916e-05}
{"train_loss": 0.312256783246994, "global_step": 23720, "epoch": 266, "lr": 9.81986497597922e-05}
{"train_loss": 0.4575684666633606, "global_step": 23721, "epoch": 266, "lr": 9.819849554654554e-05}
{"train_loss": 0.41921666264533997, "global_step": 23722, "epoch": 266, "lr": 9.819834132681917e-05}
{"train_loss": 0.3486310839653015, "global_step": 23723, "epoch": 266, "lr": 9.819818710061312e-05}
{"train_loss": 0.5065748691558838, "global_step": 23724, "epoch": 266, "lr": 9.81980328679274e-05}
{"train_loss": 0.433432400226593, "global_step": 23725, "epoch": 266, "lr": 9.819787862876205e-05}
{"train_loss": 0.4107484221458435, "global_step": 23726, "epoch": 266, "lr": 9.819772438311708e-05}
{"train_loss": 0.4329017698764801, "global_step": 23727, "epoch": 266, "lr": 9.819757013099251e-05}
{"train_loss": 0.3638899326324463, "global_step": 23728, "epoch": 266, "lr": 9.819741587238837e-05}
{"train_loss": 0.3799053132534027, "global_step": 23729, "epoch": 266, "lr": 9.819726160730469e-05}
{"train_loss": 0.46496692299842834, "global_step": 23730, "epoch": 266, "lr": 9.819710733574146e-05}
{"train_loss": 0.46370476484298706, "global_step": 23731, "epoch": 266, "lr": 9.819695305769871e-05}
{"train_loss": 0.41467010974884033, "global_step": 23732, "epoch": 266, "lr": 9.819679877317647e-05}
{"train_loss": 0.35222846269607544, "global_step": 23733, "epoch": 266, "lr": 9.819664448217476e-05}
{"train_loss": 0.6544582843780518, "global_step": 23734, "epoch": 266, "lr": 9.819649018469361e-05}
{"train_loss": 0.48649361729621887, "global_step": 23735, "epoch": 266, "lr": 9.819633588073301e-05}
{"train_loss": 0.37300509214401245, "global_step": 23736, "epoch": 266, "lr": 9.819618157029303e-05}
{"train_loss": 0.5048775672912598, "global_step": 23737, "epoch": 266, "lr": 9.819602725337362e-05}
{"train_loss": 0.5071366429328918, "global_step": 23738, "epoch": 266, "lr": 9.819587292997488e-05}
{"train_loss": 0.4854196608066559, "global_step": 23739, "epoch": 266, "lr": 9.819571860009676e-05}
{"train_loss": 0.4144178032875061, "global_step": 23740, "epoch": 266, "lr": 9.819556426373932e-05}
{"train_loss": 0.3963553309440613, "global_step": 23741, "epoch": 266, "lr": 9.819540992090259e-05}
{"train_loss": 0.5822494626045227, "global_step": 23742, "epoch": 266, "lr": 9.819525557158658e-05}
{"train_loss": 0.42417019605636597, "global_step": 23743, "epoch": 266, "lr": 9.819510121579127e-05}
{"train_loss": 0.5464202165603638, "global_step": 23744, "epoch": 266, "lr": 9.819494685351675e-05}
{"train_loss": 0.4047916531562805, "global_step": 23745, "epoch": 266, "lr": 9.819479248476298e-05}
{"train_loss": 0.3806283473968506, "global_step": 23746, "epoch": 266, "lr": 9.819463810953002e-05}
{"train_loss": 0.5046631693840027, "global_step": 23747, "epoch": 266, "lr": 9.81944837278179e-05}
{"train_loss": 0.4306015074253082, "global_step": 23748, "epoch": 266, "lr": 9.819432933962659e-05}
{"train_loss": 0.39155513048171997, "global_step": 23749, "epoch": 266, "lr": 9.819417494495615e-05}
{"train_loss": 0.3212716281414032, "global_step": 23750, "epoch": 266, "lr": 9.819402054380658e-05}
{"train_loss": 0.3902061879634857, "global_step": 23751, "epoch": 266, "lr": 9.819386613617793e-05}
{"train_loss": 0.5469042658805847, "global_step": 23752, "epoch": 266, "lr": 9.81937117220702e-05}
{"train_loss": 0.44998183846473694, "global_step": 23753, "epoch": 266, "lr": 9.819355730148339e-05}
{"train_loss": 0.42946961522102356, "global_step": 23754, "epoch": 266, "lr": 9.819340287441756e-05}
{"train_loss": 0.46816718578338623, "global_step": 23755, "epoch": 266, "lr": 9.819324844087272e-05}
{"train_loss": 0.7057093381881714, "global_step": 23756, "epoch": 266, "lr": 9.819309400084888e-05}
{"train_loss": 0.4594660699367523, "global_step": 23757, "epoch": 266, "lr": 9.819293955434607e-05}
{"train_loss": 0.47277095913887024, "global_step": 23758, "epoch": 266, "lr": 9.819278510136429e-05}
{"train_loss": 0.39260509610176086, "global_step": 23759, "epoch": 266, "lr": 9.819263064190359e-05}
{"train_loss": 0.37373271584510803, "global_step": 23760, "epoch": 266, "lr": 9.819247617596399e-05}
{"train_loss": 0.4354703724384308, "global_step": 23761, "epoch": 266, "lr": 9.819232170354547e-05}
{"train_loss": 0.43101650141598136, "global_step": 23762, "epoch": 266, "lr": 9.819216722464811e-05, "val_loss": 1.957098364830017}
{"train_loss": 0.5152596235275269, "global_step": 23763, "epoch": 267, "lr": 9.819201273927187e-05}
{"train_loss": 0.3607538342475891, "global_step": 23764, "epoch": 267, "lr": 9.819185824741683e-05}
{"train_loss": 0.41866496205329895, "global_step": 23765, "epoch": 267, "lr": 9.819170374908297e-05}
{"train_loss": 0.48489701747894287, "global_step": 23766, "epoch": 267, "lr": 9.819154924427033e-05}
{"train_loss": 0.4291552007198334, "global_step": 23767, "epoch": 267, "lr": 9.81913947329789e-05}
{"train_loss": 0.3962312638759613, "global_step": 23768, "epoch": 267, "lr": 9.819124021520874e-05}
{"train_loss": 0.3551887273788452, "global_step": 23769, "epoch": 267, "lr": 9.819108569095984e-05}
{"train_loss": 0.3258027136325836, "global_step": 23770, "epoch": 267, "lr": 9.819093116023227e-05}
{"train_loss": 0.5251544713973999, "global_step": 23771, "epoch": 267, "lr": 9.8190776623026e-05}
{"train_loss": 0.5541661977767944, "global_step": 23772, "epoch": 267, "lr": 9.819062207934106e-05}
{"train_loss": 0.4355359375476837, "global_step": 23773, "epoch": 267, "lr": 9.819046752917749e-05}
{"train_loss": 0.38306379318237305, "global_step": 23774, "epoch": 267, "lr": 9.819031297253529e-05}
{"train_loss": 0.30979007482528687, "global_step": 23775, "epoch": 267, "lr": 9.819015840941449e-05}
{"train_loss": 0.3682604432106018, "global_step": 23776, "epoch": 267, "lr": 9.819000383981511e-05}
{"train_loss": 0.4427485466003418, "global_step": 23777, "epoch": 267, "lr": 9.818984926373718e-05}
{"train_loss": 0.3841182291507721, "global_step": 23778, "epoch": 267, "lr": 9.818969468118071e-05}
{"train_loss": 0.4471876621246338, "global_step": 23779, "epoch": 267, "lr": 9.818954009214572e-05}
{"train_loss": 0.42685940861701965, "global_step": 23780, "epoch": 267, "lr": 9.818938549663222e-05}
{"train_loss": 0.39405548572540283, "global_step": 23781, "epoch": 267, "lr": 9.818923089464026e-05}
{"train_loss": 0.46539732813835144, "global_step": 23782, "epoch": 267, "lr": 9.818907628616985e-05}
{"train_loss": 0.5064368844032288, "global_step": 23783, "epoch": 267, "lr": 9.8188921671221e-05}
{"train_loss": 0.554923951625824, "global_step": 23784, "epoch": 267, "lr": 9.818876704979373e-05}
{"train_loss": 0.5145010352134705, "global_step": 23785, "epoch": 267, "lr": 9.818861242188807e-05}
{"train_loss": 0.5397776365280151, "global_step": 23786, "epoch": 267, "lr": 9.818845778750405e-05}
{"train_loss": 0.33251920342445374, "global_step": 23787, "epoch": 267, "lr": 9.818830314664168e-05}
{"train_loss": 0.4672793447971344, "global_step": 23788, "epoch": 267, "lr": 9.818814849930095e-05}
{"train_loss": 0.3907330334186554, "global_step": 23789, "epoch": 267, "lr": 9.818799384548195e-05}
{"train_loss": 0.39929184317588806, "global_step": 23790, "epoch": 267, "lr": 9.818783918518464e-05}
{"train_loss": 0.341827929019928, "global_step": 23791, "epoch": 267, "lr": 9.818768451840905e-05}
{"train_loss": 0.4108665883541107, "global_step": 23792, "epoch": 267, "lr": 9.818752984515523e-05}
{"train_loss": 0.37073132395744324, "global_step": 23793, "epoch": 267, "lr": 9.818737516542318e-05}
{"train_loss": 0.4112003445625305, "global_step": 23794, "epoch": 267, "lr": 9.818722047921294e-05}
{"train_loss": 0.4154164493083954, "global_step": 23795, "epoch": 267, "lr": 9.81870657865245e-05}
{"train_loss": 0.4666551351547241, "global_step": 23796, "epoch": 267, "lr": 9.818691108735789e-05}
{"train_loss": 0.6027491092681885, "global_step": 23797, "epoch": 267, "lr": 9.818675638171315e-05}
{"train_loss": 0.4198762774467468, "global_step": 23798, "epoch": 267, "lr": 9.818660166959029e-05}
{"train_loss": 0.3257492482662201, "global_step": 23799, "epoch": 267, "lr": 9.818644695098932e-05}
{"train_loss": 0.5652183890342712, "global_step": 23800, "epoch": 267, "lr": 9.818629222591027e-05}
{"train_loss": 0.42418140172958374, "global_step": 23801, "epoch": 267, "lr": 9.818613749435317e-05}
{"train_loss": 0.3609800934791565, "global_step": 23802, "epoch": 267, "lr": 9.818598275631801e-05}
{"train_loss": 0.4067555367946625, "global_step": 23803, "epoch": 267, "lr": 9.818582801180484e-05}
{"train_loss": 0.37813183665275574, "global_step": 23804, "epoch": 267, "lr": 9.81856732608137e-05}
{"train_loss": 0.3918328881263733, "global_step": 23805, "epoch": 267, "lr": 9.818551850334456e-05}
{"train_loss": 0.41783949732780457, "global_step": 23806, "epoch": 267, "lr": 9.818536373939746e-05}
{"train_loss": 0.37617939710617065, "global_step": 23807, "epoch": 267, "lr": 9.818520896897243e-05}
{"train_loss": 0.36162394285202026, "global_step": 23808, "epoch": 267, "lr": 9.818505419206949e-05}
{"train_loss": 0.40240898728370667, "global_step": 23809, "epoch": 267, "lr": 9.818489940868865e-05}
{"train_loss": 0.4301445186138153, "global_step": 23810, "epoch": 267, "lr": 9.818474461882994e-05}
{"train_loss": 0.4324774742126465, "global_step": 23811, "epoch": 267, "lr": 9.818458982249338e-05}
{"train_loss": 0.4582422077655792, "global_step": 23812, "epoch": 267, "lr": 9.8184435019679e-05}
{"train_loss": 0.3200646936893463, "global_step": 23813, "epoch": 267, "lr": 9.818428021038679e-05}
{"train_loss": 0.3626581132411957, "global_step": 23814, "epoch": 267, "lr": 9.81841253946168e-05}
{"train_loss": 0.3455447256565094, "global_step": 23815, "epoch": 267, "lr": 9.818397057236905e-05}
{"train_loss": 0.5372707843780518, "global_step": 23816, "epoch": 267, "lr": 9.818381574364354e-05}
{"train_loss": 0.37342116236686707, "global_step": 23817, "epoch": 267, "lr": 9.818366090844031e-05}
{"train_loss": 0.5663905739784241, "global_step": 23818, "epoch": 267, "lr": 9.818350606675937e-05}
{"train_loss": 0.4065213203430176, "global_step": 23819, "epoch": 267, "lr": 9.818335121860074e-05}
{"train_loss": 0.5994870662689209, "global_step": 23820, "epoch": 267, "lr": 9.818319636396444e-05}
{"train_loss": 0.3717312514781952, "global_step": 23821, "epoch": 267, "lr": 9.818304150285052e-05}
{"train_loss": 0.430620402097702, "global_step": 23822, "epoch": 267, "lr": 9.818288663525896e-05}
{"train_loss": 0.4720856249332428, "global_step": 23823, "epoch": 267, "lr": 9.818273176118981e-05}
{"train_loss": 0.4418739378452301, "global_step": 23824, "epoch": 267, "lr": 9.818257688064307e-05}
{"train_loss": 0.3818127512931824, "global_step": 23825, "epoch": 267, "lr": 9.818242199361878e-05}
{"train_loss": 0.4922170042991638, "global_step": 23826, "epoch": 267, "lr": 9.818226710011694e-05}
{"train_loss": 0.45874977111816406, "global_step": 23827, "epoch": 267, "lr": 9.818211220013758e-05}
{"train_loss": 0.4215797483921051, "global_step": 23828, "epoch": 267, "lr": 9.818195729368074e-05}
{"train_loss": 0.47177553176879883, "global_step": 23829, "epoch": 267, "lr": 9.81818023807464e-05}
{"train_loss": 0.3282279372215271, "global_step": 23830, "epoch": 267, "lr": 9.818164746133463e-05}
{"train_loss": 0.3922845125198364, "global_step": 23831, "epoch": 267, "lr": 9.81814925354454e-05}
{"train_loss": 0.31073638796806335, "global_step": 23832, "epoch": 267, "lr": 9.818133760307875e-05}
{"train_loss": 0.49808603525161743, "global_step": 23833, "epoch": 267, "lr": 9.818118266423473e-05}
{"train_loss": 0.43351349234580994, "global_step": 23834, "epoch": 267, "lr": 9.818102771891333e-05}
{"train_loss": 0.458821177482605, "global_step": 23835, "epoch": 267, "lr": 9.818087276711457e-05}
{"train_loss": 0.5459878444671631, "global_step": 23836, "epoch": 267, "lr": 9.818071780883849e-05}
{"train_loss": 0.5050939917564392, "global_step": 23837, "epoch": 267, "lr": 9.81805628440851e-05}
{"train_loss": 0.5144365429878235, "global_step": 23838, "epoch": 267, "lr": 9.818040787285441e-05}
{"train_loss": 0.43566107749938965, "global_step": 23839, "epoch": 267, "lr": 9.818025289514645e-05}
{"train_loss": 0.46889016032218933, "global_step": 23840, "epoch": 267, "lr": 9.818009791096125e-05}
{"train_loss": 0.3367403447628021, "global_step": 23841, "epoch": 267, "lr": 9.817994292029883e-05}
{"train_loss": 0.3665808141231537, "global_step": 23842, "epoch": 267, "lr": 9.81797879231592e-05}
{"train_loss": 0.4757827818393707, "global_step": 23843, "epoch": 267, "lr": 9.817963291954237e-05}
{"train_loss": 0.44322317838668823, "global_step": 23844, "epoch": 267, "lr": 9.81794779094484e-05}
{"train_loss": 0.5769273042678833, "global_step": 23845, "epoch": 267, "lr": 9.817932289287725e-05}
{"train_loss": 0.3414628505706787, "global_step": 23846, "epoch": 267, "lr": 9.817916786982901e-05}
{"train_loss": 0.33108624815940857, "global_step": 23847, "epoch": 267, "lr": 9.817901284030366e-05}
{"train_loss": 0.38263094425201416, "global_step": 23848, "epoch": 267, "lr": 9.817885780430123e-05}
{"train_loss": 0.4300940930843353, "global_step": 23849, "epoch": 267, "lr": 9.817870276182174e-05}
{"train_loss": 0.5469480752944946, "global_step": 23850, "epoch": 267, "lr": 9.817854771286521e-05}
{"train_loss": 0.43135326263609897, "global_step": 23851, "epoch": 267, "lr": 9.817839265743166e-05, "val_loss": 1.909077525138855}
{"train_loss": 0.39720940589904785, "global_step": 23852, "epoch": 268, "lr": 9.81782375955211e-05}
{"train_loss": 0.34553566575050354, "global_step": 23853, "epoch": 268, "lr": 9.817808252713359e-05}
{"train_loss": 0.41139519214630127, "global_step": 23854, "epoch": 268, "lr": 9.81779274522691e-05}
{"train_loss": 0.4723888337612152, "global_step": 23855, "epoch": 268, "lr": 9.817777237092768e-05}
{"train_loss": 0.46307384967803955, "global_step": 23856, "epoch": 268, "lr": 9.817761728310934e-05}
{"train_loss": 0.46645939350128174, "global_step": 23857, "epoch": 268, "lr": 9.817746218881412e-05}
{"train_loss": 0.42456865310668945, "global_step": 23858, "epoch": 268, "lr": 9.817730708804203e-05}
{"train_loss": 0.44918861985206604, "global_step": 23859, "epoch": 268, "lr": 9.817715198079308e-05}
{"train_loss": 0.39188477396965027, "global_step": 23860, "epoch": 268, "lr": 9.817699686706729e-05}
{"train_loss": 0.40614721179008484, "global_step": 23861, "epoch": 268, "lr": 9.81768417468647e-05}
{"train_loss": 0.38007253408432007, "global_step": 23862, "epoch": 268, "lr": 9.817668662018533e-05}
{"train_loss": 0.3635467290878296, "global_step": 23863, "epoch": 268, "lr": 9.81765314870292e-05}
{"train_loss": 0.45877259969711304, "global_step": 23864, "epoch": 268, "lr": 9.81763763473963e-05}
{"train_loss": 0.3689365088939667, "global_step": 23865, "epoch": 268, "lr": 9.817622120128666e-05}
{"train_loss": 0.5836494565010071, "global_step": 23866, "epoch": 268, "lr": 9.817606604870033e-05}
{"train_loss": 0.46523308753967285, "global_step": 23867, "epoch": 268, "lr": 9.817591088963733e-05}
{"train_loss": 0.3106277287006378, "global_step": 23868, "epoch": 268, "lr": 9.817575572409765e-05}
{"train_loss": 0.34885120391845703, "global_step": 23869, "epoch": 268, "lr": 9.817560055208133e-05}
{"train_loss": 0.4575517177581787, "global_step": 23870, "epoch": 268, "lr": 9.817544537358839e-05}
{"train_loss": 0.49299076199531555, "global_step": 23871, "epoch": 268, "lr": 9.817529018861885e-05}
{"train_loss": 0.4145923852920532, "global_step": 23872, "epoch": 268, "lr": 9.817513499717273e-05}
{"train_loss": 0.3746074140071869, "global_step": 23873, "epoch": 268, "lr": 9.817497979925005e-05}
{"train_loss": 0.39657095074653625, "global_step": 23874, "epoch": 268, "lr": 9.817482459485083e-05}
{"train_loss": 0.3984093964099884, "global_step": 23875, "epoch": 268, "lr": 9.817466938397509e-05}
{"train_loss": 0.36264216899871826, "global_step": 23876, "epoch": 268, "lr": 9.817451416662284e-05}
{"train_loss": 0.3875260353088379, "global_step": 23877, "epoch": 268, "lr": 9.817435894279413e-05}
{"train_loss": 0.36419638991355896, "global_step": 23878, "epoch": 268, "lr": 9.817420371248896e-05}
{"train_loss": 0.3964095413684845, "global_step": 23879, "epoch": 268, "lr": 9.817404847570736e-05}
{"train_loss": 0.44500958919525146, "global_step": 23880, "epoch": 268, "lr": 9.817389323244935e-05}
{"train_loss": 0.4945666193962097, "global_step": 23881, "epoch": 268, "lr": 9.817373798271493e-05}
{"train_loss": 0.3832525610923767, "global_step": 23882, "epoch": 268, "lr": 9.817358272650415e-05}
{"train_loss": 0.33811625838279724, "global_step": 23883, "epoch": 268, "lr": 9.817342746381701e-05}
{"train_loss": 0.4733913838863373, "global_step": 23884, "epoch": 268, "lr": 9.817327219465356e-05}
{"train_loss": 0.5630500316619873, "global_step": 23885, "epoch": 268, "lr": 9.817311691901379e-05}
{"train_loss": 0.3938480019569397, "global_step": 23886, "epoch": 268, "lr": 9.817296163689772e-05}
{"train_loss": 0.3207992911338806, "global_step": 23887, "epoch": 268, "lr": 9.81728063483054e-05}
{"train_loss": 0.4844053089618683, "global_step": 23888, "epoch": 268, "lr": 9.81726510532368e-05}
{"train_loss": 0.3070959448814392, "global_step": 23889, "epoch": 268, "lr": 9.817249575169201e-05}
{"train_loss": 0.5309131741523743, "global_step": 23890, "epoch": 268, "lr": 9.817234044367101e-05}
{"train_loss": 0.33513760566711426, "global_step": 23891, "epoch": 268, "lr": 9.817218512917383e-05}
{"train_loss": 0.41956406831741333, "global_step": 23892, "epoch": 268, "lr": 9.817202980820046e-05}
{"train_loss": 0.3578994572162628, "global_step": 23893, "epoch": 268, "lr": 9.817187448075097e-05}
{"train_loss": 0.35923999547958374, "global_step": 23894, "epoch": 268, "lr": 9.817171914682536e-05}
{"train_loss": 0.4429801106452942, "global_step": 23895, "epoch": 268, "lr": 9.817156380642363e-05}
{"train_loss": 0.5220880508422852, "global_step": 23896, "epoch": 268, "lr": 9.817140845954584e-05}
{"train_loss": 0.33805492520332336, "global_step": 23897, "epoch": 268, "lr": 9.817125310619199e-05}
{"train_loss": 0.4108949899673462, "global_step": 23898, "epoch": 268, "lr": 9.817109774636209e-05}
{"train_loss": 0.40064457058906555, "global_step": 23899, "epoch": 268, "lr": 9.817094238005618e-05}
{"train_loss": 0.37872710824012756, "global_step": 23900, "epoch": 268, "lr": 9.817078700727428e-05}
{"train_loss": 0.5232089161872864, "global_step": 23901, "epoch": 268, "lr": 9.81706316280164e-05}
{"train_loss": 0.36839592456817627, "global_step": 23902, "epoch": 268, "lr": 9.817047624228257e-05}
{"train_loss": 0.4177100360393524, "global_step": 23903, "epoch": 268, "lr": 9.817032085007278e-05}
{"train_loss": 0.4269610047340393, "global_step": 23904, "epoch": 268, "lr": 9.81701654513871e-05}
{"train_loss": 0.3940065801143646, "global_step": 23905, "epoch": 268, "lr": 9.817001004622553e-05}
{"train_loss": 0.33644556999206543, "global_step": 23906, "epoch": 268, "lr": 9.81698546345881e-05}
{"train_loss": 0.38127052783966064, "global_step": 23907, "epoch": 268, "lr": 9.81696992164748e-05}
{"train_loss": 0.38182634115219116, "global_step": 23908, "epoch": 268, "lr": 9.816954379188568e-05}
{"train_loss": 0.34528449177742004, "global_step": 23909, "epoch": 268, "lr": 9.816938836082074e-05}
{"train_loss": 0.5252763032913208, "global_step": 23910, "epoch": 268, "lr": 9.816923292328002e-05}
{"train_loss": 0.3247984051704407, "global_step": 23911, "epoch": 268, "lr": 9.816907747926354e-05}
{"train_loss": 0.506220281124115, "global_step": 23912, "epoch": 268, "lr": 9.81689220287713e-05}
{"train_loss": 0.49235257506370544, "global_step": 23913, "epoch": 268, "lr": 9.816876657180336e-05}
{"train_loss": 0.5060334801673889, "global_step": 23914, "epoch": 268, "lr": 9.81686111083597e-05}
{"train_loss": 0.3639933466911316, "global_step": 23915, "epoch": 268, "lr": 9.816845563844036e-05}
{"train_loss": 0.5193807482719421, "global_step": 23916, "epoch": 268, "lr": 9.816830016204535e-05}
{"train_loss": 0.47297802567481995, "global_step": 23917, "epoch": 268, "lr": 9.816814467917471e-05}
{"train_loss": 0.4051803946495056, "global_step": 23918, "epoch": 268, "lr": 9.816798918982846e-05}
{"train_loss": 0.5076096653938293, "global_step": 23919, "epoch": 268, "lr": 9.81678336940066e-05}
{"train_loss": 0.3298225998878479, "global_step": 23920, "epoch": 268, "lr": 9.816767819170916e-05}
{"train_loss": 0.42104849219322205, "global_step": 23921, "epoch": 268, "lr": 9.816752268293616e-05}
{"train_loss": 0.47322961688041687, "global_step": 23922, "epoch": 268, "lr": 9.816736716768762e-05}
{"train_loss": 0.4014635682106018, "global_step": 23923, "epoch": 268, "lr": 9.816721164596357e-05}
{"train_loss": 0.32998302578926086, "global_step": 23924, "epoch": 268, "lr": 9.816705611776402e-05}
{"train_loss": 0.3792288303375244, "global_step": 23925, "epoch": 268, "lr": 9.8166900583089e-05}
{"train_loss": 0.5323843359947205, "global_step": 23926, "epoch": 268, "lr": 9.816674504193854e-05}
{"train_loss": 0.40788692235946655, "global_step": 23927, "epoch": 268, "lr": 9.816658949431264e-05}
{"train_loss": 0.4620797336101532, "global_step": 23928, "epoch": 268, "lr": 9.816643394021133e-05}
{"train_loss": 0.40218329429626465, "global_step": 23929, "epoch": 268, "lr": 9.816627837963462e-05}
{"train_loss": 0.49786487221717834, "global_step": 23930, "epoch": 268, "lr": 9.816612281258255e-05}
{"train_loss": 0.3890770673751831, "global_step": 23931, "epoch": 268, "lr": 9.816596723905513e-05}
{"train_loss": 0.40108463168144226, "global_step": 23932, "epoch": 268, "lr": 9.816581165905238e-05}
{"train_loss": 0.4700745642185211, "global_step": 23933, "epoch": 268, "lr": 9.816565607257433e-05}
{"train_loss": 0.5054088830947876, "global_step": 23934, "epoch": 268, "lr": 9.816550047962098e-05}
{"train_loss": 0.3733806908130646, "global_step": 23935, "epoch": 268, "lr": 9.816534488019238e-05}
{"train_loss": 0.3496943712234497, "global_step": 23936, "epoch": 268, "lr": 9.816518927428854e-05}
{"train_loss": 0.4902816116809845, "global_step": 23937, "epoch": 268, "lr": 9.816503366190946e-05}
{"train_loss": 0.3402363359928131, "global_step": 23938, "epoch": 268, "lr": 9.816487804305519e-05}
{"train_loss": 0.3923935890197754, "global_step": 23939, "epoch": 268, "lr": 9.816472241772574e-05}
{"train_loss": 0.41999145605590904, "global_step": 23940, "epoch": 268, "lr": 9.816456678592113e-05, "val_loss": 1.9841469526290894}
{"train_loss": 0.44590070843696594, "global_step": 23941, "epoch": 269, "lr": 9.816441114764138e-05}
{"train_loss": 0.45348939299583435, "global_step": 23942, "epoch": 269, "lr": 9.81642555028865e-05}
{"train_loss": 0.4559885263442993, "global_step": 23943, "epoch": 269, "lr": 9.816409985165654e-05}
{"train_loss": 0.4642152488231659, "global_step": 23944, "epoch": 269, "lr": 9.81639441939515e-05}
{"train_loss": 0.42622825503349304, "global_step": 23945, "epoch": 269, "lr": 9.81637885297714e-05}
{"train_loss": 0.4429336190223694, "global_step": 23946, "epoch": 269, "lr": 9.816363285911627e-05}
{"train_loss": 0.30607154965400696, "global_step": 23947, "epoch": 269, "lr": 9.816347718198613e-05}
{"train_loss": 0.39465904235839844, "global_step": 23948, "epoch": 269, "lr": 9.816332149838099e-05}
{"train_loss": 0.389749675989151, "global_step": 23949, "epoch": 269, "lr": 9.816316580830087e-05}
{"train_loss": 0.5559181571006775, "global_step": 23950, "epoch": 269, "lr": 9.816301011174582e-05}
{"train_loss": 0.4528222382068634, "global_step": 23951, "epoch": 269, "lr": 9.816285440871583e-05}
{"train_loss": 0.4350212514400482, "global_step": 23952, "epoch": 269, "lr": 9.816269869921093e-05}
{"train_loss": 0.4271819591522217, "global_step": 23953, "epoch": 269, "lr": 9.816254298323114e-05}
{"train_loss": 0.470339298248291, "global_step": 23954, "epoch": 269, "lr": 9.816238726077649e-05}
{"train_loss": 0.41408464312553406, "global_step": 23955, "epoch": 269, "lr": 9.8162231531847e-05}
{"train_loss": 0.43399107456207275, "global_step": 23956, "epoch": 269, "lr": 9.816207579644268e-05}
{"train_loss": 0.4860459566116333, "global_step": 23957, "epoch": 269, "lr": 9.816192005456354e-05}
{"train_loss": 0.3589051067829132, "global_step": 23958, "epoch": 269, "lr": 9.816176430620963e-05}
{"train_loss": 0.42246443033218384, "global_step": 23959, "epoch": 269, "lr": 9.816160855138096e-05}
{"train_loss": 0.43360230326652527, "global_step": 23960, "epoch": 269, "lr": 9.816145279007754e-05}
{"train_loss": 0.4331360161304474, "global_step": 23961, "epoch": 269, "lr": 9.81612970222994e-05}
{"train_loss": 0.5495163798332214, "global_step": 23962, "epoch": 269, "lr": 9.816114124804657e-05}
{"train_loss": 0.4595766067504883, "global_step": 23963, "epoch": 269, "lr": 9.816098546731906e-05}
{"train_loss": 0.4971793591976166, "global_step": 23964, "epoch": 269, "lr": 9.816082968011688e-05}
{"train_loss": 0.4166269600391388, "global_step": 23965, "epoch": 269, "lr": 9.816067388644008e-05}
{"train_loss": 0.40174728631973267, "global_step": 23966, "epoch": 269, "lr": 9.816051808628865e-05}
{"train_loss": 0.45761778950691223, "global_step": 23967, "epoch": 269, "lr": 9.816036227966263e-05}
{"train_loss": 0.4130091071128845, "global_step": 23968, "epoch": 269, "lr": 9.816020646656205e-05}
{"train_loss": 0.39667975902557373, "global_step": 23969, "epoch": 269, "lr": 9.81600506469869e-05}
{"train_loss": 0.303577184677124, "global_step": 23970, "epoch": 269, "lr": 9.815989482093724e-05}
{"train_loss": 0.45371443033218384, "global_step": 23971, "epoch": 269, "lr": 9.815973898841304e-05}
{"train_loss": 0.5465250015258789, "global_step": 23972, "epoch": 269, "lr": 9.815958314941437e-05}
{"train_loss": 0.4612763226032257, "global_step": 23973, "epoch": 269, "lr": 9.815942730394122e-05}
{"train_loss": 0.5312288403511047, "global_step": 23974, "epoch": 269, "lr": 9.815927145199363e-05}
{"train_loss": 0.4060112535953522, "global_step": 23975, "epoch": 269, "lr": 9.81591155935716e-05}
{"train_loss": 0.51082444190979, "global_step": 23976, "epoch": 269, "lr": 9.815895972867518e-05}
{"train_loss": 0.40437281131744385, "global_step": 23977, "epoch": 269, "lr": 9.815880385730437e-05}
{"train_loss": 0.37033841013908386, "global_step": 23978, "epoch": 269, "lr": 9.81586479794592e-05}
{"train_loss": 0.5528990030288696, "global_step": 23979, "epoch": 269, "lr": 9.815849209513968e-05}
{"train_loss": 0.49461355805397034, "global_step": 23980, "epoch": 269, "lr": 9.815833620434583e-05}
{"train_loss": 0.6936676502227783, "global_step": 23981, "epoch": 269, "lr": 9.815818030707769e-05}
{"train_loss": 0.4638028144836426, "global_step": 23982, "epoch": 269, "lr": 9.815802440333527e-05}
{"train_loss": 0.4829578399658203, "global_step": 23983, "epoch": 269, "lr": 9.815786849311858e-05}
{"train_loss": 0.417637437582016, "global_step": 23984, "epoch": 269, "lr": 9.815771257642766e-05}
{"train_loss": 0.3414328396320343, "global_step": 23985, "epoch": 269, "lr": 9.815755665326252e-05}
{"train_loss": 0.4594574570655823, "global_step": 23986, "epoch": 269, "lr": 9.815740072362318e-05}
{"train_loss": 0.5293864607810974, "global_step": 23987, "epoch": 269, "lr": 9.815724478750967e-05}
{"train_loss": 0.3524065315723419, "global_step": 23988, "epoch": 269, "lr": 9.815708884492201e-05}
{"train_loss": 0.45089098811149597, "global_step": 23989, "epoch": 269, "lr": 9.815693289586021e-05}
{"train_loss": 0.48233669996261597, "global_step": 23990, "epoch": 269, "lr": 9.81567769403243e-05}
{"train_loss": 0.4144771695137024, "global_step": 23991, "epoch": 269, "lr": 9.81566209783143e-05}
{"train_loss": 0.4752381145954132, "global_step": 23992, "epoch": 269, "lr": 9.815646500983023e-05}
{"train_loss": 0.5439501404762268, "global_step": 23993, "epoch": 269, "lr": 9.81563090348721e-05}
{"train_loss": 0.40000876784324646, "global_step": 23994, "epoch": 269, "lr": 9.815615305343994e-05}
{"train_loss": 0.5872079730033875, "global_step": 23995, "epoch": 269, "lr": 9.815599706553379e-05}
{"train_loss": 0.34298643469810486, "global_step": 23996, "epoch": 269, "lr": 9.815584107115364e-05}
{"train_loss": 0.37325701117515564, "global_step": 23997, "epoch": 269, "lr": 9.815568507029953e-05}
{"train_loss": 0.503481388092041, "global_step": 23998, "epoch": 269, "lr": 9.815552906297148e-05}
{"train_loss": 0.5338581800460815, "global_step": 23999, "epoch": 269, "lr": 9.815537304916949e-05}
{"train_loss": 0.5135372281074524, "global_step": 24000, "epoch": 269, "lr": 9.81552170288936e-05}
{"train_loss": 0.3723611533641815, "global_step": 24001, "epoch": 269, "lr": 9.815506100214385e-05}
{"train_loss": 0.5136429667472839, "global_step": 24002, "epoch": 269, "lr": 9.815490496892022e-05}
{"train_loss": 0.4924963116645813, "global_step": 24003, "epoch": 269, "lr": 9.815474892922274e-05}
{"train_loss": 0.536699116230011, "global_step": 24004, "epoch": 269, "lr": 9.815459288305146e-05}
{"train_loss": 0.46830010414123535, "global_step": 24005, "epoch": 269, "lr": 9.815443683040637e-05}
{"train_loss": 0.4276524484157562, "global_step": 24006, "epoch": 269, "lr": 9.81542807712875e-05}
{"train_loss": 0.4589560329914093, "global_step": 24007, "epoch": 269, "lr": 9.815412470569489e-05}
{"train_loss": 0.42344191670417786, "global_step": 24008, "epoch": 269, "lr": 9.815396863362854e-05}
{"train_loss": 0.48047906160354614, "global_step": 24009, "epoch": 269, "lr": 9.815381255508846e-05}
{"train_loss": 0.4363938868045807, "global_step": 24010, "epoch": 269, "lr": 9.81536564700747e-05}
{"train_loss": 0.4872577488422394, "global_step": 24011, "epoch": 269, "lr": 9.815350037858727e-05}
{"train_loss": 0.584463894367218, "global_step": 24012, "epoch": 269, "lr": 9.815334428062619e-05}
{"train_loss": 0.444619357585907, "global_step": 24013, "epoch": 269, "lr": 9.815318817619147e-05}
{"train_loss": 0.4321369528770447, "global_step": 24014, "epoch": 269, "lr": 9.815303206528315e-05}
{"train_loss": 0.4701000452041626, "global_step": 24015, "epoch": 269, "lr": 9.815287594790124e-05}
{"train_loss": 0.3773862421512604, "global_step": 24016, "epoch": 269, "lr": 9.815271982404575e-05}
{"train_loss": 0.323478639125824, "global_step": 24017, "epoch": 269, "lr": 9.815256369371672e-05}
{"train_loss": 0.4290749430656433, "global_step": 24018, "epoch": 269, "lr": 9.815240755691417e-05}
{"train_loss": 0.5066238045692444, "global_step": 24019, "epoch": 269, "lr": 9.815225141363814e-05}
{"train_loss": 0.35295259952545166, "global_step": 24020, "epoch": 269, "lr": 9.815209526388858e-05}
{"train_loss": 0.3849687874317169, "global_step": 24021, "epoch": 269, "lr": 9.815193910766558e-05}
{"train_loss": 0.3534199595451355, "global_step": 24022, "epoch": 269, "lr": 9.815178294496914e-05}
{"train_loss": 0.4030653238296509, "global_step": 24023, "epoch": 269, "lr": 9.815162677579927e-05}
{"train_loss": 0.46105191111564636, "global_step": 24024, "epoch": 269, "lr": 9.815147060015601e-05}
{"train_loss": 0.42710092663764954, "global_step": 24025, "epoch": 269, "lr": 9.815131441803938e-05}
{"train_loss": 0.4122469127178192, "global_step": 24026, "epoch": 269, "lr": 9.815115822944938e-05}
{"train_loss": 0.5378268361091614, "global_step": 24027, "epoch": 269, "lr": 9.815100203438605e-05}
{"train_loss": 0.5141652226448059, "global_step": 24028, "epoch": 269, "lr": 9.815084583284939e-05}
{"train_loss": 0.4490422084090415, "global_step": 24029, "epoch": 269, "lr": 9.815068962483945e-05, "val_loss": 1.9070069789886475}
{"train_loss": 0.3282696604728699, "global_step": 24030, "epoch": 270, "lr": 9.815053341035624e-05}
{"train_loss": 0.4060657024383545, "global_step": 24031, "epoch": 270, "lr": 9.815037718939976e-05}
{"train_loss": 0.4503241181373596, "global_step": 24032, "epoch": 270, "lr": 9.815022096197006e-05}
{"train_loss": 0.4540448784828186, "global_step": 24033, "epoch": 270, "lr": 9.815006472806716e-05}
{"train_loss": 0.34424588084220886, "global_step": 24034, "epoch": 270, "lr": 9.814990848769105e-05}
{"train_loss": 0.42081376910209656, "global_step": 24035, "epoch": 270, "lr": 9.814975224084179e-05}
{"train_loss": 0.36715805530548096, "global_step": 24036, "epoch": 270, "lr": 9.814959598751937e-05}
{"train_loss": 0.3348301947116852, "global_step": 24037, "epoch": 270, "lr": 9.814943972772384e-05}
{"train_loss": 0.45808541774749756, "global_step": 24038, "epoch": 270, "lr": 9.814928346145518e-05}
{"train_loss": 0.42861855030059814, "global_step": 24039, "epoch": 270, "lr": 9.814912718871346e-05}
{"train_loss": 0.4331023395061493, "global_step": 24040, "epoch": 270, "lr": 9.814897090949866e-05}
{"train_loss": 0.4119483232498169, "global_step": 24041, "epoch": 270, "lr": 9.814881462381081e-05}
{"train_loss": 0.40569454431533813, "global_step": 24042, "epoch": 270, "lr": 9.814865833164996e-05}
{"train_loss": 0.299873024225235, "global_step": 24043, "epoch": 270, "lr": 9.81485020330161e-05}
{"train_loss": 0.4291956424713135, "global_step": 24044, "epoch": 270, "lr": 9.814834572790926e-05}
{"train_loss": 0.3778384327888489, "global_step": 24045, "epoch": 270, "lr": 9.814818941632946e-05}
{"train_loss": 0.49960777163505554, "global_step": 24046, "epoch": 270, "lr": 9.814803309827673e-05}
{"train_loss": 0.5080326199531555, "global_step": 24047, "epoch": 270, "lr": 9.814787677375108e-05}
{"train_loss": 0.30778947472572327, "global_step": 24048, "epoch": 270, "lr": 9.814772044275253e-05}
{"train_loss": 0.36081838607788086, "global_step": 24049, "epoch": 270, "lr": 9.814756410528112e-05}
{"train_loss": 0.4383278787136078, "global_step": 24050, "epoch": 270, "lr": 9.814740776133684e-05}
{"train_loss": 0.44588059186935425, "global_step": 24051, "epoch": 270, "lr": 9.814725141091973e-05}
{"train_loss": 0.3344193696975708, "global_step": 24052, "epoch": 270, "lr": 9.814709505402981e-05}
{"train_loss": 0.4749702513217926, "global_step": 24053, "epoch": 270, "lr": 9.814693869066711e-05}
{"train_loss": 0.35157567262649536, "global_step": 24054, "epoch": 270, "lr": 9.814678232083164e-05}
{"train_loss": 0.38380929827690125, "global_step": 24055, "epoch": 270, "lr": 9.814662594452342e-05}
{"train_loss": 0.49986135959625244, "global_step": 24056, "epoch": 270, "lr": 9.814646956174247e-05}
{"train_loss": 0.31231316924095154, "global_step": 24057, "epoch": 270, "lr": 9.814631317248882e-05}
{"train_loss": 0.3998124599456787, "global_step": 24058, "epoch": 270, "lr": 9.814615677676246e-05}
{"train_loss": 0.39268365502357483, "global_step": 24059, "epoch": 270, "lr": 9.814600037456346e-05}
{"train_loss": 0.48702436685562134, "global_step": 24060, "epoch": 270, "lr": 9.814584396589182e-05}
{"train_loss": 0.5560938715934753, "global_step": 24061, "epoch": 270, "lr": 9.814568755074754e-05}
{"train_loss": 0.42509177327156067, "global_step": 24062, "epoch": 270, "lr": 9.814553112913067e-05}
{"train_loss": 0.4064141511917114, "global_step": 24063, "epoch": 270, "lr": 9.814537470104122e-05}
{"train_loss": 0.4013432562351227, "global_step": 24064, "epoch": 270, "lr": 9.814521826647922e-05}
{"train_loss": 0.37653350830078125, "global_step": 24065, "epoch": 270, "lr": 9.814506182544467e-05}
{"train_loss": 0.41651561856269836, "global_step": 24066, "epoch": 270, "lr": 9.814490537793761e-05}
{"train_loss": 0.34513556957244873, "global_step": 24067, "epoch": 270, "lr": 9.814474892395805e-05}
{"train_loss": 0.46108105778694153, "global_step": 24068, "epoch": 270, "lr": 9.814459246350602e-05}
{"train_loss": 0.548019289970398, "global_step": 24069, "epoch": 270, "lr": 9.814443599658152e-05}
{"train_loss": 0.4366038143634796, "global_step": 24070, "epoch": 270, "lr": 9.81442795231846e-05}
{"train_loss": 0.4739034175872803, "global_step": 24071, "epoch": 270, "lr": 9.814412304331527e-05}
{"train_loss": 0.4252682626247406, "global_step": 24072, "epoch": 270, "lr": 9.814396655697355e-05}
{"train_loss": 0.4286317825317383, "global_step": 24073, "epoch": 270, "lr": 9.814381006415945e-05}
{"train_loss": 0.3969303071498871, "global_step": 24074, "epoch": 270, "lr": 9.814365356487302e-05}
{"train_loss": 0.5210843682289124, "global_step": 24075, "epoch": 270, "lr": 9.814349705911426e-05}
{"train_loss": 0.4770027995109558, "global_step": 24076, "epoch": 270, "lr": 9.814334054688317e-05}
{"train_loss": 0.5083613395690918, "global_step": 24077, "epoch": 270, "lr": 9.814318402817982e-05}
{"train_loss": 0.5590994358062744, "global_step": 24078, "epoch": 270, "lr": 9.814302750300419e-05}
{"train_loss": 0.39126625657081604, "global_step": 24079, "epoch": 270, "lr": 9.814287097135633e-05}
{"train_loss": 0.4826975464820862, "global_step": 24080, "epoch": 270, "lr": 9.814271443323624e-05}
{"train_loss": 0.4587200880050659, "global_step": 24081, "epoch": 270, "lr": 9.814255788864394e-05}
{"train_loss": 0.4148712158203125, "global_step": 24082, "epoch": 270, "lr": 9.814240133757947e-05}
{"train_loss": 0.38373491168022156, "global_step": 24083, "epoch": 270, "lr": 9.814224478004283e-05}
{"train_loss": 0.45120131969451904, "global_step": 24084, "epoch": 270, "lr": 9.814208821603408e-05}
{"train_loss": 0.33980637788772583, "global_step": 24085, "epoch": 270, "lr": 9.814193164555319e-05}
{"train_loss": 0.29205748438835144, "global_step": 24086, "epoch": 270, "lr": 9.814177506860021e-05}
{"train_loss": 0.4227212071418762, "global_step": 24087, "epoch": 270, "lr": 9.814161848517515e-05}
{"train_loss": 0.5877640843391418, "global_step": 24088, "epoch": 270, "lr": 9.814146189527804e-05}
{"train_loss": 0.558799147605896, "global_step": 24089, "epoch": 270, "lr": 9.814130529890891e-05}
{"train_loss": 0.3254285752773285, "global_step": 24090, "epoch": 270, "lr": 9.814114869606774e-05}
{"train_loss": 0.3911172151565552, "global_step": 24091, "epoch": 270, "lr": 9.81409920867546e-05}
{"train_loss": 0.482932984828949, "global_step": 24092, "epoch": 270, "lr": 9.814083547096949e-05}
{"train_loss": 0.3922378122806549, "global_step": 24093, "epoch": 270, "lr": 9.814067884871242e-05}
{"train_loss": 0.47854992747306824, "global_step": 24094, "epoch": 270, "lr": 9.814052221998342e-05}
{"train_loss": 0.3125176727771759, "global_step": 24095, "epoch": 270, "lr": 9.814036558478253e-05}
{"train_loss": 0.4142865538597107, "global_step": 24096, "epoch": 270, "lr": 9.814020894310975e-05}
{"train_loss": 0.5244449377059937, "global_step": 24097, "epoch": 270, "lr": 9.81400522949651e-05}
{"train_loss": 0.32777199149131775, "global_step": 24098, "epoch": 270, "lr": 9.81398956403486e-05}
{"train_loss": 0.43828460574150085, "global_step": 24099, "epoch": 270, "lr": 9.81397389792603e-05}
{"train_loss": 0.3611915707588196, "global_step": 24100, "epoch": 270, "lr": 9.813958231170018e-05}
{"train_loss": 0.386747270822525, "global_step": 24101, "epoch": 270, "lr": 9.813942563766829e-05}
{"train_loss": 0.4730467200279236, "global_step": 24102, "epoch": 270, "lr": 9.813926895716464e-05}
{"train_loss": 0.41751259565353394, "global_step": 24103, "epoch": 270, "lr": 9.813911227018924e-05}
{"train_loss": 0.5018492341041565, "global_step": 24104, "epoch": 270, "lr": 9.813895557674214e-05}
{"train_loss": 0.5269395112991333, "global_step": 24105, "epoch": 270, "lr": 9.813879887682334e-05}
{"train_loss": 0.4859445095062256, "global_step": 24106, "epoch": 270, "lr": 9.813864217043287e-05}
{"train_loss": 0.42087408900260925, "global_step": 24107, "epoch": 270, "lr": 9.813848545757072e-05}
{"train_loss": 0.38716673851013184, "global_step": 24108, "epoch": 270, "lr": 9.813832873823696e-05}
{"train_loss": 0.4093109667301178, "global_step": 24109, "epoch": 270, "lr": 9.813817201243159e-05}
{"train_loss": 0.4542137086391449, "global_step": 24110, "epoch": 270, "lr": 9.813801528015462e-05}
{"train_loss": 0.5772042274475098, "global_step": 24111, "epoch": 270, "lr": 9.813785854140609e-05}
{"train_loss": 0.5055031180381775, "global_step": 24112, "epoch": 270, "lr": 9.8137701796186e-05}
{"train_loss": 0.5801003575325012, "global_step": 24113, "epoch": 270, "lr": 9.81375450444944e-05}
{"train_loss": 0.43345388770103455, "global_step": 24114, "epoch": 270, "lr": 9.813738828633127e-05}
{"train_loss": 0.4169785976409912, "global_step": 24115, "epoch": 270, "lr": 9.813723152169667e-05}
{"train_loss": 0.29763466119766235, "global_step": 24116, "epoch": 270, "lr": 9.813707475059059e-05}
{"train_loss": 0.48471012711524963, "global_step": 24117, "epoch": 270, "lr": 9.813691797301308e-05}
{"train_loss": 0.4292453279655971, "global_step": 24118, "epoch": 270, "lr": 9.813676118896415e-05, "val_loss": 1.9433976411819458, "train_action_mse_error": 31.674814224243164}
{"train_loss": 0.4326801300048828, "global_step": 24119, "epoch": 271, "lr": 9.813660439844382e-05}
{"train_loss": 0.3899380564689636, "global_step": 24120, "epoch": 271, "lr": 9.81364476014521e-05}
{"train_loss": 0.4963369071483612, "global_step": 24121, "epoch": 271, "lr": 9.813629079798904e-05}
{"train_loss": 0.5035902857780457, "global_step": 24122, "epoch": 271, "lr": 9.813613398805463e-05}
{"train_loss": 0.41231590509414673, "global_step": 24123, "epoch": 271, "lr": 9.81359771716489e-05}
{"train_loss": 0.4627934694290161, "global_step": 24124, "epoch": 271, "lr": 9.813582034877189e-05}
{"train_loss": 0.3256368935108185, "global_step": 24125, "epoch": 271, "lr": 9.813566351942358e-05}
{"train_loss": 0.4702070951461792, "global_step": 24126, "epoch": 271, "lr": 9.813550668360404e-05}
{"train_loss": 0.41126081347465515, "global_step": 24127, "epoch": 271, "lr": 9.813534984131326e-05}
{"train_loss": 0.48985227942466736, "global_step": 24128, "epoch": 271, "lr": 9.813519299255127e-05}
{"train_loss": 0.5702865123748779, "global_step": 24129, "epoch": 271, "lr": 9.81350361373181e-05}
{"train_loss": 0.3905964195728302, "global_step": 24130, "epoch": 271, "lr": 9.813487927561374e-05}
{"train_loss": 0.40089887380599976, "global_step": 24131, "epoch": 271, "lr": 9.813472240743824e-05}
{"train_loss": 0.32869645953178406, "global_step": 24132, "epoch": 271, "lr": 9.813456553279161e-05}
{"train_loss": 0.4828759729862213, "global_step": 24133, "epoch": 271, "lr": 9.813440865167388e-05}
{"train_loss": 0.4679836928844452, "global_step": 24134, "epoch": 271, "lr": 9.813425176408506e-05}
{"train_loss": 0.409453809261322, "global_step": 24135, "epoch": 271, "lr": 9.813409487002519e-05}
{"train_loss": 0.44630923867225647, "global_step": 24136, "epoch": 271, "lr": 9.813393796949426e-05}
{"train_loss": 0.4678192436695099, "global_step": 24137, "epoch": 271, "lr": 9.813378106249232e-05}
{"train_loss": 0.4100019633769989, "global_step": 24138, "epoch": 271, "lr": 9.813362414901938e-05}
{"train_loss": 0.3679797947406769, "global_step": 24139, "epoch": 271, "lr": 9.813346722907546e-05}
{"train_loss": 0.3711335361003876, "global_step": 24140, "epoch": 271, "lr": 9.813331030266058e-05}
{"train_loss": 0.49474215507507324, "global_step": 24141, "epoch": 271, "lr": 9.813315336977476e-05}
{"train_loss": 0.4201422929763794, "global_step": 24142, "epoch": 271, "lr": 9.813299643041802e-05}
{"train_loss": 0.3726150393486023, "global_step": 24143, "epoch": 271, "lr": 9.81328394845904e-05}
{"train_loss": 0.369498610496521, "global_step": 24144, "epoch": 271, "lr": 9.81326825322919e-05}
{"train_loss": 0.4038543999195099, "global_step": 24145, "epoch": 271, "lr": 9.813252557352254e-05}
{"train_loss": 0.43729329109191895, "global_step": 24146, "epoch": 271, "lr": 9.813236860828237e-05}
{"train_loss": 0.46364885568618774, "global_step": 24147, "epoch": 271, "lr": 9.813221163657136e-05}
{"train_loss": 0.4041935205459595, "global_step": 24148, "epoch": 271, "lr": 9.813205465838958e-05}
{"train_loss": 0.534467875957489, "global_step": 24149, "epoch": 271, "lr": 9.813189767373703e-05}
{"train_loss": 0.52003413438797, "global_step": 24150, "epoch": 271, "lr": 9.813174068261373e-05}
{"train_loss": 0.37717998027801514, "global_step": 24151, "epoch": 271, "lr": 9.81315836850197e-05}
{"train_loss": 0.42871788144111633, "global_step": 24152, "epoch": 271, "lr": 9.813142668095496e-05}
{"train_loss": 0.42022326588630676, "global_step": 24153, "epoch": 271, "lr": 9.813126967041955e-05}
{"train_loss": 0.43316522240638733, "global_step": 24154, "epoch": 271, "lr": 9.813111265341348e-05}
{"train_loss": 0.36752474308013916, "global_step": 24155, "epoch": 271, "lr": 9.813095562993677e-05}
{"train_loss": 0.4856710135936737, "global_step": 24156, "epoch": 271, "lr": 9.813079859998941e-05}
{"train_loss": 0.31706029176712036, "global_step": 24157, "epoch": 271, "lr": 9.813064156357149e-05}
{"train_loss": 0.3907989263534546, "global_step": 24158, "epoch": 271, "lr": 9.813048452068296e-05}
{"train_loss": 0.3936443030834198, "global_step": 24159, "epoch": 271, "lr": 9.813032747132389e-05}
{"train_loss": 0.435556024312973, "global_step": 24160, "epoch": 271, "lr": 9.813017041549428e-05}
{"train_loss": 0.33879396319389343, "global_step": 24161, "epoch": 271, "lr": 9.813001335319416e-05}
{"train_loss": 0.3839973211288452, "global_step": 24162, "epoch": 271, "lr": 9.812985628442354e-05}
{"train_loss": 0.46815329790115356, "global_step": 24163, "epoch": 271, "lr": 9.812969920918245e-05}
{"train_loss": 0.4700009524822235, "global_step": 24164, "epoch": 271, "lr": 9.812954212747092e-05}
{"train_loss": 0.3292749524116516, "global_step": 24165, "epoch": 271, "lr": 9.812938503928894e-05}
{"train_loss": 0.4213416874408722, "global_step": 24166, "epoch": 271, "lr": 9.812922794463658e-05}
{"train_loss": 0.3728116750717163, "global_step": 24167, "epoch": 271, "lr": 9.812907084351379e-05}
{"train_loss": 0.5128390789031982, "global_step": 24168, "epoch": 271, "lr": 9.812891373592066e-05}
{"train_loss": 0.4857809543609619, "global_step": 24169, "epoch": 271, "lr": 9.812875662185719e-05}
{"train_loss": 0.2957213819026947, "global_step": 24170, "epoch": 271, "lr": 9.812859950132339e-05}
{"train_loss": 0.4717440605163574, "global_step": 24171, "epoch": 271, "lr": 9.812844237431926e-05}
{"train_loss": 0.4056837260723114, "global_step": 24172, "epoch": 271, "lr": 9.812828524084488e-05}
{"train_loss": 0.39302098751068115, "global_step": 24173, "epoch": 271, "lr": 9.812812810090021e-05}
{"train_loss": 0.4882396161556244, "global_step": 24174, "epoch": 271, "lr": 9.812797095448533e-05}
{"train_loss": 0.3629037141799927, "global_step": 24175, "epoch": 271, "lr": 9.812781380160021e-05}
{"train_loss": 0.4195441007614136, "global_step": 24176, "epoch": 271, "lr": 9.812765664224491e-05}
{"train_loss": 0.39934155344963074, "global_step": 24177, "epoch": 271, "lr": 9.812749947641942e-05}
{"train_loss": 0.48427581787109375, "global_step": 24178, "epoch": 271, "lr": 9.812734230412377e-05}
{"train_loss": 0.5497708916664124, "global_step": 24179, "epoch": 271, "lr": 9.812718512535799e-05}
{"train_loss": 0.5025079250335693, "global_step": 24180, "epoch": 271, "lr": 9.812702794012209e-05}
{"train_loss": 0.440748006105423, "global_step": 24181, "epoch": 271, "lr": 9.812687074841611e-05}
{"train_loss": 0.5222851037979126, "global_step": 24182, "epoch": 271, "lr": 9.812671355024006e-05}
{"train_loss": 0.44648388028144836, "global_step": 24183, "epoch": 271, "lr": 9.812655634559396e-05}
{"train_loss": 0.48484498262405396, "global_step": 24184, "epoch": 271, "lr": 9.812639913447782e-05}
{"train_loss": 0.4159094989299774, "global_step": 24185, "epoch": 271, "lr": 9.812624191689169e-05}
{"train_loss": 0.3957182765007019, "global_step": 24186, "epoch": 271, "lr": 9.812608469283555e-05}
{"train_loss": 0.45626121759414673, "global_step": 24187, "epoch": 271, "lr": 9.812592746230947e-05}
{"train_loss": 0.4549083113670349, "global_step": 24188, "epoch": 271, "lr": 9.812577022531342e-05}
{"train_loss": 0.39649999141693115, "global_step": 24189, "epoch": 271, "lr": 9.812561298184746e-05}
{"train_loss": 0.3587800860404968, "global_step": 24190, "epoch": 271, "lr": 9.812545573191159e-05}
{"train_loss": 0.49778181314468384, "global_step": 24191, "epoch": 271, "lr": 9.812529847550584e-05}
{"train_loss": 0.5469122529029846, "global_step": 24192, "epoch": 271, "lr": 9.812514121263024e-05}
{"train_loss": 0.4073849320411682, "global_step": 24193, "epoch": 271, "lr": 9.81249839432848e-05}
{"train_loss": 0.4280311167240143, "global_step": 24194, "epoch": 271, "lr": 9.812482666746954e-05}
{"train_loss": 0.44456058740615845, "global_step": 24195, "epoch": 271, "lr": 9.812466938518447e-05}
{"train_loss": 0.3917268216609955, "global_step": 24196, "epoch": 271, "lr": 9.812451209642966e-05}
{"train_loss": 0.5001732110977173, "global_step": 24197, "epoch": 271, "lr": 9.812435480120507e-05}
{"train_loss": 0.4188501536846161, "global_step": 24198, "epoch": 271, "lr": 9.812419749951073e-05}
{"train_loss": 0.4030935764312744, "global_step": 24199, "epoch": 271, "lr": 9.812404019134671e-05}
{"train_loss": 0.40760695934295654, "global_step": 24200, "epoch": 271, "lr": 9.812388287671299e-05}
{"train_loss": 0.4265410900115967, "global_step": 24201, "epoch": 271, "lr": 9.812372555560961e-05}
{"train_loss": 0.4472992718219757, "global_step": 24202, "epoch": 271, "lr": 9.812356822803657e-05}
{"train_loss": 0.43743419647216797, "global_step": 24203, "epoch": 271, "lr": 9.81234108939939e-05}
{"train_loss": 0.4617411494255066, "global_step": 24204, "epoch": 271, "lr": 9.812325355348162e-05}
{"train_loss": 0.3686317801475525, "global_step": 24205, "epoch": 271, "lr": 9.812309620649977e-05}
{"train_loss": 0.3422737121582031, "global_step": 24206, "epoch": 271, "lr": 9.812293885304834e-05}
{"train_loss": 0.43116249864021045, "global_step": 24207, "epoch": 271, "lr": 9.812278149312738e-05, "val_loss": 1.9326220750808716}
{"train_loss": 0.4272882640361786, "global_step": 24208, "epoch": 272, "lr": 9.81226241267369e-05}
{"train_loss": 0.535485565662384, "global_step": 24209, "epoch": 272, "lr": 9.81224667538769e-05}
{"train_loss": 0.39741265773773193, "global_step": 24210, "epoch": 272, "lr": 9.812230937454744e-05}
{"train_loss": 0.48688626289367676, "global_step": 24211, "epoch": 272, "lr": 9.812215198874853e-05}
{"train_loss": 0.5235415697097778, "global_step": 24212, "epoch": 272, "lr": 9.812199459648016e-05}
{"train_loss": 0.4521578848361969, "global_step": 24213, "epoch": 272, "lr": 9.812183719774238e-05}
{"train_loss": 0.43544822931289673, "global_step": 24214, "epoch": 272, "lr": 9.812167979253522e-05}
{"train_loss": 0.41829201579093933, "global_step": 24215, "epoch": 272, "lr": 9.812152238085868e-05}
{"train_loss": 0.3518267273902893, "global_step": 24216, "epoch": 272, "lr": 9.812136496271278e-05}
{"train_loss": 0.397792249917984, "global_step": 24217, "epoch": 272, "lr": 9.812120753809755e-05}
{"train_loss": 0.41503143310546875, "global_step": 24218, "epoch": 272, "lr": 9.812105010701301e-05}
{"train_loss": 0.4018203318119049, "global_step": 24219, "epoch": 272, "lr": 9.812089266945919e-05}
{"train_loss": 0.425381600856781, "global_step": 24220, "epoch": 272, "lr": 9.81207352254361e-05}
{"train_loss": 0.46614915132522583, "global_step": 24221, "epoch": 272, "lr": 9.812057777494375e-05}
{"train_loss": 0.5813540816307068, "global_step": 24222, "epoch": 272, "lr": 9.812042031798219e-05}
{"train_loss": 0.47566163539886475, "global_step": 24223, "epoch": 272, "lr": 9.812026285455143e-05}
{"train_loss": 0.39200615882873535, "global_step": 24224, "epoch": 272, "lr": 9.812010538465147e-05}
{"train_loss": 0.4689619839191437, "global_step": 24225, "epoch": 272, "lr": 9.811994790828236e-05}
{"train_loss": 0.35256412625312805, "global_step": 24226, "epoch": 272, "lr": 9.81197904254441e-05}
{"train_loss": 0.32411304116249084, "global_step": 24227, "epoch": 272, "lr": 9.811963293613673e-05}
{"train_loss": 0.45939764380455017, "global_step": 24228, "epoch": 272, "lr": 9.811947544036026e-05}
{"train_loss": 0.38639023900032043, "global_step": 24229, "epoch": 272, "lr": 9.811931793811472e-05}
{"train_loss": 0.47286558151245117, "global_step": 24230, "epoch": 272, "lr": 9.811916042940011e-05}
{"train_loss": 0.40326425433158875, "global_step": 24231, "epoch": 272, "lr": 9.811900291421648e-05}
{"train_loss": 0.4049665331840515, "global_step": 24232, "epoch": 272, "lr": 9.811884539256383e-05}
{"train_loss": 0.4882737994194031, "global_step": 24233, "epoch": 272, "lr": 9.811868786444219e-05}
{"train_loss": 0.437116801738739, "global_step": 24234, "epoch": 272, "lr": 9.811853032985156e-05}
{"train_loss": 0.46947139501571655, "global_step": 24235, "epoch": 272, "lr": 9.811837278879201e-05}
{"train_loss": 0.3842763900756836, "global_step": 24236, "epoch": 272, "lr": 9.811821524126352e-05}
{"train_loss": 0.3217479884624481, "global_step": 24237, "epoch": 272, "lr": 9.811805768726612e-05}
{"train_loss": 0.440557062625885, "global_step": 24238, "epoch": 272, "lr": 9.811790012679982e-05}
{"train_loss": 0.5021073222160339, "global_step": 24239, "epoch": 272, "lr": 9.811774255986468e-05}
{"train_loss": 0.317190945148468, "global_step": 24240, "epoch": 272, "lr": 9.811758498646068e-05}
{"train_loss": 0.3237876296043396, "global_step": 24241, "epoch": 272, "lr": 9.811742740658785e-05}
{"train_loss": 0.3616349399089813, "global_step": 24242, "epoch": 272, "lr": 9.811726982024623e-05}
{"train_loss": 0.5081884860992432, "global_step": 24243, "epoch": 272, "lr": 9.811711222743582e-05}
{"train_loss": 0.41402775049209595, "global_step": 24244, "epoch": 272, "lr": 9.811695462815667e-05}
{"train_loss": 0.36509808897972107, "global_step": 24245, "epoch": 272, "lr": 9.811679702240877e-05}
{"train_loss": 0.4554476737976074, "global_step": 24246, "epoch": 272, "lr": 9.811663941019216e-05}
{"train_loss": 0.4547470211982727, "global_step": 24247, "epoch": 272, "lr": 9.811648179150686e-05}
{"train_loss": 0.383131742477417, "global_step": 24248, "epoch": 272, "lr": 9.811632416635285e-05}
{"train_loss": 0.4776601791381836, "global_step": 24249, "epoch": 272, "lr": 9.811616653473021e-05}
{"train_loss": 0.4101862907409668, "global_step": 24250, "epoch": 272, "lr": 9.811600889663894e-05}
{"train_loss": 0.5147462487220764, "global_step": 24251, "epoch": 272, "lr": 9.811585125207907e-05}
{"train_loss": 0.3628804087638855, "global_step": 24252, "epoch": 272, "lr": 9.81156936010506e-05}
{"train_loss": 0.4756683111190796, "global_step": 24253, "epoch": 272, "lr": 9.811553594355357e-05}
{"train_loss": 0.4724369943141937, "global_step": 24254, "epoch": 272, "lr": 9.811537827958797e-05}
{"train_loss": 0.39176568388938904, "global_step": 24255, "epoch": 272, "lr": 9.811522060915386e-05}
{"train_loss": 0.3923787474632263, "global_step": 24256, "epoch": 272, "lr": 9.811506293225124e-05}
{"train_loss": 0.43302035331726074, "global_step": 24257, "epoch": 272, "lr": 9.811490524888013e-05}
{"train_loss": 0.46270039677619934, "global_step": 24258, "epoch": 272, "lr": 9.811474755904057e-05}
{"train_loss": 0.3503612279891968, "global_step": 24259, "epoch": 272, "lr": 9.811458986273257e-05}
{"train_loss": 0.3658509850502014, "global_step": 24260, "epoch": 272, "lr": 9.811443215995613e-05}
{"train_loss": 0.4572649896144867, "global_step": 24261, "epoch": 272, "lr": 9.81142744507113e-05}
{"train_loss": 0.47070059180259705, "global_step": 24262, "epoch": 272, "lr": 9.811411673499811e-05}
{"train_loss": 0.4753924012184143, "global_step": 24263, "epoch": 272, "lr": 9.811395901281655e-05}
{"train_loss": 0.3241143524646759, "global_step": 24264, "epoch": 272, "lr": 9.811380128416665e-05}
{"train_loss": 0.4898930490016937, "global_step": 24265, "epoch": 272, "lr": 9.811364354904844e-05}
{"train_loss": 0.40785086154937744, "global_step": 24266, "epoch": 272, "lr": 9.811348580746194e-05}
{"train_loss": 0.5660407543182373, "global_step": 24267, "epoch": 272, "lr": 9.811332805940716e-05}
{"train_loss": 0.3697202801704407, "global_step": 24268, "epoch": 272, "lr": 9.811317030488414e-05}
{"train_loss": 0.40529361367225647, "global_step": 24269, "epoch": 272, "lr": 9.811301254389287e-05}
{"train_loss": 0.3826411962509155, "global_step": 24270, "epoch": 272, "lr": 9.811285477643342e-05}
{"train_loss": 0.4379850924015045, "global_step": 24271, "epoch": 272, "lr": 9.811269700250576e-05}
{"train_loss": 0.4351402223110199, "global_step": 24272, "epoch": 272, "lr": 9.811253922210994e-05}
{"train_loss": 0.36900365352630615, "global_step": 24273, "epoch": 272, "lr": 9.811238143524599e-05}
{"train_loss": 0.4439350664615631, "global_step": 24274, "epoch": 272, "lr": 9.81122236419139e-05}
{"train_loss": 0.4898854196071625, "global_step": 24275, "epoch": 272, "lr": 9.811206584211371e-05}
{"train_loss": 0.40194186568260193, "global_step": 24276, "epoch": 272, "lr": 9.811190803584547e-05}
{"train_loss": 0.39842572808265686, "global_step": 24277, "epoch": 272, "lr": 9.811175022310914e-05}
{"train_loss": 0.4593939185142517, "global_step": 24278, "epoch": 272, "lr": 9.811159240390478e-05}
{"train_loss": 0.37125059962272644, "global_step": 24279, "epoch": 272, "lr": 9.811143457823239e-05}
{"train_loss": 0.4745575487613678, "global_step": 24280, "epoch": 272, "lr": 9.811127674609202e-05}
{"train_loss": 0.31551048159599304, "global_step": 24281, "epoch": 272, "lr": 9.811111890748365e-05}
{"train_loss": 0.37974539399147034, "global_step": 24282, "epoch": 272, "lr": 9.811096106240736e-05}
{"train_loss": 0.3856455683708191, "global_step": 24283, "epoch": 272, "lr": 9.811080321086312e-05}
{"train_loss": 0.45011523365974426, "global_step": 24284, "epoch": 272, "lr": 9.811064535285097e-05}
{"train_loss": 0.4662112891674042, "global_step": 24285, "epoch": 272, "lr": 9.811048748837093e-05}
{"train_loss": 0.49600622057914734, "global_step": 24286, "epoch": 272, "lr": 9.811032961742303e-05}
{"train_loss": 0.5834100842475891, "global_step": 24287, "epoch": 272, "lr": 9.811017174000727e-05}
{"train_loss": 0.3690875172615051, "global_step": 24288, "epoch": 272, "lr": 9.811001385612368e-05}
{"train_loss": 0.40758025646209717, "global_step": 24289, "epoch": 272, "lr": 9.810985596577231e-05}
{"train_loss": 0.607077419757843, "global_step": 24290, "epoch": 272, "lr": 9.810969806895314e-05}
{"train_loss": 0.43534213304519653, "global_step": 24291, "epoch": 272, "lr": 9.81095401656662e-05}
{"train_loss": 0.4444708228111267, "global_step": 24292, "epoch": 272, "lr": 9.810938225591153e-05}
{"train_loss": 0.39341363310813904, "global_step": 24293, "epoch": 272, "lr": 9.810922433968914e-05}
{"train_loss": 0.5001751780509949, "global_step": 24294, "epoch": 272, "lr": 9.810906641699904e-05}
{"train_loss": 0.4552580416202545, "global_step": 24295, "epoch": 272, "lr": 9.810890848784126e-05}
{"train_loss": 0.4312065846464607, "global_step": 24296, "epoch": 272, "lr": 9.810875055221584e-05, "val_loss": 1.996464490890503}
{"train_loss": 0.4292094111442566, "global_step": 24297, "epoch": 273, "lr": 9.810859261012277e-05}
{"train_loss": 0.5587239265441895, "global_step": 24298, "epoch": 273, "lr": 9.81084346615621e-05}
{"train_loss": 0.38384097814559937, "global_step": 24299, "epoch": 273, "lr": 9.810827670653384e-05}
{"train_loss": 0.4555811583995819, "global_step": 24300, "epoch": 273, "lr": 9.8108118745038e-05}
{"train_loss": 0.5843350887298584, "global_step": 24301, "epoch": 273, "lr": 9.81079607770746e-05}
{"train_loss": 0.5038654208183289, "global_step": 24302, "epoch": 273, "lr": 9.810780280264369e-05}
{"train_loss": 0.5257138013839722, "global_step": 24303, "epoch": 273, "lr": 9.810764482174526e-05}
{"train_loss": 0.447957307100296, "global_step": 24304, "epoch": 273, "lr": 9.810748683437936e-05}
{"train_loss": 0.5187791585922241, "global_step": 24305, "epoch": 273, "lr": 9.810732884054598e-05}
{"train_loss": 0.4665561616420746, "global_step": 24306, "epoch": 273, "lr": 9.810717084024515e-05}
{"train_loss": 0.4070306122303009, "global_step": 24307, "epoch": 273, "lr": 9.810701283347689e-05}
{"train_loss": 0.45325684547424316, "global_step": 24308, "epoch": 273, "lr": 9.810685482024126e-05}
{"train_loss": 0.5124548673629761, "global_step": 24309, "epoch": 273, "lr": 9.810669680053823e-05}
{"train_loss": 0.3349379897117615, "global_step": 24310, "epoch": 273, "lr": 9.810653877436784e-05}
{"train_loss": 0.4183836579322815, "global_step": 24311, "epoch": 273, "lr": 9.810638074173012e-05}
{"train_loss": 0.48509904742240906, "global_step": 24312, "epoch": 273, "lr": 9.810622270262508e-05}
{"train_loss": 0.47016018629074097, "global_step": 24313, "epoch": 273, "lr": 9.810606465705274e-05}
{"train_loss": 0.4939020574092865, "global_step": 24314, "epoch": 273, "lr": 9.810590660501313e-05}
{"train_loss": 0.48901060223579407, "global_step": 24315, "epoch": 273, "lr": 9.810574854650627e-05}
{"train_loss": 0.4077652394771576, "global_step": 24316, "epoch": 273, "lr": 9.810559048153218e-05}
{"train_loss": 0.560984194278717, "global_step": 24317, "epoch": 273, "lr": 9.810543241009087e-05}
{"train_loss": 0.4966771602630615, "global_step": 24318, "epoch": 273, "lr": 9.810527433218239e-05}
{"train_loss": 0.3754688501358032, "global_step": 24319, "epoch": 273, "lr": 9.810511624780672e-05}
{"train_loss": 0.4263755679130554, "global_step": 24320, "epoch": 273, "lr": 9.810495815696391e-05}
{"train_loss": 0.36269062757492065, "global_step": 24321, "epoch": 273, "lr": 9.810480005965398e-05}
{"train_loss": 0.3679032027721405, "global_step": 24322, "epoch": 273, "lr": 9.810464195587695e-05}
{"train_loss": 0.5539766550064087, "global_step": 24323, "epoch": 273, "lr": 9.810448384563282e-05}
{"train_loss": 0.47002342343330383, "global_step": 24324, "epoch": 273, "lr": 9.810432572892164e-05}
{"train_loss": 0.5389577150344849, "global_step": 24325, "epoch": 273, "lr": 9.810416760574342e-05}
{"train_loss": 0.3954963982105255, "global_step": 24326, "epoch": 273, "lr": 9.810400947609818e-05}
{"train_loss": 0.4350590705871582, "global_step": 24327, "epoch": 273, "lr": 9.810385133998594e-05}
{"train_loss": 0.4292950928211212, "global_step": 24328, "epoch": 273, "lr": 9.810369319740674e-05}
{"train_loss": 0.4334332346916199, "global_step": 24329, "epoch": 273, "lr": 9.810353504836056e-05}
{"train_loss": 0.4992702603340149, "global_step": 24330, "epoch": 273, "lr": 9.810337689284745e-05}
{"train_loss": 0.4170645475387573, "global_step": 24331, "epoch": 273, "lr": 9.810321873086744e-05}
{"train_loss": 0.4704091548919678, "global_step": 24332, "epoch": 273, "lr": 9.810306056242052e-05}
{"train_loss": 0.5533181428909302, "global_step": 24333, "epoch": 273, "lr": 9.810290238750675e-05}
{"train_loss": 0.42362627387046814, "global_step": 24334, "epoch": 273, "lr": 9.810274420612612e-05}
{"train_loss": 0.46045830845832825, "global_step": 24335, "epoch": 273, "lr": 9.810258601827867e-05}
{"train_loss": 0.5358217358589172, "global_step": 24336, "epoch": 273, "lr": 9.810242782396441e-05}
{"train_loss": 0.413203626871109, "global_step": 24337, "epoch": 273, "lr": 9.810226962318336e-05}
{"train_loss": 0.35266274213790894, "global_step": 24338, "epoch": 273, "lr": 9.810211141593555e-05}
{"train_loss": 0.4398511052131653, "global_step": 24339, "epoch": 273, "lr": 9.8101953202221e-05}
{"train_loss": 0.490181565284729, "global_step": 24340, "epoch": 273, "lr": 9.810179498203971e-05}
{"train_loss": 0.5247299671173096, "global_step": 24341, "epoch": 273, "lr": 9.810163675539174e-05}
{"train_loss": 0.5438780188560486, "global_step": 24342, "epoch": 273, "lr": 9.810147852227709e-05}
{"train_loss": 0.40228474140167236, "global_step": 24343, "epoch": 273, "lr": 9.810132028269578e-05}
{"train_loss": 0.48743942379951477, "global_step": 24344, "epoch": 273, "lr": 9.810116203664783e-05}
{"train_loss": 0.44229766726493835, "global_step": 24345, "epoch": 273, "lr": 9.810100378413327e-05}
{"train_loss": 0.3880503475666046, "global_step": 24346, "epoch": 273, "lr": 9.81008455251521e-05}
{"train_loss": 0.37705326080322266, "global_step": 24347, "epoch": 273, "lr": 9.810068725970439e-05}
{"train_loss": 0.46388310194015503, "global_step": 24348, "epoch": 273, "lr": 9.81005289877901e-05}
{"train_loss": 0.6221113204956055, "global_step": 24349, "epoch": 273, "lr": 9.81003707094093e-05}
{"train_loss": 0.4655361473560333, "global_step": 24350, "epoch": 273, "lr": 9.810021242456198e-05}
{"train_loss": 0.41488391160964966, "global_step": 24351, "epoch": 273, "lr": 9.810005413324816e-05}
{"train_loss": 0.39335840940475464, "global_step": 24352, "epoch": 273, "lr": 9.80998958354679e-05}
{"train_loss": 0.47559860348701477, "global_step": 24353, "epoch": 273, "lr": 9.809973753122119e-05}
{"train_loss": 0.4263279438018799, "global_step": 24354, "epoch": 273, "lr": 9.809957922050805e-05}
{"train_loss": 0.38831162452697754, "global_step": 24355, "epoch": 273, "lr": 9.80994209033285e-05}
{"train_loss": 0.48255684971809387, "global_step": 24356, "epoch": 273, "lr": 9.809926257968258e-05}
{"train_loss": 0.4490011930465698, "global_step": 24357, "epoch": 273, "lr": 9.809910424957029e-05}
{"train_loss": 0.5851262807846069, "global_step": 24358, "epoch": 273, "lr": 9.809894591299167e-05}
{"train_loss": 0.5198492407798767, "global_step": 24359, "epoch": 273, "lr": 9.809878756994675e-05}
{"train_loss": 0.4523118734359741, "global_step": 24360, "epoch": 273, "lr": 9.809862922043552e-05}
{"train_loss": 0.4986370801925659, "global_step": 24361, "epoch": 273, "lr": 9.8098470864458e-05}
{"train_loss": 0.39688077569007874, "global_step": 24362, "epoch": 273, "lr": 9.809831250201425e-05}
{"train_loss": 0.35479098558425903, "global_step": 24363, "epoch": 273, "lr": 9.809815413310425e-05}
{"train_loss": 0.3561849892139435, "global_step": 24364, "epoch": 273, "lr": 9.809799575772804e-05}
{"train_loss": 0.43461981415748596, "global_step": 24365, "epoch": 273, "lr": 9.809783737588565e-05}
{"train_loss": 0.5082690715789795, "global_step": 24366, "epoch": 273, "lr": 9.809767898757709e-05}
{"train_loss": 0.46207910776138306, "global_step": 24367, "epoch": 273, "lr": 9.809752059280238e-05}
{"train_loss": 0.4997040927410126, "global_step": 24368, "epoch": 273, "lr": 9.809736219156157e-05}
{"train_loss": 0.5012626051902771, "global_step": 24369, "epoch": 273, "lr": 9.809720378385461e-05}
{"train_loss": 0.4338841438293457, "global_step": 24370, "epoch": 273, "lr": 9.80970453696816e-05}
{"train_loss": 0.5166008472442627, "global_step": 24371, "epoch": 273, "lr": 9.809688694904251e-05}
{"train_loss": 0.4495082497596741, "global_step": 24372, "epoch": 273, "lr": 9.809672852193739e-05}
{"train_loss": 0.4050251543521881, "global_step": 24373, "epoch": 273, "lr": 9.809657008836626e-05}
{"train_loss": 0.5450568199157715, "global_step": 24374, "epoch": 273, "lr": 9.809641164832912e-05}
{"train_loss": 0.4123441278934479, "global_step": 24375, "epoch": 273, "lr": 9.8096253201826e-05}
{"train_loss": 0.38784685730934143, "global_step": 24376, "epoch": 273, "lr": 9.809609474885693e-05}
{"train_loss": 0.29482996463775635, "global_step": 24377, "epoch": 273, "lr": 9.809593628942193e-05}
{"train_loss": 0.5393363237380981, "global_step": 24378, "epoch": 273, "lr": 9.809577782352102e-05}
{"train_loss": 0.459364116191864, "global_step": 24379, "epoch": 273, "lr": 9.809561935115421e-05}
{"train_loss": 0.42179784178733826, "global_step": 24380, "epoch": 273, "lr": 9.809546087232154e-05}
{"train_loss": 0.4173855483531952, "global_step": 24381, "epoch": 273, "lr": 9.8095302387023e-05}
{"train_loss": 0.5578804612159729, "global_step": 24382, "epoch": 273, "lr": 9.809514389525864e-05}
{"train_loss": 0.44031015038490295, "global_step": 24383, "epoch": 273, "lr": 9.809498539702849e-05}
{"train_loss": 0.48070886731147766, "global_step": 24384, "epoch": 273, "lr": 9.809482689233256e-05}
{"train_loss": 0.4576190442181705, "global_step": 24385, "epoch": 273, "lr": 9.809466838117086e-05, "val_loss": 2.0110790729522705}
{"train_loss": 0.42652079463005066, "global_step": 24386, "epoch": 274, "lr": 9.809450986354341e-05}
{"train_loss": 0.4489983916282654, "global_step": 24387, "epoch": 274, "lr": 9.809435133945024e-05}
{"train_loss": 0.3948921859264374, "global_step": 24388, "epoch": 274, "lr": 9.809419280889136e-05}
{"train_loss": 0.4709090292453766, "global_step": 24389, "epoch": 274, "lr": 9.809403427186681e-05}
{"train_loss": 0.34909483790397644, "global_step": 24390, "epoch": 274, "lr": 9.80938757283766e-05}
{"train_loss": 0.36112314462661743, "global_step": 24391, "epoch": 274, "lr": 9.809371717842077e-05}
{"train_loss": 0.30678755044937134, "global_step": 24392, "epoch": 274, "lr": 9.809355862199931e-05}
{"train_loss": 0.3300836980342865, "global_step": 24393, "epoch": 274, "lr": 9.809340005911227e-05}
{"train_loss": 0.4190531075000763, "global_step": 24394, "epoch": 274, "lr": 9.809324148975966e-05}
{"train_loss": 0.4943138659000397, "global_step": 24395, "epoch": 274, "lr": 9.80930829139415e-05}
{"train_loss": 0.36816757917404175, "global_step": 24396, "epoch": 274, "lr": 9.80929243316578e-05}
{"train_loss": 0.5046965479850769, "global_step": 24397, "epoch": 274, "lr": 9.80927657429086e-05}
{"train_loss": 0.352835476398468, "global_step": 24398, "epoch": 274, "lr": 9.80926071476939e-05}
{"train_loss": 0.4607793390750885, "global_step": 24399, "epoch": 274, "lr": 9.809244854601376e-05}
{"train_loss": 0.4363718032836914, "global_step": 24400, "epoch": 274, "lr": 9.809228993786816e-05}
{"train_loss": 0.4200924038887024, "global_step": 24401, "epoch": 274, "lr": 9.809213132325714e-05}
{"train_loss": 0.4195747375488281, "global_step": 24402, "epoch": 274, "lr": 9.809197270218072e-05}
{"train_loss": 0.38243862986564636, "global_step": 24403, "epoch": 274, "lr": 9.809181407463892e-05}
{"train_loss": 0.4625605642795563, "global_step": 24404, "epoch": 274, "lr": 9.809165544063176e-05}
{"train_loss": 0.3984721302986145, "global_step": 24405, "epoch": 274, "lr": 9.809149680015926e-05}
{"train_loss": 0.4245845079421997, "global_step": 24406, "epoch": 274, "lr": 9.809133815322144e-05}
{"train_loss": 0.37829920649528503, "global_step": 24407, "epoch": 274, "lr": 9.809117949981834e-05}
{"train_loss": 0.4921291768550873, "global_step": 24408, "epoch": 274, "lr": 9.809102083994997e-05}
{"train_loss": 0.3022221326828003, "global_step": 24409, "epoch": 274, "lr": 9.809086217361633e-05}
{"train_loss": 0.4236331880092621, "global_step": 24410, "epoch": 274, "lr": 9.809070350081749e-05}
{"train_loss": 0.4388727843761444, "global_step": 24411, "epoch": 274, "lr": 9.809054482155341e-05}
{"train_loss": 0.5608633756637573, "global_step": 24412, "epoch": 274, "lr": 9.809038613582415e-05}
{"train_loss": 0.4723801612854004, "global_step": 24413, "epoch": 274, "lr": 9.809022744362972e-05}
{"train_loss": 0.4824482500553131, "global_step": 24414, "epoch": 274, "lr": 9.809006874497016e-05}
{"train_loss": 0.45602428913116455, "global_step": 24415, "epoch": 274, "lr": 9.808991003984547e-05}
{"train_loss": 0.41747137904167175, "global_step": 24416, "epoch": 274, "lr": 9.808975132825568e-05}
{"train_loss": 0.532229483127594, "global_step": 24417, "epoch": 274, "lr": 9.808959261020081e-05}
{"train_loss": 0.35713014006614685, "global_step": 24418, "epoch": 274, "lr": 9.808943388568087e-05}
{"train_loss": 0.5336015224456787, "global_step": 24419, "epoch": 274, "lr": 9.80892751546959e-05}
{"train_loss": 0.3934583067893982, "global_step": 24420, "epoch": 274, "lr": 9.808911641724591e-05}
{"train_loss": 0.3164256513118744, "global_step": 24421, "epoch": 274, "lr": 9.808895767333092e-05}
{"train_loss": 0.6026468276977539, "global_step": 24422, "epoch": 274, "lr": 9.808879892295096e-05}
{"train_loss": 0.4876053035259247, "global_step": 24423, "epoch": 274, "lr": 9.808864016610605e-05}
{"train_loss": 0.4637639820575714, "global_step": 24424, "epoch": 274, "lr": 9.808848140279621e-05}
{"train_loss": 0.4114372134208679, "global_step": 24425, "epoch": 274, "lr": 9.808832263302145e-05}
{"train_loss": 0.2984863817691803, "global_step": 24426, "epoch": 274, "lr": 9.808816385678181e-05}
{"train_loss": 0.3600939214229584, "global_step": 24427, "epoch": 274, "lr": 9.80880050740773e-05}
{"train_loss": 0.43158388137817383, "global_step": 24428, "epoch": 274, "lr": 9.808784628490795e-05}
{"train_loss": 0.4108623266220093, "global_step": 24429, "epoch": 274, "lr": 9.808768748927376e-05}
{"train_loss": 0.5326269865036011, "global_step": 24430, "epoch": 274, "lr": 9.808752868717478e-05}
{"train_loss": 0.4987698495388031, "global_step": 24431, "epoch": 274, "lr": 9.808736987861101e-05}
{"train_loss": 0.45205211639404297, "global_step": 24432, "epoch": 274, "lr": 9.808721106358249e-05}
{"train_loss": 0.5156692266464233, "global_step": 24433, "epoch": 274, "lr": 9.808705224208922e-05}
{"train_loss": 0.5292689204216003, "global_step": 24434, "epoch": 274, "lr": 9.808689341413123e-05}
{"train_loss": 0.4525339901447296, "global_step": 24435, "epoch": 274, "lr": 9.808673457970855e-05}
{"train_loss": 0.37935787439346313, "global_step": 24436, "epoch": 274, "lr": 9.808657573882118e-05}
{"train_loss": 0.5010470151901245, "global_step": 24437, "epoch": 274, "lr": 9.808641689146918e-05}
{"train_loss": 0.382461279630661, "global_step": 24438, "epoch": 274, "lr": 9.808625803765255e-05}
{"train_loss": 0.43068915605545044, "global_step": 24439, "epoch": 274, "lr": 9.808609917737128e-05}
{"train_loss": 0.4703844487667084, "global_step": 24440, "epoch": 274, "lr": 9.808594031062543e-05}
{"train_loss": 0.34395602345466614, "global_step": 24441, "epoch": 274, "lr": 9.808578143741503e-05}
{"train_loss": 0.42046216130256653, "global_step": 24442, "epoch": 274, "lr": 9.808562255774007e-05}
{"train_loss": 0.4711834192276001, "global_step": 24443, "epoch": 274, "lr": 9.808546367160058e-05}
{"train_loss": 0.37845876812934875, "global_step": 24444, "epoch": 274, "lr": 9.808530477899658e-05}
{"train_loss": 0.449647456407547, "global_step": 24445, "epoch": 274, "lr": 9.808514587992811e-05}
{"train_loss": 0.46328437328338623, "global_step": 24446, "epoch": 274, "lr": 9.808498697439517e-05}
{"train_loss": 0.3996962606906891, "global_step": 24447, "epoch": 274, "lr": 9.80848280623978e-05}
{"train_loss": 0.37936943769454956, "global_step": 24448, "epoch": 274, "lr": 9.8084669143936e-05}
{"train_loss": 0.43736353516578674, "global_step": 24449, "epoch": 274, "lr": 9.80845102190098e-05}
{"train_loss": 0.5416233539581299, "global_step": 24450, "epoch": 274, "lr": 9.808435128761924e-05}
{"train_loss": 0.3735133409500122, "global_step": 24451, "epoch": 274, "lr": 9.808419234976431e-05}
{"train_loss": 0.49772530794143677, "global_step": 24452, "epoch": 274, "lr": 9.808403340544505e-05}
{"train_loss": 0.693597674369812, "global_step": 24453, "epoch": 274, "lr": 9.808387445466149e-05}
{"train_loss": 0.46153944730758667, "global_step": 24454, "epoch": 274, "lr": 9.808371549741363e-05}
{"train_loss": 0.3558402359485626, "global_step": 24455, "epoch": 274, "lr": 9.808355653370149e-05}
{"train_loss": 0.47658637166023254, "global_step": 24456, "epoch": 274, "lr": 9.808339756352511e-05}
{"train_loss": 0.5123658776283264, "global_step": 24457, "epoch": 274, "lr": 9.80832385868845e-05}
{"train_loss": 0.45164135098457336, "global_step": 24458, "epoch": 274, "lr": 9.808307960377969e-05}
{"train_loss": 0.5540543794631958, "global_step": 24459, "epoch": 274, "lr": 9.80829206142107e-05}
{"train_loss": 0.3517216444015503, "global_step": 24460, "epoch": 274, "lr": 9.808276161817754e-05}
{"train_loss": 0.4153728783130646, "global_step": 24461, "epoch": 274, "lr": 9.808260261568026e-05}
{"train_loss": 0.46826836466789246, "global_step": 24462, "epoch": 274, "lr": 9.808244360671884e-05}
{"train_loss": 0.4994332790374756, "global_step": 24463, "epoch": 274, "lr": 9.808228459129332e-05}
{"train_loss": 0.5318647027015686, "global_step": 24464, "epoch": 274, "lr": 9.808212556940371e-05}
{"train_loss": 0.41008326411247253, "global_step": 24465, "epoch": 274, "lr": 9.808196654105006e-05}
{"train_loss": 0.4092422425746918, "global_step": 24466, "epoch": 274, "lr": 9.808180750623238e-05}
{"train_loss": 0.256557434797287, "global_step": 24467, "epoch": 274, "lr": 9.808164846495068e-05}
{"train_loss": 0.3653188943862915, "global_step": 24468, "epoch": 274, "lr": 9.808148941720499e-05}
{"train_loss": 0.2997063100337982, "global_step": 24469, "epoch": 274, "lr": 9.808133036299534e-05}
{"train_loss": 0.44111427664756775, "global_step": 24470, "epoch": 274, "lr": 9.808117130232173e-05}
{"train_loss": 0.594566285610199, "global_step": 24471, "epoch": 274, "lr": 9.808101223518419e-05}
{"train_loss": 0.39642560482025146, "global_step": 24472, "epoch": 274, "lr": 9.808085316158276e-05}
{"train_loss": 0.4503058195114136, "global_step": 24473, "epoch": 274, "lr": 9.808069408151742e-05}
{"train_loss": 0.43421474132644994, "global_step": 24474, "epoch": 274, "lr": 9.808053499498823e-05, "val_loss": 1.9649925231933594}
{"train_loss": 0.4144705533981323, "global_step": 24475, "epoch": 275, "lr": 9.808037590199521e-05}
{"train_loss": 0.43344950675964355, "global_step": 24476, "epoch": 275, "lr": 9.808021680253835e-05}
{"train_loss": 0.3468114137649536, "global_step": 24477, "epoch": 275, "lr": 9.80800576966177e-05}
{"train_loss": 0.3322424292564392, "global_step": 24478, "epoch": 275, "lr": 9.807989858423326e-05}
{"train_loss": 0.41123947501182556, "global_step": 24479, "epoch": 275, "lr": 9.807973946538509e-05}
{"train_loss": 0.3939565420150757, "global_step": 24480, "epoch": 275, "lr": 9.807958034007316e-05}
{"train_loss": 0.4306301474571228, "global_step": 24481, "epoch": 275, "lr": 9.807942120829752e-05}
{"train_loss": 0.3930210769176483, "global_step": 24482, "epoch": 275, "lr": 9.80792620700582e-05}
{"train_loss": 0.46886664628982544, "global_step": 24483, "epoch": 275, "lr": 9.80791029253552e-05}
{"train_loss": 0.435740202665329, "global_step": 24484, "epoch": 275, "lr": 9.807894377418855e-05}
{"train_loss": 0.4074454605579376, "global_step": 24485, "epoch": 275, "lr": 9.807878461655828e-05}
{"train_loss": 0.3766489624977112, "global_step": 24486, "epoch": 275, "lr": 9.80786254524644e-05}
{"train_loss": 0.4688766598701477, "global_step": 24487, "epoch": 275, "lr": 9.807846628190693e-05}
{"train_loss": 0.4645330607891083, "global_step": 24488, "epoch": 275, "lr": 9.80783071048859e-05}
{"train_loss": 0.43359506130218506, "global_step": 24489, "epoch": 275, "lr": 9.807814792140133e-05}
{"train_loss": 0.3172614574432373, "global_step": 24490, "epoch": 275, "lr": 9.807798873145324e-05}
{"train_loss": 0.41963279247283936, "global_step": 24491, "epoch": 275, "lr": 9.807782953504165e-05}
{"train_loss": 0.38957297801971436, "global_step": 24492, "epoch": 275, "lr": 9.807767033216657e-05}
{"train_loss": 0.5288033485412598, "global_step": 24493, "epoch": 275, "lr": 9.807751112282805e-05}
{"train_loss": 0.3767845332622528, "global_step": 24494, "epoch": 275, "lr": 9.807735190702609e-05}
{"train_loss": 0.5102270841598511, "global_step": 24495, "epoch": 275, "lr": 9.807719268476071e-05}
{"train_loss": 0.3237093687057495, "global_step": 24496, "epoch": 275, "lr": 9.807703345603195e-05}
{"train_loss": 0.44875368475914, "global_step": 24497, "epoch": 275, "lr": 9.80768742208398e-05}
{"train_loss": 0.45460250973701477, "global_step": 24498, "epoch": 275, "lr": 9.807671497918431e-05}
{"train_loss": 0.39497581124305725, "global_step": 24499, "epoch": 275, "lr": 9.807655573106548e-05}
{"train_loss": 0.48653361201286316, "global_step": 24500, "epoch": 275, "lr": 9.807639647648337e-05}
{"train_loss": 0.48556405305862427, "global_step": 24501, "epoch": 275, "lr": 9.807623721543795e-05}
{"train_loss": 0.47158533334732056, "global_step": 24502, "epoch": 275, "lr": 9.807607794792928e-05}
{"train_loss": 0.47881197929382324, "global_step": 24503, "epoch": 275, "lr": 9.807591867395735e-05}
{"train_loss": 0.42262130975723267, "global_step": 24504, "epoch": 275, "lr": 9.807575939352223e-05}
{"train_loss": 0.4162370562553406, "global_step": 24505, "epoch": 275, "lr": 9.807560010662388e-05}
{"train_loss": 0.47283247113227844, "global_step": 24506, "epoch": 275, "lr": 9.807544081326237e-05}
{"train_loss": 0.5304511785507202, "global_step": 24507, "epoch": 275, "lr": 9.807528151343768e-05}
{"train_loss": 0.5330657362937927, "global_step": 24508, "epoch": 275, "lr": 9.807512220714988e-05}
{"train_loss": 0.4099713861942291, "global_step": 24509, "epoch": 275, "lr": 9.807496289439895e-05}
{"train_loss": 0.5004487633705139, "global_step": 24510, "epoch": 275, "lr": 9.807480357518493e-05}
{"train_loss": 0.33072346448898315, "global_step": 24511, "epoch": 275, "lr": 9.807464424950784e-05}
{"train_loss": 0.4262818694114685, "global_step": 24512, "epoch": 275, "lr": 9.80744849173677e-05}
{"train_loss": 0.4671667814254761, "global_step": 24513, "epoch": 275, "lr": 9.807432557876453e-05}
{"train_loss": 0.41071799397468567, "global_step": 24514, "epoch": 275, "lr": 9.807416623369837e-05}
{"train_loss": 0.5477861762046814, "global_step": 24515, "epoch": 275, "lr": 9.80740068821692e-05}
{"train_loss": 0.42012014985084534, "global_step": 24516, "epoch": 275, "lr": 9.807384752417708e-05}
{"train_loss": 0.4420245289802551, "global_step": 24517, "epoch": 275, "lr": 9.8073688159722e-05}
{"train_loss": 0.41848957538604736, "global_step": 24518, "epoch": 275, "lr": 9.807352878880403e-05}
{"train_loss": 0.487494558095932, "global_step": 24519, "epoch": 275, "lr": 9.807336941142313e-05}
{"train_loss": 0.4565293788909912, "global_step": 24520, "epoch": 275, "lr": 9.807321002757936e-05}
{"train_loss": 0.46333229541778564, "global_step": 24521, "epoch": 275, "lr": 9.807305063727273e-05}
{"train_loss": 0.398648202419281, "global_step": 24522, "epoch": 275, "lr": 9.807289124050328e-05}
{"train_loss": 0.46619242429733276, "global_step": 24523, "epoch": 275, "lr": 9.807273183727099e-05}
{"train_loss": 0.3283419609069824, "global_step": 24524, "epoch": 275, "lr": 9.807257242757593e-05}
{"train_loss": 0.39919084310531616, "global_step": 24525, "epoch": 275, "lr": 9.80724130114181e-05}
{"train_loss": 0.3554530143737793, "global_step": 24526, "epoch": 275, "lr": 9.807225358879751e-05}
{"train_loss": 0.38287970423698425, "global_step": 24527, "epoch": 275, "lr": 9.807209415971418e-05}
{"train_loss": 0.587699294090271, "global_step": 24528, "epoch": 275, "lr": 9.807193472416816e-05}
{"train_loss": 0.4402042031288147, "global_step": 24529, "epoch": 275, "lr": 9.807177528215945e-05}
{"train_loss": 0.42953920364379883, "global_step": 24530, "epoch": 275, "lr": 9.807161583368807e-05}
{"train_loss": 0.3738415241241455, "global_step": 24531, "epoch": 275, "lr": 9.807145637875406e-05}
{"train_loss": 0.2887963354587555, "global_step": 24532, "epoch": 275, "lr": 9.807129691735742e-05}
{"train_loss": 0.2769151031970978, "global_step": 24533, "epoch": 275, "lr": 9.807113744949818e-05}
{"train_loss": 0.37445053458213806, "global_step": 24534, "epoch": 275, "lr": 9.807097797517637e-05}
{"train_loss": 0.447453111410141, "global_step": 24535, "epoch": 275, "lr": 9.807081849439199e-05}
{"train_loss": 0.4121803641319275, "global_step": 24536, "epoch": 275, "lr": 9.807065900714507e-05}
{"train_loss": 0.37014567852020264, "global_step": 24537, "epoch": 275, "lr": 9.807049951343565e-05}
{"train_loss": 0.4153006076812744, "global_step": 24538, "epoch": 275, "lr": 9.807034001326373e-05}
{"train_loss": 0.4630182087421417, "global_step": 24539, "epoch": 275, "lr": 9.807018050662935e-05}
{"train_loss": 0.48131656646728516, "global_step": 24540, "epoch": 275, "lr": 9.80700209935325e-05}
{"train_loss": 0.47512465715408325, "global_step": 24541, "epoch": 275, "lr": 9.806986147397324e-05}
{"train_loss": 0.40222057700157166, "global_step": 24542, "epoch": 275, "lr": 9.806970194795156e-05}
{"train_loss": 0.43427857756614685, "global_step": 24543, "epoch": 275, "lr": 9.806954241546751e-05}
{"train_loss": 0.3879328966140747, "global_step": 24544, "epoch": 275, "lr": 9.806938287652108e-05}
{"train_loss": 0.3447239398956299, "global_step": 24545, "epoch": 275, "lr": 9.806922333111232e-05}
{"train_loss": 0.3767612874507904, "global_step": 24546, "epoch": 275, "lr": 9.806906377924123e-05}
{"train_loss": 0.4931723177433014, "global_step": 24547, "epoch": 275, "lr": 9.806890422090784e-05}
{"train_loss": 0.4176865816116333, "global_step": 24548, "epoch": 275, "lr": 9.806874465611217e-05}
{"train_loss": 0.29991310834884644, "global_step": 24549, "epoch": 275, "lr": 9.806858508485426e-05}
{"train_loss": 0.43961820006370544, "global_step": 24550, "epoch": 275, "lr": 9.80684255071341e-05}
{"train_loss": 0.4412567913532257, "global_step": 24551, "epoch": 275, "lr": 9.806826592295173e-05}
{"train_loss": 0.42571502923965454, "global_step": 24552, "epoch": 275, "lr": 9.806810633230716e-05}
{"train_loss": 0.4717314541339874, "global_step": 24553, "epoch": 275, "lr": 9.806794673520043e-05}
{"train_loss": 0.4379165470600128, "global_step": 24554, "epoch": 275, "lr": 9.806778713163154e-05}
{"train_loss": 0.4431706964969635, "global_step": 24555, "epoch": 275, "lr": 9.806762752160053e-05}
{"train_loss": 0.4496091604232788, "global_step": 24556, "epoch": 275, "lr": 9.806746790510741e-05}
{"train_loss": 0.38646480441093445, "global_step": 24557, "epoch": 275, "lr": 9.80673082821522e-05}
{"train_loss": 0.39251279830932617, "global_step": 24558, "epoch": 275, "lr": 9.806714865273495e-05}
{"train_loss": 0.3990708887577057, "global_step": 24559, "epoch": 275, "lr": 9.806698901685563e-05}
{"train_loss": 0.4272899329662323, "global_step": 24560, "epoch": 275, "lr": 9.806682937451429e-05}
{"train_loss": 0.6014480590820312, "global_step": 24561, "epoch": 275, "lr": 9.806666972571097e-05}
{"train_loss": 0.429461270570755, "global_step": 24562, "epoch": 275, "lr": 9.806651007044564e-05}
{"train_loss": 0.4260085593448596, "global_step": 24563, "epoch": 275, "lr": 9.806635040871839e-05, "val_loss": 2.0654184818267822, "train_action_mse_error": 23.517139434814453}
{"train_loss": 0.4399598240852356, "global_step": 24564, "epoch": 276, "lr": 9.806619074052919e-05}
{"train_loss": 0.4481395184993744, "global_step": 24565, "epoch": 276, "lr": 9.806603106587807e-05}
{"train_loss": 0.34520596265792847, "global_step": 24566, "epoch": 276, "lr": 9.806587138476506e-05}
{"train_loss": 0.36673039197921753, "global_step": 24567, "epoch": 276, "lr": 9.806571169719018e-05}
{"train_loss": 0.6576173901557922, "global_step": 24568, "epoch": 276, "lr": 9.806555200315346e-05}
{"train_loss": 0.35811901092529297, "global_step": 24569, "epoch": 276, "lr": 9.806539230265491e-05}
{"train_loss": 0.40814897418022156, "global_step": 24570, "epoch": 276, "lr": 9.806523259569455e-05}
{"train_loss": 0.4124308228492737, "global_step": 24571, "epoch": 276, "lr": 9.806507288227242e-05}
{"train_loss": 0.4738005995750427, "global_step": 24572, "epoch": 276, "lr": 9.806491316238851e-05}
{"train_loss": 0.38935816287994385, "global_step": 24573, "epoch": 276, "lr": 9.806475343604286e-05}
{"train_loss": 0.43395259976387024, "global_step": 24574, "epoch": 276, "lr": 9.806459370323549e-05}
{"train_loss": 0.33436891436576843, "global_step": 24575, "epoch": 276, "lr": 9.806443396396642e-05}
{"train_loss": 0.4594433903694153, "global_step": 24576, "epoch": 276, "lr": 9.806427421823567e-05}
{"train_loss": 0.39665403962135315, "global_step": 24577, "epoch": 276, "lr": 9.806411446604327e-05}
{"train_loss": 0.33400267362594604, "global_step": 24578, "epoch": 276, "lr": 9.806395470738923e-05}
{"train_loss": 0.42866456508636475, "global_step": 24579, "epoch": 276, "lr": 9.806379494227357e-05}
{"train_loss": 0.5326327681541443, "global_step": 24580, "epoch": 276, "lr": 9.806363517069634e-05}
{"train_loss": 0.4523264467716217, "global_step": 24581, "epoch": 276, "lr": 9.806347539265754e-05}
{"train_loss": 0.33583498001098633, "global_step": 24582, "epoch": 276, "lr": 9.806331560815718e-05}
{"train_loss": 0.45335879921913147, "global_step": 24583, "epoch": 276, "lr": 9.80631558171953e-05}
{"train_loss": 0.45362523198127747, "global_step": 24584, "epoch": 276, "lr": 9.80629960197719e-05}
{"train_loss": 0.3978877365589142, "global_step": 24585, "epoch": 276, "lr": 9.806283621588703e-05}
{"train_loss": 0.4418487250804901, "global_step": 24586, "epoch": 276, "lr": 9.806267640554069e-05}
{"train_loss": 0.46280962228775024, "global_step": 24587, "epoch": 276, "lr": 9.806251658873292e-05}
{"train_loss": 0.555160641670227, "global_step": 24588, "epoch": 276, "lr": 9.806235676546371e-05}
{"train_loss": 0.32470566034317017, "global_step": 24589, "epoch": 276, "lr": 9.806219693573311e-05}
{"train_loss": 0.3642981946468353, "global_step": 24590, "epoch": 276, "lr": 9.806203709954114e-05}
{"train_loss": 0.4297490417957306, "global_step": 24591, "epoch": 276, "lr": 9.806187725688783e-05}
{"train_loss": 0.39573127031326294, "global_step": 24592, "epoch": 276, "lr": 9.806171740777316e-05}
{"train_loss": 0.32166391611099243, "global_step": 24593, "epoch": 276, "lr": 9.80615575521972e-05}
{"train_loss": 0.38502001762390137, "global_step": 24594, "epoch": 276, "lr": 9.806139769015992e-05}
{"train_loss": 0.4040926396846771, "global_step": 24595, "epoch": 276, "lr": 9.80612378216614e-05}
{"train_loss": 0.3358956277370453, "global_step": 24596, "epoch": 276, "lr": 9.80610779467016e-05}
{"train_loss": 0.3731667995452881, "global_step": 24597, "epoch": 276, "lr": 9.806091806528061e-05}
{"train_loss": 0.40314289927482605, "global_step": 24598, "epoch": 276, "lr": 9.806075817739839e-05}
{"train_loss": 0.5110313296318054, "global_step": 24599, "epoch": 276, "lr": 9.8060598283055e-05}
{"train_loss": 0.38172709941864014, "global_step": 24600, "epoch": 276, "lr": 9.806043838225044e-05}
{"train_loss": 0.44933372735977173, "global_step": 24601, "epoch": 276, "lr": 9.806027847498475e-05}
{"train_loss": 0.3802448809146881, "global_step": 24602, "epoch": 276, "lr": 9.806011856125793e-05}
{"train_loss": 0.4557001292705536, "global_step": 24603, "epoch": 276, "lr": 9.805995864107001e-05}
{"train_loss": 0.520287811756134, "global_step": 24604, "epoch": 276, "lr": 9.805979871442104e-05}
{"train_loss": 0.31896331906318665, "global_step": 24605, "epoch": 276, "lr": 9.805963878131099e-05}
{"train_loss": 0.4845636487007141, "global_step": 24606, "epoch": 276, "lr": 9.805947884173992e-05}
{"train_loss": 0.31929880380630493, "global_step": 24607, "epoch": 276, "lr": 9.805931889570783e-05}
{"train_loss": 0.4175911843776703, "global_step": 24608, "epoch": 276, "lr": 9.805915894321475e-05}
{"train_loss": 0.49079763889312744, "global_step": 24609, "epoch": 276, "lr": 9.805899898426071e-05}
{"train_loss": 0.44015422463417053, "global_step": 24610, "epoch": 276, "lr": 9.805883901884572e-05}
{"train_loss": 0.5097982883453369, "global_step": 24611, "epoch": 276, "lr": 9.805867904696979e-05}
{"train_loss": 0.35394707322120667, "global_step": 24612, "epoch": 276, "lr": 9.805851906863298e-05}
{"train_loss": 0.3582718074321747, "global_step": 24613, "epoch": 276, "lr": 9.805835908383527e-05}
{"train_loss": 0.3334692418575287, "global_step": 24614, "epoch": 276, "lr": 9.805819909257672e-05}
{"train_loss": 0.3967345356941223, "global_step": 24615, "epoch": 276, "lr": 9.805803909485732e-05}
{"train_loss": 0.47173774242401123, "global_step": 24616, "epoch": 276, "lr": 9.80578790906771e-05}
{"train_loss": 0.5275610685348511, "global_step": 24617, "epoch": 276, "lr": 9.805771908003608e-05}
{"train_loss": 0.4425446391105652, "global_step": 24618, "epoch": 276, "lr": 9.80575590629343e-05}
{"train_loss": 0.44433748722076416, "global_step": 24619, "epoch": 276, "lr": 9.805739903937175e-05}
{"train_loss": 0.46037915349006653, "global_step": 24620, "epoch": 276, "lr": 9.805723900934847e-05}
{"train_loss": 0.43425044417381287, "global_step": 24621, "epoch": 276, "lr": 9.805707897286449e-05}
{"train_loss": 0.5381143093109131, "global_step": 24622, "epoch": 276, "lr": 9.805691892991981e-05}
{"train_loss": 0.3499408960342407, "global_step": 24623, "epoch": 276, "lr": 9.805675888051448e-05}
{"train_loss": 0.4324711561203003, "global_step": 24624, "epoch": 276, "lr": 9.805659882464849e-05}
{"train_loss": 0.44068843126296997, "global_step": 24625, "epoch": 276, "lr": 9.805643876232188e-05}
{"train_loss": 0.3473404049873352, "global_step": 24626, "epoch": 276, "lr": 9.805627869353468e-05}
{"train_loss": 0.3473600447177887, "global_step": 24627, "epoch": 276, "lr": 9.805611861828689e-05}
{"train_loss": 0.4131016731262207, "global_step": 24628, "epoch": 276, "lr": 9.805595853657853e-05}
{"train_loss": 0.3880596458911896, "global_step": 24629, "epoch": 276, "lr": 9.805579844840964e-05}
{"train_loss": 0.3895246088504791, "global_step": 24630, "epoch": 276, "lr": 9.805563835378023e-05}
{"train_loss": 0.39300185441970825, "global_step": 24631, "epoch": 276, "lr": 9.805547825269033e-05}
{"train_loss": 0.34696492552757263, "global_step": 24632, "epoch": 276, "lr": 9.805531814513995e-05}
{"train_loss": 0.3082534968852997, "global_step": 24633, "epoch": 276, "lr": 9.805515803112914e-05}
{"train_loss": 0.45987439155578613, "global_step": 24634, "epoch": 276, "lr": 9.805499791065788e-05}
{"train_loss": 0.391961008310318, "global_step": 24635, "epoch": 276, "lr": 9.805483778372622e-05}
{"train_loss": 0.3181149959564209, "global_step": 24636, "epoch": 276, "lr": 9.805467765033416e-05}
{"train_loss": 0.4656389653682709, "global_step": 24637, "epoch": 276, "lr": 9.805451751048174e-05}
{"train_loss": 0.5684303045272827, "global_step": 24638, "epoch": 276, "lr": 9.805435736416898e-05}
{"train_loss": 0.4301363229751587, "global_step": 24639, "epoch": 276, "lr": 9.805419721139589e-05}
{"train_loss": 0.5921786427497864, "global_step": 24640, "epoch": 276, "lr": 9.80540370521625e-05}
{"train_loss": 0.4901147484779358, "global_step": 24641, "epoch": 276, "lr": 9.805387688646884e-05}
{"train_loss": 0.45175841450691223, "global_step": 24642, "epoch": 276, "lr": 9.805371671431491e-05}
{"train_loss": 0.481602281332016, "global_step": 24643, "epoch": 276, "lr": 9.805355653570076e-05}
{"train_loss": 0.3814238905906677, "global_step": 24644, "epoch": 276, "lr": 9.805339635062638e-05}
{"train_loss": 0.4928698241710663, "global_step": 24645, "epoch": 276, "lr": 9.805323615909181e-05}
{"train_loss": 0.3399078845977783, "global_step": 24646, "epoch": 276, "lr": 9.805307596109708e-05}
{"train_loss": 0.47868940234184265, "global_step": 24647, "epoch": 276, "lr": 9.805291575664218e-05}
{"train_loss": 0.3580452501773834, "global_step": 24648, "epoch": 276, "lr": 9.805275554572715e-05}
{"train_loss": 0.5443211793899536, "global_step": 24649, "epoch": 276, "lr": 9.805259532835202e-05}
{"train_loss": 0.37792572379112244, "global_step": 24650, "epoch": 276, "lr": 9.805243510451683e-05}
{"train_loss": 0.4889889359474182, "global_step": 24651, "epoch": 276, "lr": 9.805227487422154e-05}
{"train_loss": 0.42326377683810973, "global_step": 24652, "epoch": 276, "lr": 9.805211463746622e-05, "val_loss": 1.9647506475448608}
{"train_loss": 0.40484318137168884, "global_step": 24653, "epoch": 277, "lr": 9.805195439425087e-05}
{"train_loss": 0.4118526577949524, "global_step": 24654, "epoch": 277, "lr": 9.805179414457555e-05}
{"train_loss": 0.5039042830467224, "global_step": 24655, "epoch": 277, "lr": 9.805163388844023e-05}
{"train_loss": 0.46772971749305725, "global_step": 24656, "epoch": 277, "lr": 9.805147362584495e-05}
{"train_loss": 0.3851218521595001, "global_step": 24657, "epoch": 277, "lr": 9.805131335678975e-05}
{"train_loss": 0.4960075318813324, "global_step": 24658, "epoch": 277, "lr": 9.805115308127461e-05}
{"train_loss": 0.3792038559913635, "global_step": 24659, "epoch": 277, "lr": 9.80509927992996e-05}
{"train_loss": 0.4198366701602936, "global_step": 24660, "epoch": 277, "lr": 9.805083251086471e-05}
{"train_loss": 0.4524981677532196, "global_step": 24661, "epoch": 277, "lr": 9.805067221596998e-05}
{"train_loss": 0.4281883239746094, "global_step": 24662, "epoch": 277, "lr": 9.805051191461542e-05}
{"train_loss": 0.41455531120300293, "global_step": 24663, "epoch": 277, "lr": 9.805035160680105e-05}
{"train_loss": 0.4570777118206024, "global_step": 24664, "epoch": 277, "lr": 9.80501912925269e-05}
{"train_loss": 0.3068178594112396, "global_step": 24665, "epoch": 277, "lr": 9.805003097179299e-05}
{"train_loss": 0.43008747696876526, "global_step": 24666, "epoch": 277, "lr": 9.804987064459933e-05}
{"train_loss": 0.42767807841300964, "global_step": 24667, "epoch": 277, "lr": 9.804971031094595e-05}
{"train_loss": 0.4394443929195404, "global_step": 24668, "epoch": 277, "lr": 9.804954997083288e-05}
{"train_loss": 0.3621075451374054, "global_step": 24669, "epoch": 277, "lr": 9.804938962426014e-05}
{"train_loss": 0.43097636103630066, "global_step": 24670, "epoch": 277, "lr": 9.804922927122773e-05}
{"train_loss": 0.33437052369117737, "global_step": 24671, "epoch": 277, "lr": 9.804906891173569e-05}
{"train_loss": 0.4278818666934967, "global_step": 24672, "epoch": 277, "lr": 9.804890854578405e-05}
{"train_loss": 0.3816286325454712, "global_step": 24673, "epoch": 277, "lr": 9.80487481733728e-05}
{"train_loss": 0.39365431666374207, "global_step": 24674, "epoch": 277, "lr": 9.804858779450199e-05}
{"train_loss": 0.3942158818244934, "global_step": 24675, "epoch": 277, "lr": 9.804842740917163e-05}
{"train_loss": 0.4711231291294098, "global_step": 24676, "epoch": 277, "lr": 9.804826701738176e-05}
{"train_loss": 0.45051029324531555, "global_step": 24677, "epoch": 277, "lr": 9.804810661913238e-05}
{"train_loss": 0.4473751485347748, "global_step": 24678, "epoch": 277, "lr": 9.804794621442351e-05}
{"train_loss": 0.5210161805152893, "global_step": 24679, "epoch": 277, "lr": 9.80477858032552e-05}
{"train_loss": 0.3816007673740387, "global_step": 24680, "epoch": 277, "lr": 9.804762538562742e-05}
{"train_loss": 0.4741956293582916, "global_step": 24681, "epoch": 277, "lr": 9.804746496154025e-05}
{"train_loss": 0.40021681785583496, "global_step": 24682, "epoch": 277, "lr": 9.804730453099366e-05}
{"train_loss": 0.4440390169620514, "global_step": 24683, "epoch": 277, "lr": 9.80471440939877e-05}
{"train_loss": 0.37224268913269043, "global_step": 24684, "epoch": 277, "lr": 9.80469836505224e-05}
{"train_loss": 0.3294277489185333, "global_step": 24685, "epoch": 277, "lr": 9.804682320059777e-05}
{"train_loss": 0.3639267683029175, "global_step": 24686, "epoch": 277, "lr": 9.804666274421381e-05}
{"train_loss": 0.4587211012840271, "global_step": 24687, "epoch": 277, "lr": 9.804650228137058e-05}
{"train_loss": 0.5058065056800842, "global_step": 24688, "epoch": 277, "lr": 9.804634181206806e-05}
{"train_loss": 0.4210091531276703, "global_step": 24689, "epoch": 277, "lr": 9.804618133630633e-05}
{"train_loss": 0.32775890827178955, "global_step": 24690, "epoch": 277, "lr": 9.804602085408536e-05}
{"train_loss": 0.4786648750305176, "global_step": 24691, "epoch": 277, "lr": 9.804586036540517e-05}
{"train_loss": 0.41461265087127686, "global_step": 24692, "epoch": 277, "lr": 9.804569987026583e-05}
{"train_loss": 0.41440504789352417, "global_step": 24693, "epoch": 277, "lr": 9.804553936866731e-05}
{"train_loss": 0.33391377329826355, "global_step": 24694, "epoch": 277, "lr": 9.804537886060967e-05}
{"train_loss": 0.4161819815635681, "global_step": 24695, "epoch": 277, "lr": 9.80452183460929e-05}
{"train_loss": 0.4347057044506073, "global_step": 24696, "epoch": 277, "lr": 9.804505782511704e-05}
{"train_loss": 0.4197988212108612, "global_step": 24697, "epoch": 277, "lr": 9.804489729768212e-05}
{"train_loss": 0.37325140833854675, "global_step": 24698, "epoch": 277, "lr": 9.804473676378812e-05}
{"train_loss": 0.400630921125412, "global_step": 24699, "epoch": 277, "lr": 9.804457622343511e-05}
{"train_loss": 0.44966644048690796, "global_step": 24700, "epoch": 277, "lr": 9.80444156766231e-05}
{"train_loss": 0.4024047255516052, "global_step": 24701, "epoch": 277, "lr": 9.804425512335209e-05}
{"train_loss": 0.3791770935058594, "global_step": 24702, "epoch": 277, "lr": 9.804409456362211e-05}
{"train_loss": 0.4079931378364563, "global_step": 24703, "epoch": 277, "lr": 9.804393399743321e-05}
{"train_loss": 0.4730941951274872, "global_step": 24704, "epoch": 277, "lr": 9.804377342478537e-05}
{"train_loss": 0.3573451638221741, "global_step": 24705, "epoch": 277, "lr": 9.804361284567865e-05}
{"train_loss": 0.37866318225860596, "global_step": 24706, "epoch": 277, "lr": 9.804345226011302e-05}
{"train_loss": 0.42598873376846313, "global_step": 24707, "epoch": 277, "lr": 9.804329166808856e-05}
{"train_loss": 0.3849671185016632, "global_step": 24708, "epoch": 277, "lr": 9.804313106960525e-05}
{"train_loss": 0.5565309524536133, "global_step": 24709, "epoch": 277, "lr": 9.804297046466313e-05}
{"train_loss": 0.5433120131492615, "global_step": 24710, "epoch": 277, "lr": 9.804280985326223e-05}
{"train_loss": 0.4586299657821655, "global_step": 24711, "epoch": 277, "lr": 9.804264923540255e-05}
{"train_loss": 0.4670311510562897, "global_step": 24712, "epoch": 277, "lr": 9.804248861108411e-05}
{"train_loss": 0.4105378985404968, "global_step": 24713, "epoch": 277, "lr": 9.804232798030697e-05}
{"train_loss": 0.45814740657806396, "global_step": 24714, "epoch": 277, "lr": 9.80421673430711e-05}
{"train_loss": 0.42394202947616577, "global_step": 24715, "epoch": 277, "lr": 9.804200669937656e-05}
{"train_loss": 0.44220393896102905, "global_step": 24716, "epoch": 277, "lr": 9.804184604922335e-05}
{"train_loss": 0.5990860462188721, "global_step": 24717, "epoch": 277, "lr": 9.804168539261151e-05}
{"train_loss": 0.514590322971344, "global_step": 24718, "epoch": 277, "lr": 9.804152472954105e-05}
{"train_loss": 0.564915657043457, "global_step": 24719, "epoch": 277, "lr": 9.804136406001199e-05}
{"train_loss": 0.43807560205459595, "global_step": 24720, "epoch": 277, "lr": 9.804120338402434e-05}
{"train_loss": 0.4242880344390869, "global_step": 24721, "epoch": 277, "lr": 9.804104270157815e-05}
{"train_loss": 0.4568548798561096, "global_step": 24722, "epoch": 277, "lr": 9.804088201267342e-05}
{"train_loss": 0.46730732917785645, "global_step": 24723, "epoch": 277, "lr": 9.804072131731018e-05}
{"train_loss": 0.3530001640319824, "global_step": 24724, "epoch": 277, "lr": 9.804056061548846e-05}
{"train_loss": 0.41395220160484314, "global_step": 24725, "epoch": 277, "lr": 9.804039990720826e-05}
{"train_loss": 0.41512057185173035, "global_step": 24726, "epoch": 277, "lr": 9.804023919246961e-05}
{"train_loss": 0.3727150857448578, "global_step": 24727, "epoch": 277, "lr": 9.804007847127256e-05}
{"train_loss": 0.4327503740787506, "global_step": 24728, "epoch": 277, "lr": 9.803991774361709e-05}
{"train_loss": 0.5747889876365662, "global_step": 24729, "epoch": 277, "lr": 9.803975700950323e-05}
{"train_loss": 0.46566635370254517, "global_step": 24730, "epoch": 277, "lr": 9.803959626893103e-05}
{"train_loss": 0.43227270245552063, "global_step": 24731, "epoch": 277, "lr": 9.803943552190049e-05}
{"train_loss": 0.3663662374019623, "global_step": 24732, "epoch": 277, "lr": 9.803927476841161e-05}
{"train_loss": 0.42690375447273254, "global_step": 24733, "epoch": 277, "lr": 9.803911400846447e-05}
{"train_loss": 0.5020992159843445, "global_step": 24734, "epoch": 277, "lr": 9.803895324205903e-05}
{"train_loss": 0.5632919669151306, "global_step": 24735, "epoch": 277, "lr": 9.803879246919534e-05}
{"train_loss": 0.41566240787506104, "global_step": 24736, "epoch": 277, "lr": 9.803863168987342e-05}
{"train_loss": 0.4887124300003052, "global_step": 24737, "epoch": 277, "lr": 9.80384709040933e-05}
{"train_loss": 0.41272562742233276, "global_step": 24738, "epoch": 277, "lr": 9.803831011185498e-05}
{"train_loss": 0.3466934263706207, "global_step": 24739, "epoch": 277, "lr": 9.803814931315852e-05}
{"train_loss": 0.350210040807724, "global_step": 24740, "epoch": 277, "lr": 9.80379885080039e-05}
{"train_loss": 0.4304579235194774, "global_step": 24741, "epoch": 277, "lr": 9.803782769639115e-05, "val_loss": 2.024531841278076}
{"train_loss": 0.3963487446308136, "global_step": 24742, "epoch": 278, "lr": 9.803766687832032e-05}
{"train_loss": 0.43292897939682007, "global_step": 24743, "epoch": 278, "lr": 9.803750605379138e-05}
{"train_loss": 0.47475501894950867, "global_step": 24744, "epoch": 278, "lr": 9.803734522280441e-05}
{"train_loss": 0.41098108887672424, "global_step": 24745, "epoch": 278, "lr": 9.803718438535941e-05}
{"train_loss": 0.47138407826423645, "global_step": 24746, "epoch": 278, "lr": 9.803702354145637e-05}
{"train_loss": 0.37649351358413696, "global_step": 24747, "epoch": 278, "lr": 9.803686269109535e-05}
{"train_loss": 0.3123109042644501, "global_step": 24748, "epoch": 278, "lr": 9.803670183427636e-05}
{"train_loss": 0.34465304017066956, "global_step": 24749, "epoch": 278, "lr": 9.803654097099941e-05}
{"train_loss": 0.46339279413223267, "global_step": 24750, "epoch": 278, "lr": 9.803638010126454e-05}
{"train_loss": 0.4787715673446655, "global_step": 24751, "epoch": 278, "lr": 9.803621922507178e-05}
{"train_loss": 0.424198716878891, "global_step": 24752, "epoch": 278, "lr": 9.803605834242114e-05}
{"train_loss": 0.5664253234863281, "global_step": 24753, "epoch": 278, "lr": 9.803589745331261e-05}
{"train_loss": 0.45016399025917053, "global_step": 24754, "epoch": 278, "lr": 9.803573655774625e-05}
{"train_loss": 0.4404135048389435, "global_step": 24755, "epoch": 278, "lr": 9.803557565572207e-05}
{"train_loss": 0.4813064932823181, "global_step": 24756, "epoch": 278, "lr": 9.80354147472401e-05}
{"train_loss": 0.5245714783668518, "global_step": 24757, "epoch": 278, "lr": 9.803525383230033e-05}
{"train_loss": 0.39567136764526367, "global_step": 24758, "epoch": 278, "lr": 9.803509291090283e-05}
{"train_loss": 0.391087144613266, "global_step": 24759, "epoch": 278, "lr": 9.803493198304758e-05}
{"train_loss": 0.5637573003768921, "global_step": 24760, "epoch": 278, "lr": 9.803477104873464e-05}
{"train_loss": 0.3509249687194824, "global_step": 24761, "epoch": 278, "lr": 9.8034610107964e-05}
{"train_loss": 0.23751135170459747, "global_step": 24762, "epoch": 278, "lr": 9.803444916073568e-05}
{"train_loss": 0.35867440700531006, "global_step": 24763, "epoch": 278, "lr": 9.803428820704973e-05}
{"train_loss": 0.44749322533607483, "global_step": 24764, "epoch": 278, "lr": 9.803412724690614e-05}
{"train_loss": 0.4673134386539459, "global_step": 24765, "epoch": 278, "lr": 9.803396628030497e-05}
{"train_loss": 0.526034951210022, "global_step": 24766, "epoch": 278, "lr": 9.80338053072462e-05}
{"train_loss": 0.43593841791152954, "global_step": 24767, "epoch": 278, "lr": 9.803364432772988e-05}
{"train_loss": 0.46317511796951294, "global_step": 24768, "epoch": 278, "lr": 9.803348334175601e-05}
{"train_loss": 0.47147223353385925, "global_step": 24769, "epoch": 278, "lr": 9.803332234932463e-05}
{"train_loss": 0.33506888151168823, "global_step": 24770, "epoch": 278, "lr": 9.803316135043577e-05}
{"train_loss": 0.37033817172050476, "global_step": 24771, "epoch": 278, "lr": 9.803300034508942e-05}
{"train_loss": 0.539737343788147, "global_step": 24772, "epoch": 278, "lr": 9.803283933328563e-05}
{"train_loss": 0.6038984060287476, "global_step": 24773, "epoch": 278, "lr": 9.80326783150244e-05}
{"train_loss": 0.4416116178035736, "global_step": 24774, "epoch": 278, "lr": 9.803251729030576e-05}
{"train_loss": 0.4343399107456207, "global_step": 24775, "epoch": 278, "lr": 9.803235625912974e-05}
{"train_loss": 0.46918347477912903, "global_step": 24776, "epoch": 278, "lr": 9.803219522149634e-05}
{"train_loss": 0.48515522480010986, "global_step": 24777, "epoch": 278, "lr": 9.803203417740562e-05}
{"train_loss": 0.45100411772727966, "global_step": 24778, "epoch": 278, "lr": 9.803187312685757e-05}
{"train_loss": 0.4565449059009552, "global_step": 24779, "epoch": 278, "lr": 9.803171206985222e-05}
{"train_loss": 0.46068698167800903, "global_step": 24780, "epoch": 278, "lr": 9.803155100638959e-05}
{"train_loss": 0.3668644428253174, "global_step": 24781, "epoch": 278, "lr": 9.80313899364697e-05}
{"train_loss": 0.4108186960220337, "global_step": 24782, "epoch": 278, "lr": 9.803122886009258e-05}
{"train_loss": 0.48427531123161316, "global_step": 24783, "epoch": 278, "lr": 9.803106777725825e-05}
{"train_loss": 0.4325185716152191, "global_step": 24784, "epoch": 278, "lr": 9.803090668796671e-05}
{"train_loss": 0.47108858823776245, "global_step": 24785, "epoch": 278, "lr": 9.803074559221802e-05}
{"train_loss": 0.46977436542510986, "global_step": 24786, "epoch": 278, "lr": 9.803058449001218e-05}
{"train_loss": 0.6112278699874878, "global_step": 24787, "epoch": 278, "lr": 9.803042338134921e-05}
{"train_loss": 0.4177935719490051, "global_step": 24788, "epoch": 278, "lr": 9.803026226622913e-05}
{"train_loss": 0.3913077712059021, "global_step": 24789, "epoch": 278, "lr": 9.803010114465196e-05}
{"train_loss": 0.49071821570396423, "global_step": 24790, "epoch": 278, "lr": 9.802994001661775e-05}
{"train_loss": 0.42692071199417114, "global_step": 24791, "epoch": 278, "lr": 9.802977888212649e-05}
{"train_loss": 0.44804617762565613, "global_step": 24792, "epoch": 278, "lr": 9.80296177411782e-05}
{"train_loss": 0.353956013917923, "global_step": 24793, "epoch": 278, "lr": 9.802945659377293e-05}
{"train_loss": 0.46859535574913025, "global_step": 24794, "epoch": 278, "lr": 9.802929543991069e-05}
{"train_loss": 0.4286705553531647, "global_step": 24795, "epoch": 278, "lr": 9.802913427959147e-05}
{"train_loss": 0.4601307213306427, "global_step": 24796, "epoch": 278, "lr": 9.802897311281533e-05}
{"train_loss": 0.5329850316047668, "global_step": 24797, "epoch": 278, "lr": 9.802881193958229e-05}
{"train_loss": 0.441317081451416, "global_step": 24798, "epoch": 278, "lr": 9.802865075989236e-05}
{"train_loss": 0.5071887969970703, "global_step": 24799, "epoch": 278, "lr": 9.802848957374555e-05}
{"train_loss": 0.5543281435966492, "global_step": 24800, "epoch": 278, "lr": 9.80283283811419e-05}
{"train_loss": 0.4868057370185852, "global_step": 24801, "epoch": 278, "lr": 9.802816718208143e-05}
{"train_loss": 0.4722873568534851, "global_step": 24802, "epoch": 278, "lr": 9.802800597656416e-05}
{"train_loss": 0.33440667390823364, "global_step": 24803, "epoch": 278, "lr": 9.802784476459011e-05}
{"train_loss": 0.33886730670928955, "global_step": 24804, "epoch": 278, "lr": 9.802768354615929e-05}
{"train_loss": 0.4309481382369995, "global_step": 24805, "epoch": 278, "lr": 9.802752232127175e-05}
{"train_loss": 0.39406687021255493, "global_step": 24806, "epoch": 278, "lr": 9.802736108992749e-05}
{"train_loss": 0.47768110036849976, "global_step": 24807, "epoch": 278, "lr": 9.802719985212654e-05}
{"train_loss": 0.3916814625263214, "global_step": 24808, "epoch": 278, "lr": 9.80270386078689e-05}
{"train_loss": 0.3973240554332733, "global_step": 24809, "epoch": 278, "lr": 9.802687735715463e-05}
{"train_loss": 0.32504650950431824, "global_step": 24810, "epoch": 278, "lr": 9.802671609998373e-05}
{"train_loss": 0.42286017537117004, "global_step": 24811, "epoch": 278, "lr": 9.802655483635621e-05}
{"train_loss": 0.415126770734787, "global_step": 24812, "epoch": 278, "lr": 9.802639356627213e-05}
{"train_loss": 0.4387179911136627, "global_step": 24813, "epoch": 278, "lr": 9.802623228973146e-05}
{"train_loss": 0.6000494956970215, "global_step": 24814, "epoch": 278, "lr": 9.802607100673426e-05}
{"train_loss": 0.5004391670227051, "global_step": 24815, "epoch": 278, "lr": 9.802590971728053e-05}
{"train_loss": 0.33394715189933777, "global_step": 24816, "epoch": 278, "lr": 9.802574842137031e-05}
{"train_loss": 0.3771427273750305, "global_step": 24817, "epoch": 278, "lr": 9.802558711900363e-05}
{"train_loss": 0.4628373980522156, "global_step": 24818, "epoch": 278, "lr": 9.802542581018048e-05}
{"train_loss": 0.41439181566238403, "global_step": 24819, "epoch": 278, "lr": 9.802526449490089e-05}
{"train_loss": 0.5252401828765869, "global_step": 24820, "epoch": 278, "lr": 9.802510317316489e-05}
{"train_loss": 0.4419359266757965, "global_step": 24821, "epoch": 278, "lr": 9.80249418449725e-05}
{"train_loss": 0.49609145522117615, "global_step": 24822, "epoch": 278, "lr": 9.802478051032375e-05}
{"train_loss": 0.43113595247268677, "global_step": 24823, "epoch": 278, "lr": 9.802461916921866e-05}
{"train_loss": 0.5122067928314209, "global_step": 24824, "epoch": 278, "lr": 9.802445782165723e-05}
{"train_loss": 0.3401825428009033, "global_step": 24825, "epoch": 278, "lr": 9.80242964676395e-05}
{"train_loss": 0.43974944949150085, "global_step": 24826, "epoch": 278, "lr": 9.802413510716548e-05}
{"train_loss": 0.4944811761379242, "global_step": 24827, "epoch": 278, "lr": 9.802397374023522e-05}
{"train_loss": 0.4499265253543854, "global_step": 24828, "epoch": 278, "lr": 9.802381236684872e-05}
{"train_loss": 0.4294351041316986, "global_step": 24829, "epoch": 278, "lr": 9.8023650987006e-05}
{"train_loss": 0.442646777529395, "global_step": 24830, "epoch": 278, "lr": 9.802348960070707e-05, "val_loss": 1.9442628622055054}
{"train_loss": 0.3983384668827057, "global_step": 24831, "epoch": 279, "lr": 9.802332820795199e-05}
{"train_loss": 0.40911760926246643, "global_step": 24832, "epoch": 279, "lr": 9.802316680874074e-05}
{"train_loss": 0.3311910331249237, "global_step": 24833, "epoch": 279, "lr": 9.802300540307337e-05}
{"train_loss": 0.35412123799324036, "global_step": 24834, "epoch": 279, "lr": 9.80228439909499e-05}
{"train_loss": 0.45654043555259705, "global_step": 24835, "epoch": 279, "lr": 9.802268257237033e-05}
{"train_loss": 0.3642970025539398, "global_step": 24836, "epoch": 279, "lr": 9.80225211473347e-05}
{"train_loss": 0.38320043683052063, "global_step": 24837, "epoch": 279, "lr": 9.802235971584303e-05}
{"train_loss": 0.3868606686592102, "global_step": 24838, "epoch": 279, "lr": 9.802219827789534e-05}
{"train_loss": 0.43829771876335144, "global_step": 24839, "epoch": 279, "lr": 9.802203683349165e-05}
{"train_loss": 0.420648992061615, "global_step": 24840, "epoch": 279, "lr": 9.802187538263197e-05}
{"train_loss": 0.4525740146636963, "global_step": 24841, "epoch": 279, "lr": 9.802171392531635e-05}
{"train_loss": 0.38083264231681824, "global_step": 24842, "epoch": 279, "lr": 9.802155246154479e-05}
{"train_loss": 0.45407798886299133, "global_step": 24843, "epoch": 279, "lr": 9.802139099131731e-05}
{"train_loss": 0.5386362671852112, "global_step": 24844, "epoch": 279, "lr": 9.802122951463394e-05}
{"train_loss": 0.45778459310531616, "global_step": 24845, "epoch": 279, "lr": 9.802106803149472e-05}
{"train_loss": 0.5185359120368958, "global_step": 24846, "epoch": 279, "lr": 9.802090654189964e-05}
{"train_loss": 0.4118175506591797, "global_step": 24847, "epoch": 279, "lr": 9.802074504584873e-05}
{"train_loss": 0.429548978805542, "global_step": 24848, "epoch": 279, "lr": 9.802058354334202e-05}
{"train_loss": 0.36096474528312683, "global_step": 24849, "epoch": 279, "lr": 9.802042203437954e-05}
{"train_loss": 0.6151171922683716, "global_step": 24850, "epoch": 279, "lr": 9.802026051896129e-05}
{"train_loss": 0.39133548736572266, "global_step": 24851, "epoch": 279, "lr": 9.80200989970873e-05}
{"train_loss": 0.4074346721172333, "global_step": 24852, "epoch": 279, "lr": 9.80199374687576e-05}
{"train_loss": 0.34190163016319275, "global_step": 24853, "epoch": 279, "lr": 9.80197759339722e-05}
{"train_loss": 0.31586697697639465, "global_step": 24854, "epoch": 279, "lr": 9.801961439273112e-05}
{"train_loss": 0.4417323172092438, "global_step": 24855, "epoch": 279, "lr": 9.801945284503439e-05}
{"train_loss": 0.3323982059955597, "global_step": 24856, "epoch": 279, "lr": 9.801929129088204e-05}
{"train_loss": 0.4037400782108307, "global_step": 24857, "epoch": 279, "lr": 9.801912973027407e-05}
{"train_loss": 0.4556422531604767, "global_step": 24858, "epoch": 279, "lr": 9.801896816321051e-05}
{"train_loss": 0.37497657537460327, "global_step": 24859, "epoch": 279, "lr": 9.80188065896914e-05}
{"train_loss": 0.5102173686027527, "global_step": 24860, "epoch": 279, "lr": 9.801864500971674e-05}
{"train_loss": 0.5427367091178894, "global_step": 24861, "epoch": 279, "lr": 9.801848342328655e-05}
{"train_loss": 0.41366904973983765, "global_step": 24862, "epoch": 279, "lr": 9.801832183040087e-05}
{"train_loss": 0.5291763544082642, "global_step": 24863, "epoch": 279, "lr": 9.801816023105971e-05}
{"train_loss": 0.38363784551620483, "global_step": 24864, "epoch": 279, "lr": 9.801799862526308e-05}
{"train_loss": 0.5189157724380493, "global_step": 24865, "epoch": 279, "lr": 9.801783701301103e-05}
{"train_loss": 0.3115568161010742, "global_step": 24866, "epoch": 279, "lr": 9.801767539430357e-05}
{"train_loss": 0.5942260026931763, "global_step": 24867, "epoch": 279, "lr": 9.801751376914072e-05}
{"train_loss": 0.27148568630218506, "global_step": 24868, "epoch": 279, "lr": 9.80173521375225e-05}
{"train_loss": 0.5246890187263489, "global_step": 24869, "epoch": 279, "lr": 9.801719049944892e-05}
{"train_loss": 0.3788186013698578, "global_step": 24870, "epoch": 279, "lr": 9.801702885492002e-05}
{"train_loss": 0.3387574851512909, "global_step": 24871, "epoch": 279, "lr": 9.801686720393581e-05}
{"train_loss": 0.4581387937068939, "global_step": 24872, "epoch": 279, "lr": 9.801670554649634e-05}
{"train_loss": 0.3747066855430603, "global_step": 24873, "epoch": 279, "lr": 9.801654388260159e-05}
{"train_loss": 0.4209936261177063, "global_step": 24874, "epoch": 279, "lr": 9.80163822122516e-05}
{"train_loss": 0.45012837648391724, "global_step": 24875, "epoch": 279, "lr": 9.80162205354464e-05}
{"train_loss": 0.39576029777526855, "global_step": 24876, "epoch": 279, "lr": 9.8016058852186e-05}
{"train_loss": 0.5106825828552246, "global_step": 24877, "epoch": 279, "lr": 9.801589716247043e-05}
{"train_loss": 0.4074014723300934, "global_step": 24878, "epoch": 279, "lr": 9.80157354662997e-05}
{"train_loss": 0.3224058747291565, "global_step": 24879, "epoch": 279, "lr": 9.801557376367385e-05}
{"train_loss": 0.43433094024658203, "global_step": 24880, "epoch": 279, "lr": 9.801541205459288e-05}
{"train_loss": 0.39415809512138367, "global_step": 24881, "epoch": 279, "lr": 9.801525033905683e-05}
{"train_loss": 0.40397563576698303, "global_step": 24882, "epoch": 279, "lr": 9.801508861706571e-05}
{"train_loss": 0.38617539405822754, "global_step": 24883, "epoch": 279, "lr": 9.801492688861955e-05}
{"train_loss": 0.5186282396316528, "global_step": 24884, "epoch": 279, "lr": 9.801476515371837e-05}
{"train_loss": 0.3385828733444214, "global_step": 24885, "epoch": 279, "lr": 9.801460341236219e-05}
{"train_loss": 0.4333622455596924, "global_step": 24886, "epoch": 279, "lr": 9.801444166455104e-05}
{"train_loss": 0.3338905870914459, "global_step": 24887, "epoch": 279, "lr": 9.80142799102849e-05}
{"train_loss": 0.5187219381332397, "global_step": 24888, "epoch": 279, "lr": 9.801411814956388e-05}
{"train_loss": 0.31078872084617615, "global_step": 24889, "epoch": 279, "lr": 9.80139563823879e-05}
{"train_loss": 0.423873633146286, "global_step": 24890, "epoch": 279, "lr": 9.801379460875706e-05}
{"train_loss": 0.40912631154060364, "global_step": 24891, "epoch": 279, "lr": 9.801363282867133e-05}
{"train_loss": 0.5112390518188477, "global_step": 24892, "epoch": 279, "lr": 9.801347104213075e-05}
{"train_loss": 0.5050058960914612, "global_step": 24893, "epoch": 279, "lr": 9.801330924913537e-05}
{"train_loss": 0.3600350022315979, "global_step": 24894, "epoch": 279, "lr": 9.801314744968515e-05}
{"train_loss": 0.5074722766876221, "global_step": 24895, "epoch": 279, "lr": 9.801298564378016e-05}
{"train_loss": 0.29544734954833984, "global_step": 24896, "epoch": 279, "lr": 9.801282383142044e-05}
{"train_loss": 0.4582168459892273, "global_step": 24897, "epoch": 279, "lr": 9.801266201260594e-05}
{"train_loss": 0.3684924244880676, "global_step": 24898, "epoch": 279, "lr": 9.801250018733674e-05}
{"train_loss": 0.45720139145851135, "global_step": 24899, "epoch": 279, "lr": 9.801233835561284e-05}
{"train_loss": 0.3879148066043854, "global_step": 24900, "epoch": 279, "lr": 9.801217651743428e-05}
{"train_loss": 0.386125385761261, "global_step": 24901, "epoch": 279, "lr": 9.801201467280106e-05}
{"train_loss": 0.4058261215686798, "global_step": 24902, "epoch": 279, "lr": 9.801185282171319e-05}
{"train_loss": 0.3745400607585907, "global_step": 24903, "epoch": 279, "lr": 9.801169096417073e-05}
{"train_loss": 0.4323645830154419, "global_step": 24904, "epoch": 279, "lr": 9.801152910017369e-05}
{"train_loss": 0.4429820775985718, "global_step": 24905, "epoch": 279, "lr": 9.801136722972208e-05}
{"train_loss": 0.46432116627693176, "global_step": 24906, "epoch": 279, "lr": 9.801120535281592e-05}
{"train_loss": 0.5208103656768799, "global_step": 24907, "epoch": 279, "lr": 9.801104346945524e-05}
{"train_loss": 0.32591843605041504, "global_step": 24908, "epoch": 279, "lr": 9.801088157964006e-05}
{"train_loss": 0.40641340613365173, "global_step": 24909, "epoch": 279, "lr": 9.801071968337041e-05}
{"train_loss": 0.43387743830680847, "global_step": 24910, "epoch": 279, "lr": 9.801055778064629e-05}
{"train_loss": 0.4187619686126709, "global_step": 24911, "epoch": 279, "lr": 9.801039587146775e-05}
{"train_loss": 0.40748751163482666, "global_step": 24912, "epoch": 279, "lr": 9.80102339558348e-05}
{"train_loss": 0.5327414870262146, "global_step": 24913, "epoch": 279, "lr": 9.801007203374745e-05}
{"train_loss": 0.34331637620925903, "global_step": 24914, "epoch": 279, "lr": 9.800991010520573e-05}
{"train_loss": 0.48509353399276733, "global_step": 24915, "epoch": 279, "lr": 9.800974817020967e-05}
{"train_loss": 0.3918290138244629, "global_step": 24916, "epoch": 279, "lr": 9.800958622875928e-05}
{"train_loss": 0.6001960039138794, "global_step": 24917, "epoch": 279, "lr": 9.800942428085458e-05}
{"train_loss": 0.35250091552734375, "global_step": 24918, "epoch": 279, "lr": 9.800926232649561e-05}
{"train_loss": 0.4221715126814467, "global_step": 24919, "epoch": 279, "lr": 9.800910036568238e-05, "val_loss": 1.9964064359664917}
{"train_loss": 0.4392947852611542, "global_step": 24920, "epoch": 280, "lr": 9.80089383984149e-05}
{"train_loss": 0.39971262216567993, "global_step": 24921, "epoch": 280, "lr": 9.80087764246932e-05}
{"train_loss": 0.35599184036254883, "global_step": 24922, "epoch": 280, "lr": 9.800861444451732e-05}
{"train_loss": 0.3132922351360321, "global_step": 24923, "epoch": 280, "lr": 9.800845245788725e-05}
{"train_loss": 0.34893444180488586, "global_step": 24924, "epoch": 280, "lr": 9.800829046480305e-05}
{"train_loss": 0.4014897048473358, "global_step": 24925, "epoch": 280, "lr": 9.80081284652647e-05}
{"train_loss": 0.3323662579059601, "global_step": 24926, "epoch": 280, "lr": 9.800796645927226e-05}
{"train_loss": 0.36222022771835327, "global_step": 24927, "epoch": 280, "lr": 9.800780444682573e-05}
{"train_loss": 0.41859152913093567, "global_step": 24928, "epoch": 280, "lr": 9.800764242792513e-05}
{"train_loss": 0.34380337595939636, "global_step": 24929, "epoch": 280, "lr": 9.800748040257048e-05}
{"train_loss": 0.298111230134964, "global_step": 24930, "epoch": 280, "lr": 9.800731837076183e-05}
{"train_loss": 0.29916834831237793, "global_step": 24931, "epoch": 280, "lr": 9.800715633249918e-05}
{"train_loss": 0.5089552998542786, "global_step": 24932, "epoch": 280, "lr": 9.800699428778253e-05}
{"train_loss": 0.4868405759334564, "global_step": 24933, "epoch": 280, "lr": 9.800683223661195e-05}
{"train_loss": 0.5292618274688721, "global_step": 24934, "epoch": 280, "lr": 9.800667017898742e-05}
{"train_loss": 0.3989628255367279, "global_step": 24935, "epoch": 280, "lr": 9.800650811490898e-05}
{"train_loss": 0.3165620267391205, "global_step": 24936, "epoch": 280, "lr": 9.800634604437665e-05}
{"train_loss": 0.30075395107269287, "global_step": 24937, "epoch": 280, "lr": 9.800618396739046e-05}
{"train_loss": 0.41762450337409973, "global_step": 24938, "epoch": 280, "lr": 9.800602188395042e-05}
{"train_loss": 0.42289528250694275, "global_step": 24939, "epoch": 280, "lr": 9.800585979405656e-05}
{"train_loss": 0.4127082824707031, "global_step": 24940, "epoch": 280, "lr": 9.800569769770889e-05}
{"train_loss": 0.3517722189426422, "global_step": 24941, "epoch": 280, "lr": 9.800553559490744e-05}
{"train_loss": 0.3847643733024597, "global_step": 24942, "epoch": 280, "lr": 9.800537348565222e-05}
{"train_loss": 0.28721919655799866, "global_step": 24943, "epoch": 280, "lr": 9.800521136994327e-05}
{"train_loss": 0.3414081633090973, "global_step": 24944, "epoch": 280, "lr": 9.800504924778063e-05}
{"train_loss": 0.3341018855571747, "global_step": 24945, "epoch": 280, "lr": 9.800488711916427e-05}
{"train_loss": 0.43381425738334656, "global_step": 24946, "epoch": 280, "lr": 9.800472498409423e-05}
{"train_loss": 0.39425593614578247, "global_step": 24947, "epoch": 280, "lr": 9.800456284257056e-05}
{"train_loss": 0.3225574791431427, "global_step": 24948, "epoch": 280, "lr": 9.800440069459325e-05}
{"train_loss": 0.4577501118183136, "global_step": 24949, "epoch": 280, "lr": 9.800423854016235e-05}
{"train_loss": 0.4227394163608551, "global_step": 24950, "epoch": 280, "lr": 9.800407637927784e-05}
{"train_loss": 0.5139621496200562, "global_step": 24951, "epoch": 280, "lr": 9.800391421193977e-05}
{"train_loss": 0.34511664509773254, "global_step": 24952, "epoch": 280, "lr": 9.800375203814818e-05}
{"train_loss": 0.4230114221572876, "global_step": 24953, "epoch": 280, "lr": 9.800358985790306e-05}
{"train_loss": 0.3743639290332794, "global_step": 24954, "epoch": 280, "lr": 9.800342767120444e-05}
{"train_loss": 0.39065155386924744, "global_step": 24955, "epoch": 280, "lr": 9.800326547805235e-05}
{"train_loss": 0.3875496983528137, "global_step": 24956, "epoch": 280, "lr": 9.800310327844679e-05}
{"train_loss": 0.49562764167785645, "global_step": 24957, "epoch": 280, "lr": 9.800294107238781e-05}
{"train_loss": 0.4272386431694031, "global_step": 24958, "epoch": 280, "lr": 9.800277885987543e-05}
{"train_loss": 0.5120381116867065, "global_step": 24959, "epoch": 280, "lr": 9.800261664090964e-05}
{"train_loss": 0.5456417798995972, "global_step": 24960, "epoch": 280, "lr": 9.80024544154905e-05}
{"train_loss": 0.402742862701416, "global_step": 24961, "epoch": 280, "lr": 9.8002292183618e-05}
{"train_loss": 0.40922611951828003, "global_step": 24962, "epoch": 280, "lr": 9.80021299452922e-05}
{"train_loss": 0.39725083112716675, "global_step": 24963, "epoch": 280, "lr": 9.800196770051309e-05}
{"train_loss": 0.3649863302707672, "global_step": 24964, "epoch": 280, "lr": 9.800180544928068e-05}
{"train_loss": 0.39839205145835876, "global_step": 24965, "epoch": 280, "lr": 9.800164319159504e-05}
{"train_loss": 0.4676450490951538, "global_step": 24966, "epoch": 280, "lr": 9.800148092745614e-05}
{"train_loss": 0.45328161120414734, "global_step": 24967, "epoch": 280, "lr": 9.800131865686404e-05}
{"train_loss": 0.414311021566391, "global_step": 24968, "epoch": 280, "lr": 9.800115637981875e-05}
{"train_loss": 0.4541514217853546, "global_step": 24969, "epoch": 280, "lr": 9.800099409632029e-05}
{"train_loss": 0.3715037703514099, "global_step": 24970, "epoch": 280, "lr": 9.800083180636868e-05}
{"train_loss": 0.46029630303382874, "global_step": 24971, "epoch": 280, "lr": 9.800066950996394e-05}
{"train_loss": 0.42874225974082947, "global_step": 24972, "epoch": 280, "lr": 9.800050720710608e-05}
{"train_loss": 0.3687480092048645, "global_step": 24973, "epoch": 280, "lr": 9.800034489779515e-05}
{"train_loss": 0.3431450128555298, "global_step": 24974, "epoch": 280, "lr": 9.800018258203117e-05}
{"train_loss": 0.3767874240875244, "global_step": 24975, "epoch": 280, "lr": 9.800002025981414e-05}
{"train_loss": 0.2698417901992798, "global_step": 24976, "epoch": 280, "lr": 9.799985793114409e-05}
{"train_loss": 0.3956766128540039, "global_step": 24977, "epoch": 280, "lr": 9.799969559602104e-05}
{"train_loss": 0.41614875197410583, "global_step": 24978, "epoch": 280, "lr": 9.799953325444503e-05}
{"train_loss": 0.43614140152931213, "global_step": 24979, "epoch": 280, "lr": 9.799937090641606e-05}
{"train_loss": 0.46425607800483704, "global_step": 24980, "epoch": 280, "lr": 9.799920855193417e-05}
{"train_loss": 0.39940309524536133, "global_step": 24981, "epoch": 280, "lr": 9.799904619099936e-05}
{"train_loss": 0.5478470921516418, "global_step": 24982, "epoch": 280, "lr": 9.799888382361166e-05}
{"train_loss": 0.3683876395225525, "global_step": 24983, "epoch": 280, "lr": 9.799872144977109e-05}
{"train_loss": 0.4173072874546051, "global_step": 24984, "epoch": 280, "lr": 9.79985590694777e-05}
{"train_loss": 0.43831804394721985, "global_step": 24985, "epoch": 280, "lr": 9.799839668273147e-05}
{"train_loss": 0.4766498804092407, "global_step": 24986, "epoch": 280, "lr": 9.799823428953246e-05}
{"train_loss": 0.42093560099601746, "global_step": 24987, "epoch": 280, "lr": 9.799807188988064e-05}
{"train_loss": 0.5492247939109802, "global_step": 24988, "epoch": 280, "lr": 9.79979094837761e-05}
{"train_loss": 0.6759719848632812, "global_step": 24989, "epoch": 280, "lr": 9.79977470712188e-05}
{"train_loss": 0.4104439616203308, "global_step": 24990, "epoch": 280, "lr": 9.79975846522088e-05}
{"train_loss": 0.4304586350917816, "global_step": 24991, "epoch": 280, "lr": 9.799742222674611e-05}
{"train_loss": 0.4000602066516876, "global_step": 24992, "epoch": 280, "lr": 9.799725979483074e-05}
{"train_loss": 0.45410698652267456, "global_step": 24993, "epoch": 280, "lr": 9.799709735646274e-05}
{"train_loss": 0.4015037417411804, "global_step": 24994, "epoch": 280, "lr": 9.79969349116421e-05}
{"train_loss": 0.37209975719451904, "global_step": 24995, "epoch": 280, "lr": 9.799677246036886e-05}
{"train_loss": 0.40408292412757874, "global_step": 24996, "epoch": 280, "lr": 9.799661000264304e-05}
{"train_loss": 0.3996160924434662, "global_step": 24997, "epoch": 280, "lr": 9.799644753846466e-05}
{"train_loss": 0.42134764790534973, "global_step": 24998, "epoch": 280, "lr": 9.799628506783376e-05}
{"train_loss": 0.4009562134742737, "global_step": 24999, "epoch": 280, "lr": 9.799612259075032e-05}
{"train_loss": 0.6071492433547974, "global_step": 25000, "epoch": 280, "lr": 9.799596010721441e-05}
{"train_loss": 0.4658394157886505, "global_step": 25001, "epoch": 280, "lr": 9.799579761722601e-05}
{"train_loss": 0.5222952961921692, "global_step": 25002, "epoch": 280, "lr": 9.799563512078517e-05}
{"train_loss": 0.5005987882614136, "global_step": 25003, "epoch": 280, "lr": 9.79954726178919e-05}
{"train_loss": 0.3571309745311737, "global_step": 25004, "epoch": 280, "lr": 9.799531010854621e-05}
{"train_loss": 0.49627381563186646, "global_step": 25005, "epoch": 280, "lr": 9.799514759274815e-05}
{"train_loss": 0.6032232046127319, "global_step": 25006, "epoch": 280, "lr": 9.799498507049773e-05}
{"train_loss": 0.41777339577674866, "global_step": 25007, "epoch": 280, "lr": 9.799482254179497e-05}
{"train_loss": 0.4166421689344256, "global_step": 25008, "epoch": 280, "lr": 9.799466000663989e-05, "val_loss": 1.8910276889801025, "train_action_mse_error": 39.42645263671875}
{"train_loss": 0.3602348566055298, "global_step": 25009, "epoch": 281, "lr": 9.799449746503249e-05}
{"train_loss": 0.41551461815834045, "global_step": 25010, "epoch": 281, "lr": 9.799433491697283e-05}
{"train_loss": 0.35173869132995605, "global_step": 25011, "epoch": 281, "lr": 9.799417236246092e-05}
{"train_loss": 0.4809694290161133, "global_step": 25012, "epoch": 281, "lr": 9.799400980149678e-05}
{"train_loss": 0.5864102840423584, "global_step": 25013, "epoch": 281, "lr": 9.799384723408043e-05}
{"train_loss": 0.5515735149383545, "global_step": 25014, "epoch": 281, "lr": 9.799368466021189e-05}
{"train_loss": 0.46078258752822876, "global_step": 25015, "epoch": 281, "lr": 9.799352207989118e-05}
{"train_loss": 0.47822269797325134, "global_step": 25016, "epoch": 281, "lr": 9.799335949311833e-05}
{"train_loss": 0.36951547861099243, "global_step": 25017, "epoch": 281, "lr": 9.799319689989336e-05}
{"train_loss": 0.4339967370033264, "global_step": 25018, "epoch": 281, "lr": 9.79930343002163e-05}
{"train_loss": 0.4004979729652405, "global_step": 25019, "epoch": 281, "lr": 9.799287169408713e-05}
{"train_loss": 0.42718467116355896, "global_step": 25020, "epoch": 281, "lr": 9.799270908150593e-05}
{"train_loss": 0.5249430537223816, "global_step": 25021, "epoch": 281, "lr": 9.799254646247269e-05}
{"train_loss": 0.41837605834007263, "global_step": 25022, "epoch": 281, "lr": 9.799238383698743e-05}
{"train_loss": 0.45717403292655945, "global_step": 25023, "epoch": 281, "lr": 9.799222120505019e-05}
{"train_loss": 0.3555242419242859, "global_step": 25024, "epoch": 281, "lr": 9.799205856666095e-05}
{"train_loss": 0.3580282926559448, "global_step": 25025, "epoch": 281, "lr": 9.79918959218198e-05}
{"train_loss": 0.4619882106781006, "global_step": 25026, "epoch": 281, "lr": 9.799173327052671e-05}
{"train_loss": 0.5011720061302185, "global_step": 25027, "epoch": 281, "lr": 9.799157061278172e-05}
{"train_loss": 0.3788872957229614, "global_step": 25028, "epoch": 281, "lr": 9.799140794858484e-05}
{"train_loss": 0.4898417890071869, "global_step": 25029, "epoch": 281, "lr": 9.799124527793611e-05}
{"train_loss": 0.44768717885017395, "global_step": 25030, "epoch": 281, "lr": 9.799108260083553e-05}
{"train_loss": 0.4552420973777771, "global_step": 25031, "epoch": 281, "lr": 9.799091991728315e-05}
{"train_loss": 0.4585428833961487, "global_step": 25032, "epoch": 281, "lr": 9.799075722727897e-05}
{"train_loss": 0.4750813841819763, "global_step": 25033, "epoch": 281, "lr": 9.799059453082303e-05}
{"train_loss": 0.4494115710258484, "global_step": 25034, "epoch": 281, "lr": 9.79904318279153e-05}
{"train_loss": 0.37912970781326294, "global_step": 25035, "epoch": 281, "lr": 9.799026911855588e-05}
{"train_loss": 0.5810160040855408, "global_step": 25036, "epoch": 281, "lr": 9.799010640274473e-05}
{"train_loss": 0.47559046745300293, "global_step": 25037, "epoch": 281, "lr": 9.798994368048192e-05}
{"train_loss": 0.5020416378974915, "global_step": 25038, "epoch": 281, "lr": 9.798978095176742e-05}
{"train_loss": 0.3966465890407562, "global_step": 25039, "epoch": 281, "lr": 9.798961821660128e-05}
{"train_loss": 0.4151063859462738, "global_step": 25040, "epoch": 281, "lr": 9.798945547498353e-05}
{"train_loss": 0.37633392214775085, "global_step": 25041, "epoch": 281, "lr": 9.79892927269142e-05}
{"train_loss": 0.6241455078125, "global_step": 25042, "epoch": 281, "lr": 9.798912997239327e-05}
{"train_loss": 0.3721752166748047, "global_step": 25043, "epoch": 281, "lr": 9.798896721142079e-05}
{"train_loss": 0.3257671594619751, "global_step": 25044, "epoch": 281, "lr": 9.798880444399678e-05}
{"train_loss": 0.2982965111732483, "global_step": 25045, "epoch": 281, "lr": 9.798864167012128e-05}
{"train_loss": 0.47099947929382324, "global_step": 25046, "epoch": 281, "lr": 9.798847888979427e-05}
{"train_loss": 0.46111971139907837, "global_step": 25047, "epoch": 281, "lr": 9.79883161030158e-05}
{"train_loss": 0.4323943257331848, "global_step": 25048, "epoch": 281, "lr": 9.798815330978588e-05}
{"train_loss": 0.35106736421585083, "global_step": 25049, "epoch": 281, "lr": 9.798799051010454e-05}
{"train_loss": 0.49267226457595825, "global_step": 25050, "epoch": 281, "lr": 9.798782770397182e-05}
{"train_loss": 0.36178383231163025, "global_step": 25051, "epoch": 281, "lr": 9.798766489138768e-05}
{"train_loss": 0.5414479970932007, "global_step": 25052, "epoch": 281, "lr": 9.798750207235221e-05}
{"train_loss": 0.6086215972900391, "global_step": 25053, "epoch": 281, "lr": 9.79873392468654e-05}
{"train_loss": 0.38928061723709106, "global_step": 25054, "epoch": 281, "lr": 9.798717641492729e-05}
{"train_loss": 0.42495137453079224, "global_step": 25055, "epoch": 281, "lr": 9.798701357653788e-05}
{"train_loss": 0.5071505308151245, "global_step": 25056, "epoch": 281, "lr": 9.79868507316972e-05}
{"train_loss": 0.41444453597068787, "global_step": 25057, "epoch": 281, "lr": 9.798668788040526e-05}
{"train_loss": 0.511897623538971, "global_step": 25058, "epoch": 281, "lr": 9.798652502266211e-05}
{"train_loss": 0.4433140158653259, "global_step": 25059, "epoch": 281, "lr": 9.798636215846775e-05}
{"train_loss": 0.41117244958877563, "global_step": 25060, "epoch": 281, "lr": 9.798619928782223e-05}
{"train_loss": 0.41105344891548157, "global_step": 25061, "epoch": 281, "lr": 9.798603641072553e-05}
{"train_loss": 0.40822821855545044, "global_step": 25062, "epoch": 281, "lr": 9.798587352717768e-05}
{"train_loss": 0.4457852840423584, "global_step": 25063, "epoch": 281, "lr": 9.798571063717873e-05}
{"train_loss": 0.5194095373153687, "global_step": 25064, "epoch": 281, "lr": 9.798554774072867e-05}
{"train_loss": 0.455712229013443, "global_step": 25065, "epoch": 281, "lr": 9.798538483782757e-05}
{"train_loss": 0.5209367275238037, "global_step": 25066, "epoch": 281, "lr": 9.79852219284754e-05}
{"train_loss": 0.5859860181808472, "global_step": 25067, "epoch": 281, "lr": 9.798505901267221e-05}
{"train_loss": 0.5033721923828125, "global_step": 25068, "epoch": 281, "lr": 9.7984896090418e-05}
{"train_loss": 0.37758398056030273, "global_step": 25069, "epoch": 281, "lr": 9.798473316171282e-05}
{"train_loss": 0.4771258234977722, "global_step": 25070, "epoch": 281, "lr": 9.798457022655668e-05}
{"train_loss": 0.44156745076179504, "global_step": 25071, "epoch": 281, "lr": 9.798440728494957e-05}
{"train_loss": 0.40275856852531433, "global_step": 25072, "epoch": 281, "lr": 9.798424433689157e-05}
{"train_loss": 0.3896174430847168, "global_step": 25073, "epoch": 281, "lr": 9.798408138238266e-05}
{"train_loss": 0.5011893510818481, "global_step": 25074, "epoch": 281, "lr": 9.798391842142288e-05}
{"train_loss": 0.47261354327201843, "global_step": 25075, "epoch": 281, "lr": 9.798375545401225e-05}
{"train_loss": 0.5138930082321167, "global_step": 25076, "epoch": 281, "lr": 9.798359248015078e-05}
{"train_loss": 0.4890619218349457, "global_step": 25077, "epoch": 281, "lr": 9.798342949983851e-05}
{"train_loss": 0.4220527112483978, "global_step": 25078, "epoch": 281, "lr": 9.798326651307546e-05}
{"train_loss": 0.43309342861175537, "global_step": 25079, "epoch": 281, "lr": 9.798310351986161e-05}
{"train_loss": 0.5088697075843811, "global_step": 25080, "epoch": 281, "lr": 9.798294052019705e-05}
{"train_loss": 0.38297176361083984, "global_step": 25081, "epoch": 281, "lr": 9.798277751408176e-05}
{"train_loss": 0.48228392004966736, "global_step": 25082, "epoch": 281, "lr": 9.798261450151578e-05}
{"train_loss": 0.4032965302467346, "global_step": 25083, "epoch": 281, "lr": 9.79824514824991e-05}
{"train_loss": 0.4315219819545746, "global_step": 25084, "epoch": 281, "lr": 9.798228845703178e-05}
{"train_loss": 0.43981045484542847, "global_step": 25085, "epoch": 281, "lr": 9.798212542511383e-05}
{"train_loss": 0.42186427116394043, "global_step": 25086, "epoch": 281, "lr": 9.798196238674526e-05}
{"train_loss": 0.4532921314239502, "global_step": 25087, "epoch": 281, "lr": 9.798179934192609e-05}
{"train_loss": 0.4403437077999115, "global_step": 25088, "epoch": 281, "lr": 9.798163629065636e-05}
{"train_loss": 0.28092825412750244, "global_step": 25089, "epoch": 281, "lr": 9.79814732329361e-05}
{"train_loss": 0.2792867422103882, "global_step": 25090, "epoch": 281, "lr": 9.798131016876529e-05}
{"train_loss": 0.4260352551937103, "global_step": 25091, "epoch": 281, "lr": 9.798114709814399e-05}
{"train_loss": 0.5566592812538147, "global_step": 25092, "epoch": 281, "lr": 9.79809840210722e-05}
{"train_loss": 0.4476943016052246, "global_step": 25093, "epoch": 281, "lr": 9.798082093754996e-05}
{"train_loss": 0.35874441266059875, "global_step": 25094, "epoch": 281, "lr": 9.798065784757728e-05}
{"train_loss": 0.4907829463481903, "global_step": 25095, "epoch": 281, "lr": 9.798049475115421e-05}
{"train_loss": 0.30062901973724365, "global_step": 25096, "epoch": 281, "lr": 9.79803316482807e-05}
{"train_loss": 0.44275423884391785, "global_step": 25097, "epoch": 281, "lr": 9.798016853895686e-05, "val_loss": 2.004775285720825}
{"train_loss": 0.34608855843544006, "global_step": 25098, "epoch": 282, "lr": 9.798000542318266e-05}
{"train_loss": 0.48432815074920654, "global_step": 25099, "epoch": 282, "lr": 9.797984230095814e-05}
{"train_loss": 0.35472530126571655, "global_step": 25100, "epoch": 282, "lr": 9.79796791722833e-05}
{"train_loss": 0.43783804774284363, "global_step": 25101, "epoch": 282, "lr": 9.797951603715819e-05}
{"train_loss": 0.43508949875831604, "global_step": 25102, "epoch": 282, "lr": 9.79793528955828e-05}
{"train_loss": 0.3960343599319458, "global_step": 25103, "epoch": 282, "lr": 9.797918974755718e-05}
{"train_loss": 0.42514902353286743, "global_step": 25104, "epoch": 282, "lr": 9.797902659308134e-05}
{"train_loss": 0.49696412682533264, "global_step": 25105, "epoch": 282, "lr": 9.797886343215531e-05}
{"train_loss": 0.3816167414188385, "global_step": 25106, "epoch": 282, "lr": 9.797870026477911e-05}
{"train_loss": 0.3606279492378235, "global_step": 25107, "epoch": 282, "lr": 9.797853709095274e-05}
{"train_loss": 0.4017919600009918, "global_step": 25108, "epoch": 282, "lr": 9.797837391067625e-05}
{"train_loss": 0.39032283425331116, "global_step": 25109, "epoch": 282, "lr": 9.797821072394967e-05}
{"train_loss": 0.4187120795249939, "global_step": 25110, "epoch": 282, "lr": 9.797804753077298e-05}
{"train_loss": 0.28657230734825134, "global_step": 25111, "epoch": 282, "lr": 9.797788433114625e-05}
{"train_loss": 0.33097243309020996, "global_step": 25112, "epoch": 282, "lr": 9.797772112506946e-05}
{"train_loss": 0.3665010333061218, "global_step": 25113, "epoch": 282, "lr": 9.797755791254266e-05}
{"train_loss": 0.5258738994598389, "global_step": 25114, "epoch": 282, "lr": 9.797739469356586e-05}
{"train_loss": 0.48145800828933716, "global_step": 25115, "epoch": 282, "lr": 9.797723146813909e-05}
{"train_loss": 0.3839374780654907, "global_step": 25116, "epoch": 282, "lr": 9.797706823626233e-05}
{"train_loss": 0.4154912531375885, "global_step": 25117, "epoch": 282, "lr": 9.797690499793568e-05}
{"train_loss": 0.33290186524391174, "global_step": 25118, "epoch": 282, "lr": 9.797674175315911e-05}
{"train_loss": 0.3626163601875305, "global_step": 25119, "epoch": 282, "lr": 9.797657850193265e-05}
{"train_loss": 0.5069169402122498, "global_step": 25120, "epoch": 282, "lr": 9.797641524425631e-05}
{"train_loss": 0.42192840576171875, "global_step": 25121, "epoch": 282, "lr": 9.797625198013015e-05}
{"train_loss": 0.4090287983417511, "global_step": 25122, "epoch": 282, "lr": 9.797608870955415e-05}
{"train_loss": 0.4659537076950073, "global_step": 25123, "epoch": 282, "lr": 9.797592543252835e-05}
{"train_loss": 0.3625282049179077, "global_step": 25124, "epoch": 282, "lr": 9.797576214905277e-05}
{"train_loss": 0.40638312697410583, "global_step": 25125, "epoch": 282, "lr": 9.797559885912743e-05}
{"train_loss": 0.3809870183467865, "global_step": 25126, "epoch": 282, "lr": 9.797543556275238e-05}
{"train_loss": 0.4516270160675049, "global_step": 25127, "epoch": 282, "lr": 9.79752722599276e-05}
{"train_loss": 0.38464513421058655, "global_step": 25128, "epoch": 282, "lr": 9.797510895065314e-05}
{"train_loss": 0.2857637405395508, "global_step": 25129, "epoch": 282, "lr": 9.7974945634929e-05}
{"train_loss": 0.33995717763900757, "global_step": 25130, "epoch": 282, "lr": 9.79747823127552e-05}
{"train_loss": 0.6450032591819763, "global_step": 25131, "epoch": 282, "lr": 9.797461898413181e-05}
{"train_loss": 0.3963891565799713, "global_step": 25132, "epoch": 282, "lr": 9.797445564905879e-05}
{"train_loss": 0.4330836832523346, "global_step": 25133, "epoch": 282, "lr": 9.79742923075362e-05}
{"train_loss": 0.4182544946670532, "global_step": 25134, "epoch": 282, "lr": 9.797412895956405e-05}
{"train_loss": 0.3910508155822754, "global_step": 25135, "epoch": 282, "lr": 9.797396560514236e-05}
{"train_loss": 0.48967546224594116, "global_step": 25136, "epoch": 282, "lr": 9.797380224427116e-05}
{"train_loss": 0.360904335975647, "global_step": 25137, "epoch": 282, "lr": 9.797363887695045e-05}
{"train_loss": 0.4199028015136719, "global_step": 25138, "epoch": 282, "lr": 9.79734755031803e-05}
{"train_loss": 0.3819723427295685, "global_step": 25139, "epoch": 282, "lr": 9.797331212296067e-05}
{"train_loss": 0.273223340511322, "global_step": 25140, "epoch": 282, "lr": 9.797314873629164e-05}
{"train_loss": 0.5213761329650879, "global_step": 25141, "epoch": 282, "lr": 9.797298534317319e-05}
{"train_loss": 0.6068146228790283, "global_step": 25142, "epoch": 282, "lr": 9.797282194360536e-05}
{"train_loss": 0.4340389370918274, "global_step": 25143, "epoch": 282, "lr": 9.797265853758817e-05}
{"train_loss": 0.4148702919483185, "global_step": 25144, "epoch": 282, "lr": 9.797249512512164e-05}
{"train_loss": 0.4789787530899048, "global_step": 25145, "epoch": 282, "lr": 9.797233170620579e-05}
{"train_loss": 0.5216295719146729, "global_step": 25146, "epoch": 282, "lr": 9.797216828084064e-05}
{"train_loss": 0.32818660140037537, "global_step": 25147, "epoch": 282, "lr": 9.797200484902623e-05}
{"train_loss": 0.47714996337890625, "global_step": 25148, "epoch": 282, "lr": 9.797184141076254e-05}
{"train_loss": 0.450547456741333, "global_step": 25149, "epoch": 282, "lr": 9.797167796604965e-05}
{"train_loss": 0.4284420609474182, "global_step": 25150, "epoch": 282, "lr": 9.797151451488755e-05}
{"train_loss": 0.43217816948890686, "global_step": 25151, "epoch": 282, "lr": 9.797135105727627e-05}
{"train_loss": 0.5188059210777283, "global_step": 25152, "epoch": 282, "lr": 9.79711875932158e-05}
{"train_loss": 0.3937477469444275, "global_step": 25153, "epoch": 282, "lr": 9.797102412270621e-05}
{"train_loss": 0.38437318801879883, "global_step": 25154, "epoch": 282, "lr": 9.797086064574748e-05}
{"train_loss": 0.4315965175628662, "global_step": 25155, "epoch": 282, "lr": 9.797069716233967e-05}
{"train_loss": 0.2981538474559784, "global_step": 25156, "epoch": 282, "lr": 9.797053367248279e-05}
{"train_loss": 0.4053279161453247, "global_step": 25157, "epoch": 282, "lr": 9.797037017617685e-05}
{"train_loss": 0.35608816146850586, "global_step": 25158, "epoch": 282, "lr": 9.797020667342186e-05}
{"train_loss": 0.31310635805130005, "global_step": 25159, "epoch": 282, "lr": 9.797004316421789e-05}
{"train_loss": 0.5512388944625854, "global_step": 25160, "epoch": 282, "lr": 9.796987964856492e-05}
{"train_loss": 0.40779080986976624, "global_step": 25161, "epoch": 282, "lr": 9.796971612646298e-05}
{"train_loss": 0.4708326458930969, "global_step": 25162, "epoch": 282, "lr": 9.79695525979121e-05}
{"train_loss": 0.44009387493133545, "global_step": 25163, "epoch": 282, "lr": 9.796938906291228e-05}
{"train_loss": 0.4321909546852112, "global_step": 25164, "epoch": 282, "lr": 9.796922552146359e-05}
{"train_loss": 0.32670357823371887, "global_step": 25165, "epoch": 282, "lr": 9.7969061973566e-05}
{"train_loss": 0.44991597533226013, "global_step": 25166, "epoch": 282, "lr": 9.796889841921957e-05}
{"train_loss": 0.38647109270095825, "global_step": 25167, "epoch": 282, "lr": 9.796873485842431e-05}
{"train_loss": 0.4392499029636383, "global_step": 25168, "epoch": 282, "lr": 9.796857129118022e-05}
{"train_loss": 0.41101062297821045, "global_step": 25169, "epoch": 282, "lr": 9.796840771748736e-05}
{"train_loss": 0.44817766547203064, "global_step": 25170, "epoch": 282, "lr": 9.796824413734571e-05}
{"train_loss": 0.4079552888870239, "global_step": 25171, "epoch": 282, "lr": 9.796808055075532e-05}
{"train_loss": 0.4574630558490753, "global_step": 25172, "epoch": 282, "lr": 9.796791695771624e-05}
{"train_loss": 0.4528156816959381, "global_step": 25173, "epoch": 282, "lr": 9.796775335822841e-05}
{"train_loss": 0.4502124488353729, "global_step": 25174, "epoch": 282, "lr": 9.796758975229192e-05}
{"train_loss": 0.3546329438686371, "global_step": 25175, "epoch": 282, "lr": 9.796742613990679e-05}
{"train_loss": 0.2972305417060852, "global_step": 25176, "epoch": 282, "lr": 9.7967262521073e-05}
{"train_loss": 0.47148361802101135, "global_step": 25177, "epoch": 282, "lr": 9.79670988957906e-05}
{"train_loss": 0.47727346420288086, "global_step": 25178, "epoch": 282, "lr": 9.796693526405962e-05}
{"train_loss": 0.5009273290634155, "global_step": 25179, "epoch": 282, "lr": 9.796677162588004e-05}
{"train_loss": 0.3806053698062897, "global_step": 25180, "epoch": 282, "lr": 9.796660798125195e-05}
{"train_loss": 0.4235990643501282, "global_step": 25181, "epoch": 282, "lr": 9.79664443301753e-05}
{"train_loss": 0.4370083808898926, "global_step": 25182, "epoch": 282, "lr": 9.796628067265018e-05}
{"train_loss": 0.442190945148468, "global_step": 25183, "epoch": 282, "lr": 9.796611700867656e-05}
{"train_loss": 0.5454469919204712, "global_step": 25184, "epoch": 282, "lr": 9.796595333825448e-05}
{"train_loss": 0.4785638451576233, "global_step": 25185, "epoch": 282, "lr": 9.796578966138397e-05}
{"train_loss": 0.4206293851472019, "global_step": 25186, "epoch": 282, "lr": 9.796562597806504e-05, "val_loss": 1.9695085287094116}
{"train_loss": 0.5206789970397949, "global_step": 25187, "epoch": 283, "lr": 9.79654622882977e-05}
{"train_loss": 0.3624004125595093, "global_step": 25188, "epoch": 283, "lr": 9.796529859208201e-05}
{"train_loss": 0.3915998041629791, "global_step": 25189, "epoch": 283, "lr": 9.796513488941797e-05}
{"train_loss": 0.44367876648902893, "global_step": 25190, "epoch": 283, "lr": 9.796497118030558e-05}
{"train_loss": 0.4358913004398346, "global_step": 25191, "epoch": 283, "lr": 9.796480746474491e-05}
{"train_loss": 0.400192528963089, "global_step": 25192, "epoch": 283, "lr": 9.796464374273596e-05}
{"train_loss": 0.4225590229034424, "global_step": 25193, "epoch": 283, "lr": 9.796448001427872e-05}
{"train_loss": 0.4076031744480133, "global_step": 25194, "epoch": 283, "lr": 9.796431627937326e-05}
{"train_loss": 0.36619457602500916, "global_step": 25195, "epoch": 283, "lr": 9.796415253801957e-05}
{"train_loss": 0.41231393814086914, "global_step": 25196, "epoch": 283, "lr": 9.796398879021768e-05}
{"train_loss": 0.4517378807067871, "global_step": 25197, "epoch": 283, "lr": 9.796382503596764e-05}
{"train_loss": 0.48591724038124084, "global_step": 25198, "epoch": 283, "lr": 9.796366127526943e-05}
{"train_loss": 0.39478975534439087, "global_step": 25199, "epoch": 283, "lr": 9.79634975081231e-05}
{"train_loss": 0.42288875579833984, "global_step": 25200, "epoch": 283, "lr": 9.796333373452866e-05}
{"train_loss": 0.4710947275161743, "global_step": 25201, "epoch": 283, "lr": 9.796316995448614e-05}
{"train_loss": 0.4429653584957123, "global_step": 25202, "epoch": 283, "lr": 9.796300616799554e-05}
{"train_loss": 0.41870787739753723, "global_step": 25203, "epoch": 283, "lr": 9.796284237505691e-05}
{"train_loss": 0.424635648727417, "global_step": 25204, "epoch": 283, "lr": 9.796267857567026e-05}
{"train_loss": 0.4748567044734955, "global_step": 25205, "epoch": 283, "lr": 9.796251476983561e-05}
{"train_loss": 0.46856945753097534, "global_step": 25206, "epoch": 283, "lr": 9.796235095755299e-05}
{"train_loss": 0.39076462388038635, "global_step": 25207, "epoch": 283, "lr": 9.796218713882242e-05}
{"train_loss": 0.36074158549308777, "global_step": 25208, "epoch": 283, "lr": 9.79620233136439e-05}
{"train_loss": 0.4882529377937317, "global_step": 25209, "epoch": 283, "lr": 9.796185948201748e-05}
{"train_loss": 0.3880649507045746, "global_step": 25210, "epoch": 283, "lr": 9.796169564394317e-05}
{"train_loss": 0.4229716360569, "global_step": 25211, "epoch": 283, "lr": 9.796153179942101e-05}
{"train_loss": 0.5623028874397278, "global_step": 25212, "epoch": 283, "lr": 9.796136794845098e-05}
{"train_loss": 0.4961850047111511, "global_step": 25213, "epoch": 283, "lr": 9.796120409103314e-05}
{"train_loss": 0.36906906962394714, "global_step": 25214, "epoch": 283, "lr": 9.796104022716751e-05}
{"train_loss": 0.37028372287750244, "global_step": 25215, "epoch": 283, "lr": 9.796087635685409e-05}
{"train_loss": 0.40683838725090027, "global_step": 25216, "epoch": 283, "lr": 9.796071248009293e-05}
{"train_loss": 0.29261133074760437, "global_step": 25217, "epoch": 283, "lr": 9.796054859688402e-05}
{"train_loss": 0.4779544770717621, "global_step": 25218, "epoch": 283, "lr": 9.796038470722741e-05}
{"train_loss": 0.43161970376968384, "global_step": 25219, "epoch": 283, "lr": 9.79602208111231e-05}
{"train_loss": 0.30340778827667236, "global_step": 25220, "epoch": 283, "lr": 9.796005690857113e-05}
{"train_loss": 0.48075342178344727, "global_step": 25221, "epoch": 283, "lr": 9.795989299957152e-05}
{"train_loss": 0.3633780777454376, "global_step": 25222, "epoch": 283, "lr": 9.795972908412428e-05}
{"train_loss": 0.4856414198875427, "global_step": 25223, "epoch": 283, "lr": 9.795956516222945e-05}
{"train_loss": 0.46950623393058777, "global_step": 25224, "epoch": 283, "lr": 9.795940123388703e-05}
{"train_loss": 0.33224350214004517, "global_step": 25225, "epoch": 283, "lr": 9.795923729909707e-05}
{"train_loss": 0.3478787839412689, "global_step": 25226, "epoch": 283, "lr": 9.795907335785954e-05}
{"train_loss": 0.41104310750961304, "global_step": 25227, "epoch": 283, "lr": 9.795890941017451e-05}
{"train_loss": 0.5859216451644897, "global_step": 25228, "epoch": 283, "lr": 9.795874545604201e-05}
{"train_loss": 0.42557692527770996, "global_step": 25229, "epoch": 283, "lr": 9.795858149546202e-05}
{"train_loss": 0.4870646595954895, "global_step": 25230, "epoch": 283, "lr": 9.79584175284346e-05}
{"train_loss": 0.47034046053886414, "global_step": 25231, "epoch": 283, "lr": 9.795825355495974e-05}
{"train_loss": 0.4129059910774231, "global_step": 25232, "epoch": 283, "lr": 9.795808957503749e-05}
{"train_loss": 0.40652740001678467, "global_step": 25233, "epoch": 283, "lr": 9.795792558866786e-05}
{"train_loss": 0.5151397585868835, "global_step": 25234, "epoch": 283, "lr": 9.795776159585086e-05}
{"train_loss": 0.5264220833778381, "global_step": 25235, "epoch": 283, "lr": 9.795759759658653e-05}
{"train_loss": 0.44751158356666565, "global_step": 25236, "epoch": 283, "lr": 9.795743359087489e-05}
{"train_loss": 0.4609803259372711, "global_step": 25237, "epoch": 283, "lr": 9.795726957871594e-05}
{"train_loss": 0.32574597001075745, "global_step": 25238, "epoch": 283, "lr": 9.795710556010974e-05}
{"train_loss": 0.46235382556915283, "global_step": 25239, "epoch": 283, "lr": 9.795694153505628e-05}
{"train_loss": 0.31871333718299866, "global_step": 25240, "epoch": 283, "lr": 9.795677750355559e-05}
{"train_loss": 0.44173961877822876, "global_step": 25241, "epoch": 283, "lr": 9.79566134656077e-05}
{"train_loss": 0.29922136664390564, "global_step": 25242, "epoch": 283, "lr": 9.795644942121263e-05}
{"train_loss": 0.33555933833122253, "global_step": 25243, "epoch": 283, "lr": 9.795628537037041e-05}
{"train_loss": 0.47937989234924316, "global_step": 25244, "epoch": 283, "lr": 9.795612131308102e-05}
{"train_loss": 0.4349895417690277, "global_step": 25245, "epoch": 283, "lr": 9.795595724934455e-05}
{"train_loss": 0.37685757875442505, "global_step": 25246, "epoch": 283, "lr": 9.795579317916098e-05}
{"train_loss": 0.4480690062046051, "global_step": 25247, "epoch": 283, "lr": 9.795562910253032e-05}
{"train_loss": 0.33909308910369873, "global_step": 25248, "epoch": 283, "lr": 9.795546501945263e-05}
{"train_loss": 0.5157039761543274, "global_step": 25249, "epoch": 283, "lr": 9.795530092992789e-05}
{"train_loss": 0.4550630450248718, "global_step": 25250, "epoch": 283, "lr": 9.795513683395616e-05}
{"train_loss": 0.4399144649505615, "global_step": 25251, "epoch": 283, "lr": 9.795497273153744e-05}
{"train_loss": 0.501487135887146, "global_step": 25252, "epoch": 283, "lr": 9.795480862267176e-05}
{"train_loss": 0.44772666692733765, "global_step": 25253, "epoch": 283, "lr": 9.795464450735913e-05}
{"train_loss": 0.39138445258140564, "global_step": 25254, "epoch": 283, "lr": 9.79544803855996e-05}
{"train_loss": 0.41139671206474304, "global_step": 25255, "epoch": 283, "lr": 9.795431625739318e-05}
{"train_loss": 0.4452587366104126, "global_step": 25256, "epoch": 283, "lr": 9.795415212273986e-05}
{"train_loss": 0.3794901967048645, "global_step": 25257, "epoch": 283, "lr": 9.795398798163971e-05}
{"train_loss": 0.5316165685653687, "global_step": 25258, "epoch": 283, "lr": 9.795382383409273e-05}
{"train_loss": 0.43467605113983154, "global_step": 25259, "epoch": 283, "lr": 9.795365968009893e-05}
{"train_loss": 0.44711363315582275, "global_step": 25260, "epoch": 283, "lr": 9.795349551965836e-05}
{"train_loss": 0.39733392000198364, "global_step": 25261, "epoch": 283, "lr": 9.795333135277102e-05}
{"train_loss": 0.43064865469932556, "global_step": 25262, "epoch": 283, "lr": 9.795316717943695e-05}
{"train_loss": 0.3862452507019043, "global_step": 25263, "epoch": 283, "lr": 9.795300299965614e-05}
{"train_loss": 0.38641557097435, "global_step": 25264, "epoch": 283, "lr": 9.795283881342864e-05}
{"train_loss": 0.39496201276779175, "global_step": 25265, "epoch": 283, "lr": 9.795267462075447e-05}
{"train_loss": 0.3673168122768402, "global_step": 25266, "epoch": 283, "lr": 9.795251042163364e-05}
{"train_loss": 0.38281527161598206, "global_step": 25267, "epoch": 283, "lr": 9.795234621606619e-05}
{"train_loss": 0.36669209599494934, "global_step": 25268, "epoch": 283, "lr": 9.795218200405212e-05}
{"train_loss": 0.3716951012611389, "global_step": 25269, "epoch": 283, "lr": 9.795201778559148e-05}
{"train_loss": 0.35057470202445984, "global_step": 25270, "epoch": 283, "lr": 9.795185356068426e-05}
{"train_loss": 0.43356359004974365, "global_step": 25271, "epoch": 283, "lr": 9.79516893293305e-05}
{"train_loss": 0.4896751344203949, "global_step": 25272, "epoch": 283, "lr": 9.795152509153023e-05}
{"train_loss": 0.2741723954677582, "global_step": 25273, "epoch": 283, "lr": 9.795136084728345e-05}
{"train_loss": 0.45092669129371643, "global_step": 25274, "epoch": 283, "lr": 9.79511965965902e-05}
{"train_loss": 0.4223195866922314, "global_step": 25275, "epoch": 283, "lr": 9.79510323394505e-05, "val_loss": 2.0292131900787354}
{"train_loss": 0.32204169034957886, "global_step": 25276, "epoch": 284, "lr": 9.795086807586436e-05}
{"train_loss": 0.3253375291824341, "global_step": 25277, "epoch": 284, "lr": 9.795070380583182e-05}
{"train_loss": 0.3389070928096771, "global_step": 25278, "epoch": 284, "lr": 9.795053952935288e-05}
{"train_loss": 0.4113726317882538, "global_step": 25279, "epoch": 284, "lr": 9.795037524642758e-05}
{"train_loss": 0.4776875674724579, "global_step": 25280, "epoch": 284, "lr": 9.795021095705594e-05}
{"train_loss": 0.34376394748687744, "global_step": 25281, "epoch": 284, "lr": 9.795004666123797e-05}
{"train_loss": 0.38716956973075867, "global_step": 25282, "epoch": 284, "lr": 9.794988235897371e-05}
{"train_loss": 0.4486299455165863, "global_step": 25283, "epoch": 284, "lr": 9.794971805026318e-05}
{"train_loss": 0.3550971746444702, "global_step": 25284, "epoch": 284, "lr": 9.794955373510637e-05}
{"train_loss": 0.429075688123703, "global_step": 25285, "epoch": 284, "lr": 9.794938941350335e-05}
{"train_loss": 0.4191049039363861, "global_step": 25286, "epoch": 284, "lr": 9.79492250854541e-05}
{"train_loss": 0.4227069616317749, "global_step": 25287, "epoch": 284, "lr": 9.794906075095867e-05}
{"train_loss": 0.3218822181224823, "global_step": 25288, "epoch": 284, "lr": 9.794889641001707e-05}
{"train_loss": 0.4093674123287201, "global_step": 25289, "epoch": 284, "lr": 9.794873206262934e-05}
{"train_loss": 0.3974305987358093, "global_step": 25290, "epoch": 284, "lr": 9.794856770879548e-05}
{"train_loss": 0.43611034750938416, "global_step": 25291, "epoch": 284, "lr": 9.794840334851552e-05}
{"train_loss": 0.33543676137924194, "global_step": 25292, "epoch": 284, "lr": 9.794823898178947e-05}
{"train_loss": 0.5070183873176575, "global_step": 25293, "epoch": 284, "lr": 9.794807460861737e-05}
{"train_loss": 0.29619428515434265, "global_step": 25294, "epoch": 284, "lr": 9.794791022899923e-05}
{"train_loss": 0.45841118693351746, "global_step": 25295, "epoch": 284, "lr": 9.794774584293509e-05}
{"train_loss": 0.44618475437164307, "global_step": 25296, "epoch": 284, "lr": 9.794758145042495e-05}
{"train_loss": 0.46486473083496094, "global_step": 25297, "epoch": 284, "lr": 9.794741705146886e-05}
{"train_loss": 0.2955569922924042, "global_step": 25298, "epoch": 284, "lr": 9.794725264606682e-05}
{"train_loss": 0.325992614030838, "global_step": 25299, "epoch": 284, "lr": 9.794708823421883e-05}
{"train_loss": 0.38258928060531616, "global_step": 25300, "epoch": 284, "lr": 9.794692381592496e-05}
{"train_loss": 0.46688374876976013, "global_step": 25301, "epoch": 284, "lr": 9.79467593911852e-05}
{"train_loss": 0.4911593198776245, "global_step": 25302, "epoch": 284, "lr": 9.794659495999959e-05}
{"train_loss": 0.48713159561157227, "global_step": 25303, "epoch": 284, "lr": 9.794643052236815e-05}
{"train_loss": 0.414531409740448, "global_step": 25304, "epoch": 284, "lr": 9.79462660782909e-05}
{"train_loss": 0.4613841474056244, "global_step": 25305, "epoch": 284, "lr": 9.794610162776783e-05}
{"train_loss": 0.5338835716247559, "global_step": 25306, "epoch": 284, "lr": 9.794593717079901e-05}
{"train_loss": 0.45735299587249756, "global_step": 25307, "epoch": 284, "lr": 9.794577270738446e-05}
{"train_loss": 0.517730712890625, "global_step": 25308, "epoch": 284, "lr": 9.794560823752416e-05}
{"train_loss": 0.35638076066970825, "global_step": 25309, "epoch": 284, "lr": 9.794544376121817e-05}
{"train_loss": 0.5527816414833069, "global_step": 25310, "epoch": 284, "lr": 9.79452792784665e-05}
{"train_loss": 0.389841765165329, "global_step": 25311, "epoch": 284, "lr": 9.794511478926915e-05}
{"train_loss": 0.4323473274707794, "global_step": 25312, "epoch": 284, "lr": 9.794495029362619e-05}
{"train_loss": 0.40419483184814453, "global_step": 25313, "epoch": 284, "lr": 9.79447857915376e-05}
{"train_loss": 0.4112577438354492, "global_step": 25314, "epoch": 284, "lr": 9.794462128300343e-05}
{"train_loss": 0.38442450761795044, "global_step": 25315, "epoch": 284, "lr": 9.794445676802368e-05}
{"train_loss": 0.36912068724632263, "global_step": 25316, "epoch": 284, "lr": 9.79442922465984e-05}
{"train_loss": 0.4089058041572571, "global_step": 25317, "epoch": 284, "lr": 9.794412771872757e-05}
{"train_loss": 0.45831987261772156, "global_step": 25318, "epoch": 284, "lr": 9.794396318441126e-05}
{"train_loss": 0.3974147140979767, "global_step": 25319, "epoch": 284, "lr": 9.794379864364944e-05}
{"train_loss": 0.4677572250366211, "global_step": 25320, "epoch": 284, "lr": 9.794363409644218e-05}
{"train_loss": 0.45825862884521484, "global_step": 25321, "epoch": 284, "lr": 9.794346954278948e-05}
{"train_loss": 0.4126749634742737, "global_step": 25322, "epoch": 284, "lr": 9.794330498269137e-05}
{"train_loss": 0.3433142304420471, "global_step": 25323, "epoch": 284, "lr": 9.794314041614786e-05}
{"train_loss": 0.3057610094547272, "global_step": 25324, "epoch": 284, "lr": 9.794297584315898e-05}
{"train_loss": 0.44460177421569824, "global_step": 25325, "epoch": 284, "lr": 9.794281126372476e-05}
{"train_loss": 0.47180411219596863, "global_step": 25326, "epoch": 284, "lr": 9.794264667784519e-05}
{"train_loss": 0.4977213144302368, "global_step": 25327, "epoch": 284, "lr": 9.794248208552034e-05}
{"train_loss": 0.41686201095581055, "global_step": 25328, "epoch": 284, "lr": 9.79423174867502e-05}
{"train_loss": 0.3236635625362396, "global_step": 25329, "epoch": 284, "lr": 9.794215288153479e-05}
{"train_loss": 0.40842583775520325, "global_step": 25330, "epoch": 284, "lr": 9.794198826987416e-05}
{"train_loss": 0.41810399293899536, "global_step": 25331, "epoch": 284, "lr": 9.794182365176829e-05}
{"train_loss": 0.37074482440948486, "global_step": 25332, "epoch": 284, "lr": 9.794165902721724e-05}
{"train_loss": 0.383848637342453, "global_step": 25333, "epoch": 284, "lr": 9.794149439622103e-05}
{"train_loss": 0.4874471127986908, "global_step": 25334, "epoch": 284, "lr": 9.794132975877965e-05}
{"train_loss": 0.33133408427238464, "global_step": 25335, "epoch": 284, "lr": 9.794116511489316e-05}
{"train_loss": 0.4119676649570465, "global_step": 25336, "epoch": 284, "lr": 9.794100046456155e-05}
{"train_loss": 0.5488520860671997, "global_step": 25337, "epoch": 284, "lr": 9.794083580778485e-05}
{"train_loss": 0.3894554376602173, "global_step": 25338, "epoch": 284, "lr": 9.794067114456312e-05}
{"train_loss": 0.43828749656677246, "global_step": 25339, "epoch": 284, "lr": 9.794050647489632e-05}
{"train_loss": 0.5954750180244446, "global_step": 25340, "epoch": 284, "lr": 9.794034179878452e-05}
{"train_loss": 0.42372867465019226, "global_step": 25341, "epoch": 284, "lr": 9.794017711622772e-05}
{"train_loss": 0.4715973734855652, "global_step": 25342, "epoch": 284, "lr": 9.794001242722594e-05}
{"train_loss": 0.4255770146846771, "global_step": 25343, "epoch": 284, "lr": 9.793984773177922e-05}
{"train_loss": 0.3629859983921051, "global_step": 25344, "epoch": 284, "lr": 9.793968302988758e-05}
{"train_loss": 0.3916042447090149, "global_step": 25345, "epoch": 284, "lr": 9.793951832155102e-05}
{"train_loss": 0.44019177556037903, "global_step": 25346, "epoch": 284, "lr": 9.793935360676958e-05}
{"train_loss": 0.3974839150905609, "global_step": 25347, "epoch": 284, "lr": 9.793918888554327e-05}
{"train_loss": 0.4054565131664276, "global_step": 25348, "epoch": 284, "lr": 9.793902415787214e-05}
{"train_loss": 0.40285390615463257, "global_step": 25349, "epoch": 284, "lr": 9.793885942375618e-05}
{"train_loss": 0.5046840906143188, "global_step": 25350, "epoch": 284, "lr": 9.793869468319543e-05}
{"train_loss": 0.4947657883167267, "global_step": 25351, "epoch": 284, "lr": 9.793852993618991e-05}
{"train_loss": 0.5177670121192932, "global_step": 25352, "epoch": 284, "lr": 9.793836518273962e-05}
{"train_loss": 0.42569851875305176, "global_step": 25353, "epoch": 284, "lr": 9.79382004228446e-05}
{"train_loss": 0.4554736018180847, "global_step": 25354, "epoch": 284, "lr": 9.79380356565049e-05}
{"train_loss": 0.45380517840385437, "global_step": 25355, "epoch": 284, "lr": 9.79378708837205e-05}
{"train_loss": 0.3924122154712677, "global_step": 25356, "epoch": 284, "lr": 9.793770610449144e-05}
{"train_loss": 0.22956916689872742, "global_step": 25357, "epoch": 284, "lr": 9.793754131881774e-05}
{"train_loss": 0.4079277813434601, "global_step": 25358, "epoch": 284, "lr": 9.793737652669941e-05}
{"train_loss": 0.3998335897922516, "global_step": 25359, "epoch": 284, "lr": 9.79372117281365e-05}
{"train_loss": 0.5506894588470459, "global_step": 25360, "epoch": 284, "lr": 9.793704692312902e-05}
{"train_loss": 0.3773351311683655, "global_step": 25361, "epoch": 284, "lr": 9.793688211167698e-05}
{"train_loss": 0.48453202843666077, "global_step": 25362, "epoch": 284, "lr": 9.79367172937804e-05}
{"train_loss": 0.42948806285858154, "global_step": 25363, "epoch": 284, "lr": 9.793655246943932e-05}
{"train_loss": 0.4211717288145858, "global_step": 25364, "epoch": 284, "lr": 9.793638763865375e-05, "val_loss": 2.016608238220215}
{"train_loss": 0.40562695264816284, "global_step": 25365, "epoch": 285, "lr": 9.793622280142373e-05}
{"train_loss": 0.48683398962020874, "global_step": 25366, "epoch": 285, "lr": 9.793605795774926e-05}
{"train_loss": 0.3838024139404297, "global_step": 25367, "epoch": 285, "lr": 9.793589310763039e-05}
{"train_loss": 0.4014703929424286, "global_step": 25368, "epoch": 285, "lr": 9.793572825106709e-05}
{"train_loss": 0.3315730094909668, "global_step": 25369, "epoch": 285, "lr": 9.793556338805943e-05}
{"train_loss": 0.37035462260246277, "global_step": 25370, "epoch": 285, "lr": 9.793539851860743e-05}
{"train_loss": 0.3976166546344757, "global_step": 25371, "epoch": 285, "lr": 9.793523364271108e-05}
{"train_loss": 0.41512951254844666, "global_step": 25372, "epoch": 285, "lr": 9.793506876037043e-05}
{"train_loss": 0.41383737325668335, "global_step": 25373, "epoch": 285, "lr": 9.79349038715855e-05}
{"train_loss": 0.5180723667144775, "global_step": 25374, "epoch": 285, "lr": 9.79347389763563e-05}
{"train_loss": 0.5268617868423462, "global_step": 25375, "epoch": 285, "lr": 9.793457407468287e-05}
{"train_loss": 0.47205907106399536, "global_step": 25376, "epoch": 285, "lr": 9.793440916656523e-05}
{"train_loss": 0.39116647839546204, "global_step": 25377, "epoch": 285, "lr": 9.793424425200336e-05}
{"train_loss": 0.39136773347854614, "global_step": 25378, "epoch": 285, "lr": 9.793407933099733e-05}
{"train_loss": 0.3942475914955139, "global_step": 25379, "epoch": 285, "lr": 9.793391440354716e-05}
{"train_loss": 0.3815493583679199, "global_step": 25380, "epoch": 285, "lr": 9.793374946965284e-05}
{"train_loss": 0.3836536705493927, "global_step": 25381, "epoch": 285, "lr": 9.793358452931441e-05}
{"train_loss": 0.3369597792625427, "global_step": 25382, "epoch": 285, "lr": 9.79334195825319e-05}
{"train_loss": 0.5123944282531738, "global_step": 25383, "epoch": 285, "lr": 9.793325462930534e-05}
{"train_loss": 0.4731086194515228, "global_step": 25384, "epoch": 285, "lr": 9.793308966963473e-05}
{"train_loss": 0.4541926085948944, "global_step": 25385, "epoch": 285, "lr": 9.793292470352009e-05}
{"train_loss": 0.49500104784965515, "global_step": 25386, "epoch": 285, "lr": 9.793275973096146e-05}
{"train_loss": 0.41406863927841187, "global_step": 25387, "epoch": 285, "lr": 9.793259475195886e-05}
{"train_loss": 0.4637850224971771, "global_step": 25388, "epoch": 285, "lr": 9.793242976651229e-05}
{"train_loss": 0.47199398279190063, "global_step": 25389, "epoch": 285, "lr": 9.793226477462181e-05}
{"train_loss": 0.4311527907848358, "global_step": 25390, "epoch": 285, "lr": 9.793209977628741e-05}
{"train_loss": 0.3959050178527832, "global_step": 25391, "epoch": 285, "lr": 9.793193477150912e-05}
{"train_loss": 0.3967929184436798, "global_step": 25392, "epoch": 285, "lr": 9.793176976028697e-05}
{"train_loss": 0.4299849271774292, "global_step": 25393, "epoch": 285, "lr": 9.793160474262098e-05}
{"train_loss": 0.44947054982185364, "global_step": 25394, "epoch": 285, "lr": 9.793143971851118e-05}
{"train_loss": 0.5396599173545837, "global_step": 25395, "epoch": 285, "lr": 9.793127468795756e-05}
{"train_loss": 0.4070492088794708, "global_step": 25396, "epoch": 285, "lr": 9.793110965096017e-05}
{"train_loss": 0.39346370100975037, "global_step": 25397, "epoch": 285, "lr": 9.793094460751904e-05}
{"train_loss": 0.3054817020893097, "global_step": 25398, "epoch": 285, "lr": 9.793077955763416e-05}
{"train_loss": 0.3664552569389343, "global_step": 25399, "epoch": 285, "lr": 9.793061450130559e-05}
{"train_loss": 0.37140101194381714, "global_step": 25400, "epoch": 285, "lr": 9.793044943853331e-05}
{"train_loss": 0.415903776884079, "global_step": 25401, "epoch": 285, "lr": 9.793028436931739e-05}
{"train_loss": 0.5068948268890381, "global_step": 25402, "epoch": 285, "lr": 9.79301192936578e-05}
{"train_loss": 0.5214536190032959, "global_step": 25403, "epoch": 285, "lr": 9.792995421155462e-05}
{"train_loss": 0.5201613306999207, "global_step": 25404, "epoch": 285, "lr": 9.792978912300783e-05}
{"train_loss": 0.33995354175567627, "global_step": 25405, "epoch": 285, "lr": 9.792962402801747e-05}
{"train_loss": 0.5455458164215088, "global_step": 25406, "epoch": 285, "lr": 9.792945892658353e-05}
{"train_loss": 0.45967575907707214, "global_step": 25407, "epoch": 285, "lr": 9.792929381870608e-05}
{"train_loss": 0.5328015089035034, "global_step": 25408, "epoch": 285, "lr": 9.792912870438513e-05}
{"train_loss": 0.4660482406616211, "global_step": 25409, "epoch": 285, "lr": 9.792896358362067e-05}
{"train_loss": 0.39960166811943054, "global_step": 25410, "epoch": 285, "lr": 9.792879845641276e-05}
{"train_loss": 0.4152155816555023, "global_step": 25411, "epoch": 285, "lr": 9.79286333227614e-05}
{"train_loss": 0.3722535967826843, "global_step": 25412, "epoch": 285, "lr": 9.792846818266663e-05}
{"train_loss": 0.38529136776924133, "global_step": 25413, "epoch": 285, "lr": 9.792830303612845e-05}
{"train_loss": 0.4766870439052582, "global_step": 25414, "epoch": 285, "lr": 9.792813788314689e-05}
{"train_loss": 0.3966951370239258, "global_step": 25415, "epoch": 285, "lr": 9.792797272372198e-05}
{"train_loss": 0.37319129705429077, "global_step": 25416, "epoch": 285, "lr": 9.792780755785374e-05}
{"train_loss": 0.4606247544288635, "global_step": 25417, "epoch": 285, "lr": 9.792764238554219e-05}
{"train_loss": 0.3219066858291626, "global_step": 25418, "epoch": 285, "lr": 9.792747720678736e-05}
{"train_loss": 0.4627540111541748, "global_step": 25419, "epoch": 285, "lr": 9.792731202158927e-05}
{"train_loss": 0.39659279584884644, "global_step": 25420, "epoch": 285, "lr": 9.792714682994793e-05}
{"train_loss": 0.4589728116989136, "global_step": 25421, "epoch": 285, "lr": 9.792698163186336e-05}
{"train_loss": 0.4144194424152374, "global_step": 25422, "epoch": 285, "lr": 9.792681642733559e-05}
{"train_loss": 0.4896448850631714, "global_step": 25423, "epoch": 285, "lr": 9.792665121636466e-05}
{"train_loss": 0.45851707458496094, "global_step": 25424, "epoch": 285, "lr": 9.792648599895055e-05}
{"train_loss": 0.3677743375301361, "global_step": 25425, "epoch": 285, "lr": 9.792632077509333e-05}
{"train_loss": 0.446987509727478, "global_step": 25426, "epoch": 285, "lr": 9.7926155544793e-05}
{"train_loss": 0.3661007285118103, "global_step": 25427, "epoch": 285, "lr": 9.792599030804958e-05}
{"train_loss": 0.37679076194763184, "global_step": 25428, "epoch": 285, "lr": 9.792582506486309e-05}
{"train_loss": 0.4666179120540619, "global_step": 25429, "epoch": 285, "lr": 9.792565981523356e-05}
{"train_loss": 0.52407306432724, "global_step": 25430, "epoch": 285, "lr": 9.792549455916101e-05}
{"train_loss": 0.4787554442882538, "global_step": 25431, "epoch": 285, "lr": 9.792532929664546e-05}
{"train_loss": 0.47937247157096863, "global_step": 25432, "epoch": 285, "lr": 9.792516402768692e-05}
{"train_loss": 0.6151723265647888, "global_step": 25433, "epoch": 285, "lr": 9.792499875228544e-05}
{"train_loss": 0.37674176692962646, "global_step": 25434, "epoch": 285, "lr": 9.792483347044104e-05}
{"train_loss": 0.46681898832321167, "global_step": 25435, "epoch": 285, "lr": 9.792466818215371e-05}
{"train_loss": 0.42760276794433594, "global_step": 25436, "epoch": 285, "lr": 9.79245028874235e-05}
{"train_loss": 0.3287610411643982, "global_step": 25437, "epoch": 285, "lr": 9.792433758625044e-05}
{"train_loss": 0.43771788477897644, "global_step": 25438, "epoch": 285, "lr": 9.792417227863451e-05}
{"train_loss": 0.4305926561355591, "global_step": 25439, "epoch": 285, "lr": 9.792400696457577e-05}
{"train_loss": 0.49021440744400024, "global_step": 25440, "epoch": 285, "lr": 9.792384164407424e-05}
{"train_loss": 0.3843994140625, "global_step": 25441, "epoch": 285, "lr": 9.792367631712992e-05}
{"train_loss": 0.46068814396858215, "global_step": 25442, "epoch": 285, "lr": 9.792351098374284e-05}
{"train_loss": 0.4945344626903534, "global_step": 25443, "epoch": 285, "lr": 9.792334564391306e-05}
{"train_loss": 0.3575435280799866, "global_step": 25444, "epoch": 285, "lr": 9.792318029764054e-05}
{"train_loss": 0.47407960891723633, "global_step": 25445, "epoch": 285, "lr": 9.792301494492535e-05}
{"train_loss": 0.48025599122047424, "global_step": 25446, "epoch": 285, "lr": 9.792284958576748e-05}
{"train_loss": 0.47585824131965637, "global_step": 25447, "epoch": 285, "lr": 9.792268422016698e-05}
{"train_loss": 0.5027785897254944, "global_step": 25448, "epoch": 285, "lr": 9.792251884812385e-05}
{"train_loss": 0.44017279148101807, "global_step": 25449, "epoch": 285, "lr": 9.792235346963812e-05}
{"train_loss": 0.4909976124763489, "global_step": 25450, "epoch": 285, "lr": 9.792218808470982e-05}
{"train_loss": 0.41497695446014404, "global_step": 25451, "epoch": 285, "lr": 9.792202269333896e-05}
{"train_loss": 0.4064273536205292, "global_step": 25452, "epoch": 285, "lr": 9.792185729552557e-05}
{"train_loss": 0.43390551376878544, "global_step": 25453, "epoch": 285, "lr": 9.792169189126967e-05, "val_loss": 2.019747734069824, "train_action_mse_error": 31.065540313720703}
{"train_loss": 0.4017864167690277, "global_step": 25454, "epoch": 286, "lr": 9.79215264805713e-05}
{"train_loss": 0.4858415126800537, "global_step": 25455, "epoch": 286, "lr": 9.792136106343046e-05}
{"train_loss": 0.42740994691848755, "global_step": 25456, "epoch": 286, "lr": 9.792119563984716e-05}
{"train_loss": 0.3651576340198517, "global_step": 25457, "epoch": 286, "lr": 9.792103020982145e-05}
{"train_loss": 0.43676942586898804, "global_step": 25458, "epoch": 286, "lr": 9.792086477335333e-05}
{"train_loss": 0.40284809470176697, "global_step": 25459, "epoch": 286, "lr": 9.792069933044285e-05}
{"train_loss": 0.3700137138366699, "global_step": 25460, "epoch": 286, "lr": 9.792053388109e-05}
{"train_loss": 0.3827047646045685, "global_step": 25461, "epoch": 286, "lr": 9.792036842529483e-05}
{"train_loss": 0.4867386519908905, "global_step": 25462, "epoch": 286, "lr": 9.792020296305736e-05}
{"train_loss": 0.4069727659225464, "global_step": 25463, "epoch": 286, "lr": 9.792003749437757e-05}
{"train_loss": 0.4071975350379944, "global_step": 25464, "epoch": 286, "lr": 9.791987201925555e-05}
{"train_loss": 0.3325856029987335, "global_step": 25465, "epoch": 286, "lr": 9.791970653769126e-05}
{"train_loss": 0.43399858474731445, "global_step": 25466, "epoch": 286, "lr": 9.791954104968477e-05}
{"train_loss": 0.4973924458026886, "global_step": 25467, "epoch": 286, "lr": 9.791937555523607e-05}
{"train_loss": 0.4531644582748413, "global_step": 25468, "epoch": 286, "lr": 9.791921005434518e-05}
{"train_loss": 0.3257572650909424, "global_step": 25469, "epoch": 286, "lr": 9.791904454701218e-05}
{"train_loss": 0.5748729705810547, "global_step": 25470, "epoch": 286, "lr": 9.791887903323701e-05}
{"train_loss": 0.326557993888855, "global_step": 25471, "epoch": 286, "lr": 9.791871351301974e-05}
{"train_loss": 0.3903379738330841, "global_step": 25472, "epoch": 286, "lr": 9.791854798636038e-05}
{"train_loss": 0.43863409757614136, "global_step": 25473, "epoch": 286, "lr": 9.791838245325897e-05}
{"train_loss": 0.39256224036216736, "global_step": 25474, "epoch": 286, "lr": 9.791821691371551e-05}
{"train_loss": 0.4231785237789154, "global_step": 25475, "epoch": 286, "lr": 9.791805136773002e-05}
{"train_loss": 0.37332090735435486, "global_step": 25476, "epoch": 286, "lr": 9.791788581530254e-05}
{"train_loss": 0.4365432858467102, "global_step": 25477, "epoch": 286, "lr": 9.791772025643309e-05}
{"train_loss": 0.4872265160083771, "global_step": 25478, "epoch": 286, "lr": 9.791755469112168e-05}
{"train_loss": 0.3971903324127197, "global_step": 25479, "epoch": 286, "lr": 9.791738911936834e-05}
{"train_loss": 0.45501208305358887, "global_step": 25480, "epoch": 286, "lr": 9.79172235411731e-05}
{"train_loss": 0.36141443252563477, "global_step": 25481, "epoch": 286, "lr": 9.791705795653596e-05}
{"train_loss": 0.39096391201019287, "global_step": 25482, "epoch": 286, "lr": 9.791689236545696e-05}
{"train_loss": 0.33107852935791016, "global_step": 25483, "epoch": 286, "lr": 9.791672676793612e-05}
{"train_loss": 0.4074389338493347, "global_step": 25484, "epoch": 286, "lr": 9.791656116397345e-05}
{"train_loss": 0.3465307056903839, "global_step": 25485, "epoch": 286, "lr": 9.7916395553569e-05}
{"train_loss": 0.4958767890930176, "global_step": 25486, "epoch": 286, "lr": 9.791622993672275e-05}
{"train_loss": 0.37615954875946045, "global_step": 25487, "epoch": 286, "lr": 9.791606431343477e-05}
{"train_loss": 0.387484073638916, "global_step": 25488, "epoch": 286, "lr": 9.791589868370505e-05}
{"train_loss": 0.29262611269950867, "global_step": 25489, "epoch": 286, "lr": 9.791573304753363e-05}
{"train_loss": 0.33043473958969116, "global_step": 25490, "epoch": 286, "lr": 9.791556740492052e-05}
{"train_loss": 0.38406360149383545, "global_step": 25491, "epoch": 286, "lr": 9.791540175586573e-05}
{"train_loss": 0.49304312467575073, "global_step": 25492, "epoch": 286, "lr": 9.791523610036932e-05}
{"train_loss": 0.4598314166069031, "global_step": 25493, "epoch": 286, "lr": 9.791507043843128e-05}
{"train_loss": 0.5304839015007019, "global_step": 25494, "epoch": 286, "lr": 9.791490477005164e-05}
{"train_loss": 0.42491984367370605, "global_step": 25495, "epoch": 286, "lr": 9.791473909523044e-05}
{"train_loss": 0.5223977565765381, "global_step": 25496, "epoch": 286, "lr": 9.791457341396766e-05}
{"train_loss": 0.5202569365501404, "global_step": 25497, "epoch": 286, "lr": 9.791440772626337e-05}
{"train_loss": 0.5187309384346008, "global_step": 25498, "epoch": 286, "lr": 9.791424203211756e-05}
{"train_loss": 0.5035158395767212, "global_step": 25499, "epoch": 286, "lr": 9.791407633153027e-05}
{"train_loss": 0.48475584387779236, "global_step": 25500, "epoch": 286, "lr": 9.791391062450153e-05}
{"train_loss": 0.47369539737701416, "global_step": 25501, "epoch": 286, "lr": 9.791374491103133e-05}
{"train_loss": 0.4468892514705658, "global_step": 25502, "epoch": 286, "lr": 9.791357919111972e-05}
{"train_loss": 0.5552651286125183, "global_step": 25503, "epoch": 286, "lr": 9.791341346476672e-05}
{"train_loss": 0.6079800128936768, "global_step": 25504, "epoch": 286, "lr": 9.791324773197233e-05}
{"train_loss": 0.46027126908302307, "global_step": 25505, "epoch": 286, "lr": 9.79130819927366e-05}
{"train_loss": 0.506671667098999, "global_step": 25506, "epoch": 286, "lr": 9.791291624705954e-05}
{"train_loss": 0.44444796442985535, "global_step": 25507, "epoch": 286, "lr": 9.791275049494116e-05}
{"train_loss": 0.44315868616104126, "global_step": 25508, "epoch": 286, "lr": 9.79125847363815e-05}
{"train_loss": 0.4203624129295349, "global_step": 25509, "epoch": 286, "lr": 9.791241897138059e-05}
{"train_loss": 0.49124670028686523, "global_step": 25510, "epoch": 286, "lr": 9.791225319993841e-05}
{"train_loss": 0.4723265469074249, "global_step": 25511, "epoch": 286, "lr": 9.791208742205505e-05}
{"train_loss": 0.593680202960968, "global_step": 25512, "epoch": 286, "lr": 9.791192163773047e-05}
{"train_loss": 0.43195584416389465, "global_step": 25513, "epoch": 286, "lr": 9.791175584696471e-05}
{"train_loss": 0.47536906599998474, "global_step": 25514, "epoch": 286, "lr": 9.791159004975781e-05}
{"train_loss": 0.40046483278274536, "global_step": 25515, "epoch": 286, "lr": 9.791142424610978e-05}
{"train_loss": 0.4326320290565491, "global_step": 25516, "epoch": 286, "lr": 9.791125843602064e-05}
{"train_loss": 0.4597885012626648, "global_step": 25517, "epoch": 286, "lr": 9.791109261949041e-05}
{"train_loss": 0.542900562286377, "global_step": 25518, "epoch": 286, "lr": 9.791092679651912e-05}
{"train_loss": 0.5441041588783264, "global_step": 25519, "epoch": 286, "lr": 9.791076096710681e-05}
{"train_loss": 0.5135457515716553, "global_step": 25520, "epoch": 286, "lr": 9.791059513125346e-05}
{"train_loss": 0.4890758991241455, "global_step": 25521, "epoch": 286, "lr": 9.791042928895913e-05}
{"train_loss": 0.48197224736213684, "global_step": 25522, "epoch": 286, "lr": 9.791026344022381e-05}
{"train_loss": 0.501168429851532, "global_step": 25523, "epoch": 286, "lr": 9.791009758504755e-05}
{"train_loss": 0.45134663581848145, "global_step": 25524, "epoch": 286, "lr": 9.790993172343034e-05}
{"train_loss": 0.39643219113349915, "global_step": 25525, "epoch": 286, "lr": 9.790976585537226e-05}
{"train_loss": 0.6018120646476746, "global_step": 25526, "epoch": 286, "lr": 9.790959998087327e-05}
{"train_loss": 0.4708377718925476, "global_step": 25527, "epoch": 286, "lr": 9.790943409993342e-05}
{"train_loss": 0.5069396495819092, "global_step": 25528, "epoch": 286, "lr": 9.790926821255274e-05}
{"train_loss": 0.504391610622406, "global_step": 25529, "epoch": 286, "lr": 9.790910231873123e-05}
{"train_loss": 0.376578688621521, "global_step": 25530, "epoch": 286, "lr": 9.790893641846894e-05}
{"train_loss": 0.373782217502594, "global_step": 25531, "epoch": 286, "lr": 9.790877051176588e-05}
{"train_loss": 0.4019957482814789, "global_step": 25532, "epoch": 286, "lr": 9.790860459862206e-05}
{"train_loss": 0.31639307737350464, "global_step": 25533, "epoch": 286, "lr": 9.790843867903751e-05}
{"train_loss": 0.42413222789764404, "global_step": 25534, "epoch": 286, "lr": 9.790827275301226e-05}
{"train_loss": 0.37307122349739075, "global_step": 25535, "epoch": 286, "lr": 9.790810682054632e-05}
{"train_loss": 0.517413854598999, "global_step": 25536, "epoch": 286, "lr": 9.790794088163973e-05}
{"train_loss": 0.3882904052734375, "global_step": 25537, "epoch": 286, "lr": 9.790777493629249e-05}
{"train_loss": 0.3990781009197235, "global_step": 25538, "epoch": 286, "lr": 9.790760898450464e-05}
{"train_loss": 0.42552387714385986, "global_step": 25539, "epoch": 286, "lr": 9.79074430262762e-05}
{"train_loss": 0.4190019369125366, "global_step": 25540, "epoch": 286, "lr": 9.790727706160717e-05}
{"train_loss": 0.38122889399528503, "global_step": 25541, "epoch": 286, "lr": 9.790711109049761e-05}
{"train_loss": 0.44251219844550227, "global_step": 25542, "epoch": 286, "lr": 9.79069451129475e-05, "val_loss": 2.0252110958099365}
{"train_loss": 0.5459970235824585, "global_step": 25543, "epoch": 287, "lr": 9.790677912895692e-05}
{"train_loss": 0.3827349841594696, "global_step": 25544, "epoch": 287, "lr": 9.790661313852583e-05}
{"train_loss": 0.42749080061912537, "global_step": 25545, "epoch": 287, "lr": 9.790644714165429e-05}
{"train_loss": 0.4627906084060669, "global_step": 25546, "epoch": 287, "lr": 9.790628113834231e-05}
{"train_loss": 0.39508941769599915, "global_step": 25547, "epoch": 287, "lr": 9.790611512858991e-05}
{"train_loss": 0.4119216203689575, "global_step": 25548, "epoch": 287, "lr": 9.790594911239712e-05}
{"train_loss": 0.4450814723968506, "global_step": 25549, "epoch": 287, "lr": 9.790578308976396e-05}
{"train_loss": 0.3631296455860138, "global_step": 25550, "epoch": 287, "lr": 9.790561706069046e-05}
{"train_loss": 0.43726763129234314, "global_step": 25551, "epoch": 287, "lr": 9.79054510251766e-05}
{"train_loss": 0.30698835849761963, "global_step": 25552, "epoch": 287, "lr": 9.790528498322248e-05}
{"train_loss": 0.5307580232620239, "global_step": 25553, "epoch": 287, "lr": 9.790511893482806e-05}
{"train_loss": 0.42218536138534546, "global_step": 25554, "epoch": 287, "lr": 9.790495287999338e-05}
{"train_loss": 0.3366442322731018, "global_step": 25555, "epoch": 287, "lr": 9.790478681871846e-05}
{"train_loss": 0.4218873977661133, "global_step": 25556, "epoch": 287, "lr": 9.790462075100332e-05}
{"train_loss": 0.37738820910453796, "global_step": 25557, "epoch": 287, "lr": 9.790445467684801e-05}
{"train_loss": 0.30902454257011414, "global_step": 25558, "epoch": 287, "lr": 9.790428859625252e-05}
{"train_loss": 0.31051644682884216, "global_step": 25559, "epoch": 287, "lr": 9.790412250921687e-05}
{"train_loss": 0.3975038230419159, "global_step": 25560, "epoch": 287, "lr": 9.790395641574111e-05}
{"train_loss": 0.42253828048706055, "global_step": 25561, "epoch": 287, "lr": 9.790379031582522e-05}
{"train_loss": 0.41376087069511414, "global_step": 25562, "epoch": 287, "lr": 9.790362420946928e-05}
{"train_loss": 0.4089152216911316, "global_step": 25563, "epoch": 287, "lr": 9.790345809667327e-05}
{"train_loss": 0.4751778841018677, "global_step": 25564, "epoch": 287, "lr": 9.790329197743721e-05}
{"train_loss": 0.5028314590454102, "global_step": 25565, "epoch": 287, "lr": 9.790312585176116e-05}
{"train_loss": 0.4425690770149231, "global_step": 25566, "epoch": 287, "lr": 9.790295971964512e-05}
{"train_loss": 0.3728086054325104, "global_step": 25567, "epoch": 287, "lr": 9.790279358108909e-05}
{"train_loss": 0.4435043931007385, "global_step": 25568, "epoch": 287, "lr": 9.790262743609312e-05}
{"train_loss": 0.4746401906013489, "global_step": 25569, "epoch": 287, "lr": 9.790246128465723e-05}
{"train_loss": 0.4293772280216217, "global_step": 25570, "epoch": 287, "lr": 9.790229512678144e-05}
{"train_loss": 0.43897706270217896, "global_step": 25571, "epoch": 287, "lr": 9.790212896246577e-05}
{"train_loss": 0.3538135290145874, "global_step": 25572, "epoch": 287, "lr": 9.790196279171025e-05}
{"train_loss": 0.46607279777526855, "global_step": 25573, "epoch": 287, "lr": 9.790179661451487e-05}
{"train_loss": 0.39923185110092163, "global_step": 25574, "epoch": 287, "lr": 9.79016304308797e-05}
{"train_loss": 0.42464879155158997, "global_step": 25575, "epoch": 287, "lr": 9.790146424080473e-05}
{"train_loss": 0.4310626983642578, "global_step": 25576, "epoch": 287, "lr": 9.790129804429e-05}
{"train_loss": 0.4501263499259949, "global_step": 25577, "epoch": 287, "lr": 9.790113184133551e-05}
{"train_loss": 0.3597969114780426, "global_step": 25578, "epoch": 287, "lr": 9.79009656319413e-05}
{"train_loss": 0.459381103515625, "global_step": 25579, "epoch": 287, "lr": 9.790079941610739e-05}
{"train_loss": 0.5428841710090637, "global_step": 25580, "epoch": 287, "lr": 9.79006331938338e-05}
{"train_loss": 0.4356054663658142, "global_step": 25581, "epoch": 287, "lr": 9.790046696512056e-05}
{"train_loss": 0.3993948996067047, "global_step": 25582, "epoch": 287, "lr": 9.79003007299677e-05}
{"train_loss": 0.37499332427978516, "global_step": 25583, "epoch": 287, "lr": 9.790013448837521e-05}
{"train_loss": 0.3494528532028198, "global_step": 25584, "epoch": 287, "lr": 9.789996824034313e-05}
{"train_loss": 0.3629428446292877, "global_step": 25585, "epoch": 287, "lr": 9.789980198587149e-05}
{"train_loss": 0.4343850612640381, "global_step": 25586, "epoch": 287, "lr": 9.789963572496031e-05}
{"train_loss": 0.391570121049881, "global_step": 25587, "epoch": 287, "lr": 9.789946945760959e-05}
{"train_loss": 0.5315592885017395, "global_step": 25588, "epoch": 287, "lr": 9.789930318381938e-05}
{"train_loss": 0.3347713351249695, "global_step": 25589, "epoch": 287, "lr": 9.78991369035897e-05}
{"train_loss": 0.3737871050834656, "global_step": 25590, "epoch": 287, "lr": 9.789897061692055e-05}
{"train_loss": 0.5607661604881287, "global_step": 25591, "epoch": 287, "lr": 9.789880432381197e-05}
{"train_loss": 0.449258416891098, "global_step": 25592, "epoch": 287, "lr": 9.789863802426399e-05}
{"train_loss": 0.42237457633018494, "global_step": 25593, "epoch": 287, "lr": 9.789847171827661e-05}
{"train_loss": 0.39003393054008484, "global_step": 25594, "epoch": 287, "lr": 9.789830540584988e-05}
{"train_loss": 0.5121158957481384, "global_step": 25595, "epoch": 287, "lr": 9.78981390869838e-05}
{"train_loss": 0.5394574403762817, "global_step": 25596, "epoch": 287, "lr": 9.789797276167838e-05}
{"train_loss": 0.4098479151725769, "global_step": 25597, "epoch": 287, "lr": 9.78978064299337e-05}
{"train_loss": 0.47489461302757263, "global_step": 25598, "epoch": 287, "lr": 9.78976400917497e-05}
{"train_loss": 0.5387524366378784, "global_step": 25599, "epoch": 287, "lr": 9.789747374712648e-05}
{"train_loss": 0.40223464369773865, "global_step": 25600, "epoch": 287, "lr": 9.789730739606401e-05}
{"train_loss": 0.4339718222618103, "global_step": 25601, "epoch": 287, "lr": 9.789714103856234e-05}
{"train_loss": 0.4541761577129364, "global_step": 25602, "epoch": 287, "lr": 9.789697467462147e-05}
{"train_loss": 0.48102596402168274, "global_step": 25603, "epoch": 287, "lr": 9.789680830424145e-05}
{"train_loss": 0.38979387283325195, "global_step": 25604, "epoch": 287, "lr": 9.789664192742228e-05}
{"train_loss": 0.47975990176200867, "global_step": 25605, "epoch": 287, "lr": 9.7896475544164e-05}
{"train_loss": 0.46548667550086975, "global_step": 25606, "epoch": 287, "lr": 9.789630915446662e-05}
{"train_loss": 0.38510459661483765, "global_step": 25607, "epoch": 287, "lr": 9.789614275833015e-05}
{"train_loss": 0.3948381841182709, "global_step": 25608, "epoch": 287, "lr": 9.789597635575465e-05}
{"train_loss": 0.608717679977417, "global_step": 25609, "epoch": 287, "lr": 9.78958099467401e-05}
{"train_loss": 0.30930373072624207, "global_step": 25610, "epoch": 287, "lr": 9.789564353128655e-05}
{"train_loss": 0.3506922125816345, "global_step": 25611, "epoch": 287, "lr": 9.789547710939402e-05}
{"train_loss": 0.3695446848869324, "global_step": 25612, "epoch": 287, "lr": 9.789531068106252e-05}
{"train_loss": 0.45737776160240173, "global_step": 25613, "epoch": 287, "lr": 9.789514424629209e-05}
{"train_loss": 0.3135516047477722, "global_step": 25614, "epoch": 287, "lr": 9.789497780508272e-05}
{"train_loss": 0.4883710741996765, "global_step": 25615, "epoch": 287, "lr": 9.789481135743447e-05}
{"train_loss": 0.374828964471817, "global_step": 25616, "epoch": 287, "lr": 9.789464490334734e-05}
{"train_loss": 0.33706289529800415, "global_step": 25617, "epoch": 287, "lr": 9.789447844282135e-05}
{"train_loss": 0.5216367244720459, "global_step": 25618, "epoch": 287, "lr": 9.789431197585656e-05}
{"train_loss": 0.4675508439540863, "global_step": 25619, "epoch": 287, "lr": 9.789414550245294e-05}
{"train_loss": 0.3946000933647156, "global_step": 25620, "epoch": 287, "lr": 9.789397902261055e-05}
{"train_loss": 0.3545507490634918, "global_step": 25621, "epoch": 287, "lr": 9.789381253632939e-05}
{"train_loss": 0.436842143535614, "global_step": 25622, "epoch": 287, "lr": 9.789364604360948e-05}
{"train_loss": 0.3171311914920807, "global_step": 25623, "epoch": 287, "lr": 9.789347954445087e-05}
{"train_loss": 0.4265379309654236, "global_step": 25624, "epoch": 287, "lr": 9.789331303885357e-05}
{"train_loss": 0.39307746291160583, "global_step": 25625, "epoch": 287, "lr": 9.789314652681758e-05}
{"train_loss": 0.470608651638031, "global_step": 25626, "epoch": 287, "lr": 9.789298000834295e-05}
{"train_loss": 0.4926488399505615, "global_step": 25627, "epoch": 287, "lr": 9.78928134834297e-05}
{"train_loss": 0.45278093218803406, "global_step": 25628, "epoch": 287, "lr": 9.789264695207781e-05}
{"train_loss": 0.5277599692344666, "global_step": 25629, "epoch": 287, "lr": 9.789248041428736e-05}
{"train_loss": 0.39910754561424255, "global_step": 25630, "epoch": 287, "lr": 9.789231387005837e-05}
{"train_loss": 0.42590079381224816, "global_step": 25631, "epoch": 287, "lr": 9.789214731939082e-05, "val_loss": 1.9716676473617554}
{"train_loss": 0.4480765461921692, "global_step": 25632, "epoch": 288, "lr": 9.789198076228477e-05}
{"train_loss": 0.37167879939079285, "global_step": 25633, "epoch": 288, "lr": 9.789181419874021e-05}
{"train_loss": 0.3373691439628601, "global_step": 25634, "epoch": 288, "lr": 9.789164762875718e-05}
{"train_loss": 0.3815484642982483, "global_step": 25635, "epoch": 288, "lr": 9.78914810523357e-05}
{"train_loss": 0.37896305322647095, "global_step": 25636, "epoch": 288, "lr": 9.78913144694758e-05}
{"train_loss": 0.4108372628688812, "global_step": 25637, "epoch": 288, "lr": 9.78911478801775e-05}
{"train_loss": 0.412435919046402, "global_step": 25638, "epoch": 288, "lr": 9.789098128444082e-05}
{"train_loss": 0.4642418324947357, "global_step": 25639, "epoch": 288, "lr": 9.789081468226578e-05}
{"train_loss": 0.42492276430130005, "global_step": 25640, "epoch": 288, "lr": 9.78906480736524e-05}
{"train_loss": 0.3334643542766571, "global_step": 25641, "epoch": 288, "lr": 9.78904814586007e-05}
{"train_loss": 0.4925243556499481, "global_step": 25642, "epoch": 288, "lr": 9.789031483711071e-05}
{"train_loss": 0.3394519090652466, "global_step": 25643, "epoch": 288, "lr": 9.789014820918246e-05}
{"train_loss": 0.46086031198501587, "global_step": 25644, "epoch": 288, "lr": 9.788998157481597e-05}
{"train_loss": 0.3108101785182953, "global_step": 25645, "epoch": 288, "lr": 9.788981493401123e-05}
{"train_loss": 0.5252779126167297, "global_step": 25646, "epoch": 288, "lr": 9.788964828676833e-05}
{"train_loss": 0.3973233699798584, "global_step": 25647, "epoch": 288, "lr": 9.78894816330872e-05}
{"train_loss": 0.3780171871185303, "global_step": 25648, "epoch": 288, "lr": 9.788931497296794e-05}
{"train_loss": 0.4312800168991089, "global_step": 25649, "epoch": 288, "lr": 9.788914830641054e-05}
{"train_loss": 0.31403881311416626, "global_step": 25650, "epoch": 288, "lr": 9.788898163341504e-05}
{"train_loss": 0.43555665016174316, "global_step": 25651, "epoch": 288, "lr": 9.788881495398144e-05}
{"train_loss": 0.3717268407344818, "global_step": 25652, "epoch": 288, "lr": 9.788864826810977e-05}
{"train_loss": 0.4116032123565674, "global_step": 25653, "epoch": 288, "lr": 9.788848157580007e-05}
{"train_loss": 0.4670875072479248, "global_step": 25654, "epoch": 288, "lr": 9.788831487705232e-05}
{"train_loss": 0.29766514897346497, "global_step": 25655, "epoch": 288, "lr": 9.788814817186658e-05}
{"train_loss": 0.332975834608078, "global_step": 25656, "epoch": 288, "lr": 9.788798146024288e-05}
{"train_loss": 0.3468185365200043, "global_step": 25657, "epoch": 288, "lr": 9.788781474218121e-05}
{"train_loss": 0.398549884557724, "global_step": 25658, "epoch": 288, "lr": 9.78876480176816e-05}
{"train_loss": 0.3675423562526703, "global_step": 25659, "epoch": 288, "lr": 9.788748128674408e-05}
{"train_loss": 0.3326934576034546, "global_step": 25660, "epoch": 288, "lr": 9.788731454936868e-05}
{"train_loss": 0.4349316656589508, "global_step": 25661, "epoch": 288, "lr": 9.788714780555541e-05}
{"train_loss": 0.34303978085517883, "global_step": 25662, "epoch": 288, "lr": 9.788698105530431e-05}
{"train_loss": 0.4797661602497101, "global_step": 25663, "epoch": 288, "lr": 9.788681429861538e-05}
{"train_loss": 0.5088688135147095, "global_step": 25664, "epoch": 288, "lr": 9.788664753548865e-05}
{"train_loss": 0.4281221032142639, "global_step": 25665, "epoch": 288, "lr": 9.788648076592415e-05}
{"train_loss": 0.3388687074184418, "global_step": 25666, "epoch": 288, "lr": 9.788631398992189e-05}
{"train_loss": 0.37879085540771484, "global_step": 25667, "epoch": 288, "lr": 9.78861472074819e-05}
{"train_loss": 0.42204684019088745, "global_step": 25668, "epoch": 288, "lr": 9.78859804186042e-05}
{"train_loss": 0.41812264919281006, "global_step": 25669, "epoch": 288, "lr": 9.788581362328881e-05}
{"train_loss": 0.5179849863052368, "global_step": 25670, "epoch": 288, "lr": 9.788564682153576e-05}
{"train_loss": 0.3581944406032562, "global_step": 25671, "epoch": 288, "lr": 9.788548001334507e-05}
{"train_loss": 0.477273166179657, "global_step": 25672, "epoch": 288, "lr": 9.788531319871676e-05}
{"train_loss": 0.5227755904197693, "global_step": 25673, "epoch": 288, "lr": 9.788514637765085e-05}
{"train_loss": 0.37286052107810974, "global_step": 25674, "epoch": 288, "lr": 9.788497955014738e-05}
{"train_loss": 0.3422672152519226, "global_step": 25675, "epoch": 288, "lr": 9.788481271620633e-05}
{"train_loss": 0.3491669297218323, "global_step": 25676, "epoch": 288, "lr": 9.788464587582777e-05}
{"train_loss": 0.3901393711566925, "global_step": 25677, "epoch": 288, "lr": 9.78844790290117e-05}
{"train_loss": 0.3413228392601013, "global_step": 25678, "epoch": 288, "lr": 9.788431217575816e-05}
{"train_loss": 0.4258479177951813, "global_step": 25679, "epoch": 288, "lr": 9.788414531606714e-05}
{"train_loss": 0.4053240716457367, "global_step": 25680, "epoch": 288, "lr": 9.788397844993869e-05}
{"train_loss": 0.46933838725090027, "global_step": 25681, "epoch": 288, "lr": 9.788381157737281e-05}
{"train_loss": 0.4296149015426636, "global_step": 25682, "epoch": 288, "lr": 9.788364469836953e-05}
{"train_loss": 0.40697968006134033, "global_step": 25683, "epoch": 288, "lr": 9.788347781292891e-05}
{"train_loss": 0.46860745549201965, "global_step": 25684, "epoch": 288, "lr": 9.788331092105091e-05}
{"train_loss": 0.4054919481277466, "global_step": 25685, "epoch": 288, "lr": 9.788314402273559e-05}
{"train_loss": 0.41063329577445984, "global_step": 25686, "epoch": 288, "lr": 9.788297711798297e-05}
{"train_loss": 0.38957837224006653, "global_step": 25687, "epoch": 288, "lr": 9.788281020679306e-05}
{"train_loss": 0.4196244180202484, "global_step": 25688, "epoch": 288, "lr": 9.78826432891659e-05}
{"train_loss": 0.4866181015968323, "global_step": 25689, "epoch": 288, "lr": 9.788247636510148e-05}
{"train_loss": 0.3600005507469177, "global_step": 25690, "epoch": 288, "lr": 9.788230943459987e-05}
{"train_loss": 0.42372578382492065, "global_step": 25691, "epoch": 288, "lr": 9.788214249766106e-05}
{"train_loss": 0.45997652411460876, "global_step": 25692, "epoch": 288, "lr": 9.788197555428508e-05}
{"train_loss": 0.35855528712272644, "global_step": 25693, "epoch": 288, "lr": 9.788180860447195e-05}
{"train_loss": 0.44100210070610046, "global_step": 25694, "epoch": 288, "lr": 9.788164164822169e-05}
{"train_loss": 0.4142337441444397, "global_step": 25695, "epoch": 288, "lr": 9.788147468553432e-05}
{"train_loss": 0.3257477879524231, "global_step": 25696, "epoch": 288, "lr": 9.788130771640988e-05}
{"train_loss": 0.33597126603126526, "global_step": 25697, "epoch": 288, "lr": 9.78811407408484e-05}
{"train_loss": 0.40082570910453796, "global_step": 25698, "epoch": 288, "lr": 9.788097375884986e-05}
{"train_loss": 0.3901108503341675, "global_step": 25699, "epoch": 288, "lr": 9.788080677041431e-05}
{"train_loss": 0.49713125824928284, "global_step": 25700, "epoch": 288, "lr": 9.788063977554177e-05}
{"train_loss": 0.3798905313014984, "global_step": 25701, "epoch": 288, "lr": 9.788047277423227e-05}
{"train_loss": 0.38711848855018616, "global_step": 25702, "epoch": 288, "lr": 9.788030576648581e-05}
{"train_loss": 0.49446889758110046, "global_step": 25703, "epoch": 288, "lr": 9.788013875230244e-05}
{"train_loss": 0.4198162257671356, "global_step": 25704, "epoch": 288, "lr": 9.787997173168216e-05}
{"train_loss": 0.3494190275669098, "global_step": 25705, "epoch": 288, "lr": 9.787980470462501e-05}
{"train_loss": 0.41007235646247864, "global_step": 25706, "epoch": 288, "lr": 9.7879637671131e-05}
{"train_loss": 0.3884564936161041, "global_step": 25707, "epoch": 288, "lr": 9.787947063120015e-05}
{"train_loss": 0.3337653577327728, "global_step": 25708, "epoch": 288, "lr": 9.787930358483249e-05}
{"train_loss": 0.38628947734832764, "global_step": 25709, "epoch": 288, "lr": 9.787913653202805e-05}
{"train_loss": 0.42570754885673523, "global_step": 25710, "epoch": 288, "lr": 9.787896947278683e-05}
{"train_loss": 0.30180782079696655, "global_step": 25711, "epoch": 288, "lr": 9.787880240710887e-05}
{"train_loss": 0.3265698552131653, "global_step": 25712, "epoch": 288, "lr": 9.78786353349942e-05}
{"train_loss": 0.3922054171562195, "global_step": 25713, "epoch": 288, "lr": 9.787846825644282e-05}
{"train_loss": 0.387007474899292, "global_step": 25714, "epoch": 288, "lr": 9.787830117145477e-05}
{"train_loss": 0.3627934753894806, "global_step": 25715, "epoch": 288, "lr": 9.787813408003007e-05}
{"train_loss": 0.40926116704940796, "global_step": 25716, "epoch": 288, "lr": 9.787796698216872e-05}
{"train_loss": 0.40315645933151245, "global_step": 25717, "epoch": 288, "lr": 9.787779987787078e-05}
{"train_loss": 0.4730896055698395, "global_step": 25718, "epoch": 288, "lr": 9.787763276713624e-05}
{"train_loss": 0.3600151240825653, "global_step": 25719, "epoch": 288, "lr": 9.787746564996515e-05}
{"train_loss": 0.40263607562258, "global_step": 25720, "epoch": 288, "lr": 9.78772985263575e-05, "val_loss": 2.0505106449127197}
{"train_loss": 0.4717442989349365, "global_step": 25721, "epoch": 289, "lr": 9.787713139631334e-05}
{"train_loss": 0.37240639328956604, "global_step": 25722, "epoch": 289, "lr": 9.787696425983268e-05}
{"train_loss": 0.4405226409435272, "global_step": 25723, "epoch": 289, "lr": 9.787679711691555e-05}
{"train_loss": 0.3541077971458435, "global_step": 25724, "epoch": 289, "lr": 9.787662996756197e-05}
{"train_loss": 0.411742627620697, "global_step": 25725, "epoch": 289, "lr": 9.787646281177196e-05}
{"train_loss": 0.3196106553077698, "global_step": 25726, "epoch": 289, "lr": 9.787629564954554e-05}
{"train_loss": 0.505816638469696, "global_step": 25727, "epoch": 289, "lr": 9.787612848088275e-05}
{"train_loss": 0.3635987341403961, "global_step": 25728, "epoch": 289, "lr": 9.787596130578359e-05}
{"train_loss": 0.537513017654419, "global_step": 25729, "epoch": 289, "lr": 9.787579412424807e-05}
{"train_loss": 0.48728442192077637, "global_step": 25730, "epoch": 289, "lr": 9.787562693627626e-05}
{"train_loss": 0.47767049074172974, "global_step": 25731, "epoch": 289, "lr": 9.787545974186813e-05}
{"train_loss": 0.33898818492889404, "global_step": 25732, "epoch": 289, "lr": 9.787529254102375e-05}
{"train_loss": 0.5215717554092407, "global_step": 25733, "epoch": 289, "lr": 9.78751253337431e-05}
{"train_loss": 0.48983708024024963, "global_step": 25734, "epoch": 289, "lr": 9.787495812002625e-05}
{"train_loss": 0.46344661712646484, "global_step": 25735, "epoch": 289, "lr": 9.787479089987317e-05}
{"train_loss": 0.40088894963264465, "global_step": 25736, "epoch": 289, "lr": 9.787462367328392e-05}
{"train_loss": 0.45482203364372253, "global_step": 25737, "epoch": 289, "lr": 9.78744564402585e-05}
{"train_loss": 0.48455193638801575, "global_step": 25738, "epoch": 289, "lr": 9.787428920079697e-05}
{"train_loss": 0.3442329466342926, "global_step": 25739, "epoch": 289, "lr": 9.78741219548993e-05}
{"train_loss": 0.39668142795562744, "global_step": 25740, "epoch": 289, "lr": 9.787395470256555e-05}
{"train_loss": 0.3821946680545807, "global_step": 25741, "epoch": 289, "lr": 9.787378744379572e-05}
{"train_loss": 0.48564112186431885, "global_step": 25742, "epoch": 289, "lr": 9.787362017858986e-05}
{"train_loss": 0.36477038264274597, "global_step": 25743, "epoch": 289, "lr": 9.787345290694796e-05}
{"train_loss": 0.32268279790878296, "global_step": 25744, "epoch": 289, "lr": 9.787328562887006e-05}
{"train_loss": 0.4479913115501404, "global_step": 25745, "epoch": 289, "lr": 9.787311834435617e-05}
{"train_loss": 0.45031604170799255, "global_step": 25746, "epoch": 289, "lr": 9.787295105340634e-05}
{"train_loss": 0.3109743595123291, "global_step": 25747, "epoch": 289, "lr": 9.787278375602057e-05}
{"train_loss": 0.3168770968914032, "global_step": 25748, "epoch": 289, "lr": 9.787261645219889e-05}
{"train_loss": 0.4895298480987549, "global_step": 25749, "epoch": 289, "lr": 9.787244914194132e-05}
{"train_loss": 0.3502086400985718, "global_step": 25750, "epoch": 289, "lr": 9.787228182524789e-05}
{"train_loss": 0.386052668094635, "global_step": 25751, "epoch": 289, "lr": 9.787211450211859e-05}
{"train_loss": 0.3818639814853668, "global_step": 25752, "epoch": 289, "lr": 9.787194717255348e-05}
{"train_loss": 0.29210788011550903, "global_step": 25753, "epoch": 289, "lr": 9.787177983655258e-05}
{"train_loss": 0.35996943712234497, "global_step": 25754, "epoch": 289, "lr": 9.787161249411589e-05}
{"train_loss": 0.41355276107788086, "global_step": 25755, "epoch": 289, "lr": 9.787144514524345e-05}
{"train_loss": 0.4448871314525604, "global_step": 25756, "epoch": 289, "lr": 9.787127778993527e-05}
{"train_loss": 0.4180639386177063, "global_step": 25757, "epoch": 289, "lr": 9.78711104281914e-05}
{"train_loss": 0.3796653747558594, "global_step": 25758, "epoch": 289, "lr": 9.787094306001183e-05}
{"train_loss": 0.387085497379303, "global_step": 25759, "epoch": 289, "lr": 9.78707756853966e-05}
{"train_loss": 0.43530911207199097, "global_step": 25760, "epoch": 289, "lr": 9.787060830434571e-05}
{"train_loss": 0.3641023635864258, "global_step": 25761, "epoch": 289, "lr": 9.787044091685921e-05}
{"train_loss": 0.4474164843559265, "global_step": 25762, "epoch": 289, "lr": 9.787027352293714e-05}
{"train_loss": 0.39046159386634827, "global_step": 25763, "epoch": 289, "lr": 9.787010612257946e-05}
{"train_loss": 0.44830965995788574, "global_step": 25764, "epoch": 289, "lr": 9.786993871578625e-05}
{"train_loss": 0.44901934266090393, "global_step": 25765, "epoch": 289, "lr": 9.78697713025575e-05}
{"train_loss": 0.431547611951828, "global_step": 25766, "epoch": 289, "lr": 9.786960388289323e-05}
{"train_loss": 0.4724124073982239, "global_step": 25767, "epoch": 289, "lr": 9.78694364567935e-05}
{"train_loss": 0.36974260210990906, "global_step": 25768, "epoch": 289, "lr": 9.78692690242583e-05}
{"train_loss": 0.5152608752250671, "global_step": 25769, "epoch": 289, "lr": 9.786910158528765e-05}
{"train_loss": 0.38830697536468506, "global_step": 25770, "epoch": 289, "lr": 9.786893413988158e-05}
{"train_loss": 0.3825259506702423, "global_step": 25771, "epoch": 289, "lr": 9.786876668804013e-05}
{"train_loss": 0.3602202534675598, "global_step": 25772, "epoch": 289, "lr": 9.786859922976331e-05}
{"train_loss": 0.4504733979701996, "global_step": 25773, "epoch": 289, "lr": 9.786843176505111e-05}
{"train_loss": 0.3999845087528229, "global_step": 25774, "epoch": 289, "lr": 9.786826429390361e-05}
{"train_loss": 0.43947869539260864, "global_step": 25775, "epoch": 289, "lr": 9.786809681632081e-05}
{"train_loss": 0.41300052404403687, "global_step": 25776, "epoch": 289, "lr": 9.786792933230271e-05}
{"train_loss": 0.4445599913597107, "global_step": 25777, "epoch": 289, "lr": 9.786776184184936e-05}
{"train_loss": 0.436229407787323, "global_step": 25778, "epoch": 289, "lr": 9.786759434496078e-05}
{"train_loss": 0.37650373578071594, "global_step": 25779, "epoch": 289, "lr": 9.786742684163696e-05}
{"train_loss": 0.5139790773391724, "global_step": 25780, "epoch": 289, "lr": 9.786725933187797e-05}
{"train_loss": 0.3658689856529236, "global_step": 25781, "epoch": 289, "lr": 9.78670918156838e-05}
{"train_loss": 0.5218400359153748, "global_step": 25782, "epoch": 289, "lr": 9.786692429305449e-05}
{"train_loss": 0.5378871560096741, "global_step": 25783, "epoch": 289, "lr": 9.786675676399007e-05}
{"train_loss": 0.421323299407959, "global_step": 25784, "epoch": 289, "lr": 9.786658922849052e-05}
{"train_loss": 0.28578856587409973, "global_step": 25785, "epoch": 289, "lr": 9.786642168655591e-05}
{"train_loss": 0.35055214166641235, "global_step": 25786, "epoch": 289, "lr": 9.786625413818622e-05}
{"train_loss": 0.4548265337944031, "global_step": 25787, "epoch": 289, "lr": 9.786608658338152e-05}
{"train_loss": 0.49153760075569153, "global_step": 25788, "epoch": 289, "lr": 9.786591902214179e-05}
{"train_loss": 0.4491642713546753, "global_step": 25789, "epoch": 289, "lr": 9.786575145446708e-05}
{"train_loss": 0.405108243227005, "global_step": 25790, "epoch": 289, "lr": 9.78655838803574e-05}
{"train_loss": 0.38790908455848694, "global_step": 25791, "epoch": 289, "lr": 9.786541629981279e-05}
{"train_loss": 0.4632105529308319, "global_step": 25792, "epoch": 289, "lr": 9.786524871283325e-05}
{"train_loss": 0.43925920128822327, "global_step": 25793, "epoch": 289, "lr": 9.786508111941878e-05}
{"train_loss": 0.41391631960868835, "global_step": 25794, "epoch": 289, "lr": 9.786491351956948e-05}
{"train_loss": 0.41153252124786377, "global_step": 25795, "epoch": 289, "lr": 9.786474591328529e-05}
{"train_loss": 0.3842371702194214, "global_step": 25796, "epoch": 289, "lr": 9.78645783005663e-05}
{"train_loss": 0.4298691153526306, "global_step": 25797, "epoch": 289, "lr": 9.786441068141247e-05}
{"train_loss": 0.5070369839668274, "global_step": 25798, "epoch": 289, "lr": 9.786424305582386e-05}
{"train_loss": 0.3552969992160797, "global_step": 25799, "epoch": 289, "lr": 9.786407542380052e-05}
{"train_loss": 0.3597840666770935, "global_step": 25800, "epoch": 289, "lr": 9.78639077853424e-05}
{"train_loss": 0.3755064904689789, "global_step": 25801, "epoch": 289, "lr": 9.786374014044956e-05}
{"train_loss": 0.4075106382369995, "global_step": 25802, "epoch": 289, "lr": 9.786357248912206e-05}
{"train_loss": 0.4235890209674835, "global_step": 25803, "epoch": 289, "lr": 9.786340483135985e-05}
{"train_loss": 0.423762708902359, "global_step": 25804, "epoch": 289, "lr": 9.7863237167163e-05}
{"train_loss": 0.3695676326751709, "global_step": 25805, "epoch": 289, "lr": 9.786306949653152e-05}
{"train_loss": 0.32371169328689575, "global_step": 25806, "epoch": 289, "lr": 9.786290181946545e-05}
{"train_loss": 0.5628860592842102, "global_step": 25807, "epoch": 289, "lr": 9.786273413596477e-05}
{"train_loss": 0.44787269830703735, "global_step": 25808, "epoch": 289, "lr": 9.786256644602955e-05}
{"train_loss": 0.4176070224703028, "global_step": 25809, "epoch": 289, "lr": 9.786239874965978e-05, "val_loss": 2.0207695960998535}
{"train_loss": 0.3236033022403717, "global_step": 25810, "epoch": 290, "lr": 9.78622310468555e-05}
{"train_loss": 0.533920407295227, "global_step": 25811, "epoch": 290, "lr": 9.786206333761672e-05}
{"train_loss": 0.3739442527294159, "global_step": 25812, "epoch": 290, "lr": 9.786189562194346e-05}
{"train_loss": 0.4161878526210785, "global_step": 25813, "epoch": 290, "lr": 9.786172789983576e-05}
{"train_loss": 0.2915981113910675, "global_step": 25814, "epoch": 290, "lr": 9.786156017129365e-05}
{"train_loss": 0.347829669713974, "global_step": 25815, "epoch": 290, "lr": 9.786139243631713e-05}
{"train_loss": 0.4658551514148712, "global_step": 25816, "epoch": 290, "lr": 9.786122469490621e-05}
{"train_loss": 0.4180164039134979, "global_step": 25817, "epoch": 290, "lr": 9.786105694706096e-05}
{"train_loss": 0.3330046534538269, "global_step": 25818, "epoch": 290, "lr": 9.786088919278135e-05}
{"train_loss": 0.34318479895591736, "global_step": 25819, "epoch": 290, "lr": 9.786072143206743e-05}
{"train_loss": 0.33131811022758484, "global_step": 25820, "epoch": 290, "lr": 9.786055366491922e-05}
{"train_loss": 0.4320604205131531, "global_step": 25821, "epoch": 290, "lr": 9.786038589133676e-05}
{"train_loss": 0.4452328085899353, "global_step": 25822, "epoch": 290, "lr": 9.786021811132005e-05}
{"train_loss": 0.35248464345932007, "global_step": 25823, "epoch": 290, "lr": 9.786005032486908e-05}
{"train_loss": 0.3935090899467468, "global_step": 25824, "epoch": 290, "lr": 9.785988253198395e-05}
{"train_loss": 0.3718687891960144, "global_step": 25825, "epoch": 290, "lr": 9.785971473266463e-05}
{"train_loss": 0.2727401852607727, "global_step": 25826, "epoch": 290, "lr": 9.785954692691116e-05}
{"train_loss": 0.40442609786987305, "global_step": 25827, "epoch": 290, "lr": 9.785937911472355e-05}
{"train_loss": 0.45152997970581055, "global_step": 25828, "epoch": 290, "lr": 9.785921129610183e-05}
{"train_loss": 0.45099595189094543, "global_step": 25829, "epoch": 290, "lr": 9.785904347104602e-05}
{"train_loss": 0.4380075931549072, "global_step": 25830, "epoch": 290, "lr": 9.785887563955615e-05}
{"train_loss": 0.4250042736530304, "global_step": 25831, "epoch": 290, "lr": 9.785870780163224e-05}
{"train_loss": 0.3170650899410248, "global_step": 25832, "epoch": 290, "lr": 9.785853995727431e-05}
{"train_loss": 0.4422348141670227, "global_step": 25833, "epoch": 290, "lr": 9.785837210648237e-05}
{"train_loss": 0.49989181756973267, "global_step": 25834, "epoch": 290, "lr": 9.785820424925648e-05}
{"train_loss": 0.43308183550834656, "global_step": 25835, "epoch": 290, "lr": 9.785803638559661e-05}
{"train_loss": 0.4342346787452698, "global_step": 25836, "epoch": 290, "lr": 9.785786851550282e-05}
{"train_loss": 0.663908064365387, "global_step": 25837, "epoch": 290, "lr": 9.785770063897512e-05}
{"train_loss": 0.49272751808166504, "global_step": 25838, "epoch": 290, "lr": 9.785753275601355e-05}
{"train_loss": 0.4922715425491333, "global_step": 25839, "epoch": 290, "lr": 9.785736486661812e-05}
{"train_loss": 0.4035555124282837, "global_step": 25840, "epoch": 290, "lr": 9.785719697078884e-05}
{"train_loss": 0.54497891664505, "global_step": 25841, "epoch": 290, "lr": 9.785702906852574e-05}
{"train_loss": 0.38322198390960693, "global_step": 25842, "epoch": 290, "lr": 9.785686115982884e-05}
{"train_loss": 0.5097622871398926, "global_step": 25843, "epoch": 290, "lr": 9.785669324469817e-05}
{"train_loss": 0.487391859292984, "global_step": 25844, "epoch": 290, "lr": 9.785652532313377e-05}
{"train_loss": 0.5699384212493896, "global_step": 25845, "epoch": 290, "lr": 9.785635739513563e-05}
{"train_loss": 0.37351518869400024, "global_step": 25846, "epoch": 290, "lr": 9.785618946070378e-05}
{"train_loss": 0.4056534767150879, "global_step": 25847, "epoch": 290, "lr": 9.785602151983826e-05}
{"train_loss": 0.41144219040870667, "global_step": 25848, "epoch": 290, "lr": 9.785585357253908e-05}
{"train_loss": 0.43047645688056946, "global_step": 25849, "epoch": 290, "lr": 9.785568561880625e-05}
{"train_loss": 0.5882703065872192, "global_step": 25850, "epoch": 290, "lr": 9.785551765863983e-05}
{"train_loss": 0.3569321930408478, "global_step": 25851, "epoch": 290, "lr": 9.78553496920398e-05}
{"train_loss": 0.6268858313560486, "global_step": 25852, "epoch": 290, "lr": 9.785518171900621e-05}
{"train_loss": 0.39732491970062256, "global_step": 25853, "epoch": 290, "lr": 9.785501373953907e-05}
{"train_loss": 0.35978758335113525, "global_step": 25854, "epoch": 290, "lr": 9.78548457536384e-05}
{"train_loss": 0.3912414312362671, "global_step": 25855, "epoch": 290, "lr": 9.785467776130423e-05}
{"train_loss": 0.5134209990501404, "global_step": 25856, "epoch": 290, "lr": 9.78545097625366e-05}
{"train_loss": 0.33846116065979004, "global_step": 25857, "epoch": 290, "lr": 9.78543417573355e-05}
{"train_loss": 0.4777953326702118, "global_step": 25858, "epoch": 290, "lr": 9.785417374570099e-05}
{"train_loss": 0.5239615440368652, "global_step": 25859, "epoch": 290, "lr": 9.785400572763304e-05}
{"train_loss": 0.41883665323257446, "global_step": 25860, "epoch": 290, "lr": 9.78538377031317e-05}
{"train_loss": 0.5335416197776794, "global_step": 25861, "epoch": 290, "lr": 9.785366967219701e-05}
{"train_loss": 0.3726222515106201, "global_step": 25862, "epoch": 290, "lr": 9.785350163482899e-05}
{"train_loss": 0.4121244251728058, "global_step": 25863, "epoch": 290, "lr": 9.785333359102762e-05}
{"train_loss": 0.32947495579719543, "global_step": 25864, "epoch": 290, "lr": 9.785316554079297e-05}
{"train_loss": 0.3882621228694916, "global_step": 25865, "epoch": 290, "lr": 9.785299748412504e-05}
{"train_loss": 0.4899141490459442, "global_step": 25866, "epoch": 290, "lr": 9.785282942102386e-05}
{"train_loss": 0.47768187522888184, "global_step": 25867, "epoch": 290, "lr": 9.785266135148945e-05}
{"train_loss": 0.3130687475204468, "global_step": 25868, "epoch": 290, "lr": 9.785249327552184e-05}
{"train_loss": 0.49726396799087524, "global_step": 25869, "epoch": 290, "lr": 9.785232519312105e-05}
{"train_loss": 0.36318233609199524, "global_step": 25870, "epoch": 290, "lr": 9.785215710428708e-05}
{"train_loss": 0.5453659296035767, "global_step": 25871, "epoch": 290, "lr": 9.785198900901997e-05}
{"train_loss": 0.27954402565956116, "global_step": 25872, "epoch": 290, "lr": 9.785182090731975e-05}
{"train_loss": 0.34629493951797485, "global_step": 25873, "epoch": 290, "lr": 9.785165279918645e-05}
{"train_loss": 0.37333348393440247, "global_step": 25874, "epoch": 290, "lr": 9.785148468462006e-05}
{"train_loss": 0.4798842668533325, "global_step": 25875, "epoch": 290, "lr": 9.785131656362061e-05}
{"train_loss": 0.29239922761917114, "global_step": 25876, "epoch": 290, "lr": 9.785114843618815e-05}
{"train_loss": 0.3353954553604126, "global_step": 25877, "epoch": 290, "lr": 9.78509803023227e-05}
{"train_loss": 0.4915849268436432, "global_step": 25878, "epoch": 290, "lr": 9.785081216202426e-05}
{"train_loss": 0.47897806763648987, "global_step": 25879, "epoch": 290, "lr": 9.785064401529284e-05}
{"train_loss": 0.48345810174942017, "global_step": 25880, "epoch": 290, "lr": 9.78504758621285e-05}
{"train_loss": 0.4980294406414032, "global_step": 25881, "epoch": 290, "lr": 9.785030770253125e-05}
{"train_loss": 0.32976093888282776, "global_step": 25882, "epoch": 290, "lr": 9.785013953650111e-05}
{"train_loss": 0.3954590857028961, "global_step": 25883, "epoch": 290, "lr": 9.78499713640381e-05}
{"train_loss": 0.4308144450187683, "global_step": 25884, "epoch": 290, "lr": 9.784980318514225e-05}
{"train_loss": 0.48002851009368896, "global_step": 25885, "epoch": 290, "lr": 9.784963499981357e-05}
{"train_loss": 0.4882397949695587, "global_step": 25886, "epoch": 290, "lr": 9.78494668080521e-05}
{"train_loss": 0.41860151290893555, "global_step": 25887, "epoch": 290, "lr": 9.784929860985783e-05}
{"train_loss": 0.536641001701355, "global_step": 25888, "epoch": 290, "lr": 9.784913040523082e-05}
{"train_loss": 0.46761372685432434, "global_step": 25889, "epoch": 290, "lr": 9.784896219417108e-05}
{"train_loss": 0.3770972490310669, "global_step": 25890, "epoch": 290, "lr": 9.784879397667861e-05}
{"train_loss": 0.37802690267562866, "global_step": 25891, "epoch": 290, "lr": 9.784862575275348e-05}
{"train_loss": 0.40083807706832886, "global_step": 25892, "epoch": 290, "lr": 9.784845752239566e-05}
{"train_loss": 0.503381609916687, "global_step": 25893, "epoch": 290, "lr": 9.784828928560522e-05}
{"train_loss": 0.501132071018219, "global_step": 25894, "epoch": 290, "lr": 9.784812104238216e-05}
{"train_loss": 0.38011088967323303, "global_step": 25895, "epoch": 290, "lr": 9.78479527927265e-05}
{"train_loss": 0.5331194996833801, "global_step": 25896, "epoch": 290, "lr": 9.784778453663825e-05}
{"train_loss": 0.4058474898338318, "global_step": 25897, "epoch": 290, "lr": 9.784761627411747e-05}
{"train_loss": 0.4290600268358595, "global_step": 25898, "epoch": 290, "lr": 9.784744800516415e-05, "val_loss": 2.0847671031951904, "train_action_mse_error": 28.762065887451172}
{"train_loss": 0.4405645728111267, "global_step": 25899, "epoch": 291, "lr": 9.784727972977832e-05}
{"train_loss": 0.31565654277801514, "global_step": 25900, "epoch": 291, "lr": 9.784711144796002e-05}
{"train_loss": 0.31646618247032166, "global_step": 25901, "epoch": 291, "lr": 9.784694315970924e-05}
{"train_loss": 0.5102719068527222, "global_step": 25902, "epoch": 291, "lr": 9.784677486502604e-05}
{"train_loss": 0.2760123908519745, "global_step": 25903, "epoch": 291, "lr": 9.784660656391041e-05}
{"train_loss": 0.37303295731544495, "global_step": 25904, "epoch": 291, "lr": 9.784643825636239e-05}
{"train_loss": 0.37046706676483154, "global_step": 25905, "epoch": 291, "lr": 9.784626994238202e-05}
{"train_loss": 0.3700668513774872, "global_step": 25906, "epoch": 291, "lr": 9.784610162196927e-05}
{"train_loss": 0.38275739550590515, "global_step": 25907, "epoch": 291, "lr": 9.784593329512422e-05}
{"train_loss": 0.5123909115791321, "global_step": 25908, "epoch": 291, "lr": 9.784576496184686e-05}
{"train_loss": 0.3617134690284729, "global_step": 25909, "epoch": 291, "lr": 9.784559662213722e-05}
{"train_loss": 0.40199509263038635, "global_step": 25910, "epoch": 291, "lr": 9.784542827599532e-05}
{"train_loss": 0.4174691140651703, "global_step": 25911, "epoch": 291, "lr": 9.784525992342117e-05}
{"train_loss": 0.402614951133728, "global_step": 25912, "epoch": 291, "lr": 9.784509156441483e-05}
{"train_loss": 0.3204137980937958, "global_step": 25913, "epoch": 291, "lr": 9.784492319897628e-05}
{"train_loss": 0.4763370752334595, "global_step": 25914, "epoch": 291, "lr": 9.78447548271056e-05}
{"train_loss": 0.3862965404987335, "global_step": 25915, "epoch": 291, "lr": 9.784458644880275e-05}
{"train_loss": 0.3734152615070343, "global_step": 25916, "epoch": 291, "lr": 9.784441806406779e-05}
{"train_loss": 0.30603528022766113, "global_step": 25917, "epoch": 291, "lr": 9.784424967290072e-05}
{"train_loss": 0.41974419355392456, "global_step": 25918, "epoch": 291, "lr": 9.784408127530157e-05}
{"train_loss": 0.39656174182891846, "global_step": 25919, "epoch": 291, "lr": 9.784391287127037e-05}
{"train_loss": 0.5119815468788147, "global_step": 25920, "epoch": 291, "lr": 9.784374446080715e-05}
{"train_loss": 0.4541777968406677, "global_step": 25921, "epoch": 291, "lr": 9.784357604391192e-05}
{"train_loss": 0.41461050510406494, "global_step": 25922, "epoch": 291, "lr": 9.784340762058468e-05}
{"train_loss": 0.3301565647125244, "global_step": 25923, "epoch": 291, "lr": 9.78432391908255e-05}
{"train_loss": 0.4037703573703766, "global_step": 25924, "epoch": 291, "lr": 9.784307075463437e-05}
{"train_loss": 0.35293349623680115, "global_step": 25925, "epoch": 291, "lr": 9.784290231201134e-05}
{"train_loss": 0.4279380738735199, "global_step": 25926, "epoch": 291, "lr": 9.78427338629564e-05}
{"train_loss": 0.4421372413635254, "global_step": 25927, "epoch": 291, "lr": 9.784256540746959e-05}
{"train_loss": 0.3480352461338043, "global_step": 25928, "epoch": 291, "lr": 9.784239694555094e-05}
{"train_loss": 0.3611172139644623, "global_step": 25929, "epoch": 291, "lr": 9.784222847720043e-05}
{"train_loss": 0.42214539647102356, "global_step": 25930, "epoch": 291, "lr": 9.784206000241816e-05}
{"train_loss": 0.423997163772583, "global_step": 25931, "epoch": 291, "lr": 9.784189152120407e-05}
{"train_loss": 0.38641878962516785, "global_step": 25932, "epoch": 291, "lr": 9.784172303355824e-05}
{"train_loss": 0.38393673300743103, "global_step": 25933, "epoch": 291, "lr": 9.784155453948067e-05}
{"train_loss": 0.3788085877895355, "global_step": 25934, "epoch": 291, "lr": 9.784138603897139e-05}
{"train_loss": 0.3309195339679718, "global_step": 25935, "epoch": 291, "lr": 9.784121753203042e-05}
{"train_loss": 0.41169071197509766, "global_step": 25936, "epoch": 291, "lr": 9.784104901865777e-05}
{"train_loss": 0.4175255000591278, "global_step": 25937, "epoch": 291, "lr": 9.784088049885349e-05}
{"train_loss": 0.46487823128700256, "global_step": 25938, "epoch": 291, "lr": 9.784071197261757e-05}
{"train_loss": 0.3607824742794037, "global_step": 25939, "epoch": 291, "lr": 9.784054343995006e-05}
{"train_loss": 0.3413825035095215, "global_step": 25940, "epoch": 291, "lr": 9.784037490085096e-05}
{"train_loss": 0.46123334765434265, "global_step": 25941, "epoch": 291, "lr": 9.784020635532032e-05}
{"train_loss": 0.3909131586551666, "global_step": 25942, "epoch": 291, "lr": 9.784003780335814e-05}
{"train_loss": 0.28652381896972656, "global_step": 25943, "epoch": 291, "lr": 9.783986924496446e-05}
{"train_loss": 0.4743117392063141, "global_step": 25944, "epoch": 291, "lr": 9.783970068013929e-05}
{"train_loss": 0.3510248363018036, "global_step": 25945, "epoch": 291, "lr": 9.783953210888264e-05}
{"train_loss": 0.3307391107082367, "global_step": 25946, "epoch": 291, "lr": 9.783936353119456e-05}
{"train_loss": 0.3659874200820923, "global_step": 25947, "epoch": 291, "lr": 9.783919494707505e-05}
{"train_loss": 0.3664446175098419, "global_step": 25948, "epoch": 291, "lr": 9.783902635652415e-05}
{"train_loss": 0.33399084210395813, "global_step": 25949, "epoch": 291, "lr": 9.783885775954188e-05}
{"train_loss": 0.4181792140007019, "global_step": 25950, "epoch": 291, "lr": 9.783868915612826e-05}
{"train_loss": 0.31791043281555176, "global_step": 25951, "epoch": 291, "lr": 9.78385205462833e-05}
{"train_loss": 0.4028720557689667, "global_step": 25952, "epoch": 291, "lr": 9.783835193000705e-05}
{"train_loss": 0.3580907881259918, "global_step": 25953, "epoch": 291, "lr": 9.783818330729951e-05}
{"train_loss": 0.47324496507644653, "global_step": 25954, "epoch": 291, "lr": 9.78380146781607e-05}
{"train_loss": 0.34802526235580444, "global_step": 25955, "epoch": 291, "lr": 9.783784604259066e-05}
{"train_loss": 0.37647882103919983, "global_step": 25956, "epoch": 291, "lr": 9.78376774005894e-05}
{"train_loss": 0.4566691517829895, "global_step": 25957, "epoch": 291, "lr": 9.783750875215696e-05}
{"train_loss": 0.5403605699539185, "global_step": 25958, "epoch": 291, "lr": 9.783734009729332e-05}
{"train_loss": 0.4247789978981018, "global_step": 25959, "epoch": 291, "lr": 9.783717143599857e-05}
{"train_loss": 0.3927389681339264, "global_step": 25960, "epoch": 291, "lr": 9.783700276827268e-05}
{"train_loss": 0.492815226316452, "global_step": 25961, "epoch": 291, "lr": 9.783683409411568e-05}
{"train_loss": 0.27594512701034546, "global_step": 25962, "epoch": 291, "lr": 9.783666541352761e-05}
{"train_loss": 0.4063626229763031, "global_step": 25963, "epoch": 291, "lr": 9.783649672650848e-05}
{"train_loss": 0.4074251651763916, "global_step": 25964, "epoch": 291, "lr": 9.783632803305831e-05}
{"train_loss": 0.3867305815219879, "global_step": 25965, "epoch": 291, "lr": 9.783615933317715e-05}
{"train_loss": 0.32457607984542847, "global_step": 25966, "epoch": 291, "lr": 9.783599062686498e-05}
{"train_loss": 0.3894454538822174, "global_step": 25967, "epoch": 291, "lr": 9.783582191412185e-05}
{"train_loss": 0.41849443316459656, "global_step": 25968, "epoch": 291, "lr": 9.783565319494778e-05}
{"train_loss": 0.32136958837509155, "global_step": 25969, "epoch": 291, "lr": 9.783548446934279e-05}
{"train_loss": 0.3659389019012451, "global_step": 25970, "epoch": 291, "lr": 9.783531573730689e-05}
{"train_loss": 0.3265296220779419, "global_step": 25971, "epoch": 291, "lr": 9.783514699884013e-05}
{"train_loss": 0.32747483253479004, "global_step": 25972, "epoch": 291, "lr": 9.783497825394252e-05}
{"train_loss": 0.3583398759365082, "global_step": 25973, "epoch": 291, "lr": 9.783480950261407e-05}
{"train_loss": 0.45986899733543396, "global_step": 25974, "epoch": 291, "lr": 9.783464074485482e-05}
{"train_loss": 0.5130491852760315, "global_step": 25975, "epoch": 291, "lr": 9.783447198066478e-05}
{"train_loss": 0.34843748807907104, "global_step": 25976, "epoch": 291, "lr": 9.783430321004396e-05}
{"train_loss": 0.43490374088287354, "global_step": 25977, "epoch": 291, "lr": 9.783413443299243e-05}
{"train_loss": 0.43005749583244324, "global_step": 25978, "epoch": 291, "lr": 9.783396564951018e-05}
{"train_loss": 0.4055558741092682, "global_step": 25979, "epoch": 291, "lr": 9.783379685959722e-05}
{"train_loss": 0.5193752646446228, "global_step": 25980, "epoch": 291, "lr": 9.78336280632536e-05}
{"train_loss": 0.3405960500240326, "global_step": 25981, "epoch": 291, "lr": 9.783345926047933e-05}
{"train_loss": 0.5644686222076416, "global_step": 25982, "epoch": 291, "lr": 9.783329045127444e-05}
{"train_loss": 0.47816458344459534, "global_step": 25983, "epoch": 291, "lr": 9.783312163563895e-05}
{"train_loss": 0.42642441391944885, "global_step": 25984, "epoch": 291, "lr": 9.783295281357287e-05}
{"train_loss": 0.5127047300338745, "global_step": 25985, "epoch": 291, "lr": 9.783278398507622e-05}
{"train_loss": 0.39695000648498535, "global_step": 25986, "epoch": 291, "lr": 9.783261515014905e-05}
{"train_loss": 0.3981199324800727, "global_step": 25987, "epoch": 291, "lr": 9.783244630879138e-05, "val_loss": 2.051142454147339}
{"train_loss": 0.3483041524887085, "global_step": 25988, "epoch": 292, "lr": 9.783227746100321e-05}
{"train_loss": 0.4614957571029663, "global_step": 25989, "epoch": 292, "lr": 9.783210860678457e-05}
{"train_loss": 0.45960700511932373, "global_step": 25990, "epoch": 292, "lr": 9.783193974613549e-05}
{"train_loss": 0.3831969201564789, "global_step": 25991, "epoch": 292, "lr": 9.783177087905599e-05}
{"train_loss": 0.3402642011642456, "global_step": 25992, "epoch": 292, "lr": 9.783160200554609e-05}
{"train_loss": 0.5119410157203674, "global_step": 25993, "epoch": 292, "lr": 9.783143312560581e-05}
{"train_loss": 0.30186235904693604, "global_step": 25994, "epoch": 292, "lr": 9.783126423923517e-05}
{"train_loss": 0.44792917370796204, "global_step": 25995, "epoch": 292, "lr": 9.78310953464342e-05}
{"train_loss": 0.3789154291152954, "global_step": 25996, "epoch": 292, "lr": 9.783092644720293e-05}
{"train_loss": 0.41074123978614807, "global_step": 25997, "epoch": 292, "lr": 9.783075754154138e-05}
{"train_loss": 0.2934926748275757, "global_step": 25998, "epoch": 292, "lr": 9.783058862944956e-05}
{"train_loss": 0.29671013355255127, "global_step": 25999, "epoch": 292, "lr": 9.78304197109275e-05}
{"train_loss": 0.3140389025211334, "global_step": 26000, "epoch": 292, "lr": 9.783025078597523e-05}
{"train_loss": 0.41269251704216003, "global_step": 26001, "epoch": 292, "lr": 9.783008185459277e-05}
{"train_loss": 0.3611046075820923, "global_step": 26002, "epoch": 292, "lr": 9.782991291678012e-05}
{"train_loss": 0.3343384265899658, "global_step": 26003, "epoch": 292, "lr": 9.782974397253735e-05}
{"train_loss": 0.4727013409137726, "global_step": 26004, "epoch": 292, "lr": 9.782957502186442e-05}
{"train_loss": 0.4220148026943207, "global_step": 26005, "epoch": 292, "lr": 9.782940606476141e-05}
{"train_loss": 0.49161040782928467, "global_step": 26006, "epoch": 292, "lr": 9.782923710122831e-05}
{"train_loss": 0.5129244923591614, "global_step": 26007, "epoch": 292, "lr": 9.782906813126516e-05}
{"train_loss": 0.4348788857460022, "global_step": 26008, "epoch": 292, "lr": 9.782889915487196e-05}
{"train_loss": 0.4177224934101105, "global_step": 26009, "epoch": 292, "lr": 9.782873017204876e-05}
{"train_loss": 0.33369216322898865, "global_step": 26010, "epoch": 292, "lr": 9.782856118279556e-05}
{"train_loss": 0.4303973615169525, "global_step": 26011, "epoch": 292, "lr": 9.782839218711239e-05}
{"train_loss": 0.3499717116355896, "global_step": 26012, "epoch": 292, "lr": 9.782822318499929e-05}
{"train_loss": 0.3407323956489563, "global_step": 26013, "epoch": 292, "lr": 9.782805417645626e-05}
{"train_loss": 0.43565598130226135, "global_step": 26014, "epoch": 292, "lr": 9.782788516148333e-05}
{"train_loss": 0.4936297833919525, "global_step": 26015, "epoch": 292, "lr": 9.782771614008054e-05}
{"train_loss": 0.3345867097377777, "global_step": 26016, "epoch": 292, "lr": 9.782754711224789e-05}
{"train_loss": 0.348379909992218, "global_step": 26017, "epoch": 292, "lr": 9.782737807798539e-05}
{"train_loss": 0.46920329332351685, "global_step": 26018, "epoch": 292, "lr": 9.78272090372931e-05}
{"train_loss": 0.42214304208755493, "global_step": 26019, "epoch": 292, "lr": 9.782703999017102e-05}
{"train_loss": 0.48568254709243774, "global_step": 26020, "epoch": 292, "lr": 9.782687093661918e-05}
{"train_loss": 0.4578864276409149, "global_step": 26021, "epoch": 292, "lr": 9.782670187663759e-05}
{"train_loss": 0.3275831341743469, "global_step": 26022, "epoch": 292, "lr": 9.782653281022629e-05}
{"train_loss": 0.3903539478778839, "global_step": 26023, "epoch": 292, "lr": 9.78263637373853e-05}
{"train_loss": 0.45087093114852905, "global_step": 26024, "epoch": 292, "lr": 9.782619465811463e-05}
{"train_loss": 0.41854026913642883, "global_step": 26025, "epoch": 292, "lr": 9.782602557241431e-05}
{"train_loss": 0.4080066382884979, "global_step": 26026, "epoch": 292, "lr": 9.782585648028438e-05}
{"train_loss": 0.35467541217803955, "global_step": 26027, "epoch": 292, "lr": 9.782568738172484e-05}
{"train_loss": 0.4975143373012543, "global_step": 26028, "epoch": 292, "lr": 9.78255182767357e-05}
{"train_loss": 0.3139534592628479, "global_step": 26029, "epoch": 292, "lr": 9.782534916531703e-05}
{"train_loss": 0.43517404794692993, "global_step": 26030, "epoch": 292, "lr": 9.782518004746882e-05}
{"train_loss": 0.3695359528064728, "global_step": 26031, "epoch": 292, "lr": 9.782501092319107e-05}
{"train_loss": 0.35581764578819275, "global_step": 26032, "epoch": 292, "lr": 9.782484179248385e-05}
{"train_loss": 0.3630896508693695, "global_step": 26033, "epoch": 292, "lr": 9.782467265534716e-05}
{"train_loss": 0.4143156409263611, "global_step": 26034, "epoch": 292, "lr": 9.782450351178103e-05}
{"train_loss": 0.4786742329597473, "global_step": 26035, "epoch": 292, "lr": 9.782433436178548e-05}
{"train_loss": 0.3463081121444702, "global_step": 26036, "epoch": 292, "lr": 9.782416520536053e-05}
{"train_loss": 0.32325243949890137, "global_step": 26037, "epoch": 292, "lr": 9.78239960425062e-05}
{"train_loss": 0.4716424345970154, "global_step": 26038, "epoch": 292, "lr": 9.782382687322251e-05}
{"train_loss": 0.36508479714393616, "global_step": 26039, "epoch": 292, "lr": 9.782365769750949e-05}
{"train_loss": 0.4738829731941223, "global_step": 26040, "epoch": 292, "lr": 9.782348851536718e-05}
{"train_loss": 0.4879927933216095, "global_step": 26041, "epoch": 292, "lr": 9.782331932679558e-05}
{"train_loss": 0.4530200660228729, "global_step": 26042, "epoch": 292, "lr": 9.78231501317947e-05}
{"train_loss": 0.41533249616622925, "global_step": 26043, "epoch": 292, "lr": 9.782298093036459e-05}
{"train_loss": 0.5138441920280457, "global_step": 26044, "epoch": 292, "lr": 9.782281172250526e-05}
{"train_loss": 0.41103509068489075, "global_step": 26045, "epoch": 292, "lr": 9.782264250821676e-05}
{"train_loss": 0.48825782537460327, "global_step": 26046, "epoch": 292, "lr": 9.782247328749906e-05}
{"train_loss": 0.4140852093696594, "global_step": 26047, "epoch": 292, "lr": 9.782230406035223e-05}
{"train_loss": 0.40174180269241333, "global_step": 26048, "epoch": 292, "lr": 9.782213482677626e-05}
{"train_loss": 0.3605284094810486, "global_step": 26049, "epoch": 292, "lr": 9.78219655867712e-05}
{"train_loss": 0.45137524604797363, "global_step": 26050, "epoch": 292, "lr": 9.782179634033704e-05}
{"train_loss": 0.3417251408100128, "global_step": 26051, "epoch": 292, "lr": 9.782162708747385e-05}
{"train_loss": 0.4124227464199066, "global_step": 26052, "epoch": 292, "lr": 9.78214578281816e-05}
{"train_loss": 0.385842502117157, "global_step": 26053, "epoch": 292, "lr": 9.782128856246033e-05}
{"train_loss": 0.2997691333293915, "global_step": 26054, "epoch": 292, "lr": 9.78211192903101e-05}
{"train_loss": 0.46240749955177307, "global_step": 26055, "epoch": 292, "lr": 9.782095001173089e-05}
{"train_loss": 0.40339240431785583, "global_step": 26056, "epoch": 292, "lr": 9.782078072672274e-05}
{"train_loss": 0.46560969948768616, "global_step": 26057, "epoch": 292, "lr": 9.782061143528566e-05}
{"train_loss": 0.5238993167877197, "global_step": 26058, "epoch": 292, "lr": 9.782044213741969e-05}
{"train_loss": 0.37876543402671814, "global_step": 26059, "epoch": 292, "lr": 9.782027283312485e-05}
{"train_loss": 0.33424127101898193, "global_step": 26060, "epoch": 292, "lr": 9.782010352240114e-05}
{"train_loss": 0.5074463486671448, "global_step": 26061, "epoch": 292, "lr": 9.781993420524861e-05}
{"train_loss": 0.3775056302547455, "global_step": 26062, "epoch": 292, "lr": 9.781976488166726e-05}
{"train_loss": 0.36174237728118896, "global_step": 26063, "epoch": 292, "lr": 9.781959555165714e-05}
{"train_loss": 0.2656005024909973, "global_step": 26064, "epoch": 292, "lr": 9.781942621521826e-05}
{"train_loss": 0.40708935260772705, "global_step": 26065, "epoch": 292, "lr": 9.781925687235064e-05}
{"train_loss": 0.4485008120536804, "global_step": 26066, "epoch": 292, "lr": 9.78190875230543e-05}
{"train_loss": 0.40322181582450867, "global_step": 26067, "epoch": 292, "lr": 9.781891816732927e-05}
{"train_loss": 0.40946075320243835, "global_step": 26068, "epoch": 292, "lr": 9.781874880517556e-05}
{"train_loss": 0.36465519666671753, "global_step": 26069, "epoch": 292, "lr": 9.781857943659321e-05}
{"train_loss": 0.47294801473617554, "global_step": 26070, "epoch": 292, "lr": 9.781841006158223e-05}
{"train_loss": 0.38270047307014465, "global_step": 26071, "epoch": 292, "lr": 9.781824068014265e-05}
{"train_loss": 0.4777814745903015, "global_step": 26072, "epoch": 292, "lr": 9.78180712922745e-05}
{"train_loss": 0.4574713706970215, "global_step": 26073, "epoch": 292, "lr": 9.781790189797777e-05}
{"train_loss": 0.3823709785938263, "global_step": 26074, "epoch": 292, "lr": 9.781773249725253e-05}
{"train_loss": 0.3514467775821686, "global_step": 26075, "epoch": 292, "lr": 9.781756309009876e-05}
{"train_loss": 0.40675046604670834, "global_step": 26076, "epoch": 292, "lr": 9.78173936765165e-05, "val_loss": 2.1292855739593506}
{"train_loss": 0.43685728311538696, "global_step": 26077, "epoch": 293, "lr": 9.78172242565058e-05}
{"train_loss": 0.3870909810066223, "global_step": 26078, "epoch": 293, "lr": 9.781705483006664e-05}
{"train_loss": 0.41574516892433167, "global_step": 26079, "epoch": 293, "lr": 9.781688539719905e-05}
{"train_loss": 0.39195966720581055, "global_step": 26080, "epoch": 293, "lr": 9.781671595790308e-05}
{"train_loss": 0.4314712584018707, "global_step": 26081, "epoch": 293, "lr": 9.781654651217873e-05}
{"train_loss": 0.33588477969169617, "global_step": 26082, "epoch": 293, "lr": 9.781637706002602e-05}
{"train_loss": 0.519058346748352, "global_step": 26083, "epoch": 293, "lr": 9.781620760144499e-05}
{"train_loss": 0.3026953339576721, "global_step": 26084, "epoch": 293, "lr": 9.781603813643565e-05}
{"train_loss": 0.3273316025733948, "global_step": 26085, "epoch": 293, "lr": 9.781586866499803e-05}
{"train_loss": 0.40859994292259216, "global_step": 26086, "epoch": 293, "lr": 9.781569918713215e-05}
{"train_loss": 0.4443984925746918, "global_step": 26087, "epoch": 293, "lr": 9.781552970283803e-05}
{"train_loss": 0.43925976753234863, "global_step": 26088, "epoch": 293, "lr": 9.78153602121157e-05}
{"train_loss": 0.31186577677726746, "global_step": 26089, "epoch": 293, "lr": 9.781519071496517e-05}
{"train_loss": 0.47188007831573486, "global_step": 26090, "epoch": 293, "lr": 9.781502121138646e-05}
{"train_loss": 0.38854357600212097, "global_step": 26091, "epoch": 293, "lr": 9.781485170137963e-05}
{"train_loss": 0.3870106339454651, "global_step": 26092, "epoch": 293, "lr": 9.781468218494465e-05}
{"train_loss": 0.41354501247406006, "global_step": 26093, "epoch": 293, "lr": 9.781451266208158e-05}
{"train_loss": 0.3376396596431732, "global_step": 26094, "epoch": 293, "lr": 9.781434313279043e-05}
{"train_loss": 0.39764270186424255, "global_step": 26095, "epoch": 293, "lr": 9.781417359707123e-05}
{"train_loss": 0.4020477533340454, "global_step": 26096, "epoch": 293, "lr": 9.7814004054924e-05}
{"train_loss": 0.3961699903011322, "global_step": 26097, "epoch": 293, "lr": 9.781383450634874e-05}
{"train_loss": 0.35381650924682617, "global_step": 26098, "epoch": 293, "lr": 9.781366495134551e-05}
{"train_loss": 0.35262972116470337, "global_step": 26099, "epoch": 293, "lr": 9.78134953899143e-05}
{"train_loss": 0.47275587916374207, "global_step": 26100, "epoch": 293, "lr": 9.781332582205517e-05}
{"train_loss": 0.4272514879703522, "global_step": 26101, "epoch": 293, "lr": 9.781315624776811e-05}
{"train_loss": 0.32157692313194275, "global_step": 26102, "epoch": 293, "lr": 9.781298666705316e-05}
{"train_loss": 0.5381781458854675, "global_step": 26103, "epoch": 293, "lr": 9.781281707991033e-05}
{"train_loss": 0.3875730037689209, "global_step": 26104, "epoch": 293, "lr": 9.781264748633964e-05}
{"train_loss": 0.4714399576187134, "global_step": 26105, "epoch": 293, "lr": 9.781247788634114e-05}
{"train_loss": 0.29839882254600525, "global_step": 26106, "epoch": 293, "lr": 9.781230827991484e-05}
{"train_loss": 0.2792549133300781, "global_step": 26107, "epoch": 293, "lr": 9.781213866706074e-05}
{"train_loss": 0.46283942461013794, "global_step": 26108, "epoch": 293, "lr": 9.781196904777887e-05}
{"train_loss": 0.4762503206729889, "global_step": 26109, "epoch": 293, "lr": 9.781179942206929e-05}
{"train_loss": 0.4084513187408447, "global_step": 26110, "epoch": 293, "lr": 9.781162978993199e-05}
{"train_loss": 0.3605467677116394, "global_step": 26111, "epoch": 293, "lr": 9.781146015136699e-05}
{"train_loss": 0.33245745301246643, "global_step": 26112, "epoch": 293, "lr": 9.781129050637432e-05}
{"train_loss": 0.3856865465641022, "global_step": 26113, "epoch": 293, "lr": 9.781112085495402e-05}
{"train_loss": 0.3249174952507019, "global_step": 26114, "epoch": 293, "lr": 9.781095119710608e-05}
{"train_loss": 0.4275651276111603, "global_step": 26115, "epoch": 293, "lr": 9.781078153283056e-05}
{"train_loss": 0.30641037225723267, "global_step": 26116, "epoch": 293, "lr": 9.781061186212745e-05}
{"train_loss": 0.3465900719165802, "global_step": 26117, "epoch": 293, "lr": 9.78104421849968e-05}
{"train_loss": 0.31624653935432434, "global_step": 26118, "epoch": 293, "lr": 9.78102725014386e-05}
{"train_loss": 0.4231468141078949, "global_step": 26119, "epoch": 293, "lr": 9.78101028114529e-05}
{"train_loss": 0.4092613756656647, "global_step": 26120, "epoch": 293, "lr": 9.780993311503972e-05}
{"train_loss": 0.3594312369823456, "global_step": 26121, "epoch": 293, "lr": 9.780976341219907e-05}
{"train_loss": 0.4074017107486725, "global_step": 26122, "epoch": 293, "lr": 9.780959370293097e-05}
{"train_loss": 0.3438566029071808, "global_step": 26123, "epoch": 293, "lr": 9.780942398723547e-05}
{"train_loss": 0.34718239307403564, "global_step": 26124, "epoch": 293, "lr": 9.780925426511257e-05}
{"train_loss": 0.5138213634490967, "global_step": 26125, "epoch": 293, "lr": 9.78090845365623e-05}
{"train_loss": 0.3446587026119232, "global_step": 26126, "epoch": 293, "lr": 9.780891480158467e-05}
{"train_loss": 0.34687384963035583, "global_step": 26127, "epoch": 293, "lr": 9.780874506017973e-05}
{"train_loss": 0.41367366909980774, "global_step": 26128, "epoch": 293, "lr": 9.780857531234748e-05}
{"train_loss": 0.4562259912490845, "global_step": 26129, "epoch": 293, "lr": 9.780840555808795e-05}
{"train_loss": 0.32711076736450195, "global_step": 26130, "epoch": 293, "lr": 9.780823579740117e-05}
{"train_loss": 0.3838450014591217, "global_step": 26131, "epoch": 293, "lr": 9.780806603028714e-05}
{"train_loss": 0.36718934774398804, "global_step": 26132, "epoch": 293, "lr": 9.780789625674591e-05}
{"train_loss": 0.41270771622657776, "global_step": 26133, "epoch": 293, "lr": 9.780772647677748e-05}
{"train_loss": 0.4944859743118286, "global_step": 26134, "epoch": 293, "lr": 9.78075566903819e-05}
{"train_loss": 0.362373411655426, "global_step": 26135, "epoch": 293, "lr": 9.780738689755918e-05}
{"train_loss": 0.3851359188556671, "global_step": 26136, "epoch": 293, "lr": 9.780721709830931e-05}
{"train_loss": 0.425191193819046, "global_step": 26137, "epoch": 293, "lr": 9.780704729263237e-05}
{"train_loss": 0.44600579142570496, "global_step": 26138, "epoch": 293, "lr": 9.780687748052833e-05}
{"train_loss": 0.38881078362464905, "global_step": 26139, "epoch": 293, "lr": 9.780670766199727e-05}
{"train_loss": 0.3911507725715637, "global_step": 26140, "epoch": 293, "lr": 9.780653783703915e-05}
{"train_loss": 0.47045886516571045, "global_step": 26141, "epoch": 293, "lr": 9.780636800565403e-05}
{"train_loss": 0.28664159774780273, "global_step": 26142, "epoch": 293, "lr": 9.780619816784195e-05}
{"train_loss": 0.4509705603122711, "global_step": 26143, "epoch": 293, "lr": 9.780602832360288e-05}
{"train_loss": 0.45749232172966003, "global_step": 26144, "epoch": 293, "lr": 9.780585847293689e-05}
{"train_loss": 0.3865693509578705, "global_step": 26145, "epoch": 293, "lr": 9.780568861584398e-05}
{"train_loss": 0.47310054302215576, "global_step": 26146, "epoch": 293, "lr": 9.780551875232418e-05}
{"train_loss": 0.5049647688865662, "global_step": 26147, "epoch": 293, "lr": 9.780534888237751e-05}
{"train_loss": 0.47691217064857483, "global_step": 26148, "epoch": 293, "lr": 9.780517900600399e-05}
{"train_loss": 0.46720820665359497, "global_step": 26149, "epoch": 293, "lr": 9.780500912320365e-05}
{"train_loss": 0.43729597330093384, "global_step": 26150, "epoch": 293, "lr": 9.780483923397652e-05}
{"train_loss": 0.42212194204330444, "global_step": 26151, "epoch": 293, "lr": 9.78046693383226e-05}
{"train_loss": 0.34635913372039795, "global_step": 26152, "epoch": 293, "lr": 9.780449943624192e-05}
{"train_loss": 0.3096370995044708, "global_step": 26153, "epoch": 293, "lr": 9.78043295277345e-05}
{"train_loss": 0.40207144618034363, "global_step": 26154, "epoch": 293, "lr": 9.78041596128004e-05}
{"train_loss": 0.463786780834198, "global_step": 26155, "epoch": 293, "lr": 9.780398969143959e-05}
{"train_loss": 0.34853804111480713, "global_step": 26156, "epoch": 293, "lr": 9.780381976365211e-05}
{"train_loss": 0.4060208201408386, "global_step": 26157, "epoch": 293, "lr": 9.780364982943802e-05}
{"train_loss": 0.35538190603256226, "global_step": 26158, "epoch": 293, "lr": 9.780347988879729e-05}
{"train_loss": 0.4440476596355438, "global_step": 26159, "epoch": 293, "lr": 9.780330994172997e-05}
{"train_loss": 0.4119788408279419, "global_step": 26160, "epoch": 293, "lr": 9.780313998823608e-05}
{"train_loss": 0.42707088589668274, "global_step": 26161, "epoch": 293, "lr": 9.780297002831563e-05}
{"train_loss": 0.3682398498058319, "global_step": 26162, "epoch": 293, "lr": 9.780280006196866e-05}
{"train_loss": 0.32670626044273376, "global_step": 26163, "epoch": 293, "lr": 9.780263008919518e-05}
{"train_loss": 0.453343003988266, "global_step": 26164, "epoch": 293, "lr": 9.780246010999523e-05}
{"train_loss": 0.3974740937184752, "global_step": 26165, "epoch": 293, "lr": 9.780229012436881e-05, "val_loss": 2.0813143253326416}
{"train_loss": 0.36454737186431885, "global_step": 26166, "epoch": 294, "lr": 9.780212013231598e-05}
{"train_loss": 0.3674013614654541, "global_step": 26167, "epoch": 294, "lr": 9.78019501338367e-05}
{"train_loss": 0.4812966287136078, "global_step": 26168, "epoch": 294, "lr": 9.780178012893107e-05}
{"train_loss": 0.45400235056877136, "global_step": 26169, "epoch": 294, "lr": 9.780161011759904e-05}
{"train_loss": 0.48398280143737793, "global_step": 26170, "epoch": 294, "lr": 9.780144009984068e-05}
{"train_loss": 0.353405624628067, "global_step": 26171, "epoch": 294, "lr": 9.7801270075656e-05}
{"train_loss": 0.4218764901161194, "global_step": 26172, "epoch": 294, "lr": 9.780110004504501e-05}
{"train_loss": 0.44763413071632385, "global_step": 26173, "epoch": 294, "lr": 9.780093000800776e-05}
{"train_loss": 0.4499864876270294, "global_step": 26174, "epoch": 294, "lr": 9.780075996454425e-05}
{"train_loss": 0.30544811487197876, "global_step": 26175, "epoch": 294, "lr": 9.780058991465452e-05}
{"train_loss": 0.4455476999282837, "global_step": 26176, "epoch": 294, "lr": 9.780041985833857e-05}
{"train_loss": 0.4393414258956909, "global_step": 26177, "epoch": 294, "lr": 9.780024979559646e-05}
{"train_loss": 0.4150961637496948, "global_step": 26178, "epoch": 294, "lr": 9.780007972642816e-05}
{"train_loss": 0.4898429811000824, "global_step": 26179, "epoch": 294, "lr": 9.779990965083373e-05}
{"train_loss": 0.45003336668014526, "global_step": 26180, "epoch": 294, "lr": 9.779973956881319e-05}
{"train_loss": 0.5065180063247681, "global_step": 26181, "epoch": 294, "lr": 9.779956948036655e-05}
{"train_loss": 0.49708470702171326, "global_step": 26182, "epoch": 294, "lr": 9.779939938549385e-05}
{"train_loss": 0.4028228223323822, "global_step": 26183, "epoch": 294, "lr": 9.779922928419511e-05}
{"train_loss": 0.45747554302215576, "global_step": 26184, "epoch": 294, "lr": 9.779905917647031e-05}
{"train_loss": 0.43331289291381836, "global_step": 26185, "epoch": 294, "lr": 9.779888906231955e-05}
{"train_loss": 0.389550119638443, "global_step": 26186, "epoch": 294, "lr": 9.77987189417428e-05}
{"train_loss": 0.4544275999069214, "global_step": 26187, "epoch": 294, "lr": 9.779854881474009e-05}
{"train_loss": 0.4579208493232727, "global_step": 26188, "epoch": 294, "lr": 9.779837868131145e-05}
{"train_loss": 0.5525754690170288, "global_step": 26189, "epoch": 294, "lr": 9.779820854145689e-05}
{"train_loss": 0.4217201769351959, "global_step": 26190, "epoch": 294, "lr": 9.779803839517646e-05}
{"train_loss": 0.461599200963974, "global_step": 26191, "epoch": 294, "lr": 9.779786824247017e-05}
{"train_loss": 0.3714166283607483, "global_step": 26192, "epoch": 294, "lr": 9.779769808333801e-05}
{"train_loss": 0.46973294019699097, "global_step": 26193, "epoch": 294, "lr": 9.779752791778006e-05}
{"train_loss": 0.3912145793437958, "global_step": 26194, "epoch": 294, "lr": 9.77973577457963e-05}
{"train_loss": 0.36570972204208374, "global_step": 26195, "epoch": 294, "lr": 9.779718756738678e-05}
{"train_loss": 0.400047242641449, "global_step": 26196, "epoch": 294, "lr": 9.77970173825515e-05}
{"train_loss": 0.5018743872642517, "global_step": 26197, "epoch": 294, "lr": 9.779684719129049e-05}
{"train_loss": 0.40073898434638977, "global_step": 26198, "epoch": 294, "lr": 9.779667699360379e-05}
{"train_loss": 0.44405853748321533, "global_step": 26199, "epoch": 294, "lr": 9.779650678949142e-05}
{"train_loss": 0.4427564740180969, "global_step": 26200, "epoch": 294, "lr": 9.779633657895336e-05}
{"train_loss": 0.40396103262901306, "global_step": 26201, "epoch": 294, "lr": 9.779616636198969e-05}
{"train_loss": 0.36060258746147156, "global_step": 26202, "epoch": 294, "lr": 9.77959961386004e-05}
{"train_loss": 0.3730013370513916, "global_step": 26203, "epoch": 294, "lr": 9.779582590878552e-05}
{"train_loss": 0.4158489406108856, "global_step": 26204, "epoch": 294, "lr": 9.779565567254507e-05}
{"train_loss": 0.3954675793647766, "global_step": 26205, "epoch": 294, "lr": 9.779548542987908e-05}
{"train_loss": 0.3418337106704712, "global_step": 26206, "epoch": 294, "lr": 9.779531518078758e-05}
{"train_loss": 0.4769032299518585, "global_step": 26207, "epoch": 294, "lr": 9.779514492527055e-05}
{"train_loss": 0.2272561639547348, "global_step": 26208, "epoch": 294, "lr": 9.779497466332808e-05}
{"train_loss": 0.421954482793808, "global_step": 26209, "epoch": 294, "lr": 9.779480439496014e-05}
{"train_loss": 0.4645984172821045, "global_step": 26210, "epoch": 294, "lr": 9.779463412016678e-05}
{"train_loss": 0.4675760865211487, "global_step": 26211, "epoch": 294, "lr": 9.779446383894801e-05}
{"train_loss": 0.3586714565753937, "global_step": 26212, "epoch": 294, "lr": 9.779429355130385e-05}
{"train_loss": 0.30756819248199463, "global_step": 26213, "epoch": 294, "lr": 9.779412325723434e-05}
{"train_loss": 0.4886850416660309, "global_step": 26214, "epoch": 294, "lr": 9.77939529567395e-05}
{"train_loss": 0.49417170882225037, "global_step": 26215, "epoch": 294, "lr": 9.779378264981933e-05}
{"train_loss": 0.4928129017353058, "global_step": 26216, "epoch": 294, "lr": 9.779361233647387e-05}
{"train_loss": 0.32266539335250854, "global_step": 26217, "epoch": 294, "lr": 9.779344201670315e-05}
{"train_loss": 0.39191171526908875, "global_step": 26218, "epoch": 294, "lr": 9.779327169050719e-05}
{"train_loss": 0.32537636160850525, "global_step": 26219, "epoch": 294, "lr": 9.779310135788598e-05}
{"train_loss": 0.3567957580089569, "global_step": 26220, "epoch": 294, "lr": 9.77929310188396e-05}
{"train_loss": 0.37942183017730713, "global_step": 26221, "epoch": 294, "lr": 9.779276067336802e-05}
{"train_loss": 0.42983412742614746, "global_step": 26222, "epoch": 294, "lr": 9.77925903214713e-05}
{"train_loss": 0.36341750621795654, "global_step": 26223, "epoch": 294, "lr": 9.779241996314944e-05}
{"train_loss": 0.3678645193576813, "global_step": 26224, "epoch": 294, "lr": 9.779224959840247e-05}
{"train_loss": 0.3832434117794037, "global_step": 26225, "epoch": 294, "lr": 9.779207922723042e-05}
{"train_loss": 0.400189071893692, "global_step": 26226, "epoch": 294, "lr": 9.779190884963331e-05}
{"train_loss": 0.49470457434654236, "global_step": 26227, "epoch": 294, "lr": 9.779173846561116e-05}
{"train_loss": 0.4001135528087616, "global_step": 26228, "epoch": 294, "lr": 9.7791568075164e-05}
{"train_loss": 0.42241930961608887, "global_step": 26229, "epoch": 294, "lr": 9.779139767829183e-05}
{"train_loss": 0.39594393968582153, "global_step": 26230, "epoch": 294, "lr": 9.77912272749947e-05}
{"train_loss": 0.4996360242366791, "global_step": 26231, "epoch": 294, "lr": 9.779105686527261e-05}
{"train_loss": 0.4230716824531555, "global_step": 26232, "epoch": 294, "lr": 9.779088644912562e-05}
{"train_loss": 0.4251129925251007, "global_step": 26233, "epoch": 294, "lr": 9.779071602655369e-05}
{"train_loss": 0.4153068959712982, "global_step": 26234, "epoch": 294, "lr": 9.779054559755692e-05}
{"train_loss": 0.48419296741485596, "global_step": 26235, "epoch": 294, "lr": 9.779037516213527e-05}
{"train_loss": 0.3863256573677063, "global_step": 26236, "epoch": 294, "lr": 9.779020472028879e-05}
{"train_loss": 0.4574202001094818, "global_step": 26237, "epoch": 294, "lr": 9.77900342720175e-05}
{"train_loss": 0.4155879318714142, "global_step": 26238, "epoch": 294, "lr": 9.778986381732142e-05}
{"train_loss": 0.4309559762477875, "global_step": 26239, "epoch": 294, "lr": 9.778969335620059e-05}
{"train_loss": 0.5104160904884338, "global_step": 26240, "epoch": 294, "lr": 9.7789522888655e-05}
{"train_loss": 0.2751346230506897, "global_step": 26241, "epoch": 294, "lr": 9.77893524146847e-05}
{"train_loss": 0.41325995326042175, "global_step": 26242, "epoch": 294, "lr": 9.77891819342897e-05}
{"train_loss": 0.5673561096191406, "global_step": 26243, "epoch": 294, "lr": 9.778901144747004e-05}
{"train_loss": 0.35840994119644165, "global_step": 26244, "epoch": 294, "lr": 9.778884095422571e-05}
{"train_loss": 0.45757558941841125, "global_step": 26245, "epoch": 294, "lr": 9.778867045455677e-05}
{"train_loss": 0.4074896574020386, "global_step": 26246, "epoch": 294, "lr": 9.778849994846322e-05}
{"train_loss": 0.3812416195869446, "global_step": 26247, "epoch": 294, "lr": 9.778832943594508e-05}
{"train_loss": 0.35792601108551025, "global_step": 26248, "epoch": 294, "lr": 9.77881589170024e-05}
{"train_loss": 0.34648409485816956, "global_step": 26249, "epoch": 294, "lr": 9.778798839163517e-05}
{"train_loss": 0.32234305143356323, "global_step": 26250, "epoch": 294, "lr": 9.778781785984344e-05}
{"train_loss": 0.3160879909992218, "global_step": 26251, "epoch": 294, "lr": 9.77876473216272e-05}
{"train_loss": 0.3785403072834015, "global_step": 26252, "epoch": 294, "lr": 9.77874767769865e-05}
{"train_loss": 0.2736722230911255, "global_step": 26253, "epoch": 294, "lr": 9.778730622592136e-05}
{"train_loss": 0.4134942468967331, "global_step": 26254, "epoch": 294, "lr": 9.77871356684318e-05, "val_loss": 2.02510142326355}
{"train_loss": 0.4398501217365265, "global_step": 26255, "epoch": 295, "lr": 9.778696510451783e-05}
{"train_loss": 0.3169889450073242, "global_step": 26256, "epoch": 295, "lr": 9.77867945341795e-05}
{"train_loss": 0.4338010549545288, "global_step": 26257, "epoch": 295, "lr": 9.778662395741682e-05}
{"train_loss": 0.2854757010936737, "global_step": 26258, "epoch": 295, "lr": 9.77864533742298e-05}
{"train_loss": 0.3623863160610199, "global_step": 26259, "epoch": 295, "lr": 9.778628278461848e-05}
{"train_loss": 0.3108033537864685, "global_step": 26260, "epoch": 295, "lr": 9.778611218858287e-05}
{"train_loss": 0.4639165699481964, "global_step": 26261, "epoch": 295, "lr": 9.778594158612302e-05}
{"train_loss": 0.4395137429237366, "global_step": 26262, "epoch": 295, "lr": 9.77857709772389e-05}
{"train_loss": 0.5595583319664001, "global_step": 26263, "epoch": 295, "lr": 9.778560036193059e-05}
{"train_loss": 0.43172967433929443, "global_step": 26264, "epoch": 295, "lr": 9.778542974019808e-05}
{"train_loss": 0.4277461767196655, "global_step": 26265, "epoch": 295, "lr": 9.778525911204142e-05}
{"train_loss": 0.4225211441516876, "global_step": 26266, "epoch": 295, "lr": 9.778508847746059e-05}
{"train_loss": 0.49002376198768616, "global_step": 26267, "epoch": 295, "lr": 9.778491783645565e-05}
{"train_loss": 0.34516388177871704, "global_step": 26268, "epoch": 295, "lr": 9.778474718902659e-05}
{"train_loss": 0.4411507546901703, "global_step": 26269, "epoch": 295, "lr": 9.778457653517348e-05}
{"train_loss": 0.38918837904930115, "global_step": 26270, "epoch": 295, "lr": 9.77844058748963e-05}
{"train_loss": 0.3649570345878601, "global_step": 26271, "epoch": 295, "lr": 9.778423520819508e-05}
{"train_loss": 0.4895658791065216, "global_step": 26272, "epoch": 295, "lr": 9.778406453506987e-05}
{"train_loss": 0.39540812373161316, "global_step": 26273, "epoch": 295, "lr": 9.778389385552067e-05}
{"train_loss": 0.4079018831253052, "global_step": 26274, "epoch": 295, "lr": 9.778372316954752e-05}
{"train_loss": 0.4319882094860077, "global_step": 26275, "epoch": 295, "lr": 9.778355247715041e-05}
{"train_loss": 0.4432705342769623, "global_step": 26276, "epoch": 295, "lr": 9.778338177832939e-05}
{"train_loss": 0.39637184143066406, "global_step": 26277, "epoch": 295, "lr": 9.778321107308448e-05}
{"train_loss": 0.45747044682502747, "global_step": 26278, "epoch": 295, "lr": 9.77830403614157e-05}
{"train_loss": 0.42797428369522095, "global_step": 26279, "epoch": 295, "lr": 9.778286964332309e-05}
{"train_loss": 0.4661979079246521, "global_step": 26280, "epoch": 295, "lr": 9.778269891880663e-05}
{"train_loss": 0.3419622778892517, "global_step": 26281, "epoch": 295, "lr": 9.778252818786638e-05}
{"train_loss": 0.37194401025772095, "global_step": 26282, "epoch": 295, "lr": 9.778235745050234e-05}
{"train_loss": 0.35655391216278076, "global_step": 26283, "epoch": 295, "lr": 9.778218670671455e-05}
{"train_loss": 0.41072607040405273, "global_step": 26284, "epoch": 295, "lr": 9.778201595650303e-05}
{"train_loss": 0.460890531539917, "global_step": 26285, "epoch": 295, "lr": 9.77818451998678e-05}
{"train_loss": 0.264190137386322, "global_step": 26286, "epoch": 295, "lr": 9.778167443680888e-05}
{"train_loss": 0.5372572541236877, "global_step": 26287, "epoch": 295, "lr": 9.77815036673263e-05}
{"train_loss": 0.42489662766456604, "global_step": 26288, "epoch": 295, "lr": 9.778133289142007e-05}
{"train_loss": 0.33796483278274536, "global_step": 26289, "epoch": 295, "lr": 9.778116210909023e-05}
{"train_loss": 0.49938639998435974, "global_step": 26290, "epoch": 295, "lr": 9.77809913203368e-05}
{"train_loss": 0.44481509923934937, "global_step": 26291, "epoch": 295, "lr": 9.77808205251598e-05}
{"train_loss": 0.3745274543762207, "global_step": 26292, "epoch": 295, "lr": 9.778064972355923e-05}
{"train_loss": 0.3445453941822052, "global_step": 26293, "epoch": 295, "lr": 9.778047891553516e-05}
{"train_loss": 0.32986360788345337, "global_step": 26294, "epoch": 295, "lr": 9.778030810108757e-05}
{"train_loss": 0.396525114774704, "global_step": 26295, "epoch": 295, "lr": 9.778013728021651e-05}
{"train_loss": 0.5573760271072388, "global_step": 26296, "epoch": 295, "lr": 9.777996645292198e-05}
{"train_loss": 0.44903069734573364, "global_step": 26297, "epoch": 295, "lr": 9.777979561920402e-05}
{"train_loss": 0.4035079777240753, "global_step": 26298, "epoch": 295, "lr": 9.777962477906266e-05}
{"train_loss": 0.37184205651283264, "global_step": 26299, "epoch": 295, "lr": 9.77794539324979e-05}
{"train_loss": 0.4187621474266052, "global_step": 26300, "epoch": 295, "lr": 9.777928307950978e-05}
{"train_loss": 0.37163302302360535, "global_step": 26301, "epoch": 295, "lr": 9.777911222009833e-05}
{"train_loss": 0.3243119418621063, "global_step": 26302, "epoch": 295, "lr": 9.777894135426354e-05}
{"train_loss": 0.36852413415908813, "global_step": 26303, "epoch": 295, "lr": 9.777877048200545e-05}
{"train_loss": 0.3716776371002197, "global_step": 26304, "epoch": 295, "lr": 9.777859960332412e-05}
{"train_loss": 0.39677542448043823, "global_step": 26305, "epoch": 295, "lr": 9.777842871821952e-05}
{"train_loss": 0.4287652373313904, "global_step": 26306, "epoch": 295, "lr": 9.777825782669168e-05}
{"train_loss": 0.40668532252311707, "global_step": 26307, "epoch": 295, "lr": 9.777808692874064e-05}
{"train_loss": 0.4109150171279907, "global_step": 26308, "epoch": 295, "lr": 9.777791602436644e-05}
{"train_loss": 0.5389093160629272, "global_step": 26309, "epoch": 295, "lr": 9.777774511356905e-05}
{"train_loss": 0.48066404461860657, "global_step": 26310, "epoch": 295, "lr": 9.777757419634855e-05}
{"train_loss": 0.36845919489860535, "global_step": 26311, "epoch": 295, "lr": 9.777740327270493e-05}
{"train_loss": 0.3941698670387268, "global_step": 26312, "epoch": 295, "lr": 9.777723234263822e-05}
{"train_loss": 0.4579135775566101, "global_step": 26313, "epoch": 295, "lr": 9.777706140614845e-05}
{"train_loss": 0.3495848476886749, "global_step": 26314, "epoch": 295, "lr": 9.777689046323561e-05}
{"train_loss": 0.4008466601371765, "global_step": 26315, "epoch": 295, "lr": 9.777671951389978e-05}
{"train_loss": 0.3468378782272339, "global_step": 26316, "epoch": 295, "lr": 9.777654855814094e-05}
{"train_loss": 0.38123026490211487, "global_step": 26317, "epoch": 295, "lr": 9.777637759595912e-05}
{"train_loss": 0.3144104778766632, "global_step": 26318, "epoch": 295, "lr": 9.777620662735435e-05}
{"train_loss": 0.4181848168373108, "global_step": 26319, "epoch": 295, "lr": 9.777603565232665e-05}
{"train_loss": 0.3130864202976227, "global_step": 26320, "epoch": 295, "lr": 9.777586467087606e-05}
{"train_loss": 0.36172181367874146, "global_step": 26321, "epoch": 295, "lr": 9.777569368300256e-05}
{"train_loss": 0.2907123863697052, "global_step": 26322, "epoch": 295, "lr": 9.777552268870622e-05}
{"train_loss": 0.40331658720970154, "global_step": 26323, "epoch": 295, "lr": 9.777535168798705e-05}
{"train_loss": 0.367825984954834, "global_step": 26324, "epoch": 295, "lr": 9.777518068084506e-05}
{"train_loss": 0.43739551305770874, "global_step": 26325, "epoch": 295, "lr": 9.777500966728027e-05}
{"train_loss": 0.3102453351020813, "global_step": 26326, "epoch": 295, "lr": 9.777483864729273e-05}
{"train_loss": 0.4311470687389374, "global_step": 26327, "epoch": 295, "lr": 9.777466762088244e-05}
{"train_loss": 0.42329469323158264, "global_step": 26328, "epoch": 295, "lr": 9.77744965880494e-05}
{"train_loss": 0.431804358959198, "global_step": 26329, "epoch": 295, "lr": 9.777432554879368e-05}
{"train_loss": 0.598246157169342, "global_step": 26330, "epoch": 295, "lr": 9.777415450311529e-05}
{"train_loss": 0.43000322580337524, "global_step": 26331, "epoch": 295, "lr": 9.777398345101426e-05}
{"train_loss": 0.40180712938308716, "global_step": 26332, "epoch": 295, "lr": 9.777381239249057e-05}
{"train_loss": 0.4943438172340393, "global_step": 26333, "epoch": 295, "lr": 9.77736413275443e-05}
{"train_loss": 0.41325142979621887, "global_step": 26334, "epoch": 295, "lr": 9.777347025617543e-05}
{"train_loss": 0.4663708209991455, "global_step": 26335, "epoch": 295, "lr": 9.7773299178384e-05}
{"train_loss": 0.37501245737075806, "global_step": 26336, "epoch": 295, "lr": 9.777312809417002e-05}
{"train_loss": 0.36397385597229004, "global_step": 26337, "epoch": 295, "lr": 9.777295700353355e-05}
{"train_loss": 0.47463157773017883, "global_step": 26338, "epoch": 295, "lr": 9.777278590647458e-05}
{"train_loss": 0.564080536365509, "global_step": 26339, "epoch": 295, "lr": 9.777261480299314e-05}
{"train_loss": 0.23834992945194244, "global_step": 26340, "epoch": 295, "lr": 9.777244369308927e-05}
{"train_loss": 0.46981650590896606, "global_step": 26341, "epoch": 295, "lr": 9.777227257676296e-05}
{"train_loss": 0.41395628452301025, "global_step": 26342, "epoch": 295, "lr": 9.777210145401424e-05}
{"train_loss": 0.4083794959810343, "global_step": 26343, "epoch": 295, "lr": 9.777193032484316e-05, "val_loss": 2.0619900226593018, "train_action_mse_error": 25.056819915771484}
{"train_loss": 0.36918655037879944, "global_step": 26344, "epoch": 296, "lr": 9.777175918924972e-05}
{"train_loss": 0.3827877938747406, "global_step": 26345, "epoch": 296, "lr": 9.777158804723394e-05}
{"train_loss": 0.37519386410713196, "global_step": 26346, "epoch": 296, "lr": 9.777141689879588e-05}
{"train_loss": 0.403283953666687, "global_step": 26347, "epoch": 296, "lr": 9.77712457439355e-05}
{"train_loss": 0.42153504490852356, "global_step": 26348, "epoch": 296, "lr": 9.777107458265289e-05}
{"train_loss": 0.3718734383583069, "global_step": 26349, "epoch": 296, "lr": 9.777090341494803e-05}
{"train_loss": 0.4615684747695923, "global_step": 26350, "epoch": 296, "lr": 9.777073224082094e-05}
{"train_loss": 0.4421340823173523, "global_step": 26351, "epoch": 296, "lr": 9.777056106027167e-05}
{"train_loss": 0.3150536119937897, "global_step": 26352, "epoch": 296, "lr": 9.777038987330022e-05}
{"train_loss": 0.4244862496852875, "global_step": 26353, "epoch": 296, "lr": 9.777021867990663e-05}
{"train_loss": 0.3359818160533905, "global_step": 26354, "epoch": 296, "lr": 9.777004748009092e-05}
{"train_loss": 0.3192099928855896, "global_step": 26355, "epoch": 296, "lr": 9.776987627385313e-05}
{"train_loss": 0.31388846039772034, "global_step": 26356, "epoch": 296, "lr": 9.776970506119323e-05}
{"train_loss": 0.3507002890110016, "global_step": 26357, "epoch": 296, "lr": 9.776953384211128e-05}
{"train_loss": 0.4937300384044647, "global_step": 26358, "epoch": 296, "lr": 9.77693626166073e-05}
{"train_loss": 0.44635164737701416, "global_step": 26359, "epoch": 296, "lr": 9.776919138468133e-05}
{"train_loss": 0.40715762972831726, "global_step": 26360, "epoch": 296, "lr": 9.776902014633335e-05}
{"train_loss": 0.40810877084732056, "global_step": 26361, "epoch": 296, "lr": 9.776884890156341e-05}
{"train_loss": 0.3503725826740265, "global_step": 26362, "epoch": 296, "lr": 9.776867765037154e-05}
{"train_loss": 0.35696735978126526, "global_step": 26363, "epoch": 296, "lr": 9.776850639275777e-05}
{"train_loss": 0.3710179924964905, "global_step": 26364, "epoch": 296, "lr": 9.776833512872207e-05}
{"train_loss": 0.34413468837738037, "global_step": 26365, "epoch": 296, "lr": 9.776816385826453e-05}
{"train_loss": 0.3537604510784149, "global_step": 26366, "epoch": 296, "lr": 9.776799258138513e-05}
{"train_loss": 0.4553692638874054, "global_step": 26367, "epoch": 296, "lr": 9.77678212980839e-05}
{"train_loss": 0.3145497441291809, "global_step": 26368, "epoch": 296, "lr": 9.776765000836087e-05}
{"train_loss": 0.49424710869789124, "global_step": 26369, "epoch": 296, "lr": 9.776747871221608e-05}
{"train_loss": 0.4082631766796112, "global_step": 26370, "epoch": 296, "lr": 9.776730740964952e-05}
{"train_loss": 0.4543156027793884, "global_step": 26371, "epoch": 296, "lr": 9.776713610066122e-05}
{"train_loss": 0.49129992723464966, "global_step": 26372, "epoch": 296, "lr": 9.776696478525123e-05}
{"train_loss": 0.3514210879802704, "global_step": 26373, "epoch": 296, "lr": 9.776679346341953e-05}
{"train_loss": 0.4747534692287445, "global_step": 26374, "epoch": 296, "lr": 9.776662213516619e-05}
{"train_loss": 0.4923924505710602, "global_step": 26375, "epoch": 296, "lr": 9.77664508004912e-05}
{"train_loss": 0.476667195558548, "global_step": 26376, "epoch": 296, "lr": 9.776627945939459e-05}
{"train_loss": 0.491464227437973, "global_step": 26377, "epoch": 296, "lr": 9.776610811187638e-05}
{"train_loss": 0.2762872874736786, "global_step": 26378, "epoch": 296, "lr": 9.776593675793661e-05}
{"train_loss": 0.4311261773109436, "global_step": 26379, "epoch": 296, "lr": 9.77657653975753e-05}
{"train_loss": 0.48034635186195374, "global_step": 26380, "epoch": 296, "lr": 9.776559403079245e-05}
{"train_loss": 0.4275124967098236, "global_step": 26381, "epoch": 296, "lr": 9.77654226575881e-05}
{"train_loss": 0.4643622636795044, "global_step": 26382, "epoch": 296, "lr": 9.776525127796227e-05}
{"train_loss": 0.4323961138725281, "global_step": 26383, "epoch": 296, "lr": 9.776507989191499e-05}
{"train_loss": 0.3801048696041107, "global_step": 26384, "epoch": 296, "lr": 9.776490849944628e-05}
{"train_loss": 0.5069864392280579, "global_step": 26385, "epoch": 296, "lr": 9.776473710055616e-05}
{"train_loss": 0.39116141200065613, "global_step": 26386, "epoch": 296, "lr": 9.776456569524464e-05}
{"train_loss": 0.4346019923686981, "global_step": 26387, "epoch": 296, "lr": 9.776439428351176e-05}
{"train_loss": 0.38571834564208984, "global_step": 26388, "epoch": 296, "lr": 9.776422286535755e-05}
{"train_loss": 0.5088061690330505, "global_step": 26389, "epoch": 296, "lr": 9.776405144078202e-05}
{"train_loss": 0.3598312735557556, "global_step": 26390, "epoch": 296, "lr": 9.77638800097852e-05}
{"train_loss": 0.4490860104560852, "global_step": 26391, "epoch": 296, "lr": 9.77637085723671e-05}
{"train_loss": 0.30407536029815674, "global_step": 26392, "epoch": 296, "lr": 9.776353712852775e-05}
{"train_loss": 0.43285343050956726, "global_step": 26393, "epoch": 296, "lr": 9.776336567826717e-05}
{"train_loss": 0.4678577184677124, "global_step": 26394, "epoch": 296, "lr": 9.77631942215854e-05}
{"train_loss": 0.43531206250190735, "global_step": 26395, "epoch": 296, "lr": 9.776302275848244e-05}
{"train_loss": 0.3843970000743866, "global_step": 26396, "epoch": 296, "lr": 9.776285128895833e-05}
{"train_loss": 0.36344781517982483, "global_step": 26397, "epoch": 296, "lr": 9.776267981301308e-05}
{"train_loss": 0.32786068320274353, "global_step": 26398, "epoch": 296, "lr": 9.776250833064673e-05}
{"train_loss": 0.427667498588562, "global_step": 26399, "epoch": 296, "lr": 9.776233684185929e-05}
{"train_loss": 0.44615820050239563, "global_step": 26400, "epoch": 296, "lr": 9.776216534665078e-05}
{"train_loss": 0.44790971279144287, "global_step": 26401, "epoch": 296, "lr": 9.776199384502123e-05}
{"train_loss": 0.3805225193500519, "global_step": 26402, "epoch": 296, "lr": 9.776182233697067e-05}
{"train_loss": 0.42705684900283813, "global_step": 26403, "epoch": 296, "lr": 9.776165082249911e-05}
{"train_loss": 0.4057701826095581, "global_step": 26404, "epoch": 296, "lr": 9.776147930160657e-05}
{"train_loss": 0.49959132075309753, "global_step": 26405, "epoch": 296, "lr": 9.77613077742931e-05}
{"train_loss": 0.37762147188186646, "global_step": 26406, "epoch": 296, "lr": 9.77611362405587e-05}
{"train_loss": 0.3292232155799866, "global_step": 26407, "epoch": 296, "lr": 9.77609647004034e-05}
{"train_loss": 0.4085123538970947, "global_step": 26408, "epoch": 296, "lr": 9.77607931538272e-05}
{"train_loss": 0.35934579372406006, "global_step": 26409, "epoch": 296, "lr": 9.776062160083015e-05}
{"train_loss": 0.5443615317344666, "global_step": 26410, "epoch": 296, "lr": 9.776045004141227e-05}
{"train_loss": 0.4657827615737915, "global_step": 26411, "epoch": 296, "lr": 9.776027847557359e-05}
{"train_loss": 0.5307354927062988, "global_step": 26412, "epoch": 296, "lr": 9.776010690331411e-05}
{"train_loss": 0.48239365220069885, "global_step": 26413, "epoch": 296, "lr": 9.775993532463386e-05}
{"train_loss": 0.5485068559646606, "global_step": 26414, "epoch": 296, "lr": 9.775976373953289e-05}
{"train_loss": 0.3188154399394989, "global_step": 26415, "epoch": 296, "lr": 9.775959214801118e-05}
{"train_loss": 0.4284942150115967, "global_step": 26416, "epoch": 296, "lr": 9.775942055006879e-05}
{"train_loss": 0.5435227751731873, "global_step": 26417, "epoch": 296, "lr": 9.775924894570572e-05}
{"train_loss": 0.40903860330581665, "global_step": 26418, "epoch": 296, "lr": 9.7759077334922e-05}
{"train_loss": 0.45962944626808167, "global_step": 26419, "epoch": 296, "lr": 9.775890571771766e-05}
{"train_loss": 0.45573681592941284, "global_step": 26420, "epoch": 296, "lr": 9.775873409409273e-05}
{"train_loss": 0.3939675986766815, "global_step": 26421, "epoch": 296, "lr": 9.77585624640472e-05}
{"train_loss": 0.4294390082359314, "global_step": 26422, "epoch": 296, "lr": 9.775839082758112e-05}
{"train_loss": 0.523851215839386, "global_step": 26423, "epoch": 296, "lr": 9.77582191846945e-05}
{"train_loss": 0.33186712861061096, "global_step": 26424, "epoch": 296, "lr": 9.775804753538739e-05}
{"train_loss": 0.44486507773399353, "global_step": 26425, "epoch": 296, "lr": 9.775787587965977e-05}
{"train_loss": 0.39990776777267456, "global_step": 26426, "epoch": 296, "lr": 9.775770421751169e-05}
{"train_loss": 0.5355778336524963, "global_step": 26427, "epoch": 296, "lr": 9.775753254894318e-05}
{"train_loss": 0.3951822817325592, "global_step": 26428, "epoch": 296, "lr": 9.775736087395423e-05}
{"train_loss": 0.4049723744392395, "global_step": 26429, "epoch": 296, "lr": 9.775718919254491e-05}
{"train_loss": 0.5377558469772339, "global_step": 26430, "epoch": 296, "lr": 9.775701750471521e-05}
{"train_loss": 0.48081356287002563, "global_step": 26431, "epoch": 296, "lr": 9.775684581046515e-05}
{"train_loss": 0.41761451005265954, "global_step": 26432, "epoch": 296, "lr": 9.775667410979479e-05, "val_loss": 2.1483235359191895}
{"train_loss": 0.4718444347381592, "global_step": 26433, "epoch": 297, "lr": 9.77565024027041e-05}
{"train_loss": 0.5751907825469971, "global_step": 26434, "epoch": 297, "lr": 9.775633068919314e-05}
{"train_loss": 0.4296824634075165, "global_step": 26435, "epoch": 297, "lr": 9.775615896926194e-05}
{"train_loss": 0.5151426792144775, "global_step": 26436, "epoch": 297, "lr": 9.775598724291048e-05}
{"train_loss": 0.5448755025863647, "global_step": 26437, "epoch": 297, "lr": 9.775581551013882e-05}
{"train_loss": 0.40451890230178833, "global_step": 26438, "epoch": 297, "lr": 9.775564377094697e-05}
{"train_loss": 0.4611128866672516, "global_step": 26439, "epoch": 297, "lr": 9.775547202533494e-05}
{"train_loss": 0.4281255900859833, "global_step": 26440, "epoch": 297, "lr": 9.775530027330279e-05}
{"train_loss": 0.4275134801864624, "global_step": 26441, "epoch": 297, "lr": 9.775512851485054e-05}
{"train_loss": 0.5319551825523376, "global_step": 26442, "epoch": 297, "lr": 9.775495674997817e-05}
{"train_loss": 0.33196863532066345, "global_step": 26443, "epoch": 297, "lr": 9.775478497868573e-05}
{"train_loss": 0.33435869216918945, "global_step": 26444, "epoch": 297, "lr": 9.775461320097325e-05}
{"train_loss": 0.5127559304237366, "global_step": 26445, "epoch": 297, "lr": 9.775444141684074e-05}
{"train_loss": 0.41698282957077026, "global_step": 26446, "epoch": 297, "lr": 9.775426962628821e-05}
{"train_loss": 0.4847705662250519, "global_step": 26447, "epoch": 297, "lr": 9.775409782931573e-05}
{"train_loss": 0.4372541904449463, "global_step": 26448, "epoch": 297, "lr": 9.775392602592327e-05}
{"train_loss": 0.4679661691188812, "global_step": 26449, "epoch": 297, "lr": 9.775375421611088e-05}
{"train_loss": 0.4309971332550049, "global_step": 26450, "epoch": 297, "lr": 9.775358239987859e-05}
{"train_loss": 0.4801975190639496, "global_step": 26451, "epoch": 297, "lr": 9.77534105772264e-05}
{"train_loss": 0.37265169620513916, "global_step": 26452, "epoch": 297, "lr": 9.775323874815437e-05}
{"train_loss": 0.47013193368911743, "global_step": 26453, "epoch": 297, "lr": 9.775306691266249e-05}
{"train_loss": 0.414995402097702, "global_step": 26454, "epoch": 297, "lr": 9.775289507075078e-05}
{"train_loss": 0.3429095447063446, "global_step": 26455, "epoch": 297, "lr": 9.775272322241927e-05}
{"train_loss": 0.4867528975009918, "global_step": 26456, "epoch": 297, "lr": 9.775255136766802e-05}
{"train_loss": 0.3785974383354187, "global_step": 26457, "epoch": 297, "lr": 9.775237950649698e-05}
{"train_loss": 0.39941221475601196, "global_step": 26458, "epoch": 297, "lr": 9.775220763890624e-05}
{"train_loss": 0.36461347341537476, "global_step": 26459, "epoch": 297, "lr": 9.77520357648958e-05}
{"train_loss": 0.4187561273574829, "global_step": 26460, "epoch": 297, "lr": 9.775186388446566e-05}
{"train_loss": 0.5238311290740967, "global_step": 26461, "epoch": 297, "lr": 9.775169199761589e-05}
{"train_loss": 0.327886700630188, "global_step": 26462, "epoch": 297, "lr": 9.775152010434648e-05}
{"train_loss": 0.41320478916168213, "global_step": 26463, "epoch": 297, "lr": 9.775134820465744e-05}
{"train_loss": 0.4056662917137146, "global_step": 26464, "epoch": 297, "lr": 9.775117629854883e-05}
{"train_loss": 0.4807144105434418, "global_step": 26465, "epoch": 297, "lr": 9.775100438602065e-05}
{"train_loss": 0.33491772413253784, "global_step": 26466, "epoch": 297, "lr": 9.775083246707294e-05}
{"train_loss": 0.3865847587585449, "global_step": 26467, "epoch": 297, "lr": 9.77506605417057e-05}
{"train_loss": 0.43725281953811646, "global_step": 26468, "epoch": 297, "lr": 9.775048860991898e-05}
{"train_loss": 0.3181626796722412, "global_step": 26469, "epoch": 297, "lr": 9.775031667171278e-05}
{"train_loss": 0.5054531693458557, "global_step": 26470, "epoch": 297, "lr": 9.775014472708713e-05}
{"train_loss": 0.3553887903690338, "global_step": 26471, "epoch": 297, "lr": 9.774997277604205e-05}
{"train_loss": 0.3742277920246124, "global_step": 26472, "epoch": 297, "lr": 9.774980081857759e-05}
{"train_loss": 0.43738049268722534, "global_step": 26473, "epoch": 297, "lr": 9.774962885469374e-05}
{"train_loss": 0.45842254161834717, "global_step": 26474, "epoch": 297, "lr": 9.774945688439053e-05}
{"train_loss": 0.2943340539932251, "global_step": 26475, "epoch": 297, "lr": 9.774928490766799e-05}
{"train_loss": 0.4518871307373047, "global_step": 26476, "epoch": 297, "lr": 9.774911292452614e-05}
{"train_loss": 0.36364224553108215, "global_step": 26477, "epoch": 297, "lr": 9.774894093496501e-05}
{"train_loss": 0.5049152374267578, "global_step": 26478, "epoch": 297, "lr": 9.77487689389846e-05}
{"train_loss": 0.4520270824432373, "global_step": 26479, "epoch": 297, "lr": 9.774859693658496e-05}
{"train_loss": 0.3732891380786896, "global_step": 26480, "epoch": 297, "lr": 9.77484249277661e-05}
{"train_loss": 0.4083738625049591, "global_step": 26481, "epoch": 297, "lr": 9.774825291252805e-05}
{"train_loss": 0.3614405393600464, "global_step": 26482, "epoch": 297, "lr": 9.774808089087084e-05}
{"train_loss": 0.3790145516395569, "global_step": 26483, "epoch": 297, "lr": 9.774790886279447e-05}
{"train_loss": 0.380087286233902, "global_step": 26484, "epoch": 297, "lr": 9.774773682829897e-05}
{"train_loss": 0.3229913115501404, "global_step": 26485, "epoch": 297, "lr": 9.774756478738438e-05}
{"train_loss": 0.47231632471084595, "global_step": 26486, "epoch": 297, "lr": 9.774739274005071e-05}
{"train_loss": 0.4470524489879608, "global_step": 26487, "epoch": 297, "lr": 9.774722068629798e-05}
{"train_loss": 0.36190250515937805, "global_step": 26488, "epoch": 297, "lr": 9.774704862612622e-05}
{"train_loss": 0.3833925426006317, "global_step": 26489, "epoch": 297, "lr": 9.774687655953546e-05}
{"train_loss": 0.3157806396484375, "global_step": 26490, "epoch": 297, "lr": 9.77467044865257e-05}
{"train_loss": 0.44853198528289795, "global_step": 26491, "epoch": 297, "lr": 9.774653240709699e-05}
{"train_loss": 0.43329742550849915, "global_step": 26492, "epoch": 297, "lr": 9.774636032124933e-05}
{"train_loss": 0.4303620457649231, "global_step": 26493, "epoch": 297, "lr": 9.774618822898276e-05}
{"train_loss": 0.5478594899177551, "global_step": 26494, "epoch": 297, "lr": 9.77460161302973e-05}
{"train_loss": 0.48822280764579773, "global_step": 26495, "epoch": 297, "lr": 9.774584402519295e-05}
{"train_loss": 0.39134615659713745, "global_step": 26496, "epoch": 297, "lr": 9.774567191366978e-05}
{"train_loss": 0.44355735182762146, "global_step": 26497, "epoch": 297, "lr": 9.774549979572777e-05}
{"train_loss": 0.42329415678977966, "global_step": 26498, "epoch": 297, "lr": 9.774532767136698e-05}
{"train_loss": 0.4740959107875824, "global_step": 26499, "epoch": 297, "lr": 9.774515554058739e-05}
{"train_loss": 0.2844894826412201, "global_step": 26500, "epoch": 297, "lr": 9.774498340338904e-05}
{"train_loss": 0.2830131947994232, "global_step": 26501, "epoch": 297, "lr": 9.774481125977198e-05}
{"train_loss": 0.3064686059951782, "global_step": 26502, "epoch": 297, "lr": 9.774463910973621e-05}
{"train_loss": 0.36908096075057983, "global_step": 26503, "epoch": 297, "lr": 9.774446695328174e-05}
{"train_loss": 0.4632135331630707, "global_step": 26504, "epoch": 297, "lr": 9.774429479040863e-05}
{"train_loss": 0.5437538623809814, "global_step": 26505, "epoch": 297, "lr": 9.774412262111686e-05}
{"train_loss": 0.4654863774776459, "global_step": 26506, "epoch": 297, "lr": 9.774395044540649e-05}
{"train_loss": 0.3621673285961151, "global_step": 26507, "epoch": 297, "lr": 9.774377826327752e-05}
{"train_loss": 0.3641444444656372, "global_step": 26508, "epoch": 297, "lr": 9.774360607472998e-05}
{"train_loss": 0.35752537846565247, "global_step": 26509, "epoch": 297, "lr": 9.77434338797639e-05}
{"train_loss": 0.4454139769077301, "global_step": 26510, "epoch": 297, "lr": 9.77432616783793e-05}
{"train_loss": 0.4637472331523895, "global_step": 26511, "epoch": 297, "lr": 9.774308947057619e-05}
{"train_loss": 0.35546785593032837, "global_step": 26512, "epoch": 297, "lr": 9.77429172563546e-05}
{"train_loss": 0.5459004640579224, "global_step": 26513, "epoch": 297, "lr": 9.774274503571457e-05}
{"train_loss": 0.4521133005619049, "global_step": 26514, "epoch": 297, "lr": 9.774257280865611e-05}
{"train_loss": 0.3881358206272125, "global_step": 26515, "epoch": 297, "lr": 9.774240057517923e-05}
{"train_loss": 0.4078771770000458, "global_step": 26516, "epoch": 297, "lr": 9.774222833528398e-05}
{"train_loss": 0.6102273464202881, "global_step": 26517, "epoch": 297, "lr": 9.774205608897036e-05}
{"train_loss": 0.37109899520874023, "global_step": 26518, "epoch": 297, "lr": 9.774188383623841e-05}
{"train_loss": 0.4474817216396332, "global_step": 26519, "epoch": 297, "lr": 9.774171157708813e-05}
{"train_loss": 0.5398554801940918, "global_step": 26520, "epoch": 297, "lr": 9.774153931151957e-05}
{"train_loss": 0.4247495843453354, "global_step": 26521, "epoch": 297, "lr": 9.774136703953274e-05, "val_loss": 2.1174895763397217}
{"train_loss": 0.4449899196624756, "global_step": 26522, "epoch": 298, "lr": 9.774119476112767e-05}
{"train_loss": 0.4099017381668091, "global_step": 26523, "epoch": 298, "lr": 9.774102247630437e-05}
{"train_loss": 0.354125440120697, "global_step": 26524, "epoch": 298, "lr": 9.774085018506288e-05}
{"train_loss": 0.4482485353946686, "global_step": 26525, "epoch": 298, "lr": 9.774067788740322e-05}
{"train_loss": 0.39615917205810547, "global_step": 26526, "epoch": 298, "lr": 9.774050558332539e-05}
{"train_loss": 0.3769811987876892, "global_step": 26527, "epoch": 298, "lr": 9.774033327282944e-05}
{"train_loss": 0.42785799503326416, "global_step": 26528, "epoch": 298, "lr": 9.774016095591538e-05}
{"train_loss": 0.3564107120037079, "global_step": 26529, "epoch": 298, "lr": 9.773998863258325e-05}
{"train_loss": 0.39487308263778687, "global_step": 26530, "epoch": 298, "lr": 9.773981630283305e-05}
{"train_loss": 0.31347593665122986, "global_step": 26531, "epoch": 298, "lr": 9.773964396666482e-05}
{"train_loss": 0.375570684671402, "global_step": 26532, "epoch": 298, "lr": 9.773947162407857e-05}
{"train_loss": 0.49732837080955505, "global_step": 26533, "epoch": 298, "lr": 9.773929927507432e-05}
{"train_loss": 0.4097040593624115, "global_step": 26534, "epoch": 298, "lr": 9.773912691965212e-05}
{"train_loss": 0.3957338035106659, "global_step": 26535, "epoch": 298, "lr": 9.773895455781196e-05}
{"train_loss": 0.40415725111961365, "global_step": 26536, "epoch": 298, "lr": 9.77387821895539e-05}
{"train_loss": 0.33625030517578125, "global_step": 26537, "epoch": 298, "lr": 9.773860981487793e-05}
{"train_loss": 0.44617244601249695, "global_step": 26538, "epoch": 298, "lr": 9.773843743378409e-05}
{"train_loss": 0.3389962613582611, "global_step": 26539, "epoch": 298, "lr": 9.77382650462724e-05}
{"train_loss": 0.5198122262954712, "global_step": 26540, "epoch": 298, "lr": 9.773809265234289e-05}
{"train_loss": 0.37666115164756775, "global_step": 26541, "epoch": 298, "lr": 9.773792025199555e-05}
{"train_loss": 0.4413939416408539, "global_step": 26542, "epoch": 298, "lr": 9.773774784523045e-05}
{"train_loss": 0.42491310834884644, "global_step": 26543, "epoch": 298, "lr": 9.773757543204758e-05}
{"train_loss": 0.4408283829689026, "global_step": 26544, "epoch": 298, "lr": 9.773740301244699e-05}
{"train_loss": 0.47624528408050537, "global_step": 26545, "epoch": 298, "lr": 9.773723058642867e-05}
{"train_loss": 0.3277495801448822, "global_step": 26546, "epoch": 298, "lr": 9.773705815399268e-05}
{"train_loss": 0.3708516061306, "global_step": 26547, "epoch": 298, "lr": 9.773688571513901e-05}
{"train_loss": 0.39917486906051636, "global_step": 26548, "epoch": 298, "lr": 9.77367132698677e-05}
{"train_loss": 0.3902447521686554, "global_step": 26549, "epoch": 298, "lr": 9.773654081817877e-05}
{"train_loss": 0.476583331823349, "global_step": 26550, "epoch": 298, "lr": 9.773636836007225e-05}
{"train_loss": 0.42547592520713806, "global_step": 26551, "epoch": 298, "lr": 9.773619589554814e-05}
{"train_loss": 0.4905816316604614, "global_step": 26552, "epoch": 298, "lr": 9.773602342460651e-05}
{"train_loss": 0.41762638092041016, "global_step": 26553, "epoch": 298, "lr": 9.773585094724732e-05}
{"train_loss": 0.42221397161483765, "global_step": 26554, "epoch": 298, "lr": 9.773567846347065e-05}
{"train_loss": 0.4157310724258423, "global_step": 26555, "epoch": 298, "lr": 9.773550597327649e-05}
{"train_loss": 0.3905835747718811, "global_step": 26556, "epoch": 298, "lr": 9.773533347666489e-05}
{"train_loss": 0.3200647532939911, "global_step": 26557, "epoch": 298, "lr": 9.773516097363581e-05}
{"train_loss": 0.484456330537796, "global_step": 26558, "epoch": 298, "lr": 9.773498846418938e-05}
{"train_loss": 0.40088725090026855, "global_step": 26559, "epoch": 298, "lr": 9.773481594832552e-05}
{"train_loss": 0.39654016494750977, "global_step": 26560, "epoch": 298, "lr": 9.773464342604432e-05}
{"train_loss": 0.4153129756450653, "global_step": 26561, "epoch": 298, "lr": 9.773447089734577e-05}
{"train_loss": 0.4228965640068054, "global_step": 26562, "epoch": 298, "lr": 9.77342983622299e-05}
{"train_loss": 0.3721817433834076, "global_step": 26563, "epoch": 298, "lr": 9.773412582069673e-05}
{"train_loss": 0.42837247252464294, "global_step": 26564, "epoch": 298, "lr": 9.77339532727463e-05}
{"train_loss": 0.43769553303718567, "global_step": 26565, "epoch": 298, "lr": 9.773378071837861e-05}
{"train_loss": 0.5231039524078369, "global_step": 26566, "epoch": 298, "lr": 9.773360815759371e-05}
{"train_loss": 0.44935929775238037, "global_step": 26567, "epoch": 298, "lr": 9.773343559039159e-05}
{"train_loss": 0.3977832794189453, "global_step": 26568, "epoch": 298, "lr": 9.77332630167723e-05}
{"train_loss": 0.37502914667129517, "global_step": 26569, "epoch": 298, "lr": 9.773309043673587e-05}
{"train_loss": 0.37116849422454834, "global_step": 26570, "epoch": 298, "lr": 9.773291785028229e-05}
{"train_loss": 0.4027908146381378, "global_step": 26571, "epoch": 298, "lr": 9.77327452574116e-05}
{"train_loss": 0.4129761755466461, "global_step": 26572, "epoch": 298, "lr": 9.773257265812383e-05}
{"train_loss": 0.47895780205726624, "global_step": 26573, "epoch": 298, "lr": 9.7732400052419e-05}
{"train_loss": 0.42545875906944275, "global_step": 26574, "epoch": 298, "lr": 9.773222744029712e-05}
{"train_loss": 0.44908684492111206, "global_step": 26575, "epoch": 298, "lr": 9.773205482175824e-05}
{"train_loss": 0.47620922327041626, "global_step": 26576, "epoch": 298, "lr": 9.773188219680235e-05}
{"train_loss": 0.45361512899398804, "global_step": 26577, "epoch": 298, "lr": 9.77317095654295e-05}
{"train_loss": 0.3843407928943634, "global_step": 26578, "epoch": 298, "lr": 9.77315369276397e-05}
{"train_loss": 0.414114385843277, "global_step": 26579, "epoch": 298, "lr": 9.773136428343297e-05}
{"train_loss": 0.4259703457355499, "global_step": 26580, "epoch": 298, "lr": 9.773119163280936e-05}
{"train_loss": 0.36407285928726196, "global_step": 26581, "epoch": 298, "lr": 9.773101897576886e-05}
{"train_loss": 0.23608160018920898, "global_step": 26582, "epoch": 298, "lr": 9.77308463123115e-05}
{"train_loss": 0.388175368309021, "global_step": 26583, "epoch": 298, "lr": 9.773067364243732e-05}
{"train_loss": 0.4220401346683502, "global_step": 26584, "epoch": 298, "lr": 9.773050096614633e-05}
{"train_loss": 0.5120612978935242, "global_step": 26585, "epoch": 298, "lr": 9.773032828343856e-05}
{"train_loss": 0.4671214520931244, "global_step": 26586, "epoch": 298, "lr": 9.773015559431402e-05}
{"train_loss": 0.3068188428878784, "global_step": 26587, "epoch": 298, "lr": 9.772998289877275e-05}
{"train_loss": 0.3320944905281067, "global_step": 26588, "epoch": 298, "lr": 9.772981019681475e-05}
{"train_loss": 0.39337632060050964, "global_step": 26589, "epoch": 298, "lr": 9.772963748844009e-05}
{"train_loss": 0.33665597438812256, "global_step": 26590, "epoch": 298, "lr": 9.772946477364872e-05}
{"train_loss": 0.31404685974121094, "global_step": 26591, "epoch": 298, "lr": 9.772929205244074e-05}
{"train_loss": 0.5662580728530884, "global_step": 26592, "epoch": 298, "lr": 9.772911932481613e-05}
{"train_loss": 0.497839093208313, "global_step": 26593, "epoch": 298, "lr": 9.772894659077492e-05}
{"train_loss": 0.4165964424610138, "global_step": 26594, "epoch": 298, "lr": 9.772877385031713e-05}
{"train_loss": 0.3675209581851959, "global_step": 26595, "epoch": 298, "lr": 9.772860110344279e-05}
{"train_loss": 0.36432868242263794, "global_step": 26596, "epoch": 298, "lr": 9.772842835015192e-05}
{"train_loss": 0.34277012944221497, "global_step": 26597, "epoch": 298, "lr": 9.772825559044454e-05}
{"train_loss": 0.3335181176662445, "global_step": 26598, "epoch": 298, "lr": 9.772808282432068e-05}
{"train_loss": 0.3415084183216095, "global_step": 26599, "epoch": 298, "lr": 9.772791005178037e-05}
{"train_loss": 0.4781109094619751, "global_step": 26600, "epoch": 298, "lr": 9.772773727282362e-05}
{"train_loss": 0.5012126564979553, "global_step": 26601, "epoch": 298, "lr": 9.772756448745046e-05}
{"train_loss": 0.37547174096107483, "global_step": 26602, "epoch": 298, "lr": 9.772739169566091e-05}
{"train_loss": 0.41579821705818176, "global_step": 26603, "epoch": 298, "lr": 9.7727218897455e-05}
{"train_loss": 0.4143317937850952, "global_step": 26604, "epoch": 298, "lr": 9.772704609283274e-05}
{"train_loss": 0.4248109459877014, "global_step": 26605, "epoch": 298, "lr": 9.772687328179415e-05}
{"train_loss": 0.43637317419052124, "global_step": 26606, "epoch": 298, "lr": 9.772670046433927e-05}
{"train_loss": 0.46192124485969543, "global_step": 26607, "epoch": 298, "lr": 9.772652764046812e-05}
{"train_loss": 0.4416285455226898, "global_step": 26608, "epoch": 298, "lr": 9.772635481018073e-05}
{"train_loss": 0.5720647573471069, "global_step": 26609, "epoch": 298, "lr": 9.772618197347709e-05}
{"train_loss": 0.41388906387800584, "global_step": 26610, "epoch": 298, "lr": 9.772600913035727e-05, "val_loss": 2.138491630554199}
{"train_loss": 0.38168564438819885, "global_step": 26611, "epoch": 299, "lr": 9.772583628082126e-05}
{"train_loss": 0.37951526045799255, "global_step": 26612, "epoch": 299, "lr": 9.772566342486908e-05}
{"train_loss": 0.33550724387168884, "global_step": 26613, "epoch": 299, "lr": 9.772549056250078e-05}
{"train_loss": 0.4767085909843445, "global_step": 26614, "epoch": 299, "lr": 9.772531769371638e-05}
{"train_loss": 0.4969019293785095, "global_step": 26615, "epoch": 299, "lr": 9.772514481851588e-05}
{"train_loss": 0.38874319195747375, "global_step": 26616, "epoch": 299, "lr": 9.772497193689932e-05}
{"train_loss": 0.33940401673316956, "global_step": 26617, "epoch": 299, "lr": 9.772479904886671e-05}
{"train_loss": 0.44014471769332886, "global_step": 26618, "epoch": 299, "lr": 9.772462615441808e-05}
{"train_loss": 0.4079035818576813, "global_step": 26619, "epoch": 299, "lr": 9.772445325355347e-05}
{"train_loss": 0.399689644575119, "global_step": 26620, "epoch": 299, "lr": 9.772428034627289e-05}
{"train_loss": 0.3599954843521118, "global_step": 26621, "epoch": 299, "lr": 9.772410743257636e-05}
{"train_loss": 0.37520578503608704, "global_step": 26622, "epoch": 299, "lr": 9.772393451246389e-05}
{"train_loss": 0.49665746092796326, "global_step": 26623, "epoch": 299, "lr": 9.772376158593552e-05}
{"train_loss": 0.43045157194137573, "global_step": 26624, "epoch": 299, "lr": 9.77235886529913e-05}
{"train_loss": 0.31975457072257996, "global_step": 26625, "epoch": 299, "lr": 9.772341571363119e-05}
{"train_loss": 0.41315340995788574, "global_step": 26626, "epoch": 299, "lr": 9.772324276785527e-05}
{"train_loss": 0.3666275143623352, "global_step": 26627, "epoch": 299, "lr": 9.772306981566354e-05}
{"train_loss": 0.33604592084884644, "global_step": 26628, "epoch": 299, "lr": 9.772289685705603e-05}
{"train_loss": 0.40663138031959534, "global_step": 26629, "epoch": 299, "lr": 9.772272389203274e-05}
{"train_loss": 0.4688556492328644, "global_step": 26630, "epoch": 299, "lr": 9.772255092059373e-05}
{"train_loss": 0.41083022952079773, "global_step": 26631, "epoch": 299, "lr": 9.772237794273899e-05}
{"train_loss": 0.4274691045284271, "global_step": 26632, "epoch": 299, "lr": 9.772220495846857e-05}
{"train_loss": 0.41018059849739075, "global_step": 26633, "epoch": 299, "lr": 9.772203196778246e-05}
{"train_loss": 0.41753676533699036, "global_step": 26634, "epoch": 299, "lr": 9.772185897068074e-05}
{"train_loss": 0.3378497064113617, "global_step": 26635, "epoch": 299, "lr": 9.772168596716337e-05}
{"train_loss": 0.39990875124931335, "global_step": 26636, "epoch": 299, "lr": 9.77215129572304e-05}
{"train_loss": 0.4491540193557739, "global_step": 26637, "epoch": 299, "lr": 9.772133994088186e-05}
{"train_loss": 0.3605816066265106, "global_step": 26638, "epoch": 299, "lr": 9.772116691811778e-05}
{"train_loss": 0.43723076581954956, "global_step": 26639, "epoch": 299, "lr": 9.772099388893816e-05}
{"train_loss": 0.34268856048583984, "global_step": 26640, "epoch": 299, "lr": 9.772082085334303e-05}
{"train_loss": 0.5263320207595825, "global_step": 26641, "epoch": 299, "lr": 9.772064781133243e-05}
{"train_loss": 0.3688180446624756, "global_step": 26642, "epoch": 299, "lr": 9.772047476290635e-05}
{"train_loss": 0.38143888115882874, "global_step": 26643, "epoch": 299, "lr": 9.772030170806486e-05}
{"train_loss": 0.4495416283607483, "global_step": 26644, "epoch": 299, "lr": 9.772012864680793e-05}
{"train_loss": 0.42396509647369385, "global_step": 26645, "epoch": 299, "lr": 9.771995557913563e-05}
{"train_loss": 0.4263143837451935, "global_step": 26646, "epoch": 299, "lr": 9.771978250504796e-05}
{"train_loss": 0.5815008282661438, "global_step": 26647, "epoch": 299, "lr": 9.771960942454494e-05}
{"train_loss": 0.3372369706630707, "global_step": 26648, "epoch": 299, "lr": 9.771943633762661e-05}
{"train_loss": 0.45793575048446655, "global_step": 26649, "epoch": 299, "lr": 9.771926324429297e-05}
{"train_loss": 0.35672757029533386, "global_step": 26650, "epoch": 299, "lr": 9.771909014454407e-05}
{"train_loss": 0.4914495348930359, "global_step": 26651, "epoch": 299, "lr": 9.771891703837991e-05}
{"train_loss": 0.44199496507644653, "global_step": 26652, "epoch": 299, "lr": 9.771874392580053e-05}
{"train_loss": 0.42025071382522583, "global_step": 26653, "epoch": 299, "lr": 9.771857080680596e-05}
{"train_loss": 0.2898519039154053, "global_step": 26654, "epoch": 299, "lr": 9.771839768139618e-05}
{"train_loss": 0.4004574418067932, "global_step": 26655, "epoch": 299, "lr": 9.771822454957125e-05}
{"train_loss": 0.42463886737823486, "global_step": 26656, "epoch": 299, "lr": 9.771805141133121e-05}
{"train_loss": 0.3761885464191437, "global_step": 26657, "epoch": 299, "lr": 9.771787826667603e-05}
{"train_loss": 0.4125237464904785, "global_step": 26658, "epoch": 299, "lr": 9.771770511560578e-05}
{"train_loss": 0.39289650321006775, "global_step": 26659, "epoch": 299, "lr": 9.771753195812046e-05}
{"train_loss": 0.41728776693344116, "global_step": 26660, "epoch": 299, "lr": 9.771735879422011e-05}
{"train_loss": 0.40265873074531555, "global_step": 26661, "epoch": 299, "lr": 9.771718562390473e-05}
{"train_loss": 0.44225162267684937, "global_step": 26662, "epoch": 299, "lr": 9.771701244717435e-05}
{"train_loss": 0.5268028974533081, "global_step": 26663, "epoch": 299, "lr": 9.771683926402901e-05}
{"train_loss": 0.5638946294784546, "global_step": 26664, "epoch": 299, "lr": 9.771666607446872e-05}
{"train_loss": 0.6006777286529541, "global_step": 26665, "epoch": 299, "lr": 9.771649287849351e-05}
{"train_loss": 0.42945972084999084, "global_step": 26666, "epoch": 299, "lr": 9.771631967610339e-05}
{"train_loss": 0.37660154700279236, "global_step": 26667, "epoch": 299, "lr": 9.771614646729841e-05}
{"train_loss": 0.42826566100120544, "global_step": 26668, "epoch": 299, "lr": 9.771597325207856e-05}
{"train_loss": 0.36457493901252747, "global_step": 26669, "epoch": 299, "lr": 9.771580003044389e-05}
{"train_loss": 0.36189061403274536, "global_step": 26670, "epoch": 299, "lr": 9.77156268023944e-05}
{"train_loss": 0.374490350484848, "global_step": 26671, "epoch": 299, "lr": 9.771545356793013e-05}
{"train_loss": 0.45213931798934937, "global_step": 26672, "epoch": 299, "lr": 9.77152803270511e-05}
{"train_loss": 0.3550434112548828, "global_step": 26673, "epoch": 299, "lr": 9.771510707975733e-05}
{"train_loss": 0.4029948115348816, "global_step": 26674, "epoch": 299, "lr": 9.771493382604886e-05}
{"train_loss": 0.41097506880760193, "global_step": 26675, "epoch": 299, "lr": 9.771476056592568e-05}
{"train_loss": 0.4444236755371094, "global_step": 26676, "epoch": 299, "lr": 9.771458729938784e-05}
{"train_loss": 0.5268187522888184, "global_step": 26677, "epoch": 299, "lr": 9.771441402643535e-05}
{"train_loss": 0.4727432131767273, "global_step": 26678, "epoch": 299, "lr": 9.771424074706826e-05}
{"train_loss": 0.40310248732566833, "global_step": 26679, "epoch": 299, "lr": 9.771406746128654e-05}
{"train_loss": 0.40932685136795044, "global_step": 26680, "epoch": 299, "lr": 9.771389416909026e-05}
{"train_loss": 0.3001018166542053, "global_step": 26681, "epoch": 299, "lr": 9.771372087047943e-05}
{"train_loss": 0.5130492448806763, "global_step": 26682, "epoch": 299, "lr": 9.771354756545407e-05}
{"train_loss": 0.44667351245880127, "global_step": 26683, "epoch": 299, "lr": 9.77133742540142e-05}
{"train_loss": 0.4319687485694885, "global_step": 26684, "epoch": 299, "lr": 9.771320093615986e-05}
{"train_loss": 0.336013525724411, "global_step": 26685, "epoch": 299, "lr": 9.771302761189105e-05}
{"train_loss": 0.38802570104599, "global_step": 26686, "epoch": 299, "lr": 9.771285428120781e-05}
{"train_loss": 0.4537426829338074, "global_step": 26687, "epoch": 299, "lr": 9.771268094411016e-05}
{"train_loss": 0.5433271527290344, "global_step": 26688, "epoch": 299, "lr": 9.771250760059811e-05}
{"train_loss": 0.4266359508037567, "global_step": 26689, "epoch": 299, "lr": 9.771233425067171e-05}
{"train_loss": 0.39428162574768066, "global_step": 26690, "epoch": 299, "lr": 9.771216089433097e-05}
{"train_loss": 0.3452751636505127, "global_step": 26691, "epoch": 299, "lr": 9.77119875315759e-05}
{"train_loss": 0.34835493564605713, "global_step": 26692, "epoch": 299, "lr": 9.771181416240653e-05}
{"train_loss": 0.4684275686740875, "global_step": 26693, "epoch": 299, "lr": 9.771164078682291e-05}
{"train_loss": 0.4703625738620758, "global_step": 26694, "epoch": 299, "lr": 9.771146740482502e-05}
{"train_loss": 0.4114498496055603, "global_step": 26695, "epoch": 299, "lr": 9.771129401641293e-05}
{"train_loss": 0.436646968126297, "global_step": 26696, "epoch": 299, "lr": 9.77111206215866e-05}
{"train_loss": 0.37082141637802124, "global_step": 26697, "epoch": 299, "lr": 9.771094722034613e-05}
{"train_loss": 0.40073564648628235, "global_step": 26698, "epoch": 299, "lr": 9.771077381269148e-05}
{"train_loss": 0.41756194260682955, "global_step": 26699, "epoch": 299, "lr": 9.77106003986227e-05, "val_loss": 2.0598158836364746}
{"train_loss": 0.5645315647125244, "global_step": 26700, "epoch": 300, "lr": 9.771042697813983e-05}
{"train_loss": 0.3119015395641327, "global_step": 26701, "epoch": 300, "lr": 9.771025355124285e-05}
{"train_loss": 0.28065332770347595, "global_step": 26702, "epoch": 300, "lr": 9.771008011793182e-05}
{"train_loss": 0.43571001291275024, "global_step": 26703, "epoch": 300, "lr": 9.770990667820675e-05}
{"train_loss": 0.4061189889907837, "global_step": 26704, "epoch": 300, "lr": 9.770973323206767e-05}
{"train_loss": 0.4689466953277588, "global_step": 26705, "epoch": 300, "lr": 9.77095597795146e-05}
{"train_loss": 0.34292662143707275, "global_step": 26706, "epoch": 300, "lr": 9.770938632054754e-05}
{"train_loss": 0.29561230540275574, "global_step": 26707, "epoch": 300, "lr": 9.770921285516655e-05}
{"train_loss": 0.36829042434692383, "global_step": 26708, "epoch": 300, "lr": 9.770903938337164e-05}
{"train_loss": 0.41622287034988403, "global_step": 26709, "epoch": 300, "lr": 9.770886590516283e-05}
{"train_loss": 0.3105911314487457, "global_step": 26710, "epoch": 300, "lr": 9.770869242054014e-05}
{"train_loss": 0.33216291666030884, "global_step": 26711, "epoch": 300, "lr": 9.77085189295036e-05}
{"train_loss": 0.39558693766593933, "global_step": 26712, "epoch": 300, "lr": 9.770834543205322e-05}
{"train_loss": 0.5105697512626648, "global_step": 26713, "epoch": 300, "lr": 9.770817192818906e-05}
{"train_loss": 0.3396589159965515, "global_step": 26714, "epoch": 300, "lr": 9.77079984179111e-05}
{"train_loss": 0.3628469705581665, "global_step": 26715, "epoch": 300, "lr": 9.770782490121939e-05}
{"train_loss": 0.4275337755680084, "global_step": 26716, "epoch": 300, "lr": 9.770765137811393e-05}
{"train_loss": 0.48019981384277344, "global_step": 26717, "epoch": 300, "lr": 9.770747784859478e-05}
{"train_loss": 0.3756391406059265, "global_step": 26718, "epoch": 300, "lr": 9.770730431266192e-05}
{"train_loss": 0.3329029083251953, "global_step": 26719, "epoch": 300, "lr": 9.77071307703154e-05}
{"train_loss": 0.3502005934715271, "global_step": 26720, "epoch": 300, "lr": 9.770695722155522e-05}
{"train_loss": 0.5188624858856201, "global_step": 26721, "epoch": 300, "lr": 9.770678366638146e-05}
{"train_loss": 0.3979763388633728, "global_step": 26722, "epoch": 300, "lr": 9.770661010479408e-05}
{"train_loss": 0.4066007435321808, "global_step": 26723, "epoch": 300, "lr": 9.770643653679314e-05}
{"train_loss": 0.4868547320365906, "global_step": 26724, "epoch": 300, "lr": 9.770626296237864e-05}
{"train_loss": 0.3853873014450073, "global_step": 26725, "epoch": 300, "lr": 9.770608938155061e-05}
{"train_loss": 0.41220715641975403, "global_step": 26726, "epoch": 300, "lr": 9.77059157943091e-05}
{"train_loss": 0.3575386703014374, "global_step": 26727, "epoch": 300, "lr": 9.77057422006541e-05}
{"train_loss": 0.4628522992134094, "global_step": 26728, "epoch": 300, "lr": 9.770556860058564e-05}
{"train_loss": 0.36289912462234497, "global_step": 26729, "epoch": 300, "lr": 9.770539499410374e-05}
{"train_loss": 0.36233970522880554, "global_step": 26730, "epoch": 300, "lr": 9.770522138120844e-05}
{"train_loss": 0.38054177165031433, "global_step": 26731, "epoch": 300, "lr": 9.770504776189976e-05}
{"train_loss": 0.4157250225543976, "global_step": 26732, "epoch": 300, "lr": 9.77048741361777e-05}
{"train_loss": 0.4010571241378784, "global_step": 26733, "epoch": 300, "lr": 9.770470050404232e-05}
{"train_loss": 0.3115224540233612, "global_step": 26734, "epoch": 300, "lr": 9.770452686549363e-05}
{"train_loss": 0.47961971163749695, "global_step": 26735, "epoch": 300, "lr": 9.770435322053163e-05}
{"train_loss": 0.44847238063812256, "global_step": 26736, "epoch": 300, "lr": 9.770417956915636e-05}
{"train_loss": 0.3309290409088135, "global_step": 26737, "epoch": 300, "lr": 9.770400591136786e-05}
{"train_loss": 0.551238477230072, "global_step": 26738, "epoch": 300, "lr": 9.770383224716613e-05}
{"train_loss": 0.3991621136665344, "global_step": 26739, "epoch": 300, "lr": 9.77036585765512e-05}
{"train_loss": 0.33763882517814636, "global_step": 26740, "epoch": 300, "lr": 9.770348489952309e-05}
{"train_loss": 0.43035784363746643, "global_step": 26741, "epoch": 300, "lr": 9.770331121608183e-05}
{"train_loss": 0.47311708331108093, "global_step": 26742, "epoch": 300, "lr": 9.770313752622744e-05}
{"train_loss": 0.3801332116127014, "global_step": 26743, "epoch": 300, "lr": 9.770296382995996e-05}
{"train_loss": 0.3731405436992645, "global_step": 26744, "epoch": 300, "lr": 9.770279012727938e-05}
{"train_loss": 0.41333651542663574, "global_step": 26745, "epoch": 300, "lr": 9.770261641818575e-05}
{"train_loss": 0.3828529715538025, "global_step": 26746, "epoch": 300, "lr": 9.770244270267907e-05}
{"train_loss": 0.4310392737388611, "global_step": 26747, "epoch": 300, "lr": 9.77022689807594e-05}
{"train_loss": 0.39019834995269775, "global_step": 26748, "epoch": 300, "lr": 9.770209525242672e-05}
{"train_loss": 0.3911890685558319, "global_step": 26749, "epoch": 300, "lr": 9.77019215176811e-05}
{"train_loss": 0.4418601393699646, "global_step": 26750, "epoch": 300, "lr": 9.770174777652253e-05}
{"train_loss": 0.48695918917655945, "global_step": 26751, "epoch": 300, "lr": 9.770157402895103e-05}
{"train_loss": 0.463533490896225, "global_step": 26752, "epoch": 300, "lr": 9.770140027496663e-05}
{"train_loss": 0.38293787837028503, "global_step": 26753, "epoch": 300, "lr": 9.77012265145694e-05}
{"train_loss": 0.45922672748565674, "global_step": 26754, "epoch": 300, "lr": 9.770105274775926e-05}
{"train_loss": 0.36920294165611267, "global_step": 26755, "epoch": 300, "lr": 9.770087897453635e-05}
{"train_loss": 0.3729662001132965, "global_step": 26756, "epoch": 300, "lr": 9.77007051949006e-05}
{"train_loss": 0.4761216640472412, "global_step": 26757, "epoch": 300, "lr": 9.770053140885209e-05}
{"train_loss": 0.46766921877861023, "global_step": 26758, "epoch": 300, "lr": 9.770035761639082e-05}
{"train_loss": 0.37126490473747253, "global_step": 26759, "epoch": 300, "lr": 9.770018381751682e-05}
{"train_loss": 0.38763073086738586, "global_step": 26760, "epoch": 300, "lr": 9.77000100122301e-05}
{"train_loss": 0.2655844986438751, "global_step": 26761, "epoch": 300, "lr": 9.769983620053071e-05}
{"train_loss": 0.5617139935493469, "global_step": 26762, "epoch": 300, "lr": 9.769966238241865e-05}
{"train_loss": 0.4703655540943146, "global_step": 26763, "epoch": 300, "lr": 9.769948855789396e-05}
{"train_loss": 0.5179578065872192, "global_step": 26764, "epoch": 300, "lr": 9.769931472695666e-05}
{"train_loss": 0.4100029170513153, "global_step": 26765, "epoch": 300, "lr": 9.769914088960674e-05}
{"train_loss": 0.3699988126754761, "global_step": 26766, "epoch": 300, "lr": 9.769896704584428e-05}
{"train_loss": 0.3983989655971527, "global_step": 26767, "epoch": 300, "lr": 9.769879319566927e-05}
{"train_loss": 0.4771551787853241, "global_step": 26768, "epoch": 300, "lr": 9.769861933908173e-05}
{"train_loss": 0.3044286370277405, "global_step": 26769, "epoch": 300, "lr": 9.76984454760817e-05}
{"train_loss": 0.4309634268283844, "global_step": 26770, "epoch": 300, "lr": 9.769827160666919e-05}
{"train_loss": 0.4474656879901886, "global_step": 26771, "epoch": 300, "lr": 9.769809773084424e-05}
{"train_loss": 0.45033320784568787, "global_step": 26772, "epoch": 300, "lr": 9.769792384860684e-05}
{"train_loss": 0.3529489040374756, "global_step": 26773, "epoch": 300, "lr": 9.769774995995705e-05}
{"train_loss": 0.4882122576236725, "global_step": 26774, "epoch": 300, "lr": 9.769757606489488e-05}
{"train_loss": 0.4094775915145874, "global_step": 26775, "epoch": 300, "lr": 9.769740216342035e-05}
{"train_loss": 0.39961880445480347, "global_step": 26776, "epoch": 300, "lr": 9.769722825553349e-05}
{"train_loss": 0.5911990404129028, "global_step": 26777, "epoch": 300, "lr": 9.769705434123431e-05}
{"train_loss": 0.3659823536872864, "global_step": 26778, "epoch": 300, "lr": 9.769688042052286e-05}
{"train_loss": 0.3287905156612396, "global_step": 26779, "epoch": 300, "lr": 9.769670649339911e-05}
{"train_loss": 0.42001956701278687, "global_step": 26780, "epoch": 300, "lr": 9.769653255986315e-05}
{"train_loss": 0.4977172911167145, "global_step": 26781, "epoch": 300, "lr": 9.769635861991497e-05}
{"train_loss": 0.3485957980155945, "global_step": 26782, "epoch": 300, "lr": 9.769618467355459e-05}
{"train_loss": 0.36563003063201904, "global_step": 26783, "epoch": 300, "lr": 9.769601072078203e-05}
{"train_loss": 0.428500235080719, "global_step": 26784, "epoch": 300, "lr": 9.769583676159733e-05}
{"train_loss": 0.4680855870246887, "global_step": 26785, "epoch": 300, "lr": 9.76956627960005e-05}
{"train_loss": 0.41426074504852295, "global_step": 26786, "epoch": 300, "lr": 9.769548882399158e-05}
{"train_loss": 0.5074974894523621, "global_step": 26787, "epoch": 300, "lr": 9.76953148455706e-05}
{"train_loss": 0.4129692476786924, "global_step": 26788, "epoch": 300, "lr": 9.769514086073755e-05, "train/sim_max_reward_0": 0.9934092013750339, "train/sim_max_reward_1": 0.8266165117080169, "train/sim_max_reward_2": 1.0, "train/sim_max_reward_3": 0.5310999648974152, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.6581318034020569, "test/sim_max_reward_4500000": 0.6016672855290618, "test/sim_max_reward_4500001": 0.4681949649601067, "test/sim_max_reward_4500002": 0.9848855479110017, "test/sim_max_reward_4500003": 0.5582283436188725, "test/sim_max_reward_4500004": 0.8943990206912458, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.8979649765633014, "test/sim_max_reward_4500007": 0.9810286145179549, "test/sim_max_reward_4500008": 0.3008344784080148, "test/sim_max_reward_4500009": 0.2034894971183402, "test/sim_max_reward_4500010": 0.9060286162326261, "test/sim_max_reward_4500011": 0.9134186448341372, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.46740650565247993, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.704292574319787, "test/sim_max_reward_4500017": 0.6123424970700753, "test/sim_max_reward_4500018": 0.8448754530079264, "test/sim_max_reward_4500019": 0.9980406014335184, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9727148588642698, "test/sim_max_reward_4500022": 0.983396397104602, "test/sim_max_reward_4500023": 0.6740020096671223, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.6394203846315434, "test/sim_max_reward_4500026": 0.8967884311242137, "test/sim_max_reward_4500027": 0.9714786294399702, "test/sim_max_reward_4500028": 0.7235192079473869, "test/sim_max_reward_4500029": 0.5211613432229234, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6488139565727159, "test/sim_max_reward_4500033": 0.552984734582279, "test/sim_max_reward_4500034": 0.96565162359927, "test/sim_max_reward_4500035": 0.6343260065221275, "test/sim_max_reward_4500036": 0.9848640873869527, "test/sim_max_reward_4500037": 0.9221574138395607, "test/sim_max_reward_4500038": 0.5764241235479598, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9998187956618202, "test/sim_max_reward_4500041": 0.5401955476295908, "test/sim_max_reward_4500042": 0.7805701940076737, "test/sim_max_reward_4500043": 0.10332215730807642, "test/sim_max_reward_4500044": 0.5577797850396037, "test/sim_max_reward_4500045": 0.5587862169687708, "test/sim_max_reward_4500046": 0.8644975642641571, "test/sim_max_reward_4500047": 0.4702931776066185, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.5903725250928445, "train/mean_score": 0.8348762468970872, "test/mean_score": 0.6894087358700101, "val_loss": 2.0922908782958984, "train_action_mse_error": 26.63312530517578}
{"train_loss": 0.28393858671188354, "global_step": 26789, "epoch": 301, "lr": 9.769496686949246e-05}
{"train_loss": 0.4683226943016052, "global_step": 26790, "epoch": 301, "lr": 9.769479287183536e-05}
{"train_loss": 0.3849489092826843, "global_step": 26791, "epoch": 301, "lr": 9.769461886776628e-05}
{"train_loss": 0.42789214849472046, "global_step": 26792, "epoch": 301, "lr": 9.769444485728525e-05}
{"train_loss": 0.3772917687892914, "global_step": 26793, "epoch": 301, "lr": 9.769427084039229e-05}
{"train_loss": 0.4199521243572235, "global_step": 26794, "epoch": 301, "lr": 9.76940968170874e-05}
{"train_loss": 0.36375629901885986, "global_step": 26795, "epoch": 301, "lr": 9.769392278737062e-05}
{"train_loss": 0.4093077778816223, "global_step": 26796, "epoch": 301, "lr": 9.769374875124197e-05}
{"train_loss": 0.3657044470310211, "global_step": 26797, "epoch": 301, "lr": 9.76935747087015e-05}
{"train_loss": 0.4629722237586975, "global_step": 26798, "epoch": 301, "lr": 9.76934006597492e-05}
{"train_loss": 0.4375551640987396, "global_step": 26799, "epoch": 301, "lr": 9.769322660438509e-05}
{"train_loss": 0.41202983260154724, "global_step": 26800, "epoch": 301, "lr": 9.769305254260921e-05}
{"train_loss": 0.43639516830444336, "global_step": 26801, "epoch": 301, "lr": 9.76928784744216e-05}
{"train_loss": 0.3580076992511749, "global_step": 26802, "epoch": 301, "lr": 9.769270439982224e-05}
{"train_loss": 0.41318583488464355, "global_step": 26803, "epoch": 301, "lr": 9.769253031881119e-05}
{"train_loss": 0.42664846777915955, "global_step": 26804, "epoch": 301, "lr": 9.769235623138846e-05}
{"train_loss": 0.3695289194583893, "global_step": 26805, "epoch": 301, "lr": 9.769218213755406e-05}
{"train_loss": 0.34397637844085693, "global_step": 26806, "epoch": 301, "lr": 9.769200803730804e-05}
{"train_loss": 0.42846348881721497, "global_step": 26807, "epoch": 301, "lr": 9.769183393065043e-05}
{"train_loss": 0.3156343102455139, "global_step": 26808, "epoch": 301, "lr": 9.769165981758121e-05}
{"train_loss": 0.5539233088493347, "global_step": 26809, "epoch": 301, "lr": 9.769148569810043e-05}
{"train_loss": 0.4043180048465729, "global_step": 26810, "epoch": 301, "lr": 9.769131157220811e-05}
{"train_loss": 0.4260897934436798, "global_step": 26811, "epoch": 301, "lr": 9.769113743990428e-05}
{"train_loss": 0.33750438690185547, "global_step": 26812, "epoch": 301, "lr": 9.769096330118895e-05}
{"train_loss": 0.2900843024253845, "global_step": 26813, "epoch": 301, "lr": 9.769078915606216e-05}
{"train_loss": 0.38368916511535645, "global_step": 26814, "epoch": 301, "lr": 9.769061500452392e-05}
{"train_loss": 0.43216046690940857, "global_step": 26815, "epoch": 301, "lr": 9.769044084657426e-05}
{"train_loss": 0.465658038854599, "global_step": 26816, "epoch": 301, "lr": 9.76902666822132e-05}
{"train_loss": 0.3360023498535156, "global_step": 26817, "epoch": 301, "lr": 9.769009251144076e-05}
{"train_loss": 0.2667962908744812, "global_step": 26818, "epoch": 301, "lr": 9.768991833425698e-05}
{"train_loss": 0.30171477794647217, "global_step": 26819, "epoch": 301, "lr": 9.768974415066186e-05}
{"train_loss": 0.34849029779434204, "global_step": 26820, "epoch": 301, "lr": 9.768956996065545e-05}
{"train_loss": 0.4260684549808502, "global_step": 26821, "epoch": 301, "lr": 9.768939576423774e-05}
{"train_loss": 0.48198360204696655, "global_step": 26822, "epoch": 301, "lr": 9.768922156140878e-05}
{"train_loss": 0.49150821566581726, "global_step": 26823, "epoch": 301, "lr": 9.768904735216859e-05}
{"train_loss": 0.27747976779937744, "global_step": 26824, "epoch": 301, "lr": 9.76888731365172e-05}
{"train_loss": 0.34738168120384216, "global_step": 26825, "epoch": 301, "lr": 9.768869891445461e-05}
{"train_loss": 0.3545061945915222, "global_step": 26826, "epoch": 301, "lr": 9.768852468598084e-05}
{"train_loss": 0.5493384599685669, "global_step": 26827, "epoch": 301, "lr": 9.768835045109595e-05}
{"train_loss": 0.3882116675376892, "global_step": 26828, "epoch": 301, "lr": 9.768817620979993e-05}
{"train_loss": 0.4261663854122162, "global_step": 26829, "epoch": 301, "lr": 9.768800196209282e-05}
{"train_loss": 0.43246588110923767, "global_step": 26830, "epoch": 301, "lr": 9.768782770797464e-05}
{"train_loss": 0.5645802617073059, "global_step": 26831, "epoch": 301, "lr": 9.768765344744542e-05}
{"train_loss": 0.4753202199935913, "global_step": 26832, "epoch": 301, "lr": 9.768747918050517e-05}
{"train_loss": 0.3925527036190033, "global_step": 26833, "epoch": 301, "lr": 9.768730490715392e-05}
{"train_loss": 0.397058367729187, "global_step": 26834, "epoch": 301, "lr": 9.76871306273917e-05}
{"train_loss": 0.48403361439704895, "global_step": 26835, "epoch": 301, "lr": 9.768695634121851e-05}
{"train_loss": 0.3352348208427429, "global_step": 26836, "epoch": 301, "lr": 9.76867820486344e-05}
{"train_loss": 0.3121546804904938, "global_step": 26837, "epoch": 301, "lr": 9.768660774963938e-05}
{"train_loss": 0.3486032783985138, "global_step": 26838, "epoch": 301, "lr": 9.768643344423348e-05}
{"train_loss": 0.49435704946517944, "global_step": 26839, "epoch": 301, "lr": 9.768625913241671e-05}
{"train_loss": 0.48477262258529663, "global_step": 26840, "epoch": 301, "lr": 9.768608481418913e-05}
{"train_loss": 0.45701703429222107, "global_step": 26841, "epoch": 301, "lr": 9.768591048955071e-05}
{"train_loss": 0.3865254819393158, "global_step": 26842, "epoch": 301, "lr": 9.768573615850152e-05}
{"train_loss": 0.42218437790870667, "global_step": 26843, "epoch": 301, "lr": 9.768556182104155e-05}
{"train_loss": 0.39652955532073975, "global_step": 26844, "epoch": 301, "lr": 9.768538747717084e-05}
{"train_loss": 0.3744984567165375, "global_step": 26845, "epoch": 301, "lr": 9.768521312688941e-05}
{"train_loss": 0.3927711844444275, "global_step": 26846, "epoch": 301, "lr": 9.76850387701973e-05}
{"train_loss": 0.5160699486732483, "global_step": 26847, "epoch": 301, "lr": 9.76848644070945e-05}
{"train_loss": 0.41458040475845337, "global_step": 26848, "epoch": 301, "lr": 9.768469003758105e-05}
{"train_loss": 0.40472114086151123, "global_step": 26849, "epoch": 301, "lr": 9.768451566165698e-05}
{"train_loss": 0.3815310597419739, "global_step": 26850, "epoch": 301, "lr": 9.768434127932232e-05}
{"train_loss": 0.5340253114700317, "global_step": 26851, "epoch": 301, "lr": 9.768416689057706e-05}
{"train_loss": 0.3201480507850647, "global_step": 26852, "epoch": 301, "lr": 9.768399249542126e-05}
{"train_loss": 0.3497997224330902, "global_step": 26853, "epoch": 301, "lr": 9.768381809385491e-05}
{"train_loss": 0.33317476511001587, "global_step": 26854, "epoch": 301, "lr": 9.768364368587808e-05}
{"train_loss": 0.3783855736255646, "global_step": 26855, "epoch": 301, "lr": 9.768346927149075e-05}
{"train_loss": 0.5269088745117188, "global_step": 26856, "epoch": 301, "lr": 9.768329485069297e-05}
{"train_loss": 0.48230239748954773, "global_step": 26857, "epoch": 301, "lr": 9.768312042348473e-05}
{"train_loss": 0.34972912073135376, "global_step": 26858, "epoch": 301, "lr": 9.768294598986609e-05}
{"train_loss": 0.4272790849208832, "global_step": 26859, "epoch": 301, "lr": 9.768277154983704e-05}
{"train_loss": 0.3930392563343048, "global_step": 26860, "epoch": 301, "lr": 9.768259710339764e-05}
{"train_loss": 0.3257506787776947, "global_step": 26861, "epoch": 301, "lr": 9.76824226505479e-05}
{"train_loss": 0.5085749626159668, "global_step": 26862, "epoch": 301, "lr": 9.768224819128784e-05}
{"train_loss": 0.4341771900653839, "global_step": 26863, "epoch": 301, "lr": 9.768207372561748e-05}
{"train_loss": 0.4635539948940277, "global_step": 26864, "epoch": 301, "lr": 9.768189925353683e-05}
{"train_loss": 0.40685248374938965, "global_step": 26865, "epoch": 301, "lr": 9.768172477504596e-05}
{"train_loss": 0.3960443139076233, "global_step": 26866, "epoch": 301, "lr": 9.768155029014485e-05}
{"train_loss": 0.40454933047294617, "global_step": 26867, "epoch": 301, "lr": 9.768137579883351e-05}
{"train_loss": 0.5051337480545044, "global_step": 26868, "epoch": 301, "lr": 9.768120130111203e-05}
{"train_loss": 0.40086638927459717, "global_step": 26869, "epoch": 301, "lr": 9.768102679698038e-05}
{"train_loss": 0.39664098620414734, "global_step": 26870, "epoch": 301, "lr": 9.76808522864386e-05}
{"train_loss": 0.35557499527931213, "global_step": 26871, "epoch": 301, "lr": 9.768067776948669e-05}
{"train_loss": 0.3735056221485138, "global_step": 26872, "epoch": 301, "lr": 9.768050324612472e-05}
{"train_loss": 0.32877448201179504, "global_step": 26873, "epoch": 301, "lr": 9.768032871635267e-05}
{"train_loss": 0.3581688106060028, "global_step": 26874, "epoch": 301, "lr": 9.76801541801706e-05}
{"train_loss": 0.46343734860420227, "global_step": 26875, "epoch": 301, "lr": 9.76799796375785e-05}
{"train_loss": 0.41786688566207886, "global_step": 26876, "epoch": 301, "lr": 9.76798050885764e-05}
{"train_loss": 0.4058346396751618, "global_step": 26877, "epoch": 301, "lr": 9.767963053316433e-05, "val_loss": 2.0790913105010986}
{"train_loss": 0.34595271944999695, "global_step": 26878, "epoch": 302, "lr": 9.767945597134233e-05}
{"train_loss": 0.3304516077041626, "global_step": 26879, "epoch": 302, "lr": 9.76792814031104e-05}
{"train_loss": 0.3600480854511261, "global_step": 26880, "epoch": 302, "lr": 9.76791068284686e-05}
{"train_loss": 0.31864938139915466, "global_step": 26881, "epoch": 302, "lr": 9.767893224741689e-05}
{"train_loss": 0.38101261854171753, "global_step": 26882, "epoch": 302, "lr": 9.767875765995533e-05}
{"train_loss": 0.4743853211402893, "global_step": 26883, "epoch": 302, "lr": 9.767858306608396e-05}
{"train_loss": 0.46549755334854126, "global_step": 26884, "epoch": 302, "lr": 9.767840846580279e-05}
{"train_loss": 0.4327266216278076, "global_step": 26885, "epoch": 302, "lr": 9.767823385911181e-05}
{"train_loss": 0.4349670112133026, "global_step": 26886, "epoch": 302, "lr": 9.76780592460111e-05}
{"train_loss": 0.4399944543838501, "global_step": 26887, "epoch": 302, "lr": 9.767788462650064e-05}
{"train_loss": 0.31094956398010254, "global_step": 26888, "epoch": 302, "lr": 9.767771000058048e-05}
{"train_loss": 0.3415546417236328, "global_step": 26889, "epoch": 302, "lr": 9.767753536825062e-05}
{"train_loss": 0.3685312271118164, "global_step": 26890, "epoch": 302, "lr": 9.76773607295111e-05}
{"train_loss": 0.3912539780139923, "global_step": 26891, "epoch": 302, "lr": 9.767718608436194e-05}
{"train_loss": 0.29065200686454773, "global_step": 26892, "epoch": 302, "lr": 9.767701143280318e-05}
{"train_loss": 0.4790590703487396, "global_step": 26893, "epoch": 302, "lr": 9.767683677483481e-05}
{"train_loss": 0.32975995540618896, "global_step": 26894, "epoch": 302, "lr": 9.767666211045688e-05}
{"train_loss": 0.42144685983657837, "global_step": 26895, "epoch": 302, "lr": 9.767648743966939e-05}
{"train_loss": 0.2957102954387665, "global_step": 26896, "epoch": 302, "lr": 9.76763127624724e-05}
{"train_loss": 0.4437006711959839, "global_step": 26897, "epoch": 302, "lr": 9.767613807886589e-05}
{"train_loss": 0.39460521936416626, "global_step": 26898, "epoch": 302, "lr": 9.767596338884991e-05}
{"train_loss": 0.44019874930381775, "global_step": 26899, "epoch": 302, "lr": 9.767578869242448e-05}
{"train_loss": 0.4469718337059021, "global_step": 26900, "epoch": 302, "lr": 9.767561398958961e-05}
{"train_loss": 0.4345080256462097, "global_step": 26901, "epoch": 302, "lr": 9.767543928034535e-05}
{"train_loss": 0.4687226116657257, "global_step": 26902, "epoch": 302, "lr": 9.767526456469172e-05}
{"train_loss": 0.3739895224571228, "global_step": 26903, "epoch": 302, "lr": 9.76750898426287e-05}
{"train_loss": 0.3535720705986023, "global_step": 26904, "epoch": 302, "lr": 9.767491511415637e-05}
{"train_loss": 0.3572559058666229, "global_step": 26905, "epoch": 302, "lr": 9.767474037927471e-05}
{"train_loss": 0.4794090986251831, "global_step": 26906, "epoch": 302, "lr": 9.767456563798377e-05}
{"train_loss": 0.40660738945007324, "global_step": 26907, "epoch": 302, "lr": 9.767439089028357e-05}
{"train_loss": 0.37919777631759644, "global_step": 26908, "epoch": 302, "lr": 9.767421613617413e-05}
{"train_loss": 0.45144909620285034, "global_step": 26909, "epoch": 302, "lr": 9.767404137565547e-05}
{"train_loss": 0.3210870027542114, "global_step": 26910, "epoch": 302, "lr": 9.767386660872762e-05}
{"train_loss": 0.3978288471698761, "global_step": 26911, "epoch": 302, "lr": 9.76736918353906e-05}
{"train_loss": 0.4085499048233032, "global_step": 26912, "epoch": 302, "lr": 9.767351705564442e-05}
{"train_loss": 0.3670460879802704, "global_step": 26913, "epoch": 302, "lr": 9.767334226948912e-05}
{"train_loss": 0.35319435596466064, "global_step": 26914, "epoch": 302, "lr": 9.767316747692471e-05}
{"train_loss": 0.499990314245224, "global_step": 26915, "epoch": 302, "lr": 9.767299267795124e-05}
{"train_loss": 0.34734106063842773, "global_step": 26916, "epoch": 302, "lr": 9.767281787256874e-05}
{"train_loss": 0.35180145502090454, "global_step": 26917, "epoch": 302, "lr": 9.767264306077718e-05}
{"train_loss": 0.31811872124671936, "global_step": 26918, "epoch": 302, "lr": 9.767246824257663e-05}
{"train_loss": 0.4744679033756256, "global_step": 26919, "epoch": 302, "lr": 9.767229341796707e-05}
{"train_loss": 0.37823933362960815, "global_step": 26920, "epoch": 302, "lr": 9.767211858694858e-05}
{"train_loss": 0.3985895812511444, "global_step": 26921, "epoch": 302, "lr": 9.767194374952114e-05}
{"train_loss": 0.4801684021949768, "global_step": 26922, "epoch": 302, "lr": 9.767176890568479e-05}
{"train_loss": 0.4509361982345581, "global_step": 26923, "epoch": 302, "lr": 9.767159405543956e-05}
{"train_loss": 0.4166882038116455, "global_step": 26924, "epoch": 302, "lr": 9.767141919878546e-05}
{"train_loss": 0.42323875427246094, "global_step": 26925, "epoch": 302, "lr": 9.767124433572252e-05}
{"train_loss": 0.4179862141609192, "global_step": 26926, "epoch": 302, "lr": 9.767106946625075e-05}
{"train_loss": 0.4681728482246399, "global_step": 26927, "epoch": 302, "lr": 9.76708945903702e-05}
{"train_loss": 0.45876413583755493, "global_step": 26928, "epoch": 302, "lr": 9.767071970808088e-05}
{"train_loss": 0.5234781503677368, "global_step": 26929, "epoch": 302, "lr": 9.76705448193828e-05}
{"train_loss": 0.428164541721344, "global_step": 26930, "epoch": 302, "lr": 9.7670369924276e-05}
{"train_loss": 0.470209002494812, "global_step": 26931, "epoch": 302, "lr": 9.767019502276052e-05}
{"train_loss": 0.33967018127441406, "global_step": 26932, "epoch": 302, "lr": 9.767002011483634e-05}
{"train_loss": 0.38147518038749695, "global_step": 26933, "epoch": 302, "lr": 9.766984520050351e-05}
{"train_loss": 0.39147573709487915, "global_step": 26934, "epoch": 302, "lr": 9.766967027976206e-05}
{"train_loss": 0.3503410220146179, "global_step": 26935, "epoch": 302, "lr": 9.7669495352612e-05}
{"train_loss": 0.38188615441322327, "global_step": 26936, "epoch": 302, "lr": 9.766932041905335e-05}
{"train_loss": 0.48650485277175903, "global_step": 26937, "epoch": 302, "lr": 9.766914547908615e-05}
{"train_loss": 0.3028748631477356, "global_step": 26938, "epoch": 302, "lr": 9.766897053271042e-05}
{"train_loss": 0.45145121216773987, "global_step": 26939, "epoch": 302, "lr": 9.766879557992616e-05}
{"train_loss": 0.45713722705841064, "global_step": 26940, "epoch": 302, "lr": 9.766862062073343e-05}
{"train_loss": 0.31689751148223877, "global_step": 26941, "epoch": 302, "lr": 9.766844565513222e-05}
{"train_loss": 0.5062330365180969, "global_step": 26942, "epoch": 302, "lr": 9.766827068312258e-05}
{"train_loss": 0.47174566984176636, "global_step": 26943, "epoch": 302, "lr": 9.766809570470451e-05}
{"train_loss": 0.3962016701698303, "global_step": 26944, "epoch": 302, "lr": 9.766792071987805e-05}
{"train_loss": 0.35425013303756714, "global_step": 26945, "epoch": 302, "lr": 9.766774572864322e-05}
{"train_loss": 0.47902530431747437, "global_step": 26946, "epoch": 302, "lr": 9.766757073100004e-05}
{"train_loss": 0.39935600757598877, "global_step": 26947, "epoch": 302, "lr": 9.766739572694854e-05}
{"train_loss": 0.31321555376052856, "global_step": 26948, "epoch": 302, "lr": 9.766722071648874e-05}
{"train_loss": 0.387258917093277, "global_step": 26949, "epoch": 302, "lr": 9.766704569962067e-05}
{"train_loss": 0.35441792011260986, "global_step": 26950, "epoch": 302, "lr": 9.766687067634433e-05}
{"train_loss": 0.4328966736793518, "global_step": 26951, "epoch": 302, "lr": 9.766669564665978e-05}
{"train_loss": 0.374385803937912, "global_step": 26952, "epoch": 302, "lr": 9.766652061056701e-05}
{"train_loss": 0.3979761302471161, "global_step": 26953, "epoch": 302, "lr": 9.766634556806606e-05}
{"train_loss": 0.42586469650268555, "global_step": 26954, "epoch": 302, "lr": 9.766617051915694e-05}
{"train_loss": 0.38440611958503723, "global_step": 26955, "epoch": 302, "lr": 9.76659954638397e-05}
{"train_loss": 0.4896616041660309, "global_step": 26956, "epoch": 302, "lr": 9.766582040211433e-05}
{"train_loss": 0.36993327736854553, "global_step": 26957, "epoch": 302, "lr": 9.766564533398091e-05}
{"train_loss": 0.3997056484222412, "global_step": 26958, "epoch": 302, "lr": 9.76654702594394e-05}
{"train_loss": 0.3763884902000427, "global_step": 26959, "epoch": 302, "lr": 9.766529517848984e-05}
{"train_loss": 0.5166351199150085, "global_step": 26960, "epoch": 302, "lr": 9.766512009113226e-05}
{"train_loss": 0.4323670268058777, "global_step": 26961, "epoch": 302, "lr": 9.76649449973667e-05}
{"train_loss": 0.40806087851524353, "global_step": 26962, "epoch": 302, "lr": 9.766476989719316e-05}
{"train_loss": 0.4541524350643158, "global_step": 26963, "epoch": 302, "lr": 9.766459479061168e-05}
{"train_loss": 0.4206409156322479, "global_step": 26964, "epoch": 302, "lr": 9.766441967762228e-05}
{"train_loss": 0.35364991426467896, "global_step": 26965, "epoch": 302, "lr": 9.766424455822497e-05}
{"train_loss": 0.40445462137125854, "global_step": 26966, "epoch": 302, "lr": 9.76640694324198e-05, "val_loss": 2.187347888946533}
{"train_loss": 0.4054940938949585, "global_step": 26967, "epoch": 303, "lr": 9.766389430020675e-05}
{"train_loss": 0.49778711795806885, "global_step": 26968, "epoch": 303, "lr": 9.766371916158588e-05}
{"train_loss": 0.4247351884841919, "global_step": 26969, "epoch": 303, "lr": 9.766354401655722e-05}
{"train_loss": 0.36040204763412476, "global_step": 26970, "epoch": 303, "lr": 9.766336886512076e-05}
{"train_loss": 0.4530330002307892, "global_step": 26971, "epoch": 303, "lr": 9.766319370727654e-05}
{"train_loss": 0.3003302812576294, "global_step": 26972, "epoch": 303, "lr": 9.76630185430246e-05}
{"train_loss": 0.49678218364715576, "global_step": 26973, "epoch": 303, "lr": 9.766284337236493e-05}
{"train_loss": 0.5872709155082703, "global_step": 26974, "epoch": 303, "lr": 9.766266819529758e-05}
{"train_loss": 0.5101194381713867, "global_step": 26975, "epoch": 303, "lr": 9.766249301182257e-05}
{"train_loss": 0.31955352425575256, "global_step": 26976, "epoch": 303, "lr": 9.766231782193991e-05}
{"train_loss": 0.3962499499320984, "global_step": 26977, "epoch": 303, "lr": 9.766214262564963e-05}
{"train_loss": 0.41901636123657227, "global_step": 26978, "epoch": 303, "lr": 9.766196742295178e-05}
{"train_loss": 0.43349751830101013, "global_step": 26979, "epoch": 303, "lr": 9.766179221384632e-05}
{"train_loss": 0.4220382571220398, "global_step": 26980, "epoch": 303, "lr": 9.766161699833334e-05}
{"train_loss": 0.3280593156814575, "global_step": 26981, "epoch": 303, "lr": 9.766144177641283e-05}
{"train_loss": 0.6106782555580139, "global_step": 26982, "epoch": 303, "lr": 9.766126654808483e-05}
{"train_loss": 0.40860265493392944, "global_step": 26983, "epoch": 303, "lr": 9.766109131334934e-05}
{"train_loss": 0.3870256543159485, "global_step": 26984, "epoch": 303, "lr": 9.76609160722064e-05}
{"train_loss": 0.38359948992729187, "global_step": 26985, "epoch": 303, "lr": 9.766074082465604e-05}
{"train_loss": 0.38356107473373413, "global_step": 26986, "epoch": 303, "lr": 9.766056557069828e-05}
{"train_loss": 0.41741153597831726, "global_step": 26987, "epoch": 303, "lr": 9.766039031033311e-05}
{"train_loss": 0.446017324924469, "global_step": 26988, "epoch": 303, "lr": 9.76602150435606e-05}
{"train_loss": 0.4243130087852478, "global_step": 26989, "epoch": 303, "lr": 9.766003977038074e-05}
{"train_loss": 0.3892086446285248, "global_step": 26990, "epoch": 303, "lr": 9.765986449079359e-05}
{"train_loss": 0.4630413055419922, "global_step": 26991, "epoch": 303, "lr": 9.765968920479913e-05}
{"train_loss": 0.3797948956489563, "global_step": 26992, "epoch": 303, "lr": 9.765951391239743e-05}
{"train_loss": 0.36877599358558655, "global_step": 26993, "epoch": 303, "lr": 9.765933861358847e-05}
{"train_loss": 0.3230111300945282, "global_step": 26994, "epoch": 303, "lr": 9.76591633083723e-05}
{"train_loss": 0.3047417998313904, "global_step": 26995, "epoch": 303, "lr": 9.765898799674893e-05}
{"train_loss": 0.4013313949108124, "global_step": 26996, "epoch": 303, "lr": 9.765881267871841e-05}
{"train_loss": 0.4197491705417633, "global_step": 26997, "epoch": 303, "lr": 9.765863735428071e-05}
{"train_loss": 0.4512043595314026, "global_step": 26998, "epoch": 303, "lr": 9.765846202343591e-05}
{"train_loss": 0.37651118636131287, "global_step": 26999, "epoch": 303, "lr": 9.765828668618402e-05}
{"train_loss": 0.4792359471321106, "global_step": 27000, "epoch": 303, "lr": 9.765811134252504e-05}
{"train_loss": 0.4025822579860687, "global_step": 27001, "epoch": 303, "lr": 9.7657935992459e-05}
{"train_loss": 0.40784308314323425, "global_step": 27002, "epoch": 303, "lr": 9.765776063598595e-05}
{"train_loss": 0.36206722259521484, "global_step": 27003, "epoch": 303, "lr": 9.765758527310589e-05}
{"train_loss": 0.32234424352645874, "global_step": 27004, "epoch": 303, "lr": 9.765740990381884e-05}
{"train_loss": 0.4199073612689972, "global_step": 27005, "epoch": 303, "lr": 9.765723452812483e-05}
{"train_loss": 0.28814980387687683, "global_step": 27006, "epoch": 303, "lr": 9.76570591460239e-05}
{"train_loss": 0.4270578920841217, "global_step": 27007, "epoch": 303, "lr": 9.765688375751605e-05}
{"train_loss": 0.35777735710144043, "global_step": 27008, "epoch": 303, "lr": 9.76567083626013e-05}
{"train_loss": 0.3945406675338745, "global_step": 27009, "epoch": 303, "lr": 9.76565329612797e-05}
{"train_loss": 0.41896671056747437, "global_step": 27010, "epoch": 303, "lr": 9.765635755355128e-05}
{"train_loss": 0.4626363515853882, "global_step": 27011, "epoch": 303, "lr": 9.7656182139416e-05}
{"train_loss": 0.32730844616889954, "global_step": 27012, "epoch": 303, "lr": 9.765600671887397e-05}
{"train_loss": 0.3575291037559509, "global_step": 27013, "epoch": 303, "lr": 9.765583129192515e-05}
{"train_loss": 0.3246620297431946, "global_step": 27014, "epoch": 303, "lr": 9.765565585856959e-05}
{"train_loss": 0.5154963731765747, "global_step": 27015, "epoch": 303, "lr": 9.76554804188073e-05}
{"train_loss": 0.4961523413658142, "global_step": 27016, "epoch": 303, "lr": 9.76553049726383e-05}
{"train_loss": 0.3928973376750946, "global_step": 27017, "epoch": 303, "lr": 9.765512952006266e-05}
{"train_loss": 0.6126267313957214, "global_step": 27018, "epoch": 303, "lr": 9.765495406108034e-05}
{"train_loss": 0.4192779064178467, "global_step": 27019, "epoch": 303, "lr": 9.765477859569142e-05}
{"train_loss": 0.3728070557117462, "global_step": 27020, "epoch": 303, "lr": 9.765460312389586e-05}
{"train_loss": 0.3159897029399872, "global_step": 27021, "epoch": 303, "lr": 9.765442764569373e-05}
{"train_loss": 0.448612242937088, "global_step": 27022, "epoch": 303, "lr": 9.765425216108506e-05}
{"train_loss": 0.43907836079597473, "global_step": 27023, "epoch": 303, "lr": 9.765407667006985e-05}
{"train_loss": 0.38841959834098816, "global_step": 27024, "epoch": 303, "lr": 9.765390117264812e-05}
{"train_loss": 0.40415433049201965, "global_step": 27025, "epoch": 303, "lr": 9.765372566881992e-05}
{"train_loss": 0.4738887846469879, "global_step": 27026, "epoch": 303, "lr": 9.765355015858524e-05}
{"train_loss": 0.4164981544017792, "global_step": 27027, "epoch": 303, "lr": 9.765337464194413e-05}
{"train_loss": 0.3017430901527405, "global_step": 27028, "epoch": 303, "lr": 9.76531991188966e-05}
{"train_loss": 0.455883651971817, "global_step": 27029, "epoch": 303, "lr": 9.765302358944268e-05}
{"train_loss": 0.5133966207504272, "global_step": 27030, "epoch": 303, "lr": 9.765284805358239e-05}
{"train_loss": 0.2706107199192047, "global_step": 27031, "epoch": 303, "lr": 9.765267251131577e-05}
{"train_loss": 0.3060983419418335, "global_step": 27032, "epoch": 303, "lr": 9.765249696264281e-05}
{"train_loss": 0.40991270542144775, "global_step": 27033, "epoch": 303, "lr": 9.765232140756356e-05}
{"train_loss": 0.2604246437549591, "global_step": 27034, "epoch": 303, "lr": 9.765214584607805e-05}
{"train_loss": 0.4476062059402466, "global_step": 27035, "epoch": 303, "lr": 9.765197027818626e-05}
{"train_loss": 0.37223154306411743, "global_step": 27036, "epoch": 303, "lr": 9.765179470388827e-05}
{"train_loss": 0.4053380787372589, "global_step": 27037, "epoch": 303, "lr": 9.765161912318405e-05}
{"train_loss": 0.44253048300743103, "global_step": 27038, "epoch": 303, "lr": 9.765144353607368e-05}
{"train_loss": 0.39215871691703796, "global_step": 27039, "epoch": 303, "lr": 9.765126794255714e-05}
{"train_loss": 0.4288281798362732, "global_step": 27040, "epoch": 303, "lr": 9.765109234263448e-05}
{"train_loss": 0.3659358322620392, "global_step": 27041, "epoch": 303, "lr": 9.765091673630568e-05}
{"train_loss": 0.34005600214004517, "global_step": 27042, "epoch": 303, "lr": 9.765074112357083e-05}
{"train_loss": 0.2911202311515808, "global_step": 27043, "epoch": 303, "lr": 9.76505655044299e-05}
{"train_loss": 0.37109455466270447, "global_step": 27044, "epoch": 303, "lr": 9.765038987888294e-05}
{"train_loss": 0.32995086908340454, "global_step": 27045, "epoch": 303, "lr": 9.765021424692996e-05}
{"train_loss": 0.42519912123680115, "global_step": 27046, "epoch": 303, "lr": 9.765003860857099e-05}
{"train_loss": 0.4563395380973816, "global_step": 27047, "epoch": 303, "lr": 9.764986296380606e-05}
{"train_loss": 0.338142454624176, "global_step": 27048, "epoch": 303, "lr": 9.764968731263518e-05}
{"train_loss": 0.40189892053604126, "global_step": 27049, "epoch": 303, "lr": 9.764951165505838e-05}
{"train_loss": 0.37432223558425903, "global_step": 27050, "epoch": 303, "lr": 9.764933599107569e-05}
{"train_loss": 0.6194800734519958, "global_step": 27051, "epoch": 303, "lr": 9.764916032068712e-05}
{"train_loss": 0.5339572429656982, "global_step": 27052, "epoch": 303, "lr": 9.764898464389272e-05}
{"train_loss": 0.29522064328193665, "global_step": 27053, "epoch": 303, "lr": 9.764880896069246e-05}
{"train_loss": 0.3616553246974945, "global_step": 27054, "epoch": 303, "lr": 9.764863327108642e-05}
{"train_loss": 0.40550254335564173, "global_step": 27055, "epoch": 303, "lr": 9.76484575750746e-05, "val_loss": 2.2055931091308594}
{"train_loss": 0.49512165784835815, "global_step": 27056, "epoch": 304, "lr": 9.764828187265704e-05}
{"train_loss": 0.3694450855255127, "global_step": 27057, "epoch": 304, "lr": 9.764810616383373e-05}
{"train_loss": 0.33387771248817444, "global_step": 27058, "epoch": 304, "lr": 9.764793044860471e-05}
{"train_loss": 0.3704489469528198, "global_step": 27059, "epoch": 304, "lr": 9.764775472697002e-05}
{"train_loss": 0.39342379570007324, "global_step": 27060, "epoch": 304, "lr": 9.764757899892966e-05}
{"train_loss": 0.42931032180786133, "global_step": 27061, "epoch": 304, "lr": 9.764740326448368e-05}
{"train_loss": 0.3935736417770386, "global_step": 27062, "epoch": 304, "lr": 9.764722752363206e-05}
{"train_loss": 0.39058518409729004, "global_step": 27063, "epoch": 304, "lr": 9.764705177637488e-05}
{"train_loss": 0.5015541911125183, "global_step": 27064, "epoch": 304, "lr": 9.764687602271211e-05}
{"train_loss": 0.40128540992736816, "global_step": 27065, "epoch": 304, "lr": 9.764670026264382e-05}
{"train_loss": 0.35040247440338135, "global_step": 27066, "epoch": 304, "lr": 9.764652449617e-05}
{"train_loss": 0.320738285779953, "global_step": 27067, "epoch": 304, "lr": 9.764634872329067e-05}
{"train_loss": 0.3650256097316742, "global_step": 27068, "epoch": 304, "lr": 9.76461729440059e-05}
{"train_loss": 0.49360132217407227, "global_step": 27069, "epoch": 304, "lr": 9.764599715831567e-05}
{"train_loss": 0.5089235901832581, "global_step": 27070, "epoch": 304, "lr": 9.764582136622e-05}
{"train_loss": 0.3508250415325165, "global_step": 27071, "epoch": 304, "lr": 9.764564556771896e-05}
{"train_loss": 0.4495939314365387, "global_step": 27072, "epoch": 304, "lr": 9.764546976281253e-05}
{"train_loss": 0.5047273635864258, "global_step": 27073, "epoch": 304, "lr": 9.764529395150073e-05}
{"train_loss": 0.3969654142856598, "global_step": 27074, "epoch": 304, "lr": 9.764511813378363e-05}
{"train_loss": 0.4553576111793518, "global_step": 27075, "epoch": 304, "lr": 9.764494230966121e-05}
{"train_loss": 0.40763816237449646, "global_step": 27076, "epoch": 304, "lr": 9.764476647913352e-05}
{"train_loss": 0.38705068826675415, "global_step": 27077, "epoch": 304, "lr": 9.764459064220056e-05}
{"train_loss": 0.4237321615219116, "global_step": 27078, "epoch": 304, "lr": 9.764441479886235e-05}
{"train_loss": 0.2857745587825775, "global_step": 27079, "epoch": 304, "lr": 9.764423894911896e-05}
{"train_loss": 0.3705718517303467, "global_step": 27080, "epoch": 304, "lr": 9.764406309297036e-05}
{"train_loss": 0.44903531670570374, "global_step": 27081, "epoch": 304, "lr": 9.76438872304166e-05}
{"train_loss": 0.4171358644962311, "global_step": 27082, "epoch": 304, "lr": 9.764371136145771e-05}
{"train_loss": 0.4954795241355896, "global_step": 27083, "epoch": 304, "lr": 9.764353548609371e-05}
{"train_loss": 0.418743759393692, "global_step": 27084, "epoch": 304, "lr": 9.76433596043246e-05}
{"train_loss": 0.5484210252761841, "global_step": 27085, "epoch": 304, "lr": 9.764318371615042e-05}
{"train_loss": 0.40617915987968445, "global_step": 27086, "epoch": 304, "lr": 9.76430078215712e-05}
{"train_loss": 0.40423494577407837, "global_step": 27087, "epoch": 304, "lr": 9.764283192058697e-05}
{"train_loss": 0.512567937374115, "global_step": 27088, "epoch": 304, "lr": 9.764265601319773e-05}
{"train_loss": 0.4948001801967621, "global_step": 27089, "epoch": 304, "lr": 9.764248009940352e-05}
{"train_loss": 0.33857083320617676, "global_step": 27090, "epoch": 304, "lr": 9.764230417920436e-05}
{"train_loss": 0.49246490001678467, "global_step": 27091, "epoch": 304, "lr": 9.764212825260026e-05}
{"train_loss": 0.4563879668712616, "global_step": 27092, "epoch": 304, "lr": 9.764195231959128e-05}
{"train_loss": 0.3853263556957245, "global_step": 27093, "epoch": 304, "lr": 9.76417763801774e-05}
{"train_loss": 0.36387938261032104, "global_step": 27094, "epoch": 304, "lr": 9.764160043435867e-05}
{"train_loss": 0.5318782925605774, "global_step": 27095, "epoch": 304, "lr": 9.76414244821351e-05}
{"train_loss": 0.42475754022598267, "global_step": 27096, "epoch": 304, "lr": 9.764124852350673e-05}
{"train_loss": 0.24778461456298828, "global_step": 27097, "epoch": 304, "lr": 9.764107255847359e-05}
{"train_loss": 0.3128305673599243, "global_step": 27098, "epoch": 304, "lr": 9.764089658703567e-05}
{"train_loss": 0.3783353269100189, "global_step": 27099, "epoch": 304, "lr": 9.7640720609193e-05}
{"train_loss": 0.4888302683830261, "global_step": 27100, "epoch": 304, "lr": 9.764054462494564e-05}
{"train_loss": 0.3065946102142334, "global_step": 27101, "epoch": 304, "lr": 9.764036863429358e-05}
{"train_loss": 0.3370693624019623, "global_step": 27102, "epoch": 304, "lr": 9.764019263723684e-05}
{"train_loss": 0.3064582049846649, "global_step": 27103, "epoch": 304, "lr": 9.764001663377549e-05}
{"train_loss": 0.4199840724468231, "global_step": 27104, "epoch": 304, "lr": 9.763984062390949e-05}
{"train_loss": 0.48839181661605835, "global_step": 27105, "epoch": 304, "lr": 9.763966460763892e-05}
{"train_loss": 0.48325803875923157, "global_step": 27106, "epoch": 304, "lr": 9.763948858496375e-05}
{"train_loss": 0.38047170639038086, "global_step": 27107, "epoch": 304, "lr": 9.763931255588404e-05}
{"train_loss": 0.41752973198890686, "global_step": 27108, "epoch": 304, "lr": 9.763913652039981e-05}
{"train_loss": 0.33502721786499023, "global_step": 27109, "epoch": 304, "lr": 9.763896047851107e-05}
{"train_loss": 0.363116055727005, "global_step": 27110, "epoch": 304, "lr": 9.763878443021786e-05}
{"train_loss": 0.4222455620765686, "global_step": 27111, "epoch": 304, "lr": 9.76386083755202e-05}
{"train_loss": 0.42971304059028625, "global_step": 27112, "epoch": 304, "lr": 9.763843231441811e-05}
{"train_loss": 0.31327202916145325, "global_step": 27113, "epoch": 304, "lr": 9.76382562469116e-05}
{"train_loss": 0.4329147934913635, "global_step": 27114, "epoch": 304, "lr": 9.763808017300073e-05}
{"train_loss": 0.5339162349700928, "global_step": 27115, "epoch": 304, "lr": 9.763790409268547e-05}
{"train_loss": 0.3665187954902649, "global_step": 27116, "epoch": 304, "lr": 9.76377280059659e-05}
{"train_loss": 0.3425922989845276, "global_step": 27117, "epoch": 304, "lr": 9.763755191284201e-05}
{"train_loss": 0.3447628915309906, "global_step": 27118, "epoch": 304, "lr": 9.763737581331382e-05}
{"train_loss": 0.45306137204170227, "global_step": 27119, "epoch": 304, "lr": 9.763719970738138e-05}
{"train_loss": 0.4696080982685089, "global_step": 27120, "epoch": 304, "lr": 9.76370235950447e-05}
{"train_loss": 0.35664358735084534, "global_step": 27121, "epoch": 304, "lr": 9.76368474763038e-05}
{"train_loss": 0.37726789712905884, "global_step": 27122, "epoch": 304, "lr": 9.76366713511587e-05}
{"train_loss": 0.45092248916625977, "global_step": 27123, "epoch": 304, "lr": 9.763649521960945e-05}
{"train_loss": 0.41038888692855835, "global_step": 27124, "epoch": 304, "lr": 9.763631908165602e-05}
{"train_loss": 0.3845062851905823, "global_step": 27125, "epoch": 304, "lr": 9.76361429372985e-05}
{"train_loss": 0.5113428831100464, "global_step": 27126, "epoch": 304, "lr": 9.763596678653686e-05}
{"train_loss": 0.4200286269187927, "global_step": 27127, "epoch": 304, "lr": 9.763579062937116e-05}
{"train_loss": 0.31364816427230835, "global_step": 27128, "epoch": 304, "lr": 9.76356144658014e-05}
{"train_loss": 0.44298702478408813, "global_step": 27129, "epoch": 304, "lr": 9.76354382958276e-05}
{"train_loss": 0.4129321277141571, "global_step": 27130, "epoch": 304, "lr": 9.763526211944983e-05}
{"train_loss": 0.36294877529144287, "global_step": 27131, "epoch": 304, "lr": 9.763508593666804e-05}
{"train_loss": 0.41661888360977173, "global_step": 27132, "epoch": 304, "lr": 9.763490974748231e-05}
{"train_loss": 0.39788752794265747, "global_step": 27133, "epoch": 304, "lr": 9.763473355189266e-05}
{"train_loss": 0.4615453779697418, "global_step": 27134, "epoch": 304, "lr": 9.763455734989908e-05}
{"train_loss": 0.47999143600463867, "global_step": 27135, "epoch": 304, "lr": 9.763438114150163e-05}
{"train_loss": 0.3172074854373932, "global_step": 27136, "epoch": 304, "lr": 9.763420492670032e-05}
{"train_loss": 0.436472624540329, "global_step": 27137, "epoch": 304, "lr": 9.763402870549516e-05}
{"train_loss": 0.42533817887306213, "global_step": 27138, "epoch": 304, "lr": 9.763385247788619e-05}
{"train_loss": 0.4450576901435852, "global_step": 27139, "epoch": 304, "lr": 9.763367624387343e-05}
{"train_loss": 0.48653337359428406, "global_step": 27140, "epoch": 304, "lr": 9.76335000034569e-05}
{"train_loss": 0.3893630802631378, "global_step": 27141, "epoch": 304, "lr": 9.763332375663663e-05}
{"train_loss": 0.4651336669921875, "global_step": 27142, "epoch": 304, "lr": 9.763314750341263e-05}
{"train_loss": 0.5144690871238708, "global_step": 27143, "epoch": 304, "lr": 9.763297124378494e-05}
{"train_loss": 0.4135670886280831, "global_step": 27144, "epoch": 304, "lr": 9.763279497775358e-05, "val_loss": 2.0449583530426025}
{"train_loss": 0.3832612633705139, "global_step": 27145, "epoch": 305, "lr": 9.763261870531856e-05}
{"train_loss": 0.4911500811576843, "global_step": 27146, "epoch": 305, "lr": 9.763244242647994e-05}
{"train_loss": 0.43442070484161377, "global_step": 27147, "epoch": 305, "lr": 9.76322661412377e-05}
{"train_loss": 0.2965136170387268, "global_step": 27148, "epoch": 305, "lr": 9.76320898495919e-05}
{"train_loss": 0.357654869556427, "global_step": 27149, "epoch": 305, "lr": 9.763191355154252e-05}
{"train_loss": 0.42339614033699036, "global_step": 27150, "epoch": 305, "lr": 9.763173724708963e-05}
{"train_loss": 0.3636610507965088, "global_step": 27151, "epoch": 305, "lr": 9.763156093623323e-05}
{"train_loss": 0.4033144414424896, "global_step": 27152, "epoch": 305, "lr": 9.763138461897334e-05}
{"train_loss": 0.39664825797080994, "global_step": 27153, "epoch": 305, "lr": 9.763120829530998e-05}
{"train_loss": 0.4388233423233032, "global_step": 27154, "epoch": 305, "lr": 9.76310319652432e-05}
{"train_loss": 0.41533297300338745, "global_step": 27155, "epoch": 305, "lr": 9.7630855628773e-05}
{"train_loss": 0.4385174512863159, "global_step": 27156, "epoch": 305, "lr": 9.763067928589943e-05}
{"train_loss": 0.34421175718307495, "global_step": 27157, "epoch": 305, "lr": 9.763050293662248e-05}
{"train_loss": 0.413023442029953, "global_step": 27158, "epoch": 305, "lr": 9.76303265809422e-05}
{"train_loss": 0.3792862892150879, "global_step": 27159, "epoch": 305, "lr": 9.763015021885859e-05}
{"train_loss": 0.34105879068374634, "global_step": 27160, "epoch": 305, "lr": 9.762997385037169e-05}
{"train_loss": 0.42675188183784485, "global_step": 27161, "epoch": 305, "lr": 9.762979747548154e-05}
{"train_loss": 0.3084724545478821, "global_step": 27162, "epoch": 305, "lr": 9.762962109418812e-05}
{"train_loss": 0.33833521604537964, "global_step": 27163, "epoch": 305, "lr": 9.762944470649149e-05}
{"train_loss": 0.37365224957466125, "global_step": 27164, "epoch": 305, "lr": 9.762926831239166e-05}
{"train_loss": 0.28153765201568604, "global_step": 27165, "epoch": 305, "lr": 9.762909191188865e-05}
{"train_loss": 0.3729385733604431, "global_step": 27166, "epoch": 305, "lr": 9.762891550498249e-05}
{"train_loss": 0.377570778131485, "global_step": 27167, "epoch": 305, "lr": 9.76287390916732e-05}
{"train_loss": 0.3835732936859131, "global_step": 27168, "epoch": 305, "lr": 9.762856267196082e-05}
{"train_loss": 0.3732116222381592, "global_step": 27169, "epoch": 305, "lr": 9.762838624584535e-05}
{"train_loss": 0.3947432339191437, "global_step": 27170, "epoch": 305, "lr": 9.762820981332684e-05}
{"train_loss": 0.42113712430000305, "global_step": 27171, "epoch": 305, "lr": 9.762803337440527e-05}
{"train_loss": 0.3440772593021393, "global_step": 27172, "epoch": 305, "lr": 9.762785692908069e-05}
{"train_loss": 0.3805542290210724, "global_step": 27173, "epoch": 305, "lr": 9.762768047735315e-05}
{"train_loss": 0.3653452396392822, "global_step": 27174, "epoch": 305, "lr": 9.762750401922264e-05}
{"train_loss": 0.332805871963501, "global_step": 27175, "epoch": 305, "lr": 9.76273275546892e-05}
{"train_loss": 0.2816501259803772, "global_step": 27176, "epoch": 305, "lr": 9.762715108375283e-05}
{"train_loss": 0.4010055363178253, "global_step": 27177, "epoch": 305, "lr": 9.762697460641359e-05}
{"train_loss": 0.3695228099822998, "global_step": 27178, "epoch": 305, "lr": 9.762679812267146e-05}
{"train_loss": 0.3548039495944977, "global_step": 27179, "epoch": 305, "lr": 9.76266216325265e-05}
{"train_loss": 0.4499758780002594, "global_step": 27180, "epoch": 305, "lr": 9.762644513597873e-05}
{"train_loss": 0.29589271545410156, "global_step": 27181, "epoch": 305, "lr": 9.762626863302815e-05}
{"train_loss": 0.33506813645362854, "global_step": 27182, "epoch": 305, "lr": 9.762609212367481e-05}
{"train_loss": 0.366700142621994, "global_step": 27183, "epoch": 305, "lr": 9.762591560791872e-05}
{"train_loss": 0.3595665395259857, "global_step": 27184, "epoch": 305, "lr": 9.76257390857599e-05}
{"train_loss": 0.32347291707992554, "global_step": 27185, "epoch": 305, "lr": 9.762556255719838e-05}
{"train_loss": 0.4094597101211548, "global_step": 27186, "epoch": 305, "lr": 9.762538602223419e-05}
{"train_loss": 0.3646814525127411, "global_step": 27187, "epoch": 305, "lr": 9.762520948086733e-05}
{"train_loss": 0.42049479484558105, "global_step": 27188, "epoch": 305, "lr": 9.762503293309786e-05}
{"train_loss": 0.41089704632759094, "global_step": 27189, "epoch": 305, "lr": 9.762485637892578e-05}
{"train_loss": 0.3107762038707733, "global_step": 27190, "epoch": 305, "lr": 9.762467981835112e-05}
{"train_loss": 0.3203100860118866, "global_step": 27191, "epoch": 305, "lr": 9.76245032513739e-05}
{"train_loss": 0.3701326847076416, "global_step": 27192, "epoch": 305, "lr": 9.762432667799414e-05}
{"train_loss": 0.2919948995113373, "global_step": 27193, "epoch": 305, "lr": 9.762415009821188e-05}
{"train_loss": 0.3638657331466675, "global_step": 27194, "epoch": 305, "lr": 9.762397351202713e-05}
{"train_loss": 0.39191579818725586, "global_step": 27195, "epoch": 305, "lr": 9.762379691943993e-05}
{"train_loss": 0.4257034361362457, "global_step": 27196, "epoch": 305, "lr": 9.762362032045028e-05}
{"train_loss": 0.4379637539386749, "global_step": 27197, "epoch": 305, "lr": 9.762344371505821e-05}
{"train_loss": 0.5076084733009338, "global_step": 27198, "epoch": 305, "lr": 9.762326710326374e-05}
{"train_loss": 0.4077880084514618, "global_step": 27199, "epoch": 305, "lr": 9.76230904850669e-05}
{"train_loss": 0.3795652985572815, "global_step": 27200, "epoch": 305, "lr": 9.762291386046774e-05}
{"train_loss": 0.3702734112739563, "global_step": 27201, "epoch": 305, "lr": 9.762273722946624e-05}
{"train_loss": 0.43826282024383545, "global_step": 27202, "epoch": 305, "lr": 9.762256059206246e-05}
{"train_loss": 0.4303742051124573, "global_step": 27203, "epoch": 305, "lr": 9.76223839482564e-05}
{"train_loss": 0.4165082573890686, "global_step": 27204, "epoch": 305, "lr": 9.762220729804809e-05}
{"train_loss": 0.44471609592437744, "global_step": 27205, "epoch": 305, "lr": 9.762203064143756e-05}
{"train_loss": 0.25911596417427063, "global_step": 27206, "epoch": 305, "lr": 9.762185397842483e-05}
{"train_loss": 0.4468299150466919, "global_step": 27207, "epoch": 305, "lr": 9.762167730900991e-05}
{"train_loss": 0.3655238747596741, "global_step": 27208, "epoch": 305, "lr": 9.762150063319284e-05}
{"train_loss": 0.41539955139160156, "global_step": 27209, "epoch": 305, "lr": 9.762132395097364e-05}
{"train_loss": 0.32307443022727966, "global_step": 27210, "epoch": 305, "lr": 9.762114726235233e-05}
{"train_loss": 0.37097790837287903, "global_step": 27211, "epoch": 305, "lr": 9.762097056732894e-05}
{"train_loss": 0.35796526074409485, "global_step": 27212, "epoch": 305, "lr": 9.762079386590349e-05}
{"train_loss": 0.2971125841140747, "global_step": 27213, "epoch": 305, "lr": 9.762061715807602e-05}
{"train_loss": 0.28389570116996765, "global_step": 27214, "epoch": 305, "lr": 9.762044044384652e-05}
{"train_loss": 0.33840811252593994, "global_step": 27215, "epoch": 305, "lr": 9.762026372321504e-05}
{"train_loss": 0.487263560295105, "global_step": 27216, "epoch": 305, "lr": 9.76200869961816e-05}
{"train_loss": 0.5461084842681885, "global_step": 27217, "epoch": 305, "lr": 9.761991026274619e-05}
{"train_loss": 0.34519386291503906, "global_step": 27218, "epoch": 305, "lr": 9.761973352290889e-05}
{"train_loss": 0.4034724533557892, "global_step": 27219, "epoch": 305, "lr": 9.761955677666969e-05}
{"train_loss": 0.5326716303825378, "global_step": 27220, "epoch": 305, "lr": 9.761938002402865e-05}
{"train_loss": 0.41154542565345764, "global_step": 27221, "epoch": 305, "lr": 9.761920326498572e-05}
{"train_loss": 0.4665972888469696, "global_step": 27222, "epoch": 305, "lr": 9.7619026499541e-05}
{"train_loss": 0.4340435564517975, "global_step": 27223, "epoch": 305, "lr": 9.761884972769446e-05}
{"train_loss": 0.332764595746994, "global_step": 27224, "epoch": 305, "lr": 9.761867294944615e-05}
{"train_loss": 0.4761175513267517, "global_step": 27225, "epoch": 305, "lr": 9.76184961647961e-05}
{"train_loss": 0.2919362783432007, "global_step": 27226, "epoch": 305, "lr": 9.761831937374433e-05}
{"train_loss": 0.3176416754722595, "global_step": 27227, "epoch": 305, "lr": 9.761814257629085e-05}
{"train_loss": 0.39208778738975525, "global_step": 27228, "epoch": 305, "lr": 9.76179657724357e-05}
{"train_loss": 0.39686909317970276, "global_step": 27229, "epoch": 305, "lr": 9.761778896217887e-05}
{"train_loss": 0.4297492504119873, "global_step": 27230, "epoch": 305, "lr": 9.761761214552041e-05}
{"train_loss": 0.40389755368232727, "global_step": 27231, "epoch": 305, "lr": 9.761743532246037e-05}
{"train_loss": 0.48091360926628113, "global_step": 27232, "epoch": 305, "lr": 9.761725849299872e-05}
{"train_loss": 0.38487903957956293, "global_step": 27233, "epoch": 305, "lr": 9.761708165713552e-05, "val_loss": 2.100707769393921, "train_action_mse_error": 54.037139892578125}
{"train_loss": 0.35936158895492554, "global_step": 27234, "epoch": 306, "lr": 9.76169048148708e-05}
{"train_loss": 0.39310136437416077, "global_step": 27235, "epoch": 306, "lr": 9.761672796620454e-05}
{"train_loss": 0.377895325422287, "global_step": 27236, "epoch": 306, "lr": 9.761655111113681e-05}
{"train_loss": 0.3667581081390381, "global_step": 27237, "epoch": 306, "lr": 9.761637424966761e-05}
{"train_loss": 0.3823550045490265, "global_step": 27238, "epoch": 306, "lr": 9.761619738179696e-05}
{"train_loss": 0.43783900141716003, "global_step": 27239, "epoch": 306, "lr": 9.76160205075249e-05}
{"train_loss": 0.37988173961639404, "global_step": 27240, "epoch": 306, "lr": 9.761584362685145e-05}
{"train_loss": 0.3596624732017517, "global_step": 27241, "epoch": 306, "lr": 9.761566673977662e-05}
{"train_loss": 0.46193310618400574, "global_step": 27242, "epoch": 306, "lr": 9.761548984630045e-05}
{"train_loss": 0.29742294549942017, "global_step": 27243, "epoch": 306, "lr": 9.761531294642296e-05}
{"train_loss": 0.41948074102401733, "global_step": 27244, "epoch": 306, "lr": 9.761513604014418e-05}
{"train_loss": 0.39932897686958313, "global_step": 27245, "epoch": 306, "lr": 9.76149591274641e-05}
{"train_loss": 0.42455318570137024, "global_step": 27246, "epoch": 306, "lr": 9.761478220838279e-05}
{"train_loss": 0.38027575612068176, "global_step": 27247, "epoch": 306, "lr": 9.761460528290025e-05}
{"train_loss": 0.41423022747039795, "global_step": 27248, "epoch": 306, "lr": 9.761442835101651e-05}
{"train_loss": 0.3147587478160858, "global_step": 27249, "epoch": 306, "lr": 9.761425141273157e-05}
{"train_loss": 0.49358171224594116, "global_step": 27250, "epoch": 306, "lr": 9.761407446804549e-05}
{"train_loss": 0.27750733494758606, "global_step": 27251, "epoch": 306, "lr": 9.761389751695828e-05}
{"train_loss": 0.3657531142234802, "global_step": 27252, "epoch": 306, "lr": 9.761372055946996e-05}
{"train_loss": 0.4568765461444855, "global_step": 27253, "epoch": 306, "lr": 9.761354359558055e-05}
{"train_loss": 0.37585145235061646, "global_step": 27254, "epoch": 306, "lr": 9.761336662529008e-05}
{"train_loss": 0.48889535665512085, "global_step": 27255, "epoch": 306, "lr": 9.761318964859859e-05}
{"train_loss": 0.3463761806488037, "global_step": 27256, "epoch": 306, "lr": 9.761301266550607e-05}
{"train_loss": 0.25922277569770813, "global_step": 27257, "epoch": 306, "lr": 9.761283567601256e-05}
{"train_loss": 0.352395623922348, "global_step": 27258, "epoch": 306, "lr": 9.761265868011808e-05}
{"train_loss": 0.3314165472984314, "global_step": 27259, "epoch": 306, "lr": 9.761248167782267e-05}
{"train_loss": 0.3999984562397003, "global_step": 27260, "epoch": 306, "lr": 9.761230466912634e-05}
{"train_loss": 0.3641146719455719, "global_step": 27261, "epoch": 306, "lr": 9.761212765402911e-05}
{"train_loss": 0.4007211923599243, "global_step": 27262, "epoch": 306, "lr": 9.761195063253101e-05}
{"train_loss": 0.4278772473335266, "global_step": 27263, "epoch": 306, "lr": 9.761177360463208e-05}
{"train_loss": 0.3428359627723694, "global_step": 27264, "epoch": 306, "lr": 9.761159657033232e-05}
{"train_loss": 0.3455832898616791, "global_step": 27265, "epoch": 306, "lr": 9.761141952963175e-05}
{"train_loss": 0.307476669549942, "global_step": 27266, "epoch": 306, "lr": 9.761124248253042e-05}
{"train_loss": 0.3436642289161682, "global_step": 27267, "epoch": 306, "lr": 9.761106542902831e-05}
{"train_loss": 0.3620762526988983, "global_step": 27268, "epoch": 306, "lr": 9.76108883691255e-05}
{"train_loss": 0.4633890688419342, "global_step": 27269, "epoch": 306, "lr": 9.761071130282198e-05}
{"train_loss": 0.43833455443382263, "global_step": 27270, "epoch": 306, "lr": 9.76105342301178e-05}
{"train_loss": 0.35470879077911377, "global_step": 27271, "epoch": 306, "lr": 9.761035715101292e-05}
{"train_loss": 0.30054011940956116, "global_step": 27272, "epoch": 306, "lr": 9.761018006550744e-05}
{"train_loss": 0.357891321182251, "global_step": 27273, "epoch": 306, "lr": 9.761000297360134e-05}
{"train_loss": 0.35245653986930847, "global_step": 27274, "epoch": 306, "lr": 9.760982587529466e-05}
{"train_loss": 0.3828839063644409, "global_step": 27275, "epoch": 306, "lr": 9.760964877058743e-05}
{"train_loss": 0.4449147880077362, "global_step": 27276, "epoch": 306, "lr": 9.760947165947964e-05}
{"train_loss": 0.46227431297302246, "global_step": 27277, "epoch": 306, "lr": 9.760929454197135e-05}
{"train_loss": 0.38860437273979187, "global_step": 27278, "epoch": 306, "lr": 9.760911741806256e-05}
{"train_loss": 0.3617410361766815, "global_step": 27279, "epoch": 306, "lr": 9.760894028775332e-05}
{"train_loss": 0.44580593705177307, "global_step": 27280, "epoch": 306, "lr": 9.760876315104363e-05}
{"train_loss": 0.43850138783454895, "global_step": 27281, "epoch": 306, "lr": 9.760858600793352e-05}
{"train_loss": 0.5164118409156799, "global_step": 27282, "epoch": 306, "lr": 9.760840885842302e-05}
{"train_loss": 0.35170766711235046, "global_step": 27283, "epoch": 306, "lr": 9.760823170251215e-05}
{"train_loss": 0.3491794168949127, "global_step": 27284, "epoch": 306, "lr": 9.760805454020092e-05}
{"train_loss": 0.4008985459804535, "global_step": 27285, "epoch": 306, "lr": 9.760787737148939e-05}
{"train_loss": 0.5271500945091248, "global_step": 27286, "epoch": 306, "lr": 9.760770019637755e-05}
{"train_loss": 0.3440050184726715, "global_step": 27287, "epoch": 306, "lr": 9.760752301486543e-05}
{"train_loss": 0.4408017694950104, "global_step": 27288, "epoch": 306, "lr": 9.760734582695305e-05}
{"train_loss": 0.5149708390235901, "global_step": 27289, "epoch": 306, "lr": 9.760716863264047e-05}
{"train_loss": 0.30837172269821167, "global_step": 27290, "epoch": 306, "lr": 9.760699143192767e-05}
{"train_loss": 0.4352389872074127, "global_step": 27291, "epoch": 306, "lr": 9.76068142248147e-05}
{"train_loss": 0.44384288787841797, "global_step": 27292, "epoch": 306, "lr": 9.760663701130158e-05}
{"train_loss": 0.40838247537612915, "global_step": 27293, "epoch": 306, "lr": 9.76064597913883e-05}
{"train_loss": 0.45465707778930664, "global_step": 27294, "epoch": 306, "lr": 9.760628256507492e-05}
{"train_loss": 0.38060250878334045, "global_step": 27295, "epoch": 306, "lr": 9.760610533236147e-05}
{"train_loss": 0.4385162591934204, "global_step": 27296, "epoch": 306, "lr": 9.760592809324795e-05}
{"train_loss": 0.44539380073547363, "global_step": 27297, "epoch": 306, "lr": 9.76057508477344e-05}
{"train_loss": 0.316617488861084, "global_step": 27298, "epoch": 306, "lr": 9.760557359582084e-05}
{"train_loss": 0.33780401945114136, "global_step": 27299, "epoch": 306, "lr": 9.760539633750727e-05}
{"train_loss": 0.46402525901794434, "global_step": 27300, "epoch": 306, "lr": 9.760521907279374e-05}
{"train_loss": 0.42484596371650696, "global_step": 27301, "epoch": 306, "lr": 9.760504180168028e-05}
{"train_loss": 0.41088953614234924, "global_step": 27302, "epoch": 306, "lr": 9.76048645241669e-05}
{"train_loss": 0.3918658494949341, "global_step": 27303, "epoch": 306, "lr": 9.760468724025362e-05}
{"train_loss": 0.5106390118598938, "global_step": 27304, "epoch": 306, "lr": 9.760450994994048e-05}
{"train_loss": 0.6130375862121582, "global_step": 27305, "epoch": 306, "lr": 9.760433265322748e-05}
{"train_loss": 0.48101553320884705, "global_step": 27306, "epoch": 306, "lr": 9.760415535011467e-05}
{"train_loss": 0.3754171133041382, "global_step": 27307, "epoch": 306, "lr": 9.760397804060206e-05}
{"train_loss": 0.3424654006958008, "global_step": 27308, "epoch": 306, "lr": 9.760380072468967e-05}
{"train_loss": 0.4535905420780182, "global_step": 27309, "epoch": 306, "lr": 9.760362340237754e-05}
{"train_loss": 0.41701191663742065, "global_step": 27310, "epoch": 306, "lr": 9.760344607366567e-05}
{"train_loss": 0.4592481255531311, "global_step": 27311, "epoch": 306, "lr": 9.76032687385541e-05}
{"train_loss": 0.38824763894081116, "global_step": 27312, "epoch": 306, "lr": 9.760309139704287e-05}
{"train_loss": 0.38436347246170044, "global_step": 27313, "epoch": 306, "lr": 9.760291404913196e-05}
{"train_loss": 0.5334948301315308, "global_step": 27314, "epoch": 306, "lr": 9.760273669482143e-05}
{"train_loss": 0.4152454435825348, "global_step": 27315, "epoch": 306, "lr": 9.760255933411129e-05}
{"train_loss": 0.4297325909137726, "global_step": 27316, "epoch": 306, "lr": 9.760238196700156e-05}
{"train_loss": 0.4769686162471771, "global_step": 27317, "epoch": 306, "lr": 9.760220459349227e-05}
{"train_loss": 0.40823614597320557, "global_step": 27318, "epoch": 306, "lr": 9.760202721358344e-05}
{"train_loss": 0.3335936963558197, "global_step": 27319, "epoch": 306, "lr": 9.760184982727511e-05}
{"train_loss": 0.50411057472229, "global_step": 27320, "epoch": 306, "lr": 9.760167243456729e-05}
{"train_loss": 0.2779492735862732, "global_step": 27321, "epoch": 306, "lr": 9.760149503546e-05}
{"train_loss": 0.4027849633372232, "global_step": 27322, "epoch": 306, "lr": 9.760131762995328e-05, "val_loss": 2.1634953022003174}
{"train_loss": 0.4303022623062134, "global_step": 27323, "epoch": 307, "lr": 9.760114021804712e-05}
{"train_loss": 0.36833134293556213, "global_step": 27324, "epoch": 307, "lr": 9.760096279974159e-05}
{"train_loss": 0.4141786992549896, "global_step": 27325, "epoch": 307, "lr": 9.760078537503668e-05}
{"train_loss": 0.3968721032142639, "global_step": 27326, "epoch": 307, "lr": 9.760060794393242e-05}
{"train_loss": 0.463248074054718, "global_step": 27327, "epoch": 307, "lr": 9.760043050642885e-05}
{"train_loss": 0.38951653242111206, "global_step": 27328, "epoch": 307, "lr": 9.760025306252599e-05}
{"train_loss": 0.37396204471588135, "global_step": 27329, "epoch": 307, "lr": 9.760007561222384e-05}
{"train_loss": 0.3241631090641022, "global_step": 27330, "epoch": 307, "lr": 9.759989815552243e-05}
{"train_loss": 0.5313653349876404, "global_step": 27331, "epoch": 307, "lr": 9.759972069242181e-05}
{"train_loss": 0.39122557640075684, "global_step": 27332, "epoch": 307, "lr": 9.759954322292199e-05}
{"train_loss": 0.45515039563179016, "global_step": 27333, "epoch": 307, "lr": 9.759936574702298e-05}
{"train_loss": 0.5505855083465576, "global_step": 27334, "epoch": 307, "lr": 9.759918826472483e-05}
{"train_loss": 0.3326670825481415, "global_step": 27335, "epoch": 307, "lr": 9.759901077602753e-05}
{"train_loss": 0.3713083267211914, "global_step": 27336, "epoch": 307, "lr": 9.759883328093115e-05}
{"train_loss": 0.3922101557254791, "global_step": 27337, "epoch": 307, "lr": 9.759865577943566e-05}
{"train_loss": 0.3016430139541626, "global_step": 27338, "epoch": 307, "lr": 9.759847827154113e-05}
{"train_loss": 0.5380950570106506, "global_step": 27339, "epoch": 307, "lr": 9.759830075724757e-05}
{"train_loss": 0.3897712528705597, "global_step": 27340, "epoch": 307, "lr": 9.759812323655498e-05}
{"train_loss": 0.3691646456718445, "global_step": 27341, "epoch": 307, "lr": 9.75979457094634e-05}
{"train_loss": 0.4466347098350525, "global_step": 27342, "epoch": 307, "lr": 9.759776817597286e-05}
{"train_loss": 0.4879894256591797, "global_step": 27343, "epoch": 307, "lr": 9.75975906360834e-05}
{"train_loss": 0.4012090265750885, "global_step": 27344, "epoch": 307, "lr": 9.7597413089795e-05}
{"train_loss": 0.444160133600235, "global_step": 27345, "epoch": 307, "lr": 9.759723553710772e-05}
{"train_loss": 0.33011120557785034, "global_step": 27346, "epoch": 307, "lr": 9.759705797802156e-05}
{"train_loss": 0.4475068747997284, "global_step": 27347, "epoch": 307, "lr": 9.759688041253657e-05}
{"train_loss": 0.49318352341651917, "global_step": 27348, "epoch": 307, "lr": 9.759670284065275e-05}
{"train_loss": 0.46534496545791626, "global_step": 27349, "epoch": 307, "lr": 9.759652526237014e-05}
{"train_loss": 0.4242672622203827, "global_step": 27350, "epoch": 307, "lr": 9.759634767768875e-05}
{"train_loss": 0.45216619968414307, "global_step": 27351, "epoch": 307, "lr": 9.759617008660862e-05}
{"train_loss": 0.41643667221069336, "global_step": 27352, "epoch": 307, "lr": 9.759599248912976e-05}
{"train_loss": 0.45156070590019226, "global_step": 27353, "epoch": 307, "lr": 9.75958148852522e-05}
{"train_loss": 0.452330082654953, "global_step": 27354, "epoch": 307, "lr": 9.759563727497596e-05}
{"train_loss": 0.3364027738571167, "global_step": 27355, "epoch": 307, "lr": 9.759545965830108e-05}
{"train_loss": 0.37386927008628845, "global_step": 27356, "epoch": 307, "lr": 9.759528203522756e-05}
{"train_loss": 0.38839733600616455, "global_step": 27357, "epoch": 307, "lr": 9.759510440575543e-05}
{"train_loss": 0.40662315487861633, "global_step": 27358, "epoch": 307, "lr": 9.759492676988473e-05}
{"train_loss": 0.44138747453689575, "global_step": 27359, "epoch": 307, "lr": 9.759474912761546e-05}
{"train_loss": 0.3551311194896698, "global_step": 27360, "epoch": 307, "lr": 9.759457147894766e-05}
{"train_loss": 0.35681456327438354, "global_step": 27361, "epoch": 307, "lr": 9.759439382388134e-05}
{"train_loss": 0.39715898036956787, "global_step": 27362, "epoch": 307, "lr": 9.759421616241655e-05}
{"train_loss": 0.4200958013534546, "global_step": 27363, "epoch": 307, "lr": 9.759403849455329e-05}
{"train_loss": 0.539513885974884, "global_step": 27364, "epoch": 307, "lr": 9.759386082029161e-05}
{"train_loss": 0.458219438791275, "global_step": 27365, "epoch": 307, "lr": 9.759368313963149e-05}
{"train_loss": 0.41028743982315063, "global_step": 27366, "epoch": 307, "lr": 9.759350545257298e-05}
{"train_loss": 0.2317056804895401, "global_step": 27367, "epoch": 307, "lr": 9.759332775911611e-05}
{"train_loss": 0.40612101554870605, "global_step": 27368, "epoch": 307, "lr": 9.75931500592609e-05}
{"train_loss": 0.4816746115684509, "global_step": 27369, "epoch": 307, "lr": 9.759297235300737e-05}
{"train_loss": 0.33289265632629395, "global_step": 27370, "epoch": 307, "lr": 9.759279464035556e-05}
{"train_loss": 0.3987623155117035, "global_step": 27371, "epoch": 307, "lr": 9.759261692130546e-05}
{"train_loss": 0.29683780670166016, "global_step": 27372, "epoch": 307, "lr": 9.759243919585712e-05}
{"train_loss": 0.3858690857887268, "global_step": 27373, "epoch": 307, "lr": 9.759226146401055e-05}
{"train_loss": 0.3862973153591156, "global_step": 27374, "epoch": 307, "lr": 9.759208372576578e-05}
{"train_loss": 0.5173362493515015, "global_step": 27375, "epoch": 307, "lr": 9.759190598112284e-05}
{"train_loss": 0.45271772146224976, "global_step": 27376, "epoch": 307, "lr": 9.759172823008174e-05}
{"train_loss": 0.43174734711647034, "global_step": 27377, "epoch": 307, "lr": 9.759155047264252e-05}
{"train_loss": 0.3985683023929596, "global_step": 27378, "epoch": 307, "lr": 9.75913727088052e-05}
{"train_loss": 0.4172128438949585, "global_step": 27379, "epoch": 307, "lr": 9.75911949385698e-05}
{"train_loss": 0.3929717242717743, "global_step": 27380, "epoch": 307, "lr": 9.759101716193634e-05}
{"train_loss": 0.38898056745529175, "global_step": 27381, "epoch": 307, "lr": 9.759083937890483e-05}
{"train_loss": 0.34648704528808594, "global_step": 27382, "epoch": 307, "lr": 9.759066158947533e-05}
{"train_loss": 0.49406126141548157, "global_step": 27383, "epoch": 307, "lr": 9.759048379364786e-05}
{"train_loss": 0.389758437871933, "global_step": 27384, "epoch": 307, "lr": 9.75903059914224e-05}
{"train_loss": 0.4906855821609497, "global_step": 27385, "epoch": 307, "lr": 9.759012818279903e-05}
{"train_loss": 0.4489784240722656, "global_step": 27386, "epoch": 307, "lr": 9.758995036777772e-05}
{"train_loss": 0.36288338899612427, "global_step": 27387, "epoch": 307, "lr": 9.758977254635855e-05}
{"train_loss": 0.35292384028434753, "global_step": 27388, "epoch": 307, "lr": 9.75895947185415e-05}
{"train_loss": 0.4080645740032196, "global_step": 27389, "epoch": 307, "lr": 9.758941688432661e-05}
{"train_loss": 0.4271443784236908, "global_step": 27390, "epoch": 307, "lr": 9.75892390437139e-05}
{"train_loss": 0.35873913764953613, "global_step": 27391, "epoch": 307, "lr": 9.758906119670339e-05}
{"train_loss": 0.4168678820133209, "global_step": 27392, "epoch": 307, "lr": 9.758888334329512e-05}
{"train_loss": 0.39503341913223267, "global_step": 27393, "epoch": 307, "lr": 9.75887054834891e-05}
{"train_loss": 0.6045358777046204, "global_step": 27394, "epoch": 307, "lr": 9.758852761728538e-05}
{"train_loss": 0.3578563332557678, "global_step": 27395, "epoch": 307, "lr": 9.758834974468394e-05}
{"train_loss": 0.32450491189956665, "global_step": 27396, "epoch": 307, "lr": 9.758817186568483e-05}
{"train_loss": 0.3922938108444214, "global_step": 27397, "epoch": 307, "lr": 9.758799398028808e-05}
{"train_loss": 0.3895276188850403, "global_step": 27398, "epoch": 307, "lr": 9.758781608849368e-05}
{"train_loss": 0.3710933029651642, "global_step": 27399, "epoch": 307, "lr": 9.75876381903017e-05}
{"train_loss": 0.4142701029777527, "global_step": 27400, "epoch": 307, "lr": 9.758746028571215e-05}
{"train_loss": 0.3601032793521881, "global_step": 27401, "epoch": 307, "lr": 9.758728237472503e-05}
{"train_loss": 0.5388228297233582, "global_step": 27402, "epoch": 307, "lr": 9.758710445734039e-05}
{"train_loss": 0.49324533343315125, "global_step": 27403, "epoch": 307, "lr": 9.758692653355823e-05}
{"train_loss": 0.39235392212867737, "global_step": 27404, "epoch": 307, "lr": 9.75867486033786e-05}
{"train_loss": 0.4018116295337677, "global_step": 27405, "epoch": 307, "lr": 9.75865706668015e-05}
{"train_loss": 0.5335463881492615, "global_step": 27406, "epoch": 307, "lr": 9.758639272382697e-05}
{"train_loss": 0.3563334345817566, "global_step": 27407, "epoch": 307, "lr": 9.758621477445503e-05}
{"train_loss": 0.38900941610336304, "global_step": 27408, "epoch": 307, "lr": 9.75860368186857e-05}
{"train_loss": 0.6073705554008484, "global_step": 27409, "epoch": 307, "lr": 9.758585885651901e-05}
{"train_loss": 0.4339084327220917, "global_step": 27410, "epoch": 307, "lr": 9.7585680887955e-05}
{"train_loss": 0.4170810431241989, "global_step": 27411, "epoch": 307, "lr": 9.758550291299365e-05, "val_loss": 2.124603748321533}
{"train_loss": 0.4777238965034485, "global_step": 27412, "epoch": 308, "lr": 9.758532493163502e-05}
{"train_loss": 0.5235022306442261, "global_step": 27413, "epoch": 308, "lr": 9.758514694387911e-05}
{"train_loss": 0.4094254672527313, "global_step": 27414, "epoch": 308, "lr": 9.758496894972598e-05}
{"train_loss": 0.31169816851615906, "global_step": 27415, "epoch": 308, "lr": 9.758479094917561e-05}
{"train_loss": 0.48755350708961487, "global_step": 27416, "epoch": 308, "lr": 9.758461294222805e-05}
{"train_loss": 0.3947496712207794, "global_step": 27417, "epoch": 308, "lr": 9.758443492888333e-05}
{"train_loss": 0.37034347653388977, "global_step": 27418, "epoch": 308, "lr": 9.758425690914145e-05}
{"train_loss": 0.4572375416755676, "global_step": 27419, "epoch": 308, "lr": 9.758407888300244e-05}
{"train_loss": 0.4706209599971771, "global_step": 27420, "epoch": 308, "lr": 9.758390085046636e-05}
{"train_loss": 0.545261800289154, "global_step": 27421, "epoch": 308, "lr": 9.758372281153318e-05}
{"train_loss": 0.45425698161125183, "global_step": 27422, "epoch": 308, "lr": 9.758354476620295e-05}
{"train_loss": 0.4112684428691864, "global_step": 27423, "epoch": 308, "lr": 9.75833667144757e-05}
{"train_loss": 0.3956785500049591, "global_step": 27424, "epoch": 308, "lr": 9.758318865635144e-05}
{"train_loss": 0.2945297360420227, "global_step": 27425, "epoch": 308, "lr": 9.758301059183022e-05}
{"train_loss": 0.3779027462005615, "global_step": 27426, "epoch": 308, "lr": 9.758283252091202e-05}
{"train_loss": 0.4027484655380249, "global_step": 27427, "epoch": 308, "lr": 9.758265444359689e-05}
{"train_loss": 0.3268306255340576, "global_step": 27428, "epoch": 308, "lr": 9.758247635988486e-05}
{"train_loss": 0.35120704770088196, "global_step": 27429, "epoch": 308, "lr": 9.758229826977595e-05}
{"train_loss": 0.34281983971595764, "global_step": 27430, "epoch": 308, "lr": 9.758212017327018e-05}
{"train_loss": 0.32326802611351013, "global_step": 27431, "epoch": 308, "lr": 9.758194207036754e-05}
{"train_loss": 0.32188746333122253, "global_step": 27432, "epoch": 308, "lr": 9.758176396106813e-05}
{"train_loss": 0.46221286058425903, "global_step": 27433, "epoch": 308, "lr": 9.758158584537192e-05}
{"train_loss": 0.47364890575408936, "global_step": 27434, "epoch": 308, "lr": 9.758140772327895e-05}
{"train_loss": 0.41005009412765503, "global_step": 27435, "epoch": 308, "lr": 9.758122959478923e-05}
{"train_loss": 0.3433930575847626, "global_step": 27436, "epoch": 308, "lr": 9.75810514599028e-05}
{"train_loss": 0.36426690220832825, "global_step": 27437, "epoch": 308, "lr": 9.758087331861968e-05}
{"train_loss": 0.4341619610786438, "global_step": 27438, "epoch": 308, "lr": 9.758069517093987e-05}
{"train_loss": 0.2815028727054596, "global_step": 27439, "epoch": 308, "lr": 9.758051701686344e-05}
{"train_loss": 0.38513243198394775, "global_step": 27440, "epoch": 308, "lr": 9.758033885639037e-05}
{"train_loss": 0.3356943726539612, "global_step": 27441, "epoch": 308, "lr": 9.758016068952071e-05}
{"train_loss": 0.29650092124938965, "global_step": 27442, "epoch": 308, "lr": 9.757998251625448e-05}
{"train_loss": 0.40873903036117554, "global_step": 27443, "epoch": 308, "lr": 9.757980433659171e-05}
{"train_loss": 0.4073255658149719, "global_step": 27444, "epoch": 308, "lr": 9.75796261505324e-05}
{"train_loss": 0.4859949052333832, "global_step": 27445, "epoch": 308, "lr": 9.75794479580766e-05}
{"train_loss": 0.3531769812107086, "global_step": 27446, "epoch": 308, "lr": 9.757926975922433e-05}
{"train_loss": 0.44643184542655945, "global_step": 27447, "epoch": 308, "lr": 9.75790915539756e-05}
{"train_loss": 0.3864709734916687, "global_step": 27448, "epoch": 308, "lr": 9.757891334233043e-05}
{"train_loss": 0.4697020649909973, "global_step": 27449, "epoch": 308, "lr": 9.757873512428885e-05}
{"train_loss": 0.3624737858772278, "global_step": 27450, "epoch": 308, "lr": 9.757855689985092e-05}
{"train_loss": 0.3625568151473999, "global_step": 27451, "epoch": 308, "lr": 9.757837866901661e-05}
{"train_loss": 0.38771289587020874, "global_step": 27452, "epoch": 308, "lr": 9.757820043178598e-05}
{"train_loss": 0.5780377388000488, "global_step": 27453, "epoch": 308, "lr": 9.757802218815902e-05}
{"train_loss": 0.3098568320274353, "global_step": 27454, "epoch": 308, "lr": 9.75778439381358e-05}
{"train_loss": 0.3585907816886902, "global_step": 27455, "epoch": 308, "lr": 9.75776656817163e-05}
{"train_loss": 0.39971724152565, "global_step": 27456, "epoch": 308, "lr": 9.757748741890057e-05}
{"train_loss": 0.38896483182907104, "global_step": 27457, "epoch": 308, "lr": 9.757730914968863e-05}
{"train_loss": 0.360882043838501, "global_step": 27458, "epoch": 308, "lr": 9.757713087408048e-05}
{"train_loss": 0.4304856061935425, "global_step": 27459, "epoch": 308, "lr": 9.75769525920762e-05}
{"train_loss": 0.33182626962661743, "global_step": 27460, "epoch": 308, "lr": 9.757677430367576e-05}
{"train_loss": 0.3962158262729645, "global_step": 27461, "epoch": 308, "lr": 9.75765960088792e-05}
{"train_loss": 0.45562800765037537, "global_step": 27462, "epoch": 308, "lr": 9.757641770768654e-05}
{"train_loss": 0.3077702522277832, "global_step": 27463, "epoch": 308, "lr": 9.757623940009781e-05}
{"train_loss": 0.47445976734161377, "global_step": 27464, "epoch": 308, "lr": 9.757606108611306e-05}
{"train_loss": 0.31589922308921814, "global_step": 27465, "epoch": 308, "lr": 9.757588276573226e-05}
{"train_loss": 0.40541303157806396, "global_step": 27466, "epoch": 308, "lr": 9.757570443895548e-05}
{"train_loss": 0.36323297023773193, "global_step": 27467, "epoch": 308, "lr": 9.757552610578272e-05}
{"train_loss": 0.5184902548789978, "global_step": 27468, "epoch": 308, "lr": 9.757534776621402e-05}
{"train_loss": 0.4013301432132721, "global_step": 27469, "epoch": 308, "lr": 9.757516942024938e-05}
{"train_loss": 0.4539715051651001, "global_step": 27470, "epoch": 308, "lr": 9.757499106788885e-05}
{"train_loss": 0.40906259417533875, "global_step": 27471, "epoch": 308, "lr": 9.757481270913243e-05}
{"train_loss": 0.6421845555305481, "global_step": 27472, "epoch": 308, "lr": 9.757463434398016e-05}
{"train_loss": 0.4152792990207672, "global_step": 27473, "epoch": 308, "lr": 9.757445597243206e-05}
{"train_loss": 0.48328691720962524, "global_step": 27474, "epoch": 308, "lr": 9.757427759448817e-05}
{"train_loss": 0.4208873212337494, "global_step": 27475, "epoch": 308, "lr": 9.757409921014849e-05}
{"train_loss": 0.49010205268859863, "global_step": 27476, "epoch": 308, "lr": 9.757392081941304e-05}
{"train_loss": 0.32016685605049133, "global_step": 27477, "epoch": 308, "lr": 9.757374242228185e-05}
{"train_loss": 0.468305766582489, "global_step": 27478, "epoch": 308, "lr": 9.757356401875497e-05}
{"train_loss": 0.48690563440322876, "global_step": 27479, "epoch": 308, "lr": 9.75733856088324e-05}
{"train_loss": 0.41336479783058167, "global_step": 27480, "epoch": 308, "lr": 9.757320719251417e-05}
{"train_loss": 0.4468889832496643, "global_step": 27481, "epoch": 308, "lr": 9.75730287698003e-05}
{"train_loss": 0.5585553050041199, "global_step": 27482, "epoch": 308, "lr": 9.757285034069081e-05}
{"train_loss": 0.4267970323562622, "global_step": 27483, "epoch": 308, "lr": 9.757267190518573e-05}
{"train_loss": 0.4891553223133087, "global_step": 27484, "epoch": 308, "lr": 9.757249346328509e-05}
{"train_loss": 0.34402287006378174, "global_step": 27485, "epoch": 308, "lr": 9.757231501498891e-05}
{"train_loss": 0.4228152632713318, "global_step": 27486, "epoch": 308, "lr": 9.757213656029722e-05}
{"train_loss": 0.3681788444519043, "global_step": 27487, "epoch": 308, "lr": 9.757195809921001e-05}
{"train_loss": 0.43087631464004517, "global_step": 27488, "epoch": 308, "lr": 9.757177963172735e-05}
{"train_loss": 0.31188085675239563, "global_step": 27489, "epoch": 308, "lr": 9.757160115784924e-05}
{"train_loss": 0.3318949043750763, "global_step": 27490, "epoch": 308, "lr": 9.757142267757571e-05}
{"train_loss": 0.6038942337036133, "global_step": 27491, "epoch": 308, "lr": 9.757124419090677e-05}
{"train_loss": 0.4819193482398987, "global_step": 27492, "epoch": 308, "lr": 9.757106569784246e-05}
{"train_loss": 0.33818361163139343, "global_step": 27493, "epoch": 308, "lr": 9.757088719838281e-05}
{"train_loss": 0.4166228473186493, "global_step": 27494, "epoch": 308, "lr": 9.757070869252783e-05}
{"train_loss": 0.41357553005218506, "global_step": 27495, "epoch": 308, "lr": 9.757053018027756e-05}
{"train_loss": 0.44519147276878357, "global_step": 27496, "epoch": 308, "lr": 9.7570351661632e-05}
{"train_loss": 0.49416717886924744, "global_step": 27497, "epoch": 308, "lr": 9.757017313659118e-05}
{"train_loss": 0.49544674158096313, "global_step": 27498, "epoch": 308, "lr": 9.756999460515514e-05}
{"train_loss": 0.41081294417381287, "global_step": 27499, "epoch": 308, "lr": 9.756981606732388e-05}
{"train_loss": 0.4138484932063671, "global_step": 27500, "epoch": 308, "lr": 9.756963752309745e-05, "val_loss": 2.1217641830444336}
{"train_loss": 0.5292582511901855, "global_step": 27501, "epoch": 309, "lr": 9.756945897247586e-05}
{"train_loss": 0.4982426166534424, "global_step": 27502, "epoch": 309, "lr": 9.756928041545914e-05}
{"train_loss": 0.40070170164108276, "global_step": 27503, "epoch": 309, "lr": 9.75691018520473e-05}
{"train_loss": 0.31540852785110474, "global_step": 27504, "epoch": 309, "lr": 9.756892328224039e-05}
{"train_loss": 0.40501341223716736, "global_step": 27505, "epoch": 309, "lr": 9.756874470603841e-05}
{"train_loss": 0.46839261054992676, "global_step": 27506, "epoch": 309, "lr": 9.756856612344139e-05}
{"train_loss": 0.3895436227321625, "global_step": 27507, "epoch": 309, "lr": 9.756838753444934e-05}
{"train_loss": 0.45848891139030457, "global_step": 27508, "epoch": 309, "lr": 9.756820893906231e-05}
{"train_loss": 0.434042751789093, "global_step": 27509, "epoch": 309, "lr": 9.756803033728032e-05}
{"train_loss": 0.5225419998168945, "global_step": 27510, "epoch": 309, "lr": 9.75678517291034e-05}
{"train_loss": 0.440213143825531, "global_step": 27511, "epoch": 309, "lr": 9.756767311453154e-05}
{"train_loss": 0.41406598687171936, "global_step": 27512, "epoch": 309, "lr": 9.75674944935648e-05}
{"train_loss": 0.3321240246295929, "global_step": 27513, "epoch": 309, "lr": 9.756731586620318e-05}
{"train_loss": 0.44661733508110046, "global_step": 27514, "epoch": 309, "lr": 9.756713723244673e-05}
{"train_loss": 0.4368245005607605, "global_step": 27515, "epoch": 309, "lr": 9.756695859229546e-05}
{"train_loss": 0.4353383481502533, "global_step": 27516, "epoch": 309, "lr": 9.756677994574936e-05}
{"train_loss": 0.3598844110965729, "global_step": 27517, "epoch": 309, "lr": 9.756660129280852e-05}
{"train_loss": 0.38943037390708923, "global_step": 27518, "epoch": 309, "lr": 9.75664226334729e-05}
{"train_loss": 0.41163647174835205, "global_step": 27519, "epoch": 309, "lr": 9.756624396774258e-05}
{"train_loss": 0.41277068853378296, "global_step": 27520, "epoch": 309, "lr": 9.756606529561755e-05}
{"train_loss": 0.4019482731819153, "global_step": 27521, "epoch": 309, "lr": 9.756588661709785e-05}
{"train_loss": 0.5072926878929138, "global_step": 27522, "epoch": 309, "lr": 9.756570793218349e-05}
{"train_loss": 0.4575963318347931, "global_step": 27523, "epoch": 309, "lr": 9.756552924087449e-05}
{"train_loss": 0.5192155838012695, "global_step": 27524, "epoch": 309, "lr": 9.75653505431709e-05}
{"train_loss": 0.46816328167915344, "global_step": 27525, "epoch": 309, "lr": 9.756517183907272e-05}
{"train_loss": 0.3589114248752594, "global_step": 27526, "epoch": 309, "lr": 9.756499312857999e-05}
{"train_loss": 0.482560396194458, "global_step": 27527, "epoch": 309, "lr": 9.756481441169272e-05}
{"train_loss": 0.45273908972740173, "global_step": 27528, "epoch": 309, "lr": 9.756463568841093e-05}
{"train_loss": 0.4488510489463806, "global_step": 27529, "epoch": 309, "lr": 9.756445695873466e-05}
{"train_loss": 0.42201486229896545, "global_step": 27530, "epoch": 309, "lr": 9.756427822266395e-05}
{"train_loss": 0.44230878353118896, "global_step": 27531, "epoch": 309, "lr": 9.756409948019878e-05}
{"train_loss": 0.426713764667511, "global_step": 27532, "epoch": 309, "lr": 9.75639207313392e-05}
{"train_loss": 0.3668042719364166, "global_step": 27533, "epoch": 309, "lr": 9.756374197608523e-05}
{"train_loss": 0.3997558355331421, "global_step": 27534, "epoch": 309, "lr": 9.75635632144369e-05}
{"train_loss": 0.4616153836250305, "global_step": 27535, "epoch": 309, "lr": 9.756338444639422e-05}
{"train_loss": 0.3778131604194641, "global_step": 27536, "epoch": 309, "lr": 9.756320567195724e-05}
{"train_loss": 0.42545974254608154, "global_step": 27537, "epoch": 309, "lr": 9.756302689112595e-05}
{"train_loss": 0.5085523724555969, "global_step": 27538, "epoch": 309, "lr": 9.75628481039004e-05}
{"train_loss": 0.4054024815559387, "global_step": 27539, "epoch": 309, "lr": 9.756266931028059e-05}
{"train_loss": 0.47094428539276123, "global_step": 27540, "epoch": 309, "lr": 9.756249051026659e-05}
{"train_loss": 0.43558183312416077, "global_step": 27541, "epoch": 309, "lr": 9.756231170385836e-05}
{"train_loss": 0.39050352573394775, "global_step": 27542, "epoch": 309, "lr": 9.756213289105597e-05}
{"train_loss": 0.36704403162002563, "global_step": 27543, "epoch": 309, "lr": 9.756195407185944e-05}
{"train_loss": 0.4707808494567871, "global_step": 27544, "epoch": 309, "lr": 9.756177524626877e-05}
{"train_loss": 0.3957313299179077, "global_step": 27545, "epoch": 309, "lr": 9.756159641428399e-05}
{"train_loss": 0.5115271806716919, "global_step": 27546, "epoch": 309, "lr": 9.756141757590515e-05}
{"train_loss": 0.3303479254245758, "global_step": 27547, "epoch": 309, "lr": 9.756123873113227e-05}
{"train_loss": 0.4762848913669586, "global_step": 27548, "epoch": 309, "lr": 9.756105987996535e-05}
{"train_loss": 0.4941423237323761, "global_step": 27549, "epoch": 309, "lr": 9.756088102240441e-05}
{"train_loss": 0.3080190122127533, "global_step": 27550, "epoch": 309, "lr": 9.75607021584495e-05}
{"train_loss": 0.45740464329719543, "global_step": 27551, "epoch": 309, "lr": 9.756052328810063e-05}
{"train_loss": 0.38061267137527466, "global_step": 27552, "epoch": 309, "lr": 9.756034441135784e-05}
{"train_loss": 0.33272796869277954, "global_step": 27553, "epoch": 309, "lr": 9.756016552822114e-05}
{"train_loss": 0.5302911400794983, "global_step": 27554, "epoch": 309, "lr": 9.755998663869054e-05}
{"train_loss": 0.379384309053421, "global_step": 27555, "epoch": 309, "lr": 9.755980774276608e-05}
{"train_loss": 0.34932172298431396, "global_step": 27556, "epoch": 309, "lr": 9.75596288404478e-05}
{"train_loss": 0.3352358937263489, "global_step": 27557, "epoch": 309, "lr": 9.755944993173569e-05}
{"train_loss": 0.3827603757381439, "global_step": 27558, "epoch": 309, "lr": 9.755927101662981e-05}
{"train_loss": 0.44172292947769165, "global_step": 27559, "epoch": 309, "lr": 9.755909209513014e-05}
{"train_loss": 0.3794513940811157, "global_step": 27560, "epoch": 309, "lr": 9.755891316723675e-05}
{"train_loss": 0.4116549491882324, "global_step": 27561, "epoch": 309, "lr": 9.755873423294964e-05}
{"train_loss": 0.4453948438167572, "global_step": 27562, "epoch": 309, "lr": 9.755855529226885e-05}
{"train_loss": 0.5325088500976562, "global_step": 27563, "epoch": 309, "lr": 9.755837634519436e-05}
{"train_loss": 0.4152354896068573, "global_step": 27564, "epoch": 309, "lr": 9.755819739172626e-05}
{"train_loss": 0.47574424743652344, "global_step": 27565, "epoch": 309, "lr": 9.755801843186452e-05}
{"train_loss": 0.4817008972167969, "global_step": 27566, "epoch": 309, "lr": 9.75578394656092e-05}
{"train_loss": 0.45871326327323914, "global_step": 27567, "epoch": 309, "lr": 9.75576604929603e-05}
{"train_loss": 0.42066118121147156, "global_step": 27568, "epoch": 309, "lr": 9.755748151391784e-05}
{"train_loss": 0.4599965810775757, "global_step": 27569, "epoch": 309, "lr": 9.755730252848187e-05}
{"train_loss": 0.36866894364356995, "global_step": 27570, "epoch": 309, "lr": 9.755712353665239e-05}
{"train_loss": 0.3515750765800476, "global_step": 27571, "epoch": 309, "lr": 9.755694453842944e-05}
{"train_loss": 0.41509872674942017, "global_step": 27572, "epoch": 309, "lr": 9.755676553381303e-05}
{"train_loss": 0.4474930465221405, "global_step": 27573, "epoch": 309, "lr": 9.755658652280321e-05}
{"train_loss": 0.40070149302482605, "global_step": 27574, "epoch": 309, "lr": 9.755640750539999e-05}
{"train_loss": 0.3847857713699341, "global_step": 27575, "epoch": 309, "lr": 9.755622848160337e-05}
{"train_loss": 0.4049203097820282, "global_step": 27576, "epoch": 309, "lr": 9.75560494514134e-05}
{"train_loss": 0.4037822484970093, "global_step": 27577, "epoch": 309, "lr": 9.75558704148301e-05}
{"train_loss": 0.3005782961845398, "global_step": 27578, "epoch": 309, "lr": 9.75556913718535e-05}
{"train_loss": 0.5225797891616821, "global_step": 27579, "epoch": 309, "lr": 9.755551232248361e-05}
{"train_loss": 0.362397700548172, "global_step": 27580, "epoch": 309, "lr": 9.755533326672046e-05}
{"train_loss": 0.35107311606407166, "global_step": 27581, "epoch": 309, "lr": 9.755515420456407e-05}
{"train_loss": 0.4558701515197754, "global_step": 27582, "epoch": 309, "lr": 9.755497513601448e-05}
{"train_loss": 0.42222198843955994, "global_step": 27583, "epoch": 309, "lr": 9.75547960610717e-05}
{"train_loss": 0.45276105403900146, "global_step": 27584, "epoch": 309, "lr": 9.755461697973576e-05}
{"train_loss": 0.4827897846698761, "global_step": 27585, "epoch": 309, "lr": 9.755443789200668e-05}
{"train_loss": 0.28643086552619934, "global_step": 27586, "epoch": 309, "lr": 9.755425879788448e-05}
{"train_loss": 0.4840243458747864, "global_step": 27587, "epoch": 309, "lr": 9.755407969736921e-05}
{"train_loss": 0.36838287115097046, "global_step": 27588, "epoch": 309, "lr": 9.755390059046084e-05}
{"train_loss": 0.42273994041292856, "global_step": 27589, "epoch": 309, "lr": 9.755372147715945e-05, "val_loss": 2.1446309089660645}
{"train_loss": 0.2832973301410675, "global_step": 27590, "epoch": 310, "lr": 9.755354235746503e-05}
{"train_loss": 0.34069520235061646, "global_step": 27591, "epoch": 310, "lr": 9.755336323137763e-05}
{"train_loss": 0.4563651382923126, "global_step": 27592, "epoch": 310, "lr": 9.755318409889725e-05}
{"train_loss": 0.4217994213104248, "global_step": 27593, "epoch": 310, "lr": 9.755300496002393e-05}
{"train_loss": 0.3543300926685333, "global_step": 27594, "epoch": 310, "lr": 9.755282581475769e-05}
{"train_loss": 0.41180726885795593, "global_step": 27595, "epoch": 310, "lr": 9.755264666309853e-05}
{"train_loss": 0.32585644721984863, "global_step": 27596, "epoch": 310, "lr": 9.755246750504652e-05}
{"train_loss": 0.41738590598106384, "global_step": 27597, "epoch": 310, "lr": 9.755228834060166e-05}
{"train_loss": 0.4618765115737915, "global_step": 27598, "epoch": 310, "lr": 9.755210916976396e-05}
{"train_loss": 0.40613314509391785, "global_step": 27599, "epoch": 310, "lr": 9.755192999253346e-05}
{"train_loss": 0.433116614818573, "global_step": 27600, "epoch": 310, "lr": 9.755175080891019e-05}
{"train_loss": 0.5053519010543823, "global_step": 27601, "epoch": 310, "lr": 9.755157161889416e-05}
{"train_loss": 0.3998126983642578, "global_step": 27602, "epoch": 310, "lr": 9.755139242248542e-05}
{"train_loss": 0.35084500908851624, "global_step": 27603, "epoch": 310, "lr": 9.755121321968396e-05}
{"train_loss": 0.3682555854320526, "global_step": 27604, "epoch": 310, "lr": 9.755103401048982e-05}
{"train_loss": 0.32757899165153503, "global_step": 27605, "epoch": 310, "lr": 9.755085479490302e-05}
{"train_loss": 0.302862286567688, "global_step": 27606, "epoch": 310, "lr": 9.755067557292358e-05}
{"train_loss": 0.4834308624267578, "global_step": 27607, "epoch": 310, "lr": 9.755049634455156e-05}
{"train_loss": 0.34307432174682617, "global_step": 27608, "epoch": 310, "lr": 9.755031710978693e-05}
{"train_loss": 0.46435269713401794, "global_step": 27609, "epoch": 310, "lr": 9.755013786862975e-05}
{"train_loss": 0.5066258311271667, "global_step": 27610, "epoch": 310, "lr": 9.754995862108003e-05}
{"train_loss": 0.45947957038879395, "global_step": 27611, "epoch": 310, "lr": 9.75497793671378e-05}
{"train_loss": 0.4303934574127197, "global_step": 27612, "epoch": 310, "lr": 9.754960010680307e-05}
{"train_loss": 0.37557336688041687, "global_step": 27613, "epoch": 310, "lr": 9.754942084007589e-05}
{"train_loss": 0.43128570914268494, "global_step": 27614, "epoch": 310, "lr": 9.754924156695627e-05}
{"train_loss": 0.44413894414901733, "global_step": 27615, "epoch": 310, "lr": 9.754906228744423e-05}
{"train_loss": 0.5117388963699341, "global_step": 27616, "epoch": 310, "lr": 9.754888300153979e-05}
{"train_loss": 0.3855019807815552, "global_step": 27617, "epoch": 310, "lr": 9.754870370924299e-05}
{"train_loss": 0.3101451098918915, "global_step": 27618, "epoch": 310, "lr": 9.754852441055385e-05}
{"train_loss": 0.486396461725235, "global_step": 27619, "epoch": 310, "lr": 9.754834510547238e-05}
{"train_loss": 0.4274141788482666, "global_step": 27620, "epoch": 310, "lr": 9.754816579399862e-05}
{"train_loss": 0.36527132987976074, "global_step": 27621, "epoch": 310, "lr": 9.754798647613259e-05}
{"train_loss": 0.45448964834213257, "global_step": 27622, "epoch": 310, "lr": 9.75478071518743e-05}
{"train_loss": 0.48395514488220215, "global_step": 27623, "epoch": 310, "lr": 9.754762782122381e-05}
{"train_loss": 0.3571520149707794, "global_step": 27624, "epoch": 310, "lr": 9.754744848418111e-05}
{"train_loss": 0.4167688190937042, "global_step": 27625, "epoch": 310, "lr": 9.754726914074624e-05}
{"train_loss": 0.47180089354515076, "global_step": 27626, "epoch": 310, "lr": 9.754708979091921e-05}
{"train_loss": 0.4572916030883789, "global_step": 27627, "epoch": 310, "lr": 9.754691043470006e-05}
{"train_loss": 0.4754371643066406, "global_step": 27628, "epoch": 310, "lr": 9.75467310720888e-05}
{"train_loss": 0.4146350622177124, "global_step": 27629, "epoch": 310, "lr": 9.754655170308547e-05}
{"train_loss": 0.2628287076950073, "global_step": 27630, "epoch": 310, "lr": 9.754637232769009e-05}
{"train_loss": 0.3625375032424927, "global_step": 27631, "epoch": 310, "lr": 9.754619294590265e-05}
{"train_loss": 0.4209875166416168, "global_step": 27632, "epoch": 310, "lr": 9.754601355772324e-05}
{"train_loss": 0.3698374629020691, "global_step": 27633, "epoch": 310, "lr": 9.754583416315183e-05}
{"train_loss": 0.4711622893810272, "global_step": 27634, "epoch": 310, "lr": 9.754565476218847e-05}
{"train_loss": 0.411407470703125, "global_step": 27635, "epoch": 310, "lr": 9.754547535483316e-05}
{"train_loss": 0.3405747413635254, "global_step": 27636, "epoch": 310, "lr": 9.754529594108596e-05}
{"train_loss": 0.3659082353115082, "global_step": 27637, "epoch": 310, "lr": 9.754511652094686e-05}
{"train_loss": 0.4163658618927002, "global_step": 27638, "epoch": 310, "lr": 9.754493709441591e-05}
{"train_loss": 0.3572433590888977, "global_step": 27639, "epoch": 310, "lr": 9.754475766149312e-05}
{"train_loss": 0.4649757146835327, "global_step": 27640, "epoch": 310, "lr": 9.754457822217851e-05}
{"train_loss": 0.4973558485507965, "global_step": 27641, "epoch": 310, "lr": 9.754439877647213e-05}
{"train_loss": 0.416801780462265, "global_step": 27642, "epoch": 310, "lr": 9.754421932437397e-05}
{"train_loss": 0.31127339601516724, "global_step": 27643, "epoch": 310, "lr": 9.754403986588404e-05}
{"train_loss": 0.3965494930744171, "global_step": 27644, "epoch": 310, "lr": 9.754386040100244e-05}
{"train_loss": 0.41588717699050903, "global_step": 27645, "epoch": 310, "lr": 9.754368092972913e-05}
{"train_loss": 0.30203038454055786, "global_step": 27646, "epoch": 310, "lr": 9.754350145206414e-05}
{"train_loss": 0.31423479318618774, "global_step": 27647, "epoch": 310, "lr": 9.754332196800752e-05}
{"train_loss": 0.4185248613357544, "global_step": 27648, "epoch": 310, "lr": 9.754314247755929e-05}
{"train_loss": 0.3418273627758026, "global_step": 27649, "epoch": 310, "lr": 9.754296298071945e-05}
{"train_loss": 0.31735941767692566, "global_step": 27650, "epoch": 310, "lr": 9.754278347748804e-05}
{"train_loss": 0.41042232513427734, "global_step": 27651, "epoch": 310, "lr": 9.754260396786508e-05}
{"train_loss": 0.3490180969238281, "global_step": 27652, "epoch": 310, "lr": 9.75424244518506e-05}
{"train_loss": 0.4115550220012665, "global_step": 27653, "epoch": 310, "lr": 9.754224492944461e-05}
{"train_loss": 0.3551141023635864, "global_step": 27654, "epoch": 310, "lr": 9.754206540064716e-05}
{"train_loss": 0.3067685067653656, "global_step": 27655, "epoch": 310, "lr": 9.754188586545825e-05}
{"train_loss": 0.4344606399536133, "global_step": 27656, "epoch": 310, "lr": 9.754170632387792e-05}
{"train_loss": 0.4427401125431061, "global_step": 27657, "epoch": 310, "lr": 9.754152677590617e-05}
{"train_loss": 0.42722517251968384, "global_step": 27658, "epoch": 310, "lr": 9.754134722154307e-05}
{"train_loss": 0.30692407488822937, "global_step": 27659, "epoch": 310, "lr": 9.75411676607886e-05}
{"train_loss": 0.4257281720638275, "global_step": 27660, "epoch": 310, "lr": 9.754098809364279e-05}
{"train_loss": 0.3747160732746124, "global_step": 27661, "epoch": 310, "lr": 9.754080852010568e-05}
{"train_loss": 0.4300419092178345, "global_step": 27662, "epoch": 310, "lr": 9.754062894017729e-05}
{"train_loss": 0.4026903212070465, "global_step": 27663, "epoch": 310, "lr": 9.754044935385765e-05}
{"train_loss": 0.3348565399646759, "global_step": 27664, "epoch": 310, "lr": 9.754026976114676e-05}
{"train_loss": 0.5238487124443054, "global_step": 27665, "epoch": 310, "lr": 9.754009016204466e-05}
{"train_loss": 0.39116308093070984, "global_step": 27666, "epoch": 310, "lr": 9.753991055655139e-05}
{"train_loss": 0.3809157609939575, "global_step": 27667, "epoch": 310, "lr": 9.753973094466695e-05}
{"train_loss": 0.4213429093360901, "global_step": 27668, "epoch": 310, "lr": 9.753955132639136e-05}
{"train_loss": 0.40563103556632996, "global_step": 27669, "epoch": 310, "lr": 9.753937170172469e-05}
{"train_loss": 0.45476508140563965, "global_step": 27670, "epoch": 310, "lr": 9.753919207066691e-05}
{"train_loss": 0.39464038610458374, "global_step": 27671, "epoch": 310, "lr": 9.753901243321806e-05}
{"train_loss": 0.42867130041122437, "global_step": 27672, "epoch": 310, "lr": 9.753883278937816e-05}
{"train_loss": 0.47336673736572266, "global_step": 27673, "epoch": 310, "lr": 9.753865313914727e-05}
{"train_loss": 0.3697893023490906, "global_step": 27674, "epoch": 310, "lr": 9.753847348252538e-05}
{"train_loss": 0.39299044013023376, "global_step": 27675, "epoch": 310, "lr": 9.753829381951252e-05}
{"train_loss": 0.4194692373275757, "global_step": 27676, "epoch": 310, "lr": 9.75381141501087e-05}
{"train_loss": 0.4276805520057678, "global_step": 27677, "epoch": 310, "lr": 9.753793447431398e-05}
{"train_loss": 0.4031906713930409, "global_step": 27678, "epoch": 310, "lr": 9.753775479212836e-05, "val_loss": 2.2489264011383057, "train_action_mse_error": 29.83648681640625}
{"train_loss": 0.4183778762817383, "global_step": 27679, "epoch": 311, "lr": 9.753757510355186e-05}
{"train_loss": 0.3097861409187317, "global_step": 27680, "epoch": 311, "lr": 9.753739540858452e-05}
{"train_loss": 0.409089058637619, "global_step": 27681, "epoch": 311, "lr": 9.753721570722636e-05}
{"train_loss": 0.27254047989845276, "global_step": 27682, "epoch": 311, "lr": 9.753703599947739e-05}
{"train_loss": 0.3958517611026764, "global_step": 27683, "epoch": 311, "lr": 9.753685628533764e-05}
{"train_loss": 0.39939194917678833, "global_step": 27684, "epoch": 311, "lr": 9.753667656480716e-05}
{"train_loss": 0.42987537384033203, "global_step": 27685, "epoch": 311, "lr": 9.753649683788593e-05}
{"train_loss": 0.338656485080719, "global_step": 27686, "epoch": 311, "lr": 9.753631710457403e-05}
{"train_loss": 0.29590678215026855, "global_step": 27687, "epoch": 311, "lr": 9.753613736487141e-05}
{"train_loss": 0.4669567048549652, "global_step": 27688, "epoch": 311, "lr": 9.753595761877816e-05}
{"train_loss": 0.4805419445037842, "global_step": 27689, "epoch": 311, "lr": 9.753577786629429e-05}
{"train_loss": 0.41428446769714355, "global_step": 27690, "epoch": 311, "lr": 9.753559810741979e-05}
{"train_loss": 0.4125820994377136, "global_step": 27691, "epoch": 311, "lr": 9.753541834215472e-05}
{"train_loss": 0.4611738920211792, "global_step": 27692, "epoch": 311, "lr": 9.753523857049908e-05}
{"train_loss": 0.4003452658653259, "global_step": 27693, "epoch": 311, "lr": 9.753505879245293e-05}
{"train_loss": 0.41217756271362305, "global_step": 27694, "epoch": 311, "lr": 9.753487900801625e-05}
{"train_loss": 0.38479095697402954, "global_step": 27695, "epoch": 311, "lr": 9.75346992171891e-05}
{"train_loss": 0.35157322883605957, "global_step": 27696, "epoch": 311, "lr": 9.753451941997149e-05}
{"train_loss": 0.3778338134288788, "global_step": 27697, "epoch": 311, "lr": 9.753433961636344e-05}
{"train_loss": 0.4327768385410309, "global_step": 27698, "epoch": 311, "lr": 9.753415980636496e-05}
{"train_loss": 0.3997250199317932, "global_step": 27699, "epoch": 311, "lr": 9.753397998997611e-05}
{"train_loss": 0.4078773558139801, "global_step": 27700, "epoch": 311, "lr": 9.753380016719688e-05}
{"train_loss": 0.507843554019928, "global_step": 27701, "epoch": 311, "lr": 9.753362033802733e-05}
{"train_loss": 0.4111982583999634, "global_step": 27702, "epoch": 311, "lr": 9.753344050246747e-05}
{"train_loss": 0.40045443177223206, "global_step": 27703, "epoch": 311, "lr": 9.753326066051729e-05}
{"train_loss": 0.3212442100048065, "global_step": 27704, "epoch": 311, "lr": 9.753308081217686e-05}
{"train_loss": 0.4041644334793091, "global_step": 27705, "epoch": 311, "lr": 9.753290095744618e-05}
{"train_loss": 0.3264344334602356, "global_step": 27706, "epoch": 311, "lr": 9.753272109632527e-05}
{"train_loss": 0.4832911491394043, "global_step": 27707, "epoch": 311, "lr": 9.75325412288142e-05}
{"train_loss": 0.48365458846092224, "global_step": 27708, "epoch": 311, "lr": 9.753236135491293e-05}
{"train_loss": 0.48813608288764954, "global_step": 27709, "epoch": 311, "lr": 9.75321814746215e-05}
{"train_loss": 0.4040446877479553, "global_step": 27710, "epoch": 311, "lr": 9.753200158793999e-05}
{"train_loss": 0.36550867557525635, "global_step": 27711, "epoch": 311, "lr": 9.753182169486835e-05}
{"train_loss": 0.35535717010498047, "global_step": 27712, "epoch": 311, "lr": 9.753164179540665e-05}
{"train_loss": 0.4645397663116455, "global_step": 27713, "epoch": 311, "lr": 9.753146188955488e-05}
{"train_loss": 0.3467139005661011, "global_step": 27714, "epoch": 311, "lr": 9.753128197731312e-05}
{"train_loss": 0.3327259421348572, "global_step": 27715, "epoch": 311, "lr": 9.753110205868133e-05}
{"train_loss": 0.30809077620506287, "global_step": 27716, "epoch": 311, "lr": 9.753092213365957e-05}
{"train_loss": 0.4939561188220978, "global_step": 27717, "epoch": 311, "lr": 9.753074220224786e-05}
{"train_loss": 0.4644530415534973, "global_step": 27718, "epoch": 311, "lr": 9.753056226444621e-05}
{"train_loss": 0.33630651235580444, "global_step": 27719, "epoch": 311, "lr": 9.753038232025467e-05}
{"train_loss": 0.35433220863342285, "global_step": 27720, "epoch": 311, "lr": 9.753020236967325e-05}
{"train_loss": 0.4867042899131775, "global_step": 27721, "epoch": 311, "lr": 9.753002241270197e-05}
{"train_loss": 0.43523818254470825, "global_step": 27722, "epoch": 311, "lr": 9.752984244934085e-05}
{"train_loss": 0.3315272033214569, "global_step": 27723, "epoch": 311, "lr": 9.752966247958992e-05}
{"train_loss": 0.29139986634254456, "global_step": 27724, "epoch": 311, "lr": 9.752948250344922e-05}
{"train_loss": 0.4868781566619873, "global_step": 27725, "epoch": 311, "lr": 9.752930252091875e-05}
{"train_loss": 0.4599399268627167, "global_step": 27726, "epoch": 311, "lr": 9.752912253199855e-05}
{"train_loss": 0.4167099893093109, "global_step": 27727, "epoch": 311, "lr": 9.752894253668864e-05}
{"train_loss": 0.36490556597709656, "global_step": 27728, "epoch": 311, "lr": 9.752876253498904e-05}
{"train_loss": 0.46558719873428345, "global_step": 27729, "epoch": 311, "lr": 9.752858252689978e-05}
{"train_loss": 0.3862287402153015, "global_step": 27730, "epoch": 311, "lr": 9.752840251242087e-05}
{"train_loss": 0.40236371755599976, "global_step": 27731, "epoch": 311, "lr": 9.752822249155235e-05}
{"train_loss": 0.4603015184402466, "global_step": 27732, "epoch": 311, "lr": 9.752804246429425e-05}
{"train_loss": 0.37249913811683655, "global_step": 27733, "epoch": 311, "lr": 9.752786243064658e-05}
{"train_loss": 0.45190826058387756, "global_step": 27734, "epoch": 311, "lr": 9.752768239060937e-05}
{"train_loss": 0.4109455347061157, "global_step": 27735, "epoch": 311, "lr": 9.752750234418263e-05}
{"train_loss": 0.3867845833301544, "global_step": 27736, "epoch": 311, "lr": 9.752732229136641e-05}
{"train_loss": 0.4079069197177887, "global_step": 27737, "epoch": 311, "lr": 9.75271422321607e-05}
{"train_loss": 0.391020804643631, "global_step": 27738, "epoch": 311, "lr": 9.752696216656557e-05}
{"train_loss": 0.41476544737815857, "global_step": 27739, "epoch": 311, "lr": 9.752678209458102e-05}
{"train_loss": 0.28848880529403687, "global_step": 27740, "epoch": 311, "lr": 9.752660201620704e-05}
{"train_loss": 0.4395856261253357, "global_step": 27741, "epoch": 311, "lr": 9.752642193144372e-05}
{"train_loss": 0.4596286714076996, "global_step": 27742, "epoch": 311, "lr": 9.752624184029104e-05}
{"train_loss": 0.4057702124118805, "global_step": 27743, "epoch": 311, "lr": 9.752606174274906e-05}
{"train_loss": 0.31697317957878113, "global_step": 27744, "epoch": 311, "lr": 9.752588163881775e-05}
{"train_loss": 0.4085628390312195, "global_step": 27745, "epoch": 311, "lr": 9.752570152849717e-05}
{"train_loss": 0.3807644844055176, "global_step": 27746, "epoch": 311, "lr": 9.752552141178735e-05}
{"train_loss": 0.4188401401042938, "global_step": 27747, "epoch": 311, "lr": 9.752534128868828e-05}
{"train_loss": 0.3930011987686157, "global_step": 27748, "epoch": 311, "lr": 9.752516115920002e-05}
{"train_loss": 0.3981221914291382, "global_step": 27749, "epoch": 311, "lr": 9.752498102332259e-05}
{"train_loss": 0.4792611002922058, "global_step": 27750, "epoch": 311, "lr": 9.7524800881056e-05}
{"train_loss": 0.4563133418560028, "global_step": 27751, "epoch": 311, "lr": 9.752462073240028e-05}
{"train_loss": 0.4231683015823364, "global_step": 27752, "epoch": 311, "lr": 9.752444057735545e-05}
{"train_loss": 0.43973687291145325, "global_step": 27753, "epoch": 311, "lr": 9.752426041592154e-05}
{"train_loss": 0.3914475739002228, "global_step": 27754, "epoch": 311, "lr": 9.752408024809858e-05}
{"train_loss": 0.3480645418167114, "global_step": 27755, "epoch": 311, "lr": 9.752390007388658e-05}
{"train_loss": 0.3336770236492157, "global_step": 27756, "epoch": 311, "lr": 9.752371989328558e-05}
{"train_loss": 0.3805145025253296, "global_step": 27757, "epoch": 311, "lr": 9.752353970629557e-05}
{"train_loss": 0.48558470606803894, "global_step": 27758, "epoch": 311, "lr": 9.752335951291662e-05}
{"train_loss": 0.5041683316230774, "global_step": 27759, "epoch": 311, "lr": 9.752317931314873e-05}
{"train_loss": 0.38341066241264343, "global_step": 27760, "epoch": 311, "lr": 9.752299910699191e-05}
{"train_loss": 0.3587181270122528, "global_step": 27761, "epoch": 311, "lr": 9.752281889444624e-05}
{"train_loss": 0.4720059931278229, "global_step": 27762, "epoch": 311, "lr": 9.752263867551168e-05}
{"train_loss": 0.4650663435459137, "global_step": 27763, "epoch": 311, "lr": 9.752245845018829e-05}
{"train_loss": 0.35161837935447693, "global_step": 27764, "epoch": 311, "lr": 9.752227821847608e-05}
{"train_loss": 0.4493490159511566, "global_step": 27765, "epoch": 311, "lr": 9.752209798037507e-05}
{"train_loss": 0.38170576095581055, "global_step": 27766, "epoch": 311, "lr": 9.752191773588531e-05}
{"train_loss": 0.404070304016049, "global_step": 27767, "epoch": 311, "lr": 9.75217374850068e-05, "val_loss": 2.2426648139953613}
{"train_loss": 0.3021737337112427, "global_step": 27768, "epoch": 312, "lr": 9.752155722773957e-05}
{"train_loss": 0.4525570273399353, "global_step": 27769, "epoch": 312, "lr": 9.752137696408365e-05}
{"train_loss": 0.46912726759910583, "global_step": 27770, "epoch": 312, "lr": 9.752119669403906e-05}
{"train_loss": 0.487161785364151, "global_step": 27771, "epoch": 312, "lr": 9.75210164176058e-05}
{"train_loss": 0.3758743703365326, "global_step": 27772, "epoch": 312, "lr": 9.752083613478396e-05}
{"train_loss": 0.5685247778892517, "global_step": 27773, "epoch": 312, "lr": 9.752065584557349e-05}
{"train_loss": 0.3800355792045593, "global_step": 27774, "epoch": 312, "lr": 9.752047554997445e-05}
{"train_loss": 0.4434412121772766, "global_step": 27775, "epoch": 312, "lr": 9.752029524798688e-05}
{"train_loss": 0.39428430795669556, "global_step": 27776, "epoch": 312, "lr": 9.752011493961077e-05}
{"train_loss": 0.3253680467605591, "global_step": 27777, "epoch": 312, "lr": 9.751993462484617e-05}
{"train_loss": 0.3517473638057709, "global_step": 27778, "epoch": 312, "lr": 9.751975430369309e-05}
{"train_loss": 0.5052040219306946, "global_step": 27779, "epoch": 312, "lr": 9.751957397615157e-05}
{"train_loss": 0.4097125232219696, "global_step": 27780, "epoch": 312, "lr": 9.75193936422216e-05}
{"train_loss": 0.4678229093551636, "global_step": 27781, "epoch": 312, "lr": 9.751921330190324e-05}
{"train_loss": 0.3700987994670868, "global_step": 27782, "epoch": 312, "lr": 9.75190329551965e-05}
{"train_loss": 0.35699382424354553, "global_step": 27783, "epoch": 312, "lr": 9.751885260210141e-05}
{"train_loss": 0.3365299105644226, "global_step": 27784, "epoch": 312, "lr": 9.751867224261797e-05}
{"train_loss": 0.35279276967048645, "global_step": 27785, "epoch": 312, "lr": 9.751849187674624e-05}
{"train_loss": 0.3447015881538391, "global_step": 27786, "epoch": 312, "lr": 9.751831150448622e-05}
{"train_loss": 0.2930092215538025, "global_step": 27787, "epoch": 312, "lr": 9.751813112583797e-05}
{"train_loss": 0.3593634068965912, "global_step": 27788, "epoch": 312, "lr": 9.751795074080146e-05}
{"train_loss": 0.31327691674232483, "global_step": 27789, "epoch": 312, "lr": 9.751777034937673e-05}
{"train_loss": 0.3653436601161957, "global_step": 27790, "epoch": 312, "lr": 9.751758995156383e-05}
{"train_loss": 0.3373843729496002, "global_step": 27791, "epoch": 312, "lr": 9.751740954736277e-05}
{"train_loss": 0.3675728142261505, "global_step": 27792, "epoch": 312, "lr": 9.751722913677357e-05}
{"train_loss": 0.3895292282104492, "global_step": 27793, "epoch": 312, "lr": 9.751704871979626e-05}
{"train_loss": 0.481336385011673, "global_step": 27794, "epoch": 312, "lr": 9.751686829643087e-05}
{"train_loss": 0.3479636311531067, "global_step": 27795, "epoch": 312, "lr": 9.75166878666774e-05}
{"train_loss": 0.3684786558151245, "global_step": 27796, "epoch": 312, "lr": 9.75165074305359e-05}
{"train_loss": 0.40403056144714355, "global_step": 27797, "epoch": 312, "lr": 9.751632698800638e-05}
{"train_loss": 0.27660855650901794, "global_step": 27798, "epoch": 312, "lr": 9.751614653908887e-05}
{"train_loss": 0.4138288199901581, "global_step": 27799, "epoch": 312, "lr": 9.75159660837834e-05}
{"train_loss": 0.45013371109962463, "global_step": 27800, "epoch": 312, "lr": 9.751578562208998e-05}
{"train_loss": 0.44254934787750244, "global_step": 27801, "epoch": 312, "lr": 9.751560515400864e-05}
{"train_loss": 0.34166577458381653, "global_step": 27802, "epoch": 312, "lr": 9.751542467953941e-05}
{"train_loss": 0.4102660119533539, "global_step": 27803, "epoch": 312, "lr": 9.751524419868232e-05}
{"train_loss": 0.4847738444805145, "global_step": 27804, "epoch": 312, "lr": 9.751506371143737e-05}
{"train_loss": 0.3267565965652466, "global_step": 27805, "epoch": 312, "lr": 9.75148832178046e-05}
{"train_loss": 0.3622600734233856, "global_step": 27806, "epoch": 312, "lr": 9.751470271778404e-05}
{"train_loss": 0.2454889565706253, "global_step": 27807, "epoch": 312, "lr": 9.75145222113757e-05}
{"train_loss": 0.41290566325187683, "global_step": 27808, "epoch": 312, "lr": 9.751434169857963e-05}
{"train_loss": 0.40781354904174805, "global_step": 27809, "epoch": 312, "lr": 9.75141611793958e-05}
{"train_loss": 0.37435585260391235, "global_step": 27810, "epoch": 312, "lr": 9.75139806538243e-05}
{"train_loss": 0.4469221532344818, "global_step": 27811, "epoch": 312, "lr": 9.751380012186511e-05}
{"train_loss": 0.3456830084323883, "global_step": 27812, "epoch": 312, "lr": 9.751361958351828e-05}
{"train_loss": 0.4399898648262024, "global_step": 27813, "epoch": 312, "lr": 9.751343903878383e-05}
{"train_loss": 0.35313066840171814, "global_step": 27814, "epoch": 312, "lr": 9.751325848766175e-05}
{"train_loss": 0.4411954879760742, "global_step": 27815, "epoch": 312, "lr": 9.751307793015212e-05}
{"train_loss": 0.33512386679649353, "global_step": 27816, "epoch": 312, "lr": 9.751289736625491e-05}
{"train_loss": 0.3468920886516571, "global_step": 27817, "epoch": 312, "lr": 9.751271679597018e-05}
{"train_loss": 0.3220764398574829, "global_step": 27818, "epoch": 312, "lr": 9.751253621929795e-05}
{"train_loss": 0.45994672179222107, "global_step": 27819, "epoch": 312, "lr": 9.751235563623824e-05}
{"train_loss": 0.425650030374527, "global_step": 27820, "epoch": 312, "lr": 9.751217504679107e-05}
{"train_loss": 0.4289914071559906, "global_step": 27821, "epoch": 312, "lr": 9.751199445095647e-05}
{"train_loss": 0.3413205146789551, "global_step": 27822, "epoch": 312, "lr": 9.751181384873444e-05}
{"train_loss": 0.3837658762931824, "global_step": 27823, "epoch": 312, "lr": 9.751163324012505e-05}
{"train_loss": 0.3846108615398407, "global_step": 27824, "epoch": 312, "lr": 9.751145262512829e-05}
{"train_loss": 0.39896461367607117, "global_step": 27825, "epoch": 312, "lr": 9.751127200374419e-05}
{"train_loss": 0.3371279239654541, "global_step": 27826, "epoch": 312, "lr": 9.75110913759728e-05}
{"train_loss": 0.361336886882782, "global_step": 27827, "epoch": 312, "lr": 9.751091074181411e-05}
{"train_loss": 0.3959013819694519, "global_step": 27828, "epoch": 312, "lr": 9.751073010126815e-05}
{"train_loss": 0.409071683883667, "global_step": 27829, "epoch": 312, "lr": 9.751054945433496e-05}
{"train_loss": 0.25957924127578735, "global_step": 27830, "epoch": 312, "lr": 9.751036880101454e-05}
{"train_loss": 0.40236660838127136, "global_step": 27831, "epoch": 312, "lr": 9.751018814130694e-05}
{"train_loss": 0.45427805185317993, "global_step": 27832, "epoch": 312, "lr": 9.751000747521219e-05}
{"train_loss": 0.45397570729255676, "global_step": 27833, "epoch": 312, "lr": 9.750982680273029e-05}
{"train_loss": 0.40413057804107666, "global_step": 27834, "epoch": 312, "lr": 9.750964612386126e-05}
{"train_loss": 0.4430367946624756, "global_step": 27835, "epoch": 312, "lr": 9.750946543860516e-05}
{"train_loss": 0.30785712599754333, "global_step": 27836, "epoch": 312, "lr": 9.750928474696196e-05}
{"train_loss": 0.3779369592666626, "global_step": 27837, "epoch": 312, "lr": 9.750910404893175e-05}
{"train_loss": 0.41278865933418274, "global_step": 27838, "epoch": 312, "lr": 9.75089233445145e-05}
{"train_loss": 0.35147666931152344, "global_step": 27839, "epoch": 312, "lr": 9.750874263371024e-05}
{"train_loss": 0.37969866394996643, "global_step": 27840, "epoch": 312, "lr": 9.750856191651903e-05}
{"train_loss": 0.5131154656410217, "global_step": 27841, "epoch": 312, "lr": 9.750838119294086e-05}
{"train_loss": 0.4495217204093933, "global_step": 27842, "epoch": 312, "lr": 9.750820046297578e-05}
{"train_loss": 0.5126966834068298, "global_step": 27843, "epoch": 312, "lr": 9.75080197266238e-05}
{"train_loss": 0.4449062645435333, "global_step": 27844, "epoch": 312, "lr": 9.750783898388494e-05}
{"train_loss": 0.3747124969959259, "global_step": 27845, "epoch": 312, "lr": 9.750765823475923e-05}
{"train_loss": 0.2958018481731415, "global_step": 27846, "epoch": 312, "lr": 9.75074774792467e-05}
{"train_loss": 0.5891861915588379, "global_step": 27847, "epoch": 312, "lr": 9.750729671734737e-05}
{"train_loss": 0.30840611457824707, "global_step": 27848, "epoch": 312, "lr": 9.750711594906125e-05}
{"train_loss": 0.4493192136287689, "global_step": 27849, "epoch": 312, "lr": 9.750693517438839e-05}
{"train_loss": 0.35711514949798584, "global_step": 27850, "epoch": 312, "lr": 9.75067543933288e-05}
{"train_loss": 0.42187634110450745, "global_step": 27851, "epoch": 312, "lr": 9.75065736058825e-05}
{"train_loss": 0.35066157579421997, "global_step": 27852, "epoch": 312, "lr": 9.750639281204953e-05}
{"train_loss": 0.3382488489151001, "global_step": 27853, "epoch": 312, "lr": 9.75062120118299e-05}
{"train_loss": 0.4098452031612396, "global_step": 27854, "epoch": 312, "lr": 9.750603120522364e-05}
{"train_loss": 0.4191167950630188, "global_step": 27855, "epoch": 312, "lr": 9.750585039223076e-05}
{"train_loss": 0.3923139272446043, "global_step": 27856, "epoch": 312, "lr": 9.750566957285132e-05, "val_loss": 2.0850985050201416}
{"train_loss": 0.4130558967590332, "global_step": 27857, "epoch": 313, "lr": 9.750548874708529e-05}
{"train_loss": 0.3017588257789612, "global_step": 27858, "epoch": 313, "lr": 9.750530791493275e-05}
{"train_loss": 0.3434217572212219, "global_step": 27859, "epoch": 313, "lr": 9.75051270763937e-05}
{"train_loss": 0.4578222632408142, "global_step": 27860, "epoch": 313, "lr": 9.750494623146816e-05}
{"train_loss": 0.3653126657009125, "global_step": 27861, "epoch": 313, "lr": 9.750476538015617e-05}
{"train_loss": 0.3677363097667694, "global_step": 27862, "epoch": 313, "lr": 9.750458452245772e-05}
{"train_loss": 0.43972155451774597, "global_step": 27863, "epoch": 313, "lr": 9.750440365837287e-05}
{"train_loss": 0.4210470914840698, "global_step": 27864, "epoch": 313, "lr": 9.750422278790163e-05}
{"train_loss": 0.3313331604003906, "global_step": 27865, "epoch": 313, "lr": 9.750404191104403e-05}
{"train_loss": 0.5085254311561584, "global_step": 27866, "epoch": 313, "lr": 9.750386102780009e-05}
{"train_loss": 0.43562668561935425, "global_step": 27867, "epoch": 313, "lr": 9.750368013816984e-05}
{"train_loss": 0.4299590587615967, "global_step": 27868, "epoch": 313, "lr": 9.750349924215329e-05}
{"train_loss": 0.39591073989868164, "global_step": 27869, "epoch": 313, "lr": 9.750331833975048e-05}
{"train_loss": 0.4837264120578766, "global_step": 27870, "epoch": 313, "lr": 9.750313743096142e-05}
{"train_loss": 0.43258699774742126, "global_step": 27871, "epoch": 313, "lr": 9.750295651578615e-05}
{"train_loss": 0.39849352836608887, "global_step": 27872, "epoch": 313, "lr": 9.750277559422468e-05}
{"train_loss": 0.3625551462173462, "global_step": 27873, "epoch": 313, "lr": 9.750259466627705e-05}
{"train_loss": 0.4488217234611511, "global_step": 27874, "epoch": 313, "lr": 9.750241373194327e-05}
{"train_loss": 0.2907019853591919, "global_step": 27875, "epoch": 313, "lr": 9.750223279122337e-05}
{"train_loss": 0.27378326654434204, "global_step": 27876, "epoch": 313, "lr": 9.750205184411735e-05}
{"train_loss": 0.4226348400115967, "global_step": 27877, "epoch": 313, "lr": 9.750187089062531e-05}
{"train_loss": 0.3069174885749817, "global_step": 27878, "epoch": 313, "lr": 9.750168993074718e-05}
{"train_loss": 0.33981525897979736, "global_step": 27879, "epoch": 313, "lr": 9.750150896448304e-05}
{"train_loss": 0.40787574648857117, "global_step": 27880, "epoch": 313, "lr": 9.75013279918329e-05}
{"train_loss": 0.36012595891952515, "global_step": 27881, "epoch": 313, "lr": 9.750114701279678e-05}
{"train_loss": 0.3691284954547882, "global_step": 27882, "epoch": 313, "lr": 9.750096602737472e-05}
{"train_loss": 0.5020301938056946, "global_step": 27883, "epoch": 313, "lr": 9.750078503556673e-05}
{"train_loss": 0.40510764718055725, "global_step": 27884, "epoch": 313, "lr": 9.750060403737284e-05}
{"train_loss": 0.4162960648536682, "global_step": 27885, "epoch": 313, "lr": 9.750042303279308e-05}
{"train_loss": 0.41275933384895325, "global_step": 27886, "epoch": 313, "lr": 9.750024202182745e-05}
{"train_loss": 0.48173898458480835, "global_step": 27887, "epoch": 313, "lr": 9.750006100447599e-05}
{"train_loss": 0.3668091893196106, "global_step": 27888, "epoch": 313, "lr": 9.749987998073875e-05}
{"train_loss": 0.37513139843940735, "global_step": 27889, "epoch": 313, "lr": 9.74996989506157e-05}
{"train_loss": 0.4086509943008423, "global_step": 27890, "epoch": 313, "lr": 9.749951791410692e-05}
{"train_loss": 0.39260607957839966, "global_step": 27891, "epoch": 313, "lr": 9.74993368712124e-05}
{"train_loss": 0.3282310664653778, "global_step": 27892, "epoch": 313, "lr": 9.749915582193216e-05}
{"train_loss": 0.42264437675476074, "global_step": 27893, "epoch": 313, "lr": 9.749897476626626e-05}
{"train_loss": 0.41828566789627075, "global_step": 27894, "epoch": 313, "lr": 9.749879370421469e-05}
{"train_loss": 0.4447641372680664, "global_step": 27895, "epoch": 313, "lr": 9.749861263577748e-05}
{"train_loss": 0.4393860399723053, "global_step": 27896, "epoch": 313, "lr": 9.749843156095468e-05}
{"train_loss": 0.47928178310394287, "global_step": 27897, "epoch": 313, "lr": 9.749825047974627e-05}
{"train_loss": 0.42959311604499817, "global_step": 27898, "epoch": 313, "lr": 9.749806939215231e-05}
{"train_loss": 0.4171651601791382, "global_step": 27899, "epoch": 313, "lr": 9.749788829817283e-05}
{"train_loss": 0.3293522000312805, "global_step": 27900, "epoch": 313, "lr": 9.749770719780782e-05}
{"train_loss": 0.420520544052124, "global_step": 27901, "epoch": 313, "lr": 9.74975260910573e-05}
{"train_loss": 0.4594196379184723, "global_step": 27902, "epoch": 313, "lr": 9.749734497792136e-05}
{"train_loss": 0.35102230310440063, "global_step": 27903, "epoch": 313, "lr": 9.749716385839997e-05}
{"train_loss": 0.41538509726524353, "global_step": 27904, "epoch": 313, "lr": 9.749698273249315e-05}
{"train_loss": 0.4065185785293579, "global_step": 27905, "epoch": 313, "lr": 9.749680160020095e-05}
{"train_loss": 0.41863739490509033, "global_step": 27906, "epoch": 313, "lr": 9.749662046152338e-05}
{"train_loss": 0.3693970739841461, "global_step": 27907, "epoch": 313, "lr": 9.749643931646046e-05}
{"train_loss": 0.3793649971485138, "global_step": 27908, "epoch": 313, "lr": 9.749625816501224e-05}
{"train_loss": 0.3107134699821472, "global_step": 27909, "epoch": 313, "lr": 9.749607700717873e-05}
{"train_loss": 0.38568124175071716, "global_step": 27910, "epoch": 313, "lr": 9.749589584295993e-05}
{"train_loss": 0.44051042199134827, "global_step": 27911, "epoch": 313, "lr": 9.74957146723559e-05}
{"train_loss": 0.4348841607570648, "global_step": 27912, "epoch": 313, "lr": 9.749553349536665e-05}
{"train_loss": 0.31964290142059326, "global_step": 27913, "epoch": 313, "lr": 9.749535231199219e-05}
{"train_loss": 0.4038093090057373, "global_step": 27914, "epoch": 313, "lr": 9.749517112223257e-05}
{"train_loss": 0.3802858293056488, "global_step": 27915, "epoch": 313, "lr": 9.749498992608781e-05}
{"train_loss": 0.29340535402297974, "global_step": 27916, "epoch": 313, "lr": 9.749480872355791e-05}
{"train_loss": 0.3580529987812042, "global_step": 27917, "epoch": 313, "lr": 9.749462751464292e-05}
{"train_loss": 0.4597373902797699, "global_step": 27918, "epoch": 313, "lr": 9.749444629934286e-05}
{"train_loss": 0.3895130455493927, "global_step": 27919, "epoch": 313, "lr": 9.749426507765775e-05}
{"train_loss": 0.3020770251750946, "global_step": 27920, "epoch": 313, "lr": 9.749408384958762e-05}
{"train_loss": 0.2737353444099426, "global_step": 27921, "epoch": 313, "lr": 9.749390261513247e-05}
{"train_loss": 0.44652342796325684, "global_step": 27922, "epoch": 313, "lr": 9.749372137429236e-05}
{"train_loss": 0.4405030310153961, "global_step": 27923, "epoch": 313, "lr": 9.74935401270673e-05}
{"train_loss": 0.34293410181999207, "global_step": 27924, "epoch": 313, "lr": 9.74933588734573e-05}
{"train_loss": 0.38622725009918213, "global_step": 27925, "epoch": 313, "lr": 9.74931776134624e-05}
{"train_loss": 0.3370618224143982, "global_step": 27926, "epoch": 313, "lr": 9.749299634708263e-05}
{"train_loss": 0.43638575077056885, "global_step": 27927, "epoch": 313, "lr": 9.7492815074318e-05}
{"train_loss": 0.32006335258483887, "global_step": 27928, "epoch": 313, "lr": 9.749263379516854e-05}
{"train_loss": 0.3349403440952301, "global_step": 27929, "epoch": 313, "lr": 9.749245250963428e-05}
{"train_loss": 0.42771151661872864, "global_step": 27930, "epoch": 313, "lr": 9.749227121771522e-05}
{"train_loss": 0.34239205718040466, "global_step": 27931, "epoch": 313, "lr": 9.749208991941142e-05}
{"train_loss": 0.3131476044654846, "global_step": 27932, "epoch": 313, "lr": 9.749190861472288e-05}
{"train_loss": 0.44776076078414917, "global_step": 27933, "epoch": 313, "lr": 9.749172730364964e-05}
{"train_loss": 0.441648930311203, "global_step": 27934, "epoch": 313, "lr": 9.749154598619172e-05}
{"train_loss": 0.3966057002544403, "global_step": 27935, "epoch": 313, "lr": 9.749136466234911e-05}
{"train_loss": 0.3865109980106354, "global_step": 27936, "epoch": 313, "lr": 9.749118333212191e-05}
{"train_loss": 0.29206427931785583, "global_step": 27937, "epoch": 313, "lr": 9.749100199551007e-05}
{"train_loss": 0.39035290479660034, "global_step": 27938, "epoch": 313, "lr": 9.749082065251365e-05}
{"train_loss": 0.4622175395488739, "global_step": 27939, "epoch": 313, "lr": 9.749063930313267e-05}
{"train_loss": 0.39678308367729187, "global_step": 27940, "epoch": 313, "lr": 9.749045794736714e-05}
{"train_loss": 0.29879817366600037, "global_step": 27941, "epoch": 313, "lr": 9.749027658521711e-05}
{"train_loss": 0.446310430765152, "global_step": 27942, "epoch": 313, "lr": 9.749009521668259e-05}
{"train_loss": 0.45956921577453613, "global_step": 27943, "epoch": 313, "lr": 9.74899138417636e-05}
{"train_loss": 0.3761512041091919, "global_step": 27944, "epoch": 313, "lr": 9.748973246046016e-05}
{"train_loss": 0.3929634415701534, "global_step": 27945, "epoch": 313, "lr": 9.748955107277234e-05, "val_loss": 2.256232976913452}
{"train_loss": 0.3469562828540802, "global_step": 27946, "epoch": 314, "lr": 9.748936967870009e-05}
{"train_loss": 0.5501834154129028, "global_step": 27947, "epoch": 314, "lr": 9.748918827824349e-05}
{"train_loss": 0.4293616712093353, "global_step": 27948, "epoch": 314, "lr": 9.748900687140255e-05}
{"train_loss": 0.3074910044670105, "global_step": 27949, "epoch": 314, "lr": 9.748882545817729e-05}
{"train_loss": 0.3677968680858612, "global_step": 27950, "epoch": 314, "lr": 9.748864403856773e-05}
{"train_loss": 0.261881947517395, "global_step": 27951, "epoch": 314, "lr": 9.748846261257391e-05}
{"train_loss": 0.3918898105621338, "global_step": 27952, "epoch": 314, "lr": 9.748828118019584e-05}
{"train_loss": 0.34056058526039124, "global_step": 27953, "epoch": 314, "lr": 9.748809974143355e-05}
{"train_loss": 0.4666365087032318, "global_step": 27954, "epoch": 314, "lr": 9.748791829628706e-05}
{"train_loss": 0.3898831605911255, "global_step": 27955, "epoch": 314, "lr": 9.74877368447564e-05}
{"train_loss": 0.38269859552383423, "global_step": 27956, "epoch": 314, "lr": 9.74875553868416e-05}
{"train_loss": 0.2746115028858185, "global_step": 27957, "epoch": 314, "lr": 9.748737392254267e-05}
{"train_loss": 0.3548349142074585, "global_step": 27958, "epoch": 314, "lr": 9.748719245185965e-05}
{"train_loss": 0.35666728019714355, "global_step": 27959, "epoch": 314, "lr": 9.748701097479254e-05}
{"train_loss": 0.32391974329948425, "global_step": 27960, "epoch": 314, "lr": 9.748682949134139e-05}
{"train_loss": 0.47334328293800354, "global_step": 27961, "epoch": 314, "lr": 9.748664800150621e-05}
{"train_loss": 0.5181125402450562, "global_step": 27962, "epoch": 314, "lr": 9.748646650528703e-05}
{"train_loss": 0.31463319063186646, "global_step": 27963, "epoch": 314, "lr": 9.748628500268387e-05}
{"train_loss": 0.3781510591506958, "global_step": 27964, "epoch": 314, "lr": 9.748610349369676e-05}
{"train_loss": 0.39205798506736755, "global_step": 27965, "epoch": 314, "lr": 9.748592197832574e-05}
{"train_loss": 0.3652293384075165, "global_step": 27966, "epoch": 314, "lr": 9.748574045657079e-05}
{"train_loss": 0.5665351152420044, "global_step": 27967, "epoch": 314, "lr": 9.748555892843199e-05}
{"train_loss": 0.39754852652549744, "global_step": 27968, "epoch": 314, "lr": 9.74853773939093e-05}
{"train_loss": 0.3910028040409088, "global_step": 27969, "epoch": 314, "lr": 9.74851958530028e-05}
{"train_loss": 0.4327048659324646, "global_step": 27970, "epoch": 314, "lr": 9.748501430571249e-05}
{"train_loss": 0.3457774519920349, "global_step": 27971, "epoch": 314, "lr": 9.748483275203842e-05}
{"train_loss": 0.28517770767211914, "global_step": 27972, "epoch": 314, "lr": 9.748465119198056e-05}
{"train_loss": 0.41384220123291016, "global_step": 27973, "epoch": 314, "lr": 9.748446962553898e-05}
{"train_loss": 0.3098292052745819, "global_step": 27974, "epoch": 314, "lr": 9.748428805271368e-05}
{"train_loss": 0.5210751295089722, "global_step": 27975, "epoch": 314, "lr": 9.748410647350473e-05}
{"train_loss": 0.3264152407646179, "global_step": 27976, "epoch": 314, "lr": 9.74839248879121e-05}
{"train_loss": 0.40448594093322754, "global_step": 27977, "epoch": 314, "lr": 9.748374329593583e-05}
{"train_loss": 0.43334609270095825, "global_step": 27978, "epoch": 314, "lr": 9.748356169757596e-05}
{"train_loss": 0.2849121391773224, "global_step": 27979, "epoch": 314, "lr": 9.74833800928325e-05}
{"train_loss": 0.38405534625053406, "global_step": 27980, "epoch": 314, "lr": 9.748319848170547e-05}
{"train_loss": 0.3642951250076294, "global_step": 27981, "epoch": 314, "lr": 9.74830168641949e-05}
{"train_loss": 0.3957095444202423, "global_step": 27982, "epoch": 314, "lr": 9.748283524030083e-05}
{"train_loss": 0.4526987373828888, "global_step": 27983, "epoch": 314, "lr": 9.748265361002327e-05}
{"train_loss": 0.42925602197647095, "global_step": 27984, "epoch": 314, "lr": 9.748247197336225e-05}
{"train_loss": 0.38508617877960205, "global_step": 27985, "epoch": 314, "lr": 9.748229033031777e-05}
{"train_loss": 0.42696040868759155, "global_step": 27986, "epoch": 314, "lr": 9.748210868088989e-05}
{"train_loss": 0.3451433777809143, "global_step": 27987, "epoch": 314, "lr": 9.748192702507862e-05}
{"train_loss": 0.40044811367988586, "global_step": 27988, "epoch": 314, "lr": 9.748174536288399e-05}
{"train_loss": 0.3650374114513397, "global_step": 27989, "epoch": 314, "lr": 9.7481563694306e-05}
{"train_loss": 0.38282492756843567, "global_step": 27990, "epoch": 314, "lr": 9.74813820193447e-05}
{"train_loss": 0.3556728661060333, "global_step": 27991, "epoch": 314, "lr": 9.74812003380001e-05}
{"train_loss": 0.4390683174133301, "global_step": 27992, "epoch": 314, "lr": 9.748101865027226e-05}
{"train_loss": 0.39373743534088135, "global_step": 27993, "epoch": 314, "lr": 9.748083695616115e-05}
{"train_loss": 0.4139910340309143, "global_step": 27994, "epoch": 314, "lr": 9.748065525566683e-05}
{"train_loss": 0.3777911365032196, "global_step": 27995, "epoch": 314, "lr": 9.748047354878931e-05}
{"train_loss": 0.3999823033809662, "global_step": 27996, "epoch": 314, "lr": 9.748029183552861e-05}
{"train_loss": 0.5714288353919983, "global_step": 27997, "epoch": 314, "lr": 9.748011011588478e-05}
{"train_loss": 0.4475593864917755, "global_step": 27998, "epoch": 314, "lr": 9.747992838985782e-05}
{"train_loss": 0.41546863317489624, "global_step": 27999, "epoch": 314, "lr": 9.747974665744776e-05}
{"train_loss": 0.44631078839302063, "global_step": 28000, "epoch": 314, "lr": 9.747956491865464e-05}
{"train_loss": 0.34664174914360046, "global_step": 28001, "epoch": 314, "lr": 9.747938317347845e-05}
{"train_loss": 0.332086443901062, "global_step": 28002, "epoch": 314, "lr": 9.747920142191926e-05}
{"train_loss": 0.5007064938545227, "global_step": 28003, "epoch": 314, "lr": 9.747901966397705e-05}
{"train_loss": 0.3848176896572113, "global_step": 28004, "epoch": 314, "lr": 9.747883789965187e-05}
{"train_loss": 0.387042760848999, "global_step": 28005, "epoch": 314, "lr": 9.747865612894374e-05}
{"train_loss": 0.2792670428752899, "global_step": 28006, "epoch": 314, "lr": 9.747847435185268e-05}
{"train_loss": 0.4123475253582001, "global_step": 28007, "epoch": 314, "lr": 9.747829256837872e-05}
{"train_loss": 0.38716670870780945, "global_step": 28008, "epoch": 314, "lr": 9.747811077852189e-05}
{"train_loss": 0.4767445921897888, "global_step": 28009, "epoch": 314, "lr": 9.74779289822822e-05}
{"train_loss": 0.4177750051021576, "global_step": 28010, "epoch": 314, "lr": 9.747774717965968e-05}
{"train_loss": 0.36468505859375, "global_step": 28011, "epoch": 314, "lr": 9.747756537065436e-05}
{"train_loss": 0.3674718141555786, "global_step": 28012, "epoch": 314, "lr": 9.747738355526626e-05}
{"train_loss": 0.3822115659713745, "global_step": 28013, "epoch": 314, "lr": 9.74772017334954e-05}
{"train_loss": 0.393288254737854, "global_step": 28014, "epoch": 314, "lr": 9.747701990534182e-05}
{"train_loss": 0.4414149522781372, "global_step": 28015, "epoch": 314, "lr": 9.747683807080552e-05}
{"train_loss": 0.30099159479141235, "global_step": 28016, "epoch": 314, "lr": 9.747665622988653e-05}
{"train_loss": 0.40244990587234497, "global_step": 28017, "epoch": 314, "lr": 9.74764743825849e-05}
{"train_loss": 0.47372692823410034, "global_step": 28018, "epoch": 314, "lr": 9.747629252890064e-05}
{"train_loss": 0.3814902603626251, "global_step": 28019, "epoch": 314, "lr": 9.747611066883376e-05}
{"train_loss": 0.33484670519828796, "global_step": 28020, "epoch": 314, "lr": 9.74759288023843e-05}
{"train_loss": 0.339149534702301, "global_step": 28021, "epoch": 314, "lr": 9.747574692955226e-05}
{"train_loss": 0.3728639483451843, "global_step": 28022, "epoch": 314, "lr": 9.747556505033771e-05}
{"train_loss": 0.37368646264076233, "global_step": 28023, "epoch": 314, "lr": 9.747538316474066e-05}
{"train_loss": 0.4969211220741272, "global_step": 28024, "epoch": 314, "lr": 9.747520127276108e-05}
{"train_loss": 0.3507155179977417, "global_step": 28025, "epoch": 314, "lr": 9.747501937439907e-05}
{"train_loss": 0.5242732763290405, "global_step": 28026, "epoch": 314, "lr": 9.747483746965463e-05}
{"train_loss": 0.3397591710090637, "global_step": 28027, "epoch": 314, "lr": 9.747465555852775e-05}
{"train_loss": 0.3233446478843689, "global_step": 28028, "epoch": 314, "lr": 9.747447364101849e-05}
{"train_loss": 0.3171479105949402, "global_step": 28029, "epoch": 314, "lr": 9.747429171712686e-05}
{"train_loss": 0.45668959617614746, "global_step": 28030, "epoch": 314, "lr": 9.74741097868529e-05}
{"train_loss": 0.4049152731895447, "global_step": 28031, "epoch": 314, "lr": 9.747392785019662e-05}
{"train_loss": 0.4084547460079193, "global_step": 28032, "epoch": 314, "lr": 9.747374590715805e-05}
{"train_loss": 0.4023015797138214, "global_step": 28033, "epoch": 314, "lr": 9.747356395773722e-05}
{"train_loss": 0.3931373257315561, "global_step": 28034, "epoch": 314, "lr": 9.747338200193415e-05, "val_loss": 2.164558172225952}
{"train_loss": 0.46515050530433655, "global_step": 28035, "epoch": 315, "lr": 9.747320003974885e-05}
{"train_loss": 0.4109784960746765, "global_step": 28036, "epoch": 315, "lr": 9.747301807118137e-05}
{"train_loss": 0.3921566605567932, "global_step": 28037, "epoch": 315, "lr": 9.747283609623171e-05}
{"train_loss": 0.4152292311191559, "global_step": 28038, "epoch": 315, "lr": 9.74726541148999e-05}
{"train_loss": 0.3726916015148163, "global_step": 28039, "epoch": 315, "lr": 9.747247212718598e-05}
{"train_loss": 0.4472089409828186, "global_step": 28040, "epoch": 315, "lr": 9.747229013308998e-05}
{"train_loss": 0.43926262855529785, "global_step": 28041, "epoch": 315, "lr": 9.747210813261189e-05}
{"train_loss": 0.4307869076728821, "global_step": 28042, "epoch": 315, "lr": 9.747192612575176e-05}
{"train_loss": 0.44595953822135925, "global_step": 28043, "epoch": 315, "lr": 9.74717441125096e-05}
{"train_loss": 0.36938220262527466, "global_step": 28044, "epoch": 315, "lr": 9.747156209288545e-05}
{"train_loss": 0.3322519063949585, "global_step": 28045, "epoch": 315, "lr": 9.747138006687932e-05}
{"train_loss": 0.3836533725261688, "global_step": 28046, "epoch": 315, "lr": 9.747119803449125e-05}
{"train_loss": 0.39929792284965515, "global_step": 28047, "epoch": 315, "lr": 9.747101599572126e-05}
{"train_loss": 0.4325510859489441, "global_step": 28048, "epoch": 315, "lr": 9.747083395056935e-05}
{"train_loss": 0.31930088996887207, "global_step": 28049, "epoch": 315, "lr": 9.74706518990356e-05}
{"train_loss": 0.38538235425949097, "global_step": 28050, "epoch": 315, "lr": 9.747046984111996e-05}
{"train_loss": 0.40107643604278564, "global_step": 28051, "epoch": 315, "lr": 9.747028777682253e-05}
{"train_loss": 0.3439745008945465, "global_step": 28052, "epoch": 315, "lr": 9.747010570614326e-05}
{"train_loss": 0.3847655951976776, "global_step": 28053, "epoch": 315, "lr": 9.746992362908223e-05}
{"train_loss": 0.3373029828071594, "global_step": 28054, "epoch": 315, "lr": 9.746974154563945e-05}
{"train_loss": 0.32331323623657227, "global_step": 28055, "epoch": 315, "lr": 9.746955945581495e-05}
{"train_loss": 0.40508338809013367, "global_step": 28056, "epoch": 315, "lr": 9.746937735960874e-05}
{"train_loss": 0.4690534472465515, "global_step": 28057, "epoch": 315, "lr": 9.746919525702085e-05}
{"train_loss": 0.41886574029922485, "global_step": 28058, "epoch": 315, "lr": 9.74690131480513e-05}
{"train_loss": 0.3057170510292053, "global_step": 28059, "epoch": 315, "lr": 9.746883103270013e-05}
{"train_loss": 0.4615922272205353, "global_step": 28060, "epoch": 315, "lr": 9.746864891096734e-05}
{"train_loss": 0.3256663382053375, "global_step": 28061, "epoch": 315, "lr": 9.746846678285298e-05}
{"train_loss": 0.39855054020881653, "global_step": 28062, "epoch": 315, "lr": 9.746828464835707e-05}
{"train_loss": 0.3360233008861542, "global_step": 28063, "epoch": 315, "lr": 9.74681025074796e-05}
{"train_loss": 0.33001378178596497, "global_step": 28064, "epoch": 315, "lr": 9.746792036022066e-05}
{"train_loss": 0.41974422335624695, "global_step": 28065, "epoch": 315, "lr": 9.746773820658021e-05}
{"train_loss": 0.3277066946029663, "global_step": 28066, "epoch": 315, "lr": 9.74675560465583e-05}
{"train_loss": 0.33317890763282776, "global_step": 28067, "epoch": 315, "lr": 9.746737388015497e-05}
{"train_loss": 0.3641301393508911, "global_step": 28068, "epoch": 315, "lr": 9.746719170737021e-05}
{"train_loss": 0.40190404653549194, "global_step": 28069, "epoch": 315, "lr": 9.746700952820409e-05}
{"train_loss": 0.354378342628479, "global_step": 28070, "epoch": 315, "lr": 9.74668273426566e-05}
{"train_loss": 0.389464795589447, "global_step": 28071, "epoch": 315, "lr": 9.746664515072777e-05}
{"train_loss": 0.2871321141719818, "global_step": 28072, "epoch": 315, "lr": 9.746646295241763e-05}
{"train_loss": 0.35066545009613037, "global_step": 28073, "epoch": 315, "lr": 9.74662807477262e-05}
{"train_loss": 0.2627168893814087, "global_step": 28074, "epoch": 315, "lr": 9.746609853665352e-05}
{"train_loss": 0.41698887944221497, "global_step": 28075, "epoch": 315, "lr": 9.746591631919958e-05}
{"train_loss": 0.2987602949142456, "global_step": 28076, "epoch": 315, "lr": 9.746573409536444e-05}
{"train_loss": 0.4545729160308838, "global_step": 28077, "epoch": 315, "lr": 9.746555186514809e-05}
{"train_loss": 0.3555941581726074, "global_step": 28078, "epoch": 315, "lr": 9.74653696285506e-05}
{"train_loss": 0.36597850918769836, "global_step": 28079, "epoch": 315, "lr": 9.746518738557196e-05}
{"train_loss": 0.38976413011550903, "global_step": 28080, "epoch": 315, "lr": 9.746500513621222e-05}
{"train_loss": 0.40343037247657776, "global_step": 28081, "epoch": 315, "lr": 9.746482288047136e-05}
{"train_loss": 0.4385574162006378, "global_step": 28082, "epoch": 315, "lr": 9.746464061834945e-05}
{"train_loss": 0.3075574040412903, "global_step": 28083, "epoch": 315, "lr": 9.74644583498465e-05}
{"train_loss": 0.38202816247940063, "global_step": 28084, "epoch": 315, "lr": 9.746427607496252e-05}
{"train_loss": 0.43291276693344116, "global_step": 28085, "epoch": 315, "lr": 9.746409379369757e-05}
{"train_loss": 0.33924099802970886, "global_step": 28086, "epoch": 315, "lr": 9.746391150605163e-05}
{"train_loss": 0.34082573652267456, "global_step": 28087, "epoch": 315, "lr": 9.746372921202475e-05}
{"train_loss": 0.40110549330711365, "global_step": 28088, "epoch": 315, "lr": 9.746354691161694e-05}
{"train_loss": 0.3556044101715088, "global_step": 28089, "epoch": 315, "lr": 9.746336460482826e-05}
{"train_loss": 0.4529798626899719, "global_step": 28090, "epoch": 315, "lr": 9.746318229165868e-05}
{"train_loss": 0.31701982021331787, "global_step": 28091, "epoch": 315, "lr": 9.746299997210827e-05}
{"train_loss": 0.31709352135658264, "global_step": 28092, "epoch": 315, "lr": 9.746281764617704e-05}
{"train_loss": 0.45551204681396484, "global_step": 28093, "epoch": 315, "lr": 9.746263531386502e-05}
{"train_loss": 0.3912409245967865, "global_step": 28094, "epoch": 315, "lr": 9.74624529751722e-05}
{"train_loss": 0.37773987650871277, "global_step": 28095, "epoch": 315, "lr": 9.746227063009865e-05}
{"train_loss": 0.47144731879234314, "global_step": 28096, "epoch": 315, "lr": 9.746208827864439e-05}
{"train_loss": 0.404837965965271, "global_step": 28097, "epoch": 315, "lr": 9.746190592080939e-05}
{"train_loss": 0.4402914345264435, "global_step": 28098, "epoch": 315, "lr": 9.746172355659376e-05}
{"train_loss": 0.5078772306442261, "global_step": 28099, "epoch": 315, "lr": 9.746154118599744e-05}
{"train_loss": 0.4014981687068939, "global_step": 28100, "epoch": 315, "lr": 9.746135880902052e-05}
{"train_loss": 0.5166930556297302, "global_step": 28101, "epoch": 315, "lr": 9.746117642566299e-05}
{"train_loss": 0.34193381667137146, "global_step": 28102, "epoch": 315, "lr": 9.74609940359249e-05}
{"train_loss": 0.3643118739128113, "global_step": 28103, "epoch": 315, "lr": 9.746081163980623e-05}
{"train_loss": 0.41384357213974, "global_step": 28104, "epoch": 315, "lr": 9.746062923730705e-05}
{"train_loss": 0.6789349317550659, "global_step": 28105, "epoch": 315, "lr": 9.746044682842736e-05}
{"train_loss": 0.500603973865509, "global_step": 28106, "epoch": 315, "lr": 9.74602644131672e-05}
{"train_loss": 0.4114728271961212, "global_step": 28107, "epoch": 315, "lr": 9.746008199152657e-05}
{"train_loss": 0.43102410435676575, "global_step": 28108, "epoch": 315, "lr": 9.745989956350552e-05}
{"train_loss": 0.33013391494750977, "global_step": 28109, "epoch": 315, "lr": 9.745971712910407e-05}
{"train_loss": 0.4875217378139496, "global_step": 28110, "epoch": 315, "lr": 9.745953468832224e-05}
{"train_loss": 0.3961262106895447, "global_step": 28111, "epoch": 315, "lr": 9.745935224116006e-05}
{"train_loss": 0.49969249963760376, "global_step": 28112, "epoch": 315, "lr": 9.745916978761754e-05}
{"train_loss": 0.5047569274902344, "global_step": 28113, "epoch": 315, "lr": 9.74589873276947e-05}
{"train_loss": 0.3917144536972046, "global_step": 28114, "epoch": 315, "lr": 9.745880486139158e-05}
{"train_loss": 0.44692692160606384, "global_step": 28115, "epoch": 315, "lr": 9.745862238870823e-05}
{"train_loss": 0.3597404360771179, "global_step": 28116, "epoch": 315, "lr": 9.74584399096446e-05}
{"train_loss": 0.4369470179080963, "global_step": 28117, "epoch": 315, "lr": 9.74582574242008e-05}
{"train_loss": 0.36140763759613037, "global_step": 28118, "epoch": 315, "lr": 9.745807493237679e-05}
{"train_loss": 0.4052555561065674, "global_step": 28119, "epoch": 315, "lr": 9.745789243417265e-05}
{"train_loss": 0.3187488913536072, "global_step": 28120, "epoch": 315, "lr": 9.745770992958835e-05}
{"train_loss": 0.49036705493927, "global_step": 28121, "epoch": 315, "lr": 9.745752741862395e-05}
{"train_loss": 0.43057212233543396, "global_step": 28122, "epoch": 315, "lr": 9.745734490127946e-05}
{"train_loss": 0.39674667160162763, "global_step": 28123, "epoch": 315, "lr": 9.745716237755491e-05, "val_loss": 2.1254775524139404, "train_action_mse_error": 25.641357421875}
{"train_loss": 0.34131577610969543, "global_step": 28124, "epoch": 316, "lr": 9.745697984745033e-05}
{"train_loss": 0.40572527050971985, "global_step": 28125, "epoch": 316, "lr": 9.745679731096573e-05}
{"train_loss": 0.46356478333473206, "global_step": 28126, "epoch": 316, "lr": 9.745661476810116e-05}
{"train_loss": 0.40028148889541626, "global_step": 28127, "epoch": 316, "lr": 9.745643221885658e-05}
{"train_loss": 0.3955109417438507, "global_step": 28128, "epoch": 316, "lr": 9.74562496632321e-05}
{"train_loss": 0.46767622232437134, "global_step": 28129, "epoch": 316, "lr": 9.74560671012277e-05}
{"train_loss": 0.36177629232406616, "global_step": 28130, "epoch": 316, "lr": 9.74558845328434e-05}
{"train_loss": 0.5208953022956848, "global_step": 28131, "epoch": 316, "lr": 9.745570195807925e-05}
{"train_loss": 0.4221148192882538, "global_step": 28132, "epoch": 316, "lr": 9.745551937693524e-05}
{"train_loss": 0.3968124985694885, "global_step": 28133, "epoch": 316, "lr": 9.745533678941143e-05}
{"train_loss": 0.3833068013191223, "global_step": 28134, "epoch": 316, "lr": 9.745515419550782e-05}
{"train_loss": 0.4522208273410797, "global_step": 28135, "epoch": 316, "lr": 9.745497159522446e-05}
{"train_loss": 0.35218414664268494, "global_step": 28136, "epoch": 316, "lr": 9.745478898856133e-05}
{"train_loss": 0.3433946371078491, "global_step": 28137, "epoch": 316, "lr": 9.74546063755185e-05}
{"train_loss": 0.41807734966278076, "global_step": 28138, "epoch": 316, "lr": 9.745442375609597e-05}
{"train_loss": 0.39770665764808655, "global_step": 28139, "epoch": 316, "lr": 9.745424113029377e-05}
{"train_loss": 0.6312351226806641, "global_step": 28140, "epoch": 316, "lr": 9.745405849811193e-05}
{"train_loss": 0.4753052294254303, "global_step": 28141, "epoch": 316, "lr": 9.745387585955047e-05}
{"train_loss": 0.4582146108150482, "global_step": 28142, "epoch": 316, "lr": 9.745369321460942e-05}
{"train_loss": 0.4554293751716614, "global_step": 28143, "epoch": 316, "lr": 9.745351056328878e-05}
{"train_loss": 0.3457455337047577, "global_step": 28144, "epoch": 316, "lr": 9.74533279055886e-05}
{"train_loss": 0.5498772263526917, "global_step": 28145, "epoch": 316, "lr": 9.745314524150893e-05}
{"train_loss": 0.5230997800827026, "global_step": 28146, "epoch": 316, "lr": 9.745296257104972e-05}
{"train_loss": 0.39330220222473145, "global_step": 28147, "epoch": 316, "lr": 9.745277989421106e-05}
{"train_loss": 0.3615708351135254, "global_step": 28148, "epoch": 316, "lr": 9.745259721099295e-05}
{"train_loss": 0.5018218159675598, "global_step": 28149, "epoch": 316, "lr": 9.745241452139541e-05}
{"train_loss": 0.38543441891670227, "global_step": 28150, "epoch": 316, "lr": 9.745223182541848e-05}
{"train_loss": 0.4881863296031952, "global_step": 28151, "epoch": 316, "lr": 9.745204912306217e-05}
{"train_loss": 0.5171949863433838, "global_step": 28152, "epoch": 316, "lr": 9.745186641432651e-05}
{"train_loss": 0.4054584503173828, "global_step": 28153, "epoch": 316, "lr": 9.745168369921154e-05}
{"train_loss": 0.47528183460235596, "global_step": 28154, "epoch": 316, "lr": 9.745150097771725e-05}
{"train_loss": 0.4281161427497864, "global_step": 28155, "epoch": 316, "lr": 9.745131824984369e-05}
{"train_loss": 0.3837184011936188, "global_step": 28156, "epoch": 316, "lr": 9.745113551559087e-05}
{"train_loss": 0.4319756031036377, "global_step": 28157, "epoch": 316, "lr": 9.745095277495883e-05}
{"train_loss": 0.37981218099594116, "global_step": 28158, "epoch": 316, "lr": 9.74507700279476e-05}
{"train_loss": 0.38847780227661133, "global_step": 28159, "epoch": 316, "lr": 9.745058727455719e-05}
{"train_loss": 0.373874694108963, "global_step": 28160, "epoch": 316, "lr": 9.74504045147876e-05}
{"train_loss": 0.3897947072982788, "global_step": 28161, "epoch": 316, "lr": 9.745022174863891e-05}
{"train_loss": 0.41964808106422424, "global_step": 28162, "epoch": 316, "lr": 9.745003897611111e-05}
{"train_loss": 0.3880124092102051, "global_step": 28163, "epoch": 316, "lr": 9.744985619720421e-05}
{"train_loss": 0.45979970693588257, "global_step": 28164, "epoch": 316, "lr": 9.744967341191829e-05}
{"train_loss": 0.5133817195892334, "global_step": 28165, "epoch": 316, "lr": 9.744949062025331e-05}
{"train_loss": 0.3698652684688568, "global_step": 28166, "epoch": 316, "lr": 9.744930782220934e-05}
{"train_loss": 0.3929181396961212, "global_step": 28167, "epoch": 316, "lr": 9.744912501778637e-05}
{"train_loss": 0.37800559401512146, "global_step": 28168, "epoch": 316, "lr": 9.744894220698448e-05}
{"train_loss": 0.5060259699821472, "global_step": 28169, "epoch": 316, "lr": 9.744875938980364e-05}
{"train_loss": 0.4333541691303253, "global_step": 28170, "epoch": 316, "lr": 9.744857656624389e-05}
{"train_loss": 0.367881715297699, "global_step": 28171, "epoch": 316, "lr": 9.744839373630526e-05}
{"train_loss": 0.47656819224357605, "global_step": 28172, "epoch": 316, "lr": 9.744821089998776e-05}
{"train_loss": 0.41286784410476685, "global_step": 28173, "epoch": 316, "lr": 9.744802805729143e-05}
{"train_loss": 0.4641319215297699, "global_step": 28174, "epoch": 316, "lr": 9.74478452082163e-05}
{"train_loss": 0.45542794466018677, "global_step": 28175, "epoch": 316, "lr": 9.744766235276238e-05}
{"train_loss": 0.3491910994052887, "global_step": 28176, "epoch": 316, "lr": 9.74474794909297e-05}
{"train_loss": 0.4196908473968506, "global_step": 28177, "epoch": 316, "lr": 9.744729662271828e-05}
{"train_loss": 0.3257702887058258, "global_step": 28178, "epoch": 316, "lr": 9.744711374812816e-05}
{"train_loss": 0.30871090292930603, "global_step": 28179, "epoch": 316, "lr": 9.744693086715934e-05}
{"train_loss": 0.42228564620018005, "global_step": 28180, "epoch": 316, "lr": 9.744674797981188e-05}
{"train_loss": 0.3014990985393524, "global_step": 28181, "epoch": 316, "lr": 9.744656508608575e-05}
{"train_loss": 0.3786417245864868, "global_step": 28182, "epoch": 316, "lr": 9.744638218598104e-05}
{"train_loss": 0.331905335187912, "global_step": 28183, "epoch": 316, "lr": 9.744619927949774e-05}
{"train_loss": 0.4389568269252777, "global_step": 28184, "epoch": 316, "lr": 9.744601636663586e-05}
{"train_loss": 0.37194716930389404, "global_step": 28185, "epoch": 316, "lr": 9.744583344739545e-05}
{"train_loss": 0.4277407228946686, "global_step": 28186, "epoch": 316, "lr": 9.744565052177651e-05}
{"train_loss": 0.4031458795070648, "global_step": 28187, "epoch": 316, "lr": 9.744546758977911e-05}
{"train_loss": 0.5119572281837463, "global_step": 28188, "epoch": 316, "lr": 9.744528465140324e-05}
{"train_loss": 0.522152304649353, "global_step": 28189, "epoch": 316, "lr": 9.744510170664891e-05}
{"train_loss": 0.41286081075668335, "global_step": 28190, "epoch": 316, "lr": 9.744491875551618e-05}
{"train_loss": 0.5017648339271545, "global_step": 28191, "epoch": 316, "lr": 9.744473579800506e-05}
{"train_loss": 0.27226006984710693, "global_step": 28192, "epoch": 316, "lr": 9.744455283411556e-05}
{"train_loss": 0.36866456270217896, "global_step": 28193, "epoch": 316, "lr": 9.744436986384772e-05}
{"train_loss": 0.3960186243057251, "global_step": 28194, "epoch": 316, "lr": 9.744418688720157e-05}
{"train_loss": 0.4661625325679779, "global_step": 28195, "epoch": 316, "lr": 9.744400390417712e-05}
{"train_loss": 0.5011898279190063, "global_step": 28196, "epoch": 316, "lr": 9.744382091477441e-05}
{"train_loss": 0.3228943347930908, "global_step": 28197, "epoch": 316, "lr": 9.744363791899345e-05}
{"train_loss": 0.34499621391296387, "global_step": 28198, "epoch": 316, "lr": 9.744345491683427e-05}
{"train_loss": 0.38336431980133057, "global_step": 28199, "epoch": 316, "lr": 9.744327190829691e-05}
{"train_loss": 0.3655385971069336, "global_step": 28200, "epoch": 316, "lr": 9.744308889338137e-05}
{"train_loss": 0.3781367242336273, "global_step": 28201, "epoch": 316, "lr": 9.744290587208769e-05}
{"train_loss": 0.47710272669792175, "global_step": 28202, "epoch": 316, "lr": 9.744272284441588e-05}
{"train_loss": 0.32549795508384705, "global_step": 28203, "epoch": 316, "lr": 9.744253981036597e-05}
{"train_loss": 0.27803561091423035, "global_step": 28204, "epoch": 316, "lr": 9.744235676993801e-05}
{"train_loss": 0.45728182792663574, "global_step": 28205, "epoch": 316, "lr": 9.744217372313198e-05}
{"train_loss": 0.43103718757629395, "global_step": 28206, "epoch": 316, "lr": 9.744199066994795e-05}
{"train_loss": 0.3088952898979187, "global_step": 28207, "epoch": 316, "lr": 9.74418076103859e-05}
{"train_loss": 0.44315099716186523, "global_step": 28208, "epoch": 316, "lr": 9.74416245444459e-05}
{"train_loss": 0.4514371156692505, "global_step": 28209, "epoch": 316, "lr": 9.744144147212794e-05}
{"train_loss": 0.386252760887146, "global_step": 28210, "epoch": 316, "lr": 9.744125839343205e-05}
{"train_loss": 0.4471679925918579, "global_step": 28211, "epoch": 316, "lr": 9.744107530835828e-05}
{"train_loss": 0.4143201635124978, "global_step": 28212, "epoch": 316, "lr": 9.744089221690663e-05, "val_loss": 2.1679728031158447}
{"train_loss": 0.29855653643608093, "global_step": 28213, "epoch": 317, "lr": 9.744070911907711e-05}
{"train_loss": 0.4172183871269226, "global_step": 28214, "epoch": 317, "lr": 9.744052601486978e-05}
{"train_loss": 0.35707971453666687, "global_step": 28215, "epoch": 317, "lr": 9.744034290428465e-05}
{"train_loss": 0.419074684381485, "global_step": 28216, "epoch": 317, "lr": 9.744015978732175e-05}
{"train_loss": 0.5094342231750488, "global_step": 28217, "epoch": 317, "lr": 9.743997666398108e-05}
{"train_loss": 0.339621901512146, "global_step": 28218, "epoch": 317, "lr": 9.74397935342627e-05}
{"train_loss": 0.5280330777168274, "global_step": 28219, "epoch": 317, "lr": 9.743961039816661e-05}
{"train_loss": 0.3047347664833069, "global_step": 28220, "epoch": 317, "lr": 9.743942725569285e-05}
{"train_loss": 0.3080110251903534, "global_step": 28221, "epoch": 317, "lr": 9.743924410684143e-05}
{"train_loss": 0.35009074211120605, "global_step": 28222, "epoch": 317, "lr": 9.74390609516124e-05}
{"train_loss": 0.5268430709838867, "global_step": 28223, "epoch": 317, "lr": 9.743887779000573e-05}
{"train_loss": 0.4179084897041321, "global_step": 28224, "epoch": 317, "lr": 9.743869462202151e-05}
{"train_loss": 0.5495361089706421, "global_step": 28225, "epoch": 317, "lr": 9.743851144765972e-05}
{"train_loss": 0.3426300287246704, "global_step": 28226, "epoch": 317, "lr": 9.743832826692041e-05}
{"train_loss": 0.31675782799720764, "global_step": 28227, "epoch": 317, "lr": 9.74381450798036e-05}
{"train_loss": 0.30544745922088623, "global_step": 28228, "epoch": 317, "lr": 9.74379618863093e-05}
{"train_loss": 0.30268558859825134, "global_step": 28229, "epoch": 317, "lr": 9.743777868643755e-05}
{"train_loss": 0.34192031621932983, "global_step": 28230, "epoch": 317, "lr": 9.743759548018836e-05}
{"train_loss": 0.33897265791893005, "global_step": 28231, "epoch": 317, "lr": 9.743741226756177e-05}
{"train_loss": 0.5067622661590576, "global_step": 28232, "epoch": 317, "lr": 9.74372290485578e-05}
{"train_loss": 0.3803662359714508, "global_step": 28233, "epoch": 317, "lr": 9.743704582317648e-05}
{"train_loss": 0.3988029360771179, "global_step": 28234, "epoch": 317, "lr": 9.743686259141782e-05}
{"train_loss": 0.5018860101699829, "global_step": 28235, "epoch": 317, "lr": 9.743667935328185e-05}
{"train_loss": 0.296930730342865, "global_step": 28236, "epoch": 317, "lr": 9.743649610876859e-05}
{"train_loss": 0.4058879017829895, "global_step": 28237, "epoch": 317, "lr": 9.743631285787807e-05}
{"train_loss": 0.47441744804382324, "global_step": 28238, "epoch": 317, "lr": 9.743612960061032e-05}
{"train_loss": 0.5280346870422363, "global_step": 28239, "epoch": 317, "lr": 9.743594633696537e-05}
{"train_loss": 0.4396149814128876, "global_step": 28240, "epoch": 317, "lr": 9.743576306694321e-05}
{"train_loss": 0.3759760856628418, "global_step": 28241, "epoch": 317, "lr": 9.743557979054392e-05}
{"train_loss": 0.33077219128608704, "global_step": 28242, "epoch": 317, "lr": 9.74353965077675e-05}
{"train_loss": 0.4142534136772156, "global_step": 28243, "epoch": 317, "lr": 9.743521321861394e-05}
{"train_loss": 0.4136521518230438, "global_step": 28244, "epoch": 317, "lr": 9.74350299230833e-05}
{"train_loss": 0.27070152759552, "global_step": 28245, "epoch": 317, "lr": 9.74348466211756e-05}
{"train_loss": 0.4963563084602356, "global_step": 28246, "epoch": 317, "lr": 9.743466331289089e-05}
{"train_loss": 0.42080366611480713, "global_step": 28247, "epoch": 317, "lr": 9.743447999822913e-05}
{"train_loss": 0.34835126996040344, "global_step": 28248, "epoch": 317, "lr": 9.74342966771904e-05}
{"train_loss": 0.5033798813819885, "global_step": 28249, "epoch": 317, "lr": 9.74341133497747e-05}
{"train_loss": 0.36771270632743835, "global_step": 28250, "epoch": 317, "lr": 9.743393001598207e-05}
{"train_loss": 0.33699655532836914, "global_step": 28251, "epoch": 317, "lr": 9.743374667581253e-05}
{"train_loss": 0.36441946029663086, "global_step": 28252, "epoch": 317, "lr": 9.743356332926609e-05}
{"train_loss": 0.3996753692626953, "global_step": 28253, "epoch": 317, "lr": 9.743337997634279e-05}
{"train_loss": 0.4201214015483856, "global_step": 28254, "epoch": 317, "lr": 9.743319661704265e-05}
{"train_loss": 0.42057785391807556, "global_step": 28255, "epoch": 317, "lr": 9.743301325136569e-05}
{"train_loss": 0.39315032958984375, "global_step": 28256, "epoch": 317, "lr": 9.743282987931194e-05}
{"train_loss": 0.31590867042541504, "global_step": 28257, "epoch": 317, "lr": 9.743264650088143e-05}
{"train_loss": 0.45145389437675476, "global_step": 28258, "epoch": 317, "lr": 9.743246311607417e-05}
{"train_loss": 0.368810772895813, "global_step": 28259, "epoch": 317, "lr": 9.743227972489021e-05}
{"train_loss": 0.4843275845050812, "global_step": 28260, "epoch": 317, "lr": 9.743209632732955e-05}
{"train_loss": 0.43730664253234863, "global_step": 28261, "epoch": 317, "lr": 9.743191292339222e-05}
{"train_loss": 0.30746132135391235, "global_step": 28262, "epoch": 317, "lr": 9.743172951307824e-05}
{"train_loss": 0.38612136244773865, "global_step": 28263, "epoch": 317, "lr": 9.743154609638764e-05}
{"train_loss": 0.44876721501350403, "global_step": 28264, "epoch": 317, "lr": 9.743136267332048e-05}
{"train_loss": 0.40036800503730774, "global_step": 28265, "epoch": 317, "lr": 9.743117924387671e-05}
{"train_loss": 0.42873039841651917, "global_step": 28266, "epoch": 317, "lr": 9.743099580805641e-05}
{"train_loss": 0.3501853346824646, "global_step": 28267, "epoch": 317, "lr": 9.743081236585961e-05}
{"train_loss": 0.4502480924129486, "global_step": 28268, "epoch": 317, "lr": 9.74306289172863e-05}
{"train_loss": 0.36479452252388, "global_step": 28269, "epoch": 317, "lr": 9.743044546233651e-05}
{"train_loss": 0.4826824367046356, "global_step": 28270, "epoch": 317, "lr": 9.743026200101029e-05}
{"train_loss": 0.46040078997612, "global_step": 28271, "epoch": 317, "lr": 9.743007853330764e-05}
{"train_loss": 0.4178409278392792, "global_step": 28272, "epoch": 317, "lr": 9.74298950592286e-05}
{"train_loss": 0.38464030623435974, "global_step": 28273, "epoch": 317, "lr": 9.742971157877319e-05}
{"train_loss": 0.27237194776535034, "global_step": 28274, "epoch": 317, "lr": 9.742952809194144e-05}
{"train_loss": 0.3355533182621002, "global_step": 28275, "epoch": 317, "lr": 9.742934459873334e-05}
{"train_loss": 0.4603341221809387, "global_step": 28276, "epoch": 317, "lr": 9.742916109914897e-05}
{"train_loss": 0.3508010506629944, "global_step": 28277, "epoch": 317, "lr": 9.742897759318832e-05}
{"train_loss": 0.37558576464653015, "global_step": 28278, "epoch": 317, "lr": 9.742879408085141e-05}
{"train_loss": 0.4105909466743469, "global_step": 28279, "epoch": 317, "lr": 9.74286105621383e-05}
{"train_loss": 0.3089156746864319, "global_step": 28280, "epoch": 317, "lr": 9.742842703704898e-05}
{"train_loss": 0.4824994206428528, "global_step": 28281, "epoch": 317, "lr": 9.742824350558348e-05}
{"train_loss": 0.34949424862861633, "global_step": 28282, "epoch": 317, "lr": 9.742805996774183e-05}
{"train_loss": 0.3074064254760742, "global_step": 28283, "epoch": 317, "lr": 9.742787642352407e-05}
{"train_loss": 0.38895198702812195, "global_step": 28284, "epoch": 317, "lr": 9.74276928729302e-05}
{"train_loss": 0.44421935081481934, "global_step": 28285, "epoch": 317, "lr": 9.742750931596026e-05}
{"train_loss": 0.32084935903549194, "global_step": 28286, "epoch": 317, "lr": 9.742732575261427e-05}
{"train_loss": 0.38787034153938293, "global_step": 28287, "epoch": 317, "lr": 9.742714218289224e-05}
{"train_loss": 0.28476598858833313, "global_step": 28288, "epoch": 317, "lr": 9.742695860679422e-05}
{"train_loss": 0.30348968505859375, "global_step": 28289, "epoch": 317, "lr": 9.742677502432022e-05}
{"train_loss": 0.5456758141517639, "global_step": 28290, "epoch": 317, "lr": 9.742659143547026e-05}
{"train_loss": 0.44848254323005676, "global_step": 28291, "epoch": 317, "lr": 9.742640784024439e-05}
{"train_loss": 0.3534422814846039, "global_step": 28292, "epoch": 317, "lr": 9.74262242386426e-05}
{"train_loss": 0.45420172810554504, "global_step": 28293, "epoch": 317, "lr": 9.742604063066496e-05}
{"train_loss": 0.3054039478302002, "global_step": 28294, "epoch": 317, "lr": 9.742585701631144e-05}
{"train_loss": 0.3988599479198456, "global_step": 28295, "epoch": 317, "lr": 9.74256733955821e-05}
{"train_loss": 0.42901790142059326, "global_step": 28296, "epoch": 317, "lr": 9.742548976847696e-05}
{"train_loss": 0.3465937376022339, "global_step": 28297, "epoch": 317, "lr": 9.742530613499603e-05}
{"train_loss": 0.320246160030365, "global_step": 28298, "epoch": 317, "lr": 9.742512249513935e-05}
{"train_loss": 0.5776481032371521, "global_step": 28299, "epoch": 317, "lr": 9.742493884890695e-05}
{"train_loss": 0.42933031916618347, "global_step": 28300, "epoch": 317, "lr": 9.742475519629884e-05}
{"train_loss": 0.3955765182382605, "global_step": 28301, "epoch": 317, "lr": 9.742457153731505e-05, "val_loss": 2.1431686878204346}
{"train_loss": 0.4178299605846405, "global_step": 28302, "epoch": 318, "lr": 9.74243878719556e-05}
{"train_loss": 0.4292054772377014, "global_step": 28303, "epoch": 318, "lr": 9.742420420022053e-05}
{"train_loss": 0.48615944385528564, "global_step": 28304, "epoch": 318, "lr": 9.742402052210984e-05}
{"train_loss": 0.4968883991241455, "global_step": 28305, "epoch": 318, "lr": 9.742383683762358e-05}
{"train_loss": 0.34949904680252075, "global_step": 28306, "epoch": 318, "lr": 9.742365314676175e-05}
{"train_loss": 0.3855710029602051, "global_step": 28307, "epoch": 318, "lr": 9.74234694495244e-05}
{"train_loss": 0.4103999137878418, "global_step": 28308, "epoch": 318, "lr": 9.742328574591153e-05}
{"train_loss": 0.4265008866786957, "global_step": 28309, "epoch": 318, "lr": 9.74231020359232e-05}
{"train_loss": 0.39009130001068115, "global_step": 28310, "epoch": 318, "lr": 9.742291831955938e-05}
{"train_loss": 0.3576972484588623, "global_step": 28311, "epoch": 318, "lr": 9.742273459682016e-05}
{"train_loss": 0.2956300973892212, "global_step": 28312, "epoch": 318, "lr": 9.742255086770551e-05}
{"train_loss": 0.3068801462650299, "global_step": 28313, "epoch": 318, "lr": 9.742236713221547e-05}
{"train_loss": 0.3055926263332367, "global_step": 28314, "epoch": 318, "lr": 9.74221833903501e-05}
{"train_loss": 0.3677901327610016, "global_step": 28315, "epoch": 318, "lr": 9.742199964210937e-05}
{"train_loss": 0.5728799700737, "global_step": 28316, "epoch": 318, "lr": 9.742181588749332e-05}
{"train_loss": 0.4753783047199249, "global_step": 28317, "epoch": 318, "lr": 9.742163212650202e-05}
{"train_loss": 0.3391411304473877, "global_step": 28318, "epoch": 318, "lr": 9.742144835913544e-05}
{"train_loss": 0.29759955406188965, "global_step": 28319, "epoch": 318, "lr": 9.742126458539362e-05}
{"train_loss": 0.3499532639980316, "global_step": 28320, "epoch": 318, "lr": 9.74210808052766e-05}
{"train_loss": 0.4443320631980896, "global_step": 28321, "epoch": 318, "lr": 9.742089701878439e-05}
{"train_loss": 0.44020336866378784, "global_step": 28322, "epoch": 318, "lr": 9.742071322591702e-05}
{"train_loss": 0.45301729440689087, "global_step": 28323, "epoch": 318, "lr": 9.74205294266745e-05}
{"train_loss": 0.33682385087013245, "global_step": 28324, "epoch": 318, "lr": 9.742034562105688e-05}
{"train_loss": 0.3204156458377838, "global_step": 28325, "epoch": 318, "lr": 9.742016180906417e-05}
{"train_loss": 0.3269740641117096, "global_step": 28326, "epoch": 318, "lr": 9.74199779906964e-05}
{"train_loss": 0.4920952320098877, "global_step": 28327, "epoch": 318, "lr": 9.741979416595359e-05}
{"train_loss": 0.31108009815216064, "global_step": 28328, "epoch": 318, "lr": 9.741961033483577e-05}
{"train_loss": 0.5074557065963745, "global_step": 28329, "epoch": 318, "lr": 9.741942649734295e-05}
{"train_loss": 0.44541651010513306, "global_step": 28330, "epoch": 318, "lr": 9.741924265347518e-05}
{"train_loss": 0.4042797386646271, "global_step": 28331, "epoch": 318, "lr": 9.741905880323245e-05}
{"train_loss": 0.3854984939098358, "global_step": 28332, "epoch": 318, "lr": 9.741887494661484e-05}
{"train_loss": 0.3479768931865692, "global_step": 28333, "epoch": 318, "lr": 9.741869108362232e-05}
{"train_loss": 0.40639087557792664, "global_step": 28334, "epoch": 318, "lr": 9.741850721425491e-05}
{"train_loss": 0.4443526864051819, "global_step": 28335, "epoch": 318, "lr": 9.74183233385127e-05}
{"train_loss": 0.36808812618255615, "global_step": 28336, "epoch": 318, "lr": 9.741813945639566e-05}
{"train_loss": 0.38606366515159607, "global_step": 28337, "epoch": 318, "lr": 9.741795556790384e-05}
{"train_loss": 0.27867573499679565, "global_step": 28338, "epoch": 318, "lr": 9.741777167303723e-05}
{"train_loss": 0.3134138584136963, "global_step": 28339, "epoch": 318, "lr": 9.741758777179588e-05}
{"train_loss": 0.3916197121143341, "global_step": 28340, "epoch": 318, "lr": 9.741740386417983e-05}
{"train_loss": 0.4363381266593933, "global_step": 28341, "epoch": 318, "lr": 9.741721995018909e-05}
{"train_loss": 0.3495129346847534, "global_step": 28342, "epoch": 318, "lr": 9.741703602982366e-05}
{"train_loss": 0.3438465893268585, "global_step": 28343, "epoch": 318, "lr": 9.741685210308361e-05}
{"train_loss": 0.39807769656181335, "global_step": 28344, "epoch": 318, "lr": 9.741666816996893e-05}
{"train_loss": 0.4564218819141388, "global_step": 28345, "epoch": 318, "lr": 9.741648423047966e-05}
{"train_loss": 0.321834921836853, "global_step": 28346, "epoch": 318, "lr": 9.741630028461582e-05}
{"train_loss": 0.42194342613220215, "global_step": 28347, "epoch": 318, "lr": 9.741611633237744e-05}
{"train_loss": 0.3199145793914795, "global_step": 28348, "epoch": 318, "lr": 9.741593237376454e-05}
{"train_loss": 0.33549872040748596, "global_step": 28349, "epoch": 318, "lr": 9.741574840877715e-05}
{"train_loss": 0.3115716874599457, "global_step": 28350, "epoch": 318, "lr": 9.741556443741528e-05}
{"train_loss": 0.36625754833221436, "global_step": 28351, "epoch": 318, "lr": 9.741538045967897e-05}
{"train_loss": 0.4122438430786133, "global_step": 28352, "epoch": 318, "lr": 9.741519647556825e-05}
{"train_loss": 0.39912307262420654, "global_step": 28353, "epoch": 318, "lr": 9.741501248508312e-05}
{"train_loss": 0.28183382749557495, "global_step": 28354, "epoch": 318, "lr": 9.741482848822364e-05}
{"train_loss": 0.38059288263320923, "global_step": 28355, "epoch": 318, "lr": 9.741464448498979e-05}
{"train_loss": 0.41972053050994873, "global_step": 28356, "epoch": 318, "lr": 9.741446047538163e-05}
{"train_loss": 0.3608669638633728, "global_step": 28357, "epoch": 318, "lr": 9.741427645939918e-05}
{"train_loss": 0.3931398093700409, "global_step": 28358, "epoch": 318, "lr": 9.741409243704245e-05}
{"train_loss": 0.45703187584877014, "global_step": 28359, "epoch": 318, "lr": 9.741390840831146e-05}
{"train_loss": 0.5472041368484497, "global_step": 28360, "epoch": 318, "lr": 9.741372437320627e-05}
{"train_loss": 0.39237189292907715, "global_step": 28361, "epoch": 318, "lr": 9.741354033172687e-05}
{"train_loss": 0.262278288602829, "global_step": 28362, "epoch": 318, "lr": 9.741335628387332e-05}
{"train_loss": 0.31570857763290405, "global_step": 28363, "epoch": 318, "lr": 9.74131722296456e-05}
{"train_loss": 0.4986249804496765, "global_step": 28364, "epoch": 318, "lr": 9.741298816904375e-05}
{"train_loss": 0.3735504746437073, "global_step": 28365, "epoch": 318, "lr": 9.741280410206782e-05}
{"train_loss": 0.5208271741867065, "global_step": 28366, "epoch": 318, "lr": 9.74126200287178e-05}
{"train_loss": 0.44710493087768555, "global_step": 28367, "epoch": 318, "lr": 9.741243594899374e-05}
{"train_loss": 0.3828701376914978, "global_step": 28368, "epoch": 318, "lr": 9.741225186289566e-05}
{"train_loss": 0.4187220335006714, "global_step": 28369, "epoch": 318, "lr": 9.741206777042357e-05}
{"train_loss": 0.3765878975391388, "global_step": 28370, "epoch": 318, "lr": 9.741188367157751e-05}
{"train_loss": 0.3760874569416046, "global_step": 28371, "epoch": 318, "lr": 9.741169956635751e-05}
{"train_loss": 0.45629602670669556, "global_step": 28372, "epoch": 318, "lr": 9.741151545476358e-05}
{"train_loss": 0.33669883012771606, "global_step": 28373, "epoch": 318, "lr": 9.741133133679575e-05}
{"train_loss": 0.43867334723472595, "global_step": 28374, "epoch": 318, "lr": 9.741114721245404e-05}
{"train_loss": 0.43869680166244507, "global_step": 28375, "epoch": 318, "lr": 9.741096308173847e-05}
{"train_loss": 0.34719595313072205, "global_step": 28376, "epoch": 318, "lr": 9.741077894464909e-05}
{"train_loss": 0.40401455760002136, "global_step": 28377, "epoch": 318, "lr": 9.74105948011859e-05}
{"train_loss": 0.4789994955062866, "global_step": 28378, "epoch": 318, "lr": 9.741041065134895e-05}
{"train_loss": 0.5207259058952332, "global_step": 28379, "epoch": 318, "lr": 9.741022649513823e-05}
{"train_loss": 0.3522437512874603, "global_step": 28380, "epoch": 318, "lr": 9.741004233255378e-05}
{"train_loss": 0.41489800810813904, "global_step": 28381, "epoch": 318, "lr": 9.740985816359563e-05}
{"train_loss": 0.35428690910339355, "global_step": 28382, "epoch": 318, "lr": 9.740967398826382e-05}
{"train_loss": 0.3431878685951233, "global_step": 28383, "epoch": 318, "lr": 9.740948980655832e-05}
{"train_loss": 0.37927719950675964, "global_step": 28384, "epoch": 318, "lr": 9.740930561847923e-05}
{"train_loss": 0.29902783036231995, "global_step": 28385, "epoch": 318, "lr": 9.740912142402653e-05}
{"train_loss": 0.45566272735595703, "global_step": 28386, "epoch": 318, "lr": 9.740893722320024e-05}
{"train_loss": 0.4440680146217346, "global_step": 28387, "epoch": 318, "lr": 9.740875301600041e-05}
{"train_loss": 0.38126805424690247, "global_step": 28388, "epoch": 318, "lr": 9.740856880242703e-05}
{"train_loss": 0.5230687260627747, "global_step": 28389, "epoch": 318, "lr": 9.740838458248016e-05}
{"train_loss": 0.3956541901224115, "global_step": 28390, "epoch": 318, "lr": 9.74082003561598e-05, "val_loss": 2.1639504432678223}
{"train_loss": 0.43165525794029236, "global_step": 28391, "epoch": 319, "lr": 9.740801612346599e-05}
{"train_loss": 0.4182402491569519, "global_step": 28392, "epoch": 319, "lr": 9.740783188439877e-05}
{"train_loss": 0.3646179139614105, "global_step": 28393, "epoch": 319, "lr": 9.740764763895813e-05}
{"train_loss": 0.4635850787162781, "global_step": 28394, "epoch": 319, "lr": 9.740746338714411e-05}
{"train_loss": 0.39232519268989563, "global_step": 28395, "epoch": 319, "lr": 9.740727912895674e-05}
{"train_loss": 0.4748484194278717, "global_step": 28396, "epoch": 319, "lr": 9.740709486439601e-05}
{"train_loss": 0.44309568405151367, "global_step": 28397, "epoch": 319, "lr": 9.7406910593462e-05}
{"train_loss": 0.3107462525367737, "global_step": 28398, "epoch": 319, "lr": 9.74067263161547e-05}
{"train_loss": 0.3222891688346863, "global_step": 28399, "epoch": 319, "lr": 9.740654203247414e-05}
{"train_loss": 0.44604921340942383, "global_step": 28400, "epoch": 319, "lr": 9.740635774242037e-05}
{"train_loss": 0.38587698340415955, "global_step": 28401, "epoch": 319, "lr": 9.740617344599338e-05}
{"train_loss": 0.36379674077033997, "global_step": 28402, "epoch": 319, "lr": 9.74059891431932e-05}
{"train_loss": 0.4044523537158966, "global_step": 28403, "epoch": 319, "lr": 9.740580483401987e-05}
{"train_loss": 0.3910975754261017, "global_step": 28404, "epoch": 319, "lr": 9.74056205184734e-05}
{"train_loss": 0.4585285782814026, "global_step": 28405, "epoch": 319, "lr": 9.740543619655383e-05}
{"train_loss": 0.4429372251033783, "global_step": 28406, "epoch": 319, "lr": 9.740525186826117e-05}
{"train_loss": 0.36782872676849365, "global_step": 28407, "epoch": 319, "lr": 9.740506753359546e-05}
{"train_loss": 0.45718008279800415, "global_step": 28408, "epoch": 319, "lr": 9.740488319255672e-05}
{"train_loss": 0.4222795069217682, "global_step": 28409, "epoch": 319, "lr": 9.740469884514497e-05}
{"train_loss": 0.5273908376693726, "global_step": 28410, "epoch": 319, "lr": 9.740451449136023e-05}
{"train_loss": 0.4389321506023407, "global_step": 28411, "epoch": 319, "lr": 9.740433013120253e-05}
{"train_loss": 0.37567493319511414, "global_step": 28412, "epoch": 319, "lr": 9.740414576467189e-05}
{"train_loss": 0.35948312282562256, "global_step": 28413, "epoch": 319, "lr": 9.740396139176835e-05}
{"train_loss": 0.4654141664505005, "global_step": 28414, "epoch": 319, "lr": 9.740377701249192e-05}
{"train_loss": 0.4086148142814636, "global_step": 28415, "epoch": 319, "lr": 9.740359262684265e-05}
{"train_loss": 0.38712218403816223, "global_step": 28416, "epoch": 319, "lr": 9.740340823482052e-05}
{"train_loss": 0.5746726989746094, "global_step": 28417, "epoch": 319, "lr": 9.740322383642557e-05}
{"train_loss": 0.3218798339366913, "global_step": 28418, "epoch": 319, "lr": 9.740303943165787e-05}
{"train_loss": 0.3679143786430359, "global_step": 28419, "epoch": 319, "lr": 9.740285502051739e-05}
{"train_loss": 0.4306281805038452, "global_step": 28420, "epoch": 319, "lr": 9.740267060300417e-05}
{"train_loss": 0.3086775839328766, "global_step": 28421, "epoch": 319, "lr": 9.740248617911823e-05}
{"train_loss": 0.42264556884765625, "global_step": 28422, "epoch": 319, "lr": 9.740230174885962e-05}
{"train_loss": 0.380164235830307, "global_step": 28423, "epoch": 319, "lr": 9.740211731222835e-05}
{"train_loss": 0.30138498544692993, "global_step": 28424, "epoch": 319, "lr": 9.740193286922444e-05}
{"train_loss": 0.4562004804611206, "global_step": 28425, "epoch": 319, "lr": 9.74017484198479e-05}
{"train_loss": 0.43969815969467163, "global_step": 28426, "epoch": 319, "lr": 9.740156396409881e-05}
{"train_loss": 0.43131792545318604, "global_step": 28427, "epoch": 319, "lr": 9.740137950197712e-05}
{"train_loss": 0.3633759915828705, "global_step": 28428, "epoch": 319, "lr": 9.74011950334829e-05}
{"train_loss": 0.45879238843917847, "global_step": 28429, "epoch": 319, "lr": 9.74010105586162e-05}
{"train_loss": 0.43527400493621826, "global_step": 28430, "epoch": 319, "lr": 9.740082607737698e-05}
{"train_loss": 0.4493488669395447, "global_step": 28431, "epoch": 319, "lr": 9.740064158976529e-05}
{"train_loss": 0.3384931981563568, "global_step": 28432, "epoch": 319, "lr": 9.740045709578118e-05}
{"train_loss": 0.3661099374294281, "global_step": 28433, "epoch": 319, "lr": 9.740027259542466e-05}
{"train_loss": 0.2586899697780609, "global_step": 28434, "epoch": 319, "lr": 9.740008808869574e-05}
{"train_loss": 0.3764544427394867, "global_step": 28435, "epoch": 319, "lr": 9.739990357559445e-05}
{"train_loss": 0.44929611682891846, "global_step": 28436, "epoch": 319, "lr": 9.739971905612085e-05}
{"train_loss": 0.3228418231010437, "global_step": 28437, "epoch": 319, "lr": 9.739953453027491e-05}
{"train_loss": 0.3329135775566101, "global_step": 28438, "epoch": 319, "lr": 9.739934999805667e-05}
{"train_loss": 0.4752306342124939, "global_step": 28439, "epoch": 319, "lr": 9.739916545946619e-05}
{"train_loss": 0.4172382652759552, "global_step": 28440, "epoch": 319, "lr": 9.739898091450345e-05}
{"train_loss": 0.3650250732898712, "global_step": 28441, "epoch": 319, "lr": 9.739879636316852e-05}
{"train_loss": 0.4003257155418396, "global_step": 28442, "epoch": 319, "lr": 9.739861180546137e-05}
{"train_loss": 0.4362452030181885, "global_step": 28443, "epoch": 319, "lr": 9.739842724138206e-05}
{"train_loss": 0.3201849162578583, "global_step": 28444, "epoch": 319, "lr": 9.739824267093062e-05}
{"train_loss": 0.3815670311450958, "global_step": 28445, "epoch": 319, "lr": 9.739805809410707e-05}
{"train_loss": 0.5326294302940369, "global_step": 28446, "epoch": 319, "lr": 9.739787351091142e-05}
{"train_loss": 0.4107186496257782, "global_step": 28447, "epoch": 319, "lr": 9.73976889213437e-05}
{"train_loss": 0.4008098542690277, "global_step": 28448, "epoch": 319, "lr": 9.739750432540394e-05}
{"train_loss": 0.2570471167564392, "global_step": 28449, "epoch": 319, "lr": 9.739731972309216e-05}
{"train_loss": 0.5652822256088257, "global_step": 28450, "epoch": 319, "lr": 9.73971351144084e-05}
{"train_loss": 0.3274669945240021, "global_step": 28451, "epoch": 319, "lr": 9.739695049935267e-05}
{"train_loss": 0.3693932890892029, "global_step": 28452, "epoch": 319, "lr": 9.739676587792499e-05}
{"train_loss": 0.3432389497756958, "global_step": 28453, "epoch": 319, "lr": 9.739658125012541e-05}
{"train_loss": 0.31236475706100464, "global_step": 28454, "epoch": 319, "lr": 9.739639661595392e-05}
{"train_loss": 0.3660876750946045, "global_step": 28455, "epoch": 319, "lr": 9.739621197541056e-05}
{"train_loss": 0.3673865795135498, "global_step": 28456, "epoch": 319, "lr": 9.739602732849537e-05}
{"train_loss": 0.39588698744773865, "global_step": 28457, "epoch": 319, "lr": 9.739584267520835e-05}
{"train_loss": 0.4318315386772156, "global_step": 28458, "epoch": 319, "lr": 9.739565801554954e-05}
{"train_loss": 0.3707412779331207, "global_step": 28459, "epoch": 319, "lr": 9.739547334951896e-05}
{"train_loss": 0.3466263711452484, "global_step": 28460, "epoch": 319, "lr": 9.739528867711665e-05}
{"train_loss": 0.36373478174209595, "global_step": 28461, "epoch": 319, "lr": 9.73951039983426e-05}
{"train_loss": 0.3041858375072479, "global_step": 28462, "epoch": 319, "lr": 9.739491931319687e-05}
{"train_loss": 0.3054778575897217, "global_step": 28463, "epoch": 319, "lr": 9.739473462167947e-05}
{"train_loss": 0.4299618899822235, "global_step": 28464, "epoch": 319, "lr": 9.739454992379042e-05}
{"train_loss": 0.3239209055900574, "global_step": 28465, "epoch": 319, "lr": 9.739436521952976e-05}
{"train_loss": 0.39784175157546997, "global_step": 28466, "epoch": 319, "lr": 9.73941805088975e-05}
{"train_loss": 0.3963755667209625, "global_step": 28467, "epoch": 319, "lr": 9.739399579189365e-05}
{"train_loss": 0.3463931679725647, "global_step": 28468, "epoch": 319, "lr": 9.739381106851828e-05}
{"train_loss": 0.34461402893066406, "global_step": 28469, "epoch": 319, "lr": 9.739362633877138e-05}
{"train_loss": 0.37930750846862793, "global_step": 28470, "epoch": 319, "lr": 9.739344160265299e-05}
{"train_loss": 0.46554863452911377, "global_step": 28471, "epoch": 319, "lr": 9.739325686016312e-05}
{"train_loss": 0.3017621636390686, "global_step": 28472, "epoch": 319, "lr": 9.739307211130181e-05}
{"train_loss": 0.33885177969932556, "global_step": 28473, "epoch": 319, "lr": 9.739288735606908e-05}
{"train_loss": 0.36322230100631714, "global_step": 28474, "epoch": 319, "lr": 9.739270259446497e-05}
{"train_loss": 0.3736382722854614, "global_step": 28475, "epoch": 319, "lr": 9.739251782648947e-05}
{"train_loss": 0.4467017352581024, "global_step": 28476, "epoch": 319, "lr": 9.739233305214262e-05}
{"train_loss": 0.43148383498191833, "global_step": 28477, "epoch": 319, "lr": 9.739214827142447e-05}
{"train_loss": 0.35214951634407043, "global_step": 28478, "epoch": 319, "lr": 9.7391963484335e-05}
{"train_loss": 0.3920436839709121, "global_step": 28479, "epoch": 319, "lr": 9.739177869087426e-05, "val_loss": 2.2345337867736816}
{"train_loss": 0.4142788350582123, "global_step": 28480, "epoch": 320, "lr": 9.739159389104229e-05}
{"train_loss": 0.44767969846725464, "global_step": 28481, "epoch": 320, "lr": 9.739140908483909e-05}
{"train_loss": 0.36499106884002686, "global_step": 28482, "epoch": 320, "lr": 9.739122427226468e-05}
{"train_loss": 0.2879381775856018, "global_step": 28483, "epoch": 320, "lr": 9.739103945331911e-05}
{"train_loss": 0.43194952607154846, "global_step": 28484, "epoch": 320, "lr": 9.739085462800239e-05}
{"train_loss": 0.32573243975639343, "global_step": 28485, "epoch": 320, "lr": 9.739066979631455e-05}
{"train_loss": 0.3229742646217346, "global_step": 28486, "epoch": 320, "lr": 9.73904849582556e-05}
{"train_loss": 0.3771663308143616, "global_step": 28487, "epoch": 320, "lr": 9.73903001138256e-05}
{"train_loss": 0.4177180230617523, "global_step": 28488, "epoch": 320, "lr": 9.739011526302452e-05}
{"train_loss": 0.30190983414649963, "global_step": 28489, "epoch": 320, "lr": 9.738993040585244e-05}
{"train_loss": 0.3778672218322754, "global_step": 28490, "epoch": 320, "lr": 9.738974554230936e-05}
{"train_loss": 0.2808924615383148, "global_step": 28491, "epoch": 320, "lr": 9.738956067239531e-05}
{"train_loss": 0.39895349740982056, "global_step": 28492, "epoch": 320, "lr": 9.73893757961103e-05}
{"train_loss": 0.3224574625492096, "global_step": 28493, "epoch": 320, "lr": 9.738919091345437e-05}
{"train_loss": 0.37082672119140625, "global_step": 28494, "epoch": 320, "lr": 9.738900602442754e-05}
{"train_loss": 0.5808457732200623, "global_step": 28495, "epoch": 320, "lr": 9.738882112902983e-05}
{"train_loss": 0.3224186897277832, "global_step": 28496, "epoch": 320, "lr": 9.738863622726129e-05}
{"train_loss": 0.3632484972476959, "global_step": 28497, "epoch": 320, "lr": 9.738845131912191e-05}
{"train_loss": 0.38431471586227417, "global_step": 28498, "epoch": 320, "lr": 9.738826640461173e-05}
{"train_loss": 0.3374491035938263, "global_step": 28499, "epoch": 320, "lr": 9.738808148373077e-05}
{"train_loss": 0.3665160834789276, "global_step": 28500, "epoch": 320, "lr": 9.738789655647906e-05}
{"train_loss": 0.2650299370288849, "global_step": 28501, "epoch": 320, "lr": 9.738771162285665e-05}
{"train_loss": 0.40519171953201294, "global_step": 28502, "epoch": 320, "lr": 9.738752668286352e-05}
{"train_loss": 0.4016723334789276, "global_step": 28503, "epoch": 320, "lr": 9.73873417364997e-05}
{"train_loss": 0.480392187833786, "global_step": 28504, "epoch": 320, "lr": 9.738715678376526e-05}
{"train_loss": 0.3377229869365692, "global_step": 28505, "epoch": 320, "lr": 9.738697182466018e-05}
{"train_loss": 0.28583505749702454, "global_step": 28506, "epoch": 320, "lr": 9.738678685918449e-05}
{"train_loss": 0.22814428806304932, "global_step": 28507, "epoch": 320, "lr": 9.738660188733823e-05}
{"train_loss": 0.43366435170173645, "global_step": 28508, "epoch": 320, "lr": 9.738641690912143e-05}
{"train_loss": 0.3634253144264221, "global_step": 28509, "epoch": 320, "lr": 9.738623192453411e-05}
{"train_loss": 0.4047342836856842, "global_step": 28510, "epoch": 320, "lr": 9.738604693357627e-05}
{"train_loss": 0.3549274504184723, "global_step": 28511, "epoch": 320, "lr": 9.738586193624795e-05}
{"train_loss": 0.35619881749153137, "global_step": 28512, "epoch": 320, "lr": 9.73856769325492e-05}
{"train_loss": 0.41903063654899597, "global_step": 28513, "epoch": 320, "lr": 9.738549192248001e-05}
{"train_loss": 0.38399505615234375, "global_step": 28514, "epoch": 320, "lr": 9.738530690604042e-05}
{"train_loss": 0.405757337808609, "global_step": 28515, "epoch": 320, "lr": 9.738512188323044e-05}
{"train_loss": 0.3862961530685425, "global_step": 28516, "epoch": 320, "lr": 9.738493685405012e-05}
{"train_loss": 0.348843514919281, "global_step": 28517, "epoch": 320, "lr": 9.738475181849948e-05}
{"train_loss": 0.3203915059566498, "global_step": 28518, "epoch": 320, "lr": 9.738456677657853e-05}
{"train_loss": 0.4592113792896271, "global_step": 28519, "epoch": 320, "lr": 9.73843817282873e-05}
{"train_loss": 0.39132440090179443, "global_step": 28520, "epoch": 320, "lr": 9.738419667362582e-05}
{"train_loss": 0.3055788278579712, "global_step": 28521, "epoch": 320, "lr": 9.738401161259412e-05}
{"train_loss": 0.3995856046676636, "global_step": 28522, "epoch": 320, "lr": 9.73838265451922e-05}
{"train_loss": 0.49119463562965393, "global_step": 28523, "epoch": 320, "lr": 9.738364147142012e-05}
{"train_loss": 0.28647443652153015, "global_step": 28524, "epoch": 320, "lr": 9.738345639127788e-05}
{"train_loss": 0.45346635580062866, "global_step": 28525, "epoch": 320, "lr": 9.73832713047655e-05}
{"train_loss": 0.4122028648853302, "global_step": 28526, "epoch": 320, "lr": 9.738308621188304e-05}
{"train_loss": 0.4564518928527832, "global_step": 28527, "epoch": 320, "lr": 9.738290111263049e-05}
{"train_loss": 0.3094492554664612, "global_step": 28528, "epoch": 320, "lr": 9.738271600700789e-05}
{"train_loss": 0.3002236485481262, "global_step": 28529, "epoch": 320, "lr": 9.738253089501523e-05}
{"train_loss": 0.43786460161209106, "global_step": 28530, "epoch": 320, "lr": 9.73823457766526e-05}
{"train_loss": 0.39377084374427795, "global_step": 28531, "epoch": 320, "lr": 9.738216065191999e-05}
{"train_loss": 0.4991273581981659, "global_step": 28532, "epoch": 320, "lr": 9.738197552081743e-05}
{"train_loss": 0.4105623960494995, "global_step": 28533, "epoch": 320, "lr": 9.738179038334492e-05}
{"train_loss": 0.3677569329738617, "global_step": 28534, "epoch": 320, "lr": 9.738160523950251e-05}
{"train_loss": 0.42862772941589355, "global_step": 28535, "epoch": 320, "lr": 9.738142008929023e-05}
{"train_loss": 0.3144416809082031, "global_step": 28536, "epoch": 320, "lr": 9.73812349327081e-05}
{"train_loss": 0.3822038471698761, "global_step": 28537, "epoch": 320, "lr": 9.738104976975613e-05}
{"train_loss": 0.46314141154289246, "global_step": 28538, "epoch": 320, "lr": 9.738086460043436e-05}
{"train_loss": 0.4350292682647705, "global_step": 28539, "epoch": 320, "lr": 9.73806794247428e-05}
{"train_loss": 0.3963334560394287, "global_step": 28540, "epoch": 320, "lr": 9.73804942426815e-05}
{"train_loss": 0.37598639726638794, "global_step": 28541, "epoch": 320, "lr": 9.738030905425046e-05}
{"train_loss": 0.4304558038711548, "global_step": 28542, "epoch": 320, "lr": 9.738012385944972e-05}
{"train_loss": 0.33911553025245667, "global_step": 28543, "epoch": 320, "lr": 9.73799386582793e-05}
{"train_loss": 0.4435993432998657, "global_step": 28544, "epoch": 320, "lr": 9.737975345073922e-05}
{"train_loss": 0.35072141885757446, "global_step": 28545, "epoch": 320, "lr": 9.737956823682952e-05}
{"train_loss": 0.3872009515762329, "global_step": 28546, "epoch": 320, "lr": 9.73793830165502e-05}
{"train_loss": 0.43863344192504883, "global_step": 28547, "epoch": 320, "lr": 9.737919778990131e-05}
{"train_loss": 0.3342325985431671, "global_step": 28548, "epoch": 320, "lr": 9.737901255688286e-05}
{"train_loss": 0.4783172011375427, "global_step": 28549, "epoch": 320, "lr": 9.737882731749488e-05}
{"train_loss": 0.2931271493434906, "global_step": 28550, "epoch": 320, "lr": 9.737864207173739e-05}
{"train_loss": 0.335796058177948, "global_step": 28551, "epoch": 320, "lr": 9.737845681961043e-05}
{"train_loss": 0.2881335914134979, "global_step": 28552, "epoch": 320, "lr": 9.737827156111401e-05}
{"train_loss": 0.30170896649360657, "global_step": 28553, "epoch": 320, "lr": 9.737808629624816e-05}
{"train_loss": 0.4460371732711792, "global_step": 28554, "epoch": 320, "lr": 9.73779010250129e-05}
{"train_loss": 0.3269422948360443, "global_step": 28555, "epoch": 320, "lr": 9.737771574740825e-05}
{"train_loss": 0.46723616123199463, "global_step": 28556, "epoch": 320, "lr": 9.737753046343425e-05}
{"train_loss": 0.3260950744152069, "global_step": 28557, "epoch": 320, "lr": 9.737734517309093e-05}
{"train_loss": 0.409931480884552, "global_step": 28558, "epoch": 320, "lr": 9.73771598763783e-05}
{"train_loss": 0.27719250321388245, "global_step": 28559, "epoch": 320, "lr": 9.737697457329637e-05}
{"train_loss": 0.483165979385376, "global_step": 28560, "epoch": 320, "lr": 9.737678926384521e-05}
{"train_loss": 0.4012851417064667, "global_step": 28561, "epoch": 320, "lr": 9.73766039480248e-05}
{"train_loss": 0.4716740846633911, "global_step": 28562, "epoch": 320, "lr": 9.737641862583518e-05}
{"train_loss": 0.38430947065353394, "global_step": 28563, "epoch": 320, "lr": 9.737623329727639e-05}
{"train_loss": 0.4144152104854584, "global_step": 28564, "epoch": 320, "lr": 9.737604796234843e-05}
{"train_loss": 0.39252862334251404, "global_step": 28565, "epoch": 320, "lr": 9.737586262105135e-05}
{"train_loss": 0.3701910674571991, "global_step": 28566, "epoch": 320, "lr": 9.737567727338517e-05}
{"train_loss": 0.41457366943359375, "global_step": 28567, "epoch": 320, "lr": 9.73754919193499e-05}
{"train_loss": 0.38229219163401745, "global_step": 28568, "epoch": 320, "lr": 9.737530655894556e-05, "val_loss": 2.181657075881958, "train_action_mse_error": 24.272415161132812}
{"train_loss": 0.38285359740257263, "global_step": 28569, "epoch": 321, "lr": 9.737512119217219e-05}
{"train_loss": 0.3068527281284332, "global_step": 28570, "epoch": 321, "lr": 9.737493581902982e-05}
{"train_loss": 0.4004855155944824, "global_step": 28571, "epoch": 321, "lr": 9.737475043951848e-05}
{"train_loss": 0.4121711254119873, "global_step": 28572, "epoch": 321, "lr": 9.737456505363815e-05}
{"train_loss": 0.29877200722694397, "global_step": 28573, "epoch": 321, "lr": 9.737437966138891e-05}
{"train_loss": 0.44317445158958435, "global_step": 28574, "epoch": 321, "lr": 9.737419426277075e-05}
{"train_loss": 0.41570520401000977, "global_step": 28575, "epoch": 321, "lr": 9.737400885778372e-05}
{"train_loss": 0.35407862067222595, "global_step": 28576, "epoch": 321, "lr": 9.737382344642783e-05}
{"train_loss": 0.4319634735584259, "global_step": 28577, "epoch": 321, "lr": 9.737363802870309e-05}
{"train_loss": 0.36886319518089294, "global_step": 28578, "epoch": 321, "lr": 9.737345260460956e-05}
{"train_loss": 0.39640286564826965, "global_step": 28579, "epoch": 321, "lr": 9.737326717414724e-05}
{"train_loss": 0.3248254954814911, "global_step": 28580, "epoch": 321, "lr": 9.737308173731617e-05}
{"train_loss": 0.43225353956222534, "global_step": 28581, "epoch": 321, "lr": 9.737289629411636e-05}
{"train_loss": 0.3154435455799103, "global_step": 28582, "epoch": 321, "lr": 9.737271084454785e-05}
{"train_loss": 0.35907846689224243, "global_step": 28583, "epoch": 321, "lr": 9.737252538861063e-05}
{"train_loss": 0.2683001756668091, "global_step": 28584, "epoch": 321, "lr": 9.737233992630477e-05}
{"train_loss": 0.3312836289405823, "global_step": 28585, "epoch": 321, "lr": 9.737215445763028e-05}
{"train_loss": 0.33716633915901184, "global_step": 28586, "epoch": 321, "lr": 9.737196898258718e-05}
{"train_loss": 0.4025556147098541, "global_step": 28587, "epoch": 321, "lr": 9.737178350117548e-05}
{"train_loss": 0.3414957523345947, "global_step": 28588, "epoch": 321, "lr": 9.737159801339523e-05}
{"train_loss": 0.37856677174568176, "global_step": 28589, "epoch": 321, "lr": 9.737141251924646e-05}
{"train_loss": 0.35710880160331726, "global_step": 28590, "epoch": 321, "lr": 9.737122701872917e-05}
{"train_loss": 0.48582321405410767, "global_step": 28591, "epoch": 321, "lr": 9.73710415118434e-05}
{"train_loss": 0.36137527227401733, "global_step": 28592, "epoch": 321, "lr": 9.737085599858917e-05}
{"train_loss": 0.40669772028923035, "global_step": 28593, "epoch": 321, "lr": 9.73706704789665e-05}
{"train_loss": 0.41892126202583313, "global_step": 28594, "epoch": 321, "lr": 9.737048495297543e-05}
{"train_loss": 0.35483500361442566, "global_step": 28595, "epoch": 321, "lr": 9.737029942061596e-05}
{"train_loss": 0.32731351256370544, "global_step": 28596, "epoch": 321, "lr": 9.737011388188814e-05}
{"train_loss": 0.4205014407634735, "global_step": 28597, "epoch": 321, "lr": 9.7369928336792e-05}
{"train_loss": 0.3906400501728058, "global_step": 28598, "epoch": 321, "lr": 9.736974278532752e-05}
{"train_loss": 0.3591843545436859, "global_step": 28599, "epoch": 321, "lr": 9.736955722749477e-05}
{"train_loss": 0.4680129885673523, "global_step": 28600, "epoch": 321, "lr": 9.736937166329376e-05}
{"train_loss": 0.3300834000110626, "global_step": 28601, "epoch": 321, "lr": 9.736918609272452e-05}
{"train_loss": 0.39964014291763306, "global_step": 28602, "epoch": 321, "lr": 9.736900051578707e-05}
{"train_loss": 0.4280222952365875, "global_step": 28603, "epoch": 321, "lr": 9.736881493248143e-05}
{"train_loss": 0.4073435366153717, "global_step": 28604, "epoch": 321, "lr": 9.736862934280764e-05}
{"train_loss": 0.2969018518924713, "global_step": 28605, "epoch": 321, "lr": 9.73684437467657e-05}
{"train_loss": 0.4112977683544159, "global_step": 28606, "epoch": 321, "lr": 9.736825814435566e-05}
{"train_loss": 0.4242071509361267, "global_step": 28607, "epoch": 321, "lr": 9.736807253557754e-05}
{"train_loss": 0.34156638383865356, "global_step": 28608, "epoch": 321, "lr": 9.736788692043135e-05}
{"train_loss": 0.4063336253166199, "global_step": 28609, "epoch": 321, "lr": 9.736770129891712e-05}
{"train_loss": 0.41842079162597656, "global_step": 28610, "epoch": 321, "lr": 9.73675156710349e-05}
{"train_loss": 0.36061450839042664, "global_step": 28611, "epoch": 321, "lr": 9.736733003678469e-05}
{"train_loss": 0.4355909526348114, "global_step": 28612, "epoch": 321, "lr": 9.73671443961665e-05}
{"train_loss": 0.39967718720436096, "global_step": 28613, "epoch": 321, "lr": 9.736695874918038e-05}
{"train_loss": 0.3698885440826416, "global_step": 28614, "epoch": 321, "lr": 9.736677309582637e-05}
{"train_loss": 0.3371310234069824, "global_step": 28615, "epoch": 321, "lr": 9.736658743610446e-05}
{"train_loss": 0.4791482090950012, "global_step": 28616, "epoch": 321, "lr": 9.73664017700147e-05}
{"train_loss": 0.5062832832336426, "global_step": 28617, "epoch": 321, "lr": 9.736621609755709e-05}
{"train_loss": 0.4072628915309906, "global_step": 28618, "epoch": 321, "lr": 9.736603041873166e-05}
{"train_loss": 0.4338257610797882, "global_step": 28619, "epoch": 321, "lr": 9.736584473353846e-05}
{"train_loss": 0.33043354749679565, "global_step": 28620, "epoch": 321, "lr": 9.736565904197751e-05}
{"train_loss": 0.31545522809028625, "global_step": 28621, "epoch": 321, "lr": 9.736547334404881e-05}
{"train_loss": 0.32364603877067566, "global_step": 28622, "epoch": 321, "lr": 9.736528763975242e-05}
{"train_loss": 0.2903916835784912, "global_step": 28623, "epoch": 321, "lr": 9.736510192908833e-05}
{"train_loss": 0.40893417596817017, "global_step": 28624, "epoch": 321, "lr": 9.736491621205657e-05}
{"train_loss": 0.40862178802490234, "global_step": 28625, "epoch": 321, "lr": 9.736473048865718e-05}
{"train_loss": 0.47532525658607483, "global_step": 28626, "epoch": 321, "lr": 9.736454475889018e-05}
{"train_loss": 0.3501191735267639, "global_step": 28627, "epoch": 321, "lr": 9.736435902275559e-05}
{"train_loss": 0.4130772352218628, "global_step": 28628, "epoch": 321, "lr": 9.736417328025344e-05}
{"train_loss": 0.43060171604156494, "global_step": 28629, "epoch": 321, "lr": 9.736398753138376e-05}
{"train_loss": 0.37656399607658386, "global_step": 28630, "epoch": 321, "lr": 9.736380177614657e-05}
{"train_loss": 0.362663596868515, "global_step": 28631, "epoch": 321, "lr": 9.736361601454189e-05}
{"train_loss": 0.358957439661026, "global_step": 28632, "epoch": 321, "lr": 9.736343024656976e-05}
{"train_loss": 0.4217914342880249, "global_step": 28633, "epoch": 321, "lr": 9.736324447223018e-05}
{"train_loss": 0.4073411822319031, "global_step": 28634, "epoch": 321, "lr": 9.73630586915232e-05}
{"train_loss": 0.42072346806526184, "global_step": 28635, "epoch": 321, "lr": 9.736287290444882e-05}
{"train_loss": 0.3368473947048187, "global_step": 28636, "epoch": 321, "lr": 9.736268711100708e-05}
{"train_loss": 0.3557601571083069, "global_step": 28637, "epoch": 321, "lr": 9.736250131119802e-05}
{"train_loss": 0.44063466787338257, "global_step": 28638, "epoch": 321, "lr": 9.736231550502164e-05}
{"train_loss": 0.34531548619270325, "global_step": 28639, "epoch": 321, "lr": 9.736212969247798e-05}
{"train_loss": 0.31266090273857117, "global_step": 28640, "epoch": 321, "lr": 9.736194387356705e-05}
{"train_loss": 0.4219788908958435, "global_step": 28641, "epoch": 321, "lr": 9.736175804828889e-05}
{"train_loss": 0.3659438490867615, "global_step": 28642, "epoch": 321, "lr": 9.73615722166435e-05}
{"train_loss": 0.3039546012878418, "global_step": 28643, "epoch": 321, "lr": 9.736138637863095e-05}
{"train_loss": 0.4174501597881317, "global_step": 28644, "epoch": 321, "lr": 9.736120053425122e-05}
{"train_loss": 0.2764095366001129, "global_step": 28645, "epoch": 321, "lr": 9.736101468350438e-05}
{"train_loss": 0.36299359798431396, "global_step": 28646, "epoch": 321, "lr": 9.736082882639039e-05}
{"train_loss": 0.380218505859375, "global_step": 28647, "epoch": 321, "lr": 9.736064296290934e-05}
{"train_loss": 0.4163970947265625, "global_step": 28648, "epoch": 321, "lr": 9.736045709306123e-05}
{"train_loss": 0.3432120084762573, "global_step": 28649, "epoch": 321, "lr": 9.736027121684608e-05}
{"train_loss": 0.4074218273162842, "global_step": 28650, "epoch": 321, "lr": 9.736008533426391e-05}
{"train_loss": 0.38851091265678406, "global_step": 28651, "epoch": 321, "lr": 9.735989944531476e-05}
{"train_loss": 0.330441951751709, "global_step": 28652, "epoch": 321, "lr": 9.735971354999864e-05}
{"train_loss": 0.46116212010383606, "global_step": 28653, "epoch": 321, "lr": 9.73595276483156e-05}
{"train_loss": 0.30112069845199585, "global_step": 28654, "epoch": 321, "lr": 9.735934174026564e-05}
{"train_loss": 0.3637625575065613, "global_step": 28655, "epoch": 321, "lr": 9.735915582584881e-05}
{"train_loss": 0.4323088228702545, "global_step": 28656, "epoch": 321, "lr": 9.735896990506509e-05}
{"train_loss": 0.3806512791119265, "global_step": 28657, "epoch": 321, "lr": 9.735878397791454e-05, "val_loss": 2.134012460708618}
{"train_loss": 0.34393399953842163, "global_step": 28658, "epoch": 322, "lr": 9.735859804439719e-05}
{"train_loss": 0.32000160217285156, "global_step": 28659, "epoch": 322, "lr": 9.735841210451307e-05}
{"train_loss": 0.396697461605072, "global_step": 28660, "epoch": 322, "lr": 9.735822615826216e-05}
{"train_loss": 0.26969024538993835, "global_step": 28661, "epoch": 322, "lr": 9.735804020564452e-05}
{"train_loss": 0.40204474329948425, "global_step": 28662, "epoch": 322, "lr": 9.735785424666017e-05}
{"train_loss": 0.3973534405231476, "global_step": 28663, "epoch": 322, "lr": 9.735766828130914e-05}
{"train_loss": 0.33997753262519836, "global_step": 28664, "epoch": 322, "lr": 9.735748230959143e-05}
{"train_loss": 0.3603411614894867, "global_step": 28665, "epoch": 322, "lr": 9.73572963315071e-05}
{"train_loss": 0.2726067304611206, "global_step": 28666, "epoch": 322, "lr": 9.735711034705616e-05}
{"train_loss": 0.44604694843292236, "global_step": 28667, "epoch": 322, "lr": 9.735692435623863e-05}
{"train_loss": 0.3818659782409668, "global_step": 28668, "epoch": 322, "lr": 9.735673835905454e-05}
{"train_loss": 0.36567968130111694, "global_step": 28669, "epoch": 322, "lr": 9.73565523555039e-05}
{"train_loss": 0.48083093762397766, "global_step": 28670, "epoch": 322, "lr": 9.735636634558677e-05}
{"train_loss": 0.36181899905204773, "global_step": 28671, "epoch": 322, "lr": 9.735618032930313e-05}
{"train_loss": 0.48074185848236084, "global_step": 28672, "epoch": 322, "lr": 9.735599430665305e-05}
{"train_loss": 0.4410342574119568, "global_step": 28673, "epoch": 322, "lr": 9.735580827763653e-05}
{"train_loss": 0.3422892689704895, "global_step": 28674, "epoch": 322, "lr": 9.735562224225359e-05}
{"train_loss": 0.2648817002773285, "global_step": 28675, "epoch": 322, "lr": 9.735543620050428e-05}
{"train_loss": 0.4806126654148102, "global_step": 28676, "epoch": 322, "lr": 9.73552501523886e-05}
{"train_loss": 0.3031924366950989, "global_step": 28677, "epoch": 322, "lr": 9.735506409790658e-05}
{"train_loss": 0.3975818157196045, "global_step": 28678, "epoch": 322, "lr": 9.735487803705825e-05}
{"train_loss": 0.29039841890335083, "global_step": 28679, "epoch": 322, "lr": 9.735469196984364e-05}
{"train_loss": 0.4284856617450714, "global_step": 28680, "epoch": 322, "lr": 9.735450589626276e-05}
{"train_loss": 0.4735880196094513, "global_step": 28681, "epoch": 322, "lr": 9.735431981631565e-05}
{"train_loss": 0.43030163645744324, "global_step": 28682, "epoch": 322, "lr": 9.735413373000232e-05}
{"train_loss": 0.31908130645751953, "global_step": 28683, "epoch": 322, "lr": 9.735394763732282e-05}
{"train_loss": 0.30432388186454773, "global_step": 28684, "epoch": 322, "lr": 9.735376153827713e-05}
{"train_loss": 0.33871862292289734, "global_step": 28685, "epoch": 322, "lr": 9.735357543286534e-05}
{"train_loss": 0.38056808710098267, "global_step": 28686, "epoch": 322, "lr": 9.735338932108741e-05}
{"train_loss": 0.3840358257293701, "global_step": 28687, "epoch": 322, "lr": 9.73532032029434e-05}
{"train_loss": 0.43422073125839233, "global_step": 28688, "epoch": 322, "lr": 9.735301707843334e-05}
{"train_loss": 0.40456244349479675, "global_step": 28689, "epoch": 322, "lr": 9.735283094755725e-05}
{"train_loss": 0.38132309913635254, "global_step": 28690, "epoch": 322, "lr": 9.735264481031513e-05}
{"train_loss": 0.37255191802978516, "global_step": 28691, "epoch": 322, "lr": 9.735245866670703e-05}
{"train_loss": 0.42722538113594055, "global_step": 28692, "epoch": 322, "lr": 9.735227251673297e-05}
{"train_loss": 0.4797327220439911, "global_step": 28693, "epoch": 322, "lr": 9.735208636039298e-05}
{"train_loss": 0.4596533179283142, "global_step": 28694, "epoch": 322, "lr": 9.735190019768707e-05}
{"train_loss": 0.28597012162208557, "global_step": 28695, "epoch": 322, "lr": 9.735171402861528e-05}
{"train_loss": 0.4445793032646179, "global_step": 28696, "epoch": 322, "lr": 9.735152785317763e-05}
{"train_loss": 0.3841249942779541, "global_step": 28697, "epoch": 322, "lr": 9.735134167137415e-05}
{"train_loss": 0.5493810176849365, "global_step": 28698, "epoch": 322, "lr": 9.735115548320484e-05}
{"train_loss": 0.37973397970199585, "global_step": 28699, "epoch": 322, "lr": 9.735096928866976e-05}
{"train_loss": 0.4500501751899719, "global_step": 28700, "epoch": 322, "lr": 9.735078308776891e-05}
{"train_loss": 0.41369307041168213, "global_step": 28701, "epoch": 322, "lr": 9.735059688050233e-05}
{"train_loss": 0.37398380041122437, "global_step": 28702, "epoch": 322, "lr": 9.735041066687004e-05}
{"train_loss": 0.25778311491012573, "global_step": 28703, "epoch": 322, "lr": 9.735022444687207e-05}
{"train_loss": 0.4077649712562561, "global_step": 28704, "epoch": 322, "lr": 9.735003822050843e-05}
{"train_loss": 0.40673863887786865, "global_step": 28705, "epoch": 322, "lr": 9.734985198777915e-05}
{"train_loss": 0.43941551446914673, "global_step": 28706, "epoch": 322, "lr": 9.734966574868427e-05}
{"train_loss": 0.3738700747489929, "global_step": 28707, "epoch": 322, "lr": 9.73494795032238e-05}
{"train_loss": 0.4228675961494446, "global_step": 28708, "epoch": 322, "lr": 9.734929325139777e-05}
{"train_loss": 0.36061638593673706, "global_step": 28709, "epoch": 322, "lr": 9.734910699320621e-05}
{"train_loss": 0.34162309765815735, "global_step": 28710, "epoch": 322, "lr": 9.734892072864913e-05}
{"train_loss": 0.42068761587142944, "global_step": 28711, "epoch": 322, "lr": 9.734873445772656e-05}
{"train_loss": 0.33761268854141235, "global_step": 28712, "epoch": 322, "lr": 9.734854818043855e-05}
{"train_loss": 0.4551222324371338, "global_step": 28713, "epoch": 322, "lr": 9.734836189678509e-05}
{"train_loss": 0.46476221084594727, "global_step": 28714, "epoch": 322, "lr": 9.734817560676622e-05}
{"train_loss": 0.3584952652454376, "global_step": 28715, "epoch": 322, "lr": 9.734798931038198e-05}
{"train_loss": 0.3902842700481415, "global_step": 28716, "epoch": 322, "lr": 9.734780300763237e-05}
{"train_loss": 0.46357765793800354, "global_step": 28717, "epoch": 322, "lr": 9.734761669851742e-05}
{"train_loss": 0.40778374671936035, "global_step": 28718, "epoch": 322, "lr": 9.734743038303715e-05}
{"train_loss": 0.4846130907535553, "global_step": 28719, "epoch": 322, "lr": 9.734724406119163e-05}
{"train_loss": 0.46746572852134705, "global_step": 28720, "epoch": 322, "lr": 9.734705773298082e-05}
{"train_loss": 0.3032533526420593, "global_step": 28721, "epoch": 322, "lr": 9.734687139840478e-05}
{"train_loss": 0.4468141198158264, "global_step": 28722, "epoch": 322, "lr": 9.734668505746354e-05}
{"train_loss": 0.3968154191970825, "global_step": 28723, "epoch": 322, "lr": 9.734649871015711e-05}
{"train_loss": 0.3988175690174103, "global_step": 28724, "epoch": 322, "lr": 9.734631235648551e-05}
{"train_loss": 0.457364946603775, "global_step": 28725, "epoch": 322, "lr": 9.734612599644879e-05}
{"train_loss": 0.38123562932014465, "global_step": 28726, "epoch": 322, "lr": 9.734593963004696e-05}
{"train_loss": 0.46447935700416565, "global_step": 28727, "epoch": 322, "lr": 9.734575325728003e-05}
{"train_loss": 0.49059319496154785, "global_step": 28728, "epoch": 322, "lr": 9.734556687814807e-05}
{"train_loss": 0.5497991442680359, "global_step": 28729, "epoch": 322, "lr": 9.734538049265106e-05}
{"train_loss": 0.3782002329826355, "global_step": 28730, "epoch": 322, "lr": 9.734519410078901e-05}
{"train_loss": 0.3113749623298645, "global_step": 28731, "epoch": 322, "lr": 9.7345007702562e-05}
{"train_loss": 0.4282166361808777, "global_step": 28732, "epoch": 322, "lr": 9.734482129797004e-05}
{"train_loss": 0.4629342257976532, "global_step": 28733, "epoch": 322, "lr": 9.734463488701316e-05}
{"train_loss": 0.347670316696167, "global_step": 28734, "epoch": 322, "lr": 9.734444846969134e-05}
{"train_loss": 0.43688833713531494, "global_step": 28735, "epoch": 322, "lr": 9.734426204600464e-05}
{"train_loss": 0.3539562225341797, "global_step": 28736, "epoch": 322, "lr": 9.73440756159531e-05}
{"train_loss": 0.38313621282577515, "global_step": 28737, "epoch": 322, "lr": 9.73438891795367e-05}
{"train_loss": 0.33034801483154297, "global_step": 28738, "epoch": 322, "lr": 9.734370273675551e-05}
{"train_loss": 0.4179522395133972, "global_step": 28739, "epoch": 322, "lr": 9.734351628760954e-05}
{"train_loss": 0.38524776697158813, "global_step": 28740, "epoch": 322, "lr": 9.73433298320988e-05}
{"train_loss": 0.36603352427482605, "global_step": 28741, "epoch": 322, "lr": 9.734314337022332e-05}
{"train_loss": 0.3736800253391266, "global_step": 28742, "epoch": 322, "lr": 9.734295690198314e-05}
{"train_loss": 0.4336186945438385, "global_step": 28743, "epoch": 322, "lr": 9.734277042737828e-05}
{"train_loss": 0.3767749071121216, "global_step": 28744, "epoch": 322, "lr": 9.734258394640876e-05}
{"train_loss": 0.41325315833091736, "global_step": 28745, "epoch": 322, "lr": 9.734239745907461e-05}
{"train_loss": 0.3967713382136956, "global_step": 28746, "epoch": 322, "lr": 9.734221096537584e-05, "val_loss": 2.094838857650757}
{"train_loss": 0.44863593578338623, "global_step": 28747, "epoch": 323, "lr": 9.73420244653125e-05}
{"train_loss": 0.3213719129562378, "global_step": 28748, "epoch": 323, "lr": 9.73418379588846e-05}
{"train_loss": 0.42679089307785034, "global_step": 28749, "epoch": 323, "lr": 9.734165144609215e-05}
{"train_loss": 0.2816573679447174, "global_step": 28750, "epoch": 323, "lr": 9.734146492693521e-05}
{"train_loss": 0.3717254400253296, "global_step": 28751, "epoch": 323, "lr": 9.734127840141378e-05}
{"train_loss": 0.3619353771209717, "global_step": 28752, "epoch": 323, "lr": 9.734109186952788e-05}
{"train_loss": 0.4332059323787689, "global_step": 28753, "epoch": 323, "lr": 9.734090533127758e-05}
{"train_loss": 0.44237226247787476, "global_step": 28754, "epoch": 323, "lr": 9.734071878666284e-05}
{"train_loss": 0.4412047564983368, "global_step": 28755, "epoch": 323, "lr": 9.734053223568373e-05}
{"train_loss": 0.37943020462989807, "global_step": 28756, "epoch": 323, "lr": 9.734034567834027e-05}
{"train_loss": 0.3107552230358124, "global_step": 28757, "epoch": 323, "lr": 9.734015911463247e-05}
{"train_loss": 0.3065233528614044, "global_step": 28758, "epoch": 323, "lr": 9.733997254456037e-05}
{"train_loss": 0.3821873366832733, "global_step": 28759, "epoch": 323, "lr": 9.733978596812397e-05}
{"train_loss": 0.5237017273902893, "global_step": 28760, "epoch": 323, "lr": 9.733959938532333e-05}
{"train_loss": 0.3470778167247772, "global_step": 28761, "epoch": 323, "lr": 9.733941279615845e-05}
{"train_loss": 0.27488091588020325, "global_step": 28762, "epoch": 323, "lr": 9.733922620062936e-05}
{"train_loss": 0.34555891156196594, "global_step": 28763, "epoch": 323, "lr": 9.73390395987361e-05}
{"train_loss": 0.34808340668678284, "global_step": 28764, "epoch": 323, "lr": 9.733885299047867e-05}
{"train_loss": 0.42833754420280457, "global_step": 28765, "epoch": 323, "lr": 9.733866637585712e-05}
{"train_loss": 0.3466297686100006, "global_step": 28766, "epoch": 323, "lr": 9.733847975487145e-05}
{"train_loss": 0.37110796570777893, "global_step": 28767, "epoch": 323, "lr": 9.733829312752171e-05}
{"train_loss": 0.44169360399246216, "global_step": 28768, "epoch": 323, "lr": 9.73381064938079e-05}
{"train_loss": 0.3171294629573822, "global_step": 28769, "epoch": 323, "lr": 9.733791985373006e-05}
{"train_loss": 0.3249291479587555, "global_step": 28770, "epoch": 323, "lr": 9.733773320728823e-05}
{"train_loss": 0.4224674105644226, "global_step": 28771, "epoch": 323, "lr": 9.733754655448241e-05}
{"train_loss": 0.4098609387874603, "global_step": 28772, "epoch": 323, "lr": 9.733735989531263e-05}
{"train_loss": 0.34116917848587036, "global_step": 28773, "epoch": 323, "lr": 9.733717322977892e-05}
{"train_loss": 0.4128119647502899, "global_step": 28774, "epoch": 323, "lr": 9.73369865578813e-05}
{"train_loss": 0.379251092672348, "global_step": 28775, "epoch": 323, "lr": 9.73367998796198e-05}
{"train_loss": 0.46910396218299866, "global_step": 28776, "epoch": 323, "lr": 9.733661319499446e-05}
{"train_loss": 0.3735426664352417, "global_step": 28777, "epoch": 323, "lr": 9.733642650400528e-05}
{"train_loss": 0.375946968793869, "global_step": 28778, "epoch": 323, "lr": 9.73362398066523e-05}
{"train_loss": 0.35050415992736816, "global_step": 28779, "epoch": 323, "lr": 9.733605310293553e-05}
{"train_loss": 0.4162884056568146, "global_step": 28780, "epoch": 323, "lr": 9.733586639285502e-05}
{"train_loss": 0.5012899041175842, "global_step": 28781, "epoch": 323, "lr": 9.733567967641075e-05}
{"train_loss": 0.5323513746261597, "global_step": 28782, "epoch": 323, "lr": 9.73354929536028e-05}
{"train_loss": 0.5077478885650635, "global_step": 28783, "epoch": 323, "lr": 9.733530622443115e-05}
{"train_loss": 0.32498860359191895, "global_step": 28784, "epoch": 323, "lr": 9.733511948889587e-05}
{"train_loss": 0.49135419726371765, "global_step": 28785, "epoch": 323, "lr": 9.733493274699695e-05}
{"train_loss": 0.3695695698261261, "global_step": 28786, "epoch": 323, "lr": 9.733474599873441e-05}
{"train_loss": 0.3291541337966919, "global_step": 28787, "epoch": 323, "lr": 9.73345592441083e-05}
{"train_loss": 0.503435492515564, "global_step": 28788, "epoch": 323, "lr": 9.733437248311864e-05}
{"train_loss": 0.4657558798789978, "global_step": 28789, "epoch": 323, "lr": 9.733418571576545e-05}
{"train_loss": 0.27511176466941833, "global_step": 28790, "epoch": 323, "lr": 9.733399894204875e-05}
{"train_loss": 0.3545563220977783, "global_step": 28791, "epoch": 323, "lr": 9.733381216196858e-05}
{"train_loss": 0.4663936197757721, "global_step": 28792, "epoch": 323, "lr": 9.733362537552495e-05}
{"train_loss": 0.37954267859458923, "global_step": 28793, "epoch": 323, "lr": 9.73334385827179e-05}
{"train_loss": 0.48539236187934875, "global_step": 28794, "epoch": 323, "lr": 9.733325178354743e-05}
{"train_loss": 0.4350707530975342, "global_step": 28795, "epoch": 323, "lr": 9.733306497801358e-05}
{"train_loss": 0.30009639263153076, "global_step": 28796, "epoch": 323, "lr": 9.73328781661164e-05}
{"train_loss": 0.32511773705482483, "global_step": 28797, "epoch": 323, "lr": 9.733269134785587e-05}
{"train_loss": 0.4221995174884796, "global_step": 28798, "epoch": 323, "lr": 9.733250452323204e-05}
{"train_loss": 0.3515443205833435, "global_step": 28799, "epoch": 323, "lr": 9.733231769224493e-05}
{"train_loss": 0.3854225277900696, "global_step": 28800, "epoch": 323, "lr": 9.733213085489458e-05}
{"train_loss": 0.47431161999702454, "global_step": 28801, "epoch": 323, "lr": 9.7331944011181e-05}
{"train_loss": 0.3793877363204956, "global_step": 28802, "epoch": 323, "lr": 9.733175716110421e-05}
{"train_loss": 0.36097970604896545, "global_step": 28803, "epoch": 323, "lr": 9.733157030466425e-05}
{"train_loss": 0.3632427155971527, "global_step": 28804, "epoch": 323, "lr": 9.733138344186113e-05}
{"train_loss": 0.39883312582969666, "global_step": 28805, "epoch": 323, "lr": 9.733119657269487e-05}
{"train_loss": 0.29573848843574524, "global_step": 28806, "epoch": 323, "lr": 9.733100969716552e-05}
{"train_loss": 0.5112403035163879, "global_step": 28807, "epoch": 323, "lr": 9.733082281527309e-05}
{"train_loss": 0.388232946395874, "global_step": 28808, "epoch": 323, "lr": 9.73306359270176e-05}
{"train_loss": 0.35325369238853455, "global_step": 28809, "epoch": 323, "lr": 9.73304490323991e-05}
{"train_loss": 0.4051227867603302, "global_step": 28810, "epoch": 323, "lr": 9.733026213141761e-05}
{"train_loss": 0.4244906008243561, "global_step": 28811, "epoch": 323, "lr": 9.733007522407311e-05}
{"train_loss": 0.41085296869277954, "global_step": 28812, "epoch": 323, "lr": 9.732988831036567e-05}
{"train_loss": 0.39857277274131775, "global_step": 28813, "epoch": 323, "lr": 9.732970139029532e-05}
{"train_loss": 0.4626287817955017, "global_step": 28814, "epoch": 323, "lr": 9.732951446386204e-05}
{"train_loss": 0.4849618971347809, "global_step": 28815, "epoch": 323, "lr": 9.732932753106591e-05}
{"train_loss": 0.5252188444137573, "global_step": 28816, "epoch": 323, "lr": 9.732914059190692e-05}
{"train_loss": 0.37625738978385925, "global_step": 28817, "epoch": 323, "lr": 9.732895364638509e-05}
{"train_loss": 0.3760600984096527, "global_step": 28818, "epoch": 323, "lr": 9.732876669450048e-05}
{"train_loss": 0.34811702370643616, "global_step": 28819, "epoch": 323, "lr": 9.732857973625308e-05}
{"train_loss": 0.4025566577911377, "global_step": 28820, "epoch": 323, "lr": 9.732839277164293e-05}
{"train_loss": 0.4078100621700287, "global_step": 28821, "epoch": 323, "lr": 9.732820580067006e-05}
{"train_loss": 0.5017441511154175, "global_step": 28822, "epoch": 323, "lr": 9.732801882333449e-05}
{"train_loss": 0.3537050485610962, "global_step": 28823, "epoch": 323, "lr": 9.732783183963623e-05}
{"train_loss": 0.4978412687778473, "global_step": 28824, "epoch": 323, "lr": 9.732764484957533e-05}
{"train_loss": 0.5469697117805481, "global_step": 28825, "epoch": 323, "lr": 9.732745785315182e-05}
{"train_loss": 0.41218942403793335, "global_step": 28826, "epoch": 323, "lr": 9.73272708503657e-05}
{"train_loss": 0.4689561724662781, "global_step": 28827, "epoch": 323, "lr": 9.732708384121701e-05}
{"train_loss": 0.5010240077972412, "global_step": 28828, "epoch": 323, "lr": 9.732689682570575e-05}
{"train_loss": 0.40008798241615295, "global_step": 28829, "epoch": 323, "lr": 9.732670980383199e-05}
{"train_loss": 0.45986688137054443, "global_step": 28830, "epoch": 323, "lr": 9.73265227755957e-05}
{"train_loss": 0.38418373465538025, "global_step": 28831, "epoch": 323, "lr": 9.732633574099697e-05}
{"train_loss": 0.4238429069519043, "global_step": 28832, "epoch": 323, "lr": 9.732614870003576e-05}
{"train_loss": 0.4495570659637451, "global_step": 28833, "epoch": 323, "lr": 9.732596165271215e-05}
{"train_loss": 0.3492554724216461, "global_step": 28834, "epoch": 323, "lr": 9.732577459902614e-05}
{"train_loss": 0.4017505284105794, "global_step": 28835, "epoch": 323, "lr": 9.732558753897774e-05, "val_loss": 2.1706924438476562}
{"train_loss": 0.4803052246570587, "global_step": 28836, "epoch": 324, "lr": 9.7325400472567e-05}
{"train_loss": 0.3489378094673157, "global_step": 28837, "epoch": 324, "lr": 9.732521339979395e-05}
{"train_loss": 0.3984261453151703, "global_step": 28838, "epoch": 324, "lr": 9.732502632065859e-05}
{"train_loss": 0.33028608560562134, "global_step": 28839, "epoch": 324, "lr": 9.732483923516094e-05}
{"train_loss": 0.3488104045391083, "global_step": 28840, "epoch": 324, "lr": 9.732465214330106e-05}
{"train_loss": 0.5360520482063293, "global_step": 28841, "epoch": 324, "lr": 9.732446504507896e-05}
{"train_loss": 0.5010996460914612, "global_step": 28842, "epoch": 324, "lr": 9.732427794049465e-05}
{"train_loss": 0.37227484583854675, "global_step": 28843, "epoch": 324, "lr": 9.732409082954819e-05}
{"train_loss": 0.39901235699653625, "global_step": 28844, "epoch": 324, "lr": 9.732390371223957e-05}
{"train_loss": 0.3299073278903961, "global_step": 28845, "epoch": 324, "lr": 9.732371658856883e-05}
{"train_loss": 0.3326787054538727, "global_step": 28846, "epoch": 324, "lr": 9.732352945853598e-05}
{"train_loss": 0.5990046262741089, "global_step": 28847, "epoch": 324, "lr": 9.732334232214108e-05}
{"train_loss": 0.4327359199523926, "global_step": 28848, "epoch": 324, "lr": 9.732315517938412e-05}
{"train_loss": 0.5031659007072449, "global_step": 28849, "epoch": 324, "lr": 9.732296803026513e-05}
{"train_loss": 0.42099297046661377, "global_step": 28850, "epoch": 324, "lr": 9.732278087478416e-05}
{"train_loss": 0.4775064289569855, "global_step": 28851, "epoch": 324, "lr": 9.732259371294121e-05}
{"train_loss": 0.33287227153778076, "global_step": 28852, "epoch": 324, "lr": 9.732240654473633e-05}
{"train_loss": 0.4264269471168518, "global_step": 28853, "epoch": 324, "lr": 9.73222193701695e-05}
{"train_loss": 0.3761252164840698, "global_step": 28854, "epoch": 324, "lr": 9.73220321892408e-05}
{"train_loss": 0.5261967778205872, "global_step": 28855, "epoch": 324, "lr": 9.732184500195022e-05}
{"train_loss": 0.3492392599582672, "global_step": 28856, "epoch": 324, "lr": 9.73216578082978e-05}
{"train_loss": 0.4681769907474518, "global_step": 28857, "epoch": 324, "lr": 9.732147060828354e-05}
{"train_loss": 0.34363803267478943, "global_step": 28858, "epoch": 324, "lr": 9.732128340190752e-05}
{"train_loss": 0.3995094299316406, "global_step": 28859, "epoch": 324, "lr": 9.732109618916969e-05}
{"train_loss": 0.47246307134628296, "global_step": 28860, "epoch": 324, "lr": 9.732090897007015e-05}
{"train_loss": 0.3874073624610901, "global_step": 28861, "epoch": 324, "lr": 9.732072174460887e-05}
{"train_loss": 0.39732617139816284, "global_step": 28862, "epoch": 324, "lr": 9.732053451278592e-05}
{"train_loss": 0.5217118263244629, "global_step": 28863, "epoch": 324, "lr": 9.732034727460127e-05}
{"train_loss": 0.5128595232963562, "global_step": 28864, "epoch": 324, "lr": 9.732016003005499e-05}
{"train_loss": 0.45185959339141846, "global_step": 28865, "epoch": 324, "lr": 9.731997277914708e-05}
{"train_loss": 0.3071696162223816, "global_step": 28866, "epoch": 324, "lr": 9.731978552187759e-05}
{"train_loss": 0.3314286172389984, "global_step": 28867, "epoch": 324, "lr": 9.731959825824653e-05}
{"train_loss": 0.39507752656936646, "global_step": 28868, "epoch": 324, "lr": 9.731941098825392e-05}
{"train_loss": 0.4254814088344574, "global_step": 28869, "epoch": 324, "lr": 9.73192237118998e-05}
{"train_loss": 0.3546234965324402, "global_step": 28870, "epoch": 324, "lr": 9.731903642918418e-05}
{"train_loss": 0.3218713402748108, "global_step": 28871, "epoch": 324, "lr": 9.73188491401071e-05}
{"train_loss": 0.4196500778198242, "global_step": 28872, "epoch": 324, "lr": 9.731866184466858e-05}
{"train_loss": 0.4004567265510559, "global_step": 28873, "epoch": 324, "lr": 9.731847454286863e-05}
{"train_loss": 0.41480880975723267, "global_step": 28874, "epoch": 324, "lr": 9.73182872347073e-05}
{"train_loss": 0.38100090622901917, "global_step": 28875, "epoch": 324, "lr": 9.731809992018459e-05}
{"train_loss": 0.4285648763179779, "global_step": 28876, "epoch": 324, "lr": 9.731791259930055e-05}
{"train_loss": 0.37054818868637085, "global_step": 28877, "epoch": 324, "lr": 9.731772527205519e-05}
{"train_loss": 0.4874707758426666, "global_step": 28878, "epoch": 324, "lr": 9.731753793844854e-05}
{"train_loss": 0.3659328818321228, "global_step": 28879, "epoch": 324, "lr": 9.731735059848062e-05}
{"train_loss": 0.48685863614082336, "global_step": 28880, "epoch": 324, "lr": 9.731716325215145e-05}
{"train_loss": 0.33223408460617065, "global_step": 28881, "epoch": 324, "lr": 9.731697589946107e-05}
{"train_loss": 0.4280600845813751, "global_step": 28882, "epoch": 324, "lr": 9.731678854040951e-05}
{"train_loss": 0.34775617718696594, "global_step": 28883, "epoch": 324, "lr": 9.731660117499678e-05}
{"train_loss": 0.40180331468582153, "global_step": 28884, "epoch": 324, "lr": 9.73164138032229e-05}
{"train_loss": 0.4437408745288849, "global_step": 28885, "epoch": 324, "lr": 9.731622642508792e-05}
{"train_loss": 0.4837031662464142, "global_step": 28886, "epoch": 324, "lr": 9.731603904059185e-05}
{"train_loss": 0.4768129885196686, "global_step": 28887, "epoch": 324, "lr": 9.73158516497347e-05}
{"train_loss": 0.2986086905002594, "global_step": 28888, "epoch": 324, "lr": 9.731566425251653e-05}
{"train_loss": 0.27799278497695923, "global_step": 28889, "epoch": 324, "lr": 9.731547684893732e-05}
{"train_loss": 0.28760960698127747, "global_step": 28890, "epoch": 324, "lr": 9.731528943899714e-05}
{"train_loss": 0.3606909215450287, "global_step": 28891, "epoch": 324, "lr": 9.731510202269598e-05}
{"train_loss": 0.43348634243011475, "global_step": 28892, "epoch": 324, "lr": 9.73149146000339e-05}
{"train_loss": 0.4150165021419525, "global_step": 28893, "epoch": 324, "lr": 9.73147271710109e-05}
{"train_loss": 0.38065093755722046, "global_step": 28894, "epoch": 324, "lr": 9.731453973562701e-05}
{"train_loss": 0.3160156011581421, "global_step": 28895, "epoch": 324, "lr": 9.731435229388226e-05}
{"train_loss": 0.42259514331817627, "global_step": 28896, "epoch": 324, "lr": 9.731416484577666e-05}
{"train_loss": 0.382709801197052, "global_step": 28897, "epoch": 324, "lr": 9.731397739131026e-05}
{"train_loss": 0.39352521300315857, "global_step": 28898, "epoch": 324, "lr": 9.731378993048307e-05}
{"train_loss": 0.34976792335510254, "global_step": 28899, "epoch": 324, "lr": 9.731360246329511e-05}
{"train_loss": 0.2971930503845215, "global_step": 28900, "epoch": 324, "lr": 9.731341498974642e-05}
{"train_loss": 0.43607887625694275, "global_step": 28901, "epoch": 324, "lr": 9.731322750983703e-05}
{"train_loss": 0.4032205045223236, "global_step": 28902, "epoch": 324, "lr": 9.731304002356693e-05}
{"train_loss": 0.41811639070510864, "global_step": 28903, "epoch": 324, "lr": 9.731285253093618e-05}
{"train_loss": 0.3829754590988159, "global_step": 28904, "epoch": 324, "lr": 9.731266503194479e-05}
{"train_loss": 0.4103529155254364, "global_step": 28905, "epoch": 324, "lr": 9.731247752659279e-05}
{"train_loss": 0.43782538175582886, "global_step": 28906, "epoch": 324, "lr": 9.73122900148802e-05}
{"train_loss": 0.5317474603652954, "global_step": 28907, "epoch": 324, "lr": 9.731210249680707e-05}
{"train_loss": 0.34976035356521606, "global_step": 28908, "epoch": 324, "lr": 9.731191497237337e-05}
{"train_loss": 0.4631213843822479, "global_step": 28909, "epoch": 324, "lr": 9.731172744157919e-05}
{"train_loss": 0.4480931758880615, "global_step": 28910, "epoch": 324, "lr": 9.73115399044245e-05}
{"train_loss": 0.4697560966014862, "global_step": 28911, "epoch": 324, "lr": 9.731135236090938e-05}
{"train_loss": 0.3807946443557739, "global_step": 28912, "epoch": 324, "lr": 9.73111648110338e-05}
{"train_loss": 0.3309619128704071, "global_step": 28913, "epoch": 324, "lr": 9.731097725479782e-05}
{"train_loss": 0.2671095132827759, "global_step": 28914, "epoch": 324, "lr": 9.731078969220147e-05}
{"train_loss": 0.3535185754299164, "global_step": 28915, "epoch": 324, "lr": 9.731060212324473e-05}
{"train_loss": 0.33634933829307556, "global_step": 28916, "epoch": 324, "lr": 9.731041454792767e-05}
{"train_loss": 0.40732985734939575, "global_step": 28917, "epoch": 324, "lr": 9.731022696625032e-05}
{"train_loss": 0.30822286009788513, "global_step": 28918, "epoch": 324, "lr": 9.731003937821266e-05}
{"train_loss": 0.4168570339679718, "global_step": 28919, "epoch": 324, "lr": 9.730985178381476e-05}
{"train_loss": 0.36847415566444397, "global_step": 28920, "epoch": 324, "lr": 9.73096641830566e-05}
{"train_loss": 0.3454527258872986, "global_step": 28921, "epoch": 324, "lr": 9.730947657593826e-05}
{"train_loss": 0.482817679643631, "global_step": 28922, "epoch": 324, "lr": 9.730928896245973e-05}
{"train_loss": 0.41900503635406494, "global_step": 28923, "epoch": 324, "lr": 9.730910134262105e-05}
{"train_loss": 0.4024932715999946, "global_step": 28924, "epoch": 324, "lr": 9.730891371642222e-05, "val_loss": 2.206172466278076}
{"train_loss": 0.39182084798812866, "global_step": 28925, "epoch": 325, "lr": 9.73087260838633e-05}
{"train_loss": 0.3769788146018982, "global_step": 28926, "epoch": 325, "lr": 9.730853844494428e-05}
{"train_loss": 0.35179680585861206, "global_step": 28927, "epoch": 325, "lr": 9.730835079966523e-05}
{"train_loss": 0.4492245614528656, "global_step": 28928, "epoch": 325, "lr": 9.730816314802613e-05}
{"train_loss": 0.300763338804245, "global_step": 28929, "epoch": 325, "lr": 9.730797549002702e-05}
{"train_loss": 0.3746575713157654, "global_step": 28930, "epoch": 325, "lr": 9.730778782566794e-05}
{"train_loss": 0.3982369005680084, "global_step": 28931, "epoch": 325, "lr": 9.730760015494891e-05}
{"train_loss": 0.347472220659256, "global_step": 28932, "epoch": 325, "lr": 9.730741247786993e-05}
{"train_loss": 0.41607850790023804, "global_step": 28933, "epoch": 325, "lr": 9.730722479443106e-05}
{"train_loss": 0.4381929337978363, "global_step": 28934, "epoch": 325, "lr": 9.730703710463231e-05}
{"train_loss": 0.2988269031047821, "global_step": 28935, "epoch": 325, "lr": 9.73068494084737e-05}
{"train_loss": 0.44466838240623474, "global_step": 28936, "epoch": 325, "lr": 9.730666170595525e-05}
{"train_loss": 0.46150466799736023, "global_step": 28937, "epoch": 325, "lr": 9.730647399707701e-05}
{"train_loss": 0.3877376616001129, "global_step": 28938, "epoch": 325, "lr": 9.7306286281839e-05}
{"train_loss": 0.37846502661705017, "global_step": 28939, "epoch": 325, "lr": 9.730609856024122e-05}
{"train_loss": 0.2686433792114258, "global_step": 28940, "epoch": 325, "lr": 9.730591083228372e-05}
{"train_loss": 0.3722693622112274, "global_step": 28941, "epoch": 325, "lr": 9.730572309796652e-05}
{"train_loss": 0.3193191885948181, "global_step": 28942, "epoch": 325, "lr": 9.730553535728963e-05}
{"train_loss": 0.2914552390575409, "global_step": 28943, "epoch": 325, "lr": 9.73053476102531e-05}
{"train_loss": 0.3923661708831787, "global_step": 28944, "epoch": 325, "lr": 9.730515985685694e-05}
{"train_loss": 0.36934998631477356, "global_step": 28945, "epoch": 325, "lr": 9.730497209710117e-05}
{"train_loss": 0.3338264226913452, "global_step": 28946, "epoch": 325, "lr": 9.730478433098582e-05}
{"train_loss": 0.4311472475528717, "global_step": 28947, "epoch": 325, "lr": 9.730459655851093e-05}
{"train_loss": 0.37591928243637085, "global_step": 28948, "epoch": 325, "lr": 9.73044087796765e-05}
{"train_loss": 0.32536494731903076, "global_step": 28949, "epoch": 325, "lr": 9.730422099448259e-05}
{"train_loss": 0.3672671616077423, "global_step": 28950, "epoch": 325, "lr": 9.730403320292919e-05}
{"train_loss": 0.3554399609565735, "global_step": 28951, "epoch": 325, "lr": 9.730384540501634e-05}
{"train_loss": 0.3089738190174103, "global_step": 28952, "epoch": 325, "lr": 9.730365760074407e-05}
{"train_loss": 0.4190390706062317, "global_step": 28953, "epoch": 325, "lr": 9.73034697901124e-05}
{"train_loss": 0.23273339867591858, "global_step": 28954, "epoch": 325, "lr": 9.730328197312135e-05}
{"train_loss": 0.5289096832275391, "global_step": 28955, "epoch": 325, "lr": 9.730309414977095e-05}
{"train_loss": 0.2557881772518158, "global_step": 28956, "epoch": 325, "lr": 9.730290632006122e-05}
{"train_loss": 0.3424966335296631, "global_step": 28957, "epoch": 325, "lr": 9.730271848399219e-05}
{"train_loss": 0.28346461057662964, "global_step": 28958, "epoch": 325, "lr": 9.73025306415639e-05}
{"train_loss": 0.3054140508174896, "global_step": 28959, "epoch": 325, "lr": 9.730234279277635e-05}
{"train_loss": 0.4389011859893799, "global_step": 28960, "epoch": 325, "lr": 9.730215493762959e-05}
{"train_loss": 0.39067670702934265, "global_step": 28961, "epoch": 325, "lr": 9.730196707612362e-05}
{"train_loss": 0.41386622190475464, "global_step": 28962, "epoch": 325, "lr": 9.730177920825849e-05}
{"train_loss": 0.3895813226699829, "global_step": 28963, "epoch": 325, "lr": 9.73015913340342e-05}
{"train_loss": 0.28652289509773254, "global_step": 28964, "epoch": 325, "lr": 9.730140345345078e-05}
{"train_loss": 0.42287078499794006, "global_step": 28965, "epoch": 325, "lr": 9.730121556650828e-05}
{"train_loss": 0.28928816318511963, "global_step": 28966, "epoch": 325, "lr": 9.73010276732067e-05}
{"train_loss": 0.3665890097618103, "global_step": 28967, "epoch": 325, "lr": 9.730083977354606e-05}
{"train_loss": 0.3551335334777832, "global_step": 28968, "epoch": 325, "lr": 9.730065186752642e-05}
{"train_loss": 0.4253332018852234, "global_step": 28969, "epoch": 325, "lr": 9.730046395514779e-05}
{"train_loss": 0.5078532695770264, "global_step": 28970, "epoch": 325, "lr": 9.730027603641017e-05}
{"train_loss": 0.4174627959728241, "global_step": 28971, "epoch": 325, "lr": 9.730008811131361e-05}
{"train_loss": 0.4530811607837677, "global_step": 28972, "epoch": 325, "lr": 9.729990017985812e-05}
{"train_loss": 0.3342937231063843, "global_step": 28973, "epoch": 325, "lr": 9.729971224204375e-05}
{"train_loss": 0.26792800426483154, "global_step": 28974, "epoch": 325, "lr": 9.72995242978705e-05}
{"train_loss": 0.3578231632709503, "global_step": 28975, "epoch": 325, "lr": 9.729933634733843e-05}
{"train_loss": 0.42079025506973267, "global_step": 28976, "epoch": 325, "lr": 9.729914839044752e-05}
{"train_loss": 0.2410963922739029, "global_step": 28977, "epoch": 325, "lr": 9.729896042719782e-05}
{"train_loss": 0.309327632188797, "global_step": 28978, "epoch": 325, "lr": 9.729877245758934e-05}
{"train_loss": 0.3925684094429016, "global_step": 28979, "epoch": 325, "lr": 9.729858448162212e-05}
{"train_loss": 0.31196174025535583, "global_step": 28980, "epoch": 325, "lr": 9.72983964992962e-05}
{"train_loss": 0.3064706325531006, "global_step": 28981, "epoch": 325, "lr": 9.729820851061157e-05}
{"train_loss": 0.39034056663513184, "global_step": 28982, "epoch": 325, "lr": 9.729802051556826e-05}
{"train_loss": 0.4858616590499878, "global_step": 28983, "epoch": 325, "lr": 9.729783251416633e-05}
{"train_loss": 0.34020400047302246, "global_step": 28984, "epoch": 325, "lr": 9.729764450640578e-05}
{"train_loss": 0.4671911299228668, "global_step": 28985, "epoch": 325, "lr": 9.729745649228663e-05}
{"train_loss": 0.36443501710891724, "global_step": 28986, "epoch": 325, "lr": 9.729726847180893e-05}
{"train_loss": 0.360293984413147, "global_step": 28987, "epoch": 325, "lr": 9.729708044497268e-05}
{"train_loss": 0.39697638154029846, "global_step": 28988, "epoch": 325, "lr": 9.729689241177791e-05}
{"train_loss": 0.28022530674934387, "global_step": 28989, "epoch": 325, "lr": 9.729670437222465e-05}
{"train_loss": 0.4143710434436798, "global_step": 28990, "epoch": 325, "lr": 9.729651632631292e-05}
{"train_loss": 0.37617605924606323, "global_step": 28991, "epoch": 325, "lr": 9.729632827404275e-05}
{"train_loss": 0.48989829421043396, "global_step": 28992, "epoch": 325, "lr": 9.729614021541417e-05}
{"train_loss": 0.3439919054508209, "global_step": 28993, "epoch": 325, "lr": 9.729595215042721e-05}
{"train_loss": 0.34648919105529785, "global_step": 28994, "epoch": 325, "lr": 9.729576407908188e-05}
{"train_loss": 0.4472866952419281, "global_step": 28995, "epoch": 325, "lr": 9.72955760013782e-05}
{"train_loss": 0.44680365920066833, "global_step": 28996, "epoch": 325, "lr": 9.72953879173162e-05}
{"train_loss": 0.35722944140434265, "global_step": 28997, "epoch": 325, "lr": 9.729519982689594e-05}
{"train_loss": 0.4047711193561554, "global_step": 28998, "epoch": 325, "lr": 9.729501173011739e-05}
{"train_loss": 0.5516512393951416, "global_step": 28999, "epoch": 325, "lr": 9.729482362698062e-05}
{"train_loss": 0.4079754054546356, "global_step": 29000, "epoch": 325, "lr": 9.729463551748564e-05}
{"train_loss": 0.33856475353240967, "global_step": 29001, "epoch": 325, "lr": 9.729444740163245e-05}
{"train_loss": 0.4067482054233551, "global_step": 29002, "epoch": 325, "lr": 9.729425927942111e-05}
{"train_loss": 0.3212834298610687, "global_step": 29003, "epoch": 325, "lr": 9.729407115085166e-05}
{"train_loss": 0.45043936371803284, "global_step": 29004, "epoch": 325, "lr": 9.729388301592406e-05}
{"train_loss": 0.44453078508377075, "global_step": 29005, "epoch": 325, "lr": 9.729369487463838e-05}
{"train_loss": 0.3534473478794098, "global_step": 29006, "epoch": 325, "lr": 9.729350672699465e-05}
{"train_loss": 0.3905058801174164, "global_step": 29007, "epoch": 325, "lr": 9.729331857299289e-05}
{"train_loss": 0.39133116602897644, "global_step": 29008, "epoch": 325, "lr": 9.72931304126331e-05}
{"train_loss": 0.3707013428211212, "global_step": 29009, "epoch": 325, "lr": 9.729294224591534e-05}
{"train_loss": 0.5217628479003906, "global_step": 29010, "epoch": 325, "lr": 9.729275407283961e-05}
{"train_loss": 0.4646581709384918, "global_step": 29011, "epoch": 325, "lr": 9.729256589340596e-05}
{"train_loss": 0.4066837728023529, "global_step": 29012, "epoch": 325, "lr": 9.729237770761439e-05}
{"train_loss": 0.3792588769384984, "global_step": 29013, "epoch": 325, "lr": 9.729218951546494e-05, "val_loss": 2.1573054790496826, "train_action_mse_error": 51.605796813964844}
{"train_loss": 0.36937373876571655, "global_step": 29014, "epoch": 326, "lr": 9.729200131695764e-05}
{"train_loss": 0.3120899200439453, "global_step": 29015, "epoch": 326, "lr": 9.729181311209248e-05}
{"train_loss": 0.2910059690475464, "global_step": 29016, "epoch": 326, "lr": 9.729162490086954e-05}
{"train_loss": 0.3545375466346741, "global_step": 29017, "epoch": 326, "lr": 9.72914366832888e-05}
{"train_loss": 0.36016058921813965, "global_step": 29018, "epoch": 326, "lr": 9.729124845935032e-05}
{"train_loss": 0.4129050076007843, "global_step": 29019, "epoch": 326, "lr": 9.72910602290541e-05}
{"train_loss": 0.3521106541156769, "global_step": 29020, "epoch": 326, "lr": 9.729087199240017e-05}
{"train_loss": 0.39763110876083374, "global_step": 29021, "epoch": 326, "lr": 9.729068374938855e-05}
{"train_loss": 0.3917686343193054, "global_step": 29022, "epoch": 326, "lr": 9.72904955000193e-05}
{"train_loss": 0.3083620071411133, "global_step": 29023, "epoch": 326, "lr": 9.729030724429238e-05}
{"train_loss": 0.29504096508026123, "global_step": 29024, "epoch": 326, "lr": 9.729011898220789e-05}
{"train_loss": 0.3805108964443207, "global_step": 29025, "epoch": 326, "lr": 9.72899307137658e-05}
{"train_loss": 0.36696329712867737, "global_step": 29026, "epoch": 326, "lr": 9.728974243896616e-05}
{"train_loss": 0.3735649883747101, "global_step": 29027, "epoch": 326, "lr": 9.7289554157809e-05}
{"train_loss": 0.38343584537506104, "global_step": 29028, "epoch": 326, "lr": 9.728936587029433e-05}
{"train_loss": 0.3695524036884308, "global_step": 29029, "epoch": 326, "lr": 9.728917757642218e-05}
{"train_loss": 0.3809605538845062, "global_step": 29030, "epoch": 326, "lr": 9.728898927619257e-05}
{"train_loss": 0.30430763959884644, "global_step": 29031, "epoch": 326, "lr": 9.728880096960555e-05}
{"train_loss": 0.38847440481185913, "global_step": 29032, "epoch": 326, "lr": 9.728861265666112e-05}
{"train_loss": 0.4413042664527893, "global_step": 29033, "epoch": 326, "lr": 9.72884243373593e-05}
{"train_loss": 0.45853811502456665, "global_step": 29034, "epoch": 326, "lr": 9.728823601170014e-05}
{"train_loss": 0.5040062069892883, "global_step": 29035, "epoch": 326, "lr": 9.728804767968364e-05}
{"train_loss": 0.4055618941783905, "global_step": 29036, "epoch": 326, "lr": 9.728785934130985e-05}
{"train_loss": 0.36911043524742126, "global_step": 29037, "epoch": 326, "lr": 9.728767099657878e-05}
{"train_loss": 0.44527560472488403, "global_step": 29038, "epoch": 326, "lr": 9.728748264549046e-05}
{"train_loss": 0.4262196719646454, "global_step": 29039, "epoch": 326, "lr": 9.728729428804492e-05}
{"train_loss": 0.43618670105934143, "global_step": 29040, "epoch": 326, "lr": 9.728710592424217e-05}
{"train_loss": 0.5046802759170532, "global_step": 29041, "epoch": 326, "lr": 9.728691755408225e-05}
{"train_loss": 0.3952292203903198, "global_step": 29042, "epoch": 326, "lr": 9.728672917756519e-05}
{"train_loss": 0.3419196307659149, "global_step": 29043, "epoch": 326, "lr": 9.728654079469099e-05}
{"train_loss": 0.30901601910591125, "global_step": 29044, "epoch": 326, "lr": 9.728635240545969e-05}
{"train_loss": 0.289031445980072, "global_step": 29045, "epoch": 326, "lr": 9.728616400987132e-05}
{"train_loss": 0.48293209075927734, "global_step": 29046, "epoch": 326, "lr": 9.72859756079259e-05}
{"train_loss": 0.4619984030723572, "global_step": 29047, "epoch": 326, "lr": 9.728578719962347e-05}
{"train_loss": 0.39759477972984314, "global_step": 29048, "epoch": 326, "lr": 9.728559878496401e-05}
{"train_loss": 0.42464005947113037, "global_step": 29049, "epoch": 326, "lr": 9.728541036394762e-05}
{"train_loss": 0.2560686469078064, "global_step": 29050, "epoch": 326, "lr": 9.728522193657425e-05}
{"train_loss": 0.31244516372680664, "global_step": 29051, "epoch": 326, "lr": 9.728503350284399e-05}
{"train_loss": 0.38883382081985474, "global_step": 29052, "epoch": 326, "lr": 9.72848450627568e-05}
{"train_loss": 0.3660763204097748, "global_step": 29053, "epoch": 326, "lr": 9.728465661631277e-05}
{"train_loss": 0.5418585538864136, "global_step": 29054, "epoch": 326, "lr": 9.728446816351186e-05}
{"train_loss": 0.3536233901977539, "global_step": 29055, "epoch": 326, "lr": 9.728427970435416e-05}
{"train_loss": 0.41009843349456787, "global_step": 29056, "epoch": 326, "lr": 9.728409123883963e-05}
{"train_loss": 0.29972735047340393, "global_step": 29057, "epoch": 326, "lr": 9.728390276696837e-05}
{"train_loss": 0.49715977907180786, "global_step": 29058, "epoch": 326, "lr": 9.728371428874035e-05}
{"train_loss": 0.4138919711112976, "global_step": 29059, "epoch": 326, "lr": 9.728352580415561e-05}
{"train_loss": 0.3946027457714081, "global_step": 29060, "epoch": 326, "lr": 9.728333731321417e-05}
{"train_loss": 0.426719605922699, "global_step": 29061, "epoch": 326, "lr": 9.728314881591608e-05}
{"train_loss": 0.47250279784202576, "global_step": 29062, "epoch": 326, "lr": 9.728296031226132e-05}
{"train_loss": 0.5328356623649597, "global_step": 29063, "epoch": 326, "lr": 9.728277180224997e-05}
{"train_loss": 0.33252331614494324, "global_step": 29064, "epoch": 326, "lr": 9.7282583285882e-05}
{"train_loss": 0.4176367223262787, "global_step": 29065, "epoch": 326, "lr": 9.728239476315747e-05}
{"train_loss": 0.3843744397163391, "global_step": 29066, "epoch": 326, "lr": 9.728220623407641e-05}
{"train_loss": 0.38363656401634216, "global_step": 29067, "epoch": 326, "lr": 9.728201769863883e-05}
{"train_loss": 0.48228850960731506, "global_step": 29068, "epoch": 326, "lr": 9.728182915684477e-05}
{"train_loss": 0.4675676226615906, "global_step": 29069, "epoch": 326, "lr": 9.728164060869423e-05}
{"train_loss": 0.3507314920425415, "global_step": 29070, "epoch": 326, "lr": 9.728145205418723e-05}
{"train_loss": 0.4912206828594208, "global_step": 29071, "epoch": 326, "lr": 9.728126349332385e-05}
{"train_loss": 0.3231549561023712, "global_step": 29072, "epoch": 326, "lr": 9.728107492610406e-05}
{"train_loss": 0.33574920892715454, "global_step": 29073, "epoch": 326, "lr": 9.728088635252792e-05}
{"train_loss": 0.4269908666610718, "global_step": 29074, "epoch": 326, "lr": 9.728069777259543e-05}
{"train_loss": 0.44190356135368347, "global_step": 29075, "epoch": 326, "lr": 9.728050918630663e-05}
{"train_loss": 0.4179081916809082, "global_step": 29076, "epoch": 326, "lr": 9.728032059366154e-05}
{"train_loss": 0.4457242488861084, "global_step": 29077, "epoch": 326, "lr": 9.72801319946602e-05}
{"train_loss": 0.400806188583374, "global_step": 29078, "epoch": 326, "lr": 9.72799433893026e-05}
{"train_loss": 0.30476221442222595, "global_step": 29079, "epoch": 326, "lr": 9.727975477758882e-05}
{"train_loss": 0.4238564968109131, "global_step": 29080, "epoch": 326, "lr": 9.727956615951882e-05}
{"train_loss": 0.3672686517238617, "global_step": 29081, "epoch": 326, "lr": 9.727937753509268e-05}
{"train_loss": 0.4619951844215393, "global_step": 29082, "epoch": 326, "lr": 9.727918890431038e-05}
{"train_loss": 0.32005295157432556, "global_step": 29083, "epoch": 326, "lr": 9.727900026717199e-05}
{"train_loss": 0.3551926910877228, "global_step": 29084, "epoch": 326, "lr": 9.727881162367751e-05}
{"train_loss": 0.3416016399860382, "global_step": 29085, "epoch": 326, "lr": 9.727862297382697e-05}
{"train_loss": 0.302736759185791, "global_step": 29086, "epoch": 326, "lr": 9.72784343176204e-05}
{"train_loss": 0.5102904438972473, "global_step": 29087, "epoch": 326, "lr": 9.72782456550578e-05}
{"train_loss": 0.4259508550167084, "global_step": 29088, "epoch": 326, "lr": 9.727805698613925e-05}
{"train_loss": 0.3605726659297943, "global_step": 29089, "epoch": 326, "lr": 9.727786831086472e-05}
{"train_loss": 0.4072936475276947, "global_step": 29090, "epoch": 326, "lr": 9.727767962923426e-05}
{"train_loss": 0.3805459141731262, "global_step": 29091, "epoch": 326, "lr": 9.72774909412479e-05}
{"train_loss": 0.33250245451927185, "global_step": 29092, "epoch": 326, "lr": 9.727730224690564e-05}
{"train_loss": 0.329256534576416, "global_step": 29093, "epoch": 326, "lr": 9.727711354620755e-05}
{"train_loss": 0.41601473093032837, "global_step": 29094, "epoch": 326, "lr": 9.72769248391536e-05}
{"train_loss": 0.42553582787513733, "global_step": 29095, "epoch": 326, "lr": 9.727673612574385e-05}
{"train_loss": 0.39150258898735046, "global_step": 29096, "epoch": 326, "lr": 9.727654740597833e-05}
{"train_loss": 0.32937145233154297, "global_step": 29097, "epoch": 326, "lr": 9.727635867985705e-05}
{"train_loss": 0.35624030232429504, "global_step": 29098, "epoch": 326, "lr": 9.727616994738003e-05}
{"train_loss": 0.38256919384002686, "global_step": 29099, "epoch": 326, "lr": 9.727598120854732e-05}
{"train_loss": 0.29689720273017883, "global_step": 29100, "epoch": 326, "lr": 9.727579246335891e-05}
{"train_loss": 0.4938872456550598, "global_step": 29101, "epoch": 326, "lr": 9.727560371181488e-05}
{"train_loss": 0.39083850953016386, "global_step": 29102, "epoch": 326, "lr": 9.727541495391519e-05, "val_loss": 2.186410665512085}
{"train_loss": 0.33372873067855835, "global_step": 29103, "epoch": 327, "lr": 9.72752261896599e-05}
{"train_loss": 0.23996193706989288, "global_step": 29104, "epoch": 327, "lr": 9.727503741904904e-05}
{"train_loss": 0.34550154209136963, "global_step": 29105, "epoch": 327, "lr": 9.727484864208262e-05}
{"train_loss": 0.3583609461784363, "global_step": 29106, "epoch": 327, "lr": 9.72746598587607e-05}
{"train_loss": 0.402986615896225, "global_step": 29107, "epoch": 327, "lr": 9.727447106908325e-05}
{"train_loss": 0.261766254901886, "global_step": 29108, "epoch": 327, "lr": 9.727428227305033e-05}
{"train_loss": 0.43971729278564453, "global_step": 29109, "epoch": 327, "lr": 9.727409347066196e-05}
{"train_loss": 0.32134461402893066, "global_step": 29110, "epoch": 327, "lr": 9.727390466191817e-05}
{"train_loss": 0.3633127808570862, "global_step": 29111, "epoch": 327, "lr": 9.727371584681897e-05}
{"train_loss": 0.4663960635662079, "global_step": 29112, "epoch": 327, "lr": 9.72735270253644e-05}
{"train_loss": 0.3735176622867584, "global_step": 29113, "epoch": 327, "lr": 9.727333819755449e-05}
{"train_loss": 0.3161810338497162, "global_step": 29114, "epoch": 327, "lr": 9.727314936338925e-05}
{"train_loss": 0.3744911551475525, "global_step": 29115, "epoch": 327, "lr": 9.72729605228687e-05}
{"train_loss": 0.42452821135520935, "global_step": 29116, "epoch": 327, "lr": 9.72727716759929e-05}
{"train_loss": 0.41683611273765564, "global_step": 29117, "epoch": 327, "lr": 9.727258282276181e-05}
{"train_loss": 0.45996204018592834, "global_step": 29118, "epoch": 327, "lr": 9.727239396317553e-05}
{"train_loss": 0.4926284849643707, "global_step": 29119, "epoch": 327, "lr": 9.727220509723405e-05}
{"train_loss": 0.41919469833374023, "global_step": 29120, "epoch": 327, "lr": 9.72720162249374e-05}
{"train_loss": 0.34395337104797363, "global_step": 29121, "epoch": 327, "lr": 9.72718273462856e-05}
{"train_loss": 0.5237613320350647, "global_step": 29122, "epoch": 327, "lr": 9.727163846127868e-05}
{"train_loss": 0.3838474750518799, "global_step": 29123, "epoch": 327, "lr": 9.727144956991665e-05}
{"train_loss": 0.40840134024620056, "global_step": 29124, "epoch": 327, "lr": 9.727126067219956e-05}
{"train_loss": 0.3934566378593445, "global_step": 29125, "epoch": 327, "lr": 9.727107176812744e-05}
{"train_loss": 0.40565821528434753, "global_step": 29126, "epoch": 327, "lr": 9.727088285770027e-05}
{"train_loss": 0.36666765809059143, "global_step": 29127, "epoch": 327, "lr": 9.727069394091812e-05}
{"train_loss": 0.46470364928245544, "global_step": 29128, "epoch": 327, "lr": 9.727050501778101e-05}
{"train_loss": 0.35297223925590515, "global_step": 29129, "epoch": 327, "lr": 9.727031608828896e-05}
{"train_loss": 0.47303950786590576, "global_step": 29130, "epoch": 327, "lr": 9.727012715244197e-05}
{"train_loss": 0.42325589060783386, "global_step": 29131, "epoch": 327, "lr": 9.726993821024011e-05}
{"train_loss": 0.4449748992919922, "global_step": 29132, "epoch": 327, "lr": 9.726974926168337e-05}
{"train_loss": 0.40940988063812256, "global_step": 29133, "epoch": 327, "lr": 9.726956030677178e-05}
{"train_loss": 0.36374300718307495, "global_step": 29134, "epoch": 327, "lr": 9.72693713455054e-05}
{"train_loss": 0.40931642055511475, "global_step": 29135, "epoch": 327, "lr": 9.726918237788422e-05}
{"train_loss": 0.4343440532684326, "global_step": 29136, "epoch": 327, "lr": 9.726899340390826e-05}
{"train_loss": 0.46691611409187317, "global_step": 29137, "epoch": 327, "lr": 9.726880442357757e-05}
{"train_loss": 0.3843434751033783, "global_step": 29138, "epoch": 327, "lr": 9.726861543689217e-05}
{"train_loss": 0.3378353714942932, "global_step": 29139, "epoch": 327, "lr": 9.726842644385208e-05}
{"train_loss": 0.33048614859580994, "global_step": 29140, "epoch": 327, "lr": 9.726823744445732e-05}
{"train_loss": 0.4689188003540039, "global_step": 29141, "epoch": 327, "lr": 9.726804843870791e-05}
{"train_loss": 0.384368360042572, "global_step": 29142, "epoch": 327, "lr": 9.72678594266039e-05}
{"train_loss": 0.4684489369392395, "global_step": 29143, "epoch": 327, "lr": 9.72676704081453e-05}
{"train_loss": 0.3003860116004944, "global_step": 29144, "epoch": 327, "lr": 9.726748138333215e-05}
{"train_loss": 0.37482020258903503, "global_step": 29145, "epoch": 327, "lr": 9.726729235216445e-05}
{"train_loss": 0.39916306734085083, "global_step": 29146, "epoch": 327, "lr": 9.726710331464224e-05}
{"train_loss": 0.40485140681266785, "global_step": 29147, "epoch": 327, "lr": 9.726691427076555e-05}
{"train_loss": 0.29776012897491455, "global_step": 29148, "epoch": 327, "lr": 9.72667252205344e-05}
{"train_loss": 0.3913614749908447, "global_step": 29149, "epoch": 327, "lr": 9.726653616394881e-05}
{"train_loss": 0.3638453185558319, "global_step": 29150, "epoch": 327, "lr": 9.726634710100881e-05}
{"train_loss": 0.3665321469306946, "global_step": 29151, "epoch": 327, "lr": 9.726615803171442e-05}
{"train_loss": 0.3885635435581207, "global_step": 29152, "epoch": 327, "lr": 9.726596895606569e-05}
{"train_loss": 0.376930296421051, "global_step": 29153, "epoch": 327, "lr": 9.726577987406261e-05}
{"train_loss": 0.4099213182926178, "global_step": 29154, "epoch": 327, "lr": 9.726559078570523e-05}
{"train_loss": 0.39776238799095154, "global_step": 29155, "epoch": 327, "lr": 9.726540169099356e-05}
{"train_loss": 0.332235187292099, "global_step": 29156, "epoch": 327, "lr": 9.726521258992765e-05}
{"train_loss": 0.4653851389884949, "global_step": 29157, "epoch": 327, "lr": 9.726502348250748e-05}
{"train_loss": 0.36931461095809937, "global_step": 29158, "epoch": 327, "lr": 9.726483436873313e-05}
{"train_loss": 0.39108386635780334, "global_step": 29159, "epoch": 327, "lr": 9.726464524860458e-05}
{"train_loss": 0.41576650738716125, "global_step": 29160, "epoch": 327, "lr": 9.726445612212188e-05}
{"train_loss": 0.32875826954841614, "global_step": 29161, "epoch": 327, "lr": 9.726426698928506e-05}
{"train_loss": 0.3539445996284485, "global_step": 29162, "epoch": 327, "lr": 9.726407785009412e-05}
{"train_loss": 0.32042768597602844, "global_step": 29163, "epoch": 327, "lr": 9.726388870454911e-05}
{"train_loss": 0.41123759746551514, "global_step": 29164, "epoch": 327, "lr": 9.726369955265004e-05}
{"train_loss": 0.3388670086860657, "global_step": 29165, "epoch": 327, "lr": 9.726351039439696e-05}
{"train_loss": 0.32767683267593384, "global_step": 29166, "epoch": 327, "lr": 9.726332122978985e-05}
{"train_loss": 0.4250079393386841, "global_step": 29167, "epoch": 327, "lr": 9.726313205882878e-05}
{"train_loss": 0.37810173630714417, "global_step": 29168, "epoch": 327, "lr": 9.726294288151377e-05}
{"train_loss": 0.38597217202186584, "global_step": 29169, "epoch": 327, "lr": 9.726275369784481e-05}
{"train_loss": 0.4169904887676239, "global_step": 29170, "epoch": 327, "lr": 9.726256450782195e-05}
{"train_loss": 0.33391597867012024, "global_step": 29171, "epoch": 327, "lr": 9.726237531144522e-05}
{"train_loss": 0.2731072008609772, "global_step": 29172, "epoch": 327, "lr": 9.726218610871465e-05}
{"train_loss": 0.31953781843185425, "global_step": 29173, "epoch": 327, "lr": 9.726199689963023e-05}
{"train_loss": 0.3692781329154968, "global_step": 29174, "epoch": 327, "lr": 9.726180768419205e-05}
{"train_loss": 0.3684777319431305, "global_step": 29175, "epoch": 327, "lr": 9.726161846240006e-05}
{"train_loss": 0.38664746284484863, "global_step": 29176, "epoch": 327, "lr": 9.726142923425432e-05}
{"train_loss": 0.3693297207355499, "global_step": 29177, "epoch": 327, "lr": 9.726123999975487e-05}
{"train_loss": 0.3586062490940094, "global_step": 29178, "epoch": 327, "lr": 9.726105075890172e-05}
{"train_loss": 0.3620307445526123, "global_step": 29179, "epoch": 327, "lr": 9.726086151169491e-05}
{"train_loss": 0.3330771028995514, "global_step": 29180, "epoch": 327, "lr": 9.726067225813444e-05}
{"train_loss": 0.3655224144458771, "global_step": 29181, "epoch": 327, "lr": 9.726048299822034e-05}
{"train_loss": 0.4640413522720337, "global_step": 29182, "epoch": 327, "lr": 9.726029373195266e-05}
{"train_loss": 0.4687633514404297, "global_step": 29183, "epoch": 327, "lr": 9.726010445933143e-05}
{"train_loss": 0.39463889598846436, "global_step": 29184, "epoch": 327, "lr": 9.725991518035663e-05}
{"train_loss": 0.4268054664134979, "global_step": 29185, "epoch": 327, "lr": 9.72597258950283e-05}
{"train_loss": 0.40800976753234863, "global_step": 29186, "epoch": 327, "lr": 9.725953660334649e-05}
{"train_loss": 0.2996625304222107, "global_step": 29187, "epoch": 327, "lr": 9.72593473053112e-05}
{"train_loss": 0.41077202558517456, "global_step": 29188, "epoch": 327, "lr": 9.725915800092248e-05}
{"train_loss": 0.4203025996685028, "global_step": 29189, "epoch": 327, "lr": 9.725896869018034e-05}
{"train_loss": 0.44085797667503357, "global_step": 29190, "epoch": 327, "lr": 9.725877937308481e-05}
{"train_loss": 0.38598738476801453, "global_step": 29191, "epoch": 327, "lr": 9.72585900496359e-05, "val_loss": 2.220914125442505}
{"train_loss": 0.4389986991882324, "global_step": 29192, "epoch": 328, "lr": 9.725840071983365e-05}
{"train_loss": 0.327519029378891, "global_step": 29193, "epoch": 328, "lr": 9.72582113836781e-05}
{"train_loss": 0.30987975001335144, "global_step": 29194, "epoch": 328, "lr": 9.725802204116925e-05}
{"train_loss": 0.3396573066711426, "global_step": 29195, "epoch": 328, "lr": 9.725783269230712e-05}
{"train_loss": 0.4150521755218506, "global_step": 29196, "epoch": 328, "lr": 9.725764333709178e-05}
{"train_loss": 0.5134566426277161, "global_step": 29197, "epoch": 328, "lr": 9.725745397552321e-05}
{"train_loss": 0.39120304584503174, "global_step": 29198, "epoch": 328, "lr": 9.725726460760145e-05}
{"train_loss": 0.34315043687820435, "global_step": 29199, "epoch": 328, "lr": 9.725707523332653e-05}
{"train_loss": 0.3079984784126282, "global_step": 29200, "epoch": 328, "lr": 9.725688585269848e-05}
{"train_loss": 0.39279839396476746, "global_step": 29201, "epoch": 328, "lr": 9.72566964657173e-05}
{"train_loss": 0.3670404851436615, "global_step": 29202, "epoch": 328, "lr": 9.725650707238306e-05}
{"train_loss": 0.3495185077190399, "global_step": 29203, "epoch": 328, "lr": 9.725631767269572e-05}
{"train_loss": 0.3091423213481903, "global_step": 29204, "epoch": 328, "lr": 9.725612826665537e-05}
{"train_loss": 0.3952542841434479, "global_step": 29205, "epoch": 328, "lr": 9.725593885426201e-05}
{"train_loss": 0.45274782180786133, "global_step": 29206, "epoch": 328, "lr": 9.725574943551566e-05}
{"train_loss": 0.3070526719093323, "global_step": 29207, "epoch": 328, "lr": 9.725556001041634e-05}
{"train_loss": 0.3239985406398773, "global_step": 29208, "epoch": 328, "lr": 9.72553705789641e-05}
{"train_loss": 0.3860737085342407, "global_step": 29209, "epoch": 328, "lr": 9.725518114115895e-05}
{"train_loss": 0.32625800371170044, "global_step": 29210, "epoch": 328, "lr": 9.725499169700091e-05}
{"train_loss": 0.3923826813697815, "global_step": 29211, "epoch": 328, "lr": 9.725480224649001e-05}
{"train_loss": 0.32046040892601013, "global_step": 29212, "epoch": 328, "lr": 9.725461278962628e-05}
{"train_loss": 0.4261074960231781, "global_step": 29213, "epoch": 328, "lr": 9.725442332640976e-05}
{"train_loss": 0.2972508370876312, "global_step": 29214, "epoch": 328, "lr": 9.725423385684046e-05}
{"train_loss": 0.42202481627464294, "global_step": 29215, "epoch": 328, "lr": 9.725404438091837e-05}
{"train_loss": 0.39401671290397644, "global_step": 29216, "epoch": 328, "lr": 9.725385489864359e-05}
{"train_loss": 0.34826111793518066, "global_step": 29217, "epoch": 328, "lr": 9.725366541001607e-05}
{"train_loss": 0.35136041045188904, "global_step": 29218, "epoch": 328, "lr": 9.725347591503588e-05}
{"train_loss": 0.4307864010334015, "global_step": 29219, "epoch": 328, "lr": 9.725328641370304e-05}
{"train_loss": 0.3314834535121918, "global_step": 29220, "epoch": 328, "lr": 9.725309690601758e-05}
{"train_loss": 0.41652706265449524, "global_step": 29221, "epoch": 328, "lr": 9.725290739197952e-05}
{"train_loss": 0.40850895643234253, "global_step": 29222, "epoch": 328, "lr": 9.725271787158887e-05}
{"train_loss": 0.42342251539230347, "global_step": 29223, "epoch": 328, "lr": 9.725252834484567e-05}
{"train_loss": 0.3076137602329254, "global_step": 29224, "epoch": 328, "lr": 9.725233881174995e-05}
{"train_loss": 0.35517725348472595, "global_step": 29225, "epoch": 328, "lr": 9.725214927230172e-05}
{"train_loss": 0.4792073667049408, "global_step": 29226, "epoch": 328, "lr": 9.725195972650101e-05}
{"train_loss": 0.2915104627609253, "global_step": 29227, "epoch": 328, "lr": 9.725177017434786e-05}
{"train_loss": 0.3321850895881653, "global_step": 29228, "epoch": 328, "lr": 9.725158061584229e-05}
{"train_loss": 0.3503609895706177, "global_step": 29229, "epoch": 328, "lr": 9.725139105098431e-05}
{"train_loss": 0.4153677225112915, "global_step": 29230, "epoch": 328, "lr": 9.725120147977395e-05}
{"train_loss": 0.3716491162776947, "global_step": 29231, "epoch": 328, "lr": 9.725101190221127e-05}
{"train_loss": 0.40334880352020264, "global_step": 29232, "epoch": 328, "lr": 9.725082231829624e-05}
{"train_loss": 0.3739185035228729, "global_step": 29233, "epoch": 328, "lr": 9.725063272802893e-05}
{"train_loss": 0.37292423844337463, "global_step": 29234, "epoch": 328, "lr": 9.725044313140934e-05}
{"train_loss": 0.40842118859291077, "global_step": 29235, "epoch": 328, "lr": 9.725025352843751e-05}
{"train_loss": 0.42005056142807007, "global_step": 29236, "epoch": 328, "lr": 9.725006391911344e-05}
{"train_loss": 0.39048880338668823, "global_step": 29237, "epoch": 328, "lr": 9.724987430343719e-05}
{"train_loss": 0.3277917504310608, "global_step": 29238, "epoch": 328, "lr": 9.724968468140878e-05}
{"train_loss": 0.3288440406322479, "global_step": 29239, "epoch": 328, "lr": 9.724949505302821e-05}
{"train_loss": 0.38383719325065613, "global_step": 29240, "epoch": 328, "lr": 9.724930541829553e-05}
{"train_loss": 0.4458208680152893, "global_step": 29241, "epoch": 328, "lr": 9.724911577721074e-05}
{"train_loss": 0.4361644983291626, "global_step": 29242, "epoch": 328, "lr": 9.724892612977389e-05}
{"train_loss": 0.4684107005596161, "global_step": 29243, "epoch": 328, "lr": 9.724873647598503e-05}
{"train_loss": 0.4186461865901947, "global_step": 29244, "epoch": 328, "lr": 9.724854681584413e-05}
{"train_loss": 0.32968637347221375, "global_step": 29245, "epoch": 328, "lr": 9.724835714935122e-05}
{"train_loss": 0.36665016412734985, "global_step": 29246, "epoch": 328, "lr": 9.724816747650635e-05}
{"train_loss": 0.4205438792705536, "global_step": 29247, "epoch": 328, "lr": 9.724797779730956e-05}
{"train_loss": 0.5135238170623779, "global_step": 29248, "epoch": 328, "lr": 9.724778811176086e-05}
{"train_loss": 0.373676061630249, "global_step": 29249, "epoch": 328, "lr": 9.724759841986024e-05}
{"train_loss": 0.42049556970596313, "global_step": 29250, "epoch": 328, "lr": 9.724740872160777e-05}
{"train_loss": 0.42990782856941223, "global_step": 29251, "epoch": 328, "lr": 9.724721901700347e-05}
{"train_loss": 0.4680178165435791, "global_step": 29252, "epoch": 328, "lr": 9.724702930604735e-05}
{"train_loss": 0.382535457611084, "global_step": 29253, "epoch": 328, "lr": 9.724683958873943e-05}
{"train_loss": 0.31763577461242676, "global_step": 29254, "epoch": 328, "lr": 9.724664986507977e-05}
{"train_loss": 0.34911027550697327, "global_step": 29255, "epoch": 328, "lr": 9.724646013506836e-05}
{"train_loss": 0.39841166138648987, "global_step": 29256, "epoch": 328, "lr": 9.724627039870524e-05}
{"train_loss": 0.47194355726242065, "global_step": 29257, "epoch": 328, "lr": 9.724608065599046e-05}
{"train_loss": 0.42242053151130676, "global_step": 29258, "epoch": 328, "lr": 9.724589090692398e-05}
{"train_loss": 0.42504820227622986, "global_step": 29259, "epoch": 328, "lr": 9.72457011515059e-05}
{"train_loss": 0.37882891297340393, "global_step": 29260, "epoch": 328, "lr": 9.724551138973619e-05}
{"train_loss": 0.44242438673973083, "global_step": 29261, "epoch": 328, "lr": 9.72453216216149e-05}
{"train_loss": 0.4289775490760803, "global_step": 29262, "epoch": 328, "lr": 9.724513184714206e-05}
{"train_loss": 0.3303934931755066, "global_step": 29263, "epoch": 328, "lr": 9.724494206631769e-05}
{"train_loss": 0.3890819251537323, "global_step": 29264, "epoch": 328, "lr": 9.72447522791418e-05}
{"train_loss": 0.4626382291316986, "global_step": 29265, "epoch": 328, "lr": 9.724456248561445e-05}
{"train_loss": 0.38271889090538025, "global_step": 29266, "epoch": 328, "lr": 9.724437268573562e-05}
{"train_loss": 0.3440418839454651, "global_step": 29267, "epoch": 328, "lr": 9.724418287950537e-05}
{"train_loss": 0.31891292333602905, "global_step": 29268, "epoch": 328, "lr": 9.724399306692371e-05}
{"train_loss": 0.386543869972229, "global_step": 29269, "epoch": 328, "lr": 9.724380324799068e-05}
{"train_loss": 0.4520012140274048, "global_step": 29270, "epoch": 328, "lr": 9.72436134227063e-05}
{"train_loss": 0.4377210736274719, "global_step": 29271, "epoch": 328, "lr": 9.724342359107058e-05}
{"train_loss": 0.41551268100738525, "global_step": 29272, "epoch": 328, "lr": 9.724323375308357e-05}
{"train_loss": 0.5036799907684326, "global_step": 29273, "epoch": 328, "lr": 9.724304390874528e-05}
{"train_loss": 0.3631637990474701, "global_step": 29274, "epoch": 328, "lr": 9.724285405805574e-05}
{"train_loss": 0.44530895352363586, "global_step": 29275, "epoch": 328, "lr": 9.724266420101495e-05}
{"train_loss": 0.41807517409324646, "global_step": 29276, "epoch": 328, "lr": 9.724247433762299e-05}
{"train_loss": 0.35653796792030334, "global_step": 29277, "epoch": 328, "lr": 9.724228446787984e-05}
{"train_loss": 0.5299334526062012, "global_step": 29278, "epoch": 328, "lr": 9.724209459178554e-05}
{"train_loss": 0.41326743364334106, "global_step": 29279, "epoch": 328, "lr": 9.724190470934012e-05}
{"train_loss": 0.39029113630230505, "global_step": 29280, "epoch": 328, "lr": 9.72417148205436e-05, "val_loss": 2.1818277835845947}
{"train_loss": 0.412783682346344, "global_step": 29281, "epoch": 329, "lr": 9.724152492539601e-05}
{"train_loss": 0.3595212697982788, "global_step": 29282, "epoch": 329, "lr": 9.724133502389738e-05}
{"train_loss": 0.4198591113090515, "global_step": 29283, "epoch": 329, "lr": 9.724114511604772e-05}
{"train_loss": 0.501609206199646, "global_step": 29284, "epoch": 329, "lr": 9.724095520184706e-05}
{"train_loss": 0.38399070501327515, "global_step": 29285, "epoch": 329, "lr": 9.724076528129543e-05}
{"train_loss": 0.3556027114391327, "global_step": 29286, "epoch": 329, "lr": 9.724057535439286e-05}
{"train_loss": 0.3843541443347931, "global_step": 29287, "epoch": 329, "lr": 9.724038542113936e-05}
{"train_loss": 0.3685075044631958, "global_step": 29288, "epoch": 329, "lr": 9.724019548153498e-05}
{"train_loss": 0.41447266936302185, "global_step": 29289, "epoch": 329, "lr": 9.724000553557972e-05}
{"train_loss": 0.42824411392211914, "global_step": 29290, "epoch": 329, "lr": 9.723981558327363e-05}
{"train_loss": 0.42725810408592224, "global_step": 29291, "epoch": 329, "lr": 9.723962562461672e-05}
{"train_loss": 0.4280399680137634, "global_step": 29292, "epoch": 329, "lr": 9.723943565960901e-05}
{"train_loss": 0.37558773159980774, "global_step": 29293, "epoch": 329, "lr": 9.723924568825054e-05}
{"train_loss": 0.3663502335548401, "global_step": 29294, "epoch": 329, "lr": 9.723905571054133e-05}
{"train_loss": 0.4619887173175812, "global_step": 29295, "epoch": 329, "lr": 9.72388657264814e-05}
{"train_loss": 0.29080063104629517, "global_step": 29296, "epoch": 329, "lr": 9.723867573607077e-05}
{"train_loss": 0.3221271336078644, "global_step": 29297, "epoch": 329, "lr": 9.72384857393095e-05}
{"train_loss": 0.34872421622276306, "global_step": 29298, "epoch": 329, "lr": 9.723829573619758e-05}
{"train_loss": 0.439200222492218, "global_step": 29299, "epoch": 329, "lr": 9.723810572673504e-05}
{"train_loss": 0.4013682007789612, "global_step": 29300, "epoch": 329, "lr": 9.723791571092192e-05}
{"train_loss": 0.4698597192764282, "global_step": 29301, "epoch": 329, "lr": 9.723772568875824e-05}
{"train_loss": 0.25284701585769653, "global_step": 29302, "epoch": 329, "lr": 9.723753566024402e-05}
{"train_loss": 0.49743637442588806, "global_step": 29303, "epoch": 329, "lr": 9.723734562537929e-05}
{"train_loss": 0.4123353064060211, "global_step": 29304, "epoch": 329, "lr": 9.723715558416406e-05}
{"train_loss": 0.38467177748680115, "global_step": 29305, "epoch": 329, "lr": 9.723696553659839e-05}
{"train_loss": 0.3805464208126068, "global_step": 29306, "epoch": 329, "lr": 9.723677548268229e-05}
{"train_loss": 0.4188181757926941, "global_step": 29307, "epoch": 329, "lr": 9.723658542241576e-05}
{"train_loss": 0.4699086546897888, "global_step": 29308, "epoch": 329, "lr": 9.723639535579886e-05}
{"train_loss": 0.33355075120925903, "global_step": 29309, "epoch": 329, "lr": 9.723620528283159e-05}
{"train_loss": 0.2981749176979065, "global_step": 29310, "epoch": 329, "lr": 9.723601520351401e-05}
{"train_loss": 0.3319567143917084, "global_step": 29311, "epoch": 329, "lr": 9.72358251178461e-05}
{"train_loss": 0.5248386859893799, "global_step": 29312, "epoch": 329, "lr": 9.723563502582793e-05}
{"train_loss": 0.45705416798591614, "global_step": 29313, "epoch": 329, "lr": 9.723544492745949e-05}
{"train_loss": 0.38975629210472107, "global_step": 29314, "epoch": 329, "lr": 9.723525482274083e-05}
{"train_loss": 0.42748987674713135, "global_step": 29315, "epoch": 329, "lr": 9.723506471167196e-05}
{"train_loss": 0.2895506024360657, "global_step": 29316, "epoch": 329, "lr": 9.72348745942529e-05}
{"train_loss": 0.3478447496891022, "global_step": 29317, "epoch": 329, "lr": 9.723468447048372e-05}
{"train_loss": 0.49082809686660767, "global_step": 29318, "epoch": 329, "lr": 9.723449434036438e-05}
{"train_loss": 0.3434877395629883, "global_step": 29319, "epoch": 329, "lr": 9.723430420389494e-05}
{"train_loss": 0.32897862792015076, "global_step": 29320, "epoch": 329, "lr": 9.723411406107544e-05}
{"train_loss": 0.39639028906822205, "global_step": 29321, "epoch": 329, "lr": 9.72339239119059e-05}
{"train_loss": 0.34669509530067444, "global_step": 29322, "epoch": 329, "lr": 9.72337337563863e-05}
{"train_loss": 0.31561776995658875, "global_step": 29323, "epoch": 329, "lr": 9.723354359451671e-05}
{"train_loss": 0.3391244113445282, "global_step": 29324, "epoch": 329, "lr": 9.723335342629716e-05}
{"train_loss": 0.4181121289730072, "global_step": 29325, "epoch": 329, "lr": 9.723316325172766e-05}
{"train_loss": 0.42094886302948, "global_step": 29326, "epoch": 329, "lr": 9.723297307080824e-05}
{"train_loss": 0.4600607454776764, "global_step": 29327, "epoch": 329, "lr": 9.72327828835389e-05}
{"train_loss": 0.3423144817352295, "global_step": 29328, "epoch": 329, "lr": 9.723259268991969e-05}
{"train_loss": 0.42430126667022705, "global_step": 29329, "epoch": 329, "lr": 9.723240248995065e-05}
{"train_loss": 0.4022655189037323, "global_step": 29330, "epoch": 329, "lr": 9.723221228363179e-05}
{"train_loss": 0.38979649543762207, "global_step": 29331, "epoch": 329, "lr": 9.723202207096312e-05}
{"train_loss": 0.3687956929206848, "global_step": 29332, "epoch": 329, "lr": 9.723183185194469e-05}
{"train_loss": 0.4491269886493683, "global_step": 29333, "epoch": 329, "lr": 9.723164162657651e-05}
{"train_loss": 0.3275381624698639, "global_step": 29334, "epoch": 329, "lr": 9.723145139485863e-05}
{"train_loss": 0.4240275025367737, "global_step": 29335, "epoch": 329, "lr": 9.723126115679105e-05}
{"train_loss": 0.3880842626094818, "global_step": 29336, "epoch": 329, "lr": 9.723107091237378e-05}
{"train_loss": 0.37306609749794006, "global_step": 29337, "epoch": 329, "lr": 9.723088066160689e-05}
{"train_loss": 0.33909422159194946, "global_step": 29338, "epoch": 329, "lr": 9.723069040449038e-05}
{"train_loss": 0.3270934522151947, "global_step": 29339, "epoch": 329, "lr": 9.723050014102427e-05}
{"train_loss": 0.4382036030292511, "global_step": 29340, "epoch": 329, "lr": 9.72303098712086e-05}
{"train_loss": 0.3774056136608124, "global_step": 29341, "epoch": 329, "lr": 9.72301195950434e-05}
{"train_loss": 0.342107892036438, "global_step": 29342, "epoch": 329, "lr": 9.722992931252868e-05}
{"train_loss": 0.2985750734806061, "global_step": 29343, "epoch": 329, "lr": 9.722973902366446e-05}
{"train_loss": 0.36401522159576416, "global_step": 29344, "epoch": 329, "lr": 9.72295487284508e-05}
{"train_loss": 0.3197002410888672, "global_step": 29345, "epoch": 329, "lr": 9.722935842688767e-05}
{"train_loss": 0.5119127631187439, "global_step": 29346, "epoch": 329, "lr": 9.722916811897516e-05}
{"train_loss": 0.3828577399253845, "global_step": 29347, "epoch": 329, "lr": 9.722897780471325e-05}
{"train_loss": 0.4680328667163849, "global_step": 29348, "epoch": 329, "lr": 9.722878748410197e-05}
{"train_loss": 0.3575427830219269, "global_step": 29349, "epoch": 329, "lr": 9.722859715714139e-05}
{"train_loss": 0.3977316617965698, "global_step": 29350, "epoch": 329, "lr": 9.722840682383147e-05}
{"train_loss": 0.4740777313709259, "global_step": 29351, "epoch": 329, "lr": 9.722821648417227e-05}
{"train_loss": 0.44721007347106934, "global_step": 29352, "epoch": 329, "lr": 9.722802613816382e-05}
{"train_loss": 0.29681533575057983, "global_step": 29353, "epoch": 329, "lr": 9.722783578580613e-05}
{"train_loss": 0.31932029128074646, "global_step": 29354, "epoch": 329, "lr": 9.722764542709924e-05}
{"train_loss": 0.4346092939376831, "global_step": 29355, "epoch": 329, "lr": 9.722745506204317e-05}
{"train_loss": 0.2940146327018738, "global_step": 29356, "epoch": 329, "lr": 9.722726469063795e-05}
{"train_loss": 0.43704432249069214, "global_step": 29357, "epoch": 329, "lr": 9.722707431288359e-05}
{"train_loss": 0.2619115710258484, "global_step": 29358, "epoch": 329, "lr": 9.722688392878011e-05}
{"train_loss": 0.4283224642276764, "global_step": 29359, "epoch": 329, "lr": 9.722669353832756e-05}
{"train_loss": 0.39013671875, "global_step": 29360, "epoch": 329, "lr": 9.722650314152597e-05}
{"train_loss": 0.3622131943702698, "global_step": 29361, "epoch": 329, "lr": 9.722631273837535e-05}
{"train_loss": 0.3840155005455017, "global_step": 29362, "epoch": 329, "lr": 9.722612232887571e-05}
{"train_loss": 0.3420940935611725, "global_step": 29363, "epoch": 329, "lr": 9.722593191302711e-05}
{"train_loss": 0.4025905728340149, "global_step": 29364, "epoch": 329, "lr": 9.722574149082957e-05}
{"train_loss": 0.4099704325199127, "global_step": 29365, "epoch": 329, "lr": 9.722555106228308e-05}
{"train_loss": 0.44783198833465576, "global_step": 29366, "epoch": 329, "lr": 9.72253606273877e-05}
{"train_loss": 0.3656555414199829, "global_step": 29367, "epoch": 329, "lr": 9.722517018614345e-05}
{"train_loss": 0.2844727635383606, "global_step": 29368, "epoch": 329, "lr": 9.722497973855034e-05}
{"train_loss": 0.38630578959925793, "global_step": 29369, "epoch": 329, "lr": 9.722478928460843e-05, "val_loss": 2.3060898780822754}
{"train_loss": 0.37170442938804626, "global_step": 29370, "epoch": 330, "lr": 9.722459882431769e-05}
{"train_loss": 0.35456275939941406, "global_step": 29371, "epoch": 330, "lr": 9.722440835767821e-05}
{"train_loss": 0.33468344807624817, "global_step": 29372, "epoch": 330, "lr": 9.722421788468996e-05}
{"train_loss": 0.4777141809463501, "global_step": 29373, "epoch": 330, "lr": 9.722402740535301e-05}
{"train_loss": 0.34647706151008606, "global_step": 29374, "epoch": 330, "lr": 9.722383691966734e-05}
{"train_loss": 0.4075019061565399, "global_step": 29375, "epoch": 330, "lr": 9.722364642763302e-05}
{"train_loss": 0.2943430244922638, "global_step": 29376, "epoch": 330, "lr": 9.722345592925006e-05}
{"train_loss": 0.41623902320861816, "global_step": 29377, "epoch": 330, "lr": 9.722326542451848e-05}
{"train_loss": 0.38194188475608826, "global_step": 29378, "epoch": 330, "lr": 9.72230749134383e-05}
{"train_loss": 0.34835711121559143, "global_step": 29379, "epoch": 330, "lr": 9.722288439600955e-05}
{"train_loss": 0.3921701908111572, "global_step": 29380, "epoch": 330, "lr": 9.722269387223227e-05}
{"train_loss": 0.42911088466644287, "global_step": 29381, "epoch": 330, "lr": 9.722250334210645e-05}
{"train_loss": 0.3103955388069153, "global_step": 29382, "epoch": 330, "lr": 9.722231280563217e-05}
{"train_loss": 0.3794359862804413, "global_step": 29383, "epoch": 330, "lr": 9.72221222628094e-05}
{"train_loss": 0.3412143588066101, "global_step": 29384, "epoch": 330, "lr": 9.72219317136382e-05}
{"train_loss": 0.38072648644447327, "global_step": 29385, "epoch": 330, "lr": 9.72217411581186e-05}
{"train_loss": 0.38979199528694153, "global_step": 29386, "epoch": 330, "lr": 9.72215505962506e-05}
{"train_loss": 0.29803258180618286, "global_step": 29387, "epoch": 330, "lr": 9.722136002803424e-05}
{"train_loss": 0.3675368130207062, "global_step": 29388, "epoch": 330, "lr": 9.722116945346955e-05}
{"train_loss": 0.33953219652175903, "global_step": 29389, "epoch": 330, "lr": 9.722097887255655e-05}
{"train_loss": 0.3544389009475708, "global_step": 29390, "epoch": 330, "lr": 9.722078828529526e-05}
{"train_loss": 0.3163907527923584, "global_step": 29391, "epoch": 330, "lr": 9.72205976916857e-05}
{"train_loss": 0.36772188544273376, "global_step": 29392, "epoch": 330, "lr": 9.722040709172792e-05}
{"train_loss": 0.36700400710105896, "global_step": 29393, "epoch": 330, "lr": 9.722021648542192e-05}
{"train_loss": 0.4479372799396515, "global_step": 29394, "epoch": 330, "lr": 9.722002587276775e-05}
{"train_loss": 0.44238919019699097, "global_step": 29395, "epoch": 330, "lr": 9.721983525376541e-05}
{"train_loss": 0.4814191162586212, "global_step": 29396, "epoch": 330, "lr": 9.721964462841496e-05}
{"train_loss": 0.4159757196903229, "global_step": 29397, "epoch": 330, "lr": 9.721945399671639e-05}
{"train_loss": 0.42582473158836365, "global_step": 29398, "epoch": 330, "lr": 9.721926335866973e-05}
{"train_loss": 0.3055858016014099, "global_step": 29399, "epoch": 330, "lr": 9.721907271427504e-05}
{"train_loss": 0.3559826612472534, "global_step": 29400, "epoch": 330, "lr": 9.72188820635323e-05}
{"train_loss": 0.43165141344070435, "global_step": 29401, "epoch": 330, "lr": 9.721869140644157e-05}
{"train_loss": 0.360989511013031, "global_step": 29402, "epoch": 330, "lr": 9.721850074300285e-05}
{"train_loss": 0.3928675055503845, "global_step": 29403, "epoch": 330, "lr": 9.721831007321618e-05}
{"train_loss": 0.36089026927948, "global_step": 29404, "epoch": 330, "lr": 9.72181193970816e-05}
{"train_loss": 0.2693942189216614, "global_step": 29405, "epoch": 330, "lr": 9.721792871459912e-05}
{"train_loss": 0.3539009690284729, "global_step": 29406, "epoch": 330, "lr": 9.721773802576874e-05}
{"train_loss": 0.32548680901527405, "global_step": 29407, "epoch": 330, "lr": 9.721754733059052e-05}
{"train_loss": 0.3271830976009369, "global_step": 29408, "epoch": 330, "lr": 9.721735662906448e-05}
{"train_loss": 0.4083079397678375, "global_step": 29409, "epoch": 330, "lr": 9.721716592119066e-05}
{"train_loss": 0.25410398840904236, "global_step": 29410, "epoch": 330, "lr": 9.721697520696904e-05}
{"train_loss": 0.3045329451560974, "global_step": 29411, "epoch": 330, "lr": 9.721678448639968e-05}
{"train_loss": 0.32455161213874817, "global_step": 29412, "epoch": 330, "lr": 9.721659375948259e-05}
{"train_loss": 0.3790643513202667, "global_step": 29413, "epoch": 330, "lr": 9.721640302621783e-05}
{"train_loss": 0.4124380350112915, "global_step": 29414, "epoch": 330, "lr": 9.721621228660538e-05}
{"train_loss": 0.4049546718597412, "global_step": 29415, "epoch": 330, "lr": 9.72160215406453e-05}
{"train_loss": 0.4495369791984558, "global_step": 29416, "epoch": 330, "lr": 9.721583078833759e-05}
{"train_loss": 0.2767816185951233, "global_step": 29417, "epoch": 330, "lr": 9.721564002968229e-05}
{"train_loss": 0.4097084105014801, "global_step": 29418, "epoch": 330, "lr": 9.721544926467942e-05}
{"train_loss": 0.46596860885620117, "global_step": 29419, "epoch": 330, "lr": 9.721525849332902e-05}
{"train_loss": 0.37648892402648926, "global_step": 29420, "epoch": 330, "lr": 9.721506771563109e-05}
{"train_loss": 0.33289065957069397, "global_step": 29421, "epoch": 330, "lr": 9.721487693158568e-05}
{"train_loss": 0.38553786277770996, "global_step": 29422, "epoch": 330, "lr": 9.721468614119281e-05}
{"train_loss": 0.4102247357368469, "global_step": 29423, "epoch": 330, "lr": 9.721449534445249e-05}
{"train_loss": 0.4722912311553955, "global_step": 29424, "epoch": 330, "lr": 9.721430454136475e-05}
{"train_loss": 0.34289640188217163, "global_step": 29425, "epoch": 330, "lr": 9.721411373192963e-05}
{"train_loss": 0.29642269015312195, "global_step": 29426, "epoch": 330, "lr": 9.721392291614714e-05}
{"train_loss": 0.5263161659240723, "global_step": 29427, "epoch": 330, "lr": 9.721373209401732e-05}
{"train_loss": 0.35880738496780396, "global_step": 29428, "epoch": 330, "lr": 9.721354126554019e-05}
{"train_loss": 0.3621463179588318, "global_step": 29429, "epoch": 330, "lr": 9.721335043071578e-05}
{"train_loss": 0.388258695602417, "global_step": 29430, "epoch": 330, "lr": 9.721315958954411e-05}
{"train_loss": 0.4335644841194153, "global_step": 29431, "epoch": 330, "lr": 9.72129687420252e-05}
{"train_loss": 0.42112234234809875, "global_step": 29432, "epoch": 330, "lr": 9.721277788815908e-05}
{"train_loss": 0.4027901291847229, "global_step": 29433, "epoch": 330, "lr": 9.721258702794577e-05}
{"train_loss": 0.546583354473114, "global_step": 29434, "epoch": 330, "lr": 9.721239616138533e-05}
{"train_loss": 0.4061118960380554, "global_step": 29435, "epoch": 330, "lr": 9.721220528847775e-05}
{"train_loss": 0.3735998868942261, "global_step": 29436, "epoch": 330, "lr": 9.721201440922305e-05}
{"train_loss": 0.43050503730773926, "global_step": 29437, "epoch": 330, "lr": 9.721182352362129e-05}
{"train_loss": 0.3791400194168091, "global_step": 29438, "epoch": 330, "lr": 9.721163263167246e-05}
{"train_loss": 0.449546217918396, "global_step": 29439, "epoch": 330, "lr": 9.721144173337659e-05}
{"train_loss": 0.4634426236152649, "global_step": 29440, "epoch": 330, "lr": 9.721125082873373e-05}
{"train_loss": 0.35961657762527466, "global_step": 29441, "epoch": 330, "lr": 9.72110599177439e-05}
{"train_loss": 0.37406209111213684, "global_step": 29442, "epoch": 330, "lr": 9.721086900040711e-05}
{"train_loss": 0.4956939220428467, "global_step": 29443, "epoch": 330, "lr": 9.721067807672339e-05}
{"train_loss": 0.3802708685398102, "global_step": 29444, "epoch": 330, "lr": 9.721048714669279e-05}
{"train_loss": 0.3542015254497528, "global_step": 29445, "epoch": 330, "lr": 9.72102962103153e-05}
{"train_loss": 0.37040987610816956, "global_step": 29446, "epoch": 330, "lr": 9.721010526759096e-05}
{"train_loss": 0.36999672651290894, "global_step": 29447, "epoch": 330, "lr": 9.72099143185198e-05}
{"train_loss": 0.405945748090744, "global_step": 29448, "epoch": 330, "lr": 9.720972336310185e-05}
{"train_loss": 0.35943248867988586, "global_step": 29449, "epoch": 330, "lr": 9.720953240133713e-05}
{"train_loss": 0.3490147292613983, "global_step": 29450, "epoch": 330, "lr": 9.720934143322564e-05}
{"train_loss": 0.4170980453491211, "global_step": 29451, "epoch": 330, "lr": 9.720915045876744e-05}
{"train_loss": 0.32261380553245544, "global_step": 29452, "epoch": 330, "lr": 9.720895947796256e-05}
{"train_loss": 0.3377391993999481, "global_step": 29453, "epoch": 330, "lr": 9.7208768490811e-05}
{"train_loss": 0.4105350375175476, "global_step": 29454, "epoch": 330, "lr": 9.72085774973128e-05}
{"train_loss": 0.30857065320014954, "global_step": 29455, "epoch": 330, "lr": 9.720838649746798e-05}
{"train_loss": 0.4593676030635834, "global_step": 29456, "epoch": 330, "lr": 9.720819549127655e-05}
{"train_loss": 0.3123554587364197, "global_step": 29457, "epoch": 330, "lr": 9.720800447873858e-05}
{"train_loss": 0.38015852818328344, "global_step": 29458, "epoch": 330, "lr": 9.720781345985406e-05, "val_loss": 2.216219663619995, "train_action_mse_error": 32.730499267578125}
{"train_loss": 0.4045780599117279, "global_step": 29459, "epoch": 331, "lr": 9.720762243462303e-05}
{"train_loss": 0.41752246022224426, "global_step": 29460, "epoch": 331, "lr": 9.72074314030455e-05}
{"train_loss": 0.34305164217948914, "global_step": 29461, "epoch": 331, "lr": 9.72072403651215e-05}
{"train_loss": 0.3132580816745758, "global_step": 29462, "epoch": 331, "lr": 9.720704932085108e-05}
{"train_loss": 0.3722436726093292, "global_step": 29463, "epoch": 331, "lr": 9.720685827023424e-05}
{"train_loss": 0.42345693707466125, "global_step": 29464, "epoch": 331, "lr": 9.7206667213271e-05}
{"train_loss": 0.260799765586853, "global_step": 29465, "epoch": 331, "lr": 9.720647614996142e-05}
{"train_loss": 0.32825446128845215, "global_step": 29466, "epoch": 331, "lr": 9.720628508030549e-05}
{"train_loss": 0.4870363771915436, "global_step": 29467, "epoch": 331, "lr": 9.720609400430327e-05}
{"train_loss": 0.4089319407939911, "global_step": 29468, "epoch": 331, "lr": 9.720590292195475e-05}
{"train_loss": 0.3770385682582855, "global_step": 29469, "epoch": 331, "lr": 9.720571183325997e-05}
{"train_loss": 0.3528708517551422, "global_step": 29470, "epoch": 331, "lr": 9.720552073821897e-05}
{"train_loss": 0.4540867805480957, "global_step": 29471, "epoch": 331, "lr": 9.720532963683175e-05}
{"train_loss": 0.33088672161102295, "global_step": 29472, "epoch": 331, "lr": 9.720513852909836e-05}
{"train_loss": 0.33185264468193054, "global_step": 29473, "epoch": 331, "lr": 9.72049474150188e-05}
{"train_loss": 0.30054616928100586, "global_step": 29474, "epoch": 331, "lr": 9.720475629459313e-05}
{"train_loss": 0.34415844082832336, "global_step": 29475, "epoch": 331, "lr": 9.720456516782134e-05}
{"train_loss": 0.32429537177085876, "global_step": 29476, "epoch": 331, "lr": 9.720437403470348e-05}
{"train_loss": 0.5286580324172974, "global_step": 29477, "epoch": 331, "lr": 9.720418289523956e-05}
{"train_loss": 0.3867325782775879, "global_step": 29478, "epoch": 331, "lr": 9.720399174942962e-05}
{"train_loss": 0.2825792133808136, "global_step": 29479, "epoch": 331, "lr": 9.720380059727368e-05}
{"train_loss": 0.3489797115325928, "global_step": 29480, "epoch": 331, "lr": 9.720360943877176e-05}
{"train_loss": 0.41519173979759216, "global_step": 29481, "epoch": 331, "lr": 9.720341827392388e-05}
{"train_loss": 0.4234738051891327, "global_step": 29482, "epoch": 331, "lr": 9.720322710273009e-05}
{"train_loss": 0.3996503949165344, "global_step": 29483, "epoch": 331, "lr": 9.72030359251904e-05}
{"train_loss": 0.2743420898914337, "global_step": 29484, "epoch": 331, "lr": 9.720284474130483e-05}
{"train_loss": 0.35713785886764526, "global_step": 29485, "epoch": 331, "lr": 9.72026535510734e-05}
{"train_loss": 0.29688456654548645, "global_step": 29486, "epoch": 331, "lr": 9.720246235449618e-05}
{"train_loss": 0.4335886240005493, "global_step": 29487, "epoch": 331, "lr": 9.720227115157314e-05}
{"train_loss": 0.336347758769989, "global_step": 29488, "epoch": 331, "lr": 9.720207994230434e-05}
{"train_loss": 0.4144759476184845, "global_step": 29489, "epoch": 331, "lr": 9.720188872668981e-05}
{"train_loss": 0.4256161153316498, "global_step": 29490, "epoch": 331, "lr": 9.720169750472955e-05}
{"train_loss": 0.4860232174396515, "global_step": 29491, "epoch": 331, "lr": 9.72015062764236e-05}
{"train_loss": 0.37380433082580566, "global_step": 29492, "epoch": 331, "lr": 9.720131504177197e-05}
{"train_loss": 0.38405030965805054, "global_step": 29493, "epoch": 331, "lr": 9.720112380077469e-05}
{"train_loss": 0.42015084624290466, "global_step": 29494, "epoch": 331, "lr": 9.72009325534318e-05}
{"train_loss": 0.3893539607524872, "global_step": 29495, "epoch": 331, "lr": 9.720074129974334e-05}
{"train_loss": 0.31156158447265625, "global_step": 29496, "epoch": 331, "lr": 9.720055003970931e-05}
{"train_loss": 0.32016441226005554, "global_step": 29497, "epoch": 331, "lr": 9.720035877332973e-05}
{"train_loss": 0.4109767973423004, "global_step": 29498, "epoch": 331, "lr": 9.720016750060466e-05}
{"train_loss": 0.3248908817768097, "global_step": 29499, "epoch": 331, "lr": 9.719997622153407e-05}
{"train_loss": 0.4012322723865509, "global_step": 29500, "epoch": 331, "lr": 9.719978493611803e-05}
{"train_loss": 0.4116939306259155, "global_step": 29501, "epoch": 331, "lr": 9.719959364435656e-05}
{"train_loss": 0.3088485300540924, "global_step": 29502, "epoch": 331, "lr": 9.719940234624969e-05}
{"train_loss": 0.24853117763996124, "global_step": 29503, "epoch": 331, "lr": 9.719921104179742e-05}
{"train_loss": 0.47344544529914856, "global_step": 29504, "epoch": 331, "lr": 9.71990197309998e-05}
{"train_loss": 0.32879284024238586, "global_step": 29505, "epoch": 331, "lr": 9.719882841385684e-05}
{"train_loss": 0.39313599467277527, "global_step": 29506, "epoch": 331, "lr": 9.719863709036858e-05}
{"train_loss": 0.4141843616962433, "global_step": 29507, "epoch": 331, "lr": 9.719844576053502e-05}
{"train_loss": 0.3485775291919708, "global_step": 29508, "epoch": 331, "lr": 9.719825442435623e-05}
{"train_loss": 0.38826221227645874, "global_step": 29509, "epoch": 331, "lr": 9.71980630818322e-05}
{"train_loss": 0.4725717604160309, "global_step": 29510, "epoch": 331, "lr": 9.719787173296295e-05}
{"train_loss": 0.4886930286884308, "global_step": 29511, "epoch": 331, "lr": 9.719768037774855e-05}
{"train_loss": 0.4744560420513153, "global_step": 29512, "epoch": 331, "lr": 9.719748901618896e-05}
{"train_loss": 0.33865129947662354, "global_step": 29513, "epoch": 331, "lr": 9.719729764828429e-05}
{"train_loss": 0.32574737071990967, "global_step": 29514, "epoch": 331, "lr": 9.719710627403449e-05}
{"train_loss": 0.25803470611572266, "global_step": 29515, "epoch": 331, "lr": 9.719691489343961e-05}
{"train_loss": 0.4331050515174866, "global_step": 29516, "epoch": 331, "lr": 9.719672350649969e-05}
{"train_loss": 0.299045205116272, "global_step": 29517, "epoch": 331, "lr": 9.719653211321475e-05}
{"train_loss": 0.49913129210472107, "global_step": 29518, "epoch": 331, "lr": 9.71963407135848e-05}
{"train_loss": 0.4990988075733185, "global_step": 29519, "epoch": 331, "lr": 9.719614930760989e-05}
{"train_loss": 0.3030828535556793, "global_step": 29520, "epoch": 331, "lr": 9.719595789529002e-05}
{"train_loss": 0.3927636742591858, "global_step": 29521, "epoch": 331, "lr": 9.719576647662523e-05}
{"train_loss": 0.4934438467025757, "global_step": 29522, "epoch": 331, "lr": 9.719557505161554e-05}
{"train_loss": 0.45195791125297546, "global_step": 29523, "epoch": 331, "lr": 9.7195383620261e-05}
{"train_loss": 0.38646626472473145, "global_step": 29524, "epoch": 331, "lr": 9.71951921825616e-05}
{"train_loss": 0.4081817865371704, "global_step": 29525, "epoch": 331, "lr": 9.71950007385174e-05}
{"train_loss": 0.3758605420589447, "global_step": 29526, "epoch": 331, "lr": 9.719480928812839e-05}
{"train_loss": 0.3959064781665802, "global_step": 29527, "epoch": 331, "lr": 9.719461783139461e-05}
{"train_loss": 0.2984239459037781, "global_step": 29528, "epoch": 331, "lr": 9.71944263683161e-05}
{"train_loss": 0.38095805048942566, "global_step": 29529, "epoch": 331, "lr": 9.719423489889286e-05}
{"train_loss": 0.4527945816516876, "global_step": 29530, "epoch": 331, "lr": 9.719404342312495e-05}
{"train_loss": 0.5066585540771484, "global_step": 29531, "epoch": 331, "lr": 9.719385194101235e-05}
{"train_loss": 0.38044798374176025, "global_step": 29532, "epoch": 331, "lr": 9.719366045255514e-05}
{"train_loss": 0.37051331996917725, "global_step": 29533, "epoch": 331, "lr": 9.71934689577533e-05}
{"train_loss": 0.3890267610549927, "global_step": 29534, "epoch": 331, "lr": 9.719327745660689e-05}
{"train_loss": 0.35441717505455017, "global_step": 29535, "epoch": 331, "lr": 9.719308594911591e-05}
{"train_loss": 0.5356272459030151, "global_step": 29536, "epoch": 331, "lr": 9.719289443528039e-05}
{"train_loss": 0.4770950973033905, "global_step": 29537, "epoch": 331, "lr": 9.719270291510037e-05}
{"train_loss": 0.34520772099494934, "global_step": 29538, "epoch": 331, "lr": 9.719251138857585e-05}
{"train_loss": 0.32792672514915466, "global_step": 29539, "epoch": 331, "lr": 9.719231985570689e-05}
{"train_loss": 0.33594560623168945, "global_step": 29540, "epoch": 331, "lr": 9.719212831649349e-05}
{"train_loss": 0.29926708340644836, "global_step": 29541, "epoch": 331, "lr": 9.719193677093568e-05}
{"train_loss": 0.5057672262191772, "global_step": 29542, "epoch": 331, "lr": 9.71917452190335e-05}
{"train_loss": 0.38544270396232605, "global_step": 29543, "epoch": 331, "lr": 9.719155366078695e-05}
{"train_loss": 0.45875951647758484, "global_step": 29544, "epoch": 331, "lr": 9.719136209619608e-05}
{"train_loss": 0.32177260518074036, "global_step": 29545, "epoch": 331, "lr": 9.71911705252609e-05}
{"train_loss": 0.36802688241004944, "global_step": 29546, "epoch": 331, "lr": 9.719097894798146e-05}
{"train_loss": 0.3845101244998782, "global_step": 29547, "epoch": 331, "lr": 9.719078736435775e-05, "val_loss": 2.289778232574463}
{"train_loss": 0.39550426602363586, "global_step": 29548, "epoch": 332, "lr": 9.719059577438982e-05}
{"train_loss": 0.2632172703742981, "global_step": 29549, "epoch": 332, "lr": 9.71904041780777e-05}
{"train_loss": 0.4153246581554413, "global_step": 29550, "epoch": 332, "lr": 9.719021257542139e-05}
{"train_loss": 0.40492284297943115, "global_step": 29551, "epoch": 332, "lr": 9.719002096642092e-05}
{"train_loss": 0.40103405714035034, "global_step": 29552, "epoch": 332, "lr": 9.718982935107635e-05}
{"train_loss": 0.3873637616634369, "global_step": 29553, "epoch": 332, "lr": 9.718963772938767e-05}
{"train_loss": 0.4036976397037506, "global_step": 29554, "epoch": 332, "lr": 9.718944610135492e-05}
{"train_loss": 0.45951133966445923, "global_step": 29555, "epoch": 332, "lr": 9.718925446697814e-05}
{"train_loss": 0.4603588283061981, "global_step": 29556, "epoch": 332, "lr": 9.718906282625733e-05}
{"train_loss": 0.4285675287246704, "global_step": 29557, "epoch": 332, "lr": 9.718887117919251e-05}
{"train_loss": 0.5096551179885864, "global_step": 29558, "epoch": 332, "lr": 9.718867952578374e-05}
{"train_loss": 0.5422086119651794, "global_step": 29559, "epoch": 332, "lr": 9.718848786603101e-05}
{"train_loss": 0.29778972268104553, "global_step": 29560, "epoch": 332, "lr": 9.718829619993439e-05}
{"train_loss": 0.23062045872211456, "global_step": 29561, "epoch": 332, "lr": 9.718810452749386e-05}
{"train_loss": 0.4350528419017792, "global_step": 29562, "epoch": 332, "lr": 9.718791284870947e-05}
{"train_loss": 0.3192988336086273, "global_step": 29563, "epoch": 332, "lr": 9.718772116358123e-05}
{"train_loss": 0.48311424255371094, "global_step": 29564, "epoch": 332, "lr": 9.718752947210919e-05}
{"train_loss": 0.34689459204673767, "global_step": 29565, "epoch": 332, "lr": 9.718733777429335e-05}
{"train_loss": 0.3330269455909729, "global_step": 29566, "epoch": 332, "lr": 9.718714607013375e-05}
{"train_loss": 0.3470005691051483, "global_step": 29567, "epoch": 332, "lr": 9.718695435963043e-05}
{"train_loss": 0.5249716639518738, "global_step": 29568, "epoch": 332, "lr": 9.718676264278339e-05}
{"train_loss": 0.3790639340877533, "global_step": 29569, "epoch": 332, "lr": 9.718657091959265e-05}
{"train_loss": 0.3469008803367615, "global_step": 29570, "epoch": 332, "lr": 9.718637919005826e-05}
{"train_loss": 0.42880383133888245, "global_step": 29571, "epoch": 332, "lr": 9.718618745418024e-05}
{"train_loss": 0.4963279664516449, "global_step": 29572, "epoch": 332, "lr": 9.718599571195862e-05}
{"train_loss": 0.4044215679168701, "global_step": 29573, "epoch": 332, "lr": 9.71858039633934e-05}
{"train_loss": 0.4564806818962097, "global_step": 29574, "epoch": 332, "lr": 9.718561220848463e-05}
{"train_loss": 0.30586427450180054, "global_step": 29575, "epoch": 332, "lr": 9.718542044723234e-05}
{"train_loss": 0.30284905433654785, "global_step": 29576, "epoch": 332, "lr": 9.718522867963653e-05}
{"train_loss": 0.4614410400390625, "global_step": 29577, "epoch": 332, "lr": 9.718503690569725e-05}
{"train_loss": 0.4096165895462036, "global_step": 29578, "epoch": 332, "lr": 9.71848451254145e-05}
{"train_loss": 0.2492510974407196, "global_step": 29579, "epoch": 332, "lr": 9.718465333878834e-05}
{"train_loss": 0.420366108417511, "global_step": 29580, "epoch": 332, "lr": 9.718446154581879e-05}
{"train_loss": 0.3297567367553711, "global_step": 29581, "epoch": 332, "lr": 9.718426974650584e-05}
{"train_loss": 0.38191506266593933, "global_step": 29582, "epoch": 332, "lr": 9.718407794084954e-05}
{"train_loss": 0.35889166593551636, "global_step": 29583, "epoch": 332, "lr": 9.718388612884994e-05}
{"train_loss": 0.38122716546058655, "global_step": 29584, "epoch": 332, "lr": 9.718369431050702e-05}
{"train_loss": 0.24980078637599945, "global_step": 29585, "epoch": 332, "lr": 9.718350248582084e-05}
{"train_loss": 0.3959266245365143, "global_step": 29586, "epoch": 332, "lr": 9.718331065479141e-05}
{"train_loss": 0.3281504809856415, "global_step": 29587, "epoch": 332, "lr": 9.718311881741875e-05}
{"train_loss": 0.3780549466609955, "global_step": 29588, "epoch": 332, "lr": 9.718292697370291e-05}
{"train_loss": 0.40625301003456116, "global_step": 29589, "epoch": 332, "lr": 9.718273512364389e-05}
{"train_loss": 0.3617066442966461, "global_step": 29590, "epoch": 332, "lr": 9.718254326724172e-05}
{"train_loss": 0.3224059045314789, "global_step": 29591, "epoch": 332, "lr": 9.718235140449645e-05}
{"train_loss": 0.27667221426963806, "global_step": 29592, "epoch": 332, "lr": 9.718215953540808e-05}
{"train_loss": 0.2300613820552826, "global_step": 29593, "epoch": 332, "lr": 9.718196765997663e-05}
{"train_loss": 0.3714147210121155, "global_step": 29594, "epoch": 332, "lr": 9.718177577820215e-05}
{"train_loss": 0.35564351081848145, "global_step": 29595, "epoch": 332, "lr": 9.718158389008467e-05}
{"train_loss": 0.3345429599285126, "global_step": 29596, "epoch": 332, "lr": 9.718139199562418e-05}
{"train_loss": 0.34499508142471313, "global_step": 29597, "epoch": 332, "lr": 9.718120009482075e-05}
{"train_loss": 0.3236057460308075, "global_step": 29598, "epoch": 332, "lr": 9.718100818767436e-05}
{"train_loss": 0.3162258267402649, "global_step": 29599, "epoch": 332, "lr": 9.718081627418507e-05}
{"train_loss": 0.3239239454269409, "global_step": 29600, "epoch": 332, "lr": 9.718062435435289e-05}
{"train_loss": 0.4759698510169983, "global_step": 29601, "epoch": 332, "lr": 9.718043242817785e-05}
{"train_loss": 0.2883669137954712, "global_step": 29602, "epoch": 332, "lr": 9.718024049565997e-05}
{"train_loss": 0.3076382279396057, "global_step": 29603, "epoch": 332, "lr": 9.71800485567993e-05}
{"train_loss": 0.48895981907844543, "global_step": 29604, "epoch": 332, "lr": 9.717985661159584e-05}
{"train_loss": 0.31661462783813477, "global_step": 29605, "epoch": 332, "lr": 9.717966466004961e-05}
{"train_loss": 0.4127490520477295, "global_step": 29606, "epoch": 332, "lr": 9.717947270216066e-05}
{"train_loss": 0.3921799957752228, "global_step": 29607, "epoch": 332, "lr": 9.7179280737929e-05}
{"train_loss": 0.32349511981010437, "global_step": 29608, "epoch": 332, "lr": 9.717908876735466e-05}
{"train_loss": 0.37787100672721863, "global_step": 29609, "epoch": 332, "lr": 9.717889679043768e-05}
{"train_loss": 0.44400689005851746, "global_step": 29610, "epoch": 332, "lr": 9.717870480717806e-05}
{"train_loss": 0.30396363139152527, "global_step": 29611, "epoch": 332, "lr": 9.717851281757584e-05}
{"train_loss": 0.32644814252853394, "global_step": 29612, "epoch": 332, "lr": 9.717832082163104e-05}
{"train_loss": 0.3166273534297943, "global_step": 29613, "epoch": 332, "lr": 9.71781288193437e-05}
{"train_loss": 0.2655526101589203, "global_step": 29614, "epoch": 332, "lr": 9.717793681071384e-05}
{"train_loss": 0.38727840781211853, "global_step": 29615, "epoch": 332, "lr": 9.717774479574146e-05}
{"train_loss": 0.35956257581710815, "global_step": 29616, "epoch": 332, "lr": 9.717755277442662e-05}
{"train_loss": 0.4019824266433716, "global_step": 29617, "epoch": 332, "lr": 9.717736074676932e-05}
{"train_loss": 0.32087332010269165, "global_step": 29618, "epoch": 332, "lr": 9.717716871276963e-05}
{"train_loss": 0.3070671260356903, "global_step": 29619, "epoch": 332, "lr": 9.717697667242751e-05}
{"train_loss": 0.37916478514671326, "global_step": 29620, "epoch": 332, "lr": 9.717678462574305e-05}
{"train_loss": 0.4076277017593384, "global_step": 29621, "epoch": 332, "lr": 9.717659257271623e-05}
{"train_loss": 0.3364114761352539, "global_step": 29622, "epoch": 332, "lr": 9.717640051334709e-05}
{"train_loss": 0.4999544620513916, "global_step": 29623, "epoch": 332, "lr": 9.717620844763566e-05}
{"train_loss": 0.3130519688129425, "global_step": 29624, "epoch": 332, "lr": 9.717601637558197e-05}
{"train_loss": 0.31026190519332886, "global_step": 29625, "epoch": 332, "lr": 9.717582429718604e-05}
{"train_loss": 0.3566068708896637, "global_step": 29626, "epoch": 332, "lr": 9.717563221244788e-05}
{"train_loss": 0.3423413932323456, "global_step": 29627, "epoch": 332, "lr": 9.717544012136755e-05}
{"train_loss": 0.4202601909637451, "global_step": 29628, "epoch": 332, "lr": 9.717524802394504e-05}
{"train_loss": 0.32740071415901184, "global_step": 29629, "epoch": 332, "lr": 9.717505592018041e-05}
{"train_loss": 0.4195285141468048, "global_step": 29630, "epoch": 332, "lr": 9.717486381007366e-05}
{"train_loss": 0.429157555103302, "global_step": 29631, "epoch": 332, "lr": 9.717467169362484e-05}
{"train_loss": 0.3653988540172577, "global_step": 29632, "epoch": 332, "lr": 9.717447957083395e-05}
{"train_loss": 0.4496021568775177, "global_step": 29633, "epoch": 332, "lr": 9.7174287441701e-05}
{"train_loss": 0.3555656969547272, "global_step": 29634, "epoch": 332, "lr": 9.717409530622607e-05}
{"train_loss": 0.3812737464904785, "global_step": 29635, "epoch": 332, "lr": 9.717390316440916e-05}
{"train_loss": 0.37300387928994855, "global_step": 29636, "epoch": 332, "lr": 9.717371101625028e-05, "val_loss": 2.302583932876587}
{"train_loss": 0.42346587777137756, "global_step": 29637, "epoch": 333, "lr": 9.717351886174948e-05}
{"train_loss": 0.3574866056442261, "global_step": 29638, "epoch": 333, "lr": 9.717332670090678e-05}
{"train_loss": 0.38752299547195435, "global_step": 29639, "epoch": 333, "lr": 9.717313453372219e-05}
{"train_loss": 0.3735530376434326, "global_step": 29640, "epoch": 333, "lr": 9.717294236019573e-05}
{"train_loss": 0.4830807149410248, "global_step": 29641, "epoch": 333, "lr": 9.717275018032748e-05}
{"train_loss": 0.4085410237312317, "global_step": 29642, "epoch": 333, "lr": 9.71725579941174e-05}
{"train_loss": 0.5119980573654175, "global_step": 29643, "epoch": 333, "lr": 9.717236580156556e-05}
{"train_loss": 0.40343570709228516, "global_step": 29644, "epoch": 333, "lr": 9.717217360267196e-05}
{"train_loss": 0.40733712911605835, "global_step": 29645, "epoch": 333, "lr": 9.717198139743665e-05}
{"train_loss": 0.32636702060699463, "global_step": 29646, "epoch": 333, "lr": 9.717178918585963e-05}
{"train_loss": 0.3086552023887634, "global_step": 29647, "epoch": 333, "lr": 9.717159696794094e-05}
{"train_loss": 0.4513404369354248, "global_step": 29648, "epoch": 333, "lr": 9.717140474368061e-05}
{"train_loss": 0.5401601791381836, "global_step": 29649, "epoch": 333, "lr": 9.717121251307866e-05}
{"train_loss": 0.505770742893219, "global_step": 29650, "epoch": 333, "lr": 9.71710202761351e-05}
{"train_loss": 0.37651339173316956, "global_step": 29651, "epoch": 333, "lr": 9.717082803284998e-05}
{"train_loss": 0.4308222830295563, "global_step": 29652, "epoch": 333, "lr": 9.717063578322332e-05}
{"train_loss": 0.466344952583313, "global_step": 29653, "epoch": 333, "lr": 9.717044352725514e-05}
{"train_loss": 0.4236185550689697, "global_step": 29654, "epoch": 333, "lr": 9.717025126494547e-05}
{"train_loss": 0.6697841286659241, "global_step": 29655, "epoch": 333, "lr": 9.717005899629435e-05}
{"train_loss": 0.3054933547973633, "global_step": 29656, "epoch": 333, "lr": 9.716986672130178e-05}
{"train_loss": 0.3542189598083496, "global_step": 29657, "epoch": 333, "lr": 9.716967443996779e-05}
{"train_loss": 0.3918212354183197, "global_step": 29658, "epoch": 333, "lr": 9.71694821522924e-05}
{"train_loss": 0.4436420500278473, "global_step": 29659, "epoch": 333, "lr": 9.716928985827566e-05}
{"train_loss": 0.3761787712574005, "global_step": 29660, "epoch": 333, "lr": 9.716909755791759e-05}
{"train_loss": 0.4731459319591522, "global_step": 29661, "epoch": 333, "lr": 9.71689052512182e-05}
{"train_loss": 0.3828655481338501, "global_step": 29662, "epoch": 333, "lr": 9.716871293817753e-05}
{"train_loss": 0.42484238743782043, "global_step": 29663, "epoch": 333, "lr": 9.716852061879561e-05}
{"train_loss": 0.40674707293510437, "global_step": 29664, "epoch": 333, "lr": 9.716832829307246e-05}
{"train_loss": 0.323792040348053, "global_step": 29665, "epoch": 333, "lr": 9.716813596100809e-05}
{"train_loss": 0.37224385142326355, "global_step": 29666, "epoch": 333, "lr": 9.716794362260253e-05}
{"train_loss": 0.385306715965271, "global_step": 29667, "epoch": 333, "lr": 9.716775127785582e-05}
{"train_loss": 0.4100790023803711, "global_step": 29668, "epoch": 333, "lr": 9.7167558926768e-05}
{"train_loss": 0.3415571451187134, "global_step": 29669, "epoch": 333, "lr": 9.716736656933906e-05}
{"train_loss": 0.36968234181404114, "global_step": 29670, "epoch": 333, "lr": 9.716717420556905e-05}
{"train_loss": 0.3748706579208374, "global_step": 29671, "epoch": 333, "lr": 9.716698183545799e-05}
{"train_loss": 0.44530463218688965, "global_step": 29672, "epoch": 333, "lr": 9.71667894590059e-05}
{"train_loss": 0.38469889760017395, "global_step": 29673, "epoch": 333, "lr": 9.716659707621281e-05}
{"train_loss": 0.44323161244392395, "global_step": 29674, "epoch": 333, "lr": 9.716640468707874e-05}
{"train_loss": 0.4549838602542877, "global_step": 29675, "epoch": 333, "lr": 9.716621229160373e-05}
{"train_loss": 0.3729008734226227, "global_step": 29676, "epoch": 333, "lr": 9.71660198897878e-05}
{"train_loss": 0.3526284694671631, "global_step": 29677, "epoch": 333, "lr": 9.716582748163097e-05}
{"train_loss": 0.4581437408924103, "global_step": 29678, "epoch": 333, "lr": 9.716563506713325e-05}
{"train_loss": 0.4086161255836487, "global_step": 29679, "epoch": 333, "lr": 9.716544264629471e-05}
{"train_loss": 0.24618938565254211, "global_step": 29680, "epoch": 333, "lr": 9.716525021911535e-05}
{"train_loss": 0.38163259625434875, "global_step": 29681, "epoch": 333, "lr": 9.716505778559519e-05}
{"train_loss": 0.42850926518440247, "global_step": 29682, "epoch": 333, "lr": 9.716486534573426e-05}
{"train_loss": 0.3569009602069855, "global_step": 29683, "epoch": 333, "lr": 9.71646728995326e-05}
{"train_loss": 0.4081316888332367, "global_step": 29684, "epoch": 333, "lr": 9.716448044699021e-05}
{"train_loss": 0.4030778408050537, "global_step": 29685, "epoch": 333, "lr": 9.716428798810714e-05}
{"train_loss": 0.29706889390945435, "global_step": 29686, "epoch": 333, "lr": 9.71640955228834e-05}
{"train_loss": 0.4349166452884674, "global_step": 29687, "epoch": 333, "lr": 9.716390305131903e-05}
{"train_loss": 0.4351678490638733, "global_step": 29688, "epoch": 333, "lr": 9.716371057341405e-05}
{"train_loss": 0.3490239381790161, "global_step": 29689, "epoch": 333, "lr": 9.716351808916847e-05}
{"train_loss": 0.39193493127822876, "global_step": 29690, "epoch": 333, "lr": 9.716332559858235e-05}
{"train_loss": 0.34044042229652405, "global_step": 29691, "epoch": 333, "lr": 9.716313310165568e-05}
{"train_loss": 0.35765066742897034, "global_step": 29692, "epoch": 333, "lr": 9.71629405983885e-05}
{"train_loss": 0.31605497002601624, "global_step": 29693, "epoch": 333, "lr": 9.716274808878084e-05}
{"train_loss": 0.40849998593330383, "global_step": 29694, "epoch": 333, "lr": 9.716255557283272e-05}
{"train_loss": 0.39724618196487427, "global_step": 29695, "epoch": 333, "lr": 9.716236305054418e-05}
{"train_loss": 0.3231920301914215, "global_step": 29696, "epoch": 333, "lr": 9.716217052191523e-05}
{"train_loss": 0.43438997864723206, "global_step": 29697, "epoch": 333, "lr": 9.71619779869459e-05}
{"train_loss": 0.42967575788497925, "global_step": 29698, "epoch": 333, "lr": 9.716178544563623e-05}
{"train_loss": 0.3755136728286743, "global_step": 29699, "epoch": 333, "lr": 9.716159289798621e-05}
{"train_loss": 0.37743598222732544, "global_step": 29700, "epoch": 333, "lr": 9.71614003439959e-05}
{"train_loss": 0.33723780512809753, "global_step": 29701, "epoch": 333, "lr": 9.716120778366531e-05}
{"train_loss": 0.31322619318962097, "global_step": 29702, "epoch": 333, "lr": 9.716101521699448e-05}
{"train_loss": 0.4956958591938019, "global_step": 29703, "epoch": 333, "lr": 9.716082264398344e-05}
{"train_loss": 0.4501507878303528, "global_step": 29704, "epoch": 333, "lr": 9.716063006463217e-05}
{"train_loss": 0.4580456614494324, "global_step": 29705, "epoch": 333, "lr": 9.716043747894075e-05}
{"train_loss": 0.3730967342853546, "global_step": 29706, "epoch": 333, "lr": 9.716024488690916e-05}
{"train_loss": 0.3620692491531372, "global_step": 29707, "epoch": 333, "lr": 9.716005228853747e-05}
{"train_loss": 0.31260862946510315, "global_step": 29708, "epoch": 333, "lr": 9.715985968382568e-05}
{"train_loss": 0.38737595081329346, "global_step": 29709, "epoch": 333, "lr": 9.715966707277382e-05}
{"train_loss": 0.32879897952079773, "global_step": 29710, "epoch": 333, "lr": 9.715947445538193e-05}
{"train_loss": 0.3529479503631592, "global_step": 29711, "epoch": 333, "lr": 9.715928183165e-05}
{"train_loss": 0.3753357529640198, "global_step": 29712, "epoch": 333, "lr": 9.71590892015781e-05}
{"train_loss": 0.3823096454143524, "global_step": 29713, "epoch": 333, "lr": 9.715889656516623e-05}
{"train_loss": 0.3994612395763397, "global_step": 29714, "epoch": 333, "lr": 9.715870392241442e-05}
{"train_loss": 0.38063666224479675, "global_step": 29715, "epoch": 333, "lr": 9.715851127332269e-05}
{"train_loss": 0.31801894307136536, "global_step": 29716, "epoch": 333, "lr": 9.715831861789108e-05}
{"train_loss": 0.300034761428833, "global_step": 29717, "epoch": 333, "lr": 9.715812595611962e-05}
{"train_loss": 0.43750908970832825, "global_step": 29718, "epoch": 333, "lr": 9.715793328800831e-05}
{"train_loss": 0.3884255588054657, "global_step": 29719, "epoch": 333, "lr": 9.71577406135572e-05}
{"train_loss": 0.44609829783439636, "global_step": 29720, "epoch": 333, "lr": 9.715754793276629e-05}
{"train_loss": 0.42714908719062805, "global_step": 29721, "epoch": 333, "lr": 9.715735524563563e-05}
{"train_loss": 0.4537891149520874, "global_step": 29722, "epoch": 333, "lr": 9.715716255216524e-05}
{"train_loss": 0.384258896112442, "global_step": 29723, "epoch": 333, "lr": 9.715696985235515e-05}
{"train_loss": 0.46797946095466614, "global_step": 29724, "epoch": 333, "lr": 9.715677714620537e-05}
{"train_loss": 0.3977477339546332, "global_step": 29725, "epoch": 333, "lr": 9.715658443371594e-05, "val_loss": 2.2321250438690186}
{"train_loss": 0.41097620129585266, "global_step": 29726, "epoch": 334, "lr": 9.715639171488689e-05}
{"train_loss": 0.4018252491950989, "global_step": 29727, "epoch": 334, "lr": 9.715619898971823e-05}
{"train_loss": 0.339374303817749, "global_step": 29728, "epoch": 334, "lr": 9.715600625821e-05}
{"train_loss": 0.2806035876274109, "global_step": 29729, "epoch": 334, "lr": 9.71558135203622e-05}
{"train_loss": 0.40933528542518616, "global_step": 29730, "epoch": 334, "lr": 9.71556207761749e-05}
{"train_loss": 0.33145591616630554, "global_step": 29731, "epoch": 334, "lr": 9.71554280256481e-05}
{"train_loss": 0.4406115710735321, "global_step": 29732, "epoch": 334, "lr": 9.71552352687818e-05}
{"train_loss": 0.5201650261878967, "global_step": 29733, "epoch": 334, "lr": 9.715504250557609e-05}
{"train_loss": 0.37469255924224854, "global_step": 29734, "epoch": 334, "lr": 9.715484973603094e-05}
{"train_loss": 0.36802569031715393, "global_step": 29735, "epoch": 334, "lr": 9.71546569601464e-05}
{"train_loss": 0.2892721891403198, "global_step": 29736, "epoch": 334, "lr": 9.715446417792248e-05}
{"train_loss": 0.34582236409187317, "global_step": 29737, "epoch": 334, "lr": 9.715427138935923e-05}
{"train_loss": 0.3623220920562744, "global_step": 29738, "epoch": 334, "lr": 9.715407859445667e-05}
{"train_loss": 0.4034569561481476, "global_step": 29739, "epoch": 334, "lr": 9.715388579321479e-05}
{"train_loss": 0.4749438464641571, "global_step": 29740, "epoch": 334, "lr": 9.715369298563368e-05}
{"train_loss": 0.39714834094047546, "global_step": 29741, "epoch": 334, "lr": 9.71535001717133e-05}
{"train_loss": 0.39568713307380676, "global_step": 29742, "epoch": 334, "lr": 9.715330735145372e-05}
{"train_loss": 0.39224767684936523, "global_step": 29743, "epoch": 334, "lr": 9.715311452485495e-05}
{"train_loss": 0.4335595369338989, "global_step": 29744, "epoch": 334, "lr": 9.715292169191702e-05}
{"train_loss": 0.33586740493774414, "global_step": 29745, "epoch": 334, "lr": 9.715272885263995e-05}
{"train_loss": 0.3778470754623413, "global_step": 29746, "epoch": 334, "lr": 9.715253600702378e-05}
{"train_loss": 0.3819020092487335, "global_step": 29747, "epoch": 334, "lr": 9.715234315506852e-05}
{"train_loss": 0.4516785144805908, "global_step": 29748, "epoch": 334, "lr": 9.71521502967742e-05}
{"train_loss": 0.39911842346191406, "global_step": 29749, "epoch": 334, "lr": 9.715195743214086e-05}
{"train_loss": 0.3954312205314636, "global_step": 29750, "epoch": 334, "lr": 9.71517645611685e-05}
{"train_loss": 0.2798490524291992, "global_step": 29751, "epoch": 334, "lr": 9.715157168385717e-05}
{"train_loss": 0.2671179175376892, "global_step": 29752, "epoch": 334, "lr": 9.715137880020688e-05}
{"train_loss": 0.30001100897789, "global_step": 29753, "epoch": 334, "lr": 9.715118591021766e-05}
{"train_loss": 0.5366308093070984, "global_step": 29754, "epoch": 334, "lr": 9.715099301388954e-05}
{"train_loss": 0.34617528319358826, "global_step": 29755, "epoch": 334, "lr": 9.715080011122255e-05}
{"train_loss": 0.4280567169189453, "global_step": 29756, "epoch": 334, "lr": 9.71506072022167e-05}
{"train_loss": 0.3442348539829254, "global_step": 29757, "epoch": 334, "lr": 9.715041428687204e-05}
{"train_loss": 0.3208269774913788, "global_step": 29758, "epoch": 334, "lr": 9.715022136518857e-05}
{"train_loss": 0.3822326064109802, "global_step": 29759, "epoch": 334, "lr": 9.715002843716632e-05}
{"train_loss": 0.40664851665496826, "global_step": 29760, "epoch": 334, "lr": 9.714983550280534e-05}
{"train_loss": 0.38073158264160156, "global_step": 29761, "epoch": 334, "lr": 9.714964256210563e-05}
{"train_loss": 0.3266601264476776, "global_step": 29762, "epoch": 334, "lr": 9.714944961506722e-05}
{"train_loss": 0.41204071044921875, "global_step": 29763, "epoch": 334, "lr": 9.714925666169015e-05}
{"train_loss": 0.3176279664039612, "global_step": 29764, "epoch": 334, "lr": 9.714906370197444e-05}
{"train_loss": 0.3344179689884186, "global_step": 29765, "epoch": 334, "lr": 9.71488707359201e-05}
{"train_loss": 0.33836206793785095, "global_step": 29766, "epoch": 334, "lr": 9.714867776352719e-05}
{"train_loss": 0.26896190643310547, "global_step": 29767, "epoch": 334, "lr": 9.714848478479571e-05}
{"train_loss": 0.27296534180641174, "global_step": 29768, "epoch": 334, "lr": 9.714829179972567e-05}
{"train_loss": 0.3835829198360443, "global_step": 29769, "epoch": 334, "lr": 9.714809880831712e-05}
{"train_loss": 0.3148154616355896, "global_step": 29770, "epoch": 334, "lr": 9.714790581057009e-05}
{"train_loss": 0.33820274472236633, "global_step": 29771, "epoch": 334, "lr": 9.71477128064846e-05}
{"train_loss": 0.3678058981895447, "global_step": 29772, "epoch": 334, "lr": 9.71475197960607e-05}
{"train_loss": 0.3295500576496124, "global_step": 29773, "epoch": 334, "lr": 9.714732677929835e-05}
{"train_loss": 0.35715150833129883, "global_step": 29774, "epoch": 334, "lr": 9.714713375619764e-05}
{"train_loss": 0.4759134352207184, "global_step": 29775, "epoch": 334, "lr": 9.714694072675856e-05}
{"train_loss": 0.39763063192367554, "global_step": 29776, "epoch": 334, "lr": 9.714674769098116e-05}
{"train_loss": 0.45583534240722656, "global_step": 29777, "epoch": 334, "lr": 9.714655464886545e-05}
{"train_loss": 0.32160547375679016, "global_step": 29778, "epoch": 334, "lr": 9.714636160041145e-05}
{"train_loss": 0.5190012454986572, "global_step": 29779, "epoch": 334, "lr": 9.714616854561921e-05}
{"train_loss": 0.3679341971874237, "global_step": 29780, "epoch": 334, "lr": 9.714597548448874e-05}
{"train_loss": 0.4380638301372528, "global_step": 29781, "epoch": 334, "lr": 9.714578241702005e-05}
{"train_loss": 0.3087810277938843, "global_step": 29782, "epoch": 334, "lr": 9.714558934321321e-05}
{"train_loss": 0.27499470114707947, "global_step": 29783, "epoch": 334, "lr": 9.714539626306821e-05}
{"train_loss": 0.2862895429134369, "global_step": 29784, "epoch": 334, "lr": 9.714520317658509e-05}
{"train_loss": 0.3697291314601898, "global_step": 29785, "epoch": 334, "lr": 9.714501008376387e-05}
{"train_loss": 0.36631184816360474, "global_step": 29786, "epoch": 334, "lr": 9.714481698460456e-05}
{"train_loss": 0.2715565860271454, "global_step": 29787, "epoch": 334, "lr": 9.714462387910723e-05}
{"train_loss": 0.3894016444683075, "global_step": 29788, "epoch": 334, "lr": 9.714443076727186e-05}
{"train_loss": 0.36991527676582336, "global_step": 29789, "epoch": 334, "lr": 9.714423764909851e-05}
{"train_loss": 0.24858999252319336, "global_step": 29790, "epoch": 334, "lr": 9.714404452458719e-05}
{"train_loss": 0.5191153883934021, "global_step": 29791, "epoch": 334, "lr": 9.714385139373793e-05}
{"train_loss": 0.2859129309654236, "global_step": 29792, "epoch": 334, "lr": 9.714365825655074e-05}
{"train_loss": 0.39053666591644287, "global_step": 29793, "epoch": 334, "lr": 9.714346511302568e-05}
{"train_loss": 0.37153536081314087, "global_step": 29794, "epoch": 334, "lr": 9.714327196316274e-05}
{"train_loss": 0.45006611943244934, "global_step": 29795, "epoch": 334, "lr": 9.714307880696196e-05}
{"train_loss": 0.37891796231269836, "global_step": 29796, "epoch": 334, "lr": 9.714288564442337e-05}
{"train_loss": 0.2611430585384369, "global_step": 29797, "epoch": 334, "lr": 9.714269247554701e-05}
{"train_loss": 0.2629249691963196, "global_step": 29798, "epoch": 334, "lr": 9.714249930033288e-05}
{"train_loss": 0.3950454890727997, "global_step": 29799, "epoch": 334, "lr": 9.714230611878101e-05}
{"train_loss": 0.48220545053482056, "global_step": 29800, "epoch": 334, "lr": 9.714211293089144e-05}
{"train_loss": 0.32258087396621704, "global_step": 29801, "epoch": 334, "lr": 9.714191973666417e-05}
{"train_loss": 0.38346341252326965, "global_step": 29802, "epoch": 334, "lr": 9.714172653609925e-05}
{"train_loss": 0.32426348328590393, "global_step": 29803, "epoch": 334, "lr": 9.714153332919671e-05}
{"train_loss": 0.390588641166687, "global_step": 29804, "epoch": 334, "lr": 9.714134011595657e-05}
{"train_loss": 0.38416025042533875, "global_step": 29805, "epoch": 334, "lr": 9.714114689637883e-05}
{"train_loss": 0.3564264476299286, "global_step": 29806, "epoch": 334, "lr": 9.714095367046356e-05}
{"train_loss": 0.5475010871887207, "global_step": 29807, "epoch": 334, "lr": 9.714076043821074e-05}
{"train_loss": 0.4425320029258728, "global_step": 29808, "epoch": 334, "lr": 9.714056719962044e-05}
{"train_loss": 0.36194390058517456, "global_step": 29809, "epoch": 334, "lr": 9.714037395469266e-05}
{"train_loss": 0.4056105315685272, "global_step": 29810, "epoch": 334, "lr": 9.714018070342744e-05}
{"train_loss": 0.43128111958503723, "global_step": 29811, "epoch": 334, "lr": 9.713998744582478e-05}
{"train_loss": 0.3385380208492279, "global_step": 29812, "epoch": 334, "lr": 9.713979418188473e-05}
{"train_loss": 0.3616991639137268, "global_step": 29813, "epoch": 334, "lr": 9.71396009116073e-05}
{"train_loss": 0.3746099475394474, "global_step": 29814, "epoch": 334, "lr": 9.713940763499253e-05, "val_loss": 2.307398557662964}
{"train_loss": 0.46065154671669006, "global_step": 29815, "epoch": 335, "lr": 9.713921435204044e-05}
{"train_loss": 0.3098823130130768, "global_step": 29816, "epoch": 335, "lr": 9.713902106275106e-05}
{"train_loss": 0.46962714195251465, "global_step": 29817, "epoch": 335, "lr": 9.713882776712442e-05}
{"train_loss": 0.3202020525932312, "global_step": 29818, "epoch": 335, "lr": 9.713863446516054e-05}
{"train_loss": 0.2955842912197113, "global_step": 29819, "epoch": 335, "lr": 9.713844115685943e-05}
{"train_loss": 0.31173568964004517, "global_step": 29820, "epoch": 335, "lr": 9.713824784222114e-05}
{"train_loss": 0.37114885449409485, "global_step": 29821, "epoch": 335, "lr": 9.713805452124569e-05}
{"train_loss": 0.3644371032714844, "global_step": 29822, "epoch": 335, "lr": 9.71378611939331e-05}
{"train_loss": 0.42295804619789124, "global_step": 29823, "epoch": 335, "lr": 9.71376678602834e-05}
{"train_loss": 0.3838922083377838, "global_step": 29824, "epoch": 335, "lr": 9.71374745202966e-05}
{"train_loss": 0.38346245884895325, "global_step": 29825, "epoch": 335, "lr": 9.713728117397277e-05}
{"train_loss": 0.37799909710884094, "global_step": 29826, "epoch": 335, "lr": 9.71370878213119e-05}
{"train_loss": 0.3234541416168213, "global_step": 29827, "epoch": 335, "lr": 9.7136894462314e-05}
{"train_loss": 0.41412845253944397, "global_step": 29828, "epoch": 335, "lr": 9.713670109697914e-05}
{"train_loss": 0.3452416956424713, "global_step": 29829, "epoch": 335, "lr": 9.713650772530733e-05}
{"train_loss": 0.4096415638923645, "global_step": 29830, "epoch": 335, "lr": 9.713631434729857e-05}
{"train_loss": 0.30201026797294617, "global_step": 29831, "epoch": 335, "lr": 9.713612096295292e-05}
{"train_loss": 0.47190773487091064, "global_step": 29832, "epoch": 335, "lr": 9.713592757227039e-05}
{"train_loss": 0.3571271002292633, "global_step": 29833, "epoch": 335, "lr": 9.713573417525101e-05}
{"train_loss": 0.36082684993743896, "global_step": 29834, "epoch": 335, "lr": 9.71355407718948e-05}
{"train_loss": 0.4916026294231415, "global_step": 29835, "epoch": 335, "lr": 9.71353473622018e-05}
{"train_loss": 0.33089834451675415, "global_step": 29836, "epoch": 335, "lr": 9.713515394617202e-05}
{"train_loss": 0.3182482123374939, "global_step": 29837, "epoch": 335, "lr": 9.713496052380551e-05}
{"train_loss": 0.31505563855171204, "global_step": 29838, "epoch": 335, "lr": 9.713476709510225e-05}
{"train_loss": 0.44040447473526, "global_step": 29839, "epoch": 335, "lr": 9.71345736600623e-05}
{"train_loss": 0.34402868151664734, "global_step": 29840, "epoch": 335, "lr": 9.71343802186857e-05}
{"train_loss": 0.3127114772796631, "global_step": 29841, "epoch": 335, "lr": 9.713418677097246e-05}
{"train_loss": 0.3360547125339508, "global_step": 29842, "epoch": 335, "lr": 9.713399331692259e-05}
{"train_loss": 0.4581366777420044, "global_step": 29843, "epoch": 335, "lr": 9.713379985653613e-05}
{"train_loss": 0.3441224992275238, "global_step": 29844, "epoch": 335, "lr": 9.71336063898131e-05}
{"train_loss": 0.3809102475643158, "global_step": 29845, "epoch": 335, "lr": 9.713341291675354e-05}
{"train_loss": 0.47935837507247925, "global_step": 29846, "epoch": 335, "lr": 9.713321943735747e-05}
{"train_loss": 0.4081827402114868, "global_step": 29847, "epoch": 335, "lr": 9.71330259516249e-05}
{"train_loss": 0.37940549850463867, "global_step": 29848, "epoch": 335, "lr": 9.713283245955588e-05}
{"train_loss": 0.3110734820365906, "global_step": 29849, "epoch": 335, "lr": 9.713263896115042e-05}
{"train_loss": 0.37998756766319275, "global_step": 29850, "epoch": 335, "lr": 9.713244545640857e-05}
{"train_loss": 0.5127735733985901, "global_step": 29851, "epoch": 335, "lr": 9.71322519453303e-05}
{"train_loss": 0.3290930688381195, "global_step": 29852, "epoch": 335, "lr": 9.71320584279157e-05}
{"train_loss": 0.2939647138118744, "global_step": 29853, "epoch": 335, "lr": 9.713186490416474e-05}
{"train_loss": 0.31380465626716614, "global_step": 29854, "epoch": 335, "lr": 9.713167137407751e-05}
{"train_loss": 0.3825397491455078, "global_step": 29855, "epoch": 335, "lr": 9.7131477837654e-05}
{"train_loss": 0.3033900856971741, "global_step": 29856, "epoch": 335, "lr": 9.713128429489421e-05}
{"train_loss": 0.40172073245048523, "global_step": 29857, "epoch": 335, "lr": 9.713109074579822e-05}
{"train_loss": 0.4329131245613098, "global_step": 29858, "epoch": 335, "lr": 9.713089719036601e-05}
{"train_loss": 0.33257827162742615, "global_step": 29859, "epoch": 335, "lr": 9.713070362859765e-05}
{"train_loss": 0.44096699357032776, "global_step": 29860, "epoch": 335, "lr": 9.713051006049312e-05}
{"train_loss": 0.4080326557159424, "global_step": 29861, "epoch": 335, "lr": 9.713031648605247e-05}
{"train_loss": 0.42341065406799316, "global_step": 29862, "epoch": 335, "lr": 9.713012290527572e-05}
{"train_loss": 0.3265077769756317, "global_step": 29863, "epoch": 335, "lr": 9.712992931816291e-05}
{"train_loss": 0.34487679600715637, "global_step": 29864, "epoch": 335, "lr": 9.712973572471405e-05}
{"train_loss": 0.4425438642501831, "global_step": 29865, "epoch": 335, "lr": 9.712954212492917e-05}
{"train_loss": 0.34481582045555115, "global_step": 29866, "epoch": 335, "lr": 9.712934851880831e-05}
{"train_loss": 0.385690301656723, "global_step": 29867, "epoch": 335, "lr": 9.712915490635146e-05}
{"train_loss": 0.3678593039512634, "global_step": 29868, "epoch": 335, "lr": 9.712896128755868e-05}
{"train_loss": 0.3804674446582794, "global_step": 29869, "epoch": 335, "lr": 9.712876766243e-05}
{"train_loss": 0.5055666565895081, "global_step": 29870, "epoch": 335, "lr": 9.712857403096542e-05}
{"train_loss": 0.4108477234840393, "global_step": 29871, "epoch": 335, "lr": 9.712838039316496e-05}
{"train_loss": 0.4227869510650635, "global_step": 29872, "epoch": 335, "lr": 9.712818674902869e-05}
{"train_loss": 0.2787953317165375, "global_step": 29873, "epoch": 335, "lr": 9.71279930985566e-05}
{"train_loss": 0.31695830821990967, "global_step": 29874, "epoch": 335, "lr": 9.712779944174873e-05}
{"train_loss": 0.4785621166229248, "global_step": 29875, "epoch": 335, "lr": 9.712760577860509e-05}
{"train_loss": 0.5121059417724609, "global_step": 29876, "epoch": 335, "lr": 9.712741210912572e-05}
{"train_loss": 0.45358723402023315, "global_step": 29877, "epoch": 335, "lr": 9.712721843331066e-05}
{"train_loss": 0.36108556389808655, "global_step": 29878, "epoch": 335, "lr": 9.712702475115991e-05}
{"train_loss": 0.5290014743804932, "global_step": 29879, "epoch": 335, "lr": 9.71268310626735e-05}
{"train_loss": 0.4344692528247833, "global_step": 29880, "epoch": 335, "lr": 9.712663736785148e-05}
{"train_loss": 0.4854010343551636, "global_step": 29881, "epoch": 335, "lr": 9.712644366669383e-05}
{"train_loss": 0.44581395387649536, "global_step": 29882, "epoch": 335, "lr": 9.712624995920063e-05}
{"train_loss": 0.4618642330169678, "global_step": 29883, "epoch": 335, "lr": 9.712605624537188e-05}
{"train_loss": 0.4232715964317322, "global_step": 29884, "epoch": 335, "lr": 9.712586252520759e-05}
{"train_loss": 0.42359471321105957, "global_step": 29885, "epoch": 335, "lr": 9.71256687987078e-05}
{"train_loss": 0.5202600955963135, "global_step": 29886, "epoch": 335, "lr": 9.712547506587255e-05}
{"train_loss": 0.44318902492523193, "global_step": 29887, "epoch": 335, "lr": 9.712528132670185e-05}
{"train_loss": 0.44297224283218384, "global_step": 29888, "epoch": 335, "lr": 9.712508758119573e-05}
{"train_loss": 0.3564644753932953, "global_step": 29889, "epoch": 335, "lr": 9.712489382935421e-05}
{"train_loss": 0.41227245330810547, "global_step": 29890, "epoch": 335, "lr": 9.712470007117733e-05}
{"train_loss": 0.4273695647716522, "global_step": 29891, "epoch": 335, "lr": 9.712450630666512e-05}
{"train_loss": 0.38695961236953735, "global_step": 29892, "epoch": 335, "lr": 9.712431253581758e-05}
{"train_loss": 0.3620976507663727, "global_step": 29893, "epoch": 335, "lr": 9.712411875863475e-05}
{"train_loss": 0.4329272210597992, "global_step": 29894, "epoch": 335, "lr": 9.712392497511667e-05}
{"train_loss": 0.29053860902786255, "global_step": 29895, "epoch": 335, "lr": 9.712373118526334e-05}
{"train_loss": 0.42106595635414124, "global_step": 29896, "epoch": 335, "lr": 9.71235373890748e-05}
{"train_loss": 0.45266616344451904, "global_step": 29897, "epoch": 335, "lr": 9.712334358655107e-05}
{"train_loss": 0.3673878014087677, "global_step": 29898, "epoch": 335, "lr": 9.712314977769219e-05}
{"train_loss": 0.3765913248062134, "global_step": 29899, "epoch": 335, "lr": 9.712295596249816e-05}
{"train_loss": 0.45114362239837646, "global_step": 29900, "epoch": 335, "lr": 9.712276214096905e-05}
{"train_loss": 0.27600041031837463, "global_step": 29901, "epoch": 335, "lr": 9.712256831310485e-05}
{"train_loss": 0.44637978076934814, "global_step": 29902, "epoch": 335, "lr": 9.712237447890558e-05}
{"train_loss": 0.39126449388064694, "global_step": 29903, "epoch": 335, "lr": 9.712218063837128e-05, "val_loss": 2.203429937362671, "train_action_mse_error": 28.589065551757812}
{"train_loss": 0.40012267231941223, "global_step": 29904, "epoch": 336, "lr": 9.7121986791502e-05}
{"train_loss": 0.34905529022216797, "global_step": 29905, "epoch": 336, "lr": 9.712179293829773e-05}
{"train_loss": 0.40029168128967285, "global_step": 29906, "epoch": 336, "lr": 9.71215990787585e-05}
{"train_loss": 0.3965165317058563, "global_step": 29907, "epoch": 336, "lr": 9.712140521288437e-05}
{"train_loss": 0.35603076219558716, "global_step": 29908, "epoch": 336, "lr": 9.712121134067533e-05}
{"train_loss": 0.4029634892940521, "global_step": 29909, "epoch": 336, "lr": 9.712101746213141e-05}
{"train_loss": 0.45202985405921936, "global_step": 29910, "epoch": 336, "lr": 9.712082357725265e-05}
{"train_loss": 0.35257744789123535, "global_step": 29911, "epoch": 336, "lr": 9.712062968603907e-05}
{"train_loss": 0.3805273473262787, "global_step": 29912, "epoch": 336, "lr": 9.71204357884907e-05}
{"train_loss": 0.3823568522930145, "global_step": 29913, "epoch": 336, "lr": 9.712024188460755e-05}
{"train_loss": 0.336719274520874, "global_step": 29914, "epoch": 336, "lr": 9.712004797438966e-05}
{"train_loss": 0.2780150771141052, "global_step": 29915, "epoch": 336, "lr": 9.711985405783706e-05}
{"train_loss": 0.44013166427612305, "global_step": 29916, "epoch": 336, "lr": 9.711966013494977e-05}
{"train_loss": 0.3400311768054962, "global_step": 29917, "epoch": 336, "lr": 9.711946620572781e-05}
{"train_loss": 0.39460572600364685, "global_step": 29918, "epoch": 336, "lr": 9.711927227017121e-05}
{"train_loss": 0.42755064368247986, "global_step": 29919, "epoch": 336, "lr": 9.711907832828001e-05}
{"train_loss": 0.2791259288787842, "global_step": 29920, "epoch": 336, "lr": 9.711888438005422e-05}
{"train_loss": 0.31652694940567017, "global_step": 29921, "epoch": 336, "lr": 9.711869042549386e-05}
{"train_loss": 0.2841174900531769, "global_step": 29922, "epoch": 336, "lr": 9.711849646459898e-05}
{"train_loss": 0.3889619708061218, "global_step": 29923, "epoch": 336, "lr": 9.71183024973696e-05}
{"train_loss": 0.43271562457084656, "global_step": 29924, "epoch": 336, "lr": 9.711810852380572e-05}
{"train_loss": 0.29641127586364746, "global_step": 29925, "epoch": 336, "lr": 9.711791454390739e-05}
{"train_loss": 0.329240620136261, "global_step": 29926, "epoch": 336, "lr": 9.711772055767463e-05}
{"train_loss": 0.42653870582580566, "global_step": 29927, "epoch": 336, "lr": 9.711752656510746e-05}
{"train_loss": 0.289583683013916, "global_step": 29928, "epoch": 336, "lr": 9.711733256620592e-05}
{"train_loss": 0.4470537602901459, "global_step": 29929, "epoch": 336, "lr": 9.711713856097003e-05}
{"train_loss": 0.48038455843925476, "global_step": 29930, "epoch": 336, "lr": 9.711694454939982e-05}
{"train_loss": 0.40361854434013367, "global_step": 29931, "epoch": 336, "lr": 9.71167505314953e-05}
{"train_loss": 0.404058575630188, "global_step": 29932, "epoch": 336, "lr": 9.71165565072565e-05}
{"train_loss": 0.38399314880371094, "global_step": 29933, "epoch": 336, "lr": 9.711636247668347e-05}
{"train_loss": 0.5029387474060059, "global_step": 29934, "epoch": 336, "lr": 9.711616843977623e-05}
{"train_loss": 0.3450491726398468, "global_step": 29935, "epoch": 336, "lr": 9.711597439653477e-05}
{"train_loss": 0.25200533866882324, "global_step": 29936, "epoch": 336, "lr": 9.711578034695916e-05}
{"train_loss": 0.40975597500801086, "global_step": 29937, "epoch": 336, "lr": 9.71155862910494e-05}
{"train_loss": 0.35572192072868347, "global_step": 29938, "epoch": 336, "lr": 9.711539222880553e-05}
{"train_loss": 0.3761868178844452, "global_step": 29939, "epoch": 336, "lr": 9.711519816022755e-05}
{"train_loss": 0.3346313238143921, "global_step": 29940, "epoch": 336, "lr": 9.711500408531552e-05}
{"train_loss": 0.41865429282188416, "global_step": 29941, "epoch": 336, "lr": 9.711481000406945e-05}
{"train_loss": 0.33189770579338074, "global_step": 29942, "epoch": 336, "lr": 9.711461591648937e-05}
{"train_loss": 0.4280717372894287, "global_step": 29943, "epoch": 336, "lr": 9.71144218225753e-05}
{"train_loss": 0.42455071210861206, "global_step": 29944, "epoch": 336, "lr": 9.711422772232727e-05}
{"train_loss": 0.3853591978549957, "global_step": 29945, "epoch": 336, "lr": 9.711403361574531e-05}
{"train_loss": 0.40021756291389465, "global_step": 29946, "epoch": 336, "lr": 9.711383950282944e-05}
{"train_loss": 0.38650083541870117, "global_step": 29947, "epoch": 336, "lr": 9.71136453835797e-05}
{"train_loss": 0.5283889770507812, "global_step": 29948, "epoch": 336, "lr": 9.711345125799609e-05}
{"train_loss": 0.32430389523506165, "global_step": 29949, "epoch": 336, "lr": 9.711325712607867e-05}
{"train_loss": 0.39884480834007263, "global_step": 29950, "epoch": 336, "lr": 9.711306298782743e-05}
{"train_loss": 0.3103698194026947, "global_step": 29951, "epoch": 336, "lr": 9.711286884324242e-05}
{"train_loss": 0.4490647315979004, "global_step": 29952, "epoch": 336, "lr": 9.711267469232363e-05}
{"train_loss": 0.4470422565937042, "global_step": 29953, "epoch": 336, "lr": 9.711248053507115e-05}
{"train_loss": 0.43583470582962036, "global_step": 29954, "epoch": 336, "lr": 9.711228637148496e-05}
{"train_loss": 0.4224761426448822, "global_step": 29955, "epoch": 336, "lr": 9.71120922015651e-05}
{"train_loss": 0.33987027406692505, "global_step": 29956, "epoch": 336, "lr": 9.71118980253116e-05}
{"train_loss": 0.4789358079433441, "global_step": 29957, "epoch": 336, "lr": 9.711170384272446e-05}
{"train_loss": 0.36700916290283203, "global_step": 29958, "epoch": 336, "lr": 9.711150965380374e-05}
{"train_loss": 0.46467024087905884, "global_step": 29959, "epoch": 336, "lr": 9.711131545854946e-05}
{"train_loss": 0.29362019896507263, "global_step": 29960, "epoch": 336, "lr": 9.711112125696162e-05}
{"train_loss": 0.30970871448516846, "global_step": 29961, "epoch": 336, "lr": 9.711092704904027e-05}
{"train_loss": 0.40814533829689026, "global_step": 29962, "epoch": 336, "lr": 9.711073283478542e-05}
{"train_loss": 0.3920629322528839, "global_step": 29963, "epoch": 336, "lr": 9.711053861419712e-05}
{"train_loss": 0.33111101388931274, "global_step": 29964, "epoch": 336, "lr": 9.711034438727538e-05}
{"train_loss": 0.3529624938964844, "global_step": 29965, "epoch": 336, "lr": 9.711015015402022e-05}
{"train_loss": 0.5506494045257568, "global_step": 29966, "epoch": 336, "lr": 9.710995591443168e-05}
{"train_loss": 0.48034366965293884, "global_step": 29967, "epoch": 336, "lr": 9.710976166850979e-05}
{"train_loss": 0.49819037318229675, "global_step": 29968, "epoch": 336, "lr": 9.710956741625456e-05}
{"train_loss": 0.40560510754585266, "global_step": 29969, "epoch": 336, "lr": 9.710937315766602e-05}
{"train_loss": 0.34926125407218933, "global_step": 29970, "epoch": 336, "lr": 9.71091788927442e-05}
{"train_loss": 0.4358280897140503, "global_step": 29971, "epoch": 336, "lr": 9.710898462148912e-05}
{"train_loss": 0.4661110043525696, "global_step": 29972, "epoch": 336, "lr": 9.710879034390082e-05}
{"train_loss": 0.3852234184741974, "global_step": 29973, "epoch": 336, "lr": 9.710859605997933e-05}
{"train_loss": 0.3650873303413391, "global_step": 29974, "epoch": 336, "lr": 9.710840176972465e-05}
{"train_loss": 0.2900216281414032, "global_step": 29975, "epoch": 336, "lr": 9.710820747313681e-05}
{"train_loss": 0.3895651400089264, "global_step": 29976, "epoch": 336, "lr": 9.710801317021585e-05}
{"train_loss": 0.42081740498542786, "global_step": 29977, "epoch": 336, "lr": 9.71078188609618e-05}
{"train_loss": 0.37835434079170227, "global_step": 29978, "epoch": 336, "lr": 9.710762454537466e-05}
{"train_loss": 0.30523595213890076, "global_step": 29979, "epoch": 336, "lr": 9.71074302234545e-05}
{"train_loss": 0.41423287987709045, "global_step": 29980, "epoch": 336, "lr": 9.71072358952013e-05}
{"train_loss": 0.39399945735931396, "global_step": 29981, "epoch": 336, "lr": 9.710704156061511e-05}
{"train_loss": 0.3473414182662964, "global_step": 29982, "epoch": 336, "lr": 9.710684721969596e-05}
{"train_loss": 0.3758733570575714, "global_step": 29983, "epoch": 336, "lr": 9.710665287244387e-05}
{"train_loss": 0.2854655086994171, "global_step": 29984, "epoch": 336, "lr": 9.710645851885884e-05}
{"train_loss": 0.428390771150589, "global_step": 29985, "epoch": 336, "lr": 9.710626415894095e-05}
{"train_loss": 0.3936384320259094, "global_step": 29986, "epoch": 336, "lr": 9.71060697926902e-05}
{"train_loss": 0.33678361773490906, "global_step": 29987, "epoch": 336, "lr": 9.710587542010659e-05}
{"train_loss": 0.3491158187389374, "global_step": 29988, "epoch": 336, "lr": 9.710568104119017e-05}
{"train_loss": 0.42081260681152344, "global_step": 29989, "epoch": 336, "lr": 9.710548665594098e-05}
{"train_loss": 0.3684050440788269, "global_step": 29990, "epoch": 336, "lr": 9.710529226435902e-05}
{"train_loss": 0.398501992225647, "global_step": 29991, "epoch": 336, "lr": 9.710509786644435e-05}
{"train_loss": 0.38537392924340924, "global_step": 29992, "epoch": 336, "lr": 9.710490346219695e-05, "val_loss": 2.205547332763672}
{"train_loss": 0.3686167597770691, "global_step": 29993, "epoch": 337, "lr": 9.710470905161688e-05}
{"train_loss": 0.34604522585868835, "global_step": 29994, "epoch": 337, "lr": 9.710451463470415e-05}
{"train_loss": 0.39147505164146423, "global_step": 29995, "epoch": 337, "lr": 9.710432021145879e-05}
{"train_loss": 0.35510796308517456, "global_step": 29996, "epoch": 337, "lr": 9.710412578188084e-05}
{"train_loss": 0.3059920370578766, "global_step": 29997, "epoch": 337, "lr": 9.710393134597031e-05}
{"train_loss": 0.3779640793800354, "global_step": 29998, "epoch": 337, "lr": 9.710373690372723e-05}
{"train_loss": 0.3287108838558197, "global_step": 29999, "epoch": 337, "lr": 9.710354245515162e-05}
{"train_loss": 0.3735312223434448, "global_step": 30000, "epoch": 337, "lr": 9.710334800024353e-05}
{"train_loss": 0.30539020895957947, "global_step": 30001, "epoch": 337, "lr": 9.710315353900295e-05}
{"train_loss": 0.2926589250564575, "global_step": 30002, "epoch": 337, "lr": 9.710295907142992e-05}
{"train_loss": 0.33675283193588257, "global_step": 30003, "epoch": 337, "lr": 9.710276459752449e-05}
{"train_loss": 0.2948804795742035, "global_step": 30004, "epoch": 337, "lr": 9.710257011728665e-05}
{"train_loss": 0.3456408381462097, "global_step": 30005, "epoch": 337, "lr": 9.710237563071647e-05}
{"train_loss": 0.3781110942363739, "global_step": 30006, "epoch": 337, "lr": 9.710218113781393e-05}
{"train_loss": 0.3357367217540741, "global_step": 30007, "epoch": 337, "lr": 9.710198663857908e-05}
{"train_loss": 0.3378024399280548, "global_step": 30008, "epoch": 337, "lr": 9.710179213301194e-05}
{"train_loss": 0.3294902741909027, "global_step": 30009, "epoch": 337, "lr": 9.710159762111253e-05}
{"train_loss": 0.3452758193016052, "global_step": 30010, "epoch": 337, "lr": 9.71014031028809e-05}
{"train_loss": 0.37057429552078247, "global_step": 30011, "epoch": 337, "lr": 9.710120857831706e-05}
{"train_loss": 0.36036989092826843, "global_step": 30012, "epoch": 337, "lr": 9.710101404742103e-05}
{"train_loss": 0.38337546586990356, "global_step": 30013, "epoch": 337, "lr": 9.710081951019284e-05}
{"train_loss": 0.3865068554878235, "global_step": 30014, "epoch": 337, "lr": 9.710062496663252e-05}
{"train_loss": 0.3411124050617218, "global_step": 30015, "epoch": 337, "lr": 9.71004304167401e-05}
{"train_loss": 0.4436843991279602, "global_step": 30016, "epoch": 337, "lr": 9.71002358605156e-05}
{"train_loss": 0.26444461941719055, "global_step": 30017, "epoch": 337, "lr": 9.710004129795905e-05}
{"train_loss": 0.3563198745250702, "global_step": 30018, "epoch": 337, "lr": 9.709984672907047e-05}
{"train_loss": 0.36437875032424927, "global_step": 30019, "epoch": 337, "lr": 9.709965215384988e-05}
{"train_loss": 0.403719425201416, "global_step": 30020, "epoch": 337, "lr": 9.709945757229733e-05}
{"train_loss": 0.2731323540210724, "global_step": 30021, "epoch": 337, "lr": 9.709926298441282e-05}
{"train_loss": 0.3949594497680664, "global_step": 30022, "epoch": 337, "lr": 9.70990683901964e-05}
{"train_loss": 0.4506252706050873, "global_step": 30023, "epoch": 337, "lr": 9.709887378964808e-05}
{"train_loss": 0.26096048951148987, "global_step": 30024, "epoch": 337, "lr": 9.70986791827679e-05}
{"train_loss": 0.3158145248889923, "global_step": 30025, "epoch": 337, "lr": 9.709848456955587e-05}
{"train_loss": 0.37081706523895264, "global_step": 30026, "epoch": 337, "lr": 9.7098289950012e-05}
{"train_loss": 0.4131920039653778, "global_step": 30027, "epoch": 337, "lr": 9.709809532413636e-05}
{"train_loss": 0.3953331708908081, "global_step": 30028, "epoch": 337, "lr": 9.709790069192897e-05}
{"train_loss": 0.3416936695575714, "global_step": 30029, "epoch": 337, "lr": 9.709770605338981e-05}
{"train_loss": 0.30872178077697754, "global_step": 30030, "epoch": 337, "lr": 9.709751140851895e-05}
{"train_loss": 0.33004680275917053, "global_step": 30031, "epoch": 337, "lr": 9.709731675731642e-05}
{"train_loss": 0.3939264416694641, "global_step": 30032, "epoch": 337, "lr": 9.709712209978221e-05}
{"train_loss": 0.40356379747390747, "global_step": 30033, "epoch": 337, "lr": 9.709692743591637e-05}
{"train_loss": 0.29340964555740356, "global_step": 30034, "epoch": 337, "lr": 9.709673276571892e-05}
{"train_loss": 0.4479295015335083, "global_step": 30035, "epoch": 337, "lr": 9.70965380891899e-05}
{"train_loss": 0.4357471466064453, "global_step": 30036, "epoch": 337, "lr": 9.709634340632931e-05}
{"train_loss": 0.2449181228876114, "global_step": 30037, "epoch": 337, "lr": 9.70961487171372e-05}
{"train_loss": 0.35177135467529297, "global_step": 30038, "epoch": 337, "lr": 9.709595402161358e-05}
{"train_loss": 0.33075276017189026, "global_step": 30039, "epoch": 337, "lr": 9.70957593197585e-05}
{"train_loss": 0.41462069749832153, "global_step": 30040, "epoch": 337, "lr": 9.709556461157194e-05}
{"train_loss": 0.3772556185722351, "global_step": 30041, "epoch": 337, "lr": 9.709536989705399e-05}
{"train_loss": 0.30365920066833496, "global_step": 30042, "epoch": 337, "lr": 9.709517517620461e-05}
{"train_loss": 0.3704923093318939, "global_step": 30043, "epoch": 337, "lr": 9.709498044902388e-05}
{"train_loss": 0.27859750390052795, "global_step": 30044, "epoch": 337, "lr": 9.709478571551178e-05}
{"train_loss": 0.29276201128959656, "global_step": 30045, "epoch": 337, "lr": 9.709459097566839e-05}
{"train_loss": 0.40184298157691956, "global_step": 30046, "epoch": 337, "lr": 9.709439622949369e-05}
{"train_loss": 0.3288469612598419, "global_step": 30047, "epoch": 337, "lr": 9.709420147698772e-05}
{"train_loss": 0.3393798768520355, "global_step": 30048, "epoch": 337, "lr": 9.70940067181505e-05}
{"train_loss": 0.34686172008514404, "global_step": 30049, "epoch": 337, "lr": 9.709381195298208e-05}
{"train_loss": 0.26179301738739014, "global_step": 30050, "epoch": 337, "lr": 9.709361718148248e-05}
{"train_loss": 0.3784969747066498, "global_step": 30051, "epoch": 337, "lr": 9.70934224036517e-05}
{"train_loss": 0.35177484154701233, "global_step": 30052, "epoch": 337, "lr": 9.709322761948979e-05}
{"train_loss": 0.39731910824775696, "global_step": 30053, "epoch": 337, "lr": 9.709303282899678e-05}
{"train_loss": 0.2737358510494232, "global_step": 30054, "epoch": 337, "lr": 9.709283803217267e-05}
{"train_loss": 0.3464069962501526, "global_step": 30055, "epoch": 337, "lr": 9.70926432290175e-05}
{"train_loss": 0.3235015869140625, "global_step": 30056, "epoch": 337, "lr": 9.70924484195313e-05}
{"train_loss": 0.3295383155345917, "global_step": 30057, "epoch": 337, "lr": 9.70922536037141e-05}
{"train_loss": 0.34452950954437256, "global_step": 30058, "epoch": 337, "lr": 9.709205878156592e-05}
{"train_loss": 0.3799845576286316, "global_step": 30059, "epoch": 337, "lr": 9.709186395308677e-05}
{"train_loss": 0.3413519263267517, "global_step": 30060, "epoch": 337, "lr": 9.709166911827672e-05}
{"train_loss": 0.31369686126708984, "global_step": 30061, "epoch": 337, "lr": 9.709147427713576e-05}
{"train_loss": 0.48529940843582153, "global_step": 30062, "epoch": 337, "lr": 9.709127942966392e-05}
{"train_loss": 0.27207496762275696, "global_step": 30063, "epoch": 337, "lr": 9.709108457586123e-05}
{"train_loss": 0.380909264087677, "global_step": 30064, "epoch": 337, "lr": 9.709088971572774e-05}
{"train_loss": 0.36237844824790955, "global_step": 30065, "epoch": 337, "lr": 9.709069484926343e-05}
{"train_loss": 0.4394710659980774, "global_step": 30066, "epoch": 337, "lr": 9.709049997646836e-05}
{"train_loss": 0.40405088663101196, "global_step": 30067, "epoch": 337, "lr": 9.709030509734254e-05}
{"train_loss": 0.45522788166999817, "global_step": 30068, "epoch": 337, "lr": 9.709011021188598e-05}
{"train_loss": 0.32281693816185, "global_step": 30069, "epoch": 337, "lr": 9.708991532009877e-05}
{"train_loss": 0.46369102597236633, "global_step": 30070, "epoch": 337, "lr": 9.708972042198088e-05}
{"train_loss": 0.3547857999801636, "global_step": 30071, "epoch": 337, "lr": 9.708952551753234e-05}
{"train_loss": 0.3785060942173004, "global_step": 30072, "epoch": 337, "lr": 9.70893306067532e-05}
{"train_loss": 0.3688361942768097, "global_step": 30073, "epoch": 337, "lr": 9.708913568964346e-05}
{"train_loss": 0.3891541659832001, "global_step": 30074, "epoch": 337, "lr": 9.708894076620318e-05}
{"train_loss": 0.3162165880203247, "global_step": 30075, "epoch": 337, "lr": 9.708874583643234e-05}
{"train_loss": 0.3015870451927185, "global_step": 30076, "epoch": 337, "lr": 9.708855090033103e-05}
{"train_loss": 0.30932411551475525, "global_step": 30077, "epoch": 337, "lr": 9.70883559578992e-05}
{"train_loss": 0.44860467314720154, "global_step": 30078, "epoch": 337, "lr": 9.708816100913693e-05}
{"train_loss": 0.44132906198501587, "global_step": 30079, "epoch": 337, "lr": 9.708796605404423e-05}
{"train_loss": 0.42629101872444153, "global_step": 30080, "epoch": 337, "lr": 9.708777109262112e-05}
{"train_loss": 0.35873743272229525, "global_step": 30081, "epoch": 337, "lr": 9.708757612486765e-05, "val_loss": 2.2391231060028076}
{"train_loss": 0.3052011728286743, "global_step": 30082, "epoch": 338, "lr": 9.708738115078381e-05}
{"train_loss": 0.2831384837627411, "global_step": 30083, "epoch": 338, "lr": 9.708718617036968e-05}
{"train_loss": 0.3946576714515686, "global_step": 30084, "epoch": 338, "lr": 9.70869911836252e-05}
{"train_loss": 0.3876734972000122, "global_step": 30085, "epoch": 338, "lr": 9.708679619055047e-05}
{"train_loss": 0.3959420323371887, "global_step": 30086, "epoch": 338, "lr": 9.708660119114552e-05}
{"train_loss": 0.26636597514152527, "global_step": 30087, "epoch": 338, "lr": 9.708640618541031e-05}
{"train_loss": 0.3523437976837158, "global_step": 30088, "epoch": 338, "lr": 9.708621117334494e-05}
{"train_loss": 0.33058518171310425, "global_step": 30089, "epoch": 338, "lr": 9.708601615494939e-05}
{"train_loss": 0.3707863986492157, "global_step": 30090, "epoch": 338, "lr": 9.70858211302237e-05}
{"train_loss": 0.4666653573513031, "global_step": 30091, "epoch": 338, "lr": 9.70856260991679e-05}
{"train_loss": 0.3928879201412201, "global_step": 30092, "epoch": 338, "lr": 9.7085431061782e-05}
{"train_loss": 0.44699278473854065, "global_step": 30093, "epoch": 338, "lr": 9.708523601806602e-05}
{"train_loss": 0.466879665851593, "global_step": 30094, "epoch": 338, "lr": 9.708504096802004e-05}
{"train_loss": 0.43896180391311646, "global_step": 30095, "epoch": 338, "lr": 9.708484591164404e-05}
{"train_loss": 0.32112160325050354, "global_step": 30096, "epoch": 338, "lr": 9.708465084893805e-05}
{"train_loss": 0.425855815410614, "global_step": 30097, "epoch": 338, "lr": 9.708445577990211e-05}
{"train_loss": 0.38148555159568787, "global_step": 30098, "epoch": 338, "lr": 9.708426070453623e-05}
{"train_loss": 0.3400789797306061, "global_step": 30099, "epoch": 338, "lr": 9.708406562284045e-05}
{"train_loss": 0.33417895436286926, "global_step": 30100, "epoch": 338, "lr": 9.708387053481479e-05}
{"train_loss": 0.2919924557209015, "global_step": 30101, "epoch": 338, "lr": 9.708367544045928e-05}
{"train_loss": 0.6018794775009155, "global_step": 30102, "epoch": 338, "lr": 9.708348033977393e-05}
{"train_loss": 0.4092944860458374, "global_step": 30103, "epoch": 338, "lr": 9.70832852327588e-05}
{"train_loss": 0.497528076171875, "global_step": 30104, "epoch": 338, "lr": 9.708309011941388e-05}
{"train_loss": 0.5349696278572083, "global_step": 30105, "epoch": 338, "lr": 9.708289499973921e-05}
{"train_loss": 0.4123256504535675, "global_step": 30106, "epoch": 338, "lr": 9.708269987373484e-05}
{"train_loss": 0.294086217880249, "global_step": 30107, "epoch": 338, "lr": 9.708250474140076e-05}
{"train_loss": 0.5498974919319153, "global_step": 30108, "epoch": 338, "lr": 9.708230960273702e-05}
{"train_loss": 0.33304011821746826, "global_step": 30109, "epoch": 338, "lr": 9.708211445774362e-05}
{"train_loss": 0.5509541630744934, "global_step": 30110, "epoch": 338, "lr": 9.708191930642063e-05}
{"train_loss": 0.4227650463581085, "global_step": 30111, "epoch": 338, "lr": 9.708172414876802e-05}
{"train_loss": 0.5489262938499451, "global_step": 30112, "epoch": 338, "lr": 9.708152898478587e-05}
{"train_loss": 0.3741534650325775, "global_step": 30113, "epoch": 338, "lr": 9.708133381447418e-05}
{"train_loss": 0.4160749912261963, "global_step": 30114, "epoch": 338, "lr": 9.708113863783297e-05}
{"train_loss": 0.4404151439666748, "global_step": 30115, "epoch": 338, "lr": 9.708094345486227e-05}
{"train_loss": 0.4300175905227661, "global_step": 30116, "epoch": 338, "lr": 9.708074826556213e-05}
{"train_loss": 0.3838562071323395, "global_step": 30117, "epoch": 338, "lr": 9.708055306993253e-05}
{"train_loss": 0.3855617046356201, "global_step": 30118, "epoch": 338, "lr": 9.708035786797354e-05}
{"train_loss": 0.31626883149147034, "global_step": 30119, "epoch": 338, "lr": 9.708016265968517e-05}
{"train_loss": 0.4099489748477936, "global_step": 30120, "epoch": 338, "lr": 9.707996744506744e-05}
{"train_loss": 0.3759348690509796, "global_step": 30121, "epoch": 338, "lr": 9.707977222412039e-05}
{"train_loss": 0.4651342034339905, "global_step": 30122, "epoch": 338, "lr": 9.707957699684403e-05}
{"train_loss": 0.3638526499271393, "global_step": 30123, "epoch": 338, "lr": 9.707938176323841e-05}
{"train_loss": 0.3465498685836792, "global_step": 30124, "epoch": 338, "lr": 9.707918652330353e-05}
{"train_loss": 0.47092577815055847, "global_step": 30125, "epoch": 338, "lr": 9.707899127703942e-05}
{"train_loss": 0.47923922538757324, "global_step": 30126, "epoch": 338, "lr": 9.707879602444611e-05}
{"train_loss": 0.43179258704185486, "global_step": 30127, "epoch": 338, "lr": 9.707860076552365e-05}
{"train_loss": 0.4598466157913208, "global_step": 30128, "epoch": 338, "lr": 9.707840550027204e-05}
{"train_loss": 0.3868144154548645, "global_step": 30129, "epoch": 338, "lr": 9.70782102286913e-05}
{"train_loss": 0.450826495885849, "global_step": 30130, "epoch": 338, "lr": 9.707801495078148e-05}
{"train_loss": 0.4953722357749939, "global_step": 30131, "epoch": 338, "lr": 9.707781966654259e-05}
{"train_loss": 0.5275269150733948, "global_step": 30132, "epoch": 338, "lr": 9.707762437597468e-05}
{"train_loss": 0.36722421646118164, "global_step": 30133, "epoch": 338, "lr": 9.707742907907773e-05}
{"train_loss": 0.41577795147895813, "global_step": 30134, "epoch": 338, "lr": 9.70772337758518e-05}
{"train_loss": 0.38154539465904236, "global_step": 30135, "epoch": 338, "lr": 9.707703846629693e-05}
{"train_loss": 0.449521541595459, "global_step": 30136, "epoch": 338, "lr": 9.70768431504131e-05}
{"train_loss": 0.4393194913864136, "global_step": 30137, "epoch": 338, "lr": 9.707664782820036e-05}
{"train_loss": 0.3608262836933136, "global_step": 30138, "epoch": 338, "lr": 9.707645249965876e-05}
{"train_loss": 0.3249947130680084, "global_step": 30139, "epoch": 338, "lr": 9.70762571647883e-05}
{"train_loss": 0.44399046897888184, "global_step": 30140, "epoch": 338, "lr": 9.707606182358899e-05}
{"train_loss": 0.3240371346473694, "global_step": 30141, "epoch": 338, "lr": 9.70758664760609e-05}
{"train_loss": 0.2897346317768097, "global_step": 30142, "epoch": 338, "lr": 9.707567112220402e-05}
{"train_loss": 0.3795050084590912, "global_step": 30143, "epoch": 338, "lr": 9.70754757620184e-05}
{"train_loss": 0.3845934569835663, "global_step": 30144, "epoch": 338, "lr": 9.707528039550406e-05}
{"train_loss": 0.3852161467075348, "global_step": 30145, "epoch": 338, "lr": 9.7075085022661e-05}
{"train_loss": 0.42453280091285706, "global_step": 30146, "epoch": 338, "lr": 9.70748896434893e-05}
{"train_loss": 0.38241878151893616, "global_step": 30147, "epoch": 338, "lr": 9.707469425798893e-05}
{"train_loss": 0.24330957233905792, "global_step": 30148, "epoch": 338, "lr": 9.707449886615994e-05}
{"train_loss": 0.4278049170970917, "global_step": 30149, "epoch": 338, "lr": 9.707430346800237e-05}
{"train_loss": 0.49123600125312805, "global_step": 30150, "epoch": 338, "lr": 9.707410806351624e-05}
{"train_loss": 0.36552292108535767, "global_step": 30151, "epoch": 338, "lr": 9.707391265270156e-05}
{"train_loss": 0.3417729139328003, "global_step": 30152, "epoch": 338, "lr": 9.707371723555836e-05}
{"train_loss": 0.3440169394016266, "global_step": 30153, "epoch": 338, "lr": 9.707352181208667e-05}
{"train_loss": 0.4172629117965698, "global_step": 30154, "epoch": 338, "lr": 9.707332638228654e-05}
{"train_loss": 0.42467641830444336, "global_step": 30155, "epoch": 338, "lr": 9.707313094615795e-05}
{"train_loss": 0.4278348386287689, "global_step": 30156, "epoch": 338, "lr": 9.707293550370098e-05}
{"train_loss": 0.3470332622528076, "global_step": 30157, "epoch": 338, "lr": 9.70727400549156e-05}
{"train_loss": 0.4609611928462982, "global_step": 30158, "epoch": 338, "lr": 9.707254459980188e-05}
{"train_loss": 0.2659689486026764, "global_step": 30159, "epoch": 338, "lr": 9.707234913835983e-05}
{"train_loss": 0.4133894145488739, "global_step": 30160, "epoch": 338, "lr": 9.707215367058946e-05}
{"train_loss": 0.40968772768974304, "global_step": 30161, "epoch": 338, "lr": 9.707195819649083e-05}
{"train_loss": 0.266853928565979, "global_step": 30162, "epoch": 338, "lr": 9.707176271606394e-05}
{"train_loss": 0.30757200717926025, "global_step": 30163, "epoch": 338, "lr": 9.707156722930883e-05}
{"train_loss": 0.2824081778526306, "global_step": 30164, "epoch": 338, "lr": 9.707137173622551e-05}
{"train_loss": 0.3245629370212555, "global_step": 30165, "epoch": 338, "lr": 9.707117623681403e-05}
{"train_loss": 0.3949231505393982, "global_step": 30166, "epoch": 338, "lr": 9.707098073107441e-05}
{"train_loss": 0.36713460087776184, "global_step": 30167, "epoch": 338, "lr": 9.707078521900665e-05}
{"train_loss": 0.41519585251808167, "global_step": 30168, "epoch": 338, "lr": 9.707058970061083e-05}
{"train_loss": 0.42759329080581665, "global_step": 30169, "epoch": 338, "lr": 9.707039417588691e-05}
{"train_loss": 0.39566992759034875, "global_step": 30170, "epoch": 338, "lr": 9.707019864483496e-05, "val_loss": 2.3594276905059814}
{"train_loss": 0.30461761355400085, "global_step": 30171, "epoch": 339, "lr": 9.7070003107455e-05}
{"train_loss": 0.2899317741394043, "global_step": 30172, "epoch": 339, "lr": 9.706980756374704e-05}
{"train_loss": 0.3847520649433136, "global_step": 30173, "epoch": 339, "lr": 9.706961201371113e-05}
{"train_loss": 0.3674037456512451, "global_step": 30174, "epoch": 339, "lr": 9.706941645734729e-05}
{"train_loss": 0.429069459438324, "global_step": 30175, "epoch": 339, "lr": 9.706922089465553e-05}
{"train_loss": 0.41234061121940613, "global_step": 30176, "epoch": 339, "lr": 9.706902532563589e-05}
{"train_loss": 0.3844147324562073, "global_step": 30177, "epoch": 339, "lr": 9.706882975028839e-05}
{"train_loss": 0.3281877040863037, "global_step": 30178, "epoch": 339, "lr": 9.706863416861305e-05}
{"train_loss": 0.26868847012519836, "global_step": 30179, "epoch": 339, "lr": 9.706843858060993e-05}
{"train_loss": 0.33605650067329407, "global_step": 30180, "epoch": 339, "lr": 9.7068242986279e-05}
{"train_loss": 0.45442652702331543, "global_step": 30181, "epoch": 339, "lr": 9.706804738562034e-05}
{"train_loss": 0.45552441477775574, "global_step": 30182, "epoch": 339, "lr": 9.706785177863395e-05}
{"train_loss": 0.33730822801589966, "global_step": 30183, "epoch": 339, "lr": 9.706765616531987e-05}
{"train_loss": 0.3974517285823822, "global_step": 30184, "epoch": 339, "lr": 9.706746054567812e-05}
{"train_loss": 0.33390098810195923, "global_step": 30185, "epoch": 339, "lr": 9.70672649197087e-05}
{"train_loss": 0.4993797838687897, "global_step": 30186, "epoch": 339, "lr": 9.706706928741169e-05}
{"train_loss": 0.24220018088817596, "global_step": 30187, "epoch": 339, "lr": 9.706687364878706e-05}
{"train_loss": 0.3475593328475952, "global_step": 30188, "epoch": 339, "lr": 9.706667800383486e-05}
{"train_loss": 0.35344839096069336, "global_step": 30189, "epoch": 339, "lr": 9.706648235255514e-05}
{"train_loss": 0.2766577899456024, "global_step": 30190, "epoch": 339, "lr": 9.70662866949479e-05}
{"train_loss": 0.37113288044929504, "global_step": 30191, "epoch": 339, "lr": 9.706609103101317e-05}
{"train_loss": 0.349678635597229, "global_step": 30192, "epoch": 339, "lr": 9.706589536075097e-05}
{"train_loss": 0.49857985973358154, "global_step": 30193, "epoch": 339, "lr": 9.706569968416135e-05}
{"train_loss": 0.3949476182460785, "global_step": 30194, "epoch": 339, "lr": 9.70655040012443e-05}
{"train_loss": 0.3439783751964569, "global_step": 30195, "epoch": 339, "lr": 9.706530831199988e-05}
{"train_loss": 0.4351772367954254, "global_step": 30196, "epoch": 339, "lr": 9.70651126164281e-05}
{"train_loss": 0.393552303314209, "global_step": 30197, "epoch": 339, "lr": 9.7064916914529e-05}
{"train_loss": 0.36079955101013184, "global_step": 30198, "epoch": 339, "lr": 9.706472120630257e-05}
{"train_loss": 0.31352582573890686, "global_step": 30199, "epoch": 339, "lr": 9.706452549174887e-05}
{"train_loss": 0.38901469111442566, "global_step": 30200, "epoch": 339, "lr": 9.706432977086794e-05}
{"train_loss": 0.3656104803085327, "global_step": 30201, "epoch": 339, "lr": 9.706413404365976e-05}
{"train_loss": 0.28262802958488464, "global_step": 30202, "epoch": 339, "lr": 9.70639383101244e-05}
{"train_loss": 0.2590027451515198, "global_step": 30203, "epoch": 339, "lr": 9.706374257026185e-05}
{"train_loss": 0.37682461738586426, "global_step": 30204, "epoch": 339, "lr": 9.706354682407216e-05}
{"train_loss": 0.45565032958984375, "global_step": 30205, "epoch": 339, "lr": 9.706335107155535e-05}
{"train_loss": 0.5002361536026001, "global_step": 30206, "epoch": 339, "lr": 9.706315531271144e-05}
{"train_loss": 0.3659764528274536, "global_step": 30207, "epoch": 339, "lr": 9.706295954754048e-05}
{"train_loss": 0.3355955183506012, "global_step": 30208, "epoch": 339, "lr": 9.706276377604247e-05}
{"train_loss": 0.32572194933891296, "global_step": 30209, "epoch": 339, "lr": 9.706256799821743e-05}
{"train_loss": 0.367373526096344, "global_step": 30210, "epoch": 339, "lr": 9.706237221406542e-05}
{"train_loss": 0.36340758204460144, "global_step": 30211, "epoch": 339, "lr": 9.706217642358643e-05}
{"train_loss": 0.3604753613471985, "global_step": 30212, "epoch": 339, "lr": 9.706198062678053e-05}
{"train_loss": 0.283011794090271, "global_step": 30213, "epoch": 339, "lr": 9.706178482364769e-05}
{"train_loss": 0.42775073647499084, "global_step": 30214, "epoch": 339, "lr": 9.7061589014188e-05}
{"train_loss": 0.3663066625595093, "global_step": 30215, "epoch": 339, "lr": 9.706139319840142e-05}
{"train_loss": 0.417745441198349, "global_step": 30216, "epoch": 339, "lr": 9.706119737628803e-05}
{"train_loss": 0.4721992313861847, "global_step": 30217, "epoch": 339, "lr": 9.706100154784783e-05}
{"train_loss": 0.3374919891357422, "global_step": 30218, "epoch": 339, "lr": 9.706080571308084e-05}
{"train_loss": 0.41864508390426636, "global_step": 30219, "epoch": 339, "lr": 9.706060987198714e-05}
{"train_loss": 0.3802870213985443, "global_step": 30220, "epoch": 339, "lr": 9.706041402456667e-05}
{"train_loss": 0.5212454199790955, "global_step": 30221, "epoch": 339, "lr": 9.706021817081951e-05}
{"train_loss": 0.4245336949825287, "global_step": 30222, "epoch": 339, "lr": 9.70600223107457e-05}
{"train_loss": 0.29770147800445557, "global_step": 30223, "epoch": 339, "lr": 9.705982644434522e-05}
{"train_loss": 0.3826557993888855, "global_step": 30224, "epoch": 339, "lr": 9.705963057161812e-05}
{"train_loss": 0.3234902024269104, "global_step": 30225, "epoch": 339, "lr": 9.705943469256444e-05}
{"train_loss": 0.480112761259079, "global_step": 30226, "epoch": 339, "lr": 9.705923880718417e-05}
{"train_loss": 0.4513760507106781, "global_step": 30227, "epoch": 339, "lr": 9.705904291547739e-05}
{"train_loss": 0.35910382866859436, "global_step": 30228, "epoch": 339, "lr": 9.705884701744408e-05}
{"train_loss": 0.3411048650741577, "global_step": 30229, "epoch": 339, "lr": 9.705865111308427e-05}
{"train_loss": 0.315522164106369, "global_step": 30230, "epoch": 339, "lr": 9.7058455202398e-05}
{"train_loss": 0.3619649410247803, "global_step": 30231, "epoch": 339, "lr": 9.705825928538531e-05}
{"train_loss": 0.3780480921268463, "global_step": 30232, "epoch": 339, "lr": 9.705806336204622e-05}
{"train_loss": 0.37140169739723206, "global_step": 30233, "epoch": 339, "lr": 9.705786743238073e-05}
{"train_loss": 0.3984806537628174, "global_step": 30234, "epoch": 339, "lr": 9.705767149638886e-05}
{"train_loss": 0.5135560631752014, "global_step": 30235, "epoch": 339, "lr": 9.70574755540707e-05}
{"train_loss": 0.3721866309642792, "global_step": 30236, "epoch": 339, "lr": 9.70572796054262e-05}
{"train_loss": 0.332634299993515, "global_step": 30237, "epoch": 339, "lr": 9.705708365045545e-05}
{"train_loss": 0.36050257086753845, "global_step": 30238, "epoch": 339, "lr": 9.705688768915844e-05}
{"train_loss": 0.37267228960990906, "global_step": 30239, "epoch": 339, "lr": 9.705669172153521e-05}
{"train_loss": 0.3929404616355896, "global_step": 30240, "epoch": 339, "lr": 9.705649574758578e-05}
{"train_loss": 0.27258986234664917, "global_step": 30241, "epoch": 339, "lr": 9.705629976731016e-05}
{"train_loss": 0.33404552936553955, "global_step": 30242, "epoch": 339, "lr": 9.705610378070842e-05}
{"train_loss": 0.3272841274738312, "global_step": 30243, "epoch": 339, "lr": 9.705590778778056e-05}
{"train_loss": 0.466272234916687, "global_step": 30244, "epoch": 339, "lr": 9.70557117885266e-05}
{"train_loss": 0.28815144300460815, "global_step": 30245, "epoch": 339, "lr": 9.705551578294656e-05}
{"train_loss": 0.4054511487483978, "global_step": 30246, "epoch": 339, "lr": 9.705531977104048e-05}
{"train_loss": 0.581683874130249, "global_step": 30247, "epoch": 339, "lr": 9.70551237528084e-05}
{"train_loss": 0.3059164881706238, "global_step": 30248, "epoch": 339, "lr": 9.705492772825032e-05}
{"train_loss": 0.3374762237071991, "global_step": 30249, "epoch": 339, "lr": 9.705473169736628e-05}
{"train_loss": 0.32463395595550537, "global_step": 30250, "epoch": 339, "lr": 9.705453566015632e-05}
{"train_loss": 0.3842695653438568, "global_step": 30251, "epoch": 339, "lr": 9.705433961662044e-05}
{"train_loss": 0.37977299094200134, "global_step": 30252, "epoch": 339, "lr": 9.705414356675868e-05}
{"train_loss": 0.5113452672958374, "global_step": 30253, "epoch": 339, "lr": 9.705394751057107e-05}
{"train_loss": 0.4298110008239746, "global_step": 30254, "epoch": 339, "lr": 9.705375144805762e-05}
{"train_loss": 0.358305960893631, "global_step": 30255, "epoch": 339, "lr": 9.705355537921837e-05}
{"train_loss": 0.32202795147895813, "global_step": 30256, "epoch": 339, "lr": 9.705335930405333e-05}
{"train_loss": 0.37490445375442505, "global_step": 30257, "epoch": 339, "lr": 9.705316322256255e-05}
{"train_loss": 0.35989952087402344, "global_step": 30258, "epoch": 339, "lr": 9.705296713474606e-05}
{"train_loss": 0.376034852326586, "global_step": 30259, "epoch": 339, "lr": 9.705277104060385e-05, "val_loss": 2.2773196697235107}
{"train_loss": 0.4254731833934784, "global_step": 30260, "epoch": 340, "lr": 9.705257494013598e-05}
{"train_loss": 0.6148950457572937, "global_step": 30261, "epoch": 340, "lr": 9.705237883334245e-05}
{"train_loss": 0.4093310832977295, "global_step": 30262, "epoch": 340, "lr": 9.705218272022332e-05}
{"train_loss": 0.3723566234111786, "global_step": 30263, "epoch": 340, "lr": 9.705198660077859e-05}
{"train_loss": 0.44104328751564026, "global_step": 30264, "epoch": 340, "lr": 9.70517904750083e-05}
{"train_loss": 0.3728732466697693, "global_step": 30265, "epoch": 340, "lr": 9.705159434291247e-05}
{"train_loss": 0.44391554594039917, "global_step": 30266, "epoch": 340, "lr": 9.705139820449111e-05}
{"train_loss": 0.34743452072143555, "global_step": 30267, "epoch": 340, "lr": 9.705120205974428e-05}
{"train_loss": 0.3999500870704651, "global_step": 30268, "epoch": 340, "lr": 9.705100590867198e-05}
{"train_loss": 0.36909717321395874, "global_step": 30269, "epoch": 340, "lr": 9.705080975127425e-05}
{"train_loss": 0.3188979923725128, "global_step": 30270, "epoch": 340, "lr": 9.705061358755112e-05}
{"train_loss": 0.34168189764022827, "global_step": 30271, "epoch": 340, "lr": 9.70504174175026e-05}
{"train_loss": 0.5113385915756226, "global_step": 30272, "epoch": 340, "lr": 9.705022124112873e-05}
{"train_loss": 0.3376704752445221, "global_step": 30273, "epoch": 340, "lr": 9.705002505842952e-05}
{"train_loss": 0.4488461911678314, "global_step": 30274, "epoch": 340, "lr": 9.704982886940501e-05}
{"train_loss": 0.3801388144493103, "global_step": 30275, "epoch": 340, "lr": 9.704963267405522e-05}
{"train_loss": 0.39666813611984253, "global_step": 30276, "epoch": 340, "lr": 9.70494364723802e-05}
{"train_loss": 0.3932480216026306, "global_step": 30277, "epoch": 340, "lr": 9.704924026437994e-05}
{"train_loss": 0.3931373953819275, "global_step": 30278, "epoch": 340, "lr": 9.704904405005448e-05}
{"train_loss": 0.39525482058525085, "global_step": 30279, "epoch": 340, "lr": 9.704884782940386e-05}
{"train_loss": 0.3135271966457367, "global_step": 30280, "epoch": 340, "lr": 9.704865160242811e-05}
{"train_loss": 0.337972491979599, "global_step": 30281, "epoch": 340, "lr": 9.704845536912721e-05}
{"train_loss": 0.39847680926322937, "global_step": 30282, "epoch": 340, "lr": 9.704825912950125e-05}
{"train_loss": 0.40201336145401, "global_step": 30283, "epoch": 340, "lr": 9.70480628835502e-05}
{"train_loss": 0.4016236364841461, "global_step": 30284, "epoch": 340, "lr": 9.704786663127412e-05}
{"train_loss": 0.3111335039138794, "global_step": 30285, "epoch": 340, "lr": 9.704767037267303e-05}
{"train_loss": 0.2832598090171814, "global_step": 30286, "epoch": 340, "lr": 9.704747410774695e-05}
{"train_loss": 0.3662120997905731, "global_step": 30287, "epoch": 340, "lr": 9.704727783649592e-05}
{"train_loss": 0.4435810446739197, "global_step": 30288, "epoch": 340, "lr": 9.704708155891994e-05}
{"train_loss": 0.34509900212287903, "global_step": 30289, "epoch": 340, "lr": 9.704688527501905e-05}
{"train_loss": 0.3357499837875366, "global_step": 30290, "epoch": 340, "lr": 9.704668898479331e-05}
{"train_loss": 0.2903289496898651, "global_step": 30291, "epoch": 340, "lr": 9.704649268824271e-05}
{"train_loss": 0.3237607479095459, "global_step": 30292, "epoch": 340, "lr": 9.704629638536726e-05}
{"train_loss": 0.37851786613464355, "global_step": 30293, "epoch": 340, "lr": 9.704610007616702e-05}
{"train_loss": 0.40493160486221313, "global_step": 30294, "epoch": 340, "lr": 9.704590376064201e-05}
{"train_loss": 0.46388307213783264, "global_step": 30295, "epoch": 340, "lr": 9.704570743879224e-05}
{"train_loss": 0.4122364819049835, "global_step": 30296, "epoch": 340, "lr": 9.704551111061775e-05}
{"train_loss": 0.3351624011993408, "global_step": 30297, "epoch": 340, "lr": 9.704531477611859e-05}
{"train_loss": 0.28184816241264343, "global_step": 30298, "epoch": 340, "lr": 9.704511843529474e-05}
{"train_loss": 0.43144744634628296, "global_step": 30299, "epoch": 340, "lr": 9.704492208814624e-05}
{"train_loss": 0.27096062898635864, "global_step": 30300, "epoch": 340, "lr": 9.704472573467314e-05}
{"train_loss": 0.5333666801452637, "global_step": 30301, "epoch": 340, "lr": 9.704452937487543e-05}
{"train_loss": 0.35810068249702454, "global_step": 30302, "epoch": 340, "lr": 9.704433300875317e-05}
{"train_loss": 0.2756431996822357, "global_step": 30303, "epoch": 340, "lr": 9.704413663630637e-05}
{"train_loss": 0.38934311270713806, "global_step": 30304, "epoch": 340, "lr": 9.704394025753507e-05}
{"train_loss": 0.4491051733493805, "global_step": 30305, "epoch": 340, "lr": 9.704374387243926e-05}
{"train_loss": 0.35401684045791626, "global_step": 30306, "epoch": 340, "lr": 9.704354748101901e-05}
{"train_loss": 0.33797919750213623, "global_step": 30307, "epoch": 340, "lr": 9.704335108327434e-05}
{"train_loss": 0.35884830355644226, "global_step": 30308, "epoch": 340, "lr": 9.704315467920523e-05}
{"train_loss": 0.28257304430007935, "global_step": 30309, "epoch": 340, "lr": 9.704295826881178e-05}
{"train_loss": 0.3766161799430847, "global_step": 30310, "epoch": 340, "lr": 9.704276185209395e-05}
{"train_loss": 0.4185841977596283, "global_step": 30311, "epoch": 340, "lr": 9.704256542905182e-05}
{"train_loss": 0.3808786869049072, "global_step": 30312, "epoch": 340, "lr": 9.704236899968536e-05}
{"train_loss": 0.3769555985927582, "global_step": 30313, "epoch": 340, "lr": 9.704217256399464e-05}
{"train_loss": 0.3971351087093353, "global_step": 30314, "epoch": 340, "lr": 9.704197612197967e-05}
{"train_loss": 0.39089953899383545, "global_step": 30315, "epoch": 340, "lr": 9.70417796736405e-05}
{"train_loss": 0.3041459321975708, "global_step": 30316, "epoch": 340, "lr": 9.704158321897712e-05}
{"train_loss": 0.35730570554733276, "global_step": 30317, "epoch": 340, "lr": 9.704138675798957e-05}
{"train_loss": 0.36110812425613403, "global_step": 30318, "epoch": 340, "lr": 9.704119029067788e-05}
{"train_loss": 0.26406753063201904, "global_step": 30319, "epoch": 340, "lr": 9.704099381704207e-05}
{"train_loss": 0.3565307557582855, "global_step": 30320, "epoch": 340, "lr": 9.704079733708218e-05}
{"train_loss": 0.32912686467170715, "global_step": 30321, "epoch": 340, "lr": 9.704060085079823e-05}
{"train_loss": 0.36501145362854004, "global_step": 30322, "epoch": 340, "lr": 9.704040435819023e-05}
{"train_loss": 0.30926987528800964, "global_step": 30323, "epoch": 340, "lr": 9.704020785925825e-05}
{"train_loss": 0.4883764386177063, "global_step": 30324, "epoch": 340, "lr": 9.704001135400226e-05}
{"train_loss": 0.2839752435684204, "global_step": 30325, "epoch": 340, "lr": 9.703981484242233e-05}
{"train_loss": 0.3582366406917572, "global_step": 30326, "epoch": 340, "lr": 9.703961832451845e-05}
{"train_loss": 0.35401812195777893, "global_step": 30327, "epoch": 340, "lr": 9.703942180029069e-05}
{"train_loss": 0.4042772054672241, "global_step": 30328, "epoch": 340, "lr": 9.703922526973904e-05}
{"train_loss": 0.3906792104244232, "global_step": 30329, "epoch": 340, "lr": 9.703902873286355e-05}
{"train_loss": 0.46543192863464355, "global_step": 30330, "epoch": 340, "lr": 9.703883218966421e-05}
{"train_loss": 0.3687573969364166, "global_step": 30331, "epoch": 340, "lr": 9.703863564014111e-05}
{"train_loss": 0.3383724093437195, "global_step": 30332, "epoch": 340, "lr": 9.703843908429422e-05}
{"train_loss": 0.4417515695095062, "global_step": 30333, "epoch": 340, "lr": 9.703824252212357e-05}
{"train_loss": 0.34807589650154114, "global_step": 30334, "epoch": 340, "lr": 9.703804595362922e-05}
{"train_loss": 0.391319215297699, "global_step": 30335, "epoch": 340, "lr": 9.703784937881117e-05}
{"train_loss": 0.23266808688640594, "global_step": 30336, "epoch": 340, "lr": 9.703765279766948e-05}
{"train_loss": 0.4077666401863098, "global_step": 30337, "epoch": 340, "lr": 9.703745621020412e-05}
{"train_loss": 0.2559439539909363, "global_step": 30338, "epoch": 340, "lr": 9.703725961641516e-05}
{"train_loss": 0.35588306188583374, "global_step": 30339, "epoch": 340, "lr": 9.703706301630261e-05}
{"train_loss": 0.3534334897994995, "global_step": 30340, "epoch": 340, "lr": 9.70368664098665e-05}
{"train_loss": 0.36429181694984436, "global_step": 30341, "epoch": 340, "lr": 9.703666979710685e-05}
{"train_loss": 0.36011093854904175, "global_step": 30342, "epoch": 340, "lr": 9.703647317802371e-05}
{"train_loss": 0.3606248199939728, "global_step": 30343, "epoch": 340, "lr": 9.703627655261708e-05}
{"train_loss": 0.43845081329345703, "global_step": 30344, "epoch": 340, "lr": 9.703607992088699e-05}
{"train_loss": 0.5663671493530273, "global_step": 30345, "epoch": 340, "lr": 9.703588328283348e-05}
{"train_loss": 0.34470370411872864, "global_step": 30346, "epoch": 340, "lr": 9.703568663845656e-05}
{"train_loss": 0.42703327536582947, "global_step": 30347, "epoch": 340, "lr": 9.703548998775628e-05}
{"train_loss": 0.37453815290767156, "global_step": 30348, "epoch": 340, "lr": 9.703529333073265e-05, "val_loss": 2.287579298019409, "train_action_mse_error": 32.22636795043945}
{"train_loss": 0.43662071228027344, "global_step": 30349, "epoch": 341, "lr": 9.703509666738568e-05}
{"train_loss": 0.39571359753608704, "global_step": 30350, "epoch": 341, "lr": 9.703489999771544e-05}
{"train_loss": 0.38123247027397156, "global_step": 30351, "epoch": 341, "lr": 9.703470332172191e-05}
{"train_loss": 0.43772241473197937, "global_step": 30352, "epoch": 341, "lr": 9.703450663940515e-05}
{"train_loss": 0.49120569229125977, "global_step": 30353, "epoch": 341, "lr": 9.703430995076517e-05}
{"train_loss": 0.49862542748451233, "global_step": 30354, "epoch": 341, "lr": 9.7034113255802e-05}
{"train_loss": 0.37703564763069153, "global_step": 30355, "epoch": 341, "lr": 9.703391655451565e-05}
{"train_loss": 0.28909438848495483, "global_step": 30356, "epoch": 341, "lr": 9.70337198469062e-05}
{"train_loss": 0.4397051930427551, "global_step": 30357, "epoch": 341, "lr": 9.703352313297362e-05}
{"train_loss": 0.4200158715248108, "global_step": 30358, "epoch": 341, "lr": 9.703332641271794e-05}
{"train_loss": 0.423378050327301, "global_step": 30359, "epoch": 341, "lr": 9.703312968613922e-05}
{"train_loss": 0.4956513047218323, "global_step": 30360, "epoch": 341, "lr": 9.703293295323748e-05}
{"train_loss": 0.4634239375591278, "global_step": 30361, "epoch": 341, "lr": 9.703273621401272e-05}
{"train_loss": 0.388763427734375, "global_step": 30362, "epoch": 341, "lr": 9.703253946846498e-05}
{"train_loss": 0.34423360228538513, "global_step": 30363, "epoch": 341, "lr": 9.70323427165943e-05}
{"train_loss": 0.27668502926826477, "global_step": 30364, "epoch": 341, "lr": 9.703214595840068e-05}
{"train_loss": 0.3980185091495514, "global_step": 30365, "epoch": 341, "lr": 9.703194919388416e-05}
{"train_loss": 0.3168851435184479, "global_step": 30366, "epoch": 341, "lr": 9.703175242304478e-05}
{"train_loss": 0.28577369451522827, "global_step": 30367, "epoch": 341, "lr": 9.703155564588255e-05}
{"train_loss": 0.34041014313697815, "global_step": 30368, "epoch": 341, "lr": 9.703135886239751e-05}
{"train_loss": 0.34808704257011414, "global_step": 30369, "epoch": 341, "lr": 9.703116207258967e-05}
{"train_loss": 0.32263273000717163, "global_step": 30370, "epoch": 341, "lr": 9.703096527645905e-05}
{"train_loss": 0.36415326595306396, "global_step": 30371, "epoch": 341, "lr": 9.70307684740057e-05}
{"train_loss": 0.3951963782310486, "global_step": 30372, "epoch": 341, "lr": 9.703057166522964e-05}
{"train_loss": 0.3700197637081146, "global_step": 30373, "epoch": 341, "lr": 9.703037485013089e-05}
{"train_loss": 0.35633304715156555, "global_step": 30374, "epoch": 341, "lr": 9.703017802870949e-05}
{"train_loss": 0.25798165798187256, "global_step": 30375, "epoch": 341, "lr": 9.702998120096543e-05}
{"train_loss": 0.3912844955921173, "global_step": 30376, "epoch": 341, "lr": 9.702978436689878e-05}
{"train_loss": 0.30278244614601135, "global_step": 30377, "epoch": 341, "lr": 9.702958752650955e-05}
{"train_loss": 0.3595382869243622, "global_step": 30378, "epoch": 341, "lr": 9.702939067979777e-05}
{"train_loss": 0.3996084928512573, "global_step": 30379, "epoch": 341, "lr": 9.702919382676345e-05}
{"train_loss": 0.40765514969825745, "global_step": 30380, "epoch": 341, "lr": 9.702899696740663e-05}
{"train_loss": 0.3999052047729492, "global_step": 30381, "epoch": 341, "lr": 9.702880010172732e-05}
{"train_loss": 0.44171425700187683, "global_step": 30382, "epoch": 341, "lr": 9.702860322972558e-05}
{"train_loss": 0.37062469124794006, "global_step": 30383, "epoch": 341, "lr": 9.702840635140143e-05}
{"train_loss": 0.47822004556655884, "global_step": 30384, "epoch": 341, "lr": 9.702820946675486e-05}
{"train_loss": 0.4271473288536072, "global_step": 30385, "epoch": 341, "lr": 9.702801257578593e-05}
{"train_loss": 0.3446199893951416, "global_step": 30386, "epoch": 341, "lr": 9.702781567849466e-05}
{"train_loss": 0.37607458233833313, "global_step": 30387, "epoch": 341, "lr": 9.702761877488107e-05}
{"train_loss": 0.4178936183452606, "global_step": 30388, "epoch": 341, "lr": 9.702742186494518e-05}
{"train_loss": 0.4634830355644226, "global_step": 30389, "epoch": 341, "lr": 9.702722494868703e-05}
{"train_loss": 0.35765340924263, "global_step": 30390, "epoch": 341, "lr": 9.702702802610665e-05}
{"train_loss": 0.4909498691558838, "global_step": 30391, "epoch": 341, "lr": 9.702683109720407e-05}
{"train_loss": 0.306231826543808, "global_step": 30392, "epoch": 341, "lr": 9.702663416197928e-05}
{"train_loss": 0.33297473192214966, "global_step": 30393, "epoch": 341, "lr": 9.702643722043235e-05}
{"train_loss": 0.35113534331321716, "global_step": 30394, "epoch": 341, "lr": 9.702624027256328e-05}
{"train_loss": 0.5242944359779358, "global_step": 30395, "epoch": 341, "lr": 9.702604331837211e-05}
{"train_loss": 0.42528578639030457, "global_step": 30396, "epoch": 341, "lr": 9.702584635785886e-05}
{"train_loss": 0.46119171380996704, "global_step": 30397, "epoch": 341, "lr": 9.702564939102357e-05}
{"train_loss": 0.37739458680152893, "global_step": 30398, "epoch": 341, "lr": 9.702545241786623e-05}
{"train_loss": 0.39731159806251526, "global_step": 30399, "epoch": 341, "lr": 9.702525543838691e-05}
{"train_loss": 0.4198726713657379, "global_step": 30400, "epoch": 341, "lr": 9.70250584525856e-05}
{"train_loss": 0.3434460461139679, "global_step": 30401, "epoch": 341, "lr": 9.702486146046237e-05}
{"train_loss": 0.43075472116470337, "global_step": 30402, "epoch": 341, "lr": 9.70246644620172e-05}
{"train_loss": 0.37174558639526367, "global_step": 30403, "epoch": 341, "lr": 9.702446745725014e-05}
{"train_loss": 0.3619808852672577, "global_step": 30404, "epoch": 341, "lr": 9.702427044616122e-05}
{"train_loss": 0.3474757671356201, "global_step": 30405, "epoch": 341, "lr": 9.702407342875046e-05}
{"train_loss": 0.3113778233528137, "global_step": 30406, "epoch": 341, "lr": 9.702387640501789e-05}
{"train_loss": 0.40636783838272095, "global_step": 30407, "epoch": 341, "lr": 9.702367937496352e-05}
{"train_loss": 0.3213314116001129, "global_step": 30408, "epoch": 341, "lr": 9.70234823385874e-05}
{"train_loss": 0.43507397174835205, "global_step": 30409, "epoch": 341, "lr": 9.702328529588953e-05}
{"train_loss": 0.26183560490608215, "global_step": 30410, "epoch": 341, "lr": 9.702308824686997e-05}
{"train_loss": 0.3716937005519867, "global_step": 30411, "epoch": 341, "lr": 9.702289119152872e-05}
{"train_loss": 0.3796037435531616, "global_step": 30412, "epoch": 341, "lr": 9.702269412986582e-05}
{"train_loss": 0.3960684537887573, "global_step": 30413, "epoch": 341, "lr": 9.702249706188129e-05}
{"train_loss": 0.2829030454158783, "global_step": 30414, "epoch": 341, "lr": 9.702229998757517e-05}
{"train_loss": 0.32310888171195984, "global_step": 30415, "epoch": 341, "lr": 9.702210290694747e-05}
{"train_loss": 0.4490409791469574, "global_step": 30416, "epoch": 341, "lr": 9.702190581999821e-05}
{"train_loss": 0.4172147512435913, "global_step": 30417, "epoch": 341, "lr": 9.702170872672742e-05}
{"train_loss": 0.3434508740901947, "global_step": 30418, "epoch": 341, "lr": 9.702151162713515e-05}
{"train_loss": 0.4374268054962158, "global_step": 30419, "epoch": 341, "lr": 9.702131452122141e-05}
{"train_loss": 0.27380308508872986, "global_step": 30420, "epoch": 341, "lr": 9.702111740898622e-05}
{"train_loss": 0.33963480591773987, "global_step": 30421, "epoch": 341, "lr": 9.702092029042963e-05}
{"train_loss": 0.3834781348705292, "global_step": 30422, "epoch": 341, "lr": 9.702072316555164e-05}
{"train_loss": 0.4398271143436432, "global_step": 30423, "epoch": 341, "lr": 9.702052603435228e-05}
{"train_loss": 0.3591654300689697, "global_step": 30424, "epoch": 341, "lr": 9.702032889683159e-05}
{"train_loss": 0.3790692389011383, "global_step": 30425, "epoch": 341, "lr": 9.70201317529896e-05}
{"train_loss": 0.30832579731941223, "global_step": 30426, "epoch": 341, "lr": 9.701993460282631e-05}
{"train_loss": 0.355142742395401, "global_step": 30427, "epoch": 341, "lr": 9.701973744634177e-05}
{"train_loss": 0.35867416858673096, "global_step": 30428, "epoch": 341, "lr": 9.7019540283536e-05}
{"train_loss": 0.5539723634719849, "global_step": 30429, "epoch": 341, "lr": 9.701934311440902e-05}
{"train_loss": 0.4544239938259125, "global_step": 30430, "epoch": 341, "lr": 9.701914593896087e-05}
{"train_loss": 0.33910271525382996, "global_step": 30431, "epoch": 341, "lr": 9.701894875719157e-05}
{"train_loss": 0.35351794958114624, "global_step": 30432, "epoch": 341, "lr": 9.701875156910114e-05}
{"train_loss": 0.43614986538887024, "global_step": 30433, "epoch": 341, "lr": 9.701855437468961e-05}
{"train_loss": 0.35083773732185364, "global_step": 30434, "epoch": 341, "lr": 9.701835717395702e-05}
{"train_loss": 0.3751586973667145, "global_step": 30435, "epoch": 341, "lr": 9.701815996690338e-05}
{"train_loss": 0.34745922684669495, "global_step": 30436, "epoch": 341, "lr": 9.70179627535287e-05}
{"train_loss": 0.3839937448501587, "global_step": 30437, "epoch": 341, "lr": 9.701776553383306e-05, "val_loss": 2.2165474891662598}
{"train_loss": 0.28349781036376953, "global_step": 30438, "epoch": 342, "lr": 9.701756830781643e-05}
{"train_loss": 0.31855112314224243, "global_step": 30439, "epoch": 342, "lr": 9.701737107547888e-05}
{"train_loss": 0.3284876346588135, "global_step": 30440, "epoch": 342, "lr": 9.701717383682041e-05}
{"train_loss": 0.34757962822914124, "global_step": 30441, "epoch": 342, "lr": 9.701697659184104e-05}
{"train_loss": 0.2951279878616333, "global_step": 30442, "epoch": 342, "lr": 9.701677934054083e-05}
{"train_loss": 0.3451630473136902, "global_step": 30443, "epoch": 342, "lr": 9.701658208291978e-05}
{"train_loss": 0.32916003465652466, "global_step": 30444, "epoch": 342, "lr": 9.701638481897792e-05}
{"train_loss": 0.2802414894104004, "global_step": 30445, "epoch": 342, "lr": 9.701618754871528e-05}
{"train_loss": 0.3553805649280548, "global_step": 30446, "epoch": 342, "lr": 9.70159902721319e-05}
{"train_loss": 0.40534573793411255, "global_step": 30447, "epoch": 342, "lr": 9.701579298922778e-05}
{"train_loss": 0.25553396344184875, "global_step": 30448, "epoch": 342, "lr": 9.701559570000298e-05}
{"train_loss": 0.3617597222328186, "global_step": 30449, "epoch": 342, "lr": 9.701539840445747e-05}
{"train_loss": 0.557083010673523, "global_step": 30450, "epoch": 342, "lr": 9.701520110259134e-05}
{"train_loss": 0.3849758803844452, "global_step": 30451, "epoch": 342, "lr": 9.701500379440457e-05}
{"train_loss": 0.3481656610965729, "global_step": 30452, "epoch": 342, "lr": 9.701480647989723e-05}
{"train_loss": 0.2879835069179535, "global_step": 30453, "epoch": 342, "lr": 9.701460915906931e-05}
{"train_loss": 0.427232950925827, "global_step": 30454, "epoch": 342, "lr": 9.701441183192086e-05}
{"train_loss": 0.32135728001594543, "global_step": 30455, "epoch": 342, "lr": 9.701421449845187e-05}
{"train_loss": 0.27803075313568115, "global_step": 30456, "epoch": 342, "lr": 9.70140171586624e-05}
{"train_loss": 0.3104008138179779, "global_step": 30457, "epoch": 342, "lr": 9.701381981255248e-05}
{"train_loss": 0.3394630551338196, "global_step": 30458, "epoch": 342, "lr": 9.701362246012213e-05}
{"train_loss": 0.2519080936908722, "global_step": 30459, "epoch": 342, "lr": 9.701342510137136e-05}
{"train_loss": 0.3044758141040802, "global_step": 30460, "epoch": 342, "lr": 9.701322773630021e-05}
{"train_loss": 0.26848843693733215, "global_step": 30461, "epoch": 342, "lr": 9.70130303649087e-05}
{"train_loss": 0.33006876707077026, "global_step": 30462, "epoch": 342, "lr": 9.701283298719686e-05}
{"train_loss": 0.3356967270374298, "global_step": 30463, "epoch": 342, "lr": 9.701263560316473e-05}
{"train_loss": 0.41457223892211914, "global_step": 30464, "epoch": 342, "lr": 9.701243821281232e-05}
{"train_loss": 0.40012291073799133, "global_step": 30465, "epoch": 342, "lr": 9.701224081613965e-05}
{"train_loss": 0.3380773961544037, "global_step": 30466, "epoch": 342, "lr": 9.701204341314677e-05}
{"train_loss": 0.2060239464044571, "global_step": 30467, "epoch": 342, "lr": 9.701184600383369e-05}
{"train_loss": 0.31541961431503296, "global_step": 30468, "epoch": 342, "lr": 9.701164858820044e-05}
{"train_loss": 0.3778751492500305, "global_step": 30469, "epoch": 342, "lr": 9.701145116624703e-05}
{"train_loss": 0.3387880027294159, "global_step": 30470, "epoch": 342, "lr": 9.701125373797353e-05}
{"train_loss": 0.3818472623825073, "global_step": 30471, "epoch": 342, "lr": 9.701105630337992e-05}
{"train_loss": 0.4513302445411682, "global_step": 30472, "epoch": 342, "lr": 9.701085886246626e-05}
{"train_loss": 0.348754346370697, "global_step": 30473, "epoch": 342, "lr": 9.701066141523257e-05}
{"train_loss": 0.2985785901546478, "global_step": 30474, "epoch": 342, "lr": 9.701046396167886e-05}
{"train_loss": 0.36602166295051575, "global_step": 30475, "epoch": 342, "lr": 9.701026650180516e-05}
{"train_loss": 0.35394901037216187, "global_step": 30476, "epoch": 342, "lr": 9.701006903561149e-05}
{"train_loss": 0.3423948585987091, "global_step": 30477, "epoch": 342, "lr": 9.70098715630979e-05}
{"train_loss": 0.4568077623844147, "global_step": 30478, "epoch": 342, "lr": 9.700967408426442e-05}
{"train_loss": 0.4571326673030853, "global_step": 30479, "epoch": 342, "lr": 9.700947659911105e-05}
{"train_loss": 0.40905794501304626, "global_step": 30480, "epoch": 342, "lr": 9.700927910763785e-05}
{"train_loss": 0.45955702662467957, "global_step": 30481, "epoch": 342, "lr": 9.70090816098448e-05}
{"train_loss": 0.35975968837738037, "global_step": 30482, "epoch": 342, "lr": 9.700888410573196e-05}
{"train_loss": 0.4296535849571228, "global_step": 30483, "epoch": 342, "lr": 9.700868659529934e-05}
{"train_loss": 0.37085801362991333, "global_step": 30484, "epoch": 342, "lr": 9.700848907854698e-05}
{"train_loss": 0.4345844089984894, "global_step": 30485, "epoch": 342, "lr": 9.70082915554749e-05}
{"train_loss": 0.37559014558792114, "global_step": 30486, "epoch": 342, "lr": 9.700809402608315e-05}
{"train_loss": 0.43313688039779663, "global_step": 30487, "epoch": 342, "lr": 9.700789649037171e-05}
{"train_loss": 0.3591234087944031, "global_step": 30488, "epoch": 342, "lr": 9.700769894834064e-05}
{"train_loss": 0.28022846579551697, "global_step": 30489, "epoch": 342, "lr": 9.700750139998996e-05}
{"train_loss": 0.39863359928131104, "global_step": 30490, "epoch": 342, "lr": 9.700730384531969e-05}
{"train_loss": 0.3224022090435028, "global_step": 30491, "epoch": 342, "lr": 9.700710628432986e-05}
{"train_loss": 0.35762712359428406, "global_step": 30492, "epoch": 342, "lr": 9.70069087170205e-05}
{"train_loss": 0.34939512610435486, "global_step": 30493, "epoch": 342, "lr": 9.700671114339164e-05}
{"train_loss": 0.41123825311660767, "global_step": 30494, "epoch": 342, "lr": 9.700651356344329e-05}
{"train_loss": 0.40601927042007446, "global_step": 30495, "epoch": 342, "lr": 9.70063159771755e-05}
{"train_loss": 0.4374120235443115, "global_step": 30496, "epoch": 342, "lr": 9.700611838458826e-05}
{"train_loss": 0.3753724694252014, "global_step": 30497, "epoch": 342, "lr": 9.700592078568165e-05}
{"train_loss": 0.35510721802711487, "global_step": 30498, "epoch": 342, "lr": 9.700572318045565e-05}
{"train_loss": 0.4867228865623474, "global_step": 30499, "epoch": 342, "lr": 9.700552556891029e-05}
{"train_loss": 0.5250602960586548, "global_step": 30500, "epoch": 342, "lr": 9.700532795104564e-05}
{"train_loss": 0.4441356062889099, "global_step": 30501, "epoch": 342, "lr": 9.700513032686167e-05}
{"train_loss": 0.27665427327156067, "global_step": 30502, "epoch": 342, "lr": 9.700493269635846e-05}
{"train_loss": 0.3564912676811218, "global_step": 30503, "epoch": 342, "lr": 9.700473505953599e-05}
{"train_loss": 0.4840778410434723, "global_step": 30504, "epoch": 342, "lr": 9.70045374163943e-05}
{"train_loss": 0.43992045521736145, "global_step": 30505, "epoch": 342, "lr": 9.700433976693344e-05}
{"train_loss": 0.2677880525588989, "global_step": 30506, "epoch": 342, "lr": 9.700414211115342e-05}
{"train_loss": 0.34373611211776733, "global_step": 30507, "epoch": 342, "lr": 9.700394444905425e-05}
{"train_loss": 0.3947581648826599, "global_step": 30508, "epoch": 342, "lr": 9.700374678063599e-05}
{"train_loss": 0.3914867341518402, "global_step": 30509, "epoch": 342, "lr": 9.700354910589864e-05}
{"train_loss": 0.36722275614738464, "global_step": 30510, "epoch": 342, "lr": 9.700335142484225e-05}
{"train_loss": 0.4251854121685028, "global_step": 30511, "epoch": 342, "lr": 9.70031537374668e-05}
{"train_loss": 0.2691870629787445, "global_step": 30512, "epoch": 342, "lr": 9.700295604377238e-05}
{"train_loss": 0.31090015172958374, "global_step": 30513, "epoch": 342, "lr": 9.700275834375897e-05}
{"train_loss": 0.34931445121765137, "global_step": 30514, "epoch": 342, "lr": 9.700256063742662e-05}
{"train_loss": 0.332324743270874, "global_step": 30515, "epoch": 342, "lr": 9.700236292477533e-05}
{"train_loss": 0.39541998505592346, "global_step": 30516, "epoch": 342, "lr": 9.700216520580518e-05}
{"train_loss": 0.3776283860206604, "global_step": 30517, "epoch": 342, "lr": 9.700196748051613e-05}
{"train_loss": 0.3268471658229828, "global_step": 30518, "epoch": 342, "lr": 9.700176974890825e-05}
{"train_loss": 0.41231194138526917, "global_step": 30519, "epoch": 342, "lr": 9.700157201098155e-05}
{"train_loss": 0.3704676330089569, "global_step": 30520, "epoch": 342, "lr": 9.700137426673608e-05}
{"train_loss": 0.45316317677497864, "global_step": 30521, "epoch": 342, "lr": 9.700117651617183e-05}
{"train_loss": 0.390977144241333, "global_step": 30522, "epoch": 342, "lr": 9.700097875928884e-05}
{"train_loss": 0.4063725769519806, "global_step": 30523, "epoch": 342, "lr": 9.700078099608716e-05}
{"train_loss": 0.444369912147522, "global_step": 30524, "epoch": 342, "lr": 9.700058322656677e-05}
{"train_loss": 0.3517841398715973, "global_step": 30525, "epoch": 342, "lr": 9.700038545072776e-05}
{"train_loss": 0.3649274879292156, "global_step": 30526, "epoch": 342, "lr": 9.700018766857011e-05, "val_loss": 2.2576229572296143}
{"train_loss": 0.35362178087234497, "global_step": 30527, "epoch": 343, "lr": 9.699998988009384e-05}
{"train_loss": 0.395334929227829, "global_step": 30528, "epoch": 343, "lr": 9.699979208529901e-05}
{"train_loss": 0.4360998272895813, "global_step": 30529, "epoch": 343, "lr": 9.699959428418562e-05}
{"train_loss": 0.30714112520217896, "global_step": 30530, "epoch": 343, "lr": 9.699939647675371e-05}
{"train_loss": 0.31204989552497864, "global_step": 30531, "epoch": 343, "lr": 9.699919866300331e-05}
{"train_loss": 0.3783908784389496, "global_step": 30532, "epoch": 343, "lr": 9.699900084293446e-05}
{"train_loss": 0.3565227687358856, "global_step": 30533, "epoch": 343, "lr": 9.699880301654713e-05}
{"train_loss": 0.4255657196044922, "global_step": 30534, "epoch": 343, "lr": 9.699860518384139e-05}
{"train_loss": 0.45176130533218384, "global_step": 30535, "epoch": 343, "lr": 9.699840734481728e-05}
{"train_loss": 0.446158766746521, "global_step": 30536, "epoch": 343, "lr": 9.699820949947478e-05}
{"train_loss": 0.5300940275192261, "global_step": 30537, "epoch": 343, "lr": 9.699801164781397e-05}
{"train_loss": 0.2675837576389313, "global_step": 30538, "epoch": 343, "lr": 9.699781378983483e-05}
{"train_loss": 0.3914045989513397, "global_step": 30539, "epoch": 343, "lr": 9.699761592553743e-05}
{"train_loss": 0.386538565158844, "global_step": 30540, "epoch": 343, "lr": 9.699741805492177e-05}
{"train_loss": 0.42023488879203796, "global_step": 30541, "epoch": 343, "lr": 9.699722017798785e-05}
{"train_loss": 0.37908241152763367, "global_step": 30542, "epoch": 343, "lr": 9.699702229473575e-05}
{"train_loss": 0.5019447803497314, "global_step": 30543, "epoch": 343, "lr": 9.699682440516548e-05}
{"train_loss": 0.4017024338245392, "global_step": 30544, "epoch": 343, "lr": 9.699662650927704e-05}
{"train_loss": 0.42451974749565125, "global_step": 30545, "epoch": 343, "lr": 9.69964286070705e-05}
{"train_loss": 0.41237181425094604, "global_step": 30546, "epoch": 343, "lr": 9.699623069854584e-05}
{"train_loss": 0.555389404296875, "global_step": 30547, "epoch": 343, "lr": 9.699603278370312e-05}
{"train_loss": 0.3900147080421448, "global_step": 30548, "epoch": 343, "lr": 9.699583486254235e-05}
{"train_loss": 0.32289865612983704, "global_step": 30549, "epoch": 343, "lr": 9.699563693506357e-05}
{"train_loss": 0.2940830588340759, "global_step": 30550, "epoch": 343, "lr": 9.699543900126679e-05}
{"train_loss": 0.496062695980072, "global_step": 30551, "epoch": 343, "lr": 9.699524106115207e-05}
{"train_loss": 0.367535799741745, "global_step": 30552, "epoch": 343, "lr": 9.699504311471938e-05}
{"train_loss": 0.46711936593055725, "global_step": 30553, "epoch": 343, "lr": 9.699484516196881e-05}
{"train_loss": 0.3624018132686615, "global_step": 30554, "epoch": 343, "lr": 9.699464720290034e-05}
{"train_loss": 0.40824809670448303, "global_step": 30555, "epoch": 343, "lr": 9.699444923751401e-05}
{"train_loss": 0.3996773064136505, "global_step": 30556, "epoch": 343, "lr": 9.699425126580985e-05}
{"train_loss": 0.40024957060813904, "global_step": 30557, "epoch": 343, "lr": 9.699405328778789e-05}
{"train_loss": 0.3221329152584076, "global_step": 30558, "epoch": 343, "lr": 9.699385530344815e-05}
{"train_loss": 0.31953099370002747, "global_step": 30559, "epoch": 343, "lr": 9.699365731279066e-05}
{"train_loss": 0.38701844215393066, "global_step": 30560, "epoch": 343, "lr": 9.699345931581546e-05}
{"train_loss": 0.39342832565307617, "global_step": 30561, "epoch": 343, "lr": 9.699326131252254e-05}
{"train_loss": 0.35432931780815125, "global_step": 30562, "epoch": 343, "lr": 9.699306330291194e-05}
{"train_loss": 0.35930100083351135, "global_step": 30563, "epoch": 343, "lr": 9.699286528698372e-05}
{"train_loss": 0.429153710603714, "global_step": 30564, "epoch": 343, "lr": 9.699266726473787e-05}
{"train_loss": 0.294738233089447, "global_step": 30565, "epoch": 343, "lr": 9.699246923617443e-05}
{"train_loss": 0.5122153162956238, "global_step": 30566, "epoch": 343, "lr": 9.699227120129343e-05}
{"train_loss": 0.4058755338191986, "global_step": 30567, "epoch": 343, "lr": 9.699207316009488e-05}
{"train_loss": 0.5292272567749023, "global_step": 30568, "epoch": 343, "lr": 9.699187511257883e-05}
{"train_loss": 0.32846519351005554, "global_step": 30569, "epoch": 343, "lr": 9.69916770587453e-05}
{"train_loss": 0.40635326504707336, "global_step": 30570, "epoch": 343, "lr": 9.699147899859431e-05}
{"train_loss": 0.47327297925949097, "global_step": 30571, "epoch": 343, "lr": 9.699128093212588e-05}
{"train_loss": 0.2917194366455078, "global_step": 30572, "epoch": 343, "lr": 9.699108285934004e-05}
{"train_loss": 0.35214120149612427, "global_step": 30573, "epoch": 343, "lr": 9.699088478023683e-05}
{"train_loss": 0.39613407850265503, "global_step": 30574, "epoch": 343, "lr": 9.699068669481626e-05}
{"train_loss": 0.5658680200576782, "global_step": 30575, "epoch": 343, "lr": 9.699048860307837e-05}
{"train_loss": 0.4482187032699585, "global_step": 30576, "epoch": 343, "lr": 9.69902905050232e-05}
{"train_loss": 0.38490262627601624, "global_step": 30577, "epoch": 343, "lr": 9.699009240065073e-05}
{"train_loss": 0.4091404378414154, "global_step": 30578, "epoch": 343, "lr": 9.698989428996104e-05}
{"train_loss": 0.43895700573921204, "global_step": 30579, "epoch": 343, "lr": 9.698969617295411e-05}
{"train_loss": 0.3839761018753052, "global_step": 30580, "epoch": 343, "lr": 9.698949804963e-05}
{"train_loss": 0.38937440514564514, "global_step": 30581, "epoch": 343, "lr": 9.698929991998874e-05}
{"train_loss": 0.35696783661842346, "global_step": 30582, "epoch": 343, "lr": 9.698910178403032e-05}
{"train_loss": 0.31792163848876953, "global_step": 30583, "epoch": 343, "lr": 9.698890364175477e-05}
{"train_loss": 0.32383108139038086, "global_step": 30584, "epoch": 343, "lr": 9.698870549316217e-05}
{"train_loss": 0.39867278933525085, "global_step": 30585, "epoch": 343, "lr": 9.698850733825249e-05}
{"train_loss": 0.4349033832550049, "global_step": 30586, "epoch": 343, "lr": 9.698830917702578e-05}
{"train_loss": 0.41070130467414856, "global_step": 30587, "epoch": 343, "lr": 9.698811100948207e-05}
{"train_loss": 0.4495886266231537, "global_step": 30588, "epoch": 343, "lr": 9.698791283562138e-05}
{"train_loss": 0.363597571849823, "global_step": 30589, "epoch": 343, "lr": 9.698771465544374e-05}
{"train_loss": 0.4494457244873047, "global_step": 30590, "epoch": 343, "lr": 9.698751646894916e-05}
{"train_loss": 0.37313112616539, "global_step": 30591, "epoch": 343, "lr": 9.69873182761377e-05}
{"train_loss": 0.299755334854126, "global_step": 30592, "epoch": 343, "lr": 9.698712007700937e-05}
{"train_loss": 0.4520757496356964, "global_step": 30593, "epoch": 343, "lr": 9.698692187156419e-05}
{"train_loss": 0.36621421575546265, "global_step": 30594, "epoch": 343, "lr": 9.698672365980218e-05}
{"train_loss": 0.3464435040950775, "global_step": 30595, "epoch": 343, "lr": 9.698652544172338e-05}
{"train_loss": 0.3741396367549896, "global_step": 30596, "epoch": 343, "lr": 9.698632721732783e-05}
{"train_loss": 0.4482334554195404, "global_step": 30597, "epoch": 343, "lr": 9.698612898661552e-05}
{"train_loss": 0.3083285093307495, "global_step": 30598, "epoch": 343, "lr": 9.69859307495865e-05}
{"train_loss": 0.25526806712150574, "global_step": 30599, "epoch": 343, "lr": 9.698573250624081e-05}
{"train_loss": 0.3716571629047394, "global_step": 30600, "epoch": 343, "lr": 9.698553425657846e-05}
{"train_loss": 0.35219672322273254, "global_step": 30601, "epoch": 343, "lr": 9.698533600059948e-05}
{"train_loss": 0.286300927400589, "global_step": 30602, "epoch": 343, "lr": 9.698513773830387e-05}
{"train_loss": 0.3922828733921051, "global_step": 30603, "epoch": 343, "lr": 9.698493946969169e-05}
{"train_loss": 0.40321987867355347, "global_step": 30604, "epoch": 343, "lr": 9.698474119476297e-05}
{"train_loss": 0.3519054651260376, "global_step": 30605, "epoch": 343, "lr": 9.698454291351772e-05}
{"train_loss": 0.4072282612323761, "global_step": 30606, "epoch": 343, "lr": 9.698434462595597e-05}
{"train_loss": 0.36871978640556335, "global_step": 30607, "epoch": 343, "lr": 9.698414633207774e-05}
{"train_loss": 0.4164673388004303, "global_step": 30608, "epoch": 343, "lr": 9.698394803188306e-05}
{"train_loss": 0.43483060598373413, "global_step": 30609, "epoch": 343, "lr": 9.698374972537198e-05}
{"train_loss": 0.43970414996147156, "global_step": 30610, "epoch": 343, "lr": 9.69835514125445e-05}
{"train_loss": 0.4223763048648834, "global_step": 30611, "epoch": 343, "lr": 9.698335309340066e-05}
{"train_loss": 0.38742196559906006, "global_step": 30612, "epoch": 343, "lr": 9.698315476794046e-05}
{"train_loss": 0.3422049283981323, "global_step": 30613, "epoch": 343, "lr": 9.698295643616398e-05}
{"train_loss": 0.3511248528957367, "global_step": 30614, "epoch": 343, "lr": 9.698275809807118e-05}
{"train_loss": 0.39155823829468717, "global_step": 30615, "epoch": 343, "lr": 9.698255975366214e-05, "val_loss": 2.2633137702941895}
{"train_loss": 0.32150042057037354, "global_step": 30616, "epoch": 344, "lr": 9.698236140293686e-05}
{"train_loss": 0.46019846200942993, "global_step": 30617, "epoch": 344, "lr": 9.698216304589538e-05}
{"train_loss": 0.36803653836250305, "global_step": 30618, "epoch": 344, "lr": 9.698196468253772e-05}
{"train_loss": 0.34871533513069153, "global_step": 30619, "epoch": 344, "lr": 9.69817663128639e-05}
{"train_loss": 0.3888895809650421, "global_step": 30620, "epoch": 344, "lr": 9.698156793687396e-05}
{"train_loss": 0.3660929501056671, "global_step": 30621, "epoch": 344, "lr": 9.698136955456791e-05}
{"train_loss": 0.342970609664917, "global_step": 30622, "epoch": 344, "lr": 9.698117116594581e-05}
{"train_loss": 0.3217373490333557, "global_step": 30623, "epoch": 344, "lr": 9.698097277100765e-05}
{"train_loss": 0.3758642375469208, "global_step": 30624, "epoch": 344, "lr": 9.698077436975348e-05}
{"train_loss": 0.43015363812446594, "global_step": 30625, "epoch": 344, "lr": 9.698057596218331e-05}
{"train_loss": 0.39080387353897095, "global_step": 30626, "epoch": 344, "lr": 9.698037754829716e-05}
{"train_loss": 0.3746219277381897, "global_step": 30627, "epoch": 344, "lr": 9.69801791280951e-05}
{"train_loss": 0.3817408084869385, "global_step": 30628, "epoch": 344, "lr": 9.69799807015771e-05}
{"train_loss": 0.27478867769241333, "global_step": 30629, "epoch": 344, "lr": 9.697978226874324e-05}
{"train_loss": 0.367978036403656, "global_step": 30630, "epoch": 344, "lr": 9.69795838295935e-05}
{"train_loss": 0.5043801069259644, "global_step": 30631, "epoch": 344, "lr": 9.697938538412794e-05}
{"train_loss": 0.27801185846328735, "global_step": 30632, "epoch": 344, "lr": 9.697918693234656e-05}
{"train_loss": 0.4868253767490387, "global_step": 30633, "epoch": 344, "lr": 9.697898847424943e-05}
{"train_loss": 0.3438228368759155, "global_step": 30634, "epoch": 344, "lr": 9.697879000983651e-05}
{"train_loss": 0.3057215213775635, "global_step": 30635, "epoch": 344, "lr": 9.697859153910789e-05}
{"train_loss": 0.3945983350276947, "global_step": 30636, "epoch": 344, "lr": 9.697839306206356e-05}
{"train_loss": 0.3955596387386322, "global_step": 30637, "epoch": 344, "lr": 9.697819457870356e-05}
{"train_loss": 0.3068098723888397, "global_step": 30638, "epoch": 344, "lr": 9.697799608902791e-05}
{"train_loss": 0.40872183442115784, "global_step": 30639, "epoch": 344, "lr": 9.697779759303665e-05}
{"train_loss": 0.598139762878418, "global_step": 30640, "epoch": 344, "lr": 9.69775990907298e-05}
{"train_loss": 0.5485588908195496, "global_step": 30641, "epoch": 344, "lr": 9.697740058210738e-05}
{"train_loss": 0.3189590573310852, "global_step": 30642, "epoch": 344, "lr": 9.697720206716942e-05}
{"train_loss": 0.38974529504776, "global_step": 30643, "epoch": 344, "lr": 9.697700354591595e-05}
{"train_loss": 0.4189172089099884, "global_step": 30644, "epoch": 344, "lr": 9.697680501834699e-05}
{"train_loss": 0.37696102261543274, "global_step": 30645, "epoch": 344, "lr": 9.697660648446257e-05}
{"train_loss": 0.3602012097835541, "global_step": 30646, "epoch": 344, "lr": 9.697640794426272e-05}
{"train_loss": 0.3484647572040558, "global_step": 30647, "epoch": 344, "lr": 9.697620939774747e-05}
{"train_loss": 0.43130460381507874, "global_step": 30648, "epoch": 344, "lr": 9.697601084491683e-05}
{"train_loss": 0.6440749764442444, "global_step": 30649, "epoch": 344, "lr": 9.697581228577084e-05}
{"train_loss": 0.49973827600479126, "global_step": 30650, "epoch": 344, "lr": 9.697561372030954e-05}
{"train_loss": 0.39845988154411316, "global_step": 30651, "epoch": 344, "lr": 9.697541514853293e-05}
{"train_loss": 0.34027716517448425, "global_step": 30652, "epoch": 344, "lr": 9.697521657044104e-05}
{"train_loss": 0.3893806040287018, "global_step": 30653, "epoch": 344, "lr": 9.697501798603393e-05}
{"train_loss": 0.341508150100708, "global_step": 30654, "epoch": 344, "lr": 9.697481939531157e-05}
{"train_loss": 0.5274503231048584, "global_step": 30655, "epoch": 344, "lr": 9.697462079827404e-05}
{"train_loss": 0.42650583386421204, "global_step": 30656, "epoch": 344, "lr": 9.697442219492133e-05}
{"train_loss": 0.40487873554229736, "global_step": 30657, "epoch": 344, "lr": 9.69742235852535e-05}
{"train_loss": 0.3350018858909607, "global_step": 30658, "epoch": 344, "lr": 9.697402496927054e-05}
{"train_loss": 0.32668307423591614, "global_step": 30659, "epoch": 344, "lr": 9.697382634697249e-05}
{"train_loss": 0.3644290864467621, "global_step": 30660, "epoch": 344, "lr": 9.697362771835941e-05}
{"train_loss": 0.4106685519218445, "global_step": 30661, "epoch": 344, "lr": 9.697342908343128e-05}
{"train_loss": 0.33610332012176514, "global_step": 30662, "epoch": 344, "lr": 9.697323044218814e-05}
{"train_loss": 0.5120850801467896, "global_step": 30663, "epoch": 344, "lr": 9.697303179463003e-05}
{"train_loss": 0.5252667665481567, "global_step": 30664, "epoch": 344, "lr": 9.697283314075697e-05}
{"train_loss": 0.4516438841819763, "global_step": 30665, "epoch": 344, "lr": 9.697263448056897e-05}
{"train_loss": 0.3869825303554535, "global_step": 30666, "epoch": 344, "lr": 9.69724358140661e-05}
{"train_loss": 0.4084840416908264, "global_step": 30667, "epoch": 344, "lr": 9.697223714124834e-05}
{"train_loss": 0.44631651043891907, "global_step": 30668, "epoch": 344, "lr": 9.697203846211573e-05}
{"train_loss": 0.3102271556854248, "global_step": 30669, "epoch": 344, "lr": 9.69718397766683e-05}
{"train_loss": 0.4142391085624695, "global_step": 30670, "epoch": 344, "lr": 9.697164108490611e-05}
{"train_loss": 0.38810184597969055, "global_step": 30671, "epoch": 344, "lr": 9.697144238682913e-05}
{"train_loss": 0.42935165762901306, "global_step": 30672, "epoch": 344, "lr": 9.69712436824374e-05}
{"train_loss": 0.47189798951148987, "global_step": 30673, "epoch": 344, "lr": 9.697104497173099e-05}
{"train_loss": 0.3565075695514679, "global_step": 30674, "epoch": 344, "lr": 9.697084625470986e-05}
{"train_loss": 0.43208396434783936, "global_step": 30675, "epoch": 344, "lr": 9.69706475313741e-05}
{"train_loss": 0.4952523708343506, "global_step": 30676, "epoch": 344, "lr": 9.69704488017237e-05}
{"train_loss": 0.3790132701396942, "global_step": 30677, "epoch": 344, "lr": 9.697025006575871e-05}
{"train_loss": 0.3467174768447876, "global_step": 30678, "epoch": 344, "lr": 9.697005132347912e-05}
{"train_loss": 0.42708268761634827, "global_step": 30679, "epoch": 344, "lr": 9.696985257488498e-05}
{"train_loss": 0.43812140822410583, "global_step": 30680, "epoch": 344, "lr": 9.696965381997634e-05}
{"train_loss": 0.30882731080055237, "global_step": 30681, "epoch": 344, "lr": 9.696945505875318e-05}
{"train_loss": 0.44629910588264465, "global_step": 30682, "epoch": 344, "lr": 9.696925629121556e-05}
{"train_loss": 0.4131113886833191, "global_step": 30683, "epoch": 344, "lr": 9.696905751736348e-05}
{"train_loss": 0.3214549124240875, "global_step": 30684, "epoch": 344, "lr": 9.6968858737197e-05}
{"train_loss": 0.30349868535995483, "global_step": 30685, "epoch": 344, "lr": 9.696865995071613e-05}
{"train_loss": 0.3996710479259491, "global_step": 30686, "epoch": 344, "lr": 9.696846115792089e-05}
{"train_loss": 0.44063299894332886, "global_step": 30687, "epoch": 344, "lr": 9.69682623588113e-05}
{"train_loss": 0.3991103172302246, "global_step": 30688, "epoch": 344, "lr": 9.696806355338742e-05}
{"train_loss": 0.5088030695915222, "global_step": 30689, "epoch": 344, "lr": 9.696786474164925e-05}
{"train_loss": 0.28305909037590027, "global_step": 30690, "epoch": 344, "lr": 9.696766592359682e-05}
{"train_loss": 0.4433136582374573, "global_step": 30691, "epoch": 344, "lr": 9.696746709923015e-05}
{"train_loss": 0.31615811586380005, "global_step": 30692, "epoch": 344, "lr": 9.69672682685493e-05}
{"train_loss": 0.3780786395072937, "global_step": 30693, "epoch": 344, "lr": 9.696706943155426e-05}
{"train_loss": 0.33141154050827026, "global_step": 30694, "epoch": 344, "lr": 9.696687058824507e-05}
{"train_loss": 0.44019168615341187, "global_step": 30695, "epoch": 344, "lr": 9.696667173862175e-05}
{"train_loss": 0.4711562395095825, "global_step": 30696, "epoch": 344, "lr": 9.696647288268435e-05}
{"train_loss": 0.38188764452934265, "global_step": 30697, "epoch": 344, "lr": 9.696627402043287e-05}
{"train_loss": 0.3549889028072357, "global_step": 30698, "epoch": 344, "lr": 9.696607515186735e-05}
{"train_loss": 0.305936336517334, "global_step": 30699, "epoch": 344, "lr": 9.696587627698781e-05}
{"train_loss": 0.376755028963089, "global_step": 30700, "epoch": 344, "lr": 9.696567739579427e-05}
{"train_loss": 0.39428195357322693, "global_step": 30701, "epoch": 344, "lr": 9.696547850828679e-05}
{"train_loss": 0.40400710701942444, "global_step": 30702, "epoch": 344, "lr": 9.696527961446535e-05}
{"train_loss": 0.41834011673927307, "global_step": 30703, "epoch": 344, "lr": 9.696508071433003e-05}
{"train_loss": 0.3966082799970434, "global_step": 30704, "epoch": 344, "lr": 9.696488180788081e-05, "val_loss": 2.31935453414917}
{"train_loss": 0.41156840324401855, "global_step": 30705, "epoch": 345, "lr": 9.696468289511773e-05}
{"train_loss": 0.3631928265094757, "global_step": 30706, "epoch": 345, "lr": 9.696448397604083e-05}
{"train_loss": 0.31680792570114136, "global_step": 30707, "epoch": 345, "lr": 9.696428505065012e-05}
{"train_loss": 0.4785097539424896, "global_step": 30708, "epoch": 345, "lr": 9.696408611894564e-05}
{"train_loss": 0.33689868450164795, "global_step": 30709, "epoch": 345, "lr": 9.696388718092741e-05}
{"train_loss": 0.33624032139778137, "global_step": 30710, "epoch": 345, "lr": 9.696368823659545e-05}
{"train_loss": 0.3840469419956207, "global_step": 30711, "epoch": 345, "lr": 9.696348928594981e-05}
{"train_loss": 0.323663592338562, "global_step": 30712, "epoch": 345, "lr": 9.696329032899049e-05}
{"train_loss": 0.36675870418548584, "global_step": 30713, "epoch": 345, "lr": 9.696309136571752e-05}
{"train_loss": 0.4599277079105377, "global_step": 30714, "epoch": 345, "lr": 9.696289239613096e-05}
{"train_loss": 0.24688252806663513, "global_step": 30715, "epoch": 345, "lr": 9.696269342023079e-05}
{"train_loss": 0.41992849111557007, "global_step": 30716, "epoch": 345, "lr": 9.696249443801706e-05}
{"train_loss": 0.4023008942604065, "global_step": 30717, "epoch": 345, "lr": 9.696229544948981e-05}
{"train_loss": 0.350877583026886, "global_step": 30718, "epoch": 345, "lr": 9.696209645464905e-05}
{"train_loss": 0.2848408818244934, "global_step": 30719, "epoch": 345, "lr": 9.696189745349479e-05}
{"train_loss": 0.31322580575942993, "global_step": 30720, "epoch": 345, "lr": 9.696169844602708e-05}
{"train_loss": 0.5174124836921692, "global_step": 30721, "epoch": 345, "lr": 9.696149943224595e-05}
{"train_loss": 0.3777323365211487, "global_step": 30722, "epoch": 345, "lr": 9.696130041215141e-05}
{"train_loss": 0.3443208336830139, "global_step": 30723, "epoch": 345, "lr": 9.69611013857435e-05}
{"train_loss": 0.3451775908470154, "global_step": 30724, "epoch": 345, "lr": 9.696090235302225e-05}
{"train_loss": 0.3519660234451294, "global_step": 30725, "epoch": 345, "lr": 9.696070331398768e-05}
{"train_loss": 0.4527408480644226, "global_step": 30726, "epoch": 345, "lr": 9.69605042686398e-05}
{"train_loss": 0.3666044771671295, "global_step": 30727, "epoch": 345, "lr": 9.696030521697866e-05}
{"train_loss": 0.34233224391937256, "global_step": 30728, "epoch": 345, "lr": 9.696010615900428e-05}
{"train_loss": 0.42807862162590027, "global_step": 30729, "epoch": 345, "lr": 9.69599070947167e-05}
{"train_loss": 0.36180639266967773, "global_step": 30730, "epoch": 345, "lr": 9.695970802411592e-05}
{"train_loss": 0.39828234910964966, "global_step": 30731, "epoch": 345, "lr": 9.695950894720197e-05}
{"train_loss": 0.30347633361816406, "global_step": 30732, "epoch": 345, "lr": 9.69593098639749e-05}
{"train_loss": 0.3249014616012573, "global_step": 30733, "epoch": 345, "lr": 9.695911077443471e-05}
{"train_loss": 0.3908418118953705, "global_step": 30734, "epoch": 345, "lr": 9.695891167858145e-05}
{"train_loss": 0.33670341968536377, "global_step": 30735, "epoch": 345, "lr": 9.695871257641514e-05}
{"train_loss": 0.35828787088394165, "global_step": 30736, "epoch": 345, "lr": 9.69585134679358e-05}
{"train_loss": 0.34072279930114746, "global_step": 30737, "epoch": 345, "lr": 9.695831435314347e-05}
{"train_loss": 0.25285911560058594, "global_step": 30738, "epoch": 345, "lr": 9.695811523203815e-05}
{"train_loss": 0.3129820227622986, "global_step": 30739, "epoch": 345, "lr": 9.695791610461988e-05}
{"train_loss": 0.4019499123096466, "global_step": 30740, "epoch": 345, "lr": 9.69577169708887e-05}
{"train_loss": 0.2922058403491974, "global_step": 30741, "epoch": 345, "lr": 9.695751783084464e-05}
{"train_loss": 0.3883659839630127, "global_step": 30742, "epoch": 345, "lr": 9.69573186844877e-05}
{"train_loss": 0.2813548147678375, "global_step": 30743, "epoch": 345, "lr": 9.695711953181792e-05}
{"train_loss": 0.3864814341068268, "global_step": 30744, "epoch": 345, "lr": 9.695692037283533e-05}
{"train_loss": 0.4407285749912262, "global_step": 30745, "epoch": 345, "lr": 9.695672120753996e-05}
{"train_loss": 0.34363439679145813, "global_step": 30746, "epoch": 345, "lr": 9.695652203593181e-05}
{"train_loss": 0.33609214425086975, "global_step": 30747, "epoch": 345, "lr": 9.695632285801095e-05}
{"train_loss": 0.36202317476272583, "global_step": 30748, "epoch": 345, "lr": 9.695612367377738e-05}
{"train_loss": 0.3196891248226166, "global_step": 30749, "epoch": 345, "lr": 9.695592448323113e-05}
{"train_loss": 0.26187604665756226, "global_step": 30750, "epoch": 345, "lr": 9.695572528637223e-05}
{"train_loss": 0.42414578795433044, "global_step": 30751, "epoch": 345, "lr": 9.695552608320071e-05}
{"train_loss": 0.46095219254493713, "global_step": 30752, "epoch": 345, "lr": 9.695532687371658e-05}
{"train_loss": 0.3239615261554718, "global_step": 30753, "epoch": 345, "lr": 9.695512765791988e-05}
{"train_loss": 0.36977243423461914, "global_step": 30754, "epoch": 345, "lr": 9.695492843581065e-05}
{"train_loss": 0.40962502360343933, "global_step": 30755, "epoch": 345, "lr": 9.695472920738888e-05}
{"train_loss": 0.35009822249412537, "global_step": 30756, "epoch": 345, "lr": 9.695452997265462e-05}
{"train_loss": 0.41646888852119446, "global_step": 30757, "epoch": 345, "lr": 9.69543307316079e-05}
{"train_loss": 0.33539488911628723, "global_step": 30758, "epoch": 345, "lr": 9.695413148424876e-05}
{"train_loss": 0.2926286458969116, "global_step": 30759, "epoch": 345, "lr": 9.69539322305772e-05}
{"train_loss": 0.4066213369369507, "global_step": 30760, "epoch": 345, "lr": 9.695373297059326e-05}
{"train_loss": 0.423920214176178, "global_step": 30761, "epoch": 345, "lr": 9.695353370429695e-05}
{"train_loss": 0.3611409068107605, "global_step": 30762, "epoch": 345, "lr": 9.695333443168832e-05}
{"train_loss": 0.32047757506370544, "global_step": 30763, "epoch": 345, "lr": 9.695313515276739e-05}
{"train_loss": 0.32940763235092163, "global_step": 30764, "epoch": 345, "lr": 9.695293586753416e-05}
{"train_loss": 0.39483097195625305, "global_step": 30765, "epoch": 345, "lr": 9.695273657598871e-05}
{"train_loss": 0.3590949475765228, "global_step": 30766, "epoch": 345, "lr": 9.695253727813102e-05}
{"train_loss": 0.3955618143081665, "global_step": 30767, "epoch": 345, "lr": 9.695233797396113e-05}
{"train_loss": 0.5575985908508301, "global_step": 30768, "epoch": 345, "lr": 9.695213866347907e-05}
{"train_loss": 0.3268745243549347, "global_step": 30769, "epoch": 345, "lr": 9.695193934668488e-05}
{"train_loss": 0.37148958444595337, "global_step": 30770, "epoch": 345, "lr": 9.695174002357858e-05}
{"train_loss": 0.3565683662891388, "global_step": 30771, "epoch": 345, "lr": 9.695154069416017e-05}
{"train_loss": 0.3437160849571228, "global_step": 30772, "epoch": 345, "lr": 9.695134135842971e-05}
{"train_loss": 0.2929661273956299, "global_step": 30773, "epoch": 345, "lr": 9.69511420163872e-05}
{"train_loss": 0.40566638112068176, "global_step": 30774, "epoch": 345, "lr": 9.69509426680327e-05}
{"train_loss": 0.3647816777229309, "global_step": 30775, "epoch": 345, "lr": 9.695074331336622e-05}
{"train_loss": 0.34427157044410706, "global_step": 30776, "epoch": 345, "lr": 9.695054395238776e-05}
{"train_loss": 0.41631361842155457, "global_step": 30777, "epoch": 345, "lr": 9.695034458509739e-05}
{"train_loss": 0.3973664939403534, "global_step": 30778, "epoch": 345, "lr": 9.695014521149513e-05}
{"train_loss": 0.3516348898410797, "global_step": 30779, "epoch": 345, "lr": 9.694994583158097e-05}
{"train_loss": 0.4273923933506012, "global_step": 30780, "epoch": 345, "lr": 9.694974644535497e-05}
{"train_loss": 0.3360034227371216, "global_step": 30781, "epoch": 345, "lr": 9.694954705281716e-05}
{"train_loss": 0.4433595836162567, "global_step": 30782, "epoch": 345, "lr": 9.694934765396755e-05}
{"train_loss": 0.4340277314186096, "global_step": 30783, "epoch": 345, "lr": 9.694914824880617e-05}
{"train_loss": 0.3431142568588257, "global_step": 30784, "epoch": 345, "lr": 9.694894883733302e-05}
{"train_loss": 0.38928449153900146, "global_step": 30785, "epoch": 345, "lr": 9.69487494195482e-05}
{"train_loss": 0.373415470123291, "global_step": 30786, "epoch": 345, "lr": 9.694854999545168e-05}
{"train_loss": 0.40975797176361084, "global_step": 30787, "epoch": 345, "lr": 9.69483505650435e-05}
{"train_loss": 0.3885176479816437, "global_step": 30788, "epoch": 345, "lr": 9.694815112832369e-05}
{"train_loss": 0.2747313380241394, "global_step": 30789, "epoch": 345, "lr": 9.694795168529226e-05}
{"train_loss": 0.34790438413619995, "global_step": 30790, "epoch": 345, "lr": 9.694775223594925e-05}
{"train_loss": 0.4138035178184509, "global_step": 30791, "epoch": 345, "lr": 9.69475527802947e-05}
{"train_loss": 0.4416324496269226, "global_step": 30792, "epoch": 345, "lr": 9.694735331832862e-05}
{"train_loss": 0.3685949592108137, "global_step": 30793, "epoch": 345, "lr": 9.694715385005104e-05, "val_loss": 2.3039886951446533, "train_action_mse_error": 38.79277038574219}
{"train_loss": 0.3187650442123413, "global_step": 30794, "epoch": 346, "lr": 9.694695437546198e-05}
{"train_loss": 0.4100470244884491, "global_step": 30795, "epoch": 346, "lr": 9.694675489456148e-05}
{"train_loss": 0.3808440566062927, "global_step": 30796, "epoch": 346, "lr": 9.694655540734956e-05}
{"train_loss": 0.411258727312088, "global_step": 30797, "epoch": 346, "lr": 9.694635591382625e-05}
{"train_loss": 0.4040106236934662, "global_step": 30798, "epoch": 346, "lr": 9.694615641399157e-05}
{"train_loss": 0.32926759123802185, "global_step": 30799, "epoch": 346, "lr": 9.694595690784557e-05}
{"train_loss": 0.48300474882125854, "global_step": 30800, "epoch": 346, "lr": 9.694575739538823e-05}
{"train_loss": 0.3125007152557373, "global_step": 30801, "epoch": 346, "lr": 9.694555787661962e-05}
{"train_loss": 0.3948970139026642, "global_step": 30802, "epoch": 346, "lr": 9.694535835153975e-05}
{"train_loss": 0.4260690212249756, "global_step": 30803, "epoch": 346, "lr": 9.694515882014866e-05}
{"train_loss": 0.29161620140075684, "global_step": 30804, "epoch": 346, "lr": 9.694495928244635e-05}
{"train_loss": 0.44461578130722046, "global_step": 30805, "epoch": 346, "lr": 9.694475973843286e-05}
{"train_loss": 0.3996489942073822, "global_step": 30806, "epoch": 346, "lr": 9.694456018810822e-05}
{"train_loss": 0.44886887073516846, "global_step": 30807, "epoch": 346, "lr": 9.694436063147248e-05}
{"train_loss": 0.32071542739868164, "global_step": 30808, "epoch": 346, "lr": 9.694416106852562e-05}
{"train_loss": 0.4049365520477295, "global_step": 30809, "epoch": 346, "lr": 9.694396149926769e-05}
{"train_loss": 0.43157052993774414, "global_step": 30810, "epoch": 346, "lr": 9.694376192369872e-05}
{"train_loss": 0.32886651158332825, "global_step": 30811, "epoch": 346, "lr": 9.694356234181874e-05}
{"train_loss": 0.4559352397918701, "global_step": 30812, "epoch": 346, "lr": 9.694336275362776e-05}
{"train_loss": 0.41740942001342773, "global_step": 30813, "epoch": 346, "lr": 9.694316315912581e-05}
{"train_loss": 0.38432928919792175, "global_step": 30814, "epoch": 346, "lr": 9.694296355831294e-05}
{"train_loss": 0.40379297733306885, "global_step": 30815, "epoch": 346, "lr": 9.694276395118917e-05}
{"train_loss": 0.30995669960975647, "global_step": 30816, "epoch": 346, "lr": 9.69425643377545e-05}
{"train_loss": 0.3408947288990021, "global_step": 30817, "epoch": 346, "lr": 9.694236471800898e-05}
{"train_loss": 0.41418251395225525, "global_step": 30818, "epoch": 346, "lr": 9.694216509195262e-05}
{"train_loss": 0.36001503467559814, "global_step": 30819, "epoch": 346, "lr": 9.694196545958547e-05}
{"train_loss": 0.4237135052680969, "global_step": 30820, "epoch": 346, "lr": 9.694176582090755e-05}
{"train_loss": 0.5348684787750244, "global_step": 30821, "epoch": 346, "lr": 9.694156617591887e-05}
{"train_loss": 0.4024490416049957, "global_step": 30822, "epoch": 346, "lr": 9.694136652461947e-05}
{"train_loss": 0.33678343892097473, "global_step": 30823, "epoch": 346, "lr": 9.694116686700938e-05}
{"train_loss": 0.29748454689979553, "global_step": 30824, "epoch": 346, "lr": 9.694096720308863e-05}
{"train_loss": 0.3801126182079315, "global_step": 30825, "epoch": 346, "lr": 9.694076753285721e-05}
{"train_loss": 0.30075180530548096, "global_step": 30826, "epoch": 346, "lr": 9.69405678563152e-05}
{"train_loss": 0.3318629860877991, "global_step": 30827, "epoch": 346, "lr": 9.69403681734626e-05}
{"train_loss": 0.3730607330799103, "global_step": 30828, "epoch": 346, "lr": 9.694016848429945e-05}
{"train_loss": 0.2854437232017517, "global_step": 30829, "epoch": 346, "lr": 9.693996878882575e-05}
{"train_loss": 0.39947062730789185, "global_step": 30830, "epoch": 346, "lr": 9.693976908704154e-05}
{"train_loss": 0.41011160612106323, "global_step": 30831, "epoch": 346, "lr": 9.693956937894686e-05}
{"train_loss": 0.33757710456848145, "global_step": 30832, "epoch": 346, "lr": 9.693936966454172e-05}
{"train_loss": 0.4820846617221832, "global_step": 30833, "epoch": 346, "lr": 9.693916994382617e-05}
{"train_loss": 0.36953669786453247, "global_step": 30834, "epoch": 346, "lr": 9.693897021680021e-05}
{"train_loss": 0.43014875054359436, "global_step": 30835, "epoch": 346, "lr": 9.693877048346387e-05}
{"train_loss": 0.29089194536209106, "global_step": 30836, "epoch": 346, "lr": 9.69385707438172e-05}
{"train_loss": 0.4073156416416168, "global_step": 30837, "epoch": 346, "lr": 9.69383709978602e-05}
{"train_loss": 0.3627733588218689, "global_step": 30838, "epoch": 346, "lr": 9.69381712455929e-05}
{"train_loss": 0.25676557421684265, "global_step": 30839, "epoch": 346, "lr": 9.693797148701536e-05}
{"train_loss": 0.38920754194259644, "global_step": 30840, "epoch": 346, "lr": 9.693777172212755e-05}
{"train_loss": 0.31924715638160706, "global_step": 30841, "epoch": 346, "lr": 9.693757195092957e-05}
{"train_loss": 0.27158600091934204, "global_step": 30842, "epoch": 346, "lr": 9.693737217342136e-05}
{"train_loss": 0.35625770688056946, "global_step": 30843, "epoch": 346, "lr": 9.693717238960302e-05}
{"train_loss": 0.4263242781162262, "global_step": 30844, "epoch": 346, "lr": 9.693697259947454e-05}
{"train_loss": 0.40777647495269775, "global_step": 30845, "epoch": 346, "lr": 9.693677280303597e-05}
{"train_loss": 0.25614261627197266, "global_step": 30846, "epoch": 346, "lr": 9.693657300028731e-05}
{"train_loss": 0.346563458442688, "global_step": 30847, "epoch": 346, "lr": 9.693637319122859e-05}
{"train_loss": 0.4523584544658661, "global_step": 30848, "epoch": 346, "lr": 9.693617337585987e-05}
{"train_loss": 0.4133140444755554, "global_step": 30849, "epoch": 346, "lr": 9.693597355418114e-05}
{"train_loss": 0.25183290243148804, "global_step": 30850, "epoch": 346, "lr": 9.693577372619244e-05}
{"train_loss": 0.4160574972629547, "global_step": 30851, "epoch": 346, "lr": 9.693557389189381e-05}
{"train_loss": 0.39858928322792053, "global_step": 30852, "epoch": 346, "lr": 9.693537405128526e-05}
{"train_loss": 0.37217605113983154, "global_step": 30853, "epoch": 346, "lr": 9.69351742043668e-05}
{"train_loss": 0.47150999307632446, "global_step": 30854, "epoch": 346, "lr": 9.69349743511385e-05}
{"train_loss": 0.39673498272895813, "global_step": 30855, "epoch": 346, "lr": 9.693477449160036e-05}
{"train_loss": 0.4173922836780548, "global_step": 30856, "epoch": 346, "lr": 9.693457462575241e-05}
{"train_loss": 0.4824115037918091, "global_step": 30857, "epoch": 346, "lr": 9.693437475359468e-05}
{"train_loss": 0.5503338575363159, "global_step": 30858, "epoch": 346, "lr": 9.693417487512719e-05}
{"train_loss": 0.4523133337497711, "global_step": 30859, "epoch": 346, "lr": 9.693397499034997e-05}
{"train_loss": 0.40195101499557495, "global_step": 30860, "epoch": 346, "lr": 9.693377509926305e-05}
{"train_loss": 0.3950890600681305, "global_step": 30861, "epoch": 346, "lr": 9.693357520186647e-05}
{"train_loss": 0.3765057623386383, "global_step": 30862, "epoch": 346, "lr": 9.693337529816024e-05}
{"train_loss": 0.43084728717803955, "global_step": 30863, "epoch": 346, "lr": 9.693317538814437e-05}
{"train_loss": 0.43481847643852234, "global_step": 30864, "epoch": 346, "lr": 9.693297547181892e-05}
{"train_loss": 0.5191153287887573, "global_step": 30865, "epoch": 346, "lr": 9.693277554918391e-05}
{"train_loss": 0.4010910391807556, "global_step": 30866, "epoch": 346, "lr": 9.693257562023935e-05}
{"train_loss": 0.5306249260902405, "global_step": 30867, "epoch": 346, "lr": 9.693237568498529e-05}
{"train_loss": 0.38275760412216187, "global_step": 30868, "epoch": 346, "lr": 9.693217574342173e-05}
{"train_loss": 0.4922291338443756, "global_step": 30869, "epoch": 346, "lr": 9.693197579554872e-05}
{"train_loss": 0.3927196264266968, "global_step": 30870, "epoch": 346, "lr": 9.693177584136627e-05}
{"train_loss": 0.35282689332962036, "global_step": 30871, "epoch": 346, "lr": 9.693157588087442e-05}
{"train_loss": 0.37748557329177856, "global_step": 30872, "epoch": 346, "lr": 9.69313759140732e-05}
{"train_loss": 0.3374161422252655, "global_step": 30873, "epoch": 346, "lr": 9.693117594096262e-05}
{"train_loss": 0.36781832575798035, "global_step": 30874, "epoch": 346, "lr": 9.69309759615427e-05}
{"train_loss": 0.5621301531791687, "global_step": 30875, "epoch": 346, "lr": 9.693077597581351e-05}
{"train_loss": 0.41626375913619995, "global_step": 30876, "epoch": 346, "lr": 9.693057598377503e-05}
{"train_loss": 0.3680615723133087, "global_step": 30877, "epoch": 346, "lr": 9.693037598542732e-05}
{"train_loss": 0.3040598928928375, "global_step": 30878, "epoch": 346, "lr": 9.693017598077037e-05}
{"train_loss": 0.4338902235031128, "global_step": 30879, "epoch": 346, "lr": 9.692997596980425e-05}
{"train_loss": 0.3439806401729584, "global_step": 30880, "epoch": 346, "lr": 9.692977595252896e-05}
{"train_loss": 0.37373730540275574, "global_step": 30881, "epoch": 346, "lr": 9.692957592894454e-05}
{"train_loss": 0.38974376374416136, "global_step": 30882, "epoch": 346, "lr": 9.6929375899051e-05, "val_loss": 2.240999937057495}
{"train_loss": 0.31220075488090515, "global_step": 30883, "epoch": 347, "lr": 9.692917586284839e-05}
{"train_loss": 0.33221280574798584, "global_step": 30884, "epoch": 347, "lr": 9.69289758203367e-05}
{"train_loss": 0.3806558847427368, "global_step": 30885, "epoch": 347, "lr": 9.692877577151601e-05}
{"train_loss": 0.3678465187549591, "global_step": 30886, "epoch": 347, "lr": 9.692857571638631e-05}
{"train_loss": 0.5396053194999695, "global_step": 30887, "epoch": 347, "lr": 9.692837565494764e-05}
{"train_loss": 0.30620062351226807, "global_step": 30888, "epoch": 347, "lr": 9.69281755872e-05}
{"train_loss": 0.3371700346469879, "global_step": 30889, "epoch": 347, "lr": 9.692797551314345e-05}
{"train_loss": 0.26586291193962097, "global_step": 30890, "epoch": 347, "lr": 9.692777543277801e-05}
{"train_loss": 0.33853718638420105, "global_step": 30891, "epoch": 347, "lr": 9.69275753461037e-05}
{"train_loss": 0.4519979953765869, "global_step": 30892, "epoch": 347, "lr": 9.692737525312054e-05}
{"train_loss": 0.3270626664161682, "global_step": 30893, "epoch": 347, "lr": 9.692717515382859e-05}
{"train_loss": 0.3358627259731293, "global_step": 30894, "epoch": 347, "lr": 9.692697504822784e-05}
{"train_loss": 0.41048353910446167, "global_step": 30895, "epoch": 347, "lr": 9.692677493631834e-05}
{"train_loss": 0.43146756291389465, "global_step": 30896, "epoch": 347, "lr": 9.692657481810007e-05}
{"train_loss": 0.33201178908348083, "global_step": 30897, "epoch": 347, "lr": 9.692637469357312e-05}
{"train_loss": 0.35632309317588806, "global_step": 30898, "epoch": 347, "lr": 9.692617456273752e-05}
{"train_loss": 0.3294086754322052, "global_step": 30899, "epoch": 347, "lr": 9.692597442559323e-05}
{"train_loss": 0.33595117926597595, "global_step": 30900, "epoch": 347, "lr": 9.692577428214033e-05}
{"train_loss": 0.5063126087188721, "global_step": 30901, "epoch": 347, "lr": 9.692557413237882e-05}
{"train_loss": 0.407065749168396, "global_step": 30902, "epoch": 347, "lr": 9.692537397630874e-05}
{"train_loss": 0.39675581455230713, "global_step": 30903, "epoch": 347, "lr": 9.692517381393011e-05}
{"train_loss": 0.4521154761314392, "global_step": 30904, "epoch": 347, "lr": 9.692497364524298e-05}
{"train_loss": 0.31369131803512573, "global_step": 30905, "epoch": 347, "lr": 9.692477347024736e-05}
{"train_loss": 0.39584243297576904, "global_step": 30906, "epoch": 347, "lr": 9.692457328894326e-05}
{"train_loss": 0.3709625005722046, "global_step": 30907, "epoch": 347, "lr": 9.692437310133074e-05}
{"train_loss": 0.3560628890991211, "global_step": 30908, "epoch": 347, "lr": 9.69241729074098e-05}
{"train_loss": 0.4687099754810333, "global_step": 30909, "epoch": 347, "lr": 9.692397270718047e-05}
{"train_loss": 0.3726159632205963, "global_step": 30910, "epoch": 347, "lr": 9.692377250064279e-05}
{"train_loss": 0.3672950863838196, "global_step": 30911, "epoch": 347, "lr": 9.692357228779678e-05}
{"train_loss": 0.4582579731941223, "global_step": 30912, "epoch": 347, "lr": 9.692337206864248e-05}
{"train_loss": 0.3121865391731262, "global_step": 30913, "epoch": 347, "lr": 9.692317184317988e-05}
{"train_loss": 0.43788978457450867, "global_step": 30914, "epoch": 347, "lr": 9.692297161140905e-05}
{"train_loss": 0.4243679642677307, "global_step": 30915, "epoch": 347, "lr": 9.692277137332999e-05}
{"train_loss": 0.4018166661262512, "global_step": 30916, "epoch": 347, "lr": 9.692257112894274e-05}
{"train_loss": 0.341955304145813, "global_step": 30917, "epoch": 347, "lr": 9.692237087824732e-05}
{"train_loss": 0.3124534785747528, "global_step": 30918, "epoch": 347, "lr": 9.692217062124376e-05}
{"train_loss": 0.3336181342601776, "global_step": 30919, "epoch": 347, "lr": 9.692197035793208e-05}
{"train_loss": 0.4464050531387329, "global_step": 30920, "epoch": 347, "lr": 9.692177008831233e-05}
{"train_loss": 0.4872753322124481, "global_step": 30921, "epoch": 347, "lr": 9.692156981238452e-05}
{"train_loss": 0.4341781735420227, "global_step": 30922, "epoch": 347, "lr": 9.692136953014866e-05}
{"train_loss": 0.4552247226238251, "global_step": 30923, "epoch": 347, "lr": 9.69211692416048e-05}
{"train_loss": 0.3219072222709656, "global_step": 30924, "epoch": 347, "lr": 9.692096894675295e-05}
{"train_loss": 0.359998494386673, "global_step": 30925, "epoch": 347, "lr": 9.692076864559317e-05}
{"train_loss": 0.3219830095767975, "global_step": 30926, "epoch": 347, "lr": 9.692056833812544e-05}
{"train_loss": 0.3799102306365967, "global_step": 30927, "epoch": 347, "lr": 9.692036802434983e-05}
{"train_loss": 0.3665578365325928, "global_step": 30928, "epoch": 347, "lr": 9.692016770426635e-05}
{"train_loss": 0.3405635952949524, "global_step": 30929, "epoch": 347, "lr": 9.691996737787501e-05}
{"train_loss": 0.40249693393707275, "global_step": 30930, "epoch": 347, "lr": 9.691976704517587e-05}
{"train_loss": 0.30501580238342285, "global_step": 30931, "epoch": 347, "lr": 9.691956670616893e-05}
{"train_loss": 0.3589586913585663, "global_step": 30932, "epoch": 347, "lr": 9.691936636085421e-05}
{"train_loss": 0.4892503619194031, "global_step": 30933, "epoch": 347, "lr": 9.691916600923178e-05}
{"train_loss": 0.34564507007598877, "global_step": 30934, "epoch": 347, "lr": 9.691896565130163e-05}
{"train_loss": 0.40194958448410034, "global_step": 30935, "epoch": 347, "lr": 9.691876528706378e-05}
{"train_loss": 0.38906407356262207, "global_step": 30936, "epoch": 347, "lr": 9.691856491651829e-05}
{"train_loss": 0.33972540497779846, "global_step": 30937, "epoch": 347, "lr": 9.691836453966519e-05}
{"train_loss": 0.34441614151000977, "global_step": 30938, "epoch": 347, "lr": 9.691816415650445e-05}
{"train_loss": 0.3562343716621399, "global_step": 30939, "epoch": 347, "lr": 9.691796376703616e-05}
{"train_loss": 0.4080137312412262, "global_step": 30940, "epoch": 347, "lr": 9.691776337126031e-05}
{"train_loss": 0.30239641666412354, "global_step": 30941, "epoch": 347, "lr": 9.691756296917694e-05}
{"train_loss": 0.26687341928482056, "global_step": 30942, "epoch": 347, "lr": 9.691736256078609e-05}
{"train_loss": 0.32313433289527893, "global_step": 30943, "epoch": 347, "lr": 9.691716214608776e-05}
{"train_loss": 0.4007139801979065, "global_step": 30944, "epoch": 347, "lr": 9.691696172508198e-05}
{"train_loss": 0.3246910572052002, "global_step": 30945, "epoch": 347, "lr": 9.691676129776881e-05}
{"train_loss": 0.4072589874267578, "global_step": 30946, "epoch": 347, "lr": 9.691656086414823e-05}
{"train_loss": 0.279820054769516, "global_step": 30947, "epoch": 347, "lr": 9.69163604242203e-05}
{"train_loss": 0.3884052038192749, "global_step": 30948, "epoch": 347, "lr": 9.691615997798505e-05}
{"train_loss": 0.31211531162261963, "global_step": 30949, "epoch": 347, "lr": 9.691595952544246e-05}
{"train_loss": 0.4115053415298462, "global_step": 30950, "epoch": 347, "lr": 9.691575906659262e-05}
{"train_loss": 0.4450473487377167, "global_step": 30951, "epoch": 347, "lr": 9.691555860143552e-05}
{"train_loss": 0.36458247900009155, "global_step": 30952, "epoch": 347, "lr": 9.69153581299712e-05}
{"train_loss": 0.33017048239707947, "global_step": 30953, "epoch": 347, "lr": 9.691515765219967e-05}
{"train_loss": 0.38764798641204834, "global_step": 30954, "epoch": 347, "lr": 9.691495716812098e-05}
{"train_loss": 0.36865144968032837, "global_step": 30955, "epoch": 347, "lr": 9.691475667773515e-05}
{"train_loss": 0.44544491171836853, "global_step": 30956, "epoch": 347, "lr": 9.69145561810422e-05}
{"train_loss": 0.34578847885131836, "global_step": 30957, "epoch": 347, "lr": 9.691435567804214e-05}
{"train_loss": 0.4063705801963806, "global_step": 30958, "epoch": 347, "lr": 9.691415516873503e-05}
{"train_loss": 0.2607751488685608, "global_step": 30959, "epoch": 347, "lr": 9.691395465312089e-05}
{"train_loss": 0.2543606460094452, "global_step": 30960, "epoch": 347, "lr": 9.691375413119975e-05}
{"train_loss": 0.46115726232528687, "global_step": 30961, "epoch": 347, "lr": 9.69135536029716e-05}
{"train_loss": 0.4508003890514374, "global_step": 30962, "epoch": 347, "lr": 9.691335306843649e-05}
{"train_loss": 0.37830427289009094, "global_step": 30963, "epoch": 347, "lr": 9.691315252759448e-05}
{"train_loss": 0.39358043670654297, "global_step": 30964, "epoch": 347, "lr": 9.691295198044556e-05}
{"train_loss": 0.39492255449295044, "global_step": 30965, "epoch": 347, "lr": 9.691275142698977e-05}
{"train_loss": 0.36507081985473633, "global_step": 30966, "epoch": 347, "lr": 9.691255086722713e-05}
{"train_loss": 0.30804312229156494, "global_step": 30967, "epoch": 347, "lr": 9.691235030115765e-05}
{"train_loss": 0.3468565046787262, "global_step": 30968, "epoch": 347, "lr": 9.69121497287814e-05}
{"train_loss": 0.3271198570728302, "global_step": 30969, "epoch": 347, "lr": 9.691194915009837e-05}
{"train_loss": 0.2801912724971771, "global_step": 30970, "epoch": 347, "lr": 9.69117485651086e-05}
{"train_loss": 0.3714542465933253, "global_step": 30971, "epoch": 347, "lr": 9.691154797381213e-05, "val_loss": 2.27667498588562}
{"train_loss": 0.2888968288898468, "global_step": 30972, "epoch": 348, "lr": 9.691134737620896e-05}
{"train_loss": 0.3270784318447113, "global_step": 30973, "epoch": 348, "lr": 9.691114677229913e-05}
{"train_loss": 0.38352981209754944, "global_step": 30974, "epoch": 348, "lr": 9.691094616208267e-05}
{"train_loss": 0.4215633273124695, "global_step": 30975, "epoch": 348, "lr": 9.69107455455596e-05}
{"train_loss": 0.33039581775665283, "global_step": 30976, "epoch": 348, "lr": 9.691054492272998e-05}
{"train_loss": 0.44431400299072266, "global_step": 30977, "epoch": 348, "lr": 9.691034429359379e-05}
{"train_loss": 0.27064356207847595, "global_step": 30978, "epoch": 348, "lr": 9.691014365815108e-05}
{"train_loss": 0.3386426568031311, "global_step": 30979, "epoch": 348, "lr": 9.690994301640185e-05}
{"train_loss": 0.29710960388183594, "global_step": 30980, "epoch": 348, "lr": 9.690974236834619e-05}
{"train_loss": 0.30071133375167847, "global_step": 30981, "epoch": 348, "lr": 9.690954171398406e-05}
{"train_loss": 0.34159019589424133, "global_step": 30982, "epoch": 348, "lr": 9.690934105331553e-05}
{"train_loss": 0.5145400762557983, "global_step": 30983, "epoch": 348, "lr": 9.69091403863406e-05}
{"train_loss": 0.31680482625961304, "global_step": 30984, "epoch": 348, "lr": 9.69089397130593e-05}
{"train_loss": 0.39175474643707275, "global_step": 30985, "epoch": 348, "lr": 9.690873903347168e-05}
{"train_loss": 0.30493617057800293, "global_step": 30986, "epoch": 348, "lr": 9.690853834757774e-05}
{"train_loss": 0.38654568791389465, "global_step": 30987, "epoch": 348, "lr": 9.690833765537752e-05}
{"train_loss": 0.3646199405193329, "global_step": 30988, "epoch": 348, "lr": 9.690813695687107e-05}
{"train_loss": 0.344465970993042, "global_step": 30989, "epoch": 348, "lr": 9.690793625205836e-05}
{"train_loss": 0.3831106126308441, "global_step": 30990, "epoch": 348, "lr": 9.690773554093946e-05}
{"train_loss": 0.37927985191345215, "global_step": 30991, "epoch": 348, "lr": 9.690753482351439e-05}
{"train_loss": 0.453336238861084, "global_step": 30992, "epoch": 348, "lr": 9.690733409978317e-05}
{"train_loss": 0.36952704191207886, "global_step": 30993, "epoch": 348, "lr": 9.690713336974584e-05}
{"train_loss": 0.34767693281173706, "global_step": 30994, "epoch": 348, "lr": 9.69069326334024e-05}
{"train_loss": 0.34527185559272766, "global_step": 30995, "epoch": 348, "lr": 9.690673189075293e-05}
{"train_loss": 0.33395278453826904, "global_step": 30996, "epoch": 348, "lr": 9.690653114179737e-05}
{"train_loss": 0.3422209322452545, "global_step": 30997, "epoch": 348, "lr": 9.690633038653583e-05}
{"train_loss": 0.33293235301971436, "global_step": 30998, "epoch": 348, "lr": 9.690612962496831e-05}
{"train_loss": 0.3347090780735016, "global_step": 30999, "epoch": 348, "lr": 9.690592885709482e-05}
{"train_loss": 0.4675706624984741, "global_step": 31000, "epoch": 348, "lr": 9.69057280829154e-05}
{"train_loss": 0.41536638140678406, "global_step": 31001, "epoch": 348, "lr": 9.69055273024301e-05}
{"train_loss": 0.3829471170902252, "global_step": 31002, "epoch": 348, "lr": 9.690532651563889e-05}
{"train_loss": 0.30365246534347534, "global_step": 31003, "epoch": 348, "lr": 9.690512572254184e-05}
{"train_loss": 0.358428031206131, "global_step": 31004, "epoch": 348, "lr": 9.690492492313898e-05}
{"train_loss": 0.38885506987571716, "global_step": 31005, "epoch": 348, "lr": 9.690472411743032e-05}
{"train_loss": 0.33098742365837097, "global_step": 31006, "epoch": 348, "lr": 9.690452330541588e-05}
{"train_loss": 0.27698683738708496, "global_step": 31007, "epoch": 348, "lr": 9.690432248709571e-05}
{"train_loss": 0.3408835530281067, "global_step": 31008, "epoch": 348, "lr": 9.690412166246983e-05}
{"train_loss": 0.3715803325176239, "global_step": 31009, "epoch": 348, "lr": 9.690392083153825e-05}
{"train_loss": 0.3081267476081848, "global_step": 31010, "epoch": 348, "lr": 9.690371999430101e-05}
{"train_loss": 0.37798333168029785, "global_step": 31011, "epoch": 348, "lr": 9.690351915075814e-05}
{"train_loss": 0.423370897769928, "global_step": 31012, "epoch": 348, "lr": 9.690331830090967e-05}
{"train_loss": 0.4145922362804413, "global_step": 31013, "epoch": 348, "lr": 9.690311744475561e-05}
{"train_loss": 0.4462251663208008, "global_step": 31014, "epoch": 348, "lr": 9.6902916582296e-05}
{"train_loss": 0.3796357810497284, "global_step": 31015, "epoch": 348, "lr": 9.690271571353087e-05}
{"train_loss": 0.4576442539691925, "global_step": 31016, "epoch": 348, "lr": 9.690251483846024e-05}
{"train_loss": 0.38853272795677185, "global_step": 31017, "epoch": 348, "lr": 9.690231395708413e-05}
{"train_loss": 0.40558353066444397, "global_step": 31018, "epoch": 348, "lr": 9.690211306940258e-05}
{"train_loss": 0.2804645597934723, "global_step": 31019, "epoch": 348, "lr": 9.690191217541562e-05}
{"train_loss": 0.4283621609210968, "global_step": 31020, "epoch": 348, "lr": 9.690171127512326e-05}
{"train_loss": 0.3975663483142853, "global_step": 31021, "epoch": 348, "lr": 9.690151036852554e-05}
{"train_loss": 0.30950912833213806, "global_step": 31022, "epoch": 348, "lr": 9.690130945562248e-05}
{"train_loss": 0.4529248774051666, "global_step": 31023, "epoch": 348, "lr": 9.690110853641412e-05}
{"train_loss": 0.3854411244392395, "global_step": 31024, "epoch": 348, "lr": 9.690090761090047e-05}
{"train_loss": 0.36030814051628113, "global_step": 31025, "epoch": 348, "lr": 9.690070667908158e-05}
{"train_loss": 0.41450080275535583, "global_step": 31026, "epoch": 348, "lr": 9.690050574095744e-05}
{"train_loss": 0.3395947813987732, "global_step": 31027, "epoch": 348, "lr": 9.69003047965281e-05}
{"train_loss": 0.5088896751403809, "global_step": 31028, "epoch": 348, "lr": 9.69001038457936e-05}
{"train_loss": 0.48761314153671265, "global_step": 31029, "epoch": 348, "lr": 9.689990288875395e-05}
{"train_loss": 0.44557109475135803, "global_step": 31030, "epoch": 348, "lr": 9.689970192540918e-05}
{"train_loss": 0.4830528199672699, "global_step": 31031, "epoch": 348, "lr": 9.68995009557593e-05}
{"train_loss": 0.3116016387939453, "global_step": 31032, "epoch": 348, "lr": 9.689929997980437e-05}
{"train_loss": 0.32983365654945374, "global_step": 31033, "epoch": 348, "lr": 9.68990989975444e-05}
{"train_loss": 0.3606038987636566, "global_step": 31034, "epoch": 348, "lr": 9.689889800897941e-05}
{"train_loss": 0.45274192094802856, "global_step": 31035, "epoch": 348, "lr": 9.689869701410944e-05}
{"train_loss": 0.45315858721733093, "global_step": 31036, "epoch": 348, "lr": 9.689849601293451e-05}
{"train_loss": 0.455742746591568, "global_step": 31037, "epoch": 348, "lr": 9.689829500545465e-05}
{"train_loss": 0.4053463935852051, "global_step": 31038, "epoch": 348, "lr": 9.68980939916699e-05}
{"train_loss": 0.411031574010849, "global_step": 31039, "epoch": 348, "lr": 9.689789297158026e-05}
{"train_loss": 0.3864915072917938, "global_step": 31040, "epoch": 348, "lr": 9.689769194518576e-05}
{"train_loss": 0.362925261259079, "global_step": 31041, "epoch": 348, "lr": 9.689749091248645e-05}
{"train_loss": 0.3548584282398224, "global_step": 31042, "epoch": 348, "lr": 9.689728987348233e-05}
{"train_loss": 0.37499210238456726, "global_step": 31043, "epoch": 348, "lr": 9.689708882817346e-05}
{"train_loss": 0.4303077757358551, "global_step": 31044, "epoch": 348, "lr": 9.689688777655983e-05}
{"train_loss": 0.47893083095550537, "global_step": 31045, "epoch": 348, "lr": 9.68966867186415e-05}
{"train_loss": 0.43000227212905884, "global_step": 31046, "epoch": 348, "lr": 9.689648565441847e-05}
{"train_loss": 0.3133312463760376, "global_step": 31047, "epoch": 348, "lr": 9.689628458389078e-05}
{"train_loss": 0.37163975834846497, "global_step": 31048, "epoch": 348, "lr": 9.689608350705846e-05}
{"train_loss": 0.3252103924751282, "global_step": 31049, "epoch": 348, "lr": 9.689588242392153e-05}
{"train_loss": 0.31409239768981934, "global_step": 31050, "epoch": 348, "lr": 9.689568133448003e-05}
{"train_loss": 0.45128998160362244, "global_step": 31051, "epoch": 348, "lr": 9.689548023873397e-05}
{"train_loss": 0.3833214044570923, "global_step": 31052, "epoch": 348, "lr": 9.68952791366834e-05}
{"train_loss": 0.39775392413139343, "global_step": 31053, "epoch": 348, "lr": 9.689507802832832e-05}
{"train_loss": 0.3877147138118744, "global_step": 31054, "epoch": 348, "lr": 9.689487691366875e-05}
{"train_loss": 0.4297460615634918, "global_step": 31055, "epoch": 348, "lr": 9.689467579270475e-05}
{"train_loss": 0.39364832639694214, "global_step": 31056, "epoch": 348, "lr": 9.689447466543633e-05}
{"train_loss": 0.5110441446304321, "global_step": 31057, "epoch": 348, "lr": 9.689427353186354e-05}
{"train_loss": 0.3213461637496948, "global_step": 31058, "epoch": 348, "lr": 9.689407239198635e-05}
{"train_loss": 0.36450645327568054, "global_step": 31059, "epoch": 348, "lr": 9.689387124580484e-05}
{"train_loss": 0.3810689094361295, "global_step": 31060, "epoch": 348, "lr": 9.689367009331902e-05, "val_loss": 2.2183191776275635}
{"train_loss": 0.39455273747444153, "global_step": 31061, "epoch": 349, "lr": 9.689346893452893e-05}
{"train_loss": 0.3253210783004761, "global_step": 31062, "epoch": 349, "lr": 9.689326776943456e-05}
{"train_loss": 0.3567618727684021, "global_step": 31063, "epoch": 349, "lr": 9.689306659803599e-05}
{"train_loss": 0.3121885061264038, "global_step": 31064, "epoch": 349, "lr": 9.68928654203332e-05}
{"train_loss": 0.3512989282608032, "global_step": 31065, "epoch": 349, "lr": 9.689266423632625e-05}
{"train_loss": 0.35467374324798584, "global_step": 31066, "epoch": 349, "lr": 9.689246304601515e-05}
{"train_loss": 0.41936159133911133, "global_step": 31067, "epoch": 349, "lr": 9.689226184939992e-05}
{"train_loss": 0.29294833540916443, "global_step": 31068, "epoch": 349, "lr": 9.68920606464806e-05}
{"train_loss": 0.37846818566322327, "global_step": 31069, "epoch": 349, "lr": 9.689185943725722e-05}
{"train_loss": 0.4129527807235718, "global_step": 31070, "epoch": 349, "lr": 9.689165822172978e-05}
{"train_loss": 0.3168230354785919, "global_step": 31071, "epoch": 349, "lr": 9.689145699989835e-05}
{"train_loss": 0.4793929159641266, "global_step": 31072, "epoch": 349, "lr": 9.689125577176293e-05}
{"train_loss": 0.372744619846344, "global_step": 31073, "epoch": 349, "lr": 9.689105453732356e-05}
{"train_loss": 0.44533440470695496, "global_step": 31074, "epoch": 349, "lr": 9.689085329658025e-05}
{"train_loss": 0.28107911348342896, "global_step": 31075, "epoch": 349, "lr": 9.689065204953302e-05}
{"train_loss": 0.355043888092041, "global_step": 31076, "epoch": 349, "lr": 9.689045079618195e-05}
{"train_loss": 0.29207274317741394, "global_step": 31077, "epoch": 349, "lr": 9.689024953652701e-05}
{"train_loss": 0.37707552313804626, "global_step": 31078, "epoch": 349, "lr": 9.689004827056825e-05}
{"train_loss": 0.32160684466362, "global_step": 31079, "epoch": 349, "lr": 9.688984699830568e-05}
{"train_loss": 0.3325124979019165, "global_step": 31080, "epoch": 349, "lr": 9.688964571973936e-05}
{"train_loss": 0.5395063757896423, "global_step": 31081, "epoch": 349, "lr": 9.688944443486929e-05}
{"train_loss": 0.40708544850349426, "global_step": 31082, "epoch": 349, "lr": 9.688924314369551e-05}
{"train_loss": 0.36277368664741516, "global_step": 31083, "epoch": 349, "lr": 9.688904184621805e-05}
{"train_loss": 0.41787898540496826, "global_step": 31084, "epoch": 349, "lr": 9.688884054243692e-05}
{"train_loss": 0.47832581400871277, "global_step": 31085, "epoch": 349, "lr": 9.688863923235215e-05}
{"train_loss": 0.45129838585853577, "global_step": 31086, "epoch": 349, "lr": 9.688843791596378e-05}
{"train_loss": 0.4425520896911621, "global_step": 31087, "epoch": 349, "lr": 9.688823659327184e-05}
{"train_loss": 0.3273012638092041, "global_step": 31088, "epoch": 349, "lr": 9.688803526427635e-05}
{"train_loss": 0.2682018578052521, "global_step": 31089, "epoch": 349, "lr": 9.688783392897731e-05}
{"train_loss": 0.30751341581344604, "global_step": 31090, "epoch": 349, "lr": 9.688763258737479e-05}
{"train_loss": 0.3880757689476013, "global_step": 31091, "epoch": 349, "lr": 9.68874312394688e-05}
{"train_loss": 0.3601209223270416, "global_step": 31092, "epoch": 349, "lr": 9.688722988525937e-05}
{"train_loss": 0.34610238671302795, "global_step": 31093, "epoch": 349, "lr": 9.688702852474651e-05}
{"train_loss": 0.551504909992218, "global_step": 31094, "epoch": 349, "lr": 9.688682715793026e-05}
{"train_loss": 0.4423437714576721, "global_step": 31095, "epoch": 349, "lr": 9.688662578481067e-05}
{"train_loss": 0.3551454544067383, "global_step": 31096, "epoch": 349, "lr": 9.688642440538773e-05}
{"train_loss": 0.2302991896867752, "global_step": 31097, "epoch": 349, "lr": 9.688622301966149e-05}
{"train_loss": 0.33334460854530334, "global_step": 31098, "epoch": 349, "lr": 9.688602162763196e-05}
{"train_loss": 0.3875635862350464, "global_step": 31099, "epoch": 349, "lr": 9.688582022929917e-05}
{"train_loss": 0.4324083924293518, "global_step": 31100, "epoch": 349, "lr": 9.688561882466316e-05}
{"train_loss": 0.32767796516418457, "global_step": 31101, "epoch": 349, "lr": 9.688541741372394e-05}
{"train_loss": 0.4647604525089264, "global_step": 31102, "epoch": 349, "lr": 9.688521599648156e-05}
{"train_loss": 0.436013787984848, "global_step": 31103, "epoch": 349, "lr": 9.688501457293604e-05}
{"train_loss": 0.34594786167144775, "global_step": 31104, "epoch": 349, "lr": 9.68848131430874e-05}
{"train_loss": 0.33551251888275146, "global_step": 31105, "epoch": 349, "lr": 9.688461170693565e-05}
{"train_loss": 0.3001824915409088, "global_step": 31106, "epoch": 349, "lr": 9.688441026448084e-05}
{"train_loss": 0.45749807357788086, "global_step": 31107, "epoch": 349, "lr": 9.6884208815723e-05}
{"train_loss": 0.333244264125824, "global_step": 31108, "epoch": 349, "lr": 9.688400736066214e-05}
{"train_loss": 0.3709352910518646, "global_step": 31109, "epoch": 349, "lr": 9.68838058992983e-05}
{"train_loss": 0.4055641293525696, "global_step": 31110, "epoch": 349, "lr": 9.688360443163151e-05}
{"train_loss": 0.291448712348938, "global_step": 31111, "epoch": 349, "lr": 9.68834029576618e-05}
{"train_loss": 0.38859373331069946, "global_step": 31112, "epoch": 349, "lr": 9.688320147738918e-05}
{"train_loss": 0.4420641362667084, "global_step": 31113, "epoch": 349, "lr": 9.688299999081368e-05}
{"train_loss": 0.4248787462711334, "global_step": 31114, "epoch": 349, "lr": 9.688279849793533e-05}
{"train_loss": 0.4988184869289398, "global_step": 31115, "epoch": 349, "lr": 9.688259699875417e-05}
{"train_loss": 0.2787020802497864, "global_step": 31116, "epoch": 349, "lr": 9.688239549327021e-05}
{"train_loss": 0.44403016567230225, "global_step": 31117, "epoch": 349, "lr": 9.688219398148348e-05}
{"train_loss": 0.2808656096458435, "global_step": 31118, "epoch": 349, "lr": 9.688199246339402e-05}
{"train_loss": 0.4119356870651245, "global_step": 31119, "epoch": 349, "lr": 9.688179093900183e-05}
{"train_loss": 0.3043728470802307, "global_step": 31120, "epoch": 349, "lr": 9.688158940830696e-05}
{"train_loss": 0.38176754117012024, "global_step": 31121, "epoch": 349, "lr": 9.688138787130945e-05}
{"train_loss": 0.3591025769710541, "global_step": 31122, "epoch": 349, "lr": 9.688118632800929e-05}
{"train_loss": 0.4214581549167633, "global_step": 31123, "epoch": 349, "lr": 9.688098477840652e-05}
{"train_loss": 0.2881348431110382, "global_step": 31124, "epoch": 349, "lr": 9.68807832225012e-05}
{"train_loss": 0.4076922833919525, "global_step": 31125, "epoch": 349, "lr": 9.68805816602933e-05}
{"train_loss": 0.36119943857192993, "global_step": 31126, "epoch": 349, "lr": 9.688038009178291e-05}
{"train_loss": 0.40436890721321106, "global_step": 31127, "epoch": 349, "lr": 9.688017851697e-05}
{"train_loss": 0.26218941807746887, "global_step": 31128, "epoch": 349, "lr": 9.687997693585462e-05}
{"train_loss": 0.37324249744415283, "global_step": 31129, "epoch": 349, "lr": 9.687977534843681e-05}
{"train_loss": 0.35267478227615356, "global_step": 31130, "epoch": 349, "lr": 9.68795737547166e-05}
{"train_loss": 0.46044713258743286, "global_step": 31131, "epoch": 349, "lr": 9.687937215469396e-05}
{"train_loss": 0.3834860324859619, "global_step": 31132, "epoch": 349, "lr": 9.6879170548369e-05}
{"train_loss": 0.36866581439971924, "global_step": 31133, "epoch": 349, "lr": 9.687896893574169e-05}
{"train_loss": 0.33674347400665283, "global_step": 31134, "epoch": 349, "lr": 9.687876731681208e-05}
{"train_loss": 0.2589115500450134, "global_step": 31135, "epoch": 349, "lr": 9.687856569158018e-05}
{"train_loss": 0.3847080171108246, "global_step": 31136, "epoch": 349, "lr": 9.687836406004603e-05}
{"train_loss": 0.3644356429576874, "global_step": 31137, "epoch": 349, "lr": 9.687816242220966e-05}
{"train_loss": 0.373415470123291, "global_step": 31138, "epoch": 349, "lr": 9.68779607780711e-05}
{"train_loss": 0.39139318466186523, "global_step": 31139, "epoch": 349, "lr": 9.687775912763035e-05}
{"train_loss": 0.35570260882377625, "global_step": 31140, "epoch": 349, "lr": 9.687755747088747e-05}
{"train_loss": 0.2522379755973816, "global_step": 31141, "epoch": 349, "lr": 9.687735580784247e-05}
{"train_loss": 0.26975148916244507, "global_step": 31142, "epoch": 349, "lr": 9.687715413849538e-05}
{"train_loss": 0.4694872498512268, "global_step": 31143, "epoch": 349, "lr": 9.687695246284624e-05}
{"train_loss": 0.36565670371055603, "global_step": 31144, "epoch": 349, "lr": 9.687675078089505e-05}
{"train_loss": 0.4176848530769348, "global_step": 31145, "epoch": 349, "lr": 9.687654909264185e-05}
{"train_loss": 0.43038278818130493, "global_step": 31146, "epoch": 349, "lr": 9.687634739808668e-05}
{"train_loss": 0.42641398310661316, "global_step": 31147, "epoch": 349, "lr": 9.687614569722956e-05}
{"train_loss": 0.369720995426178, "global_step": 31148, "epoch": 349, "lr": 9.687594399007052e-05}
{"train_loss": 0.37281811588935637, "global_step": 31149, "epoch": 349, "lr": 9.687574227660956e-05, "val_loss": 2.36944842338562}
{"train_loss": 0.48271095752716064, "global_step": 31150, "epoch": 350, "lr": 9.687554055684673e-05}
{"train_loss": 0.34114891290664673, "global_step": 31151, "epoch": 350, "lr": 9.687533883078206e-05}
{"train_loss": 0.49301254749298096, "global_step": 31152, "epoch": 350, "lr": 9.687513709841558e-05}
{"train_loss": 0.3586852252483368, "global_step": 31153, "epoch": 350, "lr": 9.687493535974729e-05}
{"train_loss": 0.412885844707489, "global_step": 31154, "epoch": 350, "lr": 9.687473361477725e-05}
{"train_loss": 0.3640664219856262, "global_step": 31155, "epoch": 350, "lr": 9.687453186350548e-05}
{"train_loss": 0.4469548463821411, "global_step": 31156, "epoch": 350, "lr": 9.687433010593199e-05}
{"train_loss": 0.2973915934562683, "global_step": 31157, "epoch": 350, "lr": 9.687412834205683e-05}
{"train_loss": 0.4543777406215668, "global_step": 31158, "epoch": 350, "lr": 9.687392657187999e-05}
{"train_loss": 0.44112175703048706, "global_step": 31159, "epoch": 350, "lr": 9.687372479540155e-05}
{"train_loss": 0.371168851852417, "global_step": 31160, "epoch": 350, "lr": 9.68735230126215e-05}
{"train_loss": 0.4693846106529236, "global_step": 31161, "epoch": 350, "lr": 9.687332122353985e-05}
{"train_loss": 0.3349649906158447, "global_step": 31162, "epoch": 350, "lr": 9.687311942815669e-05}
{"train_loss": 0.3731227517127991, "global_step": 31163, "epoch": 350, "lr": 9.687291762647199e-05}
{"train_loss": 0.30369478464126587, "global_step": 31164, "epoch": 350, "lr": 9.68727158184858e-05}
{"train_loss": 0.44150063395500183, "global_step": 31165, "epoch": 350, "lr": 9.687251400419815e-05}
{"train_loss": 0.3180338144302368, "global_step": 31166, "epoch": 350, "lr": 9.687231218360906e-05}
{"train_loss": 0.42147010564804077, "global_step": 31167, "epoch": 350, "lr": 9.687211035671856e-05}
{"train_loss": 0.38681769371032715, "global_step": 31168, "epoch": 350, "lr": 9.687190852352668e-05}
{"train_loss": 0.30805668234825134, "global_step": 31169, "epoch": 350, "lr": 9.687170668403342e-05}
{"train_loss": 0.4208192825317383, "global_step": 31170, "epoch": 350, "lr": 9.687150483823885e-05}
{"train_loss": 0.40127646923065186, "global_step": 31171, "epoch": 350, "lr": 9.687130298614297e-05}
{"train_loss": 0.37517884373664856, "global_step": 31172, "epoch": 350, "lr": 9.687110112774583e-05}
{"train_loss": 0.38126468658447266, "global_step": 31173, "epoch": 350, "lr": 9.687089926304744e-05}
{"train_loss": 0.3123006224632263, "global_step": 31174, "epoch": 350, "lr": 9.687069739204781e-05}
{"train_loss": 0.359607994556427, "global_step": 31175, "epoch": 350, "lr": 9.6870495514747e-05}
{"train_loss": 0.31255143880844116, "global_step": 31176, "epoch": 350, "lr": 9.687029363114502e-05}
{"train_loss": 0.31792792677879333, "global_step": 31177, "epoch": 350, "lr": 9.68700917412419e-05}
{"train_loss": 0.32801273465156555, "global_step": 31178, "epoch": 350, "lr": 9.686988984503766e-05}
{"train_loss": 0.35604605078697205, "global_step": 31179, "epoch": 350, "lr": 9.686968794253235e-05}
{"train_loss": 0.3622013032436371, "global_step": 31180, "epoch": 350, "lr": 9.686948603372596e-05}
{"train_loss": 0.38082584738731384, "global_step": 31181, "epoch": 350, "lr": 9.686928411861856e-05}
{"train_loss": 0.3420644700527191, "global_step": 31182, "epoch": 350, "lr": 9.686908219721016e-05}
{"train_loss": 0.22887006402015686, "global_step": 31183, "epoch": 350, "lr": 9.686888026950076e-05}
{"train_loss": 0.42557498812675476, "global_step": 31184, "epoch": 350, "lr": 9.686867833549042e-05}
{"train_loss": 0.3062436282634735, "global_step": 31185, "epoch": 350, "lr": 9.686847639517916e-05}
{"train_loss": 0.30605027079582214, "global_step": 31186, "epoch": 350, "lr": 9.6868274448567e-05}
{"train_loss": 0.33710965514183044, "global_step": 31187, "epoch": 350, "lr": 9.686807249565398e-05}
{"train_loss": 0.365963339805603, "global_step": 31188, "epoch": 350, "lr": 9.686787053644011e-05}
{"train_loss": 0.2703467309474945, "global_step": 31189, "epoch": 350, "lr": 9.686766857092543e-05}
{"train_loss": 0.42782047390937805, "global_step": 31190, "epoch": 350, "lr": 9.686746659910997e-05}
{"train_loss": 0.28591111302375793, "global_step": 31191, "epoch": 350, "lr": 9.686726462099374e-05}
{"train_loss": 0.28623273968696594, "global_step": 31192, "epoch": 350, "lr": 9.686706263657679e-05}
{"train_loss": 0.3975357115268707, "global_step": 31193, "epoch": 350, "lr": 9.686686064585912e-05}
{"train_loss": 0.2795579433441162, "global_step": 31194, "epoch": 350, "lr": 9.68666586488408e-05}
{"train_loss": 0.36151084303855896, "global_step": 31195, "epoch": 350, "lr": 9.686645664552179e-05}
{"train_loss": 0.32127171754837036, "global_step": 31196, "epoch": 350, "lr": 9.686625463590217e-05}
{"train_loss": 0.36612120270729065, "global_step": 31197, "epoch": 350, "lr": 9.686605261998196e-05}
{"train_loss": 0.28735753893852234, "global_step": 31198, "epoch": 350, "lr": 9.686585059776117e-05}
{"train_loss": 0.29658547043800354, "global_step": 31199, "epoch": 350, "lr": 9.686564856923985e-05}
{"train_loss": 0.4219991862773895, "global_step": 31200, "epoch": 350, "lr": 9.686544653441802e-05}
{"train_loss": 0.39883995056152344, "global_step": 31201, "epoch": 350, "lr": 9.686524449329569e-05}
{"train_loss": 0.4307206869125366, "global_step": 31202, "epoch": 350, "lr": 9.68650424458729e-05}
{"train_loss": 0.28376221656799316, "global_step": 31203, "epoch": 350, "lr": 9.686484039214967e-05}
{"train_loss": 0.4074467420578003, "global_step": 31204, "epoch": 350, "lr": 9.686463833212605e-05}
{"train_loss": 0.444401353597641, "global_step": 31205, "epoch": 350, "lr": 9.686443626580203e-05}
{"train_loss": 0.5125352144241333, "global_step": 31206, "epoch": 350, "lr": 9.686423419317767e-05}
{"train_loss": 0.505700945854187, "global_step": 31207, "epoch": 350, "lr": 9.686403211425299e-05}
{"train_loss": 0.35407447814941406, "global_step": 31208, "epoch": 350, "lr": 9.6863830029028e-05}
{"train_loss": 0.2682386338710785, "global_step": 31209, "epoch": 350, "lr": 9.686362793750275e-05}
{"train_loss": 0.3277027904987335, "global_step": 31210, "epoch": 350, "lr": 9.686342583967725e-05}
{"train_loss": 0.3384491205215454, "global_step": 31211, "epoch": 350, "lr": 9.686322373555153e-05}
{"train_loss": 0.5201489925384521, "global_step": 31212, "epoch": 350, "lr": 9.686302162512561e-05}
{"train_loss": 0.4538165032863617, "global_step": 31213, "epoch": 350, "lr": 9.686281950839954e-05}
{"train_loss": 0.4508575201034546, "global_step": 31214, "epoch": 350, "lr": 9.686261738537336e-05}
{"train_loss": 0.46697577834129333, "global_step": 31215, "epoch": 350, "lr": 9.686241525604705e-05}
{"train_loss": 0.4028305113315582, "global_step": 31216, "epoch": 350, "lr": 9.686221312042065e-05}
{"train_loss": 0.3808874785900116, "global_step": 31217, "epoch": 350, "lr": 9.68620109784942e-05}
{"train_loss": 0.4472685754299164, "global_step": 31218, "epoch": 350, "lr": 9.686180883026774e-05}
{"train_loss": 0.455517441034317, "global_step": 31219, "epoch": 350, "lr": 9.686160667574127e-05}
{"train_loss": 0.46240144968032837, "global_step": 31220, "epoch": 350, "lr": 9.686140451491483e-05}
{"train_loss": 0.4233003258705139, "global_step": 31221, "epoch": 350, "lr": 9.686120234778845e-05}
{"train_loss": 0.3686385154724121, "global_step": 31222, "epoch": 350, "lr": 9.686100017436215e-05}
{"train_loss": 0.35462430119514465, "global_step": 31223, "epoch": 350, "lr": 9.686079799463595e-05}
{"train_loss": 0.3778849244117737, "global_step": 31224, "epoch": 350, "lr": 9.68605958086099e-05}
{"train_loss": 0.43693527579307556, "global_step": 31225, "epoch": 350, "lr": 9.686039361628401e-05}
{"train_loss": 0.4595767557621002, "global_step": 31226, "epoch": 350, "lr": 9.68601914176583e-05}
{"train_loss": 0.41554534435272217, "global_step": 31227, "epoch": 350, "lr": 9.685998921273282e-05}
{"train_loss": 0.3563116788864136, "global_step": 31228, "epoch": 350, "lr": 9.685978700150758e-05}
{"train_loss": 0.3464111089706421, "global_step": 31229, "epoch": 350, "lr": 9.685958478398263e-05}
{"train_loss": 0.3930508494377136, "global_step": 31230, "epoch": 350, "lr": 9.685938256015797e-05}
{"train_loss": 0.4463229179382324, "global_step": 31231, "epoch": 350, "lr": 9.685918033003362e-05}
{"train_loss": 0.4789363145828247, "global_step": 31232, "epoch": 350, "lr": 9.685897809360964e-05}
{"train_loss": 0.4129035770893097, "global_step": 31233, "epoch": 350, "lr": 9.685877585088603e-05}
{"train_loss": 0.33629217743873596, "global_step": 31234, "epoch": 350, "lr": 9.685857360186284e-05}
{"train_loss": 0.3953903913497925, "global_step": 31235, "epoch": 350, "lr": 9.685837134654009e-05}
{"train_loss": 0.3665962517261505, "global_step": 31236, "epoch": 350, "lr": 9.685816908491779e-05}
{"train_loss": 0.5262651443481445, "global_step": 31237, "epoch": 350, "lr": 9.6857966816996e-05}
{"train_loss": 0.38103252147020916, "global_step": 31238, "epoch": 350, "lr": 9.685776454277471e-05, "train/sim_max_reward_0": 0.6533966331009043, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.621290238219461, "train/sim_max_reward_3": 0.6748708146906442, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.4809740854465611, "test/sim_max_reward_4500000": 0.9756490582845931, "test/sim_max_reward_4500001": 0.9363906074995669, "test/sim_max_reward_4500002": 0.9683033166916192, "test/sim_max_reward_4500003": 0.6511279754765815, "test/sim_max_reward_4500004": 0.8045694959437665, "test/sim_max_reward_4500005": 0.9391897983930942, "test/sim_max_reward_4500006": 0.7211671284219919, "test/sim_max_reward_4500007": 0.6175773233261752, "test/sim_max_reward_4500008": 0.33691614659564706, "test/sim_max_reward_4500009": 1.0, "test/sim_max_reward_4500010": 0.9469452666892556, "test/sim_max_reward_4500011": 0.8145834806006063, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5349426930163563, "test/sim_max_reward_4500014": 0.41878810230382657, "test/sim_max_reward_4500015": 0.5655219876215833, "test/sim_max_reward_4500016": 0.47010331974338126, "test/sim_max_reward_4500017": 0.7031597766660519, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9579123921677463, "test/sim_max_reward_4500020": 0.965122467899318, "test/sim_max_reward_4500021": 0.9426415984240566, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6761631204447325, "test/sim_max_reward_4500024": 0.5909842470756176, "test/sim_max_reward_4500025": 0.7249134443185165, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 0.9498464873522551, "test/sim_max_reward_4500028": 0.6282080942927539, "test/sim_max_reward_4500029": 0.061153332114594616, "test/sim_max_reward_4500030": 0.9895883728429747, "test/sim_max_reward_4500031": 1.0, "test/sim_max_reward_4500032": 0.295179008645352, "test/sim_max_reward_4500033": 0.9165438391499274, "test/sim_max_reward_4500034": 0.6735528073051936, "test/sim_max_reward_4500035": 0.7065320103572281, "test/sim_max_reward_4500036": 0.961548517859797, "test/sim_max_reward_4500037": 0.42613248730580644, "test/sim_max_reward_4500038": 0.7774312823182941, "test/sim_max_reward_4500039": 0.5664596564841345, "test/sim_max_reward_4500040": 0.657888507489113, "test/sim_max_reward_4500041": 0.6404827951252654, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9979329019933939, "test/sim_max_reward_4500044": 0.6789436845609013, "test/sim_max_reward_4500045": 0.6782840678394229, "test/sim_max_reward_4500046": 0.5101234623535593, "test/sim_max_reward_4500047": 0.38895806637255037, "test/sim_max_reward_4500048": 0.924176434856789, "test/sim_max_reward_4500049": 0.6887936726128749, "train/mean_score": 0.7384219619095952, "test/mean_score": 0.7076086447367252, "val_loss": 2.2910194396972656, "train_action_mse_error": 28.92779541015625}
{"train_loss": 0.43447554111480713, "global_step": 31239, "epoch": 351, "lr": 9.685756226225397e-05}
{"train_loss": 0.34864553809165955, "global_step": 31240, "epoch": 351, "lr": 9.685735997543379e-05}
{"train_loss": 0.425933837890625, "global_step": 31241, "epoch": 351, "lr": 9.685715768231423e-05}
{"train_loss": 0.5288342833518982, "global_step": 31242, "epoch": 351, "lr": 9.685695538289528e-05}
{"train_loss": 0.40399956703186035, "global_step": 31243, "epoch": 351, "lr": 9.685675307717699e-05}
{"train_loss": 0.38084232807159424, "global_step": 31244, "epoch": 351, "lr": 9.685655076515939e-05}
{"train_loss": 0.4793477952480316, "global_step": 31245, "epoch": 351, "lr": 9.685634844684249e-05}
{"train_loss": 0.3764949142932892, "global_step": 31246, "epoch": 351, "lr": 9.685614612222631e-05}
{"train_loss": 0.4768702983856201, "global_step": 31247, "epoch": 351, "lr": 9.685594379131091e-05}
{"train_loss": 0.30061838030815125, "global_step": 31248, "epoch": 351, "lr": 9.685574145409629e-05}
{"train_loss": 0.37142547965049744, "global_step": 31249, "epoch": 351, "lr": 9.685553911058249e-05}
{"train_loss": 0.3951750099658966, "global_step": 31250, "epoch": 351, "lr": 9.685533676076955e-05}
{"train_loss": 0.33628395199775696, "global_step": 31251, "epoch": 351, "lr": 9.685513440465745e-05}
{"train_loss": 0.44236648082733154, "global_step": 31252, "epoch": 351, "lr": 9.685493204224627e-05}
{"train_loss": 0.34712180495262146, "global_step": 31253, "epoch": 351, "lr": 9.685472967353602e-05}
{"train_loss": 0.3551041781902313, "global_step": 31254, "epoch": 351, "lr": 9.68545272985267e-05}
{"train_loss": 0.3632623255252838, "global_step": 31255, "epoch": 351, "lr": 9.685432491721838e-05}
{"train_loss": 0.3325321674346924, "global_step": 31256, "epoch": 351, "lr": 9.685412252961106e-05}
{"train_loss": 0.395006388425827, "global_step": 31257, "epoch": 351, "lr": 9.685392013570477e-05}
{"train_loss": 0.4052734076976776, "global_step": 31258, "epoch": 351, "lr": 9.685371773549955e-05}
{"train_loss": 0.32601654529571533, "global_step": 31259, "epoch": 351, "lr": 9.685351532899542e-05}
{"train_loss": 0.3647812008857727, "global_step": 31260, "epoch": 351, "lr": 9.685331291619239e-05}
{"train_loss": 0.31296274065971375, "global_step": 31261, "epoch": 351, "lr": 9.685311049709051e-05}
{"train_loss": 0.43044695258140564, "global_step": 31262, "epoch": 351, "lr": 9.685290807168981e-05}
{"train_loss": 0.3756542503833771, "global_step": 31263, "epoch": 351, "lr": 9.68527056399903e-05}
{"train_loss": 0.28294795751571655, "global_step": 31264, "epoch": 351, "lr": 9.685250320199201e-05}
{"train_loss": 0.30147555470466614, "global_step": 31265, "epoch": 351, "lr": 9.685230075769498e-05}
{"train_loss": 0.39188799262046814, "global_step": 31266, "epoch": 351, "lr": 9.685209830709924e-05}
{"train_loss": 0.4073950946331024, "global_step": 31267, "epoch": 351, "lr": 9.68518958502048e-05}
{"train_loss": 0.3901576101779938, "global_step": 31268, "epoch": 351, "lr": 9.685169338701168e-05}
{"train_loss": 0.4232262074947357, "global_step": 31269, "epoch": 351, "lr": 9.685149091751993e-05}
{"train_loss": 0.3369189202785492, "global_step": 31270, "epoch": 351, "lr": 9.685128844172957e-05}
{"train_loss": 0.3225829005241394, "global_step": 31271, "epoch": 351, "lr": 9.685108595964061e-05}
{"train_loss": 0.3293476402759552, "global_step": 31272, "epoch": 351, "lr": 9.68508834712531e-05}
{"train_loss": 0.34685784578323364, "global_step": 31273, "epoch": 351, "lr": 9.685068097656707e-05}
{"train_loss": 0.4504415988922119, "global_step": 31274, "epoch": 351, "lr": 9.685047847558253e-05}
{"train_loss": 0.38862067461013794, "global_step": 31275, "epoch": 351, "lr": 9.685027596829952e-05}
{"train_loss": 0.3800271153450012, "global_step": 31276, "epoch": 351, "lr": 9.685007345471807e-05}
{"train_loss": 0.36257484555244446, "global_step": 31277, "epoch": 351, "lr": 9.684987093483817e-05}
{"train_loss": 0.333319753408432, "global_step": 31278, "epoch": 351, "lr": 9.68496684086599e-05}
{"train_loss": 0.2942814826965332, "global_step": 31279, "epoch": 351, "lr": 9.684946587618325e-05}
{"train_loss": 0.5054426789283752, "global_step": 31280, "epoch": 351, "lr": 9.684926333740826e-05}
{"train_loss": 0.32797494530677795, "global_step": 31281, "epoch": 351, "lr": 9.684906079233496e-05}
{"train_loss": 0.3621099591255188, "global_step": 31282, "epoch": 351, "lr": 9.684885824096338e-05}
{"train_loss": 0.2644740045070648, "global_step": 31283, "epoch": 351, "lr": 9.684865568329353e-05}
{"train_loss": 0.44444185495376587, "global_step": 31284, "epoch": 351, "lr": 9.684845311932547e-05}
{"train_loss": 0.37891101837158203, "global_step": 31285, "epoch": 351, "lr": 9.684825054905919e-05}
{"train_loss": 0.27948492765426636, "global_step": 31286, "epoch": 351, "lr": 9.684804797249474e-05}
{"train_loss": 0.3464411199092865, "global_step": 31287, "epoch": 351, "lr": 9.684784538963213e-05}
{"train_loss": 0.38621312379837036, "global_step": 31288, "epoch": 351, "lr": 9.684764280047141e-05}
{"train_loss": 0.2631162106990814, "global_step": 31289, "epoch": 351, "lr": 9.68474402050126e-05}
{"train_loss": 0.3961615562438965, "global_step": 31290, "epoch": 351, "lr": 9.684723760325572e-05}
{"train_loss": 0.3859408497810364, "global_step": 31291, "epoch": 351, "lr": 9.684703499520079e-05}
{"train_loss": 0.34011468291282654, "global_step": 31292, "epoch": 351, "lr": 9.684683238084786e-05}
{"train_loss": 0.4512600302696228, "global_step": 31293, "epoch": 351, "lr": 9.684662976019693e-05}
{"train_loss": 0.35044825077056885, "global_step": 31294, "epoch": 351, "lr": 9.684642713324806e-05}
{"train_loss": 0.3677210211753845, "global_step": 31295, "epoch": 351, "lr": 9.684622450000123e-05}
{"train_loss": 0.2325882613658905, "global_step": 31296, "epoch": 351, "lr": 9.684602186045652e-05}
{"train_loss": 0.3170258402824402, "global_step": 31297, "epoch": 351, "lr": 9.684581921461394e-05}
{"train_loss": 0.3467065989971161, "global_step": 31298, "epoch": 351, "lr": 9.684561656247348e-05}
{"train_loss": 0.3527984619140625, "global_step": 31299, "epoch": 351, "lr": 9.684541390403522e-05}
{"train_loss": 0.39017990231513977, "global_step": 31300, "epoch": 351, "lr": 9.684521123929916e-05}
{"train_loss": 0.40052151679992676, "global_step": 31301, "epoch": 351, "lr": 9.684500856826533e-05}
{"train_loss": 0.33798059821128845, "global_step": 31302, "epoch": 351, "lr": 9.684480589093376e-05}
{"train_loss": 0.3196621537208557, "global_step": 31303, "epoch": 351, "lr": 9.684460320730448e-05}
{"train_loss": 0.39229103922843933, "global_step": 31304, "epoch": 351, "lr": 9.68444005173775e-05}
{"train_loss": 0.46032825112342834, "global_step": 31305, "epoch": 351, "lr": 9.684419782115289e-05}
{"train_loss": 0.48216742277145386, "global_step": 31306, "epoch": 351, "lr": 9.684399511863063e-05}
{"train_loss": 0.48951250314712524, "global_step": 31307, "epoch": 351, "lr": 9.684379240981078e-05}
{"train_loss": 0.288544625043869, "global_step": 31308, "epoch": 351, "lr": 9.684358969469334e-05}
{"train_loss": 0.40574681758880615, "global_step": 31309, "epoch": 351, "lr": 9.684338697327835e-05}
{"train_loss": 0.3248802721500397, "global_step": 31310, "epoch": 351, "lr": 9.684318424556585e-05}
{"train_loss": 0.3903482258319855, "global_step": 31311, "epoch": 351, "lr": 9.684298151155583e-05}
{"train_loss": 0.30325189232826233, "global_step": 31312, "epoch": 351, "lr": 9.684277877124837e-05}
{"train_loss": 0.3527379631996155, "global_step": 31313, "epoch": 351, "lr": 9.684257602464344e-05}
{"train_loss": 0.3349463939666748, "global_step": 31314, "epoch": 351, "lr": 9.684237327174112e-05}
{"train_loss": 0.4299376904964447, "global_step": 31315, "epoch": 351, "lr": 9.68421705125414e-05}
{"train_loss": 0.4120672643184662, "global_step": 31316, "epoch": 351, "lr": 9.684196774704433e-05}
{"train_loss": 0.29302525520324707, "global_step": 31317, "epoch": 351, "lr": 9.684176497524992e-05}
{"train_loss": 0.3340069353580475, "global_step": 31318, "epoch": 351, "lr": 9.684156219715822e-05}
{"train_loss": 0.2761865556240082, "global_step": 31319, "epoch": 351, "lr": 9.684135941276924e-05}
{"train_loss": 0.3326530158519745, "global_step": 31320, "epoch": 351, "lr": 9.6841156622083e-05}
{"train_loss": 0.4268706440925598, "global_step": 31321, "epoch": 351, "lr": 9.684095382509954e-05}
{"train_loss": 0.4167446196079254, "global_step": 31322, "epoch": 351, "lr": 9.684075102181889e-05}
{"train_loss": 0.38487523794174194, "global_step": 31323, "epoch": 351, "lr": 9.684054821224107e-05}
{"train_loss": 0.3344092071056366, "global_step": 31324, "epoch": 351, "lr": 9.68403453963661e-05}
{"train_loss": 0.435007780790329, "global_step": 31325, "epoch": 351, "lr": 9.684014257419403e-05}
{"train_loss": 0.35869401693344116, "global_step": 31326, "epoch": 351, "lr": 9.683993974572487e-05}
{"train_loss": 0.3709735679492522, "global_step": 31327, "epoch": 351, "lr": 9.683973691095864e-05, "val_loss": 2.3151581287384033}
{"train_loss": 0.36764195561408997, "global_step": 31328, "epoch": 352, "lr": 9.683953406989539e-05}
{"train_loss": 0.317013144493103, "global_step": 31329, "epoch": 352, "lr": 9.683933122253513e-05}
{"train_loss": 0.3431970477104187, "global_step": 31330, "epoch": 352, "lr": 9.683912836887789e-05}
{"train_loss": 0.38807085156440735, "global_step": 31331, "epoch": 352, "lr": 9.68389255089237e-05}
{"train_loss": 0.32162588834762573, "global_step": 31332, "epoch": 352, "lr": 9.68387226426726e-05}
{"train_loss": 0.33186212182044983, "global_step": 31333, "epoch": 352, "lr": 9.68385197701246e-05}
{"train_loss": 0.3667435944080353, "global_step": 31334, "epoch": 352, "lr": 9.683831689127973e-05}
{"train_loss": 0.4031658470630646, "global_step": 31335, "epoch": 352, "lr": 9.683811400613802e-05}
{"train_loss": 0.36659538745880127, "global_step": 31336, "epoch": 352, "lr": 9.683791111469948e-05}
{"train_loss": 0.3004418909549713, "global_step": 31337, "epoch": 352, "lr": 9.683770821696418e-05}
{"train_loss": 0.4276215732097626, "global_step": 31338, "epoch": 352, "lr": 9.683750531293212e-05}
{"train_loss": 0.28484830260276794, "global_step": 31339, "epoch": 352, "lr": 9.683730240260332e-05}
{"train_loss": 0.3197484016418457, "global_step": 31340, "epoch": 352, "lr": 9.683709948597781e-05}
{"train_loss": 0.3287489414215088, "global_step": 31341, "epoch": 352, "lr": 9.683689656305563e-05}
{"train_loss": 0.3213198781013489, "global_step": 31342, "epoch": 352, "lr": 9.683669363383679e-05}
{"train_loss": 0.39913347363471985, "global_step": 31343, "epoch": 352, "lr": 9.683649069832132e-05}
{"train_loss": 0.3443055748939514, "global_step": 31344, "epoch": 352, "lr": 9.683628775650929e-05}
{"train_loss": 0.3316309154033661, "global_step": 31345, "epoch": 352, "lr": 9.683608480840066e-05}
{"train_loss": 0.36788204312324524, "global_step": 31346, "epoch": 352, "lr": 9.68358818539955e-05}
{"train_loss": 0.2888576090335846, "global_step": 31347, "epoch": 352, "lr": 9.683567889329384e-05}
{"train_loss": 0.345925897359848, "global_step": 31348, "epoch": 352, "lr": 9.683547592629568e-05}
{"train_loss": 0.3480762541294098, "global_step": 31349, "epoch": 352, "lr": 9.683527295300105e-05}
{"train_loss": 0.26151013374328613, "global_step": 31350, "epoch": 352, "lr": 9.683506997340999e-05}
{"train_loss": 0.477144330739975, "global_step": 31351, "epoch": 352, "lr": 9.683486698752255e-05}
{"train_loss": 0.3844528794288635, "global_step": 31352, "epoch": 352, "lr": 9.683466399533872e-05}
{"train_loss": 0.280574232339859, "global_step": 31353, "epoch": 352, "lr": 9.683446099685854e-05}
{"train_loss": 0.406487375497818, "global_step": 31354, "epoch": 352, "lr": 9.683425799208203e-05}
{"train_loss": 0.4576278328895569, "global_step": 31355, "epoch": 352, "lr": 9.683405498100922e-05}
{"train_loss": 0.2950495183467865, "global_step": 31356, "epoch": 352, "lr": 9.683385196364017e-05}
{"train_loss": 0.3762543797492981, "global_step": 31357, "epoch": 352, "lr": 9.683364893997485e-05}
{"train_loss": 0.387344092130661, "global_step": 31358, "epoch": 352, "lr": 9.683344591001333e-05}
{"train_loss": 0.4247986674308777, "global_step": 31359, "epoch": 352, "lr": 9.683324287375563e-05}
{"train_loss": 0.3729867935180664, "global_step": 31360, "epoch": 352, "lr": 9.683303983120176e-05}
{"train_loss": 0.3641713857650757, "global_step": 31361, "epoch": 352, "lr": 9.683283678235177e-05}
{"train_loss": 0.31719303131103516, "global_step": 31362, "epoch": 352, "lr": 9.683263372720565e-05}
{"train_loss": 0.44175955653190613, "global_step": 31363, "epoch": 352, "lr": 9.683243066576348e-05}
{"train_loss": 0.3237833082675934, "global_step": 31364, "epoch": 352, "lr": 9.683222759802526e-05}
{"train_loss": 0.37320423126220703, "global_step": 31365, "epoch": 352, "lr": 9.6832024523991e-05}
{"train_loss": 0.4892849624156952, "global_step": 31366, "epoch": 352, "lr": 9.683182144366077e-05}
{"train_loss": 0.35036832094192505, "global_step": 31367, "epoch": 352, "lr": 9.683161835703456e-05}
{"train_loss": 0.36255815625190735, "global_step": 31368, "epoch": 352, "lr": 9.68314152641124e-05}
{"train_loss": 0.40332362055778503, "global_step": 31369, "epoch": 352, "lr": 9.683121216489433e-05}
{"train_loss": 0.42987382411956787, "global_step": 31370, "epoch": 352, "lr": 9.683100905938038e-05}
{"train_loss": 0.4106932580471039, "global_step": 31371, "epoch": 352, "lr": 9.683080594757057e-05}
{"train_loss": 0.2673729658126831, "global_step": 31372, "epoch": 352, "lr": 9.683060282946493e-05}
{"train_loss": 0.4158966839313507, "global_step": 31373, "epoch": 352, "lr": 9.683039970506349e-05}
{"train_loss": 0.3545743227005005, "global_step": 31374, "epoch": 352, "lr": 9.683019657436627e-05}
{"train_loss": 0.42483893036842346, "global_step": 31375, "epoch": 352, "lr": 9.68299934373733e-05}
{"train_loss": 0.29686617851257324, "global_step": 31376, "epoch": 352, "lr": 9.682979029408461e-05}
{"train_loss": 0.39941248297691345, "global_step": 31377, "epoch": 352, "lr": 9.682958714450022e-05}
{"train_loss": 0.4258907437324524, "global_step": 31378, "epoch": 352, "lr": 9.682938398862017e-05}
{"train_loss": 0.3288319408893585, "global_step": 31379, "epoch": 352, "lr": 9.682918082644448e-05}
{"train_loss": 0.34732693433761597, "global_step": 31380, "epoch": 352, "lr": 9.682897765797318e-05}
{"train_loss": 0.3812439441680908, "global_step": 31381, "epoch": 352, "lr": 9.682877448320629e-05}
{"train_loss": 0.41932812333106995, "global_step": 31382, "epoch": 352, "lr": 9.682857130214383e-05}
{"train_loss": 0.39318203926086426, "global_step": 31383, "epoch": 352, "lr": 9.682836811478585e-05}
{"train_loss": 0.3600431978702545, "global_step": 31384, "epoch": 352, "lr": 9.682816492113238e-05}
{"train_loss": 0.3198744058609009, "global_step": 31385, "epoch": 352, "lr": 9.68279617211834e-05}
{"train_loss": 0.3663015365600586, "global_step": 31386, "epoch": 352, "lr": 9.682775851493901e-05}
{"train_loss": 0.3828366696834564, "global_step": 31387, "epoch": 352, "lr": 9.682755530239917e-05}
{"train_loss": 0.4201463460922241, "global_step": 31388, "epoch": 352, "lr": 9.682735208356395e-05}
{"train_loss": 0.4860515594482422, "global_step": 31389, "epoch": 352, "lr": 9.682714885843336e-05}
{"train_loss": 0.26640215516090393, "global_step": 31390, "epoch": 352, "lr": 9.682694562700743e-05}
{"train_loss": 0.370398610830307, "global_step": 31391, "epoch": 352, "lr": 9.682674238928619e-05}
{"train_loss": 0.34416285157203674, "global_step": 31392, "epoch": 352, "lr": 9.682653914526967e-05}
{"train_loss": 0.39038458466529846, "global_step": 31393, "epoch": 352, "lr": 9.682633589495788e-05}
{"train_loss": 0.509596049785614, "global_step": 31394, "epoch": 352, "lr": 9.682613263835088e-05}
{"train_loss": 0.28969186544418335, "global_step": 31395, "epoch": 352, "lr": 9.682592937544865e-05}
{"train_loss": 0.42223668098449707, "global_step": 31396, "epoch": 352, "lr": 9.682572610625125e-05}
{"train_loss": 0.40383267402648926, "global_step": 31397, "epoch": 352, "lr": 9.682552283075872e-05}
{"train_loss": 0.3724513649940491, "global_step": 31398, "epoch": 352, "lr": 9.682531954897105e-05}
{"train_loss": 0.3751428425312042, "global_step": 31399, "epoch": 352, "lr": 9.682511626088831e-05}
{"train_loss": 0.3418656587600708, "global_step": 31400, "epoch": 352, "lr": 9.682491296651049e-05}
{"train_loss": 0.29145291447639465, "global_step": 31401, "epoch": 352, "lr": 9.682470966583762e-05}
{"train_loss": 0.3791442811489105, "global_step": 31402, "epoch": 352, "lr": 9.682450635886974e-05}
{"train_loss": 0.3410191833972931, "global_step": 31403, "epoch": 352, "lr": 9.68243030456069e-05}
{"train_loss": 0.5163889527320862, "global_step": 31404, "epoch": 352, "lr": 9.68240997260491e-05}
{"train_loss": 0.33130714297294617, "global_step": 31405, "epoch": 352, "lr": 9.682389640019634e-05}
{"train_loss": 0.40128093957901, "global_step": 31406, "epoch": 352, "lr": 9.68236930680487e-05}
{"train_loss": 0.35941505432128906, "global_step": 31407, "epoch": 352, "lr": 9.682348972960617e-05}
{"train_loss": 0.4447329044342041, "global_step": 31408, "epoch": 352, "lr": 9.682328638486881e-05}
{"train_loss": 0.3328661024570465, "global_step": 31409, "epoch": 352, "lr": 9.682308303383662e-05}
{"train_loss": 0.4044934809207916, "global_step": 31410, "epoch": 352, "lr": 9.682287967650964e-05}
{"train_loss": 0.37923917174339294, "global_step": 31411, "epoch": 352, "lr": 9.682267631288791e-05}
{"train_loss": 0.3631608486175537, "global_step": 31412, "epoch": 352, "lr": 9.682247294297142e-05}
{"train_loss": 0.4200575053691864, "global_step": 31413, "epoch": 352, "lr": 9.682226956676023e-05}
{"train_loss": 0.49740883708000183, "global_step": 31414, "epoch": 352, "lr": 9.682206618425435e-05}
{"train_loss": 0.3874475359916687, "global_step": 31415, "epoch": 352, "lr": 9.682186279545382e-05}
{"train_loss": 0.37180510966965324, "global_step": 31416, "epoch": 352, "lr": 9.682165940035866e-05, "val_loss": 2.322631359100342}
{"train_loss": 0.3390926122665405, "global_step": 31417, "epoch": 353, "lr": 9.68214559989689e-05}
{"train_loss": 0.42488783597946167, "global_step": 31418, "epoch": 353, "lr": 9.682125259128456e-05}
{"train_loss": 0.33704349398612976, "global_step": 31419, "epoch": 353, "lr": 9.682104917730567e-05}
{"train_loss": 0.2670275568962097, "global_step": 31420, "epoch": 353, "lr": 9.682084575703228e-05}
{"train_loss": 0.37390196323394775, "global_step": 31421, "epoch": 353, "lr": 9.682064233046437e-05}
{"train_loss": 0.36077436804771423, "global_step": 31422, "epoch": 353, "lr": 9.682043889760202e-05}
{"train_loss": 0.35735806822776794, "global_step": 31423, "epoch": 353, "lr": 9.682023545844522e-05}
{"train_loss": 0.3492528796195984, "global_step": 31424, "epoch": 353, "lr": 9.6820032012994e-05}
{"train_loss": 0.3276600241661072, "global_step": 31425, "epoch": 353, "lr": 9.681982856124843e-05}
{"train_loss": 0.28356024622917175, "global_step": 31426, "epoch": 353, "lr": 9.681962510320848e-05}
{"train_loss": 0.3532269299030304, "global_step": 31427, "epoch": 353, "lr": 9.68194216388742e-05}
{"train_loss": 0.39313942193984985, "global_step": 31428, "epoch": 353, "lr": 9.681921816824563e-05}
{"train_loss": 0.3266282379627228, "global_step": 31429, "epoch": 353, "lr": 9.681901469132278e-05}
{"train_loss": 0.434541791677475, "global_step": 31430, "epoch": 353, "lr": 9.681881120810569e-05}
{"train_loss": 0.3490487039089203, "global_step": 31431, "epoch": 353, "lr": 9.681860771859437e-05}
{"train_loss": 0.33750787377357483, "global_step": 31432, "epoch": 353, "lr": 9.681840422278887e-05}
{"train_loss": 0.3215194642543793, "global_step": 31433, "epoch": 353, "lr": 9.681820072068921e-05}
{"train_loss": 0.38231730461120605, "global_step": 31434, "epoch": 353, "lr": 9.681799721229541e-05}
{"train_loss": 0.3993552327156067, "global_step": 31435, "epoch": 353, "lr": 9.681779369760748e-05}
{"train_loss": 0.35283929109573364, "global_step": 31436, "epoch": 353, "lr": 9.68175901766255e-05}
{"train_loss": 0.38394036889076233, "global_step": 31437, "epoch": 353, "lr": 9.681738664934945e-05}
{"train_loss": 0.2983059883117676, "global_step": 31438, "epoch": 353, "lr": 9.681718311577938e-05}
{"train_loss": 0.29318320751190186, "global_step": 31439, "epoch": 353, "lr": 9.68169795759153e-05}
{"train_loss": 0.21288515627384186, "global_step": 31440, "epoch": 353, "lr": 9.681677602975725e-05}
{"train_loss": 0.3525315821170807, "global_step": 31441, "epoch": 353, "lr": 9.681657247730526e-05}
{"train_loss": 0.26999086141586304, "global_step": 31442, "epoch": 353, "lr": 9.681636891855935e-05}
{"train_loss": 0.5665779709815979, "global_step": 31443, "epoch": 353, "lr": 9.681616535351955e-05}
{"train_loss": 0.3376794457435608, "global_step": 31444, "epoch": 353, "lr": 9.681596178218588e-05}
{"train_loss": 0.22161605954170227, "global_step": 31445, "epoch": 353, "lr": 9.68157582045584e-05}
{"train_loss": 0.37246444821357727, "global_step": 31446, "epoch": 353, "lr": 9.681555462063708e-05}
{"train_loss": 0.3389844596385956, "global_step": 31447, "epoch": 353, "lr": 9.6815351030422e-05}
{"train_loss": 0.3009483516216278, "global_step": 31448, "epoch": 353, "lr": 9.681514743391316e-05}
{"train_loss": 0.47419846057891846, "global_step": 31449, "epoch": 353, "lr": 9.681494383111058e-05}
{"train_loss": 0.4483962953090668, "global_step": 31450, "epoch": 353, "lr": 9.681474022201432e-05}
{"train_loss": 0.3541082739830017, "global_step": 31451, "epoch": 353, "lr": 9.681453660662438e-05}
{"train_loss": 0.4485805928707123, "global_step": 31452, "epoch": 353, "lr": 9.681433298494081e-05}
{"train_loss": 0.41611817479133606, "global_step": 31453, "epoch": 353, "lr": 9.681412935696362e-05}
{"train_loss": 0.3942558467388153, "global_step": 31454, "epoch": 353, "lr": 9.681392572269283e-05}
{"train_loss": 0.3726681172847748, "global_step": 31455, "epoch": 353, "lr": 9.681372208212848e-05}
{"train_loss": 0.32045361399650574, "global_step": 31456, "epoch": 353, "lr": 9.68135184352706e-05}
{"train_loss": 0.4222177565097809, "global_step": 31457, "epoch": 353, "lr": 9.681331478211922e-05}
{"train_loss": 0.33398592472076416, "global_step": 31458, "epoch": 353, "lr": 9.681311112267435e-05}
{"train_loss": 0.35595273971557617, "global_step": 31459, "epoch": 353, "lr": 9.681290745693602e-05}
{"train_loss": 0.33850371837615967, "global_step": 31460, "epoch": 353, "lr": 9.681270378490428e-05}
{"train_loss": 0.383190393447876, "global_step": 31461, "epoch": 353, "lr": 9.681250010657914e-05}
{"train_loss": 0.42515599727630615, "global_step": 31462, "epoch": 353, "lr": 9.68122964219606e-05}
{"train_loss": 0.3898870646953583, "global_step": 31463, "epoch": 353, "lr": 9.681209273104876e-05}
{"train_loss": 0.3642715811729431, "global_step": 31464, "epoch": 353, "lr": 9.681188903384359e-05}
{"train_loss": 0.2931581437587738, "global_step": 31465, "epoch": 353, "lr": 9.681168533034513e-05}
{"train_loss": 0.33390042185783386, "global_step": 31466, "epoch": 353, "lr": 9.681148162055342e-05}
{"train_loss": 0.39098671078681946, "global_step": 31467, "epoch": 353, "lr": 9.681127790446847e-05}
{"train_loss": 0.3017107844352722, "global_step": 31468, "epoch": 353, "lr": 9.68110741820903e-05}
{"train_loss": 0.4018445611000061, "global_step": 31469, "epoch": 353, "lr": 9.681087045341897e-05}
{"train_loss": 0.3388057053089142, "global_step": 31470, "epoch": 353, "lr": 9.681066671845448e-05}
{"train_loss": 0.36006173491477966, "global_step": 31471, "epoch": 353, "lr": 9.681046297719687e-05}
{"train_loss": 0.34235963225364685, "global_step": 31472, "epoch": 353, "lr": 9.681025922964616e-05}
{"train_loss": 0.43771278858184814, "global_step": 31473, "epoch": 353, "lr": 9.681005547580238e-05}
{"train_loss": 0.3843277096748352, "global_step": 31474, "epoch": 353, "lr": 9.680985171566557e-05}
{"train_loss": 0.36417341232299805, "global_step": 31475, "epoch": 353, "lr": 9.680964794923574e-05}
{"train_loss": 0.4199134409427643, "global_step": 31476, "epoch": 353, "lr": 9.680944417651291e-05}
{"train_loss": 0.321785032749176, "global_step": 31477, "epoch": 353, "lr": 9.680924039749714e-05}
{"train_loss": 0.30520951747894287, "global_step": 31478, "epoch": 353, "lr": 9.680903661218843e-05}
{"train_loss": 0.3894713521003723, "global_step": 31479, "epoch": 353, "lr": 9.680883282058681e-05}
{"train_loss": 0.3497261106967926, "global_step": 31480, "epoch": 353, "lr": 9.680862902269232e-05}
{"train_loss": 0.3806532919406891, "global_step": 31481, "epoch": 353, "lr": 9.680842521850499e-05}
{"train_loss": 0.3302513360977173, "global_step": 31482, "epoch": 353, "lr": 9.680822140802482e-05}
{"train_loss": 0.40883001685142517, "global_step": 31483, "epoch": 353, "lr": 9.680801759125186e-05}
{"train_loss": 0.3581804633140564, "global_step": 31484, "epoch": 353, "lr": 9.680781376818613e-05}
{"train_loss": 0.36218103766441345, "global_step": 31485, "epoch": 353, "lr": 9.680760993882767e-05}
{"train_loss": 0.4344453513622284, "global_step": 31486, "epoch": 353, "lr": 9.68074061031765e-05}
{"train_loss": 0.29765769839286804, "global_step": 31487, "epoch": 353, "lr": 9.680720226123263e-05}
{"train_loss": 0.31544363498687744, "global_step": 31488, "epoch": 353, "lr": 9.680699841299613e-05}
{"train_loss": 0.41460826992988586, "global_step": 31489, "epoch": 353, "lr": 9.680679455846696e-05}
{"train_loss": 0.360386461019516, "global_step": 31490, "epoch": 353, "lr": 9.680659069764521e-05}
{"train_loss": 0.3582877814769745, "global_step": 31491, "epoch": 353, "lr": 9.680638683053089e-05}
{"train_loss": 0.45883119106292725, "global_step": 31492, "epoch": 353, "lr": 9.6806182957124e-05}
{"train_loss": 0.26546892523765564, "global_step": 31493, "epoch": 353, "lr": 9.680597907742461e-05}
{"train_loss": 0.3839687407016754, "global_step": 31494, "epoch": 353, "lr": 9.680577519143271e-05}
{"train_loss": 0.39337190985679626, "global_step": 31495, "epoch": 353, "lr": 9.680557129914836e-05}
{"train_loss": 0.40797001123428345, "global_step": 31496, "epoch": 353, "lr": 9.680536740057156e-05}
{"train_loss": 0.3519453704357147, "global_step": 31497, "epoch": 353, "lr": 9.680516349570235e-05}
{"train_loss": 0.35217198729515076, "global_step": 31498, "epoch": 353, "lr": 9.680495958454075e-05}
{"train_loss": 0.4314258098602295, "global_step": 31499, "epoch": 353, "lr": 9.68047556670868e-05}
{"train_loss": 0.4148712158203125, "global_step": 31500, "epoch": 353, "lr": 9.680455174334053e-05}
{"train_loss": 0.3958083987236023, "global_step": 31501, "epoch": 353, "lr": 9.680434781330194e-05}
{"train_loss": 0.29567185044288635, "global_step": 31502, "epoch": 353, "lr": 9.68041438769711e-05}
{"train_loss": 0.35648635029792786, "global_step": 31503, "epoch": 353, "lr": 9.6803939934348e-05}
{"train_loss": 0.4004514515399933, "global_step": 31504, "epoch": 353, "lr": 9.680373598543269e-05}
{"train_loss": 0.36378848435503713, "global_step": 31505, "epoch": 353, "lr": 9.680353203022516e-05, "val_loss": 2.349275588989258}
{"train_loss": 0.35125425457954407, "global_step": 31506, "epoch": 354, "lr": 9.68033280687255e-05}
{"train_loss": 0.3190351724624634, "global_step": 31507, "epoch": 354, "lr": 9.680312410093369e-05}
{"train_loss": 0.3024211525917053, "global_step": 31508, "epoch": 354, "lr": 9.680292012684977e-05}
{"train_loss": 0.32821184396743774, "global_step": 31509, "epoch": 354, "lr": 9.680271614647376e-05}
{"train_loss": 0.3018319010734558, "global_step": 31510, "epoch": 354, "lr": 9.68025121598057e-05}
{"train_loss": 0.3500302731990814, "global_step": 31511, "epoch": 354, "lr": 9.680230816684561e-05}
{"train_loss": 0.293245404958725, "global_step": 31512, "epoch": 354, "lr": 9.680210416759352e-05}
{"train_loss": 0.34046244621276855, "global_step": 31513, "epoch": 354, "lr": 9.680190016204947e-05}
{"train_loss": 0.5120301246643066, "global_step": 31514, "epoch": 354, "lr": 9.680169615021347e-05}
{"train_loss": 0.42397013306617737, "global_step": 31515, "epoch": 354, "lr": 9.680149213208554e-05}
{"train_loss": 0.3900744318962097, "global_step": 31516, "epoch": 354, "lr": 9.680128810766573e-05}
{"train_loss": 0.3486515283584595, "global_step": 31517, "epoch": 354, "lr": 9.680108407695407e-05}
{"train_loss": 0.31856513023376465, "global_step": 31518, "epoch": 354, "lr": 9.680088003995056e-05}
{"train_loss": 0.32649463415145874, "global_step": 31519, "epoch": 354, "lr": 9.680067599665523e-05}
{"train_loss": 0.39900174736976624, "global_step": 31520, "epoch": 354, "lr": 9.680047194706814e-05}
{"train_loss": 0.33864066004753113, "global_step": 31521, "epoch": 354, "lr": 9.680026789118929e-05}
{"train_loss": 0.3911146819591522, "global_step": 31522, "epoch": 354, "lr": 9.680006382901871e-05}
{"train_loss": 0.3365185856819153, "global_step": 31523, "epoch": 354, "lr": 9.679985976055643e-05}
{"train_loss": 0.4522154629230499, "global_step": 31524, "epoch": 354, "lr": 9.67996556858025e-05}
{"train_loss": 0.35406771302223206, "global_step": 31525, "epoch": 354, "lr": 9.679945160475689e-05}
{"train_loss": 0.5197911262512207, "global_step": 31526, "epoch": 354, "lr": 9.67992475174197e-05}
{"train_loss": 0.3826214373111725, "global_step": 31527, "epoch": 354, "lr": 9.679904342379091e-05}
{"train_loss": 0.4603654146194458, "global_step": 31528, "epoch": 354, "lr": 9.679883932387055e-05}
{"train_loss": 0.43207988142967224, "global_step": 31529, "epoch": 354, "lr": 9.679863521765866e-05}
{"train_loss": 0.39767980575561523, "global_step": 31530, "epoch": 354, "lr": 9.679843110515527e-05}
{"train_loss": 0.4028473496437073, "global_step": 31531, "epoch": 354, "lr": 9.67982269863604e-05}
{"train_loss": 0.4347599148750305, "global_step": 31532, "epoch": 354, "lr": 9.679802286127409e-05}
{"train_loss": 0.26515889167785645, "global_step": 31533, "epoch": 354, "lr": 9.679781872989633e-05}
{"train_loss": 0.3885180354118347, "global_step": 31534, "epoch": 354, "lr": 9.679761459222718e-05}
{"train_loss": 0.3884698748588562, "global_step": 31535, "epoch": 354, "lr": 9.679741044826667e-05}
{"train_loss": 0.31034502387046814, "global_step": 31536, "epoch": 354, "lr": 9.679720629801482e-05}
{"train_loss": 0.34450629353523254, "global_step": 31537, "epoch": 354, "lr": 9.679700214147165e-05}
{"train_loss": 0.37420541048049927, "global_step": 31538, "epoch": 354, "lr": 9.67967979786372e-05}
{"train_loss": 0.37736499309539795, "global_step": 31539, "epoch": 354, "lr": 9.679659380951149e-05}
{"train_loss": 0.3077709674835205, "global_step": 31540, "epoch": 354, "lr": 9.679638963409455e-05}
{"train_loss": 0.3551607131958008, "global_step": 31541, "epoch": 354, "lr": 9.679618545238638e-05}
{"train_loss": 0.3599916398525238, "global_step": 31542, "epoch": 354, "lr": 9.679598126438707e-05}
{"train_loss": 0.41191020607948303, "global_step": 31543, "epoch": 354, "lr": 9.67957770700966e-05}
{"train_loss": 0.28611546754837036, "global_step": 31544, "epoch": 354, "lr": 9.679557286951499e-05}
{"train_loss": 0.41372981667518616, "global_step": 31545, "epoch": 354, "lr": 9.679536866264231e-05}
{"train_loss": 0.46491292119026184, "global_step": 31546, "epoch": 354, "lr": 9.679516444947854e-05}
{"train_loss": 0.49565884470939636, "global_step": 31547, "epoch": 354, "lr": 9.679496023002374e-05}
{"train_loss": 0.34610432386398315, "global_step": 31548, "epoch": 354, "lr": 9.679475600427795e-05}
{"train_loss": 0.29539114236831665, "global_step": 31549, "epoch": 354, "lr": 9.679455177224115e-05}
{"train_loss": 0.2559177279472351, "global_step": 31550, "epoch": 354, "lr": 9.679434753391339e-05}
{"train_loss": 0.5098050236701965, "global_step": 31551, "epoch": 354, "lr": 9.67941432892947e-05}
{"train_loss": 0.31165599822998047, "global_step": 31552, "epoch": 354, "lr": 9.679393903838512e-05}
{"train_loss": 0.3996676504611969, "global_step": 31553, "epoch": 354, "lr": 9.679373478118467e-05}
{"train_loss": 0.4681192636489868, "global_step": 31554, "epoch": 354, "lr": 9.679353051769335e-05}
{"train_loss": 0.41413551568984985, "global_step": 31555, "epoch": 354, "lr": 9.679332624791125e-05}
{"train_loss": 0.25176548957824707, "global_step": 31556, "epoch": 354, "lr": 9.679312197183832e-05}
{"train_loss": 0.49142390489578247, "global_step": 31557, "epoch": 354, "lr": 9.679291768947463e-05}
{"train_loss": 0.30626288056373596, "global_step": 31558, "epoch": 354, "lr": 9.679271340082021e-05}
{"train_loss": 0.3482917249202728, "global_step": 31559, "epoch": 354, "lr": 9.679250910587508e-05}
{"train_loss": 0.38891950249671936, "global_step": 31560, "epoch": 354, "lr": 9.679230480463928e-05}
{"train_loss": 0.5169585347175598, "global_step": 31561, "epoch": 354, "lr": 9.67921004971128e-05}
{"train_loss": 0.43840187788009644, "global_step": 31562, "epoch": 354, "lr": 9.679189618329571e-05}
{"train_loss": 0.42014169692993164, "global_step": 31563, "epoch": 354, "lr": 9.679169186318801e-05}
{"train_loss": 0.4405317008495331, "global_step": 31564, "epoch": 354, "lr": 9.679148753678975e-05}
{"train_loss": 0.2886224687099457, "global_step": 31565, "epoch": 354, "lr": 9.679128320410093e-05}
{"train_loss": 0.3852204382419586, "global_step": 31566, "epoch": 354, "lr": 9.67910788651216e-05}
{"train_loss": 0.49421998858451843, "global_step": 31567, "epoch": 354, "lr": 9.679087451985178e-05}
{"train_loss": 0.46286100149154663, "global_step": 31568, "epoch": 354, "lr": 9.679067016829148e-05}
{"train_loss": 0.3409581184387207, "global_step": 31569, "epoch": 354, "lr": 9.679046581044078e-05}
{"train_loss": 0.3753473162651062, "global_step": 31570, "epoch": 354, "lr": 9.679026144629964e-05}
{"train_loss": 0.346609890460968, "global_step": 31571, "epoch": 354, "lr": 9.679005707586813e-05}
{"train_loss": 0.2719210088253021, "global_step": 31572, "epoch": 354, "lr": 9.678985269914628e-05}
{"train_loss": 0.33250367641448975, "global_step": 31573, "epoch": 354, "lr": 9.678964831613408e-05}
{"train_loss": 0.4160307049751282, "global_step": 31574, "epoch": 354, "lr": 9.67894439268316e-05}
{"train_loss": 0.3030206263065338, "global_step": 31575, "epoch": 354, "lr": 9.678923953123885e-05}
{"train_loss": 0.406586617231369, "global_step": 31576, "epoch": 354, "lr": 9.678903512935586e-05}
{"train_loss": 0.4271373748779297, "global_step": 31577, "epoch": 354, "lr": 9.678883072118265e-05}
{"train_loss": 0.1597202718257904, "global_step": 31578, "epoch": 354, "lr": 9.678862630671924e-05}
{"train_loss": 0.2842731475830078, "global_step": 31579, "epoch": 354, "lr": 9.678842188596568e-05}
{"train_loss": 0.33092156052589417, "global_step": 31580, "epoch": 354, "lr": 9.678821745892199e-05}
{"train_loss": 0.31762999296188354, "global_step": 31581, "epoch": 354, "lr": 9.678801302558819e-05}
{"train_loss": 0.360287070274353, "global_step": 31582, "epoch": 354, "lr": 9.67878085859643e-05}
{"train_loss": 0.2761576473712921, "global_step": 31583, "epoch": 354, "lr": 9.678760414005038e-05}
{"train_loss": 0.3613273501396179, "global_step": 31584, "epoch": 354, "lr": 9.678739968784643e-05}
{"train_loss": 0.3800124228000641, "global_step": 31585, "epoch": 354, "lr": 9.678719522935248e-05}
{"train_loss": 0.30200380086898804, "global_step": 31586, "epoch": 354, "lr": 9.678699076456856e-05}
{"train_loss": 0.4242105185985565, "global_step": 31587, "epoch": 354, "lr": 9.678678629349472e-05}
{"train_loss": 0.38761603832244873, "global_step": 31588, "epoch": 354, "lr": 9.678658181613094e-05}
{"train_loss": 0.328232079744339, "global_step": 31589, "epoch": 354, "lr": 9.678637733247728e-05}
{"train_loss": 0.33944809436798096, "global_step": 31590, "epoch": 354, "lr": 9.678617284253377e-05}
{"train_loss": 0.352573037147522, "global_step": 31591, "epoch": 354, "lr": 9.678596834630043e-05}
{"train_loss": 0.31474873423576355, "global_step": 31592, "epoch": 354, "lr": 9.678576384377728e-05}
{"train_loss": 0.3438322842121124, "global_step": 31593, "epoch": 354, "lr": 9.678555933496436e-05}
{"train_loss": 0.3686178105600764, "global_step": 31594, "epoch": 354, "lr": 9.678535481986169e-05, "val_loss": 2.3223812580108643}
{"train_loss": 0.30631619691848755, "global_step": 31595, "epoch": 355, "lr": 9.678515029846928e-05}
{"train_loss": 0.30189111828804016, "global_step": 31596, "epoch": 355, "lr": 9.678494577078721e-05}
{"train_loss": 0.2669179439544678, "global_step": 31597, "epoch": 355, "lr": 9.678474123681546e-05}
{"train_loss": 0.31174910068511963, "global_step": 31598, "epoch": 355, "lr": 9.678453669655407e-05}
{"train_loss": 0.4072757661342621, "global_step": 31599, "epoch": 355, "lr": 9.678433215000306e-05}
{"train_loss": 0.41116756200790405, "global_step": 31600, "epoch": 355, "lr": 9.678412759716249e-05}
{"train_loss": 0.34991008043289185, "global_step": 31601, "epoch": 355, "lr": 9.678392303803234e-05}
{"train_loss": 0.3106289505958557, "global_step": 31602, "epoch": 355, "lr": 9.678371847261267e-05}
{"train_loss": 0.28993675112724304, "global_step": 31603, "epoch": 355, "lr": 9.67835139009035e-05}
{"train_loss": 0.47126492857933044, "global_step": 31604, "epoch": 355, "lr": 9.678330932290487e-05}
{"train_loss": 0.2720150351524353, "global_step": 31605, "epoch": 355, "lr": 9.678310473861678e-05}
{"train_loss": 0.3681761622428894, "global_step": 31606, "epoch": 355, "lr": 9.678290014803927e-05}
{"train_loss": 0.3378233313560486, "global_step": 31607, "epoch": 355, "lr": 9.678269555117239e-05}
{"train_loss": 0.36259686946868896, "global_step": 31608, "epoch": 355, "lr": 9.678249094801612e-05}
{"train_loss": 0.3131784200668335, "global_step": 31609, "epoch": 355, "lr": 9.678228633857053e-05}
{"train_loss": 0.36936286091804504, "global_step": 31610, "epoch": 355, "lr": 9.678208172283562e-05}
{"train_loss": 0.3588613271713257, "global_step": 31611, "epoch": 355, "lr": 9.678187710081144e-05}
{"train_loss": 0.32307395339012146, "global_step": 31612, "epoch": 355, "lr": 9.678167247249801e-05}
{"train_loss": 0.39925163984298706, "global_step": 31613, "epoch": 355, "lr": 9.678146783789534e-05}
{"train_loss": 0.3727792799472809, "global_step": 31614, "epoch": 355, "lr": 9.678126319700348e-05}
{"train_loss": 0.4321134686470032, "global_step": 31615, "epoch": 355, "lr": 9.678105854982244e-05}
{"train_loss": 0.36089661717414856, "global_step": 31616, "epoch": 355, "lr": 9.678085389635227e-05}
{"train_loss": 0.3032819330692291, "global_step": 31617, "epoch": 355, "lr": 9.678064923659299e-05}
{"train_loss": 0.32270240783691406, "global_step": 31618, "epoch": 355, "lr": 9.678044457054461e-05}
{"train_loss": 0.2758764326572418, "global_step": 31619, "epoch": 355, "lr": 9.678023989820717e-05}
{"train_loss": 0.33200135827064514, "global_step": 31620, "epoch": 355, "lr": 9.67800352195807e-05}
{"train_loss": 0.34239402413368225, "global_step": 31621, "epoch": 355, "lr": 9.677983053466521e-05}
{"train_loss": 0.4748927652835846, "global_step": 31622, "epoch": 355, "lr": 9.677962584346076e-05}
{"train_loss": 0.35712701082229614, "global_step": 31623, "epoch": 355, "lr": 9.677942114596736e-05}
{"train_loss": 0.37241941690444946, "global_step": 31624, "epoch": 355, "lr": 9.677921644218503e-05}
{"train_loss": 0.37959760427474976, "global_step": 31625, "epoch": 355, "lr": 9.67790117321138e-05}
{"train_loss": 0.29243800044059753, "global_step": 31626, "epoch": 355, "lr": 9.677880701575372e-05}
{"train_loss": 0.30780863761901855, "global_step": 31627, "epoch": 355, "lr": 9.677860229310478e-05}
{"train_loss": 0.5699434280395508, "global_step": 31628, "epoch": 355, "lr": 9.677839756416704e-05}
{"train_loss": 0.4122810661792755, "global_step": 31629, "epoch": 355, "lr": 9.67781928289405e-05}
{"train_loss": 0.3305462598800659, "global_step": 31630, "epoch": 355, "lr": 9.677798808742522e-05}
{"train_loss": 0.3152453899383545, "global_step": 31631, "epoch": 355, "lr": 9.67777833396212e-05}
{"train_loss": 0.3343067765235901, "global_step": 31632, "epoch": 355, "lr": 9.677757858552847e-05}
{"train_loss": 0.38083869218826294, "global_step": 31633, "epoch": 355, "lr": 9.677737382514708e-05}
{"train_loss": 0.3419385254383087, "global_step": 31634, "epoch": 355, "lr": 9.677716905847703e-05}
{"train_loss": 0.4005345404148102, "global_step": 31635, "epoch": 355, "lr": 9.677696428551837e-05}
{"train_loss": 0.33618035912513733, "global_step": 31636, "epoch": 355, "lr": 9.67767595062711e-05}
{"train_loss": 0.30623677372932434, "global_step": 31637, "epoch": 355, "lr": 9.677655472073529e-05}
{"train_loss": 0.43619370460510254, "global_step": 31638, "epoch": 355, "lr": 9.677634992891093e-05}
{"train_loss": 0.3500644862651825, "global_step": 31639, "epoch": 355, "lr": 9.677614513079804e-05}
{"train_loss": 0.27994078397750854, "global_step": 31640, "epoch": 355, "lr": 9.67759403263967e-05}
{"train_loss": 0.28587183356285095, "global_step": 31641, "epoch": 355, "lr": 9.677573551570689e-05}
{"train_loss": 0.359789103269577, "global_step": 31642, "epoch": 355, "lr": 9.677553069872865e-05}
{"train_loss": 0.33470794558525085, "global_step": 31643, "epoch": 355, "lr": 9.677532587546203e-05}
{"train_loss": 0.4072924852371216, "global_step": 31644, "epoch": 355, "lr": 9.6775121045907e-05}
{"train_loss": 0.2987152934074402, "global_step": 31645, "epoch": 355, "lr": 9.677491621006365e-05}
{"train_loss": 0.39670562744140625, "global_step": 31646, "epoch": 355, "lr": 9.677471136793199e-05}
{"train_loss": 0.25692152976989746, "global_step": 31647, "epoch": 355, "lr": 9.677450651951202e-05}
{"train_loss": 0.3467892110347748, "global_step": 31648, "epoch": 355, "lr": 9.67743016648038e-05}
{"train_loss": 0.32280483841896057, "global_step": 31649, "epoch": 355, "lr": 9.677409680380734e-05}
{"train_loss": 0.42988210916519165, "global_step": 31650, "epoch": 355, "lr": 9.677389193652268e-05}
{"train_loss": 0.2675107419490814, "global_step": 31651, "epoch": 355, "lr": 9.677368706294983e-05}
{"train_loss": 0.39719119668006897, "global_step": 31652, "epoch": 355, "lr": 9.677348218308884e-05}
{"train_loss": 0.2663347125053406, "global_step": 31653, "epoch": 355, "lr": 9.677327729693971e-05}
{"train_loss": 0.3676131069660187, "global_step": 31654, "epoch": 355, "lr": 9.677307240450248e-05}
{"train_loss": 0.3032602369785309, "global_step": 31655, "epoch": 355, "lr": 9.677286750577722e-05}
{"train_loss": 0.34501343965530396, "global_step": 31656, "epoch": 355, "lr": 9.677266260076388e-05}
{"train_loss": 0.34850582480430603, "global_step": 31657, "epoch": 355, "lr": 9.677245768946253e-05}
{"train_loss": 0.36851346492767334, "global_step": 31658, "epoch": 355, "lr": 9.67722527718732e-05}
{"train_loss": 0.3998327851295471, "global_step": 31659, "epoch": 355, "lr": 9.677204784799591e-05}
{"train_loss": 0.4179958701133728, "global_step": 31660, "epoch": 355, "lr": 9.677184291783068e-05}
{"train_loss": 0.4902963936328888, "global_step": 31661, "epoch": 355, "lr": 9.677163798137756e-05}
{"train_loss": 0.5432373285293579, "global_step": 31662, "epoch": 355, "lr": 9.677143303863656e-05}
{"train_loss": 0.3561338782310486, "global_step": 31663, "epoch": 355, "lr": 9.67712280896077e-05}
{"train_loss": 0.36076536774635315, "global_step": 31664, "epoch": 355, "lr": 9.677102313429103e-05}
{"train_loss": 0.312486469745636, "global_step": 31665, "epoch": 355, "lr": 9.677081817268657e-05}
{"train_loss": 0.30690595507621765, "global_step": 31666, "epoch": 355, "lr": 9.677061320479433e-05}
{"train_loss": 0.32766401767730713, "global_step": 31667, "epoch": 355, "lr": 9.677040823061435e-05}
{"train_loss": 0.2718755006790161, "global_step": 31668, "epoch": 355, "lr": 9.677020325014668e-05}
{"train_loss": 0.38773399591445923, "global_step": 31669, "epoch": 355, "lr": 9.676999826339131e-05}
{"train_loss": 0.3227180242538452, "global_step": 31670, "epoch": 355, "lr": 9.676979327034828e-05}
{"train_loss": 0.33432087302207947, "global_step": 31671, "epoch": 355, "lr": 9.676958827101761e-05}
{"train_loss": 0.5074732899665833, "global_step": 31672, "epoch": 355, "lr": 9.676938326539937e-05}
{"train_loss": 0.31156593561172485, "global_step": 31673, "epoch": 355, "lr": 9.676917825349354e-05}
{"train_loss": 0.4694533050060272, "global_step": 31674, "epoch": 355, "lr": 9.676897323530016e-05}
{"train_loss": 0.27047696709632874, "global_step": 31675, "epoch": 355, "lr": 9.676876821081928e-05}
{"train_loss": 0.47114333510398865, "global_step": 31676, "epoch": 355, "lr": 9.67685631800509e-05}
{"train_loss": 0.40537670254707336, "global_step": 31677, "epoch": 355, "lr": 9.676835814299504e-05}
{"train_loss": 0.3253557085990906, "global_step": 31678, "epoch": 355, "lr": 9.676815309965176e-05}
{"train_loss": 0.35433560609817505, "global_step": 31679, "epoch": 355, "lr": 9.676794805002107e-05}
{"train_loss": 0.32295864820480347, "global_step": 31680, "epoch": 355, "lr": 9.6767742994103e-05}
{"train_loss": 0.389066606760025, "global_step": 31681, "epoch": 355, "lr": 9.676753793189758e-05}
{"train_loss": 0.3356809914112091, "global_step": 31682, "epoch": 355, "lr": 9.676733286340483e-05}
{"train_loss": 0.3560934435115771, "global_step": 31683, "epoch": 355, "lr": 9.676712778862478e-05, "val_loss": 2.3656983375549316, "train_action_mse_error": 33.53449249267578}
{"train_loss": 0.29000577330589294, "global_step": 31684, "epoch": 356, "lr": 9.676692270755746e-05}
{"train_loss": 0.3516797125339508, "global_step": 31685, "epoch": 356, "lr": 9.67667176202029e-05}
{"train_loss": 0.36556899547576904, "global_step": 31686, "epoch": 356, "lr": 9.676651252656112e-05}
{"train_loss": 0.32750624418258667, "global_step": 31687, "epoch": 356, "lr": 9.676630742663217e-05}
{"train_loss": 0.4420602321624756, "global_step": 31688, "epoch": 356, "lr": 9.676610232041604e-05}
{"train_loss": 0.3672931492328644, "global_step": 31689, "epoch": 356, "lr": 9.676589720791278e-05}
{"train_loss": 0.2649276554584503, "global_step": 31690, "epoch": 356, "lr": 9.676569208912243e-05}
{"train_loss": 0.41722816228866577, "global_step": 31691, "epoch": 356, "lr": 9.676548696404499e-05}
{"train_loss": 0.40252912044525146, "global_step": 31692, "epoch": 356, "lr": 9.67652818326805e-05}
{"train_loss": 0.4179309904575348, "global_step": 31693, "epoch": 356, "lr": 9.676507669502899e-05}
{"train_loss": 0.258230060338974, "global_step": 31694, "epoch": 356, "lr": 9.676487155109049e-05}
{"train_loss": 0.3999350070953369, "global_step": 31695, "epoch": 356, "lr": 9.676466640086502e-05}
{"train_loss": 0.42070916295051575, "global_step": 31696, "epoch": 356, "lr": 9.67644612443526e-05}
{"train_loss": 0.29165682196617126, "global_step": 31697, "epoch": 356, "lr": 9.676425608155329e-05}
{"train_loss": 0.3474235534667969, "global_step": 31698, "epoch": 356, "lr": 9.676405091246707e-05}
{"train_loss": 0.3653824031352997, "global_step": 31699, "epoch": 356, "lr": 9.6763845737094e-05}
{"train_loss": 0.40091806650161743, "global_step": 31700, "epoch": 356, "lr": 9.676364055543412e-05}
{"train_loss": 0.37492045760154724, "global_step": 31701, "epoch": 356, "lr": 9.676343536748742e-05}
{"train_loss": 0.395942747592926, "global_step": 31702, "epoch": 356, "lr": 9.676323017325395e-05}
{"train_loss": 0.3571779131889343, "global_step": 31703, "epoch": 356, "lr": 9.676302497273374e-05}
{"train_loss": 0.3633420765399933, "global_step": 31704, "epoch": 356, "lr": 9.67628197659268e-05}
{"train_loss": 0.4907403290271759, "global_step": 31705, "epoch": 356, "lr": 9.676261455283318e-05}
{"train_loss": 0.2640547752380371, "global_step": 31706, "epoch": 356, "lr": 9.676240933345288e-05}
{"train_loss": 0.326271116733551, "global_step": 31707, "epoch": 356, "lr": 9.676220410778596e-05}
{"train_loss": 0.28846272826194763, "global_step": 31708, "epoch": 356, "lr": 9.676199887583243e-05}
{"train_loss": 0.3568035662174225, "global_step": 31709, "epoch": 356, "lr": 9.676179363759232e-05}
{"train_loss": 0.37766122817993164, "global_step": 31710, "epoch": 356, "lr": 9.676158839306567e-05}
{"train_loss": 0.33609023690223694, "global_step": 31711, "epoch": 356, "lr": 9.676138314225247e-05}
{"train_loss": 0.4088071286678314, "global_step": 31712, "epoch": 356, "lr": 9.676117788515278e-05}
{"train_loss": 0.3772328197956085, "global_step": 31713, "epoch": 356, "lr": 9.676097262176662e-05}
{"train_loss": 0.4133850038051605, "global_step": 31714, "epoch": 356, "lr": 9.6760767352094e-05}
{"train_loss": 0.41729483008384705, "global_step": 31715, "epoch": 356, "lr": 9.6760562076135e-05}
{"train_loss": 0.36278077960014343, "global_step": 31716, "epoch": 356, "lr": 9.67603567938896e-05}
{"train_loss": 0.3873939514160156, "global_step": 31717, "epoch": 356, "lr": 9.676015150535782e-05}
{"train_loss": 0.36075854301452637, "global_step": 31718, "epoch": 356, "lr": 9.675994621053972e-05}
{"train_loss": 0.3933870494365692, "global_step": 31719, "epoch": 356, "lr": 9.675974090943531e-05}
{"train_loss": 0.3673628866672516, "global_step": 31720, "epoch": 356, "lr": 9.675953560204464e-05}
{"train_loss": 0.3802661895751953, "global_step": 31721, "epoch": 356, "lr": 9.675933028836771e-05}
{"train_loss": 0.39455446600914, "global_step": 31722, "epoch": 356, "lr": 9.675912496840456e-05}
{"train_loss": 0.33716991543769836, "global_step": 31723, "epoch": 356, "lr": 9.67589196421552e-05}
{"train_loss": 0.2554883062839508, "global_step": 31724, "epoch": 356, "lr": 9.67587143096197e-05}
{"train_loss": 0.4843025505542755, "global_step": 31725, "epoch": 356, "lr": 9.675850897079804e-05}
{"train_loss": 0.3755779266357422, "global_step": 31726, "epoch": 356, "lr": 9.675830362569027e-05}
{"train_loss": 0.2691781520843506, "global_step": 31727, "epoch": 356, "lr": 9.675809827429643e-05}
{"train_loss": 0.42971640825271606, "global_step": 31728, "epoch": 356, "lr": 9.675789291661652e-05}
{"train_loss": 0.397722065448761, "global_step": 31729, "epoch": 356, "lr": 9.67576875526506e-05}
{"train_loss": 0.39117932319641113, "global_step": 31730, "epoch": 356, "lr": 9.675748218239866e-05}
{"train_loss": 0.4366971254348755, "global_step": 31731, "epoch": 356, "lr": 9.675727680586075e-05}
{"train_loss": 0.33453553915023804, "global_step": 31732, "epoch": 356, "lr": 9.67570714230369e-05}
{"train_loss": 0.29485952854156494, "global_step": 31733, "epoch": 356, "lr": 9.675686603392713e-05}
{"train_loss": 0.34936419129371643, "global_step": 31734, "epoch": 356, "lr": 9.675666063853146e-05}
{"train_loss": 0.4775570034980774, "global_step": 31735, "epoch": 356, "lr": 9.675645523684993e-05}
{"train_loss": 0.3164929151535034, "global_step": 31736, "epoch": 356, "lr": 9.675624982888256e-05}
{"train_loss": 0.38617026805877686, "global_step": 31737, "epoch": 356, "lr": 9.67560444146294e-05}
{"train_loss": 0.3374156653881073, "global_step": 31738, "epoch": 356, "lr": 9.675583899409045e-05}
{"train_loss": 0.3871111273765564, "global_step": 31739, "epoch": 356, "lr": 9.675563356726574e-05}
{"train_loss": 0.2715093493461609, "global_step": 31740, "epoch": 356, "lr": 9.675542813415533e-05}
{"train_loss": 0.3788414001464844, "global_step": 31741, "epoch": 356, "lr": 9.675522269475919e-05}
{"train_loss": 0.3708704710006714, "global_step": 31742, "epoch": 356, "lr": 9.67550172490774e-05}
{"train_loss": 0.411577433347702, "global_step": 31743, "epoch": 356, "lr": 9.675481179710996e-05}
{"train_loss": 0.412067174911499, "global_step": 31744, "epoch": 356, "lr": 9.67546063388569e-05}
{"train_loss": 0.4267863631248474, "global_step": 31745, "epoch": 356, "lr": 9.675440087431827e-05}
{"train_loss": 0.35589703917503357, "global_step": 31746, "epoch": 356, "lr": 9.675419540349406e-05}
{"train_loss": 0.4044981896877289, "global_step": 31747, "epoch": 356, "lr": 9.675398992638434e-05}
{"train_loss": 0.2811763882637024, "global_step": 31748, "epoch": 356, "lr": 9.675378444298911e-05}
{"train_loss": 0.40361449122428894, "global_step": 31749, "epoch": 356, "lr": 9.675357895330839e-05}
{"train_loss": 0.4034581482410431, "global_step": 31750, "epoch": 356, "lr": 9.675337345734224e-05}
{"train_loss": 0.31654754281044006, "global_step": 31751, "epoch": 356, "lr": 9.675316795509064e-05}
{"train_loss": 0.4032494127750397, "global_step": 31752, "epoch": 356, "lr": 9.675296244655367e-05}
{"train_loss": 0.3448468744754791, "global_step": 31753, "epoch": 356, "lr": 9.675275693173134e-05}
{"train_loss": 0.3929756283760071, "global_step": 31754, "epoch": 356, "lr": 9.675255141062366e-05}
{"train_loss": 0.4136215150356293, "global_step": 31755, "epoch": 356, "lr": 9.675234588323068e-05}
{"train_loss": 0.3451378643512726, "global_step": 31756, "epoch": 356, "lr": 9.67521403495524e-05}
{"train_loss": 0.28704801201820374, "global_step": 31757, "epoch": 356, "lr": 9.675193480958886e-05}
{"train_loss": 0.39774224162101746, "global_step": 31758, "epoch": 356, "lr": 9.675172926334011e-05}
{"train_loss": 0.37376320362091064, "global_step": 31759, "epoch": 356, "lr": 9.675152371080616e-05}
{"train_loss": 0.3528093993663788, "global_step": 31760, "epoch": 356, "lr": 9.675131815198704e-05}
{"train_loss": 0.4221348762512207, "global_step": 31761, "epoch": 356, "lr": 9.675111258688276e-05}
{"train_loss": 0.38282543420791626, "global_step": 31762, "epoch": 356, "lr": 9.675090701549336e-05}
{"train_loss": 0.2812176048755646, "global_step": 31763, "epoch": 356, "lr": 9.675070143781887e-05}
{"train_loss": 0.4025443196296692, "global_step": 31764, "epoch": 356, "lr": 9.675049585385935e-05}
{"train_loss": 0.3189161419868469, "global_step": 31765, "epoch": 356, "lr": 9.675029026361477e-05}
{"train_loss": 0.4461674690246582, "global_step": 31766, "epoch": 356, "lr": 9.675008466708519e-05}
{"train_loss": 0.34652838110923767, "global_step": 31767, "epoch": 356, "lr": 9.674987906427063e-05}
{"train_loss": 0.3887108564376831, "global_step": 31768, "epoch": 356, "lr": 9.674967345517111e-05}
{"train_loss": 0.3682864308357239, "global_step": 31769, "epoch": 356, "lr": 9.674946783978666e-05}
{"train_loss": 0.2180979698896408, "global_step": 31770, "epoch": 356, "lr": 9.674926221811733e-05}
{"train_loss": 0.3356948792934418, "global_step": 31771, "epoch": 356, "lr": 9.674905659016313e-05}
{"train_loss": 0.36658816340934025, "global_step": 31772, "epoch": 356, "lr": 9.67488509559241e-05, "val_loss": 2.3508567810058594}
{"train_loss": 0.31957969069480896, "global_step": 31773, "epoch": 357, "lr": 9.674864531540023e-05}
{"train_loss": 0.354662150144577, "global_step": 31774, "epoch": 357, "lr": 9.674843966859159e-05}
{"train_loss": 0.25327566266059875, "global_step": 31775, "epoch": 357, "lr": 9.67482340154982e-05}
{"train_loss": 0.4258189797401428, "global_step": 31776, "epoch": 357, "lr": 9.674802835612007e-05}
{"train_loss": 0.46640637516975403, "global_step": 31777, "epoch": 357, "lr": 9.674782269045724e-05}
{"train_loss": 0.2903881072998047, "global_step": 31778, "epoch": 357, "lr": 9.674761701850972e-05}
{"train_loss": 0.32828739285469055, "global_step": 31779, "epoch": 357, "lr": 9.674741134027755e-05}
{"train_loss": 0.3152922987937927, "global_step": 31780, "epoch": 357, "lr": 9.674720565576079e-05}
{"train_loss": 0.31559133529663086, "global_step": 31781, "epoch": 357, "lr": 9.674699996495941e-05}
{"train_loss": 0.35168614983558655, "global_step": 31782, "epoch": 357, "lr": 9.67467942678735e-05}
{"train_loss": 0.26361769437789917, "global_step": 31783, "epoch": 357, "lr": 9.674658856450303e-05}
{"train_loss": 0.2559879422187805, "global_step": 31784, "epoch": 357, "lr": 9.674638285484806e-05}
{"train_loss": 0.30503812432289124, "global_step": 31785, "epoch": 357, "lr": 9.674617713890859e-05}
{"train_loss": 0.5600847601890564, "global_step": 31786, "epoch": 357, "lr": 9.674597141668468e-05}
{"train_loss": 0.37130236625671387, "global_step": 31787, "epoch": 357, "lr": 9.674576568817634e-05}
{"train_loss": 0.33351895213127136, "global_step": 31788, "epoch": 357, "lr": 9.67455599533836e-05}
{"train_loss": 0.3391895890235901, "global_step": 31789, "epoch": 357, "lr": 9.67453542123065e-05}
{"train_loss": 0.4459080994129181, "global_step": 31790, "epoch": 357, "lr": 9.674514846494505e-05}
{"train_loss": 0.3743089735507965, "global_step": 31791, "epoch": 357, "lr": 9.674494271129928e-05}
{"train_loss": 0.264218807220459, "global_step": 31792, "epoch": 357, "lr": 9.674473695136923e-05}
{"train_loss": 0.2704296410083771, "global_step": 31793, "epoch": 357, "lr": 9.674453118515492e-05}
{"train_loss": 0.35557809472084045, "global_step": 31794, "epoch": 357, "lr": 9.674432541265637e-05}
{"train_loss": 0.33604660630226135, "global_step": 31795, "epoch": 357, "lr": 9.674411963387362e-05}
{"train_loss": 0.3638409972190857, "global_step": 31796, "epoch": 357, "lr": 9.674391384880668e-05}
{"train_loss": 0.3272402584552765, "global_step": 31797, "epoch": 357, "lr": 9.674370805745561e-05}
{"train_loss": 0.3818412125110626, "global_step": 31798, "epoch": 357, "lr": 9.674350225982041e-05}
{"train_loss": 0.3126210570335388, "global_step": 31799, "epoch": 357, "lr": 9.67432964559011e-05}
{"train_loss": 0.44827571511268616, "global_step": 31800, "epoch": 357, "lr": 9.674309064569775e-05}
{"train_loss": 0.33807337284088135, "global_step": 31801, "epoch": 357, "lr": 9.674288482921034e-05}
{"train_loss": 0.36981651186943054, "global_step": 31802, "epoch": 357, "lr": 9.674267900643894e-05}
{"train_loss": 0.45391857624053955, "global_step": 31803, "epoch": 357, "lr": 9.674247317738354e-05}
{"train_loss": 0.2919788062572479, "global_step": 31804, "epoch": 357, "lr": 9.674226734204419e-05}
{"train_loss": 0.35717710852622986, "global_step": 31805, "epoch": 357, "lr": 9.674206150042092e-05}
{"train_loss": 0.47481322288513184, "global_step": 31806, "epoch": 357, "lr": 9.674185565251374e-05}
{"train_loss": 0.5891895294189453, "global_step": 31807, "epoch": 357, "lr": 9.67416497983227e-05}
{"train_loss": 0.41626372933387756, "global_step": 31808, "epoch": 357, "lr": 9.674144393784779e-05}
{"train_loss": 0.491681307554245, "global_step": 31809, "epoch": 357, "lr": 9.674123807108908e-05}
{"train_loss": 0.30953943729400635, "global_step": 31810, "epoch": 357, "lr": 9.674103219804657e-05}
{"train_loss": 0.3579193651676178, "global_step": 31811, "epoch": 357, "lr": 9.674082631872032e-05}
{"train_loss": 0.31674468517303467, "global_step": 31812, "epoch": 357, "lr": 9.674062043311032e-05}
{"train_loss": 0.4566481411457062, "global_step": 31813, "epoch": 357, "lr": 9.67404145412166e-05}
{"train_loss": 0.44718271493911743, "global_step": 31814, "epoch": 357, "lr": 9.674020864303922e-05}
{"train_loss": 0.39078080654144287, "global_step": 31815, "epoch": 357, "lr": 9.674000273857819e-05}
{"train_loss": 0.31673529744148254, "global_step": 31816, "epoch": 357, "lr": 9.673979682783352e-05}
{"train_loss": 0.36310330033302307, "global_step": 31817, "epoch": 357, "lr": 9.673959091080527e-05}
{"train_loss": 0.3254954218864441, "global_step": 31818, "epoch": 357, "lr": 9.673938498749345e-05}
{"train_loss": 0.4199731945991516, "global_step": 31819, "epoch": 357, "lr": 9.673917905789806e-05}
{"train_loss": 0.3223436176776886, "global_step": 31820, "epoch": 357, "lr": 9.67389731220192e-05}
{"train_loss": 0.36028510332107544, "global_step": 31821, "epoch": 357, "lr": 9.673876717985684e-05}
{"train_loss": 0.3589748442173004, "global_step": 31822, "epoch": 357, "lr": 9.6738561231411e-05}
{"train_loss": 0.32307833433151245, "global_step": 31823, "epoch": 357, "lr": 9.673835527668175e-05}
{"train_loss": 0.4297909736633301, "global_step": 31824, "epoch": 357, "lr": 9.67381493156691e-05}
{"train_loss": 0.3964194655418396, "global_step": 31825, "epoch": 357, "lr": 9.673794334837307e-05}
{"train_loss": 0.4488804042339325, "global_step": 31826, "epoch": 357, "lr": 9.67377373747937e-05}
{"train_loss": 0.45228585600852966, "global_step": 31827, "epoch": 357, "lr": 9.673753139493101e-05}
{"train_loss": 0.3965521454811096, "global_step": 31828, "epoch": 357, "lr": 9.673732540878502e-05}
{"train_loss": 0.3124788701534271, "global_step": 31829, "epoch": 357, "lr": 9.673711941635577e-05}
{"train_loss": 0.42715486884117126, "global_step": 31830, "epoch": 357, "lr": 9.673691341764328e-05}
{"train_loss": 0.41177988052368164, "global_step": 31831, "epoch": 357, "lr": 9.67367074126476e-05}
{"train_loss": 0.329113632440567, "global_step": 31832, "epoch": 357, "lr": 9.673650140136872e-05}
{"train_loss": 0.3416009843349457, "global_step": 31833, "epoch": 357, "lr": 9.673629538380669e-05}
{"train_loss": 0.4210652709007263, "global_step": 31834, "epoch": 357, "lr": 9.673608935996155e-05}
{"train_loss": 0.3392053246498108, "global_step": 31835, "epoch": 357, "lr": 9.673588332983328e-05}
{"train_loss": 0.3858373761177063, "global_step": 31836, "epoch": 357, "lr": 9.673567729342196e-05}
{"train_loss": 0.3577486276626587, "global_step": 31837, "epoch": 357, "lr": 9.673547125072761e-05}
{"train_loss": 0.3577364683151245, "global_step": 31838, "epoch": 357, "lr": 9.673526520175022e-05}
{"train_loss": 0.3349500000476837, "global_step": 31839, "epoch": 357, "lr": 9.673505914648986e-05}
{"train_loss": 0.31787630915641785, "global_step": 31840, "epoch": 357, "lr": 9.673485308494654e-05}
{"train_loss": 0.42588138580322266, "global_step": 31841, "epoch": 357, "lr": 9.67346470171203e-05}
{"train_loss": 0.471153199672699, "global_step": 31842, "epoch": 357, "lr": 9.673444094301113e-05}
{"train_loss": 0.40773800015449524, "global_step": 31843, "epoch": 357, "lr": 9.67342348626191e-05}
{"train_loss": 0.4370368719100952, "global_step": 31844, "epoch": 357, "lr": 9.673402877594422e-05}
{"train_loss": 0.3535354733467102, "global_step": 31845, "epoch": 357, "lr": 9.673382268298652e-05}
{"train_loss": 0.34663286805152893, "global_step": 31846, "epoch": 357, "lr": 9.673361658374603e-05}
{"train_loss": 0.2933077812194824, "global_step": 31847, "epoch": 357, "lr": 9.673341047822275e-05}
{"train_loss": 0.36727797985076904, "global_step": 31848, "epoch": 357, "lr": 9.673320436641676e-05}
{"train_loss": 0.4074755907058716, "global_step": 31849, "epoch": 357, "lr": 9.673299824832804e-05}
{"train_loss": 0.2821577787399292, "global_step": 31850, "epoch": 357, "lr": 9.673279212395666e-05}
{"train_loss": 0.3568277359008789, "global_step": 31851, "epoch": 357, "lr": 9.673258599330261e-05}
{"train_loss": 0.45105233788490295, "global_step": 31852, "epoch": 357, "lr": 9.673237985636593e-05}
{"train_loss": 0.3816812038421631, "global_step": 31853, "epoch": 357, "lr": 9.673217371314667e-05}
{"train_loss": 0.37250855565071106, "global_step": 31854, "epoch": 357, "lr": 9.673196756364484e-05}
{"train_loss": 0.4005984961986542, "global_step": 31855, "epoch": 357, "lr": 9.673176140786045e-05}
{"train_loss": 0.3662971258163452, "global_step": 31856, "epoch": 357, "lr": 9.673155524579356e-05}
{"train_loss": 0.31768718361854553, "global_step": 31857, "epoch": 357, "lr": 9.673134907744416e-05}
{"train_loss": 0.3515196442604065, "global_step": 31858, "epoch": 357, "lr": 9.673114290281231e-05}
{"train_loss": 0.33213385939598083, "global_step": 31859, "epoch": 357, "lr": 9.673093672189804e-05}
{"train_loss": 0.34799617528915405, "global_step": 31860, "epoch": 357, "lr": 9.673073053470134e-05}
{"train_loss": 0.37000532632463434, "global_step": 31861, "epoch": 357, "lr": 9.673052434122229e-05, "val_loss": 2.360239267349243}
{"train_loss": 0.39461836218833923, "global_step": 31862, "epoch": 358, "lr": 9.673031814146088e-05}
{"train_loss": 0.4534638822078705, "global_step": 31863, "epoch": 358, "lr": 9.673011193541714e-05}
{"train_loss": 0.4548946022987366, "global_step": 31864, "epoch": 358, "lr": 9.67299057230911e-05}
{"train_loss": 0.46174076199531555, "global_step": 31865, "epoch": 358, "lr": 9.672969950448282e-05}
{"train_loss": 0.36671432852745056, "global_step": 31866, "epoch": 358, "lr": 9.672949327959229e-05}
{"train_loss": 0.37385210394859314, "global_step": 31867, "epoch": 358, "lr": 9.672928704841955e-05}
{"train_loss": 0.21775765717029572, "global_step": 31868, "epoch": 358, "lr": 9.672908081096461e-05}
{"train_loss": 0.37523332238197327, "global_step": 31869, "epoch": 358, "lr": 9.672887456722753e-05}
{"train_loss": 0.3231951892375946, "global_step": 31870, "epoch": 358, "lr": 9.672866831720832e-05}
{"train_loss": 0.3800485134124756, "global_step": 31871, "epoch": 358, "lr": 9.6728462060907e-05}
{"train_loss": 0.4373522102832794, "global_step": 31872, "epoch": 358, "lr": 9.672825579832363e-05}
{"train_loss": 0.33659374713897705, "global_step": 31873, "epoch": 358, "lr": 9.67280495294582e-05}
{"train_loss": 0.3887040317058563, "global_step": 31874, "epoch": 358, "lr": 9.672784325431076e-05}
{"train_loss": 0.45836254954338074, "global_step": 31875, "epoch": 358, "lr": 9.672763697288133e-05}
{"train_loss": 0.4374031722545624, "global_step": 31876, "epoch": 358, "lr": 9.672743068516994e-05}
{"train_loss": 0.2491181492805481, "global_step": 31877, "epoch": 358, "lr": 9.67272243911766e-05}
{"train_loss": 0.4013505280017853, "global_step": 31878, "epoch": 358, "lr": 9.672701809090137e-05}
{"train_loss": 0.3187354803085327, "global_step": 31879, "epoch": 358, "lr": 9.672681178434426e-05}
{"train_loss": 0.338203489780426, "global_step": 31880, "epoch": 358, "lr": 9.67266054715053e-05}
{"train_loss": 0.348655104637146, "global_step": 31881, "epoch": 358, "lr": 9.672639915238453e-05}
{"train_loss": 0.3331102728843689, "global_step": 31882, "epoch": 358, "lr": 9.672619282698194e-05}
{"train_loss": 0.32681331038475037, "global_step": 31883, "epoch": 358, "lr": 9.67259864952976e-05}
{"train_loss": 0.40738576650619507, "global_step": 31884, "epoch": 358, "lr": 9.672578015733151e-05}
{"train_loss": 0.3320915400981903, "global_step": 31885, "epoch": 358, "lr": 9.672557381308371e-05}
{"train_loss": 0.27526146173477173, "global_step": 31886, "epoch": 358, "lr": 9.672536746255423e-05}
{"train_loss": 0.24304358661174774, "global_step": 31887, "epoch": 358, "lr": 9.67251611057431e-05}
{"train_loss": 0.29967719316482544, "global_step": 31888, "epoch": 358, "lr": 9.672495474265033e-05}
{"train_loss": 0.3521648645401001, "global_step": 31889, "epoch": 358, "lr": 9.672474837327596e-05}
{"train_loss": 0.4315231442451477, "global_step": 31890, "epoch": 358, "lr": 9.672454199762002e-05}
{"train_loss": 0.29512330889701843, "global_step": 31891, "epoch": 358, "lr": 9.672433561568255e-05}
{"train_loss": 0.39532750844955444, "global_step": 31892, "epoch": 358, "lr": 9.672412922746355e-05}
{"train_loss": 0.3170689642429352, "global_step": 31893, "epoch": 358, "lr": 9.672392283296306e-05}
{"train_loss": 0.3835480511188507, "global_step": 31894, "epoch": 358, "lr": 9.672371643218109e-05}
{"train_loss": 0.33359864354133606, "global_step": 31895, "epoch": 358, "lr": 9.672351002511771e-05}
{"train_loss": 0.35809624195098877, "global_step": 31896, "epoch": 358, "lr": 9.672330361177292e-05}
{"train_loss": 0.34651926159858704, "global_step": 31897, "epoch": 358, "lr": 9.672309719214675e-05}
{"train_loss": 0.34075185656547546, "global_step": 31898, "epoch": 358, "lr": 9.672289076623923e-05}
{"train_loss": 0.3571561574935913, "global_step": 31899, "epoch": 358, "lr": 9.672268433405038e-05}
{"train_loss": 0.2808570861816406, "global_step": 31900, "epoch": 358, "lr": 9.672247789558025e-05}
{"train_loss": 0.3641270101070404, "global_step": 31901, "epoch": 358, "lr": 9.672227145082883e-05}
{"train_loss": 0.2640994191169739, "global_step": 31902, "epoch": 358, "lr": 9.672206499979618e-05}
{"train_loss": 0.34481164813041687, "global_step": 31903, "epoch": 358, "lr": 9.672185854248233e-05}
{"train_loss": 0.36196431517601013, "global_step": 31904, "epoch": 358, "lr": 9.672165207888728e-05}
{"train_loss": 0.33187514543533325, "global_step": 31905, "epoch": 358, "lr": 9.672144560901106e-05}
{"train_loss": 0.3315860331058502, "global_step": 31906, "epoch": 358, "lr": 9.672123913285373e-05}
{"train_loss": 0.26291802525520325, "global_step": 31907, "epoch": 358, "lr": 9.67210326504153e-05}
{"train_loss": 0.3546714186668396, "global_step": 31908, "epoch": 358, "lr": 9.67208261616958e-05}
{"train_loss": 0.3909853994846344, "global_step": 31909, "epoch": 358, "lr": 9.672061966669526e-05}
{"train_loss": 0.3637954890727997, "global_step": 31910, "epoch": 358, "lr": 9.672041316541367e-05}
{"train_loss": 0.30550965666770935, "global_step": 31911, "epoch": 358, "lr": 9.672020665785112e-05}
{"train_loss": 0.22891290485858917, "global_step": 31912, "epoch": 358, "lr": 9.672000014400759e-05}
{"train_loss": 0.35508501529693604, "global_step": 31913, "epoch": 358, "lr": 9.671979362388314e-05}
{"train_loss": 0.30528512597084045, "global_step": 31914, "epoch": 358, "lr": 9.671958709747778e-05}
{"train_loss": 0.28400784730911255, "global_step": 31915, "epoch": 358, "lr": 9.671938056479153e-05}
{"train_loss": 0.4245222508907318, "global_step": 31916, "epoch": 358, "lr": 9.671917402582444e-05}
{"train_loss": 0.37808576226234436, "global_step": 31917, "epoch": 358, "lr": 9.671896748057652e-05}
{"train_loss": 0.34843209385871887, "global_step": 31918, "epoch": 358, "lr": 9.671876092904782e-05}
{"train_loss": 0.31020301580429077, "global_step": 31919, "epoch": 358, "lr": 9.671855437123833e-05}
{"train_loss": 0.4075535833835602, "global_step": 31920, "epoch": 358, "lr": 9.67183478071481e-05}
{"train_loss": 0.3302232623100281, "global_step": 31921, "epoch": 358, "lr": 9.671814123677719e-05}
{"train_loss": 0.35294514894485474, "global_step": 31922, "epoch": 358, "lr": 9.671793466012556e-05}
{"train_loss": 0.3782941997051239, "global_step": 31923, "epoch": 358, "lr": 9.671772807719328e-05}
{"train_loss": 0.38683071732521057, "global_step": 31924, "epoch": 358, "lr": 9.671752148798038e-05}
{"train_loss": 0.365599662065506, "global_step": 31925, "epoch": 358, "lr": 9.671731489248687e-05}
{"train_loss": 0.2715681791305542, "global_step": 31926, "epoch": 358, "lr": 9.671710829071278e-05}
{"train_loss": 0.41187193989753723, "global_step": 31927, "epoch": 358, "lr": 9.671690168265817e-05}
{"train_loss": 0.306382417678833, "global_step": 31928, "epoch": 358, "lr": 9.671669506832302e-05}
{"train_loss": 0.28141599893569946, "global_step": 31929, "epoch": 358, "lr": 9.67164884477074e-05}
{"train_loss": 0.3953265845775604, "global_step": 31930, "epoch": 358, "lr": 9.671628182081129e-05}
{"train_loss": 0.4297161102294922, "global_step": 31931, "epoch": 358, "lr": 9.671607518763476e-05}
{"train_loss": 0.3598984479904175, "global_step": 31932, "epoch": 358, "lr": 9.671586854817782e-05}
{"train_loss": 0.3284289836883545, "global_step": 31933, "epoch": 358, "lr": 9.67156619024405e-05}
{"train_loss": 0.31129327416419983, "global_step": 31934, "epoch": 358, "lr": 9.671545525042285e-05}
{"train_loss": 0.3325136601924896, "global_step": 31935, "epoch": 358, "lr": 9.671524859212486e-05}
{"train_loss": 0.4141739010810852, "global_step": 31936, "epoch": 358, "lr": 9.671504192754657e-05}
{"train_loss": 0.3982336223125458, "global_step": 31937, "epoch": 358, "lr": 9.6714835256688e-05}
{"train_loss": 0.46079495549201965, "global_step": 31938, "epoch": 358, "lr": 9.67146285795492e-05}
{"train_loss": 0.3523421883583069, "global_step": 31939, "epoch": 358, "lr": 9.67144218961302e-05}
{"train_loss": 0.44472944736480713, "global_step": 31940, "epoch": 358, "lr": 9.671421520643101e-05}
{"train_loss": 0.26667818427085876, "global_step": 31941, "epoch": 358, "lr": 9.671400851045166e-05}
{"train_loss": 0.3951887786388397, "global_step": 31942, "epoch": 358, "lr": 9.671380180819218e-05}
{"train_loss": 0.26922807097435, "global_step": 31943, "epoch": 358, "lr": 9.671359509965261e-05}
{"train_loss": 0.4490842819213867, "global_step": 31944, "epoch": 358, "lr": 9.671338838483294e-05}
{"train_loss": 0.34444770216941833, "global_step": 31945, "epoch": 358, "lr": 9.671318166373324e-05}
{"train_loss": 0.21505752205848694, "global_step": 31946, "epoch": 358, "lr": 9.671297493635354e-05}
{"train_loss": 0.28081396222114563, "global_step": 31947, "epoch": 358, "lr": 9.671276820269383e-05}
{"train_loss": 0.5147626996040344, "global_step": 31948, "epoch": 358, "lr": 9.671256146275416e-05}
{"train_loss": 0.4127166271209717, "global_step": 31949, "epoch": 358, "lr": 9.671235471653455e-05}
{"train_loss": 0.3528964665163769, "global_step": 31950, "epoch": 358, "lr": 9.671214796403505e-05, "val_loss": 2.3407304286956787}
{"train_loss": 0.40869030356407166, "global_step": 31951, "epoch": 359, "lr": 9.671194120525565e-05}
{"train_loss": 0.39145174622535706, "global_step": 31952, "epoch": 359, "lr": 9.67117344401964e-05}
{"train_loss": 0.38431861996650696, "global_step": 31953, "epoch": 359, "lr": 9.671152766885736e-05}
{"train_loss": 0.5014684200286865, "global_step": 31954, "epoch": 359, "lr": 9.671132089123848e-05}
{"train_loss": 0.3373779058456421, "global_step": 31955, "epoch": 359, "lr": 9.671111410733986e-05}
{"train_loss": 0.382526695728302, "global_step": 31956, "epoch": 359, "lr": 9.67109073171615e-05}
{"train_loss": 0.32441192865371704, "global_step": 31957, "epoch": 359, "lr": 9.671070052070342e-05}
{"train_loss": 0.32228362560272217, "global_step": 31958, "epoch": 359, "lr": 9.671049371796565e-05}
{"train_loss": 0.3599109351634979, "global_step": 31959, "epoch": 359, "lr": 9.671028690894823e-05}
{"train_loss": 0.3781851828098297, "global_step": 31960, "epoch": 359, "lr": 9.671008009365118e-05}
{"train_loss": 0.3952392637729645, "global_step": 31961, "epoch": 359, "lr": 9.670987327207452e-05}
{"train_loss": 0.31347891688346863, "global_step": 31962, "epoch": 359, "lr": 9.67096664442183e-05}
{"train_loss": 0.3281482458114624, "global_step": 31963, "epoch": 359, "lr": 9.670945961008254e-05}
{"train_loss": 0.33263668417930603, "global_step": 31964, "epoch": 359, "lr": 9.670925276966724e-05}
{"train_loss": 0.371410608291626, "global_step": 31965, "epoch": 359, "lr": 9.670904592297247e-05}
{"train_loss": 0.3789100646972656, "global_step": 31966, "epoch": 359, "lr": 9.670883906999824e-05}
{"train_loss": 0.311575710773468, "global_step": 31967, "epoch": 359, "lr": 9.670863221074457e-05}
{"train_loss": 0.334349125623703, "global_step": 31968, "epoch": 359, "lr": 9.670842534521148e-05}
{"train_loss": 0.38030222058296204, "global_step": 31969, "epoch": 359, "lr": 9.670821847339902e-05}
{"train_loss": 0.32168084383010864, "global_step": 31970, "epoch": 359, "lr": 9.670801159530723e-05}
{"train_loss": 0.46259447932243347, "global_step": 31971, "epoch": 359, "lr": 9.670780471093609e-05}
{"train_loss": 0.3535771369934082, "global_step": 31972, "epoch": 359, "lr": 9.670759782028567e-05}
{"train_loss": 0.3146316707134247, "global_step": 31973, "epoch": 359, "lr": 9.670739092335596e-05}
{"train_loss": 0.32592299580574036, "global_step": 31974, "epoch": 359, "lr": 9.670718402014704e-05}
{"train_loss": 0.35375046730041504, "global_step": 31975, "epoch": 359, "lr": 9.67069771106589e-05}
{"train_loss": 0.3937133550643921, "global_step": 31976, "epoch": 359, "lr": 9.670677019489157e-05}
{"train_loss": 0.37241747975349426, "global_step": 31977, "epoch": 359, "lr": 9.670656327284508e-05}
{"train_loss": 0.384321004152298, "global_step": 31978, "epoch": 359, "lr": 9.670635634451948e-05}
{"train_loss": 0.3846224546432495, "global_step": 31979, "epoch": 359, "lr": 9.670614940991476e-05}
{"train_loss": 0.4876585304737091, "global_step": 31980, "epoch": 359, "lr": 9.670594246903098e-05}
{"train_loss": 0.39573511481285095, "global_step": 31981, "epoch": 359, "lr": 9.670573552186815e-05}
{"train_loss": 0.32026681303977966, "global_step": 31982, "epoch": 359, "lr": 9.67055285684263e-05}
{"train_loss": 0.26536044478416443, "global_step": 31983, "epoch": 359, "lr": 9.670532160870546e-05}
{"train_loss": 0.3603871166706085, "global_step": 31984, "epoch": 359, "lr": 9.670511464270566e-05}
{"train_loss": 0.3576732277870178, "global_step": 31985, "epoch": 359, "lr": 9.670490767042694e-05}
{"train_loss": 0.29440590739250183, "global_step": 31986, "epoch": 359, "lr": 9.67047006918693e-05}
{"train_loss": 0.33519142866134644, "global_step": 31987, "epoch": 359, "lr": 9.670449370703279e-05}
{"train_loss": 0.49724438786506653, "global_step": 31988, "epoch": 359, "lr": 9.670428671591742e-05}
{"train_loss": 0.3820509612560272, "global_step": 31989, "epoch": 359, "lr": 9.670407971852325e-05}
{"train_loss": 0.3489559590816498, "global_step": 31990, "epoch": 359, "lr": 9.670387271485027e-05}
{"train_loss": 0.3725815415382385, "global_step": 31991, "epoch": 359, "lr": 9.670366570489851e-05}
{"train_loss": 0.36863934993743896, "global_step": 31992, "epoch": 359, "lr": 9.670345868866803e-05}
{"train_loss": 0.4840450882911682, "global_step": 31993, "epoch": 359, "lr": 9.670325166615883e-05}
{"train_loss": 0.3180482089519501, "global_step": 31994, "epoch": 359, "lr": 9.670304463737096e-05}
{"train_loss": 0.46661075949668884, "global_step": 31995, "epoch": 359, "lr": 9.670283760230443e-05}
{"train_loss": 0.38167792558670044, "global_step": 31996, "epoch": 359, "lr": 9.670263056095927e-05}
{"train_loss": 0.3676146864891052, "global_step": 31997, "epoch": 359, "lr": 9.670242351333551e-05}
{"train_loss": 0.3214232921600342, "global_step": 31998, "epoch": 359, "lr": 9.670221645943317e-05}
{"train_loss": 0.3672140836715698, "global_step": 31999, "epoch": 359, "lr": 9.670200939925229e-05}
{"train_loss": 0.38060906529426575, "global_step": 32000, "epoch": 359, "lr": 9.670180233279292e-05}
{"train_loss": 0.46515169739723206, "global_step": 32001, "epoch": 359, "lr": 9.670159526005503e-05}
{"train_loss": 0.3233422040939331, "global_step": 32002, "epoch": 359, "lr": 9.670138818103869e-05}
{"train_loss": 0.41264769434928894, "global_step": 32003, "epoch": 359, "lr": 9.67011810957439e-05}
{"train_loss": 0.32322096824645996, "global_step": 32004, "epoch": 359, "lr": 9.670097400417073e-05}
{"train_loss": 0.3093310594558716, "global_step": 32005, "epoch": 359, "lr": 9.670076690631919e-05}
{"train_loss": 0.35991960763931274, "global_step": 32006, "epoch": 359, "lr": 9.670055980218928e-05}
{"train_loss": 0.32082095742225647, "global_step": 32007, "epoch": 359, "lr": 9.670035269178106e-05}
{"train_loss": 0.41038867831230164, "global_step": 32008, "epoch": 359, "lr": 9.670014557509453e-05}
{"train_loss": 0.2809109091758728, "global_step": 32009, "epoch": 359, "lr": 9.669993845212976e-05}
{"train_loss": 0.28791043162345886, "global_step": 32010, "epoch": 359, "lr": 9.669973132288675e-05}
{"train_loss": 0.3715071976184845, "global_step": 32011, "epoch": 359, "lr": 9.669952418736551e-05}
{"train_loss": 0.38919198513031006, "global_step": 32012, "epoch": 359, "lr": 9.66993170455661e-05}
{"train_loss": 0.3605552911758423, "global_step": 32013, "epoch": 359, "lr": 9.669910989748853e-05}
{"train_loss": 0.3811678886413574, "global_step": 32014, "epoch": 359, "lr": 9.669890274313285e-05}
{"train_loss": 0.4212811291217804, "global_step": 32015, "epoch": 359, "lr": 9.669869558249904e-05}
{"train_loss": 0.3747797906398773, "global_step": 32016, "epoch": 359, "lr": 9.669848841558719e-05}
{"train_loss": 0.2632710635662079, "global_step": 32017, "epoch": 359, "lr": 9.669828124239728e-05}
{"train_loss": 0.30302953720092773, "global_step": 32018, "epoch": 359, "lr": 9.669807406292936e-05}
{"train_loss": 0.30319830775260925, "global_step": 32019, "epoch": 359, "lr": 9.669786687718347e-05}
{"train_loss": 0.28635403513908386, "global_step": 32020, "epoch": 359, "lr": 9.66976596851596e-05}
{"train_loss": 0.35765230655670166, "global_step": 32021, "epoch": 359, "lr": 9.66974524868578e-05}
{"train_loss": 0.4737599492073059, "global_step": 32022, "epoch": 359, "lr": 9.66972452822781e-05}
{"train_loss": 0.2854141592979431, "global_step": 32023, "epoch": 359, "lr": 9.669703807142054e-05}
{"train_loss": 0.21997642517089844, "global_step": 32024, "epoch": 359, "lr": 9.669683085428511e-05}
{"train_loss": 0.3388436734676361, "global_step": 32025, "epoch": 359, "lr": 9.669662363087188e-05}
{"train_loss": 0.3456585705280304, "global_step": 32026, "epoch": 359, "lr": 9.669641640118084e-05}
{"train_loss": 0.28928929567337036, "global_step": 32027, "epoch": 359, "lr": 9.669620916521204e-05}
{"train_loss": 0.3879055082798004, "global_step": 32028, "epoch": 359, "lr": 9.66960019229655e-05}
{"train_loss": 0.330591082572937, "global_step": 32029, "epoch": 359, "lr": 9.669579467444128e-05}
{"train_loss": 0.23324009776115417, "global_step": 32030, "epoch": 359, "lr": 9.669558741963936e-05}
{"train_loss": 0.2980465590953827, "global_step": 32031, "epoch": 359, "lr": 9.669538015855976e-05}
{"train_loss": 0.35559356212615967, "global_step": 32032, "epoch": 359, "lr": 9.669517289120258e-05}
{"train_loss": 0.27558955550193787, "global_step": 32033, "epoch": 359, "lr": 9.669496561756778e-05}
{"train_loss": 0.37383362650871277, "global_step": 32034, "epoch": 359, "lr": 9.669475833765543e-05}
{"train_loss": 0.4451896548271179, "global_step": 32035, "epoch": 359, "lr": 9.66945510514655e-05}
{"train_loss": 0.2806157171726227, "global_step": 32036, "epoch": 359, "lr": 9.669434375899808e-05}
{"train_loss": 0.34817856550216675, "global_step": 32037, "epoch": 359, "lr": 9.669413646025318e-05}
{"train_loss": 0.31429606676101685, "global_step": 32038, "epoch": 359, "lr": 9.669392915523083e-05}
{"train_loss": 0.3565086836225531, "global_step": 32039, "epoch": 359, "lr": 9.669372184393103e-05, "val_loss": 2.445732355117798}
{"train_loss": 0.4106379747390747, "global_step": 32040, "epoch": 360, "lr": 9.669351452635383e-05}
{"train_loss": 0.3506943881511688, "global_step": 32041, "epoch": 360, "lr": 9.669330720249927e-05}
{"train_loss": 0.36907562613487244, "global_step": 32042, "epoch": 360, "lr": 9.669309987236735e-05}
{"train_loss": 0.34033775329589844, "global_step": 32043, "epoch": 360, "lr": 9.669289253595812e-05}
{"train_loss": 0.2708507180213928, "global_step": 32044, "epoch": 360, "lr": 9.66926851932716e-05}
{"train_loss": 0.262006014585495, "global_step": 32045, "epoch": 360, "lr": 9.669247784430781e-05}
{"train_loss": 0.5103442072868347, "global_step": 32046, "epoch": 360, "lr": 9.669227048906679e-05}
{"train_loss": 0.38431668281555176, "global_step": 32047, "epoch": 360, "lr": 9.669206312754857e-05}
{"train_loss": 0.36219802498817444, "global_step": 32048, "epoch": 360, "lr": 9.669185575975317e-05}
{"train_loss": 0.25894060730934143, "global_step": 32049, "epoch": 360, "lr": 9.669164838568061e-05}
{"train_loss": 0.28392109274864197, "global_step": 32050, "epoch": 360, "lr": 9.669144100533094e-05}
{"train_loss": 0.3716290295124054, "global_step": 32051, "epoch": 360, "lr": 9.669123361870416e-05}
{"train_loss": 0.49538466334342957, "global_step": 32052, "epoch": 360, "lr": 9.669102622580031e-05}
{"train_loss": 0.28652840852737427, "global_step": 32053, "epoch": 360, "lr": 9.669081882661943e-05}
{"train_loss": 0.31498414278030396, "global_step": 32054, "epoch": 360, "lr": 9.669061142116154e-05}
{"train_loss": 0.3536200523376465, "global_step": 32055, "epoch": 360, "lr": 9.669040400942667e-05}
{"train_loss": 0.3170925974845886, "global_step": 32056, "epoch": 360, "lr": 9.669019659141484e-05}
{"train_loss": 0.26663726568222046, "global_step": 32057, "epoch": 360, "lr": 9.668998916712608e-05}
{"train_loss": 0.375700980424881, "global_step": 32058, "epoch": 360, "lr": 9.668978173656042e-05}
{"train_loss": 0.40583324432373047, "global_step": 32059, "epoch": 360, "lr": 9.66895742997179e-05}
{"train_loss": 0.39754846692085266, "global_step": 32060, "epoch": 360, "lr": 9.668936685659852e-05}
{"train_loss": 0.371878057718277, "global_step": 32061, "epoch": 360, "lr": 9.668915940720233e-05}
{"train_loss": 0.32263368368148804, "global_step": 32062, "epoch": 360, "lr": 9.668895195152934e-05}
{"train_loss": 0.4138188660144806, "global_step": 32063, "epoch": 360, "lr": 9.668874448957961e-05}
{"train_loss": 0.3109463155269623, "global_step": 32064, "epoch": 360, "lr": 9.668853702135313e-05}
{"train_loss": 0.33468931913375854, "global_step": 32065, "epoch": 360, "lr": 9.668832954684995e-05}
{"train_loss": 0.42856571078300476, "global_step": 32066, "epoch": 360, "lr": 9.66881220660701e-05}
{"train_loss": 0.3811832070350647, "global_step": 32067, "epoch": 360, "lr": 9.66879145790136e-05}
{"train_loss": 0.4387822449207306, "global_step": 32068, "epoch": 360, "lr": 9.668770708568047e-05}
{"train_loss": 0.34328028559684753, "global_step": 32069, "epoch": 360, "lr": 9.668749958607075e-05}
{"train_loss": 0.32539692521095276, "global_step": 32070, "epoch": 360, "lr": 9.668729208018448e-05}
{"train_loss": 0.3425219655036926, "global_step": 32071, "epoch": 360, "lr": 9.668708456802164e-05}
{"train_loss": 0.5431303977966309, "global_step": 32072, "epoch": 360, "lr": 9.668687704958232e-05}
{"train_loss": 0.48694002628326416, "global_step": 32073, "epoch": 360, "lr": 9.668666952486651e-05}
{"train_loss": 0.4075908362865448, "global_step": 32074, "epoch": 360, "lr": 9.668646199387424e-05}
{"train_loss": 0.3290170729160309, "global_step": 32075, "epoch": 360, "lr": 9.668625445660556e-05}
{"train_loss": 0.5121753811836243, "global_step": 32076, "epoch": 360, "lr": 9.668604691306047e-05}
{"train_loss": 0.3711523413658142, "global_step": 32077, "epoch": 360, "lr": 9.6685839363239e-05}
{"train_loss": 0.38340941071510315, "global_step": 32078, "epoch": 360, "lr": 9.66856318071412e-05}
{"train_loss": 0.3879346549510956, "global_step": 32079, "epoch": 360, "lr": 9.668542424476708e-05}
{"train_loss": 0.4191168248653412, "global_step": 32080, "epoch": 360, "lr": 9.668521667611669e-05}
{"train_loss": 0.3475326895713806, "global_step": 32081, "epoch": 360, "lr": 9.668500910119003e-05}
{"train_loss": 0.4199371933937073, "global_step": 32082, "epoch": 360, "lr": 9.668480151998714e-05}
{"train_loss": 0.477459579706192, "global_step": 32083, "epoch": 360, "lr": 9.668459393250804e-05}
{"train_loss": 0.37248221039772034, "global_step": 32084, "epoch": 360, "lr": 9.668438633875278e-05}
{"train_loss": 0.41420137882232666, "global_step": 32085, "epoch": 360, "lr": 9.668417873872136e-05}
{"train_loss": 0.34915947914123535, "global_step": 32086, "epoch": 360, "lr": 9.668397113241383e-05}
{"train_loss": 0.38920941948890686, "global_step": 32087, "epoch": 360, "lr": 9.66837635198302e-05}
{"train_loss": 0.3934536576271057, "global_step": 32088, "epoch": 360, "lr": 9.668355590097052e-05}
{"train_loss": 0.6658868789672852, "global_step": 32089, "epoch": 360, "lr": 9.668334827583479e-05}
{"train_loss": 0.3807925879955292, "global_step": 32090, "epoch": 360, "lr": 9.668314064442306e-05}
{"train_loss": 0.2671630084514618, "global_step": 32091, "epoch": 360, "lr": 9.668293300673537e-05}
{"train_loss": 0.33174222707748413, "global_step": 32092, "epoch": 360, "lr": 9.66827253627717e-05}
{"train_loss": 0.3774378001689911, "global_step": 32093, "epoch": 360, "lr": 9.668251771253211e-05}
{"train_loss": 0.38954681158065796, "global_step": 32094, "epoch": 360, "lr": 9.668231005601664e-05}
{"train_loss": 0.44532787799835205, "global_step": 32095, "epoch": 360, "lr": 9.668210239322528e-05}
{"train_loss": 0.46751752495765686, "global_step": 32096, "epoch": 360, "lr": 9.668189472415809e-05}
{"train_loss": 0.4909331500530243, "global_step": 32097, "epoch": 360, "lr": 9.66816870488151e-05}
{"train_loss": 0.4050550162792206, "global_step": 32098, "epoch": 360, "lr": 9.668147936719632e-05}
{"train_loss": 0.3962338864803314, "global_step": 32099, "epoch": 360, "lr": 9.668127167930178e-05}
{"train_loss": 0.4053410589694977, "global_step": 32100, "epoch": 360, "lr": 9.668106398513153e-05}
{"train_loss": 0.3384084701538086, "global_step": 32101, "epoch": 360, "lr": 9.668085628468555e-05}
{"train_loss": 0.4639071822166443, "global_step": 32102, "epoch": 360, "lr": 9.668064857796392e-05}
{"train_loss": 0.40564438700675964, "global_step": 32103, "epoch": 360, "lr": 9.668044086496663e-05}
{"train_loss": 0.3490249514579773, "global_step": 32104, "epoch": 360, "lr": 9.668023314569373e-05}
{"train_loss": 0.23941347002983093, "global_step": 32105, "epoch": 360, "lr": 9.668002542014525e-05}
{"train_loss": 0.379671186208725, "global_step": 32106, "epoch": 360, "lr": 9.66798176883212e-05}
{"train_loss": 0.39092570543289185, "global_step": 32107, "epoch": 360, "lr": 9.667960995022161e-05}
{"train_loss": 0.3325921595096588, "global_step": 32108, "epoch": 360, "lr": 9.667940220584654e-05}
{"train_loss": 0.44658786058425903, "global_step": 32109, "epoch": 360, "lr": 9.667919445519598e-05}
{"train_loss": 0.3394424617290497, "global_step": 32110, "epoch": 360, "lr": 9.667898669826996e-05}
{"train_loss": 0.3536982238292694, "global_step": 32111, "epoch": 360, "lr": 9.667877893506853e-05}
{"train_loss": 0.27745696902275085, "global_step": 32112, "epoch": 360, "lr": 9.667857116559171e-05}
{"train_loss": 0.3955097496509552, "global_step": 32113, "epoch": 360, "lr": 9.667836338983953e-05}
{"train_loss": 0.31180480122566223, "global_step": 32114, "epoch": 360, "lr": 9.6678155607812e-05}
{"train_loss": 0.3956141173839569, "global_step": 32115, "epoch": 360, "lr": 9.667794781950917e-05}
{"train_loss": 0.41900888085365295, "global_step": 32116, "epoch": 360, "lr": 9.667774002493105e-05}
{"train_loss": 0.3376573920249939, "global_step": 32117, "epoch": 360, "lr": 9.667753222407767e-05}
{"train_loss": 0.31492599844932556, "global_step": 32118, "epoch": 360, "lr": 9.667732441694908e-05}
{"train_loss": 0.43886929750442505, "global_step": 32119, "epoch": 360, "lr": 9.667711660354528e-05}
{"train_loss": 0.4341605603694916, "global_step": 32120, "epoch": 360, "lr": 9.667690878386632e-05}
{"train_loss": 0.37025055289268494, "global_step": 32121, "epoch": 360, "lr": 9.667670095791222e-05}
{"train_loss": 0.39624541997909546, "global_step": 32122, "epoch": 360, "lr": 9.667649312568301e-05}
{"train_loss": 0.2845665216445923, "global_step": 32123, "epoch": 360, "lr": 9.667628528717871e-05}
{"train_loss": 0.42851704359054565, "global_step": 32124, "epoch": 360, "lr": 9.667607744239934e-05}
{"train_loss": 0.32599377632141113, "global_step": 32125, "epoch": 360, "lr": 9.667586959134495e-05}
{"train_loss": 0.40004780888557434, "global_step": 32126, "epoch": 360, "lr": 9.667566173401557e-05}
{"train_loss": 0.33882570266723633, "global_step": 32127, "epoch": 360, "lr": 9.667545387041119e-05}
{"train_loss": 0.37816411993476784, "global_step": 32128, "epoch": 360, "lr": 9.667524600053188e-05, "val_loss": 2.332780599594116, "train_action_mse_error": 33.882301330566406}
{"train_loss": 0.3716967701911926, "global_step": 32129, "epoch": 361, "lr": 9.667503812437766e-05}
{"train_loss": 0.38016068935394287, "global_step": 32130, "epoch": 361, "lr": 9.667483024194854e-05}
{"train_loss": 0.4020611047744751, "global_step": 32131, "epoch": 361, "lr": 9.667462235324456e-05}
{"train_loss": 0.3619552254676819, "global_step": 32132, "epoch": 361, "lr": 9.667441445826575e-05}
{"train_loss": 0.3146529197692871, "global_step": 32133, "epoch": 361, "lr": 9.667420655701213e-05}
{"train_loss": 0.2932734787464142, "global_step": 32134, "epoch": 361, "lr": 9.667399864948372e-05}
{"train_loss": 0.33817076683044434, "global_step": 32135, "epoch": 361, "lr": 9.667379073568057e-05}
{"train_loss": 0.4374900460243225, "global_step": 32136, "epoch": 361, "lr": 9.66735828156027e-05}
{"train_loss": 0.46716099977493286, "global_step": 32137, "epoch": 361, "lr": 9.667337488925015e-05}
{"train_loss": 0.3712407946586609, "global_step": 32138, "epoch": 361, "lr": 9.667316695662291e-05}
{"train_loss": 0.3005388379096985, "global_step": 32139, "epoch": 361, "lr": 9.667295901772104e-05}
{"train_loss": 0.3513111174106598, "global_step": 32140, "epoch": 361, "lr": 9.667275107254456e-05}
{"train_loss": 0.40950191020965576, "global_step": 32141, "epoch": 361, "lr": 9.667254312109348e-05}
{"train_loss": 0.3066238462924957, "global_step": 32142, "epoch": 361, "lr": 9.667233516336787e-05}
{"train_loss": 0.3456559181213379, "global_step": 32143, "epoch": 361, "lr": 9.667212719936774e-05}
{"train_loss": 0.39038944244384766, "global_step": 32144, "epoch": 361, "lr": 9.66719192290931e-05}
{"train_loss": 0.34451377391815186, "global_step": 32145, "epoch": 361, "lr": 9.667171125254397e-05}
{"train_loss": 0.3517579734325409, "global_step": 32146, "epoch": 361, "lr": 9.667150326972041e-05}
{"train_loss": 0.34079617261886597, "global_step": 32147, "epoch": 361, "lr": 9.667129528062245e-05}
{"train_loss": 0.26734811067581177, "global_step": 32148, "epoch": 361, "lr": 9.667108728525008e-05}
{"train_loss": 0.3699664771556854, "global_step": 32149, "epoch": 361, "lr": 9.667087928360336e-05}
{"train_loss": 0.36047205328941345, "global_step": 32150, "epoch": 361, "lr": 9.667067127568232e-05}
{"train_loss": 0.32935193181037903, "global_step": 32151, "epoch": 361, "lr": 9.667046326148696e-05}
{"train_loss": 0.39461925625801086, "global_step": 32152, "epoch": 361, "lr": 9.667025524101734e-05}
{"train_loss": 0.3805830776691437, "global_step": 32153, "epoch": 361, "lr": 9.667004721427346e-05}
{"train_loss": 0.3595619797706604, "global_step": 32154, "epoch": 361, "lr": 9.666983918125537e-05}
{"train_loss": 0.38286060094833374, "global_step": 32155, "epoch": 361, "lr": 9.666963114196309e-05}
{"train_loss": 0.34626996517181396, "global_step": 32156, "epoch": 361, "lr": 9.666942309639664e-05}
{"train_loss": 0.36182528734207153, "global_step": 32157, "epoch": 361, "lr": 9.666921504455606e-05}
{"train_loss": 0.39518120884895325, "global_step": 32158, "epoch": 361, "lr": 9.666900698644137e-05}
{"train_loss": 0.3441009223461151, "global_step": 32159, "epoch": 361, "lr": 9.666879892205261e-05}
{"train_loss": 0.36875468492507935, "global_step": 32160, "epoch": 361, "lr": 9.666859085138979e-05}
{"train_loss": 0.2861645817756653, "global_step": 32161, "epoch": 361, "lr": 9.666838277445294e-05}
{"train_loss": 0.2988355755805969, "global_step": 32162, "epoch": 361, "lr": 9.666817469124211e-05}
{"train_loss": 0.2533935606479645, "global_step": 32163, "epoch": 361, "lr": 9.66679666017573e-05}
{"train_loss": 0.3530506491661072, "global_step": 32164, "epoch": 361, "lr": 9.666775850599856e-05}
{"train_loss": 0.34731975197792053, "global_step": 32165, "epoch": 361, "lr": 9.66675504039659e-05}
{"train_loss": 0.3780539035797119, "global_step": 32166, "epoch": 361, "lr": 9.666734229565937e-05}
{"train_loss": 0.3458418548107147, "global_step": 32167, "epoch": 361, "lr": 9.666713418107898e-05}
{"train_loss": 0.4088803231716156, "global_step": 32168, "epoch": 361, "lr": 9.666692606022475e-05}
{"train_loss": 0.3212285041809082, "global_step": 32169, "epoch": 361, "lr": 9.666671793309673e-05}
{"train_loss": 0.41935110092163086, "global_step": 32170, "epoch": 361, "lr": 9.666650979969494e-05}
{"train_loss": 0.42792561650276184, "global_step": 32171, "epoch": 361, "lr": 9.66663016600194e-05}
{"train_loss": 0.3409560024738312, "global_step": 32172, "epoch": 361, "lr": 9.666609351407014e-05}
{"train_loss": 0.38800227642059326, "global_step": 32173, "epoch": 361, "lr": 9.666588536184721e-05}
{"train_loss": 0.3987275958061218, "global_step": 32174, "epoch": 361, "lr": 9.666567720335062e-05}
{"train_loss": 0.4232521951198578, "global_step": 32175, "epoch": 361, "lr": 9.666546903858038e-05}
{"train_loss": 0.3835776448249817, "global_step": 32176, "epoch": 361, "lr": 9.666526086753654e-05}
{"train_loss": 0.305033802986145, "global_step": 32177, "epoch": 361, "lr": 9.666505269021914e-05}
{"train_loss": 0.3358365297317505, "global_step": 32178, "epoch": 361, "lr": 9.666484450662818e-05}
{"train_loss": 0.36119186878204346, "global_step": 32179, "epoch": 361, "lr": 9.66646363167637e-05}
{"train_loss": 0.47234436869621277, "global_step": 32180, "epoch": 361, "lr": 9.666442812062574e-05}
{"train_loss": 0.2828117311000824, "global_step": 32181, "epoch": 361, "lr": 9.66642199182143e-05}
{"train_loss": 0.4734392464160919, "global_step": 32182, "epoch": 361, "lr": 9.666401170952943e-05}
{"train_loss": 0.35735487937927246, "global_step": 32183, "epoch": 361, "lr": 9.666380349457116e-05}
{"train_loss": 0.3714558482170105, "global_step": 32184, "epoch": 361, "lr": 9.666359527333951e-05}
{"train_loss": 0.4237089157104492, "global_step": 32185, "epoch": 361, "lr": 9.66633870458345e-05}
{"train_loss": 0.3726600408554077, "global_step": 32186, "epoch": 361, "lr": 9.666317881205618e-05}
{"train_loss": 0.28446435928344727, "global_step": 32187, "epoch": 361, "lr": 9.666297057200455e-05}
{"train_loss": 0.42300593852996826, "global_step": 32188, "epoch": 361, "lr": 9.666276232567965e-05}
{"train_loss": 0.35107654333114624, "global_step": 32189, "epoch": 361, "lr": 9.666255407308153e-05}
{"train_loss": 0.34256288409233093, "global_step": 32190, "epoch": 361, "lr": 9.666234581421018e-05}
{"train_loss": 0.41813838481903076, "global_step": 32191, "epoch": 361, "lr": 9.666213754906566e-05}
{"train_loss": 0.3823244571685791, "global_step": 32192, "epoch": 361, "lr": 9.666192927764797e-05}
{"train_loss": 0.3909071087837219, "global_step": 32193, "epoch": 361, "lr": 9.666172099995717e-05}
{"train_loss": 0.38923048973083496, "global_step": 32194, "epoch": 361, "lr": 9.666151271599324e-05}
{"train_loss": 0.4396935999393463, "global_step": 32195, "epoch": 361, "lr": 9.666130442575627e-05}
{"train_loss": 0.4409351348876953, "global_step": 32196, "epoch": 361, "lr": 9.666109612924625e-05}
{"train_loss": 0.45974820852279663, "global_step": 32197, "epoch": 361, "lr": 9.66608878264632e-05}
{"train_loss": 0.4129515290260315, "global_step": 32198, "epoch": 361, "lr": 9.666067951740717e-05}
{"train_loss": 0.36263909935951233, "global_step": 32199, "epoch": 361, "lr": 9.666047120207817e-05}
{"train_loss": 0.3781358599662781, "global_step": 32200, "epoch": 361, "lr": 9.666026288047626e-05}
{"train_loss": 0.582822859287262, "global_step": 32201, "epoch": 361, "lr": 9.666005455260144e-05}
{"train_loss": 0.31343787908554077, "global_step": 32202, "epoch": 361, "lr": 9.665984621845375e-05}
{"train_loss": 0.3983231782913208, "global_step": 32203, "epoch": 361, "lr": 9.665963787803319e-05}
{"train_loss": 0.31356388330459595, "global_step": 32204, "epoch": 361, "lr": 9.665942953133981e-05}
{"train_loss": 0.31905481219291687, "global_step": 32205, "epoch": 361, "lr": 9.665922117837367e-05}
{"train_loss": 0.3976862132549286, "global_step": 32206, "epoch": 361, "lr": 9.665901281913475e-05}
{"train_loss": 0.4236820936203003, "global_step": 32207, "epoch": 361, "lr": 9.665880445362308e-05}
{"train_loss": 0.314158171415329, "global_step": 32208, "epoch": 361, "lr": 9.665859608183871e-05}
{"train_loss": 0.344195693731308, "global_step": 32209, "epoch": 361, "lr": 9.665838770378167e-05}
{"train_loss": 0.4562768042087555, "global_step": 32210, "epoch": 361, "lr": 9.665817931945198e-05}
{"train_loss": 0.31174856424331665, "global_step": 32211, "epoch": 361, "lr": 9.665797092884966e-05}
{"train_loss": 0.3763653635978699, "global_step": 32212, "epoch": 361, "lr": 9.665776253197474e-05}
{"train_loss": 0.48420587182044983, "global_step": 32213, "epoch": 361, "lr": 9.665755412882725e-05}
{"train_loss": 0.4304639399051666, "global_step": 32214, "epoch": 361, "lr": 9.665734571940724e-05}
{"train_loss": 0.3130572736263275, "global_step": 32215, "epoch": 361, "lr": 9.665713730371471e-05}
{"train_loss": 0.425117552280426, "global_step": 32216, "epoch": 361, "lr": 9.665692888174969e-05}
{"train_loss": 0.3716903127981036, "global_step": 32217, "epoch": 361, "lr": 9.665672045351221e-05, "val_loss": 2.3199880123138428}
{"train_loss": 0.28735995292663574, "global_step": 32218, "epoch": 362, "lr": 9.665651201900231e-05}
{"train_loss": 0.32635393738746643, "global_step": 32219, "epoch": 362, "lr": 9.665630357822001e-05}
{"train_loss": 0.39351320266723633, "global_step": 32220, "epoch": 362, "lr": 9.665609513116533e-05}
{"train_loss": 0.4233841300010681, "global_step": 32221, "epoch": 362, "lr": 9.665588667783833e-05}
{"train_loss": 0.349885493516922, "global_step": 32222, "epoch": 362, "lr": 9.6655678218239e-05}
{"train_loss": 0.3053020238876343, "global_step": 32223, "epoch": 362, "lr": 9.665546975236738e-05}
{"train_loss": 0.35774511098861694, "global_step": 32224, "epoch": 362, "lr": 9.66552612802235e-05}
{"train_loss": 0.4100416600704193, "global_step": 32225, "epoch": 362, "lr": 9.66550528018074e-05}
{"train_loss": 0.3245511054992676, "global_step": 32226, "epoch": 362, "lr": 9.665484431711908e-05}
{"train_loss": 0.34670165181159973, "global_step": 32227, "epoch": 362, "lr": 9.66546358261586e-05}
{"train_loss": 0.35504209995269775, "global_step": 32228, "epoch": 362, "lr": 9.665442732892596e-05}
{"train_loss": 0.36316531896591187, "global_step": 32229, "epoch": 362, "lr": 9.665421882542122e-05}
{"train_loss": 0.3189973533153534, "global_step": 32230, "epoch": 362, "lr": 9.665401031564438e-05}
{"train_loss": 0.36087170243263245, "global_step": 32231, "epoch": 362, "lr": 9.665380179959547e-05}
{"train_loss": 0.23157614469528198, "global_step": 32232, "epoch": 362, "lr": 9.665359327727453e-05}
{"train_loss": 0.4227665662765503, "global_step": 32233, "epoch": 362, "lr": 9.665338474868157e-05}
{"train_loss": 0.5203894376754761, "global_step": 32234, "epoch": 362, "lr": 9.665317621381665e-05}
{"train_loss": 0.41714105010032654, "global_step": 32235, "epoch": 362, "lr": 9.665296767267977e-05}
{"train_loss": 0.38628479838371277, "global_step": 32236, "epoch": 362, "lr": 9.665275912527097e-05}
{"train_loss": 0.3364274501800537, "global_step": 32237, "epoch": 362, "lr": 9.665255057159027e-05}
{"train_loss": 0.4048352539539337, "global_step": 32238, "epoch": 362, "lr": 9.665234201163772e-05}
{"train_loss": 0.3232623040676117, "global_step": 32239, "epoch": 362, "lr": 9.665213344541331e-05}
{"train_loss": 0.3469228148460388, "global_step": 32240, "epoch": 362, "lr": 9.66519248729171e-05}
{"train_loss": 0.282191663980484, "global_step": 32241, "epoch": 362, "lr": 9.665171629414911e-05}
{"train_loss": 0.3218536972999573, "global_step": 32242, "epoch": 362, "lr": 9.665150770910935e-05}
{"train_loss": 0.3809753358364105, "global_step": 32243, "epoch": 362, "lr": 9.665129911779788e-05}
{"train_loss": 0.2709963321685791, "global_step": 32244, "epoch": 362, "lr": 9.66510905202147e-05}
{"train_loss": 0.3945322036743164, "global_step": 32245, "epoch": 362, "lr": 9.665088191635986e-05}
{"train_loss": 0.3443319499492645, "global_step": 32246, "epoch": 362, "lr": 9.665067330623336e-05}
{"train_loss": 0.4163632392883301, "global_step": 32247, "epoch": 362, "lr": 9.665046468983527e-05}
{"train_loss": 0.422320693731308, "global_step": 32248, "epoch": 362, "lr": 9.665025606716558e-05}
{"train_loss": 0.414159893989563, "global_step": 32249, "epoch": 362, "lr": 9.665004743822433e-05}
{"train_loss": 0.4699614942073822, "global_step": 32250, "epoch": 362, "lr": 9.664983880301155e-05}
{"train_loss": 0.3765111267566681, "global_step": 32251, "epoch": 362, "lr": 9.664963016152728e-05}
{"train_loss": 0.36239689588546753, "global_step": 32252, "epoch": 362, "lr": 9.664942151377152e-05}
{"train_loss": 0.401950865983963, "global_step": 32253, "epoch": 362, "lr": 9.664921285974432e-05}
{"train_loss": 0.3596811890602112, "global_step": 32254, "epoch": 362, "lr": 9.66490041994457e-05}
{"train_loss": 0.4399479925632477, "global_step": 32255, "epoch": 362, "lr": 9.664879553287569e-05}
{"train_loss": 0.46469423174858093, "global_step": 32256, "epoch": 362, "lr": 9.664858686003432e-05}
{"train_loss": 0.37850087881088257, "global_step": 32257, "epoch": 362, "lr": 9.664837818092162e-05}
{"train_loss": 0.29635903239250183, "global_step": 32258, "epoch": 362, "lr": 9.664816949553761e-05}
{"train_loss": 0.35335344076156616, "global_step": 32259, "epoch": 362, "lr": 9.664796080388232e-05}
{"train_loss": 0.28307852149009705, "global_step": 32260, "epoch": 362, "lr": 9.664775210595579e-05}
{"train_loss": 0.4130258560180664, "global_step": 32261, "epoch": 362, "lr": 9.664754340175804e-05}
{"train_loss": 0.36662566661834717, "global_step": 32262, "epoch": 362, "lr": 9.664733469128909e-05}
{"train_loss": 0.2858428955078125, "global_step": 32263, "epoch": 362, "lr": 9.664712597454896e-05}
{"train_loss": 0.24601931869983673, "global_step": 32264, "epoch": 362, "lr": 9.66469172515377e-05}
{"train_loss": 0.3620787560939789, "global_step": 32265, "epoch": 362, "lr": 9.664670852225535e-05}
{"train_loss": 0.24609094858169556, "global_step": 32266, "epoch": 362, "lr": 9.664649978670189e-05}
{"train_loss": 0.3774992823600769, "global_step": 32267, "epoch": 362, "lr": 9.664629104487739e-05}
{"train_loss": 0.3926599323749542, "global_step": 32268, "epoch": 362, "lr": 9.664608229678187e-05}
{"train_loss": 0.4476504325866699, "global_step": 32269, "epoch": 362, "lr": 9.664587354241534e-05}
{"train_loss": 0.3549644947052002, "global_step": 32270, "epoch": 362, "lr": 9.664566478177785e-05}
{"train_loss": 0.4872177243232727, "global_step": 32271, "epoch": 362, "lr": 9.664545601486941e-05}
{"train_loss": 0.3840488791465759, "global_step": 32272, "epoch": 362, "lr": 9.664524724169007e-05}
{"train_loss": 0.30032414197921753, "global_step": 32273, "epoch": 362, "lr": 9.664503846223984e-05}
{"train_loss": 0.26771536469459534, "global_step": 32274, "epoch": 362, "lr": 9.664482967651875e-05}
{"train_loss": 0.30217793583869934, "global_step": 32275, "epoch": 362, "lr": 9.664462088452683e-05}
{"train_loss": 0.27058663964271545, "global_step": 32276, "epoch": 362, "lr": 9.664441208626411e-05}
{"train_loss": 0.36439278721809387, "global_step": 32277, "epoch": 362, "lr": 9.664420328173063e-05}
{"train_loss": 0.4481784403324127, "global_step": 32278, "epoch": 362, "lr": 9.66439944709264e-05}
{"train_loss": 0.3307240307331085, "global_step": 32279, "epoch": 362, "lr": 9.664378565385145e-05}
{"train_loss": 0.29606014490127563, "global_step": 32280, "epoch": 362, "lr": 9.66435768305058e-05}
{"train_loss": 0.41387221217155457, "global_step": 32281, "epoch": 362, "lr": 9.664336800088952e-05}
{"train_loss": 0.3725090026855469, "global_step": 32282, "epoch": 362, "lr": 9.664315916500258e-05}
{"train_loss": 0.3206666111946106, "global_step": 32283, "epoch": 362, "lr": 9.664295032284504e-05}
{"train_loss": 0.3903309106826782, "global_step": 32284, "epoch": 362, "lr": 9.664274147441691e-05}
{"train_loss": 0.37366002798080444, "global_step": 32285, "epoch": 362, "lr": 9.664253261971827e-05}
{"train_loss": 0.3709004521369934, "global_step": 32286, "epoch": 362, "lr": 9.664232375874909e-05}
{"train_loss": 0.49601468443870544, "global_step": 32287, "epoch": 362, "lr": 9.66421148915094e-05}
{"train_loss": 0.34167012572288513, "global_step": 32288, "epoch": 362, "lr": 9.664190601799927e-05}
{"train_loss": 0.29587218165397644, "global_step": 32289, "epoch": 362, "lr": 9.66416971382187e-05}
{"train_loss": 0.38071542978286743, "global_step": 32290, "epoch": 362, "lr": 9.664148825216773e-05}
{"train_loss": 0.40278348326683044, "global_step": 32291, "epoch": 362, "lr": 9.664127935984636e-05}
{"train_loss": 0.35931938886642456, "global_step": 32292, "epoch": 362, "lr": 9.664107046125465e-05}
{"train_loss": 0.3002430200576782, "global_step": 32293, "epoch": 362, "lr": 9.664086155639262e-05}
{"train_loss": 0.2967700958251953, "global_step": 32294, "epoch": 362, "lr": 9.664065264526029e-05}
{"train_loss": 0.4284617602825165, "global_step": 32295, "epoch": 362, "lr": 9.664044372785769e-05}
{"train_loss": 0.29269158840179443, "global_step": 32296, "epoch": 362, "lr": 9.664023480418485e-05}
{"train_loss": 0.4800715446472168, "global_step": 32297, "epoch": 362, "lr": 9.664002587424182e-05}
{"train_loss": 0.3394242227077484, "global_step": 32298, "epoch": 362, "lr": 9.663981693802857e-05}
{"train_loss": 0.41550934314727783, "global_step": 32299, "epoch": 362, "lr": 9.66396079955452e-05}
{"train_loss": 0.3861449062824249, "global_step": 32300, "epoch": 362, "lr": 9.663939904679167e-05}
{"train_loss": 0.40747514367103577, "global_step": 32301, "epoch": 362, "lr": 9.663919009176807e-05}
{"train_loss": 0.31617122888565063, "global_step": 32302, "epoch": 362, "lr": 9.663898113047438e-05}
{"train_loss": 0.22807227075099945, "global_step": 32303, "epoch": 362, "lr": 9.663877216291066e-05}
{"train_loss": 0.38116469979286194, "global_step": 32304, "epoch": 362, "lr": 9.663856318907691e-05}
{"train_loss": 0.22162757813930511, "global_step": 32305, "epoch": 362, "lr": 9.663835420897319e-05}
{"train_loss": 0.35887611732724006, "global_step": 32306, "epoch": 362, "lr": 9.663814522259951e-05, "val_loss": 2.3965446949005127}
{"train_loss": 0.40619468688964844, "global_step": 32307, "epoch": 363, "lr": 9.66379362299559e-05}
{"train_loss": 0.37300270795822144, "global_step": 32308, "epoch": 363, "lr": 9.663772723104238e-05}
{"train_loss": 0.3104347884654999, "global_step": 32309, "epoch": 363, "lr": 9.6637518225859e-05}
{"train_loss": 0.3988218307495117, "global_step": 32310, "epoch": 363, "lr": 9.663730921440575e-05}
{"train_loss": 0.399120569229126, "global_step": 32311, "epoch": 363, "lr": 9.66371001966827e-05}
{"train_loss": 0.4548860192298889, "global_step": 32312, "epoch": 363, "lr": 9.663689117268986e-05}
{"train_loss": 0.4199429154396057, "global_step": 32313, "epoch": 363, "lr": 9.663668214242725e-05}
{"train_loss": 0.3114296793937683, "global_step": 32314, "epoch": 363, "lr": 9.663647310589492e-05}
{"train_loss": 0.3537779748439789, "global_step": 32315, "epoch": 363, "lr": 9.663626406309287e-05}
{"train_loss": 0.3917844891548157, "global_step": 32316, "epoch": 363, "lr": 9.663605501402116e-05}
{"train_loss": 0.33209607005119324, "global_step": 32317, "epoch": 363, "lr": 9.663584595867979e-05}
{"train_loss": 0.42219528555870056, "global_step": 32318, "epoch": 363, "lr": 9.663563689706879e-05}
{"train_loss": 0.38315367698669434, "global_step": 32319, "epoch": 363, "lr": 9.663542782918821e-05}
{"train_loss": 0.31726107001304626, "global_step": 32320, "epoch": 363, "lr": 9.663521875503806e-05}
{"train_loss": 0.3823070228099823, "global_step": 32321, "epoch": 363, "lr": 9.663500967461836e-05}
{"train_loss": 0.2562391757965088, "global_step": 32322, "epoch": 363, "lr": 9.663480058792917e-05}
{"train_loss": 0.3263321816921234, "global_step": 32323, "epoch": 363, "lr": 9.66345914949705e-05}
{"train_loss": 0.38472262024879456, "global_step": 32324, "epoch": 363, "lr": 9.663438239574237e-05}
{"train_loss": 0.3929681181907654, "global_step": 32325, "epoch": 363, "lr": 9.663417329024482e-05}
{"train_loss": 0.43135058879852295, "global_step": 32326, "epoch": 363, "lr": 9.663396417847787e-05}
{"train_loss": 0.30922484397888184, "global_step": 32327, "epoch": 363, "lr": 9.663375506044156e-05}
{"train_loss": 0.412513792514801, "global_step": 32328, "epoch": 363, "lr": 9.66335459361359e-05}
{"train_loss": 0.44264858961105347, "global_step": 32329, "epoch": 363, "lr": 9.663333680556093e-05}
{"train_loss": 0.33140867948532104, "global_step": 32330, "epoch": 363, "lr": 9.663312766871668e-05}
{"train_loss": 0.2923828959465027, "global_step": 32331, "epoch": 363, "lr": 9.663291852560317e-05}
{"train_loss": 0.3744514286518097, "global_step": 32332, "epoch": 363, "lr": 9.663270937622044e-05}
{"train_loss": 0.3409661054611206, "global_step": 32333, "epoch": 363, "lr": 9.663250022056852e-05}
{"train_loss": 0.358910471200943, "global_step": 32334, "epoch": 363, "lr": 9.663229105864739e-05}
{"train_loss": 0.34639525413513184, "global_step": 32335, "epoch": 363, "lr": 9.663208189045716e-05}
{"train_loss": 0.29665297269821167, "global_step": 32336, "epoch": 363, "lr": 9.663187271599778e-05}
{"train_loss": 0.33387500047683716, "global_step": 32337, "epoch": 363, "lr": 9.663166353526934e-05}
{"train_loss": 0.28784751892089844, "global_step": 32338, "epoch": 363, "lr": 9.663145434827183e-05}
{"train_loss": 0.3166050612926483, "global_step": 32339, "epoch": 363, "lr": 9.66312451550053e-05}
{"train_loss": 0.3101210594177246, "global_step": 32340, "epoch": 363, "lr": 9.663103595546974e-05}
{"train_loss": 0.3053574562072754, "global_step": 32341, "epoch": 363, "lr": 9.663082674966524e-05}
{"train_loss": 0.3814179599285126, "global_step": 32342, "epoch": 363, "lr": 9.663061753759177e-05}
{"train_loss": 0.3745312988758087, "global_step": 32343, "epoch": 363, "lr": 9.663040831924939e-05}
{"train_loss": 0.3376231789588928, "global_step": 32344, "epoch": 363, "lr": 9.66301990946381e-05}
{"train_loss": 0.33521443605422974, "global_step": 32345, "epoch": 363, "lr": 9.662998986375798e-05}
{"train_loss": 0.3763182759284973, "global_step": 32346, "epoch": 363, "lr": 9.662978062660901e-05}
{"train_loss": 0.33854755759239197, "global_step": 32347, "epoch": 363, "lr": 9.662957138319124e-05}
{"train_loss": 0.4691562354564667, "global_step": 32348, "epoch": 363, "lr": 9.662936213350469e-05}
{"train_loss": 0.3415268659591675, "global_step": 32349, "epoch": 363, "lr": 9.662915287754939e-05}
{"train_loss": 0.4047807455062866, "global_step": 32350, "epoch": 363, "lr": 9.662894361532537e-05}
{"train_loss": 0.28019845485687256, "global_step": 32351, "epoch": 363, "lr": 9.662873434683266e-05}
{"train_loss": 0.34401172399520874, "global_step": 32352, "epoch": 363, "lr": 9.662852507207127e-05}
{"train_loss": 0.3556339144706726, "global_step": 32353, "epoch": 363, "lr": 9.662831579104126e-05}
{"train_loss": 0.3043544292449951, "global_step": 32354, "epoch": 363, "lr": 9.662810650374263e-05}
{"train_loss": 0.3311541974544525, "global_step": 32355, "epoch": 363, "lr": 9.662789721017543e-05}
{"train_loss": 0.4126277565956116, "global_step": 32356, "epoch": 363, "lr": 9.662768791033968e-05}
{"train_loss": 0.33884087204933167, "global_step": 32357, "epoch": 363, "lr": 9.662747860423538e-05}
{"train_loss": 0.36504632234573364, "global_step": 32358, "epoch": 363, "lr": 9.662726929186261e-05}
{"train_loss": 0.34533146023750305, "global_step": 32359, "epoch": 363, "lr": 9.662705997322137e-05}
{"train_loss": 0.4450940787792206, "global_step": 32360, "epoch": 363, "lr": 9.662685064831167e-05}
{"train_loss": 0.4810883700847626, "global_step": 32361, "epoch": 363, "lr": 9.662664131713357e-05}
{"train_loss": 0.3715362250804901, "global_step": 32362, "epoch": 363, "lr": 9.66264319796871e-05}
{"train_loss": 0.30230042338371277, "global_step": 32363, "epoch": 363, "lr": 9.662622263597225e-05}
{"train_loss": 0.3822117745876312, "global_step": 32364, "epoch": 363, "lr": 9.66260132859891e-05}
{"train_loss": 0.4323103725910187, "global_step": 32365, "epoch": 363, "lr": 9.662580392973763e-05}
{"train_loss": 0.3561250567436218, "global_step": 32366, "epoch": 363, "lr": 9.662559456721789e-05}
{"train_loss": 0.4107714891433716, "global_step": 32367, "epoch": 363, "lr": 9.662538519842992e-05}
{"train_loss": 0.3682762086391449, "global_step": 32368, "epoch": 363, "lr": 9.662517582337373e-05}
{"train_loss": 0.42679792642593384, "global_step": 32369, "epoch": 363, "lr": 9.662496644204934e-05}
{"train_loss": 0.38599690794944763, "global_step": 32370, "epoch": 363, "lr": 9.66247570544568e-05}
{"train_loss": 0.3773055076599121, "global_step": 32371, "epoch": 363, "lr": 9.662454766059614e-05}
{"train_loss": 0.3387240469455719, "global_step": 32372, "epoch": 363, "lr": 9.662433826046737e-05}
{"train_loss": 0.44685184955596924, "global_step": 32373, "epoch": 363, "lr": 9.662412885407052e-05}
{"train_loss": 0.40344828367233276, "global_step": 32374, "epoch": 363, "lr": 9.662391944140563e-05}
{"train_loss": 0.452373206615448, "global_step": 32375, "epoch": 363, "lr": 9.662371002247273e-05}
{"train_loss": 0.39217594265937805, "global_step": 32376, "epoch": 363, "lr": 9.662350059727183e-05}
{"train_loss": 0.31068921089172363, "global_step": 32377, "epoch": 363, "lr": 9.662329116580297e-05}
{"train_loss": 0.43596720695495605, "global_step": 32378, "epoch": 363, "lr": 9.662308172806618e-05}
{"train_loss": 0.4588059186935425, "global_step": 32379, "epoch": 363, "lr": 9.662287228406148e-05}
{"train_loss": 0.33819013833999634, "global_step": 32380, "epoch": 363, "lr": 9.662266283378891e-05}
{"train_loss": 0.4885246157646179, "global_step": 32381, "epoch": 363, "lr": 9.662245337724849e-05}
{"train_loss": 0.3097010850906372, "global_step": 32382, "epoch": 363, "lr": 9.662224391444024e-05}
{"train_loss": 0.35001400113105774, "global_step": 32383, "epoch": 363, "lr": 9.66220344453642e-05}
{"train_loss": 0.43790724873542786, "global_step": 32384, "epoch": 363, "lr": 9.66218249700204e-05}
{"train_loss": 0.3696518838405609, "global_step": 32385, "epoch": 363, "lr": 9.662161548840886e-05}
{"train_loss": 0.3709125816822052, "global_step": 32386, "epoch": 363, "lr": 9.662140600052962e-05}
{"train_loss": 0.39776474237442017, "global_step": 32387, "epoch": 363, "lr": 9.662119650638271e-05}
{"train_loss": 0.2541298568248749, "global_step": 32388, "epoch": 363, "lr": 9.662098700596812e-05}
{"train_loss": 0.3597770929336548, "global_step": 32389, "epoch": 363, "lr": 9.662077749928593e-05}
{"train_loss": 0.3546617031097412, "global_step": 32390, "epoch": 363, "lr": 9.662056798633613e-05}
{"train_loss": 0.37095901370048523, "global_step": 32391, "epoch": 363, "lr": 9.662035846711876e-05}
{"train_loss": 0.40875929594039917, "global_step": 32392, "epoch": 363, "lr": 9.662014894163387e-05}
{"train_loss": 0.44118255376815796, "global_step": 32393, "epoch": 363, "lr": 9.661993940988146e-05}
{"train_loss": 0.3252089023590088, "global_step": 32394, "epoch": 363, "lr": 9.661972987186156e-05}
{"train_loss": 0.36917175034458716, "global_step": 32395, "epoch": 363, "lr": 9.661952032757422e-05, "val_loss": 2.381032943725586}
{"train_loss": 0.30842339992523193, "global_step": 32396, "epoch": 364, "lr": 9.661931077701944e-05}
{"train_loss": 0.4419548213481903, "global_step": 32397, "epoch": 364, "lr": 9.661910122019727e-05}
{"train_loss": 0.41952940821647644, "global_step": 32398, "epoch": 364, "lr": 9.661889165710773e-05}
{"train_loss": 0.3544921875, "global_step": 32399, "epoch": 364, "lr": 9.661868208775084e-05}
{"train_loss": 0.3047349750995636, "global_step": 32400, "epoch": 364, "lr": 9.661847251212665e-05}
{"train_loss": 0.28696638345718384, "global_step": 32401, "epoch": 364, "lr": 9.661826293023516e-05}
{"train_loss": 0.3631186783313751, "global_step": 32402, "epoch": 364, "lr": 9.661805334207642e-05}
{"train_loss": 0.38237836956977844, "global_step": 32403, "epoch": 364, "lr": 9.661784374765046e-05}
{"train_loss": 0.48197945952415466, "global_step": 32404, "epoch": 364, "lr": 9.661763414695728e-05}
{"train_loss": 0.2845815420150757, "global_step": 32405, "epoch": 364, "lr": 9.661742453999693e-05}
{"train_loss": 0.2850770056247711, "global_step": 32406, "epoch": 364, "lr": 9.661721492676946e-05}
{"train_loss": 0.35336077213287354, "global_step": 32407, "epoch": 364, "lr": 9.661700530727484e-05}
{"train_loss": 0.4064968228340149, "global_step": 32408, "epoch": 364, "lr": 9.661679568151315e-05}
{"train_loss": 0.45308753848075867, "global_step": 32409, "epoch": 364, "lr": 9.66165860494844e-05}
{"train_loss": 0.37878766655921936, "global_step": 32410, "epoch": 364, "lr": 9.661637641118862e-05}
{"train_loss": 0.38973090052604675, "global_step": 32411, "epoch": 364, "lr": 9.661616676662583e-05}
{"train_loss": 0.3244866728782654, "global_step": 32412, "epoch": 364, "lr": 9.661595711579607e-05}
{"train_loss": 0.4151163101196289, "global_step": 32413, "epoch": 364, "lr": 9.661574745869934e-05}
{"train_loss": 0.38124093413352966, "global_step": 32414, "epoch": 364, "lr": 9.661553779533573e-05}
{"train_loss": 0.37515416741371155, "global_step": 32415, "epoch": 364, "lr": 9.661532812570519e-05}
{"train_loss": 0.3388127386569977, "global_step": 32416, "epoch": 364, "lr": 9.661511844980782e-05}
{"train_loss": 0.27285999059677124, "global_step": 32417, "epoch": 364, "lr": 9.66149087676436e-05}
{"train_loss": 0.3976302444934845, "global_step": 32418, "epoch": 364, "lr": 9.661469907921256e-05}
{"train_loss": 0.3709143102169037, "global_step": 32419, "epoch": 364, "lr": 9.661448938451477e-05}
{"train_loss": 0.356023907661438, "global_step": 32420, "epoch": 364, "lr": 9.661427968355021e-05}
{"train_loss": 0.4086129367351532, "global_step": 32421, "epoch": 364, "lr": 9.661406997631895e-05}
{"train_loss": 0.355457603931427, "global_step": 32422, "epoch": 364, "lr": 9.661386026282097e-05}
{"train_loss": 0.3254100978374481, "global_step": 32423, "epoch": 364, "lr": 9.661365054305635e-05}
{"train_loss": 0.42320582270622253, "global_step": 32424, "epoch": 364, "lr": 9.661344081702509e-05}
{"train_loss": 0.2500060796737671, "global_step": 32425, "epoch": 364, "lr": 9.661323108472719e-05}
{"train_loss": 0.4374102056026459, "global_step": 32426, "epoch": 364, "lr": 9.661302134616273e-05}
{"train_loss": 0.4537518620491028, "global_step": 32427, "epoch": 364, "lr": 9.661281160133172e-05}
{"train_loss": 0.40268492698669434, "global_step": 32428, "epoch": 364, "lr": 9.661260185023419e-05}
{"train_loss": 0.3332037627696991, "global_step": 32429, "epoch": 364, "lr": 9.661239209287015e-05}
{"train_loss": 0.27849501371383667, "global_step": 32430, "epoch": 364, "lr": 9.661218232923965e-05}
{"train_loss": 0.34813475608825684, "global_step": 32431, "epoch": 364, "lr": 9.661197255934271e-05}
{"train_loss": 0.3512713313102722, "global_step": 32432, "epoch": 364, "lr": 9.661176278317937e-05}
{"train_loss": 0.3675958514213562, "global_step": 32433, "epoch": 364, "lr": 9.661155300074962e-05}
{"train_loss": 0.32924729585647583, "global_step": 32434, "epoch": 364, "lr": 9.661134321205353e-05}
{"train_loss": 0.3621918559074402, "global_step": 32435, "epoch": 364, "lr": 9.661113341709111e-05}
{"train_loss": 0.37372520565986633, "global_step": 32436, "epoch": 364, "lr": 9.661092361586239e-05}
{"train_loss": 0.45796698331832886, "global_step": 32437, "epoch": 364, "lr": 9.661071380836742e-05}
{"train_loss": 0.3682411015033722, "global_step": 32438, "epoch": 364, "lr": 9.661050399460618e-05}
{"train_loss": 0.4070528447628021, "global_step": 32439, "epoch": 364, "lr": 9.661029417457874e-05}
{"train_loss": 0.3906649053096771, "global_step": 32440, "epoch": 364, "lr": 9.66100843482851e-05}
{"train_loss": 0.3216177225112915, "global_step": 32441, "epoch": 364, "lr": 9.660987451572532e-05}
{"train_loss": 0.5476210117340088, "global_step": 32442, "epoch": 364, "lr": 9.66096646768994e-05}
{"train_loss": 0.4543951451778412, "global_step": 32443, "epoch": 364, "lr": 9.660945483180737e-05}
{"train_loss": 0.28818637132644653, "global_step": 32444, "epoch": 364, "lr": 9.66092449804493e-05}
{"train_loss": 0.3688618242740631, "global_step": 32445, "epoch": 364, "lr": 9.660903512282515e-05}
{"train_loss": 0.3832360506057739, "global_step": 32446, "epoch": 364, "lr": 9.6608825258935e-05}
{"train_loss": 0.4940139055252075, "global_step": 32447, "epoch": 364, "lr": 9.660861538877886e-05}
{"train_loss": 0.3536301553249359, "global_step": 32448, "epoch": 364, "lr": 9.660840551235675e-05}
{"train_loss": 0.38645410537719727, "global_step": 32449, "epoch": 364, "lr": 9.660819562966872e-05}
{"train_loss": 0.31776636838912964, "global_step": 32450, "epoch": 364, "lr": 9.66079857407148e-05}
{"train_loss": 0.29150915145874023, "global_step": 32451, "epoch": 364, "lr": 9.660777584549496e-05}
{"train_loss": 0.35687753558158875, "global_step": 32452, "epoch": 364, "lr": 9.66075659440093e-05}
{"train_loss": 0.469312846660614, "global_step": 32453, "epoch": 364, "lr": 9.660735603625784e-05}
{"train_loss": 0.3245021104812622, "global_step": 32454, "epoch": 364, "lr": 9.660714612224056e-05}
{"train_loss": 0.36862272024154663, "global_step": 32455, "epoch": 364, "lr": 9.660693620195753e-05}
{"train_loss": 0.39023271203041077, "global_step": 32456, "epoch": 364, "lr": 9.660672627540875e-05}
{"train_loss": 0.33521729707717896, "global_step": 32457, "epoch": 364, "lr": 9.660651634259428e-05}
{"train_loss": 0.43565648794174194, "global_step": 32458, "epoch": 364, "lr": 9.660630640351414e-05}
{"train_loss": 0.4256104826927185, "global_step": 32459, "epoch": 364, "lr": 9.660609645816833e-05}
{"train_loss": 0.4474140405654907, "global_step": 32460, "epoch": 364, "lr": 9.660588650655691e-05}
{"train_loss": 0.32511869072914124, "global_step": 32461, "epoch": 364, "lr": 9.660567654867989e-05}
{"train_loss": 0.4002862274646759, "global_step": 32462, "epoch": 364, "lr": 9.660546658453731e-05}
{"train_loss": 0.501859188079834, "global_step": 32463, "epoch": 364, "lr": 9.660525661412918e-05}
{"train_loss": 0.3994980454444885, "global_step": 32464, "epoch": 364, "lr": 9.660504663745555e-05}
{"train_loss": 0.22000212967395782, "global_step": 32465, "epoch": 364, "lr": 9.660483665451645e-05}
{"train_loss": 0.34409454464912415, "global_step": 32466, "epoch": 364, "lr": 9.66046266653119e-05}
{"train_loss": 0.48296216130256653, "global_step": 32467, "epoch": 364, "lr": 9.66044166698419e-05}
{"train_loss": 0.32223331928253174, "global_step": 32468, "epoch": 364, "lr": 9.660420666810652e-05}
{"train_loss": 0.48067715764045715, "global_step": 32469, "epoch": 364, "lr": 9.660399666010577e-05}
{"train_loss": 0.3820033371448517, "global_step": 32470, "epoch": 364, "lr": 9.660378664583969e-05}
{"train_loss": 0.3815515637397766, "global_step": 32471, "epoch": 364, "lr": 9.660357662530828e-05}
{"train_loss": 0.34028804302215576, "global_step": 32472, "epoch": 364, "lr": 9.660336659851161e-05}
{"train_loss": 0.373878538608551, "global_step": 32473, "epoch": 364, "lr": 9.660315656544967e-05}
{"train_loss": 0.3355850577354431, "global_step": 32474, "epoch": 364, "lr": 9.66029465261225e-05}
{"train_loss": 0.35460060834884644, "global_step": 32475, "epoch": 364, "lr": 9.660273648053015e-05}
{"train_loss": 0.5104864239692688, "global_step": 32476, "epoch": 364, "lr": 9.660252642867261e-05}
{"train_loss": 0.2732197642326355, "global_step": 32477, "epoch": 364, "lr": 9.660231637054994e-05}
{"train_loss": 0.39705711603164673, "global_step": 32478, "epoch": 364, "lr": 9.660210630616215e-05}
{"train_loss": 0.40184494853019714, "global_step": 32479, "epoch": 364, "lr": 9.660189623550929e-05}
{"train_loss": 0.2872695028781891, "global_step": 32480, "epoch": 364, "lr": 9.660168615859135e-05}
{"train_loss": 0.3824349045753479, "global_step": 32481, "epoch": 364, "lr": 9.660147607540839e-05}
{"train_loss": 0.3705669939517975, "global_step": 32482, "epoch": 364, "lr": 9.660126598596045e-05}
{"train_loss": 0.31412941217422485, "global_step": 32483, "epoch": 364, "lr": 9.66010558902475e-05}
{"train_loss": 0.3730728366736616, "global_step": 32484, "epoch": 364, "lr": 9.660084578826964e-05, "val_loss": 2.3145575523376465}
{"train_loss": 0.29907286167144775, "global_step": 32485, "epoch": 365, "lr": 9.660063568002684e-05}
{"train_loss": 0.3490639328956604, "global_step": 32486, "epoch": 365, "lr": 9.660042556551917e-05}
{"train_loss": 0.4156951308250427, "global_step": 32487, "epoch": 365, "lr": 9.660021544474664e-05}
{"train_loss": 0.3527064919471741, "global_step": 32488, "epoch": 365, "lr": 9.660000531770928e-05}
{"train_loss": 0.5466354489326477, "global_step": 32489, "epoch": 365, "lr": 9.659979518440709e-05}
{"train_loss": 0.259238064289093, "global_step": 32490, "epoch": 365, "lr": 9.659958504484016e-05}
{"train_loss": 0.39798638224601746, "global_step": 32491, "epoch": 365, "lr": 9.659937489900847e-05}
{"train_loss": 0.33813390135765076, "global_step": 32492, "epoch": 365, "lr": 9.659916474691206e-05}
{"train_loss": 0.2638848125934601, "global_step": 32493, "epoch": 365, "lr": 9.659895458855096e-05}
{"train_loss": 0.3151295781135559, "global_step": 32494, "epoch": 365, "lr": 9.65987444239252e-05}
{"train_loss": 0.4099401533603668, "global_step": 32495, "epoch": 365, "lr": 9.65985342530348e-05}
{"train_loss": 0.5181588530540466, "global_step": 32496, "epoch": 365, "lr": 9.65983240758798e-05}
{"train_loss": 0.39125537872314453, "global_step": 32497, "epoch": 365, "lr": 9.659811389246023e-05}
{"train_loss": 0.31093958020210266, "global_step": 32498, "epoch": 365, "lr": 9.65979037027761e-05}
{"train_loss": 0.31527236104011536, "global_step": 32499, "epoch": 365, "lr": 9.659769350682747e-05}
{"train_loss": 0.47892284393310547, "global_step": 32500, "epoch": 365, "lr": 9.659748330461433e-05}
{"train_loss": 0.4954245984554291, "global_step": 32501, "epoch": 365, "lr": 9.659727309613673e-05}
{"train_loss": 0.34926795959472656, "global_step": 32502, "epoch": 365, "lr": 9.65970628813947e-05}
{"train_loss": 0.416320264339447, "global_step": 32503, "epoch": 365, "lr": 9.659685266038825e-05}
{"train_loss": 0.37928980588912964, "global_step": 32504, "epoch": 365, "lr": 9.659664243311744e-05}
{"train_loss": 0.3488406240940094, "global_step": 32505, "epoch": 365, "lr": 9.659643219958227e-05}
{"train_loss": 0.3015598654747009, "global_step": 32506, "epoch": 365, "lr": 9.659622195978277e-05}
{"train_loss": 0.2991718351840973, "global_step": 32507, "epoch": 365, "lr": 9.659601171371899e-05}
{"train_loss": 0.33519184589385986, "global_step": 32508, "epoch": 365, "lr": 9.659580146139094e-05}
{"train_loss": 0.29742667078971863, "global_step": 32509, "epoch": 365, "lr": 9.659559120279864e-05}
{"train_loss": 0.3150772452354431, "global_step": 32510, "epoch": 365, "lr": 9.659538093794214e-05}
{"train_loss": 0.25021615624427795, "global_step": 32511, "epoch": 365, "lr": 9.659517066682146e-05}
{"train_loss": 0.28248846530914307, "global_step": 32512, "epoch": 365, "lr": 9.659496038943663e-05}
{"train_loss": 0.38554421067237854, "global_step": 32513, "epoch": 365, "lr": 9.659475010578767e-05}
{"train_loss": 0.35830599069595337, "global_step": 32514, "epoch": 365, "lr": 9.659453981587462e-05}
{"train_loss": 0.3640427887439728, "global_step": 32515, "epoch": 365, "lr": 9.659432951969749e-05}
{"train_loss": 0.3170109689235687, "global_step": 32516, "epoch": 365, "lr": 9.659411921725632e-05}
{"train_loss": 0.377311646938324, "global_step": 32517, "epoch": 365, "lr": 9.659390890855115e-05}
{"train_loss": 0.32541584968566895, "global_step": 32518, "epoch": 365, "lr": 9.6593698593582e-05}
{"train_loss": 0.3044761121273041, "global_step": 32519, "epoch": 365, "lr": 9.659348827234889e-05}
{"train_loss": 0.38037043809890747, "global_step": 32520, "epoch": 365, "lr": 9.659327794485184e-05}
{"train_loss": 0.33002570271492004, "global_step": 32521, "epoch": 365, "lr": 9.659306761109091e-05}
{"train_loss": 0.24705727398395538, "global_step": 32522, "epoch": 365, "lr": 9.65928572710661e-05}
{"train_loss": 0.31533849239349365, "global_step": 32523, "epoch": 365, "lr": 9.659264692477745e-05}
{"train_loss": 0.34802481532096863, "global_step": 32524, "epoch": 365, "lr": 9.6592436572225e-05}
{"train_loss": 0.49370628595352173, "global_step": 32525, "epoch": 365, "lr": 9.659222621340875e-05}
{"train_loss": 0.3624174892902374, "global_step": 32526, "epoch": 365, "lr": 9.659201584832874e-05}
{"train_loss": 0.353837251663208, "global_step": 32527, "epoch": 365, "lr": 9.659180547698501e-05}
{"train_loss": 0.39230239391326904, "global_step": 32528, "epoch": 365, "lr": 9.659159509937758e-05}
{"train_loss": 0.37327665090560913, "global_step": 32529, "epoch": 365, "lr": 9.659138471550648e-05}
{"train_loss": 0.44866669178009033, "global_step": 32530, "epoch": 365, "lr": 9.659117432537172e-05}
{"train_loss": 0.46694067120552063, "global_step": 32531, "epoch": 365, "lr": 9.659096392897335e-05}
{"train_loss": 0.4362494647502899, "global_step": 32532, "epoch": 365, "lr": 9.659075352631141e-05}
{"train_loss": 0.3466584086418152, "global_step": 32533, "epoch": 365, "lr": 9.659054311738589e-05}
{"train_loss": 0.28961682319641113, "global_step": 32534, "epoch": 365, "lr": 9.659033270219686e-05}
{"train_loss": 0.4707931578159332, "global_step": 32535, "epoch": 365, "lr": 9.659012228074431e-05}
{"train_loss": 0.29129713773727417, "global_step": 32536, "epoch": 365, "lr": 9.658991185302829e-05}
{"train_loss": 0.37002262473106384, "global_step": 32537, "epoch": 365, "lr": 9.658970141904883e-05}
{"train_loss": 0.31673091650009155, "global_step": 32538, "epoch": 365, "lr": 9.658949097880594e-05}
{"train_loss": 0.33999454975128174, "global_step": 32539, "epoch": 365, "lr": 9.658928053229968e-05}
{"train_loss": 0.4111199975013733, "global_step": 32540, "epoch": 365, "lr": 9.658907007953004e-05}
{"train_loss": 0.39979618787765503, "global_step": 32541, "epoch": 365, "lr": 9.658885962049709e-05}
{"train_loss": 0.31653526425361633, "global_step": 32542, "epoch": 365, "lr": 9.658864915520082e-05}
{"train_loss": 0.39573007822036743, "global_step": 32543, "epoch": 365, "lr": 9.658843868364127e-05}
{"train_loss": 0.3879883289337158, "global_step": 32544, "epoch": 365, "lr": 9.658822820581848e-05}
{"train_loss": 0.41272106766700745, "global_step": 32545, "epoch": 365, "lr": 9.658801772173247e-05}
{"train_loss": 0.34078162908554077, "global_step": 32546, "epoch": 365, "lr": 9.658780723138326e-05}
{"train_loss": 0.39071154594421387, "global_step": 32547, "epoch": 365, "lr": 9.658759673477089e-05}
{"train_loss": 0.4138364791870117, "global_step": 32548, "epoch": 365, "lr": 9.65873862318954e-05}
{"train_loss": 0.37881261110305786, "global_step": 32549, "epoch": 365, "lr": 9.65871757227568e-05}
{"train_loss": 0.36763283610343933, "global_step": 32550, "epoch": 365, "lr": 9.658696520735511e-05}
{"train_loss": 0.26741600036621094, "global_step": 32551, "epoch": 365, "lr": 9.658675468569038e-05}
{"train_loss": 0.33310699462890625, "global_step": 32552, "epoch": 365, "lr": 9.658654415776262e-05}
{"train_loss": 0.42734599113464355, "global_step": 32553, "epoch": 365, "lr": 9.658633362357187e-05}
{"train_loss": 0.46887120604515076, "global_step": 32554, "epoch": 365, "lr": 9.658612308311816e-05}
{"train_loss": 0.3236224055290222, "global_step": 32555, "epoch": 365, "lr": 9.65859125364015e-05}
{"train_loss": 0.3552137017250061, "global_step": 32556, "epoch": 365, "lr": 9.658570198342195e-05}
{"train_loss": 0.37128502130508423, "global_step": 32557, "epoch": 365, "lr": 9.65854914241795e-05}
{"train_loss": 0.3655686676502228, "global_step": 32558, "epoch": 365, "lr": 9.658528085867422e-05}
{"train_loss": 0.37454625964164734, "global_step": 32559, "epoch": 365, "lr": 9.658507028690611e-05}
{"train_loss": 0.49263933300971985, "global_step": 32560, "epoch": 365, "lr": 9.658485970887521e-05}
{"train_loss": 0.35542312264442444, "global_step": 32561, "epoch": 365, "lr": 9.658464912458153e-05}
{"train_loss": 0.3184434771537781, "global_step": 32562, "epoch": 365, "lr": 9.658443853402513e-05}
{"train_loss": 0.48366081714630127, "global_step": 32563, "epoch": 365, "lr": 9.6584227937206e-05}
{"train_loss": 0.3250615894794464, "global_step": 32564, "epoch": 365, "lr": 9.658401733412419e-05}
{"train_loss": 0.36660483479499817, "global_step": 32565, "epoch": 365, "lr": 9.658380672477973e-05}
{"train_loss": 0.36474505066871643, "global_step": 32566, "epoch": 365, "lr": 9.658359610917265e-05}
{"train_loss": 0.4086416959762573, "global_step": 32567, "epoch": 365, "lr": 9.658338548730297e-05}
{"train_loss": 0.2566528916358948, "global_step": 32568, "epoch": 365, "lr": 9.658317485917072e-05}
{"train_loss": 0.3051525056362152, "global_step": 32569, "epoch": 365, "lr": 9.658296422477594e-05}
{"train_loss": 0.4590194821357727, "global_step": 32570, "epoch": 365, "lr": 9.658275358411864e-05}
{"train_loss": 0.317137211561203, "global_step": 32571, "epoch": 365, "lr": 9.658254293719887e-05}
{"train_loss": 0.4229045808315277, "global_step": 32572, "epoch": 365, "lr": 9.658233228401662e-05}
{"train_loss": 0.3662272777115361, "global_step": 32573, "epoch": 365, "lr": 9.658212162457195e-05, "val_loss": 2.3538997173309326, "train_action_mse_error": 22.45408058166504}
{"train_loss": 0.3123449981212616, "global_step": 32574, "epoch": 366, "lr": 9.658191095886489e-05}
{"train_loss": 0.333792120218277, "global_step": 32575, "epoch": 366, "lr": 9.658170028689546e-05}
{"train_loss": 0.3228902518749237, "global_step": 32576, "epoch": 366, "lr": 9.658148960866369e-05}
{"train_loss": 0.43696892261505127, "global_step": 32577, "epoch": 366, "lr": 9.65812789241696e-05}
{"train_loss": 0.41863858699798584, "global_step": 32578, "epoch": 366, "lr": 9.658106823341323e-05}
{"train_loss": 0.42040950059890747, "global_step": 32579, "epoch": 366, "lr": 9.658085753639458e-05}
{"train_loss": 0.36331164836883545, "global_step": 32580, "epoch": 366, "lr": 9.658064683311374e-05}
{"train_loss": 0.2731577754020691, "global_step": 32581, "epoch": 366, "lr": 9.658043612357068e-05}
{"train_loss": 0.49901965260505676, "global_step": 32582, "epoch": 366, "lr": 9.658022540776544e-05}
{"train_loss": 0.4446547329425812, "global_step": 32583, "epoch": 366, "lr": 9.658001468569807e-05}
{"train_loss": 0.26773515343666077, "global_step": 32584, "epoch": 366, "lr": 9.657980395736857e-05}
{"train_loss": 0.3447212278842926, "global_step": 32585, "epoch": 366, "lr": 9.6579593222777e-05}
{"train_loss": 0.34346407651901245, "global_step": 32586, "epoch": 366, "lr": 9.657938248192337e-05}
{"train_loss": 0.4563177525997162, "global_step": 32587, "epoch": 366, "lr": 9.657917173480769e-05}
{"train_loss": 0.30250000953674316, "global_step": 32588, "epoch": 366, "lr": 9.657896098143002e-05}
{"train_loss": 0.29897353053092957, "global_step": 32589, "epoch": 366, "lr": 9.65787502217904e-05}
{"train_loss": 0.33050718903541565, "global_step": 32590, "epoch": 366, "lr": 9.65785394558888e-05}
{"train_loss": 0.345151424407959, "global_step": 32591, "epoch": 366, "lr": 9.657832868372529e-05}
{"train_loss": 0.2908948063850403, "global_step": 32592, "epoch": 366, "lr": 9.65781179052999e-05}
{"train_loss": 0.36992961168289185, "global_step": 32593, "epoch": 366, "lr": 9.657790712061266e-05}
{"train_loss": 0.4210659861564636, "global_step": 32594, "epoch": 366, "lr": 9.657769632966357e-05}
{"train_loss": 0.2677001953125, "global_step": 32595, "epoch": 366, "lr": 9.657748553245267e-05}
{"train_loss": 0.37163129448890686, "global_step": 32596, "epoch": 366, "lr": 9.657727472898001e-05}
{"train_loss": 0.39240339398384094, "global_step": 32597, "epoch": 366, "lr": 9.657706391924561e-05}
{"train_loss": 0.31736618280410767, "global_step": 32598, "epoch": 366, "lr": 9.657685310324948e-05}
{"train_loss": 0.335033655166626, "global_step": 32599, "epoch": 366, "lr": 9.657664228099166e-05}
{"train_loss": 0.4909287989139557, "global_step": 32600, "epoch": 366, "lr": 9.657643145247217e-05}
{"train_loss": 0.23670828342437744, "global_step": 32601, "epoch": 366, "lr": 9.657622061769107e-05}
{"train_loss": 0.3280879557132721, "global_step": 32602, "epoch": 366, "lr": 9.657600977664834e-05}
{"train_loss": 0.4138853847980499, "global_step": 32603, "epoch": 366, "lr": 9.657579892934406e-05}
{"train_loss": 0.31683579087257385, "global_step": 32604, "epoch": 366, "lr": 9.65755880757782e-05}
{"train_loss": 0.3936762809753418, "global_step": 32605, "epoch": 366, "lr": 9.657537721595085e-05}
{"train_loss": 0.3495998680591583, "global_step": 32606, "epoch": 366, "lr": 9.657516634986199e-05}
{"train_loss": 0.3378926217556, "global_step": 32607, "epoch": 366, "lr": 9.657495547751166e-05}
{"train_loss": 0.3912408649921417, "global_step": 32608, "epoch": 366, "lr": 9.65747445988999e-05}
{"train_loss": 0.2851748466491699, "global_step": 32609, "epoch": 366, "lr": 9.657453371402674e-05}
{"train_loss": 0.3023943305015564, "global_step": 32610, "epoch": 366, "lr": 9.65743228228922e-05}
{"train_loss": 0.3256455957889557, "global_step": 32611, "epoch": 366, "lr": 9.657411192549631e-05}
{"train_loss": 0.36266472935676575, "global_step": 32612, "epoch": 366, "lr": 9.65739010218391e-05}
{"train_loss": 0.4018624424934387, "global_step": 32613, "epoch": 366, "lr": 9.657369011192058e-05}
{"train_loss": 0.25895705819129944, "global_step": 32614, "epoch": 366, "lr": 9.657347919574083e-05}
{"train_loss": 0.2754180431365967, "global_step": 32615, "epoch": 366, "lr": 9.657326827329983e-05}
{"train_loss": 0.3638015687465668, "global_step": 32616, "epoch": 366, "lr": 9.65730573445976e-05}
{"train_loss": 0.3163946568965912, "global_step": 32617, "epoch": 366, "lr": 9.65728464096342e-05}
{"train_loss": 0.287578284740448, "global_step": 32618, "epoch": 366, "lr": 9.657263546840966e-05}
{"train_loss": 0.2687506079673767, "global_step": 32619, "epoch": 366, "lr": 9.657242452092398e-05}
{"train_loss": 0.32145199179649353, "global_step": 32620, "epoch": 366, "lr": 9.657221356717723e-05}
{"train_loss": 0.3713897168636322, "global_step": 32621, "epoch": 366, "lr": 9.65720026071694e-05}
{"train_loss": 0.3209119439125061, "global_step": 32622, "epoch": 366, "lr": 9.657179164090053e-05}
{"train_loss": 0.4403587281703949, "global_step": 32623, "epoch": 366, "lr": 9.657158066837064e-05}
{"train_loss": 0.2632319927215576, "global_step": 32624, "epoch": 366, "lr": 9.657136968957978e-05}
{"train_loss": 0.4799058437347412, "global_step": 32625, "epoch": 366, "lr": 9.657115870452797e-05}
{"train_loss": 0.36790981888771057, "global_step": 32626, "epoch": 366, "lr": 9.657094771321524e-05}
{"train_loss": 0.30407941341400146, "global_step": 32627, "epoch": 366, "lr": 9.657073671564161e-05}
{"train_loss": 0.3862920105457306, "global_step": 32628, "epoch": 366, "lr": 9.657052571180709e-05}
{"train_loss": 0.3712279796600342, "global_step": 32629, "epoch": 366, "lr": 9.657031470171176e-05}
{"train_loss": 0.4125717878341675, "global_step": 32630, "epoch": 366, "lr": 9.657010368535561e-05}
{"train_loss": 0.3818528652191162, "global_step": 32631, "epoch": 366, "lr": 9.656989266273867e-05}
{"train_loss": 0.4203258454799652, "global_step": 32632, "epoch": 366, "lr": 9.656968163386097e-05}
{"train_loss": 0.3299645483493805, "global_step": 32633, "epoch": 366, "lr": 9.656947059872256e-05}
{"train_loss": 0.3466489613056183, "global_step": 32634, "epoch": 366, "lr": 9.656925955732345e-05}
{"train_loss": 0.32535409927368164, "global_step": 32635, "epoch": 366, "lr": 9.656904850966366e-05}
{"train_loss": 0.386435329914093, "global_step": 32636, "epoch": 366, "lr": 9.656883745574324e-05}
{"train_loss": 0.3991662859916687, "global_step": 32637, "epoch": 366, "lr": 9.65686263955622e-05}
{"train_loss": 0.3069988787174225, "global_step": 32638, "epoch": 366, "lr": 9.656841532912057e-05}
{"train_loss": 0.3340912461280823, "global_step": 32639, "epoch": 366, "lr": 9.656820425641841e-05}
{"train_loss": 0.37214162945747375, "global_step": 32640, "epoch": 366, "lr": 9.65679931774557e-05}
{"train_loss": 0.32680967450141907, "global_step": 32641, "epoch": 366, "lr": 9.65677820922325e-05}
{"train_loss": 0.3683010935783386, "global_step": 32642, "epoch": 366, "lr": 9.656757100074883e-05}
{"train_loss": 0.3795093595981598, "global_step": 32643, "epoch": 366, "lr": 9.65673599030047e-05}
{"train_loss": 0.33345088362693787, "global_step": 32644, "epoch": 366, "lr": 9.656714879900018e-05}
{"train_loss": 0.29854968190193176, "global_step": 32645, "epoch": 366, "lr": 9.656693768873526e-05}
{"train_loss": 0.3852946162223816, "global_step": 32646, "epoch": 366, "lr": 9.656672657220997e-05}
{"train_loss": 0.34042105078697205, "global_step": 32647, "epoch": 366, "lr": 9.656651544942438e-05}
{"train_loss": 0.3805626630783081, "global_step": 32648, "epoch": 366, "lr": 9.656630432037848e-05}
{"train_loss": 0.5648062229156494, "global_step": 32649, "epoch": 366, "lr": 9.65660931850723e-05}
{"train_loss": 0.4243505895137787, "global_step": 32650, "epoch": 366, "lr": 9.656588204350588e-05}
{"train_loss": 0.374908447265625, "global_step": 32651, "epoch": 366, "lr": 9.656567089567926e-05}
{"train_loss": 0.3373119831085205, "global_step": 32652, "epoch": 366, "lr": 9.656545974159243e-05}
{"train_loss": 0.30387601256370544, "global_step": 32653, "epoch": 366, "lr": 9.656524858124545e-05}
{"train_loss": 0.37440934777259827, "global_step": 32654, "epoch": 366, "lr": 9.656503741463833e-05}
{"train_loss": 0.30512699484825134, "global_step": 32655, "epoch": 366, "lr": 9.656482624177112e-05}
{"train_loss": 0.32710176706314087, "global_step": 32656, "epoch": 366, "lr": 9.656461506264385e-05}
{"train_loss": 0.40991440415382385, "global_step": 32657, "epoch": 366, "lr": 9.656440387725651e-05}
{"train_loss": 0.359935462474823, "global_step": 32658, "epoch": 366, "lr": 9.656419268560917e-05}
{"train_loss": 0.3373127579689026, "global_step": 32659, "epoch": 366, "lr": 9.656398148770183e-05}
{"train_loss": 0.34965112805366516, "global_step": 32660, "epoch": 366, "lr": 9.656377028353454e-05}
{"train_loss": 0.28444021940231323, "global_step": 32661, "epoch": 366, "lr": 9.656355907310731e-05}
{"train_loss": 0.3547963687543119, "global_step": 32662, "epoch": 366, "lr": 9.656334785642019e-05, "val_loss": 2.34075665473938}
{"train_loss": 0.46027809381484985, "global_step": 32663, "epoch": 367, "lr": 9.656313663347319e-05}
{"train_loss": 0.3729511499404907, "global_step": 32664, "epoch": 367, "lr": 9.656292540426633e-05}
{"train_loss": 0.42252594232559204, "global_step": 32665, "epoch": 367, "lr": 9.656271416879967e-05}
{"train_loss": 0.4216320216655731, "global_step": 32666, "epoch": 367, "lr": 9.656250292707321e-05}
{"train_loss": 0.37759873270988464, "global_step": 32667, "epoch": 367, "lr": 9.656229167908699e-05}
{"train_loss": 0.4372938871383667, "global_step": 32668, "epoch": 367, "lr": 9.656208042484105e-05}
{"train_loss": 0.43684759736061096, "global_step": 32669, "epoch": 367, "lr": 9.656186916433538e-05}
{"train_loss": 0.43771156668663025, "global_step": 32670, "epoch": 367, "lr": 9.656165789757005e-05}
{"train_loss": 0.364563524723053, "global_step": 32671, "epoch": 367, "lr": 9.656144662454509e-05}
{"train_loss": 0.3746374249458313, "global_step": 32672, "epoch": 367, "lr": 9.656123534526049e-05}
{"train_loss": 0.3696703016757965, "global_step": 32673, "epoch": 367, "lr": 9.65610240597163e-05}
{"train_loss": 0.3105781078338623, "global_step": 32674, "epoch": 367, "lr": 9.656081276791255e-05}
{"train_loss": 0.280681848526001, "global_step": 32675, "epoch": 367, "lr": 9.656060146984927e-05}
{"train_loss": 0.3747912347316742, "global_step": 32676, "epoch": 367, "lr": 9.656039016552647e-05}
{"train_loss": 0.31692683696746826, "global_step": 32677, "epoch": 367, "lr": 9.656017885494421e-05}
{"train_loss": 0.3151666820049286, "global_step": 32678, "epoch": 367, "lr": 9.65599675381025e-05}
{"train_loss": 0.39491894841194153, "global_step": 32679, "epoch": 367, "lr": 9.655975621500137e-05}
{"train_loss": 0.3935758173465729, "global_step": 32680, "epoch": 367, "lr": 9.655954488564083e-05}
{"train_loss": 0.30738693475723267, "global_step": 32681, "epoch": 367, "lr": 9.655933355002096e-05}
{"train_loss": 0.4955193102359772, "global_step": 32682, "epoch": 367, "lr": 9.655912220814173e-05}
{"train_loss": 0.29934337735176086, "global_step": 32683, "epoch": 367, "lr": 9.655891086000319e-05}
{"train_loss": 0.38216719031333923, "global_step": 32684, "epoch": 367, "lr": 9.655869950560538e-05}
{"train_loss": 0.3685133755207062, "global_step": 32685, "epoch": 367, "lr": 9.655848814494832e-05}
{"train_loss": 0.2619116008281708, "global_step": 32686, "epoch": 367, "lr": 9.655827677803205e-05}
{"train_loss": 0.45003193616867065, "global_step": 32687, "epoch": 367, "lr": 9.655806540485659e-05}
{"train_loss": 0.39727783203125, "global_step": 32688, "epoch": 367, "lr": 9.655785402542195e-05}
{"train_loss": 0.33278486132621765, "global_step": 32689, "epoch": 367, "lr": 9.655764263972818e-05}
{"train_loss": 0.35062968730926514, "global_step": 32690, "epoch": 367, "lr": 9.655743124777529e-05}
{"train_loss": 0.3162214457988739, "global_step": 32691, "epoch": 367, "lr": 9.655721984956334e-05}
{"train_loss": 0.4296378195285797, "global_step": 32692, "epoch": 367, "lr": 9.655700844509233e-05}
{"train_loss": 0.44131967425346375, "global_step": 32693, "epoch": 367, "lr": 9.655679703436229e-05}
{"train_loss": 0.30851253867149353, "global_step": 32694, "epoch": 367, "lr": 9.655658561737327e-05}
{"train_loss": 0.2941868007183075, "global_step": 32695, "epoch": 367, "lr": 9.655637419412526e-05}
{"train_loss": 0.367678701877594, "global_step": 32696, "epoch": 367, "lr": 9.655616276461834e-05}
{"train_loss": 0.32574617862701416, "global_step": 32697, "epoch": 367, "lr": 9.65559513288525e-05}
{"train_loss": 0.29475152492523193, "global_step": 32698, "epoch": 367, "lr": 9.655573988682778e-05}
{"train_loss": 0.3679690361022949, "global_step": 32699, "epoch": 367, "lr": 9.655552843854421e-05}
{"train_loss": 0.31445983052253723, "global_step": 32700, "epoch": 367, "lr": 9.655531698400182e-05}
{"train_loss": 0.3764649033546448, "global_step": 32701, "epoch": 367, "lr": 9.655510552320064e-05}
{"train_loss": 0.35295629501342773, "global_step": 32702, "epoch": 367, "lr": 9.655489405614068e-05}
{"train_loss": 0.328662246465683, "global_step": 32703, "epoch": 367, "lr": 9.655468258282198e-05}
{"train_loss": 0.29378563165664673, "global_step": 32704, "epoch": 367, "lr": 9.655447110324459e-05}
{"train_loss": 0.32302024960517883, "global_step": 32705, "epoch": 367, "lr": 9.655425961740849e-05}
{"train_loss": 0.32210972905158997, "global_step": 32706, "epoch": 367, "lr": 9.655404812531376e-05}
{"train_loss": 0.37001311779022217, "global_step": 32707, "epoch": 367, "lr": 9.65538366269604e-05}
{"train_loss": 0.3238217830657959, "global_step": 32708, "epoch": 367, "lr": 9.655362512234844e-05}
{"train_loss": 0.3313038647174835, "global_step": 32709, "epoch": 367, "lr": 9.655341361147792e-05}
{"train_loss": 0.3768889904022217, "global_step": 32710, "epoch": 367, "lr": 9.655320209434886e-05}
{"train_loss": 0.23515461385250092, "global_step": 32711, "epoch": 367, "lr": 9.655299057096128e-05}
{"train_loss": 0.49679455161094666, "global_step": 32712, "epoch": 367, "lr": 9.655277904131523e-05}
{"train_loss": 0.3393104076385498, "global_step": 32713, "epoch": 367, "lr": 9.655256750541071e-05}
{"train_loss": 0.35176360607147217, "global_step": 32714, "epoch": 367, "lr": 9.655235596324779e-05}
{"train_loss": 0.4627288281917572, "global_step": 32715, "epoch": 367, "lr": 9.655214441482644e-05}
{"train_loss": 0.49587282538414, "global_step": 32716, "epoch": 367, "lr": 9.655193286014676e-05}
{"train_loss": 0.35910987854003906, "global_step": 32717, "epoch": 367, "lr": 9.65517212992087e-05}
{"train_loss": 0.32736292481422424, "global_step": 32718, "epoch": 367, "lr": 9.655150973201235e-05}
{"train_loss": 0.44536855816841125, "global_step": 32719, "epoch": 367, "lr": 9.655129815855773e-05}
{"train_loss": 0.3721807301044464, "global_step": 32720, "epoch": 367, "lr": 9.655108657884485e-05}
{"train_loss": 0.30578646063804626, "global_step": 32721, "epoch": 367, "lr": 9.655087499287372e-05}
{"train_loss": 0.27685099840164185, "global_step": 32722, "epoch": 367, "lr": 9.65506634006444e-05}
{"train_loss": 0.29945647716522217, "global_step": 32723, "epoch": 367, "lr": 9.655045180215691e-05}
{"train_loss": 0.25874948501586914, "global_step": 32724, "epoch": 367, "lr": 9.65502401974113e-05}
{"train_loss": 0.40572768449783325, "global_step": 32725, "epoch": 367, "lr": 9.655002858640756e-05}
{"train_loss": 0.39478570222854614, "global_step": 32726, "epoch": 367, "lr": 9.654981696914575e-05}
{"train_loss": 0.32159680128097534, "global_step": 32727, "epoch": 367, "lr": 9.654960534562587e-05}
{"train_loss": 0.46346768736839294, "global_step": 32728, "epoch": 367, "lr": 9.654939371584797e-05}
{"train_loss": 0.3524132966995239, "global_step": 32729, "epoch": 367, "lr": 9.654918207981206e-05}
{"train_loss": 0.2646924555301666, "global_step": 32730, "epoch": 367, "lr": 9.65489704375182e-05}
{"train_loss": 0.32107076048851013, "global_step": 32731, "epoch": 367, "lr": 9.654875878896639e-05}
{"train_loss": 0.36778780817985535, "global_step": 32732, "epoch": 367, "lr": 9.654854713415665e-05}
{"train_loss": 0.28276559710502625, "global_step": 32733, "epoch": 367, "lr": 9.654833547308903e-05}
{"train_loss": 0.3818366825580597, "global_step": 32734, "epoch": 367, "lr": 9.654812380576357e-05}
{"train_loss": 0.24565953016281128, "global_step": 32735, "epoch": 367, "lr": 9.654791213218029e-05}
{"train_loss": 0.3839278817176819, "global_step": 32736, "epoch": 367, "lr": 9.654770045233918e-05}
{"train_loss": 0.383224219083786, "global_step": 32737, "epoch": 367, "lr": 9.654748876624031e-05}
{"train_loss": 0.4085882604122162, "global_step": 32738, "epoch": 367, "lr": 9.654727707388372e-05}
{"train_loss": 0.33667027950286865, "global_step": 32739, "epoch": 367, "lr": 9.654706537526939e-05}
{"train_loss": 0.30684614181518555, "global_step": 32740, "epoch": 367, "lr": 9.654685367039737e-05}
{"train_loss": 0.31144896149635315, "global_step": 32741, "epoch": 367, "lr": 9.654664195926772e-05}
{"train_loss": 0.47861000895500183, "global_step": 32742, "epoch": 367, "lr": 9.654643024188042e-05}
{"train_loss": 0.36012133955955505, "global_step": 32743, "epoch": 367, "lr": 9.654621851823552e-05}
{"train_loss": 0.37091362476348877, "global_step": 32744, "epoch": 367, "lr": 9.654600678833306e-05}
{"train_loss": 0.249740332365036, "global_step": 32745, "epoch": 367, "lr": 9.654579505217305e-05}
{"train_loss": 0.33278679847717285, "global_step": 32746, "epoch": 367, "lr": 9.654558330975551e-05}
{"train_loss": 0.38803204894065857, "global_step": 32747, "epoch": 367, "lr": 9.65453715610805e-05}
{"train_loss": 0.4164978265762329, "global_step": 32748, "epoch": 367, "lr": 9.654515980614804e-05}
{"train_loss": 0.26449617743492126, "global_step": 32749, "epoch": 367, "lr": 9.654494804495815e-05}
{"train_loss": 0.32168713212013245, "global_step": 32750, "epoch": 367, "lr": 9.654473627751085e-05}
{"train_loss": 0.3571256601743484, "global_step": 32751, "epoch": 367, "lr": 9.654452450380618e-05, "val_loss": 2.3289833068847656}
{"train_loss": 0.39980506896972656, "global_step": 32752, "epoch": 368, "lr": 9.654431272384416e-05}
{"train_loss": 0.3374342918395996, "global_step": 32753, "epoch": 368, "lr": 9.654410093762482e-05}
{"train_loss": 0.4345962703227997, "global_step": 32754, "epoch": 368, "lr": 9.654388914514822e-05}
{"train_loss": 0.37667325139045715, "global_step": 32755, "epoch": 368, "lr": 9.654367734641434e-05}
{"train_loss": 0.29518765211105347, "global_step": 32756, "epoch": 368, "lr": 9.654346554142323e-05}
{"train_loss": 0.3173946440219879, "global_step": 32757, "epoch": 368, "lr": 9.654325373017492e-05}
{"train_loss": 0.3665834963321686, "global_step": 32758, "epoch": 368, "lr": 9.654304191266943e-05}
{"train_loss": 0.27903831005096436, "global_step": 32759, "epoch": 368, "lr": 9.654283008890682e-05}
{"train_loss": 0.41944319009780884, "global_step": 32760, "epoch": 368, "lr": 9.654261825888706e-05}
{"train_loss": 0.3657340705394745, "global_step": 32761, "epoch": 368, "lr": 9.654240642261023e-05}
{"train_loss": 0.2784181833267212, "global_step": 32762, "epoch": 368, "lr": 9.654219458007635e-05}
{"train_loss": 0.3160305321216583, "global_step": 32763, "epoch": 368, "lr": 9.654198273128542e-05}
{"train_loss": 0.2990228831768036, "global_step": 32764, "epoch": 368, "lr": 9.65417708762375e-05}
{"train_loss": 0.32462623715400696, "global_step": 32765, "epoch": 368, "lr": 9.65415590149326e-05}
{"train_loss": 0.2885521352291107, "global_step": 32766, "epoch": 368, "lr": 9.654134714737078e-05}
{"train_loss": 0.335492879152298, "global_step": 32767, "epoch": 368, "lr": 9.654113527355201e-05}
{"train_loss": 0.399950236082077, "global_step": 32768, "epoch": 368, "lr": 9.654092339347635e-05}
{"train_loss": 0.4446204900741577, "global_step": 32769, "epoch": 368, "lr": 9.654071150714386e-05}
{"train_loss": 0.308609277009964, "global_step": 32770, "epoch": 368, "lr": 9.654049961455451e-05}
{"train_loss": 0.3750382363796234, "global_step": 32771, "epoch": 368, "lr": 9.654028771570838e-05}
{"train_loss": 0.303501695394516, "global_step": 32772, "epoch": 368, "lr": 9.654007581060547e-05}
{"train_loss": 0.3308003842830658, "global_step": 32773, "epoch": 368, "lr": 9.65398638992458e-05}
{"train_loss": 0.27747294306755066, "global_step": 32774, "epoch": 368, "lr": 9.653965198162942e-05}
{"train_loss": 0.40472546219825745, "global_step": 32775, "epoch": 368, "lr": 9.653944005775635e-05}
{"train_loss": 0.328453004360199, "global_step": 32776, "epoch": 368, "lr": 9.653922812762663e-05}
{"train_loss": 0.3163013458251953, "global_step": 32777, "epoch": 368, "lr": 9.653901619124026e-05}
{"train_loss": 0.2948243319988251, "global_step": 32778, "epoch": 368, "lr": 9.653880424859729e-05}
{"train_loss": 0.3731142282485962, "global_step": 32779, "epoch": 368, "lr": 9.653859229969776e-05}
{"train_loss": 0.3666926622390747, "global_step": 32780, "epoch": 368, "lr": 9.653838034454169e-05}
{"train_loss": 0.41858360171318054, "global_step": 32781, "epoch": 368, "lr": 9.653816838312908e-05}
{"train_loss": 0.5174773335456848, "global_step": 32782, "epoch": 368, "lr": 9.653795641545997e-05}
{"train_loss": 0.2991432249546051, "global_step": 32783, "epoch": 368, "lr": 9.653774444153443e-05}
{"train_loss": 0.43930280208587646, "global_step": 32784, "epoch": 368, "lr": 9.653753246135245e-05}
{"train_loss": 0.26197299361228943, "global_step": 32785, "epoch": 368, "lr": 9.653732047491404e-05}
{"train_loss": 0.3657159209251404, "global_step": 32786, "epoch": 368, "lr": 9.653710848221929e-05}
{"train_loss": 0.41877394914627075, "global_step": 32787, "epoch": 368, "lr": 9.653689648326817e-05}
{"train_loss": 0.444537490606308, "global_step": 32788, "epoch": 368, "lr": 9.653668447806074e-05}
{"train_loss": 0.3666747510433197, "global_step": 32789, "epoch": 368, "lr": 9.653647246659701e-05}
{"train_loss": 0.43990761041641235, "global_step": 32790, "epoch": 368, "lr": 9.653626044887702e-05}
{"train_loss": 0.35002821683883667, "global_step": 32791, "epoch": 368, "lr": 9.653604842490082e-05}
{"train_loss": 0.4776276648044586, "global_step": 32792, "epoch": 368, "lr": 9.653583639466839e-05}
{"train_loss": 0.3709890842437744, "global_step": 32793, "epoch": 368, "lr": 9.65356243581798e-05}
{"train_loss": 0.33185818791389465, "global_step": 32794, "epoch": 368, "lr": 9.653541231543506e-05}
{"train_loss": 0.418784499168396, "global_step": 32795, "epoch": 368, "lr": 9.653520026643418e-05}
{"train_loss": 0.3476552665233612, "global_step": 32796, "epoch": 368, "lr": 9.653498821117723e-05}
{"train_loss": 0.415014386177063, "global_step": 32797, "epoch": 368, "lr": 9.653477614966421e-05}
{"train_loss": 0.40944910049438477, "global_step": 32798, "epoch": 368, "lr": 9.653456408189516e-05}
{"train_loss": 0.3289409577846527, "global_step": 32799, "epoch": 368, "lr": 9.653435200787011e-05}
{"train_loss": 0.2762744128704071, "global_step": 32800, "epoch": 368, "lr": 9.653413992758908e-05}
{"train_loss": 0.26402536034584045, "global_step": 32801, "epoch": 368, "lr": 9.653392784105211e-05}
{"train_loss": 0.24645519256591797, "global_step": 32802, "epoch": 368, "lr": 9.65337157482592e-05}
{"train_loss": 0.3743506968021393, "global_step": 32803, "epoch": 368, "lr": 9.653350364921042e-05}
{"train_loss": 0.2954179048538208, "global_step": 32804, "epoch": 368, "lr": 9.653329154390576e-05}
{"train_loss": 0.39093464612960815, "global_step": 32805, "epoch": 368, "lr": 9.653307943234528e-05}
{"train_loss": 0.40411674976348877, "global_step": 32806, "epoch": 368, "lr": 9.6532867314529e-05}
{"train_loss": 0.41554105281829834, "global_step": 32807, "epoch": 368, "lr": 9.653265519045693e-05}
{"train_loss": 0.45893847942352295, "global_step": 32808, "epoch": 368, "lr": 9.65324430601291e-05}
{"train_loss": 0.35992395877838135, "global_step": 32809, "epoch": 368, "lr": 9.653223092354557e-05}
{"train_loss": 0.2928939163684845, "global_step": 32810, "epoch": 368, "lr": 9.653201878070635e-05}
{"train_loss": 0.40492233633995056, "global_step": 32811, "epoch": 368, "lr": 9.653180663161145e-05}
{"train_loss": 0.4081054925918579, "global_step": 32812, "epoch": 368, "lr": 9.653159447626093e-05}
{"train_loss": 0.36209315061569214, "global_step": 32813, "epoch": 368, "lr": 9.65313823146548e-05}
{"train_loss": 0.34621763229370117, "global_step": 32814, "epoch": 368, "lr": 9.653117014679309e-05}
{"train_loss": 0.2957867980003357, "global_step": 32815, "epoch": 368, "lr": 9.653095797267583e-05}
{"train_loss": 0.3026385009288788, "global_step": 32816, "epoch": 368, "lr": 9.653074579230306e-05}
{"train_loss": 0.3334529995918274, "global_step": 32817, "epoch": 368, "lr": 9.65305336056748e-05}
{"train_loss": 0.34965750575065613, "global_step": 32818, "epoch": 368, "lr": 9.653032141279108e-05}
{"train_loss": 0.31439676880836487, "global_step": 32819, "epoch": 368, "lr": 9.653010921365192e-05}
{"train_loss": 0.3508550822734833, "global_step": 32820, "epoch": 368, "lr": 9.652989700825733e-05}
{"train_loss": 0.34333911538124084, "global_step": 32821, "epoch": 368, "lr": 9.652968479660738e-05}
{"train_loss": 0.323059618473053, "global_step": 32822, "epoch": 368, "lr": 9.652947257870208e-05}
{"train_loss": 0.37538227438926697, "global_step": 32823, "epoch": 368, "lr": 9.652926035454146e-05}
{"train_loss": 0.3561462163925171, "global_step": 32824, "epoch": 368, "lr": 9.652904812412556e-05}
{"train_loss": 0.3190775513648987, "global_step": 32825, "epoch": 368, "lr": 9.652883588745438e-05}
{"train_loss": 0.4093237519264221, "global_step": 32826, "epoch": 368, "lr": 9.652862364452797e-05}
{"train_loss": 0.3510608375072479, "global_step": 32827, "epoch": 368, "lr": 9.652841139534636e-05}
{"train_loss": 0.42730578780174255, "global_step": 32828, "epoch": 368, "lr": 9.652819913990955e-05}
{"train_loss": 0.3527727723121643, "global_step": 32829, "epoch": 368, "lr": 9.652798687821762e-05}
{"train_loss": 0.295978307723999, "global_step": 32830, "epoch": 368, "lr": 9.652777461027056e-05}
{"train_loss": 0.43008434772491455, "global_step": 32831, "epoch": 368, "lr": 9.652756233606841e-05}
{"train_loss": 0.26978135108947754, "global_step": 32832, "epoch": 368, "lr": 9.652735005561119e-05}
{"train_loss": 0.37235885858535767, "global_step": 32833, "epoch": 368, "lr": 9.652713776889893e-05}
{"train_loss": 0.30000996589660645, "global_step": 32834, "epoch": 368, "lr": 9.652692547593168e-05}
{"train_loss": 0.2903939485549927, "global_step": 32835, "epoch": 368, "lr": 9.652671317670944e-05}
{"train_loss": 0.421263188123703, "global_step": 32836, "epoch": 368, "lr": 9.652650087123226e-05}
{"train_loss": 0.303763210773468, "global_step": 32837, "epoch": 368, "lr": 9.652628855950015e-05}
{"train_loss": 0.36224600672721863, "global_step": 32838, "epoch": 368, "lr": 9.652607624151313e-05}
{"train_loss": 0.4548444151878357, "global_step": 32839, "epoch": 368, "lr": 9.652586391727127e-05}
{"train_loss": 0.35588636324646766, "global_step": 32840, "epoch": 368, "lr": 9.652565158677458e-05, "val_loss": 2.3911585807800293}
{"train_loss": 0.34763607382774353, "global_step": 32841, "epoch": 369, "lr": 9.652543925002306e-05}
{"train_loss": 0.22866633534431458, "global_step": 32842, "epoch": 369, "lr": 9.652522690701678e-05}
{"train_loss": 0.344862163066864, "global_step": 32843, "epoch": 369, "lr": 9.652501455775575e-05}
{"train_loss": 0.2541331350803375, "global_step": 32844, "epoch": 369, "lr": 9.652480220223998e-05}
{"train_loss": 0.3406846821308136, "global_step": 32845, "epoch": 369, "lr": 9.652458984046954e-05}
{"train_loss": 0.423778235912323, "global_step": 32846, "epoch": 369, "lr": 9.652437747244443e-05}
{"train_loss": 0.3348381519317627, "global_step": 32847, "epoch": 369, "lr": 9.652416509816469e-05}
{"train_loss": 0.2972879409790039, "global_step": 32848, "epoch": 369, "lr": 9.652395271763033e-05}
{"train_loss": 0.27285364270210266, "global_step": 32849, "epoch": 369, "lr": 9.65237403308414e-05}
{"train_loss": 0.3222663104534149, "global_step": 32850, "epoch": 369, "lr": 9.652352793779791e-05}
{"train_loss": 0.4191673696041107, "global_step": 32851, "epoch": 369, "lr": 9.652331553849991e-05}
{"train_loss": 0.37767845392227173, "global_step": 32852, "epoch": 369, "lr": 9.652310313294741e-05}
{"train_loss": 0.3120703101158142, "global_step": 32853, "epoch": 369, "lr": 9.652289072114045e-05}
{"train_loss": 0.3604511618614197, "global_step": 32854, "epoch": 369, "lr": 9.652267830307904e-05}
{"train_loss": 0.35177385807037354, "global_step": 32855, "epoch": 369, "lr": 9.652246587876323e-05}
{"train_loss": 0.3670310080051422, "global_step": 32856, "epoch": 369, "lr": 9.652225344819305e-05}
{"train_loss": 0.3100406527519226, "global_step": 32857, "epoch": 369, "lr": 9.652204101136853e-05}
{"train_loss": 0.40396857261657715, "global_step": 32858, "epoch": 369, "lr": 9.652182856828967e-05}
{"train_loss": 0.3238372504711151, "global_step": 32859, "epoch": 369, "lr": 9.652161611895653e-05}
{"train_loss": 0.40446531772613525, "global_step": 32860, "epoch": 369, "lr": 9.652140366336911e-05}
{"train_loss": 0.2518317699432373, "global_step": 32861, "epoch": 369, "lr": 9.652119120152748e-05}
{"train_loss": 0.3207190930843353, "global_step": 32862, "epoch": 369, "lr": 9.652097873343161e-05}
{"train_loss": 0.3063092529773712, "global_step": 32863, "epoch": 369, "lr": 9.652076625908158e-05}
{"train_loss": 0.29659703373908997, "global_step": 32864, "epoch": 369, "lr": 9.65205537784774e-05}
{"train_loss": 0.369655579328537, "global_step": 32865, "epoch": 369, "lr": 9.652034129161909e-05}
{"train_loss": 0.21736283600330353, "global_step": 32866, "epoch": 369, "lr": 9.65201287985067e-05}
{"train_loss": 0.3868027925491333, "global_step": 32867, "epoch": 369, "lr": 9.651991629914023e-05}
{"train_loss": 0.40652570128440857, "global_step": 32868, "epoch": 369, "lr": 9.651970379351974e-05}
{"train_loss": 0.3572641909122467, "global_step": 32869, "epoch": 369, "lr": 9.651949128164524e-05}
{"train_loss": 0.3845655024051666, "global_step": 32870, "epoch": 369, "lr": 9.651927876351673e-05}
{"train_loss": 0.3445955216884613, "global_step": 32871, "epoch": 369, "lr": 9.651906623913431e-05}
{"train_loss": 0.37865200638771057, "global_step": 32872, "epoch": 369, "lr": 9.651885370849796e-05}
{"train_loss": 0.34809795022010803, "global_step": 32873, "epoch": 369, "lr": 9.651864117160771e-05}
{"train_loss": 0.3699497580528259, "global_step": 32874, "epoch": 369, "lr": 9.651842862846361e-05}
{"train_loss": 0.4420314431190491, "global_step": 32875, "epoch": 369, "lr": 9.651821607906566e-05}
{"train_loss": 0.3303147852420807, "global_step": 32876, "epoch": 369, "lr": 9.651800352341389e-05}
{"train_loss": 0.3547114133834839, "global_step": 32877, "epoch": 369, "lr": 9.651779096150836e-05}
{"train_loss": 0.3914016783237457, "global_step": 32878, "epoch": 369, "lr": 9.651757839334908e-05}
{"train_loss": 0.2139735072851181, "global_step": 32879, "epoch": 369, "lr": 9.651736581893608e-05}
{"train_loss": 0.34329360723495483, "global_step": 32880, "epoch": 369, "lr": 9.651715323826937e-05}
{"train_loss": 0.37352657318115234, "global_step": 32881, "epoch": 369, "lr": 9.651694065134902e-05}
{"train_loss": 0.3213249146938324, "global_step": 32882, "epoch": 369, "lr": 9.651672805817502e-05}
{"train_loss": 0.3549991250038147, "global_step": 32883, "epoch": 369, "lr": 9.651651545874742e-05}
{"train_loss": 0.38452258706092834, "global_step": 32884, "epoch": 369, "lr": 9.651630285306623e-05}
{"train_loss": 0.379667192697525, "global_step": 32885, "epoch": 369, "lr": 9.651609024113151e-05}
{"train_loss": 0.3685019910335541, "global_step": 32886, "epoch": 369, "lr": 9.651587762294325e-05}
{"train_loss": 0.4396193027496338, "global_step": 32887, "epoch": 369, "lr": 9.65156649985015e-05}
{"train_loss": 0.3113158047199249, "global_step": 32888, "epoch": 369, "lr": 9.651545236780628e-05}
{"train_loss": 0.39898839592933655, "global_step": 32889, "epoch": 369, "lr": 9.651523973085764e-05}
{"train_loss": 0.23530025780200958, "global_step": 32890, "epoch": 369, "lr": 9.651502708765558e-05}
{"train_loss": 0.3294607698917389, "global_step": 32891, "epoch": 369, "lr": 9.651481443820015e-05}
{"train_loss": 0.3517324924468994, "global_step": 32892, "epoch": 369, "lr": 9.651460178249137e-05}
{"train_loss": 0.3387013077735901, "global_step": 32893, "epoch": 369, "lr": 9.651438912052926e-05}
{"train_loss": 0.27333009243011475, "global_step": 32894, "epoch": 369, "lr": 9.651417645231387e-05}
{"train_loss": 0.29531383514404297, "global_step": 32895, "epoch": 369, "lr": 9.651396377784521e-05}
{"train_loss": 0.3430192768573761, "global_step": 32896, "epoch": 369, "lr": 9.65137510971233e-05}
{"train_loss": 0.26474255323410034, "global_step": 32897, "epoch": 369, "lr": 9.651353841014821e-05}
{"train_loss": 0.3567659854888916, "global_step": 32898, "epoch": 369, "lr": 9.651332571691994e-05}
{"train_loss": 0.4436766803264618, "global_step": 32899, "epoch": 369, "lr": 9.65131130174385e-05}
{"train_loss": 0.3671143054962158, "global_step": 32900, "epoch": 369, "lr": 9.651290031170395e-05}
{"train_loss": 0.3735920488834381, "global_step": 32901, "epoch": 369, "lr": 9.65126875997163e-05}
{"train_loss": 0.37698104977607727, "global_step": 32902, "epoch": 369, "lr": 9.651247488147559e-05}
{"train_loss": 0.37676963210105896, "global_step": 32903, "epoch": 369, "lr": 9.651226215698185e-05}
{"train_loss": 0.3870657682418823, "global_step": 32904, "epoch": 369, "lr": 9.651204942623509e-05}
{"train_loss": 0.2524471879005432, "global_step": 32905, "epoch": 369, "lr": 9.651183668923536e-05}
{"train_loss": 0.33680298924446106, "global_step": 32906, "epoch": 369, "lr": 9.651162394598269e-05}
{"train_loss": 0.4018027186393738, "global_step": 32907, "epoch": 369, "lr": 9.651141119647708e-05}
{"train_loss": 0.379825234413147, "global_step": 32908, "epoch": 369, "lr": 9.65111984407186e-05}
{"train_loss": 0.3626282811164856, "global_step": 32909, "epoch": 369, "lr": 9.651098567870724e-05}
{"train_loss": 0.36057889461517334, "global_step": 32910, "epoch": 369, "lr": 9.651077291044305e-05}
{"train_loss": 0.35963594913482666, "global_step": 32911, "epoch": 369, "lr": 9.651056013592605e-05}
{"train_loss": 0.25883448123931885, "global_step": 32912, "epoch": 369, "lr": 9.651034735515627e-05}
{"train_loss": 0.2775017023086548, "global_step": 32913, "epoch": 369, "lr": 9.651013456813376e-05}
{"train_loss": 0.2747996747493744, "global_step": 32914, "epoch": 369, "lr": 9.650992177485852e-05}
{"train_loss": 0.38901734352111816, "global_step": 32915, "epoch": 369, "lr": 9.650970897533057e-05}
{"train_loss": 0.371939480304718, "global_step": 32916, "epoch": 369, "lr": 9.650949616954998e-05}
{"train_loss": 0.44653695821762085, "global_step": 32917, "epoch": 369, "lr": 9.650928335751674e-05}
{"train_loss": 0.4056185483932495, "global_step": 32918, "epoch": 369, "lr": 9.650907053923089e-05}
{"train_loss": 0.27409419417381287, "global_step": 32919, "epoch": 369, "lr": 9.650885771469247e-05}
{"train_loss": 0.39183735847473145, "global_step": 32920, "epoch": 369, "lr": 9.65086448839015e-05}
{"train_loss": 0.2555941045284271, "global_step": 32921, "epoch": 369, "lr": 9.650843204685803e-05}
{"train_loss": 0.3684333562850952, "global_step": 32922, "epoch": 369, "lr": 9.650821920356203e-05}
{"train_loss": 0.379829466342926, "global_step": 32923, "epoch": 369, "lr": 9.650800635401359e-05}
{"train_loss": 0.31438612937927246, "global_step": 32924, "epoch": 369, "lr": 9.650779349821271e-05}
{"train_loss": 0.3758029341697693, "global_step": 32925, "epoch": 369, "lr": 9.650758063615944e-05}
{"train_loss": 0.3172287344932556, "global_step": 32926, "epoch": 369, "lr": 9.650736776785376e-05}
{"train_loss": 0.3554431200027466, "global_step": 32927, "epoch": 369, "lr": 9.650715489329576e-05}
{"train_loss": 0.361309677362442, "global_step": 32928, "epoch": 369, "lr": 9.650694201248542e-05}
{"train_loss": 0.34629414607299847, "global_step": 32929, "epoch": 369, "lr": 9.65067291254228e-05, "val_loss": 2.400608777999878}
{"train_loss": 0.40196090936660767, "global_step": 32930, "epoch": 370, "lr": 9.650651623210791e-05}
{"train_loss": 0.42777207493782043, "global_step": 32931, "epoch": 370, "lr": 9.650630333254081e-05}
{"train_loss": 0.34309932589530945, "global_step": 32932, "epoch": 370, "lr": 9.650609042672148e-05}
{"train_loss": 0.24042509496212006, "global_step": 32933, "epoch": 370, "lr": 9.650587751464997e-05}
{"train_loss": 0.3535412549972534, "global_step": 32934, "epoch": 370, "lr": 9.650566459632632e-05}
{"train_loss": 0.3218575119972229, "global_step": 32935, "epoch": 370, "lr": 9.650545167175055e-05}
{"train_loss": 0.28022947907447815, "global_step": 32936, "epoch": 370, "lr": 9.650523874092269e-05}
{"train_loss": 0.3001940846443176, "global_step": 32937, "epoch": 370, "lr": 9.650502580384276e-05}
{"train_loss": 0.36983633041381836, "global_step": 32938, "epoch": 370, "lr": 9.65048128605108e-05}
{"train_loss": 0.34842050075531006, "global_step": 32939, "epoch": 370, "lr": 9.650459991092684e-05}
{"train_loss": 0.2813374400138855, "global_step": 32940, "epoch": 370, "lr": 9.650438695509089e-05}
{"train_loss": 0.32754600048065186, "global_step": 32941, "epoch": 370, "lr": 9.650417399300303e-05}
{"train_loss": 0.34642863273620605, "global_step": 32942, "epoch": 370, "lr": 9.65039610246632e-05}
{"train_loss": 0.23494499921798706, "global_step": 32943, "epoch": 370, "lr": 9.65037480500715e-05}
{"train_loss": 0.4072515070438385, "global_step": 32944, "epoch": 370, "lr": 9.650353506922795e-05}
{"train_loss": 0.409816712141037, "global_step": 32945, "epoch": 370, "lr": 9.650332208213255e-05}
{"train_loss": 0.37826964259147644, "global_step": 32946, "epoch": 370, "lr": 9.650310908878535e-05}
{"train_loss": 0.33968856930732727, "global_step": 32947, "epoch": 370, "lr": 9.650289608918638e-05}
{"train_loss": 0.3902890980243683, "global_step": 32948, "epoch": 370, "lr": 9.650268308333566e-05}
{"train_loss": 0.354356974363327, "global_step": 32949, "epoch": 370, "lr": 9.65024700712332e-05}
{"train_loss": 0.36251378059387207, "global_step": 32950, "epoch": 370, "lr": 9.650225705287908e-05}
{"train_loss": 0.40541884303092957, "global_step": 32951, "epoch": 370, "lr": 9.65020440282733e-05}
{"train_loss": 0.32854223251342773, "global_step": 32952, "epoch": 370, "lr": 9.650183099741586e-05}
{"train_loss": 0.4081059694290161, "global_step": 32953, "epoch": 370, "lr": 9.650161796030683e-05}
{"train_loss": 0.31645864248275757, "global_step": 32954, "epoch": 370, "lr": 9.650140491694622e-05}
{"train_loss": 0.24317973852157593, "global_step": 32955, "epoch": 370, "lr": 9.650119186733408e-05}
{"train_loss": 0.3711933493614197, "global_step": 32956, "epoch": 370, "lr": 9.65009788114704e-05}
{"train_loss": 0.33510369062423706, "global_step": 32957, "epoch": 370, "lr": 9.650076574935524e-05}
{"train_loss": 0.3677126169204712, "global_step": 32958, "epoch": 370, "lr": 9.650055268098861e-05}
{"train_loss": 0.3341800570487976, "global_step": 32959, "epoch": 370, "lr": 9.650033960637056e-05}
{"train_loss": 0.3621469736099243, "global_step": 32960, "epoch": 370, "lr": 9.65001265255011e-05}
{"train_loss": 0.2655216455459595, "global_step": 32961, "epoch": 370, "lr": 9.649991343838026e-05}
{"train_loss": 0.38428688049316406, "global_step": 32962, "epoch": 370, "lr": 9.649970034500809e-05}
{"train_loss": 0.4459139108657837, "global_step": 32963, "epoch": 370, "lr": 9.649948724538458e-05}
{"train_loss": 0.300197958946228, "global_step": 32964, "epoch": 370, "lr": 9.649927413950978e-05}
{"train_loss": 0.36862072348594666, "global_step": 32965, "epoch": 370, "lr": 9.649906102738375e-05}
{"train_loss": 0.42824608087539673, "global_step": 32966, "epoch": 370, "lr": 9.649884790900646e-05}
{"train_loss": 0.39337158203125, "global_step": 32967, "epoch": 370, "lr": 9.649863478437797e-05}
{"train_loss": 0.4557434320449829, "global_step": 32968, "epoch": 370, "lr": 9.649842165349831e-05}
{"train_loss": 0.30612993240356445, "global_step": 32969, "epoch": 370, "lr": 9.649820851636751e-05}
{"train_loss": 0.4415748119354248, "global_step": 32970, "epoch": 370, "lr": 9.64979953729856e-05}
{"train_loss": 0.3691452145576477, "global_step": 32971, "epoch": 370, "lr": 9.649778222335257e-05}
{"train_loss": 0.498176246881485, "global_step": 32972, "epoch": 370, "lr": 9.64975690674685e-05}
{"train_loss": 0.3315715789794922, "global_step": 32973, "epoch": 370, "lr": 9.64973559053334e-05}
{"train_loss": 0.38971927762031555, "global_step": 32974, "epoch": 370, "lr": 9.64971427369473e-05}
{"train_loss": 0.35285666584968567, "global_step": 32975, "epoch": 370, "lr": 9.649692956231023e-05}
{"train_loss": 0.3161463141441345, "global_step": 32976, "epoch": 370, "lr": 9.649671638142219e-05}
{"train_loss": 0.2807542383670807, "global_step": 32977, "epoch": 370, "lr": 9.649650319428324e-05}
{"train_loss": 0.3310079574584961, "global_step": 32978, "epoch": 370, "lr": 9.649629000089342e-05}
{"train_loss": 0.38551610708236694, "global_step": 32979, "epoch": 370, "lr": 9.649607680125273e-05}
{"train_loss": 0.3907032310962677, "global_step": 32980, "epoch": 370, "lr": 9.64958635953612e-05}
{"train_loss": 0.41937655210494995, "global_step": 32981, "epoch": 370, "lr": 9.649565038321888e-05}
{"train_loss": 0.36045050621032715, "global_step": 32982, "epoch": 370, "lr": 9.649543716482579e-05}
{"train_loss": 0.29264503717422485, "global_step": 32983, "epoch": 370, "lr": 9.649522394018195e-05}
{"train_loss": 0.3542670011520386, "global_step": 32984, "epoch": 370, "lr": 9.64950107092874e-05}
{"train_loss": 0.354597806930542, "global_step": 32985, "epoch": 370, "lr": 9.649479747214214e-05}
{"train_loss": 0.3472038805484772, "global_step": 32986, "epoch": 370, "lr": 9.649458422874625e-05}
{"train_loss": 0.37460657954216003, "global_step": 32987, "epoch": 370, "lr": 9.64943709790997e-05}
{"train_loss": 0.49473515152931213, "global_step": 32988, "epoch": 370, "lr": 9.649415772320257e-05}
{"train_loss": 0.4379287660121918, "global_step": 32989, "epoch": 370, "lr": 9.649394446105487e-05}
{"train_loss": 0.3699885308742523, "global_step": 32990, "epoch": 370, "lr": 9.64937311926566e-05}
{"train_loss": 0.38402223587036133, "global_step": 32991, "epoch": 370, "lr": 9.649351791800785e-05}
{"train_loss": 0.4196915328502655, "global_step": 32992, "epoch": 370, "lr": 9.649330463710859e-05}
{"train_loss": 0.331779420375824, "global_step": 32993, "epoch": 370, "lr": 9.649309134995887e-05}
{"train_loss": 0.36872217059135437, "global_step": 32994, "epoch": 370, "lr": 9.649287805655873e-05}
{"train_loss": 0.3577670454978943, "global_step": 32995, "epoch": 370, "lr": 9.64926647569082e-05}
{"train_loss": 0.3760063350200653, "global_step": 32996, "epoch": 370, "lr": 9.649245145100727e-05}
{"train_loss": 0.34311380982398987, "global_step": 32997, "epoch": 370, "lr": 9.649223813885602e-05}
{"train_loss": 0.41290318965911865, "global_step": 32998, "epoch": 370, "lr": 9.649202482045445e-05}
{"train_loss": 0.34593555331230164, "global_step": 32999, "epoch": 370, "lr": 9.649181149580258e-05}
{"train_loss": 0.3761337697505951, "global_step": 33000, "epoch": 370, "lr": 9.649159816490046e-05}
{"train_loss": 0.31857064366340637, "global_step": 33001, "epoch": 370, "lr": 9.649138482774813e-05}
{"train_loss": 0.32319730520248413, "global_step": 33002, "epoch": 370, "lr": 9.649117148434557e-05}
{"train_loss": 0.3860364854335785, "global_step": 33003, "epoch": 370, "lr": 9.649095813469286e-05}
{"train_loss": 0.4067579209804535, "global_step": 33004, "epoch": 370, "lr": 9.649074477879e-05}
{"train_loss": 0.35082772374153137, "global_step": 33005, "epoch": 370, "lr": 9.649053141663702e-05}
{"train_loss": 0.29883134365081787, "global_step": 33006, "epoch": 370, "lr": 9.649031804823397e-05}
{"train_loss": 0.43331968784332275, "global_step": 33007, "epoch": 370, "lr": 9.649010467358084e-05}
{"train_loss": 0.4231224060058594, "global_step": 33008, "epoch": 370, "lr": 9.648989129267771e-05}
{"train_loss": 0.40716809034347534, "global_step": 33009, "epoch": 370, "lr": 9.648967790552457e-05}
{"train_loss": 0.38614198565483093, "global_step": 33010, "epoch": 370, "lr": 9.648946451212145e-05}
{"train_loss": 0.35073235630989075, "global_step": 33011, "epoch": 370, "lr": 9.64892511124684e-05}
{"train_loss": 0.29581552743911743, "global_step": 33012, "epoch": 370, "lr": 9.648903770656543e-05}
{"train_loss": 0.3633135259151459, "global_step": 33013, "epoch": 370, "lr": 9.648882429441257e-05}
{"train_loss": 0.38054901361465454, "global_step": 33014, "epoch": 370, "lr": 9.648861087600987e-05}
{"train_loss": 0.2655602991580963, "global_step": 33015, "epoch": 370, "lr": 9.648839745135734e-05}
{"train_loss": 0.3028615713119507, "global_step": 33016, "epoch": 370, "lr": 9.648818402045499e-05}
{"train_loss": 0.3191818296909332, "global_step": 33017, "epoch": 370, "lr": 9.64879705833029e-05}
{"train_loss": 0.359423189518157, "global_step": 33018, "epoch": 370, "lr": 9.648775713990104e-05, "val_loss": 2.4538156986236572, "train_action_mse_error": 22.826370239257812}
{"train_loss": 0.3103626072406769, "global_step": 33019, "epoch": 371, "lr": 9.648754369024948e-05}
{"train_loss": 0.5130261182785034, "global_step": 33020, "epoch": 371, "lr": 9.648733023434825e-05}
{"train_loss": 0.3079282343387604, "global_step": 33021, "epoch": 371, "lr": 9.648711677219735e-05}
{"train_loss": 0.42185020446777344, "global_step": 33022, "epoch": 371, "lr": 9.648690330379684e-05}
{"train_loss": 0.4368744194507599, "global_step": 33023, "epoch": 371, "lr": 9.64866898291467e-05}
{"train_loss": 0.37969550490379333, "global_step": 33024, "epoch": 371, "lr": 9.648647634824702e-05}
{"train_loss": 0.36143022775650024, "global_step": 33025, "epoch": 371, "lr": 9.648626286109778e-05}
{"train_loss": 0.3232519030570984, "global_step": 33026, "epoch": 371, "lr": 9.648604936769905e-05}
{"train_loss": 0.3569023311138153, "global_step": 33027, "epoch": 371, "lr": 9.648583586805084e-05}
{"train_loss": 0.43111374974250793, "global_step": 33028, "epoch": 371, "lr": 9.648562236215315e-05}
{"train_loss": 0.33052197098731995, "global_step": 33029, "epoch": 371, "lr": 9.648540885000605e-05}
{"train_loss": 0.3875507414340973, "global_step": 33030, "epoch": 371, "lr": 9.648519533160955e-05}
{"train_loss": 0.35114020109176636, "global_step": 33031, "epoch": 371, "lr": 9.648498180696368e-05}
{"train_loss": 0.386567085981369, "global_step": 33032, "epoch": 371, "lr": 9.648476827606848e-05}
{"train_loss": 0.359647661447525, "global_step": 33033, "epoch": 371, "lr": 9.648455473892397e-05}
{"train_loss": 0.39127784967422485, "global_step": 33034, "epoch": 371, "lr": 9.648434119553016e-05}
{"train_loss": 0.3332539200782776, "global_step": 33035, "epoch": 371, "lr": 9.648412764588711e-05}
{"train_loss": 0.2890807092189789, "global_step": 33036, "epoch": 371, "lr": 9.648391408999484e-05}
{"train_loss": 0.37506401538848877, "global_step": 33037, "epoch": 371, "lr": 9.648370052785335e-05}
{"train_loss": 0.3316308557987213, "global_step": 33038, "epoch": 371, "lr": 9.648348695946272e-05}
{"train_loss": 0.3579160273075104, "global_step": 33039, "epoch": 371, "lr": 9.648327338482295e-05}
{"train_loss": 0.4142814874649048, "global_step": 33040, "epoch": 371, "lr": 9.648305980393407e-05}
{"train_loss": 0.43505212664604187, "global_step": 33041, "epoch": 371, "lr": 9.64828462167961e-05}
{"train_loss": 0.37043213844299316, "global_step": 33042, "epoch": 371, "lr": 9.648263262340908e-05}
{"train_loss": 0.3868677020072937, "global_step": 33043, "epoch": 371, "lr": 9.648241902377302e-05}
{"train_loss": 0.3500097095966339, "global_step": 33044, "epoch": 371, "lr": 9.6482205417888e-05}
{"train_loss": 0.435247004032135, "global_step": 33045, "epoch": 371, "lr": 9.6481991805754e-05}
{"train_loss": 0.30309778451919556, "global_step": 33046, "epoch": 371, "lr": 9.648177818737106e-05}
{"train_loss": 0.2857881784439087, "global_step": 33047, "epoch": 371, "lr": 9.64815645627392e-05}
{"train_loss": 0.23179391026496887, "global_step": 33048, "epoch": 371, "lr": 9.648135093185847e-05}
{"train_loss": 0.46061480045318604, "global_step": 33049, "epoch": 371, "lr": 9.648113729472889e-05}
{"train_loss": 0.3341571092605591, "global_step": 33050, "epoch": 371, "lr": 9.648092365135048e-05}
{"train_loss": 0.3803105652332306, "global_step": 33051, "epoch": 371, "lr": 9.648071000172329e-05}
{"train_loss": 0.4056088924407959, "global_step": 33052, "epoch": 371, "lr": 9.648049634584733e-05}
{"train_loss": 0.3497985899448395, "global_step": 33053, "epoch": 371, "lr": 9.648028268372263e-05}
{"train_loss": 0.4122212529182434, "global_step": 33054, "epoch": 371, "lr": 9.648006901534923e-05}
{"train_loss": 0.4318276345729828, "global_step": 33055, "epoch": 371, "lr": 9.647985534072714e-05}
{"train_loss": 0.3787490427494049, "global_step": 33056, "epoch": 371, "lr": 9.647964165985642e-05}
{"train_loss": 0.40569522976875305, "global_step": 33057, "epoch": 371, "lr": 9.647942797273705e-05}
{"train_loss": 0.2974925637245178, "global_step": 33058, "epoch": 371, "lr": 9.64792142793691e-05}
{"train_loss": 0.36911627650260925, "global_step": 33059, "epoch": 371, "lr": 9.647900057975259e-05}
{"train_loss": 0.4168909788131714, "global_step": 33060, "epoch": 371, "lr": 9.647878687388755e-05}
{"train_loss": 0.2591974139213562, "global_step": 33061, "epoch": 371, "lr": 9.6478573161774e-05}
{"train_loss": 0.4935959577560425, "global_step": 33062, "epoch": 371, "lr": 9.647835944341197e-05}
{"train_loss": 0.3294048607349396, "global_step": 33063, "epoch": 371, "lr": 9.64781457188015e-05}
{"train_loss": 0.31618526577949524, "global_step": 33064, "epoch": 371, "lr": 9.64779319879426e-05}
{"train_loss": 0.39129579067230225, "global_step": 33065, "epoch": 371, "lr": 9.647771825083529e-05}
{"train_loss": 0.3406943380832672, "global_step": 33066, "epoch": 371, "lr": 9.647750450747963e-05}
{"train_loss": 0.42168793082237244, "global_step": 33067, "epoch": 371, "lr": 9.647729075787567e-05}
{"train_loss": 0.3131931722164154, "global_step": 33068, "epoch": 371, "lr": 9.647707700202335e-05}
{"train_loss": 0.35844072699546814, "global_step": 33069, "epoch": 371, "lr": 9.647686323992278e-05}
{"train_loss": 0.4287671148777008, "global_step": 33070, "epoch": 371, "lr": 9.647664947157398e-05}
{"train_loss": 0.3205876052379608, "global_step": 33071, "epoch": 371, "lr": 9.647643569697693e-05}
{"train_loss": 0.312778502702713, "global_step": 33072, "epoch": 371, "lr": 9.64762219161317e-05}
{"train_loss": 0.27615103125572205, "global_step": 33073, "epoch": 371, "lr": 9.647600812903833e-05}
{"train_loss": 0.3368762135505676, "global_step": 33074, "epoch": 371, "lr": 9.64757943356968e-05}
{"train_loss": 0.4258558750152588, "global_step": 33075, "epoch": 371, "lr": 9.647558053610717e-05}
{"train_loss": 0.3840685486793518, "global_step": 33076, "epoch": 371, "lr": 9.647536673026947e-05}
{"train_loss": 0.46210771799087524, "global_step": 33077, "epoch": 371, "lr": 9.647515291818372e-05}
{"train_loss": 0.4029730260372162, "global_step": 33078, "epoch": 371, "lr": 9.647493909984997e-05}
{"train_loss": 0.3584311604499817, "global_step": 33079, "epoch": 371, "lr": 9.64747252752682e-05}
{"train_loss": 0.2810339629650116, "global_step": 33080, "epoch": 371, "lr": 9.647451144443849e-05}
{"train_loss": 0.44586896896362305, "global_step": 33081, "epoch": 371, "lr": 9.647429760736085e-05}
{"train_loss": 0.28607022762298584, "global_step": 33082, "epoch": 371, "lr": 9.64740837640353e-05}
{"train_loss": 0.3315300941467285, "global_step": 33083, "epoch": 371, "lr": 9.647386991446187e-05}
{"train_loss": 0.307789146900177, "global_step": 33084, "epoch": 371, "lr": 9.647365605864061e-05}
{"train_loss": 0.2482183575630188, "global_step": 33085, "epoch": 371, "lr": 9.647344219657152e-05}
{"train_loss": 0.31924012303352356, "global_step": 33086, "epoch": 371, "lr": 9.647322832825466e-05}
{"train_loss": 0.3554370701313019, "global_step": 33087, "epoch": 371, "lr": 9.647301445369002e-05}
{"train_loss": 0.36517420411109924, "global_step": 33088, "epoch": 371, "lr": 9.647280057287764e-05}
{"train_loss": 0.2554287016391754, "global_step": 33089, "epoch": 371, "lr": 9.647258668581758e-05}
{"train_loss": 0.29047510027885437, "global_step": 33090, "epoch": 371, "lr": 9.647237279250984e-05}
{"train_loss": 0.3838214576244354, "global_step": 33091, "epoch": 371, "lr": 9.647215889295447e-05}
{"train_loss": 0.5079047679901123, "global_step": 33092, "epoch": 371, "lr": 9.647194498715147e-05}
{"train_loss": 0.27558469772338867, "global_step": 33093, "epoch": 371, "lr": 9.647173107510088e-05}
{"train_loss": 0.30856430530548096, "global_step": 33094, "epoch": 371, "lr": 9.647151715680274e-05}
{"train_loss": 0.33095985651016235, "global_step": 33095, "epoch": 371, "lr": 9.647130323225707e-05}
{"train_loss": 0.2809741497039795, "global_step": 33096, "epoch": 371, "lr": 9.647108930146391e-05}
{"train_loss": 0.3394453823566437, "global_step": 33097, "epoch": 371, "lr": 9.647087536442326e-05}
{"train_loss": 0.4578011631965637, "global_step": 33098, "epoch": 371, "lr": 9.647066142113516e-05}
{"train_loss": 0.4046730697154999, "global_step": 33099, "epoch": 371, "lr": 9.647044747159968e-05}
{"train_loss": 0.5032235980033875, "global_step": 33100, "epoch": 371, "lr": 9.64702335158168e-05}
{"train_loss": 0.3667553663253784, "global_step": 33101, "epoch": 371, "lr": 9.647001955378654e-05}
{"train_loss": 0.4073021411895752, "global_step": 33102, "epoch": 371, "lr": 9.646980558550897e-05}
{"train_loss": 0.36506742238998413, "global_step": 33103, "epoch": 371, "lr": 9.646959161098412e-05}
{"train_loss": 0.3224748969078064, "global_step": 33104, "epoch": 371, "lr": 9.646937763021197e-05}
{"train_loss": 0.41473907232284546, "global_step": 33105, "epoch": 371, "lr": 9.646916364319261e-05}
{"train_loss": 0.33346548676490784, "global_step": 33106, "epoch": 371, "lr": 9.646894964992601e-05}
{"train_loss": 0.3627003150039844, "global_step": 33107, "epoch": 371, "lr": 9.646873565041224e-05, "val_loss": 2.4003405570983887}
{"train_loss": 0.3936828076839447, "global_step": 33108, "epoch": 372, "lr": 9.64685216446513e-05}
{"train_loss": 0.32534170150756836, "global_step": 33109, "epoch": 372, "lr": 9.646830763264325e-05}
{"train_loss": 0.3149740695953369, "global_step": 33110, "epoch": 372, "lr": 9.646809361438809e-05}
{"train_loss": 0.43619197607040405, "global_step": 33111, "epoch": 372, "lr": 9.646787958988585e-05}
{"train_loss": 0.3359769284725189, "global_step": 33112, "epoch": 372, "lr": 9.64676655591366e-05}
{"train_loss": 0.294199138879776, "global_step": 33113, "epoch": 372, "lr": 9.646745152214032e-05}
{"train_loss": 0.29501426219940186, "global_step": 33114, "epoch": 372, "lr": 9.646723747889707e-05}
{"train_loss": 0.38330700993537903, "global_step": 33115, "epoch": 372, "lr": 9.646702342940685e-05}
{"train_loss": 0.2675459086894989, "global_step": 33116, "epoch": 372, "lr": 9.646680937366972e-05}
{"train_loss": 0.33513393998146057, "global_step": 33117, "epoch": 372, "lr": 9.64665953116857e-05}
{"train_loss": 0.27822497487068176, "global_step": 33118, "epoch": 372, "lr": 9.64663812434548e-05}
{"train_loss": 0.399370014667511, "global_step": 33119, "epoch": 372, "lr": 9.646616716897705e-05}
{"train_loss": 0.2475961595773697, "global_step": 33120, "epoch": 372, "lr": 9.64659530882525e-05}
{"train_loss": 0.3454936444759369, "global_step": 33121, "epoch": 372, "lr": 9.646573900128117e-05}
{"train_loss": 0.4709186553955078, "global_step": 33122, "epoch": 372, "lr": 9.646552490806309e-05}
{"train_loss": 0.3079666793346405, "global_step": 33123, "epoch": 372, "lr": 9.646531080859829e-05}
{"train_loss": 0.3652700185775757, "global_step": 33124, "epoch": 372, "lr": 9.646509670288679e-05}
{"train_loss": 0.4945615231990814, "global_step": 33125, "epoch": 372, "lr": 9.646488259092864e-05}
{"train_loss": 0.24073661863803864, "global_step": 33126, "epoch": 372, "lr": 9.646466847272382e-05}
{"train_loss": 0.3541373014450073, "global_step": 33127, "epoch": 372, "lr": 9.64644543482724e-05}
{"train_loss": 0.3815639913082123, "global_step": 33128, "epoch": 372, "lr": 9.646424021757442e-05}
{"train_loss": 0.2746109664440155, "global_step": 33129, "epoch": 372, "lr": 9.646402608062987e-05}
{"train_loss": 0.429158091545105, "global_step": 33130, "epoch": 372, "lr": 9.64638119374388e-05}
{"train_loss": 0.37664905190467834, "global_step": 33131, "epoch": 372, "lr": 9.646359778800126e-05}
{"train_loss": 0.32749590277671814, "global_step": 33132, "epoch": 372, "lr": 9.646338363231724e-05}
{"train_loss": 0.2985711693763733, "global_step": 33133, "epoch": 372, "lr": 9.646316947038678e-05}
{"train_loss": 0.27528151869773865, "global_step": 33134, "epoch": 372, "lr": 9.646295530220992e-05}
{"train_loss": 0.2577013671398163, "global_step": 33135, "epoch": 372, "lr": 9.646274112778668e-05}
{"train_loss": 0.3180796504020691, "global_step": 33136, "epoch": 372, "lr": 9.646252694711708e-05}
{"train_loss": 0.31947773694992065, "global_step": 33137, "epoch": 372, "lr": 9.646231276020117e-05}
{"train_loss": 0.39528384804725647, "global_step": 33138, "epoch": 372, "lr": 9.646209856703897e-05}
{"train_loss": 0.22404253482818604, "global_step": 33139, "epoch": 372, "lr": 9.64618843676305e-05}
{"train_loss": 0.3457140326499939, "global_step": 33140, "epoch": 372, "lr": 9.646167016197581e-05}
{"train_loss": 0.4110182523727417, "global_step": 33141, "epoch": 372, "lr": 9.646145595007489e-05}
{"train_loss": 0.3765755593776703, "global_step": 33142, "epoch": 372, "lr": 9.646124173192782e-05}
{"train_loss": 0.3238350749015808, "global_step": 33143, "epoch": 372, "lr": 9.646102750753459e-05}
{"train_loss": 0.3025476038455963, "global_step": 33144, "epoch": 372, "lr": 9.646081327689525e-05}
{"train_loss": 0.4009245038032532, "global_step": 33145, "epoch": 372, "lr": 9.646059904000981e-05}
{"train_loss": 0.2861282527446747, "global_step": 33146, "epoch": 372, "lr": 9.646038479687832e-05}
{"train_loss": 0.3509911298751831, "global_step": 33147, "epoch": 372, "lr": 9.646017054750079e-05}
{"train_loss": 0.32496821880340576, "global_step": 33148, "epoch": 372, "lr": 9.645995629187723e-05}
{"train_loss": 0.32062214612960815, "global_step": 33149, "epoch": 372, "lr": 9.645974203000774e-05}
{"train_loss": 0.3320789337158203, "global_step": 33150, "epoch": 372, "lr": 9.645952776189226e-05}
{"train_loss": 0.2593207061290741, "global_step": 33151, "epoch": 372, "lr": 9.64593134875309e-05}
{"train_loss": 0.3765048086643219, "global_step": 33152, "epoch": 372, "lr": 9.645909920692365e-05}
{"train_loss": 0.37507012486457825, "global_step": 33153, "epoch": 372, "lr": 9.645888492007052e-05}
{"train_loss": 0.45397356152534485, "global_step": 33154, "epoch": 372, "lr": 9.645867062697156e-05}
{"train_loss": 0.39129653573036194, "global_step": 33155, "epoch": 372, "lr": 9.645845632762681e-05}
{"train_loss": 0.3284780979156494, "global_step": 33156, "epoch": 372, "lr": 9.645824202203629e-05}
{"train_loss": 0.32875749468803406, "global_step": 33157, "epoch": 372, "lr": 9.645802771020001e-05}
{"train_loss": 0.31545254588127136, "global_step": 33158, "epoch": 372, "lr": 9.645781339211803e-05}
{"train_loss": 0.40506118535995483, "global_step": 33159, "epoch": 372, "lr": 9.645759906779034e-05}
{"train_loss": 0.49871376156806946, "global_step": 33160, "epoch": 372, "lr": 9.645738473721702e-05}
{"train_loss": 0.3156282901763916, "global_step": 33161, "epoch": 372, "lr": 9.645717040039806e-05}
{"train_loss": 0.35505664348602295, "global_step": 33162, "epoch": 372, "lr": 9.645695605733349e-05}
{"train_loss": 0.38054516911506653, "global_step": 33163, "epoch": 372, "lr": 9.645674170802335e-05}
{"train_loss": 0.46511679887771606, "global_step": 33164, "epoch": 372, "lr": 9.645652735246768e-05}
{"train_loss": 0.3483821749687195, "global_step": 33165, "epoch": 372, "lr": 9.645631299066649e-05}
{"train_loss": 0.4576113224029541, "global_step": 33166, "epoch": 372, "lr": 9.645609862261982e-05}
{"train_loss": 0.33719322085380554, "global_step": 33167, "epoch": 372, "lr": 9.645588424832769e-05}
{"train_loss": 0.3159511387348175, "global_step": 33168, "epoch": 372, "lr": 9.645566986779013e-05}
{"train_loss": 0.3692963719367981, "global_step": 33169, "epoch": 372, "lr": 9.645545548100717e-05}
{"train_loss": 0.3288465142250061, "global_step": 33170, "epoch": 372, "lr": 9.645524108797884e-05}
{"train_loss": 0.3072541356086731, "global_step": 33171, "epoch": 372, "lr": 9.645502668870517e-05}
{"train_loss": 0.33467191457748413, "global_step": 33172, "epoch": 372, "lr": 9.64548122831862e-05}
{"train_loss": 0.3775424659252167, "global_step": 33173, "epoch": 372, "lr": 9.645459787142192e-05}
{"train_loss": 0.4545084238052368, "global_step": 33174, "epoch": 372, "lr": 9.645438345341242e-05}
{"train_loss": 0.2915899455547333, "global_step": 33175, "epoch": 372, "lr": 9.645416902915768e-05}
{"train_loss": 0.2955699563026428, "global_step": 33176, "epoch": 372, "lr": 9.645395459865774e-05}
{"train_loss": 0.3170326352119446, "global_step": 33177, "epoch": 372, "lr": 9.645374016191262e-05}
{"train_loss": 0.3998512923717499, "global_step": 33178, "epoch": 372, "lr": 9.645352571892238e-05}
{"train_loss": 0.42095378041267395, "global_step": 33179, "epoch": 372, "lr": 9.645331126968702e-05}
{"train_loss": 0.43894222378730774, "global_step": 33180, "epoch": 372, "lr": 9.645309681420657e-05}
{"train_loss": 0.2926504611968994, "global_step": 33181, "epoch": 372, "lr": 9.645288235248108e-05}
{"train_loss": 0.3815997242927551, "global_step": 33182, "epoch": 372, "lr": 9.645266788451057e-05}
{"train_loss": 0.2567157447338104, "global_step": 33183, "epoch": 372, "lr": 9.645245341029506e-05}
{"train_loss": 0.4408813416957855, "global_step": 33184, "epoch": 372, "lr": 9.645223892983458e-05}
{"train_loss": 0.3474687337875366, "global_step": 33185, "epoch": 372, "lr": 9.645202444312918e-05}
{"train_loss": 0.3660896122455597, "global_step": 33186, "epoch": 372, "lr": 9.645180995017884e-05}
{"train_loss": 0.3765130341053009, "global_step": 33187, "epoch": 372, "lr": 9.645159545098365e-05}
{"train_loss": 0.4694921672344208, "global_step": 33188, "epoch": 372, "lr": 9.645138094554359e-05}
{"train_loss": 0.47056493163108826, "global_step": 33189, "epoch": 372, "lr": 9.645116643385872e-05}
{"train_loss": 0.3412996232509613, "global_step": 33190, "epoch": 372, "lr": 9.645095191592904e-05}
{"train_loss": 0.34419000148773193, "global_step": 33191, "epoch": 372, "lr": 9.64507373917546e-05}
{"train_loss": 0.3852083384990692, "global_step": 33192, "epoch": 372, "lr": 9.645052286133545e-05}
{"train_loss": 0.4026235342025757, "global_step": 33193, "epoch": 372, "lr": 9.645030832467157e-05}
{"train_loss": 0.32846733927726746, "global_step": 33194, "epoch": 372, "lr": 9.645009378176302e-05}
{"train_loss": 0.4674657881259918, "global_step": 33195, "epoch": 372, "lr": 9.644987923260983e-05}
{"train_loss": 0.3562571048066857, "global_step": 33196, "epoch": 372, "lr": 9.644966467721199e-05, "val_loss": 2.3167338371276855}
{"train_loss": 0.3089132010936737, "global_step": 33197, "epoch": 373, "lr": 9.644945011556958e-05}
{"train_loss": 0.33994483947753906, "global_step": 33198, "epoch": 373, "lr": 9.644923554768261e-05}
{"train_loss": 0.3100261688232422, "global_step": 33199, "epoch": 373, "lr": 9.64490209735511e-05}
{"train_loss": 0.3166421055793762, "global_step": 33200, "epoch": 373, "lr": 9.644880639317509e-05}
{"train_loss": 0.3331592381000519, "global_step": 33201, "epoch": 373, "lr": 9.644859180655459e-05}
{"train_loss": 0.49605897068977356, "global_step": 33202, "epoch": 373, "lr": 9.644837721368966e-05}
{"train_loss": 0.3977419137954712, "global_step": 33203, "epoch": 373, "lr": 9.64481626145803e-05}
{"train_loss": 0.3434251844882965, "global_step": 33204, "epoch": 373, "lr": 9.644794800922655e-05}
{"train_loss": 0.30575889348983765, "global_step": 33205, "epoch": 373, "lr": 9.644773339762844e-05}
{"train_loss": 0.33288854360580444, "global_step": 33206, "epoch": 373, "lr": 9.644751877978602e-05}
{"train_loss": 0.3184798061847687, "global_step": 33207, "epoch": 373, "lr": 9.644730415569926e-05}
{"train_loss": 0.2500850558280945, "global_step": 33208, "epoch": 373, "lr": 9.644708952536826e-05}
{"train_loss": 0.31671562790870667, "global_step": 33209, "epoch": 373, "lr": 9.644687488879301e-05}
{"train_loss": 0.3114750385284424, "global_step": 33210, "epoch": 373, "lr": 9.644666024597353e-05}
{"train_loss": 0.4061741232872009, "global_step": 33211, "epoch": 373, "lr": 9.644644559690987e-05}
{"train_loss": 0.2939220666885376, "global_step": 33212, "epoch": 373, "lr": 9.644623094160204e-05}
{"train_loss": 0.3703307807445526, "global_step": 33213, "epoch": 373, "lr": 9.64460162800501e-05}
{"train_loss": 0.4042479395866394, "global_step": 33214, "epoch": 373, "lr": 9.644580161225405e-05}
{"train_loss": 0.33357104659080505, "global_step": 33215, "epoch": 373, "lr": 9.644558693821392e-05}
{"train_loss": 0.3881462514400482, "global_step": 33216, "epoch": 373, "lr": 9.644537225792975e-05}
{"train_loss": 0.2685478627681732, "global_step": 33217, "epoch": 373, "lr": 9.644515757140158e-05}
{"train_loss": 0.2836620807647705, "global_step": 33218, "epoch": 373, "lr": 9.644494287862941e-05}
{"train_loss": 0.3674311339855194, "global_step": 33219, "epoch": 373, "lr": 9.644472817961329e-05}
{"train_loss": 0.3526192605495453, "global_step": 33220, "epoch": 373, "lr": 9.644451347435323e-05}
{"train_loss": 0.37524425983428955, "global_step": 33221, "epoch": 373, "lr": 9.644429876284929e-05}
{"train_loss": 0.4123949706554413, "global_step": 33222, "epoch": 373, "lr": 9.644408404510149e-05}
{"train_loss": 0.3495710492134094, "global_step": 33223, "epoch": 373, "lr": 9.64438693211098e-05}
{"train_loss": 0.37575793266296387, "global_step": 33224, "epoch": 373, "lr": 9.644365459087434e-05}
{"train_loss": 0.326282799243927, "global_step": 33225, "epoch": 373, "lr": 9.644343985439508e-05}
{"train_loss": 0.4557191729545593, "global_step": 33226, "epoch": 373, "lr": 9.644322511167207e-05}
{"train_loss": 0.28245899081230164, "global_step": 33227, "epoch": 373, "lr": 9.644301036270533e-05}
{"train_loss": 0.2933754622936249, "global_step": 33228, "epoch": 373, "lr": 9.644279560749491e-05}
{"train_loss": 0.3614504337310791, "global_step": 33229, "epoch": 373, "lr": 9.64425808460408e-05}
{"train_loss": 0.3419666290283203, "global_step": 33230, "epoch": 373, "lr": 9.644236607834308e-05}
{"train_loss": 0.41876551508903503, "global_step": 33231, "epoch": 373, "lr": 9.644215130440173e-05}
{"train_loss": 0.3421756625175476, "global_step": 33232, "epoch": 373, "lr": 9.64419365242168e-05}
{"train_loss": 0.5719601511955261, "global_step": 33233, "epoch": 373, "lr": 9.644172173778832e-05}
{"train_loss": 0.38393762707710266, "global_step": 33234, "epoch": 373, "lr": 9.644150694511632e-05}
{"train_loss": 0.4052673876285553, "global_step": 33235, "epoch": 373, "lr": 9.64412921462008e-05}
{"train_loss": 0.27584606409072876, "global_step": 33236, "epoch": 373, "lr": 9.644107734104184e-05}
{"train_loss": 0.3700994849205017, "global_step": 33237, "epoch": 373, "lr": 9.644086252963945e-05}
{"train_loss": 0.3933851718902588, "global_step": 33238, "epoch": 373, "lr": 9.644064771199364e-05}
{"train_loss": 0.3430536985397339, "global_step": 33239, "epoch": 373, "lr": 9.644043288810445e-05}
{"train_loss": 0.3718101680278778, "global_step": 33240, "epoch": 373, "lr": 9.64402180579719e-05}
{"train_loss": 0.3972833752632141, "global_step": 33241, "epoch": 373, "lr": 9.644000322159604e-05}
{"train_loss": 0.4145018458366394, "global_step": 33242, "epoch": 373, "lr": 9.643978837897689e-05}
{"train_loss": 0.29313987493515015, "global_step": 33243, "epoch": 373, "lr": 9.643957353011448e-05}
{"train_loss": 0.3908941447734833, "global_step": 33244, "epoch": 373, "lr": 9.643935867500882e-05}
{"train_loss": 0.34452131390571594, "global_step": 33245, "epoch": 373, "lr": 9.643914381365997e-05}
{"train_loss": 0.2869841754436493, "global_step": 33246, "epoch": 373, "lr": 9.643892894606794e-05}
{"train_loss": 0.4080427587032318, "global_step": 33247, "epoch": 373, "lr": 9.643871407223275e-05}
{"train_loss": 0.3332187235355377, "global_step": 33248, "epoch": 373, "lr": 9.643849919215444e-05}
{"train_loss": 0.32810869812965393, "global_step": 33249, "epoch": 373, "lr": 9.643828430583305e-05}
{"train_loss": 0.3762829005718231, "global_step": 33250, "epoch": 373, "lr": 9.64380694132686e-05}
{"train_loss": 0.2728201746940613, "global_step": 33251, "epoch": 373, "lr": 9.643785451446113e-05}
{"train_loss": 0.3554995059967041, "global_step": 33252, "epoch": 373, "lr": 9.643763960941064e-05}
{"train_loss": 0.3153982162475586, "global_step": 33253, "epoch": 373, "lr": 9.643742469811717e-05}
{"train_loss": 0.31338000297546387, "global_step": 33254, "epoch": 373, "lr": 9.643720978058075e-05}
{"train_loss": 0.2675243020057678, "global_step": 33255, "epoch": 373, "lr": 9.643699485680143e-05}
{"train_loss": 0.3711875379085541, "global_step": 33256, "epoch": 373, "lr": 9.643677992677922e-05}
{"train_loss": 0.3001693785190582, "global_step": 33257, "epoch": 373, "lr": 9.643656499051414e-05}
{"train_loss": 0.34661561250686646, "global_step": 33258, "epoch": 373, "lr": 9.643635004800624e-05}
{"train_loss": 0.2987126111984253, "global_step": 33259, "epoch": 373, "lr": 9.643613509925554e-05}
{"train_loss": 0.4176028370857239, "global_step": 33260, "epoch": 373, "lr": 9.643592014426206e-05}
{"train_loss": 0.31839725375175476, "global_step": 33261, "epoch": 373, "lr": 9.643570518302585e-05}
{"train_loss": 0.38562434911727905, "global_step": 33262, "epoch": 373, "lr": 9.643549021554691e-05}
{"train_loss": 0.44745802879333496, "global_step": 33263, "epoch": 373, "lr": 9.643527524182528e-05}
{"train_loss": 0.39127230644226074, "global_step": 33264, "epoch": 373, "lr": 9.6435060261861e-05}
{"train_loss": 0.33983826637268066, "global_step": 33265, "epoch": 373, "lr": 9.643484527565411e-05}
{"train_loss": 0.3258916735649109, "global_step": 33266, "epoch": 373, "lr": 9.643463028320461e-05}
{"train_loss": 0.3335374891757965, "global_step": 33267, "epoch": 373, "lr": 9.643441528451253e-05}
{"train_loss": 0.4079345464706421, "global_step": 33268, "epoch": 373, "lr": 9.643420027957791e-05}
{"train_loss": 0.36204999685287476, "global_step": 33269, "epoch": 373, "lr": 9.643398526840078e-05}
{"train_loss": 0.2686510384082794, "global_step": 33270, "epoch": 373, "lr": 9.643377025098116e-05}
{"train_loss": 0.38025298714637756, "global_step": 33271, "epoch": 373, "lr": 9.643355522731911e-05}
{"train_loss": 0.338721364736557, "global_step": 33272, "epoch": 373, "lr": 9.643334019741461e-05}
{"train_loss": 0.34905117750167847, "global_step": 33273, "epoch": 373, "lr": 9.643312516126772e-05}
{"train_loss": 0.2844122350215912, "global_step": 33274, "epoch": 373, "lr": 9.643291011887847e-05}
{"train_loss": 0.2702884376049042, "global_step": 33275, "epoch": 373, "lr": 9.643269507024688e-05}
{"train_loss": 0.3096514940261841, "global_step": 33276, "epoch": 373, "lr": 9.643248001537298e-05}
{"train_loss": 0.3782167136669159, "global_step": 33277, "epoch": 373, "lr": 9.643226495425679e-05}
{"train_loss": 0.32396742701530457, "global_step": 33278, "epoch": 373, "lr": 9.643204988689835e-05}
{"train_loss": 0.2891913056373596, "global_step": 33279, "epoch": 373, "lr": 9.64318348132977e-05}
{"train_loss": 0.32796940207481384, "global_step": 33280, "epoch": 373, "lr": 9.643161973345483e-05}
{"train_loss": 0.32155996561050415, "global_step": 33281, "epoch": 373, "lr": 9.643140464736981e-05}
{"train_loss": 0.3573872447013855, "global_step": 33282, "epoch": 373, "lr": 9.643118955504267e-05}
{"train_loss": 0.37326106429100037, "global_step": 33283, "epoch": 373, "lr": 9.64309744564734e-05}
{"train_loss": 0.35918018221855164, "global_step": 33284, "epoch": 373, "lr": 9.643075935166206e-05}
{"train_loss": 0.34896075892984196, "global_step": 33285, "epoch": 373, "lr": 9.643054424060867e-05, "val_loss": 2.3654274940490723}
{"train_loss": 0.27510136365890503, "global_step": 33286, "epoch": 374, "lr": 9.643032912331325e-05}
{"train_loss": 0.30470556020736694, "global_step": 33287, "epoch": 374, "lr": 9.643011399977584e-05}
{"train_loss": 0.34013423323631287, "global_step": 33288, "epoch": 374, "lr": 9.642989886999647e-05}
{"train_loss": 0.2991625666618347, "global_step": 33289, "epoch": 374, "lr": 9.642968373397518e-05}
{"train_loss": 0.3188832700252533, "global_step": 33290, "epoch": 374, "lr": 9.642946859171196e-05}
{"train_loss": 0.38186725974082947, "global_step": 33291, "epoch": 374, "lr": 9.642925344320689e-05}
{"train_loss": 0.3933303654193878, "global_step": 33292, "epoch": 374, "lr": 9.642903828845995e-05}
{"train_loss": 0.4904468059539795, "global_step": 33293, "epoch": 374, "lr": 9.642882312747122e-05}
{"train_loss": 0.4169539511203766, "global_step": 33294, "epoch": 374, "lr": 9.642860796024067e-05}
{"train_loss": 0.3512071371078491, "global_step": 33295, "epoch": 374, "lr": 9.642839278676837e-05}
{"train_loss": 0.3717598617076874, "global_step": 33296, "epoch": 374, "lr": 9.642817760705434e-05}
{"train_loss": 0.32744231820106506, "global_step": 33297, "epoch": 374, "lr": 9.64279624210986e-05}
{"train_loss": 0.3403284251689911, "global_step": 33298, "epoch": 374, "lr": 9.64277472289012e-05}
{"train_loss": 0.38760098814964294, "global_step": 33299, "epoch": 374, "lr": 9.642753203046216e-05}
{"train_loss": 0.33517149090766907, "global_step": 33300, "epoch": 374, "lr": 9.642731682578147e-05}
{"train_loss": 0.3795396685600281, "global_step": 33301, "epoch": 374, "lr": 9.642710161485923e-05}
{"train_loss": 0.41714656352996826, "global_step": 33302, "epoch": 374, "lr": 9.642688639769541e-05}
{"train_loss": 0.3940548896789551, "global_step": 33303, "epoch": 374, "lr": 9.642667117429007e-05}
{"train_loss": 0.3534766137599945, "global_step": 33304, "epoch": 374, "lr": 9.642645594464323e-05}
{"train_loss": 0.32592830061912537, "global_step": 33305, "epoch": 374, "lr": 9.642624070875492e-05}
{"train_loss": 0.41842034459114075, "global_step": 33306, "epoch": 374, "lr": 9.642602546662518e-05}
{"train_loss": 0.29661262035369873, "global_step": 33307, "epoch": 374, "lr": 9.6425810218254e-05}
{"train_loss": 0.32346010208129883, "global_step": 33308, "epoch": 374, "lr": 9.642559496364145e-05}
{"train_loss": 0.43872958421707153, "global_step": 33309, "epoch": 374, "lr": 9.642537970278755e-05}
{"train_loss": 0.39591771364212036, "global_step": 33310, "epoch": 374, "lr": 9.642516443569232e-05}
{"train_loss": 0.36177918314933777, "global_step": 33311, "epoch": 374, "lr": 9.642494916235578e-05}
{"train_loss": 0.3680754005908966, "global_step": 33312, "epoch": 374, "lr": 9.642473388277799e-05}
{"train_loss": 0.42654043436050415, "global_step": 33313, "epoch": 374, "lr": 9.642451859695895e-05}
{"train_loss": 0.3898833096027374, "global_step": 33314, "epoch": 374, "lr": 9.642430330489869e-05}
{"train_loss": 0.4005843997001648, "global_step": 33315, "epoch": 374, "lr": 9.642408800659726e-05}
{"train_loss": 0.4374784827232361, "global_step": 33316, "epoch": 374, "lr": 9.642387270205469e-05}
{"train_loss": 0.2652685046195984, "global_step": 33317, "epoch": 374, "lr": 9.642365739127097e-05}
{"train_loss": 0.32083457708358765, "global_step": 33318, "epoch": 374, "lr": 9.642344207424616e-05}
{"train_loss": 0.33810073137283325, "global_step": 33319, "epoch": 374, "lr": 9.64232267509803e-05}
{"train_loss": 0.3620333671569824, "global_step": 33320, "epoch": 374, "lr": 9.642301142147338e-05}
{"train_loss": 0.2688331604003906, "global_step": 33321, "epoch": 374, "lr": 9.642279608572548e-05}
{"train_loss": 0.37451013922691345, "global_step": 33322, "epoch": 374, "lr": 9.642258074373658e-05}
{"train_loss": 0.3690083622932434, "global_step": 33323, "epoch": 374, "lr": 9.642236539550675e-05}
{"train_loss": 0.3925451338291168, "global_step": 33324, "epoch": 374, "lr": 9.642215004103597e-05}
{"train_loss": 0.3099733889102936, "global_step": 33325, "epoch": 374, "lr": 9.642193468032433e-05}
{"train_loss": 0.3682172894477844, "global_step": 33326, "epoch": 374, "lr": 9.642171931337181e-05}
{"train_loss": 0.3935128152370453, "global_step": 33327, "epoch": 374, "lr": 9.642150394017844e-05}
{"train_loss": 0.5139480233192444, "global_step": 33328, "epoch": 374, "lr": 9.642128856074429e-05}
{"train_loss": 0.3820131719112396, "global_step": 33329, "epoch": 374, "lr": 9.642107317506935e-05}
{"train_loss": 0.32856249809265137, "global_step": 33330, "epoch": 374, "lr": 9.642085778315366e-05}
{"train_loss": 0.3120373785495758, "global_step": 33331, "epoch": 374, "lr": 9.642064238499726e-05}
{"train_loss": 0.319709837436676, "global_step": 33332, "epoch": 374, "lr": 9.642042698060017e-05}
{"train_loss": 0.31172308325767517, "global_step": 33333, "epoch": 374, "lr": 9.642021156996241e-05}
{"train_loss": 0.3868998885154724, "global_step": 33334, "epoch": 374, "lr": 9.641999615308404e-05}
{"train_loss": 0.4392564594745636, "global_step": 33335, "epoch": 374, "lr": 9.641978072996505e-05}
{"train_loss": 0.4074954092502594, "global_step": 33336, "epoch": 374, "lr": 9.641956530060548e-05}
{"train_loss": 0.3892478942871094, "global_step": 33337, "epoch": 374, "lr": 9.641934986500539e-05}
{"train_loss": 0.40313875675201416, "global_step": 33338, "epoch": 374, "lr": 9.641913442316476e-05}
{"train_loss": 0.2577548325061798, "global_step": 33339, "epoch": 374, "lr": 9.641891897508365e-05}
{"train_loss": 0.3542400300502777, "global_step": 33340, "epoch": 374, "lr": 9.641870352076207e-05}
{"train_loss": 0.3630220293998718, "global_step": 33341, "epoch": 374, "lr": 9.641848806020008e-05}
{"train_loss": 0.3615840971469879, "global_step": 33342, "epoch": 374, "lr": 9.641827259339768e-05}
{"train_loss": 0.3441646695137024, "global_step": 33343, "epoch": 374, "lr": 9.641805712035493e-05}
{"train_loss": 0.4158134460449219, "global_step": 33344, "epoch": 374, "lr": 9.641784164107181e-05}
{"train_loss": 0.3662359416484833, "global_step": 33345, "epoch": 374, "lr": 9.641762615554839e-05}
{"train_loss": 0.39620649814605713, "global_step": 33346, "epoch": 374, "lr": 9.641741066378468e-05}
{"train_loss": 0.31023484468460083, "global_step": 33347, "epoch": 374, "lr": 9.641719516578073e-05}
{"train_loss": 0.3804887533187866, "global_step": 33348, "epoch": 374, "lr": 9.641697966153654e-05}
{"train_loss": 0.32879889011383057, "global_step": 33349, "epoch": 374, "lr": 9.641676415105216e-05}
{"train_loss": 0.318745881319046, "global_step": 33350, "epoch": 374, "lr": 9.64165486343276e-05}
{"train_loss": 0.32174986600875854, "global_step": 33351, "epoch": 374, "lr": 9.641633311136291e-05}
{"train_loss": 0.36265259981155396, "global_step": 33352, "epoch": 374, "lr": 9.641611758215811e-05}
{"train_loss": 0.3747555613517761, "global_step": 33353, "epoch": 374, "lr": 9.641590204671324e-05}
{"train_loss": 0.3912475109100342, "global_step": 33354, "epoch": 374, "lr": 9.64156865050283e-05}
{"train_loss": 0.3217981159687042, "global_step": 33355, "epoch": 374, "lr": 9.641547095710334e-05}
{"train_loss": 0.33908766508102417, "global_step": 33356, "epoch": 374, "lr": 9.641525540293839e-05}
{"train_loss": 0.3268715441226959, "global_step": 33357, "epoch": 374, "lr": 9.641503984253348e-05}
{"train_loss": 0.30434301495552063, "global_step": 33358, "epoch": 374, "lr": 9.641482427588862e-05}
{"train_loss": 0.2782728970050812, "global_step": 33359, "epoch": 374, "lr": 9.641460870300386e-05}
{"train_loss": 0.3156088590621948, "global_step": 33360, "epoch": 374, "lr": 9.641439312387921e-05}
{"train_loss": 0.3643174171447754, "global_step": 33361, "epoch": 374, "lr": 9.641417753851472e-05}
{"train_loss": 0.33911895751953125, "global_step": 33362, "epoch": 374, "lr": 9.641396194691042e-05}
{"train_loss": 0.366864413022995, "global_step": 33363, "epoch": 374, "lr": 9.641374634906632e-05}
{"train_loss": 0.44509759545326233, "global_step": 33364, "epoch": 374, "lr": 9.641353074498245e-05}
{"train_loss": 0.2751924693584442, "global_step": 33365, "epoch": 374, "lr": 9.641331513465886e-05}
{"train_loss": 0.31535840034484863, "global_step": 33366, "epoch": 374, "lr": 9.641309951809555e-05}
{"train_loss": 0.30082905292510986, "global_step": 33367, "epoch": 374, "lr": 9.641288389529257e-05}
{"train_loss": 0.35484254360198975, "global_step": 33368, "epoch": 374, "lr": 9.641266826624995e-05}
{"train_loss": 0.2948923707008362, "global_step": 33369, "epoch": 374, "lr": 9.64124526309677e-05}
{"train_loss": 0.3859516382217407, "global_step": 33370, "epoch": 374, "lr": 9.641223698944589e-05}
{"train_loss": 0.30071038007736206, "global_step": 33371, "epoch": 374, "lr": 9.641202134168448e-05}
{"train_loss": 0.33228015899658203, "global_step": 33372, "epoch": 374, "lr": 9.641180568768357e-05}
{"train_loss": 0.30261191725730896, "global_step": 33373, "epoch": 374, "lr": 9.641159002744315e-05}
{"train_loss": 0.35564900516124254, "global_step": 33374, "epoch": 374, "lr": 9.641137436096325e-05, "val_loss": 2.4766063690185547}
{"train_loss": 0.36027589440345764, "global_step": 33375, "epoch": 375, "lr": 9.641115868824391e-05}
{"train_loss": 0.2693859040737152, "global_step": 33376, "epoch": 375, "lr": 9.641094300928515e-05}
{"train_loss": 0.3848375976085663, "global_step": 33377, "epoch": 375, "lr": 9.641072732408702e-05}
{"train_loss": 0.33000436425209045, "global_step": 33378, "epoch": 375, "lr": 9.641051163264953e-05}
{"train_loss": 0.34313541650772095, "global_step": 33379, "epoch": 375, "lr": 9.64102959349727e-05}
{"train_loss": 0.30137887597084045, "global_step": 33380, "epoch": 375, "lr": 9.641008023105657e-05}
{"train_loss": 0.3057301640510559, "global_step": 33381, "epoch": 375, "lr": 9.640986452090119e-05}
{"train_loss": 0.4504246711730957, "global_step": 33382, "epoch": 375, "lr": 9.640964880450656e-05}
{"train_loss": 0.2021171748638153, "global_step": 33383, "epoch": 375, "lr": 9.640943308187271e-05}
{"train_loss": 0.22532163560390472, "global_step": 33384, "epoch": 375, "lr": 9.640921735299968e-05}
{"train_loss": 0.31371673941612244, "global_step": 33385, "epoch": 375, "lr": 9.640900161788751e-05}
{"train_loss": 0.4465372562408447, "global_step": 33386, "epoch": 375, "lr": 9.640878587653621e-05}
{"train_loss": 0.39155739545822144, "global_step": 33387, "epoch": 375, "lr": 9.640857012894579e-05}
{"train_loss": 0.4802541732788086, "global_step": 33388, "epoch": 375, "lr": 9.640835437511634e-05}
{"train_loss": 0.30890750885009766, "global_step": 33389, "epoch": 375, "lr": 9.640813861504783e-05}
{"train_loss": 0.3933030366897583, "global_step": 33390, "epoch": 375, "lr": 9.640792284874031e-05}
{"train_loss": 0.3186974823474884, "global_step": 33391, "epoch": 375, "lr": 9.640770707619383e-05}
{"train_loss": 0.2669907212257385, "global_step": 33392, "epoch": 375, "lr": 9.640749129740838e-05}
{"train_loss": 0.36115923523902893, "global_step": 33393, "epoch": 375, "lr": 9.640727551238403e-05}
{"train_loss": 0.4130188822746277, "global_step": 33394, "epoch": 375, "lr": 9.640705972112076e-05}
{"train_loss": 0.3436804413795471, "global_step": 33395, "epoch": 375, "lr": 9.640684392361865e-05}
{"train_loss": 0.3580417037010193, "global_step": 33396, "epoch": 375, "lr": 9.640662811987768e-05}
{"train_loss": 0.38485416769981384, "global_step": 33397, "epoch": 375, "lr": 9.640641230989793e-05}
{"train_loss": 0.3571687340736389, "global_step": 33398, "epoch": 375, "lr": 9.640619649367939e-05}
{"train_loss": 0.44335049390792847, "global_step": 33399, "epoch": 375, "lr": 9.640598067122211e-05}
{"train_loss": 0.4043674170970917, "global_step": 33400, "epoch": 375, "lr": 9.64057648425261e-05}
{"train_loss": 0.2110637128353119, "global_step": 33401, "epoch": 375, "lr": 9.64055490075914e-05}
{"train_loss": 0.3878959119319916, "global_step": 33402, "epoch": 375, "lr": 9.640533316641806e-05}
{"train_loss": 0.3223547637462616, "global_step": 33403, "epoch": 375, "lr": 9.640511731900608e-05}
{"train_loss": 0.3771611452102661, "global_step": 33404, "epoch": 375, "lr": 9.64049014653555e-05}
{"train_loss": 0.40530791878700256, "global_step": 33405, "epoch": 375, "lr": 9.640468560546633e-05}
{"train_loss": 0.3451792895793915, "global_step": 33406, "epoch": 375, "lr": 9.640446973933863e-05}
{"train_loss": 0.33186113834381104, "global_step": 33407, "epoch": 375, "lr": 9.64042538669724e-05}
{"train_loss": 0.3912521004676819, "global_step": 33408, "epoch": 375, "lr": 9.640403798836771e-05}
{"train_loss": 0.3550287187099457, "global_step": 33409, "epoch": 375, "lr": 9.640382210352453e-05}
{"train_loss": 0.3050072193145752, "global_step": 33410, "epoch": 375, "lr": 9.640360621244295e-05}
{"train_loss": 0.382814884185791, "global_step": 33411, "epoch": 375, "lr": 9.640339031512296e-05}
{"train_loss": 0.1952187716960907, "global_step": 33412, "epoch": 375, "lr": 9.64031744115646e-05}
{"train_loss": 0.38934212923049927, "global_step": 33413, "epoch": 375, "lr": 9.640295850176788e-05}
{"train_loss": 0.3246292769908905, "global_step": 33414, "epoch": 375, "lr": 9.640274258573287e-05}
{"train_loss": 0.3389357328414917, "global_step": 33415, "epoch": 375, "lr": 9.640252666345956e-05}
{"train_loss": 0.4626747667789459, "global_step": 33416, "epoch": 375, "lr": 9.640231073494802e-05}
{"train_loss": 0.3464488685131073, "global_step": 33417, "epoch": 375, "lr": 9.640209480019824e-05}
{"train_loss": 0.4188297986984253, "global_step": 33418, "epoch": 375, "lr": 9.640187885921027e-05}
{"train_loss": 0.3001711368560791, "global_step": 33419, "epoch": 375, "lr": 9.640166291198414e-05}
{"train_loss": 0.41364461183547974, "global_step": 33420, "epoch": 375, "lr": 9.640144695851985e-05}
{"train_loss": 0.4010851979255676, "global_step": 33421, "epoch": 375, "lr": 9.640123099881746e-05}
{"train_loss": 0.3795599937438965, "global_step": 33422, "epoch": 375, "lr": 9.640101503287699e-05}
{"train_loss": 0.3699183166027069, "global_step": 33423, "epoch": 375, "lr": 9.640079906069848e-05}
{"train_loss": 0.34933534264564514, "global_step": 33424, "epoch": 375, "lr": 9.640058308228193e-05}
{"train_loss": 0.26544642448425293, "global_step": 33425, "epoch": 375, "lr": 9.640036709762741e-05}
{"train_loss": 0.32546019554138184, "global_step": 33426, "epoch": 375, "lr": 9.64001511067349e-05}
{"train_loss": 0.3914381265640259, "global_step": 33427, "epoch": 375, "lr": 9.639993510960448e-05}
{"train_loss": 0.3867558240890503, "global_step": 33428, "epoch": 375, "lr": 9.639971910623614e-05}
{"train_loss": 0.42436403036117554, "global_step": 33429, "epoch": 375, "lr": 9.639950309662993e-05}
{"train_loss": 0.3564230501651764, "global_step": 33430, "epoch": 375, "lr": 9.639928708078587e-05}
{"train_loss": 0.3373072147369385, "global_step": 33431, "epoch": 375, "lr": 9.639907105870399e-05}
{"train_loss": 0.3477958142757416, "global_step": 33432, "epoch": 375, "lr": 9.639885503038431e-05}
{"train_loss": 0.264169842004776, "global_step": 33433, "epoch": 375, "lr": 9.639863899582688e-05}
{"train_loss": 0.34507864713668823, "global_step": 33434, "epoch": 375, "lr": 9.639842295503172e-05}
{"train_loss": 0.3530360460281372, "global_step": 33435, "epoch": 375, "lr": 9.639820690799885e-05}
{"train_loss": 0.46573901176452637, "global_step": 33436, "epoch": 375, "lr": 9.639799085472831e-05}
{"train_loss": 0.39507409930229187, "global_step": 33437, "epoch": 375, "lr": 9.639777479522015e-05}
{"train_loss": 0.33983471989631653, "global_step": 33438, "epoch": 375, "lr": 9.639755872947434e-05}
{"train_loss": 0.31402358412742615, "global_step": 33439, "epoch": 375, "lr": 9.639734265749097e-05}
{"train_loss": 0.351223349571228, "global_step": 33440, "epoch": 375, "lr": 9.639712657927002e-05}
{"train_loss": 0.3815331757068634, "global_step": 33441, "epoch": 375, "lr": 9.639691049481155e-05}
{"train_loss": 0.44820576906204224, "global_step": 33442, "epoch": 375, "lr": 9.639669440411558e-05}
{"train_loss": 0.429821640253067, "global_step": 33443, "epoch": 375, "lr": 9.639647830718215e-05}
{"train_loss": 0.3506360352039337, "global_step": 33444, "epoch": 375, "lr": 9.639626220401126e-05}
{"train_loss": 0.36997365951538086, "global_step": 33445, "epoch": 375, "lr": 9.639604609460298e-05}
{"train_loss": 0.3306026756763458, "global_step": 33446, "epoch": 375, "lr": 9.63958299789573e-05}
{"train_loss": 0.4426996409893036, "global_step": 33447, "epoch": 375, "lr": 9.639561385707428e-05}
{"train_loss": 0.3947857618331909, "global_step": 33448, "epoch": 375, "lr": 9.639539772895391e-05}
{"train_loss": 0.3055874705314636, "global_step": 33449, "epoch": 375, "lr": 9.639518159459627e-05}
{"train_loss": 0.3368476629257202, "global_step": 33450, "epoch": 375, "lr": 9.639496545400136e-05}
{"train_loss": 0.3075077533721924, "global_step": 33451, "epoch": 375, "lr": 9.63947493071692e-05}
{"train_loss": 0.35805612802505493, "global_step": 33452, "epoch": 375, "lr": 9.639453315409983e-05}
{"train_loss": 0.3871065378189087, "global_step": 33453, "epoch": 375, "lr": 9.639431699479329e-05}
{"train_loss": 0.3250170648097992, "global_step": 33454, "epoch": 375, "lr": 9.63941008292496e-05}
{"train_loss": 0.33653607964515686, "global_step": 33455, "epoch": 375, "lr": 9.639388465746879e-05}
{"train_loss": 0.3880624771118164, "global_step": 33456, "epoch": 375, "lr": 9.639366847945088e-05}
{"train_loss": 0.38257795572280884, "global_step": 33457, "epoch": 375, "lr": 9.639345229519592e-05}
{"train_loss": 0.3635435700416565, "global_step": 33458, "epoch": 375, "lr": 9.639323610470393e-05}
{"train_loss": 0.42261219024658203, "global_step": 33459, "epoch": 375, "lr": 9.639301990797491e-05}
{"train_loss": 0.35031798481941223, "global_step": 33460, "epoch": 375, "lr": 9.639280370500891e-05}
{"train_loss": 0.3861333727836609, "global_step": 33461, "epoch": 375, "lr": 9.6392587495806e-05}
{"train_loss": 0.22489656507968903, "global_step": 33462, "epoch": 375, "lr": 9.639237128036615e-05}
{"train_loss": 0.3557766749617759, "global_step": 33463, "epoch": 375, "lr": 9.639215505868941e-05, "val_loss": 2.3654236793518066, "train_action_mse_error": 35.2945556640625}
{"train_loss": 0.4017559885978699, "global_step": 33464, "epoch": 376, "lr": 9.639193883077581e-05}
{"train_loss": 0.31680455803871155, "global_step": 33465, "epoch": 376, "lr": 9.639172259662538e-05}
{"train_loss": 0.3864981532096863, "global_step": 33466, "epoch": 376, "lr": 9.639150635623815e-05}
{"train_loss": 0.356086403131485, "global_step": 33467, "epoch": 376, "lr": 9.639129010961414e-05}
{"train_loss": 0.25921395421028137, "global_step": 33468, "epoch": 376, "lr": 9.63910738567534e-05}
{"train_loss": 0.3378611207008362, "global_step": 33469, "epoch": 376, "lr": 9.639085759765594e-05}
{"train_loss": 0.28134962916374207, "global_step": 33470, "epoch": 376, "lr": 9.63906413323218e-05}
{"train_loss": 0.31285959482192993, "global_step": 33471, "epoch": 376, "lr": 9.6390425060751e-05}
{"train_loss": 0.27376124262809753, "global_step": 33472, "epoch": 376, "lr": 9.639020878294356e-05}
{"train_loss": 0.4192720651626587, "global_step": 33473, "epoch": 376, "lr": 9.638999249889953e-05}
{"train_loss": 0.3162948489189148, "global_step": 33474, "epoch": 376, "lr": 9.638977620861893e-05}
{"train_loss": 0.2651984989643097, "global_step": 33475, "epoch": 376, "lr": 9.63895599121018e-05}
{"train_loss": 0.28050875663757324, "global_step": 33476, "epoch": 376, "lr": 9.638934360934814e-05}
{"train_loss": 0.3579860329627991, "global_step": 33477, "epoch": 376, "lr": 9.638912730035803e-05}
{"train_loss": 0.35981887578964233, "global_step": 33478, "epoch": 376, "lr": 9.638891098513145e-05}
{"train_loss": 0.49292129278182983, "global_step": 33479, "epoch": 376, "lr": 9.638869466366844e-05}
{"train_loss": 0.34964728355407715, "global_step": 33480, "epoch": 376, "lr": 9.638847833596903e-05}
{"train_loss": 0.32983264327049255, "global_step": 33481, "epoch": 376, "lr": 9.638826200203327e-05}
{"train_loss": 0.3254331648349762, "global_step": 33482, "epoch": 376, "lr": 9.638804566186117e-05}
{"train_loss": 0.34038060903549194, "global_step": 33483, "epoch": 376, "lr": 9.638782931545276e-05}
{"train_loss": 0.4004700779914856, "global_step": 33484, "epoch": 376, "lr": 9.638761296280807e-05}
{"train_loss": 0.32418036460876465, "global_step": 33485, "epoch": 376, "lr": 9.638739660392713e-05}
{"train_loss": 0.300323486328125, "global_step": 33486, "epoch": 376, "lr": 9.638718023880997e-05}
{"train_loss": 0.37443259358406067, "global_step": 33487, "epoch": 376, "lr": 9.638696386745663e-05}
{"train_loss": 0.3468126952648163, "global_step": 33488, "epoch": 376, "lr": 9.638674748986712e-05}
{"train_loss": 0.2669096887111664, "global_step": 33489, "epoch": 376, "lr": 9.638653110604148e-05}
{"train_loss": 0.3425949215888977, "global_step": 33490, "epoch": 376, "lr": 9.638631471597973e-05}
{"train_loss": 0.36847683787345886, "global_step": 33491, "epoch": 376, "lr": 9.63860983196819e-05}
{"train_loss": 0.36189666390419006, "global_step": 33492, "epoch": 376, "lr": 9.638588191714804e-05}
{"train_loss": 0.35455507040023804, "global_step": 33493, "epoch": 376, "lr": 9.638566550837815e-05}
{"train_loss": 0.4275626540184021, "global_step": 33494, "epoch": 376, "lr": 9.638544909337228e-05}
{"train_loss": 0.26065593957901, "global_step": 33495, "epoch": 376, "lr": 9.638523267213046e-05}
{"train_loss": 0.27744680643081665, "global_step": 33496, "epoch": 376, "lr": 9.638501624465271e-05}
{"train_loss": 0.38579022884368896, "global_step": 33497, "epoch": 376, "lr": 9.638479981093904e-05}
{"train_loss": 0.31516027450561523, "global_step": 33498, "epoch": 376, "lr": 9.638458337098951e-05}
{"train_loss": 0.3827376067638397, "global_step": 33499, "epoch": 376, "lr": 9.638436692480414e-05}
{"train_loss": 0.33307841420173645, "global_step": 33500, "epoch": 376, "lr": 9.638415047238296e-05}
{"train_loss": 0.35126447677612305, "global_step": 33501, "epoch": 376, "lr": 9.638393401372598e-05}
{"train_loss": 0.3870890736579895, "global_step": 33502, "epoch": 376, "lr": 9.638371754883326e-05}
{"train_loss": 0.29672807455062866, "global_step": 33503, "epoch": 376, "lr": 9.638350107770482e-05}
{"train_loss": 0.3259700834751129, "global_step": 33504, "epoch": 376, "lr": 9.638328460034068e-05}
{"train_loss": 0.33059582114219666, "global_step": 33505, "epoch": 376, "lr": 9.638306811674086e-05}
{"train_loss": 0.35707128047943115, "global_step": 33506, "epoch": 376, "lr": 9.638285162690542e-05}
{"train_loss": 0.35389599204063416, "global_step": 33507, "epoch": 376, "lr": 9.638263513083436e-05}
{"train_loss": 0.2722610831260681, "global_step": 33508, "epoch": 376, "lr": 9.638241862852772e-05}
{"train_loss": 0.3783058226108551, "global_step": 33509, "epoch": 376, "lr": 9.638220211998552e-05}
{"train_loss": 0.3819543123245239, "global_step": 33510, "epoch": 376, "lr": 9.638198560520781e-05}
{"train_loss": 0.32977527379989624, "global_step": 33511, "epoch": 376, "lr": 9.63817690841946e-05}
{"train_loss": 0.37295758724212646, "global_step": 33512, "epoch": 376, "lr": 9.638155255694594e-05}
{"train_loss": 0.20888398587703705, "global_step": 33513, "epoch": 376, "lr": 9.638133602346185e-05}
{"train_loss": 0.4041813313961029, "global_step": 33514, "epoch": 376, "lr": 9.638111948374233e-05}
{"train_loss": 0.32485708594322205, "global_step": 33515, "epoch": 376, "lr": 9.638090293778746e-05}
{"train_loss": 0.27500268816947937, "global_step": 33516, "epoch": 376, "lr": 9.638068638559721e-05}
{"train_loss": 0.4582797586917877, "global_step": 33517, "epoch": 376, "lr": 9.638046982717168e-05}
{"train_loss": 0.507260262966156, "global_step": 33518, "epoch": 376, "lr": 9.638025326251084e-05}
{"train_loss": 0.3016892969608307, "global_step": 33519, "epoch": 376, "lr": 9.638003669161473e-05}
{"train_loss": 0.4003702402114868, "global_step": 33520, "epoch": 376, "lr": 9.637982011448341e-05}
{"train_loss": 0.3496439456939697, "global_step": 33521, "epoch": 376, "lr": 9.637960353111688e-05}
{"train_loss": 0.3904639184474945, "global_step": 33522, "epoch": 376, "lr": 9.637938694151519e-05}
{"train_loss": 0.43432995676994324, "global_step": 33523, "epoch": 376, "lr": 9.637917034567834e-05}
{"train_loss": 0.43063056468963623, "global_step": 33524, "epoch": 376, "lr": 9.637895374360637e-05}
{"train_loss": 0.3931538164615631, "global_step": 33525, "epoch": 376, "lr": 9.637873713529932e-05}
{"train_loss": 0.5715548396110535, "global_step": 33526, "epoch": 376, "lr": 9.637852052075723e-05}
{"train_loss": 0.350704550743103, "global_step": 33527, "epoch": 376, "lr": 9.637830389998011e-05}
{"train_loss": 0.31837230920791626, "global_step": 33528, "epoch": 376, "lr": 9.637808727296798e-05}
{"train_loss": 0.2908037006855011, "global_step": 33529, "epoch": 376, "lr": 9.637787063972088e-05}
{"train_loss": 0.46626368165016174, "global_step": 33530, "epoch": 376, "lr": 9.637765400023885e-05}
{"train_loss": 0.3284080922603607, "global_step": 33531, "epoch": 376, "lr": 9.637743735452192e-05}
{"train_loss": 0.43502840399742126, "global_step": 33532, "epoch": 376, "lr": 9.637722070257009e-05}
{"train_loss": 0.27422139048576355, "global_step": 33533, "epoch": 376, "lr": 9.637700404438342e-05}
{"train_loss": 0.3419571816921234, "global_step": 33534, "epoch": 376, "lr": 9.637678737996192e-05}
{"train_loss": 0.3541266620159149, "global_step": 33535, "epoch": 376, "lr": 9.637657070930562e-05}
{"train_loss": 0.3248208165168762, "global_step": 33536, "epoch": 376, "lr": 9.637635403241456e-05}
{"train_loss": 0.3366473913192749, "global_step": 33537, "epoch": 376, "lr": 9.637613734928876e-05}
{"train_loss": 0.3135039508342743, "global_step": 33538, "epoch": 376, "lr": 9.637592065992827e-05}
{"train_loss": 0.44927817583084106, "global_step": 33539, "epoch": 376, "lr": 9.637570396433309e-05}
{"train_loss": 0.34222185611724854, "global_step": 33540, "epoch": 376, "lr": 9.637548726250327e-05}
{"train_loss": 0.4073459804058075, "global_step": 33541, "epoch": 376, "lr": 9.637527055443882e-05}
{"train_loss": 0.3514571785926819, "global_step": 33542, "epoch": 376, "lr": 9.637505384013978e-05}
{"train_loss": 0.3928520083427429, "global_step": 33543, "epoch": 376, "lr": 9.63748371196062e-05}
{"train_loss": 0.4223211109638214, "global_step": 33544, "epoch": 376, "lr": 9.637462039283806e-05}
{"train_loss": 0.2382683902978897, "global_step": 33545, "epoch": 376, "lr": 9.637440365983544e-05}
{"train_loss": 0.41383635997772217, "global_step": 33546, "epoch": 376, "lr": 9.637418692059834e-05}
{"train_loss": 0.36283576488494873, "global_step": 33547, "epoch": 376, "lr": 9.637397017512679e-05}
{"train_loss": 0.34150418639183044, "global_step": 33548, "epoch": 376, "lr": 9.637375342342083e-05}
{"train_loss": 0.46009716391563416, "global_step": 33549, "epoch": 376, "lr": 9.637353666548048e-05}
{"train_loss": 0.3965626358985901, "global_step": 33550, "epoch": 376, "lr": 9.637331990130578e-05}
{"train_loss": 0.5718724727630615, "global_step": 33551, "epoch": 376, "lr": 9.637310313089675e-05}
{"train_loss": 0.35806408223141445, "global_step": 33552, "epoch": 376, "lr": 9.637288635425341e-05, "val_loss": 2.3515055179595947}
{"train_loss": 0.352949321269989, "global_step": 33553, "epoch": 377, "lr": 9.637266957137582e-05}
{"train_loss": 0.3908294141292572, "global_step": 33554, "epoch": 377, "lr": 9.637245278226397e-05}
{"train_loss": 0.37953996658325195, "global_step": 33555, "epoch": 377, "lr": 9.637223598691792e-05}
{"train_loss": 0.4534209370613098, "global_step": 33556, "epoch": 377, "lr": 9.637201918533769e-05}
{"train_loss": 0.32838964462280273, "global_step": 33557, "epoch": 377, "lr": 9.63718023775233e-05}
{"train_loss": 0.32222890853881836, "global_step": 33558, "epoch": 377, "lr": 9.63715855634748e-05}
{"train_loss": 0.33558782935142517, "global_step": 33559, "epoch": 377, "lr": 9.637136874319219e-05}
{"train_loss": 0.37947356700897217, "global_step": 33560, "epoch": 377, "lr": 9.637115191667553e-05}
{"train_loss": 0.30451658368110657, "global_step": 33561, "epoch": 377, "lr": 9.637093508392483e-05}
{"train_loss": 0.3817387819290161, "global_step": 33562, "epoch": 377, "lr": 9.63707182449401e-05}
{"train_loss": 0.3741909861564636, "global_step": 33563, "epoch": 377, "lr": 9.637050139972141e-05}
{"train_loss": 0.2731381356716156, "global_step": 33564, "epoch": 377, "lr": 9.637028454826878e-05}
{"train_loss": 0.3684524893760681, "global_step": 33565, "epoch": 377, "lr": 9.637006769058223e-05}
{"train_loss": 0.352355033159256, "global_step": 33566, "epoch": 377, "lr": 9.636985082666177e-05}
{"train_loss": 0.32198894023895264, "global_step": 33567, "epoch": 377, "lr": 9.636963395650746e-05}
{"train_loss": 0.43414637446403503, "global_step": 33568, "epoch": 377, "lr": 9.636941708011931e-05}
{"train_loss": 0.3886207342147827, "global_step": 33569, "epoch": 377, "lr": 9.636920019749737e-05}
{"train_loss": 0.3972756862640381, "global_step": 33570, "epoch": 377, "lr": 9.636898330864166e-05}
{"train_loss": 0.38180050253868103, "global_step": 33571, "epoch": 377, "lr": 9.636876641355219e-05}
{"train_loss": 0.31208327412605286, "global_step": 33572, "epoch": 377, "lr": 9.636854951222901e-05}
{"train_loss": 0.4494246244430542, "global_step": 33573, "epoch": 377, "lr": 9.636833260467215e-05}
{"train_loss": 0.344793438911438, "global_step": 33574, "epoch": 377, "lr": 9.636811569088162e-05}
{"train_loss": 0.41593223810195923, "global_step": 33575, "epoch": 377, "lr": 9.636789877085748e-05}
{"train_loss": 0.3010326325893402, "global_step": 33576, "epoch": 377, "lr": 9.636768184459972e-05}
{"train_loss": 0.41745656728744507, "global_step": 33577, "epoch": 377, "lr": 9.63674649121084e-05}
{"train_loss": 0.3741036355495453, "global_step": 33578, "epoch": 377, "lr": 9.636724797338355e-05}
{"train_loss": 0.3569025695323944, "global_step": 33579, "epoch": 377, "lr": 9.636703102842517e-05}
{"train_loss": 0.31848061084747314, "global_step": 33580, "epoch": 377, "lr": 9.636681407723331e-05}
{"train_loss": 0.4025643467903137, "global_step": 33581, "epoch": 377, "lr": 9.6366597119808e-05}
{"train_loss": 0.36560580134391785, "global_step": 33582, "epoch": 377, "lr": 9.636638015614928e-05}
{"train_loss": 0.3855096399784088, "global_step": 33583, "epoch": 377, "lr": 9.636616318625716e-05}
{"train_loss": 0.34012162685394287, "global_step": 33584, "epoch": 377, "lr": 9.636594621013167e-05}
{"train_loss": 0.3131019175052643, "global_step": 33585, "epoch": 377, "lr": 9.636572922777286e-05}
{"train_loss": 0.33088889718055725, "global_step": 33586, "epoch": 377, "lr": 9.636551223918071e-05}
{"train_loss": 0.3684885501861572, "global_step": 33587, "epoch": 377, "lr": 9.63652952443553e-05}
{"train_loss": 0.4113827049732208, "global_step": 33588, "epoch": 377, "lr": 9.636507824329664e-05}
{"train_loss": 0.3744439482688904, "global_step": 33589, "epoch": 377, "lr": 9.636486123600475e-05}
{"train_loss": 0.3254745304584503, "global_step": 33590, "epoch": 377, "lr": 9.636464422247968e-05}
{"train_loss": 0.38592010736465454, "global_step": 33591, "epoch": 377, "lr": 9.636442720272145e-05}
{"train_loss": 0.3750133812427521, "global_step": 33592, "epoch": 377, "lr": 9.636421017673008e-05}
{"train_loss": 0.3470495939254761, "global_step": 33593, "epoch": 377, "lr": 9.636399314450561e-05}
{"train_loss": 0.4856148660182953, "global_step": 33594, "epoch": 377, "lr": 9.636377610604808e-05}
{"train_loss": 0.4123659133911133, "global_step": 33595, "epoch": 377, "lr": 9.63635590613575e-05}
{"train_loss": 0.434184730052948, "global_step": 33596, "epoch": 377, "lr": 9.636334201043389e-05}
{"train_loss": 0.4571109116077423, "global_step": 33597, "epoch": 377, "lr": 9.636312495327731e-05}
{"train_loss": 0.35090702772140503, "global_step": 33598, "epoch": 377, "lr": 9.636290788988775e-05}
{"train_loss": 0.36497363448143005, "global_step": 33599, "epoch": 377, "lr": 9.636269082026529e-05}
{"train_loss": 0.26480910181999207, "global_step": 33600, "epoch": 377, "lr": 9.636247374440992e-05}
{"train_loss": 0.38804706931114197, "global_step": 33601, "epoch": 377, "lr": 9.636225666232168e-05}
{"train_loss": 0.3980560302734375, "global_step": 33602, "epoch": 377, "lr": 9.63620395740006e-05}
{"train_loss": 0.27550703287124634, "global_step": 33603, "epoch": 377, "lr": 9.636182247944671e-05}
{"train_loss": 0.2933257818222046, "global_step": 33604, "epoch": 377, "lr": 9.636160537866004e-05}
{"train_loss": 0.4093097150325775, "global_step": 33605, "epoch": 377, "lr": 9.636138827164061e-05}
{"train_loss": 0.31020650267601013, "global_step": 33606, "epoch": 377, "lr": 9.636117115838845e-05}
{"train_loss": 0.40919795632362366, "global_step": 33607, "epoch": 377, "lr": 9.63609540389036e-05}
{"train_loss": 0.34394949674606323, "global_step": 33608, "epoch": 377, "lr": 9.63607369131861e-05}
{"train_loss": 0.3809412121772766, "global_step": 33609, "epoch": 377, "lr": 9.636051978123595e-05}
{"train_loss": 0.3003159761428833, "global_step": 33610, "epoch": 377, "lr": 9.63603026430532e-05}
{"train_loss": 0.3696434497833252, "global_step": 33611, "epoch": 377, "lr": 9.636008549863787e-05}
{"train_loss": 0.41762906312942505, "global_step": 33612, "epoch": 377, "lr": 9.635986834798999e-05}
{"train_loss": 0.28898149728775024, "global_step": 33613, "epoch": 377, "lr": 9.635965119110958e-05}
{"train_loss": 0.2771826982498169, "global_step": 33614, "epoch": 377, "lr": 9.635943402799669e-05}
{"train_loss": 0.2845425307750702, "global_step": 33615, "epoch": 377, "lr": 9.635921685865135e-05}
{"train_loss": 0.2945428192615509, "global_step": 33616, "epoch": 377, "lr": 9.635899968307359e-05}
{"train_loss": 0.4018396735191345, "global_step": 33617, "epoch": 377, "lr": 9.635878250126339e-05}
{"train_loss": 0.31188517808914185, "global_step": 33618, "epoch": 377, "lr": 9.635856531322082e-05}
{"train_loss": 0.3356257975101471, "global_step": 33619, "epoch": 377, "lr": 9.635834811894593e-05}
{"train_loss": 0.38472655415534973, "global_step": 33620, "epoch": 377, "lr": 9.635813091843872e-05}
{"train_loss": 0.484387069940567, "global_step": 33621, "epoch": 377, "lr": 9.635791371169922e-05}
{"train_loss": 0.25844916701316833, "global_step": 33622, "epoch": 377, "lr": 9.635769649872745e-05}
{"train_loss": 0.40274813771247864, "global_step": 33623, "epoch": 377, "lr": 9.635747927952347e-05}
{"train_loss": 0.35889995098114014, "global_step": 33624, "epoch": 377, "lr": 9.635726205408729e-05}
{"train_loss": 0.2952904999256134, "global_step": 33625, "epoch": 377, "lr": 9.635704482241894e-05}
{"train_loss": 0.421886682510376, "global_step": 33626, "epoch": 377, "lr": 9.635682758451845e-05}
{"train_loss": 0.38193652033805847, "global_step": 33627, "epoch": 377, "lr": 9.635661034038584e-05}
{"train_loss": 0.33060285449028015, "global_step": 33628, "epoch": 377, "lr": 9.635639309002116e-05}
{"train_loss": 0.4133206605911255, "global_step": 33629, "epoch": 377, "lr": 9.635617583342442e-05}
{"train_loss": 0.34578847885131836, "global_step": 33630, "epoch": 377, "lr": 9.635595857059567e-05}
{"train_loss": 0.3541732430458069, "global_step": 33631, "epoch": 377, "lr": 9.635574130153492e-05}
{"train_loss": 0.3295828104019165, "global_step": 33632, "epoch": 377, "lr": 9.635552402624221e-05}
{"train_loss": 0.4800608456134796, "global_step": 33633, "epoch": 377, "lr": 9.635530674471755e-05}
{"train_loss": 0.3575846552848816, "global_step": 33634, "epoch": 377, "lr": 9.635508945696099e-05}
{"train_loss": 0.27949658036231995, "global_step": 33635, "epoch": 377, "lr": 9.635487216297256e-05}
{"train_loss": 0.4413212239742279, "global_step": 33636, "epoch": 377, "lr": 9.635465486275227e-05}
{"train_loss": 0.38314563035964966, "global_step": 33637, "epoch": 377, "lr": 9.635443755630019e-05}
{"train_loss": 0.4508020877838135, "global_step": 33638, "epoch": 377, "lr": 9.635422024361629e-05}
{"train_loss": 0.42737510800361633, "global_step": 33639, "epoch": 377, "lr": 9.635400292470065e-05}
{"train_loss": 0.36749738454818726, "global_step": 33640, "epoch": 377, "lr": 9.635378559955326e-05}
{"train_loss": 0.3642278126116549, "global_step": 33641, "epoch": 377, "lr": 9.635356826817418e-05, "val_loss": 2.3562161922454834}
{"train_loss": 0.34030941128730774, "global_step": 33642, "epoch": 378, "lr": 9.635335093056342e-05}
{"train_loss": 0.28378063440322876, "global_step": 33643, "epoch": 378, "lr": 9.635313358672102e-05}
{"train_loss": 0.35233983397483826, "global_step": 33644, "epoch": 378, "lr": 9.635291623664702e-05}
{"train_loss": 0.46261706948280334, "global_step": 33645, "epoch": 378, "lr": 9.635269888034142e-05}
{"train_loss": 0.33890271186828613, "global_step": 33646, "epoch": 378, "lr": 9.635248151780427e-05}
{"train_loss": 0.32660743594169617, "global_step": 33647, "epoch": 378, "lr": 9.63522641490356e-05}
{"train_loss": 0.35411974787712097, "global_step": 33648, "epoch": 378, "lr": 9.635204677403543e-05}
{"train_loss": 0.35270875692367554, "global_step": 33649, "epoch": 378, "lr": 9.635182939280378e-05}
{"train_loss": 0.38272273540496826, "global_step": 33650, "epoch": 378, "lr": 9.63516120053407e-05}
{"train_loss": 0.4422321319580078, "global_step": 33651, "epoch": 378, "lr": 9.635139461164622e-05}
{"train_loss": 0.4146836996078491, "global_step": 33652, "epoch": 378, "lr": 9.635117721172034e-05}
{"train_loss": 0.29665350914001465, "global_step": 33653, "epoch": 378, "lr": 9.635095980556312e-05}
{"train_loss": 0.29427316784858704, "global_step": 33654, "epoch": 378, "lr": 9.635074239317459e-05}
{"train_loss": 0.46591538190841675, "global_step": 33655, "epoch": 378, "lr": 9.635052497455475e-05}
{"train_loss": 0.3813488185405731, "global_step": 33656, "epoch": 378, "lr": 9.635030754970364e-05}
{"train_loss": 0.43975210189819336, "global_step": 33657, "epoch": 378, "lr": 9.635009011862132e-05}
{"train_loss": 0.44905731081962585, "global_step": 33658, "epoch": 378, "lr": 9.634987268130779e-05}
{"train_loss": 0.3039897680282593, "global_step": 33659, "epoch": 378, "lr": 9.634965523776306e-05}
{"train_loss": 0.29427939653396606, "global_step": 33660, "epoch": 378, "lr": 9.634943778798721e-05}
{"train_loss": 0.4909742474555969, "global_step": 33661, "epoch": 378, "lr": 9.634922033198024e-05}
{"train_loss": 0.39079827070236206, "global_step": 33662, "epoch": 378, "lr": 9.634900286974217e-05}
{"train_loss": 0.4343940019607544, "global_step": 33663, "epoch": 378, "lr": 9.634878540127305e-05}
{"train_loss": 0.39363810420036316, "global_step": 33664, "epoch": 378, "lr": 9.634856792657291e-05}
{"train_loss": 0.3004385232925415, "global_step": 33665, "epoch": 378, "lr": 9.634835044564175e-05}
{"train_loss": 0.3217158317565918, "global_step": 33666, "epoch": 378, "lr": 9.634813295847964e-05}
{"train_loss": 0.34964901208877563, "global_step": 33667, "epoch": 378, "lr": 9.634791546508657e-05}
{"train_loss": 0.3924350142478943, "global_step": 33668, "epoch": 378, "lr": 9.63476979654626e-05}
{"train_loss": 0.3533894717693329, "global_step": 33669, "epoch": 378, "lr": 9.634748045960775e-05}
{"train_loss": 0.35124990344047546, "global_step": 33670, "epoch": 378, "lr": 9.634726294752204e-05}
{"train_loss": 0.4022464156150818, "global_step": 33671, "epoch": 378, "lr": 9.63470454292055e-05}
{"train_loss": 0.4681446850299835, "global_step": 33672, "epoch": 378, "lr": 9.634682790465816e-05}
{"train_loss": 0.28043264150619507, "global_step": 33673, "epoch": 378, "lr": 9.634661037388006e-05}
{"train_loss": 0.4075815975666046, "global_step": 33674, "epoch": 378, "lr": 9.634639283687123e-05}
{"train_loss": 0.3464854061603546, "global_step": 33675, "epoch": 378, "lr": 9.634617529363168e-05}
{"train_loss": 0.37185290455818176, "global_step": 33676, "epoch": 378, "lr": 9.634595774416146e-05}
{"train_loss": 0.47891873121261597, "global_step": 33677, "epoch": 378, "lr": 9.634574018846059e-05}
{"train_loss": 0.39606672525405884, "global_step": 33678, "epoch": 378, "lr": 9.63455226265291e-05}
{"train_loss": 0.49060723185539246, "global_step": 33679, "epoch": 378, "lr": 9.634530505836704e-05}
{"train_loss": 0.27740776538848877, "global_step": 33680, "epoch": 378, "lr": 9.634508748397438e-05}
{"train_loss": 0.4916967451572418, "global_step": 33681, "epoch": 378, "lr": 9.634486990335121e-05}
{"train_loss": 0.32324665784835815, "global_step": 33682, "epoch": 378, "lr": 9.634465231649753e-05}
{"train_loss": 0.34492799639701843, "global_step": 33683, "epoch": 378, "lr": 9.634443472341338e-05}
{"train_loss": 0.37893593311309814, "global_step": 33684, "epoch": 378, "lr": 9.634421712409878e-05}
{"train_loss": 0.4429869055747986, "global_step": 33685, "epoch": 378, "lr": 9.634399951855378e-05}
{"train_loss": 0.4628696143627167, "global_step": 33686, "epoch": 378, "lr": 9.634378190677837e-05}
{"train_loss": 0.3922612965106964, "global_step": 33687, "epoch": 378, "lr": 9.634356428877262e-05}
{"train_loss": 0.302764892578125, "global_step": 33688, "epoch": 378, "lr": 9.634334666453655e-05}
{"train_loss": 0.3981587886810303, "global_step": 33689, "epoch": 378, "lr": 9.634312903407017e-05}
{"train_loss": 0.3953327238559723, "global_step": 33690, "epoch": 378, "lr": 9.634291139737353e-05}
{"train_loss": 0.4717869758605957, "global_step": 33691, "epoch": 378, "lr": 9.634269375444664e-05}
{"train_loss": 0.5967689156532288, "global_step": 33692, "epoch": 378, "lr": 9.634247610528955e-05}
{"train_loss": 0.5876764059066772, "global_step": 33693, "epoch": 378, "lr": 9.634225844990226e-05}
{"train_loss": 0.34011733531951904, "global_step": 33694, "epoch": 378, "lr": 9.634204078828484e-05}
{"train_loss": 0.36318883299827576, "global_step": 33695, "epoch": 378, "lr": 9.634182312043728e-05}
{"train_loss": 0.32243722677230835, "global_step": 33696, "epoch": 378, "lr": 9.634160544635964e-05}
{"train_loss": 0.35913634300231934, "global_step": 33697, "epoch": 378, "lr": 9.634138776605193e-05}
{"train_loss": 0.29228106141090393, "global_step": 33698, "epoch": 378, "lr": 9.634117007951417e-05}
{"train_loss": 0.29818424582481384, "global_step": 33699, "epoch": 378, "lr": 9.634095238674643e-05}
{"train_loss": 0.4028540253639221, "global_step": 33700, "epoch": 378, "lr": 9.634073468774871e-05}
{"train_loss": 0.5065153241157532, "global_step": 33701, "epoch": 378, "lr": 9.634051698252103e-05}
{"train_loss": 0.35190051794052124, "global_step": 33702, "epoch": 378, "lr": 9.634029927106345e-05}
{"train_loss": 0.35721370577812195, "global_step": 33703, "epoch": 378, "lr": 9.634008155337597e-05}
{"train_loss": 0.26800674200057983, "global_step": 33704, "epoch": 378, "lr": 9.633986382945862e-05}
{"train_loss": 0.39345842599868774, "global_step": 33705, "epoch": 378, "lr": 9.633964609931145e-05}
{"train_loss": 0.3539721369743347, "global_step": 33706, "epoch": 378, "lr": 9.633942836293448e-05}
{"train_loss": 0.42994043231010437, "global_step": 33707, "epoch": 378, "lr": 9.633921062032775e-05}
{"train_loss": 0.3007179796695709, "global_step": 33708, "epoch": 378, "lr": 9.633899287149127e-05}
{"train_loss": 0.26799526810646057, "global_step": 33709, "epoch": 378, "lr": 9.633877511642506e-05}
{"train_loss": 0.3430077135562897, "global_step": 33710, "epoch": 378, "lr": 9.633855735512918e-05}
{"train_loss": 0.35111182928085327, "global_step": 33711, "epoch": 378, "lr": 9.633833958760365e-05}
{"train_loss": 0.28802019357681274, "global_step": 33712, "epoch": 378, "lr": 9.633812181384849e-05}
{"train_loss": 0.2966802418231964, "global_step": 33713, "epoch": 378, "lr": 9.633790403386374e-05}
{"train_loss": 0.42419731616973877, "global_step": 33714, "epoch": 378, "lr": 9.633768624764943e-05}
{"train_loss": 0.37203696370124817, "global_step": 33715, "epoch": 378, "lr": 9.633746845520557e-05}
{"train_loss": 0.5059670209884644, "global_step": 33716, "epoch": 378, "lr": 9.63372506565322e-05}
{"train_loss": 0.3707472085952759, "global_step": 33717, "epoch": 378, "lr": 9.633703285162936e-05}
{"train_loss": 0.3965575695037842, "global_step": 33718, "epoch": 378, "lr": 9.633681504049707e-05}
{"train_loss": 0.35334527492523193, "global_step": 33719, "epoch": 378, "lr": 9.633659722313535e-05}
{"train_loss": 0.3428024649620056, "global_step": 33720, "epoch": 378, "lr": 9.633637939954425e-05}
{"train_loss": 0.41623592376708984, "global_step": 33721, "epoch": 378, "lr": 9.633616156972378e-05}
{"train_loss": 0.4275006055831909, "global_step": 33722, "epoch": 378, "lr": 9.633594373367398e-05}
{"train_loss": 0.42980480194091797, "global_step": 33723, "epoch": 378, "lr": 9.633572589139489e-05}
{"train_loss": 0.23240108788013458, "global_step": 33724, "epoch": 378, "lr": 9.633550804288652e-05}
{"train_loss": 0.3262748122215271, "global_step": 33725, "epoch": 378, "lr": 9.633529018814889e-05}
{"train_loss": 0.4265742301940918, "global_step": 33726, "epoch": 378, "lr": 9.633507232718206e-05}
{"train_loss": 0.29670947790145874, "global_step": 33727, "epoch": 378, "lr": 9.633485445998605e-05}
{"train_loss": 0.42894187569618225, "global_step": 33728, "epoch": 378, "lr": 9.633463658656088e-05}
{"train_loss": 0.431503027677536, "global_step": 33729, "epoch": 378, "lr": 9.633441870690657e-05}
{"train_loss": 0.3775834513179372, "global_step": 33730, "epoch": 378, "lr": 9.633420082102317e-05, "val_loss": 2.414066791534424}
{"train_loss": 0.4574567675590515, "global_step": 33731, "epoch": 379, "lr": 9.63339829289107e-05}
{"train_loss": 0.3433186709880829, "global_step": 33732, "epoch": 379, "lr": 9.63337650305692e-05}
{"train_loss": 0.4835411012172699, "global_step": 33733, "epoch": 379, "lr": 9.633354712599867e-05}
{"train_loss": 0.31768980622291565, "global_step": 33734, "epoch": 379, "lr": 9.633332921519917e-05}
{"train_loss": 0.36433038115501404, "global_step": 33735, "epoch": 379, "lr": 9.633311129817073e-05}
{"train_loss": 0.2548914849758148, "global_step": 33736, "epoch": 379, "lr": 9.633289337491335e-05}
{"train_loss": 0.43166834115982056, "global_step": 33737, "epoch": 379, "lr": 9.633267544542709e-05}
{"train_loss": 0.3203068971633911, "global_step": 33738, "epoch": 379, "lr": 9.633245750971196e-05}
{"train_loss": 0.3251846134662628, "global_step": 33739, "epoch": 379, "lr": 9.6332239567768e-05}
{"train_loss": 0.3038552403450012, "global_step": 33740, "epoch": 379, "lr": 9.633202161959522e-05}
{"train_loss": 0.33497798442840576, "global_step": 33741, "epoch": 379, "lr": 9.633180366519369e-05}
{"train_loss": 0.36330071091651917, "global_step": 33742, "epoch": 379, "lr": 9.63315857045634e-05}
{"train_loss": 0.37191274762153625, "global_step": 33743, "epoch": 379, "lr": 9.633136773770439e-05}
{"train_loss": 0.3150825798511505, "global_step": 33744, "epoch": 379, "lr": 9.63311497646167e-05}
{"train_loss": 0.36407071352005005, "global_step": 33745, "epoch": 379, "lr": 9.633093178530035e-05}
{"train_loss": 0.2996586561203003, "global_step": 33746, "epoch": 379, "lr": 9.633071379975535e-05}
{"train_loss": 0.24484503269195557, "global_step": 33747, "epoch": 379, "lr": 9.63304958079818e-05}
{"train_loss": 0.37976017594337463, "global_step": 33748, "epoch": 379, "lr": 9.633027780997963e-05}
{"train_loss": 0.42614954710006714, "global_step": 33749, "epoch": 379, "lr": 9.633005980574894e-05}
{"train_loss": 0.39183786511421204, "global_step": 33750, "epoch": 379, "lr": 9.632984179528973e-05}
{"train_loss": 0.2721484303474426, "global_step": 33751, "epoch": 379, "lr": 9.632962377860205e-05}
{"train_loss": 0.42133811116218567, "global_step": 33752, "epoch": 379, "lr": 9.632940575568589e-05}
{"train_loss": 0.30261415243148804, "global_step": 33753, "epoch": 379, "lr": 9.632918772654134e-05}
{"train_loss": 0.28885599970817566, "global_step": 33754, "epoch": 379, "lr": 9.632896969116838e-05}
{"train_loss": 0.43566393852233887, "global_step": 33755, "epoch": 379, "lr": 9.632875164956705e-05}
{"train_loss": 0.3559293746948242, "global_step": 33756, "epoch": 379, "lr": 9.63285336017374e-05}
{"train_loss": 0.3659394681453705, "global_step": 33757, "epoch": 379, "lr": 9.632831554767942e-05}
{"train_loss": 0.270273357629776, "global_step": 33758, "epoch": 379, "lr": 9.632809748739317e-05}
{"train_loss": 0.414654403924942, "global_step": 33759, "epoch": 379, "lr": 9.632787942087869e-05}
{"train_loss": 0.4081646502017975, "global_step": 33760, "epoch": 379, "lr": 9.632766134813598e-05}
{"train_loss": 0.41144829988479614, "global_step": 33761, "epoch": 379, "lr": 9.632744326916507e-05}
{"train_loss": 0.3146555721759796, "global_step": 33762, "epoch": 379, "lr": 9.632722518396603e-05}
{"train_loss": 0.3001146912574768, "global_step": 33763, "epoch": 379, "lr": 9.632700709253884e-05}
{"train_loss": 0.2954951822757721, "global_step": 33764, "epoch": 379, "lr": 9.632678899488353e-05}
{"train_loss": 0.4347938895225525, "global_step": 33765, "epoch": 379, "lr": 9.632657089100016e-05}
{"train_loss": 0.29479843378067017, "global_step": 33766, "epoch": 379, "lr": 9.632635278088876e-05}
{"train_loss": 0.2937992513179779, "global_step": 33767, "epoch": 379, "lr": 9.632613466454934e-05}
{"train_loss": 0.32834309339523315, "global_step": 33768, "epoch": 379, "lr": 9.632591654198193e-05}
{"train_loss": 0.396049827337265, "global_step": 33769, "epoch": 379, "lr": 9.632569841318657e-05}
{"train_loss": 0.3432576656341553, "global_step": 33770, "epoch": 379, "lr": 9.632548027816328e-05}
{"train_loss": 0.34097549319267273, "global_step": 33771, "epoch": 379, "lr": 9.632526213691211e-05}
{"train_loss": 0.3142586648464203, "global_step": 33772, "epoch": 379, "lr": 9.632504398943305e-05}
{"train_loss": 0.2969706356525421, "global_step": 33773, "epoch": 379, "lr": 9.632482583572617e-05}
{"train_loss": 0.3391972780227661, "global_step": 33774, "epoch": 379, "lr": 9.632460767579148e-05}
{"train_loss": 0.41447702050209045, "global_step": 33775, "epoch": 379, "lr": 9.632438950962901e-05}
{"train_loss": 0.34347087144851685, "global_step": 33776, "epoch": 379, "lr": 9.632417133723878e-05}
{"train_loss": 0.35315877199172974, "global_step": 33777, "epoch": 379, "lr": 9.632395315862084e-05}
{"train_loss": 0.22894342243671417, "global_step": 33778, "epoch": 379, "lr": 9.63237349737752e-05}
{"train_loss": 0.4115602970123291, "global_step": 33779, "epoch": 379, "lr": 9.632351678270191e-05}
{"train_loss": 0.37927430868148804, "global_step": 33780, "epoch": 379, "lr": 9.632329858540098e-05}
{"train_loss": 0.40624529123306274, "global_step": 33781, "epoch": 379, "lr": 9.632308038187244e-05}
{"train_loss": 0.40629372000694275, "global_step": 33782, "epoch": 379, "lr": 9.632286217211635e-05}
{"train_loss": 0.28252628445625305, "global_step": 33783, "epoch": 379, "lr": 9.632264395613268e-05}
{"train_loss": 0.32399165630340576, "global_step": 33784, "epoch": 379, "lr": 9.632242573392152e-05}
{"train_loss": 0.3723878860473633, "global_step": 33785, "epoch": 379, "lr": 9.632220750548289e-05}
{"train_loss": 0.22973422706127167, "global_step": 33786, "epoch": 379, "lr": 9.632198927081677e-05}
{"train_loss": 0.3703969419002533, "global_step": 33787, "epoch": 379, "lr": 9.632177102992323e-05}
{"train_loss": 0.2512264847755432, "global_step": 33788, "epoch": 379, "lr": 9.632155278280231e-05}
{"train_loss": 0.3740909993648529, "global_step": 33789, "epoch": 379, "lr": 9.632133452945403e-05}
{"train_loss": 0.28490757942199707, "global_step": 33790, "epoch": 379, "lr": 9.632111626987837e-05}
{"train_loss": 0.30653709173202515, "global_step": 33791, "epoch": 379, "lr": 9.632089800407544e-05}
{"train_loss": 0.3048849105834961, "global_step": 33792, "epoch": 379, "lr": 9.63206797320452e-05}
{"train_loss": 0.30785855650901794, "global_step": 33793, "epoch": 379, "lr": 9.632046145378772e-05}
{"train_loss": 0.3024243414402008, "global_step": 33794, "epoch": 379, "lr": 9.632024316930303e-05}
{"train_loss": 0.38480257987976074, "global_step": 33795, "epoch": 379, "lr": 9.632002487859113e-05}
{"train_loss": 0.3190115988254547, "global_step": 33796, "epoch": 379, "lr": 9.631980658165208e-05}
{"train_loss": 0.3743879199028015, "global_step": 33797, "epoch": 379, "lr": 9.63195882784859e-05}
{"train_loss": 0.36008647084236145, "global_step": 33798, "epoch": 379, "lr": 9.63193699690926e-05}
{"train_loss": 0.23699739575386047, "global_step": 33799, "epoch": 379, "lr": 9.631915165347223e-05}
{"train_loss": 0.3113749325275421, "global_step": 33800, "epoch": 379, "lr": 9.631893333162483e-05}
{"train_loss": 0.35903826355934143, "global_step": 33801, "epoch": 379, "lr": 9.631871500355041e-05}
{"train_loss": 0.4463171362876892, "global_step": 33802, "epoch": 379, "lr": 9.631849666924899e-05}
{"train_loss": 0.30842292308807373, "global_step": 33803, "epoch": 379, "lr": 9.631827832872061e-05}
{"train_loss": 0.35237663984298706, "global_step": 33804, "epoch": 379, "lr": 9.631805998196531e-05}
{"train_loss": 0.3485356867313385, "global_step": 33805, "epoch": 379, "lr": 9.631784162898312e-05}
{"train_loss": 0.2587515711784363, "global_step": 33806, "epoch": 379, "lr": 9.631762326977405e-05}
{"train_loss": 0.3659078776836395, "global_step": 33807, "epoch": 379, "lr": 9.631740490433813e-05}
{"train_loss": 0.30310744047164917, "global_step": 33808, "epoch": 379, "lr": 9.631718653267543e-05}
{"train_loss": 0.2939216196537018, "global_step": 33809, "epoch": 379, "lr": 9.631696815478592e-05}
{"train_loss": 0.37370771169662476, "global_step": 33810, "epoch": 379, "lr": 9.631674977066967e-05}
{"train_loss": 0.41376951336860657, "global_step": 33811, "epoch": 379, "lr": 9.631653138032671e-05}
{"train_loss": 0.454875111579895, "global_step": 33812, "epoch": 379, "lr": 9.631631298375703e-05}
{"train_loss": 0.39475783705711365, "global_step": 33813, "epoch": 379, "lr": 9.63160945809607e-05}
{"train_loss": 0.39070925116539, "global_step": 33814, "epoch": 379, "lr": 9.631587617193774e-05}
{"train_loss": 0.5446354746818542, "global_step": 33815, "epoch": 379, "lr": 9.631565775668818e-05}
{"train_loss": 0.41481807827949524, "global_step": 33816, "epoch": 379, "lr": 9.631543933521204e-05}
{"train_loss": 0.3918570876121521, "global_step": 33817, "epoch": 379, "lr": 9.631522090750933e-05}
{"train_loss": 0.38467732071876526, "global_step": 33818, "epoch": 379, "lr": 9.631500247358011e-05}
{"train_loss": 0.34928079267566126, "global_step": 33819, "epoch": 379, "lr": 9.631478403342442e-05, "val_loss": 2.4156863689422607}
{"train_loss": 0.4736325144767761, "global_step": 33820, "epoch": 380, "lr": 9.631456558704229e-05}
{"train_loss": 0.4188714325428009, "global_step": 33821, "epoch": 380, "lr": 9.631434713443369e-05}
{"train_loss": 0.43128272891044617, "global_step": 33822, "epoch": 380, "lr": 9.63141286755987e-05}
{"train_loss": 0.3098876476287842, "global_step": 33823, "epoch": 380, "lr": 9.631391021053735e-05}
{"train_loss": 0.37143319845199585, "global_step": 33824, "epoch": 380, "lr": 9.631369173924967e-05}
{"train_loss": 0.32813796401023865, "global_step": 33825, "epoch": 380, "lr": 9.631347326173565e-05}
{"train_loss": 0.3296733796596527, "global_step": 33826, "epoch": 380, "lr": 9.631325477799537e-05}
{"train_loss": 0.3221574127674103, "global_step": 33827, "epoch": 380, "lr": 9.631303628802884e-05}
{"train_loss": 0.3249669671058655, "global_step": 33828, "epoch": 380, "lr": 9.631281779183608e-05}
{"train_loss": 0.4097997844219208, "global_step": 33829, "epoch": 380, "lr": 9.631259928941712e-05}
{"train_loss": 0.48099207878112793, "global_step": 33830, "epoch": 380, "lr": 9.631238078077198e-05}
{"train_loss": 0.36028552055358887, "global_step": 33831, "epoch": 380, "lr": 9.631216226590074e-05}
{"train_loss": 0.28891628980636597, "global_step": 33832, "epoch": 380, "lr": 9.631194374480337e-05}
{"train_loss": 0.24055129289627075, "global_step": 33833, "epoch": 380, "lr": 9.631172521747993e-05}
{"train_loss": 0.25895532965660095, "global_step": 33834, "epoch": 380, "lr": 9.631150668393046e-05}
{"train_loss": 0.27193617820739746, "global_step": 33835, "epoch": 380, "lr": 9.631128814415495e-05}
{"train_loss": 0.39628100395202637, "global_step": 33836, "epoch": 380, "lr": 9.631106959815346e-05}
{"train_loss": 0.35266992449760437, "global_step": 33837, "epoch": 380, "lr": 9.631085104592601e-05}
{"train_loss": 0.41277724504470825, "global_step": 33838, "epoch": 380, "lr": 9.631063248747262e-05}
{"train_loss": 0.34876683354377747, "global_step": 33839, "epoch": 380, "lr": 9.631041392279334e-05}
{"train_loss": 0.3589668571949005, "global_step": 33840, "epoch": 380, "lr": 9.63101953518882e-05}
{"train_loss": 0.34698373079299927, "global_step": 33841, "epoch": 380, "lr": 9.630997677475719e-05}
{"train_loss": 0.37701117992401123, "global_step": 33842, "epoch": 380, "lr": 9.63097581914004e-05}
{"train_loss": 0.317958265542984, "global_step": 33843, "epoch": 380, "lr": 9.630953960181781e-05}
{"train_loss": 0.3037823736667633, "global_step": 33844, "epoch": 380, "lr": 9.630932100600947e-05}
{"train_loss": 0.3134775161743164, "global_step": 33845, "epoch": 380, "lr": 9.630910240397541e-05}
{"train_loss": 0.3029800057411194, "global_step": 33846, "epoch": 380, "lr": 9.630888379571565e-05}
{"train_loss": 0.3184819519519806, "global_step": 33847, "epoch": 380, "lr": 9.630866518123023e-05}
{"train_loss": 0.34618598222732544, "global_step": 33848, "epoch": 380, "lr": 9.630844656051917e-05}
{"train_loss": 0.4447031319141388, "global_step": 33849, "epoch": 380, "lr": 9.63082279335825e-05}
{"train_loss": 0.32398533821105957, "global_step": 33850, "epoch": 380, "lr": 9.630800930042027e-05}
{"train_loss": 0.33909696340560913, "global_step": 33851, "epoch": 380, "lr": 9.630779066103248e-05}
{"train_loss": 0.37956762313842773, "global_step": 33852, "epoch": 380, "lr": 9.630757201541917e-05}
{"train_loss": 0.3326910734176636, "global_step": 33853, "epoch": 380, "lr": 9.630735336358036e-05}
{"train_loss": 0.4218539893627167, "global_step": 33854, "epoch": 380, "lr": 9.630713470551611e-05}
{"train_loss": 0.31484681367874146, "global_step": 33855, "epoch": 380, "lr": 9.630691604122643e-05}
{"train_loss": 0.3497544825077057, "global_step": 33856, "epoch": 380, "lr": 9.630669737071136e-05}
{"train_loss": 0.4343375563621521, "global_step": 33857, "epoch": 380, "lr": 9.630647869397089e-05}
{"train_loss": 0.32192766666412354, "global_step": 33858, "epoch": 380, "lr": 9.630626001100509e-05}
{"train_loss": 0.3057533800601959, "global_step": 33859, "epoch": 380, "lr": 9.630604132181398e-05}
{"train_loss": 0.23545895516872406, "global_step": 33860, "epoch": 380, "lr": 9.63058226263976e-05}
{"train_loss": 0.2591360807418823, "global_step": 33861, "epoch": 380, "lr": 9.630560392475596e-05}
{"train_loss": 0.34496375918388367, "global_step": 33862, "epoch": 380, "lr": 9.630538521688907e-05}
{"train_loss": 0.3249184787273407, "global_step": 33863, "epoch": 380, "lr": 9.630516650279702e-05}
{"train_loss": 0.3756582736968994, "global_step": 33864, "epoch": 380, "lr": 9.630494778247978e-05}
{"train_loss": 0.30663859844207764, "global_step": 33865, "epoch": 380, "lr": 9.630472905593741e-05}
{"train_loss": 0.3440137803554535, "global_step": 33866, "epoch": 380, "lr": 9.630451032316994e-05}
{"train_loss": 0.3294123709201813, "global_step": 33867, "epoch": 380, "lr": 9.63042915841774e-05}
{"train_loss": 0.2921750247478485, "global_step": 33868, "epoch": 380, "lr": 9.630407283895979e-05}
{"train_loss": 0.3211155831813812, "global_step": 33869, "epoch": 380, "lr": 9.630385408751718e-05}
{"train_loss": 0.35993313789367676, "global_step": 33870, "epoch": 380, "lr": 9.630363532984958e-05}
{"train_loss": 0.430284321308136, "global_step": 33871, "epoch": 380, "lr": 9.630341656595701e-05}
{"train_loss": 0.3276086747646332, "global_step": 33872, "epoch": 380, "lr": 9.630319779583951e-05}
{"train_loss": 0.354997456073761, "global_step": 33873, "epoch": 380, "lr": 9.630297901949712e-05}
{"train_loss": 0.280059814453125, "global_step": 33874, "epoch": 380, "lr": 9.630276023692986e-05}
{"train_loss": 0.30024826526641846, "global_step": 33875, "epoch": 380, "lr": 9.630254144813775e-05}
{"train_loss": 0.3839346170425415, "global_step": 33876, "epoch": 380, "lr": 9.630232265312084e-05}
{"train_loss": 0.33920928835868835, "global_step": 33877, "epoch": 380, "lr": 9.630210385187914e-05}
{"train_loss": 0.3137468993663788, "global_step": 33878, "epoch": 380, "lr": 9.630188504441267e-05}
{"train_loss": 0.2754717171192169, "global_step": 33879, "epoch": 380, "lr": 9.63016662307215e-05}
{"train_loss": 0.3579329252243042, "global_step": 33880, "epoch": 380, "lr": 9.630144741080563e-05}
{"train_loss": 0.27714383602142334, "global_step": 33881, "epoch": 380, "lr": 9.630122858466508e-05}
{"train_loss": 0.419549822807312, "global_step": 33882, "epoch": 380, "lr": 9.630100975229989e-05}
{"train_loss": 0.32882562279701233, "global_step": 33883, "epoch": 380, "lr": 9.630079091371012e-05}
{"train_loss": 0.4073942303657532, "global_step": 33884, "epoch": 380, "lr": 9.630057206889575e-05}
{"train_loss": 0.3630673587322235, "global_step": 33885, "epoch": 380, "lr": 9.630035321785686e-05}
{"train_loss": 0.3070968687534332, "global_step": 33886, "epoch": 380, "lr": 9.630013436059343e-05}
{"train_loss": 0.3844161033630371, "global_step": 33887, "epoch": 380, "lr": 9.629991549710551e-05}
{"train_loss": 0.44397324323654175, "global_step": 33888, "epoch": 380, "lr": 9.629969662739315e-05}
{"train_loss": 0.3357676863670349, "global_step": 33889, "epoch": 380, "lr": 9.629947775145634e-05}
{"train_loss": 0.3445397615432739, "global_step": 33890, "epoch": 380, "lr": 9.629925886929513e-05}
{"train_loss": 0.33489152789115906, "global_step": 33891, "epoch": 380, "lr": 9.629903998090957e-05}
{"train_loss": 0.31418687105178833, "global_step": 33892, "epoch": 380, "lr": 9.629882108629964e-05}
{"train_loss": 0.37559276819229126, "global_step": 33893, "epoch": 380, "lr": 9.629860218546542e-05}
{"train_loss": 0.25093039870262146, "global_step": 33894, "epoch": 380, "lr": 9.62983832784069e-05}
{"train_loss": 0.32380977272987366, "global_step": 33895, "epoch": 380, "lr": 9.629816436512415e-05}
{"train_loss": 0.24151025712490082, "global_step": 33896, "epoch": 380, "lr": 9.629794544561716e-05}
{"train_loss": 0.386075884103775, "global_step": 33897, "epoch": 380, "lr": 9.629772651988598e-05}
{"train_loss": 0.39618831872940063, "global_step": 33898, "epoch": 380, "lr": 9.629750758793062e-05}
{"train_loss": 0.3166041970252991, "global_step": 33899, "epoch": 380, "lr": 9.629728864975114e-05}
{"train_loss": 0.35926368832588196, "global_step": 33900, "epoch": 380, "lr": 9.629706970534755e-05}
{"train_loss": 0.3777124285697937, "global_step": 33901, "epoch": 380, "lr": 9.629685075471987e-05}
{"train_loss": 0.3897857666015625, "global_step": 33902, "epoch": 380, "lr": 9.629663179786816e-05}
{"train_loss": 0.3509604334831238, "global_step": 33903, "epoch": 380, "lr": 9.629641283479243e-05}
{"train_loss": 0.39713865518569946, "global_step": 33904, "epoch": 380, "lr": 9.629619386549271e-05}
{"train_loss": 0.327907532453537, "global_step": 33905, "epoch": 380, "lr": 9.629597488996904e-05}
{"train_loss": 0.37899690866470337, "global_step": 33906, "epoch": 380, "lr": 9.629575590822143e-05}
{"train_loss": 0.39413872361183167, "global_step": 33907, "epoch": 380, "lr": 9.629553692024993e-05}
{"train_loss": 0.3484209132998177, "global_step": 33908, "epoch": 380, "lr": 9.629531792605454e-05, "val_loss": 2.4494950771331787, "train_action_mse_error": 30.38117218017578}
{"train_loss": 0.33611926436424255, "global_step": 33909, "epoch": 381, "lr": 9.629509892563533e-05}
{"train_loss": 0.39258089661598206, "global_step": 33910, "epoch": 381, "lr": 9.629487991899228e-05}
{"train_loss": 0.29218003153800964, "global_step": 33911, "epoch": 381, "lr": 9.629466090612548e-05}
{"train_loss": 0.29295262694358826, "global_step": 33912, "epoch": 381, "lr": 9.629444188703491e-05}
{"train_loss": 0.384152352809906, "global_step": 33913, "epoch": 381, "lr": 9.629422286172061e-05}
{"train_loss": 0.3006141185760498, "global_step": 33914, "epoch": 381, "lr": 9.629400383018263e-05}
{"train_loss": 0.3552856743335724, "global_step": 33915, "epoch": 381, "lr": 9.629378479242098e-05}
{"train_loss": 0.39352673292160034, "global_step": 33916, "epoch": 381, "lr": 9.62935657484357e-05}
{"train_loss": 0.40842074155807495, "global_step": 33917, "epoch": 381, "lr": 9.62933466982268e-05}
{"train_loss": 0.41783082485198975, "global_step": 33918, "epoch": 381, "lr": 9.629312764179435e-05}
{"train_loss": 0.43987157940864563, "global_step": 33919, "epoch": 381, "lr": 9.629290857913834e-05}
{"train_loss": 0.3961741626262665, "global_step": 33920, "epoch": 381, "lr": 9.629268951025879e-05}
{"train_loss": 0.38495492935180664, "global_step": 33921, "epoch": 381, "lr": 9.629247043515577e-05}
{"train_loss": 0.38326436281204224, "global_step": 33922, "epoch": 381, "lr": 9.62922513538293e-05}
{"train_loss": 0.38284194469451904, "global_step": 33923, "epoch": 381, "lr": 9.629203226627939e-05}
{"train_loss": 0.3971472978591919, "global_step": 33924, "epoch": 381, "lr": 9.629181317250609e-05}
{"train_loss": 0.3587266504764557, "global_step": 33925, "epoch": 381, "lr": 9.629159407250941e-05}
{"train_loss": 0.3169662654399872, "global_step": 33926, "epoch": 381, "lr": 9.62913749662894e-05}
{"train_loss": 0.4788317382335663, "global_step": 33927, "epoch": 381, "lr": 9.629115585384607e-05}
{"train_loss": 0.35088643431663513, "global_step": 33928, "epoch": 381, "lr": 9.629093673517946e-05}
{"train_loss": 0.3453812897205353, "global_step": 33929, "epoch": 381, "lr": 9.62907176102896e-05}
{"train_loss": 0.3401526212692261, "global_step": 33930, "epoch": 381, "lr": 9.629049847917652e-05}
{"train_loss": 0.3688497245311737, "global_step": 33931, "epoch": 381, "lr": 9.629027934184023e-05}
{"train_loss": 0.2651388347148895, "global_step": 33932, "epoch": 381, "lr": 9.62900601982808e-05}
{"train_loss": 0.32011720538139343, "global_step": 33933, "epoch": 381, "lr": 9.628984104849822e-05}
{"train_loss": 0.3134182095527649, "global_step": 33934, "epoch": 381, "lr": 9.628962189249253e-05}
{"train_loss": 0.2610076069831848, "global_step": 33935, "epoch": 381, "lr": 9.628940273026378e-05}
{"train_loss": 0.2996167540550232, "global_step": 33936, "epoch": 381, "lr": 9.628918356181198e-05}
{"train_loss": 0.2533256709575653, "global_step": 33937, "epoch": 381, "lr": 9.628896438713717e-05}
{"train_loss": 0.32319095730781555, "global_step": 33938, "epoch": 381, "lr": 9.628874520623936e-05}
{"train_loss": 0.3757752478122711, "global_step": 33939, "epoch": 381, "lr": 9.62885260191186e-05}
{"train_loss": 0.3905847370624542, "global_step": 33940, "epoch": 381, "lr": 9.62883068257749e-05}
{"train_loss": 0.25873684883117676, "global_step": 33941, "epoch": 381, "lr": 9.628808762620832e-05}
{"train_loss": 0.3140958845615387, "global_step": 33942, "epoch": 381, "lr": 9.628786842041886e-05}
{"train_loss": 0.38693663477897644, "global_step": 33943, "epoch": 381, "lr": 9.628764920840656e-05}
{"train_loss": 0.34321799874305725, "global_step": 33944, "epoch": 381, "lr": 9.628742999017143e-05}
{"train_loss": 0.24672962725162506, "global_step": 33945, "epoch": 381, "lr": 9.628721076571355e-05}
{"train_loss": 0.29887083172798157, "global_step": 33946, "epoch": 381, "lr": 9.62869915350329e-05}
{"train_loss": 0.34093615412712097, "global_step": 33947, "epoch": 381, "lr": 9.628677229812955e-05}
{"train_loss": 0.3431592285633087, "global_step": 33948, "epoch": 381, "lr": 9.628655305500349e-05}
{"train_loss": 0.36261850595474243, "global_step": 33949, "epoch": 381, "lr": 9.628633380565477e-05}
{"train_loss": 0.40241536498069763, "global_step": 33950, "epoch": 381, "lr": 9.628611455008343e-05}
{"train_loss": 0.4364122748374939, "global_step": 33951, "epoch": 381, "lr": 9.628589528828946e-05}
{"train_loss": 0.3404892385005951, "global_step": 33952, "epoch": 381, "lr": 9.628567602027294e-05}
{"train_loss": 0.33292800188064575, "global_step": 33953, "epoch": 381, "lr": 9.628545674603386e-05}
{"train_loss": 0.31729376316070557, "global_step": 33954, "epoch": 381, "lr": 9.628523746557227e-05}
{"train_loss": 0.41349488496780396, "global_step": 33955, "epoch": 381, "lr": 9.62850181788882e-05}
{"train_loss": 0.2618580460548401, "global_step": 33956, "epoch": 381, "lr": 9.628479888598165e-05}
{"train_loss": 0.43884363770484924, "global_step": 33957, "epoch": 381, "lr": 9.62845795868527e-05}
{"train_loss": 0.32128915190696716, "global_step": 33958, "epoch": 381, "lr": 9.628436028150134e-05}
{"train_loss": 0.3455355167388916, "global_step": 33959, "epoch": 381, "lr": 9.628414096992761e-05}
{"train_loss": 0.33492738008499146, "global_step": 33960, "epoch": 381, "lr": 9.628392165213155e-05}
{"train_loss": 0.3275132477283478, "global_step": 33961, "epoch": 381, "lr": 9.628370232811317e-05}
{"train_loss": 0.3597210645675659, "global_step": 33962, "epoch": 381, "lr": 9.628348299787252e-05}
{"train_loss": 0.37573933601379395, "global_step": 33963, "epoch": 381, "lr": 9.628326366140961e-05}
{"train_loss": 0.34153345227241516, "global_step": 33964, "epoch": 381, "lr": 9.628304431872448e-05}
{"train_loss": 0.33744317293167114, "global_step": 33965, "epoch": 381, "lr": 9.628282496981717e-05}
{"train_loss": 0.4253862798213959, "global_step": 33966, "epoch": 381, "lr": 9.62826056146877e-05}
{"train_loss": 0.4201939105987549, "global_step": 33967, "epoch": 381, "lr": 9.628238625333609e-05}
{"train_loss": 0.5064989328384399, "global_step": 33968, "epoch": 381, "lr": 9.628216688576237e-05}
{"train_loss": 0.30393385887145996, "global_step": 33969, "epoch": 381, "lr": 9.62819475119666e-05}
{"train_loss": 0.3735113739967346, "global_step": 33970, "epoch": 381, "lr": 9.628172813194876e-05}
{"train_loss": 0.31835633516311646, "global_step": 33971, "epoch": 381, "lr": 9.628150874570891e-05}
{"train_loss": 0.3429297208786011, "global_step": 33972, "epoch": 381, "lr": 9.62812893532471e-05}
{"train_loss": 0.5069833993911743, "global_step": 33973, "epoch": 381, "lr": 9.62810699545633e-05}
{"train_loss": 0.3507956564426422, "global_step": 33974, "epoch": 381, "lr": 9.628085054965758e-05}
{"train_loss": 0.3444899320602417, "global_step": 33975, "epoch": 381, "lr": 9.628063113852998e-05}
{"train_loss": 0.3582018315792084, "global_step": 33976, "epoch": 381, "lr": 9.628041172118053e-05}
{"train_loss": 0.38245704770088196, "global_step": 33977, "epoch": 381, "lr": 9.628019229760921e-05}
{"train_loss": 0.3850802779197693, "global_step": 33978, "epoch": 381, "lr": 9.62799728678161e-05}
{"train_loss": 0.43437597155570984, "global_step": 33979, "epoch": 381, "lr": 9.627975343180119e-05}
{"train_loss": 0.34018024802207947, "global_step": 33980, "epoch": 381, "lr": 9.627953398956455e-05}
{"train_loss": 0.33130720257759094, "global_step": 33981, "epoch": 381, "lr": 9.627931454110619e-05}
{"train_loss": 0.38528838753700256, "global_step": 33982, "epoch": 381, "lr": 9.627909508642612e-05}
{"train_loss": 0.3664992153644562, "global_step": 33983, "epoch": 381, "lr": 9.627887562552443e-05}
{"train_loss": 0.3129717707633972, "global_step": 33984, "epoch": 381, "lr": 9.627865615840107e-05}
{"train_loss": 0.3780304193496704, "global_step": 33985, "epoch": 381, "lr": 9.627843668505612e-05}
{"train_loss": 0.39975523948669434, "global_step": 33986, "epoch": 381, "lr": 9.627821720548961e-05}
{"train_loss": 0.28278419375419617, "global_step": 33987, "epoch": 381, "lr": 9.627799771970153e-05}
{"train_loss": 0.47521108388900757, "global_step": 33988, "epoch": 381, "lr": 9.627777822769198e-05}
{"train_loss": 0.4533534049987793, "global_step": 33989, "epoch": 381, "lr": 9.627755872946092e-05}
{"train_loss": 0.3244512677192688, "global_step": 33990, "epoch": 381, "lr": 9.627733922500841e-05}
{"train_loss": 0.28729093074798584, "global_step": 33991, "epoch": 381, "lr": 9.627711971433447e-05}
{"train_loss": 0.24530956149101257, "global_step": 33992, "epoch": 381, "lr": 9.627690019743914e-05}
{"train_loss": 0.29709580540657043, "global_step": 33993, "epoch": 381, "lr": 9.627668067432245e-05}
{"train_loss": 0.32316097617149353, "global_step": 33994, "epoch": 381, "lr": 9.627646114498442e-05}
{"train_loss": 0.33607885241508484, "global_step": 33995, "epoch": 381, "lr": 9.627624160942509e-05}
{"train_loss": 0.39966723322868347, "global_step": 33996, "epoch": 381, "lr": 9.627602206764448e-05}
{"train_loss": 0.3547136162439089, "global_step": 33997, "epoch": 381, "lr": 9.627580251964261e-05, "val_loss": 2.3399484157562256}
{"train_loss": 0.34158486127853394, "global_step": 33998, "epoch": 382, "lr": 9.627558296541954e-05}
{"train_loss": 0.32535985112190247, "global_step": 33999, "epoch": 382, "lr": 9.627536340497527e-05}
{"train_loss": 0.3123983144760132, "global_step": 34000, "epoch": 382, "lr": 9.627514383830984e-05}
{"train_loss": 0.30251994729042053, "global_step": 34001, "epoch": 382, "lr": 9.627492426542329e-05}
{"train_loss": 0.4385432004928589, "global_step": 34002, "epoch": 382, "lr": 9.627470468631564e-05}
{"train_loss": 0.30588987469673157, "global_step": 34003, "epoch": 382, "lr": 9.627448510098692e-05}
{"train_loss": 0.3933757543563843, "global_step": 34004, "epoch": 382, "lr": 9.627426550943716e-05}
{"train_loss": 0.29571929574012756, "global_step": 34005, "epoch": 382, "lr": 9.627404591166638e-05}
{"train_loss": 0.41941142082214355, "global_step": 34006, "epoch": 382, "lr": 9.627382630767463e-05}
{"train_loss": 0.3225032389163971, "global_step": 34007, "epoch": 382, "lr": 9.627360669746193e-05}
{"train_loss": 0.34534862637519836, "global_step": 34008, "epoch": 382, "lr": 9.627338708102829e-05}
{"train_loss": 0.43808627128601074, "global_step": 34009, "epoch": 382, "lr": 9.627316745837376e-05}
{"train_loss": 0.29747864603996277, "global_step": 34010, "epoch": 382, "lr": 9.627294782949839e-05}
{"train_loss": 0.3313291370868683, "global_step": 34011, "epoch": 382, "lr": 9.627272819440216e-05}
{"train_loss": 0.37678292393684387, "global_step": 34012, "epoch": 382, "lr": 9.627250855308514e-05}
{"train_loss": 0.42570191621780396, "global_step": 34013, "epoch": 382, "lr": 9.627228890554734e-05}
{"train_loss": 0.2727089524269104, "global_step": 34014, "epoch": 382, "lr": 9.62720692517888e-05}
{"train_loss": 0.314404159784317, "global_step": 34015, "epoch": 382, "lr": 9.627184959180953e-05}
{"train_loss": 0.31096151471138, "global_step": 34016, "epoch": 382, "lr": 9.627162992560959e-05}
{"train_loss": 0.2820882499217987, "global_step": 34017, "epoch": 382, "lr": 9.627141025318898e-05}
{"train_loss": 0.39521992206573486, "global_step": 34018, "epoch": 382, "lr": 9.627119057454775e-05}
{"train_loss": 0.31745022535324097, "global_step": 34019, "epoch": 382, "lr": 9.627097088968593e-05}
{"train_loss": 0.2669838070869446, "global_step": 34020, "epoch": 382, "lr": 9.627075119860354e-05}
{"train_loss": 0.3620961308479309, "global_step": 34021, "epoch": 382, "lr": 9.627053150130058e-05}
{"train_loss": 0.3507747948169708, "global_step": 34022, "epoch": 382, "lr": 9.627031179777714e-05}
{"train_loss": 0.3290342390537262, "global_step": 34023, "epoch": 382, "lr": 9.627009208803324e-05}
{"train_loss": 0.5047532320022583, "global_step": 34024, "epoch": 382, "lr": 9.626987237206885e-05}
{"train_loss": 0.31344133615493774, "global_step": 34025, "epoch": 382, "lr": 9.626965264988406e-05}
{"train_loss": 0.2274922877550125, "global_step": 34026, "epoch": 382, "lr": 9.626943292147888e-05}
{"train_loss": 0.34262287616729736, "global_step": 34027, "epoch": 382, "lr": 9.626921318685334e-05}
{"train_loss": 0.3353898823261261, "global_step": 34028, "epoch": 382, "lr": 9.626899344600746e-05}
{"train_loss": 0.3245191276073456, "global_step": 34029, "epoch": 382, "lr": 9.626877369894127e-05}
{"train_loss": 0.41088783740997314, "global_step": 34030, "epoch": 382, "lr": 9.626855394565483e-05}
{"train_loss": 0.39707475900650024, "global_step": 34031, "epoch": 382, "lr": 9.626833418614814e-05}
{"train_loss": 0.32508546113967896, "global_step": 34032, "epoch": 382, "lr": 9.626811442042122e-05}
{"train_loss": 0.38509321212768555, "global_step": 34033, "epoch": 382, "lr": 9.626789464847412e-05}
{"train_loss": 0.2909691333770752, "global_step": 34034, "epoch": 382, "lr": 9.626767487030687e-05}
{"train_loss": 0.3472710847854614, "global_step": 34035, "epoch": 382, "lr": 9.626745508591952e-05}
{"train_loss": 0.39162492752075195, "global_step": 34036, "epoch": 382, "lr": 9.626723529531205e-05}
{"train_loss": 0.44148343801498413, "global_step": 34037, "epoch": 382, "lr": 9.626701549848451e-05}
{"train_loss": 0.28319522738456726, "global_step": 34038, "epoch": 382, "lr": 9.626679569543694e-05}
{"train_loss": 0.3436111807823181, "global_step": 34039, "epoch": 382, "lr": 9.626657588616936e-05}
{"train_loss": 0.3817501962184906, "global_step": 34040, "epoch": 382, "lr": 9.626635607068181e-05}
{"train_loss": 0.3217969536781311, "global_step": 34041, "epoch": 382, "lr": 9.62661362489743e-05}
{"train_loss": 0.4051210582256317, "global_step": 34042, "epoch": 382, "lr": 9.626591642104689e-05}
{"train_loss": 0.3238925039768219, "global_step": 34043, "epoch": 382, "lr": 9.626569658689957e-05}
{"train_loss": 0.3584178686141968, "global_step": 34044, "epoch": 382, "lr": 9.626547674653242e-05}
{"train_loss": 0.44969525933265686, "global_step": 34045, "epoch": 382, "lr": 9.626525689994541e-05}
{"train_loss": 0.3446289002895355, "global_step": 34046, "epoch": 382, "lr": 9.626503704713862e-05}
{"train_loss": 0.45252525806427, "global_step": 34047, "epoch": 382, "lr": 9.626481718811205e-05}
{"train_loss": 0.4335208237171173, "global_step": 34048, "epoch": 382, "lr": 9.626459732286574e-05}
{"train_loss": 0.2804543375968933, "global_step": 34049, "epoch": 382, "lr": 9.626437745139972e-05}
{"train_loss": 0.31310805678367615, "global_step": 34050, "epoch": 382, "lr": 9.626415757371403e-05}
{"train_loss": 0.46188756823539734, "global_step": 34051, "epoch": 382, "lr": 9.626393768980866e-05}
{"train_loss": 0.32927459478378296, "global_step": 34052, "epoch": 382, "lr": 9.626371779968369e-05}
{"train_loss": 0.3781907856464386, "global_step": 34053, "epoch": 382, "lr": 9.626349790333913e-05}
{"train_loss": 0.34855538606643677, "global_step": 34054, "epoch": 382, "lr": 9.6263278000775e-05}
{"train_loss": 0.342299222946167, "global_step": 34055, "epoch": 382, "lr": 9.626305809199134e-05}
{"train_loss": 0.34967368841171265, "global_step": 34056, "epoch": 382, "lr": 9.626283817698815e-05}
{"train_loss": 0.42295852303504944, "global_step": 34057, "epoch": 382, "lr": 9.626261825576552e-05}
{"train_loss": 0.4128431975841522, "global_step": 34058, "epoch": 382, "lr": 9.626239832832342e-05}
{"train_loss": 0.3638432025909424, "global_step": 34059, "epoch": 382, "lr": 9.626217839466191e-05}
{"train_loss": 0.33602190017700195, "global_step": 34060, "epoch": 382, "lr": 9.626195845478103e-05}
{"train_loss": 0.3669649362564087, "global_step": 34061, "epoch": 382, "lr": 9.626173850868078e-05}
{"train_loss": 0.47619691491127014, "global_step": 34062, "epoch": 382, "lr": 9.626151855636121e-05}
{"train_loss": 0.2816280722618103, "global_step": 34063, "epoch": 382, "lr": 9.626129859782234e-05}
{"train_loss": 0.4387510418891907, "global_step": 34064, "epoch": 382, "lr": 9.626107863306419e-05}
{"train_loss": 0.4087221920490265, "global_step": 34065, "epoch": 382, "lr": 9.626085866208682e-05}
{"train_loss": 0.3440607786178589, "global_step": 34066, "epoch": 382, "lr": 9.626063868489023e-05}
{"train_loss": 0.35932931303977966, "global_step": 34067, "epoch": 382, "lr": 9.626041870147447e-05}
{"train_loss": 0.2688502073287964, "global_step": 34068, "epoch": 382, "lr": 9.626019871183956e-05}
{"train_loss": 0.4101685881614685, "global_step": 34069, "epoch": 382, "lr": 9.625997871598551e-05}
{"train_loss": 0.34326332807540894, "global_step": 34070, "epoch": 382, "lr": 9.625975871391239e-05}
{"train_loss": 0.2756485641002655, "global_step": 34071, "epoch": 382, "lr": 9.62595387056202e-05}
{"train_loss": 0.3720773160457611, "global_step": 34072, "epoch": 382, "lr": 9.625931869110899e-05}
{"train_loss": 0.4283393323421478, "global_step": 34073, "epoch": 382, "lr": 9.625909867037877e-05}
{"train_loss": 0.30541861057281494, "global_step": 34074, "epoch": 382, "lr": 9.625887864342958e-05}
{"train_loss": 0.33861637115478516, "global_step": 34075, "epoch": 382, "lr": 9.625865861026145e-05}
{"train_loss": 0.33899760246276855, "global_step": 34076, "epoch": 382, "lr": 9.62584385708744e-05}
{"train_loss": 0.3484991192817688, "global_step": 34077, "epoch": 382, "lr": 9.625821852526848e-05}
{"train_loss": 0.3906009793281555, "global_step": 34078, "epoch": 382, "lr": 9.625799847344369e-05}
{"train_loss": 0.32491829991340637, "global_step": 34079, "epoch": 382, "lr": 9.625777841540007e-05}
{"train_loss": 0.3672984838485718, "global_step": 34080, "epoch": 382, "lr": 9.625755835113767e-05}
{"train_loss": 0.36041802167892456, "global_step": 34081, "epoch": 382, "lr": 9.625733828065652e-05}
{"train_loss": 0.29935774207115173, "global_step": 34082, "epoch": 382, "lr": 9.625711820395662e-05}
{"train_loss": 0.3207979202270508, "global_step": 34083, "epoch": 382, "lr": 9.6256898121038e-05}
{"train_loss": 0.5417436361312866, "global_step": 34084, "epoch": 382, "lr": 9.625667803190072e-05}
{"train_loss": 0.4067981243133545, "global_step": 34085, "epoch": 382, "lr": 9.625645793654479e-05}
{"train_loss": 0.3574006622761823, "global_step": 34086, "epoch": 382, "lr": 9.625623783497025e-05, "val_loss": 2.3835341930389404}
{"train_loss": 0.2913792133331299, "global_step": 34087, "epoch": 383, "lr": 9.625601772717712e-05}
{"train_loss": 0.3945475220680237, "global_step": 34088, "epoch": 383, "lr": 9.625579761316541e-05}
{"train_loss": 0.4202522933483124, "global_step": 34089, "epoch": 383, "lr": 9.625557749293519e-05}
{"train_loss": 0.29426801204681396, "global_step": 34090, "epoch": 383, "lr": 9.625535736648648e-05}
{"train_loss": 0.3765556216239929, "global_step": 34091, "epoch": 383, "lr": 9.62551372338193e-05}
{"train_loss": 0.3971368372440338, "global_step": 34092, "epoch": 383, "lr": 9.625491709493366e-05}
{"train_loss": 0.3839416205883026, "global_step": 34093, "epoch": 383, "lr": 9.625469694982963e-05}
{"train_loss": 0.2866936922073364, "global_step": 34094, "epoch": 383, "lr": 9.625447679850721e-05}
{"train_loss": 0.2972540259361267, "global_step": 34095, "epoch": 383, "lr": 9.625425664096643e-05}
{"train_loss": 0.28619199991226196, "global_step": 34096, "epoch": 383, "lr": 9.625403647720734e-05}
{"train_loss": 0.35748526453971863, "global_step": 34097, "epoch": 383, "lr": 9.625381630722997e-05}
{"train_loss": 0.4040907919406891, "global_step": 34098, "epoch": 383, "lr": 9.625359613103434e-05}
{"train_loss": 0.40146318078041077, "global_step": 34099, "epoch": 383, "lr": 9.625337594862045e-05}
{"train_loss": 0.30986693501472473, "global_step": 34100, "epoch": 383, "lr": 9.625315575998836e-05}
{"train_loss": 0.3237930238246918, "global_step": 34101, "epoch": 383, "lr": 9.625293556513812e-05}
{"train_loss": 0.3242015540599823, "global_step": 34102, "epoch": 383, "lr": 9.625271536406972e-05}
{"train_loss": 0.2958393096923828, "global_step": 34103, "epoch": 383, "lr": 9.625249515678322e-05}
{"train_loss": 0.23099036514759064, "global_step": 34104, "epoch": 383, "lr": 9.625227494327863e-05}
{"train_loss": 0.42663514614105225, "global_step": 34105, "epoch": 383, "lr": 9.625205472355598e-05}
{"train_loss": 0.38791412115097046, "global_step": 34106, "epoch": 383, "lr": 9.625183449761531e-05}
{"train_loss": 0.34583723545074463, "global_step": 34107, "epoch": 383, "lr": 9.625161426545664e-05}
{"train_loss": 0.2636550962924957, "global_step": 34108, "epoch": 383, "lr": 9.625139402708e-05}
{"train_loss": 0.3207917809486389, "global_step": 34109, "epoch": 383, "lr": 9.625117378248544e-05}
{"train_loss": 0.3111851215362549, "global_step": 34110, "epoch": 383, "lr": 9.625095353167297e-05}
{"train_loss": 0.3299367427825928, "global_step": 34111, "epoch": 383, "lr": 9.62507332746426e-05}
{"train_loss": 0.42905113101005554, "global_step": 34112, "epoch": 383, "lr": 9.62505130113944e-05}
{"train_loss": 0.3719119131565094, "global_step": 34113, "epoch": 383, "lr": 9.625029274192839e-05}
{"train_loss": 0.370337575674057, "global_step": 34114, "epoch": 383, "lr": 9.625007246624459e-05}
{"train_loss": 0.2929992973804474, "global_step": 34115, "epoch": 383, "lr": 9.624985218434302e-05}
{"train_loss": 0.3828650712966919, "global_step": 34116, "epoch": 383, "lr": 9.624963189622373e-05}
{"train_loss": 0.39042550325393677, "global_step": 34117, "epoch": 383, "lr": 9.624941160188674e-05}
{"train_loss": 0.4646371603012085, "global_step": 34118, "epoch": 383, "lr": 9.624919130133208e-05}
{"train_loss": 0.4027082622051239, "global_step": 34119, "epoch": 383, "lr": 9.624897099455977e-05}
{"train_loss": 0.3092130124568939, "global_step": 34120, "epoch": 383, "lr": 9.624875068156986e-05}
{"train_loss": 0.34542131423950195, "global_step": 34121, "epoch": 383, "lr": 9.624853036236236e-05}
{"train_loss": 0.3093501627445221, "global_step": 34122, "epoch": 383, "lr": 9.624831003693733e-05}
{"train_loss": 0.35399138927459717, "global_step": 34123, "epoch": 383, "lr": 9.624808970529475e-05}
{"train_loss": 0.28262072801589966, "global_step": 34124, "epoch": 383, "lr": 9.624786936743469e-05}
{"train_loss": 0.4394282400608063, "global_step": 34125, "epoch": 383, "lr": 9.624764902335718e-05}
{"train_loss": 0.377137690782547, "global_step": 34126, "epoch": 383, "lr": 9.624742867306222e-05}
{"train_loss": 0.33750173449516296, "global_step": 34127, "epoch": 383, "lr": 9.624720831654986e-05}
{"train_loss": 0.30018943548202515, "global_step": 34128, "epoch": 383, "lr": 9.624698795382014e-05}
{"train_loss": 0.29861047863960266, "global_step": 34129, "epoch": 383, "lr": 9.624676758487305e-05}
{"train_loss": 0.31308433413505554, "global_step": 34130, "epoch": 383, "lr": 9.624654720970867e-05}
{"train_loss": 0.43431225419044495, "global_step": 34131, "epoch": 383, "lr": 9.624632682832699e-05}
{"train_loss": 0.3567809760570526, "global_step": 34132, "epoch": 383, "lr": 9.624610644072807e-05}
{"train_loss": 0.3579842150211334, "global_step": 34133, "epoch": 383, "lr": 9.624588604691192e-05}
{"train_loss": 0.37861162424087524, "global_step": 34134, "epoch": 383, "lr": 9.624566564687857e-05}
{"train_loss": 0.3173710107803345, "global_step": 34135, "epoch": 383, "lr": 9.624544524062806e-05}
{"train_loss": 0.4067939817905426, "global_step": 34136, "epoch": 383, "lr": 9.624522482816041e-05}
{"train_loss": 0.3244953751564026, "global_step": 34137, "epoch": 383, "lr": 9.624500440947563e-05}
{"train_loss": 0.4552667737007141, "global_step": 34138, "epoch": 383, "lr": 9.624478398457382e-05}
{"train_loss": 0.5064030885696411, "global_step": 34139, "epoch": 383, "lr": 9.624456355345493e-05}
{"train_loss": 0.38841545581817627, "global_step": 34140, "epoch": 383, "lr": 9.624434311611904e-05}
{"train_loss": 0.40205925703048706, "global_step": 34141, "epoch": 383, "lr": 9.624412267256612e-05}
{"train_loss": 0.3985680043697357, "global_step": 34142, "epoch": 383, "lr": 9.624390222279629e-05}
{"train_loss": 0.3988453149795532, "global_step": 34143, "epoch": 383, "lr": 9.624368176680952e-05}
{"train_loss": 0.3108502924442291, "global_step": 34144, "epoch": 383, "lr": 9.624346130460584e-05}
{"train_loss": 0.4905257821083069, "global_step": 34145, "epoch": 383, "lr": 9.624324083618528e-05}
{"train_loss": 0.3543008267879486, "global_step": 34146, "epoch": 383, "lr": 9.62430203615479e-05}
{"train_loss": 0.3607526123523712, "global_step": 34147, "epoch": 383, "lr": 9.624279988069371e-05}
{"train_loss": 0.3342026174068451, "global_step": 34148, "epoch": 383, "lr": 9.624257939362274e-05}
{"train_loss": 0.40487486124038696, "global_step": 34149, "epoch": 383, "lr": 9.624235890033499e-05}
{"train_loss": 0.32795169949531555, "global_step": 34150, "epoch": 383, "lr": 9.624213840083054e-05}
{"train_loss": 0.3871023654937744, "global_step": 34151, "epoch": 383, "lr": 9.624191789510941e-05}
{"train_loss": 0.4193041920661926, "global_step": 34152, "epoch": 383, "lr": 9.62416973831716e-05}
{"train_loss": 0.3189195394515991, "global_step": 34153, "epoch": 383, "lr": 9.624147686501716e-05}
{"train_loss": 0.35562801361083984, "global_step": 34154, "epoch": 383, "lr": 9.624125634064613e-05}
{"train_loss": 0.3555820882320404, "global_step": 34155, "epoch": 383, "lr": 9.624103581005851e-05}
{"train_loss": 0.36678168177604675, "global_step": 34156, "epoch": 383, "lr": 9.624081527325434e-05}
{"train_loss": 0.34657227993011475, "global_step": 34157, "epoch": 383, "lr": 9.624059473023368e-05}
{"train_loss": 0.3164246082305908, "global_step": 34158, "epoch": 383, "lr": 9.624037418099653e-05}
{"train_loss": 0.34099486470222473, "global_step": 34159, "epoch": 383, "lr": 9.624015362554291e-05}
{"train_loss": 0.40692663192749023, "global_step": 34160, "epoch": 383, "lr": 9.62399330638729e-05}
{"train_loss": 0.27886882424354553, "global_step": 34161, "epoch": 383, "lr": 9.623971249598644e-05}
{"train_loss": 0.4487122893333435, "global_step": 34162, "epoch": 383, "lr": 9.623949192188366e-05}
{"train_loss": 0.28457319736480713, "global_step": 34163, "epoch": 383, "lr": 9.623927134156452e-05}
{"train_loss": 0.3053825795650482, "global_step": 34164, "epoch": 383, "lr": 9.623905075502908e-05}
{"train_loss": 0.38460081815719604, "global_step": 34165, "epoch": 383, "lr": 9.623883016227738e-05}
{"train_loss": 0.29805824160575867, "global_step": 34166, "epoch": 383, "lr": 9.623860956330941e-05}
{"train_loss": 0.3302348852157593, "global_step": 34167, "epoch": 383, "lr": 9.623838895812523e-05}
{"train_loss": 0.44149699807167053, "global_step": 34168, "epoch": 383, "lr": 9.623816834672486e-05}
{"train_loss": 0.3521808981895447, "global_step": 34169, "epoch": 383, "lr": 9.623794772910832e-05}
{"train_loss": 0.3966158926486969, "global_step": 34170, "epoch": 383, "lr": 9.623772710527568e-05}
{"train_loss": 0.30448445677757263, "global_step": 34171, "epoch": 383, "lr": 9.623750647522692e-05}
{"train_loss": 0.3036532700061798, "global_step": 34172, "epoch": 383, "lr": 9.62372858389621e-05}
{"train_loss": 0.2973597049713135, "global_step": 34173, "epoch": 383, "lr": 9.623706519648124e-05}
{"train_loss": 0.40211960673332214, "global_step": 34174, "epoch": 383, "lr": 9.623684454778437e-05}
{"train_loss": 0.3578634253713522, "global_step": 34175, "epoch": 383, "lr": 9.623662389287152e-05, "val_loss": 2.4643237590789795}
{"train_loss": 0.4494397044181824, "global_step": 34176, "epoch": 384, "lr": 9.62364032317427e-05}
{"train_loss": 0.3664155602455139, "global_step": 34177, "epoch": 384, "lr": 9.6236182564398e-05}
{"train_loss": 0.42122966051101685, "global_step": 34178, "epoch": 384, "lr": 9.623596189083737e-05}
{"train_loss": 0.37722617387771606, "global_step": 34179, "epoch": 384, "lr": 9.623574121106089e-05}
{"train_loss": 0.33423084020614624, "global_step": 34180, "epoch": 384, "lr": 9.62355205250686e-05}
{"train_loss": 0.4946419298648834, "global_step": 34181, "epoch": 384, "lr": 9.623529983286047e-05}
{"train_loss": 0.41701140999794006, "global_step": 34182, "epoch": 384, "lr": 9.62350791344366e-05}
{"train_loss": 0.4745570421218872, "global_step": 34183, "epoch": 384, "lr": 9.623485842979696e-05}
{"train_loss": 0.3969470262527466, "global_step": 34184, "epoch": 384, "lr": 9.623463771894162e-05}
{"train_loss": 0.3473626673221588, "global_step": 34185, "epoch": 384, "lr": 9.62344170018706e-05}
{"train_loss": 0.38050681352615356, "global_step": 34186, "epoch": 384, "lr": 9.623419627858392e-05}
{"train_loss": 0.3455239534378052, "global_step": 34187, "epoch": 384, "lr": 9.623397554908162e-05}
{"train_loss": 0.2917189598083496, "global_step": 34188, "epoch": 384, "lr": 9.623375481336373e-05}
{"train_loss": 0.3448975682258606, "global_step": 34189, "epoch": 384, "lr": 9.623353407143028e-05}
{"train_loss": 0.4722663462162018, "global_step": 34190, "epoch": 384, "lr": 9.623331332328127e-05}
{"train_loss": 0.3878427743911743, "global_step": 34191, "epoch": 384, "lr": 9.623309256891678e-05}
{"train_loss": 0.3292960524559021, "global_step": 34192, "epoch": 384, "lr": 9.62328718083368e-05}
{"train_loss": 0.36469125747680664, "global_step": 34193, "epoch": 384, "lr": 9.623265104154138e-05}
{"train_loss": 0.3497556447982788, "global_step": 34194, "epoch": 384, "lr": 9.623243026853053e-05}
{"train_loss": 0.4236712157726288, "global_step": 34195, "epoch": 384, "lr": 9.623220948930432e-05}
{"train_loss": 0.3687058389186859, "global_step": 34196, "epoch": 384, "lr": 9.623198870386271e-05}
{"train_loss": 0.33006051182746887, "global_step": 34197, "epoch": 384, "lr": 9.623176791220582e-05}
{"train_loss": 0.3550818860530853, "global_step": 34198, "epoch": 384, "lr": 9.623154711433361e-05}
{"train_loss": 0.48998555541038513, "global_step": 34199, "epoch": 384, "lr": 9.623132631024614e-05}
{"train_loss": 0.40095582604408264, "global_step": 34200, "epoch": 384, "lr": 9.623110549994342e-05}
{"train_loss": 0.3592199683189392, "global_step": 34201, "epoch": 384, "lr": 9.62308846834255e-05}
{"train_loss": 0.2896169424057007, "global_step": 34202, "epoch": 384, "lr": 9.623066386069239e-05}
{"train_loss": 0.33350813388824463, "global_step": 34203, "epoch": 384, "lr": 9.623044303174414e-05}
{"train_loss": 0.39350029826164246, "global_step": 34204, "epoch": 384, "lr": 9.623022219658077e-05}
{"train_loss": 0.3712153732776642, "global_step": 34205, "epoch": 384, "lr": 9.62300013552023e-05}
{"train_loss": 0.4164698123931885, "global_step": 34206, "epoch": 384, "lr": 9.622978050760878e-05}
{"train_loss": 0.36395102739334106, "global_step": 34207, "epoch": 384, "lr": 9.622955965380022e-05}
{"train_loss": 0.2826281189918518, "global_step": 34208, "epoch": 384, "lr": 9.622933879377669e-05}
{"train_loss": 0.3574235141277313, "global_step": 34209, "epoch": 384, "lr": 9.622911792753815e-05}
{"train_loss": 0.406154990196228, "global_step": 34210, "epoch": 384, "lr": 9.622889705508469e-05}
{"train_loss": 0.35589665174484253, "global_step": 34211, "epoch": 384, "lr": 9.62286761764163e-05}
{"train_loss": 0.3195987641811371, "global_step": 34212, "epoch": 384, "lr": 9.622845529153306e-05}
{"train_loss": 0.2982521951198578, "global_step": 34213, "epoch": 384, "lr": 9.622823440043495e-05}
{"train_loss": 0.39176473021507263, "global_step": 34214, "epoch": 384, "lr": 9.622801350312202e-05}
{"train_loss": 0.37904155254364014, "global_step": 34215, "epoch": 384, "lr": 9.622779259959429e-05}
{"train_loss": 0.37621524930000305, "global_step": 34216, "epoch": 384, "lr": 9.622757168985179e-05}
{"train_loss": 0.33203914761543274, "global_step": 34217, "epoch": 384, "lr": 9.622735077389458e-05}
{"train_loss": 0.40930071473121643, "global_step": 34218, "epoch": 384, "lr": 9.622712985172263e-05}
{"train_loss": 0.31989777088165283, "global_step": 34219, "epoch": 384, "lr": 9.622690892333605e-05}
{"train_loss": 0.30250823497772217, "global_step": 34220, "epoch": 384, "lr": 9.62266879887348e-05}
{"train_loss": 0.3543519973754883, "global_step": 34221, "epoch": 384, "lr": 9.622646704791893e-05}
{"train_loss": 0.40115299820899963, "global_step": 34222, "epoch": 384, "lr": 9.622624610088848e-05}
{"train_loss": 0.301672101020813, "global_step": 34223, "epoch": 384, "lr": 9.62260251476435e-05}
{"train_loss": 0.38618460297584534, "global_step": 34224, "epoch": 384, "lr": 9.622580418818396e-05}
{"train_loss": 0.37492984533309937, "global_step": 34225, "epoch": 384, "lr": 9.622558322250994e-05}
{"train_loss": 0.32629621028900146, "global_step": 34226, "epoch": 384, "lr": 9.622536225062145e-05}
{"train_loss": 0.2799973785877228, "global_step": 34227, "epoch": 384, "lr": 9.622514127251853e-05}
{"train_loss": 0.3270716667175293, "global_step": 34228, "epoch": 384, "lr": 9.622492028820119e-05}
{"train_loss": 0.4208771586418152, "global_step": 34229, "epoch": 384, "lr": 9.622469929766949e-05}
{"train_loss": 0.23899170756340027, "global_step": 34230, "epoch": 384, "lr": 9.622447830092342e-05}
{"train_loss": 0.3439539074897766, "global_step": 34231, "epoch": 384, "lr": 9.622425729796305e-05}
{"train_loss": 0.31485089659690857, "global_step": 34232, "epoch": 384, "lr": 9.622403628878839e-05}
{"train_loss": 0.31513071060180664, "global_step": 34233, "epoch": 384, "lr": 9.622381527339947e-05}
{"train_loss": 0.366725355386734, "global_step": 34234, "epoch": 384, "lr": 9.622359425179632e-05}
{"train_loss": 0.3519909679889679, "global_step": 34235, "epoch": 384, "lr": 9.622337322397896e-05}
{"train_loss": 0.3489401340484619, "global_step": 34236, "epoch": 384, "lr": 9.622315218994744e-05}
{"train_loss": 0.3044753968715668, "global_step": 34237, "epoch": 384, "lr": 9.62229311497018e-05}
{"train_loss": 0.25158119201660156, "global_step": 34238, "epoch": 384, "lr": 9.622271010324202e-05}
{"train_loss": 0.30531540513038635, "global_step": 34239, "epoch": 384, "lr": 9.622248905056818e-05}
{"train_loss": 0.2510765790939331, "global_step": 34240, "epoch": 384, "lr": 9.622226799168028e-05}
{"train_loss": 0.315080463886261, "global_step": 34241, "epoch": 384, "lr": 9.622204692657837e-05}
{"train_loss": 0.3597213625907898, "global_step": 34242, "epoch": 384, "lr": 9.622182585526244e-05}
{"train_loss": 0.3476276099681854, "global_step": 34243, "epoch": 384, "lr": 9.622160477773259e-05}
{"train_loss": 0.43231475353240967, "global_step": 34244, "epoch": 384, "lr": 9.622138369398878e-05}
{"train_loss": 0.2842840552330017, "global_step": 34245, "epoch": 384, "lr": 9.622116260403108e-05}
{"train_loss": 0.35422903299331665, "global_step": 34246, "epoch": 384, "lr": 9.622094150785949e-05}
{"train_loss": 0.3042329251766205, "global_step": 34247, "epoch": 384, "lr": 9.62207204054741e-05}
{"train_loss": 0.3134844899177551, "global_step": 34248, "epoch": 384, "lr": 9.622049929687487e-05}
{"train_loss": 0.4345874488353729, "global_step": 34249, "epoch": 384, "lr": 9.622027818206186e-05}
{"train_loss": 0.29503265023231506, "global_step": 34250, "epoch": 384, "lr": 9.622005706103508e-05}
{"train_loss": 0.40052351355552673, "global_step": 34251, "epoch": 384, "lr": 9.62198359337946e-05}
{"train_loss": 0.38129347562789917, "global_step": 34252, "epoch": 384, "lr": 9.62196148003404e-05}
{"train_loss": 0.29845961928367615, "global_step": 34253, "epoch": 384, "lr": 9.621939366067256e-05}
{"train_loss": 0.3677080571651459, "global_step": 34254, "epoch": 384, "lr": 9.621917251479109e-05}
{"train_loss": 0.3160629868507385, "global_step": 34255, "epoch": 384, "lr": 9.6218951362696e-05}
{"train_loss": 0.3859044909477234, "global_step": 34256, "epoch": 384, "lr": 9.621873020438735e-05}
{"train_loss": 0.3044854402542114, "global_step": 34257, "epoch": 384, "lr": 9.621850903986513e-05}
{"train_loss": 0.28806692361831665, "global_step": 34258, "epoch": 384, "lr": 9.62182878691294e-05}
{"train_loss": 0.21784694492816925, "global_step": 34259, "epoch": 384, "lr": 9.621806669218021e-05}
{"train_loss": 0.472994327545166, "global_step": 34260, "epoch": 384, "lr": 9.621784550901755e-05}
{"train_loss": 0.27941247820854187, "global_step": 34261, "epoch": 384, "lr": 9.621762431964147e-05}
{"train_loss": 0.33311134576797485, "global_step": 34262, "epoch": 384, "lr": 9.621740312405198e-05}
{"train_loss": 0.3074233829975128, "global_step": 34263, "epoch": 384, "lr": 9.621718192224913e-05}
{"train_loss": 0.35299287468529816, "global_step": 34264, "epoch": 384, "lr": 9.621696071423294e-05, "val_loss": 2.498210906982422}
{"train_loss": 0.3057790994644165, "global_step": 34265, "epoch": 385, "lr": 9.621673950000344e-05}
{"train_loss": 0.36174172163009644, "global_step": 34266, "epoch": 385, "lr": 9.621651827956068e-05}
{"train_loss": 0.40666770935058594, "global_step": 34267, "epoch": 385, "lr": 9.621629705290466e-05}
{"train_loss": 0.3717304766178131, "global_step": 34268, "epoch": 385, "lr": 9.621607582003542e-05}
{"train_loss": 0.2943192422389984, "global_step": 34269, "epoch": 385, "lr": 9.6215854580953e-05}
{"train_loss": 0.21299991011619568, "global_step": 34270, "epoch": 385, "lr": 9.621563333565742e-05}
{"train_loss": 0.2923703193664551, "global_step": 34271, "epoch": 385, "lr": 9.62154120841487e-05}
{"train_loss": 0.2704043686389923, "global_step": 34272, "epoch": 385, "lr": 9.62151908264269e-05}
{"train_loss": 0.28025108575820923, "global_step": 34273, "epoch": 385, "lr": 9.621496956249202e-05}
{"train_loss": 0.4090735614299774, "global_step": 34274, "epoch": 385, "lr": 9.621474829234409e-05}
{"train_loss": 0.341239869594574, "global_step": 34275, "epoch": 385, "lr": 9.621452701598317e-05}
{"train_loss": 0.20279008150100708, "global_step": 34276, "epoch": 385, "lr": 9.621430573340926e-05}
{"train_loss": 0.34477266669273376, "global_step": 34277, "epoch": 385, "lr": 9.621408444462241e-05}
{"train_loss": 0.24842429161071777, "global_step": 34278, "epoch": 385, "lr": 9.621386314962261e-05}
{"train_loss": 0.3052186369895935, "global_step": 34279, "epoch": 385, "lr": 9.621364184840995e-05}
{"train_loss": 0.2869834899902344, "global_step": 34280, "epoch": 385, "lr": 9.621342054098442e-05}
{"train_loss": 0.368289589881897, "global_step": 34281, "epoch": 385, "lr": 9.621319922734607e-05}
{"train_loss": 0.49662792682647705, "global_step": 34282, "epoch": 385, "lr": 9.621297790749491e-05}
{"train_loss": 0.3181364834308624, "global_step": 34283, "epoch": 385, "lr": 9.621275658143096e-05}
{"train_loss": 0.2694318890571594, "global_step": 34284, "epoch": 385, "lr": 9.621253524915429e-05}
{"train_loss": 0.3220205307006836, "global_step": 34285, "epoch": 385, "lr": 9.62123139106649e-05}
{"train_loss": 0.27778157591819763, "global_step": 34286, "epoch": 385, "lr": 9.621209256596284e-05}
{"train_loss": 0.22667883336544037, "global_step": 34287, "epoch": 385, "lr": 9.62118712150481e-05}
{"train_loss": 0.33341896533966064, "global_step": 34288, "epoch": 385, "lr": 9.621164985792076e-05}
{"train_loss": 0.3326796889305115, "global_step": 34289, "epoch": 385, "lr": 9.621142849458082e-05}
{"train_loss": 0.2806979715824127, "global_step": 34290, "epoch": 385, "lr": 9.621120712502831e-05}
{"train_loss": 0.39672309160232544, "global_step": 34291, "epoch": 385, "lr": 9.621098574926328e-05}
{"train_loss": 0.3872729539871216, "global_step": 34292, "epoch": 385, "lr": 9.621076436728575e-05}
{"train_loss": 0.33244210481643677, "global_step": 34293, "epoch": 385, "lr": 9.621054297909573e-05}
{"train_loss": 0.3476755917072296, "global_step": 34294, "epoch": 385, "lr": 9.621032158469327e-05}
{"train_loss": 0.3502535820007324, "global_step": 34295, "epoch": 385, "lr": 9.62101001840784e-05}
{"train_loss": 0.36627426743507385, "global_step": 34296, "epoch": 385, "lr": 9.620987877725114e-05}
{"train_loss": 0.41066762804985046, "global_step": 34297, "epoch": 385, "lr": 9.620965736421153e-05}
{"train_loss": 0.4397791922092438, "global_step": 34298, "epoch": 385, "lr": 9.620943594495959e-05}
{"train_loss": 0.25310882925987244, "global_step": 34299, "epoch": 385, "lr": 9.620921451949536e-05}
{"train_loss": 0.27451130747795105, "global_step": 34300, "epoch": 385, "lr": 9.620899308781886e-05}
{"train_loss": 0.3445985019207001, "global_step": 34301, "epoch": 385, "lr": 9.620877164993012e-05}
{"train_loss": 0.39268073439598083, "global_step": 34302, "epoch": 385, "lr": 9.620855020582918e-05}
{"train_loss": 0.37700432538986206, "global_step": 34303, "epoch": 385, "lr": 9.620832875551607e-05}
{"train_loss": 0.3517930507659912, "global_step": 34304, "epoch": 385, "lr": 9.62081072989908e-05}
{"train_loss": 0.37649810314178467, "global_step": 34305, "epoch": 385, "lr": 9.620788583625343e-05}
{"train_loss": 0.3250705599784851, "global_step": 34306, "epoch": 385, "lr": 9.620766436730395e-05}
{"train_loss": 0.32215696573257446, "global_step": 34307, "epoch": 385, "lr": 9.620744289214243e-05}
{"train_loss": 0.37293651700019836, "global_step": 34308, "epoch": 385, "lr": 9.620722141076888e-05}
{"train_loss": 0.2658085227012634, "global_step": 34309, "epoch": 385, "lr": 9.620699992318332e-05}
{"train_loss": 0.4158765971660614, "global_step": 34310, "epoch": 385, "lr": 9.620677842938582e-05}
{"train_loss": 0.32820576429367065, "global_step": 34311, "epoch": 385, "lr": 9.620655692937635e-05}
{"train_loss": 0.42458924651145935, "global_step": 34312, "epoch": 385, "lr": 9.620633542315498e-05}
{"train_loss": 0.38610339164733887, "global_step": 34313, "epoch": 385, "lr": 9.620611391072174e-05}
{"train_loss": 0.2656957507133484, "global_step": 34314, "epoch": 385, "lr": 9.620589239207665e-05}
{"train_loss": 0.38783249258995056, "global_step": 34315, "epoch": 385, "lr": 9.620567086721975e-05}
{"train_loss": 0.36659422516822815, "global_step": 34316, "epoch": 385, "lr": 9.620544933615103e-05}
{"train_loss": 0.34923455119132996, "global_step": 34317, "epoch": 385, "lr": 9.620522779887058e-05}
{"train_loss": 0.3101442754268646, "global_step": 34318, "epoch": 385, "lr": 9.620500625537838e-05}
{"train_loss": 0.36019963026046753, "global_step": 34319, "epoch": 385, "lr": 9.62047847056745e-05}
{"train_loss": 0.33192604780197144, "global_step": 34320, "epoch": 385, "lr": 9.620456314975894e-05}
{"train_loss": 0.37259310483932495, "global_step": 34321, "epoch": 385, "lr": 9.620434158763175e-05}
{"train_loss": 0.38357633352279663, "global_step": 34322, "epoch": 385, "lr": 9.620412001929293e-05}
{"train_loss": 0.29218587279319763, "global_step": 34323, "epoch": 385, "lr": 9.620389844474254e-05}
{"train_loss": 0.3064354360103607, "global_step": 34324, "epoch": 385, "lr": 9.62036768639806e-05}
{"train_loss": 0.3230228126049042, "global_step": 34325, "epoch": 385, "lr": 9.620345527700713e-05}
{"train_loss": 0.32810288667678833, "global_step": 34326, "epoch": 385, "lr": 9.620323368382218e-05}
{"train_loss": 0.3620479702949524, "global_step": 34327, "epoch": 385, "lr": 9.620301208442576e-05}
{"train_loss": 0.43470966815948486, "global_step": 34328, "epoch": 385, "lr": 9.620279047881791e-05}
{"train_loss": 0.27184540033340454, "global_step": 34329, "epoch": 385, "lr": 9.620256886699866e-05}
{"train_loss": 0.2351989448070526, "global_step": 34330, "epoch": 385, "lr": 9.620234724896804e-05}
{"train_loss": 0.39040857553482056, "global_step": 34331, "epoch": 385, "lr": 9.620212562472607e-05}
{"train_loss": 0.3129599690437317, "global_step": 34332, "epoch": 385, "lr": 9.620190399427279e-05}
{"train_loss": 0.26596614718437195, "global_step": 34333, "epoch": 385, "lr": 9.620168235760824e-05}
{"train_loss": 0.33894580602645874, "global_step": 34334, "epoch": 385, "lr": 9.620146071473242e-05}
{"train_loss": 0.35693085193634033, "global_step": 34335, "epoch": 385, "lr": 9.620123906564538e-05}
{"train_loss": 0.3040831983089447, "global_step": 34336, "epoch": 385, "lr": 9.620101741034715e-05}
{"train_loss": 0.2589985728263855, "global_step": 34337, "epoch": 385, "lr": 9.620079574883776e-05}
{"train_loss": 0.39057421684265137, "global_step": 34338, "epoch": 385, "lr": 9.620057408111722e-05}
{"train_loss": 0.37759020924568176, "global_step": 34339, "epoch": 385, "lr": 9.62003524071856e-05}
{"train_loss": 0.2596258819103241, "global_step": 34340, "epoch": 385, "lr": 9.620013072704289e-05}
{"train_loss": 0.3840591013431549, "global_step": 34341, "epoch": 385, "lr": 9.619990904068914e-05}
{"train_loss": 0.3523540496826172, "global_step": 34342, "epoch": 385, "lr": 9.619968734812437e-05}
{"train_loss": 0.47571009397506714, "global_step": 34343, "epoch": 385, "lr": 9.619946564934862e-05}
{"train_loss": 0.349725604057312, "global_step": 34344, "epoch": 385, "lr": 9.619924394436192e-05}
{"train_loss": 0.35063210129737854, "global_step": 34345, "epoch": 385, "lr": 9.619902223316428e-05}
{"train_loss": 0.31807973980903625, "global_step": 34346, "epoch": 385, "lr": 9.619880051575576e-05}
{"train_loss": 0.26099693775177, "global_step": 34347, "epoch": 385, "lr": 9.619857879213636e-05}
{"train_loss": 0.3262275159358978, "global_step": 34348, "epoch": 385, "lr": 9.619835706230613e-05}
{"train_loss": 0.3081270158290863, "global_step": 34349, "epoch": 385, "lr": 9.61981353262651e-05}
{"train_loss": 0.3182241916656494, "global_step": 34350, "epoch": 385, "lr": 9.619791358401329e-05}
{"train_loss": 0.3118136525154114, "global_step": 34351, "epoch": 385, "lr": 9.619769183555074e-05}
{"train_loss": 0.5451108813285828, "global_step": 34352, "epoch": 385, "lr": 9.619747008087746e-05}
{"train_loss": 0.3368531410278899, "global_step": 34353, "epoch": 385, "lr": 9.61972483199935e-05, "val_loss": 2.5639681816101074, "train_action_mse_error": 27.805877685546875}
{"train_loss": 0.33740127086639404, "global_step": 34354, "epoch": 386, "lr": 9.619702655289888e-05}
{"train_loss": 0.4179675579071045, "global_step": 34355, "epoch": 386, "lr": 9.619680477959365e-05}
{"train_loss": 0.4378706216812134, "global_step": 34356, "epoch": 386, "lr": 9.61965830000778e-05}
{"train_loss": 0.3743158280849457, "global_step": 34357, "epoch": 386, "lr": 9.61963612143514e-05}
{"train_loss": 0.2792584300041199, "global_step": 34358, "epoch": 386, "lr": 9.619613942241445e-05}
{"train_loss": 0.3711531460285187, "global_step": 34359, "epoch": 386, "lr": 9.619591762426699e-05}
{"train_loss": 0.43018969893455505, "global_step": 34360, "epoch": 386, "lr": 9.619569581990907e-05}
{"train_loss": 0.3407566547393799, "global_step": 34361, "epoch": 386, "lr": 9.619547400934069e-05}
{"train_loss": 0.4468456208705902, "global_step": 34362, "epoch": 386, "lr": 9.619525219256187e-05}
{"train_loss": 0.33331894874572754, "global_step": 34363, "epoch": 386, "lr": 9.61950303695727e-05}
{"train_loss": 0.41602739691734314, "global_step": 34364, "epoch": 386, "lr": 9.619480854037314e-05}
{"train_loss": 0.2944521903991699, "global_step": 34365, "epoch": 386, "lr": 9.619458670496327e-05}
{"train_loss": 0.3310682773590088, "global_step": 34366, "epoch": 386, "lr": 9.619436486334309e-05}
{"train_loss": 0.35528475046157837, "global_step": 34367, "epoch": 386, "lr": 9.619414301551265e-05}
{"train_loss": 0.3970445394515991, "global_step": 34368, "epoch": 386, "lr": 9.619392116147196e-05}
{"train_loss": 0.31773290038108826, "global_step": 34369, "epoch": 386, "lr": 9.619369930122106e-05}
{"train_loss": 0.2992088198661804, "global_step": 34370, "epoch": 386, "lr": 9.619347743476e-05}
{"train_loss": 0.39030903577804565, "global_step": 34371, "epoch": 386, "lr": 9.619325556208878e-05}
{"train_loss": 0.3018036484718323, "global_step": 34372, "epoch": 386, "lr": 9.619303368320742e-05}
{"train_loss": 0.2331399917602539, "global_step": 34373, "epoch": 386, "lr": 9.619281179811598e-05}
{"train_loss": 0.32992976903915405, "global_step": 34374, "epoch": 386, "lr": 9.61925899068145e-05}
{"train_loss": 0.34902623295783997, "global_step": 34375, "epoch": 386, "lr": 9.619236800930298e-05}
{"train_loss": 0.39986279606819153, "global_step": 34376, "epoch": 386, "lr": 9.619214610558145e-05}
{"train_loss": 0.34006941318511963, "global_step": 34377, "epoch": 386, "lr": 9.619192419564996e-05}
{"train_loss": 0.23017434775829315, "global_step": 34378, "epoch": 386, "lr": 9.619170227950851e-05}
{"train_loss": 0.280607670545578, "global_step": 34379, "epoch": 386, "lr": 9.619148035715717e-05}
{"train_loss": 0.4948629140853882, "global_step": 34380, "epoch": 386, "lr": 9.619125842859593e-05}
{"train_loss": 0.3824766278266907, "global_step": 34381, "epoch": 386, "lr": 9.619103649382486e-05}
{"train_loss": 0.3408536911010742, "global_step": 34382, "epoch": 386, "lr": 9.619081455284396e-05}
{"train_loss": 0.3843628466129303, "global_step": 34383, "epoch": 386, "lr": 9.619059260565326e-05}
{"train_loss": 0.26571497321128845, "global_step": 34384, "epoch": 386, "lr": 9.61903706522528e-05}
{"train_loss": 0.5299587249755859, "global_step": 34385, "epoch": 386, "lr": 9.61901486926426e-05}
{"train_loss": 0.30365636944770813, "global_step": 34386, "epoch": 386, "lr": 9.61899267268227e-05}
{"train_loss": 0.3589398264884949, "global_step": 34387, "epoch": 386, "lr": 9.618970475479314e-05}
{"train_loss": 0.3122016191482544, "global_step": 34388, "epoch": 386, "lr": 9.618948277655393e-05}
{"train_loss": 0.350867360830307, "global_step": 34389, "epoch": 386, "lr": 9.618926079210511e-05}
{"train_loss": 0.3541184365749359, "global_step": 34390, "epoch": 386, "lr": 9.61890388014467e-05}
{"train_loss": 0.2929421365261078, "global_step": 34391, "epoch": 386, "lr": 9.618881680457873e-05}
{"train_loss": 0.3232507109642029, "global_step": 34392, "epoch": 386, "lr": 9.618859480150126e-05}
{"train_loss": 0.42336875200271606, "global_step": 34393, "epoch": 386, "lr": 9.618837279221429e-05}
{"train_loss": 0.4114888906478882, "global_step": 34394, "epoch": 386, "lr": 9.618815077671785e-05}
{"train_loss": 0.4943120777606964, "global_step": 34395, "epoch": 386, "lr": 9.618792875501196e-05}
{"train_loss": 0.4361245036125183, "global_step": 34396, "epoch": 386, "lr": 9.618770672709669e-05}
{"train_loss": 0.35107702016830444, "global_step": 34397, "epoch": 386, "lr": 9.618748469297204e-05}
{"train_loss": 0.3527756631374359, "global_step": 34398, "epoch": 386, "lr": 9.618726265263805e-05}
{"train_loss": 0.31675487756729126, "global_step": 34399, "epoch": 386, "lr": 9.618704060609473e-05}
{"train_loss": 0.3888586461544037, "global_step": 34400, "epoch": 386, "lr": 9.618681855334215e-05}
{"train_loss": 0.371236652135849, "global_step": 34401, "epoch": 386, "lr": 9.618659649438029e-05}
{"train_loss": 0.30770817399024963, "global_step": 34402, "epoch": 386, "lr": 9.618637442920921e-05}
{"train_loss": 0.30096951127052307, "global_step": 34403, "epoch": 386, "lr": 9.618615235782895e-05}
{"train_loss": 0.4055163264274597, "global_step": 34404, "epoch": 386, "lr": 9.618593028023951e-05}
{"train_loss": 0.3692299425601959, "global_step": 34405, "epoch": 386, "lr": 9.618570819644093e-05}
{"train_loss": 0.42906007170677185, "global_step": 34406, "epoch": 386, "lr": 9.618548610643327e-05}
{"train_loss": 0.3954225182533264, "global_step": 34407, "epoch": 386, "lr": 9.618526401021651e-05}
{"train_loss": 0.30173712968826294, "global_step": 34408, "epoch": 386, "lr": 9.618504190779072e-05}
{"train_loss": 0.3240416646003723, "global_step": 34409, "epoch": 386, "lr": 9.61848197991559e-05}
{"train_loss": 0.33912038803100586, "global_step": 34410, "epoch": 386, "lr": 9.618459768431209e-05}
{"train_loss": 0.23339644074440002, "global_step": 34411, "epoch": 386, "lr": 9.618437556325934e-05}
{"train_loss": 0.3828480541706085, "global_step": 34412, "epoch": 386, "lr": 9.618415343599766e-05}
{"train_loss": 0.3372439444065094, "global_step": 34413, "epoch": 386, "lr": 9.618393130252708e-05}
{"train_loss": 0.2867601811885834, "global_step": 34414, "epoch": 386, "lr": 9.618370916284764e-05}
{"train_loss": 0.28772562742233276, "global_step": 34415, "epoch": 386, "lr": 9.618348701695935e-05}
{"train_loss": 0.3042266070842743, "global_step": 34416, "epoch": 386, "lr": 9.618326486486227e-05}
{"train_loss": 0.3919747471809387, "global_step": 34417, "epoch": 386, "lr": 9.61830427065564e-05}
{"train_loss": 0.44881510734558105, "global_step": 34418, "epoch": 386, "lr": 9.618282054204178e-05}
{"train_loss": 0.32086536288261414, "global_step": 34419, "epoch": 386, "lr": 9.618259837131846e-05}
{"train_loss": 0.45491692423820496, "global_step": 34420, "epoch": 386, "lr": 9.618237619438644e-05}
{"train_loss": 0.30731815099716187, "global_step": 34421, "epoch": 386, "lr": 9.618215401124575e-05}
{"train_loss": 0.3929908275604248, "global_step": 34422, "epoch": 386, "lr": 9.618193182189645e-05}
{"train_loss": 0.4103802442550659, "global_step": 34423, "epoch": 386, "lr": 9.618170962633855e-05}
{"train_loss": 0.3744982182979584, "global_step": 34424, "epoch": 386, "lr": 9.618148742457208e-05}
{"train_loss": 0.38667693734169006, "global_step": 34425, "epoch": 386, "lr": 9.618126521659708e-05}
{"train_loss": 0.3797056972980499, "global_step": 34426, "epoch": 386, "lr": 9.618104300241355e-05}
{"train_loss": 0.30758166313171387, "global_step": 34427, "epoch": 386, "lr": 9.618082078202156e-05}
{"train_loss": 0.3339083194732666, "global_step": 34428, "epoch": 386, "lr": 9.618059855542112e-05}
{"train_loss": 0.389439195394516, "global_step": 34429, "epoch": 386, "lr": 9.618037632261225e-05}
{"train_loss": 0.42825448513031006, "global_step": 34430, "epoch": 386, "lr": 9.618015408359501e-05}
{"train_loss": 0.25886431336402893, "global_step": 34431, "epoch": 386, "lr": 9.61799318383694e-05}
{"train_loss": 0.282885879278183, "global_step": 34432, "epoch": 386, "lr": 9.617970958693546e-05}
{"train_loss": 0.4091884195804596, "global_step": 34433, "epoch": 386, "lr": 9.617948732929323e-05}
{"train_loss": 0.3870915472507477, "global_step": 34434, "epoch": 386, "lr": 9.617926506544273e-05}
{"train_loss": 0.40187710523605347, "global_step": 34435, "epoch": 386, "lr": 9.617904279538396e-05}
{"train_loss": 0.3074263334274292, "global_step": 34436, "epoch": 386, "lr": 9.617882051911702e-05}
{"train_loss": 0.4081372618675232, "global_step": 34437, "epoch": 386, "lr": 9.617859823664188e-05}
{"train_loss": 0.30019089579582214, "global_step": 34438, "epoch": 386, "lr": 9.61783759479586e-05}
{"train_loss": 0.301827996969223, "global_step": 34439, "epoch": 386, "lr": 9.61781536530672e-05}
{"train_loss": 0.3636253774166107, "global_step": 34440, "epoch": 386, "lr": 9.61779313519677e-05}
{"train_loss": 0.23436641693115234, "global_step": 34441, "epoch": 386, "lr": 9.617770904466014e-05}
{"train_loss": 0.35468314253212363, "global_step": 34442, "epoch": 386, "lr": 9.617748673114456e-05, "val_loss": 2.486767292022705}
{"train_loss": 0.367203950881958, "global_step": 34443, "epoch": 387, "lr": 9.617726441142098e-05}
{"train_loss": 0.3002098500728607, "global_step": 34444, "epoch": 387, "lr": 9.617704208548941e-05}
{"train_loss": 0.3102678060531616, "global_step": 34445, "epoch": 387, "lr": 9.617681975334993e-05}
{"train_loss": 0.35382845997810364, "global_step": 34446, "epoch": 387, "lr": 9.617659741500251e-05}
{"train_loss": 0.40587154030799866, "global_step": 34447, "epoch": 387, "lr": 9.617637507044723e-05}
{"train_loss": 0.4123084247112274, "global_step": 34448, "epoch": 387, "lr": 9.61761527196841e-05}
{"train_loss": 0.39049243927001953, "global_step": 34449, "epoch": 387, "lr": 9.617593036271313e-05}
{"train_loss": 0.2617139220237732, "global_step": 34450, "epoch": 387, "lr": 9.617570799953439e-05}
{"train_loss": 0.36390990018844604, "global_step": 34451, "epoch": 387, "lr": 9.617548563014788e-05}
{"train_loss": 0.3364143371582031, "global_step": 34452, "epoch": 387, "lr": 9.617526325455363e-05}
{"train_loss": 0.2877272069454193, "global_step": 34453, "epoch": 387, "lr": 9.617504087275168e-05}
{"train_loss": 0.32927605509757996, "global_step": 34454, "epoch": 387, "lr": 9.617481848474208e-05}
{"train_loss": 0.3822726011276245, "global_step": 34455, "epoch": 387, "lr": 9.617459609052482e-05}
{"train_loss": 0.2673075795173645, "global_step": 34456, "epoch": 387, "lr": 9.617437369009995e-05}
{"train_loss": 0.3750249147415161, "global_step": 34457, "epoch": 387, "lr": 9.61741512834675e-05}
{"train_loss": 0.3215058743953705, "global_step": 34458, "epoch": 387, "lr": 9.617392887062751e-05}
{"train_loss": 0.3983137309551239, "global_step": 34459, "epoch": 387, "lr": 9.617370645157998e-05}
{"train_loss": 0.2731190621852875, "global_step": 34460, "epoch": 387, "lr": 9.617348402632495e-05}
{"train_loss": 0.3836743235588074, "global_step": 34461, "epoch": 387, "lr": 9.617326159486249e-05}
{"train_loss": 0.449022501707077, "global_step": 34462, "epoch": 387, "lr": 9.617303915719257e-05}
{"train_loss": 0.36864781379699707, "global_step": 34463, "epoch": 387, "lr": 9.617281671331526e-05}
{"train_loss": 0.2771550118923187, "global_step": 34464, "epoch": 387, "lr": 9.617259426323057e-05}
{"train_loss": 0.3203968405723572, "global_step": 34465, "epoch": 387, "lr": 9.617237180693855e-05}
{"train_loss": 0.4428941011428833, "global_step": 34466, "epoch": 387, "lr": 9.617214934443921e-05}
{"train_loss": 0.3444781005382538, "global_step": 34467, "epoch": 387, "lr": 9.617192687573259e-05}
{"train_loss": 0.34255096316337585, "global_step": 34468, "epoch": 387, "lr": 9.61717044008187e-05}
{"train_loss": 0.3594004809856415, "global_step": 34469, "epoch": 387, "lr": 9.61714819196976e-05}
{"train_loss": 0.36693933606147766, "global_step": 34470, "epoch": 387, "lr": 9.617125943236932e-05}
{"train_loss": 0.3151967525482178, "global_step": 34471, "epoch": 387, "lr": 9.617103693883387e-05}
{"train_loss": 0.3960272967815399, "global_step": 34472, "epoch": 387, "lr": 9.617081443909128e-05}
{"train_loss": 0.23047570884227753, "global_step": 34473, "epoch": 387, "lr": 9.617059193314159e-05}
{"train_loss": 0.2728355824947357, "global_step": 34474, "epoch": 387, "lr": 9.617036942098482e-05}
{"train_loss": 0.3845185339450836, "global_step": 34475, "epoch": 387, "lr": 9.6170146902621e-05}
{"train_loss": 0.35595789551734924, "global_step": 34476, "epoch": 387, "lr": 9.616992437805019e-05}
{"train_loss": 0.31883475184440613, "global_step": 34477, "epoch": 387, "lr": 9.616970184727239e-05}
{"train_loss": 0.32609832286834717, "global_step": 34478, "epoch": 387, "lr": 9.616947931028762e-05}
{"train_loss": 0.42348822951316833, "global_step": 34479, "epoch": 387, "lr": 9.616925676709594e-05}
{"train_loss": 0.34449678659439087, "global_step": 34480, "epoch": 387, "lr": 9.616903421769737e-05}
{"train_loss": 0.31529879570007324, "global_step": 34481, "epoch": 387, "lr": 9.616881166209193e-05}
{"train_loss": 0.30538132786750793, "global_step": 34482, "epoch": 387, "lr": 9.616858910027966e-05}
{"train_loss": 0.28684157133102417, "global_step": 34483, "epoch": 387, "lr": 9.616836653226058e-05}
{"train_loss": 0.2943011522293091, "global_step": 34484, "epoch": 387, "lr": 9.616814395803473e-05}
{"train_loss": 0.37958791851997375, "global_step": 34485, "epoch": 387, "lr": 9.616792137760213e-05}
{"train_loss": 0.37141746282577515, "global_step": 34486, "epoch": 387, "lr": 9.616769879096281e-05}
{"train_loss": 0.44235530495643616, "global_step": 34487, "epoch": 387, "lr": 9.616747619811683e-05}
{"train_loss": 0.311156302690506, "global_step": 34488, "epoch": 387, "lr": 9.616725359906418e-05}
{"train_loss": 0.25145435333251953, "global_step": 34489, "epoch": 387, "lr": 9.61670309938049e-05}
{"train_loss": 0.33027184009552, "global_step": 34490, "epoch": 387, "lr": 9.616680838233903e-05}
{"train_loss": 0.34431132674217224, "global_step": 34491, "epoch": 387, "lr": 9.616658576466659e-05}
{"train_loss": 0.2627078592777252, "global_step": 34492, "epoch": 387, "lr": 9.616636314078762e-05}
{"train_loss": 0.3902761936187744, "global_step": 34493, "epoch": 387, "lr": 9.616614051070215e-05}
{"train_loss": 0.40054184198379517, "global_step": 34494, "epoch": 387, "lr": 9.616591787441021e-05}
{"train_loss": 0.3187662959098816, "global_step": 34495, "epoch": 387, "lr": 9.61656952319118e-05}
{"train_loss": 0.3438139259815216, "global_step": 34496, "epoch": 387, "lr": 9.616547258320701e-05}
{"train_loss": 0.3674093782901764, "global_step": 34497, "epoch": 387, "lr": 9.61652499282958e-05}
{"train_loss": 0.4018447697162628, "global_step": 34498, "epoch": 387, "lr": 9.616502726717826e-05}
{"train_loss": 0.3148587942123413, "global_step": 34499, "epoch": 387, "lr": 9.616480459985439e-05}
{"train_loss": 0.43275508284568787, "global_step": 34500, "epoch": 387, "lr": 9.616458192632422e-05}
{"train_loss": 0.34007373452186584, "global_step": 34501, "epoch": 387, "lr": 9.616435924658778e-05}
{"train_loss": 0.33698412775993347, "global_step": 34502, "epoch": 387, "lr": 9.616413656064511e-05}
{"train_loss": 0.2751222848892212, "global_step": 34503, "epoch": 387, "lr": 9.616391386849624e-05}
{"train_loss": 0.27469855546951294, "global_step": 34504, "epoch": 387, "lr": 9.616369117014119e-05}
{"train_loss": 0.3156682252883911, "global_step": 34505, "epoch": 387, "lr": 9.616346846558e-05}
{"train_loss": 0.30888622999191284, "global_step": 34506, "epoch": 387, "lr": 9.616324575481268e-05}
{"train_loss": 0.3434589207172394, "global_step": 34507, "epoch": 387, "lr": 9.616302303783929e-05}
{"train_loss": 0.49225467443466187, "global_step": 34508, "epoch": 387, "lr": 9.616280031465983e-05}
{"train_loss": 0.3182862401008606, "global_step": 34509, "epoch": 387, "lr": 9.616257758527435e-05}
{"train_loss": 0.29645827412605286, "global_step": 34510, "epoch": 387, "lr": 9.616235484968287e-05}
{"train_loss": 0.329169362783432, "global_step": 34511, "epoch": 387, "lr": 9.616213210788543e-05}
{"train_loss": 0.27899885177612305, "global_step": 34512, "epoch": 387, "lr": 9.616190935988205e-05}
{"train_loss": 0.3600151538848877, "global_step": 34513, "epoch": 387, "lr": 9.616168660567278e-05}
{"train_loss": 0.2882627248764038, "global_step": 34514, "epoch": 387, "lr": 9.616146384525762e-05}
{"train_loss": 0.3329143226146698, "global_step": 34515, "epoch": 387, "lr": 9.61612410786366e-05}
{"train_loss": 0.286379337310791, "global_step": 34516, "epoch": 387, "lr": 9.616101830580979e-05}
{"train_loss": 0.3251737952232361, "global_step": 34517, "epoch": 387, "lr": 9.616079552677717e-05}
{"train_loss": 0.36471930146217346, "global_step": 34518, "epoch": 387, "lr": 9.616057274153881e-05}
{"train_loss": 0.3223956227302551, "global_step": 34519, "epoch": 387, "lr": 9.616034995009471e-05}
{"train_loss": 0.2364441305398941, "global_step": 34520, "epoch": 387, "lr": 9.616012715244493e-05}
{"train_loss": 0.2872435748577118, "global_step": 34521, "epoch": 387, "lr": 9.615990434858948e-05}
{"train_loss": 0.2642931640148163, "global_step": 34522, "epoch": 387, "lr": 9.615968153852838e-05}
{"train_loss": 0.22848482429981232, "global_step": 34523, "epoch": 387, "lr": 9.615945872226169e-05}
{"train_loss": 0.26163652539253235, "global_step": 34524, "epoch": 387, "lr": 9.61592358997894e-05}
{"train_loss": 0.2616652548313141, "global_step": 34525, "epoch": 387, "lr": 9.615901307111159e-05}
{"train_loss": 0.25529932975769043, "global_step": 34526, "epoch": 387, "lr": 9.615879023622826e-05}
{"train_loss": 0.3595593273639679, "global_step": 34527, "epoch": 387, "lr": 9.615856739513944e-05}
{"train_loss": 0.2600979208946228, "global_step": 34528, "epoch": 387, "lr": 9.615834454784515e-05}
{"train_loss": 0.36152762174606323, "global_step": 34529, "epoch": 387, "lr": 9.615812169434545e-05}
{"train_loss": 0.31407079100608826, "global_step": 34530, "epoch": 387, "lr": 9.615789883464035e-05}
{"train_loss": 0.3327395984630906, "global_step": 34531, "epoch": 387, "lr": 9.615767596872986e-05, "val_loss": 2.575836420059204}
{"train_loss": 0.28331413865089417, "global_step": 34532, "epoch": 388, "lr": 9.615745309661406e-05}
{"train_loss": 0.3062993288040161, "global_step": 34533, "epoch": 388, "lr": 9.615723021829295e-05}
{"train_loss": 0.35645943880081177, "global_step": 34534, "epoch": 388, "lr": 9.615700733376656e-05}
{"train_loss": 0.37126895785331726, "global_step": 34535, "epoch": 388, "lr": 9.615678444303491e-05}
{"train_loss": 0.4184596836566925, "global_step": 34536, "epoch": 388, "lr": 9.615656154609807e-05}
{"train_loss": 0.42607975006103516, "global_step": 34537, "epoch": 388, "lr": 9.615633864295601e-05}
{"train_loss": 0.3970458209514618, "global_step": 34538, "epoch": 388, "lr": 9.61561157336088e-05}
{"train_loss": 0.2613518536090851, "global_step": 34539, "epoch": 388, "lr": 9.615589281805648e-05}
{"train_loss": 0.28174179792404175, "global_step": 34540, "epoch": 388, "lr": 9.615566989629907e-05}
{"train_loss": 0.3065873980522156, "global_step": 34541, "epoch": 388, "lr": 9.615544696833657e-05}
{"train_loss": 0.40678080916404724, "global_step": 34542, "epoch": 388, "lr": 9.615522403416903e-05}
{"train_loss": 0.30332237482070923, "global_step": 34543, "epoch": 388, "lr": 9.61550010937965e-05}
{"train_loss": 0.3429356813430786, "global_step": 34544, "epoch": 388, "lr": 9.615477814721898e-05}
{"train_loss": 0.32912909984588623, "global_step": 34545, "epoch": 388, "lr": 9.615455519443653e-05}
{"train_loss": 0.36945247650146484, "global_step": 34546, "epoch": 388, "lr": 9.615433223544914e-05}
{"train_loss": 0.3617870509624481, "global_step": 34547, "epoch": 388, "lr": 9.615410927025688e-05}
{"train_loss": 0.35968464612960815, "global_step": 34548, "epoch": 388, "lr": 9.615388629885976e-05}
{"train_loss": 0.36935487389564514, "global_step": 34549, "epoch": 388, "lr": 9.615366332125781e-05}
{"train_loss": 0.2525146007537842, "global_step": 34550, "epoch": 388, "lr": 9.615344033745106e-05}
{"train_loss": 0.3199884295463562, "global_step": 34551, "epoch": 388, "lr": 9.615321734743954e-05}
{"train_loss": 0.401122123003006, "global_step": 34552, "epoch": 388, "lr": 9.615299435122329e-05}
{"train_loss": 0.36598801612854004, "global_step": 34553, "epoch": 388, "lr": 9.615277134880232e-05}
{"train_loss": 0.3403250575065613, "global_step": 34554, "epoch": 388, "lr": 9.615254834017669e-05}
{"train_loss": 0.335026353597641, "global_step": 34555, "epoch": 388, "lr": 9.61523253253464e-05}
{"train_loss": 0.4151613116264343, "global_step": 34556, "epoch": 388, "lr": 9.615210230431149e-05}
{"train_loss": 0.37363201379776, "global_step": 34557, "epoch": 388, "lr": 9.6151879277072e-05}
{"train_loss": 0.3051293194293976, "global_step": 34558, "epoch": 388, "lr": 9.615165624362796e-05}
{"train_loss": 0.24693337082862854, "global_step": 34559, "epoch": 388, "lr": 9.615143320397937e-05}
{"train_loss": 0.3999621272087097, "global_step": 34560, "epoch": 388, "lr": 9.615121015812629e-05}
{"train_loss": 0.5066453814506531, "global_step": 34561, "epoch": 388, "lr": 9.615098710606875e-05}
{"train_loss": 0.35413920879364014, "global_step": 34562, "epoch": 388, "lr": 9.615076404780679e-05}
{"train_loss": 0.3588959872722626, "global_step": 34563, "epoch": 388, "lr": 9.615054098334039e-05}
{"train_loss": 0.42734429240226746, "global_step": 34564, "epoch": 388, "lr": 9.615031791266962e-05}
{"train_loss": 0.27937713265419006, "global_step": 34565, "epoch": 388, "lr": 9.615009483579451e-05}
{"train_loss": 0.31780463457107544, "global_step": 34566, "epoch": 388, "lr": 9.614987175271508e-05}
{"train_loss": 0.3381679356098175, "global_step": 34567, "epoch": 388, "lr": 9.614964866343136e-05}
{"train_loss": 0.3360089957714081, "global_step": 34568, "epoch": 388, "lr": 9.614942556794337e-05}
{"train_loss": 0.4474433958530426, "global_step": 34569, "epoch": 388, "lr": 9.614920246625118e-05}
{"train_loss": 0.3787573277950287, "global_step": 34570, "epoch": 388, "lr": 9.614897935835478e-05}
{"train_loss": 0.30929574370384216, "global_step": 34571, "epoch": 388, "lr": 9.614875624425422e-05}
{"train_loss": 0.4082605242729187, "global_step": 34572, "epoch": 388, "lr": 9.614853312394949e-05}
{"train_loss": 0.20071949064731598, "global_step": 34573, "epoch": 388, "lr": 9.614830999744068e-05}
{"train_loss": 0.4045266807079315, "global_step": 34574, "epoch": 388, "lr": 9.614808686472779e-05}
{"train_loss": 0.3905860483646393, "global_step": 34575, "epoch": 388, "lr": 9.614786372581084e-05}
{"train_loss": 0.4831722676753998, "global_step": 34576, "epoch": 388, "lr": 9.614764058068988e-05}
{"train_loss": 0.3886681795120239, "global_step": 34577, "epoch": 388, "lr": 9.614741742936494e-05}
{"train_loss": 0.35316330194473267, "global_step": 34578, "epoch": 388, "lr": 9.614719427183603e-05}
{"train_loss": 0.3719136416912079, "global_step": 34579, "epoch": 388, "lr": 9.61469711081032e-05}
{"train_loss": 0.4126100540161133, "global_step": 34580, "epoch": 388, "lr": 9.614674793816648e-05}
{"train_loss": 0.4267180860042572, "global_step": 34581, "epoch": 388, "lr": 9.614652476202586e-05}
{"train_loss": 0.3136698305606842, "global_step": 34582, "epoch": 388, "lr": 9.614630157968143e-05}
{"train_loss": 0.36059871315956116, "global_step": 34583, "epoch": 388, "lr": 9.614607839113318e-05}
{"train_loss": 0.4267400801181793, "global_step": 34584, "epoch": 388, "lr": 9.614585519638117e-05}
{"train_loss": 0.31231001019477844, "global_step": 34585, "epoch": 388, "lr": 9.61456319954254e-05}
{"train_loss": 0.30960729718208313, "global_step": 34586, "epoch": 388, "lr": 9.614540878826591e-05}
{"train_loss": 0.3549094796180725, "global_step": 34587, "epoch": 388, "lr": 9.614518557490273e-05}
{"train_loss": 0.3325293958187103, "global_step": 34588, "epoch": 388, "lr": 9.61449623553359e-05}
{"train_loss": 0.42503178119659424, "global_step": 34589, "epoch": 388, "lr": 9.614473912956545e-05}
{"train_loss": 0.2690843939781189, "global_step": 34590, "epoch": 388, "lr": 9.614451589759139e-05}
{"train_loss": 0.4741164743900299, "global_step": 34591, "epoch": 388, "lr": 9.614429265941376e-05}
{"train_loss": 0.36826738715171814, "global_step": 34592, "epoch": 388, "lr": 9.614406941503261e-05}
{"train_loss": 0.2739477753639221, "global_step": 34593, "epoch": 388, "lr": 9.614384616444792e-05}
{"train_loss": 0.29046961665153503, "global_step": 34594, "epoch": 388, "lr": 9.614362290765977e-05}
{"train_loss": 0.30957573652267456, "global_step": 34595, "epoch": 388, "lr": 9.614339964466818e-05}
{"train_loss": 0.3278394639492035, "global_step": 34596, "epoch": 388, "lr": 9.614317637547316e-05}
{"train_loss": 0.3990742564201355, "global_step": 34597, "epoch": 388, "lr": 9.614295310007475e-05}
{"train_loss": 0.33995360136032104, "global_step": 34598, "epoch": 388, "lr": 9.6142729818473e-05}
{"train_loss": 0.4247385263442993, "global_step": 34599, "epoch": 388, "lr": 9.61425065306679e-05}
{"train_loss": 0.39994290471076965, "global_step": 34600, "epoch": 388, "lr": 9.614228323665952e-05}
{"train_loss": 0.35152411460876465, "global_step": 34601, "epoch": 388, "lr": 9.614205993644787e-05}
{"train_loss": 0.3177717626094818, "global_step": 34602, "epoch": 388, "lr": 9.614183663003298e-05}
{"train_loss": 0.2596585154533386, "global_step": 34603, "epoch": 388, "lr": 9.61416133174149e-05}
{"train_loss": 0.27027347683906555, "global_step": 34604, "epoch": 388, "lr": 9.614138999859361e-05}
{"train_loss": 0.31641095876693726, "global_step": 34605, "epoch": 388, "lr": 9.614116667356918e-05}
{"train_loss": 0.353103905916214, "global_step": 34606, "epoch": 388, "lr": 9.614094334234164e-05}
{"train_loss": 0.3243379294872284, "global_step": 34607, "epoch": 388, "lr": 9.614072000491101e-05}
{"train_loss": 0.43515849113464355, "global_step": 34608, "epoch": 388, "lr": 9.614049666127733e-05}
{"train_loss": 0.34056028723716736, "global_step": 34609, "epoch": 388, "lr": 9.614027331144062e-05}
{"train_loss": 0.44989949464797974, "global_step": 34610, "epoch": 388, "lr": 9.614004995540091e-05}
{"train_loss": 0.4105432331562042, "global_step": 34611, "epoch": 388, "lr": 9.613982659315823e-05}
{"train_loss": 0.3767189681529999, "global_step": 34612, "epoch": 388, "lr": 9.613960322471262e-05}
{"train_loss": 0.32061612606048584, "global_step": 34613, "epoch": 388, "lr": 9.613937985006409e-05}
{"train_loss": 0.23840521275997162, "global_step": 34614, "epoch": 388, "lr": 9.61391564692127e-05}
{"train_loss": 0.34092578291893005, "global_step": 34615, "epoch": 388, "lr": 9.613893308215845e-05}
{"train_loss": 0.38648271560668945, "global_step": 34616, "epoch": 388, "lr": 9.613870968890139e-05}
{"train_loss": 0.274106502532959, "global_step": 34617, "epoch": 388, "lr": 9.613848628944153e-05}
{"train_loss": 0.4834440350532532, "global_step": 34618, "epoch": 388, "lr": 9.613826288377892e-05}
{"train_loss": 0.30065175890922546, "global_step": 34619, "epoch": 388, "lr": 9.613803947191359e-05}
{"train_loss": 0.35320745793621194, "global_step": 34620, "epoch": 388, "lr": 9.613781605384556e-05, "val_loss": 2.3988430500030518}
{"train_loss": 0.30802345275878906, "global_step": 34621, "epoch": 389, "lr": 9.613759262957485e-05}
{"train_loss": 0.33388757705688477, "global_step": 34622, "epoch": 389, "lr": 9.613736919910153e-05}
{"train_loss": 0.4420801103115082, "global_step": 34623, "epoch": 389, "lr": 9.613714576242559e-05}
{"train_loss": 0.29389506578445435, "global_step": 34624, "epoch": 389, "lr": 9.613692231954705e-05}
{"train_loss": 0.4561845660209656, "global_step": 34625, "epoch": 389, "lr": 9.613669887046599e-05}
{"train_loss": 0.3383791446685791, "global_step": 34626, "epoch": 389, "lr": 9.61364754151824e-05}
{"train_loss": 0.3400481939315796, "global_step": 34627, "epoch": 389, "lr": 9.613625195369633e-05}
{"train_loss": 0.33694735169410706, "global_step": 34628, "epoch": 389, "lr": 9.61360284860078e-05}
{"train_loss": 0.3574064373970032, "global_step": 34629, "epoch": 389, "lr": 9.613580501211685e-05}
{"train_loss": 0.3289759159088135, "global_step": 34630, "epoch": 389, "lr": 9.61355815320235e-05}
{"train_loss": 0.28617316484451294, "global_step": 34631, "epoch": 389, "lr": 9.613535804572779e-05}
{"train_loss": 0.3965764045715332, "global_step": 34632, "epoch": 389, "lr": 9.613513455322974e-05}
{"train_loss": 0.3549303710460663, "global_step": 34633, "epoch": 389, "lr": 9.613491105452937e-05}
{"train_loss": 0.3818065822124481, "global_step": 34634, "epoch": 389, "lr": 9.613468754962673e-05}
{"train_loss": 0.2537100613117218, "global_step": 34635, "epoch": 389, "lr": 9.613446403852185e-05}
{"train_loss": 0.35176289081573486, "global_step": 34636, "epoch": 389, "lr": 9.613424052121476e-05}
{"train_loss": 0.2983952760696411, "global_step": 34637, "epoch": 389, "lr": 9.613401699770547e-05}
{"train_loss": 0.4235476553440094, "global_step": 34638, "epoch": 389, "lr": 9.613379346799402e-05}
{"train_loss": 0.45336681604385376, "global_step": 34639, "epoch": 389, "lr": 9.613356993208046e-05}
{"train_loss": 0.36784303188323975, "global_step": 34640, "epoch": 389, "lr": 9.613334638996479e-05}
{"train_loss": 0.28583061695098877, "global_step": 34641, "epoch": 389, "lr": 9.613312284164706e-05}
{"train_loss": 0.38892579078674316, "global_step": 34642, "epoch": 389, "lr": 9.61328992871273e-05}
{"train_loss": 0.402171790599823, "global_step": 34643, "epoch": 389, "lr": 9.613267572640552e-05}
{"train_loss": 0.34166717529296875, "global_step": 34644, "epoch": 389, "lr": 9.613245215948177e-05}
{"train_loss": 0.4465257525444031, "global_step": 34645, "epoch": 389, "lr": 9.613222858635608e-05}
{"train_loss": 0.31369078159332275, "global_step": 34646, "epoch": 389, "lr": 9.613200500702846e-05}
{"train_loss": 0.46894606947898865, "global_step": 34647, "epoch": 389, "lr": 9.613178142149896e-05}
{"train_loss": 0.353516548871994, "global_step": 34648, "epoch": 389, "lr": 9.613155782976761e-05}
{"train_loss": 0.27662476897239685, "global_step": 34649, "epoch": 389, "lr": 9.613133423183443e-05}
{"train_loss": 0.2937631905078888, "global_step": 34650, "epoch": 389, "lr": 9.613111062769946e-05}
{"train_loss": 0.4470067024230957, "global_step": 34651, "epoch": 389, "lr": 9.613088701736272e-05}
{"train_loss": 0.2600933313369751, "global_step": 34652, "epoch": 389, "lr": 9.613066340082425e-05}
{"train_loss": 0.389538049697876, "global_step": 34653, "epoch": 389, "lr": 9.613043977808407e-05}
{"train_loss": 0.41345274448394775, "global_step": 34654, "epoch": 389, "lr": 9.613021614914222e-05}
{"train_loss": 0.28716397285461426, "global_step": 34655, "epoch": 389, "lr": 9.612999251399871e-05}
{"train_loss": 0.3417450189590454, "global_step": 34656, "epoch": 389, "lr": 9.612976887265359e-05}
{"train_loss": 0.43485042452812195, "global_step": 34657, "epoch": 389, "lr": 9.612954522510688e-05}
{"train_loss": 0.33313849568367004, "global_step": 34658, "epoch": 389, "lr": 9.612932157135863e-05}
{"train_loss": 0.44528260827064514, "global_step": 34659, "epoch": 389, "lr": 9.612909791140885e-05}
{"train_loss": 0.3249327540397644, "global_step": 34660, "epoch": 389, "lr": 9.612887424525758e-05}
{"train_loss": 0.32779353857040405, "global_step": 34661, "epoch": 389, "lr": 9.612865057290483e-05}
{"train_loss": 0.3245554268360138, "global_step": 34662, "epoch": 389, "lr": 9.612842689435066e-05}
{"train_loss": 0.3484371602535248, "global_step": 34663, "epoch": 389, "lr": 9.612820320959508e-05}
{"train_loss": 0.3770555853843689, "global_step": 34664, "epoch": 389, "lr": 9.612797951863813e-05}
{"train_loss": 0.3634266257286072, "global_step": 34665, "epoch": 389, "lr": 9.612775582147982e-05}
{"train_loss": 0.33720940351486206, "global_step": 34666, "epoch": 389, "lr": 9.61275321181202e-05}
{"train_loss": 0.3081899881362915, "global_step": 34667, "epoch": 389, "lr": 9.612730840855931e-05}
{"train_loss": 0.308580607175827, "global_step": 34668, "epoch": 389, "lr": 9.612708469279715e-05}
{"train_loss": 0.3615694046020508, "global_step": 34669, "epoch": 389, "lr": 9.612686097083378e-05}
{"train_loss": 0.31868138909339905, "global_step": 34670, "epoch": 389, "lr": 9.61266372426692e-05}
{"train_loss": 0.3873297870159149, "global_step": 34671, "epoch": 389, "lr": 9.612641350830347e-05}
{"train_loss": 0.395622193813324, "global_step": 34672, "epoch": 389, "lr": 9.61261897677366e-05}
{"train_loss": 0.32483407855033875, "global_step": 34673, "epoch": 389, "lr": 9.612596602096861e-05}
{"train_loss": 0.326517790555954, "global_step": 34674, "epoch": 389, "lr": 9.612574226799958e-05}
{"train_loss": 0.28343072533607483, "global_step": 34675, "epoch": 389, "lr": 9.612551850882948e-05}
{"train_loss": 0.38396769762039185, "global_step": 34676, "epoch": 389, "lr": 9.612529474345838e-05}
{"train_loss": 0.28260037302970886, "global_step": 34677, "epoch": 389, "lr": 9.612507097188629e-05}
{"train_loss": 0.286714106798172, "global_step": 34678, "epoch": 389, "lr": 9.612484719411325e-05}
{"train_loss": 0.2788887917995453, "global_step": 34679, "epoch": 389, "lr": 9.612462341013928e-05}
{"train_loss": 0.31278109550476074, "global_step": 34680, "epoch": 389, "lr": 9.612439961996443e-05}
{"train_loss": 0.3514999449253082, "global_step": 34681, "epoch": 389, "lr": 9.61241758235887e-05}
{"train_loss": 0.42818090319633484, "global_step": 34682, "epoch": 389, "lr": 9.612395202101214e-05}
{"train_loss": 0.2857324779033661, "global_step": 34683, "epoch": 389, "lr": 9.612372821223478e-05}
{"train_loss": 0.36133238673210144, "global_step": 34684, "epoch": 389, "lr": 9.612350439725664e-05}
{"train_loss": 0.4273509085178375, "global_step": 34685, "epoch": 389, "lr": 9.612328057607777e-05}
{"train_loss": 0.4193846583366394, "global_step": 34686, "epoch": 389, "lr": 9.612305674869818e-05}
{"train_loss": 0.3309686779975891, "global_step": 34687, "epoch": 389, "lr": 9.61228329151179e-05}
{"train_loss": 0.3375004827976227, "global_step": 34688, "epoch": 389, "lr": 9.612260907533697e-05}
{"train_loss": 0.2192317396402359, "global_step": 34689, "epoch": 389, "lr": 9.612238522935544e-05}
{"train_loss": 0.4051573872566223, "global_step": 34690, "epoch": 389, "lr": 9.612216137717329e-05}
{"train_loss": 0.3371672034263611, "global_step": 34691, "epoch": 389, "lr": 9.61219375187906e-05}
{"train_loss": 0.2570907473564148, "global_step": 34692, "epoch": 389, "lr": 9.612171365420735e-05}
{"train_loss": 0.3814290165901184, "global_step": 34693, "epoch": 389, "lr": 9.612148978342361e-05}
{"train_loss": 0.4072488248348236, "global_step": 34694, "epoch": 389, "lr": 9.61212659064394e-05}
{"train_loss": 0.2807130813598633, "global_step": 34695, "epoch": 389, "lr": 9.612104202325476e-05}
{"train_loss": 0.3992063105106354, "global_step": 34696, "epoch": 389, "lr": 9.61208181338697e-05}
{"train_loss": 0.2618660628795624, "global_step": 34697, "epoch": 389, "lr": 9.612059423828425e-05}
{"train_loss": 0.4023827016353607, "global_step": 34698, "epoch": 389, "lr": 9.612037033649845e-05}
{"train_loss": 0.3423271179199219, "global_step": 34699, "epoch": 389, "lr": 9.612014642851232e-05}
{"train_loss": 0.3087208867073059, "global_step": 34700, "epoch": 389, "lr": 9.611992251432592e-05}
{"train_loss": 0.4723842144012451, "global_step": 34701, "epoch": 389, "lr": 9.611969859393925e-05}
{"train_loss": 0.23021461069583893, "global_step": 34702, "epoch": 389, "lr": 9.611947466735235e-05}
{"train_loss": 0.3631124496459961, "global_step": 34703, "epoch": 389, "lr": 9.611925073456525e-05}
{"train_loss": 0.36612027883529663, "global_step": 34704, "epoch": 389, "lr": 9.611902679557797e-05}
{"train_loss": 0.2930102050304413, "global_step": 34705, "epoch": 389, "lr": 9.611880285039056e-05}
{"train_loss": 0.3447551727294922, "global_step": 34706, "epoch": 389, "lr": 9.611857889900302e-05}
{"train_loss": 0.4313139319419861, "global_step": 34707, "epoch": 389, "lr": 9.611835494141541e-05}
{"train_loss": 0.3708173930644989, "global_step": 34708, "epoch": 389, "lr": 9.611813097762777e-05}
{"train_loss": 0.3502445200855812, "global_step": 34709, "epoch": 389, "lr": 9.611790700764007e-05, "val_loss": 2.5245089530944824}
{"train_loss": 0.3647383749485016, "global_step": 34710, "epoch": 390, "lr": 9.611768303145242e-05}
{"train_loss": 0.37006762623786926, "global_step": 34711, "epoch": 390, "lr": 9.611745904906477e-05}
{"train_loss": 0.39417919516563416, "global_step": 34712, "epoch": 390, "lr": 9.611723506047721e-05}
{"train_loss": 0.2557283341884613, "global_step": 34713, "epoch": 390, "lr": 9.611701106568976e-05}
{"train_loss": 0.2962063252925873, "global_step": 34714, "epoch": 390, "lr": 9.611678706470243e-05}
{"train_loss": 0.28338178992271423, "global_step": 34715, "epoch": 390, "lr": 9.611656305751526e-05}
{"train_loss": 0.32336726784706116, "global_step": 34716, "epoch": 390, "lr": 9.611633904412828e-05}
{"train_loss": 0.2689965069293976, "global_step": 34717, "epoch": 390, "lr": 9.611611502454153e-05}
{"train_loss": 0.33259880542755127, "global_step": 34718, "epoch": 390, "lr": 9.6115890998755e-05}
{"train_loss": 0.34153446555137634, "global_step": 34719, "epoch": 390, "lr": 9.611566696676877e-05}
{"train_loss": 0.4158787727355957, "global_step": 34720, "epoch": 390, "lr": 9.611544292858286e-05}
{"train_loss": 0.37336960434913635, "global_step": 34721, "epoch": 390, "lr": 9.611521888419728e-05}
{"train_loss": 0.27947890758514404, "global_step": 34722, "epoch": 390, "lr": 9.611499483361208e-05}
{"train_loss": 0.3259161114692688, "global_step": 34723, "epoch": 390, "lr": 9.611477077682727e-05}
{"train_loss": 0.34200727939605713, "global_step": 34724, "epoch": 390, "lr": 9.61145467138429e-05}
{"train_loss": 0.33330366015434265, "global_step": 34725, "epoch": 390, "lr": 9.611432264465898e-05}
{"train_loss": 0.3434674143791199, "global_step": 34726, "epoch": 390, "lr": 9.611409856927555e-05}
{"train_loss": 0.3199370503425598, "global_step": 34727, "epoch": 390, "lr": 9.611387448769266e-05}
{"train_loss": 0.2770605683326721, "global_step": 34728, "epoch": 390, "lr": 9.61136503999103e-05}
{"train_loss": 0.3436075747013092, "global_step": 34729, "epoch": 390, "lr": 9.611342630592854e-05}
{"train_loss": 0.4027445614337921, "global_step": 34730, "epoch": 390, "lr": 9.611320220574738e-05}
{"train_loss": 0.3063136339187622, "global_step": 34731, "epoch": 390, "lr": 9.611297809936687e-05}
{"train_loss": 0.2928016781806946, "global_step": 34732, "epoch": 390, "lr": 9.611275398678703e-05}
{"train_loss": 0.28778839111328125, "global_step": 34733, "epoch": 390, "lr": 9.61125298680079e-05}
{"train_loss": 0.3971725106239319, "global_step": 34734, "epoch": 390, "lr": 9.611230574302949e-05}
{"train_loss": 0.2712516486644745, "global_step": 34735, "epoch": 390, "lr": 9.611208161185183e-05}
{"train_loss": 0.39161595702171326, "global_step": 34736, "epoch": 390, "lr": 9.611185747447498e-05}
{"train_loss": 0.40844133496284485, "global_step": 34737, "epoch": 390, "lr": 9.611163333089897e-05}
{"train_loss": 0.32024362683296204, "global_step": 34738, "epoch": 390, "lr": 9.611140918112377e-05}
{"train_loss": 0.344708114862442, "global_step": 34739, "epoch": 390, "lr": 9.611118502514949e-05}
{"train_loss": 0.28701460361480713, "global_step": 34740, "epoch": 390, "lr": 9.611096086297611e-05}
{"train_loss": 0.4371546804904938, "global_step": 34741, "epoch": 390, "lr": 9.611073669460368e-05}
{"train_loss": 0.38930439949035645, "global_step": 34742, "epoch": 390, "lr": 9.611051252003222e-05}
{"train_loss": 0.28393977880477905, "global_step": 34743, "epoch": 390, "lr": 9.611028833926174e-05}
{"train_loss": 0.2315162569284439, "global_step": 34744, "epoch": 390, "lr": 9.611006415229234e-05}
{"train_loss": 0.404207319021225, "global_step": 34745, "epoch": 390, "lr": 9.610983995912397e-05}
{"train_loss": 0.4128369688987732, "global_step": 34746, "epoch": 390, "lr": 9.61096157597567e-05}
{"train_loss": 0.35655921697616577, "global_step": 34747, "epoch": 390, "lr": 9.610939155419056e-05}
{"train_loss": 0.40138334035873413, "global_step": 34748, "epoch": 390, "lr": 9.610916734242557e-05}
{"train_loss": 0.40395286679267883, "global_step": 34749, "epoch": 390, "lr": 9.610894312446176e-05}
{"train_loss": 0.3980255722999573, "global_step": 34750, "epoch": 390, "lr": 9.610871890029918e-05}
{"train_loss": 0.4281545579433441, "global_step": 34751, "epoch": 390, "lr": 9.610849466993783e-05}
{"train_loss": 0.3663671612739563, "global_step": 34752, "epoch": 390, "lr": 9.610827043337775e-05}
{"train_loss": 0.2492738962173462, "global_step": 34753, "epoch": 390, "lr": 9.610804619061899e-05}
{"train_loss": 0.31846722960472107, "global_step": 34754, "epoch": 390, "lr": 9.610782194166156e-05}
{"train_loss": 0.3539113998413086, "global_step": 34755, "epoch": 390, "lr": 9.61075976865055e-05}
{"train_loss": 0.3035719394683838, "global_step": 34756, "epoch": 390, "lr": 9.610737342515082e-05}
{"train_loss": 0.3838808834552765, "global_step": 34757, "epoch": 390, "lr": 9.610714915759758e-05}
{"train_loss": 0.29456183314323425, "global_step": 34758, "epoch": 390, "lr": 9.61069248838458e-05}
{"train_loss": 0.3423973321914673, "global_step": 34759, "epoch": 390, "lr": 9.610670060389549e-05}
{"train_loss": 0.3093174695968628, "global_step": 34760, "epoch": 390, "lr": 9.61064763177467e-05}
{"train_loss": 0.31013503670692444, "global_step": 34761, "epoch": 390, "lr": 9.610625202539947e-05}
{"train_loss": 0.37867000699043274, "global_step": 34762, "epoch": 390, "lr": 9.61060277268538e-05}
{"train_loss": 0.32921302318573, "global_step": 34763, "epoch": 390, "lr": 9.610580342210975e-05}
{"train_loss": 0.3151583969593048, "global_step": 34764, "epoch": 390, "lr": 9.610557911116733e-05}
{"train_loss": 0.2753007709980011, "global_step": 34765, "epoch": 390, "lr": 9.610535479402657e-05}
{"train_loss": 0.37527456879615784, "global_step": 34766, "epoch": 390, "lr": 9.610513047068752e-05}
{"train_loss": 0.38488706946372986, "global_step": 34767, "epoch": 390, "lr": 9.610490614115019e-05}
{"train_loss": 0.3540349304676056, "global_step": 34768, "epoch": 390, "lr": 9.61046818054146e-05}
{"train_loss": 0.491324245929718, "global_step": 34769, "epoch": 390, "lr": 9.610445746348082e-05}
{"train_loss": 0.442076712846756, "global_step": 34770, "epoch": 390, "lr": 9.610423311534887e-05}
{"train_loss": 0.3502906262874603, "global_step": 34771, "epoch": 390, "lr": 9.610400876101875e-05}
{"train_loss": 0.31705358624458313, "global_step": 34772, "epoch": 390, "lr": 9.610378440049051e-05}
{"train_loss": 0.2814910411834717, "global_step": 34773, "epoch": 390, "lr": 9.610356003376417e-05}
{"train_loss": 0.32332393527030945, "global_step": 34774, "epoch": 390, "lr": 9.610333566083978e-05}
{"train_loss": 0.3859274387359619, "global_step": 34775, "epoch": 390, "lr": 9.610311128171735e-05}
{"train_loss": 0.41658762097358704, "global_step": 34776, "epoch": 390, "lr": 9.610288689639691e-05}
{"train_loss": 0.4027176797389984, "global_step": 34777, "epoch": 390, "lr": 9.610266250487852e-05}
{"train_loss": 0.27226462960243225, "global_step": 34778, "epoch": 390, "lr": 9.610243810716217e-05}
{"train_loss": 0.2864880859851837, "global_step": 34779, "epoch": 390, "lr": 9.610221370324792e-05}
{"train_loss": 0.37884199619293213, "global_step": 34780, "epoch": 390, "lr": 9.61019892931358e-05}
{"train_loss": 0.3222501277923584, "global_step": 34781, "epoch": 390, "lr": 9.610176487682582e-05}
{"train_loss": 0.38504916429519653, "global_step": 34782, "epoch": 390, "lr": 9.610154045431801e-05}
{"train_loss": 0.3241288363933563, "global_step": 34783, "epoch": 390, "lr": 9.610131602561242e-05}
{"train_loss": 0.33768853545188904, "global_step": 34784, "epoch": 390, "lr": 9.610109159070906e-05}
{"train_loss": 0.30900809168815613, "global_step": 34785, "epoch": 390, "lr": 9.610086714960797e-05}
{"train_loss": 0.30544209480285645, "global_step": 34786, "epoch": 390, "lr": 9.610064270230921e-05}
{"train_loss": 0.2440711259841919, "global_step": 34787, "epoch": 390, "lr": 9.610041824881274e-05}
{"train_loss": 0.2792462110519409, "global_step": 34788, "epoch": 390, "lr": 9.610019378911865e-05}
{"train_loss": 0.4365927577018738, "global_step": 34789, "epoch": 390, "lr": 9.609996932322695e-05}
{"train_loss": 0.3891166150569916, "global_step": 34790, "epoch": 390, "lr": 9.609974485113768e-05}
{"train_loss": 0.36608630418777466, "global_step": 34791, "epoch": 390, "lr": 9.609952037285085e-05}
{"train_loss": 0.3269573152065277, "global_step": 34792, "epoch": 390, "lr": 9.60992958883665e-05}
{"train_loss": 0.31780028343200684, "global_step": 34793, "epoch": 390, "lr": 9.609907139768467e-05}
{"train_loss": 0.2977992594242096, "global_step": 34794, "epoch": 390, "lr": 9.609884690080537e-05}
{"train_loss": 0.3045164942741394, "global_step": 34795, "epoch": 390, "lr": 9.609862239772864e-05}
{"train_loss": 0.41741544008255005, "global_step": 34796, "epoch": 390, "lr": 9.609839788845454e-05}
{"train_loss": 0.3658558130264282, "global_step": 34797, "epoch": 390, "lr": 9.609817337298304e-05}
{"train_loss": 0.34250935393103055, "global_step": 34798, "epoch": 390, "lr": 9.609794885131422e-05, "val_loss": 2.520616292953491, "train_action_mse_error": 23.762971878051758}
{"train_loss": 0.5401477217674255, "global_step": 34799, "epoch": 391, "lr": 9.609772432344807e-05}
{"train_loss": 0.35719457268714905, "global_step": 34800, "epoch": 391, "lr": 9.609749978938467e-05}
{"train_loss": 0.42244717478752136, "global_step": 34801, "epoch": 391, "lr": 9.609727524912402e-05}
{"train_loss": 0.32091373205184937, "global_step": 34802, "epoch": 391, "lr": 9.609705070266614e-05}
{"train_loss": 0.33136481046676636, "global_step": 34803, "epoch": 391, "lr": 9.609682615001108e-05}
{"train_loss": 0.38661566376686096, "global_step": 34804, "epoch": 391, "lr": 9.609660159115887e-05}
{"train_loss": 0.4342178702354431, "global_step": 34805, "epoch": 391, "lr": 9.609637702610953e-05}
{"train_loss": 0.2819736897945404, "global_step": 34806, "epoch": 391, "lr": 9.60961524548631e-05}
{"train_loss": 0.2506694495677948, "global_step": 34807, "epoch": 391, "lr": 9.609592787741959e-05}
{"train_loss": 0.27505257725715637, "global_step": 34808, "epoch": 391, "lr": 9.609570329377904e-05}
{"train_loss": 0.28800955414772034, "global_step": 34809, "epoch": 391, "lr": 9.60954787039415e-05}
{"train_loss": 0.4152562916278839, "global_step": 34810, "epoch": 391, "lr": 9.609525410790698e-05}
{"train_loss": 0.3239799737930298, "global_step": 34811, "epoch": 391, "lr": 9.609502950567551e-05}
{"train_loss": 0.29260990023612976, "global_step": 34812, "epoch": 391, "lr": 9.609480489724714e-05}
{"train_loss": 0.355496346950531, "global_step": 34813, "epoch": 391, "lr": 9.609458028262186e-05}
{"train_loss": 0.3310466408729553, "global_step": 34814, "epoch": 391, "lr": 9.609435566179974e-05}
{"train_loss": 0.26816684007644653, "global_step": 34815, "epoch": 391, "lr": 9.60941310347808e-05}
{"train_loss": 0.3181004226207733, "global_step": 34816, "epoch": 391, "lr": 9.609390640156505e-05}
{"train_loss": 0.2728961110115051, "global_step": 34817, "epoch": 391, "lr": 9.609368176215257e-05}
{"train_loss": 0.310886025428772, "global_step": 34818, "epoch": 391, "lr": 9.609345711654332e-05}
{"train_loss": 0.281980037689209, "global_step": 34819, "epoch": 391, "lr": 9.609323246473738e-05}
{"train_loss": 0.32891586422920227, "global_step": 34820, "epoch": 391, "lr": 9.609300780673478e-05}
{"train_loss": 0.3203287422657013, "global_step": 34821, "epoch": 391, "lr": 9.609278314253551e-05}
{"train_loss": 0.2913820147514343, "global_step": 34822, "epoch": 391, "lr": 9.609255847213966e-05}
{"train_loss": 0.3406234085559845, "global_step": 34823, "epoch": 391, "lr": 9.609233379554719e-05}
{"train_loss": 0.29631537199020386, "global_step": 34824, "epoch": 391, "lr": 9.609210911275819e-05}
{"train_loss": 0.25386157631874084, "global_step": 34825, "epoch": 391, "lr": 9.609188442377266e-05}
{"train_loss": 0.3311816453933716, "global_step": 34826, "epoch": 391, "lr": 9.609165972859064e-05}
{"train_loss": 0.32495561242103577, "global_step": 34827, "epoch": 391, "lr": 9.609143502721217e-05}
{"train_loss": 0.2757221758365631, "global_step": 34828, "epoch": 391, "lr": 9.609121031963726e-05}
{"train_loss": 0.38767677545547485, "global_step": 34829, "epoch": 391, "lr": 9.609098560586593e-05}
{"train_loss": 0.34615686535835266, "global_step": 34830, "epoch": 391, "lr": 9.609076088589826e-05}
{"train_loss": 0.2928802967071533, "global_step": 34831, "epoch": 391, "lr": 9.609053615973423e-05}
{"train_loss": 0.31516703963279724, "global_step": 34832, "epoch": 391, "lr": 9.609031142737389e-05}
{"train_loss": 0.3058210611343384, "global_step": 34833, "epoch": 391, "lr": 9.609008668881727e-05}
{"train_loss": 0.38599464297294617, "global_step": 34834, "epoch": 391, "lr": 9.60898619440644e-05}
{"train_loss": 0.3567497730255127, "global_step": 34835, "epoch": 391, "lr": 9.608963719311532e-05}
{"train_loss": 0.32716673612594604, "global_step": 34836, "epoch": 391, "lr": 9.608941243597004e-05}
{"train_loss": 0.38666269183158875, "global_step": 34837, "epoch": 391, "lr": 9.60891876726286e-05}
{"train_loss": 0.3463574945926666, "global_step": 34838, "epoch": 391, "lr": 9.608896290309103e-05}
{"train_loss": 0.35582420229911804, "global_step": 34839, "epoch": 391, "lr": 9.608873812735736e-05}
{"train_loss": 0.3772873282432556, "global_step": 34840, "epoch": 391, "lr": 9.608851334542763e-05}
{"train_loss": 0.40645119547843933, "global_step": 34841, "epoch": 391, "lr": 9.608828855730185e-05}
{"train_loss": 0.3874131143093109, "global_step": 34842, "epoch": 391, "lr": 9.608806376298008e-05}
{"train_loss": 0.2940128445625305, "global_step": 34843, "epoch": 391, "lr": 9.608783896246232e-05}
{"train_loss": 0.3622311055660248, "global_step": 34844, "epoch": 391, "lr": 9.608761415574861e-05}
{"train_loss": 0.3973725438117981, "global_step": 34845, "epoch": 391, "lr": 9.608738934283898e-05}
{"train_loss": 0.2838870882987976, "global_step": 34846, "epoch": 391, "lr": 9.608716452373347e-05}
{"train_loss": 0.29665690660476685, "global_step": 34847, "epoch": 391, "lr": 9.608693969843209e-05}
{"train_loss": 0.29951468110084534, "global_step": 34848, "epoch": 391, "lr": 9.60867148669349e-05}
{"train_loss": 0.29572373628616333, "global_step": 34849, "epoch": 391, "lr": 9.608649002924188e-05}
{"train_loss": 0.3757781982421875, "global_step": 34850, "epoch": 391, "lr": 9.608626518535313e-05}
{"train_loss": 0.4488745331764221, "global_step": 34851, "epoch": 391, "lr": 9.608604033526863e-05}
{"train_loss": 0.49225422739982605, "global_step": 34852, "epoch": 391, "lr": 9.608581547898841e-05}
{"train_loss": 0.3541225790977478, "global_step": 34853, "epoch": 391, "lr": 9.608559061651254e-05}
{"train_loss": 0.348924845457077, "global_step": 34854, "epoch": 391, "lr": 9.6085365747841e-05}
{"train_loss": 0.3314679265022278, "global_step": 34855, "epoch": 391, "lr": 9.608514087297386e-05}
{"train_loss": 0.3968234360218048, "global_step": 34856, "epoch": 391, "lr": 9.608491599191113e-05}
{"train_loss": 0.2997070848941803, "global_step": 34857, "epoch": 391, "lr": 9.608469110465285e-05}
{"train_loss": 0.27912136912345886, "global_step": 34858, "epoch": 391, "lr": 9.608446621119904e-05}
{"train_loss": 0.22578558325767517, "global_step": 34859, "epoch": 391, "lr": 9.608424131154973e-05}
{"train_loss": 0.43427708745002747, "global_step": 34860, "epoch": 391, "lr": 9.608401640570496e-05}
{"train_loss": 0.39553195238113403, "global_step": 34861, "epoch": 391, "lr": 9.608379149366475e-05}
{"train_loss": 0.29758012294769287, "global_step": 34862, "epoch": 391, "lr": 9.608356657542915e-05}
{"train_loss": 0.3337791860103607, "global_step": 34863, "epoch": 391, "lr": 9.608334165099816e-05}
{"train_loss": 0.2946023941040039, "global_step": 34864, "epoch": 391, "lr": 9.608311672037183e-05}
{"train_loss": 0.3245977461338043, "global_step": 34865, "epoch": 391, "lr": 9.608289178355018e-05}
{"train_loss": 0.33529677987098694, "global_step": 34866, "epoch": 391, "lr": 9.608266684053326e-05}
{"train_loss": 0.31468823552131653, "global_step": 34867, "epoch": 391, "lr": 9.60824418913211e-05}
{"train_loss": 0.30490943789482117, "global_step": 34868, "epoch": 391, "lr": 9.608221693591371e-05}
{"train_loss": 0.3977055549621582, "global_step": 34869, "epoch": 391, "lr": 9.608199197431111e-05}
{"train_loss": 0.38732844591140747, "global_step": 34870, "epoch": 391, "lr": 9.608176700651335e-05}
{"train_loss": 0.3114231526851654, "global_step": 34871, "epoch": 391, "lr": 9.608154203252046e-05}
{"train_loss": 0.2226194590330124, "global_step": 34872, "epoch": 391, "lr": 9.608131705233247e-05}
{"train_loss": 0.2796734571456909, "global_step": 34873, "epoch": 391, "lr": 9.608109206594942e-05}
{"train_loss": 0.36701300740242004, "global_step": 34874, "epoch": 391, "lr": 9.608086707337133e-05}
{"train_loss": 0.36679816246032715, "global_step": 34875, "epoch": 391, "lr": 9.608064207459822e-05}
{"train_loss": 0.30989763140678406, "global_step": 34876, "epoch": 391, "lr": 9.608041706963012e-05}
{"train_loss": 0.3163648545742035, "global_step": 34877, "epoch": 391, "lr": 9.608019205846707e-05}
{"train_loss": 0.2545091211795807, "global_step": 34878, "epoch": 391, "lr": 9.607996704110914e-05}
{"train_loss": 0.3864803910255432, "global_step": 34879, "epoch": 391, "lr": 9.607974201755627e-05}
{"train_loss": 0.3865751028060913, "global_step": 34880, "epoch": 391, "lr": 9.607951698780856e-05}
{"train_loss": 0.478278249502182, "global_step": 34881, "epoch": 391, "lr": 9.607929195186603e-05}
{"train_loss": 0.358163446187973, "global_step": 34882, "epoch": 391, "lr": 9.607906690972868e-05}
{"train_loss": 0.44705286622047424, "global_step": 34883, "epoch": 391, "lr": 9.607884186139658e-05}
{"train_loss": 0.395643949508667, "global_step": 34884, "epoch": 391, "lr": 9.607861680686972e-05}
{"train_loss": 0.3313007354736328, "global_step": 34885, "epoch": 391, "lr": 9.607839174614817e-05}
{"train_loss": 0.36493980884552, "global_step": 34886, "epoch": 391, "lr": 9.607816667923192e-05}
{"train_loss": 0.339669337935662, "global_step": 34887, "epoch": 391, "lr": 9.607794160612106e-05, "val_loss": 2.4107844829559326}
{"train_loss": 0.3344831168651581, "global_step": 34888, "epoch": 392, "lr": 9.607771652681556e-05}
{"train_loss": 0.3776782155036926, "global_step": 34889, "epoch": 392, "lr": 9.607749144131547e-05}
{"train_loss": 0.4331305921077728, "global_step": 34890, "epoch": 392, "lr": 9.607726634962082e-05}
{"train_loss": 0.3556940257549286, "global_step": 34891, "epoch": 392, "lr": 9.607704125173165e-05}
{"train_loss": 0.3528868854045868, "global_step": 34892, "epoch": 392, "lr": 9.607681614764799e-05}
{"train_loss": 0.37084683775901794, "global_step": 34893, "epoch": 392, "lr": 9.607659103736985e-05}
{"train_loss": 0.4106108546257019, "global_step": 34894, "epoch": 392, "lr": 9.607636592089728e-05}
{"train_loss": 0.3205515444278717, "global_step": 34895, "epoch": 392, "lr": 9.607614079823031e-05}
{"train_loss": 0.27067357301712036, "global_step": 34896, "epoch": 392, "lr": 9.607591566936896e-05}
{"train_loss": 0.3153901994228363, "global_step": 34897, "epoch": 392, "lr": 9.607569053431327e-05}
{"train_loss": 0.3694421947002411, "global_step": 34898, "epoch": 392, "lr": 9.607546539306324e-05}
{"train_loss": 0.4320695400238037, "global_step": 34899, "epoch": 392, "lr": 9.607524024561896e-05}
{"train_loss": 0.29679036140441895, "global_step": 34900, "epoch": 392, "lr": 9.60750150919804e-05}
{"train_loss": 0.2954656183719635, "global_step": 34901, "epoch": 392, "lr": 9.607478993214764e-05}
{"train_loss": 0.432218462228775, "global_step": 34902, "epoch": 392, "lr": 9.607456476612067e-05}
{"train_loss": 0.2840861976146698, "global_step": 34903, "epoch": 392, "lr": 9.607433959389953e-05}
{"train_loss": 0.30449309945106506, "global_step": 34904, "epoch": 392, "lr": 9.607411441548426e-05}
{"train_loss": 0.3124254047870636, "global_step": 34905, "epoch": 392, "lr": 9.60738892308749e-05}
{"train_loss": 0.38500407338142395, "global_step": 34906, "epoch": 392, "lr": 9.607366404007145e-05}
{"train_loss": 0.314052939414978, "global_step": 34907, "epoch": 392, "lr": 9.607343884307397e-05}
{"train_loss": 0.4683046042919159, "global_step": 34908, "epoch": 392, "lr": 9.607321363988248e-05}
{"train_loss": 0.393385648727417, "global_step": 34909, "epoch": 392, "lr": 9.6072988430497e-05}
{"train_loss": 0.37619978189468384, "global_step": 34910, "epoch": 392, "lr": 9.607276321491756e-05}
{"train_loss": 0.30538007616996765, "global_step": 34911, "epoch": 392, "lr": 9.607253799314421e-05}
{"train_loss": 0.3921910524368286, "global_step": 34912, "epoch": 392, "lr": 9.607231276517697e-05}
{"train_loss": 0.42352327704429626, "global_step": 34913, "epoch": 392, "lr": 9.607208753101587e-05}
{"train_loss": 0.3838024139404297, "global_step": 34914, "epoch": 392, "lr": 9.607186229066093e-05}
{"train_loss": 0.3575825095176697, "global_step": 34915, "epoch": 392, "lr": 9.60716370441122e-05}
{"train_loss": 0.35059311985969543, "global_step": 34916, "epoch": 392, "lr": 9.607141179136969e-05}
{"train_loss": 0.29966628551483154, "global_step": 34917, "epoch": 392, "lr": 9.607118653243346e-05}
{"train_loss": 0.367958664894104, "global_step": 34918, "epoch": 392, "lr": 9.607096126730349e-05}
{"train_loss": 0.471397340297699, "global_step": 34919, "epoch": 392, "lr": 9.607073599597987e-05}
{"train_loss": 0.3124363422393799, "global_step": 34920, "epoch": 392, "lr": 9.607051071846258e-05}
{"train_loss": 0.3939304053783417, "global_step": 34921, "epoch": 392, "lr": 9.607028543475169e-05}
{"train_loss": 0.3583872318267822, "global_step": 34922, "epoch": 392, "lr": 9.607006014484719e-05}
{"train_loss": 0.45364615321159363, "global_step": 34923, "epoch": 392, "lr": 9.606983484874914e-05}
{"train_loss": 0.4646872580051422, "global_step": 34924, "epoch": 392, "lr": 9.606960954645756e-05}
{"train_loss": 0.4039803147315979, "global_step": 34925, "epoch": 392, "lr": 9.60693842379725e-05}
{"train_loss": 0.3814365565776825, "global_step": 34926, "epoch": 392, "lr": 9.606915892329395e-05}
{"train_loss": 0.4875890612602234, "global_step": 34927, "epoch": 392, "lr": 9.606893360242198e-05}
{"train_loss": 0.26917973160743713, "global_step": 34928, "epoch": 392, "lr": 9.60687082753566e-05}
{"train_loss": 0.39947429299354553, "global_step": 34929, "epoch": 392, "lr": 9.606848294209784e-05}
{"train_loss": 0.3840840756893158, "global_step": 34930, "epoch": 392, "lr": 9.606825760264573e-05}
{"train_loss": 0.3800528347492218, "global_step": 34931, "epoch": 392, "lr": 9.606803225700032e-05}
{"train_loss": 0.406042218208313, "global_step": 34932, "epoch": 392, "lr": 9.60678069051616e-05}
{"train_loss": 0.20194579660892487, "global_step": 34933, "epoch": 392, "lr": 9.606758154712965e-05}
{"train_loss": 0.2964678406715393, "global_step": 34934, "epoch": 392, "lr": 9.606735618290446e-05}
{"train_loss": 0.2721339166164398, "global_step": 34935, "epoch": 392, "lr": 9.606713081248607e-05}
{"train_loss": 0.38645458221435547, "global_step": 34936, "epoch": 392, "lr": 9.606690543587454e-05}
{"train_loss": 0.4068754017353058, "global_step": 34937, "epoch": 392, "lr": 9.606668005306984e-05}
{"train_loss": 0.3743409514427185, "global_step": 34938, "epoch": 392, "lr": 9.606645466407207e-05}
{"train_loss": 0.40494000911712646, "global_step": 34939, "epoch": 392, "lr": 9.606622926888121e-05}
{"train_loss": 0.41549035906791687, "global_step": 34940, "epoch": 392, "lr": 9.606600386749732e-05}
{"train_loss": 0.4043886959552765, "global_step": 34941, "epoch": 392, "lr": 9.60657784599204e-05}
{"train_loss": 0.414272665977478, "global_step": 34942, "epoch": 392, "lr": 9.60655530461505e-05}
{"train_loss": 0.3620871305465698, "global_step": 34943, "epoch": 392, "lr": 9.606532762618765e-05}
{"train_loss": 0.43528252840042114, "global_step": 34944, "epoch": 392, "lr": 9.606510220003188e-05}
{"train_loss": 0.3768216669559479, "global_step": 34945, "epoch": 392, "lr": 9.606487676768323e-05}
{"train_loss": 0.34194332361221313, "global_step": 34946, "epoch": 392, "lr": 9.606465132914171e-05}
{"train_loss": 0.32203051447868347, "global_step": 34947, "epoch": 392, "lr": 9.606442588440735e-05}
{"train_loss": 0.3401411175727844, "global_step": 34948, "epoch": 392, "lr": 9.60642004334802e-05}
{"train_loss": 0.3437292277812958, "global_step": 34949, "epoch": 392, "lr": 9.606397497636027e-05}
{"train_loss": 0.3896535336971283, "global_step": 34950, "epoch": 392, "lr": 9.606374951304762e-05}
{"train_loss": 0.27618005871772766, "global_step": 34951, "epoch": 392, "lr": 9.606352404354224e-05}
{"train_loss": 0.3502359390258789, "global_step": 34952, "epoch": 392, "lr": 9.606329856784419e-05}
{"train_loss": 0.4065295457839966, "global_step": 34953, "epoch": 392, "lr": 9.606307308595349e-05}
{"train_loss": 0.3760666847229004, "global_step": 34954, "epoch": 392, "lr": 9.606284759787016e-05}
{"train_loss": 0.37360572814941406, "global_step": 34955, "epoch": 392, "lr": 9.606262210359425e-05}
{"train_loss": 0.2710244357585907, "global_step": 34956, "epoch": 392, "lr": 9.606239660312579e-05}
{"train_loss": 0.30841532349586487, "global_step": 34957, "epoch": 392, "lr": 9.60621710964648e-05}
{"train_loss": 0.4107532501220703, "global_step": 34958, "epoch": 392, "lr": 9.60619455836113e-05}
{"train_loss": 0.3410365879535675, "global_step": 34959, "epoch": 392, "lr": 9.606172006456535e-05}
{"train_loss": 0.28740572929382324, "global_step": 34960, "epoch": 392, "lr": 9.606149453932694e-05}
{"train_loss": 0.25883427262306213, "global_step": 34961, "epoch": 392, "lr": 9.606126900789613e-05}
{"train_loss": 0.3297621011734009, "global_step": 34962, "epoch": 392, "lr": 9.606104347027295e-05}
{"train_loss": 0.42976245284080505, "global_step": 34963, "epoch": 392, "lr": 9.606081792645745e-05}
{"train_loss": 0.30444109439849854, "global_step": 34964, "epoch": 392, "lr": 9.60605923764496e-05}
{"train_loss": 0.30997276306152344, "global_step": 34965, "epoch": 392, "lr": 9.606036682024948e-05}
{"train_loss": 0.28306522965431213, "global_step": 34966, "epoch": 392, "lr": 9.60601412578571e-05}
{"train_loss": 0.36793139576911926, "global_step": 34967, "epoch": 392, "lr": 9.605991568927248e-05}
{"train_loss": 0.286000519990921, "global_step": 34968, "epoch": 392, "lr": 9.60596901144957e-05}
{"train_loss": 0.38459092378616333, "global_step": 34969, "epoch": 392, "lr": 9.605946453352673e-05}
{"train_loss": 0.3058561086654663, "global_step": 34970, "epoch": 392, "lr": 9.605923894636562e-05}
{"train_loss": 0.25435587763786316, "global_step": 34971, "epoch": 392, "lr": 9.605901335301243e-05}
{"train_loss": 0.5025303363800049, "global_step": 34972, "epoch": 392, "lr": 9.605878775346715e-05}
{"train_loss": 0.31512799859046936, "global_step": 34973, "epoch": 392, "lr": 9.605856214772984e-05}
{"train_loss": 0.27615121006965637, "global_step": 34974, "epoch": 392, "lr": 9.605833653580052e-05}
{"train_loss": 0.3398180305957794, "global_step": 34975, "epoch": 392, "lr": 9.60581109176792e-05}
{"train_loss": 0.35705137503950785, "global_step": 34976, "epoch": 392, "lr": 9.605788529336595e-05, "val_loss": 2.485966920852661}
{"train_loss": 0.37793558835983276, "global_step": 34977, "epoch": 393, "lr": 9.605765966286076e-05}
{"train_loss": 0.33583006262779236, "global_step": 34978, "epoch": 393, "lr": 9.605743402616369e-05}
{"train_loss": 0.39374685287475586, "global_step": 34979, "epoch": 393, "lr": 9.605720838327474e-05}
{"train_loss": 0.3253270089626312, "global_step": 34980, "epoch": 393, "lr": 9.605698273419399e-05}
{"train_loss": 0.3347945809364319, "global_step": 34981, "epoch": 393, "lr": 9.605675707892141e-05}
{"train_loss": 0.3322649896144867, "global_step": 34982, "epoch": 393, "lr": 9.605653141745708e-05}
{"train_loss": 0.24357037246227264, "global_step": 34983, "epoch": 393, "lr": 9.605630574980101e-05}
{"train_loss": 0.22918391227722168, "global_step": 34984, "epoch": 393, "lr": 9.605608007595322e-05}
{"train_loss": 0.3954995572566986, "global_step": 34985, "epoch": 393, "lr": 9.605585439591376e-05}
{"train_loss": 0.23188230395317078, "global_step": 34986, "epoch": 393, "lr": 9.605562870968265e-05}
{"train_loss": 0.34654954075813293, "global_step": 34987, "epoch": 393, "lr": 9.605540301725993e-05}
{"train_loss": 0.3122994601726532, "global_step": 34988, "epoch": 393, "lr": 9.605517731864562e-05}
{"train_loss": 0.259268194437027, "global_step": 34989, "epoch": 393, "lr": 9.605495161383974e-05}
{"train_loss": 0.2862708270549774, "global_step": 34990, "epoch": 393, "lr": 9.605472590284235e-05}
{"train_loss": 0.39026787877082825, "global_step": 34991, "epoch": 393, "lr": 9.605450018565345e-05}
{"train_loss": 0.3731381297111511, "global_step": 34992, "epoch": 393, "lr": 9.605427446227308e-05}
{"train_loss": 0.29616641998291016, "global_step": 34993, "epoch": 393, "lr": 9.605404873270129e-05}
{"train_loss": 0.30821818113327026, "global_step": 34994, "epoch": 393, "lr": 9.605382299693808e-05}
{"train_loss": 0.344362735748291, "global_step": 34995, "epoch": 393, "lr": 9.60535972549835e-05}
{"train_loss": 0.26073628664016724, "global_step": 34996, "epoch": 393, "lr": 9.605337150683758e-05}
{"train_loss": 0.2715434730052948, "global_step": 34997, "epoch": 393, "lr": 9.605314575250034e-05}
{"train_loss": 0.28014233708381653, "global_step": 34998, "epoch": 393, "lr": 9.605291999197181e-05}
{"train_loss": 0.27195972204208374, "global_step": 34999, "epoch": 393, "lr": 9.605269422525204e-05}
{"train_loss": 0.2759974002838135, "global_step": 35000, "epoch": 393, "lr": 9.605246845234104e-05}
{"train_loss": 0.31629678606987, "global_step": 35001, "epoch": 393, "lr": 9.605224267323885e-05}
{"train_loss": 0.31956303119659424, "global_step": 35002, "epoch": 393, "lr": 9.605201688794549e-05}
{"train_loss": 0.2772246301174164, "global_step": 35003, "epoch": 393, "lr": 9.605179109646101e-05}
{"train_loss": 0.3167833089828491, "global_step": 35004, "epoch": 393, "lr": 9.605156529878542e-05}
{"train_loss": 0.30660519003868103, "global_step": 35005, "epoch": 393, "lr": 9.605133949491876e-05}
{"train_loss": 0.3027045726776123, "global_step": 35006, "epoch": 393, "lr": 9.605111368486105e-05}
{"train_loss": 0.2778165936470032, "global_step": 35007, "epoch": 393, "lr": 9.605088786861233e-05}
{"train_loss": 0.32804280519485474, "global_step": 35008, "epoch": 393, "lr": 9.605066204617263e-05}
{"train_loss": 0.353601336479187, "global_step": 35009, "epoch": 393, "lr": 9.605043621754199e-05}
{"train_loss": 0.24988962709903717, "global_step": 35010, "epoch": 393, "lr": 9.605021038272044e-05}
{"train_loss": 0.4191628098487854, "global_step": 35011, "epoch": 393, "lr": 9.604998454170798e-05}
{"train_loss": 0.23746180534362793, "global_step": 35012, "epoch": 393, "lr": 9.604975869450467e-05}
{"train_loss": 0.35710009932518005, "global_step": 35013, "epoch": 393, "lr": 9.604953284111052e-05}
{"train_loss": 0.28387004137039185, "global_step": 35014, "epoch": 393, "lr": 9.604930698152558e-05}
{"train_loss": 0.27653661370277405, "global_step": 35015, "epoch": 393, "lr": 9.604908111574989e-05}
{"train_loss": 0.2776333689689636, "global_step": 35016, "epoch": 393, "lr": 9.604885524378343e-05}
{"train_loss": 0.31855207681655884, "global_step": 35017, "epoch": 393, "lr": 9.604862936562628e-05}
{"train_loss": 0.3457375466823578, "global_step": 35018, "epoch": 393, "lr": 9.604840348127844e-05}
{"train_loss": 0.33589062094688416, "global_step": 35019, "epoch": 393, "lr": 9.604817759073996e-05}
{"train_loss": 0.3284366726875305, "global_step": 35020, "epoch": 393, "lr": 9.604795169401088e-05}
{"train_loss": 0.2151469886302948, "global_step": 35021, "epoch": 393, "lr": 9.60477257910912e-05}
{"train_loss": 0.265736848115921, "global_step": 35022, "epoch": 393, "lr": 9.604749988198097e-05}
{"train_loss": 0.3390038013458252, "global_step": 35023, "epoch": 393, "lr": 9.60472739666802e-05}
{"train_loss": 0.38604554533958435, "global_step": 35024, "epoch": 393, "lr": 9.604704804518895e-05}
{"train_loss": 0.31414327025413513, "global_step": 35025, "epoch": 393, "lr": 9.604682211750723e-05}
{"train_loss": 0.4178931415081024, "global_step": 35026, "epoch": 393, "lr": 9.604659618363507e-05}
{"train_loss": 0.23468267917633057, "global_step": 35027, "epoch": 393, "lr": 9.604637024357253e-05}
{"train_loss": 0.2550041973590851, "global_step": 35028, "epoch": 393, "lr": 9.60461442973196e-05}
{"train_loss": 0.30781933665275574, "global_step": 35029, "epoch": 393, "lr": 9.604591834487632e-05}
{"train_loss": 0.34942933917045593, "global_step": 35030, "epoch": 393, "lr": 9.604569238624273e-05}
{"train_loss": 0.3397580683231354, "global_step": 35031, "epoch": 393, "lr": 9.604546642141887e-05}
{"train_loss": 0.30038172006607056, "global_step": 35032, "epoch": 393, "lr": 9.604524045040475e-05}
{"train_loss": 0.4050639569759369, "global_step": 35033, "epoch": 393, "lr": 9.60450144732004e-05}
{"train_loss": 0.4584256410598755, "global_step": 35034, "epoch": 393, "lr": 9.604478848980587e-05}
{"train_loss": 0.4073287546634674, "global_step": 35035, "epoch": 393, "lr": 9.604456250022118e-05}
{"train_loss": 0.32127946615219116, "global_step": 35036, "epoch": 393, "lr": 9.604433650444637e-05}
{"train_loss": 0.3710547983646393, "global_step": 35037, "epoch": 393, "lr": 9.604411050248145e-05}
{"train_loss": 0.44127053022384644, "global_step": 35038, "epoch": 393, "lr": 9.604388449432646e-05}
{"train_loss": 0.3075178861618042, "global_step": 35039, "epoch": 393, "lr": 9.604365847998143e-05}
{"train_loss": 0.47806283831596375, "global_step": 35040, "epoch": 393, "lr": 9.604343245944637e-05}
{"train_loss": 0.32520484924316406, "global_step": 35041, "epoch": 393, "lr": 9.604320643272136e-05}
{"train_loss": 0.24930444359779358, "global_step": 35042, "epoch": 393, "lr": 9.60429803998064e-05}
{"train_loss": 0.2910782992839813, "global_step": 35043, "epoch": 393, "lr": 9.604275436070151e-05}
{"train_loss": 0.3898784816265106, "global_step": 35044, "epoch": 393, "lr": 9.604252831540675e-05}
{"train_loss": 0.37417590618133545, "global_step": 35045, "epoch": 393, "lr": 9.604230226392213e-05}
{"train_loss": 0.3331080675125122, "global_step": 35046, "epoch": 393, "lr": 9.604207620624766e-05}
{"train_loss": 0.2741236388683319, "global_step": 35047, "epoch": 393, "lr": 9.604185014238342e-05}
{"train_loss": 0.31585609912872314, "global_step": 35048, "epoch": 393, "lr": 9.60416240723294e-05}
{"train_loss": 0.3065798580646515, "global_step": 35049, "epoch": 393, "lr": 9.604139799608564e-05}
{"train_loss": 0.3370736539363861, "global_step": 35050, "epoch": 393, "lr": 9.604117191365219e-05}
{"train_loss": 0.31225335597991943, "global_step": 35051, "epoch": 393, "lr": 9.604094582502908e-05}
{"train_loss": 0.35771727561950684, "global_step": 35052, "epoch": 393, "lr": 9.604071973021628e-05}
{"train_loss": 0.29716455936431885, "global_step": 35053, "epoch": 393, "lr": 9.60404936292139e-05}
{"train_loss": 0.42195582389831543, "global_step": 35054, "epoch": 393, "lr": 9.604026752202193e-05}
{"train_loss": 0.2867635190486908, "global_step": 35055, "epoch": 393, "lr": 9.604004140864043e-05}
{"train_loss": 0.35286617279052734, "global_step": 35056, "epoch": 393, "lr": 9.603981528906938e-05}
{"train_loss": 0.3486330211162567, "global_step": 35057, "epoch": 393, "lr": 9.603958916330884e-05}
{"train_loss": 0.40479618310928345, "global_step": 35058, "epoch": 393, "lr": 9.603936303135883e-05}
{"train_loss": 0.4139673411846161, "global_step": 35059, "epoch": 393, "lr": 9.603913689321941e-05}
{"train_loss": 0.28375306725502014, "global_step": 35060, "epoch": 393, "lr": 9.603891074889057e-05}
{"train_loss": 0.26497727632522583, "global_step": 35061, "epoch": 393, "lr": 9.603868459837238e-05}
{"train_loss": 0.23774930834770203, "global_step": 35062, "epoch": 393, "lr": 9.603845844166483e-05}
{"train_loss": 0.357792466878891, "global_step": 35063, "epoch": 393, "lr": 9.603823227876797e-05}
{"train_loss": 0.32765698432922363, "global_step": 35064, "epoch": 393, "lr": 9.603800610968186e-05}
{"train_loss": 0.3233522273851245, "global_step": 35065, "epoch": 393, "lr": 9.603777993440648e-05, "val_loss": 2.541931390762329}
{"train_loss": 0.3661472201347351, "global_step": 35066, "epoch": 394, "lr": 9.603755375294187e-05}
{"train_loss": 0.23407337069511414, "global_step": 35067, "epoch": 394, "lr": 9.603732756528809e-05}
{"train_loss": 0.3763277232646942, "global_step": 35068, "epoch": 394, "lr": 9.603710137144515e-05}
{"train_loss": 0.36956557631492615, "global_step": 35069, "epoch": 394, "lr": 9.603687517141308e-05}
{"train_loss": 0.34815505146980286, "global_step": 35070, "epoch": 394, "lr": 9.603664896519191e-05}
{"train_loss": 0.26033255457878113, "global_step": 35071, "epoch": 394, "lr": 9.603642275278169e-05}
{"train_loss": 0.3126743733882904, "global_step": 35072, "epoch": 394, "lr": 9.603619653418242e-05}
{"train_loss": 0.3391646146774292, "global_step": 35073, "epoch": 394, "lr": 9.603597030939412e-05}
{"train_loss": 0.20776771008968353, "global_step": 35074, "epoch": 394, "lr": 9.60357440784169e-05}
{"train_loss": 0.29854801297187805, "global_step": 35075, "epoch": 394, "lr": 9.603551784125069e-05}
{"train_loss": 0.4282028079032898, "global_step": 35076, "epoch": 394, "lr": 9.603529159789558e-05}
{"train_loss": 0.2651897966861725, "global_step": 35077, "epoch": 394, "lr": 9.603506534835158e-05}
{"train_loss": 0.25700774788856506, "global_step": 35078, "epoch": 394, "lr": 9.603483909261875e-05}
{"train_loss": 0.50273197889328, "global_step": 35079, "epoch": 394, "lr": 9.603461283069707e-05}
{"train_loss": 0.269161581993103, "global_step": 35080, "epoch": 394, "lr": 9.603438656258661e-05}
{"train_loss": 0.3618900179862976, "global_step": 35081, "epoch": 394, "lr": 9.603416028828739e-05}
{"train_loss": 0.40107208490371704, "global_step": 35082, "epoch": 394, "lr": 9.603393400779943e-05}
{"train_loss": 0.32561999559402466, "global_step": 35083, "epoch": 394, "lr": 9.603370772112277e-05}
{"train_loss": 0.32523131370544434, "global_step": 35084, "epoch": 394, "lr": 9.603348142825746e-05}
{"train_loss": 0.36136534810066223, "global_step": 35085, "epoch": 394, "lr": 9.603325512920348e-05}
{"train_loss": 0.3183762729167938, "global_step": 35086, "epoch": 394, "lr": 9.60330288239609e-05}
{"train_loss": 0.2948712706565857, "global_step": 35087, "epoch": 394, "lr": 9.603280251252976e-05}
{"train_loss": 0.2911999821662903, "global_step": 35088, "epoch": 394, "lr": 9.603257619491004e-05}
{"train_loss": 0.510065495967865, "global_step": 35089, "epoch": 394, "lr": 9.603234987110181e-05}
{"train_loss": 0.2684885859489441, "global_step": 35090, "epoch": 394, "lr": 9.603212354110509e-05}
{"train_loss": 0.3033032715320587, "global_step": 35091, "epoch": 394, "lr": 9.603189720491991e-05}
{"train_loss": 0.44830793142318726, "global_step": 35092, "epoch": 394, "lr": 9.603167086254632e-05}
{"train_loss": 0.2866225838661194, "global_step": 35093, "epoch": 394, "lr": 9.603144451398433e-05}
{"train_loss": 0.3156448304653168, "global_step": 35094, "epoch": 394, "lr": 9.603121815923397e-05}
{"train_loss": 0.3869481086730957, "global_step": 35095, "epoch": 394, "lr": 9.603099179829527e-05}
{"train_loss": 0.39840829372406006, "global_step": 35096, "epoch": 394, "lr": 9.603076543116825e-05}
{"train_loss": 0.3062852919101715, "global_step": 35097, "epoch": 394, "lr": 9.603053905785298e-05}
{"train_loss": 0.40889716148376465, "global_step": 35098, "epoch": 394, "lr": 9.603031267834945e-05}
{"train_loss": 0.4176531732082367, "global_step": 35099, "epoch": 394, "lr": 9.603008629265771e-05}
{"train_loss": 0.3395441174507141, "global_step": 35100, "epoch": 394, "lr": 9.602985990077778e-05}
{"train_loss": 0.372749388217926, "global_step": 35101, "epoch": 394, "lr": 9.602963350270971e-05}
{"train_loss": 0.31002306938171387, "global_step": 35102, "epoch": 394, "lr": 9.602940709845351e-05}
{"train_loss": 0.2207261621952057, "global_step": 35103, "epoch": 394, "lr": 9.602918068800922e-05}
{"train_loss": 0.30074065923690796, "global_step": 35104, "epoch": 394, "lr": 9.602895427137685e-05}
{"train_loss": 0.25324442982673645, "global_step": 35105, "epoch": 394, "lr": 9.602872784855646e-05}
{"train_loss": 0.30288854241371155, "global_step": 35106, "epoch": 394, "lr": 9.602850141954808e-05}
{"train_loss": 0.3772209882736206, "global_step": 35107, "epoch": 394, "lr": 9.602827498435172e-05}
{"train_loss": 0.29079270362854004, "global_step": 35108, "epoch": 394, "lr": 9.602804854296741e-05}
{"train_loss": 0.31463751196861267, "global_step": 35109, "epoch": 394, "lr": 9.60278220953952e-05}
{"train_loss": 0.3543182909488678, "global_step": 35110, "epoch": 394, "lr": 9.60275956416351e-05}
{"train_loss": 0.5034711956977844, "global_step": 35111, "epoch": 394, "lr": 9.602736918168718e-05}
{"train_loss": 0.3513881266117096, "global_step": 35112, "epoch": 394, "lr": 9.60271427155514e-05}
{"train_loss": 0.3117121458053589, "global_step": 35113, "epoch": 394, "lr": 9.602691624322786e-05}
{"train_loss": 0.36406469345092773, "global_step": 35114, "epoch": 394, "lr": 9.602668976471655e-05}
{"train_loss": 0.3542579710483551, "global_step": 35115, "epoch": 394, "lr": 9.602646328001752e-05}
{"train_loss": 0.37117812037467957, "global_step": 35116, "epoch": 394, "lr": 9.60262367891308e-05}
{"train_loss": 0.42771196365356445, "global_step": 35117, "epoch": 394, "lr": 9.60260102920564e-05}
{"train_loss": 0.34715619683265686, "global_step": 35118, "epoch": 394, "lr": 9.602578378879437e-05}
{"train_loss": 0.3929930627346039, "global_step": 35119, "epoch": 394, "lr": 9.602555727934473e-05}
{"train_loss": 0.35667890310287476, "global_step": 35120, "epoch": 394, "lr": 9.602533076370752e-05}
{"train_loss": 0.1808381825685501, "global_step": 35121, "epoch": 394, "lr": 9.602510424188276e-05}
{"train_loss": 0.3462202250957489, "global_step": 35122, "epoch": 394, "lr": 9.602487771387047e-05}
{"train_loss": 0.448400616645813, "global_step": 35123, "epoch": 394, "lr": 9.602465117967072e-05}
{"train_loss": 0.34448349475860596, "global_step": 35124, "epoch": 394, "lr": 9.602442463928352e-05}
{"train_loss": 0.29287153482437134, "global_step": 35125, "epoch": 394, "lr": 9.602419809270889e-05}
{"train_loss": 0.30049198865890503, "global_step": 35126, "epoch": 394, "lr": 9.602397153994687e-05}
{"train_loss": 0.41476115584373474, "global_step": 35127, "epoch": 394, "lr": 9.602374498099746e-05}
{"train_loss": 0.35961395502090454, "global_step": 35128, "epoch": 394, "lr": 9.602351841586075e-05}
{"train_loss": 0.2942235469818115, "global_step": 35129, "epoch": 394, "lr": 9.602329184453674e-05}
{"train_loss": 0.41930100321769714, "global_step": 35130, "epoch": 394, "lr": 9.602306526702545e-05}
{"train_loss": 0.45737162232398987, "global_step": 35131, "epoch": 394, "lr": 9.602283868332691e-05}
{"train_loss": 0.3597264289855957, "global_step": 35132, "epoch": 394, "lr": 9.602261209344117e-05}
{"train_loss": 0.3746383786201477, "global_step": 35133, "epoch": 394, "lr": 9.602238549736825e-05}
{"train_loss": 0.2931216359138489, "global_step": 35134, "epoch": 394, "lr": 9.602215889510818e-05}
{"train_loss": 0.49010583758354187, "global_step": 35135, "epoch": 394, "lr": 9.602193228666099e-05}
{"train_loss": 0.30662649869918823, "global_step": 35136, "epoch": 394, "lr": 9.602170567202673e-05}
{"train_loss": 0.3614344894886017, "global_step": 35137, "epoch": 394, "lr": 9.602147905120538e-05}
{"train_loss": 0.37553849816322327, "global_step": 35138, "epoch": 394, "lr": 9.602125242419703e-05}
{"train_loss": 0.2618806064128876, "global_step": 35139, "epoch": 394, "lr": 9.602102579100167e-05}
{"train_loss": 0.35397467017173767, "global_step": 35140, "epoch": 394, "lr": 9.602079915161934e-05}
{"train_loss": 0.4211407005786896, "global_step": 35141, "epoch": 394, "lr": 9.60205725060501e-05}
{"train_loss": 0.25147247314453125, "global_step": 35142, "epoch": 394, "lr": 9.602034585429394e-05}
{"train_loss": 0.37258636951446533, "global_step": 35143, "epoch": 394, "lr": 9.60201191963509e-05}
{"train_loss": 0.33163943886756897, "global_step": 35144, "epoch": 394, "lr": 9.601989253222101e-05}
{"train_loss": 0.46471476554870605, "global_step": 35145, "epoch": 394, "lr": 9.601966586190431e-05}
{"train_loss": 0.473923921585083, "global_step": 35146, "epoch": 394, "lr": 9.601943918540083e-05}
{"train_loss": 0.4112682044506073, "global_step": 35147, "epoch": 394, "lr": 9.601921250271062e-05}
{"train_loss": 0.2890009582042694, "global_step": 35148, "epoch": 394, "lr": 9.601898581383366e-05}
{"train_loss": 0.33122533559799194, "global_step": 35149, "epoch": 394, "lr": 9.601875911877002e-05}
{"train_loss": 0.3379415273666382, "global_step": 35150, "epoch": 394, "lr": 9.601853241751971e-05}
{"train_loss": 0.3319311738014221, "global_step": 35151, "epoch": 394, "lr": 9.601830571008277e-05}
{"train_loss": 0.27484652400016785, "global_step": 35152, "epoch": 394, "lr": 9.601807899645924e-05}
{"train_loss": 0.3746483027935028, "global_step": 35153, "epoch": 394, "lr": 9.601785227664913e-05}
{"train_loss": 0.34622734148850604, "global_step": 35154, "epoch": 394, "lr": 9.601762555065248e-05, "val_loss": 2.4838080406188965}
{"train_loss": 0.4038662612438202, "global_step": 35155, "epoch": 395, "lr": 9.601739881846934e-05}
{"train_loss": 0.3274480998516083, "global_step": 35156, "epoch": 395, "lr": 9.601717208009971e-05}
{"train_loss": 0.34775418043136597, "global_step": 35157, "epoch": 395, "lr": 9.601694533554362e-05}
{"train_loss": 0.39410823583602905, "global_step": 35158, "epoch": 395, "lr": 9.601671858480113e-05}
{"train_loss": 0.3393738865852356, "global_step": 35159, "epoch": 395, "lr": 9.601649182787224e-05}
{"train_loss": 0.31336668133735657, "global_step": 35160, "epoch": 395, "lr": 9.601626506475698e-05}
{"train_loss": 0.2413911521434784, "global_step": 35161, "epoch": 395, "lr": 9.601603829545543e-05}
{"train_loss": 0.3043071925640106, "global_step": 35162, "epoch": 395, "lr": 9.601581151996755e-05}
{"train_loss": 0.3331226706504822, "global_step": 35163, "epoch": 395, "lr": 9.601558473829342e-05}
{"train_loss": 0.3514527678489685, "global_step": 35164, "epoch": 395, "lr": 9.601535795043307e-05}
{"train_loss": 0.30332061648368835, "global_step": 35165, "epoch": 395, "lr": 9.60151311563865e-05}
{"train_loss": 0.2596290409564972, "global_step": 35166, "epoch": 395, "lr": 9.601490435615377e-05}
{"train_loss": 0.48026517033576965, "global_step": 35167, "epoch": 395, "lr": 9.601467754973488e-05}
{"train_loss": 0.3391794264316559, "global_step": 35168, "epoch": 395, "lr": 9.601445073712988e-05}
{"train_loss": 0.32070690393447876, "global_step": 35169, "epoch": 395, "lr": 9.60142239183388e-05}
{"train_loss": 0.3723451793193817, "global_step": 35170, "epoch": 395, "lr": 9.601399709336167e-05}
{"train_loss": 0.512844443321228, "global_step": 35171, "epoch": 395, "lr": 9.601377026219852e-05}
{"train_loss": 0.3493776023387909, "global_step": 35172, "epoch": 395, "lr": 9.601354342484938e-05}
{"train_loss": 0.3081328868865967, "global_step": 35173, "epoch": 395, "lr": 9.601331658131427e-05}
{"train_loss": 0.43862184882164, "global_step": 35174, "epoch": 395, "lr": 9.601308973159324e-05}
{"train_loss": 0.2969622313976288, "global_step": 35175, "epoch": 395, "lr": 9.601286287568631e-05}
{"train_loss": 0.32842808961868286, "global_step": 35176, "epoch": 395, "lr": 9.601263601359352e-05}
{"train_loss": 0.45192745327949524, "global_step": 35177, "epoch": 395, "lr": 9.601240914531488e-05}
{"train_loss": 0.34250608086586, "global_step": 35178, "epoch": 395, "lr": 9.601218227085043e-05}
{"train_loss": 0.4049164056777954, "global_step": 35179, "epoch": 395, "lr": 9.601195539020022e-05}
{"train_loss": 0.34339019656181335, "global_step": 35180, "epoch": 395, "lr": 9.601172850336424e-05}
{"train_loss": 0.31042370200157166, "global_step": 35181, "epoch": 395, "lr": 9.601150161034256e-05}
{"train_loss": 0.2887708246707916, "global_step": 35182, "epoch": 395, "lr": 9.60112747111352e-05}
{"train_loss": 0.3023056983947754, "global_step": 35183, "epoch": 395, "lr": 9.601104780574217e-05}
{"train_loss": 0.27821096777915955, "global_step": 35184, "epoch": 395, "lr": 9.601082089416353e-05}
{"train_loss": 0.3311040699481964, "global_step": 35185, "epoch": 395, "lr": 9.601059397639927e-05}
{"train_loss": 0.3831334412097931, "global_step": 35186, "epoch": 395, "lr": 9.601036705244947e-05}
{"train_loss": 0.41697558760643005, "global_step": 35187, "epoch": 395, "lr": 9.601014012231413e-05}
{"train_loss": 0.26298144459724426, "global_step": 35188, "epoch": 395, "lr": 9.600991318599329e-05}
{"train_loss": 0.3619433641433716, "global_step": 35189, "epoch": 395, "lr": 9.600968624348697e-05}
{"train_loss": 0.2734294831752777, "global_step": 35190, "epoch": 395, "lr": 9.600945929479521e-05}
{"train_loss": 0.3253808915615082, "global_step": 35191, "epoch": 395, "lr": 9.600923233991805e-05}
{"train_loss": 0.3649056851863861, "global_step": 35192, "epoch": 395, "lr": 9.60090053788555e-05}
{"train_loss": 0.28384530544281006, "global_step": 35193, "epoch": 395, "lr": 9.60087784116076e-05}
{"train_loss": 0.26899418234825134, "global_step": 35194, "epoch": 395, "lr": 9.600855143817437e-05}
{"train_loss": 0.28455737233161926, "global_step": 35195, "epoch": 395, "lr": 9.600832445855588e-05}
{"train_loss": 0.3321535885334015, "global_step": 35196, "epoch": 395, "lr": 9.600809747275211e-05}
{"train_loss": 0.33691301941871643, "global_step": 35197, "epoch": 395, "lr": 9.600787048076312e-05}
{"train_loss": 0.3150297701358795, "global_step": 35198, "epoch": 395, "lr": 9.600764348258893e-05}
{"train_loss": 0.22873643040657043, "global_step": 35199, "epoch": 395, "lr": 9.600741647822957e-05}
{"train_loss": 0.3877604305744171, "global_step": 35200, "epoch": 395, "lr": 9.600718946768508e-05}
{"train_loss": 0.3262677788734436, "global_step": 35201, "epoch": 395, "lr": 9.600696245095548e-05}
{"train_loss": 0.44396281242370605, "global_step": 35202, "epoch": 395, "lr": 9.60067354280408e-05}
{"train_loss": 0.4071138799190521, "global_step": 35203, "epoch": 395, "lr": 9.600650839894108e-05}
{"train_loss": 0.2999749481678009, "global_step": 35204, "epoch": 395, "lr": 9.600628136365635e-05}
{"train_loss": 0.2920100688934326, "global_step": 35205, "epoch": 395, "lr": 9.600605432218663e-05}
{"train_loss": 0.42140644788742065, "global_step": 35206, "epoch": 395, "lr": 9.600582727453196e-05}
{"train_loss": 0.30311885476112366, "global_step": 35207, "epoch": 395, "lr": 9.600560022069238e-05}
{"train_loss": 0.4535008668899536, "global_step": 35208, "epoch": 395, "lr": 9.600537316066789e-05}
{"train_loss": 0.3264053165912628, "global_step": 35209, "epoch": 395, "lr": 9.600514609445854e-05}
{"train_loss": 0.34030622243881226, "global_step": 35210, "epoch": 395, "lr": 9.600491902206437e-05}
{"train_loss": 0.33380940556526184, "global_step": 35211, "epoch": 395, "lr": 9.600469194348538e-05}
{"train_loss": 0.3325074315071106, "global_step": 35212, "epoch": 395, "lr": 9.600446485872164e-05}
{"train_loss": 0.3895484209060669, "global_step": 35213, "epoch": 395, "lr": 9.600423776777315e-05}
{"train_loss": 0.2520124018192291, "global_step": 35214, "epoch": 395, "lr": 9.600401067063995e-05}
{"train_loss": 0.3373042941093445, "global_step": 35215, "epoch": 395, "lr": 9.600378356732207e-05}
{"train_loss": 0.3259715139865875, "global_step": 35216, "epoch": 395, "lr": 9.600355645781953e-05}
{"train_loss": 0.4286791980266571, "global_step": 35217, "epoch": 395, "lr": 9.600332934213239e-05}
{"train_loss": 0.40646716952323914, "global_step": 35218, "epoch": 395, "lr": 9.600310222026068e-05}
{"train_loss": 0.3472004532814026, "global_step": 35219, "epoch": 395, "lr": 9.600287509220438e-05}
{"train_loss": 0.37834614515304565, "global_step": 35220, "epoch": 395, "lr": 9.600264795796358e-05}
{"train_loss": 0.21771475672721863, "global_step": 35221, "epoch": 395, "lr": 9.600242081753827e-05}
{"train_loss": 0.32783636450767517, "global_step": 35222, "epoch": 395, "lr": 9.600219367092849e-05}
{"train_loss": 0.32829391956329346, "global_step": 35223, "epoch": 395, "lr": 9.600196651813429e-05}
{"train_loss": 0.3786861002445221, "global_step": 35224, "epoch": 395, "lr": 9.600173935915569e-05}
{"train_loss": 0.45971745252609253, "global_step": 35225, "epoch": 395, "lr": 9.60015121939927e-05}
{"train_loss": 0.33181819319725037, "global_step": 35226, "epoch": 395, "lr": 9.600128502264537e-05}
{"train_loss": 0.24843454360961914, "global_step": 35227, "epoch": 395, "lr": 9.600105784511373e-05}
{"train_loss": 0.2786097526550293, "global_step": 35228, "epoch": 395, "lr": 9.600083066139782e-05}
{"train_loss": 0.2278602123260498, "global_step": 35229, "epoch": 395, "lr": 9.600060347149764e-05}
{"train_loss": 0.2796608507633209, "global_step": 35230, "epoch": 395, "lr": 9.600037627541326e-05}
{"train_loss": 0.36554980278015137, "global_step": 35231, "epoch": 395, "lr": 9.600014907314467e-05}
{"train_loss": 0.3458336293697357, "global_step": 35232, "epoch": 395, "lr": 9.599992186469194e-05}
{"train_loss": 0.4426649808883667, "global_step": 35233, "epoch": 395, "lr": 9.599969465005507e-05}
{"train_loss": 0.3650486469268799, "global_step": 35234, "epoch": 395, "lr": 9.59994674292341e-05}
{"train_loss": 0.3006795346736908, "global_step": 35235, "epoch": 395, "lr": 9.599924020222907e-05}
{"train_loss": 0.4343099296092987, "global_step": 35236, "epoch": 395, "lr": 9.599901296903999e-05}
{"train_loss": 0.35902315378189087, "global_step": 35237, "epoch": 395, "lr": 9.599878572966692e-05}
{"train_loss": 0.40258729457855225, "global_step": 35238, "epoch": 395, "lr": 9.599855848410985e-05}
{"train_loss": 0.4109627306461334, "global_step": 35239, "epoch": 395, "lr": 9.599833123236886e-05}
{"train_loss": 0.48294728994369507, "global_step": 35240, "epoch": 395, "lr": 9.599810397444395e-05}
{"train_loss": 0.4619775116443634, "global_step": 35241, "epoch": 395, "lr": 9.599787671033514e-05}
{"train_loss": 0.4879979193210602, "global_step": 35242, "epoch": 395, "lr": 9.59976494400425e-05}
{"train_loss": 0.3478946150018928, "global_step": 35243, "epoch": 395, "lr": 9.599742216356602e-05, "val_loss": 2.5199856758117676, "train_action_mse_error": 24.518634796142578}
{"train_loss": 0.4083946645259857, "global_step": 35244, "epoch": 396, "lr": 9.599719488090575e-05}
{"train_loss": 0.3120722472667694, "global_step": 35245, "epoch": 396, "lr": 9.599696759206171e-05}
{"train_loss": 0.3190849721431732, "global_step": 35246, "epoch": 396, "lr": 9.599674029703395e-05}
{"train_loss": 0.3289735019207001, "global_step": 35247, "epoch": 396, "lr": 9.599651299582249e-05}
{"train_loss": 0.491779088973999, "global_step": 35248, "epoch": 396, "lr": 9.599628568842736e-05}
{"train_loss": 0.4041922688484192, "global_step": 35249, "epoch": 396, "lr": 9.599605837484857e-05}
{"train_loss": 0.36347392201423645, "global_step": 35250, "epoch": 396, "lr": 9.599583105508619e-05}
{"train_loss": 0.45572781562805176, "global_step": 35251, "epoch": 396, "lr": 9.599560372914023e-05}
{"train_loss": 0.3809491693973541, "global_step": 35252, "epoch": 396, "lr": 9.599537639701072e-05}
{"train_loss": 0.3852846026420593, "global_step": 35253, "epoch": 396, "lr": 9.599514905869768e-05}
{"train_loss": 0.39863121509552, "global_step": 35254, "epoch": 396, "lr": 9.599492171420117e-05}
{"train_loss": 0.4064694046974182, "global_step": 35255, "epoch": 396, "lr": 9.599469436352119e-05}
{"train_loss": 0.4267929792404175, "global_step": 35256, "epoch": 396, "lr": 9.59944670066578e-05}
{"train_loss": 0.36923250555992126, "global_step": 35257, "epoch": 396, "lr": 9.5994239643611e-05}
{"train_loss": 0.39659982919692993, "global_step": 35258, "epoch": 396, "lr": 9.599401227438086e-05}
{"train_loss": 0.4768109619617462, "global_step": 35259, "epoch": 396, "lr": 9.599378489896735e-05}
{"train_loss": 0.34393569827079773, "global_step": 35260, "epoch": 396, "lr": 9.599355751737055e-05}
{"train_loss": 0.35783833265304565, "global_step": 35261, "epoch": 396, "lr": 9.599333012959048e-05}
{"train_loss": 0.3431903123855591, "global_step": 35262, "epoch": 396, "lr": 9.599310273562718e-05}
{"train_loss": 0.5168783664703369, "global_step": 35263, "epoch": 396, "lr": 9.599287533548065e-05}
{"train_loss": 0.4258241355419159, "global_step": 35264, "epoch": 396, "lr": 9.599264792915094e-05}
{"train_loss": 0.2997983992099762, "global_step": 35265, "epoch": 396, "lr": 9.599242051663808e-05}
{"train_loss": 0.40861770510673523, "global_step": 35266, "epoch": 396, "lr": 9.59921930979421e-05}
{"train_loss": 0.3220837414264679, "global_step": 35267, "epoch": 396, "lr": 9.599196567306303e-05}
{"train_loss": 0.39236775040626526, "global_step": 35268, "epoch": 396, "lr": 9.59917382420009e-05}
{"train_loss": 0.350754052400589, "global_step": 35269, "epoch": 396, "lr": 9.599151080475575e-05}
{"train_loss": 0.3175801932811737, "global_step": 35270, "epoch": 396, "lr": 9.599128336132759e-05}
{"train_loss": 0.46670040488243103, "global_step": 35271, "epoch": 396, "lr": 9.599105591171647e-05}
{"train_loss": 0.4145978093147278, "global_step": 35272, "epoch": 396, "lr": 9.599082845592242e-05}
{"train_loss": 0.3012785315513611, "global_step": 35273, "epoch": 396, "lr": 9.599060099394545e-05}
{"train_loss": 0.33182671666145325, "global_step": 35274, "epoch": 396, "lr": 9.599037352578561e-05}
{"train_loss": 0.42607855796813965, "global_step": 35275, "epoch": 396, "lr": 9.599014605144292e-05}
{"train_loss": 0.44294795393943787, "global_step": 35276, "epoch": 396, "lr": 9.598991857091741e-05}
{"train_loss": 0.41555315256118774, "global_step": 35277, "epoch": 396, "lr": 9.598969108420914e-05}
{"train_loss": 0.3723953068256378, "global_step": 35278, "epoch": 396, "lr": 9.598946359131809e-05}
{"train_loss": 0.3678484857082367, "global_step": 35279, "epoch": 396, "lr": 9.598923609224434e-05}
{"train_loss": 0.3345944285392761, "global_step": 35280, "epoch": 396, "lr": 9.598900858698788e-05}
{"train_loss": 0.32930636405944824, "global_step": 35281, "epoch": 396, "lr": 9.598878107554877e-05}
{"train_loss": 0.30341437458992004, "global_step": 35282, "epoch": 396, "lr": 9.598855355792703e-05}
{"train_loss": 0.4187384247779846, "global_step": 35283, "epoch": 396, "lr": 9.598832603412268e-05}
{"train_loss": 0.38739287853240967, "global_step": 35284, "epoch": 396, "lr": 9.598809850413576e-05}
{"train_loss": 0.3628697991371155, "global_step": 35285, "epoch": 396, "lr": 9.598787096796632e-05}
{"train_loss": 0.362049400806427, "global_step": 35286, "epoch": 396, "lr": 9.598764342561434e-05}
{"train_loss": 0.4708086848258972, "global_step": 35287, "epoch": 396, "lr": 9.59874158770799e-05}
{"train_loss": 0.25863978266716003, "global_step": 35288, "epoch": 396, "lr": 9.598718832236301e-05}
{"train_loss": 0.3282482326030731, "global_step": 35289, "epoch": 396, "lr": 9.59869607614637e-05}
{"train_loss": 0.5013827085494995, "global_step": 35290, "epoch": 396, "lr": 9.598673319438201e-05}
{"train_loss": 0.4129084348678589, "global_step": 35291, "epoch": 396, "lr": 9.598650562111796e-05}
{"train_loss": 0.37391015887260437, "global_step": 35292, "epoch": 396, "lr": 9.598627804167158e-05}
{"train_loss": 0.35955390334129333, "global_step": 35293, "epoch": 396, "lr": 9.598605045604292e-05}
{"train_loss": 0.28464430570602417, "global_step": 35294, "epoch": 396, "lr": 9.598582286423199e-05}
{"train_loss": 0.32804393768310547, "global_step": 35295, "epoch": 396, "lr": 9.598559526623882e-05}
{"train_loss": 0.25637558102607727, "global_step": 35296, "epoch": 396, "lr": 9.598536766206346e-05}
{"train_loss": 0.26456740498542786, "global_step": 35297, "epoch": 396, "lr": 9.59851400517059e-05}
{"train_loss": 0.2974476218223572, "global_step": 35298, "epoch": 396, "lr": 9.598491243516624e-05}
{"train_loss": 0.35373082756996155, "global_step": 35299, "epoch": 396, "lr": 9.598468481244443e-05}
{"train_loss": 0.37475624680519104, "global_step": 35300, "epoch": 396, "lr": 9.598445718354056e-05}
{"train_loss": 0.31084591150283813, "global_step": 35301, "epoch": 396, "lr": 9.598422954845463e-05}
{"train_loss": 0.33882659673690796, "global_step": 35302, "epoch": 396, "lr": 9.598400190718668e-05}
{"train_loss": 0.33306676149368286, "global_step": 35303, "epoch": 396, "lr": 9.598377425973675e-05}
{"train_loss": 0.32188478112220764, "global_step": 35304, "epoch": 396, "lr": 9.598354660610487e-05}
{"train_loss": 0.3012668490409851, "global_step": 35305, "epoch": 396, "lr": 9.598331894629104e-05}
{"train_loss": 0.49274611473083496, "global_step": 35306, "epoch": 396, "lr": 9.598309128029534e-05}
{"train_loss": 0.32789021730422974, "global_step": 35307, "epoch": 396, "lr": 9.598286360811775e-05}
{"train_loss": 0.29483360052108765, "global_step": 35308, "epoch": 396, "lr": 9.598263592975833e-05}
{"train_loss": 0.34787148237228394, "global_step": 35309, "epoch": 396, "lr": 9.598240824521711e-05}
{"train_loss": 0.3514325022697449, "global_step": 35310, "epoch": 396, "lr": 9.598218055449413e-05}
{"train_loss": 0.30706825852394104, "global_step": 35311, "epoch": 396, "lr": 9.598195285758938e-05}
{"train_loss": 0.27241313457489014, "global_step": 35312, "epoch": 396, "lr": 9.598172515450293e-05}
{"train_loss": 0.27468404173851013, "global_step": 35313, "epoch": 396, "lr": 9.598149744523479e-05}
{"train_loss": 0.46118590235710144, "global_step": 35314, "epoch": 396, "lr": 9.5981269729785e-05}
{"train_loss": 0.24341490864753723, "global_step": 35315, "epoch": 396, "lr": 9.59810420081536e-05}
{"train_loss": 0.30844131112098694, "global_step": 35316, "epoch": 396, "lr": 9.59808142803406e-05}
{"train_loss": 0.27848950028419495, "global_step": 35317, "epoch": 396, "lr": 9.598058654634605e-05}
{"train_loss": 0.22585436701774597, "global_step": 35318, "epoch": 396, "lr": 9.598035880616996e-05}
{"train_loss": 0.3694525957107544, "global_step": 35319, "epoch": 396, "lr": 9.598013105981238e-05}
{"train_loss": 0.37783941626548767, "global_step": 35320, "epoch": 396, "lr": 9.597990330727331e-05}
{"train_loss": 0.4689801335334778, "global_step": 35321, "epoch": 396, "lr": 9.597967554855283e-05}
{"train_loss": 0.36348727345466614, "global_step": 35322, "epoch": 396, "lr": 9.597944778365092e-05}
{"train_loss": 0.31253644824028015, "global_step": 35323, "epoch": 396, "lr": 9.597922001256765e-05}
{"train_loss": 0.3212404251098633, "global_step": 35324, "epoch": 396, "lr": 9.597899223530302e-05}
{"train_loss": 0.36512792110443115, "global_step": 35325, "epoch": 396, "lr": 9.597876445185709e-05}
{"train_loss": 0.395355224609375, "global_step": 35326, "epoch": 396, "lr": 9.597853666222987e-05}
{"train_loss": 0.37048566341400146, "global_step": 35327, "epoch": 396, "lr": 9.59783088664214e-05}
{"train_loss": 0.44082245230674744, "global_step": 35328, "epoch": 396, "lr": 9.597808106443169e-05}
{"train_loss": 0.38374099135398865, "global_step": 35329, "epoch": 396, "lr": 9.59778532562608e-05}
{"train_loss": 0.3894353210926056, "global_step": 35330, "epoch": 396, "lr": 9.597762544190876e-05}
{"train_loss": 0.36047568917274475, "global_step": 35331, "epoch": 396, "lr": 9.597739762137557e-05}
{"train_loss": 0.3645368161496152, "global_step": 35332, "epoch": 396, "lr": 9.597716979466129e-05, "val_loss": 2.529876947402954}
{"train_loss": 0.33314380049705505, "global_step": 35333, "epoch": 397, "lr": 9.597694196176593e-05}
{"train_loss": 0.33706629276275635, "global_step": 35334, "epoch": 397, "lr": 9.597671412268953e-05}
{"train_loss": 0.3542454242706299, "global_step": 35335, "epoch": 397, "lr": 9.597648627743213e-05}
{"train_loss": 0.3639366924762726, "global_step": 35336, "epoch": 397, "lr": 9.597625842599376e-05}
{"train_loss": 0.2835729718208313, "global_step": 35337, "epoch": 397, "lr": 9.597603056837443e-05}
{"train_loss": 0.2808811068534851, "global_step": 35338, "epoch": 397, "lr": 9.597580270457418e-05}
{"train_loss": 0.37267759442329407, "global_step": 35339, "epoch": 397, "lr": 9.597557483459305e-05}
{"train_loss": 0.34672653675079346, "global_step": 35340, "epoch": 397, "lr": 9.597534695843105e-05}
{"train_loss": 0.3626555800437927, "global_step": 35341, "epoch": 397, "lr": 9.597511907608825e-05}
{"train_loss": 0.3079696595668793, "global_step": 35342, "epoch": 397, "lr": 9.597489118756463e-05}
{"train_loss": 0.49718141555786133, "global_step": 35343, "epoch": 397, "lr": 9.597466329286027e-05}
{"train_loss": 0.2869139015674591, "global_step": 35344, "epoch": 397, "lr": 9.597443539197517e-05}
{"train_loss": 0.45198312401771545, "global_step": 35345, "epoch": 397, "lr": 9.597420748490935e-05}
{"train_loss": 0.35205987095832825, "global_step": 35346, "epoch": 397, "lr": 9.597397957166288e-05}
{"train_loss": 0.4704146385192871, "global_step": 35347, "epoch": 397, "lr": 9.597375165223576e-05}
{"train_loss": 0.4228445887565613, "global_step": 35348, "epoch": 397, "lr": 9.597352372662802e-05}
{"train_loss": 0.38554713129997253, "global_step": 35349, "epoch": 397, "lr": 9.597329579483971e-05}
{"train_loss": 0.3244566023349762, "global_step": 35350, "epoch": 397, "lr": 9.597306785687086e-05}
{"train_loss": 0.3343586027622223, "global_step": 35351, "epoch": 397, "lr": 9.597283991272148e-05}
{"train_loss": 0.2908861041069031, "global_step": 35352, "epoch": 397, "lr": 9.597261196239162e-05}
{"train_loss": 0.4758712351322174, "global_step": 35353, "epoch": 397, "lr": 9.597238400588129e-05}
{"train_loss": 0.36234748363494873, "global_step": 35354, "epoch": 397, "lr": 9.597215604319054e-05}
{"train_loss": 0.29618707299232483, "global_step": 35355, "epoch": 397, "lr": 9.59719280743194e-05}
{"train_loss": 0.3501724302768707, "global_step": 35356, "epoch": 397, "lr": 9.597170009926789e-05}
{"train_loss": 0.3134428858757019, "global_step": 35357, "epoch": 397, "lr": 9.597147211803604e-05}
{"train_loss": 0.3482440114021301, "global_step": 35358, "epoch": 397, "lr": 9.597124413062389e-05}
{"train_loss": 0.3266805410385132, "global_step": 35359, "epoch": 397, "lr": 9.597101613703146e-05}
{"train_loss": 0.4761120080947876, "global_step": 35360, "epoch": 397, "lr": 9.597078813725879e-05}
{"train_loss": 0.33195576071739197, "global_step": 35361, "epoch": 397, "lr": 9.597056013130593e-05}
{"train_loss": 0.3596973121166229, "global_step": 35362, "epoch": 397, "lr": 9.597033211917285e-05}
{"train_loss": 0.2712904214859009, "global_step": 35363, "epoch": 397, "lr": 9.597010410085965e-05}
{"train_loss": 0.35262641310691833, "global_step": 35364, "epoch": 397, "lr": 9.596987607636632e-05}
{"train_loss": 0.3313671946525574, "global_step": 35365, "epoch": 397, "lr": 9.596964804569289e-05}
{"train_loss": 0.35261234641075134, "global_step": 35366, "epoch": 397, "lr": 9.596942000883942e-05}
{"train_loss": 0.3627361059188843, "global_step": 35367, "epoch": 397, "lr": 9.59691919658059e-05}
{"train_loss": 0.3852009177207947, "global_step": 35368, "epoch": 397, "lr": 9.59689639165924e-05}
{"train_loss": 0.3667484223842621, "global_step": 35369, "epoch": 397, "lr": 9.596873586119892e-05}
{"train_loss": 0.31497424840927124, "global_step": 35370, "epoch": 397, "lr": 9.596850779962552e-05}
{"train_loss": 0.37698373198509216, "global_step": 35371, "epoch": 397, "lr": 9.59682797318722e-05}
{"train_loss": 0.34925422072410583, "global_step": 35372, "epoch": 397, "lr": 9.596805165793901e-05}
{"train_loss": 0.3532218337059021, "global_step": 35373, "epoch": 397, "lr": 9.596782357782597e-05}
{"train_loss": 0.3523502051830292, "global_step": 35374, "epoch": 397, "lr": 9.596759549153313e-05}
{"train_loss": 0.2924731969833374, "global_step": 35375, "epoch": 397, "lr": 9.596736739906049e-05}
{"train_loss": 0.41144514083862305, "global_step": 35376, "epoch": 397, "lr": 9.596713930040812e-05}
{"train_loss": 0.41220927238464355, "global_step": 35377, "epoch": 397, "lr": 9.596691119557602e-05}
{"train_loss": 0.2703329920768738, "global_step": 35378, "epoch": 397, "lr": 9.596668308456422e-05}
{"train_loss": 0.32340794801712036, "global_step": 35379, "epoch": 397, "lr": 9.596645496737276e-05}
{"train_loss": 0.2610841989517212, "global_step": 35380, "epoch": 397, "lr": 9.596622684400167e-05}
{"train_loss": 0.30192700028419495, "global_step": 35381, "epoch": 397, "lr": 9.5965998714451e-05}
{"train_loss": 0.5197311043739319, "global_step": 35382, "epoch": 397, "lr": 9.596577057872074e-05}
{"train_loss": 0.3613661825656891, "global_step": 35383, "epoch": 397, "lr": 9.596554243681096e-05}
{"train_loss": 0.3207989037036896, "global_step": 35384, "epoch": 397, "lr": 9.596531428872166e-05}
{"train_loss": 0.41869428753852844, "global_step": 35385, "epoch": 397, "lr": 9.596508613445288e-05}
{"train_loss": 0.32487401366233826, "global_step": 35386, "epoch": 397, "lr": 9.596485797400466e-05}
{"train_loss": 0.40875253081321716, "global_step": 35387, "epoch": 397, "lr": 9.596462980737703e-05}
{"train_loss": 0.3394066393375397, "global_step": 35388, "epoch": 397, "lr": 9.596440163457001e-05}
{"train_loss": 0.24833926558494568, "global_step": 35389, "epoch": 397, "lr": 9.596417345558364e-05}
{"train_loss": 0.4644336700439453, "global_step": 35390, "epoch": 397, "lr": 9.596394527041795e-05}
{"train_loss": 0.3568354547023773, "global_step": 35391, "epoch": 397, "lr": 9.596371707907296e-05}
{"train_loss": 0.3314052224159241, "global_step": 35392, "epoch": 397, "lr": 9.596348888154872e-05}
{"train_loss": 0.38124212622642517, "global_step": 35393, "epoch": 397, "lr": 9.596326067784522e-05}
{"train_loss": 0.3348259925842285, "global_step": 35394, "epoch": 397, "lr": 9.596303246796254e-05}
{"train_loss": 0.2854006886482239, "global_step": 35395, "epoch": 397, "lr": 9.59628042519007e-05}
{"train_loss": 0.40741392970085144, "global_step": 35396, "epoch": 397, "lr": 9.596257602965972e-05}
{"train_loss": 0.4352937936782837, "global_step": 35397, "epoch": 397, "lr": 9.596234780123963e-05}
{"train_loss": 0.34712231159210205, "global_step": 35398, "epoch": 397, "lr": 9.596211956664046e-05}
{"train_loss": 0.41787219047546387, "global_step": 35399, "epoch": 397, "lr": 9.596189132586222e-05}
{"train_loss": 0.2984335124492645, "global_step": 35400, "epoch": 397, "lr": 9.5961663078905e-05}
{"train_loss": 0.36891067028045654, "global_step": 35401, "epoch": 397, "lr": 9.596143482576877e-05}
{"train_loss": 0.33249324560165405, "global_step": 35402, "epoch": 397, "lr": 9.59612065664536e-05}
{"train_loss": 0.3623064458370209, "global_step": 35403, "epoch": 397, "lr": 9.59609783009595e-05}
{"train_loss": 0.48822516202926636, "global_step": 35404, "epoch": 397, "lr": 9.59607500292865e-05}
{"train_loss": 0.4426420032978058, "global_step": 35405, "epoch": 397, "lr": 9.596052175143464e-05}
{"train_loss": 0.287239134311676, "global_step": 35406, "epoch": 397, "lr": 9.596029346740397e-05}
{"train_loss": 0.4236328601837158, "global_step": 35407, "epoch": 397, "lr": 9.596006517719448e-05}
{"train_loss": 0.4014894962310791, "global_step": 35408, "epoch": 397, "lr": 9.595983688080621e-05}
{"train_loss": 0.3311014473438263, "global_step": 35409, "epoch": 397, "lr": 9.59596085782392e-05}
{"train_loss": 0.3096167743206024, "global_step": 35410, "epoch": 397, "lr": 9.595938026949349e-05}
{"train_loss": 0.41846588253974915, "global_step": 35411, "epoch": 397, "lr": 9.59591519545691e-05}
{"train_loss": 0.3650326132774353, "global_step": 35412, "epoch": 397, "lr": 9.595892363346607e-05}
{"train_loss": 0.45582541823387146, "global_step": 35413, "epoch": 397, "lr": 9.595869530618442e-05}
{"train_loss": 0.31665119528770447, "global_step": 35414, "epoch": 397, "lr": 9.595846697272418e-05}
{"train_loss": 0.3770380914211273, "global_step": 35415, "epoch": 397, "lr": 9.59582386330854e-05}
{"train_loss": 0.4111415147781372, "global_step": 35416, "epoch": 397, "lr": 9.595801028726807e-05}
{"train_loss": 0.3768554925918579, "global_step": 35417, "epoch": 397, "lr": 9.595778193527224e-05}
{"train_loss": 0.2849788963794708, "global_step": 35418, "epoch": 397, "lr": 9.595755357709798e-05}
{"train_loss": 0.2737294137477875, "global_step": 35419, "epoch": 397, "lr": 9.595732521274527e-05}
{"train_loss": 0.394544392824173, "global_step": 35420, "epoch": 397, "lr": 9.595709684221416e-05}
{"train_loss": 0.35956704516089366, "global_step": 35421, "epoch": 397, "lr": 9.595686846550467e-05, "val_loss": 2.496330976486206}
{"train_loss": 0.33993595838546753, "global_step": 35422, "epoch": 398, "lr": 9.595664008261685e-05}
{"train_loss": 0.28463852405548096, "global_step": 35423, "epoch": 398, "lr": 9.595641169355071e-05}
{"train_loss": 0.44904741644859314, "global_step": 35424, "epoch": 398, "lr": 9.595618329830629e-05}
{"train_loss": 0.43999770283699036, "global_step": 35425, "epoch": 398, "lr": 9.595595489688362e-05}
{"train_loss": 0.37504610419273376, "global_step": 35426, "epoch": 398, "lr": 9.595572648928275e-05}
{"train_loss": 0.2962827980518341, "global_step": 35427, "epoch": 398, "lr": 9.595549807550368e-05}
{"train_loss": 0.38748133182525635, "global_step": 35428, "epoch": 398, "lr": 9.595526965554647e-05}
{"train_loss": 0.40551209449768066, "global_step": 35429, "epoch": 398, "lr": 9.59550412294111e-05}
{"train_loss": 0.3715876042842865, "global_step": 35430, "epoch": 398, "lr": 9.595481279709766e-05}
{"train_loss": 0.4334102272987366, "global_step": 35431, "epoch": 398, "lr": 9.595458435860615e-05}
{"train_loss": 0.36192673444747925, "global_step": 35432, "epoch": 398, "lr": 9.59543559139366e-05}
{"train_loss": 0.30525830388069153, "global_step": 35433, "epoch": 398, "lr": 9.595412746308904e-05}
{"train_loss": 0.25966450572013855, "global_step": 35434, "epoch": 398, "lr": 9.595389900606353e-05}
{"train_loss": 0.3230384290218353, "global_step": 35435, "epoch": 398, "lr": 9.595367054286007e-05}
{"train_loss": 0.2859838902950287, "global_step": 35436, "epoch": 398, "lr": 9.595344207347869e-05}
{"train_loss": 0.3581395149230957, "global_step": 35437, "epoch": 398, "lr": 9.595321359791944e-05}
{"train_loss": 0.4311639666557312, "global_step": 35438, "epoch": 398, "lr": 9.595298511618235e-05}
{"train_loss": 0.4123447835445404, "global_step": 35439, "epoch": 398, "lr": 9.595275662826742e-05}
{"train_loss": 0.34541580080986023, "global_step": 35440, "epoch": 398, "lr": 9.595252813417472e-05}
{"train_loss": 0.31989508867263794, "global_step": 35441, "epoch": 398, "lr": 9.595229963390424e-05}
{"train_loss": 0.36263275146484375, "global_step": 35442, "epoch": 398, "lr": 9.595207112745605e-05}
{"train_loss": 0.3012811541557312, "global_step": 35443, "epoch": 398, "lr": 9.595184261483017e-05}
{"train_loss": 0.3836981952190399, "global_step": 35444, "epoch": 398, "lr": 9.595161409602662e-05}
{"train_loss": 0.27219340205192566, "global_step": 35445, "epoch": 398, "lr": 9.595138557104543e-05}
{"train_loss": 0.24440523982048035, "global_step": 35446, "epoch": 398, "lr": 9.595115703988665e-05}
{"train_loss": 0.30835267901420593, "global_step": 35447, "epoch": 398, "lr": 9.595092850255028e-05}
{"train_loss": 0.38507241010665894, "global_step": 35448, "epoch": 398, "lr": 9.595069995903636e-05}
{"train_loss": 0.33804893493652344, "global_step": 35449, "epoch": 398, "lr": 9.595047140934494e-05}
{"train_loss": 0.3234730660915375, "global_step": 35450, "epoch": 398, "lr": 9.595024285347605e-05}
{"train_loss": 0.38256192207336426, "global_step": 35451, "epoch": 398, "lr": 9.59500142914297e-05}
{"train_loss": 0.3528823256492615, "global_step": 35452, "epoch": 398, "lr": 9.594978572320593e-05}
{"train_loss": 0.35789141058921814, "global_step": 35453, "epoch": 398, "lr": 9.594955714880475e-05}
{"train_loss": 0.39622703194618225, "global_step": 35454, "epoch": 398, "lr": 9.594932856822626e-05}
{"train_loss": 0.30804312229156494, "global_step": 35455, "epoch": 398, "lr": 9.594909998147041e-05}
{"train_loss": 0.411118745803833, "global_step": 35456, "epoch": 398, "lr": 9.594887138853726e-05}
{"train_loss": 0.268558531999588, "global_step": 35457, "epoch": 398, "lr": 9.594864278942685e-05}
{"train_loss": 0.33915582299232483, "global_step": 35458, "epoch": 398, "lr": 9.594841418413923e-05}
{"train_loss": 0.21149951219558716, "global_step": 35459, "epoch": 398, "lr": 9.594818557267437e-05}
{"train_loss": 0.33517715334892273, "global_step": 35460, "epoch": 398, "lr": 9.594795695503235e-05}
{"train_loss": 0.29685646295547485, "global_step": 35461, "epoch": 398, "lr": 9.594772833121319e-05}
{"train_loss": 0.3006405830383301, "global_step": 35462, "epoch": 398, "lr": 9.594749970121692e-05}
{"train_loss": 0.39719200134277344, "global_step": 35463, "epoch": 398, "lr": 9.594727106504357e-05}
{"train_loss": 0.3008940815925598, "global_step": 35464, "epoch": 398, "lr": 9.594704242269317e-05}
{"train_loss": 0.2533872425556183, "global_step": 35465, "epoch": 398, "lr": 9.594681377416575e-05}
{"train_loss": 0.37056857347488403, "global_step": 35466, "epoch": 398, "lr": 9.594658511946133e-05}
{"train_loss": 0.31298938393592834, "global_step": 35467, "epoch": 398, "lr": 9.594635645857996e-05}
{"train_loss": 0.3531970977783203, "global_step": 35468, "epoch": 398, "lr": 9.594612779152166e-05}
{"train_loss": 0.37070128321647644, "global_step": 35469, "epoch": 398, "lr": 9.594589911828646e-05}
{"train_loss": 0.34520387649536133, "global_step": 35470, "epoch": 398, "lr": 9.59456704388744e-05}
{"train_loss": 0.2551938593387604, "global_step": 35471, "epoch": 398, "lr": 9.59454417532855e-05}
{"train_loss": 0.40240344405174255, "global_step": 35472, "epoch": 398, "lr": 9.59452130615198e-05}
{"train_loss": 0.3289889097213745, "global_step": 35473, "epoch": 398, "lr": 9.594498436357733e-05}
{"train_loss": 0.330716997385025, "global_step": 35474, "epoch": 398, "lr": 9.59447556594581e-05}
{"train_loss": 0.401108980178833, "global_step": 35475, "epoch": 398, "lr": 9.594452694916216e-05}
{"train_loss": 0.2735540270805359, "global_step": 35476, "epoch": 398, "lr": 9.594429823268955e-05}
{"train_loss": 0.4113645553588867, "global_step": 35477, "epoch": 398, "lr": 9.594406951004029e-05}
{"train_loss": 0.39774540066719055, "global_step": 35478, "epoch": 398, "lr": 9.594384078121441e-05}
{"train_loss": 0.2970222234725952, "global_step": 35479, "epoch": 398, "lr": 9.594361204621193e-05}
{"train_loss": 0.2538774311542511, "global_step": 35480, "epoch": 398, "lr": 9.59433833050329e-05}
{"train_loss": 0.42560288310050964, "global_step": 35481, "epoch": 398, "lr": 9.594315455767734e-05}
{"train_loss": 0.3571823537349701, "global_step": 35482, "epoch": 398, "lr": 9.594292580414528e-05}
{"train_loss": 0.3476768434047699, "global_step": 35483, "epoch": 398, "lr": 9.594269704443676e-05}
{"train_loss": 0.3254982829093933, "global_step": 35484, "epoch": 398, "lr": 9.59424682785518e-05}
{"train_loss": 0.3605697751045227, "global_step": 35485, "epoch": 398, "lr": 9.594223950649043e-05}
{"train_loss": 0.3517535924911499, "global_step": 35486, "epoch": 398, "lr": 9.594201072825269e-05}
{"train_loss": 0.2747536897659302, "global_step": 35487, "epoch": 398, "lr": 9.59417819438386e-05}
{"train_loss": 0.4057818353176117, "global_step": 35488, "epoch": 398, "lr": 9.594155315324822e-05}
{"train_loss": 0.42219749093055725, "global_step": 35489, "epoch": 398, "lr": 9.594132435648153e-05}
{"train_loss": 0.33790838718414307, "global_step": 35490, "epoch": 398, "lr": 9.59410955535386e-05}
{"train_loss": 0.38468530774116516, "global_step": 35491, "epoch": 398, "lr": 9.594086674441946e-05}
{"train_loss": 0.3508690297603607, "global_step": 35492, "epoch": 398, "lr": 9.594063792912412e-05}
{"train_loss": 0.399314820766449, "global_step": 35493, "epoch": 398, "lr": 9.594040910765263e-05}
{"train_loss": 0.33856964111328125, "global_step": 35494, "epoch": 398, "lr": 9.5940180280005e-05}
{"train_loss": 0.35501164197921753, "global_step": 35495, "epoch": 398, "lr": 9.593995144618127e-05}
{"train_loss": 0.24847573041915894, "global_step": 35496, "epoch": 398, "lr": 9.593972260618149e-05}
{"train_loss": 0.3684503138065338, "global_step": 35497, "epoch": 398, "lr": 9.593949376000566e-05}
{"train_loss": 0.42824840545654297, "global_step": 35498, "epoch": 398, "lr": 9.593926490765384e-05}
{"train_loss": 0.3294443190097809, "global_step": 35499, "epoch": 398, "lr": 9.593903604912604e-05}
{"train_loss": 0.30678436160087585, "global_step": 35500, "epoch": 398, "lr": 9.593880718442228e-05}
{"train_loss": 0.42446810007095337, "global_step": 35501, "epoch": 398, "lr": 9.593857831354263e-05}
{"train_loss": 0.3774294853210449, "global_step": 35502, "epoch": 398, "lr": 9.593834943648707e-05}
{"train_loss": 0.45755693316459656, "global_step": 35503, "epoch": 398, "lr": 9.593812055325568e-05}
{"train_loss": 0.3401530086994171, "global_step": 35504, "epoch": 398, "lr": 9.593789166384848e-05}
{"train_loss": 0.3316247761249542, "global_step": 35505, "epoch": 398, "lr": 9.593766276826547e-05}
{"train_loss": 0.3568035960197449, "global_step": 35506, "epoch": 398, "lr": 9.593743386650671e-05}
{"train_loss": 0.27548879384994507, "global_step": 35507, "epoch": 398, "lr": 9.593720495857221e-05}
{"train_loss": 0.31032904982566833, "global_step": 35508, "epoch": 398, "lr": 9.593697604446203e-05}
{"train_loss": 0.33637556433677673, "global_step": 35509, "epoch": 398, "lr": 9.593674712417618e-05}
{"train_loss": 0.34693239277668214, "global_step": 35510, "epoch": 398, "lr": 9.593651819771469e-05, "val_loss": 2.4858555793762207}
{"train_loss": 0.34173962473869324, "global_step": 35511, "epoch": 399, "lr": 9.59362892650776e-05}
{"train_loss": 0.37600773572921753, "global_step": 35512, "epoch": 399, "lr": 9.593606032626492e-05}
{"train_loss": 0.3290907144546509, "global_step": 35513, "epoch": 399, "lr": 9.593583138127671e-05}
{"train_loss": 0.39899319410324097, "global_step": 35514, "epoch": 399, "lr": 9.593560243011299e-05}
{"train_loss": 0.23721784353256226, "global_step": 35515, "epoch": 399, "lr": 9.593537347277378e-05}
{"train_loss": 0.26135799288749695, "global_step": 35516, "epoch": 399, "lr": 9.593514450925913e-05}
{"train_loss": 0.3253824710845947, "global_step": 35517, "epoch": 399, "lr": 9.593491553956904e-05}
{"train_loss": 0.34212303161621094, "global_step": 35518, "epoch": 399, "lr": 9.593468656370357e-05}
{"train_loss": 0.33414140343666077, "global_step": 35519, "epoch": 399, "lr": 9.593445758166275e-05}
{"train_loss": 0.3403926193714142, "global_step": 35520, "epoch": 399, "lr": 9.59342285934466e-05}
{"train_loss": 0.2852049171924591, "global_step": 35521, "epoch": 399, "lr": 9.593399959905513e-05}
{"train_loss": 0.3189060091972351, "global_step": 35522, "epoch": 399, "lr": 9.593377059848843e-05}
{"train_loss": 0.23159664869308472, "global_step": 35523, "epoch": 399, "lr": 9.593354159174648e-05}
{"train_loss": 0.23124314844608307, "global_step": 35524, "epoch": 399, "lr": 9.593331257882932e-05}
{"train_loss": 0.30147045850753784, "global_step": 35525, "epoch": 399, "lr": 9.593308355973697e-05}
{"train_loss": 0.4074063003063202, "global_step": 35526, "epoch": 399, "lr": 9.59328545344695e-05}
{"train_loss": 0.31320011615753174, "global_step": 35527, "epoch": 399, "lr": 9.593262550302691e-05}
{"train_loss": 0.38041865825653076, "global_step": 35528, "epoch": 399, "lr": 9.593239646540925e-05}
{"train_loss": 0.3476143777370453, "global_step": 35529, "epoch": 399, "lr": 9.593216742161653e-05}
{"train_loss": 0.25935694575309753, "global_step": 35530, "epoch": 399, "lr": 9.593193837164878e-05}
{"train_loss": 0.2488774061203003, "global_step": 35531, "epoch": 399, "lr": 9.593170931550606e-05}
{"train_loss": 0.41801074147224426, "global_step": 35532, "epoch": 399, "lr": 9.593148025318836e-05}
{"train_loss": 0.35396766662597656, "global_step": 35533, "epoch": 399, "lr": 9.593125118469575e-05}
{"train_loss": 0.3712230920791626, "global_step": 35534, "epoch": 399, "lr": 9.593102211002823e-05}
{"train_loss": 0.37455829977989197, "global_step": 35535, "epoch": 399, "lr": 9.593079302918586e-05}
{"train_loss": 0.32006222009658813, "global_step": 35536, "epoch": 399, "lr": 9.593056394216864e-05}
{"train_loss": 0.3594609498977661, "global_step": 35537, "epoch": 399, "lr": 9.593033484897662e-05}
{"train_loss": 0.38458549976348877, "global_step": 35538, "epoch": 399, "lr": 9.593010574960983e-05}
{"train_loss": 0.2710581123828888, "global_step": 35539, "epoch": 399, "lr": 9.59298766440683e-05}
{"train_loss": 0.3642711341381073, "global_step": 35540, "epoch": 399, "lr": 9.592964753235205e-05}
{"train_loss": 0.32434988021850586, "global_step": 35541, "epoch": 399, "lr": 9.592941841446113e-05}
{"train_loss": 0.26920297741889954, "global_step": 35542, "epoch": 399, "lr": 9.592918929039554e-05}
{"train_loss": 0.42969632148742676, "global_step": 35543, "epoch": 399, "lr": 9.592896016015533e-05}
{"train_loss": 0.22119000554084778, "global_step": 35544, "epoch": 399, "lr": 9.592873102374056e-05}
{"train_loss": 0.29347559809684753, "global_step": 35545, "epoch": 399, "lr": 9.592850188115122e-05}
{"train_loss": 0.3695624768733978, "global_step": 35546, "epoch": 399, "lr": 9.592827273238734e-05}
{"train_loss": 0.40492480993270874, "global_step": 35547, "epoch": 399, "lr": 9.592804357744896e-05}
{"train_loss": 0.30095717310905457, "global_step": 35548, "epoch": 399, "lr": 9.592781441633612e-05}
{"train_loss": 0.2916202247142792, "global_step": 35549, "epoch": 399, "lr": 9.592758524904885e-05}
{"train_loss": 0.33371391892433167, "global_step": 35550, "epoch": 399, "lr": 9.592735607558718e-05}
{"train_loss": 0.3436482846736908, "global_step": 35551, "epoch": 399, "lr": 9.592712689595114e-05}
{"train_loss": 0.32428696751594543, "global_step": 35552, "epoch": 399, "lr": 9.592689771014074e-05}
{"train_loss": 0.2670161724090576, "global_step": 35553, "epoch": 399, "lr": 9.592666851815604e-05}
{"train_loss": 0.3388897478580475, "global_step": 35554, "epoch": 399, "lr": 9.592643931999705e-05}
{"train_loss": 0.28614377975463867, "global_step": 35555, "epoch": 399, "lr": 9.592621011566383e-05}
{"train_loss": 0.37851187586784363, "global_step": 35556, "epoch": 399, "lr": 9.592598090515636e-05}
{"train_loss": 0.3513503074645996, "global_step": 35557, "epoch": 399, "lr": 9.592575168847473e-05}
{"train_loss": 0.3138953447341919, "global_step": 35558, "epoch": 399, "lr": 9.592552246561893e-05}
{"train_loss": 0.34321048855781555, "global_step": 35559, "epoch": 399, "lr": 9.592529323658901e-05}
{"train_loss": 0.4547732472419739, "global_step": 35560, "epoch": 399, "lr": 9.592506400138499e-05}
{"train_loss": 0.3064807653427124, "global_step": 35561, "epoch": 399, "lr": 9.592483476000691e-05}
{"train_loss": 0.3915732800960541, "global_step": 35562, "epoch": 399, "lr": 9.592460551245479e-05}
{"train_loss": 0.2847670912742615, "global_step": 35563, "epoch": 399, "lr": 9.592437625872865e-05}
{"train_loss": 0.3471100628376007, "global_step": 35564, "epoch": 399, "lr": 9.592414699882857e-05}
{"train_loss": 0.3706052899360657, "global_step": 35565, "epoch": 399, "lr": 9.592391773275453e-05}
{"train_loss": 0.4266813099384308, "global_step": 35566, "epoch": 399, "lr": 9.592368846050658e-05}
{"train_loss": 0.3756849765777588, "global_step": 35567, "epoch": 399, "lr": 9.592345918208474e-05}
{"train_loss": 0.35397037863731384, "global_step": 35568, "epoch": 399, "lr": 9.592322989748908e-05}
{"train_loss": 0.2603924572467804, "global_step": 35569, "epoch": 399, "lr": 9.592300060671957e-05}
{"train_loss": 0.37765735387802124, "global_step": 35570, "epoch": 399, "lr": 9.59227713097763e-05}
{"train_loss": 0.32268431782722473, "global_step": 35571, "epoch": 399, "lr": 9.592254200665925e-05}
{"train_loss": 0.39973893761634827, "global_step": 35572, "epoch": 399, "lr": 9.592231269736848e-05}
{"train_loss": 0.4225623905658722, "global_step": 35573, "epoch": 399, "lr": 9.592208338190401e-05}
{"train_loss": 0.34132909774780273, "global_step": 35574, "epoch": 399, "lr": 9.592185406026589e-05}
{"train_loss": 0.35198310017585754, "global_step": 35575, "epoch": 399, "lr": 9.592162473245411e-05}
{"train_loss": 0.3859553933143616, "global_step": 35576, "epoch": 399, "lr": 9.592139539846876e-05}
{"train_loss": 0.4175778925418854, "global_step": 35577, "epoch": 399, "lr": 9.592116605830984e-05}
{"train_loss": 0.3261268734931946, "global_step": 35578, "epoch": 399, "lr": 9.592093671197736e-05}
{"train_loss": 0.31741392612457275, "global_step": 35579, "epoch": 399, "lr": 9.592070735947136e-05}
{"train_loss": 0.2662976384162903, "global_step": 35580, "epoch": 399, "lr": 9.592047800079189e-05}
{"train_loss": 0.38289207220077515, "global_step": 35581, "epoch": 399, "lr": 9.592024863593898e-05}
{"train_loss": 0.2932594120502472, "global_step": 35582, "epoch": 399, "lr": 9.592001926491265e-05}
{"train_loss": 0.27783066034317017, "global_step": 35583, "epoch": 399, "lr": 9.591978988771292e-05}
{"train_loss": 0.3786676228046417, "global_step": 35584, "epoch": 399, "lr": 9.591956050433986e-05}
{"train_loss": 0.3871835470199585, "global_step": 35585, "epoch": 399, "lr": 9.591933111479345e-05}
{"train_loss": 0.2915096580982208, "global_step": 35586, "epoch": 399, "lr": 9.591910171907377e-05}
{"train_loss": 0.3840566873550415, "global_step": 35587, "epoch": 399, "lr": 9.59188723171808e-05}
{"train_loss": 0.34108999371528625, "global_step": 35588, "epoch": 399, "lr": 9.59186429091146e-05}
{"train_loss": 0.33759379386901855, "global_step": 35589, "epoch": 399, "lr": 9.591841349487522e-05}
{"train_loss": 0.44773197174072266, "global_step": 35590, "epoch": 399, "lr": 9.591818407446265e-05}
{"train_loss": 0.33929502964019775, "global_step": 35591, "epoch": 399, "lr": 9.591795464787695e-05}
{"train_loss": 0.2958601713180542, "global_step": 35592, "epoch": 399, "lr": 9.591772521511814e-05}
{"train_loss": 0.3287092447280884, "global_step": 35593, "epoch": 399, "lr": 9.591749577618624e-05}
{"train_loss": 0.31186583638191223, "global_step": 35594, "epoch": 399, "lr": 9.59172663310813e-05}
{"train_loss": 0.3147852122783661, "global_step": 35595, "epoch": 399, "lr": 9.591703687980334e-05}
{"train_loss": 0.3710992634296417, "global_step": 35596, "epoch": 399, "lr": 9.59168074223524e-05}
{"train_loss": 0.33006951212882996, "global_step": 35597, "epoch": 399, "lr": 9.59165779587285e-05}
{"train_loss": 0.29038292169570923, "global_step": 35598, "epoch": 399, "lr": 9.591634848893169e-05}
{"train_loss": 0.3393037026182989, "global_step": 35599, "epoch": 399, "lr": 9.591611901296197e-05, "val_loss": 2.5380733013153076}
{"train_loss": 0.362456351518631, "global_step": 35600, "epoch": 400, "lr": 9.59158895308194e-05}
{"train_loss": 0.3477325439453125, "global_step": 35601, "epoch": 400, "lr": 9.591566004250399e-05}
{"train_loss": 0.34478721022605896, "global_step": 35602, "epoch": 400, "lr": 9.591543054801577e-05}
{"train_loss": 0.34230461716651917, "global_step": 35603, "epoch": 400, "lr": 9.591520104735478e-05}
{"train_loss": 0.35115498304367065, "global_step": 35604, "epoch": 400, "lr": 9.591497154052106e-05}
{"train_loss": 0.37253767251968384, "global_step": 35605, "epoch": 400, "lr": 9.591474202751463e-05}
{"train_loss": 0.37723278999328613, "global_step": 35606, "epoch": 400, "lr": 9.591451250833554e-05}
{"train_loss": 0.3146706223487854, "global_step": 35607, "epoch": 400, "lr": 9.591428298298378e-05}
{"train_loss": 0.30176398158073425, "global_step": 35608, "epoch": 400, "lr": 9.591405345145941e-05}
{"train_loss": 0.3523173928260803, "global_step": 35609, "epoch": 400, "lr": 9.591382391376247e-05}
{"train_loss": 0.3301528990268707, "global_step": 35610, "epoch": 400, "lr": 9.591359436989296e-05}
{"train_loss": 0.4661417603492737, "global_step": 35611, "epoch": 400, "lr": 9.591336481985093e-05}
{"train_loss": 0.3703932762145996, "global_step": 35612, "epoch": 400, "lr": 9.59131352636364e-05}
{"train_loss": 0.3025874197483063, "global_step": 35613, "epoch": 400, "lr": 9.591290570124942e-05}
{"train_loss": 0.39885959029197693, "global_step": 35614, "epoch": 400, "lr": 9.591267613269001e-05}
{"train_loss": 0.3026410937309265, "global_step": 35615, "epoch": 400, "lr": 9.591244655795819e-05}
{"train_loss": 0.34197700023651123, "global_step": 35616, "epoch": 400, "lr": 9.591221697705401e-05}
{"train_loss": 0.2815015912055969, "global_step": 35617, "epoch": 400, "lr": 9.591198738997749e-05}
{"train_loss": 0.306540310382843, "global_step": 35618, "epoch": 400, "lr": 9.591175779672866e-05}
{"train_loss": 0.33016255497932434, "global_step": 35619, "epoch": 400, "lr": 9.591152819730756e-05}
{"train_loss": 0.295785516500473, "global_step": 35620, "epoch": 400, "lr": 9.59112985917142e-05}
{"train_loss": 0.27547377347946167, "global_step": 35621, "epoch": 400, "lr": 9.591106897994864e-05}
{"train_loss": 0.279503732919693, "global_step": 35622, "epoch": 400, "lr": 9.59108393620109e-05}
{"train_loss": 0.37269604206085205, "global_step": 35623, "epoch": 400, "lr": 9.591060973790099e-05}
{"train_loss": 0.3712446391582489, "global_step": 35624, "epoch": 400, "lr": 9.591038010761897e-05}
{"train_loss": 0.3527178168296814, "global_step": 35625, "epoch": 400, "lr": 9.591015047116486e-05}
{"train_loss": 0.3686515688896179, "global_step": 35626, "epoch": 400, "lr": 9.59099208285387e-05}
{"train_loss": 0.34403911232948303, "global_step": 35627, "epoch": 400, "lr": 9.59096911797405e-05}
{"train_loss": 0.284136027097702, "global_step": 35628, "epoch": 400, "lr": 9.590946152477028e-05}
{"train_loss": 0.3595236837863922, "global_step": 35629, "epoch": 400, "lr": 9.590923186362812e-05}
{"train_loss": 0.3685164153575897, "global_step": 35630, "epoch": 400, "lr": 9.5909002196314e-05}
{"train_loss": 0.25531280040740967, "global_step": 35631, "epoch": 400, "lr": 9.590877252282799e-05}
{"train_loss": 0.33034178614616394, "global_step": 35632, "epoch": 400, "lr": 9.59085428431701e-05}
{"train_loss": 0.4264375865459442, "global_step": 35633, "epoch": 400, "lr": 9.590831315734038e-05}
{"train_loss": 0.2466290444135666, "global_step": 35634, "epoch": 400, "lr": 9.590808346533884e-05}
{"train_loss": 0.46035459637641907, "global_step": 35635, "epoch": 400, "lr": 9.590785376716551e-05}
{"train_loss": 0.3922213613986969, "global_step": 35636, "epoch": 400, "lr": 9.590762406282044e-05}
{"train_loss": 0.3120940327644348, "global_step": 35637, "epoch": 400, "lr": 9.590739435230364e-05}
{"train_loss": 0.35285720229148865, "global_step": 35638, "epoch": 400, "lr": 9.590716463561516e-05}
{"train_loss": 0.32197442650794983, "global_step": 35639, "epoch": 400, "lr": 9.5906934912755e-05}
{"train_loss": 0.2781798839569092, "global_step": 35640, "epoch": 400, "lr": 9.590670518372322e-05}
{"train_loss": 0.2778421938419342, "global_step": 35641, "epoch": 400, "lr": 9.590647544851986e-05}
{"train_loss": 0.30899763107299805, "global_step": 35642, "epoch": 400, "lr": 9.590624570714492e-05}
{"train_loss": 0.31338536739349365, "global_step": 35643, "epoch": 400, "lr": 9.590601595959846e-05}
{"train_loss": 0.3637130856513977, "global_step": 35644, "epoch": 400, "lr": 9.590578620588047e-05}
{"train_loss": 0.33924952149391174, "global_step": 35645, "epoch": 400, "lr": 9.590555644599102e-05}
{"train_loss": 0.33097952604293823, "global_step": 35646, "epoch": 400, "lr": 9.590532667993013e-05}
{"train_loss": 0.24853210151195526, "global_step": 35647, "epoch": 400, "lr": 9.590509690769782e-05}
{"train_loss": 0.27910271286964417, "global_step": 35648, "epoch": 400, "lr": 9.590486712929415e-05}
{"train_loss": 0.44361820816993713, "global_step": 35649, "epoch": 400, "lr": 9.590463734471911e-05}
{"train_loss": 0.3672558069229126, "global_step": 35650, "epoch": 400, "lr": 9.590440755397276e-05}
{"train_loss": 0.26954251527786255, "global_step": 35651, "epoch": 400, "lr": 9.590417775705511e-05}
{"train_loss": 0.33980274200439453, "global_step": 35652, "epoch": 400, "lr": 9.590394795396621e-05}
{"train_loss": 0.34618842601776123, "global_step": 35653, "epoch": 400, "lr": 9.590371814470609e-05}
{"train_loss": 0.4057418704032898, "global_step": 35654, "epoch": 400, "lr": 9.590348832927476e-05}
{"train_loss": 0.28950560092926025, "global_step": 35655, "epoch": 400, "lr": 9.590325850767228e-05}
{"train_loss": 0.36800292134284973, "global_step": 35656, "epoch": 400, "lr": 9.590302867989866e-05}
{"train_loss": 0.316751629114151, "global_step": 35657, "epoch": 400, "lr": 9.590279884595395e-05}
{"train_loss": 0.30106672644615173, "global_step": 35658, "epoch": 400, "lr": 9.590256900583814e-05}
{"train_loss": 0.2623191475868225, "global_step": 35659, "epoch": 400, "lr": 9.590233915955132e-05}
{"train_loss": 0.41408857703208923, "global_step": 35660, "epoch": 400, "lr": 9.590210930709346e-05}
{"train_loss": 0.33344319462776184, "global_step": 35661, "epoch": 400, "lr": 9.590187944846464e-05}
{"train_loss": 0.391809344291687, "global_step": 35662, "epoch": 400, "lr": 9.590164958366488e-05}
{"train_loss": 0.25880149006843567, "global_step": 35663, "epoch": 400, "lr": 9.590141971269417e-05}
{"train_loss": 0.302272766828537, "global_step": 35664, "epoch": 400, "lr": 9.590118983555261e-05}
{"train_loss": 0.36896875500679016, "global_step": 35665, "epoch": 400, "lr": 9.590095995224018e-05}
{"train_loss": 0.29126548767089844, "global_step": 35666, "epoch": 400, "lr": 9.590073006275691e-05}
{"train_loss": 0.3607415556907654, "global_step": 35667, "epoch": 400, "lr": 9.590050016710287e-05}
{"train_loss": 0.4855031967163086, "global_step": 35668, "epoch": 400, "lr": 9.590027026527805e-05}
{"train_loss": 0.49941813945770264, "global_step": 35669, "epoch": 400, "lr": 9.590004035728251e-05}
{"train_loss": 0.3839147686958313, "global_step": 35670, "epoch": 400, "lr": 9.589981044311626e-05}
{"train_loss": 0.2840934991836548, "global_step": 35671, "epoch": 400, "lr": 9.589958052277935e-05}
{"train_loss": 0.42938780784606934, "global_step": 35672, "epoch": 400, "lr": 9.589935059627179e-05}
{"train_loss": 0.42536816000938416, "global_step": 35673, "epoch": 400, "lr": 9.589912066359363e-05}
{"train_loss": 0.30064821243286133, "global_step": 35674, "epoch": 400, "lr": 9.58988907247449e-05}
{"train_loss": 0.42679792642593384, "global_step": 35675, "epoch": 400, "lr": 9.58986607797256e-05}
{"train_loss": 0.32779937982559204, "global_step": 35676, "epoch": 400, "lr": 9.58984308285358e-05}
{"train_loss": 0.39525407552719116, "global_step": 35677, "epoch": 400, "lr": 9.589820087117552e-05}
{"train_loss": 0.3641543984413147, "global_step": 35678, "epoch": 400, "lr": 9.589797090764479e-05}
{"train_loss": 0.38605961203575134, "global_step": 35679, "epoch": 400, "lr": 9.589774093794362e-05}
{"train_loss": 0.37366652488708496, "global_step": 35680, "epoch": 400, "lr": 9.589751096207207e-05}
{"train_loss": 0.35235023498535156, "global_step": 35681, "epoch": 400, "lr": 9.589728098003014e-05}
{"train_loss": 0.3571372926235199, "global_step": 35682, "epoch": 400, "lr": 9.58970509918179e-05}
{"train_loss": 0.442360520362854, "global_step": 35683, "epoch": 400, "lr": 9.589682099743535e-05}
{"train_loss": 0.3602660000324249, "global_step": 35684, "epoch": 400, "lr": 9.589659099688255e-05}
{"train_loss": 0.3097159266471863, "global_step": 35685, "epoch": 400, "lr": 9.58963609901595e-05}
{"train_loss": 0.3349691927433014, "global_step": 35686, "epoch": 400, "lr": 9.589613097726624e-05}
{"train_loss": 0.3139802813529968, "global_step": 35687, "epoch": 400, "lr": 9.58959009582028e-05}
{"train_loss": 0.34632252474849146, "global_step": 35688, "epoch": 400, "lr": 9.589567093296923e-05, "train/sim_max_reward_0": 0.9666579032184606, "train/sim_max_reward_1": 0.900690253645576, "train/sim_max_reward_2": 0.496005713474332, "train/sim_max_reward_3": 0.6391143102179944, "train/sim_max_reward_4": 0.7568539408891899, "train/sim_max_reward_5": 0.3651148029843588, "test/sim_max_reward_4500000": 0.42688249145611856, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9168224949078309, "test/sim_max_reward_4500003": 0.3757580804792818, "test/sim_max_reward_4500004": 0.48734702078519, "test/sim_max_reward_4500005": 0.8808825927650463, "test/sim_max_reward_4500006": 0.9669599880420586, "test/sim_max_reward_4500007": 0.302382631949792, "test/sim_max_reward_4500008": 0.23837235390300812, "test/sim_max_reward_4500009": 0.049894766516665155, "test/sim_max_reward_4500010": 0.8680092498477456, "test/sim_max_reward_4500011": 0.616557642334309, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5918802621403927, "test/sim_max_reward_4500014": 0.5821792913676793, "test/sim_max_reward_4500015": 0.3849974251860631, "test/sim_max_reward_4500016": 0.4331921927853455, "test/sim_max_reward_4500017": 0.6006096165581439, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9393061616650215, "test/sim_max_reward_4500020": 0.4281246381545683, "test/sim_max_reward_4500021": 0.14861511220147822, "test/sim_max_reward_4500022": 0.6765296424446609, "test/sim_max_reward_4500023": 0.6802991911668105, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.0942099257657801, "test/sim_max_reward_4500026": 0.5114774218431848, "test/sim_max_reward_4500027": 0.49743574265146717, "test/sim_max_reward_4500028": 0.4964407579743293, "test/sim_max_reward_4500029": 0.21939148388230822, "test/sim_max_reward_4500030": 0.9456104400225167, "test/sim_max_reward_4500031": 0.1501052587598866, "test/sim_max_reward_4500032": 0.5621253480779608, "test/sim_max_reward_4500033": 0.9217995274365437, "test/sim_max_reward_4500034": 0.40516476790586126, "test/sim_max_reward_4500035": 0.8474968611062503, "test/sim_max_reward_4500036": 0.5221584677330299, "test/sim_max_reward_4500037": 0.46808730225615597, "test/sim_max_reward_4500038": 0.46552171080411514, "test/sim_max_reward_4500039": 0.5858904776406679, "test/sim_max_reward_4500040": 0.9658774183750456, "test/sim_max_reward_4500041": 0.47680427779021284, "test/sim_max_reward_4500042": 0.684506825170685, "test/sim_max_reward_4500043": 0.9089744368310846, "test/sim_max_reward_4500044": 0.9103245242016664, "test/sim_max_reward_4500045": 0.5808358633371971, "test/sim_max_reward_4500046": 0.5226726198360299, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.7950183814166148, "test/sim_max_reward_4500049": 0.5869836846520008, "train/mean_score": 0.6874061540716521, "test/mean_score": 0.5344103274425561, "val_loss": 2.5657966136932373, "train_action_mse_error": 51.523128509521484}
{"train_loss": 0.43934372067451477, "global_step": 35689, "epoch": 401, "lr": 9.589544090156554e-05}
{"train_loss": 0.3022463321685791, "global_step": 35690, "epoch": 401, "lr": 9.589521086399177e-05}
{"train_loss": 0.30331629514694214, "global_step": 35691, "epoch": 401, "lr": 9.589498082024795e-05}
{"train_loss": 0.42988401651382446, "global_step": 35692, "epoch": 401, "lr": 9.58947507703341e-05}
{"train_loss": 0.3322802484035492, "global_step": 35693, "epoch": 401, "lr": 9.589452071425027e-05}
{"train_loss": 0.3286875784397125, "global_step": 35694, "epoch": 401, "lr": 9.589429065199646e-05}
{"train_loss": 0.39279043674468994, "global_step": 35695, "epoch": 401, "lr": 9.589406058357276e-05}
{"train_loss": 0.365989625453949, "global_step": 35696, "epoch": 401, "lr": 9.589383050897913e-05}
{"train_loss": 0.42166876792907715, "global_step": 35697, "epoch": 401, "lr": 9.589360042821565e-05}
{"train_loss": 0.28815367817878723, "global_step": 35698, "epoch": 401, "lr": 9.589337034128232e-05}
{"train_loss": 0.321339875459671, "global_step": 35699, "epoch": 401, "lr": 9.58931402481792e-05}
{"train_loss": 0.3289870321750641, "global_step": 35700, "epoch": 401, "lr": 9.58929101489063e-05}
{"train_loss": 0.3115304410457611, "global_step": 35701, "epoch": 401, "lr": 9.589268004346366e-05}
{"train_loss": 0.4121524393558502, "global_step": 35702, "epoch": 401, "lr": 9.589244993185131e-05}
{"train_loss": 0.4485158622264862, "global_step": 35703, "epoch": 401, "lr": 9.589221981406927e-05}
{"train_loss": 0.3241470754146576, "global_step": 35704, "epoch": 401, "lr": 9.589198969011759e-05}
{"train_loss": 0.25303399562835693, "global_step": 35705, "epoch": 401, "lr": 9.58917595599963e-05}
{"train_loss": 0.41207635402679443, "global_step": 35706, "epoch": 401, "lr": 9.58915294237054e-05}
{"train_loss": 0.4420435428619385, "global_step": 35707, "epoch": 401, "lr": 9.589129928124496e-05}
{"train_loss": 0.33349382877349854, "global_step": 35708, "epoch": 401, "lr": 9.589106913261499e-05}
{"train_loss": 0.33615320920944214, "global_step": 35709, "epoch": 401, "lr": 9.589083897781552e-05}
{"train_loss": 0.360408216714859, "global_step": 35710, "epoch": 401, "lr": 9.589060881684659e-05}
{"train_loss": 0.42342042922973633, "global_step": 35711, "epoch": 401, "lr": 9.589037864970822e-05}
{"train_loss": 0.455651193857193, "global_step": 35712, "epoch": 401, "lr": 9.589014847640045e-05}
{"train_loss": 0.4252660572528839, "global_step": 35713, "epoch": 401, "lr": 9.588991829692333e-05}
{"train_loss": 0.37151506543159485, "global_step": 35714, "epoch": 401, "lr": 9.588968811127685e-05}
{"train_loss": 0.3783089220523834, "global_step": 35715, "epoch": 401, "lr": 9.588945791946107e-05}
{"train_loss": 0.26453644037246704, "global_step": 35716, "epoch": 401, "lr": 9.5889227721476e-05}
{"train_loss": 0.32384222745895386, "global_step": 35717, "epoch": 401, "lr": 9.588899751732169e-05}
{"train_loss": 0.27591371536254883, "global_step": 35718, "epoch": 401, "lr": 9.588876730699816e-05}
{"train_loss": 0.3957921266555786, "global_step": 35719, "epoch": 401, "lr": 9.588853709050545e-05}
{"train_loss": 0.3690241277217865, "global_step": 35720, "epoch": 401, "lr": 9.588830686784358e-05}
{"train_loss": 0.5953290462493896, "global_step": 35721, "epoch": 401, "lr": 9.588807663901257e-05}
{"train_loss": 0.3628128170967102, "global_step": 35722, "epoch": 401, "lr": 9.588784640401251e-05}
{"train_loss": 0.38476648926734924, "global_step": 35723, "epoch": 401, "lr": 9.588761616284336e-05}
{"train_loss": 0.3698004186153412, "global_step": 35724, "epoch": 401, "lr": 9.588738591550519e-05}
{"train_loss": 0.3540545701980591, "global_step": 35725, "epoch": 401, "lr": 9.5887155661998e-05}
{"train_loss": 0.24932478368282318, "global_step": 35726, "epoch": 401, "lr": 9.588692540232187e-05}
{"train_loss": 0.39129403233528137, "global_step": 35727, "epoch": 401, "lr": 9.588669513647678e-05}
{"train_loss": 0.3778960108757019, "global_step": 35728, "epoch": 401, "lr": 9.588646486446278e-05}
{"train_loss": 0.36532390117645264, "global_step": 35729, "epoch": 401, "lr": 9.588623458627992e-05}
{"train_loss": 0.4063138961791992, "global_step": 35730, "epoch": 401, "lr": 9.588600430192821e-05}
{"train_loss": 0.3708283603191376, "global_step": 35731, "epoch": 401, "lr": 9.588577401140768e-05}
{"train_loss": 0.2748400568962097, "global_step": 35732, "epoch": 401, "lr": 9.588554371471838e-05}
{"train_loss": 0.3582787811756134, "global_step": 35733, "epoch": 401, "lr": 9.588531341186033e-05}
{"train_loss": 0.3364488184452057, "global_step": 35734, "epoch": 401, "lr": 9.588508310283354e-05}
{"train_loss": 0.36002054810523987, "global_step": 35735, "epoch": 401, "lr": 9.588485278763806e-05}
{"train_loss": 0.33361294865608215, "global_step": 35736, "epoch": 401, "lr": 9.588462246627394e-05}
{"train_loss": 0.30795201659202576, "global_step": 35737, "epoch": 401, "lr": 9.588439213874119e-05}
{"train_loss": 0.3516048491001129, "global_step": 35738, "epoch": 401, "lr": 9.588416180503983e-05}
{"train_loss": 0.2865621745586395, "global_step": 35739, "epoch": 401, "lr": 9.588393146516989e-05}
{"train_loss": 0.382222056388855, "global_step": 35740, "epoch": 401, "lr": 9.588370111913145e-05}
{"train_loss": 0.34626504778862, "global_step": 35741, "epoch": 401, "lr": 9.588347076692448e-05}
{"train_loss": 0.3847227096557617, "global_step": 35742, "epoch": 401, "lr": 9.588324040854906e-05}
{"train_loss": 0.25946104526519775, "global_step": 35743, "epoch": 401, "lr": 9.588301004400517e-05}
{"train_loss": 0.2713495194911957, "global_step": 35744, "epoch": 401, "lr": 9.588277967329289e-05}
{"train_loss": 0.318671852350235, "global_step": 35745, "epoch": 401, "lr": 9.588254929641222e-05}
{"train_loss": 0.3645201027393341, "global_step": 35746, "epoch": 401, "lr": 9.58823189133632e-05}
{"train_loss": 0.24785003066062927, "global_step": 35747, "epoch": 401, "lr": 9.588208852414586e-05}
{"train_loss": 0.345453143119812, "global_step": 35748, "epoch": 401, "lr": 9.588185812876023e-05}
{"train_loss": 0.3561108410358429, "global_step": 35749, "epoch": 401, "lr": 9.588162772720634e-05}
{"train_loss": 0.38030070066452026, "global_step": 35750, "epoch": 401, "lr": 9.588139731948425e-05}
{"train_loss": 0.37720054388046265, "global_step": 35751, "epoch": 401, "lr": 9.588116690559395e-05}
{"train_loss": 0.3396921753883362, "global_step": 35752, "epoch": 401, "lr": 9.588093648553548e-05}
{"train_loss": 0.3715381324291229, "global_step": 35753, "epoch": 401, "lr": 9.588070605930888e-05}
{"train_loss": 0.2986072599887848, "global_step": 35754, "epoch": 401, "lr": 9.588047562691418e-05}
{"train_loss": 0.41770488023757935, "global_step": 35755, "epoch": 401, "lr": 9.588024518835142e-05}
{"train_loss": 0.38330069184303284, "global_step": 35756, "epoch": 401, "lr": 9.58800147436206e-05}
{"train_loss": 0.3615771532058716, "global_step": 35757, "epoch": 401, "lr": 9.587978429272177e-05}
{"train_loss": 0.3246983587741852, "global_step": 35758, "epoch": 401, "lr": 9.587955383565498e-05}
{"train_loss": 0.2373059242963791, "global_step": 35759, "epoch": 401, "lr": 9.587932337242024e-05}
{"train_loss": 0.38927099108695984, "global_step": 35760, "epoch": 401, "lr": 9.587909290301758e-05}
{"train_loss": 0.3040550649166107, "global_step": 35761, "epoch": 401, "lr": 9.587886242744704e-05}
{"train_loss": 0.3628201484680176, "global_step": 35762, "epoch": 401, "lr": 9.587863194570863e-05}
{"train_loss": 0.3214454650878906, "global_step": 35763, "epoch": 401, "lr": 9.587840145780242e-05}
{"train_loss": 0.34299126267433167, "global_step": 35764, "epoch": 401, "lr": 9.58781709637284e-05}
{"train_loss": 0.3191013038158417, "global_step": 35765, "epoch": 401, "lr": 9.587794046348663e-05}
{"train_loss": 0.24918051064014435, "global_step": 35766, "epoch": 401, "lr": 9.587770995707713e-05}
{"train_loss": 0.38417699933052063, "global_step": 35767, "epoch": 401, "lr": 9.587747944449992e-05}
{"train_loss": 0.41630738973617554, "global_step": 35768, "epoch": 401, "lr": 9.587724892575505e-05}
{"train_loss": 0.2919134199619293, "global_step": 35769, "epoch": 401, "lr": 9.587701840084254e-05}
{"train_loss": 0.2501901686191559, "global_step": 35770, "epoch": 401, "lr": 9.587678786976242e-05}
{"train_loss": 0.3418116569519043, "global_step": 35771, "epoch": 401, "lr": 9.587655733251474e-05}
{"train_loss": 0.31224584579467773, "global_step": 35772, "epoch": 401, "lr": 9.58763267890995e-05}
{"train_loss": 0.38454878330230713, "global_step": 35773, "epoch": 401, "lr": 9.587609623951676e-05}
{"train_loss": 0.3759448230266571, "global_step": 35774, "epoch": 401, "lr": 9.587586568376653e-05}
{"train_loss": 0.31821128726005554, "global_step": 35775, "epoch": 401, "lr": 9.587563512184884e-05}
{"train_loss": 0.3718855679035187, "global_step": 35776, "epoch": 401, "lr": 9.587540455376377e-05}
{"train_loss": 0.35090613783745284, "global_step": 35777, "epoch": 401, "lr": 9.587517397951127e-05, "val_loss": 2.6107754707336426}
{"train_loss": 0.4511827230453491, "global_step": 35778, "epoch": 402, "lr": 9.587494339909141e-05}
{"train_loss": 0.38711801171302795, "global_step": 35779, "epoch": 402, "lr": 9.587471281250424e-05}
{"train_loss": 0.33990100026130676, "global_step": 35780, "epoch": 402, "lr": 9.587448221974978e-05}
{"train_loss": 0.2666766941547394, "global_step": 35781, "epoch": 402, "lr": 9.587425162082804e-05}
{"train_loss": 0.4073575437068939, "global_step": 35782, "epoch": 402, "lr": 9.587402101573907e-05}
{"train_loss": 0.3381922245025635, "global_step": 35783, "epoch": 402, "lr": 9.58737904044829e-05}
{"train_loss": 0.25573503971099854, "global_step": 35784, "epoch": 402, "lr": 9.587355978705957e-05}
{"train_loss": 0.2622016966342926, "global_step": 35785, "epoch": 402, "lr": 9.587332916346907e-05}
{"train_loss": 0.32785242795944214, "global_step": 35786, "epoch": 402, "lr": 9.587309853371149e-05}
{"train_loss": 0.28939443826675415, "global_step": 35787, "epoch": 402, "lr": 9.587286789778681e-05}
{"train_loss": 0.3757895529270172, "global_step": 35788, "epoch": 402, "lr": 9.587263725569509e-05}
{"train_loss": 0.3047941327095032, "global_step": 35789, "epoch": 402, "lr": 9.587240660743635e-05}
{"train_loss": 0.2850998044013977, "global_step": 35790, "epoch": 402, "lr": 9.587217595301063e-05}
{"train_loss": 0.33267176151275635, "global_step": 35791, "epoch": 402, "lr": 9.587194529241795e-05}
{"train_loss": 0.3554767668247223, "global_step": 35792, "epoch": 402, "lr": 9.587171462565835e-05}
{"train_loss": 0.3243710994720459, "global_step": 35793, "epoch": 402, "lr": 9.587148395273184e-05}
{"train_loss": 0.30977490544319153, "global_step": 35794, "epoch": 402, "lr": 9.587125327363849e-05}
{"train_loss": 0.3146178126335144, "global_step": 35795, "epoch": 402, "lr": 9.58710225883783e-05}
{"train_loss": 0.35915833711624146, "global_step": 35796, "epoch": 402, "lr": 9.587079189695132e-05}
{"train_loss": 0.3649793863296509, "global_step": 35797, "epoch": 402, "lr": 9.587056119935756e-05}
{"train_loss": 0.2405509352684021, "global_step": 35798, "epoch": 402, "lr": 9.587033049559706e-05}
{"train_loss": 0.31435659527778625, "global_step": 35799, "epoch": 402, "lr": 9.587009978566987e-05}
{"train_loss": 0.35975444316864014, "global_step": 35800, "epoch": 402, "lr": 9.586986906957599e-05}
{"train_loss": 0.3659895062446594, "global_step": 35801, "epoch": 402, "lr": 9.586963834731546e-05}
{"train_loss": 0.32120922207832336, "global_step": 35802, "epoch": 402, "lr": 9.586940761888834e-05}
{"train_loss": 0.28362545371055603, "global_step": 35803, "epoch": 402, "lr": 9.58691768842946e-05}
{"train_loss": 0.25283706188201904, "global_step": 35804, "epoch": 402, "lr": 9.586894614353434e-05}
{"train_loss": 0.3420758545398712, "global_step": 35805, "epoch": 402, "lr": 9.586871539660754e-05}
{"train_loss": 0.33174270391464233, "global_step": 35806, "epoch": 402, "lr": 9.586848464351427e-05}
{"train_loss": 0.2838422656059265, "global_step": 35807, "epoch": 402, "lr": 9.586825388425453e-05}
{"train_loss": 0.36278650164604187, "global_step": 35808, "epoch": 402, "lr": 9.586802311882837e-05}
{"train_loss": 0.36486098170280457, "global_step": 35809, "epoch": 402, "lr": 9.586779234723579e-05}
{"train_loss": 0.27297261357307434, "global_step": 35810, "epoch": 402, "lr": 9.586756156947687e-05}
{"train_loss": 0.3028828799724579, "global_step": 35811, "epoch": 402, "lr": 9.58673307855516e-05}
{"train_loss": 0.31507840752601624, "global_step": 35812, "epoch": 402, "lr": 9.586709999546003e-05}
{"train_loss": 0.3306255638599396, "global_step": 35813, "epoch": 402, "lr": 9.586686919920219e-05}
{"train_loss": 0.2504931390285492, "global_step": 35814, "epoch": 402, "lr": 9.58666383967781e-05}
{"train_loss": 0.32268697023391724, "global_step": 35815, "epoch": 402, "lr": 9.586640758818782e-05}
{"train_loss": 0.45388272404670715, "global_step": 35816, "epoch": 402, "lr": 9.586617677343135e-05}
{"train_loss": 0.3802543580532074, "global_step": 35817, "epoch": 402, "lr": 9.586594595250873e-05}
{"train_loss": 0.2759244740009308, "global_step": 35818, "epoch": 402, "lr": 9.586571512541999e-05}
{"train_loss": 0.33156663179397583, "global_step": 35819, "epoch": 402, "lr": 9.586548429216516e-05}
{"train_loss": 0.3573707938194275, "global_step": 35820, "epoch": 402, "lr": 9.586525345274428e-05}
{"train_loss": 0.4247584044933319, "global_step": 35821, "epoch": 402, "lr": 9.586502260715738e-05}
{"train_loss": 0.3758460283279419, "global_step": 35822, "epoch": 402, "lr": 9.586479175540449e-05}
{"train_loss": 0.38752925395965576, "global_step": 35823, "epoch": 402, "lr": 9.58645608974856e-05}
{"train_loss": 0.42093560099601746, "global_step": 35824, "epoch": 402, "lr": 9.586433003340082e-05}
{"train_loss": 0.33198922872543335, "global_step": 35825, "epoch": 402, "lr": 9.586409916315013e-05}
{"train_loss": 0.2673436403274536, "global_step": 35826, "epoch": 402, "lr": 9.586386828673357e-05}
{"train_loss": 0.32320231199264526, "global_step": 35827, "epoch": 402, "lr": 9.586363740415119e-05}
{"train_loss": 0.3608444333076477, "global_step": 35828, "epoch": 402, "lr": 9.586340651540298e-05}
{"train_loss": 0.3572205901145935, "global_step": 35829, "epoch": 402, "lr": 9.586317562048899e-05}
{"train_loss": 0.30525311827659607, "global_step": 35830, "epoch": 402, "lr": 9.586294471940927e-05}
{"train_loss": 0.4339917004108429, "global_step": 35831, "epoch": 402, "lr": 9.586271381216383e-05}
{"train_loss": 0.3062865436077118, "global_step": 35832, "epoch": 402, "lr": 9.586248289875271e-05}
{"train_loss": 0.33689287304878235, "global_step": 35833, "epoch": 402, "lr": 9.586225197917593e-05}
{"train_loss": 0.34507879614830017, "global_step": 35834, "epoch": 402, "lr": 9.586202105343353e-05}
{"train_loss": 0.4523875415325165, "global_step": 35835, "epoch": 402, "lr": 9.586179012152553e-05}
{"train_loss": 0.28052371740341187, "global_step": 35836, "epoch": 402, "lr": 9.586155918345201e-05}
{"train_loss": 0.3457755744457245, "global_step": 35837, "epoch": 402, "lr": 9.586132823921293e-05}
{"train_loss": 0.43932774662971497, "global_step": 35838, "epoch": 402, "lr": 9.586109728880837e-05}
{"train_loss": 0.39416491985321045, "global_step": 35839, "epoch": 402, "lr": 9.586086633223834e-05}
{"train_loss": 0.5002743601799011, "global_step": 35840, "epoch": 402, "lr": 9.586063536950285e-05}
{"train_loss": 0.30193278193473816, "global_step": 35841, "epoch": 402, "lr": 9.586040440060199e-05}
{"train_loss": 0.3567555248737335, "global_step": 35842, "epoch": 402, "lr": 9.586017342553575e-05}
{"train_loss": 0.3597407042980194, "global_step": 35843, "epoch": 402, "lr": 9.585994244430416e-05}
{"train_loss": 0.4414319097995758, "global_step": 35844, "epoch": 402, "lr": 9.585971145690727e-05}
{"train_loss": 0.43085286021232605, "global_step": 35845, "epoch": 402, "lr": 9.58594804633451e-05}
{"train_loss": 0.2908194065093994, "global_step": 35846, "epoch": 402, "lr": 9.585924946361766e-05}
{"train_loss": 0.23233550786972046, "global_step": 35847, "epoch": 402, "lr": 9.585901845772503e-05}
{"train_loss": 0.39090827107429504, "global_step": 35848, "epoch": 402, "lr": 9.58587874456672e-05}
{"train_loss": 0.3030374348163605, "global_step": 35849, "epoch": 402, "lr": 9.585855642744422e-05}
{"train_loss": 0.3283137381076813, "global_step": 35850, "epoch": 402, "lr": 9.58583254030561e-05}
{"train_loss": 0.42325863242149353, "global_step": 35851, "epoch": 402, "lr": 9.585809437250291e-05}
{"train_loss": 0.3302765190601349, "global_step": 35852, "epoch": 402, "lr": 9.585786333578466e-05}
{"train_loss": 0.298533171415329, "global_step": 35853, "epoch": 402, "lr": 9.585763229290137e-05}
{"train_loss": 0.2659716010093689, "global_step": 35854, "epoch": 402, "lr": 9.585740124385307e-05}
{"train_loss": 0.4359399676322937, "global_step": 35855, "epoch": 402, "lr": 9.58571701886398e-05}
{"train_loss": 0.511795163154602, "global_step": 35856, "epoch": 402, "lr": 9.585693912726162e-05}
{"train_loss": 0.3337821662425995, "global_step": 35857, "epoch": 402, "lr": 9.585670805971851e-05}
{"train_loss": 0.359825074672699, "global_step": 35858, "epoch": 402, "lr": 9.585647698601053e-05}
{"train_loss": 0.39938369393348694, "global_step": 35859, "epoch": 402, "lr": 9.585624590613771e-05}
{"train_loss": 0.3031198978424072, "global_step": 35860, "epoch": 402, "lr": 9.585601482010007e-05}
{"train_loss": 0.37789416313171387, "global_step": 35861, "epoch": 402, "lr": 9.585578372789765e-05}
{"train_loss": 0.3441336452960968, "global_step": 35862, "epoch": 402, "lr": 9.585555262953047e-05}
{"train_loss": 0.35385000705718994, "global_step": 35863, "epoch": 402, "lr": 9.585532152499858e-05}
{"train_loss": 0.3604884147644043, "global_step": 35864, "epoch": 402, "lr": 9.5855090414302e-05}
{"train_loss": 0.44022026658058167, "global_step": 35865, "epoch": 402, "lr": 9.585485929744077e-05}
{"train_loss": 0.3471918065896195, "global_step": 35866, "epoch": 402, "lr": 9.58546281744149e-05, "val_loss": 2.4286465644836426}
{"train_loss": 0.3532099723815918, "global_step": 35867, "epoch": 403, "lr": 9.585439704522444e-05}
{"train_loss": 0.3968309462070465, "global_step": 35868, "epoch": 403, "lr": 9.585416590986941e-05}
{"train_loss": 0.45897170901298523, "global_step": 35869, "epoch": 403, "lr": 9.585393476834986e-05}
{"train_loss": 0.36648568511009216, "global_step": 35870, "epoch": 403, "lr": 9.585370362066579e-05}
{"train_loss": 0.37634146213531494, "global_step": 35871, "epoch": 403, "lr": 9.585347246681726e-05}
{"train_loss": 0.3444463014602661, "global_step": 35872, "epoch": 403, "lr": 9.585324130680428e-05}
{"train_loss": 0.48943424224853516, "global_step": 35873, "epoch": 403, "lr": 9.585301014062691e-05}
{"train_loss": 0.4333087205886841, "global_step": 35874, "epoch": 403, "lr": 9.585277896828516e-05}
{"train_loss": 0.33809709548950195, "global_step": 35875, "epoch": 403, "lr": 9.585254778977904e-05}
{"train_loss": 0.3644176125526428, "global_step": 35876, "epoch": 403, "lr": 9.58523166051086e-05}
{"train_loss": 0.3357570171356201, "global_step": 35877, "epoch": 403, "lr": 9.585208541427391e-05}
{"train_loss": 0.4720311462879181, "global_step": 35878, "epoch": 403, "lr": 9.585185421727493e-05}
{"train_loss": 0.39665666222572327, "global_step": 35879, "epoch": 403, "lr": 9.585162301411175e-05}
{"train_loss": 0.473489910364151, "global_step": 35880, "epoch": 403, "lr": 9.585139180478438e-05}
{"train_loss": 0.3570370376110077, "global_step": 35881, "epoch": 403, "lr": 9.585116058929283e-05}
{"train_loss": 0.3584637939929962, "global_step": 35882, "epoch": 403, "lr": 9.585092936763715e-05}
{"train_loss": 0.2458256632089615, "global_step": 35883, "epoch": 403, "lr": 9.58506981398174e-05}
{"train_loss": 0.3569938838481903, "global_step": 35884, "epoch": 403, "lr": 9.585046690583356e-05}
{"train_loss": 0.32215791940689087, "global_step": 35885, "epoch": 403, "lr": 9.585023566568568e-05}
{"train_loss": 0.3893042802810669, "global_step": 35886, "epoch": 403, "lr": 9.58500044193738e-05}
{"train_loss": 0.35983914136886597, "global_step": 35887, "epoch": 403, "lr": 9.584977316689796e-05}
{"train_loss": 0.3900327980518341, "global_step": 35888, "epoch": 403, "lr": 9.584954190825817e-05}
{"train_loss": 0.3721458613872528, "global_step": 35889, "epoch": 403, "lr": 9.584931064345446e-05}
{"train_loss": 0.3540903925895691, "global_step": 35890, "epoch": 403, "lr": 9.584907937248686e-05}
{"train_loss": 0.4002542197704315, "global_step": 35891, "epoch": 403, "lr": 9.584884809535544e-05}
{"train_loss": 0.41480910778045654, "global_step": 35892, "epoch": 403, "lr": 9.584861681206016e-05}
{"train_loss": 0.48604080080986023, "global_step": 35893, "epoch": 403, "lr": 9.584838552260113e-05}
{"train_loss": 0.3100898861885071, "global_step": 35894, "epoch": 403, "lr": 9.584815422697832e-05}
{"train_loss": 0.30131399631500244, "global_step": 35895, "epoch": 403, "lr": 9.58479229251918e-05}
{"train_loss": 0.4531343877315521, "global_step": 35896, "epoch": 403, "lr": 9.584769161724159e-05}
{"train_loss": 0.36978963017463684, "global_step": 35897, "epoch": 403, "lr": 9.58474603031277e-05}
{"train_loss": 0.4938555359840393, "global_step": 35898, "epoch": 403, "lr": 9.584722898285018e-05}
{"train_loss": 0.3207840323448181, "global_step": 35899, "epoch": 403, "lr": 9.584699765640905e-05}
{"train_loss": 0.33015745878219604, "global_step": 35900, "epoch": 403, "lr": 9.584676632380438e-05}
{"train_loss": 0.2514151930809021, "global_step": 35901, "epoch": 403, "lr": 9.584653498503615e-05}
{"train_loss": 0.3581598103046417, "global_step": 35902, "epoch": 403, "lr": 9.584630364010441e-05}
{"train_loss": 0.29745757579803467, "global_step": 35903, "epoch": 403, "lr": 9.58460722890092e-05}
{"train_loss": 0.28737735748291016, "global_step": 35904, "epoch": 403, "lr": 9.584584093175055e-05}
{"train_loss": 0.4295858144760132, "global_step": 35905, "epoch": 403, "lr": 9.584560956832846e-05}
{"train_loss": 0.334386944770813, "global_step": 35906, "epoch": 403, "lr": 9.584537819874302e-05}
{"train_loss": 0.2750084400177002, "global_step": 35907, "epoch": 403, "lr": 9.584514682299422e-05}
{"train_loss": 0.30720189213752747, "global_step": 35908, "epoch": 403, "lr": 9.584491544108208e-05}
{"train_loss": 0.42235851287841797, "global_step": 35909, "epoch": 403, "lr": 9.584468405300666e-05}
{"train_loss": 0.3996933102607727, "global_step": 35910, "epoch": 403, "lr": 9.584445265876798e-05}
{"train_loss": 0.31342506408691406, "global_step": 35911, "epoch": 403, "lr": 9.584422125836608e-05}
{"train_loss": 0.404074490070343, "global_step": 35912, "epoch": 403, "lr": 9.584398985180099e-05}
{"train_loss": 0.2873803675174713, "global_step": 35913, "epoch": 403, "lr": 9.584375843907272e-05}
{"train_loss": 0.36273524165153503, "global_step": 35914, "epoch": 403, "lr": 9.584352702018132e-05}
{"train_loss": 0.4090443253517151, "global_step": 35915, "epoch": 403, "lr": 9.584329559512681e-05}
{"train_loss": 0.38991278409957886, "global_step": 35916, "epoch": 403, "lr": 9.584306416390922e-05}
{"train_loss": 0.35297927260398865, "global_step": 35917, "epoch": 403, "lr": 9.584283272652861e-05}
{"train_loss": 0.3033299744129181, "global_step": 35918, "epoch": 403, "lr": 9.584260128298497e-05}
{"train_loss": 0.2898910939693451, "global_step": 35919, "epoch": 403, "lr": 9.584236983327836e-05}
{"train_loss": 0.2976224422454834, "global_step": 35920, "epoch": 403, "lr": 9.584213837740881e-05}
{"train_loss": 0.4030475318431854, "global_step": 35921, "epoch": 403, "lr": 9.584190691537634e-05}
{"train_loss": 0.3532504439353943, "global_step": 35922, "epoch": 403, "lr": 9.584167544718098e-05}
{"train_loss": 0.37036383152008057, "global_step": 35923, "epoch": 403, "lr": 9.584144397282277e-05}
{"train_loss": 0.270067036151886, "global_step": 35924, "epoch": 403, "lr": 9.584121249230174e-05}
{"train_loss": 0.40456467866897583, "global_step": 35925, "epoch": 403, "lr": 9.58409810056179e-05}
{"train_loss": 0.26010969281196594, "global_step": 35926, "epoch": 403, "lr": 9.584074951277131e-05}
{"train_loss": 0.35310468077659607, "global_step": 35927, "epoch": 403, "lr": 9.584051801376199e-05}
{"train_loss": 0.3475814163684845, "global_step": 35928, "epoch": 403, "lr": 9.584028650858997e-05}
{"train_loss": 0.4170173704624176, "global_step": 35929, "epoch": 403, "lr": 9.584005499725528e-05}
{"train_loss": 0.32093966007232666, "global_step": 35930, "epoch": 403, "lr": 9.583982347975796e-05}
{"train_loss": 0.36176183819770813, "global_step": 35931, "epoch": 403, "lr": 9.583959195609805e-05}
{"train_loss": 0.4652891457080841, "global_step": 35932, "epoch": 403, "lr": 9.583936042627552e-05}
{"train_loss": 0.2204202115535736, "global_step": 35933, "epoch": 403, "lr": 9.583912889029048e-05}
{"train_loss": 0.3000074326992035, "global_step": 35934, "epoch": 403, "lr": 9.583889734814294e-05}
{"train_loss": 0.30374404788017273, "global_step": 35935, "epoch": 403, "lr": 9.583866579983288e-05}
{"train_loss": 0.4251100718975067, "global_step": 35936, "epoch": 403, "lr": 9.583843424536039e-05}
{"train_loss": 0.37002891302108765, "global_step": 35937, "epoch": 403, "lr": 9.583820268472549e-05}
{"train_loss": 0.37536951899528503, "global_step": 35938, "epoch": 403, "lr": 9.583797111792818e-05}
{"train_loss": 0.3205249607563019, "global_step": 35939, "epoch": 403, "lr": 9.583773954496853e-05}
{"train_loss": 0.27884939312934875, "global_step": 35940, "epoch": 403, "lr": 9.583750796584656e-05}
{"train_loss": 0.4245176911354065, "global_step": 35941, "epoch": 403, "lr": 9.583727638056227e-05}
{"train_loss": 0.3630446195602417, "global_step": 35942, "epoch": 403, "lr": 9.583704478911575e-05}
{"train_loss": 0.26650741696357727, "global_step": 35943, "epoch": 403, "lr": 9.583681319150698e-05}
{"train_loss": 0.3036639094352722, "global_step": 35944, "epoch": 403, "lr": 9.5836581587736e-05}
{"train_loss": 0.3477614223957062, "global_step": 35945, "epoch": 403, "lr": 9.583634997780288e-05}
{"train_loss": 0.3160833716392517, "global_step": 35946, "epoch": 403, "lr": 9.58361183617076e-05}
{"train_loss": 0.397862046957016, "global_step": 35947, "epoch": 403, "lr": 9.583588673945021e-05}
{"train_loss": 0.38139259815216064, "global_step": 35948, "epoch": 403, "lr": 9.583565511103074e-05}
{"train_loss": 0.28524965047836304, "global_step": 35949, "epoch": 403, "lr": 9.583542347644925e-05}
{"train_loss": 0.29558753967285156, "global_step": 35950, "epoch": 403, "lr": 9.583519183570573e-05}
{"train_loss": 0.2624644935131073, "global_step": 35951, "epoch": 403, "lr": 9.583496018880023e-05}
{"train_loss": 0.27533403038978577, "global_step": 35952, "epoch": 403, "lr": 9.583472853573276e-05}
{"train_loss": 0.33724281191825867, "global_step": 35953, "epoch": 403, "lr": 9.583449687650339e-05}
{"train_loss": 0.38884466886520386, "global_step": 35954, "epoch": 403, "lr": 9.583426521111212e-05}
{"train_loss": 0.3554728776216507, "global_step": 35955, "epoch": 403, "lr": 9.583403353955899e-05, "val_loss": 2.598168134689331}
{"train_loss": 0.355048269033432, "global_step": 35956, "epoch": 404, "lr": 9.583380186184406e-05}
{"train_loss": 0.3966462016105652, "global_step": 35957, "epoch": 404, "lr": 9.58335701779673e-05}
{"train_loss": 0.3029228746891022, "global_step": 35958, "epoch": 404, "lr": 9.58333384879288e-05}
{"train_loss": 0.4717603623867035, "global_step": 35959, "epoch": 404, "lr": 9.583310679172854e-05}
{"train_loss": 0.44050711393356323, "global_step": 35960, "epoch": 404, "lr": 9.58328750893666e-05}
{"train_loss": 0.45630937814712524, "global_step": 35961, "epoch": 404, "lr": 9.583264338084299e-05}
{"train_loss": 0.3125665783882141, "global_step": 35962, "epoch": 404, "lr": 9.583241166615773e-05}
{"train_loss": 0.37308722734451294, "global_step": 35963, "epoch": 404, "lr": 9.583217994531086e-05}
{"train_loss": 0.38249585032463074, "global_step": 35964, "epoch": 404, "lr": 9.583194821830242e-05}
{"train_loss": 0.33559849858283997, "global_step": 35965, "epoch": 404, "lr": 9.583171648513241e-05}
{"train_loss": 0.34080520272254944, "global_step": 35966, "epoch": 404, "lr": 9.58314847458009e-05}
{"train_loss": 0.3815056085586548, "global_step": 35967, "epoch": 404, "lr": 9.58312530003079e-05}
{"train_loss": 0.31288960576057434, "global_step": 35968, "epoch": 404, "lr": 9.583102124865347e-05}
{"train_loss": 0.4107891321182251, "global_step": 35969, "epoch": 404, "lr": 9.58307894908376e-05}
{"train_loss": 0.3726213574409485, "global_step": 35970, "epoch": 404, "lr": 9.583055772686034e-05}
{"train_loss": 0.2857210338115692, "global_step": 35971, "epoch": 404, "lr": 9.583032595672172e-05}
{"train_loss": 0.4465588927268982, "global_step": 35972, "epoch": 404, "lr": 9.583009418042177e-05}
{"train_loss": 0.4297439754009247, "global_step": 35973, "epoch": 404, "lr": 9.582986239796052e-05}
{"train_loss": 0.5067083239555359, "global_step": 35974, "epoch": 404, "lr": 9.5829630609338e-05}
{"train_loss": 0.3274851143360138, "global_step": 35975, "epoch": 404, "lr": 9.582939881455426e-05}
{"train_loss": 0.3175601065158844, "global_step": 35976, "epoch": 404, "lr": 9.582916701360932e-05}
{"train_loss": 0.3627457916736603, "global_step": 35977, "epoch": 404, "lr": 9.582893520650319e-05}
{"train_loss": 0.31995120644569397, "global_step": 35978, "epoch": 404, "lr": 9.582870339323593e-05}
{"train_loss": 0.25912851095199585, "global_step": 35979, "epoch": 404, "lr": 9.582847157380754e-05}
{"train_loss": 0.3517191410064697, "global_step": 35980, "epoch": 404, "lr": 9.582823974821808e-05}
{"train_loss": 0.41182002425193787, "global_step": 35981, "epoch": 404, "lr": 9.582800791646759e-05}
{"train_loss": 0.2861945927143097, "global_step": 35982, "epoch": 404, "lr": 9.582777607855606e-05}
{"train_loss": 0.3585900068283081, "global_step": 35983, "epoch": 404, "lr": 9.582754423448356e-05}
{"train_loss": 0.3788488209247589, "global_step": 35984, "epoch": 404, "lr": 9.582731238425009e-05}
{"train_loss": 0.2930407226085663, "global_step": 35985, "epoch": 404, "lr": 9.582708052785571e-05}
{"train_loss": 0.38026198744773865, "global_step": 35986, "epoch": 404, "lr": 9.582684866530043e-05}
{"train_loss": 0.3151266276836395, "global_step": 35987, "epoch": 404, "lr": 9.58266167965843e-05}
{"train_loss": 0.3702034652233124, "global_step": 35988, "epoch": 404, "lr": 9.582638492170733e-05}
{"train_loss": 0.3157350718975067, "global_step": 35989, "epoch": 404, "lr": 9.582615304066955e-05}
{"train_loss": 0.3084286153316498, "global_step": 35990, "epoch": 404, "lr": 9.582592115347102e-05}
{"train_loss": 0.4164433479309082, "global_step": 35991, "epoch": 404, "lr": 9.582568926011174e-05}
{"train_loss": 0.289264440536499, "global_step": 35992, "epoch": 404, "lr": 9.582545736059177e-05}
{"train_loss": 0.3509688675403595, "global_step": 35993, "epoch": 404, "lr": 9.58252254549111e-05}
{"train_loss": 0.3609289228916168, "global_step": 35994, "epoch": 404, "lr": 9.582499354306982e-05}
{"train_loss": 0.3350594639778137, "global_step": 35995, "epoch": 404, "lr": 9.58247616250679e-05}
{"train_loss": 0.3912637531757355, "global_step": 35996, "epoch": 404, "lr": 9.582452970090542e-05}
{"train_loss": 0.42069119215011597, "global_step": 35997, "epoch": 404, "lr": 9.582429777058237e-05}
{"train_loss": 0.29839983582496643, "global_step": 35998, "epoch": 404, "lr": 9.582406583409882e-05}
{"train_loss": 0.35935065150260925, "global_step": 35999, "epoch": 404, "lr": 9.582383389145477e-05}
{"train_loss": 0.45414552092552185, "global_step": 36000, "epoch": 404, "lr": 9.582360194265027e-05}
{"train_loss": 0.250537633895874, "global_step": 36001, "epoch": 404, "lr": 9.582336998768535e-05}
{"train_loss": 0.419389009475708, "global_step": 36002, "epoch": 404, "lr": 9.582313802656005e-05}
{"train_loss": 0.3302862048149109, "global_step": 36003, "epoch": 404, "lr": 9.582290605927435e-05}
{"train_loss": 0.4900565445423126, "global_step": 36004, "epoch": 404, "lr": 9.582267408582834e-05}
{"train_loss": 0.46558699011802673, "global_step": 36005, "epoch": 404, "lr": 9.582244210622204e-05}
{"train_loss": 0.4052616059780121, "global_step": 36006, "epoch": 404, "lr": 9.582221012045546e-05}
{"train_loss": 0.30898886919021606, "global_step": 36007, "epoch": 404, "lr": 9.582197812852864e-05}
{"train_loss": 0.339269757270813, "global_step": 36008, "epoch": 404, "lr": 9.582174613044161e-05}
{"train_loss": 0.3107258379459381, "global_step": 36009, "epoch": 404, "lr": 9.582151412619442e-05}
{"train_loss": 0.3516025245189667, "global_step": 36010, "epoch": 404, "lr": 9.582128211578707e-05}
{"train_loss": 0.33361178636550903, "global_step": 36011, "epoch": 404, "lr": 9.582105009921962e-05}
{"train_loss": 0.3968767523765564, "global_step": 36012, "epoch": 404, "lr": 9.582081807649207e-05}
{"train_loss": 0.3799230754375458, "global_step": 36013, "epoch": 404, "lr": 9.582058604760449e-05}
{"train_loss": 0.30165746808052063, "global_step": 36014, "epoch": 404, "lr": 9.582035401255687e-05}
{"train_loss": 0.38621795177459717, "global_step": 36015, "epoch": 404, "lr": 9.582012197134928e-05}
{"train_loss": 0.3792562782764435, "global_step": 36016, "epoch": 404, "lr": 9.581988992398173e-05}
{"train_loss": 0.40457046031951904, "global_step": 36017, "epoch": 404, "lr": 9.581965787045426e-05}
{"train_loss": 0.36686351895332336, "global_step": 36018, "epoch": 404, "lr": 9.581942581076688e-05}
{"train_loss": 0.5587298274040222, "global_step": 36019, "epoch": 404, "lr": 9.581919374491964e-05}
{"train_loss": 0.2937163710594177, "global_step": 36020, "epoch": 404, "lr": 9.581896167291258e-05}
{"train_loss": 0.3554146885871887, "global_step": 36021, "epoch": 404, "lr": 9.58187295947457e-05}
{"train_loss": 0.37395909428596497, "global_step": 36022, "epoch": 404, "lr": 9.581849751041908e-05}
{"train_loss": 0.3917901813983917, "global_step": 36023, "epoch": 404, "lr": 9.581826541993268e-05}
{"train_loss": 0.50565105676651, "global_step": 36024, "epoch": 404, "lr": 9.58180333232866e-05}
{"train_loss": 0.4132157862186432, "global_step": 36025, "epoch": 404, "lr": 9.581780122048084e-05}
{"train_loss": 0.4359358847141266, "global_step": 36026, "epoch": 404, "lr": 9.581756911151544e-05}
{"train_loss": 0.28603246808052063, "global_step": 36027, "epoch": 404, "lr": 9.581733699639042e-05}
{"train_loss": 0.34443265199661255, "global_step": 36028, "epoch": 404, "lr": 9.581710487510581e-05}
{"train_loss": 0.3373875319957733, "global_step": 36029, "epoch": 404, "lr": 9.581687274766166e-05}
{"train_loss": 0.2859359681606293, "global_step": 36030, "epoch": 404, "lr": 9.581664061405798e-05}
{"train_loss": 0.3792683482170105, "global_step": 36031, "epoch": 404, "lr": 9.581640847429483e-05}
{"train_loss": 0.3161729574203491, "global_step": 36032, "epoch": 404, "lr": 9.58161763283722e-05}
{"train_loss": 0.2857573926448822, "global_step": 36033, "epoch": 404, "lr": 9.581594417629015e-05}
{"train_loss": 0.28116974234580994, "global_step": 36034, "epoch": 404, "lr": 9.58157120180487e-05}
{"train_loss": 0.31857386231422424, "global_step": 36035, "epoch": 404, "lr": 9.581547985364791e-05}
{"train_loss": 0.5427670478820801, "global_step": 36036, "epoch": 404, "lr": 9.581524768308776e-05}
{"train_loss": 0.38671305775642395, "global_step": 36037, "epoch": 404, "lr": 9.581501550636833e-05}
{"train_loss": 0.31369590759277344, "global_step": 36038, "epoch": 404, "lr": 9.581478332348961e-05}
{"train_loss": 0.32012131810188293, "global_step": 36039, "epoch": 404, "lr": 9.581455113445166e-05}
{"train_loss": 0.23911963403224945, "global_step": 36040, "epoch": 404, "lr": 9.58143189392545e-05}
{"train_loss": 0.35353919863700867, "global_step": 36041, "epoch": 404, "lr": 9.581408673789817e-05}
{"train_loss": 0.33986154198646545, "global_step": 36042, "epoch": 404, "lr": 9.581385453038268e-05}
{"train_loss": 0.3290335536003113, "global_step": 36043, "epoch": 404, "lr": 9.581362231670809e-05}
{"train_loss": 0.3642111633265956, "global_step": 36044, "epoch": 404, "lr": 9.581339009687441e-05, "val_loss": 2.54597806930542}
{"train_loss": 0.33347630500793457, "global_step": 36045, "epoch": 405, "lr": 9.581315787088168e-05}
{"train_loss": 0.35352280735969543, "global_step": 36046, "epoch": 405, "lr": 9.581292563872993e-05}
{"train_loss": 0.3247412443161011, "global_step": 36047, "epoch": 405, "lr": 9.581269340041918e-05}
{"train_loss": 0.28828832507133484, "global_step": 36048, "epoch": 405, "lr": 9.581246115594948e-05}
{"train_loss": 0.2987099885940552, "global_step": 36049, "epoch": 405, "lr": 9.581222890532087e-05}
{"train_loss": 0.3293272852897644, "global_step": 36050, "epoch": 405, "lr": 9.581199664853334e-05}
{"train_loss": 0.27761828899383545, "global_step": 36051, "epoch": 405, "lr": 9.581176438558697e-05}
{"train_loss": 0.4224626421928406, "global_step": 36052, "epoch": 405, "lr": 9.581153211648174e-05}
{"train_loss": 0.30562102794647217, "global_step": 36053, "epoch": 405, "lr": 9.581129984121771e-05}
{"train_loss": 0.34825488924980164, "global_step": 36054, "epoch": 405, "lr": 9.581106755979492e-05}
{"train_loss": 0.28186729550361633, "global_step": 36055, "epoch": 405, "lr": 9.58108352722134e-05}
{"train_loss": 0.3580337464809418, "global_step": 36056, "epoch": 405, "lr": 9.581060297847315e-05}
{"train_loss": 0.32641661167144775, "global_step": 36057, "epoch": 405, "lr": 9.581037067857424e-05}
{"train_loss": 0.36490920186042786, "global_step": 36058, "epoch": 405, "lr": 9.581013837251667e-05}
{"train_loss": 0.31647923588752747, "global_step": 36059, "epoch": 405, "lr": 9.58099060603005e-05}
{"train_loss": 0.30879250168800354, "global_step": 36060, "epoch": 405, "lr": 9.580967374192574e-05}
{"train_loss": 0.3154706358909607, "global_step": 36061, "epoch": 405, "lr": 9.580944141739242e-05}
{"train_loss": 0.30615007877349854, "global_step": 36062, "epoch": 405, "lr": 9.580920908670058e-05}
{"train_loss": 0.41771575808525085, "global_step": 36063, "epoch": 405, "lr": 9.580897674985026e-05}
{"train_loss": 0.26204586029052734, "global_step": 36064, "epoch": 405, "lr": 9.580874440684147e-05}
{"train_loss": 0.3070376217365265, "global_step": 36065, "epoch": 405, "lr": 9.580851205767426e-05}
{"train_loss": 0.3976527452468872, "global_step": 36066, "epoch": 405, "lr": 9.580827970234865e-05}
{"train_loss": 0.31265193223953247, "global_step": 36067, "epoch": 405, "lr": 9.580804734086468e-05}
{"train_loss": 0.36971089243888855, "global_step": 36068, "epoch": 405, "lr": 9.580781497322237e-05}
{"train_loss": 0.2823646366596222, "global_step": 36069, "epoch": 405, "lr": 9.580758259942176e-05}
{"train_loss": 0.2778908908367157, "global_step": 36070, "epoch": 405, "lr": 9.580735021946287e-05}
{"train_loss": 0.25997304916381836, "global_step": 36071, "epoch": 405, "lr": 9.580711783334577e-05}
{"train_loss": 0.4439249038696289, "global_step": 36072, "epoch": 405, "lr": 9.580688544107044e-05}
{"train_loss": 0.3323763906955719, "global_step": 36073, "epoch": 405, "lr": 9.580665304263695e-05}
{"train_loss": 0.26959705352783203, "global_step": 36074, "epoch": 405, "lr": 9.580642063804528e-05}
{"train_loss": 0.3174244165420532, "global_step": 36075, "epoch": 405, "lr": 9.580618822729551e-05}
{"train_loss": 0.397228866815567, "global_step": 36076, "epoch": 405, "lr": 9.580595581038766e-05}
{"train_loss": 0.3046242594718933, "global_step": 36077, "epoch": 405, "lr": 9.580572338732177e-05}
{"train_loss": 0.3303360044956207, "global_step": 36078, "epoch": 405, "lr": 9.580549095809783e-05}
{"train_loss": 0.40270304679870605, "global_step": 36079, "epoch": 405, "lr": 9.580525852271592e-05}
{"train_loss": 0.35611915588378906, "global_step": 36080, "epoch": 405, "lr": 9.580502608117605e-05}
{"train_loss": 0.43340739607810974, "global_step": 36081, "epoch": 405, "lr": 9.580479363347825e-05}
{"train_loss": 0.32249781489372253, "global_step": 36082, "epoch": 405, "lr": 9.580456117962255e-05}
{"train_loss": 0.3573111295700073, "global_step": 36083, "epoch": 405, "lr": 9.580432871960898e-05}
{"train_loss": 0.33299514651298523, "global_step": 36084, "epoch": 405, "lr": 9.58040962534376e-05}
{"train_loss": 0.2689041495323181, "global_step": 36085, "epoch": 405, "lr": 9.58038637811084e-05}
{"train_loss": 0.3443039059638977, "global_step": 36086, "epoch": 405, "lr": 9.580363130262142e-05}
{"train_loss": 0.3512153923511505, "global_step": 36087, "epoch": 405, "lr": 9.580339881797672e-05}
{"train_loss": 0.2908937931060791, "global_step": 36088, "epoch": 405, "lr": 9.58031663271743e-05}
{"train_loss": 0.2954159080982208, "global_step": 36089, "epoch": 405, "lr": 9.580293383021421e-05}
{"train_loss": 0.297378271818161, "global_step": 36090, "epoch": 405, "lr": 9.580270132709647e-05}
{"train_loss": 0.2533578872680664, "global_step": 36091, "epoch": 405, "lr": 9.580246881782111e-05}
{"train_loss": 0.4372265636920929, "global_step": 36092, "epoch": 405, "lr": 9.580223630238817e-05}
{"train_loss": 0.31634777784347534, "global_step": 36093, "epoch": 405, "lr": 9.580200378079767e-05}
{"train_loss": 0.33897414803504944, "global_step": 36094, "epoch": 405, "lr": 9.580177125304967e-05}
{"train_loss": 0.23716256022453308, "global_step": 36095, "epoch": 405, "lr": 9.580153871914416e-05}
{"train_loss": 0.3319488763809204, "global_step": 36096, "epoch": 405, "lr": 9.580130617908121e-05}
{"train_loss": 0.2206028550863266, "global_step": 36097, "epoch": 405, "lr": 9.580107363286082e-05}
{"train_loss": 0.38611555099487305, "global_step": 36098, "epoch": 405, "lr": 9.580084108048302e-05}
{"train_loss": 0.3811991214752197, "global_step": 36099, "epoch": 405, "lr": 9.580060852194788e-05}
{"train_loss": 0.30078330636024475, "global_step": 36100, "epoch": 405, "lr": 9.58003759572554e-05}
{"train_loss": 0.3936404287815094, "global_step": 36101, "epoch": 405, "lr": 9.580014338640561e-05}
{"train_loss": 0.259143590927124, "global_step": 36102, "epoch": 405, "lr": 9.579991080939855e-05}
{"train_loss": 0.3728848099708557, "global_step": 36103, "epoch": 405, "lr": 9.579967822623426e-05}
{"train_loss": 0.29073381423950195, "global_step": 36104, "epoch": 405, "lr": 9.579944563691274e-05}
{"train_loss": 0.2416858673095703, "global_step": 36105, "epoch": 405, "lr": 9.579921304143407e-05}
{"train_loss": 0.25266033411026, "global_step": 36106, "epoch": 405, "lr": 9.579898043979823e-05}
{"train_loss": 0.3225362300872803, "global_step": 36107, "epoch": 405, "lr": 9.579874783200529e-05}
{"train_loss": 0.27792835235595703, "global_step": 36108, "epoch": 405, "lr": 9.579851521805526e-05}
{"train_loss": 0.34058713912963867, "global_step": 36109, "epoch": 405, "lr": 9.579828259794818e-05}
{"train_loss": 0.4734247028827667, "global_step": 36110, "epoch": 405, "lr": 9.579804997168409e-05}
{"train_loss": 0.3470474183559418, "global_step": 36111, "epoch": 405, "lr": 9.5797817339263e-05}
{"train_loss": 0.33934104442596436, "global_step": 36112, "epoch": 405, "lr": 9.579758470068494e-05}
{"train_loss": 0.3799874484539032, "global_step": 36113, "epoch": 405, "lr": 9.579735205594998e-05}
{"train_loss": 0.3351171612739563, "global_step": 36114, "epoch": 405, "lr": 9.57971194050581e-05}
{"train_loss": 0.2401576042175293, "global_step": 36115, "epoch": 405, "lr": 9.579688674800937e-05}
{"train_loss": 0.4836602807044983, "global_step": 36116, "epoch": 405, "lr": 9.57966540848038e-05}
{"train_loss": 0.32118192315101624, "global_step": 36117, "epoch": 405, "lr": 9.579642141544143e-05}
{"train_loss": 0.29971832036972046, "global_step": 36118, "epoch": 405, "lr": 9.57961887399223e-05}
{"train_loss": 0.27327314019203186, "global_step": 36119, "epoch": 405, "lr": 9.57959560582464e-05}
{"train_loss": 0.32360535860061646, "global_step": 36120, "epoch": 405, "lr": 9.579572337041382e-05}
{"train_loss": 0.3031282126903534, "global_step": 36121, "epoch": 405, "lr": 9.579549067642455e-05}
{"train_loss": 0.36393681168556213, "global_step": 36122, "epoch": 405, "lr": 9.579525797627865e-05}
{"train_loss": 0.36285892128944397, "global_step": 36123, "epoch": 405, "lr": 9.579502526997611e-05}
{"train_loss": 0.43387681245803833, "global_step": 36124, "epoch": 405, "lr": 9.5794792557517e-05}
{"train_loss": 0.3058129549026489, "global_step": 36125, "epoch": 405, "lr": 9.579455983890135e-05}
{"train_loss": 0.2884875237941742, "global_step": 36126, "epoch": 405, "lr": 9.579432711412917e-05}
{"train_loss": 0.36859679222106934, "global_step": 36127, "epoch": 405, "lr": 9.57940943832005e-05}
{"train_loss": 0.30474504828453064, "global_step": 36128, "epoch": 405, "lr": 9.579386164611537e-05}
{"train_loss": 0.3476370871067047, "global_step": 36129, "epoch": 405, "lr": 9.579362890287383e-05}
{"train_loss": 0.2844085395336151, "global_step": 36130, "epoch": 405, "lr": 9.579339615347586e-05}
{"train_loss": 0.3351273238658905, "global_step": 36131, "epoch": 405, "lr": 9.579316339792155e-05}
{"train_loss": 0.3756122589111328, "global_step": 36132, "epoch": 405, "lr": 9.579293063621091e-05}
{"train_loss": 0.32971186768472865, "global_step": 36133, "epoch": 405, "lr": 9.579269786834396e-05, "val_loss": 2.621347427368164, "train_action_mse_error": 22.96379852294922}
{"train_loss": 0.29344692826271057, "global_step": 36134, "epoch": 406, "lr": 9.579246509432074e-05}
{"train_loss": 0.3337385952472687, "global_step": 36135, "epoch": 406, "lr": 9.579223231414129e-05}
{"train_loss": 0.3283222019672394, "global_step": 36136, "epoch": 406, "lr": 9.579199952780562e-05}
{"train_loss": 0.29070764780044556, "global_step": 36137, "epoch": 406, "lr": 9.579176673531378e-05}
{"train_loss": 0.40300485491752625, "global_step": 36138, "epoch": 406, "lr": 9.57915339366658e-05}
{"train_loss": 0.2467813491821289, "global_step": 36139, "epoch": 406, "lr": 9.579130113186169e-05}
{"train_loss": 0.33192965388298035, "global_step": 36140, "epoch": 406, "lr": 9.579106832090153e-05}
{"train_loss": 0.2563054859638214, "global_step": 36141, "epoch": 406, "lr": 9.579083550378529e-05}
{"train_loss": 0.3795257806777954, "global_step": 36142, "epoch": 406, "lr": 9.579060268051303e-05}
{"train_loss": 0.3179107904434204, "global_step": 36143, "epoch": 406, "lr": 9.57903698510848e-05}
{"train_loss": 0.44113263487815857, "global_step": 36144, "epoch": 406, "lr": 9.579013701550059e-05}
{"train_loss": 0.3394325077533722, "global_step": 36145, "epoch": 406, "lr": 9.578990417376048e-05}
{"train_loss": 0.38119393587112427, "global_step": 36146, "epoch": 406, "lr": 9.578967132586446e-05}
{"train_loss": 0.40679293870925903, "global_step": 36147, "epoch": 406, "lr": 9.578943847181257e-05}
{"train_loss": 0.2979930639266968, "global_step": 36148, "epoch": 406, "lr": 9.578920561160488e-05}
{"train_loss": 0.2942907214164734, "global_step": 36149, "epoch": 406, "lr": 9.578897274524134e-05}
{"train_loss": 0.33547258377075195, "global_step": 36150, "epoch": 406, "lr": 9.578873987272207e-05}
{"train_loss": 0.42297160625457764, "global_step": 36151, "epoch": 406, "lr": 9.578850699404704e-05}
{"train_loss": 0.41741660237312317, "global_step": 36152, "epoch": 406, "lr": 9.578827410921632e-05}
{"train_loss": 0.32968828082084656, "global_step": 36153, "epoch": 406, "lr": 9.578804121822993e-05}
{"train_loss": 0.2886975407600403, "global_step": 36154, "epoch": 406, "lr": 9.578780832108786e-05}
{"train_loss": 0.2717685401439667, "global_step": 36155, "epoch": 406, "lr": 9.578757541779022e-05}
{"train_loss": 0.34252384305000305, "global_step": 36156, "epoch": 406, "lr": 9.578734250833697e-05}
{"train_loss": 0.3229525089263916, "global_step": 36157, "epoch": 406, "lr": 9.578710959272819e-05}
{"train_loss": 0.2407861202955246, "global_step": 36158, "epoch": 406, "lr": 9.578687667096387e-05}
{"train_loss": 0.33568650484085083, "global_step": 36159, "epoch": 406, "lr": 9.578664374304408e-05}
{"train_loss": 0.3780454695224762, "global_step": 36160, "epoch": 406, "lr": 9.578641080896881e-05}
{"train_loss": 0.3639003336429596, "global_step": 36161, "epoch": 406, "lr": 9.578617786873815e-05}
{"train_loss": 0.30808883905410767, "global_step": 36162, "epoch": 406, "lr": 9.578594492235208e-05}
{"train_loss": 0.3623623549938202, "global_step": 36163, "epoch": 406, "lr": 9.578571196981064e-05}
{"train_loss": 0.3250854015350342, "global_step": 36164, "epoch": 406, "lr": 9.578547901111387e-05}
{"train_loss": 0.3089311420917511, "global_step": 36165, "epoch": 406, "lr": 9.578524604626181e-05}
{"train_loss": 0.4002039134502411, "global_step": 36166, "epoch": 406, "lr": 9.578501307525447e-05}
{"train_loss": 0.2633591294288635, "global_step": 36167, "epoch": 406, "lr": 9.578478009809191e-05}
{"train_loss": 0.26930496096611023, "global_step": 36168, "epoch": 406, "lr": 9.578454711477414e-05}
{"train_loss": 0.36011838912963867, "global_step": 36169, "epoch": 406, "lr": 9.578431412530117e-05}
{"train_loss": 0.2601889371871948, "global_step": 36170, "epoch": 406, "lr": 9.578408112967309e-05}
{"train_loss": 0.3958841562271118, "global_step": 36171, "epoch": 406, "lr": 9.578384812788989e-05}
{"train_loss": 0.31364670395851135, "global_step": 36172, "epoch": 406, "lr": 9.57836151199516e-05}
{"train_loss": 0.40115800499916077, "global_step": 36173, "epoch": 406, "lr": 9.578338210585827e-05}
{"train_loss": 0.33820006251335144, "global_step": 36174, "epoch": 406, "lr": 9.578314908560992e-05}
{"train_loss": 0.3246011435985565, "global_step": 36175, "epoch": 406, "lr": 9.578291605920657e-05}
{"train_loss": 0.2625781297683716, "global_step": 36176, "epoch": 406, "lr": 9.578268302664827e-05}
{"train_loss": 0.45115622878074646, "global_step": 36177, "epoch": 406, "lr": 9.578244998793507e-05}
{"train_loss": 0.3041161000728607, "global_step": 36178, "epoch": 406, "lr": 9.578221694306696e-05}
{"train_loss": 0.291839599609375, "global_step": 36179, "epoch": 406, "lr": 9.578198389204399e-05}
{"train_loss": 0.3677021563053131, "global_step": 36180, "epoch": 406, "lr": 9.578175083486619e-05}
{"train_loss": 0.3272061049938202, "global_step": 36181, "epoch": 406, "lr": 9.578151777153358e-05}
{"train_loss": 0.3359583616256714, "global_step": 36182, "epoch": 406, "lr": 9.578128470204622e-05}
{"train_loss": 0.47458937764167786, "global_step": 36183, "epoch": 406, "lr": 9.578105162640413e-05}
{"train_loss": 0.3611662983894348, "global_step": 36184, "epoch": 406, "lr": 9.578081854460732e-05}
{"train_loss": 0.3367649018764496, "global_step": 36185, "epoch": 406, "lr": 9.578058545665585e-05}
{"train_loss": 0.3282622992992401, "global_step": 36186, "epoch": 406, "lr": 9.578035236254972e-05}
{"train_loss": 0.489406943321228, "global_step": 36187, "epoch": 406, "lr": 9.578011926228899e-05}
{"train_loss": 0.3518189787864685, "global_step": 36188, "epoch": 406, "lr": 9.577988615587368e-05}
{"train_loss": 0.2976231575012207, "global_step": 36189, "epoch": 406, "lr": 9.577965304330383e-05}
{"train_loss": 0.2961714267730713, "global_step": 36190, "epoch": 406, "lr": 9.577941992457945e-05}
{"train_loss": 0.34914952516555786, "global_step": 36191, "epoch": 406, "lr": 9.577918679970059e-05}
{"train_loss": 0.36162450909614563, "global_step": 36192, "epoch": 406, "lr": 9.577895366866729e-05}
{"train_loss": 0.4679867625236511, "global_step": 36193, "epoch": 406, "lr": 9.577872053147954e-05}
{"train_loss": 0.48003673553466797, "global_step": 36194, "epoch": 406, "lr": 9.577848738813742e-05}
{"train_loss": 0.4150295853614807, "global_step": 36195, "epoch": 406, "lr": 9.577825423864093e-05}
{"train_loss": 0.3751765191555023, "global_step": 36196, "epoch": 406, "lr": 9.577802108299012e-05}
{"train_loss": 0.26902616024017334, "global_step": 36197, "epoch": 406, "lr": 9.577778792118502e-05}
{"train_loss": 0.4882516860961914, "global_step": 36198, "epoch": 406, "lr": 9.577755475322563e-05}
{"train_loss": 0.384935200214386, "global_step": 36199, "epoch": 406, "lr": 9.577732157911203e-05}
{"train_loss": 0.37382861971855164, "global_step": 36200, "epoch": 406, "lr": 9.577708839884421e-05}
{"train_loss": 0.42052820324897766, "global_step": 36201, "epoch": 406, "lr": 9.577685521242223e-05}
{"train_loss": 0.3101974129676819, "global_step": 36202, "epoch": 406, "lr": 9.57766220198461e-05}
{"train_loss": 0.29827654361724854, "global_step": 36203, "epoch": 406, "lr": 9.577638882111586e-05}
{"train_loss": 0.31711187958717346, "global_step": 36204, "epoch": 406, "lr": 9.577615561623154e-05}
{"train_loss": 0.32523760199546814, "global_step": 36205, "epoch": 406, "lr": 9.57759224051932e-05}
{"train_loss": 0.250918447971344, "global_step": 36206, "epoch": 406, "lr": 9.577568918800081e-05}
{"train_loss": 0.3538629710674286, "global_step": 36207, "epoch": 406, "lr": 9.577545596465447e-05}
{"train_loss": 0.3138229548931122, "global_step": 36208, "epoch": 406, "lr": 9.577522273515416e-05}
{"train_loss": 0.285677433013916, "global_step": 36209, "epoch": 406, "lr": 9.577498949949992e-05}
{"train_loss": 0.44429391622543335, "global_step": 36210, "epoch": 406, "lr": 9.577475625769181e-05}
{"train_loss": 0.2852939963340759, "global_step": 36211, "epoch": 406, "lr": 9.577452300972984e-05}
{"train_loss": 0.36958664655685425, "global_step": 36212, "epoch": 406, "lr": 9.577428975561403e-05}
{"train_loss": 0.39314553141593933, "global_step": 36213, "epoch": 406, "lr": 9.577405649534444e-05}
{"train_loss": 0.2371370643377304, "global_step": 36214, "epoch": 406, "lr": 9.577382322892107e-05}
{"train_loss": 0.2865440547466278, "global_step": 36215, "epoch": 406, "lr": 9.577358995634397e-05}
{"train_loss": 0.4675610065460205, "global_step": 36216, "epoch": 406, "lr": 9.577335667761318e-05}
{"train_loss": 0.27804842591285706, "global_step": 36217, "epoch": 406, "lr": 9.57731233927287e-05}
{"train_loss": 0.3547045886516571, "global_step": 36218, "epoch": 406, "lr": 9.57728901016906e-05}
{"train_loss": 0.368853896856308, "global_step": 36219, "epoch": 406, "lr": 9.57726568044989e-05}
{"train_loss": 0.3869476020336151, "global_step": 36220, "epoch": 406, "lr": 9.577242350115361e-05}
{"train_loss": 0.3511471450328827, "global_step": 36221, "epoch": 406, "lr": 9.577219019165477e-05}
{"train_loss": 0.34496163451269773, "global_step": 36222, "epoch": 406, "lr": 9.577195687600243e-05, "val_loss": 2.5136849880218506}
{"train_loss": 0.2623835504055023, "global_step": 36223, "epoch": 407, "lr": 9.57717235541966e-05}
{"train_loss": 0.30101776123046875, "global_step": 36224, "epoch": 407, "lr": 9.577149022623732e-05}
{"train_loss": 0.3334941565990448, "global_step": 36225, "epoch": 407, "lr": 9.577125689212463e-05}
{"train_loss": 0.27927857637405396, "global_step": 36226, "epoch": 407, "lr": 9.577102355185854e-05}
{"train_loss": 0.42710232734680176, "global_step": 36227, "epoch": 407, "lr": 9.57707902054391e-05}
{"train_loss": 0.2562795877456665, "global_step": 36228, "epoch": 407, "lr": 9.577055685286633e-05}
{"train_loss": 0.3604091703891754, "global_step": 36229, "epoch": 407, "lr": 9.577032349414027e-05}
{"train_loss": 0.3073182702064514, "global_step": 36230, "epoch": 407, "lr": 9.577009012926095e-05}
{"train_loss": 0.2823147177696228, "global_step": 36231, "epoch": 407, "lr": 9.57698567582284e-05}
{"train_loss": 0.3410100042819977, "global_step": 36232, "epoch": 407, "lr": 9.576962338104265e-05}
{"train_loss": 0.36389344930648804, "global_step": 36233, "epoch": 407, "lr": 9.576938999770373e-05}
{"train_loss": 0.44480472803115845, "global_step": 36234, "epoch": 407, "lr": 9.576915660821168e-05}
{"train_loss": 0.35639527440071106, "global_step": 36235, "epoch": 407, "lr": 9.57689232125665e-05}
{"train_loss": 0.36572757363319397, "global_step": 36236, "epoch": 407, "lr": 9.576868981076828e-05}
{"train_loss": 0.3353349268436432, "global_step": 36237, "epoch": 407, "lr": 9.5768456402817e-05}
{"train_loss": 0.28940024971961975, "global_step": 36238, "epoch": 407, "lr": 9.576822298871271e-05}
{"train_loss": 0.35021457076072693, "global_step": 36239, "epoch": 407, "lr": 9.576798956845544e-05}
{"train_loss": 0.2902025878429413, "global_step": 36240, "epoch": 407, "lr": 9.576775614204522e-05}
{"train_loss": 0.3166670799255371, "global_step": 36241, "epoch": 407, "lr": 9.576752270948208e-05}
{"train_loss": 0.3349181115627289, "global_step": 36242, "epoch": 407, "lr": 9.576728927076607e-05}
{"train_loss": 0.247721865773201, "global_step": 36243, "epoch": 407, "lr": 9.57670558258972e-05}
{"train_loss": 0.35301920771598816, "global_step": 36244, "epoch": 407, "lr": 9.57668223748755e-05}
{"train_loss": 0.28661608695983887, "global_step": 36245, "epoch": 407, "lr": 9.5766588917701e-05}
{"train_loss": 0.31209221482276917, "global_step": 36246, "epoch": 407, "lr": 9.576635545437376e-05}
{"train_loss": 0.29011082649230957, "global_step": 36247, "epoch": 407, "lr": 9.576612198489379e-05}
{"train_loss": 0.40101805329322815, "global_step": 36248, "epoch": 407, "lr": 9.576588850926111e-05}
{"train_loss": 0.22100858390331268, "global_step": 36249, "epoch": 407, "lr": 9.576565502747577e-05}
{"train_loss": 0.44801896810531616, "global_step": 36250, "epoch": 407, "lr": 9.57654215395378e-05}
{"train_loss": 0.2944876253604889, "global_step": 36251, "epoch": 407, "lr": 9.576518804544723e-05}
{"train_loss": 0.2977611720561981, "global_step": 36252, "epoch": 407, "lr": 9.576495454520407e-05}
{"train_loss": 0.3513522148132324, "global_step": 36253, "epoch": 407, "lr": 9.576472103880839e-05}
{"train_loss": 0.3351820409297943, "global_step": 36254, "epoch": 407, "lr": 9.57644875262602e-05}
{"train_loss": 0.3327779173851013, "global_step": 36255, "epoch": 407, "lr": 9.57642540075595e-05}
{"train_loss": 0.3975740671157837, "global_step": 36256, "epoch": 407, "lr": 9.576402048270638e-05}
{"train_loss": 0.3248729705810547, "global_step": 36257, "epoch": 407, "lr": 9.576378695170085e-05}
{"train_loss": 0.2498355209827423, "global_step": 36258, "epoch": 407, "lr": 9.576355341454292e-05}
{"train_loss": 0.2631843090057373, "global_step": 36259, "epoch": 407, "lr": 9.576331987123266e-05}
{"train_loss": 0.2987235486507416, "global_step": 36260, "epoch": 407, "lr": 9.576308632177007e-05}
{"train_loss": 0.2687572240829468, "global_step": 36261, "epoch": 407, "lr": 9.576285276615517e-05}
{"train_loss": 0.3990005552768707, "global_step": 36262, "epoch": 407, "lr": 9.576261920438805e-05}
{"train_loss": 0.33169054985046387, "global_step": 36263, "epoch": 407, "lr": 9.576238563646868e-05}
{"train_loss": 0.34739741683006287, "global_step": 36264, "epoch": 407, "lr": 9.576215206239712e-05}
{"train_loss": 0.37094876170158386, "global_step": 36265, "epoch": 407, "lr": 9.57619184821734e-05}
{"train_loss": 0.25975656509399414, "global_step": 36266, "epoch": 407, "lr": 9.576168489579753e-05}
{"train_loss": 0.34598878026008606, "global_step": 36267, "epoch": 407, "lr": 9.576145130326957e-05}
{"train_loss": 0.26747044920921326, "global_step": 36268, "epoch": 407, "lr": 9.576121770458955e-05}
{"train_loss": 0.3481653332710266, "global_step": 36269, "epoch": 407, "lr": 9.576098409975749e-05}
{"train_loss": 0.35835233330726624, "global_step": 36270, "epoch": 407, "lr": 9.576075048877341e-05}
{"train_loss": 0.26871106028556824, "global_step": 36271, "epoch": 407, "lr": 9.576051687163736e-05}
{"train_loss": 0.33991768956184387, "global_step": 36272, "epoch": 407, "lr": 9.576028324834938e-05}
{"train_loss": 0.37438973784446716, "global_step": 36273, "epoch": 407, "lr": 9.576004961890948e-05}
{"train_loss": 0.24837230145931244, "global_step": 36274, "epoch": 407, "lr": 9.575981598331769e-05}
{"train_loss": 0.3459884226322174, "global_step": 36275, "epoch": 407, "lr": 9.575958234157405e-05}
{"train_loss": 0.3536165952682495, "global_step": 36276, "epoch": 407, "lr": 9.57593486936786e-05}
{"train_loss": 0.3063710033893585, "global_step": 36277, "epoch": 407, "lr": 9.575911503963136e-05}
{"train_loss": 0.3386290371417999, "global_step": 36278, "epoch": 407, "lr": 9.575888137943237e-05}
{"train_loss": 0.39120376110076904, "global_step": 36279, "epoch": 407, "lr": 9.575864771308166e-05}
{"train_loss": 0.33511826395988464, "global_step": 36280, "epoch": 407, "lr": 9.575841404057923e-05}
{"train_loss": 0.26917240023612976, "global_step": 36281, "epoch": 407, "lr": 9.575818036192517e-05}
{"train_loss": 0.2762303352355957, "global_step": 36282, "epoch": 407, "lr": 9.575794667711947e-05}
{"train_loss": 0.33416154980659485, "global_step": 36283, "epoch": 407, "lr": 9.575771298616217e-05}
{"train_loss": 0.24150580167770386, "global_step": 36284, "epoch": 407, "lr": 9.575747928905331e-05}
{"train_loss": 0.4017956852912903, "global_step": 36285, "epoch": 407, "lr": 9.575724558579291e-05}
{"train_loss": 0.33990538120269775, "global_step": 36286, "epoch": 407, "lr": 9.5757011876381e-05}
{"train_loss": 0.38087600469589233, "global_step": 36287, "epoch": 407, "lr": 9.575677816081763e-05}
{"train_loss": 0.342946320772171, "global_step": 36288, "epoch": 407, "lr": 9.57565444391028e-05}
{"train_loss": 0.3936527967453003, "global_step": 36289, "epoch": 407, "lr": 9.575631071123659e-05}
{"train_loss": 0.31021031737327576, "global_step": 36290, "epoch": 407, "lr": 9.575607697721897e-05}
{"train_loss": 0.28930696845054626, "global_step": 36291, "epoch": 407, "lr": 9.575584323705003e-05}
{"train_loss": 0.3370100259780884, "global_step": 36292, "epoch": 407, "lr": 9.575560949072974e-05}
{"train_loss": 0.41738149523735046, "global_step": 36293, "epoch": 407, "lr": 9.57553757382582e-05}
{"train_loss": 0.38775041699409485, "global_step": 36294, "epoch": 407, "lr": 9.575514197963539e-05}
{"train_loss": 0.3435174524784088, "global_step": 36295, "epoch": 407, "lr": 9.575490821486135e-05}
{"train_loss": 0.36431336402893066, "global_step": 36296, "epoch": 407, "lr": 9.575467444393616e-05}
{"train_loss": 0.3107030391693115, "global_step": 36297, "epoch": 407, "lr": 9.575444066685976e-05}
{"train_loss": 0.2369811087846756, "global_step": 36298, "epoch": 407, "lr": 9.575420688363227e-05}
{"train_loss": 0.28839272260665894, "global_step": 36299, "epoch": 407, "lr": 9.575397309425367e-05}
{"train_loss": 0.330233097076416, "global_step": 36300, "epoch": 407, "lr": 9.575373929872401e-05}
{"train_loss": 0.30081260204315186, "global_step": 36301, "epoch": 407, "lr": 9.575350549704332e-05}
{"train_loss": 0.2938833236694336, "global_step": 36302, "epoch": 407, "lr": 9.575327168921162e-05}
{"train_loss": 0.462978720664978, "global_step": 36303, "epoch": 407, "lr": 9.575303787522896e-05}
{"train_loss": 0.3411710858345032, "global_step": 36304, "epoch": 407, "lr": 9.575280405509536e-05}
{"train_loss": 0.3188028633594513, "global_step": 36305, "epoch": 407, "lr": 9.575257022881084e-05}
{"train_loss": 0.26089587807655334, "global_step": 36306, "epoch": 407, "lr": 9.575233639637546e-05}
{"train_loss": 0.35146772861480713, "global_step": 36307, "epoch": 407, "lr": 9.575210255778923e-05}
{"train_loss": 0.33876150846481323, "global_step": 36308, "epoch": 407, "lr": 9.575186871305219e-05}
{"train_loss": 0.32510727643966675, "global_step": 36309, "epoch": 407, "lr": 9.575163486216436e-05}
{"train_loss": 0.25667762756347656, "global_step": 36310, "epoch": 407, "lr": 9.575140100512579e-05}
{"train_loss": 0.32594693309805367, "global_step": 36311, "epoch": 407, "lr": 9.57511671419365e-05, "val_loss": 2.6398909091949463}
{"train_loss": 0.32279446721076965, "global_step": 36312, "epoch": 408, "lr": 9.575093327259652e-05}
{"train_loss": 0.35087403655052185, "global_step": 36313, "epoch": 408, "lr": 9.57506993971059e-05}
{"train_loss": 0.365020215511322, "global_step": 36314, "epoch": 408, "lr": 9.575046551546463e-05}
{"train_loss": 0.3096925914287567, "global_step": 36315, "epoch": 408, "lr": 9.57502316276728e-05}
{"train_loss": 0.2830446660518646, "global_step": 36316, "epoch": 408, "lr": 9.574999773373037e-05}
{"train_loss": 0.37008681893348694, "global_step": 36317, "epoch": 408, "lr": 9.574976383363743e-05}
{"train_loss": 0.3662145137786865, "global_step": 36318, "epoch": 408, "lr": 9.574952992739401e-05}
{"train_loss": 0.27577653527259827, "global_step": 36319, "epoch": 408, "lr": 9.574929601500009e-05}
{"train_loss": 0.3257114887237549, "global_step": 36320, "epoch": 408, "lr": 9.574906209645575e-05}
{"train_loss": 0.3568848967552185, "global_step": 36321, "epoch": 408, "lr": 9.5748828171761e-05}
{"train_loss": 0.32710424065589905, "global_step": 36322, "epoch": 408, "lr": 9.57485942409159e-05}
{"train_loss": 0.38093292713165283, "global_step": 36323, "epoch": 408, "lr": 9.574836030392045e-05}
{"train_loss": 0.2943614423274994, "global_step": 36324, "epoch": 408, "lr": 9.574812636077469e-05}
{"train_loss": 0.2818737030029297, "global_step": 36325, "epoch": 408, "lr": 9.574789241147863e-05}
{"train_loss": 0.32369089126586914, "global_step": 36326, "epoch": 408, "lr": 9.574765845603234e-05}
{"train_loss": 0.3429035544395447, "global_step": 36327, "epoch": 408, "lr": 9.574742449443583e-05}
{"train_loss": 0.3193054795265198, "global_step": 36328, "epoch": 408, "lr": 9.574719052668914e-05}
{"train_loss": 0.3200678527355194, "global_step": 36329, "epoch": 408, "lr": 9.57469565527923e-05}
{"train_loss": 0.2735491991043091, "global_step": 36330, "epoch": 408, "lr": 9.574672257274535e-05}
{"train_loss": 0.42927348613739014, "global_step": 36331, "epoch": 408, "lr": 9.574648858654828e-05}
{"train_loss": 0.338577538728714, "global_step": 36332, "epoch": 408, "lr": 9.574625459420118e-05}
{"train_loss": 0.3574012219905853, "global_step": 36333, "epoch": 408, "lr": 9.574602059570404e-05}
{"train_loss": 0.3293127417564392, "global_step": 36334, "epoch": 408, "lr": 9.574578659105691e-05}
{"train_loss": 0.32555150985717773, "global_step": 36335, "epoch": 408, "lr": 9.574555258025981e-05}
{"train_loss": 0.3153442144393921, "global_step": 36336, "epoch": 408, "lr": 9.574531856331279e-05}
{"train_loss": 0.3415803611278534, "global_step": 36337, "epoch": 408, "lr": 9.574508454021586e-05}
{"train_loss": 0.32685399055480957, "global_step": 36338, "epoch": 408, "lr": 9.574485051096906e-05}
{"train_loss": 0.31153133511543274, "global_step": 36339, "epoch": 408, "lr": 9.574461647557244e-05}
{"train_loss": 0.43663498759269714, "global_step": 36340, "epoch": 408, "lr": 9.574438243402599e-05}
{"train_loss": 0.3219691812992096, "global_step": 36341, "epoch": 408, "lr": 9.574414838632978e-05}
{"train_loss": 0.3247838616371155, "global_step": 36342, "epoch": 408, "lr": 9.574391433248383e-05}
{"train_loss": 0.27261877059936523, "global_step": 36343, "epoch": 408, "lr": 9.574368027248816e-05}
{"train_loss": 0.394361287355423, "global_step": 36344, "epoch": 408, "lr": 9.574344620634281e-05}
{"train_loss": 0.26735568046569824, "global_step": 36345, "epoch": 408, "lr": 9.57432121340478e-05}
{"train_loss": 0.41522225737571716, "global_step": 36346, "epoch": 408, "lr": 9.574297805560319e-05}
{"train_loss": 0.34380626678466797, "global_step": 36347, "epoch": 408, "lr": 9.574274397100899e-05}
{"train_loss": 0.33744391798973083, "global_step": 36348, "epoch": 408, "lr": 9.574250988026523e-05}
{"train_loss": 0.3533349633216858, "global_step": 36349, "epoch": 408, "lr": 9.574227578337196e-05}
{"train_loss": 0.3228709101676941, "global_step": 36350, "epoch": 408, "lr": 9.574204168032918e-05}
{"train_loss": 0.3315267264842987, "global_step": 36351, "epoch": 408, "lr": 9.574180757113696e-05}
{"train_loss": 0.2638423442840576, "global_step": 36352, "epoch": 408, "lr": 9.574157345579529e-05}
{"train_loss": 0.4112244248390198, "global_step": 36353, "epoch": 408, "lr": 9.574133933430424e-05}
{"train_loss": 0.3664218783378601, "global_step": 36354, "epoch": 408, "lr": 9.574110520666382e-05}
{"train_loss": 0.2970784604549408, "global_step": 36355, "epoch": 408, "lr": 9.574087107287406e-05}
{"train_loss": 0.31429603695869446, "global_step": 36356, "epoch": 408, "lr": 9.5740636932935e-05}
{"train_loss": 0.2863196134567261, "global_step": 36357, "epoch": 408, "lr": 9.574040278684667e-05}
{"train_loss": 0.3374565541744232, "global_step": 36358, "epoch": 408, "lr": 9.574016863460911e-05}
{"train_loss": 0.3709664046764374, "global_step": 36359, "epoch": 408, "lr": 9.573993447622234e-05}
{"train_loss": 0.42273589968681335, "global_step": 36360, "epoch": 408, "lr": 9.573970031168638e-05}
{"train_loss": 0.38456493616104126, "global_step": 36361, "epoch": 408, "lr": 9.573946614100127e-05}
{"train_loss": 0.31011471152305603, "global_step": 36362, "epoch": 408, "lr": 9.573923196416706e-05}
{"train_loss": 0.31042996048927307, "global_step": 36363, "epoch": 408, "lr": 9.573899778118376e-05}
{"train_loss": 0.3241339921951294, "global_step": 36364, "epoch": 408, "lr": 9.573876359205143e-05}
{"train_loss": 0.3507629930973053, "global_step": 36365, "epoch": 408, "lr": 9.573852939677006e-05}
{"train_loss": 0.3598004877567291, "global_step": 36366, "epoch": 408, "lr": 9.57382951953397e-05}
{"train_loss": 0.4397446811199188, "global_step": 36367, "epoch": 408, "lr": 9.57380609877604e-05}
{"train_loss": 0.4026378393173218, "global_step": 36368, "epoch": 408, "lr": 9.573782677403215e-05}
{"train_loss": 0.4024833142757416, "global_step": 36369, "epoch": 408, "lr": 9.573759255415504e-05}
{"train_loss": 0.30109456181526184, "global_step": 36370, "epoch": 408, "lr": 9.573735832812905e-05}
{"train_loss": 0.297090083360672, "global_step": 36371, "epoch": 408, "lr": 9.573712409595424e-05}
{"train_loss": 0.2821558713912964, "global_step": 36372, "epoch": 408, "lr": 9.573688985763061e-05}
{"train_loss": 0.4006633758544922, "global_step": 36373, "epoch": 408, "lr": 9.573665561315822e-05}
{"train_loss": 0.3742658793926239, "global_step": 36374, "epoch": 408, "lr": 9.57364213625371e-05}
{"train_loss": 0.28087764978408813, "global_step": 36375, "epoch": 408, "lr": 9.573618710576727e-05}
{"train_loss": 0.3571677803993225, "global_step": 36376, "epoch": 408, "lr": 9.573595284284878e-05}
{"train_loss": 0.4122580885887146, "global_step": 36377, "epoch": 408, "lr": 9.573571857378163e-05}
{"train_loss": 0.3859670162200928, "global_step": 36378, "epoch": 408, "lr": 9.57354842985659e-05}
{"train_loss": 0.2736053466796875, "global_step": 36379, "epoch": 408, "lr": 9.573525001720157e-05}
{"train_loss": 0.3244468569755554, "global_step": 36380, "epoch": 408, "lr": 9.573501572968869e-05}
{"train_loss": 0.21404539048671722, "global_step": 36381, "epoch": 408, "lr": 9.573478143602728e-05}
{"train_loss": 0.33852383494377136, "global_step": 36382, "epoch": 408, "lr": 9.573454713621741e-05}
{"train_loss": 0.3735432028770447, "global_step": 36383, "epoch": 408, "lr": 9.573431283025908e-05}
{"train_loss": 0.35446271300315857, "global_step": 36384, "epoch": 408, "lr": 9.573407851815234e-05}
{"train_loss": 0.34771546721458435, "global_step": 36385, "epoch": 408, "lr": 9.573384419989719e-05}
{"train_loss": 0.3395117223262787, "global_step": 36386, "epoch": 408, "lr": 9.573360987549371e-05}
{"train_loss": 0.28588199615478516, "global_step": 36387, "epoch": 408, "lr": 9.573337554494188e-05}
{"train_loss": 0.3609384596347809, "global_step": 36388, "epoch": 408, "lr": 9.573314120824177e-05}
{"train_loss": 0.3230849504470825, "global_step": 36389, "epoch": 408, "lr": 9.573290686539338e-05}
{"train_loss": 0.39118558168411255, "global_step": 36390, "epoch": 408, "lr": 9.573267251639678e-05}
{"train_loss": 0.37134650349617004, "global_step": 36391, "epoch": 408, "lr": 9.573243816125196e-05}
{"train_loss": 0.2597002387046814, "global_step": 36392, "epoch": 408, "lr": 9.573220379995898e-05}
{"train_loss": 0.2345508635044098, "global_step": 36393, "epoch": 408, "lr": 9.573196943251786e-05}
{"train_loss": 0.3626191020011902, "global_step": 36394, "epoch": 408, "lr": 9.573173505892862e-05}
{"train_loss": 0.3192426264286041, "global_step": 36395, "epoch": 408, "lr": 9.573150067919134e-05}
{"train_loss": 0.37100905179977417, "global_step": 36396, "epoch": 408, "lr": 9.573126629330599e-05}
{"train_loss": 0.296764612197876, "global_step": 36397, "epoch": 408, "lr": 9.573103190127263e-05}
{"train_loss": 0.42802026867866516, "global_step": 36398, "epoch": 408, "lr": 9.573079750309131e-05}
{"train_loss": 0.4256768226623535, "global_step": 36399, "epoch": 408, "lr": 9.573056309876203e-05}
{"train_loss": 0.3383700079127644, "global_step": 36400, "epoch": 408, "lr": 9.573032868828482e-05, "val_loss": 2.567014217376709}
{"train_loss": 0.39560186862945557, "global_step": 36401, "epoch": 409, "lr": 9.573009427165973e-05}
{"train_loss": 0.25229611992836, "global_step": 36402, "epoch": 409, "lr": 9.57298598488868e-05}
{"train_loss": 0.23868891596794128, "global_step": 36403, "epoch": 409, "lr": 9.572962541996604e-05}
{"train_loss": 0.27164921164512634, "global_step": 36404, "epoch": 409, "lr": 9.572939098489748e-05}
{"train_loss": 0.3841173052787781, "global_step": 36405, "epoch": 409, "lr": 9.572915654368118e-05}
{"train_loss": 0.3322530686855316, "global_step": 36406, "epoch": 409, "lr": 9.572892209631714e-05}
{"train_loss": 0.3503480553627014, "global_step": 36407, "epoch": 409, "lr": 9.57286876428054e-05}
{"train_loss": 0.352046400308609, "global_step": 36408, "epoch": 409, "lr": 9.5728453183146e-05}
{"train_loss": 0.29592928290367126, "global_step": 36409, "epoch": 409, "lr": 9.572821871733897e-05}
{"train_loss": 0.2978883683681488, "global_step": 36410, "epoch": 409, "lr": 9.572798424538435e-05}
{"train_loss": 0.3639163374900818, "global_step": 36411, "epoch": 409, "lr": 9.572774976728214e-05}
{"train_loss": 0.3265443742275238, "global_step": 36412, "epoch": 409, "lr": 9.572751528303239e-05}
{"train_loss": 0.36333203315734863, "global_step": 36413, "epoch": 409, "lr": 9.572728079263514e-05}
{"train_loss": 0.31231480836868286, "global_step": 36414, "epoch": 409, "lr": 9.572704629609042e-05}
{"train_loss": 0.32436904311180115, "global_step": 36415, "epoch": 409, "lr": 9.572681179339823e-05}
{"train_loss": 0.2938733696937561, "global_step": 36416, "epoch": 409, "lr": 9.572657728455867e-05}
{"train_loss": 0.3458239734172821, "global_step": 36417, "epoch": 409, "lr": 9.57263427695717e-05}
{"train_loss": 0.32935866713523865, "global_step": 36418, "epoch": 409, "lr": 9.572610824843739e-05}
{"train_loss": 0.24255052208900452, "global_step": 36419, "epoch": 409, "lr": 9.572587372115575e-05}
{"train_loss": 0.2753288149833679, "global_step": 36420, "epoch": 409, "lr": 9.572563918772685e-05}
{"train_loss": 0.27907952666282654, "global_step": 36421, "epoch": 409, "lr": 9.572540464815068e-05}
{"train_loss": 0.2606138288974762, "global_step": 36422, "epoch": 409, "lr": 9.572517010242727e-05}
{"train_loss": 0.34237369894981384, "global_step": 36423, "epoch": 409, "lr": 9.57249355505567e-05}
{"train_loss": 0.4022366404533386, "global_step": 36424, "epoch": 409, "lr": 9.572470099253894e-05}
{"train_loss": 0.34495770931243896, "global_step": 36425, "epoch": 409, "lr": 9.572446642837407e-05}
{"train_loss": 0.32756054401397705, "global_step": 36426, "epoch": 409, "lr": 9.57242318580621e-05}
{"train_loss": 0.4598569869995117, "global_step": 36427, "epoch": 409, "lr": 9.572399728160306e-05}
{"train_loss": 0.39962607622146606, "global_step": 36428, "epoch": 409, "lr": 9.5723762698997e-05}
{"train_loss": 0.39627522230148315, "global_step": 36429, "epoch": 409, "lr": 9.572352811024391e-05}
{"train_loss": 0.33280062675476074, "global_step": 36430, "epoch": 409, "lr": 9.572329351534386e-05}
{"train_loss": 0.3845212757587433, "global_step": 36431, "epoch": 409, "lr": 9.572305891429688e-05}
{"train_loss": 0.3351316750049591, "global_step": 36432, "epoch": 409, "lr": 9.572282430710299e-05}
{"train_loss": 0.3125682771205902, "global_step": 36433, "epoch": 409, "lr": 9.572258969376222e-05}
{"train_loss": 0.31767991185188293, "global_step": 36434, "epoch": 409, "lr": 9.57223550742746e-05}
{"train_loss": 0.371888130903244, "global_step": 36435, "epoch": 409, "lr": 9.572212044864017e-05}
{"train_loss": 0.31316953897476196, "global_step": 36436, "epoch": 409, "lr": 9.572188581685897e-05}
{"train_loss": 0.3298827111721039, "global_step": 36437, "epoch": 409, "lr": 9.5721651178931e-05}
{"train_loss": 0.3074330985546112, "global_step": 36438, "epoch": 409, "lr": 9.572141653485632e-05}
{"train_loss": 0.2134149968624115, "global_step": 36439, "epoch": 409, "lr": 9.572118188463494e-05}
{"train_loss": 0.27871158719062805, "global_step": 36440, "epoch": 409, "lr": 9.572094722826693e-05}
{"train_loss": 0.3411174714565277, "global_step": 36441, "epoch": 409, "lr": 9.572071256575229e-05}
{"train_loss": 0.3129587769508362, "global_step": 36442, "epoch": 409, "lr": 9.572047789709104e-05}
{"train_loss": 0.324223130941391, "global_step": 36443, "epoch": 409, "lr": 9.572024322228325e-05}
{"train_loss": 0.15557965636253357, "global_step": 36444, "epoch": 409, "lr": 9.572000854132892e-05}
{"train_loss": 0.3783550262451172, "global_step": 36445, "epoch": 409, "lr": 9.57197738542281e-05}
{"train_loss": 0.3786589503288269, "global_step": 36446, "epoch": 409, "lr": 9.57195391609808e-05}
{"train_loss": 0.3568280339241028, "global_step": 36447, "epoch": 409, "lr": 9.571930446158708e-05}
{"train_loss": 0.2884179949760437, "global_step": 36448, "epoch": 409, "lr": 9.571906975604694e-05}
{"train_loss": 0.22029748558998108, "global_step": 36449, "epoch": 409, "lr": 9.571883504436043e-05}
{"train_loss": 0.2727970480918884, "global_step": 36450, "epoch": 409, "lr": 9.571860032652759e-05}
{"train_loss": 0.43885862827301025, "global_step": 36451, "epoch": 409, "lr": 9.571836560254845e-05}
{"train_loss": 0.3214462399482727, "global_step": 36452, "epoch": 409, "lr": 9.5718130872423e-05}
{"train_loss": 0.30262744426727295, "global_step": 36453, "epoch": 409, "lr": 9.571789613615135e-05}
{"train_loss": 0.2604058086872101, "global_step": 36454, "epoch": 409, "lr": 9.571766139373345e-05}
{"train_loss": 0.2531205415725708, "global_step": 36455, "epoch": 409, "lr": 9.571742664516938e-05}
{"train_loss": 0.4536588191986084, "global_step": 36456, "epoch": 409, "lr": 9.571719189045916e-05}
{"train_loss": 0.38627198338508606, "global_step": 36457, "epoch": 409, "lr": 9.571695712960282e-05}
{"train_loss": 0.35580503940582275, "global_step": 36458, "epoch": 409, "lr": 9.571672236260039e-05}
{"train_loss": 0.34329599142074585, "global_step": 36459, "epoch": 409, "lr": 9.57164875894519e-05}
{"train_loss": 0.24985672533512115, "global_step": 36460, "epoch": 409, "lr": 9.571625281015739e-05}
{"train_loss": 0.2922854721546173, "global_step": 36461, "epoch": 409, "lr": 9.57160180247169e-05}
{"train_loss": 0.3064812421798706, "global_step": 36462, "epoch": 409, "lr": 9.571578323313043e-05}
{"train_loss": 0.3609725832939148, "global_step": 36463, "epoch": 409, "lr": 9.571554843539804e-05}
{"train_loss": 0.369577020406723, "global_step": 36464, "epoch": 409, "lr": 9.571531363151975e-05}
{"train_loss": 0.3734542429447174, "global_step": 36465, "epoch": 409, "lr": 9.571507882149558e-05}
{"train_loss": 0.31807371973991394, "global_step": 36466, "epoch": 409, "lr": 9.57148440053256e-05}
{"train_loss": 0.332658976316452, "global_step": 36467, "epoch": 409, "lr": 9.57146091830098e-05}
{"train_loss": 0.32325634360313416, "global_step": 36468, "epoch": 409, "lr": 9.571437435454821e-05}
{"train_loss": 0.2927927076816559, "global_step": 36469, "epoch": 409, "lr": 9.571413951994091e-05}
{"train_loss": 0.39060690999031067, "global_step": 36470, "epoch": 409, "lr": 9.571390467918791e-05}
{"train_loss": 0.3635891079902649, "global_step": 36471, "epoch": 409, "lr": 9.57136698322892e-05}
{"train_loss": 0.24010057747364044, "global_step": 36472, "epoch": 409, "lr": 9.571343497924487e-05}
{"train_loss": 0.287894070148468, "global_step": 36473, "epoch": 409, "lr": 9.571320012005491e-05}
{"train_loss": 0.2233157604932785, "global_step": 36474, "epoch": 409, "lr": 9.571296525471937e-05}
{"train_loss": 0.4064062833786011, "global_step": 36475, "epoch": 409, "lr": 9.571273038323829e-05}
{"train_loss": 0.2956238090991974, "global_step": 36476, "epoch": 409, "lr": 9.571249550561167e-05}
{"train_loss": 0.32698407769203186, "global_step": 36477, "epoch": 409, "lr": 9.571226062183957e-05}
{"train_loss": 0.3023354113101959, "global_step": 36478, "epoch": 409, "lr": 9.571202573192201e-05}
{"train_loss": 0.32162174582481384, "global_step": 36479, "epoch": 409, "lr": 9.571179083585904e-05}
{"train_loss": 0.37376219034194946, "global_step": 36480, "epoch": 409, "lr": 9.571155593365067e-05}
{"train_loss": 0.38162386417388916, "global_step": 36481, "epoch": 409, "lr": 9.571132102529694e-05}
{"train_loss": 0.32242435216903687, "global_step": 36482, "epoch": 409, "lr": 9.571108611079787e-05}
{"train_loss": 0.23582880198955536, "global_step": 36483, "epoch": 409, "lr": 9.571085119015352e-05}
{"train_loss": 0.3408971130847931, "global_step": 36484, "epoch": 409, "lr": 9.571061626336388e-05}
{"train_loss": 0.2607942819595337, "global_step": 36485, "epoch": 409, "lr": 9.571038133042902e-05}
{"train_loss": 0.3608427047729492, "global_step": 36486, "epoch": 409, "lr": 9.571014639134897e-05}
{"train_loss": 0.3173828125, "global_step": 36487, "epoch": 409, "lr": 9.570991144612372e-05}
{"train_loss": 0.24495117366313934, "global_step": 36488, "epoch": 409, "lr": 9.570967649475334e-05}
{"train_loss": 0.32325547682435324, "global_step": 36489, "epoch": 409, "lr": 9.570944153723787e-05, "val_loss": 2.6463570594787598}
{"train_loss": 0.31706783175468445, "global_step": 36490, "epoch": 410, "lr": 9.57092065735773e-05}
{"train_loss": 0.3234585225582123, "global_step": 36491, "epoch": 410, "lr": 9.570897160377168e-05}
{"train_loss": 0.27262407541275024, "global_step": 36492, "epoch": 410, "lr": 9.570873662782105e-05}
{"train_loss": 0.29980120062828064, "global_step": 36493, "epoch": 410, "lr": 9.570850164572545e-05}
{"train_loss": 0.26584476232528687, "global_step": 36494, "epoch": 410, "lr": 9.57082666574849e-05}
{"train_loss": 0.31403014063835144, "global_step": 36495, "epoch": 410, "lr": 9.57080316630994e-05}
{"train_loss": 0.39143913984298706, "global_step": 36496, "epoch": 410, "lr": 9.570779666256905e-05}
{"train_loss": 0.32312971353530884, "global_step": 36497, "epoch": 410, "lr": 9.570756165589383e-05}
{"train_loss": 0.25915467739105225, "global_step": 36498, "epoch": 410, "lr": 9.570732664307378e-05}
{"train_loss": 0.3177854120731354, "global_step": 36499, "epoch": 410, "lr": 9.570709162410893e-05}
{"train_loss": 0.2692122459411621, "global_step": 36500, "epoch": 410, "lr": 9.570685659899934e-05}
{"train_loss": 0.31846463680267334, "global_step": 36501, "epoch": 410, "lr": 9.570662156774501e-05}
{"train_loss": 0.25705742835998535, "global_step": 36502, "epoch": 410, "lr": 9.570638653034599e-05}
{"train_loss": 0.40648284554481506, "global_step": 36503, "epoch": 410, "lr": 9.570615148680229e-05}
{"train_loss": 0.30852973461151123, "global_step": 36504, "epoch": 410, "lr": 9.570591643711395e-05}
{"train_loss": 0.2825169563293457, "global_step": 36505, "epoch": 410, "lr": 9.570568138128102e-05}
{"train_loss": 0.40379372239112854, "global_step": 36506, "epoch": 410, "lr": 9.570544631930352e-05}
{"train_loss": 0.2825773358345032, "global_step": 36507, "epoch": 410, "lr": 9.570521125118148e-05}
{"train_loss": 0.23226802051067352, "global_step": 36508, "epoch": 410, "lr": 9.570497617691491e-05}
{"train_loss": 0.27317577600479126, "global_step": 36509, "epoch": 410, "lr": 9.57047410965039e-05}
{"train_loss": 0.31334179639816284, "global_step": 36510, "epoch": 410, "lr": 9.570450600994841e-05}
{"train_loss": 0.3798602223396301, "global_step": 36511, "epoch": 410, "lr": 9.570427091724852e-05}
{"train_loss": 0.3153979480266571, "global_step": 36512, "epoch": 410, "lr": 9.570403581840425e-05}
{"train_loss": 0.32723796367645264, "global_step": 36513, "epoch": 410, "lr": 9.570380071341563e-05}
{"train_loss": 0.3550358712673187, "global_step": 36514, "epoch": 410, "lr": 9.570356560228269e-05}
{"train_loss": 0.2673413157463074, "global_step": 36515, "epoch": 410, "lr": 9.570333048500547e-05}
{"train_loss": 0.3351512849330902, "global_step": 36516, "epoch": 410, "lr": 9.570309536158397e-05}
{"train_loss": 0.3747178018093109, "global_step": 36517, "epoch": 410, "lr": 9.570286023201827e-05}
{"train_loss": 0.34981587529182434, "global_step": 36518, "epoch": 410, "lr": 9.570262509630836e-05}
{"train_loss": 0.3045583963394165, "global_step": 36519, "epoch": 410, "lr": 9.570238995445429e-05}
{"train_loss": 0.3103984296321869, "global_step": 36520, "epoch": 410, "lr": 9.57021548064561e-05}
{"train_loss": 0.3683962821960449, "global_step": 36521, "epoch": 410, "lr": 9.570191965231383e-05}
{"train_loss": 0.3787992596626282, "global_step": 36522, "epoch": 410, "lr": 9.570168449202745e-05}
{"train_loss": 0.40248844027519226, "global_step": 36523, "epoch": 410, "lr": 9.570144932559707e-05}
{"train_loss": 0.2869608700275421, "global_step": 36524, "epoch": 410, "lr": 9.570121415302268e-05}
{"train_loss": 0.3553127646446228, "global_step": 36525, "epoch": 410, "lr": 9.570097897430433e-05}
{"train_loss": 0.3363257646560669, "global_step": 36526, "epoch": 410, "lr": 9.5700743789442e-05}
{"train_loss": 0.3132576048374176, "global_step": 36527, "epoch": 410, "lr": 9.57005085984358e-05}
{"train_loss": 0.2838558256626129, "global_step": 36528, "epoch": 410, "lr": 9.570027340128571e-05}
{"train_loss": 0.29673394560813904, "global_step": 36529, "epoch": 410, "lr": 9.570003819799178e-05}
{"train_loss": 0.35306379199028015, "global_step": 36530, "epoch": 410, "lr": 9.569980298855403e-05}
{"train_loss": 0.5066999197006226, "global_step": 36531, "epoch": 410, "lr": 9.569956777297251e-05}
{"train_loss": 0.319837749004364, "global_step": 36532, "epoch": 410, "lr": 9.569933255124723e-05}
{"train_loss": 0.3515697419643402, "global_step": 36533, "epoch": 410, "lr": 9.569909732337823e-05}
{"train_loss": 0.31722334027290344, "global_step": 36534, "epoch": 410, "lr": 9.569886208936556e-05}
{"train_loss": 0.39456892013549805, "global_step": 36535, "epoch": 410, "lr": 9.569862684920922e-05}
{"train_loss": 0.3870399594306946, "global_step": 36536, "epoch": 410, "lr": 9.569839160290926e-05}
{"train_loss": 0.33522549271583557, "global_step": 36537, "epoch": 410, "lr": 9.56981563504657e-05}
{"train_loss": 0.2921609878540039, "global_step": 36538, "epoch": 410, "lr": 9.569792109187858e-05}
{"train_loss": 0.3517689108848572, "global_step": 36539, "epoch": 410, "lr": 9.569768582714795e-05}
{"train_loss": 0.38206812739372253, "global_step": 36540, "epoch": 410, "lr": 9.569745055627381e-05}
{"train_loss": 0.26824507117271423, "global_step": 36541, "epoch": 410, "lr": 9.569721527925622e-05}
{"train_loss": 0.35374048352241516, "global_step": 36542, "epoch": 410, "lr": 9.569697999609517e-05}
{"train_loss": 0.3053431212902069, "global_step": 36543, "epoch": 410, "lr": 9.569674470679073e-05}
{"train_loss": 0.3685128390789032, "global_step": 36544, "epoch": 410, "lr": 9.569650941134293e-05}
{"train_loss": 0.36641642451286316, "global_step": 36545, "epoch": 410, "lr": 9.569627410975177e-05}
{"train_loss": 0.35676705837249756, "global_step": 36546, "epoch": 410, "lr": 9.569603880201732e-05}
{"train_loss": 0.43812060356140137, "global_step": 36547, "epoch": 410, "lr": 9.569580348813959e-05}
{"train_loss": 0.20212292671203613, "global_step": 36548, "epoch": 410, "lr": 9.569556816811862e-05}
{"train_loss": 0.3527556359767914, "global_step": 36549, "epoch": 410, "lr": 9.569533284195444e-05}
{"train_loss": 0.3484914004802704, "global_step": 36550, "epoch": 410, "lr": 9.569509750964707e-05}
{"train_loss": 0.3487703800201416, "global_step": 36551, "epoch": 410, "lr": 9.569486217119655e-05}
{"train_loss": 0.3658091425895691, "global_step": 36552, "epoch": 410, "lr": 9.569462682660292e-05}
{"train_loss": 0.37261447310447693, "global_step": 36553, "epoch": 410, "lr": 9.569439147586619e-05}
{"train_loss": 0.2648026645183563, "global_step": 36554, "epoch": 410, "lr": 9.569415611898643e-05}
{"train_loss": 0.37147077918052673, "global_step": 36555, "epoch": 410, "lr": 9.569392075596363e-05}
{"train_loss": 0.3102271556854248, "global_step": 36556, "epoch": 410, "lr": 9.569368538679783e-05}
{"train_loss": 0.3445076048374176, "global_step": 36557, "epoch": 410, "lr": 9.569345001148911e-05}
{"train_loss": 0.2909817695617676, "global_step": 36558, "epoch": 410, "lr": 9.569321463003742e-05}
{"train_loss": 0.3988327980041504, "global_step": 36559, "epoch": 410, "lr": 9.569297924244286e-05}
{"train_loss": 0.3829253315925598, "global_step": 36560, "epoch": 410, "lr": 9.569274384870544e-05}
{"train_loss": 0.3643750846385956, "global_step": 36561, "epoch": 410, "lr": 9.569250844882516e-05}
{"train_loss": 0.44875240325927734, "global_step": 36562, "epoch": 410, "lr": 9.569227304280209e-05}
{"train_loss": 0.30236005783081055, "global_step": 36563, "epoch": 410, "lr": 9.569203763063626e-05}
{"train_loss": 0.3846000134944916, "global_step": 36564, "epoch": 410, "lr": 9.56918022123277e-05}
{"train_loss": 0.3604047894477844, "global_step": 36565, "epoch": 410, "lr": 9.569156678787641e-05}
{"train_loss": 0.3558667302131653, "global_step": 36566, "epoch": 410, "lr": 9.569133135728245e-05}
{"train_loss": 0.28987085819244385, "global_step": 36567, "epoch": 410, "lr": 9.569109592054585e-05}
{"train_loss": 0.36237627267837524, "global_step": 36568, "epoch": 410, "lr": 9.569086047766666e-05}
{"train_loss": 0.2721007466316223, "global_step": 36569, "epoch": 410, "lr": 9.569062502864488e-05}
{"train_loss": 0.27753737568855286, "global_step": 36570, "epoch": 410, "lr": 9.569038957348053e-05}
{"train_loss": 0.3283665180206299, "global_step": 36571, "epoch": 410, "lr": 9.569015411217369e-05}
{"train_loss": 0.2824717164039612, "global_step": 36572, "epoch": 410, "lr": 9.568991864472436e-05}
{"train_loss": 0.45710188150405884, "global_step": 36573, "epoch": 410, "lr": 9.568968317113258e-05}
{"train_loss": 0.32384082674980164, "global_step": 36574, "epoch": 410, "lr": 9.568944769139836e-05}
{"train_loss": 0.342808336019516, "global_step": 36575, "epoch": 410, "lr": 9.568921220552176e-05}
{"train_loss": 0.29535552859306335, "global_step": 36576, "epoch": 410, "lr": 9.568897671350281e-05}
{"train_loss": 0.3874243497848511, "global_step": 36577, "epoch": 410, "lr": 9.568874121534153e-05}
{"train_loss": 0.3331392310308606, "global_step": 36578, "epoch": 410, "lr": 9.568850571103796e-05, "val_loss": 2.5199151039123535, "train_action_mse_error": 45.807125091552734}
{"train_loss": 0.2692095935344696, "global_step": 36579, "epoch": 411, "lr": 9.568827020059212e-05}
{"train_loss": 0.2885095477104187, "global_step": 36580, "epoch": 411, "lr": 9.568803468400406e-05}
{"train_loss": 0.2716650366783142, "global_step": 36581, "epoch": 411, "lr": 9.56877991612738e-05}
{"train_loss": 0.31986522674560547, "global_step": 36582, "epoch": 411, "lr": 9.568756363240134e-05}
{"train_loss": 0.3785483241081238, "global_step": 36583, "epoch": 411, "lr": 9.568732809738679e-05}
{"train_loss": 0.23218363523483276, "global_step": 36584, "epoch": 411, "lr": 9.568709255623009e-05}
{"train_loss": 0.325737327337265, "global_step": 36585, "epoch": 411, "lr": 9.568685700893137e-05}
{"train_loss": 0.30783212184906006, "global_step": 36586, "epoch": 411, "lr": 9.568662145549056e-05}
{"train_loss": 0.3029947876930237, "global_step": 36587, "epoch": 411, "lr": 9.568638589590778e-05}
{"train_loss": 0.43214163184165955, "global_step": 36588, "epoch": 411, "lr": 9.568615033018299e-05}
{"train_loss": 0.3837830424308777, "global_step": 36589, "epoch": 411, "lr": 9.568591475831627e-05}
{"train_loss": 0.3113674521446228, "global_step": 36590, "epoch": 411, "lr": 9.568567918030763e-05}
{"train_loss": 0.2605760097503662, "global_step": 36591, "epoch": 411, "lr": 9.568544359615711e-05}
{"train_loss": 0.2456921488046646, "global_step": 36592, "epoch": 411, "lr": 9.568520800586473e-05}
{"train_loss": 0.34052225947380066, "global_step": 36593, "epoch": 411, "lr": 9.568497240943055e-05}
{"train_loss": 0.42105576395988464, "global_step": 36594, "epoch": 411, "lr": 9.568473680685457e-05}
{"train_loss": 0.3486267030239105, "global_step": 36595, "epoch": 411, "lr": 9.568450119813683e-05}
{"train_loss": 0.30955567955970764, "global_step": 36596, "epoch": 411, "lr": 9.568426558327738e-05}
{"train_loss": 0.23356303572654724, "global_step": 36597, "epoch": 411, "lr": 9.568402996227622e-05}
{"train_loss": 0.3253706395626068, "global_step": 36598, "epoch": 411, "lr": 9.568379433513342e-05}
{"train_loss": 0.3977133631706238, "global_step": 36599, "epoch": 411, "lr": 9.568355870184897e-05}
{"train_loss": 0.24483738839626312, "global_step": 36600, "epoch": 411, "lr": 9.568332306242294e-05}
{"train_loss": 0.2999238669872284, "global_step": 36601, "epoch": 411, "lr": 9.568308741685531e-05}
{"train_loss": 0.22525916993618011, "global_step": 36602, "epoch": 411, "lr": 9.568285176514619e-05}
{"train_loss": 0.4340815842151642, "global_step": 36603, "epoch": 411, "lr": 9.568261610729554e-05}
{"train_loss": 0.30160418152809143, "global_step": 36604, "epoch": 411, "lr": 9.568238044330343e-05}
{"train_loss": 0.41332367062568665, "global_step": 36605, "epoch": 411, "lr": 9.568214477316988e-05}
{"train_loss": 0.26800283789634705, "global_step": 36606, "epoch": 411, "lr": 9.568190909689491e-05}
{"train_loss": 0.2528952956199646, "global_step": 36607, "epoch": 411, "lr": 9.568167341447858e-05}
{"train_loss": 0.42169663310050964, "global_step": 36608, "epoch": 411, "lr": 9.56814377259209e-05}
{"train_loss": 0.3056585192680359, "global_step": 36609, "epoch": 411, "lr": 9.56812020312219e-05}
{"train_loss": 0.3078872561454773, "global_step": 36610, "epoch": 411, "lr": 9.568096633038164e-05}
{"train_loss": 0.34544146060943604, "global_step": 36611, "epoch": 411, "lr": 9.568073062340011e-05}
{"train_loss": 0.29865679144859314, "global_step": 36612, "epoch": 411, "lr": 9.568049491027737e-05}
{"train_loss": 0.3441029191017151, "global_step": 36613, "epoch": 411, "lr": 9.568025919101344e-05}
{"train_loss": 0.30863964557647705, "global_step": 36614, "epoch": 411, "lr": 9.568002346560835e-05}
{"train_loss": 0.4549177289009094, "global_step": 36615, "epoch": 411, "lr": 9.567978773406216e-05}
{"train_loss": 0.2324458360671997, "global_step": 36616, "epoch": 411, "lr": 9.567955199637487e-05}
{"train_loss": 0.3559515178203583, "global_step": 36617, "epoch": 411, "lr": 9.567931625254652e-05}
{"train_loss": 0.34685295820236206, "global_step": 36618, "epoch": 411, "lr": 9.567908050257714e-05}
{"train_loss": 0.43513259291648865, "global_step": 36619, "epoch": 411, "lr": 9.567884474646676e-05}
{"train_loss": 0.2644641101360321, "global_step": 36620, "epoch": 411, "lr": 9.567860898421542e-05}
{"train_loss": 0.33284932374954224, "global_step": 36621, "epoch": 411, "lr": 9.567837321582316e-05}
{"train_loss": 0.2594190835952759, "global_step": 36622, "epoch": 411, "lr": 9.567813744128999e-05}
{"train_loss": 0.30801427364349365, "global_step": 36623, "epoch": 411, "lr": 9.567790166061595e-05}
{"train_loss": 0.3970711827278137, "global_step": 36624, "epoch": 411, "lr": 9.567766587380107e-05}
{"train_loss": 0.3575785756111145, "global_step": 36625, "epoch": 411, "lr": 9.567743008084538e-05}
{"train_loss": 0.20739373564720154, "global_step": 36626, "epoch": 411, "lr": 9.567719428174893e-05}
{"train_loss": 0.3185732662677765, "global_step": 36627, "epoch": 411, "lr": 9.567695847651172e-05}
{"train_loss": 0.23177269101142883, "global_step": 36628, "epoch": 411, "lr": 9.567672266513381e-05}
{"train_loss": 0.3269363343715668, "global_step": 36629, "epoch": 411, "lr": 9.567648684761524e-05}
{"train_loss": 0.3327322006225586, "global_step": 36630, "epoch": 411, "lr": 9.567625102395599e-05}
{"train_loss": 0.2731291651725769, "global_step": 36631, "epoch": 411, "lr": 9.567601519415614e-05}
{"train_loss": 0.2037709504365921, "global_step": 36632, "epoch": 411, "lr": 9.567577935821571e-05}
{"train_loss": 0.4362299144268036, "global_step": 36633, "epoch": 411, "lr": 9.567554351613472e-05}
{"train_loss": 0.40534719824790955, "global_step": 36634, "epoch": 411, "lr": 9.567530766791322e-05}
{"train_loss": 0.38674360513687134, "global_step": 36635, "epoch": 411, "lr": 9.567507181355122e-05}
{"train_loss": 0.33242300152778625, "global_step": 36636, "epoch": 411, "lr": 9.567483595304879e-05}
{"train_loss": 0.35748720169067383, "global_step": 36637, "epoch": 411, "lr": 9.567460008640589e-05}
{"train_loss": 0.3835172951221466, "global_step": 36638, "epoch": 411, "lr": 9.567436421362263e-05}
{"train_loss": 0.3762381374835968, "global_step": 36639, "epoch": 411, "lr": 9.5674128334699e-05}
{"train_loss": 0.25494611263275146, "global_step": 36640, "epoch": 411, "lr": 9.567389244963504e-05}
{"train_loss": 0.39057305455207825, "global_step": 36641, "epoch": 411, "lr": 9.567365655843078e-05}
{"train_loss": 0.36145859956741333, "global_step": 36642, "epoch": 411, "lr": 9.567342066108625e-05}
{"train_loss": 0.342020183801651, "global_step": 36643, "epoch": 411, "lr": 9.56731847576015e-05}
{"train_loss": 0.3245754837989807, "global_step": 36644, "epoch": 411, "lr": 9.567294884797652e-05}
{"train_loss": 0.3388385772705078, "global_step": 36645, "epoch": 411, "lr": 9.56727129322114e-05}
{"train_loss": 0.4166170358657837, "global_step": 36646, "epoch": 411, "lr": 9.567247701030613e-05}
{"train_loss": 0.3327121436595917, "global_step": 36647, "epoch": 411, "lr": 9.567224108226074e-05}
{"train_loss": 0.4103540778160095, "global_step": 36648, "epoch": 411, "lr": 9.567200514807529e-05}
{"train_loss": 0.27712705731391907, "global_step": 36649, "epoch": 411, "lr": 9.567176920774978e-05}
{"train_loss": 0.263910174369812, "global_step": 36650, "epoch": 411, "lr": 9.567153326128426e-05}
{"train_loss": 0.3383719325065613, "global_step": 36651, "epoch": 411, "lr": 9.567129730867877e-05}
{"train_loss": 0.2806301414966583, "global_step": 36652, "epoch": 411, "lr": 9.567106134993334e-05}
{"train_loss": 0.2711970806121826, "global_step": 36653, "epoch": 411, "lr": 9.567082538504797e-05}
{"train_loss": 0.3833775818347931, "global_step": 36654, "epoch": 411, "lr": 9.567058941402271e-05}
{"train_loss": 0.3308499753475189, "global_step": 36655, "epoch": 411, "lr": 9.567035343685762e-05}
{"train_loss": 0.273263156414032, "global_step": 36656, "epoch": 411, "lr": 9.567011745355269e-05}
{"train_loss": 0.3440370261669159, "global_step": 36657, "epoch": 411, "lr": 9.566988146410798e-05}
{"train_loss": 0.38064831495285034, "global_step": 36658, "epoch": 411, "lr": 9.56696454685235e-05}
{"train_loss": 0.3111657202243805, "global_step": 36659, "epoch": 411, "lr": 9.56694094667993e-05}
{"train_loss": 0.35629305243492126, "global_step": 36660, "epoch": 411, "lr": 9.56691734589354e-05}
{"train_loss": 0.39422962069511414, "global_step": 36661, "epoch": 411, "lr": 9.566893744493184e-05}
{"train_loss": 0.34086519479751587, "global_step": 36662, "epoch": 411, "lr": 9.566870142478866e-05}
{"train_loss": 0.30964088439941406, "global_step": 36663, "epoch": 411, "lr": 9.566846539850585e-05}
{"train_loss": 0.3787165582180023, "global_step": 36664, "epoch": 411, "lr": 9.566822936608348e-05}
{"train_loss": 0.41228187084198, "global_step": 36665, "epoch": 411, "lr": 9.566799332752159e-05}
{"train_loss": 0.3019358515739441, "global_step": 36666, "epoch": 411, "lr": 9.566775728282018e-05}
{"train_loss": 0.3298515369383137, "global_step": 36667, "epoch": 411, "lr": 9.566752123197931e-05, "val_loss": 2.5685713291168213}
{"train_loss": 0.33316344022750854, "global_step": 36668, "epoch": 412, "lr": 9.566728517499899e-05}
{"train_loss": 0.22168564796447754, "global_step": 36669, "epoch": 412, "lr": 9.566704911187927e-05}
{"train_loss": 0.3062278628349304, "global_step": 36670, "epoch": 412, "lr": 9.566681304262016e-05}
{"train_loss": 0.36435410380363464, "global_step": 36671, "epoch": 412, "lr": 9.56665769672217e-05}
{"train_loss": 0.28448230028152466, "global_step": 36672, "epoch": 412, "lr": 9.566634088568394e-05}
{"train_loss": 0.3115348815917969, "global_step": 36673, "epoch": 412, "lr": 9.566610479800689e-05}
{"train_loss": 0.3668391704559326, "global_step": 36674, "epoch": 412, "lr": 9.566586870419059e-05}
{"train_loss": 0.3852200210094452, "global_step": 36675, "epoch": 412, "lr": 9.566563260423507e-05}
{"train_loss": 0.3630746006965637, "global_step": 36676, "epoch": 412, "lr": 9.566539649814035e-05}
{"train_loss": 0.3148069977760315, "global_step": 36677, "epoch": 412, "lr": 9.56651603859065e-05}
{"train_loss": 0.3771108090877533, "global_step": 36678, "epoch": 412, "lr": 9.56649242675335e-05}
{"train_loss": 0.30267277359962463, "global_step": 36679, "epoch": 412, "lr": 9.566468814302142e-05}
{"train_loss": 0.3404802680015564, "global_step": 36680, "epoch": 412, "lr": 9.566445201237028e-05}
{"train_loss": 0.3140491545200348, "global_step": 36681, "epoch": 412, "lr": 9.56642158755801e-05}
{"train_loss": 0.36798998713493347, "global_step": 36682, "epoch": 412, "lr": 9.566397973265093e-05}
{"train_loss": 0.4473624527454376, "global_step": 36683, "epoch": 412, "lr": 9.56637435835828e-05}
{"train_loss": 0.354305624961853, "global_step": 36684, "epoch": 412, "lr": 9.566350742837573e-05}
{"train_loss": 0.29142215847969055, "global_step": 36685, "epoch": 412, "lr": 9.566327126702976e-05}
{"train_loss": 0.4206291437149048, "global_step": 36686, "epoch": 412, "lr": 9.566303509954493e-05}
{"train_loss": 0.28503140807151794, "global_step": 36687, "epoch": 412, "lr": 9.566279892592125e-05}
{"train_loss": 0.32786309719085693, "global_step": 36688, "epoch": 412, "lr": 9.566256274615876e-05}
{"train_loss": 0.305634081363678, "global_step": 36689, "epoch": 412, "lr": 9.56623265602575e-05}
{"train_loss": 0.31094223260879517, "global_step": 36690, "epoch": 412, "lr": 9.56620903682175e-05}
{"train_loss": 0.34120577573776245, "global_step": 36691, "epoch": 412, "lr": 9.566185417003878e-05}
{"train_loss": 0.3130050301551819, "global_step": 36692, "epoch": 412, "lr": 9.566161796572138e-05}
{"train_loss": 0.44392284750938416, "global_step": 36693, "epoch": 412, "lr": 9.566138175526533e-05}
{"train_loss": 0.2689145803451538, "global_step": 36694, "epoch": 412, "lr": 9.566114553867068e-05}
{"train_loss": 0.4479615092277527, "global_step": 36695, "epoch": 412, "lr": 9.566090931593744e-05}
{"train_loss": 0.2799314856529236, "global_step": 36696, "epoch": 412, "lr": 9.566067308706563e-05}
{"train_loss": 0.3220025897026062, "global_step": 36697, "epoch": 412, "lr": 9.566043685205532e-05}
{"train_loss": 0.27844536304473877, "global_step": 36698, "epoch": 412, "lr": 9.566020061090651e-05}
{"train_loss": 0.2948891520500183, "global_step": 36699, "epoch": 412, "lr": 9.565996436361926e-05}
{"train_loss": 0.2517581284046173, "global_step": 36700, "epoch": 412, "lr": 9.565972811019356e-05}
{"train_loss": 0.31598275899887085, "global_step": 36701, "epoch": 412, "lr": 9.565949185062948e-05}
{"train_loss": 0.3440992534160614, "global_step": 36702, "epoch": 412, "lr": 9.565925558492703e-05}
{"train_loss": 0.2861718237400055, "global_step": 36703, "epoch": 412, "lr": 9.565901931308625e-05}
{"train_loss": 0.35963138937950134, "global_step": 36704, "epoch": 412, "lr": 9.565878303510718e-05}
{"train_loss": 0.26613983511924744, "global_step": 36705, "epoch": 412, "lr": 9.565854675098984e-05}
{"train_loss": 0.28313305974006653, "global_step": 36706, "epoch": 412, "lr": 9.565831046073425e-05}
{"train_loss": 0.376414954662323, "global_step": 36707, "epoch": 412, "lr": 9.565807416434048e-05}
{"train_loss": 0.35373398661613464, "global_step": 36708, "epoch": 412, "lr": 9.565783786180853e-05}
{"train_loss": 0.21113625168800354, "global_step": 36709, "epoch": 412, "lr": 9.565760155313843e-05}
{"train_loss": 0.3769013285636902, "global_step": 36710, "epoch": 412, "lr": 9.565736523833024e-05}
{"train_loss": 0.2627031207084656, "global_step": 36711, "epoch": 412, "lr": 9.565712891738397e-05}
{"train_loss": 0.24373547732830048, "global_step": 36712, "epoch": 412, "lr": 9.565689259029964e-05}
{"train_loss": 0.35883766412734985, "global_step": 36713, "epoch": 412, "lr": 9.565665625707732e-05}
{"train_loss": 0.37711408734321594, "global_step": 36714, "epoch": 412, "lr": 9.565641991771699e-05}
{"train_loss": 0.3860889673233032, "global_step": 36715, "epoch": 412, "lr": 9.565618357221873e-05}
{"train_loss": 0.24206387996673584, "global_step": 36716, "epoch": 412, "lr": 9.565594722058255e-05}
{"train_loss": 0.3583829402923584, "global_step": 36717, "epoch": 412, "lr": 9.56557108628085e-05}
{"train_loss": 0.27034398913383484, "global_step": 36718, "epoch": 412, "lr": 9.565547449889656e-05}
{"train_loss": 0.3175335228443146, "global_step": 36719, "epoch": 412, "lr": 9.565523812884682e-05}
{"train_loss": 0.4021129608154297, "global_step": 36720, "epoch": 412, "lr": 9.565500175265929e-05}
{"train_loss": 0.3537100851535797, "global_step": 36721, "epoch": 412, "lr": 9.5654765370334e-05}
{"train_loss": 0.4091227054595947, "global_step": 36722, "epoch": 412, "lr": 9.565452898187098e-05}
{"train_loss": 0.3585778474807739, "global_step": 36723, "epoch": 412, "lr": 9.565429258727026e-05}
{"train_loss": 0.318584144115448, "global_step": 36724, "epoch": 412, "lr": 9.56540561865319e-05}
{"train_loss": 0.39407381415367126, "global_step": 36725, "epoch": 412, "lr": 9.565381977965589e-05}
{"train_loss": 0.3853423595428467, "global_step": 36726, "epoch": 412, "lr": 9.565358336664228e-05}
{"train_loss": 0.36745330691337585, "global_step": 36727, "epoch": 412, "lr": 9.56533469474911e-05}
{"train_loss": 0.31106045842170715, "global_step": 36728, "epoch": 412, "lr": 9.565311052220238e-05}
{"train_loss": 0.249204620718956, "global_step": 36729, "epoch": 412, "lr": 9.565287409077617e-05}
{"train_loss": 0.2903918921947479, "global_step": 36730, "epoch": 412, "lr": 9.565263765321248e-05}
{"train_loss": 0.4216081500053406, "global_step": 36731, "epoch": 412, "lr": 9.565240120951136e-05}
{"train_loss": 0.3076779544353485, "global_step": 36732, "epoch": 412, "lr": 9.565216475967282e-05}
{"train_loss": 0.35562971234321594, "global_step": 36733, "epoch": 412, "lr": 9.56519283036969e-05}
{"train_loss": 0.41448819637298584, "global_step": 36734, "epoch": 412, "lr": 9.565169184158365e-05}
{"train_loss": 0.34669229388237, "global_step": 36735, "epoch": 412, "lr": 9.565145537333306e-05}
{"train_loss": 0.3746253550052643, "global_step": 36736, "epoch": 412, "lr": 9.565121889894522e-05}
{"train_loss": 0.3867449164390564, "global_step": 36737, "epoch": 412, "lr": 9.565098241842011e-05}
{"train_loss": 0.4005344808101654, "global_step": 36738, "epoch": 412, "lr": 9.565074593175778e-05}
{"train_loss": 0.3897257447242737, "global_step": 36739, "epoch": 412, "lr": 9.565050943895828e-05}
{"train_loss": 0.3803512156009674, "global_step": 36740, "epoch": 412, "lr": 9.565027294002161e-05}
{"train_loss": 0.4204757809638977, "global_step": 36741, "epoch": 412, "lr": 9.565003643494783e-05}
{"train_loss": 0.40544918179512024, "global_step": 36742, "epoch": 412, "lr": 9.564979992373696e-05}
{"train_loss": 0.3722854554653168, "global_step": 36743, "epoch": 412, "lr": 9.564956340638901e-05}
{"train_loss": 0.4272708296775818, "global_step": 36744, "epoch": 412, "lr": 9.564932688290405e-05}
{"train_loss": 0.40863221883773804, "global_step": 36745, "epoch": 412, "lr": 9.564909035328209e-05}
{"train_loss": 0.33053386211395264, "global_step": 36746, "epoch": 412, "lr": 9.564885381752316e-05}
{"train_loss": 0.37298882007598877, "global_step": 36747, "epoch": 412, "lr": 9.56486172756273e-05}
{"train_loss": 0.3321288526058197, "global_step": 36748, "epoch": 412, "lr": 9.564838072759454e-05}
{"train_loss": 0.3981535732746124, "global_step": 36749, "epoch": 412, "lr": 9.564814417342491e-05}
{"train_loss": 0.3570238947868347, "global_step": 36750, "epoch": 412, "lr": 9.564790761311847e-05}
{"train_loss": 0.37187960743904114, "global_step": 36751, "epoch": 412, "lr": 9.56476710466752e-05}
{"train_loss": 0.2501792311668396, "global_step": 36752, "epoch": 412, "lr": 9.564743447409516e-05}
{"train_loss": 0.2604944407939911, "global_step": 36753, "epoch": 412, "lr": 9.564719789537838e-05}
{"train_loss": 0.3199126422405243, "global_step": 36754, "epoch": 412, "lr": 9.564696131052489e-05}
{"train_loss": 0.30212733149528503, "global_step": 36755, "epoch": 412, "lr": 9.564672471953471e-05}
{"train_loss": 0.33878887871677954, "global_step": 36756, "epoch": 412, "lr": 9.56464881224079e-05, "val_loss": 2.4897491931915283}
{"train_loss": 0.357267826795578, "global_step": 36757, "epoch": 413, "lr": 9.564625151914449e-05}
{"train_loss": 0.31594738364219666, "global_step": 36758, "epoch": 413, "lr": 9.564601490974446e-05}
{"train_loss": 0.3169851303100586, "global_step": 36759, "epoch": 413, "lr": 9.564577829420791e-05}
{"train_loss": 0.39374202489852905, "global_step": 36760, "epoch": 413, "lr": 9.564554167253484e-05}
{"train_loss": 0.32226958870887756, "global_step": 36761, "epoch": 413, "lr": 9.564530504472526e-05}
{"train_loss": 0.2992224395275116, "global_step": 36762, "epoch": 413, "lr": 9.564506841077925e-05}
{"train_loss": 0.303937703371048, "global_step": 36763, "epoch": 413, "lr": 9.56448317706968e-05}
{"train_loss": 0.32243290543556213, "global_step": 36764, "epoch": 413, "lr": 9.564459512447796e-05}
{"train_loss": 0.27658525109291077, "global_step": 36765, "epoch": 413, "lr": 9.564435847212276e-05}
{"train_loss": 0.1932789385318756, "global_step": 36766, "epoch": 413, "lr": 9.564412181363125e-05}
{"train_loss": 0.2650260329246521, "global_step": 36767, "epoch": 413, "lr": 9.564388514900343e-05}
{"train_loss": 0.37914469838142395, "global_step": 36768, "epoch": 413, "lr": 9.564364847823935e-05}
{"train_loss": 0.28851673007011414, "global_step": 36769, "epoch": 413, "lr": 9.564341180133905e-05}
{"train_loss": 0.30049267411231995, "global_step": 36770, "epoch": 413, "lr": 9.564317511830251e-05}
{"train_loss": 0.4051295518875122, "global_step": 36771, "epoch": 413, "lr": 9.564293842912983e-05}
{"train_loss": 0.26886385679244995, "global_step": 36772, "epoch": 413, "lr": 9.564270173382101e-05}
{"train_loss": 0.28702402114868164, "global_step": 36773, "epoch": 413, "lr": 9.56424650323761e-05}
{"train_loss": 0.3473643660545349, "global_step": 36774, "epoch": 413, "lr": 9.56422283247951e-05}
{"train_loss": 0.3612266778945923, "global_step": 36775, "epoch": 413, "lr": 9.564199161107805e-05}
{"train_loss": 0.27521824836730957, "global_step": 36776, "epoch": 413, "lr": 9.564175489122501e-05}
{"train_loss": 0.26471152901649475, "global_step": 36777, "epoch": 413, "lr": 9.5641518165236e-05}
{"train_loss": 0.26950040459632874, "global_step": 36778, "epoch": 413, "lr": 9.564128143311102e-05}
{"train_loss": 0.3611994981765747, "global_step": 36779, "epoch": 413, "lr": 9.564104469485013e-05}
{"train_loss": 0.3141424357891083, "global_step": 36780, "epoch": 413, "lr": 9.564080795045335e-05}
{"train_loss": 0.37871870398521423, "global_step": 36781, "epoch": 413, "lr": 9.564057119992074e-05}
{"train_loss": 0.41649577021598816, "global_step": 36782, "epoch": 413, "lr": 9.564033444325231e-05}
{"train_loss": 0.29876479506492615, "global_step": 36783, "epoch": 413, "lr": 9.564009768044808e-05}
{"train_loss": 0.31263843178749084, "global_step": 36784, "epoch": 413, "lr": 9.563986091150811e-05}
{"train_loss": 0.3163380026817322, "global_step": 36785, "epoch": 413, "lr": 9.563962413643241e-05}
{"train_loss": 0.2639075219631195, "global_step": 36786, "epoch": 413, "lr": 9.563938735522102e-05}
{"train_loss": 0.35445162653923035, "global_step": 36787, "epoch": 413, "lr": 9.563915056787396e-05}
{"train_loss": 0.31115394830703735, "global_step": 36788, "epoch": 413, "lr": 9.563891377439129e-05}
{"train_loss": 0.3984707295894623, "global_step": 36789, "epoch": 413, "lr": 9.5638676974773e-05}
{"train_loss": 0.4077901244163513, "global_step": 36790, "epoch": 413, "lr": 9.563844016901916e-05}
{"train_loss": 0.35849320888519287, "global_step": 36791, "epoch": 413, "lr": 9.563820335712978e-05}
{"train_loss": 0.3198358714580536, "global_step": 36792, "epoch": 413, "lr": 9.563796653910493e-05}
{"train_loss": 0.2832755744457245, "global_step": 36793, "epoch": 413, "lr": 9.563772971494458e-05}
{"train_loss": 0.36966627836227417, "global_step": 36794, "epoch": 413, "lr": 9.56374928846488e-05}
{"train_loss": 0.3123498260974884, "global_step": 36795, "epoch": 413, "lr": 9.563725604821763e-05}
{"train_loss": 0.2878061830997467, "global_step": 36796, "epoch": 413, "lr": 9.563701920565107e-05}
{"train_loss": 0.3010545074939728, "global_step": 36797, "epoch": 413, "lr": 9.563678235694917e-05}
{"train_loss": 0.3666258454322815, "global_step": 36798, "epoch": 413, "lr": 9.563654550211198e-05}
{"train_loss": 0.37647104263305664, "global_step": 36799, "epoch": 413, "lr": 9.563630864113949e-05}
{"train_loss": 0.32360583543777466, "global_step": 36800, "epoch": 413, "lr": 9.563607177403176e-05}
{"train_loss": 0.42072418332099915, "global_step": 36801, "epoch": 413, "lr": 9.563583490078882e-05}
{"train_loss": 0.3339439928531647, "global_step": 36802, "epoch": 413, "lr": 9.563559802141071e-05}
{"train_loss": 0.2765755355358124, "global_step": 36803, "epoch": 413, "lr": 9.563536113589743e-05}
{"train_loss": 0.2471790313720703, "global_step": 36804, "epoch": 413, "lr": 9.563512424424905e-05}
{"train_loss": 0.3311195969581604, "global_step": 36805, "epoch": 413, "lr": 9.563488734646558e-05}
{"train_loss": 0.3274620771408081, "global_step": 36806, "epoch": 413, "lr": 9.563465044254703e-05}
{"train_loss": 0.3378112316131592, "global_step": 36807, "epoch": 413, "lr": 9.563441353249349e-05}
{"train_loss": 0.33762943744659424, "global_step": 36808, "epoch": 413, "lr": 9.563417661630494e-05}
{"train_loss": 0.20074954628944397, "global_step": 36809, "epoch": 413, "lr": 9.563393969398144e-05}
{"train_loss": 0.4442616105079651, "global_step": 36810, "epoch": 413, "lr": 9.5633702765523e-05}
{"train_loss": 0.31528934836387634, "global_step": 36811, "epoch": 413, "lr": 9.563346583092968e-05}
{"train_loss": 0.27602168917655945, "global_step": 36812, "epoch": 413, "lr": 9.56332288902015e-05}
{"train_loss": 0.27645641565322876, "global_step": 36813, "epoch": 413, "lr": 9.563299194333849e-05}
{"train_loss": 0.46082618832588196, "global_step": 36814, "epoch": 413, "lr": 9.563275499034065e-05}
{"train_loss": 0.35220804810523987, "global_step": 36815, "epoch": 413, "lr": 9.563251803120807e-05}
{"train_loss": 0.3321836292743683, "global_step": 36816, "epoch": 413, "lr": 9.563228106594074e-05}
{"train_loss": 0.381301611661911, "global_step": 36817, "epoch": 413, "lr": 9.563204409453873e-05}
{"train_loss": 0.3976937532424927, "global_step": 36818, "epoch": 413, "lr": 9.563180711700201e-05}
{"train_loss": 0.2739501893520355, "global_step": 36819, "epoch": 413, "lr": 9.563157013333069e-05}
{"train_loss": 0.21012364327907562, "global_step": 36820, "epoch": 413, "lr": 9.563133314352472e-05}
{"train_loss": 0.4165838360786438, "global_step": 36821, "epoch": 413, "lr": 9.56310961475842e-05}
{"train_loss": 0.4072771370410919, "global_step": 36822, "epoch": 413, "lr": 9.563085914550913e-05}
{"train_loss": 0.4288415014743805, "global_step": 36823, "epoch": 413, "lr": 9.563062213729957e-05}
{"train_loss": 0.2716779112815857, "global_step": 36824, "epoch": 413, "lr": 9.56303851229555e-05}
{"train_loss": 0.34597206115722656, "global_step": 36825, "epoch": 413, "lr": 9.5630148102477e-05}
{"train_loss": 0.35137099027633667, "global_step": 36826, "epoch": 413, "lr": 9.562991107586406e-05}
{"train_loss": 0.348190039396286, "global_step": 36827, "epoch": 413, "lr": 9.562967404311673e-05}
{"train_loss": 0.24705088138580322, "global_step": 36828, "epoch": 413, "lr": 9.562943700423507e-05}
{"train_loss": 0.2527146637439728, "global_step": 36829, "epoch": 413, "lr": 9.562919995921908e-05}
{"train_loss": 0.3748140335083008, "global_step": 36830, "epoch": 413, "lr": 9.562896290806879e-05}
{"train_loss": 0.36225107312202454, "global_step": 36831, "epoch": 413, "lr": 9.562872585078426e-05}
{"train_loss": 0.31827497482299805, "global_step": 36832, "epoch": 413, "lr": 9.56284887873655e-05}
{"train_loss": 0.2310531586408615, "global_step": 36833, "epoch": 413, "lr": 9.562825171781254e-05}
{"train_loss": 0.2678920328617096, "global_step": 36834, "epoch": 413, "lr": 9.562801464212543e-05}
{"train_loss": 0.4375702738761902, "global_step": 36835, "epoch": 413, "lr": 9.562777756030417e-05}
{"train_loss": 0.3847157657146454, "global_step": 36836, "epoch": 413, "lr": 9.562754047234883e-05}
{"train_loss": 0.35633933544158936, "global_step": 36837, "epoch": 413, "lr": 9.56273033782594e-05}
{"train_loss": 0.3675644099712372, "global_step": 36838, "epoch": 413, "lr": 9.562706627803595e-05}
{"train_loss": 0.28493770956993103, "global_step": 36839, "epoch": 413, "lr": 9.562682917167851e-05}
{"train_loss": 0.2935219705104828, "global_step": 36840, "epoch": 413, "lr": 9.562659205918709e-05}
{"train_loss": 0.39263787865638733, "global_step": 36841, "epoch": 413, "lr": 9.562635494056172e-05}
{"train_loss": 0.2545051574707031, "global_step": 36842, "epoch": 413, "lr": 9.562611781580244e-05}
{"train_loss": 0.4014722406864166, "global_step": 36843, "epoch": 413, "lr": 9.56258806849093e-05}
{"train_loss": 0.4041762053966522, "global_step": 36844, "epoch": 413, "lr": 9.562564354788231e-05}
{"train_loss": 0.3288163680039095, "global_step": 36845, "epoch": 413, "lr": 9.56254064047215e-05, "val_loss": 2.5269834995269775}
{"train_loss": 0.3226781189441681, "global_step": 36846, "epoch": 414, "lr": 9.562516925542693e-05}
{"train_loss": 0.3953220546245575, "global_step": 36847, "epoch": 414, "lr": 9.56249320999986e-05}
{"train_loss": 0.26703184843063354, "global_step": 36848, "epoch": 414, "lr": 9.562469493843655e-05}
{"train_loss": 0.4326487183570862, "global_step": 36849, "epoch": 414, "lr": 9.562445777074083e-05}
{"train_loss": 0.35142621397972107, "global_step": 36850, "epoch": 414, "lr": 9.562422059691144e-05}
{"train_loss": 0.25938481092453003, "global_step": 36851, "epoch": 414, "lr": 9.562398341694844e-05}
{"train_loss": 0.3378828763961792, "global_step": 36852, "epoch": 414, "lr": 9.562374623085185e-05}
{"train_loss": 0.35845547914505005, "global_step": 36853, "epoch": 414, "lr": 9.56235090386217e-05}
{"train_loss": 0.34709277749061584, "global_step": 36854, "epoch": 414, "lr": 9.562327184025803e-05}
{"train_loss": 0.31564953923225403, "global_step": 36855, "epoch": 414, "lr": 9.562303463576087e-05}
{"train_loss": 0.31677350401878357, "global_step": 36856, "epoch": 414, "lr": 9.562279742513024e-05}
{"train_loss": 0.2456873059272766, "global_step": 36857, "epoch": 414, "lr": 9.562256020836619e-05}
{"train_loss": 0.3151790201663971, "global_step": 36858, "epoch": 414, "lr": 9.562232298546873e-05}
{"train_loss": 0.26649072766304016, "global_step": 36859, "epoch": 414, "lr": 9.562208575643792e-05}
{"train_loss": 0.28374335169792175, "global_step": 36860, "epoch": 414, "lr": 9.562184852127376e-05}
{"train_loss": 0.25214827060699463, "global_step": 36861, "epoch": 414, "lr": 9.562161127997631e-05}
{"train_loss": 0.331249475479126, "global_step": 36862, "epoch": 414, "lr": 9.562137403254559e-05}
{"train_loss": 0.28094789385795593, "global_step": 36863, "epoch": 414, "lr": 9.562113677898164e-05}
{"train_loss": 0.30878204107284546, "global_step": 36864, "epoch": 414, "lr": 9.562089951928447e-05}
{"train_loss": 0.3718554675579071, "global_step": 36865, "epoch": 414, "lr": 9.562066225345413e-05}
{"train_loss": 0.3274292051792145, "global_step": 36866, "epoch": 414, "lr": 9.562042498149064e-05}
{"train_loss": 0.3023780286312103, "global_step": 36867, "epoch": 414, "lr": 9.562018770339404e-05}
{"train_loss": 0.2895353138446808, "global_step": 36868, "epoch": 414, "lr": 9.561995041916438e-05}
{"train_loss": 0.3422214388847351, "global_step": 36869, "epoch": 414, "lr": 9.561971312880166e-05}
{"train_loss": 0.2884032428264618, "global_step": 36870, "epoch": 414, "lr": 9.561947583230594e-05}
{"train_loss": 0.27616819739341736, "global_step": 36871, "epoch": 414, "lr": 9.561923852967722e-05}
{"train_loss": 0.2498723566532135, "global_step": 36872, "epoch": 414, "lr": 9.561900122091556e-05}
{"train_loss": 0.27929574251174927, "global_step": 36873, "epoch": 414, "lr": 9.561876390602098e-05}
{"train_loss": 0.30774396657943726, "global_step": 36874, "epoch": 414, "lr": 9.561852658499351e-05}
{"train_loss": 0.2375423163175583, "global_step": 36875, "epoch": 414, "lr": 9.561828925783319e-05}
{"train_loss": 0.35745754837989807, "global_step": 36876, "epoch": 414, "lr": 9.561805192454004e-05}
{"train_loss": 0.22528831660747528, "global_step": 36877, "epoch": 414, "lr": 9.561781458511411e-05}
{"train_loss": 0.3052056133747101, "global_step": 36878, "epoch": 414, "lr": 9.56175772395554e-05}
{"train_loss": 0.36808082461357117, "global_step": 36879, "epoch": 414, "lr": 9.561733988786399e-05}
{"train_loss": 0.2885952293872833, "global_step": 36880, "epoch": 414, "lr": 9.561710253003985e-05}
{"train_loss": 0.2525128126144409, "global_step": 36881, "epoch": 414, "lr": 9.561686516608308e-05}
{"train_loss": 0.3165706396102905, "global_step": 36882, "epoch": 414, "lr": 9.561662779599367e-05}
{"train_loss": 0.34049126505851746, "global_step": 36883, "epoch": 414, "lr": 9.561639041977167e-05}
{"train_loss": 0.2901739776134491, "global_step": 36884, "epoch": 414, "lr": 9.561615303741706e-05}
{"train_loss": 0.26552414894104004, "global_step": 36885, "epoch": 414, "lr": 9.561591564892995e-05}
{"train_loss": 0.29735299944877625, "global_step": 36886, "epoch": 414, "lr": 9.561567825431033e-05}
{"train_loss": 0.3051917254924774, "global_step": 36887, "epoch": 414, "lr": 9.561544085355825e-05}
{"train_loss": 0.33726415038108826, "global_step": 36888, "epoch": 414, "lr": 9.56152034466737e-05}
{"train_loss": 0.25345245003700256, "global_step": 36889, "epoch": 414, "lr": 9.561496603365676e-05}
{"train_loss": 0.36942172050476074, "global_step": 36890, "epoch": 414, "lr": 9.561472861450744e-05}
{"train_loss": 0.3066292405128479, "global_step": 36891, "epoch": 414, "lr": 9.561449118922579e-05}
{"train_loss": 0.35861706733703613, "global_step": 36892, "epoch": 414, "lr": 9.561425375781181e-05}
{"train_loss": 0.3496909439563751, "global_step": 36893, "epoch": 414, "lr": 9.561401632026557e-05}
{"train_loss": 0.37473252415657043, "global_step": 36894, "epoch": 414, "lr": 9.561377887658704e-05}
{"train_loss": 0.2826850414276123, "global_step": 36895, "epoch": 414, "lr": 9.561354142677633e-05}
{"train_loss": 0.34971117973327637, "global_step": 36896, "epoch": 414, "lr": 9.561330397083343e-05}
{"train_loss": 0.2570498287677765, "global_step": 36897, "epoch": 414, "lr": 9.561306650875836e-05}
{"train_loss": 0.27886614203453064, "global_step": 36898, "epoch": 414, "lr": 9.56128290405512e-05}
{"train_loss": 0.40658244490623474, "global_step": 36899, "epoch": 414, "lr": 9.561259156621192e-05}
{"train_loss": 0.281832754611969, "global_step": 36900, "epoch": 414, "lr": 9.56123540857406e-05}
{"train_loss": 0.3369894027709961, "global_step": 36901, "epoch": 414, "lr": 9.561211659913725e-05}
{"train_loss": 0.24606281518936157, "global_step": 36902, "epoch": 414, "lr": 9.561187910640191e-05}
{"train_loss": 0.2877037227153778, "global_step": 36903, "epoch": 414, "lr": 9.56116416075346e-05}
{"train_loss": 0.2761164605617523, "global_step": 36904, "epoch": 414, "lr": 9.561140410253537e-05}
{"train_loss": 0.45085638761520386, "global_step": 36905, "epoch": 414, "lr": 9.561116659140425e-05}
{"train_loss": 0.38160163164138794, "global_step": 36906, "epoch": 414, "lr": 9.561092907414124e-05}
{"train_loss": 0.3797653019428253, "global_step": 36907, "epoch": 414, "lr": 9.561069155074642e-05}
{"train_loss": 0.3064383268356323, "global_step": 36908, "epoch": 414, "lr": 9.561045402121979e-05}
{"train_loss": 0.27900537848472595, "global_step": 36909, "epoch": 414, "lr": 9.561021648556138e-05}
{"train_loss": 0.2916144132614136, "global_step": 36910, "epoch": 414, "lr": 9.560997894377125e-05}
{"train_loss": 0.2857010066509247, "global_step": 36911, "epoch": 414, "lr": 9.56097413958494e-05}
{"train_loss": 0.2983494997024536, "global_step": 36912, "epoch": 414, "lr": 9.560950384179589e-05}
{"train_loss": 0.3490038514137268, "global_step": 36913, "epoch": 414, "lr": 9.560926628161073e-05}
{"train_loss": 0.3055708110332489, "global_step": 36914, "epoch": 414, "lr": 9.560902871529396e-05}
{"train_loss": 0.30694279074668884, "global_step": 36915, "epoch": 414, "lr": 9.560879114284561e-05}
{"train_loss": 0.33850619196891785, "global_step": 36916, "epoch": 414, "lr": 9.560855356426573e-05}
{"train_loss": 0.24336093664169312, "global_step": 36917, "epoch": 414, "lr": 9.560831597955431e-05}
{"train_loss": 0.4442901015281677, "global_step": 36918, "epoch": 414, "lr": 9.560807838871143e-05}
{"train_loss": 0.4320276379585266, "global_step": 36919, "epoch": 414, "lr": 9.560784079173708e-05}
{"train_loss": 0.4029242992401123, "global_step": 36920, "epoch": 414, "lr": 9.560760318863131e-05}
{"train_loss": 0.21975930035114288, "global_step": 36921, "epoch": 414, "lr": 9.560736557939418e-05}
{"train_loss": 0.31653672456741333, "global_step": 36922, "epoch": 414, "lr": 9.560712796402568e-05}
{"train_loss": 0.3492940366268158, "global_step": 36923, "epoch": 414, "lr": 9.560689034252585e-05}
{"train_loss": 0.28051966428756714, "global_step": 36924, "epoch": 414, "lr": 9.560665271489474e-05}
{"train_loss": 0.3930010199546814, "global_step": 36925, "epoch": 414, "lr": 9.560641508113237e-05}
{"train_loss": 0.32712724804878235, "global_step": 36926, "epoch": 414, "lr": 9.560617744123876e-05}
{"train_loss": 0.4110322892665863, "global_step": 36927, "epoch": 414, "lr": 9.560593979521397e-05}
{"train_loss": 0.2256840616464615, "global_step": 36928, "epoch": 414, "lr": 9.560570214305801e-05}
{"train_loss": 0.3710676431655884, "global_step": 36929, "epoch": 414, "lr": 9.560546448477095e-05}
{"train_loss": 0.44361284375190735, "global_step": 36930, "epoch": 414, "lr": 9.560522682035276e-05}
{"train_loss": 0.3894667327404022, "global_step": 36931, "epoch": 414, "lr": 9.56049891498035e-05}
{"train_loss": 0.28915178775787354, "global_step": 36932, "epoch": 414, "lr": 9.560475147312322e-05}
{"train_loss": 0.4844408929347992, "global_step": 36933, "epoch": 414, "lr": 9.560451379031192e-05}
{"train_loss": 0.32077039694518183, "global_step": 36934, "epoch": 414, "lr": 9.560427610136966e-05, "val_loss": 2.5137197971343994}
{"train_loss": 0.4399764835834503, "global_step": 36935, "epoch": 415, "lr": 9.560403840629647e-05}
{"train_loss": 0.22817762196063995, "global_step": 36936, "epoch": 415, "lr": 9.560380070509236e-05}
{"train_loss": 0.27684420347213745, "global_step": 36937, "epoch": 415, "lr": 9.560356299775738e-05}
{"train_loss": 0.2951536476612091, "global_step": 36938, "epoch": 415, "lr": 9.560332528429154e-05}
{"train_loss": 0.3986815810203552, "global_step": 36939, "epoch": 415, "lr": 9.560308756469491e-05}
{"train_loss": 0.4128800630569458, "global_step": 36940, "epoch": 415, "lr": 9.56028498389675e-05}
{"train_loss": 0.3214872479438782, "global_step": 36941, "epoch": 415, "lr": 9.560261210710934e-05}
{"train_loss": 0.38922756910324097, "global_step": 36942, "epoch": 415, "lr": 9.560237436912046e-05}
{"train_loss": 0.2700488567352295, "global_step": 36943, "epoch": 415, "lr": 9.56021366250009e-05}
{"train_loss": 0.33341285586357117, "global_step": 36944, "epoch": 415, "lr": 9.560189887475068e-05}
{"train_loss": 0.36704033613204956, "global_step": 36945, "epoch": 415, "lr": 9.560166111836986e-05}
{"train_loss": 0.40404391288757324, "global_step": 36946, "epoch": 415, "lr": 9.560142335585843e-05}
{"train_loss": 0.20389823615550995, "global_step": 36947, "epoch": 415, "lr": 9.560118558721646e-05}
{"train_loss": 0.2436610460281372, "global_step": 36948, "epoch": 415, "lr": 9.560094781244395e-05}
{"train_loss": 0.30457302927970886, "global_step": 36949, "epoch": 415, "lr": 9.560071003154096e-05}
{"train_loss": 0.39505907893180847, "global_step": 36950, "epoch": 415, "lr": 9.560047224450752e-05}
{"train_loss": 0.3514523208141327, "global_step": 36951, "epoch": 415, "lr": 9.560023445134363e-05}
{"train_loss": 0.3556128740310669, "global_step": 36952, "epoch": 415, "lr": 9.559999665204936e-05}
{"train_loss": 0.4531201422214508, "global_step": 36953, "epoch": 415, "lr": 9.559975884662473e-05}
{"train_loss": 0.2838938236236572, "global_step": 36954, "epoch": 415, "lr": 9.559952103506975e-05}
{"train_loss": 0.4698103964328766, "global_step": 36955, "epoch": 415, "lr": 9.559928321738449e-05}
{"train_loss": 0.2894022464752197, "global_step": 36956, "epoch": 415, "lr": 9.559904539356894e-05}
{"train_loss": 0.2871709167957306, "global_step": 36957, "epoch": 415, "lr": 9.559880756362317e-05}
{"train_loss": 0.39185965061187744, "global_step": 36958, "epoch": 415, "lr": 9.559856972754719e-05}
{"train_loss": 0.3540666103363037, "global_step": 36959, "epoch": 415, "lr": 9.559833188534104e-05}
{"train_loss": 0.3606593608856201, "global_step": 36960, "epoch": 415, "lr": 9.559809403700476e-05}
{"train_loss": 0.48473116755485535, "global_step": 36961, "epoch": 415, "lr": 9.559785618253836e-05}
{"train_loss": 0.5538638234138489, "global_step": 36962, "epoch": 415, "lr": 9.559761832194189e-05}
{"train_loss": 0.3150418996810913, "global_step": 36963, "epoch": 415, "lr": 9.559738045521537e-05}
{"train_loss": 0.35086479783058167, "global_step": 36964, "epoch": 415, "lr": 9.559714258235883e-05}
{"train_loss": 0.28308773040771484, "global_step": 36965, "epoch": 415, "lr": 9.559690470337233e-05}
{"train_loss": 0.2716238498687744, "global_step": 36966, "epoch": 415, "lr": 9.559666681825587e-05}
{"train_loss": 0.4100223183631897, "global_step": 36967, "epoch": 415, "lr": 9.559642892700951e-05}
{"train_loss": 0.4221719205379486, "global_step": 36968, "epoch": 415, "lr": 9.559619102963324e-05}
{"train_loss": 0.4199604392051697, "global_step": 36969, "epoch": 415, "lr": 9.559595312612713e-05}
{"train_loss": 0.4549176096916199, "global_step": 36970, "epoch": 415, "lr": 9.55957152164912e-05}
{"train_loss": 0.31638067960739136, "global_step": 36971, "epoch": 415, "lr": 9.55954773007255e-05}
{"train_loss": 0.3607473373413086, "global_step": 36972, "epoch": 415, "lr": 9.559523937883001e-05}
{"train_loss": 0.2698058784008026, "global_step": 36973, "epoch": 415, "lr": 9.559500145080482e-05}
{"train_loss": 0.25675877928733826, "global_step": 36974, "epoch": 415, "lr": 9.559476351664992e-05}
{"train_loss": 0.3048056960105896, "global_step": 36975, "epoch": 415, "lr": 9.559452557636537e-05}
{"train_loss": 0.270537406206131, "global_step": 36976, "epoch": 415, "lr": 9.55942876299512e-05}
{"train_loss": 0.3095840513706207, "global_step": 36977, "epoch": 415, "lr": 9.559404967740742e-05}
{"train_loss": 0.35991305112838745, "global_step": 36978, "epoch": 415, "lr": 9.559381171873408e-05}
{"train_loss": 0.3235972225666046, "global_step": 36979, "epoch": 415, "lr": 9.559357375393121e-05}
{"train_loss": 0.35826945304870605, "global_step": 36980, "epoch": 415, "lr": 9.559333578299882e-05}
{"train_loss": 0.3959396481513977, "global_step": 36981, "epoch": 415, "lr": 9.559309780593699e-05}
{"train_loss": 0.2952737510204315, "global_step": 36982, "epoch": 415, "lr": 9.559285982274572e-05}
{"train_loss": 0.39931148290634155, "global_step": 36983, "epoch": 415, "lr": 9.559262183342503e-05}
{"train_loss": 0.4556526243686676, "global_step": 36984, "epoch": 415, "lr": 9.559238383797497e-05}
{"train_loss": 0.294545978307724, "global_step": 36985, "epoch": 415, "lr": 9.559214583639559e-05}
{"train_loss": 0.3700942099094391, "global_step": 36986, "epoch": 415, "lr": 9.559190782868689e-05}
{"train_loss": 0.4590863883495331, "global_step": 36987, "epoch": 415, "lr": 9.55916698148489e-05}
{"train_loss": 0.29743126034736633, "global_step": 36988, "epoch": 415, "lr": 9.559143179488168e-05}
{"train_loss": 0.2938452959060669, "global_step": 36989, "epoch": 415, "lr": 9.559119376878525e-05}
{"train_loss": 0.2401634156703949, "global_step": 36990, "epoch": 415, "lr": 9.559095573655963e-05}
{"train_loss": 0.293263703584671, "global_step": 36991, "epoch": 415, "lr": 9.559071769820487e-05}
{"train_loss": 0.389563649892807, "global_step": 36992, "epoch": 415, "lr": 9.559047965372098e-05}
{"train_loss": 0.38698846101760864, "global_step": 36993, "epoch": 415, "lr": 9.559024160310803e-05}
{"train_loss": 0.30424779653549194, "global_step": 36994, "epoch": 415, "lr": 9.5590003546366e-05}
{"train_loss": 0.3359503448009491, "global_step": 36995, "epoch": 415, "lr": 9.558976548349497e-05}
{"train_loss": 0.32132863998413086, "global_step": 36996, "epoch": 415, "lr": 9.558952741449495e-05}
{"train_loss": 0.35532572865486145, "global_step": 36997, "epoch": 415, "lr": 9.558928933936598e-05}
{"train_loss": 0.24705146253108978, "global_step": 36998, "epoch": 415, "lr": 9.558905125810806e-05}
{"train_loss": 0.3581346273422241, "global_step": 36999, "epoch": 415, "lr": 9.558881317072126e-05}
{"train_loss": 0.24413152039051056, "global_step": 37000, "epoch": 415, "lr": 9.55885750772056e-05}
{"train_loss": 0.30700650811195374, "global_step": 37001, "epoch": 415, "lr": 9.558833697756111e-05}
{"train_loss": 0.3005738854408264, "global_step": 37002, "epoch": 415, "lr": 9.558809887178784e-05}
{"train_loss": 0.264116108417511, "global_step": 37003, "epoch": 415, "lr": 9.558786075988579e-05}
{"train_loss": 0.29520857334136963, "global_step": 37004, "epoch": 415, "lr": 9.558762264185502e-05}
{"train_loss": 0.29090964794158936, "global_step": 37005, "epoch": 415, "lr": 9.558738451769553e-05}
{"train_loss": 0.2794899642467499, "global_step": 37006, "epoch": 415, "lr": 9.558714638740738e-05}
{"train_loss": 0.3766394257545471, "global_step": 37007, "epoch": 415, "lr": 9.558690825099059e-05}
{"train_loss": 0.3117137849330902, "global_step": 37008, "epoch": 415, "lr": 9.55866701084452e-05}
{"train_loss": 0.28304609656333923, "global_step": 37009, "epoch": 415, "lr": 9.558643195977125e-05}
{"train_loss": 0.3726327121257782, "global_step": 37010, "epoch": 415, "lr": 9.558619380496874e-05}
{"train_loss": 0.39145734906196594, "global_step": 37011, "epoch": 415, "lr": 9.558595564403774e-05}
{"train_loss": 0.22741910815238953, "global_step": 37012, "epoch": 415, "lr": 9.558571747697825e-05}
{"train_loss": 0.31365966796875, "global_step": 37013, "epoch": 415, "lr": 9.558547930379033e-05}
{"train_loss": 0.3949645161628723, "global_step": 37014, "epoch": 415, "lr": 9.558524112447398e-05}
{"train_loss": 0.27586275339126587, "global_step": 37015, "epoch": 415, "lr": 9.558500293902926e-05}
{"train_loss": 0.3146427273750305, "global_step": 37016, "epoch": 415, "lr": 9.55847647474562e-05}
{"train_loss": 0.3099639415740967, "global_step": 37017, "epoch": 415, "lr": 9.55845265497548e-05}
{"train_loss": 0.35142838954925537, "global_step": 37018, "epoch": 415, "lr": 9.558428834592514e-05}
{"train_loss": 0.37190189957618713, "global_step": 37019, "epoch": 415, "lr": 9.558405013596722e-05}
{"train_loss": 0.22681528329849243, "global_step": 37020, "epoch": 415, "lr": 9.558381191988108e-05}
{"train_loss": 0.18224219977855682, "global_step": 37021, "epoch": 415, "lr": 9.558357369766675e-05}
{"train_loss": 0.2599819302558899, "global_step": 37022, "epoch": 415, "lr": 9.558333546932426e-05}
{"train_loss": 0.33472768928897517, "global_step": 37023, "epoch": 415, "lr": 9.558309723485366e-05, "val_loss": 2.670856475830078, "train_action_mse_error": 35.69165802001953}
{"train_loss": 0.3024555444717407, "global_step": 37024, "epoch": 416, "lr": 9.558285899425495e-05}
{"train_loss": 0.375186026096344, "global_step": 37025, "epoch": 416, "lr": 9.558262074752819e-05}
{"train_loss": 0.31730279326438904, "global_step": 37026, "epoch": 416, "lr": 9.55823824946734e-05}
{"train_loss": 0.30912843346595764, "global_step": 37027, "epoch": 416, "lr": 9.558214423569062e-05}
{"train_loss": 0.2612711191177368, "global_step": 37028, "epoch": 416, "lr": 9.558190597057988e-05}
{"train_loss": 0.3125598728656769, "global_step": 37029, "epoch": 416, "lr": 9.558166769934119e-05}
{"train_loss": 0.3135141134262085, "global_step": 37030, "epoch": 416, "lr": 9.558142942197461e-05}
{"train_loss": 0.24660392105579376, "global_step": 37031, "epoch": 416, "lr": 9.558119113848016e-05}
{"train_loss": 0.26729607582092285, "global_step": 37032, "epoch": 416, "lr": 9.558095284885788e-05}
{"train_loss": 0.2776614725589752, "global_step": 37033, "epoch": 416, "lr": 9.55807145531078e-05}
{"train_loss": 0.35156404972076416, "global_step": 37034, "epoch": 416, "lr": 9.558047625122994e-05}
{"train_loss": 0.25543075799942017, "global_step": 37035, "epoch": 416, "lr": 9.558023794322434e-05}
{"train_loss": 0.4353170096874237, "global_step": 37036, "epoch": 416, "lr": 9.557999962909104e-05}
{"train_loss": 0.3042440116405487, "global_step": 37037, "epoch": 416, "lr": 9.557976130883007e-05}
{"train_loss": 0.48221859335899353, "global_step": 37038, "epoch": 416, "lr": 9.557952298244144e-05}
{"train_loss": 0.3966695964336395, "global_step": 37039, "epoch": 416, "lr": 9.55792846499252e-05}
{"train_loss": 0.31377342343330383, "global_step": 37040, "epoch": 416, "lr": 9.557904631128139e-05}
{"train_loss": 0.290267676115036, "global_step": 37041, "epoch": 416, "lr": 9.557880796651003e-05}
{"train_loss": 0.32673829793930054, "global_step": 37042, "epoch": 416, "lr": 9.557856961561116e-05}
{"train_loss": 0.3009149432182312, "global_step": 37043, "epoch": 416, "lr": 9.55783312585848e-05}
{"train_loss": 0.30788928270339966, "global_step": 37044, "epoch": 416, "lr": 9.557809289543098e-05}
{"train_loss": 0.3499908745288849, "global_step": 37045, "epoch": 416, "lr": 9.557785452614976e-05}
{"train_loss": 0.49322426319122314, "global_step": 37046, "epoch": 416, "lr": 9.557761615074114e-05}
{"train_loss": 0.2838068902492523, "global_step": 37047, "epoch": 416, "lr": 9.557737776920517e-05}
{"train_loss": 0.34831565618515015, "global_step": 37048, "epoch": 416, "lr": 9.557713938154187e-05}
{"train_loss": 0.23454946279525757, "global_step": 37049, "epoch": 416, "lr": 9.557690098775128e-05}
{"train_loss": 0.22939948737621307, "global_step": 37050, "epoch": 416, "lr": 9.557666258783344e-05}
{"train_loss": 0.2773994505405426, "global_step": 37051, "epoch": 416, "lr": 9.557642418178838e-05}
{"train_loss": 0.3142618238925934, "global_step": 37052, "epoch": 416, "lr": 9.557618576961612e-05}
{"train_loss": 0.3190270662307739, "global_step": 37053, "epoch": 416, "lr": 9.557594735131668e-05}
{"train_loss": 0.38614028692245483, "global_step": 37054, "epoch": 416, "lr": 9.557570892689011e-05}
{"train_loss": 0.4135759472846985, "global_step": 37055, "epoch": 416, "lr": 9.557547049633646e-05}
{"train_loss": 0.29464372992515564, "global_step": 37056, "epoch": 416, "lr": 9.557523205965573e-05}
{"train_loss": 0.31783097982406616, "global_step": 37057, "epoch": 416, "lr": 9.557499361684798e-05}
{"train_loss": 0.38519638776779175, "global_step": 37058, "epoch": 416, "lr": 9.55747551679132e-05}
{"train_loss": 0.3720667362213135, "global_step": 37059, "epoch": 416, "lr": 9.557451671285149e-05}
{"train_loss": 0.37635475397109985, "global_step": 37060, "epoch": 416, "lr": 9.55742782516628e-05}
{"train_loss": 0.2552347481250763, "global_step": 37061, "epoch": 416, "lr": 9.557403978434725e-05}
{"train_loss": 0.2713657319545746, "global_step": 37062, "epoch": 416, "lr": 9.557380131090478e-05}
{"train_loss": 0.35200250148773193, "global_step": 37063, "epoch": 416, "lr": 9.55735628313355e-05}
{"train_loss": 0.23058032989501953, "global_step": 37064, "epoch": 416, "lr": 9.557332434563939e-05}
{"train_loss": 0.30824577808380127, "global_step": 37065, "epoch": 416, "lr": 9.557308585381652e-05}
{"train_loss": 0.424052357673645, "global_step": 37066, "epoch": 416, "lr": 9.55728473558669e-05}
{"train_loss": 0.3767523467540741, "global_step": 37067, "epoch": 416, "lr": 9.557260885179056e-05}
{"train_loss": 0.2988663911819458, "global_step": 37068, "epoch": 416, "lr": 9.557237034158753e-05}
{"train_loss": 0.41066986322402954, "global_step": 37069, "epoch": 416, "lr": 9.557213182525786e-05}
{"train_loss": 0.3424065411090851, "global_step": 37070, "epoch": 416, "lr": 9.557189330280156e-05}
{"train_loss": 0.381367027759552, "global_step": 37071, "epoch": 416, "lr": 9.55716547742187e-05}
{"train_loss": 0.2798294425010681, "global_step": 37072, "epoch": 416, "lr": 9.557141623950927e-05}
{"train_loss": 0.31636443734169006, "global_step": 37073, "epoch": 416, "lr": 9.557117769867332e-05}
{"train_loss": 0.34317833185195923, "global_step": 37074, "epoch": 416, "lr": 9.557093915171088e-05}
{"train_loss": 0.3885628581047058, "global_step": 37075, "epoch": 416, "lr": 9.557070059862197e-05}
{"train_loss": 0.1755165308713913, "global_step": 37076, "epoch": 416, "lr": 9.557046203940665e-05}
{"train_loss": 0.3043786585330963, "global_step": 37077, "epoch": 416, "lr": 9.557022347406495e-05}
{"train_loss": 0.3543734848499298, "global_step": 37078, "epoch": 416, "lr": 9.556998490259686e-05}
{"train_loss": 0.3141561448574066, "global_step": 37079, "epoch": 416, "lr": 9.556974632500247e-05}
{"train_loss": 0.2784971594810486, "global_step": 37080, "epoch": 416, "lr": 9.556950774128176e-05}
{"train_loss": 0.2636883854866028, "global_step": 37081, "epoch": 416, "lr": 9.556926915143479e-05}
{"train_loss": 0.4305868148803711, "global_step": 37082, "epoch": 416, "lr": 9.55690305554616e-05}
{"train_loss": 0.3039298355579376, "global_step": 37083, "epoch": 416, "lr": 9.556879195336219e-05}
{"train_loss": 0.253294974565506, "global_step": 37084, "epoch": 416, "lr": 9.556855334513662e-05}
{"train_loss": 0.37241482734680176, "global_step": 37085, "epoch": 416, "lr": 9.556831473078491e-05}
{"train_loss": 0.2326267659664154, "global_step": 37086, "epoch": 416, "lr": 9.556807611030711e-05}
{"train_loss": 0.34593889117240906, "global_step": 37087, "epoch": 416, "lr": 9.556783748370321e-05}
{"train_loss": 0.39395207166671753, "global_step": 37088, "epoch": 416, "lr": 9.55675988509733e-05}
{"train_loss": 0.2699296176433563, "global_step": 37089, "epoch": 416, "lr": 9.556736021211736e-05}
{"train_loss": 0.3490358889102936, "global_step": 37090, "epoch": 416, "lr": 9.556712156713546e-05}
{"train_loss": 0.283358097076416, "global_step": 37091, "epoch": 416, "lr": 9.55668829160276e-05}
{"train_loss": 0.3352771997451782, "global_step": 37092, "epoch": 416, "lr": 9.556664425879384e-05}
{"train_loss": 0.2867012023925781, "global_step": 37093, "epoch": 416, "lr": 9.556640559543419e-05}
{"train_loss": 0.4359370768070221, "global_step": 37094, "epoch": 416, "lr": 9.55661669259487e-05}
{"train_loss": 0.4684160649776459, "global_step": 37095, "epoch": 416, "lr": 9.55659282503374e-05}
{"train_loss": 0.4085735082626343, "global_step": 37096, "epoch": 416, "lr": 9.55656895686003e-05}
{"train_loss": 0.2656156122684479, "global_step": 37097, "epoch": 416, "lr": 9.556545088073747e-05}
{"train_loss": 0.23830562829971313, "global_step": 37098, "epoch": 416, "lr": 9.55652121867489e-05}
{"train_loss": 0.33761245012283325, "global_step": 37099, "epoch": 416, "lr": 9.556497348663466e-05}
{"train_loss": 0.34023517370224, "global_step": 37100, "epoch": 416, "lr": 9.556473478039475e-05}
{"train_loss": 0.35664018988609314, "global_step": 37101, "epoch": 416, "lr": 9.556449606802922e-05}
{"train_loss": 0.339224636554718, "global_step": 37102, "epoch": 416, "lr": 9.55642573495381e-05}
{"train_loss": 0.3138129413127899, "global_step": 37103, "epoch": 416, "lr": 9.556401862492142e-05}
{"train_loss": 0.2794489562511444, "global_step": 37104, "epoch": 416, "lr": 9.556377989417922e-05}
{"train_loss": 0.30160602927207947, "global_step": 37105, "epoch": 416, "lr": 9.556354115731151e-05}
{"train_loss": 0.3758501410484314, "global_step": 37106, "epoch": 416, "lr": 9.556330241431835e-05}
{"train_loss": 0.3103352189064026, "global_step": 37107, "epoch": 416, "lr": 9.556306366519976e-05}
{"train_loss": 0.2836797833442688, "global_step": 37108, "epoch": 416, "lr": 9.556282490995576e-05}
{"train_loss": 0.37053415179252625, "global_step": 37109, "epoch": 416, "lr": 9.55625861485864e-05}
{"train_loss": 0.3144471347332001, "global_step": 37110, "epoch": 416, "lr": 9.556234738109171e-05}
{"train_loss": 0.41535523533821106, "global_step": 37111, "epoch": 416, "lr": 9.556210860747172e-05}
{"train_loss": 0.32690234582745625, "global_step": 37112, "epoch": 416, "lr": 9.556186982772646e-05, "val_loss": 2.532745361328125}
{"train_loss": 0.33659854531288147, "global_step": 37113, "epoch": 417, "lr": 9.556163104185597e-05}
{"train_loss": 0.27974072098731995, "global_step": 37114, "epoch": 417, "lr": 9.556139224986025e-05}
{"train_loss": 0.32166996598243713, "global_step": 37115, "epoch": 417, "lr": 9.556115345173935e-05}
{"train_loss": 0.22967535257339478, "global_step": 37116, "epoch": 417, "lr": 9.556091464749334e-05}
{"train_loss": 0.326858788728714, "global_step": 37117, "epoch": 417, "lr": 9.55606758371222e-05}
{"train_loss": 0.5558854937553406, "global_step": 37118, "epoch": 417, "lr": 9.556043702062599e-05}
{"train_loss": 0.42533907294273376, "global_step": 37119, "epoch": 417, "lr": 9.556019819800472e-05}
{"train_loss": 0.32299062609672546, "global_step": 37120, "epoch": 417, "lr": 9.555995936925846e-05}
{"train_loss": 0.3172247111797333, "global_step": 37121, "epoch": 417, "lr": 9.55597205343872e-05}
{"train_loss": 0.4242112934589386, "global_step": 37122, "epoch": 417, "lr": 9.555948169339101e-05}
{"train_loss": 0.3470304310321808, "global_step": 37123, "epoch": 417, "lr": 9.555924284626989e-05}
{"train_loss": 0.3750517964363098, "global_step": 37124, "epoch": 417, "lr": 9.555900399302389e-05}
{"train_loss": 0.4222657084465027, "global_step": 37125, "epoch": 417, "lr": 9.555876513365303e-05}
{"train_loss": 0.3408515751361847, "global_step": 37126, "epoch": 417, "lr": 9.555852626815736e-05}
{"train_loss": 0.3072494566440582, "global_step": 37127, "epoch": 417, "lr": 9.555828739653689e-05}
{"train_loss": 0.2740270495414734, "global_step": 37128, "epoch": 417, "lr": 9.555804851879167e-05}
{"train_loss": 0.28393182158470154, "global_step": 37129, "epoch": 417, "lr": 9.555780963492171e-05}
{"train_loss": 0.3674159646034241, "global_step": 37130, "epoch": 417, "lr": 9.555757074492709e-05}
{"train_loss": 0.39253658056259155, "global_step": 37131, "epoch": 417, "lr": 9.555733184880777e-05}
{"train_loss": 0.2609255909919739, "global_step": 37132, "epoch": 417, "lr": 9.555709294656385e-05}
{"train_loss": 0.36026352643966675, "global_step": 37133, "epoch": 417, "lr": 9.555685403819533e-05}
{"train_loss": 0.2865704596042633, "global_step": 37134, "epoch": 417, "lr": 9.555661512370224e-05}
{"train_loss": 0.42694494128227234, "global_step": 37135, "epoch": 417, "lr": 9.555637620308461e-05}
{"train_loss": 0.2466835379600525, "global_step": 37136, "epoch": 417, "lr": 9.55561372763425e-05}
{"train_loss": 0.3710172176361084, "global_step": 37137, "epoch": 417, "lr": 9.555589834347592e-05}
{"train_loss": 0.2922012507915497, "global_step": 37138, "epoch": 417, "lr": 9.55556594044849e-05}
{"train_loss": 0.3945631980895996, "global_step": 37139, "epoch": 417, "lr": 9.555542045936947e-05}
{"train_loss": 0.3835090100765228, "global_step": 37140, "epoch": 417, "lr": 9.555518150812968e-05}
{"train_loss": 0.26874399185180664, "global_step": 37141, "epoch": 417, "lr": 9.555494255076556e-05}
{"train_loss": 0.35604771971702576, "global_step": 37142, "epoch": 417, "lr": 9.555470358727712e-05}
{"train_loss": 0.29435715079307556, "global_step": 37143, "epoch": 417, "lr": 9.55544646176644e-05}
{"train_loss": 0.2556408643722534, "global_step": 37144, "epoch": 417, "lr": 9.555422564192744e-05}
{"train_loss": 0.331156462430954, "global_step": 37145, "epoch": 417, "lr": 9.555398666006627e-05}
{"train_loss": 0.44196000695228577, "global_step": 37146, "epoch": 417, "lr": 9.555374767208092e-05}
{"train_loss": 0.332032710313797, "global_step": 37147, "epoch": 417, "lr": 9.555350867797143e-05}
{"train_loss": 0.25551992654800415, "global_step": 37148, "epoch": 417, "lr": 9.555326967773783e-05}
{"train_loss": 0.26611366868019104, "global_step": 37149, "epoch": 417, "lr": 9.555303067138013e-05}
{"train_loss": 0.32861727476119995, "global_step": 37150, "epoch": 417, "lr": 9.55527916588984e-05}
{"train_loss": 0.33878210186958313, "global_step": 37151, "epoch": 417, "lr": 9.555255264029264e-05}
{"train_loss": 0.29225900769233704, "global_step": 37152, "epoch": 417, "lr": 9.55523136155629e-05}
{"train_loss": 0.29709362983703613, "global_step": 37153, "epoch": 417, "lr": 9.55520745847092e-05}
{"train_loss": 0.332618772983551, "global_step": 37154, "epoch": 417, "lr": 9.55518355477316e-05}
{"train_loss": 0.34918931126594543, "global_step": 37155, "epoch": 417, "lr": 9.555159650463008e-05}
{"train_loss": 0.30843818187713623, "global_step": 37156, "epoch": 417, "lr": 9.555135745540473e-05}
{"train_loss": 0.2969991862773895, "global_step": 37157, "epoch": 417, "lr": 9.555111840005554e-05}
{"train_loss": 0.40523311495780945, "global_step": 37158, "epoch": 417, "lr": 9.555087933858256e-05}
{"train_loss": 0.2890348732471466, "global_step": 37159, "epoch": 417, "lr": 9.555064027098581e-05}
{"train_loss": 0.3389216959476471, "global_step": 37160, "epoch": 417, "lr": 9.555040119726534e-05}
{"train_loss": 0.418699711561203, "global_step": 37161, "epoch": 417, "lr": 9.555016211742119e-05}
{"train_loss": 0.3076849579811096, "global_step": 37162, "epoch": 417, "lr": 9.554992303145336e-05}
{"train_loss": 0.32678094506263733, "global_step": 37163, "epoch": 417, "lr": 9.55496839393619e-05}
{"train_loss": 0.301532119512558, "global_step": 37164, "epoch": 417, "lr": 9.554944484114684e-05}
{"train_loss": 0.31320303678512573, "global_step": 37165, "epoch": 417, "lr": 9.554920573680821e-05}
{"train_loss": 0.196612149477005, "global_step": 37166, "epoch": 417, "lr": 9.554896662634605e-05}
{"train_loss": 0.26849040389060974, "global_step": 37167, "epoch": 417, "lr": 9.554872750976039e-05}
{"train_loss": 0.34894657135009766, "global_step": 37168, "epoch": 417, "lr": 9.554848838705123e-05}
{"train_loss": 0.31681355834007263, "global_step": 37169, "epoch": 417, "lr": 9.554824925821867e-05}
{"train_loss": 0.3217991292476654, "global_step": 37170, "epoch": 417, "lr": 9.554801012326269e-05}
{"train_loss": 0.33601614832878113, "global_step": 37171, "epoch": 417, "lr": 9.554777098218333e-05}
{"train_loss": 0.29124715924263, "global_step": 37172, "epoch": 417, "lr": 9.554753183498062e-05}
{"train_loss": 0.40217065811157227, "global_step": 37173, "epoch": 417, "lr": 9.554729268165461e-05}
{"train_loss": 0.35990482568740845, "global_step": 37174, "epoch": 417, "lr": 9.554705352220532e-05}
{"train_loss": 0.29459333419799805, "global_step": 37175, "epoch": 417, "lr": 9.554681435663278e-05}
{"train_loss": 0.20382066071033478, "global_step": 37176, "epoch": 417, "lr": 9.554657518493704e-05}
{"train_loss": 0.3248385488986969, "global_step": 37177, "epoch": 417, "lr": 9.55463360071181e-05}
{"train_loss": 0.3246813714504242, "global_step": 37178, "epoch": 417, "lr": 9.5546096823176e-05}
{"train_loss": 0.31254008412361145, "global_step": 37179, "epoch": 417, "lr": 9.554585763311082e-05}
{"train_loss": 0.2566147744655609, "global_step": 37180, "epoch": 417, "lr": 9.554561843692253e-05}
{"train_loss": 0.18974241614341736, "global_step": 37181, "epoch": 417, "lr": 9.55453792346112e-05}
{"train_loss": 0.3751913011074066, "global_step": 37182, "epoch": 417, "lr": 9.554514002617683e-05}
{"train_loss": 0.3553248345851898, "global_step": 37183, "epoch": 417, "lr": 9.554490081161949e-05}
{"train_loss": 0.28974029421806335, "global_step": 37184, "epoch": 417, "lr": 9.554466159093918e-05}
{"train_loss": 0.2867755889892578, "global_step": 37185, "epoch": 417, "lr": 9.554442236413594e-05}
{"train_loss": 0.27456989884376526, "global_step": 37186, "epoch": 417, "lr": 9.554418313120983e-05}
{"train_loss": 0.2927103638648987, "global_step": 37187, "epoch": 417, "lr": 9.554394389216083e-05}
{"train_loss": 0.30898338556289673, "global_step": 37188, "epoch": 417, "lr": 9.554370464698904e-05}
{"train_loss": 0.31357541680336, "global_step": 37189, "epoch": 417, "lr": 9.554346539569442e-05}
{"train_loss": 0.2810974717140198, "global_step": 37190, "epoch": 417, "lr": 9.554322613827705e-05}
{"train_loss": 0.3004553020000458, "global_step": 37191, "epoch": 417, "lr": 9.554298687473695e-05}
{"train_loss": 0.23064136505126953, "global_step": 37192, "epoch": 417, "lr": 9.554274760507415e-05}
{"train_loss": 0.31628233194351196, "global_step": 37193, "epoch": 417, "lr": 9.554250832928868e-05}
{"train_loss": 0.32881179451942444, "global_step": 37194, "epoch": 417, "lr": 9.554226904738057e-05}
{"train_loss": 0.26112598180770874, "global_step": 37195, "epoch": 417, "lr": 9.554202975934987e-05}
{"train_loss": 0.24425415694713593, "global_step": 37196, "epoch": 417, "lr": 9.554179046519659e-05}
{"train_loss": 0.42749330401420593, "global_step": 37197, "epoch": 417, "lr": 9.554155116492078e-05}
{"train_loss": 0.29798582196235657, "global_step": 37198, "epoch": 417, "lr": 9.554131185852244e-05}
{"train_loss": 0.44968685507774353, "global_step": 37199, "epoch": 417, "lr": 9.554107254600165e-05}
{"train_loss": 0.4419480860233307, "global_step": 37200, "epoch": 417, "lr": 9.55408332273584e-05}
{"train_loss": 0.32410512379046236, "global_step": 37201, "epoch": 417, "lr": 9.554059390259275e-05, "val_loss": 2.602792501449585}
{"train_loss": 0.2773174047470093, "global_step": 37202, "epoch": 418, "lr": 9.554035457170474e-05}
{"train_loss": 0.2885485291481018, "global_step": 37203, "epoch": 418, "lr": 9.554011523469436e-05}
{"train_loss": 0.37335965037345886, "global_step": 37204, "epoch": 418, "lr": 9.553987589156167e-05}
{"train_loss": 0.384621798992157, "global_step": 37205, "epoch": 418, "lr": 9.55396365423067e-05}
{"train_loss": 0.3483334183692932, "global_step": 37206, "epoch": 418, "lr": 9.553939718692949e-05}
{"train_loss": 0.3112218976020813, "global_step": 37207, "epoch": 418, "lr": 9.553915782543006e-05}
{"train_loss": 0.41228118538856506, "global_step": 37208, "epoch": 418, "lr": 9.553891845780843e-05}
{"train_loss": 0.3399231433868408, "global_step": 37209, "epoch": 418, "lr": 9.553867908406466e-05}
{"train_loss": 0.2951536178588867, "global_step": 37210, "epoch": 418, "lr": 9.553843970419877e-05}
{"train_loss": 0.336630642414093, "global_step": 37211, "epoch": 418, "lr": 9.55382003182108e-05}
{"train_loss": 0.36257249116897583, "global_step": 37212, "epoch": 418, "lr": 9.553796092610075e-05}
{"train_loss": 0.343784362077713, "global_step": 37213, "epoch": 418, "lr": 9.553772152786869e-05}
{"train_loss": 0.33424481749534607, "global_step": 37214, "epoch": 418, "lr": 9.553748212351464e-05}
{"train_loss": 0.4464946985244751, "global_step": 37215, "epoch": 418, "lr": 9.553724271303864e-05}
{"train_loss": 0.41211438179016113, "global_step": 37216, "epoch": 418, "lr": 9.553700329644069e-05}
{"train_loss": 0.4083849787712097, "global_step": 37217, "epoch": 418, "lr": 9.553676387372086e-05}
{"train_loss": 0.36233067512512207, "global_step": 37218, "epoch": 418, "lr": 9.553652444487917e-05}
{"train_loss": 0.30434003472328186, "global_step": 37219, "epoch": 418, "lr": 9.553628500991564e-05}
{"train_loss": 0.34674766659736633, "global_step": 37220, "epoch": 418, "lr": 9.553604556883031e-05}
{"train_loss": 0.3297504782676697, "global_step": 37221, "epoch": 418, "lr": 9.553580612162323e-05}
{"train_loss": 0.31762072443962097, "global_step": 37222, "epoch": 418, "lr": 9.55355666682944e-05}
{"train_loss": 0.2857223153114319, "global_step": 37223, "epoch": 418, "lr": 9.553532720884385e-05}
{"train_loss": 0.2555299997329712, "global_step": 37224, "epoch": 418, "lr": 9.553508774327167e-05}
{"train_loss": 0.4362128973007202, "global_step": 37225, "epoch": 418, "lr": 9.553484827157784e-05}
{"train_loss": 0.32265597581863403, "global_step": 37226, "epoch": 418, "lr": 9.553460879376239e-05}
{"train_loss": 0.5166348814964294, "global_step": 37227, "epoch": 418, "lr": 9.553436930982538e-05}
{"train_loss": 0.314328134059906, "global_step": 37228, "epoch": 418, "lr": 9.553412981976683e-05}
{"train_loss": 0.293567955493927, "global_step": 37229, "epoch": 418, "lr": 9.553389032358677e-05}
{"train_loss": 0.2907356023788452, "global_step": 37230, "epoch": 418, "lr": 9.553365082128522e-05}
{"train_loss": 0.3272400200366974, "global_step": 37231, "epoch": 418, "lr": 9.553341131286224e-05}
{"train_loss": 0.325166791677475, "global_step": 37232, "epoch": 418, "lr": 9.553317179831785e-05}
{"train_loss": 0.3963054120540619, "global_step": 37233, "epoch": 418, "lr": 9.553293227765208e-05}
{"train_loss": 0.3112075626850128, "global_step": 37234, "epoch": 418, "lr": 9.553269275086494e-05}
{"train_loss": 0.28223246335983276, "global_step": 37235, "epoch": 418, "lr": 9.55324532179565e-05}
{"train_loss": 0.27162712812423706, "global_step": 37236, "epoch": 418, "lr": 9.553221367892677e-05}
{"train_loss": 0.34179943799972534, "global_step": 37237, "epoch": 418, "lr": 9.553197413377582e-05}
{"train_loss": 0.3364162743091583, "global_step": 37238, "epoch": 418, "lr": 9.553173458250362e-05}
{"train_loss": 0.3240569829940796, "global_step": 37239, "epoch": 418, "lr": 9.553149502511024e-05}
{"train_loss": 0.28172552585601807, "global_step": 37240, "epoch": 418, "lr": 9.553125546159569e-05}
{"train_loss": 0.2965802848339081, "global_step": 37241, "epoch": 418, "lr": 9.553101589196003e-05}
{"train_loss": 0.298816978931427, "global_step": 37242, "epoch": 418, "lr": 9.55307763162033e-05}
{"train_loss": 0.3061791658401489, "global_step": 37243, "epoch": 418, "lr": 9.55305367343255e-05}
{"train_loss": 0.2665185034275055, "global_step": 37244, "epoch": 418, "lr": 9.553029714632667e-05}
{"train_loss": 0.31888195872306824, "global_step": 37245, "epoch": 418, "lr": 9.553005755220683e-05}
{"train_loss": 0.2757991850376129, "global_step": 37246, "epoch": 418, "lr": 9.552981795196605e-05}
{"train_loss": 0.3091396391391754, "global_step": 37247, "epoch": 418, "lr": 9.552957834560434e-05}
{"train_loss": 0.3135417103767395, "global_step": 37248, "epoch": 418, "lr": 9.552933873312174e-05}
{"train_loss": 0.33981549739837646, "global_step": 37249, "epoch": 418, "lr": 9.552909911451826e-05}
{"train_loss": 0.24697433412075043, "global_step": 37250, "epoch": 418, "lr": 9.552885948979394e-05}
{"train_loss": 0.32301127910614014, "global_step": 37251, "epoch": 418, "lr": 9.552861985894883e-05}
{"train_loss": 0.3509543240070343, "global_step": 37252, "epoch": 418, "lr": 9.552838022198296e-05}
{"train_loss": 0.19130611419677734, "global_step": 37253, "epoch": 418, "lr": 9.552814057889634e-05}
{"train_loss": 0.36433351039886475, "global_step": 37254, "epoch": 418, "lr": 9.552790092968901e-05}
{"train_loss": 0.27495118975639343, "global_step": 37255, "epoch": 418, "lr": 9.552766127436103e-05}
{"train_loss": 0.3714682459831238, "global_step": 37256, "epoch": 418, "lr": 9.552742161291241e-05}
{"train_loss": 0.2912255525588989, "global_step": 37257, "epoch": 418, "lr": 9.552718194534319e-05}
{"train_loss": 0.2916039228439331, "global_step": 37258, "epoch": 418, "lr": 9.552694227165337e-05}
{"train_loss": 0.3456975221633911, "global_step": 37259, "epoch": 418, "lr": 9.552670259184301e-05}
{"train_loss": 0.3787015974521637, "global_step": 37260, "epoch": 418, "lr": 9.552646290591216e-05}
{"train_loss": 0.3238212466239929, "global_step": 37261, "epoch": 418, "lr": 9.552622321386083e-05}
{"train_loss": 0.1570000946521759, "global_step": 37262, "epoch": 418, "lr": 9.552598351568903e-05}
{"train_loss": 0.322647362947464, "global_step": 37263, "epoch": 418, "lr": 9.552574381139682e-05}
{"train_loss": 0.3033624589443207, "global_step": 37264, "epoch": 418, "lr": 9.552550410098425e-05}
{"train_loss": 0.26750490069389343, "global_step": 37265, "epoch": 418, "lr": 9.552526438445133e-05}
{"train_loss": 0.34550419449806213, "global_step": 37266, "epoch": 418, "lr": 9.552502466179808e-05}
{"train_loss": 0.25875571370124817, "global_step": 37267, "epoch": 418, "lr": 9.552478493302455e-05}
{"train_loss": 0.41424867510795593, "global_step": 37268, "epoch": 418, "lr": 9.552454519813077e-05}
{"train_loss": 0.3361947238445282, "global_step": 37269, "epoch": 418, "lr": 9.552430545711677e-05}
{"train_loss": 0.27784061431884766, "global_step": 37270, "epoch": 418, "lr": 9.552406570998258e-05}
{"train_loss": 0.290546178817749, "global_step": 37271, "epoch": 418, "lr": 9.552382595672822e-05}
{"train_loss": 0.3522442579269409, "global_step": 37272, "epoch": 418, "lr": 9.552358619735376e-05}
{"train_loss": 0.3350875973701477, "global_step": 37273, "epoch": 418, "lr": 9.552334643185918e-05}
{"train_loss": 0.3674989342689514, "global_step": 37274, "epoch": 418, "lr": 9.552310666024457e-05}
{"train_loss": 0.2671559453010559, "global_step": 37275, "epoch": 418, "lr": 9.552286688250993e-05}
{"train_loss": 0.30831679701805115, "global_step": 37276, "epoch": 418, "lr": 9.552262709865528e-05}
{"train_loss": 0.34108632802963257, "global_step": 37277, "epoch": 418, "lr": 9.552238730868069e-05}
{"train_loss": 0.27263203263282776, "global_step": 37278, "epoch": 418, "lr": 9.552214751258614e-05}
{"train_loss": 0.32758861780166626, "global_step": 37279, "epoch": 418, "lr": 9.55219077103717e-05}
{"train_loss": 0.36583906412124634, "global_step": 37280, "epoch": 418, "lr": 9.552166790203741e-05}
{"train_loss": 0.31309011578559875, "global_step": 37281, "epoch": 418, "lr": 9.552142808758328e-05}
{"train_loss": 0.2891508638858795, "global_step": 37282, "epoch": 418, "lr": 9.552118826700935e-05}
{"train_loss": 0.4209657311439514, "global_step": 37283, "epoch": 418, "lr": 9.552094844031564e-05}
{"train_loss": 0.2794248163700104, "global_step": 37284, "epoch": 418, "lr": 9.552070860750219e-05}
{"train_loss": 0.29192355275154114, "global_step": 37285, "epoch": 418, "lr": 9.552046876856905e-05}
{"train_loss": 0.25884440541267395, "global_step": 37286, "epoch": 418, "lr": 9.552022892351625e-05}
{"train_loss": 0.23894459009170532, "global_step": 37287, "epoch": 418, "lr": 9.551998907234378e-05}
{"train_loss": 0.2794151306152344, "global_step": 37288, "epoch": 418, "lr": 9.551974921505172e-05}
{"train_loss": 0.39029553532600403, "global_step": 37289, "epoch": 418, "lr": 9.55195093516401e-05}
{"train_loss": 0.32220238329989187, "global_step": 37290, "epoch": 418, "lr": 9.55192694821089e-05, "val_loss": 2.6791179180145264}
{"train_loss": 0.3475983440876007, "global_step": 37291, "epoch": 419, "lr": 9.551902960645821e-05}
{"train_loss": 0.36186838150024414, "global_step": 37292, "epoch": 419, "lr": 9.551878972468804e-05}
{"train_loss": 0.33394721150398254, "global_step": 37293, "epoch": 419, "lr": 9.551854983679842e-05}
{"train_loss": 0.2768872380256653, "global_step": 37294, "epoch": 419, "lr": 9.551830994278939e-05}
{"train_loss": 0.27065643668174744, "global_step": 37295, "epoch": 419, "lr": 9.551807004266098e-05}
{"train_loss": 0.4261227250099182, "global_step": 37296, "epoch": 419, "lr": 9.551783013641321e-05}
{"train_loss": 0.3350277245044708, "global_step": 37297, "epoch": 419, "lr": 9.551759022404614e-05}
{"train_loss": 0.24560518562793732, "global_step": 37298, "epoch": 419, "lr": 9.551735030555976e-05}
{"train_loss": 0.3730204105377197, "global_step": 37299, "epoch": 419, "lr": 9.551711038095415e-05}
{"train_loss": 0.30484360456466675, "global_step": 37300, "epoch": 419, "lr": 9.55168704502293e-05}
{"train_loss": 0.26076987385749817, "global_step": 37301, "epoch": 419, "lr": 9.551663051338528e-05}
{"train_loss": 0.30658459663391113, "global_step": 37302, "epoch": 419, "lr": 9.551639057042209e-05}
{"train_loss": 0.3573962152004242, "global_step": 37303, "epoch": 419, "lr": 9.551615062133978e-05}
{"train_loss": 0.3143461048603058, "global_step": 37304, "epoch": 419, "lr": 9.551591066613839e-05}
{"train_loss": 0.3408312201499939, "global_step": 37305, "epoch": 419, "lr": 9.551567070481793e-05}
{"train_loss": 0.36746251583099365, "global_step": 37306, "epoch": 419, "lr": 9.551543073737843e-05}
{"train_loss": 0.39232438802719116, "global_step": 37307, "epoch": 419, "lr": 9.551519076381995e-05}
{"train_loss": 0.3392435908317566, "global_step": 37308, "epoch": 419, "lr": 9.551495078414249e-05}
{"train_loss": 0.24339936673641205, "global_step": 37309, "epoch": 419, "lr": 9.551471079834613e-05}
{"train_loss": 0.34572935104370117, "global_step": 37310, "epoch": 419, "lr": 9.551447080643085e-05}
{"train_loss": 0.2790027856826782, "global_step": 37311, "epoch": 419, "lr": 9.551423080839672e-05}
{"train_loss": 0.2652359902858734, "global_step": 37312, "epoch": 419, "lr": 9.551399080424375e-05}
{"train_loss": 0.39006805419921875, "global_step": 37313, "epoch": 419, "lr": 9.551375079397197e-05}
{"train_loss": 0.3613649010658264, "global_step": 37314, "epoch": 419, "lr": 9.551351077758142e-05}
{"train_loss": 0.26491522789001465, "global_step": 37315, "epoch": 419, "lr": 9.551327075507215e-05}
{"train_loss": 0.39604461193084717, "global_step": 37316, "epoch": 419, "lr": 9.551303072644417e-05}
{"train_loss": 0.3487738072872162, "global_step": 37317, "epoch": 419, "lr": 9.55127906916975e-05}
{"train_loss": 0.3961378335952759, "global_step": 37318, "epoch": 419, "lr": 9.55125506508322e-05}
{"train_loss": 0.3205542266368866, "global_step": 37319, "epoch": 419, "lr": 9.551231060384831e-05}
{"train_loss": 0.346849262714386, "global_step": 37320, "epoch": 419, "lr": 9.551207055074583e-05}
{"train_loss": 0.4015044867992401, "global_step": 37321, "epoch": 419, "lr": 9.55118304915248e-05}
{"train_loss": 0.3051756024360657, "global_step": 37322, "epoch": 419, "lr": 9.551159042618527e-05}
{"train_loss": 0.3914927542209625, "global_step": 37323, "epoch": 419, "lr": 9.551135035472725e-05}
{"train_loss": 0.28145018219947815, "global_step": 37324, "epoch": 419, "lr": 9.551111027715079e-05}
{"train_loss": 0.2718036472797394, "global_step": 37325, "epoch": 419, "lr": 9.551087019345592e-05}
{"train_loss": 0.338300496339798, "global_step": 37326, "epoch": 419, "lr": 9.551063010364267e-05}
{"train_loss": 0.3155737817287445, "global_step": 37327, "epoch": 419, "lr": 9.551039000771108e-05}
{"train_loss": 0.4233436584472656, "global_step": 37328, "epoch": 419, "lr": 9.551014990566114e-05}
{"train_loss": 0.38301369547843933, "global_step": 37329, "epoch": 419, "lr": 9.550990979749294e-05}
{"train_loss": 0.27913206815719604, "global_step": 37330, "epoch": 419, "lr": 9.550966968320647e-05}
{"train_loss": 0.24845357239246368, "global_step": 37331, "epoch": 419, "lr": 9.55094295628018e-05}
{"train_loss": 0.31941503286361694, "global_step": 37332, "epoch": 419, "lr": 9.550918943627893e-05}
{"train_loss": 0.2540278732776642, "global_step": 37333, "epoch": 419, "lr": 9.55089493036379e-05}
{"train_loss": 0.32855963706970215, "global_step": 37334, "epoch": 419, "lr": 9.550870916487875e-05}
{"train_loss": 0.30053985118865967, "global_step": 37335, "epoch": 419, "lr": 9.550846902000153e-05}
{"train_loss": 0.2530261278152466, "global_step": 37336, "epoch": 419, "lr": 9.550822886900622e-05}
{"train_loss": 0.34007540345191956, "global_step": 37337, "epoch": 419, "lr": 9.550798871189291e-05}
{"train_loss": 0.39538800716400146, "global_step": 37338, "epoch": 419, "lr": 9.550774854866158e-05}
{"train_loss": 0.3694847524166107, "global_step": 37339, "epoch": 419, "lr": 9.55075083793123e-05}
{"train_loss": 0.2952507436275482, "global_step": 37340, "epoch": 419, "lr": 9.550726820384508e-05}
{"train_loss": 0.36295127868652344, "global_step": 37341, "epoch": 419, "lr": 9.550702802225999e-05}
{"train_loss": 0.4269876480102539, "global_step": 37342, "epoch": 419, "lr": 9.550678783455701e-05}
{"train_loss": 0.2701086699962616, "global_step": 37343, "epoch": 419, "lr": 9.55065476407362e-05}
{"train_loss": 0.3404653072357178, "global_step": 37344, "epoch": 419, "lr": 9.55063074407976e-05}
{"train_loss": 0.3915470242500305, "global_step": 37345, "epoch": 419, "lr": 9.550606723474122e-05}
{"train_loss": 0.26065200567245483, "global_step": 37346, "epoch": 419, "lr": 9.55058270225671e-05}
{"train_loss": 0.30728912353515625, "global_step": 37347, "epoch": 419, "lr": 9.55055868042753e-05}
{"train_loss": 0.49259671568870544, "global_step": 37348, "epoch": 419, "lr": 9.55053465798658e-05}
{"train_loss": 0.4144871234893799, "global_step": 37349, "epoch": 419, "lr": 9.550510634933868e-05}
{"train_loss": 0.4257415235042572, "global_step": 37350, "epoch": 419, "lr": 9.550486611269395e-05}
{"train_loss": 0.3633153438568115, "global_step": 37351, "epoch": 419, "lr": 9.550462586993163e-05}
{"train_loss": 0.2832297384738922, "global_step": 37352, "epoch": 419, "lr": 9.55043856210518e-05}
{"train_loss": 0.31963345408439636, "global_step": 37353, "epoch": 419, "lr": 9.550414536605442e-05}
{"train_loss": 0.30166560411453247, "global_step": 37354, "epoch": 419, "lr": 9.550390510493959e-05}
{"train_loss": 0.32921209931373596, "global_step": 37355, "epoch": 419, "lr": 9.55036648377073e-05}
{"train_loss": 0.3243659734725952, "global_step": 37356, "epoch": 419, "lr": 9.550342456435761e-05}
{"train_loss": 0.27595195174217224, "global_step": 37357, "epoch": 419, "lr": 9.550318428489054e-05}
{"train_loss": 0.3925566077232361, "global_step": 37358, "epoch": 419, "lr": 9.55029439993061e-05}
{"train_loss": 0.34300869703292847, "global_step": 37359, "epoch": 419, "lr": 9.550270370760436e-05}
{"train_loss": 0.3775652050971985, "global_step": 37360, "epoch": 419, "lr": 9.550246340978535e-05}
{"train_loss": 0.2620718479156494, "global_step": 37361, "epoch": 419, "lr": 9.550222310584907e-05}
{"train_loss": 0.2506211996078491, "global_step": 37362, "epoch": 419, "lr": 9.550198279579556e-05}
{"train_loss": 0.38410353660583496, "global_step": 37363, "epoch": 419, "lr": 9.55017424796249e-05}
{"train_loss": 0.2892191708087921, "global_step": 37364, "epoch": 419, "lr": 9.550150215733704e-05}
{"train_loss": 0.3016549348831177, "global_step": 37365, "epoch": 419, "lr": 9.550126182893209e-05}
{"train_loss": 0.36669325828552246, "global_step": 37366, "epoch": 419, "lr": 9.550102149441003e-05}
{"train_loss": 0.3554098308086395, "global_step": 37367, "epoch": 419, "lr": 9.550078115377093e-05}
{"train_loss": 0.26569679379463196, "global_step": 37368, "epoch": 419, "lr": 9.550054080701479e-05}
{"train_loss": 0.24826009571552277, "global_step": 37369, "epoch": 419, "lr": 9.550030045414167e-05}
{"train_loss": 0.3477139472961426, "global_step": 37370, "epoch": 419, "lr": 9.550006009515158e-05}
{"train_loss": 0.34380611777305603, "global_step": 37371, "epoch": 419, "lr": 9.549981973004456e-05}
{"train_loss": 0.3070908188819885, "global_step": 37372, "epoch": 419, "lr": 9.549957935882064e-05}
{"train_loss": 0.35144123435020447, "global_step": 37373, "epoch": 419, "lr": 9.549933898147986e-05}
{"train_loss": 0.31811922788619995, "global_step": 37374, "epoch": 419, "lr": 9.549909859802226e-05}
{"train_loss": 0.32585352659225464, "global_step": 37375, "epoch": 419, "lr": 9.549885820844786e-05}
{"train_loss": 0.32420220971107483, "global_step": 37376, "epoch": 419, "lr": 9.549861781275668e-05}
{"train_loss": 0.34465545415878296, "global_step": 37377, "epoch": 419, "lr": 9.549837741094878e-05}
{"train_loss": 0.3667321801185608, "global_step": 37378, "epoch": 419, "lr": 9.549813700302417e-05}
{"train_loss": 0.3326967040474495, "global_step": 37379, "epoch": 419, "lr": 9.549789658898288e-05, "val_loss": 2.6762948036193848}
{"train_loss": 0.37714970111846924, "global_step": 37380, "epoch": 420, "lr": 9.549765616882496e-05}
{"train_loss": 0.2727067768573761, "global_step": 37381, "epoch": 420, "lr": 9.549741574255045e-05}
{"train_loss": 0.2928132712841034, "global_step": 37382, "epoch": 420, "lr": 9.549717531015933e-05}
{"train_loss": 0.327958345413208, "global_step": 37383, "epoch": 420, "lr": 9.54969348716517e-05}
{"train_loss": 0.3989430069923401, "global_step": 37384, "epoch": 420, "lr": 9.549669442702755e-05}
{"train_loss": 0.35673364996910095, "global_step": 37385, "epoch": 420, "lr": 9.549645397628694e-05}
{"train_loss": 0.30578547716140747, "global_step": 37386, "epoch": 420, "lr": 9.549621351942986e-05}
{"train_loss": 0.41915491223335266, "global_step": 37387, "epoch": 420, "lr": 9.54959730564564e-05}
{"train_loss": 0.25148648023605347, "global_step": 37388, "epoch": 420, "lr": 9.549573258736654e-05}
{"train_loss": 0.324447900056839, "global_step": 37389, "epoch": 420, "lr": 9.549549211216033e-05}
{"train_loss": 0.311069130897522, "global_step": 37390, "epoch": 420, "lr": 9.549525163083781e-05}
{"train_loss": 0.3657401204109192, "global_step": 37391, "epoch": 420, "lr": 9.549501114339902e-05}
{"train_loss": 0.3261233866214752, "global_step": 37392, "epoch": 420, "lr": 9.549477064984397e-05}
{"train_loss": 0.22372563183307648, "global_step": 37393, "epoch": 420, "lr": 9.54945301501727e-05}
{"train_loss": 0.34673261642456055, "global_step": 37394, "epoch": 420, "lr": 9.549428964438525e-05}
{"train_loss": 0.3462248146533966, "global_step": 37395, "epoch": 420, "lr": 9.549404913248164e-05}
{"train_loss": 0.34508615732192993, "global_step": 37396, "epoch": 420, "lr": 9.54938086144619e-05}
{"train_loss": 0.2756510078907013, "global_step": 37397, "epoch": 420, "lr": 9.54935680903261e-05}
{"train_loss": 0.426439493894577, "global_step": 37398, "epoch": 420, "lr": 9.549332756007423e-05}
{"train_loss": 0.3830746114253998, "global_step": 37399, "epoch": 420, "lr": 9.549308702370634e-05}
{"train_loss": 0.3020866811275482, "global_step": 37400, "epoch": 420, "lr": 9.549284648122245e-05}
{"train_loss": 0.4560309946537018, "global_step": 37401, "epoch": 420, "lr": 9.549260593262261e-05}
{"train_loss": 0.26084578037261963, "global_step": 37402, "epoch": 420, "lr": 9.549236537790684e-05}
{"train_loss": 0.3812626600265503, "global_step": 37403, "epoch": 420, "lr": 9.549212481707519e-05}
{"train_loss": 0.4243277907371521, "global_step": 37404, "epoch": 420, "lr": 9.549188425012767e-05}
{"train_loss": 0.247322678565979, "global_step": 37405, "epoch": 420, "lr": 9.549164367706431e-05}
{"train_loss": 0.23854047060012817, "global_step": 37406, "epoch": 420, "lr": 9.549140309788516e-05}
{"train_loss": 0.4299117624759674, "global_step": 37407, "epoch": 420, "lr": 9.549116251259024e-05}
{"train_loss": 0.37751370668411255, "global_step": 37408, "epoch": 420, "lr": 9.549092192117958e-05}
{"train_loss": 0.3143264353275299, "global_step": 37409, "epoch": 420, "lr": 9.549068132365324e-05}
{"train_loss": 0.2952049970626831, "global_step": 37410, "epoch": 420, "lr": 9.549044072001124e-05}
{"train_loss": 0.3747754991054535, "global_step": 37411, "epoch": 420, "lr": 9.549020011025359e-05}
{"train_loss": 0.37480324506759644, "global_step": 37412, "epoch": 420, "lr": 9.548995949438033e-05}
{"train_loss": 0.4121745824813843, "global_step": 37413, "epoch": 420, "lr": 9.548971887239151e-05}
{"train_loss": 0.3212321102619171, "global_step": 37414, "epoch": 420, "lr": 9.548947824428714e-05}
{"train_loss": 0.37271130084991455, "global_step": 37415, "epoch": 420, "lr": 9.548923761006728e-05}
{"train_loss": 0.28435567021369934, "global_step": 37416, "epoch": 420, "lr": 9.548899696973193e-05}
{"train_loss": 0.44555768370628357, "global_step": 37417, "epoch": 420, "lr": 9.548875632328115e-05}
{"train_loss": 0.3636407256126404, "global_step": 37418, "epoch": 420, "lr": 9.548851567071496e-05}
{"train_loss": 0.32365453243255615, "global_step": 37419, "epoch": 420, "lr": 9.548827501203339e-05}
{"train_loss": 0.3256593346595764, "global_step": 37420, "epoch": 420, "lr": 9.548803434723647e-05}
{"train_loss": 0.24720822274684906, "global_step": 37421, "epoch": 420, "lr": 9.548779367632425e-05}
{"train_loss": 0.25419124960899353, "global_step": 37422, "epoch": 420, "lr": 9.548755299929675e-05}
{"train_loss": 0.20582586526870728, "global_step": 37423, "epoch": 420, "lr": 9.548731231615399e-05}
{"train_loss": 0.2957528531551361, "global_step": 37424, "epoch": 420, "lr": 9.548707162689603e-05}
{"train_loss": 0.3584178388118744, "global_step": 37425, "epoch": 420, "lr": 9.548683093152289e-05}
{"train_loss": 0.24280458688735962, "global_step": 37426, "epoch": 420, "lr": 9.548659023003459e-05}
{"train_loss": 0.35480812191963196, "global_step": 37427, "epoch": 420, "lr": 9.548634952243117e-05}
{"train_loss": 0.300737202167511, "global_step": 37428, "epoch": 420, "lr": 9.548610880871268e-05}
{"train_loss": 0.4048744738101959, "global_step": 37429, "epoch": 420, "lr": 9.548586808887911e-05}
{"train_loss": 0.29077085852622986, "global_step": 37430, "epoch": 420, "lr": 9.548562736293054e-05}
{"train_loss": 0.3549470901489258, "global_step": 37431, "epoch": 420, "lr": 9.548538663086698e-05}
{"train_loss": 0.37226319313049316, "global_step": 37432, "epoch": 420, "lr": 9.548514589268847e-05}
{"train_loss": 0.35907575488090515, "global_step": 37433, "epoch": 420, "lr": 9.548490514839502e-05}
{"train_loss": 0.4178706109523773, "global_step": 37434, "epoch": 420, "lr": 9.54846643979867e-05}
{"train_loss": 0.323160320520401, "global_step": 37435, "epoch": 420, "lr": 9.54844236414635e-05}
{"train_loss": 0.341621071100235, "global_step": 37436, "epoch": 420, "lr": 9.548418287882548e-05}
{"train_loss": 0.23911140859127045, "global_step": 37437, "epoch": 420, "lr": 9.548394211007266e-05}
{"train_loss": 0.29941943287849426, "global_step": 37438, "epoch": 420, "lr": 9.548370133520509e-05}
{"train_loss": 0.2230980098247528, "global_step": 37439, "epoch": 420, "lr": 9.548346055422279e-05}
{"train_loss": 0.3266385793685913, "global_step": 37440, "epoch": 420, "lr": 9.548321976712578e-05}
{"train_loss": 0.43728938698768616, "global_step": 37441, "epoch": 420, "lr": 9.548297897391412e-05}
{"train_loss": 0.3448331654071808, "global_step": 37442, "epoch": 420, "lr": 9.548273817458783e-05}
{"train_loss": 0.3689184784889221, "global_step": 37443, "epoch": 420, "lr": 9.548249736914694e-05}
{"train_loss": 0.3660595417022705, "global_step": 37444, "epoch": 420, "lr": 9.548225655759146e-05}
{"train_loss": 0.33826303482055664, "global_step": 37445, "epoch": 420, "lr": 9.548201573992146e-05}
{"train_loss": 0.2411576509475708, "global_step": 37446, "epoch": 420, "lr": 9.548177491613697e-05}
{"train_loss": 0.3472575843334198, "global_step": 37447, "epoch": 420, "lr": 9.5481534086238e-05}
{"train_loss": 0.3420839309692383, "global_step": 37448, "epoch": 420, "lr": 9.548129325022459e-05}
{"train_loss": 0.33980047702789307, "global_step": 37449, "epoch": 420, "lr": 9.548105240809677e-05}
{"train_loss": 0.3248525559902191, "global_step": 37450, "epoch": 420, "lr": 9.548081155985457e-05}
{"train_loss": 0.2902381718158722, "global_step": 37451, "epoch": 420, "lr": 9.548057070549805e-05}
{"train_loss": 0.34222733974456787, "global_step": 37452, "epoch": 420, "lr": 9.548032984502721e-05}
{"train_loss": 0.4003699719905853, "global_step": 37453, "epoch": 420, "lr": 9.54800889784421e-05}
{"train_loss": 0.4605904817581177, "global_step": 37454, "epoch": 420, "lr": 9.547984810574274e-05}
{"train_loss": 0.3658176362514496, "global_step": 37455, "epoch": 420, "lr": 9.547960722692916e-05}
{"train_loss": 0.35069528222084045, "global_step": 37456, "epoch": 420, "lr": 9.547936634200142e-05}
{"train_loss": 0.27552473545074463, "global_step": 37457, "epoch": 420, "lr": 9.547912545095954e-05}
{"train_loss": 0.3360252380371094, "global_step": 37458, "epoch": 420, "lr": 9.547888455380352e-05}
{"train_loss": 0.2755579352378845, "global_step": 37459, "epoch": 420, "lr": 9.547864365053342e-05}
{"train_loss": 0.4036199748516083, "global_step": 37460, "epoch": 420, "lr": 9.547840274114929e-05}
{"train_loss": 0.3843972086906433, "global_step": 37461, "epoch": 420, "lr": 9.547816182565114e-05}
{"train_loss": 0.46125441789627075, "global_step": 37462, "epoch": 420, "lr": 9.5477920904039e-05}
{"train_loss": 0.23546139895915985, "global_step": 37463, "epoch": 420, "lr": 9.547767997631291e-05}
{"train_loss": 0.2965054512023926, "global_step": 37464, "epoch": 420, "lr": 9.547743904247289e-05}
{"train_loss": 0.32530128955841064, "global_step": 37465, "epoch": 420, "lr": 9.547719810251899e-05}
{"train_loss": 0.2914281189441681, "global_step": 37466, "epoch": 420, "lr": 9.547695715645124e-05}
{"train_loss": 0.3236885666847229, "global_step": 37467, "epoch": 420, "lr": 9.547671620426966e-05}
{"train_loss": 0.3354058228851704, "global_step": 37468, "epoch": 420, "lr": 9.54764752459743e-05, "val_loss": 2.5127384662628174, "train_action_mse_error": 25.46997833251953}
{"train_loss": 0.34947896003723145, "global_step": 37469, "epoch": 421, "lr": 9.547623428156517e-05}
{"train_loss": 0.2924591898918152, "global_step": 37470, "epoch": 421, "lr": 9.547599331104231e-05}
{"train_loss": 0.3397557735443115, "global_step": 37471, "epoch": 421, "lr": 9.547575233440578e-05}
{"train_loss": 0.26342177391052246, "global_step": 37472, "epoch": 421, "lr": 9.547551135165556e-05}
{"train_loss": 0.3841519057750702, "global_step": 37473, "epoch": 421, "lr": 9.547527036279174e-05}
{"train_loss": 0.3582044243812561, "global_step": 37474, "epoch": 421, "lr": 9.547502936781431e-05}
{"train_loss": 0.2531820833683014, "global_step": 37475, "epoch": 421, "lr": 9.547478836672332e-05}
{"train_loss": 0.44241783022880554, "global_step": 37476, "epoch": 421, "lr": 9.54745473595188e-05}
{"train_loss": 0.3083210289478302, "global_step": 37477, "epoch": 421, "lr": 9.547430634620077e-05}
{"train_loss": 0.37252572178840637, "global_step": 37478, "epoch": 421, "lr": 9.547406532676928e-05}
{"train_loss": 0.30512571334838867, "global_step": 37479, "epoch": 421, "lr": 9.547382430122436e-05}
{"train_loss": 0.3611465096473694, "global_step": 37480, "epoch": 421, "lr": 9.547358326956604e-05}
{"train_loss": 0.31260940432548523, "global_step": 37481, "epoch": 421, "lr": 9.547334223179433e-05}
{"train_loss": 0.27205801010131836, "global_step": 37482, "epoch": 421, "lr": 9.547310118790931e-05}
{"train_loss": 0.3461134135723114, "global_step": 37483, "epoch": 421, "lr": 9.547286013791098e-05}
{"train_loss": 0.4481309950351715, "global_step": 37484, "epoch": 421, "lr": 9.547261908179937e-05}
{"train_loss": 0.35209259390830994, "global_step": 37485, "epoch": 421, "lr": 9.547237801957454e-05}
{"train_loss": 0.42867809534072876, "global_step": 37486, "epoch": 421, "lr": 9.547213695123648e-05}
{"train_loss": 0.2870325744152069, "global_step": 37487, "epoch": 421, "lr": 9.547189587678525e-05}
{"train_loss": 0.3443325459957123, "global_step": 37488, "epoch": 421, "lr": 9.547165479622089e-05}
{"train_loss": 0.3952890932559967, "global_step": 37489, "epoch": 421, "lr": 9.54714137095434e-05}
{"train_loss": 0.4735592305660248, "global_step": 37490, "epoch": 421, "lr": 9.547117261675285e-05}
{"train_loss": 0.3254548907279968, "global_step": 37491, "epoch": 421, "lr": 9.547093151784924e-05}
{"train_loss": 0.3862656354904175, "global_step": 37492, "epoch": 421, "lr": 9.547069041283264e-05}
{"train_loss": 0.4010293483734131, "global_step": 37493, "epoch": 421, "lr": 9.547044930170304e-05}
{"train_loss": 0.2877565920352936, "global_step": 37494, "epoch": 421, "lr": 9.547020818446051e-05}
{"train_loss": 0.38308605551719666, "global_step": 37495, "epoch": 421, "lr": 9.546996706110505e-05}
{"train_loss": 0.346330851316452, "global_step": 37496, "epoch": 421, "lr": 9.54697259316367e-05}
{"train_loss": 0.32571959495544434, "global_step": 37497, "epoch": 421, "lr": 9.546948479605553e-05}
{"train_loss": 0.2991699278354645, "global_step": 37498, "epoch": 421, "lr": 9.546924365436152e-05}
{"train_loss": 0.31670379638671875, "global_step": 37499, "epoch": 421, "lr": 9.546900250655473e-05}
{"train_loss": 0.460531085729599, "global_step": 37500, "epoch": 421, "lr": 9.546876135263518e-05}
{"train_loss": 0.29342642426490784, "global_step": 37501, "epoch": 421, "lr": 9.54685201926029e-05}
{"train_loss": 0.33050674200057983, "global_step": 37502, "epoch": 421, "lr": 9.546827902645795e-05}
{"train_loss": 0.29177001118659973, "global_step": 37503, "epoch": 421, "lr": 9.546803785420034e-05}
{"train_loss": 0.27840012311935425, "global_step": 37504, "epoch": 421, "lr": 9.54677966758301e-05}
{"train_loss": 0.3486067056655884, "global_step": 37505, "epoch": 421, "lr": 9.546755549134728e-05}
{"train_loss": 0.27230945229530334, "global_step": 37506, "epoch": 421, "lr": 9.54673143007519e-05}
{"train_loss": 0.41525009274482727, "global_step": 37507, "epoch": 421, "lr": 9.546707310404399e-05}
{"train_loss": 0.4606666564941406, "global_step": 37508, "epoch": 421, "lr": 9.546683190122359e-05}
{"train_loss": 0.37435969710350037, "global_step": 37509, "epoch": 421, "lr": 9.546659069229072e-05}
{"train_loss": 0.364911824464798, "global_step": 37510, "epoch": 421, "lr": 9.546634947724543e-05}
{"train_loss": 0.33312687277793884, "global_step": 37511, "epoch": 421, "lr": 9.546610825608773e-05}
{"train_loss": 0.3661847412586212, "global_step": 37512, "epoch": 421, "lr": 9.546586702881767e-05}
{"train_loss": 0.2944320738315582, "global_step": 37513, "epoch": 421, "lr": 9.546562579543528e-05}
{"train_loss": 0.37280765175819397, "global_step": 37514, "epoch": 421, "lr": 9.546538455594061e-05}
{"train_loss": 0.36930063366889954, "global_step": 37515, "epoch": 421, "lr": 9.546514331033365e-05}
{"train_loss": 0.4409120976924896, "global_step": 37516, "epoch": 421, "lr": 9.546490205861446e-05}
{"train_loss": 0.35123372077941895, "global_step": 37517, "epoch": 421, "lr": 9.546466080078306e-05}
{"train_loss": 0.43661871552467346, "global_step": 37518, "epoch": 421, "lr": 9.54644195368395e-05}
{"train_loss": 0.2952115833759308, "global_step": 37519, "epoch": 421, "lr": 9.546417826678383e-05}
{"train_loss": 0.331094890832901, "global_step": 37520, "epoch": 421, "lr": 9.546393699061603e-05}
{"train_loss": 0.3779700696468353, "global_step": 37521, "epoch": 421, "lr": 9.546369570833615e-05}
{"train_loss": 0.35431382060050964, "global_step": 37522, "epoch": 421, "lr": 9.546345441994424e-05}
{"train_loss": 0.38264039158821106, "global_step": 37523, "epoch": 421, "lr": 9.546321312544031e-05}
{"train_loss": 0.3209225833415985, "global_step": 37524, "epoch": 421, "lr": 9.546297182482442e-05}
{"train_loss": 0.4047349691390991, "global_step": 37525, "epoch": 421, "lr": 9.546273051809658e-05}
{"train_loss": 0.3518943190574646, "global_step": 37526, "epoch": 421, "lr": 9.546248920525683e-05}
{"train_loss": 0.28353315591812134, "global_step": 37527, "epoch": 421, "lr": 9.546224788630521e-05}
{"train_loss": 0.2820366621017456, "global_step": 37528, "epoch": 421, "lr": 9.546200656124173e-05}
{"train_loss": 0.4238629639148712, "global_step": 37529, "epoch": 421, "lr": 9.546176523006645e-05}
{"train_loss": 0.2665690779685974, "global_step": 37530, "epoch": 421, "lr": 9.546152389277939e-05}
{"train_loss": 0.42957890033721924, "global_step": 37531, "epoch": 421, "lr": 9.546128254938058e-05}
{"train_loss": 0.3226308524608612, "global_step": 37532, "epoch": 421, "lr": 9.546104119987005e-05}
{"train_loss": 0.26521557569503784, "global_step": 37533, "epoch": 421, "lr": 9.546079984424785e-05}
{"train_loss": 0.322037011384964, "global_step": 37534, "epoch": 421, "lr": 9.546055848251399e-05}
{"train_loss": 0.31582731008529663, "global_step": 37535, "epoch": 421, "lr": 9.54603171146685e-05}
{"train_loss": 0.34291741251945496, "global_step": 37536, "epoch": 421, "lr": 9.546007574071146e-05}
{"train_loss": 0.4141328036785126, "global_step": 37537, "epoch": 421, "lr": 9.545983436064283e-05}
{"train_loss": 0.28790491819381714, "global_step": 37538, "epoch": 421, "lr": 9.54595929744627e-05}
{"train_loss": 0.31686925888061523, "global_step": 37539, "epoch": 421, "lr": 9.545935158217107e-05}
{"train_loss": 0.4139937460422516, "global_step": 37540, "epoch": 421, "lr": 9.545911018376799e-05}
{"train_loss": 0.35143885016441345, "global_step": 37541, "epoch": 421, "lr": 9.545886877925349e-05}
{"train_loss": 0.3211241662502289, "global_step": 37542, "epoch": 421, "lr": 9.545862736862758e-05}
{"train_loss": 0.3172234296798706, "global_step": 37543, "epoch": 421, "lr": 9.545838595189034e-05}
{"train_loss": 0.33878573775291443, "global_step": 37544, "epoch": 421, "lr": 9.545814452904176e-05}
{"train_loss": 0.44594135880470276, "global_step": 37545, "epoch": 421, "lr": 9.545790310008189e-05}
{"train_loss": 0.3249208331108093, "global_step": 37546, "epoch": 421, "lr": 9.545766166501075e-05}
{"train_loss": 0.4642266631126404, "global_step": 37547, "epoch": 421, "lr": 9.54574202238284e-05}
{"train_loss": 0.36577484011650085, "global_step": 37548, "epoch": 421, "lr": 9.545717877653484e-05}
{"train_loss": 0.3446728587150574, "global_step": 37549, "epoch": 421, "lr": 9.54569373231301e-05}
{"train_loss": 0.29379892349243164, "global_step": 37550, "epoch": 421, "lr": 9.545669586361425e-05}
{"train_loss": 0.3279925584793091, "global_step": 37551, "epoch": 421, "lr": 9.545645439798731e-05}
{"train_loss": 0.3118841350078583, "global_step": 37552, "epoch": 421, "lr": 9.54562129262493e-05}
{"train_loss": 0.4154619574546814, "global_step": 37553, "epoch": 421, "lr": 9.545597144840024e-05}
{"train_loss": 0.40933236479759216, "global_step": 37554, "epoch": 421, "lr": 9.545572996444018e-05}
{"train_loss": 0.449270635843277, "global_step": 37555, "epoch": 421, "lr": 9.545548847436916e-05}
{"train_loss": 0.2607559263706207, "global_step": 37556, "epoch": 421, "lr": 9.54552469781872e-05}
{"train_loss": 0.35092368420590175, "global_step": 37557, "epoch": 421, "lr": 9.545500547589434e-05, "val_loss": 2.596853733062744}
{"train_loss": 0.2612614631652832, "global_step": 37558, "epoch": 422, "lr": 9.54547639674906e-05}
{"train_loss": 0.3928643763065338, "global_step": 37559, "epoch": 422, "lr": 9.545452245297603e-05}
{"train_loss": 0.3931938409805298, "global_step": 37560, "epoch": 422, "lr": 9.545428093235065e-05}
{"train_loss": 0.33277520537376404, "global_step": 37561, "epoch": 422, "lr": 9.54540394056145e-05}
{"train_loss": 0.2886136472225189, "global_step": 37562, "epoch": 422, "lr": 9.54537978727676e-05}
{"train_loss": 0.3334812819957733, "global_step": 37563, "epoch": 422, "lr": 9.545355633381e-05}
{"train_loss": 0.27082210779190063, "global_step": 37564, "epoch": 422, "lr": 9.545331478874173e-05}
{"train_loss": 0.27542680501937866, "global_step": 37565, "epoch": 422, "lr": 9.545307323756279e-05}
{"train_loss": 0.374186247587204, "global_step": 37566, "epoch": 422, "lr": 9.545283168027325e-05}
{"train_loss": 0.2974230647087097, "global_step": 37567, "epoch": 422, "lr": 9.545259011687314e-05}
{"train_loss": 0.3701007664203644, "global_step": 37568, "epoch": 422, "lr": 9.545234854736247e-05}
{"train_loss": 0.3452474772930145, "global_step": 37569, "epoch": 422, "lr": 9.54521069717413e-05}
{"train_loss": 0.2955682873725891, "global_step": 37570, "epoch": 422, "lr": 9.545186539000964e-05}
{"train_loss": 0.23934690654277802, "global_step": 37571, "epoch": 422, "lr": 9.545162380216753e-05}
{"train_loss": 0.29086750745773315, "global_step": 37572, "epoch": 422, "lr": 9.5451382208215e-05}
{"train_loss": 0.35495612025260925, "global_step": 37573, "epoch": 422, "lr": 9.54511406081521e-05}
{"train_loss": 0.3967004120349884, "global_step": 37574, "epoch": 422, "lr": 9.545089900197883e-05}
{"train_loss": 0.2662724256515503, "global_step": 37575, "epoch": 422, "lr": 9.545065738969527e-05}
{"train_loss": 0.32001644372940063, "global_step": 37576, "epoch": 422, "lr": 9.54504157713014e-05}
{"train_loss": 0.36030423641204834, "global_step": 37577, "epoch": 422, "lr": 9.545017414679727e-05}
{"train_loss": 0.3219052255153656, "global_step": 37578, "epoch": 422, "lr": 9.544993251618292e-05}
{"train_loss": 0.48457083106040955, "global_step": 37579, "epoch": 422, "lr": 9.544969087945839e-05}
{"train_loss": 0.26261106133461, "global_step": 37580, "epoch": 422, "lr": 9.544944923662372e-05}
{"train_loss": 0.32907670736312866, "global_step": 37581, "epoch": 422, "lr": 9.544920758767889e-05}
{"train_loss": 0.29112187027931213, "global_step": 37582, "epoch": 422, "lr": 9.544896593262398e-05}
{"train_loss": 0.3774479925632477, "global_step": 37583, "epoch": 422, "lr": 9.544872427145901e-05}
{"train_loss": 0.3485923707485199, "global_step": 37584, "epoch": 422, "lr": 9.544848260418402e-05}
{"train_loss": 0.32559651136398315, "global_step": 37585, "epoch": 422, "lr": 9.544824093079903e-05}
{"train_loss": 0.24885733425617218, "global_step": 37586, "epoch": 422, "lr": 9.544799925130409e-05}
{"train_loss": 0.24644626677036285, "global_step": 37587, "epoch": 422, "lr": 9.544775756569921e-05}
{"train_loss": 0.35960420966148376, "global_step": 37588, "epoch": 422, "lr": 9.544751587398442e-05}
{"train_loss": 0.3381389379501343, "global_step": 37589, "epoch": 422, "lr": 9.54472741761598e-05}
{"train_loss": 0.26432034373283386, "global_step": 37590, "epoch": 422, "lr": 9.54470324722253e-05}
{"train_loss": 0.3298823833465576, "global_step": 37591, "epoch": 422, "lr": 9.544679076218104e-05}
{"train_loss": 0.31055763363838196, "global_step": 37592, "epoch": 422, "lr": 9.544654904602701e-05}
{"train_loss": 0.3343732953071594, "global_step": 37593, "epoch": 422, "lr": 9.544630732376323e-05}
{"train_loss": 0.3878132998943329, "global_step": 37594, "epoch": 422, "lr": 9.544606559538975e-05}
{"train_loss": 0.3177839517593384, "global_step": 37595, "epoch": 422, "lr": 9.544582386090661e-05}
{"train_loss": 0.25578224658966064, "global_step": 37596, "epoch": 422, "lr": 9.544558212031383e-05}
{"train_loss": 0.2903279960155487, "global_step": 37597, "epoch": 422, "lr": 9.544534037361144e-05}
{"train_loss": 0.38712894916534424, "global_step": 37598, "epoch": 422, "lr": 9.544509862079947e-05}
{"train_loss": 0.33145567774772644, "global_step": 37599, "epoch": 422, "lr": 9.544485686187798e-05}
{"train_loss": 0.3412543535232544, "global_step": 37600, "epoch": 422, "lr": 9.544461509684697e-05}
{"train_loss": 0.4829915463924408, "global_step": 37601, "epoch": 422, "lr": 9.544437332570648e-05}
{"train_loss": 0.4718642234802246, "global_step": 37602, "epoch": 422, "lr": 9.544413154845657e-05}
{"train_loss": 0.2942754328250885, "global_step": 37603, "epoch": 422, "lr": 9.544388976509723e-05}
{"train_loss": 0.2909645736217499, "global_step": 37604, "epoch": 422, "lr": 9.544364797562853e-05}
{"train_loss": 0.3750576674938202, "global_step": 37605, "epoch": 422, "lr": 9.544340618005046e-05}
{"train_loss": 0.27506232261657715, "global_step": 37606, "epoch": 422, "lr": 9.54431643783631e-05}
{"train_loss": 0.35077616572380066, "global_step": 37607, "epoch": 422, "lr": 9.544292257056645e-05}
{"train_loss": 0.2755239009857178, "global_step": 37608, "epoch": 422, "lr": 9.544268075666057e-05}
{"train_loss": 0.280117005109787, "global_step": 37609, "epoch": 422, "lr": 9.544243893664545e-05}
{"train_loss": 0.34188511967658997, "global_step": 37610, "epoch": 422, "lr": 9.544219711052115e-05}
{"train_loss": 0.2787885069847107, "global_step": 37611, "epoch": 422, "lr": 9.544195527828772e-05}
{"train_loss": 0.304823637008667, "global_step": 37612, "epoch": 422, "lr": 9.544171343994517e-05}
{"train_loss": 0.30546730756759644, "global_step": 37613, "epoch": 422, "lr": 9.544147159549353e-05}
{"train_loss": 0.3303697407245636, "global_step": 37614, "epoch": 422, "lr": 9.544122974493283e-05}
{"train_loss": 0.3128613233566284, "global_step": 37615, "epoch": 422, "lr": 9.544098788826312e-05}
{"train_loss": 0.3317372500896454, "global_step": 37616, "epoch": 422, "lr": 9.544074602548441e-05}
{"train_loss": 0.33569854497909546, "global_step": 37617, "epoch": 422, "lr": 9.544050415659675e-05}
{"train_loss": 0.42438480257987976, "global_step": 37618, "epoch": 422, "lr": 9.544026228160019e-05}
{"train_loss": 0.31863757967948914, "global_step": 37619, "epoch": 422, "lr": 9.544002040049471e-05}
{"train_loss": 0.3474985957145691, "global_step": 37620, "epoch": 422, "lr": 9.54397785132804e-05}
{"train_loss": 0.363716721534729, "global_step": 37621, "epoch": 422, "lr": 9.543953661995724e-05}
{"train_loss": 0.24242106080055237, "global_step": 37622, "epoch": 422, "lr": 9.54392947205253e-05}
{"train_loss": 0.27388396859169006, "global_step": 37623, "epoch": 422, "lr": 9.543905281498462e-05}
{"train_loss": 0.37930747866630554, "global_step": 37624, "epoch": 422, "lr": 9.543881090333519e-05}
{"train_loss": 0.3197658956050873, "global_step": 37625, "epoch": 422, "lr": 9.543856898557707e-05}
{"train_loss": 0.42354997992515564, "global_step": 37626, "epoch": 422, "lr": 9.543832706171029e-05}
{"train_loss": 0.3945717513561249, "global_step": 37627, "epoch": 422, "lr": 9.543808513173488e-05}
{"train_loss": 0.2870529890060425, "global_step": 37628, "epoch": 422, "lr": 9.543784319565088e-05}
{"train_loss": 0.3099903464317322, "global_step": 37629, "epoch": 422, "lr": 9.54376012534583e-05}
{"train_loss": 0.4127323031425476, "global_step": 37630, "epoch": 422, "lr": 9.54373593051572e-05}
{"train_loss": 0.3874478042125702, "global_step": 37631, "epoch": 422, "lr": 9.54371173507476e-05}
{"train_loss": 0.29585492610931396, "global_step": 37632, "epoch": 422, "lr": 9.543687539022951e-05}
{"train_loss": 0.40333643555641174, "global_step": 37633, "epoch": 422, "lr": 9.543663342360302e-05}
{"train_loss": 0.3738701343536377, "global_step": 37634, "epoch": 422, "lr": 9.543639145086811e-05}
{"train_loss": 0.35010966658592224, "global_step": 37635, "epoch": 422, "lr": 9.543614947202484e-05}
{"train_loss": 0.2579093873500824, "global_step": 37636, "epoch": 422, "lr": 9.543590748707322e-05}
{"train_loss": 0.4205072522163391, "global_step": 37637, "epoch": 422, "lr": 9.543566549601333e-05}
{"train_loss": 0.2510235905647278, "global_step": 37638, "epoch": 422, "lr": 9.543542349884513e-05}
{"train_loss": 0.3110833466053009, "global_step": 37639, "epoch": 422, "lr": 9.543518149556872e-05}
{"train_loss": 0.42092055082321167, "global_step": 37640, "epoch": 422, "lr": 9.543493948618409e-05}
{"train_loss": 0.471912682056427, "global_step": 37641, "epoch": 422, "lr": 9.543469747069128e-05}
{"train_loss": 0.31870803236961365, "global_step": 37642, "epoch": 422, "lr": 9.543445544909033e-05}
{"train_loss": 0.47209227085113525, "global_step": 37643, "epoch": 422, "lr": 9.543421342138128e-05}
{"train_loss": 0.25307294726371765, "global_step": 37644, "epoch": 422, "lr": 9.543397138756414e-05}
{"train_loss": 0.3158619999885559, "global_step": 37645, "epoch": 422, "lr": 9.543372934763897e-05}
{"train_loss": 0.33537533460708147, "global_step": 37646, "epoch": 422, "lr": 9.54334873016058e-05, "val_loss": 2.528928518295288}
{"train_loss": 0.2778846025466919, "global_step": 37647, "epoch": 423, "lr": 9.543324524946462e-05}
{"train_loss": 0.30425378680229187, "global_step": 37648, "epoch": 423, "lr": 9.543300319121551e-05}
{"train_loss": 0.3768715262413025, "global_step": 37649, "epoch": 423, "lr": 9.543276112685848e-05}
{"train_loss": 0.33981189131736755, "global_step": 37650, "epoch": 423, "lr": 9.543251905639359e-05}
{"train_loss": 0.3351864516735077, "global_step": 37651, "epoch": 423, "lr": 9.543227697982084e-05}
{"train_loss": 0.3905550539493561, "global_step": 37652, "epoch": 423, "lr": 9.543203489714025e-05}
{"train_loss": 0.40443363785743713, "global_step": 37653, "epoch": 423, "lr": 9.54317928083519e-05}
{"train_loss": 0.39634189009666443, "global_step": 37654, "epoch": 423, "lr": 9.543155071345581e-05}
{"train_loss": 0.3496326506137848, "global_step": 37655, "epoch": 423, "lr": 9.543130861245199e-05}
{"train_loss": 0.34720873832702637, "global_step": 37656, "epoch": 423, "lr": 9.543106650534049e-05}
{"train_loss": 0.25829654932022095, "global_step": 37657, "epoch": 423, "lr": 9.543082439212133e-05}
{"train_loss": 0.44213980436325073, "global_step": 37658, "epoch": 423, "lr": 9.543058227279453e-05}
{"train_loss": 0.1877709925174713, "global_step": 37659, "epoch": 423, "lr": 9.543034014736016e-05}
{"train_loss": 0.3349350690841675, "global_step": 37660, "epoch": 423, "lr": 9.543009801581824e-05}
{"train_loss": 0.2904665470123291, "global_step": 37661, "epoch": 423, "lr": 9.542985587816879e-05}
{"train_loss": 0.28217756748199463, "global_step": 37662, "epoch": 423, "lr": 9.542961373441185e-05}
{"train_loss": 0.3761434853076935, "global_step": 37663, "epoch": 423, "lr": 9.542937158454746e-05}
{"train_loss": 0.5558802485466003, "global_step": 37664, "epoch": 423, "lr": 9.542912942857564e-05}
{"train_loss": 0.3354113698005676, "global_step": 37665, "epoch": 423, "lr": 9.542888726649642e-05}
{"train_loss": 0.32616716623306274, "global_step": 37666, "epoch": 423, "lr": 9.542864509830985e-05}
{"train_loss": 0.36418938636779785, "global_step": 37667, "epoch": 423, "lr": 9.542840292401597e-05}
{"train_loss": 0.1882132887840271, "global_step": 37668, "epoch": 423, "lr": 9.542816074361476e-05}
{"train_loss": 0.35884398221969604, "global_step": 37669, "epoch": 423, "lr": 9.54279185571063e-05}
{"train_loss": 0.4054173231124878, "global_step": 37670, "epoch": 423, "lr": 9.542767636449061e-05}
{"train_loss": 0.30736687779426575, "global_step": 37671, "epoch": 423, "lr": 9.542743416576772e-05}
{"train_loss": 0.3319498896598816, "global_step": 37672, "epoch": 423, "lr": 9.542719196093767e-05}
{"train_loss": 0.34683388471603394, "global_step": 37673, "epoch": 423, "lr": 9.542694975000048e-05}
{"train_loss": 0.33300918340682983, "global_step": 37674, "epoch": 423, "lr": 9.542670753295619e-05}
{"train_loss": 0.2578355669975281, "global_step": 37675, "epoch": 423, "lr": 9.542646530980485e-05}
{"train_loss": 0.3373672366142273, "global_step": 37676, "epoch": 423, "lr": 9.542622308054647e-05}
{"train_loss": 0.24571725726127625, "global_step": 37677, "epoch": 423, "lr": 9.542598084518108e-05}
{"train_loss": 0.26138392090797424, "global_step": 37678, "epoch": 423, "lr": 9.542573860370873e-05}
{"train_loss": 0.33970963954925537, "global_step": 37679, "epoch": 423, "lr": 9.542549635612943e-05}
{"train_loss": 0.34341782331466675, "global_step": 37680, "epoch": 423, "lr": 9.542525410244323e-05}
{"train_loss": 0.2943105101585388, "global_step": 37681, "epoch": 423, "lr": 9.542501184265015e-05}
{"train_loss": 0.37928032875061035, "global_step": 37682, "epoch": 423, "lr": 9.542476957675025e-05}
{"train_loss": 0.29611676931381226, "global_step": 37683, "epoch": 423, "lr": 9.542452730474354e-05}
{"train_loss": 0.3068619966506958, "global_step": 37684, "epoch": 423, "lr": 9.542428502663005e-05}
{"train_loss": 0.2858961522579193, "global_step": 37685, "epoch": 423, "lr": 9.542404274240982e-05}
{"train_loss": 0.40740716457366943, "global_step": 37686, "epoch": 423, "lr": 9.542380045208288e-05}
{"train_loss": 0.2820095419883728, "global_step": 37687, "epoch": 423, "lr": 9.542355815564926e-05}
{"train_loss": 0.3702796399593353, "global_step": 37688, "epoch": 423, "lr": 9.542331585310899e-05}
{"train_loss": 0.3364392817020416, "global_step": 37689, "epoch": 423, "lr": 9.542307354446213e-05}
{"train_loss": 0.3368228077888489, "global_step": 37690, "epoch": 423, "lr": 9.542283122970868e-05}
{"train_loss": 0.35457468032836914, "global_step": 37691, "epoch": 423, "lr": 9.542258890884867e-05}
{"train_loss": 0.28943806886672974, "global_step": 37692, "epoch": 423, "lr": 9.542234658188218e-05}
{"train_loss": 0.33482322096824646, "global_step": 37693, "epoch": 423, "lr": 9.542210424880918e-05}
{"train_loss": 0.3535628914833069, "global_step": 37694, "epoch": 423, "lr": 9.542186190962973e-05}
{"train_loss": 0.2842250168323517, "global_step": 37695, "epoch": 423, "lr": 9.54216195643439e-05}
{"train_loss": 0.3839418590068817, "global_step": 37696, "epoch": 423, "lr": 9.542137721295166e-05}
{"train_loss": 0.36544251441955566, "global_step": 37697, "epoch": 423, "lr": 9.542113485545308e-05}
{"train_loss": 0.29824164509773254, "global_step": 37698, "epoch": 423, "lr": 9.542089249184816e-05}
{"train_loss": 0.35273903608322144, "global_step": 37699, "epoch": 423, "lr": 9.542065012213697e-05}
{"train_loss": 0.26247474551200867, "global_step": 37700, "epoch": 423, "lr": 9.542040774631953e-05}
{"train_loss": 0.31537845730781555, "global_step": 37701, "epoch": 423, "lr": 9.542016536439586e-05}
{"train_loss": 0.2668552100658417, "global_step": 37702, "epoch": 423, "lr": 9.541992297636602e-05}
{"train_loss": 0.25348031520843506, "global_step": 37703, "epoch": 423, "lr": 9.541968058223e-05}
{"train_loss": 0.33936065435409546, "global_step": 37704, "epoch": 423, "lr": 9.541943818198788e-05}
{"train_loss": 0.2698371112346649, "global_step": 37705, "epoch": 423, "lr": 9.541919577563965e-05}
{"train_loss": 0.3754670321941376, "global_step": 37706, "epoch": 423, "lr": 9.541895336318538e-05}
{"train_loss": 0.3832527697086334, "global_step": 37707, "epoch": 423, "lr": 9.541871094462508e-05}
{"train_loss": 0.3396146297454834, "global_step": 37708, "epoch": 423, "lr": 9.541846851995878e-05}
{"train_loss": 0.26060715317726135, "global_step": 37709, "epoch": 423, "lr": 9.541822608918653e-05}
{"train_loss": 0.3774715065956116, "global_step": 37710, "epoch": 423, "lr": 9.541798365230835e-05}
{"train_loss": 0.3103076219558716, "global_step": 37711, "epoch": 423, "lr": 9.541774120932428e-05}
{"train_loss": 0.31376364827156067, "global_step": 37712, "epoch": 423, "lr": 9.541749876023434e-05}
{"train_loss": 0.38836604356765747, "global_step": 37713, "epoch": 423, "lr": 9.541725630503857e-05}
{"train_loss": 0.31935280561447144, "global_step": 37714, "epoch": 423, "lr": 9.541701384373701e-05}
{"train_loss": 0.34733250737190247, "global_step": 37715, "epoch": 423, "lr": 9.54167713763297e-05}
{"train_loss": 0.37509021162986755, "global_step": 37716, "epoch": 423, "lr": 9.541652890281663e-05}
{"train_loss": 0.2927711606025696, "global_step": 37717, "epoch": 423, "lr": 9.541628642319787e-05}
{"train_loss": 0.30319610238075256, "global_step": 37718, "epoch": 423, "lr": 9.541604393747346e-05}
{"train_loss": 0.2905120253562927, "global_step": 37719, "epoch": 423, "lr": 9.541580144564339e-05}
{"train_loss": 0.23729853332042694, "global_step": 37720, "epoch": 423, "lr": 9.541555894770775e-05}
{"train_loss": 0.2608027756214142, "global_step": 37721, "epoch": 423, "lr": 9.541531644366652e-05}
{"train_loss": 0.3285742402076721, "global_step": 37722, "epoch": 423, "lr": 9.541507393351975e-05}
{"train_loss": 0.28074297308921814, "global_step": 37723, "epoch": 423, "lr": 9.541483141726749e-05}
{"train_loss": 0.3343657851219177, "global_step": 37724, "epoch": 423, "lr": 9.541458889490976e-05}
{"train_loss": 0.3492574393749237, "global_step": 37725, "epoch": 423, "lr": 9.54143463664466e-05}
{"train_loss": 0.30226922035217285, "global_step": 37726, "epoch": 423, "lr": 9.541410383187801e-05}
{"train_loss": 0.3926903009414673, "global_step": 37727, "epoch": 423, "lr": 9.541386129120406e-05}
{"train_loss": 0.27448394894599915, "global_step": 37728, "epoch": 423, "lr": 9.541361874442477e-05}
{"train_loss": 0.28866928815841675, "global_step": 37729, "epoch": 423, "lr": 9.541337619154017e-05}
{"train_loss": 0.30056828260421753, "global_step": 37730, "epoch": 423, "lr": 9.54131336325503e-05}
{"train_loss": 0.3608548641204834, "global_step": 37731, "epoch": 423, "lr": 9.541289106745519e-05}
{"train_loss": 0.33586177229881287, "global_step": 37732, "epoch": 423, "lr": 9.541264849625487e-05}
{"train_loss": 0.2837831377983093, "global_step": 37733, "epoch": 423, "lr": 9.541240591894936e-05}
{"train_loss": 0.32746830582618713, "global_step": 37734, "epoch": 423, "lr": 9.541216333553871e-05}
{"train_loss": 0.325069027502885, "global_step": 37735, "epoch": 423, "lr": 9.541192074602295e-05, "val_loss": 2.5749027729034424}
{"train_loss": 0.34085822105407715, "global_step": 37736, "epoch": 424, "lr": 9.541167815040211e-05}
{"train_loss": 0.36763718724250793, "global_step": 37737, "epoch": 424, "lr": 9.541143554867623e-05}
{"train_loss": 0.3442358672618866, "global_step": 37738, "epoch": 424, "lr": 9.541119294084533e-05}
{"train_loss": 0.37411367893218994, "global_step": 37739, "epoch": 424, "lr": 9.541095032690946e-05}
{"train_loss": 0.2808986306190491, "global_step": 37740, "epoch": 424, "lr": 9.541070770686862e-05}
{"train_loss": 0.32582733035087585, "global_step": 37741, "epoch": 424, "lr": 9.541046508072289e-05}
{"train_loss": 0.31819674372673035, "global_step": 37742, "epoch": 424, "lr": 9.541022244847225e-05}
{"train_loss": 0.2925352156162262, "global_step": 37743, "epoch": 424, "lr": 9.540997981011677e-05}
{"train_loss": 0.25369012355804443, "global_step": 37744, "epoch": 424, "lr": 9.540973716565647e-05}
{"train_loss": 0.3000568449497223, "global_step": 37745, "epoch": 424, "lr": 9.540949451509139e-05}
{"train_loss": 0.33900269865989685, "global_step": 37746, "epoch": 424, "lr": 9.540925185842155e-05}
{"train_loss": 0.3418484032154083, "global_step": 37747, "epoch": 424, "lr": 9.540900919564699e-05}
{"train_loss": 0.2833448350429535, "global_step": 37748, "epoch": 424, "lr": 9.540876652676775e-05}
{"train_loss": 0.25356221199035645, "global_step": 37749, "epoch": 424, "lr": 9.540852385178384e-05}
{"train_loss": 0.23641762137413025, "global_step": 37750, "epoch": 424, "lr": 9.540828117069533e-05}
{"train_loss": 0.35608041286468506, "global_step": 37751, "epoch": 424, "lr": 9.540803848350221e-05}
{"train_loss": 0.3318277597427368, "global_step": 37752, "epoch": 424, "lr": 9.540779579020456e-05}
{"train_loss": 0.27172279357910156, "global_step": 37753, "epoch": 424, "lr": 9.540755309080235e-05}
{"train_loss": 0.35831302404403687, "global_step": 37754, "epoch": 424, "lr": 9.540731038529566e-05}
{"train_loss": 0.254228800535202, "global_step": 37755, "epoch": 424, "lr": 9.540706767368453e-05}
{"train_loss": 0.2748967409133911, "global_step": 37756, "epoch": 424, "lr": 9.540682495596895e-05}
{"train_loss": 0.31883132457733154, "global_step": 37757, "epoch": 424, "lr": 9.540658223214898e-05}
{"train_loss": 0.3705388307571411, "global_step": 37758, "epoch": 424, "lr": 9.540633950222466e-05}
{"train_loss": 0.2923724949359894, "global_step": 37759, "epoch": 424, "lr": 9.540609676619601e-05}
{"train_loss": 0.3520675599575043, "global_step": 37760, "epoch": 424, "lr": 9.540585402406304e-05}
{"train_loss": 0.35416746139526367, "global_step": 37761, "epoch": 424, "lr": 9.540561127582584e-05}
{"train_loss": 0.2550737261772156, "global_step": 37762, "epoch": 424, "lr": 9.540536852148438e-05}
{"train_loss": 0.3992891311645508, "global_step": 37763, "epoch": 424, "lr": 9.540512576103873e-05}
{"train_loss": 0.3536859452724457, "global_step": 37764, "epoch": 424, "lr": 9.540488299448892e-05}
{"train_loss": 0.27904942631721497, "global_step": 37765, "epoch": 424, "lr": 9.540464022183498e-05}
{"train_loss": 0.40133580565452576, "global_step": 37766, "epoch": 424, "lr": 9.540439744307693e-05}
{"train_loss": 0.39947593212127686, "global_step": 37767, "epoch": 424, "lr": 9.540415465821482e-05}
{"train_loss": 0.2749820351600647, "global_step": 37768, "epoch": 424, "lr": 9.540391186724867e-05}
{"train_loss": 0.371080607175827, "global_step": 37769, "epoch": 424, "lr": 9.540366907017853e-05}
{"train_loss": 0.2831118404865265, "global_step": 37770, "epoch": 424, "lr": 9.540342626700441e-05}
{"train_loss": 0.2835851311683655, "global_step": 37771, "epoch": 424, "lr": 9.540318345772634e-05}
{"train_loss": 0.2734268009662628, "global_step": 37772, "epoch": 424, "lr": 9.540294064234438e-05}
{"train_loss": 0.31438973546028137, "global_step": 37773, "epoch": 424, "lr": 9.540269782085853e-05}
{"train_loss": 0.32869353890419006, "global_step": 37774, "epoch": 424, "lr": 9.540245499326887e-05}
{"train_loss": 0.3549657464027405, "global_step": 37775, "epoch": 424, "lr": 9.540221215957538e-05}
{"train_loss": 0.28933805227279663, "global_step": 37776, "epoch": 424, "lr": 9.540196931977813e-05}
{"train_loss": 0.34911876916885376, "global_step": 37777, "epoch": 424, "lr": 9.540172647387713e-05}
{"train_loss": 0.2958473563194275, "global_step": 37778, "epoch": 424, "lr": 9.540148362187242e-05}
{"train_loss": 0.40627503395080566, "global_step": 37779, "epoch": 424, "lr": 9.540124076376405e-05}
{"train_loss": 0.4104624390602112, "global_step": 37780, "epoch": 424, "lr": 9.540099789955201e-05}
{"train_loss": 0.27735307812690735, "global_step": 37781, "epoch": 424, "lr": 9.540075502923639e-05}
{"train_loss": 0.4636435806751251, "global_step": 37782, "epoch": 424, "lr": 9.540051215281717e-05}
{"train_loss": 0.2911239564418793, "global_step": 37783, "epoch": 424, "lr": 9.540026927029442e-05}
{"train_loss": 0.32903897762298584, "global_step": 37784, "epoch": 424, "lr": 9.540002638166815e-05}
{"train_loss": 0.3649095296859741, "global_step": 37785, "epoch": 424, "lr": 9.539978348693839e-05}
{"train_loss": 0.40285152196884155, "global_step": 37786, "epoch": 424, "lr": 9.539954058610519e-05}
{"train_loss": 0.2957974374294281, "global_step": 37787, "epoch": 424, "lr": 9.539929767916857e-05}
{"train_loss": 0.4829787611961365, "global_step": 37788, "epoch": 424, "lr": 9.539905476612858e-05}
{"train_loss": 0.34183764457702637, "global_step": 37789, "epoch": 424, "lr": 9.539881184698522e-05}
{"train_loss": 0.3768661916255951, "global_step": 37790, "epoch": 424, "lr": 9.539856892173856e-05}
{"train_loss": 0.37168022990226746, "global_step": 37791, "epoch": 424, "lr": 9.539832599038861e-05}
{"train_loss": 0.279153972864151, "global_step": 37792, "epoch": 424, "lr": 9.539808305293542e-05}
{"train_loss": 0.2387162446975708, "global_step": 37793, "epoch": 424, "lr": 9.5397840109379e-05}
{"train_loss": 0.31780797243118286, "global_step": 37794, "epoch": 424, "lr": 9.539759715971939e-05}
{"train_loss": 0.3005155920982361, "global_step": 37795, "epoch": 424, "lr": 9.539735420395663e-05}
{"train_loss": 0.38699471950531006, "global_step": 37796, "epoch": 424, "lr": 9.539711124209077e-05}
{"train_loss": 0.25759851932525635, "global_step": 37797, "epoch": 424, "lr": 9.539686827412178e-05}
{"train_loss": 0.28843870759010315, "global_step": 37798, "epoch": 424, "lr": 9.539662530004976e-05}
{"train_loss": 0.2592170536518097, "global_step": 37799, "epoch": 424, "lr": 9.539638231987473e-05}
{"train_loss": 0.28421178460121155, "global_step": 37800, "epoch": 424, "lr": 9.539613933359668e-05}
{"train_loss": 0.3419452905654907, "global_step": 37801, "epoch": 424, "lr": 9.539589634121569e-05}
{"train_loss": 0.39639899134635925, "global_step": 37802, "epoch": 424, "lr": 9.539565334273177e-05}
{"train_loss": 0.3967148959636688, "global_step": 37803, "epoch": 424, "lr": 9.539541033814496e-05}
{"train_loss": 0.29541879892349243, "global_step": 37804, "epoch": 424, "lr": 9.539516732745528e-05}
{"train_loss": 0.4484984874725342, "global_step": 37805, "epoch": 424, "lr": 9.539492431066279e-05}
{"train_loss": 0.22863946855068207, "global_step": 37806, "epoch": 424, "lr": 9.53946812877675e-05}
{"train_loss": 0.3133779466152191, "global_step": 37807, "epoch": 424, "lr": 9.539443825876943e-05}
{"train_loss": 0.35105860233306885, "global_step": 37808, "epoch": 424, "lr": 9.539419522366865e-05}
{"train_loss": 0.2678237855434418, "global_step": 37809, "epoch": 424, "lr": 9.539395218246519e-05}
{"train_loss": 0.40029987692832947, "global_step": 37810, "epoch": 424, "lr": 9.539370913515904e-05}
{"train_loss": 0.2828855514526367, "global_step": 37811, "epoch": 424, "lr": 9.539346608175026e-05}
{"train_loss": 0.23339949548244476, "global_step": 37812, "epoch": 424, "lr": 9.53932230222389e-05}
{"train_loss": 0.3412618041038513, "global_step": 37813, "epoch": 424, "lr": 9.539297995662495e-05}
{"train_loss": 0.37503311038017273, "global_step": 37814, "epoch": 424, "lr": 9.539273688490849e-05}
{"train_loss": 0.2837883234024048, "global_step": 37815, "epoch": 424, "lr": 9.539249380708951e-05}
{"train_loss": 0.37190690636634827, "global_step": 37816, "epoch": 424, "lr": 9.539225072316808e-05}
{"train_loss": 0.3024356961250305, "global_step": 37817, "epoch": 424, "lr": 9.53920076331442e-05}
{"train_loss": 0.3293749690055847, "global_step": 37818, "epoch": 424, "lr": 9.539176453701792e-05}
{"train_loss": 0.2651565968990326, "global_step": 37819, "epoch": 424, "lr": 9.539152143478929e-05}
{"train_loss": 0.2874454855918884, "global_step": 37820, "epoch": 424, "lr": 9.539127832645831e-05}
{"train_loss": 0.35457849502563477, "global_step": 37821, "epoch": 424, "lr": 9.539103521202501e-05}
{"train_loss": 0.29769787192344666, "global_step": 37822, "epoch": 424, "lr": 9.539079209148946e-05}
{"train_loss": 0.3157186210155487, "global_step": 37823, "epoch": 424, "lr": 9.539054896485166e-05}
{"train_loss": 0.3251603804947285, "global_step": 37824, "epoch": 424, "lr": 9.539030583211166e-05, "val_loss": 2.704519033432007}
{"train_loss": 0.28046417236328125, "global_step": 37825, "epoch": 425, "lr": 9.539006269326949e-05}
{"train_loss": 0.30449894070625305, "global_step": 37826, "epoch": 425, "lr": 9.538981954832518e-05}
{"train_loss": 0.21469545364379883, "global_step": 37827, "epoch": 425, "lr": 9.538957639727875e-05}
{"train_loss": 0.3093131482601166, "global_step": 37828, "epoch": 425, "lr": 9.538933324013024e-05}
{"train_loss": 0.36851197481155396, "global_step": 37829, "epoch": 425, "lr": 9.53890900768797e-05}
{"train_loss": 0.3404030501842499, "global_step": 37830, "epoch": 425, "lr": 9.538884690752717e-05}
{"train_loss": 0.2602321207523346, "global_step": 37831, "epoch": 425, "lr": 9.538860373207264e-05}
{"train_loss": 0.26623064279556274, "global_step": 37832, "epoch": 425, "lr": 9.538836055051615e-05}
{"train_loss": 0.35324016213417053, "global_step": 37833, "epoch": 425, "lr": 9.538811736285777e-05}
{"train_loss": 0.40183302760124207, "global_step": 37834, "epoch": 425, "lr": 9.53878741690975e-05}
{"train_loss": 0.3989766836166382, "global_step": 37835, "epoch": 425, "lr": 9.53876309692354e-05}
{"train_loss": 0.4480134844779968, "global_step": 37836, "epoch": 425, "lr": 9.538738776327146e-05}
{"train_loss": 0.30892059206962585, "global_step": 37837, "epoch": 425, "lr": 9.538714455120576e-05}
{"train_loss": 0.38863441348075867, "global_step": 37838, "epoch": 425, "lr": 9.53869013330383e-05}
{"train_loss": 0.3453286290168762, "global_step": 37839, "epoch": 425, "lr": 9.538665810876914e-05}
{"train_loss": 0.373526394367218, "global_step": 37840, "epoch": 425, "lr": 9.538641487839828e-05}
{"train_loss": 0.4279824197292328, "global_step": 37841, "epoch": 425, "lr": 9.538617164192578e-05}
{"train_loss": 0.32654616236686707, "global_step": 37842, "epoch": 425, "lr": 9.538592839935165e-05}
{"train_loss": 0.3625733554363251, "global_step": 37843, "epoch": 425, "lr": 9.538568515067595e-05}
{"train_loss": 0.21138428151607513, "global_step": 37844, "epoch": 425, "lr": 9.538544189589868e-05}
{"train_loss": 0.33388620615005493, "global_step": 37845, "epoch": 425, "lr": 9.538519863501991e-05}
{"train_loss": 0.3852713704109192, "global_step": 37846, "epoch": 425, "lr": 9.538495536803965e-05}
{"train_loss": 0.3343126177787781, "global_step": 37847, "epoch": 425, "lr": 9.538471209495794e-05}
{"train_loss": 0.4529227018356323, "global_step": 37848, "epoch": 425, "lr": 9.538446881577478e-05}
{"train_loss": 0.4021557569503784, "global_step": 37849, "epoch": 425, "lr": 9.538422553049027e-05}
{"train_loss": 0.2604791820049286, "global_step": 37850, "epoch": 425, "lr": 9.538398223910439e-05}
{"train_loss": 0.34836849570274353, "global_step": 37851, "epoch": 425, "lr": 9.538373894161717e-05}
{"train_loss": 0.2754182815551758, "global_step": 37852, "epoch": 425, "lr": 9.538349563802867e-05}
{"train_loss": 0.3226933777332306, "global_step": 37853, "epoch": 425, "lr": 9.53832523283389e-05}
{"train_loss": 0.1778930127620697, "global_step": 37854, "epoch": 425, "lr": 9.538300901254793e-05}
{"train_loss": 0.2848917841911316, "global_step": 37855, "epoch": 425, "lr": 9.538276569065577e-05}
{"train_loss": 0.42855367064476013, "global_step": 37856, "epoch": 425, "lr": 9.538252236266243e-05}
{"train_loss": 0.4365795850753784, "global_step": 37857, "epoch": 425, "lr": 9.538227902856798e-05}
{"train_loss": 0.3436953127384186, "global_step": 37858, "epoch": 425, "lr": 9.538203568837241e-05}
{"train_loss": 0.27149829268455505, "global_step": 37859, "epoch": 425, "lr": 9.538179234207581e-05}
{"train_loss": 0.26036685705184937, "global_step": 37860, "epoch": 425, "lr": 9.538154898967815e-05}
{"train_loss": 0.4244758188724518, "global_step": 37861, "epoch": 425, "lr": 9.53813056311795e-05}
{"train_loss": 0.3691727817058563, "global_step": 37862, "epoch": 425, "lr": 9.53810622665799e-05}
{"train_loss": 0.27868276834487915, "global_step": 37863, "epoch": 425, "lr": 9.538081889587938e-05}
{"train_loss": 0.3089897036552429, "global_step": 37864, "epoch": 425, "lr": 9.538057551907794e-05}
{"train_loss": 0.36818060278892517, "global_step": 37865, "epoch": 425, "lr": 9.538033213617563e-05}
{"train_loss": 0.3244096040725708, "global_step": 37866, "epoch": 425, "lr": 9.53800887471725e-05}
{"train_loss": 0.29062679409980774, "global_step": 37867, "epoch": 425, "lr": 9.537984535206859e-05}
{"train_loss": 0.31920549273490906, "global_step": 37868, "epoch": 425, "lr": 9.537960195086388e-05}
{"train_loss": 0.4328292906284332, "global_step": 37869, "epoch": 425, "lr": 9.537935854355845e-05}
{"train_loss": 0.3467889726161957, "global_step": 37870, "epoch": 425, "lr": 9.537911513015232e-05}
{"train_loss": 0.33288878202438354, "global_step": 37871, "epoch": 425, "lr": 9.537887171064551e-05}
{"train_loss": 0.3228481709957123, "global_step": 37872, "epoch": 425, "lr": 9.537862828503807e-05}
{"train_loss": 0.30339208245277405, "global_step": 37873, "epoch": 425, "lr": 9.537838485333002e-05}
{"train_loss": 0.3192794620990753, "global_step": 37874, "epoch": 425, "lr": 9.537814141552142e-05}
{"train_loss": 0.28591978549957275, "global_step": 37875, "epoch": 425, "lr": 9.537789797161226e-05}
{"train_loss": 0.32797765731811523, "global_step": 37876, "epoch": 425, "lr": 9.537765452160261e-05}
{"train_loss": 0.3437807857990265, "global_step": 37877, "epoch": 425, "lr": 9.537741106549247e-05}
{"train_loss": 0.3583792448043823, "global_step": 37878, "epoch": 425, "lr": 9.53771676032819e-05}
{"train_loss": 0.30984634160995483, "global_step": 37879, "epoch": 425, "lr": 9.537692413497093e-05}
{"train_loss": 0.42989581823349, "global_step": 37880, "epoch": 425, "lr": 9.537668066055958e-05}
{"train_loss": 0.36179861426353455, "global_step": 37881, "epoch": 425, "lr": 9.537643718004789e-05}
{"train_loss": 0.32084760069847107, "global_step": 37882, "epoch": 425, "lr": 9.537619369343589e-05}
{"train_loss": 0.3220328390598297, "global_step": 37883, "epoch": 425, "lr": 9.53759502007236e-05}
{"train_loss": 0.2814135253429413, "global_step": 37884, "epoch": 425, "lr": 9.537570670191109e-05}
{"train_loss": 0.39110463857650757, "global_step": 37885, "epoch": 425, "lr": 9.537546319699835e-05}
{"train_loss": 0.36298084259033203, "global_step": 37886, "epoch": 425, "lr": 9.537521968598544e-05}
{"train_loss": 0.25582489371299744, "global_step": 37887, "epoch": 425, "lr": 9.537497616887239e-05}
{"train_loss": 0.3346710801124573, "global_step": 37888, "epoch": 425, "lr": 9.537473264565921e-05}
{"train_loss": 0.2749939560890198, "global_step": 37889, "epoch": 425, "lr": 9.537448911634598e-05}
{"train_loss": 0.3437238931655884, "global_step": 37890, "epoch": 425, "lr": 9.537424558093269e-05}
{"train_loss": 0.3156417906284332, "global_step": 37891, "epoch": 425, "lr": 9.537400203941937e-05}
{"train_loss": 0.3763069212436676, "global_step": 37892, "epoch": 425, "lr": 9.537375849180608e-05}
{"train_loss": 0.3810209035873413, "global_step": 37893, "epoch": 425, "lr": 9.537351493809284e-05}
{"train_loss": 0.40156593918800354, "global_step": 37894, "epoch": 425, "lr": 9.53732713782797e-05}
{"train_loss": 0.3184988498687744, "global_step": 37895, "epoch": 425, "lr": 9.537302781236665e-05}
{"train_loss": 0.25374260544776917, "global_step": 37896, "epoch": 425, "lr": 9.537278424035376e-05}
{"train_loss": 0.30631986260414124, "global_step": 37897, "epoch": 425, "lr": 9.537254066224106e-05}
{"train_loss": 0.27531376481056213, "global_step": 37898, "epoch": 425, "lr": 9.537229707802857e-05}
{"train_loss": 0.2457181215286255, "global_step": 37899, "epoch": 425, "lr": 9.537205348771631e-05}
{"train_loss": 0.3412950932979584, "global_step": 37900, "epoch": 425, "lr": 9.537180989130436e-05}
{"train_loss": 0.297247976064682, "global_step": 37901, "epoch": 425, "lr": 9.537156628879272e-05}
{"train_loss": 0.31017041206359863, "global_step": 37902, "epoch": 425, "lr": 9.537132268018141e-05}
{"train_loss": 0.3300662636756897, "global_step": 37903, "epoch": 425, "lr": 9.537107906547047e-05}
{"train_loss": 0.37245458364486694, "global_step": 37904, "epoch": 425, "lr": 9.537083544465996e-05}
{"train_loss": 0.2843649983406067, "global_step": 37905, "epoch": 425, "lr": 9.53705918177499e-05}
{"train_loss": 0.5248900055885315, "global_step": 37906, "epoch": 425, "lr": 9.53703481847403e-05}
{"train_loss": 0.3436128795146942, "global_step": 37907, "epoch": 425, "lr": 9.537010454563122e-05}
{"train_loss": 0.27438420057296753, "global_step": 37908, "epoch": 425, "lr": 9.536986090042267e-05}
{"train_loss": 0.42997246980667114, "global_step": 37909, "epoch": 425, "lr": 9.536961724911471e-05}
{"train_loss": 0.4266379773616791, "global_step": 37910, "epoch": 425, "lr": 9.536937359170735e-05}
{"train_loss": 0.2995104193687439, "global_step": 37911, "epoch": 425, "lr": 9.536912992820065e-05}
{"train_loss": 0.3910061717033386, "global_step": 37912, "epoch": 425, "lr": 9.536888625859459e-05}
{"train_loss": 0.33705514016446103, "global_step": 37913, "epoch": 425, "lr": 9.536864258288924e-05, "val_loss": 2.6591649055480957, "train_action_mse_error": 25.087539672851562}
{"train_loss": 0.5127078890800476, "global_step": 37914, "epoch": 426, "lr": 9.536839890108465e-05}
{"train_loss": 0.29613178968429565, "global_step": 37915, "epoch": 426, "lr": 9.536815521318082e-05}
{"train_loss": 0.4806513488292694, "global_step": 37916, "epoch": 426, "lr": 9.53679115191778e-05}
{"train_loss": 0.4160776436328888, "global_step": 37917, "epoch": 426, "lr": 9.536766781907563e-05}
{"train_loss": 0.2909390330314636, "global_step": 37918, "epoch": 426, "lr": 9.53674241128743e-05}
{"train_loss": 0.36512354016304016, "global_step": 37919, "epoch": 426, "lr": 9.536718040057388e-05}
{"train_loss": 0.48181799054145813, "global_step": 37920, "epoch": 426, "lr": 9.536693668217442e-05}
{"train_loss": 0.2751539647579193, "global_step": 37921, "epoch": 426, "lr": 9.536669295767591e-05}
{"train_loss": 0.3445747196674347, "global_step": 37922, "epoch": 426, "lr": 9.53664492270784e-05}
{"train_loss": 0.31892383098602295, "global_step": 37923, "epoch": 426, "lr": 9.536620549038192e-05}
{"train_loss": 0.3026137053966522, "global_step": 37924, "epoch": 426, "lr": 9.536596174758652e-05}
{"train_loss": 0.3206838071346283, "global_step": 37925, "epoch": 426, "lr": 9.53657179986922e-05}
{"train_loss": 0.30131399631500244, "global_step": 37926, "epoch": 426, "lr": 9.536547424369903e-05}
{"train_loss": 0.34296494722366333, "global_step": 37927, "epoch": 426, "lr": 9.536523048260702e-05}
{"train_loss": 0.31670403480529785, "global_step": 37928, "epoch": 426, "lr": 9.53649867154162e-05}
{"train_loss": 0.28018221259117126, "global_step": 37929, "epoch": 426, "lr": 9.536474294212663e-05}
{"train_loss": 0.2970527112483978, "global_step": 37930, "epoch": 426, "lr": 9.536449916273831e-05}
{"train_loss": 0.3048391044139862, "global_step": 37931, "epoch": 426, "lr": 9.536425537725129e-05}
{"train_loss": 0.33986401557922363, "global_step": 37932, "epoch": 426, "lr": 9.536401158566559e-05}
{"train_loss": 0.27925586700439453, "global_step": 37933, "epoch": 426, "lr": 9.536376778798127e-05}
{"train_loss": 0.47840338945388794, "global_step": 37934, "epoch": 426, "lr": 9.536352398419832e-05}
{"train_loss": 0.32354116439819336, "global_step": 37935, "epoch": 426, "lr": 9.53632801743168e-05}
{"train_loss": 0.2995639741420746, "global_step": 37936, "epoch": 426, "lr": 9.536303635833675e-05}
{"train_loss": 0.25724372267723083, "global_step": 37937, "epoch": 426, "lr": 9.536279253625819e-05}
{"train_loss": 0.33499467372894287, "global_step": 37938, "epoch": 426, "lr": 9.536254870808117e-05}
{"train_loss": 0.4492420554161072, "global_step": 37939, "epoch": 426, "lr": 9.536230487380568e-05}
{"train_loss": 0.31703513860702515, "global_step": 37940, "epoch": 426, "lr": 9.53620610334318e-05}
{"train_loss": 0.4464226961135864, "global_step": 37941, "epoch": 426, "lr": 9.536181718695954e-05}
{"train_loss": 0.30823609232902527, "global_step": 37942, "epoch": 426, "lr": 9.536157333438893e-05}
{"train_loss": 0.2887585759162903, "global_step": 37943, "epoch": 426, "lr": 9.536132947572002e-05}
{"train_loss": 0.4650486409664154, "global_step": 37944, "epoch": 426, "lr": 9.536108561095283e-05}
{"train_loss": 0.3403760492801666, "global_step": 37945, "epoch": 426, "lr": 9.536084174008739e-05}
{"train_loss": 0.2745853364467621, "global_step": 37946, "epoch": 426, "lr": 9.536059786312375e-05}
{"train_loss": 0.2701047956943512, "global_step": 37947, "epoch": 426, "lr": 9.53603539800619e-05}
{"train_loss": 0.3843516707420349, "global_step": 37948, "epoch": 426, "lr": 9.536011009090194e-05}
{"train_loss": 0.3784196078777313, "global_step": 37949, "epoch": 426, "lr": 9.535986619564384e-05}
{"train_loss": 0.3506614565849304, "global_step": 37950, "epoch": 426, "lr": 9.535962229428768e-05}
{"train_loss": 0.4195041358470917, "global_step": 37951, "epoch": 426, "lr": 9.535937838683346e-05}
{"train_loss": 0.35569679737091064, "global_step": 37952, "epoch": 426, "lr": 9.535913447328122e-05}
{"train_loss": 0.19803623855113983, "global_step": 37953, "epoch": 426, "lr": 9.5358890553631e-05}
{"train_loss": 0.38249313831329346, "global_step": 37954, "epoch": 426, "lr": 9.535864662788283e-05}
{"train_loss": 0.34674516320228577, "global_step": 37955, "epoch": 426, "lr": 9.535840269603676e-05}
{"train_loss": 0.3742196261882782, "global_step": 37956, "epoch": 426, "lr": 9.535815875809278e-05}
{"train_loss": 0.2905671298503876, "global_step": 37957, "epoch": 426, "lr": 9.535791481405096e-05}
{"train_loss": 0.3469608724117279, "global_step": 37958, "epoch": 426, "lr": 9.535767086391131e-05}
{"train_loss": 0.2930727005004883, "global_step": 37959, "epoch": 426, "lr": 9.535742690767389e-05}
{"train_loss": 0.316053569316864, "global_step": 37960, "epoch": 426, "lr": 9.53571829453387e-05}
{"train_loss": 0.3468663990497589, "global_step": 37961, "epoch": 426, "lr": 9.53569389769058e-05}
{"train_loss": 0.34644779562950134, "global_step": 37962, "epoch": 426, "lr": 9.535669500237523e-05}
{"train_loss": 0.3454667627811432, "global_step": 37963, "epoch": 426, "lr": 9.535645102174698e-05}
{"train_loss": 0.3242284655570984, "global_step": 37964, "epoch": 426, "lr": 9.535620703502111e-05}
{"train_loss": 0.36334097385406494, "global_step": 37965, "epoch": 426, "lr": 9.535596304219768e-05}
{"train_loss": 0.3781394362449646, "global_step": 37966, "epoch": 426, "lr": 9.535571904327665e-05}
{"train_loss": 0.29405584931373596, "global_step": 37967, "epoch": 426, "lr": 9.535547503825813e-05}
{"train_loss": 0.29205626249313354, "global_step": 37968, "epoch": 426, "lr": 9.53552310271421e-05}
{"train_loss": 0.3827548027038574, "global_step": 37969, "epoch": 426, "lr": 9.535498700992862e-05}
{"train_loss": 0.3111262917518616, "global_step": 37970, "epoch": 426, "lr": 9.53547429866177e-05}
{"train_loss": 0.3478738069534302, "global_step": 37971, "epoch": 426, "lr": 9.535449895720941e-05}
{"train_loss": 0.36611464619636536, "global_step": 37972, "epoch": 426, "lr": 9.535425492170375e-05}
{"train_loss": 0.3122686743736267, "global_step": 37973, "epoch": 426, "lr": 9.535401088010076e-05}
{"train_loss": 0.3135327398777008, "global_step": 37974, "epoch": 426, "lr": 9.535376683240048e-05}
{"train_loss": 0.36721715331077576, "global_step": 37975, "epoch": 426, "lr": 9.535352277860294e-05}
{"train_loss": 0.27768102288246155, "global_step": 37976, "epoch": 426, "lr": 9.535327871870817e-05}
{"train_loss": 0.2282484918832779, "global_step": 37977, "epoch": 426, "lr": 9.535303465271621e-05}
{"train_loss": 0.3787095844745636, "global_step": 37978, "epoch": 426, "lr": 9.535279058062708e-05}
{"train_loss": 0.4505206048488617, "global_step": 37979, "epoch": 426, "lr": 9.535254650244083e-05}
{"train_loss": 0.3080248534679413, "global_step": 37980, "epoch": 426, "lr": 9.535230241815747e-05}
{"train_loss": 0.3872297406196594, "global_step": 37981, "epoch": 426, "lr": 9.535205832777705e-05}
{"train_loss": 0.3112321197986603, "global_step": 37982, "epoch": 426, "lr": 9.53518142312996e-05}
{"train_loss": 0.33111852407455444, "global_step": 37983, "epoch": 426, "lr": 9.535157012872516e-05}
{"train_loss": 0.2878628969192505, "global_step": 37984, "epoch": 426, "lr": 9.535132602005373e-05}
{"train_loss": 0.29019612073898315, "global_step": 37985, "epoch": 426, "lr": 9.535108190528539e-05}
{"train_loss": 0.3211395740509033, "global_step": 37986, "epoch": 426, "lr": 9.535083778442014e-05}
{"train_loss": 0.34560254216194153, "global_step": 37987, "epoch": 426, "lr": 9.535059365745802e-05}
{"train_loss": 0.37088578939437866, "global_step": 37988, "epoch": 426, "lr": 9.535034952439907e-05}
{"train_loss": 0.3834129869937897, "global_step": 37989, "epoch": 426, "lr": 9.535010538524331e-05}
{"train_loss": 0.3792250454425812, "global_step": 37990, "epoch": 426, "lr": 9.53498612399908e-05}
{"train_loss": 0.26289910078048706, "global_step": 37991, "epoch": 426, "lr": 9.534961708864154e-05}
{"train_loss": 0.471583753824234, "global_step": 37992, "epoch": 426, "lr": 9.534937293119558e-05}
{"train_loss": 0.33412936329841614, "global_step": 37993, "epoch": 426, "lr": 9.534912876765294e-05}
{"train_loss": 0.3080901503562927, "global_step": 37994, "epoch": 426, "lr": 9.534888459801368e-05}
{"train_loss": 0.3467220366001129, "global_step": 37995, "epoch": 426, "lr": 9.53486404222778e-05}
{"train_loss": 0.3540574014186859, "global_step": 37996, "epoch": 426, "lr": 9.534839624044536e-05}
{"train_loss": 0.24187199771404266, "global_step": 37997, "epoch": 426, "lr": 9.534815205251637e-05}
{"train_loss": 0.28208258748054504, "global_step": 37998, "epoch": 426, "lr": 9.534790785849088e-05}
{"train_loss": 0.1938951164484024, "global_step": 37999, "epoch": 426, "lr": 9.534766365836891e-05}
{"train_loss": 0.3205322027206421, "global_step": 38000, "epoch": 426, "lr": 9.534741945215051e-05}
{"train_loss": 0.33796384930610657, "global_step": 38001, "epoch": 426, "lr": 9.53471752398357e-05}
{"train_loss": 0.33890393208921626, "global_step": 38002, "epoch": 426, "lr": 9.534693102142451e-05, "val_loss": 2.5516505241394043}
{"train_loss": 0.38224950432777405, "global_step": 38003, "epoch": 427, "lr": 9.534668679691699e-05}
{"train_loss": 0.4596504867076874, "global_step": 38004, "epoch": 427, "lr": 9.534644256631314e-05}
{"train_loss": 0.3314497172832489, "global_step": 38005, "epoch": 427, "lr": 9.534619832961302e-05}
{"train_loss": 0.3311724066734314, "global_step": 38006, "epoch": 427, "lr": 9.534595408681667e-05}
{"train_loss": 0.32970690727233887, "global_step": 38007, "epoch": 427, "lr": 9.53457098379241e-05}
{"train_loss": 0.32099324464797974, "global_step": 38008, "epoch": 427, "lr": 9.534546558293535e-05}
{"train_loss": 0.37270838022232056, "global_step": 38009, "epoch": 427, "lr": 9.534522132185047e-05}
{"train_loss": 0.3318675458431244, "global_step": 38010, "epoch": 427, "lr": 9.534497705466944e-05}
{"train_loss": 0.3449985682964325, "global_step": 38011, "epoch": 427, "lr": 9.534473278139237e-05}
{"train_loss": 0.2695339620113373, "global_step": 38012, "epoch": 427, "lr": 9.534448850201923e-05}
{"train_loss": 0.3216853439807892, "global_step": 38013, "epoch": 427, "lr": 9.534424421655009e-05}
{"train_loss": 0.241025909781456, "global_step": 38014, "epoch": 427, "lr": 9.534399992498496e-05}
{"train_loss": 0.3715093433856964, "global_step": 38015, "epoch": 427, "lr": 9.534375562732388e-05}
{"train_loss": 0.2558595836162567, "global_step": 38016, "epoch": 427, "lr": 9.534351132356689e-05}
{"train_loss": 0.30883654952049255, "global_step": 38017, "epoch": 427, "lr": 9.534326701371402e-05}
{"train_loss": 0.44368863105773926, "global_step": 38018, "epoch": 427, "lr": 9.53430226977653e-05}
{"train_loss": 0.2900312542915344, "global_step": 38019, "epoch": 427, "lr": 9.534277837572077e-05}
{"train_loss": 0.3400612771511078, "global_step": 38020, "epoch": 427, "lr": 9.534253404758045e-05}
{"train_loss": 0.34745457768440247, "global_step": 38021, "epoch": 427, "lr": 9.534228971334436e-05}
{"train_loss": 0.3132695257663727, "global_step": 38022, "epoch": 427, "lr": 9.534204537301257e-05}
{"train_loss": 0.31809765100479126, "global_step": 38023, "epoch": 427, "lr": 9.53418010265851e-05}
{"train_loss": 0.270755797624588, "global_step": 38024, "epoch": 427, "lr": 9.534155667406197e-05}
{"train_loss": 0.33016452193260193, "global_step": 38025, "epoch": 427, "lr": 9.534131231544321e-05}
{"train_loss": 0.2751767635345459, "global_step": 38026, "epoch": 427, "lr": 9.534106795072887e-05}
{"train_loss": 0.32302558422088623, "global_step": 38027, "epoch": 427, "lr": 9.534082357991898e-05}
{"train_loss": 0.39417147636413574, "global_step": 38028, "epoch": 427, "lr": 9.534057920301359e-05}
{"train_loss": 0.40230491757392883, "global_step": 38029, "epoch": 427, "lr": 9.534033482001268e-05}
{"train_loss": 0.29299086332321167, "global_step": 38030, "epoch": 427, "lr": 9.534009043091631e-05}
{"train_loss": 0.3657313883304596, "global_step": 38031, "epoch": 427, "lr": 9.533984603572454e-05}
{"train_loss": 0.35550758242607117, "global_step": 38032, "epoch": 427, "lr": 9.533960163443738e-05}
{"train_loss": 0.3326379656791687, "global_step": 38033, "epoch": 427, "lr": 9.533935722705485e-05}
{"train_loss": 0.32881778478622437, "global_step": 38034, "epoch": 427, "lr": 9.5339112813577e-05}
{"train_loss": 0.3240851163864136, "global_step": 38035, "epoch": 427, "lr": 9.533886839400387e-05}
{"train_loss": 0.2887086570262909, "global_step": 38036, "epoch": 427, "lr": 9.533862396833546e-05}
{"train_loss": 0.2933114171028137, "global_step": 38037, "epoch": 427, "lr": 9.533837953657184e-05}
{"train_loss": 0.40973100066185, "global_step": 38038, "epoch": 427, "lr": 9.533813509871303e-05}
{"train_loss": 0.26205846667289734, "global_step": 38039, "epoch": 427, "lr": 9.533789065475905e-05}
{"train_loss": 0.31452038884162903, "global_step": 38040, "epoch": 427, "lr": 9.533764620470995e-05}
{"train_loss": 0.3178517818450928, "global_step": 38041, "epoch": 427, "lr": 9.533740174856576e-05}
{"train_loss": 0.3590593636035919, "global_step": 38042, "epoch": 427, "lr": 9.53371572863265e-05}
{"train_loss": 0.3060038685798645, "global_step": 38043, "epoch": 427, "lr": 9.533691281799222e-05}
{"train_loss": 0.2993934154510498, "global_step": 38044, "epoch": 427, "lr": 9.533666834356294e-05}
{"train_loss": 0.26950109004974365, "global_step": 38045, "epoch": 427, "lr": 9.53364238630387e-05}
{"train_loss": 0.25956588983535767, "global_step": 38046, "epoch": 427, "lr": 9.533617937641952e-05}
{"train_loss": 0.2995922565460205, "global_step": 38047, "epoch": 427, "lr": 9.533593488370545e-05}
{"train_loss": 0.29519447684288025, "global_step": 38048, "epoch": 427, "lr": 9.533569038489652e-05}
{"train_loss": 0.26749616861343384, "global_step": 38049, "epoch": 427, "lr": 9.533544587999275e-05}
{"train_loss": 0.3499820828437805, "global_step": 38050, "epoch": 427, "lr": 9.533520136899419e-05}
{"train_loss": 0.45558732748031616, "global_step": 38051, "epoch": 427, "lr": 9.533495685190086e-05}
{"train_loss": 0.34861835837364197, "global_step": 38052, "epoch": 427, "lr": 9.53347123287128e-05}
{"train_loss": 0.3688233196735382, "global_step": 38053, "epoch": 427, "lr": 9.533446779943004e-05}
{"train_loss": 0.27791205048561096, "global_step": 38054, "epoch": 427, "lr": 9.533422326405261e-05}
{"train_loss": 0.300686240196228, "global_step": 38055, "epoch": 427, "lr": 9.533397872258055e-05}
{"train_loss": 0.3809008300304413, "global_step": 38056, "epoch": 427, "lr": 9.53337341750139e-05}
{"train_loss": 0.32486122846603394, "global_step": 38057, "epoch": 427, "lr": 9.533348962135266e-05}
{"train_loss": 0.2695791721343994, "global_step": 38058, "epoch": 427, "lr": 9.53332450615969e-05}
{"train_loss": 0.22540324926376343, "global_step": 38059, "epoch": 427, "lr": 9.533300049574662e-05}
{"train_loss": 0.30950355529785156, "global_step": 38060, "epoch": 427, "lr": 9.53327559238019e-05}
{"train_loss": 0.3346976339817047, "global_step": 38061, "epoch": 427, "lr": 9.533251134576273e-05}
{"train_loss": 0.2993355989456177, "global_step": 38062, "epoch": 427, "lr": 9.533226676162916e-05}
{"train_loss": 0.35754284262657166, "global_step": 38063, "epoch": 427, "lr": 9.53320221714012e-05}
{"train_loss": 0.33265358209609985, "global_step": 38064, "epoch": 427, "lr": 9.533177757507892e-05}
{"train_loss": 0.3955513834953308, "global_step": 38065, "epoch": 427, "lr": 9.533153297266233e-05}
{"train_loss": 0.2548501789569855, "global_step": 38066, "epoch": 427, "lr": 9.533128836415146e-05}
{"train_loss": 0.30252212285995483, "global_step": 38067, "epoch": 427, "lr": 9.533104374954637e-05}
{"train_loss": 0.3081870973110199, "global_step": 38068, "epoch": 427, "lr": 9.533079912884706e-05}
{"train_loss": 0.3426579535007477, "global_step": 38069, "epoch": 427, "lr": 9.533055450205356e-05}
{"train_loss": 0.2358550876379013, "global_step": 38070, "epoch": 427, "lr": 9.533030986916595e-05}
{"train_loss": 0.40173593163490295, "global_step": 38071, "epoch": 427, "lr": 9.533006523018423e-05}
{"train_loss": 0.4147487282752991, "global_step": 38072, "epoch": 427, "lr": 9.532982058510842e-05}
{"train_loss": 0.24075287580490112, "global_step": 38073, "epoch": 427, "lr": 9.532957593393858e-05}
{"train_loss": 0.4286736845970154, "global_step": 38074, "epoch": 427, "lr": 9.532933127667473e-05}
{"train_loss": 0.231061190366745, "global_step": 38075, "epoch": 427, "lr": 9.53290866133169e-05}
{"train_loss": 0.2898409366607666, "global_step": 38076, "epoch": 427, "lr": 9.532884194386512e-05}
{"train_loss": 0.30625689029693604, "global_step": 38077, "epoch": 427, "lr": 9.532859726831944e-05}
{"train_loss": 0.3383297622203827, "global_step": 38078, "epoch": 427, "lr": 9.532835258667987e-05}
{"train_loss": 0.36234354972839355, "global_step": 38079, "epoch": 427, "lr": 9.532810789894646e-05}
{"train_loss": 0.49034300446510315, "global_step": 38080, "epoch": 427, "lr": 9.532786320511926e-05}
{"train_loss": 0.3568785786628723, "global_step": 38081, "epoch": 427, "lr": 9.532761850519826e-05}
{"train_loss": 0.269443154335022, "global_step": 38082, "epoch": 427, "lr": 9.532737379918352e-05}
{"train_loss": 0.3434850871562958, "global_step": 38083, "epoch": 427, "lr": 9.532712908707506e-05}
{"train_loss": 0.3428688943386078, "global_step": 38084, "epoch": 427, "lr": 9.532688436887292e-05}
{"train_loss": 0.45002856850624084, "global_step": 38085, "epoch": 427, "lr": 9.532663964457715e-05}
{"train_loss": 0.3808531165122986, "global_step": 38086, "epoch": 427, "lr": 9.532639491418774e-05}
{"train_loss": 0.33020535111427307, "global_step": 38087, "epoch": 427, "lr": 9.532615017770476e-05}
{"train_loss": 0.322868287563324, "global_step": 38088, "epoch": 427, "lr": 9.532590543512824e-05}
{"train_loss": 0.4191160202026367, "global_step": 38089, "epoch": 427, "lr": 9.532566068645819e-05}
{"train_loss": 0.3186427056789398, "global_step": 38090, "epoch": 427, "lr": 9.532541593169467e-05}
{"train_loss": 0.32946277334449, "global_step": 38091, "epoch": 427, "lr": 9.53251711708377e-05, "val_loss": 2.5756704807281494}
{"train_loss": 0.23404450714588165, "global_step": 38092, "epoch": 428, "lr": 9.53249264038873e-05}
{"train_loss": 0.38654470443725586, "global_step": 38093, "epoch": 428, "lr": 9.532468163084354e-05}
{"train_loss": 0.3871912956237793, "global_step": 38094, "epoch": 428, "lr": 9.532443685170642e-05}
{"train_loss": 0.4124452471733093, "global_step": 38095, "epoch": 428, "lr": 9.532419206647596e-05}
{"train_loss": 0.36303746700286865, "global_step": 38096, "epoch": 428, "lr": 9.532394727515223e-05}
{"train_loss": 0.431264191865921, "global_step": 38097, "epoch": 428, "lr": 9.532370247773525e-05}
{"train_loss": 0.33891555666923523, "global_step": 38098, "epoch": 428, "lr": 9.532345767422507e-05}
{"train_loss": 0.44359290599823, "global_step": 38099, "epoch": 428, "lr": 9.532321286462166e-05}
{"train_loss": 0.33499875664711, "global_step": 38100, "epoch": 428, "lr": 9.532296804892513e-05}
{"train_loss": 0.28478023409843445, "global_step": 38101, "epoch": 428, "lr": 9.532272322713547e-05}
{"train_loss": 0.4123298227787018, "global_step": 38102, "epoch": 428, "lr": 9.532247839925271e-05}
{"train_loss": 0.45413491129875183, "global_step": 38103, "epoch": 428, "lr": 9.532223356527691e-05}
{"train_loss": 0.3839047849178314, "global_step": 38104, "epoch": 428, "lr": 9.532198872520807e-05}
{"train_loss": 0.35877907276153564, "global_step": 38105, "epoch": 428, "lr": 9.532174387904627e-05}
{"train_loss": 0.2783719003200531, "global_step": 38106, "epoch": 428, "lr": 9.532149902679149e-05}
{"train_loss": 0.33586782217025757, "global_step": 38107, "epoch": 428, "lr": 9.53212541684438e-05}
{"train_loss": 0.31108540296554565, "global_step": 38108, "epoch": 428, "lr": 9.532100930400322e-05}
{"train_loss": 0.3402794599533081, "global_step": 38109, "epoch": 428, "lr": 9.532076443346977e-05}
{"train_loss": 0.3509179949760437, "global_step": 38110, "epoch": 428, "lr": 9.532051955684351e-05}
{"train_loss": 0.29041144251823425, "global_step": 38111, "epoch": 428, "lr": 9.532027467412446e-05}
{"train_loss": 0.33228734135627747, "global_step": 38112, "epoch": 428, "lr": 9.532002978531263e-05}
{"train_loss": 0.20025934278964996, "global_step": 38113, "epoch": 428, "lr": 9.531978489040809e-05}
{"train_loss": 0.2884863018989563, "global_step": 38114, "epoch": 428, "lr": 9.531953998941085e-05}
{"train_loss": 0.30093345046043396, "global_step": 38115, "epoch": 428, "lr": 9.531929508232096e-05}
{"train_loss": 0.39861053228378296, "global_step": 38116, "epoch": 428, "lr": 9.531905016913844e-05}
{"train_loss": 0.2875806391239166, "global_step": 38117, "epoch": 428, "lr": 9.531880524986331e-05}
{"train_loss": 0.29947856068611145, "global_step": 38118, "epoch": 428, "lr": 9.531856032449565e-05}
{"train_loss": 0.3049362003803253, "global_step": 38119, "epoch": 428, "lr": 9.531831539303544e-05}
{"train_loss": 0.2682493329048157, "global_step": 38120, "epoch": 428, "lr": 9.531807045548275e-05}
{"train_loss": 0.32080593705177307, "global_step": 38121, "epoch": 428, "lr": 9.53178255118376e-05}
{"train_loss": 0.2676602602005005, "global_step": 38122, "epoch": 428, "lr": 9.531758056209999e-05}
{"train_loss": 0.3130802512168884, "global_step": 38123, "epoch": 428, "lr": 9.531733560627001e-05}
{"train_loss": 0.34257814288139343, "global_step": 38124, "epoch": 428, "lr": 9.531709064434767e-05}
{"train_loss": 0.28398871421813965, "global_step": 38125, "epoch": 428, "lr": 9.531684567633298e-05}
{"train_loss": 0.35128581523895264, "global_step": 38126, "epoch": 428, "lr": 9.531660070222599e-05}
{"train_loss": 0.2322259545326233, "global_step": 38127, "epoch": 428, "lr": 9.531635572202675e-05}
{"train_loss": 0.3270041346549988, "global_step": 38128, "epoch": 428, "lr": 9.531611073573527e-05}
{"train_loss": 0.42878156900405884, "global_step": 38129, "epoch": 428, "lr": 9.53158657433516e-05}
{"train_loss": 0.39602383971214294, "global_step": 38130, "epoch": 428, "lr": 9.531562074487577e-05}
{"train_loss": 0.27824223041534424, "global_step": 38131, "epoch": 428, "lr": 9.53153757403078e-05}
{"train_loss": 0.3759208023548126, "global_step": 38132, "epoch": 428, "lr": 9.531513072964773e-05}
{"train_loss": 0.3054084777832031, "global_step": 38133, "epoch": 428, "lr": 9.531488571289558e-05}
{"train_loss": 0.28444206714630127, "global_step": 38134, "epoch": 428, "lr": 9.53146406900514e-05}
{"train_loss": 0.36981678009033203, "global_step": 38135, "epoch": 428, "lr": 9.531439566111523e-05}
{"train_loss": 0.3306781053543091, "global_step": 38136, "epoch": 428, "lr": 9.531415062608709e-05}
{"train_loss": 0.3974382281303406, "global_step": 38137, "epoch": 428, "lr": 9.531390558496701e-05}
{"train_loss": 0.29824554920196533, "global_step": 38138, "epoch": 428, "lr": 9.531366053775504e-05}
{"train_loss": 0.39045441150665283, "global_step": 38139, "epoch": 428, "lr": 9.531341548445118e-05}
{"train_loss": 0.3089190721511841, "global_step": 38140, "epoch": 428, "lr": 9.53131704250555e-05}
{"train_loss": 0.3440747559070587, "global_step": 38141, "epoch": 428, "lr": 9.531292535956801e-05}
{"train_loss": 0.4073772430419922, "global_step": 38142, "epoch": 428, "lr": 9.531268028798874e-05}
{"train_loss": 0.4497646391391754, "global_step": 38143, "epoch": 428, "lr": 9.531243521031775e-05}
{"train_loss": 0.4417075216770172, "global_step": 38144, "epoch": 428, "lr": 9.531219012655503e-05}
{"train_loss": 0.4410359859466553, "global_step": 38145, "epoch": 428, "lr": 9.531194503670066e-05}
{"train_loss": 0.27749699354171753, "global_step": 38146, "epoch": 428, "lr": 9.531169994075466e-05}
{"train_loss": 0.4640488922595978, "global_step": 38147, "epoch": 428, "lr": 9.531145483871705e-05}
{"train_loss": 0.31686559319496155, "global_step": 38148, "epoch": 428, "lr": 9.531120973058785e-05}
{"train_loss": 0.27999627590179443, "global_step": 38149, "epoch": 428, "lr": 9.531096461636711e-05}
{"train_loss": 0.30440273880958557, "global_step": 38150, "epoch": 428, "lr": 9.531071949605488e-05}
{"train_loss": 0.4053390622138977, "global_step": 38151, "epoch": 428, "lr": 9.531047436965117e-05}
{"train_loss": 0.4072636663913727, "global_step": 38152, "epoch": 428, "lr": 9.5310229237156e-05}
{"train_loss": 0.2596847712993622, "global_step": 38153, "epoch": 428, "lr": 9.530998409856944e-05}
{"train_loss": 0.3344557285308838, "global_step": 38154, "epoch": 428, "lr": 9.53097389538915e-05}
{"train_loss": 0.29386642575263977, "global_step": 38155, "epoch": 428, "lr": 9.530949380312224e-05}
{"train_loss": 0.2619704008102417, "global_step": 38156, "epoch": 428, "lr": 9.530924864626165e-05}
{"train_loss": 0.30057650804519653, "global_step": 38157, "epoch": 428, "lr": 9.530900348330978e-05}
{"train_loss": 0.2967173457145691, "global_step": 38158, "epoch": 428, "lr": 9.530875831426666e-05}
{"train_loss": 0.38895440101623535, "global_step": 38159, "epoch": 428, "lr": 9.530851313913236e-05}
{"train_loss": 0.25402122735977173, "global_step": 38160, "epoch": 428, "lr": 9.530826795790686e-05}
{"train_loss": 0.2365403175354004, "global_step": 38161, "epoch": 428, "lr": 9.530802277059023e-05}
{"train_loss": 0.3099226653575897, "global_step": 38162, "epoch": 428, "lr": 9.530777757718248e-05}
{"train_loss": 0.3467969000339508, "global_step": 38163, "epoch": 428, "lr": 9.530753237768365e-05}
{"train_loss": 0.32072076201438904, "global_step": 38164, "epoch": 428, "lr": 9.530728717209376e-05}
{"train_loss": 0.36580368876457214, "global_step": 38165, "epoch": 428, "lr": 9.530704196041287e-05}
{"train_loss": 0.35844823718070984, "global_step": 38166, "epoch": 428, "lr": 9.5306796742641e-05}
{"train_loss": 0.2517995834350586, "global_step": 38167, "epoch": 428, "lr": 9.53065515187782e-05}
{"train_loss": 0.3624577224254608, "global_step": 38168, "epoch": 428, "lr": 9.530630628882446e-05}
{"train_loss": 0.45997354388237, "global_step": 38169, "epoch": 428, "lr": 9.530606105277985e-05}
{"train_loss": 0.32146602869033813, "global_step": 38170, "epoch": 428, "lr": 9.53058158106444e-05}
{"train_loss": 0.2791512906551361, "global_step": 38171, "epoch": 428, "lr": 9.530557056241814e-05}
{"train_loss": 0.2562353014945984, "global_step": 38172, "epoch": 428, "lr": 9.530532530810108e-05}
{"train_loss": 0.27905037999153137, "global_step": 38173, "epoch": 428, "lr": 9.530508004769328e-05}
{"train_loss": 0.25397905707359314, "global_step": 38174, "epoch": 428, "lr": 9.530483478119477e-05}
{"train_loss": 0.3470480144023895, "global_step": 38175, "epoch": 428, "lr": 9.530458950860557e-05}
{"train_loss": 0.37342536449432373, "global_step": 38176, "epoch": 428, "lr": 9.530434422992571e-05}
{"train_loss": 0.350077748298645, "global_step": 38177, "epoch": 428, "lr": 9.530409894515525e-05}
{"train_loss": 0.23069635033607483, "global_step": 38178, "epoch": 428, "lr": 9.530385365429419e-05}
{"train_loss": 0.3429560661315918, "global_step": 38179, "epoch": 428, "lr": 9.53036083573426e-05}
{"train_loss": 0.33319499887777176, "global_step": 38180, "epoch": 428, "lr": 9.530336305430048e-05, "val_loss": 2.562467336654663}
{"train_loss": 0.2565871775150299, "global_step": 38181, "epoch": 429, "lr": 9.530311774516787e-05}
{"train_loss": 0.2575022280216217, "global_step": 38182, "epoch": 429, "lr": 9.530287242994482e-05}
{"train_loss": 0.2038000524044037, "global_step": 38183, "epoch": 429, "lr": 9.530262710863134e-05}
{"train_loss": 0.3443405330181122, "global_step": 38184, "epoch": 429, "lr": 9.530238178122748e-05}
{"train_loss": 0.3237333297729492, "global_step": 38185, "epoch": 429, "lr": 9.530213644773327e-05}
{"train_loss": 0.3816944658756256, "global_step": 38186, "epoch": 429, "lr": 9.530189110814874e-05}
{"train_loss": 0.27545642852783203, "global_step": 38187, "epoch": 429, "lr": 9.530164576247392e-05}
{"train_loss": 0.276140034198761, "global_step": 38188, "epoch": 429, "lr": 9.530140041070884e-05}
{"train_loss": 0.3571232259273529, "global_step": 38189, "epoch": 429, "lr": 9.530115505285353e-05}
{"train_loss": 0.3268016576766968, "global_step": 38190, "epoch": 429, "lr": 9.530090968890806e-05}
{"train_loss": 0.31858694553375244, "global_step": 38191, "epoch": 429, "lr": 9.530066431887241e-05}
{"train_loss": 0.3219485878944397, "global_step": 38192, "epoch": 429, "lr": 9.530041894274665e-05}
{"train_loss": 0.3606162369251251, "global_step": 38193, "epoch": 429, "lr": 9.53001735605308e-05}
{"train_loss": 0.30566146969795227, "global_step": 38194, "epoch": 429, "lr": 9.52999281722249e-05}
{"train_loss": 0.3415234088897705, "global_step": 38195, "epoch": 429, "lr": 9.529968277782897e-05}
{"train_loss": 0.2530967891216278, "global_step": 38196, "epoch": 429, "lr": 9.529943737734304e-05}
{"train_loss": 0.338442325592041, "global_step": 38197, "epoch": 429, "lr": 9.529919197076716e-05}
{"train_loss": 0.3127506673336029, "global_step": 38198, "epoch": 429, "lr": 9.529894655810137e-05}
{"train_loss": 0.39899712800979614, "global_step": 38199, "epoch": 429, "lr": 9.529870113934567e-05}
{"train_loss": 0.4101991653442383, "global_step": 38200, "epoch": 429, "lr": 9.529845571450011e-05}
{"train_loss": 0.21778857707977295, "global_step": 38201, "epoch": 429, "lr": 9.529821028356474e-05}
{"train_loss": 0.32563650608062744, "global_step": 38202, "epoch": 429, "lr": 9.529796484653957e-05}
{"train_loss": 0.3240041434764862, "global_step": 38203, "epoch": 429, "lr": 9.529771940342463e-05}
{"train_loss": 0.4394189715385437, "global_step": 38204, "epoch": 429, "lr": 9.529747395421998e-05}
{"train_loss": 0.32499054074287415, "global_step": 38205, "epoch": 429, "lr": 9.529722849892564e-05}
{"train_loss": 0.3785399794578552, "global_step": 38206, "epoch": 429, "lr": 9.529698303754162e-05}
{"train_loss": 0.32650724053382874, "global_step": 38207, "epoch": 429, "lr": 9.529673757006799e-05}
{"train_loss": 0.20603910088539124, "global_step": 38208, "epoch": 429, "lr": 9.529649209650476e-05}
{"train_loss": 0.2294943779706955, "global_step": 38209, "epoch": 429, "lr": 9.529624661685197e-05}
{"train_loss": 0.40876442193984985, "global_step": 38210, "epoch": 429, "lr": 9.529600113110964e-05}
{"train_loss": 0.3427542448043823, "global_step": 38211, "epoch": 429, "lr": 9.529575563927784e-05}
{"train_loss": 0.3469923138618469, "global_step": 38212, "epoch": 429, "lr": 9.529551014135655e-05}
{"train_loss": 0.35061249136924744, "global_step": 38213, "epoch": 429, "lr": 9.529526463734585e-05}
{"train_loss": 0.37465620040893555, "global_step": 38214, "epoch": 429, "lr": 9.529501912724574e-05}
{"train_loss": 0.3971762955188751, "global_step": 38215, "epoch": 429, "lr": 9.529477361105628e-05}
{"train_loss": 0.3723377585411072, "global_step": 38216, "epoch": 429, "lr": 9.529452808877747e-05}
{"train_loss": 0.23977506160736084, "global_step": 38217, "epoch": 429, "lr": 9.529428256040938e-05}
{"train_loss": 0.28475725650787354, "global_step": 38218, "epoch": 429, "lr": 9.529403702595203e-05}
{"train_loss": 0.2893680930137634, "global_step": 38219, "epoch": 429, "lr": 9.529379148540545e-05}
{"train_loss": 0.366701602935791, "global_step": 38220, "epoch": 429, "lr": 9.529354593876965e-05}
{"train_loss": 0.33400851488113403, "global_step": 38221, "epoch": 429, "lr": 9.52933003860447e-05}
{"train_loss": 0.36170652508735657, "global_step": 38222, "epoch": 429, "lr": 9.529305482723062e-05}
{"train_loss": 0.36618053913116455, "global_step": 38223, "epoch": 429, "lr": 9.529280926232743e-05}
{"train_loss": 0.27144747972488403, "global_step": 38224, "epoch": 429, "lr": 9.529256369133518e-05}
{"train_loss": 0.28452205657958984, "global_step": 38225, "epoch": 429, "lr": 9.52923181142539e-05}
{"train_loss": 0.2889288365840912, "global_step": 38226, "epoch": 429, "lr": 9.52920725310836e-05}
{"train_loss": 0.3703431189060211, "global_step": 38227, "epoch": 429, "lr": 9.529182694182436e-05}
{"train_loss": 0.36668074131011963, "global_step": 38228, "epoch": 429, "lr": 9.529158134647617e-05}
{"train_loss": 0.25401902198791504, "global_step": 38229, "epoch": 429, "lr": 9.52913357450391e-05}
{"train_loss": 0.3059800863265991, "global_step": 38230, "epoch": 429, "lr": 9.529109013751314e-05}
{"train_loss": 0.23298318684101105, "global_step": 38231, "epoch": 429, "lr": 9.529084452389835e-05}
{"train_loss": 0.36027342081069946, "global_step": 38232, "epoch": 429, "lr": 9.529059890419476e-05}
{"train_loss": 0.33980870246887207, "global_step": 38233, "epoch": 429, "lr": 9.52903532784024e-05}
{"train_loss": 0.2983825206756592, "global_step": 38234, "epoch": 429, "lr": 9.529010764652131e-05}
{"train_loss": 0.25460249185562134, "global_step": 38235, "epoch": 429, "lr": 9.52898620085515e-05}
{"train_loss": 0.28627389669418335, "global_step": 38236, "epoch": 429, "lr": 9.528961636449304e-05}
{"train_loss": 0.34355539083480835, "global_step": 38237, "epoch": 429, "lr": 9.528937071434594e-05}
{"train_loss": 0.29690665006637573, "global_step": 38238, "epoch": 429, "lr": 9.528912505811023e-05}
{"train_loss": 0.39442041516304016, "global_step": 38239, "epoch": 429, "lr": 9.528887939578594e-05}
{"train_loss": 0.3753036558628082, "global_step": 38240, "epoch": 429, "lr": 9.528863372737314e-05}
{"train_loss": 0.24230864644050598, "global_step": 38241, "epoch": 429, "lr": 9.52883880528718e-05}
{"train_loss": 0.330927312374115, "global_step": 38242, "epoch": 429, "lr": 9.528814237228201e-05}
{"train_loss": 0.23407387733459473, "global_step": 38243, "epoch": 429, "lr": 9.528789668560378e-05}
{"train_loss": 0.36628836393356323, "global_step": 38244, "epoch": 429, "lr": 9.528765099283714e-05}
{"train_loss": 0.30465638637542725, "global_step": 38245, "epoch": 429, "lr": 9.528740529398213e-05}
{"train_loss": 0.30698052048683167, "global_step": 38246, "epoch": 429, "lr": 9.528715958903877e-05}
{"train_loss": 0.3161183297634125, "global_step": 38247, "epoch": 429, "lr": 9.52869138780071e-05}
{"train_loss": 0.41758397221565247, "global_step": 38248, "epoch": 429, "lr": 9.528666816088718e-05}
{"train_loss": 0.2864489257335663, "global_step": 38249, "epoch": 429, "lr": 9.5286422437679e-05}
{"train_loss": 0.26783454418182373, "global_step": 38250, "epoch": 429, "lr": 9.528617670838262e-05}
{"train_loss": 0.3361896574497223, "global_step": 38251, "epoch": 429, "lr": 9.528593097299807e-05}
{"train_loss": 0.3502187728881836, "global_step": 38252, "epoch": 429, "lr": 9.528568523152538e-05}
{"train_loss": 0.2783507704734802, "global_step": 38253, "epoch": 429, "lr": 9.528543948396458e-05}
{"train_loss": 0.26960262656211853, "global_step": 38254, "epoch": 429, "lr": 9.528519373031569e-05}
{"train_loss": 0.34535524249076843, "global_step": 38255, "epoch": 429, "lr": 9.528494797057876e-05}
{"train_loss": 0.3593841791152954, "global_step": 38256, "epoch": 429, "lr": 9.528470220475383e-05}
{"train_loss": 0.44777852296829224, "global_step": 38257, "epoch": 429, "lr": 9.528445643284094e-05}
{"train_loss": 0.309262752532959, "global_step": 38258, "epoch": 429, "lr": 9.528421065484008e-05}
{"train_loss": 0.2559819221496582, "global_step": 38259, "epoch": 429, "lr": 9.528396487075132e-05}
{"train_loss": 0.4403753876686096, "global_step": 38260, "epoch": 429, "lr": 9.528371908057467e-05}
{"train_loss": 0.2995714843273163, "global_step": 38261, "epoch": 429, "lr": 9.52834732843102e-05}
{"train_loss": 0.3819097876548767, "global_step": 38262, "epoch": 429, "lr": 9.52832274819579e-05}
{"train_loss": 0.3084377646446228, "global_step": 38263, "epoch": 429, "lr": 9.528298167351784e-05}
{"train_loss": 0.3940708637237549, "global_step": 38264, "epoch": 429, "lr": 9.528273585899001e-05}
{"train_loss": 0.2542588412761688, "global_step": 38265, "epoch": 429, "lr": 9.528249003837449e-05}
{"train_loss": 0.38223665952682495, "global_step": 38266, "epoch": 429, "lr": 9.528224421167128e-05}
{"train_loss": 0.33792930841445923, "global_step": 38267, "epoch": 429, "lr": 9.528199837888043e-05}
{"train_loss": 0.34116628766059875, "global_step": 38268, "epoch": 429, "lr": 9.528175254000195e-05}
{"train_loss": 0.3241233705134874, "global_step": 38269, "epoch": 429, "lr": 9.528150669503592e-05, "val_loss": 2.575895071029663}
{"train_loss": 0.37797752022743225, "global_step": 38270, "epoch": 430, "lr": 9.528126084398232e-05}
{"train_loss": 0.35398170351982117, "global_step": 38271, "epoch": 430, "lr": 9.528101498684122e-05}
{"train_loss": 0.34173157811164856, "global_step": 38272, "epoch": 430, "lr": 9.528076912361264e-05}
{"train_loss": 0.3314126133918762, "global_step": 38273, "epoch": 430, "lr": 9.528052325429659e-05}
{"train_loss": 0.2641403079032898, "global_step": 38274, "epoch": 430, "lr": 9.528027737889315e-05}
{"train_loss": 0.33669403195381165, "global_step": 38275, "epoch": 430, "lr": 9.528003149740232e-05}
{"train_loss": 0.26565951108932495, "global_step": 38276, "epoch": 430, "lr": 9.527978560982415e-05}
{"train_loss": 0.3518162667751312, "global_step": 38277, "epoch": 430, "lr": 9.527953971615865e-05}
{"train_loss": 0.22936634719371796, "global_step": 38278, "epoch": 430, "lr": 9.527929381640588e-05}
{"train_loss": 0.33515089750289917, "global_step": 38279, "epoch": 430, "lr": 9.527904791056585e-05}
{"train_loss": 0.3346145451068878, "global_step": 38280, "epoch": 430, "lr": 9.52788019986386e-05}
{"train_loss": 0.35327428579330444, "global_step": 38281, "epoch": 430, "lr": 9.52785560806242e-05}
{"train_loss": 0.3868236541748047, "global_step": 38282, "epoch": 430, "lr": 9.527831015652261e-05}
{"train_loss": 0.27561891078948975, "global_step": 38283, "epoch": 430, "lr": 9.527806422633392e-05}
{"train_loss": 0.3359402120113373, "global_step": 38284, "epoch": 430, "lr": 9.527781829005815e-05}
{"train_loss": 0.2955121099948883, "global_step": 38285, "epoch": 430, "lr": 9.527757234769532e-05}
{"train_loss": 0.37190014123916626, "global_step": 38286, "epoch": 430, "lr": 9.527732639924547e-05}
{"train_loss": 0.2046438455581665, "global_step": 38287, "epoch": 430, "lr": 9.527708044470865e-05}
{"train_loss": 0.2868915796279907, "global_step": 38288, "epoch": 430, "lr": 9.527683448408486e-05}
{"train_loss": 0.33298376202583313, "global_step": 38289, "epoch": 430, "lr": 9.527658851737417e-05}
{"train_loss": 0.32929474115371704, "global_step": 38290, "epoch": 430, "lr": 9.527634254457658e-05}
{"train_loss": 0.2616782486438751, "global_step": 38291, "epoch": 430, "lr": 9.527609656569216e-05}
{"train_loss": 0.2946925163269043, "global_step": 38292, "epoch": 430, "lr": 9.52758505807209e-05}
{"train_loss": 0.3078939616680145, "global_step": 38293, "epoch": 430, "lr": 9.527560458966285e-05}
{"train_loss": 0.27393731474876404, "global_step": 38294, "epoch": 430, "lr": 9.527535859251806e-05}
{"train_loss": 0.2585395276546478, "global_step": 38295, "epoch": 430, "lr": 9.527511258928654e-05}
{"train_loss": 0.40398094058036804, "global_step": 38296, "epoch": 430, "lr": 9.527486657996833e-05}
{"train_loss": 0.21111977100372314, "global_step": 38297, "epoch": 430, "lr": 9.527462056456349e-05}
{"train_loss": 0.32737329602241516, "global_step": 38298, "epoch": 430, "lr": 9.527437454307201e-05}
{"train_loss": 0.286480188369751, "global_step": 38299, "epoch": 430, "lr": 9.527412851549394e-05}
{"train_loss": 0.31978505849838257, "global_step": 38300, "epoch": 430, "lr": 9.527388248182932e-05}
{"train_loss": 0.3762706220149994, "global_step": 38301, "epoch": 430, "lr": 9.527363644207818e-05}
{"train_loss": 0.36345067620277405, "global_step": 38302, "epoch": 430, "lr": 9.527339039624054e-05}
{"train_loss": 0.3602689802646637, "global_step": 38303, "epoch": 430, "lr": 9.527314434431647e-05}
{"train_loss": 0.45249155163764954, "global_step": 38304, "epoch": 430, "lr": 9.527289828630597e-05}
{"train_loss": 0.4463898837566376, "global_step": 38305, "epoch": 430, "lr": 9.527265222220907e-05}
{"train_loss": 0.31240493059158325, "global_step": 38306, "epoch": 430, "lr": 9.527240615202581e-05}
{"train_loss": 0.307931125164032, "global_step": 38307, "epoch": 430, "lr": 9.527216007575623e-05}
{"train_loss": 0.321555495262146, "global_step": 38308, "epoch": 430, "lr": 9.527191399340036e-05}
{"train_loss": 0.25332555174827576, "global_step": 38309, "epoch": 430, "lr": 9.527166790495825e-05}
{"train_loss": 0.37419217824935913, "global_step": 38310, "epoch": 430, "lr": 9.527142181042989e-05}
{"train_loss": 0.3501132130622864, "global_step": 38311, "epoch": 430, "lr": 9.527117570981537e-05}
{"train_loss": 0.3271888792514801, "global_step": 38312, "epoch": 430, "lr": 9.527092960311466e-05}
{"train_loss": 0.3917565643787384, "global_step": 38313, "epoch": 430, "lr": 9.527068349032784e-05}
{"train_loss": 0.44204074144363403, "global_step": 38314, "epoch": 430, "lr": 9.527043737145492e-05}
{"train_loss": 0.4345420300960541, "global_step": 38315, "epoch": 430, "lr": 9.527019124649595e-05}
{"train_loss": 0.34754452109336853, "global_step": 38316, "epoch": 430, "lr": 9.526994511545095e-05}
{"train_loss": 0.28402620553970337, "global_step": 38317, "epoch": 430, "lr": 9.526969897831997e-05}
{"train_loss": 0.32893678545951843, "global_step": 38318, "epoch": 430, "lr": 9.526945283510302e-05}
{"train_loss": 0.29311510920524597, "global_step": 38319, "epoch": 430, "lr": 9.526920668580015e-05}
{"train_loss": 0.429606169462204, "global_step": 38320, "epoch": 430, "lr": 9.526896053041138e-05}
{"train_loss": 0.3896470367908478, "global_step": 38321, "epoch": 430, "lr": 9.526871436893674e-05}
{"train_loss": 0.22260001301765442, "global_step": 38322, "epoch": 430, "lr": 9.526846820137629e-05}
{"train_loss": 0.3422708213329315, "global_step": 38323, "epoch": 430, "lr": 9.526822202773005e-05}
{"train_loss": 0.23571819067001343, "global_step": 38324, "epoch": 430, "lr": 9.526797584799804e-05}
{"train_loss": 0.3948875069618225, "global_step": 38325, "epoch": 430, "lr": 9.52677296621803e-05}
{"train_loss": 0.29873135685920715, "global_step": 38326, "epoch": 430, "lr": 9.526748347027687e-05}
{"train_loss": 0.34966301918029785, "global_step": 38327, "epoch": 430, "lr": 9.526723727228776e-05}
{"train_loss": 0.3278854191303253, "global_step": 38328, "epoch": 430, "lr": 9.526699106821306e-05}
{"train_loss": 0.29190826416015625, "global_step": 38329, "epoch": 430, "lr": 9.526674485805274e-05}
{"train_loss": 0.33411794900894165, "global_step": 38330, "epoch": 430, "lr": 9.526649864180686e-05}
{"train_loss": 0.4910948872566223, "global_step": 38331, "epoch": 430, "lr": 9.526625241947545e-05}
{"train_loss": 0.1992386132478714, "global_step": 38332, "epoch": 430, "lr": 9.526600619105855e-05}
{"train_loss": 0.36713549494743347, "global_step": 38333, "epoch": 430, "lr": 9.526575995655618e-05}
{"train_loss": 0.3353877663612366, "global_step": 38334, "epoch": 430, "lr": 9.52655137159684e-05}
{"train_loss": 0.3085702061653137, "global_step": 38335, "epoch": 430, "lr": 9.52652674692952e-05}
{"train_loss": 0.3266327381134033, "global_step": 38336, "epoch": 430, "lr": 9.526502121653664e-05}
{"train_loss": 0.39033013582229614, "global_step": 38337, "epoch": 430, "lr": 9.526477495769275e-05}
{"train_loss": 0.33541035652160645, "global_step": 38338, "epoch": 430, "lr": 9.526452869276356e-05}
{"train_loss": 0.4387240409851074, "global_step": 38339, "epoch": 430, "lr": 9.526428242174912e-05}
{"train_loss": 0.3724369406700134, "global_step": 38340, "epoch": 430, "lr": 9.526403614464945e-05}
{"train_loss": 0.3139381408691406, "global_step": 38341, "epoch": 430, "lr": 9.526378986146456e-05}
{"train_loss": 0.2734065055847168, "global_step": 38342, "epoch": 430, "lr": 9.526354357219451e-05}
{"train_loss": 0.3941287696361542, "global_step": 38343, "epoch": 430, "lr": 9.526329727683935e-05}
{"train_loss": 0.4378342628479004, "global_step": 38344, "epoch": 430, "lr": 9.526305097539907e-05}
{"train_loss": 0.42056867480278015, "global_step": 38345, "epoch": 430, "lr": 9.526280466787374e-05}
{"train_loss": 0.38898760080337524, "global_step": 38346, "epoch": 430, "lr": 9.526255835426336e-05}
{"train_loss": 0.3514339029788971, "global_step": 38347, "epoch": 430, "lr": 9.5262312034568e-05}
{"train_loss": 0.4127185046672821, "global_step": 38348, "epoch": 430, "lr": 9.526206570878766e-05}
{"train_loss": 0.46168580651283264, "global_step": 38349, "epoch": 430, "lr": 9.52618193769224e-05}
{"train_loss": 0.3583647906780243, "global_step": 38350, "epoch": 430, "lr": 9.526157303897223e-05}
{"train_loss": 0.5074383020401001, "global_step": 38351, "epoch": 430, "lr": 9.52613266949372e-05}
{"train_loss": 0.48295414447784424, "global_step": 38352, "epoch": 430, "lr": 9.526108034481733e-05}
{"train_loss": 0.3924601674079895, "global_step": 38353, "epoch": 430, "lr": 9.526083398861265e-05}
{"train_loss": 0.2607377767562866, "global_step": 38354, "epoch": 430, "lr": 9.526058762632322e-05}
{"train_loss": 0.452285498380661, "global_step": 38355, "epoch": 430, "lr": 9.526034125794904e-05}
{"train_loss": 0.3447004556655884, "global_step": 38356, "epoch": 430, "lr": 9.526009488349016e-05}
{"train_loss": 0.45941540598869324, "global_step": 38357, "epoch": 430, "lr": 9.525984850294662e-05}
{"train_loss": 0.34436411937970796, "global_step": 38358, "epoch": 430, "lr": 9.525960211631846e-05, "val_loss": 2.6772396564483643, "train_action_mse_error": 25.66250991821289}
{"train_loss": 0.382098913192749, "global_step": 38359, "epoch": 431, "lr": 9.525935572360567e-05}
{"train_loss": 0.29975804686546326, "global_step": 38360, "epoch": 431, "lr": 9.525910932480832e-05}
{"train_loss": 0.26144513487815857, "global_step": 38361, "epoch": 431, "lr": 9.525886291992645e-05}
{"train_loss": 0.3400879502296448, "global_step": 38362, "epoch": 431, "lr": 9.525861650896007e-05}
{"train_loss": 0.2772618532180786, "global_step": 38363, "epoch": 431, "lr": 9.525837009190921e-05}
{"train_loss": 0.3225160539150238, "global_step": 38364, "epoch": 431, "lr": 9.525812366877391e-05}
{"train_loss": 0.29040154814720154, "global_step": 38365, "epoch": 431, "lr": 9.525787723955423e-05}
{"train_loss": 0.40977922081947327, "global_step": 38366, "epoch": 431, "lr": 9.525763080425017e-05}
{"train_loss": 0.3083566129207611, "global_step": 38367, "epoch": 431, "lr": 9.525738436286176e-05}
{"train_loss": 0.3303094208240509, "global_step": 38368, "epoch": 431, "lr": 9.525713791538906e-05}
{"train_loss": 0.29865771532058716, "global_step": 38369, "epoch": 431, "lr": 9.525689146183207e-05}
{"train_loss": 0.31868913769721985, "global_step": 38370, "epoch": 431, "lr": 9.525664500219087e-05}
{"train_loss": 0.34830379486083984, "global_step": 38371, "epoch": 431, "lr": 9.525639853646546e-05}
{"train_loss": 0.3514653146266937, "global_step": 38372, "epoch": 431, "lr": 9.525615206465587e-05}
{"train_loss": 0.27346643805503845, "global_step": 38373, "epoch": 431, "lr": 9.525590558676214e-05}
{"train_loss": 0.3205174505710602, "global_step": 38374, "epoch": 431, "lr": 9.525565910278431e-05}
{"train_loss": 0.3924163579940796, "global_step": 38375, "epoch": 431, "lr": 9.525541261272241e-05}
{"train_loss": 0.2858142852783203, "global_step": 38376, "epoch": 431, "lr": 9.525516611657648e-05}
{"train_loss": 0.34678155183792114, "global_step": 38377, "epoch": 431, "lr": 9.525491961434651e-05}
{"train_loss": 0.3175068199634552, "global_step": 38378, "epoch": 431, "lr": 9.52546731060326e-05}
{"train_loss": 0.34425920248031616, "global_step": 38379, "epoch": 431, "lr": 9.525442659163474e-05}
{"train_loss": 0.396796852350235, "global_step": 38380, "epoch": 431, "lr": 9.525418007115298e-05}
{"train_loss": 0.36394914984703064, "global_step": 38381, "epoch": 431, "lr": 9.525393354458735e-05}
{"train_loss": 0.30190509557724, "global_step": 38382, "epoch": 431, "lr": 9.525368701193786e-05}
{"train_loss": 0.3705610930919647, "global_step": 38383, "epoch": 431, "lr": 9.525344047320459e-05}
{"train_loss": 0.3686056137084961, "global_step": 38384, "epoch": 431, "lr": 9.525319392838753e-05}
{"train_loss": 0.3275684714317322, "global_step": 38385, "epoch": 431, "lr": 9.525294737748674e-05}
{"train_loss": 0.3271869122982025, "global_step": 38386, "epoch": 431, "lr": 9.525270082050223e-05}
{"train_loss": 0.3475710153579712, "global_step": 38387, "epoch": 431, "lr": 9.525245425743405e-05}
{"train_loss": 0.35881975293159485, "global_step": 38388, "epoch": 431, "lr": 9.525220768828222e-05}
{"train_loss": 0.3013516366481781, "global_step": 38389, "epoch": 431, "lr": 9.52519611130468e-05}
{"train_loss": 0.25902554392814636, "global_step": 38390, "epoch": 431, "lr": 9.52517145317278e-05}
{"train_loss": 0.2621028423309326, "global_step": 38391, "epoch": 431, "lr": 9.525146794432526e-05}
{"train_loss": 0.3649117648601532, "global_step": 38392, "epoch": 431, "lr": 9.52512213508392e-05}
{"train_loss": 0.28598666191101074, "global_step": 38393, "epoch": 431, "lr": 9.525097475126967e-05}
{"train_loss": 0.2536989450454712, "global_step": 38394, "epoch": 431, "lr": 9.52507281456167e-05}
{"train_loss": 0.37077587842941284, "global_step": 38395, "epoch": 431, "lr": 9.525048153388033e-05}
{"train_loss": 0.40602952241897583, "global_step": 38396, "epoch": 431, "lr": 9.525023491606057e-05}
{"train_loss": 0.4568176865577698, "global_step": 38397, "epoch": 431, "lr": 9.524998829215747e-05}
{"train_loss": 0.2809041738510132, "global_step": 38398, "epoch": 431, "lr": 9.524974166217107e-05}
{"train_loss": 0.37409693002700806, "global_step": 38399, "epoch": 431, "lr": 9.524949502610139e-05}
{"train_loss": 0.2458937019109726, "global_step": 38400, "epoch": 431, "lr": 9.524924838394847e-05}
{"train_loss": 0.347154438495636, "global_step": 38401, "epoch": 431, "lr": 9.524900173571235e-05}
{"train_loss": 0.30950069427490234, "global_step": 38402, "epoch": 431, "lr": 9.524875508139302e-05}
{"train_loss": 0.41790086030960083, "global_step": 38403, "epoch": 431, "lr": 9.524850842099056e-05}
{"train_loss": 0.29799792170524597, "global_step": 38404, "epoch": 431, "lr": 9.524826175450501e-05}
{"train_loss": 0.42705878615379333, "global_step": 38405, "epoch": 431, "lr": 9.524801508193636e-05}
{"train_loss": 0.2796570062637329, "global_step": 38406, "epoch": 431, "lr": 9.524776840328466e-05}
{"train_loss": 0.392428994178772, "global_step": 38407, "epoch": 431, "lr": 9.524752171854999e-05}
{"train_loss": 0.36511409282684326, "global_step": 38408, "epoch": 431, "lr": 9.52472750277323e-05}
{"train_loss": 0.3124000132083893, "global_step": 38409, "epoch": 431, "lr": 9.524702833083168e-05}
{"train_loss": 0.28358224034309387, "global_step": 38410, "epoch": 431, "lr": 9.524678162784814e-05}
{"train_loss": 0.4579627513885498, "global_step": 38411, "epoch": 431, "lr": 9.524653491878174e-05}
{"train_loss": 0.3876529335975647, "global_step": 38412, "epoch": 431, "lr": 9.524628820363248e-05}
{"train_loss": 0.30520957708358765, "global_step": 38413, "epoch": 431, "lr": 9.524604148240042e-05}
{"train_loss": 0.3760550022125244, "global_step": 38414, "epoch": 431, "lr": 9.524579475508557e-05}
{"train_loss": 0.3708416223526001, "global_step": 38415, "epoch": 431, "lr": 9.524554802168797e-05}
{"train_loss": 0.2827043831348419, "global_step": 38416, "epoch": 431, "lr": 9.524530128220768e-05}
{"train_loss": 0.3400036096572876, "global_step": 38417, "epoch": 431, "lr": 9.524505453664469e-05}
{"train_loss": 0.37738075852394104, "global_step": 38418, "epoch": 431, "lr": 9.524480778499906e-05}
{"train_loss": 0.3509521782398224, "global_step": 38419, "epoch": 431, "lr": 9.524456102727082e-05}
{"train_loss": 0.27276432514190674, "global_step": 38420, "epoch": 431, "lr": 9.524431426346e-05}
{"train_loss": 0.3977816104888916, "global_step": 38421, "epoch": 431, "lr": 9.524406749356663e-05}
{"train_loss": 0.36242198944091797, "global_step": 38422, "epoch": 431, "lr": 9.524382071759075e-05}
{"train_loss": 0.3676716983318329, "global_step": 38423, "epoch": 431, "lr": 9.524357393553238e-05}
{"train_loss": 0.23812542855739594, "global_step": 38424, "epoch": 431, "lr": 9.524332714739157e-05}
{"train_loss": 0.3345150351524353, "global_step": 38425, "epoch": 431, "lr": 9.524308035316834e-05}
{"train_loss": 0.3625257909297943, "global_step": 38426, "epoch": 431, "lr": 9.524283355286274e-05}
{"train_loss": 0.33126145601272583, "global_step": 38427, "epoch": 431, "lr": 9.524258674647479e-05}
{"train_loss": 0.26872149109840393, "global_step": 38428, "epoch": 431, "lr": 9.524233993400454e-05}
{"train_loss": 0.36313191056251526, "global_step": 38429, "epoch": 431, "lr": 9.524209311545197e-05}
{"train_loss": 0.22150424122810364, "global_step": 38430, "epoch": 431, "lr": 9.524184629081719e-05}
{"train_loss": 0.3095847964286804, "global_step": 38431, "epoch": 431, "lr": 9.524159946010018e-05}
{"train_loss": 0.30368924140930176, "global_step": 38432, "epoch": 431, "lr": 9.524135262330098e-05}
{"train_loss": 0.33416417241096497, "global_step": 38433, "epoch": 431, "lr": 9.524110578041965e-05}
{"train_loss": 0.2769415080547333, "global_step": 38434, "epoch": 431, "lr": 9.524085893145618e-05}
{"train_loss": 0.3490361273288727, "global_step": 38435, "epoch": 431, "lr": 9.524061207641064e-05}
{"train_loss": 0.30361345410346985, "global_step": 38436, "epoch": 431, "lr": 9.524036521528305e-05}
{"train_loss": 0.28822600841522217, "global_step": 38437, "epoch": 431, "lr": 9.524011834807346e-05}
{"train_loss": 0.33380985260009766, "global_step": 38438, "epoch": 431, "lr": 9.523987147478187e-05}
{"train_loss": 0.3683360517024994, "global_step": 38439, "epoch": 431, "lr": 9.523962459540834e-05}
{"train_loss": 0.3161970376968384, "global_step": 38440, "epoch": 431, "lr": 9.523937770995289e-05}
{"train_loss": 0.29281947016716003, "global_step": 38441, "epoch": 431, "lr": 9.523913081841556e-05}
{"train_loss": 0.4248647689819336, "global_step": 38442, "epoch": 431, "lr": 9.523888392079638e-05}
{"train_loss": 0.28378355503082275, "global_step": 38443, "epoch": 431, "lr": 9.523863701709538e-05}
{"train_loss": 0.30810368061065674, "global_step": 38444, "epoch": 431, "lr": 9.52383901073126e-05}
{"train_loss": 0.40530622005462646, "global_step": 38445, "epoch": 431, "lr": 9.523814319144806e-05}
{"train_loss": 0.46317631006240845, "global_step": 38446, "epoch": 431, "lr": 9.523789626950181e-05}
{"train_loss": 0.333891072300043, "global_step": 38447, "epoch": 431, "lr": 9.523764934147388e-05, "val_loss": 2.6700167655944824}
{"train_loss": 0.2719966769218445, "global_step": 38448, "epoch": 432, "lr": 9.523740240736429e-05}
{"train_loss": 0.3401534855365753, "global_step": 38449, "epoch": 432, "lr": 9.52371554671731e-05}
{"train_loss": 0.25744470953941345, "global_step": 38450, "epoch": 432, "lr": 9.52369085209003e-05}
{"train_loss": 0.3340928852558136, "global_step": 38451, "epoch": 432, "lr": 9.523666156854597e-05}
{"train_loss": 0.3465583324432373, "global_step": 38452, "epoch": 432, "lr": 9.523641461011011e-05}
{"train_loss": 0.3036254942417145, "global_step": 38453, "epoch": 432, "lr": 9.523616764559278e-05}
{"train_loss": 0.28090012073516846, "global_step": 38454, "epoch": 432, "lr": 9.523592067499398e-05}
{"train_loss": 0.28217780590057373, "global_step": 38455, "epoch": 432, "lr": 9.523567369831378e-05}
{"train_loss": 0.29732248187065125, "global_step": 38456, "epoch": 432, "lr": 9.523542671555218e-05}
{"train_loss": 0.33195504546165466, "global_step": 38457, "epoch": 432, "lr": 9.523517972670923e-05}
{"train_loss": 0.22501619160175323, "global_step": 38458, "epoch": 432, "lr": 9.523493273178496e-05}
{"train_loss": 0.2547730803489685, "global_step": 38459, "epoch": 432, "lr": 9.523468573077943e-05}
{"train_loss": 0.1832299828529358, "global_step": 38460, "epoch": 432, "lr": 9.523443872369262e-05}
{"train_loss": 0.2654990553855896, "global_step": 38461, "epoch": 432, "lr": 9.52341917105246e-05}
{"train_loss": 0.29292061924934387, "global_step": 38462, "epoch": 432, "lr": 9.52339446912754e-05}
{"train_loss": 0.26799502968788147, "global_step": 38463, "epoch": 432, "lr": 9.523369766594504e-05}
{"train_loss": 0.2676277458667755, "global_step": 38464, "epoch": 432, "lr": 9.523345063453356e-05}
{"train_loss": 0.33272916078567505, "global_step": 38465, "epoch": 432, "lr": 9.5233203597041e-05}
{"train_loss": 0.2877756655216217, "global_step": 38466, "epoch": 432, "lr": 9.523295655346737e-05}
{"train_loss": 0.31309229135513306, "global_step": 38467, "epoch": 432, "lr": 9.523270950381273e-05}
{"train_loss": 0.3408777713775635, "global_step": 38468, "epoch": 432, "lr": 9.52324624480771e-05}
{"train_loss": 0.3141818642616272, "global_step": 38469, "epoch": 432, "lr": 9.523221538626054e-05}
{"train_loss": 0.30086037516593933, "global_step": 38470, "epoch": 432, "lr": 9.523196831836304e-05}
{"train_loss": 0.33556240797042847, "global_step": 38471, "epoch": 432, "lr": 9.523172124438466e-05}
{"train_loss": 0.2488277554512024, "global_step": 38472, "epoch": 432, "lr": 9.523147416432542e-05}
{"train_loss": 0.2526634931564331, "global_step": 38473, "epoch": 432, "lr": 9.523122707818537e-05}
{"train_loss": 0.38978394865989685, "global_step": 38474, "epoch": 432, "lr": 9.523097998596452e-05}
{"train_loss": 0.3803544342517853, "global_step": 38475, "epoch": 432, "lr": 9.523073288766292e-05}
{"train_loss": 0.35702863335609436, "global_step": 38476, "epoch": 432, "lr": 9.52304857832806e-05}
{"train_loss": 0.2594257891178131, "global_step": 38477, "epoch": 432, "lr": 9.52302386728176e-05}
{"train_loss": 0.33104613423347473, "global_step": 38478, "epoch": 432, "lr": 9.522999155627393e-05}
{"train_loss": 0.47695812582969666, "global_step": 38479, "epoch": 432, "lr": 9.522974443364965e-05}
{"train_loss": 0.4233858287334442, "global_step": 38480, "epoch": 432, "lr": 9.522949730494478e-05}
{"train_loss": 0.2977481484413147, "global_step": 38481, "epoch": 432, "lr": 9.522925017015936e-05}
{"train_loss": 0.3572562336921692, "global_step": 38482, "epoch": 432, "lr": 9.52290030292934e-05}
{"train_loss": 0.23688213527202606, "global_step": 38483, "epoch": 432, "lr": 9.522875588234697e-05}
{"train_loss": 0.3571690320968628, "global_step": 38484, "epoch": 432, "lr": 9.522850872932008e-05}
{"train_loss": 0.44259288907051086, "global_step": 38485, "epoch": 432, "lr": 9.522826157021277e-05}
{"train_loss": 0.2591477334499359, "global_step": 38486, "epoch": 432, "lr": 9.522801440502507e-05}
{"train_loss": 0.421673446893692, "global_step": 38487, "epoch": 432, "lr": 9.522776723375701e-05}
{"train_loss": 0.4093153178691864, "global_step": 38488, "epoch": 432, "lr": 9.522752005640864e-05}
{"train_loss": 0.25802308320999146, "global_step": 38489, "epoch": 432, "lr": 9.522727287297996e-05}
{"train_loss": 0.3616991639137268, "global_step": 38490, "epoch": 432, "lr": 9.522702568347104e-05}
{"train_loss": 0.22013047337532043, "global_step": 38491, "epoch": 432, "lr": 9.52267784878819e-05}
{"train_loss": 0.2953544855117798, "global_step": 38492, "epoch": 432, "lr": 9.522653128621257e-05}
{"train_loss": 0.36040082573890686, "global_step": 38493, "epoch": 432, "lr": 9.522628407846308e-05}
{"train_loss": 0.26591214537620544, "global_step": 38494, "epoch": 432, "lr": 9.522603686463347e-05}
{"train_loss": 0.2973161041736603, "global_step": 38495, "epoch": 432, "lr": 9.522578964472378e-05}
{"train_loss": 0.3384498655796051, "global_step": 38496, "epoch": 432, "lr": 9.522554241873402e-05}
{"train_loss": 0.324670672416687, "global_step": 38497, "epoch": 432, "lr": 9.522529518666423e-05}
{"train_loss": 0.3975178301334381, "global_step": 38498, "epoch": 432, "lr": 9.522504794851447e-05}
{"train_loss": 0.3623928725719452, "global_step": 38499, "epoch": 432, "lr": 9.522480070428474e-05}
{"train_loss": 0.3713168501853943, "global_step": 38500, "epoch": 432, "lr": 9.52245534539751e-05}
{"train_loss": 0.3580918610095978, "global_step": 38501, "epoch": 432, "lr": 9.522430619758555e-05}
{"train_loss": 0.374307245016098, "global_step": 38502, "epoch": 432, "lr": 9.522405893511615e-05}
{"train_loss": 0.4123460650444031, "global_step": 38503, "epoch": 432, "lr": 9.522381166656693e-05}
{"train_loss": 0.30486026406288147, "global_step": 38504, "epoch": 432, "lr": 9.522356439193792e-05}
{"train_loss": 0.3280971050262451, "global_step": 38505, "epoch": 432, "lr": 9.522331711122915e-05}
{"train_loss": 0.3098534047603607, "global_step": 38506, "epoch": 432, "lr": 9.522306982444067e-05}
{"train_loss": 0.338641494512558, "global_step": 38507, "epoch": 432, "lr": 9.522282253157249e-05}
{"train_loss": 0.29909253120422363, "global_step": 38508, "epoch": 432, "lr": 9.522257523262464e-05}
{"train_loss": 0.3960149586200714, "global_step": 38509, "epoch": 432, "lr": 9.522232792759718e-05}
{"train_loss": 0.41275879740715027, "global_step": 38510, "epoch": 432, "lr": 9.522208061649013e-05}
{"train_loss": 0.24305953085422516, "global_step": 38511, "epoch": 432, "lr": 9.522183329930352e-05}
{"train_loss": 0.25981059670448303, "global_step": 38512, "epoch": 432, "lr": 9.522158597603739e-05}
{"train_loss": 0.3800097107887268, "global_step": 38513, "epoch": 432, "lr": 9.522133864669176e-05}
{"train_loss": 0.3614799678325653, "global_step": 38514, "epoch": 432, "lr": 9.522109131126667e-05}
{"train_loss": 0.2974138557910919, "global_step": 38515, "epoch": 432, "lr": 9.522084396976216e-05}
{"train_loss": 0.30995839834213257, "global_step": 38516, "epoch": 432, "lr": 9.522059662217826e-05}
{"train_loss": 0.39090004563331604, "global_step": 38517, "epoch": 432, "lr": 9.5220349268515e-05}
{"train_loss": 0.3220500648021698, "global_step": 38518, "epoch": 432, "lr": 9.522010190877242e-05}
{"train_loss": 0.3404870927333832, "global_step": 38519, "epoch": 432, "lr": 9.521985454295055e-05}
{"train_loss": 0.2564026117324829, "global_step": 38520, "epoch": 432, "lr": 9.521960717104941e-05}
{"train_loss": 0.3381583094596863, "global_step": 38521, "epoch": 432, "lr": 9.521935979306905e-05}
{"train_loss": 0.26159369945526123, "global_step": 38522, "epoch": 432, "lr": 9.521911240900951e-05}
{"train_loss": 0.2447734922170639, "global_step": 38523, "epoch": 432, "lr": 9.521886501887079e-05}
{"train_loss": 0.2562595307826996, "global_step": 38524, "epoch": 432, "lr": 9.521861762265297e-05}
{"train_loss": 0.3072563111782074, "global_step": 38525, "epoch": 432, "lr": 9.521837022035604e-05}
{"train_loss": 0.40659794211387634, "global_step": 38526, "epoch": 432, "lr": 9.521812281198006e-05}
{"train_loss": 0.43421587347984314, "global_step": 38527, "epoch": 432, "lr": 9.521787539752505e-05}
{"train_loss": 0.3166225254535675, "global_step": 38528, "epoch": 432, "lr": 9.521762797699104e-05}
{"train_loss": 0.2630626857280731, "global_step": 38529, "epoch": 432, "lr": 9.521738055037808e-05}
{"train_loss": 0.25934284925460815, "global_step": 38530, "epoch": 432, "lr": 9.521713311768619e-05}
{"train_loss": 0.3632639944553375, "global_step": 38531, "epoch": 432, "lr": 9.521688567891541e-05}
{"train_loss": 0.3538820147514343, "global_step": 38532, "epoch": 432, "lr": 9.521663823406577e-05}
{"train_loss": 0.2892266809940338, "global_step": 38533, "epoch": 432, "lr": 9.521639078313731e-05}
{"train_loss": 0.2240273356437683, "global_step": 38534, "epoch": 432, "lr": 9.521614332613005e-05}
{"train_loss": 0.3153512179851532, "global_step": 38535, "epoch": 432, "lr": 9.521589586304404e-05}
{"train_loss": 0.3197771063011684, "global_step": 38536, "epoch": 432, "lr": 9.52156483938793e-05, "val_loss": 2.708524703979492}
{"train_loss": 0.3420456349849701, "global_step": 38537, "epoch": 433, "lr": 9.521540091863586e-05}
{"train_loss": 0.3169976770877838, "global_step": 38538, "epoch": 433, "lr": 9.521515343731376e-05}
{"train_loss": 0.3906572163105011, "global_step": 38539, "epoch": 433, "lr": 9.521490594991304e-05}
{"train_loss": 0.29532161355018616, "global_step": 38540, "epoch": 433, "lr": 9.521465845643373e-05}
{"train_loss": 0.2427445948123932, "global_step": 38541, "epoch": 433, "lr": 9.521441095687586e-05}
{"train_loss": 0.31404438614845276, "global_step": 38542, "epoch": 433, "lr": 9.521416345123945e-05}
{"train_loss": 0.40315836668014526, "global_step": 38543, "epoch": 433, "lr": 9.521391593952456e-05}
{"train_loss": 0.3613363206386566, "global_step": 38544, "epoch": 433, "lr": 9.521366842173121e-05}
{"train_loss": 0.325992614030838, "global_step": 38545, "epoch": 433, "lr": 9.521342089785942e-05}
{"train_loss": 0.31294819712638855, "global_step": 38546, "epoch": 433, "lr": 9.521317336790926e-05}
{"train_loss": 0.30428197979927063, "global_step": 38547, "epoch": 433, "lr": 9.521292583188073e-05}
{"train_loss": 0.4039006233215332, "global_step": 38548, "epoch": 433, "lr": 9.521267828977387e-05}
{"train_loss": 0.23077721893787384, "global_step": 38549, "epoch": 433, "lr": 9.521243074158871e-05}
{"train_loss": 0.2945941984653473, "global_step": 38550, "epoch": 433, "lr": 9.52121831873253e-05}
{"train_loss": 0.3035722076892853, "global_step": 38551, "epoch": 433, "lr": 9.521193562698365e-05}
{"train_loss": 0.3078194856643677, "global_step": 38552, "epoch": 433, "lr": 9.521168806056381e-05}
{"train_loss": 0.284600168466568, "global_step": 38553, "epoch": 433, "lr": 9.521144048806583e-05}
{"train_loss": 0.2523239850997925, "global_step": 38554, "epoch": 433, "lr": 9.52111929094897e-05}
{"train_loss": 0.4780150055885315, "global_step": 38555, "epoch": 433, "lr": 9.521094532483549e-05}
{"train_loss": 0.36883074045181274, "global_step": 38556, "epoch": 433, "lr": 9.52106977341032e-05}
{"train_loss": 0.333786278963089, "global_step": 38557, "epoch": 433, "lr": 9.521045013729291e-05}
{"train_loss": 0.32609236240386963, "global_step": 38558, "epoch": 433, "lr": 9.52102025344046e-05}
{"train_loss": 0.2871604263782501, "global_step": 38559, "epoch": 433, "lr": 9.520995492543835e-05}
{"train_loss": 0.2791813015937805, "global_step": 38560, "epoch": 433, "lr": 9.520970731039416e-05}
{"train_loss": 0.2995043694972992, "global_step": 38561, "epoch": 433, "lr": 9.520945968927208e-05}
{"train_loss": 0.3110581338405609, "global_step": 38562, "epoch": 433, "lr": 9.520921206207214e-05}
{"train_loss": 0.3071804642677307, "global_step": 38563, "epoch": 433, "lr": 9.520896442879437e-05}
{"train_loss": 0.3520055115222931, "global_step": 38564, "epoch": 433, "lr": 9.520871678943881e-05}
{"train_loss": 0.2679252326488495, "global_step": 38565, "epoch": 433, "lr": 9.520846914400549e-05}
{"train_loss": 0.32367876172065735, "global_step": 38566, "epoch": 433, "lr": 9.520822149249444e-05}
{"train_loss": 0.39932942390441895, "global_step": 38567, "epoch": 433, "lr": 9.52079738349057e-05}
{"train_loss": 0.39223435521125793, "global_step": 38568, "epoch": 433, "lr": 9.520772617123928e-05}
{"train_loss": 0.35983821749687195, "global_step": 38569, "epoch": 433, "lr": 9.520747850149525e-05}
{"train_loss": 0.33390456438064575, "global_step": 38570, "epoch": 433, "lr": 9.520723082567362e-05}
{"train_loss": 0.33356156945228577, "global_step": 38571, "epoch": 433, "lr": 9.520698314377444e-05}
{"train_loss": 0.34923410415649414, "global_step": 38572, "epoch": 433, "lr": 9.520673545579771e-05}
{"train_loss": 0.37171489000320435, "global_step": 38573, "epoch": 433, "lr": 9.520648776174351e-05}
{"train_loss": 0.31189221143722534, "global_step": 38574, "epoch": 433, "lr": 9.520624006161183e-05}
{"train_loss": 0.241371288895607, "global_step": 38575, "epoch": 433, "lr": 9.520599235540273e-05}
{"train_loss": 0.4056672751903534, "global_step": 38576, "epoch": 433, "lr": 9.520574464311622e-05}
{"train_loss": 0.40408915281295776, "global_step": 38577, "epoch": 433, "lr": 9.520549692475237e-05}
{"train_loss": 0.3106168210506439, "global_step": 38578, "epoch": 433, "lr": 9.520524920031119e-05}
{"train_loss": 0.2825215458869934, "global_step": 38579, "epoch": 433, "lr": 9.520500146979269e-05}
{"train_loss": 0.41195082664489746, "global_step": 38580, "epoch": 433, "lr": 9.520475373319696e-05}
{"train_loss": 0.266512393951416, "global_step": 38581, "epoch": 433, "lr": 9.5204505990524e-05}
{"train_loss": 0.30273863673210144, "global_step": 38582, "epoch": 433, "lr": 9.520425824177382e-05}
{"train_loss": 0.3423990309238434, "global_step": 38583, "epoch": 433, "lr": 9.52040104869465e-05}
{"train_loss": 0.2232367992401123, "global_step": 38584, "epoch": 433, "lr": 9.520376272604205e-05}
{"train_loss": 0.30151528120040894, "global_step": 38585, "epoch": 433, "lr": 9.520351495906049e-05}
{"train_loss": 0.4187655448913574, "global_step": 38586, "epoch": 433, "lr": 9.520326718600188e-05}
{"train_loss": 0.33100375533103943, "global_step": 38587, "epoch": 433, "lr": 9.520301940686624e-05}
{"train_loss": 0.36350369453430176, "global_step": 38588, "epoch": 433, "lr": 9.52027716216536e-05}
{"train_loss": 0.2692943215370178, "global_step": 38589, "epoch": 433, "lr": 9.520252383036401e-05}
{"train_loss": 0.40546250343322754, "global_step": 38590, "epoch": 433, "lr": 9.520227603299749e-05}
{"train_loss": 0.30450910329818726, "global_step": 38591, "epoch": 433, "lr": 9.520202822955406e-05}
{"train_loss": 0.2701414227485657, "global_step": 38592, "epoch": 433, "lr": 9.520178042003378e-05}
{"train_loss": 0.29978248476982117, "global_step": 38593, "epoch": 433, "lr": 9.520153260443667e-05}
{"train_loss": 0.334555059671402, "global_step": 38594, "epoch": 433, "lr": 9.520128478276277e-05}
{"train_loss": 0.32240962982177734, "global_step": 38595, "epoch": 433, "lr": 9.52010369550121e-05}
{"train_loss": 0.2952529489994049, "global_step": 38596, "epoch": 433, "lr": 9.520078912118471e-05}
{"train_loss": 0.29630327224731445, "global_step": 38597, "epoch": 433, "lr": 9.520054128128062e-05}
{"train_loss": 0.3410870134830475, "global_step": 38598, "epoch": 433, "lr": 9.520029343529986e-05}
{"train_loss": 0.25319400429725647, "global_step": 38599, "epoch": 433, "lr": 9.520004558324247e-05}
{"train_loss": 0.35480669140815735, "global_step": 38600, "epoch": 433, "lr": 9.51997977251085e-05}
{"train_loss": 0.3223519027233124, "global_step": 38601, "epoch": 433, "lr": 9.519954986089796e-05}
{"train_loss": 0.2874765396118164, "global_step": 38602, "epoch": 433, "lr": 9.519930199061088e-05}
{"train_loss": 0.2727019190788269, "global_step": 38603, "epoch": 433, "lr": 9.519905411424734e-05}
{"train_loss": 0.22091194987297058, "global_step": 38604, "epoch": 433, "lr": 9.51988062318073e-05}
{"train_loss": 0.23370416462421417, "global_step": 38605, "epoch": 433, "lr": 9.519855834329085e-05}
{"train_loss": 0.291386216878891, "global_step": 38606, "epoch": 433, "lr": 9.5198310448698e-05}
{"train_loss": 0.3088444769382477, "global_step": 38607, "epoch": 433, "lr": 9.51980625480288e-05}
{"train_loss": 0.3496890664100647, "global_step": 38608, "epoch": 433, "lr": 9.519781464128325e-05}
{"train_loss": 0.2618367075920105, "global_step": 38609, "epoch": 433, "lr": 9.519756672846143e-05}
{"train_loss": 0.30292850732803345, "global_step": 38610, "epoch": 433, "lr": 9.519731880956333e-05}
{"train_loss": 0.42487192153930664, "global_step": 38611, "epoch": 433, "lr": 9.5197070884589e-05}
{"train_loss": 0.3475096523761749, "global_step": 38612, "epoch": 433, "lr": 9.519682295353849e-05}
{"train_loss": 0.3619023859500885, "global_step": 38613, "epoch": 433, "lr": 9.51965750164118e-05}
{"train_loss": 0.4443560540676117, "global_step": 38614, "epoch": 433, "lr": 9.519632707320899e-05}
{"train_loss": 0.31095457077026367, "global_step": 38615, "epoch": 433, "lr": 9.519607912393009e-05}
{"train_loss": 0.35372185707092285, "global_step": 38616, "epoch": 433, "lr": 9.519583116857511e-05}
{"train_loss": 0.3576720356941223, "global_step": 38617, "epoch": 433, "lr": 9.519558320714412e-05}
{"train_loss": 0.23420622944831848, "global_step": 38618, "epoch": 433, "lr": 9.519533523963713e-05}
{"train_loss": 0.5083097219467163, "global_step": 38619, "epoch": 433, "lr": 9.519508726605418e-05}
{"train_loss": 0.2925644516944885, "global_step": 38620, "epoch": 433, "lr": 9.51948392863953e-05}
{"train_loss": 0.25445684790611267, "global_step": 38621, "epoch": 433, "lr": 9.519459130066052e-05}
{"train_loss": 0.34760963916778564, "global_step": 38622, "epoch": 433, "lr": 9.519434330884988e-05}
{"train_loss": 0.3366342782974243, "global_step": 38623, "epoch": 433, "lr": 9.519409531096341e-05}
{"train_loss": 0.33180418610572815, "global_step": 38624, "epoch": 433, "lr": 9.519384730700114e-05}
{"train_loss": 0.32502243408326353, "global_step": 38625, "epoch": 433, "lr": 9.519359929696312e-05, "val_loss": 2.62196946144104}
{"train_loss": 0.2954183518886566, "global_step": 38626, "epoch": 434, "lr": 9.519335128084936e-05}
{"train_loss": 0.2714179456233978, "global_step": 38627, "epoch": 434, "lr": 9.51931032586599e-05}
{"train_loss": 0.38639310002326965, "global_step": 38628, "epoch": 434, "lr": 9.519285523039481e-05}
{"train_loss": 0.31242626905441284, "global_step": 38629, "epoch": 434, "lr": 9.519260719605406e-05}
{"train_loss": 0.282126247882843, "global_step": 38630, "epoch": 434, "lr": 9.519235915563773e-05}
{"train_loss": 0.35867300629615784, "global_step": 38631, "epoch": 434, "lr": 9.519211110914583e-05}
{"train_loss": 0.22974014282226562, "global_step": 38632, "epoch": 434, "lr": 9.519186305657841e-05}
{"train_loss": 0.38632655143737793, "global_step": 38633, "epoch": 434, "lr": 9.519161499793549e-05}
{"train_loss": 0.2655685544013977, "global_step": 38634, "epoch": 434, "lr": 9.51913669332171e-05}
{"train_loss": 0.4176108241081238, "global_step": 38635, "epoch": 434, "lr": 9.519111886242329e-05}
{"train_loss": 0.38456565141677856, "global_step": 38636, "epoch": 434, "lr": 9.519087078555408e-05}
{"train_loss": 0.3069484233856201, "global_step": 38637, "epoch": 434, "lr": 9.519062270260952e-05}
{"train_loss": 0.3827812075614929, "global_step": 38638, "epoch": 434, "lr": 9.519037461358962e-05}
{"train_loss": 0.2762371003627777, "global_step": 38639, "epoch": 434, "lr": 9.519012651849444e-05}
{"train_loss": 0.3616596758365631, "global_step": 38640, "epoch": 434, "lr": 9.518987841732399e-05}
{"train_loss": 0.33080101013183594, "global_step": 38641, "epoch": 434, "lr": 9.51896303100783e-05}
{"train_loss": 0.25653815269470215, "global_step": 38642, "epoch": 434, "lr": 9.518938219675741e-05}
{"train_loss": 0.412709504365921, "global_step": 38643, "epoch": 434, "lr": 9.518913407736138e-05}
{"train_loss": 0.3690459132194519, "global_step": 38644, "epoch": 434, "lr": 9.518888595189022e-05}
{"train_loss": 0.23935502767562866, "global_step": 38645, "epoch": 434, "lr": 9.518863782034396e-05}
{"train_loss": 0.3445271849632263, "global_step": 38646, "epoch": 434, "lr": 9.518838968272264e-05}
{"train_loss": 0.4234223961830139, "global_step": 38647, "epoch": 434, "lr": 9.518814153902629e-05}
{"train_loss": 0.21853996813297272, "global_step": 38648, "epoch": 434, "lr": 9.518789338925493e-05}
{"train_loss": 0.2973504960536957, "global_step": 38649, "epoch": 434, "lr": 9.518764523340862e-05}
{"train_loss": 0.2784102261066437, "global_step": 38650, "epoch": 434, "lr": 9.51873970714874e-05}
{"train_loss": 0.3345482647418976, "global_step": 38651, "epoch": 434, "lr": 9.518714890349126e-05}
{"train_loss": 0.3149919807910919, "global_step": 38652, "epoch": 434, "lr": 9.518690072942027e-05}
{"train_loss": 0.4575395882129669, "global_step": 38653, "epoch": 434, "lr": 9.518665254927446e-05}
{"train_loss": 0.3058653175830841, "global_step": 38654, "epoch": 434, "lr": 9.518640436305384e-05}
{"train_loss": 0.31567370891571045, "global_step": 38655, "epoch": 434, "lr": 9.518615617075846e-05}
{"train_loss": 0.258711576461792, "global_step": 38656, "epoch": 434, "lr": 9.518590797238836e-05}
{"train_loss": 0.26593267917633057, "global_step": 38657, "epoch": 434, "lr": 9.518565976794356e-05}
{"train_loss": 0.38690146803855896, "global_step": 38658, "epoch": 434, "lr": 9.51854115574241e-05}
{"train_loss": 0.2858048379421234, "global_step": 38659, "epoch": 434, "lr": 9.518516334083e-05}
{"train_loss": 0.3276934027671814, "global_step": 38660, "epoch": 434, "lr": 9.518491511816132e-05}
{"train_loss": 0.2840576171875, "global_step": 38661, "epoch": 434, "lr": 9.518466688941808e-05}
{"train_loss": 0.32907575368881226, "global_step": 38662, "epoch": 434, "lr": 9.518441865460031e-05}
{"train_loss": 0.26372137665748596, "global_step": 38663, "epoch": 434, "lr": 9.518417041370805e-05}
{"train_loss": 0.3739393353462219, "global_step": 38664, "epoch": 434, "lr": 9.518392216674132e-05}
{"train_loss": 0.423933207988739, "global_step": 38665, "epoch": 434, "lr": 9.518367391370018e-05}
{"train_loss": 0.36519208550453186, "global_step": 38666, "epoch": 434, "lr": 9.518342565458462e-05}
{"train_loss": 0.28395310044288635, "global_step": 38667, "epoch": 434, "lr": 9.51831773893947e-05}
{"train_loss": 0.3425234854221344, "global_step": 38668, "epoch": 434, "lr": 9.518292911813047e-05}
{"train_loss": 0.3221166729927063, "global_step": 38669, "epoch": 434, "lr": 9.518268084079195e-05}
{"train_loss": 0.3333604335784912, "global_step": 38670, "epoch": 434, "lr": 9.518243255737915e-05}
{"train_loss": 0.3596018850803375, "global_step": 38671, "epoch": 434, "lr": 9.518218426789213e-05}
{"train_loss": 0.26512256264686584, "global_step": 38672, "epoch": 434, "lr": 9.518193597233093e-05}
{"train_loss": 0.3830665946006775, "global_step": 38673, "epoch": 434, "lr": 9.518168767069556e-05}
{"train_loss": 0.33720728754997253, "global_step": 38674, "epoch": 434, "lr": 9.518143936298605e-05}
{"train_loss": 0.29146358370780945, "global_step": 38675, "epoch": 434, "lr": 9.518119104920245e-05}
{"train_loss": 0.4012514352798462, "global_step": 38676, "epoch": 434, "lr": 9.51809427293448e-05}
{"train_loss": 0.39281749725341797, "global_step": 38677, "epoch": 434, "lr": 9.518069440341311e-05}
{"train_loss": 0.3200434148311615, "global_step": 38678, "epoch": 434, "lr": 9.518044607140744e-05}
{"train_loss": 0.3244155943393707, "global_step": 38679, "epoch": 434, "lr": 9.51801977333278e-05}
{"train_loss": 0.22214947640895844, "global_step": 38680, "epoch": 434, "lr": 9.517994938917426e-05}
{"train_loss": 0.24582844972610474, "global_step": 38681, "epoch": 434, "lr": 9.517970103894679e-05}
{"train_loss": 0.42942339181900024, "global_step": 38682, "epoch": 434, "lr": 9.517945268264547e-05}
{"train_loss": 0.3436831831932068, "global_step": 38683, "epoch": 434, "lr": 9.517920432027033e-05}
{"train_loss": 0.2927260994911194, "global_step": 38684, "epoch": 434, "lr": 9.517895595182141e-05}
{"train_loss": 0.29112136363983154, "global_step": 38685, "epoch": 434, "lr": 9.51787075772987e-05}
{"train_loss": 0.34627532958984375, "global_step": 38686, "epoch": 434, "lr": 9.517845919670227e-05}
{"train_loss": 0.4503448009490967, "global_step": 38687, "epoch": 434, "lr": 9.517821081003215e-05}
{"train_loss": 0.2709381580352783, "global_step": 38688, "epoch": 434, "lr": 9.517796241728837e-05}
{"train_loss": 0.3535681962966919, "global_step": 38689, "epoch": 434, "lr": 9.517771401847098e-05}
{"train_loss": 0.32529547810554504, "global_step": 38690, "epoch": 434, "lr": 9.517746561357999e-05}
{"train_loss": 0.3452410399913788, "global_step": 38691, "epoch": 434, "lr": 9.517721720261543e-05}
{"train_loss": 0.3652234971523285, "global_step": 38692, "epoch": 434, "lr": 9.517696878557735e-05}
{"train_loss": 0.2543964385986328, "global_step": 38693, "epoch": 434, "lr": 9.517672036246578e-05}
{"train_loss": 0.3183874785900116, "global_step": 38694, "epoch": 434, "lr": 9.517647193328074e-05}
{"train_loss": 0.41542476415634155, "global_step": 38695, "epoch": 434, "lr": 9.517622349802229e-05}
{"train_loss": 0.3110343813896179, "global_step": 38696, "epoch": 434, "lr": 9.517597505669043e-05}
{"train_loss": 0.2842155694961548, "global_step": 38697, "epoch": 434, "lr": 9.517572660928521e-05}
{"train_loss": 0.2562876343727112, "global_step": 38698, "epoch": 434, "lr": 9.517547815580668e-05}
{"train_loss": 0.393707811832428, "global_step": 38699, "epoch": 434, "lr": 9.517522969625485e-05}
{"train_loss": 0.30480289459228516, "global_step": 38700, "epoch": 434, "lr": 9.517498123062976e-05}
{"train_loss": 0.3820490539073944, "global_step": 38701, "epoch": 434, "lr": 9.517473275893145e-05}
{"train_loss": 0.2897580564022064, "global_step": 38702, "epoch": 434, "lr": 9.517448428115994e-05}
{"train_loss": 0.3374454379081726, "global_step": 38703, "epoch": 434, "lr": 9.517423579731527e-05}
{"train_loss": 0.38719919323921204, "global_step": 38704, "epoch": 434, "lr": 9.517398730739748e-05}
{"train_loss": 0.2461252510547638, "global_step": 38705, "epoch": 434, "lr": 9.51737388114066e-05}
{"train_loss": 0.24863983690738678, "global_step": 38706, "epoch": 434, "lr": 9.517349030934267e-05}
{"train_loss": 0.3116922378540039, "global_step": 38707, "epoch": 434, "lr": 9.517324180120569e-05}
{"train_loss": 0.31852301955223083, "global_step": 38708, "epoch": 434, "lr": 9.517299328699573e-05}
{"train_loss": 0.39194434881210327, "global_step": 38709, "epoch": 434, "lr": 9.517274476671282e-05}
{"train_loss": 0.30517733097076416, "global_step": 38710, "epoch": 434, "lr": 9.517249624035698e-05}
{"train_loss": 0.2827831208705902, "global_step": 38711, "epoch": 434, "lr": 9.517224770792826e-05}
{"train_loss": 0.25527068972587585, "global_step": 38712, "epoch": 434, "lr": 9.517199916942665e-05}
{"train_loss": 0.2623019218444824, "global_step": 38713, "epoch": 434, "lr": 9.517175062485225e-05}
{"train_loss": 0.32373742290427177, "global_step": 38714, "epoch": 434, "lr": 9.517150207420503e-05, "val_loss": 2.656536102294922}
{"train_loss": 0.2936006188392639, "global_step": 38715, "epoch": 435, "lr": 9.517125351748505e-05}
{"train_loss": 0.22200362384319305, "global_step": 38716, "epoch": 435, "lr": 9.517100495469238e-05}
{"train_loss": 0.23730938136577606, "global_step": 38717, "epoch": 435, "lr": 9.5170756385827e-05}
{"train_loss": 0.3922748863697052, "global_step": 38718, "epoch": 435, "lr": 9.517050781088896e-05}
{"train_loss": 0.34642359614372253, "global_step": 38719, "epoch": 435, "lr": 9.51702592298783e-05}
{"train_loss": 0.3036101758480072, "global_step": 38720, "epoch": 435, "lr": 9.517001064279505e-05}
{"train_loss": 0.4471453130245209, "global_step": 38721, "epoch": 435, "lr": 9.516976204963924e-05}
{"train_loss": 0.33841076493263245, "global_step": 38722, "epoch": 435, "lr": 9.51695134504109e-05}
{"train_loss": 0.36883246898651123, "global_step": 38723, "epoch": 435, "lr": 9.516926484511006e-05}
{"train_loss": 0.26435041427612305, "global_step": 38724, "epoch": 435, "lr": 9.516901623373679e-05}
{"train_loss": 0.3047247529029846, "global_step": 38725, "epoch": 435, "lr": 9.516876761629108e-05}
{"train_loss": 0.32398125529289246, "global_step": 38726, "epoch": 435, "lr": 9.516851899277299e-05}
{"train_loss": 0.2952754497528076, "global_step": 38727, "epoch": 435, "lr": 9.516827036318253e-05}
{"train_loss": 0.2748236656188965, "global_step": 38728, "epoch": 435, "lr": 9.516802172751975e-05}
{"train_loss": 0.32247889041900635, "global_step": 38729, "epoch": 435, "lr": 9.516777308578468e-05}
{"train_loss": 0.25659191608428955, "global_step": 38730, "epoch": 435, "lr": 9.516752443797736e-05}
{"train_loss": 0.29145991802215576, "global_step": 38731, "epoch": 435, "lr": 9.516727578409782e-05}
{"train_loss": 0.29631105065345764, "global_step": 38732, "epoch": 435, "lr": 9.516702712414607e-05}
{"train_loss": 0.3321307301521301, "global_step": 38733, "epoch": 435, "lr": 9.516677845812219e-05}
{"train_loss": 0.25626975297927856, "global_step": 38734, "epoch": 435, "lr": 9.516652978602618e-05}
{"train_loss": 0.4057435691356659, "global_step": 38735, "epoch": 435, "lr": 9.516628110785808e-05}
{"train_loss": 0.36502423882484436, "global_step": 38736, "epoch": 435, "lr": 9.516603242361791e-05}
{"train_loss": 0.2484084516763687, "global_step": 38737, "epoch": 435, "lr": 9.516578373330574e-05}
{"train_loss": 0.3878386616706848, "global_step": 38738, "epoch": 435, "lr": 9.516553503692156e-05}
{"train_loss": 0.3701218068599701, "global_step": 38739, "epoch": 435, "lr": 9.516528633446542e-05}
{"train_loss": 0.27579447627067566, "global_step": 38740, "epoch": 435, "lr": 9.516503762593739e-05}
{"train_loss": 0.33148643374443054, "global_step": 38741, "epoch": 435, "lr": 9.516478891133745e-05}
{"train_loss": 0.409127801656723, "global_step": 38742, "epoch": 435, "lr": 9.516454019066564e-05}
{"train_loss": 0.3835758864879608, "global_step": 38743, "epoch": 435, "lr": 9.516429146392203e-05}
{"train_loss": 0.30324506759643555, "global_step": 38744, "epoch": 435, "lr": 9.516404273110662e-05}
{"train_loss": 0.19338881969451904, "global_step": 38745, "epoch": 435, "lr": 9.516379399221946e-05}
{"train_loss": 0.23781684041023254, "global_step": 38746, "epoch": 435, "lr": 9.516354524726059e-05}
{"train_loss": 0.3994271755218506, "global_step": 38747, "epoch": 435, "lr": 9.516329649623002e-05}
{"train_loss": 0.2575882375240326, "global_step": 38748, "epoch": 435, "lr": 9.516304773912779e-05}
{"train_loss": 0.38526058197021484, "global_step": 38749, "epoch": 435, "lr": 9.516279897595395e-05}
{"train_loss": 0.4056280851364136, "global_step": 38750, "epoch": 435, "lr": 9.51625502067085e-05}
{"train_loss": 0.3444960117340088, "global_step": 38751, "epoch": 435, "lr": 9.51623014313915e-05}
{"train_loss": 0.3217926621437073, "global_step": 38752, "epoch": 435, "lr": 9.5162052650003e-05}
{"train_loss": 0.32235631346702576, "global_step": 38753, "epoch": 435, "lr": 9.516180386254301e-05}
{"train_loss": 0.4148268401622772, "global_step": 38754, "epoch": 435, "lr": 9.516155506901157e-05}
{"train_loss": 0.29114335775375366, "global_step": 38755, "epoch": 435, "lr": 9.516130626940868e-05}
{"train_loss": 0.33777594566345215, "global_step": 38756, "epoch": 435, "lr": 9.516105746373443e-05}
{"train_loss": 0.3761144280433655, "global_step": 38757, "epoch": 435, "lr": 9.516080865198883e-05}
{"train_loss": 0.3488456606864929, "global_step": 38758, "epoch": 435, "lr": 9.516055983417189e-05}
{"train_loss": 0.3349338173866272, "global_step": 38759, "epoch": 435, "lr": 9.516031101028367e-05}
{"train_loss": 0.3025299608707428, "global_step": 38760, "epoch": 435, "lr": 9.51600621803242e-05}
{"train_loss": 0.3809213638305664, "global_step": 38761, "epoch": 435, "lr": 9.51598133442935e-05}
{"train_loss": 0.34662628173828125, "global_step": 38762, "epoch": 435, "lr": 9.515956450219163e-05}
{"train_loss": 0.327765554189682, "global_step": 38763, "epoch": 435, "lr": 9.515931565401859e-05}
{"train_loss": 0.38615724444389343, "global_step": 38764, "epoch": 435, "lr": 9.515906679977444e-05}
{"train_loss": 0.4717675745487213, "global_step": 38765, "epoch": 435, "lr": 9.51588179394592e-05}
{"train_loss": 0.47588491439819336, "global_step": 38766, "epoch": 435, "lr": 9.51585690730729e-05}
{"train_loss": 0.35120856761932373, "global_step": 38767, "epoch": 435, "lr": 9.515832020061559e-05}
{"train_loss": 0.3557029068470001, "global_step": 38768, "epoch": 435, "lr": 9.51580713220873e-05}
{"train_loss": 0.2910967767238617, "global_step": 38769, "epoch": 435, "lr": 9.515782243748805e-05}
{"train_loss": 0.2943114638328552, "global_step": 38770, "epoch": 435, "lr": 9.515757354681789e-05}
{"train_loss": 0.3457016944885254, "global_step": 38771, "epoch": 435, "lr": 9.515732465007682e-05}
{"train_loss": 0.31390705704689026, "global_step": 38772, "epoch": 435, "lr": 9.515707574726493e-05}
{"train_loss": 0.38874486088752747, "global_step": 38773, "epoch": 435, "lr": 9.515682683838221e-05}
{"train_loss": 0.29407790303230286, "global_step": 38774, "epoch": 435, "lr": 9.51565779234287e-05}
{"train_loss": 0.3226322829723358, "global_step": 38775, "epoch": 435, "lr": 9.515632900240444e-05}
{"train_loss": 0.371609091758728, "global_step": 38776, "epoch": 435, "lr": 9.515608007530947e-05}
{"train_loss": 0.3465002477169037, "global_step": 38777, "epoch": 435, "lr": 9.51558311421438e-05}
{"train_loss": 0.37980619072914124, "global_step": 38778, "epoch": 435, "lr": 9.515558220290749e-05}
{"train_loss": 0.34053927659988403, "global_step": 38779, "epoch": 435, "lr": 9.515533325760056e-05}
{"train_loss": 0.3432674705982208, "global_step": 38780, "epoch": 435, "lr": 9.515508430622305e-05}
{"train_loss": 0.3477191925048828, "global_step": 38781, "epoch": 435, "lr": 9.515483534877498e-05}
{"train_loss": 0.35078364610671997, "global_step": 38782, "epoch": 435, "lr": 9.515458638525641e-05}
{"train_loss": 0.3048911392688751, "global_step": 38783, "epoch": 435, "lr": 9.515433741566732e-05}
{"train_loss": 0.2838764190673828, "global_step": 38784, "epoch": 435, "lr": 9.515408844000782e-05}
{"train_loss": 0.3241689205169678, "global_step": 38785, "epoch": 435, "lr": 9.515383945827788e-05}
{"train_loss": 0.27682259678840637, "global_step": 38786, "epoch": 435, "lr": 9.515359047047755e-05}
{"train_loss": 0.2671836018562317, "global_step": 38787, "epoch": 435, "lr": 9.515334147660691e-05}
{"train_loss": 0.35711005330085754, "global_step": 38788, "epoch": 435, "lr": 9.515309247666592e-05}
{"train_loss": 0.3696497976779938, "global_step": 38789, "epoch": 435, "lr": 9.515284347065465e-05}
{"train_loss": 0.31545040011405945, "global_step": 38790, "epoch": 435, "lr": 9.515259445857315e-05}
{"train_loss": 0.2695525884628296, "global_step": 38791, "epoch": 435, "lr": 9.51523454404214e-05}
{"train_loss": 0.27046021819114685, "global_step": 38792, "epoch": 435, "lr": 9.515209641619949e-05}
{"train_loss": 0.2365759313106537, "global_step": 38793, "epoch": 435, "lr": 9.515184738590744e-05}
{"train_loss": 0.2613644301891327, "global_step": 38794, "epoch": 435, "lr": 9.515159834954524e-05}
{"train_loss": 0.30070415139198303, "global_step": 38795, "epoch": 435, "lr": 9.515134930711298e-05}
{"train_loss": 0.3370949923992157, "global_step": 38796, "epoch": 435, "lr": 9.515110025861067e-05}
{"train_loss": 0.32947197556495667, "global_step": 38797, "epoch": 435, "lr": 9.515085120403835e-05}
{"train_loss": 0.30351561307907104, "global_step": 38798, "epoch": 435, "lr": 9.515060214339604e-05}
{"train_loss": 0.3456622362136841, "global_step": 38799, "epoch": 435, "lr": 9.515035307668378e-05}
{"train_loss": 0.3035000264644623, "global_step": 38800, "epoch": 435, "lr": 9.515010400390162e-05}
{"train_loss": 0.3540327250957489, "global_step": 38801, "epoch": 435, "lr": 9.514985492504957e-05}
{"train_loss": 0.33105558156967163, "global_step": 38802, "epoch": 435, "lr": 9.514960584012768e-05}
{"train_loss": 0.32821267123302716, "global_step": 38803, "epoch": 435, "lr": 9.514935674913596e-05, "val_loss": 2.7808799743652344, "train_action_mse_error": 49.819732666015625}
{"train_loss": 0.2989673614501953, "global_step": 38804, "epoch": 436, "lr": 9.514910765207446e-05}
{"train_loss": 0.24639800190925598, "global_step": 38805, "epoch": 436, "lr": 9.514885854894322e-05}
{"train_loss": 0.3720475137233734, "global_step": 38806, "epoch": 436, "lr": 9.514860943974227e-05}
{"train_loss": 0.24909326434135437, "global_step": 38807, "epoch": 436, "lr": 9.514836032447164e-05}
{"train_loss": 0.37039411067962646, "global_step": 38808, "epoch": 436, "lr": 9.514811120313136e-05}
{"train_loss": 0.2356417030096054, "global_step": 38809, "epoch": 436, "lr": 9.514786207572147e-05}
{"train_loss": 0.34329742193222046, "global_step": 38810, "epoch": 436, "lr": 9.5147612942242e-05}
{"train_loss": 0.2757384479045868, "global_step": 38811, "epoch": 436, "lr": 9.514736380269298e-05}
{"train_loss": 0.5166684985160828, "global_step": 38812, "epoch": 436, "lr": 9.514711465707445e-05}
{"train_loss": 0.32674482464790344, "global_step": 38813, "epoch": 436, "lr": 9.514686550538642e-05}
{"train_loss": 0.28780943155288696, "global_step": 38814, "epoch": 436, "lr": 9.514661634762897e-05}
{"train_loss": 0.261590838432312, "global_step": 38815, "epoch": 436, "lr": 9.51463671838021e-05}
{"train_loss": 0.2904728055000305, "global_step": 38816, "epoch": 436, "lr": 9.514611801390584e-05}
{"train_loss": 0.23687484860420227, "global_step": 38817, "epoch": 436, "lr": 9.514586883794026e-05}
{"train_loss": 0.3157449960708618, "global_step": 38818, "epoch": 436, "lr": 9.514561965590535e-05}
{"train_loss": 0.3349624276161194, "global_step": 38819, "epoch": 436, "lr": 9.514537046780117e-05}
{"train_loss": 0.2716452181339264, "global_step": 38820, "epoch": 436, "lr": 9.514512127362773e-05}
{"train_loss": 0.4734008312225342, "global_step": 38821, "epoch": 436, "lr": 9.51448720733851e-05}
{"train_loss": 0.41131120920181274, "global_step": 38822, "epoch": 436, "lr": 9.514462286707328e-05}
{"train_loss": 0.34157639741897583, "global_step": 38823, "epoch": 436, "lr": 9.51443736546923e-05}
{"train_loss": 0.2907986044883728, "global_step": 38824, "epoch": 436, "lr": 9.514412443624223e-05}
{"train_loss": 0.37423139810562134, "global_step": 38825, "epoch": 436, "lr": 9.514387521172309e-05}
{"train_loss": 0.35463330149650574, "global_step": 38826, "epoch": 436, "lr": 9.514362598113488e-05}
{"train_loss": 0.4052741825580597, "global_step": 38827, "epoch": 436, "lr": 9.514337674447769e-05}
{"train_loss": 0.3972674310207367, "global_step": 38828, "epoch": 436, "lr": 9.51431275017515e-05}
{"train_loss": 0.30037006735801697, "global_step": 38829, "epoch": 436, "lr": 9.514287825295638e-05}
{"train_loss": 0.2706073522567749, "global_step": 38830, "epoch": 436, "lr": 9.514262899809235e-05}
{"train_loss": 0.38778913021087646, "global_step": 38831, "epoch": 436, "lr": 9.514237973715943e-05}
{"train_loss": 0.2831762433052063, "global_step": 38832, "epoch": 436, "lr": 9.514213047015767e-05}
{"train_loss": 0.304503470659256, "global_step": 38833, "epoch": 436, "lr": 9.514188119708712e-05}
{"train_loss": 0.30847886204719543, "global_step": 38834, "epoch": 436, "lr": 9.514163191794778e-05}
{"train_loss": 0.3435218036174774, "global_step": 38835, "epoch": 436, "lr": 9.51413826327397e-05}
{"train_loss": 0.27916789054870605, "global_step": 38836, "epoch": 436, "lr": 9.51411333414629e-05}
{"train_loss": 0.30826759338378906, "global_step": 38837, "epoch": 436, "lr": 9.514088404411745e-05}
{"train_loss": 0.3879193663597107, "global_step": 38838, "epoch": 436, "lr": 9.514063474070334e-05}
{"train_loss": 0.2675609588623047, "global_step": 38839, "epoch": 436, "lr": 9.514038543122062e-05}
{"train_loss": 0.4309212267398834, "global_step": 38840, "epoch": 436, "lr": 9.514013611566934e-05}
{"train_loss": 0.3950803577899933, "global_step": 38841, "epoch": 436, "lr": 9.51398867940495e-05}
{"train_loss": 0.2973100244998932, "global_step": 38842, "epoch": 436, "lr": 9.513963746636116e-05}
{"train_loss": 0.27186545729637146, "global_step": 38843, "epoch": 436, "lr": 9.513938813260434e-05}
{"train_loss": 0.40644967555999756, "global_step": 38844, "epoch": 436, "lr": 9.513913879277911e-05}
{"train_loss": 0.26034262776374817, "global_step": 38845, "epoch": 436, "lr": 9.513888944688544e-05}
{"train_loss": 0.3699958622455597, "global_step": 38846, "epoch": 436, "lr": 9.513864009492342e-05}
{"train_loss": 0.3726647198200226, "global_step": 38847, "epoch": 436, "lr": 9.513839073689303e-05}
{"train_loss": 0.28348472714424133, "global_step": 38848, "epoch": 436, "lr": 9.513814137279435e-05}
{"train_loss": 0.3307969868183136, "global_step": 38849, "epoch": 436, "lr": 9.513789200262741e-05}
{"train_loss": 0.31882795691490173, "global_step": 38850, "epoch": 436, "lr": 9.513764262639221e-05}
{"train_loss": 0.2940825819969177, "global_step": 38851, "epoch": 436, "lr": 9.513739324408882e-05}
{"train_loss": 0.3080182671546936, "global_step": 38852, "epoch": 436, "lr": 9.513714385571724e-05}
{"train_loss": 0.34447333216667175, "global_step": 38853, "epoch": 436, "lr": 9.513689446127754e-05}
{"train_loss": 0.3758343458175659, "global_step": 38854, "epoch": 436, "lr": 9.513664506076971e-05}
{"train_loss": 0.3235650658607483, "global_step": 38855, "epoch": 436, "lr": 9.513639565419383e-05}
{"train_loss": 0.3827834725379944, "global_step": 38856, "epoch": 436, "lr": 9.513614624154991e-05}
{"train_loss": 0.3459062874317169, "global_step": 38857, "epoch": 436, "lr": 9.513589682283797e-05}
{"train_loss": 0.24352414906024933, "global_step": 38858, "epoch": 436, "lr": 9.513564739805808e-05}
{"train_loss": 0.3481763005256653, "global_step": 38859, "epoch": 436, "lr": 9.513539796721024e-05}
{"train_loss": 0.4061782956123352, "global_step": 38860, "epoch": 436, "lr": 9.51351485302945e-05}
{"train_loss": 0.23277990520000458, "global_step": 38861, "epoch": 436, "lr": 9.513489908731088e-05}
{"train_loss": 0.3051491379737854, "global_step": 38862, "epoch": 436, "lr": 9.513464963825943e-05}
{"train_loss": 0.32741230726242065, "global_step": 38863, "epoch": 436, "lr": 9.513440018314018e-05}
{"train_loss": 0.3910951316356659, "global_step": 38864, "epoch": 436, "lr": 9.513415072195315e-05}
{"train_loss": 0.3462044596672058, "global_step": 38865, "epoch": 436, "lr": 9.513390125469839e-05}
{"train_loss": 0.40510526299476624, "global_step": 38866, "epoch": 436, "lr": 9.513365178137594e-05}
{"train_loss": 0.3271108567714691, "global_step": 38867, "epoch": 436, "lr": 9.513340230198579e-05}
{"train_loss": 0.3673771619796753, "global_step": 38868, "epoch": 436, "lr": 9.513315281652804e-05}
{"train_loss": 0.2611273229122162, "global_step": 38869, "epoch": 436, "lr": 9.513290332500266e-05}
{"train_loss": 0.354145884513855, "global_step": 38870, "epoch": 436, "lr": 9.513265382740972e-05}
{"train_loss": 0.32729867100715637, "global_step": 38871, "epoch": 436, "lr": 9.513240432374926e-05}
{"train_loss": 0.36507707834243774, "global_step": 38872, "epoch": 436, "lr": 9.513215481402127e-05}
{"train_loss": 0.33510181307792664, "global_step": 38873, "epoch": 436, "lr": 9.513190529822584e-05}
{"train_loss": 0.44181591272354126, "global_step": 38874, "epoch": 436, "lr": 9.513165577636296e-05}
{"train_loss": 0.36888206005096436, "global_step": 38875, "epoch": 436, "lr": 9.513140624843266e-05}
{"train_loss": 0.27466922998428345, "global_step": 38876, "epoch": 436, "lr": 9.513115671443502e-05}
{"train_loss": 0.35389259457588196, "global_step": 38877, "epoch": 436, "lr": 9.513090717437004e-05}
{"train_loss": 0.3513432741165161, "global_step": 38878, "epoch": 436, "lr": 9.513065762823776e-05}
{"train_loss": 0.28137263655662537, "global_step": 38879, "epoch": 436, "lr": 9.513040807603822e-05}
{"train_loss": 0.36365869641304016, "global_step": 38880, "epoch": 436, "lr": 9.513015851777142e-05}
{"train_loss": 0.365696519613266, "global_step": 38881, "epoch": 436, "lr": 9.512990895343745e-05}
{"train_loss": 0.33146563172340393, "global_step": 38882, "epoch": 436, "lr": 9.512965938303631e-05}
{"train_loss": 0.32708674669265747, "global_step": 38883, "epoch": 436, "lr": 9.512940980656801e-05}
{"train_loss": 0.3773384988307953, "global_step": 38884, "epoch": 436, "lr": 9.512916022403264e-05}
{"train_loss": 0.47073107957839966, "global_step": 38885, "epoch": 436, "lr": 9.51289106354302e-05}
{"train_loss": 0.3154047727584839, "global_step": 38886, "epoch": 436, "lr": 9.512866104076072e-05}
{"train_loss": 0.23941653966903687, "global_step": 38887, "epoch": 436, "lr": 9.512841144002425e-05}
{"train_loss": 0.2812953591346741, "global_step": 38888, "epoch": 436, "lr": 9.51281618332208e-05}
{"train_loss": 0.381587952375412, "global_step": 38889, "epoch": 436, "lr": 9.512791222035042e-05}
{"train_loss": 0.34631574153900146, "global_step": 38890, "epoch": 436, "lr": 9.512766260141318e-05}
{"train_loss": 0.2742948532104492, "global_step": 38891, "epoch": 436, "lr": 9.512741297640904e-05}
{"train_loss": 0.33189151410976153, "global_step": 38892, "epoch": 436, "lr": 9.512716334533807e-05, "val_loss": 2.6626572608947754}
{"train_loss": 0.45040789246559143, "global_step": 38893, "epoch": 437, "lr": 9.512691370820031e-05}
{"train_loss": 0.30352911353111267, "global_step": 38894, "epoch": 437, "lr": 9.512666406499578e-05}
{"train_loss": 0.3118433356285095, "global_step": 38895, "epoch": 437, "lr": 9.512641441572451e-05}
{"train_loss": 0.2748136818408966, "global_step": 38896, "epoch": 437, "lr": 9.512616476038656e-05}
{"train_loss": 0.40293705463409424, "global_step": 38897, "epoch": 437, "lr": 9.512591509898195e-05}
{"train_loss": 0.4088937044143677, "global_step": 38898, "epoch": 437, "lr": 9.512566543151071e-05}
{"train_loss": 0.38872137665748596, "global_step": 38899, "epoch": 437, "lr": 9.512541575797287e-05}
{"train_loss": 0.3063800632953644, "global_step": 38900, "epoch": 437, "lr": 9.512516607836847e-05}
{"train_loss": 0.366466224193573, "global_step": 38901, "epoch": 437, "lr": 9.512491639269754e-05}
{"train_loss": 0.2467355579137802, "global_step": 38902, "epoch": 437, "lr": 9.512466670096012e-05}
{"train_loss": 0.3406791687011719, "global_step": 38903, "epoch": 437, "lr": 9.512441700315624e-05}
{"train_loss": 0.3873383700847626, "global_step": 38904, "epoch": 437, "lr": 9.512416729928592e-05}
{"train_loss": 0.2797095775604248, "global_step": 38905, "epoch": 437, "lr": 9.51239175893492e-05}
{"train_loss": 0.33111000061035156, "global_step": 38906, "epoch": 437, "lr": 9.512366787334615e-05}
{"train_loss": 0.3054656982421875, "global_step": 38907, "epoch": 437, "lr": 9.512341815127674e-05}
{"train_loss": 0.37835368514060974, "global_step": 38908, "epoch": 437, "lr": 9.512316842314107e-05}
{"train_loss": 0.323677659034729, "global_step": 38909, "epoch": 437, "lr": 9.51229186889391e-05}
{"train_loss": 0.3325720429420471, "global_step": 38910, "epoch": 437, "lr": 9.512266894867093e-05}
{"train_loss": 0.25989219546318054, "global_step": 38911, "epoch": 437, "lr": 9.512241920233656e-05}
{"train_loss": 0.33218643069267273, "global_step": 38912, "epoch": 437, "lr": 9.512216944993603e-05}
{"train_loss": 0.3566933274269104, "global_step": 38913, "epoch": 437, "lr": 9.512191969146937e-05}
{"train_loss": 0.26927220821380615, "global_step": 38914, "epoch": 437, "lr": 9.512166992693662e-05}
{"train_loss": 0.3124648630619049, "global_step": 38915, "epoch": 437, "lr": 9.512142015633782e-05}
{"train_loss": 0.4008099138736725, "global_step": 38916, "epoch": 437, "lr": 9.512117037967299e-05}
{"train_loss": 0.2642052471637726, "global_step": 38917, "epoch": 437, "lr": 9.512092059694218e-05}
{"train_loss": 0.30588483810424805, "global_step": 38918, "epoch": 437, "lr": 9.51206708081454e-05}
{"train_loss": 0.35829052329063416, "global_step": 38919, "epoch": 437, "lr": 9.512042101328269e-05}
{"train_loss": 0.33877769112586975, "global_step": 38920, "epoch": 437, "lr": 9.51201712123541e-05}
{"train_loss": 0.345492959022522, "global_step": 38921, "epoch": 437, "lr": 9.511992140535965e-05}
{"train_loss": 0.3506452441215515, "global_step": 38922, "epoch": 437, "lr": 9.511967159229937e-05}
{"train_loss": 0.401614248752594, "global_step": 38923, "epoch": 437, "lr": 9.51194217731733e-05}
{"train_loss": 0.26372814178466797, "global_step": 38924, "epoch": 437, "lr": 9.511917194798148e-05}
{"train_loss": 0.31772977113723755, "global_step": 38925, "epoch": 437, "lr": 9.511892211672394e-05}
{"train_loss": 0.3621481657028198, "global_step": 38926, "epoch": 437, "lr": 9.51186722794007e-05}
{"train_loss": 0.24139806628227234, "global_step": 38927, "epoch": 437, "lr": 9.511842243601181e-05}
{"train_loss": 0.41247016191482544, "global_step": 38928, "epoch": 437, "lr": 9.51181725865573e-05}
{"train_loss": 0.238784521818161, "global_step": 38929, "epoch": 437, "lr": 9.511792273103721e-05}
{"train_loss": 0.3460198938846588, "global_step": 38930, "epoch": 437, "lr": 9.511767286945155e-05}
{"train_loss": 0.31130197644233704, "global_step": 38931, "epoch": 437, "lr": 9.511742300180038e-05}
{"train_loss": 0.3711794912815094, "global_step": 38932, "epoch": 437, "lr": 9.511717312808372e-05}
{"train_loss": 0.2828805148601532, "global_step": 38933, "epoch": 437, "lr": 9.51169232483016e-05}
{"train_loss": 0.48016905784606934, "global_step": 38934, "epoch": 437, "lr": 9.511667336245407e-05}
{"train_loss": 0.3221205472946167, "global_step": 38935, "epoch": 437, "lr": 9.511642347054115e-05}
{"train_loss": 0.2904636859893799, "global_step": 38936, "epoch": 437, "lr": 9.511617357256286e-05}
{"train_loss": 0.3142807185649872, "global_step": 38937, "epoch": 437, "lr": 9.511592366851927e-05}
{"train_loss": 0.3756393790245056, "global_step": 38938, "epoch": 437, "lr": 9.511567375841038e-05}
{"train_loss": 0.24276769161224365, "global_step": 38939, "epoch": 437, "lr": 9.511542384223625e-05}
{"train_loss": 0.3118034899234772, "global_step": 38940, "epoch": 437, "lr": 9.51151739199969e-05}
{"train_loss": 0.28621819615364075, "global_step": 38941, "epoch": 437, "lr": 9.511492399169235e-05}
{"train_loss": 0.2898021340370178, "global_step": 38942, "epoch": 437, "lr": 9.511467405732266e-05}
{"train_loss": 0.32992109656333923, "global_step": 38943, "epoch": 437, "lr": 9.511442411688786e-05}
{"train_loss": 0.24995319545269012, "global_step": 38944, "epoch": 437, "lr": 9.511417417038798e-05}
{"train_loss": 0.33373183012008667, "global_step": 38945, "epoch": 437, "lr": 9.511392421782301e-05}
{"train_loss": 0.2494141161441803, "global_step": 38946, "epoch": 437, "lr": 9.511367425919307e-05}
{"train_loss": 0.31753385066986084, "global_step": 38947, "epoch": 437, "lr": 9.51134242944981e-05}
{"train_loss": 0.3018026053905487, "global_step": 38948, "epoch": 437, "lr": 9.511317432373822e-05}
{"train_loss": 0.24789009988307953, "global_step": 38949, "epoch": 437, "lr": 9.511292434691342e-05}
{"train_loss": 0.20541466772556305, "global_step": 38950, "epoch": 437, "lr": 9.511267436402371e-05}
{"train_loss": 0.3129817843437195, "global_step": 38951, "epoch": 437, "lr": 9.511242437506916e-05}
{"train_loss": 0.30942586064338684, "global_step": 38952, "epoch": 437, "lr": 9.51121743800498e-05}
{"train_loss": 0.39967966079711914, "global_step": 38953, "epoch": 437, "lr": 9.511192437896567e-05}
{"train_loss": 0.3090590834617615, "global_step": 38954, "epoch": 437, "lr": 9.511167437181679e-05}
{"train_loss": 0.32915744185447693, "global_step": 38955, "epoch": 437, "lr": 9.511142435860318e-05}
{"train_loss": 0.3180733025074005, "global_step": 38956, "epoch": 437, "lr": 9.511117433932488e-05}
{"train_loss": 0.30220910906791687, "global_step": 38957, "epoch": 437, "lr": 9.511092431398195e-05}
{"train_loss": 0.2891497015953064, "global_step": 38958, "epoch": 437, "lr": 9.511067428257441e-05}
{"train_loss": 0.45357489585876465, "global_step": 38959, "epoch": 437, "lr": 9.511042424510228e-05}
{"train_loss": 0.3673098087310791, "global_step": 38960, "epoch": 437, "lr": 9.511017420156561e-05}
{"train_loss": 0.4283841848373413, "global_step": 38961, "epoch": 437, "lr": 9.510992415196441e-05}
{"train_loss": 0.3479427993297577, "global_step": 38962, "epoch": 437, "lr": 9.510967409629875e-05}
{"train_loss": 0.2518596351146698, "global_step": 38963, "epoch": 437, "lr": 9.510942403456863e-05}
{"train_loss": 0.3168751001358032, "global_step": 38964, "epoch": 437, "lr": 9.51091739667741e-05}
{"train_loss": 0.3827340602874756, "global_step": 38965, "epoch": 437, "lr": 9.510892389291521e-05}
{"train_loss": 0.3049769997596741, "global_step": 38966, "epoch": 437, "lr": 9.510867381299196e-05}
{"train_loss": 0.3450584411621094, "global_step": 38967, "epoch": 437, "lr": 9.51084237270044e-05}
{"train_loss": 0.3621678352355957, "global_step": 38968, "epoch": 437, "lr": 9.510817363495256e-05}
{"train_loss": 0.23777909576892853, "global_step": 38969, "epoch": 437, "lr": 9.510792353683648e-05}
{"train_loss": 0.37726736068725586, "global_step": 38970, "epoch": 437, "lr": 9.510767343265618e-05}
{"train_loss": 0.27551332116127014, "global_step": 38971, "epoch": 437, "lr": 9.510742332241172e-05}
{"train_loss": 0.3715834319591522, "global_step": 38972, "epoch": 437, "lr": 9.51071732061031e-05}
{"train_loss": 0.32718196511268616, "global_step": 38973, "epoch": 437, "lr": 9.510692308373038e-05}
{"train_loss": 0.343112587928772, "global_step": 38974, "epoch": 437, "lr": 9.510667295529359e-05}
{"train_loss": 0.3524564802646637, "global_step": 38975, "epoch": 437, "lr": 9.510642282079274e-05}
{"train_loss": 0.315753698348999, "global_step": 38976, "epoch": 437, "lr": 9.51061726802279e-05}
{"train_loss": 0.31755951046943665, "global_step": 38977, "epoch": 437, "lr": 9.510592253359906e-05}
{"train_loss": 0.24111151695251465, "global_step": 38978, "epoch": 437, "lr": 9.51056723809063e-05}
{"train_loss": 0.3168964087963104, "global_step": 38979, "epoch": 437, "lr": 9.510542222214962e-05}
{"train_loss": 0.2691766917705536, "global_step": 38980, "epoch": 437, "lr": 9.510517205732907e-05}
{"train_loss": 0.32418012033017835, "global_step": 38981, "epoch": 437, "lr": 9.510492188644468e-05, "val_loss": 2.5918021202087402}
{"train_loss": 0.34576454758644104, "global_step": 38982, "epoch": 438, "lr": 9.51046717094965e-05}
{"train_loss": 0.26135003566741943, "global_step": 38983, "epoch": 438, "lr": 9.510442152648454e-05}
{"train_loss": 0.3546964228153229, "global_step": 38984, "epoch": 438, "lr": 9.510417133740882e-05}
{"train_loss": 0.20362509787082672, "global_step": 38985, "epoch": 438, "lr": 9.510392114226941e-05}
{"train_loss": 0.28002050518989563, "global_step": 38986, "epoch": 438, "lr": 9.510367094106631e-05}
{"train_loss": 0.3398467004299164, "global_step": 38987, "epoch": 438, "lr": 9.510342073379958e-05}
{"train_loss": 0.2910759449005127, "global_step": 38988, "epoch": 438, "lr": 9.510317052046926e-05}
{"train_loss": 0.24997861683368683, "global_step": 38989, "epoch": 438, "lr": 9.510292030107535e-05}
{"train_loss": 0.3645254671573639, "global_step": 38990, "epoch": 438, "lr": 9.51026700756179e-05}
{"train_loss": 0.3067868947982788, "global_step": 38991, "epoch": 438, "lr": 9.510241984409697e-05}
{"train_loss": 0.34333398938179016, "global_step": 38992, "epoch": 438, "lr": 9.510216960651255e-05}
{"train_loss": 0.396205872297287, "global_step": 38993, "epoch": 438, "lr": 9.51019193628647e-05}
{"train_loss": 0.4162207543849945, "global_step": 38994, "epoch": 438, "lr": 9.510166911315343e-05}
{"train_loss": 0.30992603302001953, "global_step": 38995, "epoch": 438, "lr": 9.510141885737882e-05}
{"train_loss": 0.3149363100528717, "global_step": 38996, "epoch": 438, "lr": 9.510116859554086e-05}
{"train_loss": 0.3185747265815735, "global_step": 38997, "epoch": 438, "lr": 9.51009183276396e-05}
{"train_loss": 0.41894978284835815, "global_step": 38998, "epoch": 438, "lr": 9.510066805367506e-05}
{"train_loss": 0.32548072934150696, "global_step": 38999, "epoch": 438, "lr": 9.510041777364728e-05}
{"train_loss": 0.29972827434539795, "global_step": 39000, "epoch": 438, "lr": 9.510016748755633e-05}
{"train_loss": 0.40857428312301636, "global_step": 39001, "epoch": 438, "lr": 9.509991719540218e-05}
{"train_loss": 0.3042905330657959, "global_step": 39002, "epoch": 438, "lr": 9.50996668971849e-05}
{"train_loss": 0.34258830547332764, "global_step": 39003, "epoch": 438, "lr": 9.509941659290453e-05}
{"train_loss": 0.2724999487400055, "global_step": 39004, "epoch": 438, "lr": 9.509916628256109e-05}
{"train_loss": 0.3098185062408447, "global_step": 39005, "epoch": 438, "lr": 9.509891596615463e-05}
{"train_loss": 0.28433525562286377, "global_step": 39006, "epoch": 438, "lr": 9.509866564368515e-05}
{"train_loss": 0.34690651297569275, "global_step": 39007, "epoch": 438, "lr": 9.509841531515271e-05}
{"train_loss": 0.47998636960983276, "global_step": 39008, "epoch": 438, "lr": 9.509816498055735e-05}
{"train_loss": 0.3554410934448242, "global_step": 39009, "epoch": 438, "lr": 9.509791463989907e-05}
{"train_loss": 0.33528822660446167, "global_step": 39010, "epoch": 438, "lr": 9.509766429317795e-05}
{"train_loss": 0.41302239894866943, "global_step": 39011, "epoch": 438, "lr": 9.509741394039398e-05}
{"train_loss": 0.3810068964958191, "global_step": 39012, "epoch": 438, "lr": 9.509716358154721e-05}
{"train_loss": 0.3634783625602722, "global_step": 39013, "epoch": 438, "lr": 9.509691321663769e-05}
{"train_loss": 0.4144744277000427, "global_step": 39014, "epoch": 438, "lr": 9.509666284566543e-05}
{"train_loss": 0.2926074266433716, "global_step": 39015, "epoch": 438, "lr": 9.509641246863048e-05}
{"train_loss": 0.3091159760951996, "global_step": 39016, "epoch": 438, "lr": 9.509616208553286e-05}
{"train_loss": 0.4193696677684784, "global_step": 39017, "epoch": 438, "lr": 9.509591169637262e-05}
{"train_loss": 0.2706551253795624, "global_step": 39018, "epoch": 438, "lr": 9.509566130114977e-05}
{"train_loss": 0.4082588255405426, "global_step": 39019, "epoch": 438, "lr": 9.509541089986437e-05}
{"train_loss": 0.37938740849494934, "global_step": 39020, "epoch": 438, "lr": 9.509516049251644e-05}
{"train_loss": 0.39851948618888855, "global_step": 39021, "epoch": 438, "lr": 9.5094910079106e-05}
{"train_loss": 0.45312365889549255, "global_step": 39022, "epoch": 438, "lr": 9.50946596596331e-05}
{"train_loss": 0.34596630930900574, "global_step": 39023, "epoch": 438, "lr": 9.50944092340978e-05}
{"train_loss": 0.5164512991905212, "global_step": 39024, "epoch": 438, "lr": 9.50941588025001e-05}
{"train_loss": 0.3422679305076599, "global_step": 39025, "epoch": 438, "lr": 9.509390836484002e-05}
{"train_loss": 0.39201220870018005, "global_step": 39026, "epoch": 438, "lr": 9.509365792111763e-05}
{"train_loss": 0.34603261947631836, "global_step": 39027, "epoch": 438, "lr": 9.509340747133295e-05}
{"train_loss": 0.4855082631111145, "global_step": 39028, "epoch": 438, "lr": 9.5093157015486e-05}
{"train_loss": 0.4155249297618866, "global_step": 39029, "epoch": 438, "lr": 9.509290655357683e-05}
{"train_loss": 0.3444817364215851, "global_step": 39030, "epoch": 438, "lr": 9.509265608560546e-05}
{"train_loss": 0.32720884680747986, "global_step": 39031, "epoch": 438, "lr": 9.509240561157193e-05}
{"train_loss": 0.35101836919784546, "global_step": 39032, "epoch": 438, "lr": 9.509215513147629e-05}
{"train_loss": 0.3804458975791931, "global_step": 39033, "epoch": 438, "lr": 9.509190464531855e-05}
{"train_loss": 0.2903718650341034, "global_step": 39034, "epoch": 438, "lr": 9.509165415309876e-05}
{"train_loss": 0.41640278697013855, "global_step": 39035, "epoch": 438, "lr": 9.509140365481694e-05}
{"train_loss": 0.3721742033958435, "global_step": 39036, "epoch": 438, "lr": 9.509115315047314e-05}
{"train_loss": 0.30553779006004333, "global_step": 39037, "epoch": 438, "lr": 9.509090264006737e-05}
{"train_loss": 0.26676106452941895, "global_step": 39038, "epoch": 438, "lr": 9.50906521235997e-05}
{"train_loss": 0.31598973274230957, "global_step": 39039, "epoch": 438, "lr": 9.50904016010701e-05}
{"train_loss": 0.3263562023639679, "global_step": 39040, "epoch": 438, "lr": 9.509015107247868e-05}
{"train_loss": 0.4199751317501068, "global_step": 39041, "epoch": 438, "lr": 9.508990053782543e-05}
{"train_loss": 0.22096620500087738, "global_step": 39042, "epoch": 438, "lr": 9.50896499971104e-05}
{"train_loss": 0.30455896258354187, "global_step": 39043, "epoch": 438, "lr": 9.50893994503336e-05}
{"train_loss": 0.3022107779979706, "global_step": 39044, "epoch": 438, "lr": 9.508914889749509e-05}
{"train_loss": 0.23770645260810852, "global_step": 39045, "epoch": 438, "lr": 9.508889833859489e-05}
{"train_loss": 0.274516224861145, "global_step": 39046, "epoch": 438, "lr": 9.508864777363304e-05}
{"train_loss": 0.22678901255130768, "global_step": 39047, "epoch": 438, "lr": 9.508839720260958e-05}
{"train_loss": 0.25908708572387695, "global_step": 39048, "epoch": 438, "lr": 9.508814662552452e-05}
{"train_loss": 0.27415311336517334, "global_step": 39049, "epoch": 438, "lr": 9.508789604237792e-05}
{"train_loss": 0.2682082951068878, "global_step": 39050, "epoch": 438, "lr": 9.50876454531698e-05}
{"train_loss": 0.28905022144317627, "global_step": 39051, "epoch": 438, "lr": 9.508739485790017e-05}
{"train_loss": 0.3823509216308594, "global_step": 39052, "epoch": 438, "lr": 9.508714425656911e-05}
{"train_loss": 0.35984355211257935, "global_step": 39053, "epoch": 438, "lr": 9.508689364917665e-05}
{"train_loss": 0.30824607610702515, "global_step": 39054, "epoch": 438, "lr": 9.508664303572278e-05}
{"train_loss": 0.2681178152561188, "global_step": 39055, "epoch": 438, "lr": 9.508639241620756e-05}
{"train_loss": 0.21052519977092743, "global_step": 39056, "epoch": 438, "lr": 9.508614179063104e-05}
{"train_loss": 0.3063116669654846, "global_step": 39057, "epoch": 438, "lr": 9.508589115899324e-05}
{"train_loss": 0.41828274726867676, "global_step": 39058, "epoch": 438, "lr": 9.508564052129417e-05}
{"train_loss": 0.4040689170360565, "global_step": 39059, "epoch": 438, "lr": 9.50853898775339e-05}
{"train_loss": 0.28658270835876465, "global_step": 39060, "epoch": 438, "lr": 9.508513922771245e-05}
{"train_loss": 0.3555299937725067, "global_step": 39061, "epoch": 438, "lr": 9.508488857182984e-05}
{"train_loss": 0.2938706874847412, "global_step": 39062, "epoch": 438, "lr": 9.508463790988613e-05}
{"train_loss": 0.28098246455192566, "global_step": 39063, "epoch": 438, "lr": 9.508438724188133e-05}
{"train_loss": 0.29421266913414, "global_step": 39064, "epoch": 438, "lr": 9.50841365678155e-05}
{"train_loss": 0.20960485935211182, "global_step": 39065, "epoch": 438, "lr": 9.508388588768864e-05}
{"train_loss": 0.36066195368766785, "global_step": 39066, "epoch": 438, "lr": 9.508363520150081e-05}
{"train_loss": 0.21687805652618408, "global_step": 39067, "epoch": 438, "lr": 9.508338450925203e-05}
{"train_loss": 0.4137505888938904, "global_step": 39068, "epoch": 438, "lr": 9.508313381094233e-05}
{"train_loss": 0.34473204612731934, "global_step": 39069, "epoch": 438, "lr": 9.508288310657178e-05}
{"train_loss": 0.33438870863298353, "global_step": 39070, "epoch": 438, "lr": 9.508263239614037e-05, "val_loss": 2.8005828857421875}
{"train_loss": 0.36275410652160645, "global_step": 39071, "epoch": 439, "lr": 9.508238167964814e-05}
{"train_loss": 0.25241798162460327, "global_step": 39072, "epoch": 439, "lr": 9.508213095709515e-05}
{"train_loss": 0.29213330149650574, "global_step": 39073, "epoch": 439, "lr": 9.508188022848142e-05}
{"train_loss": 0.2699381113052368, "global_step": 39074, "epoch": 439, "lr": 9.508162949380695e-05}
{"train_loss": 0.35338452458381653, "global_step": 39075, "epoch": 439, "lr": 9.508137875307184e-05}
{"train_loss": 0.42519575357437134, "global_step": 39076, "epoch": 439, "lr": 9.508112800627606e-05}
{"train_loss": 0.33776766061782837, "global_step": 39077, "epoch": 439, "lr": 9.508087725341968e-05}
{"train_loss": 0.2502891421318054, "global_step": 39078, "epoch": 439, "lr": 9.508062649450274e-05}
{"train_loss": 0.36613836884498596, "global_step": 39079, "epoch": 439, "lr": 9.508037572952524e-05}
{"train_loss": 0.2629712224006653, "global_step": 39080, "epoch": 439, "lr": 9.508012495848723e-05}
{"train_loss": 0.3219543397426605, "global_step": 39081, "epoch": 439, "lr": 9.507987418138878e-05}
{"train_loss": 0.25867676734924316, "global_step": 39082, "epoch": 439, "lr": 9.507962339822985e-05}
{"train_loss": 0.2959946393966675, "global_step": 39083, "epoch": 439, "lr": 9.507937260901054e-05}
{"train_loss": 0.31519997119903564, "global_step": 39084, "epoch": 439, "lr": 9.507912181373086e-05}
{"train_loss": 0.2503780722618103, "global_step": 39085, "epoch": 439, "lr": 9.507887101239081e-05}
{"train_loss": 0.2768903076648712, "global_step": 39086, "epoch": 439, "lr": 9.507862020499049e-05}
{"train_loss": 0.2851342558860779, "global_step": 39087, "epoch": 439, "lr": 9.507836939152988e-05}
{"train_loss": 0.2992396652698517, "global_step": 39088, "epoch": 439, "lr": 9.507811857200905e-05}
{"train_loss": 0.32415205240249634, "global_step": 39089, "epoch": 439, "lr": 9.5077867746428e-05}
{"train_loss": 0.36944612860679626, "global_step": 39090, "epoch": 439, "lr": 9.507761691478678e-05}
{"train_loss": 0.35218560695648193, "global_step": 39091, "epoch": 439, "lr": 9.507736607708543e-05}
{"train_loss": 0.30413946509361267, "global_step": 39092, "epoch": 439, "lr": 9.507711523332398e-05}
{"train_loss": 0.32974085211753845, "global_step": 39093, "epoch": 439, "lr": 9.507686438350245e-05}
{"train_loss": 0.2859318256378174, "global_step": 39094, "epoch": 439, "lr": 9.50766135276209e-05}
{"train_loss": 0.20206746459007263, "global_step": 39095, "epoch": 439, "lr": 9.507636266567934e-05}
{"train_loss": 0.20565946400165558, "global_step": 39096, "epoch": 439, "lr": 9.507611179767781e-05}
{"train_loss": 0.3187931478023529, "global_step": 39097, "epoch": 439, "lr": 9.507586092361635e-05}
{"train_loss": 0.25371113419532776, "global_step": 39098, "epoch": 439, "lr": 9.507561004349499e-05}
{"train_loss": 0.2692601978778839, "global_step": 39099, "epoch": 439, "lr": 9.507535915731376e-05}
{"train_loss": 0.33421701192855835, "global_step": 39100, "epoch": 439, "lr": 9.507510826507271e-05}
{"train_loss": 0.34711721539497375, "global_step": 39101, "epoch": 439, "lr": 9.507485736677183e-05}
{"train_loss": 0.37522900104522705, "global_step": 39102, "epoch": 439, "lr": 9.507460646241122e-05}
{"train_loss": 0.2970579266548157, "global_step": 39103, "epoch": 439, "lr": 9.507435555199086e-05}
{"train_loss": 0.42808520793914795, "global_step": 39104, "epoch": 439, "lr": 9.50741046355108e-05}
{"train_loss": 0.27151352167129517, "global_step": 39105, "epoch": 439, "lr": 9.50738537129711e-05}
{"train_loss": 0.38076239824295044, "global_step": 39106, "epoch": 439, "lr": 9.507360278437175e-05}
{"train_loss": 0.37293490767478943, "global_step": 39107, "epoch": 439, "lr": 9.50733518497128e-05}
{"train_loss": 0.3554794490337372, "global_step": 39108, "epoch": 439, "lr": 9.507310090899431e-05}
{"train_loss": 0.3133332133293152, "global_step": 39109, "epoch": 439, "lr": 9.507284996221626e-05}
{"train_loss": 0.4032400846481323, "global_step": 39110, "epoch": 439, "lr": 9.507259900937873e-05}
{"train_loss": 0.3148689270019531, "global_step": 39111, "epoch": 439, "lr": 9.507234805048175e-05}
{"train_loss": 0.28234410285949707, "global_step": 39112, "epoch": 439, "lr": 9.507209708552533e-05}
{"train_loss": 0.41386449337005615, "global_step": 39113, "epoch": 439, "lr": 9.507184611450951e-05}
{"train_loss": 0.3235786557197571, "global_step": 39114, "epoch": 439, "lr": 9.507159513743433e-05}
{"train_loss": 0.27609777450561523, "global_step": 39115, "epoch": 439, "lr": 9.507134415429984e-05}
{"train_loss": 0.3839537799358368, "global_step": 39116, "epoch": 439, "lr": 9.507109316510604e-05}
{"train_loss": 0.35000765323638916, "global_step": 39117, "epoch": 439, "lr": 9.507084216985299e-05}
{"train_loss": 0.43753230571746826, "global_step": 39118, "epoch": 439, "lr": 9.507059116854071e-05}
{"train_loss": 0.24609963595867157, "global_step": 39119, "epoch": 439, "lr": 9.507034016116925e-05}
{"train_loss": 0.3551080524921417, "global_step": 39120, "epoch": 439, "lr": 9.507008914773861e-05}
{"train_loss": 0.4355272650718689, "global_step": 39121, "epoch": 439, "lr": 9.506983812824886e-05}
{"train_loss": 0.265715628862381, "global_step": 39122, "epoch": 439, "lr": 9.506958710270001e-05}
{"train_loss": 0.24104377627372742, "global_step": 39123, "epoch": 439, "lr": 9.506933607109212e-05}
{"train_loss": 0.24639838933944702, "global_step": 39124, "epoch": 439, "lr": 9.50690850334252e-05}
{"train_loss": 0.2992265522480011, "global_step": 39125, "epoch": 439, "lr": 9.506883398969929e-05}
{"train_loss": 0.3169573247432709, "global_step": 39126, "epoch": 439, "lr": 9.506858293991442e-05}
{"train_loss": 0.41805651783943176, "global_step": 39127, "epoch": 439, "lr": 9.506833188407063e-05}
{"train_loss": 0.3542490601539612, "global_step": 39128, "epoch": 439, "lr": 9.506808082216796e-05}
{"train_loss": 0.3066023588180542, "global_step": 39129, "epoch": 439, "lr": 9.506782975420643e-05}
{"train_loss": 0.43588072061538696, "global_step": 39130, "epoch": 439, "lr": 9.506757868018608e-05}
{"train_loss": 0.4227331578731537, "global_step": 39131, "epoch": 439, "lr": 9.506732760010693e-05}
{"train_loss": 0.3612818717956543, "global_step": 39132, "epoch": 439, "lr": 9.506707651396905e-05}
{"train_loss": 0.36292585730552673, "global_step": 39133, "epoch": 439, "lr": 9.506682542177244e-05}
{"train_loss": 0.3067648112773895, "global_step": 39134, "epoch": 439, "lr": 9.506657432351713e-05}
{"train_loss": 0.35681143403053284, "global_step": 39135, "epoch": 439, "lr": 9.50663232192032e-05}
{"train_loss": 0.37458688020706177, "global_step": 39136, "epoch": 439, "lr": 9.506607210883063e-05}
{"train_loss": 0.21008500456809998, "global_step": 39137, "epoch": 439, "lr": 9.506582099239948e-05}
{"train_loss": 0.3492388129234314, "global_step": 39138, "epoch": 439, "lr": 9.506556986990978e-05}
{"train_loss": 0.3225014805793762, "global_step": 39139, "epoch": 439, "lr": 9.506531874136156e-05}
{"train_loss": 0.3024667501449585, "global_step": 39140, "epoch": 439, "lr": 9.506506760675487e-05}
{"train_loss": 0.2305753082036972, "global_step": 39141, "epoch": 439, "lr": 9.506481646608971e-05}
{"train_loss": 0.40010085701942444, "global_step": 39142, "epoch": 439, "lr": 9.506456531936616e-05}
{"train_loss": 0.39526188373565674, "global_step": 39143, "epoch": 439, "lr": 9.506431416658421e-05}
{"train_loss": 0.3520362675189972, "global_step": 39144, "epoch": 439, "lr": 9.506406300774392e-05}
{"train_loss": 0.35831692814826965, "global_step": 39145, "epoch": 439, "lr": 9.506381184284532e-05}
{"train_loss": 0.3064161241054535, "global_step": 39146, "epoch": 439, "lr": 9.506356067188843e-05}
{"train_loss": 0.3274262249469757, "global_step": 39147, "epoch": 439, "lr": 9.50633094948733e-05}
{"train_loss": 0.33940109610557556, "global_step": 39148, "epoch": 439, "lr": 9.506305831179996e-05}
{"train_loss": 0.29112738370895386, "global_step": 39149, "epoch": 439, "lr": 9.506280712266842e-05}
{"train_loss": 0.3765568435192108, "global_step": 39150, "epoch": 439, "lr": 9.506255592747875e-05}
{"train_loss": 0.3377458453178406, "global_step": 39151, "epoch": 439, "lr": 9.506230472623096e-05}
{"train_loss": 0.33033159375190735, "global_step": 39152, "epoch": 439, "lr": 9.506205351892512e-05}
{"train_loss": 0.30490630865097046, "global_step": 39153, "epoch": 439, "lr": 9.506180230556121e-05}
{"train_loss": 0.3452867865562439, "global_step": 39154, "epoch": 439, "lr": 9.50615510861393e-05}
{"train_loss": 0.255567729473114, "global_step": 39155, "epoch": 439, "lr": 9.506129986065942e-05}
{"train_loss": 0.35218989849090576, "global_step": 39156, "epoch": 439, "lr": 9.506104862912158e-05}
{"train_loss": 0.4337626099586487, "global_step": 39157, "epoch": 439, "lr": 9.506079739152584e-05}
{"train_loss": 0.2747102975845337, "global_step": 39158, "epoch": 439, "lr": 9.506054614787222e-05}
{"train_loss": 0.32569609281052364, "global_step": 39159, "epoch": 439, "lr": 9.506029489816077e-05, "val_loss": 2.6460070610046387}
{"train_loss": 0.3729420304298401, "global_step": 39160, "epoch": 440, "lr": 9.50600436423915e-05}
{"train_loss": 0.30262845754623413, "global_step": 39161, "epoch": 440, "lr": 9.505979238056446e-05}
{"train_loss": 0.27988919615745544, "global_step": 39162, "epoch": 440, "lr": 9.505954111267969e-05}
{"train_loss": 0.3784255385398865, "global_step": 39163, "epoch": 440, "lr": 9.50592898387372e-05}
{"train_loss": 0.29058799147605896, "global_step": 39164, "epoch": 440, "lr": 9.505903855873704e-05}
{"train_loss": 0.38375458121299744, "global_step": 39165, "epoch": 440, "lr": 9.505878727267926e-05}
{"train_loss": 0.36457526683807373, "global_step": 39166, "epoch": 440, "lr": 9.505853598056386e-05}
{"train_loss": 0.2851516902446747, "global_step": 39167, "epoch": 440, "lr": 9.505828468239088e-05}
{"train_loss": 0.3583177626132965, "global_step": 39168, "epoch": 440, "lr": 9.505803337816037e-05}
{"train_loss": 0.3144363462924957, "global_step": 39169, "epoch": 440, "lr": 9.505778206787237e-05}
{"train_loss": 0.33205926418304443, "global_step": 39170, "epoch": 440, "lr": 9.505753075152688e-05}
{"train_loss": 0.348364919424057, "global_step": 39171, "epoch": 440, "lr": 9.505727942912397e-05}
{"train_loss": 0.296372652053833, "global_step": 39172, "epoch": 440, "lr": 9.505702810066365e-05}
{"train_loss": 0.2445640116930008, "global_step": 39173, "epoch": 440, "lr": 9.505677676614597e-05}
{"train_loss": 0.2766239643096924, "global_step": 39174, "epoch": 440, "lr": 9.505652542557094e-05}
{"train_loss": 0.2946363687515259, "global_step": 39175, "epoch": 440, "lr": 9.505627407893863e-05}
{"train_loss": 0.2916472256183624, "global_step": 39176, "epoch": 440, "lr": 9.505602272624903e-05}
{"train_loss": 0.358835905790329, "global_step": 39177, "epoch": 440, "lr": 9.505577136750221e-05}
{"train_loss": 0.26292744278907776, "global_step": 39178, "epoch": 440, "lr": 9.50555200026982e-05}
{"train_loss": 0.24107781052589417, "global_step": 39179, "epoch": 440, "lr": 9.505526863183701e-05}
{"train_loss": 0.21492597460746765, "global_step": 39180, "epoch": 440, "lr": 9.505501725491868e-05}
{"train_loss": 0.3040142059326172, "global_step": 39181, "epoch": 440, "lr": 9.505476587194326e-05}
{"train_loss": 0.44780001044273376, "global_step": 39182, "epoch": 440, "lr": 9.505451448291078e-05}
{"train_loss": 0.3267194330692291, "global_step": 39183, "epoch": 440, "lr": 9.505426308782128e-05}
{"train_loss": 0.32360443472862244, "global_step": 39184, "epoch": 440, "lr": 9.505401168667477e-05}
{"train_loss": 0.3423810601234436, "global_step": 39185, "epoch": 440, "lr": 9.505376027947131e-05}
{"train_loss": 0.415264368057251, "global_step": 39186, "epoch": 440, "lr": 9.50535088662109e-05}
{"train_loss": 0.31549766659736633, "global_step": 39187, "epoch": 440, "lr": 9.505325744689361e-05}
{"train_loss": 0.27473533153533936, "global_step": 39188, "epoch": 440, "lr": 9.505300602151945e-05}
{"train_loss": 0.36488771438598633, "global_step": 39189, "epoch": 440, "lr": 9.505275459008846e-05}
{"train_loss": 0.3569042682647705, "global_step": 39190, "epoch": 440, "lr": 9.505250315260069e-05}
{"train_loss": 0.30898308753967285, "global_step": 39191, "epoch": 440, "lr": 9.505225170905616e-05}
{"train_loss": 0.30888617038726807, "global_step": 39192, "epoch": 440, "lr": 9.505200025945489e-05}
{"train_loss": 0.33171793818473816, "global_step": 39193, "epoch": 440, "lr": 9.505174880379693e-05}
{"train_loss": 0.23246128857135773, "global_step": 39194, "epoch": 440, "lr": 9.505149734208232e-05}
{"train_loss": 0.2584136426448822, "global_step": 39195, "epoch": 440, "lr": 9.505124587431109e-05}
{"train_loss": 0.31087782979011536, "global_step": 39196, "epoch": 440, "lr": 9.505099440048327e-05}
{"train_loss": 0.4120183289051056, "global_step": 39197, "epoch": 440, "lr": 9.505074292059887e-05}
{"train_loss": 0.2681281864643097, "global_step": 39198, "epoch": 440, "lr": 9.505049143465796e-05}
{"train_loss": 0.2726231813430786, "global_step": 39199, "epoch": 440, "lr": 9.505023994266055e-05}
{"train_loss": 0.32071778178215027, "global_step": 39200, "epoch": 440, "lr": 9.504998844460671e-05}
{"train_loss": 0.26410606503486633, "global_step": 39201, "epoch": 440, "lr": 9.504973694049644e-05}
{"train_loss": 0.36177146434783936, "global_step": 39202, "epoch": 440, "lr": 9.504948543032977e-05}
{"train_loss": 0.31410548090934753, "global_step": 39203, "epoch": 440, "lr": 9.504923391410674e-05}
{"train_loss": 0.30180492997169495, "global_step": 39204, "epoch": 440, "lr": 9.504898239182741e-05}
{"train_loss": 0.2936524748802185, "global_step": 39205, "epoch": 440, "lr": 9.504873086349177e-05}
{"train_loss": 0.3046802282333374, "global_step": 39206, "epoch": 440, "lr": 9.50484793290999e-05}
{"train_loss": 0.24460449814796448, "global_step": 39207, "epoch": 440, "lr": 9.504822778865181e-05}
{"train_loss": 0.2956629693508148, "global_step": 39208, "epoch": 440, "lr": 9.504797624214752e-05}
{"train_loss": 0.2898118197917938, "global_step": 39209, "epoch": 440, "lr": 9.50477246895871e-05}
{"train_loss": 0.404226154088974, "global_step": 39210, "epoch": 440, "lr": 9.504747313097053e-05}
{"train_loss": 0.3897917568683624, "global_step": 39211, "epoch": 440, "lr": 9.50472215662979e-05}
{"train_loss": 0.36345362663269043, "global_step": 39212, "epoch": 440, "lr": 9.504696999556922e-05}
{"train_loss": 0.41441503167152405, "global_step": 39213, "epoch": 440, "lr": 9.504671841878452e-05}
{"train_loss": 0.25104251503944397, "global_step": 39214, "epoch": 440, "lr": 9.504646683594383e-05}
{"train_loss": 0.35298439860343933, "global_step": 39215, "epoch": 440, "lr": 9.50462152470472e-05}
{"train_loss": 0.3979685306549072, "global_step": 39216, "epoch": 440, "lr": 9.504596365209465e-05}
{"train_loss": 0.316326379776001, "global_step": 39217, "epoch": 440, "lr": 9.504571205108622e-05}
{"train_loss": 0.28862494230270386, "global_step": 39218, "epoch": 440, "lr": 9.504546044402195e-05}
{"train_loss": 0.3183114528656006, "global_step": 39219, "epoch": 440, "lr": 9.504520883090187e-05}
{"train_loss": 0.2867312729358673, "global_step": 39220, "epoch": 440, "lr": 9.5044957211726e-05}
{"train_loss": 0.31360283493995667, "global_step": 39221, "epoch": 440, "lr": 9.504470558649439e-05}
{"train_loss": 0.2861625850200653, "global_step": 39222, "epoch": 440, "lr": 9.504445395520706e-05}
{"train_loss": 0.30672886967658997, "global_step": 39223, "epoch": 440, "lr": 9.504420231786406e-05}
{"train_loss": 0.2866498827934265, "global_step": 39224, "epoch": 440, "lr": 9.504395067446541e-05}
{"train_loss": 0.25010502338409424, "global_step": 39225, "epoch": 440, "lr": 9.504369902501115e-05}
{"train_loss": 0.34951597452163696, "global_step": 39226, "epoch": 440, "lr": 9.504344736950133e-05}
{"train_loss": 0.3991599380970001, "global_step": 39227, "epoch": 440, "lr": 9.504319570793595e-05}
{"train_loss": 0.43410977721214294, "global_step": 39228, "epoch": 440, "lr": 9.504294404031507e-05}
{"train_loss": 0.24051043391227722, "global_step": 39229, "epoch": 440, "lr": 9.50426923666387e-05}
{"train_loss": 0.3113856315612793, "global_step": 39230, "epoch": 440, "lr": 9.50424406869069e-05}
{"train_loss": 0.28981271386146545, "global_step": 39231, "epoch": 440, "lr": 9.50421890011197e-05}
{"train_loss": 0.2858644127845764, "global_step": 39232, "epoch": 440, "lr": 9.504193730927712e-05}
{"train_loss": 0.2631235718727112, "global_step": 39233, "epoch": 440, "lr": 9.504168561137921e-05}
{"train_loss": 0.21653105318546295, "global_step": 39234, "epoch": 440, "lr": 9.504143390742597e-05}
{"train_loss": 0.3109126389026642, "global_step": 39235, "epoch": 440, "lr": 9.504118219741747e-05}
{"train_loss": 0.24038612842559814, "global_step": 39236, "epoch": 440, "lr": 9.504093048135374e-05}
{"train_loss": 0.20290721952915192, "global_step": 39237, "epoch": 440, "lr": 9.50406787592348e-05}
{"train_loss": 0.3505743443965912, "global_step": 39238, "epoch": 440, "lr": 9.50404270310607e-05}
{"train_loss": 0.3257373571395874, "global_step": 39239, "epoch": 440, "lr": 9.504017529683145e-05}
{"train_loss": 0.25682690739631653, "global_step": 39240, "epoch": 440, "lr": 9.50399235565471e-05}
{"train_loss": 0.30509454011917114, "global_step": 39241, "epoch": 440, "lr": 9.503967181020768e-05}
{"train_loss": 0.2099604606628418, "global_step": 39242, "epoch": 440, "lr": 9.503942005781324e-05}
{"train_loss": 0.2913946807384491, "global_step": 39243, "epoch": 440, "lr": 9.503916829936378e-05}
{"train_loss": 0.3079740107059479, "global_step": 39244, "epoch": 440, "lr": 9.503891653485935e-05}
{"train_loss": 0.35092106461524963, "global_step": 39245, "epoch": 440, "lr": 9.503866476430001e-05}
{"train_loss": 0.2835017740726471, "global_step": 39246, "epoch": 440, "lr": 9.503841298768575e-05}
{"train_loss": 0.3296617567539215, "global_step": 39247, "epoch": 440, "lr": 9.503816120501663e-05}
{"train_loss": 0.3108835977114988, "global_step": 39248, "epoch": 440, "lr": 9.503790941629268e-05, "val_loss": 2.7983334064483643, "train_action_mse_error": 31.90727996826172}
{"train_loss": 0.2854424715042114, "global_step": 39249, "epoch": 441, "lr": 9.503765762151393e-05}
{"train_loss": 0.30038851499557495, "global_step": 39250, "epoch": 441, "lr": 9.503740582068041e-05}
{"train_loss": 0.2559988498687744, "global_step": 39251, "epoch": 441, "lr": 9.503715401379219e-05}
{"train_loss": 0.3598288893699646, "global_step": 39252, "epoch": 441, "lr": 9.503690220084924e-05}
{"train_loss": 0.3828497529029846, "global_step": 39253, "epoch": 441, "lr": 9.503665038185164e-05}
{"train_loss": 0.3840293288230896, "global_step": 39254, "epoch": 441, "lr": 9.50363985567994e-05}
{"train_loss": 0.3125634491443634, "global_step": 39255, "epoch": 441, "lr": 9.503614672569259e-05}
{"train_loss": 0.3919903635978699, "global_step": 39256, "epoch": 441, "lr": 9.50358948885312e-05}
{"train_loss": 0.2957739531993866, "global_step": 39257, "epoch": 441, "lr": 9.503564304531528e-05}
{"train_loss": 0.2976968586444855, "global_step": 39258, "epoch": 441, "lr": 9.503539119604488e-05}
{"train_loss": 0.34956279397010803, "global_step": 39259, "epoch": 441, "lr": 9.503513934072001e-05}
{"train_loss": 0.22279226779937744, "global_step": 39260, "epoch": 441, "lr": 9.50348874793407e-05}
{"train_loss": 0.22152182459831238, "global_step": 39261, "epoch": 441, "lr": 9.503463561190703e-05}
{"train_loss": 0.2751789689064026, "global_step": 39262, "epoch": 441, "lr": 9.503438373841898e-05}
{"train_loss": 0.318072646856308, "global_step": 39263, "epoch": 441, "lr": 9.503413185887661e-05}
{"train_loss": 0.3574303686618805, "global_step": 39264, "epoch": 441, "lr": 9.503387997327998e-05}
{"train_loss": 0.3073230981826782, "global_step": 39265, "epoch": 441, "lr": 9.503362808162906e-05}
{"train_loss": 0.2992732524871826, "global_step": 39266, "epoch": 441, "lr": 9.503337618392392e-05}
{"train_loss": 0.27513208985328674, "global_step": 39267, "epoch": 441, "lr": 9.503312428016459e-05}
{"train_loss": 0.3813610076904297, "global_step": 39268, "epoch": 441, "lr": 9.503287237035112e-05}
{"train_loss": 0.27189499139785767, "global_step": 39269, "epoch": 441, "lr": 9.503262045448352e-05}
{"train_loss": 0.3125387132167816, "global_step": 39270, "epoch": 441, "lr": 9.503236853256183e-05}
{"train_loss": 0.38425594568252563, "global_step": 39271, "epoch": 441, "lr": 9.50321166045861e-05}
{"train_loss": 0.32164427638053894, "global_step": 39272, "epoch": 441, "lr": 9.503186467055634e-05}
{"train_loss": 0.2944260537624359, "global_step": 39273, "epoch": 441, "lr": 9.50316127304726e-05}
{"train_loss": 0.2287885844707489, "global_step": 39274, "epoch": 441, "lr": 9.50313607843349e-05}
{"train_loss": 0.41987285017967224, "global_step": 39275, "epoch": 441, "lr": 9.50311088321433e-05}
{"train_loss": 0.2087133228778839, "global_step": 39276, "epoch": 441, "lr": 9.50308568738978e-05}
{"train_loss": 0.3775402903556824, "global_step": 39277, "epoch": 441, "lr": 9.503060490959845e-05}
{"train_loss": 0.4380033612251282, "global_step": 39278, "epoch": 441, "lr": 9.50303529392453e-05}
{"train_loss": 0.2940215468406677, "global_step": 39279, "epoch": 441, "lr": 9.503010096283834e-05}
{"train_loss": 0.22329194843769073, "global_step": 39280, "epoch": 441, "lr": 9.502984898037766e-05}
{"train_loss": 0.3140157461166382, "global_step": 39281, "epoch": 441, "lr": 9.502959699186324e-05}
{"train_loss": 0.2395828515291214, "global_step": 39282, "epoch": 441, "lr": 9.502934499729515e-05}
{"train_loss": 0.22939209640026093, "global_step": 39283, "epoch": 441, "lr": 9.502909299667342e-05}
{"train_loss": 0.41665664315223694, "global_step": 39284, "epoch": 441, "lr": 9.502884098999807e-05}
{"train_loss": 0.30423375964164734, "global_step": 39285, "epoch": 441, "lr": 9.502858897726915e-05}
{"train_loss": 0.2751855254173279, "global_step": 39286, "epoch": 441, "lr": 9.502833695848668e-05}
{"train_loss": 0.20837178826332092, "global_step": 39287, "epoch": 441, "lr": 9.50280849336507e-05}
{"train_loss": 0.40670427680015564, "global_step": 39288, "epoch": 441, "lr": 9.502783290276124e-05}
{"train_loss": 0.2828492820262909, "global_step": 39289, "epoch": 441, "lr": 9.502758086581833e-05}
{"train_loss": 0.364489883184433, "global_step": 39290, "epoch": 441, "lr": 9.502732882282202e-05}
{"train_loss": 0.27821192145347595, "global_step": 39291, "epoch": 441, "lr": 9.502707677377234e-05}
{"train_loss": 0.4092230498790741, "global_step": 39292, "epoch": 441, "lr": 9.50268247186693e-05}
{"train_loss": 0.34690141677856445, "global_step": 39293, "epoch": 441, "lr": 9.502657265751298e-05}
{"train_loss": 0.28126469254493713, "global_step": 39294, "epoch": 441, "lr": 9.502632059030336e-05}
{"train_loss": 0.3376416862010956, "global_step": 39295, "epoch": 441, "lr": 9.502606851704052e-05}
{"train_loss": 0.2606121301651001, "global_step": 39296, "epoch": 441, "lr": 9.502581643772445e-05}
{"train_loss": 0.24930742383003235, "global_step": 39297, "epoch": 441, "lr": 9.502556435235523e-05}
{"train_loss": 0.30255281925201416, "global_step": 39298, "epoch": 441, "lr": 9.502531226093287e-05}
{"train_loss": 0.2713758647441864, "global_step": 39299, "epoch": 441, "lr": 9.502506016345739e-05}
{"train_loss": 0.3062215745449066, "global_step": 39300, "epoch": 441, "lr": 9.502480805992886e-05}
{"train_loss": 0.3572404980659485, "global_step": 39301, "epoch": 441, "lr": 9.502455595034727e-05}
{"train_loss": 0.24436607956886292, "global_step": 39302, "epoch": 441, "lr": 9.50243038347127e-05}
{"train_loss": 0.29452526569366455, "global_step": 39303, "epoch": 441, "lr": 9.502405171302514e-05}
{"train_loss": 0.31564241647720337, "global_step": 39304, "epoch": 441, "lr": 9.502379958528466e-05}
{"train_loss": 0.29498428106307983, "global_step": 39305, "epoch": 441, "lr": 9.502354745149127e-05}
{"train_loss": 0.32021844387054443, "global_step": 39306, "epoch": 441, "lr": 9.502329531164503e-05}
{"train_loss": 0.3065502643585205, "global_step": 39307, "epoch": 441, "lr": 9.502304316574592e-05}
{"train_loss": 0.2778661251068115, "global_step": 39308, "epoch": 441, "lr": 9.502279101379404e-05}
{"train_loss": 0.29382532835006714, "global_step": 39309, "epoch": 441, "lr": 9.50225388557894e-05}
{"train_loss": 0.2790997624397278, "global_step": 39310, "epoch": 441, "lr": 9.502228669173201e-05}
{"train_loss": 0.2992921769618988, "global_step": 39311, "epoch": 441, "lr": 9.502203452162193e-05}
{"train_loss": 0.29516416788101196, "global_step": 39312, "epoch": 441, "lr": 9.502178234545918e-05}
{"train_loss": 0.3321540653705597, "global_step": 39313, "epoch": 441, "lr": 9.502153016324382e-05}
{"train_loss": 0.26046791672706604, "global_step": 39314, "epoch": 441, "lr": 9.502127797497584e-05}
{"train_loss": 0.3447423577308655, "global_step": 39315, "epoch": 441, "lr": 9.50210257806553e-05}
{"train_loss": 0.37393301725387573, "global_step": 39316, "epoch": 441, "lr": 9.502077358028225e-05}
{"train_loss": 0.29416170716285706, "global_step": 39317, "epoch": 441, "lr": 9.502052137385668e-05}
{"train_loss": 0.29060450196266174, "global_step": 39318, "epoch": 441, "lr": 9.502026916137867e-05}
{"train_loss": 0.393856018781662, "global_step": 39319, "epoch": 441, "lr": 9.502001694284823e-05}
{"train_loss": 0.3542632460594177, "global_step": 39320, "epoch": 441, "lr": 9.501976471826539e-05}
{"train_loss": 0.37387755513191223, "global_step": 39321, "epoch": 441, "lr": 9.50195124876302e-05}
{"train_loss": 0.27602407336235046, "global_step": 39322, "epoch": 441, "lr": 9.501926025094268e-05}
{"train_loss": 0.32384729385375977, "global_step": 39323, "epoch": 441, "lr": 9.501900800820287e-05}
{"train_loss": 0.31472262740135193, "global_step": 39324, "epoch": 441, "lr": 9.501875575941081e-05}
{"train_loss": 0.33732593059539795, "global_step": 39325, "epoch": 441, "lr": 9.501850350456652e-05}
{"train_loss": 0.3067914545536041, "global_step": 39326, "epoch": 441, "lr": 9.501825124367005e-05}
{"train_loss": 0.3753068745136261, "global_step": 39327, "epoch": 441, "lr": 9.501799897672141e-05}
{"train_loss": 0.33748820424079895, "global_step": 39328, "epoch": 441, "lr": 9.501774670372065e-05}
{"train_loss": 0.36788004636764526, "global_step": 39329, "epoch": 441, "lr": 9.501749442466782e-05}
{"train_loss": 0.3388778865337372, "global_step": 39330, "epoch": 441, "lr": 9.501724213956292e-05}
{"train_loss": 0.32144269347190857, "global_step": 39331, "epoch": 441, "lr": 9.501698984840601e-05}
{"train_loss": 0.33543890714645386, "global_step": 39332, "epoch": 441, "lr": 9.50167375511971e-05}
{"train_loss": 0.3425131142139435, "global_step": 39333, "epoch": 441, "lr": 9.501648524793625e-05}
{"train_loss": 0.3970067799091339, "global_step": 39334, "epoch": 441, "lr": 9.501623293862348e-05}
{"train_loss": 0.3651082515716553, "global_step": 39335, "epoch": 441, "lr": 9.501598062325882e-05}
{"train_loss": 0.30926215648651123, "global_step": 39336, "epoch": 441, "lr": 9.501572830184233e-05}
{"train_loss": 0.3164085275336598, "global_step": 39337, "epoch": 441, "lr": 9.501547597437401e-05, "val_loss": 2.732309103012085}
{"train_loss": 0.37776556611061096, "global_step": 39338, "epoch": 442, "lr": 9.501522364085391e-05}
{"train_loss": 0.5014608502388, "global_step": 39339, "epoch": 442, "lr": 9.501497130128205e-05}
{"train_loss": 0.2814093232154846, "global_step": 39340, "epoch": 442, "lr": 9.50147189556585e-05}
{"train_loss": 0.3032185435295105, "global_step": 39341, "epoch": 442, "lr": 9.501446660398326e-05}
{"train_loss": 0.20620553195476532, "global_step": 39342, "epoch": 442, "lr": 9.501421424625636e-05}
{"train_loss": 0.2863398492336273, "global_step": 39343, "epoch": 442, "lr": 9.501396188247785e-05}
{"train_loss": 0.27863842248916626, "global_step": 39344, "epoch": 442, "lr": 9.501370951264778e-05}
{"train_loss": 0.35130155086517334, "global_step": 39345, "epoch": 442, "lr": 9.501345713676615e-05}
{"train_loss": 0.31802886724472046, "global_step": 39346, "epoch": 442, "lr": 9.501320475483302e-05}
{"train_loss": 0.2589697241783142, "global_step": 39347, "epoch": 442, "lr": 9.501295236684839e-05}
{"train_loss": 0.24097101390361786, "global_step": 39348, "epoch": 442, "lr": 9.501269997281234e-05}
{"train_loss": 0.30264100432395935, "global_step": 39349, "epoch": 442, "lr": 9.501244757272487e-05}
{"train_loss": 0.32824206352233887, "global_step": 39350, "epoch": 442, "lr": 9.501219516658604e-05}
{"train_loss": 0.2708393335342407, "global_step": 39351, "epoch": 442, "lr": 9.501194275439586e-05}
{"train_loss": 0.2430068850517273, "global_step": 39352, "epoch": 442, "lr": 9.501169033615436e-05}
{"train_loss": 0.2646513879299164, "global_step": 39353, "epoch": 442, "lr": 9.50114379118616e-05}
{"train_loss": 0.334743857383728, "global_step": 39354, "epoch": 442, "lr": 9.501118548151761e-05}
{"train_loss": 0.3411698341369629, "global_step": 39355, "epoch": 442, "lr": 9.501093304512239e-05}
{"train_loss": 0.371236115694046, "global_step": 39356, "epoch": 442, "lr": 9.501068060267601e-05}
{"train_loss": 0.27132701873779297, "global_step": 39357, "epoch": 442, "lr": 9.50104281541785e-05}
{"train_loss": 0.24279403686523438, "global_step": 39358, "epoch": 442, "lr": 9.50101756996299e-05}
{"train_loss": 0.2380256950855255, "global_step": 39359, "epoch": 442, "lr": 9.50099232390302e-05}
{"train_loss": 0.34341806173324585, "global_step": 39360, "epoch": 442, "lr": 9.500967077237948e-05}
{"train_loss": 0.4443167448043823, "global_step": 39361, "epoch": 442, "lr": 9.500941829967776e-05}
{"train_loss": 0.31471771001815796, "global_step": 39362, "epoch": 442, "lr": 9.500916582092505e-05}
{"train_loss": 0.3962889015674591, "global_step": 39363, "epoch": 442, "lr": 9.500891333612144e-05}
{"train_loss": 0.25189822912216187, "global_step": 39364, "epoch": 442, "lr": 9.50086608452669e-05}
{"train_loss": 0.2038797289133072, "global_step": 39365, "epoch": 442, "lr": 9.500840834836151e-05}
{"train_loss": 0.38259032368659973, "global_step": 39366, "epoch": 442, "lr": 9.500815584540529e-05}
{"train_loss": 0.3163865804672241, "global_step": 39367, "epoch": 442, "lr": 9.500790333639826e-05}
{"train_loss": 0.41689589619636536, "global_step": 39368, "epoch": 442, "lr": 9.500765082134048e-05}
{"train_loss": 0.25968554615974426, "global_step": 39369, "epoch": 442, "lr": 9.500739830023197e-05}
{"train_loss": 0.2193969041109085, "global_step": 39370, "epoch": 442, "lr": 9.500714577307275e-05}
{"train_loss": 0.20799456536769867, "global_step": 39371, "epoch": 442, "lr": 9.500689323986287e-05}
{"train_loss": 0.26835912466049194, "global_step": 39372, "epoch": 442, "lr": 9.500664070060237e-05}
{"train_loss": 0.2375805675983429, "global_step": 39373, "epoch": 442, "lr": 9.500638815529128e-05}
{"train_loss": 0.33364832401275635, "global_step": 39374, "epoch": 442, "lr": 9.500613560392962e-05}
{"train_loss": 0.30676230788230896, "global_step": 39375, "epoch": 442, "lr": 9.500588304651743e-05}
{"train_loss": 0.3745705485343933, "global_step": 39376, "epoch": 442, "lr": 9.500563048305475e-05}
{"train_loss": 0.2416132688522339, "global_step": 39377, "epoch": 442, "lr": 9.500537791354162e-05}
{"train_loss": 0.23003412783145905, "global_step": 39378, "epoch": 442, "lr": 9.500512533797808e-05}
{"train_loss": 0.41265708208084106, "global_step": 39379, "epoch": 442, "lr": 9.500487275636411e-05}
{"train_loss": 0.3831297755241394, "global_step": 39380, "epoch": 442, "lr": 9.500462016869981e-05}
{"train_loss": 0.3895726799964905, "global_step": 39381, "epoch": 442, "lr": 9.50043675749852e-05}
{"train_loss": 0.3703558146953583, "global_step": 39382, "epoch": 442, "lr": 9.500411497522026e-05}
{"train_loss": 0.21881674230098724, "global_step": 39383, "epoch": 442, "lr": 9.500386236940509e-05}
{"train_loss": 0.2925124764442444, "global_step": 39384, "epoch": 442, "lr": 9.50036097575397e-05}
{"train_loss": 0.32782864570617676, "global_step": 39385, "epoch": 442, "lr": 9.500335713962412e-05}
{"train_loss": 0.30351871252059937, "global_step": 39386, "epoch": 442, "lr": 9.500310451565839e-05}
{"train_loss": 0.3990216553211212, "global_step": 39387, "epoch": 442, "lr": 9.500285188564254e-05}
{"train_loss": 0.29497307538986206, "global_step": 39388, "epoch": 442, "lr": 9.500259924957659e-05}
{"train_loss": 0.32110434770584106, "global_step": 39389, "epoch": 442, "lr": 9.500234660746061e-05}
{"train_loss": 0.34114015102386475, "global_step": 39390, "epoch": 442, "lr": 9.50020939592946e-05}
{"train_loss": 0.33941876888275146, "global_step": 39391, "epoch": 442, "lr": 9.500184130507862e-05}
{"train_loss": 0.3584241569042206, "global_step": 39392, "epoch": 442, "lr": 9.500158864481268e-05}
{"train_loss": 0.3115474581718445, "global_step": 39393, "epoch": 442, "lr": 9.500133597849684e-05}
{"train_loss": 0.28859877586364746, "global_step": 39394, "epoch": 442, "lr": 9.500108330613108e-05}
{"train_loss": 0.399446964263916, "global_step": 39395, "epoch": 442, "lr": 9.500083062771553e-05}
{"train_loss": 0.3438051640987396, "global_step": 39396, "epoch": 442, "lr": 9.500057794325012e-05}
{"train_loss": 0.26708361506462097, "global_step": 39397, "epoch": 442, "lr": 9.500032525273495e-05}
{"train_loss": 0.37757590413093567, "global_step": 39398, "epoch": 442, "lr": 9.500007255617002e-05}
{"train_loss": 0.34703975915908813, "global_step": 39399, "epoch": 442, "lr": 9.499981985355541e-05}
{"train_loss": 0.2920810878276825, "global_step": 39400, "epoch": 442, "lr": 9.49995671448911e-05}
{"train_loss": 0.2253330796957016, "global_step": 39401, "epoch": 442, "lr": 9.499931443017716e-05}
{"train_loss": 0.289935827255249, "global_step": 39402, "epoch": 442, "lr": 9.49990617094136e-05}
{"train_loss": 0.2461538165807724, "global_step": 39403, "epoch": 442, "lr": 9.499880898260045e-05}
{"train_loss": 0.4205988645553589, "global_step": 39404, "epoch": 442, "lr": 9.499855624973777e-05}
{"train_loss": 0.32657960057258606, "global_step": 39405, "epoch": 442, "lr": 9.499830351082559e-05}
{"train_loss": 0.35680803656578064, "global_step": 39406, "epoch": 442, "lr": 9.499805076586394e-05}
{"train_loss": 0.31889405846595764, "global_step": 39407, "epoch": 442, "lr": 9.499779801485285e-05}
{"train_loss": 0.22392836213111877, "global_step": 39408, "epoch": 442, "lr": 9.499754525779234e-05}
{"train_loss": 0.18028542399406433, "global_step": 39409, "epoch": 442, "lr": 9.499729249468247e-05}
{"train_loss": 0.41601312160491943, "global_step": 39410, "epoch": 442, "lr": 9.499703972552326e-05}
{"train_loss": 0.2929439842700958, "global_step": 39411, "epoch": 442, "lr": 9.499678695031474e-05}
{"train_loss": 0.2449176013469696, "global_step": 39412, "epoch": 442, "lr": 9.499653416905697e-05}
{"train_loss": 0.3571770191192627, "global_step": 39413, "epoch": 442, "lr": 9.499628138174994e-05}
{"train_loss": 0.24591588973999023, "global_step": 39414, "epoch": 442, "lr": 9.499602858839372e-05}
{"train_loss": 0.35657697916030884, "global_step": 39415, "epoch": 442, "lr": 9.499577578898833e-05}
{"train_loss": 0.3530522286891937, "global_step": 39416, "epoch": 442, "lr": 9.49955229835338e-05}
{"train_loss": 0.3767186999320984, "global_step": 39417, "epoch": 442, "lr": 9.499527017203018e-05}
{"train_loss": 0.4704272449016571, "global_step": 39418, "epoch": 442, "lr": 9.49950173544775e-05}
{"train_loss": 0.3717184066772461, "global_step": 39419, "epoch": 442, "lr": 9.499476453087577e-05}
{"train_loss": 0.31959500908851624, "global_step": 39420, "epoch": 442, "lr": 9.499451170122506e-05}
{"train_loss": 0.33863577246665955, "global_step": 39421, "epoch": 442, "lr": 9.499425886552538e-05}
{"train_loss": 0.26584458351135254, "global_step": 39422, "epoch": 442, "lr": 9.499400602377677e-05}
{"train_loss": 0.2359924465417862, "global_step": 39423, "epoch": 442, "lr": 9.499375317597926e-05}
{"train_loss": 0.3783780634403229, "global_step": 39424, "epoch": 442, "lr": 9.49935003221329e-05}
{"train_loss": 0.28317925333976746, "global_step": 39425, "epoch": 442, "lr": 9.49932474622377e-05}
{"train_loss": 0.31606197357177734, "global_step": 39426, "epoch": 442, "lr": 9.499299459629371e-05, "val_loss": 2.625868082046509}
{"train_loss": 0.32904255390167236, "global_step": 39427, "epoch": 443, "lr": 9.499274172430096e-05}
{"train_loss": 0.3210655748844147, "global_step": 39428, "epoch": 443, "lr": 9.499248884625948e-05}
{"train_loss": 0.41324058175086975, "global_step": 39429, "epoch": 443, "lr": 9.499223596216932e-05}
{"train_loss": 0.28767043352127075, "global_step": 39430, "epoch": 443, "lr": 9.49919830720305e-05}
{"train_loss": 0.3616543710231781, "global_step": 39431, "epoch": 443, "lr": 9.499173017584305e-05}
{"train_loss": 0.3205242455005646, "global_step": 39432, "epoch": 443, "lr": 9.499147727360701e-05}
{"train_loss": 0.3419535160064697, "global_step": 39433, "epoch": 443, "lr": 9.499122436532242e-05}
{"train_loss": 0.27600768208503723, "global_step": 39434, "epoch": 443, "lr": 9.49909714509893e-05}
{"train_loss": 0.33215048909187317, "global_step": 39435, "epoch": 443, "lr": 9.49907185306077e-05}
{"train_loss": 0.32289594411849976, "global_step": 39436, "epoch": 443, "lr": 9.499046560417763e-05}
{"train_loss": 0.33167558908462524, "global_step": 39437, "epoch": 443, "lr": 9.499021267169916e-05}
{"train_loss": 0.2709942162036896, "global_step": 39438, "epoch": 443, "lr": 9.498995973317231e-05}
{"train_loss": 0.24585744738578796, "global_step": 39439, "epoch": 443, "lr": 9.498970678859709e-05}
{"train_loss": 0.28043875098228455, "global_step": 39440, "epoch": 443, "lr": 9.498945383797356e-05}
{"train_loss": 0.2826020419597626, "global_step": 39441, "epoch": 443, "lr": 9.498920088130174e-05}
{"train_loss": 0.3125085234642029, "global_step": 39442, "epoch": 443, "lr": 9.498894791858167e-05}
{"train_loss": 0.3428910970687866, "global_step": 39443, "epoch": 443, "lr": 9.49886949498134e-05}
{"train_loss": 0.29518333077430725, "global_step": 39444, "epoch": 443, "lr": 9.498844197499693e-05}
{"train_loss": 0.41454648971557617, "global_step": 39445, "epoch": 443, "lr": 9.498818899413232e-05}
{"train_loss": 0.31675639748573303, "global_step": 39446, "epoch": 443, "lr": 9.49879360072196e-05}
{"train_loss": 0.20500235259532928, "global_step": 39447, "epoch": 443, "lr": 9.49876830142588e-05}
{"train_loss": 0.3224758207798004, "global_step": 39448, "epoch": 443, "lr": 9.498743001524995e-05}
{"train_loss": 0.28848475217819214, "global_step": 39449, "epoch": 443, "lr": 9.498717701019309e-05}
{"train_loss": 0.3127812147140503, "global_step": 39450, "epoch": 443, "lr": 9.498692399908824e-05}
{"train_loss": 0.2771017849445343, "global_step": 39451, "epoch": 443, "lr": 9.498667098193546e-05}
{"train_loss": 0.2732968032360077, "global_step": 39452, "epoch": 443, "lr": 9.498641795873478e-05}
{"train_loss": 0.32123270630836487, "global_step": 39453, "epoch": 443, "lr": 9.49861649294862e-05}
{"train_loss": 0.22695203125476837, "global_step": 39454, "epoch": 443, "lr": 9.49859118941898e-05}
{"train_loss": 0.33807212114334106, "global_step": 39455, "epoch": 443, "lr": 9.498565885284558e-05}
{"train_loss": 0.25103315711021423, "global_step": 39456, "epoch": 443, "lr": 9.498540580545358e-05}
{"train_loss": 0.30543410778045654, "global_step": 39457, "epoch": 443, "lr": 9.498515275201385e-05}
{"train_loss": 0.25488951802253723, "global_step": 39458, "epoch": 443, "lr": 9.498489969252642e-05}
{"train_loss": 0.30001750588417053, "global_step": 39459, "epoch": 443, "lr": 9.498464662699132e-05}
{"train_loss": 0.48522722721099854, "global_step": 39460, "epoch": 443, "lr": 9.498439355540858e-05}
{"train_loss": 0.4054107367992401, "global_step": 39461, "epoch": 443, "lr": 9.498414047777822e-05}
{"train_loss": 0.29000183939933777, "global_step": 39462, "epoch": 443, "lr": 9.498388739410031e-05}
{"train_loss": 0.37203845381736755, "global_step": 39463, "epoch": 443, "lr": 9.498363430437486e-05}
{"train_loss": 0.4099145233631134, "global_step": 39464, "epoch": 443, "lr": 9.498338120860191e-05}
{"train_loss": 0.2923367917537689, "global_step": 39465, "epoch": 443, "lr": 9.498312810678148e-05}
{"train_loss": 0.3143766522407532, "global_step": 39466, "epoch": 443, "lr": 9.498287499891362e-05}
{"train_loss": 0.3012734353542328, "global_step": 39467, "epoch": 443, "lr": 9.498262188499838e-05}
{"train_loss": 0.30724647641181946, "global_step": 39468, "epoch": 443, "lr": 9.498236876503575e-05}
{"train_loss": 0.26061999797821045, "global_step": 39469, "epoch": 443, "lr": 9.49821156390258e-05}
{"train_loss": 0.29497793316841125, "global_step": 39470, "epoch": 443, "lr": 9.498186250696856e-05}
{"train_loss": 0.36862117052078247, "global_step": 39471, "epoch": 443, "lr": 9.498160936886405e-05}
{"train_loss": 0.328315794467926, "global_step": 39472, "epoch": 443, "lr": 9.498135622471232e-05}
{"train_loss": 0.36067548394203186, "global_step": 39473, "epoch": 443, "lr": 9.498110307451338e-05}
{"train_loss": 0.30201441049575806, "global_step": 39474, "epoch": 443, "lr": 9.498084991826728e-05}
{"train_loss": 0.18212729692459106, "global_step": 39475, "epoch": 443, "lr": 9.498059675597407e-05}
{"train_loss": 0.21489237248897552, "global_step": 39476, "epoch": 443, "lr": 9.498034358763374e-05}
{"train_loss": 0.418173223733902, "global_step": 39477, "epoch": 443, "lr": 9.498009041324636e-05}
{"train_loss": 0.41649001836776733, "global_step": 39478, "epoch": 443, "lr": 9.497983723281197e-05}
{"train_loss": 0.2672017812728882, "global_step": 39479, "epoch": 443, "lr": 9.497958404633059e-05}
{"train_loss": 0.22867442667484283, "global_step": 39480, "epoch": 443, "lr": 9.497933085380223e-05}
{"train_loss": 0.29651299118995667, "global_step": 39481, "epoch": 443, "lr": 9.497907765522696e-05}
{"train_loss": 0.3665732741355896, "global_step": 39482, "epoch": 443, "lr": 9.49788244506048e-05}
{"train_loss": 0.2248191088438034, "global_step": 39483, "epoch": 443, "lr": 9.497857123993579e-05}
{"train_loss": 0.3181907534599304, "global_step": 39484, "epoch": 443, "lr": 9.497831802321995e-05}
{"train_loss": 0.275295227766037, "global_step": 39485, "epoch": 443, "lr": 9.497806480045733e-05}
{"train_loss": 0.3953477144241333, "global_step": 39486, "epoch": 443, "lr": 9.497781157164796e-05}
{"train_loss": 0.28524479269981384, "global_step": 39487, "epoch": 443, "lr": 9.497755833679188e-05}
{"train_loss": 0.3708881139755249, "global_step": 39488, "epoch": 443, "lr": 9.497730509588909e-05}
{"train_loss": 0.32542386651039124, "global_step": 39489, "epoch": 443, "lr": 9.497705184893966e-05}
{"train_loss": 0.39553725719451904, "global_step": 39490, "epoch": 443, "lr": 9.497679859594361e-05}
{"train_loss": 0.4347566068172455, "global_step": 39491, "epoch": 443, "lr": 9.4976545336901e-05}
{"train_loss": 0.4059467613697052, "global_step": 39492, "epoch": 443, "lr": 9.497629207181182e-05}
{"train_loss": 0.32532376050949097, "global_step": 39493, "epoch": 443, "lr": 9.497603880067614e-05}
{"train_loss": 0.26759037375450134, "global_step": 39494, "epoch": 443, "lr": 9.497578552349395e-05}
{"train_loss": 0.21832209825515747, "global_step": 39495, "epoch": 443, "lr": 9.497553224026535e-05}
{"train_loss": 0.20167408883571625, "global_step": 39496, "epoch": 443, "lr": 9.497527895099032e-05}
{"train_loss": 0.2890346646308899, "global_step": 39497, "epoch": 443, "lr": 9.497502565566892e-05}
{"train_loss": 0.2515825927257538, "global_step": 39498, "epoch": 443, "lr": 9.497477235430116e-05}
{"train_loss": 0.2911761701107025, "global_step": 39499, "epoch": 443, "lr": 9.49745190468871e-05}
{"train_loss": 0.3159971535205841, "global_step": 39500, "epoch": 443, "lr": 9.497426573342677e-05}
{"train_loss": 0.3213959336280823, "global_step": 39501, "epoch": 443, "lr": 9.497401241392018e-05}
{"train_loss": 0.41363173723220825, "global_step": 39502, "epoch": 443, "lr": 9.497375908836741e-05}
{"train_loss": 0.35415852069854736, "global_step": 39503, "epoch": 443, "lr": 9.497350575676844e-05}
{"train_loss": 0.2860540449619293, "global_step": 39504, "epoch": 443, "lr": 9.497325241912336e-05}
{"train_loss": 0.3671293556690216, "global_step": 39505, "epoch": 443, "lr": 9.497299907543215e-05}
{"train_loss": 0.29665037989616394, "global_step": 39506, "epoch": 443, "lr": 9.497274572569488e-05}
{"train_loss": 0.4295078217983246, "global_step": 39507, "epoch": 443, "lr": 9.497249236991156e-05}
{"train_loss": 0.2949473559856415, "global_step": 39508, "epoch": 443, "lr": 9.497223900808226e-05}
{"train_loss": 0.3951528072357178, "global_step": 39509, "epoch": 443, "lr": 9.497198564020697e-05}
{"train_loss": 0.275485634803772, "global_step": 39510, "epoch": 443, "lr": 9.497173226628574e-05}
{"train_loss": 0.3686123490333557, "global_step": 39511, "epoch": 443, "lr": 9.497147888631862e-05}
{"train_loss": 0.3404158055782318, "global_step": 39512, "epoch": 443, "lr": 9.497122550030563e-05}
{"train_loss": 0.36636996269226074, "global_step": 39513, "epoch": 443, "lr": 9.49709721082468e-05}
{"train_loss": 0.4157549738883972, "global_step": 39514, "epoch": 443, "lr": 9.497071871014218e-05}
{"train_loss": 0.3181784879625513, "global_step": 39515, "epoch": 443, "lr": 9.49704653059918e-05, "val_loss": 2.563291311264038}
{"train_loss": 0.2965061068534851, "global_step": 39516, "epoch": 444, "lr": 9.497021189579566e-05}
{"train_loss": 0.3747636079788208, "global_step": 39517, "epoch": 444, "lr": 9.496995847955385e-05}
{"train_loss": 0.2600633203983307, "global_step": 39518, "epoch": 444, "lr": 9.496970505726639e-05}
{"train_loss": 0.28581947088241577, "global_step": 39519, "epoch": 444, "lr": 9.496945162893329e-05}
{"train_loss": 0.33799949288368225, "global_step": 39520, "epoch": 444, "lr": 9.496919819455458e-05}
{"train_loss": 0.2765577733516693, "global_step": 39521, "epoch": 444, "lr": 9.496894475413032e-05}
{"train_loss": 0.302715003490448, "global_step": 39522, "epoch": 444, "lr": 9.496869130766053e-05}
{"train_loss": 0.2990822196006775, "global_step": 39523, "epoch": 444, "lr": 9.496843785514524e-05}
{"train_loss": 0.28448235988616943, "global_step": 39524, "epoch": 444, "lr": 9.49681843965845e-05}
{"train_loss": 0.45033854246139526, "global_step": 39525, "epoch": 444, "lr": 9.496793093197834e-05}
{"train_loss": 0.3181381821632385, "global_step": 39526, "epoch": 444, "lr": 9.49676774613268e-05}
{"train_loss": 0.20668290555477142, "global_step": 39527, "epoch": 444, "lr": 9.496742398462988e-05}
{"train_loss": 0.3677140474319458, "global_step": 39528, "epoch": 444, "lr": 9.496717050188766e-05}
{"train_loss": 0.35053113102912903, "global_step": 39529, "epoch": 444, "lr": 9.496691701310014e-05}
{"train_loss": 0.3172970712184906, "global_step": 39530, "epoch": 444, "lr": 9.496666351826735e-05}
{"train_loss": 0.4289904832839966, "global_step": 39531, "epoch": 444, "lr": 9.496641001738936e-05}
{"train_loss": 0.32373619079589844, "global_step": 39532, "epoch": 444, "lr": 9.496615651046619e-05}
{"train_loss": 0.3443354666233063, "global_step": 39533, "epoch": 444, "lr": 9.496590299749786e-05}
{"train_loss": 0.3853216767311096, "global_step": 39534, "epoch": 444, "lr": 9.49656494784844e-05}
{"train_loss": 0.2878451943397522, "global_step": 39535, "epoch": 444, "lr": 9.496539595342587e-05}
{"train_loss": 0.2528064250946045, "global_step": 39536, "epoch": 444, "lr": 9.496514242232229e-05}
{"train_loss": 0.3073529601097107, "global_step": 39537, "epoch": 444, "lr": 9.49648888851737e-05}
{"train_loss": 0.31648799777030945, "global_step": 39538, "epoch": 444, "lr": 9.496463534198012e-05}
{"train_loss": 0.43696504831314087, "global_step": 39539, "epoch": 444, "lr": 9.496438179274158e-05}
{"train_loss": 0.45179057121276855, "global_step": 39540, "epoch": 444, "lr": 9.496412823745815e-05}
{"train_loss": 0.2924887537956238, "global_step": 39541, "epoch": 444, "lr": 9.496387467612984e-05}
{"train_loss": 0.25652340054512024, "global_step": 39542, "epoch": 444, "lr": 9.496362110875668e-05}
{"train_loss": 0.1864580512046814, "global_step": 39543, "epoch": 444, "lr": 9.49633675353387e-05}
{"train_loss": 0.2843558192253113, "global_step": 39544, "epoch": 444, "lr": 9.496311395587595e-05}
{"train_loss": 0.36267903447151184, "global_step": 39545, "epoch": 444, "lr": 9.496286037036846e-05}
{"train_loss": 0.31470003724098206, "global_step": 39546, "epoch": 444, "lr": 9.496260677881626e-05}
{"train_loss": 0.2795553207397461, "global_step": 39547, "epoch": 444, "lr": 9.496235318121938e-05}
{"train_loss": 0.32390907406806946, "global_step": 39548, "epoch": 444, "lr": 9.496209957757786e-05}
{"train_loss": 0.29860156774520874, "global_step": 39549, "epoch": 444, "lr": 9.496184596789175e-05}
{"train_loss": 0.36037424206733704, "global_step": 39550, "epoch": 444, "lr": 9.496159235216104e-05}
{"train_loss": 0.22902750968933105, "global_step": 39551, "epoch": 444, "lr": 9.496133873038582e-05}
{"train_loss": 0.29913562536239624, "global_step": 39552, "epoch": 444, "lr": 9.496108510256608e-05}
{"train_loss": 0.3478652536869049, "global_step": 39553, "epoch": 444, "lr": 9.496083146870187e-05}
{"train_loss": 0.28658032417297363, "global_step": 39554, "epoch": 444, "lr": 9.496057782879324e-05}
{"train_loss": 0.25276699662208557, "global_step": 39555, "epoch": 444, "lr": 9.49603241828402e-05}
{"train_loss": 0.32129037380218506, "global_step": 39556, "epoch": 444, "lr": 9.496007053084278e-05}
{"train_loss": 0.3351098895072937, "global_step": 39557, "epoch": 444, "lr": 9.495981687280103e-05}
{"train_loss": 0.28137317299842834, "global_step": 39558, "epoch": 444, "lr": 9.4959563208715e-05}
{"train_loss": 0.31866157054901123, "global_step": 39559, "epoch": 444, "lr": 9.495930953858468e-05}
{"train_loss": 0.33647817373275757, "global_step": 39560, "epoch": 444, "lr": 9.495905586241014e-05}
{"train_loss": 0.3294931948184967, "global_step": 39561, "epoch": 444, "lr": 9.49588021801914e-05}
{"train_loss": 0.45914530754089355, "global_step": 39562, "epoch": 444, "lr": 9.49585484919285e-05}
{"train_loss": 0.4401625692844391, "global_step": 39563, "epoch": 444, "lr": 9.495829479762148e-05}
{"train_loss": 0.2993208169937134, "global_step": 39564, "epoch": 444, "lr": 9.495804109727034e-05}
{"train_loss": 0.2667478024959564, "global_step": 39565, "epoch": 444, "lr": 9.495778739087516e-05}
{"train_loss": 0.2813425064086914, "global_step": 39566, "epoch": 444, "lr": 9.495753367843595e-05}
{"train_loss": 0.35175958275794983, "global_step": 39567, "epoch": 444, "lr": 9.495727995995274e-05}
{"train_loss": 0.2813674211502075, "global_step": 39568, "epoch": 444, "lr": 9.495702623542558e-05}
{"train_loss": 0.4009682238101959, "global_step": 39569, "epoch": 444, "lr": 9.49567725048545e-05}
{"train_loss": 0.2923227846622467, "global_step": 39570, "epoch": 444, "lr": 9.495651876823952e-05}
{"train_loss": 0.2629421353340149, "global_step": 39571, "epoch": 444, "lr": 9.495626502558068e-05}
{"train_loss": 0.3775610029697418, "global_step": 39572, "epoch": 444, "lr": 9.495601127687804e-05}
{"train_loss": 0.34724199771881104, "global_step": 39573, "epoch": 444, "lr": 9.49557575221316e-05}
{"train_loss": 0.479736328125, "global_step": 39574, "epoch": 444, "lr": 9.495550376134139e-05}
{"train_loss": 0.3200394809246063, "global_step": 39575, "epoch": 444, "lr": 9.495524999450748e-05}
{"train_loss": 0.25584322214126587, "global_step": 39576, "epoch": 444, "lr": 9.495499622162987e-05}
{"train_loss": 0.4522617757320404, "global_step": 39577, "epoch": 444, "lr": 9.495474244270862e-05}
{"train_loss": 0.312654048204422, "global_step": 39578, "epoch": 444, "lr": 9.495448865774376e-05}
{"train_loss": 0.47930023074150085, "global_step": 39579, "epoch": 444, "lr": 9.49542348667353e-05}
{"train_loss": 0.29487380385398865, "global_step": 39580, "epoch": 444, "lr": 9.49539810696833e-05}
{"train_loss": 0.32040223479270935, "global_step": 39581, "epoch": 444, "lr": 9.495372726658778e-05}
{"train_loss": 0.3767375946044922, "global_step": 39582, "epoch": 444, "lr": 9.495347345744878e-05}
{"train_loss": 0.27613675594329834, "global_step": 39583, "epoch": 444, "lr": 9.495321964226634e-05}
{"train_loss": 0.33383047580718994, "global_step": 39584, "epoch": 444, "lr": 9.495296582104049e-05}
{"train_loss": 0.41264083981513977, "global_step": 39585, "epoch": 444, "lr": 9.495271199377125e-05}
{"train_loss": 0.28291502594947815, "global_step": 39586, "epoch": 444, "lr": 9.495245816045868e-05}
{"train_loss": 0.422655314207077, "global_step": 39587, "epoch": 444, "lr": 9.495220432110278e-05}
{"train_loss": 0.28754734992980957, "global_step": 39588, "epoch": 444, "lr": 9.495195047570361e-05}
{"train_loss": 0.34252092242240906, "global_step": 39589, "epoch": 444, "lr": 9.495169662426121e-05}
{"train_loss": 0.3649199903011322, "global_step": 39590, "epoch": 444, "lr": 9.49514427667756e-05}
{"train_loss": 0.3877817690372467, "global_step": 39591, "epoch": 444, "lr": 9.495118890324681e-05}
{"train_loss": 0.43979018926620483, "global_step": 39592, "epoch": 444, "lr": 9.495093503367488e-05}
{"train_loss": 0.31718966364860535, "global_step": 39593, "epoch": 444, "lr": 9.495068115805985e-05}
{"train_loss": 0.30133262276649475, "global_step": 39594, "epoch": 444, "lr": 9.495042727640174e-05}
{"train_loss": 0.3162382245063782, "global_step": 39595, "epoch": 444, "lr": 9.49501733887006e-05}
{"train_loss": 0.22392238676548004, "global_step": 39596, "epoch": 444, "lr": 9.494991949495647e-05}
{"train_loss": 0.3204330801963806, "global_step": 39597, "epoch": 444, "lr": 9.494966559516934e-05}
{"train_loss": 0.28644686937332153, "global_step": 39598, "epoch": 444, "lr": 9.49494116893393e-05}
{"train_loss": 0.40521153807640076, "global_step": 39599, "epoch": 444, "lr": 9.494915777746636e-05}
{"train_loss": 0.3228049874305725, "global_step": 39600, "epoch": 444, "lr": 9.494890385955053e-05}
{"train_loss": 0.3766140639781952, "global_step": 39601, "epoch": 444, "lr": 9.494864993559189e-05}
{"train_loss": 0.3373704254627228, "global_step": 39602, "epoch": 444, "lr": 9.494839600559045e-05}
{"train_loss": 0.391870379447937, "global_step": 39603, "epoch": 444, "lr": 9.494814206954623e-05}
{"train_loss": 0.3305347876602344, "global_step": 39604, "epoch": 444, "lr": 9.49478881274593e-05, "val_loss": 2.606402635574341}
{"train_loss": 0.3564505875110626, "global_step": 39605, "epoch": 445, "lr": 9.494763417932965e-05}
{"train_loss": 0.3793468773365021, "global_step": 39606, "epoch": 445, "lr": 9.494738022515737e-05}
{"train_loss": 0.3744371831417084, "global_step": 39607, "epoch": 445, "lr": 9.494712626494244e-05}
{"train_loss": 0.3113453984260559, "global_step": 39608, "epoch": 445, "lr": 9.494687229868493e-05}
{"train_loss": 0.3344614505767822, "global_step": 39609, "epoch": 445, "lr": 9.494661832638485e-05}
{"train_loss": 0.24939046800136566, "global_step": 39610, "epoch": 445, "lr": 9.494636434804225e-05}
{"train_loss": 0.34308674931526184, "global_step": 39611, "epoch": 445, "lr": 9.494611036365715e-05}
{"train_loss": 0.36018651723861694, "global_step": 39612, "epoch": 445, "lr": 9.494585637322962e-05}
{"train_loss": 0.39302608370780945, "global_step": 39613, "epoch": 445, "lr": 9.494560237675963e-05}
{"train_loss": 0.3429529368877411, "global_step": 39614, "epoch": 445, "lr": 9.494534837424726e-05}
{"train_loss": 0.3084544837474823, "global_step": 39615, "epoch": 445, "lr": 9.494509436569257e-05}
{"train_loss": 0.40508782863616943, "global_step": 39616, "epoch": 445, "lr": 9.494484035109552e-05}
{"train_loss": 0.3840603530406952, "global_step": 39617, "epoch": 445, "lr": 9.494458633045621e-05}
{"train_loss": 0.2955496311187744, "global_step": 39618, "epoch": 445, "lr": 9.494433230377464e-05}
{"train_loss": 0.2884213924407959, "global_step": 39619, "epoch": 445, "lr": 9.494407827105085e-05}
{"train_loss": 0.3310989737510681, "global_step": 39620, "epoch": 445, "lr": 9.494382423228487e-05}
{"train_loss": 0.3550632894039154, "global_step": 39621, "epoch": 445, "lr": 9.494357018747674e-05}
{"train_loss": 0.4491489827632904, "global_step": 39622, "epoch": 445, "lr": 9.494331613662651e-05}
{"train_loss": 0.4954912066459656, "global_step": 39623, "epoch": 445, "lr": 9.494306207973419e-05}
{"train_loss": 0.332409530878067, "global_step": 39624, "epoch": 445, "lr": 9.494280801679982e-05}
{"train_loss": 0.36003705859184265, "global_step": 39625, "epoch": 445, "lr": 9.494255394782346e-05}
{"train_loss": 0.30658045411109924, "global_step": 39626, "epoch": 445, "lr": 9.49422998728051e-05}
{"train_loss": 0.33668142557144165, "global_step": 39627, "epoch": 445, "lr": 9.49420457917448e-05}
{"train_loss": 0.3090595304965973, "global_step": 39628, "epoch": 445, "lr": 9.494179170464258e-05}
{"train_loss": 0.4515901803970337, "global_step": 39629, "epoch": 445, "lr": 9.494153761149849e-05}
{"train_loss": 0.2579534351825714, "global_step": 39630, "epoch": 445, "lr": 9.494128351231256e-05}
{"train_loss": 0.30383387207984924, "global_step": 39631, "epoch": 445, "lr": 9.494102940708482e-05}
{"train_loss": 0.30090320110321045, "global_step": 39632, "epoch": 445, "lr": 9.494077529581531e-05}
{"train_loss": 0.3521401584148407, "global_step": 39633, "epoch": 445, "lr": 9.494052117850406e-05}
{"train_loss": 0.4059510827064514, "global_step": 39634, "epoch": 445, "lr": 9.494026705515111e-05}
{"train_loss": 0.2561300992965698, "global_step": 39635, "epoch": 445, "lr": 9.494001292575648e-05}
{"train_loss": 0.4141918420791626, "global_step": 39636, "epoch": 445, "lr": 9.493975879032022e-05}
{"train_loss": 0.4255305528640747, "global_step": 39637, "epoch": 445, "lr": 9.493950464884235e-05}
{"train_loss": 0.3330141603946686, "global_step": 39638, "epoch": 445, "lr": 9.493925050132291e-05}
{"train_loss": 0.3502309322357178, "global_step": 39639, "epoch": 445, "lr": 9.493899634776194e-05}
{"train_loss": 0.3199734091758728, "global_step": 39640, "epoch": 445, "lr": 9.493874218815947e-05}
{"train_loss": 0.2388874888420105, "global_step": 39641, "epoch": 445, "lr": 9.493848802251552e-05}
{"train_loss": 0.4195416569709778, "global_step": 39642, "epoch": 445, "lr": 9.493823385083016e-05}
{"train_loss": 0.30741941928863525, "global_step": 39643, "epoch": 445, "lr": 9.49379796731034e-05}
{"train_loss": 0.30578675866127014, "global_step": 39644, "epoch": 445, "lr": 9.493772548933526e-05}
{"train_loss": 0.35613903403282166, "global_step": 39645, "epoch": 445, "lr": 9.49374712995258e-05}
{"train_loss": 0.2881549298763275, "global_step": 39646, "epoch": 445, "lr": 9.493721710367505e-05}
{"train_loss": 0.3408471941947937, "global_step": 39647, "epoch": 445, "lr": 9.493696290178303e-05}
{"train_loss": 0.37294596433639526, "global_step": 39648, "epoch": 445, "lr": 9.493670869384978e-05}
{"train_loss": 0.4408559799194336, "global_step": 39649, "epoch": 445, "lr": 9.493645447987536e-05}
{"train_loss": 0.3895631730556488, "global_step": 39650, "epoch": 445, "lr": 9.493620025985976e-05}
{"train_loss": 0.44787293672561646, "global_step": 39651, "epoch": 445, "lr": 9.493594603380305e-05}
{"train_loss": 0.3152633607387543, "global_step": 39652, "epoch": 445, "lr": 9.493569180170523e-05}
{"train_loss": 0.28402572870254517, "global_step": 39653, "epoch": 445, "lr": 9.493543756356637e-05}
{"train_loss": 0.2543758749961853, "global_step": 39654, "epoch": 445, "lr": 9.493518331938648e-05}
{"train_loss": 0.38209813833236694, "global_step": 39655, "epoch": 445, "lr": 9.493492906916562e-05}
{"train_loss": 0.31961193680763245, "global_step": 39656, "epoch": 445, "lr": 9.493467481290378e-05}
{"train_loss": 0.24760335683822632, "global_step": 39657, "epoch": 445, "lr": 9.493442055060103e-05}
{"train_loss": 0.3569379448890686, "global_step": 39658, "epoch": 445, "lr": 9.49341662822574e-05}
{"train_loss": 0.29396918416023254, "global_step": 39659, "epoch": 445, "lr": 9.493391200787292e-05}
{"train_loss": 0.38858920335769653, "global_step": 39660, "epoch": 445, "lr": 9.493365772744761e-05}
{"train_loss": 0.26437923312187195, "global_step": 39661, "epoch": 445, "lr": 9.493340344098153e-05}
{"train_loss": 0.25559917092323303, "global_step": 39662, "epoch": 445, "lr": 9.493314914847469e-05}
{"train_loss": 0.3400561213493347, "global_step": 39663, "epoch": 445, "lr": 9.493289484992715e-05}
{"train_loss": 0.27101314067840576, "global_step": 39664, "epoch": 445, "lr": 9.493264054533893e-05}
{"train_loss": 0.3400464951992035, "global_step": 39665, "epoch": 445, "lr": 9.493238623471006e-05}
{"train_loss": 0.2824784517288208, "global_step": 39666, "epoch": 445, "lr": 9.493213191804057e-05}
{"train_loss": 0.3884502649307251, "global_step": 39667, "epoch": 445, "lr": 9.493187759533052e-05}
{"train_loss": 0.2995073199272156, "global_step": 39668, "epoch": 445, "lr": 9.49316232665799e-05}
{"train_loss": 0.2960171103477478, "global_step": 39669, "epoch": 445, "lr": 9.49313689317888e-05}
{"train_loss": 0.31476062536239624, "global_step": 39670, "epoch": 445, "lr": 9.493111459095721e-05}
{"train_loss": 0.2306007742881775, "global_step": 39671, "epoch": 445, "lr": 9.493086024408517e-05}
{"train_loss": 0.45318856835365295, "global_step": 39672, "epoch": 445, "lr": 9.493060589117272e-05}
{"train_loss": 0.3303148150444031, "global_step": 39673, "epoch": 445, "lr": 9.493035153221992e-05}
{"train_loss": 0.2944100499153137, "global_step": 39674, "epoch": 445, "lr": 9.493009716722678e-05}
{"train_loss": 0.3925292193889618, "global_step": 39675, "epoch": 445, "lr": 9.492984279619332e-05}
{"train_loss": 0.4083408713340759, "global_step": 39676, "epoch": 445, "lr": 9.492958841911961e-05}
{"train_loss": 0.3769841492176056, "global_step": 39677, "epoch": 445, "lr": 9.492933403600566e-05}
{"train_loss": 0.37952136993408203, "global_step": 39678, "epoch": 445, "lr": 9.492907964685149e-05}
{"train_loss": 0.23981735110282898, "global_step": 39679, "epoch": 445, "lr": 9.492882525165717e-05}
{"train_loss": 0.2788376808166504, "global_step": 39680, "epoch": 445, "lr": 9.492857085042271e-05}
{"train_loss": 0.35988637804985046, "global_step": 39681, "epoch": 445, "lr": 9.492831644314817e-05}
{"train_loss": 0.3345640301704407, "global_step": 39682, "epoch": 445, "lr": 9.492806202983354e-05}
{"train_loss": 0.373821496963501, "global_step": 39683, "epoch": 445, "lr": 9.49278076104789e-05}
{"train_loss": 0.4361986815929413, "global_step": 39684, "epoch": 445, "lr": 9.492755318508424e-05}
{"train_loss": 0.4770902693271637, "global_step": 39685, "epoch": 445, "lr": 9.492729875364963e-05}
{"train_loss": 0.3096405863761902, "global_step": 39686, "epoch": 445, "lr": 9.49270443161751e-05}
{"train_loss": 0.46320706605911255, "global_step": 39687, "epoch": 445, "lr": 9.492678987266066e-05}
{"train_loss": 0.44757696986198425, "global_step": 39688, "epoch": 445, "lr": 9.492653542310638e-05}
{"train_loss": 0.3539814054965973, "global_step": 39689, "epoch": 445, "lr": 9.492628096751226e-05}
{"train_loss": 0.4253884255886078, "global_step": 39690, "epoch": 445, "lr": 9.492602650587836e-05}
{"train_loss": 0.3233794867992401, "global_step": 39691, "epoch": 445, "lr": 9.492577203820469e-05}
{"train_loss": 0.408627450466156, "global_step": 39692, "epoch": 445, "lr": 9.492551756449131e-05}
{"train_loss": 0.3461237447650245, "global_step": 39693, "epoch": 445, "lr": 9.492526308473824e-05, "val_loss": 2.5797276496887207, "train_action_mse_error": 20.543724060058594}
{"train_loss": 0.3202590048313141, "global_step": 39694, "epoch": 446, "lr": 9.49250085989455e-05}
{"train_loss": 0.2979808449745178, "global_step": 39695, "epoch": 446, "lr": 9.492475410711317e-05}
{"train_loss": 0.2695987820625305, "global_step": 39696, "epoch": 446, "lr": 9.492449960924124e-05}
{"train_loss": 0.337161123752594, "global_step": 39697, "epoch": 446, "lr": 9.492424510532975e-05}
{"train_loss": 0.3358612656593323, "global_step": 39698, "epoch": 446, "lr": 9.492399059537875e-05}
{"train_loss": 0.3537476360797882, "global_step": 39699, "epoch": 446, "lr": 9.492373607938826e-05}
{"train_loss": 0.4072258770465851, "global_step": 39700, "epoch": 446, "lr": 9.492348155735833e-05}
{"train_loss": 0.3953362703323364, "global_step": 39701, "epoch": 446, "lr": 9.492322702928899e-05}
{"train_loss": 0.30214399099349976, "global_step": 39702, "epoch": 446, "lr": 9.492297249518025e-05}
{"train_loss": 0.30848586559295654, "global_step": 39703, "epoch": 446, "lr": 9.492271795503219e-05}
{"train_loss": 0.3483906090259552, "global_step": 39704, "epoch": 446, "lr": 9.492246340884481e-05}
{"train_loss": 0.3116372227668762, "global_step": 39705, "epoch": 446, "lr": 9.492220885661813e-05}
{"train_loss": 0.3361092507839203, "global_step": 39706, "epoch": 446, "lr": 9.492195429835222e-05}
{"train_loss": 0.3286135494709015, "global_step": 39707, "epoch": 446, "lr": 9.492169973404712e-05}
{"train_loss": 0.30957770347595215, "global_step": 39708, "epoch": 446, "lr": 9.492144516370282e-05}
{"train_loss": 0.29880082607269287, "global_step": 39709, "epoch": 446, "lr": 9.492119058731939e-05}
{"train_loss": 0.32285991311073303, "global_step": 39710, "epoch": 446, "lr": 9.492093600489686e-05}
{"train_loss": 0.29811012744903564, "global_step": 39711, "epoch": 446, "lr": 9.492068141643525e-05}
{"train_loss": 0.2556602656841278, "global_step": 39712, "epoch": 446, "lr": 9.49204268219346e-05}
{"train_loss": 0.32205092906951904, "global_step": 39713, "epoch": 446, "lr": 9.492017222139495e-05}
{"train_loss": 0.26421183347702026, "global_step": 39714, "epoch": 446, "lr": 9.491991761481632e-05}
{"train_loss": 0.30796462297439575, "global_step": 39715, "epoch": 446, "lr": 9.491966300219877e-05}
{"train_loss": 0.3127145767211914, "global_step": 39716, "epoch": 446, "lr": 9.491940838354229e-05}
{"train_loss": 0.33942317962646484, "global_step": 39717, "epoch": 446, "lr": 9.491915375884698e-05}
{"train_loss": 0.2306881844997406, "global_step": 39718, "epoch": 446, "lr": 9.491889912811283e-05}
{"train_loss": 0.23432515561580658, "global_step": 39719, "epoch": 446, "lr": 9.491864449133987e-05}
{"train_loss": 0.3623315691947937, "global_step": 39720, "epoch": 446, "lr": 9.491838984852815e-05}
{"train_loss": 0.3446406126022339, "global_step": 39721, "epoch": 446, "lr": 9.49181351996777e-05}
{"train_loss": 0.34360620379447937, "global_step": 39722, "epoch": 446, "lr": 9.491788054478855e-05}
{"train_loss": 0.3054794371128082, "global_step": 39723, "epoch": 446, "lr": 9.491762588386073e-05}
{"train_loss": 0.36301174759864807, "global_step": 39724, "epoch": 446, "lr": 9.49173712168943e-05}
{"train_loss": 0.23332971334457397, "global_step": 39725, "epoch": 446, "lr": 9.491711654388926e-05}
{"train_loss": 0.3534228801727295, "global_step": 39726, "epoch": 446, "lr": 9.491686186484567e-05}
{"train_loss": 0.3153996765613556, "global_step": 39727, "epoch": 446, "lr": 9.491660717976356e-05}
{"train_loss": 0.45425719022750854, "global_step": 39728, "epoch": 446, "lr": 9.491635248864297e-05}
{"train_loss": 0.24802103638648987, "global_step": 39729, "epoch": 446, "lr": 9.49160977914839e-05}
{"train_loss": 0.21003903448581696, "global_step": 39730, "epoch": 446, "lr": 9.491584308828642e-05}
{"train_loss": 0.2813883125782013, "global_step": 39731, "epoch": 446, "lr": 9.491558837905055e-05}
{"train_loss": 0.38713204860687256, "global_step": 39732, "epoch": 446, "lr": 9.491533366377635e-05}
{"train_loss": 0.2923925220966339, "global_step": 39733, "epoch": 446, "lr": 9.491507894246378e-05}
{"train_loss": 0.38585764169692993, "global_step": 39734, "epoch": 446, "lr": 9.491482421511297e-05}
{"train_loss": 0.30952131748199463, "global_step": 39735, "epoch": 446, "lr": 9.49145694817239e-05}
{"train_loss": 0.2688412070274353, "global_step": 39736, "epoch": 446, "lr": 9.491431474229659e-05}
{"train_loss": 0.38166776299476624, "global_step": 39737, "epoch": 446, "lr": 9.491405999683112e-05}
{"train_loss": 0.25701048970222473, "global_step": 39738, "epoch": 446, "lr": 9.49138052453275e-05}
{"train_loss": 0.2870858907699585, "global_step": 39739, "epoch": 446, "lr": 9.491355048778576e-05}
{"train_loss": 0.2929907739162445, "global_step": 39740, "epoch": 446, "lr": 9.491329572420594e-05}
{"train_loss": 0.30776605010032654, "global_step": 39741, "epoch": 446, "lr": 9.491304095458809e-05}
{"train_loss": 0.35787731409072876, "global_step": 39742, "epoch": 446, "lr": 9.49127861789322e-05}
{"train_loss": 0.2851063311100006, "global_step": 39743, "epoch": 446, "lr": 9.491253139723836e-05}
{"train_loss": 0.2818765938282013, "global_step": 39744, "epoch": 446, "lr": 9.491227660950657e-05}
{"train_loss": 0.27659979462623596, "global_step": 39745, "epoch": 446, "lr": 9.491202181573687e-05}
{"train_loss": 0.2877260446548462, "global_step": 39746, "epoch": 446, "lr": 9.49117670159293e-05}
{"train_loss": 0.30240511894226074, "global_step": 39747, "epoch": 446, "lr": 9.491151221008388e-05}
{"train_loss": 0.41742658615112305, "global_step": 39748, "epoch": 446, "lr": 9.491125739820065e-05}
{"train_loss": 0.2608726918697357, "global_step": 39749, "epoch": 446, "lr": 9.491100258027967e-05}
{"train_loss": 0.27596694231033325, "global_step": 39750, "epoch": 446, "lr": 9.491074775632094e-05}
{"train_loss": 0.29117700457572937, "global_step": 39751, "epoch": 446, "lr": 9.491049292632451e-05}
{"train_loss": 0.3417937755584717, "global_step": 39752, "epoch": 446, "lr": 9.491023809029041e-05}
{"train_loss": 0.38245442509651184, "global_step": 39753, "epoch": 446, "lr": 9.490998324821868e-05}
{"train_loss": 0.40847450494766235, "global_step": 39754, "epoch": 446, "lr": 9.490972840010935e-05}
{"train_loss": 0.3264070153236389, "global_step": 39755, "epoch": 446, "lr": 9.490947354596246e-05}
{"train_loss": 0.35769808292388916, "global_step": 39756, "epoch": 446, "lr": 9.490921868577802e-05}
{"train_loss": 0.3120644986629486, "global_step": 39757, "epoch": 446, "lr": 9.49089638195561e-05}
{"train_loss": 0.4552123546600342, "global_step": 39758, "epoch": 446, "lr": 9.490870894729671e-05}
{"train_loss": 0.2708238959312439, "global_step": 39759, "epoch": 446, "lr": 9.490845406899988e-05}
{"train_loss": 0.297601580619812, "global_step": 39760, "epoch": 446, "lr": 9.490819918466568e-05}
{"train_loss": 0.2254243642091751, "global_step": 39761, "epoch": 446, "lr": 9.49079442942941e-05}
{"train_loss": 0.23533228039741516, "global_step": 39762, "epoch": 446, "lr": 9.490768939788522e-05}
{"train_loss": 0.2928520441055298, "global_step": 39763, "epoch": 446, "lr": 9.490743449543902e-05}
{"train_loss": 0.20430727303028107, "global_step": 39764, "epoch": 446, "lr": 9.490717958695558e-05}
{"train_loss": 0.28800472617149353, "global_step": 39765, "epoch": 446, "lr": 9.490692467243491e-05}
{"train_loss": 0.31132742762565613, "global_step": 39766, "epoch": 446, "lr": 9.490666975187705e-05}
{"train_loss": 0.28084060549736023, "global_step": 39767, "epoch": 446, "lr": 9.490641482528204e-05}
{"train_loss": 0.34063220024108887, "global_step": 39768, "epoch": 446, "lr": 9.49061598926499e-05}
{"train_loss": 0.3432425856590271, "global_step": 39769, "epoch": 446, "lr": 9.490590495398069e-05}
{"train_loss": 0.42766889929771423, "global_step": 39770, "epoch": 446, "lr": 9.490565000927442e-05}
{"train_loss": 0.361325204372406, "global_step": 39771, "epoch": 446, "lr": 9.490539505853114e-05}
{"train_loss": 0.27843958139419556, "global_step": 39772, "epoch": 446, "lr": 9.490514010175085e-05}
{"train_loss": 0.23139876127243042, "global_step": 39773, "epoch": 446, "lr": 9.490488513893364e-05}
{"train_loss": 0.35139724612236023, "global_step": 39774, "epoch": 446, "lr": 9.490463017007951e-05}
{"train_loss": 0.2566012740135193, "global_step": 39775, "epoch": 446, "lr": 9.490437519518849e-05}
{"train_loss": 0.30765846371650696, "global_step": 39776, "epoch": 446, "lr": 9.490412021426062e-05}
{"train_loss": 0.31317320466041565, "global_step": 39777, "epoch": 446, "lr": 9.490386522729595e-05}
{"train_loss": 0.31341299414634705, "global_step": 39778, "epoch": 446, "lr": 9.49036102342945e-05}
{"train_loss": 0.2727334201335907, "global_step": 39779, "epoch": 446, "lr": 9.490335523525631e-05}
{"train_loss": 0.34166964888572693, "global_step": 39780, "epoch": 446, "lr": 9.490310023018142e-05}
{"train_loss": 0.3464180529117584, "global_step": 39781, "epoch": 446, "lr": 9.490284521906984e-05}
{"train_loss": 0.31649429275748436, "global_step": 39782, "epoch": 446, "lr": 9.490259020192163e-05, "val_loss": 2.7718684673309326}
{"train_loss": 0.33572667837142944, "global_step": 39783, "epoch": 447, "lr": 9.490233517873681e-05}
{"train_loss": 0.3339340090751648, "global_step": 39784, "epoch": 447, "lr": 9.490208014951541e-05}
{"train_loss": 0.32208532094955444, "global_step": 39785, "epoch": 447, "lr": 9.490182511425748e-05}
{"train_loss": 0.2817220687866211, "global_step": 39786, "epoch": 447, "lr": 9.490157007296306e-05}
{"train_loss": 0.3550018072128296, "global_step": 39787, "epoch": 447, "lr": 9.490131502563215e-05}
{"train_loss": 0.3036308288574219, "global_step": 39788, "epoch": 447, "lr": 9.49010599722648e-05}
{"train_loss": 0.331276535987854, "global_step": 39789, "epoch": 447, "lr": 9.490080491286106e-05}
{"train_loss": 0.32627153396606445, "global_step": 39790, "epoch": 447, "lr": 9.490054984742096e-05}
{"train_loss": 0.3340456187725067, "global_step": 39791, "epoch": 447, "lr": 9.490029477594453e-05}
{"train_loss": 0.3658064603805542, "global_step": 39792, "epoch": 447, "lr": 9.49000396984318e-05}
{"train_loss": 0.3119850158691406, "global_step": 39793, "epoch": 447, "lr": 9.48997846148828e-05}
{"train_loss": 0.2707231342792511, "global_step": 39794, "epoch": 447, "lr": 9.489952952529758e-05}
{"train_loss": 0.30501067638397217, "global_step": 39795, "epoch": 447, "lr": 9.489927442967617e-05}
{"train_loss": 0.2384984940290451, "global_step": 39796, "epoch": 447, "lr": 9.489901932801858e-05}
{"train_loss": 0.31122881174087524, "global_step": 39797, "epoch": 447, "lr": 9.489876422032488e-05}
{"train_loss": 0.3733261227607727, "global_step": 39798, "epoch": 447, "lr": 9.489850910659508e-05}
{"train_loss": 0.2929666042327881, "global_step": 39799, "epoch": 447, "lr": 9.489825398682923e-05}
{"train_loss": 0.3721652030944824, "global_step": 39800, "epoch": 447, "lr": 9.489799886102735e-05}
{"train_loss": 0.28038427233695984, "global_step": 39801, "epoch": 447, "lr": 9.489774372918949e-05}
{"train_loss": 0.3372959792613983, "global_step": 39802, "epoch": 447, "lr": 9.489748859131566e-05}
{"train_loss": 0.25387972593307495, "global_step": 39803, "epoch": 447, "lr": 9.489723344740593e-05}
{"train_loss": 0.40196529030799866, "global_step": 39804, "epoch": 447, "lr": 9.489697829746029e-05}
{"train_loss": 0.40955010056495667, "global_step": 39805, "epoch": 447, "lr": 9.489672314147882e-05}
{"train_loss": 0.2926352322101593, "global_step": 39806, "epoch": 447, "lr": 9.489646797946152e-05}
{"train_loss": 0.26451367139816284, "global_step": 39807, "epoch": 447, "lr": 9.489621281140846e-05}
{"train_loss": 0.36107364296913147, "global_step": 39808, "epoch": 447, "lr": 9.489595763731963e-05}
{"train_loss": 0.38773855566978455, "global_step": 39809, "epoch": 447, "lr": 9.489570245719508e-05}
{"train_loss": 0.2786654829978943, "global_step": 39810, "epoch": 447, "lr": 9.489544727103487e-05}
{"train_loss": 0.34807056188583374, "global_step": 39811, "epoch": 447, "lr": 9.489519207883899e-05}
{"train_loss": 0.33603012561798096, "global_step": 39812, "epoch": 447, "lr": 9.489493688060752e-05}
{"train_loss": 0.3357188403606415, "global_step": 39813, "epoch": 447, "lr": 9.489468167634046e-05}
{"train_loss": 0.23005551099777222, "global_step": 39814, "epoch": 447, "lr": 9.489442646603786e-05}
{"train_loss": 0.26857492327690125, "global_step": 39815, "epoch": 447, "lr": 9.489417124969975e-05}
{"train_loss": 0.20366477966308594, "global_step": 39816, "epoch": 447, "lr": 9.489391602732618e-05}
{"train_loss": 0.2495424896478653, "global_step": 39817, "epoch": 447, "lr": 9.489366079891716e-05}
{"train_loss": 0.41648390889167786, "global_step": 39818, "epoch": 447, "lr": 9.489340556447274e-05}
{"train_loss": 0.2965586185455322, "global_step": 39819, "epoch": 447, "lr": 9.489315032399295e-05}
{"train_loss": 0.30430930852890015, "global_step": 39820, "epoch": 447, "lr": 9.48928950774778e-05}
{"train_loss": 0.3105442225933075, "global_step": 39821, "epoch": 447, "lr": 9.489263982492737e-05}
{"train_loss": 0.3200424015522003, "global_step": 39822, "epoch": 447, "lr": 9.489238456634166e-05}
{"train_loss": 0.3761333227157593, "global_step": 39823, "epoch": 447, "lr": 9.489212930172074e-05}
{"train_loss": 0.3470708131790161, "global_step": 39824, "epoch": 447, "lr": 9.489187403106459e-05}
{"train_loss": 0.2804943323135376, "global_step": 39825, "epoch": 447, "lr": 9.48916187543733e-05}
{"train_loss": 0.3112989068031311, "global_step": 39826, "epoch": 447, "lr": 9.489136347164686e-05}
{"train_loss": 0.36329877376556396, "global_step": 39827, "epoch": 447, "lr": 9.489110818288532e-05}
{"train_loss": 0.3914456069469452, "global_step": 39828, "epoch": 447, "lr": 9.489085288808873e-05}
{"train_loss": 0.33098194003105164, "global_step": 39829, "epoch": 447, "lr": 9.489059758725711e-05}
{"train_loss": 0.24647599458694458, "global_step": 39830, "epoch": 447, "lr": 9.489034228039048e-05}
{"train_loss": 0.4130920469760895, "global_step": 39831, "epoch": 447, "lr": 9.489008696748892e-05}
{"train_loss": 0.41309165954589844, "global_step": 39832, "epoch": 447, "lr": 9.488983164855243e-05}
{"train_loss": 0.2858177423477173, "global_step": 39833, "epoch": 447, "lr": 9.488957632358103e-05}
{"train_loss": 0.30616551637649536, "global_step": 39834, "epoch": 447, "lr": 9.48893209925748e-05}
{"train_loss": 0.24340949952602386, "global_step": 39835, "epoch": 447, "lr": 9.488906565553371e-05}
{"train_loss": 0.3539283871650696, "global_step": 39836, "epoch": 447, "lr": 9.488881031245787e-05}
{"train_loss": 0.35129547119140625, "global_step": 39837, "epoch": 447, "lr": 9.488855496334725e-05}
{"train_loss": 0.3707127869129181, "global_step": 39838, "epoch": 447, "lr": 9.488829960820191e-05}
{"train_loss": 0.3801804482936859, "global_step": 39839, "epoch": 447, "lr": 9.488804424702192e-05}
{"train_loss": 0.31900376081466675, "global_step": 39840, "epoch": 447, "lr": 9.488778887980725e-05}
{"train_loss": 0.32422199845314026, "global_step": 39841, "epoch": 447, "lr": 9.488753350655797e-05}
{"train_loss": 0.32240530848503113, "global_step": 39842, "epoch": 447, "lr": 9.48872781272741e-05}
{"train_loss": 0.29937809705734253, "global_step": 39843, "epoch": 447, "lr": 9.488702274195569e-05}
{"train_loss": 0.37597212195396423, "global_step": 39844, "epoch": 447, "lr": 9.488676735060278e-05}
{"train_loss": 0.3860808312892914, "global_step": 39845, "epoch": 447, "lr": 9.488651195321538e-05}
{"train_loss": 0.38961923122406006, "global_step": 39846, "epoch": 447, "lr": 9.488625654979353e-05}
{"train_loss": 0.4029906690120697, "global_step": 39847, "epoch": 447, "lr": 9.488600114033728e-05}
{"train_loss": 0.23523928225040436, "global_step": 39848, "epoch": 447, "lr": 9.488574572484663e-05}
{"train_loss": 0.39445024728775024, "global_step": 39849, "epoch": 447, "lr": 9.488549030332166e-05}
{"train_loss": 0.4559749960899353, "global_step": 39850, "epoch": 447, "lr": 9.488523487576239e-05}
{"train_loss": 0.3687080144882202, "global_step": 39851, "epoch": 447, "lr": 9.488497944216883e-05}
{"train_loss": 0.37028709053993225, "global_step": 39852, "epoch": 447, "lr": 9.488472400254105e-05}
{"train_loss": 0.2689814865589142, "global_step": 39853, "epoch": 447, "lr": 9.488446855687904e-05}
{"train_loss": 0.31030040979385376, "global_step": 39854, "epoch": 447, "lr": 9.488421310518288e-05}
{"train_loss": 0.25504180788993835, "global_step": 39855, "epoch": 447, "lr": 9.488395764745258e-05}
{"train_loss": 0.37069985270500183, "global_step": 39856, "epoch": 447, "lr": 9.488370218368818e-05}
{"train_loss": 0.2984329164028168, "global_step": 39857, "epoch": 447, "lr": 9.488344671388971e-05}
{"train_loss": 0.3741927146911621, "global_step": 39858, "epoch": 447, "lr": 9.488319123805722e-05}
{"train_loss": 0.2660566568374634, "global_step": 39859, "epoch": 447, "lr": 9.488293575619072e-05}
{"train_loss": 0.4102138578891754, "global_step": 39860, "epoch": 447, "lr": 9.488268026829026e-05}
{"train_loss": 0.33853447437286377, "global_step": 39861, "epoch": 447, "lr": 9.488242477435586e-05}
{"train_loss": 0.3696962594985962, "global_step": 39862, "epoch": 447, "lr": 9.488216927438757e-05}
{"train_loss": 0.2898007035255432, "global_step": 39863, "epoch": 447, "lr": 9.488191376838542e-05}
{"train_loss": 0.34451916813850403, "global_step": 39864, "epoch": 447, "lr": 9.488165825634945e-05}
{"train_loss": 0.23777130246162415, "global_step": 39865, "epoch": 447, "lr": 9.488140273827968e-05}
{"train_loss": 0.258373886346817, "global_step": 39866, "epoch": 447, "lr": 9.488114721417615e-05}
{"train_loss": 0.28512701392173767, "global_step": 39867, "epoch": 447, "lr": 9.488089168403892e-05}
{"train_loss": 0.35817280411720276, "global_step": 39868, "epoch": 447, "lr": 9.488063614786798e-05}
{"train_loss": 0.3422999680042267, "global_step": 39869, "epoch": 447, "lr": 9.48803806056634e-05}
{"train_loss": 0.41372689604759216, "global_step": 39870, "epoch": 447, "lr": 9.488012505742519e-05}
{"train_loss": 0.32955895682399194, "global_step": 39871, "epoch": 447, "lr": 9.48798695031534e-05, "val_loss": 2.6633362770080566}
{"train_loss": 0.2535337805747986, "global_step": 39872, "epoch": 448, "lr": 9.487961394284804e-05}
{"train_loss": 0.28691184520721436, "global_step": 39873, "epoch": 448, "lr": 9.487935837650918e-05}
{"train_loss": 0.2880875766277313, "global_step": 39874, "epoch": 448, "lr": 9.487910280413684e-05}
{"train_loss": 0.3075582683086395, "global_step": 39875, "epoch": 448, "lr": 9.487884722573104e-05}
{"train_loss": 0.4150276482105255, "global_step": 39876, "epoch": 448, "lr": 9.487859164129182e-05}
{"train_loss": 0.3048536479473114, "global_step": 39877, "epoch": 448, "lr": 9.487833605081923e-05}
{"train_loss": 0.36254775524139404, "global_step": 39878, "epoch": 448, "lr": 9.48780804543133e-05}
{"train_loss": 0.304809033870697, "global_step": 39879, "epoch": 448, "lr": 9.487782485177405e-05}
{"train_loss": 0.27156490087509155, "global_step": 39880, "epoch": 448, "lr": 9.487756924320155e-05}
{"train_loss": 0.3082607388496399, "global_step": 39881, "epoch": 448, "lr": 9.487731362859577e-05}
{"train_loss": 0.3575695753097534, "global_step": 39882, "epoch": 448, "lr": 9.487705800795682e-05}
{"train_loss": 0.3308177590370178, "global_step": 39883, "epoch": 448, "lr": 9.487680238128467e-05}
{"train_loss": 0.3339194357395172, "global_step": 39884, "epoch": 448, "lr": 9.487654674857937e-05}
{"train_loss": 0.24893192946910858, "global_step": 39885, "epoch": 448, "lr": 9.487629110984099e-05}
{"train_loss": 0.34482482075691223, "global_step": 39886, "epoch": 448, "lr": 9.487603546506954e-05}
{"train_loss": 0.2929994761943817, "global_step": 39887, "epoch": 448, "lr": 9.487577981426503e-05}
{"train_loss": 0.3111381530761719, "global_step": 39888, "epoch": 448, "lr": 9.487552415742753e-05}
{"train_loss": 0.23583804070949554, "global_step": 39889, "epoch": 448, "lr": 9.487526849455707e-05}
{"train_loss": 0.4177635610103607, "global_step": 39890, "epoch": 448, "lr": 9.487501282565368e-05}
{"train_loss": 0.3455249071121216, "global_step": 39891, "epoch": 448, "lr": 9.487475715071738e-05}
{"train_loss": 0.4010908603668213, "global_step": 39892, "epoch": 448, "lr": 9.487450146974823e-05}
{"train_loss": 0.36815744638442993, "global_step": 39893, "epoch": 448, "lr": 9.487424578274623e-05}
{"train_loss": 0.39570721983909607, "global_step": 39894, "epoch": 448, "lr": 9.487399008971145e-05}
{"train_loss": 0.2513221800327301, "global_step": 39895, "epoch": 448, "lr": 9.48737343906439e-05}
{"train_loss": 0.24013404548168182, "global_step": 39896, "epoch": 448, "lr": 9.487347868554362e-05}
{"train_loss": 0.22894152998924255, "global_step": 39897, "epoch": 448, "lr": 9.487322297441066e-05}
{"train_loss": 0.36353933811187744, "global_step": 39898, "epoch": 448, "lr": 9.487296725724504e-05}
{"train_loss": 0.278646856546402, "global_step": 39899, "epoch": 448, "lr": 9.487271153404679e-05}
{"train_loss": 0.33355602622032166, "global_step": 39900, "epoch": 448, "lr": 9.487245580481596e-05}
{"train_loss": 0.2978174686431885, "global_step": 39901, "epoch": 448, "lr": 9.487220006955257e-05}
{"train_loss": 0.23944607377052307, "global_step": 39902, "epoch": 448, "lr": 9.487194432825666e-05}
{"train_loss": 0.305820494890213, "global_step": 39903, "epoch": 448, "lr": 9.487168858092826e-05}
{"train_loss": 0.4052746593952179, "global_step": 39904, "epoch": 448, "lr": 9.487143282756741e-05}
{"train_loss": 0.406528502702713, "global_step": 39905, "epoch": 448, "lr": 9.487117706817415e-05}
{"train_loss": 0.3059314787387848, "global_step": 39906, "epoch": 448, "lr": 9.48709213027485e-05}
{"train_loss": 0.34896931052207947, "global_step": 39907, "epoch": 448, "lr": 9.487066553129051e-05}
{"train_loss": 0.3198956251144409, "global_step": 39908, "epoch": 448, "lr": 9.487040975380019e-05}
{"train_loss": 0.2872890830039978, "global_step": 39909, "epoch": 448, "lr": 9.487015397027761e-05}
{"train_loss": 0.33585354685783386, "global_step": 39910, "epoch": 448, "lr": 9.486989818072277e-05}
{"train_loss": 0.30171072483062744, "global_step": 39911, "epoch": 448, "lr": 9.486964238513572e-05}
{"train_loss": 0.3335687518119812, "global_step": 39912, "epoch": 448, "lr": 9.486938658351649e-05}
{"train_loss": 0.2998397946357727, "global_step": 39913, "epoch": 448, "lr": 9.486913077586513e-05}
{"train_loss": 0.3248961567878723, "global_step": 39914, "epoch": 448, "lr": 9.486887496218166e-05}
{"train_loss": 0.3300328552722931, "global_step": 39915, "epoch": 448, "lr": 9.486861914246611e-05}
{"train_loss": 0.2686157524585724, "global_step": 39916, "epoch": 448, "lr": 9.486836331671852e-05}
{"train_loss": 0.26603254675865173, "global_step": 39917, "epoch": 448, "lr": 9.486810748493893e-05}
{"train_loss": 0.370126873254776, "global_step": 39918, "epoch": 448, "lr": 9.486785164712737e-05}
{"train_loss": 0.33467134833335876, "global_step": 39919, "epoch": 448, "lr": 9.486759580328388e-05}
{"train_loss": 0.3090023994445801, "global_step": 39920, "epoch": 448, "lr": 9.486733995340848e-05}
{"train_loss": 0.29463639855384827, "global_step": 39921, "epoch": 448, "lr": 9.486708409750122e-05}
{"train_loss": 0.312536358833313, "global_step": 39922, "epoch": 448, "lr": 9.486682823556212e-05}
{"train_loss": 0.36505547165870667, "global_step": 39923, "epoch": 448, "lr": 9.486657236759123e-05}
{"train_loss": 0.3356917202472687, "global_step": 39924, "epoch": 448, "lr": 9.486631649358858e-05}
{"train_loss": 0.34064796566963196, "global_step": 39925, "epoch": 448, "lr": 9.486606061355417e-05}
{"train_loss": 0.3870724141597748, "global_step": 39926, "epoch": 448, "lr": 9.486580472748809e-05}
{"train_loss": 0.4478829801082611, "global_step": 39927, "epoch": 448, "lr": 9.486554883539035e-05}
{"train_loss": 0.34485694766044617, "global_step": 39928, "epoch": 448, "lr": 9.486529293726099e-05}
{"train_loss": 0.30161499977111816, "global_step": 39929, "epoch": 448, "lr": 9.486503703310001e-05}
{"train_loss": 0.30718734860420227, "global_step": 39930, "epoch": 448, "lr": 9.48647811229075e-05}
{"train_loss": 0.41547906398773193, "global_step": 39931, "epoch": 448, "lr": 9.486452520668345e-05}
{"train_loss": 0.2826709449291229, "global_step": 39932, "epoch": 448, "lr": 9.486426928442793e-05}
{"train_loss": 0.24148176610469818, "global_step": 39933, "epoch": 448, "lr": 9.486401335614095e-05}
{"train_loss": 0.318705290555954, "global_step": 39934, "epoch": 448, "lr": 9.486375742182254e-05}
{"train_loss": 0.28330689668655396, "global_step": 39935, "epoch": 448, "lr": 9.486350148147274e-05}
{"train_loss": 0.34429991245269775, "global_step": 39936, "epoch": 448, "lr": 9.486324553509161e-05}
{"train_loss": 0.2698080837726593, "global_step": 39937, "epoch": 448, "lr": 9.486298958267913e-05}
{"train_loss": 0.2809883654117584, "global_step": 39938, "epoch": 448, "lr": 9.48627336242354e-05}
{"train_loss": 0.30342426896095276, "global_step": 39939, "epoch": 448, "lr": 9.486247765976042e-05}
{"train_loss": 0.29607123136520386, "global_step": 39940, "epoch": 448, "lr": 9.486222168925422e-05}
{"train_loss": 0.31402650475502014, "global_step": 39941, "epoch": 448, "lr": 9.486196571271684e-05}
{"train_loss": 0.2828524708747864, "global_step": 39942, "epoch": 448, "lr": 9.486170973014832e-05}
{"train_loss": 0.3529776930809021, "global_step": 39943, "epoch": 448, "lr": 9.486145374154867e-05}
{"train_loss": 0.242758646607399, "global_step": 39944, "epoch": 448, "lr": 9.486119774691796e-05}
{"train_loss": 0.34725716710090637, "global_step": 39945, "epoch": 448, "lr": 9.486094174625621e-05}
{"train_loss": 0.24956515431404114, "global_step": 39946, "epoch": 448, "lr": 9.486068573956346e-05}
{"train_loss": 0.35275375843048096, "global_step": 39947, "epoch": 448, "lr": 9.486042972683973e-05}
{"train_loss": 0.3890454173088074, "global_step": 39948, "epoch": 448, "lr": 9.486017370808505e-05}
{"train_loss": 0.3481408357620239, "global_step": 39949, "epoch": 448, "lr": 9.485991768329948e-05}
{"train_loss": 0.2805963158607483, "global_step": 39950, "epoch": 448, "lr": 9.485966165248303e-05}
{"train_loss": 0.40922725200653076, "global_step": 39951, "epoch": 448, "lr": 9.485940561563576e-05}
{"train_loss": 0.2918870151042938, "global_step": 39952, "epoch": 448, "lr": 9.485914957275768e-05}
{"train_loss": 0.27321329712867737, "global_step": 39953, "epoch": 448, "lr": 9.485889352384884e-05}
{"train_loss": 0.28445279598236084, "global_step": 39954, "epoch": 448, "lr": 9.485863746890928e-05}
{"train_loss": 0.3193417489528656, "global_step": 39955, "epoch": 448, "lr": 9.4858381407939e-05}
{"train_loss": 0.3230156898498535, "global_step": 39956, "epoch": 448, "lr": 9.485812534093806e-05}
{"train_loss": 0.2798610031604767, "global_step": 39957, "epoch": 448, "lr": 9.485786926790649e-05}
{"train_loss": 0.3546997904777527, "global_step": 39958, "epoch": 448, "lr": 9.485761318884434e-05}
{"train_loss": 0.27592790126800537, "global_step": 39959, "epoch": 448, "lr": 9.485735710375163e-05}
{"train_loss": 0.3174375143278851, "global_step": 39960, "epoch": 448, "lr": 9.485710101262839e-05, "val_loss": 2.661794900894165}
{"train_loss": 0.2873154580593109, "global_step": 39961, "epoch": 449, "lr": 9.485684491547466e-05}
{"train_loss": 0.36958742141723633, "global_step": 39962, "epoch": 449, "lr": 9.485658881229047e-05}
{"train_loss": 0.34652194380760193, "global_step": 39963, "epoch": 449, "lr": 9.485633270307587e-05}
{"train_loss": 0.33781588077545166, "global_step": 39964, "epoch": 449, "lr": 9.485607658783087e-05}
{"train_loss": 0.2721327543258667, "global_step": 39965, "epoch": 449, "lr": 9.485582046655552e-05}
{"train_loss": 0.2815524935722351, "global_step": 39966, "epoch": 449, "lr": 9.485556433924986e-05}
{"train_loss": 0.3463956117630005, "global_step": 39967, "epoch": 449, "lr": 9.48553082059139e-05}
{"train_loss": 0.23590189218521118, "global_step": 39968, "epoch": 449, "lr": 9.48550520665477e-05}
{"train_loss": 0.35665708780288696, "global_step": 39969, "epoch": 449, "lr": 9.485479592115129e-05}
{"train_loss": 0.2900424301624298, "global_step": 39970, "epoch": 449, "lr": 9.48545397697247e-05}
{"train_loss": 0.22679880261421204, "global_step": 39971, "epoch": 449, "lr": 9.485428361226796e-05}
{"train_loss": 0.18017219007015228, "global_step": 39972, "epoch": 449, "lr": 9.48540274487811e-05}
{"train_loss": 0.2334701120853424, "global_step": 39973, "epoch": 449, "lr": 9.485377127926416e-05}
{"train_loss": 0.3083694279193878, "global_step": 39974, "epoch": 449, "lr": 9.485351510371719e-05}
{"train_loss": 0.3199390769004822, "global_step": 39975, "epoch": 449, "lr": 9.48532589221402e-05}
{"train_loss": 0.2429547756910324, "global_step": 39976, "epoch": 449, "lr": 9.485300273453326e-05}
{"train_loss": 0.24933688342571259, "global_step": 39977, "epoch": 449, "lr": 9.485274654089637e-05}
{"train_loss": 0.3497414290904999, "global_step": 39978, "epoch": 449, "lr": 9.485249034122957e-05}
{"train_loss": 0.42577067017555237, "global_step": 39979, "epoch": 449, "lr": 9.48522341355329e-05}
{"train_loss": 0.3688560128211975, "global_step": 39980, "epoch": 449, "lr": 9.485197792380638e-05}
{"train_loss": 0.3315080404281616, "global_step": 39981, "epoch": 449, "lr": 9.485172170605008e-05}
{"train_loss": 0.27667877078056335, "global_step": 39982, "epoch": 449, "lr": 9.4851465482264e-05}
{"train_loss": 0.407323956489563, "global_step": 39983, "epoch": 449, "lr": 9.485120925244819e-05}
{"train_loss": 0.26362696290016174, "global_step": 39984, "epoch": 449, "lr": 9.485095301660267e-05}
{"train_loss": 0.3503068685531616, "global_step": 39985, "epoch": 449, "lr": 9.485069677472752e-05}
{"train_loss": 0.31832507252693176, "global_step": 39986, "epoch": 449, "lr": 9.48504405268227e-05}
{"train_loss": 0.33746078610420227, "global_step": 39987, "epoch": 449, "lr": 9.48501842728883e-05}
{"train_loss": 0.31886935234069824, "global_step": 39988, "epoch": 449, "lr": 9.484992801292436e-05}
{"train_loss": 0.32559674978256226, "global_step": 39989, "epoch": 449, "lr": 9.484967174693087e-05}
{"train_loss": 0.28731948137283325, "global_step": 39990, "epoch": 449, "lr": 9.48494154749079e-05}
{"train_loss": 0.28999948501586914, "global_step": 39991, "epoch": 449, "lr": 9.484915919685546e-05}
{"train_loss": 0.3433579206466675, "global_step": 39992, "epoch": 449, "lr": 9.484890291277361e-05}
{"train_loss": 0.23859386146068573, "global_step": 39993, "epoch": 449, "lr": 9.484864662266237e-05}
{"train_loss": 0.29275351762771606, "global_step": 39994, "epoch": 449, "lr": 9.484839032652178e-05}
{"train_loss": 0.20400406420230865, "global_step": 39995, "epoch": 449, "lr": 9.484813402435185e-05}
{"train_loss": 0.42824628949165344, "global_step": 39996, "epoch": 449, "lr": 9.484787771615264e-05}
{"train_loss": 0.2712094485759735, "global_step": 39997, "epoch": 449, "lr": 9.48476214019242e-05}
{"train_loss": 0.22823847830295563, "global_step": 39998, "epoch": 449, "lr": 9.484736508166653e-05}
{"train_loss": 0.3371986150741577, "global_step": 39999, "epoch": 449, "lr": 9.484710875537968e-05}
{"train_loss": 0.3076012134552002, "global_step": 40000, "epoch": 449, "lr": 9.484685242306369e-05}
{"train_loss": 0.2431909441947937, "global_step": 40001, "epoch": 449, "lr": 9.484659608471858e-05}
{"train_loss": 0.29879769682884216, "global_step": 40002, "epoch": 449, "lr": 9.484633974034438e-05}
{"train_loss": 0.3627071678638458, "global_step": 40003, "epoch": 449, "lr": 9.484608338994114e-05}
{"train_loss": 0.3464282155036926, "global_step": 40004, "epoch": 449, "lr": 9.48458270335089e-05}
{"train_loss": 0.26660430431365967, "global_step": 40005, "epoch": 449, "lr": 9.48455706710477e-05}
{"train_loss": 0.24787183105945587, "global_step": 40006, "epoch": 449, "lr": 9.484531430255754e-05}
{"train_loss": 0.4173144996166229, "global_step": 40007, "epoch": 449, "lr": 9.484505792803847e-05}
{"train_loss": 0.2858242392539978, "global_step": 40008, "epoch": 449, "lr": 9.484480154749053e-05}
{"train_loss": 0.26736220717430115, "global_step": 40009, "epoch": 449, "lr": 9.484454516091377e-05}
{"train_loss": 0.35382670164108276, "global_step": 40010, "epoch": 449, "lr": 9.48442887683082e-05}
{"train_loss": 0.3539580702781677, "global_step": 40011, "epoch": 449, "lr": 9.484403236967385e-05}
{"train_loss": 0.2727786898612976, "global_step": 40012, "epoch": 449, "lr": 9.484377596501078e-05}
{"train_loss": 0.27282220125198364, "global_step": 40013, "epoch": 449, "lr": 9.4843519554319e-05}
{"train_loss": 0.2489830106496811, "global_step": 40014, "epoch": 449, "lr": 9.484326313759857e-05}
{"train_loss": 0.4306033253669739, "global_step": 40015, "epoch": 449, "lr": 9.484300671484949e-05}
{"train_loss": 0.4129382371902466, "global_step": 40016, "epoch": 449, "lr": 9.484275028607183e-05}
{"train_loss": 0.2407996654510498, "global_step": 40017, "epoch": 449, "lr": 9.48424938512656e-05}
{"train_loss": 0.3978725075721741, "global_step": 40018, "epoch": 449, "lr": 9.484223741043086e-05}
{"train_loss": 0.31794866919517517, "global_step": 40019, "epoch": 449, "lr": 9.484198096356761e-05}
{"train_loss": 0.2963753342628479, "global_step": 40020, "epoch": 449, "lr": 9.48417245106759e-05}
{"train_loss": 0.3865586519241333, "global_step": 40021, "epoch": 449, "lr": 9.484146805175578e-05}
{"train_loss": 0.35789746046066284, "global_step": 40022, "epoch": 449, "lr": 9.484121158680726e-05}
{"train_loss": 0.31900572776794434, "global_step": 40023, "epoch": 449, "lr": 9.48409551158304e-05}
{"train_loss": 0.369155615568161, "global_step": 40024, "epoch": 449, "lr": 9.484069863882521e-05}
{"train_loss": 0.24063453078269958, "global_step": 40025, "epoch": 449, "lr": 9.484044215579174e-05}
{"train_loss": 0.2124943882226944, "global_step": 40026, "epoch": 449, "lr": 9.484018566673e-05}
{"train_loss": 0.25176674127578735, "global_step": 40027, "epoch": 449, "lr": 9.483992917164007e-05}
{"train_loss": 0.3774363696575165, "global_step": 40028, "epoch": 449, "lr": 9.483967267052194e-05}
{"train_loss": 0.34211447834968567, "global_step": 40029, "epoch": 449, "lr": 9.483941616337567e-05}
{"train_loss": 0.3315783441066742, "global_step": 40030, "epoch": 449, "lr": 9.48391596502013e-05}
{"train_loss": 0.3530227243900299, "global_step": 40031, "epoch": 449, "lr": 9.483890313099882e-05}
{"train_loss": 0.2567467987537384, "global_step": 40032, "epoch": 449, "lr": 9.483864660576831e-05}
{"train_loss": 0.4251500368118286, "global_step": 40033, "epoch": 449, "lr": 9.48383900745098e-05}
{"train_loss": 0.38902953267097473, "global_step": 40034, "epoch": 449, "lr": 9.48381335372233e-05}
{"train_loss": 0.3952835202217102, "global_step": 40035, "epoch": 449, "lr": 9.483787699390886e-05}
{"train_loss": 0.31907275319099426, "global_step": 40036, "epoch": 449, "lr": 9.483762044456655e-05}
{"train_loss": 0.42041152715682983, "global_step": 40037, "epoch": 449, "lr": 9.483736388919632e-05}
{"train_loss": 0.30345839262008667, "global_step": 40038, "epoch": 449, "lr": 9.483710732779827e-05}
{"train_loss": 0.26863572001457214, "global_step": 40039, "epoch": 449, "lr": 9.483685076037243e-05}
{"train_loss": 0.4226076900959015, "global_step": 40040, "epoch": 449, "lr": 9.483659418691881e-05}
{"train_loss": 0.34461724758148193, "global_step": 40041, "epoch": 449, "lr": 9.483633760743746e-05}
{"train_loss": 0.2973502278327942, "global_step": 40042, "epoch": 449, "lr": 9.483608102192841e-05}
{"train_loss": 0.4264865517616272, "global_step": 40043, "epoch": 449, "lr": 9.48358244303917e-05}
{"train_loss": 0.3864390552043915, "global_step": 40044, "epoch": 449, "lr": 9.483556783282736e-05}
{"train_loss": 0.2170533984899521, "global_step": 40045, "epoch": 449, "lr": 9.483531122923541e-05}
{"train_loss": 0.3337026536464691, "global_step": 40046, "epoch": 449, "lr": 9.483505461961591e-05}
{"train_loss": 0.2584301233291626, "global_step": 40047, "epoch": 449, "lr": 9.483479800396889e-05}
{"train_loss": 0.3402375280857086, "global_step": 40048, "epoch": 449, "lr": 9.483454138229438e-05}
{"train_loss": 0.31692091099331887, "global_step": 40049, "epoch": 449, "lr": 9.483428475459239e-05, "val_loss": 2.6384241580963135}
{"train_loss": 0.2968309819698334, "global_step": 40050, "epoch": 450, "lr": 9.4834028120863e-05}
{"train_loss": 0.4236429035663605, "global_step": 40051, "epoch": 450, "lr": 9.483377148110619e-05}
{"train_loss": 0.37293919920921326, "global_step": 40052, "epoch": 450, "lr": 9.483351483532206e-05}
{"train_loss": 0.5009755492210388, "global_step": 40053, "epoch": 450, "lr": 9.483325818351059e-05}
{"train_loss": 0.35296887159347534, "global_step": 40054, "epoch": 450, "lr": 9.483300152567185e-05}
{"train_loss": 0.259606271982193, "global_step": 40055, "epoch": 450, "lr": 9.483274486180585e-05}
{"train_loss": 0.297301322221756, "global_step": 40056, "epoch": 450, "lr": 9.483248819191263e-05}
{"train_loss": 0.20457980036735535, "global_step": 40057, "epoch": 450, "lr": 9.483223151599223e-05}
{"train_loss": 0.41440242528915405, "global_step": 40058, "epoch": 450, "lr": 9.483197483404469e-05}
{"train_loss": 0.37656909227371216, "global_step": 40059, "epoch": 450, "lr": 9.483171814607004e-05}
{"train_loss": 0.2665913701057434, "global_step": 40060, "epoch": 450, "lr": 9.483146145206829e-05}
{"train_loss": 0.2953013479709625, "global_step": 40061, "epoch": 450, "lr": 9.483120475203952e-05}
{"train_loss": 0.23403331637382507, "global_step": 40062, "epoch": 450, "lr": 9.483094804598373e-05}
{"train_loss": 0.2636817991733551, "global_step": 40063, "epoch": 450, "lr": 9.483069133390097e-05}
{"train_loss": 0.2403911054134369, "global_step": 40064, "epoch": 450, "lr": 9.483043461579126e-05}
{"train_loss": 0.31826016306877136, "global_step": 40065, "epoch": 450, "lr": 9.483017789165465e-05}
{"train_loss": 0.3418736457824707, "global_step": 40066, "epoch": 450, "lr": 9.482992116149118e-05}
{"train_loss": 0.25925642251968384, "global_step": 40067, "epoch": 450, "lr": 9.482966442530085e-05}
{"train_loss": 0.28311920166015625, "global_step": 40068, "epoch": 450, "lr": 9.482940768308375e-05}
{"train_loss": 0.3896864354610443, "global_step": 40069, "epoch": 450, "lr": 9.482915093483986e-05}
{"train_loss": 0.3391336500644684, "global_step": 40070, "epoch": 450, "lr": 9.482889418056925e-05}
{"train_loss": 0.288312703371048, "global_step": 40071, "epoch": 450, "lr": 9.482863742027193e-05}
{"train_loss": 0.28147387504577637, "global_step": 40072, "epoch": 450, "lr": 9.482838065394795e-05}
{"train_loss": 0.3399890661239624, "global_step": 40073, "epoch": 450, "lr": 9.482812388159734e-05}
{"train_loss": 0.4692328870296478, "global_step": 40074, "epoch": 450, "lr": 9.482786710322014e-05}
{"train_loss": 0.255215585231781, "global_step": 40075, "epoch": 450, "lr": 9.482761031881638e-05}
{"train_loss": 0.3329771161079407, "global_step": 40076, "epoch": 450, "lr": 9.482735352838608e-05}
{"train_loss": 0.22492915391921997, "global_step": 40077, "epoch": 450, "lr": 9.48270967319293e-05}
{"train_loss": 0.39158833026885986, "global_step": 40078, "epoch": 450, "lr": 9.482683992944607e-05}
{"train_loss": 0.308588445186615, "global_step": 40079, "epoch": 450, "lr": 9.48265831209364e-05}
{"train_loss": 0.29266253113746643, "global_step": 40080, "epoch": 450, "lr": 9.482632630640036e-05}
{"train_loss": 0.3395301103591919, "global_step": 40081, "epoch": 450, "lr": 9.482606948583796e-05}
{"train_loss": 0.3700254559516907, "global_step": 40082, "epoch": 450, "lr": 9.482581265924923e-05}
{"train_loss": 0.17015615105628967, "global_step": 40083, "epoch": 450, "lr": 9.482555582663423e-05}
{"train_loss": 0.26933881640434265, "global_step": 40084, "epoch": 450, "lr": 9.482529898799297e-05}
{"train_loss": 0.2721785604953766, "global_step": 40085, "epoch": 450, "lr": 9.48250421433255e-05}
{"train_loss": 0.48153626918792725, "global_step": 40086, "epoch": 450, "lr": 9.482478529263185e-05}
{"train_loss": 0.2929375469684601, "global_step": 40087, "epoch": 450, "lr": 9.482452843591205e-05}
{"train_loss": 0.2791575789451599, "global_step": 40088, "epoch": 450, "lr": 9.482427157316616e-05}
{"train_loss": 0.3186279237270355, "global_step": 40089, "epoch": 450, "lr": 9.482401470439417e-05}
{"train_loss": 0.30747824907302856, "global_step": 40090, "epoch": 450, "lr": 9.482375782959614e-05}
{"train_loss": 0.2711838185787201, "global_step": 40091, "epoch": 450, "lr": 9.48235009487721e-05}
{"train_loss": 0.3224937319755554, "global_step": 40092, "epoch": 450, "lr": 9.48232440619221e-05}
{"train_loss": 0.21510356664657593, "global_step": 40093, "epoch": 450, "lr": 9.482298716904615e-05}
{"train_loss": 0.26510199904441833, "global_step": 40094, "epoch": 450, "lr": 9.48227302701443e-05}
{"train_loss": 0.26667648553848267, "global_step": 40095, "epoch": 450, "lr": 9.482247336521658e-05}
{"train_loss": 0.35268110036849976, "global_step": 40096, "epoch": 450, "lr": 9.482221645426302e-05}
{"train_loss": 0.35061389207839966, "global_step": 40097, "epoch": 450, "lr": 9.482195953728367e-05}
{"train_loss": 0.2894591987133026, "global_step": 40098, "epoch": 450, "lr": 9.482170261427855e-05}
{"train_loss": 0.24862991273403168, "global_step": 40099, "epoch": 450, "lr": 9.482144568524769e-05}
{"train_loss": 0.26761895418167114, "global_step": 40100, "epoch": 450, "lr": 9.482118875019114e-05}
{"train_loss": 0.3042004406452179, "global_step": 40101, "epoch": 450, "lr": 9.482093180910892e-05}
{"train_loss": 0.2790869474411011, "global_step": 40102, "epoch": 450, "lr": 9.482067486200108e-05}
{"train_loss": 0.39850154519081116, "global_step": 40103, "epoch": 450, "lr": 9.482041790886764e-05}
{"train_loss": 0.23972901701927185, "global_step": 40104, "epoch": 450, "lr": 9.482016094970864e-05}
{"train_loss": 0.2783629894256592, "global_step": 40105, "epoch": 450, "lr": 9.481990398452413e-05}
{"train_loss": 0.35297852754592896, "global_step": 40106, "epoch": 450, "lr": 9.481964701331413e-05}
{"train_loss": 0.30594491958618164, "global_step": 40107, "epoch": 450, "lr": 9.481939003607866e-05}
{"train_loss": 0.22620311379432678, "global_step": 40108, "epoch": 450, "lr": 9.481913305281777e-05}
{"train_loss": 0.298648476600647, "global_step": 40109, "epoch": 450, "lr": 9.48188760635315e-05}
{"train_loss": 0.3552529513835907, "global_step": 40110, "epoch": 450, "lr": 9.481861906821988e-05}
{"train_loss": 0.18411876261234283, "global_step": 40111, "epoch": 450, "lr": 9.481836206688294e-05}
{"train_loss": 0.34759241342544556, "global_step": 40112, "epoch": 450, "lr": 9.481810505952072e-05}
{"train_loss": 0.4168744683265686, "global_step": 40113, "epoch": 450, "lr": 9.481784804613325e-05}
{"train_loss": 0.335988312959671, "global_step": 40114, "epoch": 450, "lr": 9.481759102672056e-05}
{"train_loss": 0.37227943539619446, "global_step": 40115, "epoch": 450, "lr": 9.48173340012827e-05}
{"train_loss": 0.30374279618263245, "global_step": 40116, "epoch": 450, "lr": 9.48170769698197e-05}
{"train_loss": 0.43942803144454956, "global_step": 40117, "epoch": 450, "lr": 9.481681993233158e-05}
{"train_loss": 0.36995238065719604, "global_step": 40118, "epoch": 450, "lr": 9.48165628888184e-05}
{"train_loss": 0.27838319540023804, "global_step": 40119, "epoch": 450, "lr": 9.481630583928017e-05}
{"train_loss": 0.2788698077201843, "global_step": 40120, "epoch": 450, "lr": 9.481604878371693e-05}
{"train_loss": 0.36453887820243835, "global_step": 40121, "epoch": 450, "lr": 9.481579172212871e-05}
{"train_loss": 0.28598690032958984, "global_step": 40122, "epoch": 450, "lr": 9.481553465451556e-05}
{"train_loss": 0.3954252600669861, "global_step": 40123, "epoch": 450, "lr": 9.481527758087752e-05}
{"train_loss": 0.2902334928512573, "global_step": 40124, "epoch": 450, "lr": 9.48150205012146e-05}
{"train_loss": 0.32048025727272034, "global_step": 40125, "epoch": 450, "lr": 9.481476341552684e-05}
{"train_loss": 0.37422001361846924, "global_step": 40126, "epoch": 450, "lr": 9.48145063238143e-05}
{"train_loss": 0.2782353162765503, "global_step": 40127, "epoch": 450, "lr": 9.4814249226077e-05}
{"train_loss": 0.3569214940071106, "global_step": 40128, "epoch": 450, "lr": 9.481399212231494e-05}
{"train_loss": 0.3442605137825012, "global_step": 40129, "epoch": 450, "lr": 9.48137350125282e-05}
{"train_loss": 0.29699990153312683, "global_step": 40130, "epoch": 450, "lr": 9.48134778967168e-05}
{"train_loss": 0.33976519107818604, "global_step": 40131, "epoch": 450, "lr": 9.481322077488079e-05}
{"train_loss": 0.39908984303474426, "global_step": 40132, "epoch": 450, "lr": 9.481296364702017e-05}
{"train_loss": 0.25371018052101135, "global_step": 40133, "epoch": 450, "lr": 9.4812706513135e-05}
{"train_loss": 0.37048575282096863, "global_step": 40134, "epoch": 450, "lr": 9.48124493732253e-05}
{"train_loss": 0.3355467617511749, "global_step": 40135, "epoch": 450, "lr": 9.481219222729112e-05}
{"train_loss": 0.3599158525466919, "global_step": 40136, "epoch": 450, "lr": 9.481193507533248e-05}
{"train_loss": 0.34406137466430664, "global_step": 40137, "epoch": 450, "lr": 9.481167791734942e-05}
{"train_loss": 0.31826174727986367, "global_step": 40138, "epoch": 450, "lr": 9.4811420753342e-05, "train/sim_max_reward_0": 0.5666405155599461, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.5120549582669391, "train/sim_max_reward_3": 0.39086498677247705, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.7246028203881095, "test/sim_max_reward_4500000": 1.0, "test/sim_max_reward_4500001": 0.4761636388672439, "test/sim_max_reward_4500002": 0.48621182007468855, "test/sim_max_reward_4500003": 0.5367661281608083, "test/sim_max_reward_4500004": 0.5986701308661702, "test/sim_max_reward_4500005": 0.8676942394955618, "test/sim_max_reward_4500006": 0.42019287389867105, "test/sim_max_reward_4500007": 0.9360965259655248, "test/sim_max_reward_4500008": 0.9727145410786268, "test/sim_max_reward_4500009": 0.8414490961531179, "test/sim_max_reward_4500010": 0.9919270926105419, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.8705812746857482, "test/sim_max_reward_4500014": 0.9768773189291523, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.8975668883818856, "test/sim_max_reward_4500017": 0.5453416599435945, "test/sim_max_reward_4500018": 0.11718674762243238, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.9900308904615047, "test/sim_max_reward_4500021": 0.6973745065106675, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.7085478942773377, "test/sim_max_reward_4500024": 0.6728350067551384, "test/sim_max_reward_4500025": 0.848179544603361, "test/sim_max_reward_4500026": 0.42825377761737976, "test/sim_max_reward_4500027": 0.8627393841649142, "test/sim_max_reward_4500028": 0.5917642219687066, "test/sim_max_reward_4500029": 0.2514507061336335, "test/sim_max_reward_4500030": 0.9721264144340426, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6915292493442395, "test/sim_max_reward_4500033": 0.5519916115138327, "test/sim_max_reward_4500034": 0.18472936218642727, "test/sim_max_reward_4500035": 1.0, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.32092874689473155, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.5870369738713916, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9558960500481567, "test/sim_max_reward_4500044": 0.500449730728869, "test/sim_max_reward_4500045": 0.6596701467582948, "test/sim_max_reward_4500046": 0.49021035269131175, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.6939480792744669, "test/sim_max_reward_4500049": 0.8858318916289964, "train/mean_score": 0.699027213497912, "test/mean_score": 0.6616192903720234, "val_loss": 2.643298625946045, "train_action_mse_error": 36.59684753417969}
{"train_loss": 0.34390634298324585, "global_step": 40139, "epoch": 451, "lr": 9.48111635833102e-05}
{"train_loss": 0.3795205056667328, "global_step": 40140, "epoch": 451, "lr": 9.48109064072541e-05}
{"train_loss": 0.3382640480995178, "global_step": 40141, "epoch": 451, "lr": 9.481064922517372e-05}
{"train_loss": 0.27638474106788635, "global_step": 40142, "epoch": 451, "lr": 9.481039203706909e-05}
{"train_loss": 0.24837522208690643, "global_step": 40143, "epoch": 451, "lr": 9.481013484294026e-05}
{"train_loss": 0.35204017162323, "global_step": 40144, "epoch": 451, "lr": 9.480987764278724e-05}
{"train_loss": 0.3491780161857605, "global_step": 40145, "epoch": 451, "lr": 9.480962043661008e-05}
{"train_loss": 0.34064972400665283, "global_step": 40146, "epoch": 451, "lr": 9.480936322440882e-05}
{"train_loss": 0.3687293231487274, "global_step": 40147, "epoch": 451, "lr": 9.480910600618348e-05}
{"train_loss": 0.33540281653404236, "global_step": 40148, "epoch": 451, "lr": 9.48088487819341e-05}
{"train_loss": 0.29837843775749207, "global_step": 40149, "epoch": 451, "lr": 9.480859155166073e-05}
{"train_loss": 0.24213965237140656, "global_step": 40150, "epoch": 451, "lr": 9.480833431536337e-05}
{"train_loss": 0.31357067823410034, "global_step": 40151, "epoch": 451, "lr": 9.480807707304209e-05}
{"train_loss": 0.3544026017189026, "global_step": 40152, "epoch": 451, "lr": 9.480781982469691e-05}
{"train_loss": 0.34701699018478394, "global_step": 40153, "epoch": 451, "lr": 9.480756257032786e-05}
{"train_loss": 0.3403436541557312, "global_step": 40154, "epoch": 451, "lr": 9.4807305309935e-05}
{"train_loss": 0.32337695360183716, "global_step": 40155, "epoch": 451, "lr": 9.480704804351832e-05}
{"train_loss": 0.3183218836784363, "global_step": 40156, "epoch": 451, "lr": 9.480679077107788e-05}
{"train_loss": 0.31144097447395325, "global_step": 40157, "epoch": 451, "lr": 9.480653349261372e-05}
{"train_loss": 0.2733994424343109, "global_step": 40158, "epoch": 451, "lr": 9.480627620812586e-05}
{"train_loss": 0.41756537556648254, "global_step": 40159, "epoch": 451, "lr": 9.480601891761436e-05}
{"train_loss": 0.33030468225479126, "global_step": 40160, "epoch": 451, "lr": 9.480576162107922e-05}
{"train_loss": 0.30985227227211, "global_step": 40161, "epoch": 451, "lr": 9.480550431852048e-05}
{"train_loss": 0.3267641067504883, "global_step": 40162, "epoch": 451, "lr": 9.480524700993822e-05}
{"train_loss": 0.3126489222049713, "global_step": 40163, "epoch": 451, "lr": 9.480498969533242e-05}
{"train_loss": 0.3636632263660431, "global_step": 40164, "epoch": 451, "lr": 9.480473237470313e-05}
{"train_loss": 0.3283623456954956, "global_step": 40165, "epoch": 451, "lr": 9.48044750480504e-05}
{"train_loss": 0.30768993496894836, "global_step": 40166, "epoch": 451, "lr": 9.480421771537426e-05}
{"train_loss": 0.2999615967273712, "global_step": 40167, "epoch": 451, "lr": 9.480396037667472e-05}
{"train_loss": 0.3869690001010895, "global_step": 40168, "epoch": 451, "lr": 9.480370303195185e-05}
{"train_loss": 0.30825889110565186, "global_step": 40169, "epoch": 451, "lr": 9.480344568120567e-05}
{"train_loss": 0.38275769352912903, "global_step": 40170, "epoch": 451, "lr": 9.48031883244362e-05}
{"train_loss": 0.2917807698249817, "global_step": 40171, "epoch": 451, "lr": 9.48029309616435e-05}
{"train_loss": 0.30843818187713623, "global_step": 40172, "epoch": 451, "lr": 9.480267359282757e-05}
{"train_loss": 0.3483084738254547, "global_step": 40173, "epoch": 451, "lr": 9.480241621798849e-05}
{"train_loss": 0.24647897481918335, "global_step": 40174, "epoch": 451, "lr": 9.480215883712626e-05}
{"train_loss": 0.33391624689102173, "global_step": 40175, "epoch": 451, "lr": 9.480190145024093e-05}
{"train_loss": 0.3447172939777374, "global_step": 40176, "epoch": 451, "lr": 9.480164405733253e-05}
{"train_loss": 0.28529414534568787, "global_step": 40177, "epoch": 451, "lr": 9.480138665840109e-05}
{"train_loss": 0.334495484828949, "global_step": 40178, "epoch": 451, "lr": 9.480112925344667e-05}
{"train_loss": 0.3006143271923065, "global_step": 40179, "epoch": 451, "lr": 9.480087184246926e-05}
{"train_loss": 0.37074506282806396, "global_step": 40180, "epoch": 451, "lr": 9.480061442546893e-05}
{"train_loss": 0.2645176351070404, "global_step": 40181, "epoch": 451, "lr": 9.48003570024457e-05}
{"train_loss": 0.27860546112060547, "global_step": 40182, "epoch": 451, "lr": 9.48000995733996e-05}
{"train_loss": 0.31816545128822327, "global_step": 40183, "epoch": 451, "lr": 9.479984213833068e-05}
{"train_loss": 0.24030262231826782, "global_step": 40184, "epoch": 451, "lr": 9.479958469723897e-05}
{"train_loss": 0.43357980251312256, "global_step": 40185, "epoch": 451, "lr": 9.47993272501245e-05}
{"train_loss": 0.44938865303993225, "global_step": 40186, "epoch": 451, "lr": 9.479906979698731e-05}
{"train_loss": 0.38266974687576294, "global_step": 40187, "epoch": 451, "lr": 9.479881233782743e-05}
{"train_loss": 0.2583642303943634, "global_step": 40188, "epoch": 451, "lr": 9.479855487264488e-05}
{"train_loss": 0.3506702780723572, "global_step": 40189, "epoch": 451, "lr": 9.479829740143973e-05}
{"train_loss": 0.26106947660446167, "global_step": 40190, "epoch": 451, "lr": 9.479803992421199e-05}
{"train_loss": 0.34656110405921936, "global_step": 40191, "epoch": 451, "lr": 9.479778244096169e-05}
{"train_loss": 0.34250012040138245, "global_step": 40192, "epoch": 451, "lr": 9.479752495168888e-05}
{"train_loss": 0.37868452072143555, "global_step": 40193, "epoch": 451, "lr": 9.47972674563936e-05}
{"train_loss": 0.3967866897583008, "global_step": 40194, "epoch": 451, "lr": 9.479700995507586e-05}
{"train_loss": 0.39524221420288086, "global_step": 40195, "epoch": 451, "lr": 9.479675244773572e-05}
{"train_loss": 0.34551239013671875, "global_step": 40196, "epoch": 451, "lr": 9.47964949343732e-05}
{"train_loss": 0.4477246403694153, "global_step": 40197, "epoch": 451, "lr": 9.479623741498834e-05}
{"train_loss": 0.32307299971580505, "global_step": 40198, "epoch": 451, "lr": 9.479597988958115e-05}
{"train_loss": 0.35391080379486084, "global_step": 40199, "epoch": 451, "lr": 9.47957223581517e-05}
{"train_loss": 0.2912501394748688, "global_step": 40200, "epoch": 451, "lr": 9.479546482070001e-05}
{"train_loss": 0.2605917453765869, "global_step": 40201, "epoch": 451, "lr": 9.479520727722612e-05}
{"train_loss": 0.33179163932800293, "global_step": 40202, "epoch": 451, "lr": 9.479494972773007e-05}
{"train_loss": 0.2623743712902069, "global_step": 40203, "epoch": 451, "lr": 9.479469217221187e-05}
{"train_loss": 0.3415748178958893, "global_step": 40204, "epoch": 451, "lr": 9.479443461067156e-05}
{"train_loss": 0.3365377187728882, "global_step": 40205, "epoch": 451, "lr": 9.47941770431092e-05}
{"train_loss": 0.30206194519996643, "global_step": 40206, "epoch": 451, "lr": 9.479391946952481e-05}
{"train_loss": 0.36278507113456726, "global_step": 40207, "epoch": 451, "lr": 9.479366188991842e-05}
{"train_loss": 0.355699747800827, "global_step": 40208, "epoch": 451, "lr": 9.479340430429006e-05}
{"train_loss": 0.36323049664497375, "global_step": 40209, "epoch": 451, "lr": 9.47931467126398e-05}
{"train_loss": 0.2325654774904251, "global_step": 40210, "epoch": 451, "lr": 9.479288911496761e-05}
{"train_loss": 0.46023818850517273, "global_step": 40211, "epoch": 451, "lr": 9.479263151127358e-05}
{"train_loss": 0.16953694820404053, "global_step": 40212, "epoch": 451, "lr": 9.479237390155773e-05}
{"train_loss": 0.3329876959323883, "global_step": 40213, "epoch": 451, "lr": 9.479211628582008e-05}
{"train_loss": 0.29001420736312866, "global_step": 40214, "epoch": 451, "lr": 9.479185866406068e-05}
{"train_loss": 0.33747851848602295, "global_step": 40215, "epoch": 451, "lr": 9.479160103627958e-05}
{"train_loss": 0.27462729811668396, "global_step": 40216, "epoch": 451, "lr": 9.479134340247678e-05}
{"train_loss": 0.30183494091033936, "global_step": 40217, "epoch": 451, "lr": 9.479108576265232e-05}
{"train_loss": 0.3435615003108978, "global_step": 40218, "epoch": 451, "lr": 9.479082811680625e-05}
{"train_loss": 0.34157297015190125, "global_step": 40219, "epoch": 451, "lr": 9.47905704649386e-05}
{"train_loss": 0.3199121654033661, "global_step": 40220, "epoch": 451, "lr": 9.47903128070494e-05}
{"train_loss": 0.26926979422569275, "global_step": 40221, "epoch": 451, "lr": 9.47900551431387e-05}
{"train_loss": 0.37495750188827515, "global_step": 40222, "epoch": 451, "lr": 9.478979747320652e-05}
{"train_loss": 0.2765173017978668, "global_step": 40223, "epoch": 451, "lr": 9.478953979725289e-05}
{"train_loss": 0.26824650168418884, "global_step": 40224, "epoch": 451, "lr": 9.478928211527787e-05}
{"train_loss": 0.3171992599964142, "global_step": 40225, "epoch": 451, "lr": 9.478902442728146e-05}
{"train_loss": 0.3423694968223572, "global_step": 40226, "epoch": 451, "lr": 9.478876673326371e-05}
{"train_loss": 0.3259458563635858, "global_step": 40227, "epoch": 451, "lr": 9.478850903322467e-05, "val_loss": 2.6549532413482666}
{"train_loss": 0.333000510931015, "global_step": 40228, "epoch": 452, "lr": 9.478825132716435e-05}
{"train_loss": 0.3124956488609314, "global_step": 40229, "epoch": 452, "lr": 9.478799361508279e-05}
{"train_loss": 0.33249858021736145, "global_step": 40230, "epoch": 452, "lr": 9.478773589698004e-05}
{"train_loss": 0.2652128338813782, "global_step": 40231, "epoch": 452, "lr": 9.478747817285613e-05}
{"train_loss": 0.36662521958351135, "global_step": 40232, "epoch": 452, "lr": 9.478722044271108e-05}
{"train_loss": 0.43151938915252686, "global_step": 40233, "epoch": 452, "lr": 9.478696270654493e-05}
{"train_loss": 0.26368626952171326, "global_step": 40234, "epoch": 452, "lr": 9.478670496435773e-05}
{"train_loss": 0.34210413694381714, "global_step": 40235, "epoch": 452, "lr": 9.47864472161495e-05}
{"train_loss": 0.3364911377429962, "global_step": 40236, "epoch": 452, "lr": 9.478618946192028e-05}
{"train_loss": 0.3151187598705292, "global_step": 40237, "epoch": 452, "lr": 9.47859317016701e-05}
{"train_loss": 0.29984745383262634, "global_step": 40238, "epoch": 452, "lr": 9.478567393539899e-05}
{"train_loss": 0.3270428776741028, "global_step": 40239, "epoch": 452, "lr": 9.4785416163107e-05}
{"train_loss": 0.30843445658683777, "global_step": 40240, "epoch": 452, "lr": 9.478515838479415e-05}
{"train_loss": 0.2998921871185303, "global_step": 40241, "epoch": 452, "lr": 9.47849006004605e-05}
{"train_loss": 0.3187532424926758, "global_step": 40242, "epoch": 452, "lr": 9.478464281010604e-05}
{"train_loss": 0.35285958647727966, "global_step": 40243, "epoch": 452, "lr": 9.478438501373086e-05}
{"train_loss": 0.16167256236076355, "global_step": 40244, "epoch": 452, "lr": 9.478412721133494e-05}
{"train_loss": 0.3917844295501709, "global_step": 40245, "epoch": 452, "lr": 9.478386940291834e-05}
{"train_loss": 0.34231293201446533, "global_step": 40246, "epoch": 452, "lr": 9.478361158848112e-05}
{"train_loss": 0.29758399724960327, "global_step": 40247, "epoch": 452, "lr": 9.478335376802326e-05}
{"train_loss": 0.4136420786380768, "global_step": 40248, "epoch": 452, "lr": 9.478309594154485e-05}
{"train_loss": 0.34763067960739136, "global_step": 40249, "epoch": 452, "lr": 9.478283810904587e-05}
{"train_loss": 0.2645317316055298, "global_step": 40250, "epoch": 452, "lr": 9.47825802705264e-05}
{"train_loss": 0.32676851749420166, "global_step": 40251, "epoch": 452, "lr": 9.478232242598646e-05}
{"train_loss": 0.2955181896686554, "global_step": 40252, "epoch": 452, "lr": 9.478206457542608e-05}
{"train_loss": 0.24416474997997284, "global_step": 40253, "epoch": 452, "lr": 9.478180671884529e-05}
{"train_loss": 0.3737447261810303, "global_step": 40254, "epoch": 452, "lr": 9.478154885624416e-05}
{"train_loss": 0.43823617696762085, "global_step": 40255, "epoch": 452, "lr": 9.478129098762266e-05}
{"train_loss": 0.3633556067943573, "global_step": 40256, "epoch": 452, "lr": 9.478103311298088e-05}
{"train_loss": 0.3408328592777252, "global_step": 40257, "epoch": 452, "lr": 9.478077523231883e-05}
{"train_loss": 0.39127224683761597, "global_step": 40258, "epoch": 452, "lr": 9.478051734563654e-05}
{"train_loss": 0.370147705078125, "global_step": 40259, "epoch": 452, "lr": 9.478025945293407e-05}
{"train_loss": 0.31171610951423645, "global_step": 40260, "epoch": 452, "lr": 9.478000155421146e-05}
{"train_loss": 0.36390525102615356, "global_step": 40261, "epoch": 452, "lr": 9.477974364946869e-05}
{"train_loss": 0.3089202344417572, "global_step": 40262, "epoch": 452, "lr": 9.477948573870583e-05}
{"train_loss": 0.3139743506908417, "global_step": 40263, "epoch": 452, "lr": 9.477922782192292e-05}
{"train_loss": 0.3074102997779846, "global_step": 40264, "epoch": 452, "lr": 9.477896989912001e-05}
{"train_loss": 0.43313366174697876, "global_step": 40265, "epoch": 452, "lr": 9.477871197029708e-05}
{"train_loss": 0.2912095785140991, "global_step": 40266, "epoch": 452, "lr": 9.477845403545422e-05}
{"train_loss": 0.33269307017326355, "global_step": 40267, "epoch": 452, "lr": 9.477819609459144e-05}
{"train_loss": 0.3647158443927765, "global_step": 40268, "epoch": 452, "lr": 9.477793814770877e-05}
{"train_loss": 0.36154645681381226, "global_step": 40269, "epoch": 452, "lr": 9.477768019480626e-05}
{"train_loss": 0.41250497102737427, "global_step": 40270, "epoch": 452, "lr": 9.477742223588394e-05}
{"train_loss": 0.20152893662452698, "global_step": 40271, "epoch": 452, "lr": 9.477716427094184e-05}
{"train_loss": 0.2969047427177429, "global_step": 40272, "epoch": 452, "lr": 9.477690629997998e-05}
{"train_loss": 0.26964280009269714, "global_step": 40273, "epoch": 452, "lr": 9.477664832299842e-05}
{"train_loss": 0.2923052906990051, "global_step": 40274, "epoch": 452, "lr": 9.477639033999718e-05}
{"train_loss": 0.26263532042503357, "global_step": 40275, "epoch": 452, "lr": 9.477613235097632e-05}
{"train_loss": 0.3735272288322449, "global_step": 40276, "epoch": 452, "lr": 9.477587435593584e-05}
{"train_loss": 0.3408753573894501, "global_step": 40277, "epoch": 452, "lr": 9.47756163548758e-05}
{"train_loss": 0.3306311368942261, "global_step": 40278, "epoch": 452, "lr": 9.47753583477962e-05}
{"train_loss": 0.36508041620254517, "global_step": 40279, "epoch": 452, "lr": 9.477510033469712e-05}
{"train_loss": 0.3799339532852173, "global_step": 40280, "epoch": 452, "lr": 9.477484231557857e-05}
{"train_loss": 0.36878418922424316, "global_step": 40281, "epoch": 452, "lr": 9.47745842904406e-05}
{"train_loss": 0.3056618869304657, "global_step": 40282, "epoch": 452, "lr": 9.47743262592832e-05}
{"train_loss": 0.33251211047172546, "global_step": 40283, "epoch": 452, "lr": 9.477406822210646e-05}
{"train_loss": 0.2421354055404663, "global_step": 40284, "epoch": 452, "lr": 9.477381017891038e-05}
{"train_loss": 0.2989508807659149, "global_step": 40285, "epoch": 452, "lr": 9.477355212969503e-05}
{"train_loss": 0.304047554731369, "global_step": 40286, "epoch": 452, "lr": 9.47732940744604e-05}
{"train_loss": 0.4073559641838074, "global_step": 40287, "epoch": 452, "lr": 9.477303601320655e-05}
{"train_loss": 0.3549821376800537, "global_step": 40288, "epoch": 452, "lr": 9.477277794593352e-05}
{"train_loss": 0.2799444794654846, "global_step": 40289, "epoch": 452, "lr": 9.477251987264133e-05}
{"train_loss": 0.36043259501457214, "global_step": 40290, "epoch": 452, "lr": 9.477226179333002e-05}
{"train_loss": 0.33029353618621826, "global_step": 40291, "epoch": 452, "lr": 9.477200370799964e-05}
{"train_loss": 0.34016939997673035, "global_step": 40292, "epoch": 452, "lr": 9.477174561665019e-05}
{"train_loss": 0.3293837010860443, "global_step": 40293, "epoch": 452, "lr": 9.477148751928172e-05}
{"train_loss": 0.35051295161247253, "global_step": 40294, "epoch": 452, "lr": 9.477122941589429e-05}
{"train_loss": 0.3966575562953949, "global_step": 40295, "epoch": 452, "lr": 9.477097130648789e-05}
{"train_loss": 0.39720243215560913, "global_step": 40296, "epoch": 452, "lr": 9.477071319106259e-05}
{"train_loss": 0.45461201667785645, "global_step": 40297, "epoch": 452, "lr": 9.477045506961841e-05}
{"train_loss": 0.30526843667030334, "global_step": 40298, "epoch": 452, "lr": 9.47701969421554e-05}
{"train_loss": 0.3894731402397156, "global_step": 40299, "epoch": 452, "lr": 9.476993880867357e-05}
{"train_loss": 0.24826499819755554, "global_step": 40300, "epoch": 452, "lr": 9.476968066917298e-05}
{"train_loss": 0.28360408544540405, "global_step": 40301, "epoch": 452, "lr": 9.476942252365364e-05}
{"train_loss": 0.3303733170032501, "global_step": 40302, "epoch": 452, "lr": 9.476916437211561e-05}
{"train_loss": 0.3974909484386444, "global_step": 40303, "epoch": 452, "lr": 9.476890621455889e-05}
{"train_loss": 0.3711632788181305, "global_step": 40304, "epoch": 452, "lr": 9.476864805098356e-05}
{"train_loss": 0.37588709592819214, "global_step": 40305, "epoch": 452, "lr": 9.476838988138964e-05}
{"train_loss": 0.4029593765735626, "global_step": 40306, "epoch": 452, "lr": 9.476813170577711e-05}
{"train_loss": 0.33071476221084595, "global_step": 40307, "epoch": 452, "lr": 9.476787352414609e-05}
{"train_loss": 0.22734393179416656, "global_step": 40308, "epoch": 452, "lr": 9.476761533649657e-05}
{"train_loss": 0.3855019807815552, "global_step": 40309, "epoch": 452, "lr": 9.476735714282856e-05}
{"train_loss": 0.26053744554519653, "global_step": 40310, "epoch": 452, "lr": 9.476709894314215e-05}
{"train_loss": 0.2959514260292053, "global_step": 40311, "epoch": 452, "lr": 9.476684073743734e-05}
{"train_loss": 0.3027677834033966, "global_step": 40312, "epoch": 452, "lr": 9.476658252571418e-05}
{"train_loss": 0.17160870134830475, "global_step": 40313, "epoch": 452, "lr": 9.47663243079727e-05}
{"train_loss": 0.3751057982444763, "global_step": 40314, "epoch": 452, "lr": 9.476606608421294e-05}
{"train_loss": 0.36265066266059875, "global_step": 40315, "epoch": 452, "lr": 9.476580785443492e-05}
{"train_loss": 0.3318127753694406, "global_step": 40316, "epoch": 452, "lr": 9.476554961863868e-05, "val_loss": 2.5873401165008545}
{"train_loss": 0.29605811834335327, "global_step": 40317, "epoch": 453, "lr": 9.476529137682426e-05}
{"train_loss": 0.30805349349975586, "global_step": 40318, "epoch": 453, "lr": 9.476503312899169e-05}
{"train_loss": 0.2931795120239258, "global_step": 40319, "epoch": 453, "lr": 9.476477487514101e-05}
{"train_loss": 0.4582355320453644, "global_step": 40320, "epoch": 453, "lr": 9.476451661527224e-05}
{"train_loss": 0.2637636065483093, "global_step": 40321, "epoch": 453, "lr": 9.476425834938544e-05}
{"train_loss": 0.26763635873794556, "global_step": 40322, "epoch": 453, "lr": 9.476400007748063e-05}
{"train_loss": 0.316626638174057, "global_step": 40323, "epoch": 453, "lr": 9.476374179955784e-05}
{"train_loss": 0.305876225233078, "global_step": 40324, "epoch": 453, "lr": 9.476348351561711e-05}
{"train_loss": 0.24044422805309296, "global_step": 40325, "epoch": 453, "lr": 9.476322522565848e-05}
{"train_loss": 0.317050576210022, "global_step": 40326, "epoch": 453, "lr": 9.476296692968198e-05}
{"train_loss": 0.29708555340766907, "global_step": 40327, "epoch": 453, "lr": 9.476270862768764e-05}
{"train_loss": 0.2731218636035919, "global_step": 40328, "epoch": 453, "lr": 9.47624503196755e-05}
{"train_loss": 0.4137479364871979, "global_step": 40329, "epoch": 453, "lr": 9.476219200564561e-05}
{"train_loss": 0.3784734308719635, "global_step": 40330, "epoch": 453, "lr": 9.476193368559798e-05}
{"train_loss": 0.28798967599868774, "global_step": 40331, "epoch": 453, "lr": 9.476167535953264e-05}
{"train_loss": 0.2657218277454376, "global_step": 40332, "epoch": 453, "lr": 9.476141702744965e-05}
{"train_loss": 0.36083975434303284, "global_step": 40333, "epoch": 453, "lr": 9.476115868934902e-05}
{"train_loss": 0.3176654577255249, "global_step": 40334, "epoch": 453, "lr": 9.476090034523082e-05}
{"train_loss": 0.3265538513660431, "global_step": 40335, "epoch": 453, "lr": 9.476064199509505e-05}
{"train_loss": 0.403973788022995, "global_step": 40336, "epoch": 453, "lr": 9.476038363894175e-05}
{"train_loss": 0.31057417392730713, "global_step": 40337, "epoch": 453, "lr": 9.476012527677097e-05}
{"train_loss": 0.2954579293727875, "global_step": 40338, "epoch": 453, "lr": 9.475986690858274e-05}
{"train_loss": 0.3156868517398834, "global_step": 40339, "epoch": 453, "lr": 9.475960853437708e-05}
{"train_loss": 0.2198285013437271, "global_step": 40340, "epoch": 453, "lr": 9.475935015415405e-05}
{"train_loss": 0.24204973876476288, "global_step": 40341, "epoch": 453, "lr": 9.475909176791367e-05}
{"train_loss": 0.2364046275615692, "global_step": 40342, "epoch": 453, "lr": 9.475883337565596e-05}
{"train_loss": 0.33967673778533936, "global_step": 40343, "epoch": 453, "lr": 9.475857497738099e-05}
{"train_loss": 0.26138943433761597, "global_step": 40344, "epoch": 453, "lr": 9.475831657308876e-05}
{"train_loss": 0.27008602023124695, "global_step": 40345, "epoch": 453, "lr": 9.475805816277934e-05}
{"train_loss": 0.4293636083602905, "global_step": 40346, "epoch": 453, "lr": 9.475779974645272e-05}
{"train_loss": 0.3360750377178192, "global_step": 40347, "epoch": 453, "lr": 9.475754132410897e-05}
{"train_loss": 0.2830923795700073, "global_step": 40348, "epoch": 453, "lr": 9.475728289574813e-05}
{"train_loss": 0.29864615201950073, "global_step": 40349, "epoch": 453, "lr": 9.475702446137019e-05}
{"train_loss": 0.3385632336139679, "global_step": 40350, "epoch": 453, "lr": 9.475676602097523e-05}
{"train_loss": 0.26229941844940186, "global_step": 40351, "epoch": 453, "lr": 9.475650757456327e-05}
{"train_loss": 0.19524620473384857, "global_step": 40352, "epoch": 453, "lr": 9.475624912213434e-05}
{"train_loss": 0.2471664398908615, "global_step": 40353, "epoch": 453, "lr": 9.475599066368848e-05}
{"train_loss": 0.38662225008010864, "global_step": 40354, "epoch": 453, "lr": 9.475573219922571e-05}
{"train_loss": 0.2842582166194916, "global_step": 40355, "epoch": 453, "lr": 9.475547372874609e-05}
{"train_loss": 0.2943251132965088, "global_step": 40356, "epoch": 453, "lr": 9.475521525224963e-05}
{"train_loss": 0.42410701513290405, "global_step": 40357, "epoch": 453, "lr": 9.475495676973638e-05}
{"train_loss": 0.24044948816299438, "global_step": 40358, "epoch": 453, "lr": 9.475469828120638e-05}
{"train_loss": 0.40959393978118896, "global_step": 40359, "epoch": 453, "lr": 9.475443978665965e-05}
{"train_loss": 0.21977457404136658, "global_step": 40360, "epoch": 453, "lr": 9.475418128609624e-05}
{"train_loss": 0.2569828927516937, "global_step": 40361, "epoch": 453, "lr": 9.475392277951616e-05}
{"train_loss": 0.2697610557079315, "global_step": 40362, "epoch": 453, "lr": 9.475366426691947e-05}
{"train_loss": 0.40751564502716064, "global_step": 40363, "epoch": 453, "lr": 9.47534057483062e-05}
{"train_loss": 0.2592669129371643, "global_step": 40364, "epoch": 453, "lr": 9.475314722367638e-05}
{"train_loss": 0.2749391794204712, "global_step": 40365, "epoch": 453, "lr": 9.475288869303004e-05}
{"train_loss": 0.33659788966178894, "global_step": 40366, "epoch": 453, "lr": 9.475263015636722e-05}
{"train_loss": 0.27789542078971863, "global_step": 40367, "epoch": 453, "lr": 9.475237161368795e-05}
{"train_loss": 0.1781017780303955, "global_step": 40368, "epoch": 453, "lr": 9.475211306499228e-05}
{"train_loss": 0.3022778630256653, "global_step": 40369, "epoch": 453, "lr": 9.475185451028023e-05}
{"train_loss": 0.3331780433654785, "global_step": 40370, "epoch": 453, "lr": 9.475159594955183e-05}
{"train_loss": 0.38784879446029663, "global_step": 40371, "epoch": 453, "lr": 9.475133738280713e-05}
{"train_loss": 0.3850131630897522, "global_step": 40372, "epoch": 453, "lr": 9.475107881004615e-05}
{"train_loss": 0.34155309200286865, "global_step": 40373, "epoch": 453, "lr": 9.475082023126895e-05}
{"train_loss": 0.4368036985397339, "global_step": 40374, "epoch": 453, "lr": 9.475056164647553e-05}
{"train_loss": 0.36049604415893555, "global_step": 40375, "epoch": 453, "lr": 9.475030305566594e-05}
{"train_loss": 0.34537991881370544, "global_step": 40376, "epoch": 453, "lr": 9.475004445884024e-05}
{"train_loss": 0.31444090604782104, "global_step": 40377, "epoch": 453, "lr": 9.474978585599844e-05}
{"train_loss": 0.29705625772476196, "global_step": 40378, "epoch": 453, "lr": 9.474952724714056e-05}
{"train_loss": 0.35072439908981323, "global_step": 40379, "epoch": 453, "lr": 9.474926863226666e-05}
{"train_loss": 0.36640870571136475, "global_step": 40380, "epoch": 453, "lr": 9.474901001137676e-05}
{"train_loss": 0.28980791568756104, "global_step": 40381, "epoch": 453, "lr": 9.47487513844709e-05}
{"train_loss": 0.26569533348083496, "global_step": 40382, "epoch": 453, "lr": 9.474849275154913e-05}
{"train_loss": 0.2984740734100342, "global_step": 40383, "epoch": 453, "lr": 9.474823411261145e-05}
{"train_loss": 0.3225964903831482, "global_step": 40384, "epoch": 453, "lr": 9.474797546765792e-05}
{"train_loss": 0.3644601106643677, "global_step": 40385, "epoch": 453, "lr": 9.474771681668859e-05}
{"train_loss": 0.22079315781593323, "global_step": 40386, "epoch": 453, "lr": 9.474745815970345e-05}
{"train_loss": 0.25804275274276733, "global_step": 40387, "epoch": 453, "lr": 9.474719949670257e-05}
{"train_loss": 0.3407261073589325, "global_step": 40388, "epoch": 453, "lr": 9.474694082768598e-05}
{"train_loss": 0.4525112211704254, "global_step": 40389, "epoch": 453, "lr": 9.47466821526537e-05}
{"train_loss": 0.26097071170806885, "global_step": 40390, "epoch": 453, "lr": 9.474642347160578e-05}
{"train_loss": 0.2919710576534271, "global_step": 40391, "epoch": 453, "lr": 9.474616478454226e-05}
{"train_loss": 0.36505916714668274, "global_step": 40392, "epoch": 453, "lr": 9.474590609146315e-05}
{"train_loss": 0.2532322406768799, "global_step": 40393, "epoch": 453, "lr": 9.47456473923685e-05}
{"train_loss": 0.3733503520488739, "global_step": 40394, "epoch": 453, "lr": 9.474538868725832e-05}
{"train_loss": 0.2902933955192566, "global_step": 40395, "epoch": 453, "lr": 9.474512997613271e-05}
{"train_loss": 0.3325057625770569, "global_step": 40396, "epoch": 453, "lr": 9.474487125899164e-05}
{"train_loss": 0.37011680006980896, "global_step": 40397, "epoch": 453, "lr": 9.474461253583518e-05}
{"train_loss": 0.34827256202697754, "global_step": 40398, "epoch": 453, "lr": 9.474435380666333e-05}
{"train_loss": 0.2575574517250061, "global_step": 40399, "epoch": 453, "lr": 9.474409507147616e-05}
{"train_loss": 0.31409865617752075, "global_step": 40400, "epoch": 453, "lr": 9.47438363302737e-05}
{"train_loss": 0.334070086479187, "global_step": 40401, "epoch": 453, "lr": 9.474357758305597e-05}
{"train_loss": 0.3044557273387909, "global_step": 40402, "epoch": 453, "lr": 9.474331882982299e-05}
{"train_loss": 0.3497820794582367, "global_step": 40403, "epoch": 453, "lr": 9.474306007057485e-05}
{"train_loss": 0.3502327799797058, "global_step": 40404, "epoch": 453, "lr": 9.474280130531153e-05}
{"train_loss": 0.31412596632255596, "global_step": 40405, "epoch": 453, "lr": 9.47425425340331e-05, "val_loss": 2.606027841567993}
{"train_loss": 0.3552487790584564, "global_step": 40406, "epoch": 454, "lr": 9.474228375673956e-05}
{"train_loss": 0.48657453060150146, "global_step": 40407, "epoch": 454, "lr": 9.474202497343099e-05}
{"train_loss": 0.2971181571483612, "global_step": 40408, "epoch": 454, "lr": 9.474176618410738e-05}
{"train_loss": 0.35713475942611694, "global_step": 40409, "epoch": 454, "lr": 9.47415073887688e-05}
{"train_loss": 0.22330844402313232, "global_step": 40410, "epoch": 454, "lr": 9.474124858741525e-05}
{"train_loss": 0.27524515986442566, "global_step": 40411, "epoch": 454, "lr": 9.47409897800468e-05}
{"train_loss": 0.24872520565986633, "global_step": 40412, "epoch": 454, "lr": 9.474073096666348e-05}
{"train_loss": 0.3529144525527954, "global_step": 40413, "epoch": 454, "lr": 9.474047214726529e-05}
{"train_loss": 0.32947492599487305, "global_step": 40414, "epoch": 454, "lr": 9.47402133218523e-05}
{"train_loss": 0.31421348452568054, "global_step": 40415, "epoch": 454, "lr": 9.473995449042453e-05}
{"train_loss": 0.3276318907737732, "global_step": 40416, "epoch": 454, "lr": 9.473969565298203e-05}
{"train_loss": 0.24546626210212708, "global_step": 40417, "epoch": 454, "lr": 9.473943680952479e-05}
{"train_loss": 0.3640284240245819, "global_step": 40418, "epoch": 454, "lr": 9.473917796005292e-05}
{"train_loss": 0.2776016891002655, "global_step": 40419, "epoch": 454, "lr": 9.473891910456638e-05}
{"train_loss": 0.33701708912849426, "global_step": 40420, "epoch": 454, "lr": 9.473866024306526e-05}
{"train_loss": 0.27944934368133545, "global_step": 40421, "epoch": 454, "lr": 9.473840137554956e-05}
{"train_loss": 0.3082377016544342, "global_step": 40422, "epoch": 454, "lr": 9.473814250201934e-05}
{"train_loss": 0.38028424978256226, "global_step": 40423, "epoch": 454, "lr": 9.473788362247463e-05}
{"train_loss": 0.4569871425628662, "global_step": 40424, "epoch": 454, "lr": 9.473762473691544e-05}
{"train_loss": 0.24854408204555511, "global_step": 40425, "epoch": 454, "lr": 9.473736584534181e-05}
{"train_loss": 0.36833375692367554, "global_step": 40426, "epoch": 454, "lr": 9.473710694775379e-05}
{"train_loss": 0.2913263142108917, "global_step": 40427, "epoch": 454, "lr": 9.473684804415143e-05}
{"train_loss": 0.2664305865764618, "global_step": 40428, "epoch": 454, "lr": 9.473658913453473e-05}
{"train_loss": 0.26391592621803284, "global_step": 40429, "epoch": 454, "lr": 9.473633021890374e-05}
{"train_loss": 0.40880894660949707, "global_step": 40430, "epoch": 454, "lr": 9.473607129725851e-05}
{"train_loss": 0.28506144881248474, "global_step": 40431, "epoch": 454, "lr": 9.473581236959905e-05}
{"train_loss": 0.3982900083065033, "global_step": 40432, "epoch": 454, "lr": 9.473555343592542e-05}
{"train_loss": 0.28906765580177307, "global_step": 40433, "epoch": 454, "lr": 9.473529449623762e-05}
{"train_loss": 0.39539673924446106, "global_step": 40434, "epoch": 454, "lr": 9.473503555053571e-05}
{"train_loss": 0.36196768283843994, "global_step": 40435, "epoch": 454, "lr": 9.473477659881973e-05}
{"train_loss": 0.3779715597629547, "global_step": 40436, "epoch": 454, "lr": 9.473451764108969e-05}
{"train_loss": 0.23579421639442444, "global_step": 40437, "epoch": 454, "lr": 9.473425867734566e-05}
{"train_loss": 0.32457971572875977, "global_step": 40438, "epoch": 454, "lr": 9.473399970758763e-05}
{"train_loss": 0.32568198442459106, "global_step": 40439, "epoch": 454, "lr": 9.473374073181566e-05}
{"train_loss": 0.29093101620674133, "global_step": 40440, "epoch": 454, "lr": 9.47334817500298e-05}
{"train_loss": 0.5813766717910767, "global_step": 40441, "epoch": 454, "lr": 9.473322276223006e-05}
{"train_loss": 0.300640732049942, "global_step": 40442, "epoch": 454, "lr": 9.47329637684165e-05}
{"train_loss": 0.42545637488365173, "global_step": 40443, "epoch": 454, "lr": 9.473270476858914e-05}
{"train_loss": 0.27378904819488525, "global_step": 40444, "epoch": 454, "lr": 9.473244576274798e-05}
{"train_loss": 0.2983364760875702, "global_step": 40445, "epoch": 454, "lr": 9.47321867508931e-05}
{"train_loss": 0.30699652433395386, "global_step": 40446, "epoch": 454, "lr": 9.473192773302453e-05}
{"train_loss": 0.39349764585494995, "global_step": 40447, "epoch": 454, "lr": 9.47316687091423e-05}
{"train_loss": 0.2459055483341217, "global_step": 40448, "epoch": 454, "lr": 9.473140967924643e-05}
{"train_loss": 0.33092376589775085, "global_step": 40449, "epoch": 454, "lr": 9.473115064333697e-05}
{"train_loss": 0.39558812975883484, "global_step": 40450, "epoch": 454, "lr": 9.473089160141398e-05}
{"train_loss": 0.29501938819885254, "global_step": 40451, "epoch": 454, "lr": 9.473063255347742e-05}
{"train_loss": 0.38995426893234253, "global_step": 40452, "epoch": 454, "lr": 9.47303734995274e-05}
{"train_loss": 0.3363553285598755, "global_step": 40453, "epoch": 454, "lr": 9.473011443956393e-05}
{"train_loss": 0.27433711290359497, "global_step": 40454, "epoch": 454, "lr": 9.472985537358703e-05}
{"train_loss": 0.3257709741592407, "global_step": 40455, "epoch": 454, "lr": 9.472959630159674e-05}
{"train_loss": 0.2776932418346405, "global_step": 40456, "epoch": 454, "lr": 9.472933722359312e-05}
{"train_loss": 0.3662443161010742, "global_step": 40457, "epoch": 454, "lr": 9.472907813957618e-05}
{"train_loss": 0.21764108538627625, "global_step": 40458, "epoch": 454, "lr": 9.472881904954596e-05}
{"train_loss": 0.3494422733783722, "global_step": 40459, "epoch": 454, "lr": 9.472855995350249e-05}
{"train_loss": 0.3456285297870636, "global_step": 40460, "epoch": 454, "lr": 9.472830085144581e-05}
{"train_loss": 0.4048992693424225, "global_step": 40461, "epoch": 454, "lr": 9.472804174337595e-05}
{"train_loss": 0.27981823682785034, "global_step": 40462, "epoch": 454, "lr": 9.472778262929296e-05}
{"train_loss": 0.25111427903175354, "global_step": 40463, "epoch": 454, "lr": 9.472752350919687e-05}
{"train_loss": 0.3158406615257263, "global_step": 40464, "epoch": 454, "lr": 9.47272643830877e-05}
{"train_loss": 0.4087755084037781, "global_step": 40465, "epoch": 454, "lr": 9.472700525096549e-05}
{"train_loss": 0.22276383638381958, "global_step": 40466, "epoch": 454, "lr": 9.472674611283029e-05}
{"train_loss": 0.34652474522590637, "global_step": 40467, "epoch": 454, "lr": 9.472648696868212e-05}
{"train_loss": 0.3134433627128601, "global_step": 40468, "epoch": 454, "lr": 9.472622781852102e-05}
{"train_loss": 0.3591800332069397, "global_step": 40469, "epoch": 454, "lr": 9.472596866234701e-05}
{"train_loss": 0.3147139549255371, "global_step": 40470, "epoch": 454, "lr": 9.472570950016016e-05}
{"train_loss": 0.28079721331596375, "global_step": 40471, "epoch": 454, "lr": 9.472545033196047e-05}
{"train_loss": 0.32560303807258606, "global_step": 40472, "epoch": 454, "lr": 9.472519115774801e-05}
{"train_loss": 0.27952736616134644, "global_step": 40473, "epoch": 454, "lr": 9.472493197752277e-05}
{"train_loss": 0.3355414867401123, "global_step": 40474, "epoch": 454, "lr": 9.472467279128481e-05}
{"train_loss": 0.23924727737903595, "global_step": 40475, "epoch": 454, "lr": 9.472441359903417e-05}
{"train_loss": 0.23283794522285461, "global_step": 40476, "epoch": 454, "lr": 9.472415440077087e-05}
{"train_loss": 0.42937394976615906, "global_step": 40477, "epoch": 454, "lr": 9.472389519649496e-05}
{"train_loss": 0.35114288330078125, "global_step": 40478, "epoch": 454, "lr": 9.472363598620646e-05}
{"train_loss": 0.27908751368522644, "global_step": 40479, "epoch": 454, "lr": 9.472337676990542e-05}
{"train_loss": 0.37961313128471375, "global_step": 40480, "epoch": 454, "lr": 9.472311754759187e-05}
{"train_loss": 0.30133500695228577, "global_step": 40481, "epoch": 454, "lr": 9.472285831926583e-05}
{"train_loss": 0.23811760544776917, "global_step": 40482, "epoch": 454, "lr": 9.472259908492736e-05}
{"train_loss": 0.4265953004360199, "global_step": 40483, "epoch": 454, "lr": 9.472233984457646e-05}
{"train_loss": 0.26372820138931274, "global_step": 40484, "epoch": 454, "lr": 9.472208059821321e-05}
{"train_loss": 0.29872187972068787, "global_step": 40485, "epoch": 454, "lr": 9.47218213458376e-05}
{"train_loss": 0.3008875846862793, "global_step": 40486, "epoch": 454, "lr": 9.472156208744971e-05}
{"train_loss": 0.3559251129627228, "global_step": 40487, "epoch": 454, "lr": 9.472130282304954e-05}
{"train_loss": 0.3519216477870941, "global_step": 40488, "epoch": 454, "lr": 9.472104355263714e-05}
{"train_loss": 0.3074999153614044, "global_step": 40489, "epoch": 454, "lr": 9.472078427621254e-05}
{"train_loss": 0.27952733635902405, "global_step": 40490, "epoch": 454, "lr": 9.472052499377576e-05}
{"train_loss": 0.3556499183177948, "global_step": 40491, "epoch": 454, "lr": 9.472026570532686e-05}
{"train_loss": 0.274863064289093, "global_step": 40492, "epoch": 454, "lr": 9.472000641086587e-05}
{"train_loss": 0.3190840780735016, "global_step": 40493, "epoch": 454, "lr": 9.471974711039282e-05}
{"train_loss": 0.32363960280847015, "global_step": 40494, "epoch": 454, "lr": 9.471948780390774e-05, "val_loss": 2.682948589324951}
{"train_loss": 0.28289780020713806, "global_step": 40495, "epoch": 455, "lr": 9.471922849141067e-05}
{"train_loss": 0.386902779340744, "global_step": 40496, "epoch": 455, "lr": 9.471896917290165e-05}
{"train_loss": 0.3485047519207001, "global_step": 40497, "epoch": 455, "lr": 9.471870984838072e-05}
{"train_loss": 0.3012728989124298, "global_step": 40498, "epoch": 455, "lr": 9.471845051784788e-05}
{"train_loss": 0.3148486316204071, "global_step": 40499, "epoch": 455, "lr": 9.471819118130321e-05}
{"train_loss": 0.3576129674911499, "global_step": 40500, "epoch": 455, "lr": 9.47179318387467e-05}
{"train_loss": 0.3038712441921234, "global_step": 40501, "epoch": 455, "lr": 9.471767249017843e-05}
{"train_loss": 0.3012476861476898, "global_step": 40502, "epoch": 455, "lr": 9.471741313559841e-05}
{"train_loss": 0.33641502261161804, "global_step": 40503, "epoch": 455, "lr": 9.471715377500668e-05}
{"train_loss": 0.2997548580169678, "global_step": 40504, "epoch": 455, "lr": 9.471689440840326e-05}
{"train_loss": 0.24322442710399628, "global_step": 40505, "epoch": 455, "lr": 9.471663503578822e-05}
{"train_loss": 0.2606329619884491, "global_step": 40506, "epoch": 455, "lr": 9.471637565716157e-05}
{"train_loss": 0.3884289264678955, "global_step": 40507, "epoch": 455, "lr": 9.471611627252333e-05}
{"train_loss": 0.36626550555229187, "global_step": 40508, "epoch": 455, "lr": 9.471585688187357e-05}
{"train_loss": 0.22543197870254517, "global_step": 40509, "epoch": 455, "lr": 9.471559748521231e-05}
{"train_loss": 0.2684306204319, "global_step": 40510, "epoch": 455, "lr": 9.471533808253956e-05}
{"train_loss": 0.2788563370704651, "global_step": 40511, "epoch": 455, "lr": 9.47150786738554e-05}
{"train_loss": 0.42253196239471436, "global_step": 40512, "epoch": 455, "lr": 9.471481925915984e-05}
{"train_loss": 0.24899879097938538, "global_step": 40513, "epoch": 455, "lr": 9.47145598384529e-05}
{"train_loss": 0.19783441722393036, "global_step": 40514, "epoch": 455, "lr": 9.471430041173466e-05}
{"train_loss": 0.239481121301651, "global_step": 40515, "epoch": 455, "lr": 9.47140409790051e-05}
{"train_loss": 0.24803847074508667, "global_step": 40516, "epoch": 455, "lr": 9.471378154026431e-05}
{"train_loss": 0.3128966987133026, "global_step": 40517, "epoch": 455, "lr": 9.471352209551228e-05}
{"train_loss": 0.2071678787469864, "global_step": 40518, "epoch": 455, "lr": 9.471326264474906e-05}
{"train_loss": 0.27665412425994873, "global_step": 40519, "epoch": 455, "lr": 9.471300318797469e-05}
{"train_loss": 0.337078332901001, "global_step": 40520, "epoch": 455, "lr": 9.47127437251892e-05}
{"train_loss": 0.3450927436351776, "global_step": 40521, "epoch": 455, "lr": 9.471248425639263e-05}
{"train_loss": 0.30759167671203613, "global_step": 40522, "epoch": 455, "lr": 9.471222478158501e-05}
{"train_loss": 0.29429373145103455, "global_step": 40523, "epoch": 455, "lr": 9.471196530076638e-05}
{"train_loss": 0.313730388879776, "global_step": 40524, "epoch": 455, "lr": 9.471170581393676e-05}
{"train_loss": 0.3103436827659607, "global_step": 40525, "epoch": 455, "lr": 9.47114463210962e-05}
{"train_loss": 0.4503418505191803, "global_step": 40526, "epoch": 455, "lr": 9.471118682224475e-05}
{"train_loss": 0.3478114604949951, "global_step": 40527, "epoch": 455, "lr": 9.471092731738241e-05}
{"train_loss": 0.3338756859302521, "global_step": 40528, "epoch": 455, "lr": 9.471066780650924e-05}
{"train_loss": 0.3331937789916992, "global_step": 40529, "epoch": 455, "lr": 9.471040828962525e-05}
{"train_loss": 0.2750760316848755, "global_step": 40530, "epoch": 455, "lr": 9.47101487667305e-05}
{"train_loss": 0.2669874429702759, "global_step": 40531, "epoch": 455, "lr": 9.470988923782501e-05}
{"train_loss": 0.2649587094783783, "global_step": 40532, "epoch": 455, "lr": 9.470962970290883e-05}
{"train_loss": 0.3068738281726837, "global_step": 40533, "epoch": 455, "lr": 9.470937016198198e-05}
{"train_loss": 0.36347857117652893, "global_step": 40534, "epoch": 455, "lr": 9.47091106150445e-05}
{"train_loss": 0.38650068640708923, "global_step": 40535, "epoch": 455, "lr": 9.470885106209641e-05}
{"train_loss": 0.38023751974105835, "global_step": 40536, "epoch": 455, "lr": 9.470859150313777e-05}
{"train_loss": 0.2697155773639679, "global_step": 40537, "epoch": 455, "lr": 9.470833193816862e-05}
{"train_loss": 0.2652474045753479, "global_step": 40538, "epoch": 455, "lr": 9.470807236718897e-05}
{"train_loss": 0.24630945920944214, "global_step": 40539, "epoch": 455, "lr": 9.470781279019887e-05}
{"train_loss": 0.2720951735973358, "global_step": 40540, "epoch": 455, "lr": 9.470755320719833e-05}
{"train_loss": 0.3002091944217682, "global_step": 40541, "epoch": 455, "lr": 9.470729361818742e-05}
{"train_loss": 0.2760455012321472, "global_step": 40542, "epoch": 455, "lr": 9.470703402316616e-05}
{"train_loss": 0.34650471806526184, "global_step": 40543, "epoch": 455, "lr": 9.470677442213458e-05}
{"train_loss": 0.4447177052497864, "global_step": 40544, "epoch": 455, "lr": 9.470651481509272e-05}
{"train_loss": 0.31697389483451843, "global_step": 40545, "epoch": 455, "lr": 9.470625520204062e-05}
{"train_loss": 0.3367277979850769, "global_step": 40546, "epoch": 455, "lr": 9.47059955829783e-05}
{"train_loss": 0.3217746913433075, "global_step": 40547, "epoch": 455, "lr": 9.47057359579058e-05}
{"train_loss": 0.29721370339393616, "global_step": 40548, "epoch": 455, "lr": 9.470547632682317e-05}
{"train_loss": 0.33716824650764465, "global_step": 40549, "epoch": 455, "lr": 9.470521668973043e-05}
{"train_loss": 0.2496790885925293, "global_step": 40550, "epoch": 455, "lr": 9.470495704662761e-05}
{"train_loss": 0.3066345155239105, "global_step": 40551, "epoch": 455, "lr": 9.470469739751477e-05}
{"train_loss": 0.27811312675476074, "global_step": 40552, "epoch": 455, "lr": 9.470443774239192e-05}
{"train_loss": 0.38826197385787964, "global_step": 40553, "epoch": 455, "lr": 9.47041780812591e-05}
{"train_loss": 0.3397452235221863, "global_step": 40554, "epoch": 455, "lr": 9.470391841411634e-05}
{"train_loss": 0.2594609558582306, "global_step": 40555, "epoch": 455, "lr": 9.470365874096369e-05}
{"train_loss": 0.27747678756713867, "global_step": 40556, "epoch": 455, "lr": 9.470339906180119e-05}
{"train_loss": 0.4131093919277191, "global_step": 40557, "epoch": 455, "lr": 9.470313937662885e-05}
{"train_loss": 0.25672033429145813, "global_step": 40558, "epoch": 455, "lr": 9.470287968544674e-05}
{"train_loss": 0.3415287733078003, "global_step": 40559, "epoch": 455, "lr": 9.470261998825484e-05}
{"train_loss": 0.23784518241882324, "global_step": 40560, "epoch": 455, "lr": 9.470236028505323e-05}
{"train_loss": 0.23318977653980255, "global_step": 40561, "epoch": 455, "lr": 9.470210057584194e-05}
{"train_loss": 0.350245863199234, "global_step": 40562, "epoch": 455, "lr": 9.4701840860621e-05}
{"train_loss": 0.33054524660110474, "global_step": 40563, "epoch": 455, "lr": 9.470158113939042e-05}
{"train_loss": 0.18710902333259583, "global_step": 40564, "epoch": 455, "lr": 9.470132141215027e-05}
{"train_loss": 0.3266977071762085, "global_step": 40565, "epoch": 455, "lr": 9.470106167890058e-05}
{"train_loss": 0.4116083085536957, "global_step": 40566, "epoch": 455, "lr": 9.470080193964137e-05}
{"train_loss": 0.28089800477027893, "global_step": 40567, "epoch": 455, "lr": 9.470054219437269e-05}
{"train_loss": 0.37662187218666077, "global_step": 40568, "epoch": 455, "lr": 9.470028244309456e-05}
{"train_loss": 0.30635571479797363, "global_step": 40569, "epoch": 455, "lr": 9.470002268580702e-05}
{"train_loss": 0.3811393082141876, "global_step": 40570, "epoch": 455, "lr": 9.46997629225101e-05}
{"train_loss": 0.293602854013443, "global_step": 40571, "epoch": 455, "lr": 9.469950315320386e-05}
{"train_loss": 0.30421167612075806, "global_step": 40572, "epoch": 455, "lr": 9.46992433778883e-05}
{"train_loss": 0.2924831211566925, "global_step": 40573, "epoch": 455, "lr": 9.469898359656348e-05}
{"train_loss": 0.31190812587738037, "global_step": 40574, "epoch": 455, "lr": 9.469872380922942e-05}
{"train_loss": 0.42687368392944336, "global_step": 40575, "epoch": 455, "lr": 9.469846401588617e-05}
{"train_loss": 0.37611839175224304, "global_step": 40576, "epoch": 455, "lr": 9.469820421653375e-05}
{"train_loss": 0.30414441227912903, "global_step": 40577, "epoch": 455, "lr": 9.469794441117221e-05}
{"train_loss": 0.3139340877532959, "global_step": 40578, "epoch": 455, "lr": 9.469768459980158e-05}
{"train_loss": 0.27698248624801636, "global_step": 40579, "epoch": 455, "lr": 9.469742478242188e-05}
{"train_loss": 0.3241242468357086, "global_step": 40580, "epoch": 455, "lr": 9.469716495903315e-05}
{"train_loss": 0.2704441249370575, "global_step": 40581, "epoch": 455, "lr": 9.469690512963544e-05}
{"train_loss": 0.27033862471580505, "global_step": 40582, "epoch": 455, "lr": 9.469664529422877e-05}
{"train_loss": 0.31082796514703986, "global_step": 40583, "epoch": 455, "lr": 9.469638545281318e-05, "val_loss": 2.6750450134277344, "train_action_mse_error": 44.06436538696289}
{"train_loss": 0.31765708327293396, "global_step": 40584, "epoch": 456, "lr": 9.469612560538872e-05}
{"train_loss": 0.25662824511528015, "global_step": 40585, "epoch": 456, "lr": 9.46958657519554e-05}
{"train_loss": 0.2517378032207489, "global_step": 40586, "epoch": 456, "lr": 9.469560589251326e-05}
{"train_loss": 0.39385688304901123, "global_step": 40587, "epoch": 456, "lr": 9.469534602706235e-05}
{"train_loss": 0.30580008029937744, "global_step": 40588, "epoch": 456, "lr": 9.46950861556027e-05}
{"train_loss": 0.3383359909057617, "global_step": 40589, "epoch": 456, "lr": 9.469482627813433e-05}
{"train_loss": 0.29454004764556885, "global_step": 40590, "epoch": 456, "lr": 9.469456639465728e-05}
{"train_loss": 0.2370838224887848, "global_step": 40591, "epoch": 456, "lr": 9.469430650517161e-05}
{"train_loss": 0.41076475381851196, "global_step": 40592, "epoch": 456, "lr": 9.469404660967731e-05}
{"train_loss": 0.41710570454597473, "global_step": 40593, "epoch": 456, "lr": 9.469378670817446e-05}
{"train_loss": 0.3341856002807617, "global_step": 40594, "epoch": 456, "lr": 9.469352680066307e-05}
{"train_loss": 0.24519860744476318, "global_step": 40595, "epoch": 456, "lr": 9.469326688714317e-05}
{"train_loss": 0.2751272916793823, "global_step": 40596, "epoch": 456, "lr": 9.469300696761482e-05}
{"train_loss": 0.3223205804824829, "global_step": 40597, "epoch": 456, "lr": 9.469274704207804e-05}
{"train_loss": 0.2743748128414154, "global_step": 40598, "epoch": 456, "lr": 9.469248711053285e-05}
{"train_loss": 0.2445736527442932, "global_step": 40599, "epoch": 456, "lr": 9.469222717297932e-05}
{"train_loss": 0.33200690150260925, "global_step": 40600, "epoch": 456, "lr": 9.469196722941744e-05}
{"train_loss": 0.33340731263160706, "global_step": 40601, "epoch": 456, "lr": 9.469170727984728e-05}
{"train_loss": 0.3385169506072998, "global_step": 40602, "epoch": 456, "lr": 9.469144732426889e-05}
{"train_loss": 0.35718613862991333, "global_step": 40603, "epoch": 456, "lr": 9.469118736268226e-05}
{"train_loss": 0.35084426403045654, "global_step": 40604, "epoch": 456, "lr": 9.469092739508743e-05}
{"train_loss": 0.3109288215637207, "global_step": 40605, "epoch": 456, "lr": 9.469066742148447e-05}
{"train_loss": 0.24940495193004608, "global_step": 40606, "epoch": 456, "lr": 9.469040744187339e-05}
{"train_loss": 0.4216946065425873, "global_step": 40607, "epoch": 456, "lr": 9.469014745625421e-05}
{"train_loss": 0.3755905032157898, "global_step": 40608, "epoch": 456, "lr": 9.468988746462702e-05}
{"train_loss": 0.24670593440532684, "global_step": 40609, "epoch": 456, "lr": 9.468962746699179e-05}
{"train_loss": 0.341031014919281, "global_step": 40610, "epoch": 456, "lr": 9.468936746334858e-05}
{"train_loss": 0.35889238119125366, "global_step": 40611, "epoch": 456, "lr": 9.468910745369746e-05}
{"train_loss": 0.24182240664958954, "global_step": 40612, "epoch": 456, "lr": 9.46888474380384e-05}
{"train_loss": 0.37531107664108276, "global_step": 40613, "epoch": 456, "lr": 9.46885874163715e-05}
{"train_loss": 0.24523860216140747, "global_step": 40614, "epoch": 456, "lr": 9.468832738869674e-05}
{"train_loss": 0.3568805456161499, "global_step": 40615, "epoch": 456, "lr": 9.46880673550142e-05}
{"train_loss": 0.4008553922176361, "global_step": 40616, "epoch": 456, "lr": 9.468780731532388e-05}
{"train_loss": 0.2992387115955353, "global_step": 40617, "epoch": 456, "lr": 9.468754726962583e-05}
{"train_loss": 0.2782181203365326, "global_step": 40618, "epoch": 456, "lr": 9.468728721792008e-05}
{"train_loss": 0.3488337993621826, "global_step": 40619, "epoch": 456, "lr": 9.468702716020668e-05}
{"train_loss": 0.3297135829925537, "global_step": 40620, "epoch": 456, "lr": 9.468676709648565e-05}
{"train_loss": 0.361030250787735, "global_step": 40621, "epoch": 456, "lr": 9.468650702675703e-05}
{"train_loss": 0.2865321636199951, "global_step": 40622, "epoch": 456, "lr": 9.468624695102085e-05}
{"train_loss": 0.363998681306839, "global_step": 40623, "epoch": 456, "lr": 9.468598686927715e-05}
{"train_loss": 0.2860097289085388, "global_step": 40624, "epoch": 456, "lr": 9.468572678152596e-05}
{"train_loss": 0.28885185718536377, "global_step": 40625, "epoch": 456, "lr": 9.468546668776733e-05}
{"train_loss": 0.25327277183532715, "global_step": 40626, "epoch": 456, "lr": 9.468520658800128e-05}
{"train_loss": 0.3444101810455322, "global_step": 40627, "epoch": 456, "lr": 9.468494648222784e-05}
{"train_loss": 0.3921731114387512, "global_step": 40628, "epoch": 456, "lr": 9.468468637044707e-05}
{"train_loss": 0.4481349289417267, "global_step": 40629, "epoch": 456, "lr": 9.468442625265896e-05}
{"train_loss": 0.36600053310394287, "global_step": 40630, "epoch": 456, "lr": 9.46841661288636e-05}
{"train_loss": 0.3502185046672821, "global_step": 40631, "epoch": 456, "lr": 9.468390599906099e-05}
{"train_loss": 0.48458167910575867, "global_step": 40632, "epoch": 456, "lr": 9.468364586325117e-05}
{"train_loss": 0.37932175397872925, "global_step": 40633, "epoch": 456, "lr": 9.468338572143418e-05}
{"train_loss": 0.3597182631492615, "global_step": 40634, "epoch": 456, "lr": 9.468312557361005e-05}
{"train_loss": 0.30800357460975647, "global_step": 40635, "epoch": 456, "lr": 9.468286541977883e-05}
{"train_loss": 0.27087125182151794, "global_step": 40636, "epoch": 456, "lr": 9.468260525994054e-05}
{"train_loss": 0.30533304810523987, "global_step": 40637, "epoch": 456, "lr": 9.468234509409521e-05}
{"train_loss": 0.29430586099624634, "global_step": 40638, "epoch": 456, "lr": 9.468208492224289e-05}
{"train_loss": 0.30716726183891296, "global_step": 40639, "epoch": 456, "lr": 9.468182474438361e-05}
{"train_loss": 0.3404368460178375, "global_step": 40640, "epoch": 456, "lr": 9.46815645605174e-05}
{"train_loss": 0.2676270008087158, "global_step": 40641, "epoch": 456, "lr": 9.46813043706443e-05}
{"train_loss": 0.35880234837532043, "global_step": 40642, "epoch": 456, "lr": 9.468104417476433e-05}
{"train_loss": 0.3798280358314514, "global_step": 40643, "epoch": 456, "lr": 9.468078397287756e-05}
{"train_loss": 0.39811640977859497, "global_step": 40644, "epoch": 456, "lr": 9.468052376498399e-05}
{"train_loss": 0.37703362107276917, "global_step": 40645, "epoch": 456, "lr": 9.468026355108368e-05}
{"train_loss": 0.30109015107154846, "global_step": 40646, "epoch": 456, "lr": 9.468000333117664e-05}
{"train_loss": 0.3969682455062866, "global_step": 40647, "epoch": 456, "lr": 9.46797431052629e-05}
{"train_loss": 0.2116130143404007, "global_step": 40648, "epoch": 456, "lr": 9.467948287334255e-05}
{"train_loss": 0.312858521938324, "global_step": 40649, "epoch": 456, "lr": 9.467922263541558e-05}
{"train_loss": 0.19881683588027954, "global_step": 40650, "epoch": 456, "lr": 9.467896239148202e-05}
{"train_loss": 0.3571295738220215, "global_step": 40651, "epoch": 456, "lr": 9.467870214154192e-05}
{"train_loss": 0.33919578790664673, "global_step": 40652, "epoch": 456, "lr": 9.467844188559531e-05}
{"train_loss": 0.4511319398880005, "global_step": 40653, "epoch": 456, "lr": 9.467818162364225e-05}
{"train_loss": 0.37189921736717224, "global_step": 40654, "epoch": 456, "lr": 9.467792135568274e-05}
{"train_loss": 0.2661873996257782, "global_step": 40655, "epoch": 456, "lr": 9.467766108171682e-05}
{"train_loss": 0.38817742466926575, "global_step": 40656, "epoch": 456, "lr": 9.467740080174454e-05}
{"train_loss": 0.3738032877445221, "global_step": 40657, "epoch": 456, "lr": 9.467714051576592e-05}
{"train_loss": 0.28368160128593445, "global_step": 40658, "epoch": 456, "lr": 9.467688022378102e-05}
{"train_loss": 0.4376614987850189, "global_step": 40659, "epoch": 456, "lr": 9.467661992578985e-05}
{"train_loss": 0.2821570932865143, "global_step": 40660, "epoch": 456, "lr": 9.467635962179244e-05}
{"train_loss": 0.34121543169021606, "global_step": 40661, "epoch": 456, "lr": 9.467609931178885e-05}
{"train_loss": 0.26765012741088867, "global_step": 40662, "epoch": 456, "lr": 9.467583899577911e-05}
{"train_loss": 0.30364924669265747, "global_step": 40663, "epoch": 456, "lr": 9.467557867376324e-05}
{"train_loss": 0.2803191840648651, "global_step": 40664, "epoch": 456, "lr": 9.467531834574129e-05}
{"train_loss": 0.3728611469268799, "global_step": 40665, "epoch": 456, "lr": 9.467505801171327e-05}
{"train_loss": 0.25388845801353455, "global_step": 40666, "epoch": 456, "lr": 9.467479767167923e-05}
{"train_loss": 0.2986608147621155, "global_step": 40667, "epoch": 456, "lr": 9.467453732563925e-05}
{"train_loss": 0.2604998052120209, "global_step": 40668, "epoch": 456, "lr": 9.467427697359328e-05}
{"train_loss": 0.329233855009079, "global_step": 40669, "epoch": 456, "lr": 9.467401661554142e-05}
{"train_loss": 0.30545979738235474, "global_step": 40670, "epoch": 456, "lr": 9.467375625148367e-05}
{"train_loss": 0.30740344524383545, "global_step": 40671, "epoch": 456, "lr": 9.46734958814201e-05}
{"train_loss": 0.3269558120309637, "global_step": 40672, "epoch": 456, "lr": 9.46732355053507e-05, "val_loss": 2.7216553688049316}
{"train_loss": 0.38970088958740234, "global_step": 40673, "epoch": 457, "lr": 9.467297512327553e-05}
{"train_loss": 0.30990707874298096, "global_step": 40674, "epoch": 457, "lr": 9.467271473519462e-05}
{"train_loss": 0.36926591396331787, "global_step": 40675, "epoch": 457, "lr": 9.467245434110802e-05}
{"train_loss": 0.3464452922344208, "global_step": 40676, "epoch": 457, "lr": 9.467219394101576e-05}
{"train_loss": 0.5375252962112427, "global_step": 40677, "epoch": 457, "lr": 9.467193353491784e-05}
{"train_loss": 0.20423176884651184, "global_step": 40678, "epoch": 457, "lr": 9.467167312281435e-05}
{"train_loss": 0.29468193650245667, "global_step": 40679, "epoch": 457, "lr": 9.467141270470529e-05}
{"train_loss": 0.3482716381549835, "global_step": 40680, "epoch": 457, "lr": 9.467115228059069e-05}
{"train_loss": 0.4279412031173706, "global_step": 40681, "epoch": 457, "lr": 9.467089185047062e-05}
{"train_loss": 0.3350464105606079, "global_step": 40682, "epoch": 457, "lr": 9.467063141434507e-05}
{"train_loss": 0.3088412880897522, "global_step": 40683, "epoch": 457, "lr": 9.467037097221412e-05}
{"train_loss": 0.3565926253795624, "global_step": 40684, "epoch": 457, "lr": 9.467011052407776e-05}
{"train_loss": 0.2927226126194, "global_step": 40685, "epoch": 457, "lr": 9.466985006993607e-05}
{"train_loss": 0.3719150722026825, "global_step": 40686, "epoch": 457, "lr": 9.466958960978905e-05}
{"train_loss": 0.2876833379268646, "global_step": 40687, "epoch": 457, "lr": 9.466932914363675e-05}
{"train_loss": 0.26039034128189087, "global_step": 40688, "epoch": 457, "lr": 9.466906867147921e-05}
{"train_loss": 0.2589403986930847, "global_step": 40689, "epoch": 457, "lr": 9.466880819331645e-05}
{"train_loss": 0.31472620368003845, "global_step": 40690, "epoch": 457, "lr": 9.466854770914852e-05}
{"train_loss": 0.2291399985551834, "global_step": 40691, "epoch": 457, "lr": 9.466828721897545e-05}
{"train_loss": 0.3216293752193451, "global_step": 40692, "epoch": 457, "lr": 9.466802672279727e-05}
{"train_loss": 0.3839406669139862, "global_step": 40693, "epoch": 457, "lr": 9.466776622061402e-05}
{"train_loss": 0.3550417423248291, "global_step": 40694, "epoch": 457, "lr": 9.466750571242573e-05}
{"train_loss": 0.33691859245300293, "global_step": 40695, "epoch": 457, "lr": 9.466724519823244e-05}
{"train_loss": 0.30824750661849976, "global_step": 40696, "epoch": 457, "lr": 9.466698467803417e-05}
{"train_loss": 0.28787851333618164, "global_step": 40697, "epoch": 457, "lr": 9.466672415183098e-05}
{"train_loss": 0.2927621901035309, "global_step": 40698, "epoch": 457, "lr": 9.466646361962291e-05}
{"train_loss": 0.30300086736679077, "global_step": 40699, "epoch": 457, "lr": 9.466620308140994e-05}
{"train_loss": 0.21889208257198334, "global_step": 40700, "epoch": 457, "lr": 9.466594253719216e-05}
{"train_loss": 0.29572874307632446, "global_step": 40701, "epoch": 457, "lr": 9.466568198696961e-05}
{"train_loss": 0.2790035903453827, "global_step": 40702, "epoch": 457, "lr": 9.466542143074229e-05}
{"train_loss": 0.35438063740730286, "global_step": 40703, "epoch": 457, "lr": 9.466516086851023e-05}
{"train_loss": 0.28168708086013794, "global_step": 40704, "epoch": 457, "lr": 9.46649003002735e-05}
{"train_loss": 0.2550637423992157, "global_step": 40705, "epoch": 457, "lr": 9.466463972603212e-05}
{"train_loss": 0.3193551003932953, "global_step": 40706, "epoch": 457, "lr": 9.466437914578612e-05}
{"train_loss": 0.25923845171928406, "global_step": 40707, "epoch": 457, "lr": 9.466411855953555e-05}
{"train_loss": 0.23827221989631653, "global_step": 40708, "epoch": 457, "lr": 9.466385796728042e-05}
{"train_loss": 0.2940596044063568, "global_step": 40709, "epoch": 457, "lr": 9.466359736902077e-05}
{"train_loss": 0.37292012572288513, "global_step": 40710, "epoch": 457, "lr": 9.466333676475668e-05}
{"train_loss": 0.2663610875606537, "global_step": 40711, "epoch": 457, "lr": 9.466307615448811e-05}
{"train_loss": 0.33127039670944214, "global_step": 40712, "epoch": 457, "lr": 9.466281553821515e-05}
{"train_loss": 0.26127418875694275, "global_step": 40713, "epoch": 457, "lr": 9.466255491593781e-05}
{"train_loss": 0.30912908911705017, "global_step": 40714, "epoch": 457, "lr": 9.466229428765614e-05}
{"train_loss": 0.3014364540576935, "global_step": 40715, "epoch": 457, "lr": 9.466203365337018e-05}
{"train_loss": 0.3599247336387634, "global_step": 40716, "epoch": 457, "lr": 9.466177301307993e-05}
{"train_loss": 0.374330073595047, "global_step": 40717, "epoch": 457, "lr": 9.466151236678547e-05}
{"train_loss": 0.31252798438072205, "global_step": 40718, "epoch": 457, "lr": 9.466125171448681e-05}
{"train_loss": 0.3562312722206116, "global_step": 40719, "epoch": 457, "lr": 9.466099105618398e-05}
{"train_loss": 0.39357927441596985, "global_step": 40720, "epoch": 457, "lr": 9.466073039187703e-05}
{"train_loss": 0.2840011715888977, "global_step": 40721, "epoch": 457, "lr": 9.4660469721566e-05}
{"train_loss": 0.2959703207015991, "global_step": 40722, "epoch": 457, "lr": 9.46602090452509e-05}
{"train_loss": 0.34620705246925354, "global_step": 40723, "epoch": 457, "lr": 9.465994836293177e-05}
{"train_loss": 0.3483424782752991, "global_step": 40724, "epoch": 457, "lr": 9.465968767460867e-05}
{"train_loss": 0.3914351463317871, "global_step": 40725, "epoch": 457, "lr": 9.465942698028162e-05}
{"train_loss": 0.2774483263492584, "global_step": 40726, "epoch": 457, "lr": 9.465916627995065e-05}
{"train_loss": 0.2785010039806366, "global_step": 40727, "epoch": 457, "lr": 9.46589055736158e-05}
{"train_loss": 0.21322588622570038, "global_step": 40728, "epoch": 457, "lr": 9.46586448612771e-05}
{"train_loss": 0.3125936686992645, "global_step": 40729, "epoch": 457, "lr": 9.46583841429346e-05}
{"train_loss": 0.2048925757408142, "global_step": 40730, "epoch": 457, "lr": 9.465812341858831e-05}
{"train_loss": 0.2411535084247589, "global_step": 40731, "epoch": 457, "lr": 9.465786268823828e-05}
{"train_loss": 0.34187933802604675, "global_step": 40732, "epoch": 457, "lr": 9.465760195188455e-05}
{"train_loss": 0.32080426812171936, "global_step": 40733, "epoch": 457, "lr": 9.465734120952716e-05}
{"train_loss": 0.22145356237888336, "global_step": 40734, "epoch": 457, "lr": 9.465708046116613e-05}
{"train_loss": 0.3071986734867096, "global_step": 40735, "epoch": 457, "lr": 9.46568197068015e-05}
{"train_loss": 0.35996726155281067, "global_step": 40736, "epoch": 457, "lr": 9.465655894643331e-05}
{"train_loss": 0.2960377037525177, "global_step": 40737, "epoch": 457, "lr": 9.465629818006157e-05}
{"train_loss": 0.3590088486671448, "global_step": 40738, "epoch": 457, "lr": 9.465603740768635e-05}
{"train_loss": 0.24831944704055786, "global_step": 40739, "epoch": 457, "lr": 9.465577662930767e-05}
{"train_loss": 0.3548949360847473, "global_step": 40740, "epoch": 457, "lr": 9.465551584492555e-05}
{"train_loss": 0.40533286333084106, "global_step": 40741, "epoch": 457, "lr": 9.465525505454006e-05}
{"train_loss": 0.28490427136421204, "global_step": 40742, "epoch": 457, "lr": 9.465499425815121e-05}
{"train_loss": 0.409707635641098, "global_step": 40743, "epoch": 457, "lr": 9.465473345575904e-05}
{"train_loss": 0.319778174161911, "global_step": 40744, "epoch": 457, "lr": 9.465447264736358e-05}
{"train_loss": 0.2908627986907959, "global_step": 40745, "epoch": 457, "lr": 9.465421183296487e-05}
{"train_loss": 0.2759559154510498, "global_step": 40746, "epoch": 457, "lr": 9.465395101256295e-05}
{"train_loss": 0.2912837266921997, "global_step": 40747, "epoch": 457, "lr": 9.465369018615784e-05}
{"train_loss": 0.27115753293037415, "global_step": 40748, "epoch": 457, "lr": 9.46534293537496e-05}
{"train_loss": 0.4797975718975067, "global_step": 40749, "epoch": 457, "lr": 9.465316851533826e-05}
{"train_loss": 0.39889803528785706, "global_step": 40750, "epoch": 457, "lr": 9.465290767092382e-05}
{"train_loss": 0.2657621204853058, "global_step": 40751, "epoch": 457, "lr": 9.465264682050634e-05}
{"train_loss": 0.3353310227394104, "global_step": 40752, "epoch": 457, "lr": 9.465238596408587e-05}
{"train_loss": 0.3205910921096802, "global_step": 40753, "epoch": 457, "lr": 9.465212510166244e-05}
{"train_loss": 0.29294219613075256, "global_step": 40754, "epoch": 457, "lr": 9.465186423323607e-05}
{"train_loss": 0.336362361907959, "global_step": 40755, "epoch": 457, "lr": 9.465160335880679e-05}
{"train_loss": 0.2808169722557068, "global_step": 40756, "epoch": 457, "lr": 9.465134247837465e-05}
{"train_loss": 0.32459041476249695, "global_step": 40757, "epoch": 457, "lr": 9.465108159193968e-05}
{"train_loss": 0.37214553356170654, "global_step": 40758, "epoch": 457, "lr": 9.465082069950191e-05}
{"train_loss": 0.3186471164226532, "global_step": 40759, "epoch": 457, "lr": 9.46505598010614e-05}
{"train_loss": 0.2775876820087433, "global_step": 40760, "epoch": 457, "lr": 9.465029889661816e-05}
{"train_loss": 0.3160225290260958, "global_step": 40761, "epoch": 457, "lr": 9.465003798617223e-05, "val_loss": 2.7367019653320312}
{"train_loss": 0.2410253882408142, "global_step": 40762, "epoch": 458, "lr": 9.464977706972364e-05}
{"train_loss": 0.27702605724334717, "global_step": 40763, "epoch": 458, "lr": 9.464951614727245e-05}
{"train_loss": 0.37210091948509216, "global_step": 40764, "epoch": 458, "lr": 9.464925521881865e-05}
{"train_loss": 0.34850555658340454, "global_step": 40765, "epoch": 458, "lr": 9.464899428436231e-05}
{"train_loss": 0.5683518052101135, "global_step": 40766, "epoch": 458, "lr": 9.464873334390347e-05}
{"train_loss": 0.40433403849601746, "global_step": 40767, "epoch": 458, "lr": 9.464847239744215e-05}
{"train_loss": 0.33183032274246216, "global_step": 40768, "epoch": 458, "lr": 9.464821144497838e-05}
{"train_loss": 0.2905890941619873, "global_step": 40769, "epoch": 458, "lr": 9.46479504865122e-05}
{"train_loss": 0.3702484369277954, "global_step": 40770, "epoch": 458, "lr": 9.464768952204366e-05}
{"train_loss": 0.3787946403026581, "global_step": 40771, "epoch": 458, "lr": 9.464742855157277e-05}
{"train_loss": 0.3001430332660675, "global_step": 40772, "epoch": 458, "lr": 9.464716757509957e-05}
{"train_loss": 0.3052266836166382, "global_step": 40773, "epoch": 458, "lr": 9.464690659262412e-05}
{"train_loss": 0.36878088116645813, "global_step": 40774, "epoch": 458, "lr": 9.464664560414643e-05}
{"train_loss": 0.34057095646858215, "global_step": 40775, "epoch": 458, "lr": 9.464638460966654e-05}
{"train_loss": 0.25454431772232056, "global_step": 40776, "epoch": 458, "lr": 9.46461236091845e-05}
{"train_loss": 0.2395787537097931, "global_step": 40777, "epoch": 458, "lr": 9.464586260270031e-05}
{"train_loss": 0.23955285549163818, "global_step": 40778, "epoch": 458, "lr": 9.464560159021405e-05}
{"train_loss": 0.2037646472454071, "global_step": 40779, "epoch": 458, "lr": 9.464534057172573e-05}
{"train_loss": 0.3710242211818695, "global_step": 40780, "epoch": 458, "lr": 9.464507954723539e-05}
{"train_loss": 0.40537014603614807, "global_step": 40781, "epoch": 458, "lr": 9.464481851674306e-05}
{"train_loss": 0.2532079815864563, "global_step": 40782, "epoch": 458, "lr": 9.464455748024877e-05}
{"train_loss": 0.3192043602466583, "global_step": 40783, "epoch": 458, "lr": 9.464429643775257e-05}
{"train_loss": 0.26327598094940186, "global_step": 40784, "epoch": 458, "lr": 9.464403538925449e-05}
{"train_loss": 0.3952624797821045, "global_step": 40785, "epoch": 458, "lr": 9.464377433475456e-05}
{"train_loss": 0.25783824920654297, "global_step": 40786, "epoch": 458, "lr": 9.464351327425281e-05}
{"train_loss": 0.31465375423431396, "global_step": 40787, "epoch": 458, "lr": 9.46432522077493e-05}
{"train_loss": 0.34382879734039307, "global_step": 40788, "epoch": 458, "lr": 9.464299113524403e-05}
{"train_loss": 0.2536098062992096, "global_step": 40789, "epoch": 458, "lr": 9.464273005673708e-05}
{"train_loss": 0.426410973072052, "global_step": 40790, "epoch": 458, "lr": 9.464246897222843e-05}
{"train_loss": 0.35816967487335205, "global_step": 40791, "epoch": 458, "lr": 9.464220788171815e-05}
{"train_loss": 0.3870459496974945, "global_step": 40792, "epoch": 458, "lr": 9.464194678520628e-05}
{"train_loss": 0.24517321586608887, "global_step": 40793, "epoch": 458, "lr": 9.464168568269285e-05}
{"train_loss": 0.27180540561676025, "global_step": 40794, "epoch": 458, "lr": 9.464142457417789e-05}
{"train_loss": 0.2793010473251343, "global_step": 40795, "epoch": 458, "lr": 9.464116345966142e-05}
{"train_loss": 0.22987380623817444, "global_step": 40796, "epoch": 458, "lr": 9.464090233914348e-05}
{"train_loss": 0.38140472769737244, "global_step": 40797, "epoch": 458, "lr": 9.464064121262415e-05}
{"train_loss": 0.3685024380683899, "global_step": 40798, "epoch": 458, "lr": 9.46403800801034e-05}
{"train_loss": 0.3281349241733551, "global_step": 40799, "epoch": 458, "lr": 9.464011894158132e-05}
{"train_loss": 0.16849155724048615, "global_step": 40800, "epoch": 458, "lr": 9.46398577970579e-05}
{"train_loss": 0.23776686191558838, "global_step": 40801, "epoch": 458, "lr": 9.463959664653319e-05}
{"train_loss": 0.28328731656074524, "global_step": 40802, "epoch": 458, "lr": 9.463933549000725e-05}
{"train_loss": 0.3441419005393982, "global_step": 40803, "epoch": 458, "lr": 9.463907432748008e-05}
{"train_loss": 0.31066226959228516, "global_step": 40804, "epoch": 458, "lr": 9.463881315895174e-05}
{"train_loss": 0.24033673107624054, "global_step": 40805, "epoch": 458, "lr": 9.463855198442227e-05}
{"train_loss": 0.2672014534473419, "global_step": 40806, "epoch": 458, "lr": 9.463829080389167e-05}
{"train_loss": 0.3203999102115631, "global_step": 40807, "epoch": 458, "lr": 9.463802961736e-05}
{"train_loss": 0.2910659611225128, "global_step": 40808, "epoch": 458, "lr": 9.463776842482728e-05}
{"train_loss": 0.33903422951698303, "global_step": 40809, "epoch": 458, "lr": 9.463750722629357e-05}
{"train_loss": 0.273632287979126, "global_step": 40810, "epoch": 458, "lr": 9.463724602175887e-05}
{"train_loss": 0.2607012689113617, "global_step": 40811, "epoch": 458, "lr": 9.463698481122327e-05}
{"train_loss": 0.22322982549667358, "global_step": 40812, "epoch": 458, "lr": 9.463672359468675e-05}
{"train_loss": 0.43527406454086304, "global_step": 40813, "epoch": 458, "lr": 9.463646237214938e-05}
{"train_loss": 0.3529413342475891, "global_step": 40814, "epoch": 458, "lr": 9.463620114361117e-05}
{"train_loss": 0.2778615355491638, "global_step": 40815, "epoch": 458, "lr": 9.463593990907217e-05}
{"train_loss": 0.38642388582229614, "global_step": 40816, "epoch": 458, "lr": 9.463567866853241e-05}
{"train_loss": 0.49994999170303345, "global_step": 40817, "epoch": 458, "lr": 9.463541742199193e-05}
{"train_loss": 0.2830740511417389, "global_step": 40818, "epoch": 458, "lr": 9.463515616945076e-05}
{"train_loss": 0.365633100271225, "global_step": 40819, "epoch": 458, "lr": 9.463489491090893e-05}
{"train_loss": 0.30901721119880676, "global_step": 40820, "epoch": 458, "lr": 9.46346336463665e-05}
{"train_loss": 0.30248725414276123, "global_step": 40821, "epoch": 458, "lr": 9.463437237582348e-05}
{"train_loss": 0.39307311177253723, "global_step": 40822, "epoch": 458, "lr": 9.463411109927991e-05}
{"train_loss": 0.36005738377571106, "global_step": 40823, "epoch": 458, "lr": 9.463384981673583e-05}
{"train_loss": 0.2599498927593231, "global_step": 40824, "epoch": 458, "lr": 9.463358852819127e-05}
{"train_loss": 0.3510608971118927, "global_step": 40825, "epoch": 458, "lr": 9.463332723364627e-05}
{"train_loss": 0.33324679732322693, "global_step": 40826, "epoch": 458, "lr": 9.463306593310086e-05}
{"train_loss": 0.30658572912216187, "global_step": 40827, "epoch": 458, "lr": 9.463280462655508e-05}
{"train_loss": 0.33997082710266113, "global_step": 40828, "epoch": 458, "lr": 9.463254331400897e-05}
{"train_loss": 0.4050231873989105, "global_step": 40829, "epoch": 458, "lr": 9.463228199546256e-05}
{"train_loss": 0.2415647953748703, "global_step": 40830, "epoch": 458, "lr": 9.463202067091588e-05}
{"train_loss": 0.24640801548957825, "global_step": 40831, "epoch": 458, "lr": 9.463175934036897e-05}
{"train_loss": 0.3049435019493103, "global_step": 40832, "epoch": 458, "lr": 9.463149800382185e-05}
{"train_loss": 0.27481991052627563, "global_step": 40833, "epoch": 458, "lr": 9.463123666127458e-05}
{"train_loss": 0.3464595377445221, "global_step": 40834, "epoch": 458, "lr": 9.463097531272718e-05}
{"train_loss": 0.24313686788082123, "global_step": 40835, "epoch": 458, "lr": 9.463071395817969e-05}
{"train_loss": 0.24259623885154724, "global_step": 40836, "epoch": 458, "lr": 9.463045259763215e-05}
{"train_loss": 0.24760417640209198, "global_step": 40837, "epoch": 458, "lr": 9.463019123108459e-05}
{"train_loss": 0.2452893853187561, "global_step": 40838, "epoch": 458, "lr": 9.462992985853704e-05}
{"train_loss": 0.3219241797924042, "global_step": 40839, "epoch": 458, "lr": 9.462966847998955e-05}
{"train_loss": 0.27178141474723816, "global_step": 40840, "epoch": 458, "lr": 9.462940709544212e-05}
{"train_loss": 0.29588907957077026, "global_step": 40841, "epoch": 458, "lr": 9.462914570489482e-05}
{"train_loss": 0.3166479170322418, "global_step": 40842, "epoch": 458, "lr": 9.46288843083477e-05}
{"train_loss": 0.2334963083267212, "global_step": 40843, "epoch": 458, "lr": 9.462862290580074e-05}
{"train_loss": 0.3074491620063782, "global_step": 40844, "epoch": 458, "lr": 9.462836149725402e-05}
{"train_loss": 0.3785543739795685, "global_step": 40845, "epoch": 458, "lr": 9.462810008270757e-05}
{"train_loss": 0.2681741416454315, "global_step": 40846, "epoch": 458, "lr": 9.46278386621614e-05}
{"train_loss": 0.29413461685180664, "global_step": 40847, "epoch": 458, "lr": 9.462757723561558e-05}
{"train_loss": 0.40996113419532776, "global_step": 40848, "epoch": 458, "lr": 9.46273158030701e-05}
{"train_loss": 0.2737948000431061, "global_step": 40849, "epoch": 458, "lr": 9.462705436452503e-05}
{"train_loss": 0.3146426019708762, "global_step": 40850, "epoch": 458, "lr": 9.46267929199804e-05, "val_loss": 2.701340436935425}
{"train_loss": 0.3532874584197998, "global_step": 40851, "epoch": 459, "lr": 9.462653146943625e-05}
{"train_loss": 0.27454689145088196, "global_step": 40852, "epoch": 459, "lr": 9.462627001289259e-05}
{"train_loss": 0.29014691710472107, "global_step": 40853, "epoch": 459, "lr": 9.462600855034949e-05}
{"train_loss": 0.3124266266822815, "global_step": 40854, "epoch": 459, "lr": 9.462574708180695e-05}
{"train_loss": 0.29045185446739197, "global_step": 40855, "epoch": 459, "lr": 9.462548560726503e-05}
{"train_loss": 0.2841743230819702, "global_step": 40856, "epoch": 459, "lr": 9.462522412672375e-05}
{"train_loss": 0.22688740491867065, "global_step": 40857, "epoch": 459, "lr": 9.462496264018316e-05}
{"train_loss": 0.25696080923080444, "global_step": 40858, "epoch": 459, "lr": 9.462470114764329e-05}
{"train_loss": 0.2505210340023041, "global_step": 40859, "epoch": 459, "lr": 9.462443964910417e-05}
{"train_loss": 0.29329994320869446, "global_step": 40860, "epoch": 459, "lr": 9.462417814456584e-05}
{"train_loss": 0.3579535186290741, "global_step": 40861, "epoch": 459, "lr": 9.462391663402832e-05}
{"train_loss": 0.32574063539505005, "global_step": 40862, "epoch": 459, "lr": 9.462365511749167e-05}
{"train_loss": 0.22080963850021362, "global_step": 40863, "epoch": 459, "lr": 9.462339359495592e-05}
{"train_loss": 0.39228200912475586, "global_step": 40864, "epoch": 459, "lr": 9.462313206642108e-05}
{"train_loss": 0.27567631006240845, "global_step": 40865, "epoch": 459, "lr": 9.462287053188723e-05}
{"train_loss": 0.3483358919620514, "global_step": 40866, "epoch": 459, "lr": 9.462260899135435e-05}
{"train_loss": 0.33448249101638794, "global_step": 40867, "epoch": 459, "lr": 9.462234744482252e-05}
{"train_loss": 0.30656543374061584, "global_step": 40868, "epoch": 459, "lr": 9.462208589229176e-05}
{"train_loss": 0.2613237500190735, "global_step": 40869, "epoch": 459, "lr": 9.46218243337621e-05}
{"train_loss": 0.3717358112335205, "global_step": 40870, "epoch": 459, "lr": 9.462156276923359e-05}
{"train_loss": 0.3500441312789917, "global_step": 40871, "epoch": 459, "lr": 9.462130119870625e-05}
{"train_loss": 0.3091222047805786, "global_step": 40872, "epoch": 459, "lr": 9.46210396221801e-05}
{"train_loss": 0.35713309049606323, "global_step": 40873, "epoch": 459, "lr": 9.462077803965522e-05}
{"train_loss": 0.2926787734031677, "global_step": 40874, "epoch": 459, "lr": 9.462051645113161e-05}
{"train_loss": 0.31619343161582947, "global_step": 40875, "epoch": 459, "lr": 9.462025485660933e-05}
{"train_loss": 0.3279849588871002, "global_step": 40876, "epoch": 459, "lr": 9.461999325608836e-05}
{"train_loss": 0.25869205594062805, "global_step": 40877, "epoch": 459, "lr": 9.461973164956881e-05}
{"train_loss": 0.28927353024482727, "global_step": 40878, "epoch": 459, "lr": 9.461947003705069e-05}
{"train_loss": 0.2583915889263153, "global_step": 40879, "epoch": 459, "lr": 9.4619208418534e-05}
{"train_loss": 0.4237896203994751, "global_step": 40880, "epoch": 459, "lr": 9.461894679401882e-05}
{"train_loss": 0.33678966760635376, "global_step": 40881, "epoch": 459, "lr": 9.461868516350515e-05}
{"train_loss": 0.3377268314361572, "global_step": 40882, "epoch": 459, "lr": 9.461842352699305e-05}
{"train_loss": 0.30968227982521057, "global_step": 40883, "epoch": 459, "lr": 9.461816188448254e-05}
{"train_loss": 0.3580436110496521, "global_step": 40884, "epoch": 459, "lr": 9.461790023597368e-05}
{"train_loss": 0.295545756816864, "global_step": 40885, "epoch": 459, "lr": 9.461763858146647e-05}
{"train_loss": 0.312119722366333, "global_step": 40886, "epoch": 459, "lr": 9.461737692096096e-05}
{"train_loss": 0.3016766905784607, "global_step": 40887, "epoch": 459, "lr": 9.46171152544572e-05}
{"train_loss": 0.36089247465133667, "global_step": 40888, "epoch": 459, "lr": 9.46168535819552e-05}
{"train_loss": 0.2521957755088806, "global_step": 40889, "epoch": 459, "lr": 9.461659190345502e-05}
{"train_loss": 0.4307355582714081, "global_step": 40890, "epoch": 459, "lr": 9.461633021895669e-05}
{"train_loss": 0.3143451511859894, "global_step": 40891, "epoch": 459, "lr": 9.461606852846021e-05}
{"train_loss": 0.2776714563369751, "global_step": 40892, "epoch": 459, "lr": 9.461580683196567e-05}
{"train_loss": 0.3585434854030609, "global_step": 40893, "epoch": 459, "lr": 9.461554512947306e-05}
{"train_loss": 0.3876902163028717, "global_step": 40894, "epoch": 459, "lr": 9.461528342098244e-05}
{"train_loss": 0.3020385503768921, "global_step": 40895, "epoch": 459, "lr": 9.461502170649386e-05}
{"train_loss": 0.28624606132507324, "global_step": 40896, "epoch": 459, "lr": 9.461475998600731e-05}
{"train_loss": 0.2185918539762497, "global_step": 40897, "epoch": 459, "lr": 9.461449825952285e-05}
{"train_loss": 0.37094077467918396, "global_step": 40898, "epoch": 459, "lr": 9.46142365270405e-05}
{"train_loss": 0.25301501154899597, "global_step": 40899, "epoch": 459, "lr": 9.461397478856035e-05}
{"train_loss": 0.23688583076000214, "global_step": 40900, "epoch": 459, "lr": 9.461371304408238e-05}
{"train_loss": 0.30862635374069214, "global_step": 40901, "epoch": 459, "lr": 9.461345129360662e-05}
{"train_loss": 0.3313233256340027, "global_step": 40902, "epoch": 459, "lr": 9.461318953713314e-05}
{"train_loss": 0.35010430216789246, "global_step": 40903, "epoch": 459, "lr": 9.461292777466197e-05}
{"train_loss": 0.2885895073413849, "global_step": 40904, "epoch": 459, "lr": 9.461266600619312e-05}
{"train_loss": 0.26259171962738037, "global_step": 40905, "epoch": 459, "lr": 9.461240423172663e-05}
{"train_loss": 0.26208752393722534, "global_step": 40906, "epoch": 459, "lr": 9.461214245126257e-05}
{"train_loss": 0.3409256935119629, "global_step": 40907, "epoch": 459, "lr": 9.461188066480095e-05}
{"train_loss": 0.27089446783065796, "global_step": 40908, "epoch": 459, "lr": 9.461161887234181e-05}
{"train_loss": 0.3200019299983978, "global_step": 40909, "epoch": 459, "lr": 9.461135707388516e-05}
{"train_loss": 0.26357370615005493, "global_step": 40910, "epoch": 459, "lr": 9.461109526943106e-05}
{"train_loss": 0.3038596510887146, "global_step": 40911, "epoch": 459, "lr": 9.461083345897956e-05}
{"train_loss": 0.3190920650959015, "global_step": 40912, "epoch": 459, "lr": 9.461057164253066e-05}
{"train_loss": 0.48043331503868103, "global_step": 40913, "epoch": 459, "lr": 9.461030982008441e-05}
{"train_loss": 0.3918587565422058, "global_step": 40914, "epoch": 459, "lr": 9.461004799164086e-05}
{"train_loss": 0.3832637071609497, "global_step": 40915, "epoch": 459, "lr": 9.460978615720004e-05}
{"train_loss": 0.3261220157146454, "global_step": 40916, "epoch": 459, "lr": 9.460952431676197e-05}
{"train_loss": 0.2579067051410675, "global_step": 40917, "epoch": 459, "lr": 9.460926247032668e-05}
{"train_loss": 0.2442471832036972, "global_step": 40918, "epoch": 459, "lr": 9.460900061789423e-05}
{"train_loss": 0.3604498505592346, "global_step": 40919, "epoch": 459, "lr": 9.460873875946464e-05}
{"train_loss": 0.3802323043346405, "global_step": 40920, "epoch": 459, "lr": 9.460847689503796e-05}
{"train_loss": 0.2863878011703491, "global_step": 40921, "epoch": 459, "lr": 9.460821502461422e-05}
{"train_loss": 0.3015102744102478, "global_step": 40922, "epoch": 459, "lr": 9.460795314819341e-05}
{"train_loss": 0.19864223897457123, "global_step": 40923, "epoch": 459, "lr": 9.460769126577564e-05}
{"train_loss": 0.30342990159988403, "global_step": 40924, "epoch": 459, "lr": 9.46074293773609e-05}
{"train_loss": 0.3893863558769226, "global_step": 40925, "epoch": 459, "lr": 9.460716748294922e-05}
{"train_loss": 0.2917743921279907, "global_step": 40926, "epoch": 459, "lr": 9.460690558254067e-05}
{"train_loss": 0.3208217918872833, "global_step": 40927, "epoch": 459, "lr": 9.460664367613526e-05}
{"train_loss": 0.2536197304725647, "global_step": 40928, "epoch": 459, "lr": 9.460638176373302e-05}
{"train_loss": 0.2090827226638794, "global_step": 40929, "epoch": 459, "lr": 9.460611984533401e-05}
{"train_loss": 0.32889294624328613, "global_step": 40930, "epoch": 459, "lr": 9.460585792093826e-05}
{"train_loss": 0.27514562010765076, "global_step": 40931, "epoch": 459, "lr": 9.460559599054577e-05}
{"train_loss": 0.3015619218349457, "global_step": 40932, "epoch": 459, "lr": 9.460533405415662e-05}
{"train_loss": 0.30030298233032227, "global_step": 40933, "epoch": 459, "lr": 9.460507211177081e-05}
{"train_loss": 0.2276008576154709, "global_step": 40934, "epoch": 459, "lr": 9.46048101633884e-05}
{"train_loss": 0.30164387822151184, "global_step": 40935, "epoch": 459, "lr": 9.460454820900943e-05}
{"train_loss": 0.3825686275959015, "global_step": 40936, "epoch": 459, "lr": 9.46042862486339e-05}
{"train_loss": 0.36482614278793335, "global_step": 40937, "epoch": 459, "lr": 9.460402428226187e-05}
{"train_loss": 0.27605387568473816, "global_step": 40938, "epoch": 459, "lr": 9.460376230989338e-05}
{"train_loss": 0.31098703738678707, "global_step": 40939, "epoch": 459, "lr": 9.460350033152848e-05, "val_loss": 2.8316798210144043}
{"train_loss": 0.2638411223888397, "global_step": 40940, "epoch": 460, "lr": 9.460323834716716e-05}
{"train_loss": 0.25747978687286377, "global_step": 40941, "epoch": 460, "lr": 9.460297635680947e-05}
{"train_loss": 0.27648794651031494, "global_step": 40942, "epoch": 460, "lr": 9.460271436045546e-05}
{"train_loss": 0.3117011487483978, "global_step": 40943, "epoch": 460, "lr": 9.460245235810517e-05}
{"train_loss": 0.37386634945869446, "global_step": 40944, "epoch": 460, "lr": 9.46021903497586e-05}
{"train_loss": 0.29921406507492065, "global_step": 40945, "epoch": 460, "lr": 9.460192833541583e-05}
{"train_loss": 0.2862900495529175, "global_step": 40946, "epoch": 460, "lr": 9.460166631507687e-05}
{"train_loss": 0.3492324948310852, "global_step": 40947, "epoch": 460, "lr": 9.460140428874175e-05}
{"train_loss": 0.2565068304538727, "global_step": 40948, "epoch": 460, "lr": 9.460114225641052e-05}
{"train_loss": 0.27206942439079285, "global_step": 40949, "epoch": 460, "lr": 9.460088021808322e-05}
{"train_loss": 0.3226364254951477, "global_step": 40950, "epoch": 460, "lr": 9.460061817375987e-05}
{"train_loss": 0.28400522470474243, "global_step": 40951, "epoch": 460, "lr": 9.460035612344052e-05}
{"train_loss": 0.37981805205345154, "global_step": 40952, "epoch": 460, "lr": 9.460009406712518e-05}
{"train_loss": 0.29112347960472107, "global_step": 40953, "epoch": 460, "lr": 9.45998320048139e-05}
{"train_loss": 0.26434600353240967, "global_step": 40954, "epoch": 460, "lr": 9.459956993650671e-05}
{"train_loss": 0.27685290575027466, "global_step": 40955, "epoch": 460, "lr": 9.459930786220367e-05}
{"train_loss": 0.2592289447784424, "global_step": 40956, "epoch": 460, "lr": 9.459904578190479e-05}
{"train_loss": 0.18292897939682007, "global_step": 40957, "epoch": 460, "lr": 9.459878369561012e-05}
{"train_loss": 0.32602667808532715, "global_step": 40958, "epoch": 460, "lr": 9.459852160331968e-05}
{"train_loss": 0.43174219131469727, "global_step": 40959, "epoch": 460, "lr": 9.459825950503351e-05}
{"train_loss": 0.3192550837993622, "global_step": 40960, "epoch": 460, "lr": 9.459799740075165e-05}
{"train_loss": 0.2281329482793808, "global_step": 40961, "epoch": 460, "lr": 9.459773529047413e-05}
{"train_loss": 0.2837282717227936, "global_step": 40962, "epoch": 460, "lr": 9.4597473174201e-05}
{"train_loss": 0.3486618399620056, "global_step": 40963, "epoch": 460, "lr": 9.459721105193227e-05}
{"train_loss": 0.2446906715631485, "global_step": 40964, "epoch": 460, "lr": 9.459694892366798e-05}
{"train_loss": 0.37057095766067505, "global_step": 40965, "epoch": 460, "lr": 9.459668678940818e-05}
{"train_loss": 0.40428534150123596, "global_step": 40966, "epoch": 460, "lr": 9.459642464915292e-05}
{"train_loss": 0.3734712302684784, "global_step": 40967, "epoch": 460, "lr": 9.45961625029022e-05}
{"train_loss": 0.2753106951713562, "global_step": 40968, "epoch": 460, "lr": 9.459590035065606e-05}
{"train_loss": 0.27274295687675476, "global_step": 40969, "epoch": 460, "lr": 9.459563819241455e-05}
{"train_loss": 0.29131147265434265, "global_step": 40970, "epoch": 460, "lr": 9.459537602817771e-05}
{"train_loss": 0.39061975479125977, "global_step": 40971, "epoch": 460, "lr": 9.459511385794554e-05}
{"train_loss": 0.4543740153312683, "global_step": 40972, "epoch": 460, "lr": 9.459485168171813e-05}
{"train_loss": 0.36917757987976074, "global_step": 40973, "epoch": 460, "lr": 9.459458949949547e-05}
{"train_loss": 0.3247111439704895, "global_step": 40974, "epoch": 460, "lr": 9.459432731127762e-05}
{"train_loss": 0.2536816895008087, "global_step": 40975, "epoch": 460, "lr": 9.45940651170646e-05}
{"train_loss": 0.2341919094324112, "global_step": 40976, "epoch": 460, "lr": 9.459380291685645e-05}
{"train_loss": 0.3263057470321655, "global_step": 40977, "epoch": 460, "lr": 9.459354071065322e-05}
{"train_loss": 0.301605224609375, "global_step": 40978, "epoch": 460, "lr": 9.459327849845492e-05}
{"train_loss": 0.3216441869735718, "global_step": 40979, "epoch": 460, "lr": 9.45930162802616e-05}
{"train_loss": 0.2689448893070221, "global_step": 40980, "epoch": 460, "lr": 9.459275405607328e-05}
{"train_loss": 0.41961944103240967, "global_step": 40981, "epoch": 460, "lr": 9.459249182589002e-05}
{"train_loss": 0.3190447986125946, "global_step": 40982, "epoch": 460, "lr": 9.459222958971184e-05}
{"train_loss": 0.32971635460853577, "global_step": 40983, "epoch": 460, "lr": 9.459196734753879e-05}
{"train_loss": 0.2830592691898346, "global_step": 40984, "epoch": 460, "lr": 9.459170509937087e-05}
{"train_loss": 0.3594508767127991, "global_step": 40985, "epoch": 460, "lr": 9.459144284520815e-05}
{"train_loss": 0.25771084427833557, "global_step": 40986, "epoch": 460, "lr": 9.459118058505067e-05}
{"train_loss": 0.4551222324371338, "global_step": 40987, "epoch": 460, "lr": 9.459091831889844e-05}
{"train_loss": 0.3237536549568176, "global_step": 40988, "epoch": 460, "lr": 9.45906560467515e-05}
{"train_loss": 0.3731766939163208, "global_step": 40989, "epoch": 460, "lr": 9.459039376860988e-05}
{"train_loss": 0.2435634434223175, "global_step": 40990, "epoch": 460, "lr": 9.459013148447365e-05}
{"train_loss": 0.3135550022125244, "global_step": 40991, "epoch": 460, "lr": 9.45898691943428e-05}
{"train_loss": 0.27733081579208374, "global_step": 40992, "epoch": 460, "lr": 9.458960689821739e-05}
{"train_loss": 0.28894343972206116, "global_step": 40993, "epoch": 460, "lr": 9.458934459609745e-05}
{"train_loss": 0.22289778292179108, "global_step": 40994, "epoch": 460, "lr": 9.458908228798303e-05}
{"train_loss": 0.37547048926353455, "global_step": 40995, "epoch": 460, "lr": 9.458881997387412e-05}
{"train_loss": 0.2847835421562195, "global_step": 40996, "epoch": 460, "lr": 9.458855765377081e-05}
{"train_loss": 0.27727627754211426, "global_step": 40997, "epoch": 460, "lr": 9.45882953276731e-05}
{"train_loss": 0.2656121253967285, "global_step": 40998, "epoch": 460, "lr": 9.458803299558105e-05}
{"train_loss": 0.3858081102371216, "global_step": 40999, "epoch": 460, "lr": 9.458777065749468e-05}
{"train_loss": 0.3026895225048065, "global_step": 41000, "epoch": 460, "lr": 9.458750831341403e-05}
{"train_loss": 0.3475119471549988, "global_step": 41001, "epoch": 460, "lr": 9.458724596333913e-05}
{"train_loss": 0.2502153813838959, "global_step": 41002, "epoch": 460, "lr": 9.458698360727e-05}
{"train_loss": 0.3079400360584259, "global_step": 41003, "epoch": 460, "lr": 9.458672124520671e-05}
{"train_loss": 0.35296952724456787, "global_step": 41004, "epoch": 460, "lr": 9.458645887714928e-05}
{"train_loss": 0.2724764943122864, "global_step": 41005, "epoch": 460, "lr": 9.458619650309775e-05}
{"train_loss": 0.2769101858139038, "global_step": 41006, "epoch": 460, "lr": 9.458593412305215e-05}
{"train_loss": 0.309645414352417, "global_step": 41007, "epoch": 460, "lr": 9.45856717370125e-05}
{"train_loss": 0.23772495985031128, "global_step": 41008, "epoch": 460, "lr": 9.458540934497886e-05}
{"train_loss": 0.22856968641281128, "global_step": 41009, "epoch": 460, "lr": 9.458514694695125e-05}
{"train_loss": 0.2774265706539154, "global_step": 41010, "epoch": 460, "lr": 9.458488454292971e-05}
{"train_loss": 0.2858128845691681, "global_step": 41011, "epoch": 460, "lr": 9.458462213291428e-05}
{"train_loss": 0.2667638659477234, "global_step": 41012, "epoch": 460, "lr": 9.458435971690499e-05}
{"train_loss": 0.2859365940093994, "global_step": 41013, "epoch": 460, "lr": 9.458409729490187e-05}
{"train_loss": 0.28501197695732117, "global_step": 41014, "epoch": 460, "lr": 9.458383486690496e-05}
{"train_loss": 0.35887011885643005, "global_step": 41015, "epoch": 460, "lr": 9.458357243291431e-05}
{"train_loss": 0.2687550485134125, "global_step": 41016, "epoch": 460, "lr": 9.458330999292992e-05}
{"train_loss": 0.2525217831134796, "global_step": 41017, "epoch": 460, "lr": 9.458304754695188e-05}
{"train_loss": 0.2421104460954666, "global_step": 41018, "epoch": 460, "lr": 9.458278509498018e-05}
{"train_loss": 0.3933370113372803, "global_step": 41019, "epoch": 460, "lr": 9.458252263701485e-05}
{"train_loss": 0.22739937901496887, "global_step": 41020, "epoch": 460, "lr": 9.458226017305596e-05}
{"train_loss": 0.3028548061847687, "global_step": 41021, "epoch": 460, "lr": 9.458199770310352e-05}
{"train_loss": 0.30219021439552307, "global_step": 41022, "epoch": 460, "lr": 9.458173522715759e-05}
{"train_loss": 0.25391513109207153, "global_step": 41023, "epoch": 460, "lr": 9.458147274521817e-05}
{"train_loss": 0.3217690587043762, "global_step": 41024, "epoch": 460, "lr": 9.458121025728532e-05}
{"train_loss": 0.24706526100635529, "global_step": 41025, "epoch": 460, "lr": 9.458094776335908e-05}
{"train_loss": 0.3066628873348236, "global_step": 41026, "epoch": 460, "lr": 9.458068526343946e-05}
{"train_loss": 0.3107820749282837, "global_step": 41027, "epoch": 460, "lr": 9.458042275752652e-05}
{"train_loss": 0.3040083988328998, "global_step": 41028, "epoch": 460, "lr": 9.458016024562027e-05, "val_loss": 2.854783773422241, "train_action_mse_error": 19.098796844482422}
{"train_loss": 0.24930337071418762, "global_step": 41029, "epoch": 461, "lr": 9.457989772772078e-05}
{"train_loss": 0.32798057794570923, "global_step": 41030, "epoch": 461, "lr": 9.457963520382807e-05}
{"train_loss": 0.28433823585510254, "global_step": 41031, "epoch": 461, "lr": 9.457937267394215e-05}
{"train_loss": 0.46448877453804016, "global_step": 41032, "epoch": 461, "lr": 9.457911013806309e-05}
{"train_loss": 0.2925887107849121, "global_step": 41033, "epoch": 461, "lr": 9.457884759619091e-05}
{"train_loss": 0.2963099777698517, "global_step": 41034, "epoch": 461, "lr": 9.457858504832564e-05}
{"train_loss": 0.1867571920156479, "global_step": 41035, "epoch": 461, "lr": 9.457832249446733e-05}
{"train_loss": 0.4031949043273926, "global_step": 41036, "epoch": 461, "lr": 9.457805993461601e-05}
{"train_loss": 0.19740909337997437, "global_step": 41037, "epoch": 461, "lr": 9.45777973687717e-05}
{"train_loss": 0.26192185282707214, "global_step": 41038, "epoch": 461, "lr": 9.457753479693447e-05}
{"train_loss": 0.3246445953845978, "global_step": 41039, "epoch": 461, "lr": 9.457727221910431e-05}
{"train_loss": 0.3608972132205963, "global_step": 41040, "epoch": 461, "lr": 9.45770096352813e-05}
{"train_loss": 0.3938859701156616, "global_step": 41041, "epoch": 461, "lr": 9.457674704546544e-05}
{"train_loss": 0.20654471218585968, "global_step": 41042, "epoch": 461, "lr": 9.457648444965679e-05}
{"train_loss": 0.2836408317089081, "global_step": 41043, "epoch": 461, "lr": 9.457622184785538e-05}
{"train_loss": 0.34156733751296997, "global_step": 41044, "epoch": 461, "lr": 9.457595924006122e-05}
{"train_loss": 0.22684727609157562, "global_step": 41045, "epoch": 461, "lr": 9.457569662627437e-05}
{"train_loss": 0.29495030641555786, "global_step": 41046, "epoch": 461, "lr": 9.457543400649488e-05}
{"train_loss": 0.34729549288749695, "global_step": 41047, "epoch": 461, "lr": 9.457517138072275e-05}
{"train_loss": 0.24829120934009552, "global_step": 41048, "epoch": 461, "lr": 9.457490874895803e-05}
{"train_loss": 0.23671896755695343, "global_step": 41049, "epoch": 461, "lr": 9.457464611120078e-05}
{"train_loss": 0.32822751998901367, "global_step": 41050, "epoch": 461, "lr": 9.457438346745098e-05}
{"train_loss": 0.26967760920524597, "global_step": 41051, "epoch": 461, "lr": 9.457412081770871e-05}
{"train_loss": 0.188980370759964, "global_step": 41052, "epoch": 461, "lr": 9.4573858161974e-05}
{"train_loss": 0.3140455484390259, "global_step": 41053, "epoch": 461, "lr": 9.457359550024687e-05}
{"train_loss": 0.22615043818950653, "global_step": 41054, "epoch": 461, "lr": 9.457333283252736e-05}
{"train_loss": 0.2847243547439575, "global_step": 41055, "epoch": 461, "lr": 9.457307015881552e-05}
{"train_loss": 0.27672621607780457, "global_step": 41056, "epoch": 461, "lr": 9.457280747911137e-05}
{"train_loss": 0.30077823996543884, "global_step": 41057, "epoch": 461, "lr": 9.457254479341495e-05}
{"train_loss": 0.23121608793735504, "global_step": 41058, "epoch": 461, "lr": 9.457228210172629e-05}
{"train_loss": 0.26999059319496155, "global_step": 41059, "epoch": 461, "lr": 9.457201940404543e-05}
{"train_loss": 0.323377400636673, "global_step": 41060, "epoch": 461, "lr": 9.45717567003724e-05}
{"train_loss": 0.33289623260498047, "global_step": 41061, "epoch": 461, "lr": 9.457149399070725e-05}
{"train_loss": 0.2538605332374573, "global_step": 41062, "epoch": 461, "lr": 9.457123127505001e-05}
{"train_loss": 0.26158708333969116, "global_step": 41063, "epoch": 461, "lr": 9.45709685534007e-05}
{"train_loss": 0.21531915664672852, "global_step": 41064, "epoch": 461, "lr": 9.457070582575938e-05}
{"train_loss": 0.34252771735191345, "global_step": 41065, "epoch": 461, "lr": 9.457044309212606e-05}
{"train_loss": 0.2606809437274933, "global_step": 41066, "epoch": 461, "lr": 9.457018035250078e-05}
{"train_loss": 0.3054058253765106, "global_step": 41067, "epoch": 461, "lr": 9.456991760688359e-05}
{"train_loss": 0.3131054639816284, "global_step": 41068, "epoch": 461, "lr": 9.456965485527452e-05}
{"train_loss": 0.42023104429244995, "global_step": 41069, "epoch": 461, "lr": 9.456939209767362e-05}
{"train_loss": 0.3829282522201538, "global_step": 41070, "epoch": 461, "lr": 9.456912933408089e-05}
{"train_loss": 0.3671373724937439, "global_step": 41071, "epoch": 461, "lr": 9.456886656449637e-05}
{"train_loss": 0.33002781867980957, "global_step": 41072, "epoch": 461, "lr": 9.456860378892014e-05}
{"train_loss": 0.29051128029823303, "global_step": 41073, "epoch": 461, "lr": 9.456834100735219e-05}
{"train_loss": 0.3905910849571228, "global_step": 41074, "epoch": 461, "lr": 9.456807821979256e-05}
{"train_loss": 0.4751569628715515, "global_step": 41075, "epoch": 461, "lr": 9.456781542624132e-05}
{"train_loss": 0.23094384372234344, "global_step": 41076, "epoch": 461, "lr": 9.456755262669845e-05}
{"train_loss": 0.2471458464860916, "global_step": 41077, "epoch": 461, "lr": 9.456728982116403e-05}
{"train_loss": 0.3867340683937073, "global_step": 41078, "epoch": 461, "lr": 9.456702700963809e-05}
{"train_loss": 0.3728187680244446, "global_step": 41079, "epoch": 461, "lr": 9.456676419212065e-05}
{"train_loss": 0.3332262933254242, "global_step": 41080, "epoch": 461, "lr": 9.456650136861174e-05}
{"train_loss": 0.36299148201942444, "global_step": 41081, "epoch": 461, "lr": 9.456623853911141e-05}
{"train_loss": 0.32415255904197693, "global_step": 41082, "epoch": 461, "lr": 9.456597570361972e-05}
{"train_loss": 0.28296926617622375, "global_step": 41083, "epoch": 461, "lr": 9.456571286213664e-05}
{"train_loss": 0.34406423568725586, "global_step": 41084, "epoch": 461, "lr": 9.456545001466227e-05}
{"train_loss": 0.2397228330373764, "global_step": 41085, "epoch": 461, "lr": 9.45651871611966e-05}
{"train_loss": 0.3041720390319824, "global_step": 41086, "epoch": 461, "lr": 9.45649243017397e-05}
{"train_loss": 0.27607908844947815, "global_step": 41087, "epoch": 461, "lr": 9.456466143629157e-05}
{"train_loss": 0.3062235414981842, "global_step": 41088, "epoch": 461, "lr": 9.456439856485229e-05}
{"train_loss": 0.2975766062736511, "global_step": 41089, "epoch": 461, "lr": 9.456413568742185e-05}
{"train_loss": 0.3584848642349243, "global_step": 41090, "epoch": 461, "lr": 9.456387280400031e-05}
{"train_loss": 0.3150651156902313, "global_step": 41091, "epoch": 461, "lr": 9.45636099145877e-05}
{"train_loss": 0.3838766813278198, "global_step": 41092, "epoch": 461, "lr": 9.456334701918405e-05}
{"train_loss": 0.3299233913421631, "global_step": 41093, "epoch": 461, "lr": 9.456308411778942e-05}
{"train_loss": 0.39300891757011414, "global_step": 41094, "epoch": 461, "lr": 9.45628212104038e-05}
{"train_loss": 0.22484901547431946, "global_step": 41095, "epoch": 461, "lr": 9.456255829702727e-05}
{"train_loss": 0.46343541145324707, "global_step": 41096, "epoch": 461, "lr": 9.456229537765987e-05}
{"train_loss": 0.3292981684207916, "global_step": 41097, "epoch": 461, "lr": 9.456203245230157e-05}
{"train_loss": 0.30299150943756104, "global_step": 41098, "epoch": 461, "lr": 9.456176952095246e-05}
{"train_loss": 0.45704883337020874, "global_step": 41099, "epoch": 461, "lr": 9.456150658361257e-05}
{"train_loss": 0.40368348360061646, "global_step": 41100, "epoch": 461, "lr": 9.456124364028192e-05}
{"train_loss": 0.2713364362716675, "global_step": 41101, "epoch": 461, "lr": 9.456098069096057e-05}
{"train_loss": 0.24332380294799805, "global_step": 41102, "epoch": 461, "lr": 9.456071773564852e-05}
{"train_loss": 0.2879998981952667, "global_step": 41103, "epoch": 461, "lr": 9.456045477434583e-05}
{"train_loss": 0.30322590470314026, "global_step": 41104, "epoch": 461, "lr": 9.456019180705253e-05}
{"train_loss": 0.26869824528694153, "global_step": 41105, "epoch": 461, "lr": 9.455992883376867e-05}
{"train_loss": 0.5096192359924316, "global_step": 41106, "epoch": 461, "lr": 9.455966585449426e-05}
{"train_loss": 0.38426703214645386, "global_step": 41107, "epoch": 461, "lr": 9.455940286922935e-05}
{"train_loss": 0.3561757504940033, "global_step": 41108, "epoch": 461, "lr": 9.455913987797395e-05}
{"train_loss": 0.3219473958015442, "global_step": 41109, "epoch": 461, "lr": 9.455887688072814e-05}
{"train_loss": 0.26810914278030396, "global_step": 41110, "epoch": 461, "lr": 9.455861387749193e-05}
{"train_loss": 0.24149979650974274, "global_step": 41111, "epoch": 461, "lr": 9.455835086826535e-05}
{"train_loss": 0.30645424127578735, "global_step": 41112, "epoch": 461, "lr": 9.455808785304844e-05}
{"train_loss": 0.3119658827781677, "global_step": 41113, "epoch": 461, "lr": 9.455782483184125e-05}
{"train_loss": 0.3696492910385132, "global_step": 41114, "epoch": 461, "lr": 9.45575618046438e-05}
{"train_loss": 0.32386717200279236, "global_step": 41115, "epoch": 461, "lr": 9.455729877145614e-05}
{"train_loss": 0.39862096309661865, "global_step": 41116, "epoch": 461, "lr": 9.455703573227827e-05}
{"train_loss": 0.31388602313700686, "global_step": 41117, "epoch": 461, "lr": 9.455677268711027e-05, "val_loss": 2.61445951461792}
{"train_loss": 0.31547582149505615, "global_step": 41118, "epoch": 462, "lr": 9.455650963595213e-05}
{"train_loss": 0.37441933155059814, "global_step": 41119, "epoch": 462, "lr": 9.455624657880393e-05}
{"train_loss": 0.3528999090194702, "global_step": 41120, "epoch": 462, "lr": 9.455598351566568e-05}
{"train_loss": 0.24803343415260315, "global_step": 41121, "epoch": 462, "lr": 9.455572044653742e-05}
{"train_loss": 0.25410276651382446, "global_step": 41122, "epoch": 462, "lr": 9.45554573714192e-05}
{"train_loss": 0.33069148659706116, "global_step": 41123, "epoch": 462, "lr": 9.455519429031103e-05}
{"train_loss": 0.3419770896434784, "global_step": 41124, "epoch": 462, "lr": 9.455493120321294e-05}
{"train_loss": 0.24418285489082336, "global_step": 41125, "epoch": 462, "lr": 9.455466811012501e-05}
{"train_loss": 0.3232713043689728, "global_step": 41126, "epoch": 462, "lr": 9.455440501104725e-05}
{"train_loss": 0.2577955424785614, "global_step": 41127, "epoch": 462, "lr": 9.455414190597967e-05}
{"train_loss": 0.3647403419017792, "global_step": 41128, "epoch": 462, "lr": 9.455387879492234e-05}
{"train_loss": 0.25160086154937744, "global_step": 41129, "epoch": 462, "lr": 9.455361567787529e-05}
{"train_loss": 0.2343699336051941, "global_step": 41130, "epoch": 462, "lr": 9.455335255483855e-05}
{"train_loss": 0.20680825412273407, "global_step": 41131, "epoch": 462, "lr": 9.455308942581215e-05}
{"train_loss": 0.3512670695781708, "global_step": 41132, "epoch": 462, "lr": 9.455282629079612e-05}
{"train_loss": 0.46370115876197815, "global_step": 41133, "epoch": 462, "lr": 9.455256314979052e-05}
{"train_loss": 0.3930836617946625, "global_step": 41134, "epoch": 462, "lr": 9.455230000279536e-05}
{"train_loss": 0.2855101227760315, "global_step": 41135, "epoch": 462, "lr": 9.45520368498107e-05}
{"train_loss": 0.256559818983078, "global_step": 41136, "epoch": 462, "lr": 9.455177369083655e-05}
{"train_loss": 0.2642922103404999, "global_step": 41137, "epoch": 462, "lr": 9.455151052587297e-05}
{"train_loss": 0.31349655985832214, "global_step": 41138, "epoch": 462, "lr": 9.455124735491996e-05}
{"train_loss": 0.22850359976291656, "global_step": 41139, "epoch": 462, "lr": 9.455098417797758e-05}
{"train_loss": 0.3876998722553253, "global_step": 41140, "epoch": 462, "lr": 9.455072099504589e-05}
{"train_loss": 0.4400806725025177, "global_step": 41141, "epoch": 462, "lr": 9.455045780612489e-05}
{"train_loss": 0.3358716666698456, "global_step": 41142, "epoch": 462, "lr": 9.45501946112146e-05}
{"train_loss": 0.32782623171806335, "global_step": 41143, "epoch": 462, "lr": 9.454993141031509e-05}
{"train_loss": 0.25803542137145996, "global_step": 41144, "epoch": 462, "lr": 9.454966820342641e-05}
{"train_loss": 0.37244486808776855, "global_step": 41145, "epoch": 462, "lr": 9.454940499054854e-05}
{"train_loss": 0.31590068340301514, "global_step": 41146, "epoch": 462, "lr": 9.454914177168154e-05}
{"train_loss": 0.3064556419849396, "global_step": 41147, "epoch": 462, "lr": 9.454887854682548e-05}
{"train_loss": 0.23849532008171082, "global_step": 41148, "epoch": 462, "lr": 9.454861531598035e-05}
{"train_loss": 0.30142271518707275, "global_step": 41149, "epoch": 462, "lr": 9.454835207914621e-05}
{"train_loss": 0.4067496061325073, "global_step": 41150, "epoch": 462, "lr": 9.45480888363231e-05}
{"train_loss": 0.23007313907146454, "global_step": 41151, "epoch": 462, "lr": 9.454782558751101e-05}
{"train_loss": 0.3166952133178711, "global_step": 41152, "epoch": 462, "lr": 9.454756233271002e-05}
{"train_loss": 0.2536245286464691, "global_step": 41153, "epoch": 462, "lr": 9.454729907192014e-05}
{"train_loss": 0.25797778367996216, "global_step": 41154, "epoch": 462, "lr": 9.454703580514145e-05}
{"train_loss": 0.279778391122818, "global_step": 41155, "epoch": 462, "lr": 9.454677253237394e-05}
{"train_loss": 0.23905660212039948, "global_step": 41156, "epoch": 462, "lr": 9.454650925361765e-05}
{"train_loss": 0.25493329763412476, "global_step": 41157, "epoch": 462, "lr": 9.454624596887264e-05}
{"train_loss": 0.36507847905158997, "global_step": 41158, "epoch": 462, "lr": 9.454598267813892e-05}
{"train_loss": 0.2621409595012665, "global_step": 41159, "epoch": 462, "lr": 9.454571938141655e-05}
{"train_loss": 0.3577464520931244, "global_step": 41160, "epoch": 462, "lr": 9.454545607870553e-05}
{"train_loss": 0.435904860496521, "global_step": 41161, "epoch": 462, "lr": 9.454519277000591e-05}
{"train_loss": 0.30175840854644775, "global_step": 41162, "epoch": 462, "lr": 9.454492945531776e-05}
{"train_loss": 0.3403920829296112, "global_step": 41163, "epoch": 462, "lr": 9.454466613464107e-05}
{"train_loss": 0.20749688148498535, "global_step": 41164, "epoch": 462, "lr": 9.45444028079759e-05}
{"train_loss": 0.3715655207633972, "global_step": 41165, "epoch": 462, "lr": 9.454413947532226e-05}
{"train_loss": 0.3321864902973175, "global_step": 41166, "epoch": 462, "lr": 9.454387613668023e-05}
{"train_loss": 0.386926531791687, "global_step": 41167, "epoch": 462, "lr": 9.45436127920498e-05}
{"train_loss": 0.35041847825050354, "global_step": 41168, "epoch": 462, "lr": 9.454334944143104e-05}
{"train_loss": 0.5739864706993103, "global_step": 41169, "epoch": 462, "lr": 9.454308608482397e-05}
{"train_loss": 0.3481464385986328, "global_step": 41170, "epoch": 462, "lr": 9.45428227222286e-05}
{"train_loss": 0.32256460189819336, "global_step": 41171, "epoch": 462, "lr": 9.4542559353645e-05}
{"train_loss": 0.3286382853984833, "global_step": 41172, "epoch": 462, "lr": 9.45422959790732e-05}
{"train_loss": 0.3647240698337555, "global_step": 41173, "epoch": 462, "lr": 9.454203259851325e-05}
{"train_loss": 0.4466444253921509, "global_step": 41174, "epoch": 462, "lr": 9.454176921196514e-05}
{"train_loss": 0.4278087615966797, "global_step": 41175, "epoch": 462, "lr": 9.454150581942893e-05}
{"train_loss": 0.41395512223243713, "global_step": 41176, "epoch": 462, "lr": 9.454124242090467e-05}
{"train_loss": 0.3373384177684784, "global_step": 41177, "epoch": 462, "lr": 9.454097901639238e-05}
{"train_loss": 0.2156469076871872, "global_step": 41178, "epoch": 462, "lr": 9.45407156058921e-05}
{"train_loss": 0.32401061058044434, "global_step": 41179, "epoch": 462, "lr": 9.454045218940387e-05}
{"train_loss": 0.3944975435733795, "global_step": 41180, "epoch": 462, "lr": 9.454018876692772e-05}
{"train_loss": 0.37654751539230347, "global_step": 41181, "epoch": 462, "lr": 9.453992533846367e-05}
{"train_loss": 0.3336847722530365, "global_step": 41182, "epoch": 462, "lr": 9.453966190401178e-05}
{"train_loss": 0.46026507019996643, "global_step": 41183, "epoch": 462, "lr": 9.453939846357207e-05}
{"train_loss": 0.35146164894104004, "global_step": 41184, "epoch": 462, "lr": 9.453913501714458e-05}
{"train_loss": 0.2690398097038269, "global_step": 41185, "epoch": 462, "lr": 9.453887156472936e-05}
{"train_loss": 0.36395424604415894, "global_step": 41186, "epoch": 462, "lr": 9.453860810632642e-05}
{"train_loss": 0.3734225928783417, "global_step": 41187, "epoch": 462, "lr": 9.453834464193581e-05}
{"train_loss": 0.28193363547325134, "global_step": 41188, "epoch": 462, "lr": 9.453808117155756e-05}
{"train_loss": 0.3406188189983368, "global_step": 41189, "epoch": 462, "lr": 9.45378176951917e-05}
{"train_loss": 0.3985093832015991, "global_step": 41190, "epoch": 462, "lr": 9.453755421283828e-05}
{"train_loss": 0.3294598162174225, "global_step": 41191, "epoch": 462, "lr": 9.453729072449733e-05}
{"train_loss": 0.34261104464530945, "global_step": 41192, "epoch": 462, "lr": 9.453702723016888e-05}
{"train_loss": 0.24526503682136536, "global_step": 41193, "epoch": 462, "lr": 9.453676372985298e-05}
{"train_loss": 0.2842782437801361, "global_step": 41194, "epoch": 462, "lr": 9.453650022354965e-05}
{"train_loss": 0.3622024953365326, "global_step": 41195, "epoch": 462, "lr": 9.453623671125893e-05}
{"train_loss": 0.34540191292762756, "global_step": 41196, "epoch": 462, "lr": 9.453597319298085e-05}
{"train_loss": 0.36063873767852783, "global_step": 41197, "epoch": 462, "lr": 9.453570966871545e-05}
{"train_loss": 0.2681165933609009, "global_step": 41198, "epoch": 462, "lr": 9.453544613846278e-05}
{"train_loss": 0.2996504306793213, "global_step": 41199, "epoch": 462, "lr": 9.453518260222285e-05}
{"train_loss": 0.34626153111457825, "global_step": 41200, "epoch": 462, "lr": 9.453491905999572e-05}
{"train_loss": 0.35027599334716797, "global_step": 41201, "epoch": 462, "lr": 9.45346555117814e-05}
{"train_loss": 0.405253529548645, "global_step": 41202, "epoch": 462, "lr": 9.453439195757993e-05}
{"train_loss": 0.3432762324810028, "global_step": 41203, "epoch": 462, "lr": 9.453412839739138e-05}
{"train_loss": 0.2835087180137634, "global_step": 41204, "epoch": 462, "lr": 9.453386483121574e-05}
{"train_loss": 0.30515167117118835, "global_step": 41205, "epoch": 462, "lr": 9.453360125905308e-05}
{"train_loss": 0.3266739244206568, "global_step": 41206, "epoch": 462, "lr": 9.45333376809034e-05, "val_loss": 2.6100709438323975}
{"train_loss": 0.18647019565105438, "global_step": 41207, "epoch": 463, "lr": 9.453307409676677e-05}
{"train_loss": 0.317323237657547, "global_step": 41208, "epoch": 463, "lr": 9.453281050664321e-05}
{"train_loss": 0.3431805968284607, "global_step": 41209, "epoch": 463, "lr": 9.453254691053275e-05}
{"train_loss": 0.3038279414176941, "global_step": 41210, "epoch": 463, "lr": 9.453228330843545e-05}
{"train_loss": 0.36264780163764954, "global_step": 41211, "epoch": 463, "lr": 9.45320197003513e-05}
{"train_loss": 0.33055341243743896, "global_step": 41212, "epoch": 463, "lr": 9.453175608628037e-05}
{"train_loss": 0.4255478084087372, "global_step": 41213, "epoch": 463, "lr": 9.45314924662227e-05}
{"train_loss": 0.2820238769054413, "global_step": 41214, "epoch": 463, "lr": 9.453122884017831e-05}
{"train_loss": 0.3352610766887665, "global_step": 41215, "epoch": 463, "lr": 9.453096520814723e-05}
{"train_loss": 0.2673020660877228, "global_step": 41216, "epoch": 463, "lr": 9.453070157012952e-05}
{"train_loss": 0.2830282747745514, "global_step": 41217, "epoch": 463, "lr": 9.453043792612519e-05}
{"train_loss": 0.2241165190935135, "global_step": 41218, "epoch": 463, "lr": 9.453017427613429e-05}
{"train_loss": 0.33032798767089844, "global_step": 41219, "epoch": 463, "lr": 9.452991062015685e-05}
{"train_loss": 0.20246490836143494, "global_step": 41220, "epoch": 463, "lr": 9.45296469581929e-05}
{"train_loss": 0.2613322138786316, "global_step": 41221, "epoch": 463, "lr": 9.452938329024248e-05}
{"train_loss": 0.3063459098339081, "global_step": 41222, "epoch": 463, "lr": 9.452911961630563e-05}
{"train_loss": 0.2909442186355591, "global_step": 41223, "epoch": 463, "lr": 9.452885593638239e-05}
{"train_loss": 0.3194305896759033, "global_step": 41224, "epoch": 463, "lr": 9.452859225047279e-05}
{"train_loss": 0.2546478509902954, "global_step": 41225, "epoch": 463, "lr": 9.452832855857686e-05}
{"train_loss": 0.24776041507720947, "global_step": 41226, "epoch": 463, "lr": 9.452806486069463e-05}
{"train_loss": 0.29316118359565735, "global_step": 41227, "epoch": 463, "lr": 9.452780115682614e-05}
{"train_loss": 0.3206220865249634, "global_step": 41228, "epoch": 463, "lr": 9.452753744697145e-05}
{"train_loss": 0.32428327202796936, "global_step": 41229, "epoch": 463, "lr": 9.452727373113056e-05}
{"train_loss": 0.2952459752559662, "global_step": 41230, "epoch": 463, "lr": 9.452701000930352e-05}
{"train_loss": 0.3324809968471527, "global_step": 41231, "epoch": 463, "lr": 9.452674628149039e-05}
{"train_loss": 0.3047057092189789, "global_step": 41232, "epoch": 463, "lr": 9.452648254769116e-05}
{"train_loss": 0.29857367277145386, "global_step": 41233, "epoch": 463, "lr": 9.452621880790589e-05}
{"train_loss": 0.3242000937461853, "global_step": 41234, "epoch": 463, "lr": 9.452595506213461e-05}
{"train_loss": 0.29957616329193115, "global_step": 41235, "epoch": 463, "lr": 9.452569131037736e-05}
{"train_loss": 0.34164610505104065, "global_step": 41236, "epoch": 463, "lr": 9.452542755263417e-05}
{"train_loss": 0.23843319714069366, "global_step": 41237, "epoch": 463, "lr": 9.452516378890508e-05}
{"train_loss": 0.2596457600593567, "global_step": 41238, "epoch": 463, "lr": 9.452490001919013e-05}
{"train_loss": 0.3223814368247986, "global_step": 41239, "epoch": 463, "lr": 9.452463624348935e-05}
{"train_loss": 0.3230660855770111, "global_step": 41240, "epoch": 463, "lr": 9.452437246180278e-05}
{"train_loss": 0.2788110673427582, "global_step": 41241, "epoch": 463, "lr": 9.452410867413044e-05}
{"train_loss": 0.3247945308685303, "global_step": 41242, "epoch": 463, "lr": 9.452384488047239e-05}
{"train_loss": 0.3321925401687622, "global_step": 41243, "epoch": 463, "lr": 9.452358108082864e-05}
{"train_loss": 0.33597052097320557, "global_step": 41244, "epoch": 463, "lr": 9.452331727519923e-05}
{"train_loss": 0.34301355481147766, "global_step": 41245, "epoch": 463, "lr": 9.452305346358422e-05}
{"train_loss": 0.19125200808048248, "global_step": 41246, "epoch": 463, "lr": 9.452278964598362e-05}
{"train_loss": 0.2780592441558838, "global_step": 41247, "epoch": 463, "lr": 9.452252582239747e-05}
{"train_loss": 0.33571094274520874, "global_step": 41248, "epoch": 463, "lr": 9.452226199282582e-05}
{"train_loss": 0.3081909418106079, "global_step": 41249, "epoch": 463, "lr": 9.452199815726867e-05}
{"train_loss": 0.23416365683078766, "global_step": 41250, "epoch": 463, "lr": 9.45217343157261e-05}
{"train_loss": 0.2504396140575409, "global_step": 41251, "epoch": 463, "lr": 9.452147046819813e-05}
{"train_loss": 0.2768903374671936, "global_step": 41252, "epoch": 463, "lr": 9.452120661468478e-05}
{"train_loss": 0.4179690182209015, "global_step": 41253, "epoch": 463, "lr": 9.45209427551861e-05}
{"train_loss": 0.28290805220603943, "global_step": 41254, "epoch": 463, "lr": 9.452067888970213e-05}
{"train_loss": 0.32961505651474, "global_step": 41255, "epoch": 463, "lr": 9.452041501823288e-05}
{"train_loss": 0.3440488278865814, "global_step": 41256, "epoch": 463, "lr": 9.452015114077841e-05}
{"train_loss": 0.3488416373729706, "global_step": 41257, "epoch": 463, "lr": 9.451988725733876e-05}
{"train_loss": 0.287492036819458, "global_step": 41258, "epoch": 463, "lr": 9.451962336791395e-05}
{"train_loss": 0.3554295599460602, "global_step": 41259, "epoch": 463, "lr": 9.4519359472504e-05}
{"train_loss": 0.2621467113494873, "global_step": 41260, "epoch": 463, "lr": 9.451909557110899e-05}
{"train_loss": 0.28647086024284363, "global_step": 41261, "epoch": 463, "lr": 9.451883166372891e-05}
{"train_loss": 0.24243494868278503, "global_step": 41262, "epoch": 463, "lr": 9.451856775036383e-05}
{"train_loss": 0.23332414031028748, "global_step": 41263, "epoch": 463, "lr": 9.451830383101377e-05}
{"train_loss": 0.3289385735988617, "global_step": 41264, "epoch": 463, "lr": 9.451803990567876e-05}
{"train_loss": 0.3364323675632477, "global_step": 41265, "epoch": 463, "lr": 9.451777597435884e-05}
{"train_loss": 0.3587678074836731, "global_step": 41266, "epoch": 463, "lr": 9.451751203705406e-05}
{"train_loss": 0.3111724257469177, "global_step": 41267, "epoch": 463, "lr": 9.451724809376443e-05}
{"train_loss": 0.34803634881973267, "global_step": 41268, "epoch": 463, "lr": 9.451698414449e-05}
{"train_loss": 0.317586213350296, "global_step": 41269, "epoch": 463, "lr": 9.451672018923082e-05}
{"train_loss": 0.3024170994758606, "global_step": 41270, "epoch": 463, "lr": 9.451645622798688e-05}
{"train_loss": 0.21823111176490784, "global_step": 41271, "epoch": 463, "lr": 9.451619226075827e-05}
{"train_loss": 0.4036673605442047, "global_step": 41272, "epoch": 463, "lr": 9.451592828754499e-05}
{"train_loss": 0.34598153829574585, "global_step": 41273, "epoch": 463, "lr": 9.451566430834709e-05}
{"train_loss": 0.32002145051956177, "global_step": 41274, "epoch": 463, "lr": 9.451540032316459e-05}
{"train_loss": 0.34120863676071167, "global_step": 41275, "epoch": 463, "lr": 9.451513633199756e-05}
{"train_loss": 0.358266144990921, "global_step": 41276, "epoch": 463, "lr": 9.4514872334846e-05}
{"train_loss": 0.4928204417228699, "global_step": 41277, "epoch": 463, "lr": 9.451460833170995e-05}
{"train_loss": 0.18729016184806824, "global_step": 41278, "epoch": 463, "lr": 9.451434432258947e-05}
{"train_loss": 0.2730899453163147, "global_step": 41279, "epoch": 463, "lr": 9.451408030748456e-05}
{"train_loss": 0.29914620518684387, "global_step": 41280, "epoch": 463, "lr": 9.451381628639528e-05}
{"train_loss": 0.3784301280975342, "global_step": 41281, "epoch": 463, "lr": 9.451355225932168e-05}
{"train_loss": 0.3160548508167267, "global_step": 41282, "epoch": 463, "lr": 9.451328822626375e-05}
{"train_loss": 0.32001805305480957, "global_step": 41283, "epoch": 463, "lr": 9.451302418722157e-05}
{"train_loss": 0.4287681579589844, "global_step": 41284, "epoch": 463, "lr": 9.451276014219514e-05}
{"train_loss": 0.39041775465011597, "global_step": 41285, "epoch": 463, "lr": 9.451249609118452e-05}
{"train_loss": 0.3412640690803528, "global_step": 41286, "epoch": 463, "lr": 9.451223203418973e-05}
{"train_loss": 0.36397606134414673, "global_step": 41287, "epoch": 463, "lr": 9.451196797121083e-05}
{"train_loss": 0.26301127672195435, "global_step": 41288, "epoch": 463, "lr": 9.451170390224783e-05}
{"train_loss": 0.33283376693725586, "global_step": 41289, "epoch": 463, "lr": 9.451143982730078e-05}
{"train_loss": 0.2868121862411499, "global_step": 41290, "epoch": 463, "lr": 9.451117574636969e-05}
{"train_loss": 0.28715428709983826, "global_step": 41291, "epoch": 463, "lr": 9.451091165945462e-05}
{"train_loss": 0.33190619945526123, "global_step": 41292, "epoch": 463, "lr": 9.451064756655561e-05}
{"train_loss": 0.43294262886047363, "global_step": 41293, "epoch": 463, "lr": 9.451038346767268e-05}
{"train_loss": 0.29205870628356934, "global_step": 41294, "epoch": 463, "lr": 9.451011936280587e-05}
{"train_loss": 0.3112881195344282, "global_step": 41295, "epoch": 463, "lr": 9.450985525195523e-05, "val_loss": 2.5327646732330322}
{"train_loss": 0.30080121755599976, "global_step": 41296, "epoch": 464, "lr": 9.450959113512077e-05}
{"train_loss": 0.30737313628196716, "global_step": 41297, "epoch": 464, "lr": 9.450932701230256e-05}
{"train_loss": 0.28497588634490967, "global_step": 41298, "epoch": 464, "lr": 9.450906288350058e-05}
{"train_loss": 0.32259079813957214, "global_step": 41299, "epoch": 464, "lr": 9.450879874871491e-05}
{"train_loss": 0.3170371353626251, "global_step": 41300, "epoch": 464, "lr": 9.450853460794557e-05}
{"train_loss": 0.2788415551185608, "global_step": 41301, "epoch": 464, "lr": 9.450827046119261e-05}
{"train_loss": 0.38464707136154175, "global_step": 41302, "epoch": 464, "lr": 9.450800630845606e-05}
{"train_loss": 0.3002345561981201, "global_step": 41303, "epoch": 464, "lr": 9.450774214973595e-05}
{"train_loss": 0.30052080750465393, "global_step": 41304, "epoch": 464, "lr": 9.450747798503231e-05}
{"train_loss": 0.3584555685520172, "global_step": 41305, "epoch": 464, "lr": 9.450721381434518e-05}
{"train_loss": 0.3283604085445404, "global_step": 41306, "epoch": 464, "lr": 9.450694963767461e-05}
{"train_loss": 0.2934955060482025, "global_step": 41307, "epoch": 464, "lr": 9.450668545502062e-05}
{"train_loss": 0.4227871596813202, "global_step": 41308, "epoch": 464, "lr": 9.450642126638323e-05}
{"train_loss": 0.25981757044792175, "global_step": 41309, "epoch": 464, "lr": 9.450615707176252e-05}
{"train_loss": 0.35266995429992676, "global_step": 41310, "epoch": 464, "lr": 9.450589287115847e-05}
{"train_loss": 0.37767645716667175, "global_step": 41311, "epoch": 464, "lr": 9.450562866457117e-05}
{"train_loss": 0.2883185148239136, "global_step": 41312, "epoch": 464, "lr": 9.450536445200062e-05}
{"train_loss": 0.21952183544635773, "global_step": 41313, "epoch": 464, "lr": 9.450510023344687e-05}
{"train_loss": 0.35882776975631714, "global_step": 41314, "epoch": 464, "lr": 9.450483600890994e-05}
{"train_loss": 0.2997996211051941, "global_step": 41315, "epoch": 464, "lr": 9.45045717783899e-05}
{"train_loss": 0.39981919527053833, "global_step": 41316, "epoch": 464, "lr": 9.450430754188675e-05}
{"train_loss": 0.4042026996612549, "global_step": 41317, "epoch": 464, "lr": 9.450404329940054e-05}
{"train_loss": 0.28692537546157837, "global_step": 41318, "epoch": 464, "lr": 9.450377905093131e-05}
{"train_loss": 0.2641189396381378, "global_step": 41319, "epoch": 464, "lr": 9.450351479647907e-05}
{"train_loss": 0.4077821373939514, "global_step": 41320, "epoch": 464, "lr": 9.450325053604388e-05}
{"train_loss": 0.2933046221733093, "global_step": 41321, "epoch": 464, "lr": 9.45029862696258e-05}
{"train_loss": 0.3247142434120178, "global_step": 41322, "epoch": 464, "lr": 9.45027219972248e-05}
{"train_loss": 0.3056192100048065, "global_step": 41323, "epoch": 464, "lr": 9.450245771884096e-05}
{"train_loss": 0.28555893898010254, "global_step": 41324, "epoch": 464, "lr": 9.450219343447431e-05}
{"train_loss": 0.28017523884773254, "global_step": 41325, "epoch": 464, "lr": 9.450192914412489e-05}
{"train_loss": 0.2769656181335449, "global_step": 41326, "epoch": 464, "lr": 9.450166484779273e-05}
{"train_loss": 0.27849018573760986, "global_step": 41327, "epoch": 464, "lr": 9.450140054547784e-05}
{"train_loss": 0.3713446855545044, "global_step": 41328, "epoch": 464, "lr": 9.45011362371803e-05}
{"train_loss": 0.3489789068698883, "global_step": 41329, "epoch": 464, "lr": 9.450087192290011e-05}
{"train_loss": 0.34833478927612305, "global_step": 41330, "epoch": 464, "lr": 9.450060760263733e-05}
{"train_loss": 0.3816929757595062, "global_step": 41331, "epoch": 464, "lr": 9.450034327639199e-05}
{"train_loss": 0.4298485815525055, "global_step": 41332, "epoch": 464, "lr": 9.450007894416412e-05}
{"train_loss": 0.2731013298034668, "global_step": 41333, "epoch": 464, "lr": 9.449981460595375e-05}
{"train_loss": 0.3529067635536194, "global_step": 41334, "epoch": 464, "lr": 9.449955026176092e-05}
{"train_loss": 0.29438892006874084, "global_step": 41335, "epoch": 464, "lr": 9.449928591158567e-05}
{"train_loss": 0.2744646966457367, "global_step": 41336, "epoch": 464, "lr": 9.449902155542805e-05}
{"train_loss": 0.3758283853530884, "global_step": 41337, "epoch": 464, "lr": 9.449875719328807e-05}
{"train_loss": 0.2925432622432709, "global_step": 41338, "epoch": 464, "lr": 9.449849282516576e-05}
{"train_loss": 0.2758030295372009, "global_step": 41339, "epoch": 464, "lr": 9.449822845106117e-05}
{"train_loss": 0.31324824690818787, "global_step": 41340, "epoch": 464, "lr": 9.449796407097435e-05}
{"train_loss": 0.23338985443115234, "global_step": 41341, "epoch": 464, "lr": 9.449769968490532e-05}
{"train_loss": 0.37377893924713135, "global_step": 41342, "epoch": 464, "lr": 9.449743529285412e-05}
{"train_loss": 0.30859076976776123, "global_step": 41343, "epoch": 464, "lr": 9.449717089482077e-05}
{"train_loss": 0.3221200108528137, "global_step": 41344, "epoch": 464, "lr": 9.449690649080532e-05}
{"train_loss": 0.34558746218681335, "global_step": 41345, "epoch": 464, "lr": 9.449664208080782e-05}
{"train_loss": 0.3362426161766052, "global_step": 41346, "epoch": 464, "lr": 9.449637766482825e-05}
{"train_loss": 0.3180179297924042, "global_step": 41347, "epoch": 464, "lr": 9.449611324286673e-05}
{"train_loss": 0.3582112789154053, "global_step": 41348, "epoch": 464, "lr": 9.449584881492322e-05}
{"train_loss": 0.3029947876930237, "global_step": 41349, "epoch": 464, "lr": 9.44955843809978e-05}
{"train_loss": 0.33354467153549194, "global_step": 41350, "epoch": 464, "lr": 9.449531994109049e-05}
{"train_loss": 0.3411090075969696, "global_step": 41351, "epoch": 464, "lr": 9.449505549520133e-05}
{"train_loss": 0.2837271988391876, "global_step": 41352, "epoch": 464, "lr": 9.449479104333033e-05}
{"train_loss": 0.30397728085517883, "global_step": 41353, "epoch": 464, "lr": 9.449452658547757e-05}
{"train_loss": 0.3046395182609558, "global_step": 41354, "epoch": 464, "lr": 9.449426212164306e-05}
{"train_loss": 0.2491869181394577, "global_step": 41355, "epoch": 464, "lr": 9.449399765182682e-05}
{"train_loss": 0.24319350719451904, "global_step": 41356, "epoch": 464, "lr": 9.449373317602893e-05}
{"train_loss": 0.24920803308486938, "global_step": 41357, "epoch": 464, "lr": 9.449346869424938e-05}
{"train_loss": 0.2705629765987396, "global_step": 41358, "epoch": 464, "lr": 9.449320420648824e-05}
{"train_loss": 0.4236721694469452, "global_step": 41359, "epoch": 464, "lr": 9.449293971274551e-05}
{"train_loss": 0.3078547716140747, "global_step": 41360, "epoch": 464, "lr": 9.449267521302128e-05}
{"train_loss": 0.33189672231674194, "global_step": 41361, "epoch": 464, "lr": 9.449241070731553e-05}
{"train_loss": 0.2809169590473175, "global_step": 41362, "epoch": 464, "lr": 9.449214619562832e-05}
{"train_loss": 0.3989853262901306, "global_step": 41363, "epoch": 464, "lr": 9.449188167795969e-05}
{"train_loss": 0.305145800113678, "global_step": 41364, "epoch": 464, "lr": 9.449161715430966e-05}
{"train_loss": 0.2898593246936798, "global_step": 41365, "epoch": 464, "lr": 9.449135262467828e-05}
{"train_loss": 0.27609845995903015, "global_step": 41366, "epoch": 464, "lr": 9.449108808906557e-05}
{"train_loss": 0.22030043601989746, "global_step": 41367, "epoch": 464, "lr": 9.449082354747159e-05}
{"train_loss": 0.23016497492790222, "global_step": 41368, "epoch": 464, "lr": 9.449055899989635e-05}
{"train_loss": 0.27073153853416443, "global_step": 41369, "epoch": 464, "lr": 9.44902944463399e-05}
{"train_loss": 0.39798662066459656, "global_step": 41370, "epoch": 464, "lr": 9.449002988680227e-05}
{"train_loss": 0.3947036862373352, "global_step": 41371, "epoch": 464, "lr": 9.44897653212835e-05}
{"train_loss": 0.24088603258132935, "global_step": 41372, "epoch": 464, "lr": 9.448950074978362e-05}
{"train_loss": 0.3168678283691406, "global_step": 41373, "epoch": 464, "lr": 9.448923617230267e-05}
{"train_loss": 0.27435779571533203, "global_step": 41374, "epoch": 464, "lr": 9.448897158884069e-05}
{"train_loss": 0.3670939803123474, "global_step": 41375, "epoch": 464, "lr": 9.448870699939771e-05}
{"train_loss": 0.36872053146362305, "global_step": 41376, "epoch": 464, "lr": 9.448844240397376e-05}
{"train_loss": 0.38085561990737915, "global_step": 41377, "epoch": 464, "lr": 9.448817780256889e-05}
{"train_loss": 0.36430439352989197, "global_step": 41378, "epoch": 464, "lr": 9.448791319518312e-05}
{"train_loss": 0.35017848014831543, "global_step": 41379, "epoch": 464, "lr": 9.448764858181649e-05}
{"train_loss": 0.4389706552028656, "global_step": 41380, "epoch": 464, "lr": 9.448738396246904e-05}
{"train_loss": 0.23231132328510284, "global_step": 41381, "epoch": 464, "lr": 9.44871193371408e-05}
{"train_loss": 0.30819207429885864, "global_step": 41382, "epoch": 464, "lr": 9.448685470583182e-05}
{"train_loss": 0.46808314323425293, "global_step": 41383, "epoch": 464, "lr": 9.448659006854213e-05}
{"train_loss": 0.32086687292275806, "global_step": 41384, "epoch": 464, "lr": 9.448632542527173e-05, "val_loss": 2.757232189178467}
{"train_loss": 0.18522514402866364, "global_step": 41385, "epoch": 465, "lr": 9.448606077602071e-05}
{"train_loss": 0.25711822509765625, "global_step": 41386, "epoch": 465, "lr": 9.448579612078907e-05}
{"train_loss": 0.2942618727684021, "global_step": 41387, "epoch": 465, "lr": 9.448553145957685e-05}
{"train_loss": 0.31400230526924133, "global_step": 41388, "epoch": 465, "lr": 9.448526679238412e-05}
{"train_loss": 0.3172989785671234, "global_step": 41389, "epoch": 465, "lr": 9.448500211921088e-05}
{"train_loss": 0.2622130811214447, "global_step": 41390, "epoch": 465, "lr": 9.448473744005717e-05}
{"train_loss": 0.42888739705085754, "global_step": 41391, "epoch": 465, "lr": 9.448447275492302e-05}
{"train_loss": 0.3086872696876526, "global_step": 41392, "epoch": 465, "lr": 9.448420806380848e-05}
{"train_loss": 0.2260054051876068, "global_step": 41393, "epoch": 465, "lr": 9.448394336671357e-05}
{"train_loss": 0.2984139621257782, "global_step": 41394, "epoch": 465, "lr": 9.448367866363836e-05}
{"train_loss": 0.31264734268188477, "global_step": 41395, "epoch": 465, "lr": 9.448341395458285e-05}
{"train_loss": 0.2477370947599411, "global_step": 41396, "epoch": 465, "lr": 9.448314923954708e-05}
{"train_loss": 0.30411002039909363, "global_step": 41397, "epoch": 465, "lr": 9.448288451853109e-05}
{"train_loss": 0.355373352766037, "global_step": 41398, "epoch": 465, "lr": 9.448261979153493e-05}
{"train_loss": 0.22247269749641418, "global_step": 41399, "epoch": 465, "lr": 9.448235505855863e-05}
{"train_loss": 0.32378026843070984, "global_step": 41400, "epoch": 465, "lr": 9.44820903196022e-05}
{"train_loss": 0.25277310609817505, "global_step": 41401, "epoch": 465, "lr": 9.448182557466571e-05}
{"train_loss": 0.3272014260292053, "global_step": 41402, "epoch": 465, "lr": 9.448156082374917e-05}
{"train_loss": 0.2913341522216797, "global_step": 41403, "epoch": 465, "lr": 9.448129606685263e-05}
{"train_loss": 0.3691210150718689, "global_step": 41404, "epoch": 465, "lr": 9.448103130397613e-05}
{"train_loss": 0.26845136284828186, "global_step": 41405, "epoch": 465, "lr": 9.44807665351197e-05}
{"train_loss": 0.2858569920063019, "global_step": 41406, "epoch": 465, "lr": 9.448050176028336e-05}
{"train_loss": 0.24903546273708344, "global_step": 41407, "epoch": 465, "lr": 9.448023697946717e-05}
{"train_loss": 0.29349252581596375, "global_step": 41408, "epoch": 465, "lr": 9.447997219267114e-05}
{"train_loss": 0.2530859112739563, "global_step": 41409, "epoch": 465, "lr": 9.447970739989532e-05}
{"train_loss": 0.3255443274974823, "global_step": 41410, "epoch": 465, "lr": 9.447944260113975e-05}
{"train_loss": 0.2563272714614868, "global_step": 41411, "epoch": 465, "lr": 9.447917779640447e-05}
{"train_loss": 0.2660253942012787, "global_step": 41412, "epoch": 465, "lr": 9.447891298568949e-05}
{"train_loss": 0.25344106554985046, "global_step": 41413, "epoch": 465, "lr": 9.447864816899488e-05}
{"train_loss": 0.25088536739349365, "global_step": 41414, "epoch": 465, "lr": 9.447838334632066e-05}
{"train_loss": 0.27723604440689087, "global_step": 41415, "epoch": 465, "lr": 9.447811851766686e-05}
{"train_loss": 0.2984468340873718, "global_step": 41416, "epoch": 465, "lr": 9.447785368303351e-05}
{"train_loss": 0.23912817239761353, "global_step": 41417, "epoch": 465, "lr": 9.447758884242065e-05}
{"train_loss": 0.22006632387638092, "global_step": 41418, "epoch": 465, "lr": 9.447732399582833e-05}
{"train_loss": 0.31294023990631104, "global_step": 41419, "epoch": 465, "lr": 9.447705914325657e-05}
{"train_loss": 0.2806738317012787, "global_step": 41420, "epoch": 465, "lr": 9.447679428470541e-05}
{"train_loss": 0.2544509768486023, "global_step": 41421, "epoch": 465, "lr": 9.44765294201749e-05}
{"train_loss": 0.37665706872940063, "global_step": 41422, "epoch": 465, "lr": 9.447626454966505e-05}
{"train_loss": 0.33271607756614685, "global_step": 41423, "epoch": 465, "lr": 9.447599967317592e-05}
{"train_loss": 0.31916990876197815, "global_step": 41424, "epoch": 465, "lr": 9.447573479070752e-05}
{"train_loss": 0.22202789783477783, "global_step": 41425, "epoch": 465, "lr": 9.447546990225992e-05}
{"train_loss": 0.30327966809272766, "global_step": 41426, "epoch": 465, "lr": 9.447520500783312e-05}
{"train_loss": 0.40195080637931824, "global_step": 41427, "epoch": 465, "lr": 9.447494010742717e-05}
{"train_loss": 0.3079676926136017, "global_step": 41428, "epoch": 465, "lr": 9.447467520104212e-05}
{"train_loss": 0.23205885291099548, "global_step": 41429, "epoch": 465, "lr": 9.447441028867797e-05}
{"train_loss": 0.2261614054441452, "global_step": 41430, "epoch": 465, "lr": 9.447414537033479e-05}
{"train_loss": 0.2221779227256775, "global_step": 41431, "epoch": 465, "lr": 9.447388044601261e-05}
{"train_loss": 0.318606972694397, "global_step": 41432, "epoch": 465, "lr": 9.447361551571145e-05}
{"train_loss": 0.2584950923919678, "global_step": 41433, "epoch": 465, "lr": 9.447335057943136e-05}
{"train_loss": 0.2727728486061096, "global_step": 41434, "epoch": 465, "lr": 9.447308563717236e-05}
{"train_loss": 0.4725106358528137, "global_step": 41435, "epoch": 465, "lr": 9.447282068893451e-05}
{"train_loss": 0.17156776785850525, "global_step": 41436, "epoch": 465, "lr": 9.44725557347178e-05}
{"train_loss": 0.4451692998409271, "global_step": 41437, "epoch": 465, "lr": 9.447229077452233e-05}
{"train_loss": 0.3104425370693207, "global_step": 41438, "epoch": 465, "lr": 9.447202580834809e-05}
{"train_loss": 0.18935564160346985, "global_step": 41439, "epoch": 465, "lr": 9.447176083619514e-05}
{"train_loss": 0.18869934976100922, "global_step": 41440, "epoch": 465, "lr": 9.44714958580635e-05}
{"train_loss": 0.39813441038131714, "global_step": 41441, "epoch": 465, "lr": 9.44712308739532e-05}
{"train_loss": 0.26405102014541626, "global_step": 41442, "epoch": 465, "lr": 9.447096588386429e-05}
{"train_loss": 0.29525884985923767, "global_step": 41443, "epoch": 465, "lr": 9.44707008877968e-05}
{"train_loss": 0.31423601508140564, "global_step": 41444, "epoch": 465, "lr": 9.447043588575075e-05}
{"train_loss": 0.2926543056964874, "global_step": 41445, "epoch": 465, "lr": 9.447017087772622e-05}
{"train_loss": 0.24194255471229553, "global_step": 41446, "epoch": 465, "lr": 9.446990586372319e-05}
{"train_loss": 0.25202691555023193, "global_step": 41447, "epoch": 465, "lr": 9.446964084374174e-05}
{"train_loss": 0.34424176812171936, "global_step": 41448, "epoch": 465, "lr": 9.44693758177819e-05}
{"train_loss": 0.3026491403579712, "global_step": 41449, "epoch": 465, "lr": 9.446911078584368e-05}
{"train_loss": 0.29972711205482483, "global_step": 41450, "epoch": 465, "lr": 9.446884574792713e-05}
{"train_loss": 0.32255327701568604, "global_step": 41451, "epoch": 465, "lr": 9.446858070403228e-05}
{"train_loss": 0.2689133286476135, "global_step": 41452, "epoch": 465, "lr": 9.446831565415919e-05}
{"train_loss": 0.32191672921180725, "global_step": 41453, "epoch": 465, "lr": 9.446805059830786e-05}
{"train_loss": 0.21836720407009125, "global_step": 41454, "epoch": 465, "lr": 9.446778553647834e-05}
{"train_loss": 0.3185039758682251, "global_step": 41455, "epoch": 465, "lr": 9.446752046867069e-05}
{"train_loss": 0.3921661376953125, "global_step": 41456, "epoch": 465, "lr": 9.44672553948849e-05}
{"train_loss": 0.2855915129184723, "global_step": 41457, "epoch": 465, "lr": 9.446699031512105e-05}
{"train_loss": 0.2972545027732849, "global_step": 41458, "epoch": 465, "lr": 9.446672522937914e-05}
{"train_loss": 0.3770095109939575, "global_step": 41459, "epoch": 465, "lr": 9.446646013765923e-05}
{"train_loss": 0.27609559893608093, "global_step": 41460, "epoch": 465, "lr": 9.446619503996134e-05}
{"train_loss": 0.28575658798217773, "global_step": 41461, "epoch": 465, "lr": 9.446592993628551e-05}
{"train_loss": 0.2684202194213867, "global_step": 41462, "epoch": 465, "lr": 9.446566482663179e-05}
{"train_loss": 0.3108328580856323, "global_step": 41463, "epoch": 465, "lr": 9.446539971100019e-05}
{"train_loss": 0.3976203501224518, "global_step": 41464, "epoch": 465, "lr": 9.446513458939076e-05}
{"train_loss": 0.34703510999679565, "global_step": 41465, "epoch": 465, "lr": 9.446486946180353e-05}
{"train_loss": 0.23041903972625732, "global_step": 41466, "epoch": 465, "lr": 9.446460432823856e-05}
{"train_loss": 0.1824558973312378, "global_step": 41467, "epoch": 465, "lr": 9.446433918869586e-05}
{"train_loss": 0.2997971177101135, "global_step": 41468, "epoch": 465, "lr": 9.446407404317545e-05}
{"train_loss": 0.32218074798583984, "global_step": 41469, "epoch": 465, "lr": 9.44638088916774e-05}
{"train_loss": 0.5048698782920837, "global_step": 41470, "epoch": 465, "lr": 9.446354373420174e-05}
{"train_loss": 0.35308003425598145, "global_step": 41471, "epoch": 465, "lr": 9.44632785707485e-05}
{"train_loss": 0.33585524559020996, "global_step": 41472, "epoch": 465, "lr": 9.446301340131771e-05}
{"train_loss": 0.294031096978134, "global_step": 41473, "epoch": 465, "lr": 9.446274822590942e-05, "val_loss": 2.7946624755859375, "train_action_mse_error": 27.79039764404297}
{"train_loss": 0.29846516251564026, "global_step": 41474, "epoch": 466, "lr": 9.446248304452363e-05}
{"train_loss": 0.36469390988349915, "global_step": 41475, "epoch": 466, "lr": 9.446221785716042e-05}
{"train_loss": 0.3689926266670227, "global_step": 41476, "epoch": 466, "lr": 9.446195266381981e-05}
{"train_loss": 0.37914153933525085, "global_step": 41477, "epoch": 466, "lr": 9.446168746450182e-05}
{"train_loss": 0.23391564190387726, "global_step": 41478, "epoch": 466, "lr": 9.44614222592065e-05}
{"train_loss": 0.3143616020679474, "global_step": 41479, "epoch": 466, "lr": 9.446115704793388e-05}
{"train_loss": 0.4245552122592926, "global_step": 41480, "epoch": 466, "lr": 9.446089183068401e-05}
{"train_loss": 0.31940537691116333, "global_step": 41481, "epoch": 466, "lr": 9.446062660745692e-05}
{"train_loss": 0.282133549451828, "global_step": 41482, "epoch": 466, "lr": 9.446036137825263e-05}
{"train_loss": 0.36578816175460815, "global_step": 41483, "epoch": 466, "lr": 9.446009614307119e-05}
{"train_loss": 0.26889768242836, "global_step": 41484, "epoch": 466, "lr": 9.445983090191262e-05}
{"train_loss": 0.2598060667514801, "global_step": 41485, "epoch": 466, "lr": 9.445956565477698e-05}
{"train_loss": 0.2766192853450775, "global_step": 41486, "epoch": 466, "lr": 9.445930040166429e-05}
{"train_loss": 0.1410028040409088, "global_step": 41487, "epoch": 466, "lr": 9.44590351425746e-05}
{"train_loss": 0.27560433745384216, "global_step": 41488, "epoch": 466, "lr": 9.44587698775079e-05}
{"train_loss": 0.3554968237876892, "global_step": 41489, "epoch": 466, "lr": 9.445850460646429e-05}
{"train_loss": 0.38355591893196106, "global_step": 41490, "epoch": 466, "lr": 9.445823932944378e-05}
{"train_loss": 0.27203017473220825, "global_step": 41491, "epoch": 466, "lr": 9.445797404644639e-05}
{"train_loss": 0.25858065485954285, "global_step": 41492, "epoch": 466, "lr": 9.445770875747217e-05}
{"train_loss": 0.2904140055179596, "global_step": 41493, "epoch": 466, "lr": 9.445744346252115e-05}
{"train_loss": 0.2434384822845459, "global_step": 41494, "epoch": 466, "lr": 9.445717816159337e-05}
{"train_loss": 0.2558884620666504, "global_step": 41495, "epoch": 466, "lr": 9.445691285468886e-05}
{"train_loss": 0.3625792860984802, "global_step": 41496, "epoch": 466, "lr": 9.445664754180767e-05}
{"train_loss": 0.2840997278690338, "global_step": 41497, "epoch": 466, "lr": 9.445638222294982e-05}
{"train_loss": 0.2677460312843323, "global_step": 41498, "epoch": 466, "lr": 9.445611689811534e-05}
{"train_loss": 0.2862144112586975, "global_step": 41499, "epoch": 466, "lr": 9.44558515673043e-05}
{"train_loss": 0.3559679090976715, "global_step": 41500, "epoch": 466, "lr": 9.445558623051669e-05}
{"train_loss": 0.31698623299598694, "global_step": 41501, "epoch": 466, "lr": 9.445532088775258e-05}
{"train_loss": 0.29741042852401733, "global_step": 41502, "epoch": 466, "lr": 9.445505553901199e-05}
{"train_loss": 0.32297283411026, "global_step": 41503, "epoch": 466, "lr": 9.445479018429498e-05}
{"train_loss": 0.2918740510940552, "global_step": 41504, "epoch": 466, "lr": 9.445452482360153e-05}
{"train_loss": 0.24458175897598267, "global_step": 41505, "epoch": 466, "lr": 9.445425945693174e-05}
{"train_loss": 0.31735536456108093, "global_step": 41506, "epoch": 466, "lr": 9.44539940842856e-05}
{"train_loss": 0.3718164265155792, "global_step": 41507, "epoch": 466, "lr": 9.445372870566317e-05}
{"train_loss": 0.40587496757507324, "global_step": 41508, "epoch": 466, "lr": 9.445346332106447e-05}
{"train_loss": 0.3436959385871887, "global_step": 41509, "epoch": 466, "lr": 9.445319793048956e-05}
{"train_loss": 0.14345552027225494, "global_step": 41510, "epoch": 466, "lr": 9.445293253393844e-05}
{"train_loss": 0.24492813646793365, "global_step": 41511, "epoch": 466, "lr": 9.445266713141117e-05}
{"train_loss": 0.31074410676956177, "global_step": 41512, "epoch": 466, "lr": 9.44524017229078e-05}
{"train_loss": 0.3709867596626282, "global_step": 41513, "epoch": 466, "lr": 9.445213630842833e-05}
{"train_loss": 0.3672489821910858, "global_step": 41514, "epoch": 466, "lr": 9.44518708879728e-05}
{"train_loss": 0.30586162209510803, "global_step": 41515, "epoch": 466, "lr": 9.445160546154128e-05}
{"train_loss": 0.28466084599494934, "global_step": 41516, "epoch": 466, "lr": 9.445134002913378e-05}
{"train_loss": 0.22971664369106293, "global_step": 41517, "epoch": 466, "lr": 9.445107459075032e-05}
{"train_loss": 0.2435748279094696, "global_step": 41518, "epoch": 466, "lr": 9.445080914639097e-05}
{"train_loss": 0.23696160316467285, "global_step": 41519, "epoch": 466, "lr": 9.445054369605575e-05}
{"train_loss": 0.17261111736297607, "global_step": 41520, "epoch": 466, "lr": 9.44502782397447e-05}
{"train_loss": 0.3283023238182068, "global_step": 41521, "epoch": 466, "lr": 9.445001277745784e-05}
{"train_loss": 0.3718099594116211, "global_step": 41522, "epoch": 466, "lr": 9.444974730919523e-05}
{"train_loss": 0.3134797513484955, "global_step": 41523, "epoch": 466, "lr": 9.444948183495688e-05}
{"train_loss": 0.25419268012046814, "global_step": 41524, "epoch": 466, "lr": 9.444921635474285e-05}
{"train_loss": 0.27050724625587463, "global_step": 41525, "epoch": 466, "lr": 9.444895086855317e-05}
{"train_loss": 0.37824177742004395, "global_step": 41526, "epoch": 466, "lr": 9.444868537638785e-05}
{"train_loss": 0.29855895042419434, "global_step": 41527, "epoch": 466, "lr": 9.444841987824695e-05}
{"train_loss": 0.21578624844551086, "global_step": 41528, "epoch": 466, "lr": 9.444815437413053e-05}
{"train_loss": 0.24188095331192017, "global_step": 41529, "epoch": 466, "lr": 9.444788886403856e-05}
{"train_loss": 0.32039910554885864, "global_step": 41530, "epoch": 466, "lr": 9.444762334797114e-05}
{"train_loss": 0.2899891138076782, "global_step": 41531, "epoch": 466, "lr": 9.444735782592826e-05}
{"train_loss": 0.3224881589412689, "global_step": 41532, "epoch": 466, "lr": 9.444709229791e-05}
{"train_loss": 0.28520911931991577, "global_step": 41533, "epoch": 466, "lr": 9.444682676391634e-05}
{"train_loss": 0.3039794862270355, "global_step": 41534, "epoch": 466, "lr": 9.444656122394736e-05}
{"train_loss": 0.35163185000419617, "global_step": 41535, "epoch": 466, "lr": 9.444629567800307e-05}
{"train_loss": 0.2403637170791626, "global_step": 41536, "epoch": 466, "lr": 9.444603012608352e-05}
{"train_loss": 0.24727453291416168, "global_step": 41537, "epoch": 466, "lr": 9.444576456818877e-05}
{"train_loss": 0.30277010798454285, "global_step": 41538, "epoch": 466, "lr": 9.44454990043188e-05}
{"train_loss": 0.2943435311317444, "global_step": 41539, "epoch": 466, "lr": 9.444523343447369e-05}
{"train_loss": 0.3519735336303711, "global_step": 41540, "epoch": 466, "lr": 9.444496785865346e-05}
{"train_loss": 0.32667332887649536, "global_step": 41541, "epoch": 466, "lr": 9.444470227685813e-05}
{"train_loss": 0.2602919638156891, "global_step": 41542, "epoch": 466, "lr": 9.444443668908776e-05}
{"train_loss": 0.31471359729766846, "global_step": 41543, "epoch": 466, "lr": 9.444417109534238e-05}
{"train_loss": 0.3127981126308441, "global_step": 41544, "epoch": 466, "lr": 9.444390549562202e-05}
{"train_loss": 0.2796189486980438, "global_step": 41545, "epoch": 466, "lr": 9.444363988992672e-05}
{"train_loss": 0.310525119304657, "global_step": 41546, "epoch": 466, "lr": 9.444337427825652e-05}
{"train_loss": 0.28276076912879944, "global_step": 41547, "epoch": 466, "lr": 9.444310866061144e-05}
{"train_loss": 0.2947264611721039, "global_step": 41548, "epoch": 466, "lr": 9.444284303699154e-05}
{"train_loss": 0.2966454029083252, "global_step": 41549, "epoch": 466, "lr": 9.444257740739681e-05}
{"train_loss": 0.3846825957298279, "global_step": 41550, "epoch": 466, "lr": 9.444231177182735e-05}
{"train_loss": 0.33430731296539307, "global_step": 41551, "epoch": 466, "lr": 9.444204613028316e-05}
{"train_loss": 0.40534234046936035, "global_step": 41552, "epoch": 466, "lr": 9.444178048276426e-05}
{"train_loss": 0.321469783782959, "global_step": 41553, "epoch": 466, "lr": 9.444151482927071e-05}
{"train_loss": 0.3429478406906128, "global_step": 41554, "epoch": 466, "lr": 9.444124916980255e-05}
{"train_loss": 0.2768721878528595, "global_step": 41555, "epoch": 466, "lr": 9.444098350435979e-05}
{"train_loss": 0.2892610430717468, "global_step": 41556, "epoch": 466, "lr": 9.44407178329425e-05}
{"train_loss": 0.30036070942878723, "global_step": 41557, "epoch": 466, "lr": 9.444045215555068e-05}
{"train_loss": 0.3033866286277771, "global_step": 41558, "epoch": 466, "lr": 9.44401864721844e-05}
{"train_loss": 0.32927003502845764, "global_step": 41559, "epoch": 466, "lr": 9.443992078284367e-05}
{"train_loss": 0.3852306008338928, "global_step": 41560, "epoch": 466, "lr": 9.443965508752853e-05}
{"train_loss": 0.39926397800445557, "global_step": 41561, "epoch": 466, "lr": 9.443938938623903e-05}
{"train_loss": 0.30468001911479436, "global_step": 41562, "epoch": 466, "lr": 9.443912367897518e-05, "val_loss": 2.6464743614196777}
{"train_loss": 0.1989823877811432, "global_step": 41563, "epoch": 467, "lr": 9.443885796573705e-05}
{"train_loss": 0.42879027128219604, "global_step": 41564, "epoch": 467, "lr": 9.443859224652464e-05}
{"train_loss": 0.3203674852848053, "global_step": 41565, "epoch": 467, "lr": 9.443832652133802e-05}
{"train_loss": 0.3304738700389862, "global_step": 41566, "epoch": 467, "lr": 9.443806079017719e-05}
{"train_loss": 0.3392721712589264, "global_step": 41567, "epoch": 467, "lr": 9.443779505304222e-05}
{"train_loss": 0.3810760974884033, "global_step": 41568, "epoch": 467, "lr": 9.443752930993313e-05}
{"train_loss": 0.3181368410587311, "global_step": 41569, "epoch": 467, "lr": 9.443726356084995e-05}
{"train_loss": 0.2734004259109497, "global_step": 41570, "epoch": 467, "lr": 9.443699780579272e-05}
{"train_loss": 0.3228663206100464, "global_step": 41571, "epoch": 467, "lr": 9.443673204476148e-05}
{"train_loss": 0.3377784788608551, "global_step": 41572, "epoch": 467, "lr": 9.443646627775626e-05}
{"train_loss": 0.22749266028404236, "global_step": 41573, "epoch": 467, "lr": 9.44362005047771e-05}
{"train_loss": 0.24454104900360107, "global_step": 41574, "epoch": 467, "lr": 9.443593472582403e-05}
{"train_loss": 0.24307787418365479, "global_step": 41575, "epoch": 467, "lr": 9.44356689408971e-05}
{"train_loss": 0.32881271839141846, "global_step": 41576, "epoch": 467, "lr": 9.443540314999632e-05}
{"train_loss": 0.3898267149925232, "global_step": 41577, "epoch": 467, "lr": 9.443513735312177e-05}
{"train_loss": 0.3061995506286621, "global_step": 41578, "epoch": 467, "lr": 9.443487155027344e-05}
{"train_loss": 0.3033613860607147, "global_step": 41579, "epoch": 467, "lr": 9.443460574145138e-05}
{"train_loss": 0.345203697681427, "global_step": 41580, "epoch": 467, "lr": 9.443433992665563e-05}
{"train_loss": 0.2934185862541199, "global_step": 41581, "epoch": 467, "lr": 9.443407410588623e-05}
{"train_loss": 0.28972864151000977, "global_step": 41582, "epoch": 467, "lr": 9.443380827914321e-05}
{"train_loss": 0.2381887286901474, "global_step": 41583, "epoch": 467, "lr": 9.44335424464266e-05}
{"train_loss": 0.3958794176578522, "global_step": 41584, "epoch": 467, "lr": 9.443327660773644e-05}
{"train_loss": 0.3385089337825775, "global_step": 41585, "epoch": 467, "lr": 9.443301076307279e-05}
{"train_loss": 0.2512024939060211, "global_step": 41586, "epoch": 467, "lr": 9.443274491243563e-05}
{"train_loss": 0.34091782569885254, "global_step": 41587, "epoch": 467, "lr": 9.443247905582504e-05}
{"train_loss": 0.2454545795917511, "global_step": 41588, "epoch": 467, "lr": 9.443221319324106e-05}
{"train_loss": 0.29332342743873596, "global_step": 41589, "epoch": 467, "lr": 9.44319473246837e-05}
{"train_loss": 0.2864861786365509, "global_step": 41590, "epoch": 467, "lr": 9.4431681450153e-05}
{"train_loss": 0.3311808705329895, "global_step": 41591, "epoch": 467, "lr": 9.443141556964901e-05}
{"train_loss": 0.31286394596099854, "global_step": 41592, "epoch": 467, "lr": 9.443114968317175e-05}
{"train_loss": 0.3240184187889099, "global_step": 41593, "epoch": 467, "lr": 9.443088379072127e-05}
{"train_loss": 0.27525025606155396, "global_step": 41594, "epoch": 467, "lr": 9.44306178922976e-05}
{"train_loss": 0.2945159077644348, "global_step": 41595, "epoch": 467, "lr": 9.443035198790078e-05}
{"train_loss": 0.19806602597236633, "global_step": 41596, "epoch": 467, "lr": 9.443008607753083e-05}
{"train_loss": 0.4113375246524811, "global_step": 41597, "epoch": 467, "lr": 9.44298201611878e-05}
{"train_loss": 0.3078770339488983, "global_step": 41598, "epoch": 467, "lr": 9.442955423887172e-05}
{"train_loss": 0.2362353801727295, "global_step": 41599, "epoch": 467, "lr": 9.442928831058264e-05}
{"train_loss": 0.32976657152175903, "global_step": 41600, "epoch": 467, "lr": 9.442902237632056e-05}
{"train_loss": 0.31655991077423096, "global_step": 41601, "epoch": 467, "lr": 9.442875643608556e-05}
{"train_loss": 0.26359614729881287, "global_step": 41602, "epoch": 467, "lr": 9.442849048987765e-05}
{"train_loss": 0.19268859922885895, "global_step": 41603, "epoch": 467, "lr": 9.442822453769688e-05}
{"train_loss": 0.3217478394508362, "global_step": 41604, "epoch": 467, "lr": 9.442795857954327e-05}
{"train_loss": 0.24934549629688263, "global_step": 41605, "epoch": 467, "lr": 9.442769261541685e-05}
{"train_loss": 0.31016620993614197, "global_step": 41606, "epoch": 467, "lr": 9.442742664531768e-05}
{"train_loss": 0.339909166097641, "global_step": 41607, "epoch": 467, "lr": 9.442716066924578e-05}
{"train_loss": 0.18855629861354828, "global_step": 41608, "epoch": 467, "lr": 9.442689468720119e-05}
{"train_loss": 0.25575369596481323, "global_step": 41609, "epoch": 467, "lr": 9.442662869918395e-05}
{"train_loss": 0.25171783566474915, "global_step": 41610, "epoch": 467, "lr": 9.442636270519409e-05}
{"train_loss": 0.4097791016101837, "global_step": 41611, "epoch": 467, "lr": 9.442609670523164e-05}
{"train_loss": 0.4137625992298126, "global_step": 41612, "epoch": 467, "lr": 9.442583069929664e-05}
{"train_loss": 0.3200271427631378, "global_step": 41613, "epoch": 467, "lr": 9.442556468738915e-05}
{"train_loss": 0.3432523012161255, "global_step": 41614, "epoch": 467, "lr": 9.442529866950916e-05}
{"train_loss": 0.21633721888065338, "global_step": 41615, "epoch": 467, "lr": 9.442503264565674e-05}
{"train_loss": 0.27067863941192627, "global_step": 41616, "epoch": 467, "lr": 9.442476661583193e-05}
{"train_loss": 0.2276322990655899, "global_step": 41617, "epoch": 467, "lr": 9.442450058003474e-05}
{"train_loss": 0.26432302594184875, "global_step": 41618, "epoch": 467, "lr": 9.442423453826521e-05}
{"train_loss": 0.29348403215408325, "global_step": 41619, "epoch": 467, "lr": 9.442396849052339e-05}
{"train_loss": 0.33867916464805603, "global_step": 41620, "epoch": 467, "lr": 9.442370243680931e-05}
{"train_loss": 0.25952690839767456, "global_step": 41621, "epoch": 467, "lr": 9.442343637712298e-05}
{"train_loss": 0.3033560812473297, "global_step": 41622, "epoch": 467, "lr": 9.44231703114645e-05}
{"train_loss": 0.18385516107082367, "global_step": 41623, "epoch": 467, "lr": 9.442290423983385e-05}
{"train_loss": 0.33354753255844116, "global_step": 41624, "epoch": 467, "lr": 9.442263816223107e-05}
{"train_loss": 0.3201045095920563, "global_step": 41625, "epoch": 467, "lr": 9.442237207865623e-05}
{"train_loss": 0.2202402949333191, "global_step": 41626, "epoch": 467, "lr": 9.442210598910932e-05}
{"train_loss": 0.299734890460968, "global_step": 41627, "epoch": 467, "lr": 9.442183989359041e-05}
{"train_loss": 0.2789766788482666, "global_step": 41628, "epoch": 467, "lr": 9.442157379209952e-05}
{"train_loss": 0.2581692039966583, "global_step": 41629, "epoch": 467, "lr": 9.44213076846367e-05}
{"train_loss": 0.3786279857158661, "global_step": 41630, "epoch": 467, "lr": 9.442104157120199e-05}
{"train_loss": 0.22905997931957245, "global_step": 41631, "epoch": 467, "lr": 9.44207754517954e-05}
{"train_loss": 0.39152273535728455, "global_step": 41632, "epoch": 467, "lr": 9.442050932641697e-05}
{"train_loss": 0.22220952808856964, "global_step": 41633, "epoch": 467, "lr": 9.442024319506675e-05}
{"train_loss": 0.2772827744483948, "global_step": 41634, "epoch": 467, "lr": 9.441997705774477e-05}
{"train_loss": 0.3491475582122803, "global_step": 41635, "epoch": 467, "lr": 9.441971091445105e-05}
{"train_loss": 0.34727323055267334, "global_step": 41636, "epoch": 467, "lr": 9.441944476518566e-05}
{"train_loss": 0.4163714647293091, "global_step": 41637, "epoch": 467, "lr": 9.441917860994861e-05}
{"train_loss": 0.2760758101940155, "global_step": 41638, "epoch": 467, "lr": 9.441891244873995e-05}
{"train_loss": 0.4993053674697876, "global_step": 41639, "epoch": 467, "lr": 9.44186462815597e-05}
{"train_loss": 0.2917575240135193, "global_step": 41640, "epoch": 467, "lr": 9.441838010840791e-05}
{"train_loss": 0.30387914180755615, "global_step": 41641, "epoch": 467, "lr": 9.441811392928462e-05}
{"train_loss": 0.22120678424835205, "global_step": 41642, "epoch": 467, "lr": 9.441784774418983e-05}
{"train_loss": 0.3575373589992523, "global_step": 41643, "epoch": 467, "lr": 9.441758155312361e-05}
{"train_loss": 0.20551970601081848, "global_step": 41644, "epoch": 467, "lr": 9.4417315356086e-05}
{"train_loss": 0.34586629271507263, "global_step": 41645, "epoch": 467, "lr": 9.441704915307701e-05}
{"train_loss": 0.4503282904624939, "global_step": 41646, "epoch": 467, "lr": 9.441678294409671e-05}
{"train_loss": 0.34419357776641846, "global_step": 41647, "epoch": 467, "lr": 9.44165167291451e-05}
{"train_loss": 0.3638194501399994, "global_step": 41648, "epoch": 467, "lr": 9.441625050822222e-05}
{"train_loss": 0.2975921630859375, "global_step": 41649, "epoch": 467, "lr": 9.441598428132812e-05}
{"train_loss": 0.4344109296798706, "global_step": 41650, "epoch": 467, "lr": 9.441571804846285e-05}
{"train_loss": 0.3046865999028924, "global_step": 41651, "epoch": 467, "lr": 9.441545180962641e-05, "val_loss": 2.786609411239624}
{"train_loss": 0.30123451352119446, "global_step": 41652, "epoch": 468, "lr": 9.441518556481888e-05}
{"train_loss": 0.2292199432849884, "global_step": 41653, "epoch": 468, "lr": 9.441491931404024e-05}
{"train_loss": 0.27592507004737854, "global_step": 41654, "epoch": 468, "lr": 9.441465305729056e-05}
{"train_loss": 0.22618159651756287, "global_step": 41655, "epoch": 468, "lr": 9.441438679456989e-05}
{"train_loss": 0.39996960759162903, "global_step": 41656, "epoch": 468, "lr": 9.441412052587824e-05}
{"train_loss": 0.34450557827949524, "global_step": 41657, "epoch": 468, "lr": 9.441385425121563e-05}
{"train_loss": 0.2629733383655548, "global_step": 41658, "epoch": 468, "lr": 9.441358797058215e-05}
{"train_loss": 0.33097776770591736, "global_step": 41659, "epoch": 468, "lr": 9.441332168397779e-05}
{"train_loss": 0.31571218371391296, "global_step": 41660, "epoch": 468, "lr": 9.44130553914026e-05}
{"train_loss": 0.3560734987258911, "global_step": 41661, "epoch": 468, "lr": 9.441278909285661e-05}
{"train_loss": 0.32074347138404846, "global_step": 41662, "epoch": 468, "lr": 9.441252278833987e-05}
{"train_loss": 0.25393620133399963, "global_step": 41663, "epoch": 468, "lr": 9.441225647785241e-05}
{"train_loss": 0.42529919743537903, "global_step": 41664, "epoch": 468, "lr": 9.441199016139426e-05}
{"train_loss": 0.3226633667945862, "global_step": 41665, "epoch": 468, "lr": 9.441172383896547e-05}
{"train_loss": 0.25595951080322266, "global_step": 41666, "epoch": 468, "lr": 9.441145751056604e-05}
{"train_loss": 0.30704087018966675, "global_step": 41667, "epoch": 468, "lr": 9.441119117619605e-05}
{"train_loss": 0.37477925419807434, "global_step": 41668, "epoch": 468, "lr": 9.441092483585551e-05}
{"train_loss": 0.23875318467617035, "global_step": 41669, "epoch": 468, "lr": 9.441065848954447e-05}
{"train_loss": 0.2707807421684265, "global_step": 41670, "epoch": 468, "lr": 9.441039213726295e-05}
{"train_loss": 0.2490064650774002, "global_step": 41671, "epoch": 468, "lr": 9.441012577901098e-05}
{"train_loss": 0.1809420883655548, "global_step": 41672, "epoch": 468, "lr": 9.440985941478864e-05}
{"train_loss": 0.3331088125705719, "global_step": 41673, "epoch": 468, "lr": 9.440959304459592e-05}
{"train_loss": 0.2523476779460907, "global_step": 41674, "epoch": 468, "lr": 9.440932666843287e-05}
{"train_loss": 0.24069994688034058, "global_step": 41675, "epoch": 468, "lr": 9.440906028629954e-05}
{"train_loss": 0.3196423649787903, "global_step": 41676, "epoch": 468, "lr": 9.440879389819593e-05}
{"train_loss": 0.2881790101528168, "global_step": 41677, "epoch": 468, "lr": 9.440852750412212e-05}
{"train_loss": 0.22414085268974304, "global_step": 41678, "epoch": 468, "lr": 9.440826110407812e-05}
{"train_loss": 0.27118757367134094, "global_step": 41679, "epoch": 468, "lr": 9.440799469806397e-05}
{"train_loss": 0.2649570405483246, "global_step": 41680, "epoch": 468, "lr": 9.44077282860797e-05}
{"train_loss": 0.3167111277580261, "global_step": 41681, "epoch": 468, "lr": 9.440746186812536e-05}
{"train_loss": 0.34384146332740784, "global_step": 41682, "epoch": 468, "lr": 9.440719544420098e-05}
{"train_loss": 0.3061486780643463, "global_step": 41683, "epoch": 468, "lr": 9.44069290143066e-05}
{"train_loss": 0.3443059027194977, "global_step": 41684, "epoch": 468, "lr": 9.440666257844222e-05}
{"train_loss": 0.31429556012153625, "global_step": 41685, "epoch": 468, "lr": 9.440639613660792e-05}
{"train_loss": 0.38402247428894043, "global_step": 41686, "epoch": 468, "lr": 9.440612968880374e-05}
{"train_loss": 0.29882508516311646, "global_step": 41687, "epoch": 468, "lr": 9.440586323502967e-05}
{"train_loss": 0.2193533182144165, "global_step": 41688, "epoch": 468, "lr": 9.440559677528579e-05}
{"train_loss": 0.33649659156799316, "global_step": 41689, "epoch": 468, "lr": 9.440533030957211e-05}
{"train_loss": 0.32661938667297363, "global_step": 41690, "epoch": 468, "lr": 9.440506383788867e-05}
{"train_loss": 0.3010651171207428, "global_step": 41691, "epoch": 468, "lr": 9.440479736023554e-05}
{"train_loss": 0.3035613000392914, "global_step": 41692, "epoch": 468, "lr": 9.44045308766127e-05}
{"train_loss": 0.2432059794664383, "global_step": 41693, "epoch": 468, "lr": 9.440426438702021e-05}
{"train_loss": 0.35424521565437317, "global_step": 41694, "epoch": 468, "lr": 9.44039978914581e-05}
{"train_loss": 0.2867203950881958, "global_step": 41695, "epoch": 468, "lr": 9.440373138992644e-05}
{"train_loss": 0.2389651983976364, "global_step": 41696, "epoch": 468, "lr": 9.440346488242524e-05}
{"train_loss": 0.31199610233306885, "global_step": 41697, "epoch": 468, "lr": 9.440319836895452e-05}
{"train_loss": 0.3620358109474182, "global_step": 41698, "epoch": 468, "lr": 9.440293184951432e-05}
{"train_loss": 0.2667391300201416, "global_step": 41699, "epoch": 468, "lr": 9.440266532410471e-05}
{"train_loss": 0.27165719866752625, "global_step": 41700, "epoch": 468, "lr": 9.44023987927257e-05}
{"train_loss": 0.42674362659454346, "global_step": 41701, "epoch": 468, "lr": 9.440213225537731e-05}
{"train_loss": 0.2264205664396286, "global_step": 41702, "epoch": 468, "lr": 9.440186571205961e-05}
{"train_loss": 0.29203730821609497, "global_step": 41703, "epoch": 468, "lr": 9.440159916277263e-05}
{"train_loss": 0.29005110263824463, "global_step": 41704, "epoch": 468, "lr": 9.440133260751638e-05}
{"train_loss": 0.2605978846549988, "global_step": 41705, "epoch": 468, "lr": 9.44010660462909e-05}
{"train_loss": 0.4166063368320465, "global_step": 41706, "epoch": 468, "lr": 9.440079947909626e-05}
{"train_loss": 0.355198472738266, "global_step": 41707, "epoch": 468, "lr": 9.440053290593247e-05}
{"train_loss": 0.3179810047149658, "global_step": 41708, "epoch": 468, "lr": 9.440026632679957e-05}
{"train_loss": 0.2898885905742645, "global_step": 41709, "epoch": 468, "lr": 9.439999974169759e-05}
{"train_loss": 0.25382906198501587, "global_step": 41710, "epoch": 468, "lr": 9.439973315062658e-05}
{"train_loss": 0.25872281193733215, "global_step": 41711, "epoch": 468, "lr": 9.439946655358655e-05}
{"train_loss": 0.3513757884502411, "global_step": 41712, "epoch": 468, "lr": 9.439919995057758e-05}
{"train_loss": 0.337565541267395, "global_step": 41713, "epoch": 468, "lr": 9.439893334159966e-05}
{"train_loss": 0.3075191080570221, "global_step": 41714, "epoch": 468, "lr": 9.439866672665284e-05}
{"train_loss": 0.315341055393219, "global_step": 41715, "epoch": 468, "lr": 9.439840010573717e-05}
{"train_loss": 0.26013126969337463, "global_step": 41716, "epoch": 468, "lr": 9.439813347885267e-05}
{"train_loss": 0.32291024923324585, "global_step": 41717, "epoch": 468, "lr": 9.439786684599938e-05}
{"train_loss": 0.3261156380176544, "global_step": 41718, "epoch": 468, "lr": 9.439760020717735e-05}
{"train_loss": 0.30660539865493774, "global_step": 41719, "epoch": 468, "lr": 9.43973335623866e-05}
{"train_loss": 0.24559848010540009, "global_step": 41720, "epoch": 468, "lr": 9.439706691162717e-05}
{"train_loss": 0.3371381461620331, "global_step": 41721, "epoch": 468, "lr": 9.439680025489909e-05}
{"train_loss": 0.35620710253715515, "global_step": 41722, "epoch": 468, "lr": 9.43965335922024e-05}
{"train_loss": 0.3481464684009552, "global_step": 41723, "epoch": 468, "lr": 9.439626692353713e-05}
{"train_loss": 0.3188633322715759, "global_step": 41724, "epoch": 468, "lr": 9.439600024890334e-05}
{"train_loss": 0.37385889887809753, "global_step": 41725, "epoch": 468, "lr": 9.439573356830105e-05}
{"train_loss": 0.4054119884967804, "global_step": 41726, "epoch": 468, "lr": 9.439546688173029e-05}
{"train_loss": 0.33147189021110535, "global_step": 41727, "epoch": 468, "lr": 9.439520018919109e-05}
{"train_loss": 0.3053603768348694, "global_step": 41728, "epoch": 468, "lr": 9.43949334906835e-05}
{"train_loss": 0.3981492221355438, "global_step": 41729, "epoch": 468, "lr": 9.439466678620757e-05}
{"train_loss": 0.3102916479110718, "global_step": 41730, "epoch": 468, "lr": 9.43944000757633e-05}
{"train_loss": 0.295904278755188, "global_step": 41731, "epoch": 468, "lr": 9.439413335935075e-05}
{"train_loss": 0.2948850691318512, "global_step": 41732, "epoch": 468, "lr": 9.439386663696994e-05}
{"train_loss": 0.3507286012172699, "global_step": 41733, "epoch": 468, "lr": 9.439359990862094e-05}
{"train_loss": 0.24664558470249176, "global_step": 41734, "epoch": 468, "lr": 9.439333317430374e-05}
{"train_loss": 0.2584238648414612, "global_step": 41735, "epoch": 468, "lr": 9.439306643401839e-05}
{"train_loss": 0.26678967475891113, "global_step": 41736, "epoch": 468, "lr": 9.439279968776496e-05}
{"train_loss": 0.2935798168182373, "global_step": 41737, "epoch": 468, "lr": 9.439253293554345e-05}
{"train_loss": 0.31629106402397156, "global_step": 41738, "epoch": 468, "lr": 9.439226617735391e-05}
{"train_loss": 0.2774175703525543, "global_step": 41739, "epoch": 468, "lr": 9.439199941319634e-05}
{"train_loss": 0.3044304983334595, "global_step": 41740, "epoch": 468, "lr": 9.439173264307084e-05, "val_loss": 2.7088358402252197}
{"train_loss": 0.41475608944892883, "global_step": 41741, "epoch": 469, "lr": 9.43914658669774e-05}
{"train_loss": 0.2590987980365753, "global_step": 41742, "epoch": 469, "lr": 9.439119908491607e-05}
{"train_loss": 0.33081555366516113, "global_step": 41743, "epoch": 469, "lr": 9.439093229688688e-05}
{"train_loss": 0.30105212330818176, "global_step": 41744, "epoch": 469, "lr": 9.439066550288988e-05}
{"train_loss": 0.28062084317207336, "global_step": 41745, "epoch": 469, "lr": 9.439039870292508e-05}
{"train_loss": 0.27324268221855164, "global_step": 41746, "epoch": 469, "lr": 9.439013189699255e-05}
{"train_loss": 0.2959083914756775, "global_step": 41747, "epoch": 469, "lr": 9.43898650850923e-05}
{"train_loss": 0.3950887620449066, "global_step": 41748, "epoch": 469, "lr": 9.438959826722438e-05}
{"train_loss": 0.31713616847991943, "global_step": 41749, "epoch": 469, "lr": 9.43893314433888e-05}
{"train_loss": 0.3632463812828064, "global_step": 41750, "epoch": 469, "lr": 9.438906461358563e-05}
{"train_loss": 0.3233698308467865, "global_step": 41751, "epoch": 469, "lr": 9.43887977778149e-05}
{"train_loss": 0.3371567130088806, "global_step": 41752, "epoch": 469, "lr": 9.438853093607663e-05}
{"train_loss": 0.2913617491722107, "global_step": 41753, "epoch": 469, "lr": 9.438826408837085e-05}
{"train_loss": 0.30506083369255066, "global_step": 41754, "epoch": 469, "lr": 9.438799723469763e-05}
{"train_loss": 0.3443937301635742, "global_step": 41755, "epoch": 469, "lr": 9.438773037505697e-05}
{"train_loss": 0.31121399998664856, "global_step": 41756, "epoch": 469, "lr": 9.438746350944891e-05}
{"train_loss": 0.36097246408462524, "global_step": 41757, "epoch": 469, "lr": 9.438719663787352e-05}
{"train_loss": 0.25371265411376953, "global_step": 41758, "epoch": 469, "lr": 9.43869297603308e-05}
{"train_loss": 0.3187236487865448, "global_step": 41759, "epoch": 469, "lr": 9.43866628768208e-05}
{"train_loss": 0.31984996795654297, "global_step": 41760, "epoch": 469, "lr": 9.438639598734356e-05}
{"train_loss": 0.32767194509506226, "global_step": 41761, "epoch": 469, "lr": 9.438612909189912e-05}
{"train_loss": 0.28122037649154663, "global_step": 41762, "epoch": 469, "lr": 9.438586219048749e-05}
{"train_loss": 0.270731121301651, "global_step": 41763, "epoch": 469, "lr": 9.438559528310871e-05}
{"train_loss": 0.3004155457019806, "global_step": 41764, "epoch": 469, "lr": 9.438532836976284e-05}
{"train_loss": 0.2675369679927826, "global_step": 41765, "epoch": 469, "lr": 9.438506145044992e-05}
{"train_loss": 0.3759814202785492, "global_step": 41766, "epoch": 469, "lr": 9.438479452516995e-05}
{"train_loss": 0.2743043601512909, "global_step": 41767, "epoch": 469, "lr": 9.438452759392299e-05}
{"train_loss": 0.31270551681518555, "global_step": 41768, "epoch": 469, "lr": 9.438426065670906e-05}
{"train_loss": 0.36445942521095276, "global_step": 41769, "epoch": 469, "lr": 9.438399371352823e-05}
{"train_loss": 0.34211692214012146, "global_step": 41770, "epoch": 469, "lr": 9.43837267643805e-05}
{"train_loss": 0.26707732677459717, "global_step": 41771, "epoch": 469, "lr": 9.438345980926592e-05}
{"train_loss": 0.3093982934951782, "global_step": 41772, "epoch": 469, "lr": 9.438319284818452e-05}
{"train_loss": 0.3109334409236908, "global_step": 41773, "epoch": 469, "lr": 9.438292588113635e-05}
{"train_loss": 0.3190303444862366, "global_step": 41774, "epoch": 469, "lr": 9.438265890812143e-05}
{"train_loss": 0.26306530833244324, "global_step": 41775, "epoch": 469, "lr": 9.43823919291398e-05}
{"train_loss": 0.2941785156726837, "global_step": 41776, "epoch": 469, "lr": 9.438212494419151e-05}
{"train_loss": 0.3014310300350189, "global_step": 41777, "epoch": 469, "lr": 9.438185795327658e-05}
{"train_loss": 0.42568835616111755, "global_step": 41778, "epoch": 469, "lr": 9.438159095639506e-05}
{"train_loss": 0.29321590065956116, "global_step": 41779, "epoch": 469, "lr": 9.438132395354695e-05}
{"train_loss": 0.32207024097442627, "global_step": 41780, "epoch": 469, "lr": 9.438105694473231e-05}
{"train_loss": 0.3941788375377655, "global_step": 41781, "epoch": 469, "lr": 9.43807899299512e-05}
{"train_loss": 0.24007461965084076, "global_step": 41782, "epoch": 469, "lr": 9.438052290920362e-05}
{"train_loss": 0.3039587438106537, "global_step": 41783, "epoch": 469, "lr": 9.438025588248964e-05}
{"train_loss": 0.32749950885772705, "global_step": 41784, "epoch": 469, "lr": 9.437998884980926e-05}
{"train_loss": 0.263200581073761, "global_step": 41785, "epoch": 469, "lr": 9.437972181116253e-05}
{"train_loss": 0.4654083251953125, "global_step": 41786, "epoch": 469, "lr": 9.43794547665495e-05}
{"train_loss": 0.33262068033218384, "global_step": 41787, "epoch": 469, "lr": 9.437918771597018e-05}
{"train_loss": 0.4457145929336548, "global_step": 41788, "epoch": 469, "lr": 9.437892065942462e-05}
{"train_loss": 0.30919787287712097, "global_step": 41789, "epoch": 469, "lr": 9.437865359691286e-05}
{"train_loss": 0.4925735890865326, "global_step": 41790, "epoch": 469, "lr": 9.437838652843493e-05}
{"train_loss": 0.37447190284729004, "global_step": 41791, "epoch": 469, "lr": 9.437811945399087e-05}
{"train_loss": 0.3079856336116791, "global_step": 41792, "epoch": 469, "lr": 9.43778523735807e-05}
{"train_loss": 0.30151233077049255, "global_step": 41793, "epoch": 469, "lr": 9.437758528720448e-05}
{"train_loss": 0.3811092972755432, "global_step": 41794, "epoch": 469, "lr": 9.437731819486224e-05}
{"train_loss": 0.4841752052307129, "global_step": 41795, "epoch": 469, "lr": 9.4377051096554e-05}
{"train_loss": 0.3533477187156677, "global_step": 41796, "epoch": 469, "lr": 9.43767839922798e-05}
{"train_loss": 0.43731188774108887, "global_step": 41797, "epoch": 469, "lr": 9.43765168820397e-05}
{"train_loss": 0.36297744512557983, "global_step": 41798, "epoch": 469, "lr": 9.43762497658337e-05}
{"train_loss": 0.32145801186561584, "global_step": 41799, "epoch": 469, "lr": 9.437598264366186e-05}
{"train_loss": 0.3341101408004761, "global_step": 41800, "epoch": 469, "lr": 9.437571551552422e-05}
{"train_loss": 0.36918044090270996, "global_step": 41801, "epoch": 469, "lr": 9.437544838142081e-05}
{"train_loss": 0.3802039921283722, "global_step": 41802, "epoch": 469, "lr": 9.437518124135164e-05}
{"train_loss": 0.3205163776874542, "global_step": 41803, "epoch": 469, "lr": 9.437491409531678e-05}
{"train_loss": 0.31224504113197327, "global_step": 41804, "epoch": 469, "lr": 9.437464694331626e-05}
{"train_loss": 0.30614200234413147, "global_step": 41805, "epoch": 469, "lr": 9.437437978535009e-05}
{"train_loss": 0.356056809425354, "global_step": 41806, "epoch": 469, "lr": 9.437411262141834e-05}
{"train_loss": 0.2369980365037918, "global_step": 41807, "epoch": 469, "lr": 9.437384545152102e-05}
{"train_loss": 0.3341146409511566, "global_step": 41808, "epoch": 469, "lr": 9.437357827565819e-05}
{"train_loss": 0.3571688234806061, "global_step": 41809, "epoch": 469, "lr": 9.437331109382988e-05}
{"train_loss": 0.3665403425693512, "global_step": 41810, "epoch": 469, "lr": 9.43730439060361e-05}
{"train_loss": 0.35403329133987427, "global_step": 41811, "epoch": 469, "lr": 9.437277671227692e-05}
{"train_loss": 0.2742699384689331, "global_step": 41812, "epoch": 469, "lr": 9.437250951255234e-05}
{"train_loss": 0.31718751788139343, "global_step": 41813, "epoch": 469, "lr": 9.437224230686244e-05}
{"train_loss": 0.25176021456718445, "global_step": 41814, "epoch": 469, "lr": 9.437197509520724e-05}
{"train_loss": 0.2590278089046478, "global_step": 41815, "epoch": 469, "lr": 9.437170787758673e-05}
{"train_loss": 0.3209792673587799, "global_step": 41816, "epoch": 469, "lr": 9.437144065400101e-05}
{"train_loss": 0.5192986130714417, "global_step": 41817, "epoch": 469, "lr": 9.43711734244501e-05}
{"train_loss": 0.4778588116168976, "global_step": 41818, "epoch": 469, "lr": 9.437090618893401e-05}
{"train_loss": 0.3090141713619232, "global_step": 41819, "epoch": 469, "lr": 9.43706389474528e-05}
{"train_loss": 0.25930672883987427, "global_step": 41820, "epoch": 469, "lr": 9.437037170000649e-05}
{"train_loss": 0.4100054204463959, "global_step": 41821, "epoch": 469, "lr": 9.437010444659514e-05}
{"train_loss": 0.2882347106933594, "global_step": 41822, "epoch": 469, "lr": 9.436983718721875e-05}
{"train_loss": 0.4136715531349182, "global_step": 41823, "epoch": 469, "lr": 9.43695699218774e-05}
{"train_loss": 0.37813693284988403, "global_step": 41824, "epoch": 469, "lr": 9.436930265057109e-05}
{"train_loss": 0.4496001899242401, "global_step": 41825, "epoch": 469, "lr": 9.436903537329986e-05}
{"train_loss": 0.4165433645248413, "global_step": 41826, "epoch": 469, "lr": 9.436876809006376e-05}
{"train_loss": 0.29928287863731384, "global_step": 41827, "epoch": 469, "lr": 9.436850080086282e-05}
{"train_loss": 0.35814929008483887, "global_step": 41828, "epoch": 469, "lr": 9.436823350569708e-05}
{"train_loss": 0.3350447751163097, "global_step": 41829, "epoch": 469, "lr": 9.436796620456656e-05, "val_loss": 2.5983080863952637}
{"train_loss": 0.25751447677612305, "global_step": 41830, "epoch": 470, "lr": 9.436769889747132e-05}
{"train_loss": 0.37046507000923157, "global_step": 41831, "epoch": 470, "lr": 9.436743158441138e-05}
{"train_loss": 0.31767189502716064, "global_step": 41832, "epoch": 470, "lr": 9.436716426538677e-05}
{"train_loss": 0.3599843680858612, "global_step": 41833, "epoch": 470, "lr": 9.436689694039755e-05}
{"train_loss": 0.3461982011795044, "global_step": 41834, "epoch": 470, "lr": 9.436662960944373e-05}
{"train_loss": 0.3298546075820923, "global_step": 41835, "epoch": 470, "lr": 9.436636227252537e-05}
{"train_loss": 0.3102688789367676, "global_step": 41836, "epoch": 470, "lr": 9.436609492964249e-05}
{"train_loss": 0.3143504559993744, "global_step": 41837, "epoch": 470, "lr": 9.436582758079513e-05}
{"train_loss": 0.27119743824005127, "global_step": 41838, "epoch": 470, "lr": 9.436556022598332e-05}
{"train_loss": 0.3231958746910095, "global_step": 41839, "epoch": 470, "lr": 9.43652928652071e-05}
{"train_loss": 0.36869651079177856, "global_step": 41840, "epoch": 470, "lr": 9.43650254984665e-05}
{"train_loss": 0.2584625482559204, "global_step": 41841, "epoch": 470, "lr": 9.436475812576157e-05}
{"train_loss": 0.2799678444862366, "global_step": 41842, "epoch": 470, "lr": 9.436449074709234e-05}
{"train_loss": 0.29628610610961914, "global_step": 41843, "epoch": 470, "lr": 9.436422336245885e-05}
{"train_loss": 0.3801131248474121, "global_step": 41844, "epoch": 470, "lr": 9.436395597186114e-05}
{"train_loss": 0.3594586253166199, "global_step": 41845, "epoch": 470, "lr": 9.436368857529923e-05}
{"train_loss": 0.3276207149028778, "global_step": 41846, "epoch": 470, "lr": 9.436342117277315e-05}
{"train_loss": 0.33064600825309753, "global_step": 41847, "epoch": 470, "lr": 9.436315376428295e-05}
{"train_loss": 0.30406999588012695, "global_step": 41848, "epoch": 470, "lr": 9.436288634982869e-05}
{"train_loss": 0.2512284815311432, "global_step": 41849, "epoch": 470, "lr": 9.436261892941034e-05}
{"train_loss": 0.4368678331375122, "global_step": 41850, "epoch": 470, "lr": 9.436235150302801e-05}
{"train_loss": 0.2721225321292877, "global_step": 41851, "epoch": 470, "lr": 9.436208407068168e-05}
{"train_loss": 0.4091491401195526, "global_step": 41852, "epoch": 470, "lr": 9.436181663237143e-05}
{"train_loss": 0.286050945520401, "global_step": 41853, "epoch": 470, "lr": 9.436154918809725e-05}
{"train_loss": 0.2528790831565857, "global_step": 41854, "epoch": 470, "lr": 9.436128173785923e-05}
{"train_loss": 0.23646773397922516, "global_step": 41855, "epoch": 470, "lr": 9.436101428165736e-05}
{"train_loss": 0.33968862891197205, "global_step": 41856, "epoch": 470, "lr": 9.436074681949169e-05}
{"train_loss": 0.36542439460754395, "global_step": 41857, "epoch": 470, "lr": 9.436047935136226e-05}
{"train_loss": 0.3305702209472656, "global_step": 41858, "epoch": 470, "lr": 9.43602118772691e-05}
{"train_loss": 0.3329819440841675, "global_step": 41859, "epoch": 470, "lr": 9.435994439721226e-05}
{"train_loss": 0.40599310398101807, "global_step": 41860, "epoch": 470, "lr": 9.435967691119177e-05}
{"train_loss": 0.3107500374317169, "global_step": 41861, "epoch": 470, "lr": 9.435940941920764e-05}
{"train_loss": 0.2700903117656708, "global_step": 41862, "epoch": 470, "lr": 9.435914192125994e-05}
{"train_loss": 0.3099418580532074, "global_step": 41863, "epoch": 470, "lr": 9.43588744173487e-05}
{"train_loss": 0.32895028591156006, "global_step": 41864, "epoch": 470, "lr": 9.435860690747394e-05}
{"train_loss": 0.3951604664325714, "global_step": 41865, "epoch": 470, "lr": 9.435833939163572e-05}
{"train_loss": 0.3614811897277832, "global_step": 41866, "epoch": 470, "lr": 9.435807186983405e-05}
{"train_loss": 0.39600130915641785, "global_step": 41867, "epoch": 470, "lr": 9.435780434206897e-05}
{"train_loss": 0.40493690967559814, "global_step": 41868, "epoch": 470, "lr": 9.435753680834054e-05}
{"train_loss": 0.27709537744522095, "global_step": 41869, "epoch": 470, "lr": 9.435726926864878e-05}
{"train_loss": 0.2794419229030609, "global_step": 41870, "epoch": 470, "lr": 9.435700172299371e-05}
{"train_loss": 0.3432454764842987, "global_step": 41871, "epoch": 470, "lr": 9.43567341713754e-05}
{"train_loss": 0.36581113934516907, "global_step": 41872, "epoch": 470, "lr": 9.435646661379386e-05}
{"train_loss": 0.2885481119155884, "global_step": 41873, "epoch": 470, "lr": 9.435619905024913e-05}
{"train_loss": 0.31698092818260193, "global_step": 41874, "epoch": 470, "lr": 9.435593148074127e-05}
{"train_loss": 0.2582723796367645, "global_step": 41875, "epoch": 470, "lr": 9.435566390527026e-05}
{"train_loss": 0.4107103645801544, "global_step": 41876, "epoch": 470, "lr": 9.43553963238362e-05}
{"train_loss": 0.34501999616622925, "global_step": 41877, "epoch": 470, "lr": 9.435512873643908e-05}
{"train_loss": 0.3152295649051666, "global_step": 41878, "epoch": 470, "lr": 9.435486114307896e-05}
{"train_loss": 0.27974167466163635, "global_step": 41879, "epoch": 470, "lr": 9.435459354375589e-05}
{"train_loss": 0.3594166040420532, "global_step": 41880, "epoch": 470, "lr": 9.435432593846986e-05}
{"train_loss": 0.28790509700775146, "global_step": 41881, "epoch": 470, "lr": 9.435405832722093e-05}
{"train_loss": 0.34552979469299316, "global_step": 41882, "epoch": 470, "lr": 9.435379071000914e-05}
{"train_loss": 0.3299521207809448, "global_step": 41883, "epoch": 470, "lr": 9.435352308683454e-05}
{"train_loss": 0.2807951271533966, "global_step": 41884, "epoch": 470, "lr": 9.435325545769713e-05}
{"train_loss": 0.25764840841293335, "global_step": 41885, "epoch": 470, "lr": 9.435298782259699e-05}
{"train_loss": 0.31865018606185913, "global_step": 41886, "epoch": 470, "lr": 9.435272018153411e-05}
{"train_loss": 0.2521136403083801, "global_step": 41887, "epoch": 470, "lr": 9.435245253450854e-05}
{"train_loss": 0.2515213191509247, "global_step": 41888, "epoch": 470, "lr": 9.435218488152035e-05}
{"train_loss": 0.291359543800354, "global_step": 41889, "epoch": 470, "lr": 9.435191722256954e-05}
{"train_loss": 0.31693121790885925, "global_step": 41890, "epoch": 470, "lr": 9.435164955765614e-05}
{"train_loss": 0.29197460412979126, "global_step": 41891, "epoch": 470, "lr": 9.43513818867802e-05}
{"train_loss": 0.28792819380760193, "global_step": 41892, "epoch": 470, "lr": 9.435111420994179e-05}
{"train_loss": 0.29963043332099915, "global_step": 41893, "epoch": 470, "lr": 9.43508465271409e-05}
{"train_loss": 0.3937501609325409, "global_step": 41894, "epoch": 470, "lr": 9.435057883837757e-05}
{"train_loss": 0.3503994345664978, "global_step": 41895, "epoch": 470, "lr": 9.435031114365183e-05}
{"train_loss": 0.2744899094104767, "global_step": 41896, "epoch": 470, "lr": 9.435004344296375e-05}
{"train_loss": 0.2882547080516815, "global_step": 41897, "epoch": 470, "lr": 9.434977573631335e-05}
{"train_loss": 0.3025900721549988, "global_step": 41898, "epoch": 470, "lr": 9.434950802370066e-05}
{"train_loss": 0.21026960015296936, "global_step": 41899, "epoch": 470, "lr": 9.434924030512572e-05}
{"train_loss": 0.20282867550849915, "global_step": 41900, "epoch": 470, "lr": 9.434897258058857e-05}
{"train_loss": 0.26145121455192566, "global_step": 41901, "epoch": 470, "lr": 9.434870485008922e-05}
{"train_loss": 0.2634449601173401, "global_step": 41902, "epoch": 470, "lr": 9.434843711362775e-05}
{"train_loss": 0.27356117963790894, "global_step": 41903, "epoch": 470, "lr": 9.434816937120417e-05}
{"train_loss": 0.29373493790626526, "global_step": 41904, "epoch": 470, "lr": 9.43479016228185e-05}
{"train_loss": 0.2713281214237213, "global_step": 41905, "epoch": 470, "lr": 9.434763386847081e-05}
{"train_loss": 0.2880627512931824, "global_step": 41906, "epoch": 470, "lr": 9.434736610816113e-05}
{"train_loss": 0.3293663561344147, "global_step": 41907, "epoch": 470, "lr": 9.434709834188948e-05}
{"train_loss": 0.3557162284851074, "global_step": 41908, "epoch": 470, "lr": 9.434683056965589e-05}
{"train_loss": 0.3049248158931732, "global_step": 41909, "epoch": 470, "lr": 9.434656279146041e-05}
{"train_loss": 0.3350880444049835, "global_step": 41910, "epoch": 470, "lr": 9.43462950073031e-05}
{"train_loss": 0.29059094190597534, "global_step": 41911, "epoch": 470, "lr": 9.434602721718396e-05}
{"train_loss": 0.30357426404953003, "global_step": 41912, "epoch": 470, "lr": 9.434575942110302e-05}
{"train_loss": 0.2859501540660858, "global_step": 41913, "epoch": 470, "lr": 9.434549161906034e-05}
{"train_loss": 0.17485865950584412, "global_step": 41914, "epoch": 470, "lr": 9.434522381105597e-05}
{"train_loss": 0.34061864018440247, "global_step": 41915, "epoch": 470, "lr": 9.43449559970899e-05}
{"train_loss": 0.305423766374588, "global_step": 41916, "epoch": 470, "lr": 9.43446881771622e-05}
{"train_loss": 0.30666282773017883, "global_step": 41917, "epoch": 470, "lr": 9.434442035127292e-05}
{"train_loss": 0.3125810166088383, "global_step": 41918, "epoch": 470, "lr": 9.434415251942204e-05, "val_loss": 2.7873241901397705, "train_action_mse_error": 17.927833557128906}
{"train_loss": 0.29085853695869446, "global_step": 41919, "epoch": 471, "lr": 9.434388468160964e-05}
{"train_loss": 0.2411024123430252, "global_step": 41920, "epoch": 471, "lr": 9.434361683783575e-05}
{"train_loss": 0.2548311948776245, "global_step": 41921, "epoch": 471, "lr": 9.43433489881004e-05}
{"train_loss": 0.28053757548332214, "global_step": 41922, "epoch": 471, "lr": 9.434308113240363e-05}
{"train_loss": 0.22659625113010406, "global_step": 41923, "epoch": 471, "lr": 9.434281327074545e-05}
{"train_loss": 0.2287876456975937, "global_step": 41924, "epoch": 471, "lr": 9.434254540312595e-05}
{"train_loss": 0.3202593922615051, "global_step": 41925, "epoch": 471, "lr": 9.434227752954512e-05}
{"train_loss": 0.30359044671058655, "global_step": 41926, "epoch": 471, "lr": 9.434200965000302e-05}
{"train_loss": 0.2955043613910675, "global_step": 41927, "epoch": 471, "lr": 9.434174176449966e-05}
{"train_loss": 0.26481008529663086, "global_step": 41928, "epoch": 471, "lr": 9.43414738730351e-05}
{"train_loss": 0.2587440609931946, "global_step": 41929, "epoch": 471, "lr": 9.434120597560938e-05}
{"train_loss": 0.2637612819671631, "global_step": 41930, "epoch": 471, "lr": 9.434093807222252e-05}
{"train_loss": 0.2879018783569336, "global_step": 41931, "epoch": 471, "lr": 9.434067016287456e-05}
{"train_loss": 0.31063780188560486, "global_step": 41932, "epoch": 471, "lr": 9.434040224756553e-05}
{"train_loss": 0.2901085317134857, "global_step": 41933, "epoch": 471, "lr": 9.434013432629548e-05}
{"train_loss": 0.22367697954177856, "global_step": 41934, "epoch": 471, "lr": 9.433986639906445e-05}
{"train_loss": 0.3756045997142792, "global_step": 41935, "epoch": 471, "lr": 9.433959846587245e-05}
{"train_loss": 0.2883448898792267, "global_step": 41936, "epoch": 471, "lr": 9.433933052671953e-05}
{"train_loss": 0.2752853333950043, "global_step": 41937, "epoch": 471, "lr": 9.433906258160574e-05}
{"train_loss": 0.3599337935447693, "global_step": 41938, "epoch": 471, "lr": 9.43387946305311e-05}
{"train_loss": 0.2742423117160797, "global_step": 41939, "epoch": 471, "lr": 9.433852667349566e-05}
{"train_loss": 0.23619423806667328, "global_step": 41940, "epoch": 471, "lr": 9.433825871049941e-05}
{"train_loss": 0.1490887999534607, "global_step": 41941, "epoch": 471, "lr": 9.433799074154246e-05}
{"train_loss": 0.24597309529781342, "global_step": 41942, "epoch": 471, "lr": 9.433772276662478e-05}
{"train_loss": 0.2619267404079437, "global_step": 41943, "epoch": 471, "lr": 9.433745478574644e-05}
{"train_loss": 0.3014986515045166, "global_step": 41944, "epoch": 471, "lr": 9.433718679890748e-05}
{"train_loss": 0.2701033055782318, "global_step": 41945, "epoch": 471, "lr": 9.433691880610792e-05}
{"train_loss": 0.24241286516189575, "global_step": 41946, "epoch": 471, "lr": 9.43366508073478e-05}
{"train_loss": 0.37705928087234497, "global_step": 41947, "epoch": 471, "lr": 9.433638280262716e-05}
{"train_loss": 0.2147900015115738, "global_step": 41948, "epoch": 471, "lr": 9.433611479194604e-05}
{"train_loss": 0.20886608958244324, "global_step": 41949, "epoch": 471, "lr": 9.433584677530447e-05}
{"train_loss": 0.32138657569885254, "global_step": 41950, "epoch": 471, "lr": 9.433557875270247e-05}
{"train_loss": 0.2678581774234772, "global_step": 41951, "epoch": 471, "lr": 9.43353107241401e-05}
{"train_loss": 0.27970343828201294, "global_step": 41952, "epoch": 471, "lr": 9.43350426896174e-05}
{"train_loss": 0.20542922616004944, "global_step": 41953, "epoch": 471, "lr": 9.433477464913437e-05}
{"train_loss": 0.17611390352249146, "global_step": 41954, "epoch": 471, "lr": 9.433450660269108e-05}
{"train_loss": 0.27937427163124084, "global_step": 41955, "epoch": 471, "lr": 9.433423855028755e-05}
{"train_loss": 0.2118825912475586, "global_step": 41956, "epoch": 471, "lr": 9.433397049192385e-05}
{"train_loss": 0.2602528929710388, "global_step": 41957, "epoch": 471, "lr": 9.433370242759995e-05}
{"train_loss": 0.21954190731048584, "global_step": 41958, "epoch": 471, "lr": 9.433343435731596e-05}
{"train_loss": 0.43493857979774475, "global_step": 41959, "epoch": 471, "lr": 9.433316628107186e-05}
{"train_loss": 0.2973462641239166, "global_step": 41960, "epoch": 471, "lr": 9.43328981988677e-05}
{"train_loss": 0.30303025245666504, "global_step": 41961, "epoch": 471, "lr": 9.433263011070354e-05}
{"train_loss": 0.21571412682533264, "global_step": 41962, "epoch": 471, "lr": 9.433236201657938e-05}
{"train_loss": 0.28119081258773804, "global_step": 41963, "epoch": 471, "lr": 9.433209391649529e-05}
{"train_loss": 0.37087202072143555, "global_step": 41964, "epoch": 471, "lr": 9.433182581045127e-05}
{"train_loss": 0.24609161913394928, "global_step": 41965, "epoch": 471, "lr": 9.43315576984474e-05}
{"train_loss": 0.26405447721481323, "global_step": 41966, "epoch": 471, "lr": 9.43312895804837e-05}
{"train_loss": 0.22500824928283691, "global_step": 41967, "epoch": 471, "lr": 9.433102145656016e-05}
{"train_loss": 0.26341232657432556, "global_step": 41968, "epoch": 471, "lr": 9.433075332667689e-05}
{"train_loss": 0.2873344421386719, "global_step": 41969, "epoch": 471, "lr": 9.433048519083387e-05}
{"train_loss": 0.25635138154029846, "global_step": 41970, "epoch": 471, "lr": 9.433021704903116e-05}
{"train_loss": 0.28029319643974304, "global_step": 41971, "epoch": 471, "lr": 9.432994890126881e-05}
{"train_loss": 0.3011677861213684, "global_step": 41972, "epoch": 471, "lr": 9.432968074754682e-05}
{"train_loss": 0.21530264616012573, "global_step": 41973, "epoch": 471, "lr": 9.432941258786524e-05}
{"train_loss": 0.3388848602771759, "global_step": 41974, "epoch": 471, "lr": 9.432914442222411e-05}
{"train_loss": 0.3755255341529846, "global_step": 41975, "epoch": 471, "lr": 9.432887625062349e-05}
{"train_loss": 0.2907046377658844, "global_step": 41976, "epoch": 471, "lr": 9.432860807306337e-05}
{"train_loss": 0.3122779428958893, "global_step": 41977, "epoch": 471, "lr": 9.432833988954383e-05}
{"train_loss": 0.3832343816757202, "global_step": 41978, "epoch": 471, "lr": 9.432807170006487e-05}
{"train_loss": 0.2935240566730499, "global_step": 41979, "epoch": 471, "lr": 9.432780350462653e-05}
{"train_loss": 0.24672512710094452, "global_step": 41980, "epoch": 471, "lr": 9.432753530322889e-05}
{"train_loss": 0.2603357434272766, "global_step": 41981, "epoch": 471, "lr": 9.432726709587193e-05}
{"train_loss": 0.3208440840244293, "global_step": 41982, "epoch": 471, "lr": 9.432699888255572e-05}
{"train_loss": 0.29172998666763306, "global_step": 41983, "epoch": 471, "lr": 9.432673066328028e-05}
{"train_loss": 0.32172828912734985, "global_step": 41984, "epoch": 471, "lr": 9.432646243804564e-05}
{"train_loss": 0.2538016140460968, "global_step": 41985, "epoch": 471, "lr": 9.432619420685186e-05}
{"train_loss": 0.40058234333992004, "global_step": 41986, "epoch": 471, "lr": 9.432592596969896e-05}
{"train_loss": 0.30148884654045105, "global_step": 41987, "epoch": 471, "lr": 9.432565772658698e-05}
{"train_loss": 0.25776153802871704, "global_step": 41988, "epoch": 471, "lr": 9.432538947751594e-05}
{"train_loss": 0.2648060619831085, "global_step": 41989, "epoch": 471, "lr": 9.432512122248591e-05}
{"train_loss": 0.26655128598213196, "global_step": 41990, "epoch": 471, "lr": 9.432485296149691e-05}
{"train_loss": 0.3486374318599701, "global_step": 41991, "epoch": 471, "lr": 9.432458469454898e-05}
{"train_loss": 0.31010398268699646, "global_step": 41992, "epoch": 471, "lr": 9.432431642164213e-05}
{"train_loss": 0.17302696406841278, "global_step": 41993, "epoch": 471, "lr": 9.432404814277643e-05}
{"train_loss": 0.2975747883319855, "global_step": 41994, "epoch": 471, "lr": 9.43237798579519e-05}
{"train_loss": 0.280986487865448, "global_step": 41995, "epoch": 471, "lr": 9.432351156716856e-05}
{"train_loss": 0.285247802734375, "global_step": 41996, "epoch": 471, "lr": 9.432324327042649e-05}
{"train_loss": 0.32391002774238586, "global_step": 41997, "epoch": 471, "lr": 9.432297496772569e-05}
{"train_loss": 0.21539273858070374, "global_step": 41998, "epoch": 471, "lr": 9.43227066590662e-05}
{"train_loss": 0.45452916622161865, "global_step": 41999, "epoch": 471, "lr": 9.432243834444807e-05}
{"train_loss": 0.28725871443748474, "global_step": 42000, "epoch": 471, "lr": 9.432217002387132e-05}
{"train_loss": 0.2980578541755676, "global_step": 42001, "epoch": 471, "lr": 9.432190169733601e-05}
{"train_loss": 0.4172784686088562, "global_step": 42002, "epoch": 471, "lr": 9.432163336484215e-05}
{"train_loss": 0.2989733815193176, "global_step": 42003, "epoch": 471, "lr": 9.432136502638979e-05}
{"train_loss": 0.35543176531791687, "global_step": 42004, "epoch": 471, "lr": 9.432109668197896e-05}
{"train_loss": 0.2956503629684448, "global_step": 42005, "epoch": 471, "lr": 9.43208283316097e-05}
{"train_loss": 0.2689129114151001, "global_step": 42006, "epoch": 471, "lr": 9.432055997528204e-05}
{"train_loss": 0.28398387803790276, "global_step": 42007, "epoch": 471, "lr": 9.432029161299603e-05, "val_loss": 2.835275650024414}
{"train_loss": 0.3091801404953003, "global_step": 42008, "epoch": 472, "lr": 9.432002324475168e-05}
{"train_loss": 0.20543250441551208, "global_step": 42009, "epoch": 472, "lr": 9.431975487054906e-05}
{"train_loss": 0.3015161454677582, "global_step": 42010, "epoch": 472, "lr": 9.431948649038819e-05}
{"train_loss": 0.4105149209499359, "global_step": 42011, "epoch": 472, "lr": 9.431921810426909e-05}
{"train_loss": 0.33549964427948, "global_step": 42012, "epoch": 472, "lr": 9.431894971219182e-05}
{"train_loss": 0.28950685262680054, "global_step": 42013, "epoch": 472, "lr": 9.431868131415642e-05}
{"train_loss": 0.24108557403087616, "global_step": 42014, "epoch": 472, "lr": 9.43184129101629e-05}
{"train_loss": 0.27003014087677, "global_step": 42015, "epoch": 472, "lr": 9.43181445002113e-05}
{"train_loss": 0.4343079626560211, "global_step": 42016, "epoch": 472, "lr": 9.431787608430169e-05}
{"train_loss": 0.2897240221500397, "global_step": 42017, "epoch": 472, "lr": 9.431760766243406e-05}
{"train_loss": 0.32464680075645447, "global_step": 42018, "epoch": 472, "lr": 9.431733923460849e-05}
{"train_loss": 0.3230248987674713, "global_step": 42019, "epoch": 472, "lr": 9.431707080082497e-05}
{"train_loss": 0.26023274660110474, "global_step": 42020, "epoch": 472, "lr": 9.431680236108356e-05}
{"train_loss": 0.29926714301109314, "global_step": 42021, "epoch": 472, "lr": 9.431653391538431e-05}
{"train_loss": 0.34473758935928345, "global_step": 42022, "epoch": 472, "lr": 9.431626546372725e-05}
{"train_loss": 0.2995467782020569, "global_step": 42023, "epoch": 472, "lr": 9.431599700611239e-05}
{"train_loss": 0.44280365109443665, "global_step": 42024, "epoch": 472, "lr": 9.431572854253978e-05}
{"train_loss": 0.31627213954925537, "global_step": 42025, "epoch": 472, "lr": 9.431546007300948e-05}
{"train_loss": 0.3422682583332062, "global_step": 42026, "epoch": 472, "lr": 9.431519159752149e-05}
{"train_loss": 0.3739077150821686, "global_step": 42027, "epoch": 472, "lr": 9.431492311607588e-05}
{"train_loss": 0.31009599566459656, "global_step": 42028, "epoch": 472, "lr": 9.431465462867266e-05}
{"train_loss": 0.32937130331993103, "global_step": 42029, "epoch": 472, "lr": 9.431438613531187e-05}
{"train_loss": 0.430000901222229, "global_step": 42030, "epoch": 472, "lr": 9.431411763599356e-05}
{"train_loss": 0.3165811598300934, "global_step": 42031, "epoch": 472, "lr": 9.431384913071774e-05}
{"train_loss": 0.2773396968841553, "global_step": 42032, "epoch": 472, "lr": 9.431358061948447e-05}
{"train_loss": 0.2252615988254547, "global_step": 42033, "epoch": 472, "lr": 9.431331210229379e-05}
{"train_loss": 0.31686633825302124, "global_step": 42034, "epoch": 472, "lr": 9.431304357914572e-05}
{"train_loss": 0.3305111229419708, "global_step": 42035, "epoch": 472, "lr": 9.43127750500403e-05}
{"train_loss": 0.31889545917510986, "global_step": 42036, "epoch": 472, "lr": 9.431250651497757e-05}
{"train_loss": 0.33786696195602417, "global_step": 42037, "epoch": 472, "lr": 9.431223797395756e-05}
{"train_loss": 0.26732710003852844, "global_step": 42038, "epoch": 472, "lr": 9.431196942698031e-05}
{"train_loss": 0.29309093952178955, "global_step": 42039, "epoch": 472, "lr": 9.431170087404587e-05}
{"train_loss": 0.3068459928035736, "global_step": 42040, "epoch": 472, "lr": 9.431143231515424e-05}
{"train_loss": 0.3033764958381653, "global_step": 42041, "epoch": 472, "lr": 9.431116375030548e-05}
{"train_loss": 0.2843395173549652, "global_step": 42042, "epoch": 472, "lr": 9.431089517949964e-05}
{"train_loss": 0.2561587393283844, "global_step": 42043, "epoch": 472, "lr": 9.431062660273674e-05}
{"train_loss": 0.2270570546388626, "global_step": 42044, "epoch": 472, "lr": 9.43103580200168e-05}
{"train_loss": 0.2843058109283447, "global_step": 42045, "epoch": 472, "lr": 9.431008943133988e-05}
{"train_loss": 0.29648512601852417, "global_step": 42046, "epoch": 472, "lr": 9.430982083670601e-05}
{"train_loss": 0.25915637612342834, "global_step": 42047, "epoch": 472, "lr": 9.430955223611522e-05}
{"train_loss": 0.3024353086948395, "global_step": 42048, "epoch": 472, "lr": 9.430928362956755e-05}
{"train_loss": 0.29423049092292786, "global_step": 42049, "epoch": 472, "lr": 9.430901501706305e-05}
{"train_loss": 0.2385648638010025, "global_step": 42050, "epoch": 472, "lr": 9.430874639860173e-05}
{"train_loss": 0.30212271213531494, "global_step": 42051, "epoch": 472, "lr": 9.430847777418365e-05}
{"train_loss": 0.24618907272815704, "global_step": 42052, "epoch": 472, "lr": 9.430820914380882e-05}
{"train_loss": 0.3433048725128174, "global_step": 42053, "epoch": 472, "lr": 9.430794050747731e-05}
{"train_loss": 0.2809441387653351, "global_step": 42054, "epoch": 472, "lr": 9.430767186518911e-05}
{"train_loss": 0.33176711201667786, "global_step": 42055, "epoch": 472, "lr": 9.430740321694431e-05}
{"train_loss": 0.27927619218826294, "global_step": 42056, "epoch": 472, "lr": 9.430713456274291e-05}
{"train_loss": 0.3450988233089447, "global_step": 42057, "epoch": 472, "lr": 9.430686590258495e-05}
{"train_loss": 0.3072524070739746, "global_step": 42058, "epoch": 472, "lr": 9.430659723647049e-05}
{"train_loss": 0.31019309163093567, "global_step": 42059, "epoch": 472, "lr": 9.430632856439954e-05}
{"train_loss": 0.38566967844963074, "global_step": 42060, "epoch": 472, "lr": 9.430605988637213e-05}
{"train_loss": 0.2945237159729004, "global_step": 42061, "epoch": 472, "lr": 9.430579120238831e-05}
{"train_loss": 0.2882629334926605, "global_step": 42062, "epoch": 472, "lr": 9.430552251244814e-05}
{"train_loss": 0.3361401855945587, "global_step": 42063, "epoch": 472, "lr": 9.430525381655162e-05}
{"train_loss": 0.36414945125579834, "global_step": 42064, "epoch": 472, "lr": 9.43049851146988e-05}
{"train_loss": 0.3634108006954193, "global_step": 42065, "epoch": 472, "lr": 9.430471640688971e-05}
{"train_loss": 0.1930931657552719, "global_step": 42066, "epoch": 472, "lr": 9.430444769312439e-05}
{"train_loss": 0.3406347632408142, "global_step": 42067, "epoch": 472, "lr": 9.43041789734029e-05}
{"train_loss": 0.3175862729549408, "global_step": 42068, "epoch": 472, "lr": 9.430391024772523e-05}
{"train_loss": 0.35727658867836, "global_step": 42069, "epoch": 472, "lr": 9.430364151609144e-05}
{"train_loss": 0.3287252187728882, "global_step": 42070, "epoch": 472, "lr": 9.430337277850157e-05}
{"train_loss": 0.43187811970710754, "global_step": 42071, "epoch": 472, "lr": 9.430310403495564e-05}
{"train_loss": 0.30352702736854553, "global_step": 42072, "epoch": 472, "lr": 9.430283528545371e-05}
{"train_loss": 0.3293881118297577, "global_step": 42073, "epoch": 472, "lr": 9.43025665299958e-05}
{"train_loss": 0.23724742233753204, "global_step": 42074, "epoch": 472, "lr": 9.430229776858195e-05}
{"train_loss": 0.4046388268470764, "global_step": 42075, "epoch": 472, "lr": 9.43020290012122e-05}
{"train_loss": 0.45335790514945984, "global_step": 42076, "epoch": 472, "lr": 9.430176022788657e-05}
{"train_loss": 0.2440485656261444, "global_step": 42077, "epoch": 472, "lr": 9.430149144860512e-05}
{"train_loss": 0.24032852053642273, "global_step": 42078, "epoch": 472, "lr": 9.430122266336788e-05}
{"train_loss": 0.2890191376209259, "global_step": 42079, "epoch": 472, "lr": 9.430095387217487e-05}
{"train_loss": 0.3428318500518799, "global_step": 42080, "epoch": 472, "lr": 9.430068507502614e-05}
{"train_loss": 0.3951706886291504, "global_step": 42081, "epoch": 472, "lr": 9.430041627192172e-05}
{"train_loss": 0.211109921336174, "global_step": 42082, "epoch": 472, "lr": 9.430014746286165e-05}
{"train_loss": 0.3336670398712158, "global_step": 42083, "epoch": 472, "lr": 9.429987864784597e-05}
{"train_loss": 0.4010498821735382, "global_step": 42084, "epoch": 472, "lr": 9.429960982687471e-05}
{"train_loss": 0.2698178291320801, "global_step": 42085, "epoch": 472, "lr": 9.429934099994791e-05}
{"train_loss": 0.23420146107673645, "global_step": 42086, "epoch": 472, "lr": 9.429907216706559e-05}
{"train_loss": 0.20677651464939117, "global_step": 42087, "epoch": 472, "lr": 9.42988033282278e-05}
{"train_loss": 0.3719371557235718, "global_step": 42088, "epoch": 472, "lr": 9.42985344834346e-05}
{"train_loss": 0.22082850337028503, "global_step": 42089, "epoch": 472, "lr": 9.429826563268598e-05}
{"train_loss": 0.33825260400772095, "global_step": 42090, "epoch": 472, "lr": 9.4297996775982e-05}
{"train_loss": 0.3133600652217865, "global_step": 42091, "epoch": 472, "lr": 9.42977279133227e-05}
{"train_loss": 0.38868531584739685, "global_step": 42092, "epoch": 472, "lr": 9.42974590447081e-05}
{"train_loss": 0.3234461843967438, "global_step": 42093, "epoch": 472, "lr": 9.429719017013827e-05}
{"train_loss": 0.2994152903556824, "global_step": 42094, "epoch": 472, "lr": 9.429692128961321e-05}
{"train_loss": 0.29879826307296753, "global_step": 42095, "epoch": 472, "lr": 9.429665240313296e-05}
{"train_loss": 0.3108192151517011, "global_step": 42096, "epoch": 472, "lr": 9.429638351069757e-05, "val_loss": 2.7406084537506104}
{"train_loss": 0.34593304991722107, "global_step": 42097, "epoch": 473, "lr": 9.429611461230708e-05}
{"train_loss": 0.30640146136283875, "global_step": 42098, "epoch": 473, "lr": 9.429584570796151e-05}
{"train_loss": 0.30079731345176697, "global_step": 42099, "epoch": 473, "lr": 9.42955767976609e-05}
{"train_loss": 0.2948850095272064, "global_step": 42100, "epoch": 473, "lr": 9.42953078814053e-05}
{"train_loss": 0.3037589490413666, "global_step": 42101, "epoch": 473, "lr": 9.429503895919472e-05}
{"train_loss": 0.31940892338752747, "global_step": 42102, "epoch": 473, "lr": 9.429477003102922e-05}
{"train_loss": 0.31550922989845276, "global_step": 42103, "epoch": 473, "lr": 9.429450109690883e-05}
{"train_loss": 0.2642431855201721, "global_step": 42104, "epoch": 473, "lr": 9.429423215683358e-05}
{"train_loss": 0.21797633171081543, "global_step": 42105, "epoch": 473, "lr": 9.429396321080352e-05}
{"train_loss": 0.23877206444740295, "global_step": 42106, "epoch": 473, "lr": 9.429369425881866e-05}
{"train_loss": 0.3673611879348755, "global_step": 42107, "epoch": 473, "lr": 9.429342530087907e-05}
{"train_loss": 0.2359398454427719, "global_step": 42108, "epoch": 473, "lr": 9.429315633698477e-05}
{"train_loss": 0.29523730278015137, "global_step": 42109, "epoch": 473, "lr": 9.429288736713577e-05}
{"train_loss": 0.33944669365882874, "global_step": 42110, "epoch": 473, "lr": 9.429261839133216e-05}
{"train_loss": 0.37395307421684265, "global_step": 42111, "epoch": 473, "lr": 9.429234940957393e-05}
{"train_loss": 0.268912136554718, "global_step": 42112, "epoch": 473, "lr": 9.429208042186114e-05}
{"train_loss": 0.3249148726463318, "global_step": 42113, "epoch": 473, "lr": 9.429181142819381e-05}
{"train_loss": 0.35306480526924133, "global_step": 42114, "epoch": 473, "lr": 9.4291542428572e-05}
{"train_loss": 0.2855184078216553, "global_step": 42115, "epoch": 473, "lr": 9.429127342299573e-05}
{"train_loss": 0.29745766520500183, "global_step": 42116, "epoch": 473, "lr": 9.429100441146503e-05}
{"train_loss": 0.24987798929214478, "global_step": 42117, "epoch": 473, "lr": 9.429073539397994e-05}
{"train_loss": 0.3971115052700043, "global_step": 42118, "epoch": 473, "lr": 9.429046637054051e-05}
{"train_loss": 0.22482579946517944, "global_step": 42119, "epoch": 473, "lr": 9.429019734114677e-05}
{"train_loss": 0.2937726378440857, "global_step": 42120, "epoch": 473, "lr": 9.428992830579874e-05}
{"train_loss": 0.2971276640892029, "global_step": 42121, "epoch": 473, "lr": 9.428965926449649e-05}
{"train_loss": 0.26620596647262573, "global_step": 42122, "epoch": 473, "lr": 9.428939021724001e-05}
{"train_loss": 0.32334405183792114, "global_step": 42123, "epoch": 473, "lr": 9.428912116402938e-05}
{"train_loss": 0.3003760874271393, "global_step": 42124, "epoch": 473, "lr": 9.428885210486462e-05}
{"train_loss": 0.4165019690990448, "global_step": 42125, "epoch": 473, "lr": 9.428858303974574e-05}
{"train_loss": 0.3853769302368164, "global_step": 42126, "epoch": 473, "lr": 9.428831396867281e-05}
{"train_loss": 0.2971828281879425, "global_step": 42127, "epoch": 473, "lr": 9.428804489164586e-05}
{"train_loss": 0.28367695212364197, "global_step": 42128, "epoch": 473, "lr": 9.428777580866494e-05}
{"train_loss": 0.39763325452804565, "global_step": 42129, "epoch": 473, "lr": 9.428750671973005e-05}
{"train_loss": 0.2794881761074066, "global_step": 42130, "epoch": 473, "lr": 9.428723762484125e-05}
{"train_loss": 0.33699798583984375, "global_step": 42131, "epoch": 473, "lr": 9.428696852399856e-05}
{"train_loss": 0.3000917136669159, "global_step": 42132, "epoch": 473, "lr": 9.428669941720202e-05}
{"train_loss": 0.2919042110443115, "global_step": 42133, "epoch": 473, "lr": 9.42864303044517e-05}
{"train_loss": 0.276417076587677, "global_step": 42134, "epoch": 473, "lr": 9.42861611857476e-05}
{"train_loss": 0.38916468620300293, "global_step": 42135, "epoch": 473, "lr": 9.428589206108975e-05}
{"train_loss": 0.28196436166763306, "global_step": 42136, "epoch": 473, "lr": 9.42856229304782e-05}
{"train_loss": 0.29451072216033936, "global_step": 42137, "epoch": 473, "lr": 9.428535379391301e-05}
{"train_loss": 0.3696199357509613, "global_step": 42138, "epoch": 473, "lr": 9.428508465139418e-05}
{"train_loss": 0.2953921854496002, "global_step": 42139, "epoch": 473, "lr": 9.428481550292178e-05}
{"train_loss": 0.28075143694877625, "global_step": 42140, "epoch": 473, "lr": 9.42845463484958e-05}
{"train_loss": 0.3242241144180298, "global_step": 42141, "epoch": 473, "lr": 9.428427718811633e-05}
{"train_loss": 0.27752685546875, "global_step": 42142, "epoch": 473, "lr": 9.428400802178335e-05}
{"train_loss": 0.43461742997169495, "global_step": 42143, "epoch": 473, "lr": 9.428373884949693e-05}
{"train_loss": 0.24405069649219513, "global_step": 42144, "epoch": 473, "lr": 9.428346967125712e-05}
{"train_loss": 0.33124664425849915, "global_step": 42145, "epoch": 473, "lr": 9.428320048706391e-05}
{"train_loss": 0.32266438007354736, "global_step": 42146, "epoch": 473, "lr": 9.428293129691738e-05}
{"train_loss": 0.28900307416915894, "global_step": 42147, "epoch": 473, "lr": 9.428266210081755e-05}
{"train_loss": 0.32244089245796204, "global_step": 42148, "epoch": 473, "lr": 9.428239289876446e-05}
{"train_loss": 0.3114818036556244, "global_step": 42149, "epoch": 473, "lr": 9.428212369075813e-05}
{"train_loss": 0.24365860223770142, "global_step": 42150, "epoch": 473, "lr": 9.428185447679862e-05}
{"train_loss": 0.33144447207450867, "global_step": 42151, "epoch": 473, "lr": 9.428158525688595e-05}
{"train_loss": 0.28884127736091614, "global_step": 42152, "epoch": 473, "lr": 9.428131603102015e-05}
{"train_loss": 0.41939204931259155, "global_step": 42153, "epoch": 473, "lr": 9.428104679920127e-05}
{"train_loss": 0.3367045223712921, "global_step": 42154, "epoch": 473, "lr": 9.428077756142936e-05}
{"train_loss": 0.2846720516681671, "global_step": 42155, "epoch": 473, "lr": 9.428050831770443e-05}
{"train_loss": 0.3825978934764862, "global_step": 42156, "epoch": 473, "lr": 9.428023906802652e-05}
{"train_loss": 0.37141746282577515, "global_step": 42157, "epoch": 473, "lr": 9.427996981239567e-05}
{"train_loss": 0.23145265877246857, "global_step": 42158, "epoch": 473, "lr": 9.427970055081192e-05}
{"train_loss": 0.333473801612854, "global_step": 42159, "epoch": 473, "lr": 9.427943128327531e-05}
{"train_loss": 0.32382717728614807, "global_step": 42160, "epoch": 473, "lr": 9.427916200978586e-05}
{"train_loss": 0.29716187715530396, "global_step": 42161, "epoch": 473, "lr": 9.427889273034362e-05}
{"train_loss": 0.27934178709983826, "global_step": 42162, "epoch": 473, "lr": 9.427862344494863e-05}
{"train_loss": 0.23245127499103546, "global_step": 42163, "epoch": 473, "lr": 9.427835415360092e-05}
{"train_loss": 0.2882883846759796, "global_step": 42164, "epoch": 473, "lr": 9.427808485630052e-05}
{"train_loss": 0.3228047490119934, "global_step": 42165, "epoch": 473, "lr": 9.427781555304746e-05}
{"train_loss": 0.470070481300354, "global_step": 42166, "epoch": 473, "lr": 9.427754624384179e-05}
{"train_loss": 0.35765549540519714, "global_step": 42167, "epoch": 473, "lr": 9.427727692868356e-05}
{"train_loss": 0.2844978868961334, "global_step": 42168, "epoch": 473, "lr": 9.427700760757278e-05}
{"train_loss": 0.2966274917125702, "global_step": 42169, "epoch": 473, "lr": 9.42767382805095e-05}
{"train_loss": 0.21853113174438477, "global_step": 42170, "epoch": 473, "lr": 9.427646894749376e-05}
{"train_loss": 0.3008197546005249, "global_step": 42171, "epoch": 473, "lr": 9.427619960852557e-05}
{"train_loss": 0.26249638199806213, "global_step": 42172, "epoch": 473, "lr": 9.4275930263605e-05}
{"train_loss": 0.35891589522361755, "global_step": 42173, "epoch": 473, "lr": 9.427566091273205e-05}
{"train_loss": 0.31735363602638245, "global_step": 42174, "epoch": 473, "lr": 9.427539155590679e-05}
{"train_loss": 0.3936253488063812, "global_step": 42175, "epoch": 473, "lr": 9.427512219312925e-05}
{"train_loss": 0.2990821599960327, "global_step": 42176, "epoch": 473, "lr": 9.427485282439945e-05}
{"train_loss": 0.3711479604244232, "global_step": 42177, "epoch": 473, "lr": 9.427458344971745e-05}
{"train_loss": 0.3255999982357025, "global_step": 42178, "epoch": 473, "lr": 9.427431406908325e-05}
{"train_loss": 0.27362826466560364, "global_step": 42179, "epoch": 473, "lr": 9.427404468249692e-05}
{"train_loss": 0.2575814723968506, "global_step": 42180, "epoch": 473, "lr": 9.427377528995848e-05}
{"train_loss": 0.3948833644390106, "global_step": 42181, "epoch": 473, "lr": 9.427350589146797e-05}
{"train_loss": 0.3851901590824127, "global_step": 42182, "epoch": 473, "lr": 9.427323648702543e-05}
{"train_loss": 0.2624266743659973, "global_step": 42183, "epoch": 473, "lr": 9.427296707663089e-05}
{"train_loss": 0.2089712768793106, "global_step": 42184, "epoch": 473, "lr": 9.42726976602844e-05}
{"train_loss": 0.312172420741467, "global_step": 42185, "epoch": 473, "lr": 9.427242823798597e-05, "val_loss": 2.7383084297180176}
{"train_loss": 0.3238492012023926, "global_step": 42186, "epoch": 474, "lr": 9.427215880973566e-05}
{"train_loss": 0.261751264333725, "global_step": 42187, "epoch": 474, "lr": 9.427188937553351e-05}
{"train_loss": 0.4037608504295349, "global_step": 42188, "epoch": 474, "lr": 9.427161993537951e-05}
{"train_loss": 0.3651553988456726, "global_step": 42189, "epoch": 474, "lr": 9.427135048927376e-05}
{"train_loss": 0.3205803334712982, "global_step": 42190, "epoch": 474, "lr": 9.427108103721626e-05}
{"train_loss": 0.2450808584690094, "global_step": 42191, "epoch": 474, "lr": 9.427081157920705e-05}
{"train_loss": 0.3714626133441925, "global_step": 42192, "epoch": 474, "lr": 9.427054211524617e-05}
{"train_loss": 0.28415536880493164, "global_step": 42193, "epoch": 474, "lr": 9.427027264533365e-05}
{"train_loss": 0.2805808186531067, "global_step": 42194, "epoch": 474, "lr": 9.427000316946955e-05}
{"train_loss": 0.22939786314964294, "global_step": 42195, "epoch": 474, "lr": 9.426973368765386e-05}
{"train_loss": 0.23743310570716858, "global_step": 42196, "epoch": 474, "lr": 9.426946419988667e-05}
{"train_loss": 0.25348615646362305, "global_step": 42197, "epoch": 474, "lr": 9.426919470616798e-05}
{"train_loss": 0.393801748752594, "global_step": 42198, "epoch": 474, "lr": 9.426892520649783e-05}
{"train_loss": 0.31289976835250854, "global_step": 42199, "epoch": 474, "lr": 9.426865570087626e-05}
{"train_loss": 0.3351885974407196, "global_step": 42200, "epoch": 474, "lr": 9.426838618930331e-05}
{"train_loss": 0.21318607032299042, "global_step": 42201, "epoch": 474, "lr": 9.426811667177902e-05}
{"train_loss": 0.24507832527160645, "global_step": 42202, "epoch": 474, "lr": 9.426784714830344e-05}
{"train_loss": 0.3854793906211853, "global_step": 42203, "epoch": 474, "lr": 9.426757761887657e-05}
{"train_loss": 0.24399887025356293, "global_step": 42204, "epoch": 474, "lr": 9.426730808349846e-05}
{"train_loss": 0.2829113304615021, "global_step": 42205, "epoch": 474, "lr": 9.426703854216914e-05}
{"train_loss": 0.31731224060058594, "global_step": 42206, "epoch": 474, "lr": 9.426676899488868e-05}
{"train_loss": 0.2684021294116974, "global_step": 42207, "epoch": 474, "lr": 9.426649944165708e-05}
{"train_loss": 0.27255380153656006, "global_step": 42208, "epoch": 474, "lr": 9.426622988247439e-05}
{"train_loss": 0.14770950376987457, "global_step": 42209, "epoch": 474, "lr": 9.426596031734064e-05}
{"train_loss": 0.3340492248535156, "global_step": 42210, "epoch": 474, "lr": 9.426569074625588e-05}
{"train_loss": 0.23798343539237976, "global_step": 42211, "epoch": 474, "lr": 9.426542116922014e-05}
{"train_loss": 0.27647659182548523, "global_step": 42212, "epoch": 474, "lr": 9.426515158623344e-05}
{"train_loss": 0.22108998894691467, "global_step": 42213, "epoch": 474, "lr": 9.426488199729584e-05}
{"train_loss": 0.3328562080860138, "global_step": 42214, "epoch": 474, "lr": 9.426461240240735e-05}
{"train_loss": 0.27454572916030884, "global_step": 42215, "epoch": 474, "lr": 9.426434280156803e-05}
{"train_loss": 0.3940505385398865, "global_step": 42216, "epoch": 474, "lr": 9.426407319477792e-05}
{"train_loss": 0.33058199286460876, "global_step": 42217, "epoch": 474, "lr": 9.426380358203706e-05}
{"train_loss": 0.3200918436050415, "global_step": 42218, "epoch": 474, "lr": 9.426353396334544e-05}
{"train_loss": 0.2311922311782837, "global_step": 42219, "epoch": 474, "lr": 9.426326433870314e-05}
{"train_loss": 0.2365099936723709, "global_step": 42220, "epoch": 474, "lr": 9.426299470811018e-05}
{"train_loss": 0.25023165345191956, "global_step": 42221, "epoch": 474, "lr": 9.42627250715666e-05}
{"train_loss": 0.3531818389892578, "global_step": 42222, "epoch": 474, "lr": 9.426245542907244e-05}
{"train_loss": 0.296598881483078, "global_step": 42223, "epoch": 474, "lr": 9.426218578062774e-05}
{"train_loss": 0.34253931045532227, "global_step": 42224, "epoch": 474, "lr": 9.42619161262325e-05}
{"train_loss": 0.28458866477012634, "global_step": 42225, "epoch": 474, "lr": 9.426164646588681e-05}
{"train_loss": 0.29124534130096436, "global_step": 42226, "epoch": 474, "lr": 9.426137679959069e-05}
{"train_loss": 0.23540396988391876, "global_step": 42227, "epoch": 474, "lr": 9.426110712734414e-05}
{"train_loss": 0.28781968355178833, "global_step": 42228, "epoch": 474, "lr": 9.426083744914723e-05}
{"train_loss": 0.2838172912597656, "global_step": 42229, "epoch": 474, "lr": 9.4260567765e-05}
{"train_loss": 0.3534356951713562, "global_step": 42230, "epoch": 474, "lr": 9.426029807490247e-05}
{"train_loss": 0.28372830152511597, "global_step": 42231, "epoch": 474, "lr": 9.426002837885468e-05}
{"train_loss": 0.30242031812667847, "global_step": 42232, "epoch": 474, "lr": 9.425975867685668e-05}
{"train_loss": 0.34580355882644653, "global_step": 42233, "epoch": 474, "lr": 9.425948896890849e-05}
{"train_loss": 0.20337001979351044, "global_step": 42234, "epoch": 474, "lr": 9.425921925501014e-05}
{"train_loss": 0.37322714924812317, "global_step": 42235, "epoch": 474, "lr": 9.425894953516168e-05}
{"train_loss": 0.3611392676830292, "global_step": 42236, "epoch": 474, "lr": 9.425867980936316e-05}
{"train_loss": 0.39398789405822754, "global_step": 42237, "epoch": 474, "lr": 9.425841007761458e-05}
{"train_loss": 0.27380508184432983, "global_step": 42238, "epoch": 474, "lr": 9.425814033991602e-05}
{"train_loss": 0.404188871383667, "global_step": 42239, "epoch": 474, "lr": 9.425787059626747e-05}
{"train_loss": 0.3791984021663666, "global_step": 42240, "epoch": 474, "lr": 9.4257600846669e-05}
{"train_loss": 0.301151305437088, "global_step": 42241, "epoch": 474, "lr": 9.425733109112063e-05}
{"train_loss": 0.25152528285980225, "global_step": 42242, "epoch": 474, "lr": 9.42570613296224e-05}
{"train_loss": 0.2838391363620758, "global_step": 42243, "epoch": 474, "lr": 9.425679156217435e-05}
{"train_loss": 0.3265216648578644, "global_step": 42244, "epoch": 474, "lr": 9.42565217887765e-05}
{"train_loss": 0.3792264759540558, "global_step": 42245, "epoch": 474, "lr": 9.425625200942894e-05}
{"train_loss": 0.29971614480018616, "global_step": 42246, "epoch": 474, "lr": 9.425598222413163e-05}
{"train_loss": 0.33894068002700806, "global_step": 42247, "epoch": 474, "lr": 9.425571243288464e-05}
{"train_loss": 0.2934597432613373, "global_step": 42248, "epoch": 474, "lr": 9.425544263568804e-05}
{"train_loss": 0.22688987851142883, "global_step": 42249, "epoch": 474, "lr": 9.425517283254181e-05}
{"train_loss": 0.22882990539073944, "global_step": 42250, "epoch": 474, "lr": 9.425490302344601e-05}
{"train_loss": 0.4772023558616638, "global_step": 42251, "epoch": 474, "lr": 9.42546332084007e-05}
{"train_loss": 0.4363376200199127, "global_step": 42252, "epoch": 474, "lr": 9.425436338740587e-05}
{"train_loss": 0.33836689591407776, "global_step": 42253, "epoch": 474, "lr": 9.425409356046159e-05}
{"train_loss": 0.3343386948108673, "global_step": 42254, "epoch": 474, "lr": 9.42538237275679e-05}
{"train_loss": 0.31128641963005066, "global_step": 42255, "epoch": 474, "lr": 9.425355388872481e-05}
{"train_loss": 0.37834838032722473, "global_step": 42256, "epoch": 474, "lr": 9.425328404393237e-05}
{"train_loss": 0.3404703736305237, "global_step": 42257, "epoch": 474, "lr": 9.425301419319062e-05}
{"train_loss": 0.3589269518852234, "global_step": 42258, "epoch": 474, "lr": 9.425274433649957e-05}
{"train_loss": 0.4196082055568695, "global_step": 42259, "epoch": 474, "lr": 9.42524744738593e-05}
{"train_loss": 0.47608068585395813, "global_step": 42260, "epoch": 474, "lr": 9.425220460526982e-05}
{"train_loss": 0.28086861968040466, "global_step": 42261, "epoch": 474, "lr": 9.425193473073116e-05}
{"train_loss": 0.3245830535888672, "global_step": 42262, "epoch": 474, "lr": 9.425166485024338e-05}
{"train_loss": 0.3145218789577484, "global_step": 42263, "epoch": 474, "lr": 9.425139496380651e-05}
{"train_loss": 0.41353484988212585, "global_step": 42264, "epoch": 474, "lr": 9.425112507142055e-05}
{"train_loss": 0.544237494468689, "global_step": 42265, "epoch": 474, "lr": 9.425085517308558e-05}
{"train_loss": 0.33075618743896484, "global_step": 42266, "epoch": 474, "lr": 9.425058526880164e-05}
{"train_loss": 0.3475857973098755, "global_step": 42267, "epoch": 474, "lr": 9.425031535856871e-05}
{"train_loss": 0.3180606961250305, "global_step": 42268, "epoch": 474, "lr": 9.42500454423869e-05}
{"train_loss": 0.29043781757354736, "global_step": 42269, "epoch": 474, "lr": 9.42497755202562e-05}
{"train_loss": 0.3314864933490753, "global_step": 42270, "epoch": 474, "lr": 9.424950559217666e-05}
{"train_loss": 0.2504943311214447, "global_step": 42271, "epoch": 474, "lr": 9.424923565814831e-05}
{"train_loss": 0.3112901449203491, "global_step": 42272, "epoch": 474, "lr": 9.424896571817118e-05}
{"train_loss": 0.3159644901752472, "global_step": 42273, "epoch": 474, "lr": 9.424869577224533e-05}
{"train_loss": 0.31436955677659323, "global_step": 42274, "epoch": 474, "lr": 9.424842582037077e-05, "val_loss": 2.694105625152588}
{"train_loss": 0.29733917117118835, "global_step": 42275, "epoch": 475, "lr": 9.424815586254756e-05}
{"train_loss": 0.3785821795463562, "global_step": 42276, "epoch": 475, "lr": 9.424788589877572e-05}
{"train_loss": 0.3869975209236145, "global_step": 42277, "epoch": 475, "lr": 9.424761592905528e-05}
{"train_loss": 0.2877974510192871, "global_step": 42278, "epoch": 475, "lr": 9.424734595338631e-05}
{"train_loss": 0.32831814885139465, "global_step": 42279, "epoch": 475, "lr": 9.42470759717688e-05}
{"train_loss": 0.36856159567832947, "global_step": 42280, "epoch": 475, "lr": 9.424680598420283e-05}
{"train_loss": 0.30598217248916626, "global_step": 42281, "epoch": 475, "lr": 9.42465359906884e-05}
{"train_loss": 0.22265681624412537, "global_step": 42282, "epoch": 475, "lr": 9.424626599122556e-05}
{"train_loss": 0.2559318542480469, "global_step": 42283, "epoch": 475, "lr": 9.424599598581439e-05}
{"train_loss": 0.3418135941028595, "global_step": 42284, "epoch": 475, "lr": 9.424572597445484e-05}
{"train_loss": 0.28335559368133545, "global_step": 42285, "epoch": 475, "lr": 9.4245455957147e-05}
{"train_loss": 0.2427646815776825, "global_step": 42286, "epoch": 475, "lr": 9.424518593389091e-05}
{"train_loss": 0.34125110507011414, "global_step": 42287, "epoch": 475, "lr": 9.424491590468658e-05}
{"train_loss": 0.3586643934249878, "global_step": 42288, "epoch": 475, "lr": 9.424464586953407e-05}
{"train_loss": 0.30746373534202576, "global_step": 42289, "epoch": 475, "lr": 9.424437582843341e-05}
{"train_loss": 0.31325072050094604, "global_step": 42290, "epoch": 475, "lr": 9.424410578138463e-05}
{"train_loss": 0.2577246129512787, "global_step": 42291, "epoch": 475, "lr": 9.424383572838775e-05}
{"train_loss": 0.3131258189678192, "global_step": 42292, "epoch": 475, "lr": 9.424356566944286e-05}
{"train_loss": 0.23944957554340363, "global_step": 42293, "epoch": 475, "lr": 9.424329560454994e-05}
{"train_loss": 0.32104507088661194, "global_step": 42294, "epoch": 475, "lr": 9.424302553370906e-05}
{"train_loss": 0.2790350019931793, "global_step": 42295, "epoch": 475, "lr": 9.424275545692023e-05}
{"train_loss": 0.27654123306274414, "global_step": 42296, "epoch": 475, "lr": 9.42424853741835e-05}
{"train_loss": 0.34227216243743896, "global_step": 42297, "epoch": 475, "lr": 9.424221528549891e-05}
{"train_loss": 0.27117061614990234, "global_step": 42298, "epoch": 475, "lr": 9.424194519086651e-05}
{"train_loss": 0.34945037961006165, "global_step": 42299, "epoch": 475, "lr": 9.424167509028629e-05}
{"train_loss": 0.32652392983436584, "global_step": 42300, "epoch": 475, "lr": 9.424140498375834e-05}
{"train_loss": 0.2738766074180603, "global_step": 42301, "epoch": 475, "lr": 9.424113487128267e-05}
{"train_loss": 0.32654136419296265, "global_step": 42302, "epoch": 475, "lr": 9.424086475285932e-05}
{"train_loss": 0.3137878477573395, "global_step": 42303, "epoch": 475, "lr": 9.424059462848831e-05}
{"train_loss": 0.4491240978240967, "global_step": 42304, "epoch": 475, "lr": 9.424032449816971e-05}
{"train_loss": 0.38879671692848206, "global_step": 42305, "epoch": 475, "lr": 9.424005436190352e-05}
{"train_loss": 0.3772696256637573, "global_step": 42306, "epoch": 475, "lr": 9.42397842196898e-05}
{"train_loss": 0.37897852063179016, "global_step": 42307, "epoch": 475, "lr": 9.423951407152858e-05}
{"train_loss": 0.32787230610847473, "global_step": 42308, "epoch": 475, "lr": 9.423924391741989e-05}
{"train_loss": 0.2920028567314148, "global_step": 42309, "epoch": 475, "lr": 9.423897375736379e-05}
{"train_loss": 0.2579997479915619, "global_step": 42310, "epoch": 475, "lr": 9.423870359136029e-05}
{"train_loss": 0.3591614067554474, "global_step": 42311, "epoch": 475, "lr": 9.423843341940944e-05}
{"train_loss": 0.332612007856369, "global_step": 42312, "epoch": 475, "lr": 9.423816324151127e-05}
{"train_loss": 0.30448025465011597, "global_step": 42313, "epoch": 475, "lr": 9.423789305766582e-05}
{"train_loss": 0.36035844683647156, "global_step": 42314, "epoch": 475, "lr": 9.423762286787312e-05}
{"train_loss": 0.37080419063568115, "global_step": 42315, "epoch": 475, "lr": 9.423735267213321e-05}
{"train_loss": 0.2933069169521332, "global_step": 42316, "epoch": 475, "lr": 9.423708247044612e-05}
{"train_loss": 0.3605020046234131, "global_step": 42317, "epoch": 475, "lr": 9.423681226281191e-05}
{"train_loss": 0.38300544023513794, "global_step": 42318, "epoch": 475, "lr": 9.42365420492306e-05}
{"train_loss": 0.3396926820278168, "global_step": 42319, "epoch": 475, "lr": 9.42362718297022e-05}
{"train_loss": 0.31985095143318176, "global_step": 42320, "epoch": 475, "lr": 9.42360016042268e-05}
{"train_loss": 0.33218497037887573, "global_step": 42321, "epoch": 475, "lr": 9.42357313728044e-05}
{"train_loss": 0.32364532351493835, "global_step": 42322, "epoch": 475, "lr": 9.423546113543505e-05}
{"train_loss": 0.3720858693122864, "global_step": 42323, "epoch": 475, "lr": 9.423519089211876e-05}
{"train_loss": 0.37522462010383606, "global_step": 42324, "epoch": 475, "lr": 9.42349206428556e-05}
{"train_loss": 0.3003053367137909, "global_step": 42325, "epoch": 475, "lr": 9.42346503876456e-05}
{"train_loss": 0.25174301862716675, "global_step": 42326, "epoch": 475, "lr": 9.423438012648879e-05}
{"train_loss": 0.2998253405094147, "global_step": 42327, "epoch": 475, "lr": 9.423410985938521e-05}
{"train_loss": 0.259869247674942, "global_step": 42328, "epoch": 475, "lr": 9.42338395863349e-05}
{"train_loss": 0.33378034830093384, "global_step": 42329, "epoch": 475, "lr": 9.423356930733786e-05}
{"train_loss": 0.33354872465133667, "global_step": 42330, "epoch": 475, "lr": 9.423329902239417e-05}
{"train_loss": 0.3868701457977295, "global_step": 42331, "epoch": 475, "lr": 9.423302873150386e-05}
{"train_loss": 0.3692675232887268, "global_step": 42332, "epoch": 475, "lr": 9.423275843466696e-05}
{"train_loss": 0.30269157886505127, "global_step": 42333, "epoch": 475, "lr": 9.42324881318835e-05}
{"train_loss": 0.3415316641330719, "global_step": 42334, "epoch": 475, "lr": 9.423221782315352e-05}
{"train_loss": 0.38729915022850037, "global_step": 42335, "epoch": 475, "lr": 9.423194750847705e-05}
{"train_loss": 0.2653292119503021, "global_step": 42336, "epoch": 475, "lr": 9.423167718785415e-05}
{"train_loss": 0.3380281627178192, "global_step": 42337, "epoch": 475, "lr": 9.423140686128482e-05}
{"train_loss": 0.401481568813324, "global_step": 42338, "epoch": 475, "lr": 9.423113652876912e-05}
{"train_loss": 0.37456515431404114, "global_step": 42339, "epoch": 475, "lr": 9.42308661903071e-05}
{"train_loss": 0.3406720757484436, "global_step": 42340, "epoch": 475, "lr": 9.423059584589875e-05}
{"train_loss": 0.3360125720500946, "global_step": 42341, "epoch": 475, "lr": 9.423032549554417e-05}
{"train_loss": 0.3461453914642334, "global_step": 42342, "epoch": 475, "lr": 9.423005513924335e-05}
{"train_loss": 0.25694993138313293, "global_step": 42343, "epoch": 475, "lr": 9.422978477699633e-05}
{"train_loss": 0.23083940148353577, "global_step": 42344, "epoch": 475, "lr": 9.422951440880316e-05}
{"train_loss": 0.37269270420074463, "global_step": 42345, "epoch": 475, "lr": 9.422924403466386e-05}
{"train_loss": 0.33800479769706726, "global_step": 42346, "epoch": 475, "lr": 9.422897365457849e-05}
{"train_loss": 0.31962868571281433, "global_step": 42347, "epoch": 475, "lr": 9.422870326854707e-05}
{"train_loss": 0.35248738527297974, "global_step": 42348, "epoch": 475, "lr": 9.422843287656964e-05}
{"train_loss": 0.3478579819202423, "global_step": 42349, "epoch": 475, "lr": 9.422816247864623e-05}
{"train_loss": 0.3507407605648041, "global_step": 42350, "epoch": 475, "lr": 9.422789207477689e-05}
{"train_loss": 0.2309580296278, "global_step": 42351, "epoch": 475, "lr": 9.422762166496165e-05}
{"train_loss": 0.2398844063282013, "global_step": 42352, "epoch": 475, "lr": 9.422735124920055e-05}
{"train_loss": 0.3983229398727417, "global_step": 42353, "epoch": 475, "lr": 9.42270808274936e-05}
{"train_loss": 0.32133522629737854, "global_step": 42354, "epoch": 475, "lr": 9.422681039984089e-05}
{"train_loss": 0.3154531717300415, "global_step": 42355, "epoch": 475, "lr": 9.42265399662424e-05}
{"train_loss": 0.3477202355861664, "global_step": 42356, "epoch": 475, "lr": 9.422626952669819e-05}
{"train_loss": 0.26070746779441833, "global_step": 42357, "epoch": 475, "lr": 9.422599908120832e-05}
{"train_loss": 0.23271618783473969, "global_step": 42358, "epoch": 475, "lr": 9.42257286297728e-05}
{"train_loss": 0.32344281673431396, "global_step": 42359, "epoch": 475, "lr": 9.422545817239164e-05}
{"train_loss": 0.29862532019615173, "global_step": 42360, "epoch": 475, "lr": 9.422518770906493e-05}
{"train_loss": 0.19537822902202606, "global_step": 42361, "epoch": 475, "lr": 9.422491723979268e-05}
{"train_loss": 0.29682302474975586, "global_step": 42362, "epoch": 475, "lr": 9.422464676457492e-05}
{"train_loss": 0.32016943277937643, "global_step": 42363, "epoch": 475, "lr": 9.42243762834117e-05, "val_loss": 2.726980209350586, "train_action_mse_error": 24.321102142333984}
{"train_loss": 0.24658113718032837, "global_step": 42364, "epoch": 476, "lr": 9.422410579630306e-05}
{"train_loss": 0.23803585767745972, "global_step": 42365, "epoch": 476, "lr": 9.422383530324902e-05}
{"train_loss": 0.33704254031181335, "global_step": 42366, "epoch": 476, "lr": 9.422356480424963e-05}
{"train_loss": 0.38070645928382874, "global_step": 42367, "epoch": 476, "lr": 9.42232942993049e-05}
{"train_loss": 0.3326015770435333, "global_step": 42368, "epoch": 476, "lr": 9.422302378841492e-05}
{"train_loss": 0.2853028178215027, "global_step": 42369, "epoch": 476, "lr": 9.422275327157968e-05}
{"train_loss": 0.3542448580265045, "global_step": 42370, "epoch": 476, "lr": 9.422248274879921e-05}
{"train_loss": 0.21812129020690918, "global_step": 42371, "epoch": 476, "lr": 9.422221222007359e-05}
{"train_loss": 0.37636616826057434, "global_step": 42372, "epoch": 476, "lr": 9.422194168540282e-05}
{"train_loss": 0.39420294761657715, "global_step": 42373, "epoch": 476, "lr": 9.422167114478697e-05}
{"train_loss": 0.35517045855522156, "global_step": 42374, "epoch": 476, "lr": 9.422140059822602e-05}
{"train_loss": 0.2655096650123596, "global_step": 42375, "epoch": 476, "lr": 9.422113004572007e-05}
{"train_loss": 0.2899596393108368, "global_step": 42376, "epoch": 476, "lr": 9.422085948726912e-05}
{"train_loss": 0.2972775399684906, "global_step": 42377, "epoch": 476, "lr": 9.422058892287323e-05}
{"train_loss": 0.394900381565094, "global_step": 42378, "epoch": 476, "lr": 9.42203183525324e-05}
{"train_loss": 0.22780445218086243, "global_step": 42379, "epoch": 476, "lr": 9.422004777624668e-05}
{"train_loss": 0.366326242685318, "global_step": 42380, "epoch": 476, "lr": 9.421977719401612e-05}
{"train_loss": 0.28146904706954956, "global_step": 42381, "epoch": 476, "lr": 9.421950660584076e-05}
{"train_loss": 0.39187222719192505, "global_step": 42382, "epoch": 476, "lr": 9.421923601172062e-05}
{"train_loss": 0.3515799939632416, "global_step": 42383, "epoch": 476, "lr": 9.421896541165574e-05}
{"train_loss": 0.380642294883728, "global_step": 42384, "epoch": 476, "lr": 9.421869480564616e-05}
{"train_loss": 0.34392452239990234, "global_step": 42385, "epoch": 476, "lr": 9.421842419369193e-05}
{"train_loss": 0.29874828457832336, "global_step": 42386, "epoch": 476, "lr": 9.421815357579307e-05}
{"train_loss": 0.36615172028541565, "global_step": 42387, "epoch": 476, "lr": 9.42178829519496e-05}
{"train_loss": 0.3176443874835968, "global_step": 42388, "epoch": 476, "lr": 9.421761232216157e-05}
{"train_loss": 0.28222575783729553, "global_step": 42389, "epoch": 476, "lr": 9.421734168642904e-05}
{"train_loss": 0.30135607719421387, "global_step": 42390, "epoch": 476, "lr": 9.421707104475203e-05}
{"train_loss": 0.276987224817276, "global_step": 42391, "epoch": 476, "lr": 9.421680039713057e-05}
{"train_loss": 0.24585166573524475, "global_step": 42392, "epoch": 476, "lr": 9.421652974356468e-05}
{"train_loss": 0.2670043110847473, "global_step": 42393, "epoch": 476, "lr": 9.421625908405444e-05}
{"train_loss": 0.33595749735832214, "global_step": 42394, "epoch": 476, "lr": 9.421598841859985e-05}
{"train_loss": 0.36948540806770325, "global_step": 42395, "epoch": 476, "lr": 9.421571774720097e-05}
{"train_loss": 0.29374486207962036, "global_step": 42396, "epoch": 476, "lr": 9.421544706985781e-05}
{"train_loss": 0.353260338306427, "global_step": 42397, "epoch": 476, "lr": 9.421517638657043e-05}
{"train_loss": 0.3457101583480835, "global_step": 42398, "epoch": 476, "lr": 9.421490569733887e-05}
{"train_loss": 0.29562726616859436, "global_step": 42399, "epoch": 476, "lr": 9.421463500216315e-05}
{"train_loss": 0.2674250304698944, "global_step": 42400, "epoch": 476, "lr": 9.42143643010433e-05}
{"train_loss": 0.305266797542572, "global_step": 42401, "epoch": 476, "lr": 9.421409359397937e-05}
{"train_loss": 0.36534062027931213, "global_step": 42402, "epoch": 476, "lr": 9.42138228809714e-05}
{"train_loss": 0.31789687275886536, "global_step": 42403, "epoch": 476, "lr": 9.421355216201941e-05}
{"train_loss": 0.3287051022052765, "global_step": 42404, "epoch": 476, "lr": 9.421328143712347e-05}
{"train_loss": 0.2937013804912567, "global_step": 42405, "epoch": 476, "lr": 9.421301070628357e-05}
{"train_loss": 0.2695586085319519, "global_step": 42406, "epoch": 476, "lr": 9.421273996949978e-05}
{"train_loss": 0.26745766401290894, "global_step": 42407, "epoch": 476, "lr": 9.421246922677213e-05}
{"train_loss": 0.26940035820007324, "global_step": 42408, "epoch": 476, "lr": 9.421219847810064e-05}
{"train_loss": 0.3267093300819397, "global_step": 42409, "epoch": 476, "lr": 9.421192772348538e-05}
{"train_loss": 0.35126546025276184, "global_step": 42410, "epoch": 476, "lr": 9.421165696292633e-05}
{"train_loss": 0.2888430953025818, "global_step": 42411, "epoch": 476, "lr": 9.421138619642359e-05}
{"train_loss": 0.393513023853302, "global_step": 42412, "epoch": 476, "lr": 9.421111542397716e-05}
{"train_loss": 0.3535061478614807, "global_step": 42413, "epoch": 476, "lr": 9.421084464558708e-05}
{"train_loss": 0.34162160754203796, "global_step": 42414, "epoch": 476, "lr": 9.42105738612534e-05}
{"train_loss": 0.20892615616321564, "global_step": 42415, "epoch": 476, "lr": 9.421030307097613e-05}
{"train_loss": 0.26247212290763855, "global_step": 42416, "epoch": 476, "lr": 9.421003227475535e-05}
{"train_loss": 0.31802231073379517, "global_step": 42417, "epoch": 476, "lr": 9.420976147259106e-05}
{"train_loss": 0.386016845703125, "global_step": 42418, "epoch": 476, "lr": 9.420949066448329e-05}
{"train_loss": 0.2849622666835785, "global_step": 42419, "epoch": 476, "lr": 9.420921985043211e-05}
{"train_loss": 0.33155950903892517, "global_step": 42420, "epoch": 476, "lr": 9.420894903043753e-05}
{"train_loss": 0.3976251184940338, "global_step": 42421, "epoch": 476, "lr": 9.42086782044996e-05}
{"train_loss": 0.39206141233444214, "global_step": 42422, "epoch": 476, "lr": 9.420840737261836e-05}
{"train_loss": 0.3275543749332428, "global_step": 42423, "epoch": 476, "lr": 9.420813653479382e-05}
{"train_loss": 0.3174380660057068, "global_step": 42424, "epoch": 476, "lr": 9.420786569102604e-05}
{"train_loss": 0.2921638786792755, "global_step": 42425, "epoch": 476, "lr": 9.420759484131507e-05}
{"train_loss": 0.33116570115089417, "global_step": 42426, "epoch": 476, "lr": 9.42073239856609e-05}
{"train_loss": 0.29766732454299927, "global_step": 42427, "epoch": 476, "lr": 9.420705312406362e-05}
{"train_loss": 0.3650260269641876, "global_step": 42428, "epoch": 476, "lr": 9.420678225652323e-05}
{"train_loss": 0.39902809262275696, "global_step": 42429, "epoch": 476, "lr": 9.420651138303978e-05}
{"train_loss": 0.2978944480419159, "global_step": 42430, "epoch": 476, "lr": 9.420624050361328e-05}
{"train_loss": 0.32459545135498047, "global_step": 42431, "epoch": 476, "lr": 9.420596961824382e-05}
{"train_loss": 0.32113611698150635, "global_step": 42432, "epoch": 476, "lr": 9.420569872693139e-05}
{"train_loss": 0.3783182203769684, "global_step": 42433, "epoch": 476, "lr": 9.420542782967605e-05}
{"train_loss": 0.2328062802553177, "global_step": 42434, "epoch": 476, "lr": 9.420515692647782e-05}
{"train_loss": 0.36446332931518555, "global_step": 42435, "epoch": 476, "lr": 9.420488601733676e-05}
{"train_loss": 0.2324451208114624, "global_step": 42436, "epoch": 476, "lr": 9.420461510225289e-05}
{"train_loss": 0.3960455656051636, "global_step": 42437, "epoch": 476, "lr": 9.420434418122624e-05}
{"train_loss": 0.22446539998054504, "global_step": 42438, "epoch": 476, "lr": 9.420407325425685e-05}
{"train_loss": 0.3641168475151062, "global_step": 42439, "epoch": 476, "lr": 9.420380232134478e-05}
{"train_loss": 0.30089759826660156, "global_step": 42440, "epoch": 476, "lr": 9.420353138249003e-05}
{"train_loss": 0.2638963758945465, "global_step": 42441, "epoch": 476, "lr": 9.420326043769266e-05}
{"train_loss": 0.31924137473106384, "global_step": 42442, "epoch": 476, "lr": 9.420298948695269e-05}
{"train_loss": 0.24008965492248535, "global_step": 42443, "epoch": 476, "lr": 9.420271853027019e-05}
{"train_loss": 0.38805145025253296, "global_step": 42444, "epoch": 476, "lr": 9.420244756764515e-05}
{"train_loss": 0.24526222050189972, "global_step": 42445, "epoch": 476, "lr": 9.420217659907764e-05}
{"train_loss": 0.2627522945404053, "global_step": 42446, "epoch": 476, "lr": 9.420190562456769e-05}
{"train_loss": 0.2820155918598175, "global_step": 42447, "epoch": 476, "lr": 9.420163464411534e-05}
{"train_loss": 0.34379756450653076, "global_step": 42448, "epoch": 476, "lr": 9.42013636577206e-05}
{"train_loss": 0.3379382789134979, "global_step": 42449, "epoch": 476, "lr": 9.420109266538353e-05}
{"train_loss": 0.2588604986667633, "global_step": 42450, "epoch": 476, "lr": 9.420082166710416e-05}
{"train_loss": 0.383425772190094, "global_step": 42451, "epoch": 476, "lr": 9.420055066288253e-05}
{"train_loss": 0.3159975762782472, "global_step": 42452, "epoch": 476, "lr": 9.420027965271868e-05, "val_loss": 2.7911159992218018}
{"train_loss": 0.2755720317363739, "global_step": 42453, "epoch": 477, "lr": 9.420000863661265e-05}
{"train_loss": 0.3162361979484558, "global_step": 42454, "epoch": 477, "lr": 9.419973761456444e-05}
{"train_loss": 0.33169153332710266, "global_step": 42455, "epoch": 477, "lr": 9.419946658657414e-05}
{"train_loss": 0.28003057837486267, "global_step": 42456, "epoch": 477, "lr": 9.419919555264174e-05}
{"train_loss": 0.27679309248924255, "global_step": 42457, "epoch": 477, "lr": 9.419892451276732e-05}
{"train_loss": 0.386567622423172, "global_step": 42458, "epoch": 477, "lr": 9.419865346695087e-05}
{"train_loss": 0.28907230496406555, "global_step": 42459, "epoch": 477, "lr": 9.419838241519247e-05}
{"train_loss": 0.2878546416759491, "global_step": 42460, "epoch": 477, "lr": 9.419811135749212e-05}
{"train_loss": 0.3171589970588684, "global_step": 42461, "epoch": 477, "lr": 9.419784029384988e-05}
{"train_loss": 0.3552234172821045, "global_step": 42462, "epoch": 477, "lr": 9.419756922426578e-05}
{"train_loss": 0.30299967527389526, "global_step": 42463, "epoch": 477, "lr": 9.419729814873985e-05}
{"train_loss": 0.24481411278247833, "global_step": 42464, "epoch": 477, "lr": 9.419702706727214e-05}
{"train_loss": 0.3042689859867096, "global_step": 42465, "epoch": 477, "lr": 9.419675597986268e-05}
{"train_loss": 0.22751161456108093, "global_step": 42466, "epoch": 477, "lr": 9.41964848865115e-05}
{"train_loss": 0.3705933392047882, "global_step": 42467, "epoch": 477, "lr": 9.419621378721863e-05}
{"train_loss": 0.37451595067977905, "global_step": 42468, "epoch": 477, "lr": 9.419594268198412e-05}
{"train_loss": 0.2516867220401764, "global_step": 42469, "epoch": 477, "lr": 9.419567157080801e-05}
{"train_loss": 0.22093112766742706, "global_step": 42470, "epoch": 477, "lr": 9.419540045369035e-05}
{"train_loss": 0.31230753660202026, "global_step": 42471, "epoch": 477, "lr": 9.419512933063114e-05}
{"train_loss": 0.23561163246631622, "global_step": 42472, "epoch": 477, "lr": 9.419485820163043e-05}
{"train_loss": 0.21946793794631958, "global_step": 42473, "epoch": 477, "lr": 9.419458706668827e-05}
{"train_loss": 0.2864936888217926, "global_step": 42474, "epoch": 477, "lr": 9.419431592580468e-05}
{"train_loss": 0.34652459621429443, "global_step": 42475, "epoch": 477, "lr": 9.41940447789797e-05}
{"train_loss": 0.26832979917526245, "global_step": 42476, "epoch": 477, "lr": 9.419377362621338e-05}
{"train_loss": 0.3623100817203522, "global_step": 42477, "epoch": 477, "lr": 9.419350246750574e-05}
{"train_loss": 0.2914794683456421, "global_step": 42478, "epoch": 477, "lr": 9.41932313028568e-05}
{"train_loss": 0.282148540019989, "global_step": 42479, "epoch": 477, "lr": 9.419296013226664e-05}
{"train_loss": 0.259891152381897, "global_step": 42480, "epoch": 477, "lr": 9.419268895573527e-05}
{"train_loss": 0.32263827323913574, "global_step": 42481, "epoch": 477, "lr": 9.419241777326276e-05}
{"train_loss": 0.2191445678472519, "global_step": 42482, "epoch": 477, "lr": 9.419214658484907e-05}
{"train_loss": 0.2769065201282501, "global_step": 42483, "epoch": 477, "lr": 9.419187539049433e-05}
{"train_loss": 0.2555059492588043, "global_step": 42484, "epoch": 477, "lr": 9.41916041901985e-05}
{"train_loss": 0.3449322581291199, "global_step": 42485, "epoch": 477, "lr": 9.419133298396166e-05}
{"train_loss": 0.21049924194812775, "global_step": 42486, "epoch": 477, "lr": 9.419106177178383e-05}
{"train_loss": 0.332071453332901, "global_step": 42487, "epoch": 477, "lr": 9.419079055366504e-05}
{"train_loss": 0.26904091238975525, "global_step": 42488, "epoch": 477, "lr": 9.419051932960535e-05}
{"train_loss": 0.2707691788673401, "global_step": 42489, "epoch": 477, "lr": 9.419024809960479e-05}
{"train_loss": 0.3618265986442566, "global_step": 42490, "epoch": 477, "lr": 9.418997686366338e-05}
{"train_loss": 0.2315627783536911, "global_step": 42491, "epoch": 477, "lr": 9.418970562178115e-05}
{"train_loss": 0.29624614119529724, "global_step": 42492, "epoch": 477, "lr": 9.418943437395818e-05}
{"train_loss": 0.3158862292766571, "global_step": 42493, "epoch": 477, "lr": 9.418916312019446e-05}
{"train_loss": 0.2810337543487549, "global_step": 42494, "epoch": 477, "lr": 9.418889186049007e-05}
{"train_loss": 0.2145538479089737, "global_step": 42495, "epoch": 477, "lr": 9.4188620594845e-05}
{"train_loss": 0.2512825131416321, "global_step": 42496, "epoch": 477, "lr": 9.418834932325931e-05}
{"train_loss": 0.3098456859588623, "global_step": 42497, "epoch": 477, "lr": 9.418807804573304e-05}
{"train_loss": 0.2933264672756195, "global_step": 42498, "epoch": 477, "lr": 9.418780676226623e-05}
{"train_loss": 0.4005320370197296, "global_step": 42499, "epoch": 477, "lr": 9.41875354728589e-05}
{"train_loss": 0.3286033868789673, "global_step": 42500, "epoch": 477, "lr": 9.41872641775111e-05}
{"train_loss": 0.2668754756450653, "global_step": 42501, "epoch": 477, "lr": 9.418699287622284e-05}
{"train_loss": 0.2162313312292099, "global_step": 42502, "epoch": 477, "lr": 9.41867215689942e-05}
{"train_loss": 0.339679479598999, "global_step": 42503, "epoch": 477, "lr": 9.418645025582517e-05}
{"train_loss": 0.3060667812824249, "global_step": 42504, "epoch": 477, "lr": 9.418617893671584e-05}
{"train_loss": 0.25090867280960083, "global_step": 42505, "epoch": 477, "lr": 9.41859076116662e-05}
{"train_loss": 0.30581456422805786, "global_step": 42506, "epoch": 477, "lr": 9.418563628067632e-05}
{"train_loss": 0.2542772591114044, "global_step": 42507, "epoch": 477, "lr": 9.41853649437462e-05}
{"train_loss": 0.32543060183525085, "global_step": 42508, "epoch": 477, "lr": 9.418509360087589e-05}
{"train_loss": 0.31171005964279175, "global_step": 42509, "epoch": 477, "lr": 9.418482225206546e-05}
{"train_loss": 0.24203436076641083, "global_step": 42510, "epoch": 477, "lr": 9.41845508973149e-05}
{"train_loss": 0.3249949812889099, "global_step": 42511, "epoch": 477, "lr": 9.418427953662427e-05}
{"train_loss": 0.45604559779167175, "global_step": 42512, "epoch": 477, "lr": 9.41840081699936e-05}
{"train_loss": 0.27751052379608154, "global_step": 42513, "epoch": 477, "lr": 9.418373679742292e-05}
{"train_loss": 0.25243204832077026, "global_step": 42514, "epoch": 477, "lr": 9.41834654189123e-05}
{"train_loss": 0.3093898892402649, "global_step": 42515, "epoch": 477, "lr": 9.418319403446173e-05}
{"train_loss": 0.40029484033584595, "global_step": 42516, "epoch": 477, "lr": 9.418292264407127e-05}
{"train_loss": 0.2719874978065491, "global_step": 42517, "epoch": 477, "lr": 9.418265124774097e-05}
{"train_loss": 0.37881872057914734, "global_step": 42518, "epoch": 477, "lr": 9.418237984547083e-05}
{"train_loss": 0.3066541850566864, "global_step": 42519, "epoch": 477, "lr": 9.418210843726092e-05}
{"train_loss": 0.18847186863422394, "global_step": 42520, "epoch": 477, "lr": 9.418183702311127e-05}
{"train_loss": 0.33967486023902893, "global_step": 42521, "epoch": 477, "lr": 9.418156560302191e-05}
{"train_loss": 0.1808370053768158, "global_step": 42522, "epoch": 477, "lr": 9.418129417699286e-05}
{"train_loss": 0.36443379521369934, "global_step": 42523, "epoch": 477, "lr": 9.418102274502418e-05}
{"train_loss": 0.34163451194763184, "global_step": 42524, "epoch": 477, "lr": 9.41807513071159e-05}
{"train_loss": 0.2932316064834595, "global_step": 42525, "epoch": 477, "lr": 9.418047986326807e-05}
{"train_loss": 0.22576399147510529, "global_step": 42526, "epoch": 477, "lr": 9.41802084134807e-05}
{"train_loss": 0.31491759419441223, "global_step": 42527, "epoch": 477, "lr": 9.417993695775385e-05}
{"train_loss": 0.2911064326763153, "global_step": 42528, "epoch": 477, "lr": 9.417966549608754e-05}
{"train_loss": 0.3744654357433319, "global_step": 42529, "epoch": 477, "lr": 9.41793940284818e-05}
{"train_loss": 0.3387161195278168, "global_step": 42530, "epoch": 477, "lr": 9.417912255493669e-05}
{"train_loss": 0.28196948766708374, "global_step": 42531, "epoch": 477, "lr": 9.417885107545224e-05}
{"train_loss": 0.36669206619262695, "global_step": 42532, "epoch": 477, "lr": 9.417857959002847e-05}
{"train_loss": 0.22910156846046448, "global_step": 42533, "epoch": 477, "lr": 9.417830809866544e-05}
{"train_loss": 0.3455618917942047, "global_step": 42534, "epoch": 477, "lr": 9.417803660136317e-05}
{"train_loss": 0.30338189005851746, "global_step": 42535, "epoch": 477, "lr": 9.41777650981217e-05}
{"train_loss": 0.28547215461730957, "global_step": 42536, "epoch": 477, "lr": 9.417749358894108e-05}
{"train_loss": 0.31143292784690857, "global_step": 42537, "epoch": 477, "lr": 9.417722207382132e-05}
{"train_loss": 0.24093712866306305, "global_step": 42538, "epoch": 477, "lr": 9.417695055276248e-05}
{"train_loss": 0.33895131945610046, "global_step": 42539, "epoch": 477, "lr": 9.417667902576457e-05}
{"train_loss": 0.3208189010620117, "global_step": 42540, "epoch": 477, "lr": 9.417640749282766e-05}
{"train_loss": 0.2972059711981355, "global_step": 42541, "epoch": 477, "lr": 9.417613595395178e-05, "val_loss": 2.7018866539001465}
{"train_loss": 0.40901264548301697, "global_step": 42542, "epoch": 478, "lr": 9.417586440913693e-05}
{"train_loss": 0.3591172993183136, "global_step": 42543, "epoch": 478, "lr": 9.41755928583832e-05}
{"train_loss": 0.3236478567123413, "global_step": 42544, "epoch": 478, "lr": 9.417532130169057e-05}
{"train_loss": 0.22545026242733002, "global_step": 42545, "epoch": 478, "lr": 9.417504973905913e-05}
{"train_loss": 0.30851489305496216, "global_step": 42546, "epoch": 478, "lr": 9.41747781704889e-05}
{"train_loss": 0.3370770514011383, "global_step": 42547, "epoch": 478, "lr": 9.417450659597988e-05}
{"train_loss": 0.3107147812843323, "global_step": 42548, "epoch": 478, "lr": 9.417423501553215e-05}
{"train_loss": 0.3434556722640991, "global_step": 42549, "epoch": 478, "lr": 9.417396342914576e-05}
{"train_loss": 0.276166707277298, "global_step": 42550, "epoch": 478, "lr": 9.417369183682068e-05}
{"train_loss": 0.3741156756877899, "global_step": 42551, "epoch": 478, "lr": 9.4173420238557e-05}
{"train_loss": 0.2602795958518982, "global_step": 42552, "epoch": 478, "lr": 9.417314863435474e-05}
{"train_loss": 0.2496480792760849, "global_step": 42553, "epoch": 478, "lr": 9.417287702421393e-05}
{"train_loss": 0.3114405870437622, "global_step": 42554, "epoch": 478, "lr": 9.417260540813464e-05}
{"train_loss": 0.33411961793899536, "global_step": 42555, "epoch": 478, "lr": 9.417233378611686e-05}
{"train_loss": 0.26755139231681824, "global_step": 42556, "epoch": 478, "lr": 9.417206215816065e-05}
{"train_loss": 0.2702333629131317, "global_step": 42557, "epoch": 478, "lr": 9.417179052426606e-05}
{"train_loss": 0.2178536355495453, "global_step": 42558, "epoch": 478, "lr": 9.417151888443309e-05}
{"train_loss": 0.2608432471752167, "global_step": 42559, "epoch": 478, "lr": 9.417124723866183e-05}
{"train_loss": 0.3125598728656769, "global_step": 42560, "epoch": 478, "lr": 9.417097558695225e-05}
{"train_loss": 0.41373008489608765, "global_step": 42561, "epoch": 478, "lr": 9.417070392930443e-05}
{"train_loss": 0.26632338762283325, "global_step": 42562, "epoch": 478, "lr": 9.41704322657184e-05}
{"train_loss": 0.2865435481071472, "global_step": 42563, "epoch": 478, "lr": 9.41701605961942e-05}
{"train_loss": 0.26685330271720886, "global_step": 42564, "epoch": 478, "lr": 9.416988892073185e-05}
{"train_loss": 0.300976037979126, "global_step": 42565, "epoch": 478, "lr": 9.41696172393314e-05}
{"train_loss": 0.2105702906847, "global_step": 42566, "epoch": 478, "lr": 9.416934555199289e-05}
{"train_loss": 0.22508186101913452, "global_step": 42567, "epoch": 478, "lr": 9.416907385871634e-05}
{"train_loss": 0.24407407641410828, "global_step": 42568, "epoch": 478, "lr": 9.41688021595018e-05}
{"train_loss": 0.2685146927833557, "global_step": 42569, "epoch": 478, "lr": 9.41685304543493e-05}
{"train_loss": 0.3229905068874359, "global_step": 42570, "epoch": 478, "lr": 9.416825874325888e-05}
{"train_loss": 0.23739364743232727, "global_step": 42571, "epoch": 478, "lr": 9.416798702623059e-05}
{"train_loss": 0.22488611936569214, "global_step": 42572, "epoch": 478, "lr": 9.416771530326444e-05}
{"train_loss": 0.30946600437164307, "global_step": 42573, "epoch": 478, "lr": 9.416744357436048e-05}
{"train_loss": 0.2497415542602539, "global_step": 42574, "epoch": 478, "lr": 9.416717183951875e-05}
{"train_loss": 0.24225161969661713, "global_step": 42575, "epoch": 478, "lr": 9.416690009873927e-05}
{"train_loss": 0.2791825830936432, "global_step": 42576, "epoch": 478, "lr": 9.41666283520221e-05}
{"train_loss": 0.35842856764793396, "global_step": 42577, "epoch": 478, "lr": 9.416635659936726e-05}
{"train_loss": 0.2816712260246277, "global_step": 42578, "epoch": 478, "lr": 9.416608484077479e-05}
{"train_loss": 0.3487692177295685, "global_step": 42579, "epoch": 478, "lr": 9.416581307624474e-05}
{"train_loss": 0.32838091254234314, "global_step": 42580, "epoch": 478, "lr": 9.416554130577713e-05}
{"train_loss": 0.36120718717575073, "global_step": 42581, "epoch": 478, "lr": 9.4165269529372e-05}
{"train_loss": 0.29699477553367615, "global_step": 42582, "epoch": 478, "lr": 9.416499774702938e-05}
{"train_loss": 0.2480587363243103, "global_step": 42583, "epoch": 478, "lr": 9.416472595874932e-05}
{"train_loss": 0.2508307099342346, "global_step": 42584, "epoch": 478, "lr": 9.416445416453186e-05}
{"train_loss": 0.19062906503677368, "global_step": 42585, "epoch": 478, "lr": 9.4164182364377e-05}
{"train_loss": 0.3072040379047394, "global_step": 42586, "epoch": 478, "lr": 9.416391055828483e-05}
{"train_loss": 0.3038918673992157, "global_step": 42587, "epoch": 478, "lr": 9.416363874625535e-05}
{"train_loss": 0.30069369077682495, "global_step": 42588, "epoch": 478, "lr": 9.416336692828862e-05}
{"train_loss": 0.35604384541511536, "global_step": 42589, "epoch": 478, "lr": 9.416309510438464e-05}
{"train_loss": 0.31938791275024414, "global_step": 42590, "epoch": 478, "lr": 9.416282327454349e-05}
{"train_loss": 0.24294701218605042, "global_step": 42591, "epoch": 478, "lr": 9.416255143876517e-05}
{"train_loss": 0.2650500237941742, "global_step": 42592, "epoch": 478, "lr": 9.416227959704975e-05}
{"train_loss": 0.2970223128795624, "global_step": 42593, "epoch": 478, "lr": 9.416200774939726e-05}
{"train_loss": 0.29789283871650696, "global_step": 42594, "epoch": 478, "lr": 9.41617358958077e-05}
{"train_loss": 0.3003685772418976, "global_step": 42595, "epoch": 478, "lr": 9.416146403628114e-05}
{"train_loss": 0.285040020942688, "global_step": 42596, "epoch": 478, "lr": 9.41611921708176e-05}
{"train_loss": 0.2762390077114105, "global_step": 42597, "epoch": 478, "lr": 9.416092029941714e-05}
{"train_loss": 0.2869769334793091, "global_step": 42598, "epoch": 478, "lr": 9.416064842207979e-05}
{"train_loss": 0.34025731682777405, "global_step": 42599, "epoch": 478, "lr": 9.416037653880557e-05}
{"train_loss": 0.2920905649662018, "global_step": 42600, "epoch": 478, "lr": 9.416010464959452e-05}
{"train_loss": 0.31688210368156433, "global_step": 42601, "epoch": 478, "lr": 9.415983275444669e-05}
{"train_loss": 0.23996196687221527, "global_step": 42602, "epoch": 478, "lr": 9.41595608533621e-05}
{"train_loss": 0.2569994330406189, "global_step": 42603, "epoch": 478, "lr": 9.41592889463408e-05}
{"train_loss": 0.3026079833507538, "global_step": 42604, "epoch": 478, "lr": 9.415901703338283e-05}
{"train_loss": 0.38061800599098206, "global_step": 42605, "epoch": 478, "lr": 9.41587451144882e-05}
{"train_loss": 0.2162812352180481, "global_step": 42606, "epoch": 478, "lr": 9.415847318965699e-05}
{"train_loss": 0.3211020827293396, "global_step": 42607, "epoch": 478, "lr": 9.41582012588892e-05}
{"train_loss": 0.3118823766708374, "global_step": 42608, "epoch": 478, "lr": 9.415792932218487e-05}
{"train_loss": 0.3437518775463104, "global_step": 42609, "epoch": 478, "lr": 9.415765737954406e-05}
{"train_loss": 0.5080761909484863, "global_step": 42610, "epoch": 478, "lr": 9.415738543096677e-05}
{"train_loss": 0.3598150908946991, "global_step": 42611, "epoch": 478, "lr": 9.415711347645306e-05}
{"train_loss": 0.28093206882476807, "global_step": 42612, "epoch": 478, "lr": 9.4156841516003e-05}
{"train_loss": 0.3099423050880432, "global_step": 42613, "epoch": 478, "lr": 9.415656954961655e-05}
{"train_loss": 0.2637684941291809, "global_step": 42614, "epoch": 478, "lr": 9.41562975772938e-05}
{"train_loss": 0.2783651351928711, "global_step": 42615, "epoch": 478, "lr": 9.415602559903478e-05}
{"train_loss": 0.29354891180992126, "global_step": 42616, "epoch": 478, "lr": 9.41557536148395e-05}
{"train_loss": 0.4647715091705322, "global_step": 42617, "epoch": 478, "lr": 9.415548162470805e-05}
{"train_loss": 0.25734540820121765, "global_step": 42618, "epoch": 478, "lr": 9.415520962864042e-05}
{"train_loss": 0.20437321066856384, "global_step": 42619, "epoch": 478, "lr": 9.415493762663666e-05}
{"train_loss": 0.27104952931404114, "global_step": 42620, "epoch": 478, "lr": 9.415466561869679e-05}
{"train_loss": 0.2469707429409027, "global_step": 42621, "epoch": 478, "lr": 9.415439360482088e-05}
{"train_loss": 0.32603195309638977, "global_step": 42622, "epoch": 478, "lr": 9.415412158500895e-05}
{"train_loss": 0.33620700240135193, "global_step": 42623, "epoch": 478, "lr": 9.415384955926103e-05}
{"train_loss": 0.33907297253608704, "global_step": 42624, "epoch": 478, "lr": 9.415357752757716e-05}
{"train_loss": 0.3504904508590698, "global_step": 42625, "epoch": 478, "lr": 9.41533054899574e-05}
{"train_loss": 0.41395512223243713, "global_step": 42626, "epoch": 478, "lr": 9.415303344640175e-05}
{"train_loss": 0.23897191882133484, "global_step": 42627, "epoch": 478, "lr": 9.415276139691027e-05}
{"train_loss": 0.3170640766620636, "global_step": 42628, "epoch": 478, "lr": 9.415248934148297e-05}
{"train_loss": 0.3749901354312897, "global_step": 42629, "epoch": 478, "lr": 9.415221728011994e-05}
{"train_loss": 0.29865840546200784, "global_step": 42630, "epoch": 478, "lr": 9.415194521282115e-05, "val_loss": 2.6679389476776123}
{"train_loss": 0.3824661076068878, "global_step": 42631, "epoch": 479, "lr": 9.415167313958669e-05}
{"train_loss": 0.27035850286483765, "global_step": 42632, "epoch": 479, "lr": 9.415140106041657e-05}
{"train_loss": 0.35024723410606384, "global_step": 42633, "epoch": 479, "lr": 9.415112897531082e-05}
{"train_loss": 0.3340602219104767, "global_step": 42634, "epoch": 479, "lr": 9.41508568842695e-05}
{"train_loss": 0.3145325779914856, "global_step": 42635, "epoch": 479, "lr": 9.415058478729264e-05}
{"train_loss": 0.2875129282474518, "global_step": 42636, "epoch": 479, "lr": 9.415031268438027e-05}
{"train_loss": 0.2811662554740906, "global_step": 42637, "epoch": 479, "lr": 9.415004057553243e-05}
{"train_loss": 0.3417319357395172, "global_step": 42638, "epoch": 479, "lr": 9.414976846074915e-05}
{"train_loss": 0.44512444734573364, "global_step": 42639, "epoch": 479, "lr": 9.414949634003047e-05}
{"train_loss": 0.33902812004089355, "global_step": 42640, "epoch": 479, "lr": 9.414922421337643e-05}
{"train_loss": 0.24325835704803467, "global_step": 42641, "epoch": 479, "lr": 9.414895208078707e-05}
{"train_loss": 0.3285364508628845, "global_step": 42642, "epoch": 479, "lr": 9.414867994226241e-05}
{"train_loss": 0.3269274830818176, "global_step": 42643, "epoch": 479, "lr": 9.41484077978025e-05}
{"train_loss": 0.21662290394306183, "global_step": 42644, "epoch": 479, "lr": 9.414813564740739e-05}
{"train_loss": 0.2759944498538971, "global_step": 42645, "epoch": 479, "lr": 9.41478634910771e-05}
{"train_loss": 0.3366427719593048, "global_step": 42646, "epoch": 479, "lr": 9.414759132881165e-05}
{"train_loss": 0.2728974223136902, "global_step": 42647, "epoch": 479, "lr": 9.41473191606111e-05}
{"train_loss": 0.3016808032989502, "global_step": 42648, "epoch": 479, "lr": 9.414704698647548e-05}
{"train_loss": 0.27038154006004333, "global_step": 42649, "epoch": 479, "lr": 9.414677480640484e-05}
{"train_loss": 0.2801503539085388, "global_step": 42650, "epoch": 479, "lr": 9.414650262039919e-05}
{"train_loss": 0.2381482571363449, "global_step": 42651, "epoch": 479, "lr": 9.414623042845858e-05}
{"train_loss": 0.3738982081413269, "global_step": 42652, "epoch": 479, "lr": 9.414595823058307e-05}
{"train_loss": 0.3164980709552765, "global_step": 42653, "epoch": 479, "lr": 9.414568602677265e-05}
{"train_loss": 0.29340341687202454, "global_step": 42654, "epoch": 479, "lr": 9.41454138170274e-05}
{"train_loss": 0.24732783436775208, "global_step": 42655, "epoch": 479, "lr": 9.414514160134732e-05}
{"train_loss": 0.26331955194473267, "global_step": 42656, "epoch": 479, "lr": 9.414486937973247e-05}
{"train_loss": 0.3267713189125061, "global_step": 42657, "epoch": 479, "lr": 9.414459715218288e-05}
{"train_loss": 0.29757964611053467, "global_step": 42658, "epoch": 479, "lr": 9.414432491869858e-05}
{"train_loss": 0.24761927127838135, "global_step": 42659, "epoch": 479, "lr": 9.414405267927962e-05}
{"train_loss": 0.23430894315242767, "global_step": 42660, "epoch": 479, "lr": 9.414378043392604e-05}
{"train_loss": 0.30463501811027527, "global_step": 42661, "epoch": 479, "lr": 9.414350818263785e-05}
{"train_loss": 0.3236438035964966, "global_step": 42662, "epoch": 479, "lr": 9.414323592541512e-05}
{"train_loss": 0.3659418821334839, "global_step": 42663, "epoch": 479, "lr": 9.414296366225785e-05}
{"train_loss": 0.2563445568084717, "global_step": 42664, "epoch": 479, "lr": 9.414269139316612e-05}
{"train_loss": 0.29173949360847473, "global_step": 42665, "epoch": 479, "lr": 9.414241911813993e-05}
{"train_loss": 0.250345915555954, "global_step": 42666, "epoch": 479, "lr": 9.414214683717932e-05}
{"train_loss": 0.3027239739894867, "global_step": 42667, "epoch": 479, "lr": 9.414187455028436e-05}
{"train_loss": 0.22941172122955322, "global_step": 42668, "epoch": 479, "lr": 9.414160225745504e-05}
{"train_loss": 0.2680434584617615, "global_step": 42669, "epoch": 479, "lr": 9.414132995869142e-05}
{"train_loss": 0.22496074438095093, "global_step": 42670, "epoch": 479, "lr": 9.414105765399356e-05}
{"train_loss": 0.3264319598674774, "global_step": 42671, "epoch": 479, "lr": 9.414078534336146e-05}
{"train_loss": 0.23977984488010406, "global_step": 42672, "epoch": 479, "lr": 9.414051302679516e-05}
{"train_loss": 0.24398207664489746, "global_step": 42673, "epoch": 479, "lr": 9.414024070429473e-05}
{"train_loss": 0.2717333137989044, "global_step": 42674, "epoch": 479, "lr": 9.413996837586016e-05}
{"train_loss": 0.2902545928955078, "global_step": 42675, "epoch": 479, "lr": 9.413969604149151e-05}
{"train_loss": 0.27173927426338196, "global_step": 42676, "epoch": 479, "lr": 9.413942370118883e-05}
{"train_loss": 0.25051823258399963, "global_step": 42677, "epoch": 479, "lr": 9.413915135495213e-05}
{"train_loss": 0.28465762734413147, "global_step": 42678, "epoch": 479, "lr": 9.413887900278146e-05}
{"train_loss": 0.34425798058509827, "global_step": 42679, "epoch": 479, "lr": 9.413860664467686e-05}
{"train_loss": 0.3026164770126343, "global_step": 42680, "epoch": 479, "lr": 9.413833428063837e-05}
{"train_loss": 0.26620444655418396, "global_step": 42681, "epoch": 479, "lr": 9.413806191066599e-05}
{"train_loss": 0.3891734182834625, "global_step": 42682, "epoch": 479, "lr": 9.41377895347598e-05}
{"train_loss": 0.3027395009994507, "global_step": 42683, "epoch": 479, "lr": 9.413751715291985e-05}
{"train_loss": 0.3480426073074341, "global_step": 42684, "epoch": 479, "lr": 9.413724476514611e-05}
{"train_loss": 0.3205218017101288, "global_step": 42685, "epoch": 479, "lr": 9.413697237143868e-05}
{"train_loss": 0.28010284900665283, "global_step": 42686, "epoch": 479, "lr": 9.413669997179757e-05}
{"train_loss": 0.25934457778930664, "global_step": 42687, "epoch": 479, "lr": 9.41364275662228e-05}
{"train_loss": 0.3456725478172302, "global_step": 42688, "epoch": 479, "lr": 9.413615515471444e-05}
{"train_loss": 0.3273952901363373, "global_step": 42689, "epoch": 479, "lr": 9.413588273727251e-05}
{"train_loss": 0.45061904191970825, "global_step": 42690, "epoch": 479, "lr": 9.413561031389706e-05}
{"train_loss": 0.36490774154663086, "global_step": 42691, "epoch": 479, "lr": 9.41353378845881e-05}
{"train_loss": 0.35528478026390076, "global_step": 42692, "epoch": 479, "lr": 9.413506544934568e-05}
{"train_loss": 0.317119836807251, "global_step": 42693, "epoch": 479, "lr": 9.413479300816985e-05}
{"train_loss": 0.26456883549690247, "global_step": 42694, "epoch": 479, "lr": 9.413452056106062e-05}
{"train_loss": 0.2837572693824768, "global_step": 42695, "epoch": 479, "lr": 9.413424810801805e-05}
{"train_loss": 0.3316444456577301, "global_step": 42696, "epoch": 479, "lr": 9.413397564904218e-05}
{"train_loss": 0.3479290008544922, "global_step": 42697, "epoch": 479, "lr": 9.413370318413302e-05}
{"train_loss": 0.28231123089790344, "global_step": 42698, "epoch": 479, "lr": 9.413343071329061e-05}
{"train_loss": 0.2573971450328827, "global_step": 42699, "epoch": 479, "lr": 9.413315823651502e-05}
{"train_loss": 0.3320619463920593, "global_step": 42700, "epoch": 479, "lr": 9.413288575380625e-05}
{"train_loss": 0.39649295806884766, "global_step": 42701, "epoch": 479, "lr": 9.413261326516437e-05}
{"train_loss": 0.43571069836616516, "global_step": 42702, "epoch": 479, "lr": 9.413234077058939e-05}
{"train_loss": 0.21330736577510834, "global_step": 42703, "epoch": 479, "lr": 9.413206827008135e-05}
{"train_loss": 0.35966384410858154, "global_step": 42704, "epoch": 479, "lr": 9.41317957636403e-05}
{"train_loss": 0.4421282708644867, "global_step": 42705, "epoch": 479, "lr": 9.413152325126626e-05}
{"train_loss": 0.2523243725299835, "global_step": 42706, "epoch": 479, "lr": 9.413125073295928e-05}
{"train_loss": 0.32464224100112915, "global_step": 42707, "epoch": 479, "lr": 9.413097820871938e-05}
{"train_loss": 0.24924840033054352, "global_step": 42708, "epoch": 479, "lr": 9.413070567854661e-05}
{"train_loss": 0.31336602568626404, "global_step": 42709, "epoch": 479, "lr": 9.4130433142441e-05}
{"train_loss": 0.25843676924705505, "global_step": 42710, "epoch": 479, "lr": 9.41301606004026e-05}
{"train_loss": 0.3854508101940155, "global_step": 42711, "epoch": 479, "lr": 9.412988805243146e-05}
{"train_loss": 0.33120787143707275, "global_step": 42712, "epoch": 479, "lr": 9.412961549852757e-05}
{"train_loss": 0.3063019812107086, "global_step": 42713, "epoch": 479, "lr": 9.412934293869099e-05}
{"train_loss": 0.3647403120994568, "global_step": 42714, "epoch": 479, "lr": 9.412907037292176e-05}
{"train_loss": 0.2616768479347229, "global_step": 42715, "epoch": 479, "lr": 9.412879780121992e-05}
{"train_loss": 0.263946533203125, "global_step": 42716, "epoch": 479, "lr": 9.41285252235855e-05}
{"train_loss": 0.310092568397522, "global_step": 42717, "epoch": 479, "lr": 9.412825264001852e-05}
{"train_loss": 0.2774260640144348, "global_step": 42718, "epoch": 479, "lr": 9.412798005051905e-05}
{"train_loss": 0.3049635244219491, "global_step": 42719, "epoch": 479, "lr": 9.412770745508712e-05, "val_loss": 2.7668511867523193}
{"train_loss": 0.30669543147087097, "global_step": 42720, "epoch": 480, "lr": 9.412743485372275e-05}
{"train_loss": 0.25210249423980713, "global_step": 42721, "epoch": 480, "lr": 9.412716224642597e-05}
{"train_loss": 0.23253649473190308, "global_step": 42722, "epoch": 480, "lr": 9.412688963319685e-05}
{"train_loss": 0.2000695914030075, "global_step": 42723, "epoch": 480, "lr": 9.41266170140354e-05}
{"train_loss": 0.3085639774799347, "global_step": 42724, "epoch": 480, "lr": 9.412634438894165e-05}
{"train_loss": 0.29331129789352417, "global_step": 42725, "epoch": 480, "lr": 9.412607175791568e-05}
{"train_loss": 0.2484108954668045, "global_step": 42726, "epoch": 480, "lr": 9.412579912095749e-05}
{"train_loss": 0.3010515868663788, "global_step": 42727, "epoch": 480, "lr": 9.41255264780671e-05}
{"train_loss": 0.24072065949440002, "global_step": 42728, "epoch": 480, "lr": 9.41252538292446e-05}
{"train_loss": 0.30139145255088806, "global_step": 42729, "epoch": 480, "lr": 9.412498117448999e-05}
{"train_loss": 0.27080416679382324, "global_step": 42730, "epoch": 480, "lr": 9.41247085138033e-05}
{"train_loss": 0.24446521699428558, "global_step": 42731, "epoch": 480, "lr": 9.412443584718459e-05}
{"train_loss": 0.31051555275917053, "global_step": 42732, "epoch": 480, "lr": 9.412416317463389e-05}
{"train_loss": 0.3374077081680298, "global_step": 42733, "epoch": 480, "lr": 9.412389049615122e-05}
{"train_loss": 0.2804294228553772, "global_step": 42734, "epoch": 480, "lr": 9.412361781173664e-05}
{"train_loss": 0.3709530532360077, "global_step": 42735, "epoch": 480, "lr": 9.412334512139019e-05}
{"train_loss": 0.2794790267944336, "global_step": 42736, "epoch": 480, "lr": 9.412307242511188e-05}
{"train_loss": 0.30083030462265015, "global_step": 42737, "epoch": 480, "lr": 9.412279972290177e-05}
{"train_loss": 0.33520299196243286, "global_step": 42738, "epoch": 480, "lr": 9.412252701475988e-05}
{"train_loss": 0.2240588366985321, "global_step": 42739, "epoch": 480, "lr": 9.412225430068624e-05}
{"train_loss": 0.3060392737388611, "global_step": 42740, "epoch": 480, "lr": 9.412198158068093e-05}
{"train_loss": 0.33236411213874817, "global_step": 42741, "epoch": 480, "lr": 9.412170885474393e-05}
{"train_loss": 0.3344534635543823, "global_step": 42742, "epoch": 480, "lr": 9.412143612287531e-05}
{"train_loss": 0.26334792375564575, "global_step": 42743, "epoch": 480, "lr": 9.412116338507512e-05}
{"train_loss": 0.3707598149776459, "global_step": 42744, "epoch": 480, "lr": 9.412089064134338e-05}
{"train_loss": 0.4000302255153656, "global_step": 42745, "epoch": 480, "lr": 9.41206178916801e-05}
{"train_loss": 0.26864340901374817, "global_step": 42746, "epoch": 480, "lr": 9.412034513608535e-05}
{"train_loss": 0.36206623911857605, "global_step": 42747, "epoch": 480, "lr": 9.412007237455916e-05}
{"train_loss": 0.31523340940475464, "global_step": 42748, "epoch": 480, "lr": 9.411979960710156e-05}
{"train_loss": 0.2905581295490265, "global_step": 42749, "epoch": 480, "lr": 9.411952683371259e-05}
{"train_loss": 0.27789220213890076, "global_step": 42750, "epoch": 480, "lr": 9.411925405439228e-05}
{"train_loss": 0.2410171777009964, "global_step": 42751, "epoch": 480, "lr": 9.41189812691407e-05}
{"train_loss": 0.22883391380310059, "global_step": 42752, "epoch": 480, "lr": 9.411870847795783e-05}
{"train_loss": 0.3869653046131134, "global_step": 42753, "epoch": 480, "lr": 9.411843568084375e-05}
{"train_loss": 0.26715192198753357, "global_step": 42754, "epoch": 480, "lr": 9.411816287779848e-05}
{"train_loss": 0.2912796139717102, "global_step": 42755, "epoch": 480, "lr": 9.411789006882208e-05}
{"train_loss": 0.31887105107307434, "global_step": 42756, "epoch": 480, "lr": 9.411761725391455e-05}
{"train_loss": 0.4236050248146057, "global_step": 42757, "epoch": 480, "lr": 9.411734443307595e-05}
{"train_loss": 0.21772707998752594, "global_step": 42758, "epoch": 480, "lr": 9.41170716063063e-05}
{"train_loss": 0.3498160243034363, "global_step": 42759, "epoch": 480, "lr": 9.411679877360567e-05}
{"train_loss": 0.2667769193649292, "global_step": 42760, "epoch": 480, "lr": 9.411652593497406e-05}
{"train_loss": 0.3319248855113983, "global_step": 42761, "epoch": 480, "lr": 9.411625309041152e-05}
{"train_loss": 0.289196640253067, "global_step": 42762, "epoch": 480, "lr": 9.411598023991807e-05}
{"train_loss": 0.28427553176879883, "global_step": 42763, "epoch": 480, "lr": 9.41157073834938e-05}
{"train_loss": 0.26323115825653076, "global_step": 42764, "epoch": 480, "lr": 9.411543452113869e-05}
{"train_loss": 0.252639502286911, "global_step": 42765, "epoch": 480, "lr": 9.41151616528528e-05}
{"train_loss": 0.307471364736557, "global_step": 42766, "epoch": 480, "lr": 9.411488877863616e-05}
{"train_loss": 0.28490111231803894, "global_step": 42767, "epoch": 480, "lr": 9.411461589848883e-05}
{"train_loss": 0.32491534948349, "global_step": 42768, "epoch": 480, "lr": 9.41143430124108e-05}
{"train_loss": 0.2859058082103729, "global_step": 42769, "epoch": 480, "lr": 9.411407012040215e-05}
{"train_loss": 0.35180073976516724, "global_step": 42770, "epoch": 480, "lr": 9.41137972224629e-05}
{"train_loss": 0.25482791662216187, "global_step": 42771, "epoch": 480, "lr": 9.411352431859308e-05}
{"train_loss": 0.26091843843460083, "global_step": 42772, "epoch": 480, "lr": 9.411325140879275e-05}
{"train_loss": 0.19706818461418152, "global_step": 42773, "epoch": 480, "lr": 9.411297849306191e-05}
{"train_loss": 0.33836784958839417, "global_step": 42774, "epoch": 480, "lr": 9.411270557140065e-05}
{"train_loss": 0.3414161503314972, "global_step": 42775, "epoch": 480, "lr": 9.411243264380893e-05}
{"train_loss": 0.37425658106803894, "global_step": 42776, "epoch": 480, "lr": 9.411215971028686e-05}
{"train_loss": 0.23266072571277618, "global_step": 42777, "epoch": 480, "lr": 9.411188677083445e-05}
{"train_loss": 0.3405912220478058, "global_step": 42778, "epoch": 480, "lr": 9.411161382545172e-05}
{"train_loss": 0.29547372460365295, "global_step": 42779, "epoch": 480, "lr": 9.411134087413874e-05}
{"train_loss": 0.239543154835701, "global_step": 42780, "epoch": 480, "lr": 9.411106791689552e-05}
{"train_loss": 0.33365464210510254, "global_step": 42781, "epoch": 480, "lr": 9.411079495372209e-05}
{"train_loss": 0.4667692482471466, "global_step": 42782, "epoch": 480, "lr": 9.411052198461851e-05}
{"train_loss": 0.2987666726112366, "global_step": 42783, "epoch": 480, "lr": 9.411024900958482e-05}
{"train_loss": 0.18548011779785156, "global_step": 42784, "epoch": 480, "lr": 9.410997602862103e-05}
{"train_loss": 0.2410092055797577, "global_step": 42785, "epoch": 480, "lr": 9.410970304172719e-05}
{"train_loss": 0.3495323359966278, "global_step": 42786, "epoch": 480, "lr": 9.410943004890335e-05}
{"train_loss": 0.24554292857646942, "global_step": 42787, "epoch": 480, "lr": 9.410915705014953e-05}
{"train_loss": 0.2531591057777405, "global_step": 42788, "epoch": 480, "lr": 9.410888404546577e-05}
{"train_loss": 0.4819888174533844, "global_step": 42789, "epoch": 480, "lr": 9.410861103485211e-05}
{"train_loss": 0.3961465060710907, "global_step": 42790, "epoch": 480, "lr": 9.410833801830859e-05}
{"train_loss": 0.33187928795814514, "global_step": 42791, "epoch": 480, "lr": 9.410806499583523e-05}
{"train_loss": 0.2882072925567627, "global_step": 42792, "epoch": 480, "lr": 9.410779196743206e-05}
{"train_loss": 0.35394126176834106, "global_step": 42793, "epoch": 480, "lr": 9.410751893309917e-05}
{"train_loss": 0.3197045624256134, "global_step": 42794, "epoch": 480, "lr": 9.410724589283655e-05}
{"train_loss": 0.21851101517677307, "global_step": 42795, "epoch": 480, "lr": 9.410697284664425e-05}
{"train_loss": 0.2644164562225342, "global_step": 42796, "epoch": 480, "lr": 9.410669979452231e-05}
{"train_loss": 0.2545306384563446, "global_step": 42797, "epoch": 480, "lr": 9.410642673647074e-05}
{"train_loss": 0.33707091212272644, "global_step": 42798, "epoch": 480, "lr": 9.410615367248963e-05}
{"train_loss": 0.2884095013141632, "global_step": 42799, "epoch": 480, "lr": 9.410588060257897e-05}
{"train_loss": 0.27922993898391724, "global_step": 42800, "epoch": 480, "lr": 9.41056075267388e-05}
{"train_loss": 0.3110003173351288, "global_step": 42801, "epoch": 480, "lr": 9.410533444496919e-05}
{"train_loss": 0.32354968786239624, "global_step": 42802, "epoch": 480, "lr": 9.410506135727013e-05}
{"train_loss": 0.303606778383255, "global_step": 42803, "epoch": 480, "lr": 9.41047882636417e-05}
{"train_loss": 0.34859228134155273, "global_step": 42804, "epoch": 480, "lr": 9.410451516408392e-05}
{"train_loss": 0.2949921786785126, "global_step": 42805, "epoch": 480, "lr": 9.410424205859682e-05}
{"train_loss": 0.2730029225349426, "global_step": 42806, "epoch": 480, "lr": 9.410396894718044e-05}
{"train_loss": 0.34005776047706604, "global_step": 42807, "epoch": 480, "lr": 9.410369582983482e-05}
{"train_loss": 0.3005507122934534, "global_step": 42808, "epoch": 480, "lr": 9.410342270656001e-05, "val_loss": 2.947449207305908, "train_action_mse_error": 30.00444793701172}
{"train_loss": 0.3405219316482544, "global_step": 42809, "epoch": 481, "lr": 9.410314957735601e-05}
{"train_loss": 0.3693472743034363, "global_step": 42810, "epoch": 481, "lr": 9.410287644222289e-05}
{"train_loss": 0.24702709913253784, "global_step": 42811, "epoch": 481, "lr": 9.410260330116067e-05}
{"train_loss": 0.27758482098579407, "global_step": 42812, "epoch": 481, "lr": 9.41023301541694e-05}
{"train_loss": 0.2483607828617096, "global_step": 42813, "epoch": 481, "lr": 9.410205700124911e-05}
{"train_loss": 0.41828086972236633, "global_step": 42814, "epoch": 481, "lr": 9.410178384239983e-05}
{"train_loss": 0.3966985046863556, "global_step": 42815, "epoch": 481, "lr": 9.410151067762161e-05}
{"train_loss": 0.3125946521759033, "global_step": 42816, "epoch": 481, "lr": 9.410123750691446e-05}
{"train_loss": 0.28755638003349304, "global_step": 42817, "epoch": 481, "lr": 9.410096433027847e-05}
{"train_loss": 0.28470760583877563, "global_step": 42818, "epoch": 481, "lr": 9.410069114771362e-05}
{"train_loss": 0.40593791007995605, "global_step": 42819, "epoch": 481, "lr": 9.410041795921998e-05}
{"train_loss": 0.27288809418678284, "global_step": 42820, "epoch": 481, "lr": 9.410014476479756e-05}
{"train_loss": 0.34435808658599854, "global_step": 42821, "epoch": 481, "lr": 9.409987156444642e-05}
{"train_loss": 0.3051193952560425, "global_step": 42822, "epoch": 481, "lr": 9.40995983581666e-05}
{"train_loss": 0.44674229621887207, "global_step": 42823, "epoch": 481, "lr": 9.409932514595811e-05}
{"train_loss": 0.37412029504776, "global_step": 42824, "epoch": 481, "lr": 9.409905192782102e-05}
{"train_loss": 0.3277644217014313, "global_step": 42825, "epoch": 481, "lr": 9.409877870375535e-05}
{"train_loss": 0.31820088624954224, "global_step": 42826, "epoch": 481, "lr": 9.40985054737611e-05}
{"train_loss": 0.2912534177303314, "global_step": 42827, "epoch": 481, "lr": 9.40982322378384e-05}
{"train_loss": 0.29628515243530273, "global_step": 42828, "epoch": 481, "lr": 9.409795899598719e-05}
{"train_loss": 0.22524669766426086, "global_step": 42829, "epoch": 481, "lr": 9.409768574820757e-05}
{"train_loss": 0.19772610068321228, "global_step": 42830, "epoch": 481, "lr": 9.409741249449953e-05}
{"train_loss": 0.30738818645477295, "global_step": 42831, "epoch": 481, "lr": 9.409713923486314e-05}
{"train_loss": 0.319296270608902, "global_step": 42832, "epoch": 481, "lr": 9.409686596929843e-05}
{"train_loss": 0.3339592516422272, "global_step": 42833, "epoch": 481, "lr": 9.409659269780544e-05}
{"train_loss": 0.28996676206588745, "global_step": 42834, "epoch": 481, "lr": 9.409631942038419e-05}
{"train_loss": 0.38515451550483704, "global_step": 42835, "epoch": 481, "lr": 9.409604613703473e-05}
{"train_loss": 0.27254581451416016, "global_step": 42836, "epoch": 481, "lr": 9.40957728477571e-05}
{"train_loss": 0.2739710509777069, "global_step": 42837, "epoch": 481, "lr": 9.409549955255133e-05}
{"train_loss": 0.318042129278183, "global_step": 42838, "epoch": 481, "lr": 9.409522625141746e-05}
{"train_loss": 0.3264242708683014, "global_step": 42839, "epoch": 481, "lr": 9.40949529443555e-05}
{"train_loss": 0.31047797203063965, "global_step": 42840, "epoch": 481, "lr": 9.409467963136555e-05}
{"train_loss": 0.2547729015350342, "global_step": 42841, "epoch": 481, "lr": 9.409440631244758e-05}
{"train_loss": 0.35715803503990173, "global_step": 42842, "epoch": 481, "lr": 9.409413298760165e-05}
{"train_loss": 0.23144134879112244, "global_step": 42843, "epoch": 481, "lr": 9.409385965682781e-05}
{"train_loss": 0.22318443655967712, "global_step": 42844, "epoch": 481, "lr": 9.40935863201261e-05}
{"train_loss": 0.3445989787578583, "global_step": 42845, "epoch": 481, "lr": 9.409331297749655e-05}
{"train_loss": 0.24390332400798798, "global_step": 42846, "epoch": 481, "lr": 9.409303962893918e-05}
{"train_loss": 0.2749309539794922, "global_step": 42847, "epoch": 481, "lr": 9.409276627445402e-05}
{"train_loss": 0.288076251745224, "global_step": 42848, "epoch": 481, "lr": 9.409249291404115e-05}
{"train_loss": 0.3483181595802307, "global_step": 42849, "epoch": 481, "lr": 9.409221954770058e-05}
{"train_loss": 0.3077373504638672, "global_step": 42850, "epoch": 481, "lr": 9.409194617543233e-05}
{"train_loss": 0.3749162256717682, "global_step": 42851, "epoch": 481, "lr": 9.409167279723648e-05}
{"train_loss": 0.21613596379756927, "global_step": 42852, "epoch": 481, "lr": 9.409139941311302e-05}
{"train_loss": 0.26065731048583984, "global_step": 42853, "epoch": 481, "lr": 9.409112602306201e-05}
{"train_loss": 0.3240198493003845, "global_step": 42854, "epoch": 481, "lr": 9.409085262708351e-05}
{"train_loss": 0.20313628017902374, "global_step": 42855, "epoch": 481, "lr": 9.40905792251775e-05}
{"train_loss": 0.26302045583724976, "global_step": 42856, "epoch": 481, "lr": 9.409030581734407e-05}
{"train_loss": 0.277538001537323, "global_step": 42857, "epoch": 481, "lr": 9.409003240358322e-05}
{"train_loss": 0.3049200177192688, "global_step": 42858, "epoch": 481, "lr": 9.408975898389502e-05}
{"train_loss": 0.28785476088523865, "global_step": 42859, "epoch": 481, "lr": 9.408948555827948e-05}
{"train_loss": 0.3500883877277374, "global_step": 42860, "epoch": 481, "lr": 9.408921212673665e-05}
{"train_loss": 0.3595614731311798, "global_step": 42861, "epoch": 481, "lr": 9.408893868926656e-05}
{"train_loss": 0.31043320894241333, "global_step": 42862, "epoch": 481, "lr": 9.408866524586924e-05}
{"train_loss": 0.26412224769592285, "global_step": 42863, "epoch": 481, "lr": 9.408839179654475e-05}
{"train_loss": 0.2765255272388458, "global_step": 42864, "epoch": 481, "lr": 9.40881183412931e-05}
{"train_loss": 0.26520588994026184, "global_step": 42865, "epoch": 481, "lr": 9.408784488011437e-05}
{"train_loss": 0.31425741314888, "global_step": 42866, "epoch": 481, "lr": 9.408757141300854e-05}
{"train_loss": 0.286981463432312, "global_step": 42867, "epoch": 481, "lr": 9.408729793997566e-05}
{"train_loss": 0.30790966749191284, "global_step": 42868, "epoch": 481, "lr": 9.408702446101581e-05}
{"train_loss": 0.2594228982925415, "global_step": 42869, "epoch": 481, "lr": 9.408675097612899e-05}
{"train_loss": 0.3679143488407135, "global_step": 42870, "epoch": 481, "lr": 9.408647748531524e-05}
{"train_loss": 0.30380409955978394, "global_step": 42871, "epoch": 481, "lr": 9.40862039885746e-05}
{"train_loss": 0.26260048151016235, "global_step": 42872, "epoch": 481, "lr": 9.40859304859071e-05}
{"train_loss": 0.3111873269081116, "global_step": 42873, "epoch": 481, "lr": 9.40856569773128e-05}
{"train_loss": 0.2904796302318573, "global_step": 42874, "epoch": 481, "lr": 9.40853834627917e-05}
{"train_loss": 0.22130483388900757, "global_step": 42875, "epoch": 481, "lr": 9.408510994234388e-05}
{"train_loss": 0.2916179299354553, "global_step": 42876, "epoch": 481, "lr": 9.408483641596935e-05}
{"train_loss": 0.23085762560367584, "global_step": 42877, "epoch": 481, "lr": 9.408456288366813e-05}
{"train_loss": 0.29335466027259827, "global_step": 42878, "epoch": 481, "lr": 9.40842893454403e-05}
{"train_loss": 0.2954869270324707, "global_step": 42879, "epoch": 481, "lr": 9.408401580128587e-05}
{"train_loss": 0.2909698486328125, "global_step": 42880, "epoch": 481, "lr": 9.408374225120488e-05}
{"train_loss": 0.4645872414112091, "global_step": 42881, "epoch": 481, "lr": 9.408346869519736e-05}
{"train_loss": 0.20639966428279877, "global_step": 42882, "epoch": 481, "lr": 9.408319513326336e-05}
{"train_loss": 0.21357986330986023, "global_step": 42883, "epoch": 481, "lr": 9.408292156540293e-05}
{"train_loss": 0.2263537347316742, "global_step": 42884, "epoch": 481, "lr": 9.408264799161605e-05}
{"train_loss": 0.2718375623226166, "global_step": 42885, "epoch": 481, "lr": 9.408237441190283e-05}
{"train_loss": 0.40444615483283997, "global_step": 42886, "epoch": 481, "lr": 9.408210082626326e-05}
{"train_loss": 0.36009112000465393, "global_step": 42887, "epoch": 481, "lr": 9.408182723469738e-05}
{"train_loss": 0.44680723547935486, "global_step": 42888, "epoch": 481, "lr": 9.408155363720525e-05}
{"train_loss": 0.285176157951355, "global_step": 42889, "epoch": 481, "lr": 9.408128003378689e-05}
{"train_loss": 0.31301742792129517, "global_step": 42890, "epoch": 481, "lr": 9.408100642444233e-05}
{"train_loss": 0.22318194806575775, "global_step": 42891, "epoch": 481, "lr": 9.408073280917163e-05}
{"train_loss": 0.4120565950870514, "global_step": 42892, "epoch": 481, "lr": 9.408045918797479e-05}
{"train_loss": 0.24473007023334503, "global_step": 42893, "epoch": 481, "lr": 9.40801855608519e-05}
{"train_loss": 0.26542550325393677, "global_step": 42894, "epoch": 481, "lr": 9.407991192780294e-05}
{"train_loss": 0.24239659309387207, "global_step": 42895, "epoch": 481, "lr": 9.407963828882799e-05}
{"train_loss": 0.28516918420791626, "global_step": 42896, "epoch": 481, "lr": 9.407936464392706e-05}
{"train_loss": 0.3012791686178593, "global_step": 42897, "epoch": 481, "lr": 9.40790909931002e-05, "val_loss": 2.7823410034179688}
{"train_loss": 0.33665764331817627, "global_step": 42898, "epoch": 482, "lr": 9.407881733634745e-05}
{"train_loss": 0.29692623019218445, "global_step": 42899, "epoch": 482, "lr": 9.407854367366882e-05}
{"train_loss": 0.382814884185791, "global_step": 42900, "epoch": 482, "lr": 9.407827000506438e-05}
{"train_loss": 0.3015917241573334, "global_step": 42901, "epoch": 482, "lr": 9.407799633053416e-05}
{"train_loss": 0.26407474279403687, "global_step": 42902, "epoch": 482, "lr": 9.407772265007818e-05}
{"train_loss": 0.2023947536945343, "global_step": 42903, "epoch": 482, "lr": 9.40774489636965e-05}
{"train_loss": 0.25923681259155273, "global_step": 42904, "epoch": 482, "lr": 9.407717527138913e-05}
{"train_loss": 0.25799036026000977, "global_step": 42905, "epoch": 482, "lr": 9.407690157315612e-05}
{"train_loss": 0.19957609474658966, "global_step": 42906, "epoch": 482, "lr": 9.407662786899753e-05}
{"train_loss": 0.332135945558548, "global_step": 42907, "epoch": 482, "lr": 9.407635415891336e-05}
{"train_loss": 0.4376147389411926, "global_step": 42908, "epoch": 482, "lr": 9.407608044290365e-05}
{"train_loss": 0.2930007874965668, "global_step": 42909, "epoch": 482, "lr": 9.407580672096847e-05}
{"train_loss": 0.30824515223503113, "global_step": 42910, "epoch": 482, "lr": 9.407553299310782e-05}
{"train_loss": 0.24171485006809235, "global_step": 42911, "epoch": 482, "lr": 9.407525925932176e-05}
{"train_loss": 0.402590811252594, "global_step": 42912, "epoch": 482, "lr": 9.407498551961032e-05}
{"train_loss": 0.32921111583709717, "global_step": 42913, "epoch": 482, "lr": 9.407471177397352e-05}
{"train_loss": 0.3104615807533264, "global_step": 42914, "epoch": 482, "lr": 9.407443802241143e-05}
{"train_loss": 0.29040029644966125, "global_step": 42915, "epoch": 482, "lr": 9.407416426492406e-05}
{"train_loss": 0.330194890499115, "global_step": 42916, "epoch": 482, "lr": 9.407389050151146e-05}
{"train_loss": 0.29718267917633057, "global_step": 42917, "epoch": 482, "lr": 9.407361673217365e-05}
{"train_loss": 0.3367057740688324, "global_step": 42918, "epoch": 482, "lr": 9.407334295691069e-05}
{"train_loss": 0.3150002360343933, "global_step": 42919, "epoch": 482, "lr": 9.407306917572261e-05}
{"train_loss": 0.30335739254951477, "global_step": 42920, "epoch": 482, "lr": 9.407279538860943e-05}
{"train_loss": 0.33584439754486084, "global_step": 42921, "epoch": 482, "lr": 9.407252159557121e-05}
{"train_loss": 0.3917080760002136, "global_step": 42922, "epoch": 482, "lr": 9.407224779660798e-05}
{"train_loss": 0.3775155246257782, "global_step": 42923, "epoch": 482, "lr": 9.407197399171977e-05}
{"train_loss": 0.33982759714126587, "global_step": 42924, "epoch": 482, "lr": 9.407170018090661e-05}
{"train_loss": 0.2682681679725647, "global_step": 42925, "epoch": 482, "lr": 9.407142636416855e-05}
{"train_loss": 0.3653494119644165, "global_step": 42926, "epoch": 482, "lr": 9.407115254150565e-05}
{"train_loss": 0.35718291997909546, "global_step": 42927, "epoch": 482, "lr": 9.407087871291788e-05}
{"train_loss": 0.3149493932723999, "global_step": 42928, "epoch": 482, "lr": 9.407060487840535e-05}
{"train_loss": 0.4277433454990387, "global_step": 42929, "epoch": 482, "lr": 9.407033103796805e-05}
{"train_loss": 0.28985336422920227, "global_step": 42930, "epoch": 482, "lr": 9.407005719160604e-05}
{"train_loss": 0.32860928773880005, "global_step": 42931, "epoch": 482, "lr": 9.406978333931932e-05}
{"train_loss": 0.2512248158454895, "global_step": 42932, "epoch": 482, "lr": 9.4069509481108e-05}
{"train_loss": 0.22425922751426697, "global_step": 42933, "epoch": 482, "lr": 9.406923561697204e-05}
{"train_loss": 0.23080328106880188, "global_step": 42934, "epoch": 482, "lr": 9.406896174691152e-05}
{"train_loss": 0.23263958096504211, "global_step": 42935, "epoch": 482, "lr": 9.406868787092646e-05}
{"train_loss": 0.3225483000278473, "global_step": 42936, "epoch": 482, "lr": 9.40684139890169e-05}
{"train_loss": 0.29280176758766174, "global_step": 42937, "epoch": 482, "lr": 9.406814010118287e-05}
{"train_loss": 0.3274325728416443, "global_step": 42938, "epoch": 482, "lr": 9.406786620742444e-05}
{"train_loss": 0.2646048665046692, "global_step": 42939, "epoch": 482, "lr": 9.40675923077416e-05}
{"train_loss": 0.3377035856246948, "global_step": 42940, "epoch": 482, "lr": 9.406731840213443e-05}
{"train_loss": 0.4026792347431183, "global_step": 42941, "epoch": 482, "lr": 9.406704449060293e-05}
{"train_loss": 0.2654029428958893, "global_step": 42942, "epoch": 482, "lr": 9.406677057314716e-05}
{"train_loss": 0.22060152888298035, "global_step": 42943, "epoch": 482, "lr": 9.406649664976715e-05}
{"train_loss": 0.2910695970058441, "global_step": 42944, "epoch": 482, "lr": 9.406622272046294e-05}
{"train_loss": 0.2788909673690796, "global_step": 42945, "epoch": 482, "lr": 9.406594878523456e-05}
{"train_loss": 0.29522863030433655, "global_step": 42946, "epoch": 482, "lr": 9.406567484408204e-05}
{"train_loss": 0.2581125497817993, "global_step": 42947, "epoch": 482, "lr": 9.406540089700544e-05}
{"train_loss": 0.3638985753059387, "global_step": 42948, "epoch": 482, "lr": 9.406512694400478e-05}
{"train_loss": 0.23953846096992493, "global_step": 42949, "epoch": 482, "lr": 9.40648529850801e-05}
{"train_loss": 0.257283478975296, "global_step": 42950, "epoch": 482, "lr": 9.406457902023145e-05}
{"train_loss": 0.37767404317855835, "global_step": 42951, "epoch": 482, "lr": 9.406430504945884e-05}
{"train_loss": 0.2738149166107178, "global_step": 42952, "epoch": 482, "lr": 9.406403107276231e-05}
{"train_loss": 0.3352981507778168, "global_step": 42953, "epoch": 482, "lr": 9.406375709014192e-05}
{"train_loss": 0.3269751965999603, "global_step": 42954, "epoch": 482, "lr": 9.40634831015977e-05}
{"train_loss": 0.31188058853149414, "global_step": 42955, "epoch": 482, "lr": 9.40632091071297e-05}
{"train_loss": 0.29328471422195435, "global_step": 42956, "epoch": 482, "lr": 9.40629351067379e-05}
{"train_loss": 0.35757604241371155, "global_step": 42957, "epoch": 482, "lr": 9.406266110042239e-05}
{"train_loss": 0.32610660791397095, "global_step": 42958, "epoch": 482, "lr": 9.40623870881832e-05}
{"train_loss": 0.31220895051956177, "global_step": 42959, "epoch": 482, "lr": 9.406211307002036e-05}
{"train_loss": 0.2926987409591675, "global_step": 42960, "epoch": 482, "lr": 9.406183904593388e-05}
{"train_loss": 0.3405960202217102, "global_step": 42961, "epoch": 482, "lr": 9.406156501592385e-05}
{"train_loss": 0.33952879905700684, "global_step": 42962, "epoch": 482, "lr": 9.406129097999029e-05}
{"train_loss": 0.36313098669052124, "global_step": 42963, "epoch": 482, "lr": 9.40610169381332e-05}
{"train_loss": 0.3275674879550934, "global_step": 42964, "epoch": 482, "lr": 9.406074289035265e-05}
{"train_loss": 0.3819158673286438, "global_step": 42965, "epoch": 482, "lr": 9.406046883664868e-05}
{"train_loss": 0.35623660683631897, "global_step": 42966, "epoch": 482, "lr": 9.40601947770213e-05}
{"train_loss": 0.38069388270378113, "global_step": 42967, "epoch": 482, "lr": 9.405992071147058e-05}
{"train_loss": 0.24460762739181519, "global_step": 42968, "epoch": 482, "lr": 9.405964663999653e-05}
{"train_loss": 0.28001725673675537, "global_step": 42969, "epoch": 482, "lr": 9.405937256259921e-05}
{"train_loss": 0.419036328792572, "global_step": 42970, "epoch": 482, "lr": 9.405909847927863e-05}
{"train_loss": 0.20493732392787933, "global_step": 42971, "epoch": 482, "lr": 9.405882439003483e-05}
{"train_loss": 0.1970439851284027, "global_step": 42972, "epoch": 482, "lr": 9.405855029486789e-05}
{"train_loss": 0.3398389518260956, "global_step": 42973, "epoch": 482, "lr": 9.405827619377781e-05}
{"train_loss": 0.3656725287437439, "global_step": 42974, "epoch": 482, "lr": 9.405800208676461e-05}
{"train_loss": 0.2164413332939148, "global_step": 42975, "epoch": 482, "lr": 9.405772797382836e-05}
{"train_loss": 0.4159744381904602, "global_step": 42976, "epoch": 482, "lr": 9.40574538549691e-05}
{"train_loss": 0.28440338373184204, "global_step": 42977, "epoch": 482, "lr": 9.405717973018681e-05}
{"train_loss": 0.3011115491390228, "global_step": 42978, "epoch": 482, "lr": 9.405690559948162e-05}
{"train_loss": 0.3167714774608612, "global_step": 42979, "epoch": 482, "lr": 9.405663146285348e-05}
{"train_loss": 0.32205265760421753, "global_step": 42980, "epoch": 482, "lr": 9.405635732030248e-05}
{"train_loss": 0.2663428783416748, "global_step": 42981, "epoch": 482, "lr": 9.405608317182863e-05}
{"train_loss": 0.3223017156124115, "global_step": 42982, "epoch": 482, "lr": 9.405580901743198e-05}
{"train_loss": 0.30011728405952454, "global_step": 42983, "epoch": 482, "lr": 9.405553485711259e-05}
{"train_loss": 0.3577217161655426, "global_step": 42984, "epoch": 482, "lr": 9.405526069087043e-05}
{"train_loss": 0.17581383883953094, "global_step": 42985, "epoch": 482, "lr": 9.405498651870558e-05}
{"train_loss": 0.3085509843370888, "global_step": 42986, "epoch": 482, "lr": 9.40547123406181e-05, "val_loss": 2.7515718936920166}
{"train_loss": 0.29406386613845825, "global_step": 42987, "epoch": 483, "lr": 9.405443815660799e-05}
{"train_loss": 0.1896142065525055, "global_step": 42988, "epoch": 483, "lr": 9.40541639666753e-05}
{"train_loss": 0.28395479917526245, "global_step": 42989, "epoch": 483, "lr": 9.405388977082005e-05}
{"train_loss": 0.2857872247695923, "global_step": 42990, "epoch": 483, "lr": 9.40536155690423e-05}
{"train_loss": 0.2395753562450409, "global_step": 42991, "epoch": 483, "lr": 9.40533413613421e-05}
{"train_loss": 0.24456393718719482, "global_step": 42992, "epoch": 483, "lr": 9.405306714771943e-05}
{"train_loss": 0.3434300422668457, "global_step": 42993, "epoch": 483, "lr": 9.405279292817439e-05}
{"train_loss": 0.3305663764476776, "global_step": 42994, "epoch": 483, "lr": 9.405251870270697e-05}
{"train_loss": 0.374965101480484, "global_step": 42995, "epoch": 483, "lr": 9.405224447131723e-05}
{"train_loss": 0.3704133629798889, "global_step": 42996, "epoch": 483, "lr": 9.405197023400521e-05}
{"train_loss": 0.2764098048210144, "global_step": 42997, "epoch": 483, "lr": 9.405169599077093e-05}
{"train_loss": 0.36921826004981995, "global_step": 42998, "epoch": 483, "lr": 9.405142174161445e-05}
{"train_loss": 0.305705726146698, "global_step": 42999, "epoch": 483, "lr": 9.405114748653577e-05}
{"train_loss": 0.31944626569747925, "global_step": 43000, "epoch": 483, "lr": 9.405087322553497e-05}
{"train_loss": 0.34182295203208923, "global_step": 43001, "epoch": 483, "lr": 9.405059895861206e-05}
{"train_loss": 0.3204267919063568, "global_step": 43002, "epoch": 483, "lr": 9.405032468576709e-05}
{"train_loss": 0.37106508016586304, "global_step": 43003, "epoch": 483, "lr": 9.405005040700008e-05}
{"train_loss": 0.3735572397708893, "global_step": 43004, "epoch": 483, "lr": 9.404977612231108e-05}
{"train_loss": 0.28744029998779297, "global_step": 43005, "epoch": 483, "lr": 9.404950183170014e-05}
{"train_loss": 0.3346805274486542, "global_step": 43006, "epoch": 483, "lr": 9.404922753516726e-05}
{"train_loss": 0.2905237078666687, "global_step": 43007, "epoch": 483, "lr": 9.404895323271251e-05}
{"train_loss": 0.2951291799545288, "global_step": 43008, "epoch": 483, "lr": 9.40486789243359e-05}
{"train_loss": 0.24659815430641174, "global_step": 43009, "epoch": 483, "lr": 9.40484046100375e-05}
{"train_loss": 0.309190958738327, "global_step": 43010, "epoch": 483, "lr": 9.404813028981734e-05}
{"train_loss": 0.2685858905315399, "global_step": 43011, "epoch": 483, "lr": 9.404785596367543e-05}
{"train_loss": 0.49480631947517395, "global_step": 43012, "epoch": 483, "lr": 9.404758163161181e-05}
{"train_loss": 0.42458146810531616, "global_step": 43013, "epoch": 483, "lr": 9.404730729362654e-05}
{"train_loss": 0.5309507250785828, "global_step": 43014, "epoch": 483, "lr": 9.404703294971964e-05}
{"train_loss": 0.3374035656452179, "global_step": 43015, "epoch": 483, "lr": 9.404675859989117e-05}
{"train_loss": 0.36185961961746216, "global_step": 43016, "epoch": 483, "lr": 9.404648424414114e-05}
{"train_loss": 0.19537678360939026, "global_step": 43017, "epoch": 483, "lr": 9.40462098824696e-05}
{"train_loss": 0.325700044631958, "global_step": 43018, "epoch": 483, "lr": 9.40459355148766e-05}
{"train_loss": 0.2870965301990509, "global_step": 43019, "epoch": 483, "lr": 9.404566114136213e-05}
{"train_loss": 0.3448903262615204, "global_step": 43020, "epoch": 483, "lr": 9.404538676192626e-05}
{"train_loss": 0.2906011939048767, "global_step": 43021, "epoch": 483, "lr": 9.404511237656904e-05}
{"train_loss": 0.35113847255706787, "global_step": 43022, "epoch": 483, "lr": 9.404483798529049e-05}
{"train_loss": 0.34507429599761963, "global_step": 43023, "epoch": 483, "lr": 9.404456358809063e-05}
{"train_loss": 0.2731432616710663, "global_step": 43024, "epoch": 483, "lr": 9.404428918496955e-05}
{"train_loss": 0.28670164942741394, "global_step": 43025, "epoch": 483, "lr": 9.404401477592724e-05}
{"train_loss": 0.3313300609588623, "global_step": 43026, "epoch": 483, "lr": 9.404374036096373e-05}
{"train_loss": 0.3389361500740051, "global_step": 43027, "epoch": 483, "lr": 9.404346594007909e-05}
{"train_loss": 0.4081767201423645, "global_step": 43028, "epoch": 483, "lr": 9.404319151327335e-05}
{"train_loss": 0.3302328288555145, "global_step": 43029, "epoch": 483, "lr": 9.404291708054652e-05}
{"train_loss": 0.3467698097229004, "global_step": 43030, "epoch": 483, "lr": 9.404264264189866e-05}
{"train_loss": 0.315097838640213, "global_step": 43031, "epoch": 483, "lr": 9.404236819732981e-05}
{"train_loss": 0.3516022264957428, "global_step": 43032, "epoch": 483, "lr": 9.404209374684001e-05}
{"train_loss": 0.19755996763706207, "global_step": 43033, "epoch": 483, "lr": 9.404181929042928e-05}
{"train_loss": 0.3883804976940155, "global_step": 43034, "epoch": 483, "lr": 9.404154482809767e-05}
{"train_loss": 0.34525346755981445, "global_step": 43035, "epoch": 483, "lr": 9.404127035984519e-05}
{"train_loss": 0.2240906059741974, "global_step": 43036, "epoch": 483, "lr": 9.404099588567192e-05}
{"train_loss": 0.35628417134284973, "global_step": 43037, "epoch": 483, "lr": 9.404072140557785e-05}
{"train_loss": 0.32553723454475403, "global_step": 43038, "epoch": 483, "lr": 9.404044691956306e-05}
{"train_loss": 0.19253334403038025, "global_step": 43039, "epoch": 483, "lr": 9.404017242762757e-05}
{"train_loss": 0.25494158267974854, "global_step": 43040, "epoch": 483, "lr": 9.403989792977142e-05}
{"train_loss": 0.25283128023147583, "global_step": 43041, "epoch": 483, "lr": 9.403962342599462e-05}
{"train_loss": 0.24176819622516632, "global_step": 43042, "epoch": 483, "lr": 9.403934891629725e-05}
{"train_loss": 0.30730554461479187, "global_step": 43043, "epoch": 483, "lr": 9.403907440067931e-05}
{"train_loss": 0.30446991324424744, "global_step": 43044, "epoch": 483, "lr": 9.403879987914088e-05}
{"train_loss": 0.355659544467926, "global_step": 43045, "epoch": 483, "lr": 9.403852535168195e-05}
{"train_loss": 0.3350353538990021, "global_step": 43046, "epoch": 483, "lr": 9.403825081830258e-05}
{"train_loss": 0.3151490390300751, "global_step": 43047, "epoch": 483, "lr": 9.40379762790028e-05}
{"train_loss": 0.2824990153312683, "global_step": 43048, "epoch": 483, "lr": 9.403770173378266e-05}
{"train_loss": 0.34423545002937317, "global_step": 43049, "epoch": 483, "lr": 9.403742718264219e-05}
{"train_loss": 0.26972687244415283, "global_step": 43050, "epoch": 483, "lr": 9.403715262558141e-05}
{"train_loss": 0.28797653317451477, "global_step": 43051, "epoch": 483, "lr": 9.403687806260038e-05}
{"train_loss": 0.23605652153491974, "global_step": 43052, "epoch": 483, "lr": 9.403660349369912e-05}
{"train_loss": 0.2805565297603607, "global_step": 43053, "epoch": 483, "lr": 9.40363289188777e-05}
{"train_loss": 0.4129769206047058, "global_step": 43054, "epoch": 483, "lr": 9.403605433813611e-05}
{"train_loss": 0.41788575053215027, "global_step": 43055, "epoch": 483, "lr": 9.403577975147441e-05}
{"train_loss": 0.2006743997335434, "global_step": 43056, "epoch": 483, "lr": 9.403550515889264e-05}
{"train_loss": 0.26824894547462463, "global_step": 43057, "epoch": 483, "lr": 9.403523056039084e-05}
{"train_loss": 0.355412095785141, "global_step": 43058, "epoch": 483, "lr": 9.403495595596903e-05}
{"train_loss": 0.2954021990299225, "global_step": 43059, "epoch": 483, "lr": 9.403468134562726e-05}
{"train_loss": 0.30563461780548096, "global_step": 43060, "epoch": 483, "lr": 9.403440672936556e-05}
{"train_loss": 0.309735506772995, "global_step": 43061, "epoch": 483, "lr": 9.403413210718399e-05}
{"train_loss": 0.3661561906337738, "global_step": 43062, "epoch": 483, "lr": 9.403385747908254e-05}
{"train_loss": 0.21989110112190247, "global_step": 43063, "epoch": 483, "lr": 9.40335828450613e-05}
{"train_loss": 0.26058951020240784, "global_step": 43064, "epoch": 483, "lr": 9.403330820512026e-05}
{"train_loss": 0.33518385887145996, "global_step": 43065, "epoch": 483, "lr": 9.40330335592595e-05}
{"train_loss": 0.26742884516716003, "global_step": 43066, "epoch": 483, "lr": 9.403275890747901e-05}
{"train_loss": 0.3550746738910675, "global_step": 43067, "epoch": 483, "lr": 9.403248424977886e-05}
{"train_loss": 0.3993898928165436, "global_step": 43068, "epoch": 483, "lr": 9.40322095861591e-05}
{"train_loss": 0.2632785439491272, "global_step": 43069, "epoch": 483, "lr": 9.403193491661973e-05}
{"train_loss": 0.35498300194740295, "global_step": 43070, "epoch": 483, "lr": 9.40316602411608e-05}
{"train_loss": 0.24852490425109863, "global_step": 43071, "epoch": 483, "lr": 9.403138555978235e-05}
{"train_loss": 0.3101135194301605, "global_step": 43072, "epoch": 483, "lr": 9.403111087248442e-05}
{"train_loss": 0.22747614979743958, "global_step": 43073, "epoch": 483, "lr": 9.403083617926704e-05}
{"train_loss": 0.2707905173301697, "global_step": 43074, "epoch": 483, "lr": 9.403056148013026e-05}
{"train_loss": 0.3112449232447013, "global_step": 43075, "epoch": 483, "lr": 9.403028677507411e-05, "val_loss": 2.836364507675171}
{"train_loss": 0.29799583554267883, "global_step": 43076, "epoch": 484, "lr": 9.40300120640986e-05}
{"train_loss": 0.2936321794986725, "global_step": 43077, "epoch": 484, "lr": 9.402973734720382e-05}
{"train_loss": 0.3295082151889801, "global_step": 43078, "epoch": 484, "lr": 9.402946262438977e-05}
{"train_loss": 0.24491342902183533, "global_step": 43079, "epoch": 484, "lr": 9.402918789565649e-05}
{"train_loss": 0.3134220838546753, "global_step": 43080, "epoch": 484, "lr": 9.402891316100402e-05}
{"train_loss": 0.4010487198829651, "global_step": 43081, "epoch": 484, "lr": 9.40286384204324e-05}
{"train_loss": 0.3079378604888916, "global_step": 43082, "epoch": 484, "lr": 9.402836367394167e-05}
{"train_loss": 0.34002166986465454, "global_step": 43083, "epoch": 484, "lr": 9.402808892153187e-05}
{"train_loss": 0.17490899562835693, "global_step": 43084, "epoch": 484, "lr": 9.402781416320301e-05}
{"train_loss": 0.3417513966560364, "global_step": 43085, "epoch": 484, "lr": 9.402753939895516e-05}
{"train_loss": 0.2513570487499237, "global_step": 43086, "epoch": 484, "lr": 9.402726462878835e-05}
{"train_loss": 0.42090269923210144, "global_step": 43087, "epoch": 484, "lr": 9.402698985270262e-05}
{"train_loss": 0.33686354756355286, "global_step": 43088, "epoch": 484, "lr": 9.402671507069798e-05}
{"train_loss": 0.26457372307777405, "global_step": 43089, "epoch": 484, "lr": 9.402644028277447e-05}
{"train_loss": 0.2818547487258911, "global_step": 43090, "epoch": 484, "lr": 9.402616548893217e-05}
{"train_loss": 0.2832450568675995, "global_step": 43091, "epoch": 484, "lr": 9.402589068917108e-05}
{"train_loss": 0.33088356256484985, "global_step": 43092, "epoch": 484, "lr": 9.402561588349124e-05}
{"train_loss": 0.3279419243335724, "global_step": 43093, "epoch": 484, "lr": 9.402534107189269e-05}
{"train_loss": 0.26884496212005615, "global_step": 43094, "epoch": 484, "lr": 9.402506625437549e-05}
{"train_loss": 0.39091619849205017, "global_step": 43095, "epoch": 484, "lr": 9.402479143093963e-05}
{"train_loss": 0.33538514375686646, "global_step": 43096, "epoch": 484, "lr": 9.402451660158519e-05}
{"train_loss": 0.25751352310180664, "global_step": 43097, "epoch": 484, "lr": 9.402424176631219e-05}
{"train_loss": 0.2675681710243225, "global_step": 43098, "epoch": 484, "lr": 9.402396692512065e-05}
{"train_loss": 0.2927241623401642, "global_step": 43099, "epoch": 484, "lr": 9.402369207801065e-05}
{"train_loss": 0.17750951647758484, "global_step": 43100, "epoch": 484, "lr": 9.402341722498217e-05}
{"train_loss": 0.3977665901184082, "global_step": 43101, "epoch": 484, "lr": 9.40231423660353e-05}
{"train_loss": 0.3339797854423523, "global_step": 43102, "epoch": 484, "lr": 9.402286750117006e-05}
{"train_loss": 0.3654865324497223, "global_step": 43103, "epoch": 484, "lr": 9.402259263038648e-05}
{"train_loss": 0.36626243591308594, "global_step": 43104, "epoch": 484, "lr": 9.40223177536846e-05}
{"train_loss": 0.33239713311195374, "global_step": 43105, "epoch": 484, "lr": 9.402204287106445e-05}
{"train_loss": 0.3378029763698578, "global_step": 43106, "epoch": 484, "lr": 9.402176798252607e-05}
{"train_loss": 0.25865069031715393, "global_step": 43107, "epoch": 484, "lr": 9.40214930880695e-05}
{"train_loss": 0.3808577060699463, "global_step": 43108, "epoch": 484, "lr": 9.402121818769477e-05}
{"train_loss": 0.2499922662973404, "global_step": 43109, "epoch": 484, "lr": 9.402094328140195e-05}
{"train_loss": 0.3971025347709656, "global_step": 43110, "epoch": 484, "lr": 9.402066836919102e-05}
{"train_loss": 0.3075888156890869, "global_step": 43111, "epoch": 484, "lr": 9.402039345106207e-05}
{"train_loss": 0.3620106875896454, "global_step": 43112, "epoch": 484, "lr": 9.402011852701509e-05}
{"train_loss": 0.3357229232788086, "global_step": 43113, "epoch": 484, "lr": 9.401984359705016e-05}
{"train_loss": 0.2549421191215515, "global_step": 43114, "epoch": 484, "lr": 9.40195686611673e-05}
{"train_loss": 0.4218057096004486, "global_step": 43115, "epoch": 484, "lr": 9.401929371936653e-05}
{"train_loss": 0.32616764307022095, "global_step": 43116, "epoch": 484, "lr": 9.401901877164791e-05}
{"train_loss": 0.29877346754074097, "global_step": 43117, "epoch": 484, "lr": 9.401874381801148e-05}
{"train_loss": 0.2610092759132385, "global_step": 43118, "epoch": 484, "lr": 9.401846885845726e-05}
{"train_loss": 0.30221158266067505, "global_step": 43119, "epoch": 484, "lr": 9.401819389298529e-05}
{"train_loss": 0.36977246403694153, "global_step": 43120, "epoch": 484, "lr": 9.401791892159561e-05}
{"train_loss": 0.38779717683792114, "global_step": 43121, "epoch": 484, "lr": 9.401764394428826e-05}
{"train_loss": 0.29212552309036255, "global_step": 43122, "epoch": 484, "lr": 9.401736896106329e-05}
{"train_loss": 0.29190272092819214, "global_step": 43123, "epoch": 484, "lr": 9.401709397192069e-05}
{"train_loss": 0.23788957297801971, "global_step": 43124, "epoch": 484, "lr": 9.401681897686055e-05}
{"train_loss": 0.3665464520454407, "global_step": 43125, "epoch": 484, "lr": 9.401654397588288e-05}
{"train_loss": 0.23184821009635925, "global_step": 43126, "epoch": 484, "lr": 9.401626896898772e-05}
{"train_loss": 0.3610506057739258, "global_step": 43127, "epoch": 484, "lr": 9.401599395617511e-05}
{"train_loss": 0.3086986243724823, "global_step": 43128, "epoch": 484, "lr": 9.401571893744509e-05}
{"train_loss": 0.3523244261741638, "global_step": 43129, "epoch": 484, "lr": 9.401544391279769e-05}
{"train_loss": 0.27054014801979065, "global_step": 43130, "epoch": 484, "lr": 9.401516888223295e-05}
{"train_loss": 0.20788493752479553, "global_step": 43131, "epoch": 484, "lr": 9.401489384575091e-05}
{"train_loss": 0.2227299064397812, "global_step": 43132, "epoch": 484, "lr": 9.401461880335159e-05}
{"train_loss": 0.23160193860530853, "global_step": 43133, "epoch": 484, "lr": 9.401434375503506e-05}
{"train_loss": 0.2774553894996643, "global_step": 43134, "epoch": 484, "lr": 9.401406870080131e-05}
{"train_loss": 0.4070642292499542, "global_step": 43135, "epoch": 484, "lr": 9.401379364065044e-05}
{"train_loss": 0.3931118845939636, "global_step": 43136, "epoch": 484, "lr": 9.401351857458243e-05}
{"train_loss": 0.3691450357437134, "global_step": 43137, "epoch": 484, "lr": 9.401324350259735e-05}
{"train_loss": 0.3415406346321106, "global_step": 43138, "epoch": 484, "lr": 9.401296842469522e-05}
{"train_loss": 0.2819828987121582, "global_step": 43139, "epoch": 484, "lr": 9.401269334087608e-05}
{"train_loss": 0.27425745129585266, "global_step": 43140, "epoch": 484, "lr": 9.401241825113997e-05}
{"train_loss": 0.36851972341537476, "global_step": 43141, "epoch": 484, "lr": 9.401214315548694e-05}
{"train_loss": 0.3430822491645813, "global_step": 43142, "epoch": 484, "lr": 9.4011868053917e-05}
{"train_loss": 0.29280146956443787, "global_step": 43143, "epoch": 484, "lr": 9.40115929464302e-05}
{"train_loss": 0.33864012360572815, "global_step": 43144, "epoch": 484, "lr": 9.401131783302659e-05}
{"train_loss": 0.31589096784591675, "global_step": 43145, "epoch": 484, "lr": 9.401104271370618e-05}
{"train_loss": 0.24660208821296692, "global_step": 43146, "epoch": 484, "lr": 9.401076758846903e-05}
{"train_loss": 0.33571940660476685, "global_step": 43147, "epoch": 484, "lr": 9.401049245731516e-05}
{"train_loss": 0.3246966302394867, "global_step": 43148, "epoch": 484, "lr": 9.401021732024465e-05}
{"train_loss": 0.39514684677124023, "global_step": 43149, "epoch": 484, "lr": 9.400994217725746e-05}
{"train_loss": 0.32462289929389954, "global_step": 43150, "epoch": 484, "lr": 9.400966702835368e-05}
{"train_loss": 0.3810771107673645, "global_step": 43151, "epoch": 484, "lr": 9.400939187353335e-05}
{"train_loss": 0.38203778862953186, "global_step": 43152, "epoch": 484, "lr": 9.400911671279649e-05}
{"train_loss": 0.23048710823059082, "global_step": 43153, "epoch": 484, "lr": 9.400884154614314e-05}
{"train_loss": 0.3097536265850067, "global_step": 43154, "epoch": 484, "lr": 9.400856637357332e-05}
{"train_loss": 0.28730276226997375, "global_step": 43155, "epoch": 484, "lr": 9.400829119508711e-05}
{"train_loss": 0.26318272948265076, "global_step": 43156, "epoch": 484, "lr": 9.40080160106845e-05}
{"train_loss": 0.3280152678489685, "global_step": 43157, "epoch": 484, "lr": 9.400774082036557e-05}
{"train_loss": 0.32648351788520813, "global_step": 43158, "epoch": 484, "lr": 9.400746562413031e-05}
{"train_loss": 0.2837607264518738, "global_step": 43159, "epoch": 484, "lr": 9.400719042197881e-05}
{"train_loss": 0.37497156858444214, "global_step": 43160, "epoch": 484, "lr": 9.400691521391107e-05}
{"train_loss": 0.25869667530059814, "global_step": 43161, "epoch": 484, "lr": 9.400663999992714e-05}
{"train_loss": 0.26991137862205505, "global_step": 43162, "epoch": 484, "lr": 9.400636478002705e-05}
{"train_loss": 0.2924099862575531, "global_step": 43163, "epoch": 484, "lr": 9.400608955421082e-05}
{"train_loss": 0.31367858172802443, "global_step": 43164, "epoch": 484, "lr": 9.400581432247853e-05, "val_loss": 2.7937588691711426}
{"train_loss": 0.29386141896247864, "global_step": 43165, "epoch": 485, "lr": 9.40055390848302e-05}
{"train_loss": 0.2900693118572235, "global_step": 43166, "epoch": 485, "lr": 9.400526384126584e-05}
{"train_loss": 0.23646821081638336, "global_step": 43167, "epoch": 485, "lr": 9.400498859178553e-05}
{"train_loss": 0.34528470039367676, "global_step": 43168, "epoch": 485, "lr": 9.400471333638927e-05}
{"train_loss": 0.3061816692352295, "global_step": 43169, "epoch": 485, "lr": 9.400443807507713e-05}
{"train_loss": 0.2827768325805664, "global_step": 43170, "epoch": 485, "lr": 9.400416280784911e-05}
{"train_loss": 0.2762519419193268, "global_step": 43171, "epoch": 485, "lr": 9.400388753470528e-05}
{"train_loss": 0.3102562129497528, "global_step": 43172, "epoch": 485, "lr": 9.400361225564565e-05}
{"train_loss": 0.2651573717594147, "global_step": 43173, "epoch": 485, "lr": 9.400333697067028e-05}
{"train_loss": 0.3300570547580719, "global_step": 43174, "epoch": 485, "lr": 9.40030616797792e-05}
{"train_loss": 0.3626917898654938, "global_step": 43175, "epoch": 485, "lr": 9.400278638297244e-05}
{"train_loss": 0.2670460641384125, "global_step": 43176, "epoch": 485, "lr": 9.400251108025005e-05}
{"train_loss": 0.2837938368320465, "global_step": 43177, "epoch": 485, "lr": 9.400223577161204e-05}
{"train_loss": 0.25187966227531433, "global_step": 43178, "epoch": 485, "lr": 9.400196045705849e-05}
{"train_loss": 0.25783371925354004, "global_step": 43179, "epoch": 485, "lr": 9.40016851365894e-05}
{"train_loss": 0.35218796133995056, "global_step": 43180, "epoch": 485, "lr": 9.400140981020482e-05}
{"train_loss": 0.2568506896495819, "global_step": 43181, "epoch": 485, "lr": 9.400113447790479e-05}
{"train_loss": 0.2480468600988388, "global_step": 43182, "epoch": 485, "lr": 9.400085913968933e-05}
{"train_loss": 0.31625479459762573, "global_step": 43183, "epoch": 485, "lr": 9.400058379555851e-05}
{"train_loss": 0.2962349057197571, "global_step": 43184, "epoch": 485, "lr": 9.400030844551233e-05}
{"train_loss": 0.29058876633644104, "global_step": 43185, "epoch": 485, "lr": 9.400003308955087e-05}
{"train_loss": 0.3581167459487915, "global_step": 43186, "epoch": 485, "lr": 9.399975772767411e-05}
{"train_loss": 0.28730183839797974, "global_step": 43187, "epoch": 485, "lr": 9.399948235988215e-05}
{"train_loss": 0.26263463497161865, "global_step": 43188, "epoch": 485, "lr": 9.399920698617496e-05}
{"train_loss": 0.25127384066581726, "global_step": 43189, "epoch": 485, "lr": 9.399893160655265e-05}
{"train_loss": 0.28433674573898315, "global_step": 43190, "epoch": 485, "lr": 9.399865622101521e-05}
{"train_loss": 0.2976803779602051, "global_step": 43191, "epoch": 485, "lr": 9.399838082956267e-05}
{"train_loss": 0.2837340831756592, "global_step": 43192, "epoch": 485, "lr": 9.39981054321951e-05}
{"train_loss": 0.2543196976184845, "global_step": 43193, "epoch": 485, "lr": 9.399783002891252e-05}
{"train_loss": 0.2910810708999634, "global_step": 43194, "epoch": 485, "lr": 9.399755461971496e-05}
{"train_loss": 0.18585021793842316, "global_step": 43195, "epoch": 485, "lr": 9.399727920460246e-05}
{"train_loss": 0.2525230944156647, "global_step": 43196, "epoch": 485, "lr": 9.399700378357507e-05}
{"train_loss": 0.21253681182861328, "global_step": 43197, "epoch": 485, "lr": 9.399672835663282e-05}
{"train_loss": 0.35053151845932007, "global_step": 43198, "epoch": 485, "lr": 9.399645292377574e-05}
{"train_loss": 0.3181576430797577, "global_step": 43199, "epoch": 485, "lr": 9.399617748500389e-05}
{"train_loss": 0.24576425552368164, "global_step": 43200, "epoch": 485, "lr": 9.399590204031728e-05}
{"train_loss": 0.23425664007663727, "global_step": 43201, "epoch": 485, "lr": 9.399562658971596e-05}
{"train_loss": 0.3550197184085846, "global_step": 43202, "epoch": 485, "lr": 9.399535113319995e-05}
{"train_loss": 0.2373160719871521, "global_step": 43203, "epoch": 485, "lr": 9.399507567076932e-05}
{"train_loss": 0.44797569513320923, "global_step": 43204, "epoch": 485, "lr": 9.399480020242408e-05}
{"train_loss": 0.21373562514781952, "global_step": 43205, "epoch": 485, "lr": 9.399452472816428e-05}
{"train_loss": 0.23360374569892883, "global_step": 43206, "epoch": 485, "lr": 9.399424924798995e-05}
{"train_loss": 0.3201809525489807, "global_step": 43207, "epoch": 485, "lr": 9.399397376190112e-05}
{"train_loss": 0.28136494755744934, "global_step": 43208, "epoch": 485, "lr": 9.399369826989784e-05}
{"train_loss": 0.2564992606639862, "global_step": 43209, "epoch": 485, "lr": 9.399342277198015e-05}
{"train_loss": 0.3740430772304535, "global_step": 43210, "epoch": 485, "lr": 9.399314726814807e-05}
{"train_loss": 0.3769893944263458, "global_step": 43211, "epoch": 485, "lr": 9.399287175840166e-05}
{"train_loss": 0.24301908910274506, "global_step": 43212, "epoch": 485, "lr": 9.399259624274095e-05}
{"train_loss": 0.37064069509506226, "global_step": 43213, "epoch": 485, "lr": 9.399232072116595e-05}
{"train_loss": 0.1711353361606598, "global_step": 43214, "epoch": 485, "lr": 9.399204519367674e-05}
{"train_loss": 0.22163839638233185, "global_step": 43215, "epoch": 485, "lr": 9.399176966027331e-05}
{"train_loss": 0.23141910135746002, "global_step": 43216, "epoch": 485, "lr": 9.399149412095574e-05}
{"train_loss": 0.2590879499912262, "global_step": 43217, "epoch": 485, "lr": 9.399121857572406e-05}
{"train_loss": 0.3426472544670105, "global_step": 43218, "epoch": 485, "lr": 9.39909430245783e-05}
{"train_loss": 0.22389642894268036, "global_step": 43219, "epoch": 485, "lr": 9.399066746751846e-05}
{"train_loss": 0.29928895831108093, "global_step": 43220, "epoch": 485, "lr": 9.399039190454464e-05}
{"train_loss": 0.264354944229126, "global_step": 43221, "epoch": 485, "lr": 9.399011633565683e-05}
{"train_loss": 0.29938170313835144, "global_step": 43222, "epoch": 485, "lr": 9.39898407608551e-05}
{"train_loss": 0.34695249795913696, "global_step": 43223, "epoch": 485, "lr": 9.398956518013946e-05}
{"train_loss": 0.48756781220436096, "global_step": 43224, "epoch": 485, "lr": 9.398928959350996e-05}
{"train_loss": 0.34059205651283264, "global_step": 43225, "epoch": 485, "lr": 9.398901400096664e-05}
{"train_loss": 0.3141131103038788, "global_step": 43226, "epoch": 485, "lr": 9.398873840250954e-05}
{"train_loss": 0.3317534029483795, "global_step": 43227, "epoch": 485, "lr": 9.398846279813869e-05}
{"train_loss": 0.2915317714214325, "global_step": 43228, "epoch": 485, "lr": 9.398818718785412e-05}
{"train_loss": 0.2760838568210602, "global_step": 43229, "epoch": 485, "lr": 9.398791157165587e-05}
{"train_loss": 0.32236427068710327, "global_step": 43230, "epoch": 485, "lr": 9.398763594954398e-05}
{"train_loss": 0.22312627732753754, "global_step": 43231, "epoch": 485, "lr": 9.39873603215185e-05}
{"train_loss": 0.3013421893119812, "global_step": 43232, "epoch": 485, "lr": 9.398708468757945e-05}
{"train_loss": 0.3433554470539093, "global_step": 43233, "epoch": 485, "lr": 9.398680904772688e-05}
{"train_loss": 0.19718025624752045, "global_step": 43234, "epoch": 485, "lr": 9.39865334019608e-05}
{"train_loss": 0.24490518867969513, "global_step": 43235, "epoch": 485, "lr": 9.398625775028129e-05}
{"train_loss": 0.323495090007782, "global_step": 43236, "epoch": 485, "lr": 9.398598209268836e-05}
{"train_loss": 0.29061761498451233, "global_step": 43237, "epoch": 485, "lr": 9.398570642918204e-05}
{"train_loss": 0.306740403175354, "global_step": 43238, "epoch": 485, "lr": 9.398543075976239e-05}
{"train_loss": 0.28658416867256165, "global_step": 43239, "epoch": 485, "lr": 9.398515508442942e-05}
{"train_loss": 0.27777355909347534, "global_step": 43240, "epoch": 485, "lr": 9.39848794031832e-05}
{"train_loss": 0.3007517457008362, "global_step": 43241, "epoch": 485, "lr": 9.398460371602373e-05}
{"train_loss": 0.24759744107723236, "global_step": 43242, "epoch": 485, "lr": 9.398432802295107e-05}
{"train_loss": 0.2909104526042938, "global_step": 43243, "epoch": 485, "lr": 9.398405232396526e-05}
{"train_loss": 0.3258557915687561, "global_step": 43244, "epoch": 485, "lr": 9.398377661906632e-05}
{"train_loss": 0.36015668511390686, "global_step": 43245, "epoch": 485, "lr": 9.398350090825431e-05}
{"train_loss": 0.26956427097320557, "global_step": 43246, "epoch": 485, "lr": 9.398322519152925e-05}
{"train_loss": 0.5166894197463989, "global_step": 43247, "epoch": 485, "lr": 9.398294946889119e-05}
{"train_loss": 0.23150253295898438, "global_step": 43248, "epoch": 485, "lr": 9.398267374034014e-05}
{"train_loss": 0.3518389165401459, "global_step": 43249, "epoch": 485, "lr": 9.398239800587618e-05}
{"train_loss": 0.27785030007362366, "global_step": 43250, "epoch": 485, "lr": 9.398212226549929e-05}
{"train_loss": 0.31366994976997375, "global_step": 43251, "epoch": 485, "lr": 9.398184651920955e-05}
{"train_loss": 0.40282830595970154, "global_step": 43252, "epoch": 485, "lr": 9.398157076700702e-05}
{"train_loss": 0.2942622740951817, "global_step": 43253, "epoch": 485, "lr": 9.398129500889167e-05, "val_loss": 2.8291432857513428, "train_action_mse_error": 51.07925796508789}
{"train_loss": 0.1695762276649475, "global_step": 43254, "epoch": 486, "lr": 9.398101924486358e-05}
{"train_loss": 0.391159325838089, "global_step": 43255, "epoch": 486, "lr": 9.398074347492279e-05}
{"train_loss": 0.30534809827804565, "global_step": 43256, "epoch": 486, "lr": 9.39804676990693e-05}
{"train_loss": 0.3485138714313507, "global_step": 43257, "epoch": 486, "lr": 9.398019191730318e-05}
{"train_loss": 0.2581549882888794, "global_step": 43258, "epoch": 486, "lr": 9.397991612962446e-05}
{"train_loss": 0.3476446270942688, "global_step": 43259, "epoch": 486, "lr": 9.39796403360332e-05}
{"train_loss": 0.2888309061527252, "global_step": 43260, "epoch": 486, "lr": 9.397936453652938e-05}
{"train_loss": 0.3033432364463806, "global_step": 43261, "epoch": 486, "lr": 9.397908873111309e-05}
{"train_loss": 0.3052004873752594, "global_step": 43262, "epoch": 486, "lr": 9.397881291978434e-05}
{"train_loss": 0.2749840021133423, "global_step": 43263, "epoch": 486, "lr": 9.397853710254317e-05}
{"train_loss": 0.2769821882247925, "global_step": 43264, "epoch": 486, "lr": 9.397826127938962e-05}
{"train_loss": 0.2756236493587494, "global_step": 43265, "epoch": 486, "lr": 9.397798545032374e-05}
{"train_loss": 0.28862500190734863, "global_step": 43266, "epoch": 486, "lr": 9.397770961534555e-05}
{"train_loss": 0.37063512206077576, "global_step": 43267, "epoch": 486, "lr": 9.39774337744551e-05}
{"train_loss": 0.3156614899635315, "global_step": 43268, "epoch": 486, "lr": 9.397715792765243e-05}
{"train_loss": 0.3955532908439636, "global_step": 43269, "epoch": 486, "lr": 9.397688207493754e-05}
{"train_loss": 0.28785794973373413, "global_step": 43270, "epoch": 486, "lr": 9.397660621631052e-05}
{"train_loss": 0.24414335191249847, "global_step": 43271, "epoch": 486, "lr": 9.397633035177135e-05}
{"train_loss": 0.39165228605270386, "global_step": 43272, "epoch": 486, "lr": 9.397605448132011e-05}
{"train_loss": 0.3750816583633423, "global_step": 43273, "epoch": 486, "lr": 9.397577860495686e-05}
{"train_loss": 0.36852291226387024, "global_step": 43274, "epoch": 486, "lr": 9.397550272268157e-05}
{"train_loss": 0.38197430968284607, "global_step": 43275, "epoch": 486, "lr": 9.397522683449431e-05}
{"train_loss": 0.3076719343662262, "global_step": 43276, "epoch": 486, "lr": 9.397495094039512e-05}
{"train_loss": 0.31902801990509033, "global_step": 43277, "epoch": 486, "lr": 9.397467504038403e-05}
{"train_loss": 0.3493342697620392, "global_step": 43278, "epoch": 486, "lr": 9.397439913446108e-05}
{"train_loss": 0.32279568910598755, "global_step": 43279, "epoch": 486, "lr": 9.397412322262634e-05}
{"train_loss": 0.28845906257629395, "global_step": 43280, "epoch": 486, "lr": 9.397384730487977e-05}
{"train_loss": 0.25344058871269226, "global_step": 43281, "epoch": 486, "lr": 9.397357138122147e-05}
{"train_loss": 0.3157190680503845, "global_step": 43282, "epoch": 486, "lr": 9.397329545165147e-05}
{"train_loss": 0.2888186275959015, "global_step": 43283, "epoch": 486, "lr": 9.397301951616978e-05}
{"train_loss": 0.33567652106285095, "global_step": 43284, "epoch": 486, "lr": 9.397274357477646e-05}
{"train_loss": 0.28217121958732605, "global_step": 43285, "epoch": 486, "lr": 9.397246762747154e-05}
{"train_loss": 0.21405363082885742, "global_step": 43286, "epoch": 486, "lr": 9.397219167425505e-05}
{"train_loss": 0.4585336446762085, "global_step": 43287, "epoch": 486, "lr": 9.397191571512704e-05}
{"train_loss": 0.2641335427761078, "global_step": 43288, "epoch": 486, "lr": 9.397163975008755e-05}
{"train_loss": 0.24529224634170532, "global_step": 43289, "epoch": 486, "lr": 9.397136377913661e-05}
{"train_loss": 0.2592657506465912, "global_step": 43290, "epoch": 486, "lr": 9.397108780227427e-05}
{"train_loss": 0.2682541608810425, "global_step": 43291, "epoch": 486, "lr": 9.397081181950051e-05}
{"train_loss": 0.29314690828323364, "global_step": 43292, "epoch": 486, "lr": 9.397053583081544e-05}
{"train_loss": 0.39582422375679016, "global_step": 43293, "epoch": 486, "lr": 9.397025983621905e-05}
{"train_loss": 0.23469242453575134, "global_step": 43294, "epoch": 486, "lr": 9.396998383571142e-05}
{"train_loss": 0.2689529061317444, "global_step": 43295, "epoch": 486, "lr": 9.396970782929254e-05}
{"train_loss": 0.3703593909740448, "global_step": 43296, "epoch": 486, "lr": 9.396943181696247e-05}
{"train_loss": 0.2544909119606018, "global_step": 43297, "epoch": 486, "lr": 9.396915579872125e-05}
{"train_loss": 0.29825854301452637, "global_step": 43298, "epoch": 486, "lr": 9.396887977456892e-05}
{"train_loss": 0.33992189168930054, "global_step": 43299, "epoch": 486, "lr": 9.396860374450551e-05}
{"train_loss": 0.2781585454940796, "global_step": 43300, "epoch": 486, "lr": 9.396832770853105e-05}
{"train_loss": 0.2827755808830261, "global_step": 43301, "epoch": 486, "lr": 9.396805166664559e-05}
{"train_loss": 0.369124174118042, "global_step": 43302, "epoch": 486, "lr": 9.396777561884915e-05}
{"train_loss": 0.32489141821861267, "global_step": 43303, "epoch": 486, "lr": 9.39674995651418e-05}
{"train_loss": 0.3275420665740967, "global_step": 43304, "epoch": 486, "lr": 9.396722350552354e-05}
{"train_loss": 0.2959795892238617, "global_step": 43305, "epoch": 486, "lr": 9.396694743999441e-05}
{"train_loss": 0.2657081186771393, "global_step": 43306, "epoch": 486, "lr": 9.396667136855449e-05}
{"train_loss": 0.33968329429626465, "global_step": 43307, "epoch": 486, "lr": 9.396639529120377e-05}
{"train_loss": 0.32945913076400757, "global_step": 43308, "epoch": 486, "lr": 9.396611920794229e-05}
{"train_loss": 0.2724725902080536, "global_step": 43309, "epoch": 486, "lr": 9.396584311877012e-05}
{"train_loss": 0.3129364550113678, "global_step": 43310, "epoch": 486, "lr": 9.396556702368728e-05}
{"train_loss": 0.22864679992198944, "global_step": 43311, "epoch": 486, "lr": 9.39652909226938e-05}
{"train_loss": 0.29701194167137146, "global_step": 43312, "epoch": 486, "lr": 9.396501481578972e-05}
{"train_loss": 0.31051546335220337, "global_step": 43313, "epoch": 486, "lr": 9.396473870297509e-05}
{"train_loss": 0.24725182354450226, "global_step": 43314, "epoch": 486, "lr": 9.396446258424991e-05}
{"train_loss": 0.3074922561645508, "global_step": 43315, "epoch": 486, "lr": 9.396418645961427e-05}
{"train_loss": 0.34614819288253784, "global_step": 43316, "epoch": 486, "lr": 9.396391032906817e-05}
{"train_loss": 0.3268186151981354, "global_step": 43317, "epoch": 486, "lr": 9.396363419261167e-05}
{"train_loss": 0.3828255236148834, "global_step": 43318, "epoch": 486, "lr": 9.396335805024479e-05}
{"train_loss": 0.24677656590938568, "global_step": 43319, "epoch": 486, "lr": 9.396308190196757e-05}
{"train_loss": 0.31640520691871643, "global_step": 43320, "epoch": 486, "lr": 9.396280574778005e-05}
{"train_loss": 0.20704738795757294, "global_step": 43321, "epoch": 486, "lr": 9.396252958768225e-05}
{"train_loss": 0.2661885619163513, "global_step": 43322, "epoch": 486, "lr": 9.396225342167424e-05}
{"train_loss": 0.29808494448661804, "global_step": 43323, "epoch": 486, "lr": 9.396197724975604e-05}
{"train_loss": 0.28830233216285706, "global_step": 43324, "epoch": 486, "lr": 9.396170107192769e-05}
{"train_loss": 0.2346513420343399, "global_step": 43325, "epoch": 486, "lr": 9.396142488818922e-05}
{"train_loss": 0.32278287410736084, "global_step": 43326, "epoch": 486, "lr": 9.396114869854067e-05}
{"train_loss": 0.3523537516593933, "global_step": 43327, "epoch": 486, "lr": 9.396087250298208e-05}
{"train_loss": 0.3536228835582733, "global_step": 43328, "epoch": 486, "lr": 9.39605963015135e-05}
{"train_loss": 0.351341187953949, "global_step": 43329, "epoch": 486, "lr": 9.396032009413494e-05}
{"train_loss": 0.24922972917556763, "global_step": 43330, "epoch": 486, "lr": 9.396004388084645e-05}
{"train_loss": 0.3270086646080017, "global_step": 43331, "epoch": 486, "lr": 9.395976766164809e-05}
{"train_loss": 0.32581910490989685, "global_step": 43332, "epoch": 486, "lr": 9.395949143653986e-05}
{"train_loss": 0.33365511894226074, "global_step": 43333, "epoch": 486, "lr": 9.39592152055218e-05}
{"train_loss": 0.23898132145404816, "global_step": 43334, "epoch": 486, "lr": 9.395893896859398e-05}
{"train_loss": 0.29993948340415955, "global_step": 43335, "epoch": 486, "lr": 9.395866272575639e-05}
{"train_loss": 0.35126233100891113, "global_step": 43336, "epoch": 486, "lr": 9.395838647700912e-05}
{"train_loss": 0.2759862542152405, "global_step": 43337, "epoch": 486, "lr": 9.395811022235217e-05}
{"train_loss": 0.22629421949386597, "global_step": 43338, "epoch": 486, "lr": 9.39578339617856e-05}
{"train_loss": 0.29127126932144165, "global_step": 43339, "epoch": 486, "lr": 9.39575576953094e-05}
{"train_loss": 0.31509125232696533, "global_step": 43340, "epoch": 486, "lr": 9.395728142292368e-05}
{"train_loss": 0.4186908006668091, "global_step": 43341, "epoch": 486, "lr": 9.395700514462842e-05}
{"train_loss": 0.30564621409003656, "global_step": 43342, "epoch": 486, "lr": 9.395672886042368e-05, "val_loss": 2.6698784828186035}
{"train_loss": 0.3008574843406677, "global_step": 43343, "epoch": 487, "lr": 9.39564525703095e-05}
{"train_loss": 0.25888004899024963, "global_step": 43344, "epoch": 487, "lr": 9.39561762742859e-05}
{"train_loss": 0.22239983081817627, "global_step": 43345, "epoch": 487, "lr": 9.395589997235294e-05}
{"train_loss": 0.27100586891174316, "global_step": 43346, "epoch": 487, "lr": 9.395562366451062e-05}
{"train_loss": 0.28996509313583374, "global_step": 43347, "epoch": 487, "lr": 9.395534735075904e-05}
{"train_loss": 0.3510579466819763, "global_step": 43348, "epoch": 487, "lr": 9.395507103109818e-05}
{"train_loss": 0.26380008459091187, "global_step": 43349, "epoch": 487, "lr": 9.395479470552808e-05}
{"train_loss": 0.3184661865234375, "global_step": 43350, "epoch": 487, "lr": 9.395451837404882e-05}
{"train_loss": 0.27738574147224426, "global_step": 43351, "epoch": 487, "lr": 9.395424203666038e-05}
{"train_loss": 0.2365686446428299, "global_step": 43352, "epoch": 487, "lr": 9.395396569336287e-05}
{"train_loss": 0.23483537137508392, "global_step": 43353, "epoch": 487, "lr": 9.395368934415626e-05}
{"train_loss": 0.3051603436470032, "global_step": 43354, "epoch": 487, "lr": 9.395341298904062e-05}
{"train_loss": 0.2949927747249603, "global_step": 43355, "epoch": 487, "lr": 9.395313662801598e-05}
{"train_loss": 0.2549622356891632, "global_step": 43356, "epoch": 487, "lr": 9.395286026108236e-05}
{"train_loss": 0.32944610714912415, "global_step": 43357, "epoch": 487, "lr": 9.395258388823983e-05}
{"train_loss": 0.24529330432415009, "global_step": 43358, "epoch": 487, "lr": 9.39523075094884e-05}
{"train_loss": 0.2590710520744324, "global_step": 43359, "epoch": 487, "lr": 9.395203112482812e-05}
{"train_loss": 0.32638639211654663, "global_step": 43360, "epoch": 487, "lr": 9.395175473425904e-05}
{"train_loss": 0.3109095096588135, "global_step": 43361, "epoch": 487, "lr": 9.395147833778116e-05}
{"train_loss": 0.29114073514938354, "global_step": 43362, "epoch": 487, "lr": 9.395120193539455e-05}
{"train_loss": 0.2701634168624878, "global_step": 43363, "epoch": 487, "lr": 9.395092552709925e-05}
{"train_loss": 0.3208076059818268, "global_step": 43364, "epoch": 487, "lr": 9.395064911289526e-05}
{"train_loss": 0.2837528884410858, "global_step": 43365, "epoch": 487, "lr": 9.395037269278266e-05}
{"train_loss": 0.3357861340045929, "global_step": 43366, "epoch": 487, "lr": 9.395009626676146e-05}
{"train_loss": 0.30057820677757263, "global_step": 43367, "epoch": 487, "lr": 9.39498198348317e-05}
{"train_loss": 0.2675895690917969, "global_step": 43368, "epoch": 487, "lr": 9.394954339699342e-05}
{"train_loss": 0.3210137188434601, "global_step": 43369, "epoch": 487, "lr": 9.394926695324668e-05}
{"train_loss": 0.24832583963871002, "global_step": 43370, "epoch": 487, "lr": 9.394899050359148e-05}
{"train_loss": 0.4175466001033783, "global_step": 43371, "epoch": 487, "lr": 9.394871404802787e-05}
{"train_loss": 0.39676448702812195, "global_step": 43372, "epoch": 487, "lr": 9.39484375865559e-05}
{"train_loss": 0.24845655262470245, "global_step": 43373, "epoch": 487, "lr": 9.39481611191756e-05}
{"train_loss": 0.19448938965797424, "global_step": 43374, "epoch": 487, "lr": 9.3947884645887e-05}
{"train_loss": 0.2700459957122803, "global_step": 43375, "epoch": 487, "lr": 9.394760816669014e-05}
{"train_loss": 0.22777746617794037, "global_step": 43376, "epoch": 487, "lr": 9.394733168158507e-05}
{"train_loss": 0.24935828149318695, "global_step": 43377, "epoch": 487, "lr": 9.394705519057181e-05}
{"train_loss": 0.28617069125175476, "global_step": 43378, "epoch": 487, "lr": 9.394677869365041e-05}
{"train_loss": 0.37946438789367676, "global_step": 43379, "epoch": 487, "lr": 9.39465021908209e-05}
{"train_loss": 0.3059380352497101, "global_step": 43380, "epoch": 487, "lr": 9.394622568208332e-05}
{"train_loss": 0.26802390813827515, "global_step": 43381, "epoch": 487, "lr": 9.394594916743769e-05}
{"train_loss": 0.37824270129203796, "global_step": 43382, "epoch": 487, "lr": 9.394567264688407e-05}
{"train_loss": 0.24510127305984497, "global_step": 43383, "epoch": 487, "lr": 9.39453961204225e-05}
{"train_loss": 0.24646179378032684, "global_step": 43384, "epoch": 487, "lr": 9.3945119588053e-05}
{"train_loss": 0.2402293086051941, "global_step": 43385, "epoch": 487, "lr": 9.394484304977561e-05}
{"train_loss": 0.36295780539512634, "global_step": 43386, "epoch": 487, "lr": 9.394456650559038e-05}
{"train_loss": 0.3595486581325531, "global_step": 43387, "epoch": 487, "lr": 9.394428995549734e-05}
{"train_loss": 0.395253449678421, "global_step": 43388, "epoch": 487, "lr": 9.394401339949653e-05}
{"train_loss": 0.33323389291763306, "global_step": 43389, "epoch": 487, "lr": 9.394373683758797e-05}
{"train_loss": 0.24032153189182281, "global_step": 43390, "epoch": 487, "lr": 9.39434602697717e-05}
{"train_loss": 0.33677369356155396, "global_step": 43391, "epoch": 487, "lr": 9.394318369604779e-05}
{"train_loss": 0.2880108654499054, "global_step": 43392, "epoch": 487, "lr": 9.394290711641625e-05}
{"train_loss": 0.29244911670684814, "global_step": 43393, "epoch": 487, "lr": 9.39426305308771e-05}
{"train_loss": 0.3590950071811676, "global_step": 43394, "epoch": 487, "lr": 9.394235393943043e-05}
{"train_loss": 0.37407386302948, "global_step": 43395, "epoch": 487, "lr": 9.394207734207623e-05}
{"train_loss": 0.353192001581192, "global_step": 43396, "epoch": 487, "lr": 9.394180073881456e-05}
{"train_loss": 0.3109318017959595, "global_step": 43397, "epoch": 487, "lr": 9.394152412964543e-05}
{"train_loss": 0.25647813081741333, "global_step": 43398, "epoch": 487, "lr": 9.394124751456891e-05}
{"train_loss": 0.29541948437690735, "global_step": 43399, "epoch": 487, "lr": 9.394097089358505e-05}
{"train_loss": 0.31455445289611816, "global_step": 43400, "epoch": 487, "lr": 9.394069426669383e-05}
{"train_loss": 0.23849357664585114, "global_step": 43401, "epoch": 487, "lr": 9.394041763389533e-05}
{"train_loss": 0.2207845002412796, "global_step": 43402, "epoch": 487, "lr": 9.394014099518958e-05}
{"train_loss": 0.2613569498062134, "global_step": 43403, "epoch": 487, "lr": 9.393986435057661e-05}
{"train_loss": 0.26533207297325134, "global_step": 43404, "epoch": 487, "lr": 9.393958770005646e-05}
{"train_loss": 0.23293691873550415, "global_step": 43405, "epoch": 487, "lr": 9.393931104362916e-05}
{"train_loss": 0.3099212348461151, "global_step": 43406, "epoch": 487, "lr": 9.393903438129476e-05}
{"train_loss": 0.3742772936820984, "global_step": 43407, "epoch": 487, "lr": 9.393875771305332e-05}
{"train_loss": 0.32606562972068787, "global_step": 43408, "epoch": 487, "lr": 9.393848103890482e-05}
{"train_loss": 0.3495499789714813, "global_step": 43409, "epoch": 487, "lr": 9.393820435884932e-05}
{"train_loss": 0.29667404294013977, "global_step": 43410, "epoch": 487, "lr": 9.393792767288688e-05}
{"train_loss": 0.3191623389720917, "global_step": 43411, "epoch": 487, "lr": 9.393765098101752e-05}
{"train_loss": 0.386218398809433, "global_step": 43412, "epoch": 487, "lr": 9.393737428324127e-05}
{"train_loss": 0.3860229551792145, "global_step": 43413, "epoch": 487, "lr": 9.39370975795582e-05}
{"train_loss": 0.3301810324192047, "global_step": 43414, "epoch": 487, "lr": 9.393682086996829e-05}
{"train_loss": 0.3561384677886963, "global_step": 43415, "epoch": 487, "lr": 9.393654415447163e-05}
{"train_loss": 0.2658052146434784, "global_step": 43416, "epoch": 487, "lr": 9.393626743306822e-05}
{"train_loss": 0.1948537826538086, "global_step": 43417, "epoch": 487, "lr": 9.393599070575814e-05}
{"train_loss": 0.35308659076690674, "global_step": 43418, "epoch": 487, "lr": 9.393571397254138e-05}
{"train_loss": 0.36064714193344116, "global_step": 43419, "epoch": 487, "lr": 9.393543723341801e-05}
{"train_loss": 0.2622954845428467, "global_step": 43420, "epoch": 487, "lr": 9.393516048838806e-05}
{"train_loss": 0.30452510714530945, "global_step": 43421, "epoch": 487, "lr": 9.393488373745156e-05}
{"train_loss": 0.1905219852924347, "global_step": 43422, "epoch": 487, "lr": 9.393460698060854e-05}
{"train_loss": 0.26570913195610046, "global_step": 43423, "epoch": 487, "lr": 9.393433021785905e-05}
{"train_loss": 0.37244337797164917, "global_step": 43424, "epoch": 487, "lr": 9.393405344920312e-05}
{"train_loss": 0.38071000576019287, "global_step": 43425, "epoch": 487, "lr": 9.39337766746408e-05}
{"train_loss": 0.2766571044921875, "global_step": 43426, "epoch": 487, "lr": 9.393349989417213e-05}
{"train_loss": 0.47183382511138916, "global_step": 43427, "epoch": 487, "lr": 9.393322310779711e-05}
{"train_loss": 0.2940206825733185, "global_step": 43428, "epoch": 487, "lr": 9.393294631551584e-05}
{"train_loss": 0.29451504349708557, "global_step": 43429, "epoch": 487, "lr": 9.393266951732829e-05}
{"train_loss": 0.36871063709259033, "global_step": 43430, "epoch": 487, "lr": 9.393239271323454e-05}
{"train_loss": 0.2995914954147982, "global_step": 43431, "epoch": 487, "lr": 9.393211590323461e-05, "val_loss": 2.8774654865264893}
{"train_loss": 0.30342036485671997, "global_step": 43432, "epoch": 488, "lr": 9.393183908732854e-05}
{"train_loss": 0.3681493401527405, "global_step": 43433, "epoch": 488, "lr": 9.393156226551638e-05}
{"train_loss": 0.2122741937637329, "global_step": 43434, "epoch": 488, "lr": 9.393128543779816e-05}
{"train_loss": 0.3187633156776428, "global_step": 43435, "epoch": 488, "lr": 9.39310086041739e-05}
{"train_loss": 0.2425714135169983, "global_step": 43436, "epoch": 488, "lr": 9.393073176464366e-05}
{"train_loss": 0.23841430246829987, "global_step": 43437, "epoch": 488, "lr": 9.393045491920746e-05}
{"train_loss": 0.30501747131347656, "global_step": 43438, "epoch": 488, "lr": 9.393017806786536e-05}
{"train_loss": 0.3058544099330902, "global_step": 43439, "epoch": 488, "lr": 9.392990121061738e-05}
{"train_loss": 0.3556171953678131, "global_step": 43440, "epoch": 488, "lr": 9.392962434746356e-05}
{"train_loss": 0.22620941698551178, "global_step": 43441, "epoch": 488, "lr": 9.392934747840392e-05}
{"train_loss": 0.3945900499820709, "global_step": 43442, "epoch": 488, "lr": 9.392907060343853e-05}
{"train_loss": 0.3051663935184479, "global_step": 43443, "epoch": 488, "lr": 9.392879372256741e-05}
{"train_loss": 0.3615747392177582, "global_step": 43444, "epoch": 488, "lr": 9.392851683579061e-05}
{"train_loss": 0.2779040038585663, "global_step": 43445, "epoch": 488, "lr": 9.392823994310813e-05}
{"train_loss": 0.2869444787502289, "global_step": 43446, "epoch": 488, "lr": 9.392796304452005e-05}
{"train_loss": 0.34696975350379944, "global_step": 43447, "epoch": 488, "lr": 9.39276861400264e-05}
{"train_loss": 0.2643490433692932, "global_step": 43448, "epoch": 488, "lr": 9.392740922962719e-05}
{"train_loss": 0.3145714998245239, "global_step": 43449, "epoch": 488, "lr": 9.392713231332249e-05}
{"train_loss": 0.3312976062297821, "global_step": 43450, "epoch": 488, "lr": 9.392685539111231e-05}
{"train_loss": 0.22867687046527863, "global_step": 43451, "epoch": 488, "lr": 9.392657846299671e-05}
{"train_loss": 0.33595311641693115, "global_step": 43452, "epoch": 488, "lr": 9.39263015289757e-05}
{"train_loss": 0.3014574944972992, "global_step": 43453, "epoch": 488, "lr": 9.392602458904935e-05}
{"train_loss": 0.2682342231273651, "global_step": 43454, "epoch": 488, "lr": 9.392574764321767e-05}
{"train_loss": 0.32408127188682556, "global_step": 43455, "epoch": 488, "lr": 9.392547069148073e-05}
{"train_loss": 0.2835340201854706, "global_step": 43456, "epoch": 488, "lr": 9.392519373383852e-05}
{"train_loss": 0.2833109200000763, "global_step": 43457, "epoch": 488, "lr": 9.392491677029111e-05}
{"train_loss": 0.31830474734306335, "global_step": 43458, "epoch": 488, "lr": 9.392463980083855e-05}
{"train_loss": 0.3483944237232208, "global_step": 43459, "epoch": 488, "lr": 9.392436282548083e-05}
{"train_loss": 0.3172878921031952, "global_step": 43460, "epoch": 488, "lr": 9.392408584421804e-05}
{"train_loss": 0.28348878026008606, "global_step": 43461, "epoch": 488, "lr": 9.392380885705016e-05}
{"train_loss": 0.3120090067386627, "global_step": 43462, "epoch": 488, "lr": 9.392353186397729e-05}
{"train_loss": 0.3449501395225525, "global_step": 43463, "epoch": 488, "lr": 9.392325486499943e-05}
{"train_loss": 0.3846193850040436, "global_step": 43464, "epoch": 488, "lr": 9.392297786011661e-05}
{"train_loss": 0.3271462917327881, "global_step": 43465, "epoch": 488, "lr": 9.392270084932889e-05}
{"train_loss": 0.3551352918148041, "global_step": 43466, "epoch": 488, "lr": 9.392242383263629e-05}
{"train_loss": 0.23198571801185608, "global_step": 43467, "epoch": 488, "lr": 9.392214681003888e-05}
{"train_loss": 0.30066385865211487, "global_step": 43468, "epoch": 488, "lr": 9.392186978153665e-05}
{"train_loss": 0.2668934762477875, "global_step": 43469, "epoch": 488, "lr": 9.392159274712966e-05}
{"train_loss": 0.36296093463897705, "global_step": 43470, "epoch": 488, "lr": 9.392131570681796e-05}
{"train_loss": 0.2678040564060211, "global_step": 43471, "epoch": 488, "lr": 9.392103866060157e-05}
{"train_loss": 0.19459374248981476, "global_step": 43472, "epoch": 488, "lr": 9.392076160848053e-05}
{"train_loss": 0.254410058259964, "global_step": 43473, "epoch": 488, "lr": 9.392048455045488e-05}
{"train_loss": 0.3048916757106781, "global_step": 43474, "epoch": 488, "lr": 9.392020748652465e-05}
{"train_loss": 0.3398457467556, "global_step": 43475, "epoch": 488, "lr": 9.39199304166899e-05}
{"train_loss": 0.4279516339302063, "global_step": 43476, "epoch": 488, "lr": 9.391965334095063e-05}
{"train_loss": 0.35593217611312866, "global_step": 43477, "epoch": 488, "lr": 9.391937625930691e-05}
{"train_loss": 0.35324257612228394, "global_step": 43478, "epoch": 488, "lr": 9.391909917175876e-05}
{"train_loss": 0.3399021625518799, "global_step": 43479, "epoch": 488, "lr": 9.391882207830622e-05}
{"train_loss": 0.3064868152141571, "global_step": 43480, "epoch": 488, "lr": 9.391854497894933e-05}
{"train_loss": 0.3836114704608917, "global_step": 43481, "epoch": 488, "lr": 9.391826787368813e-05}
{"train_loss": 0.27462953329086304, "global_step": 43482, "epoch": 488, "lr": 9.391799076252266e-05}
{"train_loss": 0.31760740280151367, "global_step": 43483, "epoch": 488, "lr": 9.391771364545294e-05}
{"train_loss": 0.2360825538635254, "global_step": 43484, "epoch": 488, "lr": 9.391743652247903e-05}
{"train_loss": 0.23997749388217926, "global_step": 43485, "epoch": 488, "lr": 9.391715939360095e-05}
{"train_loss": 0.23203939199447632, "global_step": 43486, "epoch": 488, "lr": 9.391688225881874e-05}
{"train_loss": 0.2466803789138794, "global_step": 43487, "epoch": 488, "lr": 9.391660511813245e-05}
{"train_loss": 0.30680516362190247, "global_step": 43488, "epoch": 488, "lr": 9.39163279715421e-05}
{"train_loss": 0.3123261332511902, "global_step": 43489, "epoch": 488, "lr": 9.391605081904773e-05}
{"train_loss": 0.28254082798957825, "global_step": 43490, "epoch": 488, "lr": 9.39157736606494e-05}
{"train_loss": 0.2958706319332123, "global_step": 43491, "epoch": 488, "lr": 9.391549649634713e-05}
{"train_loss": 0.4621110260486603, "global_step": 43492, "epoch": 488, "lr": 9.391521932614093e-05}
{"train_loss": 0.24407751858234406, "global_step": 43493, "epoch": 488, "lr": 9.391494215003087e-05}
{"train_loss": 0.21602731943130493, "global_step": 43494, "epoch": 488, "lr": 9.391466496801699e-05}
{"train_loss": 0.2907375395298004, "global_step": 43495, "epoch": 488, "lr": 9.391438778009933e-05}
{"train_loss": 0.2788555920124054, "global_step": 43496, "epoch": 488, "lr": 9.391411058627791e-05}
{"train_loss": 0.34944701194763184, "global_step": 43497, "epoch": 488, "lr": 9.391383338655277e-05}
{"train_loss": 0.2999795377254486, "global_step": 43498, "epoch": 488, "lr": 9.391355618092394e-05}
{"train_loss": 0.3129526376724243, "global_step": 43499, "epoch": 488, "lr": 9.391327896939149e-05}
{"train_loss": 0.3955829441547394, "global_step": 43500, "epoch": 488, "lr": 9.391300175195541e-05}
{"train_loss": 0.30032384395599365, "global_step": 43501, "epoch": 488, "lr": 9.391272452861576e-05}
{"train_loss": 0.2829759120941162, "global_step": 43502, "epoch": 488, "lr": 9.391244729937261e-05}
{"train_loss": 0.3298513889312744, "global_step": 43503, "epoch": 488, "lr": 9.391217006422594e-05}
{"train_loss": 0.27891358733177185, "global_step": 43504, "epoch": 488, "lr": 9.391189282317583e-05}
{"train_loss": 0.361043244600296, "global_step": 43505, "epoch": 488, "lr": 9.39116155762223e-05}
{"train_loss": 0.34070488810539246, "global_step": 43506, "epoch": 488, "lr": 9.391133832336538e-05}
{"train_loss": 0.18454931676387787, "global_step": 43507, "epoch": 488, "lr": 9.391106106460512e-05}
{"train_loss": 0.2788533866405487, "global_step": 43508, "epoch": 488, "lr": 9.391078379994155e-05}
{"train_loss": 0.2617138624191284, "global_step": 43509, "epoch": 488, "lr": 9.39105065293747e-05}
{"train_loss": 0.31348931789398193, "global_step": 43510, "epoch": 488, "lr": 9.391022925290464e-05}
{"train_loss": 0.328593909740448, "global_step": 43511, "epoch": 488, "lr": 9.390995197053137e-05}
{"train_loss": 0.35746991634368896, "global_step": 43512, "epoch": 488, "lr": 9.390967468225495e-05}
{"train_loss": 0.41061288118362427, "global_step": 43513, "epoch": 488, "lr": 9.39093973880754e-05}
{"train_loss": 0.2996346354484558, "global_step": 43514, "epoch": 488, "lr": 9.390912008799277e-05}
{"train_loss": 0.28325769305229187, "global_step": 43515, "epoch": 488, "lr": 9.390884278200711e-05}
{"train_loss": 0.260250985622406, "global_step": 43516, "epoch": 488, "lr": 9.390856547011841e-05}
{"train_loss": 0.2975350618362427, "global_step": 43517, "epoch": 488, "lr": 9.390828815232677e-05}
{"train_loss": 0.26024797558784485, "global_step": 43518, "epoch": 488, "lr": 9.390801082863219e-05}
{"train_loss": 0.4197530746459961, "global_step": 43519, "epoch": 488, "lr": 9.39077334990347e-05}
{"train_loss": 0.30719632496324817, "global_step": 43520, "epoch": 488, "lr": 9.390745616353436e-05, "val_loss": 2.8541839122772217}
{"train_loss": 0.3139735460281372, "global_step": 43521, "epoch": 489, "lr": 9.390717882213118e-05}
{"train_loss": 0.26978743076324463, "global_step": 43522, "epoch": 489, "lr": 9.390690147482524e-05}
{"train_loss": 0.25134673714637756, "global_step": 43523, "epoch": 489, "lr": 9.390662412161654e-05}
{"train_loss": 0.26894044876098633, "global_step": 43524, "epoch": 489, "lr": 9.390634676250512e-05}
{"train_loss": 0.2048371583223343, "global_step": 43525, "epoch": 489, "lr": 9.390606939749104e-05}
{"train_loss": 0.2697423994541168, "global_step": 43526, "epoch": 489, "lr": 9.390579202657432e-05}
{"train_loss": 0.25524675846099854, "global_step": 43527, "epoch": 489, "lr": 9.390551464975501e-05}
{"train_loss": 0.3530142307281494, "global_step": 43528, "epoch": 489, "lr": 9.390523726703313e-05}
{"train_loss": 0.2434769719839096, "global_step": 43529, "epoch": 489, "lr": 9.390495987840873e-05}
{"train_loss": 0.2900744676589966, "global_step": 43530, "epoch": 489, "lr": 9.390468248388184e-05}
{"train_loss": 0.34596630930900574, "global_step": 43531, "epoch": 489, "lr": 9.39044050834525e-05}
{"train_loss": 0.2497415840625763, "global_step": 43532, "epoch": 489, "lr": 9.390412767712075e-05}
{"train_loss": 0.23874853551387787, "global_step": 43533, "epoch": 489, "lr": 9.390385026488661e-05}
{"train_loss": 0.3928523063659668, "global_step": 43534, "epoch": 489, "lr": 9.390357284675015e-05}
{"train_loss": 0.24549530446529388, "global_step": 43535, "epoch": 489, "lr": 9.390329542271138e-05}
{"train_loss": 0.2550348937511444, "global_step": 43536, "epoch": 489, "lr": 9.390301799277037e-05}
{"train_loss": 0.3619404733181, "global_step": 43537, "epoch": 489, "lr": 9.390274055692711e-05}
{"train_loss": 0.3801540732383728, "global_step": 43538, "epoch": 489, "lr": 9.390246311518167e-05}
{"train_loss": 0.3625629246234894, "global_step": 43539, "epoch": 489, "lr": 9.390218566753408e-05}
{"train_loss": 0.2565722167491913, "global_step": 43540, "epoch": 489, "lr": 9.390190821398437e-05}
{"train_loss": 0.294109582901001, "global_step": 43541, "epoch": 489, "lr": 9.39016307545326e-05}
{"train_loss": 0.3075988292694092, "global_step": 43542, "epoch": 489, "lr": 9.390135328917878e-05}
{"train_loss": 0.340795636177063, "global_step": 43543, "epoch": 489, "lr": 9.390107581792295e-05}
{"train_loss": 0.1887105256319046, "global_step": 43544, "epoch": 489, "lr": 9.390079834076516e-05}
{"train_loss": 0.35999640822410583, "global_step": 43545, "epoch": 489, "lr": 9.390052085770544e-05}
{"train_loss": 0.19537843763828278, "global_step": 43546, "epoch": 489, "lr": 9.390024336874385e-05}
{"train_loss": 0.283830851316452, "global_step": 43547, "epoch": 489, "lr": 9.389996587388039e-05}
{"train_loss": 0.23709183931350708, "global_step": 43548, "epoch": 489, "lr": 9.389968837311512e-05}
{"train_loss": 0.247674822807312, "global_step": 43549, "epoch": 489, "lr": 9.389941086644807e-05}
{"train_loss": 0.31007182598114014, "global_step": 43550, "epoch": 489, "lr": 9.389913335387928e-05}
{"train_loss": 0.3084545135498047, "global_step": 43551, "epoch": 489, "lr": 9.389885583540878e-05}
{"train_loss": 0.3488571345806122, "global_step": 43552, "epoch": 489, "lr": 9.38985783110366e-05}
{"train_loss": 0.2773587107658386, "global_step": 43553, "epoch": 489, "lr": 9.389830078076283e-05}
{"train_loss": 0.2776695489883423, "global_step": 43554, "epoch": 489, "lr": 9.389802324458745e-05}
{"train_loss": 0.2758398950099945, "global_step": 43555, "epoch": 489, "lr": 9.389774570251051e-05}
{"train_loss": 0.2933672070503235, "global_step": 43556, "epoch": 489, "lr": 9.389746815453207e-05}
{"train_loss": 0.3161105513572693, "global_step": 43557, "epoch": 489, "lr": 9.389719060065212e-05}
{"train_loss": 0.3306829035282135, "global_step": 43558, "epoch": 489, "lr": 9.389691304087074e-05}
{"train_loss": 0.2835354506969452, "global_step": 43559, "epoch": 489, "lr": 9.389663547518797e-05}
{"train_loss": 0.2919907867908478, "global_step": 43560, "epoch": 489, "lr": 9.389635790360383e-05}
{"train_loss": 0.2165403813123703, "global_step": 43561, "epoch": 489, "lr": 9.389608032611834e-05}
{"train_loss": 0.24518422782421112, "global_step": 43562, "epoch": 489, "lr": 9.389580274273158e-05}
{"train_loss": 0.25111550092697144, "global_step": 43563, "epoch": 489, "lr": 9.389552515344355e-05}
{"train_loss": 0.39352095127105713, "global_step": 43564, "epoch": 489, "lr": 9.389524755825431e-05}
{"train_loss": 0.2585916221141815, "global_step": 43565, "epoch": 489, "lr": 9.389496995716388e-05}
{"train_loss": 0.20485049486160278, "global_step": 43566, "epoch": 489, "lr": 9.389469235017232e-05}
{"train_loss": 0.3118007779121399, "global_step": 43567, "epoch": 489, "lr": 9.389441473727964e-05}
{"train_loss": 0.3015279173851013, "global_step": 43568, "epoch": 489, "lr": 9.38941371184859e-05}
{"train_loss": 0.29484570026397705, "global_step": 43569, "epoch": 489, "lr": 9.389385949379112e-05}
{"train_loss": 0.2943059802055359, "global_step": 43570, "epoch": 489, "lr": 9.389358186319535e-05}
{"train_loss": 0.33790990710258484, "global_step": 43571, "epoch": 489, "lr": 9.389330422669862e-05}
{"train_loss": 0.18576782941818237, "global_step": 43572, "epoch": 489, "lr": 9.389302658430097e-05}
{"train_loss": 0.356751024723053, "global_step": 43573, "epoch": 489, "lr": 9.389274893600244e-05}
{"train_loss": 0.3964063227176666, "global_step": 43574, "epoch": 489, "lr": 9.389247128180307e-05}
{"train_loss": 0.3189193904399872, "global_step": 43575, "epoch": 489, "lr": 9.389219362170289e-05}
{"train_loss": 0.31709563732147217, "global_step": 43576, "epoch": 489, "lr": 9.389191595570192e-05}
{"train_loss": 0.2964279353618622, "global_step": 43577, "epoch": 489, "lr": 9.389163828380026e-05}
{"train_loss": 0.2788255512714386, "global_step": 43578, "epoch": 489, "lr": 9.389136060599786e-05}
{"train_loss": 0.3068459630012512, "global_step": 43579, "epoch": 489, "lr": 9.389108292229483e-05}
{"train_loss": 0.336806058883667, "global_step": 43580, "epoch": 489, "lr": 9.389080523269116e-05}
{"train_loss": 0.2899840176105499, "global_step": 43581, "epoch": 489, "lr": 9.389052753718692e-05}
{"train_loss": 0.2942808270454407, "global_step": 43582, "epoch": 489, "lr": 9.389024983578212e-05}
{"train_loss": 0.4458130896091461, "global_step": 43583, "epoch": 489, "lr": 9.388997212847683e-05}
{"train_loss": 0.24919095635414124, "global_step": 43584, "epoch": 489, "lr": 9.388969441527105e-05}
{"train_loss": 0.38185757398605347, "global_step": 43585, "epoch": 489, "lr": 9.388941669616485e-05}
{"train_loss": 0.3331824839115143, "global_step": 43586, "epoch": 489, "lr": 9.388913897115824e-05}
{"train_loss": 0.32036644220352173, "global_step": 43587, "epoch": 489, "lr": 9.388886124025127e-05}
{"train_loss": 0.3019576370716095, "global_step": 43588, "epoch": 489, "lr": 9.388858350344399e-05}
{"train_loss": 0.32779568433761597, "global_step": 43589, "epoch": 489, "lr": 9.388830576073641e-05}
{"train_loss": 0.2817007303237915, "global_step": 43590, "epoch": 489, "lr": 9.388802801212859e-05}
{"train_loss": 0.27678579092025757, "global_step": 43591, "epoch": 489, "lr": 9.388775025762056e-05}
{"train_loss": 0.2832445204257965, "global_step": 43592, "epoch": 489, "lr": 9.388747249721237e-05}
{"train_loss": 0.29733267426490784, "global_step": 43593, "epoch": 489, "lr": 9.388719473090403e-05}
{"train_loss": 0.3743864893913269, "global_step": 43594, "epoch": 489, "lr": 9.388691695869558e-05}
{"train_loss": 0.33510738611221313, "global_step": 43595, "epoch": 489, "lr": 9.388663918058708e-05}
{"train_loss": 0.32197532057762146, "global_step": 43596, "epoch": 489, "lr": 9.388636139657856e-05}
{"train_loss": 0.3508839011192322, "global_step": 43597, "epoch": 489, "lr": 9.388608360667006e-05}
{"train_loss": 0.5230914354324341, "global_step": 43598, "epoch": 489, "lr": 9.38858058108616e-05}
{"train_loss": 0.4011768400669098, "global_step": 43599, "epoch": 489, "lr": 9.388552800915323e-05}
{"train_loss": 0.29180896282196045, "global_step": 43600, "epoch": 489, "lr": 9.388525020154499e-05}
{"train_loss": 0.3219265341758728, "global_step": 43601, "epoch": 489, "lr": 9.388497238803691e-05}
{"train_loss": 0.2554924190044403, "global_step": 43602, "epoch": 489, "lr": 9.388469456862902e-05}
{"train_loss": 0.33350303769111633, "global_step": 43603, "epoch": 489, "lr": 9.388441674332138e-05}
{"train_loss": 0.25319451093673706, "global_step": 43604, "epoch": 489, "lr": 9.388413891211401e-05}
{"train_loss": 0.28725001215934753, "global_step": 43605, "epoch": 489, "lr": 9.388386107500696e-05}
{"train_loss": 0.22267358005046844, "global_step": 43606, "epoch": 489, "lr": 9.388358323200026e-05}
{"train_loss": 0.3366391062736511, "global_step": 43607, "epoch": 489, "lr": 9.388330538309393e-05}
{"train_loss": 0.34457018971443176, "global_step": 43608, "epoch": 489, "lr": 9.388302752828803e-05}
{"train_loss": 0.30179581732562416, "global_step": 43609, "epoch": 489, "lr": 9.388274966758262e-05, "val_loss": 2.7559375762939453}
{"train_loss": 0.31808069348335266, "global_step": 43610, "epoch": 490, "lr": 9.388247180097767e-05}
{"train_loss": 0.25640398263931274, "global_step": 43611, "epoch": 490, "lr": 9.388219392847329e-05}
{"train_loss": 0.30178943276405334, "global_step": 43612, "epoch": 490, "lr": 9.388191605006946e-05}
{"train_loss": 0.23918913304805756, "global_step": 43613, "epoch": 490, "lr": 9.388163816576625e-05}
{"train_loss": 0.2618584930896759, "global_step": 43614, "epoch": 490, "lr": 9.388136027556369e-05}
{"train_loss": 0.3709348440170288, "global_step": 43615, "epoch": 490, "lr": 9.38810823794618e-05}
{"train_loss": 0.3567509353160858, "global_step": 43616, "epoch": 490, "lr": 9.388080447746066e-05}
{"train_loss": 0.3837724030017853, "global_step": 43617, "epoch": 490, "lr": 9.388052656956026e-05}
{"train_loss": 0.28494516015052795, "global_step": 43618, "epoch": 490, "lr": 9.388024865576066e-05}
{"train_loss": 0.28934457898139954, "global_step": 43619, "epoch": 490, "lr": 9.387997073606191e-05}
{"train_loss": 0.3724774718284607, "global_step": 43620, "epoch": 490, "lr": 9.387969281046403e-05}
{"train_loss": 0.3206928074359894, "global_step": 43621, "epoch": 490, "lr": 9.387941487896704e-05}
{"train_loss": 0.35441505908966064, "global_step": 43622, "epoch": 490, "lr": 9.387913694157103e-05}
{"train_loss": 0.2947057783603668, "global_step": 43623, "epoch": 490, "lr": 9.387885899827598e-05}
{"train_loss": 0.3098681569099426, "global_step": 43624, "epoch": 490, "lr": 9.387858104908196e-05}
{"train_loss": 0.28933820128440857, "global_step": 43625, "epoch": 490, "lr": 9.387830309398899e-05}
{"train_loss": 0.297692209482193, "global_step": 43626, "epoch": 490, "lr": 9.387802513299712e-05}
{"train_loss": 0.2655012309551239, "global_step": 43627, "epoch": 490, "lr": 9.38777471661064e-05}
{"train_loss": 0.3471226692199707, "global_step": 43628, "epoch": 490, "lr": 9.387746919331683e-05}
{"train_loss": 0.22801409661769867, "global_step": 43629, "epoch": 490, "lr": 9.387719121462848e-05}
{"train_loss": 0.19436587393283844, "global_step": 43630, "epoch": 490, "lr": 9.387691323004137e-05}
{"train_loss": 0.20175322890281677, "global_step": 43631, "epoch": 490, "lr": 9.387663523955556e-05}
{"train_loss": 0.3285978436470032, "global_step": 43632, "epoch": 490, "lr": 9.387635724317105e-05}
{"train_loss": 0.3017810583114624, "global_step": 43633, "epoch": 490, "lr": 9.387607924088792e-05}
{"train_loss": 0.4127715528011322, "global_step": 43634, "epoch": 490, "lr": 9.387580123270618e-05}
{"train_loss": 0.3358771502971649, "global_step": 43635, "epoch": 490, "lr": 9.387552321862587e-05}
{"train_loss": 0.37881287932395935, "global_step": 43636, "epoch": 490, "lr": 9.387524519864702e-05}
{"train_loss": 0.3054046034812927, "global_step": 43637, "epoch": 490, "lr": 9.387496717276969e-05}
{"train_loss": 0.25632765889167786, "global_step": 43638, "epoch": 490, "lr": 9.387468914099391e-05}
{"train_loss": 0.2043718546628952, "global_step": 43639, "epoch": 490, "lr": 9.38744111033197e-05}
{"train_loss": 0.2129848450422287, "global_step": 43640, "epoch": 490, "lr": 9.387413305974712e-05}
{"train_loss": 0.2361622005701065, "global_step": 43641, "epoch": 490, "lr": 9.387385501027619e-05}
{"train_loss": 0.2809647023677826, "global_step": 43642, "epoch": 490, "lr": 9.387357695490695e-05}
{"train_loss": 0.20806613564491272, "global_step": 43643, "epoch": 490, "lr": 9.387329889363945e-05}
{"train_loss": 0.28170809149742126, "global_step": 43644, "epoch": 490, "lr": 9.387302082647374e-05}
{"train_loss": 0.2776852250099182, "global_step": 43645, "epoch": 490, "lr": 9.38727427534098e-05}
{"train_loss": 0.2604183256626129, "global_step": 43646, "epoch": 490, "lr": 9.387246467444772e-05}
{"train_loss": 0.3227997422218323, "global_step": 43647, "epoch": 490, "lr": 9.387218658958754e-05}
{"train_loss": 0.387119859457016, "global_step": 43648, "epoch": 490, "lr": 9.387190849882925e-05}
{"train_loss": 0.35374143719673157, "global_step": 43649, "epoch": 490, "lr": 9.387163040217292e-05}
{"train_loss": 0.18594302237033844, "global_step": 43650, "epoch": 490, "lr": 9.38713522996186e-05}
{"train_loss": 0.2594640552997589, "global_step": 43651, "epoch": 490, "lr": 9.387107419116632e-05}
{"train_loss": 0.3416531980037689, "global_step": 43652, "epoch": 490, "lr": 9.387079607681607e-05}
{"train_loss": 0.2745753526687622, "global_step": 43653, "epoch": 490, "lr": 9.387051795656797e-05}
{"train_loss": 0.32400843501091003, "global_step": 43654, "epoch": 490, "lr": 9.3870239830422e-05}
{"train_loss": 0.27405640482902527, "global_step": 43655, "epoch": 490, "lr": 9.38699616983782e-05}
{"train_loss": 0.24258193373680115, "global_step": 43656, "epoch": 490, "lr": 9.386968356043661e-05}
{"train_loss": 0.22210322320461273, "global_step": 43657, "epoch": 490, "lr": 9.38694054165973e-05}
{"train_loss": 0.32521089911460876, "global_step": 43658, "epoch": 490, "lr": 9.386912726686026e-05}
{"train_loss": 0.22915560007095337, "global_step": 43659, "epoch": 490, "lr": 9.386884911122558e-05}
{"train_loss": 0.3476424813270569, "global_step": 43660, "epoch": 490, "lr": 9.386857094969323e-05}
{"train_loss": 0.3294501006603241, "global_step": 43661, "epoch": 490, "lr": 9.386829278226332e-05}
{"train_loss": 0.3562971353530884, "global_step": 43662, "epoch": 490, "lr": 9.386801460893584e-05}
{"train_loss": 0.27599677443504333, "global_step": 43663, "epoch": 490, "lr": 9.386773642971083e-05}
{"train_loss": 0.33149051666259766, "global_step": 43664, "epoch": 490, "lr": 9.386745824458835e-05}
{"train_loss": 0.30114343762397766, "global_step": 43665, "epoch": 490, "lr": 9.386718005356842e-05}
{"train_loss": 0.3555828034877777, "global_step": 43666, "epoch": 490, "lr": 9.386690185665108e-05}
{"train_loss": 0.3084733188152313, "global_step": 43667, "epoch": 490, "lr": 9.386662365383638e-05}
{"train_loss": 0.3418269157409668, "global_step": 43668, "epoch": 490, "lr": 9.386634544512434e-05}
{"train_loss": 0.2537584602832794, "global_step": 43669, "epoch": 490, "lr": 9.3866067230515e-05}
{"train_loss": 0.23809130489826202, "global_step": 43670, "epoch": 490, "lr": 9.386578901000841e-05}
{"train_loss": 0.30182844400405884, "global_step": 43671, "epoch": 490, "lr": 9.386551078360459e-05}
{"train_loss": 0.2612302005290985, "global_step": 43672, "epoch": 490, "lr": 9.38652325513036e-05}
{"train_loss": 0.29266858100891113, "global_step": 43673, "epoch": 490, "lr": 9.386495431310546e-05}
{"train_loss": 0.4217202067375183, "global_step": 43674, "epoch": 490, "lr": 9.38646760690102e-05}
{"train_loss": 0.30872026085853577, "global_step": 43675, "epoch": 490, "lr": 9.386439781901788e-05}
{"train_loss": 0.3289787471294403, "global_step": 43676, "epoch": 490, "lr": 9.386411956312853e-05}
{"train_loss": 0.31743666529655457, "global_step": 43677, "epoch": 490, "lr": 9.386384130134217e-05}
{"train_loss": 0.2237323522567749, "global_step": 43678, "epoch": 490, "lr": 9.386356303365886e-05}
{"train_loss": 0.28828391432762146, "global_step": 43679, "epoch": 490, "lr": 9.386328476007861e-05}
{"train_loss": 0.3210352957248688, "global_step": 43680, "epoch": 490, "lr": 9.386300648060152e-05}
{"train_loss": 0.2901672422885895, "global_step": 43681, "epoch": 490, "lr": 9.386272819522755e-05}
{"train_loss": 0.292644739151001, "global_step": 43682, "epoch": 490, "lr": 9.386244990395677e-05}
{"train_loss": 0.3780791163444519, "global_step": 43683, "epoch": 490, "lr": 9.386217160678923e-05}
{"train_loss": 0.3322466313838959, "global_step": 43684, "epoch": 490, "lr": 9.386189330372495e-05}
{"train_loss": 0.35317692160606384, "global_step": 43685, "epoch": 490, "lr": 9.386161499476396e-05}
{"train_loss": 0.2773679792881012, "global_step": 43686, "epoch": 490, "lr": 9.386133667990633e-05}
{"train_loss": 0.26967504620552063, "global_step": 43687, "epoch": 490, "lr": 9.386105835915208e-05}
{"train_loss": 0.34393686056137085, "global_step": 43688, "epoch": 490, "lr": 9.386078003250124e-05}
{"train_loss": 0.2586473822593689, "global_step": 43689, "epoch": 490, "lr": 9.386050169995384e-05}
{"train_loss": 0.2687404453754425, "global_step": 43690, "epoch": 490, "lr": 9.386022336150995e-05}
{"train_loss": 0.2559002637863159, "global_step": 43691, "epoch": 490, "lr": 9.385994501716958e-05}
{"train_loss": 0.27170106768608093, "global_step": 43692, "epoch": 490, "lr": 9.385966666693278e-05}
{"train_loss": 0.28049030900001526, "global_step": 43693, "epoch": 490, "lr": 9.385938831079957e-05}
{"train_loss": 0.24463219940662384, "global_step": 43694, "epoch": 490, "lr": 9.385910994877e-05}
{"train_loss": 0.2777901887893677, "global_step": 43695, "epoch": 490, "lr": 9.385883158084411e-05}
{"train_loss": 0.3033643662929535, "global_step": 43696, "epoch": 490, "lr": 9.385855320702193e-05}
{"train_loss": 0.2346847802400589, "global_step": 43697, "epoch": 490, "lr": 9.385827482730351e-05}
{"train_loss": 0.29603317296237086, "global_step": 43698, "epoch": 490, "lr": 9.385799644168889e-05, "val_loss": 2.841660976409912, "train_action_mse_error": 25.690826416015625}
{"train_loss": 0.21422775089740753, "global_step": 43699, "epoch": 491, "lr": 9.385771805017809e-05}
{"train_loss": 0.32498034834861755, "global_step": 43700, "epoch": 491, "lr": 9.385743965277115e-05}
{"train_loss": 0.26376473903656006, "global_step": 43701, "epoch": 491, "lr": 9.38571612494681e-05}
{"train_loss": 0.3263953626155853, "global_step": 43702, "epoch": 491, "lr": 9.3856882840269e-05}
{"train_loss": 0.20640119910240173, "global_step": 43703, "epoch": 491, "lr": 9.385660442517388e-05}
{"train_loss": 0.3178155720233917, "global_step": 43704, "epoch": 491, "lr": 9.385632600418277e-05}
{"train_loss": 0.23645544052124023, "global_step": 43705, "epoch": 491, "lr": 9.385604757729572e-05}
{"train_loss": 0.3277113437652588, "global_step": 43706, "epoch": 491, "lr": 9.385576914451276e-05}
{"train_loss": 0.2863706350326538, "global_step": 43707, "epoch": 491, "lr": 9.38554907058339e-05}
{"train_loss": 0.28934618830680847, "global_step": 43708, "epoch": 491, "lr": 9.385521226125923e-05}
{"train_loss": 0.34839025139808655, "global_step": 43709, "epoch": 491, "lr": 9.385493381078876e-05}
{"train_loss": 0.22683709859848022, "global_step": 43710, "epoch": 491, "lr": 9.385465535442252e-05}
{"train_loss": 0.36074966192245483, "global_step": 43711, "epoch": 491, "lr": 9.385437689216056e-05}
{"train_loss": 0.2803616225719452, "global_step": 43712, "epoch": 491, "lr": 9.38540984240029e-05}
{"train_loss": 0.30510973930358887, "global_step": 43713, "epoch": 491, "lr": 9.385381994994962e-05}
{"train_loss": 0.44319555163383484, "global_step": 43714, "epoch": 491, "lr": 9.38535414700007e-05}
{"train_loss": 0.39937081933021545, "global_step": 43715, "epoch": 491, "lr": 9.385326298415622e-05}
{"train_loss": 0.23828956484794617, "global_step": 43716, "epoch": 491, "lr": 9.38529844924162e-05}
{"train_loss": 0.3425079882144928, "global_step": 43717, "epoch": 491, "lr": 9.385270599478068e-05}
{"train_loss": 0.2918679714202881, "global_step": 43718, "epoch": 491, "lr": 9.38524274912497e-05}
{"train_loss": 0.22539874911308289, "global_step": 43719, "epoch": 491, "lr": 9.385214898182329e-05}
{"train_loss": 0.22144350409507751, "global_step": 43720, "epoch": 491, "lr": 9.38518704665015e-05}
{"train_loss": 0.3931479752063751, "global_step": 43721, "epoch": 491, "lr": 9.385159194528437e-05}
{"train_loss": 0.3607161045074463, "global_step": 43722, "epoch": 491, "lr": 9.38513134181719e-05}
{"train_loss": 0.22256210446357727, "global_step": 43723, "epoch": 491, "lr": 9.385103488516418e-05}
{"train_loss": 0.29100748896598816, "global_step": 43724, "epoch": 491, "lr": 9.385075634626122e-05}
{"train_loss": 0.29294633865356445, "global_step": 43725, "epoch": 491, "lr": 9.385047780146306e-05}
{"train_loss": 0.31748145818710327, "global_step": 43726, "epoch": 491, "lr": 9.385019925076972e-05}
{"train_loss": 0.29913946986198425, "global_step": 43727, "epoch": 491, "lr": 9.384992069418127e-05}
{"train_loss": 0.3055127263069153, "global_step": 43728, "epoch": 491, "lr": 9.384964213169774e-05}
{"train_loss": 0.37554970383644104, "global_step": 43729, "epoch": 491, "lr": 9.384936356331914e-05}
{"train_loss": 0.2817259132862091, "global_step": 43730, "epoch": 491, "lr": 9.384908498904555e-05}
{"train_loss": 0.36926424503326416, "global_step": 43731, "epoch": 491, "lr": 9.384880640887697e-05}
{"train_loss": 0.30007457733154297, "global_step": 43732, "epoch": 491, "lr": 9.384852782281344e-05}
{"train_loss": 0.23554526269435883, "global_step": 43733, "epoch": 491, "lr": 9.384824923085503e-05}
{"train_loss": 0.2572520673274994, "global_step": 43734, "epoch": 491, "lr": 9.384797063300175e-05}
{"train_loss": 0.29712459444999695, "global_step": 43735, "epoch": 491, "lr": 9.384769202925366e-05}
{"train_loss": 0.3074225187301636, "global_step": 43736, "epoch": 491, "lr": 9.384741341961076e-05}
{"train_loss": 0.347883015871048, "global_step": 43737, "epoch": 491, "lr": 9.384713480407312e-05}
{"train_loss": 0.3192146122455597, "global_step": 43738, "epoch": 491, "lr": 9.384685618264076e-05}
{"train_loss": 0.3253878057003021, "global_step": 43739, "epoch": 491, "lr": 9.384657755531374e-05}
{"train_loss": 0.37742340564727783, "global_step": 43740, "epoch": 491, "lr": 9.384629892209207e-05}
{"train_loss": 0.3457605838775635, "global_step": 43741, "epoch": 491, "lr": 9.38460202829758e-05}
{"train_loss": 0.30412545800209045, "global_step": 43742, "epoch": 491, "lr": 9.384574163796498e-05}
{"train_loss": 0.24597816169261932, "global_step": 43743, "epoch": 491, "lr": 9.384546298705961e-05}
{"train_loss": 0.4121239483356476, "global_step": 43744, "epoch": 491, "lr": 9.384518433025978e-05}
{"train_loss": 0.26612356305122375, "global_step": 43745, "epoch": 491, "lr": 9.384490566756547e-05}
{"train_loss": 0.23953059315681458, "global_step": 43746, "epoch": 491, "lr": 9.384462699897676e-05}
{"train_loss": 0.31338027119636536, "global_step": 43747, "epoch": 491, "lr": 9.38443483244937e-05}
{"train_loss": 0.34229737520217896, "global_step": 43748, "epoch": 491, "lr": 9.384406964411626e-05}
{"train_loss": 0.3648016154766083, "global_step": 43749, "epoch": 491, "lr": 9.384379095784455e-05}
{"train_loss": 0.2798173725605011, "global_step": 43750, "epoch": 491, "lr": 9.384351226567855e-05}
{"train_loss": 0.3147192895412445, "global_step": 43751, "epoch": 491, "lr": 9.384323356761835e-05}
{"train_loss": 0.31723618507385254, "global_step": 43752, "epoch": 491, "lr": 9.384295486366394e-05}
{"train_loss": 0.3218437135219574, "global_step": 43753, "epoch": 491, "lr": 9.384267615381539e-05}
{"train_loss": 0.31111782789230347, "global_step": 43754, "epoch": 491, "lr": 9.384239743807273e-05}
{"train_loss": 0.4239349067211151, "global_step": 43755, "epoch": 491, "lr": 9.384211871643597e-05}
{"train_loss": 0.26011645793914795, "global_step": 43756, "epoch": 491, "lr": 9.384183998890521e-05}
{"train_loss": 0.24368953704833984, "global_step": 43757, "epoch": 491, "lr": 9.384156125548043e-05}
{"train_loss": 0.28953468799591064, "global_step": 43758, "epoch": 491, "lr": 9.38412825161617e-05}
{"train_loss": 0.2762545645236969, "global_step": 43759, "epoch": 491, "lr": 9.384100377094902e-05}
{"train_loss": 0.23084932565689087, "global_step": 43760, "epoch": 491, "lr": 9.384072501984246e-05}
{"train_loss": 0.38176754117012024, "global_step": 43761, "epoch": 491, "lr": 9.384044626284205e-05}
{"train_loss": 0.2987993657588959, "global_step": 43762, "epoch": 491, "lr": 9.384016749994781e-05}
{"train_loss": 0.3288441300392151, "global_step": 43763, "epoch": 491, "lr": 9.383988873115983e-05}
{"train_loss": 0.2613145709037781, "global_step": 43764, "epoch": 491, "lr": 9.383960995647809e-05}
{"train_loss": 0.20205660164356232, "global_step": 43765, "epoch": 491, "lr": 9.383933117590264e-05}
{"train_loss": 0.3506326377391815, "global_step": 43766, "epoch": 491, "lr": 9.383905238943354e-05}
{"train_loss": 0.2480066567659378, "global_step": 43767, "epoch": 491, "lr": 9.383877359707082e-05}
{"train_loss": 0.33429738879203796, "global_step": 43768, "epoch": 491, "lr": 9.38384947988145e-05}
{"train_loss": 0.21522215008735657, "global_step": 43769, "epoch": 491, "lr": 9.383821599466462e-05}
{"train_loss": 0.36543458700180054, "global_step": 43770, "epoch": 491, "lr": 9.383793718462124e-05}
{"train_loss": 0.2677548825740814, "global_step": 43771, "epoch": 491, "lr": 9.383765836868439e-05}
{"train_loss": 0.22677284479141235, "global_step": 43772, "epoch": 491, "lr": 9.383737954685409e-05}
{"train_loss": 0.36976656317710876, "global_step": 43773, "epoch": 491, "lr": 9.383710071913039e-05}
{"train_loss": 0.2372765690088272, "global_step": 43774, "epoch": 491, "lr": 9.383682188551333e-05}
{"train_loss": 0.2353348582983017, "global_step": 43775, "epoch": 491, "lr": 9.383654304600293e-05}
{"train_loss": 0.32424843311309814, "global_step": 43776, "epoch": 491, "lr": 9.383626420059925e-05}
{"train_loss": 0.24336926639080048, "global_step": 43777, "epoch": 491, "lr": 9.383598534930233e-05}
{"train_loss": 0.24064169824123383, "global_step": 43778, "epoch": 491, "lr": 9.383570649211218e-05}
{"train_loss": 0.34759286046028137, "global_step": 43779, "epoch": 491, "lr": 9.383542762902887e-05}
{"train_loss": 0.43048009276390076, "global_step": 43780, "epoch": 491, "lr": 9.38351487600524e-05}
{"train_loss": 0.2932937443256378, "global_step": 43781, "epoch": 491, "lr": 9.383486988518284e-05}
{"train_loss": 0.3421979248523712, "global_step": 43782, "epoch": 491, "lr": 9.383459100442022e-05}
{"train_loss": 0.3061019778251648, "global_step": 43783, "epoch": 491, "lr": 9.383431211776457e-05}
{"train_loss": 0.34776219725608826, "global_step": 43784, "epoch": 491, "lr": 9.383403322521592e-05}
{"train_loss": 0.2543151378631592, "global_step": 43785, "epoch": 491, "lr": 9.383375432677432e-05}
{"train_loss": 0.32595378160476685, "global_step": 43786, "epoch": 491, "lr": 9.383347542243981e-05}
{"train_loss": 0.3016403925217939, "global_step": 43787, "epoch": 491, "lr": 9.383319651221243e-05, "val_loss": 2.849627733230591}
{"train_loss": 0.32111045718193054, "global_step": 43788, "epoch": 492, "lr": 9.383291759609221e-05}
{"train_loss": 0.3778555691242218, "global_step": 43789, "epoch": 492, "lr": 9.383263867407919e-05}
{"train_loss": 0.286873996257782, "global_step": 43790, "epoch": 492, "lr": 9.383235974617342e-05}
{"train_loss": 0.3617997169494629, "global_step": 43791, "epoch": 492, "lr": 9.383208081237489e-05}
{"train_loss": 0.3589307963848114, "global_step": 43792, "epoch": 492, "lr": 9.38318018726837e-05}
{"train_loss": 0.2580506205558777, "global_step": 43793, "epoch": 492, "lr": 9.383152292709983e-05}
{"train_loss": 0.32637497782707214, "global_step": 43794, "epoch": 492, "lr": 9.383124397562337e-05}
{"train_loss": 0.33972305059432983, "global_step": 43795, "epoch": 492, "lr": 9.383096501825433e-05}
{"train_loss": 0.21786852180957794, "global_step": 43796, "epoch": 492, "lr": 9.383068605499275e-05}
{"train_loss": 0.24704615771770477, "global_step": 43797, "epoch": 492, "lr": 9.383040708583866e-05}
{"train_loss": 0.2887667715549469, "global_step": 43798, "epoch": 492, "lr": 9.38301281107921e-05}
{"train_loss": 0.35814595222473145, "global_step": 43799, "epoch": 492, "lr": 9.382984912985313e-05}
{"train_loss": 0.3130618631839752, "global_step": 43800, "epoch": 492, "lr": 9.382957014302176e-05}
{"train_loss": 0.33774423599243164, "global_step": 43801, "epoch": 492, "lr": 9.382929115029804e-05}
{"train_loss": 0.3636620044708252, "global_step": 43802, "epoch": 492, "lr": 9.382901215168202e-05}
{"train_loss": 0.3463815450668335, "global_step": 43803, "epoch": 492, "lr": 9.38287331471737e-05}
{"train_loss": 0.17696039378643036, "global_step": 43804, "epoch": 492, "lr": 9.382845413677316e-05}
{"train_loss": 0.2818433940410614, "global_step": 43805, "epoch": 492, "lr": 9.382817512048041e-05}
{"train_loss": 0.39765310287475586, "global_step": 43806, "epoch": 492, "lr": 9.38278960982955e-05}
{"train_loss": 0.3060992360115051, "global_step": 43807, "epoch": 492, "lr": 9.382761707021847e-05}
{"train_loss": 0.3398931324481964, "global_step": 43808, "epoch": 492, "lr": 9.382733803624935e-05}
{"train_loss": 0.2899394631385803, "global_step": 43809, "epoch": 492, "lr": 9.382705899638816e-05}
{"train_loss": 0.262004554271698, "global_step": 43810, "epoch": 492, "lr": 9.382677995063497e-05}
{"train_loss": 0.3328383266925812, "global_step": 43811, "epoch": 492, "lr": 9.38265008989898e-05}
{"train_loss": 0.33763518929481506, "global_step": 43812, "epoch": 492, "lr": 9.382622184145268e-05}
{"train_loss": 0.2088378220796585, "global_step": 43813, "epoch": 492, "lr": 9.382594277802368e-05}
{"train_loss": 0.3072710931301117, "global_step": 43814, "epoch": 492, "lr": 9.38256637087028e-05}
{"train_loss": 0.32017961144447327, "global_step": 43815, "epoch": 492, "lr": 9.382538463349011e-05}
{"train_loss": 0.17547455430030823, "global_step": 43816, "epoch": 492, "lr": 9.382510555238563e-05}
{"train_loss": 0.21116404235363007, "global_step": 43817, "epoch": 492, "lr": 9.38248264653894e-05}
{"train_loss": 0.3277364671230316, "global_step": 43818, "epoch": 492, "lr": 9.382454737250144e-05}
{"train_loss": 0.32632067799568176, "global_step": 43819, "epoch": 492, "lr": 9.38242682737218e-05}
{"train_loss": 0.26372429728507996, "global_step": 43820, "epoch": 492, "lr": 9.382398916905053e-05}
{"train_loss": 0.40186843276023865, "global_step": 43821, "epoch": 492, "lr": 9.382371005848767e-05}
{"train_loss": 0.32731497287750244, "global_step": 43822, "epoch": 492, "lr": 9.382343094203322e-05}
{"train_loss": 0.2309410125017166, "global_step": 43823, "epoch": 492, "lr": 9.382315181968729e-05}
{"train_loss": 0.354265034198761, "global_step": 43824, "epoch": 492, "lr": 9.382287269144982e-05}
{"train_loss": 0.29060792922973633, "global_step": 43825, "epoch": 492, "lr": 9.382259355732093e-05}
{"train_loss": 0.2589534819126129, "global_step": 43826, "epoch": 492, "lr": 9.382231441730062e-05}
{"train_loss": 0.3175961971282959, "global_step": 43827, "epoch": 492, "lr": 9.382203527138893e-05}
{"train_loss": 0.37804874777793884, "global_step": 43828, "epoch": 492, "lr": 9.38217561195859e-05}
{"train_loss": 0.21078498661518097, "global_step": 43829, "epoch": 492, "lr": 9.382147696189158e-05}
{"train_loss": 0.24013906717300415, "global_step": 43830, "epoch": 492, "lr": 9.382119779830598e-05}
{"train_loss": 0.23302248120307922, "global_step": 43831, "epoch": 492, "lr": 9.382091862882917e-05}
{"train_loss": 0.4070741534233093, "global_step": 43832, "epoch": 492, "lr": 9.382063945346116e-05}
{"train_loss": 0.42629221081733704, "global_step": 43833, "epoch": 492, "lr": 9.3820360272202e-05}
{"train_loss": 0.2823065519332886, "global_step": 43834, "epoch": 492, "lr": 9.382008108505174e-05}
{"train_loss": 0.30052632093429565, "global_step": 43835, "epoch": 492, "lr": 9.38198018920104e-05}
{"train_loss": 0.28499066829681396, "global_step": 43836, "epoch": 492, "lr": 9.3819522693078e-05}
{"train_loss": 0.31599631905555725, "global_step": 43837, "epoch": 492, "lr": 9.381924348825463e-05}
{"train_loss": 0.2725576162338257, "global_step": 43838, "epoch": 492, "lr": 9.381896427754028e-05}
{"train_loss": 0.37983423471450806, "global_step": 43839, "epoch": 492, "lr": 9.381868506093501e-05}
{"train_loss": 0.24894094467163086, "global_step": 43840, "epoch": 492, "lr": 9.381840583843884e-05}
{"train_loss": 0.3599889278411865, "global_step": 43841, "epoch": 492, "lr": 9.381812661005183e-05}
{"train_loss": 0.30799704790115356, "global_step": 43842, "epoch": 492, "lr": 9.3817847375774e-05}
{"train_loss": 0.29773759841918945, "global_step": 43843, "epoch": 492, "lr": 9.38175681356054e-05}
{"train_loss": 0.30954211950302124, "global_step": 43844, "epoch": 492, "lr": 9.381728888954606e-05}
{"train_loss": 0.35223081707954407, "global_step": 43845, "epoch": 492, "lr": 9.381700963759603e-05}
{"train_loss": 0.32017940282821655, "global_step": 43846, "epoch": 492, "lr": 9.381673037975533e-05}
{"train_loss": 0.28967082500457764, "global_step": 43847, "epoch": 492, "lr": 9.3816451116024e-05}
{"train_loss": 0.20021003484725952, "global_step": 43848, "epoch": 492, "lr": 9.381617184640208e-05}
{"train_loss": 0.248223215341568, "global_step": 43849, "epoch": 492, "lr": 9.38158925708896e-05}
{"train_loss": 0.3875831365585327, "global_step": 43850, "epoch": 492, "lr": 9.381561328948664e-05}
{"train_loss": 0.35540321469306946, "global_step": 43851, "epoch": 492, "lr": 9.381533400219318e-05}
{"train_loss": 0.29412660002708435, "global_step": 43852, "epoch": 492, "lr": 9.381505470900929e-05}
{"train_loss": 0.2866629660129547, "global_step": 43853, "epoch": 492, "lr": 9.3814775409935e-05}
{"train_loss": 0.31707704067230225, "global_step": 43854, "epoch": 492, "lr": 9.381449610497036e-05}
{"train_loss": 0.42271173000335693, "global_step": 43855, "epoch": 492, "lr": 9.381421679411537e-05}
{"train_loss": 0.1690772920846939, "global_step": 43856, "epoch": 492, "lr": 9.381393747737011e-05}
{"train_loss": 0.38735681772232056, "global_step": 43857, "epoch": 492, "lr": 9.38136581547346e-05}
{"train_loss": 0.2616291642189026, "global_step": 43858, "epoch": 492, "lr": 9.381337882620886e-05}
{"train_loss": 0.3012339174747467, "global_step": 43859, "epoch": 492, "lr": 9.381309949179296e-05}
{"train_loss": 0.2503153085708618, "global_step": 43860, "epoch": 492, "lr": 9.381282015148692e-05}
{"train_loss": 0.3920946419239044, "global_step": 43861, "epoch": 492, "lr": 9.381254080529078e-05}
{"train_loss": 0.3210216164588928, "global_step": 43862, "epoch": 492, "lr": 9.381226145320457e-05}
{"train_loss": 0.3305756747722626, "global_step": 43863, "epoch": 492, "lr": 9.381198209522834e-05}
{"train_loss": 0.24305161833763123, "global_step": 43864, "epoch": 492, "lr": 9.381170273136214e-05}
{"train_loss": 0.3343437612056732, "global_step": 43865, "epoch": 492, "lr": 9.381142336160597e-05}
{"train_loss": 0.2760562002658844, "global_step": 43866, "epoch": 492, "lr": 9.381114398595989e-05}
{"train_loss": 0.2549748718738556, "global_step": 43867, "epoch": 492, "lr": 9.381086460442394e-05}
{"train_loss": 0.23858924210071564, "global_step": 43868, "epoch": 492, "lr": 9.381058521699816e-05}
{"train_loss": 0.2246447205543518, "global_step": 43869, "epoch": 492, "lr": 9.381030582368258e-05}
{"train_loss": 0.27685546875, "global_step": 43870, "epoch": 492, "lr": 9.381002642447723e-05}
{"train_loss": 0.23840151727199554, "global_step": 43871, "epoch": 492, "lr": 9.380974701938217e-05}
{"train_loss": 0.3009505867958069, "global_step": 43872, "epoch": 492, "lr": 9.38094676083974e-05}
{"train_loss": 0.22370503842830658, "global_step": 43873, "epoch": 492, "lr": 9.3809188191523e-05}
{"train_loss": 0.22980903089046478, "global_step": 43874, "epoch": 492, "lr": 9.380890876875898e-05}
{"train_loss": 0.3034862279891968, "global_step": 43875, "epoch": 492, "lr": 9.380862934010539e-05}
{"train_loss": 0.30169103319725293, "global_step": 43876, "epoch": 492, "lr": 9.380834990556227e-05, "val_loss": 2.8918955326080322}
{"train_loss": 0.28693798184394836, "global_step": 43877, "epoch": 493, "lr": 9.380807046512963e-05}
{"train_loss": 0.2185947299003601, "global_step": 43878, "epoch": 493, "lr": 9.380779101880755e-05}
{"train_loss": 0.16524915397167206, "global_step": 43879, "epoch": 493, "lr": 9.380751156659604e-05}
{"train_loss": 0.25010308623313904, "global_step": 43880, "epoch": 493, "lr": 9.380723210849514e-05}
{"train_loss": 0.334918737411499, "global_step": 43881, "epoch": 493, "lr": 9.38069526445049e-05}
{"train_loss": 0.21661752462387085, "global_step": 43882, "epoch": 493, "lr": 9.380667317462534e-05}
{"train_loss": 0.22382256388664246, "global_step": 43883, "epoch": 493, "lr": 9.38063936988565e-05}
{"train_loss": 0.3480888307094574, "global_step": 43884, "epoch": 493, "lr": 9.380611421719845e-05}
{"train_loss": 0.23860429227352142, "global_step": 43885, "epoch": 493, "lr": 9.380583472965118e-05}
{"train_loss": 0.39610230922698975, "global_step": 43886, "epoch": 493, "lr": 9.380555523621475e-05}
{"train_loss": 0.29766568541526794, "global_step": 43887, "epoch": 493, "lr": 9.380527573688921e-05}
{"train_loss": 0.3196493685245514, "global_step": 43888, "epoch": 493, "lr": 9.380499623167459e-05}
{"train_loss": 0.2683759927749634, "global_step": 43889, "epoch": 493, "lr": 9.38047167205709e-05}
{"train_loss": 0.24857844412326813, "global_step": 43890, "epoch": 493, "lr": 9.380443720357822e-05}
{"train_loss": 0.21732798218727112, "global_step": 43891, "epoch": 493, "lr": 9.380415768069654e-05}
{"train_loss": 0.3214627206325531, "global_step": 43892, "epoch": 493, "lr": 9.380387815192596e-05}
{"train_loss": 0.2737031877040863, "global_step": 43893, "epoch": 493, "lr": 9.380359861726644e-05}
{"train_loss": 0.2436472326517105, "global_step": 43894, "epoch": 493, "lr": 9.380331907671809e-05}
{"train_loss": 0.3121490478515625, "global_step": 43895, "epoch": 493, "lr": 9.380303953028091e-05}
{"train_loss": 0.2854415774345398, "global_step": 43896, "epoch": 493, "lr": 9.380275997795495e-05}
{"train_loss": 0.2510901987552643, "global_step": 43897, "epoch": 493, "lr": 9.380248041974025e-05}
{"train_loss": 0.2957639694213867, "global_step": 43898, "epoch": 493, "lr": 9.380220085563682e-05}
{"train_loss": 0.35554248094558716, "global_step": 43899, "epoch": 493, "lr": 9.380192128564473e-05}
{"train_loss": 0.3133150339126587, "global_step": 43900, "epoch": 493, "lr": 9.3801641709764e-05}
{"train_loss": 0.21778276562690735, "global_step": 43901, "epoch": 493, "lr": 9.380136212799468e-05}
{"train_loss": 0.25336772203445435, "global_step": 43902, "epoch": 493, "lr": 9.38010825403368e-05}
{"train_loss": 0.31502217054367065, "global_step": 43903, "epoch": 493, "lr": 9.380080294679039e-05}
{"train_loss": 0.2830730676651001, "global_step": 43904, "epoch": 493, "lr": 9.38005233473555e-05}
{"train_loss": 0.30912962555885315, "global_step": 43905, "epoch": 493, "lr": 9.380024374203216e-05}
{"train_loss": 0.269899845123291, "global_step": 43906, "epoch": 493, "lr": 9.379996413082043e-05}
{"train_loss": 0.1810224950313568, "global_step": 43907, "epoch": 493, "lr": 9.379968451372031e-05}
{"train_loss": 0.4793532192707062, "global_step": 43908, "epoch": 493, "lr": 9.379940489073186e-05}
{"train_loss": 0.28907310962677, "global_step": 43909, "epoch": 493, "lr": 9.379912526185513e-05}
{"train_loss": 0.2965860962867737, "global_step": 43910, "epoch": 493, "lr": 9.379884562709012e-05}
{"train_loss": 0.2513405382633209, "global_step": 43911, "epoch": 493, "lr": 9.379856598643688e-05}
{"train_loss": 0.2774658203125, "global_step": 43912, "epoch": 493, "lr": 9.379828633989548e-05}
{"train_loss": 0.2917148470878601, "global_step": 43913, "epoch": 493, "lr": 9.379800668746593e-05}
{"train_loss": 0.34922534227371216, "global_step": 43914, "epoch": 493, "lr": 9.379772702914827e-05}
{"train_loss": 0.284327894449234, "global_step": 43915, "epoch": 493, "lr": 9.379744736494253e-05}
{"train_loss": 0.2719617784023285, "global_step": 43916, "epoch": 493, "lr": 9.379716769484877e-05}
{"train_loss": 0.2830018997192383, "global_step": 43917, "epoch": 493, "lr": 9.379688801886701e-05}
{"train_loss": 0.15828926861286163, "global_step": 43918, "epoch": 493, "lr": 9.379660833699728e-05}
{"train_loss": 0.2920294404029846, "global_step": 43919, "epoch": 493, "lr": 9.379632864923965e-05}
{"train_loss": 0.331880658864975, "global_step": 43920, "epoch": 493, "lr": 9.379604895559413e-05}
{"train_loss": 0.36457380652427673, "global_step": 43921, "epoch": 493, "lr": 9.379576925606077e-05}
{"train_loss": 0.3498334288597107, "global_step": 43922, "epoch": 493, "lr": 9.379548955063959e-05}
{"train_loss": 0.26059985160827637, "global_step": 43923, "epoch": 493, "lr": 9.379520983933065e-05}
{"train_loss": 0.3554876744747162, "global_step": 43924, "epoch": 493, "lr": 9.379493012213397e-05}
{"train_loss": 0.32635819911956787, "global_step": 43925, "epoch": 493, "lr": 9.37946503990496e-05}
{"train_loss": 0.28370028734207153, "global_step": 43926, "epoch": 493, "lr": 9.379437067007758e-05}
{"train_loss": 0.2385302186012268, "global_step": 43927, "epoch": 493, "lr": 9.379409093521794e-05}
{"train_loss": 0.1925235241651535, "global_step": 43928, "epoch": 493, "lr": 9.37938111944707e-05}
{"train_loss": 0.28459933400154114, "global_step": 43929, "epoch": 493, "lr": 9.379353144783594e-05}
{"train_loss": 0.16973279416561127, "global_step": 43930, "epoch": 493, "lr": 9.379325169531365e-05}
{"train_loss": 0.3411259353160858, "global_step": 43931, "epoch": 493, "lr": 9.379297193690391e-05}
{"train_loss": 0.2688869833946228, "global_step": 43932, "epoch": 493, "lr": 9.379269217260672e-05}
{"train_loss": 0.28156396746635437, "global_step": 43933, "epoch": 493, "lr": 9.379241240242216e-05}
{"train_loss": 0.3765925467014313, "global_step": 43934, "epoch": 493, "lr": 9.379213262635022e-05}
{"train_loss": 0.3854841887950897, "global_step": 43935, "epoch": 493, "lr": 9.379185284439098e-05}
{"train_loss": 0.2762620449066162, "global_step": 43936, "epoch": 493, "lr": 9.379157305654445e-05}
{"train_loss": 0.2824946939945221, "global_step": 43937, "epoch": 493, "lr": 9.379129326281068e-05}
{"train_loss": 0.26026880741119385, "global_step": 43938, "epoch": 493, "lr": 9.379101346318971e-05}
{"train_loss": 0.3157930076122284, "global_step": 43939, "epoch": 493, "lr": 9.379073365768156e-05}
{"train_loss": 0.29518821835517883, "global_step": 43940, "epoch": 493, "lr": 9.379045384628629e-05}
{"train_loss": 0.40221115946769714, "global_step": 43941, "epoch": 493, "lr": 9.37901740290039e-05}
{"train_loss": 0.28940796852111816, "global_step": 43942, "epoch": 493, "lr": 9.378989420583448e-05}
{"train_loss": 0.2556022107601166, "global_step": 43943, "epoch": 493, "lr": 9.378961437677804e-05}
{"train_loss": 0.37104716897010803, "global_step": 43944, "epoch": 493, "lr": 9.378933454183463e-05}
{"train_loss": 0.27742335200309753, "global_step": 43945, "epoch": 493, "lr": 9.378905470100426e-05}
{"train_loss": 0.20433345437049866, "global_step": 43946, "epoch": 493, "lr": 9.378877485428699e-05}
{"train_loss": 0.22285515069961548, "global_step": 43947, "epoch": 493, "lr": 9.378849500168285e-05}
{"train_loss": 0.41071298718452454, "global_step": 43948, "epoch": 493, "lr": 9.37882151431919e-05}
{"train_loss": 0.29731571674346924, "global_step": 43949, "epoch": 493, "lr": 9.378793527881413e-05}
{"train_loss": 0.3306456506252289, "global_step": 43950, "epoch": 493, "lr": 9.378765540854963e-05}
{"train_loss": 0.32318055629730225, "global_step": 43951, "epoch": 493, "lr": 9.37873755323984e-05}
{"train_loss": 0.2708313763141632, "global_step": 43952, "epoch": 493, "lr": 9.37870956503605e-05}
{"train_loss": 0.28586921095848083, "global_step": 43953, "epoch": 493, "lr": 9.378681576243595e-05}
{"train_loss": 0.27399903535842896, "global_step": 43954, "epoch": 493, "lr": 9.378653586862481e-05}
{"train_loss": 0.320284903049469, "global_step": 43955, "epoch": 493, "lr": 9.378625596892707e-05}
{"train_loss": 0.23767676949501038, "global_step": 43956, "epoch": 493, "lr": 9.378597606334283e-05}
{"train_loss": 0.2284451574087143, "global_step": 43957, "epoch": 493, "lr": 9.37856961518721e-05}
{"train_loss": 0.3736550509929657, "global_step": 43958, "epoch": 493, "lr": 9.378541623451491e-05}
{"train_loss": 0.2800350487232208, "global_step": 43959, "epoch": 493, "lr": 9.378513631127131e-05}
{"train_loss": 0.31197598576545715, "global_step": 43960, "epoch": 493, "lr": 9.378485638214132e-05}
{"train_loss": 0.33281993865966797, "global_step": 43961, "epoch": 493, "lr": 9.3784576447125e-05}
{"train_loss": 0.23815353214740753, "global_step": 43962, "epoch": 493, "lr": 9.378429650622237e-05}
{"train_loss": 0.3103293776512146, "global_step": 43963, "epoch": 493, "lr": 9.378401655943348e-05}
{"train_loss": 0.31503531336784363, "global_step": 43964, "epoch": 493, "lr": 9.378373660675837e-05}
{"train_loss": 0.2879644791396816, "global_step": 43965, "epoch": 493, "lr": 9.378345664819706e-05, "val_loss": 2.8626456260681152}
{"train_loss": 0.36355704069137573, "global_step": 43966, "epoch": 494, "lr": 9.378317668374959e-05}
{"train_loss": 0.31087160110473633, "global_step": 43967, "epoch": 494, "lr": 9.378289671341602e-05}
{"train_loss": 0.32460835576057434, "global_step": 43968, "epoch": 494, "lr": 9.378261673719636e-05}
{"train_loss": 0.3276618719100952, "global_step": 43969, "epoch": 494, "lr": 9.378233675509067e-05}
{"train_loss": 0.2801622152328491, "global_step": 43970, "epoch": 494, "lr": 9.378205676709899e-05}
{"train_loss": 0.28440260887145996, "global_step": 43971, "epoch": 494, "lr": 9.378177677322134e-05}
{"train_loss": 0.2539060711860657, "global_step": 43972, "epoch": 494, "lr": 9.378149677345774e-05}
{"train_loss": 0.31860798597335815, "global_step": 43973, "epoch": 494, "lr": 9.378121676780828e-05}
{"train_loss": 0.31084877252578735, "global_step": 43974, "epoch": 494, "lr": 9.378093675627295e-05}
{"train_loss": 0.3326972424983978, "global_step": 43975, "epoch": 494, "lr": 9.378065673885182e-05}
{"train_loss": 0.36592456698417664, "global_step": 43976, "epoch": 494, "lr": 9.378037671554491e-05}
{"train_loss": 0.3112643361091614, "global_step": 43977, "epoch": 494, "lr": 9.378009668635226e-05}
{"train_loss": 0.28058725595474243, "global_step": 43978, "epoch": 494, "lr": 9.37798166512739e-05}
{"train_loss": 0.3191581070423126, "global_step": 43979, "epoch": 494, "lr": 9.37795366103099e-05}
{"train_loss": 0.22660024464130402, "global_step": 43980, "epoch": 494, "lr": 9.377925656346027e-05}
{"train_loss": 0.28574562072753906, "global_step": 43981, "epoch": 494, "lr": 9.377897651072503e-05}
{"train_loss": 0.3916078209877014, "global_step": 43982, "epoch": 494, "lr": 9.377869645210426e-05}
{"train_loss": 0.3860393166542053, "global_step": 43983, "epoch": 494, "lr": 9.377841638759798e-05}
{"train_loss": 0.31570446491241455, "global_step": 43984, "epoch": 494, "lr": 9.377813631720622e-05}
{"train_loss": 0.29579395055770874, "global_step": 43985, "epoch": 494, "lr": 9.377785624092901e-05}
{"train_loss": 0.24852579832077026, "global_step": 43986, "epoch": 494, "lr": 9.377757615876641e-05}
{"train_loss": 0.31660348176956177, "global_step": 43987, "epoch": 494, "lr": 9.377729607071846e-05}
{"train_loss": 0.3100365102291107, "global_step": 43988, "epoch": 494, "lr": 9.377701597678518e-05}
{"train_loss": 0.26740339398384094, "global_step": 43989, "epoch": 494, "lr": 9.37767358769666e-05}
{"train_loss": 0.34425732493400574, "global_step": 43990, "epoch": 494, "lr": 9.377645577126279e-05}
{"train_loss": 0.302427738904953, "global_step": 43991, "epoch": 494, "lr": 9.377617565967376e-05}
{"train_loss": 0.25444498658180237, "global_step": 43992, "epoch": 494, "lr": 9.377589554219956e-05}
{"train_loss": 0.41298994421958923, "global_step": 43993, "epoch": 494, "lr": 9.377561541884023e-05}
{"train_loss": 0.3278433084487915, "global_step": 43994, "epoch": 494, "lr": 9.377533528959579e-05}
{"train_loss": 0.30010315775871277, "global_step": 43995, "epoch": 494, "lr": 9.37750551544663e-05}
{"train_loss": 0.26076826453208923, "global_step": 43996, "epoch": 494, "lr": 9.377477501345177e-05}
{"train_loss": 0.28629884123802185, "global_step": 43997, "epoch": 494, "lr": 9.377449486655228e-05}
{"train_loss": 0.27213117480278015, "global_step": 43998, "epoch": 494, "lr": 9.377421471376782e-05}
{"train_loss": 0.37572693824768066, "global_step": 43999, "epoch": 494, "lr": 9.377393455509847e-05}
{"train_loss": 0.20622093975543976, "global_step": 44000, "epoch": 494, "lr": 9.377365439054423e-05}
{"train_loss": 0.27496787905693054, "global_step": 44001, "epoch": 494, "lr": 9.377337422010517e-05}
{"train_loss": 0.27552926540374756, "global_step": 44002, "epoch": 494, "lr": 9.377309404378129e-05}
{"train_loss": 0.305249959230423, "global_step": 44003, "epoch": 494, "lr": 9.377281386157268e-05}
{"train_loss": 0.2514835000038147, "global_step": 44004, "epoch": 494, "lr": 9.377253367347933e-05}
{"train_loss": 0.2982426583766937, "global_step": 44005, "epoch": 494, "lr": 9.377225347950131e-05}
{"train_loss": 0.3417653739452362, "global_step": 44006, "epoch": 494, "lr": 9.377197327963865e-05}
{"train_loss": 0.3211485743522644, "global_step": 44007, "epoch": 494, "lr": 9.377169307389135e-05}
{"train_loss": 0.35717862844467163, "global_step": 44008, "epoch": 494, "lr": 9.37714128622595e-05}
{"train_loss": 0.4246523976325989, "global_step": 44009, "epoch": 494, "lr": 9.377113264474312e-05}
{"train_loss": 0.3254254460334778, "global_step": 44010, "epoch": 494, "lr": 9.377085242134224e-05}
{"train_loss": 0.3341313600540161, "global_step": 44011, "epoch": 494, "lr": 9.377057219205691e-05}
{"train_loss": 0.290466845035553, "global_step": 44012, "epoch": 494, "lr": 9.377029195688715e-05}
{"train_loss": 0.21432740986347198, "global_step": 44013, "epoch": 494, "lr": 9.377001171583301e-05}
{"train_loss": 0.30918648838996887, "global_step": 44014, "epoch": 494, "lr": 9.376973146889452e-05}
{"train_loss": 0.3524617850780487, "global_step": 44015, "epoch": 494, "lr": 9.376945121607172e-05}
{"train_loss": 0.3123094439506531, "global_step": 44016, "epoch": 494, "lr": 9.376917095736467e-05}
{"train_loss": 0.3501342236995697, "global_step": 44017, "epoch": 494, "lr": 9.376889069277338e-05}
{"train_loss": 0.23365499079227448, "global_step": 44018, "epoch": 494, "lr": 9.37686104222979e-05}
{"train_loss": 0.33648818731307983, "global_step": 44019, "epoch": 494, "lr": 9.376833014593826e-05}
{"train_loss": 0.34748631715774536, "global_step": 44020, "epoch": 494, "lr": 9.37680498636945e-05}
{"train_loss": 0.24194148182868958, "global_step": 44021, "epoch": 494, "lr": 9.376776957556666e-05}
{"train_loss": 0.3695034086704254, "global_step": 44022, "epoch": 494, "lr": 9.376748928155476e-05}
{"train_loss": 0.23211000859737396, "global_step": 44023, "epoch": 494, "lr": 9.376720898165887e-05}
{"train_loss": 0.2958669662475586, "global_step": 44024, "epoch": 494, "lr": 9.376692867587902e-05}
{"train_loss": 0.33290523290634155, "global_step": 44025, "epoch": 494, "lr": 9.376664836421523e-05}
{"train_loss": 0.32696202397346497, "global_step": 44026, "epoch": 494, "lr": 9.376636804666755e-05}
{"train_loss": 0.2899535000324249, "global_step": 44027, "epoch": 494, "lr": 9.3766087723236e-05}
{"train_loss": 0.30776748061180115, "global_step": 44028, "epoch": 494, "lr": 9.376580739392066e-05}
{"train_loss": 0.3733537793159485, "global_step": 44029, "epoch": 494, "lr": 9.376552705872153e-05}
{"train_loss": 0.2924500107765198, "global_step": 44030, "epoch": 494, "lr": 9.376524671763865e-05}
{"train_loss": 0.19907799363136292, "global_step": 44031, "epoch": 494, "lr": 9.376496637067207e-05}
{"train_loss": 0.29247212409973145, "global_step": 44032, "epoch": 494, "lr": 9.376468601782183e-05}
{"train_loss": 0.39313510060310364, "global_step": 44033, "epoch": 494, "lr": 9.376440565908796e-05}
{"train_loss": 0.3670930862426758, "global_step": 44034, "epoch": 494, "lr": 9.37641252944705e-05}
{"train_loss": 0.22702595591545105, "global_step": 44035, "epoch": 494, "lr": 9.376384492396948e-05}
{"train_loss": 0.40247198939323425, "global_step": 44036, "epoch": 494, "lr": 9.376356454758495e-05}
{"train_loss": 0.22298969328403473, "global_step": 44037, "epoch": 494, "lr": 9.376328416531695e-05}
{"train_loss": 0.3232044279575348, "global_step": 44038, "epoch": 494, "lr": 9.376300377716549e-05}
{"train_loss": 0.41916874051094055, "global_step": 44039, "epoch": 494, "lr": 9.376272338313063e-05}
{"train_loss": 0.30083346366882324, "global_step": 44040, "epoch": 494, "lr": 9.376244298321242e-05}
{"train_loss": 0.26958486437797546, "global_step": 44041, "epoch": 494, "lr": 9.376216257741089e-05}
{"train_loss": 0.28091421723365784, "global_step": 44042, "epoch": 494, "lr": 9.376188216572605e-05}
{"train_loss": 0.2981666922569275, "global_step": 44043, "epoch": 494, "lr": 9.376160174815796e-05}
{"train_loss": 0.27072352170944214, "global_step": 44044, "epoch": 494, "lr": 9.376132132470666e-05}
{"train_loss": 0.3428196310997009, "global_step": 44045, "epoch": 494, "lr": 9.376104089537218e-05}
{"train_loss": 0.3580140471458435, "global_step": 44046, "epoch": 494, "lr": 9.376076046015457e-05}
{"train_loss": 0.28665220737457275, "global_step": 44047, "epoch": 494, "lr": 9.376048001905385e-05}
{"train_loss": 0.38876187801361084, "global_step": 44048, "epoch": 494, "lr": 9.376019957207008e-05}
{"train_loss": 0.3352157175540924, "global_step": 44049, "epoch": 494, "lr": 9.375991911920328e-05}
{"train_loss": 0.35713839530944824, "global_step": 44050, "epoch": 494, "lr": 9.375963866045347e-05}
{"train_loss": 0.3120517134666443, "global_step": 44051, "epoch": 494, "lr": 9.375935819582074e-05}
{"train_loss": 0.2686082720756531, "global_step": 44052, "epoch": 494, "lr": 9.375907772530508e-05}
{"train_loss": 0.298579603433609, "global_step": 44053, "epoch": 494, "lr": 9.375879724890655e-05}
{"train_loss": 0.3104153426175707, "global_step": 44054, "epoch": 494, "lr": 9.37585167666252e-05, "val_loss": 2.8084962368011475}
{"train_loss": 0.37177756428718567, "global_step": 44055, "epoch": 495, "lr": 9.375823627846104e-05}
{"train_loss": 0.23768293857574463, "global_step": 44056, "epoch": 495, "lr": 9.375795578441411e-05}
{"train_loss": 0.2482243776321411, "global_step": 44057, "epoch": 495, "lr": 9.375767528448445e-05}
{"train_loss": 0.3003997802734375, "global_step": 44058, "epoch": 495, "lr": 9.375739477867212e-05}
{"train_loss": 0.34572163224220276, "global_step": 44059, "epoch": 495, "lr": 9.375711426697714e-05}
{"train_loss": 0.2084592878818512, "global_step": 44060, "epoch": 495, "lr": 9.375683374939955e-05}
{"train_loss": 0.4035830497741699, "global_step": 44061, "epoch": 495, "lr": 9.375655322593937e-05}
{"train_loss": 0.34952864050865173, "global_step": 44062, "epoch": 495, "lr": 9.375627269659666e-05}
{"train_loss": 0.3449103534221649, "global_step": 44063, "epoch": 495, "lr": 9.375599216137147e-05}
{"train_loss": 0.34805867075920105, "global_step": 44064, "epoch": 495, "lr": 9.37557116202638e-05}
{"train_loss": 0.4048539102077484, "global_step": 44065, "epoch": 495, "lr": 9.375543107327372e-05}
{"train_loss": 0.39633485674858093, "global_step": 44066, "epoch": 495, "lr": 9.375515052040125e-05}
{"train_loss": 0.31394851207733154, "global_step": 44067, "epoch": 495, "lr": 9.375486996164645e-05}
{"train_loss": 0.29051321744918823, "global_step": 44068, "epoch": 495, "lr": 9.375458939700932e-05}
{"train_loss": 0.40136101841926575, "global_step": 44069, "epoch": 495, "lr": 9.375430882648992e-05}
{"train_loss": 0.2276388555765152, "global_step": 44070, "epoch": 495, "lr": 9.37540282500883e-05}
{"train_loss": 0.22834187746047974, "global_step": 44071, "epoch": 495, "lr": 9.375374766780447e-05}
{"train_loss": 0.2842905521392822, "global_step": 44072, "epoch": 495, "lr": 9.375346707963848e-05}
{"train_loss": 0.30140265822410583, "global_step": 44073, "epoch": 495, "lr": 9.375318648559038e-05}
{"train_loss": 0.28745076060295105, "global_step": 44074, "epoch": 495, "lr": 9.37529058856602e-05}
{"train_loss": 0.323132187128067, "global_step": 44075, "epoch": 495, "lr": 9.375262527984797e-05}
{"train_loss": 0.4219798445701599, "global_step": 44076, "epoch": 495, "lr": 9.375234466815374e-05}
{"train_loss": 0.2925507128238678, "global_step": 44077, "epoch": 495, "lr": 9.375206405057752e-05}
{"train_loss": 0.4038141965866089, "global_step": 44078, "epoch": 495, "lr": 9.375178342711939e-05}
{"train_loss": 0.27925702929496765, "global_step": 44079, "epoch": 495, "lr": 9.375150279777935e-05}
{"train_loss": 0.3075607717037201, "global_step": 44080, "epoch": 495, "lr": 9.375122216255748e-05}
{"train_loss": 0.1818947196006775, "global_step": 44081, "epoch": 495, "lr": 9.375094152145377e-05}
{"train_loss": 0.25078096985816956, "global_step": 44082, "epoch": 495, "lr": 9.375066087446827e-05}
{"train_loss": 0.2646960914134979, "global_step": 44083, "epoch": 495, "lr": 9.375038022160106e-05}
{"train_loss": 0.2863154113292694, "global_step": 44084, "epoch": 495, "lr": 9.37500995628521e-05}
{"train_loss": 0.430909126996994, "global_step": 44085, "epoch": 495, "lr": 9.37498188982215e-05}
{"train_loss": 0.31912654638290405, "global_step": 44086, "epoch": 495, "lr": 9.374953822770927e-05}
{"train_loss": 0.33242154121398926, "global_step": 44087, "epoch": 495, "lr": 9.374925755131545e-05}
{"train_loss": 0.36534249782562256, "global_step": 44088, "epoch": 495, "lr": 9.374897686904006e-05}
{"train_loss": 0.2956662178039551, "global_step": 44089, "epoch": 495, "lr": 9.374869618088316e-05}
{"train_loss": 0.3193708658218384, "global_step": 44090, "epoch": 495, "lr": 9.374841548684478e-05}
{"train_loss": 0.22711002826690674, "global_step": 44091, "epoch": 495, "lr": 9.374813478692497e-05}
{"train_loss": 0.36800646781921387, "global_step": 44092, "epoch": 495, "lr": 9.374785408112375e-05}
{"train_loss": 0.26253578066825867, "global_step": 44093, "epoch": 495, "lr": 9.374757336944117e-05}
{"train_loss": 0.3376334011554718, "global_step": 44094, "epoch": 495, "lr": 9.374729265187726e-05}
{"train_loss": 0.26676225662231445, "global_step": 44095, "epoch": 495, "lr": 9.374701192843205e-05}
{"train_loss": 0.31793951988220215, "global_step": 44096, "epoch": 495, "lr": 9.374673119910558e-05}
{"train_loss": 0.3037402927875519, "global_step": 44097, "epoch": 495, "lr": 9.374645046389792e-05}
{"train_loss": 0.19457657635211945, "global_step": 44098, "epoch": 495, "lr": 9.374616972280907e-05}
{"train_loss": 0.2837023138999939, "global_step": 44099, "epoch": 495, "lr": 9.374588897583907e-05}
{"train_loss": 0.3467957377433777, "global_step": 44100, "epoch": 495, "lr": 9.3745608222988e-05}
{"train_loss": 0.3522343337535858, "global_step": 44101, "epoch": 495, "lr": 9.374532746425584e-05}
{"train_loss": 0.3323127329349518, "global_step": 44102, "epoch": 495, "lr": 9.374504669964265e-05}
{"train_loss": 0.36191806197166443, "global_step": 44103, "epoch": 495, "lr": 9.374476592914848e-05}
{"train_loss": 0.3768182098865509, "global_step": 44104, "epoch": 495, "lr": 9.374448515277336e-05}
{"train_loss": 0.28101107478141785, "global_step": 44105, "epoch": 495, "lr": 9.374420437051732e-05}
{"train_loss": 0.2726612389087677, "global_step": 44106, "epoch": 495, "lr": 9.374392358238043e-05}
{"train_loss": 0.34146472811698914, "global_step": 44107, "epoch": 495, "lr": 9.374364278836267e-05}
{"train_loss": 0.2813287675380707, "global_step": 44108, "epoch": 495, "lr": 9.374336198846413e-05}
{"train_loss": 0.3152671754360199, "global_step": 44109, "epoch": 495, "lr": 9.374308118268483e-05}
{"train_loss": 0.4078536629676819, "global_step": 44110, "epoch": 495, "lr": 9.37428003710248e-05}
{"train_loss": 0.2777373194694519, "global_step": 44111, "epoch": 495, "lr": 9.374251955348408e-05}
{"train_loss": 0.2616056203842163, "global_step": 44112, "epoch": 495, "lr": 9.374223873006272e-05}
{"train_loss": 0.26072052121162415, "global_step": 44113, "epoch": 495, "lr": 9.374195790076074e-05}
{"train_loss": 0.30188313126564026, "global_step": 44114, "epoch": 495, "lr": 9.374167706557819e-05}
{"train_loss": 0.24755313992500305, "global_step": 44115, "epoch": 495, "lr": 9.374139622451511e-05}
{"train_loss": 0.21908721327781677, "global_step": 44116, "epoch": 495, "lr": 9.374111537757153e-05}
{"train_loss": 0.3728625774383545, "global_step": 44117, "epoch": 495, "lr": 9.374083452474749e-05}
{"train_loss": 0.2831386923789978, "global_step": 44118, "epoch": 495, "lr": 9.374055366604302e-05}
{"train_loss": 0.24424384534358978, "global_step": 44119, "epoch": 495, "lr": 9.374027280145817e-05}
{"train_loss": 0.2785983085632324, "global_step": 44120, "epoch": 495, "lr": 9.373999193099297e-05}
{"train_loss": 0.39367517828941345, "global_step": 44121, "epoch": 495, "lr": 9.373971105464747e-05}
{"train_loss": 0.21726499497890472, "global_step": 44122, "epoch": 495, "lr": 9.373943017242171e-05}
{"train_loss": 0.2625841498374939, "global_step": 44123, "epoch": 495, "lr": 9.373914928431568e-05}
{"train_loss": 0.3350472152233124, "global_step": 44124, "epoch": 495, "lr": 9.373886839032948e-05}
{"train_loss": 0.30582574009895325, "global_step": 44125, "epoch": 495, "lr": 9.373858749046313e-05}
{"train_loss": 0.23590008914470673, "global_step": 44126, "epoch": 495, "lr": 9.373830658471666e-05}
{"train_loss": 0.2815043032169342, "global_step": 44127, "epoch": 495, "lr": 9.37380256730901e-05}
{"train_loss": 0.34027916193008423, "global_step": 44128, "epoch": 495, "lr": 9.373774475558348e-05}
{"train_loss": 0.2884747087955475, "global_step": 44129, "epoch": 495, "lr": 9.373746383219685e-05}
{"train_loss": 0.30819466710090637, "global_step": 44130, "epoch": 495, "lr": 9.373718290293028e-05}
{"train_loss": 0.3204424977302551, "global_step": 44131, "epoch": 495, "lr": 9.373690196778376e-05}
{"train_loss": 0.16482874751091003, "global_step": 44132, "epoch": 495, "lr": 9.373662102675735e-05}
{"train_loss": 0.28376105427742004, "global_step": 44133, "epoch": 495, "lr": 9.373634007985109e-05}
{"train_loss": 0.34705570340156555, "global_step": 44134, "epoch": 495, "lr": 9.3736059127065e-05}
{"train_loss": 0.2795502841472626, "global_step": 44135, "epoch": 495, "lr": 9.373577816839914e-05}
{"train_loss": 0.22532965242862701, "global_step": 44136, "epoch": 495, "lr": 9.373549720385355e-05}
{"train_loss": 0.238280788064003, "global_step": 44137, "epoch": 495, "lr": 9.373521623342824e-05}
{"train_loss": 0.21895672380924225, "global_step": 44138, "epoch": 495, "lr": 9.373493525712326e-05}
{"train_loss": 0.2909894585609436, "global_step": 44139, "epoch": 495, "lr": 9.373465427493866e-05}
{"train_loss": 0.2115006148815155, "global_step": 44140, "epoch": 495, "lr": 9.373437328687447e-05}
{"train_loss": 0.29005885124206543, "global_step": 44141, "epoch": 495, "lr": 9.373409229293071e-05}
{"train_loss": 0.2695508301258087, "global_step": 44142, "epoch": 495, "lr": 9.373381129310745e-05}
{"train_loss": 0.3007244846124328, "global_step": 44143, "epoch": 495, "lr": 9.37335302874047e-05, "val_loss": 2.8741061687469482, "train_action_mse_error": 25.75765609741211}
{"train_loss": 0.29047641158103943, "global_step": 44144, "epoch": 496, "lr": 9.373324927582252e-05}
{"train_loss": 0.2859705686569214, "global_step": 44145, "epoch": 496, "lr": 9.373296825836093e-05}
{"train_loss": 0.2235860377550125, "global_step": 44146, "epoch": 496, "lr": 9.373268723502e-05}
{"train_loss": 0.3254496455192566, "global_step": 44147, "epoch": 496, "lr": 9.37324062057997e-05}
{"train_loss": 0.2341063767671585, "global_step": 44148, "epoch": 496, "lr": 9.373212517070014e-05}
{"train_loss": 0.33531710505485535, "global_step": 44149, "epoch": 496, "lr": 9.373184412972134e-05}
{"train_loss": 0.37271347641944885, "global_step": 44150, "epoch": 496, "lr": 9.373156308286331e-05}
{"train_loss": 0.28302595019340515, "global_step": 44151, "epoch": 496, "lr": 9.373128203012609e-05}
{"train_loss": 0.30502477288246155, "global_step": 44152, "epoch": 496, "lr": 9.373100097150975e-05}
{"train_loss": 0.2948205769062042, "global_step": 44153, "epoch": 496, "lr": 9.37307199070143e-05}
{"train_loss": 0.18935629725456238, "global_step": 44154, "epoch": 496, "lr": 9.37304388366398e-05}
{"train_loss": 0.3206328749656677, "global_step": 44155, "epoch": 496, "lr": 9.373015776038627e-05}
{"train_loss": 0.25583600997924805, "global_step": 44156, "epoch": 496, "lr": 9.372987667825375e-05}
{"train_loss": 0.2665129601955414, "global_step": 44157, "epoch": 496, "lr": 9.372959559024228e-05}
{"train_loss": 0.3415187895298004, "global_step": 44158, "epoch": 496, "lr": 9.372931449635192e-05}
{"train_loss": 0.423912912607193, "global_step": 44159, "epoch": 496, "lr": 9.372903339658266e-05}
{"train_loss": 0.3154090344905853, "global_step": 44160, "epoch": 496, "lr": 9.372875229093458e-05}
{"train_loss": 0.3113939166069031, "global_step": 44161, "epoch": 496, "lr": 9.37284711794077e-05}
{"train_loss": 0.3190918266773224, "global_step": 44162, "epoch": 496, "lr": 9.372819006200205e-05}
{"train_loss": 0.3830329179763794, "global_step": 44163, "epoch": 496, "lr": 9.372790893871767e-05}
{"train_loss": 0.3589368164539337, "global_step": 44164, "epoch": 496, "lr": 9.372762780955465e-05}
{"train_loss": 0.3386567533016205, "global_step": 44165, "epoch": 496, "lr": 9.372734667451294e-05}
{"train_loss": 0.1584133505821228, "global_step": 44166, "epoch": 496, "lr": 9.372706553359262e-05}
{"train_loss": 0.32897400856018066, "global_step": 44167, "epoch": 496, "lr": 9.372678438679375e-05}
{"train_loss": 0.2983463406562805, "global_step": 44168, "epoch": 496, "lr": 9.372650323411634e-05}
{"train_loss": 0.23935358226299286, "global_step": 44169, "epoch": 496, "lr": 9.372622207556043e-05}
{"train_loss": 0.2927189767360687, "global_step": 44170, "epoch": 496, "lr": 9.372594091112607e-05}
{"train_loss": 0.2560163140296936, "global_step": 44171, "epoch": 496, "lr": 9.372565974081329e-05}
{"train_loss": 0.25338098406791687, "global_step": 44172, "epoch": 496, "lr": 9.372537856462213e-05}
{"train_loss": 0.31486624479293823, "global_step": 44173, "epoch": 496, "lr": 9.372509738255261e-05}
{"train_loss": 0.23354753851890564, "global_step": 44174, "epoch": 496, "lr": 9.372481619460481e-05}
{"train_loss": 0.2584710717201233, "global_step": 44175, "epoch": 496, "lr": 9.372453500077872e-05}
{"train_loss": 0.2948133945465088, "global_step": 44176, "epoch": 496, "lr": 9.372425380107441e-05}
{"train_loss": 0.26223281025886536, "global_step": 44177, "epoch": 496, "lr": 9.37239725954919e-05}
{"train_loss": 0.28180205821990967, "global_step": 44178, "epoch": 496, "lr": 9.372369138403123e-05}
{"train_loss": 0.29907068610191345, "global_step": 44179, "epoch": 496, "lr": 9.372341016669246e-05}
{"train_loss": 0.30473875999450684, "global_step": 44180, "epoch": 496, "lr": 9.372312894347559e-05}
{"train_loss": 0.39386987686157227, "global_step": 44181, "epoch": 496, "lr": 9.372284771438069e-05}
{"train_loss": 0.30685776472091675, "global_step": 44182, "epoch": 496, "lr": 9.37225664794078e-05}
{"train_loss": 0.4068421721458435, "global_step": 44183, "epoch": 496, "lr": 9.372228523855692e-05}
{"train_loss": 0.32950693368911743, "global_step": 44184, "epoch": 496, "lr": 9.372200399182812e-05}
{"train_loss": 0.3332647681236267, "global_step": 44185, "epoch": 496, "lr": 9.372172273922142e-05}
{"train_loss": 0.30492809414863586, "global_step": 44186, "epoch": 496, "lr": 9.372144148073689e-05}
{"train_loss": 0.3766500949859619, "global_step": 44187, "epoch": 496, "lr": 9.372116021637453e-05}
{"train_loss": 0.2723611295223236, "global_step": 44188, "epoch": 496, "lr": 9.37208789461344e-05}
{"train_loss": 0.26782920956611633, "global_step": 44189, "epoch": 496, "lr": 9.372059767001652e-05}
{"train_loss": 0.3552393615245819, "global_step": 44190, "epoch": 496, "lr": 9.372031638802094e-05}
{"train_loss": 0.3194332718849182, "global_step": 44191, "epoch": 496, "lr": 9.372003510014771e-05}
{"train_loss": 0.31350961327552795, "global_step": 44192, "epoch": 496, "lr": 9.371975380639683e-05}
{"train_loss": 0.3167361319065094, "global_step": 44193, "epoch": 496, "lr": 9.371947250676837e-05}
{"train_loss": 0.21732047200202942, "global_step": 44194, "epoch": 496, "lr": 9.371919120126238e-05}
{"train_loss": 0.2894206941127777, "global_step": 44195, "epoch": 496, "lr": 9.371890988987885e-05}
{"train_loss": 0.3872072696685791, "global_step": 44196, "epoch": 496, "lr": 9.371862857261786e-05}
{"train_loss": 0.2657151520252228, "global_step": 44197, "epoch": 496, "lr": 9.371834724947944e-05}
{"train_loss": 0.2634737491607666, "global_step": 44198, "epoch": 496, "lr": 9.371806592046361e-05}
{"train_loss": 0.302470862865448, "global_step": 44199, "epoch": 496, "lr": 9.371778458557043e-05}
{"train_loss": 0.3058692216873169, "global_step": 44200, "epoch": 496, "lr": 9.371750324479991e-05}
{"train_loss": 0.32435449957847595, "global_step": 44201, "epoch": 496, "lr": 9.371722189815211e-05}
{"train_loss": 0.23111332952976227, "global_step": 44202, "epoch": 496, "lr": 9.371694054562708e-05}
{"train_loss": 0.25941842794418335, "global_step": 44203, "epoch": 496, "lr": 9.371665918722482e-05}
{"train_loss": 0.27509352564811707, "global_step": 44204, "epoch": 496, "lr": 9.37163778229454e-05}
{"train_loss": 0.2670905590057373, "global_step": 44205, "epoch": 496, "lr": 9.371609645278883e-05}
{"train_loss": 0.32901135087013245, "global_step": 44206, "epoch": 496, "lr": 9.371581507675517e-05}
{"train_loss": 0.29694390296936035, "global_step": 44207, "epoch": 496, "lr": 9.371553369484447e-05}
{"train_loss": 0.31026601791381836, "global_step": 44208, "epoch": 496, "lr": 9.371525230705674e-05}
{"train_loss": 0.305467426776886, "global_step": 44209, "epoch": 496, "lr": 9.3714970913392e-05}
{"train_loss": 0.30764707922935486, "global_step": 44210, "epoch": 496, "lr": 9.371468951385036e-05}
{"train_loss": 0.2642920911312103, "global_step": 44211, "epoch": 496, "lr": 9.371440810843178e-05}
{"train_loss": 0.2810713052749634, "global_step": 44212, "epoch": 496, "lr": 9.371412669713633e-05}
{"train_loss": 0.3690960705280304, "global_step": 44213, "epoch": 496, "lr": 9.371384527996407e-05}
{"train_loss": 0.2653995156288147, "global_step": 44214, "epoch": 496, "lr": 9.3713563856915e-05}
{"train_loss": 0.3782741129398346, "global_step": 44215, "epoch": 496, "lr": 9.371328242798917e-05}
{"train_loss": 0.4099634289741516, "global_step": 44216, "epoch": 496, "lr": 9.371300099318664e-05}
{"train_loss": 0.3574378788471222, "global_step": 44217, "epoch": 496, "lr": 9.371271955250742e-05}
{"train_loss": 0.33817392587661743, "global_step": 44218, "epoch": 496, "lr": 9.371243810595155e-05}
{"train_loss": 0.23098188638687134, "global_step": 44219, "epoch": 496, "lr": 9.37121566535191e-05}
{"train_loss": 0.33358317613601685, "global_step": 44220, "epoch": 496, "lr": 9.371187519521006e-05}
{"train_loss": 0.2533087432384491, "global_step": 44221, "epoch": 496, "lr": 9.37115937310245e-05}
{"train_loss": 0.3351903259754181, "global_step": 44222, "epoch": 496, "lr": 9.371131226096243e-05}
{"train_loss": 0.2943670153617859, "global_step": 44223, "epoch": 496, "lr": 9.371103078502394e-05}
{"train_loss": 0.32533571124076843, "global_step": 44224, "epoch": 496, "lr": 9.371074930320902e-05}
{"train_loss": 0.2786831855773926, "global_step": 44225, "epoch": 496, "lr": 9.371046781551771e-05}
{"train_loss": 0.3080718219280243, "global_step": 44226, "epoch": 496, "lr": 9.371018632195008e-05}
{"train_loss": 0.24857500195503235, "global_step": 44227, "epoch": 496, "lr": 9.370990482250614e-05}
{"train_loss": 0.3634849488735199, "global_step": 44228, "epoch": 496, "lr": 9.370962331718593e-05}
{"train_loss": 0.2070983201265335, "global_step": 44229, "epoch": 496, "lr": 9.37093418059895e-05}
{"train_loss": 0.36328402161598206, "global_step": 44230, "epoch": 496, "lr": 9.370906028891688e-05}
{"train_loss": 0.39889049530029297, "global_step": 44231, "epoch": 496, "lr": 9.370877876596811e-05}
{"train_loss": 0.3031707126437948, "global_step": 44232, "epoch": 496, "lr": 9.370849723714323e-05, "val_loss": 3.0277163982391357}
{"train_loss": 0.34139296412467957, "global_step": 44233, "epoch": 497, "lr": 9.370821570244228e-05}
{"train_loss": 0.3046905994415283, "global_step": 44234, "epoch": 497, "lr": 9.370793416186528e-05}
{"train_loss": 0.3082944452762604, "global_step": 44235, "epoch": 497, "lr": 9.37076526154123e-05}
{"train_loss": 0.25526583194732666, "global_step": 44236, "epoch": 497, "lr": 9.370737106308334e-05}
{"train_loss": 0.2774466872215271, "global_step": 44237, "epoch": 497, "lr": 9.370708950487846e-05}
{"train_loss": 0.31261691451072693, "global_step": 44238, "epoch": 497, "lr": 9.370680794079771e-05}
{"train_loss": 0.2614094913005829, "global_step": 44239, "epoch": 497, "lr": 9.370652637084109e-05}
{"train_loss": 0.2998608350753784, "global_step": 44240, "epoch": 497, "lr": 9.370624479500867e-05}
{"train_loss": 0.23895834386348724, "global_step": 44241, "epoch": 497, "lr": 9.370596321330047e-05}
{"train_loss": 0.43139728903770447, "global_step": 44242, "epoch": 497, "lr": 9.370568162571656e-05}
{"train_loss": 0.34767773747444153, "global_step": 44243, "epoch": 497, "lr": 9.370540003225694e-05}
{"train_loss": 0.3522464334964752, "global_step": 44244, "epoch": 497, "lr": 9.370511843292166e-05}
{"train_loss": 0.3085394501686096, "global_step": 44245, "epoch": 497, "lr": 9.370483682771076e-05}
{"train_loss": 0.2096746563911438, "global_step": 44246, "epoch": 497, "lr": 9.370455521662427e-05}
{"train_loss": 0.2815644145011902, "global_step": 44247, "epoch": 497, "lr": 9.370427359966225e-05}
{"train_loss": 0.2230735421180725, "global_step": 44248, "epoch": 497, "lr": 9.370399197682472e-05}
{"train_loss": 0.2640051543712616, "global_step": 44249, "epoch": 497, "lr": 9.37037103481117e-05}
{"train_loss": 0.32053881883621216, "global_step": 44250, "epoch": 497, "lr": 9.370342871352327e-05}
{"train_loss": 0.2950962781906128, "global_step": 44251, "epoch": 497, "lr": 9.370314707305944e-05}
{"train_loss": 0.33686789870262146, "global_step": 44252, "epoch": 497, "lr": 9.370286542672027e-05}
{"train_loss": 0.281191349029541, "global_step": 44253, "epoch": 497, "lr": 9.370258377450576e-05}
{"train_loss": 0.274032324552536, "global_step": 44254, "epoch": 497, "lr": 9.370230211641597e-05}
{"train_loss": 0.3836797773838043, "global_step": 44255, "epoch": 497, "lr": 9.370202045245095e-05}
{"train_loss": 0.2867278754711151, "global_step": 44256, "epoch": 497, "lr": 9.370173878261072e-05}
{"train_loss": 0.30977046489715576, "global_step": 44257, "epoch": 497, "lr": 9.370145710689533e-05}
{"train_loss": 0.29543787240982056, "global_step": 44258, "epoch": 497, "lr": 9.37011754253048e-05}
{"train_loss": 0.3009503483772278, "global_step": 44259, "epoch": 497, "lr": 9.37008937378392e-05}
{"train_loss": 0.3033064901828766, "global_step": 44260, "epoch": 497, "lr": 9.370061204449854e-05}
{"train_loss": 0.3330960273742676, "global_step": 44261, "epoch": 497, "lr": 9.370033034528285e-05}
{"train_loss": 0.35624250769615173, "global_step": 44262, "epoch": 497, "lr": 9.370004864019219e-05}
{"train_loss": 0.23212505877017975, "global_step": 44263, "epoch": 497, "lr": 9.369976692922661e-05}
{"train_loss": 0.29715004563331604, "global_step": 44264, "epoch": 497, "lr": 9.369948521238612e-05}
{"train_loss": 0.2363176941871643, "global_step": 44265, "epoch": 497, "lr": 9.369920348967075e-05}
{"train_loss": 0.35428497195243835, "global_step": 44266, "epoch": 497, "lr": 9.369892176108057e-05}
{"train_loss": 0.27521002292633057, "global_step": 44267, "epoch": 497, "lr": 9.369864002661559e-05}
{"train_loss": 0.3757919371128082, "global_step": 44268, "epoch": 497, "lr": 9.369835828627586e-05}
{"train_loss": 0.3014109432697296, "global_step": 44269, "epoch": 497, "lr": 9.369807654006144e-05}
{"train_loss": 0.2617073953151703, "global_step": 44270, "epoch": 497, "lr": 9.369779478797234e-05}
{"train_loss": 0.30279305577278137, "global_step": 44271, "epoch": 497, "lr": 9.369751303000859e-05}
{"train_loss": 0.23208306729793549, "global_step": 44272, "epoch": 497, "lr": 9.369723126617024e-05}
{"train_loss": 0.21877090632915497, "global_step": 44273, "epoch": 497, "lr": 9.369694949645735e-05}
{"train_loss": 0.3144876956939697, "global_step": 44274, "epoch": 497, "lr": 9.369666772086993e-05}
{"train_loss": 0.261275976896286, "global_step": 44275, "epoch": 497, "lr": 9.369638593940801e-05}
{"train_loss": 0.2351294904947281, "global_step": 44276, "epoch": 497, "lr": 9.369610415207166e-05}
{"train_loss": 0.3781082332134247, "global_step": 44277, "epoch": 497, "lr": 9.36958223588609e-05}
{"train_loss": 0.303052693605423, "global_step": 44278, "epoch": 497, "lr": 9.369554055977578e-05}
{"train_loss": 0.3491728603839874, "global_step": 44279, "epoch": 497, "lr": 9.369525875481631e-05}
{"train_loss": 0.27017319202423096, "global_step": 44280, "epoch": 497, "lr": 9.369497694398255e-05}
{"train_loss": 0.2913582921028137, "global_step": 44281, "epoch": 497, "lr": 9.369469512727454e-05}
{"train_loss": 0.4003813564777374, "global_step": 44282, "epoch": 497, "lr": 9.369441330469228e-05}
{"train_loss": 0.3141702711582184, "global_step": 44283, "epoch": 497, "lr": 9.369413147623586e-05}
{"train_loss": 0.27906277775764465, "global_step": 44284, "epoch": 497, "lr": 9.36938496419053e-05}
{"train_loss": 0.4230528771877289, "global_step": 44285, "epoch": 497, "lr": 9.369356780170065e-05}
{"train_loss": 0.3535926938056946, "global_step": 44286, "epoch": 497, "lr": 9.36932859556219e-05}
{"train_loss": 0.2569095492362976, "global_step": 44287, "epoch": 497, "lr": 9.369300410366914e-05}
{"train_loss": 0.22686012089252472, "global_step": 44288, "epoch": 497, "lr": 9.369272224584238e-05}
{"train_loss": 0.2977582812309265, "global_step": 44289, "epoch": 497, "lr": 9.369244038214168e-05}
{"train_loss": 0.2373627871274948, "global_step": 44290, "epoch": 497, "lr": 9.369215851256706e-05}
{"train_loss": 0.35088810324668884, "global_step": 44291, "epoch": 497, "lr": 9.369187663711854e-05}
{"train_loss": 0.31359347701072693, "global_step": 44292, "epoch": 497, "lr": 9.36915947557962e-05}
{"train_loss": 0.3671509623527527, "global_step": 44293, "epoch": 497, "lr": 9.369131286860004e-05}
{"train_loss": 0.23784412443637848, "global_step": 44294, "epoch": 497, "lr": 9.369103097553013e-05}
{"train_loss": 0.3851138949394226, "global_step": 44295, "epoch": 497, "lr": 9.36907490765865e-05}
{"train_loss": 0.22795164585113525, "global_step": 44296, "epoch": 497, "lr": 9.369046717176917e-05}
{"train_loss": 0.26991283893585205, "global_step": 44297, "epoch": 497, "lr": 9.36901852610782e-05}
{"train_loss": 0.31409141421318054, "global_step": 44298, "epoch": 497, "lr": 9.36899033445136e-05}
{"train_loss": 0.31028541922569275, "global_step": 44299, "epoch": 497, "lr": 9.368962142207544e-05}
{"train_loss": 0.2582777440547943, "global_step": 44300, "epoch": 497, "lr": 9.368933949376373e-05}
{"train_loss": 0.3631259500980377, "global_step": 44301, "epoch": 497, "lr": 9.368905755957853e-05}
{"train_loss": 0.3452562987804413, "global_step": 44302, "epoch": 497, "lr": 9.368877561951988e-05}
{"train_loss": 0.3694290816783905, "global_step": 44303, "epoch": 497, "lr": 9.368849367358778e-05}
{"train_loss": 0.26424577832221985, "global_step": 44304, "epoch": 497, "lr": 9.36882117217823e-05}
{"train_loss": 0.25812169909477234, "global_step": 44305, "epoch": 497, "lr": 9.368792976410349e-05}
{"train_loss": 0.2925395667552948, "global_step": 44306, "epoch": 497, "lr": 9.368764780055136e-05}
{"train_loss": 0.2686291038990021, "global_step": 44307, "epoch": 497, "lr": 9.368736583112596e-05}
{"train_loss": 0.276110976934433, "global_step": 44308, "epoch": 497, "lr": 9.368708385582732e-05}
{"train_loss": 0.3884337544441223, "global_step": 44309, "epoch": 497, "lr": 9.368680187465549e-05}
{"train_loss": 0.3113086521625519, "global_step": 44310, "epoch": 497, "lr": 9.36865198876105e-05}
{"train_loss": 0.4528861939907074, "global_step": 44311, "epoch": 497, "lr": 9.36862378946924e-05}
{"train_loss": 0.3458538353443146, "global_step": 44312, "epoch": 497, "lr": 9.36859558959012e-05}
{"train_loss": 0.4431077837944031, "global_step": 44313, "epoch": 497, "lr": 9.368567389123697e-05}
{"train_loss": 0.2744932770729065, "global_step": 44314, "epoch": 497, "lr": 9.368539188069972e-05}
{"train_loss": 0.38363298773765564, "global_step": 44315, "epoch": 497, "lr": 9.368510986428951e-05}
{"train_loss": 0.3310447931289673, "global_step": 44316, "epoch": 497, "lr": 9.368482784200636e-05}
{"train_loss": 0.27846333384513855, "global_step": 44317, "epoch": 497, "lr": 9.368454581385033e-05}
{"train_loss": 0.4076346755027771, "global_step": 44318, "epoch": 497, "lr": 9.368426377982145e-05}
{"train_loss": 0.23212438821792603, "global_step": 44319, "epoch": 497, "lr": 9.368398173991975e-05}
{"train_loss": 0.392659991979599, "global_step": 44320, "epoch": 497, "lr": 9.368369969414525e-05}
{"train_loss": 0.30764343698373003, "global_step": 44321, "epoch": 497, "lr": 9.368341764249802e-05, "val_loss": 2.9757919311523438}
{"train_loss": 0.31948933005332947, "global_step": 44322, "epoch": 498, "lr": 9.36831355849781e-05}
{"train_loss": 0.2273484766483307, "global_step": 44323, "epoch": 498, "lr": 9.36828535215855e-05}
{"train_loss": 0.41616350412368774, "global_step": 44324, "epoch": 498, "lr": 9.36825714523203e-05}
{"train_loss": 0.3238314092159271, "global_step": 44325, "epoch": 498, "lr": 9.368228937718247e-05}
{"train_loss": 0.39159491658210754, "global_step": 44326, "epoch": 498, "lr": 9.368200729617212e-05}
{"train_loss": 0.3243716061115265, "global_step": 44327, "epoch": 498, "lr": 9.368172520928924e-05}
{"train_loss": 0.3254021406173706, "global_step": 44328, "epoch": 498, "lr": 9.368144311653389e-05}
{"train_loss": 0.28197529911994934, "global_step": 44329, "epoch": 498, "lr": 9.36811610179061e-05}
{"train_loss": 0.35105612874031067, "global_step": 44330, "epoch": 498, "lr": 9.36808789134059e-05}
{"train_loss": 0.2899312376976013, "global_step": 44331, "epoch": 498, "lr": 9.368059680303335e-05}
{"train_loss": 0.21467949450016022, "global_step": 44332, "epoch": 498, "lr": 9.368031468678848e-05}
{"train_loss": 0.27380087971687317, "global_step": 44333, "epoch": 498, "lr": 9.368003256467132e-05}
{"train_loss": 0.28679412603378296, "global_step": 44334, "epoch": 498, "lr": 9.367975043668191e-05}
{"train_loss": 0.21467265486717224, "global_step": 44335, "epoch": 498, "lr": 9.36794683028203e-05}
{"train_loss": 0.43032142519950867, "global_step": 44336, "epoch": 498, "lr": 9.36791861630865e-05}
{"train_loss": 0.3759484589099884, "global_step": 44337, "epoch": 498, "lr": 9.367890401748058e-05}
{"train_loss": 0.30935996770858765, "global_step": 44338, "epoch": 498, "lr": 9.367862186600256e-05}
{"train_loss": 0.28383034467697144, "global_step": 44339, "epoch": 498, "lr": 9.367833970865249e-05}
{"train_loss": 0.25404325127601624, "global_step": 44340, "epoch": 498, "lr": 9.367805754543038e-05}
{"train_loss": 0.3729366660118103, "global_step": 44341, "epoch": 498, "lr": 9.36777753763363e-05}
{"train_loss": 0.3220561742782593, "global_step": 44342, "epoch": 498, "lr": 9.367749320137028e-05}
{"train_loss": 0.41227370500564575, "global_step": 44343, "epoch": 498, "lr": 9.367721102053235e-05}
{"train_loss": 0.3347407579421997, "global_step": 44344, "epoch": 498, "lr": 9.367692883382255e-05}
{"train_loss": 0.2592953145503998, "global_step": 44345, "epoch": 498, "lr": 9.367664664124092e-05}
{"train_loss": 0.2176215499639511, "global_step": 44346, "epoch": 498, "lr": 9.367636444278749e-05}
{"train_loss": 0.267814576625824, "global_step": 44347, "epoch": 498, "lr": 9.36760822384623e-05}
{"train_loss": 0.33546221256256104, "global_step": 44348, "epoch": 498, "lr": 9.367580002826541e-05}
{"train_loss": 0.24641868472099304, "global_step": 44349, "epoch": 498, "lr": 9.367551781219683e-05}
{"train_loss": 0.3387294411659241, "global_step": 44350, "epoch": 498, "lr": 9.367523559025661e-05}
{"train_loss": 0.24945202469825745, "global_step": 44351, "epoch": 498, "lr": 9.36749533624448e-05}
{"train_loss": 0.33744558691978455, "global_step": 44352, "epoch": 498, "lr": 9.367467112876141e-05}
{"train_loss": 0.25536471605300903, "global_step": 44353, "epoch": 498, "lr": 9.36743888892065e-05}
{"train_loss": 0.306761234998703, "global_step": 44354, "epoch": 498, "lr": 9.367410664378011e-05}
{"train_loss": 0.2812429964542389, "global_step": 44355, "epoch": 498, "lr": 9.367382439248224e-05}
{"train_loss": 0.4002341330051422, "global_step": 44356, "epoch": 498, "lr": 9.367354213531298e-05}
{"train_loss": 0.2963164746761322, "global_step": 44357, "epoch": 498, "lr": 9.367325987227234e-05}
{"train_loss": 0.30997321009635925, "global_step": 44358, "epoch": 498, "lr": 9.367297760336036e-05}
{"train_loss": 0.3165362477302551, "global_step": 44359, "epoch": 498, "lr": 9.367269532857707e-05}
{"train_loss": 0.2694191336631775, "global_step": 44360, "epoch": 498, "lr": 9.367241304792253e-05}
{"train_loss": 0.40313956141471863, "global_step": 44361, "epoch": 498, "lr": 9.367213076139678e-05}
{"train_loss": 0.2912064790725708, "global_step": 44362, "epoch": 498, "lr": 9.367184846899983e-05}
{"train_loss": 0.4041566848754883, "global_step": 44363, "epoch": 498, "lr": 9.367156617073174e-05}
{"train_loss": 0.23857741057872772, "global_step": 44364, "epoch": 498, "lr": 9.367128386659252e-05}
{"train_loss": 0.28160446882247925, "global_step": 44365, "epoch": 498, "lr": 9.367100155658223e-05}
{"train_loss": 0.32273122668266296, "global_step": 44366, "epoch": 498, "lr": 9.367071924070092e-05}
{"train_loss": 0.2947085499763489, "global_step": 44367, "epoch": 498, "lr": 9.367043691894861e-05}
{"train_loss": 0.4510335624217987, "global_step": 44368, "epoch": 498, "lr": 9.367015459132535e-05}
{"train_loss": 0.27517229318618774, "global_step": 44369, "epoch": 498, "lr": 9.366987225783118e-05}
{"train_loss": 0.3358426094055176, "global_step": 44370, "epoch": 498, "lr": 9.36695899184661e-05}
{"train_loss": 0.35401397943496704, "global_step": 44371, "epoch": 498, "lr": 9.366930757323018e-05}
{"train_loss": 0.318032443523407, "global_step": 44372, "epoch": 498, "lr": 9.366902522212346e-05}
{"train_loss": 0.2506379187107086, "global_step": 44373, "epoch": 498, "lr": 9.366874286514597e-05}
{"train_loss": 0.23375672101974487, "global_step": 44374, "epoch": 498, "lr": 9.366846050229776e-05}
{"train_loss": 0.32341334223747253, "global_step": 44375, "epoch": 498, "lr": 9.366817813357885e-05}
{"train_loss": 0.1824464648962021, "global_step": 44376, "epoch": 498, "lr": 9.366789575898929e-05}
{"train_loss": 0.3158416152000427, "global_step": 44377, "epoch": 498, "lr": 9.36676133785291e-05}
{"train_loss": 0.2877802550792694, "global_step": 44378, "epoch": 498, "lr": 9.366733099219835e-05}
{"train_loss": 0.3586622178554535, "global_step": 44379, "epoch": 498, "lr": 9.366704859999704e-05}
{"train_loss": 0.27848225831985474, "global_step": 44380, "epoch": 498, "lr": 9.366676620192525e-05}
{"train_loss": 0.32365718483924866, "global_step": 44381, "epoch": 498, "lr": 9.366648379798298e-05}
{"train_loss": 0.3666495084762573, "global_step": 44382, "epoch": 498, "lr": 9.366620138817028e-05}
{"train_loss": 0.311425119638443, "global_step": 44383, "epoch": 498, "lr": 9.36659189724872e-05}
{"train_loss": 0.3002607524394989, "global_step": 44384, "epoch": 498, "lr": 9.366563655093377e-05}
{"train_loss": 0.3384374976158142, "global_step": 44385, "epoch": 498, "lr": 9.366535412351003e-05}
{"train_loss": 0.2299432009458542, "global_step": 44386, "epoch": 498, "lr": 9.366507169021601e-05}
{"train_loss": 0.3724784255027771, "global_step": 44387, "epoch": 498, "lr": 9.366478925105175e-05}
{"train_loss": 0.3044463098049164, "global_step": 44388, "epoch": 498, "lr": 9.366450680601731e-05}
{"train_loss": 0.3138478696346283, "global_step": 44389, "epoch": 498, "lr": 9.366422435511268e-05}
{"train_loss": 0.3683522343635559, "global_step": 44390, "epoch": 498, "lr": 9.366394189833794e-05}
{"train_loss": 0.2659716010093689, "global_step": 44391, "epoch": 498, "lr": 9.366365943569313e-05}
{"train_loss": 0.3119611144065857, "global_step": 44392, "epoch": 498, "lr": 9.366337696717824e-05}
{"train_loss": 0.2894727885723114, "global_step": 44393, "epoch": 498, "lr": 9.366309449279338e-05}
{"train_loss": 0.4727761149406433, "global_step": 44394, "epoch": 498, "lr": 9.366281201253852e-05}
{"train_loss": 0.25332972407341003, "global_step": 44395, "epoch": 498, "lr": 9.366252952641375e-05}
{"train_loss": 0.27823832631111145, "global_step": 44396, "epoch": 498, "lr": 9.366224703441906e-05}
{"train_loss": 0.31165170669555664, "global_step": 44397, "epoch": 498, "lr": 9.366196453655453e-05}
{"train_loss": 0.24633780121803284, "global_step": 44398, "epoch": 498, "lr": 9.366168203282018e-05}
{"train_loss": 0.22894763946533203, "global_step": 44399, "epoch": 498, "lr": 9.366139952321605e-05}
{"train_loss": 0.22203803062438965, "global_step": 44400, "epoch": 498, "lr": 9.366111700774217e-05}
{"train_loss": 0.2784896790981293, "global_step": 44401, "epoch": 498, "lr": 9.366083448639858e-05}
{"train_loss": 0.2781078517436981, "global_step": 44402, "epoch": 498, "lr": 9.366055195918533e-05}
{"train_loss": 0.3007185757160187, "global_step": 44403, "epoch": 498, "lr": 9.366026942610246e-05}
{"train_loss": 0.32440921664237976, "global_step": 44404, "epoch": 498, "lr": 9.365998688714998e-05}
{"train_loss": 0.27746227383613586, "global_step": 44405, "epoch": 498, "lr": 9.365970434232796e-05}
{"train_loss": 0.25568699836730957, "global_step": 44406, "epoch": 498, "lr": 9.365942179163643e-05}
{"train_loss": 0.31372302770614624, "global_step": 44407, "epoch": 498, "lr": 9.36591392350754e-05}
{"train_loss": 0.35697805881500244, "global_step": 44408, "epoch": 498, "lr": 9.365885667264496e-05}
{"train_loss": 0.2556734085083008, "global_step": 44409, "epoch": 498, "lr": 9.365857410434511e-05}
{"train_loss": 0.30627198316408005, "global_step": 44410, "epoch": 498, "lr": 9.365829153017589e-05, "val_loss": 2.9224398136138916}
{"train_loss": 0.2408406138420105, "global_step": 44411, "epoch": 499, "lr": 9.365800895013735e-05}
{"train_loss": 0.3246685266494751, "global_step": 44412, "epoch": 499, "lr": 9.365772636422952e-05}
{"train_loss": 0.3222348093986511, "global_step": 44413, "epoch": 499, "lr": 9.365744377245242e-05}
{"train_loss": 0.2694612145423889, "global_step": 44414, "epoch": 499, "lr": 9.365716117480614e-05}
{"train_loss": 0.24888022243976593, "global_step": 44415, "epoch": 499, "lr": 9.365687857129067e-05}
{"train_loss": 0.23274777829647064, "global_step": 44416, "epoch": 499, "lr": 9.365659596190609e-05}
{"train_loss": 0.27976906299591064, "global_step": 44417, "epoch": 499, "lr": 9.365631334665238e-05}
{"train_loss": 0.2915632724761963, "global_step": 44418, "epoch": 499, "lr": 9.365603072552962e-05}
{"train_loss": 0.26891130208969116, "global_step": 44419, "epoch": 499, "lr": 9.365574809853785e-05}
{"train_loss": 0.2426142543554306, "global_step": 44420, "epoch": 499, "lr": 9.36554654656771e-05}
{"train_loss": 0.30458328127861023, "global_step": 44421, "epoch": 499, "lr": 9.365518282694739e-05}
{"train_loss": 0.2905503511428833, "global_step": 44422, "epoch": 499, "lr": 9.365490018234877e-05}
{"train_loss": 0.23967887461185455, "global_step": 44423, "epoch": 499, "lr": 9.365461753188128e-05}
{"train_loss": 0.3172891139984131, "global_step": 44424, "epoch": 499, "lr": 9.365433487554499e-05}
{"train_loss": 0.38077935576438904, "global_step": 44425, "epoch": 499, "lr": 9.365405221333987e-05}
{"train_loss": 0.34751182794570923, "global_step": 44426, "epoch": 499, "lr": 9.365376954526602e-05}
{"train_loss": 0.2516178488731384, "global_step": 44427, "epoch": 499, "lr": 9.365348687132343e-05}
{"train_loss": 0.30973052978515625, "global_step": 44428, "epoch": 499, "lr": 9.365320419151216e-05}
{"train_loss": 0.23689310252666473, "global_step": 44429, "epoch": 499, "lr": 9.365292150583227e-05}
{"train_loss": 0.1883213371038437, "global_step": 44430, "epoch": 499, "lr": 9.365263881428377e-05}
{"train_loss": 0.2696782350540161, "global_step": 44431, "epoch": 499, "lr": 9.365235611686669e-05}
{"train_loss": 0.2044069468975067, "global_step": 44432, "epoch": 499, "lr": 9.36520734135811e-05}
{"train_loss": 0.322623610496521, "global_step": 44433, "epoch": 499, "lr": 9.365179070442703e-05}
{"train_loss": 0.3016662299633026, "global_step": 44434, "epoch": 499, "lr": 9.365150798940448e-05}
{"train_loss": 0.29959365725517273, "global_step": 44435, "epoch": 499, "lr": 9.365122526851353e-05}
{"train_loss": 0.3179599940776825, "global_step": 44436, "epoch": 499, "lr": 9.365094254175421e-05}
{"train_loss": 0.31930091977119446, "global_step": 44437, "epoch": 499, "lr": 9.365065980912655e-05}
{"train_loss": 0.28768712282180786, "global_step": 44438, "epoch": 499, "lr": 9.365037707063058e-05}
{"train_loss": 0.16383036971092224, "global_step": 44439, "epoch": 499, "lr": 9.365009432626637e-05}
{"train_loss": 0.38933584094047546, "global_step": 44440, "epoch": 499, "lr": 9.364981157603391e-05}
{"train_loss": 0.4691237211227417, "global_step": 44441, "epoch": 499, "lr": 9.36495288199333e-05}
{"train_loss": 0.2524029314517975, "global_step": 44442, "epoch": 499, "lr": 9.364924605796451e-05}
{"train_loss": 0.2224034070968628, "global_step": 44443, "epoch": 499, "lr": 9.364896329012762e-05}
{"train_loss": 0.31626108288764954, "global_step": 44444, "epoch": 499, "lr": 9.364868051642265e-05}
{"train_loss": 0.3475085496902466, "global_step": 44445, "epoch": 499, "lr": 9.364839773684968e-05}
{"train_loss": 0.2721598744392395, "global_step": 44446, "epoch": 499, "lr": 9.36481149514087e-05}
{"train_loss": 0.28684502840042114, "global_step": 44447, "epoch": 499, "lr": 9.364783216009975e-05}
{"train_loss": 0.32289445400238037, "global_step": 44448, "epoch": 499, "lr": 9.364754936292288e-05}
{"train_loss": 0.21342894434928894, "global_step": 44449, "epoch": 499, "lr": 9.364726655987814e-05}
{"train_loss": 0.36850517988204956, "global_step": 44450, "epoch": 499, "lr": 9.364698375096557e-05}
{"train_loss": 0.24660438299179077, "global_step": 44451, "epoch": 499, "lr": 9.364670093618517e-05}
{"train_loss": 0.4663078188896179, "global_step": 44452, "epoch": 499, "lr": 9.3646418115537e-05}
{"train_loss": 0.3304803669452667, "global_step": 44453, "epoch": 499, "lr": 9.364613528902113e-05}
{"train_loss": 0.19685909152030945, "global_step": 44454, "epoch": 499, "lr": 9.364585245663755e-05}
{"train_loss": 0.2320103645324707, "global_step": 44455, "epoch": 499, "lr": 9.364556961838632e-05}
{"train_loss": 0.25520575046539307, "global_step": 44456, "epoch": 499, "lr": 9.364528677426748e-05}
{"train_loss": 0.30813226103782654, "global_step": 44457, "epoch": 499, "lr": 9.364500392428105e-05}
{"train_loss": 0.3259752690792084, "global_step": 44458, "epoch": 499, "lr": 9.36447210684271e-05}
{"train_loss": 0.22950734198093414, "global_step": 44459, "epoch": 499, "lr": 9.364443820670564e-05}
{"train_loss": 0.4206058084964752, "global_step": 44460, "epoch": 499, "lr": 9.36441553391167e-05}
{"train_loss": 0.3340659439563751, "global_step": 44461, "epoch": 499, "lr": 9.364387246566036e-05}
{"train_loss": 0.31344011425971985, "global_step": 44462, "epoch": 499, "lr": 9.36435895863366e-05}
{"train_loss": 0.2524213194847107, "global_step": 44463, "epoch": 499, "lr": 9.364330670114554e-05}
{"train_loss": 0.23312830924987793, "global_step": 44464, "epoch": 499, "lr": 9.364302381008713e-05}
{"train_loss": 0.4017382860183716, "global_step": 44465, "epoch": 499, "lr": 9.364274091316148e-05}
{"train_loss": 0.28068092465400696, "global_step": 44466, "epoch": 499, "lr": 9.364245801036857e-05}
{"train_loss": 0.3098472058773041, "global_step": 44467, "epoch": 499, "lr": 9.364217510170848e-05}
{"train_loss": 0.3670210838317871, "global_step": 44468, "epoch": 499, "lr": 9.36418921871812e-05}
{"train_loss": 0.28862476348876953, "global_step": 44469, "epoch": 499, "lr": 9.364160926678684e-05}
{"train_loss": 0.3492821753025055, "global_step": 44470, "epoch": 499, "lr": 9.364132634052537e-05}
{"train_loss": 0.35062143206596375, "global_step": 44471, "epoch": 499, "lr": 9.364104340839687e-05}
{"train_loss": 0.2863081693649292, "global_step": 44472, "epoch": 499, "lr": 9.364076047040135e-05}
{"train_loss": 0.2777103781700134, "global_step": 44473, "epoch": 499, "lr": 9.364047752653888e-05}
{"train_loss": 0.2622104585170746, "global_step": 44474, "epoch": 499, "lr": 9.364019457680946e-05}
{"train_loss": 0.31981512904167175, "global_step": 44475, "epoch": 499, "lr": 9.363991162121316e-05}
{"train_loss": 0.3640086352825165, "global_step": 44476, "epoch": 499, "lr": 9.363962865975e-05}
{"train_loss": 0.28784307837486267, "global_step": 44477, "epoch": 499, "lr": 9.363934569242003e-05}
{"train_loss": 0.278375506401062, "global_step": 44478, "epoch": 499, "lr": 9.363906271922328e-05}
{"train_loss": 0.24959534406661987, "global_step": 44479, "epoch": 499, "lr": 9.36387797401598e-05}
{"train_loss": 0.27724552154541016, "global_step": 44480, "epoch": 499, "lr": 9.36384967552296e-05}
{"train_loss": 0.277717262506485, "global_step": 44481, "epoch": 499, "lr": 9.363821376443274e-05}
{"train_loss": 0.41904258728027344, "global_step": 44482, "epoch": 499, "lr": 9.363793076776925e-05}
{"train_loss": 0.3705529272556305, "global_step": 44483, "epoch": 499, "lr": 9.363764776523919e-05}
{"train_loss": 0.39710256457328796, "global_step": 44484, "epoch": 499, "lr": 9.363736475684257e-05}
{"train_loss": 0.37684330344200134, "global_step": 44485, "epoch": 499, "lr": 9.363708174257943e-05}
{"train_loss": 0.38318490982055664, "global_step": 44486, "epoch": 499, "lr": 9.363679872244984e-05}
{"train_loss": 0.26179182529449463, "global_step": 44487, "epoch": 499, "lr": 9.363651569645378e-05}
{"train_loss": 0.26371538639068604, "global_step": 44488, "epoch": 499, "lr": 9.363623266459135e-05}
{"train_loss": 0.3350701928138733, "global_step": 44489, "epoch": 499, "lr": 9.363594962686254e-05}
{"train_loss": 0.29786941409111023, "global_step": 44490, "epoch": 499, "lr": 9.363566658326743e-05}
{"train_loss": 0.2875586152076721, "global_step": 44491, "epoch": 499, "lr": 9.363538353380603e-05}
{"train_loss": 0.3671739101409912, "global_step": 44492, "epoch": 499, "lr": 9.363510047847837e-05}
{"train_loss": 0.35460013151168823, "global_step": 44493, "epoch": 499, "lr": 9.363481741728451e-05}
{"train_loss": 0.3873865306377411, "global_step": 44494, "epoch": 499, "lr": 9.363453435022449e-05}
{"train_loss": 0.2714202404022217, "global_step": 44495, "epoch": 499, "lr": 9.363425127729832e-05}
{"train_loss": 0.39336326718330383, "global_step": 44496, "epoch": 499, "lr": 9.363396819850607e-05}
{"train_loss": 0.2432820200920105, "global_step": 44497, "epoch": 499, "lr": 9.363368511384778e-05}
{"train_loss": 0.3045210540294647, "global_step": 44498, "epoch": 499, "lr": 9.363340202332345e-05}
{"train_loss": 0.30578402806533855, "global_step": 44499, "epoch": 499, "lr": 9.363311892693315e-05, "val_loss": 2.842109203338623}
{"train_loss": 0.3603265583515167, "global_step": 44500, "epoch": 500, "lr": 9.36328358246769e-05}
{"train_loss": 0.3047018349170685, "global_step": 44501, "epoch": 500, "lr": 9.363255271655475e-05}
{"train_loss": 0.2651045024394989, "global_step": 44502, "epoch": 500, "lr": 9.363226960256675e-05}
{"train_loss": 0.29055970907211304, "global_step": 44503, "epoch": 500, "lr": 9.363198648271291e-05}
{"train_loss": 0.22207173705101013, "global_step": 44504, "epoch": 500, "lr": 9.363170335699327e-05}
{"train_loss": 0.30192679166793823, "global_step": 44505, "epoch": 500, "lr": 9.36314202254079e-05}
{"train_loss": 0.45979195833206177, "global_step": 44506, "epoch": 500, "lr": 9.363113708795682e-05}
{"train_loss": 0.2566719949245453, "global_step": 44507, "epoch": 500, "lr": 9.363085394464004e-05}
{"train_loss": 0.47786232829093933, "global_step": 44508, "epoch": 500, "lr": 9.363057079545765e-05}
{"train_loss": 0.2975544333457947, "global_step": 44509, "epoch": 500, "lr": 9.363028764040964e-05}
{"train_loss": 0.24854134023189545, "global_step": 44510, "epoch": 500, "lr": 9.363000447949608e-05}
{"train_loss": 0.2726965546607971, "global_step": 44511, "epoch": 500, "lr": 9.3629721312717e-05}
{"train_loss": 0.2838887870311737, "global_step": 44512, "epoch": 500, "lr": 9.362943814007243e-05}
{"train_loss": 0.31883618235588074, "global_step": 44513, "epoch": 500, "lr": 9.362915496156242e-05}
{"train_loss": 0.3176264464855194, "global_step": 44514, "epoch": 500, "lr": 9.3628871777187e-05}
{"train_loss": 0.28389424085617065, "global_step": 44515, "epoch": 500, "lr": 9.362858858694621e-05}
{"train_loss": 0.28696930408477783, "global_step": 44516, "epoch": 500, "lr": 9.362830539084009e-05}
{"train_loss": 0.28190913796424866, "global_step": 44517, "epoch": 500, "lr": 9.362802218886867e-05}
{"train_loss": 0.27356866002082825, "global_step": 44518, "epoch": 500, "lr": 9.362773898103198e-05}
{"train_loss": 0.28795325756073, "global_step": 44519, "epoch": 500, "lr": 9.36274557673301e-05}
{"train_loss": 0.30824196338653564, "global_step": 44520, "epoch": 500, "lr": 9.362717254776303e-05}
{"train_loss": 0.37766119837760925, "global_step": 44521, "epoch": 500, "lr": 9.362688932233081e-05}
{"train_loss": 0.3463834822177887, "global_step": 44522, "epoch": 500, "lr": 9.362660609103351e-05}
{"train_loss": 0.30014461278915405, "global_step": 44523, "epoch": 500, "lr": 9.362632285387112e-05}
{"train_loss": 0.25562238693237305, "global_step": 44524, "epoch": 500, "lr": 9.362603961084371e-05}
{"train_loss": 0.27456045150756836, "global_step": 44525, "epoch": 500, "lr": 9.362575636195133e-05}
{"train_loss": 0.26596489548683167, "global_step": 44526, "epoch": 500, "lr": 9.362547310719396e-05}
{"train_loss": 0.2541705369949341, "global_step": 44527, "epoch": 500, "lr": 9.36251898465717e-05}
{"train_loss": 0.2367808073759079, "global_step": 44528, "epoch": 500, "lr": 9.362490658008456e-05}
{"train_loss": 0.26596587896347046, "global_step": 44529, "epoch": 500, "lr": 9.362462330773259e-05}
{"train_loss": 0.22584374248981476, "global_step": 44530, "epoch": 500, "lr": 9.362434002951583e-05}
{"train_loss": 0.24898909032344818, "global_step": 44531, "epoch": 500, "lr": 9.362405674543429e-05}
{"train_loss": 0.29719990491867065, "global_step": 44532, "epoch": 500, "lr": 9.362377345548803e-05}
{"train_loss": 0.3108394742012024, "global_step": 44533, "epoch": 500, "lr": 9.362349015967708e-05}
{"train_loss": 0.2669542729854584, "global_step": 44534, "epoch": 500, "lr": 9.36232068580015e-05}
{"train_loss": 0.25320568680763245, "global_step": 44535, "epoch": 500, "lr": 9.36229235504613e-05}
{"train_loss": 0.31329023838043213, "global_step": 44536, "epoch": 500, "lr": 9.362264023705653e-05}
{"train_loss": 0.22580650448799133, "global_step": 44537, "epoch": 500, "lr": 9.362235691778723e-05}
{"train_loss": 0.3727940618991852, "global_step": 44538, "epoch": 500, "lr": 9.362207359265345e-05}
{"train_loss": 0.31277957558631897, "global_step": 44539, "epoch": 500, "lr": 9.362179026165519e-05}
{"train_loss": 0.3132040500640869, "global_step": 44540, "epoch": 500, "lr": 9.362150692479252e-05}
{"train_loss": 0.38796141743659973, "global_step": 44541, "epoch": 500, "lr": 9.362122358206549e-05}
{"train_loss": 0.24633431434631348, "global_step": 44542, "epoch": 500, "lr": 9.36209402334741e-05}
{"train_loss": 0.25237518548965454, "global_step": 44543, "epoch": 500, "lr": 9.36206568790184e-05}
{"train_loss": 0.3439239263534546, "global_step": 44544, "epoch": 500, "lr": 9.362037351869844e-05}
{"train_loss": 0.30720531940460205, "global_step": 44545, "epoch": 500, "lr": 9.362009015251425e-05}
{"train_loss": 0.31550079584121704, "global_step": 44546, "epoch": 500, "lr": 9.361980678046588e-05}
{"train_loss": 0.29534438252449036, "global_step": 44547, "epoch": 500, "lr": 9.361952340255335e-05}
{"train_loss": 0.19114543497562408, "global_step": 44548, "epoch": 500, "lr": 9.361924001877672e-05}
{"train_loss": 0.26043879985809326, "global_step": 44549, "epoch": 500, "lr": 9.3618956629136e-05}
{"train_loss": 0.40640392899513245, "global_step": 44550, "epoch": 500, "lr": 9.361867323363125e-05}
{"train_loss": 0.3037331700325012, "global_step": 44551, "epoch": 500, "lr": 9.36183898322625e-05}
{"train_loss": 0.36731135845184326, "global_step": 44552, "epoch": 500, "lr": 9.361810642502977e-05}
{"train_loss": 0.23247435688972473, "global_step": 44553, "epoch": 500, "lr": 9.361782301193314e-05}
{"train_loss": 0.475542277097702, "global_step": 44554, "epoch": 500, "lr": 9.361753959297262e-05}
{"train_loss": 0.30705076456069946, "global_step": 44555, "epoch": 500, "lr": 9.361725616814825e-05}
{"train_loss": 0.36638936400413513, "global_step": 44556, "epoch": 500, "lr": 9.361697273746009e-05}
{"train_loss": 0.4538479149341583, "global_step": 44557, "epoch": 500, "lr": 9.361668930090813e-05}
{"train_loss": 0.328732967376709, "global_step": 44558, "epoch": 500, "lr": 9.361640585849245e-05}
{"train_loss": 0.23424911499023438, "global_step": 44559, "epoch": 500, "lr": 9.361612241021309e-05}
{"train_loss": 0.29421260952949524, "global_step": 44560, "epoch": 500, "lr": 9.361583895607007e-05}
{"train_loss": 0.2510611414909363, "global_step": 44561, "epoch": 500, "lr": 9.361555549606342e-05}
{"train_loss": 0.3735542893409729, "global_step": 44562, "epoch": 500, "lr": 9.361527203019319e-05}
{"train_loss": 0.32139062881469727, "global_step": 44563, "epoch": 500, "lr": 9.361498855845943e-05}
{"train_loss": 0.34964221715927124, "global_step": 44564, "epoch": 500, "lr": 9.361470508086215e-05}
{"train_loss": 0.30633315443992615, "global_step": 44565, "epoch": 500, "lr": 9.361442159740141e-05}
{"train_loss": 0.19425299763679504, "global_step": 44566, "epoch": 500, "lr": 9.361413810807725e-05}
{"train_loss": 0.32349011301994324, "global_step": 44567, "epoch": 500, "lr": 9.361385461288969e-05}
{"train_loss": 0.26601290702819824, "global_step": 44568, "epoch": 500, "lr": 9.36135711118388e-05}
{"train_loss": 0.3519063889980316, "global_step": 44569, "epoch": 500, "lr": 9.361328760492458e-05}
{"train_loss": 0.31841081380844116, "global_step": 44570, "epoch": 500, "lr": 9.361300409214709e-05}
{"train_loss": 0.3161334991455078, "global_step": 44571, "epoch": 500, "lr": 9.361272057350636e-05}
{"train_loss": 0.3121482729911804, "global_step": 44572, "epoch": 500, "lr": 9.361243704900243e-05}
{"train_loss": 0.30257660150527954, "global_step": 44573, "epoch": 500, "lr": 9.361215351863534e-05}
{"train_loss": 0.2504883110523224, "global_step": 44574, "epoch": 500, "lr": 9.361186998240512e-05}
{"train_loss": 0.25680461525917053, "global_step": 44575, "epoch": 500, "lr": 9.361158644031182e-05}
{"train_loss": 0.17139612138271332, "global_step": 44576, "epoch": 500, "lr": 9.361130289235549e-05}
{"train_loss": 0.3329199552536011, "global_step": 44577, "epoch": 500, "lr": 9.361101933853614e-05}
{"train_loss": 0.2577059268951416, "global_step": 44578, "epoch": 500, "lr": 9.361073577885383e-05}
{"train_loss": 0.2913898527622223, "global_step": 44579, "epoch": 500, "lr": 9.361045221330858e-05}
{"train_loss": 0.29405108094215393, "global_step": 44580, "epoch": 500, "lr": 9.361016864190042e-05}
{"train_loss": 0.22505620121955872, "global_step": 44581, "epoch": 500, "lr": 9.360988506462942e-05}
{"train_loss": 0.34290099143981934, "global_step": 44582, "epoch": 500, "lr": 9.36096014814956e-05}
{"train_loss": 0.32461804151535034, "global_step": 44583, "epoch": 500, "lr": 9.3609317892499e-05}
{"train_loss": 0.26572495698928833, "global_step": 44584, "epoch": 500, "lr": 9.360903429763967e-05}
{"train_loss": 0.30386024713516235, "global_step": 44585, "epoch": 500, "lr": 9.360875069691762e-05}
{"train_loss": 0.3152998983860016, "global_step": 44586, "epoch": 500, "lr": 9.360846709033291e-05}
{"train_loss": 0.28761523962020874, "global_step": 44587, "epoch": 500, "lr": 9.360818347788558e-05}
{"train_loss": 0.29960997051067567, "global_step": 44588, "epoch": 500, "lr": 9.360789985957566e-05, "train/sim_max_reward_0": 0.6699293029697042, "train/sim_max_reward_1": 0.5420557349218095, "train/sim_max_reward_2": 0.3960876620467213, "train/sim_max_reward_3": 0.5914917295406921, "train/sim_max_reward_4": 0.9911369178155945, "train/sim_max_reward_5": 0.7197589656478467, "test/sim_max_reward_4500000": 0.8582525777909374, "test/sim_max_reward_4500001": 0.9168509388171616, "test/sim_max_reward_4500002": 0.9478757279217201, "test/sim_max_reward_4500003": 0.3557511439461016, "test/sim_max_reward_4500004": 0.5088600779551792, "test/sim_max_reward_4500005": 0.1671235566907441, "test/sim_max_reward_4500006": 0.09684017027104046, "test/sim_max_reward_4500007": 0.0, "test/sim_max_reward_4500008": 0.15463443322879078, "test/sim_max_reward_4500009": 0.3726307011977666, "test/sim_max_reward_4500010": 0.0011085005590369362, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.7235310058643417, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9839333551435122, "test/sim_max_reward_4500016": 0.49347044362404785, "test/sim_max_reward_4500017": 0.9991948954596053, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9689475813530511, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.738056908025605, "test/sim_max_reward_4500022": 0.9734271129022554, "test/sim_max_reward_4500023": 0.7026866865906035, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9936019477648755, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 0.44685763912424953, "test/sim_max_reward_4500028": 0.6777839476386021, "test/sim_max_reward_4500029": 0.316598357174545, "test/sim_max_reward_4500030": 0.7377532314470066, "test/sim_max_reward_4500031": 0.35697024334511995, "test/sim_max_reward_4500032": 1.0, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.5756498814294904, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.694923807650072, "test/sim_max_reward_4500038": 0.16219654909558698, "test/sim_max_reward_4500039": 0.8761043543355825, "test/sim_max_reward_4500040": 0.94547019414488, "test/sim_max_reward_4500041": 0.3329896126113719, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.43054515566461715, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.7219090258848201, "test/sim_max_reward_4500046": 0.3732556556720406, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.4951700646677293, "test/sim_max_reward_4500049": 0.7233203090410177, "train/mean_score": 0.6517433854903948, "test/mean_score": 0.5764855158806622, "val_loss": 2.8415091037750244, "train_action_mse_error": 36.53310775756836}
{"train_loss": 0.21826863288879395, "global_step": 44589, "epoch": 501, "lr": 9.36076162354032e-05}
{"train_loss": 0.22447605431079865, "global_step": 44590, "epoch": 501, "lr": 9.36073326053682e-05}
{"train_loss": 0.2635130286216736, "global_step": 44591, "epoch": 501, "lr": 9.360704896947074e-05}
{"train_loss": 0.3074054718017578, "global_step": 44592, "epoch": 501, "lr": 9.360676532771084e-05}
{"train_loss": 0.3708694279193878, "global_step": 44593, "epoch": 501, "lr": 9.360648168008854e-05}
{"train_loss": 0.30445122718811035, "global_step": 44594, "epoch": 501, "lr": 9.360619802660389e-05}
{"train_loss": 0.30646127462387085, "global_step": 44595, "epoch": 501, "lr": 9.360591436725691e-05}
{"train_loss": 0.2671937346458435, "global_step": 44596, "epoch": 501, "lr": 9.360563070204766e-05}
{"train_loss": 0.23840320110321045, "global_step": 44597, "epoch": 501, "lr": 9.360534703097615e-05}
{"train_loss": 0.1548546552658081, "global_step": 44598, "epoch": 501, "lr": 9.360506335404244e-05}
{"train_loss": 0.29490336775779724, "global_step": 44599, "epoch": 501, "lr": 9.360477967124656e-05}
{"train_loss": 0.3448624312877655, "global_step": 44600, "epoch": 501, "lr": 9.360449598258854e-05}
{"train_loss": 0.34809550642967224, "global_step": 44601, "epoch": 501, "lr": 9.360421228806844e-05}
{"train_loss": 0.2795780897140503, "global_step": 44602, "epoch": 501, "lr": 9.360392858768627e-05}
{"train_loss": 0.24843664467334747, "global_step": 44603, "epoch": 501, "lr": 9.360364488144209e-05}
{"train_loss": 0.26746153831481934, "global_step": 44604, "epoch": 501, "lr": 9.360336116933594e-05}
{"train_loss": 0.22318816184997559, "global_step": 44605, "epoch": 501, "lr": 9.360307745136785e-05}
{"train_loss": 0.2569526731967926, "global_step": 44606, "epoch": 501, "lr": 9.360279372753784e-05}
{"train_loss": 0.40951937437057495, "global_step": 44607, "epoch": 501, "lr": 9.360250999784597e-05}
{"train_loss": 0.19644023478031158, "global_step": 44608, "epoch": 501, "lr": 9.36022262622923e-05}
{"train_loss": 0.33650124073028564, "global_step": 44609, "epoch": 501, "lr": 9.360194252087682e-05}
{"train_loss": 0.34220170974731445, "global_step": 44610, "epoch": 501, "lr": 9.360165877359958e-05}
{"train_loss": 0.48791804909706116, "global_step": 44611, "epoch": 501, "lr": 9.360137502046065e-05}
{"train_loss": 0.32942214608192444, "global_step": 44612, "epoch": 501, "lr": 9.360109126146003e-05}
{"train_loss": 0.21921643614768982, "global_step": 44613, "epoch": 501, "lr": 9.360080749659778e-05}
{"train_loss": 0.21827960014343262, "global_step": 44614, "epoch": 501, "lr": 9.360052372587393e-05}
{"train_loss": 0.2181198000907898, "global_step": 44615, "epoch": 501, "lr": 9.360023994928853e-05}
{"train_loss": 0.22654592990875244, "global_step": 44616, "epoch": 501, "lr": 9.359995616684161e-05}
{"train_loss": 0.2049754410982132, "global_step": 44617, "epoch": 501, "lr": 9.35996723785332e-05}
{"train_loss": 0.24992673099040985, "global_step": 44618, "epoch": 501, "lr": 9.359938858436335e-05}
{"train_loss": 0.25968965888023376, "global_step": 44619, "epoch": 501, "lr": 9.359910478433209e-05}
{"train_loss": 0.32424497604370117, "global_step": 44620, "epoch": 501, "lr": 9.359882097843947e-05}
{"train_loss": 0.32613083720207214, "global_step": 44621, "epoch": 501, "lr": 9.35985371666855e-05}
{"train_loss": 0.3596591651439667, "global_step": 44622, "epoch": 501, "lr": 9.359825334907026e-05}
{"train_loss": 0.24753542244434357, "global_step": 44623, "epoch": 501, "lr": 9.359796952559375e-05}
{"train_loss": 0.17996153235435486, "global_step": 44624, "epoch": 501, "lr": 9.359768569625605e-05}
{"train_loss": 0.24544821679592133, "global_step": 44625, "epoch": 501, "lr": 9.359740186105715e-05}
{"train_loss": 0.30424919724464417, "global_step": 44626, "epoch": 501, "lr": 9.359711801999711e-05}
{"train_loss": 0.3212502896785736, "global_step": 44627, "epoch": 501, "lr": 9.359683417307598e-05}
{"train_loss": 0.34200525283813477, "global_step": 44628, "epoch": 501, "lr": 9.359655032029377e-05}
{"train_loss": 0.2962215840816498, "global_step": 44629, "epoch": 501, "lr": 9.359626646165054e-05}
{"train_loss": 0.3124412000179291, "global_step": 44630, "epoch": 501, "lr": 9.359598259714634e-05}
{"train_loss": 0.27994459867477417, "global_step": 44631, "epoch": 501, "lr": 9.359569872678117e-05}
{"train_loss": 0.21975824236869812, "global_step": 44632, "epoch": 501, "lr": 9.359541485055511e-05}
{"train_loss": 0.2592177093029022, "global_step": 44633, "epoch": 501, "lr": 9.359513096846817e-05}
{"train_loss": 0.2827751338481903, "global_step": 44634, "epoch": 501, "lr": 9.35948470805204e-05}
{"train_loss": 0.25950077176094055, "global_step": 44635, "epoch": 501, "lr": 9.359456318671182e-05}
{"train_loss": 0.312980055809021, "global_step": 44636, "epoch": 501, "lr": 9.35942792870425e-05}
{"train_loss": 0.3268187642097473, "global_step": 44637, "epoch": 501, "lr": 9.359399538151244e-05}
{"train_loss": 0.2229137122631073, "global_step": 44638, "epoch": 501, "lr": 9.359371147012172e-05}
{"train_loss": 0.2915993332862854, "global_step": 44639, "epoch": 501, "lr": 9.359342755287034e-05}
{"train_loss": 0.2768978178501129, "global_step": 44640, "epoch": 501, "lr": 9.359314362975837e-05}
{"train_loss": 0.33679404854774475, "global_step": 44641, "epoch": 501, "lr": 9.359285970078581e-05}
{"train_loss": 0.27452945709228516, "global_step": 44642, "epoch": 501, "lr": 9.359257576595275e-05}
{"train_loss": 0.3115604519844055, "global_step": 44643, "epoch": 501, "lr": 9.359229182525917e-05}
{"train_loss": 0.23601233959197998, "global_step": 44644, "epoch": 501, "lr": 9.359200787870517e-05}
{"train_loss": 0.31970441341400146, "global_step": 44645, "epoch": 501, "lr": 9.359172392629073e-05}
{"train_loss": 0.2720370292663574, "global_step": 44646, "epoch": 501, "lr": 9.359143996801592e-05}
{"train_loss": 0.29367128014564514, "global_step": 44647, "epoch": 501, "lr": 9.359115600388077e-05}
{"train_loss": 0.36291995644569397, "global_step": 44648, "epoch": 501, "lr": 9.359087203388533e-05}
{"train_loss": 0.2919287085533142, "global_step": 44649, "epoch": 501, "lr": 9.359058805802963e-05}
{"train_loss": 0.34499359130859375, "global_step": 44650, "epoch": 501, "lr": 9.35903040763137e-05}
{"train_loss": 0.21568001806735992, "global_step": 44651, "epoch": 501, "lr": 9.359002008873759e-05}
{"train_loss": 0.24485188722610474, "global_step": 44652, "epoch": 501, "lr": 9.358973609530133e-05}
{"train_loss": 0.36691489815711975, "global_step": 44653, "epoch": 501, "lr": 9.358945209600496e-05}
{"train_loss": 0.28851422667503357, "global_step": 44654, "epoch": 501, "lr": 9.358916809084852e-05}
{"train_loss": 0.3969319760799408, "global_step": 44655, "epoch": 501, "lr": 9.358888407983205e-05}
{"train_loss": 0.2841385304927826, "global_step": 44656, "epoch": 501, "lr": 9.358860006295558e-05}
{"train_loss": 0.2838202118873596, "global_step": 44657, "epoch": 501, "lr": 9.358831604021915e-05}
{"train_loss": 0.2822154462337494, "global_step": 44658, "epoch": 501, "lr": 9.358803201162282e-05}
{"train_loss": 0.21136413514614105, "global_step": 44659, "epoch": 501, "lr": 9.35877479771666e-05}
{"train_loss": 0.29708170890808105, "global_step": 44660, "epoch": 501, "lr": 9.358746393685054e-05}
{"train_loss": 0.31439414620399475, "global_step": 44661, "epoch": 501, "lr": 9.358717989067467e-05}
{"train_loss": 0.33452439308166504, "global_step": 44662, "epoch": 501, "lr": 9.358689583863905e-05}
{"train_loss": 0.3290227949619293, "global_step": 44663, "epoch": 501, "lr": 9.358661178074368e-05}
{"train_loss": 0.2703932523727417, "global_step": 44664, "epoch": 501, "lr": 9.358632771698863e-05}
{"train_loss": 0.40474703907966614, "global_step": 44665, "epoch": 501, "lr": 9.358604364737394e-05}
{"train_loss": 0.29111868143081665, "global_step": 44666, "epoch": 501, "lr": 9.358575957189963e-05}
{"train_loss": 0.2925329804420471, "global_step": 44667, "epoch": 501, "lr": 9.358547549056575e-05}
{"train_loss": 0.35806575417518616, "global_step": 44668, "epoch": 501, "lr": 9.358519140337234e-05}
{"train_loss": 0.3198418915271759, "global_step": 44669, "epoch": 501, "lr": 9.358490731031941e-05}
{"train_loss": 0.19239485263824463, "global_step": 44670, "epoch": 501, "lr": 9.358462321140704e-05}
{"train_loss": 0.172834575176239, "global_step": 44671, "epoch": 501, "lr": 9.358433910663525e-05}
{"train_loss": 0.3649277091026306, "global_step": 44672, "epoch": 501, "lr": 9.358405499600407e-05}
{"train_loss": 0.2466459572315216, "global_step": 44673, "epoch": 501, "lr": 9.358377087951353e-05}
{"train_loss": 0.2290368676185608, "global_step": 44674, "epoch": 501, "lr": 9.358348675716371e-05}
{"train_loss": 0.39646807312965393, "global_step": 44675, "epoch": 501, "lr": 9.358320262895462e-05}
{"train_loss": 0.42381981015205383, "global_step": 44676, "epoch": 501, "lr": 9.358291849488627e-05}
{"train_loss": 0.2891373528858249, "global_step": 44677, "epoch": 501, "lr": 9.358263435495875e-05, "val_loss": 2.943723201751709}
{"train_loss": 0.2365817278623581, "global_step": 44678, "epoch": 502, "lr": 9.358235020917207e-05}
{"train_loss": 0.3672265410423279, "global_step": 44679, "epoch": 502, "lr": 9.358206605752628e-05}
{"train_loss": 0.25321656465530396, "global_step": 44680, "epoch": 502, "lr": 9.358178190002141e-05}
{"train_loss": 0.33678680658340454, "global_step": 44681, "epoch": 502, "lr": 9.35814977366575e-05}
{"train_loss": 0.31370919942855835, "global_step": 44682, "epoch": 502, "lr": 9.35812135674346e-05}
{"train_loss": 0.24303089082241058, "global_step": 44683, "epoch": 502, "lr": 9.358092939235272e-05}
{"train_loss": 0.28156325221061707, "global_step": 44684, "epoch": 502, "lr": 9.358064521141192e-05}
{"train_loss": 0.251259982585907, "global_step": 44685, "epoch": 502, "lr": 9.358036102461223e-05}
{"train_loss": 0.38403448462486267, "global_step": 44686, "epoch": 502, "lr": 9.35800768319537e-05}
{"train_loss": 0.31416240334510803, "global_step": 44687, "epoch": 502, "lr": 9.357979263343635e-05}
{"train_loss": 0.34276339411735535, "global_step": 44688, "epoch": 502, "lr": 9.357950842906025e-05}
{"train_loss": 0.4418638348579407, "global_step": 44689, "epoch": 502, "lr": 9.357922421882538e-05}
{"train_loss": 0.31875598430633545, "global_step": 44690, "epoch": 502, "lr": 9.357894000273184e-05}
{"train_loss": 0.29207122325897217, "global_step": 44691, "epoch": 502, "lr": 9.357865578077965e-05}
{"train_loss": 0.2593059241771698, "global_step": 44692, "epoch": 502, "lr": 9.357837155296883e-05}
{"train_loss": 0.31495702266693115, "global_step": 44693, "epoch": 502, "lr": 9.357808731929942e-05}
{"train_loss": 0.21654221415519714, "global_step": 44694, "epoch": 502, "lr": 9.357780307977147e-05}
{"train_loss": 0.267796128988266, "global_step": 44695, "epoch": 502, "lr": 9.357751883438503e-05}
{"train_loss": 0.3047254681587219, "global_step": 44696, "epoch": 502, "lr": 9.35772345831401e-05}
{"train_loss": 0.3521527647972107, "global_step": 44697, "epoch": 502, "lr": 9.357695032603676e-05}
{"train_loss": 0.28100234270095825, "global_step": 44698, "epoch": 502, "lr": 9.357666606307503e-05}
{"train_loss": 0.28395020961761475, "global_step": 44699, "epoch": 502, "lr": 9.357638179425494e-05}
{"train_loss": 0.31971997022628784, "global_step": 44700, "epoch": 502, "lr": 9.357609751957655e-05}
{"train_loss": 0.26429736614227295, "global_step": 44701, "epoch": 502, "lr": 9.357581323903987e-05}
{"train_loss": 0.26164039969444275, "global_step": 44702, "epoch": 502, "lr": 9.357552895264496e-05}
{"train_loss": 0.26540228724479675, "global_step": 44703, "epoch": 502, "lr": 9.357524466039185e-05}
{"train_loss": 0.34314048290252686, "global_step": 44704, "epoch": 502, "lr": 9.357496036228057e-05}
{"train_loss": 0.21631492674350739, "global_step": 44705, "epoch": 502, "lr": 9.357467605831119e-05}
{"train_loss": 0.18973277509212494, "global_step": 44706, "epoch": 502, "lr": 9.35743917484837e-05}
{"train_loss": 0.3045981526374817, "global_step": 44707, "epoch": 502, "lr": 9.357410743279818e-05}
{"train_loss": 0.29293909668922424, "global_step": 44708, "epoch": 502, "lr": 9.357382311125466e-05}
{"train_loss": 0.33648258447647095, "global_step": 44709, "epoch": 502, "lr": 9.357353878385315e-05}
{"train_loss": 0.329085111618042, "global_step": 44710, "epoch": 502, "lr": 9.357325445059373e-05}
{"train_loss": 0.32951730489730835, "global_step": 44711, "epoch": 502, "lr": 9.35729701114764e-05}
{"train_loss": 0.25624576210975647, "global_step": 44712, "epoch": 502, "lr": 9.357268576650122e-05}
{"train_loss": 0.2419663667678833, "global_step": 44713, "epoch": 502, "lr": 9.357240141566823e-05}
{"train_loss": 0.23494085669517517, "global_step": 44714, "epoch": 502, "lr": 9.357211705897747e-05}
{"train_loss": 0.2873559594154358, "global_step": 44715, "epoch": 502, "lr": 9.357183269642894e-05}
{"train_loss": 0.2769942879676819, "global_step": 44716, "epoch": 502, "lr": 9.357154832802273e-05}
{"train_loss": 0.24809813499450684, "global_step": 44717, "epoch": 502, "lr": 9.357126395375885e-05}
{"train_loss": 0.3557296395301819, "global_step": 44718, "epoch": 502, "lr": 9.357097957363733e-05}
{"train_loss": 0.2871055603027344, "global_step": 44719, "epoch": 502, "lr": 9.357069518765824e-05}
{"train_loss": 0.3318631649017334, "global_step": 44720, "epoch": 502, "lr": 9.357041079582161e-05}
{"train_loss": 0.3311181366443634, "global_step": 44721, "epoch": 502, "lr": 9.357012639812745e-05}
{"train_loss": 0.3609102964401245, "global_step": 44722, "epoch": 502, "lr": 9.356984199457584e-05}
{"train_loss": 0.2684326171875, "global_step": 44723, "epoch": 502, "lr": 9.356955758516678e-05}
{"train_loss": 0.35909631848335266, "global_step": 44724, "epoch": 502, "lr": 9.356927316990031e-05}
{"train_loss": 0.3034393787384033, "global_step": 44725, "epoch": 502, "lr": 9.356898874877651e-05}
{"train_loss": 0.2697775661945343, "global_step": 44726, "epoch": 502, "lr": 9.356870432179538e-05}
{"train_loss": 0.31413498520851135, "global_step": 44727, "epoch": 502, "lr": 9.356841988895697e-05}
{"train_loss": 0.23234522342681885, "global_step": 44728, "epoch": 502, "lr": 9.356813545026131e-05}
{"train_loss": 0.23750397562980652, "global_step": 44729, "epoch": 502, "lr": 9.356785100570845e-05}
{"train_loss": 0.24898521602153778, "global_step": 44730, "epoch": 502, "lr": 9.356756655529842e-05}
{"train_loss": 0.2779441773891449, "global_step": 44731, "epoch": 502, "lr": 9.356728209903128e-05}
{"train_loss": 0.3311154842376709, "global_step": 44732, "epoch": 502, "lr": 9.356699763690702e-05}
{"train_loss": 0.3191717863082886, "global_step": 44733, "epoch": 502, "lr": 9.356671316892573e-05}
{"train_loss": 0.43942126631736755, "global_step": 44734, "epoch": 502, "lr": 9.356642869508741e-05}
{"train_loss": 0.25135332345962524, "global_step": 44735, "epoch": 502, "lr": 9.356614421539214e-05}
{"train_loss": 0.26236429810523987, "global_step": 44736, "epoch": 502, "lr": 9.356585972983991e-05}
{"train_loss": 0.20547738671302795, "global_step": 44737, "epoch": 502, "lr": 9.356557523843078e-05}
{"train_loss": 0.27083587646484375, "global_step": 44738, "epoch": 502, "lr": 9.35652907411648e-05}
{"train_loss": 0.3119264841079712, "global_step": 44739, "epoch": 502, "lr": 9.3565006238042e-05}
{"train_loss": 0.3310125172138214, "global_step": 44740, "epoch": 502, "lr": 9.35647217290624e-05}
{"train_loss": 0.2937961518764496, "global_step": 44741, "epoch": 502, "lr": 9.356443721422607e-05}
{"train_loss": 0.29636359214782715, "global_step": 44742, "epoch": 502, "lr": 9.356415269353303e-05}
{"train_loss": 0.270179808139801, "global_step": 44743, "epoch": 502, "lr": 9.356386816698333e-05}
{"train_loss": 0.2708282768726349, "global_step": 44744, "epoch": 502, "lr": 9.356358363457699e-05}
{"train_loss": 0.2877497375011444, "global_step": 44745, "epoch": 502, "lr": 9.356329909631405e-05}
{"train_loss": 0.27214205265045166, "global_step": 44746, "epoch": 502, "lr": 9.356301455219454e-05}
{"train_loss": 0.40377816557884216, "global_step": 44747, "epoch": 502, "lr": 9.356273000221854e-05}
{"train_loss": 0.27871453762054443, "global_step": 44748, "epoch": 502, "lr": 9.356244544638606e-05}
{"train_loss": 0.3286137878894806, "global_step": 44749, "epoch": 502, "lr": 9.356216088469714e-05}
{"train_loss": 0.23964789509773254, "global_step": 44750, "epoch": 502, "lr": 9.356187631715181e-05}
{"train_loss": 0.1805220991373062, "global_step": 44751, "epoch": 502, "lr": 9.356159174375012e-05}
{"train_loss": 0.2627546489238739, "global_step": 44752, "epoch": 502, "lr": 9.35613071644921e-05}
{"train_loss": 0.4278537631034851, "global_step": 44753, "epoch": 502, "lr": 9.35610225793778e-05}
{"train_loss": 0.26255178451538086, "global_step": 44754, "epoch": 502, "lr": 9.356073798840726e-05}
{"train_loss": 0.32978367805480957, "global_step": 44755, "epoch": 502, "lr": 9.35604533915805e-05}
{"train_loss": 0.34856510162353516, "global_step": 44756, "epoch": 502, "lr": 9.356016878889756e-05}
{"train_loss": 0.336759090423584, "global_step": 44757, "epoch": 502, "lr": 9.35598841803585e-05}
{"train_loss": 0.28870731592178345, "global_step": 44758, "epoch": 502, "lr": 9.355959956596334e-05}
{"train_loss": 0.29136115312576294, "global_step": 44759, "epoch": 502, "lr": 9.355931494571212e-05}
{"train_loss": 0.23909783363342285, "global_step": 44760, "epoch": 502, "lr": 9.355903031960488e-05}
{"train_loss": 0.23241166770458221, "global_step": 44761, "epoch": 502, "lr": 9.355874568764167e-05}
{"train_loss": 0.21516171097755432, "global_step": 44762, "epoch": 502, "lr": 9.355846104982251e-05}
{"train_loss": 0.3247627913951874, "global_step": 44763, "epoch": 502, "lr": 9.355817640614744e-05}
{"train_loss": 0.22632017731666565, "global_step": 44764, "epoch": 502, "lr": 9.35578917566165e-05}
{"train_loss": 0.3805469870567322, "global_step": 44765, "epoch": 502, "lr": 9.355760710122973e-05}
{"train_loss": 0.29347256981254966, "global_step": 44766, "epoch": 502, "lr": 9.35573224399872e-05, "val_loss": 2.8157527446746826}
{"train_loss": 0.2896312475204468, "global_step": 44767, "epoch": 503, "lr": 9.35570377728889e-05}
{"train_loss": 0.25615638494491577, "global_step": 44768, "epoch": 503, "lr": 9.355675309993488e-05}
{"train_loss": 0.3136630058288574, "global_step": 44769, "epoch": 503, "lr": 9.35564684211252e-05}
{"train_loss": 0.28408700227737427, "global_step": 44770, "epoch": 503, "lr": 9.355618373645988e-05}
{"train_loss": 0.1860089749097824, "global_step": 44771, "epoch": 503, "lr": 9.355589904593895e-05}
{"train_loss": 0.36591455340385437, "global_step": 44772, "epoch": 503, "lr": 9.355561434956248e-05}
{"train_loss": 0.2958255410194397, "global_step": 44773, "epoch": 503, "lr": 9.355532964733046e-05}
{"train_loss": 0.22708295285701752, "global_step": 44774, "epoch": 503, "lr": 9.355504493924298e-05}
{"train_loss": 0.2617488503456116, "global_step": 44775, "epoch": 503, "lr": 9.355476022530005e-05}
{"train_loss": 0.24495859444141388, "global_step": 44776, "epoch": 503, "lr": 9.355447550550172e-05}
{"train_loss": 0.38314107060432434, "global_step": 44777, "epoch": 503, "lr": 9.355419077984802e-05}
{"train_loss": 0.24550557136535645, "global_step": 44778, "epoch": 503, "lr": 9.355390604833898e-05}
{"train_loss": 0.2617247700691223, "global_step": 44779, "epoch": 503, "lr": 9.355362131097465e-05}
{"train_loss": 0.24199025332927704, "global_step": 44780, "epoch": 503, "lr": 9.355333656775507e-05}
{"train_loss": 0.28052541613578796, "global_step": 44781, "epoch": 503, "lr": 9.355305181868028e-05}
{"train_loss": 0.3305531442165375, "global_step": 44782, "epoch": 503, "lr": 9.35527670637503e-05}
{"train_loss": 0.27356618642807007, "global_step": 44783, "epoch": 503, "lr": 9.35524823029652e-05}
{"train_loss": 0.24022474884986877, "global_step": 44784, "epoch": 503, "lr": 9.355219753632499e-05}
{"train_loss": 0.3576880693435669, "global_step": 44785, "epoch": 503, "lr": 9.35519127638297e-05}
{"train_loss": 0.1971348375082016, "global_step": 44786, "epoch": 503, "lr": 9.355162798547941e-05}
{"train_loss": 0.2304912954568863, "global_step": 44787, "epoch": 503, "lr": 9.355134320127412e-05}
{"train_loss": 0.29414549469947815, "global_step": 44788, "epoch": 503, "lr": 9.35510584112139e-05}
{"train_loss": 0.30416783690452576, "global_step": 44789, "epoch": 503, "lr": 9.355077361529876e-05}
{"train_loss": 0.29830819368362427, "global_step": 44790, "epoch": 503, "lr": 9.355048881352874e-05}
{"train_loss": 0.20356781780719757, "global_step": 44791, "epoch": 503, "lr": 9.355020400590391e-05}
{"train_loss": 0.2544107735157013, "global_step": 44792, "epoch": 503, "lr": 9.354991919242426e-05}
{"train_loss": 0.2853795886039734, "global_step": 44793, "epoch": 503, "lr": 9.354963437308987e-05}
{"train_loss": 0.28225481510162354, "global_step": 44794, "epoch": 503, "lr": 9.354934954790075e-05}
{"train_loss": 0.21923828125, "global_step": 44795, "epoch": 503, "lr": 9.354906471685698e-05}
{"train_loss": 0.3372843861579895, "global_step": 44796, "epoch": 503, "lr": 9.354877987995854e-05}
{"train_loss": 0.3215115964412689, "global_step": 44797, "epoch": 503, "lr": 9.35484950372055e-05}
{"train_loss": 0.287594735622406, "global_step": 44798, "epoch": 503, "lr": 9.354821018859792e-05}
{"train_loss": 0.20811176300048828, "global_step": 44799, "epoch": 503, "lr": 9.354792533413579e-05}
{"train_loss": 0.3004913330078125, "global_step": 44800, "epoch": 503, "lr": 9.354764047381918e-05}
{"train_loss": 0.25140246748924255, "global_step": 44801, "epoch": 503, "lr": 9.354735560764811e-05}
{"train_loss": 0.34051966667175293, "global_step": 44802, "epoch": 503, "lr": 9.354707073562263e-05}
{"train_loss": 0.25159403681755066, "global_step": 44803, "epoch": 503, "lr": 9.354678585774279e-05}
{"train_loss": 0.3820842206478119, "global_step": 44804, "epoch": 503, "lr": 9.354650097400862e-05}
{"train_loss": 0.35412585735321045, "global_step": 44805, "epoch": 503, "lr": 9.354621608442013e-05}
{"train_loss": 0.3194897174835205, "global_step": 44806, "epoch": 503, "lr": 9.354593118897741e-05}
{"train_loss": 0.2380346804857254, "global_step": 44807, "epoch": 503, "lr": 9.354564628768045e-05}
{"train_loss": 0.2527955174446106, "global_step": 44808, "epoch": 503, "lr": 9.354536138052931e-05}
{"train_loss": 0.2966160476207733, "global_step": 44809, "epoch": 503, "lr": 9.354507646752404e-05}
{"train_loss": 0.29378318786621094, "global_step": 44810, "epoch": 503, "lr": 9.354479154866465e-05}
{"train_loss": 0.3926045596599579, "global_step": 44811, "epoch": 503, "lr": 9.35445066239512e-05}
{"train_loss": 0.26960593461990356, "global_step": 44812, "epoch": 503, "lr": 9.354422169338373e-05}
{"train_loss": 0.29379749298095703, "global_step": 44813, "epoch": 503, "lr": 9.354393675696226e-05}
{"train_loss": 0.2997078001499176, "global_step": 44814, "epoch": 503, "lr": 9.354365181468684e-05}
{"train_loss": 0.31991174817085266, "global_step": 44815, "epoch": 503, "lr": 9.354336686655751e-05}
{"train_loss": 0.3781674802303314, "global_step": 44816, "epoch": 503, "lr": 9.354308191257431e-05}
{"train_loss": 0.31306740641593933, "global_step": 44817, "epoch": 503, "lr": 9.354279695273725e-05}
{"train_loss": 0.3034781813621521, "global_step": 44818, "epoch": 503, "lr": 9.354251198704642e-05}
{"train_loss": 0.3525483310222626, "global_step": 44819, "epoch": 503, "lr": 9.354222701550182e-05}
{"train_loss": 0.28107789158821106, "global_step": 44820, "epoch": 503, "lr": 9.35419420381035e-05}
{"train_loss": 0.2810746133327484, "global_step": 44821, "epoch": 503, "lr": 9.354165705485148e-05}
{"train_loss": 0.20282283425331116, "global_step": 44822, "epoch": 503, "lr": 9.354137206574583e-05}
{"train_loss": 0.3141120970249176, "global_step": 44823, "epoch": 503, "lr": 9.354108707078657e-05}
{"train_loss": 0.33913567662239075, "global_step": 44824, "epoch": 503, "lr": 9.354080206997374e-05}
{"train_loss": 0.3521081507205963, "global_step": 44825, "epoch": 503, "lr": 9.354051706330737e-05}
{"train_loss": 0.25679129362106323, "global_step": 44826, "epoch": 503, "lr": 9.354023205078754e-05}
{"train_loss": 0.3228890895843506, "global_step": 44827, "epoch": 503, "lr": 9.353994703241423e-05}
{"train_loss": 0.3606831133365631, "global_step": 44828, "epoch": 503, "lr": 9.353966200818751e-05}
{"train_loss": 0.3578592836856842, "global_step": 44829, "epoch": 503, "lr": 9.353937697810742e-05}
{"train_loss": 0.23092208802700043, "global_step": 44830, "epoch": 503, "lr": 9.353909194217398e-05}
{"train_loss": 0.3073730766773224, "global_step": 44831, "epoch": 503, "lr": 9.353880690038726e-05}
{"train_loss": 0.2712346613407135, "global_step": 44832, "epoch": 503, "lr": 9.353852185274725e-05}
{"train_loss": 0.2956140637397766, "global_step": 44833, "epoch": 503, "lr": 9.353823679925403e-05}
{"train_loss": 0.2929946184158325, "global_step": 44834, "epoch": 503, "lr": 9.353795173990762e-05}
{"train_loss": 0.20735591650009155, "global_step": 44835, "epoch": 503, "lr": 9.353766667470807e-05}
{"train_loss": 0.29921582341194153, "global_step": 44836, "epoch": 503, "lr": 9.353738160365542e-05}
{"train_loss": 0.3203204274177551, "global_step": 44837, "epoch": 503, "lr": 9.35370965267497e-05}
{"train_loss": 0.231421560049057, "global_step": 44838, "epoch": 503, "lr": 9.353681144399092e-05}
{"train_loss": 0.34398865699768066, "global_step": 44839, "epoch": 503, "lr": 9.353652635537916e-05}
{"train_loss": 0.21313677728176117, "global_step": 44840, "epoch": 503, "lr": 9.353624126091445e-05}
{"train_loss": 0.3733372688293457, "global_step": 44841, "epoch": 503, "lr": 9.353595616059683e-05}
{"train_loss": 0.17100048065185547, "global_step": 44842, "epoch": 503, "lr": 9.35356710544263e-05}
{"train_loss": 0.3128482699394226, "global_step": 44843, "epoch": 503, "lr": 9.353538594240296e-05}
{"train_loss": 0.27909931540489197, "global_step": 44844, "epoch": 503, "lr": 9.35351008245268e-05}
{"train_loss": 0.312141478061676, "global_step": 44845, "epoch": 503, "lr": 9.35348157007979e-05}
{"train_loss": 0.2539411783218384, "global_step": 44846, "epoch": 503, "lr": 9.353453057121626e-05}
{"train_loss": 0.24641193449497223, "global_step": 44847, "epoch": 503, "lr": 9.353424543578193e-05}
{"train_loss": 0.3445401191711426, "global_step": 44848, "epoch": 503, "lr": 9.353396029449496e-05}
{"train_loss": 0.31443384289741516, "global_step": 44849, "epoch": 503, "lr": 9.353367514735538e-05}
{"train_loss": 0.34511661529541016, "global_step": 44850, "epoch": 503, "lr": 9.353338999436322e-05}
{"train_loss": 0.38382917642593384, "global_step": 44851, "epoch": 503, "lr": 9.353310483551854e-05}
{"train_loss": 0.20708350837230682, "global_step": 44852, "epoch": 503, "lr": 9.353281967082134e-05}
{"train_loss": 0.20678897202014923, "global_step": 44853, "epoch": 503, "lr": 9.35325345002717e-05}
{"train_loss": 0.16982409358024597, "global_step": 44854, "epoch": 503, "lr": 9.353224932386964e-05}
{"train_loss": 0.2864896907900157, "global_step": 44855, "epoch": 503, "lr": 9.35319641416152e-05, "val_loss": 2.919102191925049}
{"train_loss": 0.4009697139263153, "global_step": 44856, "epoch": 504, "lr": 9.35316789535084e-05}
{"train_loss": 0.23302331566810608, "global_step": 44857, "epoch": 504, "lr": 9.353139375954932e-05}
{"train_loss": 0.2946338355541229, "global_step": 44858, "epoch": 504, "lr": 9.353110855973796e-05}
{"train_loss": 0.33557695150375366, "global_step": 44859, "epoch": 504, "lr": 9.35308233540744e-05}
{"train_loss": 0.24956807494163513, "global_step": 44860, "epoch": 504, "lr": 9.353053814255863e-05}
{"train_loss": 0.3047732412815094, "global_step": 44861, "epoch": 504, "lr": 9.353025292519071e-05}
{"train_loss": 0.2831454277038574, "global_step": 44862, "epoch": 504, "lr": 9.352996770197067e-05}
{"train_loss": 0.24893438816070557, "global_step": 44863, "epoch": 504, "lr": 9.352968247289857e-05}
{"train_loss": 0.30483871698379517, "global_step": 44864, "epoch": 504, "lr": 9.352939723797443e-05}
{"train_loss": 0.1949983388185501, "global_step": 44865, "epoch": 504, "lr": 9.352911199719829e-05}
{"train_loss": 0.20953969657421112, "global_step": 44866, "epoch": 504, "lr": 9.35288267505702e-05}
{"train_loss": 0.18520008027553558, "global_step": 44867, "epoch": 504, "lr": 9.352854149809019e-05}
{"train_loss": 0.28627827763557434, "global_step": 44868, "epoch": 504, "lr": 9.352825623975828e-05}
{"train_loss": 0.25551921129226685, "global_step": 44869, "epoch": 504, "lr": 9.352797097557453e-05}
{"train_loss": 0.43466007709503174, "global_step": 44870, "epoch": 504, "lr": 9.352768570553901e-05}
{"train_loss": 0.19562062621116638, "global_step": 44871, "epoch": 504, "lr": 9.352740042965169e-05}
{"train_loss": 0.31795212626457214, "global_step": 44872, "epoch": 504, "lr": 9.352711514791264e-05}
{"train_loss": 0.38215529918670654, "global_step": 44873, "epoch": 504, "lr": 9.352682986032191e-05}
{"train_loss": 0.1992933303117752, "global_step": 44874, "epoch": 504, "lr": 9.352654456687953e-05}
{"train_loss": 0.291603684425354, "global_step": 44875, "epoch": 504, "lr": 9.352625926758554e-05}
{"train_loss": 0.30339741706848145, "global_step": 44876, "epoch": 504, "lr": 9.352597396243996e-05}
{"train_loss": 0.31924206018447876, "global_step": 44877, "epoch": 504, "lr": 9.352568865144285e-05}
{"train_loss": 0.26036733388900757, "global_step": 44878, "epoch": 504, "lr": 9.352540333459424e-05}
{"train_loss": 0.2509104609489441, "global_step": 44879, "epoch": 504, "lr": 9.352511801189417e-05}
{"train_loss": 0.21416646242141724, "global_step": 44880, "epoch": 504, "lr": 9.35248326833427e-05}
{"train_loss": 0.20319367945194244, "global_step": 44881, "epoch": 504, "lr": 9.352454734893982e-05}
{"train_loss": 0.23887978494167328, "global_step": 44882, "epoch": 504, "lr": 9.35242620086856e-05}
{"train_loss": 0.30886274576187134, "global_step": 44883, "epoch": 504, "lr": 9.352397666258008e-05}
{"train_loss": 0.26204344630241394, "global_step": 44884, "epoch": 504, "lr": 9.352369131062328e-05}
{"train_loss": 0.274355947971344, "global_step": 44885, "epoch": 504, "lr": 9.352340595281526e-05}
{"train_loss": 0.37351182103157043, "global_step": 44886, "epoch": 504, "lr": 9.352312058915606e-05}
{"train_loss": 0.32798993587493896, "global_step": 44887, "epoch": 504, "lr": 9.352283521964569e-05}
{"train_loss": 0.26599615812301636, "global_step": 44888, "epoch": 504, "lr": 9.352254984428422e-05}
{"train_loss": 0.20167428255081177, "global_step": 44889, "epoch": 504, "lr": 9.352226446307166e-05}
{"train_loss": 0.24059394001960754, "global_step": 44890, "epoch": 504, "lr": 9.352197907600805e-05}
{"train_loss": 0.3123849630355835, "global_step": 44891, "epoch": 504, "lr": 9.352169368309346e-05}
{"train_loss": 0.21017520129680634, "global_step": 44892, "epoch": 504, "lr": 9.352140828432791e-05}
{"train_loss": 0.24535252153873444, "global_step": 44893, "epoch": 504, "lr": 9.352112287971143e-05}
{"train_loss": 0.20161771774291992, "global_step": 44894, "epoch": 504, "lr": 9.352083746924407e-05}
{"train_loss": 0.2423069179058075, "global_step": 44895, "epoch": 504, "lr": 9.352055205292585e-05}
{"train_loss": 0.26008322834968567, "global_step": 44896, "epoch": 504, "lr": 9.352026663075683e-05}
{"train_loss": 0.38796466588974, "global_step": 44897, "epoch": 504, "lr": 9.351998120273704e-05}
{"train_loss": 0.32727715373039246, "global_step": 44898, "epoch": 504, "lr": 9.351969576886654e-05}
{"train_loss": 0.24211561679840088, "global_step": 44899, "epoch": 504, "lr": 9.351941032914533e-05}
{"train_loss": 0.2972891926765442, "global_step": 44900, "epoch": 504, "lr": 9.351912488357346e-05}
{"train_loss": 0.24967865645885468, "global_step": 44901, "epoch": 504, "lr": 9.351883943215098e-05}
{"train_loss": 0.37540575861930847, "global_step": 44902, "epoch": 504, "lr": 9.351855397487791e-05}
{"train_loss": 0.3381742238998413, "global_step": 44903, "epoch": 504, "lr": 9.351826851175433e-05}
{"train_loss": 0.31533756852149963, "global_step": 44904, "epoch": 504, "lr": 9.351798304278023e-05}
{"train_loss": 0.3069482147693634, "global_step": 44905, "epoch": 504, "lr": 9.351769756795568e-05}
{"train_loss": 0.20279699563980103, "global_step": 44906, "epoch": 504, "lr": 9.351741208728069e-05}
{"train_loss": 0.279051274061203, "global_step": 44907, "epoch": 504, "lr": 9.351712660075532e-05}
{"train_loss": 0.35621941089630127, "global_step": 44908, "epoch": 504, "lr": 9.351684110837959e-05}
{"train_loss": 0.3798924386501312, "global_step": 44909, "epoch": 504, "lr": 9.351655561015357e-05}
{"train_loss": 0.29652103781700134, "global_step": 44910, "epoch": 504, "lr": 9.351627010607727e-05}
{"train_loss": 0.2832319736480713, "global_step": 44911, "epoch": 504, "lr": 9.351598459615075e-05}
{"train_loss": 0.3647380769252777, "global_step": 44912, "epoch": 504, "lr": 9.351569908037404e-05}
{"train_loss": 0.23248395323753357, "global_step": 44913, "epoch": 504, "lr": 9.351541355874715e-05}
{"train_loss": 0.22496278584003448, "global_step": 44914, "epoch": 504, "lr": 9.351512803127014e-05}
{"train_loss": 0.2680509686470032, "global_step": 44915, "epoch": 504, "lr": 9.351484249794307e-05}
{"train_loss": 0.31009653210639954, "global_step": 44916, "epoch": 504, "lr": 9.351455695876596e-05}
{"train_loss": 0.3489918112754822, "global_step": 44917, "epoch": 504, "lr": 9.351427141373885e-05}
{"train_loss": 0.33850571513175964, "global_step": 44918, "epoch": 504, "lr": 9.351398586286178e-05}
{"train_loss": 0.3611029088497162, "global_step": 44919, "epoch": 504, "lr": 9.351370030613476e-05}
{"train_loss": 0.2146085500717163, "global_step": 44920, "epoch": 504, "lr": 9.351341474355788e-05}
{"train_loss": 0.3304373025894165, "global_step": 44921, "epoch": 504, "lr": 9.351312917513113e-05}
{"train_loss": 0.23802773654460907, "global_step": 44922, "epoch": 504, "lr": 9.35128436008546e-05}
{"train_loss": 0.25567394495010376, "global_step": 44923, "epoch": 504, "lr": 9.351255802072827e-05}
{"train_loss": 0.3418925404548645, "global_step": 44924, "epoch": 504, "lr": 9.351227243475223e-05}
{"train_loss": 0.29314732551574707, "global_step": 44925, "epoch": 504, "lr": 9.351198684292647e-05}
{"train_loss": 0.2669713497161865, "global_step": 44926, "epoch": 504, "lr": 9.351170124525106e-05}
{"train_loss": 0.2428535372018814, "global_step": 44927, "epoch": 504, "lr": 9.351141564172604e-05}
{"train_loss": 0.3327431082725525, "global_step": 44928, "epoch": 504, "lr": 9.351113003235143e-05}
{"train_loss": 0.23705442249774933, "global_step": 44929, "epoch": 504, "lr": 9.351084441712729e-05}
{"train_loss": 0.3030155301094055, "global_step": 44930, "epoch": 504, "lr": 9.351055879605365e-05}
{"train_loss": 0.35552823543548584, "global_step": 44931, "epoch": 504, "lr": 9.351027316913055e-05}
{"train_loss": 0.26315778493881226, "global_step": 44932, "epoch": 504, "lr": 9.3509987536358e-05}
{"train_loss": 0.2988147437572479, "global_step": 44933, "epoch": 504, "lr": 9.350970189773609e-05}
{"train_loss": 0.38668811321258545, "global_step": 44934, "epoch": 504, "lr": 9.350941625326481e-05}
{"train_loss": 0.37760135531425476, "global_step": 44935, "epoch": 504, "lr": 9.350913060294423e-05}
{"train_loss": 0.4452168345451355, "global_step": 44936, "epoch": 504, "lr": 9.350884494677437e-05}
{"train_loss": 0.4412061870098114, "global_step": 44937, "epoch": 504, "lr": 9.350855928475528e-05}
{"train_loss": 0.3423275351524353, "global_step": 44938, "epoch": 504, "lr": 9.3508273616887e-05}
{"train_loss": 0.29005172848701477, "global_step": 44939, "epoch": 504, "lr": 9.350798794316954e-05}
{"train_loss": 0.28475600481033325, "global_step": 44940, "epoch": 504, "lr": 9.3507702263603e-05}
{"train_loss": 0.26268482208251953, "global_step": 44941, "epoch": 504, "lr": 9.350741657818736e-05}
{"train_loss": 0.30763235688209534, "global_step": 44942, "epoch": 504, "lr": 9.350713088692268e-05}
{"train_loss": 0.2787514925003052, "global_step": 44943, "epoch": 504, "lr": 9.350684518980899e-05}
{"train_loss": 0.29231599157445887, "global_step": 44944, "epoch": 504, "lr": 9.350655948684635e-05, "val_loss": 2.8352134227752686}
{"train_loss": 0.31841444969177246, "global_step": 44945, "epoch": 505, "lr": 9.350627377803476e-05}
{"train_loss": 0.3046019673347473, "global_step": 44946, "epoch": 505, "lr": 9.350598806337429e-05}
{"train_loss": 0.2918183207511902, "global_step": 44947, "epoch": 505, "lr": 9.350570234286497e-05}
{"train_loss": 0.24275588989257812, "global_step": 44948, "epoch": 505, "lr": 9.350541661650684e-05}
{"train_loss": 0.31062546372413635, "global_step": 44949, "epoch": 505, "lr": 9.350513088429995e-05}
{"train_loss": 0.3469787836074829, "global_step": 44950, "epoch": 505, "lr": 9.350484514624431e-05}
{"train_loss": 0.4351882338523865, "global_step": 44951, "epoch": 505, "lr": 9.350455940233998e-05}
{"train_loss": 0.46123260259628296, "global_step": 44952, "epoch": 505, "lr": 9.3504273652587e-05}
{"train_loss": 0.37460896372795105, "global_step": 44953, "epoch": 505, "lr": 9.35039878969854e-05}
{"train_loss": 0.28773754835128784, "global_step": 44954, "epoch": 505, "lr": 9.35037021355352e-05}
{"train_loss": 0.2731691896915436, "global_step": 44955, "epoch": 505, "lr": 9.350341636823647e-05}
{"train_loss": 0.27370205521583557, "global_step": 44956, "epoch": 505, "lr": 9.350313059508924e-05}
{"train_loss": 0.2955245077610016, "global_step": 44957, "epoch": 505, "lr": 9.350284481609353e-05}
{"train_loss": 0.2900243401527405, "global_step": 44958, "epoch": 505, "lr": 9.35025590312494e-05}
{"train_loss": 0.3254990577697754, "global_step": 44959, "epoch": 505, "lr": 9.350227324055688e-05}
{"train_loss": 0.3559713065624237, "global_step": 44960, "epoch": 505, "lr": 9.350198744401601e-05}
{"train_loss": 0.2765829861164093, "global_step": 44961, "epoch": 505, "lr": 9.350170164162684e-05}
{"train_loss": 0.30847644805908203, "global_step": 44962, "epoch": 505, "lr": 9.350141583338938e-05}
{"train_loss": 0.43821126222610474, "global_step": 44963, "epoch": 505, "lr": 9.35011300193037e-05}
{"train_loss": 0.33637309074401855, "global_step": 44964, "epoch": 505, "lr": 9.35008441993698e-05}
{"train_loss": 0.2795726954936981, "global_step": 44965, "epoch": 505, "lr": 9.350055837358776e-05}
{"train_loss": 0.22618895769119263, "global_step": 44966, "epoch": 505, "lr": 9.35002725419576e-05}
{"train_loss": 0.2731762230396271, "global_step": 44967, "epoch": 505, "lr": 9.349998670447936e-05}
{"train_loss": 0.32995590567588806, "global_step": 44968, "epoch": 505, "lr": 9.349970086115307e-05}
{"train_loss": 0.4114192724227905, "global_step": 44969, "epoch": 505, "lr": 9.349941501197877e-05}
{"train_loss": 0.3313600718975067, "global_step": 44970, "epoch": 505, "lr": 9.349912915695653e-05}
{"train_loss": 0.25334951281547546, "global_step": 44971, "epoch": 505, "lr": 9.349884329608634e-05}
{"train_loss": 0.29876992106437683, "global_step": 44972, "epoch": 505, "lr": 9.349855742936827e-05}
{"train_loss": 0.23863951861858368, "global_step": 44973, "epoch": 505, "lr": 9.349827155680233e-05}
{"train_loss": 0.32953518629074097, "global_step": 44974, "epoch": 505, "lr": 9.349798567838859e-05}
{"train_loss": 0.3540264964103699, "global_step": 44975, "epoch": 505, "lr": 9.349769979412708e-05}
{"train_loss": 0.2692021131515503, "global_step": 44976, "epoch": 505, "lr": 9.349741390401783e-05}
{"train_loss": 0.2784414291381836, "global_step": 44977, "epoch": 505, "lr": 9.349712800806087e-05}
{"train_loss": 0.291826993227005, "global_step": 44978, "epoch": 505, "lr": 9.349684210625628e-05}
{"train_loss": 0.3643297255039215, "global_step": 44979, "epoch": 505, "lr": 9.349655619860405e-05}
{"train_loss": 0.26390889286994934, "global_step": 44980, "epoch": 505, "lr": 9.349627028510425e-05}
{"train_loss": 0.3204585909843445, "global_step": 44981, "epoch": 505, "lr": 9.34959843657569e-05}
{"train_loss": 0.4051561951637268, "global_step": 44982, "epoch": 505, "lr": 9.349569844056204e-05}
{"train_loss": 0.3728052079677582, "global_step": 44983, "epoch": 505, "lr": 9.349541250951972e-05}
{"train_loss": 0.334073007106781, "global_step": 44984, "epoch": 505, "lr": 9.349512657262998e-05}
{"train_loss": 0.30972954630851746, "global_step": 44985, "epoch": 505, "lr": 9.349484062989283e-05}
{"train_loss": 0.2852638363838196, "global_step": 44986, "epoch": 505, "lr": 9.349455468130834e-05}
{"train_loss": 0.35579773783683777, "global_step": 44987, "epoch": 505, "lr": 9.349426872687655e-05}
{"train_loss": 0.22050483524799347, "global_step": 44988, "epoch": 505, "lr": 9.349398276659747e-05}
{"train_loss": 0.2897631824016571, "global_step": 44989, "epoch": 505, "lr": 9.349369680047116e-05}
{"train_loss": 0.2839904725551605, "global_step": 44990, "epoch": 505, "lr": 9.349341082849764e-05}
{"train_loss": 0.35353684425354004, "global_step": 44991, "epoch": 505, "lr": 9.349312485067696e-05}
{"train_loss": 0.2799966037273407, "global_step": 44992, "epoch": 505, "lr": 9.349283886700919e-05}
{"train_loss": 0.26998698711395264, "global_step": 44993, "epoch": 505, "lr": 9.34925528774943e-05}
{"train_loss": 0.25897273421287537, "global_step": 44994, "epoch": 505, "lr": 9.349226688213239e-05}
{"train_loss": 0.25705933570861816, "global_step": 44995, "epoch": 505, "lr": 9.349198088092347e-05}
{"train_loss": 0.3618152141571045, "global_step": 44996, "epoch": 505, "lr": 9.349169487386757e-05}
{"train_loss": 0.28669100999832153, "global_step": 44997, "epoch": 505, "lr": 9.349140886096477e-05}
{"train_loss": 0.3528822660446167, "global_step": 44998, "epoch": 505, "lr": 9.349112284221507e-05}
{"train_loss": 0.22642965614795685, "global_step": 44999, "epoch": 505, "lr": 9.34908368176185e-05}
{"train_loss": 0.31390655040740967, "global_step": 45000, "epoch": 505, "lr": 9.349055078717512e-05}
{"train_loss": 0.4368554949760437, "global_step": 45001, "epoch": 505, "lr": 9.349026475088498e-05}
{"train_loss": 0.26020121574401855, "global_step": 45002, "epoch": 505, "lr": 9.34899787087481e-05}
{"train_loss": 0.26861658692359924, "global_step": 45003, "epoch": 505, "lr": 9.348969266076453e-05}
{"train_loss": 0.28058817982673645, "global_step": 45004, "epoch": 505, "lr": 9.348940660693428e-05}
{"train_loss": 0.21921281516551971, "global_step": 45005, "epoch": 505, "lr": 9.348912054725742e-05}
{"train_loss": 0.417741060256958, "global_step": 45006, "epoch": 505, "lr": 9.3488834481734e-05}
{"train_loss": 0.23960213363170624, "global_step": 45007, "epoch": 505, "lr": 9.3488548410364e-05}
{"train_loss": 0.35660406947135925, "global_step": 45008, "epoch": 505, "lr": 9.348826233314751e-05}
{"train_loss": 0.3501625955104828, "global_step": 45009, "epoch": 505, "lr": 9.348797625008456e-05}
{"train_loss": 0.26747918128967285, "global_step": 45010, "epoch": 505, "lr": 9.348769016117518e-05}
{"train_loss": 0.2763464152812958, "global_step": 45011, "epoch": 505, "lr": 9.348740406641941e-05}
{"train_loss": 0.238093301653862, "global_step": 45012, "epoch": 505, "lr": 9.348711796581728e-05}
{"train_loss": 0.3118738532066345, "global_step": 45013, "epoch": 505, "lr": 9.348683185936884e-05}
{"train_loss": 0.24972806870937347, "global_step": 45014, "epoch": 505, "lr": 9.348654574707411e-05}
{"train_loss": 0.2824327051639557, "global_step": 45015, "epoch": 505, "lr": 9.348625962893319e-05}
{"train_loss": 0.3395971357822418, "global_step": 45016, "epoch": 505, "lr": 9.348597350494604e-05}
{"train_loss": 0.3036438226699829, "global_step": 45017, "epoch": 505, "lr": 9.348568737511272e-05}
{"train_loss": 0.32321009039878845, "global_step": 45018, "epoch": 505, "lr": 9.34854012394333e-05}
{"train_loss": 0.2489694207906723, "global_step": 45019, "epoch": 505, "lr": 9.348511509790778e-05}
{"train_loss": 0.275076687335968, "global_step": 45020, "epoch": 505, "lr": 9.348482895053623e-05}
{"train_loss": 0.2972396910190582, "global_step": 45021, "epoch": 505, "lr": 9.348454279731868e-05}
{"train_loss": 0.28158271312713623, "global_step": 45022, "epoch": 505, "lr": 9.348425663825515e-05}
{"train_loss": 0.2892969846725464, "global_step": 45023, "epoch": 505, "lr": 9.34839704733457e-05}
{"train_loss": 0.2863491177558899, "global_step": 45024, "epoch": 505, "lr": 9.348368430259035e-05}
{"train_loss": 0.2816314995288849, "global_step": 45025, "epoch": 505, "lr": 9.348339812598917e-05}
{"train_loss": 0.3661339581012726, "global_step": 45026, "epoch": 505, "lr": 9.348311194354216e-05}
{"train_loss": 0.3366980254650116, "global_step": 45027, "epoch": 505, "lr": 9.348282575524939e-05}
{"train_loss": 0.3127295970916748, "global_step": 45028, "epoch": 505, "lr": 9.348253956111086e-05}
{"train_loss": 0.37381622195243835, "global_step": 45029, "epoch": 505, "lr": 9.348225336112663e-05}
{"train_loss": 0.22709186375141144, "global_step": 45030, "epoch": 505, "lr": 9.348196715529676e-05}
{"train_loss": 0.2625763416290283, "global_step": 45031, "epoch": 505, "lr": 9.348168094362128e-05}
{"train_loss": 0.27000364661216736, "global_step": 45032, "epoch": 505, "lr": 9.34813947261002e-05}
{"train_loss": 0.3062071641174595, "global_step": 45033, "epoch": 505, "lr": 9.348110850273357e-05, "val_loss": 2.910419225692749, "train_action_mse_error": 17.24040985107422}
{"train_loss": 0.244510218501091, "global_step": 45034, "epoch": 506, "lr": 9.348082227352145e-05}
{"train_loss": 0.3022359311580658, "global_step": 45035, "epoch": 506, "lr": 9.348053603846387e-05}
{"train_loss": 0.2754229009151459, "global_step": 45036, "epoch": 506, "lr": 9.348024979756084e-05}
{"train_loss": 0.30919498205184937, "global_step": 45037, "epoch": 506, "lr": 9.347996355081244e-05}
{"train_loss": 0.3317459523677826, "global_step": 45038, "epoch": 506, "lr": 9.347967729821867e-05}
{"train_loss": 0.3753967583179474, "global_step": 45039, "epoch": 506, "lr": 9.34793910397796e-05}
{"train_loss": 0.3460124135017395, "global_step": 45040, "epoch": 506, "lr": 9.347910477549525e-05}
{"train_loss": 0.320631206035614, "global_step": 45041, "epoch": 506, "lr": 9.347881850536566e-05}
{"train_loss": 0.2453659176826477, "global_step": 45042, "epoch": 506, "lr": 9.347853222939089e-05}
{"train_loss": 0.25304901599884033, "global_step": 45043, "epoch": 506, "lr": 9.347824594757096e-05}
{"train_loss": 0.29503172636032104, "global_step": 45044, "epoch": 506, "lr": 9.347795965990589e-05}
{"train_loss": 0.3846416473388672, "global_step": 45045, "epoch": 506, "lr": 9.347767336639575e-05}
{"train_loss": 0.2900681793689728, "global_step": 45046, "epoch": 506, "lr": 9.347738706704057e-05}
{"train_loss": 0.22190749645233154, "global_step": 45047, "epoch": 506, "lr": 9.347710076184037e-05}
{"train_loss": 0.2302631139755249, "global_step": 45048, "epoch": 506, "lr": 9.347681445079521e-05}
{"train_loss": 0.24152088165283203, "global_step": 45049, "epoch": 506, "lr": 9.347652813390512e-05}
{"train_loss": 0.2653392553329468, "global_step": 45050, "epoch": 506, "lr": 9.347624181117017e-05}
{"train_loss": 0.39451298117637634, "global_step": 45051, "epoch": 506, "lr": 9.347595548259034e-05}
{"train_loss": 0.2791637182235718, "global_step": 45052, "epoch": 506, "lr": 9.34756691481657e-05}
{"train_loss": 0.2466445416212082, "global_step": 45053, "epoch": 506, "lr": 9.34753828078963e-05}
{"train_loss": 0.21499022841453552, "global_step": 45054, "epoch": 506, "lr": 9.347509646178215e-05}
{"train_loss": 0.23858153820037842, "global_step": 45055, "epoch": 506, "lr": 9.34748101098233e-05}
{"train_loss": 0.3502424657344818, "global_step": 45056, "epoch": 506, "lr": 9.34745237520198e-05}
{"train_loss": 0.29825422167778015, "global_step": 45057, "epoch": 506, "lr": 9.347423738837169e-05}
{"train_loss": 0.3896947205066681, "global_step": 45058, "epoch": 506, "lr": 9.3473951018879e-05}
{"train_loss": 0.20903104543685913, "global_step": 45059, "epoch": 506, "lr": 9.347366464354175e-05}
{"train_loss": 0.24284955859184265, "global_step": 45060, "epoch": 506, "lr": 9.347337826236e-05}
{"train_loss": 0.31991201639175415, "global_step": 45061, "epoch": 506, "lr": 9.347309187533377e-05}
{"train_loss": 0.4053077697753906, "global_step": 45062, "epoch": 506, "lr": 9.347280548246313e-05}
{"train_loss": 0.24279361963272095, "global_step": 45063, "epoch": 506, "lr": 9.347251908374811e-05}
{"train_loss": 0.2835524082183838, "global_step": 45064, "epoch": 506, "lr": 9.347223267918872e-05}
{"train_loss": 0.2850341796875, "global_step": 45065, "epoch": 506, "lr": 9.347194626878504e-05}
{"train_loss": 0.2108626365661621, "global_step": 45066, "epoch": 506, "lr": 9.347165985253707e-05}
{"train_loss": 0.34820806980133057, "global_step": 45067, "epoch": 506, "lr": 9.347137343044488e-05}
{"train_loss": 0.2912348210811615, "global_step": 45068, "epoch": 506, "lr": 9.347108700250847e-05}
{"train_loss": 0.2700909674167633, "global_step": 45069, "epoch": 506, "lr": 9.347080056872793e-05}
{"train_loss": 0.34535881876945496, "global_step": 45070, "epoch": 506, "lr": 9.347051412910323e-05}
{"train_loss": 0.3718448579311371, "global_step": 45071, "epoch": 506, "lr": 9.347022768363449e-05}
{"train_loss": 0.18542978167533875, "global_step": 45072, "epoch": 506, "lr": 9.346994123232169e-05}
{"train_loss": 0.3664076030254364, "global_step": 45073, "epoch": 506, "lr": 9.34696547751649e-05}
{"train_loss": 0.35035374760627747, "global_step": 45074, "epoch": 506, "lr": 9.346936831216413e-05}
{"train_loss": 0.2821686565876007, "global_step": 45075, "epoch": 506, "lr": 9.346908184331944e-05}
{"train_loss": 0.35513758659362793, "global_step": 45076, "epoch": 506, "lr": 9.346879536863084e-05}
{"train_loss": 0.3213278353214264, "global_step": 45077, "epoch": 506, "lr": 9.346850888809842e-05}
{"train_loss": 0.28125378489494324, "global_step": 45078, "epoch": 506, "lr": 9.346822240172217e-05}
{"train_loss": 0.25836631655693054, "global_step": 45079, "epoch": 506, "lr": 9.346793590950215e-05}
{"train_loss": 0.2973703444004059, "global_step": 45080, "epoch": 506, "lr": 9.346764941143842e-05}
{"train_loss": 0.22440670430660248, "global_step": 45081, "epoch": 506, "lr": 9.346736290753098e-05}
{"train_loss": 0.31117483973503113, "global_step": 45082, "epoch": 506, "lr": 9.346707639777986e-05}
{"train_loss": 0.2981249988079071, "global_step": 45083, "epoch": 506, "lr": 9.346678988218515e-05}
{"train_loss": 0.19760902225971222, "global_step": 45084, "epoch": 506, "lr": 9.346650336074684e-05}
{"train_loss": 0.32791730761528015, "global_step": 45085, "epoch": 506, "lr": 9.3466216833465e-05}
{"train_loss": 0.2756192088127136, "global_step": 45086, "epoch": 506, "lr": 9.346593030033965e-05}
{"train_loss": 0.24135762453079224, "global_step": 45087, "epoch": 506, "lr": 9.346564376137084e-05}
{"train_loss": 0.2372225672006607, "global_step": 45088, "epoch": 506, "lr": 9.34653572165586e-05}
{"train_loss": 0.31207025051116943, "global_step": 45089, "epoch": 506, "lr": 9.3465070665903e-05}
{"train_loss": 0.33966752886772156, "global_step": 45090, "epoch": 506, "lr": 9.346478410940402e-05}
{"train_loss": 0.29282376170158386, "global_step": 45091, "epoch": 506, "lr": 9.346449754706172e-05}
{"train_loss": 0.39224773645401, "global_step": 45092, "epoch": 506, "lr": 9.346421097887617e-05}
{"train_loss": 0.3131735920906067, "global_step": 45093, "epoch": 506, "lr": 9.34639244048474e-05}
{"train_loss": 0.3731590211391449, "global_step": 45094, "epoch": 506, "lr": 9.346363782497541e-05}
{"train_loss": 0.27449801564216614, "global_step": 45095, "epoch": 506, "lr": 9.346335123926026e-05}
{"train_loss": 0.3403126299381256, "global_step": 45096, "epoch": 506, "lr": 9.346306464770201e-05}
{"train_loss": 0.29340943694114685, "global_step": 45097, "epoch": 506, "lr": 9.346277805030068e-05}
{"train_loss": 0.27138257026672363, "global_step": 45098, "epoch": 506, "lr": 9.34624914470563e-05}
{"train_loss": 0.28810590505599976, "global_step": 45099, "epoch": 506, "lr": 9.346220483796893e-05}
{"train_loss": 0.26874881982803345, "global_step": 45100, "epoch": 506, "lr": 9.346191822303858e-05}
{"train_loss": 0.23555083572864532, "global_step": 45101, "epoch": 506, "lr": 9.346163160226531e-05}
{"train_loss": 0.19228121638298035, "global_step": 45102, "epoch": 506, "lr": 9.346134497564915e-05}
{"train_loss": 0.23725904524326324, "global_step": 45103, "epoch": 506, "lr": 9.346105834319015e-05}
{"train_loss": 0.29524677991867065, "global_step": 45104, "epoch": 506, "lr": 9.346077170488834e-05}
{"train_loss": 0.3332657516002655, "global_step": 45105, "epoch": 506, "lr": 9.346048506074376e-05}
{"train_loss": 0.374835729598999, "global_step": 45106, "epoch": 506, "lr": 9.346019841075644e-05}
{"train_loss": 0.2745824158191681, "global_step": 45107, "epoch": 506, "lr": 9.345991175492643e-05}
{"train_loss": 0.2016013115644455, "global_step": 45108, "epoch": 506, "lr": 9.345962509325377e-05}
{"train_loss": 0.210547536611557, "global_step": 45109, "epoch": 506, "lr": 9.345933842573847e-05}
{"train_loss": 0.210677370429039, "global_step": 45110, "epoch": 506, "lr": 9.345905175238061e-05}
{"train_loss": 0.2446914166212082, "global_step": 45111, "epoch": 506, "lr": 9.345876507318021e-05}
{"train_loss": 0.21697980165481567, "global_step": 45112, "epoch": 506, "lr": 9.345847838813732e-05}
{"train_loss": 0.23929227888584137, "global_step": 45113, "epoch": 506, "lr": 9.345819169725196e-05}
{"train_loss": 0.3199458718299866, "global_step": 45114, "epoch": 506, "lr": 9.345790500052416e-05}
{"train_loss": 0.3338641822338104, "global_step": 45115, "epoch": 506, "lr": 9.345761829795398e-05}
{"train_loss": 0.25551697611808777, "global_step": 45116, "epoch": 506, "lr": 9.345733158954147e-05}
{"train_loss": 0.3414958715438843, "global_step": 45117, "epoch": 506, "lr": 9.345704487528664e-05}
{"train_loss": 0.33098530769348145, "global_step": 45118, "epoch": 506, "lr": 9.345675815518953e-05}
{"train_loss": 0.38879215717315674, "global_step": 45119, "epoch": 506, "lr": 9.34564714292502e-05}
{"train_loss": 0.2574472427368164, "global_step": 45120, "epoch": 506, "lr": 9.345618469746867e-05}
{"train_loss": 0.2567095160484314, "global_step": 45121, "epoch": 506, "lr": 9.3455897959845e-05}
{"train_loss": 0.2902075107512849, "global_step": 45122, "epoch": 506, "lr": 9.34556112163792e-05, "val_loss": 2.9079291820526123}
{"train_loss": 0.19278278946876526, "global_step": 45123, "epoch": 507, "lr": 9.345532446707133e-05}
{"train_loss": 0.2650306820869446, "global_step": 45124, "epoch": 507, "lr": 9.345503771192143e-05}
{"train_loss": 0.2675589919090271, "global_step": 45125, "epoch": 507, "lr": 9.345475095092951e-05}
{"train_loss": 0.27469590306282043, "global_step": 45126, "epoch": 507, "lr": 9.345446418409564e-05}
{"train_loss": 0.22445957362651825, "global_step": 45127, "epoch": 507, "lr": 9.345417741141984e-05}
{"train_loss": 0.24836724996566772, "global_step": 45128, "epoch": 507, "lr": 9.345389063290215e-05}
{"train_loss": 0.3319009244441986, "global_step": 45129, "epoch": 507, "lr": 9.345360384854262e-05}
{"train_loss": 0.3455502688884735, "global_step": 45130, "epoch": 507, "lr": 9.345331705834131e-05}
{"train_loss": 0.2604597508907318, "global_step": 45131, "epoch": 507, "lr": 9.345303026229819e-05}
{"train_loss": 0.20784197747707367, "global_step": 45132, "epoch": 507, "lr": 9.345274346041338e-05}
{"train_loss": 0.2861242890357971, "global_step": 45133, "epoch": 507, "lr": 9.345245665268685e-05}
{"train_loss": 0.26126110553741455, "global_step": 45134, "epoch": 507, "lr": 9.345216983911867e-05}
{"train_loss": 0.27381423115730286, "global_step": 45135, "epoch": 507, "lr": 9.345188301970886e-05}
{"train_loss": 0.30550381541252136, "global_step": 45136, "epoch": 507, "lr": 9.34515961944575e-05}
{"train_loss": 0.2920002341270447, "global_step": 45137, "epoch": 507, "lr": 9.34513093633646e-05}
{"train_loss": 0.3307848870754242, "global_step": 45138, "epoch": 507, "lr": 9.345102252643018e-05}
{"train_loss": 0.3308869004249573, "global_step": 45139, "epoch": 507, "lr": 9.345073568365432e-05}
{"train_loss": 0.34783607721328735, "global_step": 45140, "epoch": 507, "lr": 9.345044883503704e-05}
{"train_loss": 0.3413107693195343, "global_step": 45141, "epoch": 507, "lr": 9.345016198057836e-05}
{"train_loss": 0.2788065969944, "global_step": 45142, "epoch": 507, "lr": 9.344987512027836e-05}
{"train_loss": 0.16240835189819336, "global_step": 45143, "epoch": 507, "lr": 9.344958825413703e-05}
{"train_loss": 0.38117077946662903, "global_step": 45144, "epoch": 507, "lr": 9.344930138215444e-05}
{"train_loss": 0.2037612497806549, "global_step": 45145, "epoch": 507, "lr": 9.344901450433062e-05}
{"train_loss": 0.28683462738990784, "global_step": 45146, "epoch": 507, "lr": 9.344872762066561e-05}
{"train_loss": 0.24651791155338287, "global_step": 45147, "epoch": 507, "lr": 9.344844073115946e-05}
{"train_loss": 0.30956053733825684, "global_step": 45148, "epoch": 507, "lr": 9.344815383581218e-05}
{"train_loss": 0.21312014758586884, "global_step": 45149, "epoch": 507, "lr": 9.344786693462384e-05}
{"train_loss": 0.3218834698200226, "global_step": 45150, "epoch": 507, "lr": 9.344758002759445e-05}
{"train_loss": 0.20491500198841095, "global_step": 45151, "epoch": 507, "lr": 9.344729311472406e-05}
{"train_loss": 0.22004511952400208, "global_step": 45152, "epoch": 507, "lr": 9.344700619601273e-05}
{"train_loss": 0.22328145802021027, "global_step": 45153, "epoch": 507, "lr": 9.344671927146047e-05}
{"train_loss": 0.25918737053871155, "global_step": 45154, "epoch": 507, "lr": 9.344643234106732e-05}
{"train_loss": 0.249787375330925, "global_step": 45155, "epoch": 507, "lr": 9.344614540483334e-05}
{"train_loss": 0.2633631229400635, "global_step": 45156, "epoch": 507, "lr": 9.344585846275855e-05}
{"train_loss": 0.3320131301879883, "global_step": 45157, "epoch": 507, "lr": 9.344557151484299e-05}
{"train_loss": 0.2785149812698364, "global_step": 45158, "epoch": 507, "lr": 9.34452845610867e-05}
{"train_loss": 0.21905264258384705, "global_step": 45159, "epoch": 507, "lr": 9.344499760148972e-05}
{"train_loss": 0.21951107680797577, "global_step": 45160, "epoch": 507, "lr": 9.344471063605211e-05}
{"train_loss": 0.2656440734863281, "global_step": 45161, "epoch": 507, "lr": 9.344442366477387e-05}
{"train_loss": 0.2808750569820404, "global_step": 45162, "epoch": 507, "lr": 9.344413668765506e-05}
{"train_loss": 0.2192697823047638, "global_step": 45163, "epoch": 507, "lr": 9.344384970469572e-05}
{"train_loss": 0.25153225660324097, "global_step": 45164, "epoch": 507, "lr": 9.344356271589587e-05}
{"train_loss": 0.3069397807121277, "global_step": 45165, "epoch": 507, "lr": 9.344327572125558e-05}
{"train_loss": 0.26233112812042236, "global_step": 45166, "epoch": 507, "lr": 9.344298872077485e-05}
{"train_loss": 0.33406713604927063, "global_step": 45167, "epoch": 507, "lr": 9.344270171445377e-05}
{"train_loss": 0.30416810512542725, "global_step": 45168, "epoch": 507, "lr": 9.344241470229234e-05}
{"train_loss": 0.23985759913921356, "global_step": 45169, "epoch": 507, "lr": 9.34421276842906e-05}
{"train_loss": 0.31465503573417664, "global_step": 45170, "epoch": 507, "lr": 9.344184066044859e-05}
{"train_loss": 0.21676036715507507, "global_step": 45171, "epoch": 507, "lr": 9.344155363076634e-05}
{"train_loss": 0.2416398525238037, "global_step": 45172, "epoch": 507, "lr": 9.344126659524393e-05}
{"train_loss": 0.21827605366706848, "global_step": 45173, "epoch": 507, "lr": 9.344097955388135e-05}
{"train_loss": 0.30898910760879517, "global_step": 45174, "epoch": 507, "lr": 9.344069250667868e-05}
{"train_loss": 0.33625704050064087, "global_step": 45175, "epoch": 507, "lr": 9.344040545363593e-05}
{"train_loss": 0.4275428354740143, "global_step": 45176, "epoch": 507, "lr": 9.344011839475315e-05}
{"train_loss": 0.34355536103248596, "global_step": 45177, "epoch": 507, "lr": 9.343983133003037e-05}
{"train_loss": 0.23778265714645386, "global_step": 45178, "epoch": 507, "lr": 9.343954425946764e-05}
{"train_loss": 0.2598516047000885, "global_step": 45179, "epoch": 507, "lr": 9.343925718306499e-05}
{"train_loss": 0.3238762617111206, "global_step": 45180, "epoch": 507, "lr": 9.343897010082245e-05}
{"train_loss": 0.2457788586616516, "global_step": 45181, "epoch": 507, "lr": 9.343868301274008e-05}
{"train_loss": 0.3297971487045288, "global_step": 45182, "epoch": 507, "lr": 9.343839591881791e-05}
{"train_loss": 0.27565333247184753, "global_step": 45183, "epoch": 507, "lr": 9.343810881905598e-05}
{"train_loss": 0.27965474128723145, "global_step": 45184, "epoch": 507, "lr": 9.343782171345433e-05}
{"train_loss": 0.296863317489624, "global_step": 45185, "epoch": 507, "lr": 9.343753460201297e-05}
{"train_loss": 0.23083874583244324, "global_step": 45186, "epoch": 507, "lr": 9.343724748473199e-05}
{"train_loss": 0.28183460235595703, "global_step": 45187, "epoch": 507, "lr": 9.343696036161138e-05}
{"train_loss": 0.2744830250740051, "global_step": 45188, "epoch": 507, "lr": 9.343667323265122e-05}
{"train_loss": 0.33305272459983826, "global_step": 45189, "epoch": 507, "lr": 9.343638609785152e-05}
{"train_loss": 0.32773077487945557, "global_step": 45190, "epoch": 507, "lr": 9.343609895721233e-05}
{"train_loss": 0.3798089027404785, "global_step": 45191, "epoch": 507, "lr": 9.343581181073367e-05}
{"train_loss": 0.2743361294269562, "global_step": 45192, "epoch": 507, "lr": 9.34355246584156e-05}
{"train_loss": 0.21906507015228271, "global_step": 45193, "epoch": 507, "lr": 9.343523750025818e-05}
{"train_loss": 0.29672759771347046, "global_step": 45194, "epoch": 507, "lr": 9.34349503362614e-05}
{"train_loss": 0.25907102227211, "global_step": 45195, "epoch": 507, "lr": 9.343466316642531e-05}
{"train_loss": 0.29247385263442993, "global_step": 45196, "epoch": 507, "lr": 9.343437599074998e-05}
{"train_loss": 0.3037853538990021, "global_step": 45197, "epoch": 507, "lr": 9.343408880923541e-05}
{"train_loss": 0.3528072237968445, "global_step": 45198, "epoch": 507, "lr": 9.343380162188167e-05}
{"train_loss": 0.3294113576412201, "global_step": 45199, "epoch": 507, "lr": 9.343351442868877e-05}
{"train_loss": 0.3178698718547821, "global_step": 45200, "epoch": 507, "lr": 9.343322722965677e-05}
{"train_loss": 0.25984257459640503, "global_step": 45201, "epoch": 507, "lr": 9.343294002478571e-05}
{"train_loss": 0.326010137796402, "global_step": 45202, "epoch": 507, "lr": 9.343265281407562e-05}
{"train_loss": 0.2549099922180176, "global_step": 45203, "epoch": 507, "lr": 9.343236559752653e-05}
{"train_loss": 0.3881015479564667, "global_step": 45204, "epoch": 507, "lr": 9.34320783751385e-05}
{"train_loss": 0.3687911033630371, "global_step": 45205, "epoch": 507, "lr": 9.343179114691154e-05}
{"train_loss": 0.2864748239517212, "global_step": 45206, "epoch": 507, "lr": 9.343150391284572e-05}
{"train_loss": 0.2696489691734314, "global_step": 45207, "epoch": 507, "lr": 9.343121667294106e-05}
{"train_loss": 0.24542701244354248, "global_step": 45208, "epoch": 507, "lr": 9.34309294271976e-05}
{"train_loss": 0.3938078284263611, "global_step": 45209, "epoch": 507, "lr": 9.343064217561537e-05}
{"train_loss": 0.21207085251808167, "global_step": 45210, "epoch": 507, "lr": 9.343035491819443e-05}
{"train_loss": 0.28395429270321065, "global_step": 45211, "epoch": 507, "lr": 9.343006765493481e-05, "val_loss": 2.952089548110962}
{"train_loss": 0.2684859037399292, "global_step": 45212, "epoch": 508, "lr": 9.342978038583653e-05}
{"train_loss": 0.274164080619812, "global_step": 45213, "epoch": 508, "lr": 9.342949311089968e-05}
{"train_loss": 0.40802985429763794, "global_step": 45214, "epoch": 508, "lr": 9.342920583012424e-05}
{"train_loss": 0.34303799271583557, "global_step": 45215, "epoch": 508, "lr": 9.342891854351028e-05}
{"train_loss": 0.336910218000412, "global_step": 45216, "epoch": 508, "lr": 9.342863125105783e-05}
{"train_loss": 0.2669731378555298, "global_step": 45217, "epoch": 508, "lr": 9.342834395276693e-05}
{"train_loss": 0.20220385491847992, "global_step": 45218, "epoch": 508, "lr": 9.34280566486376e-05}
{"train_loss": 0.3242335617542267, "global_step": 45219, "epoch": 508, "lr": 9.342776933866992e-05}
{"train_loss": 0.3490636944770813, "global_step": 45220, "epoch": 508, "lr": 9.342748202286391e-05}
{"train_loss": 0.27632516622543335, "global_step": 45221, "epoch": 508, "lr": 9.34271947012196e-05}
{"train_loss": 0.25368326902389526, "global_step": 45222, "epoch": 508, "lr": 9.342690737373702e-05}
{"train_loss": 0.239637553691864, "global_step": 45223, "epoch": 508, "lr": 9.342662004041622e-05}
{"train_loss": 0.2260051667690277, "global_step": 45224, "epoch": 508, "lr": 9.342633270125725e-05}
{"train_loss": 0.340247243642807, "global_step": 45225, "epoch": 508, "lr": 9.342604535626014e-05}
{"train_loss": 0.2328435778617859, "global_step": 45226, "epoch": 508, "lr": 9.342575800542493e-05}
{"train_loss": 0.3138415813446045, "global_step": 45227, "epoch": 508, "lr": 9.342547064875164e-05}
{"train_loss": 0.31834158301353455, "global_step": 45228, "epoch": 508, "lr": 9.342518328624035e-05}
{"train_loss": 0.37465178966522217, "global_step": 45229, "epoch": 508, "lr": 9.342489591789106e-05}
{"train_loss": 0.3185839354991913, "global_step": 45230, "epoch": 508, "lr": 9.342460854370381e-05}
{"train_loss": 0.37482136487960815, "global_step": 45231, "epoch": 508, "lr": 9.342432116367866e-05}
{"train_loss": 0.3648103177547455, "global_step": 45232, "epoch": 508, "lr": 9.342403377781565e-05}
{"train_loss": 0.3028404414653778, "global_step": 45233, "epoch": 508, "lr": 9.34237463861148e-05}
{"train_loss": 0.3954997658729553, "global_step": 45234, "epoch": 508, "lr": 9.342345898857614e-05}
{"train_loss": 0.2683996260166168, "global_step": 45235, "epoch": 508, "lr": 9.342317158519976e-05}
{"train_loss": 0.25967904925346375, "global_step": 45236, "epoch": 508, "lr": 9.342288417598564e-05}
{"train_loss": 0.2843351364135742, "global_step": 45237, "epoch": 508, "lr": 9.342259676093383e-05}
{"train_loss": 0.21982967853546143, "global_step": 45238, "epoch": 508, "lr": 9.342230934004441e-05}
{"train_loss": 0.3390145003795624, "global_step": 45239, "epoch": 508, "lr": 9.342202191331737e-05}
{"train_loss": 0.2607071101665497, "global_step": 45240, "epoch": 508, "lr": 9.342173448075277e-05}
{"train_loss": 0.3406871259212494, "global_step": 45241, "epoch": 508, "lr": 9.342144704235066e-05}
{"train_loss": 0.3406354486942291, "global_step": 45242, "epoch": 508, "lr": 9.342115959811107e-05}
{"train_loss": 0.21498095989227295, "global_step": 45243, "epoch": 508, "lr": 9.3420872148034e-05}
{"train_loss": 0.37879326939582825, "global_step": 45244, "epoch": 508, "lr": 9.342058469211956e-05}
{"train_loss": 0.23836587369441986, "global_step": 45245, "epoch": 508, "lr": 9.342029723036774e-05}
{"train_loss": 0.2307928204536438, "global_step": 45246, "epoch": 508, "lr": 9.342000976277858e-05}
{"train_loss": 0.459272563457489, "global_step": 45247, "epoch": 508, "lr": 9.341972228935214e-05}
{"train_loss": 0.45643001794815063, "global_step": 45248, "epoch": 508, "lr": 9.341943481008846e-05}
{"train_loss": 0.2632065415382385, "global_step": 45249, "epoch": 508, "lr": 9.341914732498753e-05}
{"train_loss": 0.25784748792648315, "global_step": 45250, "epoch": 508, "lr": 9.341885983404945e-05}
{"train_loss": 0.22996903955936432, "global_step": 45251, "epoch": 508, "lr": 9.341857233727423e-05}
{"train_loss": 0.3380116820335388, "global_step": 45252, "epoch": 508, "lr": 9.341828483466191e-05}
{"train_loss": 0.2728460431098938, "global_step": 45253, "epoch": 508, "lr": 9.341799732621253e-05}
{"train_loss": 0.21162670850753784, "global_step": 45254, "epoch": 508, "lr": 9.341770981192613e-05}
{"train_loss": 0.2526160478591919, "global_step": 45255, "epoch": 508, "lr": 9.341742229180275e-05}
{"train_loss": 0.2996427118778229, "global_step": 45256, "epoch": 508, "lr": 9.341713476584244e-05}
{"train_loss": 0.2820700407028198, "global_step": 45257, "epoch": 508, "lr": 9.34168472340452e-05}
{"train_loss": 0.30967339873313904, "global_step": 45258, "epoch": 508, "lr": 9.341655969641113e-05}
{"train_loss": 0.27224472165107727, "global_step": 45259, "epoch": 508, "lr": 9.34162721529402e-05}
{"train_loss": 0.2946036458015442, "global_step": 45260, "epoch": 508, "lr": 9.341598460363248e-05}
{"train_loss": 0.2549157440662384, "global_step": 45261, "epoch": 508, "lr": 9.341569704848802e-05}
{"train_loss": 0.3039957582950592, "global_step": 45262, "epoch": 508, "lr": 9.341540948750685e-05}
{"train_loss": 0.2592846155166626, "global_step": 45263, "epoch": 508, "lr": 9.3415121920689e-05}
{"train_loss": 0.26291975378990173, "global_step": 45264, "epoch": 508, "lr": 9.341483434803453e-05}
{"train_loss": 0.2339775711297989, "global_step": 45265, "epoch": 508, "lr": 9.341454676954346e-05}
{"train_loss": 0.25255823135375977, "global_step": 45266, "epoch": 508, "lr": 9.341425918521582e-05}
{"train_loss": 0.30974239110946655, "global_step": 45267, "epoch": 508, "lr": 9.341397159505167e-05}
{"train_loss": 0.26023128628730774, "global_step": 45268, "epoch": 508, "lr": 9.341368399905105e-05}
{"train_loss": 0.26292577385902405, "global_step": 45269, "epoch": 508, "lr": 9.341339639721398e-05}
{"train_loss": 0.2178281843662262, "global_step": 45270, "epoch": 508, "lr": 9.341310878954051e-05}
{"train_loss": 0.2503577172756195, "global_step": 45271, "epoch": 508, "lr": 9.341282117603068e-05}
{"train_loss": 0.2802942991256714, "global_step": 45272, "epoch": 508, "lr": 9.341253355668452e-05}
{"train_loss": 0.36595815420150757, "global_step": 45273, "epoch": 508, "lr": 9.341224593150207e-05}
{"train_loss": 0.371648907661438, "global_step": 45274, "epoch": 508, "lr": 9.341195830048338e-05}
{"train_loss": 0.2699694037437439, "global_step": 45275, "epoch": 508, "lr": 9.341167066362848e-05}
{"train_loss": 0.31082290410995483, "global_step": 45276, "epoch": 508, "lr": 9.341138302093741e-05}
{"train_loss": 0.26932477951049805, "global_step": 45277, "epoch": 508, "lr": 9.341109537241022e-05}
{"train_loss": 0.32093721628189087, "global_step": 45278, "epoch": 508, "lr": 9.341080771804691e-05}
{"train_loss": 0.30599409341812134, "global_step": 45279, "epoch": 508, "lr": 9.341052005784758e-05}
{"train_loss": 0.28218552470207214, "global_step": 45280, "epoch": 508, "lr": 9.341023239181222e-05}
{"train_loss": 0.20159323513507843, "global_step": 45281, "epoch": 508, "lr": 9.340994471994087e-05}
{"train_loss": 0.30388158559799194, "global_step": 45282, "epoch": 508, "lr": 9.34096570422336e-05}
{"train_loss": 0.36239203810691833, "global_step": 45283, "epoch": 508, "lr": 9.34093693586904e-05}
{"train_loss": 0.32954585552215576, "global_step": 45284, "epoch": 508, "lr": 9.340908166931137e-05}
{"train_loss": 0.42854681611061096, "global_step": 45285, "epoch": 508, "lr": 9.340879397409652e-05}
{"train_loss": 0.23406361043453217, "global_step": 45286, "epoch": 508, "lr": 9.340850627304588e-05}
{"train_loss": 0.22148767113685608, "global_step": 45287, "epoch": 508, "lr": 9.340821856615949e-05}
{"train_loss": 0.3615313172340393, "global_step": 45288, "epoch": 508, "lr": 9.34079308534374e-05}
{"train_loss": 0.31593191623687744, "global_step": 45289, "epoch": 508, "lr": 9.340764313487963e-05}
{"train_loss": 0.23032701015472412, "global_step": 45290, "epoch": 508, "lr": 9.340735541048625e-05}
{"train_loss": 0.21345411241054535, "global_step": 45291, "epoch": 508, "lr": 9.340706768025727e-05}
{"train_loss": 0.3456653952598572, "global_step": 45292, "epoch": 508, "lr": 9.340677994419273e-05}
{"train_loss": 0.2263140231370926, "global_step": 45293, "epoch": 508, "lr": 9.34064922022927e-05}
{"train_loss": 0.33930715918540955, "global_step": 45294, "epoch": 508, "lr": 9.340620445455717e-05}
{"train_loss": 0.2503867745399475, "global_step": 45295, "epoch": 508, "lr": 9.340591670098622e-05}
{"train_loss": 0.20032402873039246, "global_step": 45296, "epoch": 508, "lr": 9.340562894157986e-05}
{"train_loss": 0.3164469003677368, "global_step": 45297, "epoch": 508, "lr": 9.340534117633816e-05}
{"train_loss": 0.3092147707939148, "global_step": 45298, "epoch": 508, "lr": 9.340505340526113e-05}
{"train_loss": 0.28823330998420715, "global_step": 45299, "epoch": 508, "lr": 9.340476562834884e-05}
{"train_loss": 0.2942327495371358, "global_step": 45300, "epoch": 508, "lr": 9.340447784560128e-05, "val_loss": 2.937858819961548}
{"train_loss": 0.24654991924762726, "global_step": 45301, "epoch": 509, "lr": 9.340419005701852e-05}
{"train_loss": 0.273489773273468, "global_step": 45302, "epoch": 509, "lr": 9.340390226260061e-05}
{"train_loss": 0.23462221026420593, "global_step": 45303, "epoch": 509, "lr": 9.340361446234758e-05}
{"train_loss": 0.31332528591156006, "global_step": 45304, "epoch": 509, "lr": 9.340332665625946e-05}
{"train_loss": 0.2560461461544037, "global_step": 45305, "epoch": 509, "lr": 9.340303884433628e-05}
{"train_loss": 0.2754307687282562, "global_step": 45306, "epoch": 509, "lr": 9.34027510265781e-05}
{"train_loss": 0.21682602167129517, "global_step": 45307, "epoch": 509, "lr": 9.340246320298493e-05}
{"train_loss": 0.2520718574523926, "global_step": 45308, "epoch": 509, "lr": 9.340217537355685e-05}
{"train_loss": 0.2872053384780884, "global_step": 45309, "epoch": 509, "lr": 9.340188753829386e-05}
{"train_loss": 0.2885439991950989, "global_step": 45310, "epoch": 509, "lr": 9.340159969719602e-05}
{"train_loss": 0.27667662501335144, "global_step": 45311, "epoch": 509, "lr": 9.340131185026337e-05}
{"train_loss": 0.3144908547401428, "global_step": 45312, "epoch": 509, "lr": 9.340102399749595e-05}
{"train_loss": 0.3132449984550476, "global_step": 45313, "epoch": 509, "lr": 9.340073613889377e-05}
{"train_loss": 0.27945733070373535, "global_step": 45314, "epoch": 509, "lr": 9.340044827445691e-05}
{"train_loss": 0.3474307954311371, "global_step": 45315, "epoch": 509, "lr": 9.340016040418538e-05}
{"train_loss": 0.333296000957489, "global_step": 45316, "epoch": 509, "lr": 9.339987252807922e-05}
{"train_loss": 0.1929592788219452, "global_step": 45317, "epoch": 509, "lr": 9.33995846461385e-05}
{"train_loss": 0.29196205735206604, "global_step": 45318, "epoch": 509, "lr": 9.339929675836321e-05}
{"train_loss": 0.2846916615962982, "global_step": 45319, "epoch": 509, "lr": 9.339900886475342e-05}
{"train_loss": 0.22351539134979248, "global_step": 45320, "epoch": 509, "lr": 9.339872096530916e-05}
{"train_loss": 0.2466607689857483, "global_step": 45321, "epoch": 509, "lr": 9.339843306003048e-05}
{"train_loss": 0.305751234292984, "global_step": 45322, "epoch": 509, "lr": 9.339814514891741e-05}
{"train_loss": 0.2731596827507019, "global_step": 45323, "epoch": 509, "lr": 9.339785723196998e-05}
{"train_loss": 0.3012421429157257, "global_step": 45324, "epoch": 509, "lr": 9.339756930918824e-05}
{"train_loss": 0.2948015034198761, "global_step": 45325, "epoch": 509, "lr": 9.339728138057222e-05}
{"train_loss": 0.2794535458087921, "global_step": 45326, "epoch": 509, "lr": 9.339699344612197e-05}
{"train_loss": 0.3369775712490082, "global_step": 45327, "epoch": 509, "lr": 9.339670550583752e-05}
{"train_loss": 0.2691022455692291, "global_step": 45328, "epoch": 509, "lr": 9.339641755971892e-05}
{"train_loss": 0.3394797146320343, "global_step": 45329, "epoch": 509, "lr": 9.33961296077662e-05}
{"train_loss": 0.4583739936351776, "global_step": 45330, "epoch": 509, "lr": 9.339584164997939e-05}
{"train_loss": 0.30069640278816223, "global_step": 45331, "epoch": 509, "lr": 9.339555368635854e-05}
{"train_loss": 0.2932800352573395, "global_step": 45332, "epoch": 509, "lr": 9.339526571690369e-05}
{"train_loss": 0.1914299875497818, "global_step": 45333, "epoch": 509, "lr": 9.339497774161488e-05}
{"train_loss": 0.22083081305027008, "global_step": 45334, "epoch": 509, "lr": 9.339468976049214e-05}
{"train_loss": 0.4025418758392334, "global_step": 45335, "epoch": 509, "lr": 9.339440177353551e-05}
{"train_loss": 0.17332567274570465, "global_step": 45336, "epoch": 509, "lr": 9.339411378074503e-05}
{"train_loss": 0.35319793224334717, "global_step": 45337, "epoch": 509, "lr": 9.339382578212075e-05}
{"train_loss": 0.23972664773464203, "global_step": 45338, "epoch": 509, "lr": 9.339353777766268e-05}
{"train_loss": 0.375837504863739, "global_step": 45339, "epoch": 509, "lr": 9.339324976737091e-05}
{"train_loss": 0.3112484812736511, "global_step": 45340, "epoch": 509, "lr": 9.339296175124542e-05}
{"train_loss": 0.18658442795276642, "global_step": 45341, "epoch": 509, "lr": 9.339267372928628e-05}
{"train_loss": 0.3959878087043762, "global_step": 45342, "epoch": 509, "lr": 9.339238570149353e-05}
{"train_loss": 0.39020979404449463, "global_step": 45343, "epoch": 509, "lr": 9.33920976678672e-05}
{"train_loss": 0.31367623805999756, "global_step": 45344, "epoch": 509, "lr": 9.339180962840734e-05}
{"train_loss": 0.24149103462696075, "global_step": 45345, "epoch": 509, "lr": 9.339152158311397e-05}
{"train_loss": 0.20500805974006653, "global_step": 45346, "epoch": 509, "lr": 9.339123353198714e-05}
{"train_loss": 0.32544946670532227, "global_step": 45347, "epoch": 509, "lr": 9.339094547502687e-05}
{"train_loss": 0.15674100816249847, "global_step": 45348, "epoch": 509, "lr": 9.339065741223324e-05}
{"train_loss": 0.4015278220176697, "global_step": 45349, "epoch": 509, "lr": 9.339036934360625e-05}
{"train_loss": 0.37852346897125244, "global_step": 45350, "epoch": 509, "lr": 9.339008126914597e-05}
{"train_loss": 0.2572488784790039, "global_step": 45351, "epoch": 509, "lr": 9.338979318885242e-05}
{"train_loss": 0.2255372703075409, "global_step": 45352, "epoch": 509, "lr": 9.338950510272563e-05}
{"train_loss": 0.25689318776130676, "global_step": 45353, "epoch": 509, "lr": 9.338921701076566e-05}
{"train_loss": 0.32138895988464355, "global_step": 45354, "epoch": 509, "lr": 9.338892891297252e-05}
{"train_loss": 0.30438101291656494, "global_step": 45355, "epoch": 509, "lr": 9.338864080934628e-05}
{"train_loss": 0.3604177236557007, "global_step": 45356, "epoch": 509, "lr": 9.338835269988698e-05}
{"train_loss": 0.3455500304698944, "global_step": 45357, "epoch": 509, "lr": 9.338806458459462e-05}
{"train_loss": 0.24654093384742737, "global_step": 45358, "epoch": 509, "lr": 9.338777646346928e-05}
{"train_loss": 0.33393099904060364, "global_step": 45359, "epoch": 509, "lr": 9.338748833651099e-05}
{"train_loss": 0.3885992765426636, "global_step": 45360, "epoch": 509, "lr": 9.338720020371976e-05}
{"train_loss": 0.267641544342041, "global_step": 45361, "epoch": 509, "lr": 9.338691206509567e-05}
{"train_loss": 0.24405136704444885, "global_step": 45362, "epoch": 509, "lr": 9.338662392063872e-05}
{"train_loss": 0.25491249561309814, "global_step": 45363, "epoch": 509, "lr": 9.338633577034896e-05}
{"train_loss": 0.2535523474216461, "global_step": 45364, "epoch": 509, "lr": 9.338604761422646e-05}
{"train_loss": 0.4061262011528015, "global_step": 45365, "epoch": 509, "lr": 9.338575945227123e-05}
{"train_loss": 0.28873133659362793, "global_step": 45366, "epoch": 509, "lr": 9.338547128448331e-05}
{"train_loss": 0.22118951380252838, "global_step": 45367, "epoch": 509, "lr": 9.338518311086272e-05}
{"train_loss": 0.31419244408607483, "global_step": 45368, "epoch": 509, "lr": 9.338489493140956e-05}
{"train_loss": 0.282185822725296, "global_step": 45369, "epoch": 509, "lr": 9.338460674612382e-05}
{"train_loss": 0.2981105446815491, "global_step": 45370, "epoch": 509, "lr": 9.338431855500553e-05}
{"train_loss": 0.3087264895439148, "global_step": 45371, "epoch": 509, "lr": 9.338403035805476e-05}
{"train_loss": 0.25489187240600586, "global_step": 45372, "epoch": 509, "lr": 9.338374215527152e-05}
{"train_loss": 0.30669108033180237, "global_step": 45373, "epoch": 509, "lr": 9.338345394665588e-05}
{"train_loss": 0.27440786361694336, "global_step": 45374, "epoch": 509, "lr": 9.338316573220786e-05}
{"train_loss": 0.28819578886032104, "global_step": 45375, "epoch": 509, "lr": 9.338287751192751e-05}
{"train_loss": 0.3700055480003357, "global_step": 45376, "epoch": 509, "lr": 9.338258928581484e-05}
{"train_loss": 0.43996235728263855, "global_step": 45377, "epoch": 509, "lr": 9.338230105386994e-05}
{"train_loss": 0.3399334251880646, "global_step": 45378, "epoch": 509, "lr": 9.33820128160928e-05}
{"train_loss": 0.20068512856960297, "global_step": 45379, "epoch": 509, "lr": 9.338172457248348e-05}
{"train_loss": 0.23665592074394226, "global_step": 45380, "epoch": 509, "lr": 9.338143632304202e-05}
{"train_loss": 0.18410545587539673, "global_step": 45381, "epoch": 509, "lr": 9.338114806776845e-05}
{"train_loss": 0.18738631904125214, "global_step": 45382, "epoch": 509, "lr": 9.33808598066628e-05}
{"train_loss": 0.24929551780223846, "global_step": 45383, "epoch": 509, "lr": 9.338057153972514e-05}
{"train_loss": 0.39629530906677246, "global_step": 45384, "epoch": 509, "lr": 9.338028326695548e-05}
{"train_loss": 0.2122679352760315, "global_step": 45385, "epoch": 509, "lr": 9.337999498835388e-05}
{"train_loss": 0.21250884234905243, "global_step": 45386, "epoch": 509, "lr": 9.337970670392036e-05}
{"train_loss": 0.3404967784881592, "global_step": 45387, "epoch": 509, "lr": 9.337941841365498e-05}
{"train_loss": 0.25331997871398926, "global_step": 45388, "epoch": 509, "lr": 9.337913011755775e-05}
{"train_loss": 0.2885554230280137, "global_step": 45389, "epoch": 509, "lr": 9.337884181562872e-05, "val_loss": 2.8898708820343018}
{"train_loss": 0.32560819387435913, "global_step": 45390, "epoch": 510, "lr": 9.337855350786796e-05}
{"train_loss": 0.2468002587556839, "global_step": 45391, "epoch": 510, "lr": 9.337826519427546e-05}
{"train_loss": 0.246639221906662, "global_step": 45392, "epoch": 510, "lr": 9.337797687485129e-05}
{"train_loss": 0.2615988552570343, "global_step": 45393, "epoch": 510, "lr": 9.337768854959547e-05}
{"train_loss": 0.23710386455059052, "global_step": 45394, "epoch": 510, "lr": 9.337740021850807e-05}
{"train_loss": 0.24340525269508362, "global_step": 45395, "epoch": 510, "lr": 9.337711188158909e-05}
{"train_loss": 0.3053232729434967, "global_step": 45396, "epoch": 510, "lr": 9.337682353883859e-05}
{"train_loss": 0.34074392914772034, "global_step": 45397, "epoch": 510, "lr": 9.33765351902566e-05}
{"train_loss": 0.3104860782623291, "global_step": 45398, "epoch": 510, "lr": 9.337624683584317e-05}
{"train_loss": 0.3476370573043823, "global_step": 45399, "epoch": 510, "lr": 9.337595847559833e-05}
{"train_loss": 0.30515119433403015, "global_step": 45400, "epoch": 510, "lr": 9.337567010952213e-05}
{"train_loss": 0.3683084547519684, "global_step": 45401, "epoch": 510, "lr": 9.337538173761459e-05}
{"train_loss": 0.24935241043567657, "global_step": 45402, "epoch": 510, "lr": 9.337509335987576e-05}
{"train_loss": 0.31852424144744873, "global_step": 45403, "epoch": 510, "lr": 9.337480497630567e-05}
{"train_loss": 0.2680075764656067, "global_step": 45404, "epoch": 510, "lr": 9.337451658690437e-05}
{"train_loss": 0.2259373515844345, "global_step": 45405, "epoch": 510, "lr": 9.33742281916719e-05}
{"train_loss": 0.23182542622089386, "global_step": 45406, "epoch": 510, "lr": 9.337393979060829e-05}
{"train_loss": 0.3016999661922455, "global_step": 45407, "epoch": 510, "lr": 9.337365138371359e-05}
{"train_loss": 0.4274686276912689, "global_step": 45408, "epoch": 510, "lr": 9.337336297098782e-05}
{"train_loss": 0.34915873408317566, "global_step": 45409, "epoch": 510, "lr": 9.337307455243104e-05}
{"train_loss": 0.25828972458839417, "global_step": 45410, "epoch": 510, "lr": 9.337278612804328e-05}
{"train_loss": 0.1861739456653595, "global_step": 45411, "epoch": 510, "lr": 9.337249769782456e-05}
{"train_loss": 0.2307249903678894, "global_step": 45412, "epoch": 510, "lr": 9.337220926177495e-05}
{"train_loss": 0.30825555324554443, "global_step": 45413, "epoch": 510, "lr": 9.337192081989448e-05}
{"train_loss": 0.2568708062171936, "global_step": 45414, "epoch": 510, "lr": 9.337163237218318e-05}
{"train_loss": 0.2665002942085266, "global_step": 45415, "epoch": 510, "lr": 9.337134391864108e-05}
{"train_loss": 0.2554211914539337, "global_step": 45416, "epoch": 510, "lr": 9.337105545926825e-05}
{"train_loss": 0.27501964569091797, "global_step": 45417, "epoch": 510, "lr": 9.337076699406469e-05}
{"train_loss": 0.3966866731643677, "global_step": 45418, "epoch": 510, "lr": 9.337047852303048e-05}
{"train_loss": 0.2620697319507599, "global_step": 45419, "epoch": 510, "lr": 9.337019004616564e-05}
{"train_loss": 0.3121022582054138, "global_step": 45420, "epoch": 510, "lr": 9.336990156347018e-05}
{"train_loss": 0.25193703174591064, "global_step": 45421, "epoch": 510, "lr": 9.336961307494418e-05}
{"train_loss": 0.32410627603530884, "global_step": 45422, "epoch": 510, "lr": 9.336932458058767e-05}
{"train_loss": 0.3286389410495758, "global_step": 45423, "epoch": 510, "lr": 9.336903608040068e-05}
{"train_loss": 0.25994718074798584, "global_step": 45424, "epoch": 510, "lr": 9.336874757438325e-05}
{"train_loss": 0.32216179370880127, "global_step": 45425, "epoch": 510, "lr": 9.336845906253541e-05}
{"train_loss": 0.3986913859844208, "global_step": 45426, "epoch": 510, "lr": 9.336817054485723e-05}
{"train_loss": 0.26339784264564514, "global_step": 45427, "epoch": 510, "lr": 9.33678820213487e-05}
{"train_loss": 0.24485252797603607, "global_step": 45428, "epoch": 510, "lr": 9.336759349200991e-05}
{"train_loss": 0.348528653383255, "global_step": 45429, "epoch": 510, "lr": 9.336730495684087e-05}
{"train_loss": 0.22707682847976685, "global_step": 45430, "epoch": 510, "lr": 9.336701641584163e-05}
{"train_loss": 0.2703408896923065, "global_step": 45431, "epoch": 510, "lr": 9.33667278690122e-05}
{"train_loss": 0.2856847047805786, "global_step": 45432, "epoch": 510, "lr": 9.336643931635268e-05}
{"train_loss": 0.32725799083709717, "global_step": 45433, "epoch": 510, "lr": 9.336615075786303e-05}
{"train_loss": 0.17945489287376404, "global_step": 45434, "epoch": 510, "lr": 9.336586219354335e-05}
{"train_loss": 0.36833739280700684, "global_step": 45435, "epoch": 510, "lr": 9.336557362339367e-05}
{"train_loss": 0.3342626094818115, "global_step": 45436, "epoch": 510, "lr": 9.3365285047414e-05}
{"train_loss": 0.301141619682312, "global_step": 45437, "epoch": 510, "lr": 9.336499646560439e-05}
{"train_loss": 0.27098751068115234, "global_step": 45438, "epoch": 510, "lr": 9.33647078779649e-05}
{"train_loss": 0.2973364591598511, "global_step": 45439, "epoch": 510, "lr": 9.336441928449555e-05}
{"train_loss": 0.34883785247802734, "global_step": 45440, "epoch": 510, "lr": 9.336413068519638e-05}
{"train_loss": 0.3680534362792969, "global_step": 45441, "epoch": 510, "lr": 9.336384208006743e-05}
{"train_loss": 0.30365341901779175, "global_step": 45442, "epoch": 510, "lr": 9.336355346910874e-05}
{"train_loss": 0.33971115946769714, "global_step": 45443, "epoch": 510, "lr": 9.336326485232037e-05}
{"train_loss": 0.31685271859169006, "global_step": 45444, "epoch": 510, "lr": 9.336297622970232e-05}
{"train_loss": 0.27434471249580383, "global_step": 45445, "epoch": 510, "lr": 9.336268760125463e-05}
{"train_loss": 0.2988240718841553, "global_step": 45446, "epoch": 510, "lr": 9.336239896697737e-05}
{"train_loss": 0.43007534742355347, "global_step": 45447, "epoch": 510, "lr": 9.336211032687057e-05}
{"train_loss": 0.3441988229751587, "global_step": 45448, "epoch": 510, "lr": 9.336182168093426e-05}
{"train_loss": 0.34089037775993347, "global_step": 45449, "epoch": 510, "lr": 9.336153302916848e-05}
{"train_loss": 0.29053372144699097, "global_step": 45450, "epoch": 510, "lr": 9.336124437157326e-05}
{"train_loss": 0.36066144704818726, "global_step": 45451, "epoch": 510, "lr": 9.336095570814866e-05}
{"train_loss": 0.3398681581020355, "global_step": 45452, "epoch": 510, "lr": 9.33606670388947e-05}
{"train_loss": 0.26078730821609497, "global_step": 45453, "epoch": 510, "lr": 9.336037836381145e-05}
{"train_loss": 0.2376612424850464, "global_step": 45454, "epoch": 510, "lr": 9.336008968289888e-05}
{"train_loss": 0.2534891366958618, "global_step": 45455, "epoch": 510, "lr": 9.335980099615712e-05}
{"train_loss": 0.25907331705093384, "global_step": 45456, "epoch": 510, "lr": 9.335951230358615e-05}
{"train_loss": 0.30941516160964966, "global_step": 45457, "epoch": 510, "lr": 9.3359223605186e-05}
{"train_loss": 0.3438200354576111, "global_step": 45458, "epoch": 510, "lr": 9.335893490095676e-05}
{"train_loss": 0.2625490128993988, "global_step": 45459, "epoch": 510, "lr": 9.335864619089843e-05}
{"train_loss": 0.3556019067764282, "global_step": 45460, "epoch": 510, "lr": 9.335835747501106e-05}
{"train_loss": 0.23053506016731262, "global_step": 45461, "epoch": 510, "lr": 9.33580687532947e-05}
{"train_loss": 0.2441367357969284, "global_step": 45462, "epoch": 510, "lr": 9.335778002574935e-05}
{"train_loss": 0.24212123453617096, "global_step": 45463, "epoch": 510, "lr": 9.335749129237508e-05}
{"train_loss": 0.26297855377197266, "global_step": 45464, "epoch": 510, "lr": 9.335720255317193e-05}
{"train_loss": 0.23394367098808289, "global_step": 45465, "epoch": 510, "lr": 9.335691380813993e-05}
{"train_loss": 0.2548152208328247, "global_step": 45466, "epoch": 510, "lr": 9.335662505727913e-05}
{"train_loss": 0.2781316339969635, "global_step": 45467, "epoch": 510, "lr": 9.335633630058955e-05}
{"train_loss": 0.4167247712612152, "global_step": 45468, "epoch": 510, "lr": 9.335604753807125e-05}
{"train_loss": 0.27771076560020447, "global_step": 45469, "epoch": 510, "lr": 9.335575876972425e-05}
{"train_loss": 0.23883552849292755, "global_step": 45470, "epoch": 510, "lr": 9.335546999554859e-05}
{"train_loss": 0.3093743920326233, "global_step": 45471, "epoch": 510, "lr": 9.335518121554432e-05}
{"train_loss": 0.23810946941375732, "global_step": 45472, "epoch": 510, "lr": 9.335489242971147e-05}
{"train_loss": 0.28140756487846375, "global_step": 45473, "epoch": 510, "lr": 9.33546036380501e-05}
{"train_loss": 0.3331286311149597, "global_step": 45474, "epoch": 510, "lr": 9.335431484056023e-05}
{"train_loss": 0.28270187973976135, "global_step": 45475, "epoch": 510, "lr": 9.335402603724188e-05}
{"train_loss": 0.3152686655521393, "global_step": 45476, "epoch": 510, "lr": 9.335373722809513e-05}
{"train_loss": 0.31479740142822266, "global_step": 45477, "epoch": 510, "lr": 9.335344841311999e-05}
{"train_loss": 0.2943755698003126, "global_step": 45478, "epoch": 510, "lr": 9.33531595923165e-05, "val_loss": 2.79852557182312, "train_action_mse_error": 15.81869125366211}
{"train_loss": 0.32200121879577637, "global_step": 45479, "epoch": 511, "lr": 9.335287076568472e-05}
{"train_loss": 0.274411678314209, "global_step": 45480, "epoch": 511, "lr": 9.335258193322467e-05}
{"train_loss": 0.2786428928375244, "global_step": 45481, "epoch": 511, "lr": 9.335229309493638e-05}
{"train_loss": 0.27784496545791626, "global_step": 45482, "epoch": 511, "lr": 9.335200425081992e-05}
{"train_loss": 0.2546900808811188, "global_step": 45483, "epoch": 511, "lr": 9.335171540087531e-05}
{"train_loss": 0.26738443970680237, "global_step": 45484, "epoch": 511, "lr": 9.335142654510259e-05}
{"train_loss": 0.22060787677764893, "global_step": 45485, "epoch": 511, "lr": 9.33511376835018e-05}
{"train_loss": 0.2836851179599762, "global_step": 45486, "epoch": 511, "lr": 9.335084881607296e-05}
{"train_loss": 0.37386074662208557, "global_step": 45487, "epoch": 511, "lr": 9.335055994281616e-05}
{"train_loss": 0.35322684049606323, "global_step": 45488, "epoch": 511, "lr": 9.335027106373138e-05}
{"train_loss": 0.25021985173225403, "global_step": 45489, "epoch": 511, "lr": 9.33499821788187e-05}
{"train_loss": 0.24231941998004913, "global_step": 45490, "epoch": 511, "lr": 9.334969328807813e-05}
{"train_loss": 0.4076964259147644, "global_step": 45491, "epoch": 511, "lr": 9.334940439150972e-05}
{"train_loss": 0.3127720355987549, "global_step": 45492, "epoch": 511, "lr": 9.334911548911352e-05}
{"train_loss": 0.2775536775588989, "global_step": 45493, "epoch": 511, "lr": 9.334882658088955e-05}
{"train_loss": 0.3448145091533661, "global_step": 45494, "epoch": 511, "lr": 9.334853766683787e-05}
{"train_loss": 0.2918919026851654, "global_step": 45495, "epoch": 511, "lr": 9.334824874695851e-05}
{"train_loss": 0.2445523738861084, "global_step": 45496, "epoch": 511, "lr": 9.334795982125149e-05}
{"train_loss": 0.19239430129528046, "global_step": 45497, "epoch": 511, "lr": 9.334767088971688e-05}
{"train_loss": 0.28223928809165955, "global_step": 45498, "epoch": 511, "lr": 9.33473819523547e-05}
{"train_loss": 0.2904440760612488, "global_step": 45499, "epoch": 511, "lr": 9.334709300916498e-05}
{"train_loss": 0.27433153986930847, "global_step": 45500, "epoch": 511, "lr": 9.334680406014779e-05}
{"train_loss": 0.34502333402633667, "global_step": 45501, "epoch": 511, "lr": 9.334651510530315e-05}
{"train_loss": 0.2898857593536377, "global_step": 45502, "epoch": 511, "lr": 9.33462261446311e-05}
{"train_loss": 0.2850764989852905, "global_step": 45503, "epoch": 511, "lr": 9.334593717813166e-05}
{"train_loss": 0.272808700799942, "global_step": 45504, "epoch": 511, "lr": 9.334564820580489e-05}
{"train_loss": 0.27724823355674744, "global_step": 45505, "epoch": 511, "lr": 9.334535922765084e-05}
{"train_loss": 0.2531777322292328, "global_step": 45506, "epoch": 511, "lr": 9.334507024366953e-05}
{"train_loss": 0.26954323053359985, "global_step": 45507, "epoch": 511, "lr": 9.334478125386101e-05}
{"train_loss": 0.2574775516986847, "global_step": 45508, "epoch": 511, "lr": 9.33444922582253e-05}
{"train_loss": 0.307590126991272, "global_step": 45509, "epoch": 511, "lr": 9.334420325676245e-05}
{"train_loss": 0.31274911761283875, "global_step": 45510, "epoch": 511, "lr": 9.33439142494725e-05}
{"train_loss": 0.2677382528781891, "global_step": 45511, "epoch": 511, "lr": 9.33436252363555e-05}
{"train_loss": 0.2872525155544281, "global_step": 45512, "epoch": 511, "lr": 9.334333621741147e-05}
{"train_loss": 0.36463233828544617, "global_step": 45513, "epoch": 511, "lr": 9.334304719264047e-05}
{"train_loss": 0.2804594039916992, "global_step": 45514, "epoch": 511, "lr": 9.334275816204252e-05}
{"train_loss": 0.1891806423664093, "global_step": 45515, "epoch": 511, "lr": 9.334246912561765e-05}
{"train_loss": 0.3210124671459198, "global_step": 45516, "epoch": 511, "lr": 9.334218008336593e-05}
{"train_loss": 0.33450087904930115, "global_step": 45517, "epoch": 511, "lr": 9.334189103528737e-05}
{"train_loss": 0.2748069167137146, "global_step": 45518, "epoch": 511, "lr": 9.334160198138202e-05}
{"train_loss": 0.36278122663497925, "global_step": 45519, "epoch": 511, "lr": 9.334131292164994e-05}
{"train_loss": 0.2531646490097046, "global_step": 45520, "epoch": 511, "lr": 9.334102385609112e-05}
{"train_loss": 0.4108148515224457, "global_step": 45521, "epoch": 511, "lr": 9.334073478470565e-05}
{"train_loss": 0.337297648191452, "global_step": 45522, "epoch": 511, "lr": 9.334044570749353e-05}
{"train_loss": 0.2804478108882904, "global_step": 45523, "epoch": 511, "lr": 9.334015662445483e-05}
{"train_loss": 0.2841627597808838, "global_step": 45524, "epoch": 511, "lr": 9.333986753558957e-05}
{"train_loss": 0.35877367854118347, "global_step": 45525, "epoch": 511, "lr": 9.33395784408978e-05}
{"train_loss": 0.41697898507118225, "global_step": 45526, "epoch": 511, "lr": 9.333928934037955e-05}
{"train_loss": 0.3069709539413452, "global_step": 45527, "epoch": 511, "lr": 9.333900023403483e-05}
{"train_loss": 0.334897518157959, "global_step": 45528, "epoch": 511, "lr": 9.333871112186374e-05}
{"train_loss": 0.2080754041671753, "global_step": 45529, "epoch": 511, "lr": 9.333842200386629e-05}
{"train_loss": 0.26035407185554504, "global_step": 45530, "epoch": 511, "lr": 9.33381328800425e-05}
{"train_loss": 0.33452892303466797, "global_step": 45531, "epoch": 511, "lr": 9.333784375039245e-05}
{"train_loss": 0.45134320855140686, "global_step": 45532, "epoch": 511, "lr": 9.333755461491613e-05}
{"train_loss": 0.35691678524017334, "global_step": 45533, "epoch": 511, "lr": 9.333726547361362e-05}
{"train_loss": 0.30687785148620605, "global_step": 45534, "epoch": 511, "lr": 9.333697632648495e-05}
{"train_loss": 0.4010921120643616, "global_step": 45535, "epoch": 511, "lr": 9.333668717353013e-05}
{"train_loss": 0.20266152918338776, "global_step": 45536, "epoch": 511, "lr": 9.333639801474923e-05}
{"train_loss": 0.33337247371673584, "global_step": 45537, "epoch": 511, "lr": 9.333610885014228e-05}
{"train_loss": 0.40479710698127747, "global_step": 45538, "epoch": 511, "lr": 9.333581967970932e-05}
{"train_loss": 0.4368143081665039, "global_step": 45539, "epoch": 511, "lr": 9.333553050345039e-05}
{"train_loss": 0.28319108486175537, "global_step": 45540, "epoch": 511, "lr": 9.33352413213655e-05}
{"train_loss": 0.28838473558425903, "global_step": 45541, "epoch": 511, "lr": 9.333495213345476e-05}
{"train_loss": 0.21217215061187744, "global_step": 45542, "epoch": 511, "lr": 9.333466293971813e-05}
{"train_loss": 0.2174534797668457, "global_step": 45543, "epoch": 511, "lr": 9.33343737401557e-05}
{"train_loss": 0.2976765036582947, "global_step": 45544, "epoch": 511, "lr": 9.333408453476749e-05}
{"train_loss": 0.264138400554657, "global_step": 45545, "epoch": 511, "lr": 9.333379532355353e-05}
{"train_loss": 0.32954296469688416, "global_step": 45546, "epoch": 511, "lr": 9.333350610651387e-05}
{"train_loss": 0.28395169973373413, "global_step": 45547, "epoch": 511, "lr": 9.333321688364856e-05}
{"train_loss": 0.31169259548187256, "global_step": 45548, "epoch": 511, "lr": 9.333292765495762e-05}
{"train_loss": 0.3663749694824219, "global_step": 45549, "epoch": 511, "lr": 9.333263842044109e-05}
{"train_loss": 0.3871265947818756, "global_step": 45550, "epoch": 511, "lr": 9.333234918009903e-05}
{"train_loss": 0.41612735390663147, "global_step": 45551, "epoch": 511, "lr": 9.333205993393145e-05}
{"train_loss": 0.2880016267299652, "global_step": 45552, "epoch": 511, "lr": 9.333177068193841e-05}
{"train_loss": 0.2920522689819336, "global_step": 45553, "epoch": 511, "lr": 9.333148142411993e-05}
{"train_loss": 0.3435910642147064, "global_step": 45554, "epoch": 511, "lr": 9.333119216047608e-05}
{"train_loss": 0.31721457839012146, "global_step": 45555, "epoch": 511, "lr": 9.333090289100685e-05}
{"train_loss": 0.4241931736469269, "global_step": 45556, "epoch": 511, "lr": 9.333061361571233e-05}
{"train_loss": 0.3225918412208557, "global_step": 45557, "epoch": 511, "lr": 9.333032433459255e-05}
{"train_loss": 0.2587379217147827, "global_step": 45558, "epoch": 511, "lr": 9.333003504764751e-05}
{"train_loss": 0.3309144079685211, "global_step": 45559, "epoch": 511, "lr": 9.33297457548773e-05}
{"train_loss": 0.26870837807655334, "global_step": 45560, "epoch": 511, "lr": 9.332945645628191e-05}
{"train_loss": 0.257301926612854, "global_step": 45561, "epoch": 511, "lr": 9.33291671518614e-05}
{"train_loss": 0.337286114692688, "global_step": 45562, "epoch": 511, "lr": 9.332887784161583e-05}
{"train_loss": 0.28655433654785156, "global_step": 45563, "epoch": 511, "lr": 9.332858852554522e-05}
{"train_loss": 0.27625954151153564, "global_step": 45564, "epoch": 511, "lr": 9.332829920364961e-05}
{"train_loss": 0.420855313539505, "global_step": 45565, "epoch": 511, "lr": 9.332800987592903e-05}
{"train_loss": 0.302500456571579, "global_step": 45566, "epoch": 511, "lr": 9.332772054238353e-05}
{"train_loss": 0.30605074780040914, "global_step": 45567, "epoch": 511, "lr": 9.332743120301312e-05, "val_loss": 2.831273317337036}
{"train_loss": 0.24494674801826477, "global_step": 45568, "epoch": 512, "lr": 9.33271418578179e-05}
{"train_loss": 0.3422749638557434, "global_step": 45569, "epoch": 512, "lr": 9.332685250679787e-05}
{"train_loss": 0.2529260516166687, "global_step": 45570, "epoch": 512, "lr": 9.332656314995306e-05}
{"train_loss": 0.22074612975120544, "global_step": 45571, "epoch": 512, "lr": 9.332627378728354e-05}
{"train_loss": 0.32408061623573303, "global_step": 45572, "epoch": 512, "lr": 9.33259844187893e-05}
{"train_loss": 0.2729623317718506, "global_step": 45573, "epoch": 512, "lr": 9.332569504447044e-05}
{"train_loss": 0.3081921339035034, "global_step": 45574, "epoch": 512, "lr": 9.332540566432695e-05}
{"train_loss": 0.23406095802783966, "global_step": 45575, "epoch": 512, "lr": 9.33251162783589e-05}
{"train_loss": 0.2874862551689148, "global_step": 45576, "epoch": 512, "lr": 9.33248268865663e-05}
{"train_loss": 0.44702985882759094, "global_step": 45577, "epoch": 512, "lr": 9.332453748894923e-05}
{"train_loss": 0.3459585905075073, "global_step": 45578, "epoch": 512, "lr": 9.332424808550768e-05}
{"train_loss": 0.21289046108722687, "global_step": 45579, "epoch": 512, "lr": 9.332395867624173e-05}
{"train_loss": 0.2437993884086609, "global_step": 45580, "epoch": 512, "lr": 9.33236692611514e-05}
{"train_loss": 0.25554776191711426, "global_step": 45581, "epoch": 512, "lr": 9.332337984023671e-05}
{"train_loss": 0.3912317454814911, "global_step": 45582, "epoch": 512, "lr": 9.332309041349773e-05}
{"train_loss": 0.37204504013061523, "global_step": 45583, "epoch": 512, "lr": 9.33228009809345e-05}
{"train_loss": 0.35726791620254517, "global_step": 45584, "epoch": 512, "lr": 9.332251154254704e-05}
{"train_loss": 0.29755666851997375, "global_step": 45585, "epoch": 512, "lr": 9.332222209833538e-05}
{"train_loss": 0.33322781324386597, "global_step": 45586, "epoch": 512, "lr": 9.332193264829961e-05}
{"train_loss": 0.24124661087989807, "global_step": 45587, "epoch": 512, "lr": 9.332164319243971e-05}
{"train_loss": 0.3881683051586151, "global_step": 45588, "epoch": 512, "lr": 9.332135373075574e-05}
{"train_loss": 0.4371299743652344, "global_step": 45589, "epoch": 512, "lr": 9.332106426324776e-05}
{"train_loss": 0.24293120205402374, "global_step": 45590, "epoch": 512, "lr": 9.332077478991579e-05}
{"train_loss": 0.35733217000961304, "global_step": 45591, "epoch": 512, "lr": 9.332048531075985e-05}
{"train_loss": 0.24745893478393555, "global_step": 45592, "epoch": 512, "lr": 9.332019582578002e-05}
{"train_loss": 0.3036656379699707, "global_step": 45593, "epoch": 512, "lr": 9.33199063349763e-05}
{"train_loss": 0.26058849692344666, "global_step": 45594, "epoch": 512, "lr": 9.331961683834876e-05}
{"train_loss": 0.31673941016197205, "global_step": 45595, "epoch": 512, "lr": 9.331932733589741e-05}
{"train_loss": 0.23026663064956665, "global_step": 45596, "epoch": 512, "lr": 9.331903782762232e-05}
{"train_loss": 0.2833946943283081, "global_step": 45597, "epoch": 512, "lr": 9.331874831352351e-05}
{"train_loss": 0.2525395154953003, "global_step": 45598, "epoch": 512, "lr": 9.331845879360101e-05}
{"train_loss": 0.38630029559135437, "global_step": 45599, "epoch": 512, "lr": 9.331816926785488e-05}
{"train_loss": 0.3431377410888672, "global_step": 45600, "epoch": 512, "lr": 9.331787973628516e-05}
{"train_loss": 0.26525095105171204, "global_step": 45601, "epoch": 512, "lr": 9.331759019889186e-05}
{"train_loss": 0.28528541326522827, "global_step": 45602, "epoch": 512, "lr": 9.331730065567505e-05}
{"train_loss": 0.3447728157043457, "global_step": 45603, "epoch": 512, "lr": 9.331701110663476e-05}
{"train_loss": 0.2500945031642914, "global_step": 45604, "epoch": 512, "lr": 9.331672155177102e-05}
{"train_loss": 0.2300650179386139, "global_step": 45605, "epoch": 512, "lr": 9.331643199108388e-05}
{"train_loss": 0.3059978187084198, "global_step": 45606, "epoch": 512, "lr": 9.331614242457337e-05}
{"train_loss": 0.3306143581867218, "global_step": 45607, "epoch": 512, "lr": 9.331585285223954e-05}
{"train_loss": 0.3101017475128174, "global_step": 45608, "epoch": 512, "lr": 9.33155632740824e-05}
{"train_loss": 0.4489189684391022, "global_step": 45609, "epoch": 512, "lr": 9.331527369010202e-05}
{"train_loss": 0.3761633336544037, "global_step": 45610, "epoch": 512, "lr": 9.331498410029844e-05}
{"train_loss": 0.30919212102890015, "global_step": 45611, "epoch": 512, "lr": 9.331469450467169e-05}
{"train_loss": 0.4350089132785797, "global_step": 45612, "epoch": 512, "lr": 9.33144049032218e-05}
{"train_loss": 0.2707696855068207, "global_step": 45613, "epoch": 512, "lr": 9.331411529594881e-05}
{"train_loss": 0.28568875789642334, "global_step": 45614, "epoch": 512, "lr": 9.331382568285277e-05}
{"train_loss": 0.3237653374671936, "global_step": 45615, "epoch": 512, "lr": 9.331353606393371e-05}
{"train_loss": 0.2981738746166229, "global_step": 45616, "epoch": 512, "lr": 9.331324643919168e-05}
{"train_loss": 0.30440691113471985, "global_step": 45617, "epoch": 512, "lr": 9.33129568086267e-05}
{"train_loss": 0.1939345896244049, "global_step": 45618, "epoch": 512, "lr": 9.331266717223885e-05}
{"train_loss": 0.3562127351760864, "global_step": 45619, "epoch": 512, "lr": 9.33123775300281e-05}
{"train_loss": 0.2938117980957031, "global_step": 45620, "epoch": 512, "lr": 9.331208788199456e-05}
{"train_loss": 0.31858471035957336, "global_step": 45621, "epoch": 512, "lr": 9.331179822813823e-05}
{"train_loss": 0.2696745991706848, "global_step": 45622, "epoch": 512, "lr": 9.331150856845914e-05}
{"train_loss": 0.22579315304756165, "global_step": 45623, "epoch": 512, "lr": 9.331121890295738e-05}
{"train_loss": 0.3720502257347107, "global_step": 45624, "epoch": 512, "lr": 9.331092923163291e-05}
{"train_loss": 0.28007593750953674, "global_step": 45625, "epoch": 512, "lr": 9.331063955448584e-05}
{"train_loss": 0.3180371820926666, "global_step": 45626, "epoch": 512, "lr": 9.331034987151618e-05}
{"train_loss": 0.44390878081321716, "global_step": 45627, "epoch": 512, "lr": 9.331006018272398e-05}
{"train_loss": 0.39972418546676636, "global_step": 45628, "epoch": 512, "lr": 9.330977048810927e-05}
{"train_loss": 0.28033366799354553, "global_step": 45629, "epoch": 512, "lr": 9.330948078767208e-05}
{"train_loss": 0.35777899622917175, "global_step": 45630, "epoch": 512, "lr": 9.330919108141245e-05}
{"train_loss": 0.26872801780700684, "global_step": 45631, "epoch": 512, "lr": 9.330890136933043e-05}
{"train_loss": 0.27008143067359924, "global_step": 45632, "epoch": 512, "lr": 9.330861165142606e-05}
{"train_loss": 0.2874421179294586, "global_step": 45633, "epoch": 512, "lr": 9.330832192769938e-05}
{"train_loss": 0.3417806029319763, "global_step": 45634, "epoch": 512, "lr": 9.330803219815042e-05}
{"train_loss": 0.2804587781429291, "global_step": 45635, "epoch": 512, "lr": 9.330774246277922e-05}
{"train_loss": 0.400672972202301, "global_step": 45636, "epoch": 512, "lr": 9.330745272158583e-05}
{"train_loss": 0.3437623083591461, "global_step": 45637, "epoch": 512, "lr": 9.330716297457027e-05}
{"train_loss": 0.24376726150512695, "global_step": 45638, "epoch": 512, "lr": 9.33068732217326e-05}
{"train_loss": 0.25487905740737915, "global_step": 45639, "epoch": 512, "lr": 9.330658346307284e-05}
{"train_loss": 0.27841103076934814, "global_step": 45640, "epoch": 512, "lr": 9.330629369859105e-05}
{"train_loss": 0.29936906695365906, "global_step": 45641, "epoch": 512, "lr": 9.330600392828724e-05}
{"train_loss": 0.228213831782341, "global_step": 45642, "epoch": 512, "lr": 9.330571415216147e-05}
{"train_loss": 0.3201235234737396, "global_step": 45643, "epoch": 512, "lr": 9.330542437021379e-05}
{"train_loss": 0.3133930265903473, "global_step": 45644, "epoch": 512, "lr": 9.33051345824442e-05}
{"train_loss": 0.3123619556427002, "global_step": 45645, "epoch": 512, "lr": 9.330484478885278e-05}
{"train_loss": 0.30851298570632935, "global_step": 45646, "epoch": 512, "lr": 9.330455498943955e-05}
{"train_loss": 0.26602721214294434, "global_step": 45647, "epoch": 512, "lr": 9.330426518420454e-05}
{"train_loss": 0.28037208318710327, "global_step": 45648, "epoch": 512, "lr": 9.33039753731478e-05}
{"train_loss": 0.26457101106643677, "global_step": 45649, "epoch": 512, "lr": 9.330368555626938e-05}
{"train_loss": 0.33039677143096924, "global_step": 45650, "epoch": 512, "lr": 9.330339573356931e-05}
{"train_loss": 0.2522601783275604, "global_step": 45651, "epoch": 512, "lr": 9.330310590504761e-05}
{"train_loss": 0.3391976058483124, "global_step": 45652, "epoch": 512, "lr": 9.330281607070435e-05}
{"train_loss": 0.35438403487205505, "global_step": 45653, "epoch": 512, "lr": 9.330252623053954e-05}
{"train_loss": 0.27889007329940796, "global_step": 45654, "epoch": 512, "lr": 9.330223638455324e-05}
{"train_loss": 0.27011385560035706, "global_step": 45655, "epoch": 512, "lr": 9.330194653274549e-05}
{"train_loss": 0.3054755407102992, "global_step": 45656, "epoch": 512, "lr": 9.330165667511631e-05, "val_loss": 2.8598384857177734}
{"train_loss": 0.2636721432209015, "global_step": 45657, "epoch": 513, "lr": 9.330136681166575e-05}
{"train_loss": 0.22304105758666992, "global_step": 45658, "epoch": 513, "lr": 9.330107694239386e-05}
{"train_loss": 0.2897893786430359, "global_step": 45659, "epoch": 513, "lr": 9.330078706730066e-05}
{"train_loss": 0.4054826498031616, "global_step": 45660, "epoch": 513, "lr": 9.330049718638621e-05}
{"train_loss": 0.3029921352863312, "global_step": 45661, "epoch": 513, "lr": 9.330020729965051e-05}
{"train_loss": 0.3406504988670349, "global_step": 45662, "epoch": 513, "lr": 9.329991740709363e-05}
{"train_loss": 0.25701841711997986, "global_step": 45663, "epoch": 513, "lr": 9.329962750871562e-05}
{"train_loss": 0.2908482253551483, "global_step": 45664, "epoch": 513, "lr": 9.32993376045165e-05}
{"train_loss": 0.27233293652534485, "global_step": 45665, "epoch": 513, "lr": 9.32990476944963e-05}
{"train_loss": 0.2059854418039322, "global_step": 45666, "epoch": 513, "lr": 9.329875777865509e-05}
{"train_loss": 0.4076169729232788, "global_step": 45667, "epoch": 513, "lr": 9.329846785699287e-05}
{"train_loss": 0.28600820899009705, "global_step": 45668, "epoch": 513, "lr": 9.32981779295097e-05}
{"train_loss": 0.24092505872249603, "global_step": 45669, "epoch": 513, "lr": 9.329788799620562e-05}
{"train_loss": 0.28728801012039185, "global_step": 45670, "epoch": 513, "lr": 9.329759805708068e-05}
{"train_loss": 0.25125089287757874, "global_step": 45671, "epoch": 513, "lr": 9.329730811213488e-05}
{"train_loss": 0.3552858531475067, "global_step": 45672, "epoch": 513, "lr": 9.329701816136831e-05}
{"train_loss": 0.32108134031295776, "global_step": 45673, "epoch": 513, "lr": 9.329672820478098e-05}
{"train_loss": 0.22914926707744598, "global_step": 45674, "epoch": 513, "lr": 9.329643824237293e-05}
{"train_loss": 0.3179683983325958, "global_step": 45675, "epoch": 513, "lr": 9.329614827414419e-05}
{"train_loss": 0.3389124274253845, "global_step": 45676, "epoch": 513, "lr": 9.329585830009482e-05}
{"train_loss": 0.2429652214050293, "global_step": 45677, "epoch": 513, "lr": 9.329556832022486e-05}
{"train_loss": 0.3391954302787781, "global_step": 45678, "epoch": 513, "lr": 9.329527833453431e-05}
{"train_loss": 0.3878937065601349, "global_step": 45679, "epoch": 513, "lr": 9.329498834302326e-05}
{"train_loss": 0.24071693420410156, "global_step": 45680, "epoch": 513, "lr": 9.329469834569172e-05}
{"train_loss": 0.38606348633766174, "global_step": 45681, "epoch": 513, "lr": 9.329440834253975e-05}
{"train_loss": 0.3381090462207794, "global_step": 45682, "epoch": 513, "lr": 9.329411833356734e-05}
{"train_loss": 0.32449841499328613, "global_step": 45683, "epoch": 513, "lr": 9.32938283187746e-05}
{"train_loss": 0.30327656865119934, "global_step": 45684, "epoch": 513, "lr": 9.329353829816151e-05}
{"train_loss": 0.3244170844554901, "global_step": 45685, "epoch": 513, "lr": 9.329324827172814e-05}
{"train_loss": 0.37826794385910034, "global_step": 45686, "epoch": 513, "lr": 9.329295823947453e-05}
{"train_loss": 0.36182618141174316, "global_step": 45687, "epoch": 513, "lr": 9.329266820140069e-05}
{"train_loss": 0.2607742249965668, "global_step": 45688, "epoch": 513, "lr": 9.329237815750669e-05}
{"train_loss": 0.32397568225860596, "global_step": 45689, "epoch": 513, "lr": 9.329208810779255e-05}
{"train_loss": 0.28534674644470215, "global_step": 45690, "epoch": 513, "lr": 9.329179805225832e-05}
{"train_loss": 0.43821585178375244, "global_step": 45691, "epoch": 513, "lr": 9.329150799090403e-05}
{"train_loss": 0.26383066177368164, "global_step": 45692, "epoch": 513, "lr": 9.329121792372973e-05}
{"train_loss": 0.30409786105155945, "global_step": 45693, "epoch": 513, "lr": 9.329092785073545e-05}
{"train_loss": 0.27386564016342163, "global_step": 45694, "epoch": 513, "lr": 9.329063777192124e-05}
{"train_loss": 0.34622442722320557, "global_step": 45695, "epoch": 513, "lr": 9.329034768728713e-05}
{"train_loss": 0.29856935143470764, "global_step": 45696, "epoch": 513, "lr": 9.329005759683316e-05}
{"train_loss": 0.3589765727519989, "global_step": 45697, "epoch": 513, "lr": 9.328976750055937e-05}
{"train_loss": 0.28812554478645325, "global_step": 45698, "epoch": 513, "lr": 9.328947739846579e-05}
{"train_loss": 0.3110637664794922, "global_step": 45699, "epoch": 513, "lr": 9.328918729055247e-05}
{"train_loss": 0.3048659563064575, "global_step": 45700, "epoch": 513, "lr": 9.328889717681945e-05}
{"train_loss": 0.28934624791145325, "global_step": 45701, "epoch": 513, "lr": 9.328860705726678e-05}
{"train_loss": 0.3749719560146332, "global_step": 45702, "epoch": 513, "lr": 9.328831693189446e-05}
{"train_loss": 0.26128920912742615, "global_step": 45703, "epoch": 513, "lr": 9.328802680070257e-05}
{"train_loss": 0.2924235165119171, "global_step": 45704, "epoch": 513, "lr": 9.328773666369112e-05}
{"train_loss": 0.27743715047836304, "global_step": 45705, "epoch": 513, "lr": 9.328744652086017e-05}
{"train_loss": 0.29887524247169495, "global_step": 45706, "epoch": 513, "lr": 9.328715637220975e-05}
{"train_loss": 0.221368208527565, "global_step": 45707, "epoch": 513, "lr": 9.32868662177399e-05}
{"train_loss": 0.42894312739372253, "global_step": 45708, "epoch": 513, "lr": 9.328657605745064e-05}
{"train_loss": 0.30787160992622375, "global_step": 45709, "epoch": 513, "lr": 9.328628589134205e-05}
{"train_loss": 0.25896936655044556, "global_step": 45710, "epoch": 513, "lr": 9.328599571941415e-05}
{"train_loss": 0.28596803545951843, "global_step": 45711, "epoch": 513, "lr": 9.328570554166696e-05}
{"train_loss": 0.3461250960826874, "global_step": 45712, "epoch": 513, "lr": 9.328541535810053e-05}
{"train_loss": 0.3387472331523895, "global_step": 45713, "epoch": 513, "lr": 9.328512516871491e-05}
{"train_loss": 0.32977479696273804, "global_step": 45714, "epoch": 513, "lr": 9.328483497351014e-05}
{"train_loss": 0.37409457564353943, "global_step": 45715, "epoch": 513, "lr": 9.328454477248625e-05}
{"train_loss": 0.44517797231674194, "global_step": 45716, "epoch": 513, "lr": 9.328425456564327e-05}
{"train_loss": 0.3972252607345581, "global_step": 45717, "epoch": 513, "lr": 9.328396435298125e-05}
{"train_loss": 0.22229358553886414, "global_step": 45718, "epoch": 513, "lr": 9.328367413450024e-05}
{"train_loss": 0.23979130387306213, "global_step": 45719, "epoch": 513, "lr": 9.328338391020026e-05}
{"train_loss": 0.3354354798793793, "global_step": 45720, "epoch": 513, "lr": 9.328309368008137e-05}
{"train_loss": 0.2739384174346924, "global_step": 45721, "epoch": 513, "lr": 9.328280344414358e-05}
{"train_loss": 0.34058448672294617, "global_step": 45722, "epoch": 513, "lr": 9.328251320238695e-05}
{"train_loss": 0.24269139766693115, "global_step": 45723, "epoch": 513, "lr": 9.32822229548115e-05}
{"train_loss": 0.3076048195362091, "global_step": 45724, "epoch": 513, "lr": 9.328193270141731e-05}
{"train_loss": 0.2788735628128052, "global_step": 45725, "epoch": 513, "lr": 9.328164244220438e-05}
{"train_loss": 0.2437593787908554, "global_step": 45726, "epoch": 513, "lr": 9.328135217717275e-05}
{"train_loss": 0.23506709933280945, "global_step": 45727, "epoch": 513, "lr": 9.328106190632248e-05}
{"train_loss": 0.3408960998058319, "global_step": 45728, "epoch": 513, "lr": 9.328077162965358e-05}
{"train_loss": 0.22961978614330292, "global_step": 45729, "epoch": 513, "lr": 9.328048134716613e-05}
{"train_loss": 0.31755876541137695, "global_step": 45730, "epoch": 513, "lr": 9.328019105886014e-05}
{"train_loss": 0.3154550790786743, "global_step": 45731, "epoch": 513, "lr": 9.327990076473567e-05}
{"train_loss": 0.3865828812122345, "global_step": 45732, "epoch": 513, "lr": 9.327961046479272e-05}
{"train_loss": 0.42171820998191833, "global_step": 45733, "epoch": 513, "lr": 9.327932015903138e-05}
{"train_loss": 0.222355455160141, "global_step": 45734, "epoch": 513, "lr": 9.327902984745164e-05}
{"train_loss": 0.29810619354248047, "global_step": 45735, "epoch": 513, "lr": 9.327873953005357e-05}
{"train_loss": 0.23299548029899597, "global_step": 45736, "epoch": 513, "lr": 9.32784492068372e-05}
{"train_loss": 0.26163944602012634, "global_step": 45737, "epoch": 513, "lr": 9.327815887780258e-05}
{"train_loss": 0.2827444076538086, "global_step": 45738, "epoch": 513, "lr": 9.327786854294971e-05}
{"train_loss": 0.38900548219680786, "global_step": 45739, "epoch": 513, "lr": 9.32775782022787e-05}
{"train_loss": 0.37115663290023804, "global_step": 45740, "epoch": 513, "lr": 9.32772878557895e-05}
{"train_loss": 0.29269903898239136, "global_step": 45741, "epoch": 513, "lr": 9.327699750348225e-05}
{"train_loss": 0.34791797399520874, "global_step": 45742, "epoch": 513, "lr": 9.32767071453569e-05}
{"train_loss": 0.2947997748851776, "global_step": 45743, "epoch": 513, "lr": 9.327641678141352e-05}
{"train_loss": 0.3269484341144562, "global_step": 45744, "epoch": 513, "lr": 9.327612641165218e-05}
{"train_loss": 0.3089563461501947, "global_step": 45745, "epoch": 513, "lr": 9.327583603607287e-05, "val_loss": 2.8494513034820557}
{"train_loss": 0.31384944915771484, "global_step": 45746, "epoch": 514, "lr": 9.327554565467565e-05}
{"train_loss": 0.2465052455663681, "global_step": 45747, "epoch": 514, "lr": 9.327525526746058e-05}
{"train_loss": 0.24192510545253754, "global_step": 45748, "epoch": 514, "lr": 9.327496487442766e-05}
{"train_loss": 0.24209272861480713, "global_step": 45749, "epoch": 514, "lr": 9.327467447557697e-05}
{"train_loss": 0.3140809237957001, "global_step": 45750, "epoch": 514, "lr": 9.327438407090849e-05}
{"train_loss": 0.25900158286094666, "global_step": 45751, "epoch": 514, "lr": 9.327409366042233e-05}
{"train_loss": 0.3437747359275818, "global_step": 45752, "epoch": 514, "lr": 9.327380324411848e-05}
{"train_loss": 0.31321442127227783, "global_step": 45753, "epoch": 514, "lr": 9.3273512821997e-05}
{"train_loss": 0.2774336040019989, "global_step": 45754, "epoch": 514, "lr": 9.327322239405792e-05}
{"train_loss": 0.23053570091724396, "global_step": 45755, "epoch": 514, "lr": 9.327293196030129e-05}
{"train_loss": 0.30724477767944336, "global_step": 45756, "epoch": 514, "lr": 9.327264152072715e-05}
{"train_loss": 0.30544692277908325, "global_step": 45757, "epoch": 514, "lr": 9.327235107533551e-05}
{"train_loss": 0.22656795382499695, "global_step": 45758, "epoch": 514, "lr": 9.327206062412643e-05}
{"train_loss": 0.3838762044906616, "global_step": 45759, "epoch": 514, "lr": 9.327177016709996e-05}
{"train_loss": 0.2858991324901581, "global_step": 45760, "epoch": 514, "lr": 9.327147970425613e-05}
{"train_loss": 0.2588334083557129, "global_step": 45761, "epoch": 514, "lr": 9.327118923559496e-05}
{"train_loss": 0.2844122350215912, "global_step": 45762, "epoch": 514, "lr": 9.327089876111653e-05}
{"train_loss": 0.38229113817214966, "global_step": 45763, "epoch": 514, "lr": 9.327060828082083e-05}
{"train_loss": 0.26638442277908325, "global_step": 45764, "epoch": 514, "lr": 9.327031779470795e-05}
{"train_loss": 0.3461220860481262, "global_step": 45765, "epoch": 514, "lr": 9.32700273027779e-05}
{"train_loss": 0.2799585461616516, "global_step": 45766, "epoch": 514, "lr": 9.32697368050307e-05}
{"train_loss": 0.2923276424407959, "global_step": 45767, "epoch": 514, "lr": 9.326944630146644e-05}
{"train_loss": 0.2539726793766022, "global_step": 45768, "epoch": 514, "lr": 9.32691557920851e-05}
{"train_loss": 0.3214481770992279, "global_step": 45769, "epoch": 514, "lr": 9.326886527688676e-05}
{"train_loss": 0.35684362053871155, "global_step": 45770, "epoch": 514, "lr": 9.326857475587145e-05}
{"train_loss": 0.18029722571372986, "global_step": 45771, "epoch": 514, "lr": 9.326828422903922e-05}
{"train_loss": 0.32371288537979126, "global_step": 45772, "epoch": 514, "lr": 9.326799369639008e-05}
{"train_loss": 0.2967525124549866, "global_step": 45773, "epoch": 514, "lr": 9.32677031579241e-05}
{"train_loss": 0.26334384083747864, "global_step": 45774, "epoch": 514, "lr": 9.326741261364129e-05}
{"train_loss": 0.2884668707847595, "global_step": 45775, "epoch": 514, "lr": 9.326712206354172e-05}
{"train_loss": 0.24954012036323547, "global_step": 45776, "epoch": 514, "lr": 9.326683150762539e-05}
{"train_loss": 0.25786325335502625, "global_step": 45777, "epoch": 514, "lr": 9.326654094589238e-05}
{"train_loss": 0.28753045201301575, "global_step": 45778, "epoch": 514, "lr": 9.326625037834272e-05}
{"train_loss": 0.25997015833854675, "global_step": 45779, "epoch": 514, "lr": 9.32659598049764e-05}
{"train_loss": 0.273934006690979, "global_step": 45780, "epoch": 514, "lr": 9.326566922579354e-05}
{"train_loss": 0.2293863445520401, "global_step": 45781, "epoch": 514, "lr": 9.326537864079412e-05}
{"train_loss": 0.2779230773448944, "global_step": 45782, "epoch": 514, "lr": 9.32650880499782e-05}
{"train_loss": 0.3869621157646179, "global_step": 45783, "epoch": 514, "lr": 9.326479745334582e-05}
{"train_loss": 0.31187447905540466, "global_step": 45784, "epoch": 514, "lr": 9.326450685089701e-05}
{"train_loss": 0.21114715933799744, "global_step": 45785, "epoch": 514, "lr": 9.326421624263181e-05}
{"train_loss": 0.26472046971321106, "global_step": 45786, "epoch": 514, "lr": 9.326392562855028e-05}
{"train_loss": 0.3687528371810913, "global_step": 45787, "epoch": 514, "lr": 9.326363500865242e-05}
{"train_loss": 0.3239981532096863, "global_step": 45788, "epoch": 514, "lr": 9.326334438293831e-05}
{"train_loss": 0.2862008213996887, "global_step": 45789, "epoch": 514, "lr": 9.326305375140795e-05}
{"train_loss": 0.30109351873397827, "global_step": 45790, "epoch": 514, "lr": 9.326276311406143e-05}
{"train_loss": 0.2691090703010559, "global_step": 45791, "epoch": 514, "lr": 9.326247247089874e-05}
{"train_loss": 0.3512822389602661, "global_step": 45792, "epoch": 514, "lr": 9.326218182191995e-05}
{"train_loss": 0.29578834772109985, "global_step": 45793, "epoch": 514, "lr": 9.326189116712507e-05}
{"train_loss": 0.24549683928489685, "global_step": 45794, "epoch": 514, "lr": 9.326160050651416e-05}
{"train_loss": 0.28504911065101624, "global_step": 45795, "epoch": 514, "lr": 9.326130984008726e-05}
{"train_loss": 0.2407914400100708, "global_step": 45796, "epoch": 514, "lr": 9.32610191678444e-05}
{"train_loss": 0.3220979869365692, "global_step": 45797, "epoch": 514, "lr": 9.326072848978564e-05}
{"train_loss": 0.33199799060821533, "global_step": 45798, "epoch": 514, "lr": 9.326043780591099e-05}
{"train_loss": 0.2552359104156494, "global_step": 45799, "epoch": 514, "lr": 9.326014711622049e-05}
{"train_loss": 0.2982543706893921, "global_step": 45800, "epoch": 514, "lr": 9.32598564207142e-05}
{"train_loss": 0.22850705683231354, "global_step": 45801, "epoch": 514, "lr": 9.325956571939214e-05}
{"train_loss": 0.2405126988887787, "global_step": 45802, "epoch": 514, "lr": 9.325927501225438e-05}
{"train_loss": 0.2961606979370117, "global_step": 45803, "epoch": 514, "lr": 9.325898429930093e-05}
{"train_loss": 0.23200452327728271, "global_step": 45804, "epoch": 514, "lr": 9.325869358053183e-05}
{"train_loss": 0.2883584499359131, "global_step": 45805, "epoch": 514, "lr": 9.325840285594712e-05}
{"train_loss": 0.30520111322402954, "global_step": 45806, "epoch": 514, "lr": 9.325811212554686e-05}
{"train_loss": 0.27082502841949463, "global_step": 45807, "epoch": 514, "lr": 9.325782138933107e-05}
{"train_loss": 0.28875476121902466, "global_step": 45808, "epoch": 514, "lr": 9.325753064729979e-05}
{"train_loss": 0.380607545375824, "global_step": 45809, "epoch": 514, "lr": 9.325723989945306e-05}
{"train_loss": 0.270285040140152, "global_step": 45810, "epoch": 514, "lr": 9.325694914579092e-05}
{"train_loss": 0.2916561961174011, "global_step": 45811, "epoch": 514, "lr": 9.325665838631342e-05}
{"train_loss": 0.304660826921463, "global_step": 45812, "epoch": 514, "lr": 9.32563676210206e-05}
{"train_loss": 0.29523634910583496, "global_step": 45813, "epoch": 514, "lr": 9.325607684991246e-05}
{"train_loss": 0.35041019320487976, "global_step": 45814, "epoch": 514, "lr": 9.325578607298908e-05}
{"train_loss": 0.29214969277381897, "global_step": 45815, "epoch": 514, "lr": 9.325549529025048e-05}
{"train_loss": 0.22308871150016785, "global_step": 45816, "epoch": 514, "lr": 9.325520450169672e-05}
{"train_loss": 0.34379443526268005, "global_step": 45817, "epoch": 514, "lr": 9.325491370732781e-05}
{"train_loss": 0.26011666655540466, "global_step": 45818, "epoch": 514, "lr": 9.325462290714382e-05}
{"train_loss": 0.29900142550468445, "global_step": 45819, "epoch": 514, "lr": 9.325433210114477e-05}
{"train_loss": 0.25020766258239746, "global_step": 45820, "epoch": 514, "lr": 9.32540412893307e-05}
{"train_loss": 0.33475667238235474, "global_step": 45821, "epoch": 514, "lr": 9.325375047170165e-05}
{"train_loss": 0.20764759182929993, "global_step": 45822, "epoch": 514, "lr": 9.325345964825766e-05}
{"train_loss": 0.3627139925956726, "global_step": 45823, "epoch": 514, "lr": 9.325316881899878e-05}
{"train_loss": 0.2646084427833557, "global_step": 45824, "epoch": 514, "lr": 9.325287798392502e-05}
{"train_loss": 0.36091136932373047, "global_step": 45825, "epoch": 514, "lr": 9.325258714303644e-05}
{"train_loss": 0.3262605667114258, "global_step": 45826, "epoch": 514, "lr": 9.325229629633307e-05}
{"train_loss": 0.3102231025695801, "global_step": 45827, "epoch": 514, "lr": 9.325200544381499e-05}
{"train_loss": 0.2219347506761551, "global_step": 45828, "epoch": 514, "lr": 9.325171458548217e-05}
{"train_loss": 0.19497033953666687, "global_step": 45829, "epoch": 514, "lr": 9.32514237213347e-05}
{"train_loss": 0.24605874717235565, "global_step": 45830, "epoch": 514, "lr": 9.325113285137261e-05}
{"train_loss": 0.25508856773376465, "global_step": 45831, "epoch": 514, "lr": 9.325084197559592e-05}
{"train_loss": 0.26455894112586975, "global_step": 45832, "epoch": 514, "lr": 9.325055109400469e-05}
{"train_loss": 0.2762434780597687, "global_step": 45833, "epoch": 514, "lr": 9.325026020659893e-05}
{"train_loss": 0.2859709102116274, "global_step": 45834, "epoch": 514, "lr": 9.324996931337872e-05, "val_loss": 2.951582431793213}
{"train_loss": 0.24945029616355896, "global_step": 45835, "epoch": 515, "lr": 9.324967841434407e-05}
{"train_loss": 0.32195499539375305, "global_step": 45836, "epoch": 515, "lr": 9.324938750949503e-05}
{"train_loss": 0.3328717052936554, "global_step": 45837, "epoch": 515, "lr": 9.324909659883165e-05}
{"train_loss": 0.39710748195648193, "global_step": 45838, "epoch": 515, "lr": 9.324880568235395e-05}
{"train_loss": 0.31322744488716125, "global_step": 45839, "epoch": 515, "lr": 9.324851476006195e-05}
{"train_loss": 0.23547987639904022, "global_step": 45840, "epoch": 515, "lr": 9.324822383195575e-05}
{"train_loss": 0.24051858484745026, "global_step": 45841, "epoch": 515, "lr": 9.324793289803533e-05}
{"train_loss": 0.27708786725997925, "global_step": 45842, "epoch": 515, "lr": 9.324764195830076e-05}
{"train_loss": 0.2988641858100891, "global_step": 45843, "epoch": 515, "lr": 9.324735101275208e-05}
{"train_loss": 0.23726876080036163, "global_step": 45844, "epoch": 515, "lr": 9.32470600613893e-05}
{"train_loss": 0.32390308380126953, "global_step": 45845, "epoch": 515, "lr": 9.324676910421251e-05}
{"train_loss": 0.2706094980239868, "global_step": 45846, "epoch": 515, "lr": 9.324647814122169e-05}
{"train_loss": 0.34548744559288025, "global_step": 45847, "epoch": 515, "lr": 9.324618717241692e-05}
{"train_loss": 0.20699141919612885, "global_step": 45848, "epoch": 515, "lr": 9.324589619779822e-05}
{"train_loss": 0.2244848608970642, "global_step": 45849, "epoch": 515, "lr": 9.324560521736565e-05}
{"train_loss": 0.15345145761966705, "global_step": 45850, "epoch": 515, "lr": 9.324531423111923e-05}
{"train_loss": 0.3504042625427246, "global_step": 45851, "epoch": 515, "lr": 9.3245023239059e-05}
{"train_loss": 0.31381818652153015, "global_step": 45852, "epoch": 515, "lr": 9.324473224118502e-05}
{"train_loss": 0.16409924626350403, "global_step": 45853, "epoch": 515, "lr": 9.32444412374973e-05}
{"train_loss": 0.37346747517585754, "global_step": 45854, "epoch": 515, "lr": 9.324415022799589e-05}
{"train_loss": 0.2709403932094574, "global_step": 45855, "epoch": 515, "lr": 9.324385921268082e-05}
{"train_loss": 0.32423216104507446, "global_step": 45856, "epoch": 515, "lr": 9.324356819155216e-05}
{"train_loss": 0.25466105341911316, "global_step": 45857, "epoch": 515, "lr": 9.324327716460992e-05}
{"train_loss": 0.37911370396614075, "global_step": 45858, "epoch": 515, "lr": 9.324298613185415e-05}
{"train_loss": 0.26770350337028503, "global_step": 45859, "epoch": 515, "lr": 9.324269509328489e-05}
{"train_loss": 0.41281798481941223, "global_step": 45860, "epoch": 515, "lr": 9.324240404890216e-05}
{"train_loss": 0.42467039823532104, "global_step": 45861, "epoch": 515, "lr": 9.324211299870604e-05}
{"train_loss": 0.3002767264842987, "global_step": 45862, "epoch": 515, "lr": 9.324182194269653e-05}
{"train_loss": 0.31117409467697144, "global_step": 45863, "epoch": 515, "lr": 9.32415308808737e-05}
{"train_loss": 0.2035072147846222, "global_step": 45864, "epoch": 515, "lr": 9.324123981323756e-05}
{"train_loss": 0.2323092520236969, "global_step": 45865, "epoch": 515, "lr": 9.324094873978817e-05}
{"train_loss": 0.37967953085899353, "global_step": 45866, "epoch": 515, "lr": 9.324065766052556e-05}
{"train_loss": 0.21006643772125244, "global_step": 45867, "epoch": 515, "lr": 9.324036657544977e-05}
{"train_loss": 0.2597652077674866, "global_step": 45868, "epoch": 515, "lr": 9.324007548456083e-05}
{"train_loss": 0.3018779456615448, "global_step": 45869, "epoch": 515, "lr": 9.32397843878588e-05}
{"train_loss": 0.31176474690437317, "global_step": 45870, "epoch": 515, "lr": 9.323949328534369e-05}
{"train_loss": 0.23807291686534882, "global_step": 45871, "epoch": 515, "lr": 9.323920217701559e-05}
{"train_loss": 0.26670175790786743, "global_step": 45872, "epoch": 515, "lr": 9.323891106287448e-05}
{"train_loss": 0.27533334493637085, "global_step": 45873, "epoch": 515, "lr": 9.323861994292042e-05}
{"train_loss": 0.3003624379634857, "global_step": 45874, "epoch": 515, "lr": 9.323832881715348e-05}
{"train_loss": 0.27183836698532104, "global_step": 45875, "epoch": 515, "lr": 9.323803768557366e-05}
{"train_loss": 0.34860703349113464, "global_step": 45876, "epoch": 515, "lr": 9.323774654818101e-05}
{"train_loss": 0.24569904804229736, "global_step": 45877, "epoch": 515, "lr": 9.323745540497559e-05}
{"train_loss": 0.4047139286994934, "global_step": 45878, "epoch": 515, "lr": 9.32371642559574e-05}
{"train_loss": 0.24984890222549438, "global_step": 45879, "epoch": 515, "lr": 9.32368731011265e-05}
{"train_loss": 0.24193012714385986, "global_step": 45880, "epoch": 515, "lr": 9.323658194048294e-05}
{"train_loss": 0.26784369349479675, "global_step": 45881, "epoch": 515, "lr": 9.323629077402675e-05}
{"train_loss": 0.2510756552219391, "global_step": 45882, "epoch": 515, "lr": 9.323599960175795e-05}
{"train_loss": 0.305213063955307, "global_step": 45883, "epoch": 515, "lr": 9.323570842367661e-05}
{"train_loss": 0.29545387625694275, "global_step": 45884, "epoch": 515, "lr": 9.323541723978276e-05}
{"train_loss": 0.3209596574306488, "global_step": 45885, "epoch": 515, "lr": 9.323512605007643e-05}
{"train_loss": 0.224701926112175, "global_step": 45886, "epoch": 515, "lr": 9.323483485455767e-05}
{"train_loss": 0.35971999168395996, "global_step": 45887, "epoch": 515, "lr": 9.323454365322652e-05}
{"train_loss": 0.26499903202056885, "global_step": 45888, "epoch": 515, "lr": 9.3234252446083e-05}
{"train_loss": 0.27051717042922974, "global_step": 45889, "epoch": 515, "lr": 9.323396123312716e-05}
{"train_loss": 0.30732426047325134, "global_step": 45890, "epoch": 515, "lr": 9.323367001435904e-05}
{"train_loss": 0.1750924289226532, "global_step": 45891, "epoch": 515, "lr": 9.323337878977868e-05}
{"train_loss": 0.2530820071697235, "global_step": 45892, "epoch": 515, "lr": 9.323308755938613e-05}
{"train_loss": 0.38100501894950867, "global_step": 45893, "epoch": 515, "lr": 9.32327963231814e-05}
{"train_loss": 0.36317282915115356, "global_step": 45894, "epoch": 515, "lr": 9.323250508116457e-05}
{"train_loss": 0.3822923004627228, "global_step": 45895, "epoch": 515, "lr": 9.323221383333565e-05}
{"train_loss": 0.273330956697464, "global_step": 45896, "epoch": 515, "lr": 9.323192257969468e-05}
{"train_loss": 0.2925560176372528, "global_step": 45897, "epoch": 515, "lr": 9.323163132024169e-05}
{"train_loss": 0.31362593173980713, "global_step": 45898, "epoch": 515, "lr": 9.323134005497676e-05}
{"train_loss": 0.21010160446166992, "global_step": 45899, "epoch": 515, "lr": 9.323104878389988e-05}
{"train_loss": 0.2797846496105194, "global_step": 45900, "epoch": 515, "lr": 9.323075750701114e-05}
{"train_loss": 0.19113270938396454, "global_step": 45901, "epoch": 515, "lr": 9.323046622431053e-05}
{"train_loss": 0.2532956898212433, "global_step": 45902, "epoch": 515, "lr": 9.323017493579812e-05}
{"train_loss": 0.2808203399181366, "global_step": 45903, "epoch": 515, "lr": 9.322988364147394e-05}
{"train_loss": 0.31043028831481934, "global_step": 45904, "epoch": 515, "lr": 9.322959234133802e-05}
{"train_loss": 0.2688804864883423, "global_step": 45905, "epoch": 515, "lr": 9.322930103539042e-05}
{"train_loss": 0.2989518344402313, "global_step": 45906, "epoch": 515, "lr": 9.322900972363117e-05}
{"train_loss": 0.4011704623699188, "global_step": 45907, "epoch": 515, "lr": 9.322871840606029e-05}
{"train_loss": 0.2637351453304291, "global_step": 45908, "epoch": 515, "lr": 9.322842708267785e-05}
{"train_loss": 0.36781778931617737, "global_step": 45909, "epoch": 515, "lr": 9.322813575348386e-05}
{"train_loss": 0.24685229361057281, "global_step": 45910, "epoch": 515, "lr": 9.322784441847839e-05}
{"train_loss": 0.24048149585723877, "global_step": 45911, "epoch": 515, "lr": 9.322755307766146e-05}
{"train_loss": 0.29332366585731506, "global_step": 45912, "epoch": 515, "lr": 9.322726173103311e-05}
{"train_loss": 0.3462737202644348, "global_step": 45913, "epoch": 515, "lr": 9.322697037859337e-05}
{"train_loss": 0.29745161533355713, "global_step": 45914, "epoch": 515, "lr": 9.32266790203423e-05}
{"train_loss": 0.2865361273288727, "global_step": 45915, "epoch": 515, "lr": 9.322638765627994e-05}
{"train_loss": 0.23542985320091248, "global_step": 45916, "epoch": 515, "lr": 9.32260962864063e-05}
{"train_loss": 0.21460993587970734, "global_step": 45917, "epoch": 515, "lr": 9.322580491072147e-05}
{"train_loss": 0.3186785876750946, "global_step": 45918, "epoch": 515, "lr": 9.322551352922541e-05}
{"train_loss": 0.30679407715797424, "global_step": 45919, "epoch": 515, "lr": 9.322522214191824e-05}
{"train_loss": 0.35549840331077576, "global_step": 45920, "epoch": 515, "lr": 9.322493074879996e-05}
{"train_loss": 0.2985213100910187, "global_step": 45921, "epoch": 515, "lr": 9.322463934987063e-05}
{"train_loss": 0.26540014147758484, "global_step": 45922, "epoch": 515, "lr": 9.322434794513024e-05}
{"train_loss": 0.2888235459836681, "global_step": 45923, "epoch": 515, "lr": 9.322405653457889e-05, "val_loss": 2.879579544067383, "train_action_mse_error": 22.66867446899414}
{"train_loss": 0.31369391083717346, "global_step": 45924, "epoch": 516, "lr": 9.322376511821657e-05}
{"train_loss": 0.2794353663921356, "global_step": 45925, "epoch": 516, "lr": 9.322347369604337e-05}
{"train_loss": 0.20564191043376923, "global_step": 45926, "epoch": 516, "lr": 9.322318226805929e-05}
{"train_loss": 0.24867187440395355, "global_step": 45927, "epoch": 516, "lr": 9.322289083426437e-05}
{"train_loss": 0.2819229066371918, "global_step": 45928, "epoch": 516, "lr": 9.322259939465865e-05}
{"train_loss": 0.37658315896987915, "global_step": 45929, "epoch": 516, "lr": 9.32223079492422e-05}
{"train_loss": 0.24508914351463318, "global_step": 45930, "epoch": 516, "lr": 9.322201649801503e-05}
{"train_loss": 0.2462710291147232, "global_step": 45931, "epoch": 516, "lr": 9.322172504097719e-05}
{"train_loss": 0.33540984988212585, "global_step": 45932, "epoch": 516, "lr": 9.322143357812872e-05}
{"train_loss": 0.355002760887146, "global_step": 45933, "epoch": 516, "lr": 9.322114210946964e-05}
{"train_loss": 0.388049453496933, "global_step": 45934, "epoch": 516, "lr": 9.322085063500001e-05}
{"train_loss": 0.29786157608032227, "global_step": 45935, "epoch": 516, "lr": 9.322055915471987e-05}
{"train_loss": 0.2794168293476105, "global_step": 45936, "epoch": 516, "lr": 9.322026766862924e-05}
{"train_loss": 0.3538728952407837, "global_step": 45937, "epoch": 516, "lr": 9.321997617672819e-05}
{"train_loss": 0.248135507106781, "global_step": 45938, "epoch": 516, "lr": 9.321968467901672e-05}
{"train_loss": 0.31277593970298767, "global_step": 45939, "epoch": 516, "lr": 9.32193931754949e-05}
{"train_loss": 0.21008658409118652, "global_step": 45940, "epoch": 516, "lr": 9.321910166616277e-05}
{"train_loss": 0.25045737624168396, "global_step": 45941, "epoch": 516, "lr": 9.321881015102034e-05}
{"train_loss": 0.2489880621433258, "global_step": 45942, "epoch": 516, "lr": 9.321851863006767e-05}
{"train_loss": 0.25567686557769775, "global_step": 45943, "epoch": 516, "lr": 9.321822710330482e-05}
{"train_loss": 0.2401028275489807, "global_step": 45944, "epoch": 516, "lr": 9.321793557073178e-05}
{"train_loss": 0.3013674318790436, "global_step": 45945, "epoch": 516, "lr": 9.321764403234862e-05}
{"train_loss": 0.27605581283569336, "global_step": 45946, "epoch": 516, "lr": 9.32173524881554e-05}
{"train_loss": 0.3305375277996063, "global_step": 45947, "epoch": 516, "lr": 9.32170609381521e-05}
{"train_loss": 0.3154709041118622, "global_step": 45948, "epoch": 516, "lr": 9.321676938233881e-05}
{"train_loss": 0.2804805338382721, "global_step": 45949, "epoch": 516, "lr": 9.321647782071554e-05}
{"train_loss": 0.3571944236755371, "global_step": 45950, "epoch": 516, "lr": 9.321618625328235e-05}
{"train_loss": 0.47938016057014465, "global_step": 45951, "epoch": 516, "lr": 9.321589468003928e-05}
{"train_loss": 0.34567734599113464, "global_step": 45952, "epoch": 516, "lr": 9.321560310098634e-05}
{"train_loss": 0.30128008127212524, "global_step": 45953, "epoch": 516, "lr": 9.32153115161236e-05}
{"train_loss": 0.2300001084804535, "global_step": 45954, "epoch": 516, "lr": 9.32150199254511e-05}
{"train_loss": 0.21003207564353943, "global_step": 45955, "epoch": 516, "lr": 9.321472832896885e-05}
{"train_loss": 0.19417288899421692, "global_step": 45956, "epoch": 516, "lr": 9.32144367266769e-05}
{"train_loss": 0.3182482421398163, "global_step": 45957, "epoch": 516, "lr": 9.321414511857531e-05}
{"train_loss": 0.27115175127983093, "global_step": 45958, "epoch": 516, "lr": 9.32138535046641e-05}
{"train_loss": 0.24417367577552795, "global_step": 45959, "epoch": 516, "lr": 9.321356188494332e-05}
{"train_loss": 0.34787115454673767, "global_step": 45960, "epoch": 516, "lr": 9.3213270259413e-05}
{"train_loss": 0.27250272035598755, "global_step": 45961, "epoch": 516, "lr": 9.321297862807318e-05}
{"train_loss": 0.2785465717315674, "global_step": 45962, "epoch": 516, "lr": 9.32126869909239e-05}
{"train_loss": 0.3481750190258026, "global_step": 45963, "epoch": 516, "lr": 9.32123953479652e-05}
{"train_loss": 0.24918903410434723, "global_step": 45964, "epoch": 516, "lr": 9.321210369919713e-05}
{"train_loss": 0.37655505537986755, "global_step": 45965, "epoch": 516, "lr": 9.321181204461971e-05}
{"train_loss": 0.2520698010921478, "global_step": 45966, "epoch": 516, "lr": 9.321152038423299e-05}
{"train_loss": 0.31727272272109985, "global_step": 45967, "epoch": 516, "lr": 9.321122871803701e-05}
{"train_loss": 0.2839140295982361, "global_step": 45968, "epoch": 516, "lr": 9.321093704603182e-05}
{"train_loss": 0.2068445384502411, "global_step": 45969, "epoch": 516, "lr": 9.321064536821741e-05}
{"train_loss": 0.2667264938354492, "global_step": 45970, "epoch": 516, "lr": 9.321035368459388e-05}
{"train_loss": 0.32331183552742004, "global_step": 45971, "epoch": 516, "lr": 9.321006199516125e-05}
{"train_loss": 0.2720222473144531, "global_step": 45972, "epoch": 516, "lr": 9.320977029991955e-05}
{"train_loss": 0.25963956117630005, "global_step": 45973, "epoch": 516, "lr": 9.320947859886881e-05}
{"train_loss": 0.3580794334411621, "global_step": 45974, "epoch": 516, "lr": 9.320918689200909e-05}
{"train_loss": 0.34284815192222595, "global_step": 45975, "epoch": 516, "lr": 9.320889517934042e-05}
{"train_loss": 0.3016287386417389, "global_step": 45976, "epoch": 516, "lr": 9.320860346086286e-05}
{"train_loss": 0.2692539691925049, "global_step": 45977, "epoch": 516, "lr": 9.32083117365764e-05}
{"train_loss": 0.3053473234176636, "global_step": 45978, "epoch": 516, "lr": 9.320802000648112e-05}
{"train_loss": 0.4739699065685272, "global_step": 45979, "epoch": 516, "lr": 9.320772827057707e-05}
{"train_loss": 0.2945733964443207, "global_step": 45980, "epoch": 516, "lr": 9.320743652886423e-05}
{"train_loss": 0.2923329174518585, "global_step": 45981, "epoch": 516, "lr": 9.320714478134271e-05}
{"train_loss": 0.32159408926963806, "global_step": 45982, "epoch": 516, "lr": 9.32068530280125e-05}
{"train_loss": 0.2873845398426056, "global_step": 45983, "epoch": 516, "lr": 9.320656126887366e-05}
{"train_loss": 0.274649441242218, "global_step": 45984, "epoch": 516, "lr": 9.320626950392622e-05}
{"train_loss": 0.37283438444137573, "global_step": 45985, "epoch": 516, "lr": 9.320597773317021e-05}
{"train_loss": 0.2630443871021271, "global_step": 45986, "epoch": 516, "lr": 9.32056859566057e-05}
{"train_loss": 0.42090094089508057, "global_step": 45987, "epoch": 516, "lr": 9.320539417423271e-05}
{"train_loss": 0.23824256658554077, "global_step": 45988, "epoch": 516, "lr": 9.32051023860513e-05}
{"train_loss": 0.40687623620033264, "global_step": 45989, "epoch": 516, "lr": 9.320481059206145e-05}
{"train_loss": 0.2703448235988617, "global_step": 45990, "epoch": 516, "lr": 9.320451879226326e-05}
{"train_loss": 0.3106272220611572, "global_step": 45991, "epoch": 516, "lr": 9.320422698665676e-05}
{"train_loss": 0.31648480892181396, "global_step": 45992, "epoch": 516, "lr": 9.320393517524197e-05}
{"train_loss": 0.23916678130626678, "global_step": 45993, "epoch": 516, "lr": 9.320364335801894e-05}
{"train_loss": 0.3179347515106201, "global_step": 45994, "epoch": 516, "lr": 9.320335153498769e-05}
{"train_loss": 0.3248554468154907, "global_step": 45995, "epoch": 516, "lr": 9.32030597061483e-05}
{"train_loss": 0.29640769958496094, "global_step": 45996, "epoch": 516, "lr": 9.320276787150078e-05}
{"train_loss": 0.29610106348991394, "global_step": 45997, "epoch": 516, "lr": 9.320247603104516e-05}
{"train_loss": 0.2780763804912567, "global_step": 45998, "epoch": 516, "lr": 9.32021841847815e-05}
{"train_loss": 0.31462162733078003, "global_step": 45999, "epoch": 516, "lr": 9.320189233270984e-05}
{"train_loss": 0.2930319607257843, "global_step": 46000, "epoch": 516, "lr": 9.320160047483021e-05}
{"train_loss": 0.2633391320705414, "global_step": 46001, "epoch": 516, "lr": 9.320130861114264e-05}
{"train_loss": 0.3389379382133484, "global_step": 46002, "epoch": 516, "lr": 9.320101674164719e-05}
{"train_loss": 0.33258765935897827, "global_step": 46003, "epoch": 516, "lr": 9.320072486634388e-05}
{"train_loss": 0.30645179748535156, "global_step": 46004, "epoch": 516, "lr": 9.320043298523278e-05}
{"train_loss": 0.2491777539253235, "global_step": 46005, "epoch": 516, "lr": 9.32001410983139e-05}
{"train_loss": 0.2366236448287964, "global_step": 46006, "epoch": 516, "lr": 9.319984920558729e-05}
{"train_loss": 0.3350868821144104, "global_step": 46007, "epoch": 516, "lr": 9.319955730705296e-05}
{"train_loss": 0.30619630217552185, "global_step": 46008, "epoch": 516, "lr": 9.319926540271101e-05}
{"train_loss": 0.32695087790489197, "global_step": 46009, "epoch": 516, "lr": 9.319897349256144e-05}
{"train_loss": 0.38795122504234314, "global_step": 46010, "epoch": 516, "lr": 9.319868157660428e-05}
{"train_loss": 0.2869330644607544, "global_step": 46011, "epoch": 516, "lr": 9.319838965483959e-05}
{"train_loss": 0.299395217152124, "global_step": 46012, "epoch": 516, "lr": 9.31980977272674e-05, "val_loss": 2.821725368499756}
{"train_loss": 0.295791894197464, "global_step": 46013, "epoch": 517, "lr": 9.319780579388776e-05}
{"train_loss": 0.33437633514404297, "global_step": 46014, "epoch": 517, "lr": 9.31975138547007e-05}
{"train_loss": 0.2879524827003479, "global_step": 46015, "epoch": 517, "lr": 9.319722190970625e-05}
{"train_loss": 0.3626250922679901, "global_step": 46016, "epoch": 517, "lr": 9.319692995890447e-05}
{"train_loss": 0.3636041283607483, "global_step": 46017, "epoch": 517, "lr": 9.319663800229538e-05}
{"train_loss": 0.31561633944511414, "global_step": 46018, "epoch": 517, "lr": 9.319634603987905e-05}
{"train_loss": 0.30180075764656067, "global_step": 46019, "epoch": 517, "lr": 9.319605407165549e-05}
{"train_loss": 0.2109639197587967, "global_step": 46020, "epoch": 517, "lr": 9.319576209762473e-05}
{"train_loss": 0.19075173139572144, "global_step": 46021, "epoch": 517, "lr": 9.319547011778684e-05}
{"train_loss": 0.30288803577423096, "global_step": 46022, "epoch": 517, "lr": 9.319517813214184e-05}
{"train_loss": 0.27905169129371643, "global_step": 46023, "epoch": 517, "lr": 9.319488614068979e-05}
{"train_loss": 0.29054099321365356, "global_step": 46024, "epoch": 517, "lr": 9.319459414343069e-05}
{"train_loss": 0.26305055618286133, "global_step": 46025, "epoch": 517, "lr": 9.319430214036462e-05}
{"train_loss": 0.3254581391811371, "global_step": 46026, "epoch": 517, "lr": 9.319401013149158e-05}
{"train_loss": 0.2772708535194397, "global_step": 46027, "epoch": 517, "lr": 9.319371811681166e-05}
{"train_loss": 0.3136458992958069, "global_step": 46028, "epoch": 517, "lr": 9.319342609632486e-05}
{"train_loss": 0.31215885281562805, "global_step": 46029, "epoch": 517, "lr": 9.319313407003123e-05}
{"train_loss": 0.30046209692955017, "global_step": 46030, "epoch": 517, "lr": 9.319284203793082e-05}
{"train_loss": 0.3596641719341278, "global_step": 46031, "epoch": 517, "lr": 9.319255000002365e-05}
{"train_loss": 0.27638110518455505, "global_step": 46032, "epoch": 517, "lr": 9.319225795630975e-05}
{"train_loss": 0.23777419328689575, "global_step": 46033, "epoch": 517, "lr": 9.31919659067892e-05}
{"train_loss": 0.27943888306617737, "global_step": 46034, "epoch": 517, "lr": 9.319167385146201e-05}
{"train_loss": 0.2611992657184601, "global_step": 46035, "epoch": 517, "lr": 9.319138179032823e-05}
{"train_loss": 0.28538504242897034, "global_step": 46036, "epoch": 517, "lr": 9.319108972338788e-05}
{"train_loss": 0.28217613697052, "global_step": 46037, "epoch": 517, "lr": 9.319079765064103e-05}
{"train_loss": 0.31350189447402954, "global_step": 46038, "epoch": 517, "lr": 9.31905055720877e-05}
{"train_loss": 0.24202223122119904, "global_step": 46039, "epoch": 517, "lr": 9.319021348772794e-05}
{"train_loss": 0.2955191731452942, "global_step": 46040, "epoch": 517, "lr": 9.318992139756177e-05}
{"train_loss": 0.3258734345436096, "global_step": 46041, "epoch": 517, "lr": 9.318962930158925e-05}
{"train_loss": 0.3755776584148407, "global_step": 46042, "epoch": 517, "lr": 9.318933719981041e-05}
{"train_loss": 0.24454143643379211, "global_step": 46043, "epoch": 517, "lr": 9.318904509222527e-05}
{"train_loss": 0.2234783172607422, "global_step": 46044, "epoch": 517, "lr": 9.318875297883391e-05}
{"train_loss": 0.2875717282295227, "global_step": 46045, "epoch": 517, "lr": 9.318846085963634e-05}
{"train_loss": 0.28767144680023193, "global_step": 46046, "epoch": 517, "lr": 9.318816873463263e-05}
{"train_loss": 0.22018174827098846, "global_step": 46047, "epoch": 517, "lr": 9.318787660382277e-05}
{"train_loss": 0.3002132475376129, "global_step": 46048, "epoch": 517, "lr": 9.318758446720685e-05}
{"train_loss": 0.30531468987464905, "global_step": 46049, "epoch": 517, "lr": 9.318729232478488e-05}
{"train_loss": 0.30824118852615356, "global_step": 46050, "epoch": 517, "lr": 9.318700017655688e-05}
{"train_loss": 0.34483015537261963, "global_step": 46051, "epoch": 517, "lr": 9.318670802252293e-05}
{"train_loss": 0.41784733533859253, "global_step": 46052, "epoch": 517, "lr": 9.318641586268306e-05}
{"train_loss": 0.26149705052375793, "global_step": 46053, "epoch": 517, "lr": 9.31861236970373e-05}
{"train_loss": 0.32346609234809875, "global_step": 46054, "epoch": 517, "lr": 9.318583152558569e-05}
{"train_loss": 0.23951725661754608, "global_step": 46055, "epoch": 517, "lr": 9.318553934832827e-05}
{"train_loss": 0.30139482021331787, "global_step": 46056, "epoch": 517, "lr": 9.318524716526508e-05}
{"train_loss": 0.26474159955978394, "global_step": 46057, "epoch": 517, "lr": 9.318495497639616e-05}
{"train_loss": 0.253133088350296, "global_step": 46058, "epoch": 517, "lr": 9.318466278172155e-05}
{"train_loss": 0.24546372890472412, "global_step": 46059, "epoch": 517, "lr": 9.318437058124129e-05}
{"train_loss": 0.3021329343318939, "global_step": 46060, "epoch": 517, "lr": 9.318407837495542e-05}
{"train_loss": 0.2513912618160248, "global_step": 46061, "epoch": 517, "lr": 9.318378616286397e-05}
{"train_loss": 0.3000775873661041, "global_step": 46062, "epoch": 517, "lr": 9.318349394496699e-05}
{"train_loss": 0.3146604001522064, "global_step": 46063, "epoch": 517, "lr": 9.318320172126451e-05}
{"train_loss": 0.3157464265823364, "global_step": 46064, "epoch": 517, "lr": 9.318290949175658e-05}
{"train_loss": 0.24955303966999054, "global_step": 46065, "epoch": 517, "lr": 9.318261725644324e-05}
{"train_loss": 0.28231561183929443, "global_step": 46066, "epoch": 517, "lr": 9.318232501532453e-05}
{"train_loss": 0.2887778878211975, "global_step": 46067, "epoch": 517, "lr": 9.318203276840046e-05}
{"train_loss": 0.30375415086746216, "global_step": 46068, "epoch": 517, "lr": 9.31817405156711e-05}
{"train_loss": 0.25482627749443054, "global_step": 46069, "epoch": 517, "lr": 9.318144825713649e-05}
{"train_loss": 0.2923717796802521, "global_step": 46070, "epoch": 517, "lr": 9.318115599279666e-05}
{"train_loss": 0.30135008692741394, "global_step": 46071, "epoch": 517, "lr": 9.318086372265164e-05}
{"train_loss": 0.2742128074169159, "global_step": 46072, "epoch": 517, "lr": 9.318057144670148e-05}
{"train_loss": 0.409027636051178, "global_step": 46073, "epoch": 517, "lr": 9.318027916494623e-05}
{"train_loss": 0.288819283246994, "global_step": 46074, "epoch": 517, "lr": 9.31799868773859e-05}
{"train_loss": 0.21876682341098785, "global_step": 46075, "epoch": 517, "lr": 9.317969458402056e-05}
{"train_loss": 0.25303226709365845, "global_step": 46076, "epoch": 517, "lr": 9.317940228485023e-05}
{"train_loss": 0.3871489465236664, "global_step": 46077, "epoch": 517, "lr": 9.317910997987496e-05}
{"train_loss": 0.36757510900497437, "global_step": 46078, "epoch": 517, "lr": 9.317881766909478e-05}
{"train_loss": 0.31260135769844055, "global_step": 46079, "epoch": 517, "lr": 9.317852535250975e-05}
{"train_loss": 0.3115012049674988, "global_step": 46080, "epoch": 517, "lr": 9.317823303011988e-05}
{"train_loss": 0.26093876361846924, "global_step": 46081, "epoch": 517, "lr": 9.317794070192523e-05}
{"train_loss": 0.3158130645751953, "global_step": 46082, "epoch": 517, "lr": 9.317764836792581e-05}
{"train_loss": 0.31992533802986145, "global_step": 46083, "epoch": 517, "lr": 9.31773560281217e-05}
{"train_loss": 0.24083642661571503, "global_step": 46084, "epoch": 517, "lr": 9.317706368251292e-05}
{"train_loss": 0.2619425654411316, "global_step": 46085, "epoch": 517, "lr": 9.317677133109953e-05}
{"train_loss": 0.2525472044944763, "global_step": 46086, "epoch": 517, "lr": 9.31764789738815e-05}
{"train_loss": 0.32167375087738037, "global_step": 46087, "epoch": 517, "lr": 9.317618661085896e-05}
{"train_loss": 0.2963191568851471, "global_step": 46088, "epoch": 517, "lr": 9.31758942420319e-05}
{"train_loss": 0.26828816533088684, "global_step": 46089, "epoch": 517, "lr": 9.317560186740035e-05}
{"train_loss": 0.4449794888496399, "global_step": 46090, "epoch": 517, "lr": 9.317530948696438e-05}
{"train_loss": 0.25671645998954773, "global_step": 46091, "epoch": 517, "lr": 9.317501710072401e-05}
{"train_loss": 0.23150528967380524, "global_step": 46092, "epoch": 517, "lr": 9.31747247086793e-05}
{"train_loss": 0.2113339900970459, "global_step": 46093, "epoch": 517, "lr": 9.317443231083027e-05}
{"train_loss": 0.2342921942472458, "global_step": 46094, "epoch": 517, "lr": 9.317413990717695e-05}
{"train_loss": 0.23414617776870728, "global_step": 46095, "epoch": 517, "lr": 9.31738474977194e-05}
{"train_loss": 0.33519938588142395, "global_step": 46096, "epoch": 517, "lr": 9.317355508245767e-05}
{"train_loss": 0.24129100143909454, "global_step": 46097, "epoch": 517, "lr": 9.317326266139175e-05}
{"train_loss": 0.3533363938331604, "global_step": 46098, "epoch": 517, "lr": 9.317297023452173e-05}
{"train_loss": 0.2595241665840149, "global_step": 46099, "epoch": 517, "lr": 9.317267780184762e-05}
{"train_loss": 0.387424498796463, "global_step": 46100, "epoch": 517, "lr": 9.317238536336948e-05}
{"train_loss": 0.29221745923663794, "global_step": 46101, "epoch": 517, "lr": 9.317209291908732e-05, "val_loss": 2.8875269889831543}
{"train_loss": 0.3248303234577179, "global_step": 46102, "epoch": 518, "lr": 9.317180046900123e-05}
{"train_loss": 0.28411799669265747, "global_step": 46103, "epoch": 518, "lr": 9.31715080131112e-05}
{"train_loss": 0.2888464033603668, "global_step": 46104, "epoch": 518, "lr": 9.317121555141728e-05}
{"train_loss": 0.2974313199520111, "global_step": 46105, "epoch": 518, "lr": 9.317092308391954e-05}
{"train_loss": 0.2825082838535309, "global_step": 46106, "epoch": 518, "lr": 9.317063061061797e-05}
{"train_loss": 0.3353675603866577, "global_step": 46107, "epoch": 518, "lr": 9.317033813151265e-05}
{"train_loss": 0.38768836855888367, "global_step": 46108, "epoch": 518, "lr": 9.31700456466036e-05}
{"train_loss": 0.18402981758117676, "global_step": 46109, "epoch": 518, "lr": 9.316975315589086e-05}
{"train_loss": 0.24239569902420044, "global_step": 46110, "epoch": 518, "lr": 9.316946065937447e-05}
{"train_loss": 0.3219471871852875, "global_step": 46111, "epoch": 518, "lr": 9.316916815705447e-05}
{"train_loss": 0.3523234724998474, "global_step": 46112, "epoch": 518, "lr": 9.316887564893091e-05}
{"train_loss": 0.24996601045131683, "global_step": 46113, "epoch": 518, "lr": 9.316858313500383e-05}
{"train_loss": 0.28332948684692383, "global_step": 46114, "epoch": 518, "lr": 9.316829061527326e-05}
{"train_loss": 0.31368499994277954, "global_step": 46115, "epoch": 518, "lr": 9.316799808973922e-05}
{"train_loss": 0.3096354901790619, "global_step": 46116, "epoch": 518, "lr": 9.316770555840178e-05}
{"train_loss": 0.3385150134563446, "global_step": 46117, "epoch": 518, "lr": 9.316741302126096e-05}
{"train_loss": 0.3532116711139679, "global_step": 46118, "epoch": 518, "lr": 9.316712047831681e-05}
{"train_loss": 0.2675417959690094, "global_step": 46119, "epoch": 518, "lr": 9.316682792956938e-05}
{"train_loss": 0.3165602385997772, "global_step": 46120, "epoch": 518, "lr": 9.31665353750187e-05}
{"train_loss": 0.325030118227005, "global_step": 46121, "epoch": 518, "lr": 9.316624281466477e-05}
{"train_loss": 0.2675340473651886, "global_step": 46122, "epoch": 518, "lr": 9.31659502485077e-05}
{"train_loss": 0.2485160529613495, "global_step": 46123, "epoch": 518, "lr": 9.316565767654748e-05}
{"train_loss": 0.36972805857658386, "global_step": 46124, "epoch": 518, "lr": 9.316536509878417e-05}
{"train_loss": 0.29601794481277466, "global_step": 46125, "epoch": 518, "lr": 9.316507251521778e-05}
{"train_loss": 0.4134765565395355, "global_step": 46126, "epoch": 518, "lr": 9.31647799258484e-05}
{"train_loss": 0.4760536551475525, "global_step": 46127, "epoch": 518, "lr": 9.316448733067603e-05}
{"train_loss": 0.4226304590702057, "global_step": 46128, "epoch": 518, "lr": 9.316419472970073e-05}
{"train_loss": 0.2444460093975067, "global_step": 46129, "epoch": 518, "lr": 9.316390212292252e-05}
{"train_loss": 0.24782253801822662, "global_step": 46130, "epoch": 518, "lr": 9.316360951034144e-05}
{"train_loss": 0.196153923869133, "global_step": 46131, "epoch": 518, "lr": 9.316331689195756e-05}
{"train_loss": 0.2778302729129791, "global_step": 46132, "epoch": 518, "lr": 9.316302426777089e-05}
{"train_loss": 0.3357888460159302, "global_step": 46133, "epoch": 518, "lr": 9.316273163778146e-05}
{"train_loss": 0.2769155204296112, "global_step": 46134, "epoch": 518, "lr": 9.316243900198936e-05}
{"train_loss": 0.3273234963417053, "global_step": 46135, "epoch": 518, "lr": 9.316214636039458e-05}
{"train_loss": 0.3187571167945862, "global_step": 46136, "epoch": 518, "lr": 9.316185371299717e-05}
{"train_loss": 0.24596653878688812, "global_step": 46137, "epoch": 518, "lr": 9.316156105979717e-05}
{"train_loss": 0.2301594465970993, "global_step": 46138, "epoch": 518, "lr": 9.316126840079464e-05}
{"train_loss": 0.3873986303806305, "global_step": 46139, "epoch": 518, "lr": 9.316097573598959e-05}
{"train_loss": 0.29389551281929016, "global_step": 46140, "epoch": 518, "lr": 9.316068306538208e-05}
{"train_loss": 0.3282528817653656, "global_step": 46141, "epoch": 518, "lr": 9.316039038897215e-05}
{"train_loss": 0.356644868850708, "global_step": 46142, "epoch": 518, "lr": 9.316009770675982e-05}
{"train_loss": 0.2759777009487152, "global_step": 46143, "epoch": 518, "lr": 9.315980501874515e-05}
{"train_loss": 0.2608141601085663, "global_step": 46144, "epoch": 518, "lr": 9.315951232492816e-05}
{"train_loss": 0.2925572693347931, "global_step": 46145, "epoch": 518, "lr": 9.31592196253089e-05}
{"train_loss": 0.3371121287345886, "global_step": 46146, "epoch": 518, "lr": 9.315892691988741e-05}
{"train_loss": 0.2765965461730957, "global_step": 46147, "epoch": 518, "lr": 9.315863420866373e-05}
{"train_loss": 0.27762076258659363, "global_step": 46148, "epoch": 518, "lr": 9.315834149163791e-05}
{"train_loss": 0.38287684321403503, "global_step": 46149, "epoch": 518, "lr": 9.315804876880996e-05}
{"train_loss": 0.23037134110927582, "global_step": 46150, "epoch": 518, "lr": 9.315775604017994e-05}
{"train_loss": 0.35945531725883484, "global_step": 46151, "epoch": 518, "lr": 9.31574633057479e-05}
{"train_loss": 0.24923989176750183, "global_step": 46152, "epoch": 518, "lr": 9.315717056551385e-05}
{"train_loss": 0.3324299156665802, "global_step": 46153, "epoch": 518, "lr": 9.315687781947783e-05}
{"train_loss": 0.29000115394592285, "global_step": 46154, "epoch": 518, "lr": 9.315658506763993e-05}
{"train_loss": 0.2074577510356903, "global_step": 46155, "epoch": 518, "lr": 9.315629231000013e-05}
{"train_loss": 0.24882033467292786, "global_step": 46156, "epoch": 518, "lr": 9.315599954655849e-05}
{"train_loss": 0.26557108759880066, "global_step": 46157, "epoch": 518, "lr": 9.315570677731507e-05}
{"train_loss": 0.38445597887039185, "global_step": 46158, "epoch": 518, "lr": 9.315541400226988e-05}
{"train_loss": 0.2554546594619751, "global_step": 46159, "epoch": 518, "lr": 9.315512122142295e-05}
{"train_loss": 0.24929431080818176, "global_step": 46160, "epoch": 518, "lr": 9.315482843477438e-05}
{"train_loss": 0.2782846689224243, "global_step": 46161, "epoch": 518, "lr": 9.315453564232415e-05}
{"train_loss": 0.2615402340888977, "global_step": 46162, "epoch": 518, "lr": 9.315424284407232e-05}
{"train_loss": 0.29130375385284424, "global_step": 46163, "epoch": 518, "lr": 9.315395004001892e-05}
{"train_loss": 0.25259867310523987, "global_step": 46164, "epoch": 518, "lr": 9.3153657230164e-05}
{"train_loss": 0.18632085621356964, "global_step": 46165, "epoch": 518, "lr": 9.315336441450761e-05}
{"train_loss": 0.3511080741882324, "global_step": 46166, "epoch": 518, "lr": 9.315307159304976e-05}
{"train_loss": 0.2682047486305237, "global_step": 46167, "epoch": 518, "lr": 9.315277876579052e-05}
{"train_loss": 0.2118833065032959, "global_step": 46168, "epoch": 518, "lr": 9.31524859327299e-05}
{"train_loss": 0.31620538234710693, "global_step": 46169, "epoch": 518, "lr": 9.315219309386797e-05}
{"train_loss": 0.3369002938270569, "global_step": 46170, "epoch": 518, "lr": 9.315190024920474e-05}
{"train_loss": 0.2733660936355591, "global_step": 46171, "epoch": 518, "lr": 9.315160739874028e-05}
{"train_loss": 0.252858430147171, "global_step": 46172, "epoch": 518, "lr": 9.315131454247459e-05}
{"train_loss": 0.22668330371379852, "global_step": 46173, "epoch": 518, "lr": 9.315102168040774e-05}
{"train_loss": 0.3455798923969269, "global_step": 46174, "epoch": 518, "lr": 9.315072881253977e-05}
{"train_loss": 0.24989734590053558, "global_step": 46175, "epoch": 518, "lr": 9.31504359388707e-05}
{"train_loss": 0.32590410113334656, "global_step": 46176, "epoch": 518, "lr": 9.315014305940058e-05}
{"train_loss": 0.22164605557918549, "global_step": 46177, "epoch": 518, "lr": 9.314985017412945e-05}
{"train_loss": 0.2710399925708771, "global_step": 46178, "epoch": 518, "lr": 9.314955728305736e-05}
{"train_loss": 0.35881319642066956, "global_step": 46179, "epoch": 518, "lr": 9.314926438618432e-05}
{"train_loss": 0.3824215829372406, "global_step": 46180, "epoch": 518, "lr": 9.31489714835104e-05}
{"train_loss": 0.32572901248931885, "global_step": 46181, "epoch": 518, "lr": 9.314867857503562e-05}
{"train_loss": 0.32746705412864685, "global_step": 46182, "epoch": 518, "lr": 9.314838566076002e-05}
{"train_loss": 0.3230123221874237, "global_step": 46183, "epoch": 518, "lr": 9.314809274068367e-05}
{"train_loss": 0.21760612726211548, "global_step": 46184, "epoch": 518, "lr": 9.314779981480657e-05}
{"train_loss": 0.22599059343338013, "global_step": 46185, "epoch": 518, "lr": 9.314750688312877e-05}
{"train_loss": 0.34365493059158325, "global_step": 46186, "epoch": 518, "lr": 9.314721394565031e-05}
{"train_loss": 0.3454487919807434, "global_step": 46187, "epoch": 518, "lr": 9.314692100237124e-05}
{"train_loss": 0.3324085474014282, "global_step": 46188, "epoch": 518, "lr": 9.314662805329159e-05}
{"train_loss": 0.34163686633110046, "global_step": 46189, "epoch": 518, "lr": 9.314633509841141e-05}
{"train_loss": 0.2981439631306723, "global_step": 46190, "epoch": 518, "lr": 9.314604213773073e-05, "val_loss": 2.9001240730285645}
{"train_loss": 0.32018938660621643, "global_step": 46191, "epoch": 519, "lr": 9.314574917124959e-05}
{"train_loss": 0.31136423349380493, "global_step": 46192, "epoch": 519, "lr": 9.314545619896802e-05}
{"train_loss": 0.3526431918144226, "global_step": 46193, "epoch": 519, "lr": 9.314516322088608e-05}
{"train_loss": 0.23633144795894623, "global_step": 46194, "epoch": 519, "lr": 9.314487023700378e-05}
{"train_loss": 0.23328468203544617, "global_step": 46195, "epoch": 519, "lr": 9.314457724732119e-05}
{"train_loss": 0.27271580696105957, "global_step": 46196, "epoch": 519, "lr": 9.314428425183835e-05}
{"train_loss": 0.2420811802148819, "global_step": 46197, "epoch": 519, "lr": 9.314399125055528e-05}
{"train_loss": 0.30772659182548523, "global_step": 46198, "epoch": 519, "lr": 9.314369824347202e-05}
{"train_loss": 0.27965477108955383, "global_step": 46199, "epoch": 519, "lr": 9.314340523058863e-05}
{"train_loss": 0.3271166682243347, "global_step": 46200, "epoch": 519, "lr": 9.314311221190512e-05}
{"train_loss": 0.2805485427379608, "global_step": 46201, "epoch": 519, "lr": 9.314281918742153e-05}
{"train_loss": 0.2784876227378845, "global_step": 46202, "epoch": 519, "lr": 9.314252615713794e-05}
{"train_loss": 0.2623045742511749, "global_step": 46203, "epoch": 519, "lr": 9.314223312105436e-05}
{"train_loss": 0.29601386189460754, "global_step": 46204, "epoch": 519, "lr": 9.314194007917083e-05}
{"train_loss": 0.2612302601337433, "global_step": 46205, "epoch": 519, "lr": 9.314164703148738e-05}
{"train_loss": 0.30788734555244446, "global_step": 46206, "epoch": 519, "lr": 9.314135397800408e-05}
{"train_loss": 0.2967119514942169, "global_step": 46207, "epoch": 519, "lr": 9.314106091872092e-05}
{"train_loss": 0.3462841808795929, "global_step": 46208, "epoch": 519, "lr": 9.3140767853638e-05}
{"train_loss": 0.31369906663894653, "global_step": 46209, "epoch": 519, "lr": 9.314047478275533e-05}
{"train_loss": 0.3351387083530426, "global_step": 46210, "epoch": 519, "lr": 9.314018170607294e-05}
{"train_loss": 0.39637359976768494, "global_step": 46211, "epoch": 519, "lr": 9.313988862359088e-05}
{"train_loss": 0.3976999521255493, "global_step": 46212, "epoch": 519, "lr": 9.313959553530918e-05}
{"train_loss": 0.2797849774360657, "global_step": 46213, "epoch": 519, "lr": 9.31393024412279e-05}
{"train_loss": 0.22164641320705414, "global_step": 46214, "epoch": 519, "lr": 9.313900934134705e-05}
{"train_loss": 0.2945444583892822, "global_step": 46215, "epoch": 519, "lr": 9.31387162356667e-05}
{"train_loss": 0.3207797706127167, "global_step": 46216, "epoch": 519, "lr": 9.313842312418687e-05}
{"train_loss": 0.3322157859802246, "global_step": 46217, "epoch": 519, "lr": 9.313813000690761e-05}
{"train_loss": 0.18943457305431366, "global_step": 46218, "epoch": 519, "lr": 9.313783688382895e-05}
{"train_loss": 0.20098020136356354, "global_step": 46219, "epoch": 519, "lr": 9.313754375495092e-05}
{"train_loss": 0.383633017539978, "global_step": 46220, "epoch": 519, "lr": 9.31372506202736e-05}
{"train_loss": 0.31161773204803467, "global_step": 46221, "epoch": 519, "lr": 9.313695747979698e-05}
{"train_loss": 0.22365953028202057, "global_step": 46222, "epoch": 519, "lr": 9.313666433352113e-05}
{"train_loss": 0.23614482581615448, "global_step": 46223, "epoch": 519, "lr": 9.313637118144609e-05}
{"train_loss": 0.297229528427124, "global_step": 46224, "epoch": 519, "lr": 9.313607802357188e-05}
{"train_loss": 0.25065889954566956, "global_step": 46225, "epoch": 519, "lr": 9.313578485989856e-05}
{"train_loss": 0.3504006862640381, "global_step": 46226, "epoch": 519, "lr": 9.313549169042615e-05}
{"train_loss": 0.35371533036231995, "global_step": 46227, "epoch": 519, "lr": 9.31351985151547e-05}
{"train_loss": 0.1549813598394394, "global_step": 46228, "epoch": 519, "lr": 9.313490533408425e-05}
{"train_loss": 0.32469815015792847, "global_step": 46229, "epoch": 519, "lr": 9.313461214721484e-05}
{"train_loss": 0.2191588580608368, "global_step": 46230, "epoch": 519, "lr": 9.31343189545465e-05}
{"train_loss": 0.27389079332351685, "global_step": 46231, "epoch": 519, "lr": 9.313402575607928e-05}
{"train_loss": 0.2954730689525604, "global_step": 46232, "epoch": 519, "lr": 9.313373255181321e-05}
{"train_loss": 0.21393944323062897, "global_step": 46233, "epoch": 519, "lr": 9.313343934174836e-05}
{"train_loss": 0.2537091374397278, "global_step": 46234, "epoch": 519, "lr": 9.313314612588471e-05}
{"train_loss": 0.25411540269851685, "global_step": 46235, "epoch": 519, "lr": 9.313285290422236e-05}
{"train_loss": 0.3484607934951782, "global_step": 46236, "epoch": 519, "lr": 9.31325596767613e-05}
{"train_loss": 0.2355698049068451, "global_step": 46237, "epoch": 519, "lr": 9.313226644350161e-05}
{"train_loss": 0.23794838786125183, "global_step": 46238, "epoch": 519, "lr": 9.313197320444333e-05}
{"train_loss": 0.3165397346019745, "global_step": 46239, "epoch": 519, "lr": 9.313167995958647e-05}
{"train_loss": 0.24223729968070984, "global_step": 46240, "epoch": 519, "lr": 9.313138670893107e-05}
{"train_loss": 0.21976909041404724, "global_step": 46241, "epoch": 519, "lr": 9.313109345247717e-05}
{"train_loss": 0.30220770835876465, "global_step": 46242, "epoch": 519, "lr": 9.313080019022484e-05}
{"train_loss": 0.26066675782203674, "global_step": 46243, "epoch": 519, "lr": 9.313050692217409e-05}
{"train_loss": 0.35220736265182495, "global_step": 46244, "epoch": 519, "lr": 9.313021364832497e-05}
{"train_loss": 0.31730908155441284, "global_step": 46245, "epoch": 519, "lr": 9.312992036867753e-05}
{"train_loss": 0.2571519613265991, "global_step": 46246, "epoch": 519, "lr": 9.312962708323179e-05}
{"train_loss": 0.3621291220188141, "global_step": 46247, "epoch": 519, "lr": 9.312933379198779e-05}
{"train_loss": 0.22820623219013214, "global_step": 46248, "epoch": 519, "lr": 9.31290404949456e-05}
{"train_loss": 0.24072332680225372, "global_step": 46249, "epoch": 519, "lr": 9.312874719210522e-05}
{"train_loss": 0.3753739595413208, "global_step": 46250, "epoch": 519, "lr": 9.312845388346669e-05}
{"train_loss": 0.24043887853622437, "global_step": 46251, "epoch": 519, "lr": 9.312816056903008e-05}
{"train_loss": 0.3069376051425934, "global_step": 46252, "epoch": 519, "lr": 9.312786724879542e-05}
{"train_loss": 0.24280931055545807, "global_step": 46253, "epoch": 519, "lr": 9.312757392276274e-05}
{"train_loss": 0.23817364871501923, "global_step": 46254, "epoch": 519, "lr": 9.312728059093207e-05}
{"train_loss": 0.2849481999874115, "global_step": 46255, "epoch": 519, "lr": 9.312698725330348e-05}
{"train_loss": 0.22547616064548492, "global_step": 46256, "epoch": 519, "lr": 9.312669390987699e-05}
{"train_loss": 0.3446740508079529, "global_step": 46257, "epoch": 519, "lr": 9.312640056065264e-05}
{"train_loss": 0.3579685688018799, "global_step": 46258, "epoch": 519, "lr": 9.312610720563048e-05}
{"train_loss": 0.23646226525306702, "global_step": 46259, "epoch": 519, "lr": 9.312581384481053e-05}
{"train_loss": 0.29340049624443054, "global_step": 46260, "epoch": 519, "lr": 9.312552047819284e-05}
{"train_loss": 0.24532148241996765, "global_step": 46261, "epoch": 519, "lr": 9.312522710577745e-05}
{"train_loss": 0.24057871103286743, "global_step": 46262, "epoch": 519, "lr": 9.31249337275644e-05}
{"train_loss": 0.3041948676109314, "global_step": 46263, "epoch": 519, "lr": 9.312464034355372e-05}
{"train_loss": 0.36608991026878357, "global_step": 46264, "epoch": 519, "lr": 9.312434695374546e-05}
{"train_loss": 0.23456411063671112, "global_step": 46265, "epoch": 519, "lr": 9.312405355813967e-05}
{"train_loss": 0.32793429493904114, "global_step": 46266, "epoch": 519, "lr": 9.312376015673637e-05}
{"train_loss": 0.23158730566501617, "global_step": 46267, "epoch": 519, "lr": 9.312346674953561e-05}
{"train_loss": 0.23379889130592346, "global_step": 46268, "epoch": 519, "lr": 9.312317333653742e-05}
{"train_loss": 0.2330276221036911, "global_step": 46269, "epoch": 519, "lr": 9.312287991774184e-05}
{"train_loss": 0.28029900789260864, "global_step": 46270, "epoch": 519, "lr": 9.312258649314891e-05}
{"train_loss": 0.36102136969566345, "global_step": 46271, "epoch": 519, "lr": 9.312229306275869e-05}
{"train_loss": 0.28425711393356323, "global_step": 46272, "epoch": 519, "lr": 9.31219996265712e-05}
{"train_loss": 0.39820346236228943, "global_step": 46273, "epoch": 519, "lr": 9.312170618458648e-05}
{"train_loss": 0.2674482464790344, "global_step": 46274, "epoch": 519, "lr": 9.312141273680456e-05}
{"train_loss": 0.32820242643356323, "global_step": 46275, "epoch": 519, "lr": 9.312111928322551e-05}
{"train_loss": 0.26004505157470703, "global_step": 46276, "epoch": 519, "lr": 9.312082582384934e-05}
{"train_loss": 0.3196602165699005, "global_step": 46277, "epoch": 519, "lr": 9.312053235867612e-05}
{"train_loss": 0.30301815271377563, "global_step": 46278, "epoch": 519, "lr": 9.312023888770584e-05}
{"train_loss": 0.2868840565507332, "global_step": 46279, "epoch": 519, "lr": 9.31199454109386e-05, "val_loss": 2.954742431640625}
{"train_loss": 0.19574956595897675, "global_step": 46280, "epoch": 520, "lr": 9.311965192837439e-05}
{"train_loss": 0.29442185163497925, "global_step": 46281, "epoch": 520, "lr": 9.311935844001327e-05}
{"train_loss": 0.37798213958740234, "global_step": 46282, "epoch": 520, "lr": 9.311906494585527e-05}
{"train_loss": 0.2788907587528229, "global_step": 46283, "epoch": 520, "lr": 9.311877144590045e-05}
{"train_loss": 0.33654388785362244, "global_step": 46284, "epoch": 520, "lr": 9.311847794014883e-05}
{"train_loss": 0.16564127802848816, "global_step": 46285, "epoch": 520, "lr": 9.311818442860047e-05}
{"train_loss": 0.28186991810798645, "global_step": 46286, "epoch": 520, "lr": 9.311789091125538e-05}
{"train_loss": 0.21702712774276733, "global_step": 46287, "epoch": 520, "lr": 9.311759738811362e-05}
{"train_loss": 0.3783552348613739, "global_step": 46288, "epoch": 520, "lr": 9.311730385917523e-05}
{"train_loss": 0.2315959632396698, "global_step": 46289, "epoch": 520, "lr": 9.311701032444023e-05}
{"train_loss": 0.33110734820365906, "global_step": 46290, "epoch": 520, "lr": 9.311671678390868e-05}
{"train_loss": 0.28664180636405945, "global_step": 46291, "epoch": 520, "lr": 9.311642323758062e-05}
{"train_loss": 0.3716354966163635, "global_step": 46292, "epoch": 520, "lr": 9.311612968545608e-05}
{"train_loss": 0.3080529570579529, "global_step": 46293, "epoch": 520, "lr": 9.31158361275351e-05}
{"train_loss": 0.3953931927680969, "global_step": 46294, "epoch": 520, "lr": 9.311554256381772e-05}
{"train_loss": 0.30376335978507996, "global_step": 46295, "epoch": 520, "lr": 9.311524899430397e-05}
{"train_loss": 0.2304723709821701, "global_step": 46296, "epoch": 520, "lr": 9.311495541899394e-05}
{"train_loss": 0.30053892731666565, "global_step": 46297, "epoch": 520, "lr": 9.311466183788758e-05}
{"train_loss": 0.35813087224960327, "global_step": 46298, "epoch": 520, "lr": 9.3114368250985e-05}
{"train_loss": 0.2410004436969757, "global_step": 46299, "epoch": 520, "lr": 9.311407465828622e-05}
{"train_loss": 0.32658517360687256, "global_step": 46300, "epoch": 520, "lr": 9.311378105979128e-05}
{"train_loss": 0.32139790058135986, "global_step": 46301, "epoch": 520, "lr": 9.311348745550021e-05}
{"train_loss": 0.17232300341129303, "global_step": 46302, "epoch": 520, "lr": 9.311319384541307e-05}
{"train_loss": 0.3005734980106354, "global_step": 46303, "epoch": 520, "lr": 9.311290022952988e-05}
{"train_loss": 0.2902337610721588, "global_step": 46304, "epoch": 520, "lr": 9.311260660785069e-05}
{"train_loss": 0.35793188214302063, "global_step": 46305, "epoch": 520, "lr": 9.311231298037553e-05}
{"train_loss": 0.24619629979133606, "global_step": 46306, "epoch": 520, "lr": 9.311201934710444e-05}
{"train_loss": 0.3041684925556183, "global_step": 46307, "epoch": 520, "lr": 9.311172570803747e-05}
{"train_loss": 0.24660389125347137, "global_step": 46308, "epoch": 520, "lr": 9.311143206317465e-05}
{"train_loss": 0.23306059837341309, "global_step": 46309, "epoch": 520, "lr": 9.311113841251602e-05}
{"train_loss": 0.27769333124160767, "global_step": 46310, "epoch": 520, "lr": 9.311084475606165e-05}
{"train_loss": 0.27022191882133484, "global_step": 46311, "epoch": 520, "lr": 9.311055109381153e-05}
{"train_loss": 0.26113104820251465, "global_step": 46312, "epoch": 520, "lr": 9.311025742576572e-05}
{"train_loss": 0.3432408571243286, "global_step": 46313, "epoch": 520, "lr": 9.310996375192427e-05}
{"train_loss": 0.2654955983161926, "global_step": 46314, "epoch": 520, "lr": 9.31096700722872e-05}
{"train_loss": 0.26720795035362244, "global_step": 46315, "epoch": 520, "lr": 9.310937638685458e-05}
{"train_loss": 0.3400332033634186, "global_step": 46316, "epoch": 520, "lr": 9.310908269562641e-05}
{"train_loss": 0.2768445611000061, "global_step": 46317, "epoch": 520, "lr": 9.310878899860274e-05}
{"train_loss": 0.2861146330833435, "global_step": 46318, "epoch": 520, "lr": 9.310849529578365e-05}
{"train_loss": 0.2696550786495209, "global_step": 46319, "epoch": 520, "lr": 9.310820158716913e-05}
{"train_loss": 0.23448723554611206, "global_step": 46320, "epoch": 520, "lr": 9.310790787275923e-05}
{"train_loss": 0.2915612459182739, "global_step": 46321, "epoch": 520, "lr": 9.310761415255402e-05}
{"train_loss": 0.3108522891998291, "global_step": 46322, "epoch": 520, "lr": 9.31073204265535e-05}
{"train_loss": 0.2867763936519623, "global_step": 46323, "epoch": 520, "lr": 9.310702669475772e-05}
{"train_loss": 0.305603563785553, "global_step": 46324, "epoch": 520, "lr": 9.310673295716675e-05}
{"train_loss": 0.31951695680618286, "global_step": 46325, "epoch": 520, "lr": 9.310643921378059e-05}
{"train_loss": 0.182908833026886, "global_step": 46326, "epoch": 520, "lr": 9.31061454645993e-05}
{"train_loss": 0.34156879782676697, "global_step": 46327, "epoch": 520, "lr": 9.310585170962291e-05}
{"train_loss": 0.2695057690143585, "global_step": 46328, "epoch": 520, "lr": 9.310555794885145e-05}
{"train_loss": 0.2910187542438507, "global_step": 46329, "epoch": 520, "lr": 9.3105264182285e-05}
{"train_loss": 0.310126394033432, "global_step": 46330, "epoch": 520, "lr": 9.310497040992355e-05}
{"train_loss": 0.2810780704021454, "global_step": 46331, "epoch": 520, "lr": 9.310467663176719e-05}
{"train_loss": 0.2471802979707718, "global_step": 46332, "epoch": 520, "lr": 9.31043828478159e-05}
{"train_loss": 0.3225370943546295, "global_step": 46333, "epoch": 520, "lr": 9.310408905806977e-05}
{"train_loss": 0.3535677194595337, "global_step": 46334, "epoch": 520, "lr": 9.310379526252881e-05}
{"train_loss": 0.29102790355682373, "global_step": 46335, "epoch": 520, "lr": 9.310350146119308e-05}
{"train_loss": 0.2642635107040405, "global_step": 46336, "epoch": 520, "lr": 9.31032076540626e-05}
{"train_loss": 0.25903913378715515, "global_step": 46337, "epoch": 520, "lr": 9.310291384113742e-05}
{"train_loss": 0.30387264490127563, "global_step": 46338, "epoch": 520, "lr": 9.310262002241758e-05}
{"train_loss": 0.2879112660884857, "global_step": 46339, "epoch": 520, "lr": 9.310232619790314e-05}
{"train_loss": 0.2206205129623413, "global_step": 46340, "epoch": 520, "lr": 9.310203236759408e-05}
{"train_loss": 0.47075217962265015, "global_step": 46341, "epoch": 520, "lr": 9.31017385314905e-05}
{"train_loss": 0.4171574115753174, "global_step": 46342, "epoch": 520, "lr": 9.310144468959242e-05}
{"train_loss": 0.2744869291782379, "global_step": 46343, "epoch": 520, "lr": 9.310115084189986e-05}
{"train_loss": 0.2798614501953125, "global_step": 46344, "epoch": 520, "lr": 9.310085698841289e-05}
{"train_loss": 0.2925739586353302, "global_step": 46345, "epoch": 520, "lr": 9.310056312913151e-05}
{"train_loss": 0.254463255405426, "global_step": 46346, "epoch": 520, "lr": 9.310026926405582e-05}
{"train_loss": 0.2585260272026062, "global_step": 46347, "epoch": 520, "lr": 9.30999753931858e-05}
{"train_loss": 0.24540655314922333, "global_step": 46348, "epoch": 520, "lr": 9.309968151652152e-05}
{"train_loss": 0.3248269259929657, "global_step": 46349, "epoch": 520, "lr": 9.309938763406302e-05}
{"train_loss": 0.26876944303512573, "global_step": 46350, "epoch": 520, "lr": 9.309909374581032e-05}
{"train_loss": 0.2605839967727661, "global_step": 46351, "epoch": 520, "lr": 9.309879985176349e-05}
{"train_loss": 0.21516336500644684, "global_step": 46352, "epoch": 520, "lr": 9.309850595192254e-05}
{"train_loss": 0.31880995631217957, "global_step": 46353, "epoch": 520, "lr": 9.309821204628752e-05}
{"train_loss": 0.32338520884513855, "global_step": 46354, "epoch": 520, "lr": 9.309791813485846e-05}
{"train_loss": 0.2686397135257721, "global_step": 46355, "epoch": 520, "lr": 9.309762421763543e-05}
{"train_loss": 0.3576658368110657, "global_step": 46356, "epoch": 520, "lr": 9.309733029461844e-05}
{"train_loss": 0.29166197776794434, "global_step": 46357, "epoch": 520, "lr": 9.309703636580755e-05}
{"train_loss": 0.25097477436065674, "global_step": 46358, "epoch": 520, "lr": 9.309674243120277e-05}
{"train_loss": 0.26276078820228577, "global_step": 46359, "epoch": 520, "lr": 9.309644849080416e-05}
{"train_loss": 0.17209142446517944, "global_step": 46360, "epoch": 520, "lr": 9.309615454461179e-05}
{"train_loss": 0.2422007918357849, "global_step": 46361, "epoch": 520, "lr": 9.309586059262563e-05}
{"train_loss": 0.29523178935050964, "global_step": 46362, "epoch": 520, "lr": 9.309556663484578e-05}
{"train_loss": 0.24429334700107574, "global_step": 46363, "epoch": 520, "lr": 9.309527267127224e-05}
{"train_loss": 0.2811722755432129, "global_step": 46364, "epoch": 520, "lr": 9.309497870190507e-05}
{"train_loss": 0.27358755469322205, "global_step": 46365, "epoch": 520, "lr": 9.309468472674431e-05}
{"train_loss": 0.23396626114845276, "global_step": 46366, "epoch": 520, "lr": 9.309439074578999e-05}
{"train_loss": 0.25022998452186584, "global_step": 46367, "epoch": 520, "lr": 9.309409675904214e-05}
{"train_loss": 0.28561928901779515, "global_step": 46368, "epoch": 520, "lr": 9.309380276650085e-05, "val_loss": 3.0137057304382324, "train_action_mse_error": 41.39859390258789}
{"train_loss": 0.32893291115760803, "global_step": 46369, "epoch": 521, "lr": 9.30935087681661e-05}
{"train_loss": 0.19784210622310638, "global_step": 46370, "epoch": 521, "lr": 9.309321476403796e-05}
{"train_loss": 0.35460200905799866, "global_step": 46371, "epoch": 521, "lr": 9.309292075411646e-05}
{"train_loss": 0.3375729024410248, "global_step": 46372, "epoch": 521, "lr": 9.309262673840165e-05}
{"train_loss": 0.3351515531539917, "global_step": 46373, "epoch": 521, "lr": 9.309233271689354e-05}
{"train_loss": 0.2978767454624176, "global_step": 46374, "epoch": 521, "lr": 9.309203868959222e-05}
{"train_loss": 0.26961278915405273, "global_step": 46375, "epoch": 521, "lr": 9.309174465649769e-05}
{"train_loss": 0.20855943858623505, "global_step": 46376, "epoch": 521, "lr": 9.309145061760998e-05}
{"train_loss": 0.4214029610157013, "global_step": 46377, "epoch": 521, "lr": 9.309115657292918e-05}
{"train_loss": 0.2851606011390686, "global_step": 46378, "epoch": 521, "lr": 9.309086252245526e-05}
{"train_loss": 0.2743755877017975, "global_step": 46379, "epoch": 521, "lr": 9.309056846618834e-05}
{"train_loss": 0.33441126346588135, "global_step": 46380, "epoch": 521, "lr": 9.30902744041284e-05}
{"train_loss": 0.28347182273864746, "global_step": 46381, "epoch": 521, "lr": 9.30899803362755e-05}
{"train_loss": 0.35405731201171875, "global_step": 46382, "epoch": 521, "lr": 9.308968626262967e-05}
{"train_loss": 0.24285204708576202, "global_step": 46383, "epoch": 521, "lr": 9.308939218319097e-05}
{"train_loss": 0.2608531713485718, "global_step": 46384, "epoch": 521, "lr": 9.308909809795942e-05}
{"train_loss": 0.2904859185218811, "global_step": 46385, "epoch": 521, "lr": 9.308880400693506e-05}
{"train_loss": 0.3252342641353607, "global_step": 46386, "epoch": 521, "lr": 9.308850991011794e-05}
{"train_loss": 0.23374351859092712, "global_step": 46387, "epoch": 521, "lr": 9.30882158075081e-05}
{"train_loss": 0.24366146326065063, "global_step": 46388, "epoch": 521, "lr": 9.308792169910556e-05}
{"train_loss": 0.27849966287612915, "global_step": 46389, "epoch": 521, "lr": 9.308762758491039e-05}
{"train_loss": 0.2825043201446533, "global_step": 46390, "epoch": 521, "lr": 9.308733346492261e-05}
{"train_loss": 0.1961732804775238, "global_step": 46391, "epoch": 521, "lr": 9.308703933914227e-05}
{"train_loss": 0.3178236782550812, "global_step": 46392, "epoch": 521, "lr": 9.30867452075694e-05}
{"train_loss": 0.27205565571784973, "global_step": 46393, "epoch": 521, "lr": 9.308645107020401e-05}
{"train_loss": 0.2628800868988037, "global_step": 46394, "epoch": 521, "lr": 9.308615692704621e-05}
{"train_loss": 0.27336904406547546, "global_step": 46395, "epoch": 521, "lr": 9.308586277809599e-05}
{"train_loss": 0.3761431872844696, "global_step": 46396, "epoch": 521, "lr": 9.308556862335341e-05}
{"train_loss": 0.2915341854095459, "global_step": 46397, "epoch": 521, "lr": 9.308527446281847e-05}
{"train_loss": 0.24947677552700043, "global_step": 46398, "epoch": 521, "lr": 9.308498029649126e-05}
{"train_loss": 0.2671959400177002, "global_step": 46399, "epoch": 521, "lr": 9.308468612437181e-05}
{"train_loss": 0.2253425270318985, "global_step": 46400, "epoch": 521, "lr": 9.308439194646014e-05}
{"train_loss": 0.1942797303199768, "global_step": 46401, "epoch": 521, "lr": 9.308409776275631e-05}
{"train_loss": 0.23872879147529602, "global_step": 46402, "epoch": 521, "lr": 9.308380357326033e-05}
{"train_loss": 0.24904559552669525, "global_step": 46403, "epoch": 521, "lr": 9.308350937797226e-05}
{"train_loss": 0.3040229082107544, "global_step": 46404, "epoch": 521, "lr": 9.308321517689215e-05}
{"train_loss": 0.3668360412120819, "global_step": 46405, "epoch": 521, "lr": 9.308292097002001e-05}
{"train_loss": 0.28258275985717773, "global_step": 46406, "epoch": 521, "lr": 9.308262675735591e-05}
{"train_loss": 0.22687824070453644, "global_step": 46407, "epoch": 521, "lr": 9.308233253889986e-05}
{"train_loss": 0.29521068930625916, "global_step": 46408, "epoch": 521, "lr": 9.308203831465193e-05}
{"train_loss": 0.28345465660095215, "global_step": 46409, "epoch": 521, "lr": 9.308174408461215e-05}
{"train_loss": 0.2294130027294159, "global_step": 46410, "epoch": 521, "lr": 9.308144984878054e-05}
{"train_loss": 0.3586589992046356, "global_step": 46411, "epoch": 521, "lr": 9.308115560715715e-05}
{"train_loss": 0.23404304683208466, "global_step": 46412, "epoch": 521, "lr": 9.308086135974205e-05}
{"train_loss": 0.24157872796058655, "global_step": 46413, "epoch": 521, "lr": 9.308056710653523e-05}
{"train_loss": 0.24532198905944824, "global_step": 46414, "epoch": 521, "lr": 9.308027284753677e-05}
{"train_loss": 0.24860627949237823, "global_step": 46415, "epoch": 521, "lr": 9.307997858274668e-05}
{"train_loss": 0.3011247217655182, "global_step": 46416, "epoch": 521, "lr": 9.3079684312165e-05}
{"train_loss": 0.20398159325122833, "global_step": 46417, "epoch": 521, "lr": 9.30793900357918e-05}
{"train_loss": 0.22774842381477356, "global_step": 46418, "epoch": 521, "lr": 9.30790957536271e-05}
{"train_loss": 0.21905666589736938, "global_step": 46419, "epoch": 521, "lr": 9.307880146567094e-05}
{"train_loss": 0.27608948945999146, "global_step": 46420, "epoch": 521, "lr": 9.307850717192336e-05}
{"train_loss": 0.3173981308937073, "global_step": 46421, "epoch": 521, "lr": 9.307821287238438e-05}
{"train_loss": 0.2164551466703415, "global_step": 46422, "epoch": 521, "lr": 9.307791856705409e-05}
{"train_loss": 0.2547016441822052, "global_step": 46423, "epoch": 521, "lr": 9.307762425593249e-05}
{"train_loss": 0.30990704894065857, "global_step": 46424, "epoch": 521, "lr": 9.30773299390196e-05}
{"train_loss": 0.23626849055290222, "global_step": 46425, "epoch": 521, "lr": 9.307703561631554e-05}
{"train_loss": 0.333869993686676, "global_step": 46426, "epoch": 521, "lr": 9.307674128782026e-05}
{"train_loss": 0.37926825881004333, "global_step": 46427, "epoch": 521, "lr": 9.307644695353385e-05}
{"train_loss": 0.21816299855709076, "global_step": 46428, "epoch": 521, "lr": 9.307615261345632e-05}
{"train_loss": 0.27869752049446106, "global_step": 46429, "epoch": 521, "lr": 9.307585826758775e-05}
{"train_loss": 0.26097023487091064, "global_step": 46430, "epoch": 521, "lr": 9.307556391592815e-05}
{"train_loss": 0.18518513441085815, "global_step": 46431, "epoch": 521, "lr": 9.307526955847756e-05}
{"train_loss": 0.25314605236053467, "global_step": 46432, "epoch": 521, "lr": 9.307497519523603e-05}
{"train_loss": 0.33787021040916443, "global_step": 46433, "epoch": 521, "lr": 9.307468082620358e-05}
{"train_loss": 0.31640198826789856, "global_step": 46434, "epoch": 521, "lr": 9.307438645138027e-05}
{"train_loss": 0.2677697539329529, "global_step": 46435, "epoch": 521, "lr": 9.307409207076614e-05}
{"train_loss": 0.3250746726989746, "global_step": 46436, "epoch": 521, "lr": 9.307379768436123e-05}
{"train_loss": 0.312811940908432, "global_step": 46437, "epoch": 521, "lr": 9.307350329216556e-05}
{"train_loss": 0.2150679975748062, "global_step": 46438, "epoch": 521, "lr": 9.307320889417918e-05}
{"train_loss": 0.2323673665523529, "global_step": 46439, "epoch": 521, "lr": 9.307291449040215e-05}
{"train_loss": 0.21015939116477966, "global_step": 46440, "epoch": 521, "lr": 9.307262008083448e-05}
{"train_loss": 0.2663908898830414, "global_step": 46441, "epoch": 521, "lr": 9.307232566547623e-05}
{"train_loss": 0.2620287239551544, "global_step": 46442, "epoch": 521, "lr": 9.307203124432742e-05}
{"train_loss": 0.29426470398902893, "global_step": 46443, "epoch": 521, "lr": 9.30717368173881e-05}
{"train_loss": 0.30058079957962036, "global_step": 46444, "epoch": 521, "lr": 9.307144238465831e-05}
{"train_loss": 0.27766868472099304, "global_step": 46445, "epoch": 521, "lr": 9.307114794613809e-05}
{"train_loss": 0.2999196946620941, "global_step": 46446, "epoch": 521, "lr": 9.30708535018275e-05}
{"train_loss": 0.3054957985877991, "global_step": 46447, "epoch": 521, "lr": 9.307055905172653e-05}
{"train_loss": 0.3251416087150574, "global_step": 46448, "epoch": 521, "lr": 9.307026459583526e-05}
{"train_loss": 0.2850051820278168, "global_step": 46449, "epoch": 521, "lr": 9.306997013415371e-05}
{"train_loss": 0.3127840757369995, "global_step": 46450, "epoch": 521, "lr": 9.306967566668195e-05}
{"train_loss": 0.444684773683548, "global_step": 46451, "epoch": 521, "lr": 9.306938119341997e-05}
{"train_loss": 0.2589173913002014, "global_step": 46452, "epoch": 521, "lr": 9.306908671436785e-05}
{"train_loss": 0.34737148880958557, "global_step": 46453, "epoch": 521, "lr": 9.306879222952562e-05}
{"train_loss": 0.2326279729604721, "global_step": 46454, "epoch": 521, "lr": 9.30684977388933e-05}
{"train_loss": 0.2103380411863327, "global_step": 46455, "epoch": 521, "lr": 9.306820324247096e-05}
{"train_loss": 0.17763547599315643, "global_step": 46456, "epoch": 521, "lr": 9.306790874025862e-05}
{"train_loss": 0.2779903853877207, "global_step": 46457, "epoch": 521, "lr": 9.306761423225632e-05, "val_loss": 2.927882671356201}
{"train_loss": 0.41159698367118835, "global_step": 46458, "epoch": 522, "lr": 9.306731971846411e-05}
{"train_loss": 0.3284459114074707, "global_step": 46459, "epoch": 522, "lr": 9.306702519888203e-05}
{"train_loss": 0.34130486845970154, "global_step": 46460, "epoch": 522, "lr": 9.30667306735101e-05}
{"train_loss": 0.3115362226963043, "global_step": 46461, "epoch": 522, "lr": 9.306643614234839e-05}
{"train_loss": 0.22107729315757751, "global_step": 46462, "epoch": 522, "lr": 9.30661416053969e-05}
{"train_loss": 0.20824791491031647, "global_step": 46463, "epoch": 522, "lr": 9.306584706265571e-05}
{"train_loss": 0.3421562612056732, "global_step": 46464, "epoch": 522, "lr": 9.306555251412483e-05}
{"train_loss": 0.2801997661590576, "global_step": 46465, "epoch": 522, "lr": 9.306525795980432e-05}
{"train_loss": 0.21747806668281555, "global_step": 46466, "epoch": 522, "lr": 9.306496339969421e-05}
{"train_loss": 0.2512173652648926, "global_step": 46467, "epoch": 522, "lr": 9.306466883379454e-05}
{"train_loss": 0.35485711693763733, "global_step": 46468, "epoch": 522, "lr": 9.306437426210535e-05}
{"train_loss": 0.32764115929603577, "global_step": 46469, "epoch": 522, "lr": 9.306407968462668e-05}
{"train_loss": 0.21623291075229645, "global_step": 46470, "epoch": 522, "lr": 9.306378510135856e-05}
{"train_loss": 0.25690287351608276, "global_step": 46471, "epoch": 522, "lr": 9.306349051230105e-05}
{"train_loss": 0.22908245027065277, "global_step": 46472, "epoch": 522, "lr": 9.306319591745419e-05}
{"train_loss": 0.3707562983036041, "global_step": 46473, "epoch": 522, "lr": 9.306290131681799e-05}
{"train_loss": 0.45828115940093994, "global_step": 46474, "epoch": 522, "lr": 9.306260671039251e-05}
{"train_loss": 0.3544657826423645, "global_step": 46475, "epoch": 522, "lr": 9.306231209817778e-05}
{"train_loss": 0.384016215801239, "global_step": 46476, "epoch": 522, "lr": 9.306201748017385e-05}
{"train_loss": 0.41876447200775146, "global_step": 46477, "epoch": 522, "lr": 9.306172285638077e-05}
{"train_loss": 0.31063112616539, "global_step": 46478, "epoch": 522, "lr": 9.306142822679857e-05}
{"train_loss": 0.32219961285591125, "global_step": 46479, "epoch": 522, "lr": 9.306113359142726e-05}
{"train_loss": 0.36495959758758545, "global_step": 46480, "epoch": 522, "lr": 9.306083895026693e-05}
{"train_loss": 0.2829608917236328, "global_step": 46481, "epoch": 522, "lr": 9.30605443033176e-05}
{"train_loss": 0.21591076254844666, "global_step": 46482, "epoch": 522, "lr": 9.306024965057927e-05}
{"train_loss": 0.3243021070957184, "global_step": 46483, "epoch": 522, "lr": 9.305995499205204e-05}
{"train_loss": 0.2912496328353882, "global_step": 46484, "epoch": 522, "lr": 9.305966032773591e-05}
{"train_loss": 0.2571476101875305, "global_step": 46485, "epoch": 522, "lr": 9.305936565763094e-05}
{"train_loss": 0.32941821217536926, "global_step": 46486, "epoch": 522, "lr": 9.305907098173716e-05}
{"train_loss": 0.30495157837867737, "global_step": 46487, "epoch": 522, "lr": 9.305877630005462e-05}
{"train_loss": 0.4052042067050934, "global_step": 46488, "epoch": 522, "lr": 9.305848161258335e-05}
{"train_loss": 0.3213250935077667, "global_step": 46489, "epoch": 522, "lr": 9.305818691932339e-05}
{"train_loss": 0.3642074465751648, "global_step": 46490, "epoch": 522, "lr": 9.30578922202748e-05}
{"train_loss": 0.2891604006290436, "global_step": 46491, "epoch": 522, "lr": 9.305759751543757e-05}
{"train_loss": 0.2785283029079437, "global_step": 46492, "epoch": 522, "lr": 9.305730280481178e-05}
{"train_loss": 0.22259946167469025, "global_step": 46493, "epoch": 522, "lr": 9.305700808839748e-05}
{"train_loss": 0.2976255714893341, "global_step": 46494, "epoch": 522, "lr": 9.305671336619467e-05}
{"train_loss": 0.3727802038192749, "global_step": 46495, "epoch": 522, "lr": 9.30564186382034e-05}
{"train_loss": 0.361381471157074, "global_step": 46496, "epoch": 522, "lr": 9.305612390442375e-05}
{"train_loss": 0.3653731346130371, "global_step": 46497, "epoch": 522, "lr": 9.305582916485571e-05}
{"train_loss": 0.27687162160873413, "global_step": 46498, "epoch": 522, "lr": 9.305553441949935e-05}
{"train_loss": 0.2949449121952057, "global_step": 46499, "epoch": 522, "lr": 9.305523966835467e-05}
{"train_loss": 0.2885168790817261, "global_step": 46500, "epoch": 522, "lr": 9.305494491142177e-05}
{"train_loss": 0.39754512906074524, "global_step": 46501, "epoch": 522, "lr": 9.305465014870064e-05}
{"train_loss": 0.355991005897522, "global_step": 46502, "epoch": 522, "lr": 9.305435538019135e-05}
{"train_loss": 0.31642699241638184, "global_step": 46503, "epoch": 522, "lr": 9.305406060589393e-05}
{"train_loss": 0.40287554264068604, "global_step": 46504, "epoch": 522, "lr": 9.30537658258084e-05}
{"train_loss": 0.17601437866687775, "global_step": 46505, "epoch": 522, "lr": 9.305347103993482e-05}
{"train_loss": 0.2977893054485321, "global_step": 46506, "epoch": 522, "lr": 9.305317624827322e-05}
{"train_loss": 0.34430205821990967, "global_step": 46507, "epoch": 522, "lr": 9.305288145082364e-05}
{"train_loss": 0.36848515272140503, "global_step": 46508, "epoch": 522, "lr": 9.305258664758616e-05}
{"train_loss": 0.3417479395866394, "global_step": 46509, "epoch": 522, "lr": 9.305229183856075e-05}
{"train_loss": 0.412504643201828, "global_step": 46510, "epoch": 522, "lr": 9.30519970237475e-05}
{"train_loss": 0.24036985635757446, "global_step": 46511, "epoch": 522, "lr": 9.305170220314643e-05}
{"train_loss": 0.36883988976478577, "global_step": 46512, "epoch": 522, "lr": 9.305140737675759e-05}
{"train_loss": 0.3591987192630768, "global_step": 46513, "epoch": 522, "lr": 9.3051112544581e-05}
{"train_loss": 0.3102676272392273, "global_step": 46514, "epoch": 522, "lr": 9.305081770661672e-05}
{"train_loss": 0.3320258557796478, "global_step": 46515, "epoch": 522, "lr": 9.305052286286478e-05}
{"train_loss": 0.31419143080711365, "global_step": 46516, "epoch": 522, "lr": 9.305022801332522e-05}
{"train_loss": 0.3142741322517395, "global_step": 46517, "epoch": 522, "lr": 9.304993315799808e-05}
{"train_loss": 0.331384539604187, "global_step": 46518, "epoch": 522, "lr": 9.30496382968834e-05}
{"train_loss": 0.2891351282596588, "global_step": 46519, "epoch": 522, "lr": 9.304934342998124e-05}
{"train_loss": 0.29261672496795654, "global_step": 46520, "epoch": 522, "lr": 9.304904855729159e-05}
{"train_loss": 0.3224914073944092, "global_step": 46521, "epoch": 522, "lr": 9.304875367881454e-05}
{"train_loss": 0.34369972348213196, "global_step": 46522, "epoch": 522, "lr": 9.30484587945501e-05}
{"train_loss": 0.42536309361457825, "global_step": 46523, "epoch": 522, "lr": 9.304816390449833e-05}
{"train_loss": 0.22918517887592316, "global_step": 46524, "epoch": 522, "lr": 9.304786900865927e-05}
{"train_loss": 0.32470065355300903, "global_step": 46525, "epoch": 522, "lr": 9.304757410703293e-05}
{"train_loss": 0.2939945161342621, "global_step": 46526, "epoch": 522, "lr": 9.304727919961938e-05}
{"train_loss": 0.3332148492336273, "global_step": 46527, "epoch": 522, "lr": 9.304698428641862e-05}
{"train_loss": 0.2591398358345032, "global_step": 46528, "epoch": 522, "lr": 9.304668936743075e-05}
{"train_loss": 0.2726249694824219, "global_step": 46529, "epoch": 522, "lr": 9.304639444265576e-05}
{"train_loss": 0.2531108260154724, "global_step": 46530, "epoch": 522, "lr": 9.30460995120937e-05}
{"train_loss": 0.2272477000951767, "global_step": 46531, "epoch": 522, "lr": 9.304580457574464e-05}
{"train_loss": 0.27458655834198, "global_step": 46532, "epoch": 522, "lr": 9.304550963360859e-05}
{"train_loss": 0.26999735832214355, "global_step": 46533, "epoch": 522, "lr": 9.304521468568558e-05}
{"train_loss": 0.31102439761161804, "global_step": 46534, "epoch": 522, "lr": 9.304491973197569e-05}
{"train_loss": 0.3722812533378601, "global_step": 46535, "epoch": 522, "lr": 9.304462477247891e-05}
{"train_loss": 0.34761810302734375, "global_step": 46536, "epoch": 522, "lr": 9.304432980719532e-05}
{"train_loss": 0.3914678692817688, "global_step": 46537, "epoch": 522, "lr": 9.304403483612494e-05}
{"train_loss": 0.37344890832901, "global_step": 46538, "epoch": 522, "lr": 9.304373985926781e-05}
{"train_loss": 0.3055133819580078, "global_step": 46539, "epoch": 522, "lr": 9.304344487662398e-05}
{"train_loss": 0.29988157749176025, "global_step": 46540, "epoch": 522, "lr": 9.304314988819349e-05}
{"train_loss": 0.2196914404630661, "global_step": 46541, "epoch": 522, "lr": 9.304285489397636e-05}
{"train_loss": 0.2297976016998291, "global_step": 46542, "epoch": 522, "lr": 9.304255989397265e-05}
{"train_loss": 0.3055798411369324, "global_step": 46543, "epoch": 522, "lr": 9.30422648881824e-05}
{"train_loss": 0.3118031322956085, "global_step": 46544, "epoch": 522, "lr": 9.304196987660563e-05}
{"train_loss": 0.26179665327072144, "global_step": 46545, "epoch": 522, "lr": 9.304167485924237e-05}
{"train_loss": 0.31363955274056854, "global_step": 46546, "epoch": 522, "lr": 9.30413798360927e-05, "val_loss": 2.798156261444092}
{"train_loss": 0.22949224710464478, "global_step": 46547, "epoch": 523, "lr": 9.304108480715666e-05}
{"train_loss": 0.2644129693508148, "global_step": 46548, "epoch": 523, "lr": 9.304078977243426e-05}
{"train_loss": 0.36138349771499634, "global_step": 46549, "epoch": 523, "lr": 9.304049473192556e-05}
{"train_loss": 0.298015296459198, "global_step": 46550, "epoch": 523, "lr": 9.304019968563057e-05}
{"train_loss": 0.34553831815719604, "global_step": 46551, "epoch": 523, "lr": 9.303990463354935e-05}
{"train_loss": 0.3491433262825012, "global_step": 46552, "epoch": 523, "lr": 9.303960957568195e-05}
{"train_loss": 0.27297359704971313, "global_step": 46553, "epoch": 523, "lr": 9.30393145120284e-05}
{"train_loss": 0.2924901843070984, "global_step": 46554, "epoch": 523, "lr": 9.303901944258872e-05}
{"train_loss": 0.35059890151023865, "global_step": 46555, "epoch": 523, "lr": 9.3038724367363e-05}
{"train_loss": 0.23310169577598572, "global_step": 46556, "epoch": 523, "lr": 9.303842928635123e-05}
{"train_loss": 0.2787119150161743, "global_step": 46557, "epoch": 523, "lr": 9.303813419955347e-05}
{"train_loss": 0.2897526025772095, "global_step": 46558, "epoch": 523, "lr": 9.303783910696975e-05}
{"train_loss": 0.34353870153427124, "global_step": 46559, "epoch": 523, "lr": 9.303754400860011e-05}
{"train_loss": 0.29656195640563965, "global_step": 46560, "epoch": 523, "lr": 9.303724890444463e-05}
{"train_loss": 0.23420368134975433, "global_step": 46561, "epoch": 523, "lr": 9.303695379450329e-05}
{"train_loss": 0.2663048505783081, "global_step": 46562, "epoch": 523, "lr": 9.303665867877616e-05}
{"train_loss": 0.2613663375377655, "global_step": 46563, "epoch": 523, "lr": 9.303636355726328e-05}
{"train_loss": 0.2072516828775406, "global_step": 46564, "epoch": 523, "lr": 9.303606842996468e-05}
{"train_loss": 0.32858192920684814, "global_step": 46565, "epoch": 523, "lr": 9.303577329688042e-05}
{"train_loss": 0.24344922602176666, "global_step": 46566, "epoch": 523, "lr": 9.303547815801052e-05}
{"train_loss": 0.3318018317222595, "global_step": 46567, "epoch": 523, "lr": 9.3035183013355e-05}
{"train_loss": 0.3760996460914612, "global_step": 46568, "epoch": 523, "lr": 9.303488786291396e-05}
{"train_loss": 0.27525436878204346, "global_step": 46569, "epoch": 523, "lr": 9.303459270668738e-05}
{"train_loss": 0.3618553578853607, "global_step": 46570, "epoch": 523, "lr": 9.303429754467532e-05}
{"train_loss": 0.2516491115093231, "global_step": 46571, "epoch": 523, "lr": 9.303400237687785e-05}
{"train_loss": 0.34050416946411133, "global_step": 46572, "epoch": 523, "lr": 9.303370720329496e-05}
{"train_loss": 0.2575280964374542, "global_step": 46573, "epoch": 523, "lr": 9.303341202392672e-05}
{"train_loss": 0.26693791151046753, "global_step": 46574, "epoch": 523, "lr": 9.303311683877316e-05}
{"train_loss": 0.2805464267730713, "global_step": 46575, "epoch": 523, "lr": 9.303282164783432e-05}
{"train_loss": 0.30174344778060913, "global_step": 46576, "epoch": 523, "lr": 9.303252645111024e-05}
{"train_loss": 0.21713513135910034, "global_step": 46577, "epoch": 523, "lr": 9.303223124860097e-05}
{"train_loss": 0.312823086977005, "global_step": 46578, "epoch": 523, "lr": 9.303193604030653e-05}
{"train_loss": 0.2942561209201813, "global_step": 46579, "epoch": 523, "lr": 9.303164082622699e-05}
{"train_loss": 0.2367067039012909, "global_step": 46580, "epoch": 523, "lr": 9.303134560636235e-05}
{"train_loss": 0.22758838534355164, "global_step": 46581, "epoch": 523, "lr": 9.303105038071267e-05}
{"train_loss": 0.2597053647041321, "global_step": 46582, "epoch": 523, "lr": 9.303075514927801e-05}
{"train_loss": 0.23436108231544495, "global_step": 46583, "epoch": 523, "lr": 9.303045991205837e-05}
{"train_loss": 0.3106289803981781, "global_step": 46584, "epoch": 523, "lr": 9.303016466905383e-05}
{"train_loss": 0.3463285565376282, "global_step": 46585, "epoch": 523, "lr": 9.30298694202644e-05}
{"train_loss": 0.24039658904075623, "global_step": 46586, "epoch": 523, "lr": 9.302957416569011e-05}
{"train_loss": 0.3302995562553406, "global_step": 46587, "epoch": 523, "lr": 9.302927890533105e-05}
{"train_loss": 0.33743518590927124, "global_step": 46588, "epoch": 523, "lr": 9.30289836391872e-05}
{"train_loss": 0.36557525396347046, "global_step": 46589, "epoch": 523, "lr": 9.302868836725863e-05}
{"train_loss": 0.2173299491405487, "global_step": 46590, "epoch": 523, "lr": 9.302839308954539e-05}
{"train_loss": 0.3358243703842163, "global_step": 46591, "epoch": 523, "lr": 9.30280978060475e-05}
{"train_loss": 0.22730642557144165, "global_step": 46592, "epoch": 523, "lr": 9.302780251676501e-05}
{"train_loss": 0.24802124500274658, "global_step": 46593, "epoch": 523, "lr": 9.302750722169795e-05}
{"train_loss": 0.23316289484500885, "global_step": 46594, "epoch": 523, "lr": 9.302721192084637e-05}
{"train_loss": 0.30522671341896057, "global_step": 46595, "epoch": 523, "lr": 9.30269166142103e-05}
{"train_loss": 0.3075520992279053, "global_step": 46596, "epoch": 523, "lr": 9.30266213017898e-05}
{"train_loss": 0.32205942273139954, "global_step": 46597, "epoch": 523, "lr": 9.302632598358488e-05}
{"train_loss": 0.3233661353588104, "global_step": 46598, "epoch": 523, "lr": 9.30260306595956e-05}
{"train_loss": 0.29558515548706055, "global_step": 46599, "epoch": 523, "lr": 9.3025735329822e-05}
{"train_loss": 0.24216967821121216, "global_step": 46600, "epoch": 523, "lr": 9.302543999426412e-05}
{"train_loss": 0.26116976141929626, "global_step": 46601, "epoch": 523, "lr": 9.302514465292198e-05}
{"train_loss": 0.229244202375412, "global_step": 46602, "epoch": 523, "lr": 9.302484930579565e-05}
{"train_loss": 0.3117615580558777, "global_step": 46603, "epoch": 523, "lr": 9.302455395288515e-05}
{"train_loss": 0.13659149408340454, "global_step": 46604, "epoch": 523, "lr": 9.30242585941905e-05}
{"train_loss": 0.21406865119934082, "global_step": 46605, "epoch": 523, "lr": 9.302396322971178e-05}
{"train_loss": 0.37188172340393066, "global_step": 46606, "epoch": 523, "lr": 9.3023667859449e-05}
{"train_loss": 0.22871185839176178, "global_step": 46607, "epoch": 523, "lr": 9.302337248340223e-05}
{"train_loss": 0.3436965346336365, "global_step": 46608, "epoch": 523, "lr": 9.302307710157147e-05}
{"train_loss": 0.24775658547878265, "global_step": 46609, "epoch": 523, "lr": 9.302278171395681e-05}
{"train_loss": 0.31835460662841797, "global_step": 46610, "epoch": 523, "lr": 9.302248632055825e-05}
{"train_loss": 0.3097813129425049, "global_step": 46611, "epoch": 523, "lr": 9.302219092137584e-05}
{"train_loss": 0.2880518436431885, "global_step": 46612, "epoch": 523, "lr": 9.302189551640962e-05}
{"train_loss": 0.3167499005794525, "global_step": 46613, "epoch": 523, "lr": 9.302160010565964e-05}
{"train_loss": 0.31196892261505127, "global_step": 46614, "epoch": 523, "lr": 9.302130468912592e-05}
{"train_loss": 0.18476246297359467, "global_step": 46615, "epoch": 523, "lr": 9.30210092668085e-05}
{"train_loss": 0.293162077665329, "global_step": 46616, "epoch": 523, "lr": 9.302071383870742e-05}
{"train_loss": 0.23254436254501343, "global_step": 46617, "epoch": 523, "lr": 9.302041840482277e-05}
{"train_loss": 0.37441250681877136, "global_step": 46618, "epoch": 523, "lr": 9.302012296515453e-05}
{"train_loss": 0.2902933359146118, "global_step": 46619, "epoch": 523, "lr": 9.301982751970276e-05}
{"train_loss": 0.23525039851665497, "global_step": 46620, "epoch": 523, "lr": 9.30195320684675e-05}
{"train_loss": 0.3067944347858429, "global_step": 46621, "epoch": 523, "lr": 9.301923661144878e-05}
{"train_loss": 0.2721445560455322, "global_step": 46622, "epoch": 523, "lr": 9.301894114864664e-05}
{"train_loss": 0.26772308349609375, "global_step": 46623, "epoch": 523, "lr": 9.301864568006114e-05}
{"train_loss": 0.3342019021511078, "global_step": 46624, "epoch": 523, "lr": 9.301835020569232e-05}
{"train_loss": 0.3005565106868744, "global_step": 46625, "epoch": 523, "lr": 9.301805472554018e-05}
{"train_loss": 0.26135560870170593, "global_step": 46626, "epoch": 523, "lr": 9.30177592396048e-05}
{"train_loss": 0.17543204128742218, "global_step": 46627, "epoch": 523, "lr": 9.301746374788621e-05}
{"train_loss": 0.27563437819480896, "global_step": 46628, "epoch": 523, "lr": 9.301716825038444e-05}
{"train_loss": 0.24220572412014008, "global_step": 46629, "epoch": 523, "lr": 9.301687274709954e-05}
{"train_loss": 0.44420915842056274, "global_step": 46630, "epoch": 523, "lr": 9.301657723803155e-05}
{"train_loss": 0.3630374073982239, "global_step": 46631, "epoch": 523, "lr": 9.301628172318052e-05}
{"train_loss": 0.3272207975387573, "global_step": 46632, "epoch": 523, "lr": 9.301598620254644e-05}
{"train_loss": 0.32269519567489624, "global_step": 46633, "epoch": 523, "lr": 9.301569067612941e-05}
{"train_loss": 0.269103467464447, "global_step": 46634, "epoch": 523, "lr": 9.301539514392943e-05}
{"train_loss": 0.28684730107864637, "global_step": 46635, "epoch": 523, "lr": 9.301509960594657e-05, "val_loss": 2.9021191596984863}
{"train_loss": 0.259889155626297, "global_step": 46636, "epoch": 524, "lr": 9.301480406218084e-05}
{"train_loss": 0.2640100419521332, "global_step": 46637, "epoch": 524, "lr": 9.30145085126323e-05}
{"train_loss": 0.23453542590141296, "global_step": 46638, "epoch": 524, "lr": 9.301421295730098e-05}
{"train_loss": 0.33358854055404663, "global_step": 46639, "epoch": 524, "lr": 9.301391739618693e-05}
{"train_loss": 0.3111708164215088, "global_step": 46640, "epoch": 524, "lr": 9.301362182929017e-05}
{"train_loss": 0.32883337140083313, "global_step": 46641, "epoch": 524, "lr": 9.301332625661077e-05}
{"train_loss": 0.26580485701560974, "global_step": 46642, "epoch": 524, "lr": 9.301303067814873e-05}
{"train_loss": 0.2691883146762848, "global_step": 46643, "epoch": 524, "lr": 9.301273509390413e-05}
{"train_loss": 0.21019287407398224, "global_step": 46644, "epoch": 524, "lr": 9.301243950387699e-05}
{"train_loss": 0.37626516819000244, "global_step": 46645, "epoch": 524, "lr": 9.301214390806736e-05}
{"train_loss": 0.35523879528045654, "global_step": 46646, "epoch": 524, "lr": 9.301184830647524e-05}
{"train_loss": 0.32028451561927795, "global_step": 46647, "epoch": 524, "lr": 9.301155269910074e-05}
{"train_loss": 0.3803189992904663, "global_step": 46648, "epoch": 524, "lr": 9.301125708594384e-05}
{"train_loss": 0.38657352328300476, "global_step": 46649, "epoch": 524, "lr": 9.301096146700461e-05}
{"train_loss": 0.3004509508609772, "global_step": 46650, "epoch": 524, "lr": 9.301066584228306e-05}
{"train_loss": 0.46488669514656067, "global_step": 46651, "epoch": 524, "lr": 9.301037021177927e-05}
{"train_loss": 0.31430864334106445, "global_step": 46652, "epoch": 524, "lr": 9.301007457549326e-05}
{"train_loss": 0.2738352119922638, "global_step": 46653, "epoch": 524, "lr": 9.300977893342506e-05}
{"train_loss": 0.23445194959640503, "global_step": 46654, "epoch": 524, "lr": 9.300948328557473e-05}
{"train_loss": 0.2686612010002136, "global_step": 46655, "epoch": 524, "lr": 9.300918763194228e-05}
{"train_loss": 0.36475393176078796, "global_step": 46656, "epoch": 524, "lr": 9.30088919725278e-05}
{"train_loss": 0.21034298837184906, "global_step": 46657, "epoch": 524, "lr": 9.300859630733127e-05}
{"train_loss": 0.3208009600639343, "global_step": 46658, "epoch": 524, "lr": 9.300830063635277e-05}
{"train_loss": 0.23583410680294037, "global_step": 46659, "epoch": 524, "lr": 9.300800495959234e-05}
{"train_loss": 0.24119389057159424, "global_step": 46660, "epoch": 524, "lr": 9.300770927704999e-05}
{"train_loss": 0.31601643562316895, "global_step": 46661, "epoch": 524, "lr": 9.300741358872578e-05}
{"train_loss": 0.338815838098526, "global_step": 46662, "epoch": 524, "lr": 9.300711789461975e-05}
{"train_loss": 0.30106866359710693, "global_step": 46663, "epoch": 524, "lr": 9.300682219473193e-05}
{"train_loss": 0.3145391643047333, "global_step": 46664, "epoch": 524, "lr": 9.300652648906239e-05}
{"train_loss": 0.26647865772247314, "global_step": 46665, "epoch": 524, "lr": 9.300623077761113e-05}
{"train_loss": 0.35257434844970703, "global_step": 46666, "epoch": 524, "lr": 9.30059350603782e-05}
{"train_loss": 0.24485629796981812, "global_step": 46667, "epoch": 524, "lr": 9.300563933736366e-05}
{"train_loss": 0.23412318527698517, "global_step": 46668, "epoch": 524, "lr": 9.300534360856752e-05}
{"train_loss": 0.25221168994903564, "global_step": 46669, "epoch": 524, "lr": 9.300504787398984e-05}
{"train_loss": 0.2734270989894867, "global_step": 46670, "epoch": 524, "lr": 9.300475213363067e-05}
{"train_loss": 0.20734739303588867, "global_step": 46671, "epoch": 524, "lr": 9.300445638749002e-05}
{"train_loss": 0.2715345323085785, "global_step": 46672, "epoch": 524, "lr": 9.300416063556796e-05}
{"train_loss": 0.3282933831214905, "global_step": 46673, "epoch": 524, "lr": 9.300386487786449e-05}
{"train_loss": 0.28341880440711975, "global_step": 46674, "epoch": 524, "lr": 9.30035691143797e-05}
{"train_loss": 0.2375747114419937, "global_step": 46675, "epoch": 524, "lr": 9.300327334511361e-05}
{"train_loss": 0.20617985725402832, "global_step": 46676, "epoch": 524, "lr": 9.300297757006623e-05}
{"train_loss": 0.46160054206848145, "global_step": 46677, "epoch": 524, "lr": 9.300268178923763e-05}
{"train_loss": 0.2596640884876251, "global_step": 46678, "epoch": 524, "lr": 9.300238600262785e-05}
{"train_loss": 0.32377690076828003, "global_step": 46679, "epoch": 524, "lr": 9.300209021023692e-05}
{"train_loss": 0.2657192051410675, "global_step": 46680, "epoch": 524, "lr": 9.300179441206488e-05}
{"train_loss": 0.27245938777923584, "global_step": 46681, "epoch": 524, "lr": 9.300149860811178e-05}
{"train_loss": 0.29321712255477905, "global_step": 46682, "epoch": 524, "lr": 9.300120279837765e-05}
{"train_loss": 0.22374875843524933, "global_step": 46683, "epoch": 524, "lr": 9.300090698286252e-05}
{"train_loss": 0.2549583315849304, "global_step": 46684, "epoch": 524, "lr": 9.300061116156646e-05}
{"train_loss": 0.4472096264362335, "global_step": 46685, "epoch": 524, "lr": 9.300031533448949e-05}
{"train_loss": 0.27631187438964844, "global_step": 46686, "epoch": 524, "lr": 9.300001950163164e-05}
{"train_loss": 0.2966291904449463, "global_step": 46687, "epoch": 524, "lr": 9.299972366299297e-05}
{"train_loss": 0.2846927046775818, "global_step": 46688, "epoch": 524, "lr": 9.299942781857352e-05}
{"train_loss": 0.2434452474117279, "global_step": 46689, "epoch": 524, "lr": 9.29991319683733e-05}
{"train_loss": 0.2849271297454834, "global_step": 46690, "epoch": 524, "lr": 9.299883611239237e-05}
{"train_loss": 0.2647404074668884, "global_step": 46691, "epoch": 524, "lr": 9.299854025063079e-05}
{"train_loss": 0.2910122573375702, "global_step": 46692, "epoch": 524, "lr": 9.299824438308857e-05}
{"train_loss": 0.28529417514801025, "global_step": 46693, "epoch": 524, "lr": 9.299794850976577e-05}
{"train_loss": 0.28806665539741516, "global_step": 46694, "epoch": 524, "lr": 9.299765263066241e-05}
{"train_loss": 0.21951311826705933, "global_step": 46695, "epoch": 524, "lr": 9.299735674577854e-05}
{"train_loss": 0.3609757721424103, "global_step": 46696, "epoch": 524, "lr": 9.299706085511419e-05}
{"train_loss": 0.26018229126930237, "global_step": 46697, "epoch": 524, "lr": 9.299676495866942e-05}
{"train_loss": 0.3097110986709595, "global_step": 46698, "epoch": 524, "lr": 9.299646905644427e-05}
{"train_loss": 0.2936341166496277, "global_step": 46699, "epoch": 524, "lr": 9.299617314843876e-05}
{"train_loss": 0.3045520484447479, "global_step": 46700, "epoch": 524, "lr": 9.299587723465293e-05}
{"train_loss": 0.25107789039611816, "global_step": 46701, "epoch": 524, "lr": 9.299558131508683e-05}
{"train_loss": 0.2825106382369995, "global_step": 46702, "epoch": 524, "lr": 9.299528538974051e-05}
{"train_loss": 0.2513580620288849, "global_step": 46703, "epoch": 524, "lr": 9.299498945861398e-05}
{"train_loss": 0.30727025866508484, "global_step": 46704, "epoch": 524, "lr": 9.299469352170732e-05}
{"train_loss": 0.3021780848503113, "global_step": 46705, "epoch": 524, "lr": 9.299439757902053e-05}
{"train_loss": 0.3535816967487335, "global_step": 46706, "epoch": 524, "lr": 9.299410163055369e-05}
{"train_loss": 0.3462110161781311, "global_step": 46707, "epoch": 524, "lr": 9.29938056763068e-05}
{"train_loss": 0.31517493724823, "global_step": 46708, "epoch": 524, "lr": 9.299350971627991e-05}
{"train_loss": 0.2965448200702667, "global_step": 46709, "epoch": 524, "lr": 9.299321375047307e-05}
{"train_loss": 0.29437118768692017, "global_step": 46710, "epoch": 524, "lr": 9.299291777888633e-05}
{"train_loss": 0.3981603980064392, "global_step": 46711, "epoch": 524, "lr": 9.299262180151969e-05}
{"train_loss": 0.3190775513648987, "global_step": 46712, "epoch": 524, "lr": 9.299232581837325e-05}
{"train_loss": 0.4739784598350525, "global_step": 46713, "epoch": 524, "lr": 9.2992029829447e-05}
{"train_loss": 0.27300041913986206, "global_step": 46714, "epoch": 524, "lr": 9.299173383474098e-05}
{"train_loss": 0.26459741592407227, "global_step": 46715, "epoch": 524, "lr": 9.299143783425525e-05}
{"train_loss": 0.24359869956970215, "global_step": 46716, "epoch": 524, "lr": 9.299114182798987e-05}
{"train_loss": 0.29078370332717896, "global_step": 46717, "epoch": 524, "lr": 9.299084581594483e-05}
{"train_loss": 0.30069202184677124, "global_step": 46718, "epoch": 524, "lr": 9.299054979812021e-05}
{"train_loss": 0.3107677400112152, "global_step": 46719, "epoch": 524, "lr": 9.299025377451604e-05}
{"train_loss": 0.2985357642173767, "global_step": 46720, "epoch": 524, "lr": 9.298995774513233e-05}
{"train_loss": 0.3441656827926636, "global_step": 46721, "epoch": 524, "lr": 9.298966170996917e-05}
{"train_loss": 0.24520301818847656, "global_step": 46722, "epoch": 524, "lr": 9.298936566902655e-05}
{"train_loss": 0.2635747790336609, "global_step": 46723, "epoch": 524, "lr": 9.298906962230454e-05}
{"train_loss": 0.2957701786850276, "global_step": 46724, "epoch": 524, "lr": 9.298877356980318e-05, "val_loss": 2.8637397289276123}
{"train_loss": 0.2346029430627823, "global_step": 46725, "epoch": 525, "lr": 9.298847751152251e-05}
{"train_loss": 0.28292879462242126, "global_step": 46726, "epoch": 525, "lr": 9.298818144746255e-05}
{"train_loss": 0.25795552134513855, "global_step": 46727, "epoch": 525, "lr": 9.298788537762336e-05}
{"train_loss": 0.23045098781585693, "global_step": 46728, "epoch": 525, "lr": 9.298758930200498e-05}
{"train_loss": 0.2589375674724579, "global_step": 46729, "epoch": 525, "lr": 9.298729322060743e-05}
{"train_loss": 0.35024064779281616, "global_step": 46730, "epoch": 525, "lr": 9.298699713343076e-05}
{"train_loss": 0.32092368602752686, "global_step": 46731, "epoch": 525, "lr": 9.298670104047503e-05}
{"train_loss": 0.3409743309020996, "global_step": 46732, "epoch": 525, "lr": 9.298640494174025e-05}
{"train_loss": 0.3232823312282562, "global_step": 46733, "epoch": 525, "lr": 9.298610883722647e-05}
{"train_loss": 0.2575799822807312, "global_step": 46734, "epoch": 525, "lr": 9.298581272693372e-05}
{"train_loss": 0.3364972174167633, "global_step": 46735, "epoch": 525, "lr": 9.298551661086207e-05}
{"train_loss": 0.2377956509590149, "global_step": 46736, "epoch": 525, "lr": 9.298522048901153e-05}
{"train_loss": 0.1685430109500885, "global_step": 46737, "epoch": 525, "lr": 9.298492436138216e-05}
{"train_loss": 0.2813965976238251, "global_step": 46738, "epoch": 525, "lr": 9.298462822797399e-05}
{"train_loss": 0.25912559032440186, "global_step": 46739, "epoch": 525, "lr": 9.298433208878707e-05}
{"train_loss": 0.2407892793416977, "global_step": 46740, "epoch": 525, "lr": 9.298403594382141e-05}
{"train_loss": 0.2862967848777771, "global_step": 46741, "epoch": 525, "lr": 9.298373979307709e-05}
{"train_loss": 0.31742000579833984, "global_step": 46742, "epoch": 525, "lr": 9.298344363655412e-05}
{"train_loss": 0.25594374537467957, "global_step": 46743, "epoch": 525, "lr": 9.298314747425254e-05}
{"train_loss": 0.37046119570732117, "global_step": 46744, "epoch": 525, "lr": 9.298285130617242e-05}
{"train_loss": 0.29059627652168274, "global_step": 46745, "epoch": 525, "lr": 9.298255513231376e-05}
{"train_loss": 0.39356935024261475, "global_step": 46746, "epoch": 525, "lr": 9.298225895267663e-05}
{"train_loss": 0.28420475125312805, "global_step": 46747, "epoch": 525, "lr": 9.298196276726105e-05}
{"train_loss": 0.37023839354515076, "global_step": 46748, "epoch": 525, "lr": 9.298166657606709e-05}
{"train_loss": 0.29444620013237, "global_step": 46749, "epoch": 525, "lr": 9.298137037909474e-05}
{"train_loss": 0.2152358740568161, "global_step": 46750, "epoch": 525, "lr": 9.298107417634409e-05}
{"train_loss": 0.28565919399261475, "global_step": 46751, "epoch": 525, "lr": 9.298077796781515e-05}
{"train_loss": 0.2603716254234314, "global_step": 46752, "epoch": 525, "lr": 9.298048175350797e-05}
{"train_loss": 0.32687297463417053, "global_step": 46753, "epoch": 525, "lr": 9.298018553342258e-05}
{"train_loss": 0.2776370346546173, "global_step": 46754, "epoch": 525, "lr": 9.297988930755905e-05}
{"train_loss": 0.3086603581905365, "global_step": 46755, "epoch": 525, "lr": 9.297959307591738e-05}
{"train_loss": 0.26700708270072937, "global_step": 46756, "epoch": 525, "lr": 9.297929683849762e-05}
{"train_loss": 0.32457655668258667, "global_step": 46757, "epoch": 525, "lr": 9.297900059529984e-05}
{"train_loss": 0.28253957629203796, "global_step": 46758, "epoch": 525, "lr": 9.297870434632404e-05}
{"train_loss": 0.293766051530838, "global_step": 46759, "epoch": 525, "lr": 9.297840809157028e-05}
{"train_loss": 0.34837600588798523, "global_step": 46760, "epoch": 525, "lr": 9.29781118310386e-05}
{"train_loss": 0.2778882086277008, "global_step": 46761, "epoch": 525, "lr": 9.297781556472905e-05}
{"train_loss": 0.22985953092575073, "global_step": 46762, "epoch": 525, "lr": 9.297751929264163e-05}
{"train_loss": 0.2976098358631134, "global_step": 46763, "epoch": 525, "lr": 9.297722301477643e-05}
{"train_loss": 0.24675805866718292, "global_step": 46764, "epoch": 525, "lr": 9.297692673113345e-05}
{"train_loss": 0.23646849393844604, "global_step": 46765, "epoch": 525, "lr": 9.297663044171274e-05}
{"train_loss": 0.27367013692855835, "global_step": 46766, "epoch": 525, "lr": 9.297633414651437e-05}
{"train_loss": 0.20529639720916748, "global_step": 46767, "epoch": 525, "lr": 9.297603784553835e-05}
{"train_loss": 0.3586135804653168, "global_step": 46768, "epoch": 525, "lr": 9.29757415387847e-05}
{"train_loss": 0.2308332473039627, "global_step": 46769, "epoch": 525, "lr": 9.297544522625351e-05}
{"train_loss": 0.23058949410915375, "global_step": 46770, "epoch": 525, "lr": 9.297514890794478e-05}
{"train_loss": 0.2884737253189087, "global_step": 46771, "epoch": 525, "lr": 9.297485258385858e-05}
{"train_loss": 0.21353627741336823, "global_step": 46772, "epoch": 525, "lr": 9.297455625399494e-05}
{"train_loss": 0.30925410985946655, "global_step": 46773, "epoch": 525, "lr": 9.297425991835388e-05}
{"train_loss": 0.3645026683807373, "global_step": 46774, "epoch": 525, "lr": 9.297396357693547e-05}
{"train_loss": 0.3315153419971466, "global_step": 46775, "epoch": 525, "lr": 9.29736672297397e-05}
{"train_loss": 0.3068876564502716, "global_step": 46776, "epoch": 525, "lr": 9.297337087676667e-05}
{"train_loss": 0.3704759478569031, "global_step": 46777, "epoch": 525, "lr": 9.29730745180164e-05}
{"train_loss": 0.3316701352596283, "global_step": 46778, "epoch": 525, "lr": 9.29727781534889e-05}
{"train_loss": 0.2517680823802948, "global_step": 46779, "epoch": 525, "lr": 9.297248178318425e-05}
{"train_loss": 0.3435247242450714, "global_step": 46780, "epoch": 525, "lr": 9.297218540710249e-05}
{"train_loss": 0.31345951557159424, "global_step": 46781, "epoch": 525, "lr": 9.297188902524361e-05}
{"train_loss": 0.2877447307109833, "global_step": 46782, "epoch": 525, "lr": 9.29715926376077e-05}
{"train_loss": 0.303716778755188, "global_step": 46783, "epoch": 525, "lr": 9.297129624419479e-05}
{"train_loss": 0.3131831884384155, "global_step": 46784, "epoch": 525, "lr": 9.29709998450049e-05}
{"train_loss": 0.3413050174713135, "global_step": 46785, "epoch": 525, "lr": 9.297070344003808e-05}
{"train_loss": 0.2986370325088501, "global_step": 46786, "epoch": 525, "lr": 9.29704070292944e-05}
{"train_loss": 0.3830595314502716, "global_step": 46787, "epoch": 525, "lr": 9.297011061277384e-05}
{"train_loss": 0.32354894280433655, "global_step": 46788, "epoch": 525, "lr": 9.296981419047648e-05}
{"train_loss": 0.2450161874294281, "global_step": 46789, "epoch": 525, "lr": 9.296951776240237e-05}
{"train_loss": 0.27189964056015015, "global_step": 46790, "epoch": 525, "lr": 9.296922132855151e-05}
{"train_loss": 0.37815093994140625, "global_step": 46791, "epoch": 525, "lr": 9.296892488892397e-05}
{"train_loss": 0.2593444883823395, "global_step": 46792, "epoch": 525, "lr": 9.296862844351979e-05}
{"train_loss": 0.3427496552467346, "global_step": 46793, "epoch": 525, "lr": 9.2968331992339e-05}
{"train_loss": 0.30151793360710144, "global_step": 46794, "epoch": 525, "lr": 9.296803553538164e-05}
{"train_loss": 0.2685364782810211, "global_step": 46795, "epoch": 525, "lr": 9.296773907264775e-05}
{"train_loss": 0.17403079569339752, "global_step": 46796, "epoch": 525, "lr": 9.296744260413736e-05}
{"train_loss": 0.3326784670352936, "global_step": 46797, "epoch": 525, "lr": 9.296714612985054e-05}
{"train_loss": 0.35141128301620483, "global_step": 46798, "epoch": 525, "lr": 9.296684964978728e-05}
{"train_loss": 0.3292224705219269, "global_step": 46799, "epoch": 525, "lr": 9.296655316394769e-05}
{"train_loss": 0.3643340766429901, "global_step": 46800, "epoch": 525, "lr": 9.296625667233174e-05}
{"train_loss": 0.17465034127235413, "global_step": 46801, "epoch": 525, "lr": 9.296596017493953e-05}
{"train_loss": 0.24261438846588135, "global_step": 46802, "epoch": 525, "lr": 9.296566367177104e-05}
{"train_loss": 0.28239691257476807, "global_step": 46803, "epoch": 525, "lr": 9.296536716282636e-05}
{"train_loss": 0.3697940707206726, "global_step": 46804, "epoch": 525, "lr": 9.29650706481055e-05}
{"train_loss": 0.325375497341156, "global_step": 46805, "epoch": 525, "lr": 9.29647741276085e-05}
{"train_loss": 0.2862136960029602, "global_step": 46806, "epoch": 525, "lr": 9.296447760133543e-05}
{"train_loss": 0.369537889957428, "global_step": 46807, "epoch": 525, "lr": 9.29641810692863e-05}
{"train_loss": 0.31233444809913635, "global_step": 46808, "epoch": 525, "lr": 9.296388453146116e-05}
{"train_loss": 0.3050723373889923, "global_step": 46809, "epoch": 525, "lr": 9.296358798786007e-05}
{"train_loss": 0.3526621162891388, "global_step": 46810, "epoch": 525, "lr": 9.296329143848303e-05}
{"train_loss": 0.29998594522476196, "global_step": 46811, "epoch": 525, "lr": 9.296299488333009e-05}
{"train_loss": 0.27894243597984314, "global_step": 46812, "epoch": 525, "lr": 9.296269832240132e-05}
{"train_loss": 0.2945186241289203, "global_step": 46813, "epoch": 525, "lr": 9.296240175569673e-05, "val_loss": 2.8475184440612793, "train_action_mse_error": 31.056636810302734}
{"train_loss": 0.30721741914749146, "global_step": 46814, "epoch": 526, "lr": 9.296210518321635e-05}
{"train_loss": 0.3432149887084961, "global_step": 46815, "epoch": 526, "lr": 9.296180860496027e-05}
{"train_loss": 0.2768217921257019, "global_step": 46816, "epoch": 526, "lr": 9.296151202092847e-05}
{"train_loss": 0.3022935688495636, "global_step": 46817, "epoch": 526, "lr": 9.296121543112104e-05}
{"train_loss": 0.271633118391037, "global_step": 46818, "epoch": 526, "lr": 9.296091883553799e-05}
{"train_loss": 0.2474302053451538, "global_step": 46819, "epoch": 526, "lr": 9.296062223417936e-05}
{"train_loss": 0.30593734979629517, "global_step": 46820, "epoch": 526, "lr": 9.296032562704522e-05}
{"train_loss": 0.30070778727531433, "global_step": 46821, "epoch": 526, "lr": 9.296002901413557e-05}
{"train_loss": 0.23558080196380615, "global_step": 46822, "epoch": 526, "lr": 9.295973239545045e-05}
{"train_loss": 0.25759240984916687, "global_step": 46823, "epoch": 526, "lr": 9.295943577098996e-05}
{"train_loss": 0.3431774973869324, "global_step": 46824, "epoch": 526, "lr": 9.295913914075407e-05}
{"train_loss": 0.36407744884490967, "global_step": 46825, "epoch": 526, "lr": 9.295884250474285e-05}
{"train_loss": 0.327918142080307, "global_step": 46826, "epoch": 526, "lr": 9.295854586295634e-05}
{"train_loss": 0.2579229474067688, "global_step": 46827, "epoch": 526, "lr": 9.295824921539458e-05}
{"train_loss": 0.23838315904140472, "global_step": 46828, "epoch": 526, "lr": 9.295795256205762e-05}
{"train_loss": 0.3481173515319824, "global_step": 46829, "epoch": 526, "lr": 9.295765590294547e-05}
{"train_loss": 0.25919368863105774, "global_step": 46830, "epoch": 526, "lr": 9.295735923805819e-05}
{"train_loss": 0.14060170948505402, "global_step": 46831, "epoch": 526, "lr": 9.29570625673958e-05}
{"train_loss": 0.17994378507137299, "global_step": 46832, "epoch": 526, "lr": 9.295676589095838e-05}
{"train_loss": 0.3214714229106903, "global_step": 46833, "epoch": 526, "lr": 9.295646920874595e-05}
{"train_loss": 0.33470475673675537, "global_step": 46834, "epoch": 526, "lr": 9.295617252075853e-05}
{"train_loss": 0.269178569316864, "global_step": 46835, "epoch": 526, "lr": 9.295587582699618e-05}
{"train_loss": 0.30583620071411133, "global_step": 46836, "epoch": 526, "lr": 9.295557912745894e-05}
{"train_loss": 0.3251870572566986, "global_step": 46837, "epoch": 526, "lr": 9.295528242214683e-05}
{"train_loss": 0.30579012632369995, "global_step": 46838, "epoch": 526, "lr": 9.295498571105991e-05}
{"train_loss": 0.2798863351345062, "global_step": 46839, "epoch": 526, "lr": 9.295468899419823e-05}
{"train_loss": 0.35679081082344055, "global_step": 46840, "epoch": 526, "lr": 9.295439227156182e-05}
{"train_loss": 0.31002679467201233, "global_step": 46841, "epoch": 526, "lr": 9.295409554315071e-05}
{"train_loss": 0.23592130839824677, "global_step": 46842, "epoch": 526, "lr": 9.295379880896493e-05}
{"train_loss": 0.3392440378665924, "global_step": 46843, "epoch": 526, "lr": 9.295350206900455e-05}
{"train_loss": 0.281825453042984, "global_step": 46844, "epoch": 526, "lr": 9.295320532326959e-05}
{"train_loss": 0.24462692439556122, "global_step": 46845, "epoch": 526, "lr": 9.295290857176009e-05}
{"train_loss": 0.3999015688896179, "global_step": 46846, "epoch": 526, "lr": 9.29526118144761e-05}
{"train_loss": 0.32996654510498047, "global_step": 46847, "epoch": 526, "lr": 9.295231505141765e-05}
{"train_loss": 0.33227410912513733, "global_step": 46848, "epoch": 526, "lr": 9.295201828258481e-05}
{"train_loss": 0.3150727450847626, "global_step": 46849, "epoch": 526, "lr": 9.295172150797758e-05}
{"train_loss": 0.3074665367603302, "global_step": 46850, "epoch": 526, "lr": 9.2951424727596e-05}
{"train_loss": 0.2894199788570404, "global_step": 46851, "epoch": 526, "lr": 9.295112794144014e-05}
{"train_loss": 0.17762881517410278, "global_step": 46852, "epoch": 526, "lr": 9.295083114951003e-05}
{"train_loss": 0.26180359721183777, "global_step": 46853, "epoch": 526, "lr": 9.295053435180568e-05}
{"train_loss": 0.30170124769210815, "global_step": 46854, "epoch": 526, "lr": 9.295023754832717e-05}
{"train_loss": 0.32410240173339844, "global_step": 46855, "epoch": 526, "lr": 9.294994073907452e-05}
{"train_loss": 0.2437743842601776, "global_step": 46856, "epoch": 526, "lr": 9.294964392404777e-05}
{"train_loss": 0.3369421660900116, "global_step": 46857, "epoch": 526, "lr": 9.294934710324698e-05}
{"train_loss": 0.28868669271469116, "global_step": 46858, "epoch": 526, "lr": 9.294905027667216e-05}
{"train_loss": 0.30160075426101685, "global_step": 46859, "epoch": 526, "lr": 9.294875344432336e-05}
{"train_loss": 0.4241544306278229, "global_step": 46860, "epoch": 526, "lr": 9.294845660620064e-05}
{"train_loss": 0.321031779050827, "global_step": 46861, "epoch": 526, "lr": 9.2948159762304e-05}
{"train_loss": 0.36594653129577637, "global_step": 46862, "epoch": 526, "lr": 9.294786291263354e-05}
{"train_loss": 0.2741350531578064, "global_step": 46863, "epoch": 526, "lr": 9.294756605718922e-05}
{"train_loss": 0.29745402932167053, "global_step": 46864, "epoch": 526, "lr": 9.294726919597114e-05}
{"train_loss": 0.26913732290267944, "global_step": 46865, "epoch": 526, "lr": 9.294697232897934e-05}
{"train_loss": 0.29183638095855713, "global_step": 46866, "epoch": 526, "lr": 9.294667545621383e-05}
{"train_loss": 0.3448300063610077, "global_step": 46867, "epoch": 526, "lr": 9.294637857767465e-05}
{"train_loss": 0.21146872639656067, "global_step": 46868, "epoch": 526, "lr": 9.294608169336187e-05}
{"train_loss": 0.2312903255224228, "global_step": 46869, "epoch": 526, "lr": 9.294578480327552e-05}
{"train_loss": 0.3090476095676422, "global_step": 46870, "epoch": 526, "lr": 9.294548790741561e-05}
{"train_loss": 0.2684142589569092, "global_step": 46871, "epoch": 526, "lr": 9.294519100578222e-05}
{"train_loss": 0.1918366700410843, "global_step": 46872, "epoch": 526, "lr": 9.294489409837537e-05}
{"train_loss": 0.21055972576141357, "global_step": 46873, "epoch": 526, "lr": 9.294459718519509e-05}
{"train_loss": 0.29373621940612793, "global_step": 46874, "epoch": 526, "lr": 9.294430026624143e-05}
{"train_loss": 0.32189422845840454, "global_step": 46875, "epoch": 526, "lr": 9.294400334151445e-05}
{"train_loss": 0.38485288619995117, "global_step": 46876, "epoch": 526, "lr": 9.294370641101418e-05}
{"train_loss": 0.3330596685409546, "global_step": 46877, "epoch": 526, "lr": 9.294340947474062e-05}
{"train_loss": 0.27718088030815125, "global_step": 46878, "epoch": 526, "lr": 9.294311253269386e-05}
{"train_loss": 0.2828824818134308, "global_step": 46879, "epoch": 526, "lr": 9.294281558487392e-05}
{"train_loss": 0.26780015230178833, "global_step": 46880, "epoch": 526, "lr": 9.294251863128084e-05}
{"train_loss": 0.3029141426086426, "global_step": 46881, "epoch": 526, "lr": 9.294222167191468e-05}
{"train_loss": 0.24265673756599426, "global_step": 46882, "epoch": 526, "lr": 9.294192470677544e-05}
{"train_loss": 0.3770558834075928, "global_step": 46883, "epoch": 526, "lr": 9.294162773586319e-05}
{"train_loss": 0.2633693814277649, "global_step": 46884, "epoch": 526, "lr": 9.294133075917797e-05}
{"train_loss": 0.25385481119155884, "global_step": 46885, "epoch": 526, "lr": 9.294103377671979e-05}
{"train_loss": 0.25817668437957764, "global_step": 46886, "epoch": 526, "lr": 9.294073678848873e-05}
{"train_loss": 0.37633511424064636, "global_step": 46887, "epoch": 526, "lr": 9.294043979448481e-05}
{"train_loss": 0.2880938649177551, "global_step": 46888, "epoch": 526, "lr": 9.294014279470805e-05}
{"train_loss": 0.24693654477596283, "global_step": 46889, "epoch": 526, "lr": 9.293984578915853e-05}
{"train_loss": 0.2932751178741455, "global_step": 46890, "epoch": 526, "lr": 9.293954877783626e-05}
{"train_loss": 0.2596791684627533, "global_step": 46891, "epoch": 526, "lr": 9.293925176074131e-05}
{"train_loss": 0.25757652521133423, "global_step": 46892, "epoch": 526, "lr": 9.293895473787369e-05}
{"train_loss": 0.21038562059402466, "global_step": 46893, "epoch": 526, "lr": 9.293865770923344e-05}
{"train_loss": 0.4120315611362457, "global_step": 46894, "epoch": 526, "lr": 9.293836067482063e-05}
{"train_loss": 0.33768776059150696, "global_step": 46895, "epoch": 526, "lr": 9.293806363463526e-05}
{"train_loss": 0.25492602586746216, "global_step": 46896, "epoch": 526, "lr": 9.29377665886774e-05}
{"train_loss": 0.24505507946014404, "global_step": 46897, "epoch": 526, "lr": 9.293746953694709e-05}
{"train_loss": 0.36866021156311035, "global_step": 46898, "epoch": 526, "lr": 9.293717247944435e-05}
{"train_loss": 0.28907814621925354, "global_step": 46899, "epoch": 526, "lr": 9.293687541616925e-05}
{"train_loss": 0.2810405194759369, "global_step": 46900, "epoch": 526, "lr": 9.29365783471218e-05}
{"train_loss": 0.33698198199272156, "global_step": 46901, "epoch": 526, "lr": 9.293628127230203e-05}
{"train_loss": 0.29228185837188464, "global_step": 46902, "epoch": 526, "lr": 9.293598419171001e-05, "val_loss": 2.9287350177764893}
{"train_loss": 0.3000718057155609, "global_step": 46903, "epoch": 527, "lr": 9.293568710534578e-05}
{"train_loss": 0.29825812578201294, "global_step": 46904, "epoch": 527, "lr": 9.293539001320937e-05}
{"train_loss": 0.22006189823150635, "global_step": 46905, "epoch": 527, "lr": 9.293509291530081e-05}
{"train_loss": 0.3161068856716156, "global_step": 46906, "epoch": 527, "lr": 9.293479581162016e-05}
{"train_loss": 0.24679933488368988, "global_step": 46907, "epoch": 527, "lr": 9.293449870216744e-05}
{"train_loss": 0.24694682657718658, "global_step": 46908, "epoch": 527, "lr": 9.293420158694271e-05}
{"train_loss": 0.19202467799186707, "global_step": 46909, "epoch": 527, "lr": 9.2933904465946e-05}
{"train_loss": 0.26268279552459717, "global_step": 46910, "epoch": 527, "lr": 9.293360733917735e-05}
{"train_loss": 0.20690999925136566, "global_step": 46911, "epoch": 527, "lr": 9.293331020663679e-05}
{"train_loss": 0.2661561965942383, "global_step": 46912, "epoch": 527, "lr": 9.293301306832438e-05}
{"train_loss": 0.23262542486190796, "global_step": 46913, "epoch": 527, "lr": 9.293271592424014e-05}
{"train_loss": 0.3057200312614441, "global_step": 46914, "epoch": 527, "lr": 9.293241877438412e-05}
{"train_loss": 0.24156272411346436, "global_step": 46915, "epoch": 527, "lr": 9.293212161875638e-05}
{"train_loss": 0.286603718996048, "global_step": 46916, "epoch": 527, "lr": 9.293182445735691e-05}
{"train_loss": 0.2828274667263031, "global_step": 46917, "epoch": 527, "lr": 9.29315272901858e-05}
{"train_loss": 0.368551105260849, "global_step": 46918, "epoch": 527, "lr": 9.293123011724305e-05}
{"train_loss": 0.25886407494544983, "global_step": 46919, "epoch": 527, "lr": 9.293093293852874e-05}
{"train_loss": 0.20620162785053253, "global_step": 46920, "epoch": 527, "lr": 9.293063575404288e-05}
{"train_loss": 0.22514483332633972, "global_step": 46921, "epoch": 527, "lr": 9.293033856378552e-05}
{"train_loss": 0.21163718402385712, "global_step": 46922, "epoch": 527, "lr": 9.293004136775669e-05}
{"train_loss": 0.24135777354240417, "global_step": 46923, "epoch": 527, "lr": 9.292974416595645e-05}
{"train_loss": 0.27055624127388, "global_step": 46924, "epoch": 527, "lr": 9.292944695838483e-05}
{"train_loss": 0.26077163219451904, "global_step": 46925, "epoch": 527, "lr": 9.292914974504188e-05}
{"train_loss": 0.27699241042137146, "global_step": 46926, "epoch": 527, "lr": 9.29288525259276e-05}
{"train_loss": 0.31920069456100464, "global_step": 46927, "epoch": 527, "lr": 9.292855530104209e-05}
{"train_loss": 0.23070289194583893, "global_step": 46928, "epoch": 527, "lr": 9.292825807038535e-05}
{"train_loss": 0.2498679757118225, "global_step": 46929, "epoch": 527, "lr": 9.29279608339574e-05}
{"train_loss": 0.2908249497413635, "global_step": 46930, "epoch": 527, "lr": 9.292766359175834e-05}
{"train_loss": 0.32497501373291016, "global_step": 46931, "epoch": 527, "lr": 9.292736634378818e-05}
{"train_loss": 0.3163539469242096, "global_step": 46932, "epoch": 527, "lr": 9.292706909004694e-05}
{"train_loss": 0.28868260979652405, "global_step": 46933, "epoch": 527, "lr": 9.292677183053469e-05}
{"train_loss": 0.35768023133277893, "global_step": 46934, "epoch": 527, "lr": 9.292647456525146e-05}
{"train_loss": 0.24798601865768433, "global_step": 46935, "epoch": 527, "lr": 9.292617729419728e-05}
{"train_loss": 0.23404310643672943, "global_step": 46936, "epoch": 527, "lr": 9.292588001737221e-05}
{"train_loss": 0.20629651844501495, "global_step": 46937, "epoch": 527, "lr": 9.292558273477627e-05}
{"train_loss": 0.39919403195381165, "global_step": 46938, "epoch": 527, "lr": 9.29252854464095e-05}
{"train_loss": 0.2721264958381653, "global_step": 46939, "epoch": 527, "lr": 9.292498815227197e-05}
{"train_loss": 0.3389164209365845, "global_step": 46940, "epoch": 527, "lr": 9.292469085236369e-05}
{"train_loss": 0.29659342765808105, "global_step": 46941, "epoch": 527, "lr": 9.292439354668471e-05}
{"train_loss": 0.24048744142055511, "global_step": 46942, "epoch": 527, "lr": 9.292409623523505e-05}
{"train_loss": 0.27005764842033386, "global_step": 46943, "epoch": 527, "lr": 9.29237989180148e-05}
{"train_loss": 0.24341407418251038, "global_step": 46944, "epoch": 527, "lr": 9.292350159502394e-05}
{"train_loss": 0.24449117481708527, "global_step": 46945, "epoch": 527, "lr": 9.292320426626256e-05}
{"train_loss": 0.22514808177947998, "global_step": 46946, "epoch": 527, "lr": 9.292290693173067e-05}
{"train_loss": 0.29487231373786926, "global_step": 46947, "epoch": 527, "lr": 9.292260959142831e-05}
{"train_loss": 0.23802322149276733, "global_step": 46948, "epoch": 527, "lr": 9.292231224535555e-05}
{"train_loss": 0.2262878566980362, "global_step": 46949, "epoch": 527, "lr": 9.292201489351238e-05}
{"train_loss": 0.2177286148071289, "global_step": 46950, "epoch": 527, "lr": 9.292171753589889e-05}
{"train_loss": 0.397909015417099, "global_step": 46951, "epoch": 527, "lr": 9.292142017251508e-05}
{"train_loss": 0.3807459771633148, "global_step": 46952, "epoch": 527, "lr": 9.292112280336103e-05}
{"train_loss": 0.25497978925704956, "global_step": 46953, "epoch": 527, "lr": 9.292082542843674e-05}
{"train_loss": 0.271377831697464, "global_step": 46954, "epoch": 527, "lr": 9.292052804774228e-05}
{"train_loss": 0.33045023679733276, "global_step": 46955, "epoch": 527, "lr": 9.292023066127767e-05}
{"train_loss": 0.26323410868644714, "global_step": 46956, "epoch": 527, "lr": 9.291993326904296e-05}
{"train_loss": 0.22101224958896637, "global_step": 46957, "epoch": 527, "lr": 9.29196358710382e-05}
{"train_loss": 0.20046783983707428, "global_step": 46958, "epoch": 527, "lr": 9.29193384672634e-05}
{"train_loss": 0.2712562680244446, "global_step": 46959, "epoch": 527, "lr": 9.291904105771862e-05}
{"train_loss": 0.22209453582763672, "global_step": 46960, "epoch": 527, "lr": 9.29187436424039e-05}
{"train_loss": 0.2761848568916321, "global_step": 46961, "epoch": 527, "lr": 9.291844622131929e-05}
{"train_loss": 0.2779327929019928, "global_step": 46962, "epoch": 527, "lr": 9.291814879446481e-05}
{"train_loss": 0.34769320487976074, "global_step": 46963, "epoch": 527, "lr": 9.29178513618405e-05}
{"train_loss": 0.30851855874061584, "global_step": 46964, "epoch": 527, "lr": 9.291755392344644e-05}
{"train_loss": 0.2925046682357788, "global_step": 46965, "epoch": 527, "lr": 9.29172564792826e-05}
{"train_loss": 0.32017388939857483, "global_step": 46966, "epoch": 527, "lr": 9.291695902934907e-05}
{"train_loss": 0.31505632400512695, "global_step": 46967, "epoch": 527, "lr": 9.291666157364589e-05}
{"train_loss": 0.31007006764411926, "global_step": 46968, "epoch": 527, "lr": 9.291636411217306e-05}
{"train_loss": 0.2650749981403351, "global_step": 46969, "epoch": 527, "lr": 9.291606664493067e-05}
{"train_loss": 0.3545747995376587, "global_step": 46970, "epoch": 527, "lr": 9.291576917191873e-05}
{"train_loss": 0.2732061445713043, "global_step": 46971, "epoch": 527, "lr": 9.29154716931373e-05}
{"train_loss": 0.3831082582473755, "global_step": 46972, "epoch": 527, "lr": 9.29151742085864e-05}
{"train_loss": 0.20660991966724396, "global_step": 46973, "epoch": 527, "lr": 9.291487671826606e-05}
{"train_loss": 0.3362825810909271, "global_step": 46974, "epoch": 527, "lr": 9.291457922217636e-05}
{"train_loss": 0.28754565119743347, "global_step": 46975, "epoch": 527, "lr": 9.291428172031732e-05}
{"train_loss": 0.31705448031425476, "global_step": 46976, "epoch": 527, "lr": 9.291398421268898e-05}
{"train_loss": 0.3100183606147766, "global_step": 46977, "epoch": 527, "lr": 9.291368669929135e-05}
{"train_loss": 0.2964233160018921, "global_step": 46978, "epoch": 527, "lr": 9.291338918012452e-05}
{"train_loss": 0.2464636117219925, "global_step": 46979, "epoch": 527, "lr": 9.291309165518851e-05}
{"train_loss": 0.30630776286125183, "global_step": 46980, "epoch": 527, "lr": 9.291279412448334e-05}
{"train_loss": 0.32091084122657776, "global_step": 46981, "epoch": 527, "lr": 9.291249658800908e-05}
{"train_loss": 0.2790515720844269, "global_step": 46982, "epoch": 527, "lr": 9.291219904576577e-05}
{"train_loss": 0.33037546277046204, "global_step": 46983, "epoch": 527, "lr": 9.291190149775342e-05}
{"train_loss": 0.27982473373413086, "global_step": 46984, "epoch": 527, "lr": 9.29116039439721e-05}
{"train_loss": 0.1756497323513031, "global_step": 46985, "epoch": 527, "lr": 9.291130638442183e-05}
{"train_loss": 0.2694535255432129, "global_step": 46986, "epoch": 527, "lr": 9.291100881910267e-05}
{"train_loss": 0.38533177971839905, "global_step": 46987, "epoch": 527, "lr": 9.291071124801463e-05}
{"train_loss": 0.19502487778663635, "global_step": 46988, "epoch": 527, "lr": 9.291041367115776e-05}
{"train_loss": 0.24193008244037628, "global_step": 46989, "epoch": 527, "lr": 9.291011608853213e-05}
{"train_loss": 0.3202771842479706, "global_step": 46990, "epoch": 527, "lr": 9.290981850013774e-05}
{"train_loss": 0.27801584929562684, "global_step": 46991, "epoch": 527, "lr": 9.290952090597467e-05, "val_loss": 2.939399242401123}
{"train_loss": 0.3153674602508545, "global_step": 46992, "epoch": 528, "lr": 9.290922330604293e-05}
{"train_loss": 0.218031108379364, "global_step": 46993, "epoch": 528, "lr": 9.290892570034256e-05}
{"train_loss": 0.34253618121147156, "global_step": 46994, "epoch": 528, "lr": 9.290862808887362e-05}
{"train_loss": 0.24235743284225464, "global_step": 46995, "epoch": 528, "lr": 9.290833047163611e-05}
{"train_loss": 0.1735127866268158, "global_step": 46996, "epoch": 528, "lr": 9.290803284863012e-05}
{"train_loss": 0.22549033164978027, "global_step": 46997, "epoch": 528, "lr": 9.290773521985567e-05}
{"train_loss": 0.3641209006309509, "global_step": 46998, "epoch": 528, "lr": 9.290743758531279e-05}
{"train_loss": 0.2505562901496887, "global_step": 46999, "epoch": 528, "lr": 9.290713994500155e-05}
{"train_loss": 0.29578453302383423, "global_step": 47000, "epoch": 528, "lr": 9.290684229892194e-05}
{"train_loss": 0.23910541832447052, "global_step": 47001, "epoch": 528, "lr": 9.290654464707404e-05}
{"train_loss": 0.19000943005084991, "global_step": 47002, "epoch": 528, "lr": 9.290624698945788e-05}
{"train_loss": 0.26885366439819336, "global_step": 47003, "epoch": 528, "lr": 9.29059493260735e-05}
{"train_loss": 0.24388815462589264, "global_step": 47004, "epoch": 528, "lr": 9.290565165692094e-05}
{"train_loss": 0.3776032328605652, "global_step": 47005, "epoch": 528, "lr": 9.290535398200023e-05}
{"train_loss": 0.22626544535160065, "global_step": 47006, "epoch": 528, "lr": 9.29050563013114e-05}
{"train_loss": 0.26472216844558716, "global_step": 47007, "epoch": 528, "lr": 9.290475861485453e-05}
{"train_loss": 0.2834685146808624, "global_step": 47008, "epoch": 528, "lr": 9.290446092262963e-05}
{"train_loss": 0.18980015814304352, "global_step": 47009, "epoch": 528, "lr": 9.290416322463676e-05}
{"train_loss": 0.1813422292470932, "global_step": 47010, "epoch": 528, "lr": 9.290386552087593e-05}
{"train_loss": 0.271987646818161, "global_step": 47011, "epoch": 528, "lr": 9.290356781134722e-05}
{"train_loss": 0.3597240746021271, "global_step": 47012, "epoch": 528, "lr": 9.290327009605064e-05}
{"train_loss": 0.29935961961746216, "global_step": 47013, "epoch": 528, "lr": 9.290297237498623e-05}
{"train_loss": 0.2582133114337921, "global_step": 47014, "epoch": 528, "lr": 9.290267464815405e-05}
{"train_loss": 0.2728642523288727, "global_step": 47015, "epoch": 528, "lr": 9.290237691555412e-05}
{"train_loss": 0.24766838550567627, "global_step": 47016, "epoch": 528, "lr": 9.290207917718648e-05}
{"train_loss": 0.35090792179107666, "global_step": 47017, "epoch": 528, "lr": 9.29017814330512e-05}
{"train_loss": 0.3320325016975403, "global_step": 47018, "epoch": 528, "lr": 9.290148368314829e-05}
{"train_loss": 0.22493310272693634, "global_step": 47019, "epoch": 528, "lr": 9.290118592747778e-05}
{"train_loss": 0.4282919764518738, "global_step": 47020, "epoch": 528, "lr": 9.290088816603975e-05}
{"train_loss": 0.26477953791618347, "global_step": 47021, "epoch": 528, "lr": 9.290059039883421e-05}
{"train_loss": 0.26729756593704224, "global_step": 47022, "epoch": 528, "lr": 9.290029262586121e-05}
{"train_loss": 0.2915457785129547, "global_step": 47023, "epoch": 528, "lr": 9.28999948471208e-05}
{"train_loss": 0.29033616185188293, "global_step": 47024, "epoch": 528, "lr": 9.289969706261299e-05}
{"train_loss": 0.4040742814540863, "global_step": 47025, "epoch": 528, "lr": 9.289939927233785e-05}
{"train_loss": 0.3833553194999695, "global_step": 47026, "epoch": 528, "lr": 9.289910147629542e-05}
{"train_loss": 0.3743674159049988, "global_step": 47027, "epoch": 528, "lr": 9.28988036744857e-05}
{"train_loss": 0.23473255336284637, "global_step": 47028, "epoch": 528, "lr": 9.289850586690878e-05}
{"train_loss": 0.3812862038612366, "global_step": 47029, "epoch": 528, "lr": 9.289820805356466e-05}
{"train_loss": 0.31453976035118103, "global_step": 47030, "epoch": 528, "lr": 9.289791023445342e-05}
{"train_loss": 0.3588656485080719, "global_step": 47031, "epoch": 528, "lr": 9.289761240957507e-05}
{"train_loss": 0.4903874695301056, "global_step": 47032, "epoch": 528, "lr": 9.289731457892965e-05}
{"train_loss": 0.29565492272377014, "global_step": 47033, "epoch": 528, "lr": 9.289701674251723e-05}
{"train_loss": 0.22022180259227753, "global_step": 47034, "epoch": 528, "lr": 9.28967189003378e-05}
{"train_loss": 0.32905691862106323, "global_step": 47035, "epoch": 528, "lr": 9.289642105239145e-05}
{"train_loss": 0.3683249354362488, "global_step": 47036, "epoch": 528, "lr": 9.28961231986782e-05}
{"train_loss": 0.44764474034309387, "global_step": 47037, "epoch": 528, "lr": 9.289582533919808e-05}
{"train_loss": 0.40211009979248047, "global_step": 47038, "epoch": 528, "lr": 9.289552747395114e-05}
{"train_loss": 0.2594911754131317, "global_step": 47039, "epoch": 528, "lr": 9.289522960293742e-05}
{"train_loss": 0.2926930785179138, "global_step": 47040, "epoch": 528, "lr": 9.289493172615697e-05}
{"train_loss": 0.30817797780036926, "global_step": 47041, "epoch": 528, "lr": 9.28946338436098e-05}
{"train_loss": 0.3129078149795532, "global_step": 47042, "epoch": 528, "lr": 9.289433595529598e-05}
{"train_loss": 0.26322680711746216, "global_step": 47043, "epoch": 528, "lr": 9.289403806121555e-05}
{"train_loss": 0.28997018933296204, "global_step": 47044, "epoch": 528, "lr": 9.289374016136851e-05}
{"train_loss": 0.2732091546058655, "global_step": 47045, "epoch": 528, "lr": 9.289344225575496e-05}
{"train_loss": 0.3980844020843506, "global_step": 47046, "epoch": 528, "lr": 9.289314434437489e-05}
{"train_loss": 0.29610922932624817, "global_step": 47047, "epoch": 528, "lr": 9.289284642722837e-05}
{"train_loss": 0.26803866028785706, "global_step": 47048, "epoch": 528, "lr": 9.289254850431542e-05}
{"train_loss": 0.4142758548259735, "global_step": 47049, "epoch": 528, "lr": 9.28922505756361e-05}
{"train_loss": 0.2861998379230499, "global_step": 47050, "epoch": 528, "lr": 9.289195264119044e-05}
{"train_loss": 0.3303326964378357, "global_step": 47051, "epoch": 528, "lr": 9.289165470097846e-05}
{"train_loss": 0.365363210439682, "global_step": 47052, "epoch": 528, "lr": 9.289135675500025e-05}
{"train_loss": 0.28309008479118347, "global_step": 47053, "epoch": 528, "lr": 9.289105880325579e-05}
{"train_loss": 0.3287245035171509, "global_step": 47054, "epoch": 528, "lr": 9.289076084574517e-05}
{"train_loss": 0.2695797085762024, "global_step": 47055, "epoch": 528, "lr": 9.289046288246841e-05}
{"train_loss": 0.31952348351478577, "global_step": 47056, "epoch": 528, "lr": 9.289016491342555e-05}
{"train_loss": 0.30868396162986755, "global_step": 47057, "epoch": 528, "lr": 9.288986693861662e-05}
{"train_loss": 0.3105770945549011, "global_step": 47058, "epoch": 528, "lr": 9.288956895804168e-05}
{"train_loss": 0.3274235725402832, "global_step": 47059, "epoch": 528, "lr": 9.288927097170075e-05}
{"train_loss": 0.2967571020126343, "global_step": 47060, "epoch": 528, "lr": 9.288897297959389e-05}
{"train_loss": 0.29683735966682434, "global_step": 47061, "epoch": 528, "lr": 9.288867498172112e-05}
{"train_loss": 0.28679943084716797, "global_step": 47062, "epoch": 528, "lr": 9.288837697808251e-05}
{"train_loss": 0.305866539478302, "global_step": 47063, "epoch": 528, "lr": 9.288807896867805e-05}
{"train_loss": 0.36479634046554565, "global_step": 47064, "epoch": 528, "lr": 9.288778095350783e-05}
{"train_loss": 0.2318427860736847, "global_step": 47065, "epoch": 528, "lr": 9.288748293257187e-05}
{"train_loss": 0.2022591531276703, "global_step": 47066, "epoch": 528, "lr": 9.28871849058702e-05}
{"train_loss": 0.32975345849990845, "global_step": 47067, "epoch": 528, "lr": 9.288688687340287e-05}
{"train_loss": 0.3462895154953003, "global_step": 47068, "epoch": 528, "lr": 9.288658883516995e-05}
{"train_loss": 0.23695454001426697, "global_step": 47069, "epoch": 528, "lr": 9.288629079117142e-05}
{"train_loss": 0.23717013001441956, "global_step": 47070, "epoch": 528, "lr": 9.288599274140737e-05}
{"train_loss": 0.22439025342464447, "global_step": 47071, "epoch": 528, "lr": 9.28856946858778e-05}
{"train_loss": 0.3116944134235382, "global_step": 47072, "epoch": 528, "lr": 9.288539662458279e-05}
{"train_loss": 0.19808965921401978, "global_step": 47073, "epoch": 528, "lr": 9.288509855752235e-05}
{"train_loss": 0.27862006425857544, "global_step": 47074, "epoch": 528, "lr": 9.288480048469654e-05}
{"train_loss": 0.34291794896125793, "global_step": 47075, "epoch": 528, "lr": 9.288450240610539e-05}
{"train_loss": 0.29165804386138916, "global_step": 47076, "epoch": 528, "lr": 9.288420432174893e-05}
{"train_loss": 0.3358961045742035, "global_step": 47077, "epoch": 528, "lr": 9.288390623162723e-05}
{"train_loss": 0.2122194766998291, "global_step": 47078, "epoch": 528, "lr": 9.288360813574029e-05}
{"train_loss": 0.2396610975265503, "global_step": 47079, "epoch": 528, "lr": 9.288331003408819e-05}
{"train_loss": 0.2964865454796995, "global_step": 47080, "epoch": 528, "lr": 9.288301192667095e-05, "val_loss": 2.909986734390259}
{"train_loss": 0.2553369402885437, "global_step": 47081, "epoch": 529, "lr": 9.288271381348859e-05}
{"train_loss": 0.2363569587469101, "global_step": 47082, "epoch": 529, "lr": 9.288241569454119e-05}
{"train_loss": 0.33742043375968933, "global_step": 47083, "epoch": 529, "lr": 9.288211756982879e-05}
{"train_loss": 0.2331220507621765, "global_step": 47084, "epoch": 529, "lr": 9.288181943935138e-05}
{"train_loss": 0.2747606039047241, "global_step": 47085, "epoch": 529, "lr": 9.288152130310905e-05}
{"train_loss": 0.34567445516586304, "global_step": 47086, "epoch": 529, "lr": 9.288122316110181e-05}
{"train_loss": 0.2569982409477234, "global_step": 47087, "epoch": 529, "lr": 9.288092501332972e-05}
{"train_loss": 0.23013466596603394, "global_step": 47088, "epoch": 529, "lr": 9.288062685979281e-05}
{"train_loss": 0.3164720833301544, "global_step": 47089, "epoch": 529, "lr": 9.288032870049112e-05}
{"train_loss": 0.20444561541080475, "global_step": 47090, "epoch": 529, "lr": 9.288003053542468e-05}
{"train_loss": 0.2802346348762512, "global_step": 47091, "epoch": 529, "lr": 9.287973236459357e-05}
{"train_loss": 0.226487398147583, "global_step": 47092, "epoch": 529, "lr": 9.287943418799779e-05}
{"train_loss": 0.23117069900035858, "global_step": 47093, "epoch": 529, "lr": 9.28791360056374e-05}
{"train_loss": 0.2577349841594696, "global_step": 47094, "epoch": 529, "lr": 9.287883781751242e-05}
{"train_loss": 0.2569188177585602, "global_step": 47095, "epoch": 529, "lr": 9.287853962362291e-05}
{"train_loss": 0.2437601387500763, "global_step": 47096, "epoch": 529, "lr": 9.28782414239689e-05}
{"train_loss": 0.26110121607780457, "global_step": 47097, "epoch": 529, "lr": 9.287794321855044e-05}
{"train_loss": 0.18549497425556183, "global_step": 47098, "epoch": 529, "lr": 9.287764500736756e-05}
{"train_loss": 0.33299100399017334, "global_step": 47099, "epoch": 529, "lr": 9.28773467904203e-05}
{"train_loss": 0.3692033290863037, "global_step": 47100, "epoch": 529, "lr": 9.28770485677087e-05}
{"train_loss": 0.2116725593805313, "global_step": 47101, "epoch": 529, "lr": 9.287675033923281e-05}
{"train_loss": 0.24995587766170502, "global_step": 47102, "epoch": 529, "lr": 9.287645210499267e-05}
{"train_loss": 0.35887569189071655, "global_step": 47103, "epoch": 529, "lr": 9.287615386498829e-05}
{"train_loss": 0.29238349199295044, "global_step": 47104, "epoch": 529, "lr": 9.287585561921976e-05}
{"train_loss": 0.23169662058353424, "global_step": 47105, "epoch": 529, "lr": 9.287555736768707e-05}
{"train_loss": 0.27874505519866943, "global_step": 47106, "epoch": 529, "lr": 9.28752591103903e-05}
{"train_loss": 0.30801454186439514, "global_step": 47107, "epoch": 529, "lr": 9.287496084732947e-05}
{"train_loss": 0.2457938939332962, "global_step": 47108, "epoch": 529, "lr": 9.287466257850464e-05}
{"train_loss": 0.311570942401886, "global_step": 47109, "epoch": 529, "lr": 9.28743643039158e-05}
{"train_loss": 0.2573937177658081, "global_step": 47110, "epoch": 529, "lr": 9.287406602356306e-05}
{"train_loss": 0.19833652675151825, "global_step": 47111, "epoch": 529, "lr": 9.28737677374464e-05}
{"train_loss": 0.2065446674823761, "global_step": 47112, "epoch": 529, "lr": 9.287346944556588e-05}
{"train_loss": 0.24357150495052338, "global_step": 47113, "epoch": 529, "lr": 9.287317114792156e-05}
{"train_loss": 0.25344184041023254, "global_step": 47114, "epoch": 529, "lr": 9.287287284451346e-05}
{"train_loss": 0.27961868047714233, "global_step": 47115, "epoch": 529, "lr": 9.287257453534164e-05}
{"train_loss": 0.3526347577571869, "global_step": 47116, "epoch": 529, "lr": 9.28722762204061e-05}
{"train_loss": 0.2851192057132721, "global_step": 47117, "epoch": 529, "lr": 9.287197789970692e-05}
{"train_loss": 0.19698664546012878, "global_step": 47118, "epoch": 529, "lr": 9.287167957324411e-05}
{"train_loss": 0.3342035412788391, "global_step": 47119, "epoch": 529, "lr": 9.287138124101776e-05}
{"train_loss": 0.2882702052593231, "global_step": 47120, "epoch": 529, "lr": 9.287108290302784e-05}
{"train_loss": 0.29231762886047363, "global_step": 47121, "epoch": 529, "lr": 9.287078455927442e-05}
{"train_loss": 0.3036314845085144, "global_step": 47122, "epoch": 529, "lr": 9.287048620975758e-05}
{"train_loss": 0.2236844152212143, "global_step": 47123, "epoch": 529, "lr": 9.287018785447729e-05}
{"train_loss": 0.24429330229759216, "global_step": 47124, "epoch": 529, "lr": 9.286988949343366e-05}
{"train_loss": 0.3528570234775543, "global_step": 47125, "epoch": 529, "lr": 9.286959112662667e-05}
{"train_loss": 0.3360421657562256, "global_step": 47126, "epoch": 529, "lr": 9.28692927540564e-05}
{"train_loss": 0.18668945133686066, "global_step": 47127, "epoch": 529, "lr": 9.286899437572287e-05}
{"train_loss": 0.22792114317417145, "global_step": 47128, "epoch": 529, "lr": 9.286869599162613e-05}
{"train_loss": 0.17581430077552795, "global_step": 47129, "epoch": 529, "lr": 9.286839760176623e-05}
{"train_loss": 0.31085965037345886, "global_step": 47130, "epoch": 529, "lr": 9.286809920614317e-05}
{"train_loss": 0.27687743306159973, "global_step": 47131, "epoch": 529, "lr": 9.286780080475704e-05}
{"train_loss": 0.32962915301322937, "global_step": 47132, "epoch": 529, "lr": 9.286750239760781e-05}
{"train_loss": 0.2622937560081482, "global_step": 47133, "epoch": 529, "lr": 9.286720398469562e-05}
{"train_loss": 0.34124431014060974, "global_step": 47134, "epoch": 529, "lr": 9.286690556602043e-05}
{"train_loss": 0.26764553785324097, "global_step": 47135, "epoch": 529, "lr": 9.286660714158233e-05}
{"train_loss": 0.2681313157081604, "global_step": 47136, "epoch": 529, "lr": 9.286630871138132e-05}
{"train_loss": 0.3114470839500427, "global_step": 47137, "epoch": 529, "lr": 9.286601027541745e-05}
{"train_loss": 0.2718903720378876, "global_step": 47138, "epoch": 529, "lr": 9.286571183369077e-05}
{"train_loss": 0.2633296251296997, "global_step": 47139, "epoch": 529, "lr": 9.286541338620134e-05}
{"train_loss": 0.2472383975982666, "global_step": 47140, "epoch": 529, "lr": 9.286511493294916e-05}
{"train_loss": 0.3039187788963318, "global_step": 47141, "epoch": 529, "lr": 9.286481647393428e-05}
{"train_loss": 0.1994166076183319, "global_step": 47142, "epoch": 529, "lr": 9.286451800915677e-05}
{"train_loss": 0.27618345618247986, "global_step": 47143, "epoch": 529, "lr": 9.286421953861661e-05}
{"train_loss": 0.21777726709842682, "global_step": 47144, "epoch": 529, "lr": 9.286392106231392e-05}
{"train_loss": 0.28031614422798157, "global_step": 47145, "epoch": 529, "lr": 9.286362258024868e-05}
{"train_loss": 0.26674994826316833, "global_step": 47146, "epoch": 529, "lr": 9.286332409242096e-05}
{"train_loss": 0.231736421585083, "global_step": 47147, "epoch": 529, "lr": 9.286302559883076e-05}
{"train_loss": 0.18841034173965454, "global_step": 47148, "epoch": 529, "lr": 9.286272709947818e-05}
{"train_loss": 0.2813901901245117, "global_step": 47149, "epoch": 529, "lr": 9.28624285943632e-05}
{"train_loss": 0.23251473903656006, "global_step": 47150, "epoch": 529, "lr": 9.286213008348591e-05}
{"train_loss": 0.287318617105484, "global_step": 47151, "epoch": 529, "lr": 9.286183156684632e-05}
{"train_loss": 0.2478051483631134, "global_step": 47152, "epoch": 529, "lr": 9.28615330444445e-05}
{"train_loss": 0.2540728449821472, "global_step": 47153, "epoch": 529, "lr": 9.286123451628044e-05}
{"train_loss": 0.30038145184516907, "global_step": 47154, "epoch": 529, "lr": 9.286093598235422e-05}
{"train_loss": 0.2131650745868683, "global_step": 47155, "epoch": 529, "lr": 9.286063744266587e-05}
{"train_loss": 0.2670210301876068, "global_step": 47156, "epoch": 529, "lr": 9.286033889721544e-05}
{"train_loss": 0.3100159466266632, "global_step": 47157, "epoch": 529, "lr": 9.286004034600295e-05}
{"train_loss": 0.37972766160964966, "global_step": 47158, "epoch": 529, "lr": 9.285974178902845e-05}
{"train_loss": 0.26834607124328613, "global_step": 47159, "epoch": 529, "lr": 9.285944322629197e-05}
{"train_loss": 0.2593105435371399, "global_step": 47160, "epoch": 529, "lr": 9.285914465779357e-05}
{"train_loss": 0.40855148434638977, "global_step": 47161, "epoch": 529, "lr": 9.28588460835333e-05}
{"train_loss": 0.24827873706817627, "global_step": 47162, "epoch": 529, "lr": 9.285854750351115e-05}
{"train_loss": 0.3705002963542938, "global_step": 47163, "epoch": 529, "lr": 9.285824891772721e-05}
{"train_loss": 0.35311928391456604, "global_step": 47164, "epoch": 529, "lr": 9.28579503261815e-05}
{"train_loss": 0.26284146308898926, "global_step": 47165, "epoch": 529, "lr": 9.285765172887404e-05}
{"train_loss": 0.2513534426689148, "global_step": 47166, "epoch": 529, "lr": 9.28573531258049e-05}
{"train_loss": 0.2538074254989624, "global_step": 47167, "epoch": 529, "lr": 9.285705451697413e-05}
{"train_loss": 0.38176169991493225, "global_step": 47168, "epoch": 529, "lr": 9.285675590238174e-05}
{"train_loss": 0.2729631079716629, "global_step": 47169, "epoch": 529, "lr": 9.285645728202776e-05, "val_loss": 2.9453773498535156}
{"train_loss": 0.2949703335762024, "global_step": 47170, "epoch": 530, "lr": 9.285615865591228e-05}
{"train_loss": 0.28284451365470886, "global_step": 47171, "epoch": 530, "lr": 9.28558600240353e-05}
{"train_loss": 0.2089817076921463, "global_step": 47172, "epoch": 530, "lr": 9.285556138639689e-05}
{"train_loss": 0.20376142859458923, "global_step": 47173, "epoch": 530, "lr": 9.285526274299706e-05}
{"train_loss": 0.2933604419231415, "global_step": 47174, "epoch": 530, "lr": 9.285496409383585e-05}
{"train_loss": 0.275465726852417, "global_step": 47175, "epoch": 530, "lr": 9.285466543891333e-05}
{"train_loss": 0.3274233937263489, "global_step": 47176, "epoch": 530, "lr": 9.285436677822951e-05}
{"train_loss": 0.21289587020874023, "global_step": 47177, "epoch": 530, "lr": 9.285406811178443e-05}
{"train_loss": 0.20805130898952484, "global_step": 47178, "epoch": 530, "lr": 9.285376943957817e-05}
{"train_loss": 0.1974596232175827, "global_step": 47179, "epoch": 530, "lr": 9.285347076161072e-05}
{"train_loss": 0.2921225130558014, "global_step": 47180, "epoch": 530, "lr": 9.285317207788216e-05}
{"train_loss": 0.23346583545207977, "global_step": 47181, "epoch": 530, "lr": 9.285287338839253e-05}
{"train_loss": 0.25704821944236755, "global_step": 47182, "epoch": 530, "lr": 9.285257469314183e-05}
{"train_loss": 0.3362793028354645, "global_step": 47183, "epoch": 530, "lr": 9.285227599213011e-05}
{"train_loss": 0.3712117075920105, "global_step": 47184, "epoch": 530, "lr": 9.285197728535745e-05}
{"train_loss": 0.3094104826450348, "global_step": 47185, "epoch": 530, "lr": 9.285167857282385e-05}
{"train_loss": 0.22467070817947388, "global_step": 47186, "epoch": 530, "lr": 9.285137985452936e-05}
{"train_loss": 0.2663799822330475, "global_step": 47187, "epoch": 530, "lr": 9.285108113047405e-05}
{"train_loss": 0.27808845043182373, "global_step": 47188, "epoch": 530, "lr": 9.285078240065791e-05}
{"train_loss": 0.27179110050201416, "global_step": 47189, "epoch": 530, "lr": 9.285048366508102e-05}
{"train_loss": 0.31356313824653625, "global_step": 47190, "epoch": 530, "lr": 9.285018492374339e-05}
{"train_loss": 0.2576943039894104, "global_step": 47191, "epoch": 530, "lr": 9.284988617664508e-05}
{"train_loss": 0.3659951984882355, "global_step": 47192, "epoch": 530, "lr": 9.284958742378613e-05}
{"train_loss": 0.3048011362552643, "global_step": 47193, "epoch": 530, "lr": 9.284928866516656e-05}
{"train_loss": 0.20212838053703308, "global_step": 47194, "epoch": 530, "lr": 9.284898990078644e-05}
{"train_loss": 0.26817041635513306, "global_step": 47195, "epoch": 530, "lr": 9.28486911306458e-05}
{"train_loss": 0.2683369517326355, "global_step": 47196, "epoch": 530, "lr": 9.284839235474466e-05}
{"train_loss": 0.3926451504230499, "global_step": 47197, "epoch": 530, "lr": 9.284809357308308e-05}
{"train_loss": 0.19902703166007996, "global_step": 47198, "epoch": 530, "lr": 9.28477947856611e-05}
{"train_loss": 0.35312214493751526, "global_step": 47199, "epoch": 530, "lr": 9.284749599247876e-05}
{"train_loss": 0.17768411338329315, "global_step": 47200, "epoch": 530, "lr": 9.284719719353609e-05}
{"train_loss": 0.2546972632408142, "global_step": 47201, "epoch": 530, "lr": 9.284689838883315e-05}
{"train_loss": 0.3053942620754242, "global_step": 47202, "epoch": 530, "lr": 9.284659957836996e-05}
{"train_loss": 0.2473388910293579, "global_step": 47203, "epoch": 530, "lr": 9.284630076214655e-05}
{"train_loss": 0.18353013694286346, "global_step": 47204, "epoch": 530, "lr": 9.2846001940163e-05}
{"train_loss": 0.2379089593887329, "global_step": 47205, "epoch": 530, "lr": 9.284570311241932e-05}
{"train_loss": 0.30915898084640503, "global_step": 47206, "epoch": 530, "lr": 9.284540427891554e-05}
{"train_loss": 0.2557392716407776, "global_step": 47207, "epoch": 530, "lr": 9.284510543965174e-05}
{"train_loss": 0.34457045793533325, "global_step": 47208, "epoch": 530, "lr": 9.284480659462793e-05}
{"train_loss": 0.27041780948638916, "global_step": 47209, "epoch": 530, "lr": 9.284450774384416e-05}
{"train_loss": 0.17843066155910492, "global_step": 47210, "epoch": 530, "lr": 9.284420888730046e-05}
{"train_loss": 0.17403897643089294, "global_step": 47211, "epoch": 530, "lr": 9.284391002499689e-05}
{"train_loss": 0.3021642863750458, "global_step": 47212, "epoch": 530, "lr": 9.284361115693347e-05}
{"train_loss": 0.29197052121162415, "global_step": 47213, "epoch": 530, "lr": 9.284331228311025e-05}
{"train_loss": 0.327547162771225, "global_step": 47214, "epoch": 530, "lr": 9.284301340352728e-05}
{"train_loss": 0.3130744397640228, "global_step": 47215, "epoch": 530, "lr": 9.284271451818459e-05}
{"train_loss": 0.2583909332752228, "global_step": 47216, "epoch": 530, "lr": 9.284241562708219e-05}
{"train_loss": 0.2363208830356598, "global_step": 47217, "epoch": 530, "lr": 9.284211673022017e-05}
{"train_loss": 0.217678502202034, "global_step": 47218, "epoch": 530, "lr": 9.284181782759856e-05}
{"train_loss": 0.27873495221138, "global_step": 47219, "epoch": 530, "lr": 9.284151891921737e-05}
{"train_loss": 0.18015138804912567, "global_step": 47220, "epoch": 530, "lr": 9.284122000507667e-05}
{"train_loss": 0.3690480589866638, "global_step": 47221, "epoch": 530, "lr": 9.284092108517649e-05}
{"train_loss": 0.24696047604084015, "global_step": 47222, "epoch": 530, "lr": 9.284062215951686e-05}
{"train_loss": 0.24135971069335938, "global_step": 47223, "epoch": 530, "lr": 9.284032322809784e-05}
{"train_loss": 0.3421320617198944, "global_step": 47224, "epoch": 530, "lr": 9.284002429091945e-05}
{"train_loss": 0.2644995152950287, "global_step": 47225, "epoch": 530, "lr": 9.283972534798175e-05}
{"train_loss": 0.2134065181016922, "global_step": 47226, "epoch": 530, "lr": 9.283942639928477e-05}
{"train_loss": 0.2574576437473297, "global_step": 47227, "epoch": 530, "lr": 9.283912744482857e-05}
{"train_loss": 0.28073373436927795, "global_step": 47228, "epoch": 530, "lr": 9.283882848461314e-05}
{"train_loss": 0.32322266697883606, "global_step": 47229, "epoch": 530, "lr": 9.283852951863857e-05}
{"train_loss": 0.3470236659049988, "global_step": 47230, "epoch": 530, "lr": 9.283823054690488e-05}
{"train_loss": 0.28064918518066406, "global_step": 47231, "epoch": 530, "lr": 9.283793156941211e-05}
{"train_loss": 0.27073991298675537, "global_step": 47232, "epoch": 530, "lr": 9.28376325861603e-05}
{"train_loss": 0.24614271521568298, "global_step": 47233, "epoch": 530, "lr": 9.283733359714949e-05}
{"train_loss": 0.25765347480773926, "global_step": 47234, "epoch": 530, "lr": 9.283703460237974e-05}
{"train_loss": 0.2521640956401825, "global_step": 47235, "epoch": 530, "lr": 9.283673560185104e-05}
{"train_loss": 0.30637869238853455, "global_step": 47236, "epoch": 530, "lr": 9.28364365955635e-05}
{"train_loss": 0.2904902398586273, "global_step": 47237, "epoch": 530, "lr": 9.28361375835171e-05}
{"train_loss": 0.40192389488220215, "global_step": 47238, "epoch": 530, "lr": 9.283583856571189e-05}
{"train_loss": 0.2757275104522705, "global_step": 47239, "epoch": 530, "lr": 9.283553954214797e-05}
{"train_loss": 0.3085523247718811, "global_step": 47240, "epoch": 530, "lr": 9.28352405128253e-05}
{"train_loss": 0.26598939299583435, "global_step": 47241, "epoch": 530, "lr": 9.283494147774395e-05}
{"train_loss": 0.2187061309814453, "global_step": 47242, "epoch": 530, "lr": 9.283464243690398e-05}
{"train_loss": 0.2505311369895935, "global_step": 47243, "epoch": 530, "lr": 9.283434339030541e-05}
{"train_loss": 0.35297098755836487, "global_step": 47244, "epoch": 530, "lr": 9.283404433794828e-05}
{"train_loss": 0.19771718978881836, "global_step": 47245, "epoch": 530, "lr": 9.283374527983264e-05}
{"train_loss": 0.22370842099189758, "global_step": 47246, "epoch": 530, "lr": 9.283344621595853e-05}
{"train_loss": 0.28035157918930054, "global_step": 47247, "epoch": 530, "lr": 9.283314714632598e-05}
{"train_loss": 0.3171100914478302, "global_step": 47248, "epoch": 530, "lr": 9.283284807093502e-05}
{"train_loss": 0.2655787765979767, "global_step": 47249, "epoch": 530, "lr": 9.283254898978573e-05}
{"train_loss": 0.19734470546245575, "global_step": 47250, "epoch": 530, "lr": 9.283224990287812e-05}
{"train_loss": 0.20571038126945496, "global_step": 47251, "epoch": 530, "lr": 9.283195081021223e-05}
{"train_loss": 0.36393964290618896, "global_step": 47252, "epoch": 530, "lr": 9.28316517117881e-05}
{"train_loss": 0.3603535592556, "global_step": 47253, "epoch": 530, "lr": 9.283135260760579e-05}
{"train_loss": 0.2651771903038025, "global_step": 47254, "epoch": 530, "lr": 9.283105349766532e-05}
{"train_loss": 0.21665963530540466, "global_step": 47255, "epoch": 530, "lr": 9.283075438196673e-05}
{"train_loss": 0.3246084749698639, "global_step": 47256, "epoch": 530, "lr": 9.283045526051008e-05}
{"train_loss": 0.2236834317445755, "global_step": 47257, "epoch": 530, "lr": 9.28301561332954e-05}
{"train_loss": 0.27169100739312974, "global_step": 47258, "epoch": 530, "lr": 9.282985700032272e-05, "val_loss": 2.96323561668396, "train_action_mse_error": 35.13245391845703}
{"train_loss": 0.3982933461666107, "global_step": 47259, "epoch": 531, "lr": 9.282955786159209e-05}
{"train_loss": 0.2078365981578827, "global_step": 47260, "epoch": 531, "lr": 9.282925871710353e-05}
{"train_loss": 0.2832086384296417, "global_step": 47261, "epoch": 531, "lr": 9.282895956685712e-05}
{"train_loss": 0.24920965731143951, "global_step": 47262, "epoch": 531, "lr": 9.282866041085288e-05}
{"train_loss": 0.2945047616958618, "global_step": 47263, "epoch": 531, "lr": 9.282836124909083e-05}
{"train_loss": 0.23524032533168793, "global_step": 47264, "epoch": 531, "lr": 9.282806208157104e-05}
{"train_loss": 0.24105261266231537, "global_step": 47265, "epoch": 531, "lr": 9.282776290829353e-05}
{"train_loss": 0.28302985429763794, "global_step": 47266, "epoch": 531, "lr": 9.282746372925837e-05}
{"train_loss": 0.3872038722038269, "global_step": 47267, "epoch": 531, "lr": 9.282716454446556e-05}
{"train_loss": 0.29002073407173157, "global_step": 47268, "epoch": 531, "lr": 9.282686535391517e-05}
{"train_loss": 0.2223512828350067, "global_step": 47269, "epoch": 531, "lr": 9.282656615760723e-05}
{"train_loss": 0.35117194056510925, "global_step": 47270, "epoch": 531, "lr": 9.282626695554177e-05}
{"train_loss": 0.21786148846149445, "global_step": 47271, "epoch": 531, "lr": 9.282596774771885e-05}
{"train_loss": 0.2242542952299118, "global_step": 47272, "epoch": 531, "lr": 9.28256685341385e-05}
{"train_loss": 0.3871079981327057, "global_step": 47273, "epoch": 531, "lr": 9.282536931480077e-05}
{"train_loss": 0.2608877718448639, "global_step": 47274, "epoch": 531, "lr": 9.282507008970567e-05}
{"train_loss": 0.2839576005935669, "global_step": 47275, "epoch": 531, "lr": 9.282477085885329e-05}
{"train_loss": 0.29371288418769836, "global_step": 47276, "epoch": 531, "lr": 9.282447162224361e-05}
{"train_loss": 0.2520480453968048, "global_step": 47277, "epoch": 531, "lr": 9.282417237987671e-05}
{"train_loss": 0.1891867071390152, "global_step": 47278, "epoch": 531, "lr": 9.282387313175262e-05}
{"train_loss": 0.24252505600452423, "global_step": 47279, "epoch": 531, "lr": 9.28235738778714e-05}
{"train_loss": 0.315192312002182, "global_step": 47280, "epoch": 531, "lr": 9.282327461823306e-05}
{"train_loss": 0.34777942299842834, "global_step": 47281, "epoch": 531, "lr": 9.282297535283767e-05}
{"train_loss": 0.2209096997976303, "global_step": 47282, "epoch": 531, "lr": 9.282267608168522e-05}
{"train_loss": 0.2646621763706207, "global_step": 47283, "epoch": 531, "lr": 9.28223768047758e-05}
{"train_loss": 0.20885589718818665, "global_step": 47284, "epoch": 531, "lr": 9.282207752210943e-05}
{"train_loss": 0.21362139284610748, "global_step": 47285, "epoch": 531, "lr": 9.282177823368616e-05}
{"train_loss": 0.2593454420566559, "global_step": 47286, "epoch": 531, "lr": 9.282147893950603e-05}
{"train_loss": 0.29828131198883057, "global_step": 47287, "epoch": 531, "lr": 9.282117963956906e-05}
{"train_loss": 0.3085192143917084, "global_step": 47288, "epoch": 531, "lr": 9.28208803338753e-05}
{"train_loss": 0.3022511601448059, "global_step": 47289, "epoch": 531, "lr": 9.282058102242482e-05}
{"train_loss": 0.2079414278268814, "global_step": 47290, "epoch": 531, "lr": 9.282028170521762e-05}
{"train_loss": 0.29095378518104553, "global_step": 47291, "epoch": 531, "lr": 9.281998238225374e-05}
{"train_loss": 0.23692749440670013, "global_step": 47292, "epoch": 531, "lr": 9.281968305353326e-05}
{"train_loss": 0.16928820312023163, "global_step": 47293, "epoch": 531, "lr": 9.281938371905617e-05}
{"train_loss": 0.2525867223739624, "global_step": 47294, "epoch": 531, "lr": 9.281908437882257e-05}
{"train_loss": 0.261001318693161, "global_step": 47295, "epoch": 531, "lr": 9.281878503283243e-05}
{"train_loss": 0.30269932746887207, "global_step": 47296, "epoch": 531, "lr": 9.281848568108585e-05}
{"train_loss": 0.24887613952159882, "global_step": 47297, "epoch": 531, "lr": 9.281818632358283e-05}
{"train_loss": 0.23391395807266235, "global_step": 47298, "epoch": 531, "lr": 9.281788696032344e-05}
{"train_loss": 0.20455677807331085, "global_step": 47299, "epoch": 531, "lr": 9.28175875913077e-05}
{"train_loss": 0.20031051337718964, "global_step": 47300, "epoch": 531, "lr": 9.281728821653566e-05}
{"train_loss": 0.2904502749443054, "global_step": 47301, "epoch": 531, "lr": 9.281698883600735e-05}
{"train_loss": 0.35444706678390503, "global_step": 47302, "epoch": 531, "lr": 9.281668944972283e-05}
{"train_loss": 0.31703296303749084, "global_step": 47303, "epoch": 531, "lr": 9.281639005768212e-05}
{"train_loss": 0.35259613394737244, "global_step": 47304, "epoch": 531, "lr": 9.281609065988528e-05}
{"train_loss": 0.23109695315361023, "global_step": 47305, "epoch": 531, "lr": 9.281579125633233e-05}
{"train_loss": 0.2997797727584839, "global_step": 47306, "epoch": 531, "lr": 9.281549184702331e-05}
{"train_loss": 0.38786134123802185, "global_step": 47307, "epoch": 531, "lr": 9.281519243195829e-05}
{"train_loss": 0.2921365797519684, "global_step": 47308, "epoch": 531, "lr": 9.281489301113727e-05}
{"train_loss": 0.2582230269908905, "global_step": 47309, "epoch": 531, "lr": 9.281459358456031e-05}
{"train_loss": 0.2122565507888794, "global_step": 47310, "epoch": 531, "lr": 9.281429415222746e-05}
{"train_loss": 0.2538439631462097, "global_step": 47311, "epoch": 531, "lr": 9.281399471413874e-05}
{"train_loss": 0.3744523823261261, "global_step": 47312, "epoch": 531, "lr": 9.281369527029423e-05}
{"train_loss": 0.2500195801258087, "global_step": 47313, "epoch": 531, "lr": 9.281339582069391e-05}
{"train_loss": 0.31998148560523987, "global_step": 47314, "epoch": 531, "lr": 9.281309636533784e-05}
{"train_loss": 0.3619837462902069, "global_step": 47315, "epoch": 531, "lr": 9.28127969042261e-05}
{"train_loss": 0.267686128616333, "global_step": 47316, "epoch": 531, "lr": 9.28124974373587e-05}
{"train_loss": 0.26407694816589355, "global_step": 47317, "epoch": 531, "lr": 9.281219796473566e-05}
{"train_loss": 0.41224175691604614, "global_step": 47318, "epoch": 531, "lr": 9.281189848635705e-05}
{"train_loss": 0.41059866547584534, "global_step": 47319, "epoch": 531, "lr": 9.281159900222291e-05}
{"train_loss": 0.3464565575122833, "global_step": 47320, "epoch": 531, "lr": 9.281129951233326e-05}
{"train_loss": 0.2566034495830536, "global_step": 47321, "epoch": 531, "lr": 9.281100001668816e-05}
{"train_loss": 0.2893196642398834, "global_step": 47322, "epoch": 531, "lr": 9.281070051528764e-05}
{"train_loss": 0.33343690633773804, "global_step": 47323, "epoch": 531, "lr": 9.281040100813174e-05}
{"train_loss": 0.4050247371196747, "global_step": 47324, "epoch": 531, "lr": 9.281010149522051e-05}
{"train_loss": 0.22280092537403107, "global_step": 47325, "epoch": 531, "lr": 9.280980197655398e-05}
{"train_loss": 0.3527974486351013, "global_step": 47326, "epoch": 531, "lr": 9.28095024521322e-05}
{"train_loss": 0.3362043499946594, "global_step": 47327, "epoch": 531, "lr": 9.280920292195521e-05}
{"train_loss": 0.35096287727355957, "global_step": 47328, "epoch": 531, "lr": 9.280890338602303e-05}
{"train_loss": 0.3120148479938507, "global_step": 47329, "epoch": 531, "lr": 9.280860384433572e-05}
{"train_loss": 0.4026072323322296, "global_step": 47330, "epoch": 531, "lr": 9.280830429689332e-05}
{"train_loss": 0.300089031457901, "global_step": 47331, "epoch": 531, "lr": 9.280800474369587e-05}
{"train_loss": 0.33457908034324646, "global_step": 47332, "epoch": 531, "lr": 9.280770518474337e-05}
{"train_loss": 0.24454334378242493, "global_step": 47333, "epoch": 531, "lr": 9.280740562003594e-05}
{"train_loss": 0.2926429808139801, "global_step": 47334, "epoch": 531, "lr": 9.280710604957355e-05}
{"train_loss": 0.2929118275642395, "global_step": 47335, "epoch": 531, "lr": 9.280680647335628e-05}
{"train_loss": 0.24690991640090942, "global_step": 47336, "epoch": 531, "lr": 9.280650689138415e-05}
{"train_loss": 0.31200218200683594, "global_step": 47337, "epoch": 531, "lr": 9.280620730365721e-05}
{"train_loss": 0.32608798146247864, "global_step": 47338, "epoch": 531, "lr": 9.28059077101755e-05}
{"train_loss": 0.309318870306015, "global_step": 47339, "epoch": 531, "lr": 9.280560811093905e-05}
{"train_loss": 0.2729772627353668, "global_step": 47340, "epoch": 531, "lr": 9.280530850594791e-05}
{"train_loss": 0.28866666555404663, "global_step": 47341, "epoch": 531, "lr": 9.280500889520211e-05}
{"train_loss": 0.22216877341270447, "global_step": 47342, "epoch": 531, "lr": 9.28047092787017e-05}
{"train_loss": 0.25082746148109436, "global_step": 47343, "epoch": 531, "lr": 9.280440965644672e-05}
{"train_loss": 0.3699359595775604, "global_step": 47344, "epoch": 531, "lr": 9.280411002843724e-05}
{"train_loss": 0.2790570557117462, "global_step": 47345, "epoch": 531, "lr": 9.280381039467323e-05}
{"train_loss": 0.2725676894187927, "global_step": 47346, "epoch": 531, "lr": 9.280351075515479e-05}
{"train_loss": 0.28831901409652794, "global_step": 47347, "epoch": 531, "lr": 9.280321110988192e-05, "val_loss": 2.98189377784729}
{"train_loss": 0.25563302636146545, "global_step": 47348, "epoch": 532, "lr": 9.280291145885469e-05}
{"train_loss": 0.33928602933883667, "global_step": 47349, "epoch": 532, "lr": 9.280261180207314e-05}
{"train_loss": 0.2819523513317108, "global_step": 47350, "epoch": 532, "lr": 9.280231213953728e-05}
{"train_loss": 0.27870693802833557, "global_step": 47351, "epoch": 532, "lr": 9.280201247124718e-05}
{"train_loss": 0.2842637598514557, "global_step": 47352, "epoch": 532, "lr": 9.280171279720289e-05}
{"train_loss": 0.2262001484632492, "global_step": 47353, "epoch": 532, "lr": 9.28014131174044e-05}
{"train_loss": 0.37496519088745117, "global_step": 47354, "epoch": 532, "lr": 9.280111343185181e-05}
{"train_loss": 0.32466110587120056, "global_step": 47355, "epoch": 532, "lr": 9.280081374054513e-05}
{"train_loss": 0.22221508622169495, "global_step": 47356, "epoch": 532, "lr": 9.280051404348438e-05}
{"train_loss": 0.1935567408800125, "global_step": 47357, "epoch": 532, "lr": 9.280021434066964e-05}
{"train_loss": 0.21874813735485077, "global_step": 47358, "epoch": 532, "lr": 9.279991463210091e-05}
{"train_loss": 0.2971099019050598, "global_step": 47359, "epoch": 532, "lr": 9.279961491777828e-05}
{"train_loss": 0.21283192932605743, "global_step": 47360, "epoch": 532, "lr": 9.279931519770175e-05}
{"train_loss": 0.37271344661712646, "global_step": 47361, "epoch": 532, "lr": 9.279901547187137e-05}
{"train_loss": 0.2840454578399658, "global_step": 47362, "epoch": 532, "lr": 9.27987157402872e-05}
{"train_loss": 0.2583640217781067, "global_step": 47363, "epoch": 532, "lr": 9.279841600294925e-05}
{"train_loss": 0.2811720669269562, "global_step": 47364, "epoch": 532, "lr": 9.279811625985759e-05}
{"train_loss": 0.19605205953121185, "global_step": 47365, "epoch": 532, "lr": 9.279781651101223e-05}
{"train_loss": 0.498237669467926, "global_step": 47366, "epoch": 532, "lr": 9.279751675641324e-05}
{"train_loss": 0.3661001920700073, "global_step": 47367, "epoch": 532, "lr": 9.279721699606063e-05}
{"train_loss": 0.3082961142063141, "global_step": 47368, "epoch": 532, "lr": 9.279691722995445e-05}
{"train_loss": 0.24703209102153778, "global_step": 47369, "epoch": 532, "lr": 9.279661745809476e-05}
{"train_loss": 0.2611270248889923, "global_step": 47370, "epoch": 532, "lr": 9.279631768048158e-05}
{"train_loss": 0.2438308298587799, "global_step": 47371, "epoch": 532, "lr": 9.279601789711497e-05}
{"train_loss": 0.23737743496894836, "global_step": 47372, "epoch": 532, "lr": 9.279571810799494e-05}
{"train_loss": 0.3032948672771454, "global_step": 47373, "epoch": 532, "lr": 9.279541831312158e-05}
{"train_loss": 0.2184099406003952, "global_step": 47374, "epoch": 532, "lr": 9.279511851249486e-05}
{"train_loss": 0.2905789315700531, "global_step": 47375, "epoch": 532, "lr": 9.279481870611486e-05}
{"train_loss": 0.27138298749923706, "global_step": 47376, "epoch": 532, "lr": 9.279451889398164e-05}
{"train_loss": 0.24455885589122772, "global_step": 47377, "epoch": 532, "lr": 9.27942190760952e-05}
{"train_loss": 0.24797701835632324, "global_step": 47378, "epoch": 532, "lr": 9.279391925245561e-05}
{"train_loss": 0.2628527879714966, "global_step": 47379, "epoch": 532, "lr": 9.27936194230629e-05}
{"train_loss": 0.2897382378578186, "global_step": 47380, "epoch": 532, "lr": 9.27933195879171e-05}
{"train_loss": 0.23476657271385193, "global_step": 47381, "epoch": 532, "lr": 9.279301974701826e-05}
{"train_loss": 0.2813911437988281, "global_step": 47382, "epoch": 532, "lr": 9.279271990036643e-05}
{"train_loss": 0.2345181107521057, "global_step": 47383, "epoch": 532, "lr": 9.279242004796164e-05}
{"train_loss": 0.26050472259521484, "global_step": 47384, "epoch": 532, "lr": 9.279212018980393e-05}
{"train_loss": 0.2823635935783386, "global_step": 47385, "epoch": 532, "lr": 9.279182032589333e-05}
{"train_loss": 0.3425023555755615, "global_step": 47386, "epoch": 532, "lr": 9.27915204562299e-05}
{"train_loss": 0.2830893397331238, "global_step": 47387, "epoch": 532, "lr": 9.279122058081366e-05}
{"train_loss": 0.20720411837100983, "global_step": 47388, "epoch": 532, "lr": 9.279092069964468e-05}
{"train_loss": 0.2800605297088623, "global_step": 47389, "epoch": 532, "lr": 9.279062081272296e-05}
{"train_loss": 0.2297632247209549, "global_step": 47390, "epoch": 532, "lr": 9.279032092004858e-05}
{"train_loss": 0.29715123772621155, "global_step": 47391, "epoch": 532, "lr": 9.279002102162156e-05}
{"train_loss": 0.22341066598892212, "global_step": 47392, "epoch": 532, "lr": 9.278972111744195e-05}
{"train_loss": 0.2826268672943115, "global_step": 47393, "epoch": 532, "lr": 9.278942120750977e-05}
{"train_loss": 0.23604920506477356, "global_step": 47394, "epoch": 532, "lr": 9.278912129182509e-05}
{"train_loss": 0.3688981235027313, "global_step": 47395, "epoch": 532, "lr": 9.278882137038792e-05}
{"train_loss": 0.22380000352859497, "global_step": 47396, "epoch": 532, "lr": 9.278852144319832e-05}
{"train_loss": 0.2740704417228699, "global_step": 47397, "epoch": 532, "lr": 9.278822151025632e-05}
{"train_loss": 0.3079894185066223, "global_step": 47398, "epoch": 532, "lr": 9.278792157156197e-05}
{"train_loss": 0.32529205083847046, "global_step": 47399, "epoch": 532, "lr": 9.278762162711529e-05}
{"train_loss": 0.29126182198524475, "global_step": 47400, "epoch": 532, "lr": 9.278732167691636e-05}
{"train_loss": 0.29082608222961426, "global_step": 47401, "epoch": 532, "lr": 9.278702172096519e-05}
{"train_loss": 0.28527942299842834, "global_step": 47402, "epoch": 532, "lr": 9.278672175926182e-05}
{"train_loss": 0.38632455468177795, "global_step": 47403, "epoch": 532, "lr": 9.27864217918063e-05}
{"train_loss": 0.2596398890018463, "global_step": 47404, "epoch": 532, "lr": 9.278612181859867e-05}
{"train_loss": 0.25477123260498047, "global_step": 47405, "epoch": 532, "lr": 9.278582183963896e-05}
{"train_loss": 0.2718125283718109, "global_step": 47406, "epoch": 532, "lr": 9.278552185492722e-05}
{"train_loss": 0.2569439113140106, "global_step": 47407, "epoch": 532, "lr": 9.278522186446349e-05}
{"train_loss": 0.28451642394065857, "global_step": 47408, "epoch": 532, "lr": 9.27849218682478e-05}
{"train_loss": 0.32397523522377014, "global_step": 47409, "epoch": 532, "lr": 9.27846218662802e-05}
{"train_loss": 0.33304619789123535, "global_step": 47410, "epoch": 532, "lr": 9.278432185856076e-05}
{"train_loss": 0.3384818434715271, "global_step": 47411, "epoch": 532, "lr": 9.278402184508946e-05}
{"train_loss": 0.23776569962501526, "global_step": 47412, "epoch": 532, "lr": 9.278372182586637e-05}
{"train_loss": 0.2452162206172943, "global_step": 47413, "epoch": 532, "lr": 9.278342180089153e-05}
{"train_loss": 0.3041205108165741, "global_step": 47414, "epoch": 532, "lr": 9.278312177016499e-05}
{"train_loss": 0.27200841903686523, "global_step": 47415, "epoch": 532, "lr": 9.278282173368677e-05}
{"train_loss": 0.22011147439479828, "global_step": 47416, "epoch": 532, "lr": 9.278252169145694e-05}
{"train_loss": 0.30861926078796387, "global_step": 47417, "epoch": 532, "lr": 9.27822216434755e-05}
{"train_loss": 0.254790335893631, "global_step": 47418, "epoch": 532, "lr": 9.278192158974253e-05}
{"train_loss": 0.3223925828933716, "global_step": 47419, "epoch": 532, "lr": 9.278162153025804e-05}
{"train_loss": 0.30671221017837524, "global_step": 47420, "epoch": 532, "lr": 9.278132146502207e-05}
{"train_loss": 0.3070845603942871, "global_step": 47421, "epoch": 532, "lr": 9.27810213940347e-05}
{"train_loss": 0.37548011541366577, "global_step": 47422, "epoch": 532, "lr": 9.278072131729592e-05}
{"train_loss": 0.2129565179347992, "global_step": 47423, "epoch": 532, "lr": 9.278042123480581e-05}
{"train_loss": 0.3292774558067322, "global_step": 47424, "epoch": 532, "lr": 9.278012114656438e-05}
{"train_loss": 0.2942969799041748, "global_step": 47425, "epoch": 532, "lr": 9.27798210525717e-05}
{"train_loss": 0.20997396111488342, "global_step": 47426, "epoch": 532, "lr": 9.277952095282778e-05}
{"train_loss": 0.344844788312912, "global_step": 47427, "epoch": 532, "lr": 9.277922084733268e-05}
{"train_loss": 0.24661725759506226, "global_step": 47428, "epoch": 532, "lr": 9.277892073608645e-05}
{"train_loss": 0.3554214835166931, "global_step": 47429, "epoch": 532, "lr": 9.277862061908909e-05}
{"train_loss": 0.3541703522205353, "global_step": 47430, "epoch": 532, "lr": 9.277832049634067e-05}
{"train_loss": 0.2954951226711273, "global_step": 47431, "epoch": 532, "lr": 9.277802036784124e-05}
{"train_loss": 0.27711209654808044, "global_step": 47432, "epoch": 532, "lr": 9.277772023359082e-05}
{"train_loss": 0.23771634697914124, "global_step": 47433, "epoch": 532, "lr": 9.277742009358947e-05}
{"train_loss": 0.2147868275642395, "global_step": 47434, "epoch": 532, "lr": 9.27771199478372e-05}
{"train_loss": 0.36624953150749207, "global_step": 47435, "epoch": 532, "lr": 9.277681979633407e-05}
{"train_loss": 0.28233498619513564, "global_step": 47436, "epoch": 532, "lr": 9.277651963908012e-05, "val_loss": 2.956310272216797}
{"train_loss": 0.3642320930957794, "global_step": 47437, "epoch": 533, "lr": 9.277621947607538e-05}
{"train_loss": 0.26941776275634766, "global_step": 47438, "epoch": 533, "lr": 9.277591930731992e-05}
{"train_loss": 0.22934982180595398, "global_step": 47439, "epoch": 533, "lr": 9.277561913281373e-05}
{"train_loss": 0.29125940799713135, "global_step": 47440, "epoch": 533, "lr": 9.277531895255692e-05}
{"train_loss": 0.29735881090164185, "global_step": 47441, "epoch": 533, "lr": 9.277501876654946e-05}
{"train_loss": 0.21809637546539307, "global_step": 47442, "epoch": 533, "lr": 9.277471857479142e-05}
{"train_loss": 0.398693710565567, "global_step": 47443, "epoch": 533, "lr": 9.277441837728284e-05}
{"train_loss": 0.2503412365913391, "global_step": 47444, "epoch": 533, "lr": 9.277411817402377e-05}
{"train_loss": 0.3309199810028076, "global_step": 47445, "epoch": 533, "lr": 9.277381796501423e-05}
{"train_loss": 0.30608487129211426, "global_step": 47446, "epoch": 533, "lr": 9.277351775025429e-05}
{"train_loss": 0.21543283760547638, "global_step": 47447, "epoch": 533, "lr": 9.277321752974395e-05}
{"train_loss": 0.19848889112472534, "global_step": 47448, "epoch": 533, "lr": 9.277291730348329e-05}
{"train_loss": 0.36019831895828247, "global_step": 47449, "epoch": 533, "lr": 9.277261707147233e-05}
{"train_loss": 0.31823334097862244, "global_step": 47450, "epoch": 533, "lr": 9.277231683371111e-05}
{"train_loss": 0.32497739791870117, "global_step": 47451, "epoch": 533, "lr": 9.277201659019967e-05}
{"train_loss": 0.3261485993862152, "global_step": 47452, "epoch": 533, "lr": 9.277171634093805e-05}
{"train_loss": 0.459488183259964, "global_step": 47453, "epoch": 533, "lr": 9.27714160859263e-05}
{"train_loss": 0.3264651298522949, "global_step": 47454, "epoch": 533, "lr": 9.277111582516445e-05}
{"train_loss": 0.20346005260944366, "global_step": 47455, "epoch": 533, "lr": 9.277081555865256e-05}
{"train_loss": 0.2834351360797882, "global_step": 47456, "epoch": 533, "lr": 9.277051528639065e-05}
{"train_loss": 0.3228210508823395, "global_step": 47457, "epoch": 533, "lr": 9.277021500837874e-05}
{"train_loss": 0.20818543434143066, "global_step": 47458, "epoch": 533, "lr": 9.276991472461693e-05}
{"train_loss": 0.24259428679943085, "global_step": 47459, "epoch": 533, "lr": 9.276961443510522e-05}
{"train_loss": 0.2001633197069168, "global_step": 47460, "epoch": 533, "lr": 9.276931413984365e-05}
{"train_loss": 0.324788898229599, "global_step": 47461, "epoch": 533, "lr": 9.276901383883225e-05}
{"train_loss": 0.38485950231552124, "global_step": 47462, "epoch": 533, "lr": 9.276871353207109e-05}
{"train_loss": 0.2898249626159668, "global_step": 47463, "epoch": 533, "lr": 9.27684132195602e-05}
{"train_loss": 0.36910736560821533, "global_step": 47464, "epoch": 533, "lr": 9.276811290129963e-05}
{"train_loss": 0.27351659536361694, "global_step": 47465, "epoch": 533, "lr": 9.276781257728939e-05}
{"train_loss": 0.31114667654037476, "global_step": 47466, "epoch": 533, "lr": 9.276751224752955e-05}
{"train_loss": 0.3039630055427551, "global_step": 47467, "epoch": 533, "lr": 9.276721191202013e-05}
{"train_loss": 0.19427940249443054, "global_step": 47468, "epoch": 533, "lr": 9.276691157076119e-05}
{"train_loss": 0.24136440455913544, "global_step": 47469, "epoch": 533, "lr": 9.276661122375275e-05}
{"train_loss": 0.392156183719635, "global_step": 47470, "epoch": 533, "lr": 9.276631087099487e-05}
{"train_loss": 0.3174900412559509, "global_step": 47471, "epoch": 533, "lr": 9.276601051248757e-05}
{"train_loss": 0.33442577719688416, "global_step": 47472, "epoch": 533, "lr": 9.276571014823091e-05}
{"train_loss": 0.2728416919708252, "global_step": 47473, "epoch": 533, "lr": 9.276540977822493e-05}
{"train_loss": 0.2439090758562088, "global_step": 47474, "epoch": 533, "lr": 9.276510940246964e-05}
{"train_loss": 0.27121561765670776, "global_step": 47475, "epoch": 533, "lr": 9.276480902096511e-05}
{"train_loss": 0.34276753664016724, "global_step": 47476, "epoch": 533, "lr": 9.276450863371138e-05}
{"train_loss": 0.24285341799259186, "global_step": 47477, "epoch": 533, "lr": 9.276420824070849e-05}
{"train_loss": 0.23288631439208984, "global_step": 47478, "epoch": 533, "lr": 9.276390784195646e-05}
{"train_loss": 0.24914775788784027, "global_step": 47479, "epoch": 533, "lr": 9.276360743745534e-05}
{"train_loss": 0.24023878574371338, "global_step": 47480, "epoch": 533, "lr": 9.276330702720517e-05}
{"train_loss": 0.2546350061893463, "global_step": 47481, "epoch": 533, "lr": 9.2763006611206e-05}
{"train_loss": 0.32402995228767395, "global_step": 47482, "epoch": 533, "lr": 9.276270618945788e-05}
{"train_loss": 0.2193986475467682, "global_step": 47483, "epoch": 533, "lr": 9.276240576196082e-05}
{"train_loss": 0.31735295057296753, "global_step": 47484, "epoch": 533, "lr": 9.276210532871488e-05}
{"train_loss": 0.28522589802742004, "global_step": 47485, "epoch": 533, "lr": 9.27618048897201e-05}
{"train_loss": 0.21605738997459412, "global_step": 47486, "epoch": 533, "lr": 9.276150444497651e-05}
{"train_loss": 0.2782912850379944, "global_step": 47487, "epoch": 533, "lr": 9.276120399448416e-05}
{"train_loss": 0.2779009938240051, "global_step": 47488, "epoch": 533, "lr": 9.27609035382431e-05}
{"train_loss": 0.2739264667034149, "global_step": 47489, "epoch": 533, "lr": 9.276060307625334e-05}
{"train_loss": 0.3621099889278412, "global_step": 47490, "epoch": 533, "lr": 9.276030260851493e-05}
{"train_loss": 0.2458338886499405, "global_step": 47491, "epoch": 533, "lr": 9.276000213502794e-05}
{"train_loss": 0.27861326932907104, "global_step": 47492, "epoch": 533, "lr": 9.275970165579238e-05}
{"train_loss": 0.3484632074832916, "global_step": 47493, "epoch": 533, "lr": 9.275940117080831e-05}
{"train_loss": 0.2830807566642761, "global_step": 47494, "epoch": 533, "lr": 9.275910068007574e-05}
{"train_loss": 0.2639693319797516, "global_step": 47495, "epoch": 533, "lr": 9.275880018359474e-05}
{"train_loss": 0.21880914270877838, "global_step": 47496, "epoch": 533, "lr": 9.275849968136534e-05}
{"train_loss": 0.32478633522987366, "global_step": 47497, "epoch": 533, "lr": 9.275819917338759e-05}
{"train_loss": 0.29797065258026123, "global_step": 47498, "epoch": 533, "lr": 9.275789865966152e-05}
{"train_loss": 0.26126372814178467, "global_step": 47499, "epoch": 533, "lr": 9.275759814018715e-05}
{"train_loss": 0.3164338171482086, "global_step": 47500, "epoch": 533, "lr": 9.275729761496455e-05}
{"train_loss": 0.2419663965702057, "global_step": 47501, "epoch": 533, "lr": 9.275699708399377e-05}
{"train_loss": 0.4646368622779846, "global_step": 47502, "epoch": 533, "lr": 9.275669654727482e-05}
{"train_loss": 0.23375533521175385, "global_step": 47503, "epoch": 533, "lr": 9.275639600480775e-05}
{"train_loss": 0.24067524075508118, "global_step": 47504, "epoch": 533, "lr": 9.275609545659261e-05}
{"train_loss": 0.207988902926445, "global_step": 47505, "epoch": 533, "lr": 9.275579490262944e-05}
{"train_loss": 0.31921863555908203, "global_step": 47506, "epoch": 533, "lr": 9.275549434291826e-05}
{"train_loss": 0.3318113684654236, "global_step": 47507, "epoch": 533, "lr": 9.275519377745914e-05}
{"train_loss": 0.22535745799541473, "global_step": 47508, "epoch": 533, "lr": 9.275489320625209e-05}
{"train_loss": 0.2878400385379791, "global_step": 47509, "epoch": 533, "lr": 9.275459262929717e-05}
{"train_loss": 0.18216028809547424, "global_step": 47510, "epoch": 533, "lr": 9.275429204659443e-05}
{"train_loss": 0.3307260274887085, "global_step": 47511, "epoch": 533, "lr": 9.275399145814388e-05}
{"train_loss": 0.30289196968078613, "global_step": 47512, "epoch": 533, "lr": 9.275369086394558e-05}
{"train_loss": 0.26793983578681946, "global_step": 47513, "epoch": 533, "lr": 9.275339026399957e-05}
{"train_loss": 0.24038733541965485, "global_step": 47514, "epoch": 533, "lr": 9.27530896583059e-05}
{"train_loss": 0.39526015520095825, "global_step": 47515, "epoch": 533, "lr": 9.275278904686457e-05}
{"train_loss": 0.39025771617889404, "global_step": 47516, "epoch": 533, "lr": 9.275248842967567e-05}
{"train_loss": 0.2286258339881897, "global_step": 47517, "epoch": 533, "lr": 9.27521878067392e-05}
{"train_loss": 0.3426376283168793, "global_step": 47518, "epoch": 533, "lr": 9.275188717805526e-05}
{"train_loss": 0.24570217728614807, "global_step": 47519, "epoch": 533, "lr": 9.275158654362382e-05}
{"train_loss": 0.2980319857597351, "global_step": 47520, "epoch": 533, "lr": 9.275128590344494e-05}
{"train_loss": 0.25723695755004883, "global_step": 47521, "epoch": 533, "lr": 9.275098525751868e-05}
{"train_loss": 0.23998643457889557, "global_step": 47522, "epoch": 533, "lr": 9.275068460584508e-05}
{"train_loss": 0.3411407470703125, "global_step": 47523, "epoch": 533, "lr": 9.275038394842417e-05}
{"train_loss": 0.3489269018173218, "global_step": 47524, "epoch": 533, "lr": 9.275008328525599e-05}
{"train_loss": 0.28685467564657835, "global_step": 47525, "epoch": 533, "lr": 9.274978261634058e-05, "val_loss": 2.912341833114624}
{"train_loss": 0.2679064869880676, "global_step": 47526, "epoch": 534, "lr": 9.274948194167799e-05}
{"train_loss": 0.30707865953445435, "global_step": 47527, "epoch": 534, "lr": 9.274918126126823e-05}
{"train_loss": 0.25927379727363586, "global_step": 47528, "epoch": 534, "lr": 9.274888057511139e-05}
{"train_loss": 0.3282032310962677, "global_step": 47529, "epoch": 534, "lr": 9.274857988320747e-05}
{"train_loss": 0.3764210343360901, "global_step": 47530, "epoch": 534, "lr": 9.274827918555654e-05}
{"train_loss": 0.2782323956489563, "global_step": 47531, "epoch": 534, "lr": 9.274797848215859e-05}
{"train_loss": 0.25005850195884705, "global_step": 47532, "epoch": 534, "lr": 9.274767777301372e-05}
{"train_loss": 0.31943103671073914, "global_step": 47533, "epoch": 534, "lr": 9.274737705812196e-05}
{"train_loss": 0.25292524695396423, "global_step": 47534, "epoch": 534, "lr": 9.274707633748332e-05}
{"train_loss": 0.21804745495319366, "global_step": 47535, "epoch": 534, "lr": 9.274677561109786e-05}
{"train_loss": 0.26688650250434875, "global_step": 47536, "epoch": 534, "lr": 9.27464748789656e-05}
{"train_loss": 0.27862927317619324, "global_step": 47537, "epoch": 534, "lr": 9.274617414108661e-05}
{"train_loss": 0.2544325590133667, "global_step": 47538, "epoch": 534, "lr": 9.274587339746093e-05}
{"train_loss": 0.2809024751186371, "global_step": 47539, "epoch": 534, "lr": 9.274557264808857e-05}
{"train_loss": 0.26479050517082214, "global_step": 47540, "epoch": 534, "lr": 9.27452718929696e-05}
{"train_loss": 0.4071153700351715, "global_step": 47541, "epoch": 534, "lr": 9.274497113210406e-05}
{"train_loss": 0.34684664011001587, "global_step": 47542, "epoch": 534, "lr": 9.274467036549195e-05}
{"train_loss": 0.24296924471855164, "global_step": 47543, "epoch": 534, "lr": 9.274436959313335e-05}
{"train_loss": 0.2507174611091614, "global_step": 47544, "epoch": 534, "lr": 9.27440688150283e-05}
{"train_loss": 0.2647772431373596, "global_step": 47545, "epoch": 534, "lr": 9.274376803117684e-05}
{"train_loss": 0.261190265417099, "global_step": 47546, "epoch": 534, "lr": 9.274346724157898e-05}
{"train_loss": 0.35103949904441833, "global_step": 47547, "epoch": 534, "lr": 9.27431664462348e-05}
{"train_loss": 0.2985502779483795, "global_step": 47548, "epoch": 534, "lr": 9.274286564514431e-05}
{"train_loss": 0.3989863395690918, "global_step": 47549, "epoch": 534, "lr": 9.274256483830757e-05}
{"train_loss": 0.2889579236507416, "global_step": 47550, "epoch": 534, "lr": 9.274226402572458e-05}
{"train_loss": 0.22182480990886688, "global_step": 47551, "epoch": 534, "lr": 9.274196320739546e-05}
{"train_loss": 0.22088320553302765, "global_step": 47552, "epoch": 534, "lr": 9.274166238332019e-05}
{"train_loss": 0.24175158143043518, "global_step": 47553, "epoch": 534, "lr": 9.274136155349882e-05}
{"train_loss": 0.2955515682697296, "global_step": 47554, "epoch": 534, "lr": 9.27410607179314e-05}
{"train_loss": 0.3957374691963196, "global_step": 47555, "epoch": 534, "lr": 9.274075987661795e-05}
{"train_loss": 0.2913498282432556, "global_step": 47556, "epoch": 534, "lr": 9.274045902955854e-05}
{"train_loss": 0.25978416204452515, "global_step": 47557, "epoch": 534, "lr": 9.274015817675319e-05}
{"train_loss": 0.31627368927001953, "global_step": 47558, "epoch": 534, "lr": 9.273985731820195e-05}
{"train_loss": 0.3599269688129425, "global_step": 47559, "epoch": 534, "lr": 9.273955645390485e-05}
{"train_loss": 0.4318215250968933, "global_step": 47560, "epoch": 534, "lr": 9.273925558386196e-05}
{"train_loss": 0.3234947919845581, "global_step": 47561, "epoch": 534, "lr": 9.273895470807328e-05}
{"train_loss": 0.26646316051483154, "global_step": 47562, "epoch": 534, "lr": 9.273865382653885e-05}
{"train_loss": 0.31284308433532715, "global_step": 47563, "epoch": 534, "lr": 9.273835293925877e-05}
{"train_loss": 0.38067442178726196, "global_step": 47564, "epoch": 534, "lr": 9.2738052046233e-05}
{"train_loss": 0.31185296177864075, "global_step": 47565, "epoch": 534, "lr": 9.273775114746164e-05}
{"train_loss": 0.2523651421070099, "global_step": 47566, "epoch": 534, "lr": 9.273745024294471e-05}
{"train_loss": 0.2966954708099365, "global_step": 47567, "epoch": 534, "lr": 9.273714933268225e-05}
{"train_loss": 0.3534995913505554, "global_step": 47568, "epoch": 534, "lr": 9.27368484166743e-05}
{"train_loss": 0.5191720128059387, "global_step": 47569, "epoch": 534, "lr": 9.273654749492091e-05}
{"train_loss": 0.28180569410324097, "global_step": 47570, "epoch": 534, "lr": 9.273624656742211e-05}
{"train_loss": 0.3282572031021118, "global_step": 47571, "epoch": 534, "lr": 9.273594563417794e-05}
{"train_loss": 0.3691428303718567, "global_step": 47572, "epoch": 534, "lr": 9.273564469518843e-05}
{"train_loss": 0.24069862067699432, "global_step": 47573, "epoch": 534, "lr": 9.273534375045364e-05}
{"train_loss": 0.3918473422527313, "global_step": 47574, "epoch": 534, "lr": 9.27350427999736e-05}
{"train_loss": 0.34678199887275696, "global_step": 47575, "epoch": 534, "lr": 9.273474184374837e-05}
{"train_loss": 0.3843853771686554, "global_step": 47576, "epoch": 534, "lr": 9.273444088177797e-05}
{"train_loss": 0.35847941040992737, "global_step": 47577, "epoch": 534, "lr": 9.273413991406243e-05}
{"train_loss": 0.29801684617996216, "global_step": 47578, "epoch": 534, "lr": 9.273383894060182e-05}
{"train_loss": 0.33672383427619934, "global_step": 47579, "epoch": 534, "lr": 9.273353796139617e-05}
{"train_loss": 0.27174484729766846, "global_step": 47580, "epoch": 534, "lr": 9.273323697644551e-05}
{"train_loss": 0.28855839371681213, "global_step": 47581, "epoch": 534, "lr": 9.273293598574988e-05}
{"train_loss": 0.2992008328437805, "global_step": 47582, "epoch": 534, "lr": 9.273263498930934e-05}
{"train_loss": 0.27491986751556396, "global_step": 47583, "epoch": 534, "lr": 9.27323339871239e-05}
{"train_loss": 0.2929644286632538, "global_step": 47584, "epoch": 534, "lr": 9.273203297919365e-05}
{"train_loss": 0.29206088185310364, "global_step": 47585, "epoch": 534, "lr": 9.273173196551858e-05}
{"train_loss": 0.2848000228404999, "global_step": 47586, "epoch": 534, "lr": 9.273143094609876e-05}
{"train_loss": 0.29616427421569824, "global_step": 47587, "epoch": 534, "lr": 9.27311299209342e-05}
{"train_loss": 0.2323714941740036, "global_step": 47588, "epoch": 534, "lr": 9.273082889002496e-05}
{"train_loss": 0.27573007345199585, "global_step": 47589, "epoch": 534, "lr": 9.27305278533711e-05}
{"train_loss": 0.3054850399494171, "global_step": 47590, "epoch": 534, "lr": 9.273022681097263e-05}
{"train_loss": 0.21885204315185547, "global_step": 47591, "epoch": 534, "lr": 9.272992576282961e-05}
{"train_loss": 0.3431229889392853, "global_step": 47592, "epoch": 534, "lr": 9.272962470894206e-05}
{"train_loss": 0.32541710138320923, "global_step": 47593, "epoch": 534, "lr": 9.272932364931004e-05}
{"train_loss": 0.3274344801902771, "global_step": 47594, "epoch": 534, "lr": 9.272902258393358e-05}
{"train_loss": 0.299039363861084, "global_step": 47595, "epoch": 534, "lr": 9.272872151281274e-05}
{"train_loss": 0.26059719920158386, "global_step": 47596, "epoch": 534, "lr": 9.272842043594752e-05}
{"train_loss": 0.2812551259994507, "global_step": 47597, "epoch": 534, "lr": 9.2728119353338e-05}
{"train_loss": 0.28546324372291565, "global_step": 47598, "epoch": 534, "lr": 9.27278182649842e-05}
{"train_loss": 0.23537959158420563, "global_step": 47599, "epoch": 534, "lr": 9.272751717088616e-05}
{"train_loss": 0.46589869260787964, "global_step": 47600, "epoch": 534, "lr": 9.272721607104393e-05}
{"train_loss": 0.29599568247795105, "global_step": 47601, "epoch": 534, "lr": 9.272691496545756e-05}
{"train_loss": 0.4064823091030121, "global_step": 47602, "epoch": 534, "lr": 9.272661385412707e-05}
{"train_loss": 0.22704628109931946, "global_step": 47603, "epoch": 534, "lr": 9.272631273705249e-05}
{"train_loss": 0.3310323655605316, "global_step": 47604, "epoch": 534, "lr": 9.272601161423389e-05}
{"train_loss": 0.25333938002586365, "global_step": 47605, "epoch": 534, "lr": 9.27257104856713e-05}
{"train_loss": 0.40550655126571655, "global_step": 47606, "epoch": 534, "lr": 9.272540935136475e-05}
{"train_loss": 0.31148964166641235, "global_step": 47607, "epoch": 534, "lr": 9.27251082113143e-05}
{"train_loss": 0.3223268389701843, "global_step": 47608, "epoch": 534, "lr": 9.272480706551998e-05}
{"train_loss": 0.30941125750541687, "global_step": 47609, "epoch": 534, "lr": 9.272450591398182e-05}
{"train_loss": 0.2622009813785553, "global_step": 47610, "epoch": 534, "lr": 9.272420475669989e-05}
{"train_loss": 0.19393786787986755, "global_step": 47611, "epoch": 534, "lr": 9.27239035936742e-05}
{"train_loss": 0.3796774446964264, "global_step": 47612, "epoch": 534, "lr": 9.272360242490481e-05}
{"train_loss": 0.3076122999191284, "global_step": 47613, "epoch": 534, "lr": 9.272330125039175e-05}
{"train_loss": 0.30468423379940934, "global_step": 47614, "epoch": 534, "lr": 9.272300007013505e-05, "val_loss": 2.8160526752471924}
{"train_loss": 0.2554227411746979, "global_step": 47615, "epoch": 535, "lr": 9.272269888413478e-05}
{"train_loss": 0.29667747020721436, "global_step": 47616, "epoch": 535, "lr": 9.272239769239095e-05}
{"train_loss": 0.2376016527414322, "global_step": 47617, "epoch": 535, "lr": 9.272209649490362e-05}
{"train_loss": 0.26834577322006226, "global_step": 47618, "epoch": 535, "lr": 9.272179529167282e-05}
{"train_loss": 0.3198878765106201, "global_step": 47619, "epoch": 535, "lr": 9.27214940826986e-05}
{"train_loss": 0.26213473081588745, "global_step": 47620, "epoch": 535, "lr": 9.2721192867981e-05}
{"train_loss": 0.33835136890411377, "global_step": 47621, "epoch": 535, "lr": 9.272089164752005e-05}
{"train_loss": 0.28507840633392334, "global_step": 47622, "epoch": 535, "lr": 9.27205904213158e-05}
{"train_loss": 0.21764619648456573, "global_step": 47623, "epoch": 535, "lr": 9.272028918936829e-05}
{"train_loss": 0.2429925501346588, "global_step": 47624, "epoch": 535, "lr": 9.271998795167754e-05}
{"train_loss": 0.22406260669231415, "global_step": 47625, "epoch": 535, "lr": 9.271968670824362e-05}
{"train_loss": 0.3060765564441681, "global_step": 47626, "epoch": 535, "lr": 9.271938545906657e-05}
{"train_loss": 0.22727949917316437, "global_step": 47627, "epoch": 535, "lr": 9.271908420414641e-05}
{"train_loss": 0.30762356519699097, "global_step": 47628, "epoch": 535, "lr": 9.27187829434832e-05}
{"train_loss": 0.29027971625328064, "global_step": 47629, "epoch": 535, "lr": 9.271848167707696e-05}
{"train_loss": 0.19957804679870605, "global_step": 47630, "epoch": 535, "lr": 9.271818040492772e-05}
{"train_loss": 0.21970194578170776, "global_step": 47631, "epoch": 535, "lr": 9.271787912703557e-05}
{"train_loss": 0.29685521125793457, "global_step": 47632, "epoch": 535, "lr": 9.271757784340052e-05}
{"train_loss": 0.29935574531555176, "global_step": 47633, "epoch": 535, "lr": 9.27172765540226e-05}
{"train_loss": 0.2638658881187439, "global_step": 47634, "epoch": 535, "lr": 9.271697525890187e-05}
{"train_loss": 0.27949827909469604, "global_step": 47635, "epoch": 535, "lr": 9.271667395803836e-05}
{"train_loss": 0.18376530706882477, "global_step": 47636, "epoch": 535, "lr": 9.271637265143212e-05}
{"train_loss": 0.27627238631248474, "global_step": 47637, "epoch": 535, "lr": 9.271607133908317e-05}
{"train_loss": 0.2835906744003296, "global_step": 47638, "epoch": 535, "lr": 9.271577002099159e-05}
{"train_loss": 0.29173919558525085, "global_step": 47639, "epoch": 535, "lr": 9.271546869715738e-05}
{"train_loss": 0.21557404100894928, "global_step": 47640, "epoch": 535, "lr": 9.271516736758059e-05}
{"train_loss": 0.33818042278289795, "global_step": 47641, "epoch": 535, "lr": 9.271486603226127e-05}
{"train_loss": 0.23213288187980652, "global_step": 47642, "epoch": 535, "lr": 9.271456469119947e-05}
{"train_loss": 0.293293833732605, "global_step": 47643, "epoch": 535, "lr": 9.271426334439521e-05}
{"train_loss": 0.2803272604942322, "global_step": 47644, "epoch": 535, "lr": 9.271396199184853e-05}
{"train_loss": 0.25157925486564636, "global_step": 47645, "epoch": 535, "lr": 9.271366063355949e-05}
{"train_loss": 0.23626399040222168, "global_step": 47646, "epoch": 535, "lr": 9.271335926952811e-05}
{"train_loss": 0.24242550134658813, "global_step": 47647, "epoch": 535, "lr": 9.271305789975444e-05}
{"train_loss": 0.2551073431968689, "global_step": 47648, "epoch": 535, "lr": 9.271275652423852e-05}
{"train_loss": 0.16562005877494812, "global_step": 47649, "epoch": 535, "lr": 9.27124551429804e-05}
{"train_loss": 0.27208343148231506, "global_step": 47650, "epoch": 535, "lr": 9.27121537559801e-05}
{"train_loss": 0.27251875400543213, "global_step": 47651, "epoch": 535, "lr": 9.271185236323768e-05}
{"train_loss": 0.2033083289861679, "global_step": 47652, "epoch": 535, "lr": 9.271155096475315e-05}
{"train_loss": 0.3139936625957489, "global_step": 47653, "epoch": 535, "lr": 9.27112495605266e-05}
{"train_loss": 0.2718629837036133, "global_step": 47654, "epoch": 535, "lr": 9.271094815055802e-05}
{"train_loss": 0.2507275640964508, "global_step": 47655, "epoch": 535, "lr": 9.27106467348475e-05}
{"train_loss": 0.28550195693969727, "global_step": 47656, "epoch": 535, "lr": 9.271034531339504e-05}
{"train_loss": 0.270949125289917, "global_step": 47657, "epoch": 535, "lr": 9.271004388620069e-05}
{"train_loss": 0.22090022265911102, "global_step": 47658, "epoch": 535, "lr": 9.27097424532645e-05}
{"train_loss": 0.24978727102279663, "global_step": 47659, "epoch": 535, "lr": 9.270944101458651e-05}
{"train_loss": 0.27073684334754944, "global_step": 47660, "epoch": 535, "lr": 9.270913957016675e-05}
{"train_loss": 0.2870188355445862, "global_step": 47661, "epoch": 535, "lr": 9.270883812000528e-05}
{"train_loss": 0.26481232047080994, "global_step": 47662, "epoch": 535, "lr": 9.270853666410211e-05}
{"train_loss": 0.2522825002670288, "global_step": 47663, "epoch": 535, "lr": 9.27082352024573e-05}
{"train_loss": 0.20853255689144135, "global_step": 47664, "epoch": 535, "lr": 9.270793373507089e-05}
{"train_loss": 0.35921409726142883, "global_step": 47665, "epoch": 535, "lr": 9.270763226194292e-05}
{"train_loss": 0.33223557472229004, "global_step": 47666, "epoch": 535, "lr": 9.270733078307344e-05}
{"train_loss": 0.2353004664182663, "global_step": 47667, "epoch": 535, "lr": 9.270702929846247e-05}
{"train_loss": 0.2137685865163803, "global_step": 47668, "epoch": 535, "lr": 9.270672780811005e-05}
{"train_loss": 0.23831596970558167, "global_step": 47669, "epoch": 535, "lr": 9.270642631201625e-05}
{"train_loss": 0.23832039535045624, "global_step": 47670, "epoch": 535, "lr": 9.270612481018109e-05}
{"train_loss": 0.26300668716430664, "global_step": 47671, "epoch": 535, "lr": 9.27058233026046e-05}
{"train_loss": 0.23778972029685974, "global_step": 47672, "epoch": 535, "lr": 9.270552178928684e-05}
{"train_loss": 0.3134627938270569, "global_step": 47673, "epoch": 535, "lr": 9.270522027022783e-05}
{"train_loss": 0.229157954454422, "global_step": 47674, "epoch": 535, "lr": 9.270491874542764e-05}
{"train_loss": 0.30609163641929626, "global_step": 47675, "epoch": 535, "lr": 9.27046172148863e-05}
{"train_loss": 0.30195027589797974, "global_step": 47676, "epoch": 535, "lr": 9.270431567860382e-05}
{"train_loss": 0.28577542304992676, "global_step": 47677, "epoch": 535, "lr": 9.270401413658028e-05}
{"train_loss": 0.2808186113834381, "global_step": 47678, "epoch": 535, "lr": 9.270371258881571e-05}
{"train_loss": 0.24924886226654053, "global_step": 47679, "epoch": 535, "lr": 9.270341103531014e-05}
{"train_loss": 0.23475924134254456, "global_step": 47680, "epoch": 535, "lr": 9.270310947606363e-05}
{"train_loss": 0.2602477967739105, "global_step": 47681, "epoch": 535, "lr": 9.270280791107618e-05}
{"train_loss": 0.26630479097366333, "global_step": 47682, "epoch": 535, "lr": 9.270250634034787e-05}
{"train_loss": 0.3356221318244934, "global_step": 47683, "epoch": 535, "lr": 9.270220476387874e-05}
{"train_loss": 0.32553979754447937, "global_step": 47684, "epoch": 535, "lr": 9.27019031816688e-05}
{"train_loss": 0.46783027052879333, "global_step": 47685, "epoch": 535, "lr": 9.270160159371813e-05}
{"train_loss": 0.23551519215106964, "global_step": 47686, "epoch": 535, "lr": 9.270130000002674e-05}
{"train_loss": 0.3186246156692505, "global_step": 47687, "epoch": 535, "lr": 9.270099840059468e-05}
{"train_loss": 0.24736717343330383, "global_step": 47688, "epoch": 535, "lr": 9.2700696795422e-05}
{"train_loss": 0.3402027189731598, "global_step": 47689, "epoch": 535, "lr": 9.270039518450871e-05}
{"train_loss": 0.19589991867542267, "global_step": 47690, "epoch": 535, "lr": 9.27000935678549e-05}
{"train_loss": 0.253251314163208, "global_step": 47691, "epoch": 535, "lr": 9.269979194546057e-05}
{"train_loss": 0.25068870186805725, "global_step": 47692, "epoch": 535, "lr": 9.269949031732576e-05}
{"train_loss": 0.32143905758857727, "global_step": 47693, "epoch": 535, "lr": 9.269918868345054e-05}
{"train_loss": 0.30977779626846313, "global_step": 47694, "epoch": 535, "lr": 9.269888704383494e-05}
{"train_loss": 0.26009809970855713, "global_step": 47695, "epoch": 535, "lr": 9.269858539847897e-05}
{"train_loss": 0.31066107749938965, "global_step": 47696, "epoch": 535, "lr": 9.269828374738271e-05}
{"train_loss": 0.25112682580947876, "global_step": 47697, "epoch": 535, "lr": 9.26979820905462e-05}
{"train_loss": 0.36105188727378845, "global_step": 47698, "epoch": 535, "lr": 9.269768042796944e-05}
{"train_loss": 0.27810025215148926, "global_step": 47699, "epoch": 535, "lr": 9.269737875965252e-05}
{"train_loss": 0.3521658778190613, "global_step": 47700, "epoch": 535, "lr": 9.269707708559545e-05}
{"train_loss": 0.2874962389469147, "global_step": 47701, "epoch": 535, "lr": 9.269677540579828e-05}
{"train_loss": 0.31052064895629883, "global_step": 47702, "epoch": 535, "lr": 9.269647372026105e-05}
{"train_loss": 0.2716230071997375, "global_step": 47703, "epoch": 535, "lr": 9.26961720289838e-05, "val_loss": 2.9708573818206787, "train_action_mse_error": 27.253173828125}
{"train_loss": 0.22543594241142273, "global_step": 47704, "epoch": 536, "lr": 9.269587033196656e-05}
{"train_loss": 0.1312725991010666, "global_step": 47705, "epoch": 536, "lr": 9.269556862920939e-05}
{"train_loss": 0.30975663661956787, "global_step": 47706, "epoch": 536, "lr": 9.26952669207123e-05}
{"train_loss": 0.24363750219345093, "global_step": 47707, "epoch": 536, "lr": 9.269496520647538e-05}
{"train_loss": 0.33470332622528076, "global_step": 47708, "epoch": 536, "lr": 9.269466348649864e-05}
{"train_loss": 0.26191452145576477, "global_step": 47709, "epoch": 536, "lr": 9.269436176078213e-05}
{"train_loss": 0.4471275508403778, "global_step": 47710, "epoch": 536, "lr": 9.269406002932586e-05}
{"train_loss": 0.34219056367874146, "global_step": 47711, "epoch": 536, "lr": 9.26937582921299e-05}
{"train_loss": 0.2432868480682373, "global_step": 47712, "epoch": 536, "lr": 9.269345654919428e-05}
{"train_loss": 0.3097206652164459, "global_step": 47713, "epoch": 536, "lr": 9.269315480051906e-05}
{"train_loss": 0.2303621917963028, "global_step": 47714, "epoch": 536, "lr": 9.269285304610425e-05}
{"train_loss": 0.25628724694252014, "global_step": 47715, "epoch": 536, "lr": 9.269255128594991e-05}
{"train_loss": 0.2918129861354828, "global_step": 47716, "epoch": 536, "lr": 9.269224952005609e-05}
{"train_loss": 0.2781921923160553, "global_step": 47717, "epoch": 536, "lr": 9.269194774842282e-05}
{"train_loss": 0.3271350562572479, "global_step": 47718, "epoch": 536, "lr": 9.269164597105011e-05}
{"train_loss": 0.3366508483886719, "global_step": 47719, "epoch": 536, "lr": 9.269134418793806e-05}
{"train_loss": 0.3405395746231079, "global_step": 47720, "epoch": 536, "lr": 9.269104239908665e-05}
{"train_loss": 0.21290774643421173, "global_step": 47721, "epoch": 536, "lr": 9.269074060449598e-05}
{"train_loss": 0.20260393619537354, "global_step": 47722, "epoch": 536, "lr": 9.269043880416604e-05}
{"train_loss": 0.2639560103416443, "global_step": 47723, "epoch": 536, "lr": 9.269013699809689e-05}
{"train_loss": 0.3405621647834778, "global_step": 47724, "epoch": 536, "lr": 9.268983518628857e-05}
{"train_loss": 0.29364919662475586, "global_step": 47725, "epoch": 536, "lr": 9.268953336874113e-05}
{"train_loss": 0.2993098497390747, "global_step": 47726, "epoch": 536, "lr": 9.268923154545461e-05}
{"train_loss": 0.3363155126571655, "global_step": 47727, "epoch": 536, "lr": 9.268892971642902e-05}
{"train_loss": 0.22461852431297302, "global_step": 47728, "epoch": 536, "lr": 9.268862788166444e-05}
{"train_loss": 0.2760600745677948, "global_step": 47729, "epoch": 536, "lr": 9.26883260411609e-05}
{"train_loss": 0.2336554229259491, "global_step": 47730, "epoch": 536, "lr": 9.268802419491843e-05}
{"train_loss": 0.2705516219139099, "global_step": 47731, "epoch": 536, "lr": 9.268772234293707e-05}
{"train_loss": 0.29307588934898376, "global_step": 47732, "epoch": 536, "lr": 9.268742048521687e-05}
{"train_loss": 0.3057306110858917, "global_step": 47733, "epoch": 536, "lr": 9.268711862175786e-05}
{"train_loss": 0.23189964890480042, "global_step": 47734, "epoch": 536, "lr": 9.26868167525601e-05}
{"train_loss": 0.2705959975719452, "global_step": 47735, "epoch": 536, "lr": 9.268651487762361e-05}
{"train_loss": 0.3135088384151459, "global_step": 47736, "epoch": 536, "lr": 9.268621299694843e-05}
{"train_loss": 0.2707091271877289, "global_step": 47737, "epoch": 536, "lr": 9.268591111053462e-05}
{"train_loss": 0.31848472356796265, "global_step": 47738, "epoch": 536, "lr": 9.268560921838221e-05}
{"train_loss": 0.23415039479732513, "global_step": 47739, "epoch": 536, "lr": 9.268530732049124e-05}
{"train_loss": 0.39305436611175537, "global_step": 47740, "epoch": 536, "lr": 9.268500541686174e-05}
{"train_loss": 0.37181729078292847, "global_step": 47741, "epoch": 536, "lr": 9.268470350749376e-05}
{"train_loss": 0.2149554193019867, "global_step": 47742, "epoch": 536, "lr": 9.268440159238736e-05}
{"train_loss": 0.20553092658519745, "global_step": 47743, "epoch": 536, "lr": 9.268409967154255e-05}
{"train_loss": 0.2378505915403366, "global_step": 47744, "epoch": 536, "lr": 9.268379774495939e-05}
{"train_loss": 0.187961146235466, "global_step": 47745, "epoch": 536, "lr": 9.268349581263791e-05}
{"train_loss": 0.33426254987716675, "global_step": 47746, "epoch": 536, "lr": 9.268319387457816e-05}
{"train_loss": 0.35900577902793884, "global_step": 47747, "epoch": 536, "lr": 9.268289193078017e-05}
{"train_loss": 0.3082523047924042, "global_step": 47748, "epoch": 536, "lr": 9.268258998124399e-05}
{"train_loss": 0.2865845561027527, "global_step": 47749, "epoch": 536, "lr": 9.268228802596963e-05}
{"train_loss": 0.2878269851207733, "global_step": 47750, "epoch": 536, "lr": 9.268198606495719e-05}
{"train_loss": 0.36386576294898987, "global_step": 47751, "epoch": 536, "lr": 9.268168409820666e-05}
{"train_loss": 0.23519229888916016, "global_step": 47752, "epoch": 536, "lr": 9.268138212571811e-05}
{"train_loss": 0.44311758875846863, "global_step": 47753, "epoch": 536, "lr": 9.268108014749155e-05}
{"train_loss": 0.22186782956123352, "global_step": 47754, "epoch": 536, "lr": 9.268077816352705e-05}
{"train_loss": 0.3434722423553467, "global_step": 47755, "epoch": 536, "lr": 9.268047617382465e-05}
{"train_loss": 0.2615889310836792, "global_step": 47756, "epoch": 536, "lr": 9.268017417838436e-05}
{"train_loss": 0.3759957253932953, "global_step": 47757, "epoch": 536, "lr": 9.267987217720626e-05}
{"train_loss": 0.26842761039733887, "global_step": 47758, "epoch": 536, "lr": 9.267957017029035e-05}
{"train_loss": 0.2989494204521179, "global_step": 47759, "epoch": 536, "lr": 9.26792681576367e-05}
{"train_loss": 0.30602192878723145, "global_step": 47760, "epoch": 536, "lr": 9.267896613924534e-05}
{"train_loss": 0.24789080023765564, "global_step": 47761, "epoch": 536, "lr": 9.267866411511633e-05}
{"train_loss": 0.2789995074272156, "global_step": 47762, "epoch": 536, "lr": 9.267836208524968e-05}
{"train_loss": 0.2668150067329407, "global_step": 47763, "epoch": 536, "lr": 9.267806004964545e-05}
{"train_loss": 0.1937173306941986, "global_step": 47764, "epoch": 536, "lr": 9.267775800830367e-05}
{"train_loss": 0.27616190910339355, "global_step": 47765, "epoch": 536, "lr": 9.267745596122439e-05}
{"train_loss": 0.21541765332221985, "global_step": 47766, "epoch": 536, "lr": 9.267715390840765e-05}
{"train_loss": 0.3460726737976074, "global_step": 47767, "epoch": 536, "lr": 9.267685184985348e-05}
{"train_loss": 0.23675373196601868, "global_step": 47768, "epoch": 536, "lr": 9.267654978556193e-05}
{"train_loss": 0.24695950746536255, "global_step": 47769, "epoch": 536, "lr": 9.267624771553303e-05}
{"train_loss": 0.20658937096595764, "global_step": 47770, "epoch": 536, "lr": 9.267594563976683e-05}
{"train_loss": 0.16658085584640503, "global_step": 47771, "epoch": 536, "lr": 9.267564355826336e-05}
{"train_loss": 0.29198506474494934, "global_step": 47772, "epoch": 536, "lr": 9.26753414710227e-05}
{"train_loss": 0.3216579258441925, "global_step": 47773, "epoch": 536, "lr": 9.267503937804485e-05}
{"train_loss": 0.20852039754390717, "global_step": 47774, "epoch": 536, "lr": 9.267473727932985e-05}
{"train_loss": 0.1990852952003479, "global_step": 47775, "epoch": 536, "lr": 9.267443517487776e-05}
{"train_loss": 0.35135817527770996, "global_step": 47776, "epoch": 536, "lr": 9.267413306468861e-05}
{"train_loss": 0.3078083097934723, "global_step": 47777, "epoch": 536, "lr": 9.267383094876245e-05}
{"train_loss": 0.20692558586597443, "global_step": 47778, "epoch": 536, "lr": 9.26735288270993e-05}
{"train_loss": 0.2236141562461853, "global_step": 47779, "epoch": 536, "lr": 9.267322669969923e-05}
{"train_loss": 0.23036278784275055, "global_step": 47780, "epoch": 536, "lr": 9.267292456656225e-05}
{"train_loss": 0.23221157491207123, "global_step": 47781, "epoch": 536, "lr": 9.267262242768842e-05}
{"train_loss": 0.25401467084884644, "global_step": 47782, "epoch": 536, "lr": 9.267232028307778e-05}
{"train_loss": 0.33623290061950684, "global_step": 47783, "epoch": 536, "lr": 9.267201813273035e-05}
{"train_loss": 0.36212876439094543, "global_step": 47784, "epoch": 536, "lr": 9.267171597664621e-05}
{"train_loss": 0.3852337598800659, "global_step": 47785, "epoch": 536, "lr": 9.267141381482536e-05}
{"train_loss": 0.18822914361953735, "global_step": 47786, "epoch": 536, "lr": 9.267111164726787e-05}
{"train_loss": 0.23585842549800873, "global_step": 47787, "epoch": 536, "lr": 9.267080947397376e-05}
{"train_loss": 0.25752297043800354, "global_step": 47788, "epoch": 536, "lr": 9.267050729494309e-05}
{"train_loss": 0.3490372896194458, "global_step": 47789, "epoch": 536, "lr": 9.267020511017588e-05}
{"train_loss": 0.26726239919662476, "global_step": 47790, "epoch": 536, "lr": 9.266990291967219e-05}
{"train_loss": 0.2149718850851059, "global_step": 47791, "epoch": 536, "lr": 9.266960072343205e-05}
{"train_loss": 0.2776141454664509, "global_step": 47792, "epoch": 536, "lr": 9.266929852145551e-05, "val_loss": 2.9523327350616455}
{"train_loss": 0.26203131675720215, "global_step": 47793, "epoch": 537, "lr": 9.266899631374258e-05}
{"train_loss": 0.2681647837162018, "global_step": 47794, "epoch": 537, "lr": 9.266869410029336e-05}
{"train_loss": 0.20903752744197845, "global_step": 47795, "epoch": 537, "lr": 9.26683918811078e-05}
{"train_loss": 0.32251396775245667, "global_step": 47796, "epoch": 537, "lr": 9.266808965618604e-05}
{"train_loss": 0.26466038823127747, "global_step": 47797, "epoch": 537, "lr": 9.266778742552806e-05}
{"train_loss": 0.2874998152256012, "global_step": 47798, "epoch": 537, "lr": 9.266748518913393e-05}
{"train_loss": 0.22853586077690125, "global_step": 47799, "epoch": 537, "lr": 9.266718294700366e-05}
{"train_loss": 0.34226900339126587, "global_step": 47800, "epoch": 537, "lr": 9.266688069913731e-05}
{"train_loss": 0.42299962043762207, "global_step": 47801, "epoch": 537, "lr": 9.266657844553493e-05}
{"train_loss": 0.3082987368106842, "global_step": 47802, "epoch": 537, "lr": 9.266627618619652e-05}
{"train_loss": 0.22875425219535828, "global_step": 47803, "epoch": 537, "lr": 9.266597392112217e-05}
{"train_loss": 0.259001225233078, "global_step": 47804, "epoch": 537, "lr": 9.26656716503119e-05}
{"train_loss": 0.22543947398662567, "global_step": 47805, "epoch": 537, "lr": 9.266536937376575e-05}
{"train_loss": 0.26200199127197266, "global_step": 47806, "epoch": 537, "lr": 9.266506709148374e-05}
{"train_loss": 0.2683730721473694, "global_step": 47807, "epoch": 537, "lr": 9.266476480346597e-05}
{"train_loss": 0.3427959382534027, "global_step": 47808, "epoch": 537, "lr": 9.266446250971242e-05}
{"train_loss": 0.2749123275279999, "global_step": 47809, "epoch": 537, "lr": 9.266416021022315e-05}
{"train_loss": 0.2459825724363327, "global_step": 47810, "epoch": 537, "lr": 9.266385790499821e-05}
{"train_loss": 0.3327224552631378, "global_step": 47811, "epoch": 537, "lr": 9.266355559403763e-05}
{"train_loss": 0.2090488076210022, "global_step": 47812, "epoch": 537, "lr": 9.266325327734147e-05}
{"train_loss": 0.2947057783603668, "global_step": 47813, "epoch": 537, "lr": 9.266295095490974e-05}
{"train_loss": 0.2835051119327545, "global_step": 47814, "epoch": 537, "lr": 9.26626486267425e-05}
{"train_loss": 0.29858264327049255, "global_step": 47815, "epoch": 537, "lr": 9.266234629283978e-05}
{"train_loss": 0.2861102819442749, "global_step": 47816, "epoch": 537, "lr": 9.266204395320163e-05}
{"train_loss": 0.2703847587108612, "global_step": 47817, "epoch": 537, "lr": 9.26617416078281e-05}
{"train_loss": 0.3275277018547058, "global_step": 47818, "epoch": 537, "lr": 9.266143925671921e-05}
{"train_loss": 0.2861690819263458, "global_step": 47819, "epoch": 537, "lr": 9.266113689987502e-05}
{"train_loss": 0.29647237062454224, "global_step": 47820, "epoch": 537, "lr": 9.266083453729555e-05}
{"train_loss": 0.27532723546028137, "global_step": 47821, "epoch": 537, "lr": 9.266053216898085e-05}
{"train_loss": 0.2984631359577179, "global_step": 47822, "epoch": 537, "lr": 9.266022979493097e-05}
{"train_loss": 0.26768141984939575, "global_step": 47823, "epoch": 537, "lr": 9.265992741514594e-05}
{"train_loss": 0.2613963484764099, "global_step": 47824, "epoch": 537, "lr": 9.26596250296258e-05}
{"train_loss": 0.2756480276584625, "global_step": 47825, "epoch": 537, "lr": 9.265932263837058e-05}
{"train_loss": 0.2579282522201538, "global_step": 47826, "epoch": 537, "lr": 9.265902024138036e-05}
{"train_loss": 0.28315702080726624, "global_step": 47827, "epoch": 537, "lr": 9.265871783865513e-05}
{"train_loss": 0.2976519465446472, "global_step": 47828, "epoch": 537, "lr": 9.265841543019497e-05}
{"train_loss": 0.3954807221889496, "global_step": 47829, "epoch": 537, "lr": 9.26581130159999e-05}
{"train_loss": 0.30378061532974243, "global_step": 47830, "epoch": 537, "lr": 9.265781059606997e-05}
{"train_loss": 0.31359368562698364, "global_step": 47831, "epoch": 537, "lr": 9.265750817040521e-05}
{"train_loss": 0.25419238209724426, "global_step": 47832, "epoch": 537, "lr": 9.265720573900567e-05}
{"train_loss": 0.22457030415534973, "global_step": 47833, "epoch": 537, "lr": 9.26569033018714e-05}
{"train_loss": 0.2857114374637604, "global_step": 47834, "epoch": 537, "lr": 9.265660085900242e-05}
{"train_loss": 0.3457714915275574, "global_step": 47835, "epoch": 537, "lr": 9.265629841039878e-05}
{"train_loss": 0.2118997424840927, "global_step": 47836, "epoch": 537, "lr": 9.265599595606052e-05}
{"train_loss": 0.3490317463874817, "global_step": 47837, "epoch": 537, "lr": 9.265569349598769e-05}
{"train_loss": 0.3780515193939209, "global_step": 47838, "epoch": 537, "lr": 9.265539103018031e-05}
{"train_loss": 0.20622481405735016, "global_step": 47839, "epoch": 537, "lr": 9.265508855863845e-05}
{"train_loss": 0.30907386541366577, "global_step": 47840, "epoch": 537, "lr": 9.265478608136211e-05}
{"train_loss": 0.34174659848213196, "global_step": 47841, "epoch": 537, "lr": 9.265448359835136e-05}
{"train_loss": 0.2861744463443756, "global_step": 47842, "epoch": 537, "lr": 9.265418110960625e-05}
{"train_loss": 0.2552109956741333, "global_step": 47843, "epoch": 537, "lr": 9.265387861512679e-05}
{"train_loss": 0.20681384205818176, "global_step": 47844, "epoch": 537, "lr": 9.265357611491303e-05}
{"train_loss": 0.29017114639282227, "global_step": 47845, "epoch": 537, "lr": 9.265327360896503e-05}
{"train_loss": 0.26339051127433777, "global_step": 47846, "epoch": 537, "lr": 9.265297109728282e-05}
{"train_loss": 0.2713181972503662, "global_step": 47847, "epoch": 537, "lr": 9.265266857986643e-05}
{"train_loss": 0.32782289385795593, "global_step": 47848, "epoch": 537, "lr": 9.265236605671592e-05}
{"train_loss": 0.18596534430980682, "global_step": 47849, "epoch": 537, "lr": 9.265206352783132e-05}
{"train_loss": 0.2417529970407486, "global_step": 47850, "epoch": 537, "lr": 9.265176099321265e-05}
{"train_loss": 0.32442066073417664, "global_step": 47851, "epoch": 537, "lr": 9.265145845285998e-05}
{"train_loss": 0.30378660559654236, "global_step": 47852, "epoch": 537, "lr": 9.265115590677334e-05}
{"train_loss": 0.21201196312904358, "global_step": 47853, "epoch": 537, "lr": 9.265085335495277e-05}
{"train_loss": 0.2728896141052246, "global_step": 47854, "epoch": 537, "lr": 9.265055079739832e-05}
{"train_loss": 0.34313738346099854, "global_step": 47855, "epoch": 537, "lr": 9.265024823411001e-05}
{"train_loss": 0.2946834862232208, "global_step": 47856, "epoch": 537, "lr": 9.264994566508792e-05}
{"train_loss": 0.2974230647087097, "global_step": 47857, "epoch": 537, "lr": 9.264964309033204e-05}
{"train_loss": 0.27780649065971375, "global_step": 47858, "epoch": 537, "lr": 9.264934050984244e-05}
{"train_loss": 0.278509259223938, "global_step": 47859, "epoch": 537, "lr": 9.264903792361916e-05}
{"train_loss": 0.25855758786201477, "global_step": 47860, "epoch": 537, "lr": 9.264873533166224e-05}
{"train_loss": 0.2836986482143402, "global_step": 47861, "epoch": 537, "lr": 9.264843273397171e-05}
{"train_loss": 0.2782320976257324, "global_step": 47862, "epoch": 537, "lr": 9.264813013054763e-05}
{"train_loss": 0.3042457401752472, "global_step": 47863, "epoch": 537, "lr": 9.264782752139002e-05}
{"train_loss": 0.3078919053077698, "global_step": 47864, "epoch": 537, "lr": 9.264752490649893e-05}
{"train_loss": 0.35573455691337585, "global_step": 47865, "epoch": 537, "lr": 9.26472222858744e-05}
{"train_loss": 0.31262683868408203, "global_step": 47866, "epoch": 537, "lr": 9.264691965951646e-05}
{"train_loss": 0.21156254410743713, "global_step": 47867, "epoch": 537, "lr": 9.264661702742517e-05}
{"train_loss": 0.27992507815361023, "global_step": 47868, "epoch": 537, "lr": 9.264631438960057e-05}
{"train_loss": 0.3339117765426636, "global_step": 47869, "epoch": 537, "lr": 9.264601174604268e-05}
{"train_loss": 0.27613505721092224, "global_step": 47870, "epoch": 537, "lr": 9.264570909675156e-05}
{"train_loss": 0.3674701750278473, "global_step": 47871, "epoch": 537, "lr": 9.264540644172726e-05}
{"train_loss": 0.2534157335758209, "global_step": 47872, "epoch": 537, "lr": 9.264510378096977e-05}
{"train_loss": 0.3295639157295227, "global_step": 47873, "epoch": 537, "lr": 9.26448011144792e-05}
{"train_loss": 0.2948142886161804, "global_step": 47874, "epoch": 537, "lr": 9.264449844225553e-05}
{"train_loss": 0.22779105603694916, "global_step": 47875, "epoch": 537, "lr": 9.264419576429884e-05}
{"train_loss": 0.2960745692253113, "global_step": 47876, "epoch": 537, "lr": 9.264389308060917e-05}
{"train_loss": 0.24799472093582153, "global_step": 47877, "epoch": 537, "lr": 9.264359039118653e-05}
{"train_loss": 0.2999747395515442, "global_step": 47878, "epoch": 537, "lr": 9.2643287696031e-05}
{"train_loss": 0.2641722857952118, "global_step": 47879, "epoch": 537, "lr": 9.264298499514256e-05}
{"train_loss": 0.26732099056243896, "global_step": 47880, "epoch": 537, "lr": 9.264268228852132e-05}
{"train_loss": 0.2847157214464766, "global_step": 47881, "epoch": 537, "lr": 9.264237957616728e-05, "val_loss": 3.0816941261291504}
{"train_loss": 0.29524096846580505, "global_step": 47882, "epoch": 538, "lr": 9.264207685808051e-05}
{"train_loss": 0.21656061708927155, "global_step": 47883, "epoch": 538, "lr": 9.2641774134261e-05}
{"train_loss": 0.3186660706996918, "global_step": 47884, "epoch": 538, "lr": 9.264147140470884e-05}
{"train_loss": 0.24702687561511993, "global_step": 47885, "epoch": 538, "lr": 9.264116866942407e-05}
{"train_loss": 0.26040127873420715, "global_step": 47886, "epoch": 538, "lr": 9.264086592840669e-05}
{"train_loss": 0.38053715229034424, "global_step": 47887, "epoch": 538, "lr": 9.264056318165678e-05}
{"train_loss": 0.24265405535697937, "global_step": 47888, "epoch": 538, "lr": 9.264026042917437e-05}
{"train_loss": 0.3002752363681793, "global_step": 47889, "epoch": 538, "lr": 9.263995767095948e-05}
{"train_loss": 0.2917724847793579, "global_step": 47890, "epoch": 538, "lr": 9.263965490701218e-05}
{"train_loss": 0.30203157663345337, "global_step": 47891, "epoch": 538, "lr": 9.26393521373325e-05}
{"train_loss": 0.32099318504333496, "global_step": 47892, "epoch": 538, "lr": 9.263904936192047e-05}
{"train_loss": 0.24453750252723694, "global_step": 47893, "epoch": 538, "lr": 9.263874658077614e-05}
{"train_loss": 0.23450875282287598, "global_step": 47894, "epoch": 538, "lr": 9.263844379389954e-05}
{"train_loss": 0.3093230724334717, "global_step": 47895, "epoch": 538, "lr": 9.263814100129074e-05}
{"train_loss": 0.4981915354728699, "global_step": 47896, "epoch": 538, "lr": 9.263783820294976e-05}
{"train_loss": 0.3761007487773895, "global_step": 47897, "epoch": 538, "lr": 9.263753539887664e-05}
{"train_loss": 0.3803032338619232, "global_step": 47898, "epoch": 538, "lr": 9.263723258907142e-05}
{"train_loss": 0.2886860966682434, "global_step": 47899, "epoch": 538, "lr": 9.263692977353414e-05}
{"train_loss": 0.2812751829624176, "global_step": 47900, "epoch": 538, "lr": 9.263662695226484e-05}
{"train_loss": 0.3314114212989807, "global_step": 47901, "epoch": 538, "lr": 9.263632412526359e-05}
{"train_loss": 0.35540369153022766, "global_step": 47902, "epoch": 538, "lr": 9.263602129253039e-05}
{"train_loss": 0.22509066760540009, "global_step": 47903, "epoch": 538, "lr": 9.26357184540653e-05}
{"train_loss": 0.25564152002334595, "global_step": 47904, "epoch": 538, "lr": 9.263541560986835e-05}
{"train_loss": 0.318899929523468, "global_step": 47905, "epoch": 538, "lr": 9.263511275993959e-05}
{"train_loss": 0.22390379011631012, "global_step": 47906, "epoch": 538, "lr": 9.263480990427907e-05}
{"train_loss": 0.23639510571956635, "global_step": 47907, "epoch": 538, "lr": 9.26345070428868e-05}
{"train_loss": 0.27952200174331665, "global_step": 47908, "epoch": 538, "lr": 9.263420417576285e-05}
{"train_loss": 0.23233263194561005, "global_step": 47909, "epoch": 538, "lr": 9.263390130290726e-05}
{"train_loss": 0.25453823804855347, "global_step": 47910, "epoch": 538, "lr": 9.263359842432005e-05}
{"train_loss": 0.28653526306152344, "global_step": 47911, "epoch": 538, "lr": 9.263329554000126e-05}
{"train_loss": 0.3789544999599457, "global_step": 47912, "epoch": 538, "lr": 9.263299264995096e-05}
{"train_loss": 0.2903973460197449, "global_step": 47913, "epoch": 538, "lr": 9.263268975416916e-05}
{"train_loss": 0.25389769673347473, "global_step": 47914, "epoch": 538, "lr": 9.263238685265594e-05}
{"train_loss": 0.2586263418197632, "global_step": 47915, "epoch": 538, "lr": 9.26320839454113e-05}
{"train_loss": 0.20009854435920715, "global_step": 47916, "epoch": 538, "lr": 9.26317810324353e-05}
{"train_loss": 0.408374160528183, "global_step": 47917, "epoch": 538, "lr": 9.263147811372797e-05}
{"train_loss": 0.23879975080490112, "global_step": 47918, "epoch": 538, "lr": 9.263117518928936e-05}
{"train_loss": 0.38540276885032654, "global_step": 47919, "epoch": 538, "lr": 9.263087225911952e-05}
{"train_loss": 0.36180543899536133, "global_step": 47920, "epoch": 538, "lr": 9.263056932321846e-05}
{"train_loss": 0.3444342017173767, "global_step": 47921, "epoch": 538, "lr": 9.263026638158625e-05}
{"train_loss": 0.2836844325065613, "global_step": 47922, "epoch": 538, "lr": 9.262996343422293e-05}
{"train_loss": 0.1827615648508072, "global_step": 47923, "epoch": 538, "lr": 9.262966048112851e-05}
{"train_loss": 0.25484734773635864, "global_step": 47924, "epoch": 538, "lr": 9.262935752230306e-05}
{"train_loss": 0.24779976904392242, "global_step": 47925, "epoch": 538, "lr": 9.262905455774661e-05}
{"train_loss": 0.2963055968284607, "global_step": 47926, "epoch": 538, "lr": 9.26287515874592e-05}
{"train_loss": 0.29435062408447266, "global_step": 47927, "epoch": 538, "lr": 9.262844861144089e-05}
{"train_loss": 0.26444190740585327, "global_step": 47928, "epoch": 538, "lr": 9.26281456296917e-05}
{"train_loss": 0.22144059836864471, "global_step": 47929, "epoch": 538, "lr": 9.262784264221167e-05}
{"train_loss": 0.35806456208229065, "global_step": 47930, "epoch": 538, "lr": 9.262753964900084e-05}
{"train_loss": 0.2648531496524811, "global_step": 47931, "epoch": 538, "lr": 9.262723665005926e-05}
{"train_loss": 0.24520686268806458, "global_step": 47932, "epoch": 538, "lr": 9.262693364538697e-05}
{"train_loss": 0.3004080057144165, "global_step": 47933, "epoch": 538, "lr": 9.262663063498402e-05}
{"train_loss": 0.2691074013710022, "global_step": 47934, "epoch": 538, "lr": 9.262632761885041e-05}
{"train_loss": 0.27525702118873596, "global_step": 47935, "epoch": 538, "lr": 9.262602459698624e-05}
{"train_loss": 0.342617392539978, "global_step": 47936, "epoch": 538, "lr": 9.26257215693915e-05}
{"train_loss": 0.30804720520973206, "global_step": 47937, "epoch": 538, "lr": 9.262541853606628e-05}
{"train_loss": 0.2287314236164093, "global_step": 47938, "epoch": 538, "lr": 9.262511549701056e-05}
{"train_loss": 0.3352220058441162, "global_step": 47939, "epoch": 538, "lr": 9.262481245222443e-05}
{"train_loss": 0.24783886969089508, "global_step": 47940, "epoch": 538, "lr": 9.262450940170791e-05}
{"train_loss": 0.27530404925346375, "global_step": 47941, "epoch": 538, "lr": 9.262420634546104e-05}
{"train_loss": 0.2063315510749817, "global_step": 47942, "epoch": 538, "lr": 9.262390328348387e-05}
{"train_loss": 0.25917279720306396, "global_step": 47943, "epoch": 538, "lr": 9.262360021577642e-05}
{"train_loss": 0.2541954815387726, "global_step": 47944, "epoch": 538, "lr": 9.262329714233877e-05}
{"train_loss": 0.2311856597661972, "global_step": 47945, "epoch": 538, "lr": 9.262299406317093e-05}
{"train_loss": 0.24880295991897583, "global_step": 47946, "epoch": 538, "lr": 9.262269097827294e-05}
{"train_loss": 0.26303648948669434, "global_step": 47947, "epoch": 538, "lr": 9.262238788764485e-05}
{"train_loss": 0.3072148561477661, "global_step": 47948, "epoch": 538, "lr": 9.262208479128671e-05}
{"train_loss": 0.3261668384075165, "global_step": 47949, "epoch": 538, "lr": 9.262178168919854e-05}
{"train_loss": 0.24205484986305237, "global_step": 47950, "epoch": 538, "lr": 9.26214785813804e-05}
{"train_loss": 0.21419993042945862, "global_step": 47951, "epoch": 538, "lr": 9.262117546783231e-05}
{"train_loss": 0.32498547434806824, "global_step": 47952, "epoch": 538, "lr": 9.262087234855432e-05}
{"train_loss": 0.20030610263347626, "global_step": 47953, "epoch": 538, "lr": 9.262056922354649e-05}
{"train_loss": 0.3296978771686554, "global_step": 47954, "epoch": 538, "lr": 9.262026609280883e-05}
{"train_loss": 0.3127140402793884, "global_step": 47955, "epoch": 538, "lr": 9.26199629563414e-05}
{"train_loss": 0.26556360721588135, "global_step": 47956, "epoch": 538, "lr": 9.261965981414424e-05}
{"train_loss": 0.2531942129135132, "global_step": 47957, "epoch": 538, "lr": 9.261935666621737e-05}
{"train_loss": 0.38494229316711426, "global_step": 47958, "epoch": 538, "lr": 9.261905351256087e-05}
{"train_loss": 0.2502170503139496, "global_step": 47959, "epoch": 538, "lr": 9.261875035317475e-05}
{"train_loss": 0.30642110109329224, "global_step": 47960, "epoch": 538, "lr": 9.261844718805906e-05}
{"train_loss": 0.25047194957733154, "global_step": 47961, "epoch": 538, "lr": 9.261814401721383e-05}
{"train_loss": 0.2359301894903183, "global_step": 47962, "epoch": 538, "lr": 9.261784084063911e-05}
{"train_loss": 0.30417299270629883, "global_step": 47963, "epoch": 538, "lr": 9.261753765833494e-05}
{"train_loss": 0.23816336691379547, "global_step": 47964, "epoch": 538, "lr": 9.261723447030137e-05}
{"train_loss": 0.2754177153110504, "global_step": 47965, "epoch": 538, "lr": 9.261693127653844e-05}
{"train_loss": 0.2658783197402954, "global_step": 47966, "epoch": 538, "lr": 9.261662807704616e-05}
{"train_loss": 0.3666898310184479, "global_step": 47967, "epoch": 538, "lr": 9.261632487182461e-05}
{"train_loss": 0.2794402539730072, "global_step": 47968, "epoch": 538, "lr": 9.26160216608738e-05}
{"train_loss": 0.24555394053459167, "global_step": 47969, "epoch": 538, "lr": 9.26157184441938e-05}
{"train_loss": 0.28687630041261736, "global_step": 47970, "epoch": 538, "lr": 9.261541522178463e-05, "val_loss": 2.9130101203918457}
{"train_loss": 0.23365521430969238, "global_step": 47971, "epoch": 539, "lr": 9.261511199364634e-05}
{"train_loss": 0.28023460507392883, "global_step": 47972, "epoch": 539, "lr": 9.261480875977897e-05}
{"train_loss": 0.2741484045982361, "global_step": 47973, "epoch": 539, "lr": 9.261450552018255e-05}
{"train_loss": 0.4070945680141449, "global_step": 47974, "epoch": 539, "lr": 9.261420227485713e-05}
{"train_loss": 0.5000311136245728, "global_step": 47975, "epoch": 539, "lr": 9.261389902380275e-05}
{"train_loss": 0.42989644408226013, "global_step": 47976, "epoch": 539, "lr": 9.261359576701946e-05}
{"train_loss": 0.3335581421852112, "global_step": 47977, "epoch": 539, "lr": 9.261329250450727e-05}
{"train_loss": 0.3199733793735504, "global_step": 47978, "epoch": 539, "lr": 9.261298923626627e-05}
{"train_loss": 0.24271421134471893, "global_step": 47979, "epoch": 539, "lr": 9.261268596229646e-05}
{"train_loss": 0.27742260694503784, "global_step": 47980, "epoch": 539, "lr": 9.261238268259789e-05}
{"train_loss": 0.30462726950645447, "global_step": 47981, "epoch": 539, "lr": 9.26120793971706e-05}
{"train_loss": 0.2797122597694397, "global_step": 47982, "epoch": 539, "lr": 9.261177610601466e-05}
{"train_loss": 0.33434435725212097, "global_step": 47983, "epoch": 539, "lr": 9.261147280913008e-05}
{"train_loss": 0.35709473490715027, "global_step": 47984, "epoch": 539, "lr": 9.26111695065169e-05}
{"train_loss": 0.3575678765773773, "global_step": 47985, "epoch": 539, "lr": 9.261086619817514e-05}
{"train_loss": 0.2968164086341858, "global_step": 47986, "epoch": 539, "lr": 9.26105628841049e-05}
{"train_loss": 0.28028061985969543, "global_step": 47987, "epoch": 539, "lr": 9.261025956430617e-05}
{"train_loss": 0.2821262776851654, "global_step": 47988, "epoch": 539, "lr": 9.260995623877904e-05}
{"train_loss": 0.2677757441997528, "global_step": 47989, "epoch": 539, "lr": 9.26096529075235e-05}
{"train_loss": 0.23896929621696472, "global_step": 47990, "epoch": 539, "lr": 9.260934957053962e-05}
{"train_loss": 0.39597034454345703, "global_step": 47991, "epoch": 539, "lr": 9.260904622782743e-05}
{"train_loss": 0.3327866196632385, "global_step": 47992, "epoch": 539, "lr": 9.260874287938696e-05}
{"train_loss": 0.3655412197113037, "global_step": 47993, "epoch": 539, "lr": 9.260843952521827e-05}
{"train_loss": 0.28615108132362366, "global_step": 47994, "epoch": 539, "lr": 9.26081361653214e-05}
{"train_loss": 0.2515639364719391, "global_step": 47995, "epoch": 539, "lr": 9.260783279969638e-05}
{"train_loss": 0.2863665223121643, "global_step": 47996, "epoch": 539, "lr": 9.260752942834325e-05}
{"train_loss": 0.21918654441833496, "global_step": 47997, "epoch": 539, "lr": 9.260722605126208e-05}
{"train_loss": 0.29503509402275085, "global_step": 47998, "epoch": 539, "lr": 9.260692266845287e-05}
{"train_loss": 0.2942173480987549, "global_step": 47999, "epoch": 539, "lr": 9.260661927991568e-05}
{"train_loss": 0.24013814330101013, "global_step": 48000, "epoch": 539, "lr": 9.260631588565056e-05}
{"train_loss": 0.3009621500968933, "global_step": 48001, "epoch": 539, "lr": 9.260601248565753e-05}
{"train_loss": 0.28741076588630676, "global_step": 48002, "epoch": 539, "lr": 9.260570907993664e-05}
{"train_loss": 0.44376978278160095, "global_step": 48003, "epoch": 539, "lr": 9.260540566848795e-05}
{"train_loss": 0.42748454213142395, "global_step": 48004, "epoch": 539, "lr": 9.260510225131145e-05}
{"train_loss": 0.2781468629837036, "global_step": 48005, "epoch": 539, "lr": 9.260479882840724e-05}
{"train_loss": 0.35669440031051636, "global_step": 48006, "epoch": 539, "lr": 9.260449539977532e-05}
{"train_loss": 0.26777294278144836, "global_step": 48007, "epoch": 539, "lr": 9.260419196541574e-05}
{"train_loss": 0.27097395062446594, "global_step": 48008, "epoch": 539, "lr": 9.260388852532857e-05}
{"train_loss": 0.3079555630683899, "global_step": 48009, "epoch": 539, "lr": 9.260358507951382e-05}
{"train_loss": 0.2675205171108246, "global_step": 48010, "epoch": 539, "lr": 9.260328162797152e-05}
{"train_loss": 0.36540186405181885, "global_step": 48011, "epoch": 539, "lr": 9.260297817070174e-05}
{"train_loss": 0.3713839650154114, "global_step": 48012, "epoch": 539, "lr": 9.26026747077045e-05}
{"train_loss": 0.3121866285800934, "global_step": 48013, "epoch": 539, "lr": 9.260237123897986e-05}
{"train_loss": 0.20395711064338684, "global_step": 48014, "epoch": 539, "lr": 9.260206776452784e-05}
{"train_loss": 0.17560377717018127, "global_step": 48015, "epoch": 539, "lr": 9.26017642843485e-05}
{"train_loss": 0.21728838980197906, "global_step": 48016, "epoch": 539, "lr": 9.260146079844186e-05}
{"train_loss": 0.22747835516929626, "global_step": 48017, "epoch": 539, "lr": 9.260115730680798e-05}
{"train_loss": 0.29580748081207275, "global_step": 48018, "epoch": 539, "lr": 9.260085380944689e-05}
{"train_loss": 0.3106982707977295, "global_step": 48019, "epoch": 539, "lr": 9.260055030635866e-05}
{"train_loss": 0.2658191919326782, "global_step": 48020, "epoch": 539, "lr": 9.260024679754327e-05}
{"train_loss": 0.3798498213291168, "global_step": 48021, "epoch": 539, "lr": 9.259994328300082e-05}
{"train_loss": 0.22361721098423004, "global_step": 48022, "epoch": 539, "lr": 9.259963976273132e-05}
{"train_loss": 0.3616214692592621, "global_step": 48023, "epoch": 539, "lr": 9.259933623673481e-05}
{"train_loss": 0.28581416606903076, "global_step": 48024, "epoch": 539, "lr": 9.259903270501136e-05}
{"train_loss": 0.29848727583885193, "global_step": 48025, "epoch": 539, "lr": 9.259872916756096e-05}
{"train_loss": 0.27724364399909973, "global_step": 48026, "epoch": 539, "lr": 9.25984256243837e-05}
{"train_loss": 0.3731190860271454, "global_step": 48027, "epoch": 539, "lr": 9.25981220754796e-05}
{"train_loss": 0.3227231502532959, "global_step": 48028, "epoch": 539, "lr": 9.259781852084868e-05}
{"train_loss": 0.48678869009017944, "global_step": 48029, "epoch": 539, "lr": 9.259751496049103e-05}
{"train_loss": 0.4037308394908905, "global_step": 48030, "epoch": 539, "lr": 9.259721139440664e-05}
{"train_loss": 0.3822799623012543, "global_step": 48031, "epoch": 539, "lr": 9.25969078225956e-05}
{"train_loss": 0.3221707344055176, "global_step": 48032, "epoch": 539, "lr": 9.25966042450579e-05}
{"train_loss": 0.20860260725021362, "global_step": 48033, "epoch": 539, "lr": 9.259630066179362e-05}
{"train_loss": 0.19239439070224762, "global_step": 48034, "epoch": 539, "lr": 9.259599707280278e-05}
{"train_loss": 0.2520291209220886, "global_step": 48035, "epoch": 539, "lr": 9.259569347808544e-05}
{"train_loss": 0.19742706418037415, "global_step": 48036, "epoch": 539, "lr": 9.259538987764161e-05}
{"train_loss": 0.2998495399951935, "global_step": 48037, "epoch": 539, "lr": 9.259508627147136e-05}
{"train_loss": 0.3400937020778656, "global_step": 48038, "epoch": 539, "lr": 9.259478265957472e-05}
{"train_loss": 0.3386780619621277, "global_step": 48039, "epoch": 539, "lr": 9.259447904195174e-05}
{"train_loss": 0.2806064784526825, "global_step": 48040, "epoch": 539, "lr": 9.259417541860243e-05}
{"train_loss": 0.2887842357158661, "global_step": 48041, "epoch": 539, "lr": 9.259387178952685e-05}
{"train_loss": 0.23106293380260468, "global_step": 48042, "epoch": 539, "lr": 9.259356815472506e-05}
{"train_loss": 0.2696785628795624, "global_step": 48043, "epoch": 539, "lr": 9.259326451419708e-05}
{"train_loss": 0.3280196487903595, "global_step": 48044, "epoch": 539, "lr": 9.259296086794296e-05}
{"train_loss": 0.27346256375312805, "global_step": 48045, "epoch": 539, "lr": 9.259265721596273e-05}
{"train_loss": 0.23668204247951508, "global_step": 48046, "epoch": 539, "lr": 9.259235355825644e-05}
{"train_loss": 0.36210691928863525, "global_step": 48047, "epoch": 539, "lr": 9.259204989482412e-05}
{"train_loss": 0.2842423915863037, "global_step": 48048, "epoch": 539, "lr": 9.259174622566582e-05}
{"train_loss": 0.2903634011745453, "global_step": 48049, "epoch": 539, "lr": 9.259144255078158e-05}
{"train_loss": 0.2708304822444916, "global_step": 48050, "epoch": 539, "lr": 9.259113887017143e-05}
{"train_loss": 0.2764531373977661, "global_step": 48051, "epoch": 539, "lr": 9.259083518383544e-05}
{"train_loss": 0.25278016924858093, "global_step": 48052, "epoch": 539, "lr": 9.259053149177361e-05}
{"train_loss": 0.2746337354183197, "global_step": 48053, "epoch": 539, "lr": 9.259022779398603e-05}
{"train_loss": 0.2673940360546112, "global_step": 48054, "epoch": 539, "lr": 9.258992409047269e-05}
{"train_loss": 0.38046836853027344, "global_step": 48055, "epoch": 539, "lr": 9.258962038123367e-05}
{"train_loss": 0.38575050234794617, "global_step": 48056, "epoch": 539, "lr": 9.258931666626898e-05}
{"train_loss": 0.2645259499549866, "global_step": 48057, "epoch": 539, "lr": 9.258901294557867e-05}
{"train_loss": 0.253105491399765, "global_step": 48058, "epoch": 539, "lr": 9.258870921916281e-05}
{"train_loss": 0.3032678660382046, "global_step": 48059, "epoch": 539, "lr": 9.258840548702139e-05, "val_loss": 3.0045583248138428}
{"train_loss": 0.20263957977294922, "global_step": 48060, "epoch": 540, "lr": 9.25881017491545e-05}
{"train_loss": 0.18776389956474304, "global_step": 48061, "epoch": 540, "lr": 9.258779800556214e-05}
{"train_loss": 0.3327786326408386, "global_step": 48062, "epoch": 540, "lr": 9.258749425624438e-05}
{"train_loss": 0.269117146730423, "global_step": 48063, "epoch": 540, "lr": 9.258719050120127e-05}
{"train_loss": 0.23947986960411072, "global_step": 48064, "epoch": 540, "lr": 9.25868867404328e-05}
{"train_loss": 0.21228326857089996, "global_step": 48065, "epoch": 540, "lr": 9.258658297393904e-05}
{"train_loss": 0.2505454123020172, "global_step": 48066, "epoch": 540, "lr": 9.258627920172005e-05}
{"train_loss": 0.3289937376976013, "global_step": 48067, "epoch": 540, "lr": 9.258597542377584e-05}
{"train_loss": 0.2313072234392166, "global_step": 48068, "epoch": 540, "lr": 9.25856716401065e-05}
{"train_loss": 0.18729156255722046, "global_step": 48069, "epoch": 540, "lr": 9.258536785071202e-05}
{"train_loss": 0.29850971698760986, "global_step": 48070, "epoch": 540, "lr": 9.258506405559244e-05}
{"train_loss": 0.27408796548843384, "global_step": 48071, "epoch": 540, "lr": 9.258476025474782e-05}
{"train_loss": 0.209473118185997, "global_step": 48072, "epoch": 540, "lr": 9.25844564481782e-05}
{"train_loss": 0.30852842330932617, "global_step": 48073, "epoch": 540, "lr": 9.258415263588362e-05}
{"train_loss": 0.2358495593070984, "global_step": 48074, "epoch": 540, "lr": 9.258384881786413e-05}
{"train_loss": 0.18831229209899902, "global_step": 48075, "epoch": 540, "lr": 9.258354499411974e-05}
{"train_loss": 0.18793494999408722, "global_step": 48076, "epoch": 540, "lr": 9.258324116465052e-05}
{"train_loss": 0.334454208612442, "global_step": 48077, "epoch": 540, "lr": 9.258293732945651e-05}
{"train_loss": 0.314723938703537, "global_step": 48078, "epoch": 540, "lr": 9.258263348853774e-05}
{"train_loss": 0.27499324083328247, "global_step": 48079, "epoch": 540, "lr": 9.258232964189425e-05}
{"train_loss": 0.2432376742362976, "global_step": 48080, "epoch": 540, "lr": 9.258202578952609e-05}
{"train_loss": 0.32462671399116516, "global_step": 48081, "epoch": 540, "lr": 9.25817219314333e-05}
{"train_loss": 0.2505943775177002, "global_step": 48082, "epoch": 540, "lr": 9.258141806761591e-05}
{"train_loss": 0.28770050406455994, "global_step": 48083, "epoch": 540, "lr": 9.258111419807396e-05}
{"train_loss": 0.2508578896522522, "global_step": 48084, "epoch": 540, "lr": 9.25808103228075e-05}
{"train_loss": 0.3051547110080719, "global_step": 48085, "epoch": 540, "lr": 9.258050644181657e-05}
{"train_loss": 0.25077518820762634, "global_step": 48086, "epoch": 540, "lr": 9.258020255510123e-05}
{"train_loss": 0.20845888555049896, "global_step": 48087, "epoch": 540, "lr": 9.257989866266148e-05}
{"train_loss": 0.3451046645641327, "global_step": 48088, "epoch": 540, "lr": 9.257959476449737e-05}
{"train_loss": 0.22133232653141022, "global_step": 48089, "epoch": 540, "lr": 9.257929086060898e-05}
{"train_loss": 0.27450504899024963, "global_step": 48090, "epoch": 540, "lr": 9.257898695099632e-05}
{"train_loss": 0.23610445857048035, "global_step": 48091, "epoch": 540, "lr": 9.257868303565943e-05}
{"train_loss": 0.27279913425445557, "global_step": 48092, "epoch": 540, "lr": 9.257837911459835e-05}
{"train_loss": 0.3532567620277405, "global_step": 48093, "epoch": 540, "lr": 9.257807518781312e-05}
{"train_loss": 0.22606293857097626, "global_step": 48094, "epoch": 540, "lr": 9.257777125530378e-05}
{"train_loss": 0.23718972504138947, "global_step": 48095, "epoch": 540, "lr": 9.25774673170704e-05}
{"train_loss": 0.2631520628929138, "global_step": 48096, "epoch": 540, "lr": 9.2577163373113e-05}
{"train_loss": 0.24328124523162842, "global_step": 48097, "epoch": 540, "lr": 9.25768594234316e-05}
{"train_loss": 0.2871550917625427, "global_step": 48098, "epoch": 540, "lr": 9.257655546802627e-05}
{"train_loss": 0.29595938324928284, "global_step": 48099, "epoch": 540, "lr": 9.257625150689704e-05}
{"train_loss": 0.27377578616142273, "global_step": 48100, "epoch": 540, "lr": 9.257594754004395e-05}
{"train_loss": 0.23629577457904816, "global_step": 48101, "epoch": 540, "lr": 9.257564356746705e-05}
{"train_loss": 0.3442675769329071, "global_step": 48102, "epoch": 540, "lr": 9.257533958916637e-05}
{"train_loss": 0.23370607197284698, "global_step": 48103, "epoch": 540, "lr": 9.257503560514195e-05}
{"train_loss": 0.2459183633327484, "global_step": 48104, "epoch": 540, "lr": 9.257473161539382e-05}
{"train_loss": 0.3537147045135498, "global_step": 48105, "epoch": 540, "lr": 9.257442761992207e-05}
{"train_loss": 0.29190436005592346, "global_step": 48106, "epoch": 540, "lr": 9.257412361872668e-05}
{"train_loss": 0.1948021650314331, "global_step": 48107, "epoch": 540, "lr": 9.257381961180773e-05}
{"train_loss": 0.3125331401824951, "global_step": 48108, "epoch": 540, "lr": 9.257351559916526e-05}
{"train_loss": 0.20829761028289795, "global_step": 48109, "epoch": 540, "lr": 9.257321158079929e-05}
{"train_loss": 0.23238562047481537, "global_step": 48110, "epoch": 540, "lr": 9.257290755670986e-05}
{"train_loss": 0.24938835203647614, "global_step": 48111, "epoch": 540, "lr": 9.257260352689704e-05}
{"train_loss": 0.23559682071208954, "global_step": 48112, "epoch": 540, "lr": 9.257229949136084e-05}
{"train_loss": 0.2990334630012512, "global_step": 48113, "epoch": 540, "lr": 9.257199545010131e-05}
{"train_loss": 0.2260403335094452, "global_step": 48114, "epoch": 540, "lr": 9.257169140311849e-05}
{"train_loss": 0.25361448526382446, "global_step": 48115, "epoch": 540, "lr": 9.257138735041245e-05}
{"train_loss": 0.2800315022468567, "global_step": 48116, "epoch": 540, "lr": 9.257108329198318e-05}
{"train_loss": 0.22650396823883057, "global_step": 48117, "epoch": 540, "lr": 9.257077922783077e-05}
{"train_loss": 0.2101583182811737, "global_step": 48118, "epoch": 540, "lr": 9.257047515795522e-05}
{"train_loss": 0.3490995764732361, "global_step": 48119, "epoch": 540, "lr": 9.257017108235659e-05}
{"train_loss": 0.35863226652145386, "global_step": 48120, "epoch": 540, "lr": 9.256986700103493e-05}
{"train_loss": 0.24153198301792145, "global_step": 48121, "epoch": 540, "lr": 9.256956291399026e-05}
{"train_loss": 0.3292061686515808, "global_step": 48122, "epoch": 540, "lr": 9.256925882122263e-05}
{"train_loss": 0.26341143250465393, "global_step": 48123, "epoch": 540, "lr": 9.25689547227321e-05}
{"train_loss": 0.3715182840824127, "global_step": 48124, "epoch": 540, "lr": 9.256865061851867e-05}
{"train_loss": 0.2719298005104065, "global_step": 48125, "epoch": 540, "lr": 9.256834650858242e-05}
{"train_loss": 0.29493287205696106, "global_step": 48126, "epoch": 540, "lr": 9.256804239292336e-05}
{"train_loss": 0.2957943081855774, "global_step": 48127, "epoch": 540, "lr": 9.256773827154155e-05}
{"train_loss": 0.2716306149959564, "global_step": 48128, "epoch": 540, "lr": 9.256743414443704e-05}
{"train_loss": 0.2347860038280487, "global_step": 48129, "epoch": 540, "lr": 9.256713001160985e-05}
{"train_loss": 0.22222097218036652, "global_step": 48130, "epoch": 540, "lr": 9.256682587306002e-05}
{"train_loss": 0.39646464586257935, "global_step": 48131, "epoch": 540, "lr": 9.256652172878761e-05}
{"train_loss": 0.3073103725910187, "global_step": 48132, "epoch": 540, "lr": 9.256621757879265e-05}
{"train_loss": 0.26963984966278076, "global_step": 48133, "epoch": 540, "lr": 9.256591342307517e-05}
{"train_loss": 0.35129353404045105, "global_step": 48134, "epoch": 540, "lr": 9.256560926163522e-05}
{"train_loss": 0.30752184987068176, "global_step": 48135, "epoch": 540, "lr": 9.256530509447286e-05}
{"train_loss": 0.2983042299747467, "global_step": 48136, "epoch": 540, "lr": 9.25650009215881e-05}
{"train_loss": 0.21154528856277466, "global_step": 48137, "epoch": 540, "lr": 9.2564696742981e-05}
{"train_loss": 0.3254568874835968, "global_step": 48138, "epoch": 540, "lr": 9.256439255865159e-05}
{"train_loss": 0.3670062720775604, "global_step": 48139, "epoch": 540, "lr": 9.256408836859992e-05}
{"train_loss": 0.243296280503273, "global_step": 48140, "epoch": 540, "lr": 9.256378417282601e-05}
{"train_loss": 0.2635458707809448, "global_step": 48141, "epoch": 540, "lr": 9.256347997132995e-05}
{"train_loss": 0.2932952642440796, "global_step": 48142, "epoch": 540, "lr": 9.256317576411174e-05}
{"train_loss": 0.2447720617055893, "global_step": 48143, "epoch": 540, "lr": 9.25628715511714e-05}
{"train_loss": 0.23834702372550964, "global_step": 48144, "epoch": 540, "lr": 9.256256733250903e-05}
{"train_loss": 0.2618717551231384, "global_step": 48145, "epoch": 540, "lr": 9.256226310812464e-05}
{"train_loss": 0.3493673503398895, "global_step": 48146, "epoch": 540, "lr": 9.256195887801826e-05}
{"train_loss": 0.38674482703208923, "global_step": 48147, "epoch": 540, "lr": 9.256165464218995e-05}
{"train_loss": 0.27156566987546643, "global_step": 48148, "epoch": 540, "lr": 9.256135040063976e-05, "val_loss": 2.9773218631744385, "train_action_mse_error": 22.01711654663086}
{"train_loss": 0.25145918130874634, "global_step": 48149, "epoch": 541, "lr": 9.256104615336769e-05}
{"train_loss": 0.2595369815826416, "global_step": 48150, "epoch": 541, "lr": 9.256074190037381e-05}
{"train_loss": 0.35187244415283203, "global_step": 48151, "epoch": 541, "lr": 9.256043764165816e-05}
{"train_loss": 0.36500510573387146, "global_step": 48152, "epoch": 541, "lr": 9.256013337722079e-05}
{"train_loss": 0.20681458711624146, "global_step": 48153, "epoch": 541, "lr": 9.255982910706173e-05}
{"train_loss": 0.3454371988773346, "global_step": 48154, "epoch": 541, "lr": 9.2559524831181e-05}
{"train_loss": 0.24971303343772888, "global_step": 48155, "epoch": 541, "lr": 9.255922054957867e-05}
{"train_loss": 0.34600120782852173, "global_step": 48156, "epoch": 541, "lr": 9.255891626225477e-05}
{"train_loss": 0.39459994435310364, "global_step": 48157, "epoch": 541, "lr": 9.255861196920935e-05}
{"train_loss": 0.26159584522247314, "global_step": 48158, "epoch": 541, "lr": 9.255830767044243e-05}
{"train_loss": 0.23711852729320526, "global_step": 48159, "epoch": 541, "lr": 9.255800336595409e-05}
{"train_loss": 0.2702289819717407, "global_step": 48160, "epoch": 541, "lr": 9.255769905574432e-05}
{"train_loss": 0.3950689136981964, "global_step": 48161, "epoch": 541, "lr": 9.25573947398132e-05}
{"train_loss": 0.30718907713890076, "global_step": 48162, "epoch": 541, "lr": 9.255709041816074e-05}
{"train_loss": 0.2684136927127838, "global_step": 48163, "epoch": 541, "lr": 9.2556786090787e-05}
{"train_loss": 0.3594726622104645, "global_step": 48164, "epoch": 541, "lr": 9.255648175769202e-05}
{"train_loss": 0.28690364956855774, "global_step": 48165, "epoch": 541, "lr": 9.255617741887585e-05}
{"train_loss": 0.2819582223892212, "global_step": 48166, "epoch": 541, "lr": 9.255587307433853e-05}
{"train_loss": 0.3362812101840973, "global_step": 48167, "epoch": 541, "lr": 9.255556872408008e-05}
{"train_loss": 0.3934642970561981, "global_step": 48168, "epoch": 541, "lr": 9.255526436810054e-05}
{"train_loss": 0.31116318702697754, "global_step": 48169, "epoch": 541, "lr": 9.255496000639999e-05}
{"train_loss": 0.3931327760219574, "global_step": 48170, "epoch": 541, "lr": 9.255465563897843e-05}
{"train_loss": 0.2699379324913025, "global_step": 48171, "epoch": 541, "lr": 9.25543512658359e-05}
{"train_loss": 0.2746815085411072, "global_step": 48172, "epoch": 541, "lr": 9.255404688697246e-05}
{"train_loss": 0.23901623487472534, "global_step": 48173, "epoch": 541, "lr": 9.255374250238818e-05}
{"train_loss": 0.2829281687736511, "global_step": 48174, "epoch": 541, "lr": 9.255343811208304e-05}
{"train_loss": 0.23580731451511383, "global_step": 48175, "epoch": 541, "lr": 9.255313371605711e-05}
{"train_loss": 0.29824039340019226, "global_step": 48176, "epoch": 541, "lr": 9.255282931431044e-05}
{"train_loss": 0.28983792662620544, "global_step": 48177, "epoch": 541, "lr": 9.255252490684305e-05}
{"train_loss": 0.30706873536109924, "global_step": 48178, "epoch": 541, "lr": 9.2552220493655e-05}
{"train_loss": 0.29499301314353943, "global_step": 48179, "epoch": 541, "lr": 9.255191607474631e-05}
{"train_loss": 0.24399608373641968, "global_step": 48180, "epoch": 541, "lr": 9.255161165011706e-05}
{"train_loss": 0.18752744793891907, "global_step": 48181, "epoch": 541, "lr": 9.255130721976723e-05}
{"train_loss": 0.1806185096502304, "global_step": 48182, "epoch": 541, "lr": 9.255100278369691e-05}
{"train_loss": 0.28731778264045715, "global_step": 48183, "epoch": 541, "lr": 9.255069834190614e-05}
{"train_loss": 0.3550797998905182, "global_step": 48184, "epoch": 541, "lr": 9.255039389439492e-05}
{"train_loss": 0.35348644852638245, "global_step": 48185, "epoch": 541, "lr": 9.255008944116335e-05}
{"train_loss": 0.3337325155735016, "global_step": 48186, "epoch": 541, "lr": 9.254978498221142e-05}
{"train_loss": 0.3690257966518402, "global_step": 48187, "epoch": 541, "lr": 9.25494805175392e-05}
{"train_loss": 0.356408953666687, "global_step": 48188, "epoch": 541, "lr": 9.254917604714671e-05}
{"train_loss": 0.3228799104690552, "global_step": 48189, "epoch": 541, "lr": 9.2548871571034e-05}
{"train_loss": 0.3274676203727722, "global_step": 48190, "epoch": 541, "lr": 9.254856708920113e-05}
{"train_loss": 0.3561404347419739, "global_step": 48191, "epoch": 541, "lr": 9.25482626016481e-05}
{"train_loss": 0.24644450843334198, "global_step": 48192, "epoch": 541, "lr": 9.254795810837499e-05}
{"train_loss": 0.26114609837532043, "global_step": 48193, "epoch": 541, "lr": 9.254765360938182e-05}
{"train_loss": 0.27643442153930664, "global_step": 48194, "epoch": 541, "lr": 9.254734910466864e-05}
{"train_loss": 0.28348031640052795, "global_step": 48195, "epoch": 541, "lr": 9.254704459423548e-05}
{"train_loss": 0.29386425018310547, "global_step": 48196, "epoch": 541, "lr": 9.25467400780824e-05}
{"train_loss": 0.3944065272808075, "global_step": 48197, "epoch": 541, "lr": 9.254643555620943e-05}
{"train_loss": 0.28547394275665283, "global_step": 48198, "epoch": 541, "lr": 9.25461310286166e-05}
{"train_loss": 0.28355446457862854, "global_step": 48199, "epoch": 541, "lr": 9.254582649530396e-05}
{"train_loss": 0.22487576305866241, "global_step": 48200, "epoch": 541, "lr": 9.254552195627157e-05}
{"train_loss": 0.24211406707763672, "global_step": 48201, "epoch": 541, "lr": 9.254521741151944e-05}
{"train_loss": 0.28927236795425415, "global_step": 48202, "epoch": 541, "lr": 9.254491286104763e-05}
{"train_loss": 0.2729763388633728, "global_step": 48203, "epoch": 541, "lr": 9.254460830485617e-05}
{"train_loss": 0.31541019678115845, "global_step": 48204, "epoch": 541, "lr": 9.25443037429451e-05}
{"train_loss": 0.32815662026405334, "global_step": 48205, "epoch": 541, "lr": 9.254399917531448e-05}
{"train_loss": 0.336265504360199, "global_step": 48206, "epoch": 541, "lr": 9.254369460196432e-05}
{"train_loss": 0.22548265755176544, "global_step": 48207, "epoch": 541, "lr": 9.254339002289469e-05}
{"train_loss": 0.22207801043987274, "global_step": 48208, "epoch": 541, "lr": 9.254308543810563e-05}
{"train_loss": 0.2555183470249176, "global_step": 48209, "epoch": 541, "lr": 9.254278084759715e-05}
{"train_loss": 0.35263538360595703, "global_step": 48210, "epoch": 541, "lr": 9.254247625136934e-05}
{"train_loss": 0.3025110065937042, "global_step": 48211, "epoch": 541, "lr": 9.254217164942218e-05}
{"train_loss": 0.22586116194725037, "global_step": 48212, "epoch": 541, "lr": 9.254186704175576e-05}
{"train_loss": 0.24620811641216278, "global_step": 48213, "epoch": 541, "lr": 9.25415624283701e-05}
{"train_loss": 0.3434854745864868, "global_step": 48214, "epoch": 541, "lr": 9.254125780926526e-05}
{"train_loss": 0.2804288864135742, "global_step": 48215, "epoch": 541, "lr": 9.254095318444125e-05}
{"train_loss": 0.22641943395137787, "global_step": 48216, "epoch": 541, "lr": 9.254064855389813e-05}
{"train_loss": 0.20580096542835236, "global_step": 48217, "epoch": 541, "lr": 9.254034391763595e-05}
{"train_loss": 0.35128119587898254, "global_step": 48218, "epoch": 541, "lr": 9.254003927565473e-05}
{"train_loss": 0.32679370045661926, "global_step": 48219, "epoch": 541, "lr": 9.253973462795451e-05}
{"train_loss": 0.2695344090461731, "global_step": 48220, "epoch": 541, "lr": 9.253942997453535e-05}
{"train_loss": 0.29112061858177185, "global_step": 48221, "epoch": 541, "lr": 9.253912531539729e-05}
{"train_loss": 0.2613180875778198, "global_step": 48222, "epoch": 541, "lr": 9.253882065054036e-05}
{"train_loss": 0.34617048501968384, "global_step": 48223, "epoch": 541, "lr": 9.253851597996459e-05}
{"train_loss": 0.2642691731452942, "global_step": 48224, "epoch": 541, "lr": 9.253821130367005e-05}
{"train_loss": 0.22725528478622437, "global_step": 48225, "epoch": 541, "lr": 9.253790662165676e-05}
{"train_loss": 0.17386671900749207, "global_step": 48226, "epoch": 541, "lr": 9.253760193392478e-05}
{"train_loss": 0.3427569568157196, "global_step": 48227, "epoch": 541, "lr": 9.253729724047411e-05}
{"train_loss": 0.3369169235229492, "global_step": 48228, "epoch": 541, "lr": 9.253699254130484e-05}
{"train_loss": 0.3775344491004944, "global_step": 48229, "epoch": 541, "lr": 9.253668783641698e-05}
{"train_loss": 0.2852399945259094, "global_step": 48230, "epoch": 541, "lr": 9.253638312581059e-05}
{"train_loss": 0.43070995807647705, "global_step": 48231, "epoch": 541, "lr": 9.25360784094857e-05}
{"train_loss": 0.2136463224887848, "global_step": 48232, "epoch": 541, "lr": 9.253577368744235e-05}
{"train_loss": 0.2669927775859833, "global_step": 48233, "epoch": 541, "lr": 9.253546895968059e-05}
{"train_loss": 0.26137882471084595, "global_step": 48234, "epoch": 541, "lr": 9.253516422620044e-05}
{"train_loss": 0.27498432993888855, "global_step": 48235, "epoch": 541, "lr": 9.253485948700196e-05}
{"train_loss": 0.31660908460617065, "global_step": 48236, "epoch": 541, "lr": 9.253455474208519e-05}
{"train_loss": 0.2936673874265692, "global_step": 48237, "epoch": 541, "lr": 9.253424999145017e-05, "val_loss": 3.077147960662842}
{"train_loss": 0.39333444833755493, "global_step": 48238, "epoch": 542, "lr": 9.253394523509693e-05}
{"train_loss": 0.2542805075645447, "global_step": 48239, "epoch": 542, "lr": 9.253364047302553e-05}
{"train_loss": 0.3287879526615143, "global_step": 48240, "epoch": 542, "lr": 9.2533335705236e-05}
{"train_loss": 0.23391805589199066, "global_step": 48241, "epoch": 542, "lr": 9.253303093172839e-05}
{"train_loss": 0.2380383014678955, "global_step": 48242, "epoch": 542, "lr": 9.25327261525027e-05}
{"train_loss": 0.3367442488670349, "global_step": 48243, "epoch": 542, "lr": 9.253242136755902e-05}
{"train_loss": 0.3129393458366394, "global_step": 48244, "epoch": 542, "lr": 9.253211657689739e-05}
{"train_loss": 0.38325077295303345, "global_step": 48245, "epoch": 542, "lr": 9.253181178051782e-05}
{"train_loss": 0.2953427731990814, "global_step": 48246, "epoch": 542, "lr": 9.253150697842037e-05}
{"train_loss": 0.2844395935535431, "global_step": 48247, "epoch": 542, "lr": 9.253120217060508e-05}
{"train_loss": 0.2811283469200134, "global_step": 48248, "epoch": 542, "lr": 9.253089735707197e-05}
{"train_loss": 0.2475966513156891, "global_step": 48249, "epoch": 542, "lr": 9.253059253782112e-05}
{"train_loss": 0.3393225371837616, "global_step": 48250, "epoch": 542, "lr": 9.253028771285255e-05}
{"train_loss": 0.20108477771282196, "global_step": 48251, "epoch": 542, "lr": 9.25299828821663e-05}
{"train_loss": 0.2751808166503906, "global_step": 48252, "epoch": 542, "lr": 9.252967804576241e-05}
{"train_loss": 0.1877911537885666, "global_step": 48253, "epoch": 542, "lr": 9.252937320364092e-05}
{"train_loss": 0.2119656205177307, "global_step": 48254, "epoch": 542, "lr": 9.252906835580188e-05}
{"train_loss": 0.21020080149173737, "global_step": 48255, "epoch": 542, "lr": 9.25287635022453e-05}
{"train_loss": 0.2807021737098694, "global_step": 48256, "epoch": 542, "lr": 9.252845864297128e-05}
{"train_loss": 0.2246665209531784, "global_step": 48257, "epoch": 542, "lr": 9.25281537779798e-05}
{"train_loss": 0.33093559741973877, "global_step": 48258, "epoch": 542, "lr": 9.252784890727095e-05}
{"train_loss": 0.24716055393218994, "global_step": 48259, "epoch": 542, "lr": 9.252754403084474e-05}
{"train_loss": 0.26403552293777466, "global_step": 48260, "epoch": 542, "lr": 9.252723914870123e-05}
{"train_loss": 0.17423510551452637, "global_step": 48261, "epoch": 542, "lr": 9.252693426084043e-05}
{"train_loss": 0.2860363721847534, "global_step": 48262, "epoch": 542, "lr": 9.252662936726242e-05}
{"train_loss": 0.24181461334228516, "global_step": 48263, "epoch": 542, "lr": 9.252632446796721e-05}
{"train_loss": 0.2210903912782669, "global_step": 48264, "epoch": 542, "lr": 9.252601956295485e-05}
{"train_loss": 0.2735382914543152, "global_step": 48265, "epoch": 542, "lr": 9.252571465222542e-05}
{"train_loss": 0.33237704634666443, "global_step": 48266, "epoch": 542, "lr": 9.252540973577889e-05}
{"train_loss": 0.2296302318572998, "global_step": 48267, "epoch": 542, "lr": 9.252510481361535e-05}
{"train_loss": 0.3130810558795929, "global_step": 48268, "epoch": 542, "lr": 9.252479988573482e-05}
{"train_loss": 0.30322977900505066, "global_step": 48269, "epoch": 542, "lr": 9.252449495213736e-05}
{"train_loss": 0.31934499740600586, "global_step": 48270, "epoch": 542, "lr": 9.252419001282299e-05}
{"train_loss": 0.35467422008514404, "global_step": 48271, "epoch": 542, "lr": 9.252388506779177e-05}
{"train_loss": 0.372253954410553, "global_step": 48272, "epoch": 542, "lr": 9.252358011704374e-05}
{"train_loss": 0.2954801321029663, "global_step": 48273, "epoch": 542, "lr": 9.252327516057891e-05}
{"train_loss": 0.27916470170021057, "global_step": 48274, "epoch": 542, "lr": 9.252297019839735e-05}
{"train_loss": 0.2934277057647705, "global_step": 48275, "epoch": 542, "lr": 9.252266523049911e-05}
{"train_loss": 0.26442569494247437, "global_step": 48276, "epoch": 542, "lr": 9.25223602568842e-05}
{"train_loss": 0.28283217549324036, "global_step": 48277, "epoch": 542, "lr": 9.252205527755268e-05}
{"train_loss": 0.3120749294757843, "global_step": 48278, "epoch": 542, "lr": 9.252175029250459e-05}
{"train_loss": 0.29523977637290955, "global_step": 48279, "epoch": 542, "lr": 9.252144530173998e-05}
{"train_loss": 0.30179324746131897, "global_step": 48280, "epoch": 542, "lr": 9.252114030525887e-05}
{"train_loss": 0.23659715056419373, "global_step": 48281, "epoch": 542, "lr": 9.25208353030613e-05}
{"train_loss": 0.18745291233062744, "global_step": 48282, "epoch": 542, "lr": 9.252053029514733e-05}
{"train_loss": 0.30673834681510925, "global_step": 48283, "epoch": 542, "lr": 9.252022528151701e-05}
{"train_loss": 0.2807292938232422, "global_step": 48284, "epoch": 542, "lr": 9.251992026217034e-05}
{"train_loss": 0.21574029326438904, "global_step": 48285, "epoch": 542, "lr": 9.25196152371074e-05}
{"train_loss": 0.30735427141189575, "global_step": 48286, "epoch": 542, "lr": 9.251931020632821e-05}
{"train_loss": 0.22336821258068085, "global_step": 48287, "epoch": 542, "lr": 9.251900516983282e-05}
{"train_loss": 0.29601556062698364, "global_step": 48288, "epoch": 542, "lr": 9.251870012762127e-05}
{"train_loss": 0.2550666630268097, "global_step": 48289, "epoch": 542, "lr": 9.25183950796936e-05}
{"train_loss": 0.28326845169067383, "global_step": 48290, "epoch": 542, "lr": 9.251809002604983e-05}
{"train_loss": 0.3751766085624695, "global_step": 48291, "epoch": 542, "lr": 9.251778496669003e-05}
{"train_loss": 0.22777050733566284, "global_step": 48292, "epoch": 542, "lr": 9.251747990161425e-05}
{"train_loss": 0.21997341513633728, "global_step": 48293, "epoch": 542, "lr": 9.251717483082249e-05}
{"train_loss": 0.3162378966808319, "global_step": 48294, "epoch": 542, "lr": 9.251686975431483e-05}
{"train_loss": 0.21784912049770355, "global_step": 48295, "epoch": 542, "lr": 9.251656467209129e-05}
{"train_loss": 0.2612653970718384, "global_step": 48296, "epoch": 542, "lr": 9.251625958415191e-05}
{"train_loss": 0.3313228487968445, "global_step": 48297, "epoch": 542, "lr": 9.251595449049675e-05}
{"train_loss": 0.251583456993103, "global_step": 48298, "epoch": 542, "lr": 9.251564939112583e-05}
{"train_loss": 0.2752201557159424, "global_step": 48299, "epoch": 542, "lr": 9.25153442860392e-05}
{"train_loss": 0.2474403977394104, "global_step": 48300, "epoch": 542, "lr": 9.25150391752369e-05}
{"train_loss": 0.3507470488548279, "global_step": 48301, "epoch": 542, "lr": 9.251473405871898e-05}
{"train_loss": 0.31887972354888916, "global_step": 48302, "epoch": 542, "lr": 9.251442893648545e-05}
{"train_loss": 0.36840859055519104, "global_step": 48303, "epoch": 542, "lr": 9.25141238085364e-05}
{"train_loss": 0.3444637954235077, "global_step": 48304, "epoch": 542, "lr": 9.251381867487182e-05}
{"train_loss": 0.22155754268169403, "global_step": 48305, "epoch": 542, "lr": 9.251351353549179e-05}
{"train_loss": 0.281599760055542, "global_step": 48306, "epoch": 542, "lr": 9.251320839039633e-05}
{"train_loss": 0.3714657425880432, "global_step": 48307, "epoch": 542, "lr": 9.251290323958549e-05}
{"train_loss": 0.2575201988220215, "global_step": 48308, "epoch": 542, "lr": 9.251259808305933e-05}
{"train_loss": 0.2897248864173889, "global_step": 48309, "epoch": 542, "lr": 9.251229292081783e-05}
{"train_loss": 0.26668208837509155, "global_step": 48310, "epoch": 542, "lr": 9.25119877528611e-05}
{"train_loss": 0.24324961006641388, "global_step": 48311, "epoch": 542, "lr": 9.251168257918912e-05}
{"train_loss": 0.31887227296829224, "global_step": 48312, "epoch": 542, "lr": 9.251137739980198e-05}
{"train_loss": 0.32290324568748474, "global_step": 48313, "epoch": 542, "lr": 9.251107221469971e-05}
{"train_loss": 0.30383726954460144, "global_step": 48314, "epoch": 542, "lr": 9.251076702388235e-05}
{"train_loss": 0.3054671585559845, "global_step": 48315, "epoch": 542, "lr": 9.251046182734993e-05}
{"train_loss": 0.3022093176841736, "global_step": 48316, "epoch": 542, "lr": 9.251015662510248e-05}
{"train_loss": 0.28902193903923035, "global_step": 48317, "epoch": 542, "lr": 9.250985141714007e-05}
{"train_loss": 0.21263349056243896, "global_step": 48318, "epoch": 542, "lr": 9.250954620346273e-05}
{"train_loss": 0.19351407885551453, "global_step": 48319, "epoch": 542, "lr": 9.25092409840705e-05}
{"train_loss": 0.23700262606143951, "global_step": 48320, "epoch": 542, "lr": 9.250893575896342e-05}
{"train_loss": 0.29941147565841675, "global_step": 48321, "epoch": 542, "lr": 9.250863052814152e-05}
{"train_loss": 0.3038010001182556, "global_step": 48322, "epoch": 542, "lr": 9.250832529160487e-05}
{"train_loss": 0.27259084582328796, "global_step": 48323, "epoch": 542, "lr": 9.250802004935348e-05}
{"train_loss": 0.3234482407569885, "global_step": 48324, "epoch": 542, "lr": 9.250771480138742e-05}
{"train_loss": 0.19130323827266693, "global_step": 48325, "epoch": 542, "lr": 9.25074095477067e-05}
{"train_loss": 0.2785946604576004, "global_step": 48326, "epoch": 542, "lr": 9.250710428831139e-05, "val_loss": 2.9289181232452393}
{"train_loss": 0.2932952046394348, "global_step": 48327, "epoch": 543, "lr": 9.250679902320151e-05}
{"train_loss": 0.2675515413284302, "global_step": 48328, "epoch": 543, "lr": 9.25064937523771e-05}
{"train_loss": 0.25345975160598755, "global_step": 48329, "epoch": 543, "lr": 9.250618847583823e-05}
{"train_loss": 0.27559560537338257, "global_step": 48330, "epoch": 543, "lr": 9.250588319358491e-05}
{"train_loss": 0.20049342513084412, "global_step": 48331, "epoch": 543, "lr": 9.25055779056172e-05}
{"train_loss": 0.24889975786209106, "global_step": 48332, "epoch": 543, "lr": 9.250527261193511e-05}
{"train_loss": 0.40308332443237305, "global_step": 48333, "epoch": 543, "lr": 9.250496731253872e-05}
{"train_loss": 0.26843616366386414, "global_step": 48334, "epoch": 543, "lr": 9.250466200742806e-05}
{"train_loss": 0.31313827633857727, "global_step": 48335, "epoch": 543, "lr": 9.250435669660318e-05}
{"train_loss": 0.30768364667892456, "global_step": 48336, "epoch": 543, "lr": 9.250405138006408e-05}
{"train_loss": 0.27030321955680847, "global_step": 48337, "epoch": 543, "lr": 9.250374605781084e-05}
{"train_loss": 0.27033793926239014, "global_step": 48338, "epoch": 543, "lr": 9.25034407298435e-05}
{"train_loss": 0.4494844675064087, "global_step": 48339, "epoch": 543, "lr": 9.250313539616206e-05}
{"train_loss": 0.3940119445323944, "global_step": 48340, "epoch": 543, "lr": 9.250283005676661e-05}
{"train_loss": 0.22460466623306274, "global_step": 48341, "epoch": 543, "lr": 9.250252471165718e-05}
{"train_loss": 0.2989281117916107, "global_step": 48342, "epoch": 543, "lr": 9.250221936083378e-05}
{"train_loss": 0.28606799244880676, "global_step": 48343, "epoch": 543, "lr": 9.250191400429651e-05}
{"train_loss": 0.278582900762558, "global_step": 48344, "epoch": 543, "lr": 9.250160864204536e-05}
{"train_loss": 0.34257543087005615, "global_step": 48345, "epoch": 543, "lr": 9.250130327408037e-05}
{"train_loss": 0.3246995210647583, "global_step": 48346, "epoch": 543, "lr": 9.25009979004016e-05}
{"train_loss": 0.4058627784252167, "global_step": 48347, "epoch": 543, "lr": 9.250069252100911e-05}
{"train_loss": 0.33261600136756897, "global_step": 48348, "epoch": 543, "lr": 9.250038713590291e-05}
{"train_loss": 0.19962960481643677, "global_step": 48349, "epoch": 543, "lr": 9.250008174508305e-05}
{"train_loss": 0.36019080877304077, "global_step": 48350, "epoch": 543, "lr": 9.249977634854957e-05}
{"train_loss": 0.4012615382671356, "global_step": 48351, "epoch": 543, "lr": 9.249947094630251e-05}
{"train_loss": 0.24994823336601257, "global_step": 48352, "epoch": 543, "lr": 9.249916553834193e-05}
{"train_loss": 0.3761567175388336, "global_step": 48353, "epoch": 543, "lr": 9.249886012466784e-05}
{"train_loss": 0.27614855766296387, "global_step": 48354, "epoch": 543, "lr": 9.24985547052803e-05}
{"train_loss": 0.25288718938827515, "global_step": 48355, "epoch": 543, "lr": 9.249824928017937e-05}
{"train_loss": 0.23339328169822693, "global_step": 48356, "epoch": 543, "lr": 9.249794384936503e-05}
{"train_loss": 0.18116243183612823, "global_step": 48357, "epoch": 543, "lr": 9.249763841283737e-05}
{"train_loss": 0.2498675435781479, "global_step": 48358, "epoch": 543, "lr": 9.249733297059644e-05}
{"train_loss": 0.157516211271286, "global_step": 48359, "epoch": 543, "lr": 9.249702752264224e-05}
{"train_loss": 0.28228679299354553, "global_step": 48360, "epoch": 543, "lr": 9.249672206897483e-05}
{"train_loss": 0.23755736649036407, "global_step": 48361, "epoch": 543, "lr": 9.249641660959427e-05}
{"train_loss": 0.25087201595306396, "global_step": 48362, "epoch": 543, "lr": 9.249611114450059e-05}
{"train_loss": 0.3230817914009094, "global_step": 48363, "epoch": 543, "lr": 9.24958056736938e-05}
{"train_loss": 0.4040527045726776, "global_step": 48364, "epoch": 543, "lr": 9.249550019717399e-05}
{"train_loss": 0.3379215896129608, "global_step": 48365, "epoch": 543, "lr": 9.249519471494117e-05}
{"train_loss": 0.22339481115341187, "global_step": 48366, "epoch": 543, "lr": 9.249488922699538e-05}
{"train_loss": 0.29080575704574585, "global_step": 48367, "epoch": 543, "lr": 9.249458373333668e-05}
{"train_loss": 0.2908308804035187, "global_step": 48368, "epoch": 543, "lr": 9.249427823396511e-05}
{"train_loss": 0.2914813458919525, "global_step": 48369, "epoch": 543, "lr": 9.249397272888069e-05}
{"train_loss": 0.24736693501472473, "global_step": 48370, "epoch": 543, "lr": 9.249366721808347e-05}
{"train_loss": 0.2680589258670807, "global_step": 48371, "epoch": 543, "lr": 9.24933617015735e-05}
{"train_loss": 0.400973916053772, "global_step": 48372, "epoch": 543, "lr": 9.249305617935083e-05}
{"train_loss": 0.2932305335998535, "global_step": 48373, "epoch": 543, "lr": 9.249275065141547e-05}
{"train_loss": 0.3432720899581909, "global_step": 48374, "epoch": 543, "lr": 9.249244511776747e-05}
{"train_loss": 0.21575871109962463, "global_step": 48375, "epoch": 543, "lr": 9.249213957840689e-05}
{"train_loss": 0.3267069458961487, "global_step": 48376, "epoch": 543, "lr": 9.249183403333376e-05}
{"train_loss": 0.27430668473243713, "global_step": 48377, "epoch": 543, "lr": 9.249152848254812e-05}
{"train_loss": 0.4152170717716217, "global_step": 48378, "epoch": 543, "lr": 9.249122292605001e-05}
{"train_loss": 0.43208545446395874, "global_step": 48379, "epoch": 543, "lr": 9.249091736383948e-05}
{"train_loss": 0.30975833535194397, "global_step": 48380, "epoch": 543, "lr": 9.249061179591654e-05}
{"train_loss": 0.250847190618515, "global_step": 48381, "epoch": 543, "lr": 9.249030622228127e-05}
{"train_loss": 0.41548585891723633, "global_step": 48382, "epoch": 543, "lr": 9.249000064293372e-05}
{"train_loss": 0.299990177154541, "global_step": 48383, "epoch": 543, "lr": 9.248969505787389e-05}
{"train_loss": 0.2598876357078552, "global_step": 48384, "epoch": 543, "lr": 9.248938946710181e-05}
{"train_loss": 0.3003154993057251, "global_step": 48385, "epoch": 543, "lr": 9.248908387061759e-05}
{"train_loss": 0.28016233444213867, "global_step": 48386, "epoch": 543, "lr": 9.248877826842121e-05}
{"train_loss": 0.31770941615104675, "global_step": 48387, "epoch": 543, "lr": 9.248847266051274e-05}
{"train_loss": 0.4354446828365326, "global_step": 48388, "epoch": 543, "lr": 9.24881670468922e-05}
{"train_loss": 0.29467108845710754, "global_step": 48389, "epoch": 543, "lr": 9.248786142755965e-05}
{"train_loss": 0.25613734126091003, "global_step": 48390, "epoch": 543, "lr": 9.248755580251513e-05}
{"train_loss": 0.3200850188732147, "global_step": 48391, "epoch": 543, "lr": 9.248725017175867e-05}
{"train_loss": 0.3416840732097626, "global_step": 48392, "epoch": 543, "lr": 9.248694453529034e-05}
{"train_loss": 0.23453596234321594, "global_step": 48393, "epoch": 543, "lr": 9.248663889311014e-05}
{"train_loss": 0.22579818964004517, "global_step": 48394, "epoch": 543, "lr": 9.248633324521812e-05}
{"train_loss": 0.20314699411392212, "global_step": 48395, "epoch": 543, "lr": 9.248602759161433e-05}
{"train_loss": 0.26415038108825684, "global_step": 48396, "epoch": 543, "lr": 9.248572193229883e-05}
{"train_loss": 0.314870685338974, "global_step": 48397, "epoch": 543, "lr": 9.248541626727164e-05}
{"train_loss": 0.34280872344970703, "global_step": 48398, "epoch": 543, "lr": 9.248511059653279e-05}
{"train_loss": 0.2955007553100586, "global_step": 48399, "epoch": 543, "lr": 9.248480492008233e-05}
{"train_loss": 0.2887280285358429, "global_step": 48400, "epoch": 543, "lr": 9.248449923792032e-05}
{"train_loss": 0.2749372124671936, "global_step": 48401, "epoch": 543, "lr": 9.248419355004679e-05}
{"train_loss": 0.36201468110084534, "global_step": 48402, "epoch": 543, "lr": 9.248388785646176e-05}
{"train_loss": 0.28637173771858215, "global_step": 48403, "epoch": 543, "lr": 9.24835821571653e-05}
{"train_loss": 0.24945974349975586, "global_step": 48404, "epoch": 543, "lr": 9.248327645215745e-05}
{"train_loss": 0.3508918583393097, "global_step": 48405, "epoch": 543, "lr": 9.248297074143823e-05}
{"train_loss": 0.19120851159095764, "global_step": 48406, "epoch": 543, "lr": 9.248266502500769e-05}
{"train_loss": 0.3394288122653961, "global_step": 48407, "epoch": 543, "lr": 9.248235930286589e-05}
{"train_loss": 0.35651177167892456, "global_step": 48408, "epoch": 543, "lr": 9.248205357501283e-05}
{"train_loss": 0.2381506860256195, "global_step": 48409, "epoch": 543, "lr": 9.248174784144858e-05}
{"train_loss": 0.22892054915428162, "global_step": 48410, "epoch": 543, "lr": 9.24814421021732e-05}
{"train_loss": 0.2612002193927765, "global_step": 48411, "epoch": 543, "lr": 9.248113635718669e-05}
{"train_loss": 0.2105623334646225, "global_step": 48412, "epoch": 543, "lr": 9.248083060648911e-05}
{"train_loss": 0.2069869339466095, "global_step": 48413, "epoch": 543, "lr": 9.248052485008049e-05}
{"train_loss": 0.29759344458580017, "global_step": 48414, "epoch": 543, "lr": 9.24802190879609e-05}
{"train_loss": 0.29478427365924537, "global_step": 48415, "epoch": 543, "lr": 9.247991332013036e-05, "val_loss": 2.9769022464752197}
{"train_loss": 0.22481992840766907, "global_step": 48416, "epoch": 544, "lr": 9.24796075465889e-05}
{"train_loss": 0.2829032838344574, "global_step": 48417, "epoch": 544, "lr": 9.247930176733659e-05}
{"train_loss": 0.3393242061138153, "global_step": 48418, "epoch": 544, "lr": 9.247899598237346e-05}
{"train_loss": 0.26174408197402954, "global_step": 48419, "epoch": 544, "lr": 9.247869019169952e-05}
{"train_loss": 0.34785711765289307, "global_step": 48420, "epoch": 544, "lr": 9.247838439531486e-05}
{"train_loss": 0.23582623898983002, "global_step": 48421, "epoch": 544, "lr": 9.247807859321949e-05}
{"train_loss": 0.31580764055252075, "global_step": 48422, "epoch": 544, "lr": 9.247777278541346e-05}
{"train_loss": 0.19672134518623352, "global_step": 48423, "epoch": 544, "lr": 9.247746697189682e-05}
{"train_loss": 0.2763209342956543, "global_step": 48424, "epoch": 544, "lr": 9.24771611526696e-05}
{"train_loss": 0.33350297808647156, "global_step": 48425, "epoch": 544, "lr": 9.247685532773183e-05}
{"train_loss": 0.1588958501815796, "global_step": 48426, "epoch": 544, "lr": 9.247654949708358e-05}
{"train_loss": 0.2996361553668976, "global_step": 48427, "epoch": 544, "lr": 9.247624366072487e-05}
{"train_loss": 0.25440698862075806, "global_step": 48428, "epoch": 544, "lr": 9.247593781865574e-05}
{"train_loss": 0.2862570881843567, "global_step": 48429, "epoch": 544, "lr": 9.247563197087624e-05}
{"train_loss": 0.2724534571170807, "global_step": 48430, "epoch": 544, "lr": 9.247532611738642e-05}
{"train_loss": 0.3256168067455292, "global_step": 48431, "epoch": 544, "lr": 9.247502025818631e-05}
{"train_loss": 0.29863405227661133, "global_step": 48432, "epoch": 544, "lr": 9.247471439327594e-05}
{"train_loss": 0.3566078245639801, "global_step": 48433, "epoch": 544, "lr": 9.247440852265537e-05}
{"train_loss": 0.22030195593833923, "global_step": 48434, "epoch": 544, "lr": 9.247410264632462e-05}
{"train_loss": 0.18806593120098114, "global_step": 48435, "epoch": 544, "lr": 9.247379676428376e-05}
{"train_loss": 0.29540547728538513, "global_step": 48436, "epoch": 544, "lr": 9.247349087653282e-05}
{"train_loss": 0.3104954957962036, "global_step": 48437, "epoch": 544, "lr": 9.247318498307181e-05}
{"train_loss": 0.20943276584148407, "global_step": 48438, "epoch": 544, "lr": 9.247287908390083e-05}
{"train_loss": 0.26878660917282104, "global_step": 48439, "epoch": 544, "lr": 9.247257317901986e-05}
{"train_loss": 0.41717612743377686, "global_step": 48440, "epoch": 544, "lr": 9.247226726842899e-05}
{"train_loss": 0.311697781085968, "global_step": 48441, "epoch": 544, "lr": 9.247196135212824e-05}
{"train_loss": 0.3128114342689514, "global_step": 48442, "epoch": 544, "lr": 9.247165543011764e-05}
{"train_loss": 0.32927316427230835, "global_step": 48443, "epoch": 544, "lr": 9.247134950239727e-05}
{"train_loss": 0.3027935326099396, "global_step": 48444, "epoch": 544, "lr": 9.247104356896712e-05}
{"train_loss": 0.20468005537986755, "global_step": 48445, "epoch": 544, "lr": 9.247073762982726e-05}
{"train_loss": 0.22752076387405396, "global_step": 48446, "epoch": 544, "lr": 9.247043168497774e-05}
{"train_loss": 0.3628372848033905, "global_step": 48447, "epoch": 544, "lr": 9.247012573441855e-05}
{"train_loss": 0.2846902906894684, "global_step": 48448, "epoch": 544, "lr": 9.246981977814981e-05}
{"train_loss": 0.2370544672012329, "global_step": 48449, "epoch": 544, "lr": 9.246951381617151e-05}
{"train_loss": 0.2636550962924957, "global_step": 48450, "epoch": 544, "lr": 9.24692078484837e-05}
{"train_loss": 0.29544004797935486, "global_step": 48451, "epoch": 544, "lr": 9.246890187508641e-05}
{"train_loss": 0.24076469242572784, "global_step": 48452, "epoch": 544, "lr": 9.246859589597972e-05}
{"train_loss": 0.22007158398628235, "global_step": 48453, "epoch": 544, "lr": 9.246828991116363e-05}
{"train_loss": 0.3506295382976532, "global_step": 48454, "epoch": 544, "lr": 9.24679839206382e-05}
{"train_loss": 0.361406534910202, "global_step": 48455, "epoch": 544, "lr": 9.246767792440346e-05}
{"train_loss": 0.2785806655883789, "global_step": 48456, "epoch": 544, "lr": 9.246737192245944e-05}
{"train_loss": 0.26545238494873047, "global_step": 48457, "epoch": 544, "lr": 9.246706591480623e-05}
{"train_loss": 0.23395994305610657, "global_step": 48458, "epoch": 544, "lr": 9.246675990144383e-05}
{"train_loss": 0.3038892149925232, "global_step": 48459, "epoch": 544, "lr": 9.24664538823723e-05}
{"train_loss": 0.20056310296058655, "global_step": 48460, "epoch": 544, "lr": 9.246614785759167e-05}
{"train_loss": 0.1912176012992859, "global_step": 48461, "epoch": 544, "lr": 9.246584182710197e-05}
{"train_loss": 0.3644343316555023, "global_step": 48462, "epoch": 544, "lr": 9.246553579090327e-05}
{"train_loss": 0.24830913543701172, "global_step": 48463, "epoch": 544, "lr": 9.246522974899559e-05}
{"train_loss": 0.2201206535100937, "global_step": 48464, "epoch": 544, "lr": 9.246492370137898e-05}
{"train_loss": 0.35011136531829834, "global_step": 48465, "epoch": 544, "lr": 9.246461764805347e-05}
{"train_loss": 0.2003970742225647, "global_step": 48466, "epoch": 544, "lr": 9.246431158901913e-05}
{"train_loss": 0.23805928230285645, "global_step": 48467, "epoch": 544, "lr": 9.246400552427597e-05}
{"train_loss": 0.3456658124923706, "global_step": 48468, "epoch": 544, "lr": 9.246369945382403e-05}
{"train_loss": 0.25987738370895386, "global_step": 48469, "epoch": 544, "lr": 9.246339337766338e-05}
{"train_loss": 0.29065918922424316, "global_step": 48470, "epoch": 544, "lr": 9.246308729579403e-05}
{"train_loss": 0.33478835225105286, "global_step": 48471, "epoch": 544, "lr": 9.246278120821606e-05}
{"train_loss": 0.24962398409843445, "global_step": 48472, "epoch": 544, "lr": 9.246247511492946e-05}
{"train_loss": 0.31663641333580017, "global_step": 48473, "epoch": 544, "lr": 9.24621690159343e-05}
{"train_loss": 0.23346510529518127, "global_step": 48474, "epoch": 544, "lr": 9.246186291123064e-05}
{"train_loss": 0.30063295364379883, "global_step": 48475, "epoch": 544, "lr": 9.246155680081847e-05}
{"train_loss": 0.2544458508491516, "global_step": 48476, "epoch": 544, "lr": 9.246125068469789e-05}
{"train_loss": 0.3062160909175873, "global_step": 48477, "epoch": 544, "lr": 9.246094456286889e-05}
{"train_loss": 0.35804054141044617, "global_step": 48478, "epoch": 544, "lr": 9.246063843533154e-05}
{"train_loss": 0.28740131855010986, "global_step": 48479, "epoch": 544, "lr": 9.246033230208588e-05}
{"train_loss": 0.2889784872531891, "global_step": 48480, "epoch": 544, "lr": 9.246002616313196e-05}
{"train_loss": 0.2200736403465271, "global_step": 48481, "epoch": 544, "lr": 9.245972001846979e-05}
{"train_loss": 0.293766051530838, "global_step": 48482, "epoch": 544, "lr": 9.245941386809942e-05}
{"train_loss": 0.25625813007354736, "global_step": 48483, "epoch": 544, "lr": 9.24591077120209e-05}
{"train_loss": 0.33408769965171814, "global_step": 48484, "epoch": 544, "lr": 9.245880155023428e-05}
{"train_loss": 0.35044869780540466, "global_step": 48485, "epoch": 544, "lr": 9.24584953827396e-05}
{"train_loss": 0.3376833200454712, "global_step": 48486, "epoch": 544, "lr": 9.245818920953689e-05}
{"train_loss": 0.280178964138031, "global_step": 48487, "epoch": 544, "lr": 9.245788303062619e-05}
{"train_loss": 0.261452853679657, "global_step": 48488, "epoch": 544, "lr": 9.245757684600754e-05}
{"train_loss": 0.29054784774780273, "global_step": 48489, "epoch": 544, "lr": 9.245727065568099e-05}
{"train_loss": 0.39238879084587097, "global_step": 48490, "epoch": 544, "lr": 9.245696445964658e-05}
{"train_loss": 0.3063601851463318, "global_step": 48491, "epoch": 544, "lr": 9.245665825790434e-05}
{"train_loss": 0.3413158357143402, "global_step": 48492, "epoch": 544, "lr": 9.245635205045432e-05}
{"train_loss": 0.2677404284477234, "global_step": 48493, "epoch": 544, "lr": 9.245604583729656e-05}
{"train_loss": 0.2783106863498688, "global_step": 48494, "epoch": 544, "lr": 9.24557396184311e-05}
{"train_loss": 0.40510252118110657, "global_step": 48495, "epoch": 544, "lr": 9.2455433393858e-05}
{"train_loss": 0.34827956557273865, "global_step": 48496, "epoch": 544, "lr": 9.245512716357727e-05}
{"train_loss": 0.26035022735595703, "global_step": 48497, "epoch": 544, "lr": 9.245482092758897e-05}
{"train_loss": 0.22557465732097626, "global_step": 48498, "epoch": 544, "lr": 9.245451468589314e-05}
{"train_loss": 0.255885511636734, "global_step": 48499, "epoch": 544, "lr": 9.245420843848982e-05}
{"train_loss": 0.2315705269575119, "global_step": 48500, "epoch": 544, "lr": 9.245390218537905e-05}
{"train_loss": 0.31898823380470276, "global_step": 48501, "epoch": 544, "lr": 9.245359592656087e-05}
{"train_loss": 0.24517522752285004, "global_step": 48502, "epoch": 544, "lr": 9.24532896620353e-05}
{"train_loss": 0.31758806109428406, "global_step": 48503, "epoch": 544, "lr": 9.245298339180243e-05}
{"train_loss": 0.2835348755790946, "global_step": 48504, "epoch": 544, "lr": 9.245267711586227e-05, "val_loss": 3.121528387069702}
{"train_loss": 0.38234078884124756, "global_step": 48505, "epoch": 545, "lr": 9.245237083421485e-05}
{"train_loss": 0.2976875305175781, "global_step": 48506, "epoch": 545, "lr": 9.245206454686024e-05}
{"train_loss": 0.23332926630973816, "global_step": 48507, "epoch": 545, "lr": 9.245175825379845e-05}
{"train_loss": 0.26137977838516235, "global_step": 48508, "epoch": 545, "lr": 9.245145195502955e-05}
{"train_loss": 0.32032930850982666, "global_step": 48509, "epoch": 545, "lr": 9.245114565055358e-05}
{"train_loss": 0.24607330560684204, "global_step": 48510, "epoch": 545, "lr": 9.245083934037056e-05}
{"train_loss": 0.34790539741516113, "global_step": 48511, "epoch": 545, "lr": 9.245053302448055e-05}
{"train_loss": 0.34480592608451843, "global_step": 48512, "epoch": 545, "lr": 9.245022670288357e-05}
{"train_loss": 0.31574922800064087, "global_step": 48513, "epoch": 545, "lr": 9.244992037557967e-05}
{"train_loss": 0.2652589976787567, "global_step": 48514, "epoch": 545, "lr": 9.244961404256892e-05}
{"train_loss": 0.28581491112709045, "global_step": 48515, "epoch": 545, "lr": 9.244930770385131e-05}
{"train_loss": 0.38502925634384155, "global_step": 48516, "epoch": 545, "lr": 9.244900135942693e-05}
{"train_loss": 0.3208865225315094, "global_step": 48517, "epoch": 545, "lr": 9.244869500929578e-05}
{"train_loss": 0.2719253897666931, "global_step": 48518, "epoch": 545, "lr": 9.244838865345793e-05}
{"train_loss": 0.24979180097579956, "global_step": 48519, "epoch": 545, "lr": 9.244808229191342e-05}
{"train_loss": 0.3004997670650482, "global_step": 48520, "epoch": 545, "lr": 9.244777592466227e-05}
{"train_loss": 0.2328072041273117, "global_step": 48521, "epoch": 545, "lr": 9.244746955170453e-05}
{"train_loss": 0.31054070591926575, "global_step": 48522, "epoch": 545, "lr": 9.244716317304027e-05}
{"train_loss": 0.30002665519714355, "global_step": 48523, "epoch": 545, "lr": 9.244685678866949e-05}
{"train_loss": 0.273061603307724, "global_step": 48524, "epoch": 545, "lr": 9.244655039859225e-05}
{"train_loss": 0.2837066948413849, "global_step": 48525, "epoch": 545, "lr": 9.244624400280857e-05}
{"train_loss": 0.3750326931476593, "global_step": 48526, "epoch": 545, "lr": 9.244593760131854e-05}
{"train_loss": 0.2857199013233185, "global_step": 48527, "epoch": 545, "lr": 9.244563119412216e-05}
{"train_loss": 0.3243897259235382, "global_step": 48528, "epoch": 545, "lr": 9.244532478121948e-05}
{"train_loss": 0.2911047339439392, "global_step": 48529, "epoch": 545, "lr": 9.244501836261053e-05}
{"train_loss": 0.2794138789176941, "global_step": 48530, "epoch": 545, "lr": 9.244471193829538e-05}
{"train_loss": 0.31060314178466797, "global_step": 48531, "epoch": 545, "lr": 9.244440550827405e-05}
{"train_loss": 0.3612521290779114, "global_step": 48532, "epoch": 545, "lr": 9.244409907254659e-05}
{"train_loss": 0.2001020908355713, "global_step": 48533, "epoch": 545, "lr": 9.244379263111303e-05}
{"train_loss": 0.2921999990940094, "global_step": 48534, "epoch": 545, "lr": 9.244348618397342e-05}
{"train_loss": 0.25483545660972595, "global_step": 48535, "epoch": 545, "lr": 9.244317973112782e-05}
{"train_loss": 0.3002675175666809, "global_step": 48536, "epoch": 545, "lr": 9.244287327257623e-05}
{"train_loss": 0.27498775720596313, "global_step": 48537, "epoch": 545, "lr": 9.244256680831873e-05}
{"train_loss": 0.31675785779953003, "global_step": 48538, "epoch": 545, "lr": 9.244226033835535e-05}
{"train_loss": 0.3417353928089142, "global_step": 48539, "epoch": 545, "lr": 9.244195386268611e-05}
{"train_loss": 0.3061293065547943, "global_step": 48540, "epoch": 545, "lr": 9.244164738131107e-05}
{"train_loss": 0.2013881653547287, "global_step": 48541, "epoch": 545, "lr": 9.244134089423026e-05}
{"train_loss": 0.23517411947250366, "global_step": 48542, "epoch": 545, "lr": 9.244103440144373e-05}
{"train_loss": 0.2042647749185562, "global_step": 48543, "epoch": 545, "lr": 9.244072790295153e-05}
{"train_loss": 0.18840079009532928, "global_step": 48544, "epoch": 545, "lr": 9.244042139875369e-05}
{"train_loss": 0.23630139231681824, "global_step": 48545, "epoch": 545, "lr": 9.244011488885025e-05}
{"train_loss": 0.26368916034698486, "global_step": 48546, "epoch": 545, "lr": 9.243980837324126e-05}
{"train_loss": 0.24920782446861267, "global_step": 48547, "epoch": 545, "lr": 9.243950185192674e-05}
{"train_loss": 0.25883448123931885, "global_step": 48548, "epoch": 545, "lr": 9.243919532490676e-05}
{"train_loss": 0.23842068016529083, "global_step": 48549, "epoch": 545, "lr": 9.243888879218134e-05}
{"train_loss": 0.27217015624046326, "global_step": 48550, "epoch": 545, "lr": 9.243858225375053e-05}
{"train_loss": 0.15694443881511688, "global_step": 48551, "epoch": 545, "lr": 9.243827570961436e-05}
{"train_loss": 0.4007687568664551, "global_step": 48552, "epoch": 545, "lr": 9.24379691597729e-05}
{"train_loss": 0.2188435196876526, "global_step": 48553, "epoch": 545, "lr": 9.243766260422618e-05}
{"train_loss": 0.2770138084888458, "global_step": 48554, "epoch": 545, "lr": 9.24373560429742e-05}
{"train_loss": 0.3618655204772949, "global_step": 48555, "epoch": 545, "lr": 9.243704947601706e-05}
{"train_loss": 0.22636069357395172, "global_step": 48556, "epoch": 545, "lr": 9.243674290335477e-05}
{"train_loss": 0.21955107152462006, "global_step": 48557, "epoch": 545, "lr": 9.243643632498738e-05}
{"train_loss": 0.21986070275306702, "global_step": 48558, "epoch": 545, "lr": 9.243612974091492e-05}
{"train_loss": 0.20770597457885742, "global_step": 48559, "epoch": 545, "lr": 9.243582315113744e-05}
{"train_loss": 0.34143680334091187, "global_step": 48560, "epoch": 545, "lr": 9.243551655565499e-05}
{"train_loss": 0.3009495139122009, "global_step": 48561, "epoch": 545, "lr": 9.24352099544676e-05}
{"train_loss": 0.18867312371730804, "global_step": 48562, "epoch": 545, "lr": 9.24349033475753e-05}
{"train_loss": 0.27880486845970154, "global_step": 48563, "epoch": 545, "lr": 9.243459673497817e-05}
{"train_loss": 0.27226418256759644, "global_step": 48564, "epoch": 545, "lr": 9.24342901166762e-05}
{"train_loss": 0.3274728059768677, "global_step": 48565, "epoch": 545, "lr": 9.243398349266947e-05}
{"train_loss": 0.26830485463142395, "global_step": 48566, "epoch": 545, "lr": 9.243367686295799e-05}
{"train_loss": 0.20313231647014618, "global_step": 48567, "epoch": 545, "lr": 9.243337022754183e-05}
{"train_loss": 0.19395974278450012, "global_step": 48568, "epoch": 545, "lr": 9.243306358642103e-05}
{"train_loss": 0.3244587779045105, "global_step": 48569, "epoch": 545, "lr": 9.243275693959562e-05}
{"train_loss": 0.19518935680389404, "global_step": 48570, "epoch": 545, "lr": 9.243245028706564e-05}
{"train_loss": 0.23851975798606873, "global_step": 48571, "epoch": 545, "lr": 9.243214362883113e-05}
{"train_loss": 0.2410963773727417, "global_step": 48572, "epoch": 545, "lr": 9.243183696489215e-05}
{"train_loss": 0.2316499501466751, "global_step": 48573, "epoch": 545, "lr": 9.243153029524871e-05}
{"train_loss": 0.3059893250465393, "global_step": 48574, "epoch": 545, "lr": 9.243122361990087e-05}
{"train_loss": 0.20003262162208557, "global_step": 48575, "epoch": 545, "lr": 9.243091693884868e-05}
{"train_loss": 0.200308158993721, "global_step": 48576, "epoch": 545, "lr": 9.243061025209217e-05}
{"train_loss": 0.34616371989250183, "global_step": 48577, "epoch": 545, "lr": 9.243030355963136e-05}
{"train_loss": 0.3221381604671478, "global_step": 48578, "epoch": 545, "lr": 9.242999686146634e-05}
{"train_loss": 0.2957350015640259, "global_step": 48579, "epoch": 545, "lr": 9.24296901575971e-05}
{"train_loss": 0.2815517783164978, "global_step": 48580, "epoch": 545, "lr": 9.242938344802372e-05}
{"train_loss": 0.28129759430885315, "global_step": 48581, "epoch": 545, "lr": 9.242907673274623e-05}
{"train_loss": 0.2682006359100342, "global_step": 48582, "epoch": 545, "lr": 9.242877001176467e-05}
{"train_loss": 0.2806316316127777, "global_step": 48583, "epoch": 545, "lr": 9.242846328507906e-05}
{"train_loss": 0.26751184463500977, "global_step": 48584, "epoch": 545, "lr": 9.242815655268948e-05}
{"train_loss": 0.21123656630516052, "global_step": 48585, "epoch": 545, "lr": 9.242784981459593e-05}
{"train_loss": 0.35295581817626953, "global_step": 48586, "epoch": 545, "lr": 9.242754307079848e-05}
{"train_loss": 0.22633381187915802, "global_step": 48587, "epoch": 545, "lr": 9.242723632129717e-05}
{"train_loss": 0.36705997586250305, "global_step": 48588, "epoch": 545, "lr": 9.242692956609204e-05}
{"train_loss": 0.30164387822151184, "global_step": 48589, "epoch": 545, "lr": 9.242662280518313e-05}
{"train_loss": 0.34082460403442383, "global_step": 48590, "epoch": 545, "lr": 9.242631603857046e-05}
{"train_loss": 0.2966613173484802, "global_step": 48591, "epoch": 545, "lr": 9.24260092662541e-05}
{"train_loss": 0.2399618774652481, "global_step": 48592, "epoch": 545, "lr": 9.242570248823408e-05}
{"train_loss": 0.2771996407026655, "global_step": 48593, "epoch": 545, "lr": 9.242539570451043e-05, "val_loss": 3.0633656978607178, "train_action_mse_error": 27.04095458984375}
{"train_loss": 0.3135310113430023, "global_step": 48594, "epoch": 546, "lr": 9.242508891508322e-05}
{"train_loss": 0.269544780254364, "global_step": 48595, "epoch": 546, "lr": 9.242478211995246e-05}
{"train_loss": 0.19787226617336273, "global_step": 48596, "epoch": 546, "lr": 9.24244753191182e-05}
{"train_loss": 0.3390304446220398, "global_step": 48597, "epoch": 546, "lr": 9.242416851258051e-05}
{"train_loss": 0.28146418929100037, "global_step": 48598, "epoch": 546, "lr": 9.242386170033938e-05}
{"train_loss": 0.23435574769973755, "global_step": 48599, "epoch": 546, "lr": 9.242355488239491e-05}
{"train_loss": 0.30410036444664, "global_step": 48600, "epoch": 546, "lr": 9.242324805874709e-05}
{"train_loss": 0.28283196687698364, "global_step": 48601, "epoch": 546, "lr": 9.242294122939598e-05}
{"train_loss": 0.2590360939502716, "global_step": 48602, "epoch": 546, "lr": 9.242263439434163e-05}
{"train_loss": 0.23790140450000763, "global_step": 48603, "epoch": 546, "lr": 9.242232755358406e-05}
{"train_loss": 0.31876230239868164, "global_step": 48604, "epoch": 546, "lr": 9.242202070712334e-05}
{"train_loss": 0.30675241351127625, "global_step": 48605, "epoch": 546, "lr": 9.242171385495951e-05}
{"train_loss": 0.23431774973869324, "global_step": 48606, "epoch": 546, "lr": 9.242140699709256e-05}
{"train_loss": 0.3046661913394928, "global_step": 48607, "epoch": 546, "lr": 9.24211001335226e-05}
{"train_loss": 0.2649371027946472, "global_step": 48608, "epoch": 546, "lr": 9.242079326424963e-05}
{"train_loss": 0.22713065147399902, "global_step": 48609, "epoch": 546, "lr": 9.242048638927369e-05}
{"train_loss": 0.21364650130271912, "global_step": 48610, "epoch": 546, "lr": 9.242017950859485e-05}
{"train_loss": 0.2164677381515503, "global_step": 48611, "epoch": 546, "lr": 9.241987262221312e-05}
{"train_loss": 0.3006322979927063, "global_step": 48612, "epoch": 546, "lr": 9.241956573012858e-05}
{"train_loss": 0.28945741057395935, "global_step": 48613, "epoch": 546, "lr": 9.241925883234122e-05}
{"train_loss": 0.27140599489212036, "global_step": 48614, "epoch": 546, "lr": 9.241895192885112e-05}
{"train_loss": 0.2822737395763397, "global_step": 48615, "epoch": 546, "lr": 9.24186450196583e-05}
{"train_loss": 0.2899560332298279, "global_step": 48616, "epoch": 546, "lr": 9.241833810476281e-05}
{"train_loss": 0.19384072721004486, "global_step": 48617, "epoch": 546, "lr": 9.24180311841647e-05}
{"train_loss": 0.2751484513282776, "global_step": 48618, "epoch": 546, "lr": 9.241772425786402e-05}
{"train_loss": 0.28569841384887695, "global_step": 48619, "epoch": 546, "lr": 9.241741732586075e-05}
{"train_loss": 0.2510225772857666, "global_step": 48620, "epoch": 546, "lr": 9.2417110388155e-05}
{"train_loss": 0.2656048834323883, "global_step": 48621, "epoch": 546, "lr": 9.241680344474679e-05}
{"train_loss": 0.35480302572250366, "global_step": 48622, "epoch": 546, "lr": 9.241649649563615e-05}
{"train_loss": 0.2141844928264618, "global_step": 48623, "epoch": 546, "lr": 9.241618954082315e-05}
{"train_loss": 0.23806990683078766, "global_step": 48624, "epoch": 546, "lr": 9.241588258030778e-05}
{"train_loss": 0.20615191757678986, "global_step": 48625, "epoch": 546, "lr": 9.241557561409013e-05}
{"train_loss": 0.24753449857234955, "global_step": 48626, "epoch": 546, "lr": 9.241526864217021e-05}
{"train_loss": 0.212143212556839, "global_step": 48627, "epoch": 546, "lr": 9.241496166454808e-05}
{"train_loss": 0.262521892786026, "global_step": 48628, "epoch": 546, "lr": 9.241465468122378e-05}
{"train_loss": 0.2493181675672531, "global_step": 48629, "epoch": 546, "lr": 9.241434769219734e-05}
{"train_loss": 0.31103888154029846, "global_step": 48630, "epoch": 546, "lr": 9.241404069746882e-05}
{"train_loss": 0.2033587098121643, "global_step": 48631, "epoch": 546, "lr": 9.241373369703823e-05}
{"train_loss": 0.26957952976226807, "global_step": 48632, "epoch": 546, "lr": 9.241342669090564e-05}
{"train_loss": 0.23960749804973602, "global_step": 48633, "epoch": 546, "lr": 9.24131196790711e-05}
{"train_loss": 0.20841598510742188, "global_step": 48634, "epoch": 546, "lr": 9.24128126615346e-05}
{"train_loss": 0.21507565677165985, "global_step": 48635, "epoch": 546, "lr": 9.241250563829623e-05}
{"train_loss": 0.25255313515663147, "global_step": 48636, "epoch": 546, "lr": 9.241219860935602e-05}
{"train_loss": 0.24159666895866394, "global_step": 48637, "epoch": 546, "lr": 9.2411891574714e-05}
{"train_loss": 0.29417967796325684, "global_step": 48638, "epoch": 546, "lr": 9.241158453437023e-05}
{"train_loss": 0.264473557472229, "global_step": 48639, "epoch": 546, "lr": 9.241127748832472e-05}
{"train_loss": 0.25906985998153687, "global_step": 48640, "epoch": 546, "lr": 9.241097043657754e-05}
{"train_loss": 0.2530120015144348, "global_step": 48641, "epoch": 546, "lr": 9.241066337912872e-05}
{"train_loss": 0.2737616002559662, "global_step": 48642, "epoch": 546, "lr": 9.241035631597831e-05}
{"train_loss": 0.31733423471450806, "global_step": 48643, "epoch": 546, "lr": 9.241004924712633e-05}
{"train_loss": 0.29957878589630127, "global_step": 48644, "epoch": 546, "lr": 9.240974217257286e-05}
{"train_loss": 0.2242424041032791, "global_step": 48645, "epoch": 546, "lr": 9.24094350923179e-05}
{"train_loss": 0.3419249653816223, "global_step": 48646, "epoch": 546, "lr": 9.24091280063615e-05}
{"train_loss": 0.2749760150909424, "global_step": 48647, "epoch": 546, "lr": 9.240882091470371e-05}
{"train_loss": 0.28418639302253723, "global_step": 48648, "epoch": 546, "lr": 9.240851381734459e-05}
{"train_loss": 0.19364076852798462, "global_step": 48649, "epoch": 546, "lr": 9.240820671428416e-05}
{"train_loss": 0.29639381170272827, "global_step": 48650, "epoch": 546, "lr": 9.240789960552245e-05}
{"train_loss": 0.27459800243377686, "global_step": 48651, "epoch": 546, "lr": 9.240759249105952e-05}
{"train_loss": 0.25458285212516785, "global_step": 48652, "epoch": 546, "lr": 9.240728537089542e-05}
{"train_loss": 0.3142824172973633, "global_step": 48653, "epoch": 546, "lr": 9.240697824503016e-05}
{"train_loss": 0.20929685235023499, "global_step": 48654, "epoch": 546, "lr": 9.24066711134638e-05}
{"train_loss": 0.23222844302654266, "global_step": 48655, "epoch": 546, "lr": 9.24063639761964e-05}
{"train_loss": 0.30055856704711914, "global_step": 48656, "epoch": 546, "lr": 9.240605683322797e-05}
{"train_loss": 0.255154013633728, "global_step": 48657, "epoch": 546, "lr": 9.240574968455854e-05}
{"train_loss": 0.3385292887687683, "global_step": 48658, "epoch": 546, "lr": 9.240544253018819e-05}
{"train_loss": 0.23860856890678406, "global_step": 48659, "epoch": 546, "lr": 9.240513537011695e-05}
{"train_loss": 0.26303401589393616, "global_step": 48660, "epoch": 546, "lr": 9.240482820434485e-05}
{"train_loss": 0.3136049807071686, "global_step": 48661, "epoch": 546, "lr": 9.240452103287195e-05}
{"train_loss": 0.35512250661849976, "global_step": 48662, "epoch": 546, "lr": 9.240421385569826e-05}
{"train_loss": 0.2991933822631836, "global_step": 48663, "epoch": 546, "lr": 9.240390667282386e-05}
{"train_loss": 0.3027132451534271, "global_step": 48664, "epoch": 546, "lr": 9.240359948424876e-05}
{"train_loss": 0.28843095898628235, "global_step": 48665, "epoch": 546, "lr": 9.2403292289973e-05}
{"train_loss": 0.4047188460826874, "global_step": 48666, "epoch": 546, "lr": 9.240298508999665e-05}
{"train_loss": 0.3320012092590332, "global_step": 48667, "epoch": 546, "lr": 9.240267788431975e-05}
{"train_loss": 0.30442631244659424, "global_step": 48668, "epoch": 546, "lr": 9.24023706729423e-05}
{"train_loss": 0.29920104146003723, "global_step": 48669, "epoch": 546, "lr": 9.240206345586437e-05}
{"train_loss": 0.28438305854797363, "global_step": 48670, "epoch": 546, "lr": 9.240175623308602e-05}
{"train_loss": 0.25674986839294434, "global_step": 48671, "epoch": 546, "lr": 9.240144900460726e-05}
{"train_loss": 0.29535651206970215, "global_step": 48672, "epoch": 546, "lr": 9.240114177042815e-05}
{"train_loss": 0.31973013281822205, "global_step": 48673, "epoch": 546, "lr": 9.24008345305487e-05}
{"train_loss": 0.2624218761920929, "global_step": 48674, "epoch": 546, "lr": 9.240052728496899e-05}
{"train_loss": 0.3690606653690338, "global_step": 48675, "epoch": 546, "lr": 9.240022003368905e-05}
{"train_loss": 0.34906676411628723, "global_step": 48676, "epoch": 546, "lr": 9.23999127767089e-05}
{"train_loss": 0.2732639014720917, "global_step": 48677, "epoch": 546, "lr": 9.239960551402861e-05}
{"train_loss": 0.31333908438682556, "global_step": 48678, "epoch": 546, "lr": 9.239929824564821e-05}
{"train_loss": 0.16833414137363434, "global_step": 48679, "epoch": 546, "lr": 9.239899097156775e-05}
{"train_loss": 0.3344399929046631, "global_step": 48680, "epoch": 546, "lr": 9.239868369178726e-05}
{"train_loss": 0.29531344771385193, "global_step": 48681, "epoch": 546, "lr": 9.239837640630678e-05}
{"train_loss": 0.27476772297634167, "global_step": 48682, "epoch": 546, "lr": 9.239806911512635e-05, "val_loss": 2.882242441177368}
{"train_loss": 0.33568692207336426, "global_step": 48683, "epoch": 547, "lr": 9.239776181824602e-05}
{"train_loss": 0.24559909105300903, "global_step": 48684, "epoch": 547, "lr": 9.239745451566582e-05}
{"train_loss": 0.26525479555130005, "global_step": 48685, "epoch": 547, "lr": 9.239714720738581e-05}
{"train_loss": 0.33101996779441833, "global_step": 48686, "epoch": 547, "lr": 9.239683989340602e-05}
{"train_loss": 0.21418912708759308, "global_step": 48687, "epoch": 547, "lr": 9.239653257372648e-05}
{"train_loss": 0.2589528560638428, "global_step": 48688, "epoch": 547, "lr": 9.239622524834725e-05}
{"train_loss": 0.2333804965019226, "global_step": 48689, "epoch": 547, "lr": 9.239591791726836e-05}
{"train_loss": 0.31469660997390747, "global_step": 48690, "epoch": 547, "lr": 9.239561058048987e-05}
{"train_loss": 0.2662324607372284, "global_step": 48691, "epoch": 547, "lr": 9.23953032380118e-05}
{"train_loss": 0.256413072347641, "global_step": 48692, "epoch": 547, "lr": 9.239499588983419e-05}
{"train_loss": 0.25093013048171997, "global_step": 48693, "epoch": 547, "lr": 9.23946885359571e-05}
{"train_loss": 0.224671870470047, "global_step": 48694, "epoch": 547, "lr": 9.239438117638055e-05}
{"train_loss": 0.30794209241867065, "global_step": 48695, "epoch": 547, "lr": 9.23940738111046e-05}
{"train_loss": 0.23290807008743286, "global_step": 48696, "epoch": 547, "lr": 9.239376644012928e-05}
{"train_loss": 0.28287285566329956, "global_step": 48697, "epoch": 547, "lr": 9.239345906345463e-05}
{"train_loss": 0.2500797510147095, "global_step": 48698, "epoch": 547, "lr": 9.239315168108071e-05}
{"train_loss": 0.3794282376766205, "global_step": 48699, "epoch": 547, "lr": 9.239284429300754e-05}
{"train_loss": 0.2652692496776581, "global_step": 48700, "epoch": 547, "lr": 9.239253689923517e-05}
{"train_loss": 0.33024168014526367, "global_step": 48701, "epoch": 547, "lr": 9.239222949976364e-05}
{"train_loss": 0.33009275794029236, "global_step": 48702, "epoch": 547, "lr": 9.239192209459299e-05}
{"train_loss": 0.26758214831352234, "global_step": 48703, "epoch": 547, "lr": 9.239161468372326e-05}
{"train_loss": 0.31708046793937683, "global_step": 48704, "epoch": 547, "lr": 9.23913072671545e-05}
{"train_loss": 0.2663920223712921, "global_step": 48705, "epoch": 547, "lr": 9.239099984488674e-05}
{"train_loss": 0.2968842685222626, "global_step": 48706, "epoch": 547, "lr": 9.239069241692003e-05}
{"train_loss": 0.25787046551704407, "global_step": 48707, "epoch": 547, "lr": 9.239038498325441e-05}
{"train_loss": 0.2571600377559662, "global_step": 48708, "epoch": 547, "lr": 9.23900775438899e-05}
{"train_loss": 0.28127023577690125, "global_step": 48709, "epoch": 547, "lr": 9.238977009882659e-05}
{"train_loss": 0.30776068568229675, "global_step": 48710, "epoch": 547, "lr": 9.238946264806447e-05}
{"train_loss": 0.2257584035396576, "global_step": 48711, "epoch": 547, "lr": 9.238915519160363e-05}
{"train_loss": 0.28761929273605347, "global_step": 48712, "epoch": 547, "lr": 9.238884772944407e-05}
{"train_loss": 0.2205003798007965, "global_step": 48713, "epoch": 547, "lr": 9.238854026158583e-05}
{"train_loss": 0.3024269938468933, "global_step": 48714, "epoch": 547, "lr": 9.238823278802899e-05}
{"train_loss": 0.3438689410686493, "global_step": 48715, "epoch": 547, "lr": 9.238792530877355e-05}
{"train_loss": 0.37841513752937317, "global_step": 48716, "epoch": 547, "lr": 9.238761782381958e-05}
{"train_loss": 0.24522143602371216, "global_step": 48717, "epoch": 547, "lr": 9.238731033316712e-05}
{"train_loss": 0.23616927862167358, "global_step": 48718, "epoch": 547, "lr": 9.238700283681619e-05}
{"train_loss": 0.24526169896125793, "global_step": 48719, "epoch": 547, "lr": 9.238669533476685e-05}
{"train_loss": 0.18056629598140717, "global_step": 48720, "epoch": 547, "lr": 9.238638782701914e-05}
{"train_loss": 0.2575678825378418, "global_step": 48721, "epoch": 547, "lr": 9.238608031357307e-05}
{"train_loss": 0.25765883922576904, "global_step": 48722, "epoch": 547, "lr": 9.238577279442875e-05}
{"train_loss": 0.29521745443344116, "global_step": 48723, "epoch": 547, "lr": 9.238546526958615e-05}
{"train_loss": 0.27541178464889526, "global_step": 48724, "epoch": 547, "lr": 9.238515773904535e-05}
{"train_loss": 0.24924542009830475, "global_step": 48725, "epoch": 547, "lr": 9.238485020280638e-05}
{"train_loss": 0.316193163394928, "global_step": 48726, "epoch": 547, "lr": 9.238454266086929e-05}
{"train_loss": 0.32849714159965515, "global_step": 48727, "epoch": 547, "lr": 9.238423511323411e-05}
{"train_loss": 0.21886979043483734, "global_step": 48728, "epoch": 547, "lr": 9.238392755990089e-05}
{"train_loss": 0.23094163835048676, "global_step": 48729, "epoch": 547, "lr": 9.238362000086965e-05}
{"train_loss": 0.2843534052371979, "global_step": 48730, "epoch": 547, "lr": 9.238331243614046e-05}
{"train_loss": 0.2978285253047943, "global_step": 48731, "epoch": 547, "lr": 9.238300486571337e-05}
{"train_loss": 0.2403191179037094, "global_step": 48732, "epoch": 547, "lr": 9.238269728958837e-05}
{"train_loss": 0.31344074010849, "global_step": 48733, "epoch": 547, "lr": 9.238238970776555e-05}
{"train_loss": 0.24553433060646057, "global_step": 48734, "epoch": 547, "lr": 9.238208212024493e-05}
{"train_loss": 0.27503702044487, "global_step": 48735, "epoch": 547, "lr": 9.238177452702655e-05}
{"train_loss": 0.23457910120487213, "global_step": 48736, "epoch": 547, "lr": 9.238146692811047e-05}
{"train_loss": 0.20536744594573975, "global_step": 48737, "epoch": 547, "lr": 9.238115932349672e-05}
{"train_loss": 0.29984596371650696, "global_step": 48738, "epoch": 547, "lr": 9.238085171318532e-05}
{"train_loss": 0.24080698192119598, "global_step": 48739, "epoch": 547, "lr": 9.238054409717634e-05}
{"train_loss": 0.24578094482421875, "global_step": 48740, "epoch": 547, "lr": 9.238023647546981e-05}
{"train_loss": 0.304324746131897, "global_step": 48741, "epoch": 547, "lr": 9.23799288480658e-05}
{"train_loss": 0.28078311681747437, "global_step": 48742, "epoch": 547, "lr": 9.237962121496428e-05}
{"train_loss": 0.223896324634552, "global_step": 48743, "epoch": 547, "lr": 9.237931357616538e-05}
{"train_loss": 0.27053260803222656, "global_step": 48744, "epoch": 547, "lr": 9.237900593166907e-05}
{"train_loss": 0.3523084223270416, "global_step": 48745, "epoch": 547, "lr": 9.237869828147542e-05}
{"train_loss": 0.2791689336299896, "global_step": 48746, "epoch": 547, "lr": 9.23783906255845e-05}
{"train_loss": 0.29932305216789246, "global_step": 48747, "epoch": 547, "lr": 9.237808296399628e-05}
{"train_loss": 0.3051701784133911, "global_step": 48748, "epoch": 547, "lr": 9.237777529671086e-05}
{"train_loss": 0.30646494030952454, "global_step": 48749, "epoch": 547, "lr": 9.237746762372827e-05}
{"train_loss": 0.3172532618045807, "global_step": 48750, "epoch": 547, "lr": 9.237715994504854e-05}
{"train_loss": 0.27919867634773254, "global_step": 48751, "epoch": 547, "lr": 9.237685226067173e-05}
{"train_loss": 0.17959070205688477, "global_step": 48752, "epoch": 547, "lr": 9.237654457059786e-05}
{"train_loss": 0.39766886830329895, "global_step": 48753, "epoch": 547, "lr": 9.237623687482697e-05}
{"train_loss": 0.3496704399585724, "global_step": 48754, "epoch": 547, "lr": 9.237592917335913e-05}
{"train_loss": 0.26347967982292175, "global_step": 48755, "epoch": 547, "lr": 9.237562146619435e-05}
{"train_loss": 0.4055347442626953, "global_step": 48756, "epoch": 547, "lr": 9.237531375333269e-05}
{"train_loss": 0.3509191870689392, "global_step": 48757, "epoch": 547, "lr": 9.237500603477418e-05}
{"train_loss": 0.30269715189933777, "global_step": 48758, "epoch": 547, "lr": 9.237469831051889e-05}
{"train_loss": 0.3874100148677826, "global_step": 48759, "epoch": 547, "lr": 9.237439058056681e-05}
{"train_loss": 0.25679075717926025, "global_step": 48760, "epoch": 547, "lr": 9.237408284491803e-05}
{"train_loss": 0.22960563004016876, "global_step": 48761, "epoch": 547, "lr": 9.237377510357254e-05}
{"train_loss": 0.3041955828666687, "global_step": 48762, "epoch": 547, "lr": 9.237346735653043e-05}
{"train_loss": 0.3575420677661896, "global_step": 48763, "epoch": 547, "lr": 9.237315960379174e-05}
{"train_loss": 0.2530271112918854, "global_step": 48764, "epoch": 547, "lr": 9.237285184535646e-05}
{"train_loss": 0.24992695450782776, "global_step": 48765, "epoch": 547, "lr": 9.23725440812247e-05}
{"train_loss": 0.394604355096817, "global_step": 48766, "epoch": 547, "lr": 9.237223631139645e-05}
{"train_loss": 0.3505827784538269, "global_step": 48767, "epoch": 547, "lr": 9.237192853587178e-05}
{"train_loss": 0.22672292590141296, "global_step": 48768, "epoch": 547, "lr": 9.237162075465071e-05}
{"train_loss": 0.2586188316345215, "global_step": 48769, "epoch": 547, "lr": 9.23713129677333e-05}
{"train_loss": 0.2864435315132141, "global_step": 48770, "epoch": 547, "lr": 9.237100517511958e-05}
{"train_loss": 0.28325759276245416, "global_step": 48771, "epoch": 547, "lr": 9.237069737680959e-05, "val_loss": 2.9887917041778564}
{"train_loss": 0.30130016803741455, "global_step": 48772, "epoch": 548, "lr": 9.237038957280339e-05}
{"train_loss": 0.19235336780548096, "global_step": 48773, "epoch": 548, "lr": 9.237008176310101e-05}
{"train_loss": 0.31416353583335876, "global_step": 48774, "epoch": 548, "lr": 9.236977394770247e-05}
{"train_loss": 0.27370572090148926, "global_step": 48775, "epoch": 548, "lr": 9.236946612660783e-05}
{"train_loss": 0.2176962047815323, "global_step": 48776, "epoch": 548, "lr": 9.236915829981714e-05}
{"train_loss": 0.32439738512039185, "global_step": 48777, "epoch": 548, "lr": 9.236885046733044e-05}
{"train_loss": 0.24977757036685944, "global_step": 48778, "epoch": 548, "lr": 9.236854262914776e-05}
{"train_loss": 0.29929205775260925, "global_step": 48779, "epoch": 548, "lr": 9.236823478526914e-05}
{"train_loss": 0.26996922492980957, "global_step": 48780, "epoch": 548, "lr": 9.236792693569463e-05}
{"train_loss": 0.25060656666755676, "global_step": 48781, "epoch": 548, "lr": 9.236761908042426e-05}
{"train_loss": 0.2682429850101471, "global_step": 48782, "epoch": 548, "lr": 9.23673112194581e-05}
{"train_loss": 0.2627429962158203, "global_step": 48783, "epoch": 548, "lr": 9.236700335279615e-05}
{"train_loss": 0.2503792345523834, "global_step": 48784, "epoch": 548, "lr": 9.236669548043849e-05}
{"train_loss": 0.3709152340888977, "global_step": 48785, "epoch": 548, "lr": 9.236638760238514e-05}
{"train_loss": 0.2885839641094208, "global_step": 48786, "epoch": 548, "lr": 9.236607971863614e-05}
{"train_loss": 0.20138995349407196, "global_step": 48787, "epoch": 548, "lr": 9.236577182919154e-05}
{"train_loss": 0.3039266765117645, "global_step": 48788, "epoch": 548, "lr": 9.236546393405139e-05}
{"train_loss": 0.32613852620124817, "global_step": 48789, "epoch": 548, "lr": 9.23651560332157e-05}
{"train_loss": 0.23828516900539398, "global_step": 48790, "epoch": 548, "lr": 9.236484812668454e-05}
{"train_loss": 0.27021440863609314, "global_step": 48791, "epoch": 548, "lr": 9.236454021445794e-05}
{"train_loss": 0.24244004487991333, "global_step": 48792, "epoch": 548, "lr": 9.236423229653594e-05}
{"train_loss": 0.28360116481781006, "global_step": 48793, "epoch": 548, "lr": 9.236392437291859e-05}
{"train_loss": 0.1882149875164032, "global_step": 48794, "epoch": 548, "lr": 9.236361644360594e-05}
{"train_loss": 0.28655537962913513, "global_step": 48795, "epoch": 548, "lr": 9.2363308508598e-05}
{"train_loss": 0.36664822697639465, "global_step": 48796, "epoch": 548, "lr": 9.236300056789484e-05}
{"train_loss": 0.2303311824798584, "global_step": 48797, "epoch": 548, "lr": 9.236269262149649e-05}
{"train_loss": 0.3193443715572357, "global_step": 48798, "epoch": 548, "lr": 9.236238466940299e-05}
{"train_loss": 0.2644728720188141, "global_step": 48799, "epoch": 548, "lr": 9.236207671161438e-05}
{"train_loss": 0.284228652715683, "global_step": 48800, "epoch": 548, "lr": 9.23617687481307e-05}
{"train_loss": 0.2920053005218506, "global_step": 48801, "epoch": 548, "lr": 9.2361460778952e-05}
{"train_loss": 0.27219170331954956, "global_step": 48802, "epoch": 548, "lr": 9.236115280407833e-05}
{"train_loss": 0.2904621362686157, "global_step": 48803, "epoch": 548, "lr": 9.236084482350971e-05}
{"train_loss": 0.1655386984348297, "global_step": 48804, "epoch": 548, "lr": 9.23605368372462e-05}
{"train_loss": 0.29993340373039246, "global_step": 48805, "epoch": 548, "lr": 9.236022884528782e-05}
{"train_loss": 0.22282257676124573, "global_step": 48806, "epoch": 548, "lr": 9.235992084763462e-05}
{"train_loss": 0.39476877450942993, "global_step": 48807, "epoch": 548, "lr": 9.235961284428664e-05}
{"train_loss": 0.3526735007762909, "global_step": 48808, "epoch": 548, "lr": 9.235930483524395e-05}
{"train_loss": 0.2729203999042511, "global_step": 48809, "epoch": 548, "lr": 9.235899682050656e-05}
{"train_loss": 0.38959619402885437, "global_step": 48810, "epoch": 548, "lr": 9.235868880007451e-05}
{"train_loss": 0.29508253931999207, "global_step": 48811, "epoch": 548, "lr": 9.235838077394787e-05}
{"train_loss": 0.2542017102241516, "global_step": 48812, "epoch": 548, "lr": 9.235807274212664e-05}
{"train_loss": 0.29598185420036316, "global_step": 48813, "epoch": 548, "lr": 9.23577647046109e-05}
{"train_loss": 0.2655091881752014, "global_step": 48814, "epoch": 548, "lr": 9.235745666140065e-05}
{"train_loss": 0.2907487452030182, "global_step": 48815, "epoch": 548, "lr": 9.235714861249598e-05}
{"train_loss": 0.3799993097782135, "global_step": 48816, "epoch": 548, "lr": 9.235684055789691e-05}
{"train_loss": 0.2834486663341522, "global_step": 48817, "epoch": 548, "lr": 9.235653249760346e-05}
{"train_loss": 0.21792474389076233, "global_step": 48818, "epoch": 548, "lr": 9.23562244316157e-05}
{"train_loss": 0.23443283140659332, "global_step": 48819, "epoch": 548, "lr": 9.235591635993366e-05}
{"train_loss": 0.29057684540748596, "global_step": 48820, "epoch": 548, "lr": 9.235560828255739e-05}
{"train_loss": 0.28075680136680603, "global_step": 48821, "epoch": 548, "lr": 9.235530019948691e-05}
{"train_loss": 0.3418982923030853, "global_step": 48822, "epoch": 548, "lr": 9.235499211072229e-05}
{"train_loss": 0.3017726540565491, "global_step": 48823, "epoch": 548, "lr": 9.235468401626356e-05}
{"train_loss": 0.31014975905418396, "global_step": 48824, "epoch": 548, "lr": 9.235437591611076e-05}
{"train_loss": 0.3412998616695404, "global_step": 48825, "epoch": 548, "lr": 9.23540678102639e-05}
{"train_loss": 0.3038958013057709, "global_step": 48826, "epoch": 548, "lr": 9.235375969872308e-05}
{"train_loss": 0.3095396161079407, "global_step": 48827, "epoch": 548, "lr": 9.235345158148831e-05}
{"train_loss": 0.25916787981987, "global_step": 48828, "epoch": 548, "lr": 9.235314345855964e-05}
{"train_loss": 0.36390313506126404, "global_step": 48829, "epoch": 548, "lr": 9.23528353299371e-05}
{"train_loss": 0.2705931067466736, "global_step": 48830, "epoch": 548, "lr": 9.235252719562074e-05}
{"train_loss": 0.37521517276763916, "global_step": 48831, "epoch": 548, "lr": 9.235221905561058e-05}
{"train_loss": 0.24140262603759766, "global_step": 48832, "epoch": 548, "lr": 9.23519109099067e-05}
{"train_loss": 0.20728784799575806, "global_step": 48833, "epoch": 548, "lr": 9.235160275850912e-05}
{"train_loss": 0.2633201479911804, "global_step": 48834, "epoch": 548, "lr": 9.235129460141788e-05}
{"train_loss": 0.23657751083374023, "global_step": 48835, "epoch": 548, "lr": 9.235098643863303e-05}
{"train_loss": 0.2597680389881134, "global_step": 48836, "epoch": 548, "lr": 9.23506782701546e-05}
{"train_loss": 0.29439932107925415, "global_step": 48837, "epoch": 548, "lr": 9.235037009598264e-05}
{"train_loss": 0.20869095623493195, "global_step": 48838, "epoch": 548, "lr": 9.235006191611718e-05}
{"train_loss": 0.3636573851108551, "global_step": 48839, "epoch": 548, "lr": 9.234975373055828e-05}
{"train_loss": 0.2495783418416977, "global_step": 48840, "epoch": 548, "lr": 9.234944553930596e-05}
{"train_loss": 0.301551878452301, "global_step": 48841, "epoch": 548, "lr": 9.234913734236028e-05}
{"train_loss": 0.49238428473472595, "global_step": 48842, "epoch": 548, "lr": 9.23488291397213e-05}
{"train_loss": 0.3236631751060486, "global_step": 48843, "epoch": 548, "lr": 9.2348520931389e-05}
{"train_loss": 0.23958274722099304, "global_step": 48844, "epoch": 548, "lr": 9.234821271736346e-05}
{"train_loss": 0.21403174102306366, "global_step": 48845, "epoch": 548, "lr": 9.234790449764474e-05}
{"train_loss": 0.25579240918159485, "global_step": 48846, "epoch": 548, "lr": 9.234759627223284e-05}
{"train_loss": 0.26801228523254395, "global_step": 48847, "epoch": 548, "lr": 9.234728804112784e-05}
{"train_loss": 0.47855299711227417, "global_step": 48848, "epoch": 548, "lr": 9.234697980432975e-05}
{"train_loss": 0.2440105676651001, "global_step": 48849, "epoch": 548, "lr": 9.234667156183864e-05}
{"train_loss": 0.27657222747802734, "global_step": 48850, "epoch": 548, "lr": 9.234636331365452e-05}
{"train_loss": 0.33248811960220337, "global_step": 48851, "epoch": 548, "lr": 9.234605505977746e-05}
{"train_loss": 0.34076163172721863, "global_step": 48852, "epoch": 548, "lr": 9.234574680020749e-05}
{"train_loss": 0.32006555795669556, "global_step": 48853, "epoch": 548, "lr": 9.234543853494464e-05}
{"train_loss": 0.28154444694519043, "global_step": 48854, "epoch": 548, "lr": 9.234513026398896e-05}
{"train_loss": 0.3260752856731415, "global_step": 48855, "epoch": 548, "lr": 9.234482198734051e-05}
{"train_loss": 0.30169355869293213, "global_step": 48856, "epoch": 548, "lr": 9.23445137049993e-05}
{"train_loss": 0.3166413903236389, "global_step": 48857, "epoch": 548, "lr": 9.234420541696537e-05}
{"train_loss": 0.3053095042705536, "global_step": 48858, "epoch": 548, "lr": 9.234389712323881e-05}
{"train_loss": 0.3260437548160553, "global_step": 48859, "epoch": 548, "lr": 9.234358882381961e-05}
{"train_loss": 0.28863026720754215, "global_step": 48860, "epoch": 548, "lr": 9.234328051870784e-05, "val_loss": 2.929929733276367}
{"train_loss": 0.25143083930015564, "global_step": 48861, "epoch": 549, "lr": 9.234297220790353e-05}
{"train_loss": 0.2034258395433426, "global_step": 48862, "epoch": 549, "lr": 9.234266389140674e-05}
{"train_loss": 0.28087472915649414, "global_step": 48863, "epoch": 549, "lr": 9.234235556921747e-05}
{"train_loss": 0.3460206389427185, "global_step": 48864, "epoch": 549, "lr": 9.23420472413358e-05}
{"train_loss": 0.3966901898384094, "global_step": 48865, "epoch": 549, "lr": 9.234173890776175e-05}
{"train_loss": 0.2919926047325134, "global_step": 48866, "epoch": 549, "lr": 9.234143056849536e-05}
{"train_loss": 0.2755372226238251, "global_step": 48867, "epoch": 549, "lr": 9.234112222353669e-05}
{"train_loss": 0.3923037350177765, "global_step": 48868, "epoch": 549, "lr": 9.234081387288578e-05}
{"train_loss": 0.3191319406032562, "global_step": 48869, "epoch": 549, "lr": 9.234050551654265e-05}
{"train_loss": 0.268772691488266, "global_step": 48870, "epoch": 549, "lr": 9.234019715450737e-05}
{"train_loss": 0.29978686571121216, "global_step": 48871, "epoch": 549, "lr": 9.233988878677994e-05}
{"train_loss": 0.3295924961566925, "global_step": 48872, "epoch": 549, "lr": 9.233958041336047e-05}
{"train_loss": 0.3754121661186218, "global_step": 48873, "epoch": 549, "lr": 9.233927203424893e-05}
{"train_loss": 0.39025139808654785, "global_step": 48874, "epoch": 549, "lr": 9.233896364944539e-05}
{"train_loss": 0.337022602558136, "global_step": 48875, "epoch": 549, "lr": 9.233865525894992e-05}
{"train_loss": 0.33431097865104675, "global_step": 48876, "epoch": 549, "lr": 9.233834686276251e-05}
{"train_loss": 0.2669874131679535, "global_step": 48877, "epoch": 549, "lr": 9.233803846088322e-05}
{"train_loss": 0.3256821632385254, "global_step": 48878, "epoch": 549, "lr": 9.23377300533121e-05}
{"train_loss": 0.3261209726333618, "global_step": 48879, "epoch": 549, "lr": 9.233742164004919e-05}
{"train_loss": 0.29482918977737427, "global_step": 48880, "epoch": 549, "lr": 9.233711322109454e-05}
{"train_loss": 0.28628480434417725, "global_step": 48881, "epoch": 549, "lr": 9.233680479644817e-05}
{"train_loss": 0.26067930459976196, "global_step": 48882, "epoch": 549, "lr": 9.233649636611015e-05}
{"train_loss": 0.23692506551742554, "global_step": 48883, "epoch": 549, "lr": 9.233618793008049e-05}
{"train_loss": 0.34635403752326965, "global_step": 48884, "epoch": 549, "lr": 9.233587948835925e-05}
{"train_loss": 0.29310035705566406, "global_step": 48885, "epoch": 549, "lr": 9.233557104094645e-05}
{"train_loss": 0.2745826542377472, "global_step": 48886, "epoch": 549, "lr": 9.233526258784218e-05}
{"train_loss": 0.2573467195034027, "global_step": 48887, "epoch": 549, "lr": 9.233495412904643e-05}
{"train_loss": 0.317119300365448, "global_step": 48888, "epoch": 549, "lr": 9.233464566455925e-05}
{"train_loss": 0.2920788824558258, "global_step": 48889, "epoch": 549, "lr": 9.233433719438072e-05}
{"train_loss": 0.3307563066482544, "global_step": 48890, "epoch": 549, "lr": 9.233402871851083e-05}
{"train_loss": 0.276202529668808, "global_step": 48891, "epoch": 549, "lr": 9.233372023694966e-05}
{"train_loss": 0.3220119774341583, "global_step": 48892, "epoch": 549, "lr": 9.233341174969723e-05}
{"train_loss": 0.31731048226356506, "global_step": 48893, "epoch": 549, "lr": 9.23331032567536e-05}
{"train_loss": 0.2546231150627136, "global_step": 48894, "epoch": 549, "lr": 9.23327947581188e-05}
{"train_loss": 0.2520864009857178, "global_step": 48895, "epoch": 549, "lr": 9.233248625379286e-05}
{"train_loss": 0.27877575159072876, "global_step": 48896, "epoch": 549, "lr": 9.233217774377583e-05}
{"train_loss": 0.2768237292766571, "global_step": 48897, "epoch": 549, "lr": 9.233186922806776e-05}
{"train_loss": 0.21707512438297272, "global_step": 48898, "epoch": 549, "lr": 9.23315607066687e-05}
{"train_loss": 0.19787552952766418, "global_step": 48899, "epoch": 549, "lr": 9.233125217957866e-05}
{"train_loss": 0.24035820364952087, "global_step": 48900, "epoch": 549, "lr": 9.233094364679771e-05}
{"train_loss": 0.22187690436840057, "global_step": 48901, "epoch": 549, "lr": 9.233063510832587e-05}
{"train_loss": 0.2602454125881195, "global_step": 48902, "epoch": 549, "lr": 9.233032656416321e-05}
{"train_loss": 0.2928144633769989, "global_step": 48903, "epoch": 549, "lr": 9.233001801430974e-05}
{"train_loss": 0.21377937495708466, "global_step": 48904, "epoch": 549, "lr": 9.232970945876552e-05}
{"train_loss": 0.24246633052825928, "global_step": 48905, "epoch": 549, "lr": 9.232940089753058e-05}
{"train_loss": 0.30701643228530884, "global_step": 48906, "epoch": 549, "lr": 9.232909233060496e-05}
{"train_loss": 0.26802384853363037, "global_step": 48907, "epoch": 549, "lr": 9.232878375798873e-05}
{"train_loss": 0.25408199429512024, "global_step": 48908, "epoch": 549, "lr": 9.23284751796819e-05}
{"train_loss": 0.28011858463287354, "global_step": 48909, "epoch": 549, "lr": 9.232816659568451e-05}
{"train_loss": 0.19398240745067596, "global_step": 48910, "epoch": 549, "lr": 9.232785800599664e-05}
{"train_loss": 0.38692358136177063, "global_step": 48911, "epoch": 549, "lr": 9.23275494106183e-05}
{"train_loss": 0.2959541082382202, "global_step": 48912, "epoch": 549, "lr": 9.232724080954952e-05}
{"train_loss": 0.24530532956123352, "global_step": 48913, "epoch": 549, "lr": 9.232693220279037e-05}
{"train_loss": 0.2078409641981125, "global_step": 48914, "epoch": 549, "lr": 9.232662359034087e-05}
{"train_loss": 0.30418550968170166, "global_step": 48915, "epoch": 549, "lr": 9.23263149722011e-05}
{"train_loss": 0.21973657608032227, "global_step": 48916, "epoch": 549, "lr": 9.232600634837104e-05}
{"train_loss": 0.284259170293808, "global_step": 48917, "epoch": 549, "lr": 9.232569771885077e-05}
{"train_loss": 0.29005369544029236, "global_step": 48918, "epoch": 549, "lr": 9.232538908364034e-05}
{"train_loss": 0.18600492179393768, "global_step": 48919, "epoch": 549, "lr": 9.232508044273977e-05}
{"train_loss": 0.43903857469558716, "global_step": 48920, "epoch": 549, "lr": 9.232477179614912e-05}
{"train_loss": 0.22870676219463348, "global_step": 48921, "epoch": 549, "lr": 9.23244631438684e-05}
{"train_loss": 0.32075685262680054, "global_step": 48922, "epoch": 549, "lr": 9.232415448589769e-05}
{"train_loss": 0.24073579907417297, "global_step": 48923, "epoch": 549, "lr": 9.232384582223699e-05}
{"train_loss": 0.31340813636779785, "global_step": 48924, "epoch": 549, "lr": 9.232353715288638e-05}
{"train_loss": 0.29676976799964905, "global_step": 48925, "epoch": 549, "lr": 9.232322847784589e-05}
{"train_loss": 0.2123594582080841, "global_step": 48926, "epoch": 549, "lr": 9.232291979711556e-05}
{"train_loss": 0.31953731179237366, "global_step": 48927, "epoch": 549, "lr": 9.232261111069541e-05}
{"train_loss": 0.21817152202129364, "global_step": 48928, "epoch": 549, "lr": 9.232230241858551e-05}
{"train_loss": 0.24452592432498932, "global_step": 48929, "epoch": 549, "lr": 9.23219937207859e-05}
{"train_loss": 0.3523344099521637, "global_step": 48930, "epoch": 549, "lr": 9.232168501729662e-05}
{"train_loss": 0.30398812890052795, "global_step": 48931, "epoch": 549, "lr": 9.232137630811768e-05}
{"train_loss": 0.3352549076080322, "global_step": 48932, "epoch": 549, "lr": 9.232106759324917e-05}
{"train_loss": 0.2744283378124237, "global_step": 48933, "epoch": 549, "lr": 9.23207588726911e-05}
{"train_loss": 0.22461377084255219, "global_step": 48934, "epoch": 549, "lr": 9.232045014644351e-05}
{"train_loss": 0.2712593376636505, "global_step": 48935, "epoch": 549, "lr": 9.232014141450647e-05}
{"train_loss": 0.3001396358013153, "global_step": 48936, "epoch": 549, "lr": 9.231983267688e-05}
{"train_loss": 0.22414326667785645, "global_step": 48937, "epoch": 549, "lr": 9.231952393356413e-05}
{"train_loss": 0.23644831776618958, "global_step": 48938, "epoch": 549, "lr": 9.231921518455893e-05}
{"train_loss": 0.2825582027435303, "global_step": 48939, "epoch": 549, "lr": 9.231890642986441e-05}
{"train_loss": 0.3148269057273865, "global_step": 48940, "epoch": 549, "lr": 9.231859766948064e-05}
{"train_loss": 0.27512457966804504, "global_step": 48941, "epoch": 549, "lr": 9.231828890340766e-05}
{"train_loss": 0.3436501920223236, "global_step": 48942, "epoch": 549, "lr": 9.231798013164549e-05}
{"train_loss": 0.262113481760025, "global_step": 48943, "epoch": 549, "lr": 9.231767135419418e-05}
{"train_loss": 0.2826346457004547, "global_step": 48944, "epoch": 549, "lr": 9.231736257105378e-05}
{"train_loss": 0.3398972749710083, "global_step": 48945, "epoch": 549, "lr": 9.231705378222432e-05}
{"train_loss": 0.39532744884490967, "global_step": 48946, "epoch": 549, "lr": 9.231674498770586e-05}
{"train_loss": 0.23685811460018158, "global_step": 48947, "epoch": 549, "lr": 9.231643618749842e-05}
{"train_loss": 0.28027838468551636, "global_step": 48948, "epoch": 549, "lr": 9.231612738160205e-05}
{"train_loss": 0.28559399789638734, "global_step": 48949, "epoch": 549, "lr": 9.231581857001681e-05, "val_loss": 2.941741943359375}
{"train_loss": 0.2817799746990204, "global_step": 48950, "epoch": 550, "lr": 9.23155097527427e-05}
{"train_loss": 0.26322123408317566, "global_step": 48951, "epoch": 550, "lr": 9.231520092977981e-05}
{"train_loss": 0.26101189851760864, "global_step": 48952, "epoch": 550, "lr": 9.231489210112814e-05}
{"train_loss": 0.34267768263816833, "global_step": 48953, "epoch": 550, "lr": 9.231458326678774e-05}
{"train_loss": 0.3569241762161255, "global_step": 48954, "epoch": 550, "lr": 9.231427442675869e-05}
{"train_loss": 0.33218494057655334, "global_step": 48955, "epoch": 550, "lr": 9.231396558104097e-05}
{"train_loss": 0.3051029145717621, "global_step": 48956, "epoch": 550, "lr": 9.231365672963469e-05}
{"train_loss": 0.3061637580394745, "global_step": 48957, "epoch": 550, "lr": 9.231334787253981e-05}
{"train_loss": 0.2102697789669037, "global_step": 48958, "epoch": 550, "lr": 9.231303900975644e-05}
{"train_loss": 0.2907177805900574, "global_step": 48959, "epoch": 550, "lr": 9.23127301412846e-05}
{"train_loss": 0.2886229455471039, "global_step": 48960, "epoch": 550, "lr": 9.231242126712431e-05}
{"train_loss": 0.30025407671928406, "global_step": 48961, "epoch": 550, "lr": 9.231211238727567e-05}
{"train_loss": 0.34046390652656555, "global_step": 48962, "epoch": 550, "lr": 9.231180350173865e-05}
{"train_loss": 0.3362709879875183, "global_step": 48963, "epoch": 550, "lr": 9.231149461051332e-05}
{"train_loss": 0.2257552444934845, "global_step": 48964, "epoch": 550, "lr": 9.231118571359974e-05}
{"train_loss": 0.2940284311771393, "global_step": 48965, "epoch": 550, "lr": 9.231087681099793e-05}
{"train_loss": 0.40030795335769653, "global_step": 48966, "epoch": 550, "lr": 9.231056790270795e-05}
{"train_loss": 0.3296366333961487, "global_step": 48967, "epoch": 550, "lr": 9.231025898872981e-05}
{"train_loss": 0.24407637119293213, "global_step": 48968, "epoch": 550, "lr": 9.230995006906358e-05}
{"train_loss": 0.28610849380493164, "global_step": 48969, "epoch": 550, "lr": 9.23096411437093e-05}
{"train_loss": 0.20013554394245148, "global_step": 48970, "epoch": 550, "lr": 9.2309332212667e-05}
{"train_loss": 0.2611757814884186, "global_step": 48971, "epoch": 550, "lr": 9.230902327593672e-05}
{"train_loss": 0.34907838702201843, "global_step": 48972, "epoch": 550, "lr": 9.230871433351852e-05}
{"train_loss": 0.3393529951572418, "global_step": 48973, "epoch": 550, "lr": 9.230840538541241e-05}
{"train_loss": 0.2348669320344925, "global_step": 48974, "epoch": 550, "lr": 9.230809643161847e-05}
{"train_loss": 0.3430129587650299, "global_step": 48975, "epoch": 550, "lr": 9.230778747213672e-05}
{"train_loss": 0.36126989126205444, "global_step": 48976, "epoch": 550, "lr": 9.23074785069672e-05}
{"train_loss": 0.21673916280269623, "global_step": 48977, "epoch": 550, "lr": 9.230716953610995e-05}
{"train_loss": 0.244229257106781, "global_step": 48978, "epoch": 550, "lr": 9.230686055956502e-05}
{"train_loss": 0.30525702238082886, "global_step": 48979, "epoch": 550, "lr": 9.230655157733245e-05}
{"train_loss": 0.2150803804397583, "global_step": 48980, "epoch": 550, "lr": 9.230624258941227e-05}
{"train_loss": 0.19149495661258698, "global_step": 48981, "epoch": 550, "lr": 9.230593359580455e-05}
{"train_loss": 0.30721086263656616, "global_step": 48982, "epoch": 550, "lr": 9.23056245965093e-05}
{"train_loss": 0.2939921021461487, "global_step": 48983, "epoch": 550, "lr": 9.230531559152658e-05}
{"train_loss": 0.23097890615463257, "global_step": 48984, "epoch": 550, "lr": 9.230500658085642e-05}
{"train_loss": 0.26816779375076294, "global_step": 48985, "epoch": 550, "lr": 9.230469756449887e-05}
{"train_loss": 0.30827435851097107, "global_step": 48986, "epoch": 550, "lr": 9.230438854245396e-05}
{"train_loss": 0.22981052100658417, "global_step": 48987, "epoch": 550, "lr": 9.230407951472175e-05}
{"train_loss": 0.33305448293685913, "global_step": 48988, "epoch": 550, "lr": 9.230377048130228e-05}
{"train_loss": 0.2876010835170746, "global_step": 48989, "epoch": 550, "lr": 9.230346144219557e-05}
{"train_loss": 0.30294182896614075, "global_step": 48990, "epoch": 550, "lr": 9.230315239740168e-05}
{"train_loss": 0.2808495759963989, "global_step": 48991, "epoch": 550, "lr": 9.230284334692066e-05}
{"train_loss": 0.2757290005683899, "global_step": 48992, "epoch": 550, "lr": 9.230253429075252e-05}
{"train_loss": 0.28609827160835266, "global_step": 48993, "epoch": 550, "lr": 9.230222522889733e-05}
{"train_loss": 0.2122703641653061, "global_step": 48994, "epoch": 550, "lr": 9.230191616135512e-05}
{"train_loss": 0.39768970012664795, "global_step": 48995, "epoch": 550, "lr": 9.230160708812591e-05}
{"train_loss": 0.2591073215007782, "global_step": 48996, "epoch": 550, "lr": 9.230129800920981e-05}
{"train_loss": 0.2802855670452118, "global_step": 48997, "epoch": 550, "lr": 9.230098892460678e-05}
{"train_loss": 0.27554264664649963, "global_step": 48998, "epoch": 550, "lr": 9.230067983431691e-05}
{"train_loss": 0.26724663376808167, "global_step": 48999, "epoch": 550, "lr": 9.230037073834022e-05}
{"train_loss": 0.28541088104248047, "global_step": 49000, "epoch": 550, "lr": 9.230006163667677e-05}
{"train_loss": 0.25148460268974304, "global_step": 49001, "epoch": 550, "lr": 9.229975252932659e-05}
{"train_loss": 0.25279295444488525, "global_step": 49002, "epoch": 550, "lr": 9.229944341628973e-05}
{"train_loss": 0.24565699696540833, "global_step": 49003, "epoch": 550, "lr": 9.229913429756622e-05}
{"train_loss": 0.220316544175148, "global_step": 49004, "epoch": 550, "lr": 9.229882517315611e-05}
{"train_loss": 0.20594747364521027, "global_step": 49005, "epoch": 550, "lr": 9.229851604305943e-05}
{"train_loss": 0.3467704653739929, "global_step": 49006, "epoch": 550, "lr": 9.229820690727624e-05}
{"train_loss": 0.2650887370109558, "global_step": 49007, "epoch": 550, "lr": 9.229789776580657e-05}
{"train_loss": 0.2767760157585144, "global_step": 49008, "epoch": 550, "lr": 9.229758861865046e-05}
{"train_loss": 0.233684241771698, "global_step": 49009, "epoch": 550, "lr": 9.229727946580795e-05}
{"train_loss": 0.2086670696735382, "global_step": 49010, "epoch": 550, "lr": 9.229697030727909e-05}
{"train_loss": 0.25034937262535095, "global_step": 49011, "epoch": 550, "lr": 9.229666114306392e-05}
{"train_loss": 0.2623615264892578, "global_step": 49012, "epoch": 550, "lr": 9.229635197316247e-05}
{"train_loss": 0.19200442731380463, "global_step": 49013, "epoch": 550, "lr": 9.229604279757481e-05}
{"train_loss": 0.18033306300640106, "global_step": 49014, "epoch": 550, "lr": 9.229573361630095e-05}
{"train_loss": 0.2627647817134857, "global_step": 49015, "epoch": 550, "lr": 9.229542442934095e-05}
{"train_loss": 0.2462100237607956, "global_step": 49016, "epoch": 550, "lr": 9.229511523669484e-05}
{"train_loss": 0.34990644454956055, "global_step": 49017, "epoch": 550, "lr": 9.229480603836266e-05}
{"train_loss": 0.21415852010250092, "global_step": 49018, "epoch": 550, "lr": 9.229449683434448e-05}
{"train_loss": 0.25364989042282104, "global_step": 49019, "epoch": 550, "lr": 9.22941876246403e-05}
{"train_loss": 0.22189973294734955, "global_step": 49020, "epoch": 550, "lr": 9.229387840925018e-05}
{"train_loss": 0.24466153979301453, "global_step": 49021, "epoch": 550, "lr": 9.229356918817418e-05}
{"train_loss": 0.21507754921913147, "global_step": 49022, "epoch": 550, "lr": 9.229325996141232e-05}
{"train_loss": 0.25895974040031433, "global_step": 49023, "epoch": 550, "lr": 9.229295072896466e-05}
{"train_loss": 0.3122910261154175, "global_step": 49024, "epoch": 550, "lr": 9.229264149083122e-05}
{"train_loss": 0.27459952235221863, "global_step": 49025, "epoch": 550, "lr": 9.229233224701204e-05}
{"train_loss": 0.295809268951416, "global_step": 49026, "epoch": 550, "lr": 9.229202299750717e-05}
{"train_loss": 0.3258258104324341, "global_step": 49027, "epoch": 550, "lr": 9.229171374231666e-05}
{"train_loss": 0.18368445336818695, "global_step": 49028, "epoch": 550, "lr": 9.229140448144053e-05}
{"train_loss": 0.22770902514457703, "global_step": 49029, "epoch": 550, "lr": 9.229109521487886e-05}
{"train_loss": 0.31040632724761963, "global_step": 49030, "epoch": 550, "lr": 9.229078594263165e-05}
{"train_loss": 0.21414977312088013, "global_step": 49031, "epoch": 550, "lr": 9.229047666469896e-05}
{"train_loss": 0.2751085162162781, "global_step": 49032, "epoch": 550, "lr": 9.229016738108085e-05}
{"train_loss": 0.2335088700056076, "global_step": 49033, "epoch": 550, "lr": 9.228985809177733e-05}
{"train_loss": 0.22619019448757172, "global_step": 49034, "epoch": 550, "lr": 9.228954879678845e-05}
{"train_loss": 0.23556628823280334, "global_step": 49035, "epoch": 550, "lr": 9.228923949611425e-05}
{"train_loss": 0.30605408549308777, "global_step": 49036, "epoch": 550, "lr": 9.228893018975479e-05}
{"train_loss": 0.278401643037796, "global_step": 49037, "epoch": 550, "lr": 9.228862087771009e-05}
{"train_loss": 0.27171315554152714, "global_step": 49038, "epoch": 550, "lr": 9.228831155998022e-05, "train/sim_max_reward_0": 0.5332832853218983, "train/sim_max_reward_1": 0.9834067856985069, "train/sim_max_reward_2": 0.4533963206566581, "train/sim_max_reward_3": 0.5685260113162178, "train/sim_max_reward_4": 0.9287326634727086, "train/sim_max_reward_5": 0.622262341278169, "test/sim_max_reward_4500000": 0.7899763114853648, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.6990351534781758, "test/sim_max_reward_4500003": 0.35730185072918214, "test/sim_max_reward_4500004": 0.5706079258909327, "test/sim_max_reward_4500005": 0.9550126778594714, "test/sim_max_reward_4500006": 0.9241886010342512, "test/sim_max_reward_4500007": 0.43908873288927275, "test/sim_max_reward_4500008": 0.047584194039662533, "test/sim_max_reward_4500009": 0.050397489115897524, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5314111544626223, "test/sim_max_reward_4500014": 0.8612403155770084, "test/sim_max_reward_4500015": 0.5628735047070829, "test/sim_max_reward_4500016": 0.3810951723407696, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 0.15065393655508955, "test/sim_max_reward_4500019": 0.4105550860915122, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9316365986508279, "test/sim_max_reward_4500023": 0.6626666194367918, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.8938053888064353, "test/sim_max_reward_4500026": 0.5619917198761547, "test/sim_max_reward_4500027": 0.990901244390178, "test/sim_max_reward_4500028": 0.6943302654227346, "test/sim_max_reward_4500029": 0.8238477565254312, "test/sim_max_reward_4500030": 0.7212064946367909, "test/sim_max_reward_4500031": 0.5086101781419692, "test/sim_max_reward_4500032": 0.6854172116498992, "test/sim_max_reward_4500033": 0.25391863966540595, "test/sim_max_reward_4500034": 0.9576373663696565, "test/sim_max_reward_4500035": 0.9191792399579714, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.7206241326908203, "test/sim_max_reward_4500038": 0.5211029865619988, "test/sim_max_reward_4500039": 0.8599739387446219, "test/sim_max_reward_4500040": 0.17778359714540354, "test/sim_max_reward_4500041": 0.9151015832131421, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.2553619459672395, "test/sim_max_reward_4500044": 0.18323707312367193, "test/sim_max_reward_4500045": 0.34813908698120993, "test/sim_max_reward_4500046": 0.374797410040079, "test/sim_max_reward_4500047": 0.9479671754842317, "test/sim_max_reward_4500048": 0.7743586911386549, "test/sim_max_reward_4500049": 0.4354397263142853, "train/mean_score": 0.6816012346240264, "test/mean_score": 0.537001163543838, "val_loss": 3.2345073223114014, "train_action_mse_error": 49.52708435058594}
{"train_loss": 0.3009859621524811, "global_step": 49039, "epoch": 551, "lr": 9.228800223656518e-05}
{"train_loss": 0.24964435398578644, "global_step": 49040, "epoch": 551, "lr": 9.228769290746504e-05}
{"train_loss": 0.23850177228450775, "global_step": 49041, "epoch": 551, "lr": 9.228738357267984e-05}
{"train_loss": 0.1924363076686859, "global_step": 49042, "epoch": 551, "lr": 9.228707423220962e-05}
{"train_loss": 0.2552329897880554, "global_step": 49043, "epoch": 551, "lr": 9.22867648860544e-05}
{"train_loss": 0.285359263420105, "global_step": 49044, "epoch": 551, "lr": 9.228645553421425e-05}
{"train_loss": 0.2252652645111084, "global_step": 49045, "epoch": 551, "lr": 9.22861461766892e-05}
{"train_loss": 0.2930745482444763, "global_step": 49046, "epoch": 551, "lr": 9.228583681347928e-05}
{"train_loss": 0.2563920319080353, "global_step": 49047, "epoch": 551, "lr": 9.228552744458458e-05}
{"train_loss": 0.314102441072464, "global_step": 49048, "epoch": 551, "lr": 9.228521807000508e-05}
{"train_loss": 0.35832396149635315, "global_step": 49049, "epoch": 551, "lr": 9.228490868974084e-05}
{"train_loss": 0.3169935941696167, "global_step": 49050, "epoch": 551, "lr": 9.228459930379194e-05}
{"train_loss": 0.3082728981971741, "global_step": 49051, "epoch": 551, "lr": 9.228428991215837e-05}
{"train_loss": 0.2680419683456421, "global_step": 49052, "epoch": 551, "lr": 9.228398051484019e-05}
{"train_loss": 0.3171333074569702, "global_step": 49053, "epoch": 551, "lr": 9.228367111183747e-05}
{"train_loss": 0.29777881503105164, "global_step": 49054, "epoch": 551, "lr": 9.228336170315019e-05}
{"train_loss": 0.30980297923088074, "global_step": 49055, "epoch": 551, "lr": 9.228305228877845e-05}
{"train_loss": 0.27910712361335754, "global_step": 49056, "epoch": 551, "lr": 9.228274286872226e-05}
{"train_loss": 0.2521545886993408, "global_step": 49057, "epoch": 551, "lr": 9.228243344298167e-05}
{"train_loss": 0.31008797883987427, "global_step": 49058, "epoch": 551, "lr": 9.228212401155674e-05}
{"train_loss": 0.243687242269516, "global_step": 49059, "epoch": 551, "lr": 9.228181457444746e-05}
{"train_loss": 0.3334601819515228, "global_step": 49060, "epoch": 551, "lr": 9.228150513165393e-05}
{"train_loss": 0.2111324816942215, "global_step": 49061, "epoch": 551, "lr": 9.228119568317617e-05}
{"train_loss": 0.2414560317993164, "global_step": 49062, "epoch": 551, "lr": 9.228088622901421e-05}
{"train_loss": 0.2996004819869995, "global_step": 49063, "epoch": 551, "lr": 9.22805767691681e-05}
{"train_loss": 0.2247437834739685, "global_step": 49064, "epoch": 551, "lr": 9.228026730363787e-05}
{"train_loss": 0.26891255378723145, "global_step": 49065, "epoch": 551, "lr": 9.227995783242361e-05}
{"train_loss": 0.20648884773254395, "global_step": 49066, "epoch": 551, "lr": 9.227964835552529e-05}
{"train_loss": 0.27798742055892944, "global_step": 49067, "epoch": 551, "lr": 9.2279338872943e-05}
{"train_loss": 0.2755374014377594, "global_step": 49068, "epoch": 551, "lr": 9.227902938467677e-05}
{"train_loss": 0.2135804444551468, "global_step": 49069, "epoch": 551, "lr": 9.227871989072663e-05}
{"train_loss": 0.21016190946102142, "global_step": 49070, "epoch": 551, "lr": 9.227841039109262e-05}
{"train_loss": 0.2758502662181854, "global_step": 49071, "epoch": 551, "lr": 9.227810088577482e-05}
{"train_loss": 0.25733429193496704, "global_step": 49072, "epoch": 551, "lr": 9.227779137477323e-05}
{"train_loss": 0.17093093693256378, "global_step": 49073, "epoch": 551, "lr": 9.22774818580879e-05}
{"train_loss": 0.22464855015277863, "global_step": 49074, "epoch": 551, "lr": 9.227717233571891e-05}
{"train_loss": 0.2851218581199646, "global_step": 49075, "epoch": 551, "lr": 9.227686280766624e-05}
{"train_loss": 0.28151652216911316, "global_step": 49076, "epoch": 551, "lr": 9.227655327392996e-05}
{"train_loss": 0.3221774399280548, "global_step": 49077, "epoch": 551, "lr": 9.227624373451014e-05}
{"train_loss": 0.24704772233963013, "global_step": 49078, "epoch": 551, "lr": 9.227593418940676e-05}
{"train_loss": 0.33058100938796997, "global_step": 49079, "epoch": 551, "lr": 9.227562463861991e-05}
{"train_loss": 0.32128092646598816, "global_step": 49080, "epoch": 551, "lr": 9.227531508214964e-05}
{"train_loss": 0.30583465099334717, "global_step": 49081, "epoch": 551, "lr": 9.227500551999595e-05}
{"train_loss": 0.27583643794059753, "global_step": 49082, "epoch": 551, "lr": 9.227469595215889e-05}
{"train_loss": 0.28015443682670593, "global_step": 49083, "epoch": 551, "lr": 9.227438637863852e-05}
{"train_loss": 0.2195042222738266, "global_step": 49084, "epoch": 551, "lr": 9.227407679943487e-05}
{"train_loss": 0.27249422669410706, "global_step": 49085, "epoch": 551, "lr": 9.227376721454798e-05}
{"train_loss": 0.2743663191795349, "global_step": 49086, "epoch": 551, "lr": 9.227345762397793e-05}
{"train_loss": 0.24022424221038818, "global_step": 49087, "epoch": 551, "lr": 9.227314802772471e-05}
{"train_loss": 0.2671716511249542, "global_step": 49088, "epoch": 551, "lr": 9.227283842578838e-05}
{"train_loss": 0.3332000970840454, "global_step": 49089, "epoch": 551, "lr": 9.227252881816897e-05}
{"train_loss": 0.49766430258750916, "global_step": 49090, "epoch": 551, "lr": 9.227221920486655e-05}
{"train_loss": 0.25835421681404114, "global_step": 49091, "epoch": 551, "lr": 9.227190958588113e-05}
{"train_loss": 0.31932324171066284, "global_step": 49092, "epoch": 551, "lr": 9.227159996121278e-05}
{"train_loss": 0.2950589954853058, "global_step": 49093, "epoch": 551, "lr": 9.227129033086152e-05}
{"train_loss": 0.23233644664287567, "global_step": 49094, "epoch": 551, "lr": 9.22709806948274e-05}
{"train_loss": 0.3308138847351074, "global_step": 49095, "epoch": 551, "lr": 9.227067105311048e-05}
{"train_loss": 0.1555452048778534, "global_step": 49096, "epoch": 551, "lr": 9.227036140571077e-05}
{"train_loss": 0.3556562066078186, "global_step": 49097, "epoch": 551, "lr": 9.227005175262833e-05}
{"train_loss": 0.24317821860313416, "global_step": 49098, "epoch": 551, "lr": 9.226974209386319e-05}
{"train_loss": 0.2789301574230194, "global_step": 49099, "epoch": 551, "lr": 9.226943242941539e-05}
{"train_loss": 0.22926832735538483, "global_step": 49100, "epoch": 551, "lr": 9.2269122759285e-05}
{"train_loss": 0.37330204248428345, "global_step": 49101, "epoch": 551, "lr": 9.226881308347203e-05}
{"train_loss": 0.2940366864204407, "global_step": 49102, "epoch": 551, "lr": 9.226850340197654e-05}
{"train_loss": 0.37636706233024597, "global_step": 49103, "epoch": 551, "lr": 9.226819371479856e-05}
{"train_loss": 0.18762047588825226, "global_step": 49104, "epoch": 551, "lr": 9.226788402193813e-05}
{"train_loss": 0.1952785700559616, "global_step": 49105, "epoch": 551, "lr": 9.22675743233953e-05}
{"train_loss": 0.25515836477279663, "global_step": 49106, "epoch": 551, "lr": 9.226726461917013e-05}
{"train_loss": 0.31996268033981323, "global_step": 49107, "epoch": 551, "lr": 9.226695490926263e-05}
{"train_loss": 0.4565775692462921, "global_step": 49108, "epoch": 551, "lr": 9.226664519367283e-05}
{"train_loss": 0.3074888586997986, "global_step": 49109, "epoch": 551, "lr": 9.226633547240082e-05}
{"train_loss": 0.3638320565223694, "global_step": 49110, "epoch": 551, "lr": 9.22660257454466e-05}
{"train_loss": 0.31153351068496704, "global_step": 49111, "epoch": 551, "lr": 9.226571601281024e-05}
{"train_loss": 0.2959921658039093, "global_step": 49112, "epoch": 551, "lr": 9.226540627449179e-05}
{"train_loss": 0.26820504665374756, "global_step": 49113, "epoch": 551, "lr": 9.226509653049124e-05}
{"train_loss": 0.28212231397628784, "global_step": 49114, "epoch": 551, "lr": 9.226478678080867e-05}
{"train_loss": 0.32047855854034424, "global_step": 49115, "epoch": 551, "lr": 9.226447702544411e-05}
{"train_loss": 0.3496670722961426, "global_step": 49116, "epoch": 551, "lr": 9.226416726439761e-05}
{"train_loss": 0.3806207478046417, "global_step": 49117, "epoch": 551, "lr": 9.226385749766922e-05}
{"train_loss": 0.3306456506252289, "global_step": 49118, "epoch": 551, "lr": 9.226354772525895e-05}
{"train_loss": 0.2852310538291931, "global_step": 49119, "epoch": 551, "lr": 9.226323794716687e-05}
{"train_loss": 0.18089531362056732, "global_step": 49120, "epoch": 551, "lr": 9.226292816339303e-05}
{"train_loss": 0.2658528983592987, "global_step": 49121, "epoch": 551, "lr": 9.226261837393744e-05}
{"train_loss": 0.21990443766117096, "global_step": 49122, "epoch": 551, "lr": 9.226230857880015e-05}
{"train_loss": 0.25135117769241333, "global_step": 49123, "epoch": 551, "lr": 9.226199877798121e-05}
{"train_loss": 0.27323585748672485, "global_step": 49124, "epoch": 551, "lr": 9.226168897148066e-05}
{"train_loss": 0.43074172735214233, "global_step": 49125, "epoch": 551, "lr": 9.226137915929855e-05}
{"train_loss": 0.3499060869216919, "global_step": 49126, "epoch": 551, "lr": 9.22610693414349e-05}
{"train_loss": 0.28275082185027306, "global_step": 49127, "epoch": 551, "lr": 9.226075951788977e-05, "val_loss": 2.9442615509033203}
{"train_loss": 0.19149334728717804, "global_step": 49128, "epoch": 552, "lr": 9.226044968866319e-05}
{"train_loss": 0.3083706796169281, "global_step": 49129, "epoch": 552, "lr": 9.226013985375522e-05}
{"train_loss": 0.2949608564376831, "global_step": 49130, "epoch": 552, "lr": 9.225983001316589e-05}
{"train_loss": 0.3445761203765869, "global_step": 49131, "epoch": 552, "lr": 9.225952016689523e-05}
{"train_loss": 0.2987581491470337, "global_step": 49132, "epoch": 552, "lr": 9.225921031494329e-05}
{"train_loss": 0.38060280680656433, "global_step": 49133, "epoch": 552, "lr": 9.225890045731013e-05}
{"train_loss": 0.2808704376220703, "global_step": 49134, "epoch": 552, "lr": 9.225859059399577e-05}
{"train_loss": 0.28068891167640686, "global_step": 49135, "epoch": 552, "lr": 9.225828072500025e-05}
{"train_loss": 0.28348708152770996, "global_step": 49136, "epoch": 552, "lr": 9.225797085032362e-05}
{"train_loss": 0.2280951738357544, "global_step": 49137, "epoch": 552, "lr": 9.225766096996592e-05}
{"train_loss": 0.21412108838558197, "global_step": 49138, "epoch": 552, "lr": 9.22573510839272e-05}
{"train_loss": 0.34159016609191895, "global_step": 49139, "epoch": 552, "lr": 9.225704119220748e-05}
{"train_loss": 0.3228728473186493, "global_step": 49140, "epoch": 552, "lr": 9.225673129480683e-05}
{"train_loss": 0.2836376428604126, "global_step": 49141, "epoch": 552, "lr": 9.225642139172528e-05}
{"train_loss": 0.20932301878929138, "global_step": 49142, "epoch": 552, "lr": 9.225611148296287e-05}
{"train_loss": 0.30202269554138184, "global_step": 49143, "epoch": 552, "lr": 9.225580156851963e-05}
{"train_loss": 0.30981606245040894, "global_step": 49144, "epoch": 552, "lr": 9.225549164839562e-05}
{"train_loss": 0.246730238199234, "global_step": 49145, "epoch": 552, "lr": 9.225518172259086e-05}
{"train_loss": 0.3143678903579712, "global_step": 49146, "epoch": 552, "lr": 9.225487179110543e-05}
{"train_loss": 0.2803308963775635, "global_step": 49147, "epoch": 552, "lr": 9.225456185393932e-05}
{"train_loss": 0.3189030885696411, "global_step": 49148, "epoch": 552, "lr": 9.225425191109263e-05}
{"train_loss": 0.26531878113746643, "global_step": 49149, "epoch": 552, "lr": 9.225394196256533e-05}
{"train_loss": 0.25123268365859985, "global_step": 49150, "epoch": 552, "lr": 9.225363200835754e-05}
{"train_loss": 0.13686759769916534, "global_step": 49151, "epoch": 552, "lr": 9.225332204846924e-05}
{"train_loss": 0.18355056643486023, "global_step": 49152, "epoch": 552, "lr": 9.225301208290052e-05}
{"train_loss": 0.29348933696746826, "global_step": 49153, "epoch": 552, "lr": 9.225270211165138e-05}
{"train_loss": 0.29143601655960083, "global_step": 49154, "epoch": 552, "lr": 9.225239213472188e-05}
{"train_loss": 0.28369787335395813, "global_step": 49155, "epoch": 552, "lr": 9.225208215211207e-05}
{"train_loss": 0.2641851603984833, "global_step": 49156, "epoch": 552, "lr": 9.225177216382197e-05}
{"train_loss": 0.20019114017486572, "global_step": 49157, "epoch": 552, "lr": 9.225146216985166e-05}
{"train_loss": 0.2451956570148468, "global_step": 49158, "epoch": 552, "lr": 9.225115217020113e-05}
{"train_loss": 0.2468312680721283, "global_step": 49159, "epoch": 552, "lr": 9.225084216487045e-05}
{"train_loss": 0.29620611667633057, "global_step": 49160, "epoch": 552, "lr": 9.225053215385965e-05}
{"train_loss": 0.30370891094207764, "global_step": 49161, "epoch": 552, "lr": 9.225022213716881e-05}
{"train_loss": 0.270927369594574, "global_step": 49162, "epoch": 552, "lr": 9.224991211479792e-05}
{"train_loss": 0.22007226943969727, "global_step": 49163, "epoch": 552, "lr": 9.224960208674705e-05}
{"train_loss": 0.30659133195877075, "global_step": 49164, "epoch": 552, "lr": 9.224929205301624e-05}
{"train_loss": 0.3010105490684509, "global_step": 49165, "epoch": 552, "lr": 9.224898201360553e-05}
{"train_loss": 0.2781922221183777, "global_step": 49166, "epoch": 552, "lr": 9.224867196851495e-05}
{"train_loss": 0.31902745366096497, "global_step": 49167, "epoch": 552, "lr": 9.224836191774455e-05}
{"train_loss": 0.27694302797317505, "global_step": 49168, "epoch": 552, "lr": 9.224805186129438e-05}
{"train_loss": 0.21658620238304138, "global_step": 49169, "epoch": 552, "lr": 9.224774179916448e-05}
{"train_loss": 0.23848700523376465, "global_step": 49170, "epoch": 552, "lr": 9.224743173135489e-05}
{"train_loss": 0.23753796517848969, "global_step": 49171, "epoch": 552, "lr": 9.224712165786563e-05}
{"train_loss": 0.2979947030544281, "global_step": 49172, "epoch": 552, "lr": 9.224681157869675e-05}
{"train_loss": 0.19892364740371704, "global_step": 49173, "epoch": 552, "lr": 9.224650149384833e-05}
{"train_loss": 0.2349788248538971, "global_step": 49174, "epoch": 552, "lr": 9.224619140332036e-05}
{"train_loss": 0.31815579533576965, "global_step": 49175, "epoch": 552, "lr": 9.224588130711291e-05}
{"train_loss": 0.2790830433368683, "global_step": 49176, "epoch": 552, "lr": 9.224557120522603e-05}
{"train_loss": 0.33066537976264954, "global_step": 49177, "epoch": 552, "lr": 9.224526109765974e-05}
{"train_loss": 0.23835442960262299, "global_step": 49178, "epoch": 552, "lr": 9.224495098441408e-05}
{"train_loss": 0.20272648334503174, "global_step": 49179, "epoch": 552, "lr": 9.224464086548911e-05}
{"train_loss": 0.27076226472854614, "global_step": 49180, "epoch": 552, "lr": 9.224433074088485e-05}
{"train_loss": 0.24540871381759644, "global_step": 49181, "epoch": 552, "lr": 9.224402061060138e-05}
{"train_loss": 0.26702970266342163, "global_step": 49182, "epoch": 552, "lr": 9.22437104746387e-05}
{"train_loss": 0.3245391249656677, "global_step": 49183, "epoch": 552, "lr": 9.224340033299687e-05}
{"train_loss": 0.2929512858390808, "global_step": 49184, "epoch": 552, "lr": 9.224309018567593e-05}
{"train_loss": 0.30973488092422485, "global_step": 49185, "epoch": 552, "lr": 9.224278003267592e-05}
{"train_loss": 0.257635235786438, "global_step": 49186, "epoch": 552, "lr": 9.224246987399689e-05}
{"train_loss": 0.23922693729400635, "global_step": 49187, "epoch": 552, "lr": 9.224215970963888e-05}
{"train_loss": 0.3147800862789154, "global_step": 49188, "epoch": 552, "lr": 9.224184953960191e-05}
{"train_loss": 0.1965787261724472, "global_step": 49189, "epoch": 552, "lr": 9.224153936388605e-05}
{"train_loss": 0.2315911501646042, "global_step": 49190, "epoch": 552, "lr": 9.224122918249132e-05}
{"train_loss": 0.33760717511177063, "global_step": 49191, "epoch": 552, "lr": 9.224091899541775e-05}
{"train_loss": 0.2478104680776596, "global_step": 49192, "epoch": 552, "lr": 9.224060880266545e-05}
{"train_loss": 0.25989028811454773, "global_step": 49193, "epoch": 552, "lr": 9.224029860423438e-05}
{"train_loss": 0.2510179579257965, "global_step": 49194, "epoch": 552, "lr": 9.223998840012464e-05}
{"train_loss": 0.3000800311565399, "global_step": 49195, "epoch": 552, "lr": 9.223967819033624e-05}
{"train_loss": 0.29049548506736755, "global_step": 49196, "epoch": 552, "lr": 9.223936797486923e-05}
{"train_loss": 0.3817610740661621, "global_step": 49197, "epoch": 552, "lr": 9.223905775372366e-05}
{"train_loss": 0.19783484935760498, "global_step": 49198, "epoch": 552, "lr": 9.223874752689953e-05}
{"train_loss": 0.3951018154621124, "global_step": 49199, "epoch": 552, "lr": 9.223843729439696e-05}
{"train_loss": 0.35218706727027893, "global_step": 49200, "epoch": 552, "lr": 9.223812705621591e-05}
{"train_loss": 0.32201722264289856, "global_step": 49201, "epoch": 552, "lr": 9.223781681235649e-05}
{"train_loss": 0.3195779025554657, "global_step": 49202, "epoch": 552, "lr": 9.223750656281869e-05}
{"train_loss": 0.27041417360305786, "global_step": 49203, "epoch": 552, "lr": 9.223719630760257e-05}
{"train_loss": 0.28219079971313477, "global_step": 49204, "epoch": 552, "lr": 9.223688604670819e-05}
{"train_loss": 0.3729633092880249, "global_step": 49205, "epoch": 552, "lr": 9.223657578013557e-05}
{"train_loss": 0.41511404514312744, "global_step": 49206, "epoch": 552, "lr": 9.223626550788475e-05}
{"train_loss": 0.2907736301422119, "global_step": 49207, "epoch": 552, "lr": 9.223595522995579e-05}
{"train_loss": 0.3432537913322449, "global_step": 49208, "epoch": 552, "lr": 9.223564494634871e-05}
{"train_loss": 0.3247649073600769, "global_step": 49209, "epoch": 552, "lr": 9.223533465706355e-05}
{"train_loss": 0.2830049991607666, "global_step": 49210, "epoch": 552, "lr": 9.223502436210039e-05}
{"train_loss": 0.25702768564224243, "global_step": 49211, "epoch": 552, "lr": 9.223471406145924e-05}
{"train_loss": 0.2916925251483917, "global_step": 49212, "epoch": 552, "lr": 9.223440375514015e-05}
{"train_loss": 0.35608547925949097, "global_step": 49213, "epoch": 552, "lr": 9.223409344314314e-05}
{"train_loss": 0.230585977435112, "global_step": 49214, "epoch": 552, "lr": 9.223378312546829e-05}
{"train_loss": 0.3014654815196991, "global_step": 49215, "epoch": 552, "lr": 9.223347280211561e-05}
{"train_loss": 0.28090239826882824, "global_step": 49216, "epoch": 552, "lr": 9.223316247308515e-05, "val_loss": 3.0367560386657715}
{"train_loss": 0.2132543921470642, "global_step": 49217, "epoch": 553, "lr": 9.223285213837699e-05}
{"train_loss": 0.23028220236301422, "global_step": 49218, "epoch": 553, "lr": 9.22325417979911e-05}
{"train_loss": 0.29957807064056396, "global_step": 49219, "epoch": 553, "lr": 9.223223145192758e-05}
{"train_loss": 0.29279497265815735, "global_step": 49220, "epoch": 553, "lr": 9.223192110018643e-05}
{"train_loss": 0.24913306534290314, "global_step": 49221, "epoch": 553, "lr": 9.223161074276774e-05}
{"train_loss": 0.4047175943851471, "global_step": 49222, "epoch": 553, "lr": 9.223130037967151e-05}
{"train_loss": 0.26332053542137146, "global_step": 49223, "epoch": 553, "lr": 9.22309900108978e-05}
{"train_loss": 0.2947380840778351, "global_step": 49224, "epoch": 553, "lr": 9.223067963644665e-05}
{"train_loss": 0.2551640570163727, "global_step": 49225, "epoch": 553, "lr": 9.22303692563181e-05}
{"train_loss": 0.31772682070732117, "global_step": 49226, "epoch": 553, "lr": 9.223005887051219e-05}
{"train_loss": 0.27275440096855164, "global_step": 49227, "epoch": 553, "lr": 9.222974847902896e-05}
{"train_loss": 0.3869982361793518, "global_step": 49228, "epoch": 553, "lr": 9.222943808186844e-05}
{"train_loss": 0.20607154071331024, "global_step": 49229, "epoch": 553, "lr": 9.222912767903072e-05}
{"train_loss": 0.28720590472221375, "global_step": 49230, "epoch": 553, "lr": 9.222881727051577e-05}
{"train_loss": 0.3681025505065918, "global_step": 49231, "epoch": 553, "lr": 9.22285068563237e-05}
{"train_loss": 0.31398317217826843, "global_step": 49232, "epoch": 553, "lr": 9.222819643645452e-05}
{"train_loss": 0.4591405987739563, "global_step": 49233, "epoch": 553, "lr": 9.222788601090826e-05}
{"train_loss": 0.28972896933555603, "global_step": 49234, "epoch": 553, "lr": 9.222757557968498e-05}
{"train_loss": 0.32340896129608154, "global_step": 49235, "epoch": 553, "lr": 9.222726514278472e-05}
{"train_loss": 0.24942268431186676, "global_step": 49236, "epoch": 553, "lr": 9.222695470020753e-05}
{"train_loss": 0.2734901010990143, "global_step": 49237, "epoch": 553, "lr": 9.222664425195342e-05}
{"train_loss": 0.3116784393787384, "global_step": 49238, "epoch": 553, "lr": 9.222633379802247e-05}
{"train_loss": 0.2854086458683014, "global_step": 49239, "epoch": 553, "lr": 9.222602333841469e-05}
{"train_loss": 0.37158605456352234, "global_step": 49240, "epoch": 553, "lr": 9.222571287313015e-05}
{"train_loss": 0.23502855002880096, "global_step": 49241, "epoch": 553, "lr": 9.222540240216886e-05}
{"train_loss": 0.3065602481365204, "global_step": 49242, "epoch": 553, "lr": 9.222509192553089e-05}
{"train_loss": 0.35658836364746094, "global_step": 49243, "epoch": 553, "lr": 9.222478144321626e-05}
{"train_loss": 0.24070203304290771, "global_step": 49244, "epoch": 553, "lr": 9.222447095522506e-05}
{"train_loss": 0.36157602071762085, "global_step": 49245, "epoch": 553, "lr": 9.222416046155727e-05}
{"train_loss": 0.30870845913887024, "global_step": 49246, "epoch": 553, "lr": 9.222384996221293e-05}
{"train_loss": 0.3504636287689209, "global_step": 49247, "epoch": 553, "lr": 9.222353945719213e-05}
{"train_loss": 0.3588865399360657, "global_step": 49248, "epoch": 553, "lr": 9.22232289464949e-05}
{"train_loss": 0.33366677165031433, "global_step": 49249, "epoch": 553, "lr": 9.222291843012126e-05}
{"train_loss": 0.2497854083776474, "global_step": 49250, "epoch": 553, "lr": 9.222260790807127e-05}
{"train_loss": 0.37639319896698, "global_step": 49251, "epoch": 553, "lr": 9.222229738034496e-05}
{"train_loss": 0.2623547613620758, "global_step": 49252, "epoch": 553, "lr": 9.222198684694238e-05}
{"train_loss": 0.3470001518726349, "global_step": 49253, "epoch": 553, "lr": 9.222167630786356e-05}
{"train_loss": 0.33332350850105286, "global_step": 49254, "epoch": 553, "lr": 9.222136576310856e-05}
{"train_loss": 0.24801144003868103, "global_step": 49255, "epoch": 553, "lr": 9.22210552126774e-05}
{"train_loss": 0.22661785781383514, "global_step": 49256, "epoch": 553, "lr": 9.222074465657015e-05}
{"train_loss": 0.2337912619113922, "global_step": 49257, "epoch": 553, "lr": 9.222043409478682e-05}
{"train_loss": 0.3093350827693939, "global_step": 49258, "epoch": 553, "lr": 9.222012352732747e-05}
{"train_loss": 0.3817042410373688, "global_step": 49259, "epoch": 553, "lr": 9.221981295419216e-05}
{"train_loss": 0.3884318470954895, "global_step": 49260, "epoch": 553, "lr": 9.221950237538088e-05}
{"train_loss": 0.35765767097473145, "global_step": 49261, "epoch": 553, "lr": 9.221919179089372e-05}
{"train_loss": 0.24886134266853333, "global_step": 49262, "epoch": 553, "lr": 9.221888120073072e-05}
{"train_loss": 0.37476882338523865, "global_step": 49263, "epoch": 553, "lr": 9.221857060489187e-05}
{"train_loss": 0.2566812336444855, "global_step": 49264, "epoch": 553, "lr": 9.221826000337727e-05}
{"train_loss": 0.30085211992263794, "global_step": 49265, "epoch": 553, "lr": 9.221794939618693e-05}
{"train_loss": 0.17902423441410065, "global_step": 49266, "epoch": 553, "lr": 9.22176387833209e-05}
{"train_loss": 0.3237266540527344, "global_step": 49267, "epoch": 553, "lr": 9.221732816477923e-05}
{"train_loss": 0.2777685821056366, "global_step": 49268, "epoch": 553, "lr": 9.221701754056195e-05}
{"train_loss": 0.27005088329315186, "global_step": 49269, "epoch": 553, "lr": 9.221670691066912e-05}
{"train_loss": 0.2657843828201294, "global_step": 49270, "epoch": 553, "lr": 9.221639627510076e-05}
{"train_loss": 0.32204028964042664, "global_step": 49271, "epoch": 553, "lr": 9.221608563385692e-05}
{"train_loss": 0.3103419840335846, "global_step": 49272, "epoch": 553, "lr": 9.221577498693764e-05}
{"train_loss": 0.17201396822929382, "global_step": 49273, "epoch": 553, "lr": 9.221546433434296e-05}
{"train_loss": 0.22427520155906677, "global_step": 49274, "epoch": 553, "lr": 9.221515367607294e-05}
{"train_loss": 0.2976553440093994, "global_step": 49275, "epoch": 553, "lr": 9.221484301212759e-05}
{"train_loss": 0.2170376479625702, "global_step": 49276, "epoch": 553, "lr": 9.221453234250697e-05}
{"train_loss": 0.3097061216831207, "global_step": 49277, "epoch": 553, "lr": 9.221422166721113e-05}
{"train_loss": 0.25870802998542786, "global_step": 49278, "epoch": 553, "lr": 9.221391098624009e-05}
{"train_loss": 0.26529669761657715, "global_step": 49279, "epoch": 553, "lr": 9.221360029959392e-05}
{"train_loss": 0.23323124647140503, "global_step": 49280, "epoch": 553, "lr": 9.221328960727263e-05}
{"train_loss": 0.18678945302963257, "global_step": 49281, "epoch": 553, "lr": 9.221297890927629e-05}
{"train_loss": 0.20366546511650085, "global_step": 49282, "epoch": 553, "lr": 9.221266820560492e-05}
{"train_loss": 0.23679877817630768, "global_step": 49283, "epoch": 553, "lr": 9.221235749625858e-05}
{"train_loss": 0.27503761649131775, "global_step": 49284, "epoch": 553, "lr": 9.221204678123731e-05}
{"train_loss": 0.3937619924545288, "global_step": 49285, "epoch": 553, "lr": 9.221173606054113e-05}
{"train_loss": 0.2624048590660095, "global_step": 49286, "epoch": 553, "lr": 9.22114253341701e-05}
{"train_loss": 0.30088070034980774, "global_step": 49287, "epoch": 553, "lr": 9.221111460212426e-05}
{"train_loss": 0.29810774326324463, "global_step": 49288, "epoch": 553, "lr": 9.221080386440364e-05}
{"train_loss": 0.22379399836063385, "global_step": 49289, "epoch": 553, "lr": 9.221049312100831e-05}
{"train_loss": 0.17903146147727966, "global_step": 49290, "epoch": 553, "lr": 9.221018237193828e-05}
{"train_loss": 0.20016027987003326, "global_step": 49291, "epoch": 553, "lr": 9.22098716171936e-05}
{"train_loss": 0.21993622183799744, "global_step": 49292, "epoch": 553, "lr": 9.220956085677435e-05}
{"train_loss": 0.23251865804195404, "global_step": 49293, "epoch": 553, "lr": 9.220925009068052e-05}
{"train_loss": 0.2184126228094101, "global_step": 49294, "epoch": 553, "lr": 9.220893931891216e-05}
{"train_loss": 0.24772164225578308, "global_step": 49295, "epoch": 553, "lr": 9.220862854146933e-05}
{"train_loss": 0.3704158067703247, "global_step": 49296, "epoch": 553, "lr": 9.220831775835206e-05}
{"train_loss": 0.28042227029800415, "global_step": 49297, "epoch": 553, "lr": 9.22080069695604e-05}
{"train_loss": 0.26982420682907104, "global_step": 49298, "epoch": 553, "lr": 9.22076961750944e-05}
{"train_loss": 0.354557603597641, "global_step": 49299, "epoch": 553, "lr": 9.220738537495408e-05}
{"train_loss": 0.2172997146844864, "global_step": 49300, "epoch": 553, "lr": 9.220707456913948e-05}
{"train_loss": 0.2126895934343338, "global_step": 49301, "epoch": 553, "lr": 9.220676375765067e-05}
{"train_loss": 0.30376484990119934, "global_step": 49302, "epoch": 553, "lr": 9.220645294048767e-05}
{"train_loss": 0.17299434542655945, "global_step": 49303, "epoch": 553, "lr": 9.220614211765052e-05}
{"train_loss": 0.2747787833213806, "global_step": 49304, "epoch": 553, "lr": 9.220583128913927e-05}
{"train_loss": 0.2846960363428244, "global_step": 49305, "epoch": 553, "lr": 9.220552045495397e-05, "val_loss": 3.123030185699463}
{"train_loss": 0.26720789074897766, "global_step": 49306, "epoch": 554, "lr": 9.220520961509464e-05}
{"train_loss": 0.28918296098709106, "global_step": 49307, "epoch": 554, "lr": 9.220489876956134e-05}
{"train_loss": 0.3492881655693054, "global_step": 49308, "epoch": 554, "lr": 9.220458791835411e-05}
{"train_loss": 0.33075910806655884, "global_step": 49309, "epoch": 554, "lr": 9.2204277061473e-05}
{"train_loss": 0.21896705031394958, "global_step": 49310, "epoch": 554, "lr": 9.2203966198918e-05}
{"train_loss": 0.2763778269290924, "global_step": 49311, "epoch": 554, "lr": 9.220365533068922e-05}
{"train_loss": 0.3214270770549774, "global_step": 49312, "epoch": 554, "lr": 9.220334445678667e-05}
{"train_loss": 0.2913885712623596, "global_step": 49313, "epoch": 554, "lr": 9.220303357721039e-05}
{"train_loss": 0.2663436830043793, "global_step": 49314, "epoch": 554, "lr": 9.220272269196042e-05}
{"train_loss": 0.22720777988433838, "global_step": 49315, "epoch": 554, "lr": 9.220241180103683e-05}
{"train_loss": 0.32267847657203674, "global_step": 49316, "epoch": 554, "lr": 9.220210090443961e-05}
{"train_loss": 0.27240875363349915, "global_step": 49317, "epoch": 554, "lr": 9.220179000216886e-05}
{"train_loss": 0.3090510666370392, "global_step": 49318, "epoch": 554, "lr": 9.220147909422459e-05}
{"train_loss": 0.18550525605678558, "global_step": 49319, "epoch": 554, "lr": 9.220116818060682e-05}
{"train_loss": 0.2592436671257019, "global_step": 49320, "epoch": 554, "lr": 9.220085726131564e-05}
{"train_loss": 0.25515875220298767, "global_step": 49321, "epoch": 554, "lr": 9.220054633635107e-05}
{"train_loss": 0.2685362696647644, "global_step": 49322, "epoch": 554, "lr": 9.220023540571314e-05}
{"train_loss": 0.3163362741470337, "global_step": 49323, "epoch": 554, "lr": 9.219992446940191e-05}
{"train_loss": 0.22923046350479126, "global_step": 49324, "epoch": 554, "lr": 9.219961352741742e-05}
{"train_loss": 0.34995073080062866, "global_step": 49325, "epoch": 554, "lr": 9.219930257975969e-05}
{"train_loss": 0.3656907081604004, "global_step": 49326, "epoch": 554, "lr": 9.219899162642878e-05}
{"train_loss": 0.20318153500556946, "global_step": 49327, "epoch": 554, "lr": 9.219868066742474e-05}
{"train_loss": 0.41281944513320923, "global_step": 49328, "epoch": 554, "lr": 9.21983697027476e-05}
{"train_loss": 0.279392272233963, "global_step": 49329, "epoch": 554, "lr": 9.219805873239739e-05}
{"train_loss": 0.22487033903598785, "global_step": 49330, "epoch": 554, "lr": 9.219774775637419e-05}
{"train_loss": 0.33220207691192627, "global_step": 49331, "epoch": 554, "lr": 9.2197436774678e-05}
{"train_loss": 0.21566364169120789, "global_step": 49332, "epoch": 554, "lr": 9.219712578730888e-05}
{"train_loss": 0.29425621032714844, "global_step": 49333, "epoch": 554, "lr": 9.219681479426688e-05}
{"train_loss": 0.24270252883434296, "global_step": 49334, "epoch": 554, "lr": 9.219650379555203e-05}
{"train_loss": 0.2717570960521698, "global_step": 49335, "epoch": 554, "lr": 9.219619279116438e-05}
{"train_loss": 0.25575578212738037, "global_step": 49336, "epoch": 554, "lr": 9.219588178110394e-05}
{"train_loss": 0.2613726854324341, "global_step": 49337, "epoch": 554, "lr": 9.219557076537079e-05}
{"train_loss": 0.34389033913612366, "global_step": 49338, "epoch": 554, "lr": 9.219525974396496e-05}
{"train_loss": 0.18365532159805298, "global_step": 49339, "epoch": 554, "lr": 9.219494871688651e-05}
{"train_loss": 0.26382479071617126, "global_step": 49340, "epoch": 554, "lr": 9.219463768413545e-05}
{"train_loss": 0.19482557475566864, "global_step": 49341, "epoch": 554, "lr": 9.219432664571184e-05}
{"train_loss": 0.3216160535812378, "global_step": 49342, "epoch": 554, "lr": 9.219401560161571e-05}
{"train_loss": 0.267312228679657, "global_step": 49343, "epoch": 554, "lr": 9.219370455184711e-05}
{"train_loss": 0.26138901710510254, "global_step": 49344, "epoch": 554, "lr": 9.219339349640608e-05}
{"train_loss": 0.30025437474250793, "global_step": 49345, "epoch": 554, "lr": 9.219308243529266e-05}
{"train_loss": 0.3310946822166443, "global_step": 49346, "epoch": 554, "lr": 9.21927713685069e-05}
{"train_loss": 0.1992015242576599, "global_step": 49347, "epoch": 554, "lr": 9.219246029604885e-05}
{"train_loss": 0.29305070638656616, "global_step": 49348, "epoch": 554, "lr": 9.219214921791851e-05}
{"train_loss": 0.21891914308071136, "global_step": 49349, "epoch": 554, "lr": 9.219183813411596e-05}
{"train_loss": 0.3350248634815216, "global_step": 49350, "epoch": 554, "lr": 9.219152704464123e-05}
{"train_loss": 0.2717030942440033, "global_step": 49351, "epoch": 554, "lr": 9.219121594949438e-05}
{"train_loss": 0.3493705987930298, "global_step": 49352, "epoch": 554, "lr": 9.219090484867542e-05}
{"train_loss": 0.39149051904678345, "global_step": 49353, "epoch": 554, "lr": 9.21905937421844e-05}
{"train_loss": 0.3338145315647125, "global_step": 49354, "epoch": 554, "lr": 9.219028263002138e-05}
{"train_loss": 0.11057352274656296, "global_step": 49355, "epoch": 554, "lr": 9.21899715121864e-05}
{"train_loss": 0.1643296331167221, "global_step": 49356, "epoch": 554, "lr": 9.218966038867947e-05}
{"train_loss": 0.32303959131240845, "global_step": 49357, "epoch": 554, "lr": 9.218934925950067e-05}
{"train_loss": 0.33626338839530945, "global_step": 49358, "epoch": 554, "lr": 9.218903812465002e-05}
{"train_loss": 0.3222866952419281, "global_step": 49359, "epoch": 554, "lr": 9.218872698412757e-05}
{"train_loss": 0.23745781183242798, "global_step": 49360, "epoch": 554, "lr": 9.218841583793336e-05}
{"train_loss": 0.27731412649154663, "global_step": 49361, "epoch": 554, "lr": 9.218810468606745e-05}
{"train_loss": 0.20966671407222748, "global_step": 49362, "epoch": 554, "lr": 9.218779352852984e-05}
{"train_loss": 0.2514974772930145, "global_step": 49363, "epoch": 554, "lr": 9.21874823653206e-05}
{"train_loss": 0.31223028898239136, "global_step": 49364, "epoch": 554, "lr": 9.218717119643976e-05}
{"train_loss": 0.3195206820964813, "global_step": 49365, "epoch": 554, "lr": 9.218686002188738e-05}
{"train_loss": 0.284293532371521, "global_step": 49366, "epoch": 554, "lr": 9.218654884166348e-05}
{"train_loss": 0.2745353579521179, "global_step": 49367, "epoch": 554, "lr": 9.218623765576813e-05}
{"train_loss": 0.19944019615650177, "global_step": 49368, "epoch": 554, "lr": 9.218592646420135e-05}
{"train_loss": 0.2521081268787384, "global_step": 49369, "epoch": 554, "lr": 9.218561526696318e-05}
{"train_loss": 0.2695639431476593, "global_step": 49370, "epoch": 554, "lr": 9.218530406405367e-05}
{"train_loss": 0.4638220965862274, "global_step": 49371, "epoch": 554, "lr": 9.218499285547287e-05}
{"train_loss": 0.2595641314983368, "global_step": 49372, "epoch": 554, "lr": 9.218468164122079e-05}
{"train_loss": 0.19665086269378662, "global_step": 49373, "epoch": 554, "lr": 9.21843704212975e-05}
{"train_loss": 0.3323492109775543, "global_step": 49374, "epoch": 554, "lr": 9.218405919570304e-05}
{"train_loss": 0.2578524947166443, "global_step": 49375, "epoch": 554, "lr": 9.218374796443745e-05}
{"train_loss": 0.271673321723938, "global_step": 49376, "epoch": 554, "lr": 9.218343672750077e-05}
{"train_loss": 0.3070374131202698, "global_step": 49377, "epoch": 554, "lr": 9.218312548489305e-05}
{"train_loss": 0.3437192738056183, "global_step": 49378, "epoch": 554, "lr": 9.21828142366143e-05}
{"train_loss": 0.3775879442691803, "global_step": 49379, "epoch": 554, "lr": 9.21825029826646e-05}
{"train_loss": 0.29190996289253235, "global_step": 49380, "epoch": 554, "lr": 9.218219172304397e-05}
{"train_loss": 0.32108885049819946, "global_step": 49381, "epoch": 554, "lr": 9.218188045775247e-05}
{"train_loss": 0.3250989615917206, "global_step": 49382, "epoch": 554, "lr": 9.218156918679011e-05}
{"train_loss": 0.2602165639400482, "global_step": 49383, "epoch": 554, "lr": 9.218125791015697e-05}
{"train_loss": 0.3010605573654175, "global_step": 49384, "epoch": 554, "lr": 9.218094662785306e-05}
{"train_loss": 0.3430100679397583, "global_step": 49385, "epoch": 554, "lr": 9.218063533987845e-05}
{"train_loss": 0.3001621663570404, "global_step": 49386, "epoch": 554, "lr": 9.218032404623315e-05}
{"train_loss": 0.3884549140930176, "global_step": 49387, "epoch": 554, "lr": 9.218001274691724e-05}
{"train_loss": 0.29087862372398376, "global_step": 49388, "epoch": 554, "lr": 9.217970144193073e-05}
{"train_loss": 0.24854867160320282, "global_step": 49389, "epoch": 554, "lr": 9.217939013127368e-05}
{"train_loss": 0.3040739595890045, "global_step": 49390, "epoch": 554, "lr": 9.217907881494612e-05}
{"train_loss": 0.3481771945953369, "global_step": 49391, "epoch": 554, "lr": 9.21787674929481e-05}
{"train_loss": 0.40752771496772766, "global_step": 49392, "epoch": 554, "lr": 9.217845616527965e-05}
{"train_loss": 0.2764662802219391, "global_step": 49393, "epoch": 554, "lr": 9.217814483194083e-05}
{"train_loss": 0.28589448575558285, "global_step": 49394, "epoch": 554, "lr": 9.217783349293168e-05, "val_loss": 3.00704026222229}
{"train_loss": 0.23309612274169922, "global_step": 49395, "epoch": 555, "lr": 9.217752214825222e-05}
{"train_loss": 0.3123222887516022, "global_step": 49396, "epoch": 555, "lr": 9.217721079790251e-05}
{"train_loss": 0.3059275448322296, "global_step": 49397, "epoch": 555, "lr": 9.217689944188259e-05}
{"train_loss": 0.2954694330692291, "global_step": 49398, "epoch": 555, "lr": 9.21765880801925e-05}
{"train_loss": 0.37430673837661743, "global_step": 49399, "epoch": 555, "lr": 9.217627671283227e-05}
{"train_loss": 0.2503129243850708, "global_step": 49400, "epoch": 555, "lr": 9.217596533980197e-05}
{"train_loss": 0.24635565280914307, "global_step": 49401, "epoch": 555, "lr": 9.217565396110161e-05}
{"train_loss": 0.23800154030323029, "global_step": 49402, "epoch": 555, "lr": 9.217534257673126e-05}
{"train_loss": 0.29335469007492065, "global_step": 49403, "epoch": 555, "lr": 9.217503118669096e-05}
{"train_loss": 0.29036134481430054, "global_step": 49404, "epoch": 555, "lr": 9.217471979098071e-05}
{"train_loss": 0.3414267301559448, "global_step": 49405, "epoch": 555, "lr": 9.217440838960061e-05}
{"train_loss": 0.25027668476104736, "global_step": 49406, "epoch": 555, "lr": 9.217409698255067e-05}
{"train_loss": 0.19834600389003754, "global_step": 49407, "epoch": 555, "lr": 9.217378556983092e-05}
{"train_loss": 0.30321958661079407, "global_step": 49408, "epoch": 555, "lr": 9.217347415144143e-05}
{"train_loss": 0.2563731372356415, "global_step": 49409, "epoch": 555, "lr": 9.217316272738223e-05}
{"train_loss": 0.2637413442134857, "global_step": 49410, "epoch": 555, "lr": 9.217285129765337e-05}
{"train_loss": 0.22525139153003693, "global_step": 49411, "epoch": 555, "lr": 9.217253986225487e-05}
{"train_loss": 0.3131738007068634, "global_step": 49412, "epoch": 555, "lr": 9.217222842118679e-05}
{"train_loss": 0.1845436543226242, "global_step": 49413, "epoch": 555, "lr": 9.217191697444918e-05}
{"train_loss": 0.3671680986881256, "global_step": 49414, "epoch": 555, "lr": 9.217160552204205e-05}
{"train_loss": 0.2567947804927826, "global_step": 49415, "epoch": 555, "lr": 9.217129406396548e-05}
{"train_loss": 0.31016841530799866, "global_step": 49416, "epoch": 555, "lr": 9.217098260021949e-05}
{"train_loss": 0.2815605700016022, "global_step": 49417, "epoch": 555, "lr": 9.217067113080411e-05}
{"train_loss": 0.3491857349872589, "global_step": 49418, "epoch": 555, "lr": 9.217035965571941e-05}
{"train_loss": 0.2614906132221222, "global_step": 49419, "epoch": 555, "lr": 9.217004817496542e-05}
{"train_loss": 0.2574908435344696, "global_step": 49420, "epoch": 555, "lr": 9.216973668854218e-05}
{"train_loss": 0.37582045793533325, "global_step": 49421, "epoch": 555, "lr": 9.216942519644974e-05}
{"train_loss": 0.29084229469299316, "global_step": 49422, "epoch": 555, "lr": 9.216911369868811e-05}
{"train_loss": 0.3392963409423828, "global_step": 49423, "epoch": 555, "lr": 9.216880219525738e-05}
{"train_loss": 0.17462538182735443, "global_step": 49424, "epoch": 555, "lr": 9.216849068615756e-05}
{"train_loss": 0.307033896446228, "global_step": 49425, "epoch": 555, "lr": 9.21681791713887e-05}
{"train_loss": 0.22458568215370178, "global_step": 49426, "epoch": 555, "lr": 9.216786765095084e-05}
{"train_loss": 0.3017495572566986, "global_step": 49427, "epoch": 555, "lr": 9.216755612484402e-05}
{"train_loss": 0.37518537044525146, "global_step": 49428, "epoch": 555, "lr": 9.21672445930683e-05}
{"train_loss": 0.3625798225402832, "global_step": 49429, "epoch": 555, "lr": 9.21669330556237e-05}
{"train_loss": 0.30529817938804626, "global_step": 49430, "epoch": 555, "lr": 9.216662151251028e-05}
{"train_loss": 0.39885061979293823, "global_step": 49431, "epoch": 555, "lr": 9.216630996372806e-05}
{"train_loss": 0.3080427348613739, "global_step": 49432, "epoch": 555, "lr": 9.21659984092771e-05}
{"train_loss": 0.21727854013442993, "global_step": 49433, "epoch": 555, "lr": 9.216568684915743e-05}
{"train_loss": 0.28705158829689026, "global_step": 49434, "epoch": 555, "lr": 9.21653752833691e-05}
{"train_loss": 0.30503740906715393, "global_step": 49435, "epoch": 555, "lr": 9.216506371191215e-05}
{"train_loss": 0.24877925217151642, "global_step": 49436, "epoch": 555, "lr": 9.216475213478662e-05}
{"train_loss": 0.2512340545654297, "global_step": 49437, "epoch": 555, "lr": 9.216444055199256e-05}
{"train_loss": 0.2552950978279114, "global_step": 49438, "epoch": 555, "lr": 9.216412896353e-05}
{"train_loss": 0.30764150619506836, "global_step": 49439, "epoch": 555, "lr": 9.216381736939898e-05}
{"train_loss": 0.3367898464202881, "global_step": 49440, "epoch": 555, "lr": 9.216350576959956e-05}
{"train_loss": 0.2099504917860031, "global_step": 49441, "epoch": 555, "lr": 9.216319416413175e-05}
{"train_loss": 0.2948956787586212, "global_step": 49442, "epoch": 555, "lr": 9.216288255299563e-05}
{"train_loss": 0.18975311517715454, "global_step": 49443, "epoch": 555, "lr": 9.216257093619123e-05}
{"train_loss": 0.26820865273475647, "global_step": 49444, "epoch": 555, "lr": 9.216225931371857e-05}
{"train_loss": 0.3631507158279419, "global_step": 49445, "epoch": 555, "lr": 9.21619476855777e-05}
{"train_loss": 0.30993759632110596, "global_step": 49446, "epoch": 555, "lr": 9.21616360517687e-05}
{"train_loss": 0.2850833237171173, "global_step": 49447, "epoch": 555, "lr": 9.216132441229156e-05}
{"train_loss": 0.3283761143684387, "global_step": 49448, "epoch": 555, "lr": 9.216101276714635e-05}
{"train_loss": 0.37228846549987793, "global_step": 49449, "epoch": 555, "lr": 9.216070111633311e-05}
{"train_loss": 0.296846479177475, "global_step": 49450, "epoch": 555, "lr": 9.216038945985189e-05}
{"train_loss": 0.2967666983604431, "global_step": 49451, "epoch": 555, "lr": 9.216007779770269e-05}
{"train_loss": 0.3480041027069092, "global_step": 49452, "epoch": 555, "lr": 9.215976612988559e-05}
{"train_loss": 0.33706361055374146, "global_step": 49453, "epoch": 555, "lr": 9.215945445640064e-05}
{"train_loss": 0.28104397654533386, "global_step": 49454, "epoch": 555, "lr": 9.215914277724784e-05}
{"train_loss": 0.3397923707962036, "global_step": 49455, "epoch": 555, "lr": 9.215883109242728e-05}
{"train_loss": 0.3476274907588959, "global_step": 49456, "epoch": 555, "lr": 9.215851940193896e-05}
{"train_loss": 0.25756922364234924, "global_step": 49457, "epoch": 555, "lr": 9.215820770578296e-05}
{"train_loss": 0.18880578875541687, "global_step": 49458, "epoch": 555, "lr": 9.215789600395929e-05}
{"train_loss": 0.27671676874160767, "global_step": 49459, "epoch": 555, "lr": 9.215758429646801e-05}
{"train_loss": 0.31718143820762634, "global_step": 49460, "epoch": 555, "lr": 9.215727258330915e-05}
{"train_loss": 0.2542508542537689, "global_step": 49461, "epoch": 555, "lr": 9.215696086448276e-05}
{"train_loss": 0.2900473475456238, "global_step": 49462, "epoch": 555, "lr": 9.215664913998889e-05}
{"train_loss": 0.33457818627357483, "global_step": 49463, "epoch": 555, "lr": 9.215633740982757e-05}
{"train_loss": 0.29286113381385803, "global_step": 49464, "epoch": 555, "lr": 9.215602567399883e-05}
{"train_loss": 0.3385085165500641, "global_step": 49465, "epoch": 555, "lr": 9.215571393250274e-05}
{"train_loss": 0.3552435338497162, "global_step": 49466, "epoch": 555, "lr": 9.215540218533933e-05}
{"train_loss": 0.2251933366060257, "global_step": 49467, "epoch": 555, "lr": 9.215509043250862e-05}
{"train_loss": 0.28150540590286255, "global_step": 49468, "epoch": 555, "lr": 9.21547786740107e-05}
{"train_loss": 0.37055113911628723, "global_step": 49469, "epoch": 555, "lr": 9.215446690984557e-05}
{"train_loss": 0.22167617082595825, "global_step": 49470, "epoch": 555, "lr": 9.215415514001329e-05}
{"train_loss": 0.3160885274410248, "global_step": 49471, "epoch": 555, "lr": 9.21538433645139e-05}
{"train_loss": 0.3508259654045105, "global_step": 49472, "epoch": 555, "lr": 9.215353158334742e-05}
{"train_loss": 0.33842986822128296, "global_step": 49473, "epoch": 555, "lr": 9.215321979651393e-05}
{"train_loss": 0.25929057598114014, "global_step": 49474, "epoch": 555, "lr": 9.215290800401344e-05}
{"train_loss": 0.2979245185852051, "global_step": 49475, "epoch": 555, "lr": 9.215259620584603e-05}
{"train_loss": 0.3853774666786194, "global_step": 49476, "epoch": 555, "lr": 9.215228440201169e-05}
{"train_loss": 0.24698549509048462, "global_step": 49477, "epoch": 555, "lr": 9.21519725925105e-05}
{"train_loss": 0.31962794065475464, "global_step": 49478, "epoch": 555, "lr": 9.215166077734249e-05}
{"train_loss": 0.2586171329021454, "global_step": 49479, "epoch": 555, "lr": 9.21513489565077e-05}
{"train_loss": 0.3752152621746063, "global_step": 49480, "epoch": 555, "lr": 9.215103713000617e-05}
{"train_loss": 0.2736159563064575, "global_step": 49481, "epoch": 555, "lr": 9.215072529783796e-05}
{"train_loss": 0.3208960294723511, "global_step": 49482, "epoch": 555, "lr": 9.21504134600031e-05}
{"train_loss": 0.2951715802878476, "global_step": 49483, "epoch": 555, "lr": 9.215010161650162e-05, "val_loss": 3.107287645339966, "train_action_mse_error": 32.92238235473633}
{"train_loss": 0.3883083164691925, "global_step": 49484, "epoch": 556, "lr": 9.214978976733358e-05}
{"train_loss": 0.30215513706207275, "global_step": 49485, "epoch": 556, "lr": 9.2149477912499e-05}
{"train_loss": 0.3407362699508667, "global_step": 49486, "epoch": 556, "lr": 9.214916605199795e-05}
{"train_loss": 0.2759373188018799, "global_step": 49487, "epoch": 556, "lr": 9.214885418583045e-05}
{"train_loss": 0.3086079955101013, "global_step": 49488, "epoch": 556, "lr": 9.214854231399657e-05}
{"train_loss": 0.2787632942199707, "global_step": 49489, "epoch": 556, "lr": 9.214823043649631e-05}
{"train_loss": 0.29359865188598633, "global_step": 49490, "epoch": 556, "lr": 9.214791855332975e-05}
{"train_loss": 0.33916446566581726, "global_step": 49491, "epoch": 556, "lr": 9.214760666449692e-05}
{"train_loss": 0.32215961813926697, "global_step": 49492, "epoch": 556, "lr": 9.214729476999784e-05}
{"train_loss": 0.35401052236557007, "global_step": 49493, "epoch": 556, "lr": 9.214698286983257e-05}
{"train_loss": 0.2956547439098358, "global_step": 49494, "epoch": 556, "lr": 9.214667096400116e-05}
{"train_loss": 0.33723652362823486, "global_step": 49495, "epoch": 556, "lr": 9.214635905250364e-05}
{"train_loss": 0.26472267508506775, "global_step": 49496, "epoch": 556, "lr": 9.214604713534006e-05}
{"train_loss": 0.26874953508377075, "global_step": 49497, "epoch": 556, "lr": 9.214573521251047e-05}
{"train_loss": 0.30942028760910034, "global_step": 49498, "epoch": 556, "lr": 9.214542328401487e-05}
{"train_loss": 0.2892984449863434, "global_step": 49499, "epoch": 556, "lr": 9.214511134985335e-05}
{"train_loss": 0.23666810989379883, "global_step": 49500, "epoch": 556, "lr": 9.214479941002595e-05}
{"train_loss": 0.2722799777984619, "global_step": 49501, "epoch": 556, "lr": 9.214448746453266e-05}
{"train_loss": 0.37125164270401, "global_step": 49502, "epoch": 556, "lr": 9.214417551337357e-05}
{"train_loss": 0.2403556853532791, "global_step": 49503, "epoch": 556, "lr": 9.214386355654871e-05}
{"train_loss": 0.31744384765625, "global_step": 49504, "epoch": 556, "lr": 9.214355159405812e-05}
{"train_loss": 0.36804264783859253, "global_step": 49505, "epoch": 556, "lr": 9.214323962590186e-05}
{"train_loss": 0.20574475824832916, "global_step": 49506, "epoch": 556, "lr": 9.214292765207993e-05}
{"train_loss": 0.2644137144088745, "global_step": 49507, "epoch": 556, "lr": 9.214261567259242e-05}
{"train_loss": 0.34383177757263184, "global_step": 49508, "epoch": 556, "lr": 9.214230368743933e-05}
{"train_loss": 0.274020791053772, "global_step": 49509, "epoch": 556, "lr": 9.214199169662073e-05}
{"train_loss": 0.24468332529067993, "global_step": 49510, "epoch": 556, "lr": 9.214167970013665e-05}
{"train_loss": 0.2524246275424957, "global_step": 49511, "epoch": 556, "lr": 9.214136769798714e-05}
{"train_loss": 0.24102827906608582, "global_step": 49512, "epoch": 556, "lr": 9.214105569017224e-05}
{"train_loss": 0.29407721757888794, "global_step": 49513, "epoch": 556, "lr": 9.214074367669197e-05}
{"train_loss": 0.31179726123809814, "global_step": 49514, "epoch": 556, "lr": 9.21404316575464e-05}
{"train_loss": 0.3367297947406769, "global_step": 49515, "epoch": 556, "lr": 9.214011963273557e-05}
{"train_loss": 0.3146982491016388, "global_step": 49516, "epoch": 556, "lr": 9.213980760225952e-05}
{"train_loss": 0.23235879838466644, "global_step": 49517, "epoch": 556, "lr": 9.213949556611827e-05}
{"train_loss": 0.27389150857925415, "global_step": 49518, "epoch": 556, "lr": 9.213918352431188e-05}
{"train_loss": 0.21362356841564178, "global_step": 49519, "epoch": 556, "lr": 9.213887147684039e-05}
{"train_loss": 0.2358168512582779, "global_step": 49520, "epoch": 556, "lr": 9.213855942370386e-05}
{"train_loss": 0.2700119912624359, "global_step": 49521, "epoch": 556, "lr": 9.21382473649023e-05}
{"train_loss": 0.32543572783470154, "global_step": 49522, "epoch": 556, "lr": 9.213793530043578e-05}
{"train_loss": 0.22787834703922272, "global_step": 49523, "epoch": 556, "lr": 9.213762323030431e-05}
{"train_loss": 0.25764575600624084, "global_step": 49524, "epoch": 556, "lr": 9.213731115450795e-05}
{"train_loss": 0.23495370149612427, "global_step": 49525, "epoch": 556, "lr": 9.213699907304675e-05}
{"train_loss": 0.32804691791534424, "global_step": 49526, "epoch": 556, "lr": 9.213668698592073e-05}
{"train_loss": 0.25792181491851807, "global_step": 49527, "epoch": 556, "lr": 9.213637489312997e-05}
{"train_loss": 0.21190984547138214, "global_step": 49528, "epoch": 556, "lr": 9.213606279467448e-05}
{"train_loss": 0.3130626380443573, "global_step": 49529, "epoch": 556, "lr": 9.21357506905543e-05}
{"train_loss": 0.28134360909461975, "global_step": 49530, "epoch": 556, "lr": 9.213543858076949e-05}
{"train_loss": 0.2638837397098541, "global_step": 49531, "epoch": 556, "lr": 9.213512646532008e-05}
{"train_loss": 0.2331741601228714, "global_step": 49532, "epoch": 556, "lr": 9.213481434420612e-05}
{"train_loss": 0.2754218280315399, "global_step": 49533, "epoch": 556, "lr": 9.213450221742765e-05}
{"train_loss": 0.3818613290786743, "global_step": 49534, "epoch": 556, "lr": 9.21341900849847e-05}
{"train_loss": 0.31855660676956177, "global_step": 49535, "epoch": 556, "lr": 9.213387794687733e-05}
{"train_loss": 0.18029211461544037, "global_step": 49536, "epoch": 556, "lr": 9.213356580310556e-05}
{"train_loss": 0.31792783737182617, "global_step": 49537, "epoch": 556, "lr": 9.213325365366946e-05}
{"train_loss": 0.40825480222702026, "global_step": 49538, "epoch": 556, "lr": 9.213294149856905e-05}
{"train_loss": 0.3288082480430603, "global_step": 49539, "epoch": 556, "lr": 9.213262933780438e-05}
{"train_loss": 0.2587505877017975, "global_step": 49540, "epoch": 556, "lr": 9.21323171713755e-05}
{"train_loss": 0.28404197096824646, "global_step": 49541, "epoch": 556, "lr": 9.213200499928245e-05}
{"train_loss": 0.2819465398788452, "global_step": 49542, "epoch": 556, "lr": 9.213169282152525e-05}
{"train_loss": 0.31762373447418213, "global_step": 49543, "epoch": 556, "lr": 9.213138063810397e-05}
{"train_loss": 0.24831560254096985, "global_step": 49544, "epoch": 556, "lr": 9.213106844901862e-05}
{"train_loss": 0.2955814003944397, "global_step": 49545, "epoch": 556, "lr": 9.213075625426926e-05}
{"train_loss": 0.23544366657733917, "global_step": 49546, "epoch": 556, "lr": 9.213044405385596e-05}
{"train_loss": 0.25547856092453003, "global_step": 49547, "epoch": 556, "lr": 9.213013184777872e-05}
{"train_loss": 0.26588085293769836, "global_step": 49548, "epoch": 556, "lr": 9.21298196360376e-05}
{"train_loss": 0.29359620809555054, "global_step": 49549, "epoch": 556, "lr": 9.212950741863262e-05}
{"train_loss": 0.22798393666744232, "global_step": 49550, "epoch": 556, "lr": 9.212919519556387e-05}
{"train_loss": 0.30249252915382385, "global_step": 49551, "epoch": 556, "lr": 9.212888296683135e-05}
{"train_loss": 0.25442346930503845, "global_step": 49552, "epoch": 556, "lr": 9.212857073243512e-05}
{"train_loss": 0.33456721901893616, "global_step": 49553, "epoch": 556, "lr": 9.21282584923752e-05}
{"train_loss": 0.2564506232738495, "global_step": 49554, "epoch": 556, "lr": 9.212794624665167e-05}
{"train_loss": 0.31907543540000916, "global_step": 49555, "epoch": 556, "lr": 9.212763399526454e-05}
{"train_loss": 0.34506478905677795, "global_step": 49556, "epoch": 556, "lr": 9.212732173821386e-05}
{"train_loss": 0.30877405405044556, "global_step": 49557, "epoch": 556, "lr": 9.21270094754997e-05}
{"train_loss": 0.2307586669921875, "global_step": 49558, "epoch": 556, "lr": 9.212669720712206e-05}
{"train_loss": 0.18335652351379395, "global_step": 49559, "epoch": 556, "lr": 9.2126384933081e-05}
{"train_loss": 0.38976773619651794, "global_step": 49560, "epoch": 556, "lr": 9.212607265337655e-05}
{"train_loss": 0.222954660654068, "global_step": 49561, "epoch": 556, "lr": 9.212576036800877e-05}
{"train_loss": 0.3143353760242462, "global_step": 49562, "epoch": 556, "lr": 9.212544807697769e-05}
{"train_loss": 0.2665110230445862, "global_step": 49563, "epoch": 556, "lr": 9.212513578028337e-05}
{"train_loss": 0.3345038890838623, "global_step": 49564, "epoch": 556, "lr": 9.212482347792584e-05}
{"train_loss": 0.23052482306957245, "global_step": 49565, "epoch": 556, "lr": 9.212451116990512e-05}
{"train_loss": 0.3687741160392761, "global_step": 49566, "epoch": 556, "lr": 9.21241988562213e-05}
{"train_loss": 0.2438788115978241, "global_step": 49567, "epoch": 556, "lr": 9.212388653687439e-05}
{"train_loss": 0.20113982260227203, "global_step": 49568, "epoch": 556, "lr": 9.212357421186442e-05}
{"train_loss": 0.3464272916316986, "global_step": 49569, "epoch": 556, "lr": 9.212326188119144e-05}
{"train_loss": 0.2807755470275879, "global_step": 49570, "epoch": 556, "lr": 9.212294954485554e-05}
{"train_loss": 0.26246127486228943, "global_step": 49571, "epoch": 556, "lr": 9.212263720285669e-05}
{"train_loss": 0.28688884735777137, "global_step": 49572, "epoch": 556, "lr": 9.212232485519497e-05, "val_loss": 3.012869358062744}
{"train_loss": 0.2481539100408554, "global_step": 49573, "epoch": 557, "lr": 9.212201250187043e-05}
{"train_loss": 0.3077174723148346, "global_step": 49574, "epoch": 557, "lr": 9.212170014288309e-05}
{"train_loss": 0.3392372727394104, "global_step": 49575, "epoch": 557, "lr": 9.212138777823301e-05}
{"train_loss": 0.35798609256744385, "global_step": 49576, "epoch": 557, "lr": 9.21210754079202e-05}
{"train_loss": 0.33159491419792175, "global_step": 49577, "epoch": 557, "lr": 9.212076303194475e-05}
{"train_loss": 0.3209269046783447, "global_step": 49578, "epoch": 557, "lr": 9.212045065030667e-05}
{"train_loss": 0.2615959346294403, "global_step": 49579, "epoch": 557, "lr": 9.2120138263006e-05}
{"train_loss": 0.3267214894294739, "global_step": 49580, "epoch": 557, "lr": 9.21198258700428e-05}
{"train_loss": 0.30562111735343933, "global_step": 49581, "epoch": 557, "lr": 9.21195134714171e-05}
{"train_loss": 0.3706733286380768, "global_step": 49582, "epoch": 557, "lr": 9.211920106712895e-05}
{"train_loss": 0.274056077003479, "global_step": 49583, "epoch": 557, "lr": 9.211888865717838e-05}
{"train_loss": 0.28061819076538086, "global_step": 49584, "epoch": 557, "lr": 9.211857624156544e-05}
{"train_loss": 0.20531901717185974, "global_step": 49585, "epoch": 557, "lr": 9.211826382029016e-05}
{"train_loss": 0.23424941301345825, "global_step": 49586, "epoch": 557, "lr": 9.211795139335261e-05}
{"train_loss": 0.3719353675842285, "global_step": 49587, "epoch": 557, "lr": 9.211763896075281e-05}
{"train_loss": 0.18602019548416138, "global_step": 49588, "epoch": 557, "lr": 9.21173265224908e-05}
{"train_loss": 0.21467725932598114, "global_step": 49589, "epoch": 557, "lr": 9.211701407856663e-05}
{"train_loss": 0.2904057502746582, "global_step": 49590, "epoch": 557, "lr": 9.211670162898034e-05}
{"train_loss": 0.3002125918865204, "global_step": 49591, "epoch": 557, "lr": 9.211638917373197e-05}
{"train_loss": 0.23325373232364655, "global_step": 49592, "epoch": 557, "lr": 9.211607671282158e-05}
{"train_loss": 0.2402082234621048, "global_step": 49593, "epoch": 557, "lr": 9.211576424624919e-05}
{"train_loss": 0.24280408024787903, "global_step": 49594, "epoch": 557, "lr": 9.211545177401485e-05}
{"train_loss": 0.33066514134407043, "global_step": 49595, "epoch": 557, "lr": 9.21151392961186e-05}
{"train_loss": 0.21303781867027283, "global_step": 49596, "epoch": 557, "lr": 9.211482681256047e-05}
{"train_loss": 0.25402042269706726, "global_step": 49597, "epoch": 557, "lr": 9.211451432334053e-05}
{"train_loss": 0.25118955969810486, "global_step": 49598, "epoch": 557, "lr": 9.21142018284588e-05}
{"train_loss": 0.3158853352069855, "global_step": 49599, "epoch": 557, "lr": 9.211388932791534e-05}
{"train_loss": 0.30514174699783325, "global_step": 49600, "epoch": 557, "lr": 9.211357682171016e-05}
{"train_loss": 0.17058026790618896, "global_step": 49601, "epoch": 557, "lr": 9.211326430984334e-05}
{"train_loss": 0.22909800708293915, "global_step": 49602, "epoch": 557, "lr": 9.21129517923149e-05}
{"train_loss": 0.32265669107437134, "global_step": 49603, "epoch": 557, "lr": 9.211263926912489e-05}
{"train_loss": 0.2777617275714874, "global_step": 49604, "epoch": 557, "lr": 9.211232674027335e-05}
{"train_loss": 0.25626879930496216, "global_step": 49605, "epoch": 557, "lr": 9.211201420576032e-05}
{"train_loss": 0.23748716711997986, "global_step": 49606, "epoch": 557, "lr": 9.211170166558584e-05}
{"train_loss": 0.18825186789035797, "global_step": 49607, "epoch": 557, "lr": 9.211138911974994e-05}
{"train_loss": 0.3739205598831177, "global_step": 49608, "epoch": 557, "lr": 9.211107656825271e-05}
{"train_loss": 0.27244332432746887, "global_step": 49609, "epoch": 557, "lr": 9.211076401109413e-05}
{"train_loss": 0.2315243035554886, "global_step": 49610, "epoch": 557, "lr": 9.211045144827428e-05}
{"train_loss": 0.2634769082069397, "global_step": 49611, "epoch": 557, "lr": 9.21101388797932e-05}
{"train_loss": 0.31124013662338257, "global_step": 49612, "epoch": 557, "lr": 9.210982630565093e-05}
{"train_loss": 0.23666098713874817, "global_step": 49613, "epoch": 557, "lr": 9.210951372584749e-05}
{"train_loss": 0.19277311861515045, "global_step": 49614, "epoch": 557, "lr": 9.210920114038295e-05}
{"train_loss": 0.24852938950061798, "global_step": 49615, "epoch": 557, "lr": 9.210888854925733e-05}
{"train_loss": 0.2995254099369049, "global_step": 49616, "epoch": 557, "lr": 9.210857595247069e-05}
{"train_loss": 0.2115877866744995, "global_step": 49617, "epoch": 557, "lr": 9.210826335002307e-05}
{"train_loss": 0.25698715448379517, "global_step": 49618, "epoch": 557, "lr": 9.21079507419145e-05}
{"train_loss": 0.19718466699123383, "global_step": 49619, "epoch": 557, "lr": 9.210763812814503e-05}
{"train_loss": 0.1711144596338272, "global_step": 49620, "epoch": 557, "lr": 9.21073255087147e-05}
{"train_loss": 0.21574480831623077, "global_step": 49621, "epoch": 557, "lr": 9.210701288362356e-05}
{"train_loss": 0.2814984619617462, "global_step": 49622, "epoch": 557, "lr": 9.210670025287164e-05}
{"train_loss": 0.2563318610191345, "global_step": 49623, "epoch": 557, "lr": 9.2106387616459e-05}
{"train_loss": 0.3113885223865509, "global_step": 49624, "epoch": 557, "lr": 9.210607497438566e-05}
{"train_loss": 0.32334449887275696, "global_step": 49625, "epoch": 557, "lr": 9.210576232665167e-05}
{"train_loss": 0.3378983736038208, "global_step": 49626, "epoch": 557, "lr": 9.210544967325707e-05}
{"train_loss": 0.2833535969257355, "global_step": 49627, "epoch": 557, "lr": 9.210513701420192e-05}
{"train_loss": 0.3210039734840393, "global_step": 49628, "epoch": 557, "lr": 9.210482434948621e-05}
{"train_loss": 0.2736876606941223, "global_step": 49629, "epoch": 557, "lr": 9.210451167911006e-05}
{"train_loss": 0.15912285447120667, "global_step": 49630, "epoch": 557, "lr": 9.210419900307346e-05}
{"train_loss": 0.3114020824432373, "global_step": 49631, "epoch": 557, "lr": 9.210388632137645e-05}
{"train_loss": 0.21181516349315643, "global_step": 49632, "epoch": 557, "lr": 9.210357363401912e-05}
{"train_loss": 0.3793100118637085, "global_step": 49633, "epoch": 557, "lr": 9.210326094100144e-05}
{"train_loss": 0.1807159185409546, "global_step": 49634, "epoch": 557, "lr": 9.210294824232351e-05}
{"train_loss": 0.26420608162879944, "global_step": 49635, "epoch": 557, "lr": 9.210263553798535e-05}
{"train_loss": 0.2256559580564499, "global_step": 49636, "epoch": 557, "lr": 9.210232282798699e-05}
{"train_loss": 0.2823037803173065, "global_step": 49637, "epoch": 557, "lr": 9.21020101123285e-05}
{"train_loss": 0.23322580754756927, "global_step": 49638, "epoch": 557, "lr": 9.21016973910099e-05}
{"train_loss": 0.29027244448661804, "global_step": 49639, "epoch": 557, "lr": 9.210138466403124e-05}
{"train_loss": 0.21776923537254333, "global_step": 49640, "epoch": 557, "lr": 9.210107193139257e-05}
{"train_loss": 0.2539316713809967, "global_step": 49641, "epoch": 557, "lr": 9.210075919309392e-05}
{"train_loss": 0.23225057125091553, "global_step": 49642, "epoch": 557, "lr": 9.210044644913534e-05}
{"train_loss": 0.2465582638978958, "global_step": 49643, "epoch": 557, "lr": 9.210013369951686e-05}
{"train_loss": 0.2505365312099457, "global_step": 49644, "epoch": 557, "lr": 9.209982094423855e-05}
{"train_loss": 0.3122081160545349, "global_step": 49645, "epoch": 557, "lr": 9.20995081833004e-05}
{"train_loss": 0.29325687885284424, "global_step": 49646, "epoch": 557, "lr": 9.209919541670251e-05}
{"train_loss": 0.21297921240329742, "global_step": 49647, "epoch": 557, "lr": 9.209888264444489e-05}
{"train_loss": 0.23055166006088257, "global_step": 49648, "epoch": 557, "lr": 9.209856986652758e-05}
{"train_loss": 0.23700743913650513, "global_step": 49649, "epoch": 557, "lr": 9.209825708295063e-05}
{"train_loss": 0.28277674317359924, "global_step": 49650, "epoch": 557, "lr": 9.209794429371409e-05}
{"train_loss": 0.2996472418308258, "global_step": 49651, "epoch": 557, "lr": 9.2097631498818e-05}
{"train_loss": 0.29141858220100403, "global_step": 49652, "epoch": 557, "lr": 9.209731869826238e-05}
{"train_loss": 0.32867249846458435, "global_step": 49653, "epoch": 557, "lr": 9.20970058920473e-05}
{"train_loss": 0.3042255938053131, "global_step": 49654, "epoch": 557, "lr": 9.20966930801728e-05}
{"train_loss": 0.2565251886844635, "global_step": 49655, "epoch": 557, "lr": 9.209638026263889e-05}
{"train_loss": 0.2490730881690979, "global_step": 49656, "epoch": 557, "lr": 9.209606743944564e-05}
{"train_loss": 0.21308554708957672, "global_step": 49657, "epoch": 557, "lr": 9.209575461059309e-05}
{"train_loss": 0.25978702306747437, "global_step": 49658, "epoch": 557, "lr": 9.20954417760813e-05}
{"train_loss": 0.28063347935676575, "global_step": 49659, "epoch": 557, "lr": 9.209512893591025e-05}
{"train_loss": 0.2910456657409668, "global_step": 49660, "epoch": 557, "lr": 9.209481609008007e-05}
{"train_loss": 0.2673967807145601, "global_step": 49661, "epoch": 557, "lr": 9.209450323859072e-05, "val_loss": 3.1002743244171143}
{"train_loss": 0.21173034608364105, "global_step": 49662, "epoch": 558, "lr": 9.209419038144229e-05}
{"train_loss": 0.22247783839702606, "global_step": 49663, "epoch": 558, "lr": 9.209387751863481e-05}
{"train_loss": 0.26436227560043335, "global_step": 49664, "epoch": 558, "lr": 9.209356465016832e-05}
{"train_loss": 0.2621992230415344, "global_step": 49665, "epoch": 558, "lr": 9.209325177604286e-05}
{"train_loss": 0.22891944646835327, "global_step": 49666, "epoch": 558, "lr": 9.209293889625847e-05}
{"train_loss": 0.26293686032295227, "global_step": 49667, "epoch": 558, "lr": 9.20926260108152e-05}
{"train_loss": 0.29525357484817505, "global_step": 49668, "epoch": 558, "lr": 9.209231311971309e-05}
{"train_loss": 0.23161792755126953, "global_step": 49669, "epoch": 558, "lr": 9.209200022295219e-05}
{"train_loss": 0.3145485520362854, "global_step": 49670, "epoch": 558, "lr": 9.209168732053253e-05}
{"train_loss": 0.2666880488395691, "global_step": 49671, "epoch": 558, "lr": 9.209137441245416e-05}
{"train_loss": 0.26806405186653137, "global_step": 49672, "epoch": 558, "lr": 9.209106149871711e-05}
{"train_loss": 0.270235538482666, "global_step": 49673, "epoch": 558, "lr": 9.209074857932142e-05}
{"train_loss": 0.2473142296075821, "global_step": 49674, "epoch": 558, "lr": 9.209043565426716e-05}
{"train_loss": 0.36092469096183777, "global_step": 49675, "epoch": 558, "lr": 9.209012272355435e-05}
{"train_loss": 0.2546689510345459, "global_step": 49676, "epoch": 558, "lr": 9.208980978718303e-05}
{"train_loss": 0.25663936138153076, "global_step": 49677, "epoch": 558, "lr": 9.208949684515325e-05}
{"train_loss": 0.33076995611190796, "global_step": 49678, "epoch": 558, "lr": 9.208918389746506e-05}
{"train_loss": 0.3133851885795593, "global_step": 49679, "epoch": 558, "lr": 9.208887094411849e-05}
{"train_loss": 0.27333635091781616, "global_step": 49680, "epoch": 558, "lr": 9.208855798511358e-05}
{"train_loss": 0.2513454556465149, "global_step": 49681, "epoch": 558, "lr": 9.208824502045037e-05}
{"train_loss": 0.2347169667482376, "global_step": 49682, "epoch": 558, "lr": 9.20879320501289e-05}
{"train_loss": 0.30772659182548523, "global_step": 49683, "epoch": 558, "lr": 9.208761907414924e-05}
{"train_loss": 0.372100830078125, "global_step": 49684, "epoch": 558, "lr": 9.208730609251142e-05}
{"train_loss": 0.32154780626296997, "global_step": 49685, "epoch": 558, "lr": 9.208699310521545e-05}
{"train_loss": 0.18522994220256805, "global_step": 49686, "epoch": 558, "lr": 9.20866801122614e-05}
{"train_loss": 0.31943926215171814, "global_step": 49687, "epoch": 558, "lr": 9.208636711364933e-05}
{"train_loss": 0.2927621901035309, "global_step": 49688, "epoch": 558, "lr": 9.208605410937923e-05}
{"train_loss": 0.268810510635376, "global_step": 49689, "epoch": 558, "lr": 9.20857410994512e-05}
{"train_loss": 0.2359788864850998, "global_step": 49690, "epoch": 558, "lr": 9.208542808386524e-05}
{"train_loss": 0.24055615067481995, "global_step": 49691, "epoch": 558, "lr": 9.208511506262141e-05}
{"train_loss": 0.3873538374900818, "global_step": 49692, "epoch": 558, "lr": 9.208480203571976e-05}
{"train_loss": 0.26662343740463257, "global_step": 49693, "epoch": 558, "lr": 9.20844890031603e-05}
{"train_loss": 0.28577253222465515, "global_step": 49694, "epoch": 558, "lr": 9.208417596494312e-05}
{"train_loss": 0.3354308307170868, "global_step": 49695, "epoch": 558, "lr": 9.208386292106822e-05}
{"train_loss": 0.26350945234298706, "global_step": 49696, "epoch": 558, "lr": 9.208354987153567e-05}
{"train_loss": 0.26962214708328247, "global_step": 49697, "epoch": 558, "lr": 9.208323681634547e-05}
{"train_loss": 0.2516024112701416, "global_step": 49698, "epoch": 558, "lr": 9.208292375549772e-05}
{"train_loss": 0.2655292749404907, "global_step": 49699, "epoch": 558, "lr": 9.208261068899242e-05}
{"train_loss": 0.2756645083427429, "global_step": 49700, "epoch": 558, "lr": 9.208229761682964e-05}
{"train_loss": 0.29881566762924194, "global_step": 49701, "epoch": 558, "lr": 9.20819845390094e-05}
{"train_loss": 0.31880563497543335, "global_step": 49702, "epoch": 558, "lr": 9.208167145553175e-05}
{"train_loss": 0.3511377274990082, "global_step": 49703, "epoch": 558, "lr": 9.208135836639673e-05}
{"train_loss": 0.2674250602722168, "global_step": 49704, "epoch": 558, "lr": 9.208104527160438e-05}
{"train_loss": 0.2601591646671295, "global_step": 49705, "epoch": 558, "lr": 9.208073217115476e-05}
{"train_loss": 0.28652042150497437, "global_step": 49706, "epoch": 558, "lr": 9.208041906504788e-05}
{"train_loss": 0.20889361202716827, "global_step": 49707, "epoch": 558, "lr": 9.208010595328381e-05}
{"train_loss": 0.25277501344680786, "global_step": 49708, "epoch": 558, "lr": 9.207979283586259e-05}
{"train_loss": 0.24963755905628204, "global_step": 49709, "epoch": 558, "lr": 9.207947971278425e-05}
{"train_loss": 0.3749583959579468, "global_step": 49710, "epoch": 558, "lr": 9.207916658404883e-05}
{"train_loss": 0.29587092995643616, "global_step": 49711, "epoch": 558, "lr": 9.207885344965639e-05}
{"train_loss": 0.31075334548950195, "global_step": 49712, "epoch": 558, "lr": 9.207854030960695e-05}
{"train_loss": 0.22892361879348755, "global_step": 49713, "epoch": 558, "lr": 9.207822716390056e-05}
{"train_loss": 0.35471418499946594, "global_step": 49714, "epoch": 558, "lr": 9.207791401253728e-05}
{"train_loss": 0.28480270504951477, "global_step": 49715, "epoch": 558, "lr": 9.207760085551714e-05}
{"train_loss": 0.2563621401786804, "global_step": 49716, "epoch": 558, "lr": 9.207728769284017e-05}
{"train_loss": 0.25906699895858765, "global_step": 49717, "epoch": 558, "lr": 9.207697452450641e-05}
{"train_loss": 0.26250943541526794, "global_step": 49718, "epoch": 558, "lr": 9.207666135051593e-05}
{"train_loss": 0.2575473487377167, "global_step": 49719, "epoch": 558, "lr": 9.207634817086874e-05}
{"train_loss": 0.23674225807189941, "global_step": 49720, "epoch": 558, "lr": 9.207603498556492e-05}
{"train_loss": 0.2960911989212036, "global_step": 49721, "epoch": 558, "lr": 9.207572179460448e-05}
{"train_loss": 0.3410397171974182, "global_step": 49722, "epoch": 558, "lr": 9.207540859798746e-05}
{"train_loss": 0.2758345901966095, "global_step": 49723, "epoch": 558, "lr": 9.207509539571393e-05}
{"train_loss": 0.22937442362308502, "global_step": 49724, "epoch": 558, "lr": 9.207478218778392e-05}
{"train_loss": 0.26105478405952454, "global_step": 49725, "epoch": 558, "lr": 9.207446897419745e-05}
{"train_loss": 0.30893075466156006, "global_step": 49726, "epoch": 558, "lr": 9.207415575495459e-05}
{"train_loss": 0.18394751846790314, "global_step": 49727, "epoch": 558, "lr": 9.207384253005538e-05}
{"train_loss": 0.3927406966686249, "global_step": 49728, "epoch": 558, "lr": 9.207352929949984e-05}
{"train_loss": 0.2702644467353821, "global_step": 49729, "epoch": 558, "lr": 9.207321606328803e-05}
{"train_loss": 0.23217171430587769, "global_step": 49730, "epoch": 558, "lr": 9.207290282142e-05}
{"train_loss": 0.3483302593231201, "global_step": 49731, "epoch": 558, "lr": 9.207258957389578e-05}
{"train_loss": 0.2739764451980591, "global_step": 49732, "epoch": 558, "lr": 9.20722763207154e-05}
{"train_loss": 0.30728328227996826, "global_step": 49733, "epoch": 558, "lr": 9.207196306187892e-05}
{"train_loss": 0.385149210691452, "global_step": 49734, "epoch": 558, "lr": 9.207164979738638e-05}
{"train_loss": 0.3048759996891022, "global_step": 49735, "epoch": 558, "lr": 9.207133652723782e-05}
{"train_loss": 0.33480724692344666, "global_step": 49736, "epoch": 558, "lr": 9.207102325143327e-05}
{"train_loss": 0.29388776421546936, "global_step": 49737, "epoch": 558, "lr": 9.20707099699728e-05}
{"train_loss": 0.23463335633277893, "global_step": 49738, "epoch": 558, "lr": 9.207039668285643e-05}
{"train_loss": 0.3017672896385193, "global_step": 49739, "epoch": 558, "lr": 9.20700833900842e-05}
{"train_loss": 0.3253760039806366, "global_step": 49740, "epoch": 558, "lr": 9.206977009165618e-05}
{"train_loss": 0.2682759463787079, "global_step": 49741, "epoch": 558, "lr": 9.206945678757237e-05}
{"train_loss": 0.2021288275718689, "global_step": 49742, "epoch": 558, "lr": 9.206914347783285e-05}
{"train_loss": 0.29189103841781616, "global_step": 49743, "epoch": 558, "lr": 9.206883016243764e-05}
{"train_loss": 0.20131823420524597, "global_step": 49744, "epoch": 558, "lr": 9.20685168413868e-05}
{"train_loss": 0.24595461785793304, "global_step": 49745, "epoch": 558, "lr": 9.206820351468034e-05}
{"train_loss": 0.31941893696784973, "global_step": 49746, "epoch": 558, "lr": 9.206789018231833e-05}
{"train_loss": 0.3122437298297882, "global_step": 49747, "epoch": 558, "lr": 9.206757684430081e-05}
{"train_loss": 0.340146541595459, "global_step": 49748, "epoch": 558, "lr": 9.206726350062782e-05}
{"train_loss": 0.36633655428886414, "global_step": 49749, "epoch": 558, "lr": 9.206695015129939e-05}
{"train_loss": 0.2814327279168568, "global_step": 49750, "epoch": 558, "lr": 9.206663679631557e-05, "val_loss": 2.963848114013672}
{"train_loss": 0.18660865724086761, "global_step": 49751, "epoch": 559, "lr": 9.20663234356764e-05}
{"train_loss": 0.24849933385849, "global_step": 49752, "epoch": 559, "lr": 9.206601006938194e-05}
{"train_loss": 0.2726927697658539, "global_step": 49753, "epoch": 559, "lr": 9.206569669743221e-05}
{"train_loss": 0.18134856224060059, "global_step": 49754, "epoch": 559, "lr": 9.206538331982728e-05}
{"train_loss": 0.3428008258342743, "global_step": 49755, "epoch": 559, "lr": 9.206506993656714e-05}
{"train_loss": 0.2240559160709381, "global_step": 49756, "epoch": 559, "lr": 9.206475654765188e-05}
{"train_loss": 0.348232239484787, "global_step": 49757, "epoch": 559, "lr": 9.206444315308152e-05}
{"train_loss": 0.31987887620925903, "global_step": 49758, "epoch": 559, "lr": 9.206412975285612e-05}
{"train_loss": 0.24668216705322266, "global_step": 49759, "epoch": 559, "lr": 9.20638163469757e-05}
{"train_loss": 0.34535667300224304, "global_step": 49760, "epoch": 559, "lr": 9.206350293544031e-05}
{"train_loss": 0.33413055539131165, "global_step": 49761, "epoch": 559, "lr": 9.206318951825e-05}
{"train_loss": 0.23117683827877045, "global_step": 49762, "epoch": 559, "lr": 9.20628760954048e-05}
{"train_loss": 0.2504255473613739, "global_step": 49763, "epoch": 559, "lr": 9.206256266690477e-05}
{"train_loss": 0.23832286894321442, "global_step": 49764, "epoch": 559, "lr": 9.206224923274994e-05}
{"train_loss": 0.2245381474494934, "global_step": 49765, "epoch": 559, "lr": 9.206193579294036e-05}
{"train_loss": 0.32228758931159973, "global_step": 49766, "epoch": 559, "lr": 9.206162234747604e-05}
{"train_loss": 0.2558825612068176, "global_step": 49767, "epoch": 559, "lr": 9.206130889635707e-05}
{"train_loss": 0.2865833342075348, "global_step": 49768, "epoch": 559, "lr": 9.206099543958346e-05}
{"train_loss": 0.32370325922966003, "global_step": 49769, "epoch": 559, "lr": 9.206068197715527e-05}
{"train_loss": 0.306034117937088, "global_step": 49770, "epoch": 559, "lr": 9.206036850907252e-05}
{"train_loss": 0.29166489839553833, "global_step": 49771, "epoch": 559, "lr": 9.206005503533528e-05}
{"train_loss": 0.2385091930627823, "global_step": 49772, "epoch": 559, "lr": 9.205974155594357e-05}
{"train_loss": 0.30534279346466064, "global_step": 49773, "epoch": 559, "lr": 9.205942807089745e-05}
{"train_loss": 0.21916988492012024, "global_step": 49774, "epoch": 559, "lr": 9.205911458019694e-05}
{"train_loss": 0.27933987975120544, "global_step": 49775, "epoch": 559, "lr": 9.20588010838421e-05}
{"train_loss": 0.31953370571136475, "global_step": 49776, "epoch": 559, "lr": 9.205848758183296e-05}
{"train_loss": 0.2692979872226715, "global_step": 49777, "epoch": 559, "lr": 9.20581740741696e-05}
{"train_loss": 0.3839106857776642, "global_step": 49778, "epoch": 559, "lr": 9.205786056085199e-05}
{"train_loss": 0.2819197475910187, "global_step": 49779, "epoch": 559, "lr": 9.205754704188024e-05}
{"train_loss": 0.284116268157959, "global_step": 49780, "epoch": 559, "lr": 9.205723351725435e-05}
{"train_loss": 0.29406723380088806, "global_step": 49781, "epoch": 559, "lr": 9.205691998697438e-05}
{"train_loss": 0.33261212706565857, "global_step": 49782, "epoch": 559, "lr": 9.205660645104037e-05}
{"train_loss": 0.2995298504829407, "global_step": 49783, "epoch": 559, "lr": 9.205629290945236e-05}
{"train_loss": 0.3065996468067169, "global_step": 49784, "epoch": 559, "lr": 9.20559793622104e-05}
{"train_loss": 0.22097475826740265, "global_step": 49785, "epoch": 559, "lr": 9.205566580931453e-05}
{"train_loss": 0.2188781350851059, "global_step": 49786, "epoch": 559, "lr": 9.205535225076478e-05}
{"train_loss": 0.1850624680519104, "global_step": 49787, "epoch": 559, "lr": 9.205503868656121e-05}
{"train_loss": 0.3032322824001312, "global_step": 49788, "epoch": 559, "lr": 9.205472511670385e-05}
{"train_loss": 0.34931984543800354, "global_step": 49789, "epoch": 559, "lr": 9.205441154119274e-05}
{"train_loss": 0.3593273162841797, "global_step": 49790, "epoch": 559, "lr": 9.205409796002793e-05}
{"train_loss": 0.2551285922527313, "global_step": 49791, "epoch": 559, "lr": 9.205378437320944e-05}
{"train_loss": 0.284734308719635, "global_step": 49792, "epoch": 559, "lr": 9.205347078073736e-05}
{"train_loss": 0.24040944874286652, "global_step": 49793, "epoch": 559, "lr": 9.205315718261169e-05}
{"train_loss": 0.26320186257362366, "global_step": 49794, "epoch": 559, "lr": 9.205284357883249e-05}
{"train_loss": 0.2003180831670761, "global_step": 49795, "epoch": 559, "lr": 9.205252996939979e-05}
{"train_loss": 0.3233405649662018, "global_step": 49796, "epoch": 559, "lr": 9.205221635431365e-05}
{"train_loss": 0.3256763219833374, "global_step": 49797, "epoch": 559, "lr": 9.20519027335741e-05}
{"train_loss": 0.3464040756225586, "global_step": 49798, "epoch": 559, "lr": 9.205158910718119e-05}
{"train_loss": 0.19030722975730896, "global_step": 49799, "epoch": 559, "lr": 9.205127547513493e-05}
{"train_loss": 0.21091654896736145, "global_step": 49800, "epoch": 559, "lr": 9.205096183743541e-05}
{"train_loss": 0.2469482421875, "global_step": 49801, "epoch": 559, "lr": 9.205064819408264e-05}
{"train_loss": 0.21001170575618744, "global_step": 49802, "epoch": 559, "lr": 9.205033454507669e-05}
{"train_loss": 0.29046136140823364, "global_step": 49803, "epoch": 559, "lr": 9.205002089041757e-05}
{"train_loss": 0.2691795527935028, "global_step": 49804, "epoch": 559, "lr": 9.204970723010536e-05}
{"train_loss": 0.28680309653282166, "global_step": 49805, "epoch": 559, "lr": 9.204939356414004e-05}
{"train_loss": 0.2884086072444916, "global_step": 49806, "epoch": 559, "lr": 9.204907989252172e-05}
{"train_loss": 0.2491888403892517, "global_step": 49807, "epoch": 559, "lr": 9.204876621525041e-05}
{"train_loss": 0.36931312084198, "global_step": 49808, "epoch": 559, "lr": 9.204845253232614e-05}
{"train_loss": 0.29857122898101807, "global_step": 49809, "epoch": 559, "lr": 9.204813884374898e-05}
{"train_loss": 0.2631407678127289, "global_step": 49810, "epoch": 559, "lr": 9.204782514951898e-05}
{"train_loss": 0.1921733170747757, "global_step": 49811, "epoch": 559, "lr": 9.204751144963614e-05}
{"train_loss": 0.28912633657455444, "global_step": 49812, "epoch": 559, "lr": 9.204719774410052e-05}
{"train_loss": 0.23326623439788818, "global_step": 49813, "epoch": 559, "lr": 9.204688403291217e-05}
{"train_loss": 0.2162352055311203, "global_step": 49814, "epoch": 559, "lr": 9.204657031607114e-05}
{"train_loss": 0.2525269687175751, "global_step": 49815, "epoch": 559, "lr": 9.204625659357746e-05}
{"train_loss": 0.30795079469680786, "global_step": 49816, "epoch": 559, "lr": 9.204594286543116e-05}
{"train_loss": 0.24160011112689972, "global_step": 49817, "epoch": 559, "lr": 9.20456291316323e-05}
{"train_loss": 0.2835763096809387, "global_step": 49818, "epoch": 559, "lr": 9.204531539218094e-05}
{"train_loss": 0.203080415725708, "global_step": 49819, "epoch": 559, "lr": 9.204500164707708e-05}
{"train_loss": 0.2644897699356079, "global_step": 49820, "epoch": 559, "lr": 9.204468789632077e-05}
{"train_loss": 0.26241353154182434, "global_step": 49821, "epoch": 559, "lr": 9.204437413991209e-05}
{"train_loss": 0.25784575939178467, "global_step": 49822, "epoch": 559, "lr": 9.204406037785102e-05}
{"train_loss": 0.2843647301197052, "global_step": 49823, "epoch": 559, "lr": 9.204374661013769e-05}
{"train_loss": 0.2822335958480835, "global_step": 49824, "epoch": 559, "lr": 9.204343283677205e-05}
{"train_loss": 0.23759956657886505, "global_step": 49825, "epoch": 559, "lr": 9.20431190577542e-05}
{"train_loss": 0.33366698026657104, "global_step": 49826, "epoch": 559, "lr": 9.204280527308416e-05}
{"train_loss": 0.33265313506126404, "global_step": 49827, "epoch": 559, "lr": 9.204249148276197e-05}
{"train_loss": 0.19925402104854584, "global_step": 49828, "epoch": 559, "lr": 9.204217768678768e-05}
{"train_loss": 0.2161134034395218, "global_step": 49829, "epoch": 559, "lr": 9.204186388516135e-05}
{"train_loss": 0.28063690662384033, "global_step": 49830, "epoch": 559, "lr": 9.204155007788299e-05}
{"train_loss": 0.24829214811325073, "global_step": 49831, "epoch": 559, "lr": 9.204123626495265e-05}
{"train_loss": 0.30263766646385193, "global_step": 49832, "epoch": 559, "lr": 9.20409224463704e-05}
{"train_loss": 0.28342485427856445, "global_step": 49833, "epoch": 559, "lr": 9.204060862213624e-05}
{"train_loss": 0.1695583611726761, "global_step": 49834, "epoch": 559, "lr": 9.204029479225024e-05}
{"train_loss": 0.24937932193279266, "global_step": 49835, "epoch": 559, "lr": 9.203998095671244e-05}
{"train_loss": 0.28301820158958435, "global_step": 49836, "epoch": 559, "lr": 9.203966711552288e-05}
{"train_loss": 0.22880364954471588, "global_step": 49837, "epoch": 559, "lr": 9.203935326868158e-05}
{"train_loss": 0.2843272089958191, "global_step": 49838, "epoch": 559, "lr": 9.20390394161886e-05}
{"train_loss": 0.270646263709229, "global_step": 49839, "epoch": 559, "lr": 9.2038725558044e-05, "val_loss": 3.0520379543304443}
{"train_loss": 0.28829124569892883, "global_step": 49840, "epoch": 560, "lr": 9.20384116942478e-05}
{"train_loss": 0.24889157712459564, "global_step": 49841, "epoch": 560, "lr": 9.203809782480006e-05}
{"train_loss": 0.2163248211145401, "global_step": 49842, "epoch": 560, "lr": 9.20377839497008e-05}
{"train_loss": 0.2830646336078644, "global_step": 49843, "epoch": 560, "lr": 9.203747006895007e-05}
{"train_loss": 0.19806784391403198, "global_step": 49844, "epoch": 560, "lr": 9.20371561825479e-05}
{"train_loss": 0.3393852412700653, "global_step": 49845, "epoch": 560, "lr": 9.203684229049439e-05}
{"train_loss": 0.271139532327652, "global_step": 49846, "epoch": 560, "lr": 9.20365283927895e-05}
{"train_loss": 0.33218711614608765, "global_step": 49847, "epoch": 560, "lr": 9.203621448943332e-05}
{"train_loss": 0.23562917113304138, "global_step": 49848, "epoch": 560, "lr": 9.203590058042589e-05}
{"train_loss": 0.21216712892055511, "global_step": 49849, "epoch": 560, "lr": 9.203558666576723e-05}
{"train_loss": 0.2658593952655792, "global_step": 49850, "epoch": 560, "lr": 9.203527274545743e-05}
{"train_loss": 0.31167465448379517, "global_step": 49851, "epoch": 560, "lr": 9.203495881949648e-05}
{"train_loss": 0.2513868808746338, "global_step": 49852, "epoch": 560, "lr": 9.203464488788443e-05}
{"train_loss": 0.3232804238796234, "global_step": 49853, "epoch": 560, "lr": 9.203433095062135e-05}
{"train_loss": 0.3544967770576477, "global_step": 49854, "epoch": 560, "lr": 9.203401700770726e-05}
{"train_loss": 0.2725331485271454, "global_step": 49855, "epoch": 560, "lr": 9.203370305914221e-05}
{"train_loss": 0.24670051038265228, "global_step": 49856, "epoch": 560, "lr": 9.203338910492625e-05}
{"train_loss": 0.1716296672821045, "global_step": 49857, "epoch": 560, "lr": 9.20330751450594e-05}
{"train_loss": 0.2575565576553345, "global_step": 49858, "epoch": 560, "lr": 9.203276117954171e-05}
{"train_loss": 0.32975730299949646, "global_step": 49859, "epoch": 560, "lr": 9.203244720837325e-05}
{"train_loss": 0.20493900775909424, "global_step": 49860, "epoch": 560, "lr": 9.203213323155403e-05}
{"train_loss": 0.2533276379108429, "global_step": 49861, "epoch": 560, "lr": 9.20318192490841e-05}
{"train_loss": 0.29158204793930054, "global_step": 49862, "epoch": 560, "lr": 9.203150526096351e-05}
{"train_loss": 0.2658725380897522, "global_step": 49863, "epoch": 560, "lr": 9.203119126719228e-05}
{"train_loss": 0.3411104083061218, "global_step": 49864, "epoch": 560, "lr": 9.203087726777048e-05}
{"train_loss": 0.3450365662574768, "global_step": 49865, "epoch": 560, "lr": 9.203056326269815e-05}
{"train_loss": 0.21600663661956787, "global_step": 49866, "epoch": 560, "lr": 9.20302492519753e-05}
{"train_loss": 0.2963753938674927, "global_step": 49867, "epoch": 560, "lr": 9.202993523560201e-05}
{"train_loss": 0.18332768976688385, "global_step": 49868, "epoch": 560, "lr": 9.202962121357831e-05}
{"train_loss": 0.30517706274986267, "global_step": 49869, "epoch": 560, "lr": 9.202930718590423e-05}
{"train_loss": 0.35577869415283203, "global_step": 49870, "epoch": 560, "lr": 9.202899315257984e-05}
{"train_loss": 0.2725759744644165, "global_step": 49871, "epoch": 560, "lr": 9.202867911360515e-05}
{"train_loss": 0.3134267032146454, "global_step": 49872, "epoch": 560, "lr": 9.202836506898021e-05}
{"train_loss": 0.4378811717033386, "global_step": 49873, "epoch": 560, "lr": 9.202805101870508e-05}
{"train_loss": 0.27123790979385376, "global_step": 49874, "epoch": 560, "lr": 9.202773696277979e-05}
{"train_loss": 0.29538244009017944, "global_step": 49875, "epoch": 560, "lr": 9.202742290120438e-05}
{"train_loss": 0.20516790449619293, "global_step": 49876, "epoch": 560, "lr": 9.202710883397889e-05}
{"train_loss": 0.2647460997104645, "global_step": 49877, "epoch": 560, "lr": 9.202679476110338e-05}
{"train_loss": 0.26421236991882324, "global_step": 49878, "epoch": 560, "lr": 9.202648068257786e-05}
{"train_loss": 0.24269536137580872, "global_step": 49879, "epoch": 560, "lr": 9.202616659840242e-05}
{"train_loss": 0.36745375394821167, "global_step": 49880, "epoch": 560, "lr": 9.202585250857705e-05}
{"train_loss": 0.2187773734331131, "global_step": 49881, "epoch": 560, "lr": 9.202553841310182e-05}
{"train_loss": 0.22836558520793915, "global_step": 49882, "epoch": 560, "lr": 9.202522431197678e-05}
{"train_loss": 0.26570263504981995, "global_step": 49883, "epoch": 560, "lr": 9.202491020520195e-05}
{"train_loss": 0.21555957198143005, "global_step": 49884, "epoch": 560, "lr": 9.202459609277739e-05}
{"train_loss": 0.2423708438873291, "global_step": 49885, "epoch": 560, "lr": 9.202428197470315e-05}
{"train_loss": 0.22341300547122955, "global_step": 49886, "epoch": 560, "lr": 9.202396785097922e-05}
{"train_loss": 0.20264939963817596, "global_step": 49887, "epoch": 560, "lr": 9.202365372160571e-05}
{"train_loss": 0.2899759113788605, "global_step": 49888, "epoch": 560, "lr": 9.202333958658262e-05}
{"train_loss": 0.18876181542873383, "global_step": 49889, "epoch": 560, "lr": 9.202302544591e-05}
{"train_loss": 0.3794203996658325, "global_step": 49890, "epoch": 560, "lr": 9.202271129958791e-05}
{"train_loss": 0.34549233317375183, "global_step": 49891, "epoch": 560, "lr": 9.202239714761636e-05}
{"train_loss": 0.3711782395839691, "global_step": 49892, "epoch": 560, "lr": 9.202208298999543e-05}
{"train_loss": 0.25417715311050415, "global_step": 49893, "epoch": 560, "lr": 9.202176882672514e-05}
{"train_loss": 0.29036837816238403, "global_step": 49894, "epoch": 560, "lr": 9.202145465780552e-05}
{"train_loss": 0.3021283447742462, "global_step": 49895, "epoch": 560, "lr": 9.202114048323665e-05}
{"train_loss": 0.29265135526657104, "global_step": 49896, "epoch": 560, "lr": 9.202082630301854e-05}
{"train_loss": 0.28816917538642883, "global_step": 49897, "epoch": 560, "lr": 9.202051211715123e-05}
{"train_loss": 0.4212115406990051, "global_step": 49898, "epoch": 560, "lr": 9.202019792563479e-05}
{"train_loss": 0.4025793671607971, "global_step": 49899, "epoch": 560, "lr": 9.201988372846924e-05}
{"train_loss": 0.2788483500480652, "global_step": 49900, "epoch": 560, "lr": 9.201956952565462e-05}
{"train_loss": 0.27200931310653687, "global_step": 49901, "epoch": 560, "lr": 9.2019255317191e-05}
{"train_loss": 0.18250907957553864, "global_step": 49902, "epoch": 560, "lr": 9.201894110307839e-05}
{"train_loss": 0.2729204297065735, "global_step": 49903, "epoch": 560, "lr": 9.201862688331685e-05}
{"train_loss": 0.28669634461402893, "global_step": 49904, "epoch": 560, "lr": 9.201831265790641e-05}
{"train_loss": 0.28328248858451843, "global_step": 49905, "epoch": 560, "lr": 9.201799842684714e-05}
{"train_loss": 0.238884836435318, "global_step": 49906, "epoch": 560, "lr": 9.201768419013904e-05}
{"train_loss": 0.24980786442756653, "global_step": 49907, "epoch": 560, "lr": 9.201736994778219e-05}
{"train_loss": 0.3252338469028473, "global_step": 49908, "epoch": 560, "lr": 9.201705569977662e-05}
{"train_loss": 0.3134903311729431, "global_step": 49909, "epoch": 560, "lr": 9.201674144612235e-05}
{"train_loss": 0.21435244381427765, "global_step": 49910, "epoch": 560, "lr": 9.201642718681944e-05}
{"train_loss": 0.248226597905159, "global_step": 49911, "epoch": 560, "lr": 9.201611292186796e-05}
{"train_loss": 0.2825445234775543, "global_step": 49912, "epoch": 560, "lr": 9.20157986512679e-05}
{"train_loss": 0.32705074548721313, "global_step": 49913, "epoch": 560, "lr": 9.201548437501935e-05}
{"train_loss": 0.2290073186159134, "global_step": 49914, "epoch": 560, "lr": 9.20151700931223e-05}
{"train_loss": 0.2801342308521271, "global_step": 49915, "epoch": 560, "lr": 9.201485580557685e-05}
{"train_loss": 0.3027247190475464, "global_step": 49916, "epoch": 560, "lr": 9.2014541512383e-05}
{"train_loss": 0.20631885528564453, "global_step": 49917, "epoch": 560, "lr": 9.201422721354082e-05}
{"train_loss": 0.31377771496772766, "global_step": 49918, "epoch": 560, "lr": 9.201391290905033e-05}
{"train_loss": 0.2126665711402893, "global_step": 49919, "epoch": 560, "lr": 9.201359859891158e-05}
{"train_loss": 0.235774427652359, "global_step": 49920, "epoch": 560, "lr": 9.201328428312462e-05}
{"train_loss": 0.2949674129486084, "global_step": 49921, "epoch": 560, "lr": 9.201296996168949e-05}
{"train_loss": 0.17804661393165588, "global_step": 49922, "epoch": 560, "lr": 9.201265563460622e-05}
{"train_loss": 0.36710911989212036, "global_step": 49923, "epoch": 560, "lr": 9.201234130187486e-05}
{"train_loss": 0.19194161891937256, "global_step": 49924, "epoch": 560, "lr": 9.201202696349544e-05}
{"train_loss": 0.3092995285987854, "global_step": 49925, "epoch": 560, "lr": 9.201171261946804e-05}
{"train_loss": 0.2671181559562683, "global_step": 49926, "epoch": 560, "lr": 9.201139826979266e-05}
{"train_loss": 0.3163641095161438, "global_step": 49927, "epoch": 560, "lr": 9.201108391446936e-05}
{"train_loss": 0.2754305030187864, "global_step": 49928, "epoch": 560, "lr": 9.201076955349819e-05, "val_loss": 3.0181078910827637, "train_action_mse_error": 22.927370071411133}
{"train_loss": 0.2543178200721741, "global_step": 49929, "epoch": 561, "lr": 9.201045518687919e-05}
{"train_loss": 0.37005850672721863, "global_step": 49930, "epoch": 561, "lr": 9.201014081461239e-05}
{"train_loss": 0.35141465067863464, "global_step": 49931, "epoch": 561, "lr": 9.200982643669785e-05}
{"train_loss": 0.32548046112060547, "global_step": 49932, "epoch": 561, "lr": 9.200951205313558e-05}
{"train_loss": 0.28053271770477295, "global_step": 49933, "epoch": 561, "lr": 9.200919766392564e-05}
{"train_loss": 0.24478593468666077, "global_step": 49934, "epoch": 561, "lr": 9.200888326906809e-05}
{"train_loss": 0.24076080322265625, "global_step": 49935, "epoch": 561, "lr": 9.200856886856295e-05}
{"train_loss": 0.1677432358264923, "global_step": 49936, "epoch": 561, "lr": 9.200825446241028e-05}
{"train_loss": 0.2869647145271301, "global_step": 49937, "epoch": 561, "lr": 9.200794005061011e-05}
{"train_loss": 0.20705044269561768, "global_step": 49938, "epoch": 561, "lr": 9.200762563316248e-05}
{"train_loss": 0.2557549774646759, "global_step": 49939, "epoch": 561, "lr": 9.200731121006743e-05}
{"train_loss": 0.3144141137599945, "global_step": 49940, "epoch": 561, "lr": 9.200699678132503e-05}
{"train_loss": 0.392464816570282, "global_step": 49941, "epoch": 561, "lr": 9.200668234693527e-05}
{"train_loss": 0.2551511526107788, "global_step": 49942, "epoch": 561, "lr": 9.200636790689825e-05}
{"train_loss": 0.22792254388332367, "global_step": 49943, "epoch": 561, "lr": 9.200605346121398e-05}
{"train_loss": 0.3040093779563904, "global_step": 49944, "epoch": 561, "lr": 9.20057390098825e-05}
{"train_loss": 0.3695833683013916, "global_step": 49945, "epoch": 561, "lr": 9.200542455290386e-05}
{"train_loss": 0.275513231754303, "global_step": 49946, "epoch": 561, "lr": 9.200511009027812e-05}
{"train_loss": 0.20371749997138977, "global_step": 49947, "epoch": 561, "lr": 9.200479562200529e-05}
{"train_loss": 0.1680723875761032, "global_step": 49948, "epoch": 561, "lr": 9.200448114808543e-05}
{"train_loss": 0.2894035577774048, "global_step": 49949, "epoch": 561, "lr": 9.200416666851858e-05}
{"train_loss": 0.351696252822876, "global_step": 49950, "epoch": 561, "lr": 9.200385218330478e-05}
{"train_loss": 0.3683864176273346, "global_step": 49951, "epoch": 561, "lr": 9.200353769244407e-05}
{"train_loss": 0.24310554563999176, "global_step": 49952, "epoch": 561, "lr": 9.200322319593651e-05}
{"train_loss": 0.2870866656303406, "global_step": 49953, "epoch": 561, "lr": 9.200290869378212e-05}
{"train_loss": 0.3180937170982361, "global_step": 49954, "epoch": 561, "lr": 9.200259418598094e-05}
{"train_loss": 0.27773308753967285, "global_step": 49955, "epoch": 561, "lr": 9.200227967253305e-05}
{"train_loss": 0.23190326988697052, "global_step": 49956, "epoch": 561, "lr": 9.200196515343844e-05}
{"train_loss": 0.2454967200756073, "global_step": 49957, "epoch": 561, "lr": 9.200165062869719e-05}
{"train_loss": 0.238030806183815, "global_step": 49958, "epoch": 561, "lr": 9.200133609830933e-05}
{"train_loss": 0.4074295163154602, "global_step": 49959, "epoch": 561, "lr": 9.20010215622749e-05}
{"train_loss": 0.2634887099266052, "global_step": 49960, "epoch": 561, "lr": 9.200070702059394e-05}
{"train_loss": 0.21685051918029785, "global_step": 49961, "epoch": 561, "lr": 9.20003924732665e-05}
{"train_loss": 0.2681557238101959, "global_step": 49962, "epoch": 561, "lr": 9.200007792029261e-05}
{"train_loss": 0.24403700232505798, "global_step": 49963, "epoch": 561, "lr": 9.199976336167235e-05}
{"train_loss": 0.35320812463760376, "global_step": 49964, "epoch": 561, "lr": 9.199944879740572e-05}
{"train_loss": 0.26484236121177673, "global_step": 49965, "epoch": 561, "lr": 9.199913422749275e-05}
{"train_loss": 0.2756238579750061, "global_step": 49966, "epoch": 561, "lr": 9.199881965193355e-05}
{"train_loss": 0.2906332314014435, "global_step": 49967, "epoch": 561, "lr": 9.199850507072809e-05}
{"train_loss": 0.20198754966259003, "global_step": 49968, "epoch": 561, "lr": 9.199819048387646e-05}
{"train_loss": 0.16663256287574768, "global_step": 49969, "epoch": 561, "lr": 9.199787589137867e-05}
{"train_loss": 0.3337862193584442, "global_step": 49970, "epoch": 561, "lr": 9.19975612932348e-05}
{"train_loss": 0.371601939201355, "global_step": 49971, "epoch": 561, "lr": 9.199724668944485e-05}
{"train_loss": 0.3151884377002716, "global_step": 49972, "epoch": 561, "lr": 9.199693208000888e-05}
{"train_loss": 0.26245054602622986, "global_step": 49973, "epoch": 561, "lr": 9.199661746492694e-05}
{"train_loss": 0.2269330471754074, "global_step": 49974, "epoch": 561, "lr": 9.199630284419908e-05}
{"train_loss": 0.20867785811424255, "global_step": 49975, "epoch": 561, "lr": 9.199598821782531e-05}
{"train_loss": 0.24759501218795776, "global_step": 49976, "epoch": 561, "lr": 9.199567358580569e-05}
{"train_loss": 0.23996448516845703, "global_step": 49977, "epoch": 561, "lr": 9.199535894814028e-05}
{"train_loss": 0.2988431453704834, "global_step": 49978, "epoch": 561, "lr": 9.199504430482911e-05}
{"train_loss": 0.19975247979164124, "global_step": 49979, "epoch": 561, "lr": 9.199472965587221e-05}
{"train_loss": 0.3613898456096649, "global_step": 49980, "epoch": 561, "lr": 9.199441500126962e-05}
{"train_loss": 0.455662339925766, "global_step": 49981, "epoch": 561, "lr": 9.199410034102141e-05}
{"train_loss": 0.3078610599040985, "global_step": 49982, "epoch": 561, "lr": 9.199378567512758e-05}
{"train_loss": 0.28517022728919983, "global_step": 49983, "epoch": 561, "lr": 9.199347100358822e-05}
{"train_loss": 0.3280993103981018, "global_step": 49984, "epoch": 561, "lr": 9.199315632640333e-05}
{"train_loss": 0.23221972584724426, "global_step": 49985, "epoch": 561, "lr": 9.199284164357299e-05}
{"train_loss": 0.2919456958770752, "global_step": 49986, "epoch": 561, "lr": 9.199252695509722e-05}
{"train_loss": 0.2651672661304474, "global_step": 49987, "epoch": 561, "lr": 9.199221226097607e-05}
{"train_loss": 0.29026278853416443, "global_step": 49988, "epoch": 561, "lr": 9.199189756120957e-05}
{"train_loss": 0.2812456786632538, "global_step": 49989, "epoch": 561, "lr": 9.199158285579777e-05}
{"train_loss": 0.24179834127426147, "global_step": 49990, "epoch": 561, "lr": 9.199126814474073e-05}
{"train_loss": 0.31194373965263367, "global_step": 49991, "epoch": 561, "lr": 9.199095342803846e-05}
{"train_loss": 0.27994048595428467, "global_step": 49992, "epoch": 561, "lr": 9.199063870569103e-05}
{"train_loss": 0.20825204253196716, "global_step": 49993, "epoch": 561, "lr": 9.199032397769845e-05}
{"train_loss": 0.23935313522815704, "global_step": 49994, "epoch": 561, "lr": 9.199000924406079e-05}
{"train_loss": 0.2215183824300766, "global_step": 49995, "epoch": 561, "lr": 9.19896945047781e-05}
{"train_loss": 0.44423583149909973, "global_step": 49996, "epoch": 561, "lr": 9.19893797598504e-05}
{"train_loss": 0.3195175230503082, "global_step": 49997, "epoch": 561, "lr": 9.198906500927774e-05}
{"train_loss": 0.3077523410320282, "global_step": 49998, "epoch": 561, "lr": 9.198875025306015e-05}
{"train_loss": 0.29065054655075073, "global_step": 49999, "epoch": 561, "lr": 9.198843549119768e-05}
{"train_loss": 0.2989845871925354, "global_step": 50000, "epoch": 561, "lr": 9.19881207236904e-05}
{"train_loss": 0.2730519473552704, "global_step": 50001, "epoch": 561, "lr": 9.198780595053832e-05}
{"train_loss": 0.3060254752635956, "global_step": 50002, "epoch": 561, "lr": 9.198749117174149e-05}
{"train_loss": 0.2611905634403229, "global_step": 50003, "epoch": 561, "lr": 9.198717638729996e-05}
{"train_loss": 0.3077191114425659, "global_step": 50004, "epoch": 561, "lr": 9.198686159721376e-05}
{"train_loss": 0.25051045417785645, "global_step": 50005, "epoch": 561, "lr": 9.198654680148293e-05}
{"train_loss": 0.3033987879753113, "global_step": 50006, "epoch": 561, "lr": 9.198623200010754e-05}
{"train_loss": 0.32894811034202576, "global_step": 50007, "epoch": 561, "lr": 9.19859171930876e-05}
{"train_loss": 0.25552013516426086, "global_step": 50008, "epoch": 561, "lr": 9.198560238042315e-05}
{"train_loss": 0.2613925635814667, "global_step": 50009, "epoch": 561, "lr": 9.198528756211428e-05}
{"train_loss": 0.27932459115982056, "global_step": 50010, "epoch": 561, "lr": 9.198497273816098e-05}
{"train_loss": 0.24632631242275238, "global_step": 50011, "epoch": 561, "lr": 9.198465790856333e-05}
{"train_loss": 0.3170926868915558, "global_step": 50012, "epoch": 561, "lr": 9.198434307332133e-05}
{"train_loss": 0.29369300603866577, "global_step": 50013, "epoch": 561, "lr": 9.198402823243506e-05}
{"train_loss": 0.30773696303367615, "global_step": 50014, "epoch": 561, "lr": 9.198371338590455e-05}
{"train_loss": 0.3318933844566345, "global_step": 50015, "epoch": 561, "lr": 9.198339853372982e-05}
{"train_loss": 0.33395421504974365, "global_step": 50016, "epoch": 561, "lr": 9.198308367591096e-05}
{"train_loss": 0.2829522316375475, "global_step": 50017, "epoch": 561, "lr": 9.198276881244798e-05, "val_loss": 2.9306492805480957}
{"train_loss": 0.3861471712589264, "global_step": 50018, "epoch": 562, "lr": 9.198245394334093e-05}
{"train_loss": 0.2713207006454468, "global_step": 50019, "epoch": 562, "lr": 9.198213906858983e-05}
{"train_loss": 0.2875611484050751, "global_step": 50020, "epoch": 562, "lr": 9.198182418819477e-05}
{"train_loss": 0.28751814365386963, "global_step": 50021, "epoch": 562, "lr": 9.198150930215576e-05}
{"train_loss": 0.26172831654548645, "global_step": 50022, "epoch": 562, "lr": 9.198119441047284e-05}
{"train_loss": 0.20276787877082825, "global_step": 50023, "epoch": 562, "lr": 9.198087951314607e-05}
{"train_loss": 0.2028120905160904, "global_step": 50024, "epoch": 562, "lr": 9.198056461017546e-05}
{"train_loss": 0.2878338694572449, "global_step": 50025, "epoch": 562, "lr": 9.198024970156109e-05}
{"train_loss": 0.2734498381614685, "global_step": 50026, "epoch": 562, "lr": 9.1979934787303e-05}
{"train_loss": 0.29572153091430664, "global_step": 50027, "epoch": 562, "lr": 9.19796198674012e-05}
{"train_loss": 0.26944127678871155, "global_step": 50028, "epoch": 562, "lr": 9.197930494185576e-05}
{"train_loss": 0.31076908111572266, "global_step": 50029, "epoch": 562, "lr": 9.197899001066673e-05}
{"train_loss": 0.1707383096218109, "global_step": 50030, "epoch": 562, "lr": 9.19786750738341e-05}
{"train_loss": 0.22033020853996277, "global_step": 50031, "epoch": 562, "lr": 9.197836013135797e-05}
{"train_loss": 0.2806069850921631, "global_step": 50032, "epoch": 562, "lr": 9.197804518323836e-05}
{"train_loss": 0.25052788853645325, "global_step": 50033, "epoch": 562, "lr": 9.197773022947531e-05}
{"train_loss": 0.31396031379699707, "global_step": 50034, "epoch": 562, "lr": 9.197741527006886e-05}
{"train_loss": 0.3367157280445099, "global_step": 50035, "epoch": 562, "lr": 9.197710030501908e-05}
{"train_loss": 0.20628036558628082, "global_step": 50036, "epoch": 562, "lr": 9.197678533432597e-05}
{"train_loss": 0.2687089741230011, "global_step": 50037, "epoch": 562, "lr": 9.19764703579896e-05}
{"train_loss": 0.2032073587179184, "global_step": 50038, "epoch": 562, "lr": 9.197615537601e-05}
{"train_loss": 0.23669783771038055, "global_step": 50039, "epoch": 562, "lr": 9.197584038838723e-05}
{"train_loss": 0.20980407297611237, "global_step": 50040, "epoch": 562, "lr": 9.197552539512132e-05}
{"train_loss": 0.2921975255012512, "global_step": 50041, "epoch": 562, "lr": 9.19752103962123e-05}
{"train_loss": 0.2702410817146301, "global_step": 50042, "epoch": 562, "lr": 9.197489539166023e-05}
{"train_loss": 0.24619731307029724, "global_step": 50043, "epoch": 562, "lr": 9.197458038146515e-05}
{"train_loss": 0.22205159068107605, "global_step": 50044, "epoch": 562, "lr": 9.19742653656271e-05}
{"train_loss": 0.21313153207302094, "global_step": 50045, "epoch": 562, "lr": 9.19739503441461e-05}
{"train_loss": 0.3270924687385559, "global_step": 50046, "epoch": 562, "lr": 9.197363531702224e-05}
{"train_loss": 0.27383285760879517, "global_step": 50047, "epoch": 562, "lr": 9.197332028425553e-05}
{"train_loss": 0.21150949597358704, "global_step": 50048, "epoch": 562, "lr": 9.197300524584602e-05}
{"train_loss": 0.2813258171081543, "global_step": 50049, "epoch": 562, "lr": 9.197269020179374e-05}
{"train_loss": 0.2789004147052765, "global_step": 50050, "epoch": 562, "lr": 9.197237515209875e-05}
{"train_loss": 0.2886066138744354, "global_step": 50051, "epoch": 562, "lr": 9.19720600967611e-05}
{"train_loss": 0.2120845764875412, "global_step": 50052, "epoch": 562, "lr": 9.197174503578078e-05}
{"train_loss": 0.17489010095596313, "global_step": 50053, "epoch": 562, "lr": 9.197142996915791e-05}
{"train_loss": 0.24331559240818024, "global_step": 50054, "epoch": 562, "lr": 9.197111489689247e-05}
{"train_loss": 0.2244034707546234, "global_step": 50055, "epoch": 562, "lr": 9.197079981898454e-05}
{"train_loss": 0.2715587913990021, "global_step": 50056, "epoch": 562, "lr": 9.197048473543414e-05}
{"train_loss": 0.17737802863121033, "global_step": 50057, "epoch": 562, "lr": 9.197016964624132e-05}
{"train_loss": 0.31471580266952515, "global_step": 50058, "epoch": 562, "lr": 9.196985455140612e-05}
{"train_loss": 0.30525025725364685, "global_step": 50059, "epoch": 562, "lr": 9.196953945092859e-05}
{"train_loss": 0.18363144993782043, "global_step": 50060, "epoch": 562, "lr": 9.196922434480876e-05}
{"train_loss": 0.192225381731987, "global_step": 50061, "epoch": 562, "lr": 9.196890923304668e-05}
{"train_loss": 0.2964669167995453, "global_step": 50062, "epoch": 562, "lr": 9.196859411564241e-05}
{"train_loss": 0.2891993522644043, "global_step": 50063, "epoch": 562, "lr": 9.196827899259595e-05}
{"train_loss": 0.2000371366739273, "global_step": 50064, "epoch": 562, "lr": 9.196796386390738e-05}
{"train_loss": 0.24465855956077576, "global_step": 50065, "epoch": 562, "lr": 9.196764872957673e-05}
{"train_loss": 0.35208210349082947, "global_step": 50066, "epoch": 562, "lr": 9.196733358960403e-05}
{"train_loss": 0.21015387773513794, "global_step": 50067, "epoch": 562, "lr": 9.196701844398933e-05}
{"train_loss": 0.21969056129455566, "global_step": 50068, "epoch": 562, "lr": 9.196670329273269e-05}
{"train_loss": 0.26566311717033386, "global_step": 50069, "epoch": 562, "lr": 9.196638813583414e-05}
{"train_loss": 0.2681429982185364, "global_step": 50070, "epoch": 562, "lr": 9.19660729732937e-05}
{"train_loss": 0.24196703732013702, "global_step": 50071, "epoch": 562, "lr": 9.196575780511145e-05}
{"train_loss": 0.24138310551643372, "global_step": 50072, "epoch": 562, "lr": 9.196544263128741e-05}
{"train_loss": 0.26397138833999634, "global_step": 50073, "epoch": 562, "lr": 9.196512745182163e-05}
{"train_loss": 0.22474294900894165, "global_step": 50074, "epoch": 562, "lr": 9.196481226671415e-05}
{"train_loss": 0.2706531584262848, "global_step": 50075, "epoch": 562, "lr": 9.196449707596501e-05}
{"train_loss": 0.32143309712409973, "global_step": 50076, "epoch": 562, "lr": 9.196418187957426e-05}
{"train_loss": 0.3011997640132904, "global_step": 50077, "epoch": 562, "lr": 9.196386667754193e-05}
{"train_loss": 0.3519473969936371, "global_step": 50078, "epoch": 562, "lr": 9.196355146986808e-05}
{"train_loss": 0.28836193680763245, "global_step": 50079, "epoch": 562, "lr": 9.196323625655275e-05}
{"train_loss": 0.3120412230491638, "global_step": 50080, "epoch": 562, "lr": 9.196292103759593e-05}
{"train_loss": 0.3217623233795166, "global_step": 50081, "epoch": 562, "lr": 9.196260581299776e-05}
{"train_loss": 0.2998602092266083, "global_step": 50082, "epoch": 562, "lr": 9.19622905827582e-05}
{"train_loss": 0.2780472934246063, "global_step": 50083, "epoch": 562, "lr": 9.196197534687733e-05}
{"train_loss": 0.2979069948196411, "global_step": 50084, "epoch": 562, "lr": 9.196166010535518e-05}
{"train_loss": 0.21851471066474915, "global_step": 50085, "epoch": 562, "lr": 9.196134485819179e-05}
{"train_loss": 0.25650015473365784, "global_step": 50086, "epoch": 562, "lr": 9.196102960538722e-05}
{"train_loss": 0.2612214982509613, "global_step": 50087, "epoch": 562, "lr": 9.19607143469415e-05}
{"train_loss": 0.2844233810901642, "global_step": 50088, "epoch": 562, "lr": 9.196039908285467e-05}
{"train_loss": 0.3286155164241791, "global_step": 50089, "epoch": 562, "lr": 9.196008381312677e-05}
{"train_loss": 0.20802272856235504, "global_step": 50090, "epoch": 562, "lr": 9.195976853775785e-05}
{"train_loss": 0.26455965638160706, "global_step": 50091, "epoch": 562, "lr": 9.195945325674798e-05}
{"train_loss": 0.3311125338077545, "global_step": 50092, "epoch": 562, "lr": 9.195913797009714e-05}
{"train_loss": 0.25647425651550293, "global_step": 50093, "epoch": 562, "lr": 9.19588226778054e-05}
{"train_loss": 0.23435890674591064, "global_step": 50094, "epoch": 562, "lr": 9.195850737987283e-05}
{"train_loss": 0.30744290351867676, "global_step": 50095, "epoch": 562, "lr": 9.195819207629943e-05}
{"train_loss": 0.32082048058509827, "global_step": 50096, "epoch": 562, "lr": 9.195787676708529e-05}
{"train_loss": 0.2706927955150604, "global_step": 50097, "epoch": 562, "lr": 9.19575614522304e-05}
{"train_loss": 0.27745380997657776, "global_step": 50098, "epoch": 562, "lr": 9.195724613173484e-05}
{"train_loss": 0.36942604184150696, "global_step": 50099, "epoch": 562, "lr": 9.195693080559863e-05}
{"train_loss": 0.2435164749622345, "global_step": 50100, "epoch": 562, "lr": 9.195661547382182e-05}
{"train_loss": 0.21588213741779327, "global_step": 50101, "epoch": 562, "lr": 9.195630013640447e-05}
{"train_loss": 0.40468519926071167, "global_step": 50102, "epoch": 562, "lr": 9.19559847933466e-05}
{"train_loss": 0.2035621851682663, "global_step": 50103, "epoch": 562, "lr": 9.195566944464826e-05}
{"train_loss": 0.25404882431030273, "global_step": 50104, "epoch": 562, "lr": 9.195535409030949e-05}
{"train_loss": 0.20862765610218048, "global_step": 50105, "epoch": 562, "lr": 9.195503873033034e-05}
{"train_loss": 0.26456677428122316, "global_step": 50106, "epoch": 562, "lr": 9.195472336471083e-05, "val_loss": 3.081756114959717}
{"train_loss": 0.29956570267677307, "global_step": 50107, "epoch": 563, "lr": 9.195440799345103e-05}
{"train_loss": 0.2322106808423996, "global_step": 50108, "epoch": 563, "lr": 9.195409261655096e-05}
{"train_loss": 0.26919975876808167, "global_step": 50109, "epoch": 563, "lr": 9.195377723401069e-05}
{"train_loss": 0.2901289165019989, "global_step": 50110, "epoch": 563, "lr": 9.195346184583025e-05}
{"train_loss": 0.28546056151390076, "global_step": 50111, "epoch": 563, "lr": 9.195314645200965e-05}
{"train_loss": 0.2936411499977112, "global_step": 50112, "epoch": 563, "lr": 9.195283105254899e-05}
{"train_loss": 0.24463285505771637, "global_step": 50113, "epoch": 563, "lr": 9.195251564744827e-05}
{"train_loss": 0.18446707725524902, "global_step": 50114, "epoch": 563, "lr": 9.195220023670755e-05}
{"train_loss": 0.35040083527565, "global_step": 50115, "epoch": 563, "lr": 9.195188482032686e-05}
{"train_loss": 0.2619948387145996, "global_step": 50116, "epoch": 563, "lr": 9.195156939830626e-05}
{"train_loss": 0.2977306544780731, "global_step": 50117, "epoch": 563, "lr": 9.195125397064577e-05}
{"train_loss": 0.27267736196517944, "global_step": 50118, "epoch": 563, "lr": 9.195093853734547e-05}
{"train_loss": 0.22972902655601501, "global_step": 50119, "epoch": 563, "lr": 9.195062309840534e-05}
{"train_loss": 0.2302696853876114, "global_step": 50120, "epoch": 563, "lr": 9.19503076538255e-05}
{"train_loss": 0.2937447130680084, "global_step": 50121, "epoch": 563, "lr": 9.194999220360593e-05}
{"train_loss": 0.22210410237312317, "global_step": 50122, "epoch": 563, "lr": 9.19496767477467e-05}
{"train_loss": 0.28755131363868713, "global_step": 50123, "epoch": 563, "lr": 9.194936128624785e-05}
{"train_loss": 0.27663761377334595, "global_step": 50124, "epoch": 563, "lr": 9.194904581910941e-05}
{"train_loss": 0.3567329943180084, "global_step": 50125, "epoch": 563, "lr": 9.194873034633143e-05}
{"train_loss": 0.27323976159095764, "global_step": 50126, "epoch": 563, "lr": 9.194841486791397e-05}
{"train_loss": 0.2584529221057892, "global_step": 50127, "epoch": 563, "lr": 9.194809938385704e-05}
{"train_loss": 0.21331337094306946, "global_step": 50128, "epoch": 563, "lr": 9.194778389416072e-05}
{"train_loss": 0.2630545198917389, "global_step": 50129, "epoch": 563, "lr": 9.194746839882502e-05}
{"train_loss": 0.2959638237953186, "global_step": 50130, "epoch": 563, "lr": 9.194715289784999e-05}
{"train_loss": 0.2055196464061737, "global_step": 50131, "epoch": 563, "lr": 9.194683739123569e-05}
{"train_loss": 0.2774607837200165, "global_step": 50132, "epoch": 563, "lr": 9.194652187898214e-05}
{"train_loss": 0.19160550832748413, "global_step": 50133, "epoch": 563, "lr": 9.194620636108938e-05}
{"train_loss": 0.2551000416278839, "global_step": 50134, "epoch": 563, "lr": 9.194589083755748e-05}
{"train_loss": 0.2882246673107147, "global_step": 50135, "epoch": 563, "lr": 9.194557530838646e-05}
{"train_loss": 0.24953311681747437, "global_step": 50136, "epoch": 563, "lr": 9.194525977357636e-05}
{"train_loss": 0.3484080731868744, "global_step": 50137, "epoch": 563, "lr": 9.194494423312725e-05}
{"train_loss": 0.23176300525665283, "global_step": 50138, "epoch": 563, "lr": 9.194462868703914e-05}
{"train_loss": 0.32859063148498535, "global_step": 50139, "epoch": 563, "lr": 9.194431313531209e-05}
{"train_loss": 0.360687792301178, "global_step": 50140, "epoch": 563, "lr": 9.194399757794614e-05}
{"train_loss": 0.2776165008544922, "global_step": 50141, "epoch": 563, "lr": 9.194368201494133e-05}
{"train_loss": 0.28813832998275757, "global_step": 50142, "epoch": 563, "lr": 9.19433664462977e-05}
{"train_loss": 0.32189247012138367, "global_step": 50143, "epoch": 563, "lr": 9.194305087201531e-05}
{"train_loss": 0.24479669332504272, "global_step": 50144, "epoch": 563, "lr": 9.194273529209416e-05}
{"train_loss": 0.2165277749300003, "global_step": 50145, "epoch": 563, "lr": 9.194241970653435e-05}
{"train_loss": 0.2619907557964325, "global_step": 50146, "epoch": 563, "lr": 9.194210411533588e-05}
{"train_loss": 0.2350492626428604, "global_step": 50147, "epoch": 563, "lr": 9.194178851849879e-05}
{"train_loss": 0.187595933675766, "global_step": 50148, "epoch": 563, "lr": 9.194147291602316e-05}
{"train_loss": 0.37184563279151917, "global_step": 50149, "epoch": 563, "lr": 9.194115730790901e-05}
{"train_loss": 0.3442343473434448, "global_step": 50150, "epoch": 563, "lr": 9.194084169415638e-05}
{"train_loss": 0.2807895243167877, "global_step": 50151, "epoch": 563, "lr": 9.194052607476531e-05}
{"train_loss": 0.2925671637058258, "global_step": 50152, "epoch": 563, "lr": 9.194021044973585e-05}
{"train_loss": 0.30117711424827576, "global_step": 50153, "epoch": 563, "lr": 9.193989481906803e-05}
{"train_loss": 0.31734609603881836, "global_step": 50154, "epoch": 563, "lr": 9.19395791827619e-05}
{"train_loss": 0.44382748007774353, "global_step": 50155, "epoch": 563, "lr": 9.193926354081752e-05}
{"train_loss": 0.23341822624206543, "global_step": 50156, "epoch": 563, "lr": 9.193894789323492e-05}
{"train_loss": 0.22981493175029755, "global_step": 50157, "epoch": 563, "lr": 9.193863224001414e-05}
{"train_loss": 0.2151583731174469, "global_step": 50158, "epoch": 563, "lr": 9.19383165811552e-05}
{"train_loss": 0.1839158833026886, "global_step": 50159, "epoch": 563, "lr": 9.193800091665819e-05}
{"train_loss": 0.13947713375091553, "global_step": 50160, "epoch": 563, "lr": 9.193768524652311e-05}
{"train_loss": 0.18198162317276, "global_step": 50161, "epoch": 563, "lr": 9.193736957075002e-05}
{"train_loss": 0.4034942090511322, "global_step": 50162, "epoch": 563, "lr": 9.193705388933897e-05}
{"train_loss": 0.24386419355869293, "global_step": 50163, "epoch": 563, "lr": 9.193673820228998e-05}
{"train_loss": 0.24076175689697266, "global_step": 50164, "epoch": 563, "lr": 9.193642250960313e-05}
{"train_loss": 0.19569888710975647, "global_step": 50165, "epoch": 563, "lr": 9.193610681127842e-05}
{"train_loss": 0.28397005796432495, "global_step": 50166, "epoch": 563, "lr": 9.193579110731592e-05}
{"train_loss": 0.2967996597290039, "global_step": 50167, "epoch": 563, "lr": 9.193547539771565e-05}
{"train_loss": 0.32145294547080994, "global_step": 50168, "epoch": 563, "lr": 9.193515968247768e-05}
{"train_loss": 0.26544836163520813, "global_step": 50169, "epoch": 563, "lr": 9.193484396160205e-05}
{"train_loss": 0.3351304829120636, "global_step": 50170, "epoch": 563, "lr": 9.193452823508877e-05}
{"train_loss": 0.28642427921295166, "global_step": 50171, "epoch": 563, "lr": 9.193421250293793e-05}
{"train_loss": 0.2507193088531494, "global_step": 50172, "epoch": 563, "lr": 9.193389676514951e-05}
{"train_loss": 0.2823978364467621, "global_step": 50173, "epoch": 563, "lr": 9.19335810217236e-05}
{"train_loss": 0.25067242980003357, "global_step": 50174, "epoch": 563, "lr": 9.193326527266024e-05}
{"train_loss": 0.29972043633461, "global_step": 50175, "epoch": 563, "lr": 9.193294951795947e-05}
{"train_loss": 0.28904709219932556, "global_step": 50176, "epoch": 563, "lr": 9.19326337576213e-05}
{"train_loss": 0.34657952189445496, "global_step": 50177, "epoch": 563, "lr": 9.193231799164583e-05}
{"train_loss": 0.2864110469818115, "global_step": 50178, "epoch": 563, "lr": 9.193200222003306e-05}
{"train_loss": 0.3348033130168915, "global_step": 50179, "epoch": 563, "lr": 9.193168644278305e-05}
{"train_loss": 0.2986895740032196, "global_step": 50180, "epoch": 563, "lr": 9.193137065989583e-05}
{"train_loss": 0.3652544319629669, "global_step": 50181, "epoch": 563, "lr": 9.193105487137144e-05}
{"train_loss": 0.31047505140304565, "global_step": 50182, "epoch": 563, "lr": 9.193073907720992e-05}
{"train_loss": 0.31295567750930786, "global_step": 50183, "epoch": 563, "lr": 9.193042327741135e-05}
{"train_loss": 0.2240496575832367, "global_step": 50184, "epoch": 563, "lr": 9.193010747197573e-05}
{"train_loss": 0.39175671339035034, "global_step": 50185, "epoch": 563, "lr": 9.192979166090313e-05}
{"train_loss": 0.34014004468917847, "global_step": 50186, "epoch": 563, "lr": 9.192947584419356e-05}
{"train_loss": 0.33458584547042847, "global_step": 50187, "epoch": 563, "lr": 9.19291600218471e-05}
{"train_loss": 0.2777647078037262, "global_step": 50188, "epoch": 563, "lr": 9.192884419386379e-05}
{"train_loss": 0.37073662877082825, "global_step": 50189, "epoch": 563, "lr": 9.192852836024362e-05}
{"train_loss": 0.22173015773296356, "global_step": 50190, "epoch": 563, "lr": 9.19282125209867e-05}
{"train_loss": 0.26982781291007996, "global_step": 50191, "epoch": 563, "lr": 9.192789667609303e-05}
{"train_loss": 0.2757794260978699, "global_step": 50192, "epoch": 563, "lr": 9.192758082556267e-05}
{"train_loss": 0.21157412230968475, "global_step": 50193, "epoch": 563, "lr": 9.192726496939566e-05}
{"train_loss": 0.27782049775123596, "global_step": 50194, "epoch": 563, "lr": 9.192694910759203e-05}
{"train_loss": 0.27801115509499325, "global_step": 50195, "epoch": 563, "lr": 9.192663324015184e-05, "val_loss": 3.1364195346832275}
{"train_loss": 0.2673009932041168, "global_step": 50196, "epoch": 564, "lr": 9.192631736707514e-05}
{"train_loss": 0.27609118819236755, "global_step": 50197, "epoch": 564, "lr": 9.192600148836192e-05}
{"train_loss": 0.280626118183136, "global_step": 50198, "epoch": 564, "lr": 9.19256856040123e-05}
{"train_loss": 0.2913154065608978, "global_step": 50199, "epoch": 564, "lr": 9.192536971402625e-05}
{"train_loss": 0.1956060528755188, "global_step": 50200, "epoch": 564, "lr": 9.192505381840386e-05}
{"train_loss": 0.27496764063835144, "global_step": 50201, "epoch": 564, "lr": 9.192473791714515e-05}
{"train_loss": 0.2671284079551697, "global_step": 50202, "epoch": 564, "lr": 9.192442201025019e-05}
{"train_loss": 0.2135171890258789, "global_step": 50203, "epoch": 564, "lr": 9.192410609771899e-05}
{"train_loss": 0.24225188791751862, "global_step": 50204, "epoch": 564, "lr": 9.19237901795516e-05}
{"train_loss": 0.18622176349163055, "global_step": 50205, "epoch": 564, "lr": 9.192347425574807e-05}
{"train_loss": 0.34934133291244507, "global_step": 50206, "epoch": 564, "lr": 9.192315832630845e-05}
{"train_loss": 0.22052250802516937, "global_step": 50207, "epoch": 564, "lr": 9.192284239123277e-05}
{"train_loss": 0.2648472785949707, "global_step": 50208, "epoch": 564, "lr": 9.192252645052107e-05}
{"train_loss": 0.2478678822517395, "global_step": 50209, "epoch": 564, "lr": 9.192221050417338e-05}
{"train_loss": 0.33412396907806396, "global_step": 50210, "epoch": 564, "lr": 9.192189455218979e-05}
{"train_loss": 0.2229519635438919, "global_step": 50211, "epoch": 564, "lr": 9.192157859457029e-05}
{"train_loss": 0.16846486926078796, "global_step": 50212, "epoch": 564, "lr": 9.192126263131495e-05}
{"train_loss": 0.20508728921413422, "global_step": 50213, "epoch": 564, "lr": 9.192094666242381e-05}
{"train_loss": 0.22126364707946777, "global_step": 50214, "epoch": 564, "lr": 9.192063068789691e-05}
{"train_loss": 0.3628399670124054, "global_step": 50215, "epoch": 564, "lr": 9.192031470773429e-05}
{"train_loss": 0.3350439667701721, "global_step": 50216, "epoch": 564, "lr": 9.1919998721936e-05}
{"train_loss": 0.2011464685201645, "global_step": 50217, "epoch": 564, "lr": 9.191968273050207e-05}
{"train_loss": 0.29810914397239685, "global_step": 50218, "epoch": 564, "lr": 9.191936673343256e-05}
{"train_loss": 0.2600475549697876, "global_step": 50219, "epoch": 564, "lr": 9.191905073072748e-05}
{"train_loss": 0.33607199788093567, "global_step": 50220, "epoch": 564, "lr": 9.191873472238692e-05}
{"train_loss": 0.2197647988796234, "global_step": 50221, "epoch": 564, "lr": 9.191841870841089e-05}
{"train_loss": 0.2339690923690796, "global_step": 50222, "epoch": 564, "lr": 9.191810268879944e-05}
{"train_loss": 0.2549271583557129, "global_step": 50223, "epoch": 564, "lr": 9.19177866635526e-05}
{"train_loss": 0.2848477363586426, "global_step": 50224, "epoch": 564, "lr": 9.191747063267043e-05}
{"train_loss": 0.22595646977424622, "global_step": 50225, "epoch": 564, "lr": 9.191715459615297e-05}
{"train_loss": 0.27292031049728394, "global_step": 50226, "epoch": 564, "lr": 9.191683855400026e-05}
{"train_loss": 0.26639559864997864, "global_step": 50227, "epoch": 564, "lr": 9.191652250621235e-05}
{"train_loss": 0.30298495292663574, "global_step": 50228, "epoch": 564, "lr": 9.191620645278926e-05}
{"train_loss": 0.34422487020492554, "global_step": 50229, "epoch": 564, "lr": 9.191589039373105e-05}
{"train_loss": 0.26188111305236816, "global_step": 50230, "epoch": 564, "lr": 9.191557432903777e-05}
{"train_loss": 0.3437935709953308, "global_step": 50231, "epoch": 564, "lr": 9.191525825870944e-05}
{"train_loss": 0.2216121405363083, "global_step": 50232, "epoch": 564, "lr": 9.191494218274612e-05}
{"train_loss": 0.22005751729011536, "global_step": 50233, "epoch": 564, "lr": 9.191462610114784e-05}
{"train_loss": 0.19473165273666382, "global_step": 50234, "epoch": 564, "lr": 9.191431001391466e-05}
{"train_loss": 0.32297956943511963, "global_step": 50235, "epoch": 564, "lr": 9.191399392104661e-05}
{"train_loss": 0.2271556705236435, "global_step": 50236, "epoch": 564, "lr": 9.191367782254372e-05}
{"train_loss": 0.2958478629589081, "global_step": 50237, "epoch": 564, "lr": 9.191336171840605e-05}
{"train_loss": 0.22081592679023743, "global_step": 50238, "epoch": 564, "lr": 9.191304560863365e-05}
{"train_loss": 0.24061959981918335, "global_step": 50239, "epoch": 564, "lr": 9.191272949322655e-05}
{"train_loss": 0.2773727774620056, "global_step": 50240, "epoch": 564, "lr": 9.191241337218479e-05}
{"train_loss": 0.28257322311401367, "global_step": 50241, "epoch": 564, "lr": 9.191209724550842e-05}
{"train_loss": 0.2568712830543518, "global_step": 50242, "epoch": 564, "lr": 9.191178111319749e-05}
{"train_loss": 0.3073153495788574, "global_step": 50243, "epoch": 564, "lr": 9.191146497525202e-05}
{"train_loss": 0.22472260892391205, "global_step": 50244, "epoch": 564, "lr": 9.191114883167207e-05}
{"train_loss": 0.34228554368019104, "global_step": 50245, "epoch": 564, "lr": 9.191083268245767e-05}
{"train_loss": 0.329515278339386, "global_step": 50246, "epoch": 564, "lr": 9.191051652760887e-05}
{"train_loss": 0.24974015355110168, "global_step": 50247, "epoch": 564, "lr": 9.191020036712572e-05}
{"train_loss": 0.19285406172275543, "global_step": 50248, "epoch": 564, "lr": 9.190988420100825e-05}
{"train_loss": 0.2671259343624115, "global_step": 50249, "epoch": 564, "lr": 9.19095680292565e-05}
{"train_loss": 0.28547802567481995, "global_step": 50250, "epoch": 564, "lr": 9.190925185187051e-05}
{"train_loss": 0.22636331617832184, "global_step": 50251, "epoch": 564, "lr": 9.190893566885036e-05}
{"train_loss": 0.1962386965751648, "global_step": 50252, "epoch": 564, "lr": 9.190861948019605e-05}
{"train_loss": 0.3449171483516693, "global_step": 50253, "epoch": 564, "lr": 9.190830328590763e-05}
{"train_loss": 0.3852938413619995, "global_step": 50254, "epoch": 564, "lr": 9.190798708598516e-05}
{"train_loss": 0.27623724937438965, "global_step": 50255, "epoch": 564, "lr": 9.190767088042867e-05}
{"train_loss": 0.3214896619319916, "global_step": 50256, "epoch": 564, "lr": 9.19073546692382e-05}
{"train_loss": 0.31377944350242615, "global_step": 50257, "epoch": 564, "lr": 9.19070384524138e-05}
{"train_loss": 0.23021817207336426, "global_step": 50258, "epoch": 564, "lr": 9.190672222995551e-05}
{"train_loss": 0.40293988585472107, "global_step": 50259, "epoch": 564, "lr": 9.190640600186338e-05}
{"train_loss": 0.28452587127685547, "global_step": 50260, "epoch": 564, "lr": 9.190608976813743e-05}
{"train_loss": 0.2940581440925598, "global_step": 50261, "epoch": 564, "lr": 9.190577352877773e-05}
{"train_loss": 0.2758640944957733, "global_step": 50262, "epoch": 564, "lr": 9.19054572837843e-05}
{"train_loss": 0.2945319712162018, "global_step": 50263, "epoch": 564, "lr": 9.19051410331572e-05}
{"train_loss": 0.321077823638916, "global_step": 50264, "epoch": 564, "lr": 9.190482477689646e-05}
{"train_loss": 0.2829432189464569, "global_step": 50265, "epoch": 564, "lr": 9.190450851500211e-05}
{"train_loss": 0.37860703468322754, "global_step": 50266, "epoch": 564, "lr": 9.190419224747424e-05}
{"train_loss": 0.42406725883483887, "global_step": 50267, "epoch": 564, "lr": 9.190387597431285e-05}
{"train_loss": 0.27456945180892944, "global_step": 50268, "epoch": 564, "lr": 9.190355969551798e-05}
{"train_loss": 0.3124050796031952, "global_step": 50269, "epoch": 564, "lr": 9.190324341108971e-05}
{"train_loss": 0.32942995429039, "global_step": 50270, "epoch": 564, "lr": 9.190292712102805e-05}
{"train_loss": 0.26704657077789307, "global_step": 50271, "epoch": 564, "lr": 9.190261082533305e-05}
{"train_loss": 0.30918070673942566, "global_step": 50272, "epoch": 564, "lr": 9.190229452400476e-05}
{"train_loss": 0.1988334357738495, "global_step": 50273, "epoch": 564, "lr": 9.19019782170432e-05}
{"train_loss": 0.33079636096954346, "global_step": 50274, "epoch": 564, "lr": 9.190166190444845e-05}
{"train_loss": 0.3114513158798218, "global_step": 50275, "epoch": 564, "lr": 9.190134558622052e-05}
{"train_loss": 0.2788548171520233, "global_step": 50276, "epoch": 564, "lr": 9.190102926235949e-05}
{"train_loss": 0.24476373195648193, "global_step": 50277, "epoch": 564, "lr": 9.190071293286536e-05}
{"train_loss": 0.25281378626823425, "global_step": 50278, "epoch": 564, "lr": 9.190039659773817e-05}
{"train_loss": 0.2717403471469879, "global_step": 50279, "epoch": 564, "lr": 9.190008025697801e-05}
{"train_loss": 0.23501741886138916, "global_step": 50280, "epoch": 564, "lr": 9.189976391058488e-05}
{"train_loss": 0.38241398334503174, "global_step": 50281, "epoch": 564, "lr": 9.189944755855884e-05}
{"train_loss": 0.31479570269584656, "global_step": 50282, "epoch": 564, "lr": 9.189913120089993e-05}
{"train_loss": 0.3691083788871765, "global_step": 50283, "epoch": 564, "lr": 9.18988148376082e-05}
{"train_loss": 0.2773668375912677, "global_step": 50284, "epoch": 564, "lr": 9.189849846868368e-05, "val_loss": 2.8852386474609375}
{"train_loss": 0.20515258610248566, "global_step": 50285, "epoch": 565, "lr": 9.189818209412642e-05}
{"train_loss": 0.2449323534965515, "global_step": 50286, "epoch": 565, "lr": 9.189786571393644e-05}
{"train_loss": 0.2753504812717438, "global_step": 50287, "epoch": 565, "lr": 9.189754932811383e-05}
{"train_loss": 0.24521666765213013, "global_step": 50288, "epoch": 565, "lr": 9.18972329366586e-05}
{"train_loss": 0.3127087354660034, "global_step": 50289, "epoch": 565, "lr": 9.189691653957079e-05}
{"train_loss": 0.2668377459049225, "global_step": 50290, "epoch": 565, "lr": 9.189660013685044e-05}
{"train_loss": 0.27677759528160095, "global_step": 50291, "epoch": 565, "lr": 9.189628372849764e-05}
{"train_loss": 0.2862685024738312, "global_step": 50292, "epoch": 565, "lr": 9.189596731451236e-05}
{"train_loss": 0.30757418274879456, "global_step": 50293, "epoch": 565, "lr": 9.189565089489469e-05}
{"train_loss": 0.1962251514196396, "global_step": 50294, "epoch": 565, "lr": 9.189533446964466e-05}
{"train_loss": 0.2411889284849167, "global_step": 50295, "epoch": 565, "lr": 9.18950180387623e-05}
{"train_loss": 0.2886257469654083, "global_step": 50296, "epoch": 565, "lr": 9.189470160224769e-05}
{"train_loss": 0.29947608709335327, "global_step": 50297, "epoch": 565, "lr": 9.189438516010083e-05}
{"train_loss": 0.19620554149150848, "global_step": 50298, "epoch": 565, "lr": 9.189406871232177e-05}
{"train_loss": 0.192450150847435, "global_step": 50299, "epoch": 565, "lr": 9.18937522589106e-05}
{"train_loss": 0.19719602167606354, "global_step": 50300, "epoch": 565, "lr": 9.18934357998673e-05}
{"train_loss": 0.25378358364105225, "global_step": 50301, "epoch": 565, "lr": 9.189311933519195e-05}
{"train_loss": 0.3025367558002472, "global_step": 50302, "epoch": 565, "lr": 9.189280286488458e-05}
{"train_loss": 0.3044779896736145, "global_step": 50303, "epoch": 565, "lr": 9.189248638894521e-05}
{"train_loss": 0.4108918607234955, "global_step": 50304, "epoch": 565, "lr": 9.189216990737393e-05}
{"train_loss": 0.28333577513694763, "global_step": 50305, "epoch": 565, "lr": 9.189185342017075e-05}
{"train_loss": 0.1980780065059662, "global_step": 50306, "epoch": 565, "lr": 9.189153692733571e-05}
{"train_loss": 0.28066205978393555, "global_step": 50307, "epoch": 565, "lr": 9.189122042886889e-05}
{"train_loss": 0.21354752779006958, "global_step": 50308, "epoch": 565, "lr": 9.189090392477028e-05}
{"train_loss": 0.26135900616645813, "global_step": 50309, "epoch": 565, "lr": 9.189058741503997e-05}
{"train_loss": 0.2520389258861542, "global_step": 50310, "epoch": 565, "lr": 9.189027089967796e-05}
{"train_loss": 0.23484766483306885, "global_step": 50311, "epoch": 565, "lr": 9.188995437868433e-05}
{"train_loss": 0.293648362159729, "global_step": 50312, "epoch": 565, "lr": 9.188963785205911e-05}
{"train_loss": 0.2657703161239624, "global_step": 50313, "epoch": 565, "lr": 9.188932131980232e-05}
{"train_loss": 0.1385578215122223, "global_step": 50314, "epoch": 565, "lr": 9.188900478191403e-05}
{"train_loss": 0.3954756557941437, "global_step": 50315, "epoch": 565, "lr": 9.188868823839428e-05}
{"train_loss": 0.31875118613243103, "global_step": 50316, "epoch": 565, "lr": 9.18883716892431e-05}
{"train_loss": 0.2716737687587738, "global_step": 50317, "epoch": 565, "lr": 9.188805513446054e-05}
{"train_loss": 0.28903865814208984, "global_step": 50318, "epoch": 565, "lr": 9.188773857404664e-05}
{"train_loss": 0.29485756158828735, "global_step": 50319, "epoch": 565, "lr": 9.188742200800144e-05}
{"train_loss": 0.2528603971004486, "global_step": 50320, "epoch": 565, "lr": 9.1887105436325e-05}
{"train_loss": 0.17785194516181946, "global_step": 50321, "epoch": 565, "lr": 9.188678885901734e-05}
{"train_loss": 0.2651958763599396, "global_step": 50322, "epoch": 565, "lr": 9.188647227607852e-05}
{"train_loss": 0.22457225620746613, "global_step": 50323, "epoch": 565, "lr": 9.188615568750857e-05}
{"train_loss": 0.3367166221141815, "global_step": 50324, "epoch": 565, "lr": 9.188583909330752e-05}
{"train_loss": 0.2929897606372833, "global_step": 50325, "epoch": 565, "lr": 9.188552249347545e-05}
{"train_loss": 0.25235500931739807, "global_step": 50326, "epoch": 565, "lr": 9.188520588801237e-05}
{"train_loss": 0.22383815050125122, "global_step": 50327, "epoch": 565, "lr": 9.188488927691836e-05}
{"train_loss": 0.22945614159107208, "global_step": 50328, "epoch": 565, "lr": 9.188457266019341e-05}
{"train_loss": 0.16311617195606232, "global_step": 50329, "epoch": 565, "lr": 9.18842560378376e-05}
{"train_loss": 0.2406974583864212, "global_step": 50330, "epoch": 565, "lr": 9.188393940985096e-05}
{"train_loss": 0.3115982711315155, "global_step": 50331, "epoch": 565, "lr": 9.188362277623354e-05}
{"train_loss": 0.3954373896121979, "global_step": 50332, "epoch": 565, "lr": 9.188330613698536e-05}
{"train_loss": 0.22121088206768036, "global_step": 50333, "epoch": 565, "lr": 9.18829894921065e-05}
{"train_loss": 0.1964278668165207, "global_step": 50334, "epoch": 565, "lr": 9.188267284159697e-05}
{"train_loss": 0.2929084897041321, "global_step": 50335, "epoch": 565, "lr": 9.188235618545683e-05}
{"train_loss": 0.3211553990840912, "global_step": 50336, "epoch": 565, "lr": 9.188203952368611e-05}
{"train_loss": 0.25960057973861694, "global_step": 50337, "epoch": 565, "lr": 9.188172285628486e-05}
{"train_loss": 0.30869168043136597, "global_step": 50338, "epoch": 565, "lr": 9.188140618325314e-05}
{"train_loss": 0.18100132048130035, "global_step": 50339, "epoch": 565, "lr": 9.188108950459098e-05}
{"train_loss": 0.20364654064178467, "global_step": 50340, "epoch": 565, "lr": 9.18807728202984e-05}
{"train_loss": 0.2908615171909332, "global_step": 50341, "epoch": 565, "lr": 9.188045613037546e-05}
{"train_loss": 0.26868680119514465, "global_step": 50342, "epoch": 565, "lr": 9.18801394348222e-05}
{"train_loss": 0.300947904586792, "global_step": 50343, "epoch": 565, "lr": 9.187982273363868e-05}
{"train_loss": 0.38163718581199646, "global_step": 50344, "epoch": 565, "lr": 9.187950602682493e-05}
{"train_loss": 0.26774320006370544, "global_step": 50345, "epoch": 565, "lr": 9.187918931438098e-05}
{"train_loss": 0.203800767660141, "global_step": 50346, "epoch": 565, "lr": 9.187887259630687e-05}
{"train_loss": 0.2446855902671814, "global_step": 50347, "epoch": 565, "lr": 9.187855587260268e-05}
{"train_loss": 0.2604076862335205, "global_step": 50348, "epoch": 565, "lr": 9.187823914326842e-05}
{"train_loss": 0.3712176978588104, "global_step": 50349, "epoch": 565, "lr": 9.187792240830414e-05}
{"train_loss": 0.2903626561164856, "global_step": 50350, "epoch": 565, "lr": 9.187760566770989e-05}
{"train_loss": 0.38924264907836914, "global_step": 50351, "epoch": 565, "lr": 9.187728892148571e-05}
{"train_loss": 0.26378965377807617, "global_step": 50352, "epoch": 565, "lr": 9.187697216963162e-05}
{"train_loss": 0.3182435631752014, "global_step": 50353, "epoch": 565, "lr": 9.187665541214769e-05}
{"train_loss": 0.3045920133590698, "global_step": 50354, "epoch": 565, "lr": 9.187633864903396e-05}
{"train_loss": 0.361221581697464, "global_step": 50355, "epoch": 565, "lr": 9.187602188029047e-05}
{"train_loss": 0.20895186066627502, "global_step": 50356, "epoch": 565, "lr": 9.187570510591725e-05}
{"train_loss": 0.3438807725906372, "global_step": 50357, "epoch": 565, "lr": 9.187538832591435e-05}
{"train_loss": 0.2807965576648712, "global_step": 50358, "epoch": 565, "lr": 9.18750715402818e-05}
{"train_loss": 0.26695516705513, "global_step": 50359, "epoch": 565, "lr": 9.187475474901969e-05}
{"train_loss": 0.19484448432922363, "global_step": 50360, "epoch": 565, "lr": 9.187443795212802e-05}
{"train_loss": 0.26957178115844727, "global_step": 50361, "epoch": 565, "lr": 9.187412114960683e-05}
{"train_loss": 0.17948471009731293, "global_step": 50362, "epoch": 565, "lr": 9.187380434145618e-05}
{"train_loss": 0.24824625253677368, "global_step": 50363, "epoch": 565, "lr": 9.187348752767611e-05}
{"train_loss": 0.3102477788925171, "global_step": 50364, "epoch": 565, "lr": 9.187317070826666e-05}
{"train_loss": 0.15598513185977936, "global_step": 50365, "epoch": 565, "lr": 9.187285388322788e-05}
{"train_loss": 0.31705912947654724, "global_step": 50366, "epoch": 565, "lr": 9.187253705255979e-05}
{"train_loss": 0.3187609314918518, "global_step": 50367, "epoch": 565, "lr": 9.187222021626247e-05}
{"train_loss": 0.27579084038734436, "global_step": 50368, "epoch": 565, "lr": 9.187190337433592e-05}
{"train_loss": 0.21520917117595673, "global_step": 50369, "epoch": 565, "lr": 9.18715865267802e-05}
{"train_loss": 0.23518134653568268, "global_step": 50370, "epoch": 565, "lr": 9.187126967359537e-05}
{"train_loss": 0.25178056955337524, "global_step": 50371, "epoch": 565, "lr": 9.187095281478145e-05}
{"train_loss": 0.29267385601997375, "global_step": 50372, "epoch": 565, "lr": 9.187063595033849e-05}
{"train_loss": 0.2679195154583856, "global_step": 50373, "epoch": 565, "lr": 9.187031908026654e-05, "val_loss": 3.041661024093628, "train_action_mse_error": 22.827707290649414}
{"train_loss": 0.2741454243659973, "global_step": 50374, "epoch": 566, "lr": 9.187000220456563e-05}
{"train_loss": 0.23902423679828644, "global_step": 50375, "epoch": 566, "lr": 9.186968532323581e-05}
{"train_loss": 0.26080551743507385, "global_step": 50376, "epoch": 566, "lr": 9.186936843627712e-05}
{"train_loss": 0.25957080721855164, "global_step": 50377, "epoch": 566, "lr": 9.186905154368962e-05}
{"train_loss": 0.3492630124092102, "global_step": 50378, "epoch": 566, "lr": 9.186873464547332e-05}
{"train_loss": 0.24173331260681152, "global_step": 50379, "epoch": 566, "lr": 9.186841774162827e-05}
{"train_loss": 0.27035272121429443, "global_step": 50380, "epoch": 566, "lr": 9.186810083215454e-05}
{"train_loss": 0.2857847511768341, "global_step": 50381, "epoch": 566, "lr": 9.186778391705213e-05}
{"train_loss": 0.2957252860069275, "global_step": 50382, "epoch": 566, "lr": 9.186746699632112e-05}
{"train_loss": 0.2981249988079071, "global_step": 50383, "epoch": 566, "lr": 9.186715006996155e-05}
{"train_loss": 0.18033120036125183, "global_step": 50384, "epoch": 566, "lr": 9.186683313797344e-05}
{"train_loss": 0.2555907070636749, "global_step": 50385, "epoch": 566, "lr": 9.186651620035685e-05}
{"train_loss": 0.28988656401634216, "global_step": 50386, "epoch": 566, "lr": 9.186619925711182e-05}
{"train_loss": 0.4290272295475006, "global_step": 50387, "epoch": 566, "lr": 9.186588230823838e-05}
{"train_loss": 0.33686327934265137, "global_step": 50388, "epoch": 566, "lr": 9.18655653537366e-05}
{"train_loss": 0.34330126643180847, "global_step": 50389, "epoch": 566, "lr": 9.186524839360648e-05}
{"train_loss": 0.27959659695625305, "global_step": 50390, "epoch": 566, "lr": 9.186493142784811e-05}
{"train_loss": 0.40463945269584656, "global_step": 50391, "epoch": 566, "lr": 9.18646144564615e-05}
{"train_loss": 0.2186712920665741, "global_step": 50392, "epoch": 566, "lr": 9.18642974794467e-05}
{"train_loss": 0.2388932704925537, "global_step": 50393, "epoch": 566, "lr": 9.186398049680376e-05}
{"train_loss": 0.25618648529052734, "global_step": 50394, "epoch": 566, "lr": 9.186366350853273e-05}
{"train_loss": 0.31709742546081543, "global_step": 50395, "epoch": 566, "lr": 9.186334651463362e-05}
{"train_loss": 0.36759835481643677, "global_step": 50396, "epoch": 566, "lr": 9.18630295151065e-05}
{"train_loss": 0.29116329550743103, "global_step": 50397, "epoch": 566, "lr": 9.18627125099514e-05}
{"train_loss": 0.3759413957595825, "global_step": 50398, "epoch": 566, "lr": 9.186239549916838e-05}
{"train_loss": 0.418206125497818, "global_step": 50399, "epoch": 566, "lr": 9.186207848275748e-05}
{"train_loss": 0.21878860890865326, "global_step": 50400, "epoch": 566, "lr": 9.18617614607187e-05}
{"train_loss": 0.259246289730072, "global_step": 50401, "epoch": 566, "lr": 9.186144443305215e-05}
{"train_loss": 0.2574721574783325, "global_step": 50402, "epoch": 566, "lr": 9.186112739975782e-05}
{"train_loss": 0.2956436276435852, "global_step": 50403, "epoch": 566, "lr": 9.186081036083579e-05}
{"train_loss": 0.3040917217731476, "global_step": 50404, "epoch": 566, "lr": 9.186049331628606e-05}
{"train_loss": 0.3148627281188965, "global_step": 50405, "epoch": 566, "lr": 9.186017626610872e-05}
{"train_loss": 0.23801569640636444, "global_step": 50406, "epoch": 566, "lr": 9.185985921030378e-05}
{"train_loss": 0.28459206223487854, "global_step": 50407, "epoch": 566, "lr": 9.185954214887128e-05}
{"train_loss": 0.2639533579349518, "global_step": 50408, "epoch": 566, "lr": 9.185922508181128e-05}
{"train_loss": 0.314396470785141, "global_step": 50409, "epoch": 566, "lr": 9.185890800912382e-05}
{"train_loss": 0.26048988103866577, "global_step": 50410, "epoch": 566, "lr": 9.185859093080895e-05}
{"train_loss": 0.28252267837524414, "global_step": 50411, "epoch": 566, "lr": 9.185827384686669e-05}
{"train_loss": 0.35023021697998047, "global_step": 50412, "epoch": 566, "lr": 9.185795675729711e-05}
{"train_loss": 0.26312339305877686, "global_step": 50413, "epoch": 566, "lr": 9.185763966210022e-05}
{"train_loss": 0.32935580611228943, "global_step": 50414, "epoch": 566, "lr": 9.18573225612761e-05}
{"train_loss": 0.25728222727775574, "global_step": 50415, "epoch": 566, "lr": 9.185700545482477e-05}
{"train_loss": 0.2669975459575653, "global_step": 50416, "epoch": 566, "lr": 9.185668834274626e-05}
{"train_loss": 0.3143109083175659, "global_step": 50417, "epoch": 566, "lr": 9.185637122504063e-05}
{"train_loss": 0.29911890625953674, "global_step": 50418, "epoch": 566, "lr": 9.185605410170792e-05}
{"train_loss": 0.2891892194747925, "global_step": 50419, "epoch": 566, "lr": 9.185573697274819e-05}
{"train_loss": 0.2637917697429657, "global_step": 50420, "epoch": 566, "lr": 9.185541983816146e-05}
{"train_loss": 0.2520153224468231, "global_step": 50421, "epoch": 566, "lr": 9.185510269794777e-05}
{"train_loss": 0.2852121591567993, "global_step": 50422, "epoch": 566, "lr": 9.185478555210719e-05}
{"train_loss": 0.2808532118797302, "global_step": 50423, "epoch": 566, "lr": 9.185446840063973e-05}
{"train_loss": 0.3079320788383484, "global_step": 50424, "epoch": 566, "lr": 9.185415124354544e-05}
{"train_loss": 0.24116286635398865, "global_step": 50425, "epoch": 566, "lr": 9.185383408082438e-05}
{"train_loss": 0.2610648274421692, "global_step": 50426, "epoch": 566, "lr": 9.185351691247658e-05}
{"train_loss": 0.2741954028606415, "global_step": 50427, "epoch": 566, "lr": 9.185319973850208e-05}
{"train_loss": 0.21688546240329742, "global_step": 50428, "epoch": 566, "lr": 9.185288255890095e-05}
{"train_loss": 0.3052472174167633, "global_step": 50429, "epoch": 566, "lr": 9.185256537367319e-05}
{"train_loss": 0.3374670743942261, "global_step": 50430, "epoch": 566, "lr": 9.185224818281886e-05}
{"train_loss": 0.32317861914634705, "global_step": 50431, "epoch": 566, "lr": 9.185193098633801e-05}
{"train_loss": 0.27262768149375916, "global_step": 50432, "epoch": 566, "lr": 9.185161378423069e-05}
{"train_loss": 0.32094860076904297, "global_step": 50433, "epoch": 566, "lr": 9.185129657649692e-05}
{"train_loss": 0.2876099944114685, "global_step": 50434, "epoch": 566, "lr": 9.185097936313674e-05}
{"train_loss": 0.2918343245983124, "global_step": 50435, "epoch": 566, "lr": 9.185066214415022e-05}
{"train_loss": 0.20516832172870636, "global_step": 50436, "epoch": 566, "lr": 9.185034491953739e-05}
{"train_loss": 0.2741563618183136, "global_step": 50437, "epoch": 566, "lr": 9.185002768929829e-05}
{"train_loss": 0.19433625042438507, "global_step": 50438, "epoch": 566, "lr": 9.184971045343296e-05}
{"train_loss": 0.25859832763671875, "global_step": 50439, "epoch": 566, "lr": 9.184939321194145e-05}
{"train_loss": 0.21693629026412964, "global_step": 50440, "epoch": 566, "lr": 9.18490759648238e-05}
{"train_loss": 0.30423685908317566, "global_step": 50441, "epoch": 566, "lr": 9.184875871208005e-05}
{"train_loss": 0.2883414626121521, "global_step": 50442, "epoch": 566, "lr": 9.184844145371024e-05}
{"train_loss": 0.24464178085327148, "global_step": 50443, "epoch": 566, "lr": 9.184812418971443e-05}
{"train_loss": 0.24204866588115692, "global_step": 50444, "epoch": 566, "lr": 9.184780692009262e-05}
{"train_loss": 0.20607686042785645, "global_step": 50445, "epoch": 566, "lr": 9.184748964484491e-05}
{"train_loss": 0.3083551526069641, "global_step": 50446, "epoch": 566, "lr": 9.184717236397131e-05}
{"train_loss": 0.20034536719322205, "global_step": 50447, "epoch": 566, "lr": 9.184685507747187e-05}
{"train_loss": 0.1893806755542755, "global_step": 50448, "epoch": 566, "lr": 9.184653778534662e-05}
{"train_loss": 0.28244513273239136, "global_step": 50449, "epoch": 566, "lr": 9.184622048759563e-05}
{"train_loss": 0.2453286051750183, "global_step": 50450, "epoch": 566, "lr": 9.18459031842189e-05}
{"train_loss": 0.2698182761669159, "global_step": 50451, "epoch": 566, "lr": 9.184558587521652e-05}
{"train_loss": 0.38668546080589294, "global_step": 50452, "epoch": 566, "lr": 9.184526856058851e-05}
{"train_loss": 0.31730958819389343, "global_step": 50453, "epoch": 566, "lr": 9.184495124033491e-05}
{"train_loss": 0.22708354890346527, "global_step": 50454, "epoch": 566, "lr": 9.184463391445577e-05}
{"train_loss": 0.2964118719100952, "global_step": 50455, "epoch": 566, "lr": 9.184431658295112e-05}
{"train_loss": 0.22780565917491913, "global_step": 50456, "epoch": 566, "lr": 9.184399924582101e-05}
{"train_loss": 0.20964829623699188, "global_step": 50457, "epoch": 566, "lr": 9.184368190306549e-05}
{"train_loss": 0.2102247178554535, "global_step": 50458, "epoch": 566, "lr": 9.18433645546846e-05}
{"train_loss": 0.33202365040779114, "global_step": 50459, "epoch": 566, "lr": 9.184304720067838e-05}
{"train_loss": 0.24963653087615967, "global_step": 50460, "epoch": 566, "lr": 9.184272984104687e-05}
{"train_loss": 0.27140820026397705, "global_step": 50461, "epoch": 566, "lr": 9.18424124757901e-05}
{"train_loss": 0.27862028191598615, "global_step": 50462, "epoch": 566, "lr": 9.184209510490814e-05, "val_loss": 3.093940019607544}
{"train_loss": 0.34059759974479675, "global_step": 50463, "epoch": 567, "lr": 9.184177772840104e-05}
{"train_loss": 0.17577049136161804, "global_step": 50464, "epoch": 567, "lr": 9.18414603462688e-05}
{"train_loss": 0.2560141384601593, "global_step": 50465, "epoch": 567, "lr": 9.184114295851149e-05}
{"train_loss": 0.27979642152786255, "global_step": 50466, "epoch": 567, "lr": 9.184082556512915e-05}
{"train_loss": 0.24311929941177368, "global_step": 50467, "epoch": 567, "lr": 9.184050816612181e-05}
{"train_loss": 0.2077290415763855, "global_step": 50468, "epoch": 567, "lr": 9.184019076148953e-05}
{"train_loss": 0.3500196933746338, "global_step": 50469, "epoch": 567, "lr": 9.183987335123236e-05}
{"train_loss": 0.18611189723014832, "global_step": 50470, "epoch": 567, "lr": 9.183955593535031e-05}
{"train_loss": 0.18744821846485138, "global_step": 50471, "epoch": 567, "lr": 9.183923851384345e-05}
{"train_loss": 0.2903585135936737, "global_step": 50472, "epoch": 567, "lr": 9.183892108671181e-05}
{"train_loss": 0.2577259838581085, "global_step": 50473, "epoch": 567, "lr": 9.183860365395545e-05}
{"train_loss": 0.21190525591373444, "global_step": 50474, "epoch": 567, "lr": 9.183828621557439e-05}
{"train_loss": 0.19622816145420074, "global_step": 50475, "epoch": 567, "lr": 9.183796877156868e-05}
{"train_loss": 0.284993052482605, "global_step": 50476, "epoch": 567, "lr": 9.183765132193838e-05}
{"train_loss": 0.29233217239379883, "global_step": 50477, "epoch": 567, "lr": 9.18373338666835e-05}
{"train_loss": 0.28935912251472473, "global_step": 50478, "epoch": 567, "lr": 9.18370164058041e-05}
{"train_loss": 0.24693605303764343, "global_step": 50479, "epoch": 567, "lr": 9.183669893930023e-05}
{"train_loss": 0.26454436779022217, "global_step": 50480, "epoch": 567, "lr": 9.183638146717193e-05}
{"train_loss": 0.24107186496257782, "global_step": 50481, "epoch": 567, "lr": 9.183606398941922e-05}
{"train_loss": 0.218715101480484, "global_step": 50482, "epoch": 567, "lr": 9.183574650604218e-05}
{"train_loss": 0.19845062494277954, "global_step": 50483, "epoch": 567, "lr": 9.183542901704084e-05}
{"train_loss": 0.29576608538627625, "global_step": 50484, "epoch": 567, "lr": 9.183511152241522e-05}
{"train_loss": 0.28394952416419983, "global_step": 50485, "epoch": 567, "lr": 9.183479402216539e-05}
{"train_loss": 0.23944422602653503, "global_step": 50486, "epoch": 567, "lr": 9.183447651629138e-05}
{"train_loss": 0.19878527522087097, "global_step": 50487, "epoch": 567, "lr": 9.183415900479322e-05}
{"train_loss": 0.30281850695610046, "global_step": 50488, "epoch": 567, "lr": 9.183384148767098e-05}
{"train_loss": 0.3323352634906769, "global_step": 50489, "epoch": 567, "lr": 9.183352396492468e-05}
{"train_loss": 0.3064612150192261, "global_step": 50490, "epoch": 567, "lr": 9.183320643655439e-05}
{"train_loss": 0.2430024892091751, "global_step": 50491, "epoch": 567, "lr": 9.183288890256012e-05}
{"train_loss": 0.31891581416130066, "global_step": 50492, "epoch": 567, "lr": 9.183257136294194e-05}
{"train_loss": 0.30806154012680054, "global_step": 50493, "epoch": 567, "lr": 9.183225381769986e-05}
{"train_loss": 0.2526153028011322, "global_step": 50494, "epoch": 567, "lr": 9.183193626683397e-05}
{"train_loss": 0.23884686827659607, "global_step": 50495, "epoch": 567, "lr": 9.183161871034427e-05}
{"train_loss": 0.27370190620422363, "global_step": 50496, "epoch": 567, "lr": 9.183130114823082e-05}
{"train_loss": 0.26597440242767334, "global_step": 50497, "epoch": 567, "lr": 9.183098358049367e-05}
{"train_loss": 0.18684601783752441, "global_step": 50498, "epoch": 567, "lr": 9.183066600713283e-05}
{"train_loss": 0.2574375867843628, "global_step": 50499, "epoch": 567, "lr": 9.183034842814839e-05}
{"train_loss": 0.2323436141014099, "global_step": 50500, "epoch": 567, "lr": 9.183003084354038e-05}
{"train_loss": 0.3012610971927643, "global_step": 50501, "epoch": 567, "lr": 9.182971325330881e-05}
{"train_loss": 0.2637130320072174, "global_step": 50502, "epoch": 567, "lr": 9.182939565745376e-05}
{"train_loss": 0.28527843952178955, "global_step": 50503, "epoch": 567, "lr": 9.182907805597525e-05}
{"train_loss": 0.2690073847770691, "global_step": 50504, "epoch": 567, "lr": 9.182876044887332e-05}
{"train_loss": 0.26549074053764343, "global_step": 50505, "epoch": 567, "lr": 9.182844283614803e-05}
{"train_loss": 0.2801420986652374, "global_step": 50506, "epoch": 567, "lr": 9.182812521779942e-05}
{"train_loss": 0.2876315414905548, "global_step": 50507, "epoch": 567, "lr": 9.182780759382752e-05}
{"train_loss": 0.3108940124511719, "global_step": 50508, "epoch": 567, "lr": 9.182748996423238e-05}
{"train_loss": 0.32016727328300476, "global_step": 50509, "epoch": 567, "lr": 9.182717232901406e-05}
{"train_loss": 0.24904297292232513, "global_step": 50510, "epoch": 567, "lr": 9.182685468817258e-05}
{"train_loss": 0.2668456733226776, "global_step": 50511, "epoch": 567, "lr": 9.182653704170798e-05}
{"train_loss": 0.24789057672023773, "global_step": 50512, "epoch": 567, "lr": 9.182621938962033e-05}
{"train_loss": 0.3028053939342499, "global_step": 50513, "epoch": 567, "lr": 9.182590173190964e-05}
{"train_loss": 0.2984429597854614, "global_step": 50514, "epoch": 567, "lr": 9.182558406857597e-05}
{"train_loss": 0.32250112295150757, "global_step": 50515, "epoch": 567, "lr": 9.182526639961936e-05}
{"train_loss": 0.4214821457862854, "global_step": 50516, "epoch": 567, "lr": 9.182494872503985e-05}
{"train_loss": 0.26891589164733887, "global_step": 50517, "epoch": 567, "lr": 9.18246310448375e-05}
{"train_loss": 0.21297460794448853, "global_step": 50518, "epoch": 567, "lr": 9.182431335901232e-05}
{"train_loss": 0.2656688988208771, "global_step": 50519, "epoch": 567, "lr": 9.182399566756437e-05}
{"train_loss": 0.2411719560623169, "global_step": 50520, "epoch": 567, "lr": 9.18236779704937e-05}
{"train_loss": 0.22339366376399994, "global_step": 50521, "epoch": 567, "lr": 9.182336026780035e-05}
{"train_loss": 0.2746870219707489, "global_step": 50522, "epoch": 567, "lr": 9.182304255948436e-05}
{"train_loss": 0.34699711203575134, "global_step": 50523, "epoch": 567, "lr": 9.182272484554577e-05}
{"train_loss": 0.1757640391588211, "global_step": 50524, "epoch": 567, "lr": 9.182240712598462e-05}
{"train_loss": 0.2497946172952652, "global_step": 50525, "epoch": 567, "lr": 9.182208940080095e-05}
{"train_loss": 0.22635386884212494, "global_step": 50526, "epoch": 567, "lr": 9.182177166999483e-05}
{"train_loss": 0.29482629895210266, "global_step": 50527, "epoch": 567, "lr": 9.182145393356627e-05}
{"train_loss": 0.2501998543739319, "global_step": 50528, "epoch": 567, "lr": 9.182113619151533e-05}
{"train_loss": 0.26461294293403625, "global_step": 50529, "epoch": 567, "lr": 9.182081844384203e-05}
{"train_loss": 0.25189000368118286, "global_step": 50530, "epoch": 567, "lr": 9.182050069054646e-05}
{"train_loss": 0.22581526637077332, "global_step": 50531, "epoch": 567, "lr": 9.182018293162862e-05}
{"train_loss": 0.22665539383888245, "global_step": 50532, "epoch": 567, "lr": 9.181986516708856e-05}
{"train_loss": 0.30387240648269653, "global_step": 50533, "epoch": 567, "lr": 9.181954739692633e-05}
{"train_loss": 0.1643042415380478, "global_step": 50534, "epoch": 567, "lr": 9.181922962114198e-05}
{"train_loss": 0.3830884099006653, "global_step": 50535, "epoch": 567, "lr": 9.181891183973555e-05}
{"train_loss": 0.2897411286830902, "global_step": 50536, "epoch": 567, "lr": 9.181859405270706e-05}
{"train_loss": 0.24823708832263947, "global_step": 50537, "epoch": 567, "lr": 9.181827626005657e-05}
{"train_loss": 0.25525057315826416, "global_step": 50538, "epoch": 567, "lr": 9.181795846178413e-05}
{"train_loss": 0.3730112314224243, "global_step": 50539, "epoch": 567, "lr": 9.181764065788978e-05}
{"train_loss": 0.24660331010818481, "global_step": 50540, "epoch": 567, "lr": 9.181732284837354e-05}
{"train_loss": 0.2999587655067444, "global_step": 50541, "epoch": 567, "lr": 9.18170050332355e-05}
{"train_loss": 0.25678592920303345, "global_step": 50542, "epoch": 567, "lr": 9.181668721247565e-05}
{"train_loss": 0.17083033919334412, "global_step": 50543, "epoch": 567, "lr": 9.181636938609406e-05}
{"train_loss": 0.29085227847099304, "global_step": 50544, "epoch": 567, "lr": 9.181605155409077e-05}
{"train_loss": 0.27692946791648865, "global_step": 50545, "epoch": 567, "lr": 9.181573371646583e-05}
{"train_loss": 0.20849376916885376, "global_step": 50546, "epoch": 567, "lr": 9.181541587321926e-05}
{"train_loss": 0.2702963948249817, "global_step": 50547, "epoch": 567, "lr": 9.181509802435112e-05}
{"train_loss": 0.26573917269706726, "global_step": 50548, "epoch": 567, "lr": 9.181478016986145e-05}
{"train_loss": 0.2985168397426605, "global_step": 50549, "epoch": 567, "lr": 9.181446230975031e-05}
{"train_loss": 0.32158905267715454, "global_step": 50550, "epoch": 567, "lr": 9.18141444440177e-05}
{"train_loss": 0.26481543096263754, "global_step": 50551, "epoch": 567, "lr": 9.18138265726637e-05, "val_loss": 3.095334529876709}
{"train_loss": 0.26875409483909607, "global_step": 50552, "epoch": 568, "lr": 9.181350869568834e-05}
{"train_loss": 0.16732631623744965, "global_step": 50553, "epoch": 568, "lr": 9.181319081309167e-05}
{"train_loss": 0.26139405369758606, "global_step": 50554, "epoch": 568, "lr": 9.181287292487371e-05}
{"train_loss": 0.2059437334537506, "global_step": 50555, "epoch": 568, "lr": 9.181255503103453e-05}
{"train_loss": 0.21254996955394745, "global_step": 50556, "epoch": 568, "lr": 9.181223713157416e-05}
{"train_loss": 0.2566579580307007, "global_step": 50557, "epoch": 568, "lr": 9.181191922649263e-05}
{"train_loss": 0.23750291764736176, "global_step": 50558, "epoch": 568, "lr": 9.181160131579003e-05}
{"train_loss": 0.29916396737098694, "global_step": 50559, "epoch": 568, "lr": 9.181128339946635e-05}
{"train_loss": 0.27440178394317627, "global_step": 50560, "epoch": 568, "lr": 9.181096547752166e-05}
{"train_loss": 0.2177484929561615, "global_step": 50561, "epoch": 568, "lr": 9.181064754995599e-05}
{"train_loss": 0.22851380705833435, "global_step": 50562, "epoch": 568, "lr": 9.181032961676938e-05}
{"train_loss": 0.29392653703689575, "global_step": 50563, "epoch": 568, "lr": 9.181001167796188e-05}
{"train_loss": 0.2880042493343353, "global_step": 50564, "epoch": 568, "lr": 9.180969373353354e-05}
{"train_loss": 0.23045481741428375, "global_step": 50565, "epoch": 568, "lr": 9.180937578348441e-05}
{"train_loss": 0.28096064925193787, "global_step": 50566, "epoch": 568, "lr": 9.18090578278145e-05}
{"train_loss": 0.23076999187469482, "global_step": 50567, "epoch": 568, "lr": 9.180873986652387e-05}
{"train_loss": 0.20745393633842468, "global_step": 50568, "epoch": 568, "lr": 9.180842189961259e-05}
{"train_loss": 0.23803041875362396, "global_step": 50569, "epoch": 568, "lr": 9.180810392708066e-05}
{"train_loss": 0.3574570417404175, "global_step": 50570, "epoch": 568, "lr": 9.180778594892814e-05}
{"train_loss": 0.28619056940078735, "global_step": 50571, "epoch": 568, "lr": 9.180746796515507e-05}
{"train_loss": 0.24648848176002502, "global_step": 50572, "epoch": 568, "lr": 9.18071499757615e-05}
{"train_loss": 0.1872624009847641, "global_step": 50573, "epoch": 568, "lr": 9.180683198074746e-05}
{"train_loss": 0.2763933539390564, "global_step": 50574, "epoch": 568, "lr": 9.1806513980113e-05}
{"train_loss": 0.265816867351532, "global_step": 50575, "epoch": 568, "lr": 9.180619597385818e-05}
{"train_loss": 0.30811169743537903, "global_step": 50576, "epoch": 568, "lr": 9.180587796198301e-05}
{"train_loss": 0.20680813491344452, "global_step": 50577, "epoch": 568, "lr": 9.180555994448757e-05}
{"train_loss": 0.29472771286964417, "global_step": 50578, "epoch": 568, "lr": 9.180524192137186e-05}
{"train_loss": 0.2805987596511841, "global_step": 50579, "epoch": 568, "lr": 9.180492389263596e-05}
{"train_loss": 0.30211588740348816, "global_step": 50580, "epoch": 568, "lr": 9.180460585827988e-05}
{"train_loss": 0.2871091663837433, "global_step": 50581, "epoch": 568, "lr": 9.18042878183037e-05}
{"train_loss": 0.23892633616924286, "global_step": 50582, "epoch": 568, "lr": 9.180396977270742e-05}
{"train_loss": 0.35444366931915283, "global_step": 50583, "epoch": 568, "lr": 9.180365172149112e-05}
{"train_loss": 0.2795962989330292, "global_step": 50584, "epoch": 568, "lr": 9.180333366465484e-05}
{"train_loss": 0.26622238755226135, "global_step": 50585, "epoch": 568, "lr": 9.18030156021986e-05}
{"train_loss": 0.24698497354984283, "global_step": 50586, "epoch": 568, "lr": 9.180269753412245e-05}
{"train_loss": 0.22095054388046265, "global_step": 50587, "epoch": 568, "lr": 9.180237946042645e-05}
{"train_loss": 0.32366764545440674, "global_step": 50588, "epoch": 568, "lr": 9.180206138111062e-05}
{"train_loss": 0.1715909242630005, "global_step": 50589, "epoch": 568, "lr": 9.180174329617501e-05}
{"train_loss": 0.3098894953727722, "global_step": 50590, "epoch": 568, "lr": 9.180142520561967e-05}
{"train_loss": 0.30473005771636963, "global_step": 50591, "epoch": 568, "lr": 9.180110710944463e-05}
{"train_loss": 0.285739541053772, "global_step": 50592, "epoch": 568, "lr": 9.180078900764994e-05}
{"train_loss": 0.3417820930480957, "global_step": 50593, "epoch": 568, "lr": 9.180047090023566e-05}
{"train_loss": 0.21823987364768982, "global_step": 50594, "epoch": 568, "lr": 9.18001527872018e-05}
{"train_loss": 0.31766077876091003, "global_step": 50595, "epoch": 568, "lr": 9.179983466854842e-05}
{"train_loss": 0.3043912649154663, "global_step": 50596, "epoch": 568, "lr": 9.179951654427558e-05}
{"train_loss": 0.2793120741844177, "global_step": 50597, "epoch": 568, "lr": 9.179919841438329e-05}
{"train_loss": 0.22124817967414856, "global_step": 50598, "epoch": 568, "lr": 9.17988802788716e-05}
{"train_loss": 0.23457977175712585, "global_step": 50599, "epoch": 568, "lr": 9.179856213774058e-05}
{"train_loss": 0.29539182782173157, "global_step": 50600, "epoch": 568, "lr": 9.179824399099024e-05}
{"train_loss": 0.2760612666606903, "global_step": 50601, "epoch": 568, "lr": 9.179792583862064e-05}
{"train_loss": 0.41291043162345886, "global_step": 50602, "epoch": 568, "lr": 9.179760768063181e-05}
{"train_loss": 0.30773070454597473, "global_step": 50603, "epoch": 568, "lr": 9.17972895170238e-05}
{"train_loss": 0.2952434718608856, "global_step": 50604, "epoch": 568, "lr": 9.179697134779666e-05}
{"train_loss": 0.20725807547569275, "global_step": 50605, "epoch": 568, "lr": 9.179665317295043e-05}
{"train_loss": 0.25884872674942017, "global_step": 50606, "epoch": 568, "lr": 9.179633499248514e-05}
{"train_loss": 0.3712798058986664, "global_step": 50607, "epoch": 568, "lr": 9.179601680640085e-05}
{"train_loss": 0.23870572447776794, "global_step": 50608, "epoch": 568, "lr": 9.179569861469758e-05}
{"train_loss": 0.21017642319202423, "global_step": 50609, "epoch": 568, "lr": 9.179538041737541e-05}
{"train_loss": 0.28694862127304077, "global_step": 50610, "epoch": 568, "lr": 9.179506221443435e-05}
{"train_loss": 0.20385146141052246, "global_step": 50611, "epoch": 568, "lr": 9.179474400587447e-05}
{"train_loss": 0.2918321192264557, "global_step": 50612, "epoch": 568, "lr": 9.179442579169577e-05}
{"train_loss": 0.2636786103248596, "global_step": 50613, "epoch": 568, "lr": 9.179410757189832e-05}
{"train_loss": 0.29820647835731506, "global_step": 50614, "epoch": 568, "lr": 9.179378934648219e-05}
{"train_loss": 0.31311431527137756, "global_step": 50615, "epoch": 568, "lr": 9.179347111544737e-05}
{"train_loss": 0.21435609459877014, "global_step": 50616, "epoch": 568, "lr": 9.179315287879393e-05}
{"train_loss": 0.2352350503206253, "global_step": 50617, "epoch": 568, "lr": 9.17928346365219e-05}
{"train_loss": 0.2378702312707901, "global_step": 50618, "epoch": 568, "lr": 9.179251638863135e-05}
{"train_loss": 0.19358307123184204, "global_step": 50619, "epoch": 568, "lr": 9.179219813512231e-05}
{"train_loss": 0.2672111392021179, "global_step": 50620, "epoch": 568, "lr": 9.17918798759948e-05}
{"train_loss": 0.23900437355041504, "global_step": 50621, "epoch": 568, "lr": 9.17915616112489e-05}
{"train_loss": 0.2807976305484772, "global_step": 50622, "epoch": 568, "lr": 9.179124334088462e-05}
{"train_loss": 0.24223048985004425, "global_step": 50623, "epoch": 568, "lr": 9.179092506490204e-05}
{"train_loss": 0.23551657795906067, "global_step": 50624, "epoch": 568, "lr": 9.179060678330115e-05}
{"train_loss": 0.2518698573112488, "global_step": 50625, "epoch": 568, "lr": 9.179028849608203e-05}
{"train_loss": 0.17799024283885956, "global_step": 50626, "epoch": 568, "lr": 9.178997020324471e-05}
{"train_loss": 0.25874534249305725, "global_step": 50627, "epoch": 568, "lr": 9.178965190478925e-05}
{"train_loss": 0.2869865894317627, "global_step": 50628, "epoch": 568, "lr": 9.17893336007157e-05}
{"train_loss": 0.2407216727733612, "global_step": 50629, "epoch": 568, "lr": 9.178901529102403e-05}
{"train_loss": 0.347618967294693, "global_step": 50630, "epoch": 568, "lr": 9.178869697571438e-05}
{"train_loss": 0.2769683301448822, "global_step": 50631, "epoch": 568, "lr": 9.178837865478673e-05}
{"train_loss": 0.19806520640850067, "global_step": 50632, "epoch": 568, "lr": 9.178806032824114e-05}
{"train_loss": 0.28805413842201233, "global_step": 50633, "epoch": 568, "lr": 9.178774199607767e-05}
{"train_loss": 0.26011720299720764, "global_step": 50634, "epoch": 568, "lr": 9.178742365829633e-05}
{"train_loss": 0.24402473866939545, "global_step": 50635, "epoch": 568, "lr": 9.17871053148972e-05}
{"train_loss": 0.2423226535320282, "global_step": 50636, "epoch": 568, "lr": 9.17867869658803e-05}
{"train_loss": 0.2521416246891022, "global_step": 50637, "epoch": 568, "lr": 9.178646861124567e-05}
{"train_loss": 0.343059778213501, "global_step": 50638, "epoch": 568, "lr": 9.178615025099335e-05}
{"train_loss": 0.3461475074291229, "global_step": 50639, "epoch": 568, "lr": 9.17858318851234e-05}
{"train_loss": 0.26491747545392325, "global_step": 50640, "epoch": 568, "lr": 9.178551351363585e-05, "val_loss": 2.9892327785491943}
{"train_loss": 0.2362554520368576, "global_step": 50641, "epoch": 569, "lr": 9.178519513653074e-05}
{"train_loss": 0.2472044974565506, "global_step": 50642, "epoch": 569, "lr": 9.178487675380813e-05}
{"train_loss": 0.19958946108818054, "global_step": 50643, "epoch": 569, "lr": 9.178455836546805e-05}
{"train_loss": 0.21288950741291046, "global_step": 50644, "epoch": 569, "lr": 9.178423997151056e-05}
{"train_loss": 0.2889961004257202, "global_step": 50645, "epoch": 569, "lr": 9.178392157193566e-05}
{"train_loss": 0.26731258630752563, "global_step": 50646, "epoch": 569, "lr": 9.178360316674345e-05}
{"train_loss": 0.2163660079240799, "global_step": 50647, "epoch": 569, "lr": 9.178328475593393e-05}
{"train_loss": 0.2403319925069809, "global_step": 50648, "epoch": 569, "lr": 9.178296633950716e-05}
{"train_loss": 0.2170870155096054, "global_step": 50649, "epoch": 569, "lr": 9.178264791746317e-05}
{"train_loss": 0.2624148428440094, "global_step": 50650, "epoch": 569, "lr": 9.178232948980202e-05}
{"train_loss": 0.23671270906925201, "global_step": 50651, "epoch": 569, "lr": 9.178201105652375e-05}
{"train_loss": 0.23273782432079315, "global_step": 50652, "epoch": 569, "lr": 9.178169261762838e-05}
{"train_loss": 0.2534131705760956, "global_step": 50653, "epoch": 569, "lr": 9.178137417311599e-05}
{"train_loss": 0.2546325922012329, "global_step": 50654, "epoch": 569, "lr": 9.178105572298659e-05}
{"train_loss": 0.2341521829366684, "global_step": 50655, "epoch": 569, "lr": 9.178073726724025e-05}
{"train_loss": 0.2658846080303192, "global_step": 50656, "epoch": 569, "lr": 9.178041880587697e-05}
{"train_loss": 0.2430388182401657, "global_step": 50657, "epoch": 569, "lr": 9.178010033889685e-05}
{"train_loss": 0.2868621349334717, "global_step": 50658, "epoch": 569, "lr": 9.177978186629991e-05}
{"train_loss": 0.3023044764995575, "global_step": 50659, "epoch": 569, "lr": 9.177946338808618e-05}
{"train_loss": 0.19195224344730377, "global_step": 50660, "epoch": 569, "lr": 9.17791449042557e-05}
{"train_loss": 0.30119746923446655, "global_step": 50661, "epoch": 569, "lr": 9.177882641480852e-05}
{"train_loss": 0.18751557171344757, "global_step": 50662, "epoch": 569, "lr": 9.177850791974469e-05}
{"train_loss": 0.22012478113174438, "global_step": 50663, "epoch": 569, "lr": 9.177818941906427e-05}
{"train_loss": 0.2083008736371994, "global_step": 50664, "epoch": 569, "lr": 9.177787091276726e-05}
{"train_loss": 0.3404487669467926, "global_step": 50665, "epoch": 569, "lr": 9.177755240085372e-05}
{"train_loss": 0.2917717397212982, "global_step": 50666, "epoch": 569, "lr": 9.17772338833237e-05}
{"train_loss": 0.21132834255695343, "global_step": 50667, "epoch": 569, "lr": 9.177691536017726e-05}
{"train_loss": 0.18658801913261414, "global_step": 50668, "epoch": 569, "lr": 9.177659683141442e-05}
{"train_loss": 0.25574883818626404, "global_step": 50669, "epoch": 569, "lr": 9.177627829703521e-05}
{"train_loss": 0.2967628538608551, "global_step": 50670, "epoch": 569, "lr": 9.177595975703969e-05}
{"train_loss": 0.2336047887802124, "global_step": 50671, "epoch": 569, "lr": 9.17756412114279e-05}
{"train_loss": 0.26617416739463806, "global_step": 50672, "epoch": 569, "lr": 9.17753226601999e-05}
{"train_loss": 0.2865517735481262, "global_step": 50673, "epoch": 569, "lr": 9.177500410335571e-05}
{"train_loss": 0.3557400703430176, "global_step": 50674, "epoch": 569, "lr": 9.177468554089537e-05}
{"train_loss": 0.22472962737083435, "global_step": 50675, "epoch": 569, "lr": 9.177436697281893e-05}
{"train_loss": 0.2436448633670807, "global_step": 50676, "epoch": 569, "lr": 9.177404839912646e-05}
{"train_loss": 0.21939891576766968, "global_step": 50677, "epoch": 569, "lr": 9.177372981981796e-05}
{"train_loss": 0.23127368092536926, "global_step": 50678, "epoch": 569, "lr": 9.177341123489349e-05}
{"train_loss": 0.16890181601047516, "global_step": 50679, "epoch": 569, "lr": 9.17730926443531e-05}
{"train_loss": 0.2535950839519501, "global_step": 50680, "epoch": 569, "lr": 9.177277404819682e-05}
{"train_loss": 0.2700636088848114, "global_step": 50681, "epoch": 569, "lr": 9.177245544642472e-05}
{"train_loss": 0.4010002017021179, "global_step": 50682, "epoch": 569, "lr": 9.177213683903681e-05}
{"train_loss": 0.25780877470970154, "global_step": 50683, "epoch": 569, "lr": 9.177181822603314e-05}
{"train_loss": 0.25466927886009216, "global_step": 50684, "epoch": 569, "lr": 9.177149960741376e-05}
{"train_loss": 0.2687424421310425, "global_step": 50685, "epoch": 569, "lr": 9.17711809831787e-05}
{"train_loss": 0.30616533756256104, "global_step": 50686, "epoch": 569, "lr": 9.177086235332804e-05}
{"train_loss": 0.29495149850845337, "global_step": 50687, "epoch": 569, "lr": 9.177054371786177e-05}
{"train_loss": 0.24973635375499725, "global_step": 50688, "epoch": 569, "lr": 9.177022507677996e-05}
{"train_loss": 0.3593296408653259, "global_step": 50689, "epoch": 569, "lr": 9.176990643008268e-05}
{"train_loss": 0.3419491946697235, "global_step": 50690, "epoch": 569, "lr": 9.176958777776993e-05}
{"train_loss": 0.25851204991340637, "global_step": 50691, "epoch": 569, "lr": 9.176926911984176e-05}
{"train_loss": 0.3554912507534027, "global_step": 50692, "epoch": 569, "lr": 9.176895045629823e-05}
{"train_loss": 0.28932201862335205, "global_step": 50693, "epoch": 569, "lr": 9.176863178713938e-05}
{"train_loss": 0.30487748980522156, "global_step": 50694, "epoch": 569, "lr": 9.176831311236524e-05}
{"train_loss": 0.32109853625297546, "global_step": 50695, "epoch": 569, "lr": 9.176799443197586e-05}
{"train_loss": 0.3159753084182739, "global_step": 50696, "epoch": 569, "lr": 9.176767574597127e-05}
{"train_loss": 0.21590836346149445, "global_step": 50697, "epoch": 569, "lr": 9.176735705435153e-05}
{"train_loss": 0.3332962989807129, "global_step": 50698, "epoch": 569, "lr": 9.176703835711668e-05}
{"train_loss": 0.2905987501144409, "global_step": 50699, "epoch": 569, "lr": 9.176671965426678e-05}
{"train_loss": 0.305098295211792, "global_step": 50700, "epoch": 569, "lr": 9.176640094580183e-05}
{"train_loss": 0.335603803396225, "global_step": 50701, "epoch": 569, "lr": 9.176608223172191e-05}
{"train_loss": 0.36523404717445374, "global_step": 50702, "epoch": 569, "lr": 9.176576351202704e-05}
{"train_loss": 0.3044382333755493, "global_step": 50703, "epoch": 569, "lr": 9.176544478671728e-05}
{"train_loss": 0.2629333436489105, "global_step": 50704, "epoch": 569, "lr": 9.176512605579266e-05}
{"train_loss": 0.27792027592658997, "global_step": 50705, "epoch": 569, "lr": 9.176480731925324e-05}
{"train_loss": 0.4424933195114136, "global_step": 50706, "epoch": 569, "lr": 9.176448857709902e-05}
{"train_loss": 0.35966143012046814, "global_step": 50707, "epoch": 569, "lr": 9.17641698293301e-05}
{"train_loss": 0.34635820984840393, "global_step": 50708, "epoch": 569, "lr": 9.176385107594649e-05}
{"train_loss": 0.32026663422584534, "global_step": 50709, "epoch": 569, "lr": 9.176353231694823e-05}
{"train_loss": 0.26782622933387756, "global_step": 50710, "epoch": 569, "lr": 9.17632135523354e-05}
{"train_loss": 0.24275077879428864, "global_step": 50711, "epoch": 569, "lr": 9.176289478210798e-05}
{"train_loss": 0.24606141448020935, "global_step": 50712, "epoch": 569, "lr": 9.176257600626606e-05}
{"train_loss": 0.28252285718917847, "global_step": 50713, "epoch": 569, "lr": 9.176225722480968e-05}
{"train_loss": 0.3534058928489685, "global_step": 50714, "epoch": 569, "lr": 9.176193843773887e-05}
{"train_loss": 0.2886936664581299, "global_step": 50715, "epoch": 569, "lr": 9.176161964505367e-05}
{"train_loss": 0.3494213819503784, "global_step": 50716, "epoch": 569, "lr": 9.176130084675414e-05}
{"train_loss": 0.32051193714141846, "global_step": 50717, "epoch": 569, "lr": 9.176098204284029e-05}
{"train_loss": 0.2431672364473343, "global_step": 50718, "epoch": 569, "lr": 9.17606632333122e-05}
{"train_loss": 0.30191612243652344, "global_step": 50719, "epoch": 569, "lr": 9.17603444181699e-05}
{"train_loss": 0.22636303305625916, "global_step": 50720, "epoch": 569, "lr": 9.176002559741343e-05}
{"train_loss": 0.2676578164100647, "global_step": 50721, "epoch": 569, "lr": 9.175970677104284e-05}
{"train_loss": 0.24084992706775665, "global_step": 50722, "epoch": 569, "lr": 9.175938793905816e-05}
{"train_loss": 0.3269707262516022, "global_step": 50723, "epoch": 569, "lr": 9.175906910145945e-05}
{"train_loss": 0.2910218834877014, "global_step": 50724, "epoch": 569, "lr": 9.175875025824672e-05}
{"train_loss": 0.3104746341705322, "global_step": 50725, "epoch": 569, "lr": 9.175843140942004e-05}
{"train_loss": 0.3352796137332916, "global_step": 50726, "epoch": 569, "lr": 9.175811255497944e-05}
{"train_loss": 0.20819061994552612, "global_step": 50727, "epoch": 569, "lr": 9.175779369492499e-05}
{"train_loss": 0.3276987671852112, "global_step": 50728, "epoch": 569, "lr": 9.175747482925672e-05}
{"train_loss": 0.2766975455739525, "global_step": 50729, "epoch": 569, "lr": 9.175715595797465e-05, "val_loss": 2.956345319747925}
{"train_loss": 0.2867191731929779, "global_step": 50730, "epoch": 570, "lr": 9.175683708107884e-05}
{"train_loss": 0.31175801157951355, "global_step": 50731, "epoch": 570, "lr": 9.175651819856933e-05}
{"train_loss": 0.41173863410949707, "global_step": 50732, "epoch": 570, "lr": 9.175619931044618e-05}
{"train_loss": 0.29135262966156006, "global_step": 50733, "epoch": 570, "lr": 9.175588041670939e-05}
{"train_loss": 0.3012148439884186, "global_step": 50734, "epoch": 570, "lr": 9.175556151735906e-05}
{"train_loss": 0.26699724793434143, "global_step": 50735, "epoch": 570, "lr": 9.175524261239519e-05}
{"train_loss": 0.30392932891845703, "global_step": 50736, "epoch": 570, "lr": 9.175492370181784e-05}
{"train_loss": 0.33643022179603577, "global_step": 50737, "epoch": 570, "lr": 9.175460478562706e-05}
{"train_loss": 0.2679067850112915, "global_step": 50738, "epoch": 570, "lr": 9.175428586382286e-05}
{"train_loss": 0.2438139021396637, "global_step": 50739, "epoch": 570, "lr": 9.175396693640532e-05}
{"train_loss": 0.2680855393409729, "global_step": 50740, "epoch": 570, "lr": 9.175364800337447e-05}
{"train_loss": 0.4475943148136139, "global_step": 50741, "epoch": 570, "lr": 9.175332906473033e-05}
{"train_loss": 0.2916244566440582, "global_step": 50742, "epoch": 570, "lr": 9.1753010120473e-05}
{"train_loss": 0.20799961686134338, "global_step": 50743, "epoch": 570, "lr": 9.175269117060245e-05}
{"train_loss": 0.2204616218805313, "global_step": 50744, "epoch": 570, "lr": 9.175237221511878e-05}
{"train_loss": 0.21469098329544067, "global_step": 50745, "epoch": 570, "lr": 9.175205325402201e-05}
{"train_loss": 0.32256019115448, "global_step": 50746, "epoch": 570, "lr": 9.175173428731217e-05}
{"train_loss": 0.22711269557476044, "global_step": 50747, "epoch": 570, "lr": 9.175141531498935e-05}
{"train_loss": 0.23015744984149933, "global_step": 50748, "epoch": 570, "lr": 9.175109633705352e-05}
{"train_loss": 0.23961368203163147, "global_step": 50749, "epoch": 570, "lr": 9.175077735350479e-05}
{"train_loss": 0.1834147572517395, "global_step": 50750, "epoch": 570, "lr": 9.175045836434318e-05}
{"train_loss": 0.29571500420570374, "global_step": 50751, "epoch": 570, "lr": 9.175013936956874e-05}
{"train_loss": 0.34162449836730957, "global_step": 50752, "epoch": 570, "lr": 9.174982036918147e-05}
{"train_loss": 0.24785764515399933, "global_step": 50753, "epoch": 570, "lr": 9.174950136318146e-05}
{"train_loss": 0.23322348296642303, "global_step": 50754, "epoch": 570, "lr": 9.174918235156873e-05}
{"train_loss": 0.33806532621383667, "global_step": 50755, "epoch": 570, "lr": 9.174886333434335e-05}
{"train_loss": 0.24381649494171143, "global_step": 50756, "epoch": 570, "lr": 9.174854431150532e-05}
{"train_loss": 0.22706864774227142, "global_step": 50757, "epoch": 570, "lr": 9.174822528305473e-05}
{"train_loss": 0.2532607913017273, "global_step": 50758, "epoch": 570, "lr": 9.174790624899158e-05}
{"train_loss": 0.3255847096443176, "global_step": 50759, "epoch": 570, "lr": 9.174758720931595e-05}
{"train_loss": 0.29335200786590576, "global_step": 50760, "epoch": 570, "lr": 9.174726816402788e-05}
{"train_loss": 0.335370272397995, "global_step": 50761, "epoch": 570, "lr": 9.174694911312736e-05}
{"train_loss": 0.35601526498794556, "global_step": 50762, "epoch": 570, "lr": 9.17466300566145e-05}
{"train_loss": 0.25856393575668335, "global_step": 50763, "epoch": 570, "lr": 9.17463109944893e-05}
{"train_loss": 0.26558706164360046, "global_step": 50764, "epoch": 570, "lr": 9.174599192675183e-05}
{"train_loss": 0.3380994498729706, "global_step": 50765, "epoch": 570, "lr": 9.174567285340211e-05}
{"train_loss": 0.209679514169693, "global_step": 50766, "epoch": 570, "lr": 9.17453537744402e-05}
{"train_loss": 0.19150696694850922, "global_step": 50767, "epoch": 570, "lr": 9.174503468986613e-05}
{"train_loss": 0.3905893564224243, "global_step": 50768, "epoch": 570, "lr": 9.174471559967994e-05}
{"train_loss": 0.3663908541202545, "global_step": 50769, "epoch": 570, "lr": 9.17443965038817e-05}
{"train_loss": 0.3186490535736084, "global_step": 50770, "epoch": 570, "lr": 9.174407740247143e-05}
{"train_loss": 0.28095972537994385, "global_step": 50771, "epoch": 570, "lr": 9.174375829544917e-05}
{"train_loss": 0.2811182141304016, "global_step": 50772, "epoch": 570, "lr": 9.174343918281499e-05}
{"train_loss": 0.22658391296863556, "global_step": 50773, "epoch": 570, "lr": 9.174312006456889e-05}
{"train_loss": 0.23177975416183472, "global_step": 50774, "epoch": 570, "lr": 9.174280094071093e-05}
{"train_loss": 0.23195713758468628, "global_step": 50775, "epoch": 570, "lr": 9.174248181124119e-05}
{"train_loss": 0.24290244281291962, "global_step": 50776, "epoch": 570, "lr": 9.174216267615966e-05}
{"train_loss": 0.3727901577949524, "global_step": 50777, "epoch": 570, "lr": 9.174184353546641e-05}
{"train_loss": 0.24428129196166992, "global_step": 50778, "epoch": 570, "lr": 9.174152438916148e-05}
{"train_loss": 0.33305346965789795, "global_step": 50779, "epoch": 570, "lr": 9.174120523724491e-05}
{"train_loss": 0.270133912563324, "global_step": 50780, "epoch": 570, "lr": 9.174088607971673e-05}
{"train_loss": 0.3023870885372162, "global_step": 50781, "epoch": 570, "lr": 9.174056691657702e-05}
{"train_loss": 0.26904022693634033, "global_step": 50782, "epoch": 570, "lr": 9.17402477478258e-05}
{"train_loss": 0.21804887056350708, "global_step": 50783, "epoch": 570, "lr": 9.17399285734631e-05}
{"train_loss": 0.2976938784122467, "global_step": 50784, "epoch": 570, "lr": 9.173960939348897e-05}
{"train_loss": 0.24074456095695496, "global_step": 50785, "epoch": 570, "lr": 9.173929020790345e-05}
{"train_loss": 0.3662281930446625, "global_step": 50786, "epoch": 570, "lr": 9.173897101670662e-05}
{"train_loss": 0.3166309595108032, "global_step": 50787, "epoch": 570, "lr": 9.173865181989848e-05}
{"train_loss": 0.35776352882385254, "global_step": 50788, "epoch": 570, "lr": 9.173833261747909e-05}
{"train_loss": 0.35836949944496155, "global_step": 50789, "epoch": 570, "lr": 9.173801340944849e-05}
{"train_loss": 0.27872252464294434, "global_step": 50790, "epoch": 570, "lr": 9.173769419580671e-05}
{"train_loss": 0.19286102056503296, "global_step": 50791, "epoch": 570, "lr": 9.173737497655383e-05}
{"train_loss": 0.2632390558719635, "global_step": 50792, "epoch": 570, "lr": 9.173705575168985e-05}
{"train_loss": 0.41600698232650757, "global_step": 50793, "epoch": 570, "lr": 9.173673652121484e-05}
{"train_loss": 0.34405234456062317, "global_step": 50794, "epoch": 570, "lr": 9.173641728512883e-05}
{"train_loss": 0.20785067975521088, "global_step": 50795, "epoch": 570, "lr": 9.173609804343186e-05}
{"train_loss": 0.18475516140460968, "global_step": 50796, "epoch": 570, "lr": 9.173577879612399e-05}
{"train_loss": 0.2664363980293274, "global_step": 50797, "epoch": 570, "lr": 9.173545954320526e-05}
{"train_loss": 0.31785184144973755, "global_step": 50798, "epoch": 570, "lr": 9.17351402846757e-05}
{"train_loss": 0.1796267181634903, "global_step": 50799, "epoch": 570, "lr": 9.173482102053534e-05}
{"train_loss": 0.15148666501045227, "global_step": 50800, "epoch": 570, "lr": 9.173450175078426e-05}
{"train_loss": 0.326224684715271, "global_step": 50801, "epoch": 570, "lr": 9.17341824754225e-05}
{"train_loss": 0.2904420793056488, "global_step": 50802, "epoch": 570, "lr": 9.173386319445005e-05}
{"train_loss": 0.20689716935157776, "global_step": 50803, "epoch": 570, "lr": 9.173354390786703e-05}
{"train_loss": 0.23469455540180206, "global_step": 50804, "epoch": 570, "lr": 9.173322461567342e-05}
{"train_loss": 0.4308914542198181, "global_step": 50805, "epoch": 570, "lr": 9.173290531786929e-05}
{"train_loss": 0.20698383450508118, "global_step": 50806, "epoch": 570, "lr": 9.173258601445469e-05}
{"train_loss": 0.24986429512500763, "global_step": 50807, "epoch": 570, "lr": 9.173226670542964e-05}
{"train_loss": 0.22727379202842712, "global_step": 50808, "epoch": 570, "lr": 9.17319473907942e-05}
{"train_loss": 0.30842751264572144, "global_step": 50809, "epoch": 570, "lr": 9.17316280705484e-05}
{"train_loss": 0.23590520024299622, "global_step": 50810, "epoch": 570, "lr": 9.173130874469231e-05}
{"train_loss": 0.2449881136417389, "global_step": 50811, "epoch": 570, "lr": 9.173098941322594e-05}
{"train_loss": 0.2512570023536682, "global_step": 50812, "epoch": 570, "lr": 9.173067007614934e-05}
{"train_loss": 0.18949241936206818, "global_step": 50813, "epoch": 570, "lr": 9.17303507334626e-05}
{"train_loss": 0.23480117321014404, "global_step": 50814, "epoch": 570, "lr": 9.173003138516567e-05}
{"train_loss": 0.21198925375938416, "global_step": 50815, "epoch": 570, "lr": 9.172971203125868e-05}
{"train_loss": 0.3265208303928375, "global_step": 50816, "epoch": 570, "lr": 9.172939267174163e-05}
{"train_loss": 0.19460797309875488, "global_step": 50817, "epoch": 570, "lr": 9.172907330661456e-05}
{"train_loss": 0.2758891684620568, "global_step": 50818, "epoch": 570, "lr": 9.172875393587754e-05, "val_loss": 3.005641222000122, "train_action_mse_error": 28.560035705566406}
{"train_loss": 0.27306824922561646, "global_step": 50819, "epoch": 571, "lr": 9.17284345595306e-05}
{"train_loss": 0.2605798542499542, "global_step": 50820, "epoch": 571, "lr": 9.172811517757376e-05}
{"train_loss": 0.27612391114234924, "global_step": 50821, "epoch": 571, "lr": 9.17277957900071e-05}
{"train_loss": 0.32909131050109863, "global_step": 50822, "epoch": 571, "lr": 9.172747639683065e-05}
{"train_loss": 0.3182235062122345, "global_step": 50823, "epoch": 571, "lr": 9.172715699804445e-05}
{"train_loss": 0.23447541892528534, "global_step": 50824, "epoch": 571, "lr": 9.172683759364853e-05}
{"train_loss": 0.2477588653564453, "global_step": 50825, "epoch": 571, "lr": 9.172651818364295e-05}
{"train_loss": 0.3200990557670593, "global_step": 50826, "epoch": 571, "lr": 9.172619876802775e-05}
{"train_loss": 0.22395206987857819, "global_step": 50827, "epoch": 571, "lr": 9.172587934680297e-05}
{"train_loss": 0.26361873745918274, "global_step": 50828, "epoch": 571, "lr": 9.172555991996865e-05}
{"train_loss": 0.24818463623523712, "global_step": 50829, "epoch": 571, "lr": 9.172524048752484e-05}
{"train_loss": 0.2527742087841034, "global_step": 50830, "epoch": 571, "lr": 9.172492104947157e-05}
{"train_loss": 0.2927604019641876, "global_step": 50831, "epoch": 571, "lr": 9.172460160580892e-05}
{"train_loss": 0.29868999123573303, "global_step": 50832, "epoch": 571, "lr": 9.172428215653688e-05}
{"train_loss": 0.3209310472011566, "global_step": 50833, "epoch": 571, "lr": 9.172396270165552e-05}
{"train_loss": 0.22034037113189697, "global_step": 50834, "epoch": 571, "lr": 9.17236432411649e-05}
{"train_loss": 0.2537596821784973, "global_step": 50835, "epoch": 571, "lr": 9.172332377506503e-05}
{"train_loss": 0.32166871428489685, "global_step": 50836, "epoch": 571, "lr": 9.172300430335597e-05}
{"train_loss": 0.27896013855934143, "global_step": 50837, "epoch": 571, "lr": 9.172268482603777e-05}
{"train_loss": 0.23766663670539856, "global_step": 50838, "epoch": 571, "lr": 9.172236534311045e-05}
{"train_loss": 0.26627710461616516, "global_step": 50839, "epoch": 571, "lr": 9.172204585457409e-05}
{"train_loss": 0.3032929301261902, "global_step": 50840, "epoch": 571, "lr": 9.172172636042868e-05}
{"train_loss": 0.32510706782341003, "global_step": 50841, "epoch": 571, "lr": 9.172140686067431e-05}
{"train_loss": 0.2504020929336548, "global_step": 50842, "epoch": 571, "lr": 9.1721087355311e-05}
{"train_loss": 0.3141172230243683, "global_step": 50843, "epoch": 571, "lr": 9.172076784433878e-05}
{"train_loss": 0.2550644278526306, "global_step": 50844, "epoch": 571, "lr": 9.172044832775774e-05}
{"train_loss": 0.18778499960899353, "global_step": 50845, "epoch": 571, "lr": 9.172012880556788e-05}
{"train_loss": 0.3136047422885895, "global_step": 50846, "epoch": 571, "lr": 9.171980927776926e-05}
{"train_loss": 0.2387678325176239, "global_step": 50847, "epoch": 571, "lr": 9.171948974436193e-05}
{"train_loss": 0.37472015619277954, "global_step": 50848, "epoch": 571, "lr": 9.171917020534592e-05}
{"train_loss": 0.4022064805030823, "global_step": 50849, "epoch": 571, "lr": 9.171885066072127e-05}
{"train_loss": 0.4463414251804352, "global_step": 50850, "epoch": 571, "lr": 9.171853111048803e-05}
{"train_loss": 0.26252520084381104, "global_step": 50851, "epoch": 571, "lr": 9.171821155464624e-05}
{"train_loss": 0.32519766688346863, "global_step": 50852, "epoch": 571, "lr": 9.171789199319594e-05}
{"train_loss": 0.2761443853378296, "global_step": 50853, "epoch": 571, "lr": 9.171757242613719e-05}
{"train_loss": 0.36897796392440796, "global_step": 50854, "epoch": 571, "lr": 9.171725285347002e-05}
{"train_loss": 0.21181878447532654, "global_step": 50855, "epoch": 571, "lr": 9.171693327519446e-05}
{"train_loss": 0.3636733293533325, "global_step": 50856, "epoch": 571, "lr": 9.171661369131057e-05}
{"train_loss": 0.2501319646835327, "global_step": 50857, "epoch": 571, "lr": 9.171629410181841e-05}
{"train_loss": 0.27406346797943115, "global_step": 50858, "epoch": 571, "lr": 9.171597450671798e-05}
{"train_loss": 0.2935667932033539, "global_step": 50859, "epoch": 571, "lr": 9.171565490600935e-05}
{"train_loss": 0.3469265103340149, "global_step": 50860, "epoch": 571, "lr": 9.171533529969256e-05}
{"train_loss": 0.4206847846508026, "global_step": 50861, "epoch": 571, "lr": 9.171501568776765e-05}
{"train_loss": 0.2714051306247711, "global_step": 50862, "epoch": 571, "lr": 9.171469607023468e-05}
{"train_loss": 0.30066797137260437, "global_step": 50863, "epoch": 571, "lr": 9.171437644709367e-05}
{"train_loss": 0.37560558319091797, "global_step": 50864, "epoch": 571, "lr": 9.171405681834466e-05}
{"train_loss": 0.2966720759868622, "global_step": 50865, "epoch": 571, "lr": 9.171373718398772e-05}
{"train_loss": 0.2842361032962799, "global_step": 50866, "epoch": 571, "lr": 9.171341754402286e-05}
{"train_loss": 0.271369069814682, "global_step": 50867, "epoch": 571, "lr": 9.171309789845015e-05}
{"train_loss": 0.28554633259773254, "global_step": 50868, "epoch": 571, "lr": 9.171277824726961e-05}
{"train_loss": 0.2638711929321289, "global_step": 50869, "epoch": 571, "lr": 9.171245859048131e-05}
{"train_loss": 0.2945825457572937, "global_step": 50870, "epoch": 571, "lr": 9.171213892808527e-05}
{"train_loss": 0.31691014766693115, "global_step": 50871, "epoch": 571, "lr": 9.171181926008154e-05}
{"train_loss": 0.25996536016464233, "global_step": 50872, "epoch": 571, "lr": 9.171149958647018e-05}
{"train_loss": 0.3175334334373474, "global_step": 50873, "epoch": 571, "lr": 9.17111799072512e-05}
{"train_loss": 0.3756745755672455, "global_step": 50874, "epoch": 571, "lr": 9.171086022242467e-05}
{"train_loss": 0.33126702904701233, "global_step": 50875, "epoch": 571, "lr": 9.171054053199061e-05}
{"train_loss": 0.26766517758369446, "global_step": 50876, "epoch": 571, "lr": 9.171022083594908e-05}
{"train_loss": 0.27116450667381287, "global_step": 50877, "epoch": 571, "lr": 9.170990113430014e-05}
{"train_loss": 0.3354869484901428, "global_step": 50878, "epoch": 571, "lr": 9.170958142704378e-05}
{"train_loss": 0.3029799461364746, "global_step": 50879, "epoch": 571, "lr": 9.17092617141801e-05}
{"train_loss": 0.31714439392089844, "global_step": 50880, "epoch": 571, "lr": 9.170894199570911e-05}
{"train_loss": 0.2989250123500824, "global_step": 50881, "epoch": 571, "lr": 9.170862227163087e-05}
{"train_loss": 0.33289775252342224, "global_step": 50882, "epoch": 571, "lr": 9.170830254194539e-05}
{"train_loss": 0.2359708845615387, "global_step": 50883, "epoch": 571, "lr": 9.170798280665276e-05}
{"train_loss": 0.2780374586582184, "global_step": 50884, "epoch": 571, "lr": 9.170766306575297e-05}
{"train_loss": 0.22243018448352814, "global_step": 50885, "epoch": 571, "lr": 9.170734331924613e-05}
{"train_loss": 0.42684441804885864, "global_step": 50886, "epoch": 571, "lr": 9.170702356713224e-05}
{"train_loss": 0.3191643953323364, "global_step": 50887, "epoch": 571, "lr": 9.170670380941134e-05}
{"train_loss": 0.20792384445667267, "global_step": 50888, "epoch": 571, "lr": 9.170638404608347e-05}
{"train_loss": 0.31436383724212646, "global_step": 50889, "epoch": 571, "lr": 9.17060642771487e-05}
{"train_loss": 0.28889578580856323, "global_step": 50890, "epoch": 571, "lr": 9.170574450260706e-05}
{"train_loss": 0.356679767370224, "global_step": 50891, "epoch": 571, "lr": 9.170542472245858e-05}
{"train_loss": 0.3102880120277405, "global_step": 50892, "epoch": 571, "lr": 9.170510493670332e-05}
{"train_loss": 0.30541402101516724, "global_step": 50893, "epoch": 571, "lr": 9.170478514534132e-05}
{"train_loss": 0.3037723898887634, "global_step": 50894, "epoch": 571, "lr": 9.170446534837262e-05}
{"train_loss": 0.3178437054157257, "global_step": 50895, "epoch": 571, "lr": 9.170414554579724e-05}
{"train_loss": 0.2803443372249603, "global_step": 50896, "epoch": 571, "lr": 9.170382573761529e-05}
{"train_loss": 0.25897303223609924, "global_step": 50897, "epoch": 571, "lr": 9.170350592382673e-05}
{"train_loss": 0.3837090730667114, "global_step": 50898, "epoch": 571, "lr": 9.170318610443166e-05}
{"train_loss": 0.3195136487483978, "global_step": 50899, "epoch": 571, "lr": 9.17028662794301e-05}
{"train_loss": 0.23290498554706573, "global_step": 50900, "epoch": 571, "lr": 9.17025464488221e-05}
{"train_loss": 0.30390864610671997, "global_step": 50901, "epoch": 571, "lr": 9.17022266126077e-05}
{"train_loss": 0.31334829330444336, "global_step": 50902, "epoch": 571, "lr": 9.170190677078695e-05}
{"train_loss": 0.26320353150367737, "global_step": 50903, "epoch": 571, "lr": 9.170158692335988e-05}
{"train_loss": 0.36038100719451904, "global_step": 50904, "epoch": 571, "lr": 9.170126707032653e-05}
{"train_loss": 0.21437205374240875, "global_step": 50905, "epoch": 571, "lr": 9.170094721168697e-05}
{"train_loss": 0.24424085021018982, "global_step": 50906, "epoch": 571, "lr": 9.170062734744123e-05}
{"train_loss": 0.29504122643658287, "global_step": 50907, "epoch": 571, "lr": 9.170030747758934e-05, "val_loss": 2.9082696437835693}
{"train_loss": 0.332968145608902, "global_step": 50908, "epoch": 572, "lr": 9.169998760213135e-05}
{"train_loss": 0.26583531498908997, "global_step": 50909, "epoch": 572, "lr": 9.16996677210673e-05}
{"train_loss": 0.19717296957969666, "global_step": 50910, "epoch": 572, "lr": 9.169934783439725e-05}
{"train_loss": 0.2633664608001709, "global_step": 50911, "epoch": 572, "lr": 9.169902794212122e-05}
{"train_loss": 0.3076365292072296, "global_step": 50912, "epoch": 572, "lr": 9.169870804423926e-05}
{"train_loss": 0.3865528702735901, "global_step": 50913, "epoch": 572, "lr": 9.169838814075142e-05}
{"train_loss": 0.3021584749221802, "global_step": 50914, "epoch": 572, "lr": 9.169806823165775e-05}
{"train_loss": 0.24415373802185059, "global_step": 50915, "epoch": 572, "lr": 9.169774831695827e-05}
{"train_loss": 0.21841606497764587, "global_step": 50916, "epoch": 572, "lr": 9.169742839665304e-05}
{"train_loss": 0.22758398950099945, "global_step": 50917, "epoch": 572, "lr": 9.169710847074211e-05}
{"train_loss": 0.26173335313796997, "global_step": 50918, "epoch": 572, "lr": 9.16967885392255e-05}
{"train_loss": 0.35312607884407043, "global_step": 50919, "epoch": 572, "lr": 9.169646860210329e-05}
{"train_loss": 0.22601957619190216, "global_step": 50920, "epoch": 572, "lr": 9.169614865937547e-05}
{"train_loss": 0.2728038430213928, "global_step": 50921, "epoch": 572, "lr": 9.169582871104213e-05}
{"train_loss": 0.3109011948108673, "global_step": 50922, "epoch": 572, "lr": 9.169550875710328e-05}
{"train_loss": 0.2033221274614334, "global_step": 50923, "epoch": 572, "lr": 9.169518879755898e-05}
{"train_loss": 0.23690232634544373, "global_step": 50924, "epoch": 572, "lr": 9.169486883240926e-05}
{"train_loss": 0.24326463043689728, "global_step": 50925, "epoch": 572, "lr": 9.169454886165419e-05}
{"train_loss": 0.212909534573555, "global_step": 50926, "epoch": 572, "lr": 9.169422888529378e-05}
{"train_loss": 0.2618342638015747, "global_step": 50927, "epoch": 572, "lr": 9.16939089033281e-05}
{"train_loss": 0.27157124876976013, "global_step": 50928, "epoch": 572, "lr": 9.169358891575719e-05}
{"train_loss": 0.27615565061569214, "global_step": 50929, "epoch": 572, "lr": 9.169326892258108e-05}
{"train_loss": 0.22113165259361267, "global_step": 50930, "epoch": 572, "lr": 9.169294892379981e-05}
{"train_loss": 0.2628321051597595, "global_step": 50931, "epoch": 572, "lr": 9.169262891941343e-05}
{"train_loss": 0.2986977696418762, "global_step": 50932, "epoch": 572, "lr": 9.1692308909422e-05}
{"train_loss": 0.395774781703949, "global_step": 50933, "epoch": 572, "lr": 9.169198889382554e-05}
{"train_loss": 0.30094772577285767, "global_step": 50934, "epoch": 572, "lr": 9.169166887262408e-05}
{"train_loss": 0.3130878210067749, "global_step": 50935, "epoch": 572, "lr": 9.16913488458177e-05}
{"train_loss": 0.3007103204727173, "global_step": 50936, "epoch": 572, "lr": 9.169102881340643e-05}
{"train_loss": 0.22555455565452576, "global_step": 50937, "epoch": 572, "lr": 9.16907087753903e-05}
{"train_loss": 0.31460750102996826, "global_step": 50938, "epoch": 572, "lr": 9.169038873176937e-05}
{"train_loss": 0.3067708909511566, "global_step": 50939, "epoch": 572, "lr": 9.169006868254367e-05}
{"train_loss": 0.28167724609375, "global_step": 50940, "epoch": 572, "lr": 9.168974862771326e-05}
{"train_loss": 0.22319835424423218, "global_step": 50941, "epoch": 572, "lr": 9.168942856727815e-05}
{"train_loss": 0.3704563081264496, "global_step": 50942, "epoch": 572, "lr": 9.168910850123841e-05}
{"train_loss": 0.27216458320617676, "global_step": 50943, "epoch": 572, "lr": 9.16887884295941e-05}
{"train_loss": 0.2585967779159546, "global_step": 50944, "epoch": 572, "lr": 9.168846835234524e-05}
{"train_loss": 0.3311128616333008, "global_step": 50945, "epoch": 572, "lr": 9.168814826949185e-05}
{"train_loss": 0.33111220598220825, "global_step": 50946, "epoch": 572, "lr": 9.168782818103401e-05}
{"train_loss": 0.25180837512016296, "global_step": 50947, "epoch": 572, "lr": 9.168750808697174e-05}
{"train_loss": 0.2527945041656494, "global_step": 50948, "epoch": 572, "lr": 9.168718798730509e-05}
{"train_loss": 0.2091473788022995, "global_step": 50949, "epoch": 572, "lr": 9.168686788203412e-05}
{"train_loss": 0.390106737613678, "global_step": 50950, "epoch": 572, "lr": 9.168654777115887e-05}
{"train_loss": 0.3755333125591278, "global_step": 50951, "epoch": 572, "lr": 9.168622765467933e-05}
{"train_loss": 0.28424176573753357, "global_step": 50952, "epoch": 572, "lr": 9.168590753259561e-05}
{"train_loss": 0.205415278673172, "global_step": 50953, "epoch": 572, "lr": 9.168558740490774e-05}
{"train_loss": 0.25240907073020935, "global_step": 50954, "epoch": 572, "lr": 9.168526727161573e-05}
{"train_loss": 0.2617504894733429, "global_step": 50955, "epoch": 572, "lr": 9.168494713271965e-05}
{"train_loss": 0.28808364272117615, "global_step": 50956, "epoch": 572, "lr": 9.168462698821953e-05}
{"train_loss": 0.42968612909317017, "global_step": 50957, "epoch": 572, "lr": 9.168430683811545e-05}
{"train_loss": 0.5222525596618652, "global_step": 50958, "epoch": 572, "lr": 9.168398668240739e-05}
{"train_loss": 0.3835301697254181, "global_step": 50959, "epoch": 572, "lr": 9.168366652109543e-05}
{"train_loss": 0.24612711369991302, "global_step": 50960, "epoch": 572, "lr": 9.168334635417963e-05}
{"train_loss": 0.22975139319896698, "global_step": 50961, "epoch": 572, "lr": 9.168302618165999e-05}
{"train_loss": 0.2656322121620178, "global_step": 50962, "epoch": 572, "lr": 9.168270600353659e-05}
{"train_loss": 0.32886171340942383, "global_step": 50963, "epoch": 572, "lr": 9.168238581980944e-05}
{"train_loss": 0.3704690933227539, "global_step": 50964, "epoch": 572, "lr": 9.168206563047863e-05}
{"train_loss": 0.3757232427597046, "global_step": 50965, "epoch": 572, "lr": 9.168174543554415e-05}
{"train_loss": 0.33975210785865784, "global_step": 50966, "epoch": 572, "lr": 9.168142523500607e-05}
{"train_loss": 0.20723126828670502, "global_step": 50967, "epoch": 572, "lr": 9.168110502886445e-05}
{"train_loss": 0.2265802025794983, "global_step": 50968, "epoch": 572, "lr": 9.168078481711928e-05}
{"train_loss": 0.25267350673675537, "global_step": 50969, "epoch": 572, "lr": 9.168046459977066e-05}
{"train_loss": 0.30010056495666504, "global_step": 50970, "epoch": 572, "lr": 9.16801443768186e-05}
{"train_loss": 0.21695081889629364, "global_step": 50971, "epoch": 572, "lr": 9.167982414826317e-05}
{"train_loss": 0.2683942914009094, "global_step": 50972, "epoch": 572, "lr": 9.167950391410439e-05}
{"train_loss": 0.24935516715049744, "global_step": 50973, "epoch": 572, "lr": 9.16791836743423e-05}
{"train_loss": 0.42393434047698975, "global_step": 50974, "epoch": 572, "lr": 9.167886342897694e-05}
{"train_loss": 0.35702377557754517, "global_step": 50975, "epoch": 572, "lr": 9.167854317800838e-05}
{"train_loss": 0.24219642579555511, "global_step": 50976, "epoch": 572, "lr": 9.167822292143664e-05}
{"train_loss": 0.2727441191673279, "global_step": 50977, "epoch": 572, "lr": 9.167790265926179e-05}
{"train_loss": 0.2899155914783478, "global_step": 50978, "epoch": 572, "lr": 9.167758239148383e-05}
{"train_loss": 0.2671196460723877, "global_step": 50979, "epoch": 572, "lr": 9.167726211810284e-05}
{"train_loss": 0.28901028633117676, "global_step": 50980, "epoch": 572, "lr": 9.167694183911886e-05}
{"train_loss": 0.2902352809906006, "global_step": 50981, "epoch": 572, "lr": 9.167662155453189e-05}
{"train_loss": 0.3284143805503845, "global_step": 50982, "epoch": 572, "lr": 9.167630126434204e-05}
{"train_loss": 0.3261313736438751, "global_step": 50983, "epoch": 572, "lr": 9.16759809685493e-05}
{"train_loss": 0.3008006811141968, "global_step": 50984, "epoch": 572, "lr": 9.167566066715373e-05}
{"train_loss": 0.2664644718170166, "global_step": 50985, "epoch": 572, "lr": 9.167534036015539e-05}
{"train_loss": 0.28504976630210876, "global_step": 50986, "epoch": 572, "lr": 9.167502004755431e-05}
{"train_loss": 0.25484567880630493, "global_step": 50987, "epoch": 572, "lr": 9.167469972935052e-05}
{"train_loss": 0.16875037550926208, "global_step": 50988, "epoch": 572, "lr": 9.167437940554408e-05}
{"train_loss": 0.26424017548561096, "global_step": 50989, "epoch": 572, "lr": 9.167405907613503e-05}
{"train_loss": 0.3031140863895416, "global_step": 50990, "epoch": 572, "lr": 9.167373874112341e-05}
{"train_loss": 0.17768199741840363, "global_step": 50991, "epoch": 572, "lr": 9.167341840050925e-05}
{"train_loss": 0.251970499753952, "global_step": 50992, "epoch": 572, "lr": 9.167309805429263e-05}
{"train_loss": 0.3022872507572174, "global_step": 50993, "epoch": 572, "lr": 9.167277770247356e-05}
{"train_loss": 0.26795467734336853, "global_step": 50994, "epoch": 572, "lr": 9.167245734505209e-05}
{"train_loss": 0.33028313517570496, "global_step": 50995, "epoch": 572, "lr": 9.167213698202826e-05}
{"train_loss": 0.2852658870514859, "global_step": 50996, "epoch": 572, "lr": 9.167181661340213e-05, "val_loss": 3.068530559539795}
{"train_loss": 0.31642696261405945, "global_step": 50997, "epoch": 573, "lr": 9.167149623917373e-05}
{"train_loss": 0.26215416193008423, "global_step": 50998, "epoch": 573, "lr": 9.16711758593431e-05}
{"train_loss": 0.37424710392951965, "global_step": 50999, "epoch": 573, "lr": 9.16708554739103e-05}
{"train_loss": 0.2573084235191345, "global_step": 51000, "epoch": 573, "lr": 9.167053508287536e-05}
{"train_loss": 0.289425253868103, "global_step": 51001, "epoch": 573, "lr": 9.16702146862383e-05}
{"train_loss": 0.2064652293920517, "global_step": 51002, "epoch": 573, "lr": 9.166989428399921e-05}
{"train_loss": 0.2927379012107849, "global_step": 51003, "epoch": 573, "lr": 9.166957387615811e-05}
{"train_loss": 0.292355477809906, "global_step": 51004, "epoch": 573, "lr": 9.166925346271503e-05}
{"train_loss": 0.22054652869701385, "global_step": 51005, "epoch": 573, "lr": 9.166893304367004e-05}
{"train_loss": 0.27215811610221863, "global_step": 51006, "epoch": 573, "lr": 9.166861261902317e-05}
{"train_loss": 0.31474027037620544, "global_step": 51007, "epoch": 573, "lr": 9.166829218877445e-05}
{"train_loss": 0.2529810070991516, "global_step": 51008, "epoch": 573, "lr": 9.166797175292394e-05}
{"train_loss": 0.21327336132526398, "global_step": 51009, "epoch": 573, "lr": 9.166765131147169e-05}
{"train_loss": 0.26048168540000916, "global_step": 51010, "epoch": 573, "lr": 9.166733086441772e-05}
{"train_loss": 0.19321687519550323, "global_step": 51011, "epoch": 573, "lr": 9.16670104117621e-05}
{"train_loss": 0.1844101846218109, "global_step": 51012, "epoch": 573, "lr": 9.166668995350485e-05}
{"train_loss": 0.3189133107662201, "global_step": 51013, "epoch": 573, "lr": 9.166636948964602e-05}
{"train_loss": 0.22646377980709076, "global_step": 51014, "epoch": 573, "lr": 9.166604902018566e-05}
{"train_loss": 0.26367276906967163, "global_step": 51015, "epoch": 573, "lr": 9.166572854512379e-05}
{"train_loss": 0.268837571144104, "global_step": 51016, "epoch": 573, "lr": 9.166540806446049e-05}
{"train_loss": 0.28553104400634766, "global_step": 51017, "epoch": 573, "lr": 9.166508757819578e-05}
{"train_loss": 0.23340797424316406, "global_step": 51018, "epoch": 573, "lr": 9.16647670863297e-05}
{"train_loss": 0.30811652541160583, "global_step": 51019, "epoch": 573, "lr": 9.166444658886231e-05}
{"train_loss": 0.2702815532684326, "global_step": 51020, "epoch": 573, "lr": 9.166412608579363e-05}
{"train_loss": 0.22244791686534882, "global_step": 51021, "epoch": 573, "lr": 9.166380557712373e-05}
{"train_loss": 0.279976487159729, "global_step": 51022, "epoch": 573, "lr": 9.166348506285263e-05}
{"train_loss": 0.32567816972732544, "global_step": 51023, "epoch": 573, "lr": 9.166316454298039e-05}
{"train_loss": 0.261759877204895, "global_step": 51024, "epoch": 573, "lr": 9.166284401750703e-05}
{"train_loss": 0.19380201399326324, "global_step": 51025, "epoch": 573, "lr": 9.166252348643262e-05}
{"train_loss": 0.2837468683719635, "global_step": 51026, "epoch": 573, "lr": 9.166220294975718e-05}
{"train_loss": 0.22344890236854553, "global_step": 51027, "epoch": 573, "lr": 9.166188240748079e-05}
{"train_loss": 0.278865784406662, "global_step": 51028, "epoch": 573, "lr": 9.166156185960344e-05}
{"train_loss": 0.22236894071102142, "global_step": 51029, "epoch": 573, "lr": 9.166124130612522e-05}
{"train_loss": 0.31563979387283325, "global_step": 51030, "epoch": 573, "lr": 9.166092074704614e-05}
{"train_loss": 0.3181472718715668, "global_step": 51031, "epoch": 573, "lr": 9.166060018236626e-05}
{"train_loss": 0.3351077735424042, "global_step": 51032, "epoch": 573, "lr": 9.166027961208563e-05}
{"train_loss": 0.2554934620857239, "global_step": 51033, "epoch": 573, "lr": 9.165995903620428e-05}
{"train_loss": 0.4270150661468506, "global_step": 51034, "epoch": 573, "lr": 9.165963845472225e-05}
{"train_loss": 0.2976382076740265, "global_step": 51035, "epoch": 573, "lr": 9.16593178676396e-05}
{"train_loss": 0.24327966570854187, "global_step": 51036, "epoch": 573, "lr": 9.165899727495635e-05}
{"train_loss": 0.25508275628089905, "global_step": 51037, "epoch": 573, "lr": 9.165867667667256e-05}
{"train_loss": 0.3030659854412079, "global_step": 51038, "epoch": 573, "lr": 9.165835607278826e-05}
{"train_loss": 0.2308799773454666, "global_step": 51039, "epoch": 573, "lr": 9.165803546330351e-05}
{"train_loss": 0.30830833315849304, "global_step": 51040, "epoch": 573, "lr": 9.165771484821835e-05}
{"train_loss": 0.24412846565246582, "global_step": 51041, "epoch": 573, "lr": 9.165739422753281e-05}
{"train_loss": 0.21103134751319885, "global_step": 51042, "epoch": 573, "lr": 9.165707360124694e-05}
{"train_loss": 0.3601316511631012, "global_step": 51043, "epoch": 573, "lr": 9.165675296936079e-05}
{"train_loss": 0.2953939735889435, "global_step": 51044, "epoch": 573, "lr": 9.165643233187439e-05}
{"train_loss": 0.25985288619995117, "global_step": 51045, "epoch": 573, "lr": 9.165611168878781e-05}
{"train_loss": 0.21818239986896515, "global_step": 51046, "epoch": 573, "lr": 9.165579104010107e-05}
{"train_loss": 0.4406225085258484, "global_step": 51047, "epoch": 573, "lr": 9.165547038581419e-05}
{"train_loss": 0.2209712564945221, "global_step": 51048, "epoch": 573, "lr": 9.165514972592726e-05}
{"train_loss": 0.2781703472137451, "global_step": 51049, "epoch": 573, "lr": 9.165482906044031e-05}
{"train_loss": 0.32330647110939026, "global_step": 51050, "epoch": 573, "lr": 9.165450838935336e-05}
{"train_loss": 0.28574061393737793, "global_step": 51051, "epoch": 573, "lr": 9.165418771266647e-05}
{"train_loss": 0.2659727931022644, "global_step": 51052, "epoch": 573, "lr": 9.165386703037969e-05}
{"train_loss": 0.2947176992893219, "global_step": 51053, "epoch": 573, "lr": 9.165354634249304e-05}
{"train_loss": 0.3139195442199707, "global_step": 51054, "epoch": 573, "lr": 9.16532256490066e-05}
{"train_loss": 0.2631889283657074, "global_step": 51055, "epoch": 573, "lr": 9.165290494992039e-05}
{"train_loss": 0.26550740003585815, "global_step": 51056, "epoch": 573, "lr": 9.165258424523444e-05}
{"train_loss": 0.3356277346611023, "global_step": 51057, "epoch": 573, "lr": 9.165226353494881e-05}
{"train_loss": 0.2300221174955368, "global_step": 51058, "epoch": 573, "lr": 9.165194281906355e-05}
{"train_loss": 0.1630096584558487, "global_step": 51059, "epoch": 573, "lr": 9.16516220975787e-05}
{"train_loss": 0.22325900197029114, "global_step": 51060, "epoch": 573, "lr": 9.165130137049427e-05}
{"train_loss": 0.32208052277565, "global_step": 51061, "epoch": 573, "lr": 9.165098063781034e-05}
{"train_loss": 0.2793121039867401, "global_step": 51062, "epoch": 573, "lr": 9.165065989952694e-05}
{"train_loss": 0.2233509123325348, "global_step": 51063, "epoch": 573, "lr": 9.165033915564414e-05}
{"train_loss": 0.24644960463047028, "global_step": 51064, "epoch": 573, "lr": 9.165001840616194e-05}
{"train_loss": 0.22367653250694275, "global_step": 51065, "epoch": 573, "lr": 9.164969765108041e-05}
{"train_loss": 0.2955343425273895, "global_step": 51066, "epoch": 573, "lr": 9.164937689039959e-05}
{"train_loss": 0.2462933510541916, "global_step": 51067, "epoch": 573, "lr": 9.164905612411951e-05}
{"train_loss": 0.36180171370506287, "global_step": 51068, "epoch": 573, "lr": 9.164873535224023e-05}
{"train_loss": 0.2868160903453827, "global_step": 51069, "epoch": 573, "lr": 9.164841457476177e-05}
{"train_loss": 0.23111805319786072, "global_step": 51070, "epoch": 573, "lr": 9.164809379168419e-05}
{"train_loss": 0.20410951972007751, "global_step": 51071, "epoch": 573, "lr": 9.164777300300755e-05}
{"train_loss": 0.21281155943870544, "global_step": 51072, "epoch": 573, "lr": 9.164745220873187e-05}
{"train_loss": 0.19569210708141327, "global_step": 51073, "epoch": 573, "lr": 9.164713140885718e-05}
{"train_loss": 0.2810610234737396, "global_step": 51074, "epoch": 573, "lr": 9.164681060338355e-05}
{"train_loss": 0.2805648148059845, "global_step": 51075, "epoch": 573, "lr": 9.164648979231103e-05}
{"train_loss": 0.26989391446113586, "global_step": 51076, "epoch": 573, "lr": 9.164616897563963e-05}
{"train_loss": 0.2864190936088562, "global_step": 51077, "epoch": 573, "lr": 9.16458481533694e-05}
{"train_loss": 0.34432846307754517, "global_step": 51078, "epoch": 573, "lr": 9.164552732550043e-05}
{"train_loss": 0.31018248200416565, "global_step": 51079, "epoch": 573, "lr": 9.16452064920327e-05}
{"train_loss": 0.2694104313850403, "global_step": 51080, "epoch": 573, "lr": 9.164488565296627e-05}
{"train_loss": 0.2639697790145874, "global_step": 51081, "epoch": 573, "lr": 9.164456480830121e-05}
{"train_loss": 0.27276411652565, "global_step": 51082, "epoch": 573, "lr": 9.164424395803755e-05}
{"train_loss": 0.24671271443367004, "global_step": 51083, "epoch": 573, "lr": 9.164392310217531e-05}
{"train_loss": 0.279806911945343, "global_step": 51084, "epoch": 573, "lr": 9.164360224071456e-05}
{"train_loss": 0.27105252310801087, "global_step": 51085, "epoch": 573, "lr": 9.164328137365535e-05, "val_loss": 3.1141586303710938}
{"train_loss": 0.2785199284553528, "global_step": 51086, "epoch": 574, "lr": 9.16429605009977e-05}
{"train_loss": 0.24220417439937592, "global_step": 51087, "epoch": 574, "lr": 9.164263962274164e-05}
{"train_loss": 0.30729565024375916, "global_step": 51088, "epoch": 574, "lr": 9.164231873888725e-05}
{"train_loss": 0.2591325640678406, "global_step": 51089, "epoch": 574, "lr": 9.164199784943456e-05}
{"train_loss": 0.34673991799354553, "global_step": 51090, "epoch": 574, "lr": 9.164167695438362e-05}
{"train_loss": 0.21187223494052887, "global_step": 51091, "epoch": 574, "lr": 9.164135605373446e-05}
{"train_loss": 0.29773998260498047, "global_step": 51092, "epoch": 574, "lr": 9.164103514748712e-05}
{"train_loss": 0.2315811663866043, "global_step": 51093, "epoch": 574, "lr": 9.164071423564164e-05}
{"train_loss": 0.2645197808742523, "global_step": 51094, "epoch": 574, "lr": 9.164039331819809e-05}
{"train_loss": 0.19883671402931213, "global_step": 51095, "epoch": 574, "lr": 9.16400723951565e-05}
{"train_loss": 0.27610617876052856, "global_step": 51096, "epoch": 574, "lr": 9.16397514665169e-05}
{"train_loss": 0.2849850654602051, "global_step": 51097, "epoch": 574, "lr": 9.163943053227935e-05}
{"train_loss": 0.20934779942035675, "global_step": 51098, "epoch": 574, "lr": 9.163910959244388e-05}
{"train_loss": 0.3098980486392975, "global_step": 51099, "epoch": 574, "lr": 9.163878864701054e-05}
{"train_loss": 0.35090792179107666, "global_step": 51100, "epoch": 574, "lr": 9.163846769597936e-05}
{"train_loss": 0.15006870031356812, "global_step": 51101, "epoch": 574, "lr": 9.163814673935043e-05}
{"train_loss": 0.21621887385845184, "global_step": 51102, "epoch": 574, "lr": 9.163782577712371e-05}
{"train_loss": 0.2613857686519623, "global_step": 51103, "epoch": 574, "lr": 9.163750480929933e-05}
{"train_loss": 0.3543042838573456, "global_step": 51104, "epoch": 574, "lr": 9.16371838358773e-05}
{"train_loss": 0.27121347188949585, "global_step": 51105, "epoch": 574, "lr": 9.163686285685765e-05}
{"train_loss": 0.19878318905830383, "global_step": 51106, "epoch": 574, "lr": 9.16365418722404e-05}
{"train_loss": 0.32407039403915405, "global_step": 51107, "epoch": 574, "lr": 9.163622088202566e-05}
{"train_loss": 0.21864427626132965, "global_step": 51108, "epoch": 574, "lr": 9.163589988621341e-05}
{"train_loss": 0.2732287645339966, "global_step": 51109, "epoch": 574, "lr": 9.163557888480374e-05}
{"train_loss": 0.20527493953704834, "global_step": 51110, "epoch": 574, "lr": 9.163525787779667e-05}
{"train_loss": 0.30108916759490967, "global_step": 51111, "epoch": 574, "lr": 9.163493686519226e-05}
{"train_loss": 0.23171880841255188, "global_step": 51112, "epoch": 574, "lr": 9.163461584699052e-05}
{"train_loss": 0.23124094307422638, "global_step": 51113, "epoch": 574, "lr": 9.163429482319152e-05}
{"train_loss": 0.252778559923172, "global_step": 51114, "epoch": 574, "lr": 9.16339737937953e-05}
{"train_loss": 0.22207507491111755, "global_step": 51115, "epoch": 574, "lr": 9.16336527588019e-05}
{"train_loss": 0.28341972827911377, "global_step": 51116, "epoch": 574, "lr": 9.163333171821135e-05}
{"train_loss": 0.1916564255952835, "global_step": 51117, "epoch": 574, "lr": 9.163301067202373e-05}
{"train_loss": 0.16922663152217865, "global_step": 51118, "epoch": 574, "lr": 9.163268962023903e-05}
{"train_loss": 0.2988833785057068, "global_step": 51119, "epoch": 574, "lr": 9.163236856285735e-05}
{"train_loss": 0.22485920786857605, "global_step": 51120, "epoch": 574, "lr": 9.163204749987868e-05}
{"train_loss": 0.2583465874195099, "global_step": 51121, "epoch": 574, "lr": 9.163172643130311e-05}
{"train_loss": 0.25898635387420654, "global_step": 51122, "epoch": 574, "lr": 9.163140535713066e-05}
{"train_loss": 0.2383924424648285, "global_step": 51123, "epoch": 574, "lr": 9.163108427736135e-05}
{"train_loss": 0.17013143002986908, "global_step": 51124, "epoch": 574, "lr": 9.163076319199528e-05}
{"train_loss": 0.21820829808712006, "global_step": 51125, "epoch": 574, "lr": 9.163044210103245e-05}
{"train_loss": 0.2995455265045166, "global_step": 51126, "epoch": 574, "lr": 9.16301210044729e-05}
{"train_loss": 0.18222910165786743, "global_step": 51127, "epoch": 574, "lr": 9.16297999023167e-05}
{"train_loss": 0.24679355323314667, "global_step": 51128, "epoch": 574, "lr": 9.162947879456388e-05}
{"train_loss": 0.30509111285209656, "global_step": 51129, "epoch": 574, "lr": 9.16291576812145e-05}
{"train_loss": 0.22480908036231995, "global_step": 51130, "epoch": 574, "lr": 9.162883656226857e-05}
{"train_loss": 0.17091363668441772, "global_step": 51131, "epoch": 574, "lr": 9.162851543772615e-05}
{"train_loss": 0.2871350347995758, "global_step": 51132, "epoch": 574, "lr": 9.162819430758728e-05}
{"train_loss": 0.2094845473766327, "global_step": 51133, "epoch": 574, "lr": 9.162787317185202e-05}
{"train_loss": 0.3108057379722595, "global_step": 51134, "epoch": 574, "lr": 9.162755203052038e-05}
{"train_loss": 0.2600332200527191, "global_step": 51135, "epoch": 574, "lr": 9.162723088359245e-05}
{"train_loss": 0.2744731307029724, "global_step": 51136, "epoch": 574, "lr": 9.162690973106822e-05}
{"train_loss": 0.29272645711898804, "global_step": 51137, "epoch": 574, "lr": 9.162658857294776e-05}
{"train_loss": 0.3459762632846832, "global_step": 51138, "epoch": 574, "lr": 9.162626740923113e-05}
{"train_loss": 0.3242751359939575, "global_step": 51139, "epoch": 574, "lr": 9.162594623991835e-05}
{"train_loss": 0.23443497717380524, "global_step": 51140, "epoch": 574, "lr": 9.162562506500947e-05}
{"train_loss": 0.30146098136901855, "global_step": 51141, "epoch": 574, "lr": 9.162530388450453e-05}
{"train_loss": 0.25824204087257385, "global_step": 51142, "epoch": 574, "lr": 9.162498269840357e-05}
{"train_loss": 0.23842865228652954, "global_step": 51143, "epoch": 574, "lr": 9.162466150670665e-05}
{"train_loss": 0.18075034022331238, "global_step": 51144, "epoch": 574, "lr": 9.162434030941379e-05}
{"train_loss": 0.2298031598329544, "global_step": 51145, "epoch": 574, "lr": 9.162401910652505e-05}
{"train_loss": 0.253663569688797, "global_step": 51146, "epoch": 574, "lr": 9.162369789804047e-05}
{"train_loss": 0.3019247055053711, "global_step": 51147, "epoch": 574, "lr": 9.162337668396008e-05}
{"train_loss": 0.2788800299167633, "global_step": 51148, "epoch": 574, "lr": 9.162305546428395e-05}
{"train_loss": 0.2026304304599762, "global_step": 51149, "epoch": 574, "lr": 9.16227342390121e-05}
{"train_loss": 0.24257704615592957, "global_step": 51150, "epoch": 574, "lr": 9.162241300814458e-05}
{"train_loss": 0.20432355999946594, "global_step": 51151, "epoch": 574, "lr": 9.162209177168144e-05}
{"train_loss": 0.2862987816333771, "global_step": 51152, "epoch": 574, "lr": 9.162177052962271e-05}
{"train_loss": 0.2779611349105835, "global_step": 51153, "epoch": 574, "lr": 9.162144928196844e-05}
{"train_loss": 0.3333239257335663, "global_step": 51154, "epoch": 574, "lr": 9.162112802871867e-05}
{"train_loss": 0.2797676920890808, "global_step": 51155, "epoch": 574, "lr": 9.162080676987346e-05}
{"train_loss": 0.23871365189552307, "global_step": 51156, "epoch": 574, "lr": 9.162048550543283e-05}
{"train_loss": 0.3671722710132599, "global_step": 51157, "epoch": 574, "lr": 9.162016423539683e-05}
{"train_loss": 0.4039006531238556, "global_step": 51158, "epoch": 574, "lr": 9.161984295976551e-05}
{"train_loss": 0.24075251817703247, "global_step": 51159, "epoch": 574, "lr": 9.161952167853891e-05}
{"train_loss": 0.2298537641763687, "global_step": 51160, "epoch": 574, "lr": 9.161920039171707e-05}
{"train_loss": 0.24039891362190247, "global_step": 51161, "epoch": 574, "lr": 9.161887909930004e-05}
{"train_loss": 0.42226582765579224, "global_step": 51162, "epoch": 574, "lr": 9.161855780128787e-05}
{"train_loss": 0.18980760872364044, "global_step": 51163, "epoch": 574, "lr": 9.161823649768057e-05}
{"train_loss": 0.28779393434524536, "global_step": 51164, "epoch": 574, "lr": 9.161791518847822e-05}
{"train_loss": 0.217779740691185, "global_step": 51165, "epoch": 574, "lr": 9.161759387368085e-05}
{"train_loss": 0.28636643290519714, "global_step": 51166, "epoch": 574, "lr": 9.161727255328849e-05}
{"train_loss": 0.3705917298793793, "global_step": 51167, "epoch": 574, "lr": 9.16169512273012e-05}
{"train_loss": 0.23947441577911377, "global_step": 51168, "epoch": 574, "lr": 9.161662989571902e-05}
{"train_loss": 0.3000379204750061, "global_step": 51169, "epoch": 574, "lr": 9.1616308558542e-05}
{"train_loss": 0.23025955259799957, "global_step": 51170, "epoch": 574, "lr": 9.161598721577016e-05}
{"train_loss": 0.2792571485042572, "global_step": 51171, "epoch": 574, "lr": 9.161566586740356e-05}
{"train_loss": 0.2772372364997864, "global_step": 51172, "epoch": 574, "lr": 9.161534451344225e-05}
{"train_loss": 0.19728820025920868, "global_step": 51173, "epoch": 574, "lr": 9.161502315388626e-05}
{"train_loss": 0.2605907302559092, "global_step": 51174, "epoch": 574, "lr": 9.161470178873564e-05, "val_loss": 3.1178314685821533}
{"train_loss": 0.22413882613182068, "global_step": 51175, "epoch": 575, "lr": 9.161438041799042e-05}
{"train_loss": 0.28800660371780396, "global_step": 51176, "epoch": 575, "lr": 9.161405904165066e-05}
{"train_loss": 0.26235252618789673, "global_step": 51177, "epoch": 575, "lr": 9.16137376597164e-05}
{"train_loss": 0.20296616852283478, "global_step": 51178, "epoch": 575, "lr": 9.161341627218766e-05}
{"train_loss": 0.18234851956367493, "global_step": 51179, "epoch": 575, "lr": 9.161309487906453e-05}
{"train_loss": 0.2714901566505432, "global_step": 51180, "epoch": 575, "lr": 9.161277348034701e-05}
{"train_loss": 0.3786754906177521, "global_step": 51181, "epoch": 575, "lr": 9.161245207603517e-05}
{"train_loss": 0.31384170055389404, "global_step": 51182, "epoch": 575, "lr": 9.161213066612904e-05}
{"train_loss": 0.29321014881134033, "global_step": 51183, "epoch": 575, "lr": 9.161180925062867e-05}
{"train_loss": 0.27889561653137207, "global_step": 51184, "epoch": 575, "lr": 9.161148782953408e-05}
{"train_loss": 0.27886974811553955, "global_step": 51185, "epoch": 575, "lr": 9.161116640284535e-05}
{"train_loss": 0.222660630941391, "global_step": 51186, "epoch": 575, "lr": 9.161084497056251e-05}
{"train_loss": 0.2236008197069168, "global_step": 51187, "epoch": 575, "lr": 9.16105235326856e-05}
{"train_loss": 0.24104370176792145, "global_step": 51188, "epoch": 575, "lr": 9.161020208921465e-05}
{"train_loss": 0.22195811569690704, "global_step": 51189, "epoch": 575, "lr": 9.160988064014972e-05}
{"train_loss": 0.2770088016986847, "global_step": 51190, "epoch": 575, "lr": 9.160955918549085e-05}
{"train_loss": 0.36321747303009033, "global_step": 51191, "epoch": 575, "lr": 9.160923772523808e-05}
{"train_loss": 0.25861072540283203, "global_step": 51192, "epoch": 575, "lr": 9.160891625939145e-05}
{"train_loss": 0.2563551962375641, "global_step": 51193, "epoch": 575, "lr": 9.160859478795102e-05}
{"train_loss": 0.29652684926986694, "global_step": 51194, "epoch": 575, "lr": 9.160827331091682e-05}
{"train_loss": 0.23033082485198975, "global_step": 51195, "epoch": 575, "lr": 9.160795182828888e-05}
{"train_loss": 0.17407453060150146, "global_step": 51196, "epoch": 575, "lr": 9.160763034006727e-05}
{"train_loss": 0.318212628364563, "global_step": 51197, "epoch": 575, "lr": 9.160730884625202e-05}
{"train_loss": 0.2759091258049011, "global_step": 51198, "epoch": 575, "lr": 9.160698734684318e-05}
{"train_loss": 0.3445727825164795, "global_step": 51199, "epoch": 575, "lr": 9.160666584184077e-05}
{"train_loss": 0.22027750313282013, "global_step": 51200, "epoch": 575, "lr": 9.160634433124486e-05}
{"train_loss": 0.30695557594299316, "global_step": 51201, "epoch": 575, "lr": 9.16060228150555e-05}
{"train_loss": 0.29156550765037537, "global_step": 51202, "epoch": 575, "lr": 9.160570129327271e-05}
{"train_loss": 0.20106932520866394, "global_step": 51203, "epoch": 575, "lr": 9.160537976589652e-05}
{"train_loss": 0.3130686283111572, "global_step": 51204, "epoch": 575, "lr": 9.160505823292703e-05}
{"train_loss": 0.21902920305728912, "global_step": 51205, "epoch": 575, "lr": 9.160473669436422e-05}
{"train_loss": 0.3461463749408722, "global_step": 51206, "epoch": 575, "lr": 9.160441515020816e-05}
{"train_loss": 0.24627096951007843, "global_step": 51207, "epoch": 575, "lr": 9.16040936004589e-05}
{"train_loss": 0.1736767292022705, "global_step": 51208, "epoch": 575, "lr": 9.160377204511647e-05}
{"train_loss": 0.21502424776554108, "global_step": 51209, "epoch": 575, "lr": 9.160345048418093e-05}
{"train_loss": 0.3025714159011841, "global_step": 51210, "epoch": 575, "lr": 9.160312891765231e-05}
{"train_loss": 0.35898274183273315, "global_step": 51211, "epoch": 575, "lr": 9.160280734553066e-05}
{"train_loss": 0.3783840835094452, "global_step": 51212, "epoch": 575, "lr": 9.160248576781601e-05}
{"train_loss": 0.30968838930130005, "global_step": 51213, "epoch": 575, "lr": 9.160216418450842e-05}
{"train_loss": 0.20892339944839478, "global_step": 51214, "epoch": 575, "lr": 9.160184259560792e-05}
{"train_loss": 0.38579997420310974, "global_step": 51215, "epoch": 575, "lr": 9.160152100111457e-05}
{"train_loss": 0.3419356346130371, "global_step": 51216, "epoch": 575, "lr": 9.16011994010284e-05}
{"train_loss": 0.33990082144737244, "global_step": 51217, "epoch": 575, "lr": 9.160087779534945e-05}
{"train_loss": 0.1772868037223816, "global_step": 51218, "epoch": 575, "lr": 9.160055618407777e-05}
{"train_loss": 0.29426273703575134, "global_step": 51219, "epoch": 575, "lr": 9.160023456721341e-05}
{"train_loss": 0.292258083820343, "global_step": 51220, "epoch": 575, "lr": 9.159991294475639e-05}
{"train_loss": 0.31233909726142883, "global_step": 51221, "epoch": 575, "lr": 9.159959131670677e-05}
{"train_loss": 0.2813628315925598, "global_step": 51222, "epoch": 575, "lr": 9.15992696830646e-05}
{"train_loss": 0.34002771973609924, "global_step": 51223, "epoch": 575, "lr": 9.159894804382991e-05}
{"train_loss": 0.30429938435554504, "global_step": 51224, "epoch": 575, "lr": 9.159862639900276e-05}
{"train_loss": 0.2430712729692459, "global_step": 51225, "epoch": 575, "lr": 9.159830474858318e-05}
{"train_loss": 0.2858739495277405, "global_step": 51226, "epoch": 575, "lr": 9.159798309257121e-05}
{"train_loss": 0.22093427181243896, "global_step": 51227, "epoch": 575, "lr": 9.159766143096689e-05}
{"train_loss": 0.3408096730709076, "global_step": 51228, "epoch": 575, "lr": 9.159733976377028e-05}
{"train_loss": 0.32382670044898987, "global_step": 51229, "epoch": 575, "lr": 9.159701809098143e-05}
{"train_loss": 0.2398681938648224, "global_step": 51230, "epoch": 575, "lr": 9.159669641260034e-05}
{"train_loss": 0.2815631926059723, "global_step": 51231, "epoch": 575, "lr": 9.15963747286271e-05}
{"train_loss": 0.22590330243110657, "global_step": 51232, "epoch": 575, "lr": 9.159605303906172e-05}
{"train_loss": 0.3280484080314636, "global_step": 51233, "epoch": 575, "lr": 9.159573134390427e-05}
{"train_loss": 0.21136821806430817, "global_step": 51234, "epoch": 575, "lr": 9.159540964315476e-05}
{"train_loss": 0.28949448466300964, "global_step": 51235, "epoch": 575, "lr": 9.159508793681329e-05}
{"train_loss": 0.26096636056900024, "global_step": 51236, "epoch": 575, "lr": 9.159476622487984e-05}
{"train_loss": 0.25329139828681946, "global_step": 51237, "epoch": 575, "lr": 9.159444450735449e-05}
{"train_loss": 0.26321980357170105, "global_step": 51238, "epoch": 575, "lr": 9.159412278423726e-05}
{"train_loss": 0.394639790058136, "global_step": 51239, "epoch": 575, "lr": 9.159380105552823e-05}
{"train_loss": 0.3534979522228241, "global_step": 51240, "epoch": 575, "lr": 9.159347932122742e-05}
{"train_loss": 0.25747019052505493, "global_step": 51241, "epoch": 575, "lr": 9.159315758133484e-05}
{"train_loss": 0.2875479757785797, "global_step": 51242, "epoch": 575, "lr": 9.15928358358506e-05}
{"train_loss": 0.2502007782459259, "global_step": 51243, "epoch": 575, "lr": 9.15925140847747e-05}
{"train_loss": 0.3091186285018921, "global_step": 51244, "epoch": 575, "lr": 9.159219232810719e-05}
{"train_loss": 0.29350733757019043, "global_step": 51245, "epoch": 575, "lr": 9.159187056584813e-05}
{"train_loss": 0.30229121446609497, "global_step": 51246, "epoch": 575, "lr": 9.159154879799754e-05}
{"train_loss": 0.26995936036109924, "global_step": 51247, "epoch": 575, "lr": 9.159122702455547e-05}
{"train_loss": 0.32694169878959656, "global_step": 51248, "epoch": 575, "lr": 9.159090524552197e-05}
{"train_loss": 0.26525020599365234, "global_step": 51249, "epoch": 575, "lr": 9.159058346089708e-05}
{"train_loss": 0.25401559472084045, "global_step": 51250, "epoch": 575, "lr": 9.159026167068085e-05}
{"train_loss": 0.33174580335617065, "global_step": 51251, "epoch": 575, "lr": 9.15899398748733e-05}
{"train_loss": 0.3628895580768585, "global_step": 51252, "epoch": 575, "lr": 9.15896180734745e-05}
{"train_loss": 0.2569964826107025, "global_step": 51253, "epoch": 575, "lr": 9.158929626648448e-05}
{"train_loss": 0.2557316720485687, "global_step": 51254, "epoch": 575, "lr": 9.158897445390329e-05}
{"train_loss": 0.2049986571073532, "global_step": 51255, "epoch": 575, "lr": 9.158865263573097e-05}
{"train_loss": 0.22166192531585693, "global_step": 51256, "epoch": 575, "lr": 9.158833081196756e-05}
{"train_loss": 0.26710766553878784, "global_step": 51257, "epoch": 575, "lr": 9.15880089826131e-05}
{"train_loss": 0.3295542597770691, "global_step": 51258, "epoch": 575, "lr": 9.158768714766764e-05}
{"train_loss": 0.36276480555534363, "global_step": 51259, "epoch": 575, "lr": 9.158736530713124e-05}
{"train_loss": 0.28625839948654175, "global_step": 51260, "epoch": 575, "lr": 9.158704346100391e-05}
{"train_loss": 0.24031677842140198, "global_step": 51261, "epoch": 575, "lr": 9.15867216092857e-05}
{"train_loss": 0.21791857481002808, "global_step": 51262, "epoch": 575, "lr": 9.158639975197667e-05}
{"train_loss": 0.2798629612400291, "global_step": 51263, "epoch": 575, "lr": 9.158607788907686e-05, "val_loss": 3.1434361934661865, "train_action_mse_error": 40.446678161621094}
{"train_loss": 0.2806236147880554, "global_step": 51264, "epoch": 576, "lr": 9.158575602058632e-05}
{"train_loss": 0.3238770067691803, "global_step": 51265, "epoch": 576, "lr": 9.158543414650507e-05}
{"train_loss": 0.2648341953754425, "global_step": 51266, "epoch": 576, "lr": 9.158511226683316e-05}
{"train_loss": 0.28750181198120117, "global_step": 51267, "epoch": 576, "lr": 9.158479038157064e-05}
{"train_loss": 0.25055253505706787, "global_step": 51268, "epoch": 576, "lr": 9.158446849071756e-05}
{"train_loss": 0.26410314440727234, "global_step": 51269, "epoch": 576, "lr": 9.158414659427395e-05}
{"train_loss": 0.24172212183475494, "global_step": 51270, "epoch": 576, "lr": 9.158382469223985e-05}
{"train_loss": 0.30527010560035706, "global_step": 51271, "epoch": 576, "lr": 9.158350278461531e-05}
{"train_loss": 0.32505419850349426, "global_step": 51272, "epoch": 576, "lr": 9.158318087140038e-05}
{"train_loss": 0.251948744058609, "global_step": 51273, "epoch": 576, "lr": 9.15828589525951e-05}
{"train_loss": 0.23444253206253052, "global_step": 51274, "epoch": 576, "lr": 9.158253702819953e-05}
{"train_loss": 0.35836395621299744, "global_step": 51275, "epoch": 576, "lr": 9.158221509821367e-05}
{"train_loss": 0.18968608975410461, "global_step": 51276, "epoch": 576, "lr": 9.158189316263762e-05}
{"train_loss": 0.26443490386009216, "global_step": 51277, "epoch": 576, "lr": 9.158157122147136e-05}
{"train_loss": 0.22888600826263428, "global_step": 51278, "epoch": 576, "lr": 9.158124927471497e-05}
{"train_loss": 0.29340478777885437, "global_step": 51279, "epoch": 576, "lr": 9.15809273223685e-05}
{"train_loss": 0.25454890727996826, "global_step": 51280, "epoch": 576, "lr": 9.158060536443196e-05}
{"train_loss": 0.26772233843803406, "global_step": 51281, "epoch": 576, "lr": 9.158028340090543e-05}
{"train_loss": 0.2318696230649948, "global_step": 51282, "epoch": 576, "lr": 9.157996143178893e-05}
{"train_loss": 0.22509810328483582, "global_step": 51283, "epoch": 576, "lr": 9.157963945708251e-05}
{"train_loss": 0.3357540965080261, "global_step": 51284, "epoch": 576, "lr": 9.157931747678623e-05}
{"train_loss": 0.30953559279441833, "global_step": 51285, "epoch": 576, "lr": 9.15789954909001e-05}
{"train_loss": 0.3467593491077423, "global_step": 51286, "epoch": 576, "lr": 9.15786734994242e-05}
{"train_loss": 0.3464506268501282, "global_step": 51287, "epoch": 576, "lr": 9.157835150235852e-05}
{"train_loss": 0.3050403594970703, "global_step": 51288, "epoch": 576, "lr": 9.157802949970317e-05}
{"train_loss": 0.24041461944580078, "global_step": 51289, "epoch": 576, "lr": 9.157770749145814e-05}
{"train_loss": 0.29744696617126465, "global_step": 51290, "epoch": 576, "lr": 9.157738547762352e-05}
{"train_loss": 0.1824905276298523, "global_step": 51291, "epoch": 576, "lr": 9.15770634581993e-05}
{"train_loss": 0.25454971194267273, "global_step": 51292, "epoch": 576, "lr": 9.157674143318555e-05}
{"train_loss": 0.27001953125, "global_step": 51293, "epoch": 576, "lr": 9.157641940258233e-05}
{"train_loss": 0.24845336377620697, "global_step": 51294, "epoch": 576, "lr": 9.157609736638967e-05}
{"train_loss": 0.20243974030017853, "global_step": 51295, "epoch": 576, "lr": 9.15757753246076e-05}
{"train_loss": 0.3225436508655548, "global_step": 51296, "epoch": 576, "lr": 9.157545327723618e-05}
{"train_loss": 0.2750346064567566, "global_step": 51297, "epoch": 576, "lr": 9.157513122427544e-05}
{"train_loss": 0.36801713705062866, "global_step": 51298, "epoch": 576, "lr": 9.157480916572543e-05}
{"train_loss": 0.18877941370010376, "global_step": 51299, "epoch": 576, "lr": 9.15744871015862e-05}
{"train_loss": 0.19631333649158478, "global_step": 51300, "epoch": 576, "lr": 9.157416503185778e-05}
{"train_loss": 0.28941574692726135, "global_step": 51301, "epoch": 576, "lr": 9.157384295654024e-05}
{"train_loss": 0.39112889766693115, "global_step": 51302, "epoch": 576, "lr": 9.157352087563357e-05}
{"train_loss": 0.31305694580078125, "global_step": 51303, "epoch": 576, "lr": 9.157319878913789e-05}
{"train_loss": 0.333538293838501, "global_step": 51304, "epoch": 576, "lr": 9.157287669705317e-05}
{"train_loss": 0.20686385035514832, "global_step": 51305, "epoch": 576, "lr": 9.157255459937948e-05}
{"train_loss": 0.2818617522716522, "global_step": 51306, "epoch": 576, "lr": 9.157223249611688e-05}
{"train_loss": 0.3153948187828064, "global_step": 51307, "epoch": 576, "lr": 9.157191038726538e-05}
{"train_loss": 0.27189457416534424, "global_step": 51308, "epoch": 576, "lr": 9.157158827282506e-05}
{"train_loss": 0.29768916964530945, "global_step": 51309, "epoch": 576, "lr": 9.157126615279595e-05}
{"train_loss": 0.2391405999660492, "global_step": 51310, "epoch": 576, "lr": 9.157094402717809e-05}
{"train_loss": 0.29123613238334656, "global_step": 51311, "epoch": 576, "lr": 9.157062189597152e-05}
{"train_loss": 0.23746493458747864, "global_step": 51312, "epoch": 576, "lr": 9.157029975917628e-05}
{"train_loss": 0.2164345681667328, "global_step": 51313, "epoch": 576, "lr": 9.156997761679243e-05}
{"train_loss": 0.24209263920783997, "global_step": 51314, "epoch": 576, "lr": 9.156965546882e-05}
{"train_loss": 0.26954367756843567, "global_step": 51315, "epoch": 576, "lr": 9.156933331525904e-05}
{"train_loss": 0.2396741509437561, "global_step": 51316, "epoch": 576, "lr": 9.156901115610957e-05}
{"train_loss": 0.20654691755771637, "global_step": 51317, "epoch": 576, "lr": 9.156868899137167e-05}
{"train_loss": 0.24870795011520386, "global_step": 51318, "epoch": 576, "lr": 9.156836682104537e-05}
{"train_loss": 0.30073800683021545, "global_step": 51319, "epoch": 576, "lr": 9.15680446451307e-05}
{"train_loss": 0.22468000650405884, "global_step": 51320, "epoch": 576, "lr": 9.156772246362771e-05}
{"train_loss": 0.22291439771652222, "global_step": 51321, "epoch": 576, "lr": 9.156740027653646e-05}
{"train_loss": 0.21526457369327545, "global_step": 51322, "epoch": 576, "lr": 9.156707808385697e-05}
{"train_loss": 0.3691073954105377, "global_step": 51323, "epoch": 576, "lr": 9.156675588558929e-05}
{"train_loss": 0.3413718044757843, "global_step": 51324, "epoch": 576, "lr": 9.156643368173348e-05}
{"train_loss": 0.2683652341365814, "global_step": 51325, "epoch": 576, "lr": 9.156611147228957e-05}
{"train_loss": 0.27287691831588745, "global_step": 51326, "epoch": 576, "lr": 9.156578925725758e-05}
{"train_loss": 0.3064059615135193, "global_step": 51327, "epoch": 576, "lr": 9.15654670366376e-05}
{"train_loss": 0.27879825234413147, "global_step": 51328, "epoch": 576, "lr": 9.156514481042963e-05}
{"train_loss": 0.26283329725265503, "global_step": 51329, "epoch": 576, "lr": 9.156482257863374e-05}
{"train_loss": 0.28916752338409424, "global_step": 51330, "epoch": 576, "lr": 9.156450034124998e-05}
{"train_loss": 0.27106398344039917, "global_step": 51331, "epoch": 576, "lr": 9.156417809827837e-05}
{"train_loss": 0.3093181848526001, "global_step": 51332, "epoch": 576, "lr": 9.156385584971897e-05}
{"train_loss": 0.27525413036346436, "global_step": 51333, "epoch": 576, "lr": 9.15635335955718e-05}
{"train_loss": 0.15438614785671234, "global_step": 51334, "epoch": 576, "lr": 9.156321133583692e-05}
{"train_loss": 0.27039802074432373, "global_step": 51335, "epoch": 576, "lr": 9.15628890705144e-05}
{"train_loss": 0.3258058428764343, "global_step": 51336, "epoch": 576, "lr": 9.156256679960422e-05}
{"train_loss": 0.3315276503562927, "global_step": 51337, "epoch": 576, "lr": 9.156224452310649e-05}
{"train_loss": 0.29291775822639465, "global_step": 51338, "epoch": 576, "lr": 9.156192224102121e-05}
{"train_loss": 0.23096325993537903, "global_step": 51339, "epoch": 576, "lr": 9.156159995334843e-05}
{"train_loss": 0.4180724322795868, "global_step": 51340, "epoch": 576, "lr": 9.15612776600882e-05}
{"train_loss": 0.2868691682815552, "global_step": 51341, "epoch": 576, "lr": 9.156095536124057e-05}
{"train_loss": 0.233265221118927, "global_step": 51342, "epoch": 576, "lr": 9.156063305680557e-05}
{"train_loss": 0.2508832812309265, "global_step": 51343, "epoch": 576, "lr": 9.156031074678326e-05}
{"train_loss": 0.25615906715393066, "global_step": 51344, "epoch": 576, "lr": 9.155998843117366e-05}
{"train_loss": 0.23357701301574707, "global_step": 51345, "epoch": 576, "lr": 9.155966610997682e-05}
{"train_loss": 0.2808099687099457, "global_step": 51346, "epoch": 576, "lr": 9.155934378319281e-05}
{"train_loss": 0.17269830405712128, "global_step": 51347, "epoch": 576, "lr": 9.155902145082165e-05}
{"train_loss": 0.278093159198761, "global_step": 51348, "epoch": 576, "lr": 9.155869911286336e-05}
{"train_loss": 0.2647532522678375, "global_step": 51349, "epoch": 576, "lr": 9.155837676931805e-05}
{"train_loss": 0.28908222913742065, "global_step": 51350, "epoch": 576, "lr": 9.15580544201857e-05}
{"train_loss": 0.2387886643409729, "global_step": 51351, "epoch": 576, "lr": 9.155773206546637e-05}
{"train_loss": 0.2726139762763227, "global_step": 51352, "epoch": 576, "lr": 9.155740970516012e-05, "val_loss": 2.980773687362671}
{"train_loss": 0.2809791564941406, "global_step": 51353, "epoch": 577, "lr": 9.155708733926698e-05}
{"train_loss": 0.2246125042438507, "global_step": 51354, "epoch": 577, "lr": 9.1556764967787e-05}
{"train_loss": 0.26216065883636475, "global_step": 51355, "epoch": 577, "lr": 9.155644259072022e-05}
{"train_loss": 0.2516128122806549, "global_step": 51356, "epoch": 577, "lr": 9.155612020806667e-05}
{"train_loss": 0.433265745639801, "global_step": 51357, "epoch": 577, "lr": 9.155579781982641e-05}
{"train_loss": 0.32520297169685364, "global_step": 51358, "epoch": 577, "lr": 9.15554754259995e-05}
{"train_loss": 0.19508400559425354, "global_step": 51359, "epoch": 577, "lr": 9.155515302658594e-05}
{"train_loss": 0.197363018989563, "global_step": 51360, "epoch": 577, "lr": 9.15548306215858e-05}
{"train_loss": 0.3084297180175781, "global_step": 51361, "epoch": 577, "lr": 9.155450821099912e-05}
{"train_loss": 0.3150097131729126, "global_step": 51362, "epoch": 577, "lr": 9.155418579482595e-05}
{"train_loss": 0.30074477195739746, "global_step": 51363, "epoch": 577, "lr": 9.155386337306633e-05}
{"train_loss": 0.26923778653144836, "global_step": 51364, "epoch": 577, "lr": 9.155354094572029e-05}
{"train_loss": 0.3627200126647949, "global_step": 51365, "epoch": 577, "lr": 9.155321851278788e-05}
{"train_loss": 0.3217819035053253, "global_step": 51366, "epoch": 577, "lr": 9.155289607426915e-05}
{"train_loss": 0.3870581388473511, "global_step": 51367, "epoch": 577, "lr": 9.155257363016414e-05}
{"train_loss": 0.23055438697338104, "global_step": 51368, "epoch": 577, "lr": 9.155225118047289e-05}
{"train_loss": 0.24073438346385956, "global_step": 51369, "epoch": 577, "lr": 9.155192872519544e-05}
{"train_loss": 0.23173102736473083, "global_step": 51370, "epoch": 577, "lr": 9.155160626433186e-05}
{"train_loss": 0.26479730010032654, "global_step": 51371, "epoch": 577, "lr": 9.155128379788216e-05}
{"train_loss": 0.20461329817771912, "global_step": 51372, "epoch": 577, "lr": 9.155096132584639e-05}
{"train_loss": 0.322126567363739, "global_step": 51373, "epoch": 577, "lr": 9.155063884822461e-05}
{"train_loss": 0.25903719663619995, "global_step": 51374, "epoch": 577, "lr": 9.155031636501687e-05}
{"train_loss": 0.2871270179748535, "global_step": 51375, "epoch": 577, "lr": 9.154999387622316e-05}
{"train_loss": 0.222925066947937, "global_step": 51376, "epoch": 577, "lr": 9.154967138184358e-05}
{"train_loss": 0.30906862020492554, "global_step": 51377, "epoch": 577, "lr": 9.154934888187817e-05}
{"train_loss": 0.24922941625118256, "global_step": 51378, "epoch": 577, "lr": 9.154902637632692e-05}
{"train_loss": 0.283892422914505, "global_step": 51379, "epoch": 577, "lr": 9.154870386518993e-05}
{"train_loss": 0.2948818504810333, "global_step": 51380, "epoch": 577, "lr": 9.154838134846722e-05}
{"train_loss": 0.40085867047309875, "global_step": 51381, "epoch": 577, "lr": 9.154805882615883e-05}
{"train_loss": 0.2975064814090729, "global_step": 51382, "epoch": 577, "lr": 9.154773629826483e-05}
{"train_loss": 0.32779455184936523, "global_step": 51383, "epoch": 577, "lr": 9.154741376478522e-05}
{"train_loss": 0.29409849643707275, "global_step": 51384, "epoch": 577, "lr": 9.154709122572008e-05}
{"train_loss": 0.22116105258464813, "global_step": 51385, "epoch": 577, "lr": 9.154676868106943e-05}
{"train_loss": 0.2663269639015198, "global_step": 51386, "epoch": 577, "lr": 9.154644613083335e-05}
{"train_loss": 0.34551355242729187, "global_step": 51387, "epoch": 577, "lr": 9.154612357501182e-05}
{"train_loss": 0.1556035578250885, "global_step": 51388, "epoch": 577, "lr": 9.154580101360494e-05}
{"train_loss": 0.29680919647216797, "global_step": 51389, "epoch": 577, "lr": 9.154547844661272e-05}
{"train_loss": 0.3149346113204956, "global_step": 51390, "epoch": 577, "lr": 9.154515587403523e-05}
{"train_loss": 0.266207218170166, "global_step": 51391, "epoch": 577, "lr": 9.154483329587249e-05}
{"train_loss": 0.24980562925338745, "global_step": 51392, "epoch": 577, "lr": 9.154451071212457e-05}
{"train_loss": 0.28646236658096313, "global_step": 51393, "epoch": 577, "lr": 9.154418812279149e-05}
{"train_loss": 0.29791465401649475, "global_step": 51394, "epoch": 577, "lr": 9.154386552787329e-05}
{"train_loss": 0.21798133850097656, "global_step": 51395, "epoch": 577, "lr": 9.154354292737004e-05}
{"train_loss": 0.1926422119140625, "global_step": 51396, "epoch": 577, "lr": 9.154322032128175e-05}
{"train_loss": 0.2793653607368469, "global_step": 51397, "epoch": 577, "lr": 9.154289770960848e-05}
{"train_loss": 0.29756981134414673, "global_step": 51398, "epoch": 577, "lr": 9.154257509235029e-05}
{"train_loss": 0.1974356472492218, "global_step": 51399, "epoch": 577, "lr": 9.15422524695072e-05}
{"train_loss": 0.23032525181770325, "global_step": 51400, "epoch": 577, "lr": 9.154192984107925e-05}
{"train_loss": 0.24457165598869324, "global_step": 51401, "epoch": 577, "lr": 9.15416072070665e-05}
{"train_loss": 0.27840688824653625, "global_step": 51402, "epoch": 577, "lr": 9.1541284567469e-05}
{"train_loss": 0.3419056832790375, "global_step": 51403, "epoch": 577, "lr": 9.154096192228677e-05}
{"train_loss": 0.26165154576301575, "global_step": 51404, "epoch": 577, "lr": 9.154063927151987e-05}
{"train_loss": 0.22865824401378632, "global_step": 51405, "epoch": 577, "lr": 9.154031661516834e-05}
{"train_loss": 0.21335597336292267, "global_step": 51406, "epoch": 577, "lr": 9.153999395323221e-05}
{"train_loss": 0.3279798626899719, "global_step": 51407, "epoch": 577, "lr": 9.153967128571153e-05}
{"train_loss": 0.21782730519771576, "global_step": 51408, "epoch": 577, "lr": 9.153934861260636e-05}
{"train_loss": 0.23310786485671997, "global_step": 51409, "epoch": 577, "lr": 9.153902593391672e-05}
{"train_loss": 0.2758825421333313, "global_step": 51410, "epoch": 577, "lr": 9.153870324964266e-05}
{"train_loss": 0.18442493677139282, "global_step": 51411, "epoch": 577, "lr": 9.153838055978425e-05}
{"train_loss": 0.2710554301738739, "global_step": 51412, "epoch": 577, "lr": 9.15380578643415e-05}
{"train_loss": 0.293661892414093, "global_step": 51413, "epoch": 577, "lr": 9.153773516331446e-05}
{"train_loss": 0.2580573558807373, "global_step": 51414, "epoch": 577, "lr": 9.153741245670318e-05}
{"train_loss": 0.23409779369831085, "global_step": 51415, "epoch": 577, "lr": 9.15370897445077e-05}
{"train_loss": 0.22021079063415527, "global_step": 51416, "epoch": 577, "lr": 9.153676702672806e-05}
{"train_loss": 0.25627607107162476, "global_step": 51417, "epoch": 577, "lr": 9.153644430336432e-05}
{"train_loss": 0.3292846381664276, "global_step": 51418, "epoch": 577, "lr": 9.153612157441651e-05}
{"train_loss": 0.23014730215072632, "global_step": 51419, "epoch": 577, "lr": 9.153579883988467e-05}
{"train_loss": 0.22388887405395508, "global_step": 51420, "epoch": 577, "lr": 9.153547609976885e-05}
{"train_loss": 0.20032650232315063, "global_step": 51421, "epoch": 577, "lr": 9.153515335406909e-05}
{"train_loss": 0.3377009332180023, "global_step": 51422, "epoch": 577, "lr": 9.153483060278544e-05}
{"train_loss": 0.21401873230934143, "global_step": 51423, "epoch": 577, "lr": 9.153450784591793e-05}
{"train_loss": 0.20225611329078674, "global_step": 51424, "epoch": 577, "lr": 9.15341850834666e-05}
{"train_loss": 0.2825109660625458, "global_step": 51425, "epoch": 577, "lr": 9.153386231543152e-05}
{"train_loss": 0.315163791179657, "global_step": 51426, "epoch": 577, "lr": 9.153353954181274e-05}
{"train_loss": 0.22127145528793335, "global_step": 51427, "epoch": 577, "lr": 9.153321676261025e-05}
{"train_loss": 0.21996982395648956, "global_step": 51428, "epoch": 577, "lr": 9.153289397782414e-05}
{"train_loss": 0.24168166518211365, "global_step": 51429, "epoch": 577, "lr": 9.153257118745443e-05}
{"train_loss": 0.20817671716213226, "global_step": 51430, "epoch": 577, "lr": 9.153224839150119e-05}
{"train_loss": 0.23909471929073334, "global_step": 51431, "epoch": 577, "lr": 9.153192558996442e-05}
{"train_loss": 0.20807500183582306, "global_step": 51432, "epoch": 577, "lr": 9.153160278284419e-05}
{"train_loss": 0.23571711778640747, "global_step": 51433, "epoch": 577, "lr": 9.153127997014056e-05}
{"train_loss": 0.25484293699264526, "global_step": 51434, "epoch": 577, "lr": 9.153095715185354e-05}
{"train_loss": 0.21887581050395966, "global_step": 51435, "epoch": 577, "lr": 9.15306343279832e-05}
{"train_loss": 0.27772584557533264, "global_step": 51436, "epoch": 577, "lr": 9.153031149852958e-05}
{"train_loss": 0.256197065114975, "global_step": 51437, "epoch": 577, "lr": 9.15299886634927e-05}
{"train_loss": 0.2423402965068817, "global_step": 51438, "epoch": 577, "lr": 9.152966582287264e-05}
{"train_loss": 0.27199459075927734, "global_step": 51439, "epoch": 577, "lr": 9.15293429766694e-05}
{"train_loss": 0.1752750724554062, "global_step": 51440, "epoch": 577, "lr": 9.152902012488305e-05}
{"train_loss": 0.2644323773598403, "global_step": 51441, "epoch": 577, "lr": 9.152869726751363e-05, "val_loss": 3.144641637802124}
{"train_loss": 0.21334314346313477, "global_step": 51442, "epoch": 578, "lr": 9.15283744045612e-05}
{"train_loss": 0.27918311953544617, "global_step": 51443, "epoch": 578, "lr": 9.152805153602575e-05}
{"train_loss": 0.25492313504219055, "global_step": 51444, "epoch": 578, "lr": 9.152772866190739e-05}
{"train_loss": 0.3830769658088684, "global_step": 51445, "epoch": 578, "lr": 9.152740578220612e-05}
{"train_loss": 0.2778315544128418, "global_step": 51446, "epoch": 578, "lr": 9.1527082896922e-05}
{"train_loss": 0.21544405817985535, "global_step": 51447, "epoch": 578, "lr": 9.152676000605507e-05}
{"train_loss": 0.2615284323692322, "global_step": 51448, "epoch": 578, "lr": 9.152643710960538e-05}
{"train_loss": 0.2627655863761902, "global_step": 51449, "epoch": 578, "lr": 9.152611420757295e-05}
{"train_loss": 0.3191315233707428, "global_step": 51450, "epoch": 578, "lr": 9.152579129995785e-05}
{"train_loss": 0.24736207723617554, "global_step": 51451, "epoch": 578, "lr": 9.152546838676012e-05}
{"train_loss": 0.2678886950016022, "global_step": 51452, "epoch": 578, "lr": 9.152514546797979e-05}
{"train_loss": 0.4247630536556244, "global_step": 51453, "epoch": 578, "lr": 9.152482254361692e-05}
{"train_loss": 0.34205639362335205, "global_step": 51454, "epoch": 578, "lr": 9.152449961367153e-05}
{"train_loss": 0.3035433888435364, "global_step": 51455, "epoch": 578, "lr": 9.152417667814369e-05}
{"train_loss": 0.2587072253227234, "global_step": 51456, "epoch": 578, "lr": 9.152385373703342e-05}
{"train_loss": 0.18474169075489044, "global_step": 51457, "epoch": 578, "lr": 9.152353079034077e-05}
{"train_loss": 0.1836378127336502, "global_step": 51458, "epoch": 578, "lr": 9.152320783806579e-05}
{"train_loss": 0.29645130038261414, "global_step": 51459, "epoch": 578, "lr": 9.152288488020853e-05}
{"train_loss": 0.3149450421333313, "global_step": 51460, "epoch": 578, "lr": 9.152256191676902e-05}
{"train_loss": 0.3538469970226288, "global_step": 51461, "epoch": 578, "lr": 9.15222389477473e-05}
{"train_loss": 0.2555612623691559, "global_step": 51462, "epoch": 578, "lr": 9.152191597314344e-05}
{"train_loss": 0.3048410415649414, "global_step": 51463, "epoch": 578, "lr": 9.152159299295744e-05}
{"train_loss": 0.2837923467159271, "global_step": 51464, "epoch": 578, "lr": 9.152127000718937e-05}
{"train_loss": 0.25809210538864136, "global_step": 51465, "epoch": 578, "lr": 9.15209470158393e-05}
{"train_loss": 0.37403491139411926, "global_step": 51466, "epoch": 578, "lr": 9.152062401890722e-05}
{"train_loss": 0.36695003509521484, "global_step": 51467, "epoch": 578, "lr": 9.152030101639321e-05}
{"train_loss": 0.335165411233902, "global_step": 51468, "epoch": 578, "lr": 9.151997800829729e-05}
{"train_loss": 0.2048386037349701, "global_step": 51469, "epoch": 578, "lr": 9.151965499461953e-05}
{"train_loss": 0.3906131982803345, "global_step": 51470, "epoch": 578, "lr": 9.151933197535994e-05}
{"train_loss": 0.27127712965011597, "global_step": 51471, "epoch": 578, "lr": 9.151900895051859e-05}
{"train_loss": 0.22542224824428558, "global_step": 51472, "epoch": 578, "lr": 9.151868592009552e-05}
{"train_loss": 0.261503130197525, "global_step": 51473, "epoch": 578, "lr": 9.151836288409076e-05}
{"train_loss": 0.2525070309638977, "global_step": 51474, "epoch": 578, "lr": 9.151803984250438e-05}
{"train_loss": 0.1837015450000763, "global_step": 51475, "epoch": 578, "lr": 9.151771679533638e-05}
{"train_loss": 0.3374840021133423, "global_step": 51476, "epoch": 578, "lr": 9.151739374258686e-05}
{"train_loss": 0.2162238210439682, "global_step": 51477, "epoch": 578, "lr": 9.15170706842558e-05}
{"train_loss": 0.2520127296447754, "global_step": 51478, "epoch": 578, "lr": 9.151674762034329e-05}
{"train_loss": 0.3861410915851593, "global_step": 51479, "epoch": 578, "lr": 9.151642455084936e-05}
{"train_loss": 0.2526969313621521, "global_step": 51480, "epoch": 578, "lr": 9.151610147577406e-05}
{"train_loss": 0.2472466379404068, "global_step": 51481, "epoch": 578, "lr": 9.151577839511741e-05}
{"train_loss": 0.26265543699264526, "global_step": 51482, "epoch": 578, "lr": 9.151545530887949e-05}
{"train_loss": 0.23746366798877716, "global_step": 51483, "epoch": 578, "lr": 9.151513221706029e-05}
{"train_loss": 0.2075025588274002, "global_step": 51484, "epoch": 578, "lr": 9.151480911965992e-05}
{"train_loss": 0.3053704500198364, "global_step": 51485, "epoch": 578, "lr": 9.151448601667838e-05}
{"train_loss": 0.2803116738796234, "global_step": 51486, "epoch": 578, "lr": 9.151416290811571e-05}
{"train_loss": 0.3168763816356659, "global_step": 51487, "epoch": 578, "lr": 9.1513839793972e-05}
{"train_loss": 0.2706015706062317, "global_step": 51488, "epoch": 578, "lr": 9.151351667424723e-05}
{"train_loss": 0.24080199003219604, "global_step": 51489, "epoch": 578, "lr": 9.151319354894148e-05}
{"train_loss": 0.2499839961528778, "global_step": 51490, "epoch": 578, "lr": 9.151287041805479e-05}
{"train_loss": 0.1828368902206421, "global_step": 51491, "epoch": 578, "lr": 9.15125472815872e-05}
{"train_loss": 0.1518266797065735, "global_step": 51492, "epoch": 578, "lr": 9.151222413953874e-05}
{"train_loss": 0.2188492864370346, "global_step": 51493, "epoch": 578, "lr": 9.151190099190948e-05}
{"train_loss": 0.31723901629447937, "global_step": 51494, "epoch": 578, "lr": 9.151157783869945e-05}
{"train_loss": 0.22733207046985626, "global_step": 51495, "epoch": 578, "lr": 9.151125467990871e-05}
{"train_loss": 0.3167014718055725, "global_step": 51496, "epoch": 578, "lr": 9.151093151553727e-05}
{"train_loss": 0.3383464217185974, "global_step": 51497, "epoch": 578, "lr": 9.15106083455852e-05}
{"train_loss": 0.29148244857788086, "global_step": 51498, "epoch": 578, "lr": 9.151028517005251e-05}
{"train_loss": 0.20597022771835327, "global_step": 51499, "epoch": 578, "lr": 9.15099619889393e-05}
{"train_loss": 0.4107990562915802, "global_step": 51500, "epoch": 578, "lr": 9.150963880224556e-05}
{"train_loss": 0.2549841105937958, "global_step": 51501, "epoch": 578, "lr": 9.150931560997137e-05}
{"train_loss": 0.2416767179965973, "global_step": 51502, "epoch": 578, "lr": 9.150899241211675e-05}
{"train_loss": 0.2608565390110016, "global_step": 51503, "epoch": 578, "lr": 9.150866920868175e-05}
{"train_loss": 0.24729660153388977, "global_step": 51504, "epoch": 578, "lr": 9.150834599966642e-05}
{"train_loss": 0.2903044521808624, "global_step": 51505, "epoch": 578, "lr": 9.15080227850708e-05}
{"train_loss": 0.2563799321651459, "global_step": 51506, "epoch": 578, "lr": 9.150769956489493e-05}
{"train_loss": 0.38998138904571533, "global_step": 51507, "epoch": 578, "lr": 9.150737633913886e-05}
{"train_loss": 0.27776259183883667, "global_step": 51508, "epoch": 578, "lr": 9.150705310780264e-05}
{"train_loss": 0.14969316124916077, "global_step": 51509, "epoch": 578, "lr": 9.150672987088628e-05}
{"train_loss": 0.31774699687957764, "global_step": 51510, "epoch": 578, "lr": 9.150640662838986e-05}
{"train_loss": 0.26814645528793335, "global_step": 51511, "epoch": 578, "lr": 9.150608338031338e-05}
{"train_loss": 0.24590075016021729, "global_step": 51512, "epoch": 578, "lr": 9.150576012665694e-05}
{"train_loss": 0.18502061069011688, "global_step": 51513, "epoch": 578, "lr": 9.150543686742057e-05}
{"train_loss": 0.28788307309150696, "global_step": 51514, "epoch": 578, "lr": 9.150511360260428e-05}
{"train_loss": 0.23246875405311584, "global_step": 51515, "epoch": 578, "lr": 9.150479033220814e-05}
{"train_loss": 0.2065798044204712, "global_step": 51516, "epoch": 578, "lr": 9.150446705623217e-05}
{"train_loss": 0.31319159269332886, "global_step": 51517, "epoch": 578, "lr": 9.150414377467645e-05}
{"train_loss": 0.3336442708969116, "global_step": 51518, "epoch": 578, "lr": 9.150382048754101e-05}
{"train_loss": 0.19725735485553741, "global_step": 51519, "epoch": 578, "lr": 9.150349719482586e-05}
{"train_loss": 0.3137447237968445, "global_step": 51520, "epoch": 578, "lr": 9.15031738965311e-05}
{"train_loss": 0.22419396042823792, "global_step": 51521, "epoch": 578, "lr": 9.150285059265671e-05}
{"train_loss": 0.30686572194099426, "global_step": 51522, "epoch": 578, "lr": 9.150252728320278e-05}
{"train_loss": 0.3055499792098999, "global_step": 51523, "epoch": 578, "lr": 9.150220396816936e-05}
{"train_loss": 0.23015694320201874, "global_step": 51524, "epoch": 578, "lr": 9.150188064755647e-05}
{"train_loss": 0.28201520442962646, "global_step": 51525, "epoch": 578, "lr": 9.150155732136414e-05}
{"train_loss": 0.20137083530426025, "global_step": 51526, "epoch": 578, "lr": 9.150123398959245e-05}
{"train_loss": 0.3329803943634033, "global_step": 51527, "epoch": 578, "lr": 9.150091065224141e-05}
{"train_loss": 0.24524636566638947, "global_step": 51528, "epoch": 578, "lr": 9.150058730931108e-05}
{"train_loss": 0.28019601106643677, "global_step": 51529, "epoch": 578, "lr": 9.150026396080151e-05}
{"train_loss": 0.2721554305446282, "global_step": 51530, "epoch": 578, "lr": 9.149994060671274e-05, "val_loss": 2.997612237930298}
{"train_loss": 0.2504550814628601, "global_step": 51531, "epoch": 579, "lr": 9.14996172470448e-05}
{"train_loss": 0.29208964109420776, "global_step": 51532, "epoch": 579, "lr": 9.149929388179775e-05}
{"train_loss": 0.24031007289886475, "global_step": 51533, "epoch": 579, "lr": 9.14989705109716e-05}
{"train_loss": 0.2541393041610718, "global_step": 51534, "epoch": 579, "lr": 9.149864713456644e-05}
{"train_loss": 0.2763873338699341, "global_step": 51535, "epoch": 579, "lr": 9.149832375258229e-05}
{"train_loss": 0.22722205519676208, "global_step": 51536, "epoch": 579, "lr": 9.149800036501921e-05}
{"train_loss": 0.2773497998714447, "global_step": 51537, "epoch": 579, "lr": 9.14976769718772e-05}
{"train_loss": 0.23102129995822906, "global_step": 51538, "epoch": 579, "lr": 9.149735357315636e-05}
{"train_loss": 0.20264217257499695, "global_step": 51539, "epoch": 579, "lr": 9.14970301688567e-05}
{"train_loss": 0.21223150193691254, "global_step": 51540, "epoch": 579, "lr": 9.149670675897827e-05}
{"train_loss": 0.32065707445144653, "global_step": 51541, "epoch": 579, "lr": 9.14963833435211e-05}
{"train_loss": 0.15792721509933472, "global_step": 51542, "epoch": 579, "lr": 9.149605992248526e-05}
{"train_loss": 0.3565267324447632, "global_step": 51543, "epoch": 579, "lr": 9.149573649587078e-05}
{"train_loss": 0.22633133828639984, "global_step": 51544, "epoch": 579, "lr": 9.149541306367771e-05}
{"train_loss": 0.2453676462173462, "global_step": 51545, "epoch": 579, "lr": 9.149508962590607e-05}
{"train_loss": 0.2584879696369171, "global_step": 51546, "epoch": 579, "lr": 9.149476618255593e-05}
{"train_loss": 0.21742530167102814, "global_step": 51547, "epoch": 579, "lr": 9.149444273362732e-05}
{"train_loss": 0.27543017268180847, "global_step": 51548, "epoch": 579, "lr": 9.14941192791203e-05}
{"train_loss": 0.16954535245895386, "global_step": 51549, "epoch": 579, "lr": 9.14937958190349e-05}
{"train_loss": 0.257301390171051, "global_step": 51550, "epoch": 579, "lr": 9.149347235337115e-05}
{"train_loss": 0.2896546423435211, "global_step": 51551, "epoch": 579, "lr": 9.149314888212912e-05}
{"train_loss": 0.2677273452281952, "global_step": 51552, "epoch": 579, "lr": 9.149282540530884e-05}
{"train_loss": 0.2656354308128357, "global_step": 51553, "epoch": 579, "lr": 9.149250192291035e-05}
{"train_loss": 0.1784256100654602, "global_step": 51554, "epoch": 579, "lr": 9.149217843493371e-05}
{"train_loss": 0.13651862740516663, "global_step": 51555, "epoch": 579, "lr": 9.149185494137894e-05}
{"train_loss": 0.29777202010154724, "global_step": 51556, "epoch": 579, "lr": 9.149153144224611e-05}
{"train_loss": 0.203687846660614, "global_step": 51557, "epoch": 579, "lr": 9.149120793753523e-05}
{"train_loss": 0.22529737651348114, "global_step": 51558, "epoch": 579, "lr": 9.149088442724638e-05}
{"train_loss": 0.22091327607631683, "global_step": 51559, "epoch": 579, "lr": 9.149056091137958e-05}
{"train_loss": 0.23258748650550842, "global_step": 51560, "epoch": 579, "lr": 9.149023738993488e-05}
{"train_loss": 0.2415601909160614, "global_step": 51561, "epoch": 579, "lr": 9.148991386291232e-05}
{"train_loss": 0.29695725440979004, "global_step": 51562, "epoch": 579, "lr": 9.148959033031194e-05}
{"train_loss": 0.34881383180618286, "global_step": 51563, "epoch": 579, "lr": 9.148926679213381e-05}
{"train_loss": 0.2620962858200073, "global_step": 51564, "epoch": 579, "lr": 9.148894324837794e-05}
{"train_loss": 0.28616613149642944, "global_step": 51565, "epoch": 579, "lr": 9.148861969904439e-05}
{"train_loss": 0.22521686553955078, "global_step": 51566, "epoch": 579, "lr": 9.148829614413321e-05}
{"train_loss": 0.21287687122821808, "global_step": 51567, "epoch": 579, "lr": 9.148797258364443e-05}
{"train_loss": 0.2798777222633362, "global_step": 51568, "epoch": 579, "lr": 9.148764901757808e-05}
{"train_loss": 0.2775443196296692, "global_step": 51569, "epoch": 579, "lr": 9.148732544593424e-05}
{"train_loss": 0.26663538813591003, "global_step": 51570, "epoch": 579, "lr": 9.148700186871293e-05}
{"train_loss": 0.3455834984779358, "global_step": 51571, "epoch": 579, "lr": 9.14866782859142e-05}
{"train_loss": 0.21338705718517303, "global_step": 51572, "epoch": 579, "lr": 9.148635469753809e-05}
{"train_loss": 0.2340182363986969, "global_step": 51573, "epoch": 579, "lr": 9.148603110358464e-05}
{"train_loss": 0.3080626428127289, "global_step": 51574, "epoch": 579, "lr": 9.148570750405391e-05}
{"train_loss": 0.18835760653018951, "global_step": 51575, "epoch": 579, "lr": 9.148538389894593e-05}
{"train_loss": 0.2309110313653946, "global_step": 51576, "epoch": 579, "lr": 9.148506028826073e-05}
{"train_loss": 0.246607705950737, "global_step": 51577, "epoch": 579, "lr": 9.148473667199838e-05}
{"train_loss": 0.3547433018684387, "global_step": 51578, "epoch": 579, "lr": 9.148441305015892e-05}
{"train_loss": 0.24870531260967255, "global_step": 51579, "epoch": 579, "lr": 9.148408942274238e-05}
{"train_loss": 0.22318311035633087, "global_step": 51580, "epoch": 579, "lr": 9.148376578974881e-05}
{"train_loss": 0.18004079163074493, "global_step": 51581, "epoch": 579, "lr": 9.148344215117825e-05}
{"train_loss": 0.2766263782978058, "global_step": 51582, "epoch": 579, "lr": 9.148311850703075e-05}
{"train_loss": 0.28314009308815, "global_step": 51583, "epoch": 579, "lr": 9.148279485730635e-05}
{"train_loss": 0.170759379863739, "global_step": 51584, "epoch": 579, "lr": 9.148247120200511e-05}
{"train_loss": 0.1828770488500595, "global_step": 51585, "epoch": 579, "lr": 9.148214754112702e-05}
{"train_loss": 0.2464888095855713, "global_step": 51586, "epoch": 579, "lr": 9.148182387467219e-05}
{"train_loss": 0.22134824097156525, "global_step": 51587, "epoch": 579, "lr": 9.148150020264062e-05}
{"train_loss": 0.2837207019329071, "global_step": 51588, "epoch": 579, "lr": 9.148117652503238e-05}
{"train_loss": 0.23826175928115845, "global_step": 51589, "epoch": 579, "lr": 9.14808528418475e-05}
{"train_loss": 0.3566994071006775, "global_step": 51590, "epoch": 579, "lr": 9.148052915308603e-05}
{"train_loss": 0.2782224714756012, "global_step": 51591, "epoch": 579, "lr": 9.148020545874798e-05}
{"train_loss": 0.266478031873703, "global_step": 51592, "epoch": 579, "lr": 9.147988175883345e-05}
{"train_loss": 0.40601691603660583, "global_step": 51593, "epoch": 579, "lr": 9.147955805334245e-05}
{"train_loss": 0.26462191343307495, "global_step": 51594, "epoch": 579, "lr": 9.147923434227502e-05}
{"train_loss": 0.26134705543518066, "global_step": 51595, "epoch": 579, "lr": 9.147891062563122e-05}
{"train_loss": 0.47160571813583374, "global_step": 51596, "epoch": 579, "lr": 9.147858690341109e-05}
{"train_loss": 0.22844524681568146, "global_step": 51597, "epoch": 579, "lr": 9.147826317561465e-05}
{"train_loss": 0.2315637618303299, "global_step": 51598, "epoch": 579, "lr": 9.1477939442242e-05}
{"train_loss": 0.22322651743888855, "global_step": 51599, "epoch": 579, "lr": 9.147761570329312e-05}
{"train_loss": 0.26978394389152527, "global_step": 51600, "epoch": 579, "lr": 9.147729195876808e-05}
{"train_loss": 0.2626652121543884, "global_step": 51601, "epoch": 579, "lr": 9.147696820866692e-05}
{"train_loss": 0.26954755187034607, "global_step": 51602, "epoch": 579, "lr": 9.147664445298969e-05}
{"train_loss": 0.2837274670600891, "global_step": 51603, "epoch": 579, "lr": 9.147632069173643e-05}
{"train_loss": 0.28077608346939087, "global_step": 51604, "epoch": 579, "lr": 9.14759969249072e-05}
{"train_loss": 0.25286951661109924, "global_step": 51605, "epoch": 579, "lr": 9.1475673152502e-05}
{"train_loss": 0.21723490953445435, "global_step": 51606, "epoch": 579, "lr": 9.147534937452093e-05}
{"train_loss": 0.2951321005821228, "global_step": 51607, "epoch": 579, "lr": 9.147502559096399e-05}
{"train_loss": 0.3023337423801422, "global_step": 51608, "epoch": 579, "lr": 9.147470180183125e-05}
{"train_loss": 0.5113040804862976, "global_step": 51609, "epoch": 579, "lr": 9.147437800712273e-05}
{"train_loss": 0.3772042691707611, "global_step": 51610, "epoch": 579, "lr": 9.147405420683849e-05}
{"train_loss": 0.3317655026912689, "global_step": 51611, "epoch": 579, "lr": 9.147373040097857e-05}
{"train_loss": 0.24161244928836823, "global_step": 51612, "epoch": 579, "lr": 9.147340658954301e-05}
{"train_loss": 0.262166291475296, "global_step": 51613, "epoch": 579, "lr": 9.147308277253185e-05}
{"train_loss": 0.18124356865882874, "global_step": 51614, "epoch": 579, "lr": 9.147275894994515e-05}
{"train_loss": 0.26945364475250244, "global_step": 51615, "epoch": 579, "lr": 9.147243512178295e-05}
{"train_loss": 0.28361523151397705, "global_step": 51616, "epoch": 579, "lr": 9.147211128804526e-05}
{"train_loss": 0.25535741448402405, "global_step": 51617, "epoch": 579, "lr": 9.147178744873218e-05}
{"train_loss": 0.2792646884918213, "global_step": 51618, "epoch": 579, "lr": 9.14714636038437e-05}
{"train_loss": 0.26188934066992126, "global_step": 51619, "epoch": 579, "lr": 9.147113975337988e-05, "val_loss": 2.9367916584014893}
{"train_loss": 0.17008887231349945, "global_step": 51620, "epoch": 580, "lr": 9.14708158973408e-05}
{"train_loss": 0.23294849693775177, "global_step": 51621, "epoch": 580, "lr": 9.147049203572646e-05}
{"train_loss": 0.2467191070318222, "global_step": 51622, "epoch": 580, "lr": 9.147016816853692e-05}
{"train_loss": 0.29469916224479675, "global_step": 51623, "epoch": 580, "lr": 9.146984429577221e-05}
{"train_loss": 0.27631595730781555, "global_step": 51624, "epoch": 580, "lr": 9.14695204174324e-05}
{"train_loss": 0.3526204824447632, "global_step": 51625, "epoch": 580, "lr": 9.146919653351752e-05}
{"train_loss": 0.2600824534893036, "global_step": 51626, "epoch": 580, "lr": 9.146887264402761e-05}
{"train_loss": 0.24428698420524597, "global_step": 51627, "epoch": 580, "lr": 9.146854874896272e-05}
{"train_loss": 0.22853362560272217, "global_step": 51628, "epoch": 580, "lr": 9.146822484832288e-05}
{"train_loss": 0.29191625118255615, "global_step": 51629, "epoch": 580, "lr": 9.146790094210813e-05}
{"train_loss": 0.32625144720077515, "global_step": 51630, "epoch": 580, "lr": 9.146757703031854e-05}
{"train_loss": 0.2530052065849304, "global_step": 51631, "epoch": 580, "lr": 9.146725311295414e-05}
{"train_loss": 0.3519861400127411, "global_step": 51632, "epoch": 580, "lr": 9.146692919001496e-05}
{"train_loss": 0.20301620662212372, "global_step": 51633, "epoch": 580, "lr": 9.146660526150107e-05}
{"train_loss": 0.1642364263534546, "global_step": 51634, "epoch": 580, "lr": 9.14662813274125e-05}
{"train_loss": 0.2297961711883545, "global_step": 51635, "epoch": 580, "lr": 9.14659573877493e-05}
{"train_loss": 0.2255931794643402, "global_step": 51636, "epoch": 580, "lr": 9.146563344251149e-05}
{"train_loss": 0.3137201964855194, "global_step": 51637, "epoch": 580, "lr": 9.146530949169915e-05}
{"train_loss": 0.2923470735549927, "global_step": 51638, "epoch": 580, "lr": 9.146498553531228e-05}
{"train_loss": 0.23064804077148438, "global_step": 51639, "epoch": 580, "lr": 9.146466157335098e-05}
{"train_loss": 0.30375170707702637, "global_step": 51640, "epoch": 580, "lr": 9.146433760581524e-05}
{"train_loss": 0.31325995922088623, "global_step": 51641, "epoch": 580, "lr": 9.146401363270513e-05}
{"train_loss": 0.26432642340660095, "global_step": 51642, "epoch": 580, "lr": 9.146368965402069e-05}
{"train_loss": 0.26009443402290344, "global_step": 51643, "epoch": 580, "lr": 9.146336566976196e-05}
{"train_loss": 0.3345186114311218, "global_step": 51644, "epoch": 580, "lr": 9.146304167992899e-05}
{"train_loss": 0.23672401905059814, "global_step": 51645, "epoch": 580, "lr": 9.146271768452181e-05}
{"train_loss": 0.2963544726371765, "global_step": 51646, "epoch": 580, "lr": 9.146239368354048e-05}
{"train_loss": 0.3083169162273407, "global_step": 51647, "epoch": 580, "lr": 9.146206967698505e-05}
{"train_loss": 0.33723223209381104, "global_step": 51648, "epoch": 580, "lr": 9.146174566485553e-05}
{"train_loss": 0.3293696641921997, "global_step": 51649, "epoch": 580, "lr": 9.146142164715199e-05}
{"train_loss": 0.38403093814849854, "global_step": 51650, "epoch": 580, "lr": 9.146109762387448e-05}
{"train_loss": 0.2936638295650482, "global_step": 51651, "epoch": 580, "lr": 9.146077359502301e-05}
{"train_loss": 0.20403359830379486, "global_step": 51652, "epoch": 580, "lr": 9.146044956059765e-05}
{"train_loss": 0.2629762887954712, "global_step": 51653, "epoch": 580, "lr": 9.146012552059844e-05}
{"train_loss": 0.28082945942878723, "global_step": 51654, "epoch": 580, "lr": 9.145980147502542e-05}
{"train_loss": 0.18351958692073822, "global_step": 51655, "epoch": 580, "lr": 9.145947742387864e-05}
{"train_loss": 0.26686638593673706, "global_step": 51656, "epoch": 580, "lr": 9.145915336715813e-05}
{"train_loss": 0.2703303396701813, "global_step": 51657, "epoch": 580, "lr": 9.145882930486395e-05}
{"train_loss": 0.25687944889068604, "global_step": 51658, "epoch": 580, "lr": 9.145850523699612e-05}
{"train_loss": 0.21904630959033966, "global_step": 51659, "epoch": 580, "lr": 9.145818116355473e-05}
{"train_loss": 0.19852116703987122, "global_step": 51660, "epoch": 580, "lr": 9.145785708453976e-05}
{"train_loss": 0.32636576890945435, "global_step": 51661, "epoch": 580, "lr": 9.14575329999513e-05}
{"train_loss": 0.2147221714258194, "global_step": 51662, "epoch": 580, "lr": 9.14572089097894e-05}
{"train_loss": 0.27182766795158386, "global_step": 51663, "epoch": 580, "lr": 9.145688481405406e-05}
{"train_loss": 0.24511811137199402, "global_step": 51664, "epoch": 580, "lr": 9.145656071274536e-05}
{"train_loss": 0.28338995575904846, "global_step": 51665, "epoch": 580, "lr": 9.14562366058633e-05}
{"train_loss": 0.21799306571483612, "global_step": 51666, "epoch": 580, "lr": 9.1455912493408e-05}
{"train_loss": 0.20425842702388763, "global_step": 51667, "epoch": 580, "lr": 9.145558837537942e-05}
{"train_loss": 0.3079919219017029, "global_step": 51668, "epoch": 580, "lr": 9.145526425177766e-05}
{"train_loss": 0.22717177867889404, "global_step": 51669, "epoch": 580, "lr": 9.145494012260274e-05}
{"train_loss": 0.17511123418807983, "global_step": 51670, "epoch": 580, "lr": 9.145461598785471e-05}
{"train_loss": 0.32402747869491577, "global_step": 51671, "epoch": 580, "lr": 9.145429184753362e-05}
{"train_loss": 0.28488054871559143, "global_step": 51672, "epoch": 580, "lr": 9.14539677016395e-05}
{"train_loss": 0.2109595537185669, "global_step": 51673, "epoch": 580, "lr": 9.145364355017239e-05}
{"train_loss": 0.2363957315683365, "global_step": 51674, "epoch": 580, "lr": 9.145331939313235e-05}
{"train_loss": 0.2110758125782013, "global_step": 51675, "epoch": 580, "lr": 9.145299523051942e-05}
{"train_loss": 0.3024030029773712, "global_step": 51676, "epoch": 580, "lr": 9.145267106233364e-05}
{"train_loss": 0.19806823134422302, "global_step": 51677, "epoch": 580, "lr": 9.145234688857505e-05}
{"train_loss": 0.3582156300544739, "global_step": 51678, "epoch": 580, "lr": 9.145202270924368e-05}
{"train_loss": 0.2756354808807373, "global_step": 51679, "epoch": 580, "lr": 9.145169852433961e-05}
{"train_loss": 0.23920786380767822, "global_step": 51680, "epoch": 580, "lr": 9.145137433386287e-05}
{"train_loss": 0.27686476707458496, "global_step": 51681, "epoch": 580, "lr": 9.145105013781349e-05}
{"train_loss": 0.2621954679489136, "global_step": 51682, "epoch": 580, "lr": 9.145072593619153e-05}
{"train_loss": 0.3423364460468292, "global_step": 51683, "epoch": 580, "lr": 9.145040172899702e-05}
{"train_loss": 0.27168095111846924, "global_step": 51684, "epoch": 580, "lr": 9.145007751623e-05}
{"train_loss": 0.2946118414402008, "global_step": 51685, "epoch": 580, "lr": 9.144975329789052e-05}
{"train_loss": 0.38571321964263916, "global_step": 51686, "epoch": 580, "lr": 9.144942907397864e-05}
{"train_loss": 0.2913544774055481, "global_step": 51687, "epoch": 580, "lr": 9.144910484449439e-05}
{"train_loss": 0.2656550705432892, "global_step": 51688, "epoch": 580, "lr": 9.14487806094378e-05}
{"train_loss": 0.24719804525375366, "global_step": 51689, "epoch": 580, "lr": 9.144845636880894e-05}
{"train_loss": 0.27569693326950073, "global_step": 51690, "epoch": 580, "lr": 9.144813212260783e-05}
{"train_loss": 0.36521705985069275, "global_step": 51691, "epoch": 580, "lr": 9.144780787083453e-05}
{"train_loss": 0.253525972366333, "global_step": 51692, "epoch": 580, "lr": 9.144748361348907e-05}
{"train_loss": 0.2041279375553131, "global_step": 51693, "epoch": 580, "lr": 9.14471593505715e-05}
{"train_loss": 0.21854810416698456, "global_step": 51694, "epoch": 580, "lr": 9.144683508208187e-05}
{"train_loss": 0.3049828112125397, "global_step": 51695, "epoch": 580, "lr": 9.144651080802022e-05}
{"train_loss": 0.2450360357761383, "global_step": 51696, "epoch": 580, "lr": 9.144618652838659e-05}
{"train_loss": 0.2181720733642578, "global_step": 51697, "epoch": 580, "lr": 9.144586224318102e-05}
{"train_loss": 0.2181413769721985, "global_step": 51698, "epoch": 580, "lr": 9.144553795240357e-05}
{"train_loss": 0.2302389144897461, "global_step": 51699, "epoch": 580, "lr": 9.144521365605428e-05}
{"train_loss": 0.2529107928276062, "global_step": 51700, "epoch": 580, "lr": 9.144488935413316e-05}
{"train_loss": 0.23523354530334473, "global_step": 51701, "epoch": 580, "lr": 9.14445650466403e-05}
{"train_loss": 0.27249130606651306, "global_step": 51702, "epoch": 580, "lr": 9.144424073357572e-05}
{"train_loss": 0.2840341329574585, "global_step": 51703, "epoch": 580, "lr": 9.144391641493946e-05}
{"train_loss": 0.19740447402000427, "global_step": 51704, "epoch": 580, "lr": 9.144359209073157e-05}
{"train_loss": 0.3443428575992584, "global_step": 51705, "epoch": 580, "lr": 9.14432677609521e-05}
{"train_loss": 0.2137017697095871, "global_step": 51706, "epoch": 580, "lr": 9.144294342560108e-05}
{"train_loss": 0.34549134969711304, "global_step": 51707, "epoch": 580, "lr": 9.144261908467858e-05}
{"train_loss": 0.2661285274819042, "global_step": 51708, "epoch": 580, "lr": 9.144229473818459e-05, "val_loss": 3.1533868312835693, "train_action_mse_error": 37.20065689086914}
{"train_loss": 0.266186386346817, "global_step": 51709, "epoch": 581, "lr": 9.144197038611922e-05}
{"train_loss": 0.2957313060760498, "global_step": 51710, "epoch": 581, "lr": 9.144164602848248e-05}
{"train_loss": 0.15565133094787598, "global_step": 51711, "epoch": 581, "lr": 9.144132166527441e-05}
{"train_loss": 0.31634363532066345, "global_step": 51712, "epoch": 581, "lr": 9.144099729649504e-05}
{"train_loss": 0.22762905061244965, "global_step": 51713, "epoch": 581, "lr": 9.144067292214447e-05}
{"train_loss": 0.17799733579158783, "global_step": 51714, "epoch": 581, "lr": 9.144034854222268e-05}
{"train_loss": 0.31789395213127136, "global_step": 51715, "epoch": 581, "lr": 9.144002415672974e-05}
{"train_loss": 0.3207775354385376, "global_step": 51716, "epoch": 581, "lr": 9.14396997656657e-05}
{"train_loss": 0.20623931288719177, "global_step": 51717, "epoch": 581, "lr": 9.143937536903061e-05}
{"train_loss": 0.3597298562526703, "global_step": 51718, "epoch": 581, "lr": 9.143905096682449e-05}
{"train_loss": 0.2979789674282074, "global_step": 51719, "epoch": 581, "lr": 9.14387265590474e-05}
{"train_loss": 0.23413591086864471, "global_step": 51720, "epoch": 581, "lr": 9.143840214569936e-05}
{"train_loss": 0.2046556919813156, "global_step": 51721, "epoch": 581, "lr": 9.143807772678046e-05}
{"train_loss": 0.18212154507637024, "global_step": 51722, "epoch": 581, "lr": 9.14377533022907e-05}
{"train_loss": 0.28453874588012695, "global_step": 51723, "epoch": 581, "lr": 9.143742887223014e-05}
{"train_loss": 0.17942747473716736, "global_step": 51724, "epoch": 581, "lr": 9.143710443659881e-05}
{"train_loss": 0.25480690598487854, "global_step": 51725, "epoch": 581, "lr": 9.143677999539679e-05}
{"train_loss": 0.26156866550445557, "global_step": 51726, "epoch": 581, "lr": 9.143645554862407e-05}
{"train_loss": 0.3227633535861969, "global_step": 51727, "epoch": 581, "lr": 9.143613109628075e-05}
{"train_loss": 0.3319080173969269, "global_step": 51728, "epoch": 581, "lr": 9.143580663836684e-05}
{"train_loss": 0.2386375516653061, "global_step": 51729, "epoch": 581, "lr": 9.14354821748824e-05}
{"train_loss": 0.20173807442188263, "global_step": 51730, "epoch": 581, "lr": 9.143515770582745e-05}
{"train_loss": 0.2939719557762146, "global_step": 51731, "epoch": 581, "lr": 9.143483323120206e-05}
{"train_loss": 0.24938178062438965, "global_step": 51732, "epoch": 581, "lr": 9.143450875100625e-05}
{"train_loss": 0.2883012294769287, "global_step": 51733, "epoch": 581, "lr": 9.143418426524009e-05}
{"train_loss": 0.1877094954252243, "global_step": 51734, "epoch": 581, "lr": 9.143385977390359e-05}
{"train_loss": 0.2653483748435974, "global_step": 51735, "epoch": 581, "lr": 9.143353527699682e-05}
{"train_loss": 0.2022351324558258, "global_step": 51736, "epoch": 581, "lr": 9.143321077451982e-05}
{"train_loss": 0.27924346923828125, "global_step": 51737, "epoch": 581, "lr": 9.143288626647262e-05}
{"train_loss": 0.22631196677684784, "global_step": 51738, "epoch": 581, "lr": 9.143256175285528e-05}
{"train_loss": 0.2338099628686905, "global_step": 51739, "epoch": 581, "lr": 9.143223723366785e-05}
{"train_loss": 0.2577131390571594, "global_step": 51740, "epoch": 581, "lr": 9.143191270891033e-05}
{"train_loss": 0.24586784839630127, "global_step": 51741, "epoch": 581, "lr": 9.143158817858282e-05}
{"train_loss": 0.2625548243522644, "global_step": 51742, "epoch": 581, "lr": 9.143126364268533e-05}
{"train_loss": 0.21951718628406525, "global_step": 51743, "epoch": 581, "lr": 9.143093910121792e-05}
{"train_loss": 0.24100551009178162, "global_step": 51744, "epoch": 581, "lr": 9.143061455418061e-05}
{"train_loss": 0.1835877001285553, "global_step": 51745, "epoch": 581, "lr": 9.143029000157348e-05}
{"train_loss": 0.28146862983703613, "global_step": 51746, "epoch": 581, "lr": 9.142996544339653e-05}
{"train_loss": 0.2598305642604828, "global_step": 51747, "epoch": 581, "lr": 9.142964087964984e-05}
{"train_loss": 0.1819814145565033, "global_step": 51748, "epoch": 581, "lr": 9.142931631033343e-05}
{"train_loss": 0.2541443407535553, "global_step": 51749, "epoch": 581, "lr": 9.142899173544736e-05}
{"train_loss": 0.21808409690856934, "global_step": 51750, "epoch": 581, "lr": 9.142866715499167e-05}
{"train_loss": 0.24041412770748138, "global_step": 51751, "epoch": 581, "lr": 9.142834256896639e-05}
{"train_loss": 0.3121512532234192, "global_step": 51752, "epoch": 581, "lr": 9.142801797737158e-05}
{"train_loss": 0.24895232915878296, "global_step": 51753, "epoch": 581, "lr": 9.142769338020729e-05}
{"train_loss": 0.2395649403333664, "global_step": 51754, "epoch": 581, "lr": 9.142736877747353e-05}
{"train_loss": 0.3122226893901825, "global_step": 51755, "epoch": 581, "lr": 9.142704416917039e-05}
{"train_loss": 0.28981876373291016, "global_step": 51756, "epoch": 581, "lr": 9.142671955529788e-05}
{"train_loss": 0.19715365767478943, "global_step": 51757, "epoch": 581, "lr": 9.142639493585603e-05}
{"train_loss": 0.24736709892749786, "global_step": 51758, "epoch": 581, "lr": 9.142607031084492e-05}
{"train_loss": 0.1938413381576538, "global_step": 51759, "epoch": 581, "lr": 9.142574568026458e-05}
{"train_loss": 0.306923508644104, "global_step": 51760, "epoch": 581, "lr": 9.142542104411506e-05}
{"train_loss": 0.4144565761089325, "global_step": 51761, "epoch": 581, "lr": 9.142509640239639e-05}
{"train_loss": 0.29784610867500305, "global_step": 51762, "epoch": 581, "lr": 9.142477175510862e-05}
{"train_loss": 0.262702614068985, "global_step": 51763, "epoch": 581, "lr": 9.14244471022518e-05}
{"train_loss": 0.24705727398395538, "global_step": 51764, "epoch": 581, "lr": 9.142412244382596e-05}
{"train_loss": 0.2706603705883026, "global_step": 51765, "epoch": 581, "lr": 9.142379777983116e-05}
{"train_loss": 0.27442485094070435, "global_step": 51766, "epoch": 581, "lr": 9.142347311026744e-05}
{"train_loss": 0.3313775658607483, "global_step": 51767, "epoch": 581, "lr": 9.142314843513482e-05}
{"train_loss": 0.17364028096199036, "global_step": 51768, "epoch": 581, "lr": 9.142282375443338e-05}
{"train_loss": 0.307731032371521, "global_step": 51769, "epoch": 581, "lr": 9.142249906816313e-05}
{"train_loss": 0.22268302738666534, "global_step": 51770, "epoch": 581, "lr": 9.142217437632414e-05}
{"train_loss": 0.2676260769367218, "global_step": 51771, "epoch": 581, "lr": 9.142184967891645e-05}
{"train_loss": 0.23332031071186066, "global_step": 51772, "epoch": 581, "lr": 9.14215249759401e-05}
{"train_loss": 0.38218221068382263, "global_step": 51773, "epoch": 581, "lr": 9.14212002673951e-05}
{"train_loss": 0.2510237693786621, "global_step": 51774, "epoch": 581, "lr": 9.142087555328156e-05}
{"train_loss": 0.30244117975234985, "global_step": 51775, "epoch": 581, "lr": 9.142055083359948e-05}
{"train_loss": 0.24011282622814178, "global_step": 51776, "epoch": 581, "lr": 9.142022610834891e-05}
{"train_loss": 0.2496543824672699, "global_step": 51777, "epoch": 581, "lr": 9.141990137752988e-05}
{"train_loss": 0.2093677669763565, "global_step": 51778, "epoch": 581, "lr": 9.141957664114247e-05}
{"train_loss": 0.24583391845226288, "global_step": 51779, "epoch": 581, "lr": 9.141925189918669e-05}
{"train_loss": 0.2582192122936249, "global_step": 51780, "epoch": 581, "lr": 9.14189271516626e-05}
{"train_loss": 0.34347936511039734, "global_step": 51781, "epoch": 581, "lr": 9.141860239857025e-05}
{"train_loss": 0.2656905949115753, "global_step": 51782, "epoch": 581, "lr": 9.141827763990965e-05}
{"train_loss": 0.22840538620948792, "global_step": 51783, "epoch": 581, "lr": 9.141795287568089e-05}
{"train_loss": 0.2565386891365051, "global_step": 51784, "epoch": 581, "lr": 9.141762810588396e-05}
{"train_loss": 0.32067179679870605, "global_step": 51785, "epoch": 581, "lr": 9.141730333051897e-05}
{"train_loss": 0.2966141104698181, "global_step": 51786, "epoch": 581, "lr": 9.14169785495859e-05}
{"train_loss": 0.27013099193573, "global_step": 51787, "epoch": 581, "lr": 9.141665376308485e-05}
{"train_loss": 0.1853034496307373, "global_step": 51788, "epoch": 581, "lr": 9.141632897101581e-05}
{"train_loss": 0.2285892218351364, "global_step": 51789, "epoch": 581, "lr": 9.141600417337888e-05}
{"train_loss": 0.2019226998090744, "global_step": 51790, "epoch": 581, "lr": 9.141567937017405e-05}
{"train_loss": 0.23550331592559814, "global_step": 51791, "epoch": 581, "lr": 9.14153545614014e-05}
{"train_loss": 0.27764642238616943, "global_step": 51792, "epoch": 581, "lr": 9.141502974706095e-05}
{"train_loss": 0.29419758915901184, "global_step": 51793, "epoch": 581, "lr": 9.141470492715276e-05}
{"train_loss": 0.21762102842330933, "global_step": 51794, "epoch": 581, "lr": 9.141438010167686e-05}
{"train_loss": 0.24011074006557465, "global_step": 51795, "epoch": 581, "lr": 9.141405527063329e-05}
{"train_loss": 0.2988467812538147, "global_step": 51796, "epoch": 581, "lr": 9.141373043402213e-05}
{"train_loss": 0.2576583406228698, "global_step": 51797, "epoch": 581, "lr": 9.14134055918434e-05, "val_loss": 3.1611993312835693}
{"train_loss": 0.23301251232624054, "global_step": 51798, "epoch": 582, "lr": 9.141308074409712e-05}
{"train_loss": 0.21968699991703033, "global_step": 51799, "epoch": 582, "lr": 9.141275589078337e-05}
{"train_loss": 0.23293966054916382, "global_step": 51800, "epoch": 582, "lr": 9.141243103190219e-05}
{"train_loss": 0.23832251131534576, "global_step": 51801, "epoch": 582, "lr": 9.14121061674536e-05}
{"train_loss": 0.19658619165420532, "global_step": 51802, "epoch": 582, "lr": 9.141178129743764e-05}
{"train_loss": 0.2092026323080063, "global_step": 51803, "epoch": 582, "lr": 9.141145642185441e-05}
{"train_loss": 0.16670365631580353, "global_step": 51804, "epoch": 582, "lr": 9.141113154070388e-05}
{"train_loss": 0.2485961616039276, "global_step": 51805, "epoch": 582, "lr": 9.141080665398614e-05}
{"train_loss": 0.2416556179523468, "global_step": 51806, "epoch": 582, "lr": 9.141048176170122e-05}
{"train_loss": 0.1733657419681549, "global_step": 51807, "epoch": 582, "lr": 9.141015686384917e-05}
{"train_loss": 0.2555725872516632, "global_step": 51808, "epoch": 582, "lr": 9.140983196043002e-05}
{"train_loss": 0.3185606002807617, "global_step": 51809, "epoch": 582, "lr": 9.140950705144385e-05}
{"train_loss": 0.2132137566804886, "global_step": 51810, "epoch": 582, "lr": 9.140918213689064e-05}
{"train_loss": 0.2701290249824524, "global_step": 51811, "epoch": 582, "lr": 9.140885721677048e-05}
{"train_loss": 0.21949847042560577, "global_step": 51812, "epoch": 582, "lr": 9.140853229108342e-05}
{"train_loss": 0.21374854445457458, "global_step": 51813, "epoch": 582, "lr": 9.140820735982948e-05}
{"train_loss": 0.27882054448127747, "global_step": 51814, "epoch": 582, "lr": 9.14078824230087e-05}
{"train_loss": 0.2704584300518036, "global_step": 51815, "epoch": 582, "lr": 9.140755748062116e-05}
{"train_loss": 0.23315155506134033, "global_step": 51816, "epoch": 582, "lr": 9.140723253266686e-05}
{"train_loss": 0.2277597188949585, "global_step": 51817, "epoch": 582, "lr": 9.140690757914586e-05}
{"train_loss": 0.2266806662082672, "global_step": 51818, "epoch": 582, "lr": 9.140658262005821e-05}
{"train_loss": 0.3327013850212097, "global_step": 51819, "epoch": 582, "lr": 9.140625765540396e-05}
{"train_loss": 0.2224837839603424, "global_step": 51820, "epoch": 582, "lr": 9.140593268518312e-05}
{"train_loss": 0.31938621401786804, "global_step": 51821, "epoch": 582, "lr": 9.140560770939578e-05}
{"train_loss": 0.25576820969581604, "global_step": 51822, "epoch": 582, "lr": 9.140528272804194e-05}
{"train_loss": 0.24308240413665771, "global_step": 51823, "epoch": 582, "lr": 9.140495774112167e-05}
{"train_loss": 0.22057704627513885, "global_step": 51824, "epoch": 582, "lr": 9.140463274863502e-05}
{"train_loss": 0.3123284578323364, "global_step": 51825, "epoch": 582, "lr": 9.140430775058202e-05}
{"train_loss": 0.24543248116970062, "global_step": 51826, "epoch": 582, "lr": 9.140398274696271e-05}
{"train_loss": 0.20032396912574768, "global_step": 51827, "epoch": 582, "lr": 9.140365773777714e-05}
{"train_loss": 0.2613499164581299, "global_step": 51828, "epoch": 582, "lr": 9.140333272302535e-05}
{"train_loss": 0.193618044257164, "global_step": 51829, "epoch": 582, "lr": 9.14030077027074e-05}
{"train_loss": 0.15539444983005524, "global_step": 51830, "epoch": 582, "lr": 9.14026826768233e-05}
{"train_loss": 0.25928086042404175, "global_step": 51831, "epoch": 582, "lr": 9.140235764537313e-05}
{"train_loss": 0.26488056778907776, "global_step": 51832, "epoch": 582, "lr": 9.140203260835691e-05}
{"train_loss": 0.24681280553340912, "global_step": 51833, "epoch": 582, "lr": 9.14017075657747e-05}
{"train_loss": 0.4125775992870331, "global_step": 51834, "epoch": 582, "lr": 9.140138251762653e-05}
{"train_loss": 0.3634403347969055, "global_step": 51835, "epoch": 582, "lr": 9.140105746391244e-05}
{"train_loss": 0.265586793422699, "global_step": 51836, "epoch": 582, "lr": 9.140073240463249e-05}
{"train_loss": 0.31283000111579895, "global_step": 51837, "epoch": 582, "lr": 9.140040733978672e-05}
{"train_loss": 0.24514302611351013, "global_step": 51838, "epoch": 582, "lr": 9.140008226937516e-05}
{"train_loss": 0.19592539966106415, "global_step": 51839, "epoch": 582, "lr": 9.139975719339788e-05}
{"train_loss": 0.17641285061836243, "global_step": 51840, "epoch": 582, "lr": 9.13994321118549e-05}
{"train_loss": 0.23770031332969666, "global_step": 51841, "epoch": 582, "lr": 9.139910702474625e-05}
{"train_loss": 0.2781675159931183, "global_step": 51842, "epoch": 582, "lr": 9.139878193207202e-05}
{"train_loss": 0.34751641750335693, "global_step": 51843, "epoch": 582, "lr": 9.139845683383222e-05}
{"train_loss": 0.170871302485466, "global_step": 51844, "epoch": 582, "lr": 9.139813173002689e-05}
{"train_loss": 0.26376840472221375, "global_step": 51845, "epoch": 582, "lr": 9.139780662065611e-05}
{"train_loss": 0.24758034944534302, "global_step": 51846, "epoch": 582, "lr": 9.139748150571988e-05}
{"train_loss": 0.3138163685798645, "global_step": 51847, "epoch": 582, "lr": 9.139715638521828e-05}
{"train_loss": 0.25995251536369324, "global_step": 51848, "epoch": 582, "lr": 9.139683125915133e-05}
{"train_loss": 0.23801162838935852, "global_step": 51849, "epoch": 582, "lr": 9.139650612751906e-05}
{"train_loss": 0.35169851779937744, "global_step": 51850, "epoch": 582, "lr": 9.139618099032157e-05}
{"train_loss": 0.2623473107814789, "global_step": 51851, "epoch": 582, "lr": 9.139585584755884e-05}
{"train_loss": 0.2775549590587616, "global_step": 51852, "epoch": 582, "lr": 9.139553069923095e-05}
{"train_loss": 0.26442408561706543, "global_step": 51853, "epoch": 582, "lr": 9.139520554533794e-05}
{"train_loss": 0.2811221182346344, "global_step": 51854, "epoch": 582, "lr": 9.139488038587984e-05}
{"train_loss": 0.2714174687862396, "global_step": 51855, "epoch": 582, "lr": 9.139455522085672e-05}
{"train_loss": 0.2254074215888977, "global_step": 51856, "epoch": 582, "lr": 9.139423005026859e-05}
{"train_loss": 0.25169283151626587, "global_step": 51857, "epoch": 582, "lr": 9.139390487411552e-05}
{"train_loss": 0.26551368832588196, "global_step": 51858, "epoch": 582, "lr": 9.139357969239754e-05}
{"train_loss": 0.1829701066017151, "global_step": 51859, "epoch": 582, "lr": 9.139325450511471e-05}
{"train_loss": 0.24080385267734528, "global_step": 51860, "epoch": 582, "lr": 9.139292931226704e-05}
{"train_loss": 0.2858022153377533, "global_step": 51861, "epoch": 582, "lr": 9.139260411385461e-05}
{"train_loss": 0.43864160776138306, "global_step": 51862, "epoch": 582, "lr": 9.139227890987743e-05}
{"train_loss": 0.33781617879867554, "global_step": 51863, "epoch": 582, "lr": 9.139195370033559e-05}
{"train_loss": 0.27125605940818787, "global_step": 51864, "epoch": 582, "lr": 9.139162848522909e-05}
{"train_loss": 0.343556672334671, "global_step": 51865, "epoch": 582, "lr": 9.139130326455799e-05}
{"train_loss": 0.2952771782875061, "global_step": 51866, "epoch": 582, "lr": 9.139097803832234e-05}
{"train_loss": 0.2297224998474121, "global_step": 51867, "epoch": 582, "lr": 9.139065280652217e-05}
{"train_loss": 0.25775033235549927, "global_step": 51868, "epoch": 582, "lr": 9.139032756915753e-05}
{"train_loss": 0.19430238008499146, "global_step": 51869, "epoch": 582, "lr": 9.139000232622847e-05}
{"train_loss": 0.3256302773952484, "global_step": 51870, "epoch": 582, "lr": 9.138967707773504e-05}
{"train_loss": 0.32369452714920044, "global_step": 51871, "epoch": 582, "lr": 9.138935182367725e-05}
{"train_loss": 0.24325545132160187, "global_step": 51872, "epoch": 582, "lr": 9.138902656405518e-05}
{"train_loss": 0.28290313482284546, "global_step": 51873, "epoch": 582, "lr": 9.138870129886884e-05}
{"train_loss": 0.2812455892562866, "global_step": 51874, "epoch": 582, "lr": 9.138837602811832e-05}
{"train_loss": 0.24117890000343323, "global_step": 51875, "epoch": 582, "lr": 9.138805075180364e-05}
{"train_loss": 0.29683196544647217, "global_step": 51876, "epoch": 582, "lr": 9.138772546992483e-05}
{"train_loss": 0.24753080308437347, "global_step": 51877, "epoch": 582, "lr": 9.138740018248194e-05}
{"train_loss": 0.2531545162200928, "global_step": 51878, "epoch": 582, "lr": 9.138707488947502e-05}
{"train_loss": 0.23818492889404297, "global_step": 51879, "epoch": 582, "lr": 9.138674959090411e-05}
{"train_loss": 0.30759355425834656, "global_step": 51880, "epoch": 582, "lr": 9.138642428676927e-05}
{"train_loss": 0.27279946208000183, "global_step": 51881, "epoch": 582, "lr": 9.138609897707051e-05}
{"train_loss": 0.36081716418266296, "global_step": 51882, "epoch": 582, "lr": 9.138577366180791e-05}
{"train_loss": 0.2232072949409485, "global_step": 51883, "epoch": 582, "lr": 9.138544834098148e-05}
{"train_loss": 0.24832689762115479, "global_step": 51884, "epoch": 582, "lr": 9.13851230145913e-05}
{"train_loss": 0.1812714785337448, "global_step": 51885, "epoch": 582, "lr": 9.13847976826374e-05}
{"train_loss": 0.25944005304507994, "global_step": 51886, "epoch": 582, "lr": 9.13844723451198e-05, "val_loss": 3.2301018238067627}
{"train_loss": 0.3354410231113434, "global_step": 51887, "epoch": 583, "lr": 9.138414700203857e-05}
{"train_loss": 0.3120693266391754, "global_step": 51888, "epoch": 583, "lr": 9.138382165339374e-05}
{"train_loss": 0.26099124550819397, "global_step": 51889, "epoch": 583, "lr": 9.138349629918536e-05}
{"train_loss": 0.2216564565896988, "global_step": 51890, "epoch": 583, "lr": 9.138317093941346e-05}
{"train_loss": 0.35151660442352295, "global_step": 51891, "epoch": 583, "lr": 9.138284557407812e-05}
{"train_loss": 0.20929551124572754, "global_step": 51892, "epoch": 583, "lr": 9.138252020317936e-05}
{"train_loss": 0.27990323305130005, "global_step": 51893, "epoch": 583, "lr": 9.138219482671722e-05}
{"train_loss": 0.43510907888412476, "global_step": 51894, "epoch": 583, "lr": 9.138186944469174e-05}
{"train_loss": 0.28378498554229736, "global_step": 51895, "epoch": 583, "lr": 9.138154405710297e-05}
{"train_loss": 0.31348323822021484, "global_step": 51896, "epoch": 583, "lr": 9.138121866395097e-05}
{"train_loss": 0.2827335000038147, "global_step": 51897, "epoch": 583, "lr": 9.138089326523576e-05}
{"train_loss": 0.30822253227233887, "global_step": 51898, "epoch": 583, "lr": 9.13805678609574e-05}
{"train_loss": 0.24897177517414093, "global_step": 51899, "epoch": 583, "lr": 9.138024245111592e-05}
{"train_loss": 0.2871542274951935, "global_step": 51900, "epoch": 583, "lr": 9.137991703571139e-05}
{"train_loss": 0.2596611976623535, "global_step": 51901, "epoch": 583, "lr": 9.137959161474381e-05}
{"train_loss": 0.29093730449676514, "global_step": 51902, "epoch": 583, "lr": 9.137926618821326e-05}
{"train_loss": 0.2174789309501648, "global_step": 51903, "epoch": 583, "lr": 9.137894075611977e-05}
{"train_loss": 0.345605731010437, "global_step": 51904, "epoch": 583, "lr": 9.137861531846338e-05}
{"train_loss": 0.23784275352954865, "global_step": 51905, "epoch": 583, "lr": 9.137828987524413e-05}
{"train_loss": 0.28561076521873474, "global_step": 51906, "epoch": 583, "lr": 9.137796442646209e-05}
{"train_loss": 0.2470065951347351, "global_step": 51907, "epoch": 583, "lr": 9.137763897211729e-05}
{"train_loss": 0.2835294306278229, "global_step": 51908, "epoch": 583, "lr": 9.137731351220977e-05}
{"train_loss": 0.3506253659725189, "global_step": 51909, "epoch": 583, "lr": 9.137698804673957e-05}
{"train_loss": 0.347728431224823, "global_step": 51910, "epoch": 583, "lr": 9.137666257570674e-05}
{"train_loss": 0.27189183235168457, "global_step": 51911, "epoch": 583, "lr": 9.137633709911131e-05}
{"train_loss": 0.3572412133216858, "global_step": 51912, "epoch": 583, "lr": 9.137601161695334e-05}
{"train_loss": 0.23446397483348846, "global_step": 51913, "epoch": 583, "lr": 9.137568612923288e-05}
{"train_loss": 0.27187851071357727, "global_step": 51914, "epoch": 583, "lr": 9.137536063594998e-05}
{"train_loss": 0.21399198472499847, "global_step": 51915, "epoch": 583, "lr": 9.137503513710464e-05}
{"train_loss": 0.2151152342557907, "global_step": 51916, "epoch": 583, "lr": 9.137470963269693e-05}
{"train_loss": 0.28626012802124023, "global_step": 51917, "epoch": 583, "lr": 9.13743841227269e-05}
{"train_loss": 0.3280147314071655, "global_step": 51918, "epoch": 583, "lr": 9.137405860719459e-05}
{"train_loss": 0.3317948877811432, "global_step": 51919, "epoch": 583, "lr": 9.137373308610004e-05}
{"train_loss": 0.3143661916255951, "global_step": 51920, "epoch": 583, "lr": 9.13734075594433e-05}
{"train_loss": 0.2590250074863434, "global_step": 51921, "epoch": 583, "lr": 9.137308202722442e-05}
{"train_loss": 0.22167940437793732, "global_step": 51922, "epoch": 583, "lr": 9.137275648944341e-05}
{"train_loss": 0.25448736548423767, "global_step": 51923, "epoch": 583, "lr": 9.137243094610036e-05}
{"train_loss": 0.21496054530143738, "global_step": 51924, "epoch": 583, "lr": 9.137210539719526e-05}
{"train_loss": 0.29861339926719666, "global_step": 51925, "epoch": 583, "lr": 9.13717798427282e-05}
{"train_loss": 0.27433207631111145, "global_step": 51926, "epoch": 583, "lr": 9.137145428269922e-05}
{"train_loss": 0.30668172240257263, "global_step": 51927, "epoch": 583, "lr": 9.137112871710835e-05}
{"train_loss": 0.31638863682746887, "global_step": 51928, "epoch": 583, "lr": 9.137080314595563e-05}
{"train_loss": 0.253581702709198, "global_step": 51929, "epoch": 583, "lr": 9.137047756924111e-05}
{"train_loss": 0.217189222574234, "global_step": 51930, "epoch": 583, "lr": 9.137015198696483e-05}
{"train_loss": 0.21208080649375916, "global_step": 51931, "epoch": 583, "lr": 9.136982639912684e-05}
{"train_loss": 0.17758263647556305, "global_step": 51932, "epoch": 583, "lr": 9.136950080572717e-05}
{"train_loss": 0.3348866105079651, "global_step": 51933, "epoch": 583, "lr": 9.136917520676588e-05}
{"train_loss": 0.350212961435318, "global_step": 51934, "epoch": 583, "lr": 9.136884960224303e-05}
{"train_loss": 0.3787842392921448, "global_step": 51935, "epoch": 583, "lr": 9.136852399215863e-05}
{"train_loss": 0.3251833915710449, "global_step": 51936, "epoch": 583, "lr": 9.136819837651274e-05}
{"train_loss": 0.2584388256072998, "global_step": 51937, "epoch": 583, "lr": 9.136787275530538e-05}
{"train_loss": 0.24889826774597168, "global_step": 51938, "epoch": 583, "lr": 9.136754712853663e-05}
{"train_loss": 0.32698914408683777, "global_step": 51939, "epoch": 583, "lr": 9.136722149620652e-05}
{"train_loss": 0.22974611818790436, "global_step": 51940, "epoch": 583, "lr": 9.136689585831508e-05}
{"train_loss": 0.30067166686058044, "global_step": 51941, "epoch": 583, "lr": 9.136657021486237e-05}
{"train_loss": 0.34579944610595703, "global_step": 51942, "epoch": 583, "lr": 9.136624456584844e-05}
{"train_loss": 0.3181458115577698, "global_step": 51943, "epoch": 583, "lr": 9.13659189112733e-05}
{"train_loss": 0.27864333987236023, "global_step": 51944, "epoch": 583, "lr": 9.136559325113702e-05}
{"train_loss": 0.3074788451194763, "global_step": 51945, "epoch": 583, "lr": 9.136526758543966e-05}
{"train_loss": 0.2832080125808716, "global_step": 51946, "epoch": 583, "lr": 9.136494191418123e-05}
{"train_loss": 0.3259667754173279, "global_step": 51947, "epoch": 583, "lr": 9.136461623736179e-05}
{"train_loss": 0.3335328698158264, "global_step": 51948, "epoch": 583, "lr": 9.136429055498139e-05}
{"train_loss": 0.2870767414569855, "global_step": 51949, "epoch": 583, "lr": 9.136396486704005e-05}
{"train_loss": 0.3249857425689697, "global_step": 51950, "epoch": 583, "lr": 9.136363917353784e-05}
{"train_loss": 0.2805320918560028, "global_step": 51951, "epoch": 583, "lr": 9.13633134744748e-05}
{"train_loss": 0.2233145833015442, "global_step": 51952, "epoch": 583, "lr": 9.136298776985095e-05}
{"train_loss": 0.3532180190086365, "global_step": 51953, "epoch": 583, "lr": 9.136266205966635e-05}
{"train_loss": 0.2679743766784668, "global_step": 51954, "epoch": 583, "lr": 9.136233634392107e-05}
{"train_loss": 0.2987734377384186, "global_step": 51955, "epoch": 583, "lr": 9.136201062261509e-05}
{"train_loss": 0.23059406876564026, "global_step": 51956, "epoch": 583, "lr": 9.136168489574852e-05}
{"train_loss": 0.2237693965435028, "global_step": 51957, "epoch": 583, "lr": 9.136135916332135e-05}
{"train_loss": 0.303780198097229, "global_step": 51958, "epoch": 583, "lr": 9.136103342533368e-05}
{"train_loss": 0.31263476610183716, "global_step": 51959, "epoch": 583, "lr": 9.13607076817855e-05}
{"train_loss": 0.31439894437789917, "global_step": 51960, "epoch": 583, "lr": 9.136038193267688e-05}
{"train_loss": 0.2275548279285431, "global_step": 51961, "epoch": 583, "lr": 9.136005617800788e-05}
{"train_loss": 0.22827522456645966, "global_step": 51962, "epoch": 583, "lr": 9.135973041777851e-05}
{"train_loss": 0.3138582408428192, "global_step": 51963, "epoch": 583, "lr": 9.135940465198883e-05}
{"train_loss": 0.3112604320049286, "global_step": 51964, "epoch": 583, "lr": 9.135907888063888e-05}
{"train_loss": 0.27970457077026367, "global_step": 51965, "epoch": 583, "lr": 9.13587531037287e-05}
{"train_loss": 0.2755117416381836, "global_step": 51966, "epoch": 583, "lr": 9.135842732125835e-05}
{"train_loss": 0.31202489137649536, "global_step": 51967, "epoch": 583, "lr": 9.135810153322786e-05}
{"train_loss": 0.27858206629753113, "global_step": 51968, "epoch": 583, "lr": 9.135777573963728e-05}
{"train_loss": 0.2013300359249115, "global_step": 51969, "epoch": 583, "lr": 9.135744994048666e-05}
{"train_loss": 0.35836198925971985, "global_step": 51970, "epoch": 583, "lr": 9.135712413577601e-05}
{"train_loss": 0.33555078506469727, "global_step": 51971, "epoch": 583, "lr": 9.135679832550542e-05}
{"train_loss": 0.2333875596523285, "global_step": 51972, "epoch": 583, "lr": 9.135647250967491e-05}
{"train_loss": 0.3215351998806, "global_step": 51973, "epoch": 583, "lr": 9.135614668828452e-05}
{"train_loss": 0.23907555639743805, "global_step": 51974, "epoch": 583, "lr": 9.135582086133429e-05}
{"train_loss": 0.28625694201903396, "global_step": 51975, "epoch": 583, "lr": 9.13554950288243e-05, "val_loss": 3.1844427585601807}
{"train_loss": 0.3467060625553131, "global_step": 51976, "epoch": 584, "lr": 9.135516919075455e-05}
{"train_loss": 0.24486106634140015, "global_step": 51977, "epoch": 584, "lr": 9.135484334712509e-05}
{"train_loss": 0.375888854265213, "global_step": 51978, "epoch": 584, "lr": 9.1354517497936e-05}
{"train_loss": 0.3069654703140259, "global_step": 51979, "epoch": 584, "lr": 9.135419164318729e-05}
{"train_loss": 0.3124973773956299, "global_step": 51980, "epoch": 584, "lr": 9.135386578287901e-05}
{"train_loss": 0.39170828461647034, "global_step": 51981, "epoch": 584, "lr": 9.135353991701121e-05}
{"train_loss": 0.34292739629745483, "global_step": 51982, "epoch": 584, "lr": 9.135321404558393e-05}
{"train_loss": 0.19999569654464722, "global_step": 51983, "epoch": 584, "lr": 9.135288816859722e-05}
{"train_loss": 0.2559141218662262, "global_step": 51984, "epoch": 584, "lr": 9.135256228605111e-05}
{"train_loss": 0.2017182856798172, "global_step": 51985, "epoch": 584, "lr": 9.135223639794565e-05}
{"train_loss": 0.25295233726501465, "global_step": 51986, "epoch": 584, "lr": 9.135191050428088e-05}
{"train_loss": 0.276556134223938, "global_step": 51987, "epoch": 584, "lr": 9.135158460505687e-05}
{"train_loss": 0.24189352989196777, "global_step": 51988, "epoch": 584, "lr": 9.135125870027363e-05}
{"train_loss": 0.2748071253299713, "global_step": 51989, "epoch": 584, "lr": 9.135093278993123e-05}
{"train_loss": 0.32841727137565613, "global_step": 51990, "epoch": 584, "lr": 9.135060687402969e-05}
{"train_loss": 0.20232531428337097, "global_step": 51991, "epoch": 584, "lr": 9.135028095256907e-05}
{"train_loss": 0.3009423613548279, "global_step": 51992, "epoch": 584, "lr": 9.13499550255494e-05}
{"train_loss": 0.22070196270942688, "global_step": 51993, "epoch": 584, "lr": 9.134962909297075e-05}
{"train_loss": 0.3325383961200714, "global_step": 51994, "epoch": 584, "lr": 9.134930315483313e-05}
{"train_loss": 0.26684311032295227, "global_step": 51995, "epoch": 584, "lr": 9.13489772111366e-05}
{"train_loss": 0.31830835342407227, "global_step": 51996, "epoch": 584, "lr": 9.134865126188121e-05}
{"train_loss": 0.28099969029426575, "global_step": 51997, "epoch": 584, "lr": 9.134832530706699e-05}
{"train_loss": 0.2833751142024994, "global_step": 51998, "epoch": 584, "lr": 9.1347999346694e-05}
{"train_loss": 0.2930351495742798, "global_step": 51999, "epoch": 584, "lr": 9.134767338076227e-05}
{"train_loss": 0.3655970096588135, "global_step": 52000, "epoch": 584, "lr": 9.134734740927186e-05}
{"train_loss": 0.24418886005878448, "global_step": 52001, "epoch": 584, "lr": 9.13470214322228e-05}
{"train_loss": 0.3539963662624359, "global_step": 52002, "epoch": 584, "lr": 9.134669544961512e-05}
{"train_loss": 0.3702172338962555, "global_step": 52003, "epoch": 584, "lr": 9.13463694614489e-05}
{"train_loss": 0.24162492156028748, "global_step": 52004, "epoch": 584, "lr": 9.134604346772415e-05}
{"train_loss": 0.21940897405147552, "global_step": 52005, "epoch": 584, "lr": 9.134571746844094e-05}
{"train_loss": 0.3396291434764862, "global_step": 52006, "epoch": 584, "lr": 9.13453914635993e-05}
{"train_loss": 0.24422717094421387, "global_step": 52007, "epoch": 584, "lr": 9.134506545319927e-05}
{"train_loss": 0.2949305474758148, "global_step": 52008, "epoch": 584, "lr": 9.134473943724091e-05}
{"train_loss": 0.319903701543808, "global_step": 52009, "epoch": 584, "lr": 9.134441341572426e-05}
{"train_loss": 0.30226632952690125, "global_step": 52010, "epoch": 584, "lr": 9.134408738864935e-05}
{"train_loss": 0.25757718086242676, "global_step": 52011, "epoch": 584, "lr": 9.13437613560162e-05}
{"train_loss": 0.2755662500858307, "global_step": 52012, "epoch": 584, "lr": 9.134343531782491e-05}
{"train_loss": 0.24956487119197845, "global_step": 52013, "epoch": 584, "lr": 9.134310927407552e-05}
{"train_loss": 0.3557114899158478, "global_step": 52014, "epoch": 584, "lr": 9.134278322476804e-05}
{"train_loss": 0.21692079305648804, "global_step": 52015, "epoch": 584, "lr": 9.13424571699025e-05}
{"train_loss": 0.2740527391433716, "global_step": 52016, "epoch": 584, "lr": 9.134213110947901e-05}
{"train_loss": 0.25226515531539917, "global_step": 52017, "epoch": 584, "lr": 9.134180504349754e-05}
{"train_loss": 0.3130204975605011, "global_step": 52018, "epoch": 584, "lr": 9.13414789719582e-05}
{"train_loss": 0.2725697159767151, "global_step": 52019, "epoch": 584, "lr": 9.134115289486098e-05}
{"train_loss": 0.22785842418670654, "global_step": 52020, "epoch": 584, "lr": 9.134082681220594e-05}
{"train_loss": 0.25191730260849, "global_step": 52021, "epoch": 584, "lr": 9.134050072399316e-05}
{"train_loss": 0.22343070805072784, "global_step": 52022, "epoch": 584, "lr": 9.134017463022263e-05}
{"train_loss": 0.24142684042453766, "global_step": 52023, "epoch": 584, "lr": 9.133984853089442e-05}
{"train_loss": 0.36466750502586365, "global_step": 52024, "epoch": 584, "lr": 9.133952242600857e-05}
{"train_loss": 0.2153291404247284, "global_step": 52025, "epoch": 584, "lr": 9.133919631556514e-05}
{"train_loss": 0.2149653136730194, "global_step": 52026, "epoch": 584, "lr": 9.133887019956414e-05}
{"train_loss": 0.19634506106376648, "global_step": 52027, "epoch": 584, "lr": 9.133854407800564e-05}
{"train_loss": 0.3410615921020508, "global_step": 52028, "epoch": 584, "lr": 9.133821795088968e-05}
{"train_loss": 0.2588593363761902, "global_step": 52029, "epoch": 584, "lr": 9.13378918182163e-05}
{"train_loss": 0.25092917680740356, "global_step": 52030, "epoch": 584, "lr": 9.133756567998554e-05}
{"train_loss": 0.27242302894592285, "global_step": 52031, "epoch": 584, "lr": 9.133723953619745e-05}
{"train_loss": 0.2004697322845459, "global_step": 52032, "epoch": 584, "lr": 9.133691338685207e-05}
{"train_loss": 0.29682281613349915, "global_step": 52033, "epoch": 584, "lr": 9.133658723194945e-05}
{"train_loss": 0.21337687969207764, "global_step": 52034, "epoch": 584, "lr": 9.133626107148963e-05}
{"train_loss": 0.30352285504341125, "global_step": 52035, "epoch": 584, "lr": 9.133593490547265e-05}
{"train_loss": 0.28990310430526733, "global_step": 52036, "epoch": 584, "lr": 9.133560873389855e-05}
{"train_loss": 0.20405666530132294, "global_step": 52037, "epoch": 584, "lr": 9.13352825567674e-05}
{"train_loss": 0.29559245705604553, "global_step": 52038, "epoch": 584, "lr": 9.13349563740792e-05}
{"train_loss": 0.30386921763420105, "global_step": 52039, "epoch": 584, "lr": 9.133463018583405e-05}
{"train_loss": 0.23535935580730438, "global_step": 52040, "epoch": 584, "lr": 9.133430399203194e-05}
{"train_loss": 0.2913571894168854, "global_step": 52041, "epoch": 584, "lr": 9.133397779267295e-05}
{"train_loss": 0.23387524485588074, "global_step": 52042, "epoch": 584, "lr": 9.13336515877571e-05}
{"train_loss": 0.26751136779785156, "global_step": 52043, "epoch": 584, "lr": 9.133332537728444e-05}
{"train_loss": 0.19118408858776093, "global_step": 52044, "epoch": 584, "lr": 9.133299916125503e-05}
{"train_loss": 0.28966444730758667, "global_step": 52045, "epoch": 584, "lr": 9.13326729396689e-05}
{"train_loss": 0.2730833888053894, "global_step": 52046, "epoch": 584, "lr": 9.133234671252608e-05}
{"train_loss": 0.3003292679786682, "global_step": 52047, "epoch": 584, "lr": 9.133202047982666e-05}
{"train_loss": 0.24097222089767456, "global_step": 52048, "epoch": 584, "lr": 9.133169424157064e-05}
{"train_loss": 0.2146906554698944, "global_step": 52049, "epoch": 584, "lr": 9.133136799775806e-05}
{"train_loss": 0.2774624824523926, "global_step": 52050, "epoch": 584, "lr": 9.1331041748389e-05}
{"train_loss": 0.29192855954170227, "global_step": 52051, "epoch": 584, "lr": 9.133071549346348e-05}
{"train_loss": 0.2675071656703949, "global_step": 52052, "epoch": 584, "lr": 9.133038923298155e-05}
{"train_loss": 0.19758254289627075, "global_step": 52053, "epoch": 584, "lr": 9.133006296694326e-05}
{"train_loss": 0.24606940150260925, "global_step": 52054, "epoch": 584, "lr": 9.132973669534864e-05}
{"train_loss": 0.1777656376361847, "global_step": 52055, "epoch": 584, "lr": 9.132941041819774e-05}
{"train_loss": 0.3097870945930481, "global_step": 52056, "epoch": 584, "lr": 9.13290841354906e-05}
{"train_loss": 0.24866551160812378, "global_step": 52057, "epoch": 584, "lr": 9.132875784722728e-05}
{"train_loss": 0.19787245988845825, "global_step": 52058, "epoch": 584, "lr": 9.13284315534078e-05}
{"train_loss": 0.23210638761520386, "global_step": 52059, "epoch": 584, "lr": 9.132810525403222e-05}
{"train_loss": 0.24764803051948547, "global_step": 52060, "epoch": 584, "lr": 9.132777894910059e-05}
{"train_loss": 0.3788883090019226, "global_step": 52061, "epoch": 584, "lr": 9.132745263861293e-05}
{"train_loss": 0.30809056758880615, "global_step": 52062, "epoch": 584, "lr": 9.13271263225693e-05}
{"train_loss": 0.1797827035188675, "global_step": 52063, "epoch": 584, "lr": 9.132680000096975e-05}
{"train_loss": 0.2703012507618143, "global_step": 52064, "epoch": 584, "lr": 9.132647367381431e-05, "val_loss": 3.224576234817505}
{"train_loss": 0.22995935380458832, "global_step": 52065, "epoch": 585, "lr": 9.132614734110304e-05}
{"train_loss": 0.29591619968414307, "global_step": 52066, "epoch": 585, "lr": 9.132582100283596e-05}
{"train_loss": 0.26005834341049194, "global_step": 52067, "epoch": 585, "lr": 9.132549465901312e-05}
{"train_loss": 0.26865917444229126, "global_step": 52068, "epoch": 585, "lr": 9.132516830963458e-05}
{"train_loss": 0.3127434253692627, "global_step": 52069, "epoch": 585, "lr": 9.132484195470038e-05}
{"train_loss": 0.26812124252319336, "global_step": 52070, "epoch": 585, "lr": 9.132451559421056e-05}
{"train_loss": 0.21778850257396698, "global_step": 52071, "epoch": 585, "lr": 9.132418922816514e-05}
{"train_loss": 0.28300246596336365, "global_step": 52072, "epoch": 585, "lr": 9.132386285656421e-05}
{"train_loss": 0.25331971049308777, "global_step": 52073, "epoch": 585, "lr": 9.132353647940779e-05}
{"train_loss": 0.31712865829467773, "global_step": 52074, "epoch": 585, "lr": 9.13232100966959e-05}
{"train_loss": 0.28803107142448425, "global_step": 52075, "epoch": 585, "lr": 9.132288370842864e-05}
{"train_loss": 0.25494086742401123, "global_step": 52076, "epoch": 585, "lr": 9.132255731460601e-05}
{"train_loss": 0.27067235112190247, "global_step": 52077, "epoch": 585, "lr": 9.132223091522805e-05}
{"train_loss": 0.24026958644390106, "global_step": 52078, "epoch": 585, "lr": 9.132190451029484e-05}
{"train_loss": 0.19875019788742065, "global_step": 52079, "epoch": 585, "lr": 9.132157809980639e-05}
{"train_loss": 0.3451940417289734, "global_step": 52080, "epoch": 585, "lr": 9.132125168376278e-05}
{"train_loss": 0.23527733981609344, "global_step": 52081, "epoch": 585, "lr": 9.1320925262164e-05}
{"train_loss": 0.37611860036849976, "global_step": 52082, "epoch": 585, "lr": 9.132059883501016e-05}
{"train_loss": 0.29628241062164307, "global_step": 52083, "epoch": 585, "lr": 9.132027240230125e-05}
{"train_loss": 0.2093314677476883, "global_step": 52084, "epoch": 585, "lr": 9.131994596403732e-05}
{"train_loss": 0.3051689863204956, "global_step": 52085, "epoch": 585, "lr": 9.131961952021845e-05}
{"train_loss": 0.2581389248371124, "global_step": 52086, "epoch": 585, "lr": 9.131929307084465e-05}
{"train_loss": 0.14605221152305603, "global_step": 52087, "epoch": 585, "lr": 9.131896661591597e-05}
{"train_loss": 0.23930270969867706, "global_step": 52088, "epoch": 585, "lr": 9.131864015543247e-05}
{"train_loss": 0.2612435221672058, "global_step": 52089, "epoch": 585, "lr": 9.131831368939416e-05}
{"train_loss": 0.1876652091741562, "global_step": 52090, "epoch": 585, "lr": 9.131798721780114e-05}
{"train_loss": 0.2382090538740158, "global_step": 52091, "epoch": 585, "lr": 9.13176607406534e-05}
{"train_loss": 0.15124721825122833, "global_step": 52092, "epoch": 585, "lr": 9.131733425795101e-05}
{"train_loss": 0.3855237662792206, "global_step": 52093, "epoch": 585, "lr": 9.1317007769694e-05}
{"train_loss": 0.25981569290161133, "global_step": 52094, "epoch": 585, "lr": 9.131668127588242e-05}
{"train_loss": 0.1891959011554718, "global_step": 52095, "epoch": 585, "lr": 9.131635477651632e-05}
{"train_loss": 0.262756884098053, "global_step": 52096, "epoch": 585, "lr": 9.131602827159575e-05}
{"train_loss": 0.3447684049606323, "global_step": 52097, "epoch": 585, "lr": 9.131570176112075e-05}
{"train_loss": 0.1786729395389557, "global_step": 52098, "epoch": 585, "lr": 9.131537524509132e-05}
{"train_loss": 0.2305373102426529, "global_step": 52099, "epoch": 585, "lr": 9.131504872350759e-05}
{"train_loss": 0.21244820952415466, "global_step": 52100, "epoch": 585, "lr": 9.131472219636952e-05}
{"train_loss": 0.26259809732437134, "global_step": 52101, "epoch": 585, "lr": 9.131439566367718e-05}
{"train_loss": 0.20679229497909546, "global_step": 52102, "epoch": 585, "lr": 9.131406912543065e-05}
{"train_loss": 0.18571272492408752, "global_step": 52103, "epoch": 585, "lr": 9.131374258162995e-05}
{"train_loss": 0.2544423043727875, "global_step": 52104, "epoch": 585, "lr": 9.13134160322751e-05}
{"train_loss": 0.23392146825790405, "global_step": 52105, "epoch": 585, "lr": 9.131308947736617e-05}
{"train_loss": 0.32524046301841736, "global_step": 52106, "epoch": 585, "lr": 9.13127629169032e-05}
{"train_loss": 0.37468039989471436, "global_step": 52107, "epoch": 585, "lr": 9.131243635088625e-05}
{"train_loss": 0.26569175720214844, "global_step": 52108, "epoch": 585, "lr": 9.131210977931533e-05}
{"train_loss": 0.2789003252983093, "global_step": 52109, "epoch": 585, "lr": 9.13117832021905e-05}
{"train_loss": 0.21750888228416443, "global_step": 52110, "epoch": 585, "lr": 9.131145661951182e-05}
{"train_loss": 0.3095395565032959, "global_step": 52111, "epoch": 585, "lr": 9.131113003127929e-05}
{"train_loss": 0.1873788833618164, "global_step": 52112, "epoch": 585, "lr": 9.131080343749301e-05}
{"train_loss": 0.33667823672294617, "global_step": 52113, "epoch": 585, "lr": 9.131047683815297e-05}
{"train_loss": 0.27991658449172974, "global_step": 52114, "epoch": 585, "lr": 9.131015023325926e-05}
{"train_loss": 0.2811211943626404, "global_step": 52115, "epoch": 585, "lr": 9.13098236228119e-05}
{"train_loss": 0.16876685619354248, "global_step": 52116, "epoch": 585, "lr": 9.130949700681093e-05}
{"train_loss": 0.2642660140991211, "global_step": 52117, "epoch": 585, "lr": 9.130917038525642e-05}
{"train_loss": 0.22788798809051514, "global_step": 52118, "epoch": 585, "lr": 9.130884375814837e-05}
{"train_loss": 0.2053319215774536, "global_step": 52119, "epoch": 585, "lr": 9.130851712548686e-05}
{"train_loss": 0.29300373792648315, "global_step": 52120, "epoch": 585, "lr": 9.130819048727193e-05}
{"train_loss": 0.23402488231658936, "global_step": 52121, "epoch": 585, "lr": 9.13078638435036e-05}
{"train_loss": 0.20608483254909515, "global_step": 52122, "epoch": 585, "lr": 9.130753719418194e-05}
{"train_loss": 0.15005460381507874, "global_step": 52123, "epoch": 585, "lr": 9.1307210539307e-05}
{"train_loss": 0.2586597204208374, "global_step": 52124, "epoch": 585, "lr": 9.13068838788788e-05}
{"train_loss": 0.3762553036212921, "global_step": 52125, "epoch": 585, "lr": 9.130655721289739e-05}
{"train_loss": 0.2563454210758209, "global_step": 52126, "epoch": 585, "lr": 9.130623054136281e-05}
{"train_loss": 0.3124365508556366, "global_step": 52127, "epoch": 585, "lr": 9.130590386427513e-05}
{"train_loss": 0.23552608489990234, "global_step": 52128, "epoch": 585, "lr": 9.130557718163435e-05}
{"train_loss": 0.25934600830078125, "global_step": 52129, "epoch": 585, "lr": 9.130525049344056e-05}
{"train_loss": 0.30926984548568726, "global_step": 52130, "epoch": 585, "lr": 9.130492379969377e-05}
{"train_loss": 0.25850728154182434, "global_step": 52131, "epoch": 585, "lr": 9.130459710039402e-05}
{"train_loss": 0.2394631952047348, "global_step": 52132, "epoch": 585, "lr": 9.130427039554141e-05}
{"train_loss": 0.28729069232940674, "global_step": 52133, "epoch": 585, "lr": 9.130394368513592e-05}
{"train_loss": 0.37801533937454224, "global_step": 52134, "epoch": 585, "lr": 9.130361696917761e-05}
{"train_loss": 0.2694575786590576, "global_step": 52135, "epoch": 585, "lr": 9.130329024766655e-05}
{"train_loss": 0.1813875138759613, "global_step": 52136, "epoch": 585, "lr": 9.130296352060277e-05}
{"train_loss": 0.3636467456817627, "global_step": 52137, "epoch": 585, "lr": 9.13026367879863e-05}
{"train_loss": 0.3599526286125183, "global_step": 52138, "epoch": 585, "lr": 9.13023100498172e-05}
{"train_loss": 0.27672651410102844, "global_step": 52139, "epoch": 585, "lr": 9.13019833060955e-05}
{"train_loss": 0.38326430320739746, "global_step": 52140, "epoch": 585, "lr": 9.130165655682125e-05}
{"train_loss": 0.38749265670776367, "global_step": 52141, "epoch": 585, "lr": 9.130132980199451e-05}
{"train_loss": 0.2832433581352234, "global_step": 52142, "epoch": 585, "lr": 9.130100304161528e-05}
{"train_loss": 0.22365078330039978, "global_step": 52143, "epoch": 585, "lr": 9.130067627568366e-05}
{"train_loss": 0.1634443998336792, "global_step": 52144, "epoch": 585, "lr": 9.130034950419967e-05}
{"train_loss": 0.28824982047080994, "global_step": 52145, "epoch": 585, "lr": 9.130002272716334e-05}
{"train_loss": 0.31885066628456116, "global_step": 52146, "epoch": 585, "lr": 9.129969594457472e-05}
{"train_loss": 0.36190107464790344, "global_step": 52147, "epoch": 585, "lr": 9.129936915643386e-05}
{"train_loss": 0.32023781538009644, "global_step": 52148, "epoch": 585, "lr": 9.129904236274083e-05}
{"train_loss": 0.3255201280117035, "global_step": 52149, "epoch": 585, "lr": 9.12987155634956e-05}
{"train_loss": 0.2874278128147125, "global_step": 52150, "epoch": 585, "lr": 9.12983887586983e-05}
{"train_loss": 0.2874048352241516, "global_step": 52151, "epoch": 585, "lr": 9.129806194834894e-05}
{"train_loss": 0.2567426860332489, "global_step": 52152, "epoch": 585, "lr": 9.129773513244753e-05}
{"train_loss": 0.26855785156903644, "global_step": 52153, "epoch": 585, "lr": 9.129740831099416e-05, "val_loss": 3.220308303833008, "train_action_mse_error": 23.039548873901367}
{"train_loss": 0.2584904134273529, "global_step": 52154, "epoch": 586, "lr": 9.129708148398884e-05}
{"train_loss": 0.29040271043777466, "global_step": 52155, "epoch": 586, "lr": 9.129675465143165e-05}
{"train_loss": 0.31222307682037354, "global_step": 52156, "epoch": 586, "lr": 9.129642781332261e-05}
{"train_loss": 0.42007049918174744, "global_step": 52157, "epoch": 586, "lr": 9.129610096966176e-05}
{"train_loss": 0.34190869331359863, "global_step": 52158, "epoch": 586, "lr": 9.129577412044917e-05}
{"train_loss": 0.23516428470611572, "global_step": 52159, "epoch": 586, "lr": 9.129544726568484e-05}
{"train_loss": 0.26921114325523376, "global_step": 52160, "epoch": 586, "lr": 9.129512040536885e-05}
{"train_loss": 0.35062670707702637, "global_step": 52161, "epoch": 586, "lr": 9.129479353950125e-05}
{"train_loss": 0.290363609790802, "global_step": 52162, "epoch": 586, "lr": 9.129446666808206e-05}
{"train_loss": 0.2617684006690979, "global_step": 52163, "epoch": 586, "lr": 9.129413979111133e-05}
{"train_loss": 0.2804265320301056, "global_step": 52164, "epoch": 586, "lr": 9.12938129085891e-05}
{"train_loss": 0.26124149560928345, "global_step": 52165, "epoch": 586, "lr": 9.129348602051543e-05}
{"train_loss": 0.29828760027885437, "global_step": 52166, "epoch": 586, "lr": 9.129315912689034e-05}
{"train_loss": 0.23337818682193756, "global_step": 52167, "epoch": 586, "lr": 9.12928322277139e-05}
{"train_loss": 0.19945767521858215, "global_step": 52168, "epoch": 586, "lr": 9.129250532298614e-05}
{"train_loss": 0.3992873430252075, "global_step": 52169, "epoch": 586, "lr": 9.129217841270711e-05}
{"train_loss": 0.2560756206512451, "global_step": 52170, "epoch": 586, "lr": 9.129185149687685e-05}
{"train_loss": 0.2742089033126831, "global_step": 52171, "epoch": 586, "lr": 9.12915245754954e-05}
{"train_loss": 0.30845195055007935, "global_step": 52172, "epoch": 586, "lr": 9.12911976485628e-05}
{"train_loss": 0.2822677195072174, "global_step": 52173, "epoch": 586, "lr": 9.129087071607911e-05}
{"train_loss": 0.38384097814559937, "global_step": 52174, "epoch": 586, "lr": 9.129054377804438e-05}
{"train_loss": 0.20060229301452637, "global_step": 52175, "epoch": 586, "lr": 9.129021683445862e-05}
{"train_loss": 0.3234223425388336, "global_step": 52176, "epoch": 586, "lr": 9.12898898853219e-05}
{"train_loss": 0.15459880232810974, "global_step": 52177, "epoch": 586, "lr": 9.128956293063425e-05}
{"train_loss": 0.29908257722854614, "global_step": 52178, "epoch": 586, "lr": 9.128923597039573e-05}
{"train_loss": 0.22891315817832947, "global_step": 52179, "epoch": 586, "lr": 9.128890900460637e-05}
{"train_loss": 0.2685801684856415, "global_step": 52180, "epoch": 586, "lr": 9.128858203326623e-05}
{"train_loss": 0.3453904986381531, "global_step": 52181, "epoch": 586, "lr": 9.128825505637532e-05}
{"train_loss": 0.1832553595304489, "global_step": 52182, "epoch": 586, "lr": 9.128792807393374e-05}
{"train_loss": 0.35690152645111084, "global_step": 52183, "epoch": 586, "lr": 9.128760108594147e-05}
{"train_loss": 0.17103607952594757, "global_step": 52184, "epoch": 586, "lr": 9.128727409239861e-05}
{"train_loss": 0.20601874589920044, "global_step": 52185, "epoch": 586, "lr": 9.128694709330517e-05}
{"train_loss": 0.1976875513792038, "global_step": 52186, "epoch": 586, "lr": 9.12866200886612e-05}
{"train_loss": 0.2930476665496826, "global_step": 52187, "epoch": 586, "lr": 9.128629307846675e-05}
{"train_loss": 0.30995047092437744, "global_step": 52188, "epoch": 586, "lr": 9.128596606272186e-05}
{"train_loss": 0.1991988867521286, "global_step": 52189, "epoch": 586, "lr": 9.128563904142658e-05}
{"train_loss": 0.2743668854236603, "global_step": 52190, "epoch": 586, "lr": 9.128531201458095e-05}
{"train_loss": 0.32876598834991455, "global_step": 52191, "epoch": 586, "lr": 9.128498498218501e-05}
{"train_loss": 0.17686057090759277, "global_step": 52192, "epoch": 586, "lr": 9.12846579442388e-05}
{"train_loss": 0.23789440095424652, "global_step": 52193, "epoch": 586, "lr": 9.128433090074238e-05}
{"train_loss": 0.1849193572998047, "global_step": 52194, "epoch": 586, "lr": 9.128400385169577e-05}
{"train_loss": 0.24625957012176514, "global_step": 52195, "epoch": 586, "lr": 9.128367679709904e-05}
{"train_loss": 0.25253036618232727, "global_step": 52196, "epoch": 586, "lr": 9.128334973695222e-05}
{"train_loss": 0.25060948729515076, "global_step": 52197, "epoch": 586, "lr": 9.128302267125536e-05}
{"train_loss": 0.25998762249946594, "global_step": 52198, "epoch": 586, "lr": 9.12826956000085e-05}
{"train_loss": 0.4146941900253296, "global_step": 52199, "epoch": 586, "lr": 9.128236852321169e-05}
{"train_loss": 0.2856505811214447, "global_step": 52200, "epoch": 586, "lr": 9.128204144086497e-05}
{"train_loss": 0.25547224283218384, "global_step": 52201, "epoch": 586, "lr": 9.128171435296836e-05}
{"train_loss": 0.23045258224010468, "global_step": 52202, "epoch": 586, "lr": 9.128138725952194e-05}
{"train_loss": 0.25002631545066833, "global_step": 52203, "epoch": 586, "lr": 9.128106016052574e-05}
{"train_loss": 0.2501477301120758, "global_step": 52204, "epoch": 586, "lr": 9.128073305597982e-05}
{"train_loss": 0.3206159472465515, "global_step": 52205, "epoch": 586, "lr": 9.128040594588419e-05}
{"train_loss": 0.30680882930755615, "global_step": 52206, "epoch": 586, "lr": 9.128007883023892e-05}
{"train_loss": 0.26455187797546387, "global_step": 52207, "epoch": 586, "lr": 9.127975170904404e-05}
{"train_loss": 0.20877346396446228, "global_step": 52208, "epoch": 586, "lr": 9.127942458229961e-05}
{"train_loss": 0.16212666034698486, "global_step": 52209, "epoch": 586, "lr": 9.127909745000567e-05}
{"train_loss": 0.16797621548175812, "global_step": 52210, "epoch": 586, "lr": 9.127877031216224e-05}
{"train_loss": 0.25207406282424927, "global_step": 52211, "epoch": 586, "lr": 9.127844316876938e-05}
{"train_loss": 0.22489185631275177, "global_step": 52212, "epoch": 586, "lr": 9.127811601982717e-05}
{"train_loss": 0.35305917263031006, "global_step": 52213, "epoch": 586, "lr": 9.127778886533558e-05}
{"train_loss": 0.2576051950454712, "global_step": 52214, "epoch": 586, "lr": 9.127746170529472e-05}
{"train_loss": 0.30837282538414, "global_step": 52215, "epoch": 586, "lr": 9.12771345397046e-05}
{"train_loss": 0.1453792005777359, "global_step": 52216, "epoch": 586, "lr": 9.127680736856527e-05}
{"train_loss": 0.1699289232492447, "global_step": 52217, "epoch": 586, "lr": 9.127648019187678e-05}
{"train_loss": 0.24294333159923553, "global_step": 52218, "epoch": 586, "lr": 9.127615300963916e-05}
{"train_loss": 0.318583220243454, "global_step": 52219, "epoch": 586, "lr": 9.12758258218525e-05}
{"train_loss": 0.19546450674533844, "global_step": 52220, "epoch": 586, "lr": 9.127549862851677e-05}
{"train_loss": 0.28295087814331055, "global_step": 52221, "epoch": 586, "lr": 9.127517142963207e-05}
{"train_loss": 0.3257457911968231, "global_step": 52222, "epoch": 586, "lr": 9.127484422519841e-05}
{"train_loss": 0.3263971209526062, "global_step": 52223, "epoch": 586, "lr": 9.127451701521587e-05}
{"train_loss": 0.32479819655418396, "global_step": 52224, "epoch": 586, "lr": 9.127418979968447e-05}
{"train_loss": 0.31470584869384766, "global_step": 52225, "epoch": 586, "lr": 9.127386257860424e-05}
{"train_loss": 0.20765087008476257, "global_step": 52226, "epoch": 586, "lr": 9.127353535197525e-05}
{"train_loss": 0.21982735395431519, "global_step": 52227, "epoch": 586, "lr": 9.127320811979756e-05}
{"train_loss": 0.4423050582408905, "global_step": 52228, "epoch": 586, "lr": 9.127288088207117e-05}
{"train_loss": 0.30452045798301697, "global_step": 52229, "epoch": 586, "lr": 9.127255363879615e-05}
{"train_loss": 0.3964441120624542, "global_step": 52230, "epoch": 586, "lr": 9.127222638997253e-05}
{"train_loss": 0.3451973497867584, "global_step": 52231, "epoch": 586, "lr": 9.127189913560037e-05}
{"train_loss": 0.2831491231918335, "global_step": 52232, "epoch": 586, "lr": 9.127157187567969e-05}
{"train_loss": 0.28174811601638794, "global_step": 52233, "epoch": 586, "lr": 9.127124461021057e-05}
{"train_loss": 0.2167513370513916, "global_step": 52234, "epoch": 586, "lr": 9.127091733919302e-05}
{"train_loss": 0.29576873779296875, "global_step": 52235, "epoch": 586, "lr": 9.127059006262711e-05}
{"train_loss": 0.32541733980178833, "global_step": 52236, "epoch": 586, "lr": 9.127026278051287e-05}
{"train_loss": 0.2564378082752228, "global_step": 52237, "epoch": 586, "lr": 9.126993549285035e-05}
{"train_loss": 0.29187700152397156, "global_step": 52238, "epoch": 586, "lr": 9.126960819963959e-05}
{"train_loss": 0.2791966199874878, "global_step": 52239, "epoch": 586, "lr": 9.126928090088062e-05}
{"train_loss": 0.30670738220214844, "global_step": 52240, "epoch": 586, "lr": 9.12689535965735e-05}
{"train_loss": 0.26465532183647156, "global_step": 52241, "epoch": 586, "lr": 9.126862628671828e-05}
{"train_loss": 0.27358180301243, "global_step": 52242, "epoch": 586, "lr": 9.1268298971315e-05, "val_loss": 3.1456451416015625}
{"train_loss": 0.301654189825058, "global_step": 52243, "epoch": 587, "lr": 9.126797165036369e-05}
{"train_loss": 0.20419518649578094, "global_step": 52244, "epoch": 587, "lr": 9.126764432386441e-05}
{"train_loss": 0.2457432746887207, "global_step": 52245, "epoch": 587, "lr": 9.126731699181721e-05}
{"train_loss": 0.2591588497161865, "global_step": 52246, "epoch": 587, "lr": 9.126698965422211e-05}
{"train_loss": 0.28846275806427, "global_step": 52247, "epoch": 587, "lr": 9.126666231107917e-05}
{"train_loss": 0.184050053358078, "global_step": 52248, "epoch": 587, "lr": 9.126633496238841e-05}
{"train_loss": 0.2858581840991974, "global_step": 52249, "epoch": 587, "lr": 9.126600760814991e-05}
{"train_loss": 0.2834767699241638, "global_step": 52250, "epoch": 587, "lr": 9.12656802483637e-05}
{"train_loss": 0.37422508001327515, "global_step": 52251, "epoch": 587, "lr": 9.126535288302983e-05}
{"train_loss": 0.31822511553764343, "global_step": 52252, "epoch": 587, "lr": 9.126502551214833e-05}
{"train_loss": 0.20974308252334595, "global_step": 52253, "epoch": 587, "lr": 9.126469813571925e-05}
{"train_loss": 0.2847626805305481, "global_step": 52254, "epoch": 587, "lr": 9.126437075374263e-05}
{"train_loss": 0.2498071789741516, "global_step": 52255, "epoch": 587, "lr": 9.126404336621852e-05}
{"train_loss": 0.4071286916732788, "global_step": 52256, "epoch": 587, "lr": 9.126371597314696e-05}
{"train_loss": 0.2547944188117981, "global_step": 52257, "epoch": 587, "lr": 9.126338857452799e-05}
{"train_loss": 0.250583678483963, "global_step": 52258, "epoch": 587, "lr": 9.126306117036167e-05}
{"train_loss": 0.2175302356481552, "global_step": 52259, "epoch": 587, "lr": 9.126273376064805e-05}
{"train_loss": 0.32943177223205566, "global_step": 52260, "epoch": 587, "lr": 9.126240634538713e-05}
{"train_loss": 0.2588028907775879, "global_step": 52261, "epoch": 587, "lr": 9.1262078924579e-05}
{"train_loss": 0.27232497930526733, "global_step": 52262, "epoch": 587, "lr": 9.126175149822368e-05}
{"train_loss": 0.26592332124710083, "global_step": 52263, "epoch": 587, "lr": 9.126142406632122e-05}
{"train_loss": 0.2584916949272156, "global_step": 52264, "epoch": 587, "lr": 9.126109662887167e-05}
{"train_loss": 0.2805682122707367, "global_step": 52265, "epoch": 587, "lr": 9.126076918587506e-05}
{"train_loss": 0.22452761232852936, "global_step": 52266, "epoch": 587, "lr": 9.126044173733145e-05}
{"train_loss": 0.24828475713729858, "global_step": 52267, "epoch": 587, "lr": 9.126011428324086e-05}
{"train_loss": 0.36572158336639404, "global_step": 52268, "epoch": 587, "lr": 9.125978682360337e-05}
{"train_loss": 0.24791206419467926, "global_step": 52269, "epoch": 587, "lr": 9.1259459358419e-05}
{"train_loss": 0.30963563919067383, "global_step": 52270, "epoch": 587, "lr": 9.12591318876878e-05}
{"train_loss": 0.3091672956943512, "global_step": 52271, "epoch": 587, "lr": 9.125880441140981e-05}
{"train_loss": 0.17124265432357788, "global_step": 52272, "epoch": 587, "lr": 9.125847692958507e-05}
{"train_loss": 0.277513325214386, "global_step": 52273, "epoch": 587, "lr": 9.125814944221364e-05}
{"train_loss": 0.29306289553642273, "global_step": 52274, "epoch": 587, "lr": 9.125782194929556e-05}
{"train_loss": 0.27715733647346497, "global_step": 52275, "epoch": 587, "lr": 9.125749445083085e-05}
{"train_loss": 0.2931731343269348, "global_step": 52276, "epoch": 587, "lr": 9.125716694681958e-05}
{"train_loss": 0.3239363729953766, "global_step": 52277, "epoch": 587, "lr": 9.12568394372618e-05}
{"train_loss": 0.2557850778102875, "global_step": 52278, "epoch": 587, "lr": 9.125651192215752e-05}
{"train_loss": 0.2571924328804016, "global_step": 52279, "epoch": 587, "lr": 9.125618440150682e-05}
{"train_loss": 0.23133300244808197, "global_step": 52280, "epoch": 587, "lr": 9.125585687530971e-05}
{"train_loss": 0.2884553074836731, "global_step": 52281, "epoch": 587, "lr": 9.125552934356627e-05}
{"train_loss": 0.35754889249801636, "global_step": 52282, "epoch": 587, "lr": 9.125520180627653e-05}
{"train_loss": 0.3649176061153412, "global_step": 52283, "epoch": 587, "lr": 9.125487426344052e-05}
{"train_loss": 0.22227196395397186, "global_step": 52284, "epoch": 587, "lr": 9.12545467150583e-05}
{"train_loss": 0.1884000301361084, "global_step": 52285, "epoch": 587, "lr": 9.125421916112992e-05}
{"train_loss": 0.2637992799282074, "global_step": 52286, "epoch": 587, "lr": 9.12538916016554e-05}
{"train_loss": 0.3060852289199829, "global_step": 52287, "epoch": 587, "lr": 9.12535640366348e-05}
{"train_loss": 0.2768401801586151, "global_step": 52288, "epoch": 587, "lr": 9.125323646606816e-05}
{"train_loss": 0.21868176758289337, "global_step": 52289, "epoch": 587, "lr": 9.125290888995553e-05}
{"train_loss": 0.22963373363018036, "global_step": 52290, "epoch": 587, "lr": 9.125258130829694e-05}
{"train_loss": 0.3169170022010803, "global_step": 52291, "epoch": 587, "lr": 9.125225372109243e-05}
{"train_loss": 0.3167451322078705, "global_step": 52292, "epoch": 587, "lr": 9.125192612834208e-05}
{"train_loss": 0.2723580002784729, "global_step": 52293, "epoch": 587, "lr": 9.125159853004592e-05}
{"train_loss": 0.27306365966796875, "global_step": 52294, "epoch": 587, "lr": 9.125127092620395e-05}
{"train_loss": 0.37566855549812317, "global_step": 52295, "epoch": 587, "lr": 9.125094331681628e-05}
{"train_loss": 0.21142461895942688, "global_step": 52296, "epoch": 587, "lr": 9.125061570188292e-05}
{"train_loss": 0.2247564047574997, "global_step": 52297, "epoch": 587, "lr": 9.125028808140392e-05}
{"train_loss": 0.2542301118373871, "global_step": 52298, "epoch": 587, "lr": 9.12499604553793e-05}
{"train_loss": 0.28242966532707214, "global_step": 52299, "epoch": 587, "lr": 9.124963282380914e-05}
{"train_loss": 0.25509077310562134, "global_step": 52300, "epoch": 587, "lr": 9.124930518669348e-05}
{"train_loss": 0.2391013652086258, "global_step": 52301, "epoch": 587, "lr": 9.124897754403234e-05}
{"train_loss": 0.24957533180713654, "global_step": 52302, "epoch": 587, "lr": 9.124864989582577e-05}
{"train_loss": 0.21724466979503632, "global_step": 52303, "epoch": 587, "lr": 9.124832224207384e-05}
{"train_loss": 0.23948726058006287, "global_step": 52304, "epoch": 587, "lr": 9.124799458277656e-05}
{"train_loss": 0.2404201477766037, "global_step": 52305, "epoch": 587, "lr": 9.124766691793401e-05}
{"train_loss": 0.24845774471759796, "global_step": 52306, "epoch": 587, "lr": 9.12473392475462e-05}
{"train_loss": 0.26186826825141907, "global_step": 52307, "epoch": 587, "lr": 9.124701157161318e-05}
{"train_loss": 0.22869527339935303, "global_step": 52308, "epoch": 587, "lr": 9.124668389013502e-05}
{"train_loss": 0.20669730007648468, "global_step": 52309, "epoch": 587, "lr": 9.124635620311173e-05}
{"train_loss": 0.27214401960372925, "global_step": 52310, "epoch": 587, "lr": 9.124602851054339e-05}
{"train_loss": 0.27787619829177856, "global_step": 52311, "epoch": 587, "lr": 9.124570081243e-05}
{"train_loss": 0.33571290969848633, "global_step": 52312, "epoch": 587, "lr": 9.124537310877163e-05}
{"train_loss": 0.284561425447464, "global_step": 52313, "epoch": 587, "lr": 9.124504539956834e-05}
{"train_loss": 0.33278021216392517, "global_step": 52314, "epoch": 587, "lr": 9.124471768482015e-05}
{"train_loss": 0.27292701601982117, "global_step": 52315, "epoch": 587, "lr": 9.124438996452711e-05}
{"train_loss": 0.1592482477426529, "global_step": 52316, "epoch": 587, "lr": 9.124406223868927e-05}
{"train_loss": 0.29205867648124695, "global_step": 52317, "epoch": 587, "lr": 9.124373450730665e-05}
{"train_loss": 0.20413850247859955, "global_step": 52318, "epoch": 587, "lr": 9.124340677037932e-05}
{"train_loss": 0.22793222963809967, "global_step": 52319, "epoch": 587, "lr": 9.124307902790733e-05}
{"train_loss": 0.26085278391838074, "global_step": 52320, "epoch": 587, "lr": 9.12427512798907e-05}
{"train_loss": 0.21458445489406586, "global_step": 52321, "epoch": 587, "lr": 9.124242352632949e-05}
{"train_loss": 0.2533281445503235, "global_step": 52322, "epoch": 587, "lr": 9.124209576722373e-05}
{"train_loss": 0.3987550437450409, "global_step": 52323, "epoch": 587, "lr": 9.124176800257349e-05}
{"train_loss": 0.29406309127807617, "global_step": 52324, "epoch": 587, "lr": 9.124144023237878e-05}
{"train_loss": 0.3464388847351074, "global_step": 52325, "epoch": 587, "lr": 9.124111245663967e-05}
{"train_loss": 0.2303139716386795, "global_step": 52326, "epoch": 587, "lr": 9.12407846753562e-05}
{"train_loss": 0.21409136056900024, "global_step": 52327, "epoch": 587, "lr": 9.12404568885284e-05}
{"train_loss": 0.2481505423784256, "global_step": 52328, "epoch": 587, "lr": 9.124012909615632e-05}
{"train_loss": 0.2522253692150116, "global_step": 52329, "epoch": 587, "lr": 9.123980129824e-05}
{"train_loss": 0.24081067740917206, "global_step": 52330, "epoch": 587, "lr": 9.12394734947795e-05}
{"train_loss": 0.2692689688018199, "global_step": 52331, "epoch": 587, "lr": 9.123914568577486e-05, "val_loss": 3.221778154373169}
{"train_loss": 0.36919599771499634, "global_step": 52332, "epoch": 588, "lr": 9.123881787122612e-05}
{"train_loss": 0.21966499090194702, "global_step": 52333, "epoch": 588, "lr": 9.123849005113332e-05}
{"train_loss": 0.25986263155937195, "global_step": 52334, "epoch": 588, "lr": 9.12381622254965e-05}
{"train_loss": 0.27015307545661926, "global_step": 52335, "epoch": 588, "lr": 9.123783439431573e-05}
{"train_loss": 0.257479190826416, "global_step": 52336, "epoch": 588, "lr": 9.123750655759102e-05}
{"train_loss": 0.23222199082374573, "global_step": 52337, "epoch": 588, "lr": 9.123717871532244e-05}
{"train_loss": 0.23484012484550476, "global_step": 52338, "epoch": 588, "lr": 9.123685086751003e-05}
{"train_loss": 0.2776136100292206, "global_step": 52339, "epoch": 588, "lr": 9.12365230141538e-05}
{"train_loss": 0.2730913758277893, "global_step": 52340, "epoch": 588, "lr": 9.123619515525386e-05}
{"train_loss": 0.30626219511032104, "global_step": 52341, "epoch": 588, "lr": 9.123586729081019e-05}
{"train_loss": 0.2544744610786438, "global_step": 52342, "epoch": 588, "lr": 9.123553942082287e-05}
{"train_loss": 0.21490631997585297, "global_step": 52343, "epoch": 588, "lr": 9.123521154529193e-05}
{"train_loss": 0.25493955612182617, "global_step": 52344, "epoch": 588, "lr": 9.123488366421742e-05}
{"train_loss": 0.2547328472137451, "global_step": 52345, "epoch": 588, "lr": 9.123455577759939e-05}
{"train_loss": 0.2516058683395386, "global_step": 52346, "epoch": 588, "lr": 9.123422788543786e-05}
{"train_loss": 0.2880726754665375, "global_step": 52347, "epoch": 588, "lr": 9.12338999877329e-05}
{"train_loss": 0.3477735221385956, "global_step": 52348, "epoch": 588, "lr": 9.123357208448453e-05}
{"train_loss": 0.3988151252269745, "global_step": 52349, "epoch": 588, "lr": 9.123324417569283e-05}
{"train_loss": 0.29213982820510864, "global_step": 52350, "epoch": 588, "lr": 9.123291626135781e-05}
{"train_loss": 0.28708815574645996, "global_step": 52351, "epoch": 588, "lr": 9.123258834147954e-05}
{"train_loss": 0.2877095639705658, "global_step": 52352, "epoch": 588, "lr": 9.123226041605803e-05}
{"train_loss": 0.3594297170639038, "global_step": 52353, "epoch": 588, "lr": 9.123193248509337e-05}
{"train_loss": 0.2986791133880615, "global_step": 52354, "epoch": 588, "lr": 9.123160454858557e-05}
{"train_loss": 0.33513328433036804, "global_step": 52355, "epoch": 588, "lr": 9.123127660653468e-05}
{"train_loss": 0.4564163088798523, "global_step": 52356, "epoch": 588, "lr": 9.123094865894076e-05}
{"train_loss": 0.26213976740837097, "global_step": 52357, "epoch": 588, "lr": 9.123062070580382e-05}
{"train_loss": 0.2246110886335373, "global_step": 52358, "epoch": 588, "lr": 9.123029274712394e-05}
{"train_loss": 0.31955429911613464, "global_step": 52359, "epoch": 588, "lr": 9.122996478290113e-05}
{"train_loss": 0.2859765589237213, "global_step": 52360, "epoch": 588, "lr": 9.122963681313547e-05}
{"train_loss": 0.2718830108642578, "global_step": 52361, "epoch": 588, "lr": 9.122930883782699e-05}
{"train_loss": 0.3989683985710144, "global_step": 52362, "epoch": 588, "lr": 9.122898085697573e-05}
{"train_loss": 0.28799307346343994, "global_step": 52363, "epoch": 588, "lr": 9.122865287058174e-05}
{"train_loss": 0.24469348788261414, "global_step": 52364, "epoch": 588, "lr": 9.122832487864505e-05}
{"train_loss": 0.2683508098125458, "global_step": 52365, "epoch": 588, "lr": 9.122799688116572e-05}
{"train_loss": 0.3719295561313629, "global_step": 52366, "epoch": 588, "lr": 9.122766887814378e-05}
{"train_loss": 0.30682456493377686, "global_step": 52367, "epoch": 588, "lr": 9.122734086957929e-05}
{"train_loss": 0.29622042179107666, "global_step": 52368, "epoch": 588, "lr": 9.122701285547229e-05}
{"train_loss": 0.22055920958518982, "global_step": 52369, "epoch": 588, "lr": 9.122668483582281e-05}
{"train_loss": 0.19210831820964813, "global_step": 52370, "epoch": 588, "lr": 9.122635681063092e-05}
{"train_loss": 0.2191382348537445, "global_step": 52371, "epoch": 588, "lr": 9.122602877989665e-05}
{"train_loss": 0.22062598168849945, "global_step": 52372, "epoch": 588, "lr": 9.122570074362003e-05}
{"train_loss": 0.2778642177581787, "global_step": 52373, "epoch": 588, "lr": 9.122537270180111e-05}
{"train_loss": 0.31502875685691833, "global_step": 52374, "epoch": 588, "lr": 9.122504465443994e-05}
{"train_loss": 0.40103310346603394, "global_step": 52375, "epoch": 588, "lr": 9.12247166015366e-05}
{"train_loss": 0.3788659870624542, "global_step": 52376, "epoch": 588, "lr": 9.122438854309107e-05}
{"train_loss": 0.2507631480693817, "global_step": 52377, "epoch": 588, "lr": 9.122406047910342e-05}
{"train_loss": 0.2785625159740448, "global_step": 52378, "epoch": 588, "lr": 9.122373240957371e-05}
{"train_loss": 0.3409023880958557, "global_step": 52379, "epoch": 588, "lr": 9.122340433450197e-05}
{"train_loss": 0.2809971868991852, "global_step": 52380, "epoch": 588, "lr": 9.122307625388824e-05}
{"train_loss": 0.2947056293487549, "global_step": 52381, "epoch": 588, "lr": 9.122274816773259e-05}
{"train_loss": 0.2731976807117462, "global_step": 52382, "epoch": 588, "lr": 9.122242007603502e-05}
{"train_loss": 0.2765464186668396, "global_step": 52383, "epoch": 588, "lr": 9.122209197879561e-05}
{"train_loss": 0.22585655748844147, "global_step": 52384, "epoch": 588, "lr": 9.122176387601438e-05}
{"train_loss": 0.35257846117019653, "global_step": 52385, "epoch": 588, "lr": 9.12214357676914e-05}
{"train_loss": 0.2739589214324951, "global_step": 52386, "epoch": 588, "lr": 9.12211076538267e-05}
{"train_loss": 0.20370937883853912, "global_step": 52387, "epoch": 588, "lr": 9.122077953442032e-05}
{"train_loss": 0.27934566140174866, "global_step": 52388, "epoch": 588, "lr": 9.122045140947231e-05}
{"train_loss": 0.3811754286289215, "global_step": 52389, "epoch": 588, "lr": 9.122012327898271e-05}
{"train_loss": 0.3556828200817108, "global_step": 52390, "epoch": 588, "lr": 9.121979514295156e-05}
{"train_loss": 0.31633174419403076, "global_step": 52391, "epoch": 588, "lr": 9.121946700137891e-05}
{"train_loss": 0.25184157490730286, "global_step": 52392, "epoch": 588, "lr": 9.121913885426483e-05}
{"train_loss": 0.2715263068675995, "global_step": 52393, "epoch": 588, "lr": 9.121881070160931e-05}
{"train_loss": 0.26265063881874084, "global_step": 52394, "epoch": 588, "lr": 9.121848254341243e-05}
{"train_loss": 0.27112701535224915, "global_step": 52395, "epoch": 588, "lr": 9.121815437967425e-05}
{"train_loss": 0.2237682044506073, "global_step": 52396, "epoch": 588, "lr": 9.121782621039477e-05}
{"train_loss": 0.16614210605621338, "global_step": 52397, "epoch": 588, "lr": 9.121749803557406e-05}
{"train_loss": 0.3029831647872925, "global_step": 52398, "epoch": 588, "lr": 9.121716985521216e-05}
{"train_loss": 0.3274368643760681, "global_step": 52399, "epoch": 588, "lr": 9.121684166930911e-05}
{"train_loss": 0.34176403284072876, "global_step": 52400, "epoch": 588, "lr": 9.121651347786498e-05}
{"train_loss": 0.3678993284702301, "global_step": 52401, "epoch": 588, "lr": 9.121618528087977e-05}
{"train_loss": 0.3093607425689697, "global_step": 52402, "epoch": 588, "lr": 9.121585707835353e-05}
{"train_loss": 0.3050342798233032, "global_step": 52403, "epoch": 588, "lr": 9.121552887028635e-05}
{"train_loss": 0.359051913022995, "global_step": 52404, "epoch": 588, "lr": 9.121520065667823e-05}
{"train_loss": 0.35461220145225525, "global_step": 52405, "epoch": 588, "lr": 9.121487243752924e-05}
{"train_loss": 0.3074323534965515, "global_step": 52406, "epoch": 588, "lr": 9.121454421283941e-05}
{"train_loss": 0.31240859627723694, "global_step": 52407, "epoch": 588, "lr": 9.121421598260878e-05}
{"train_loss": 0.30699422955513, "global_step": 52408, "epoch": 588, "lr": 9.12138877468374e-05}
{"train_loss": 0.3207727074623108, "global_step": 52409, "epoch": 588, "lr": 9.121355950552532e-05}
{"train_loss": 0.25675323605537415, "global_step": 52410, "epoch": 588, "lr": 9.121323125867258e-05}
{"train_loss": 0.19588036835193634, "global_step": 52411, "epoch": 588, "lr": 9.121290300627923e-05}
{"train_loss": 0.2665969431400299, "global_step": 52412, "epoch": 588, "lr": 9.121257474834529e-05}
{"train_loss": 0.27748772501945496, "global_step": 52413, "epoch": 588, "lr": 9.121224648487084e-05}
{"train_loss": 0.3399878740310669, "global_step": 52414, "epoch": 588, "lr": 9.121191821585588e-05}
{"train_loss": 0.24227268993854523, "global_step": 52415, "epoch": 588, "lr": 9.121158994130051e-05}
{"train_loss": 0.22400836646556854, "global_step": 52416, "epoch": 588, "lr": 9.121126166120473e-05}
{"train_loss": 0.23747128248214722, "global_step": 52417, "epoch": 588, "lr": 9.121093337556861e-05}
{"train_loss": 0.30856239795684814, "global_step": 52418, "epoch": 588, "lr": 9.121060508439217e-05}
{"train_loss": 0.25505587458610535, "global_step": 52419, "epoch": 588, "lr": 9.121027678767546e-05}
{"train_loss": 0.2878583644548159, "global_step": 52420, "epoch": 588, "lr": 9.120994848541856e-05, "val_loss": 3.1913905143737793}
{"train_loss": 0.301142156124115, "global_step": 52421, "epoch": 589, "lr": 9.120962017762144e-05}
{"train_loss": 0.21658846735954285, "global_step": 52422, "epoch": 589, "lr": 9.120929186428422e-05}
{"train_loss": 0.3535291850566864, "global_step": 52423, "epoch": 589, "lr": 9.12089635454069e-05}
{"train_loss": 0.2705058157444, "global_step": 52424, "epoch": 589, "lr": 9.120863522098955e-05}
{"train_loss": 0.2948461174964905, "global_step": 52425, "epoch": 589, "lr": 9.12083068910322e-05}
{"train_loss": 0.26155152916908264, "global_step": 52426, "epoch": 589, "lr": 9.120797855553487e-05}
{"train_loss": 0.2940307557582855, "global_step": 52427, "epoch": 589, "lr": 9.120765021449765e-05}
{"train_loss": 0.29005011916160583, "global_step": 52428, "epoch": 589, "lr": 9.120732186792056e-05}
{"train_loss": 0.23203997313976288, "global_step": 52429, "epoch": 589, "lr": 9.120699351580366e-05}
{"train_loss": 0.24299529194831848, "global_step": 52430, "epoch": 589, "lr": 9.120666515814698e-05}
{"train_loss": 0.24445214867591858, "global_step": 52431, "epoch": 589, "lr": 9.120633679495054e-05}
{"train_loss": 0.2715502083301544, "global_step": 52432, "epoch": 589, "lr": 9.120600842621443e-05}
{"train_loss": 0.24331507086753845, "global_step": 52433, "epoch": 589, "lr": 9.120568005193868e-05}
{"train_loss": 0.21231330931186676, "global_step": 52434, "epoch": 589, "lr": 9.120535167212331e-05}
{"train_loss": 0.3135935068130493, "global_step": 52435, "epoch": 589, "lr": 9.12050232867684e-05}
{"train_loss": 0.3054059147834778, "global_step": 52436, "epoch": 589, "lr": 9.120469489587397e-05}
{"train_loss": 0.3044223189353943, "global_step": 52437, "epoch": 589, "lr": 9.120436649944008e-05}
{"train_loss": 0.264608234167099, "global_step": 52438, "epoch": 589, "lr": 9.120403809746675e-05}
{"train_loss": 0.28790420293807983, "global_step": 52439, "epoch": 589, "lr": 9.120370968995405e-05}
{"train_loss": 0.28437158465385437, "global_step": 52440, "epoch": 589, "lr": 9.1203381276902e-05}
{"train_loss": 0.23628737032413483, "global_step": 52441, "epoch": 589, "lr": 9.120305285831068e-05}
{"train_loss": 0.24002565443515778, "global_step": 52442, "epoch": 589, "lr": 9.12027244341801e-05}
{"train_loss": 0.324373334646225, "global_step": 52443, "epoch": 589, "lr": 9.120239600451031e-05}
{"train_loss": 0.2552623450756073, "global_step": 52444, "epoch": 589, "lr": 9.120206756930136e-05}
{"train_loss": 0.2518640160560608, "global_step": 52445, "epoch": 589, "lr": 9.12017391285533e-05}
{"train_loss": 0.1746440827846527, "global_step": 52446, "epoch": 589, "lr": 9.120141068226617e-05}
{"train_loss": 0.3365110158920288, "global_step": 52447, "epoch": 589, "lr": 9.120108223044001e-05}
{"train_loss": 0.16558657586574554, "global_step": 52448, "epoch": 589, "lr": 9.120075377307486e-05}
{"train_loss": 0.19811072945594788, "global_step": 52449, "epoch": 589, "lr": 9.120042531017078e-05}
{"train_loss": 0.318355917930603, "global_step": 52450, "epoch": 589, "lr": 9.120009684172781e-05}
{"train_loss": 0.25805819034576416, "global_step": 52451, "epoch": 589, "lr": 9.119976836774598e-05}
{"train_loss": 0.25811082124710083, "global_step": 52452, "epoch": 589, "lr": 9.119943988822535e-05}
{"train_loss": 0.22537000477313995, "global_step": 52453, "epoch": 589, "lr": 9.119911140316595e-05}
{"train_loss": 0.23519933223724365, "global_step": 52454, "epoch": 589, "lr": 9.119878291256783e-05}
{"train_loss": 0.3777036666870117, "global_step": 52455, "epoch": 589, "lr": 9.119845441643104e-05}
{"train_loss": 0.2719005048274994, "global_step": 52456, "epoch": 589, "lr": 9.119812591475562e-05}
{"train_loss": 0.2791095972061157, "global_step": 52457, "epoch": 589, "lr": 9.119779740754162e-05}
{"train_loss": 0.19979465007781982, "global_step": 52458, "epoch": 589, "lr": 9.119746889478906e-05}
{"train_loss": 0.2622522711753845, "global_step": 52459, "epoch": 589, "lr": 9.119714037649802e-05}
{"train_loss": 0.18528114259243011, "global_step": 52460, "epoch": 589, "lr": 9.119681185266852e-05}
{"train_loss": 0.19089607894420624, "global_step": 52461, "epoch": 589, "lr": 9.119648332330062e-05}
{"train_loss": 0.1816246211528778, "global_step": 52462, "epoch": 589, "lr": 9.119615478839434e-05}
{"train_loss": 0.3004331588745117, "global_step": 52463, "epoch": 589, "lr": 9.119582624794975e-05}
{"train_loss": 0.30547797679901123, "global_step": 52464, "epoch": 589, "lr": 9.119549770196688e-05}
{"train_loss": 0.26875919103622437, "global_step": 52465, "epoch": 589, "lr": 9.119516915044576e-05}
{"train_loss": 0.2301596999168396, "global_step": 52466, "epoch": 589, "lr": 9.119484059338648e-05}
{"train_loss": 0.22735072672367096, "global_step": 52467, "epoch": 589, "lr": 9.119451203078903e-05}
{"train_loss": 0.20278958976268768, "global_step": 52468, "epoch": 589, "lr": 9.11941834626535e-05}
{"train_loss": 0.3066861629486084, "global_step": 52469, "epoch": 589, "lr": 9.119385488897991e-05}
{"train_loss": 0.1681157350540161, "global_step": 52470, "epoch": 589, "lr": 9.119352630976832e-05}
{"train_loss": 0.23059888184070587, "global_step": 52471, "epoch": 589, "lr": 9.119319772501874e-05}
{"train_loss": 0.23179009556770325, "global_step": 52472, "epoch": 589, "lr": 9.119286913473124e-05}
{"train_loss": 0.213497593998909, "global_step": 52473, "epoch": 589, "lr": 9.119254053890586e-05}
{"train_loss": 0.3240724503993988, "global_step": 52474, "epoch": 589, "lr": 9.119221193754266e-05}
{"train_loss": 0.26768171787261963, "global_step": 52475, "epoch": 589, "lr": 9.119188333064165e-05}
{"train_loss": 0.207956001162529, "global_step": 52476, "epoch": 589, "lr": 9.11915547182029e-05}
{"train_loss": 0.24698887765407562, "global_step": 52477, "epoch": 589, "lr": 9.119122610022647e-05}
{"train_loss": 0.2523329257965088, "global_step": 52478, "epoch": 589, "lr": 9.119089747671237e-05}
{"train_loss": 0.2719886600971222, "global_step": 52479, "epoch": 589, "lr": 9.119056884766064e-05}
{"train_loss": 0.2219039648771286, "global_step": 52480, "epoch": 589, "lr": 9.119024021307134e-05}
{"train_loss": 0.27081310749053955, "global_step": 52481, "epoch": 589, "lr": 9.118991157294453e-05}
{"train_loss": 0.29435858130455017, "global_step": 52482, "epoch": 589, "lr": 9.118958292728024e-05}
{"train_loss": 0.21471470594406128, "global_step": 52483, "epoch": 589, "lr": 9.118925427607849e-05}
{"train_loss": 0.26613593101501465, "global_step": 52484, "epoch": 589, "lr": 9.118892561933936e-05}
{"train_loss": 0.3605808615684509, "global_step": 52485, "epoch": 589, "lr": 9.118859695706289e-05}
{"train_loss": 0.2553948163986206, "global_step": 52486, "epoch": 589, "lr": 9.118826828924912e-05}
{"train_loss": 0.2780906558036804, "global_step": 52487, "epoch": 589, "lr": 9.118793961589806e-05}
{"train_loss": 0.18913839757442474, "global_step": 52488, "epoch": 589, "lr": 9.118761093700981e-05}
{"train_loss": 0.37637871503829956, "global_step": 52489, "epoch": 589, "lr": 9.118728225258438e-05}
{"train_loss": 0.2879054844379425, "global_step": 52490, "epoch": 589, "lr": 9.118695356262181e-05}
{"train_loss": 0.37375688552856445, "global_step": 52491, "epoch": 589, "lr": 9.118662486712217e-05}
{"train_loss": 0.29304102063179016, "global_step": 52492, "epoch": 589, "lr": 9.118629616608549e-05}
{"train_loss": 0.28240659832954407, "global_step": 52493, "epoch": 589, "lr": 9.11859674595118e-05}
{"train_loss": 0.18357743322849274, "global_step": 52494, "epoch": 589, "lr": 9.118563874740116e-05}
{"train_loss": 0.29381975531578064, "global_step": 52495, "epoch": 589, "lr": 9.118531002975363e-05}
{"train_loss": 0.28456705808639526, "global_step": 52496, "epoch": 589, "lr": 9.118498130656923e-05}
{"train_loss": 0.33845603466033936, "global_step": 52497, "epoch": 589, "lr": 9.1184652577848e-05}
{"train_loss": 0.25046372413635254, "global_step": 52498, "epoch": 589, "lr": 9.118432384359002e-05}
{"train_loss": 0.3137960731983185, "global_step": 52499, "epoch": 589, "lr": 9.118399510379529e-05}
{"train_loss": 0.2810641825199127, "global_step": 52500, "epoch": 589, "lr": 9.118366635846388e-05}
{"train_loss": 0.25533175468444824, "global_step": 52501, "epoch": 589, "lr": 9.118333760759583e-05}
{"train_loss": 0.23686200380325317, "global_step": 52502, "epoch": 589, "lr": 9.118300885119118e-05}
{"train_loss": 0.2684023380279541, "global_step": 52503, "epoch": 589, "lr": 9.118268008924997e-05}
{"train_loss": 0.38234829902648926, "global_step": 52504, "epoch": 589, "lr": 9.118235132177226e-05}
{"train_loss": 0.3518150746822357, "global_step": 52505, "epoch": 589, "lr": 9.118202254875809e-05}
{"train_loss": 0.2912675142288208, "global_step": 52506, "epoch": 589, "lr": 9.118169377020748e-05}
{"train_loss": 0.30126217007637024, "global_step": 52507, "epoch": 589, "lr": 9.118136498612052e-05}
{"train_loss": 0.2643719017505646, "global_step": 52508, "epoch": 589, "lr": 9.118103619649721e-05}
{"train_loss": 0.2670315435428298, "global_step": 52509, "epoch": 589, "lr": 9.11807074013376e-05, "val_loss": 3.1061270236968994}
{"train_loss": 0.264274537563324, "global_step": 52510, "epoch": 590, "lr": 9.118037860064175e-05}
{"train_loss": 0.32180649042129517, "global_step": 52511, "epoch": 590, "lr": 9.11800497944097e-05}
{"train_loss": 0.24997924268245697, "global_step": 52512, "epoch": 590, "lr": 9.117972098264151e-05}
{"train_loss": 0.36082980036735535, "global_step": 52513, "epoch": 590, "lr": 9.117939216533721e-05}
{"train_loss": 0.41454583406448364, "global_step": 52514, "epoch": 590, "lr": 9.117906334249685e-05}
{"train_loss": 0.37414464354515076, "global_step": 52515, "epoch": 590, "lr": 9.117873451412044e-05}
{"train_loss": 0.31625062227249146, "global_step": 52516, "epoch": 590, "lr": 9.117840568020805e-05}
{"train_loss": 0.2978731095790863, "global_step": 52517, "epoch": 590, "lr": 9.117807684075973e-05}
{"train_loss": 0.3216984272003174, "global_step": 52518, "epoch": 590, "lr": 9.117774799577555e-05}
{"train_loss": 0.31353995203971863, "global_step": 52519, "epoch": 590, "lr": 9.117741914525549e-05}
{"train_loss": 0.26582708954811096, "global_step": 52520, "epoch": 590, "lr": 9.117709028919963e-05}
{"train_loss": 0.1825501173734665, "global_step": 52521, "epoch": 590, "lr": 9.117676142760803e-05}
{"train_loss": 0.2718271017074585, "global_step": 52522, "epoch": 590, "lr": 9.117643256048069e-05}
{"train_loss": 0.3071509301662445, "global_step": 52523, "epoch": 590, "lr": 9.11761036878177e-05}
{"train_loss": 0.2091689109802246, "global_step": 52524, "epoch": 590, "lr": 9.117577480961909e-05}
{"train_loss": 0.349311888217926, "global_step": 52525, "epoch": 590, "lr": 9.117544592588489e-05}
{"train_loss": 0.2257702648639679, "global_step": 52526, "epoch": 590, "lr": 9.117511703661515e-05}
{"train_loss": 0.30150991678237915, "global_step": 52527, "epoch": 590, "lr": 9.117478814180992e-05}
{"train_loss": 0.24516822397708893, "global_step": 52528, "epoch": 590, "lr": 9.117445924146924e-05}
{"train_loss": 0.25752902030944824, "global_step": 52529, "epoch": 590, "lr": 9.117413033559317e-05}
{"train_loss": 0.39546409249305725, "global_step": 52530, "epoch": 590, "lr": 9.117380142418171e-05}
{"train_loss": 0.2519224286079407, "global_step": 52531, "epoch": 590, "lr": 9.117347250723496e-05}
{"train_loss": 0.37738311290740967, "global_step": 52532, "epoch": 590, "lr": 9.117314358475292e-05}
{"train_loss": 0.23374725878238678, "global_step": 52533, "epoch": 590, "lr": 9.117281465673568e-05}
{"train_loss": 0.23017661273479462, "global_step": 52534, "epoch": 590, "lr": 9.117248572318325e-05}
{"train_loss": 0.2596816122531891, "global_step": 52535, "epoch": 590, "lr": 9.117215678409565e-05}
{"train_loss": 0.18380889296531677, "global_step": 52536, "epoch": 590, "lr": 9.117182783947298e-05}
{"train_loss": 0.26538655161857605, "global_step": 52537, "epoch": 590, "lr": 9.117149888931526e-05}
{"train_loss": 0.24288755655288696, "global_step": 52538, "epoch": 590, "lr": 9.117116993362254e-05}
{"train_loss": 0.25679534673690796, "global_step": 52539, "epoch": 590, "lr": 9.117084097239483e-05}
{"train_loss": 0.31217125058174133, "global_step": 52540, "epoch": 590, "lr": 9.117051200563222e-05}
{"train_loss": 0.30795833468437195, "global_step": 52541, "epoch": 590, "lr": 9.117018303333474e-05}
{"train_loss": 0.3417758047580719, "global_step": 52542, "epoch": 590, "lr": 9.116985405550243e-05}
{"train_loss": 0.26165756583213806, "global_step": 52543, "epoch": 590, "lr": 9.116952507213534e-05}
{"train_loss": 0.25994354486465454, "global_step": 52544, "epoch": 590, "lr": 9.116919608323349e-05}
{"train_loss": 0.3553279638290405, "global_step": 52545, "epoch": 590, "lr": 9.116886708879697e-05}
{"train_loss": 0.24831831455230713, "global_step": 52546, "epoch": 590, "lr": 9.116853808882577e-05}
{"train_loss": 0.27178752422332764, "global_step": 52547, "epoch": 590, "lr": 9.116820908331997e-05}
{"train_loss": 0.3047335743904114, "global_step": 52548, "epoch": 590, "lr": 9.116788007227962e-05}
{"train_loss": 0.23248949646949768, "global_step": 52549, "epoch": 590, "lr": 9.116755105570473e-05}
{"train_loss": 0.2534714937210083, "global_step": 52550, "epoch": 590, "lr": 9.116722203359538e-05}
{"train_loss": 0.3013674020767212, "global_step": 52551, "epoch": 590, "lr": 9.11668930059516e-05}
{"train_loss": 0.28817108273506165, "global_step": 52552, "epoch": 590, "lr": 9.116656397277343e-05}
{"train_loss": 0.38194793462753296, "global_step": 52553, "epoch": 590, "lr": 9.116623493406091e-05}
{"train_loss": 0.25531837344169617, "global_step": 52554, "epoch": 590, "lr": 9.11659058898141e-05}
{"train_loss": 0.28133928775787354, "global_step": 52555, "epoch": 590, "lr": 9.116557684003303e-05}
{"train_loss": 0.28038597106933594, "global_step": 52556, "epoch": 590, "lr": 9.116524778471775e-05}
{"train_loss": 0.20384886860847473, "global_step": 52557, "epoch": 590, "lr": 9.11649187238683e-05}
{"train_loss": 0.3777559995651245, "global_step": 52558, "epoch": 590, "lr": 9.116458965748473e-05}
{"train_loss": 0.30659839510917664, "global_step": 52559, "epoch": 590, "lr": 9.116426058556709e-05}
{"train_loss": 0.19358494877815247, "global_step": 52560, "epoch": 590, "lr": 9.116393150811542e-05}
{"train_loss": 0.2892408072948456, "global_step": 52561, "epoch": 590, "lr": 9.116360242512975e-05}
{"train_loss": 0.24906907975673676, "global_step": 52562, "epoch": 590, "lr": 9.116327333661015e-05}
{"train_loss": 0.2476189136505127, "global_step": 52563, "epoch": 590, "lr": 9.116294424255664e-05}
{"train_loss": 0.3159656524658203, "global_step": 52564, "epoch": 590, "lr": 9.116261514296927e-05}
{"train_loss": 0.30566293001174927, "global_step": 52565, "epoch": 590, "lr": 9.116228603784809e-05}
{"train_loss": 0.3369307816028595, "global_step": 52566, "epoch": 590, "lr": 9.116195692719314e-05}
{"train_loss": 0.20144182443618774, "global_step": 52567, "epoch": 590, "lr": 9.116162781100448e-05}
{"train_loss": 0.2556663453578949, "global_step": 52568, "epoch": 590, "lr": 9.116129868928212e-05}
{"train_loss": 0.328011691570282, "global_step": 52569, "epoch": 590, "lr": 9.116096956202614e-05}
{"train_loss": 0.27770501375198364, "global_step": 52570, "epoch": 590, "lr": 9.116064042923657e-05}
{"train_loss": 0.19580078125, "global_step": 52571, "epoch": 590, "lr": 9.116031129091344e-05}
{"train_loss": 0.32392051815986633, "global_step": 52572, "epoch": 590, "lr": 9.115998214705684e-05}
{"train_loss": 0.2668927013874054, "global_step": 52573, "epoch": 590, "lr": 9.115965299766675e-05}
{"train_loss": 0.30180662870407104, "global_step": 52574, "epoch": 590, "lr": 9.115932384274324e-05}
{"train_loss": 0.1729656457901001, "global_step": 52575, "epoch": 590, "lr": 9.115899468228638e-05}
{"train_loss": 0.2595643401145935, "global_step": 52576, "epoch": 590, "lr": 9.115866551629619e-05}
{"train_loss": 0.27179640531539917, "global_step": 52577, "epoch": 590, "lr": 9.115833634477272e-05}
{"train_loss": 0.29559099674224854, "global_step": 52578, "epoch": 590, "lr": 9.1158007167716e-05}
{"train_loss": 0.38385963439941406, "global_step": 52579, "epoch": 590, "lr": 9.115767798512611e-05}
{"train_loss": 0.2633706033229828, "global_step": 52580, "epoch": 590, "lr": 9.115734879700308e-05}
{"train_loss": 0.31379401683807373, "global_step": 52581, "epoch": 590, "lr": 9.115701960334692e-05}
{"train_loss": 0.255241334438324, "global_step": 52582, "epoch": 590, "lr": 9.11566904041577e-05}
{"train_loss": 0.35596519708633423, "global_step": 52583, "epoch": 590, "lr": 9.11563611994355e-05}
{"train_loss": 0.22830933332443237, "global_step": 52584, "epoch": 590, "lr": 9.115603198918029e-05}
{"train_loss": 0.25836649537086487, "global_step": 52585, "epoch": 590, "lr": 9.115570277339216e-05}
{"train_loss": 0.28781434893608093, "global_step": 52586, "epoch": 590, "lr": 9.115537355207116e-05}
{"train_loss": 0.2390938401222229, "global_step": 52587, "epoch": 590, "lr": 9.115504432521732e-05}
{"train_loss": 0.2650148868560791, "global_step": 52588, "epoch": 590, "lr": 9.115471509283067e-05}
{"train_loss": 0.22735586762428284, "global_step": 52589, "epoch": 590, "lr": 9.115438585491127e-05}
{"train_loss": 0.28323906660079956, "global_step": 52590, "epoch": 590, "lr": 9.115405661145919e-05}
{"train_loss": 0.32410117983818054, "global_step": 52591, "epoch": 590, "lr": 9.115372736247442e-05}
{"train_loss": 0.34515196084976196, "global_step": 52592, "epoch": 590, "lr": 9.115339810795707e-05}
{"train_loss": 0.36379504203796387, "global_step": 52593, "epoch": 590, "lr": 9.115306884790711e-05}
{"train_loss": 0.34002944827079773, "global_step": 52594, "epoch": 590, "lr": 9.115273958232465e-05}
{"train_loss": 0.25252214074134827, "global_step": 52595, "epoch": 590, "lr": 9.115241031120966e-05}
{"train_loss": 0.24334639310836792, "global_step": 52596, "epoch": 590, "lr": 9.115208103456228e-05}
{"train_loss": 0.3154156506061554, "global_step": 52597, "epoch": 590, "lr": 9.115175175238248e-05}
{"train_loss": 0.28305945202205957, "global_step": 52598, "epoch": 590, "lr": 9.115142246467034e-05, "val_loss": 3.05078387260437, "train_action_mse_error": 21.719894409179688}
{"train_loss": 0.27424386143684387, "global_step": 52599, "epoch": 591, "lr": 9.115109317142588e-05}
{"train_loss": 0.27117854356765747, "global_step": 52600, "epoch": 591, "lr": 9.115076387264917e-05}
{"train_loss": 0.28491759300231934, "global_step": 52601, "epoch": 591, "lr": 9.115043456834024e-05}
{"train_loss": 0.27354833483695984, "global_step": 52602, "epoch": 591, "lr": 9.115010525849912e-05}
{"train_loss": 0.2781783938407898, "global_step": 52603, "epoch": 591, "lr": 9.114977594312588e-05}
{"train_loss": 0.2867409586906433, "global_step": 52604, "epoch": 591, "lr": 9.114944662222056e-05}
{"train_loss": 0.26267683506011963, "global_step": 52605, "epoch": 591, "lr": 9.11491172957832e-05}
{"train_loss": 0.30272728204727173, "global_step": 52606, "epoch": 591, "lr": 9.114878796381383e-05}
{"train_loss": 0.28320249915122986, "global_step": 52607, "epoch": 591, "lr": 9.114845862631252e-05}
{"train_loss": 0.2845833897590637, "global_step": 52608, "epoch": 591, "lr": 9.114812928327929e-05}
{"train_loss": 0.2380635291337967, "global_step": 52609, "epoch": 591, "lr": 9.11477999347142e-05}
{"train_loss": 0.19505172967910767, "global_step": 52610, "epoch": 591, "lr": 9.114747058061729e-05}
{"train_loss": 0.21160587668418884, "global_step": 52611, "epoch": 591, "lr": 9.11471412209886e-05}
{"train_loss": 0.22942110896110535, "global_step": 52612, "epoch": 591, "lr": 9.114681185582818e-05}
{"train_loss": 0.35447874665260315, "global_step": 52613, "epoch": 591, "lr": 9.114648248513608e-05}
{"train_loss": 0.3458898365497589, "global_step": 52614, "epoch": 591, "lr": 9.114615310891232e-05}
{"train_loss": 0.3104228675365448, "global_step": 52615, "epoch": 591, "lr": 9.114582372715698e-05}
{"train_loss": 0.22419221699237823, "global_step": 52616, "epoch": 591, "lr": 9.114549433987007e-05}
{"train_loss": 0.27480849623680115, "global_step": 52617, "epoch": 591, "lr": 9.114516494705167e-05}
{"train_loss": 0.27408725023269653, "global_step": 52618, "epoch": 591, "lr": 9.11448355487018e-05}
{"train_loss": 0.18966393172740936, "global_step": 52619, "epoch": 591, "lr": 9.11445061448205e-05}
{"train_loss": 0.34591221809387207, "global_step": 52620, "epoch": 591, "lr": 9.114417673540781e-05}
{"train_loss": 0.22341963648796082, "global_step": 52621, "epoch": 591, "lr": 9.114384732046381e-05}
{"train_loss": 0.26151588559150696, "global_step": 52622, "epoch": 591, "lr": 9.114351789998851e-05}
{"train_loss": 0.2716454863548279, "global_step": 52623, "epoch": 591, "lr": 9.114318847398196e-05}
{"train_loss": 0.24131399393081665, "global_step": 52624, "epoch": 591, "lr": 9.114285904244422e-05}
{"train_loss": 0.2686576545238495, "global_step": 52625, "epoch": 591, "lr": 9.114252960537533e-05}
{"train_loss": 0.2672770619392395, "global_step": 52626, "epoch": 591, "lr": 9.114220016277531e-05}
{"train_loss": 0.23128370940685272, "global_step": 52627, "epoch": 591, "lr": 9.114187071464424e-05}
{"train_loss": 0.32796511054039, "global_step": 52628, "epoch": 591, "lr": 9.114154126098213e-05}
{"train_loss": 0.2767767012119293, "global_step": 52629, "epoch": 591, "lr": 9.114121180178905e-05}
{"train_loss": 0.21063083410263062, "global_step": 52630, "epoch": 591, "lr": 9.114088233706504e-05}
{"train_loss": 0.17593586444854736, "global_step": 52631, "epoch": 591, "lr": 9.114055286681014e-05}
{"train_loss": 0.21848760545253754, "global_step": 52632, "epoch": 591, "lr": 9.11402233910244e-05}
{"train_loss": 0.2824360132217407, "global_step": 52633, "epoch": 591, "lr": 9.113989390970783e-05}
{"train_loss": 0.31615015864372253, "global_step": 52634, "epoch": 591, "lr": 9.113956442286052e-05}
{"train_loss": 0.2708367109298706, "global_step": 52635, "epoch": 591, "lr": 9.113923493048251e-05}
{"train_loss": 0.31492558121681213, "global_step": 52636, "epoch": 591, "lr": 9.113890543257381e-05}
{"train_loss": 0.274320513010025, "global_step": 52637, "epoch": 591, "lr": 9.11385759291345e-05}
{"train_loss": 0.19758659601211548, "global_step": 52638, "epoch": 591, "lr": 9.11382464201646e-05}
{"train_loss": 0.3344137668609619, "global_step": 52639, "epoch": 591, "lr": 9.113791690566417e-05}
{"train_loss": 0.21177341043949127, "global_step": 52640, "epoch": 591, "lr": 9.113758738563325e-05}
{"train_loss": 0.22989657521247864, "global_step": 52641, "epoch": 591, "lr": 9.113725786007188e-05}
{"train_loss": 0.24099832773208618, "global_step": 52642, "epoch": 591, "lr": 9.11369283289801e-05}
{"train_loss": 0.3286563754081726, "global_step": 52643, "epoch": 591, "lr": 9.113659879235796e-05}
{"train_loss": 0.3205486238002777, "global_step": 52644, "epoch": 591, "lr": 9.113626925020552e-05}
{"train_loss": 0.26901301741600037, "global_step": 52645, "epoch": 591, "lr": 9.113593970252281e-05}
{"train_loss": 0.21768975257873535, "global_step": 52646, "epoch": 591, "lr": 9.113561014930987e-05}
{"train_loss": 0.2869188189506531, "global_step": 52647, "epoch": 591, "lr": 9.113528059056675e-05}
{"train_loss": 0.2532930374145508, "global_step": 52648, "epoch": 591, "lr": 9.113495102629348e-05}
{"train_loss": 0.28984272480010986, "global_step": 52649, "epoch": 591, "lr": 9.113462145649012e-05}
{"train_loss": 0.20815318822860718, "global_step": 52650, "epoch": 591, "lr": 9.113429188115672e-05}
{"train_loss": 0.3010292649269104, "global_step": 52651, "epoch": 591, "lr": 9.113396230029332e-05}
{"train_loss": 0.1930496245622635, "global_step": 52652, "epoch": 591, "lr": 9.113363271389995e-05}
{"train_loss": 0.2654125690460205, "global_step": 52653, "epoch": 591, "lr": 9.113330312197668e-05}
{"train_loss": 0.24692818522453308, "global_step": 52654, "epoch": 591, "lr": 9.113297352452352e-05}
{"train_loss": 0.18219947814941406, "global_step": 52655, "epoch": 591, "lr": 9.113264392154054e-05}
{"train_loss": 0.2004391849040985, "global_step": 52656, "epoch": 591, "lr": 9.113231431302777e-05}
{"train_loss": 0.24852189421653748, "global_step": 52657, "epoch": 591, "lr": 9.113198469898528e-05}
{"train_loss": 0.2738756537437439, "global_step": 52658, "epoch": 591, "lr": 9.113165507941308e-05}
{"train_loss": 0.27924197912216187, "global_step": 52659, "epoch": 591, "lr": 9.113132545431124e-05}
{"train_loss": 0.2950909733772278, "global_step": 52660, "epoch": 591, "lr": 9.113099582367979e-05}
{"train_loss": 0.21777160465717316, "global_step": 52661, "epoch": 591, "lr": 9.113066618751878e-05}
{"train_loss": 0.2726956009864807, "global_step": 52662, "epoch": 591, "lr": 9.113033654582825e-05}
{"train_loss": 0.3111720681190491, "global_step": 52663, "epoch": 591, "lr": 9.113000689860825e-05}
{"train_loss": 0.30927374958992004, "global_step": 52664, "epoch": 591, "lr": 9.112967724585883e-05}
{"train_loss": 0.2580692172050476, "global_step": 52665, "epoch": 591, "lr": 9.112934758758e-05}
{"train_loss": 0.24594202637672424, "global_step": 52666, "epoch": 591, "lr": 9.112901792377187e-05}
{"train_loss": 0.22910398244857788, "global_step": 52667, "epoch": 591, "lr": 9.112868825443441e-05}
{"train_loss": 0.3452112674713135, "global_step": 52668, "epoch": 591, "lr": 9.112835857956772e-05}
{"train_loss": 0.33831238746643066, "global_step": 52669, "epoch": 591, "lr": 9.112802889917182e-05}
{"train_loss": 0.3173975646495819, "global_step": 52670, "epoch": 591, "lr": 9.112769921324675e-05}
{"train_loss": 0.3101372718811035, "global_step": 52671, "epoch": 591, "lr": 9.112736952179256e-05}
{"train_loss": 0.22572647035121918, "global_step": 52672, "epoch": 591, "lr": 9.11270398248093e-05}
{"train_loss": 0.2165822982788086, "global_step": 52673, "epoch": 591, "lr": 9.1126710122297e-05}
{"train_loss": 0.2619549632072449, "global_step": 52674, "epoch": 591, "lr": 9.112638041425574e-05}
{"train_loss": 0.17293362319469452, "global_step": 52675, "epoch": 591, "lr": 9.112605070068553e-05}
{"train_loss": 0.1838299036026001, "global_step": 52676, "epoch": 591, "lr": 9.112572098158642e-05}
{"train_loss": 0.29341602325439453, "global_step": 52677, "epoch": 591, "lr": 9.112539125695846e-05}
{"train_loss": 0.25814154744148254, "global_step": 52678, "epoch": 591, "lr": 9.112506152680169e-05}
{"train_loss": 0.2512558400630951, "global_step": 52679, "epoch": 591, "lr": 9.112473179111615e-05}
{"train_loss": 0.2819446325302124, "global_step": 52680, "epoch": 591, "lr": 9.112440204990192e-05}
{"train_loss": 0.3446550667285919, "global_step": 52681, "epoch": 591, "lr": 9.112407230315899e-05}
{"train_loss": 0.2675970494747162, "global_step": 52682, "epoch": 591, "lr": 9.112374255088744e-05}
{"train_loss": 0.21969887614250183, "global_step": 52683, "epoch": 591, "lr": 9.11234127930873e-05}
{"train_loss": 0.22479131817817688, "global_step": 52684, "epoch": 591, "lr": 9.112308302975861e-05}
{"train_loss": 0.32977408170700073, "global_step": 52685, "epoch": 591, "lr": 9.112275326090143e-05}
{"train_loss": 0.34414932131767273, "global_step": 52686, "epoch": 591, "lr": 9.112242348651579e-05}
{"train_loss": 0.2657722814029522, "global_step": 52687, "epoch": 591, "lr": 9.112209370660176e-05, "val_loss": 3.1372711658477783}
{"train_loss": 0.24807989597320557, "global_step": 52688, "epoch": 592, "lr": 9.112176392115935e-05}
{"train_loss": 0.3248730003833771, "global_step": 52689, "epoch": 592, "lr": 9.112143413018862e-05}
{"train_loss": 0.31915828585624695, "global_step": 52690, "epoch": 592, "lr": 9.112110433368962e-05}
{"train_loss": 0.2386656105518341, "global_step": 52691, "epoch": 592, "lr": 9.11207745316624e-05}
{"train_loss": 0.29325070977211, "global_step": 52692, "epoch": 592, "lr": 9.112044472410697e-05}
{"train_loss": 0.2534273862838745, "global_step": 52693, "epoch": 592, "lr": 9.11201149110234e-05}
{"train_loss": 0.245836079120636, "global_step": 52694, "epoch": 592, "lr": 9.111978509241174e-05}
{"train_loss": 0.3091222941875458, "global_step": 52695, "epoch": 592, "lr": 9.111945526827204e-05}
{"train_loss": 0.31470271944999695, "global_step": 52696, "epoch": 592, "lr": 9.111912543860431e-05}
{"train_loss": 0.30504709482192993, "global_step": 52697, "epoch": 592, "lr": 9.111879560340861e-05}
{"train_loss": 0.3239530622959137, "global_step": 52698, "epoch": 592, "lr": 9.111846576268502e-05}
{"train_loss": 0.2503112554550171, "global_step": 52699, "epoch": 592, "lr": 9.111813591643352e-05}
{"train_loss": 0.2198277860879898, "global_step": 52700, "epoch": 592, "lr": 9.111780606465421e-05}
{"train_loss": 0.2400376945734024, "global_step": 52701, "epoch": 592, "lr": 9.11174762073471e-05}
{"train_loss": 0.22633379697799683, "global_step": 52702, "epoch": 592, "lr": 9.111714634451225e-05}
{"train_loss": 0.3088032007217407, "global_step": 52703, "epoch": 592, "lr": 9.11168164761497e-05}
{"train_loss": 0.19932083785533905, "global_step": 52704, "epoch": 592, "lr": 9.111648660225949e-05}
{"train_loss": 0.2885173559188843, "global_step": 52705, "epoch": 592, "lr": 9.111615672284168e-05}
{"train_loss": 0.3295467793941498, "global_step": 52706, "epoch": 592, "lr": 9.11158268378963e-05}
{"train_loss": 0.18528275191783905, "global_step": 52707, "epoch": 592, "lr": 9.111549694742338e-05}
{"train_loss": 0.2808278203010559, "global_step": 52708, "epoch": 592, "lr": 9.111516705142301e-05}
{"train_loss": 0.29830247163772583, "global_step": 52709, "epoch": 592, "lr": 9.111483714989519e-05}
{"train_loss": 0.16637541353702545, "global_step": 52710, "epoch": 592, "lr": 9.111450724284e-05}
{"train_loss": 0.17698881030082703, "global_step": 52711, "epoch": 592, "lr": 9.111417733025745e-05}
{"train_loss": 0.30898988246917725, "global_step": 52712, "epoch": 592, "lr": 9.111384741214761e-05}
{"train_loss": 0.2913860082626343, "global_step": 52713, "epoch": 592, "lr": 9.111351748851049e-05}
{"train_loss": 0.27397266030311584, "global_step": 52714, "epoch": 592, "lr": 9.11131875593462e-05}
{"train_loss": 0.28613924980163574, "global_step": 52715, "epoch": 592, "lr": 9.111285762465472e-05}
{"train_loss": 0.2924480736255646, "global_step": 52716, "epoch": 592, "lr": 9.11125276844361e-05}
{"train_loss": 0.18671205639839172, "global_step": 52717, "epoch": 592, "lr": 9.111219773869042e-05}
{"train_loss": 0.2384939342737198, "global_step": 52718, "epoch": 592, "lr": 9.111186778741772e-05}
{"train_loss": 0.19468072056770325, "global_step": 52719, "epoch": 592, "lr": 9.111153783061802e-05}
{"train_loss": 0.34532105922698975, "global_step": 52720, "epoch": 592, "lr": 9.111120786829135e-05}
{"train_loss": 0.29877033829689026, "global_step": 52721, "epoch": 592, "lr": 9.11108779004378e-05}
{"train_loss": 0.24424557387828827, "global_step": 52722, "epoch": 592, "lr": 9.11105479270574e-05}
{"train_loss": 0.32083114981651306, "global_step": 52723, "epoch": 592, "lr": 9.111021794815018e-05}
{"train_loss": 0.26150256395339966, "global_step": 52724, "epoch": 592, "lr": 9.110988796371619e-05}
{"train_loss": 0.2151489406824112, "global_step": 52725, "epoch": 592, "lr": 9.110955797375548e-05}
{"train_loss": 0.24391087889671326, "global_step": 52726, "epoch": 592, "lr": 9.110922797826808e-05}
{"train_loss": 0.22444196045398712, "global_step": 52727, "epoch": 592, "lr": 9.110889797725406e-05}
{"train_loss": 0.25494521856307983, "global_step": 52728, "epoch": 592, "lr": 9.110856797071345e-05}
{"train_loss": 0.28768742084503174, "global_step": 52729, "epoch": 592, "lr": 9.110823795864629e-05}
{"train_loss": 0.2782261073589325, "global_step": 52730, "epoch": 592, "lr": 9.110790794105262e-05}
{"train_loss": 0.2610672414302826, "global_step": 52731, "epoch": 592, "lr": 9.110757791793251e-05}
{"train_loss": 0.1924320012331009, "global_step": 52732, "epoch": 592, "lr": 9.110724788928597e-05}
{"train_loss": 0.31520140171051025, "global_step": 52733, "epoch": 592, "lr": 9.110691785511307e-05}
{"train_loss": 0.2835586369037628, "global_step": 52734, "epoch": 592, "lr": 9.110658781541386e-05}
{"train_loss": 0.23663616180419922, "global_step": 52735, "epoch": 592, "lr": 9.110625777018835e-05}
{"train_loss": 0.2812487781047821, "global_step": 52736, "epoch": 592, "lr": 9.110592771943661e-05}
{"train_loss": 0.34970611333847046, "global_step": 52737, "epoch": 592, "lr": 9.110559766315867e-05}
{"train_loss": 0.2757120728492737, "global_step": 52738, "epoch": 592, "lr": 9.11052676013546e-05}
{"train_loss": 0.3079889714717865, "global_step": 52739, "epoch": 592, "lr": 9.110493753402441e-05}
{"train_loss": 0.3474634885787964, "global_step": 52740, "epoch": 592, "lr": 9.110460746116817e-05}
{"train_loss": 0.22935901582241058, "global_step": 52741, "epoch": 592, "lr": 9.110427738278593e-05}
{"train_loss": 0.23341435194015503, "global_step": 52742, "epoch": 592, "lr": 9.110394729887771e-05}
{"train_loss": 0.30285608768463135, "global_step": 52743, "epoch": 592, "lr": 9.110361720944357e-05}
{"train_loss": 0.26920345425605774, "global_step": 52744, "epoch": 592, "lr": 9.110328711448355e-05}
{"train_loss": 0.26155030727386475, "global_step": 52745, "epoch": 592, "lr": 9.110295701399768e-05}
{"train_loss": 0.25103217363357544, "global_step": 52746, "epoch": 592, "lr": 9.110262690798604e-05}
{"train_loss": 0.24866566061973572, "global_step": 52747, "epoch": 592, "lr": 9.110229679644863e-05}
{"train_loss": 0.21751749515533447, "global_step": 52748, "epoch": 592, "lr": 9.110196667938553e-05}
{"train_loss": 0.17779338359832764, "global_step": 52749, "epoch": 592, "lr": 9.110163655679677e-05}
{"train_loss": 0.344279021024704, "global_step": 52750, "epoch": 592, "lr": 9.11013064286824e-05}
{"train_loss": 0.25127097964286804, "global_step": 52751, "epoch": 592, "lr": 9.110097629504244e-05}
{"train_loss": 0.2840518057346344, "global_step": 52752, "epoch": 592, "lr": 9.110064615587697e-05}
{"train_loss": 0.2013903260231018, "global_step": 52753, "epoch": 592, "lr": 9.110031601118603e-05}
{"train_loss": 0.2997162640094757, "global_step": 52754, "epoch": 592, "lr": 9.109998586096965e-05}
{"train_loss": 0.36547669768333435, "global_step": 52755, "epoch": 592, "lr": 9.109965570522786e-05}
{"train_loss": 0.2310880720615387, "global_step": 52756, "epoch": 592, "lr": 9.109932554396073e-05}
{"train_loss": 0.2669042646884918, "global_step": 52757, "epoch": 592, "lr": 9.10989953771683e-05}
{"train_loss": 0.13497097790241241, "global_step": 52758, "epoch": 592, "lr": 9.109866520485061e-05}
{"train_loss": 0.3323326110839844, "global_step": 52759, "epoch": 592, "lr": 9.109833502700771e-05}
{"train_loss": 0.29546812176704407, "global_step": 52760, "epoch": 592, "lr": 9.109800484363962e-05}
{"train_loss": 0.35806456208229065, "global_step": 52761, "epoch": 592, "lr": 9.109767465474643e-05}
{"train_loss": 0.3545389771461487, "global_step": 52762, "epoch": 592, "lr": 9.109734446032815e-05}
{"train_loss": 0.281074196100235, "global_step": 52763, "epoch": 592, "lr": 9.109701426038484e-05}
{"train_loss": 0.34047412872314453, "global_step": 52764, "epoch": 592, "lr": 9.109668405491651e-05}
{"train_loss": 0.31670042872428894, "global_step": 52765, "epoch": 592, "lr": 9.109635384392326e-05}
{"train_loss": 0.2300216406583786, "global_step": 52766, "epoch": 592, "lr": 9.10960236274051e-05}
{"train_loss": 0.30294638872146606, "global_step": 52767, "epoch": 592, "lr": 9.109569340536207e-05}
{"train_loss": 0.30589696764945984, "global_step": 52768, "epoch": 592, "lr": 9.109536317779423e-05}
{"train_loss": 0.26365041732788086, "global_step": 52769, "epoch": 592, "lr": 9.109503294470163e-05}
{"train_loss": 0.24787268042564392, "global_step": 52770, "epoch": 592, "lr": 9.109470270608429e-05}
{"train_loss": 0.2344948798418045, "global_step": 52771, "epoch": 592, "lr": 9.109437246194226e-05}
{"train_loss": 0.2692352533340454, "global_step": 52772, "epoch": 592, "lr": 9.109404221227562e-05}
{"train_loss": 0.3041266202926636, "global_step": 52773, "epoch": 592, "lr": 9.109371195708437e-05}
{"train_loss": 0.29667985439300537, "global_step": 52774, "epoch": 592, "lr": 9.109338169636858e-05}
{"train_loss": 0.343588262796402, "global_step": 52775, "epoch": 592, "lr": 9.109305143012827e-05}
{"train_loss": 0.27205007966984524, "global_step": 52776, "epoch": 592, "lr": 9.109272115836351e-05, "val_loss": 3.0559990406036377}
{"train_loss": 0.27246731519699097, "global_step": 52777, "epoch": 593, "lr": 9.109239088107434e-05}
{"train_loss": 0.24933625757694244, "global_step": 52778, "epoch": 593, "lr": 9.109206059826079e-05}
{"train_loss": 0.30170994997024536, "global_step": 52779, "epoch": 593, "lr": 9.10917303099229e-05}
{"train_loss": 0.22365806996822357, "global_step": 52780, "epoch": 593, "lr": 9.109140001606076e-05}
{"train_loss": 0.27092546224594116, "global_step": 52781, "epoch": 593, "lr": 9.109106971667435e-05}
{"train_loss": 0.3379260003566742, "global_step": 52782, "epoch": 593, "lr": 9.109073941176377e-05}
{"train_loss": 0.2313353270292282, "global_step": 52783, "epoch": 593, "lr": 9.109040910132905e-05}
{"train_loss": 0.2412382960319519, "global_step": 52784, "epoch": 593, "lr": 9.109007878537019e-05}
{"train_loss": 0.30204707384109497, "global_step": 52785, "epoch": 593, "lr": 9.10897484638873e-05}
{"train_loss": 0.26543208956718445, "global_step": 52786, "epoch": 593, "lr": 9.108941813688037e-05}
{"train_loss": 0.22977501153945923, "global_step": 52787, "epoch": 593, "lr": 9.108908780434948e-05}
{"train_loss": 0.19040489196777344, "global_step": 52788, "epoch": 593, "lr": 9.108875746629468e-05}
{"train_loss": 0.24392320215702057, "global_step": 52789, "epoch": 593, "lr": 9.108842712271597e-05}
{"train_loss": 0.23391222953796387, "global_step": 52790, "epoch": 593, "lr": 9.108809677361343e-05}
{"train_loss": 0.26515817642211914, "global_step": 52791, "epoch": 593, "lr": 9.10877664189871e-05}
{"train_loss": 0.20749914646148682, "global_step": 52792, "epoch": 593, "lr": 9.108743605883703e-05}
{"train_loss": 0.2856903374195099, "global_step": 52793, "epoch": 593, "lr": 9.108710569316324e-05}
{"train_loss": 0.16261352598667145, "global_step": 52794, "epoch": 593, "lr": 9.108677532196578e-05}
{"train_loss": 0.19233191013336182, "global_step": 52795, "epoch": 593, "lr": 9.108644494524473e-05}
{"train_loss": 0.20549368858337402, "global_step": 52796, "epoch": 593, "lr": 9.108611456300009e-05}
{"train_loss": 0.2113274186849594, "global_step": 52797, "epoch": 593, "lr": 9.108578417523194e-05}
{"train_loss": 0.2121204137802124, "global_step": 52798, "epoch": 593, "lr": 9.108545378194029e-05}
{"train_loss": 0.2374913990497589, "global_step": 52799, "epoch": 593, "lr": 9.108512338312521e-05}
{"train_loss": 0.221418097615242, "global_step": 52800, "epoch": 593, "lr": 9.108479297878673e-05}
{"train_loss": 0.22674210369586945, "global_step": 52801, "epoch": 593, "lr": 9.10844625689249e-05}
{"train_loss": 0.2887827157974243, "global_step": 52802, "epoch": 593, "lr": 9.108413215353976e-05}
{"train_loss": 0.22694844007492065, "global_step": 52803, "epoch": 593, "lr": 9.108380173263138e-05}
{"train_loss": 0.2533505856990814, "global_step": 52804, "epoch": 593, "lr": 9.108347130619976e-05}
{"train_loss": 0.22924691438674927, "global_step": 52805, "epoch": 593, "lr": 9.108314087424499e-05}
{"train_loss": 0.27410295605659485, "global_step": 52806, "epoch": 593, "lr": 9.108281043676707e-05}
{"train_loss": 0.2576388120651245, "global_step": 52807, "epoch": 593, "lr": 9.10824799937661e-05}
{"train_loss": 0.23704174160957336, "global_step": 52808, "epoch": 593, "lr": 9.108214954524205e-05}
{"train_loss": 0.24874386191368103, "global_step": 52809, "epoch": 593, "lr": 9.108181909119505e-05}
{"train_loss": 0.2376246154308319, "global_step": 52810, "epoch": 593, "lr": 9.108148863162506e-05}
{"train_loss": 0.2359766960144043, "global_step": 52811, "epoch": 593, "lr": 9.108115816653218e-05}
{"train_loss": 0.4314049780368805, "global_step": 52812, "epoch": 593, "lr": 9.108082769591643e-05}
{"train_loss": 0.31888091564178467, "global_step": 52813, "epoch": 593, "lr": 9.108049721977789e-05}
{"train_loss": 0.18587641417980194, "global_step": 52814, "epoch": 593, "lr": 9.108016673811656e-05}
{"train_loss": 0.23090486228466034, "global_step": 52815, "epoch": 593, "lr": 9.10798362509325e-05}
{"train_loss": 0.30158939957618713, "global_step": 52816, "epoch": 593, "lr": 9.107950575822576e-05}
{"train_loss": 0.24518483877182007, "global_step": 52817, "epoch": 593, "lr": 9.107917525999639e-05}
{"train_loss": 0.2754819393157959, "global_step": 52818, "epoch": 593, "lr": 9.107884475624439e-05}
{"train_loss": 0.23175574839115143, "global_step": 52819, "epoch": 593, "lr": 9.107851424696989e-05}
{"train_loss": 0.19836100935935974, "global_step": 52820, "epoch": 593, "lr": 9.107818373217285e-05}
{"train_loss": 0.21199709177017212, "global_step": 52821, "epoch": 593, "lr": 9.107785321185336e-05}
{"train_loss": 0.21636147797107697, "global_step": 52822, "epoch": 593, "lr": 9.107752268601145e-05}
{"train_loss": 0.2901677191257477, "global_step": 52823, "epoch": 593, "lr": 9.107719215464717e-05}
{"train_loss": 0.2508488595485687, "global_step": 52824, "epoch": 593, "lr": 9.107686161776057e-05}
{"train_loss": 0.2388591766357422, "global_step": 52825, "epoch": 593, "lr": 9.107653107535168e-05}
{"train_loss": 0.30134955048561096, "global_step": 52826, "epoch": 593, "lr": 9.107620052742054e-05}
{"train_loss": 0.3323820233345032, "global_step": 52827, "epoch": 593, "lr": 9.107586997396722e-05}
{"train_loss": 0.31463950872421265, "global_step": 52828, "epoch": 593, "lr": 9.107553941499175e-05}
{"train_loss": 0.29163628816604614, "global_step": 52829, "epoch": 593, "lr": 9.107520885049417e-05}
{"train_loss": 0.22316263616085052, "global_step": 52830, "epoch": 593, "lr": 9.107487828047453e-05}
{"train_loss": 0.17420917749404907, "global_step": 52831, "epoch": 593, "lr": 9.107454770493287e-05}
{"train_loss": 0.28315427899360657, "global_step": 52832, "epoch": 593, "lr": 9.107421712386923e-05}
{"train_loss": 0.32259249687194824, "global_step": 52833, "epoch": 593, "lr": 9.107388653728368e-05}
{"train_loss": 0.38331660628318787, "global_step": 52834, "epoch": 593, "lr": 9.107355594517623e-05}
{"train_loss": 0.23853081464767456, "global_step": 52835, "epoch": 593, "lr": 9.107322534754697e-05}
{"train_loss": 0.2501128613948822, "global_step": 52836, "epoch": 593, "lr": 9.107289474439588e-05}
{"train_loss": 0.27965328097343445, "global_step": 52837, "epoch": 593, "lr": 9.107256413572306e-05}
{"train_loss": 0.16724753379821777, "global_step": 52838, "epoch": 593, "lr": 9.107223352152853e-05}
{"train_loss": 0.1814703494310379, "global_step": 52839, "epoch": 593, "lr": 9.107190290181234e-05}
{"train_loss": 0.21732649207115173, "global_step": 52840, "epoch": 593, "lr": 9.107157227657453e-05}
{"train_loss": 0.2641458809375763, "global_step": 52841, "epoch": 593, "lr": 9.107124164581516e-05}
{"train_loss": 0.19221587479114532, "global_step": 52842, "epoch": 593, "lr": 9.107091100953424e-05}
{"train_loss": 0.3308773934841156, "global_step": 52843, "epoch": 593, "lr": 9.107058036773185e-05}
{"train_loss": 0.21857993304729462, "global_step": 52844, "epoch": 593, "lr": 9.107024972040801e-05}
{"train_loss": 0.2378719300031662, "global_step": 52845, "epoch": 593, "lr": 9.10699190675628e-05}
{"train_loss": 0.3303202688694, "global_step": 52846, "epoch": 593, "lr": 9.106958840919621e-05}
{"train_loss": 0.19645003974437714, "global_step": 52847, "epoch": 593, "lr": 9.106925774530834e-05}
{"train_loss": 0.22292351722717285, "global_step": 52848, "epoch": 593, "lr": 9.106892707589919e-05}
{"train_loss": 0.29106390476226807, "global_step": 52849, "epoch": 593, "lr": 9.106859640096883e-05}
{"train_loss": 0.24675661325454712, "global_step": 52850, "epoch": 593, "lr": 9.10682657205173e-05}
{"train_loss": 0.3368605673313141, "global_step": 52851, "epoch": 593, "lr": 9.106793503454465e-05}
{"train_loss": 0.2010684758424759, "global_step": 52852, "epoch": 593, "lr": 9.10676043430509e-05}
{"train_loss": 0.20531094074249268, "global_step": 52853, "epoch": 593, "lr": 9.106727364603612e-05}
{"train_loss": 0.20110078155994415, "global_step": 52854, "epoch": 593, "lr": 9.106694294350034e-05}
{"train_loss": 0.1878214031457901, "global_step": 52855, "epoch": 593, "lr": 9.106661223544362e-05}
{"train_loss": 0.1991947889328003, "global_step": 52856, "epoch": 593, "lr": 9.106628152186598e-05}
{"train_loss": 0.23319238424301147, "global_step": 52857, "epoch": 593, "lr": 9.10659508027675e-05}
{"train_loss": 0.19445431232452393, "global_step": 52858, "epoch": 593, "lr": 9.106562007814818e-05}
{"train_loss": 0.23131269216537476, "global_step": 52859, "epoch": 593, "lr": 9.106528934800811e-05}
{"train_loss": 0.33903318643569946, "global_step": 52860, "epoch": 593, "lr": 9.106495861234729e-05}
{"train_loss": 0.2743350565433502, "global_step": 52861, "epoch": 593, "lr": 9.10646278711658e-05}
{"train_loss": 0.32217174768447876, "global_step": 52862, "epoch": 593, "lr": 9.106429712446367e-05}
{"train_loss": 0.2741076648235321, "global_step": 52863, "epoch": 593, "lr": 9.106396637224096e-05}
{"train_loss": 0.3218911290168762, "global_step": 52864, "epoch": 593, "lr": 9.106363561449767e-05}
{"train_loss": 0.2515125462178434, "global_step": 52865, "epoch": 593, "lr": 9.106330485123389e-05, "val_loss": 3.226041316986084}
{"train_loss": 0.1480129063129425, "global_step": 52866, "epoch": 594, "lr": 9.106297408244966e-05}
{"train_loss": 0.2316579669713974, "global_step": 52867, "epoch": 594, "lr": 9.1062643308145e-05}
{"train_loss": 0.3174572288990021, "global_step": 52868, "epoch": 594, "lr": 9.106231252831997e-05}
{"train_loss": 0.3130689859390259, "global_step": 52869, "epoch": 594, "lr": 9.106198174297462e-05}
{"train_loss": 0.27510860562324524, "global_step": 52870, "epoch": 594, "lr": 9.106165095210898e-05}
{"train_loss": 0.3432801067829132, "global_step": 52871, "epoch": 594, "lr": 9.10613201557231e-05}
{"train_loss": 0.25231173634529114, "global_step": 52872, "epoch": 594, "lr": 9.106098935381702e-05}
{"train_loss": 0.24883055686950684, "global_step": 52873, "epoch": 594, "lr": 9.106065854639081e-05}
{"train_loss": 0.24645569920539856, "global_step": 52874, "epoch": 594, "lr": 9.106032773344448e-05}
{"train_loss": 0.22859878838062286, "global_step": 52875, "epoch": 594, "lr": 9.105999691497808e-05}
{"train_loss": 0.2759626805782318, "global_step": 52876, "epoch": 594, "lr": 9.105966609099169e-05}
{"train_loss": 0.24952886998653412, "global_step": 52877, "epoch": 594, "lr": 9.105933526148532e-05}
{"train_loss": 0.21201753616333008, "global_step": 52878, "epoch": 594, "lr": 9.105900442645902e-05}
{"train_loss": 0.2956241965293884, "global_step": 52879, "epoch": 594, "lr": 9.105867358591282e-05}
{"train_loss": 0.32014140486717224, "global_step": 52880, "epoch": 594, "lr": 9.105834273984682e-05}
{"train_loss": 0.2521134912967682, "global_step": 52881, "epoch": 594, "lr": 9.1058011888261e-05}
{"train_loss": 0.20493276417255402, "global_step": 52882, "epoch": 594, "lr": 9.105768103115541e-05}
{"train_loss": 0.2606965899467468, "global_step": 52883, "epoch": 594, "lr": 9.105735016853014e-05}
{"train_loss": 0.2349446415901184, "global_step": 52884, "epoch": 594, "lr": 9.105701930038522e-05}
{"train_loss": 0.23625588417053223, "global_step": 52885, "epoch": 594, "lr": 9.105668842672068e-05}
{"train_loss": 0.29253917932510376, "global_step": 52886, "epoch": 594, "lr": 9.105635754753656e-05}
{"train_loss": 0.32094207406044006, "global_step": 52887, "epoch": 594, "lr": 9.10560266628329e-05}
{"train_loss": 0.25540903210639954, "global_step": 52888, "epoch": 594, "lr": 9.105569577260979e-05}
{"train_loss": 0.18212847411632538, "global_step": 52889, "epoch": 594, "lr": 9.105536487686722e-05}
{"train_loss": 0.20093056559562683, "global_step": 52890, "epoch": 594, "lr": 9.105503397560526e-05}
{"train_loss": 0.31433361768722534, "global_step": 52891, "epoch": 594, "lr": 9.105470306882396e-05}
{"train_loss": 0.26204001903533936, "global_step": 52892, "epoch": 594, "lr": 9.105437215652334e-05}
{"train_loss": 0.23130665719509125, "global_step": 52893, "epoch": 594, "lr": 9.105404123870346e-05}
{"train_loss": 0.3182279169559479, "global_step": 52894, "epoch": 594, "lr": 9.105371031536439e-05}
{"train_loss": 0.17117100954055786, "global_step": 52895, "epoch": 594, "lr": 9.105337938650612e-05}
{"train_loss": 0.22238536179065704, "global_step": 52896, "epoch": 594, "lr": 9.105304845212874e-05}
{"train_loss": 0.2360721379518509, "global_step": 52897, "epoch": 594, "lr": 9.105271751223228e-05}
{"train_loss": 0.17885176837444305, "global_step": 52898, "epoch": 594, "lr": 9.105238656681676e-05}
{"train_loss": 0.1931290626525879, "global_step": 52899, "epoch": 594, "lr": 9.105205561588227e-05}
{"train_loss": 0.2762685716152191, "global_step": 52900, "epoch": 594, "lr": 9.105172465942881e-05}
{"train_loss": 0.24843648076057434, "global_step": 52901, "epoch": 594, "lr": 9.105139369745648e-05}
{"train_loss": 0.26010704040527344, "global_step": 52902, "epoch": 594, "lr": 9.105106272996526e-05}
{"train_loss": 0.23332926630973816, "global_step": 52903, "epoch": 594, "lr": 9.105073175695525e-05}
{"train_loss": 0.32289040088653564, "global_step": 52904, "epoch": 594, "lr": 9.105040077842644e-05}
{"train_loss": 0.2671218812465668, "global_step": 52905, "epoch": 594, "lr": 9.105006979437893e-05}
{"train_loss": 0.3155140280723572, "global_step": 52906, "epoch": 594, "lr": 9.104973880481273e-05}
{"train_loss": 0.23409588634967804, "global_step": 52907, "epoch": 594, "lr": 9.104940780972788e-05}
{"train_loss": 0.41741639375686646, "global_step": 52908, "epoch": 594, "lr": 9.104907680912445e-05}
{"train_loss": 0.26906800270080566, "global_step": 52909, "epoch": 594, "lr": 9.104874580300247e-05}
{"train_loss": 0.2273833006620407, "global_step": 52910, "epoch": 594, "lr": 9.1048414791362e-05}
{"train_loss": 0.311620831489563, "global_step": 52911, "epoch": 594, "lr": 9.104808377420304e-05}
{"train_loss": 0.22045648097991943, "global_step": 52912, "epoch": 594, "lr": 9.104775275152569e-05}
{"train_loss": 0.2798735499382019, "global_step": 52913, "epoch": 594, "lr": 9.104742172332995e-05}
{"train_loss": 0.2913129925727844, "global_step": 52914, "epoch": 594, "lr": 9.10470906896159e-05}
{"train_loss": 0.24382109940052032, "global_step": 52915, "epoch": 594, "lr": 9.104675965038357e-05}
{"train_loss": 0.2716193199157715, "global_step": 52916, "epoch": 594, "lr": 9.1046428605633e-05}
{"train_loss": 0.25155937671661377, "global_step": 52917, "epoch": 594, "lr": 9.104609755536424e-05}
{"train_loss": 0.21286419034004211, "global_step": 52918, "epoch": 594, "lr": 9.104576649957732e-05}
{"train_loss": 0.22256717085838318, "global_step": 52919, "epoch": 594, "lr": 9.10454354382723e-05}
{"train_loss": 0.2533871531486511, "global_step": 52920, "epoch": 594, "lr": 9.104510437144925e-05}
{"train_loss": 0.34250637888908386, "global_step": 52921, "epoch": 594, "lr": 9.104477329910814e-05}
{"train_loss": 0.3128824234008789, "global_step": 52922, "epoch": 594, "lr": 9.10444422212491e-05}
{"train_loss": 0.34062159061431885, "global_step": 52923, "epoch": 594, "lr": 9.104411113787212e-05}
{"train_loss": 0.22437605261802673, "global_step": 52924, "epoch": 594, "lr": 9.104378004897725e-05}
{"train_loss": 0.28227800130844116, "global_step": 52925, "epoch": 594, "lr": 9.104344895456455e-05}
{"train_loss": 0.2499949336051941, "global_step": 52926, "epoch": 594, "lr": 9.104311785463408e-05}
{"train_loss": 0.2778044044971466, "global_step": 52927, "epoch": 594, "lr": 9.104278674918583e-05}
{"train_loss": 0.28062722086906433, "global_step": 52928, "epoch": 594, "lr": 9.104245563821989e-05}
{"train_loss": 0.25021687150001526, "global_step": 52929, "epoch": 594, "lr": 9.10421245217363e-05}
{"train_loss": 0.31844785809516907, "global_step": 52930, "epoch": 594, "lr": 9.104179339973508e-05}
{"train_loss": 0.20317816734313965, "global_step": 52931, "epoch": 594, "lr": 9.104146227221631e-05}
{"train_loss": 0.23383623361587524, "global_step": 52932, "epoch": 594, "lr": 9.104113113918001e-05}
{"train_loss": 0.3242636024951935, "global_step": 52933, "epoch": 594, "lr": 9.104080000062622e-05}
{"train_loss": 0.2542148232460022, "global_step": 52934, "epoch": 594, "lr": 9.104046885655502e-05}
{"train_loss": 0.2678318917751312, "global_step": 52935, "epoch": 594, "lr": 9.10401377069664e-05}
{"train_loss": 0.2717457711696625, "global_step": 52936, "epoch": 594, "lr": 9.103980655186045e-05}
{"train_loss": 0.2360052913427353, "global_step": 52937, "epoch": 594, "lr": 9.103947539123719e-05}
{"train_loss": 0.3255060911178589, "global_step": 52938, "epoch": 594, "lr": 9.103914422509668e-05}
{"train_loss": 0.2690383195877075, "global_step": 52939, "epoch": 594, "lr": 9.103881305343895e-05}
{"train_loss": 0.17590191960334778, "global_step": 52940, "epoch": 594, "lr": 9.103848187626406e-05}
{"train_loss": 0.2920328378677368, "global_step": 52941, "epoch": 594, "lr": 9.103815069357205e-05}
{"train_loss": 0.2208755612373352, "global_step": 52942, "epoch": 594, "lr": 9.103781950536295e-05}
{"train_loss": 0.2585945427417755, "global_step": 52943, "epoch": 594, "lr": 9.103748831163682e-05}
{"train_loss": 0.37388160824775696, "global_step": 52944, "epoch": 594, "lr": 9.10371571123937e-05}
{"train_loss": 0.27172166109085083, "global_step": 52945, "epoch": 594, "lr": 9.103682590763363e-05}
{"train_loss": 0.23108968138694763, "global_step": 52946, "epoch": 594, "lr": 9.103649469735666e-05}
{"train_loss": 0.33367329835891724, "global_step": 52947, "epoch": 594, "lr": 9.103616348156283e-05}
{"train_loss": 0.24562884867191315, "global_step": 52948, "epoch": 594, "lr": 9.103583226025218e-05}
{"train_loss": 0.25938907265663147, "global_step": 52949, "epoch": 594, "lr": 9.103550103342478e-05}
{"train_loss": 0.3238641917705536, "global_step": 52950, "epoch": 594, "lr": 9.103516980108065e-05}
{"train_loss": 0.20855772495269775, "global_step": 52951, "epoch": 594, "lr": 9.103483856321985e-05}
{"train_loss": 0.2210042029619217, "global_step": 52952, "epoch": 594, "lr": 9.103450731984241e-05}
{"train_loss": 0.27523815631866455, "global_step": 52953, "epoch": 594, "lr": 9.103417607094837e-05}
{"train_loss": 0.2636679332242923, "global_step": 52954, "epoch": 594, "lr": 9.10338448165378e-05, "val_loss": 3.092069625854492}
{"train_loss": 0.22766023874282837, "global_step": 52955, "epoch": 595, "lr": 9.103351355661074e-05}
{"train_loss": 0.26912084221839905, "global_step": 52956, "epoch": 595, "lr": 9.10331822911672e-05}
{"train_loss": 0.27949339151382446, "global_step": 52957, "epoch": 595, "lr": 9.103285102020726e-05}
{"train_loss": 0.2503320574760437, "global_step": 52958, "epoch": 595, "lr": 9.103251974373094e-05}
{"train_loss": 0.2019822746515274, "global_step": 52959, "epoch": 595, "lr": 9.103218846173831e-05}
{"train_loss": 0.2722562551498413, "global_step": 52960, "epoch": 595, "lr": 9.103185717422941e-05}
{"train_loss": 0.3035048544406891, "global_step": 52961, "epoch": 595, "lr": 9.103152588120427e-05}
{"train_loss": 0.24447815120220184, "global_step": 52962, "epoch": 595, "lr": 9.103119458266293e-05}
{"train_loss": 0.3296527862548828, "global_step": 52963, "epoch": 595, "lr": 9.103086327860546e-05}
{"train_loss": 0.30956122279167175, "global_step": 52964, "epoch": 595, "lr": 9.10305319690319e-05}
{"train_loss": 0.3673650026321411, "global_step": 52965, "epoch": 595, "lr": 9.103020065394226e-05}
{"train_loss": 0.3123844861984253, "global_step": 52966, "epoch": 595, "lr": 9.102986933333661e-05}
{"train_loss": 0.2480297088623047, "global_step": 52967, "epoch": 595, "lr": 9.102953800721502e-05}
{"train_loss": 0.288390576839447, "global_step": 52968, "epoch": 595, "lr": 9.102920667557749e-05}
{"train_loss": 0.22839757800102234, "global_step": 52969, "epoch": 595, "lr": 9.10288753384241e-05}
{"train_loss": 0.2261257916688919, "global_step": 52970, "epoch": 595, "lr": 9.102854399575485e-05}
{"train_loss": 0.35287344455718994, "global_step": 52971, "epoch": 595, "lr": 9.102821264756984e-05}
{"train_loss": 0.23977293074131012, "global_step": 52972, "epoch": 595, "lr": 9.102788129386908e-05}
{"train_loss": 0.15099433064460754, "global_step": 52973, "epoch": 595, "lr": 9.10275499346526e-05}
{"train_loss": 0.3381526470184326, "global_step": 52974, "epoch": 595, "lr": 9.102721856992048e-05}
{"train_loss": 0.23533938825130463, "global_step": 52975, "epoch": 595, "lr": 9.102688719967276e-05}
{"train_loss": 0.20109321177005768, "global_step": 52976, "epoch": 595, "lr": 9.102655582390946e-05}
{"train_loss": 0.25068968534469604, "global_step": 52977, "epoch": 595, "lr": 9.102622444263065e-05}
{"train_loss": 0.21512474119663239, "global_step": 52978, "epoch": 595, "lr": 9.102589305583637e-05}
{"train_loss": 0.26149383187294006, "global_step": 52979, "epoch": 595, "lr": 9.102556166352665e-05}
{"train_loss": 0.3086671531200409, "global_step": 52980, "epoch": 595, "lr": 9.102523026570153e-05}
{"train_loss": 0.25144481658935547, "global_step": 52981, "epoch": 595, "lr": 9.102489886236109e-05}
{"train_loss": 0.19663478434085846, "global_step": 52982, "epoch": 595, "lr": 9.102456745350534e-05}
{"train_loss": 0.25472012162208557, "global_step": 52983, "epoch": 595, "lr": 9.102423603913434e-05}
{"train_loss": 0.2608000338077545, "global_step": 52984, "epoch": 595, "lr": 9.102390461924812e-05}
{"train_loss": 0.19186687469482422, "global_step": 52985, "epoch": 595, "lr": 9.102357319384676e-05}
{"train_loss": 0.21670138835906982, "global_step": 52986, "epoch": 595, "lr": 9.102324176293027e-05}
{"train_loss": 0.32086867094039917, "global_step": 52987, "epoch": 595, "lr": 9.10229103264987e-05}
{"train_loss": 0.2052599936723709, "global_step": 52988, "epoch": 595, "lr": 9.102257888455211e-05}
{"train_loss": 0.3550378978252411, "global_step": 52989, "epoch": 595, "lr": 9.102224743709051e-05}
{"train_loss": 0.23030772805213928, "global_step": 52990, "epoch": 595, "lr": 9.1021915984114e-05}
{"train_loss": 0.26186004281044006, "global_step": 52991, "epoch": 595, "lr": 9.102158452562257e-05}
{"train_loss": 0.24502263963222504, "global_step": 52992, "epoch": 595, "lr": 9.102125306161631e-05}
{"train_loss": 0.3819376826286316, "global_step": 52993, "epoch": 595, "lr": 9.102092159209522e-05}
{"train_loss": 0.2594398260116577, "global_step": 52994, "epoch": 595, "lr": 9.102059011705938e-05}
{"train_loss": 0.26359283924102783, "global_step": 52995, "epoch": 595, "lr": 9.102025863650882e-05}
{"train_loss": 0.22888174653053284, "global_step": 52996, "epoch": 595, "lr": 9.101992715044358e-05}
{"train_loss": 0.3284527361392975, "global_step": 52997, "epoch": 595, "lr": 9.101959565886372e-05}
{"train_loss": 0.21954292058944702, "global_step": 52998, "epoch": 595, "lr": 9.101926416176927e-05}
{"train_loss": 0.3616429269313812, "global_step": 52999, "epoch": 595, "lr": 9.101893265916029e-05}
{"train_loss": 0.2781408131122589, "global_step": 53000, "epoch": 595, "lr": 9.10186011510368e-05}
{"train_loss": 0.24115300178527832, "global_step": 53001, "epoch": 595, "lr": 9.101826963739885e-05}
{"train_loss": 0.2754763960838318, "global_step": 53002, "epoch": 595, "lr": 9.101793811824652e-05}
{"train_loss": 0.26047375798225403, "global_step": 53003, "epoch": 595, "lr": 9.101760659357981e-05}
{"train_loss": 0.2818034291267395, "global_step": 53004, "epoch": 595, "lr": 9.101727506339879e-05}
{"train_loss": 0.22779309749603271, "global_step": 53005, "epoch": 595, "lr": 9.10169435277035e-05}
{"train_loss": 0.27744507789611816, "global_step": 53006, "epoch": 595, "lr": 9.101661198649398e-05}
{"train_loss": 0.18148234486579895, "global_step": 53007, "epoch": 595, "lr": 9.101628043977028e-05}
{"train_loss": 0.33857351541519165, "global_step": 53008, "epoch": 595, "lr": 9.101594888753243e-05}
{"train_loss": 0.21065151691436768, "global_step": 53009, "epoch": 595, "lr": 9.101561732978048e-05}
{"train_loss": 0.2715322971343994, "global_step": 53010, "epoch": 595, "lr": 9.10152857665145e-05}
{"train_loss": 0.245056614279747, "global_step": 53011, "epoch": 595, "lr": 9.101495419773452e-05}
{"train_loss": 0.2225222885608673, "global_step": 53012, "epoch": 595, "lr": 9.101462262344056e-05}
{"train_loss": 0.290818452835083, "global_step": 53013, "epoch": 595, "lr": 9.101429104363268e-05}
{"train_loss": 0.3921014070510864, "global_step": 53014, "epoch": 595, "lr": 9.101395945831094e-05}
{"train_loss": 0.17544445395469666, "global_step": 53015, "epoch": 595, "lr": 9.101362786747537e-05}
{"train_loss": 0.19575384259223938, "global_step": 53016, "epoch": 595, "lr": 9.101329627112603e-05}
{"train_loss": 0.27810245752334595, "global_step": 53017, "epoch": 595, "lr": 9.101296466926293e-05}
{"train_loss": 0.3157062828540802, "global_step": 53018, "epoch": 595, "lr": 9.101263306188615e-05}
{"train_loss": 0.28337275981903076, "global_step": 53019, "epoch": 595, "lr": 9.10123014489957e-05}
{"train_loss": 0.28324636816978455, "global_step": 53020, "epoch": 595, "lr": 9.101196983059168e-05}
{"train_loss": 0.295349657535553, "global_step": 53021, "epoch": 595, "lr": 9.101163820667409e-05}
{"train_loss": 0.2786674201488495, "global_step": 53022, "epoch": 595, "lr": 9.101130657724298e-05}
{"train_loss": 0.2325003743171692, "global_step": 53023, "epoch": 595, "lr": 9.10109749422984e-05}
{"train_loss": 0.2695361375808716, "global_step": 53024, "epoch": 595, "lr": 9.10106433018404e-05}
{"train_loss": 0.293623685836792, "global_step": 53025, "epoch": 595, "lr": 9.1010311655869e-05}
{"train_loss": 0.17115385830402374, "global_step": 53026, "epoch": 595, "lr": 9.100998000438428e-05}
{"train_loss": 0.23255793750286102, "global_step": 53027, "epoch": 595, "lr": 9.100964834738626e-05}
{"train_loss": 0.20180551707744598, "global_step": 53028, "epoch": 595, "lr": 9.1009316684875e-05}
{"train_loss": 0.2758547067642212, "global_step": 53029, "epoch": 595, "lr": 9.100898501685054e-05}
{"train_loss": 0.3470260798931122, "global_step": 53030, "epoch": 595, "lr": 9.100865334331292e-05}
{"train_loss": 0.2523914873600006, "global_step": 53031, "epoch": 595, "lr": 9.100832166426217e-05}
{"train_loss": 0.3044964671134949, "global_step": 53032, "epoch": 595, "lr": 9.100798997969837e-05}
{"train_loss": 0.2485112100839615, "global_step": 53033, "epoch": 595, "lr": 9.100765828962154e-05}
{"train_loss": 0.22335144877433777, "global_step": 53034, "epoch": 595, "lr": 9.100732659403173e-05}
{"train_loss": 0.356732577085495, "global_step": 53035, "epoch": 595, "lr": 9.100699489292898e-05}
{"train_loss": 0.36029252409935, "global_step": 53036, "epoch": 595, "lr": 9.100666318631335e-05}
{"train_loss": 0.29614177346229553, "global_step": 53037, "epoch": 595, "lr": 9.100633147418485e-05}
{"train_loss": 0.3026394546031952, "global_step": 53038, "epoch": 595, "lr": 9.100599975654358e-05}
{"train_loss": 0.2190014272928238, "global_step": 53039, "epoch": 595, "lr": 9.100566803338952e-05}
{"train_loss": 0.23968227207660675, "global_step": 53040, "epoch": 595, "lr": 9.100533630472277e-05}
{"train_loss": 0.21270689368247986, "global_step": 53041, "epoch": 595, "lr": 9.100500457054335e-05}
{"train_loss": 0.25143224000930786, "global_step": 53042, "epoch": 595, "lr": 9.100467283085131e-05}
{"train_loss": 0.26496310029806713, "global_step": 53043, "epoch": 595, "lr": 9.100434108564668e-05, "val_loss": 3.260566473007202, "train_action_mse_error": 22.007692337036133}
{"train_loss": 0.23677390813827515, "global_step": 53044, "epoch": 596, "lr": 9.100400933492954e-05}
{"train_loss": 0.2602129280567169, "global_step": 53045, "epoch": 596, "lr": 9.100367757869989e-05}
{"train_loss": 0.26973825693130493, "global_step": 53046, "epoch": 596, "lr": 9.10033458169578e-05}
{"train_loss": 0.323742151260376, "global_step": 53047, "epoch": 596, "lr": 9.10030140497033e-05}
{"train_loss": 0.3649803102016449, "global_step": 53048, "epoch": 596, "lr": 9.100268227693647e-05}
{"train_loss": 0.280146986246109, "global_step": 53049, "epoch": 596, "lr": 9.100235049865731e-05}
{"train_loss": 0.27102261781692505, "global_step": 53050, "epoch": 596, "lr": 9.100201871486587e-05}
{"train_loss": 0.28116148710250854, "global_step": 53051, "epoch": 596, "lr": 9.100168692556224e-05}
{"train_loss": 0.22939980030059814, "global_step": 53052, "epoch": 596, "lr": 9.100135513074642e-05}
{"train_loss": 0.2471657693386078, "global_step": 53053, "epoch": 596, "lr": 9.100102333041847e-05}
{"train_loss": 0.2148931920528412, "global_step": 53054, "epoch": 596, "lr": 9.100069152457842e-05}
{"train_loss": 0.2700774073600769, "global_step": 53055, "epoch": 596, "lr": 9.100035971322633e-05}
{"train_loss": 0.21755744516849518, "global_step": 53056, "epoch": 596, "lr": 9.100002789636224e-05}
{"train_loss": 0.2159232795238495, "global_step": 53057, "epoch": 596, "lr": 9.099969607398621e-05}
{"train_loss": 0.23449738323688507, "global_step": 53058, "epoch": 596, "lr": 9.099936424609826e-05}
{"train_loss": 0.21469824016094208, "global_step": 53059, "epoch": 596, "lr": 9.099903241269845e-05}
{"train_loss": 0.22967471182346344, "global_step": 53060, "epoch": 596, "lr": 9.09987005737868e-05}
{"train_loss": 0.32674792408943176, "global_step": 53061, "epoch": 596, "lr": 9.09983687293634e-05}
{"train_loss": 0.3148818910121918, "global_step": 53062, "epoch": 596, "lr": 9.099803687942827e-05}
{"train_loss": 0.2977004647254944, "global_step": 53063, "epoch": 596, "lr": 9.099770502398144e-05}
{"train_loss": 0.2616124153137207, "global_step": 53064, "epoch": 596, "lr": 9.099737316302297e-05}
{"train_loss": 0.3257640600204468, "global_step": 53065, "epoch": 596, "lr": 9.099704129655291e-05}
{"train_loss": 0.3423532545566559, "global_step": 53066, "epoch": 596, "lr": 9.099670942457129e-05}
{"train_loss": 0.28441059589385986, "global_step": 53067, "epoch": 596, "lr": 9.099637754707818e-05}
{"train_loss": 0.24620197713375092, "global_step": 53068, "epoch": 596, "lr": 9.099604566407358e-05}
{"train_loss": 0.26810765266418457, "global_step": 53069, "epoch": 596, "lr": 9.099571377555757e-05}
{"train_loss": 0.2597844898700714, "global_step": 53070, "epoch": 596, "lr": 9.09953818815302e-05}
{"train_loss": 0.2844938635826111, "global_step": 53071, "epoch": 596, "lr": 9.099504998199149e-05}
{"train_loss": 0.3005291819572449, "global_step": 53072, "epoch": 596, "lr": 9.099471807694149e-05}
{"train_loss": 0.22046597301959991, "global_step": 53073, "epoch": 596, "lr": 9.099438616638025e-05}
{"train_loss": 0.2345724105834961, "global_step": 53074, "epoch": 596, "lr": 9.099405425030782e-05}
{"train_loss": 0.2939620018005371, "global_step": 53075, "epoch": 596, "lr": 9.099372232872423e-05}
{"train_loss": 0.2409171462059021, "global_step": 53076, "epoch": 596, "lr": 9.099339040162955e-05}
{"train_loss": 0.26143401861190796, "global_step": 53077, "epoch": 596, "lr": 9.09930584690238e-05}
{"train_loss": 0.2795528173446655, "global_step": 53078, "epoch": 596, "lr": 9.099272653090704e-05}
{"train_loss": 0.263721764087677, "global_step": 53079, "epoch": 596, "lr": 9.09923945872793e-05}
{"train_loss": 0.18415609002113342, "global_step": 53080, "epoch": 596, "lr": 9.099206263814063e-05}
{"train_loss": 0.2629321813583374, "global_step": 53081, "epoch": 596, "lr": 9.099173068349107e-05}
{"train_loss": 0.3036421239376068, "global_step": 53082, "epoch": 596, "lr": 9.099139872333069e-05}
{"train_loss": 0.2830241322517395, "global_step": 53083, "epoch": 596, "lr": 9.099106675765951e-05}
{"train_loss": 0.2595759332180023, "global_step": 53084, "epoch": 596, "lr": 9.099073478647757e-05}
{"train_loss": 0.33210474252700806, "global_step": 53085, "epoch": 596, "lr": 9.099040280978493e-05}
{"train_loss": 0.24893125891685486, "global_step": 53086, "epoch": 596, "lr": 9.099007082758164e-05}
{"train_loss": 0.27569788694381714, "global_step": 53087, "epoch": 596, "lr": 9.098973883986772e-05}
{"train_loss": 0.19996897876262665, "global_step": 53088, "epoch": 596, "lr": 9.098940684664324e-05}
{"train_loss": 0.28567391633987427, "global_step": 53089, "epoch": 596, "lr": 9.098907484790825e-05}
{"train_loss": 0.24849097430706024, "global_step": 53090, "epoch": 596, "lr": 9.098874284366274e-05}
{"train_loss": 0.1890345960855484, "global_step": 53091, "epoch": 596, "lr": 9.098841083390681e-05}
{"train_loss": 0.3139013350009918, "global_step": 53092, "epoch": 596, "lr": 9.098807881864049e-05}
{"train_loss": 0.34118887782096863, "global_step": 53093, "epoch": 596, "lr": 9.098774679786383e-05}
{"train_loss": 0.2658267915248871, "global_step": 53094, "epoch": 596, "lr": 9.098741477157685e-05}
{"train_loss": 0.246820867061615, "global_step": 53095, "epoch": 596, "lr": 9.098708273977963e-05}
{"train_loss": 0.30145150423049927, "global_step": 53096, "epoch": 596, "lr": 9.098675070247218e-05}
{"train_loss": 0.2441502809524536, "global_step": 53097, "epoch": 596, "lr": 9.098641865965456e-05}
{"train_loss": 0.24816495180130005, "global_step": 53098, "epoch": 596, "lr": 9.098608661132683e-05}
{"train_loss": 0.4080672562122345, "global_step": 53099, "epoch": 596, "lr": 9.098575455748903e-05}
{"train_loss": 0.29256144165992737, "global_step": 53100, "epoch": 596, "lr": 9.098542249814117e-05}
{"train_loss": 0.2640039622783661, "global_step": 53101, "epoch": 596, "lr": 9.098509043328332e-05}
{"train_loss": 0.24837936460971832, "global_step": 53102, "epoch": 596, "lr": 9.098475836291554e-05}
{"train_loss": 0.20008009672164917, "global_step": 53103, "epoch": 596, "lr": 9.098442628703785e-05}
{"train_loss": 0.23934975266456604, "global_step": 53104, "epoch": 596, "lr": 9.09840942056503e-05}
{"train_loss": 0.21690045297145844, "global_step": 53105, "epoch": 596, "lr": 9.098376211875295e-05}
{"train_loss": 0.1631384640932083, "global_step": 53106, "epoch": 596, "lr": 9.098343002634582e-05}
{"train_loss": 0.22139190137386322, "global_step": 53107, "epoch": 596, "lr": 9.098309792842899e-05}
{"train_loss": 0.23567719757556915, "global_step": 53108, "epoch": 596, "lr": 9.098276582500244e-05}
{"train_loss": 0.19492048025131226, "global_step": 53109, "epoch": 596, "lr": 9.09824337160663e-05}
{"train_loss": 0.1553371697664261, "global_step": 53110, "epoch": 596, "lr": 9.098210160162055e-05}
{"train_loss": 0.3119398057460785, "global_step": 53111, "epoch": 596, "lr": 9.098176948166526e-05}
{"train_loss": 0.29509487748146057, "global_step": 53112, "epoch": 596, "lr": 9.098143735620047e-05}
{"train_loss": 0.30077266693115234, "global_step": 53113, "epoch": 596, "lr": 9.098110522522621e-05}
{"train_loss": 0.2319163680076599, "global_step": 53114, "epoch": 596, "lr": 9.098077308874256e-05}
{"train_loss": 0.27765244245529175, "global_step": 53115, "epoch": 596, "lr": 9.098044094674953e-05}
{"train_loss": 0.2795712649822235, "global_step": 53116, "epoch": 596, "lr": 9.09801087992472e-05}
{"train_loss": 0.19534896314144135, "global_step": 53117, "epoch": 596, "lr": 9.097977664623559e-05}
{"train_loss": 0.18972541391849518, "global_step": 53118, "epoch": 596, "lr": 9.097944448771474e-05}
{"train_loss": 0.19644761085510254, "global_step": 53119, "epoch": 596, "lr": 9.09791123236847e-05}
{"train_loss": 0.26768168807029724, "global_step": 53120, "epoch": 596, "lr": 9.097878015414552e-05}
{"train_loss": 0.18551288545131683, "global_step": 53121, "epoch": 596, "lr": 9.097844797909724e-05}
{"train_loss": 0.34862270951271057, "global_step": 53122, "epoch": 596, "lr": 9.097811579853991e-05}
{"train_loss": 0.2479063719511032, "global_step": 53123, "epoch": 596, "lr": 9.097778361247358e-05}
{"train_loss": 0.34249162673950195, "global_step": 53124, "epoch": 596, "lr": 9.097745142089827e-05}
{"train_loss": 0.30802857875823975, "global_step": 53125, "epoch": 596, "lr": 9.097711922381406e-05}
{"train_loss": 0.3489932417869568, "global_step": 53126, "epoch": 596, "lr": 9.097678702122097e-05}
{"train_loss": 0.2534249424934387, "global_step": 53127, "epoch": 596, "lr": 9.097645481311904e-05}
{"train_loss": 0.23069895803928375, "global_step": 53128, "epoch": 596, "lr": 9.097612259950832e-05}
{"train_loss": 0.23318329453468323, "global_step": 53129, "epoch": 596, "lr": 9.097579038038888e-05}
{"train_loss": 0.27707618474960327, "global_step": 53130, "epoch": 596, "lr": 9.097545815576074e-05}
{"train_loss": 0.28559643030166626, "global_step": 53131, "epoch": 596, "lr": 9.097512592562393e-05}
{"train_loss": 0.26269314178590025, "global_step": 53132, "epoch": 596, "lr": 9.097479368997854e-05, "val_loss": 3.2891924381256104}
{"train_loss": 0.2775248885154724, "global_step": 53133, "epoch": 597, "lr": 9.097446144882457e-05}
{"train_loss": 0.19238944351673126, "global_step": 53134, "epoch": 597, "lr": 9.09741292021621e-05}
{"train_loss": 0.352638304233551, "global_step": 53135, "epoch": 597, "lr": 9.097379694999115e-05}
{"train_loss": 0.34941890835762024, "global_step": 53136, "epoch": 597, "lr": 9.097346469231175e-05}
{"train_loss": 0.18242235481739044, "global_step": 53137, "epoch": 597, "lr": 9.0973132429124e-05}
{"train_loss": 0.33778268098831177, "global_step": 53138, "epoch": 597, "lr": 9.09728001604279e-05}
{"train_loss": 0.24155525863170624, "global_step": 53139, "epoch": 597, "lr": 9.09724678862235e-05}
{"train_loss": 0.2303580343723297, "global_step": 53140, "epoch": 597, "lr": 9.097213560651085e-05}
{"train_loss": 0.2896220088005066, "global_step": 53141, "epoch": 597, "lr": 9.097180332129e-05}
{"train_loss": 0.24995478987693787, "global_step": 53142, "epoch": 597, "lr": 9.097147103056101e-05}
{"train_loss": 0.31283143162727356, "global_step": 53143, "epoch": 597, "lr": 9.097113873432388e-05}
{"train_loss": 0.18308299779891968, "global_step": 53144, "epoch": 597, "lr": 9.097080643257869e-05}
{"train_loss": 0.26836034655570984, "global_step": 53145, "epoch": 597, "lr": 9.097047412532547e-05}
{"train_loss": 0.3108851909637451, "global_step": 53146, "epoch": 597, "lr": 9.097014181256427e-05}
{"train_loss": 0.17895901203155518, "global_step": 53147, "epoch": 597, "lr": 9.096980949429514e-05}
{"train_loss": 0.21815910935401917, "global_step": 53148, "epoch": 597, "lr": 9.096947717051811e-05}
{"train_loss": 0.27286574244499207, "global_step": 53149, "epoch": 597, "lr": 9.096914484123323e-05}
{"train_loss": 0.3829719126224518, "global_step": 53150, "epoch": 597, "lr": 9.096881250644056e-05}
{"train_loss": 0.28858238458633423, "global_step": 53151, "epoch": 597, "lr": 9.096848016614013e-05}
{"train_loss": 0.22346970438957214, "global_step": 53152, "epoch": 597, "lr": 9.096814782033198e-05}
{"train_loss": 0.31542882323265076, "global_step": 53153, "epoch": 597, "lr": 9.096781546901616e-05}
{"train_loss": 0.2090977430343628, "global_step": 53154, "epoch": 597, "lr": 9.096748311219273e-05}
{"train_loss": 0.3082174062728882, "global_step": 53155, "epoch": 597, "lr": 9.09671507498617e-05}
{"train_loss": 0.25499770045280457, "global_step": 53156, "epoch": 597, "lr": 9.096681838202317e-05}
{"train_loss": 0.22958552837371826, "global_step": 53157, "epoch": 597, "lr": 9.096648600867712e-05}
{"train_loss": 0.30397528409957886, "global_step": 53158, "epoch": 597, "lr": 9.096615362982363e-05}
{"train_loss": 0.33347877860069275, "global_step": 53159, "epoch": 597, "lr": 9.096582124546276e-05}
{"train_loss": 0.17849022150039673, "global_step": 53160, "epoch": 597, "lr": 9.096548885559451e-05}
{"train_loss": 0.19268304109573364, "global_step": 53161, "epoch": 597, "lr": 9.096515646021897e-05}
{"train_loss": 0.23202399909496307, "global_step": 53162, "epoch": 597, "lr": 9.096482405933615e-05}
{"train_loss": 0.2576565444469452, "global_step": 53163, "epoch": 597, "lr": 9.096449165294612e-05}
{"train_loss": 0.2623877227306366, "global_step": 53164, "epoch": 597, "lr": 9.09641592410489e-05}
{"train_loss": 0.2033451348543167, "global_step": 53165, "epoch": 597, "lr": 9.096382682364455e-05}
{"train_loss": 0.2562308609485626, "global_step": 53166, "epoch": 597, "lr": 9.096349440073311e-05}
{"train_loss": 0.28116124868392944, "global_step": 53167, "epoch": 597, "lr": 9.096316197231463e-05}
{"train_loss": 0.2436227649450302, "global_step": 53168, "epoch": 597, "lr": 9.096282953838917e-05}
{"train_loss": 0.20706187188625336, "global_step": 53169, "epoch": 597, "lr": 9.096249709895674e-05}
{"train_loss": 0.20930345356464386, "global_step": 53170, "epoch": 597, "lr": 9.096216465401739e-05}
{"train_loss": 0.2591215670108795, "global_step": 53171, "epoch": 597, "lr": 9.096183220357119e-05}
{"train_loss": 0.2844044864177704, "global_step": 53172, "epoch": 597, "lr": 9.096149974761818e-05}
{"train_loss": 0.27886661887168884, "global_step": 53173, "epoch": 597, "lr": 9.09611672861584e-05}
{"train_loss": 0.27721548080444336, "global_step": 53174, "epoch": 597, "lr": 9.096083481919187e-05}
{"train_loss": 0.27687758207321167, "global_step": 53175, "epoch": 597, "lr": 9.096050234671866e-05}
{"train_loss": 0.2448062300682068, "global_step": 53176, "epoch": 597, "lr": 9.096016986873881e-05}
{"train_loss": 0.3036098778247833, "global_step": 53177, "epoch": 597, "lr": 9.095983738525236e-05}
{"train_loss": 0.2728578448295593, "global_step": 53178, "epoch": 597, "lr": 9.095950489625937e-05}
{"train_loss": 0.4090261459350586, "global_step": 53179, "epoch": 597, "lr": 9.095917240175987e-05}
{"train_loss": 0.2872503399848938, "global_step": 53180, "epoch": 597, "lr": 9.09588399017539e-05}
{"train_loss": 0.221464604139328, "global_step": 53181, "epoch": 597, "lr": 9.095850739624152e-05}
{"train_loss": 0.240571528673172, "global_step": 53182, "epoch": 597, "lr": 9.095817488522278e-05}
{"train_loss": 0.23289383947849274, "global_step": 53183, "epoch": 597, "lr": 9.09578423686977e-05}
{"train_loss": 0.2188921421766281, "global_step": 53184, "epoch": 597, "lr": 9.095750984666633e-05}
{"train_loss": 0.3024730086326599, "global_step": 53185, "epoch": 597, "lr": 9.095717731912874e-05}
{"train_loss": 0.2306094765663147, "global_step": 53186, "epoch": 597, "lr": 9.095684478608495e-05}
{"train_loss": 0.34828200936317444, "global_step": 53187, "epoch": 597, "lr": 9.095651224753501e-05}
{"train_loss": 0.23532013595104218, "global_step": 53188, "epoch": 597, "lr": 9.095617970347896e-05}
{"train_loss": 0.2186618149280548, "global_step": 53189, "epoch": 597, "lr": 9.095584715391685e-05}
{"train_loss": 0.20658931136131287, "global_step": 53190, "epoch": 597, "lr": 9.095551459884873e-05}
{"train_loss": 0.3083505630493164, "global_step": 53191, "epoch": 597, "lr": 9.095518203827464e-05}
{"train_loss": 0.2589847445487976, "global_step": 53192, "epoch": 597, "lr": 9.095484947219464e-05}
{"train_loss": 0.20378455519676208, "global_step": 53193, "epoch": 597, "lr": 9.095451690060874e-05}
{"train_loss": 0.23358574509620667, "global_step": 53194, "epoch": 597, "lr": 9.095418432351702e-05}
{"train_loss": 0.2812175750732422, "global_step": 53195, "epoch": 597, "lr": 9.095385174091949e-05}
{"train_loss": 0.2611600160598755, "global_step": 53196, "epoch": 597, "lr": 9.095351915281623e-05}
{"train_loss": 0.2521076798439026, "global_step": 53197, "epoch": 597, "lr": 9.095318655920726e-05}
{"train_loss": 0.3655071258544922, "global_step": 53198, "epoch": 597, "lr": 9.095285396009263e-05}
{"train_loss": 0.2723473310470581, "global_step": 53199, "epoch": 597, "lr": 9.095252135547239e-05}
{"train_loss": 0.3872371017932892, "global_step": 53200, "epoch": 597, "lr": 9.095218874534659e-05}
{"train_loss": 0.4366742670536041, "global_step": 53201, "epoch": 597, "lr": 9.095185612971526e-05}
{"train_loss": 0.34547972679138184, "global_step": 53202, "epoch": 597, "lr": 9.095152350857845e-05}
{"train_loss": 0.3664894700050354, "global_step": 53203, "epoch": 597, "lr": 9.095119088193622e-05}
{"train_loss": 0.20779205858707428, "global_step": 53204, "epoch": 597, "lr": 9.09508582497886e-05}
{"train_loss": 0.3676556348800659, "global_step": 53205, "epoch": 597, "lr": 9.095052561213562e-05}
{"train_loss": 0.27202099561691284, "global_step": 53206, "epoch": 597, "lr": 9.095019296897736e-05}
{"train_loss": 0.3995295763015747, "global_step": 53207, "epoch": 597, "lr": 9.094986032031382e-05}
{"train_loss": 0.26164373755455017, "global_step": 53208, "epoch": 597, "lr": 9.094952766614508e-05}
{"train_loss": 0.3762192726135254, "global_step": 53209, "epoch": 597, "lr": 9.094919500647119e-05}
{"train_loss": 0.2911478579044342, "global_step": 53210, "epoch": 597, "lr": 9.094886234129218e-05}
{"train_loss": 0.35318678617477417, "global_step": 53211, "epoch": 597, "lr": 9.094852967060807e-05}
{"train_loss": 0.3289647400379181, "global_step": 53212, "epoch": 597, "lr": 9.094819699441896e-05}
{"train_loss": 0.4264650344848633, "global_step": 53213, "epoch": 597, "lr": 9.094786431272483e-05}
{"train_loss": 0.31390058994293213, "global_step": 53214, "epoch": 597, "lr": 9.09475316255258e-05}
{"train_loss": 0.2988041043281555, "global_step": 53215, "epoch": 597, "lr": 9.094719893282184e-05}
{"train_loss": 0.28845229744911194, "global_step": 53216, "epoch": 597, "lr": 9.094686623461303e-05}
{"train_loss": 0.243210569024086, "global_step": 53217, "epoch": 597, "lr": 9.094653353089944e-05}
{"train_loss": 0.20959016680717468, "global_step": 53218, "epoch": 597, "lr": 9.094620082168105e-05}
{"train_loss": 0.28055739402770996, "global_step": 53219, "epoch": 597, "lr": 9.094586810695797e-05}
{"train_loss": 0.3273720443248749, "global_step": 53220, "epoch": 597, "lr": 9.094553538673022e-05}
{"train_loss": 0.2781752399849088, "global_step": 53221, "epoch": 597, "lr": 9.094520266099782e-05, "val_loss": 3.282210111618042}
{"train_loss": 0.2741413116455078, "global_step": 53222, "epoch": 598, "lr": 9.094486992976085e-05}
{"train_loss": 0.2532575726509094, "global_step": 53223, "epoch": 598, "lr": 9.094453719301934e-05}
{"train_loss": 0.27621573209762573, "global_step": 53224, "epoch": 598, "lr": 9.094420445077334e-05}
{"train_loss": 0.21767091751098633, "global_step": 53225, "epoch": 598, "lr": 9.09438717030229e-05}
{"train_loss": 0.28087857365608215, "global_step": 53226, "epoch": 598, "lr": 9.094353894976803e-05}
{"train_loss": 0.3259375989437103, "global_step": 53227, "epoch": 598, "lr": 9.09432061910088e-05}
{"train_loss": 0.2336251437664032, "global_step": 53228, "epoch": 598, "lr": 9.094287342674527e-05}
{"train_loss": 0.25032734870910645, "global_step": 53229, "epoch": 598, "lr": 9.094254065697746e-05}
{"train_loss": 0.38801223039627075, "global_step": 53230, "epoch": 598, "lr": 9.094220788170545e-05}
{"train_loss": 0.27757352590560913, "global_step": 53231, "epoch": 598, "lr": 9.094187510092924e-05}
{"train_loss": 0.32086488604545593, "global_step": 53232, "epoch": 598, "lr": 9.094154231464889e-05}
{"train_loss": 0.20301376283168793, "global_step": 53233, "epoch": 598, "lr": 9.094120952286445e-05}
{"train_loss": 0.30943894386291504, "global_step": 53234, "epoch": 598, "lr": 9.094087672557596e-05}
{"train_loss": 0.20958465337753296, "global_step": 53235, "epoch": 598, "lr": 9.094054392278349e-05}
{"train_loss": 0.3131399154663086, "global_step": 53236, "epoch": 598, "lr": 9.094021111448705e-05}
{"train_loss": 0.22136284410953522, "global_step": 53237, "epoch": 598, "lr": 9.093987830068669e-05}
{"train_loss": 0.2511727511882782, "global_step": 53238, "epoch": 598, "lr": 9.093954548138247e-05}
{"train_loss": 0.20271654427051544, "global_step": 53239, "epoch": 598, "lr": 9.093921265657442e-05}
{"train_loss": 0.26277971267700195, "global_step": 53240, "epoch": 598, "lr": 9.093887982626259e-05}
{"train_loss": 0.2437247484922409, "global_step": 53241, "epoch": 598, "lr": 9.093854699044704e-05}
{"train_loss": 0.3596861660480499, "global_step": 53242, "epoch": 598, "lr": 9.09382141491278e-05}
{"train_loss": 0.21291673183441162, "global_step": 53243, "epoch": 598, "lr": 9.09378813023049e-05}
{"train_loss": 0.23860570788383484, "global_step": 53244, "epoch": 598, "lr": 9.093754844997841e-05}
{"train_loss": 0.2899836301803589, "global_step": 53245, "epoch": 598, "lr": 9.093721559214835e-05}
{"train_loss": 0.22150756418704987, "global_step": 53246, "epoch": 598, "lr": 9.09368827288148e-05}
{"train_loss": 0.28380319476127625, "global_step": 53247, "epoch": 598, "lr": 9.093654985997779e-05}
{"train_loss": 0.2486853003501892, "global_step": 53248, "epoch": 598, "lr": 9.093621698563736e-05}
{"train_loss": 0.21258072555065155, "global_step": 53249, "epoch": 598, "lr": 9.093588410579355e-05}
{"train_loss": 0.27093106508255005, "global_step": 53250, "epoch": 598, "lr": 9.093555122044639e-05}
{"train_loss": 0.260193794965744, "global_step": 53251, "epoch": 598, "lr": 9.093521832959598e-05}
{"train_loss": 0.25954562425613403, "global_step": 53252, "epoch": 598, "lr": 9.093488543324229e-05}
{"train_loss": 0.28384581208229065, "global_step": 53253, "epoch": 598, "lr": 9.093455253138543e-05}
{"train_loss": 0.1559140831232071, "global_step": 53254, "epoch": 598, "lr": 9.093421962402541e-05}
{"train_loss": 0.2297792285680771, "global_step": 53255, "epoch": 598, "lr": 9.093388671116228e-05}
{"train_loss": 0.1821148842573166, "global_step": 53256, "epoch": 598, "lr": 9.093355379279609e-05}
{"train_loss": 0.2555682361125946, "global_step": 53257, "epoch": 598, "lr": 9.093322086892689e-05}
{"train_loss": 0.25461655855178833, "global_step": 53258, "epoch": 598, "lr": 9.09328879395547e-05}
{"train_loss": 0.17664986848831177, "global_step": 53259, "epoch": 598, "lr": 9.09325550046796e-05}
{"train_loss": 0.2543322741985321, "global_step": 53260, "epoch": 598, "lr": 9.093222206430162e-05}
{"train_loss": 0.3834640085697174, "global_step": 53261, "epoch": 598, "lr": 9.093188911842079e-05}
{"train_loss": 0.22548545897006989, "global_step": 53262, "epoch": 598, "lr": 9.093155616703716e-05}
{"train_loss": 0.21133330464363098, "global_step": 53263, "epoch": 598, "lr": 9.093122321015079e-05}
{"train_loss": 0.2534613609313965, "global_step": 53264, "epoch": 598, "lr": 9.093089024776171e-05}
{"train_loss": 0.21993237733840942, "global_step": 53265, "epoch": 598, "lr": 9.093055727986998e-05}
{"train_loss": 0.2100953757762909, "global_step": 53266, "epoch": 598, "lr": 9.093022430647563e-05}
{"train_loss": 0.2594383656978607, "global_step": 53267, "epoch": 598, "lr": 9.092989132757872e-05}
{"train_loss": 0.20278054475784302, "global_step": 53268, "epoch": 598, "lr": 9.092955834317926e-05}
{"train_loss": 0.20684552192687988, "global_step": 53269, "epoch": 598, "lr": 9.092922535327735e-05}
{"train_loss": 0.343462198972702, "global_step": 53270, "epoch": 598, "lr": 9.092889235787299e-05}
{"train_loss": 0.1744457483291626, "global_step": 53271, "epoch": 598, "lr": 9.092855935696623e-05}
{"train_loss": 0.20625993609428406, "global_step": 53272, "epoch": 598, "lr": 9.092822635055713e-05}
{"train_loss": 0.2868232727050781, "global_step": 53273, "epoch": 598, "lr": 9.092789333864574e-05}
{"train_loss": 0.24736949801445007, "global_step": 53274, "epoch": 598, "lr": 9.092756032123209e-05}
{"train_loss": 0.2633530795574188, "global_step": 53275, "epoch": 598, "lr": 9.092722729831621e-05}
{"train_loss": 0.2843243181705475, "global_step": 53276, "epoch": 598, "lr": 9.09268942698982e-05}
{"train_loss": 0.29662519693374634, "global_step": 53277, "epoch": 598, "lr": 9.092656123597804e-05}
{"train_loss": 0.13291418552398682, "global_step": 53278, "epoch": 598, "lr": 9.092622819655581e-05}
{"train_loss": 0.36021938920021057, "global_step": 53279, "epoch": 598, "lr": 9.092589515163154e-05}
{"train_loss": 0.25914737582206726, "global_step": 53280, "epoch": 598, "lr": 9.092556210120529e-05}
{"train_loss": 0.3360424339771271, "global_step": 53281, "epoch": 598, "lr": 9.092522904527709e-05}
{"train_loss": 0.2504160702228546, "global_step": 53282, "epoch": 598, "lr": 9.0924895983847e-05}
{"train_loss": 0.21060873568058014, "global_step": 53283, "epoch": 598, "lr": 9.092456291691505e-05}
{"train_loss": 0.21530193090438843, "global_step": 53284, "epoch": 598, "lr": 9.09242298444813e-05}
{"train_loss": 0.2173883616924286, "global_step": 53285, "epoch": 598, "lr": 9.092389676654578e-05}
{"train_loss": 0.3714081346988678, "global_step": 53286, "epoch": 598, "lr": 9.092356368310854e-05}
{"train_loss": 0.28259381651878357, "global_step": 53287, "epoch": 598, "lr": 9.092323059416962e-05}
{"train_loss": 0.31411513686180115, "global_step": 53288, "epoch": 598, "lr": 9.092289749972908e-05}
{"train_loss": 0.2928175926208496, "global_step": 53289, "epoch": 598, "lr": 9.092256439978696e-05}
{"train_loss": 0.2056368738412857, "global_step": 53290, "epoch": 598, "lr": 9.092223129434329e-05}
{"train_loss": 0.2690810561180115, "global_step": 53291, "epoch": 598, "lr": 9.092189818339813e-05}
{"train_loss": 0.1673084944486618, "global_step": 53292, "epoch": 598, "lr": 9.092156506695151e-05}
{"train_loss": 0.32103174924850464, "global_step": 53293, "epoch": 598, "lr": 9.09212319450035e-05}
{"train_loss": 0.34698253870010376, "global_step": 53294, "epoch": 598, "lr": 9.092089881755412e-05}
{"train_loss": 0.3203585147857666, "global_step": 53295, "epoch": 598, "lr": 9.092056568460342e-05}
{"train_loss": 0.23980584740638733, "global_step": 53296, "epoch": 598, "lr": 9.092023254615146e-05}
{"train_loss": 0.2539094388484955, "global_step": 53297, "epoch": 598, "lr": 9.091989940219827e-05}
{"train_loss": 0.2828393280506134, "global_step": 53298, "epoch": 598, "lr": 9.091956625274389e-05}
{"train_loss": 0.23242001235485077, "global_step": 53299, "epoch": 598, "lr": 9.091923309778839e-05}
{"train_loss": 0.26470544934272766, "global_step": 53300, "epoch": 598, "lr": 9.091889993733178e-05}
{"train_loss": 0.23287896811962128, "global_step": 53301, "epoch": 598, "lr": 9.091856677137412e-05}
{"train_loss": 0.24841229617595673, "global_step": 53302, "epoch": 598, "lr": 9.091823359991547e-05}
{"train_loss": 0.29516005516052246, "global_step": 53303, "epoch": 598, "lr": 9.091790042295587e-05}
{"train_loss": 0.35072413086891174, "global_step": 53304, "epoch": 598, "lr": 9.091756724049534e-05}
{"train_loss": 0.2678771913051605, "global_step": 53305, "epoch": 598, "lr": 9.091723405253394e-05}
{"train_loss": 0.20674213767051697, "global_step": 53306, "epoch": 598, "lr": 9.091690085907173e-05}
{"train_loss": 0.29377567768096924, "global_step": 53307, "epoch": 598, "lr": 9.091656766010873e-05}
{"train_loss": 0.22939279675483704, "global_step": 53308, "epoch": 598, "lr": 9.091623445564501e-05}
{"train_loss": 0.22853277623653412, "global_step": 53309, "epoch": 598, "lr": 9.091590124568059e-05}
{"train_loss": 0.25847066336133506, "global_step": 53310, "epoch": 598, "lr": 9.091556803021552e-05, "val_loss": 3.2913196086883545}
{"train_loss": 0.24326860904693604, "global_step": 53311, "epoch": 599, "lr": 9.091523480924986e-05}
{"train_loss": 0.2880348861217499, "global_step": 53312, "epoch": 599, "lr": 9.091490158278364e-05}
{"train_loss": 0.2856614887714386, "global_step": 53313, "epoch": 599, "lr": 9.091456835081692e-05}
{"train_loss": 0.24024640023708344, "global_step": 53314, "epoch": 599, "lr": 9.091423511334973e-05}
{"train_loss": 0.3640826344490051, "global_step": 53315, "epoch": 599, "lr": 9.091390187038211e-05}
{"train_loss": 0.3171186149120331, "global_step": 53316, "epoch": 599, "lr": 9.091356862191414e-05}
{"train_loss": 0.15707820653915405, "global_step": 53317, "epoch": 599, "lr": 9.091323536794581e-05}
{"train_loss": 0.19672253727912903, "global_step": 53318, "epoch": 599, "lr": 9.09129021084772e-05}
{"train_loss": 0.26254191994667053, "global_step": 53319, "epoch": 599, "lr": 9.091256884350836e-05}
{"train_loss": 0.2528877854347229, "global_step": 53320, "epoch": 599, "lr": 9.091223557303932e-05}
{"train_loss": 0.2570539116859436, "global_step": 53321, "epoch": 599, "lr": 9.091190229707013e-05}
{"train_loss": 0.34246158599853516, "global_step": 53322, "epoch": 599, "lr": 9.091156901560083e-05}
{"train_loss": 0.225931316614151, "global_step": 53323, "epoch": 599, "lr": 9.091123572863148e-05}
{"train_loss": 0.2223898321390152, "global_step": 53324, "epoch": 599, "lr": 9.091090243616209e-05}
{"train_loss": 0.1832030862569809, "global_step": 53325, "epoch": 599, "lr": 9.091056913819276e-05}
{"train_loss": 0.15484710037708282, "global_step": 53326, "epoch": 599, "lr": 9.091023583472346e-05}
{"train_loss": 0.31131982803344727, "global_step": 53327, "epoch": 599, "lr": 9.09099025257543e-05}
{"train_loss": 0.27929532527923584, "global_step": 53328, "epoch": 599, "lr": 9.090956921128532e-05}
{"train_loss": 0.21882449090480804, "global_step": 53329, "epoch": 599, "lr": 9.090923589131652e-05}
{"train_loss": 0.22889837622642517, "global_step": 53330, "epoch": 599, "lr": 9.090890256584799e-05}
{"train_loss": 0.21722804009914398, "global_step": 53331, "epoch": 599, "lr": 9.090856923487975e-05}
{"train_loss": 0.1947481334209442, "global_step": 53332, "epoch": 599, "lr": 9.090823589841184e-05}
{"train_loss": 0.2391071617603302, "global_step": 53333, "epoch": 599, "lr": 9.090790255644433e-05}
{"train_loss": 0.21594615280628204, "global_step": 53334, "epoch": 599, "lr": 9.090756920897725e-05}
{"train_loss": 0.2850363850593567, "global_step": 53335, "epoch": 599, "lr": 9.090723585601064e-05}
{"train_loss": 0.2721066474914551, "global_step": 53336, "epoch": 599, "lr": 9.090690249754457e-05}
{"train_loss": 0.29332804679870605, "global_step": 53337, "epoch": 599, "lr": 9.090656913357904e-05}
{"train_loss": 0.19820907711982727, "global_step": 53338, "epoch": 599, "lr": 9.090623576411412e-05}
{"train_loss": 0.289188951253891, "global_step": 53339, "epoch": 599, "lr": 9.090590238914987e-05}
{"train_loss": 0.37349236011505127, "global_step": 53340, "epoch": 599, "lr": 9.090556900868632e-05}
{"train_loss": 0.22528864443302155, "global_step": 53341, "epoch": 599, "lr": 9.09052356227235e-05}
{"train_loss": 0.2759421765804291, "global_step": 53342, "epoch": 599, "lr": 9.090490223126149e-05}
{"train_loss": 0.24475739896297455, "global_step": 53343, "epoch": 599, "lr": 9.09045688343003e-05}
{"train_loss": 0.26473575830459595, "global_step": 53344, "epoch": 599, "lr": 9.090423543184e-05}
{"train_loss": 0.3073091506958008, "global_step": 53345, "epoch": 599, "lr": 9.090390202388062e-05}
{"train_loss": 0.24202656745910645, "global_step": 53346, "epoch": 599, "lr": 9.09035686104222e-05}
{"train_loss": 0.2819962203502655, "global_step": 53347, "epoch": 599, "lr": 9.09032351914648e-05}
{"train_loss": 0.3535066545009613, "global_step": 53348, "epoch": 599, "lr": 9.090290176700846e-05}
{"train_loss": 0.27052733302116394, "global_step": 53349, "epoch": 599, "lr": 9.090256833705322e-05}
{"train_loss": 0.24394722282886505, "global_step": 53350, "epoch": 599, "lr": 9.090223490159913e-05}
{"train_loss": 0.2826574444770813, "global_step": 53351, "epoch": 599, "lr": 9.090190146064623e-05}
{"train_loss": 0.1557057946920395, "global_step": 53352, "epoch": 599, "lr": 9.090156801419457e-05}
{"train_loss": 0.1900256723165512, "global_step": 53353, "epoch": 599, "lr": 9.09012345622442e-05}
{"train_loss": 0.21033430099487305, "global_step": 53354, "epoch": 599, "lr": 9.090090110479514e-05}
{"train_loss": 0.2890227437019348, "global_step": 53355, "epoch": 599, "lr": 9.090056764184747e-05}
{"train_loss": 0.3129730820655823, "global_step": 53356, "epoch": 599, "lr": 9.090023417340121e-05}
{"train_loss": 0.2827913761138916, "global_step": 53357, "epoch": 599, "lr": 9.089990069945642e-05}
{"train_loss": 0.2491304874420166, "global_step": 53358, "epoch": 599, "lr": 9.089956722001312e-05}
{"train_loss": 0.2784791588783264, "global_step": 53359, "epoch": 599, "lr": 9.089923373507139e-05}
{"train_loss": 0.3283000886440277, "global_step": 53360, "epoch": 599, "lr": 9.089890024463124e-05}
{"train_loss": 0.2862775921821594, "global_step": 53361, "epoch": 599, "lr": 9.089856674869275e-05}
{"train_loss": 0.34943437576293945, "global_step": 53362, "epoch": 599, "lr": 9.089823324725592e-05}
{"train_loss": 0.37473776936531067, "global_step": 53363, "epoch": 599, "lr": 9.089789974032084e-05}
{"train_loss": 0.32775235176086426, "global_step": 53364, "epoch": 599, "lr": 9.089756622788754e-05}
{"train_loss": 0.28735873103141785, "global_step": 53365, "epoch": 599, "lr": 9.089723270995605e-05}
{"train_loss": 0.3335515260696411, "global_step": 53366, "epoch": 599, "lr": 9.089689918652643e-05}
{"train_loss": 0.1854284256696701, "global_step": 53367, "epoch": 599, "lr": 9.089656565759872e-05}
{"train_loss": 0.1696256697177887, "global_step": 53368, "epoch": 599, "lr": 9.089623212317296e-05}
{"train_loss": 0.22743937373161316, "global_step": 53369, "epoch": 599, "lr": 9.089589858324921e-05}
{"train_loss": 0.28141969442367554, "global_step": 53370, "epoch": 599, "lr": 9.08955650378275e-05}
{"train_loss": 0.2806342542171478, "global_step": 53371, "epoch": 599, "lr": 9.089523148690788e-05}
{"train_loss": 0.22407974302768707, "global_step": 53372, "epoch": 599, "lr": 9.089489793049038e-05}
{"train_loss": 0.3061337172985077, "global_step": 53373, "epoch": 599, "lr": 9.089456436857508e-05}
{"train_loss": 0.28076237440109253, "global_step": 53374, "epoch": 599, "lr": 9.0894230801162e-05}
{"train_loss": 0.265794038772583, "global_step": 53375, "epoch": 599, "lr": 9.08938972282512e-05}
{"train_loss": 0.2570905089378357, "global_step": 53376, "epoch": 599, "lr": 9.089356364984269e-05}
{"train_loss": 0.2790391743183136, "global_step": 53377, "epoch": 599, "lr": 9.089323006593654e-05}
{"train_loss": 0.23408152163028717, "global_step": 53378, "epoch": 599, "lr": 9.08928964765328e-05}
{"train_loss": 0.2600192427635193, "global_step": 53379, "epoch": 599, "lr": 9.089256288163152e-05}
{"train_loss": 0.26057547330856323, "global_step": 53380, "epoch": 599, "lr": 9.089222928123272e-05}
{"train_loss": 0.36778873205184937, "global_step": 53381, "epoch": 599, "lr": 9.089189567533646e-05}
{"train_loss": 0.20442497730255127, "global_step": 53382, "epoch": 599, "lr": 9.089156206394279e-05}
{"train_loss": 0.25790661573410034, "global_step": 53383, "epoch": 599, "lr": 9.089122844705173e-05}
{"train_loss": 0.24262720346450806, "global_step": 53384, "epoch": 599, "lr": 9.089089482466337e-05}
{"train_loss": 0.25621047616004944, "global_step": 53385, "epoch": 599, "lr": 9.089056119677771e-05}
{"train_loss": 0.19838079810142517, "global_step": 53386, "epoch": 599, "lr": 9.089022756339482e-05}
{"train_loss": 0.3692574203014374, "global_step": 53387, "epoch": 599, "lr": 9.088989392451472e-05}
{"train_loss": 0.4010416865348816, "global_step": 53388, "epoch": 599, "lr": 9.08895602801375e-05}
{"train_loss": 0.31616634130477905, "global_step": 53389, "epoch": 599, "lr": 9.088922663026316e-05}
{"train_loss": 0.301620215177536, "global_step": 53390, "epoch": 599, "lr": 9.088889297489176e-05}
{"train_loss": 0.21130381524562836, "global_step": 53391, "epoch": 599, "lr": 9.088855931402336e-05}
{"train_loss": 0.1961309164762497, "global_step": 53392, "epoch": 599, "lr": 9.088822564765798e-05}
{"train_loss": 0.2299850583076477, "global_step": 53393, "epoch": 599, "lr": 9.088789197579569e-05}
{"train_loss": 0.32951125502586365, "global_step": 53394, "epoch": 599, "lr": 9.088755829843649e-05}
{"train_loss": 0.22574111819267273, "global_step": 53395, "epoch": 599, "lr": 9.088722461558047e-05}
{"train_loss": 0.22755448520183563, "global_step": 53396, "epoch": 599, "lr": 9.088689092722769e-05}
{"train_loss": 0.29637855291366577, "global_step": 53397, "epoch": 599, "lr": 9.088655723337813e-05}
{"train_loss": 0.2564389109611511, "global_step": 53398, "epoch": 599, "lr": 9.088622353403188e-05}
{"train_loss": 0.2642321318722843, "global_step": 53399, "epoch": 599, "lr": 9.088588982918899e-05, "val_loss": 3.2003607749938965}
{"train_loss": 0.22847287356853485, "global_step": 53400, "epoch": 600, "lr": 9.088555611884948e-05}
{"train_loss": 0.27984321117401123, "global_step": 53401, "epoch": 600, "lr": 9.08852224030134e-05}
{"train_loss": 0.23117953538894653, "global_step": 53402, "epoch": 600, "lr": 9.088488868168081e-05}
{"train_loss": 0.24247562885284424, "global_step": 53403, "epoch": 600, "lr": 9.088455495485175e-05}
{"train_loss": 0.20878033339977264, "global_step": 53404, "epoch": 600, "lr": 9.088422122252624e-05}
{"train_loss": 0.17571806907653809, "global_step": 53405, "epoch": 600, "lr": 9.088388748470436e-05}
{"train_loss": 0.22080345451831818, "global_step": 53406, "epoch": 600, "lr": 9.088355374138614e-05}
{"train_loss": 0.29292285442352295, "global_step": 53407, "epoch": 600, "lr": 9.088321999257161e-05}
{"train_loss": 0.17753063142299652, "global_step": 53408, "epoch": 600, "lr": 9.088288623826084e-05}
{"train_loss": 0.29570120573043823, "global_step": 53409, "epoch": 600, "lr": 9.088255247845387e-05}
{"train_loss": 0.2884005606174469, "global_step": 53410, "epoch": 600, "lr": 9.088221871315073e-05}
{"train_loss": 0.23282060027122498, "global_step": 53411, "epoch": 600, "lr": 9.088188494235147e-05}
{"train_loss": 0.252474844455719, "global_step": 53412, "epoch": 600, "lr": 9.088155116605614e-05}
{"train_loss": 0.3374778628349304, "global_step": 53413, "epoch": 600, "lr": 9.088121738426479e-05}
{"train_loss": 0.1712924838066101, "global_step": 53414, "epoch": 600, "lr": 9.088088359697746e-05}
{"train_loss": 0.2282315045595169, "global_step": 53415, "epoch": 600, "lr": 9.08805498041942e-05}
{"train_loss": 0.20704032480716705, "global_step": 53416, "epoch": 600, "lr": 9.088021600591503e-05}
{"train_loss": 0.22360306978225708, "global_step": 53417, "epoch": 600, "lr": 9.087988220214002e-05}
{"train_loss": 0.2992037534713745, "global_step": 53418, "epoch": 600, "lr": 9.087954839286921e-05}
{"train_loss": 0.2658083140850067, "global_step": 53419, "epoch": 600, "lr": 9.087921457810265e-05}
{"train_loss": 0.24260246753692627, "global_step": 53420, "epoch": 600, "lr": 9.087888075784036e-05}
{"train_loss": 0.2490536868572235, "global_step": 53421, "epoch": 600, "lr": 9.087854693208242e-05}
{"train_loss": 0.2548742890357971, "global_step": 53422, "epoch": 600, "lr": 9.087821310082885e-05}
{"train_loss": 0.2562274634838104, "global_step": 53423, "epoch": 600, "lr": 9.087787926407971e-05}
{"train_loss": 0.1404801905155182, "global_step": 53424, "epoch": 600, "lr": 9.087754542183503e-05}
{"train_loss": 0.23872224986553192, "global_step": 53425, "epoch": 600, "lr": 9.087721157409485e-05}
{"train_loss": 0.2933920621871948, "global_step": 53426, "epoch": 600, "lr": 9.087687772085925e-05}
{"train_loss": 0.28606727719306946, "global_step": 53427, "epoch": 600, "lr": 9.087654386212825e-05}
{"train_loss": 0.23514796793460846, "global_step": 53428, "epoch": 600, "lr": 9.087620999790188e-05}
{"train_loss": 0.23482652008533478, "global_step": 53429, "epoch": 600, "lr": 9.087587612818022e-05}
{"train_loss": 0.2236742079257965, "global_step": 53430, "epoch": 600, "lr": 9.087554225296329e-05}
{"train_loss": 0.2002193182706833, "global_step": 53431, "epoch": 600, "lr": 9.087520837225113e-05}
{"train_loss": 0.2304021120071411, "global_step": 53432, "epoch": 600, "lr": 9.087487448604381e-05}
{"train_loss": 0.2952885925769806, "global_step": 53433, "epoch": 600, "lr": 9.087454059434136e-05}
{"train_loss": 0.24235638976097107, "global_step": 53434, "epoch": 600, "lr": 9.087420669714381e-05}
{"train_loss": 0.23756471276283264, "global_step": 53435, "epoch": 600, "lr": 9.087387279445125e-05}
{"train_loss": 0.2428768426179886, "global_step": 53436, "epoch": 600, "lr": 9.087353888626368e-05}
{"train_loss": 0.26333293318748474, "global_step": 53437, "epoch": 600, "lr": 9.087320497258117e-05}
{"train_loss": 0.24987486004829407, "global_step": 53438, "epoch": 600, "lr": 9.087287105340374e-05}
{"train_loss": 0.2216920107603073, "global_step": 53439, "epoch": 600, "lr": 9.087253712873145e-05}
{"train_loss": 0.26368144154548645, "global_step": 53440, "epoch": 600, "lr": 9.087220319856436e-05}
{"train_loss": 0.2535398006439209, "global_step": 53441, "epoch": 600, "lr": 9.08718692629025e-05}
{"train_loss": 0.2841750979423523, "global_step": 53442, "epoch": 600, "lr": 9.087153532174591e-05}
{"train_loss": 0.2988428473472595, "global_step": 53443, "epoch": 600, "lr": 9.087120137509464e-05}
{"train_loss": 0.27148890495300293, "global_step": 53444, "epoch": 600, "lr": 9.087086742294873e-05}
{"train_loss": 0.2970757484436035, "global_step": 53445, "epoch": 600, "lr": 9.087053346530823e-05}
{"train_loss": 0.2890922427177429, "global_step": 53446, "epoch": 600, "lr": 9.08701995021732e-05}
{"train_loss": 0.2839593291282654, "global_step": 53447, "epoch": 600, "lr": 9.086986553354364e-05}
{"train_loss": 0.19228318333625793, "global_step": 53448, "epoch": 600, "lr": 9.086953155941966e-05}
{"train_loss": 0.2855359613895416, "global_step": 53449, "epoch": 600, "lr": 9.086919757980125e-05}
{"train_loss": 0.17037531733512878, "global_step": 53450, "epoch": 600, "lr": 9.086886359468848e-05}
{"train_loss": 0.2152191400527954, "global_step": 53451, "epoch": 600, "lr": 9.086852960408139e-05}
{"train_loss": 0.22693587839603424, "global_step": 53452, "epoch": 600, "lr": 9.086819560798001e-05}
{"train_loss": 0.20737053453922272, "global_step": 53453, "epoch": 600, "lr": 9.086786160638442e-05}
{"train_loss": 0.28262850642204285, "global_step": 53454, "epoch": 600, "lr": 9.086752759929464e-05}
{"train_loss": 0.24169199168682098, "global_step": 53455, "epoch": 600, "lr": 9.08671935867107e-05}
{"train_loss": 0.21220286190509796, "global_step": 53456, "epoch": 600, "lr": 9.086685956863268e-05}
{"train_loss": 0.2109186202287674, "global_step": 53457, "epoch": 600, "lr": 9.086652554506062e-05}
{"train_loss": 0.22194962203502655, "global_step": 53458, "epoch": 600, "lr": 9.086619151599453e-05}
{"train_loss": 0.25610461831092834, "global_step": 53459, "epoch": 600, "lr": 9.08658574814345e-05}
{"train_loss": 0.2823885679244995, "global_step": 53460, "epoch": 600, "lr": 9.086552344138052e-05}
{"train_loss": 0.32240378856658936, "global_step": 53461, "epoch": 600, "lr": 9.086518939583271e-05}
{"train_loss": 0.17674541473388672, "global_step": 53462, "epoch": 600, "lr": 9.086485534479105e-05}
{"train_loss": 0.2594136595726013, "global_step": 53463, "epoch": 600, "lr": 9.086452128825562e-05}
{"train_loss": 0.3886587619781494, "global_step": 53464, "epoch": 600, "lr": 9.086418722622644e-05}
{"train_loss": 0.22476916015148163, "global_step": 53465, "epoch": 600, "lr": 9.086385315870357e-05}
{"train_loss": 0.19245614111423492, "global_step": 53466, "epoch": 600, "lr": 9.086351908568706e-05}
{"train_loss": 0.3310723602771759, "global_step": 53467, "epoch": 600, "lr": 9.086318500717694e-05}
{"train_loss": 0.23279184103012085, "global_step": 53468, "epoch": 600, "lr": 9.086285092317328e-05}
{"train_loss": 0.30113518238067627, "global_step": 53469, "epoch": 600, "lr": 9.086251683367609e-05}
{"train_loss": 0.2617330551147461, "global_step": 53470, "epoch": 600, "lr": 9.086218273868544e-05}
{"train_loss": 0.22946485877037048, "global_step": 53471, "epoch": 600, "lr": 9.086184863820137e-05}
{"train_loss": 0.29559898376464844, "global_step": 53472, "epoch": 600, "lr": 9.086151453222391e-05}
{"train_loss": 0.35601356625556946, "global_step": 53473, "epoch": 600, "lr": 9.086118042075314e-05}
{"train_loss": 0.3096383810043335, "global_step": 53474, "epoch": 600, "lr": 9.086084630378906e-05}
{"train_loss": 0.25892600417137146, "global_step": 53475, "epoch": 600, "lr": 9.086051218133174e-05}
{"train_loss": 0.3139221668243408, "global_step": 53476, "epoch": 600, "lr": 9.086017805338122e-05}
{"train_loss": 0.32157570123672485, "global_step": 53477, "epoch": 600, "lr": 9.085984391993757e-05}
{"train_loss": 0.21094462275505066, "global_step": 53478, "epoch": 600, "lr": 9.08595097810008e-05}
{"train_loss": 0.19955302774906158, "global_step": 53479, "epoch": 600, "lr": 9.085917563657095e-05}
{"train_loss": 0.2666320502758026, "global_step": 53480, "epoch": 600, "lr": 9.08588414866481e-05}
{"train_loss": 0.3145666718482971, "global_step": 53481, "epoch": 600, "lr": 9.085850733123228e-05}
{"train_loss": 0.24089112877845764, "global_step": 53482, "epoch": 600, "lr": 9.085817317032353e-05}
{"train_loss": 0.23638702929019928, "global_step": 53483, "epoch": 600, "lr": 9.085783900392188e-05}
{"train_loss": 0.2728608548641205, "global_step": 53484, "epoch": 600, "lr": 9.085750483202739e-05}
{"train_loss": 0.33658674359321594, "global_step": 53485, "epoch": 600, "lr": 9.085717065464012e-05}
{"train_loss": 0.22276204824447632, "global_step": 53486, "epoch": 600, "lr": 9.085683647176012e-05}
{"train_loss": 0.29555442929267883, "global_step": 53487, "epoch": 600, "lr": 9.08565022833874e-05}
{"train_loss": 0.25461311922984176, "global_step": 53488, "epoch": 600, "lr": 9.085616808952201e-05, "train/sim_max_reward_0": 0.5890700793426988, "train/sim_max_reward_1": 0.002738050832760117, "train/sim_max_reward_2": 0.36903296502258, "train/sim_max_reward_3": 0.46912077600249225, "train/sim_max_reward_4": 0.9958649876214339, "train/sim_max_reward_5": 0.638688570124886, "test/sim_max_reward_4500000": 0.9826880140364384, "test/sim_max_reward_4500001": 0.8940730524902509, "test/sim_max_reward_4500002": 0.8683787694692802, "test/sim_max_reward_4500003": 0.6993852347755463, "test/sim_max_reward_4500004": 0.6556490972739348, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.40086083493358704, "test/sim_max_reward_4500007": 0.4655416138919242, "test/sim_max_reward_4500008": 0.1753928333297681, "test/sim_max_reward_4500009": 0.05044903116462404, "test/sim_max_reward_4500010": 0.01504909034507207, "test/sim_max_reward_4500011": 0.9684222085028575, "test/sim_max_reward_4500012": 0.10999560292403553, "test/sim_max_reward_4500013": 0.40705761200670326, "test/sim_max_reward_4500014": 0.3200072094400632, "test/sim_max_reward_4500015": 0.8724636966128193, "test/sim_max_reward_4500016": 0.5547538291958901, "test/sim_max_reward_4500017": 0.6998029368942259, "test/sim_max_reward_4500018": 0.3628112943015595, "test/sim_max_reward_4500019": 0.48134220907343855, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.947319929979637, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.4961035388638116, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.4642443307180779, "test/sim_max_reward_4500026": 0.8278944819945457, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.4238871942389865, "test/sim_max_reward_4500029": 0.5313430894269521, "test/sim_max_reward_4500030": 0.7041426450418983, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.45983712065198207, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.8076433015806245, "test/sim_max_reward_4500035": 0.6776792459848825, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.4638704550645408, "test/sim_max_reward_4500039": 0.9750811198069996, "test/sim_max_reward_4500040": 0.9590497672473858, "test/sim_max_reward_4500041": 0.9749483490710941, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.5738513680823981, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.2896953510887086, "test/sim_max_reward_4500046": 0.42663224154095314, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.8841195534465972, "test/sim_max_reward_4500049": 0.19427312075972347, "train/mean_score": 0.510752571491142, "test/mean_score": 0.5613148075050365, "val_loss": 3.1972415447235107, "train_action_mse_error": 24.427518844604492}
{"train_loss": 0.2593018710613251, "global_step": 53489, "epoch": 601, "lr": 9.085583389016402e-05}
{"train_loss": 0.25289517641067505, "global_step": 53490, "epoch": 601, "lr": 9.085549968531344e-05}
{"train_loss": 0.3775455355644226, "global_step": 53491, "epoch": 601, "lr": 9.085516547497035e-05}
{"train_loss": 0.27394819259643555, "global_step": 53492, "epoch": 601, "lr": 9.085483125913479e-05}
{"train_loss": 0.20838341116905212, "global_step": 53493, "epoch": 601, "lr": 9.085449703780677e-05}
{"train_loss": 0.23706978559494019, "global_step": 53494, "epoch": 601, "lr": 9.08541628109864e-05}
{"train_loss": 0.32770586013793945, "global_step": 53495, "epoch": 601, "lr": 9.085382857867366e-05}
{"train_loss": 0.2620336711406708, "global_step": 53496, "epoch": 601, "lr": 9.085349434086863e-05}
{"train_loss": 0.33813658356666565, "global_step": 53497, "epoch": 601, "lr": 9.085316009757133e-05}
{"train_loss": 0.28534767031669617, "global_step": 53498, "epoch": 601, "lr": 9.085282584878183e-05}
{"train_loss": 0.25290995836257935, "global_step": 53499, "epoch": 601, "lr": 9.085249159450018e-05}
{"train_loss": 0.2583768665790558, "global_step": 53500, "epoch": 601, "lr": 9.085215733472639e-05}
{"train_loss": 0.334011435508728, "global_step": 53501, "epoch": 601, "lr": 9.085182306946053e-05}
{"train_loss": 0.23704692721366882, "global_step": 53502, "epoch": 601, "lr": 9.085148879870265e-05}
{"train_loss": 0.2362479567527771, "global_step": 53503, "epoch": 601, "lr": 9.085115452245276e-05}
{"train_loss": 0.23131172358989716, "global_step": 53504, "epoch": 601, "lr": 9.085082024071095e-05}
{"train_loss": 0.38861918449401855, "global_step": 53505, "epoch": 601, "lr": 9.085048595347725e-05}
{"train_loss": 0.2522648274898529, "global_step": 53506, "epoch": 601, "lr": 9.08501516607517e-05}
{"train_loss": 0.26425984501838684, "global_step": 53507, "epoch": 601, "lr": 9.084981736253433e-05}
{"train_loss": 0.20698021352291107, "global_step": 53508, "epoch": 601, "lr": 9.084948305882521e-05}
{"train_loss": 0.26039406657218933, "global_step": 53509, "epoch": 601, "lr": 9.084914874962438e-05}
{"train_loss": 0.20803213119506836, "global_step": 53510, "epoch": 601, "lr": 9.084881443493187e-05}
{"train_loss": 0.2277975082397461, "global_step": 53511, "epoch": 601, "lr": 9.084848011474773e-05}
{"train_loss": 0.2513364851474762, "global_step": 53512, "epoch": 601, "lr": 9.084814578907201e-05}
{"train_loss": 0.2561894357204437, "global_step": 53513, "epoch": 601, "lr": 9.084781145790477e-05}
{"train_loss": 0.30917441844940186, "global_step": 53514, "epoch": 601, "lr": 9.084747712124604e-05}
{"train_loss": 0.323269248008728, "global_step": 53515, "epoch": 601, "lr": 9.084714277909584e-05}
{"train_loss": 0.18645751476287842, "global_step": 53516, "epoch": 601, "lr": 9.084680843145428e-05}
{"train_loss": 0.26164475083351135, "global_step": 53517, "epoch": 601, "lr": 9.084647407832134e-05}
{"train_loss": 0.19451704621315002, "global_step": 53518, "epoch": 601, "lr": 9.084613971969708e-05}
{"train_loss": 0.2303357571363449, "global_step": 53519, "epoch": 601, "lr": 9.084580535558156e-05}
{"train_loss": 0.3273323178291321, "global_step": 53520, "epoch": 601, "lr": 9.084547098597482e-05}
{"train_loss": 0.2495419830083847, "global_step": 53521, "epoch": 601, "lr": 9.084513661087691e-05}
{"train_loss": 0.1796562671661377, "global_step": 53522, "epoch": 601, "lr": 9.084480223028787e-05}
{"train_loss": 0.2782129943370819, "global_step": 53523, "epoch": 601, "lr": 9.084446784420774e-05}
{"train_loss": 0.2516227066516876, "global_step": 53524, "epoch": 601, "lr": 9.084413345263657e-05}
{"train_loss": 0.30386340618133545, "global_step": 53525, "epoch": 601, "lr": 9.08437990555744e-05}
{"train_loss": 0.23275209963321686, "global_step": 53526, "epoch": 601, "lr": 9.08434646530213e-05}
{"train_loss": 0.3792687952518463, "global_step": 53527, "epoch": 601, "lr": 9.084313024497726e-05}
{"train_loss": 0.20951615273952484, "global_step": 53528, "epoch": 601, "lr": 9.084279583144237e-05}
{"train_loss": 0.17363081872463226, "global_step": 53529, "epoch": 601, "lr": 9.084246141241668e-05}
{"train_loss": 0.26006829738616943, "global_step": 53530, "epoch": 601, "lr": 9.084212698790021e-05}
{"train_loss": 0.3279125988483429, "global_step": 53531, "epoch": 601, "lr": 9.084179255789302e-05}
{"train_loss": 0.32585084438323975, "global_step": 53532, "epoch": 601, "lr": 9.084145812239513e-05}
{"train_loss": 0.26140108704566956, "global_step": 53533, "epoch": 601, "lr": 9.084112368140662e-05}
{"train_loss": 0.2318139225244522, "global_step": 53534, "epoch": 601, "lr": 9.084078923492751e-05}
{"train_loss": 0.16224604845046997, "global_step": 53535, "epoch": 601, "lr": 9.084045478295787e-05}
{"train_loss": 0.2720882296562195, "global_step": 53536, "epoch": 601, "lr": 9.084012032549771e-05}
{"train_loss": 0.323000431060791, "global_step": 53537, "epoch": 601, "lr": 9.083978586254709e-05}
{"train_loss": 0.2706175446510315, "global_step": 53538, "epoch": 601, "lr": 9.083945139410608e-05}
{"train_loss": 0.19891662895679474, "global_step": 53539, "epoch": 601, "lr": 9.083911692017469e-05}
{"train_loss": 0.20034940540790558, "global_step": 53540, "epoch": 601, "lr": 9.083878244075298e-05}
{"train_loss": 0.2527623772621155, "global_step": 53541, "epoch": 601, "lr": 9.0838447955841e-05}
{"train_loss": 0.1775999665260315, "global_step": 53542, "epoch": 601, "lr": 9.083811346543877e-05}
{"train_loss": 0.21376030147075653, "global_step": 53543, "epoch": 601, "lr": 9.083777896954638e-05}
{"train_loss": 0.31769630312919617, "global_step": 53544, "epoch": 601, "lr": 9.083744446816382e-05}
{"train_loss": 0.2661837339401245, "global_step": 53545, "epoch": 601, "lr": 9.083710996129118e-05}
{"train_loss": 0.26127511262893677, "global_step": 53546, "epoch": 601, "lr": 9.083677544892848e-05}
{"train_loss": 0.282016783952713, "global_step": 53547, "epoch": 601, "lr": 9.083644093107579e-05}
{"train_loss": 0.24312172830104828, "global_step": 53548, "epoch": 601, "lr": 9.083610640773312e-05}
{"train_loss": 0.24891553819179535, "global_step": 53549, "epoch": 601, "lr": 9.083577187890054e-05}
{"train_loss": 0.22342629730701447, "global_step": 53550, "epoch": 601, "lr": 9.08354373445781e-05}
{"train_loss": 0.29988160729408264, "global_step": 53551, "epoch": 601, "lr": 9.083510280476582e-05}
{"train_loss": 0.3621574640274048, "global_step": 53552, "epoch": 601, "lr": 9.083476825946375e-05}
{"train_loss": 0.312970906496048, "global_step": 53553, "epoch": 601, "lr": 9.083443370867195e-05}
{"train_loss": 0.19646859169006348, "global_step": 53554, "epoch": 601, "lr": 9.083409915239047e-05}
{"train_loss": 0.310031920671463, "global_step": 53555, "epoch": 601, "lr": 9.083376459061934e-05}
{"train_loss": 0.27523529529571533, "global_step": 53556, "epoch": 601, "lr": 9.08334300233586e-05}
{"train_loss": 0.27015647292137146, "global_step": 53557, "epoch": 601, "lr": 9.08330954506083e-05}
{"train_loss": 0.30339041352272034, "global_step": 53558, "epoch": 601, "lr": 9.083276087236849e-05}
{"train_loss": 0.30781450867652893, "global_step": 53559, "epoch": 601, "lr": 9.083242628863922e-05}
{"train_loss": 0.22316189110279083, "global_step": 53560, "epoch": 601, "lr": 9.083209169942053e-05}
{"train_loss": 0.30608251690864563, "global_step": 53561, "epoch": 601, "lr": 9.083175710471246e-05}
{"train_loss": 0.21649020910263062, "global_step": 53562, "epoch": 601, "lr": 9.083142250451504e-05}
{"train_loss": 0.2620566487312317, "global_step": 53563, "epoch": 601, "lr": 9.083108789882835e-05}
{"train_loss": 0.19812773168087006, "global_step": 53564, "epoch": 601, "lr": 9.083075328765243e-05}
{"train_loss": 0.2934935688972473, "global_step": 53565, "epoch": 601, "lr": 9.083041867098729e-05}
{"train_loss": 0.28812137246131897, "global_step": 53566, "epoch": 601, "lr": 9.0830084048833e-05}
{"train_loss": 0.31072282791137695, "global_step": 53567, "epoch": 601, "lr": 9.082974942118961e-05}
{"train_loss": 0.23722529411315918, "global_step": 53568, "epoch": 601, "lr": 9.082941478805715e-05}
{"train_loss": 0.22771649062633514, "global_step": 53569, "epoch": 601, "lr": 9.08290801494357e-05}
{"train_loss": 0.2261466681957245, "global_step": 53570, "epoch": 601, "lr": 9.082874550532526e-05}
{"train_loss": 0.20947502553462982, "global_step": 53571, "epoch": 601, "lr": 9.082841085572588e-05}
{"train_loss": 0.21577255427837372, "global_step": 53572, "epoch": 601, "lr": 9.082807620063763e-05}
{"train_loss": 0.18703855574131012, "global_step": 53573, "epoch": 601, "lr": 9.082774154006055e-05}
{"train_loss": 0.2947199046611786, "global_step": 53574, "epoch": 601, "lr": 9.082740687399466e-05}
{"train_loss": 0.2538924217224121, "global_step": 53575, "epoch": 601, "lr": 9.082707220244003e-05}
{"train_loss": 0.23488742113113403, "global_step": 53576, "epoch": 601, "lr": 9.08267375253967e-05}
{"train_loss": 0.2610475541500563, "global_step": 53577, "epoch": 601, "lr": 9.082640284286473e-05, "val_loss": 3.2241623401641846}
{"train_loss": 0.2638871371746063, "global_step": 53578, "epoch": 602, "lr": 9.082606815484411e-05}
{"train_loss": 0.2449750006198883, "global_step": 53579, "epoch": 602, "lr": 9.082573346133493e-05}
{"train_loss": 0.200270414352417, "global_step": 53580, "epoch": 602, "lr": 9.082539876233725e-05}
{"train_loss": 0.2513445317745209, "global_step": 53581, "epoch": 602, "lr": 9.082506405785109e-05}
{"train_loss": 0.3421015441417694, "global_step": 53582, "epoch": 602, "lr": 9.08247293478765e-05}
{"train_loss": 0.32406994700431824, "global_step": 53583, "epoch": 602, "lr": 9.08243946324135e-05}
{"train_loss": 0.2648116946220398, "global_step": 53584, "epoch": 602, "lr": 9.082405991146218e-05}
{"train_loss": 0.15615861117839813, "global_step": 53585, "epoch": 602, "lr": 9.082372518502255e-05}
{"train_loss": 0.24226915836334229, "global_step": 53586, "epoch": 602, "lr": 9.082339045309467e-05}
{"train_loss": 0.3513393700122833, "global_step": 53587, "epoch": 602, "lr": 9.082305571567859e-05}
{"train_loss": 0.3359518349170685, "global_step": 53588, "epoch": 602, "lr": 9.082272097277434e-05}
{"train_loss": 0.3296782970428467, "global_step": 53589, "epoch": 602, "lr": 9.082238622438198e-05}
{"train_loss": 0.28855186700820923, "global_step": 53590, "epoch": 602, "lr": 9.082205147050153e-05}
{"train_loss": 0.23335953056812286, "global_step": 53591, "epoch": 602, "lr": 9.082171671113307e-05}
{"train_loss": 0.24535757303237915, "global_step": 53592, "epoch": 602, "lr": 9.082138194627663e-05}
{"train_loss": 0.20775331556797028, "global_step": 53593, "epoch": 602, "lr": 9.082104717593225e-05}
{"train_loss": 0.23158082365989685, "global_step": 53594, "epoch": 602, "lr": 9.082071240009997e-05}
{"train_loss": 0.28460538387298584, "global_step": 53595, "epoch": 602, "lr": 9.082037761877983e-05}
{"train_loss": 0.2923407554626465, "global_step": 53596, "epoch": 602, "lr": 9.082004283197192e-05}
{"train_loss": 0.29851874709129333, "global_step": 53597, "epoch": 602, "lr": 9.081970803967623e-05}
{"train_loss": 0.29525554180145264, "global_step": 53598, "epoch": 602, "lr": 9.081937324189285e-05}
{"train_loss": 0.27856722474098206, "global_step": 53599, "epoch": 602, "lr": 9.081903843862178e-05}
{"train_loss": 0.2732224762439728, "global_step": 53600, "epoch": 602, "lr": 9.081870362986308e-05}
{"train_loss": 0.23189371824264526, "global_step": 53601, "epoch": 602, "lr": 9.081836881561682e-05}
{"train_loss": 0.2747870683670044, "global_step": 53602, "epoch": 602, "lr": 9.081803399588304e-05}
{"train_loss": 0.2246922254562378, "global_step": 53603, "epoch": 602, "lr": 9.081769917066176e-05}
{"train_loss": 0.2010951191186905, "global_step": 53604, "epoch": 602, "lr": 9.081736433995303e-05}
{"train_loss": 0.25002217292785645, "global_step": 53605, "epoch": 602, "lr": 9.081702950375691e-05}
{"train_loss": 0.1565150022506714, "global_step": 53606, "epoch": 602, "lr": 9.081669466207343e-05}
{"train_loss": 0.351182758808136, "global_step": 53607, "epoch": 602, "lr": 9.081635981490266e-05}
{"train_loss": 0.20125429332256317, "global_step": 53608, "epoch": 602, "lr": 9.081602496224461e-05}
{"train_loss": 0.215728297829628, "global_step": 53609, "epoch": 602, "lr": 9.081569010409936e-05}
{"train_loss": 0.22471100091934204, "global_step": 53610, "epoch": 602, "lr": 9.081535524046693e-05}
{"train_loss": 0.19805653393268585, "global_step": 53611, "epoch": 602, "lr": 9.081502037134737e-05}
{"train_loss": 0.24644321203231812, "global_step": 53612, "epoch": 602, "lr": 9.081468549674073e-05}
{"train_loss": 0.356327086687088, "global_step": 53613, "epoch": 602, "lr": 9.081435061664705e-05}
{"train_loss": 0.23821164667606354, "global_step": 53614, "epoch": 602, "lr": 9.081401573106638e-05}
{"train_loss": 0.2710590660572052, "global_step": 53615, "epoch": 602, "lr": 9.081368083999877e-05}
{"train_loss": 0.20963238179683685, "global_step": 53616, "epoch": 602, "lr": 9.081334594344425e-05}
{"train_loss": 0.27373456954956055, "global_step": 53617, "epoch": 602, "lr": 9.081301104140287e-05}
{"train_loss": 0.2571733295917511, "global_step": 53618, "epoch": 602, "lr": 9.081267613387469e-05}
{"train_loss": 0.2765916585922241, "global_step": 53619, "epoch": 602, "lr": 9.081234122085974e-05}
{"train_loss": 0.29587358236312866, "global_step": 53620, "epoch": 602, "lr": 9.081200630235806e-05}
{"train_loss": 0.2791644036769867, "global_step": 53621, "epoch": 602, "lr": 9.08116713783697e-05}
{"train_loss": 0.24314023554325104, "global_step": 53622, "epoch": 602, "lr": 9.081133644889473e-05}
{"train_loss": 0.23508690297603607, "global_step": 53623, "epoch": 602, "lr": 9.081100151393315e-05}
{"train_loss": 0.23470903933048248, "global_step": 53624, "epoch": 602, "lr": 9.081066657348504e-05}
{"train_loss": 0.3134962320327759, "global_step": 53625, "epoch": 602, "lr": 9.081033162755043e-05}
{"train_loss": 0.2630980610847473, "global_step": 53626, "epoch": 602, "lr": 9.080999667612937e-05}
{"train_loss": 0.2691913843154907, "global_step": 53627, "epoch": 602, "lr": 9.080966171922189e-05}
{"train_loss": 0.23749037086963654, "global_step": 53628, "epoch": 602, "lr": 9.080932675682807e-05}
{"train_loss": 0.2086237221956253, "global_step": 53629, "epoch": 602, "lr": 9.080899178894793e-05}
{"train_loss": 0.2319062054157257, "global_step": 53630, "epoch": 602, "lr": 9.08086568155815e-05}
{"train_loss": 0.32622742652893066, "global_step": 53631, "epoch": 602, "lr": 9.080832183672887e-05}
{"train_loss": 0.2488347440958023, "global_step": 53632, "epoch": 602, "lr": 9.080798685239005e-05}
{"train_loss": 0.23565351963043213, "global_step": 53633, "epoch": 602, "lr": 9.080765186256508e-05}
{"train_loss": 0.16413037478923798, "global_step": 53634, "epoch": 602, "lr": 9.080731686725403e-05}
{"train_loss": 0.24790717661380768, "global_step": 53635, "epoch": 602, "lr": 9.080698186645693e-05}
{"train_loss": 0.30568209290504456, "global_step": 53636, "epoch": 602, "lr": 9.080664686017382e-05}
{"train_loss": 0.23857849836349487, "global_step": 53637, "epoch": 602, "lr": 9.080631184840476e-05}
{"train_loss": 0.20137415826320648, "global_step": 53638, "epoch": 602, "lr": 9.080597683114979e-05}
{"train_loss": 0.20574891567230225, "global_step": 53639, "epoch": 602, "lr": 9.080564180840896e-05}
{"train_loss": 0.2607092261314392, "global_step": 53640, "epoch": 602, "lr": 9.080530678018231e-05}
{"train_loss": 0.2671942710876465, "global_step": 53641, "epoch": 602, "lr": 9.080497174646986e-05}
{"train_loss": 0.294009268283844, "global_step": 53642, "epoch": 602, "lr": 9.080463670727171e-05}
{"train_loss": 0.2149963676929474, "global_step": 53643, "epoch": 602, "lr": 9.080430166258786e-05}
{"train_loss": 0.263016015291214, "global_step": 53644, "epoch": 602, "lr": 9.080396661241837e-05}
{"train_loss": 0.27430033683776855, "global_step": 53645, "epoch": 602, "lr": 9.080363155676328e-05}
{"train_loss": 0.18762435019016266, "global_step": 53646, "epoch": 602, "lr": 9.080329649562264e-05}
{"train_loss": 0.16438888013362885, "global_step": 53647, "epoch": 602, "lr": 9.08029614289965e-05}
{"train_loss": 0.23169681429862976, "global_step": 53648, "epoch": 602, "lr": 9.08026263568849e-05}
{"train_loss": 0.21136008203029633, "global_step": 53649, "epoch": 602, "lr": 9.080229127928788e-05}
{"train_loss": 0.3004346489906311, "global_step": 53650, "epoch": 602, "lr": 9.08019561962055e-05}
{"train_loss": 0.20904234051704407, "global_step": 53651, "epoch": 602, "lr": 9.080162110763777e-05}
{"train_loss": 0.24423480033874512, "global_step": 53652, "epoch": 602, "lr": 9.080128601358478e-05}
{"train_loss": 0.2577302157878876, "global_step": 53653, "epoch": 602, "lr": 9.080095091404655e-05}
{"train_loss": 0.35868439078330994, "global_step": 53654, "epoch": 602, "lr": 9.080061580902313e-05}
{"train_loss": 0.27004796266555786, "global_step": 53655, "epoch": 602, "lr": 9.080028069851456e-05}
{"train_loss": 0.1997823417186737, "global_step": 53656, "epoch": 602, "lr": 9.079994558252089e-05}
{"train_loss": 0.3125908076763153, "global_step": 53657, "epoch": 602, "lr": 9.079961046104216e-05}
{"train_loss": 0.26229891180992126, "global_step": 53658, "epoch": 602, "lr": 9.079927533407842e-05}
{"train_loss": 0.25197634100914, "global_step": 53659, "epoch": 602, "lr": 9.079894020162973e-05}
{"train_loss": 0.2159147560596466, "global_step": 53660, "epoch": 602, "lr": 9.079860506369611e-05}
{"train_loss": 0.25495821237564087, "global_step": 53661, "epoch": 602, "lr": 9.079826992027761e-05}
{"train_loss": 0.16074129939079285, "global_step": 53662, "epoch": 602, "lr": 9.079793477137428e-05}
{"train_loss": 0.19673468172550201, "global_step": 53663, "epoch": 602, "lr": 9.079759961698617e-05}
{"train_loss": 0.2987912893295288, "global_step": 53664, "epoch": 602, "lr": 9.079726445711332e-05}
{"train_loss": 0.14844070374965668, "global_step": 53665, "epoch": 602, "lr": 9.079692929175579e-05}
{"train_loss": 0.25251267816913264, "global_step": 53666, "epoch": 602, "lr": 9.079659412091357e-05, "val_loss": 3.3092942237854004}
{"train_loss": 0.2809644937515259, "global_step": 53667, "epoch": 603, "lr": 9.079625894458679e-05}
{"train_loss": 0.21439681947231293, "global_step": 53668, "epoch": 603, "lr": 9.079592376277542e-05}
{"train_loss": 0.3116876780986786, "global_step": 53669, "epoch": 603, "lr": 9.079558857547955e-05}
{"train_loss": 0.2653694152832031, "global_step": 53670, "epoch": 603, "lr": 9.079525338269921e-05}
{"train_loss": 0.2648938298225403, "global_step": 53671, "epoch": 603, "lr": 9.079491818443445e-05}
{"train_loss": 0.2396029531955719, "global_step": 53672, "epoch": 603, "lr": 9.079458298068528e-05}
{"train_loss": 0.24952593445777893, "global_step": 53673, "epoch": 603, "lr": 9.07942477714518e-05}
{"train_loss": 0.25052037835121155, "global_step": 53674, "epoch": 603, "lr": 9.079391255673403e-05}
{"train_loss": 0.22333699464797974, "global_step": 53675, "epoch": 603, "lr": 9.079357733653201e-05}
{"train_loss": 0.3505372703075409, "global_step": 53676, "epoch": 603, "lr": 9.079324211084579e-05}
{"train_loss": 0.21858243644237518, "global_step": 53677, "epoch": 603, "lr": 9.079290687967543e-05}
{"train_loss": 0.22955407202243805, "global_step": 53678, "epoch": 603, "lr": 9.079257164302096e-05}
{"train_loss": 0.13400787115097046, "global_step": 53679, "epoch": 603, "lr": 9.079223640088241e-05}
{"train_loss": 0.22870558500289917, "global_step": 53680, "epoch": 603, "lr": 9.079190115325983e-05}
{"train_loss": 0.20564879477024078, "global_step": 53681, "epoch": 603, "lr": 9.079156590015331e-05}
{"train_loss": 0.17397603392601013, "global_step": 53682, "epoch": 603, "lr": 9.079123064156285e-05}
{"train_loss": 0.1772710233926773, "global_step": 53683, "epoch": 603, "lr": 9.079089537748848e-05}
{"train_loss": 0.3170950412750244, "global_step": 53684, "epoch": 603, "lr": 9.07905601079303e-05}
{"train_loss": 0.4065932035446167, "global_step": 53685, "epoch": 603, "lr": 9.079022483288833e-05}
{"train_loss": 0.3674078583717346, "global_step": 53686, "epoch": 603, "lr": 9.078988955236258e-05}
{"train_loss": 0.31731677055358887, "global_step": 53687, "epoch": 603, "lr": 9.078955426635317e-05}
{"train_loss": 0.22444254159927368, "global_step": 53688, "epoch": 603, "lr": 9.078921897486007e-05}
{"train_loss": 0.26112881302833557, "global_step": 53689, "epoch": 603, "lr": 9.078888367788336e-05}
{"train_loss": 0.3285612463951111, "global_step": 53690, "epoch": 603, "lr": 9.078854837542309e-05}
{"train_loss": 0.2999007999897003, "global_step": 53691, "epoch": 603, "lr": 9.07882130674793e-05}
{"train_loss": 0.29716020822525024, "global_step": 53692, "epoch": 603, "lr": 9.078787775405202e-05}
{"train_loss": 0.340300977230072, "global_step": 53693, "epoch": 603, "lr": 9.078754243514132e-05}
{"train_loss": 0.2404487282037735, "global_step": 53694, "epoch": 603, "lr": 9.078720711074722e-05}
{"train_loss": 0.2612968981266022, "global_step": 53695, "epoch": 603, "lr": 9.078687178086978e-05}
{"train_loss": 0.3370286822319031, "global_step": 53696, "epoch": 603, "lr": 9.078653644550903e-05}
{"train_loss": 0.28919902443885803, "global_step": 53697, "epoch": 603, "lr": 9.078620110466505e-05}
{"train_loss": 0.1767059713602066, "global_step": 53698, "epoch": 603, "lr": 9.078586575833786e-05}
{"train_loss": 0.3173104226589203, "global_step": 53699, "epoch": 603, "lr": 9.07855304065275e-05}
{"train_loss": 0.2865612804889679, "global_step": 53700, "epoch": 603, "lr": 9.078519504923401e-05}
{"train_loss": 0.18254578113555908, "global_step": 53701, "epoch": 603, "lr": 9.078485968645747e-05}
{"train_loss": 0.4058226943016052, "global_step": 53702, "epoch": 603, "lr": 9.078452431819788e-05}
{"train_loss": 0.281772643327713, "global_step": 53703, "epoch": 603, "lr": 9.078418894445533e-05}
{"train_loss": 0.37998315691947937, "global_step": 53704, "epoch": 603, "lr": 9.078385356522984e-05}
{"train_loss": 0.2253195196390152, "global_step": 53705, "epoch": 603, "lr": 9.078351818052144e-05}
{"train_loss": 0.21511133015155792, "global_step": 53706, "epoch": 603, "lr": 9.078318279033021e-05}
{"train_loss": 0.2579042315483093, "global_step": 53707, "epoch": 603, "lr": 9.078284739465618e-05}
{"train_loss": 0.24387948215007782, "global_step": 53708, "epoch": 603, "lr": 9.078251199349938e-05}
{"train_loss": 0.2428753823041916, "global_step": 53709, "epoch": 603, "lr": 9.078217658685988e-05}
{"train_loss": 0.21627165377140045, "global_step": 53710, "epoch": 603, "lr": 9.078184117473772e-05}
{"train_loss": 0.17994743585586548, "global_step": 53711, "epoch": 603, "lr": 9.078150575713291e-05}
{"train_loss": 0.3554410934448242, "global_step": 53712, "epoch": 603, "lr": 9.078117033404554e-05}
{"train_loss": 0.3020809292793274, "global_step": 53713, "epoch": 603, "lr": 9.078083490547564e-05}
{"train_loss": 0.20138804614543915, "global_step": 53714, "epoch": 603, "lr": 9.078049947142327e-05}
{"train_loss": 0.32315972447395325, "global_step": 53715, "epoch": 603, "lr": 9.078016403188842e-05}
{"train_loss": 0.21125736832618713, "global_step": 53716, "epoch": 603, "lr": 9.077982858687121e-05}
{"train_loss": 0.28885024785995483, "global_step": 53717, "epoch": 603, "lr": 9.077949313637164e-05}
{"train_loss": 0.24129122495651245, "global_step": 53718, "epoch": 603, "lr": 9.077915768038975e-05}
{"train_loss": 0.18064339458942413, "global_step": 53719, "epoch": 603, "lr": 9.077882221892561e-05}
{"train_loss": 0.2768286466598511, "global_step": 53720, "epoch": 603, "lr": 9.077848675197925e-05}
{"train_loss": 0.271175354719162, "global_step": 53721, "epoch": 603, "lr": 9.077815127955071e-05}
{"train_loss": 0.2918820381164551, "global_step": 53722, "epoch": 603, "lr": 9.077781580164007e-05}
{"train_loss": 0.20149238407611847, "global_step": 53723, "epoch": 603, "lr": 9.077748031824734e-05}
{"train_loss": 0.25638407468795776, "global_step": 53724, "epoch": 603, "lr": 9.077714482937256e-05}
{"train_loss": 0.32601094245910645, "global_step": 53725, "epoch": 603, "lr": 9.077680933501581e-05}
{"train_loss": 0.34975665807724, "global_step": 53726, "epoch": 603, "lr": 9.077647383517711e-05}
{"train_loss": 0.23580223321914673, "global_step": 53727, "epoch": 603, "lr": 9.07761383298565e-05}
{"train_loss": 0.21980349719524384, "global_step": 53728, "epoch": 603, "lr": 9.077580281905404e-05}
{"train_loss": 0.27393776178359985, "global_step": 53729, "epoch": 603, "lr": 9.077546730276977e-05}
{"train_loss": 0.2587277591228485, "global_step": 53730, "epoch": 603, "lr": 9.077513178100374e-05}
{"train_loss": 0.3564072251319885, "global_step": 53731, "epoch": 603, "lr": 9.077479625375599e-05}
{"train_loss": 0.23111726343631744, "global_step": 53732, "epoch": 603, "lr": 9.077446072102656e-05}
{"train_loss": 0.14143960177898407, "global_step": 53733, "epoch": 603, "lr": 9.077412518281551e-05}
{"train_loss": 0.31788137555122375, "global_step": 53734, "epoch": 603, "lr": 9.077378963912287e-05}
{"train_loss": 0.21195881068706512, "global_step": 53735, "epoch": 603, "lr": 9.077345408994869e-05}
{"train_loss": 0.28088489174842834, "global_step": 53736, "epoch": 603, "lr": 9.077311853529302e-05}
{"train_loss": 0.24446992576122284, "global_step": 53737, "epoch": 603, "lr": 9.07727829751559e-05}
{"train_loss": 0.2737380266189575, "global_step": 53738, "epoch": 603, "lr": 9.077244740953736e-05}
{"train_loss": 0.27289363741874695, "global_step": 53739, "epoch": 603, "lr": 9.077211183843748e-05}
{"train_loss": 0.23373520374298096, "global_step": 53740, "epoch": 603, "lr": 9.077177626185628e-05}
{"train_loss": 0.3427930474281311, "global_step": 53741, "epoch": 603, "lr": 9.077144067979382e-05}
{"train_loss": 0.3473779857158661, "global_step": 53742, "epoch": 603, "lr": 9.077110509225013e-05}
{"train_loss": 0.26136577129364014, "global_step": 53743, "epoch": 603, "lr": 9.077076949922526e-05}
{"train_loss": 0.17314587533473969, "global_step": 53744, "epoch": 603, "lr": 9.077043390071927e-05}
{"train_loss": 0.324319452047348, "global_step": 53745, "epoch": 603, "lr": 9.077009829673218e-05}
{"train_loss": 0.2543083131313324, "global_step": 53746, "epoch": 603, "lr": 9.076976268726405e-05}
{"train_loss": 0.2782990336418152, "global_step": 53747, "epoch": 603, "lr": 9.076942707231492e-05}
{"train_loss": 0.31541308760643005, "global_step": 53748, "epoch": 603, "lr": 9.076909145188484e-05}
{"train_loss": 0.30746006965637207, "global_step": 53749, "epoch": 603, "lr": 9.076875582597384e-05}
{"train_loss": 0.2656322419643402, "global_step": 53750, "epoch": 603, "lr": 9.076842019458199e-05}
{"train_loss": 0.22992829978466034, "global_step": 53751, "epoch": 603, "lr": 9.076808455770933e-05}
{"train_loss": 0.2868103086948395, "global_step": 53752, "epoch": 603, "lr": 9.076774891535588e-05}
{"train_loss": 0.26479730010032654, "global_step": 53753, "epoch": 603, "lr": 9.076741326752172e-05}
{"train_loss": 0.25050780177116394, "global_step": 53754, "epoch": 603, "lr": 9.076707761420686e-05}
{"train_loss": 0.2659625581810983, "global_step": 53755, "epoch": 603, "lr": 9.076674195541138e-05, "val_loss": 3.242594003677368}
{"train_loss": 0.2581035792827606, "global_step": 53756, "epoch": 604, "lr": 9.07664062911353e-05}
{"train_loss": 0.2503792345523834, "global_step": 53757, "epoch": 604, "lr": 9.076607062137868e-05}
{"train_loss": 0.22825461626052856, "global_step": 53758, "epoch": 604, "lr": 9.076573494614156e-05}
{"train_loss": 0.2087918221950531, "global_step": 53759, "epoch": 604, "lr": 9.076539926542397e-05}
{"train_loss": 0.1997915357351303, "global_step": 53760, "epoch": 604, "lr": 9.076506357922598e-05}
{"train_loss": 0.23369958996772766, "global_step": 53761, "epoch": 604, "lr": 9.076472788754761e-05}
{"train_loss": 0.23732000589370728, "global_step": 53762, "epoch": 604, "lr": 9.076439219038894e-05}
{"train_loss": 0.2546796202659607, "global_step": 53763, "epoch": 604, "lr": 9.076405648775e-05}
{"train_loss": 0.15596753358840942, "global_step": 53764, "epoch": 604, "lr": 9.07637207796308e-05}
{"train_loss": 0.24368713796138763, "global_step": 53765, "epoch": 604, "lr": 9.076338506603144e-05}
{"train_loss": 0.2477918565273285, "global_step": 53766, "epoch": 604, "lr": 9.076304934695192e-05}
{"train_loss": 0.24195928871631622, "global_step": 53767, "epoch": 604, "lr": 9.076271362239231e-05}
{"train_loss": 0.3634716272354126, "global_step": 53768, "epoch": 604, "lr": 9.076237789235266e-05}
{"train_loss": 0.21957889199256897, "global_step": 53769, "epoch": 604, "lr": 9.076204215683299e-05}
{"train_loss": 0.2051142156124115, "global_step": 53770, "epoch": 604, "lr": 9.076170641583338e-05}
{"train_loss": 0.21820451319217682, "global_step": 53771, "epoch": 604, "lr": 9.076137066935383e-05}
{"train_loss": 0.25609683990478516, "global_step": 53772, "epoch": 604, "lr": 9.076103491739442e-05}
{"train_loss": 0.2169167399406433, "global_step": 53773, "epoch": 604, "lr": 9.07606991599552e-05}
{"train_loss": 0.28177279233932495, "global_step": 53774, "epoch": 604, "lr": 9.07603633970362e-05}
{"train_loss": 0.2946661412715912, "global_step": 53775, "epoch": 604, "lr": 9.076002762863745e-05}
{"train_loss": 0.3299980163574219, "global_step": 53776, "epoch": 604, "lr": 9.075969185475902e-05}
{"train_loss": 0.28104129433631897, "global_step": 53777, "epoch": 604, "lr": 9.075935607540094e-05}
{"train_loss": 0.23496025800704956, "global_step": 53778, "epoch": 604, "lr": 9.075902029056327e-05}
{"train_loss": 0.2638418674468994, "global_step": 53779, "epoch": 604, "lr": 9.075868450024605e-05}
{"train_loss": 0.24433203041553497, "global_step": 53780, "epoch": 604, "lr": 9.075834870444931e-05}
{"train_loss": 0.22314453125, "global_step": 53781, "epoch": 604, "lr": 9.07580129031731e-05}
{"train_loss": 0.3546924293041229, "global_step": 53782, "epoch": 604, "lr": 9.07576770964175e-05}
{"train_loss": 0.3060980439186096, "global_step": 53783, "epoch": 604, "lr": 9.07573412841825e-05}
{"train_loss": 0.25517407059669495, "global_step": 53784, "epoch": 604, "lr": 9.075700546646818e-05}
{"train_loss": 0.2947279214859009, "global_step": 53785, "epoch": 604, "lr": 9.075666964327458e-05}
{"train_loss": 0.23333418369293213, "global_step": 53786, "epoch": 604, "lr": 9.075633381460173e-05}
{"train_loss": 0.21660660207271576, "global_step": 53787, "epoch": 604, "lr": 9.075599798044971e-05}
{"train_loss": 0.26465439796447754, "global_step": 53788, "epoch": 604, "lr": 9.075566214081854e-05}
{"train_loss": 0.2634837031364441, "global_step": 53789, "epoch": 604, "lr": 9.075532629570825e-05}
{"train_loss": 0.35280731320381165, "global_step": 53790, "epoch": 604, "lr": 9.075499044511892e-05}
{"train_loss": 0.27004873752593994, "global_step": 53791, "epoch": 604, "lr": 9.075465458905057e-05}
{"train_loss": 0.2170766144990921, "global_step": 53792, "epoch": 604, "lr": 9.075431872750326e-05}
{"train_loss": 0.280364453792572, "global_step": 53793, "epoch": 604, "lr": 9.075398286047702e-05}
{"train_loss": 0.2929791510105133, "global_step": 53794, "epoch": 604, "lr": 9.075364698797191e-05}
{"train_loss": 0.20433063805103302, "global_step": 53795, "epoch": 604, "lr": 9.075331110998797e-05}
{"train_loss": 0.24478548765182495, "global_step": 53796, "epoch": 604, "lr": 9.075297522652524e-05}
{"train_loss": 0.2115674912929535, "global_step": 53797, "epoch": 604, "lr": 9.075263933758376e-05}
{"train_loss": 0.30261823534965515, "global_step": 53798, "epoch": 604, "lr": 9.07523034431636e-05}
{"train_loss": 0.26517295837402344, "global_step": 53799, "epoch": 604, "lr": 9.075196754326479e-05}
{"train_loss": 0.2584397792816162, "global_step": 53800, "epoch": 604, "lr": 9.075163163788735e-05}
{"train_loss": 0.15847556293010712, "global_step": 53801, "epoch": 604, "lr": 9.075129572703138e-05}
{"train_loss": 0.27148571610450745, "global_step": 53802, "epoch": 604, "lr": 9.075095981069687e-05}
{"train_loss": 0.2802131772041321, "global_step": 53803, "epoch": 604, "lr": 9.075062388888391e-05}
{"train_loss": 0.1733214557170868, "global_step": 53804, "epoch": 604, "lr": 9.075028796159253e-05}
{"train_loss": 0.22716304659843445, "global_step": 53805, "epoch": 604, "lr": 9.074995202882275e-05}
{"train_loss": 0.3843390941619873, "global_step": 53806, "epoch": 604, "lr": 9.074961609057464e-05}
{"train_loss": 0.32343775033950806, "global_step": 53807, "epoch": 604, "lr": 9.074928014684825e-05}
{"train_loss": 0.18698465824127197, "global_step": 53808, "epoch": 604, "lr": 9.074894419764362e-05}
{"train_loss": 0.3007889688014984, "global_step": 53809, "epoch": 604, "lr": 9.074860824296076e-05}
{"train_loss": 0.1860024780035019, "global_step": 53810, "epoch": 604, "lr": 9.074827228279979e-05}
{"train_loss": 0.26311132311820984, "global_step": 53811, "epoch": 604, "lr": 9.074793631716068e-05}
{"train_loss": 0.21271973848342896, "global_step": 53812, "epoch": 604, "lr": 9.074760034604352e-05}
{"train_loss": 0.20116335153579712, "global_step": 53813, "epoch": 604, "lr": 9.074726436944833e-05}
{"train_loss": 0.24632267653942108, "global_step": 53814, "epoch": 604, "lr": 9.074692838737517e-05}
{"train_loss": 0.270272821187973, "global_step": 53815, "epoch": 604, "lr": 9.07465923998241e-05}
{"train_loss": 0.3269912302494049, "global_step": 53816, "epoch": 604, "lr": 9.074625640679513e-05}
{"train_loss": 0.23171356320381165, "global_step": 53817, "epoch": 604, "lr": 9.074592040828834e-05}
{"train_loss": 0.2005854845046997, "global_step": 53818, "epoch": 604, "lr": 9.074558440430374e-05}
{"train_loss": 0.2508881688117981, "global_step": 53819, "epoch": 604, "lr": 9.07452483948414e-05}
{"train_loss": 0.2726244032382965, "global_step": 53820, "epoch": 604, "lr": 9.074491237990135e-05}
{"train_loss": 0.2557812035083771, "global_step": 53821, "epoch": 604, "lr": 9.074457635948364e-05}
{"train_loss": 0.1734822690486908, "global_step": 53822, "epoch": 604, "lr": 9.074424033358834e-05}
{"train_loss": 0.3368184268474579, "global_step": 53823, "epoch": 604, "lr": 9.074390430221545e-05}
{"train_loss": 0.2789292633533478, "global_step": 53824, "epoch": 604, "lr": 9.074356826536506e-05}
{"train_loss": 0.2705746293067932, "global_step": 53825, "epoch": 604, "lr": 9.074323222303718e-05}
{"train_loss": 0.21774427592754364, "global_step": 53826, "epoch": 604, "lr": 9.074289617523186e-05}
{"train_loss": 0.27835172414779663, "global_step": 53827, "epoch": 604, "lr": 9.074256012194917e-05}
{"train_loss": 0.3106186091899872, "global_step": 53828, "epoch": 604, "lr": 9.074222406318914e-05}
{"train_loss": 0.2150861620903015, "global_step": 53829, "epoch": 604, "lr": 9.07418879989518e-05}
{"train_loss": 0.2585286498069763, "global_step": 53830, "epoch": 604, "lr": 9.074155192923724e-05}
{"train_loss": 0.268736869096756, "global_step": 53831, "epoch": 604, "lr": 9.074121585404544e-05}
{"train_loss": 0.2680186927318573, "global_step": 53832, "epoch": 604, "lr": 9.074087977337651e-05}
{"train_loss": 0.38816604018211365, "global_step": 53833, "epoch": 604, "lr": 9.074054368723044e-05}
{"train_loss": 0.24795681238174438, "global_step": 53834, "epoch": 604, "lr": 9.07402075956073e-05}
{"train_loss": 0.27369895577430725, "global_step": 53835, "epoch": 604, "lr": 9.073987149850716e-05}
{"train_loss": 0.20136113464832306, "global_step": 53836, "epoch": 604, "lr": 9.073953539593004e-05}
{"train_loss": 0.20839941501617432, "global_step": 53837, "epoch": 604, "lr": 9.073919928787596e-05}
{"train_loss": 0.23974354565143585, "global_step": 53838, "epoch": 604, "lr": 9.0738863174345e-05}
{"train_loss": 0.26972562074661255, "global_step": 53839, "epoch": 604, "lr": 9.073852705533721e-05}
{"train_loss": 0.3088774085044861, "global_step": 53840, "epoch": 604, "lr": 9.07381909308526e-05}
{"train_loss": 0.26841360330581665, "global_step": 53841, "epoch": 604, "lr": 9.073785480089125e-05}
{"train_loss": 0.23441779613494873, "global_step": 53842, "epoch": 604, "lr": 9.073751866545319e-05}
{"train_loss": 0.26573818922042847, "global_step": 53843, "epoch": 604, "lr": 9.073718252453846e-05}
{"train_loss": 0.2557149161113782, "global_step": 53844, "epoch": 604, "lr": 9.073684637814714e-05, "val_loss": 3.1929895877838135}
{"train_loss": 0.4017094671726227, "global_step": 53845, "epoch": 605, "lr": 9.073651022627923e-05}
{"train_loss": 0.23502618074417114, "global_step": 53846, "epoch": 605, "lr": 9.073617406893477e-05}
{"train_loss": 0.2912587523460388, "global_step": 53847, "epoch": 605, "lr": 9.073583790611387e-05}
{"train_loss": 0.25459519028663635, "global_step": 53848, "epoch": 605, "lr": 9.073550173781651e-05}
{"train_loss": 0.24807141721248627, "global_step": 53849, "epoch": 605, "lr": 9.073516556404276e-05}
{"train_loss": 0.19107168912887573, "global_step": 53850, "epoch": 605, "lr": 9.073482938479266e-05}
{"train_loss": 0.24414211511611938, "global_step": 53851, "epoch": 605, "lr": 9.073449320006627e-05}
{"train_loss": 0.17954237759113312, "global_step": 53852, "epoch": 605, "lr": 9.073415700986361e-05}
{"train_loss": 0.3282620906829834, "global_step": 53853, "epoch": 605, "lr": 9.073382081418473e-05}
{"train_loss": 0.2558531165122986, "global_step": 53854, "epoch": 605, "lr": 9.07334846130297e-05}
{"train_loss": 0.18122628331184387, "global_step": 53855, "epoch": 605, "lr": 9.073314840639855e-05}
{"train_loss": 0.2998514175415039, "global_step": 53856, "epoch": 605, "lr": 9.073281219429133e-05}
{"train_loss": 0.23638032376766205, "global_step": 53857, "epoch": 605, "lr": 9.073247597670806e-05}
{"train_loss": 0.23297441005706787, "global_step": 53858, "epoch": 605, "lr": 9.073213975364882e-05}
{"train_loss": 0.2812758684158325, "global_step": 53859, "epoch": 605, "lr": 9.073180352511366e-05}
{"train_loss": 0.18845567107200623, "global_step": 53860, "epoch": 605, "lr": 9.073146729110258e-05}
{"train_loss": 0.30646294355392456, "global_step": 53861, "epoch": 605, "lr": 9.073113105161565e-05}
{"train_loss": 0.27821341156959534, "global_step": 53862, "epoch": 605, "lr": 9.073079480665291e-05}
{"train_loss": 0.21678924560546875, "global_step": 53863, "epoch": 605, "lr": 9.073045855621442e-05}
{"train_loss": 0.32746073603630066, "global_step": 53864, "epoch": 605, "lr": 9.073012230030022e-05}
{"train_loss": 0.28589263558387756, "global_step": 53865, "epoch": 605, "lr": 9.072978603891036e-05}
{"train_loss": 0.2567870616912842, "global_step": 53866, "epoch": 605, "lr": 9.072944977204486e-05}
{"train_loss": 0.26994600892066956, "global_step": 53867, "epoch": 605, "lr": 9.07291134997038e-05}
{"train_loss": 0.2075137197971344, "global_step": 53868, "epoch": 605, "lr": 9.07287772218872e-05}
{"train_loss": 0.28965163230895996, "global_step": 53869, "epoch": 605, "lr": 9.072844093859509e-05}
{"train_loss": 0.2595694661140442, "global_step": 53870, "epoch": 605, "lr": 9.072810464982756e-05}
{"train_loss": 0.25439345836639404, "global_step": 53871, "epoch": 605, "lr": 9.072776835558463e-05}
{"train_loss": 0.24981112778186798, "global_step": 53872, "epoch": 605, "lr": 9.072743205586633e-05}
{"train_loss": 0.22383920848369598, "global_step": 53873, "epoch": 605, "lr": 9.072709575067276e-05}
{"train_loss": 0.1888587325811386, "global_step": 53874, "epoch": 605, "lr": 9.072675944000391e-05}
{"train_loss": 0.26664069294929504, "global_step": 53875, "epoch": 605, "lr": 9.072642312385984e-05}
{"train_loss": 0.2178076207637787, "global_step": 53876, "epoch": 605, "lr": 9.072608680224059e-05}
{"train_loss": 0.23210600018501282, "global_step": 53877, "epoch": 605, "lr": 9.072575047514622e-05}
{"train_loss": 0.24464431405067444, "global_step": 53878, "epoch": 605, "lr": 9.072541414257678e-05}
{"train_loss": 0.3384826183319092, "global_step": 53879, "epoch": 605, "lr": 9.07250778045323e-05}
{"train_loss": 0.2988499104976654, "global_step": 53880, "epoch": 605, "lr": 9.072474146101282e-05}
{"train_loss": 0.24197211861610413, "global_step": 53881, "epoch": 605, "lr": 9.07244051120184e-05}
{"train_loss": 0.2945379614830017, "global_step": 53882, "epoch": 605, "lr": 9.072406875754909e-05}
{"train_loss": 0.27237942814826965, "global_step": 53883, "epoch": 605, "lr": 9.072373239760492e-05}
{"train_loss": 0.245807945728302, "global_step": 53884, "epoch": 605, "lr": 9.072339603218593e-05}
{"train_loss": 0.3115682899951935, "global_step": 53885, "epoch": 605, "lr": 9.072305966129217e-05}
{"train_loss": 0.22426678240299225, "global_step": 53886, "epoch": 605, "lr": 9.072272328492371e-05}
{"train_loss": 0.30370989441871643, "global_step": 53887, "epoch": 605, "lr": 9.072238690308058e-05}
{"train_loss": 0.2362688034772873, "global_step": 53888, "epoch": 605, "lr": 9.072205051576281e-05}
{"train_loss": 0.24935071170330048, "global_step": 53889, "epoch": 605, "lr": 9.072171412297046e-05}
{"train_loss": 0.21539393067359924, "global_step": 53890, "epoch": 605, "lr": 9.072137772470357e-05}
{"train_loss": 0.2655555307865143, "global_step": 53891, "epoch": 605, "lr": 9.072104132096218e-05}
{"train_loss": 0.2815311849117279, "global_step": 53892, "epoch": 605, "lr": 9.072070491174635e-05}
{"train_loss": 0.2968064546585083, "global_step": 53893, "epoch": 605, "lr": 9.072036849705612e-05}
{"train_loss": 0.1983717530965805, "global_step": 53894, "epoch": 605, "lr": 9.072003207689154e-05}
{"train_loss": 0.19626781344413757, "global_step": 53895, "epoch": 605, "lr": 9.071969565125264e-05}
{"train_loss": 0.352039098739624, "global_step": 53896, "epoch": 605, "lr": 9.071935922013947e-05}
{"train_loss": 0.26325085759162903, "global_step": 53897, "epoch": 605, "lr": 9.071902278355207e-05}
{"train_loss": 0.36845606565475464, "global_step": 53898, "epoch": 605, "lr": 9.071868634149051e-05}
{"train_loss": 0.20433834195137024, "global_step": 53899, "epoch": 605, "lr": 9.071834989395481e-05}
{"train_loss": 0.266966313123703, "global_step": 53900, "epoch": 605, "lr": 9.071801344094502e-05}
{"train_loss": 0.236977681517601, "global_step": 53901, "epoch": 605, "lr": 9.071767698246121e-05}
{"train_loss": 0.20099101960659027, "global_step": 53902, "epoch": 605, "lr": 9.071734051850337e-05}
{"train_loss": 0.18407438695430756, "global_step": 53903, "epoch": 605, "lr": 9.071700404907162e-05}
{"train_loss": 0.23152336478233337, "global_step": 53904, "epoch": 605, "lr": 9.071666757416595e-05}
{"train_loss": 0.23771822452545166, "global_step": 53905, "epoch": 605, "lr": 9.071633109378641e-05}
{"train_loss": 0.2616111636161804, "global_step": 53906, "epoch": 605, "lr": 9.071599460793306e-05}
{"train_loss": 0.30175793170928955, "global_step": 53907, "epoch": 605, "lr": 9.071565811660594e-05}
{"train_loss": 0.21497729420661926, "global_step": 53908, "epoch": 605, "lr": 9.071532161980509e-05}
{"train_loss": 0.34437650442123413, "global_step": 53909, "epoch": 605, "lr": 9.071498511753059e-05}
{"train_loss": 0.3381730318069458, "global_step": 53910, "epoch": 605, "lr": 9.071464860978243e-05}
{"train_loss": 0.24273408949375153, "global_step": 53911, "epoch": 605, "lr": 9.071431209656068e-05}
{"train_loss": 0.34723228216171265, "global_step": 53912, "epoch": 605, "lr": 9.07139755778654e-05}
{"train_loss": 0.24163126945495605, "global_step": 53913, "epoch": 605, "lr": 9.071363905369661e-05}
{"train_loss": 0.29747775197029114, "global_step": 53914, "epoch": 605, "lr": 9.071330252405437e-05}
{"train_loss": 0.2322894036769867, "global_step": 53915, "epoch": 605, "lr": 9.071296598893874e-05}
{"train_loss": 0.261370450258255, "global_step": 53916, "epoch": 605, "lr": 9.071262944834973e-05}
{"train_loss": 0.3075602352619171, "global_step": 53917, "epoch": 605, "lr": 9.071229290228741e-05}
{"train_loss": 0.25168874859809875, "global_step": 53918, "epoch": 605, "lr": 9.071195635075181e-05}
{"train_loss": 0.26521751284599304, "global_step": 53919, "epoch": 605, "lr": 9.0711619793743e-05}
{"train_loss": 0.20684170722961426, "global_step": 53920, "epoch": 605, "lr": 9.071128323126099e-05}
{"train_loss": 0.26820310950279236, "global_step": 53921, "epoch": 605, "lr": 9.071094666330587e-05}
{"train_loss": 0.298271119594574, "global_step": 53922, "epoch": 605, "lr": 9.071061008987763e-05}
{"train_loss": 0.307277113199234, "global_step": 53923, "epoch": 605, "lr": 9.071027351097636e-05}
{"train_loss": 0.24177949130535126, "global_step": 53924, "epoch": 605, "lr": 9.070993692660207e-05}
{"train_loss": 0.2057424932718277, "global_step": 53925, "epoch": 605, "lr": 9.070960033675486e-05}
{"train_loss": 0.2841740548610687, "global_step": 53926, "epoch": 605, "lr": 9.070926374143473e-05}
{"train_loss": 0.32285556197166443, "global_step": 53927, "epoch": 605, "lr": 9.070892714064173e-05}
{"train_loss": 0.28816622495651245, "global_step": 53928, "epoch": 605, "lr": 9.07085905343759e-05}
{"train_loss": 0.2042127549648285, "global_step": 53929, "epoch": 605, "lr": 9.070825392263731e-05}
{"train_loss": 0.22464922070503235, "global_step": 53930, "epoch": 605, "lr": 9.070791730542598e-05}
{"train_loss": 0.22317589819431305, "global_step": 53931, "epoch": 605, "lr": 9.070758068274197e-05}
{"train_loss": 0.1828446239233017, "global_step": 53932, "epoch": 605, "lr": 9.070724405458532e-05}
{"train_loss": 0.25900622666551826, "global_step": 53933, "epoch": 605, "lr": 9.07069074209561e-05, "val_loss": 3.1790428161621094, "train_action_mse_error": 41.02153778076172}
{"train_loss": 0.29360437393188477, "global_step": 53934, "epoch": 606, "lr": 9.070657078185431e-05}
{"train_loss": 0.19847512245178223, "global_step": 53935, "epoch": 606, "lr": 9.070623413728003e-05}
{"train_loss": 0.28972429037094116, "global_step": 53936, "epoch": 606, "lr": 9.070589748723329e-05}
{"train_loss": 0.2853524684906006, "global_step": 53937, "epoch": 606, "lr": 9.070556083171414e-05}
{"train_loss": 0.18414399027824402, "global_step": 53938, "epoch": 606, "lr": 9.07052241707226e-05}
{"train_loss": 0.28571629524230957, "global_step": 53939, "epoch": 606, "lr": 9.070488750425877e-05}
{"train_loss": 0.3200698494911194, "global_step": 53940, "epoch": 606, "lr": 9.070455083232264e-05}
{"train_loss": 0.20992812514305115, "global_step": 53941, "epoch": 606, "lr": 9.070421415491429e-05}
{"train_loss": 0.23113492131233215, "global_step": 53942, "epoch": 606, "lr": 9.070387747203374e-05}
{"train_loss": 0.21287471055984497, "global_step": 53943, "epoch": 606, "lr": 9.070354078368109e-05}
{"train_loss": 0.1933305710554123, "global_step": 53944, "epoch": 606, "lr": 9.07032040898563e-05}
{"train_loss": 0.1919146329164505, "global_step": 53945, "epoch": 606, "lr": 9.070286739055948e-05}
{"train_loss": 0.2655760645866394, "global_step": 53946, "epoch": 606, "lr": 9.070253068579067e-05}
{"train_loss": 0.15329448878765106, "global_step": 53947, "epoch": 606, "lr": 9.070219397554988e-05}
{"train_loss": 0.21639512479305267, "global_step": 53948, "epoch": 606, "lr": 9.070185725983717e-05}
{"train_loss": 0.29752472043037415, "global_step": 53949, "epoch": 606, "lr": 9.070152053865261e-05}
{"train_loss": 0.22839894890785217, "global_step": 53950, "epoch": 606, "lr": 9.070118381199622e-05}
{"train_loss": 0.29253384470939636, "global_step": 53951, "epoch": 606, "lr": 9.070084707986805e-05}
{"train_loss": 0.1817135214805603, "global_step": 53952, "epoch": 606, "lr": 9.070051034226817e-05}
{"train_loss": 0.21620652079582214, "global_step": 53953, "epoch": 606, "lr": 9.070017359919656e-05}
{"train_loss": 0.2140744924545288, "global_step": 53954, "epoch": 606, "lr": 9.069983685065333e-05}
{"train_loss": 0.3094058334827423, "global_step": 53955, "epoch": 606, "lr": 9.069950009663852e-05}
{"train_loss": 0.269081175327301, "global_step": 53956, "epoch": 606, "lr": 9.069916333715213e-05}
{"train_loss": 0.3319138288497925, "global_step": 53957, "epoch": 606, "lr": 9.069882657219424e-05}
{"train_loss": 0.24408388137817383, "global_step": 53958, "epoch": 606, "lr": 9.06984898017649e-05}
{"train_loss": 0.31065887212753296, "global_step": 53959, "epoch": 606, "lr": 9.069815302586413e-05}
{"train_loss": 0.1854664534330368, "global_step": 53960, "epoch": 606, "lr": 9.0697816244492e-05}
{"train_loss": 0.18136152625083923, "global_step": 53961, "epoch": 606, "lr": 9.069747945764855e-05}
{"train_loss": 0.22211550176143646, "global_step": 53962, "epoch": 606, "lr": 9.06971426653338e-05}
{"train_loss": 0.2890074849128723, "global_step": 53963, "epoch": 606, "lr": 9.069680586754782e-05}
{"train_loss": 0.1917819380760193, "global_step": 53964, "epoch": 606, "lr": 9.069646906429066e-05}
{"train_loss": 0.3983263373374939, "global_step": 53965, "epoch": 606, "lr": 9.069613225556236e-05}
{"train_loss": 0.359038382768631, "global_step": 53966, "epoch": 606, "lr": 9.069579544136295e-05}
{"train_loss": 0.2582075893878937, "global_step": 53967, "epoch": 606, "lr": 9.069545862169249e-05}
{"train_loss": 0.2244112342596054, "global_step": 53968, "epoch": 606, "lr": 9.069512179655102e-05}
{"train_loss": 0.239597886800766, "global_step": 53969, "epoch": 606, "lr": 9.06947849659386e-05}
{"train_loss": 0.2513912618160248, "global_step": 53970, "epoch": 606, "lr": 9.069444812985524e-05}
{"train_loss": 0.18780003488063812, "global_step": 53971, "epoch": 606, "lr": 9.069411128830103e-05}
{"train_loss": 0.32809877395629883, "global_step": 53972, "epoch": 606, "lr": 9.069377444127597e-05}
{"train_loss": 0.21637612581253052, "global_step": 53973, "epoch": 606, "lr": 9.069343758878013e-05}
{"train_loss": 0.30984535813331604, "global_step": 53974, "epoch": 606, "lr": 9.069310073081356e-05}
{"train_loss": 0.23765714466571808, "global_step": 53975, "epoch": 606, "lr": 9.069276386737631e-05}
{"train_loss": 0.26939839124679565, "global_step": 53976, "epoch": 606, "lr": 9.069242699846841e-05}
{"train_loss": 0.1643432378768921, "global_step": 53977, "epoch": 606, "lr": 9.06920901240899e-05}
{"train_loss": 0.2658293843269348, "global_step": 53978, "epoch": 606, "lr": 9.069175324424084e-05}
{"train_loss": 0.2321961522102356, "global_step": 53979, "epoch": 606, "lr": 9.069141635892126e-05}
{"train_loss": 0.21870329976081848, "global_step": 53980, "epoch": 606, "lr": 9.069107946813123e-05}
{"train_loss": 0.3028411567211151, "global_step": 53981, "epoch": 606, "lr": 9.069074257187077e-05}
{"train_loss": 0.3222121000289917, "global_step": 53982, "epoch": 606, "lr": 9.069040567013994e-05}
{"train_loss": 0.31025275588035583, "global_step": 53983, "epoch": 606, "lr": 9.069006876293877e-05}
{"train_loss": 0.2798997759819031, "global_step": 53984, "epoch": 606, "lr": 9.068973185026733e-05}
{"train_loss": 0.30736982822418213, "global_step": 53985, "epoch": 606, "lr": 9.068939493212564e-05}
{"train_loss": 0.33275794982910156, "global_step": 53986, "epoch": 606, "lr": 9.068905800851378e-05}
{"train_loss": 0.22422948479652405, "global_step": 53987, "epoch": 606, "lr": 9.068872107943176e-05}
{"train_loss": 0.3103981018066406, "global_step": 53988, "epoch": 606, "lr": 9.068838414487963e-05}
{"train_loss": 0.19109739363193512, "global_step": 53989, "epoch": 606, "lr": 9.068804720485744e-05}
{"train_loss": 0.1894306242465973, "global_step": 53990, "epoch": 606, "lr": 9.068771025936525e-05}
{"train_loss": 0.28053638339042664, "global_step": 53991, "epoch": 606, "lr": 9.068737330840308e-05}
{"train_loss": 0.22077308595180511, "global_step": 53992, "epoch": 606, "lr": 9.0687036351971e-05}
{"train_loss": 0.3475031554698944, "global_step": 53993, "epoch": 606, "lr": 9.068669939006904e-05}
{"train_loss": 0.25656378269195557, "global_step": 53994, "epoch": 606, "lr": 9.068636242269725e-05}
{"train_loss": 0.16799305379390717, "global_step": 53995, "epoch": 606, "lr": 9.068602544985568e-05}
{"train_loss": 0.20119526982307434, "global_step": 53996, "epoch": 606, "lr": 9.068568847154434e-05}
{"train_loss": 0.16751597821712494, "global_step": 53997, "epoch": 606, "lr": 9.068535148776334e-05}
{"train_loss": 0.2711314260959625, "global_step": 53998, "epoch": 606, "lr": 9.068501449851268e-05}
{"train_loss": 0.2412603497505188, "global_step": 53999, "epoch": 606, "lr": 9.068467750379241e-05}
{"train_loss": 0.22745995223522186, "global_step": 54000, "epoch": 606, "lr": 9.06843405036026e-05}
{"train_loss": 0.2854062616825104, "global_step": 54001, "epoch": 606, "lr": 9.068400349794326e-05}
{"train_loss": 0.2907625436782837, "global_step": 54002, "epoch": 606, "lr": 9.068366648681444e-05}
{"train_loss": 0.25382035970687866, "global_step": 54003, "epoch": 606, "lr": 9.068332947021622e-05}
{"train_loss": 0.3296036124229431, "global_step": 54004, "epoch": 606, "lr": 9.06829924481486e-05}
{"train_loss": 0.2131652683019638, "global_step": 54005, "epoch": 606, "lr": 9.068265542061167e-05}
{"train_loss": 0.3262578845024109, "global_step": 54006, "epoch": 606, "lr": 9.068231838760543e-05}
{"train_loss": 0.30952247977256775, "global_step": 54007, "epoch": 606, "lr": 9.068198134912998e-05}
{"train_loss": 0.24217259883880615, "global_step": 54008, "epoch": 606, "lr": 9.06816443051853e-05}
{"train_loss": 0.20724356174468994, "global_step": 54009, "epoch": 606, "lr": 9.068130725577148e-05}
{"train_loss": 0.23808355629444122, "global_step": 54010, "epoch": 606, "lr": 9.068097020088857e-05}
{"train_loss": 0.24532350897789001, "global_step": 54011, "epoch": 606, "lr": 9.068063314053659e-05}
{"train_loss": 0.28706711530685425, "global_step": 54012, "epoch": 606, "lr": 9.068029607471559e-05}
{"train_loss": 0.2772991359233856, "global_step": 54013, "epoch": 606, "lr": 9.067995900342563e-05}
{"train_loss": 0.33591428399086, "global_step": 54014, "epoch": 606, "lr": 9.067962192666672e-05}
{"train_loss": 0.266829252243042, "global_step": 54015, "epoch": 606, "lr": 9.067928484443895e-05}
{"train_loss": 0.4335585832595825, "global_step": 54016, "epoch": 606, "lr": 9.067894775674235e-05}
{"train_loss": 0.3310062885284424, "global_step": 54017, "epoch": 606, "lr": 9.067861066357694e-05}
{"train_loss": 0.3361291289329529, "global_step": 54018, "epoch": 606, "lr": 9.067827356494281e-05}
{"train_loss": 0.24141302704811096, "global_step": 54019, "epoch": 606, "lr": 9.067793646083997e-05}
{"train_loss": 0.27408844232559204, "global_step": 54020, "epoch": 606, "lr": 9.067759935126847e-05}
{"train_loss": 0.23120912909507751, "global_step": 54021, "epoch": 606, "lr": 9.067726223622839e-05}
{"train_loss": 0.257945981923114, "global_step": 54022, "epoch": 606, "lr": 9.067692511571972e-05, "val_loss": 3.146235227584839}
{"train_loss": 0.22100336849689484, "global_step": 54023, "epoch": 607, "lr": 9.067658798974255e-05}
{"train_loss": 0.2867297828197479, "global_step": 54024, "epoch": 607, "lr": 9.06762508582969e-05}
{"train_loss": 0.23390422761440277, "global_step": 54025, "epoch": 607, "lr": 9.067591372138283e-05}
{"train_loss": 0.2371579110622406, "global_step": 54026, "epoch": 607, "lr": 9.067557657900036e-05}
{"train_loss": 0.1961584836244583, "global_step": 54027, "epoch": 607, "lr": 9.067523943114958e-05}
{"train_loss": 0.27725949883461, "global_step": 54028, "epoch": 607, "lr": 9.06749022778305e-05}
{"train_loss": 0.24676722288131714, "global_step": 54029, "epoch": 607, "lr": 9.067456511904317e-05}
{"train_loss": 0.2919250726699829, "global_step": 54030, "epoch": 607, "lr": 9.067422795478766e-05}
{"train_loss": 0.23153358697891235, "global_step": 54031, "epoch": 607, "lr": 9.067389078506397e-05}
{"train_loss": 0.25517547130584717, "global_step": 54032, "epoch": 607, "lr": 9.067355360987218e-05}
{"train_loss": 0.25824180245399475, "global_step": 54033, "epoch": 607, "lr": 9.067321642921234e-05}
{"train_loss": 0.23801852762699127, "global_step": 54034, "epoch": 607, "lr": 9.067287924308447e-05}
{"train_loss": 0.17707961797714233, "global_step": 54035, "epoch": 607, "lr": 9.067254205148863e-05}
{"train_loss": 0.19476254284381866, "global_step": 54036, "epoch": 607, "lr": 9.067220485442485e-05}
{"train_loss": 0.21634958684444427, "global_step": 54037, "epoch": 607, "lr": 9.067186765189319e-05}
{"train_loss": 0.24165762960910797, "global_step": 54038, "epoch": 607, "lr": 9.067153044389371e-05}
{"train_loss": 0.22535189986228943, "global_step": 54039, "epoch": 607, "lr": 9.067119323042644e-05}
{"train_loss": 0.3390534520149231, "global_step": 54040, "epoch": 607, "lr": 9.067085601149141e-05}
{"train_loss": 0.19602181017398834, "global_step": 54041, "epoch": 607, "lr": 9.067051878708867e-05}
{"train_loss": 0.1783519834280014, "global_step": 54042, "epoch": 607, "lr": 9.067018155721829e-05}
{"train_loss": 0.24266788363456726, "global_step": 54043, "epoch": 607, "lr": 9.066984432188031e-05}
{"train_loss": 0.29316723346710205, "global_step": 54044, "epoch": 607, "lr": 9.066950708107474e-05}
{"train_loss": 0.1930219531059265, "global_step": 54045, "epoch": 607, "lr": 9.066916983480167e-05}
{"train_loss": 0.20303310453891754, "global_step": 54046, "epoch": 607, "lr": 9.066883258306111e-05}
{"train_loss": 0.17970259487628937, "global_step": 54047, "epoch": 607, "lr": 9.066849532585314e-05}
{"train_loss": 0.20136559009552002, "global_step": 54048, "epoch": 607, "lr": 9.066815806317776e-05}
{"train_loss": 0.25826218724250793, "global_step": 54049, "epoch": 607, "lr": 9.066782079503507e-05}
{"train_loss": 0.3156385123729706, "global_step": 54050, "epoch": 607, "lr": 9.066748352142507e-05}
{"train_loss": 0.2788296341896057, "global_step": 54051, "epoch": 607, "lr": 9.066714624234783e-05}
{"train_loss": 0.35553714632987976, "global_step": 54052, "epoch": 607, "lr": 9.066680895780339e-05}
{"train_loss": 0.3366563320159912, "global_step": 54053, "epoch": 607, "lr": 9.066647166779178e-05}
{"train_loss": 0.26985469460487366, "global_step": 54054, "epoch": 607, "lr": 9.066613437231306e-05}
{"train_loss": 0.3216647505760193, "global_step": 54055, "epoch": 607, "lr": 9.066579707136728e-05}
{"train_loss": 0.27095678448677063, "global_step": 54056, "epoch": 607, "lr": 9.066545976495449e-05}
{"train_loss": 0.24868355691432953, "global_step": 54057, "epoch": 607, "lr": 9.06651224530747e-05}
{"train_loss": 0.23396103084087372, "global_step": 54058, "epoch": 607, "lr": 9.066478513572801e-05}
{"train_loss": 0.25906458497047424, "global_step": 54059, "epoch": 607, "lr": 9.06644478129144e-05}
{"train_loss": 0.3141877055168152, "global_step": 54060, "epoch": 607, "lr": 9.066411048463399e-05}
{"train_loss": 0.188296377658844, "global_step": 54061, "epoch": 607, "lr": 9.066377315088676e-05}
{"train_loss": 0.24354399740695953, "global_step": 54062, "epoch": 607, "lr": 9.06634358116728e-05}
{"train_loss": 0.24950656294822693, "global_step": 54063, "epoch": 607, "lr": 9.066309846699211e-05}
{"train_loss": 0.23349574208259583, "global_step": 54064, "epoch": 607, "lr": 9.066276111684477e-05}
{"train_loss": 0.18945132195949554, "global_step": 54065, "epoch": 607, "lr": 9.066242376123082e-05}
{"train_loss": 0.21259862184524536, "global_step": 54066, "epoch": 607, "lr": 9.066208640015032e-05}
{"train_loss": 0.30938947200775146, "global_step": 54067, "epoch": 607, "lr": 9.066174903360328e-05}
{"train_loss": 0.2368747740983963, "global_step": 54068, "epoch": 607, "lr": 9.066141166158976e-05}
{"train_loss": 0.2573384940624237, "global_step": 54069, "epoch": 607, "lr": 9.066107428410983e-05}
{"train_loss": 0.2945047616958618, "global_step": 54070, "epoch": 607, "lr": 9.06607369011635e-05}
{"train_loss": 0.26984143257141113, "global_step": 54071, "epoch": 607, "lr": 9.066039951275083e-05}
{"train_loss": 0.22241178154945374, "global_step": 54072, "epoch": 607, "lr": 9.066006211887186e-05}
{"train_loss": 0.22491078078746796, "global_step": 54073, "epoch": 607, "lr": 9.065972471952665e-05}
{"train_loss": 0.28240275382995605, "global_step": 54074, "epoch": 607, "lr": 9.065938731471523e-05}
{"train_loss": 0.20256507396697998, "global_step": 54075, "epoch": 607, "lr": 9.065904990443765e-05}
{"train_loss": 0.22692185640335083, "global_step": 54076, "epoch": 607, "lr": 9.065871248869395e-05}
{"train_loss": 0.25958535075187683, "global_step": 54077, "epoch": 607, "lr": 9.06583750674842e-05}
{"train_loss": 0.26571759581565857, "global_step": 54078, "epoch": 607, "lr": 9.065803764080842e-05}
{"train_loss": 0.31032341718673706, "global_step": 54079, "epoch": 607, "lr": 9.065770020866667e-05}
{"train_loss": 0.2691956162452698, "global_step": 54080, "epoch": 607, "lr": 9.065736277105896e-05}
{"train_loss": 0.22932207584381104, "global_step": 54081, "epoch": 607, "lr": 9.065702532798539e-05}
{"train_loss": 0.20396459102630615, "global_step": 54082, "epoch": 607, "lr": 9.065668787944597e-05}
{"train_loss": 0.26569461822509766, "global_step": 54083, "epoch": 607, "lr": 9.065635042544073e-05}
{"train_loss": 0.2414143830537796, "global_step": 54084, "epoch": 607, "lr": 9.065601296596978e-05}
{"train_loss": 0.27692484855651855, "global_step": 54085, "epoch": 607, "lr": 9.06556755010331e-05}
{"train_loss": 0.25939488410949707, "global_step": 54086, "epoch": 607, "lr": 9.065533803063078e-05}
{"train_loss": 0.23356446623802185, "global_step": 54087, "epoch": 607, "lr": 9.065500055476281e-05}
{"train_loss": 0.25398755073547363, "global_step": 54088, "epoch": 607, "lr": 9.065466307342931e-05}
{"train_loss": 0.2995623052120209, "global_step": 54089, "epoch": 607, "lr": 9.065432558663026e-05}
{"train_loss": 0.2558850646018982, "global_step": 54090, "epoch": 607, "lr": 9.065398809436574e-05}
{"train_loss": 0.19727903604507446, "global_step": 54091, "epoch": 607, "lr": 9.065365059663578e-05}
{"train_loss": 0.21727950870990753, "global_step": 54092, "epoch": 607, "lr": 9.065331309344046e-05}
{"train_loss": 0.268025666475296, "global_step": 54093, "epoch": 607, "lr": 9.065297558477976e-05}
{"train_loss": 0.21684131026268005, "global_step": 54094, "epoch": 607, "lr": 9.065263807065378e-05}
{"train_loss": 0.3788372576236725, "global_step": 54095, "epoch": 607, "lr": 9.065230055106255e-05}
{"train_loss": 0.3673546314239502, "global_step": 54096, "epoch": 607, "lr": 9.065196302600612e-05}
{"train_loss": 0.21753880381584167, "global_step": 54097, "epoch": 607, "lr": 9.065162549548452e-05}
{"train_loss": 0.22883805632591248, "global_step": 54098, "epoch": 607, "lr": 9.06512879594978e-05}
{"train_loss": 0.21089497208595276, "global_step": 54099, "epoch": 607, "lr": 9.065095041804602e-05}
{"train_loss": 0.293417364358902, "global_step": 54100, "epoch": 607, "lr": 9.06506128711292e-05}
{"train_loss": 0.2622479498386383, "global_step": 54101, "epoch": 607, "lr": 9.065027531874741e-05}
{"train_loss": 0.22843462228775024, "global_step": 54102, "epoch": 607, "lr": 9.064993776090068e-05}
{"train_loss": 0.3234429657459259, "global_step": 54103, "epoch": 607, "lr": 9.064960019758908e-05}
{"train_loss": 0.2715173661708832, "global_step": 54104, "epoch": 607, "lr": 9.064926262881261e-05}
{"train_loss": 0.2247825264930725, "global_step": 54105, "epoch": 607, "lr": 9.064892505457137e-05}
{"train_loss": 0.3164386451244354, "global_step": 54106, "epoch": 607, "lr": 9.064858747486535e-05}
{"train_loss": 0.3204532861709595, "global_step": 54107, "epoch": 607, "lr": 9.064824988969465e-05}
{"train_loss": 0.20880964398384094, "global_step": 54108, "epoch": 607, "lr": 9.064791229905927e-05}
{"train_loss": 0.20427429676055908, "global_step": 54109, "epoch": 607, "lr": 9.064757470295928e-05}
{"train_loss": 0.3031167984008789, "global_step": 54110, "epoch": 607, "lr": 9.064723710139471e-05}
{"train_loss": 0.25290343182140523, "global_step": 54111, "epoch": 607, "lr": 9.064689949436563e-05, "val_loss": 3.21317720413208}
{"train_loss": 0.24358540773391724, "global_step": 54112, "epoch": 608, "lr": 9.064656188187206e-05}
{"train_loss": 0.21044406294822693, "global_step": 54113, "epoch": 608, "lr": 9.064622426391405e-05}
{"train_loss": 0.23297736048698425, "global_step": 54114, "epoch": 608, "lr": 9.064588664049167e-05}
{"train_loss": 0.30563002824783325, "global_step": 54115, "epoch": 608, "lr": 9.064554901160492e-05}
{"train_loss": 0.2043294757604599, "global_step": 54116, "epoch": 608, "lr": 9.064521137725389e-05}
{"train_loss": 0.25331753492355347, "global_step": 54117, "epoch": 608, "lr": 9.06448737374386e-05}
{"train_loss": 0.21247607469558716, "global_step": 54118, "epoch": 608, "lr": 9.06445360921591e-05}
{"train_loss": 0.24372410774230957, "global_step": 54119, "epoch": 608, "lr": 9.064419844141545e-05}
{"train_loss": 0.333871066570282, "global_step": 54120, "epoch": 608, "lr": 9.064386078520767e-05}
{"train_loss": 0.2448539286851883, "global_step": 54121, "epoch": 608, "lr": 9.064352312353583e-05}
{"train_loss": 0.36726677417755127, "global_step": 54122, "epoch": 608, "lr": 9.064318545639994e-05}
{"train_loss": 0.22322985529899597, "global_step": 54123, "epoch": 608, "lr": 9.06428477838001e-05}
{"train_loss": 0.3018968999385834, "global_step": 54124, "epoch": 608, "lr": 9.06425101057363e-05}
{"train_loss": 0.18456381559371948, "global_step": 54125, "epoch": 608, "lr": 9.064217242220862e-05}
{"train_loss": 0.22398412227630615, "global_step": 54126, "epoch": 608, "lr": 9.06418347332171e-05}
{"train_loss": 0.2537000775337219, "global_step": 54127, "epoch": 608, "lr": 9.064149703876177e-05}
{"train_loss": 0.2739434838294983, "global_step": 54128, "epoch": 608, "lr": 9.06411593388427e-05}
{"train_loss": 0.20845623314380646, "global_step": 54129, "epoch": 608, "lr": 9.06408216334599e-05}
{"train_loss": 0.3057975172996521, "global_step": 54130, "epoch": 608, "lr": 9.064048392261346e-05}
{"train_loss": 0.19252879917621613, "global_step": 54131, "epoch": 608, "lr": 9.064014620630338e-05}
{"train_loss": 0.30866000056266785, "global_step": 54132, "epoch": 608, "lr": 9.063980848452976e-05}
{"train_loss": 0.20029769837856293, "global_step": 54133, "epoch": 608, "lr": 9.063947075729257e-05}
{"train_loss": 0.320446640253067, "global_step": 54134, "epoch": 608, "lr": 9.063913302459192e-05}
{"train_loss": 0.2844713628292084, "global_step": 54135, "epoch": 608, "lr": 9.063879528642784e-05}
{"train_loss": 0.3191385865211487, "global_step": 54136, "epoch": 608, "lr": 9.063845754280037e-05}
{"train_loss": 0.29276779294013977, "global_step": 54137, "epoch": 608, "lr": 9.063811979370955e-05}
{"train_loss": 0.2579859793186188, "global_step": 54138, "epoch": 608, "lr": 9.063778203915543e-05}
{"train_loss": 0.31324419379234314, "global_step": 54139, "epoch": 608, "lr": 9.063744427913805e-05}
{"train_loss": 0.15888075530529022, "global_step": 54140, "epoch": 608, "lr": 9.063710651365747e-05}
{"train_loss": 0.20588921010494232, "global_step": 54141, "epoch": 608, "lr": 9.063676874271372e-05}
{"train_loss": 0.23329538106918335, "global_step": 54142, "epoch": 608, "lr": 9.063643096630685e-05}
{"train_loss": 0.18751947581768036, "global_step": 54143, "epoch": 608, "lr": 9.063609318443691e-05}
{"train_loss": 0.35656365752220154, "global_step": 54144, "epoch": 608, "lr": 9.063575539710395e-05}
{"train_loss": 0.2979161739349365, "global_step": 54145, "epoch": 608, "lr": 9.0635417604308e-05}
{"train_loss": 0.36045336723327637, "global_step": 54146, "epoch": 608, "lr": 9.063507980604911e-05}
{"train_loss": 0.3122706711292267, "global_step": 54147, "epoch": 608, "lr": 9.063474200232734e-05}
{"train_loss": 0.23775281012058258, "global_step": 54148, "epoch": 608, "lr": 9.063440419314271e-05}
{"train_loss": 0.28633058071136475, "global_step": 54149, "epoch": 608, "lr": 9.06340663784953e-05}
{"train_loss": 0.2858767509460449, "global_step": 54150, "epoch": 608, "lr": 9.063372855838513e-05}
{"train_loss": 0.19209107756614685, "global_step": 54151, "epoch": 608, "lr": 9.063339073281223e-05}
{"train_loss": 0.24330958724021912, "global_step": 54152, "epoch": 608, "lr": 9.063305290177668e-05}
{"train_loss": 0.2293284386396408, "global_step": 54153, "epoch": 608, "lr": 9.063271506527851e-05}
{"train_loss": 0.3041309118270874, "global_step": 54154, "epoch": 608, "lr": 9.063237722331776e-05}
{"train_loss": 0.28910741209983826, "global_step": 54155, "epoch": 608, "lr": 9.063203937589448e-05}
{"train_loss": 0.2583150565624237, "global_step": 54156, "epoch": 608, "lr": 9.063170152300874e-05}
{"train_loss": 0.2515694499015808, "global_step": 54157, "epoch": 608, "lr": 9.063136366466054e-05}
{"train_loss": 0.2600329518318176, "global_step": 54158, "epoch": 608, "lr": 9.063102580084996e-05}
{"train_loss": 0.23339638113975525, "global_step": 54159, "epoch": 608, "lr": 9.063068793157702e-05}
{"train_loss": 0.2428654134273529, "global_step": 54160, "epoch": 608, "lr": 9.063035005684181e-05}
{"train_loss": 0.2862679362297058, "global_step": 54161, "epoch": 608, "lr": 9.063001217664433e-05}
{"train_loss": 0.3234986662864685, "global_step": 54162, "epoch": 608, "lr": 9.062967429098462e-05}
{"train_loss": 0.2640756368637085, "global_step": 54163, "epoch": 608, "lr": 9.062933639986278e-05}
{"train_loss": 0.3708442151546478, "global_step": 54164, "epoch": 608, "lr": 9.06289985032788e-05}
{"train_loss": 0.2607860863208771, "global_step": 54165, "epoch": 608, "lr": 9.062866060123273e-05}
{"train_loss": 0.19616524875164032, "global_step": 54166, "epoch": 608, "lr": 9.062832269372466e-05}
{"train_loss": 0.26861244440078735, "global_step": 54167, "epoch": 608, "lr": 9.062798478075461e-05}
{"train_loss": 0.274733304977417, "global_step": 54168, "epoch": 608, "lr": 9.062764686232259e-05}
{"train_loss": 0.368496298789978, "global_step": 54169, "epoch": 608, "lr": 9.06273089384287e-05}
{"train_loss": 0.3283970355987549, "global_step": 54170, "epoch": 608, "lr": 9.062697100907296e-05}
{"train_loss": 0.3380194306373596, "global_step": 54171, "epoch": 608, "lr": 9.062663307425544e-05}
{"train_loss": 0.28502926230430603, "global_step": 54172, "epoch": 608, "lr": 9.062629513397612e-05}
{"train_loss": 0.3423546254634857, "global_step": 54173, "epoch": 608, "lr": 9.062595718823512e-05}
{"train_loss": 0.24215497076511383, "global_step": 54174, "epoch": 608, "lr": 9.062561923703245e-05}
{"train_loss": 0.24661390483379364, "global_step": 54175, "epoch": 608, "lr": 9.062528128036816e-05}
{"train_loss": 0.200234055519104, "global_step": 54176, "epoch": 608, "lr": 9.062494331824229e-05}
{"train_loss": 0.28891506791114807, "global_step": 54177, "epoch": 608, "lr": 9.062460535065492e-05}
{"train_loss": 0.3420153260231018, "global_step": 54178, "epoch": 608, "lr": 9.062426737760604e-05}
{"train_loss": 0.34103408455848694, "global_step": 54179, "epoch": 608, "lr": 9.062392939909572e-05}
{"train_loss": 0.23386183381080627, "global_step": 54180, "epoch": 608, "lr": 9.062359141512402e-05}
{"train_loss": 0.25409504771232605, "global_step": 54181, "epoch": 608, "lr": 9.062325342569096e-05}
{"train_loss": 0.2742089629173279, "global_step": 54182, "epoch": 608, "lr": 9.062291543079662e-05}
{"train_loss": 0.2924385368824005, "global_step": 54183, "epoch": 608, "lr": 9.0622577430441e-05}
{"train_loss": 0.2755736708641052, "global_step": 54184, "epoch": 608, "lr": 9.062223942462418e-05}
{"train_loss": 0.2516537308692932, "global_step": 54185, "epoch": 608, "lr": 9.06219014133462e-05}
{"train_loss": 0.2718089818954468, "global_step": 54186, "epoch": 608, "lr": 9.06215633966071e-05}
{"train_loss": 0.24702678620815277, "global_step": 54187, "epoch": 608, "lr": 9.062122537440692e-05}
{"train_loss": 0.26318472623825073, "global_step": 54188, "epoch": 608, "lr": 9.06208873467457e-05}
{"train_loss": 0.26291194558143616, "global_step": 54189, "epoch": 608, "lr": 9.062054931362351e-05}
{"train_loss": 0.2133367955684662, "global_step": 54190, "epoch": 608, "lr": 9.062021127504039e-05}
{"train_loss": 0.23638884723186493, "global_step": 54191, "epoch": 608, "lr": 9.061987323099636e-05}
{"train_loss": 0.36097151041030884, "global_step": 54192, "epoch": 608, "lr": 9.061953518149151e-05}
{"train_loss": 0.2078576385974884, "global_step": 54193, "epoch": 608, "lr": 9.061919712652584e-05}
{"train_loss": 0.2817735970020294, "global_step": 54194, "epoch": 608, "lr": 9.061885906609942e-05}
{"train_loss": 0.2357582449913025, "global_step": 54195, "epoch": 608, "lr": 9.061852100021228e-05}
{"train_loss": 0.31962019205093384, "global_step": 54196, "epoch": 608, "lr": 9.061818292886449e-05}
{"train_loss": 0.2501932680606842, "global_step": 54197, "epoch": 608, "lr": 9.061784485205606e-05}
{"train_loss": 0.2550298273563385, "global_step": 54198, "epoch": 608, "lr": 9.061750676978707e-05}
{"train_loss": 0.38035398721694946, "global_step": 54199, "epoch": 608, "lr": 9.061716868205754e-05}
{"train_loss": 0.2689359548051706, "global_step": 54200, "epoch": 608, "lr": 9.061683058886754e-05, "val_loss": 3.138427495956421}
{"train_loss": 0.21035079658031464, "global_step": 54201, "epoch": 609, "lr": 9.061649249021709e-05}
{"train_loss": 0.30549630522727966, "global_step": 54202, "epoch": 609, "lr": 9.061615438610626e-05}
{"train_loss": 0.2566473186016083, "global_step": 54203, "epoch": 609, "lr": 9.061581627653508e-05}
{"train_loss": 0.23459431529045105, "global_step": 54204, "epoch": 609, "lr": 9.06154781615036e-05}
{"train_loss": 0.16869528591632843, "global_step": 54205, "epoch": 609, "lr": 9.061514004101185e-05}
{"train_loss": 0.28176090121269226, "global_step": 54206, "epoch": 609, "lr": 9.06148019150599e-05}
{"train_loss": 0.31166255474090576, "global_step": 54207, "epoch": 609, "lr": 9.06144637836478e-05}
{"train_loss": 0.22976166009902954, "global_step": 54208, "epoch": 609, "lr": 9.061412564677555e-05}
{"train_loss": 0.1483345478773117, "global_step": 54209, "epoch": 609, "lr": 9.061378750444324e-05}
{"train_loss": 0.16895262897014618, "global_step": 54210, "epoch": 609, "lr": 9.061344935665091e-05}
{"train_loss": 0.27011802792549133, "global_step": 54211, "epoch": 609, "lr": 9.061311120339859e-05}
{"train_loss": 0.33538147807121277, "global_step": 54212, "epoch": 609, "lr": 9.061277304468632e-05}
{"train_loss": 0.25594592094421387, "global_step": 54213, "epoch": 609, "lr": 9.061243488051417e-05}
{"train_loss": 0.26645615696907043, "global_step": 54214, "epoch": 609, "lr": 9.061209671088217e-05}
{"train_loss": 0.24616514146327972, "global_step": 54215, "epoch": 609, "lr": 9.061175853579038e-05}
{"train_loss": 0.23090548813343048, "global_step": 54216, "epoch": 609, "lr": 9.061142035523882e-05}
{"train_loss": 0.25676628947257996, "global_step": 54217, "epoch": 609, "lr": 9.061108216922756e-05}
{"train_loss": 0.25990206003189087, "global_step": 54218, "epoch": 609, "lr": 9.061074397775661e-05}
{"train_loss": 0.24587373435497284, "global_step": 54219, "epoch": 609, "lr": 9.061040578082607e-05}
{"train_loss": 0.24252201616764069, "global_step": 54220, "epoch": 609, "lr": 9.061006757843594e-05}
{"train_loss": 0.250213623046875, "global_step": 54221, "epoch": 609, "lr": 9.060972937058629e-05}
{"train_loss": 0.24616208672523499, "global_step": 54222, "epoch": 609, "lr": 9.060939115727715e-05}
{"train_loss": 0.1878836750984192, "global_step": 54223, "epoch": 609, "lr": 9.060905293850858e-05}
{"train_loss": 0.32416704297065735, "global_step": 54224, "epoch": 609, "lr": 9.060871471428062e-05}
{"train_loss": 0.25788065791130066, "global_step": 54225, "epoch": 609, "lr": 9.060837648459331e-05}
{"train_loss": 0.1802264153957367, "global_step": 54226, "epoch": 609, "lr": 9.060803824944668e-05}
{"train_loss": 0.2434675395488739, "global_step": 54227, "epoch": 609, "lr": 9.060770000884082e-05}
{"train_loss": 0.2089352309703827, "global_step": 54228, "epoch": 609, "lr": 9.060736176277574e-05}
{"train_loss": 0.17840488255023956, "global_step": 54229, "epoch": 609, "lr": 9.06070235112515e-05}
{"train_loss": 0.2884407639503479, "global_step": 54230, "epoch": 609, "lr": 9.060668525426812e-05}
{"train_loss": 0.28310272097587585, "global_step": 54231, "epoch": 609, "lr": 9.060634699182569e-05}
{"train_loss": 0.3100021481513977, "global_step": 54232, "epoch": 609, "lr": 9.060600872392422e-05}
{"train_loss": 0.2756494879722595, "global_step": 54233, "epoch": 609, "lr": 9.060567045056376e-05}
{"train_loss": 0.3927918076515198, "global_step": 54234, "epoch": 609, "lr": 9.060533217174438e-05}
{"train_loss": 0.18859314918518066, "global_step": 54235, "epoch": 609, "lr": 9.06049938874661e-05}
{"train_loss": 0.2642078399658203, "global_step": 54236, "epoch": 609, "lr": 9.060465559772897e-05}
{"train_loss": 0.2938344478607178, "global_step": 54237, "epoch": 609, "lr": 9.060431730253305e-05}
{"train_loss": 0.292127788066864, "global_step": 54238, "epoch": 609, "lr": 9.060397900187836e-05}
{"train_loss": 0.22663529217243195, "global_step": 54239, "epoch": 609, "lr": 9.060364069576497e-05}
{"train_loss": 0.18129608035087585, "global_step": 54240, "epoch": 609, "lr": 9.060330238419292e-05}
{"train_loss": 0.21953235566616058, "global_step": 54241, "epoch": 609, "lr": 9.060296406716224e-05}
{"train_loss": 0.21757544577121735, "global_step": 54242, "epoch": 609, "lr": 9.060262574467298e-05}
{"train_loss": 0.12772293388843536, "global_step": 54243, "epoch": 609, "lr": 9.060228741672521e-05}
{"train_loss": 0.2560601234436035, "global_step": 54244, "epoch": 609, "lr": 9.060194908331894e-05}
{"train_loss": 0.20784328877925873, "global_step": 54245, "epoch": 609, "lr": 9.060161074445425e-05}
{"train_loss": 0.22684265673160553, "global_step": 54246, "epoch": 609, "lr": 9.060127240013116e-05}
{"train_loss": 0.2317349761724472, "global_step": 54247, "epoch": 609, "lr": 9.060093405034973e-05}
{"train_loss": 0.24521832168102264, "global_step": 54248, "epoch": 609, "lr": 9.060059569510999e-05}
{"train_loss": 0.3184433877468109, "global_step": 54249, "epoch": 609, "lr": 9.060025733441199e-05}
{"train_loss": 0.2331543117761612, "global_step": 54250, "epoch": 609, "lr": 9.05999189682558e-05}
{"train_loss": 0.22653789818286896, "global_step": 54251, "epoch": 609, "lr": 9.059958059664142e-05}
{"train_loss": 0.2402038276195526, "global_step": 54252, "epoch": 609, "lr": 9.059924221956895e-05}
{"train_loss": 0.22054731845855713, "global_step": 54253, "epoch": 609, "lr": 9.059890383703838e-05}
{"train_loss": 0.3184412717819214, "global_step": 54254, "epoch": 609, "lr": 9.059856544904978e-05}
{"train_loss": 0.22376342117786407, "global_step": 54255, "epoch": 609, "lr": 9.059822705560323e-05}
{"train_loss": 0.17438726127147675, "global_step": 54256, "epoch": 609, "lr": 9.059788865669871e-05}
{"train_loss": 0.2552359402179718, "global_step": 54257, "epoch": 609, "lr": 9.059755025233632e-05}
{"train_loss": 0.28743359446525574, "global_step": 54258, "epoch": 609, "lr": 9.059721184251608e-05}
{"train_loss": 0.1860370934009552, "global_step": 54259, "epoch": 609, "lr": 9.059687342723804e-05}
{"train_loss": 0.21710608899593353, "global_step": 54260, "epoch": 609, "lr": 9.059653500650225e-05}
{"train_loss": 0.2665974795818329, "global_step": 54261, "epoch": 609, "lr": 9.059619658030874e-05}
{"train_loss": 0.28961533308029175, "global_step": 54262, "epoch": 609, "lr": 9.059585814865756e-05}
{"train_loss": 0.21211960911750793, "global_step": 54263, "epoch": 609, "lr": 9.059551971154878e-05}
{"train_loss": 0.3026113212108612, "global_step": 54264, "epoch": 609, "lr": 9.05951812689824e-05}
{"train_loss": 0.2766379117965698, "global_step": 54265, "epoch": 609, "lr": 9.059484282095853e-05}
{"train_loss": 0.35742953419685364, "global_step": 54266, "epoch": 609, "lr": 9.059450436747715e-05}
{"train_loss": 0.2584938704967499, "global_step": 54267, "epoch": 609, "lr": 9.059416590853835e-05}
{"train_loss": 0.25924012064933777, "global_step": 54268, "epoch": 609, "lr": 9.059382744414215e-05}
{"train_loss": 0.3367902338504791, "global_step": 54269, "epoch": 609, "lr": 9.059348897428861e-05}
{"train_loss": 0.24600628018379211, "global_step": 54270, "epoch": 609, "lr": 9.059315049897776e-05}
{"train_loss": 0.26596009731292725, "global_step": 54271, "epoch": 609, "lr": 9.059281201820966e-05}
{"train_loss": 0.2221485823392868, "global_step": 54272, "epoch": 609, "lr": 9.059247353198435e-05}
{"train_loss": 0.2931857109069824, "global_step": 54273, "epoch": 609, "lr": 9.059213504030189e-05}
{"train_loss": 0.232107013463974, "global_step": 54274, "epoch": 609, "lr": 9.05917965431623e-05}
{"train_loss": 0.23656225204467773, "global_step": 54275, "epoch": 609, "lr": 9.059145804056564e-05}
{"train_loss": 0.2482423335313797, "global_step": 54276, "epoch": 609, "lr": 9.059111953251195e-05}
{"train_loss": 0.28518447279930115, "global_step": 54277, "epoch": 609, "lr": 9.059078101900129e-05}
{"train_loss": 0.20771515369415283, "global_step": 54278, "epoch": 609, "lr": 9.059044250003368e-05}
{"train_loss": 0.1972995400428772, "global_step": 54279, "epoch": 609, "lr": 9.059010397560918e-05}
{"train_loss": 0.3388403356075287, "global_step": 54280, "epoch": 609, "lr": 9.058976544572785e-05}
{"train_loss": 0.153367280960083, "global_step": 54281, "epoch": 609, "lr": 9.058942691038971e-05}
{"train_loss": 0.23306044936180115, "global_step": 54282, "epoch": 609, "lr": 9.058908836959483e-05}
{"train_loss": 0.2390187382698059, "global_step": 54283, "epoch": 609, "lr": 9.058874982334322e-05}
{"train_loss": 0.3064660429954529, "global_step": 54284, "epoch": 609, "lr": 9.058841127163496e-05}
{"train_loss": 0.3617435097694397, "global_step": 54285, "epoch": 609, "lr": 9.058807271447007e-05}
{"train_loss": 0.24841924011707306, "global_step": 54286, "epoch": 609, "lr": 9.058773415184863e-05}
{"train_loss": 0.31314006447792053, "global_step": 54287, "epoch": 609, "lr": 9.058739558377065e-05}
{"train_loss": 0.27414417266845703, "global_step": 54288, "epoch": 609, "lr": 9.05870570102362e-05}
{"train_loss": 0.25233399482925284, "global_step": 54289, "epoch": 609, "lr": 9.05867184312453e-05, "val_loss": 3.3011531829833984}
{"train_loss": 0.2727014720439911, "global_step": 54290, "epoch": 610, "lr": 9.058637984679801e-05}
{"train_loss": 0.23923486471176147, "global_step": 54291, "epoch": 610, "lr": 9.058604125689439e-05}
{"train_loss": 0.3135634958744049, "global_step": 54292, "epoch": 610, "lr": 9.058570266153445e-05}
{"train_loss": 0.26909103989601135, "global_step": 54293, "epoch": 610, "lr": 9.058536406071829e-05}
{"train_loss": 0.15517030656337738, "global_step": 54294, "epoch": 610, "lr": 9.05850254544459e-05}
{"train_loss": 0.2848603129386902, "global_step": 54295, "epoch": 610, "lr": 9.058468684271734e-05}
{"train_loss": 0.2485564649105072, "global_step": 54296, "epoch": 610, "lr": 9.058434822553268e-05}
{"train_loss": 0.19887371361255646, "global_step": 54297, "epoch": 610, "lr": 9.058400960289195e-05}
{"train_loss": 0.3384273648262024, "global_step": 54298, "epoch": 610, "lr": 9.058367097479518e-05}
{"train_loss": 0.3629123866558075, "global_step": 54299, "epoch": 610, "lr": 9.058333234124244e-05}
{"train_loss": 0.2165738344192505, "global_step": 54300, "epoch": 610, "lr": 9.058299370223377e-05}
{"train_loss": 0.19262732565402985, "global_step": 54301, "epoch": 610, "lr": 9.058265505776921e-05}
{"train_loss": 0.20642544329166412, "global_step": 54302, "epoch": 610, "lr": 9.05823164078488e-05}
{"train_loss": 0.2970321476459503, "global_step": 54303, "epoch": 610, "lr": 9.058197775247259e-05}
{"train_loss": 0.19729311764240265, "global_step": 54304, "epoch": 610, "lr": 9.058163909164063e-05}
{"train_loss": 0.3036026954650879, "global_step": 54305, "epoch": 610, "lr": 9.058130042535298e-05}
{"train_loss": 0.21390430629253387, "global_step": 54306, "epoch": 610, "lr": 9.058096175360964e-05}
{"train_loss": 0.21628105640411377, "global_step": 54307, "epoch": 610, "lr": 9.058062307641071e-05}
{"train_loss": 0.27972009778022766, "global_step": 54308, "epoch": 610, "lr": 9.058028439375619e-05}
{"train_loss": 0.21103404462337494, "global_step": 54309, "epoch": 610, "lr": 9.057994570564615e-05}
{"train_loss": 0.2768632769584656, "global_step": 54310, "epoch": 610, "lr": 9.057960701208063e-05}
{"train_loss": 0.30490896105766296, "global_step": 54311, "epoch": 610, "lr": 9.057926831305969e-05}
{"train_loss": 0.3099654018878937, "global_step": 54312, "epoch": 610, "lr": 9.057892960858335e-05}
{"train_loss": 0.22898365557193756, "global_step": 54313, "epoch": 610, "lr": 9.057859089865168e-05}
{"train_loss": 0.35004478693008423, "global_step": 54314, "epoch": 610, "lr": 9.05782521832647e-05}
{"train_loss": 0.20946809649467468, "global_step": 54315, "epoch": 610, "lr": 9.057791346242247e-05}
{"train_loss": 0.20352230966091156, "global_step": 54316, "epoch": 610, "lr": 9.057757473612504e-05}
{"train_loss": 0.39438891410827637, "global_step": 54317, "epoch": 610, "lr": 9.057723600437244e-05}
{"train_loss": 0.299204558134079, "global_step": 54318, "epoch": 610, "lr": 9.057689726716474e-05}
{"train_loss": 0.22402621805667877, "global_step": 54319, "epoch": 610, "lr": 9.057655852450195e-05}
{"train_loss": 0.2089984118938446, "global_step": 54320, "epoch": 610, "lr": 9.057621977638415e-05}
{"train_loss": 0.26160284876823425, "global_step": 54321, "epoch": 610, "lr": 9.057588102281137e-05}
{"train_loss": 0.28118500113487244, "global_step": 54322, "epoch": 610, "lr": 9.057554226378366e-05}
{"train_loss": 0.26669541001319885, "global_step": 54323, "epoch": 610, "lr": 9.057520349930106e-05}
{"train_loss": 0.2897205352783203, "global_step": 54324, "epoch": 610, "lr": 9.057486472936361e-05}
{"train_loss": 0.1929214894771576, "global_step": 54325, "epoch": 610, "lr": 9.057452595397138e-05}
{"train_loss": 0.27324655652046204, "global_step": 54326, "epoch": 610, "lr": 9.057418717312439e-05}
{"train_loss": 0.2979046106338501, "global_step": 54327, "epoch": 610, "lr": 9.05738483868227e-05}
{"train_loss": 0.25749632716178894, "global_step": 54328, "epoch": 610, "lr": 9.057350959506634e-05}
{"train_loss": 0.23832379281520844, "global_step": 54329, "epoch": 610, "lr": 9.05731707978554e-05}
{"train_loss": 0.21015313267707825, "global_step": 54330, "epoch": 610, "lr": 9.057283199518984e-05}
{"train_loss": 0.21183453500270844, "global_step": 54331, "epoch": 610, "lr": 9.05724931870698e-05}
{"train_loss": 0.27591413259506226, "global_step": 54332, "epoch": 610, "lr": 9.057215437349526e-05}
{"train_loss": 0.2390085756778717, "global_step": 54333, "epoch": 610, "lr": 9.057181555446628e-05}
{"train_loss": 0.21089677512645721, "global_step": 54334, "epoch": 610, "lr": 9.057147672998293e-05}
{"train_loss": 0.20470277965068817, "global_step": 54335, "epoch": 610, "lr": 9.057113790004525e-05}
{"train_loss": 0.30509239435195923, "global_step": 54336, "epoch": 610, "lr": 9.057079906465325e-05}
{"train_loss": 0.22184793651103973, "global_step": 54337, "epoch": 610, "lr": 9.057046022380702e-05}
{"train_loss": 0.26768073439598083, "global_step": 54338, "epoch": 610, "lr": 9.057012137750658e-05}
{"train_loss": 0.21804295480251312, "global_step": 54339, "epoch": 610, "lr": 9.056978252575198e-05}
{"train_loss": 0.21322381496429443, "global_step": 54340, "epoch": 610, "lr": 9.056944366854328e-05}
{"train_loss": 0.1784965991973877, "global_step": 54341, "epoch": 610, "lr": 9.056910480588049e-05}
{"train_loss": 0.26249364018440247, "global_step": 54342, "epoch": 610, "lr": 9.056876593776368e-05}
{"train_loss": 0.18063338100910187, "global_step": 54343, "epoch": 610, "lr": 9.05684270641929e-05}
{"train_loss": 0.3053320646286011, "global_step": 54344, "epoch": 610, "lr": 9.05680881851682e-05}
{"train_loss": 0.20058374106884003, "global_step": 54345, "epoch": 610, "lr": 9.056774930068961e-05}
{"train_loss": 0.2120974063873291, "global_step": 54346, "epoch": 610, "lr": 9.056741041075718e-05}
{"train_loss": 0.21580365300178528, "global_step": 54347, "epoch": 610, "lr": 9.056707151537095e-05}
{"train_loss": 0.2198924720287323, "global_step": 54348, "epoch": 610, "lr": 9.056673261453097e-05}
{"train_loss": 0.1933063417673111, "global_step": 54349, "epoch": 610, "lr": 9.056639370823729e-05}
{"train_loss": 0.3478308618068695, "global_step": 54350, "epoch": 610, "lr": 9.056605479648996e-05}
{"train_loss": 0.16726043820381165, "global_step": 54351, "epoch": 610, "lr": 9.056571587928902e-05}
{"train_loss": 0.34265759587287903, "global_step": 54352, "epoch": 610, "lr": 9.056537695663451e-05}
{"train_loss": 0.17067091166973114, "global_step": 54353, "epoch": 610, "lr": 9.056503802852647e-05}
{"train_loss": 0.2476399689912796, "global_step": 54354, "epoch": 610, "lr": 9.056469909496496e-05}
{"train_loss": 0.2295302301645279, "global_step": 54355, "epoch": 610, "lr": 9.056436015595002e-05}
{"train_loss": 0.2415933459997177, "global_step": 54356, "epoch": 610, "lr": 9.05640212114817e-05}
{"train_loss": 0.2936954200267792, "global_step": 54357, "epoch": 610, "lr": 9.056368226156004e-05}
{"train_loss": 0.28760600090026855, "global_step": 54358, "epoch": 610, "lr": 9.05633433061851e-05}
{"train_loss": 0.273145467042923, "global_step": 54359, "epoch": 610, "lr": 9.05630043453569e-05}
{"train_loss": 0.29837486147880554, "global_step": 54360, "epoch": 610, "lr": 9.056266537907549e-05}
{"train_loss": 0.23124319314956665, "global_step": 54361, "epoch": 610, "lr": 9.056232640734095e-05}
{"train_loss": 0.21940883994102478, "global_step": 54362, "epoch": 610, "lr": 9.056198743015327e-05}
{"train_loss": 0.26246505975723267, "global_step": 54363, "epoch": 610, "lr": 9.056164844751254e-05}
{"train_loss": 0.3478004038333893, "global_step": 54364, "epoch": 610, "lr": 9.05613094594188e-05}
{"train_loss": 0.324245423078537, "global_step": 54365, "epoch": 610, "lr": 9.056097046587207e-05}
{"train_loss": 0.3474405109882355, "global_step": 54366, "epoch": 610, "lr": 9.056063146687242e-05}
{"train_loss": 0.2314295470714569, "global_step": 54367, "epoch": 610, "lr": 9.056029246241988e-05}
{"train_loss": 0.30895522236824036, "global_step": 54368, "epoch": 610, "lr": 9.055995345251452e-05}
{"train_loss": 0.21973565220832825, "global_step": 54369, "epoch": 610, "lr": 9.055961443715634e-05}
{"train_loss": 0.22565957903862, "global_step": 54370, "epoch": 610, "lr": 9.055927541634544e-05}
{"train_loss": 0.2819873094558716, "global_step": 54371, "epoch": 610, "lr": 9.055893639008183e-05}
{"train_loss": 0.2811209559440613, "global_step": 54372, "epoch": 610, "lr": 9.055859735836555e-05}
{"train_loss": 0.2507103383541107, "global_step": 54373, "epoch": 610, "lr": 9.055825832119669e-05}
{"train_loss": 0.2692156136035919, "global_step": 54374, "epoch": 610, "lr": 9.055791927857525e-05}
{"train_loss": 0.25539836287498474, "global_step": 54375, "epoch": 610, "lr": 9.055758023050131e-05}
{"train_loss": 0.3131861984729767, "global_step": 54376, "epoch": 610, "lr": 9.055724117697488e-05}
{"train_loss": 0.2241772711277008, "global_step": 54377, "epoch": 610, "lr": 9.055690211799602e-05}
{"train_loss": 0.2549806771318564, "global_step": 54378, "epoch": 610, "lr": 9.05565630535648e-05, "val_loss": 3.287916421890259, "train_action_mse_error": 38.05010223388672}
{"train_loss": 0.2274945080280304, "global_step": 54379, "epoch": 611, "lr": 9.055622398368122e-05}
{"train_loss": 0.26445308327674866, "global_step": 54380, "epoch": 611, "lr": 9.055588490834535e-05}
{"train_loss": 0.2931593358516693, "global_step": 54381, "epoch": 611, "lr": 9.055554582755725e-05}
{"train_loss": 0.19563673436641693, "global_step": 54382, "epoch": 611, "lr": 9.055520674131694e-05}
{"train_loss": 0.3877531886100769, "global_step": 54383, "epoch": 611, "lr": 9.055486764962448e-05}
{"train_loss": 0.26624107360839844, "global_step": 54384, "epoch": 611, "lr": 9.055452855247993e-05}
{"train_loss": 0.22491295635700226, "global_step": 54385, "epoch": 611, "lr": 9.05541894498833e-05}
{"train_loss": 0.1988457441329956, "global_step": 54386, "epoch": 611, "lr": 9.055385034183465e-05}
{"train_loss": 0.2482612133026123, "global_step": 54387, "epoch": 611, "lr": 9.055351122833404e-05}
{"train_loss": 0.20585604012012482, "global_step": 54388, "epoch": 611, "lr": 9.05531721093815e-05}
{"train_loss": 0.35224175453186035, "global_step": 54389, "epoch": 611, "lr": 9.055283298497708e-05}
{"train_loss": 0.2263457477092743, "global_step": 54390, "epoch": 611, "lr": 9.055249385512083e-05}
{"train_loss": 0.200496107339859, "global_step": 54391, "epoch": 611, "lr": 9.055215471981279e-05}
{"train_loss": 0.2064833641052246, "global_step": 54392, "epoch": 611, "lr": 9.0551815579053e-05}
{"train_loss": 0.20096084475517273, "global_step": 54393, "epoch": 611, "lr": 9.055147643284152e-05}
{"train_loss": 0.25069668889045715, "global_step": 54394, "epoch": 611, "lr": 9.055113728117839e-05}
{"train_loss": 0.27901872992515564, "global_step": 54395, "epoch": 611, "lr": 9.055079812406366e-05}
{"train_loss": 0.2725169062614441, "global_step": 54396, "epoch": 611, "lr": 9.055045896149734e-05}
{"train_loss": 0.32806017994880676, "global_step": 54397, "epoch": 611, "lr": 9.055011979347954e-05}
{"train_loss": 0.21641813218593597, "global_step": 54398, "epoch": 611, "lr": 9.054978062001025e-05}
{"train_loss": 0.25426316261291504, "global_step": 54399, "epoch": 611, "lr": 9.054944144108955e-05}
{"train_loss": 0.18798112869262695, "global_step": 54400, "epoch": 611, "lr": 9.054910225671745e-05}
{"train_loss": 0.21898671984672546, "global_step": 54401, "epoch": 611, "lr": 9.054876306689404e-05}
{"train_loss": 0.3606914281845093, "global_step": 54402, "epoch": 611, "lr": 9.054842387161932e-05}
{"train_loss": 0.29907989501953125, "global_step": 54403, "epoch": 611, "lr": 9.054808467089337e-05}
{"train_loss": 0.3086923360824585, "global_step": 54404, "epoch": 611, "lr": 9.054774546471624e-05}
{"train_loss": 0.26657184958457947, "global_step": 54405, "epoch": 611, "lr": 9.054740625308795e-05}
{"train_loss": 0.18481267988681793, "global_step": 54406, "epoch": 611, "lr": 9.054706703600855e-05}
{"train_loss": 0.1565198451280594, "global_step": 54407, "epoch": 611, "lr": 9.054672781347809e-05}
{"train_loss": 0.2050584852695465, "global_step": 54408, "epoch": 611, "lr": 9.054638858549662e-05}
{"train_loss": 0.2617431581020355, "global_step": 54409, "epoch": 611, "lr": 9.054604935206418e-05}
{"train_loss": 0.22936603426933289, "global_step": 54410, "epoch": 611, "lr": 9.054571011318082e-05}
{"train_loss": 0.2003517746925354, "global_step": 54411, "epoch": 611, "lr": 9.054537086884657e-05}
{"train_loss": 0.23304003477096558, "global_step": 54412, "epoch": 611, "lr": 9.054503161906152e-05}
{"train_loss": 0.26484566926956177, "global_step": 54413, "epoch": 611, "lr": 9.054469236382566e-05}
{"train_loss": 0.21956971287727356, "global_step": 54414, "epoch": 611, "lr": 9.054435310313907e-05}
{"train_loss": 0.24659910798072815, "global_step": 54415, "epoch": 611, "lr": 9.054401383700177e-05}
{"train_loss": 0.2198960781097412, "global_step": 54416, "epoch": 611, "lr": 9.054367456541384e-05}
{"train_loss": 0.23681075870990753, "global_step": 54417, "epoch": 611, "lr": 9.054333528837531e-05}
{"train_loss": 0.23663410544395447, "global_step": 54418, "epoch": 611, "lr": 9.054299600588621e-05}
{"train_loss": 0.272830069065094, "global_step": 54419, "epoch": 611, "lr": 9.05426567179466e-05}
{"train_loss": 0.20120498538017273, "global_step": 54420, "epoch": 611, "lr": 9.054231742455652e-05}
{"train_loss": 0.23178060352802277, "global_step": 54421, "epoch": 611, "lr": 9.054197812571603e-05}
{"train_loss": 0.2824082672595978, "global_step": 54422, "epoch": 611, "lr": 9.054163882142517e-05}
{"train_loss": 0.2780514359474182, "global_step": 54423, "epoch": 611, "lr": 9.054129951168398e-05}
{"train_loss": 0.2210102677345276, "global_step": 54424, "epoch": 611, "lr": 9.054096019649249e-05}
{"train_loss": 0.199043869972229, "global_step": 54425, "epoch": 611, "lr": 9.054062087585077e-05}
{"train_loss": 0.23362119495868683, "global_step": 54426, "epoch": 611, "lr": 9.054028154975886e-05}
{"train_loss": 0.36034095287323, "global_step": 54427, "epoch": 611, "lr": 9.05399422182168e-05}
{"train_loss": 0.2620849609375, "global_step": 54428, "epoch": 611, "lr": 9.053960288122463e-05}
{"train_loss": 0.26656439900398254, "global_step": 54429, "epoch": 611, "lr": 9.053926353878241e-05}
{"train_loss": 0.21401122212409973, "global_step": 54430, "epoch": 611, "lr": 9.053892419089019e-05}
{"train_loss": 0.2804569602012634, "global_step": 54431, "epoch": 611, "lr": 9.053858483754801e-05}
{"train_loss": 0.26861444115638733, "global_step": 54432, "epoch": 611, "lr": 9.05382454787559e-05}
{"train_loss": 0.2809925973415375, "global_step": 54433, "epoch": 611, "lr": 9.053790611451391e-05}
{"train_loss": 0.20002582669258118, "global_step": 54434, "epoch": 611, "lr": 9.053756674482211e-05}
{"train_loss": 0.18911771476268768, "global_step": 54435, "epoch": 611, "lr": 9.053722736968053e-05}
{"train_loss": 0.3456370532512665, "global_step": 54436, "epoch": 611, "lr": 9.053688798908919e-05}
{"train_loss": 0.2926867604255676, "global_step": 54437, "epoch": 611, "lr": 9.053654860304818e-05}
{"train_loss": 0.2648259997367859, "global_step": 54438, "epoch": 611, "lr": 9.053620921155752e-05}
{"train_loss": 0.23301447927951813, "global_step": 54439, "epoch": 611, "lr": 9.053586981461726e-05}
{"train_loss": 0.20513375103473663, "global_step": 54440, "epoch": 611, "lr": 9.053553041222745e-05}
{"train_loss": 0.2877715528011322, "global_step": 54441, "epoch": 611, "lr": 9.053519100438811e-05}
{"train_loss": 0.2110045701265335, "global_step": 54442, "epoch": 611, "lr": 9.053485159109934e-05}
{"train_loss": 0.25267118215560913, "global_step": 54443, "epoch": 611, "lr": 9.053451217236113e-05}
{"train_loss": 0.16473151743412018, "global_step": 54444, "epoch": 611, "lr": 9.053417274817355e-05}
{"train_loss": 0.271456778049469, "global_step": 54445, "epoch": 611, "lr": 9.053383331853666e-05}
{"train_loss": 0.2694215178489685, "global_step": 54446, "epoch": 611, "lr": 9.053349388345048e-05}
{"train_loss": 0.20247673988342285, "global_step": 54447, "epoch": 611, "lr": 9.053315444291506e-05}
{"train_loss": 0.21022404730319977, "global_step": 54448, "epoch": 611, "lr": 9.053281499693048e-05}
{"train_loss": 0.2069401741027832, "global_step": 54449, "epoch": 611, "lr": 9.053247554549673e-05}
{"train_loss": 0.22562596201896667, "global_step": 54450, "epoch": 611, "lr": 9.053213608861388e-05}
{"train_loss": 0.2947706878185272, "global_step": 54451, "epoch": 611, "lr": 9.0531796626282e-05}
{"train_loss": 0.23644877970218658, "global_step": 54452, "epoch": 611, "lr": 9.053145715850108e-05}
{"train_loss": 0.27248048782348633, "global_step": 54453, "epoch": 611, "lr": 9.053111768527123e-05}
{"train_loss": 0.2425432950258255, "global_step": 54454, "epoch": 611, "lr": 9.053077820659246e-05}
{"train_loss": 0.2609633505344391, "global_step": 54455, "epoch": 611, "lr": 9.053043872246481e-05}
{"train_loss": 0.28453490138053894, "global_step": 54456, "epoch": 611, "lr": 9.053009923288836e-05}
{"train_loss": 0.30158358812332153, "global_step": 54457, "epoch": 611, "lr": 9.05297597378631e-05}
{"train_loss": 0.335387259721756, "global_step": 54458, "epoch": 611, "lr": 9.052942023738913e-05}
{"train_loss": 0.2969239354133606, "global_step": 54459, "epoch": 611, "lr": 9.052908073146646e-05}
{"train_loss": 0.3395387530326843, "global_step": 54460, "epoch": 611, "lr": 9.052874122009516e-05}
{"train_loss": 0.22226189076900482, "global_step": 54461, "epoch": 611, "lr": 9.052840170327526e-05}
{"train_loss": 0.26036497950553894, "global_step": 54462, "epoch": 611, "lr": 9.052806218100681e-05}
{"train_loss": 0.294861763715744, "global_step": 54463, "epoch": 611, "lr": 9.052772265328987e-05}
{"train_loss": 0.2692025601863861, "global_step": 54464, "epoch": 611, "lr": 9.052738312012445e-05}
{"train_loss": 0.2559841275215149, "global_step": 54465, "epoch": 611, "lr": 9.052704358151063e-05}
{"train_loss": 0.284037321805954, "global_step": 54466, "epoch": 611, "lr": 9.052670403744842e-05}
{"train_loss": 0.251424878165963, "global_step": 54467, "epoch": 611, "lr": 9.052636448793793e-05, "val_loss": 3.3722147941589355}
{"train_loss": 0.36958447098731995, "global_step": 54468, "epoch": 612, "lr": 9.052602493297912e-05}
{"train_loss": 0.24171096086502075, "global_step": 54469, "epoch": 612, "lr": 9.052568537257211e-05}
{"train_loss": 0.21955707669258118, "global_step": 54470, "epoch": 612, "lr": 9.052534580671691e-05}
{"train_loss": 0.2438606470823288, "global_step": 54471, "epoch": 612, "lr": 9.052500623541357e-05}
{"train_loss": 0.21764448285102844, "global_step": 54472, "epoch": 612, "lr": 9.052466665866213e-05}
{"train_loss": 0.2141202688217163, "global_step": 54473, "epoch": 612, "lr": 9.052432707646265e-05}
{"train_loss": 0.2432912290096283, "global_step": 54474, "epoch": 612, "lr": 9.052398748881517e-05}
{"train_loss": 0.30670398473739624, "global_step": 54475, "epoch": 612, "lr": 9.052364789571973e-05}
{"train_loss": 0.24813632667064667, "global_step": 54476, "epoch": 612, "lr": 9.052330829717636e-05}
{"train_loss": 0.19379110634326935, "global_step": 54477, "epoch": 612, "lr": 9.052296869318515e-05}
{"train_loss": 0.27313846349716187, "global_step": 54478, "epoch": 612, "lr": 9.052262908374613e-05}
{"train_loss": 0.2115703672170639, "global_step": 54479, "epoch": 612, "lr": 9.052228946885932e-05}
{"train_loss": 0.2236015498638153, "global_step": 54480, "epoch": 612, "lr": 9.052194984852478e-05}
{"train_loss": 0.22608992457389832, "global_step": 54481, "epoch": 612, "lr": 9.052161022274256e-05}
{"train_loss": 0.25814294815063477, "global_step": 54482, "epoch": 612, "lr": 9.05212705915127e-05}
{"train_loss": 0.2965247929096222, "global_step": 54483, "epoch": 612, "lr": 9.052093095483526e-05}
{"train_loss": 0.27153289318084717, "global_step": 54484, "epoch": 612, "lr": 9.052059131271027e-05}
{"train_loss": 0.24665671586990356, "global_step": 54485, "epoch": 612, "lr": 9.052025166513778e-05}
{"train_loss": 0.23667362332344055, "global_step": 54486, "epoch": 612, "lr": 9.051991201211782e-05}
{"train_loss": 0.3400128185749054, "global_step": 54487, "epoch": 612, "lr": 9.051957235365047e-05}
{"train_loss": 0.2716844081878662, "global_step": 54488, "epoch": 612, "lr": 9.051923268973577e-05}
{"train_loss": 0.21287840604782104, "global_step": 54489, "epoch": 612, "lr": 9.051889302037373e-05}
{"train_loss": 0.1973910927772522, "global_step": 54490, "epoch": 612, "lr": 9.051855334556443e-05}
{"train_loss": 0.22287234663963318, "global_step": 54491, "epoch": 612, "lr": 9.051821366530789e-05}
{"train_loss": 0.35749563574790955, "global_step": 54492, "epoch": 612, "lr": 9.051787397960418e-05}
{"train_loss": 0.23731182515621185, "global_step": 54493, "epoch": 612, "lr": 9.051753428845334e-05}
{"train_loss": 0.15175829827785492, "global_step": 54494, "epoch": 612, "lr": 9.05171945918554e-05}
{"train_loss": 0.3048199415206909, "global_step": 54495, "epoch": 612, "lr": 9.051685488981043e-05}
{"train_loss": 0.37582704424858093, "global_step": 54496, "epoch": 612, "lr": 9.051651518231845e-05}
{"train_loss": 0.266703337430954, "global_step": 54497, "epoch": 612, "lr": 9.051617546937953e-05}
{"train_loss": 0.3566822111606598, "global_step": 54498, "epoch": 612, "lr": 9.051583575099369e-05}
{"train_loss": 0.22470273077487946, "global_step": 54499, "epoch": 612, "lr": 9.0515496027161e-05}
{"train_loss": 0.2936762571334839, "global_step": 54500, "epoch": 612, "lr": 9.05151562978815e-05}
{"train_loss": 0.2642824947834015, "global_step": 54501, "epoch": 612, "lr": 9.051481656315522e-05}
{"train_loss": 0.21972794830799103, "global_step": 54502, "epoch": 612, "lr": 9.051447682298222e-05}
{"train_loss": 0.3964937627315521, "global_step": 54503, "epoch": 612, "lr": 9.051413707736254e-05}
{"train_loss": 0.39926600456237793, "global_step": 54504, "epoch": 612, "lr": 9.051379732629625e-05}
{"train_loss": 0.26195207238197327, "global_step": 54505, "epoch": 612, "lr": 9.051345756978335e-05}
{"train_loss": 0.2888408303260803, "global_step": 54506, "epoch": 612, "lr": 9.051311780782393e-05}
{"train_loss": 0.2448900192975998, "global_step": 54507, "epoch": 612, "lr": 9.051277804041798e-05}
{"train_loss": 0.27594563364982605, "global_step": 54508, "epoch": 612, "lr": 9.051243826756561e-05}
{"train_loss": 0.2461104840040207, "global_step": 54509, "epoch": 612, "lr": 9.051209848926682e-05}
{"train_loss": 0.23810884356498718, "global_step": 54510, "epoch": 612, "lr": 9.051175870552168e-05}
{"train_loss": 0.31365782022476196, "global_step": 54511, "epoch": 612, "lr": 9.051141891633024e-05}
{"train_loss": 0.23813237249851227, "global_step": 54512, "epoch": 612, "lr": 9.051107912169252e-05}
{"train_loss": 0.17769666016101837, "global_step": 54513, "epoch": 612, "lr": 9.051073932160857e-05}
{"train_loss": 0.22750376164913177, "global_step": 54514, "epoch": 612, "lr": 9.051039951607846e-05}
{"train_loss": 0.2123359739780426, "global_step": 54515, "epoch": 612, "lr": 9.051005970510222e-05}
{"train_loss": 0.23004043102264404, "global_step": 54516, "epoch": 612, "lr": 9.050971988867989e-05}
{"train_loss": 0.19935809075832367, "global_step": 54517, "epoch": 612, "lr": 9.050938006681152e-05}
{"train_loss": 0.2760397493839264, "global_step": 54518, "epoch": 612, "lr": 9.050904023949716e-05}
{"train_loss": 0.24705921113491058, "global_step": 54519, "epoch": 612, "lr": 9.050870040673686e-05}
{"train_loss": 0.24253953993320465, "global_step": 54520, "epoch": 612, "lr": 9.050836056853065e-05}
{"train_loss": 0.23037362098693848, "global_step": 54521, "epoch": 612, "lr": 9.050802072487858e-05}
{"train_loss": 0.3315773904323578, "global_step": 54522, "epoch": 612, "lr": 9.050768087578072e-05}
{"train_loss": 0.3299327492713928, "global_step": 54523, "epoch": 612, "lr": 9.050734102123708e-05}
{"train_loss": 0.19097000360488892, "global_step": 54524, "epoch": 612, "lr": 9.050700116124773e-05}
{"train_loss": 0.3031357526779175, "global_step": 54525, "epoch": 612, "lr": 9.05066612958127e-05}
{"train_loss": 0.2481124997138977, "global_step": 54526, "epoch": 612, "lr": 9.050632142493204e-05}
{"train_loss": 0.28227564692497253, "global_step": 54527, "epoch": 612, "lr": 9.050598154860581e-05}
{"train_loss": 0.3323732316493988, "global_step": 54528, "epoch": 612, "lr": 9.050564166683403e-05}
{"train_loss": 0.36202770471572876, "global_step": 54529, "epoch": 612, "lr": 9.050530177961677e-05}
{"train_loss": 0.23607923090457916, "global_step": 54530, "epoch": 612, "lr": 9.050496188695408e-05}
{"train_loss": 0.294200599193573, "global_step": 54531, "epoch": 612, "lr": 9.050462198884597e-05}
{"train_loss": 0.2762729227542877, "global_step": 54532, "epoch": 612, "lr": 9.050428208529253e-05}
{"train_loss": 0.3829237222671509, "global_step": 54533, "epoch": 612, "lr": 9.050394217629377e-05}
{"train_loss": 0.2529338002204895, "global_step": 54534, "epoch": 612, "lr": 9.050360226184972e-05}
{"train_loss": 0.20458132028579712, "global_step": 54535, "epoch": 612, "lr": 9.050326234196049e-05}
{"train_loss": 0.29482734203338623, "global_step": 54536, "epoch": 612, "lr": 9.05029224166261e-05}
{"train_loss": 0.1872486025094986, "global_step": 54537, "epoch": 612, "lr": 9.050258248584655e-05}
{"train_loss": 0.26257359981536865, "global_step": 54538, "epoch": 612, "lr": 9.050224254962194e-05}
{"train_loss": 0.2159501165151596, "global_step": 54539, "epoch": 612, "lr": 9.050190260795231e-05}
{"train_loss": 0.2666510343551636, "global_step": 54540, "epoch": 612, "lr": 9.050156266083767e-05}
{"train_loss": 0.2768653333187103, "global_step": 54541, "epoch": 612, "lr": 9.05012227082781e-05}
{"train_loss": 0.281624972820282, "global_step": 54542, "epoch": 612, "lr": 9.050088275027363e-05}
{"train_loss": 0.22937117516994476, "global_step": 54543, "epoch": 612, "lr": 9.050054278682431e-05}
{"train_loss": 0.28177687525749207, "global_step": 54544, "epoch": 612, "lr": 9.050020281793018e-05}
{"train_loss": 0.24832631647586823, "global_step": 54545, "epoch": 612, "lr": 9.049986284359131e-05}
{"train_loss": 0.2566554546356201, "global_step": 54546, "epoch": 612, "lr": 9.049952286380771e-05}
{"train_loss": 0.27273422479629517, "global_step": 54547, "epoch": 612, "lr": 9.049918287857944e-05}
{"train_loss": 0.30359020829200745, "global_step": 54548, "epoch": 612, "lr": 9.049884288790655e-05}
{"train_loss": 0.23351922631263733, "global_step": 54549, "epoch": 612, "lr": 9.049850289178911e-05}
{"train_loss": 0.26526719331741333, "global_step": 54550, "epoch": 612, "lr": 9.049816289022712e-05}
{"train_loss": 0.26960647106170654, "global_step": 54551, "epoch": 612, "lr": 9.049782288322064e-05}
{"train_loss": 0.1964908242225647, "global_step": 54552, "epoch": 612, "lr": 9.049748287076973e-05}
{"train_loss": 0.2282586693763733, "global_step": 54553, "epoch": 612, "lr": 9.049714285287441e-05}
{"train_loss": 0.27833297848701477, "global_step": 54554, "epoch": 612, "lr": 9.049680282953477e-05}
{"train_loss": 0.28058040142059326, "global_step": 54555, "epoch": 612, "lr": 9.049646280075082e-05}
{"train_loss": 0.26345429778768775, "global_step": 54556, "epoch": 612, "lr": 9.04961227665226e-05, "val_loss": 3.1838815212249756}
{"train_loss": 0.2716258764266968, "global_step": 54557, "epoch": 613, "lr": 9.049578272685018e-05}
{"train_loss": 0.2526187598705292, "global_step": 54558, "epoch": 613, "lr": 9.04954426817336e-05}
{"train_loss": 0.25563663244247437, "global_step": 54559, "epoch": 613, "lr": 9.04951026311729e-05}
{"train_loss": 0.20612719655036926, "global_step": 54560, "epoch": 613, "lr": 9.049476257516811e-05}
{"train_loss": 0.24518804252147675, "global_step": 54561, "epoch": 613, "lr": 9.049442251371932e-05}
{"train_loss": 0.26365089416503906, "global_step": 54562, "epoch": 613, "lr": 9.049408244682652e-05}
{"train_loss": 0.1828562319278717, "global_step": 54563, "epoch": 613, "lr": 9.049374237448979e-05}
{"train_loss": 0.20884686708450317, "global_step": 54564, "epoch": 613, "lr": 9.04934022967092e-05}
{"train_loss": 0.2782938778400421, "global_step": 54565, "epoch": 613, "lr": 9.049306221348475e-05}
{"train_loss": 0.2510983943939209, "global_step": 54566, "epoch": 613, "lr": 9.049272212481649e-05}
{"train_loss": 0.2125663310289383, "global_step": 54567, "epoch": 613, "lr": 9.049238203070448e-05}
{"train_loss": 0.23899255692958832, "global_step": 54568, "epoch": 613, "lr": 9.049204193114877e-05}
{"train_loss": 0.24658651649951935, "global_step": 54569, "epoch": 613, "lr": 9.049170182614939e-05}
{"train_loss": 0.2898150086402893, "global_step": 54570, "epoch": 613, "lr": 9.04913617157064e-05}
{"train_loss": 0.22769197821617126, "global_step": 54571, "epoch": 613, "lr": 9.049102159981983e-05}
{"train_loss": 0.2795840799808502, "global_step": 54572, "epoch": 613, "lr": 9.049068147848974e-05}
{"train_loss": 0.3133668899536133, "global_step": 54573, "epoch": 613, "lr": 9.049034135171618e-05}
{"train_loss": 0.2908696234226227, "global_step": 54574, "epoch": 613, "lr": 9.049000121949917e-05}
{"train_loss": 0.23732170462608337, "global_step": 54575, "epoch": 613, "lr": 9.04896610818388e-05}
{"train_loss": 0.28064751625061035, "global_step": 54576, "epoch": 613, "lr": 9.048932093873508e-05}
{"train_loss": 0.2675161361694336, "global_step": 54577, "epoch": 613, "lr": 9.048898079018804e-05}
{"train_loss": 0.2998676002025604, "global_step": 54578, "epoch": 613, "lr": 9.048864063619777e-05}
{"train_loss": 0.16393527388572693, "global_step": 54579, "epoch": 613, "lr": 9.048830047676431e-05}
{"train_loss": 0.23007355630397797, "global_step": 54580, "epoch": 613, "lr": 9.048796031188767e-05}
{"train_loss": 0.25884199142456055, "global_step": 54581, "epoch": 613, "lr": 9.048762014156792e-05}
{"train_loss": 0.36222895979881287, "global_step": 54582, "epoch": 613, "lr": 9.04872799658051e-05}
{"train_loss": 0.29224324226379395, "global_step": 54583, "epoch": 613, "lr": 9.048693978459926e-05}
{"train_loss": 0.2906140983104706, "global_step": 54584, "epoch": 613, "lr": 9.048659959795046e-05}
{"train_loss": 0.2917265295982361, "global_step": 54585, "epoch": 613, "lr": 9.04862594058587e-05}
{"train_loss": 0.32070666551589966, "global_step": 54586, "epoch": 613, "lr": 9.04859192083241e-05}
{"train_loss": 0.261446475982666, "global_step": 54587, "epoch": 613, "lr": 9.048557900534662e-05}
{"train_loss": 0.2098551094532013, "global_step": 54588, "epoch": 613, "lr": 9.048523879692636e-05}
{"train_loss": 0.2496250867843628, "global_step": 54589, "epoch": 613, "lr": 9.048489858306337e-05}
{"train_loss": 0.19837048649787903, "global_step": 54590, "epoch": 613, "lr": 9.048455836375767e-05}
{"train_loss": 0.23994313180446625, "global_step": 54591, "epoch": 613, "lr": 9.04842181390093e-05}
{"train_loss": 0.21976719796657562, "global_step": 54592, "epoch": 613, "lr": 9.048387790881834e-05}
{"train_loss": 0.24150556325912476, "global_step": 54593, "epoch": 613, "lr": 9.04835376731848e-05}
{"train_loss": 0.31447339057922363, "global_step": 54594, "epoch": 613, "lr": 9.048319743210874e-05}
{"train_loss": 0.23019351065158844, "global_step": 54595, "epoch": 613, "lr": 9.048285718559021e-05}
{"train_loss": 0.32009169459342957, "global_step": 54596, "epoch": 613, "lr": 9.048251693362928e-05}
{"train_loss": 0.22351302206516266, "global_step": 54597, "epoch": 613, "lr": 9.048217667622594e-05}
{"train_loss": 0.2567419409751892, "global_step": 54598, "epoch": 613, "lr": 9.048183641338026e-05}
{"train_loss": 0.37945133447647095, "global_step": 54599, "epoch": 613, "lr": 9.048149614509231e-05}
{"train_loss": 0.24345068633556366, "global_step": 54600, "epoch": 613, "lr": 9.048115587136211e-05}
{"train_loss": 0.3322274684906006, "global_step": 54601, "epoch": 613, "lr": 9.04808155921897e-05}
{"train_loss": 0.27083075046539307, "global_step": 54602, "epoch": 613, "lr": 9.048047530757516e-05}
{"train_loss": 0.21908481419086456, "global_step": 54603, "epoch": 613, "lr": 9.04801350175185e-05}
{"train_loss": 0.19399675726890564, "global_step": 54604, "epoch": 613, "lr": 9.04797947220198e-05}
{"train_loss": 0.28458932042121887, "global_step": 54605, "epoch": 613, "lr": 9.047945442107905e-05}
{"train_loss": 0.18116796016693115, "global_step": 54606, "epoch": 613, "lr": 9.047911411469636e-05}
{"train_loss": 0.2183896154165268, "global_step": 54607, "epoch": 613, "lr": 9.047877380287172e-05}
{"train_loss": 0.27318987250328064, "global_step": 54608, "epoch": 613, "lr": 9.047843348560522e-05}
{"train_loss": 0.2950037717819214, "global_step": 54609, "epoch": 613, "lr": 9.047809316289688e-05}
{"train_loss": 0.2761332392692566, "global_step": 54610, "epoch": 613, "lr": 9.047775283474677e-05}
{"train_loss": 0.3420339822769165, "global_step": 54611, "epoch": 613, "lr": 9.04774125011549e-05}
{"train_loss": 0.27217039465904236, "global_step": 54612, "epoch": 613, "lr": 9.047707216212135e-05}
{"train_loss": 0.237102672457695, "global_step": 54613, "epoch": 613, "lr": 9.047673181764613e-05}
{"train_loss": 0.22283491492271423, "global_step": 54614, "epoch": 613, "lr": 9.047639146772933e-05}
{"train_loss": 0.21505196392536163, "global_step": 54615, "epoch": 613, "lr": 9.047605111237096e-05}
{"train_loss": 0.30908527970314026, "global_step": 54616, "epoch": 613, "lr": 9.047571075157109e-05}
{"train_loss": 0.3050405979156494, "global_step": 54617, "epoch": 613, "lr": 9.047537038532974e-05}
{"train_loss": 0.32310712337493896, "global_step": 54618, "epoch": 613, "lr": 9.047503001364697e-05}
{"train_loss": 0.3442436456680298, "global_step": 54619, "epoch": 613, "lr": 9.047468963652283e-05}
{"train_loss": 0.2084687501192093, "global_step": 54620, "epoch": 613, "lr": 9.047434925395737e-05}
{"train_loss": 0.22604791820049286, "global_step": 54621, "epoch": 613, "lr": 9.04740088659506e-05}
{"train_loss": 0.17269983887672424, "global_step": 54622, "epoch": 613, "lr": 9.047366847250263e-05}
{"train_loss": 0.3150116801261902, "global_step": 54623, "epoch": 613, "lr": 9.047332807361344e-05}
{"train_loss": 0.3068650960922241, "global_step": 54624, "epoch": 613, "lr": 9.047298766928312e-05}
{"train_loss": 0.23241424560546875, "global_step": 54625, "epoch": 613, "lr": 9.04726472595117e-05}
{"train_loss": 0.2381761074066162, "global_step": 54626, "epoch": 613, "lr": 9.047230684429922e-05}
{"train_loss": 0.23490367829799652, "global_step": 54627, "epoch": 613, "lr": 9.047196642364572e-05}
{"train_loss": 0.20586803555488586, "global_step": 54628, "epoch": 613, "lr": 9.047162599755128e-05}
{"train_loss": 0.311562180519104, "global_step": 54629, "epoch": 613, "lr": 9.047128556601591e-05}
{"train_loss": 0.27063611149787903, "global_step": 54630, "epoch": 613, "lr": 9.047094512903967e-05}
{"train_loss": 0.3586331307888031, "global_step": 54631, "epoch": 613, "lr": 9.047060468662261e-05}
{"train_loss": 0.29224899411201477, "global_step": 54632, "epoch": 613, "lr": 9.047026423876477e-05}
{"train_loss": 0.3320043087005615, "global_step": 54633, "epoch": 613, "lr": 9.046992378546619e-05}
{"train_loss": 0.3022666573524475, "global_step": 54634, "epoch": 613, "lr": 9.046958332672692e-05}
{"train_loss": 0.3258115351200104, "global_step": 54635, "epoch": 613, "lr": 9.046924286254701e-05}
{"train_loss": 0.24392388761043549, "global_step": 54636, "epoch": 613, "lr": 9.046890239292651e-05}
{"train_loss": 0.22221475839614868, "global_step": 54637, "epoch": 613, "lr": 9.046856191786546e-05}
{"train_loss": 0.24061787128448486, "global_step": 54638, "epoch": 613, "lr": 9.04682214373639e-05}
{"train_loss": 0.2852637767791748, "global_step": 54639, "epoch": 613, "lr": 9.046788095142188e-05}
{"train_loss": 0.243197500705719, "global_step": 54640, "epoch": 613, "lr": 9.046754046003945e-05}
{"train_loss": 0.23159466683864594, "global_step": 54641, "epoch": 613, "lr": 9.046719996321665e-05}
{"train_loss": 0.16827984154224396, "global_step": 54642, "epoch": 613, "lr": 9.046685946095352e-05}
{"train_loss": 0.24058575928211212, "global_step": 54643, "epoch": 613, "lr": 9.046651895325013e-05}
{"train_loss": 0.2021038979291916, "global_step": 54644, "epoch": 613, "lr": 9.046617844010649e-05}
{"train_loss": 0.2595876425169827, "global_step": 54645, "epoch": 613, "lr": 9.046583792152269e-05, "val_loss": 3.259321451187134}
{"train_loss": 0.17959140241146088, "global_step": 54646, "epoch": 614, "lr": 9.046549739749874e-05}
{"train_loss": 0.26009437441825867, "global_step": 54647, "epoch": 614, "lr": 9.04651568680347e-05}
{"train_loss": 0.27312105894088745, "global_step": 54648, "epoch": 614, "lr": 9.04648163331306e-05}
{"train_loss": 0.23536260426044464, "global_step": 54649, "epoch": 614, "lr": 9.046447579278652e-05}
{"train_loss": 0.28029412031173706, "global_step": 54650, "epoch": 614, "lr": 9.046413524700247e-05}
{"train_loss": 0.20905272662639618, "global_step": 54651, "epoch": 614, "lr": 9.046379469577852e-05}
{"train_loss": 0.24948899447917938, "global_step": 54652, "epoch": 614, "lr": 9.046345413911469e-05}
{"train_loss": 0.25152459740638733, "global_step": 54653, "epoch": 614, "lr": 9.046311357701105e-05}
{"train_loss": 0.23531463742256165, "global_step": 54654, "epoch": 614, "lr": 9.046277300946764e-05}
{"train_loss": 0.35408738255500793, "global_step": 54655, "epoch": 614, "lr": 9.046243243648449e-05}
{"train_loss": 0.2899913191795349, "global_step": 54656, "epoch": 614, "lr": 9.046209185806167e-05}
{"train_loss": 0.30401694774627686, "global_step": 54657, "epoch": 614, "lr": 9.046175127419922e-05}
{"train_loss": 0.2509276270866394, "global_step": 54658, "epoch": 614, "lr": 9.046141068489717e-05}
{"train_loss": 0.2536545395851135, "global_step": 54659, "epoch": 614, "lr": 9.046107009015559e-05}
{"train_loss": 0.18521519005298615, "global_step": 54660, "epoch": 614, "lr": 9.04607294899745e-05}
{"train_loss": 0.24652442336082458, "global_step": 54661, "epoch": 614, "lr": 9.046038888435397e-05}
{"train_loss": 0.2086450755596161, "global_step": 54662, "epoch": 614, "lr": 9.046004827329402e-05}
{"train_loss": 0.18877533078193665, "global_step": 54663, "epoch": 614, "lr": 9.045970765679472e-05}
{"train_loss": 0.2893438935279846, "global_step": 54664, "epoch": 614, "lr": 9.045936703485612e-05}
{"train_loss": 0.295598566532135, "global_step": 54665, "epoch": 614, "lr": 9.045902640747822e-05}
{"train_loss": 0.2630046606063843, "global_step": 54666, "epoch": 614, "lr": 9.045868577466112e-05}
{"train_loss": 0.2743387818336487, "global_step": 54667, "epoch": 614, "lr": 9.045834513640481e-05}
{"train_loss": 0.22531913220882416, "global_step": 54668, "epoch": 614, "lr": 9.045800449270941e-05}
{"train_loss": 0.20371511578559875, "global_step": 54669, "epoch": 614, "lr": 9.04576638435749e-05}
{"train_loss": 0.18332888185977936, "global_step": 54670, "epoch": 614, "lr": 9.045732318900135e-05}
{"train_loss": 0.26297640800476074, "global_step": 54671, "epoch": 614, "lr": 9.045698252898882e-05}
{"train_loss": 0.3018113374710083, "global_step": 54672, "epoch": 614, "lr": 9.045664186353733e-05}
{"train_loss": 0.1966743916273117, "global_step": 54673, "epoch": 614, "lr": 9.045630119264694e-05}
{"train_loss": 0.2543138265609741, "global_step": 54674, "epoch": 614, "lr": 9.04559605163177e-05}
{"train_loss": 0.2546396255493164, "global_step": 54675, "epoch": 614, "lr": 9.045561983454964e-05}
{"train_loss": 0.2294161468744278, "global_step": 54676, "epoch": 614, "lr": 9.045527914734281e-05}
{"train_loss": 0.2296675145626068, "global_step": 54677, "epoch": 614, "lr": 9.045493845469728e-05}
{"train_loss": 0.2815876007080078, "global_step": 54678, "epoch": 614, "lr": 9.045459775661307e-05}
{"train_loss": 0.26381173729896545, "global_step": 54679, "epoch": 614, "lr": 9.045425705309022e-05}
{"train_loss": 0.21340548992156982, "global_step": 54680, "epoch": 614, "lr": 9.045391634412881e-05}
{"train_loss": 0.21818499267101288, "global_step": 54681, "epoch": 614, "lr": 9.045357562972884e-05}
{"train_loss": 0.22288267314434052, "global_step": 54682, "epoch": 614, "lr": 9.045323490989039e-05}
{"train_loss": 0.3425176441669464, "global_step": 54683, "epoch": 614, "lr": 9.04528941846135e-05}
{"train_loss": 0.3135685920715332, "global_step": 54684, "epoch": 614, "lr": 9.04525534538982e-05}
{"train_loss": 0.29726889729499817, "global_step": 54685, "epoch": 614, "lr": 9.045221271774456e-05}
{"train_loss": 0.16713038086891174, "global_step": 54686, "epoch": 614, "lr": 9.045187197615261e-05}
{"train_loss": 0.23496627807617188, "global_step": 54687, "epoch": 614, "lr": 9.04515312291224e-05}
{"train_loss": 0.242460235953331, "global_step": 54688, "epoch": 614, "lr": 9.045119047665399e-05}
{"train_loss": 0.319914847612381, "global_step": 54689, "epoch": 614, "lr": 9.045084971874738e-05}
{"train_loss": 0.3255368769168854, "global_step": 54690, "epoch": 614, "lr": 9.045050895540266e-05}
{"train_loss": 0.2932571768760681, "global_step": 54691, "epoch": 614, "lr": 9.045016818661985e-05}
{"train_loss": 0.29973870515823364, "global_step": 54692, "epoch": 614, "lr": 9.044982741239903e-05}
{"train_loss": 0.3202851414680481, "global_step": 54693, "epoch": 614, "lr": 9.04494866327402e-05}
{"train_loss": 0.25274768471717834, "global_step": 54694, "epoch": 614, "lr": 9.044914584764345e-05}
{"train_loss": 0.2630523443222046, "global_step": 54695, "epoch": 614, "lr": 9.04488050571088e-05}
{"train_loss": 0.26440754532814026, "global_step": 54696, "epoch": 614, "lr": 9.044846426113629e-05}
{"train_loss": 0.25495490431785583, "global_step": 54697, "epoch": 614, "lr": 9.0448123459726e-05}
{"train_loss": 0.3213653564453125, "global_step": 54698, "epoch": 614, "lr": 9.044778265287793e-05}
{"train_loss": 0.23289360105991364, "global_step": 54699, "epoch": 614, "lr": 9.044744184059215e-05}
{"train_loss": 0.2535206377506256, "global_step": 54700, "epoch": 614, "lr": 9.044710102286871e-05}
{"train_loss": 0.19006817042827606, "global_step": 54701, "epoch": 614, "lr": 9.044676019970766e-05}
{"train_loss": 0.37581613659858704, "global_step": 54702, "epoch": 614, "lr": 9.044641937110903e-05}
{"train_loss": 0.18420475721359253, "global_step": 54703, "epoch": 614, "lr": 9.044607853707286e-05}
{"train_loss": 0.27481719851493835, "global_step": 54704, "epoch": 614, "lr": 9.044573769759922e-05}
{"train_loss": 0.2425479292869568, "global_step": 54705, "epoch": 614, "lr": 9.044539685268814e-05}
{"train_loss": 0.36998218297958374, "global_step": 54706, "epoch": 614, "lr": 9.044505600233968e-05}
{"train_loss": 0.2910830080509186, "global_step": 54707, "epoch": 614, "lr": 9.044471514655386e-05}
{"train_loss": 0.30230650305747986, "global_step": 54708, "epoch": 614, "lr": 9.044437428533075e-05}
{"train_loss": 0.3081311285495758, "global_step": 54709, "epoch": 614, "lr": 9.044403341867038e-05}
{"train_loss": 0.1794929802417755, "global_step": 54710, "epoch": 614, "lr": 9.04436925465728e-05}
{"train_loss": 0.27494004368782043, "global_step": 54711, "epoch": 614, "lr": 9.044335166903806e-05}
{"train_loss": 0.19992877542972565, "global_step": 54712, "epoch": 614, "lr": 9.04430107860662e-05}
{"train_loss": 0.25053369998931885, "global_step": 54713, "epoch": 614, "lr": 9.04426698976573e-05}
{"train_loss": 0.3598901927471161, "global_step": 54714, "epoch": 614, "lr": 9.044232900381135e-05}
{"train_loss": 0.1862466037273407, "global_step": 54715, "epoch": 614, "lr": 9.044198810452843e-05}
{"train_loss": 0.3326703906059265, "global_step": 54716, "epoch": 614, "lr": 9.044164719980857e-05}
{"train_loss": 0.3305411636829376, "global_step": 54717, "epoch": 614, "lr": 9.044130628965181e-05}
{"train_loss": 0.2950695753097534, "global_step": 54718, "epoch": 614, "lr": 9.044096537405823e-05}
{"train_loss": 0.255258172750473, "global_step": 54719, "epoch": 614, "lr": 9.044062445302786e-05}
{"train_loss": 0.20676420629024506, "global_step": 54720, "epoch": 614, "lr": 9.044028352656073e-05}
{"train_loss": 0.18336597084999084, "global_step": 54721, "epoch": 614, "lr": 9.043994259465692e-05}
{"train_loss": 0.31045135855674744, "global_step": 54722, "epoch": 614, "lr": 9.043960165731642e-05}
{"train_loss": 0.3547075390815735, "global_step": 54723, "epoch": 614, "lr": 9.043926071453932e-05}
{"train_loss": 0.2996627986431122, "global_step": 54724, "epoch": 614, "lr": 9.043891976632566e-05}
{"train_loss": 0.23385852575302124, "global_step": 54725, "epoch": 614, "lr": 9.043857881267546e-05}
{"train_loss": 0.2124604731798172, "global_step": 54726, "epoch": 614, "lr": 9.04382378535888e-05}
{"train_loss": 0.2636132538318634, "global_step": 54727, "epoch": 614, "lr": 9.043789688906572e-05}
{"train_loss": 0.25173357129096985, "global_step": 54728, "epoch": 614, "lr": 9.043755591910626e-05}
{"train_loss": 0.33098024129867554, "global_step": 54729, "epoch": 614, "lr": 9.043721494371044e-05}
{"train_loss": 0.2601870894432068, "global_step": 54730, "epoch": 614, "lr": 9.043687396287834e-05}
{"train_loss": 0.2152801901102066, "global_step": 54731, "epoch": 614, "lr": 9.043653297661002e-05}
{"train_loss": 0.26391732692718506, "global_step": 54732, "epoch": 614, "lr": 9.043619198490547e-05}
{"train_loss": 0.2445109784603119, "global_step": 54733, "epoch": 614, "lr": 9.043585098776476e-05}
{"train_loss": 0.2624623598007674, "global_step": 54734, "epoch": 614, "lr": 9.043550998518797e-05, "val_loss": 3.238865375518799}
{"train_loss": 0.25581610202789307, "global_step": 54735, "epoch": 615, "lr": 9.04351689771751e-05}
{"train_loss": 0.28020578622817993, "global_step": 54736, "epoch": 615, "lr": 9.043482796372623e-05}
{"train_loss": 0.19658079743385315, "global_step": 54737, "epoch": 615, "lr": 9.043448694484138e-05}
{"train_loss": 0.20842087268829346, "global_step": 54738, "epoch": 615, "lr": 9.04341459205206e-05}
{"train_loss": 0.38991281390190125, "global_step": 54739, "epoch": 615, "lr": 9.043380489076396e-05}
{"train_loss": 0.25158315896987915, "global_step": 54740, "epoch": 615, "lr": 9.043346385557147e-05}
{"train_loss": 0.25904449820518494, "global_step": 54741, "epoch": 615, "lr": 9.04331228149432e-05}
{"train_loss": 0.3943461775779724, "global_step": 54742, "epoch": 615, "lr": 9.04327817688792e-05}
{"train_loss": 0.2929103374481201, "global_step": 54743, "epoch": 615, "lr": 9.043244071737948e-05}
{"train_loss": 0.20130307972431183, "global_step": 54744, "epoch": 615, "lr": 9.043209966044414e-05}
{"train_loss": 0.2112685590982437, "global_step": 54745, "epoch": 615, "lr": 9.043175859807318e-05}
{"train_loss": 0.22635504603385925, "global_step": 54746, "epoch": 615, "lr": 9.043141753026667e-05}
{"train_loss": 0.26735731959342957, "global_step": 54747, "epoch": 615, "lr": 9.043107645702465e-05}
{"train_loss": 0.27795976400375366, "global_step": 54748, "epoch": 615, "lr": 9.043073537834716e-05}
{"train_loss": 0.27634531259536743, "global_step": 54749, "epoch": 615, "lr": 9.043039429423425e-05}
{"train_loss": 0.2064858078956604, "global_step": 54750, "epoch": 615, "lr": 9.043005320468596e-05}
{"train_loss": 0.322800874710083, "global_step": 54751, "epoch": 615, "lr": 9.042971210970235e-05}
{"train_loss": 0.3175090253353119, "global_step": 54752, "epoch": 615, "lr": 9.042937100928346e-05}
{"train_loss": 0.289448618888855, "global_step": 54753, "epoch": 615, "lr": 9.042902990342933e-05}
{"train_loss": 0.3336714208126068, "global_step": 54754, "epoch": 615, "lr": 9.042868879214002e-05}
{"train_loss": 0.2753141224384308, "global_step": 54755, "epoch": 615, "lr": 9.042834767541555e-05}
{"train_loss": 0.25066670775413513, "global_step": 54756, "epoch": 615, "lr": 9.042800655325598e-05}
{"train_loss": 0.23131708800792694, "global_step": 54757, "epoch": 615, "lr": 9.042766542566137e-05}
{"train_loss": 0.43394702672958374, "global_step": 54758, "epoch": 615, "lr": 9.042732429263175e-05}
{"train_loss": 0.28721457719802856, "global_step": 54759, "epoch": 615, "lr": 9.042698315416717e-05}
{"train_loss": 0.2148822546005249, "global_step": 54760, "epoch": 615, "lr": 9.042664201026767e-05}
{"train_loss": 0.21440155804157257, "global_step": 54761, "epoch": 615, "lr": 9.042630086093328e-05}
{"train_loss": 0.3200565278530121, "global_step": 54762, "epoch": 615, "lr": 9.04259597061641e-05}
{"train_loss": 0.20731382071971893, "global_step": 54763, "epoch": 615, "lr": 9.042561854596014e-05}
{"train_loss": 0.31200554966926575, "global_step": 54764, "epoch": 615, "lr": 9.042527738032143e-05}
{"train_loss": 0.28355827927589417, "global_step": 54765, "epoch": 615, "lr": 9.042493620924805e-05}
{"train_loss": 0.2568208575248718, "global_step": 54766, "epoch": 615, "lr": 9.042459503274004e-05}
{"train_loss": 0.2402278333902359, "global_step": 54767, "epoch": 615, "lr": 9.04242538507974e-05}
{"train_loss": 0.1905498504638672, "global_step": 54768, "epoch": 615, "lr": 9.042391266342025e-05}
{"train_loss": 0.2399396151304245, "global_step": 54769, "epoch": 615, "lr": 9.042357147060857e-05}
{"train_loss": 0.30293700098991394, "global_step": 54770, "epoch": 615, "lr": 9.042323027236244e-05}
{"train_loss": 0.32866549491882324, "global_step": 54771, "epoch": 615, "lr": 9.04228890686819e-05}
{"train_loss": 0.25936612486839294, "global_step": 54772, "epoch": 615, "lr": 9.0422547859567e-05}
{"train_loss": 0.3154115080833435, "global_step": 54773, "epoch": 615, "lr": 9.042220664501777e-05}
{"train_loss": 0.19259683787822723, "global_step": 54774, "epoch": 615, "lr": 9.042186542503428e-05}
{"train_loss": 0.26322460174560547, "global_step": 54775, "epoch": 615, "lr": 9.042152419961656e-05}
{"train_loss": 0.22671186923980713, "global_step": 54776, "epoch": 615, "lr": 9.042118296876466e-05}
{"train_loss": 0.3122943341732025, "global_step": 54777, "epoch": 615, "lr": 9.042084173247862e-05}
{"train_loss": 0.15282240509986877, "global_step": 54778, "epoch": 615, "lr": 9.04205004907585e-05}
{"train_loss": 0.24914996325969696, "global_step": 54779, "epoch": 615, "lr": 9.042015924360432e-05}
{"train_loss": 0.18845213949680328, "global_step": 54780, "epoch": 615, "lr": 9.041981799101615e-05}
{"train_loss": 0.252729207277298, "global_step": 54781, "epoch": 615, "lr": 9.041947673299403e-05}
{"train_loss": 0.32965582609176636, "global_step": 54782, "epoch": 615, "lr": 9.0419135469538e-05}
{"train_loss": 0.3179022967815399, "global_step": 54783, "epoch": 615, "lr": 9.041879420064812e-05}
{"train_loss": 0.2971533536911011, "global_step": 54784, "epoch": 615, "lr": 9.041845292632442e-05}
{"train_loss": 0.2604737877845764, "global_step": 54785, "epoch": 615, "lr": 9.041811164656695e-05}
{"train_loss": 0.23971834778785706, "global_step": 54786, "epoch": 615, "lr": 9.041777036137575e-05}
{"train_loss": 0.18411938846111298, "global_step": 54787, "epoch": 615, "lr": 9.041742907075088e-05}
{"train_loss": 0.18253754079341888, "global_step": 54788, "epoch": 615, "lr": 9.041708777469238e-05}
{"train_loss": 0.25629425048828125, "global_step": 54789, "epoch": 615, "lr": 9.041674647320029e-05}
{"train_loss": 0.29345616698265076, "global_step": 54790, "epoch": 615, "lr": 9.041640516627467e-05}
{"train_loss": 0.36385098099708557, "global_step": 54791, "epoch": 615, "lr": 9.041606385391556e-05}
{"train_loss": 0.24321380257606506, "global_step": 54792, "epoch": 615, "lr": 9.0415722536123e-05}
{"train_loss": 0.23708274960517883, "global_step": 54793, "epoch": 615, "lr": 9.041538121289703e-05}
{"train_loss": 0.22088277339935303, "global_step": 54794, "epoch": 615, "lr": 9.041503988423771e-05}
{"train_loss": 0.28726616501808167, "global_step": 54795, "epoch": 615, "lr": 9.041469855014509e-05}
{"train_loss": 0.24355636537075043, "global_step": 54796, "epoch": 615, "lr": 9.041435721061919e-05}
{"train_loss": 0.3547433018684387, "global_step": 54797, "epoch": 615, "lr": 9.041401586566007e-05}
{"train_loss": 0.3984754681587219, "global_step": 54798, "epoch": 615, "lr": 9.04136745152678e-05}
{"train_loss": 0.30826887488365173, "global_step": 54799, "epoch": 615, "lr": 9.041333315944238e-05}
{"train_loss": 0.3273482620716095, "global_step": 54800, "epoch": 615, "lr": 9.04129917981839e-05}
{"train_loss": 0.35159456729888916, "global_step": 54801, "epoch": 615, "lr": 9.041265043149236e-05}
{"train_loss": 0.26454871892929077, "global_step": 54802, "epoch": 615, "lr": 9.041230905936786e-05}
{"train_loss": 0.3979673981666565, "global_step": 54803, "epoch": 615, "lr": 9.04119676818104e-05}
{"train_loss": 0.2770833671092987, "global_step": 54804, "epoch": 615, "lr": 9.041162629882006e-05}
{"train_loss": 0.29531899094581604, "global_step": 54805, "epoch": 615, "lr": 9.041128491039686e-05}
{"train_loss": 0.22331145405769348, "global_step": 54806, "epoch": 615, "lr": 9.041094351654085e-05}
{"train_loss": 0.28232547640800476, "global_step": 54807, "epoch": 615, "lr": 9.041060211725208e-05}
{"train_loss": 0.28802186250686646, "global_step": 54808, "epoch": 615, "lr": 9.041026071253061e-05}
{"train_loss": 0.2475658655166626, "global_step": 54809, "epoch": 615, "lr": 9.040991930237646e-05}
{"train_loss": 0.181054025888443, "global_step": 54810, "epoch": 615, "lr": 9.04095778867897e-05}
{"train_loss": 0.26374688744544983, "global_step": 54811, "epoch": 615, "lr": 9.040923646577035e-05}
{"train_loss": 0.3826121687889099, "global_step": 54812, "epoch": 615, "lr": 9.040889503931849e-05}
{"train_loss": 0.3048624098300934, "global_step": 54813, "epoch": 615, "lr": 9.040855360743414e-05}
{"train_loss": 0.21185527741909027, "global_step": 54814, "epoch": 615, "lr": 9.040821217011735e-05}
{"train_loss": 0.238056942820549, "global_step": 54815, "epoch": 615, "lr": 9.040787072736815e-05}
{"train_loss": 0.2359565943479538, "global_step": 54816, "epoch": 615, "lr": 9.040752927918663e-05}
{"train_loss": 0.30366626381874084, "global_step": 54817, "epoch": 615, "lr": 9.040718782557281e-05}
{"train_loss": 0.20533333718776703, "global_step": 54818, "epoch": 615, "lr": 9.040684636652672e-05}
{"train_loss": 0.36589643359184265, "global_step": 54819, "epoch": 615, "lr": 9.040650490204842e-05}
{"train_loss": 0.3052337169647217, "global_step": 54820, "epoch": 615, "lr": 9.040616343213797e-05}
{"train_loss": 0.2900370955467224, "global_step": 54821, "epoch": 615, "lr": 9.04058219567954e-05}
{"train_loss": 0.31151437759399414, "global_step": 54822, "epoch": 615, "lr": 9.040548047602076e-05}
{"train_loss": 0.27307746634724434, "global_step": 54823, "epoch": 615, "lr": 9.040513898981411e-05, "val_loss": 3.313063383102417, "train_action_mse_error": 28.790624618530273}
{"train_loss": 0.1669130027294159, "global_step": 54824, "epoch": 616, "lr": 9.040479749817546e-05}
{"train_loss": 0.2292913943529129, "global_step": 54825, "epoch": 616, "lr": 9.040445600110489e-05}
{"train_loss": 0.17318594455718994, "global_step": 54826, "epoch": 616, "lr": 9.040411449860242e-05}
{"train_loss": 0.2478662133216858, "global_step": 54827, "epoch": 616, "lr": 9.040377299066812e-05}
{"train_loss": 0.183068186044693, "global_step": 54828, "epoch": 616, "lr": 9.040343147730203e-05}
{"train_loss": 0.31490960717201233, "global_step": 54829, "epoch": 616, "lr": 9.040308995850418e-05}
{"train_loss": 0.23924671113491058, "global_step": 54830, "epoch": 616, "lr": 9.040274843427463e-05}
{"train_loss": 0.30631646513938904, "global_step": 54831, "epoch": 616, "lr": 9.040240690461343e-05}
{"train_loss": 0.2889362871646881, "global_step": 54832, "epoch": 616, "lr": 9.040206536952062e-05}
{"train_loss": 0.20223024487495422, "global_step": 54833, "epoch": 616, "lr": 9.040172382899623e-05}
{"train_loss": 0.192466139793396, "global_step": 54834, "epoch": 616, "lr": 9.040138228304034e-05}
{"train_loss": 0.2784140706062317, "global_step": 54835, "epoch": 616, "lr": 9.040104073165296e-05}
{"train_loss": 0.30055180191993713, "global_step": 54836, "epoch": 616, "lr": 9.040069917483416e-05}
{"train_loss": 0.2826243042945862, "global_step": 54837, "epoch": 616, "lr": 9.040035761258397e-05}
{"train_loss": 0.3379632532596588, "global_step": 54838, "epoch": 616, "lr": 9.040001604490244e-05}
{"train_loss": 0.25357452034950256, "global_step": 54839, "epoch": 616, "lr": 9.039967447178964e-05}
{"train_loss": 0.3918459415435791, "global_step": 54840, "epoch": 616, "lr": 9.039933289324558e-05}
{"train_loss": 0.3340010344982147, "global_step": 54841, "epoch": 616, "lr": 9.039899130927035e-05}
{"train_loss": 0.27131524682044983, "global_step": 54842, "epoch": 616, "lr": 9.039864971986394e-05}
{"train_loss": 0.28826385736465454, "global_step": 54843, "epoch": 616, "lr": 9.039830812502643e-05}
{"train_loss": 0.33327749371528625, "global_step": 54844, "epoch": 616, "lr": 9.039796652475785e-05}
{"train_loss": 0.24434590339660645, "global_step": 54845, "epoch": 616, "lr": 9.039762491905828e-05}
{"train_loss": 0.39214345812797546, "global_step": 54846, "epoch": 616, "lr": 9.039728330792772e-05}
{"train_loss": 0.27224934101104736, "global_step": 54847, "epoch": 616, "lr": 9.039694169136625e-05}
{"train_loss": 0.32103291153907776, "global_step": 54848, "epoch": 616, "lr": 9.039660006937389e-05}
{"train_loss": 0.2606542706489563, "global_step": 54849, "epoch": 616, "lr": 9.039625844195072e-05}
{"train_loss": 0.24405229091644287, "global_step": 54850, "epoch": 616, "lr": 9.039591680909674e-05}
{"train_loss": 0.24652844667434692, "global_step": 54851, "epoch": 616, "lr": 9.039557517081205e-05}
{"train_loss": 0.38568392395973206, "global_step": 54852, "epoch": 616, "lr": 9.039523352709664e-05}
{"train_loss": 0.4158100485801697, "global_step": 54853, "epoch": 616, "lr": 9.03948918779506e-05}
{"train_loss": 0.28706714510917664, "global_step": 54854, "epoch": 616, "lr": 9.039455022337396e-05}
{"train_loss": 0.27060678601264954, "global_step": 54855, "epoch": 616, "lr": 9.039420856336676e-05}
{"train_loss": 0.2784748077392578, "global_step": 54856, "epoch": 616, "lr": 9.039386689792905e-05}
{"train_loss": 0.2490612268447876, "global_step": 54857, "epoch": 616, "lr": 9.039352522706089e-05}
{"train_loss": 0.21333646774291992, "global_step": 54858, "epoch": 616, "lr": 9.039318355076229e-05}
{"train_loss": 0.2943020462989807, "global_step": 54859, "epoch": 616, "lr": 9.039284186903335e-05}
{"train_loss": 0.24384573101997375, "global_step": 54860, "epoch": 616, "lr": 9.039250018187404e-05}
{"train_loss": 0.26910993456840515, "global_step": 54861, "epoch": 616, "lr": 9.039215848928448e-05}
{"train_loss": 0.3211926221847534, "global_step": 54862, "epoch": 616, "lr": 9.039181679126468e-05}
{"train_loss": 0.37708133459091187, "global_step": 54863, "epoch": 616, "lr": 9.039147508781469e-05}
{"train_loss": 0.31369221210479736, "global_step": 54864, "epoch": 616, "lr": 9.039113337893456e-05}
{"train_loss": 0.38090643286705017, "global_step": 54865, "epoch": 616, "lr": 9.039079166462434e-05}
{"train_loss": 0.30923891067504883, "global_step": 54866, "epoch": 616, "lr": 9.039044994488408e-05}
{"train_loss": 0.17422334849834442, "global_step": 54867, "epoch": 616, "lr": 9.039010821971381e-05}
{"train_loss": 0.30533361434936523, "global_step": 54868, "epoch": 616, "lr": 9.038976648911357e-05}
{"train_loss": 0.22134150564670563, "global_step": 54869, "epoch": 616, "lr": 9.038942475308343e-05}
{"train_loss": 0.2474149614572525, "global_step": 54870, "epoch": 616, "lr": 9.03890830116234e-05}
{"train_loss": 0.25191277265548706, "global_step": 54871, "epoch": 616, "lr": 9.038874126473358e-05}
{"train_loss": 0.28780680894851685, "global_step": 54872, "epoch": 616, "lr": 9.038839951241398e-05}
{"train_loss": 0.23844441771507263, "global_step": 54873, "epoch": 616, "lr": 9.038805775466463e-05}
{"train_loss": 0.3631182312965393, "global_step": 54874, "epoch": 616, "lr": 9.038771599148564e-05}
{"train_loss": 0.3025457561016083, "global_step": 54875, "epoch": 616, "lr": 9.038737422287697e-05}
{"train_loss": 0.28187474608421326, "global_step": 54876, "epoch": 616, "lr": 9.038703244883873e-05}
{"train_loss": 0.26351186633110046, "global_step": 54877, "epoch": 616, "lr": 9.038669066937095e-05}
{"train_loss": 0.22530516982078552, "global_step": 54878, "epoch": 616, "lr": 9.038634888447366e-05}
{"train_loss": 0.30331486463546753, "global_step": 54879, "epoch": 616, "lr": 9.038600709414692e-05}
{"train_loss": 0.23643162846565247, "global_step": 54880, "epoch": 616, "lr": 9.038566529839078e-05}
{"train_loss": 0.33144354820251465, "global_step": 54881, "epoch": 616, "lr": 9.038532349720527e-05}
{"train_loss": 0.22761894762516022, "global_step": 54882, "epoch": 616, "lr": 9.038498169059047e-05}
{"train_loss": 0.26526039838790894, "global_step": 54883, "epoch": 616, "lr": 9.038463987854637e-05}
{"train_loss": 0.21444198489189148, "global_step": 54884, "epoch": 616, "lr": 9.038429806107306e-05}
{"train_loss": 0.22604045271873474, "global_step": 54885, "epoch": 616, "lr": 9.038395623817057e-05}
{"train_loss": 0.208440363407135, "global_step": 54886, "epoch": 616, "lr": 9.038361440983896e-05}
{"train_loss": 0.20373129844665527, "global_step": 54887, "epoch": 616, "lr": 9.038327257607826e-05}
{"train_loss": 0.22674612700939178, "global_step": 54888, "epoch": 616, "lr": 9.038293073688852e-05}
{"train_loss": 0.2557274103164673, "global_step": 54889, "epoch": 616, "lr": 9.038258889226978e-05}
{"train_loss": 0.29122722148895264, "global_step": 54890, "epoch": 616, "lr": 9.038224704222212e-05}
{"train_loss": 0.3150871992111206, "global_step": 54891, "epoch": 616, "lr": 9.038190518674552e-05}
{"train_loss": 0.14865492284297943, "global_step": 54892, "epoch": 616, "lr": 9.038156332584009e-05}
{"train_loss": 0.19709429144859314, "global_step": 54893, "epoch": 616, "lr": 9.038122145950585e-05}
{"train_loss": 0.2527492940425873, "global_step": 54894, "epoch": 616, "lr": 9.038087958774284e-05}
{"train_loss": 0.21126499772071838, "global_step": 54895, "epoch": 616, "lr": 9.038053771055112e-05}
{"train_loss": 0.19943854212760925, "global_step": 54896, "epoch": 616, "lr": 9.038019582793072e-05}
{"train_loss": 0.2300931066274643, "global_step": 54897, "epoch": 616, "lr": 9.03798539398817e-05}
{"train_loss": 0.284498393535614, "global_step": 54898, "epoch": 616, "lr": 9.03795120464041e-05}
{"train_loss": 0.269506573677063, "global_step": 54899, "epoch": 616, "lr": 9.037917014749796e-05}
{"train_loss": 0.2393275648355484, "global_step": 54900, "epoch": 616, "lr": 9.037882824316335e-05}
{"train_loss": 0.24828460812568665, "global_step": 54901, "epoch": 616, "lr": 9.037848633340028e-05}
{"train_loss": 0.19274428486824036, "global_step": 54902, "epoch": 616, "lr": 9.037814441820882e-05}
{"train_loss": 0.3008955121040344, "global_step": 54903, "epoch": 616, "lr": 9.037780249758902e-05}
{"train_loss": 0.2361375242471695, "global_step": 54904, "epoch": 616, "lr": 9.037746057154091e-05}
{"train_loss": 0.30499234795570374, "global_step": 54905, "epoch": 616, "lr": 9.037711864006453e-05}
{"train_loss": 0.21787628531455994, "global_step": 54906, "epoch": 616, "lr": 9.037677670315996e-05}
{"train_loss": 0.25107941031455994, "global_step": 54907, "epoch": 616, "lr": 9.037643476082721e-05}
{"train_loss": 0.204001784324646, "global_step": 54908, "epoch": 616, "lr": 9.037609281306635e-05}
{"train_loss": 0.2689785957336426, "global_step": 54909, "epoch": 616, "lr": 9.037575085987742e-05}
{"train_loss": 0.3216540217399597, "global_step": 54910, "epoch": 616, "lr": 9.037540890126044e-05}
{"train_loss": 0.2242286056280136, "global_step": 54911, "epoch": 616, "lr": 9.03750669372155e-05}
{"train_loss": 0.2663092805763309, "global_step": 54912, "epoch": 616, "lr": 9.037472496774262e-05, "val_loss": 3.3672215938568115}
{"train_loss": 0.2141990065574646, "global_step": 54913, "epoch": 617, "lr": 9.037438299284184e-05}
{"train_loss": 0.22931784391403198, "global_step": 54914, "epoch": 617, "lr": 9.037404101251324e-05}
{"train_loss": 0.15993444621562958, "global_step": 54915, "epoch": 617, "lr": 9.03736990267568e-05}
{"train_loss": 0.24533814191818237, "global_step": 54916, "epoch": 617, "lr": 9.037335703557264e-05}
{"train_loss": 0.18210534751415253, "global_step": 54917, "epoch": 617, "lr": 9.037301503896077e-05}
{"train_loss": 0.29927754402160645, "global_step": 54918, "epoch": 617, "lr": 9.037267303692125e-05}
{"train_loss": 0.2934560775756836, "global_step": 54919, "epoch": 617, "lr": 9.03723310294541e-05}
{"train_loss": 0.240143820643425, "global_step": 54920, "epoch": 617, "lr": 9.037198901655938e-05}
{"train_loss": 0.30542391538619995, "global_step": 54921, "epoch": 617, "lr": 9.037164699823715e-05}
{"train_loss": 0.27352309226989746, "global_step": 54922, "epoch": 617, "lr": 9.037130497448744e-05}
{"train_loss": 0.3017064929008484, "global_step": 54923, "epoch": 617, "lr": 9.03709629453103e-05}
{"train_loss": 0.2575172483921051, "global_step": 54924, "epoch": 617, "lr": 9.037062091070578e-05}
{"train_loss": 0.2497207671403885, "global_step": 54925, "epoch": 617, "lr": 9.037027887067391e-05}
{"train_loss": 0.20412388443946838, "global_step": 54926, "epoch": 617, "lr": 9.036993682521476e-05}
{"train_loss": 0.20407482981681824, "global_step": 54927, "epoch": 617, "lr": 9.036959477432836e-05}
{"train_loss": 0.2434936910867691, "global_step": 54928, "epoch": 617, "lr": 9.036925271801476e-05}
{"train_loss": 0.25516024231910706, "global_step": 54929, "epoch": 617, "lr": 9.036891065627401e-05}
{"train_loss": 0.4225444495677948, "global_step": 54930, "epoch": 617, "lr": 9.036856858910614e-05}
{"train_loss": 0.2560412287712097, "global_step": 54931, "epoch": 617, "lr": 9.036822651651122e-05}
{"train_loss": 0.35153499245643616, "global_step": 54932, "epoch": 617, "lr": 9.036788443848928e-05}
{"train_loss": 0.2893019914627075, "global_step": 54933, "epoch": 617, "lr": 9.036754235504036e-05}
{"train_loss": 0.29170429706573486, "global_step": 54934, "epoch": 617, "lr": 9.036720026616453e-05}
{"train_loss": 0.21523964405059814, "global_step": 54935, "epoch": 617, "lr": 9.036685817186181e-05}
{"train_loss": 0.32631850242614746, "global_step": 54936, "epoch": 617, "lr": 9.036651607213228e-05}
{"train_loss": 0.2732888460159302, "global_step": 54937, "epoch": 617, "lr": 9.036617396697595e-05}
{"train_loss": 0.2577197849750519, "global_step": 54938, "epoch": 617, "lr": 9.036583185639287e-05}
{"train_loss": 0.18777309358119965, "global_step": 54939, "epoch": 617, "lr": 9.036548974038311e-05}
{"train_loss": 0.2780950665473938, "global_step": 54940, "epoch": 617, "lr": 9.03651476189467e-05}
{"train_loss": 0.15781733393669128, "global_step": 54941, "epoch": 617, "lr": 9.036480549208368e-05}
{"train_loss": 0.21677199006080627, "global_step": 54942, "epoch": 617, "lr": 9.036446335979411e-05}
{"train_loss": 0.23976175487041473, "global_step": 54943, "epoch": 617, "lr": 9.036412122207802e-05}
{"train_loss": 0.3070652484893799, "global_step": 54944, "epoch": 617, "lr": 9.036377907893548e-05}
{"train_loss": 0.31555861234664917, "global_step": 54945, "epoch": 617, "lr": 9.036343693036652e-05}
{"train_loss": 0.2937646806240082, "global_step": 54946, "epoch": 617, "lr": 9.036309477637119e-05}
{"train_loss": 0.20285312831401825, "global_step": 54947, "epoch": 617, "lr": 9.036275261694953e-05}
{"train_loss": 0.2887433171272278, "global_step": 54948, "epoch": 617, "lr": 9.036241045210159e-05}
{"train_loss": 0.2794630825519562, "global_step": 54949, "epoch": 617, "lr": 9.03620682818274e-05}
{"train_loss": 0.2613622546195984, "global_step": 54950, "epoch": 617, "lr": 9.036172610612705e-05}
{"train_loss": 0.2878042459487915, "global_step": 54951, "epoch": 617, "lr": 9.036138392500053e-05}
{"train_loss": 0.21767054498195648, "global_step": 54952, "epoch": 617, "lr": 9.036104173844793e-05}
{"train_loss": 0.21982553601264954, "global_step": 54953, "epoch": 617, "lr": 9.036069954646929e-05}
{"train_loss": 0.26478978991508484, "global_step": 54954, "epoch": 617, "lr": 9.036035734906463e-05}
{"train_loss": 0.37680327892303467, "global_step": 54955, "epoch": 617, "lr": 9.036001514623401e-05}
{"train_loss": 0.25909414887428284, "global_step": 54956, "epoch": 617, "lr": 9.03596729379775e-05}
{"train_loss": 0.17625702917575836, "global_step": 54957, "epoch": 617, "lr": 9.035933072429508e-05}
{"train_loss": 0.25231537222862244, "global_step": 54958, "epoch": 617, "lr": 9.035898850518688e-05}
{"train_loss": 0.3515298068523407, "global_step": 54959, "epoch": 617, "lr": 9.035864628065289e-05}
{"train_loss": 0.27131959795951843, "global_step": 54960, "epoch": 617, "lr": 9.035830405069318e-05}
{"train_loss": 0.20681944489479065, "global_step": 54961, "epoch": 617, "lr": 9.035796181530778e-05}
{"train_loss": 0.34905025362968445, "global_step": 54962, "epoch": 617, "lr": 9.035761957449673e-05}
{"train_loss": 0.27821364998817444, "global_step": 54963, "epoch": 617, "lr": 9.035727732826012e-05}
{"train_loss": 0.24360163509845734, "global_step": 54964, "epoch": 617, "lr": 9.035693507659795e-05}
{"train_loss": 0.26441672444343567, "global_step": 54965, "epoch": 617, "lr": 9.035659281951029e-05}
{"train_loss": 0.22337859869003296, "global_step": 54966, "epoch": 617, "lr": 9.035625055699717e-05}
{"train_loss": 0.268497109413147, "global_step": 54967, "epoch": 617, "lr": 9.035590828905864e-05}
{"train_loss": 0.2545481324195862, "global_step": 54968, "epoch": 617, "lr": 9.035556601569476e-05}
{"train_loss": 0.23258744180202484, "global_step": 54969, "epoch": 617, "lr": 9.035522373690557e-05}
{"train_loss": 0.187479630112648, "global_step": 54970, "epoch": 617, "lr": 9.035488145269109e-05}
{"train_loss": 0.3476320803165436, "global_step": 54971, "epoch": 617, "lr": 9.035453916305141e-05}
{"train_loss": 0.30125465989112854, "global_step": 54972, "epoch": 617, "lr": 9.035419686798654e-05}
{"train_loss": 0.16658470034599304, "global_step": 54973, "epoch": 617, "lr": 9.035385456749654e-05}
{"train_loss": 0.3945797085762024, "global_step": 54974, "epoch": 617, "lr": 9.035351226158147e-05}
{"train_loss": 0.2299317568540573, "global_step": 54975, "epoch": 617, "lr": 9.035316995024137e-05}
{"train_loss": 0.136933833360672, "global_step": 54976, "epoch": 617, "lr": 9.035282763347626e-05}
{"train_loss": 0.34421035647392273, "global_step": 54977, "epoch": 617, "lr": 9.03524853112862e-05}
{"train_loss": 0.2581901550292969, "global_step": 54978, "epoch": 617, "lr": 9.035214298367125e-05}
{"train_loss": 0.2879117429256439, "global_step": 54979, "epoch": 617, "lr": 9.035180065063145e-05}
{"train_loss": 0.20470866560935974, "global_step": 54980, "epoch": 617, "lr": 9.035145831216683e-05}
{"train_loss": 0.2797504663467407, "global_step": 54981, "epoch": 617, "lr": 9.035111596827745e-05}
{"train_loss": 0.3191014528274536, "global_step": 54982, "epoch": 617, "lr": 9.035077361896335e-05}
{"train_loss": 0.3071429133415222, "global_step": 54983, "epoch": 617, "lr": 9.03504312642246e-05}
{"train_loss": 0.40205857157707214, "global_step": 54984, "epoch": 617, "lr": 9.035008890406121e-05}
{"train_loss": 0.34012624621391296, "global_step": 54985, "epoch": 617, "lr": 9.034974653847325e-05}
{"train_loss": 0.24626876413822174, "global_step": 54986, "epoch": 617, "lr": 9.034940416746076e-05}
{"train_loss": 0.20638862252235413, "global_step": 54987, "epoch": 617, "lr": 9.034906179102378e-05}
{"train_loss": 0.2132667601108551, "global_step": 54988, "epoch": 617, "lr": 9.034871940916235e-05}
{"train_loss": 0.206892192363739, "global_step": 54989, "epoch": 617, "lr": 9.034837702187654e-05}
{"train_loss": 0.2533673644065857, "global_step": 54990, "epoch": 617, "lr": 9.034803462916638e-05}
{"train_loss": 0.2823757827281952, "global_step": 54991, "epoch": 617, "lr": 9.034769223103192e-05}
{"train_loss": 0.1993085891008377, "global_step": 54992, "epoch": 617, "lr": 9.034734982747319e-05}
{"train_loss": 0.21557007730007172, "global_step": 54993, "epoch": 617, "lr": 9.034700741849027e-05}
{"train_loss": 0.22508655488491058, "global_step": 54994, "epoch": 617, "lr": 9.034666500408319e-05}
{"train_loss": 0.18175584077835083, "global_step": 54995, "epoch": 617, "lr": 9.034632258425197e-05}
{"train_loss": 0.2160208821296692, "global_step": 54996, "epoch": 617, "lr": 9.03459801589967e-05}
{"train_loss": 0.22252076864242554, "global_step": 54997, "epoch": 617, "lr": 9.03456377283174e-05}
{"train_loss": 0.2810207009315491, "global_step": 54998, "epoch": 617, "lr": 9.03452952922141e-05}
{"train_loss": 0.23108144104480743, "global_step": 54999, "epoch": 617, "lr": 9.034495285068689e-05}
{"train_loss": 0.19856639206409454, "global_step": 55000, "epoch": 617, "lr": 9.034461040373579e-05}
{"train_loss": 0.25853919949424403, "global_step": 55001, "epoch": 617, "lr": 9.034426795136083e-05, "val_loss": 3.197866678237915}
{"train_loss": 0.20315977931022644, "global_step": 55002, "epoch": 618, "lr": 9.03439254935621e-05}
{"train_loss": 0.2790750563144684, "global_step": 55003, "epoch": 618, "lr": 9.034358303033961e-05}
{"train_loss": 0.2399764358997345, "global_step": 55004, "epoch": 618, "lr": 9.034324056169342e-05}
{"train_loss": 0.25232407450675964, "global_step": 55005, "epoch": 618, "lr": 9.034289808762358e-05}
{"train_loss": 0.2742493748664856, "global_step": 55006, "epoch": 618, "lr": 9.034255560813011e-05}
{"train_loss": 0.23784105479717255, "global_step": 55007, "epoch": 618, "lr": 9.034221312321308e-05}
{"train_loss": 0.32654234766960144, "global_step": 55008, "epoch": 618, "lr": 9.034187063287254e-05}
{"train_loss": 0.223237544298172, "global_step": 55009, "epoch": 618, "lr": 9.034152813710852e-05}
{"train_loss": 0.36474329233169556, "global_step": 55010, "epoch": 618, "lr": 9.034118563592107e-05}
{"train_loss": 0.2728264033794403, "global_step": 55011, "epoch": 618, "lr": 9.034084312931026e-05}
{"train_loss": 0.2836184501647949, "global_step": 55012, "epoch": 618, "lr": 9.03405006172761e-05}
{"train_loss": 0.2598005533218384, "global_step": 55013, "epoch": 618, "lr": 9.034015809981866e-05}
{"train_loss": 0.22137942910194397, "global_step": 55014, "epoch": 618, "lr": 9.033981557693795e-05}
{"train_loss": 0.18755482137203217, "global_step": 55015, "epoch": 618, "lr": 9.033947304863407e-05}
{"train_loss": 0.23358002305030823, "global_step": 55016, "epoch": 618, "lr": 9.033913051490703e-05}
{"train_loss": 0.16578832268714905, "global_step": 55017, "epoch": 618, "lr": 9.033878797575688e-05}
{"train_loss": 0.27689170837402344, "global_step": 55018, "epoch": 618, "lr": 9.033844543118369e-05}
{"train_loss": 0.29506048560142517, "global_step": 55019, "epoch": 618, "lr": 9.033810288118748e-05}
{"train_loss": 0.2497226595878601, "global_step": 55020, "epoch": 618, "lr": 9.03377603257683e-05}
{"train_loss": 0.19443273544311523, "global_step": 55021, "epoch": 618, "lr": 9.033741776492621e-05}
{"train_loss": 0.17247892916202545, "global_step": 55022, "epoch": 618, "lr": 9.033707519866123e-05}
{"train_loss": 0.27101436257362366, "global_step": 55023, "epoch": 618, "lr": 9.033673262697343e-05}
{"train_loss": 0.20388895273208618, "global_step": 55024, "epoch": 618, "lr": 9.033639004986285e-05}
{"train_loss": 0.22260665893554688, "global_step": 55025, "epoch": 618, "lr": 9.033604746732953e-05}
{"train_loss": 0.2179165780544281, "global_step": 55026, "epoch": 618, "lr": 9.033570487937352e-05}
{"train_loss": 0.2118426114320755, "global_step": 55027, "epoch": 618, "lr": 9.033536228599486e-05}
{"train_loss": 0.2694195508956909, "global_step": 55028, "epoch": 618, "lr": 9.033501968719362e-05}
{"train_loss": 0.29423120617866516, "global_step": 55029, "epoch": 618, "lr": 9.033467708296982e-05}
{"train_loss": 0.2426729053258896, "global_step": 55030, "epoch": 618, "lr": 9.03343344733235e-05}
{"train_loss": 0.162281796336174, "global_step": 55031, "epoch": 618, "lr": 9.033399185825473e-05}
{"train_loss": 0.26340481638908386, "global_step": 55032, "epoch": 618, "lr": 9.033364923776356e-05}
{"train_loss": 0.22848013043403625, "global_step": 55033, "epoch": 618, "lr": 9.033330661185e-05}
{"train_loss": 0.2889329493045807, "global_step": 55034, "epoch": 618, "lr": 9.033296398051413e-05}
{"train_loss": 0.28854212164878845, "global_step": 55035, "epoch": 618, "lr": 9.033262134375599e-05}
{"train_loss": 0.3030603528022766, "global_step": 55036, "epoch": 618, "lr": 9.033227870157561e-05}
{"train_loss": 0.25195109844207764, "global_step": 55037, "epoch": 618, "lr": 9.033193605397305e-05}
{"train_loss": 0.20609033107757568, "global_step": 55038, "epoch": 618, "lr": 9.033159340094836e-05}
{"train_loss": 0.15733899176120758, "global_step": 55039, "epoch": 618, "lr": 9.033125074250158e-05}
{"train_loss": 0.407839834690094, "global_step": 55040, "epoch": 618, "lr": 9.033090807863274e-05}
{"train_loss": 0.3663971424102783, "global_step": 55041, "epoch": 618, "lr": 9.03305654093419e-05}
{"train_loss": 0.3110460340976715, "global_step": 55042, "epoch": 618, "lr": 9.033022273462912e-05}
{"train_loss": 0.2535710036754608, "global_step": 55043, "epoch": 618, "lr": 9.032988005449443e-05}
{"train_loss": 0.2594468295574188, "global_step": 55044, "epoch": 618, "lr": 9.032953736893788e-05}
{"train_loss": 0.19298729300498962, "global_step": 55045, "epoch": 618, "lr": 9.032919467795951e-05}
{"train_loss": 0.2318602204322815, "global_step": 55046, "epoch": 618, "lr": 9.032885198155938e-05}
{"train_loss": 0.15817052125930786, "global_step": 55047, "epoch": 618, "lr": 9.032850927973752e-05}
{"train_loss": 0.2945125102996826, "global_step": 55048, "epoch": 618, "lr": 9.032816657249399e-05}
{"train_loss": 0.3208339512348175, "global_step": 55049, "epoch": 618, "lr": 9.03278238598288e-05}
{"train_loss": 0.3358062505722046, "global_step": 55050, "epoch": 618, "lr": 9.032748114174207e-05}
{"train_loss": 0.27456212043762207, "global_step": 55051, "epoch": 618, "lr": 9.032713841823377e-05}
{"train_loss": 0.30328357219696045, "global_step": 55052, "epoch": 618, "lr": 9.0326795689304e-05}
{"train_loss": 0.21712417900562286, "global_step": 55053, "epoch": 618, "lr": 9.032645295495276e-05}
{"train_loss": 0.28495511412620544, "global_step": 55054, "epoch": 618, "lr": 9.032611021518015e-05}
{"train_loss": 0.2532278895378113, "global_step": 55055, "epoch": 618, "lr": 9.032576746998616e-05}
{"train_loss": 0.29585397243499756, "global_step": 55056, "epoch": 618, "lr": 9.032542471937088e-05}
{"train_loss": 0.26158756017684937, "global_step": 55057, "epoch": 618, "lr": 9.032508196333434e-05}
{"train_loss": 0.31323203444480896, "global_step": 55058, "epoch": 618, "lr": 9.032473920187656e-05}
{"train_loss": 0.30451101064682007, "global_step": 55059, "epoch": 618, "lr": 9.032439643499762e-05}
{"train_loss": 0.2995431423187256, "global_step": 55060, "epoch": 618, "lr": 9.032405366269756e-05}
{"train_loss": 0.31625062227249146, "global_step": 55061, "epoch": 618, "lr": 9.032371088497644e-05}
{"train_loss": 0.18551188707351685, "global_step": 55062, "epoch": 618, "lr": 9.032336810183426e-05}
{"train_loss": 0.22076573967933655, "global_step": 55063, "epoch": 618, "lr": 9.03230253132711e-05}
{"train_loss": 0.3410128057003021, "global_step": 55064, "epoch": 618, "lr": 9.032268251928701e-05}
{"train_loss": 0.3044651746749878, "global_step": 55065, "epoch": 618, "lr": 9.032233971988203e-05}
{"train_loss": 0.3552536964416504, "global_step": 55066, "epoch": 618, "lr": 9.032199691505621e-05}
{"train_loss": 0.2853061556816101, "global_step": 55067, "epoch": 618, "lr": 9.032165410480957e-05}
{"train_loss": 0.31771320104599, "global_step": 55068, "epoch": 618, "lr": 9.032131128914217e-05}
{"train_loss": 0.268616259098053, "global_step": 55069, "epoch": 618, "lr": 9.032096846805407e-05}
{"train_loss": 0.280604749917984, "global_step": 55070, "epoch": 618, "lr": 9.032062564154531e-05}
{"train_loss": 0.40971386432647705, "global_step": 55071, "epoch": 618, "lr": 9.032028280961594e-05}
{"train_loss": 0.2890397310256958, "global_step": 55072, "epoch": 618, "lr": 9.0319939972266e-05}
{"train_loss": 0.21123473346233368, "global_step": 55073, "epoch": 618, "lr": 9.031959712949551e-05}
{"train_loss": 0.29133477807044983, "global_step": 55074, "epoch": 618, "lr": 9.031925428130456e-05}
{"train_loss": 0.3439828157424927, "global_step": 55075, "epoch": 618, "lr": 9.031891142769319e-05}
{"train_loss": 0.2874285876750946, "global_step": 55076, "epoch": 618, "lr": 9.031856856866142e-05}
{"train_loss": 0.21674661338329315, "global_step": 55077, "epoch": 618, "lr": 9.031822570420931e-05}
{"train_loss": 0.245383158326149, "global_step": 55078, "epoch": 618, "lr": 9.031788283433691e-05}
{"train_loss": 0.31181883811950684, "global_step": 55079, "epoch": 618, "lr": 9.031753995904426e-05}
{"train_loss": 0.2702915072441101, "global_step": 55080, "epoch": 618, "lr": 9.031719707833139e-05}
{"train_loss": 0.28993701934814453, "global_step": 55081, "epoch": 618, "lr": 9.03168541921984e-05}
{"train_loss": 0.3288680613040924, "global_step": 55082, "epoch": 618, "lr": 9.031651130064526e-05}
{"train_loss": 0.22402818500995636, "global_step": 55083, "epoch": 618, "lr": 9.031616840367209e-05}
{"train_loss": 0.27457016706466675, "global_step": 55084, "epoch": 618, "lr": 9.031582550127888e-05}
{"train_loss": 0.29143786430358887, "global_step": 55085, "epoch": 618, "lr": 9.031548259346572e-05}
{"train_loss": 0.29031533002853394, "global_step": 55086, "epoch": 618, "lr": 9.031513968023262e-05}
{"train_loss": 0.2069360911846161, "global_step": 55087, "epoch": 618, "lr": 9.031479676157965e-05}
{"train_loss": 0.2587181031703949, "global_step": 55088, "epoch": 618, "lr": 9.031445383750683e-05}
{"train_loss": 0.3240560293197632, "global_step": 55089, "epoch": 618, "lr": 9.031411090801425e-05}
{"train_loss": 0.26752113157443785, "global_step": 55090, "epoch": 618, "lr": 9.03137679731019e-05, "val_loss": 3.109632968902588}
{"train_loss": 0.2672424018383026, "global_step": 55091, "epoch": 619, "lr": 9.031342503276987e-05}
{"train_loss": 0.2692376375198364, "global_step": 55092, "epoch": 619, "lr": 9.03130820870182e-05}
{"train_loss": 0.24976769089698792, "global_step": 55093, "epoch": 619, "lr": 9.031273913584691e-05}
{"train_loss": 0.3926187753677368, "global_step": 55094, "epoch": 619, "lr": 9.031239617925608e-05}
{"train_loss": 0.2957065999507904, "global_step": 55095, "epoch": 619, "lr": 9.031205321724571e-05}
{"train_loss": 0.381618469953537, "global_step": 55096, "epoch": 619, "lr": 9.031171024981591e-05}
{"train_loss": 0.3182724714279175, "global_step": 55097, "epoch": 619, "lr": 9.031136727696668e-05}
{"train_loss": 0.29156044125556946, "global_step": 55098, "epoch": 619, "lr": 9.031102429869807e-05}
{"train_loss": 0.17847615480422974, "global_step": 55099, "epoch": 619, "lr": 9.031068131501015e-05}
{"train_loss": 0.3241628110408783, "global_step": 55100, "epoch": 619, "lr": 9.031033832590295e-05}
{"train_loss": 0.24019108712673187, "global_step": 55101, "epoch": 619, "lr": 9.03099953313765e-05}
{"train_loss": 0.27711448073387146, "global_step": 55102, "epoch": 619, "lr": 9.030965233143088e-05}
{"train_loss": 0.43449434638023376, "global_step": 55103, "epoch": 619, "lr": 9.03093093260661e-05}
{"train_loss": 0.3818514347076416, "global_step": 55104, "epoch": 619, "lr": 9.030896631528223e-05}
{"train_loss": 0.31075263023376465, "global_step": 55105, "epoch": 619, "lr": 9.03086232990793e-05}
{"train_loss": 0.23849816620349884, "global_step": 55106, "epoch": 619, "lr": 9.030828027745739e-05}
{"train_loss": 0.20307870209217072, "global_step": 55107, "epoch": 619, "lr": 9.03079372504165e-05}
{"train_loss": 0.2935296297073364, "global_step": 55108, "epoch": 619, "lr": 9.030759421795673e-05}
{"train_loss": 0.27688074111938477, "global_step": 55109, "epoch": 619, "lr": 9.030725118007808e-05}
{"train_loss": 0.30722030997276306, "global_step": 55110, "epoch": 619, "lr": 9.03069081367806e-05}
{"train_loss": 0.31977829337120056, "global_step": 55111, "epoch": 619, "lr": 9.030656508806437e-05}
{"train_loss": 0.22947171330451965, "global_step": 55112, "epoch": 619, "lr": 9.03062220339294e-05}
{"train_loss": 0.20708589255809784, "global_step": 55113, "epoch": 619, "lr": 9.030587897437574e-05}
{"train_loss": 0.23642568290233612, "global_step": 55114, "epoch": 619, "lr": 9.030553590940345e-05}
{"train_loss": 0.26420995593070984, "global_step": 55115, "epoch": 619, "lr": 9.030519283901258e-05}
{"train_loss": 0.21456147730350494, "global_step": 55116, "epoch": 619, "lr": 9.030484976320317e-05}
{"train_loss": 0.21742643415927887, "global_step": 55117, "epoch": 619, "lr": 9.030450668197525e-05}
{"train_loss": 0.27056342363357544, "global_step": 55118, "epoch": 619, "lr": 9.030416359532889e-05}
{"train_loss": 0.3006613850593567, "global_step": 55119, "epoch": 619, "lr": 9.030382050326413e-05}
{"train_loss": 0.23819531500339508, "global_step": 55120, "epoch": 619, "lr": 9.030347740578102e-05}
{"train_loss": 0.2509460151195526, "global_step": 55121, "epoch": 619, "lr": 9.030313430287959e-05}
{"train_loss": 0.24417494237422943, "global_step": 55122, "epoch": 619, "lr": 9.030279119455989e-05}
{"train_loss": 0.2507990300655365, "global_step": 55123, "epoch": 619, "lr": 9.030244808082198e-05}
{"train_loss": 0.2454664260149002, "global_step": 55124, "epoch": 619, "lr": 9.030210496166588e-05}
{"train_loss": 0.28035831451416016, "global_step": 55125, "epoch": 619, "lr": 9.030176183709166e-05}
{"train_loss": 0.26429226994514465, "global_step": 55126, "epoch": 619, "lr": 9.030141870709936e-05}
{"train_loss": 0.22031763195991516, "global_step": 55127, "epoch": 619, "lr": 9.030107557168904e-05}
{"train_loss": 0.26426199078559875, "global_step": 55128, "epoch": 619, "lr": 9.030073243086073e-05}
{"train_loss": 0.17868109047412872, "global_step": 55129, "epoch": 619, "lr": 9.030038928461446e-05}
{"train_loss": 0.28788629174232483, "global_step": 55130, "epoch": 619, "lr": 9.03000461329503e-05}
{"train_loss": 0.24345213174819946, "global_step": 55131, "epoch": 619, "lr": 9.02997029758683e-05}
{"train_loss": 0.2460605949163437, "global_step": 55132, "epoch": 619, "lr": 9.029935981336848e-05}
{"train_loss": 0.22902192175388336, "global_step": 55133, "epoch": 619, "lr": 9.029901664545091e-05}
{"train_loss": 0.17579546570777893, "global_step": 55134, "epoch": 619, "lr": 9.029867347211562e-05}
{"train_loss": 0.25304800271987915, "global_step": 55135, "epoch": 619, "lr": 9.029833029336268e-05}
{"train_loss": 0.23474256694316864, "global_step": 55136, "epoch": 619, "lr": 9.029798710919212e-05}
{"train_loss": 0.29092779755592346, "global_step": 55137, "epoch": 619, "lr": 9.029764391960398e-05}
{"train_loss": 0.22946719825267792, "global_step": 55138, "epoch": 619, "lr": 9.029730072459832e-05}
{"train_loss": 0.3158632814884186, "global_step": 55139, "epoch": 619, "lr": 9.029695752417517e-05}
{"train_loss": 0.24154214560985565, "global_step": 55140, "epoch": 619, "lr": 9.029661431833458e-05}
{"train_loss": 0.16898593306541443, "global_step": 55141, "epoch": 619, "lr": 9.029627110707661e-05}
{"train_loss": 0.29119378328323364, "global_step": 55142, "epoch": 619, "lr": 9.02959278904013e-05}
{"train_loss": 0.18066515028476715, "global_step": 55143, "epoch": 619, "lr": 9.02955846683087e-05}
{"train_loss": 0.21050287783145905, "global_step": 55144, "epoch": 619, "lr": 9.029524144079883e-05}
{"train_loss": 0.3014562427997589, "global_step": 55145, "epoch": 619, "lr": 9.029489820787175e-05}
{"train_loss": 0.21896469593048096, "global_step": 55146, "epoch": 619, "lr": 9.029455496952754e-05}
{"train_loss": 0.2056759148836136, "global_step": 55147, "epoch": 619, "lr": 9.029421172576622e-05}
{"train_loss": 0.2941986918449402, "global_step": 55148, "epoch": 619, "lr": 9.029386847658782e-05}
{"train_loss": 0.25177010893821716, "global_step": 55149, "epoch": 619, "lr": 9.029352522199239e-05}
{"train_loss": 0.210555762052536, "global_step": 55150, "epoch": 619, "lr": 9.029318196198e-05}
{"train_loss": 0.12915155291557312, "global_step": 55151, "epoch": 619, "lr": 9.029283869655068e-05}
{"train_loss": 0.3001660704612732, "global_step": 55152, "epoch": 619, "lr": 9.029249542570448e-05}
{"train_loss": 0.1635109931230545, "global_step": 55153, "epoch": 619, "lr": 9.029215214944144e-05}
{"train_loss": 0.423579603433609, "global_step": 55154, "epoch": 619, "lr": 9.029180886776162e-05}
{"train_loss": 0.21074597537517548, "global_step": 55155, "epoch": 619, "lr": 9.029146558066505e-05}
{"train_loss": 0.2886682450771332, "global_step": 55156, "epoch": 619, "lr": 9.029112228815179e-05}
{"train_loss": 0.19652627408504486, "global_step": 55157, "epoch": 619, "lr": 9.029077899022187e-05}
{"train_loss": 0.183814138174057, "global_step": 55158, "epoch": 619, "lr": 9.029043568687536e-05}
{"train_loss": 0.22908280789852142, "global_step": 55159, "epoch": 619, "lr": 9.029009237811229e-05}
{"train_loss": 0.23400786519050598, "global_step": 55160, "epoch": 619, "lr": 9.028974906393271e-05}
{"train_loss": 0.19722695648670197, "global_step": 55161, "epoch": 619, "lr": 9.028940574433664e-05}
{"train_loss": 0.2650980055332184, "global_step": 55162, "epoch": 619, "lr": 9.028906241932417e-05}
{"train_loss": 0.2386413961648941, "global_step": 55163, "epoch": 619, "lr": 9.028871908889534e-05}
{"train_loss": 0.21530728042125702, "global_step": 55164, "epoch": 619, "lr": 9.028837575305016e-05}
{"train_loss": 0.1836496889591217, "global_step": 55165, "epoch": 619, "lr": 9.028803241178871e-05}
{"train_loss": 0.24065947532653809, "global_step": 55166, "epoch": 619, "lr": 9.028768906511101e-05}
{"train_loss": 0.29158756136894226, "global_step": 55167, "epoch": 619, "lr": 9.028734571301713e-05}
{"train_loss": 0.1887531876564026, "global_step": 55168, "epoch": 619, "lr": 9.028700235550712e-05}
{"train_loss": 0.21509826183319092, "global_step": 55169, "epoch": 619, "lr": 9.0286658992581e-05}
{"train_loss": 0.2550395429134369, "global_step": 55170, "epoch": 619, "lr": 9.028631562423883e-05}
{"train_loss": 0.4065858721733093, "global_step": 55171, "epoch": 619, "lr": 9.028597225048065e-05}
{"train_loss": 0.2096419334411621, "global_step": 55172, "epoch": 619, "lr": 9.028562887130653e-05}
{"train_loss": 0.301123708486557, "global_step": 55173, "epoch": 619, "lr": 9.028528548671648e-05}
{"train_loss": 0.30072107911109924, "global_step": 55174, "epoch": 619, "lr": 9.028494209671058e-05}
{"train_loss": 0.20680469274520874, "global_step": 55175, "epoch": 619, "lr": 9.028459870128885e-05}
{"train_loss": 0.29051685333251953, "global_step": 55176, "epoch": 619, "lr": 9.028425530045136e-05}
{"train_loss": 0.24907685816287994, "global_step": 55177, "epoch": 619, "lr": 9.028391189419814e-05}
{"train_loss": 0.2181658148765564, "global_step": 55178, "epoch": 619, "lr": 9.028356848252921e-05}
{"train_loss": 0.2557852350259095, "global_step": 55179, "epoch": 619, "lr": 9.028322506544468e-05, "val_loss": 3.3576138019561768}
{"train_loss": 0.2255765199661255, "global_step": 55180, "epoch": 620, "lr": 9.028288164294454e-05}
{"train_loss": 0.24762387573719025, "global_step": 55181, "epoch": 620, "lr": 9.028253821502887e-05}
{"train_loss": 0.15659524500370026, "global_step": 55182, "epoch": 620, "lr": 9.02821947816977e-05}
{"train_loss": 0.3071568012237549, "global_step": 55183, "epoch": 620, "lr": 9.028185134295109e-05}
{"train_loss": 0.21840231120586395, "global_step": 55184, "epoch": 620, "lr": 9.028150789878906e-05}
{"train_loss": 0.25694960355758667, "global_step": 55185, "epoch": 620, "lr": 9.028116444921168e-05}
{"train_loss": 0.26697686314582825, "global_step": 55186, "epoch": 620, "lr": 9.028082099421899e-05}
{"train_loss": 0.36752164363861084, "global_step": 55187, "epoch": 620, "lr": 9.028047753381103e-05}
{"train_loss": 0.1872614324092865, "global_step": 55188, "epoch": 620, "lr": 9.028013406798785e-05}
{"train_loss": 0.308849036693573, "global_step": 55189, "epoch": 620, "lr": 9.027979059674949e-05}
{"train_loss": 0.16303659975528717, "global_step": 55190, "epoch": 620, "lr": 9.027944712009601e-05}
{"train_loss": 0.25346365571022034, "global_step": 55191, "epoch": 620, "lr": 9.027910363802746e-05}
{"train_loss": 0.2567169666290283, "global_step": 55192, "epoch": 620, "lr": 9.027876015054386e-05}
{"train_loss": 0.2532881200313568, "global_step": 55193, "epoch": 620, "lr": 9.02784166576453e-05}
{"train_loss": 0.23246750235557556, "global_step": 55194, "epoch": 620, "lr": 9.027807315933176e-05}
{"train_loss": 0.22604334354400635, "global_step": 55195, "epoch": 620, "lr": 9.027772965560334e-05}
{"train_loss": 0.23462238907814026, "global_step": 55196, "epoch": 620, "lr": 9.027738614646008e-05}
{"train_loss": 0.24781876802444458, "global_step": 55197, "epoch": 620, "lr": 9.0277042631902e-05}
{"train_loss": 0.20715923607349396, "global_step": 55198, "epoch": 620, "lr": 9.027669911192917e-05}
{"train_loss": 0.27671632170677185, "global_step": 55199, "epoch": 620, "lr": 9.027635558654163e-05}
{"train_loss": 0.2847122550010681, "global_step": 55200, "epoch": 620, "lr": 9.027601205573942e-05}
{"train_loss": 0.3083305060863495, "global_step": 55201, "epoch": 620, "lr": 9.027566851952259e-05}
{"train_loss": 0.2808498740196228, "global_step": 55202, "epoch": 620, "lr": 9.02753249778912e-05}
{"train_loss": 0.1824222356081009, "global_step": 55203, "epoch": 620, "lr": 9.027498143084527e-05}
{"train_loss": 0.2654695510864258, "global_step": 55204, "epoch": 620, "lr": 9.027463787838487e-05}
{"train_loss": 0.38405174016952515, "global_step": 55205, "epoch": 620, "lr": 9.027429432051002e-05}
{"train_loss": 0.24837549030780792, "global_step": 55206, "epoch": 620, "lr": 9.02739507572208e-05}
{"train_loss": 0.217665433883667, "global_step": 55207, "epoch": 620, "lr": 9.027360718851723e-05}
{"train_loss": 0.22161397337913513, "global_step": 55208, "epoch": 620, "lr": 9.027326361439936e-05}
{"train_loss": 0.2894953489303589, "global_step": 55209, "epoch": 620, "lr": 9.027292003486725e-05}
{"train_loss": 0.26873481273651123, "global_step": 55210, "epoch": 620, "lr": 9.027257644992094e-05}
{"train_loss": 0.2481856346130371, "global_step": 55211, "epoch": 620, "lr": 9.027223285956046e-05}
{"train_loss": 0.26771530508995056, "global_step": 55212, "epoch": 620, "lr": 9.027188926378588e-05}
{"train_loss": 0.23587381839752197, "global_step": 55213, "epoch": 620, "lr": 9.027154566259721e-05}
{"train_loss": 0.20359016954898834, "global_step": 55214, "epoch": 620, "lr": 9.027120205599455e-05}
{"train_loss": 0.20960895717144012, "global_step": 55215, "epoch": 620, "lr": 9.027085844397792e-05}
{"train_loss": 0.26905742287635803, "global_step": 55216, "epoch": 620, "lr": 9.027051482654733e-05}
{"train_loss": 0.2394171804189682, "global_step": 55217, "epoch": 620, "lr": 9.02701712037029e-05}
{"train_loss": 0.21323169767856598, "global_step": 55218, "epoch": 620, "lr": 9.026982757544461e-05}
{"train_loss": 0.2514878809452057, "global_step": 55219, "epoch": 620, "lr": 9.026948394177255e-05}
{"train_loss": 0.26847270131111145, "global_step": 55220, "epoch": 620, "lr": 9.026914030268674e-05}
{"train_loss": 0.2791338264942169, "global_step": 55221, "epoch": 620, "lr": 9.026879665818724e-05}
{"train_loss": 0.2768651843070984, "global_step": 55222, "epoch": 620, "lr": 9.026845300827408e-05}
{"train_loss": 0.35512635111808777, "global_step": 55223, "epoch": 620, "lr": 9.026810935294732e-05}
{"train_loss": 0.22142910957336426, "global_step": 55224, "epoch": 620, "lr": 9.026776569220701e-05}
{"train_loss": 0.21758393943309784, "global_step": 55225, "epoch": 620, "lr": 9.026742202605318e-05}
{"train_loss": 0.29496127367019653, "global_step": 55226, "epoch": 620, "lr": 9.026707835448589e-05}
{"train_loss": 0.23362255096435547, "global_step": 55227, "epoch": 620, "lr": 9.02667346775052e-05}
{"train_loss": 0.2601184844970703, "global_step": 55228, "epoch": 620, "lr": 9.026639099511112e-05}
{"train_loss": 0.24417376518249512, "global_step": 55229, "epoch": 620, "lr": 9.026604730730371e-05}
{"train_loss": 0.20790207386016846, "global_step": 55230, "epoch": 620, "lr": 9.026570361408303e-05}
{"train_loss": 0.26036936044692993, "global_step": 55231, "epoch": 620, "lr": 9.02653599154491e-05}
{"train_loss": 0.2807753384113312, "global_step": 55232, "epoch": 620, "lr": 9.026501621140202e-05}
{"train_loss": 0.3046499192714691, "global_step": 55233, "epoch": 620, "lr": 9.026467250194178e-05}
{"train_loss": 0.24484801292419434, "global_step": 55234, "epoch": 620, "lr": 9.026432878706843e-05}
{"train_loss": 0.20375728607177734, "global_step": 55235, "epoch": 620, "lr": 9.026398506678205e-05}
{"train_loss": 0.3297651708126068, "global_step": 55236, "epoch": 620, "lr": 9.026364134108267e-05}
{"train_loss": 0.3132403790950775, "global_step": 55237, "epoch": 620, "lr": 9.026329760997032e-05}
{"train_loss": 0.3324761688709259, "global_step": 55238, "epoch": 620, "lr": 9.026295387344507e-05}
{"train_loss": 0.28308409452438354, "global_step": 55239, "epoch": 620, "lr": 9.026261013150695e-05}
{"train_loss": 0.24456153810024261, "global_step": 55240, "epoch": 620, "lr": 9.026226638415602e-05}
{"train_loss": 0.161864772439003, "global_step": 55241, "epoch": 620, "lr": 9.026192263139232e-05}
{"train_loss": 0.23889338970184326, "global_step": 55242, "epoch": 620, "lr": 9.026157887321589e-05}
{"train_loss": 0.2754020392894745, "global_step": 55243, "epoch": 620, "lr": 9.026123510962679e-05}
{"train_loss": 0.4779062271118164, "global_step": 55244, "epoch": 620, "lr": 9.026089134062503e-05}
{"train_loss": 0.31850430369377136, "global_step": 55245, "epoch": 620, "lr": 9.026054756621072e-05}
{"train_loss": 0.3175610601902008, "global_step": 55246, "epoch": 620, "lr": 9.026020378638385e-05}
{"train_loss": 0.25963863730430603, "global_step": 55247, "epoch": 620, "lr": 9.025986000114448e-05}
{"train_loss": 0.24577641487121582, "global_step": 55248, "epoch": 620, "lr": 9.025951621049268e-05}
{"train_loss": 0.26356416940689087, "global_step": 55249, "epoch": 620, "lr": 9.025917241442847e-05}
{"train_loss": 0.24381962418556213, "global_step": 55250, "epoch": 620, "lr": 9.025882861295191e-05}
{"train_loss": 0.22857040166854858, "global_step": 55251, "epoch": 620, "lr": 9.025848480606304e-05}
{"train_loss": 0.3351389765739441, "global_step": 55252, "epoch": 620, "lr": 9.02581409937619e-05}
{"train_loss": 0.29442909359931946, "global_step": 55253, "epoch": 620, "lr": 9.025779717604855e-05}
{"train_loss": 0.2180115282535553, "global_step": 55254, "epoch": 620, "lr": 9.025745335292303e-05}
{"train_loss": 0.2955995500087738, "global_step": 55255, "epoch": 620, "lr": 9.025710952438538e-05}
{"train_loss": 0.20672446489334106, "global_step": 55256, "epoch": 620, "lr": 9.025676569043567e-05}
{"train_loss": 0.2416106015443802, "global_step": 55257, "epoch": 620, "lr": 9.02564218510739e-05}
{"train_loss": 0.2871013879776001, "global_step": 55258, "epoch": 620, "lr": 9.025607800630016e-05}
{"train_loss": 0.18455755710601807, "global_step": 55259, "epoch": 620, "lr": 9.025573415611449e-05}
{"train_loss": 0.3092139959335327, "global_step": 55260, "epoch": 620, "lr": 9.025539030051692e-05}
{"train_loss": 0.2906284034252167, "global_step": 55261, "epoch": 620, "lr": 9.025504643950751e-05}
{"train_loss": 0.28305208683013916, "global_step": 55262, "epoch": 620, "lr": 9.025470257308628e-05}
{"train_loss": 0.24942141771316528, "global_step": 55263, "epoch": 620, "lr": 9.025435870125332e-05}
{"train_loss": 0.30550307035446167, "global_step": 55264, "epoch": 620, "lr": 9.025401482400863e-05}
{"train_loss": 0.32884645462036133, "global_step": 55265, "epoch": 620, "lr": 9.025367094135229e-05}
{"train_loss": 0.25350648164749146, "global_step": 55266, "epoch": 620, "lr": 9.025332705328433e-05}
{"train_loss": 0.24540945887565613, "global_step": 55267, "epoch": 620, "lr": 9.025298315980479e-05}
{"train_loss": 0.2614581017346864, "global_step": 55268, "epoch": 620, "lr": 9.025263926091375e-05, "val_loss": 3.236055850982666, "train_action_mse_error": 24.29067611694336}
{"train_loss": 0.24744565784931183, "global_step": 55269, "epoch": 621, "lr": 9.025229535661123e-05}
{"train_loss": 0.25632888078689575, "global_step": 55270, "epoch": 621, "lr": 9.025195144689726e-05}
{"train_loss": 0.24909241497516632, "global_step": 55271, "epoch": 621, "lr": 9.025160753177193e-05}
{"train_loss": 0.3216072618961334, "global_step": 55272, "epoch": 621, "lr": 9.025126361123524e-05}
{"train_loss": 0.24878934025764465, "global_step": 55273, "epoch": 621, "lr": 9.025091968528727e-05}
{"train_loss": 0.27069002389907837, "global_step": 55274, "epoch": 621, "lr": 9.025057575392805e-05}
{"train_loss": 0.1920502632856369, "global_step": 55275, "epoch": 621, "lr": 9.025023181715763e-05}
{"train_loss": 0.3275226354598999, "global_step": 55276, "epoch": 621, "lr": 9.024988787497608e-05}
{"train_loss": 0.24267229437828064, "global_step": 55277, "epoch": 621, "lr": 9.024954392738341e-05}
{"train_loss": 0.21146278083324432, "global_step": 55278, "epoch": 621, "lr": 9.024919997437966e-05}
{"train_loss": 0.21649815142154694, "global_step": 55279, "epoch": 621, "lr": 9.02488560159649e-05}
{"train_loss": 0.22878499329090118, "global_step": 55280, "epoch": 621, "lr": 9.02485120521392e-05}
{"train_loss": 0.22583314776420593, "global_step": 55281, "epoch": 621, "lr": 9.024816808290256e-05}
{"train_loss": 0.2044423669576645, "global_step": 55282, "epoch": 621, "lr": 9.024782410825503e-05}
{"train_loss": 0.2749726176261902, "global_step": 55283, "epoch": 621, "lr": 9.024748012819669e-05}
{"train_loss": 0.23634253442287445, "global_step": 55284, "epoch": 621, "lr": 9.024713614272756e-05}
{"train_loss": 0.3526345491409302, "global_step": 55285, "epoch": 621, "lr": 9.024679215184769e-05}
{"train_loss": 0.2538328170776367, "global_step": 55286, "epoch": 621, "lr": 9.024644815555714e-05}
{"train_loss": 0.2451837658882141, "global_step": 55287, "epoch": 621, "lr": 9.024610415385593e-05}
{"train_loss": 0.2586522698402405, "global_step": 55288, "epoch": 621, "lr": 9.024576014674413e-05}
{"train_loss": 0.28610002994537354, "global_step": 55289, "epoch": 621, "lr": 9.02454161342218e-05}
{"train_loss": 0.2828173339366913, "global_step": 55290, "epoch": 621, "lr": 9.024507211628893e-05}
{"train_loss": 0.21768032014369965, "global_step": 55291, "epoch": 621, "lr": 9.024472809294562e-05}
{"train_loss": 0.3146592676639557, "global_step": 55292, "epoch": 621, "lr": 9.024438406419189e-05}
{"train_loss": 0.2864567041397095, "global_step": 55293, "epoch": 621, "lr": 9.02440400300278e-05}
{"train_loss": 0.28099679946899414, "global_step": 55294, "epoch": 621, "lr": 9.024369599045336e-05}
{"train_loss": 0.24983824789524078, "global_step": 55295, "epoch": 621, "lr": 9.024335194546868e-05}
{"train_loss": 0.17166215181350708, "global_step": 55296, "epoch": 621, "lr": 9.024300789507376e-05}
{"train_loss": 0.334255188703537, "global_step": 55297, "epoch": 621, "lr": 9.024266383926865e-05}
{"train_loss": 0.2554922103881836, "global_step": 55298, "epoch": 621, "lr": 9.024231977805342e-05}
{"train_loss": 0.28604957461357117, "global_step": 55299, "epoch": 621, "lr": 9.02419757114281e-05}
{"train_loss": 0.22621068358421326, "global_step": 55300, "epoch": 621, "lr": 9.024163163939272e-05}
{"train_loss": 0.35177770256996155, "global_step": 55301, "epoch": 621, "lr": 9.024128756194736e-05}
{"train_loss": 0.2504490613937378, "global_step": 55302, "epoch": 621, "lr": 9.024094347909204e-05}
{"train_loss": 0.255156934261322, "global_step": 55303, "epoch": 621, "lr": 9.024059939082682e-05}
{"train_loss": 0.2976306080818176, "global_step": 55304, "epoch": 621, "lr": 9.024025529715175e-05}
{"train_loss": 0.18561780452728271, "global_step": 55305, "epoch": 621, "lr": 9.023991119806686e-05}
{"train_loss": 0.2450161576271057, "global_step": 55306, "epoch": 621, "lr": 9.023956709357219e-05}
{"train_loss": 0.27928635478019714, "global_step": 55307, "epoch": 621, "lr": 9.02392229836678e-05}
{"train_loss": 0.2914183735847473, "global_step": 55308, "epoch": 621, "lr": 9.023887886835376e-05}
{"train_loss": 0.18657755851745605, "global_step": 55309, "epoch": 621, "lr": 9.023853474763009e-05}
{"train_loss": 0.26344573497772217, "global_step": 55310, "epoch": 621, "lr": 9.023819062149683e-05}
{"train_loss": 0.25631532073020935, "global_step": 55311, "epoch": 621, "lr": 9.023784648995404e-05}
{"train_loss": 0.2734881341457367, "global_step": 55312, "epoch": 621, "lr": 9.023750235300176e-05}
{"train_loss": 0.3545737862586975, "global_step": 55313, "epoch": 621, "lr": 9.023715821064003e-05}
{"train_loss": 0.3015340268611908, "global_step": 55314, "epoch": 621, "lr": 9.023681406286892e-05}
{"train_loss": 0.19784560799598694, "global_step": 55315, "epoch": 621, "lr": 9.023646990968845e-05}
{"train_loss": 0.29648327827453613, "global_step": 55316, "epoch": 621, "lr": 9.023612575109868e-05}
{"train_loss": 0.2530769407749176, "global_step": 55317, "epoch": 621, "lr": 9.023578158709965e-05}
{"train_loss": 0.22524306178092957, "global_step": 55318, "epoch": 621, "lr": 9.023543741769143e-05}
{"train_loss": 0.387085497379303, "global_step": 55319, "epoch": 621, "lr": 9.023509324287403e-05}
{"train_loss": 0.21833641827106476, "global_step": 55320, "epoch": 621, "lr": 9.023474906264751e-05}
{"train_loss": 0.20664642751216888, "global_step": 55321, "epoch": 621, "lr": 9.023440487701191e-05}
{"train_loss": 0.3263731002807617, "global_step": 55322, "epoch": 621, "lr": 9.023406068596731e-05}
{"train_loss": 0.27613094449043274, "global_step": 55323, "epoch": 621, "lr": 9.02337164895137e-05}
{"train_loss": 0.21699826419353485, "global_step": 55324, "epoch": 621, "lr": 9.023337228765118e-05}
{"train_loss": 0.29558947682380676, "global_step": 55325, "epoch": 621, "lr": 9.023302808037977e-05}
{"train_loss": 0.2810356914997101, "global_step": 55326, "epoch": 621, "lr": 9.02326838676995e-05}
{"train_loss": 0.2645376920700073, "global_step": 55327, "epoch": 621, "lr": 9.023233964961047e-05}
{"train_loss": 0.3001171946525574, "global_step": 55328, "epoch": 621, "lr": 9.023199542611267e-05}
{"train_loss": 0.274116188287735, "global_step": 55329, "epoch": 621, "lr": 9.023165119720618e-05}
{"train_loss": 0.3233034312725067, "global_step": 55330, "epoch": 621, "lr": 9.023130696289102e-05}
{"train_loss": 0.2852884531021118, "global_step": 55331, "epoch": 621, "lr": 9.023096272316727e-05}
{"train_loss": 0.2557768225669861, "global_step": 55332, "epoch": 621, "lr": 9.023061847803494e-05}
{"train_loss": 0.2300310879945755, "global_step": 55333, "epoch": 621, "lr": 9.023027422749409e-05}
{"train_loss": 0.2544526755809784, "global_step": 55334, "epoch": 621, "lr": 9.022992997154478e-05}
{"train_loss": 0.3350284993648529, "global_step": 55335, "epoch": 621, "lr": 9.022958571018705e-05}
{"train_loss": 0.2766186594963074, "global_step": 55336, "epoch": 621, "lr": 9.022924144342094e-05}
{"train_loss": 0.24719999730587006, "global_step": 55337, "epoch": 621, "lr": 9.022889717124649e-05}
{"train_loss": 0.33041805028915405, "global_step": 55338, "epoch": 621, "lr": 9.022855289366377e-05}
{"train_loss": 0.26245617866516113, "global_step": 55339, "epoch": 621, "lr": 9.022820861067281e-05}
{"train_loss": 0.2163587212562561, "global_step": 55340, "epoch": 621, "lr": 9.022786432227365e-05}
{"train_loss": 0.21224012970924377, "global_step": 55341, "epoch": 621, "lr": 9.022752002846635e-05}
{"train_loss": 0.32555297017097473, "global_step": 55342, "epoch": 621, "lr": 9.022717572925095e-05}
{"train_loss": 0.29783058166503906, "global_step": 55343, "epoch": 621, "lr": 9.02268314246275e-05}
{"train_loss": 0.23475812375545502, "global_step": 55344, "epoch": 621, "lr": 9.022648711459601e-05}
{"train_loss": 0.2805003821849823, "global_step": 55345, "epoch": 621, "lr": 9.02261427991566e-05}
{"train_loss": 0.26930859684944153, "global_step": 55346, "epoch": 621, "lr": 9.022579847830925e-05}
{"train_loss": 0.2823185622692108, "global_step": 55347, "epoch": 621, "lr": 9.022545415205404e-05}
{"train_loss": 0.1708412617444992, "global_step": 55348, "epoch": 621, "lr": 9.022510982039102e-05}
{"train_loss": 0.35107284784317017, "global_step": 55349, "epoch": 621, "lr": 9.022476548332021e-05}
{"train_loss": 0.30842798948287964, "global_step": 55350, "epoch": 621, "lr": 9.022442114084168e-05}
{"train_loss": 0.1925017237663269, "global_step": 55351, "epoch": 621, "lr": 9.022407679295545e-05}
{"train_loss": 0.20200468599796295, "global_step": 55352, "epoch": 621, "lr": 9.02237324396616e-05}
{"train_loss": 0.25216686725616455, "global_step": 55353, "epoch": 621, "lr": 9.022338808096016e-05}
{"train_loss": 0.21884335577487946, "global_step": 55354, "epoch": 621, "lr": 9.022304371685115e-05}
{"train_loss": 0.2557939887046814, "global_step": 55355, "epoch": 621, "lr": 9.022269934733467e-05}
{"train_loss": 0.2713943421840668, "global_step": 55356, "epoch": 621, "lr": 9.022235497241074e-05}
{"train_loss": 0.2642089766397905, "global_step": 55357, "epoch": 621, "lr": 9.022201059207938e-05, "val_loss": 3.114452838897705}
{"train_loss": 0.22203217446804047, "global_step": 55358, "epoch": 622, "lr": 9.022166620634068e-05}
{"train_loss": 0.21593576669692993, "global_step": 55359, "epoch": 622, "lr": 9.022132181519467e-05}
{"train_loss": 0.22575798630714417, "global_step": 55360, "epoch": 622, "lr": 9.022097741864137e-05}
{"train_loss": 0.24371103942394257, "global_step": 55361, "epoch": 622, "lr": 9.022063301668088e-05}
{"train_loss": 0.14378443360328674, "global_step": 55362, "epoch": 622, "lr": 9.022028860931319e-05}
{"train_loss": 0.2802039384841919, "global_step": 55363, "epoch": 622, "lr": 9.021994419653839e-05}
{"train_loss": 0.2517976462841034, "global_step": 55364, "epoch": 622, "lr": 9.02195997783565e-05}
{"train_loss": 0.19067294895648956, "global_step": 55365, "epoch": 622, "lr": 9.021925535476759e-05}
{"train_loss": 0.24301527440547943, "global_step": 55366, "epoch": 622, "lr": 9.021891092577169e-05}
{"train_loss": 0.30544766783714294, "global_step": 55367, "epoch": 622, "lr": 9.021856649136881e-05}
{"train_loss": 0.18470774590969086, "global_step": 55368, "epoch": 622, "lr": 9.021822205155907e-05}
{"train_loss": 0.15939052402973175, "global_step": 55369, "epoch": 622, "lr": 9.021787760634248e-05}
{"train_loss": 0.17746330797672272, "global_step": 55370, "epoch": 622, "lr": 9.021753315571907e-05}
{"train_loss": 0.20158031582832336, "global_step": 55371, "epoch": 622, "lr": 9.021718869968891e-05}
{"train_loss": 0.16059280931949615, "global_step": 55372, "epoch": 622, "lr": 9.021684423825204e-05}
{"train_loss": 0.22149835526943207, "global_step": 55373, "epoch": 622, "lr": 9.021649977140851e-05}
{"train_loss": 0.31993985176086426, "global_step": 55374, "epoch": 622, "lr": 9.021615529915836e-05}
{"train_loss": 0.21243984997272491, "global_step": 55375, "epoch": 622, "lr": 9.021581082150163e-05}
{"train_loss": 0.27310454845428467, "global_step": 55376, "epoch": 622, "lr": 9.021546633843838e-05}
{"train_loss": 0.1580343097448349, "global_step": 55377, "epoch": 622, "lr": 9.021512184996865e-05}
{"train_loss": 0.28334152698516846, "global_step": 55378, "epoch": 622, "lr": 9.02147773560925e-05}
{"train_loss": 0.2893678545951843, "global_step": 55379, "epoch": 622, "lr": 9.021443285680994e-05}
{"train_loss": 0.19766628742218018, "global_step": 55380, "epoch": 622, "lr": 9.021408835212106e-05}
{"train_loss": 0.1465502828359604, "global_step": 55381, "epoch": 622, "lr": 9.021374384202587e-05}
{"train_loss": 0.1931557059288025, "global_step": 55382, "epoch": 622, "lr": 9.021339932652444e-05}
{"train_loss": 0.23353566229343414, "global_step": 55383, "epoch": 622, "lr": 9.02130548056168e-05}
{"train_loss": 0.16913171112537384, "global_step": 55384, "epoch": 622, "lr": 9.021271027930302e-05}
{"train_loss": 0.2678624093532562, "global_step": 55385, "epoch": 622, "lr": 9.021236574758312e-05}
{"train_loss": 0.252003937959671, "global_step": 55386, "epoch": 622, "lr": 9.021202121045715e-05}
{"train_loss": 0.24600879848003387, "global_step": 55387, "epoch": 622, "lr": 9.021167666792518e-05}
{"train_loss": 0.26493164896965027, "global_step": 55388, "epoch": 622, "lr": 9.021133211998724e-05}
{"train_loss": 0.16629572212696075, "global_step": 55389, "epoch": 622, "lr": 9.021098756664336e-05}
{"train_loss": 0.271091490983963, "global_step": 55390, "epoch": 622, "lr": 9.021064300789363e-05}
{"train_loss": 0.21687456965446472, "global_step": 55391, "epoch": 622, "lr": 9.021029844373804e-05}
{"train_loss": 0.2972266674041748, "global_step": 55392, "epoch": 622, "lr": 9.020995387417668e-05}
{"train_loss": 0.3216090798377991, "global_step": 55393, "epoch": 622, "lr": 9.020960929920959e-05}
{"train_loss": 0.3032552897930145, "global_step": 55394, "epoch": 622, "lr": 9.020926471883679e-05}
{"train_loss": 0.17105108499526978, "global_step": 55395, "epoch": 622, "lr": 9.020892013305836e-05}
{"train_loss": 0.2658572494983673, "global_step": 55396, "epoch": 622, "lr": 9.020857554187432e-05}
{"train_loss": 0.30216214060783386, "global_step": 55397, "epoch": 622, "lr": 9.020823094528473e-05}
{"train_loss": 0.2766304314136505, "global_step": 55398, "epoch": 622, "lr": 9.020788634328964e-05}
{"train_loss": 0.2635539770126343, "global_step": 55399, "epoch": 622, "lr": 9.020754173588908e-05}
{"train_loss": 0.20868165791034698, "global_step": 55400, "epoch": 622, "lr": 9.020719712308311e-05}
{"train_loss": 0.3266581594944, "global_step": 55401, "epoch": 622, "lr": 9.020685250487177e-05}
{"train_loss": 0.37475356459617615, "global_step": 55402, "epoch": 622, "lr": 9.020650788125511e-05}
{"train_loss": 0.2991349995136261, "global_step": 55403, "epoch": 622, "lr": 9.02061632522332e-05}
{"train_loss": 0.21486292779445648, "global_step": 55404, "epoch": 622, "lr": 9.020581861780601e-05}
{"train_loss": 0.2063221037387848, "global_step": 55405, "epoch": 622, "lr": 9.020547397797367e-05}
{"train_loss": 0.19038359820842743, "global_step": 55406, "epoch": 622, "lr": 9.020512933273619e-05}
{"train_loss": 0.2251281440258026, "global_step": 55407, "epoch": 622, "lr": 9.020478468209363e-05}
{"train_loss": 0.30661946535110474, "global_step": 55408, "epoch": 622, "lr": 9.020444002604602e-05}
{"train_loss": 0.22667983174324036, "global_step": 55409, "epoch": 622, "lr": 9.02040953645934e-05}
{"train_loss": 0.26657843589782715, "global_step": 55410, "epoch": 622, "lr": 9.020375069773583e-05}
{"train_loss": 0.3282758593559265, "global_step": 55411, "epoch": 622, "lr": 9.020340602547336e-05}
{"train_loss": 0.2011181116104126, "global_step": 55412, "epoch": 622, "lr": 9.020306134780606e-05}
{"train_loss": 0.29097363352775574, "global_step": 55413, "epoch": 622, "lr": 9.020271666473392e-05}
{"train_loss": 0.2884971797466278, "global_step": 55414, "epoch": 622, "lr": 9.020237197625702e-05}
{"train_loss": 0.1761966496706009, "global_step": 55415, "epoch": 622, "lr": 9.02020272823754e-05}
{"train_loss": 0.31240713596343994, "global_step": 55416, "epoch": 622, "lr": 9.02016825830891e-05}
{"train_loss": 0.20532086491584778, "global_step": 55417, "epoch": 622, "lr": 9.020133787839819e-05}
{"train_loss": 0.1765914261341095, "global_step": 55418, "epoch": 622, "lr": 9.02009931683027e-05}
{"train_loss": 0.4278261661529541, "global_step": 55419, "epoch": 622, "lr": 9.020064845280267e-05}
{"train_loss": 0.2985784411430359, "global_step": 55420, "epoch": 622, "lr": 9.020030373189814e-05}
{"train_loss": 0.21251092851161957, "global_step": 55421, "epoch": 622, "lr": 9.019995900558918e-05}
{"train_loss": 0.2368844747543335, "global_step": 55422, "epoch": 622, "lr": 9.019961427387583e-05}
{"train_loss": 0.19675174355506897, "global_step": 55423, "epoch": 622, "lr": 9.019926953675813e-05}
{"train_loss": 0.2520350217819214, "global_step": 55424, "epoch": 622, "lr": 9.019892479423612e-05}
{"train_loss": 0.11144457012414932, "global_step": 55425, "epoch": 622, "lr": 9.019858004630987e-05}
{"train_loss": 0.3440554738044739, "global_step": 55426, "epoch": 622, "lr": 9.019823529297939e-05}
{"train_loss": 0.23877853155136108, "global_step": 55427, "epoch": 622, "lr": 9.019789053424477e-05}
{"train_loss": 0.25532132387161255, "global_step": 55428, "epoch": 622, "lr": 9.019754577010602e-05}
{"train_loss": 0.2055092751979828, "global_step": 55429, "epoch": 622, "lr": 9.01972010005632e-05}
{"train_loss": 0.26902955770492554, "global_step": 55430, "epoch": 622, "lr": 9.019685622561637e-05}
{"train_loss": 0.3590104579925537, "global_step": 55431, "epoch": 622, "lr": 9.019651144526555e-05}
{"train_loss": 0.18845586478710175, "global_step": 55432, "epoch": 622, "lr": 9.019616665951079e-05}
{"train_loss": 0.23152492940425873, "global_step": 55433, "epoch": 622, "lr": 9.019582186835216e-05}
{"train_loss": 0.24409396946430206, "global_step": 55434, "epoch": 622, "lr": 9.019547707178969e-05}
{"train_loss": 0.17932242155075073, "global_step": 55435, "epoch": 622, "lr": 9.019513226982343e-05}
{"train_loss": 0.31576475501060486, "global_step": 55436, "epoch": 622, "lr": 9.019478746245341e-05}
{"train_loss": 0.28522369265556335, "global_step": 55437, "epoch": 622, "lr": 9.019444264967971e-05}
{"train_loss": 0.2587625980377197, "global_step": 55438, "epoch": 622, "lr": 9.019409783150235e-05}
{"train_loss": 0.3129732012748718, "global_step": 55439, "epoch": 622, "lr": 9.019375300792137e-05}
{"train_loss": 0.20686455070972443, "global_step": 55440, "epoch": 622, "lr": 9.019340817893684e-05}
{"train_loss": 0.29954805970191956, "global_step": 55441, "epoch": 622, "lr": 9.01930633445488e-05}
{"train_loss": 0.23418669402599335, "global_step": 55442, "epoch": 622, "lr": 9.019271850475728e-05}
{"train_loss": 0.3404359221458435, "global_step": 55443, "epoch": 622, "lr": 9.019237365956235e-05}
{"train_loss": 0.2552056908607483, "global_step": 55444, "epoch": 622, "lr": 9.019202880896405e-05}
{"train_loss": 0.32230690121650696, "global_step": 55445, "epoch": 622, "lr": 9.019168395296243e-05}
{"train_loss": 0.24644399852900023, "global_step": 55446, "epoch": 622, "lr": 9.01913390915575e-05, "val_loss": 3.2464988231658936}
{"train_loss": 0.22330768406391144, "global_step": 55447, "epoch": 623, "lr": 9.019099422474935e-05}
{"train_loss": 0.18868115544319153, "global_step": 55448, "epoch": 623, "lr": 9.019064935253801e-05}
{"train_loss": 0.16317246854305267, "global_step": 55449, "epoch": 623, "lr": 9.019030447492352e-05}
{"train_loss": 0.26946789026260376, "global_step": 55450, "epoch": 623, "lr": 9.018995959190595e-05}
{"train_loss": 0.24638976156711578, "global_step": 55451, "epoch": 623, "lr": 9.01896147034853e-05}
{"train_loss": 0.37692907452583313, "global_step": 55452, "epoch": 623, "lr": 9.018926980966167e-05}
{"train_loss": 0.21854442358016968, "global_step": 55453, "epoch": 623, "lr": 9.018892491043508e-05}
{"train_loss": 0.30678728222846985, "global_step": 55454, "epoch": 623, "lr": 9.018858000580558e-05}
{"train_loss": 0.3081356883049011, "global_step": 55455, "epoch": 623, "lr": 9.018823509577321e-05}
{"train_loss": 0.24956651031970978, "global_step": 55456, "epoch": 623, "lr": 9.018789018033802e-05}
{"train_loss": 0.20885565876960754, "global_step": 55457, "epoch": 623, "lr": 9.018754525950006e-05}
{"train_loss": 0.16577836871147156, "global_step": 55458, "epoch": 623, "lr": 9.018720033325938e-05}
{"train_loss": 0.2521538734436035, "global_step": 55459, "epoch": 623, "lr": 9.018685540161602e-05}
{"train_loss": 0.1947556585073471, "global_step": 55460, "epoch": 623, "lr": 9.018651046457002e-05}
{"train_loss": 0.25610432028770447, "global_step": 55461, "epoch": 623, "lr": 9.018616552212142e-05}
{"train_loss": 0.24973446130752563, "global_step": 55462, "epoch": 623, "lr": 9.018582057427031e-05}
{"train_loss": 0.30739477276802063, "global_step": 55463, "epoch": 623, "lr": 9.01854756210167e-05}
{"train_loss": 0.2328353375196457, "global_step": 55464, "epoch": 623, "lr": 9.018513066236063e-05}
{"train_loss": 0.1914958655834198, "global_step": 55465, "epoch": 623, "lr": 9.018478569830216e-05}
{"train_loss": 0.20875222980976105, "global_step": 55466, "epoch": 623, "lr": 9.018444072884134e-05}
{"train_loss": 0.23965026438236237, "global_step": 55467, "epoch": 623, "lr": 9.018409575397821e-05}
{"train_loss": 0.2325156331062317, "global_step": 55468, "epoch": 623, "lr": 9.018375077371281e-05}
{"train_loss": 0.230958491563797, "global_step": 55469, "epoch": 623, "lr": 9.01834057880452e-05}
{"train_loss": 0.3775533139705658, "global_step": 55470, "epoch": 623, "lr": 9.018306079697543e-05}
{"train_loss": 0.18435876071453094, "global_step": 55471, "epoch": 623, "lr": 9.018271580050354e-05}
{"train_loss": 0.19611644744873047, "global_step": 55472, "epoch": 623, "lr": 9.018237079862956e-05}
{"train_loss": 0.35621577501296997, "global_step": 55473, "epoch": 623, "lr": 9.018202579135356e-05}
{"train_loss": 0.28279396891593933, "global_step": 55474, "epoch": 623, "lr": 9.018168077867558e-05}
{"train_loss": 0.17939560115337372, "global_step": 55475, "epoch": 623, "lr": 9.018133576059565e-05}
{"train_loss": 0.2850237488746643, "global_step": 55476, "epoch": 623, "lr": 9.018099073711383e-05}
{"train_loss": 0.20604117214679718, "global_step": 55477, "epoch": 623, "lr": 9.018064570823018e-05}
{"train_loss": 0.23194707930088043, "global_step": 55478, "epoch": 623, "lr": 9.01803006739447e-05}
{"train_loss": 0.1966826617717743, "global_step": 55479, "epoch": 623, "lr": 9.01799556342575e-05}
{"train_loss": 0.2745739221572876, "global_step": 55480, "epoch": 623, "lr": 9.017961058916859e-05}
{"train_loss": 0.25837019085884094, "global_step": 55481, "epoch": 623, "lr": 9.017926553867802e-05}
{"train_loss": 0.2274351567029953, "global_step": 55482, "epoch": 623, "lr": 9.017892048278582e-05}
{"train_loss": 0.24231116473674774, "global_step": 55483, "epoch": 623, "lr": 9.017857542149207e-05}
{"train_loss": 0.17601148784160614, "global_step": 55484, "epoch": 623, "lr": 9.01782303547968e-05}
{"train_loss": 0.24849960207939148, "global_step": 55485, "epoch": 623, "lr": 9.017788528270005e-05}
{"train_loss": 0.2310553640127182, "global_step": 55486, "epoch": 623, "lr": 9.017754020520189e-05}
{"train_loss": 0.2053965926170349, "global_step": 55487, "epoch": 623, "lr": 9.017719512230232e-05}
{"train_loss": 0.24158187210559845, "global_step": 55488, "epoch": 623, "lr": 9.017685003400142e-05}
{"train_loss": 0.2711532711982727, "global_step": 55489, "epoch": 623, "lr": 9.017650494029926e-05}
{"train_loss": 0.24054163694381714, "global_step": 55490, "epoch": 623, "lr": 9.017615984119583e-05}
{"train_loss": 0.21245889365673065, "global_step": 55491, "epoch": 623, "lr": 9.017581473669124e-05}
{"train_loss": 0.25946876406669617, "global_step": 55492, "epoch": 623, "lr": 9.017546962678547e-05}
{"train_loss": 0.2507951259613037, "global_step": 55493, "epoch": 623, "lr": 9.01751245114786e-05}
{"train_loss": 0.24102208018302917, "global_step": 55494, "epoch": 623, "lr": 9.017477939077068e-05}
{"train_loss": 0.2463325709104538, "global_step": 55495, "epoch": 623, "lr": 9.017443426466175e-05}
{"train_loss": 0.25070565938949585, "global_step": 55496, "epoch": 623, "lr": 9.017408913315186e-05}
{"train_loss": 0.25034382939338684, "global_step": 55497, "epoch": 623, "lr": 9.017374399624106e-05}
{"train_loss": 0.27612051367759705, "global_step": 55498, "epoch": 623, "lr": 9.017339885392937e-05}
{"train_loss": 0.3291068971157074, "global_step": 55499, "epoch": 623, "lr": 9.017305370621687e-05}
{"train_loss": 0.23966512084007263, "global_step": 55500, "epoch": 623, "lr": 9.01727085531036e-05}
{"train_loss": 0.3085961937904358, "global_step": 55501, "epoch": 623, "lr": 9.017236339458958e-05}
{"train_loss": 0.22028450667858124, "global_step": 55502, "epoch": 623, "lr": 9.01720182306749e-05}
{"train_loss": 0.2089853733778, "global_step": 55503, "epoch": 623, "lr": 9.017167306135956e-05}
{"train_loss": 0.21275658905506134, "global_step": 55504, "epoch": 623, "lr": 9.017132788664363e-05}
{"train_loss": 0.1265036165714264, "global_step": 55505, "epoch": 623, "lr": 9.017098270652715e-05}
{"train_loss": 0.1743052452802658, "global_step": 55506, "epoch": 623, "lr": 9.017063752101018e-05}
{"train_loss": 0.18134494125843048, "global_step": 55507, "epoch": 623, "lr": 9.017029233009276e-05}
{"train_loss": 0.17225292325019836, "global_step": 55508, "epoch": 623, "lr": 9.016994713377493e-05}
{"train_loss": 0.21143066883087158, "global_step": 55509, "epoch": 623, "lr": 9.016960193205674e-05}
{"train_loss": 0.21022726595401764, "global_step": 55510, "epoch": 623, "lr": 9.016925672493826e-05}
{"train_loss": 0.2565862536430359, "global_step": 55511, "epoch": 623, "lr": 9.016891151241948e-05}
{"train_loss": 0.2044883817434311, "global_step": 55512, "epoch": 623, "lr": 9.016856629450049e-05}
{"train_loss": 0.23988483846187592, "global_step": 55513, "epoch": 623, "lr": 9.016822107118132e-05}
{"train_loss": 0.24340184032917023, "global_step": 55514, "epoch": 623, "lr": 9.016787584246204e-05}
{"train_loss": 0.2525479793548584, "global_step": 55515, "epoch": 623, "lr": 9.016753060834267e-05}
{"train_loss": 0.2894572615623474, "global_step": 55516, "epoch": 623, "lr": 9.016718536882324e-05}
{"train_loss": 0.20818868279457092, "global_step": 55517, "epoch": 623, "lr": 9.016684012390387e-05}
{"train_loss": 0.2577936053276062, "global_step": 55518, "epoch": 623, "lr": 9.016649487358452e-05}
{"train_loss": 0.24546602368354797, "global_step": 55519, "epoch": 623, "lr": 9.016614961786529e-05}
{"train_loss": 0.2280207723379135, "global_step": 55520, "epoch": 623, "lr": 9.016580435674621e-05}
{"train_loss": 0.2111223042011261, "global_step": 55521, "epoch": 623, "lr": 9.016545909022732e-05}
{"train_loss": 0.25035780668258667, "global_step": 55522, "epoch": 623, "lr": 9.016511381830867e-05}
{"train_loss": 0.2506055235862732, "global_step": 55523, "epoch": 623, "lr": 9.016476854099031e-05}
{"train_loss": 0.26919227838516235, "global_step": 55524, "epoch": 623, "lr": 9.01644232582723e-05}
{"train_loss": 0.20358847081661224, "global_step": 55525, "epoch": 623, "lr": 9.016407797015466e-05}
{"train_loss": 0.2458839863538742, "global_step": 55526, "epoch": 623, "lr": 9.016373267663747e-05}
{"train_loss": 0.23527708649635315, "global_step": 55527, "epoch": 623, "lr": 9.016338737772075e-05}
{"train_loss": 0.21247446537017822, "global_step": 55528, "epoch": 623, "lr": 9.016304207340453e-05}
{"train_loss": 0.23925450444221497, "global_step": 55529, "epoch": 623, "lr": 9.016269676368891e-05}
{"train_loss": 0.16171672940254211, "global_step": 55530, "epoch": 623, "lr": 9.016235144857388e-05}
{"train_loss": 0.19568170607089996, "global_step": 55531, "epoch": 623, "lr": 9.016200612805952e-05}
{"train_loss": 0.29560044407844543, "global_step": 55532, "epoch": 623, "lr": 9.016166080214585e-05}
{"train_loss": 0.2328079789876938, "global_step": 55533, "epoch": 623, "lr": 9.016131547083296e-05}
{"train_loss": 0.3901980221271515, "global_step": 55534, "epoch": 623, "lr": 9.016097013412085e-05}
{"train_loss": 0.23787591664978627, "global_step": 55535, "epoch": 623, "lr": 9.016062479200962e-05, "val_loss": 3.5094141960144043}
{"train_loss": 0.2173197716474533, "global_step": 55536, "epoch": 624, "lr": 9.016027944449926e-05}
{"train_loss": 0.20072022080421448, "global_step": 55537, "epoch": 624, "lr": 9.015993409158983e-05}
{"train_loss": 0.3344677984714508, "global_step": 55538, "epoch": 624, "lr": 9.01595887332814e-05}
{"train_loss": 0.385532945394516, "global_step": 55539, "epoch": 624, "lr": 9.015924336957401e-05}
{"train_loss": 0.28849178552627563, "global_step": 55540, "epoch": 624, "lr": 9.015889800046769e-05}
{"train_loss": 0.2019554078578949, "global_step": 55541, "epoch": 624, "lr": 9.015855262596249e-05}
{"train_loss": 0.35503214597702026, "global_step": 55542, "epoch": 624, "lr": 9.015820724605846e-05}
{"train_loss": 0.288705438375473, "global_step": 55543, "epoch": 624, "lr": 9.015786186075567e-05}
{"train_loss": 0.22893604636192322, "global_step": 55544, "epoch": 624, "lr": 9.015751647005413e-05}
{"train_loss": 0.20146945118904114, "global_step": 55545, "epoch": 624, "lr": 9.01571710739539e-05}
{"train_loss": 0.3362383544445038, "global_step": 55546, "epoch": 624, "lr": 9.015682567245503e-05}
{"train_loss": 0.24944758415222168, "global_step": 55547, "epoch": 624, "lr": 9.015648026555757e-05}
{"train_loss": 0.2609756290912628, "global_step": 55548, "epoch": 624, "lr": 9.015613485326156e-05}
{"train_loss": 0.2033132016658783, "global_step": 55549, "epoch": 624, "lr": 9.015578943556704e-05}
{"train_loss": 0.2447270154953003, "global_step": 55550, "epoch": 624, "lr": 9.015544401247405e-05}
{"train_loss": 0.29228320717811584, "global_step": 55551, "epoch": 624, "lr": 9.015509858398267e-05}
{"train_loss": 0.28011855483055115, "global_step": 55552, "epoch": 624, "lr": 9.015475315009291e-05}
{"train_loss": 0.15101999044418335, "global_step": 55553, "epoch": 624, "lr": 9.015440771080486e-05}
{"train_loss": 0.26754364371299744, "global_step": 55554, "epoch": 624, "lr": 9.015406226611854e-05}
{"train_loss": 0.23799698054790497, "global_step": 55555, "epoch": 624, "lr": 9.015371681603397e-05}
{"train_loss": 0.24661298096179962, "global_step": 55556, "epoch": 624, "lr": 9.015337136055123e-05}
{"train_loss": 0.2680968642234802, "global_step": 55557, "epoch": 624, "lr": 9.015302589967036e-05}
{"train_loss": 0.1921427994966507, "global_step": 55558, "epoch": 624, "lr": 9.015268043339141e-05}
{"train_loss": 0.19762687385082245, "global_step": 55559, "epoch": 624, "lr": 9.015233496171441e-05}
{"train_loss": 0.2744933068752289, "global_step": 55560, "epoch": 624, "lr": 9.015198948463943e-05}
{"train_loss": 0.2902417480945587, "global_step": 55561, "epoch": 624, "lr": 9.015164400216649e-05}
{"train_loss": 0.28481414914131165, "global_step": 55562, "epoch": 624, "lr": 9.015129851429566e-05}
{"train_loss": 0.27283141016960144, "global_step": 55563, "epoch": 624, "lr": 9.015095302102699e-05}
{"train_loss": 0.22265426814556122, "global_step": 55564, "epoch": 624, "lr": 9.01506075223605e-05}
{"train_loss": 0.2832099199295044, "global_step": 55565, "epoch": 624, "lr": 9.015026201829626e-05}
{"train_loss": 0.27963921427726746, "global_step": 55566, "epoch": 624, "lr": 9.014991650883429e-05}
{"train_loss": 0.20059801638126373, "global_step": 55567, "epoch": 624, "lr": 9.014957099397466e-05}
{"train_loss": 0.3176456093788147, "global_step": 55568, "epoch": 624, "lr": 9.014922547371742e-05}
{"train_loss": 0.22685104608535767, "global_step": 55569, "epoch": 624, "lr": 9.01488799480626e-05}
{"train_loss": 0.25196388363838196, "global_step": 55570, "epoch": 624, "lr": 9.014853441701023e-05}
{"train_loss": 0.17911973595619202, "global_step": 55571, "epoch": 624, "lr": 9.014818888056041e-05}
{"train_loss": 0.27921566367149353, "global_step": 55572, "epoch": 624, "lr": 9.014784333871314e-05}
{"train_loss": 0.3441772758960724, "global_step": 55573, "epoch": 624, "lr": 9.014749779146849e-05}
{"train_loss": 0.2226361185312271, "global_step": 55574, "epoch": 624, "lr": 9.014715223882649e-05}
{"train_loss": 0.292648583650589, "global_step": 55575, "epoch": 624, "lr": 9.01468066807872e-05}
{"train_loss": 0.3059608042240143, "global_step": 55576, "epoch": 624, "lr": 9.014646111735067e-05}
{"train_loss": 0.3213149905204773, "global_step": 55577, "epoch": 624, "lr": 9.014611554851694e-05}
{"train_loss": 0.2596762180328369, "global_step": 55578, "epoch": 624, "lr": 9.014576997428602e-05}
{"train_loss": 0.31202155351638794, "global_step": 55579, "epoch": 624, "lr": 9.0145424394658e-05}
{"train_loss": 0.28506243228912354, "global_step": 55580, "epoch": 624, "lr": 9.014507880963294e-05}
{"train_loss": 0.3132418692111969, "global_step": 55581, "epoch": 624, "lr": 9.014473321921085e-05}
{"train_loss": 0.18757425248622894, "global_step": 55582, "epoch": 624, "lr": 9.01443876233918e-05}
{"train_loss": 0.2613040804862976, "global_step": 55583, "epoch": 624, "lr": 9.014404202217581e-05}
{"train_loss": 0.26793360710144043, "global_step": 55584, "epoch": 624, "lr": 9.014369641556295e-05}
{"train_loss": 0.22368115186691284, "global_step": 55585, "epoch": 624, "lr": 9.014335080355327e-05}
{"train_loss": 0.1791456639766693, "global_step": 55586, "epoch": 624, "lr": 9.01430051861468e-05}
{"train_loss": 0.2840023338794708, "global_step": 55587, "epoch": 624, "lr": 9.014265956334358e-05}
{"train_loss": 0.29000070691108704, "global_step": 55588, "epoch": 624, "lr": 9.014231393514367e-05}
{"train_loss": 0.22412964701652527, "global_step": 55589, "epoch": 624, "lr": 9.014196830154713e-05}
{"train_loss": 0.2625961899757385, "global_step": 55590, "epoch": 624, "lr": 9.0141622662554e-05}
{"train_loss": 0.24518777430057526, "global_step": 55591, "epoch": 624, "lr": 9.01412770181643e-05}
{"train_loss": 0.3064364194869995, "global_step": 55592, "epoch": 624, "lr": 9.014093136837809e-05}
{"train_loss": 0.28027793765068054, "global_step": 55593, "epoch": 624, "lr": 9.014058571319542e-05}
{"train_loss": 0.2395653873682022, "global_step": 55594, "epoch": 624, "lr": 9.014024005261635e-05}
{"train_loss": 0.3082861006259918, "global_step": 55595, "epoch": 624, "lr": 9.01398943866409e-05}
{"train_loss": 0.2098112404346466, "global_step": 55596, "epoch": 624, "lr": 9.013954871526916e-05}
{"train_loss": 0.27509212493896484, "global_step": 55597, "epoch": 624, "lr": 9.013920303850112e-05}
{"train_loss": 0.2125302255153656, "global_step": 55598, "epoch": 624, "lr": 9.013885735633685e-05}
{"train_loss": 0.25643616914749146, "global_step": 55599, "epoch": 624, "lr": 9.013851166877642e-05}
{"train_loss": 0.2762492597103119, "global_step": 55600, "epoch": 624, "lr": 9.013816597581985e-05}
{"train_loss": 0.36835211515426636, "global_step": 55601, "epoch": 624, "lr": 9.013782027746717e-05}
{"train_loss": 0.2855836749076843, "global_step": 55602, "epoch": 624, "lr": 9.013747457371848e-05}
{"train_loss": 0.1469009518623352, "global_step": 55603, "epoch": 624, "lr": 9.013712886457379e-05}
{"train_loss": 0.24903737008571625, "global_step": 55604, "epoch": 624, "lr": 9.013678315003314e-05}
{"train_loss": 0.24068085849285126, "global_step": 55605, "epoch": 624, "lr": 9.01364374300966e-05}
{"train_loss": 0.3201836347579956, "global_step": 55606, "epoch": 624, "lr": 9.01360917047642e-05}
{"train_loss": 0.19938693940639496, "global_step": 55607, "epoch": 624, "lr": 9.013574597403599e-05}
{"train_loss": 0.30627748370170593, "global_step": 55608, "epoch": 624, "lr": 9.013540023791203e-05}
{"train_loss": 0.25551697611808777, "global_step": 55609, "epoch": 624, "lr": 9.013505449639235e-05}
{"train_loss": 0.21490852534770966, "global_step": 55610, "epoch": 624, "lr": 9.013470874947698e-05}
{"train_loss": 0.32054388523101807, "global_step": 55611, "epoch": 624, "lr": 9.0134362997166e-05}
{"train_loss": 0.2764829397201538, "global_step": 55612, "epoch": 624, "lr": 9.013401723945945e-05}
{"train_loss": 0.24813564121723175, "global_step": 55613, "epoch": 624, "lr": 9.013367147635737e-05}
{"train_loss": 0.20626170933246613, "global_step": 55614, "epoch": 624, "lr": 9.013332570785981e-05}
{"train_loss": 0.20109549164772034, "global_step": 55615, "epoch": 624, "lr": 9.013297993396681e-05}
{"train_loss": 0.28107690811157227, "global_step": 55616, "epoch": 624, "lr": 9.01326341546784e-05}
{"train_loss": 0.27431851625442505, "global_step": 55617, "epoch": 624, "lr": 9.013228836999468e-05}
{"train_loss": 0.1740368902683258, "global_step": 55618, "epoch": 624, "lr": 9.013194257991566e-05}
{"train_loss": 0.338574081659317, "global_step": 55619, "epoch": 624, "lr": 9.013159678444136e-05}
{"train_loss": 0.3273041844367981, "global_step": 55620, "epoch": 624, "lr": 9.013125098357188e-05}
{"train_loss": 0.32951587438583374, "global_step": 55621, "epoch": 624, "lr": 9.013090517730723e-05}
{"train_loss": 0.23331648111343384, "global_step": 55622, "epoch": 624, "lr": 9.013055936564748e-05}
{"train_loss": 0.3085451126098633, "global_step": 55623, "epoch": 624, "lr": 9.013021354859266e-05}
{"train_loss": 0.2617891693717978, "global_step": 55624, "epoch": 624, "lr": 9.012986772614283e-05, "val_loss": 3.29936146736145}
{"train_loss": 0.26279520988464355, "global_step": 55625, "epoch": 625, "lr": 9.012952189829802e-05}
{"train_loss": 0.24152646958827972, "global_step": 55626, "epoch": 625, "lr": 9.012917606505829e-05}
{"train_loss": 0.2756500244140625, "global_step": 55627, "epoch": 625, "lr": 9.012883022642366e-05}
{"train_loss": 0.20164266228675842, "global_step": 55628, "epoch": 625, "lr": 9.012848438239424e-05}
{"train_loss": 0.3387520909309387, "global_step": 55629, "epoch": 625, "lr": 9.012813853297e-05}
{"train_loss": 0.29624757170677185, "global_step": 55630, "epoch": 625, "lr": 9.012779267815101e-05}
{"train_loss": 0.2637879252433777, "global_step": 55631, "epoch": 625, "lr": 9.012744681793737e-05}
{"train_loss": 0.28016188740730286, "global_step": 55632, "epoch": 625, "lr": 9.012710095232906e-05}
{"train_loss": 0.25514593720436096, "global_step": 55633, "epoch": 625, "lr": 9.012675508132614e-05}
{"train_loss": 0.2853274941444397, "global_step": 55634, "epoch": 625, "lr": 9.012640920492867e-05}
{"train_loss": 0.22624003887176514, "global_step": 55635, "epoch": 625, "lr": 9.012606332313672e-05}
{"train_loss": 0.25278547406196594, "global_step": 55636, "epoch": 625, "lr": 9.012571743595028e-05}
{"train_loss": 0.34202003479003906, "global_step": 55637, "epoch": 625, "lr": 9.012537154336944e-05}
{"train_loss": 0.2504313588142395, "global_step": 55638, "epoch": 625, "lr": 9.012502564539424e-05}
{"train_loss": 0.24341043829917908, "global_step": 55639, "epoch": 625, "lr": 9.01246797420247e-05}
{"train_loss": 0.22328227758407593, "global_step": 55640, "epoch": 625, "lr": 9.01243338332609e-05}
{"train_loss": 0.21137338876724243, "global_step": 55641, "epoch": 625, "lr": 9.012398791910287e-05}
{"train_loss": 0.19176450371742249, "global_step": 55642, "epoch": 625, "lr": 9.012364199955065e-05}
{"train_loss": 0.3103165924549103, "global_step": 55643, "epoch": 625, "lr": 9.01232960746043e-05}
{"train_loss": 0.3232048749923706, "global_step": 55644, "epoch": 625, "lr": 9.012295014426387e-05}
{"train_loss": 0.2426518052816391, "global_step": 55645, "epoch": 625, "lr": 9.012260420852939e-05}
{"train_loss": 0.2402278333902359, "global_step": 55646, "epoch": 625, "lr": 9.012225826740093e-05}
{"train_loss": 0.25468218326568604, "global_step": 55647, "epoch": 625, "lr": 9.01219123208785e-05}
{"train_loss": 0.2528764009475708, "global_step": 55648, "epoch": 625, "lr": 9.012156636896219e-05}
{"train_loss": 0.2635149359703064, "global_step": 55649, "epoch": 625, "lr": 9.012122041165201e-05}
{"train_loss": 0.2667596638202667, "global_step": 55650, "epoch": 625, "lr": 9.012087444894801e-05}
{"train_loss": 0.29396429657936096, "global_step": 55651, "epoch": 625, "lr": 9.012052848085027e-05}
{"train_loss": 0.21179145574569702, "global_step": 55652, "epoch": 625, "lr": 9.012018250735879e-05}
{"train_loss": 0.26372790336608887, "global_step": 55653, "epoch": 625, "lr": 9.011983652847367e-05}
{"train_loss": 0.2867218554019928, "global_step": 55654, "epoch": 625, "lr": 9.01194905441949e-05}
{"train_loss": 0.29032260179519653, "global_step": 55655, "epoch": 625, "lr": 9.011914455452258e-05}
{"train_loss": 0.21399357914924622, "global_step": 55656, "epoch": 625, "lr": 9.011879855945671e-05}
{"train_loss": 0.18550334870815277, "global_step": 55657, "epoch": 625, "lr": 9.011845255899735e-05}
{"train_loss": 0.20951254665851593, "global_step": 55658, "epoch": 625, "lr": 9.011810655314457e-05}
{"train_loss": 0.27421998977661133, "global_step": 55659, "epoch": 625, "lr": 9.011776054189838e-05}
{"train_loss": 0.2526244819164276, "global_step": 55660, "epoch": 625, "lr": 9.011741452525886e-05}
{"train_loss": 0.24220579862594604, "global_step": 55661, "epoch": 625, "lr": 9.011706850322606e-05}
{"train_loss": 0.24755437672138214, "global_step": 55662, "epoch": 625, "lr": 9.011672247579999e-05}
{"train_loss": 0.15177671611309052, "global_step": 55663, "epoch": 625, "lr": 9.011637644298073e-05}
{"train_loss": 0.20885848999023438, "global_step": 55664, "epoch": 625, "lr": 9.011603040476828e-05}
{"train_loss": 0.30188679695129395, "global_step": 55665, "epoch": 625, "lr": 9.011568436116274e-05}
{"train_loss": 0.24645976722240448, "global_step": 55666, "epoch": 625, "lr": 9.011533831216413e-05}
{"train_loss": 0.2455100268125534, "global_step": 55667, "epoch": 625, "lr": 9.011499225777251e-05}
{"train_loss": 0.21124625205993652, "global_step": 55668, "epoch": 625, "lr": 9.011464619798791e-05}
{"train_loss": 0.2300843596458435, "global_step": 55669, "epoch": 625, "lr": 9.01143001328104e-05}
{"train_loss": 0.16215240955352783, "global_step": 55670, "epoch": 625, "lr": 9.011395406223999e-05}
{"train_loss": 0.26177728176116943, "global_step": 55671, "epoch": 625, "lr": 9.011360798627674e-05}
{"train_loss": 0.2346334010362625, "global_step": 55672, "epoch": 625, "lr": 9.011326190492073e-05}
{"train_loss": 0.24960801005363464, "global_step": 55673, "epoch": 625, "lr": 9.011291581817198e-05}
{"train_loss": 0.28632986545562744, "global_step": 55674, "epoch": 625, "lr": 9.011256972603052e-05}
{"train_loss": 0.24363337457180023, "global_step": 55675, "epoch": 625, "lr": 9.011222362849642e-05}
{"train_loss": 0.32266783714294434, "global_step": 55676, "epoch": 625, "lr": 9.011187752556974e-05}
{"train_loss": 0.215076744556427, "global_step": 55677, "epoch": 625, "lr": 9.011153141725049e-05}
{"train_loss": 0.2554519474506378, "global_step": 55678, "epoch": 625, "lr": 9.011118530353873e-05}
{"train_loss": 0.18441684544086456, "global_step": 55679, "epoch": 625, "lr": 9.011083918443451e-05}
{"train_loss": 0.2804952561855316, "global_step": 55680, "epoch": 625, "lr": 9.011049305993788e-05}
{"train_loss": 0.25201621651649475, "global_step": 55681, "epoch": 625, "lr": 9.011014693004889e-05}
{"train_loss": 0.22289535403251648, "global_step": 55682, "epoch": 625, "lr": 9.010980079476757e-05}
{"train_loss": 0.2028832882642746, "global_step": 55683, "epoch": 625, "lr": 9.010945465409398e-05}
{"train_loss": 0.23540589213371277, "global_step": 55684, "epoch": 625, "lr": 9.010910850802817e-05}
{"train_loss": 0.3159085512161255, "global_step": 55685, "epoch": 625, "lr": 9.010876235657018e-05}
{"train_loss": 0.19700337946414948, "global_step": 55686, "epoch": 625, "lr": 9.010841619972003e-05}
{"train_loss": 0.1911165863275528, "global_step": 55687, "epoch": 625, "lr": 9.010807003747781e-05}
{"train_loss": 0.18799982964992523, "global_step": 55688, "epoch": 625, "lr": 9.010772386984355e-05}
{"train_loss": 0.2708718478679657, "global_step": 55689, "epoch": 625, "lr": 9.01073776968173e-05}
{"train_loss": 0.33022090792655945, "global_step": 55690, "epoch": 625, "lr": 9.010703151839909e-05}
{"train_loss": 0.2225290834903717, "global_step": 55691, "epoch": 625, "lr": 9.010668533458899e-05}
{"train_loss": 0.25818613171577454, "global_step": 55692, "epoch": 625, "lr": 9.010633914538702e-05}
{"train_loss": 0.21019592881202698, "global_step": 55693, "epoch": 625, "lr": 9.010599295079325e-05}
{"train_loss": 0.21224896609783173, "global_step": 55694, "epoch": 625, "lr": 9.010564675080773e-05}
{"train_loss": 0.2648088037967682, "global_step": 55695, "epoch": 625, "lr": 9.010530054543048e-05}
{"train_loss": 0.24961772561073303, "global_step": 55696, "epoch": 625, "lr": 9.010495433466157e-05}
{"train_loss": 0.2504464089870453, "global_step": 55697, "epoch": 625, "lr": 9.010460811850102e-05}
{"train_loss": 0.30880486965179443, "global_step": 55698, "epoch": 625, "lr": 9.010426189694891e-05}
{"train_loss": 0.18408384919166565, "global_step": 55699, "epoch": 625, "lr": 9.010391567000527e-05}
{"train_loss": 0.17835579812526703, "global_step": 55700, "epoch": 625, "lr": 9.010356943767015e-05}
{"train_loss": 0.19273005425930023, "global_step": 55701, "epoch": 625, "lr": 9.010322319994358e-05}
{"train_loss": 0.2413533627986908, "global_step": 55702, "epoch": 625, "lr": 9.010287695682564e-05}
{"train_loss": 0.35266247391700745, "global_step": 55703, "epoch": 625, "lr": 9.010253070831634e-05}
{"train_loss": 0.24632304906845093, "global_step": 55704, "epoch": 625, "lr": 9.010218445441576e-05}
{"train_loss": 0.2754707336425781, "global_step": 55705, "epoch": 625, "lr": 9.01018381951239e-05}
{"train_loss": 0.29002803564071655, "global_step": 55706, "epoch": 625, "lr": 9.010149193044086e-05}
{"train_loss": 0.25148600339889526, "global_step": 55707, "epoch": 625, "lr": 9.010114566036666e-05}
{"train_loss": 0.30475783348083496, "global_step": 55708, "epoch": 625, "lr": 9.010079938490135e-05}
{"train_loss": 0.25776350498199463, "global_step": 55709, "epoch": 625, "lr": 9.010045310404497e-05}
{"train_loss": 0.21818028390407562, "global_step": 55710, "epoch": 625, "lr": 9.010010681779759e-05}
{"train_loss": 0.23281972110271454, "global_step": 55711, "epoch": 625, "lr": 9.009976052615923e-05}
{"train_loss": 0.21326902508735657, "global_step": 55712, "epoch": 625, "lr": 9.009941422912994e-05}
{"train_loss": 0.24870100667637385, "global_step": 55713, "epoch": 625, "lr": 9.009906792670977e-05, "val_loss": 3.2269058227539062, "train_action_mse_error": 28.73790740966797}
{"train_loss": 0.2750736474990845, "global_step": 55714, "epoch": 626, "lr": 9.009872161889878e-05}
{"train_loss": 0.22973030805587769, "global_step": 55715, "epoch": 626, "lr": 9.0098375305697e-05}
{"train_loss": 0.2667282521724701, "global_step": 55716, "epoch": 626, "lr": 9.009802898710448e-05}
{"train_loss": 0.19158107042312622, "global_step": 55717, "epoch": 626, "lr": 9.009768266312128e-05}
{"train_loss": 0.2757570445537567, "global_step": 55718, "epoch": 626, "lr": 9.009733633374742e-05}
{"train_loss": 0.19160014390945435, "global_step": 55719, "epoch": 626, "lr": 9.009698999898298e-05}
{"train_loss": 0.3208172023296356, "global_step": 55720, "epoch": 626, "lr": 9.009664365882798e-05}
{"train_loss": 0.29375791549682617, "global_step": 55721, "epoch": 626, "lr": 9.009629731328246e-05}
{"train_loss": 0.20858153700828552, "global_step": 55722, "epoch": 626, "lr": 9.009595096234649e-05}
{"train_loss": 0.24880293011665344, "global_step": 55723, "epoch": 626, "lr": 9.009560460602011e-05}
{"train_loss": 0.24224169552326202, "global_step": 55724, "epoch": 626, "lr": 9.009525824430337e-05}
{"train_loss": 0.21516452729701996, "global_step": 55725, "epoch": 626, "lr": 9.00949118771963e-05}
{"train_loss": 0.33213233947753906, "global_step": 55726, "epoch": 626, "lr": 9.009456550469897e-05}
{"train_loss": 0.20606662333011627, "global_step": 55727, "epoch": 626, "lr": 9.009421912681142e-05}
{"train_loss": 0.24078218638896942, "global_step": 55728, "epoch": 626, "lr": 9.009387274353367e-05}
{"train_loss": 0.2692534923553467, "global_step": 55729, "epoch": 626, "lr": 9.009352635486581e-05}
{"train_loss": 0.2616116404533386, "global_step": 55730, "epoch": 626, "lr": 9.009317996080785e-05}
{"train_loss": 0.21436655521392822, "global_step": 55731, "epoch": 626, "lr": 9.009283356135985e-05}
{"train_loss": 0.2701043486595154, "global_step": 55732, "epoch": 626, "lr": 9.009248715652186e-05}
{"train_loss": 0.14015164971351624, "global_step": 55733, "epoch": 626, "lr": 9.009214074629392e-05}
{"train_loss": 0.2810891270637512, "global_step": 55734, "epoch": 626, "lr": 9.009179433067608e-05}
{"train_loss": 0.2686570882797241, "global_step": 55735, "epoch": 626, "lr": 9.00914479096684e-05}
{"train_loss": 0.22197408974170685, "global_step": 55736, "epoch": 626, "lr": 9.00911014832709e-05}
{"train_loss": 0.2427676022052765, "global_step": 55737, "epoch": 626, "lr": 9.009075505148364e-05}
{"train_loss": 0.2825520932674408, "global_step": 55738, "epoch": 626, "lr": 9.009040861430668e-05}
{"train_loss": 0.2647566795349121, "global_step": 55739, "epoch": 626, "lr": 9.009006217174005e-05}
{"train_loss": 0.16308560967445374, "global_step": 55740, "epoch": 626, "lr": 9.00897157237838e-05}
{"train_loss": 0.2513992488384247, "global_step": 55741, "epoch": 626, "lr": 9.008936927043796e-05}
{"train_loss": 0.24431517720222473, "global_step": 55742, "epoch": 626, "lr": 9.00890228117026e-05}
{"train_loss": 0.31863319873809814, "global_step": 55743, "epoch": 626, "lr": 9.008867634757777e-05}
{"train_loss": 0.27189865708351135, "global_step": 55744, "epoch": 626, "lr": 9.00883298780635e-05}
{"train_loss": 0.3042796552181244, "global_step": 55745, "epoch": 626, "lr": 9.008798340315983e-05}
{"train_loss": 0.21256549656391144, "global_step": 55746, "epoch": 626, "lr": 9.008763692286683e-05}
{"train_loss": 0.1777300089597702, "global_step": 55747, "epoch": 626, "lr": 9.008729043718454e-05}
{"train_loss": 0.18192815780639648, "global_step": 55748, "epoch": 626, "lr": 9.0086943946113e-05}
{"train_loss": 0.28844258189201355, "global_step": 55749, "epoch": 626, "lr": 9.008659744965227e-05}
{"train_loss": 0.2260761260986328, "global_step": 55750, "epoch": 626, "lr": 9.008625094780236e-05}
{"train_loss": 0.2960774302482605, "global_step": 55751, "epoch": 626, "lr": 9.008590444056335e-05}
{"train_loss": 0.29011523723602295, "global_step": 55752, "epoch": 626, "lr": 9.00855579279353e-05}
{"train_loss": 0.1893763542175293, "global_step": 55753, "epoch": 626, "lr": 9.008521140991822e-05}
{"train_loss": 0.2326681911945343, "global_step": 55754, "epoch": 626, "lr": 9.008486488651217e-05}
{"train_loss": 0.42976510524749756, "global_step": 55755, "epoch": 626, "lr": 9.00845183577172e-05}
{"train_loss": 0.21631474792957306, "global_step": 55756, "epoch": 626, "lr": 9.008417182353335e-05}
{"train_loss": 0.23411737382411957, "global_step": 55757, "epoch": 626, "lr": 9.008382528396068e-05}
{"train_loss": 0.2678438723087311, "global_step": 55758, "epoch": 626, "lr": 9.008347873899923e-05}
{"train_loss": 0.31140172481536865, "global_step": 55759, "epoch": 626, "lr": 9.008313218864904e-05}
{"train_loss": 0.22728943824768066, "global_step": 55760, "epoch": 626, "lr": 9.008278563291015e-05}
{"train_loss": 0.288356214761734, "global_step": 55761, "epoch": 626, "lr": 9.008243907178263e-05}
{"train_loss": 0.2672094702720642, "global_step": 55762, "epoch": 626, "lr": 9.008209250526652e-05}
{"train_loss": 0.25125592947006226, "global_step": 55763, "epoch": 626, "lr": 9.008174593336185e-05}
{"train_loss": 0.25406575202941895, "global_step": 55764, "epoch": 626, "lr": 9.008139935606869e-05}
{"train_loss": 0.1695912778377533, "global_step": 55765, "epoch": 626, "lr": 9.008105277338705e-05}
{"train_loss": 0.2802307903766632, "global_step": 55766, "epoch": 626, "lr": 9.008070618531703e-05}
{"train_loss": 0.19654886424541473, "global_step": 55767, "epoch": 626, "lr": 9.008035959185863e-05}
{"train_loss": 0.24819940328598022, "global_step": 55768, "epoch": 626, "lr": 9.008001299301192e-05}
{"train_loss": 0.2498195320367813, "global_step": 55769, "epoch": 626, "lr": 9.007966638877694e-05}
{"train_loss": 0.21943306922912598, "global_step": 55770, "epoch": 626, "lr": 9.007931977915375e-05}
{"train_loss": 0.2680867910385132, "global_step": 55771, "epoch": 626, "lr": 9.007897316414237e-05}
{"train_loss": 0.2185203731060028, "global_step": 55772, "epoch": 626, "lr": 9.007862654374286e-05}
{"train_loss": 0.24875465035438538, "global_step": 55773, "epoch": 626, "lr": 9.007827991795528e-05}
{"train_loss": 0.27979767322540283, "global_step": 55774, "epoch": 626, "lr": 9.007793328677964e-05}
{"train_loss": 0.2855041027069092, "global_step": 55775, "epoch": 626, "lr": 9.007758665021603e-05}
{"train_loss": 0.27766406536102295, "global_step": 55776, "epoch": 626, "lr": 9.007724000826447e-05}
{"train_loss": 0.24660396575927734, "global_step": 55777, "epoch": 626, "lr": 9.007689336092501e-05}
{"train_loss": 0.3406605124473572, "global_step": 55778, "epoch": 626, "lr": 9.007654670819771e-05}
{"train_loss": 0.2132137268781662, "global_step": 55779, "epoch": 626, "lr": 9.007620005008259e-05}
{"train_loss": 0.2310880869626999, "global_step": 55780, "epoch": 626, "lr": 9.007585338657974e-05}
{"train_loss": 0.24488380551338196, "global_step": 55781, "epoch": 626, "lr": 9.007550671768916e-05}
{"train_loss": 0.2352619469165802, "global_step": 55782, "epoch": 626, "lr": 9.007516004341092e-05}
{"train_loss": 0.22024504840373993, "global_step": 55783, "epoch": 626, "lr": 9.007481336374505e-05}
{"train_loss": 0.308805912733078, "global_step": 55784, "epoch": 626, "lr": 9.007446667869164e-05}
{"train_loss": 0.24874457716941833, "global_step": 55785, "epoch": 626, "lr": 9.007411998825068e-05}
{"train_loss": 0.25697624683380127, "global_step": 55786, "epoch": 626, "lr": 9.007377329242226e-05}
{"train_loss": 0.28322839736938477, "global_step": 55787, "epoch": 626, "lr": 9.007342659120641e-05}
{"train_loss": 0.2874803841114044, "global_step": 55788, "epoch": 626, "lr": 9.007307988460317e-05}
{"train_loss": 0.3001345694065094, "global_step": 55789, "epoch": 626, "lr": 9.007273317261259e-05}
{"train_loss": 0.27328380942344666, "global_step": 55790, "epoch": 626, "lr": 9.007238645523473e-05}
{"train_loss": 0.3621719479560852, "global_step": 55791, "epoch": 626, "lr": 9.007203973246961e-05}
{"train_loss": 0.20934218168258667, "global_step": 55792, "epoch": 626, "lr": 9.00716930043173e-05}
{"train_loss": 0.27955302596092224, "global_step": 55793, "epoch": 626, "lr": 9.007134627077785e-05}
{"train_loss": 0.31401997804641724, "global_step": 55794, "epoch": 626, "lr": 9.007099953185127e-05}
{"train_loss": 0.21968771517276764, "global_step": 55795, "epoch": 626, "lr": 9.007065278753766e-05}
{"train_loss": 0.24251359701156616, "global_step": 55796, "epoch": 626, "lr": 9.007030603783701e-05}
{"train_loss": 0.2106761485338211, "global_step": 55797, "epoch": 626, "lr": 9.006995928274942e-05}
{"train_loss": 0.23987098038196564, "global_step": 55798, "epoch": 626, "lr": 9.00696125222749e-05}
{"train_loss": 0.2726614773273468, "global_step": 55799, "epoch": 626, "lr": 9.006926575641352e-05}
{"train_loss": 0.29783421754837036, "global_step": 55800, "epoch": 626, "lr": 9.006891898516532e-05}
{"train_loss": 0.2808314859867096, "global_step": 55801, "epoch": 626, "lr": 9.006857220853033e-05}
{"train_loss": 0.2550183700376682, "global_step": 55802, "epoch": 626, "lr": 9.006822542650862e-05, "val_loss": 3.2277042865753174}
{"train_loss": 0.22286923229694366, "global_step": 55803, "epoch": 627, "lr": 9.00678786391002e-05}
{"train_loss": 0.2170807123184204, "global_step": 55804, "epoch": 627, "lr": 9.006753184630516e-05}
{"train_loss": 0.2610408663749695, "global_step": 55805, "epoch": 627, "lr": 9.006718504812353e-05}
{"train_loss": 0.2804807424545288, "global_step": 55806, "epoch": 627, "lr": 9.006683824455535e-05}
{"train_loss": 0.22609089314937592, "global_step": 55807, "epoch": 627, "lr": 9.006649143560067e-05}
{"train_loss": 0.19971810281276703, "global_step": 55808, "epoch": 627, "lr": 9.006614462125955e-05}
{"train_loss": 0.21616488695144653, "global_step": 55809, "epoch": 627, "lr": 9.006579780153202e-05}
{"train_loss": 0.19332171976566315, "global_step": 55810, "epoch": 627, "lr": 9.006545097641813e-05}
{"train_loss": 0.330014169216156, "global_step": 55811, "epoch": 627, "lr": 9.006510414591794e-05}
{"train_loss": 0.30722612142562866, "global_step": 55812, "epoch": 627, "lr": 9.006475731003145e-05}
{"train_loss": 0.2687510848045349, "global_step": 55813, "epoch": 627, "lr": 9.006441046875878e-05}
{"train_loss": 0.20156928896903992, "global_step": 55814, "epoch": 627, "lr": 9.006406362209991e-05}
{"train_loss": 0.2637220621109009, "global_step": 55815, "epoch": 627, "lr": 9.006371677005494e-05}
{"train_loss": 0.17157910764217377, "global_step": 55816, "epoch": 627, "lr": 9.006336991262387e-05}
{"train_loss": 0.21683959662914276, "global_step": 55817, "epoch": 627, "lr": 9.006302304980678e-05}
{"train_loss": 0.2608998417854309, "global_step": 55818, "epoch": 627, "lr": 9.006267618160372e-05}
{"train_loss": 0.2667541801929474, "global_step": 55819, "epoch": 627, "lr": 9.00623293080147e-05}
{"train_loss": 0.2739691734313965, "global_step": 55820, "epoch": 627, "lr": 9.006198242903978e-05}
{"train_loss": 0.2564767599105835, "global_step": 55821, "epoch": 627, "lr": 9.006163554467904e-05}
{"train_loss": 0.19813652336597443, "global_step": 55822, "epoch": 627, "lr": 9.006128865493248e-05}
{"train_loss": 0.23736456036567688, "global_step": 55823, "epoch": 627, "lr": 9.006094175980018e-05}
{"train_loss": 0.26536330580711365, "global_step": 55824, "epoch": 627, "lr": 9.006059485928218e-05}
{"train_loss": 0.182170569896698, "global_step": 55825, "epoch": 627, "lr": 9.006024795337851e-05}
{"train_loss": 0.24466630816459656, "global_step": 55826, "epoch": 627, "lr": 9.005990104208923e-05}
{"train_loss": 0.31430840492248535, "global_step": 55827, "epoch": 627, "lr": 9.005955412541438e-05}
{"train_loss": 0.18898417055606842, "global_step": 55828, "epoch": 627, "lr": 9.005920720335402e-05}
{"train_loss": 0.1949089616537094, "global_step": 55829, "epoch": 627, "lr": 9.005886027590819e-05}
{"train_loss": 0.2285025268793106, "global_step": 55830, "epoch": 627, "lr": 9.005851334307692e-05}
{"train_loss": 0.2950055003166199, "global_step": 55831, "epoch": 627, "lr": 9.00581664048603e-05}
{"train_loss": 0.28928595781326294, "global_step": 55832, "epoch": 627, "lr": 9.005781946125831e-05}
{"train_loss": 0.27650776505470276, "global_step": 55833, "epoch": 627, "lr": 9.005747251227106e-05}
{"train_loss": 0.19417087733745575, "global_step": 55834, "epoch": 627, "lr": 9.005712555789856e-05}
{"train_loss": 0.3104785084724426, "global_step": 55835, "epoch": 627, "lr": 9.005677859814087e-05}
{"train_loss": 0.26935240626335144, "global_step": 55836, "epoch": 627, "lr": 9.005643163299804e-05}
{"train_loss": 0.26843300461769104, "global_step": 55837, "epoch": 627, "lr": 9.00560846624701e-05}
{"train_loss": 0.19377201795578003, "global_step": 55838, "epoch": 627, "lr": 9.005573768655712e-05}
{"train_loss": 0.24716685712337494, "global_step": 55839, "epoch": 627, "lr": 9.005539070525912e-05}
{"train_loss": 0.21407799422740936, "global_step": 55840, "epoch": 627, "lr": 9.005504371857617e-05}
{"train_loss": 0.23822224140167236, "global_step": 55841, "epoch": 627, "lr": 9.00546967265083e-05}
{"train_loss": 0.2456289678812027, "global_step": 55842, "epoch": 627, "lr": 9.005434972905557e-05}
{"train_loss": 0.31575900316238403, "global_step": 55843, "epoch": 627, "lr": 9.005400272621803e-05}
{"train_loss": 0.2913139760494232, "global_step": 55844, "epoch": 627, "lr": 9.005365571799572e-05}
{"train_loss": 0.26252105832099915, "global_step": 55845, "epoch": 627, "lr": 9.005330870438866e-05}
{"train_loss": 0.22755543887615204, "global_step": 55846, "epoch": 627, "lr": 9.005296168539694e-05}
{"train_loss": 0.1548948884010315, "global_step": 55847, "epoch": 627, "lr": 9.005261466102058e-05}
{"train_loss": 0.2519647479057312, "global_step": 55848, "epoch": 627, "lr": 9.005226763125964e-05}
{"train_loss": 0.22025948762893677, "global_step": 55849, "epoch": 627, "lr": 9.005192059611416e-05}
{"train_loss": 0.2763568162918091, "global_step": 55850, "epoch": 627, "lr": 9.005157355558418e-05}
{"train_loss": 0.21656844019889832, "global_step": 55851, "epoch": 627, "lr": 9.005122650966975e-05}
{"train_loss": 0.2961494028568268, "global_step": 55852, "epoch": 627, "lr": 9.005087945837094e-05}
{"train_loss": 0.24369944632053375, "global_step": 55853, "epoch": 627, "lr": 9.005053240168777e-05}
{"train_loss": 0.2276330292224884, "global_step": 55854, "epoch": 627, "lr": 9.005018533962029e-05}
{"train_loss": 0.20686471462249756, "global_step": 55855, "epoch": 627, "lr": 9.004983827216855e-05}
{"train_loss": 0.2203240692615509, "global_step": 55856, "epoch": 627, "lr": 9.00494911993326e-05}
{"train_loss": 0.27795735001564026, "global_step": 55857, "epoch": 627, "lr": 9.004914412111247e-05}
{"train_loss": 0.26515087485313416, "global_step": 55858, "epoch": 627, "lr": 9.004879703750823e-05}
{"train_loss": 0.2135392725467682, "global_step": 55859, "epoch": 627, "lr": 9.004844994851993e-05}
{"train_loss": 0.22914396226406097, "global_step": 55860, "epoch": 627, "lr": 9.004810285414759e-05}
{"train_loss": 0.23015817999839783, "global_step": 55861, "epoch": 627, "lr": 9.004775575439128e-05}
{"train_loss": 0.17467260360717773, "global_step": 55862, "epoch": 627, "lr": 9.004740864925103e-05}
{"train_loss": 0.30568429827690125, "global_step": 55863, "epoch": 627, "lr": 9.00470615387269e-05}
{"train_loss": 0.36453577876091003, "global_step": 55864, "epoch": 627, "lr": 9.004671442281893e-05}
{"train_loss": 0.24604806303977966, "global_step": 55865, "epoch": 627, "lr": 9.004636730152716e-05}
{"train_loss": 0.1645144671201706, "global_step": 55866, "epoch": 627, "lr": 9.004602017485166e-05}
{"train_loss": 0.2110893875360489, "global_step": 55867, "epoch": 627, "lr": 9.004567304279245e-05}
{"train_loss": 0.25150150060653687, "global_step": 55868, "epoch": 627, "lr": 9.004532590534958e-05}
{"train_loss": 0.33600735664367676, "global_step": 55869, "epoch": 627, "lr": 9.004497876252312e-05}
{"train_loss": 0.31323257088661194, "global_step": 55870, "epoch": 627, "lr": 9.004463161431308e-05}
{"train_loss": 0.21307922899723053, "global_step": 55871, "epoch": 627, "lr": 9.004428446071954e-05}
{"train_loss": 0.23896947503089905, "global_step": 55872, "epoch": 627, "lr": 9.004393730174255e-05}
{"train_loss": 0.250622034072876, "global_step": 55873, "epoch": 627, "lr": 9.004359013738211e-05}
{"train_loss": 0.2615009546279907, "global_step": 55874, "epoch": 627, "lr": 9.004324296763831e-05}
{"train_loss": 0.29073867201805115, "global_step": 55875, "epoch": 627, "lr": 9.004289579251119e-05}
{"train_loss": 0.2568577229976654, "global_step": 55876, "epoch": 627, "lr": 9.004254861200079e-05}
{"train_loss": 0.33743375539779663, "global_step": 55877, "epoch": 627, "lr": 9.004220142610715e-05}
{"train_loss": 0.1903778314590454, "global_step": 55878, "epoch": 627, "lr": 9.004185423483032e-05}
{"train_loss": 0.2664138972759247, "global_step": 55879, "epoch": 627, "lr": 9.004150703817036e-05}
{"train_loss": 0.27551883459091187, "global_step": 55880, "epoch": 627, "lr": 9.004115983612731e-05}
{"train_loss": 0.24305570125579834, "global_step": 55881, "epoch": 627, "lr": 9.004081262870121e-05}
{"train_loss": 0.23116259276866913, "global_step": 55882, "epoch": 627, "lr": 9.00404654158921e-05}
{"train_loss": 0.27257075905799866, "global_step": 55883, "epoch": 627, "lr": 9.004011819770005e-05}
{"train_loss": 0.24172727763652802, "global_step": 55884, "epoch": 627, "lr": 9.003977097412509e-05}
{"train_loss": 0.2668614387512207, "global_step": 55885, "epoch": 627, "lr": 9.003942374516727e-05}
{"train_loss": 0.30995675921440125, "global_step": 55886, "epoch": 627, "lr": 9.003907651082664e-05}
{"train_loss": 0.2848058044910431, "global_step": 55887, "epoch": 627, "lr": 9.003872927110324e-05}
{"train_loss": 0.28081420063972473, "global_step": 55888, "epoch": 627, "lr": 9.003838202599713e-05}
{"train_loss": 0.21249018609523773, "global_step": 55889, "epoch": 627, "lr": 9.003803477550834e-05}
{"train_loss": 0.25086942315101624, "global_step": 55890, "epoch": 627, "lr": 9.003768751963694e-05}
{"train_loss": 0.24927401659863718, "global_step": 55891, "epoch": 627, "lr": 9.003734025838296e-05, "val_loss": 3.1479203701019287}
{"train_loss": 0.24728108942508698, "global_step": 55892, "epoch": 628, "lr": 9.003699299174642e-05}
{"train_loss": 0.30148229002952576, "global_step": 55893, "epoch": 628, "lr": 9.00366457197274e-05}
{"train_loss": 0.3017723858356476, "global_step": 55894, "epoch": 628, "lr": 9.003629844232596e-05}
{"train_loss": 0.2058461606502533, "global_step": 55895, "epoch": 628, "lr": 9.003595115954211e-05}
{"train_loss": 0.41661471128463745, "global_step": 55896, "epoch": 628, "lr": 9.003560387137593e-05}
{"train_loss": 0.23718233406543732, "global_step": 55897, "epoch": 628, "lr": 9.003525657782744e-05}
{"train_loss": 0.2308349907398224, "global_step": 55898, "epoch": 628, "lr": 9.003490927889671e-05}
{"train_loss": 0.30573514103889465, "global_step": 55899, "epoch": 628, "lr": 9.003456197458376e-05}
{"train_loss": 0.2543785870075226, "global_step": 55900, "epoch": 628, "lr": 9.003421466488866e-05}
{"train_loss": 0.20378601551055908, "global_step": 55901, "epoch": 628, "lr": 9.003386734981145e-05}
{"train_loss": 0.2824631333351135, "global_step": 55902, "epoch": 628, "lr": 9.003352002935216e-05}
{"train_loss": 0.19989895820617676, "global_step": 55903, "epoch": 628, "lr": 9.003317270351086e-05}
{"train_loss": 0.2698208689689636, "global_step": 55904, "epoch": 628, "lr": 9.003282537228759e-05}
{"train_loss": 0.3149077594280243, "global_step": 55905, "epoch": 628, "lr": 9.00324780356824e-05}
{"train_loss": 0.26211437582969666, "global_step": 55906, "epoch": 628, "lr": 9.003213069369532e-05}
{"train_loss": 0.2363688200712204, "global_step": 55907, "epoch": 628, "lr": 9.003178334632641e-05}
{"train_loss": 0.2576635777950287, "global_step": 55908, "epoch": 628, "lr": 9.003143599357572e-05}
{"train_loss": 0.2821074426174164, "global_step": 55909, "epoch": 628, "lr": 9.003108863544328e-05}
{"train_loss": 0.18325963616371155, "global_step": 55910, "epoch": 628, "lr": 9.003074127192916e-05}
{"train_loss": 0.2651943862438202, "global_step": 55911, "epoch": 628, "lr": 9.003039390303339e-05}
{"train_loss": 0.2745307385921478, "global_step": 55912, "epoch": 628, "lr": 9.003004652875603e-05}
{"train_loss": 0.30020764470100403, "global_step": 55913, "epoch": 628, "lr": 9.00296991490971e-05}
{"train_loss": 0.24727027118206024, "global_step": 55914, "epoch": 628, "lr": 9.002935176405667e-05}
{"train_loss": 0.23226837813854218, "global_step": 55915, "epoch": 628, "lr": 9.00290043736348e-05}
{"train_loss": 0.3345496952533722, "global_step": 55916, "epoch": 628, "lr": 9.00286569778315e-05}
{"train_loss": 0.3049306273460388, "global_step": 55917, "epoch": 628, "lr": 9.002830957664684e-05}
{"train_loss": 0.23322656750679016, "global_step": 55918, "epoch": 628, "lr": 9.002796217008087e-05}
{"train_loss": 0.24160800874233246, "global_step": 55919, "epoch": 628, "lr": 9.002761475813361e-05}
{"train_loss": 0.21788786351680756, "global_step": 55920, "epoch": 628, "lr": 9.002726734080514e-05}
{"train_loss": 0.19015945494174957, "global_step": 55921, "epoch": 628, "lr": 9.002691991809548e-05}
{"train_loss": 0.18637865781784058, "global_step": 55922, "epoch": 628, "lr": 9.00265724900047e-05}
{"train_loss": 0.31320300698280334, "global_step": 55923, "epoch": 628, "lr": 9.002622505653283e-05}
{"train_loss": 0.20188261568546295, "global_step": 55924, "epoch": 628, "lr": 9.002587761767992e-05}
{"train_loss": 0.28984981775283813, "global_step": 55925, "epoch": 628, "lr": 9.002553017344603e-05}
{"train_loss": 0.19487792253494263, "global_step": 55926, "epoch": 628, "lr": 9.002518272383118e-05}
{"train_loss": 0.24950452148914337, "global_step": 55927, "epoch": 628, "lr": 9.002483526883544e-05}
{"train_loss": 0.24787931144237518, "global_step": 55928, "epoch": 628, "lr": 9.002448780845885e-05}
{"train_loss": 0.1553831398487091, "global_step": 55929, "epoch": 628, "lr": 9.002414034270145e-05}
{"train_loss": 0.23311640322208405, "global_step": 55930, "epoch": 628, "lr": 9.00237928715633e-05}
{"train_loss": 0.2921169698238373, "global_step": 55931, "epoch": 628, "lr": 9.002344539504443e-05}
{"train_loss": 0.27026090025901794, "global_step": 55932, "epoch": 628, "lr": 9.002309791314489e-05}
{"train_loss": 0.23433104157447815, "global_step": 55933, "epoch": 628, "lr": 9.002275042586475e-05}
{"train_loss": 0.24862389266490936, "global_step": 55934, "epoch": 628, "lr": 9.002240293320403e-05}
{"train_loss": 0.27352023124694824, "global_step": 55935, "epoch": 628, "lr": 9.002205543516278e-05}
{"train_loss": 0.3403535485267639, "global_step": 55936, "epoch": 628, "lr": 9.002170793174106e-05}
{"train_loss": 0.342463880777359, "global_step": 55937, "epoch": 628, "lr": 9.002136042293891e-05}
{"train_loss": 0.26068422198295593, "global_step": 55938, "epoch": 628, "lr": 9.002101290875638e-05}
{"train_loss": 0.34721237421035767, "global_step": 55939, "epoch": 628, "lr": 9.00206653891935e-05}
{"train_loss": 0.30943211913108826, "global_step": 55940, "epoch": 628, "lr": 9.002031786425034e-05}
{"train_loss": 0.20723876357078552, "global_step": 55941, "epoch": 628, "lr": 9.001997033392693e-05}
{"train_loss": 0.1535581350326538, "global_step": 55942, "epoch": 628, "lr": 9.001962279822332e-05}
{"train_loss": 0.2427416294813156, "global_step": 55943, "epoch": 628, "lr": 9.001927525713957e-05}
{"train_loss": 0.29095458984375, "global_step": 55944, "epoch": 628, "lr": 9.00189277106757e-05}
{"train_loss": 0.17430827021598816, "global_step": 55945, "epoch": 628, "lr": 9.001858015883179e-05}
{"train_loss": 0.30172085762023926, "global_step": 55946, "epoch": 628, "lr": 9.001823260160786e-05}
{"train_loss": 0.2627318203449249, "global_step": 55947, "epoch": 628, "lr": 9.001788503900397e-05}
{"train_loss": 0.34315112233161926, "global_step": 55948, "epoch": 628, "lr": 9.001753747102018e-05}
{"train_loss": 0.33739978075027466, "global_step": 55949, "epoch": 628, "lr": 9.001718989765648e-05}
{"train_loss": 0.3174741864204407, "global_step": 55950, "epoch": 628, "lr": 9.001684231891298e-05}
{"train_loss": 0.22202138602733612, "global_step": 55951, "epoch": 628, "lr": 9.00164947347897e-05}
{"train_loss": 0.32870960235595703, "global_step": 55952, "epoch": 628, "lr": 9.00161471452867e-05}
{"train_loss": 0.2953924834728241, "global_step": 55953, "epoch": 628, "lr": 9.0015799550404e-05}
{"train_loss": 0.34999018907546997, "global_step": 55954, "epoch": 628, "lr": 9.001545195014167e-05}
{"train_loss": 0.20009544491767883, "global_step": 55955, "epoch": 628, "lr": 9.001510434449975e-05}
{"train_loss": 0.30362874269485474, "global_step": 55956, "epoch": 628, "lr": 9.00147567334783e-05}
{"train_loss": 0.26652055978775024, "global_step": 55957, "epoch": 628, "lr": 9.001440911707734e-05}
{"train_loss": 0.12535324692726135, "global_step": 55958, "epoch": 628, "lr": 9.001406149529694e-05}
{"train_loss": 0.25717565417289734, "global_step": 55959, "epoch": 628, "lr": 9.001371386813712e-05}
{"train_loss": 0.3438512980937958, "global_step": 55960, "epoch": 628, "lr": 9.001336623559797e-05}
{"train_loss": 0.285231351852417, "global_step": 55961, "epoch": 628, "lr": 9.001301859767949e-05}
{"train_loss": 0.21523383259773254, "global_step": 55962, "epoch": 628, "lr": 9.001267095438175e-05}
{"train_loss": 0.25478050112724304, "global_step": 55963, "epoch": 628, "lr": 9.001232330570479e-05}
{"train_loss": 0.19510141015052795, "global_step": 55964, "epoch": 628, "lr": 9.001197565164867e-05}
{"train_loss": 0.2484508752822876, "global_step": 55965, "epoch": 628, "lr": 9.001162799221344e-05}
{"train_loss": 0.27936288714408875, "global_step": 55966, "epoch": 628, "lr": 9.00112803273991e-05}
{"train_loss": 0.27058354020118713, "global_step": 55967, "epoch": 628, "lr": 9.001093265720575e-05}
{"train_loss": 0.2540452480316162, "global_step": 55968, "epoch": 628, "lr": 9.001058498163342e-05}
{"train_loss": 0.28072890639305115, "global_step": 55969, "epoch": 628, "lr": 9.001023730068216e-05}
{"train_loss": 0.1752910017967224, "global_step": 55970, "epoch": 628, "lr": 9.000988961435199e-05}
{"train_loss": 0.3227008879184723, "global_step": 55971, "epoch": 628, "lr": 9.000954192264299e-05}
{"train_loss": 0.2942578196525574, "global_step": 55972, "epoch": 628, "lr": 9.00091942255552e-05}
{"train_loss": 0.28870031237602234, "global_step": 55973, "epoch": 628, "lr": 9.000884652308867e-05}
{"train_loss": 0.2774566113948822, "global_step": 55974, "epoch": 628, "lr": 9.000849881524342e-05}
{"train_loss": 0.3171825110912323, "global_step": 55975, "epoch": 628, "lr": 9.000815110201952e-05}
{"train_loss": 0.1946854591369629, "global_step": 55976, "epoch": 628, "lr": 9.000780338341701e-05}
{"train_loss": 0.20926854014396667, "global_step": 55977, "epoch": 628, "lr": 9.000745565943594e-05}
{"train_loss": 0.3121567666530609, "global_step": 55978, "epoch": 628, "lr": 9.000710793007636e-05}
{"train_loss": 0.22635115683078766, "global_step": 55979, "epoch": 628, "lr": 9.000676019533831e-05}
{"train_loss": 0.2609563967150249, "global_step": 55980, "epoch": 628, "lr": 9.000641245522184e-05, "val_loss": 3.304286479949951}
{"train_loss": 0.2914796471595764, "global_step": 55981, "epoch": 629, "lr": 9.000606470972699e-05}
{"train_loss": 0.27941375970840454, "global_step": 55982, "epoch": 629, "lr": 9.000571695885382e-05}
{"train_loss": 0.22002433240413666, "global_step": 55983, "epoch": 629, "lr": 9.000536920260236e-05}
{"train_loss": 0.2835921347141266, "global_step": 55984, "epoch": 629, "lr": 9.000502144097267e-05}
{"train_loss": 0.2105325609445572, "global_step": 55985, "epoch": 629, "lr": 9.000467367396479e-05}
{"train_loss": 0.27716580033302307, "global_step": 55986, "epoch": 629, "lr": 9.000432590157877e-05}
{"train_loss": 0.2630332112312317, "global_step": 55987, "epoch": 629, "lr": 9.000397812381466e-05}
{"train_loss": 0.31803637742996216, "global_step": 55988, "epoch": 629, "lr": 9.000363034067249e-05}
{"train_loss": 0.21347685158252716, "global_step": 55989, "epoch": 629, "lr": 9.000328255215233e-05}
{"train_loss": 0.1912376880645752, "global_step": 55990, "epoch": 629, "lr": 9.000293475825422e-05}
{"train_loss": 0.23568658530712128, "global_step": 55991, "epoch": 629, "lr": 9.00025869589782e-05}
{"train_loss": 0.2563086748123169, "global_step": 55992, "epoch": 629, "lr": 9.000223915432432e-05}
{"train_loss": 0.26852792501449585, "global_step": 55993, "epoch": 629, "lr": 9.000189134429263e-05}
{"train_loss": 0.19107945263385773, "global_step": 55994, "epoch": 629, "lr": 9.000154352888316e-05}
{"train_loss": 0.23181386291980743, "global_step": 55995, "epoch": 629, "lr": 9.000119570809598e-05}
{"train_loss": 0.20371633768081665, "global_step": 55996, "epoch": 629, "lr": 9.000084788193112e-05}
{"train_loss": 0.21828749775886536, "global_step": 55997, "epoch": 629, "lr": 9.000050005038865e-05}
{"train_loss": 0.2233644425868988, "global_step": 55998, "epoch": 629, "lr": 9.000015221346857e-05}
{"train_loss": 0.2525588572025299, "global_step": 55999, "epoch": 629, "lr": 8.999980437117099e-05}
{"train_loss": 0.35302603244781494, "global_step": 56000, "epoch": 629, "lr": 8.99994565234959e-05}
{"train_loss": 0.2568426728248596, "global_step": 56001, "epoch": 629, "lr": 8.999910867044339e-05}
{"train_loss": 0.2564208507537842, "global_step": 56002, "epoch": 629, "lr": 8.999876081201347e-05}
{"train_loss": 0.17416326701641083, "global_step": 56003, "epoch": 629, "lr": 8.999841294820622e-05}
{"train_loss": 0.2391129434108734, "global_step": 56004, "epoch": 629, "lr": 8.999806507902165e-05}
{"train_loss": 0.20779792964458466, "global_step": 56005, "epoch": 629, "lr": 8.999771720445984e-05}
{"train_loss": 0.2486031949520111, "global_step": 56006, "epoch": 629, "lr": 8.999736932452083e-05}
{"train_loss": 0.20976611971855164, "global_step": 56007, "epoch": 629, "lr": 8.999702143920466e-05}
{"train_loss": 0.18306341767311096, "global_step": 56008, "epoch": 629, "lr": 8.999667354851137e-05}
{"train_loss": 0.21810230612754822, "global_step": 56009, "epoch": 629, "lr": 8.999632565244102e-05}
{"train_loss": 0.2425583153963089, "global_step": 56010, "epoch": 629, "lr": 8.999597775099365e-05}
{"train_loss": 0.24721983075141907, "global_step": 56011, "epoch": 629, "lr": 8.99956298441693e-05}
{"train_loss": 0.20079757273197174, "global_step": 56012, "epoch": 629, "lr": 8.999528193196804e-05}
{"train_loss": 0.2662206292152405, "global_step": 56013, "epoch": 629, "lr": 8.99949340143899e-05}
{"train_loss": 0.1563374400138855, "global_step": 56014, "epoch": 629, "lr": 8.999458609143491e-05}
{"train_loss": 0.2704329788684845, "global_step": 56015, "epoch": 629, "lr": 8.999423816310315e-05}
{"train_loss": 0.34415102005004883, "global_step": 56016, "epoch": 629, "lr": 8.999389022939467e-05}
{"train_loss": 0.26169124245643616, "global_step": 56017, "epoch": 629, "lr": 8.999354229030948e-05}
{"train_loss": 0.3148793578147888, "global_step": 56018, "epoch": 629, "lr": 8.999319434584763e-05}
{"train_loss": 0.3228057622909546, "global_step": 56019, "epoch": 629, "lr": 8.99928463960092e-05}
{"train_loss": 0.26955923438072205, "global_step": 56020, "epoch": 629, "lr": 8.999249844079423e-05}
{"train_loss": 0.3099624514579773, "global_step": 56021, "epoch": 629, "lr": 8.999215048020274e-05}
{"train_loss": 0.250129759311676, "global_step": 56022, "epoch": 629, "lr": 8.999180251423478e-05}
{"train_loss": 0.3182331621646881, "global_step": 56023, "epoch": 629, "lr": 8.999145454289044e-05}
{"train_loss": 0.23046569526195526, "global_step": 56024, "epoch": 629, "lr": 8.999110656616972e-05}
{"train_loss": 0.3063317537307739, "global_step": 56025, "epoch": 629, "lr": 8.999075858407268e-05}
{"train_loss": 0.18892312049865723, "global_step": 56026, "epoch": 629, "lr": 8.999041059659939e-05}
{"train_loss": 0.28362664580345154, "global_step": 56027, "epoch": 629, "lr": 8.999006260374985e-05}
{"train_loss": 0.3129415810108185, "global_step": 56028, "epoch": 629, "lr": 8.998971460552415e-05}
{"train_loss": 0.29944002628326416, "global_step": 56029, "epoch": 629, "lr": 8.998936660192232e-05}
{"train_loss": 0.21111243963241577, "global_step": 56030, "epoch": 629, "lr": 8.99890185929444e-05}
{"train_loss": 0.2558538019657135, "global_step": 56031, "epoch": 629, "lr": 8.998867057859044e-05}
{"train_loss": 0.2709883749485016, "global_step": 56032, "epoch": 629, "lr": 8.998832255886049e-05}
{"train_loss": 0.24878433346748352, "global_step": 56033, "epoch": 629, "lr": 8.998797453375462e-05}
{"train_loss": 0.38574352860450745, "global_step": 56034, "epoch": 629, "lr": 8.998762650327284e-05}
{"train_loss": 0.31526151299476624, "global_step": 56035, "epoch": 629, "lr": 8.99872784674152e-05}
{"train_loss": 0.2621193826198578, "global_step": 56036, "epoch": 629, "lr": 8.998693042618177e-05}
{"train_loss": 0.19066932797431946, "global_step": 56037, "epoch": 629, "lr": 8.998658237957259e-05}
{"train_loss": 0.20851047337055206, "global_step": 56038, "epoch": 629, "lr": 8.998623432758768e-05}
{"train_loss": 0.2663288116455078, "global_step": 56039, "epoch": 629, "lr": 8.998588627022713e-05}
{"train_loss": 0.20638589560985565, "global_step": 56040, "epoch": 629, "lr": 8.998553820749094e-05}
{"train_loss": 0.42029818892478943, "global_step": 56041, "epoch": 629, "lr": 8.99851901393792e-05}
{"train_loss": 0.2643369138240814, "global_step": 56042, "epoch": 629, "lr": 8.998484206589195e-05}
{"train_loss": 0.32967981696128845, "global_step": 56043, "epoch": 629, "lr": 8.998449398702922e-05}
{"train_loss": 0.2550520896911621, "global_step": 56044, "epoch": 629, "lr": 8.998414590279105e-05}
{"train_loss": 0.24867098033428192, "global_step": 56045, "epoch": 629, "lr": 8.99837978131775e-05}
{"train_loss": 0.3196861147880554, "global_step": 56046, "epoch": 629, "lr": 8.998344971818861e-05}
{"train_loss": 0.30595362186431885, "global_step": 56047, "epoch": 629, "lr": 8.998310161782445e-05}
{"train_loss": 0.20010493695735931, "global_step": 56048, "epoch": 629, "lr": 8.998275351208505e-05}
{"train_loss": 0.23849570751190186, "global_step": 56049, "epoch": 629, "lr": 8.998240540097044e-05}
{"train_loss": 0.3081372082233429, "global_step": 56050, "epoch": 629, "lr": 8.998205728448069e-05}
{"train_loss": 0.188520148396492, "global_step": 56051, "epoch": 629, "lr": 8.998170916261583e-05}
{"train_loss": 0.2357311248779297, "global_step": 56052, "epoch": 629, "lr": 8.998136103537593e-05}
{"train_loss": 0.24647365510463715, "global_step": 56053, "epoch": 629, "lr": 8.998101290276102e-05}
{"train_loss": 0.2660472095012665, "global_step": 56054, "epoch": 629, "lr": 8.998066476477116e-05}
{"train_loss": 0.3535681664943695, "global_step": 56055, "epoch": 629, "lr": 8.998031662140638e-05}
{"train_loss": 0.25091439485549927, "global_step": 56056, "epoch": 629, "lr": 8.997996847266671e-05}
{"train_loss": 0.18040646612644196, "global_step": 56057, "epoch": 629, "lr": 8.997962031855225e-05}
{"train_loss": 0.28382495045661926, "global_step": 56058, "epoch": 629, "lr": 8.997927215906301e-05}
{"train_loss": 0.3197610378265381, "global_step": 56059, "epoch": 629, "lr": 8.997892399419904e-05}
{"train_loss": 0.19910043478012085, "global_step": 56060, "epoch": 629, "lr": 8.997857582396038e-05}
{"train_loss": 0.2506343424320221, "global_step": 56061, "epoch": 629, "lr": 8.997822764834711e-05}
{"train_loss": 0.2502638101577759, "global_step": 56062, "epoch": 629, "lr": 8.997787946735925e-05}
{"train_loss": 0.21971221268177032, "global_step": 56063, "epoch": 629, "lr": 8.997753128099685e-05}
{"train_loss": 0.25867146253585815, "global_step": 56064, "epoch": 629, "lr": 8.997718308925995e-05}
{"train_loss": 0.2521650195121765, "global_step": 56065, "epoch": 629, "lr": 8.997683489214861e-05}
{"train_loss": 0.2583039104938507, "global_step": 56066, "epoch": 629, "lr": 8.997648668966288e-05}
{"train_loss": 0.2722800374031067, "global_step": 56067, "epoch": 629, "lr": 8.997613848180279e-05}
{"train_loss": 0.20216260850429535, "global_step": 56068, "epoch": 629, "lr": 8.997579026856839e-05}
{"train_loss": 0.25625543781880583, "global_step": 56069, "epoch": 629, "lr": 8.997544204995974e-05, "val_loss": 3.1549220085144043}
{"train_loss": 0.21246594190597534, "global_step": 56070, "epoch": 630, "lr": 8.997509382597687e-05}
{"train_loss": 0.18025067448616028, "global_step": 56071, "epoch": 630, "lr": 8.997474559661985e-05}
{"train_loss": 0.23953409492969513, "global_step": 56072, "epoch": 630, "lr": 8.99743973618887e-05}
{"train_loss": 0.16311150789260864, "global_step": 56073, "epoch": 630, "lr": 8.99740491217835e-05}
{"train_loss": 0.24858590960502625, "global_step": 56074, "epoch": 630, "lr": 8.997370087630426e-05}
{"train_loss": 0.31253060698509216, "global_step": 56075, "epoch": 630, "lr": 8.997335262545103e-05}
{"train_loss": 0.3476039171218872, "global_step": 56076, "epoch": 630, "lr": 8.99730043692239e-05}
{"train_loss": 0.24378782510757446, "global_step": 56077, "epoch": 630, "lr": 8.997265610762286e-05}
{"train_loss": 0.17249305546283722, "global_step": 56078, "epoch": 630, "lr": 8.997230784064799e-05}
{"train_loss": 0.20068958401679993, "global_step": 56079, "epoch": 630, "lr": 8.997195956829934e-05}
{"train_loss": 0.20168665051460266, "global_step": 56080, "epoch": 630, "lr": 8.997161129057694e-05}
{"train_loss": 0.279529333114624, "global_step": 56081, "epoch": 630, "lr": 8.997126300748085e-05}
{"train_loss": 0.2236204445362091, "global_step": 56082, "epoch": 630, "lr": 8.997091471901111e-05}
{"train_loss": 0.2616710066795349, "global_step": 56083, "epoch": 630, "lr": 8.997056642516776e-05}
{"train_loss": 0.24534335732460022, "global_step": 56084, "epoch": 630, "lr": 8.997021812595086e-05}
{"train_loss": 0.24524375796318054, "global_step": 56085, "epoch": 630, "lr": 8.996986982136047e-05}
{"train_loss": 0.19347865879535675, "global_step": 56086, "epoch": 630, "lr": 8.996952151139659e-05}
{"train_loss": 0.19815528392791748, "global_step": 56087, "epoch": 630, "lr": 8.996917319605929e-05}
{"train_loss": 0.24534514546394348, "global_step": 56088, "epoch": 630, "lr": 8.996882487534864e-05}
{"train_loss": 0.218724325299263, "global_step": 56089, "epoch": 630, "lr": 8.996847654926467e-05}
{"train_loss": 0.236414834856987, "global_step": 56090, "epoch": 630, "lr": 8.996812821780743e-05}
{"train_loss": 0.318238765001297, "global_step": 56091, "epoch": 630, "lr": 8.996777988097695e-05}
{"train_loss": 0.4021894335746765, "global_step": 56092, "epoch": 630, "lr": 8.996743153877329e-05}
{"train_loss": 0.3251384198665619, "global_step": 56093, "epoch": 630, "lr": 8.996708319119649e-05}
{"train_loss": 0.1964791715145111, "global_step": 56094, "epoch": 630, "lr": 8.996673483824662e-05}
{"train_loss": 0.23312917351722717, "global_step": 56095, "epoch": 630, "lr": 8.996638647992369e-05}
{"train_loss": 0.18645790219306946, "global_step": 56096, "epoch": 630, "lr": 8.996603811622778e-05}
{"train_loss": 0.36401915550231934, "global_step": 56097, "epoch": 630, "lr": 8.996568974715892e-05}
{"train_loss": 0.2755928635597229, "global_step": 56098, "epoch": 630, "lr": 8.996534137271716e-05}
{"train_loss": 0.23170402646064758, "global_step": 56099, "epoch": 630, "lr": 8.996499299290255e-05}
{"train_loss": 0.2925543189048767, "global_step": 56100, "epoch": 630, "lr": 8.996464460771513e-05}
{"train_loss": 0.24719004333019257, "global_step": 56101, "epoch": 630, "lr": 8.996429621715495e-05}
{"train_loss": 0.19881707429885864, "global_step": 56102, "epoch": 630, "lr": 8.996394782122207e-05}
{"train_loss": 0.1695461869239807, "global_step": 56103, "epoch": 630, "lr": 8.99635994199165e-05}
{"train_loss": 0.29297372698783875, "global_step": 56104, "epoch": 630, "lr": 8.996325101323833e-05}
{"train_loss": 0.2198801040649414, "global_step": 56105, "epoch": 630, "lr": 8.996290260118759e-05}
{"train_loss": 0.2765624225139618, "global_step": 56106, "epoch": 630, "lr": 8.99625541837643e-05}
{"train_loss": 0.3197185695171356, "global_step": 56107, "epoch": 630, "lr": 8.996220576096856e-05}
{"train_loss": 0.1787981241941452, "global_step": 56108, "epoch": 630, "lr": 8.996185733280038e-05}
{"train_loss": 0.22850771248340607, "global_step": 56109, "epoch": 630, "lr": 8.996150889925981e-05}
{"train_loss": 0.22899556159973145, "global_step": 56110, "epoch": 630, "lr": 8.99611604603469e-05}
{"train_loss": 0.18242546916007996, "global_step": 56111, "epoch": 630, "lr": 8.99608120160617e-05}
{"train_loss": 0.2541850209236145, "global_step": 56112, "epoch": 630, "lr": 8.996046356640426e-05}
{"train_loss": 0.25214195251464844, "global_step": 56113, "epoch": 630, "lr": 8.996011511137461e-05}
{"train_loss": 0.2381167858839035, "global_step": 56114, "epoch": 630, "lr": 8.995976665097283e-05}
{"train_loss": 0.3324412703514099, "global_step": 56115, "epoch": 630, "lr": 8.995941818519894e-05}
{"train_loss": 0.27327781915664673, "global_step": 56116, "epoch": 630, "lr": 8.9959069714053e-05}
{"train_loss": 0.3507128357887268, "global_step": 56117, "epoch": 630, "lr": 8.995872123753504e-05}
{"train_loss": 0.23970812559127808, "global_step": 56118, "epoch": 630, "lr": 8.995837275564511e-05}
{"train_loss": 0.1590978354215622, "global_step": 56119, "epoch": 630, "lr": 8.995802426838328e-05}
{"train_loss": 0.24080540239810944, "global_step": 56120, "epoch": 630, "lr": 8.995767577574957e-05}
{"train_loss": 0.2296833097934723, "global_step": 56121, "epoch": 630, "lr": 8.995732727774404e-05}
{"train_loss": 0.28856536746025085, "global_step": 56122, "epoch": 630, "lr": 8.995697877436674e-05}
{"train_loss": 0.34428632259368896, "global_step": 56123, "epoch": 630, "lr": 8.995663026561771e-05}
{"train_loss": 0.3474501669406891, "global_step": 56124, "epoch": 630, "lr": 8.995628175149699e-05}
{"train_loss": 0.2677873373031616, "global_step": 56125, "epoch": 630, "lr": 8.995593323200464e-05}
{"train_loss": 0.2424067109823227, "global_step": 56126, "epoch": 630, "lr": 8.99555847071407e-05}
{"train_loss": 0.31993457674980164, "global_step": 56127, "epoch": 630, "lr": 8.995523617690522e-05}
{"train_loss": 0.27412495017051697, "global_step": 56128, "epoch": 630, "lr": 8.995488764129825e-05}
{"train_loss": 0.21590949594974518, "global_step": 56129, "epoch": 630, "lr": 8.995453910031982e-05}
{"train_loss": 0.36919841170310974, "global_step": 56130, "epoch": 630, "lr": 8.995419055397e-05}
{"train_loss": 0.23979021608829498, "global_step": 56131, "epoch": 630, "lr": 8.99538420022488e-05}
{"train_loss": 0.2636057138442993, "global_step": 56132, "epoch": 630, "lr": 8.995349344515633e-05}
{"train_loss": 0.2055283486843109, "global_step": 56133, "epoch": 630, "lr": 8.995314488269259e-05}
{"train_loss": 0.28898948431015015, "global_step": 56134, "epoch": 630, "lr": 8.995279631485762e-05}
{"train_loss": 0.2548324763774872, "global_step": 56135, "epoch": 630, "lr": 8.99524477416515e-05}
{"train_loss": 0.25424373149871826, "global_step": 56136, "epoch": 630, "lr": 8.995209916307426e-05}
{"train_loss": 0.31293603777885437, "global_step": 56137, "epoch": 630, "lr": 8.995175057912594e-05}
{"train_loss": 0.24630685150623322, "global_step": 56138, "epoch": 630, "lr": 8.99514019898066e-05}
{"train_loss": 0.2712598145008087, "global_step": 56139, "epoch": 630, "lr": 8.99510533951163e-05}
{"train_loss": 0.2089749425649643, "global_step": 56140, "epoch": 630, "lr": 8.995070479505503e-05}
{"train_loss": 0.19802553951740265, "global_step": 56141, "epoch": 630, "lr": 8.995035618962289e-05}
{"train_loss": 0.32533931732177734, "global_step": 56142, "epoch": 630, "lr": 8.99500075788199e-05}
{"train_loss": 0.25427308678627014, "global_step": 56143, "epoch": 630, "lr": 8.994965896264613e-05}
{"train_loss": 0.30350226163864136, "global_step": 56144, "epoch": 630, "lr": 8.994931034110163e-05}
{"train_loss": 0.2830972373485565, "global_step": 56145, "epoch": 630, "lr": 8.994896171418641e-05}
{"train_loss": 0.3081972301006317, "global_step": 56146, "epoch": 630, "lr": 8.994861308190054e-05}
{"train_loss": 0.2626442313194275, "global_step": 56147, "epoch": 630, "lr": 8.994826444424409e-05}
{"train_loss": 0.2540665566921234, "global_step": 56148, "epoch": 630, "lr": 8.994791580121706e-05}
{"train_loss": 0.19954751431941986, "global_step": 56149, "epoch": 630, "lr": 8.994756715281953e-05}
{"train_loss": 0.3087345361709595, "global_step": 56150, "epoch": 630, "lr": 8.994721849905152e-05}
{"train_loss": 0.2491982877254486, "global_step": 56151, "epoch": 630, "lr": 8.994686983991312e-05}
{"train_loss": 0.35725152492523193, "global_step": 56152, "epoch": 630, "lr": 8.994652117540432e-05}
{"train_loss": 0.2810702919960022, "global_step": 56153, "epoch": 630, "lr": 8.994617250552521e-05}
{"train_loss": 0.2589367926120758, "global_step": 56154, "epoch": 630, "lr": 8.994582383027583e-05}
{"train_loss": 0.23826289176940918, "global_step": 56155, "epoch": 630, "lr": 8.994547514965623e-05}
{"train_loss": 0.31846195459365845, "global_step": 56156, "epoch": 630, "lr": 8.994512646366642e-05}
{"train_loss": 0.23774726688861847, "global_step": 56157, "epoch": 630, "lr": 8.99447777723065e-05}
{"train_loss": 0.2578945653827003, "global_step": 56158, "epoch": 630, "lr": 8.994442907557648e-05, "val_loss": 3.1425039768218994, "train_action_mse_error": 18.14788055419922}
{"train_loss": 0.22277317941188812, "global_step": 56159, "epoch": 631, "lr": 8.994408037347642e-05}
{"train_loss": 0.23990599811077118, "global_step": 56160, "epoch": 631, "lr": 8.994373166600636e-05}
{"train_loss": 0.32108449935913086, "global_step": 56161, "epoch": 631, "lr": 8.994338295316636e-05}
{"train_loss": 0.28016096353530884, "global_step": 56162, "epoch": 631, "lr": 8.994303423495645e-05}
{"train_loss": 0.23730207979679108, "global_step": 56163, "epoch": 631, "lr": 8.99426855113767e-05}
{"train_loss": 0.2583823800086975, "global_step": 56164, "epoch": 631, "lr": 8.994233678242713e-05}
{"train_loss": 0.23215222358703613, "global_step": 56165, "epoch": 631, "lr": 8.99419880481078e-05}
{"train_loss": 0.30054959654808044, "global_step": 56166, "epoch": 631, "lr": 8.994163930841877e-05}
{"train_loss": 0.3659409284591675, "global_step": 56167, "epoch": 631, "lr": 8.994129056336006e-05}
{"train_loss": 0.3304198682308197, "global_step": 56168, "epoch": 631, "lr": 8.994094181293173e-05}
{"train_loss": 0.31701430678367615, "global_step": 56169, "epoch": 631, "lr": 8.994059305713382e-05}
{"train_loss": 0.2506019175052643, "global_step": 56170, "epoch": 631, "lr": 8.994024429596641e-05}
{"train_loss": 0.28154832124710083, "global_step": 56171, "epoch": 631, "lr": 8.993989552942951e-05}
{"train_loss": 0.19786445796489716, "global_step": 56172, "epoch": 631, "lr": 8.993954675752316e-05}
{"train_loss": 0.21592538058757782, "global_step": 56173, "epoch": 631, "lr": 8.993919798024745e-05}
{"train_loss": 0.32450559735298157, "global_step": 56174, "epoch": 631, "lr": 8.993884919760239e-05}
{"train_loss": 0.2196846306324005, "global_step": 56175, "epoch": 631, "lr": 8.993850040958804e-05}
{"train_loss": 0.20267702639102936, "global_step": 56176, "epoch": 631, "lr": 8.993815161620445e-05}
{"train_loss": 0.33393949270248413, "global_step": 56177, "epoch": 631, "lr": 8.993780281745165e-05}
{"train_loss": 0.23618243634700775, "global_step": 56178, "epoch": 631, "lr": 8.993745401332972e-05}
{"train_loss": 0.22739438712596893, "global_step": 56179, "epoch": 631, "lr": 8.993710520383867e-05}
{"train_loss": 0.23311296105384827, "global_step": 56180, "epoch": 631, "lr": 8.993675638897857e-05}
{"train_loss": 0.2539496123790741, "global_step": 56181, "epoch": 631, "lr": 8.993640756874943e-05}
{"train_loss": 0.3032926023006439, "global_step": 56182, "epoch": 631, "lr": 8.993605874315136e-05}
{"train_loss": 0.23370906710624695, "global_step": 56183, "epoch": 631, "lr": 8.993570991218436e-05}
{"train_loss": 0.191355898976326, "global_step": 56184, "epoch": 631, "lr": 8.99353610758485e-05}
{"train_loss": 0.2836269438266754, "global_step": 56185, "epoch": 631, "lr": 8.993501223414382e-05}
{"train_loss": 0.2978374660015106, "global_step": 56186, "epoch": 631, "lr": 8.993466338707036e-05}
{"train_loss": 0.2856614589691162, "global_step": 56187, "epoch": 631, "lr": 8.993431453462816e-05}
{"train_loss": 0.2180560678243637, "global_step": 56188, "epoch": 631, "lr": 8.993396567681729e-05}
{"train_loss": 0.26803669333457947, "global_step": 56189, "epoch": 631, "lr": 8.993361681363779e-05}
{"train_loss": 0.31898945569992065, "global_step": 56190, "epoch": 631, "lr": 8.99332679450897e-05}
{"train_loss": 0.23393958806991577, "global_step": 56191, "epoch": 631, "lr": 8.993291907117305e-05}
{"train_loss": 0.3346233367919922, "global_step": 56192, "epoch": 631, "lr": 8.993257019188792e-05}
{"train_loss": 0.24709032475948334, "global_step": 56193, "epoch": 631, "lr": 8.993222130723433e-05}
{"train_loss": 0.21353164315223694, "global_step": 56194, "epoch": 631, "lr": 8.993187241721236e-05}
{"train_loss": 0.17964331805706024, "global_step": 56195, "epoch": 631, "lr": 8.993152352182203e-05}
{"train_loss": 0.238285630941391, "global_step": 56196, "epoch": 631, "lr": 8.99311746210634e-05}
{"train_loss": 0.2579531669616699, "global_step": 56197, "epoch": 631, "lr": 8.993082571493649e-05}
{"train_loss": 0.16969306766986847, "global_step": 56198, "epoch": 631, "lr": 8.993047680344138e-05}
{"train_loss": 0.1652994453907013, "global_step": 56199, "epoch": 631, "lr": 8.99301278865781e-05}
{"train_loss": 0.21393455564975739, "global_step": 56200, "epoch": 631, "lr": 8.99297789643467e-05}
{"train_loss": 0.2951950430870056, "global_step": 56201, "epoch": 631, "lr": 8.992943003674723e-05}
{"train_loss": 0.24571402370929718, "global_step": 56202, "epoch": 631, "lr": 8.992908110377974e-05}
{"train_loss": 0.3043045401573181, "global_step": 56203, "epoch": 631, "lr": 8.992873216544426e-05}
{"train_loss": 0.21093159914016724, "global_step": 56204, "epoch": 631, "lr": 8.992838322174086e-05}
{"train_loss": 0.2795928716659546, "global_step": 56205, "epoch": 631, "lr": 8.992803427266958e-05}
{"train_loss": 0.30034559965133667, "global_step": 56206, "epoch": 631, "lr": 8.992768531823046e-05}
{"train_loss": 0.21958817541599274, "global_step": 56207, "epoch": 631, "lr": 8.992733635842354e-05}
{"train_loss": 0.24359793961048126, "global_step": 56208, "epoch": 631, "lr": 8.992698739324888e-05}
{"train_loss": 0.20768100023269653, "global_step": 56209, "epoch": 631, "lr": 8.992663842270652e-05}
{"train_loss": 0.19457130134105682, "global_step": 56210, "epoch": 631, "lr": 8.99262894467965e-05}
{"train_loss": 0.1825157105922699, "global_step": 56211, "epoch": 631, "lr": 8.99259404655189e-05}
{"train_loss": 0.2544456720352173, "global_step": 56212, "epoch": 631, "lr": 8.992559147887374e-05}
{"train_loss": 0.23836715519428253, "global_step": 56213, "epoch": 631, "lr": 8.992524248686105e-05}
{"train_loss": 0.25226056575775146, "global_step": 56214, "epoch": 631, "lr": 8.992489348948092e-05}
{"train_loss": 0.2729766070842743, "global_step": 56215, "epoch": 631, "lr": 8.992454448673337e-05}
{"train_loss": 0.15953272581100464, "global_step": 56216, "epoch": 631, "lr": 8.992419547861846e-05}
{"train_loss": 0.25365501642227173, "global_step": 56217, "epoch": 631, "lr": 8.992384646513621e-05}
{"train_loss": 0.22152115404605865, "global_step": 56218, "epoch": 631, "lr": 8.99234974462867e-05}
{"train_loss": 0.2214203178882599, "global_step": 56219, "epoch": 631, "lr": 8.992314842206995e-05}
{"train_loss": 0.35552793741226196, "global_step": 56220, "epoch": 631, "lr": 8.992279939248603e-05}
{"train_loss": 0.39607372879981995, "global_step": 56221, "epoch": 631, "lr": 8.992245035753495e-05}
{"train_loss": 0.3470924198627472, "global_step": 56222, "epoch": 631, "lr": 8.992210131721681e-05}
{"train_loss": 0.23640775680541992, "global_step": 56223, "epoch": 631, "lr": 8.992175227153163e-05}
{"train_loss": 0.29068470001220703, "global_step": 56224, "epoch": 631, "lr": 8.992140322047945e-05}
{"train_loss": 0.25548112392425537, "global_step": 56225, "epoch": 631, "lr": 8.992105416406033e-05}
{"train_loss": 0.2548674941062927, "global_step": 56226, "epoch": 631, "lr": 8.99207051022743e-05}
{"train_loss": 0.25438186526298523, "global_step": 56227, "epoch": 631, "lr": 8.992035603512142e-05}
{"train_loss": 0.3678564131259918, "global_step": 56228, "epoch": 631, "lr": 8.992000696260174e-05}
{"train_loss": 0.2107946276664734, "global_step": 56229, "epoch": 631, "lr": 8.99196578847153e-05}
{"train_loss": 0.19874024391174316, "global_step": 56230, "epoch": 631, "lr": 8.991930880146215e-05}
{"train_loss": 0.23712685704231262, "global_step": 56231, "epoch": 631, "lr": 8.991895971284233e-05}
{"train_loss": 0.1956072598695755, "global_step": 56232, "epoch": 631, "lr": 8.99186106188559e-05}
{"train_loss": 0.22688911855220795, "global_step": 56233, "epoch": 631, "lr": 8.991826151950289e-05}
{"train_loss": 0.21086576581001282, "global_step": 56234, "epoch": 631, "lr": 8.991791241478336e-05}
{"train_loss": 0.27215760946273804, "global_step": 56235, "epoch": 631, "lr": 8.991756330469736e-05}
{"train_loss": 0.1997130811214447, "global_step": 56236, "epoch": 631, "lr": 8.991721418924493e-05}
{"train_loss": 0.2902029752731323, "global_step": 56237, "epoch": 631, "lr": 8.991686506842611e-05}
{"train_loss": 0.2692740857601166, "global_step": 56238, "epoch": 631, "lr": 8.991651594224095e-05}
{"train_loss": 0.3242839276790619, "global_step": 56239, "epoch": 631, "lr": 8.991616681068951e-05}
{"train_loss": 0.3444026708602905, "global_step": 56240, "epoch": 631, "lr": 8.991581767377183e-05}
{"train_loss": 0.261593759059906, "global_step": 56241, "epoch": 631, "lr": 8.991546853148795e-05}
{"train_loss": 0.1623769849538803, "global_step": 56242, "epoch": 631, "lr": 8.991511938383792e-05}
{"train_loss": 0.3156379461288452, "global_step": 56243, "epoch": 631, "lr": 8.99147702308218e-05}
{"train_loss": 0.2616141140460968, "global_step": 56244, "epoch": 631, "lr": 8.991442107243962e-05}
{"train_loss": 0.1859324872493744, "global_step": 56245, "epoch": 631, "lr": 8.991407190869141e-05}
{"train_loss": 0.24037396907806396, "global_step": 56246, "epoch": 631, "lr": 8.991372273957727e-05}
{"train_loss": 0.2552344378795517, "global_step": 56247, "epoch": 631, "lr": 8.991337356509721e-05, "val_loss": 3.1771676540374756}
{"train_loss": 0.30585506558418274, "global_step": 56248, "epoch": 632, "lr": 8.991302438525127e-05}
{"train_loss": 0.19823656976222992, "global_step": 56249, "epoch": 632, "lr": 8.991267520003952e-05}
{"train_loss": 0.2501133382320404, "global_step": 56250, "epoch": 632, "lr": 8.991232600946198e-05}
{"train_loss": 0.21298393607139587, "global_step": 56251, "epoch": 632, "lr": 8.991197681351874e-05}
{"train_loss": 0.2335895448923111, "global_step": 56252, "epoch": 632, "lr": 8.991162761220982e-05}
{"train_loss": 0.1708691567182541, "global_step": 56253, "epoch": 632, "lr": 8.991127840553527e-05}
{"train_loss": 0.30083951354026794, "global_step": 56254, "epoch": 632, "lr": 8.991092919349511e-05}
{"train_loss": 0.24813377857208252, "global_step": 56255, "epoch": 632, "lr": 8.991057997608942e-05}
{"train_loss": 0.21826981008052826, "global_step": 56256, "epoch": 632, "lr": 8.991023075331825e-05}
{"train_loss": 0.26059937477111816, "global_step": 56257, "epoch": 632, "lr": 8.990988152518162e-05}
{"train_loss": 0.334671288728714, "global_step": 56258, "epoch": 632, "lr": 8.990953229167962e-05}
{"train_loss": 0.23397791385650635, "global_step": 56259, "epoch": 632, "lr": 8.990918305281224e-05}
{"train_loss": 0.23049268126487732, "global_step": 56260, "epoch": 632, "lr": 8.990883380857957e-05}
{"train_loss": 0.25236043334007263, "global_step": 56261, "epoch": 632, "lr": 8.990848455898164e-05}
{"train_loss": 0.22755467891693115, "global_step": 56262, "epoch": 632, "lr": 8.990813530401851e-05}
{"train_loss": 0.29414066672325134, "global_step": 56263, "epoch": 632, "lr": 8.990778604369023e-05}
{"train_loss": 0.2020585983991623, "global_step": 56264, "epoch": 632, "lr": 8.99074367779968e-05}
{"train_loss": 0.21384108066558838, "global_step": 56265, "epoch": 632, "lr": 8.990708750693832e-05}
{"train_loss": 0.22573502361774445, "global_step": 56266, "epoch": 632, "lr": 8.990673823051483e-05}
{"train_loss": 0.24127154052257538, "global_step": 56267, "epoch": 632, "lr": 8.990638894872634e-05}
{"train_loss": 0.34423941373825073, "global_step": 56268, "epoch": 632, "lr": 8.990603966157294e-05}
{"train_loss": 0.2041199803352356, "global_step": 56269, "epoch": 632, "lr": 8.990569036905465e-05}
{"train_loss": 0.22574764490127563, "global_step": 56270, "epoch": 632, "lr": 8.990534107117154e-05}
{"train_loss": 0.273064523935318, "global_step": 56271, "epoch": 632, "lr": 8.990499176792363e-05}
{"train_loss": 0.19242000579833984, "global_step": 56272, "epoch": 632, "lr": 8.990464245931098e-05}
{"train_loss": 0.2557935416698456, "global_step": 56273, "epoch": 632, "lr": 8.990429314533364e-05}
{"train_loss": 0.2572750151157379, "global_step": 56274, "epoch": 632, "lr": 8.990394382599166e-05}
{"train_loss": 0.2843610644340515, "global_step": 56275, "epoch": 632, "lr": 8.990359450128507e-05}
{"train_loss": 0.2753438651561737, "global_step": 56276, "epoch": 632, "lr": 8.990324517121393e-05}
{"train_loss": 0.235540509223938, "global_step": 56277, "epoch": 632, "lr": 8.99028958357783e-05}
{"train_loss": 0.2623668313026428, "global_step": 56278, "epoch": 632, "lr": 8.99025464949782e-05}
{"train_loss": 0.31474342942237854, "global_step": 56279, "epoch": 632, "lr": 8.99021971488137e-05}
{"train_loss": 0.3341694176197052, "global_step": 56280, "epoch": 632, "lr": 8.990184779728481e-05}
{"train_loss": 0.2810479402542114, "global_step": 56281, "epoch": 632, "lr": 8.990149844039163e-05}
{"train_loss": 0.2753731906414032, "global_step": 56282, "epoch": 632, "lr": 8.990114907813417e-05}
{"train_loss": 0.20552852749824524, "global_step": 56283, "epoch": 632, "lr": 8.990079971051248e-05}
{"train_loss": 0.3033163845539093, "global_step": 56284, "epoch": 632, "lr": 8.990045033752663e-05}
{"train_loss": 0.3296714425086975, "global_step": 56285, "epoch": 632, "lr": 8.990010095917664e-05}
{"train_loss": 0.2429628074169159, "global_step": 56286, "epoch": 632, "lr": 8.989975157546257e-05}
{"train_loss": 0.33155572414398193, "global_step": 56287, "epoch": 632, "lr": 8.989940218638446e-05}
{"train_loss": 0.23260369896888733, "global_step": 56288, "epoch": 632, "lr": 8.989905279194237e-05}
{"train_loss": 0.26020723581314087, "global_step": 56289, "epoch": 632, "lr": 8.989870339213632e-05}
{"train_loss": 0.2102736532688141, "global_step": 56290, "epoch": 632, "lr": 8.989835398696639e-05}
{"train_loss": 0.21155187487602234, "global_step": 56291, "epoch": 632, "lr": 8.989800457643263e-05}
{"train_loss": 0.3138587474822998, "global_step": 56292, "epoch": 632, "lr": 8.989765516053505e-05}
{"train_loss": 0.2639780044555664, "global_step": 56293, "epoch": 632, "lr": 8.989730573927371e-05}
{"train_loss": 0.2550296485424042, "global_step": 56294, "epoch": 632, "lr": 8.989695631264867e-05}
{"train_loss": 0.20392730832099915, "global_step": 56295, "epoch": 632, "lr": 8.989660688065998e-05}
{"train_loss": 0.24636465311050415, "global_step": 56296, "epoch": 632, "lr": 8.989625744330765e-05}
{"train_loss": 0.21255818009376526, "global_step": 56297, "epoch": 632, "lr": 8.989590800059178e-05}
{"train_loss": 0.22996525466442108, "global_step": 56298, "epoch": 632, "lr": 8.98955585525124e-05}
{"train_loss": 0.15719494223594666, "global_step": 56299, "epoch": 632, "lr": 8.989520909906953e-05}
{"train_loss": 0.2814842164516449, "global_step": 56300, "epoch": 632, "lr": 8.989485964026323e-05}
{"train_loss": 0.1796634942293167, "global_step": 56301, "epoch": 632, "lr": 8.989451017609355e-05}
{"train_loss": 0.359279990196228, "global_step": 56302, "epoch": 632, "lr": 8.989416070656056e-05}
{"train_loss": 0.2526235580444336, "global_step": 56303, "epoch": 632, "lr": 8.989381123166428e-05}
{"train_loss": 0.22120197117328644, "global_step": 56304, "epoch": 632, "lr": 8.989346175140476e-05}
{"train_loss": 0.21881848573684692, "global_step": 56305, "epoch": 632, "lr": 8.989311226578205e-05}
{"train_loss": 0.3480243980884552, "global_step": 56306, "epoch": 632, "lr": 8.98927627747962e-05}
{"train_loss": 0.2533903121948242, "global_step": 56307, "epoch": 632, "lr": 8.989241327844726e-05}
{"train_loss": 0.250015527009964, "global_step": 56308, "epoch": 632, "lr": 8.989206377673526e-05}
{"train_loss": 0.2893555760383606, "global_step": 56309, "epoch": 632, "lr": 8.989171426966026e-05}
{"train_loss": 0.16689427196979523, "global_step": 56310, "epoch": 632, "lr": 8.989136475722231e-05}
{"train_loss": 0.3055291175842285, "global_step": 56311, "epoch": 632, "lr": 8.989101523942144e-05}
{"train_loss": 0.15820759534835815, "global_step": 56312, "epoch": 632, "lr": 8.989066571625772e-05}
{"train_loss": 0.2414429932832718, "global_step": 56313, "epoch": 632, "lr": 8.989031618773118e-05}
{"train_loss": 0.44861364364624023, "global_step": 56314, "epoch": 632, "lr": 8.988996665384188e-05}
{"train_loss": 0.3723045885562897, "global_step": 56315, "epoch": 632, "lr": 8.988961711458986e-05}
{"train_loss": 0.14676548540592194, "global_step": 56316, "epoch": 632, "lr": 8.988926756997518e-05}
{"train_loss": 0.32278722524642944, "global_step": 56317, "epoch": 632, "lr": 8.988891801999785e-05}
{"train_loss": 0.26776960492134094, "global_step": 56318, "epoch": 632, "lr": 8.988856846465795e-05}
{"train_loss": 0.1789959967136383, "global_step": 56319, "epoch": 632, "lr": 8.988821890395551e-05}
{"train_loss": 0.2875368893146515, "global_step": 56320, "epoch": 632, "lr": 8.98878693378906e-05}
{"train_loss": 0.34457483887672424, "global_step": 56321, "epoch": 632, "lr": 8.988751976646324e-05}
{"train_loss": 0.21385417878627777, "global_step": 56322, "epoch": 632, "lr": 8.98871701896735e-05}
{"train_loss": 0.23695248365402222, "global_step": 56323, "epoch": 632, "lr": 8.988682060752141e-05}
{"train_loss": 0.28884488344192505, "global_step": 56324, "epoch": 632, "lr": 8.988647102000703e-05}
{"train_loss": 0.21081338822841644, "global_step": 56325, "epoch": 632, "lr": 8.98861214271304e-05}
{"train_loss": 0.3162718713283539, "global_step": 56326, "epoch": 632, "lr": 8.988577182889155e-05}
{"train_loss": 0.2854078412055969, "global_step": 56327, "epoch": 632, "lr": 8.988542222529056e-05}
{"train_loss": 0.2813083231449127, "global_step": 56328, "epoch": 632, "lr": 8.988507261632746e-05}
{"train_loss": 0.34655144810676575, "global_step": 56329, "epoch": 632, "lr": 8.98847230020023e-05}
{"train_loss": 0.2638205587863922, "global_step": 56330, "epoch": 632, "lr": 8.988437338231512e-05}
{"train_loss": 0.20296192169189453, "global_step": 56331, "epoch": 632, "lr": 8.988402375726597e-05}
{"train_loss": 0.2726644277572632, "global_step": 56332, "epoch": 632, "lr": 8.98836741268549e-05}
{"train_loss": 0.1891641467809677, "global_step": 56333, "epoch": 632, "lr": 8.988332449108196e-05}
{"train_loss": 0.23579561710357666, "global_step": 56334, "epoch": 632, "lr": 8.988297484994719e-05}
{"train_loss": 0.29546359181404114, "global_step": 56335, "epoch": 632, "lr": 8.988262520345063e-05}
{"train_loss": 0.25791304232029433, "global_step": 56336, "epoch": 632, "lr": 8.988227555159236e-05, "val_loss": 3.3685286045074463}
{"train_loss": 0.2278580516576767, "global_step": 56337, "epoch": 633, "lr": 8.988192589437239e-05}
{"train_loss": 0.17173409461975098, "global_step": 56338, "epoch": 633, "lr": 8.988157623179077e-05}
{"train_loss": 0.298078715801239, "global_step": 56339, "epoch": 633, "lr": 8.988122656384759e-05}
{"train_loss": 0.2855428457260132, "global_step": 56340, "epoch": 633, "lr": 8.988087689054283e-05}
{"train_loss": 0.2513378858566284, "global_step": 56341, "epoch": 633, "lr": 8.98805272118766e-05}
{"train_loss": 0.2268269807100296, "global_step": 56342, "epoch": 633, "lr": 8.98801775278489e-05}
{"train_loss": 0.2744119167327881, "global_step": 56343, "epoch": 633, "lr": 8.987982783845981e-05}
{"train_loss": 0.2739470303058624, "global_step": 56344, "epoch": 633, "lr": 8.987947814370936e-05}
{"train_loss": 0.2727756202220917, "global_step": 56345, "epoch": 633, "lr": 8.98791284435976e-05}
{"train_loss": 0.23825940489768982, "global_step": 56346, "epoch": 633, "lr": 8.987877873812457e-05}
{"train_loss": 0.24240367114543915, "global_step": 56347, "epoch": 633, "lr": 8.987842902729033e-05}
{"train_loss": 0.2581118941307068, "global_step": 56348, "epoch": 633, "lr": 8.98780793110949e-05}
{"train_loss": 0.25882676243782043, "global_step": 56349, "epoch": 633, "lr": 8.987772958953838e-05}
{"train_loss": 0.348615437746048, "global_step": 56350, "epoch": 633, "lr": 8.987737986262078e-05}
{"train_loss": 0.283966988325119, "global_step": 56351, "epoch": 633, "lr": 8.987703013034214e-05}
{"train_loss": 0.28470027446746826, "global_step": 56352, "epoch": 633, "lr": 8.987668039270252e-05}
{"train_loss": 0.2912668287754059, "global_step": 56353, "epoch": 633, "lr": 8.987633064970197e-05}
{"train_loss": 0.28871363401412964, "global_step": 56354, "epoch": 633, "lr": 8.987598090134056e-05}
{"train_loss": 0.20232278108596802, "global_step": 56355, "epoch": 633, "lr": 8.987563114761827e-05}
{"train_loss": 0.2764996290206909, "global_step": 56356, "epoch": 633, "lr": 8.987528138853521e-05}
{"train_loss": 0.31860795617103577, "global_step": 56357, "epoch": 633, "lr": 8.98749316240914e-05}
{"train_loss": 0.25327813625335693, "global_step": 56358, "epoch": 633, "lr": 8.987458185428688e-05}
{"train_loss": 0.3266168236732483, "global_step": 56359, "epoch": 633, "lr": 8.987423207912171e-05}
{"train_loss": 0.26211899518966675, "global_step": 56360, "epoch": 633, "lr": 8.987388229859594e-05}
{"train_loss": 0.21747906506061554, "global_step": 56361, "epoch": 633, "lr": 8.987353251270963e-05}
{"train_loss": 0.31366705894470215, "global_step": 56362, "epoch": 633, "lr": 8.98731827214628e-05}
{"train_loss": 0.2836190462112427, "global_step": 56363, "epoch": 633, "lr": 8.98728329248555e-05}
{"train_loss": 0.3923879861831665, "global_step": 56364, "epoch": 633, "lr": 8.987248312288779e-05}
{"train_loss": 0.28296589851379395, "global_step": 56365, "epoch": 633, "lr": 8.98721333155597e-05}
{"train_loss": 0.27985140681266785, "global_step": 56366, "epoch": 633, "lr": 8.987178350287129e-05}
{"train_loss": 0.31509703397750854, "global_step": 56367, "epoch": 633, "lr": 8.98714336848226e-05}
{"train_loss": 0.4099663197994232, "global_step": 56368, "epoch": 633, "lr": 8.98710838614137e-05}
{"train_loss": 0.28224828839302063, "global_step": 56369, "epoch": 633, "lr": 8.987073403264461e-05}
{"train_loss": 0.28647565841674805, "global_step": 56370, "epoch": 633, "lr": 8.987038419851537e-05}
{"train_loss": 0.3095632791519165, "global_step": 56371, "epoch": 633, "lr": 8.987003435902607e-05}
{"train_loss": 0.23799526691436768, "global_step": 56372, "epoch": 633, "lr": 8.986968451417671e-05}
{"train_loss": 0.24457484483718872, "global_step": 56373, "epoch": 633, "lr": 8.986933466396736e-05}
{"train_loss": 0.24505822360515594, "global_step": 56374, "epoch": 633, "lr": 8.986898480839807e-05}
{"train_loss": 0.20250672101974487, "global_step": 56375, "epoch": 633, "lr": 8.986863494746888e-05}
{"train_loss": 0.2925737798213959, "global_step": 56376, "epoch": 633, "lr": 8.986828508117984e-05}
{"train_loss": 0.30008965730667114, "global_step": 56377, "epoch": 633, "lr": 8.986793520953099e-05}
{"train_loss": 0.2119794487953186, "global_step": 56378, "epoch": 633, "lr": 8.986758533252239e-05}
{"train_loss": 0.23455175757408142, "global_step": 56379, "epoch": 633, "lr": 8.986723545015407e-05}
{"train_loss": 0.2842961549758911, "global_step": 56380, "epoch": 633, "lr": 8.986688556242608e-05}
{"train_loss": 0.22455433011054993, "global_step": 56381, "epoch": 633, "lr": 8.986653566933849e-05}
{"train_loss": 0.34438657760620117, "global_step": 56382, "epoch": 633, "lr": 8.986618577089132e-05}
{"train_loss": 0.2505161762237549, "global_step": 56383, "epoch": 633, "lr": 8.986583586708463e-05}
{"train_loss": 0.3306453227996826, "global_step": 56384, "epoch": 633, "lr": 8.986548595791846e-05}
{"train_loss": 0.327160120010376, "global_step": 56385, "epoch": 633, "lr": 8.986513604339287e-05}
{"train_loss": 0.21806679666042328, "global_step": 56386, "epoch": 633, "lr": 8.986478612350789e-05}
{"train_loss": 0.27697718143463135, "global_step": 56387, "epoch": 633, "lr": 8.986443619826358e-05}
{"train_loss": 0.19348418712615967, "global_step": 56388, "epoch": 633, "lr": 8.986408626765999e-05}
{"train_loss": 0.2977060377597809, "global_step": 56389, "epoch": 633, "lr": 8.986373633169713e-05}
{"train_loss": 0.24296943843364716, "global_step": 56390, "epoch": 633, "lr": 8.98633863903751e-05}
{"train_loss": 0.25625118613243103, "global_step": 56391, "epoch": 633, "lr": 8.986303644369393e-05}
{"train_loss": 0.32033950090408325, "global_step": 56392, "epoch": 633, "lr": 8.986268649165364e-05}
{"train_loss": 0.2666708827018738, "global_step": 56393, "epoch": 633, "lr": 8.986233653425432e-05}
{"train_loss": 0.24432294070720673, "global_step": 56394, "epoch": 633, "lr": 8.986198657149598e-05}
{"train_loss": 0.26671478152275085, "global_step": 56395, "epoch": 633, "lr": 8.986163660337868e-05}
{"train_loss": 0.28844353556632996, "global_step": 56396, "epoch": 633, "lr": 8.986128662990247e-05}
{"train_loss": 0.2677408456802368, "global_step": 56397, "epoch": 633, "lr": 8.986093665106741e-05}
{"train_loss": 0.2511347830295563, "global_step": 56398, "epoch": 633, "lr": 8.986058666687353e-05}
{"train_loss": 0.32513687014579773, "global_step": 56399, "epoch": 633, "lr": 8.986023667732086e-05}
{"train_loss": 0.2932913899421692, "global_step": 56400, "epoch": 633, "lr": 8.985988668240948e-05}
{"train_loss": 0.29913556575775146, "global_step": 56401, "epoch": 633, "lr": 8.985953668213944e-05}
{"train_loss": 0.25570183992385864, "global_step": 56402, "epoch": 633, "lr": 8.985918667651074e-05}
{"train_loss": 0.18861833214759827, "global_step": 56403, "epoch": 633, "lr": 8.985883666552348e-05}
{"train_loss": 0.2640194594860077, "global_step": 56404, "epoch": 633, "lr": 8.985848664917768e-05}
{"train_loss": 0.22203293442726135, "global_step": 56405, "epoch": 633, "lr": 8.985813662747339e-05}
{"train_loss": 0.2586043179035187, "global_step": 56406, "epoch": 633, "lr": 8.985778660041067e-05}
{"train_loss": 0.25239166617393494, "global_step": 56407, "epoch": 633, "lr": 8.985743656798953e-05}
{"train_loss": 0.389996737241745, "global_step": 56408, "epoch": 633, "lr": 8.985708653021007e-05}
{"train_loss": 0.1946900635957718, "global_step": 56409, "epoch": 633, "lr": 8.985673648707228e-05}
{"train_loss": 0.3301122486591339, "global_step": 56410, "epoch": 633, "lr": 8.985638643857628e-05}
{"train_loss": 0.14694002270698547, "global_step": 56411, "epoch": 633, "lr": 8.985603638472205e-05}
{"train_loss": 0.3313521444797516, "global_step": 56412, "epoch": 633, "lr": 8.985568632550966e-05}
{"train_loss": 0.12834058701992035, "global_step": 56413, "epoch": 633, "lr": 8.985533626093916e-05}
{"train_loss": 0.19152097404003143, "global_step": 56414, "epoch": 633, "lr": 8.98549861910106e-05}
{"train_loss": 0.21681398153305054, "global_step": 56415, "epoch": 633, "lr": 8.985463611572402e-05}
{"train_loss": 0.2833702862262726, "global_step": 56416, "epoch": 633, "lr": 8.985428603507948e-05}
{"train_loss": 0.2347470372915268, "global_step": 56417, "epoch": 633, "lr": 8.985393594907701e-05}
{"train_loss": 0.1931556910276413, "global_step": 56418, "epoch": 633, "lr": 8.985358585771666e-05}
{"train_loss": 0.25099480152130127, "global_step": 56419, "epoch": 633, "lr": 8.985323576099847e-05}
{"train_loss": 0.1955389380455017, "global_step": 56420, "epoch": 633, "lr": 8.985288565892251e-05}
{"train_loss": 0.24178370833396912, "global_step": 56421, "epoch": 633, "lr": 8.985253555148882e-05}
{"train_loss": 0.3434576392173767, "global_step": 56422, "epoch": 633, "lr": 8.985218543869744e-05}
{"train_loss": 0.2575950026512146, "global_step": 56423, "epoch": 633, "lr": 8.985183532054841e-05}
{"train_loss": 0.22791847586631775, "global_step": 56424, "epoch": 633, "lr": 8.985148519704181e-05}
{"train_loss": 0.2659379007776132, "global_step": 56425, "epoch": 633, "lr": 8.985113506817765e-05, "val_loss": 3.2801997661590576}
{"train_loss": 0.23588724434375763, "global_step": 56426, "epoch": 634, "lr": 8.985078493395598e-05}
{"train_loss": 0.3127976059913635, "global_step": 56427, "epoch": 634, "lr": 8.985043479437688e-05}
{"train_loss": 0.2497512698173523, "global_step": 56428, "epoch": 634, "lr": 8.985008464944036e-05}
{"train_loss": 0.192427396774292, "global_step": 56429, "epoch": 634, "lr": 8.984973449914648e-05}
{"train_loss": 0.22510501742362976, "global_step": 56430, "epoch": 634, "lr": 8.98493843434953e-05}
{"train_loss": 0.3016637861728668, "global_step": 56431, "epoch": 634, "lr": 8.984903418248685e-05}
{"train_loss": 0.26275503635406494, "global_step": 56432, "epoch": 634, "lr": 8.984868401612118e-05}
{"train_loss": 0.2637535333633423, "global_step": 56433, "epoch": 634, "lr": 8.984833384439833e-05}
{"train_loss": 0.25108590722084045, "global_step": 56434, "epoch": 634, "lr": 8.984798366731838e-05}
{"train_loss": 0.25937312841415405, "global_step": 56435, "epoch": 634, "lr": 8.984763348488134e-05}
{"train_loss": 0.26055264472961426, "global_step": 56436, "epoch": 634, "lr": 8.984728329708727e-05}
{"train_loss": 0.20144639909267426, "global_step": 56437, "epoch": 634, "lr": 8.984693310393623e-05}
{"train_loss": 0.32224488258361816, "global_step": 56438, "epoch": 634, "lr": 8.984658290542825e-05}
{"train_loss": 0.2360275834798813, "global_step": 56439, "epoch": 634, "lr": 8.984623270156339e-05}
{"train_loss": 0.17803417146205902, "global_step": 56440, "epoch": 634, "lr": 8.984588249234167e-05}
{"train_loss": 0.21536044776439667, "global_step": 56441, "epoch": 634, "lr": 8.984553227776317e-05}
{"train_loss": 0.24829508364200592, "global_step": 56442, "epoch": 634, "lr": 8.984518205782791e-05}
{"train_loss": 0.25215673446655273, "global_step": 56443, "epoch": 634, "lr": 8.984483183253597e-05}
{"train_loss": 0.27089497447013855, "global_step": 56444, "epoch": 634, "lr": 8.984448160188736e-05}
{"train_loss": 0.2332475483417511, "global_step": 56445, "epoch": 634, "lr": 8.984413136588216e-05}
{"train_loss": 0.20594921708106995, "global_step": 56446, "epoch": 634, "lr": 8.984378112452041e-05}
{"train_loss": 0.3601625859737396, "global_step": 56447, "epoch": 634, "lr": 8.984343087780213e-05}
{"train_loss": 0.2577207386493683, "global_step": 56448, "epoch": 634, "lr": 8.98430806257274e-05}
{"train_loss": 0.40133050084114075, "global_step": 56449, "epoch": 634, "lr": 8.984273036829625e-05}
{"train_loss": 0.3073362112045288, "global_step": 56450, "epoch": 634, "lr": 8.984238010550871e-05}
{"train_loss": 0.2472607046365738, "global_step": 56451, "epoch": 634, "lr": 8.984202983736488e-05}
{"train_loss": 0.2448650747537613, "global_step": 56452, "epoch": 634, "lr": 8.984167956386476e-05}
{"train_loss": 0.22245730459690094, "global_step": 56453, "epoch": 634, "lr": 8.984132928500841e-05}
{"train_loss": 0.2131521850824356, "global_step": 56454, "epoch": 634, "lr": 8.984097900079588e-05}
{"train_loss": 0.21491505205631256, "global_step": 56455, "epoch": 634, "lr": 8.98406287112272e-05}
{"train_loss": 0.2623390555381775, "global_step": 56456, "epoch": 634, "lr": 8.984027841630246e-05}
{"train_loss": 0.17677031457424164, "global_step": 56457, "epoch": 634, "lr": 8.983992811602166e-05}
{"train_loss": 0.2606028616428375, "global_step": 56458, "epoch": 634, "lr": 8.983957781038488e-05}
{"train_loss": 0.20318926870822906, "global_step": 56459, "epoch": 634, "lr": 8.983922749939216e-05}
{"train_loss": 0.2912999987602234, "global_step": 56460, "epoch": 634, "lr": 8.983887718304353e-05}
{"train_loss": 0.20774491131305695, "global_step": 56461, "epoch": 634, "lr": 8.983852686133905e-05}
{"train_loss": 0.3168136179447174, "global_step": 56462, "epoch": 634, "lr": 8.983817653427876e-05}
{"train_loss": 0.30505692958831787, "global_step": 56463, "epoch": 634, "lr": 8.983782620186272e-05}
{"train_loss": 0.1776895970106125, "global_step": 56464, "epoch": 634, "lr": 8.983747586409097e-05}
{"train_loss": 0.27752020955085754, "global_step": 56465, "epoch": 634, "lr": 8.983712552096356e-05}
{"train_loss": 0.2782704830169678, "global_step": 56466, "epoch": 634, "lr": 8.983677517248052e-05}
{"train_loss": 0.2730211913585663, "global_step": 56467, "epoch": 634, "lr": 8.983642481864193e-05}
{"train_loss": 0.21329839527606964, "global_step": 56468, "epoch": 634, "lr": 8.983607445944782e-05}
{"train_loss": 0.25921595096588135, "global_step": 56469, "epoch": 634, "lr": 8.983572409489821e-05}
{"train_loss": 0.31700536608695984, "global_step": 56470, "epoch": 634, "lr": 8.983537372499318e-05}
{"train_loss": 0.20987102389335632, "global_step": 56471, "epoch": 634, "lr": 8.983502334973279e-05}
{"train_loss": 0.25744080543518066, "global_step": 56472, "epoch": 634, "lr": 8.983467296911705e-05}
{"train_loss": 0.36462050676345825, "global_step": 56473, "epoch": 634, "lr": 8.983432258314603e-05}
{"train_loss": 0.19324566423892975, "global_step": 56474, "epoch": 634, "lr": 8.983397219181977e-05}
{"train_loss": 0.31025275588035583, "global_step": 56475, "epoch": 634, "lr": 8.98336217951383e-05}
{"train_loss": 0.2972894012928009, "global_step": 56476, "epoch": 634, "lr": 8.983327139310171e-05}
{"train_loss": 0.23806315660476685, "global_step": 56477, "epoch": 634, "lr": 8.983292098571001e-05}
{"train_loss": 0.3155192732810974, "global_step": 56478, "epoch": 634, "lr": 8.983257057296326e-05}
{"train_loss": 0.2906789779663086, "global_step": 56479, "epoch": 634, "lr": 8.983222015486151e-05}
{"train_loss": 0.25956737995147705, "global_step": 56480, "epoch": 634, "lr": 8.98318697314048e-05}
{"train_loss": 0.2737407684326172, "global_step": 56481, "epoch": 634, "lr": 8.98315193025932e-05}
{"train_loss": 0.2543257772922516, "global_step": 56482, "epoch": 634, "lr": 8.98311688684267e-05}
{"train_loss": 0.2773703336715698, "global_step": 56483, "epoch": 634, "lr": 8.983081842890542e-05}
{"train_loss": 0.3634217083454132, "global_step": 56484, "epoch": 634, "lr": 8.983046798402935e-05}
{"train_loss": 0.2825269103050232, "global_step": 56485, "epoch": 634, "lr": 8.983011753379857e-05}
{"train_loss": 0.2735007107257843, "global_step": 56486, "epoch": 634, "lr": 8.98297670782131e-05}
{"train_loss": 0.34027668833732605, "global_step": 56487, "epoch": 634, "lr": 8.982941661727303e-05}
{"train_loss": 0.22024127840995789, "global_step": 56488, "epoch": 634, "lr": 8.982906615097837e-05}
{"train_loss": 0.346871942281723, "global_step": 56489, "epoch": 634, "lr": 8.982871567932916e-05}
{"train_loss": 0.40427154302597046, "global_step": 56490, "epoch": 634, "lr": 8.982836520232549e-05}
{"train_loss": 0.19919045269489288, "global_step": 56491, "epoch": 634, "lr": 8.982801471996736e-05}
{"train_loss": 0.303855836391449, "global_step": 56492, "epoch": 634, "lr": 8.982766423225486e-05}
{"train_loss": 0.2219020575284958, "global_step": 56493, "epoch": 634, "lr": 8.9827313739188e-05}
{"train_loss": 0.234036386013031, "global_step": 56494, "epoch": 634, "lr": 8.982696324076684e-05}
{"train_loss": 0.3170768618583679, "global_step": 56495, "epoch": 634, "lr": 8.982661273699144e-05}
{"train_loss": 0.3204154670238495, "global_step": 56496, "epoch": 634, "lr": 8.982626222786184e-05}
{"train_loss": 0.2981533110141754, "global_step": 56497, "epoch": 634, "lr": 8.982591171337808e-05}
{"train_loss": 0.19110707938671112, "global_step": 56498, "epoch": 634, "lr": 8.982556119354021e-05}
{"train_loss": 0.2507611811161041, "global_step": 56499, "epoch": 634, "lr": 8.982521066834829e-05}
{"train_loss": 0.29532843828201294, "global_step": 56500, "epoch": 634, "lr": 8.982486013780233e-05}
{"train_loss": 0.2989368736743927, "global_step": 56501, "epoch": 634, "lr": 8.982450960190242e-05}
{"train_loss": 0.2752668261528015, "global_step": 56502, "epoch": 634, "lr": 8.98241590606486e-05}
{"train_loss": 0.3009311258792877, "global_step": 56503, "epoch": 634, "lr": 8.982380851404088e-05}
{"train_loss": 0.27699998021125793, "global_step": 56504, "epoch": 634, "lr": 8.982345796207936e-05}
{"train_loss": 0.3039941191673279, "global_step": 56505, "epoch": 634, "lr": 8.982310740476405e-05}
{"train_loss": 0.299763023853302, "global_step": 56506, "epoch": 634, "lr": 8.982275684209501e-05}
{"train_loss": 0.33996304869651794, "global_step": 56507, "epoch": 634, "lr": 8.98224062740723e-05}
{"train_loss": 0.31098511815071106, "global_step": 56508, "epoch": 634, "lr": 8.982205570069592e-05}
{"train_loss": 0.3277891278266907, "global_step": 56509, "epoch": 634, "lr": 8.982170512196598e-05}
{"train_loss": 0.3580436408519745, "global_step": 56510, "epoch": 634, "lr": 8.982135453788248e-05}
{"train_loss": 0.325141966342926, "global_step": 56511, "epoch": 634, "lr": 8.98210039484455e-05}
{"train_loss": 0.30999305844306946, "global_step": 56512, "epoch": 634, "lr": 8.982065335365506e-05}
{"train_loss": 0.2810191512107849, "global_step": 56513, "epoch": 634, "lr": 8.982030275351122e-05}
{"train_loss": 0.2714698220906633, "global_step": 56514, "epoch": 634, "lr": 8.981995214801401e-05, "val_loss": 3.3400378227233887}
{"train_loss": 0.1908307671546936, "global_step": 56515, "epoch": 635, "lr": 8.981960153716352e-05}
{"train_loss": 0.19272199273109436, "global_step": 56516, "epoch": 635, "lr": 8.981925092095975e-05}
{"train_loss": 0.2661799192428589, "global_step": 56517, "epoch": 635, "lr": 8.981890029940279e-05}
{"train_loss": 0.15638992190361023, "global_step": 56518, "epoch": 635, "lr": 8.981854967249265e-05}
{"train_loss": 0.3008422255516052, "global_step": 56519, "epoch": 635, "lr": 8.98181990402294e-05}
{"train_loss": 0.39927390217781067, "global_step": 56520, "epoch": 635, "lr": 8.981784840261305e-05}
{"train_loss": 0.2714177072048187, "global_step": 56521, "epoch": 635, "lr": 8.98174977596437e-05}
{"train_loss": 0.23391753435134888, "global_step": 56522, "epoch": 635, "lr": 8.981714711132136e-05}
{"train_loss": 0.25116872787475586, "global_step": 56523, "epoch": 635, "lr": 8.981679645764611e-05}
{"train_loss": 0.2775178551673889, "global_step": 56524, "epoch": 635, "lr": 8.981644579861796e-05}
{"train_loss": 0.3120264410972595, "global_step": 56525, "epoch": 635, "lr": 8.981609513423699e-05}
{"train_loss": 0.21999025344848633, "global_step": 56526, "epoch": 635, "lr": 8.981574446450321e-05}
{"train_loss": 0.28727129101753235, "global_step": 56527, "epoch": 635, "lr": 8.98153937894167e-05}
{"train_loss": 0.2546897530555725, "global_step": 56528, "epoch": 635, "lr": 8.98150431089775e-05}
{"train_loss": 0.290014386177063, "global_step": 56529, "epoch": 635, "lr": 8.981469242318565e-05}
{"train_loss": 0.19984112679958344, "global_step": 56530, "epoch": 635, "lr": 8.98143417320412e-05}
{"train_loss": 0.23431706428527832, "global_step": 56531, "epoch": 635, "lr": 8.981399103554417e-05}
{"train_loss": 0.24860486388206482, "global_step": 56532, "epoch": 635, "lr": 8.981364033369467e-05}
{"train_loss": 0.20632533729076385, "global_step": 56533, "epoch": 635, "lr": 8.98132896264927e-05}
{"train_loss": 0.2337932288646698, "global_step": 56534, "epoch": 635, "lr": 8.981293891393831e-05}
{"train_loss": 0.3624318838119507, "global_step": 56535, "epoch": 635, "lr": 8.981258819603158e-05}
{"train_loss": 0.3049326539039612, "global_step": 56536, "epoch": 635, "lr": 8.981223747277251e-05}
{"train_loss": 0.31018680334091187, "global_step": 56537, "epoch": 635, "lr": 8.981188674416117e-05}
{"train_loss": 0.22734133899211884, "global_step": 56538, "epoch": 635, "lr": 8.981153601019761e-05}
{"train_loss": 0.2539425790309906, "global_step": 56539, "epoch": 635, "lr": 8.981118527088189e-05}
{"train_loss": 0.20149783790111542, "global_step": 56540, "epoch": 635, "lr": 8.981083452621401e-05}
{"train_loss": 0.32707664370536804, "global_step": 56541, "epoch": 635, "lr": 8.981048377619408e-05}
{"train_loss": 0.3249424993991852, "global_step": 56542, "epoch": 635, "lr": 8.981013302082211e-05}
{"train_loss": 0.22567982971668243, "global_step": 56543, "epoch": 635, "lr": 8.980978226009814e-05}
{"train_loss": 0.3707643747329712, "global_step": 56544, "epoch": 635, "lr": 8.980943149402223e-05}
{"train_loss": 0.24394644796848297, "global_step": 56545, "epoch": 635, "lr": 8.980908072259444e-05}
{"train_loss": 0.2500515282154083, "global_step": 56546, "epoch": 635, "lr": 8.980872994581479e-05}
{"train_loss": 0.3144594430923462, "global_step": 56547, "epoch": 635, "lr": 8.980837916368335e-05}
{"train_loss": 0.32962191104888916, "global_step": 56548, "epoch": 635, "lr": 8.980802837620016e-05}
{"train_loss": 0.27405187487602234, "global_step": 56549, "epoch": 635, "lr": 8.980767758336526e-05}
{"train_loss": 0.3256482481956482, "global_step": 56550, "epoch": 635, "lr": 8.980732678517871e-05}
{"train_loss": 0.20730268955230713, "global_step": 56551, "epoch": 635, "lr": 8.980697598164054e-05}
{"train_loss": 0.257665753364563, "global_step": 56552, "epoch": 635, "lr": 8.980662517275082e-05}
{"train_loss": 0.25949302315711975, "global_step": 56553, "epoch": 635, "lr": 8.980627435850958e-05}
{"train_loss": 0.2942323088645935, "global_step": 56554, "epoch": 635, "lr": 8.980592353891688e-05}
{"train_loss": 0.31357356905937195, "global_step": 56555, "epoch": 635, "lr": 8.980557271397273e-05}
{"train_loss": 0.12621915340423584, "global_step": 56556, "epoch": 635, "lr": 8.980522188367724e-05}
{"train_loss": 0.23410539329051971, "global_step": 56557, "epoch": 635, "lr": 8.98048710480304e-05}
{"train_loss": 0.2891847789287567, "global_step": 56558, "epoch": 635, "lr": 8.980452020703229e-05}
{"train_loss": 0.28714388608932495, "global_step": 56559, "epoch": 635, "lr": 8.980416936068294e-05}
{"train_loss": 0.2564268112182617, "global_step": 56560, "epoch": 635, "lr": 8.98038185089824e-05}
{"train_loss": 0.22257240116596222, "global_step": 56561, "epoch": 635, "lr": 8.980346765193074e-05}
{"train_loss": 0.3021564185619354, "global_step": 56562, "epoch": 635, "lr": 8.980311678952798e-05}
{"train_loss": 0.2519068717956543, "global_step": 56563, "epoch": 635, "lr": 8.980276592177418e-05}
{"train_loss": 0.23455332219600677, "global_step": 56564, "epoch": 635, "lr": 8.980241504866938e-05}
{"train_loss": 0.21467329561710358, "global_step": 56565, "epoch": 635, "lr": 8.980206417021362e-05}
{"train_loss": 0.30905088782310486, "global_step": 56566, "epoch": 635, "lr": 8.980171328640698e-05}
{"train_loss": 0.21846672892570496, "global_step": 56567, "epoch": 635, "lr": 8.980136239724946e-05}
{"train_loss": 0.38706088066101074, "global_step": 56568, "epoch": 635, "lr": 8.980101150274115e-05}
{"train_loss": 0.30100277066230774, "global_step": 56569, "epoch": 635, "lr": 8.980066060288208e-05}
{"train_loss": 0.24042920768260956, "global_step": 56570, "epoch": 635, "lr": 8.980030969767228e-05}
{"train_loss": 0.2613338232040405, "global_step": 56571, "epoch": 635, "lr": 8.979995878711182e-05}
{"train_loss": 0.1872500628232956, "global_step": 56572, "epoch": 635, "lr": 8.979960787120075e-05}
{"train_loss": 0.2243916541337967, "global_step": 56573, "epoch": 635, "lr": 8.97992569499391e-05}
{"train_loss": 0.20921148359775543, "global_step": 56574, "epoch": 635, "lr": 8.979890602332693e-05}
{"train_loss": 0.2719716727733612, "global_step": 56575, "epoch": 635, "lr": 8.979855509136428e-05}
{"train_loss": 0.29559075832366943, "global_step": 56576, "epoch": 635, "lr": 8.979820415405119e-05}
{"train_loss": 0.2344471663236618, "global_step": 56577, "epoch": 635, "lr": 8.979785321138773e-05}
{"train_loss": 0.24238881468772888, "global_step": 56578, "epoch": 635, "lr": 8.979750226337394e-05}
{"train_loss": 0.2927989661693573, "global_step": 56579, "epoch": 635, "lr": 8.979715131000983e-05}
{"train_loss": 0.30090153217315674, "global_step": 56580, "epoch": 635, "lr": 8.97968003512955e-05}
{"train_loss": 0.28230786323547363, "global_step": 56581, "epoch": 635, "lr": 8.979644938723098e-05}
{"train_loss": 0.3645533621311188, "global_step": 56582, "epoch": 635, "lr": 8.97960984178163e-05}
{"train_loss": 0.26052024960517883, "global_step": 56583, "epoch": 635, "lr": 8.979574744305152e-05}
{"train_loss": 0.2820015549659729, "global_step": 56584, "epoch": 635, "lr": 8.979539646293669e-05}
{"train_loss": 0.24122047424316406, "global_step": 56585, "epoch": 635, "lr": 8.979504547747186e-05}
{"train_loss": 0.21500973403453827, "global_step": 56586, "epoch": 635, "lr": 8.979469448665707e-05}
{"train_loss": 0.36487993597984314, "global_step": 56587, "epoch": 635, "lr": 8.979434349049236e-05}
{"train_loss": 0.3140074908733368, "global_step": 56588, "epoch": 635, "lr": 8.97939924889778e-05}
{"train_loss": 0.30556243658065796, "global_step": 56589, "epoch": 635, "lr": 8.97936414821134e-05}
{"train_loss": 0.2371162325143814, "global_step": 56590, "epoch": 635, "lr": 8.979329046989924e-05}
{"train_loss": 0.2720784544944763, "global_step": 56591, "epoch": 635, "lr": 8.979293945233537e-05}
{"train_loss": 0.20556318759918213, "global_step": 56592, "epoch": 635, "lr": 8.97925884294218e-05}
{"train_loss": 0.23290567100048065, "global_step": 56593, "epoch": 635, "lr": 8.979223740115862e-05}
{"train_loss": 0.23029792308807373, "global_step": 56594, "epoch": 635, "lr": 8.979188636754586e-05}
{"train_loss": 0.23750150203704834, "global_step": 56595, "epoch": 635, "lr": 8.979153532858355e-05}
{"train_loss": 0.18834999203681946, "global_step": 56596, "epoch": 635, "lr": 8.979118428427178e-05}
{"train_loss": 0.23674815893173218, "global_step": 56597, "epoch": 635, "lr": 8.979083323461055e-05}
{"train_loss": 0.14852218329906464, "global_step": 56598, "epoch": 635, "lr": 8.979048217959994e-05}
{"train_loss": 0.2325619012117386, "global_step": 56599, "epoch": 635, "lr": 8.979013111923997e-05}
{"train_loss": 0.19590623676776886, "global_step": 56600, "epoch": 635, "lr": 8.97897800535307e-05}
{"train_loss": 0.24226327240467072, "global_step": 56601, "epoch": 635, "lr": 8.978942898247219e-05}
{"train_loss": 0.26860323548316956, "global_step": 56602, "epoch": 635, "lr": 8.978907790606448e-05}
{"train_loss": 0.2614612632922912, "global_step": 56603, "epoch": 635, "lr": 8.97887268243076e-05, "val_loss": 3.2814886569976807, "train_action_mse_error": 46.396400451660156}
{"train_loss": 0.19281955063343048, "global_step": 56604, "epoch": 636, "lr": 8.978837573720161e-05}
{"train_loss": 0.3123941421508789, "global_step": 56605, "epoch": 636, "lr": 8.978802464474657e-05}
{"train_loss": 0.29137471318244934, "global_step": 56606, "epoch": 636, "lr": 8.97876735469425e-05}
{"train_loss": 0.24965347349643707, "global_step": 56607, "epoch": 636, "lr": 8.97873224437895e-05}
{"train_loss": 0.19812515377998352, "global_step": 56608, "epoch": 636, "lr": 8.978697133528752e-05}
{"train_loss": 0.2570958733558655, "global_step": 56609, "epoch": 636, "lr": 8.97866202214367e-05}
{"train_loss": 0.2613992393016815, "global_step": 56610, "epoch": 636, "lr": 8.978626910223706e-05}
{"train_loss": 0.1949876844882965, "global_step": 56611, "epoch": 636, "lr": 8.978591797768864e-05}
{"train_loss": 0.2828620672225952, "global_step": 56612, "epoch": 636, "lr": 8.978556684779147e-05}
{"train_loss": 0.20413890480995178, "global_step": 56613, "epoch": 636, "lr": 8.978521571254564e-05}
{"train_loss": 0.25690263509750366, "global_step": 56614, "epoch": 636, "lr": 8.978486457195115e-05}
{"train_loss": 0.21308951079845428, "global_step": 56615, "epoch": 636, "lr": 8.978451342600808e-05}
{"train_loss": 0.2826404869556427, "global_step": 56616, "epoch": 636, "lr": 8.978416227471647e-05}
{"train_loss": 0.3245658278465271, "global_step": 56617, "epoch": 636, "lr": 8.978381111807637e-05}
{"train_loss": 0.3516022861003876, "global_step": 56618, "epoch": 636, "lr": 8.97834599560878e-05}
{"train_loss": 0.19304803013801575, "global_step": 56619, "epoch": 636, "lr": 8.978310878875085e-05}
{"train_loss": 0.18859289586544037, "global_step": 56620, "epoch": 636, "lr": 8.978275761606553e-05}
{"train_loss": 0.18466514348983765, "global_step": 56621, "epoch": 636, "lr": 8.978240643803191e-05}
{"train_loss": 0.2669121325016022, "global_step": 56622, "epoch": 636, "lr": 8.978205525465002e-05}
{"train_loss": 0.19530101120471954, "global_step": 56623, "epoch": 636, "lr": 8.978170406591993e-05}
{"train_loss": 0.20211046934127808, "global_step": 56624, "epoch": 636, "lr": 8.978135287184168e-05}
{"train_loss": 0.2872854471206665, "global_step": 56625, "epoch": 636, "lr": 8.97810016724153e-05}
{"train_loss": 0.2979592978954315, "global_step": 56626, "epoch": 636, "lr": 8.978065046764086e-05}
{"train_loss": 0.296674519777298, "global_step": 56627, "epoch": 636, "lr": 8.978029925751838e-05}
{"train_loss": 0.2812820374965668, "global_step": 56628, "epoch": 636, "lr": 8.977994804204795e-05}
{"train_loss": 0.18965770304203033, "global_step": 56629, "epoch": 636, "lr": 8.977959682122956e-05}
{"train_loss": 0.16368742287158966, "global_step": 56630, "epoch": 636, "lr": 8.97792455950633e-05}
{"train_loss": 0.17598901689052582, "global_step": 56631, "epoch": 636, "lr": 8.977889436354921e-05}
{"train_loss": 0.2072293758392334, "global_step": 56632, "epoch": 636, "lr": 8.977854312668735e-05}
{"train_loss": 0.24878621101379395, "global_step": 56633, "epoch": 636, "lr": 8.977819188447771e-05}
{"train_loss": 0.2227851301431656, "global_step": 56634, "epoch": 636, "lr": 8.97778406369204e-05}
{"train_loss": 0.24199770390987396, "global_step": 56635, "epoch": 636, "lr": 8.977748938401544e-05}
{"train_loss": 0.2662225663661957, "global_step": 56636, "epoch": 636, "lr": 8.97771381257629e-05}
{"train_loss": 0.2083396464586258, "global_step": 56637, "epoch": 636, "lr": 8.977678686216278e-05}
{"train_loss": 0.31887680292129517, "global_step": 56638, "epoch": 636, "lr": 8.977643559321516e-05}
{"train_loss": 0.27228739857673645, "global_step": 56639, "epoch": 636, "lr": 8.97760843189201e-05}
{"train_loss": 0.27075067162513733, "global_step": 56640, "epoch": 636, "lr": 8.977573303927762e-05}
{"train_loss": 0.2109275907278061, "global_step": 56641, "epoch": 636, "lr": 8.977538175428777e-05}
{"train_loss": 0.27929186820983887, "global_step": 56642, "epoch": 636, "lr": 8.977503046395062e-05}
{"train_loss": 0.2601315379142761, "global_step": 56643, "epoch": 636, "lr": 8.977467916826618e-05}
{"train_loss": 0.2850377559661865, "global_step": 56644, "epoch": 636, "lr": 8.977432786723454e-05}
{"train_loss": 0.27728471159935, "global_step": 56645, "epoch": 636, "lr": 8.977397656085572e-05}
{"train_loss": 0.21668767929077148, "global_step": 56646, "epoch": 636, "lr": 8.977362524912978e-05}
{"train_loss": 0.1833796203136444, "global_step": 56647, "epoch": 636, "lr": 8.977327393205675e-05}
{"train_loss": 0.35004720091819763, "global_step": 56648, "epoch": 636, "lr": 8.977292260963669e-05}
{"train_loss": 0.2652641236782074, "global_step": 56649, "epoch": 636, "lr": 8.977257128186965e-05}
{"train_loss": 0.26227614283561707, "global_step": 56650, "epoch": 636, "lr": 8.977221994875567e-05}
{"train_loss": 0.22225356101989746, "global_step": 56651, "epoch": 636, "lr": 8.977186861029481e-05}
{"train_loss": 0.255990207195282, "global_step": 56652, "epoch": 636, "lr": 8.97715172664871e-05}
{"train_loss": 0.22519351541996002, "global_step": 56653, "epoch": 636, "lr": 8.977116591733257e-05}
{"train_loss": 0.2328599989414215, "global_step": 56654, "epoch": 636, "lr": 8.977081456283132e-05}
{"train_loss": 0.21540695428848267, "global_step": 56655, "epoch": 636, "lr": 8.977046320298336e-05}
{"train_loss": 0.3286895453929901, "global_step": 56656, "epoch": 636, "lr": 8.977011183778875e-05}
{"train_loss": 0.21333490312099457, "global_step": 56657, "epoch": 636, "lr": 8.976976046724754e-05}
{"train_loss": 0.2670809030532837, "global_step": 56658, "epoch": 636, "lr": 8.976940909135976e-05}
{"train_loss": 0.25641489028930664, "global_step": 56659, "epoch": 636, "lr": 8.976905771012546e-05}
{"train_loss": 0.09561700373888016, "global_step": 56660, "epoch": 636, "lr": 8.976870632354471e-05}
{"train_loss": 0.3049582540988922, "global_step": 56661, "epoch": 636, "lr": 8.976835493161753e-05}
{"train_loss": 0.2994043231010437, "global_step": 56662, "epoch": 636, "lr": 8.976800353434399e-05}
{"train_loss": 0.26320505142211914, "global_step": 56663, "epoch": 636, "lr": 8.976765213172412e-05}
{"train_loss": 0.2141071856021881, "global_step": 56664, "epoch": 636, "lr": 8.976730072375797e-05}
{"train_loss": 0.22270089387893677, "global_step": 56665, "epoch": 636, "lr": 8.976694931044561e-05}
{"train_loss": 0.21560163795948029, "global_step": 56666, "epoch": 636, "lr": 8.976659789178706e-05}
{"train_loss": 0.2628755569458008, "global_step": 56667, "epoch": 636, "lr": 8.976624646778236e-05}
{"train_loss": 0.16310657560825348, "global_step": 56668, "epoch": 636, "lr": 8.976589503843159e-05}
{"train_loss": 0.18758738040924072, "global_step": 56669, "epoch": 636, "lr": 8.976554360373477e-05}
{"train_loss": 0.22300775349140167, "global_step": 56670, "epoch": 636, "lr": 8.976519216369195e-05}
{"train_loss": 0.27059996128082275, "global_step": 56671, "epoch": 636, "lr": 8.97648407183032e-05}
{"train_loss": 0.19061224162578583, "global_step": 56672, "epoch": 636, "lr": 8.976448926756855e-05}
{"train_loss": 0.2309994101524353, "global_step": 56673, "epoch": 636, "lr": 8.976413781148804e-05}
{"train_loss": 0.24428609013557434, "global_step": 56674, "epoch": 636, "lr": 8.976378635006174e-05}
{"train_loss": 0.2776421308517456, "global_step": 56675, "epoch": 636, "lr": 8.976343488328967e-05}
{"train_loss": 0.15810680389404297, "global_step": 56676, "epoch": 636, "lr": 8.97630834111719e-05}
{"train_loss": 0.2141660451889038, "global_step": 56677, "epoch": 636, "lr": 8.976273193370847e-05}
{"train_loss": 0.18670348823070526, "global_step": 56678, "epoch": 636, "lr": 8.976238045089942e-05}
{"train_loss": 0.14397071301937103, "global_step": 56679, "epoch": 636, "lr": 8.97620289627448e-05}
{"train_loss": 0.3141673803329468, "global_step": 56680, "epoch": 636, "lr": 8.976167746924467e-05}
{"train_loss": 0.3068138659000397, "global_step": 56681, "epoch": 636, "lr": 8.976132597039905e-05}
{"train_loss": 0.2336895912885666, "global_step": 56682, "epoch": 636, "lr": 8.976097446620801e-05}
{"train_loss": 0.23160704970359802, "global_step": 56683, "epoch": 636, "lr": 8.976062295667161e-05}
{"train_loss": 0.26449739933013916, "global_step": 56684, "epoch": 636, "lr": 8.976027144178987e-05}
{"train_loss": 0.25397148728370667, "global_step": 56685, "epoch": 636, "lr": 8.975991992156283e-05}
{"train_loss": 0.3592897951602936, "global_step": 56686, "epoch": 636, "lr": 8.975956839599057e-05}
{"train_loss": 0.26728659868240356, "global_step": 56687, "epoch": 636, "lr": 8.975921686507312e-05}
{"train_loss": 0.23266537487506866, "global_step": 56688, "epoch": 636, "lr": 8.975886532881051e-05}
{"train_loss": 0.24042248725891113, "global_step": 56689, "epoch": 636, "lr": 8.975851378720282e-05}
{"train_loss": 0.196366086602211, "global_step": 56690, "epoch": 636, "lr": 8.975816224025009e-05}
{"train_loss": 0.18182791769504547, "global_step": 56691, "epoch": 636, "lr": 8.975781068795234e-05}
{"train_loss": 0.242645465674695, "global_step": 56692, "epoch": 636, "lr": 8.975745913030966e-05, "val_loss": 3.2779617309570312}
{"train_loss": 0.1950986534357071, "global_step": 56693, "epoch": 637, "lr": 8.975710756732205e-05}
{"train_loss": 0.19156649708747864, "global_step": 56694, "epoch": 637, "lr": 8.975675599898959e-05}
{"train_loss": 0.14255604147911072, "global_step": 56695, "epoch": 637, "lr": 8.975640442531233e-05}
{"train_loss": 0.21042782068252563, "global_step": 56696, "epoch": 637, "lr": 8.975605284629029e-05}
{"train_loss": 0.21108581125736237, "global_step": 56697, "epoch": 637, "lr": 8.975570126192354e-05}
{"train_loss": 0.23493453860282898, "global_step": 56698, "epoch": 637, "lr": 8.975534967221214e-05}
{"train_loss": 0.2835184931755066, "global_step": 56699, "epoch": 637, "lr": 8.975499807715609e-05}
{"train_loss": 0.24659010767936707, "global_step": 56700, "epoch": 637, "lr": 8.975464647675546e-05}
{"train_loss": 0.18868282437324524, "global_step": 56701, "epoch": 637, "lr": 8.975429487101032e-05}
{"train_loss": 0.2162197232246399, "global_step": 56702, "epoch": 637, "lr": 8.97539432599207e-05}
{"train_loss": 0.24932712316513062, "global_step": 56703, "epoch": 637, "lr": 8.975359164348665e-05}
{"train_loss": 0.24063622951507568, "global_step": 56704, "epoch": 637, "lr": 8.975324002170819e-05}
{"train_loss": 0.22874245047569275, "global_step": 56705, "epoch": 637, "lr": 8.975288839458543e-05}
{"train_loss": 0.2399166077375412, "global_step": 56706, "epoch": 637, "lr": 8.975253676211835e-05}
{"train_loss": 0.3660350441932678, "global_step": 56707, "epoch": 637, "lr": 8.975218512430703e-05}
{"train_loss": 0.20127104222774506, "global_step": 56708, "epoch": 637, "lr": 8.975183348115153e-05}
{"train_loss": 0.22014367580413818, "global_step": 56709, "epoch": 637, "lr": 8.975148183265185e-05}
{"train_loss": 0.21470947563648224, "global_step": 56710, "epoch": 637, "lr": 8.97511301788081e-05}
{"train_loss": 0.2490312159061432, "global_step": 56711, "epoch": 637, "lr": 8.975077851962027e-05}
{"train_loss": 0.252508282661438, "global_step": 56712, "epoch": 637, "lr": 8.975042685508844e-05}
{"train_loss": 0.200219988822937, "global_step": 56713, "epoch": 637, "lr": 8.975007518521266e-05}
{"train_loss": 0.2764446437358856, "global_step": 56714, "epoch": 637, "lr": 8.974972350999295e-05}
{"train_loss": 0.24792133271694183, "global_step": 56715, "epoch": 637, "lr": 8.974937182942939e-05}
{"train_loss": 0.1633366197347641, "global_step": 56716, "epoch": 637, "lr": 8.9749020143522e-05}
{"train_loss": 0.28665095567703247, "global_step": 56717, "epoch": 637, "lr": 8.974866845227084e-05}
{"train_loss": 0.3450912833213806, "global_step": 56718, "epoch": 637, "lr": 8.974831675567596e-05}
{"train_loss": 0.35617271065711975, "global_step": 56719, "epoch": 637, "lr": 8.974796505373741e-05}
{"train_loss": 0.2109251469373703, "global_step": 56720, "epoch": 637, "lr": 8.974761334645522e-05}
{"train_loss": 0.18786528706550598, "global_step": 56721, "epoch": 637, "lr": 8.974726163382945e-05}
{"train_loss": 0.2583903670310974, "global_step": 56722, "epoch": 637, "lr": 8.974690991586014e-05}
{"train_loss": 0.2928234338760376, "global_step": 56723, "epoch": 637, "lr": 8.974655819254735e-05}
{"train_loss": 0.2259049117565155, "global_step": 56724, "epoch": 637, "lr": 8.974620646389111e-05}
{"train_loss": 0.22800526022911072, "global_step": 56725, "epoch": 637, "lr": 8.97458547298915e-05}
{"train_loss": 0.32140859961509705, "global_step": 56726, "epoch": 637, "lr": 8.974550299054852e-05}
{"train_loss": 0.17790192365646362, "global_step": 56727, "epoch": 637, "lr": 8.974515124586224e-05}
{"train_loss": 0.21145249903202057, "global_step": 56728, "epoch": 637, "lr": 8.974479949583274e-05}
{"train_loss": 0.2859245240688324, "global_step": 56729, "epoch": 637, "lr": 8.974444774046e-05}
{"train_loss": 0.2299177646636963, "global_step": 56730, "epoch": 637, "lr": 8.974409597974413e-05}
{"train_loss": 0.20327232778072357, "global_step": 56731, "epoch": 637, "lr": 8.974374421368514e-05}
{"train_loss": 0.2299235463142395, "global_step": 56732, "epoch": 637, "lr": 8.974339244228308e-05}
{"train_loss": 0.2462409883737564, "global_step": 56733, "epoch": 637, "lr": 8.974304066553802e-05}
{"train_loss": 0.24065279960632324, "global_step": 56734, "epoch": 637, "lr": 8.974268888344998e-05}
{"train_loss": 0.27500155568122864, "global_step": 56735, "epoch": 637, "lr": 8.974233709601902e-05}
{"train_loss": 0.25922903418540955, "global_step": 56736, "epoch": 637, "lr": 8.97419853032452e-05}
{"train_loss": 0.2504133880138397, "global_step": 56737, "epoch": 637, "lr": 8.974163350512854e-05}
{"train_loss": 0.25093162059783936, "global_step": 56738, "epoch": 637, "lr": 8.97412817016691e-05}
{"train_loss": 0.3214121162891388, "global_step": 56739, "epoch": 637, "lr": 8.974092989286695e-05}
{"train_loss": 0.29695162177085876, "global_step": 56740, "epoch": 637, "lr": 8.974057807872212e-05}
{"train_loss": 0.128925621509552, "global_step": 56741, "epoch": 637, "lr": 8.974022625923462e-05}
{"train_loss": 0.24377913773059845, "global_step": 56742, "epoch": 637, "lr": 8.973987443440455e-05}
{"train_loss": 0.3736119270324707, "global_step": 56743, "epoch": 637, "lr": 8.973952260423195e-05}
{"train_loss": 0.30070993304252625, "global_step": 56744, "epoch": 637, "lr": 8.973917076871682e-05}
{"train_loss": 0.34209585189819336, "global_step": 56745, "epoch": 637, "lr": 8.973881892785928e-05}
{"train_loss": 0.16857507824897766, "global_step": 56746, "epoch": 637, "lr": 8.973846708165932e-05}
{"train_loss": 0.21392467617988586, "global_step": 56747, "epoch": 637, "lr": 8.973811523011702e-05}
{"train_loss": 0.27326294779777527, "global_step": 56748, "epoch": 637, "lr": 8.973776337323241e-05}
{"train_loss": 0.21949270367622375, "global_step": 56749, "epoch": 637, "lr": 8.973741151100554e-05}
{"train_loss": 0.18317946791648865, "global_step": 56750, "epoch": 637, "lr": 8.973705964343646e-05}
{"train_loss": 0.23515523970127106, "global_step": 56751, "epoch": 637, "lr": 8.973670777052521e-05}
{"train_loss": 0.22623573243618011, "global_step": 56752, "epoch": 637, "lr": 8.973635589227185e-05}
{"train_loss": 0.28675365447998047, "global_step": 56753, "epoch": 637, "lr": 8.973600400867642e-05}
{"train_loss": 0.2592015862464905, "global_step": 56754, "epoch": 637, "lr": 8.973565211973898e-05}
{"train_loss": 0.2286626696586609, "global_step": 56755, "epoch": 637, "lr": 8.973530022545954e-05}
{"train_loss": 0.3002055883407593, "global_step": 56756, "epoch": 637, "lr": 8.973494832583819e-05}
{"train_loss": 0.2382001429796219, "global_step": 56757, "epoch": 637, "lr": 8.973459642087496e-05}
{"train_loss": 0.2287842482328415, "global_step": 56758, "epoch": 637, "lr": 8.97342445105699e-05}
{"train_loss": 0.27980557084083557, "global_step": 56759, "epoch": 637, "lr": 8.973389259492305e-05}
{"train_loss": 0.2637563943862915, "global_step": 56760, "epoch": 637, "lr": 8.973354067393447e-05}
{"train_loss": 0.3001982271671295, "global_step": 56761, "epoch": 637, "lr": 8.973318874760419e-05}
{"train_loss": 0.2293335199356079, "global_step": 56762, "epoch": 637, "lr": 8.973283681593228e-05}
{"train_loss": 0.2768629193305969, "global_step": 56763, "epoch": 637, "lr": 8.973248487891877e-05}
{"train_loss": 0.2507131099700928, "global_step": 56764, "epoch": 637, "lr": 8.97321329365637e-05}
{"train_loss": 0.27713704109191895, "global_step": 56765, "epoch": 637, "lr": 8.973178098886714e-05}
{"train_loss": 0.346914678812027, "global_step": 56766, "epoch": 637, "lr": 8.973142903582912e-05}
{"train_loss": 0.19929762184619904, "global_step": 56767, "epoch": 637, "lr": 8.973107707744969e-05}
{"train_loss": 0.2749231457710266, "global_step": 56768, "epoch": 637, "lr": 8.97307251137289e-05}
{"train_loss": 0.2150636613368988, "global_step": 56769, "epoch": 637, "lr": 8.973037314466681e-05}
{"train_loss": 0.2561517059803009, "global_step": 56770, "epoch": 637, "lr": 8.973002117026346e-05}
{"train_loss": 0.2731441855430603, "global_step": 56771, "epoch": 637, "lr": 8.972966919051888e-05}
{"train_loss": 0.20029479265213013, "global_step": 56772, "epoch": 637, "lr": 8.972931720543314e-05}
{"train_loss": 0.25444427132606506, "global_step": 56773, "epoch": 637, "lr": 8.972896521500626e-05}
{"train_loss": 0.29914337396621704, "global_step": 56774, "epoch": 637, "lr": 8.972861321923832e-05}
{"train_loss": 0.24336367845535278, "global_step": 56775, "epoch": 637, "lr": 8.972826121812934e-05}
{"train_loss": 0.2858670651912689, "global_step": 56776, "epoch": 637, "lr": 8.972790921167939e-05}
{"train_loss": 0.2582056224346161, "global_step": 56777, "epoch": 637, "lr": 8.972755719988851e-05}
{"train_loss": 0.30231383442878723, "global_step": 56778, "epoch": 637, "lr": 8.972720518275675e-05}
{"train_loss": 0.1961878389120102, "global_step": 56779, "epoch": 637, "lr": 8.972685316028413e-05}
{"train_loss": 0.23583872616291046, "global_step": 56780, "epoch": 637, "lr": 8.972650113247073e-05}
{"train_loss": 0.2482470798023631, "global_step": 56781, "epoch": 637, "lr": 8.972614909931658e-05, "val_loss": 3.305448293685913}
{"train_loss": 0.2751704752445221, "global_step": 56782, "epoch": 638, "lr": 8.972579706082175e-05}
{"train_loss": 0.2510036528110504, "global_step": 56783, "epoch": 638, "lr": 8.972544501698626e-05}
{"train_loss": 0.22213591635227203, "global_step": 56784, "epoch": 638, "lr": 8.972509296781015e-05}
{"train_loss": 0.2297835350036621, "global_step": 56785, "epoch": 638, "lr": 8.972474091329351e-05}
{"train_loss": 0.30168503522872925, "global_step": 56786, "epoch": 638, "lr": 8.972438885343637e-05}
{"train_loss": 0.19534899294376373, "global_step": 56787, "epoch": 638, "lr": 8.972403678823875e-05}
{"train_loss": 0.21906767785549164, "global_step": 56788, "epoch": 638, "lr": 8.972368471770072e-05}
{"train_loss": 0.3824044167995453, "global_step": 56789, "epoch": 638, "lr": 8.972333264182234e-05}
{"train_loss": 0.28414222598075867, "global_step": 56790, "epoch": 638, "lr": 8.972298056060363e-05}
{"train_loss": 0.3218986988067627, "global_step": 56791, "epoch": 638, "lr": 8.972262847404465e-05}
{"train_loss": 0.1793152093887329, "global_step": 56792, "epoch": 638, "lr": 8.972227638214543e-05}
{"train_loss": 0.26578518748283386, "global_step": 56793, "epoch": 638, "lr": 8.972192428490606e-05}
{"train_loss": 0.24749936163425446, "global_step": 56794, "epoch": 638, "lr": 8.972157218232655e-05}
{"train_loss": 0.2761196494102478, "global_step": 56795, "epoch": 638, "lr": 8.972122007440696e-05}
{"train_loss": 0.22533360123634338, "global_step": 56796, "epoch": 638, "lr": 8.972086796114733e-05}
{"train_loss": 0.23036342859268188, "global_step": 56797, "epoch": 638, "lr": 8.972051584254773e-05}
{"train_loss": 0.4140915274620056, "global_step": 56798, "epoch": 638, "lr": 8.972016371860818e-05}
{"train_loss": 0.32635870575904846, "global_step": 56799, "epoch": 638, "lr": 8.971981158932874e-05}
{"train_loss": 0.3129380941390991, "global_step": 56800, "epoch": 638, "lr": 8.971945945470945e-05}
{"train_loss": 0.39211106300354004, "global_step": 56801, "epoch": 638, "lr": 8.971910731475037e-05}
{"train_loss": 0.325346976518631, "global_step": 56802, "epoch": 638, "lr": 8.971875516945155e-05}
{"train_loss": 0.3566938638687134, "global_step": 56803, "epoch": 638, "lr": 8.971840301881301e-05}
{"train_loss": 0.25550976395606995, "global_step": 56804, "epoch": 638, "lr": 8.971805086283482e-05}
{"train_loss": 0.3125230669975281, "global_step": 56805, "epoch": 638, "lr": 8.971769870151701e-05}
{"train_loss": 0.2712468206882477, "global_step": 56806, "epoch": 638, "lr": 8.971734653485965e-05}
{"train_loss": 0.2716579735279083, "global_step": 56807, "epoch": 638, "lr": 8.971699436286278e-05}
{"train_loss": 0.2949891984462738, "global_step": 56808, "epoch": 638, "lr": 8.971664218552644e-05}
{"train_loss": 0.2006242573261261, "global_step": 56809, "epoch": 638, "lr": 8.971629000285069e-05}
{"train_loss": 0.22626636922359467, "global_step": 56810, "epoch": 638, "lr": 8.971593781483556e-05}
{"train_loss": 0.33119940757751465, "global_step": 56811, "epoch": 638, "lr": 8.971558562148111e-05}
{"train_loss": 0.31346505880355835, "global_step": 56812, "epoch": 638, "lr": 8.971523342278738e-05}
{"train_loss": 0.26433029770851135, "global_step": 56813, "epoch": 638, "lr": 8.971488121875442e-05}
{"train_loss": 0.24042777717113495, "global_step": 56814, "epoch": 638, "lr": 8.971452900938228e-05}
{"train_loss": 0.1711384505033493, "global_step": 56815, "epoch": 638, "lr": 8.971417679467103e-05}
{"train_loss": 0.2862059473991394, "global_step": 56816, "epoch": 638, "lr": 8.971382457462066e-05}
{"train_loss": 0.2633783221244812, "global_step": 56817, "epoch": 638, "lr": 8.971347234923126e-05}
{"train_loss": 0.21246188879013062, "global_step": 56818, "epoch": 638, "lr": 8.971312011850289e-05}
{"train_loss": 0.1959550976753235, "global_step": 56819, "epoch": 638, "lr": 8.971276788243553e-05}
{"train_loss": 0.25206777453422546, "global_step": 56820, "epoch": 638, "lr": 8.97124156410293e-05}
{"train_loss": 0.21062758564949036, "global_step": 56821, "epoch": 638, "lr": 8.971206339428422e-05}
{"train_loss": 0.23558561503887177, "global_step": 56822, "epoch": 638, "lr": 8.971171114220033e-05}
{"train_loss": 0.2608499228954315, "global_step": 56823, "epoch": 638, "lr": 8.971135888477769e-05}
{"train_loss": 0.2013079971075058, "global_step": 56824, "epoch": 638, "lr": 8.971100662201634e-05}
{"train_loss": 0.26186397671699524, "global_step": 56825, "epoch": 638, "lr": 8.971065435391633e-05}
{"train_loss": 0.33422526717185974, "global_step": 56826, "epoch": 638, "lr": 8.97103020804777e-05}
{"train_loss": 0.33449697494506836, "global_step": 56827, "epoch": 638, "lr": 8.970994980170051e-05}
{"train_loss": 0.3148130774497986, "global_step": 56828, "epoch": 638, "lr": 8.97095975175848e-05}
{"train_loss": 0.3155118525028229, "global_step": 56829, "epoch": 638, "lr": 8.970924522813063e-05}
{"train_loss": 0.21091030538082123, "global_step": 56830, "epoch": 638, "lr": 8.9708892933338e-05}
{"train_loss": 0.31796857714653015, "global_step": 56831, "epoch": 638, "lr": 8.970854063320703e-05}
{"train_loss": 0.18116052448749542, "global_step": 56832, "epoch": 638, "lr": 8.970818832773772e-05}
{"train_loss": 0.2547447085380554, "global_step": 56833, "epoch": 638, "lr": 8.970783601693012e-05}
{"train_loss": 0.28909504413604736, "global_step": 56834, "epoch": 638, "lr": 8.97074837007843e-05}
{"train_loss": 0.23483842611312866, "global_step": 56835, "epoch": 638, "lr": 8.970713137930029e-05}
{"train_loss": 0.22913411259651184, "global_step": 56836, "epoch": 638, "lr": 8.970677905247813e-05}
{"train_loss": 0.1557464450597763, "global_step": 56837, "epoch": 638, "lr": 8.970642672031788e-05}
{"train_loss": 0.20518618822097778, "global_step": 56838, "epoch": 638, "lr": 8.970607438281959e-05}
{"train_loss": 0.2509492337703705, "global_step": 56839, "epoch": 638, "lr": 8.970572203998329e-05}
{"train_loss": 0.2571581304073334, "global_step": 56840, "epoch": 638, "lr": 8.970536969180905e-05}
{"train_loss": 0.3294306993484497, "global_step": 56841, "epoch": 638, "lr": 8.97050173382969e-05}
{"train_loss": 0.24830232560634613, "global_step": 56842, "epoch": 638, "lr": 8.970466497944692e-05}
{"train_loss": 0.20053409039974213, "global_step": 56843, "epoch": 638, "lr": 8.97043126152591e-05}
{"train_loss": 0.2586919665336609, "global_step": 56844, "epoch": 638, "lr": 8.970396024573353e-05}
{"train_loss": 0.3105999529361725, "global_step": 56845, "epoch": 638, "lr": 8.970360787087025e-05}
{"train_loss": 0.29035133123397827, "global_step": 56846, "epoch": 638, "lr": 8.970325549066932e-05}
{"train_loss": 0.27957525849342346, "global_step": 56847, "epoch": 638, "lr": 8.970290310513075e-05}
{"train_loss": 0.2656026780605316, "global_step": 56848, "epoch": 638, "lr": 8.97025507142546e-05}
{"train_loss": 0.1930856555700302, "global_step": 56849, "epoch": 638, "lr": 8.970219831804095e-05}
{"train_loss": 0.1674509048461914, "global_step": 56850, "epoch": 638, "lr": 8.97018459164898e-05}
{"train_loss": 0.22225621342658997, "global_step": 56851, "epoch": 638, "lr": 8.970149350960123e-05}
{"train_loss": 0.22712872922420502, "global_step": 56852, "epoch": 638, "lr": 8.970114109737528e-05}
{"train_loss": 0.23169302940368652, "global_step": 56853, "epoch": 638, "lr": 8.9700788679812e-05}
{"train_loss": 0.28883421421051025, "global_step": 56854, "epoch": 638, "lr": 8.970043625691143e-05}
{"train_loss": 0.2627198100090027, "global_step": 56855, "epoch": 638, "lr": 8.970008382867362e-05}
{"train_loss": 0.17916953563690186, "global_step": 56856, "epoch": 638, "lr": 8.96997313950986e-05}
{"train_loss": 0.2698569893836975, "global_step": 56857, "epoch": 638, "lr": 8.969937895618645e-05}
{"train_loss": 0.18818299472332, "global_step": 56858, "epoch": 638, "lr": 8.969902651193721e-05}
{"train_loss": 0.21227823197841644, "global_step": 56859, "epoch": 638, "lr": 8.96986740623509e-05}
{"train_loss": 0.29411986470222473, "global_step": 56860, "epoch": 638, "lr": 8.969832160742761e-05}
{"train_loss": 0.253153920173645, "global_step": 56861, "epoch": 638, "lr": 8.969796914716735e-05}
{"train_loss": 0.2577942907810211, "global_step": 56862, "epoch": 638, "lr": 8.96976166815702e-05}
{"train_loss": 0.26844730973243713, "global_step": 56863, "epoch": 638, "lr": 8.969726421063616e-05}
{"train_loss": 0.3352024257183075, "global_step": 56864, "epoch": 638, "lr": 8.969691173436534e-05}
{"train_loss": 0.28277769684791565, "global_step": 56865, "epoch": 638, "lr": 8.969655925275773e-05}
{"train_loss": 0.2842632532119751, "global_step": 56866, "epoch": 638, "lr": 8.96962067658134e-05}
{"train_loss": 0.2510010898113251, "global_step": 56867, "epoch": 638, "lr": 8.969585427353242e-05}
{"train_loss": 0.29095420241355896, "global_step": 56868, "epoch": 638, "lr": 8.969550177591479e-05}
{"train_loss": 0.2787024974822998, "global_step": 56869, "epoch": 638, "lr": 8.969514927296061e-05}
{"train_loss": 0.26138134032822724, "global_step": 56870, "epoch": 638, "lr": 8.969479676466988e-05, "val_loss": 3.2029190063476562}
{"train_loss": 0.27730241417884827, "global_step": 56871, "epoch": 639, "lr": 8.969444425104268e-05}
{"train_loss": 0.1949475109577179, "global_step": 56872, "epoch": 639, "lr": 8.969409173207903e-05}
{"train_loss": 0.23691712319850922, "global_step": 56873, "epoch": 639, "lr": 8.969373920777901e-05}
{"train_loss": 0.25917601585388184, "global_step": 56874, "epoch": 639, "lr": 8.969338667814264e-05}
{"train_loss": 0.28335294127464294, "global_step": 56875, "epoch": 639, "lr": 8.969303414316999e-05}
{"train_loss": 0.2458227127790451, "global_step": 56876, "epoch": 639, "lr": 8.969268160286109e-05}
{"train_loss": 0.2936440706253052, "global_step": 56877, "epoch": 639, "lr": 8.969232905721599e-05}
{"train_loss": 0.19972790777683258, "global_step": 56878, "epoch": 639, "lr": 8.969197650623474e-05}
{"train_loss": 0.23344771564006805, "global_step": 56879, "epoch": 639, "lr": 8.969162394991739e-05}
{"train_loss": 0.3012513518333435, "global_step": 56880, "epoch": 639, "lr": 8.969127138826399e-05}
{"train_loss": 0.21464337408542633, "global_step": 56881, "epoch": 639, "lr": 8.969091882127458e-05}
{"train_loss": 0.24883760511875153, "global_step": 56882, "epoch": 639, "lr": 8.969056624894921e-05}
{"train_loss": 0.18051154911518097, "global_step": 56883, "epoch": 639, "lr": 8.969021367128791e-05}
{"train_loss": 0.1715128868818283, "global_step": 56884, "epoch": 639, "lr": 8.968986108829076e-05}
{"train_loss": 0.3033369481563568, "global_step": 56885, "epoch": 639, "lr": 8.968950849995779e-05}
{"train_loss": 0.219697505235672, "global_step": 56886, "epoch": 639, "lr": 8.968915590628906e-05}
{"train_loss": 0.2343541979789734, "global_step": 56887, "epoch": 639, "lr": 8.968880330728458e-05}
{"train_loss": 0.3186070919036865, "global_step": 56888, "epoch": 639, "lr": 8.968845070294445e-05}
{"train_loss": 0.2367059588432312, "global_step": 56889, "epoch": 639, "lr": 8.968809809326868e-05}
{"train_loss": 0.30988064408302307, "global_step": 56890, "epoch": 639, "lr": 8.968774547825733e-05}
{"train_loss": 0.22939535975456238, "global_step": 56891, "epoch": 639, "lr": 8.968739285791045e-05}
{"train_loss": 0.23882733285427094, "global_step": 56892, "epoch": 639, "lr": 8.968704023222807e-05}
{"train_loss": 0.2763243317604065, "global_step": 56893, "epoch": 639, "lr": 8.968668760121026e-05}
{"train_loss": 0.314823180437088, "global_step": 56894, "epoch": 639, "lr": 8.968633496485706e-05}
{"train_loss": 0.22850674390792847, "global_step": 56895, "epoch": 639, "lr": 8.968598232316852e-05}
{"train_loss": 0.22061537206172943, "global_step": 56896, "epoch": 639, "lr": 8.968562967614469e-05}
{"train_loss": 0.20841462910175323, "global_step": 56897, "epoch": 639, "lr": 8.968527702378559e-05}
{"train_loss": 0.16119512915611267, "global_step": 56898, "epoch": 639, "lr": 8.968492436609131e-05}
{"train_loss": 0.15240558981895447, "global_step": 56899, "epoch": 639, "lr": 8.968457170306186e-05}
{"train_loss": 0.24951483309268951, "global_step": 56900, "epoch": 639, "lr": 8.96842190346973e-05}
{"train_loss": 0.28268131613731384, "global_step": 56901, "epoch": 639, "lr": 8.968386636099769e-05}
{"train_loss": 0.23734399676322937, "global_step": 56902, "epoch": 639, "lr": 8.968351368196306e-05}
{"train_loss": 0.22247812151908875, "global_step": 56903, "epoch": 639, "lr": 8.968316099759349e-05}
{"train_loss": 0.18150411546230316, "global_step": 56904, "epoch": 639, "lr": 8.968280830788898e-05}
{"train_loss": 0.22181861102581024, "global_step": 56905, "epoch": 639, "lr": 8.968245561284959e-05}
{"train_loss": 0.20680736005306244, "global_step": 56906, "epoch": 639, "lr": 8.96821029124754e-05}
{"train_loss": 0.2885144352912903, "global_step": 56907, "epoch": 639, "lr": 8.968175020676642e-05}
{"train_loss": 0.2599136233329773, "global_step": 56908, "epoch": 639, "lr": 8.968139749572272e-05}
{"train_loss": 0.2706819772720337, "global_step": 56909, "epoch": 639, "lr": 8.968104477934433e-05}
{"train_loss": 0.21840205788612366, "global_step": 56910, "epoch": 639, "lr": 8.968069205763132e-05}
{"train_loss": 0.2362460345029831, "global_step": 56911, "epoch": 639, "lr": 8.968033933058373e-05}
{"train_loss": 0.24044428765773773, "global_step": 56912, "epoch": 639, "lr": 8.967998659820158e-05}
{"train_loss": 0.27222469449043274, "global_step": 56913, "epoch": 639, "lr": 8.967963386048495e-05}
{"train_loss": 0.2872757911682129, "global_step": 56914, "epoch": 639, "lr": 8.967928111743389e-05}
{"train_loss": 0.18862847983837128, "global_step": 56915, "epoch": 639, "lr": 8.967892836904842e-05}
{"train_loss": 0.29608154296875, "global_step": 56916, "epoch": 639, "lr": 8.96785756153286e-05}
{"train_loss": 0.2580321729183197, "global_step": 56917, "epoch": 639, "lr": 8.967822285627448e-05}
{"train_loss": 0.2247323840856552, "global_step": 56918, "epoch": 639, "lr": 8.96778700918861e-05}
{"train_loss": 0.22129710018634796, "global_step": 56919, "epoch": 639, "lr": 8.967751732216354e-05}
{"train_loss": 0.2308649867773056, "global_step": 56920, "epoch": 639, "lr": 8.967716454710679e-05}
{"train_loss": 0.2090330719947815, "global_step": 56921, "epoch": 639, "lr": 8.967681176671597e-05}
{"train_loss": 0.21886196732521057, "global_step": 56922, "epoch": 639, "lr": 8.967645898099104e-05}
{"train_loss": 0.3350386619567871, "global_step": 56923, "epoch": 639, "lr": 8.967610618993212e-05}
{"train_loss": 0.21428313851356506, "global_step": 56924, "epoch": 639, "lr": 8.967575339353922e-05}
{"train_loss": 0.22039902210235596, "global_step": 56925, "epoch": 639, "lr": 8.96754005918124e-05}
{"train_loss": 0.26292479038238525, "global_step": 56926, "epoch": 639, "lr": 8.967504778475173e-05}
{"train_loss": 0.47692152857780457, "global_step": 56927, "epoch": 639, "lr": 8.967469497235719e-05}
{"train_loss": 0.21701525151729584, "global_step": 56928, "epoch": 639, "lr": 8.967434215462891e-05}
{"train_loss": 0.2700818181037903, "global_step": 56929, "epoch": 639, "lr": 8.967398933156687e-05}
{"train_loss": 0.31409865617752075, "global_step": 56930, "epoch": 639, "lr": 8.967363650317117e-05}
{"train_loss": 0.2219979465007782, "global_step": 56931, "epoch": 639, "lr": 8.967328366944181e-05}
{"train_loss": 0.22584082186222076, "global_step": 56932, "epoch": 639, "lr": 8.967293083037887e-05}
{"train_loss": 0.31589075922966003, "global_step": 56933, "epoch": 639, "lr": 8.96725779859824e-05}
{"train_loss": 0.3251431882381439, "global_step": 56934, "epoch": 639, "lr": 8.967222513625242e-05}
{"train_loss": 0.20513732731342316, "global_step": 56935, "epoch": 639, "lr": 8.9671872281189e-05}
{"train_loss": 0.23670820891857147, "global_step": 56936, "epoch": 639, "lr": 8.967151942079217e-05}
{"train_loss": 0.19491802155971527, "global_step": 56937, "epoch": 639, "lr": 8.967116655506201e-05}
{"train_loss": 0.32309800386428833, "global_step": 56938, "epoch": 639, "lr": 8.967081368399852e-05}
{"train_loss": 0.28307995200157166, "global_step": 56939, "epoch": 639, "lr": 8.967046080760178e-05}
{"train_loss": 0.24469704926013947, "global_step": 56940, "epoch": 639, "lr": 8.967010792587184e-05}
{"train_loss": 0.17356163263320923, "global_step": 56941, "epoch": 639, "lr": 8.966975503880872e-05}
{"train_loss": 0.17899936437606812, "global_step": 56942, "epoch": 639, "lr": 8.966940214641251e-05}
{"train_loss": 0.2141033411026001, "global_step": 56943, "epoch": 639, "lr": 8.966904924868323e-05}
{"train_loss": 0.2528285086154938, "global_step": 56944, "epoch": 639, "lr": 8.96686963456209e-05}
{"train_loss": 0.24024565517902374, "global_step": 56945, "epoch": 639, "lr": 8.966834343722562e-05}
{"train_loss": 0.26368749141693115, "global_step": 56946, "epoch": 639, "lr": 8.966799052349742e-05}
{"train_loss": 0.42709168791770935, "global_step": 56947, "epoch": 639, "lr": 8.966763760443631e-05}
{"train_loss": 0.22546593844890594, "global_step": 56948, "epoch": 639, "lr": 8.966728468004239e-05}
{"train_loss": 0.23291011154651642, "global_step": 56949, "epoch": 639, "lr": 8.96669317503157e-05}
{"train_loss": 0.27237099409103394, "global_step": 56950, "epoch": 639, "lr": 8.966657881525627e-05}
{"train_loss": 0.29554712772369385, "global_step": 56951, "epoch": 639, "lr": 8.966622587486415e-05}
{"train_loss": 0.2940514087677002, "global_step": 56952, "epoch": 639, "lr": 8.966587292913938e-05}
{"train_loss": 0.26216691732406616, "global_step": 56953, "epoch": 639, "lr": 8.966551997808201e-05}
{"train_loss": 0.3784135580062866, "global_step": 56954, "epoch": 639, "lr": 8.966516702169212e-05}
{"train_loss": 0.2153046727180481, "global_step": 56955, "epoch": 639, "lr": 8.966481405996972e-05}
{"train_loss": 0.4118421673774719, "global_step": 56956, "epoch": 639, "lr": 8.966446109291485e-05}
{"train_loss": 0.329891562461853, "global_step": 56957, "epoch": 639, "lr": 8.96641081205276e-05}
{"train_loss": 0.26680701971054077, "global_step": 56958, "epoch": 639, "lr": 8.9663755142808e-05}
{"train_loss": 0.25339228701725436, "global_step": 56959, "epoch": 639, "lr": 8.966340215975607e-05, "val_loss": 3.207460641860962}
{"train_loss": 0.21482183039188385, "global_step": 56960, "epoch": 640, "lr": 8.966304917137189e-05}
{"train_loss": 0.2221205085515976, "global_step": 56961, "epoch": 640, "lr": 8.966269617765549e-05}
{"train_loss": 0.2512976825237274, "global_step": 56962, "epoch": 640, "lr": 8.966234317860693e-05}
{"train_loss": 0.2452676147222519, "global_step": 56963, "epoch": 640, "lr": 8.966199017422626e-05}
{"train_loss": 0.34161120653152466, "global_step": 56964, "epoch": 640, "lr": 8.96616371645135e-05}
{"train_loss": 0.1489098072052002, "global_step": 56965, "epoch": 640, "lr": 8.966128414946873e-05}
{"train_loss": 0.23622192442417145, "global_step": 56966, "epoch": 640, "lr": 8.966093112909198e-05}
{"train_loss": 0.25949689745903015, "global_step": 56967, "epoch": 640, "lr": 8.966057810338328e-05}
{"train_loss": 0.19981960952281952, "global_step": 56968, "epoch": 640, "lr": 8.966022507234273e-05}
{"train_loss": 0.2788582444190979, "global_step": 56969, "epoch": 640, "lr": 8.965987203597033e-05}
{"train_loss": 0.2475779503583908, "global_step": 56970, "epoch": 640, "lr": 8.965951899426616e-05}
{"train_loss": 0.2949104607105255, "global_step": 56971, "epoch": 640, "lr": 8.965916594723023e-05}
{"train_loss": 0.23747777938842773, "global_step": 56972, "epoch": 640, "lr": 8.965881289486262e-05}
{"train_loss": 0.2618461549282074, "global_step": 56973, "epoch": 640, "lr": 8.965845983716337e-05}
{"train_loss": 0.17023007571697235, "global_step": 56974, "epoch": 640, "lr": 8.96581067741325e-05}
{"train_loss": 0.23392900824546814, "global_step": 56975, "epoch": 640, "lr": 8.96577537057701e-05}
{"train_loss": 0.287150114774704, "global_step": 56976, "epoch": 640, "lr": 8.96574006320762e-05}
{"train_loss": 0.22161319851875305, "global_step": 56977, "epoch": 640, "lr": 8.965704755305085e-05}
{"train_loss": 0.23363108932971954, "global_step": 56978, "epoch": 640, "lr": 8.96566944686941e-05}
{"train_loss": 0.20052944123744965, "global_step": 56979, "epoch": 640, "lr": 8.965634137900597e-05}
{"train_loss": 0.3676718771457672, "global_step": 56980, "epoch": 640, "lr": 8.965598828398654e-05}
{"train_loss": 0.25051751732826233, "global_step": 56981, "epoch": 640, "lr": 8.965563518363583e-05}
{"train_loss": 0.18297263979911804, "global_step": 56982, "epoch": 640, "lr": 8.965528207795391e-05}
{"train_loss": 0.2157527506351471, "global_step": 56983, "epoch": 640, "lr": 8.965492896694083e-05}
{"train_loss": 0.2643463611602783, "global_step": 56984, "epoch": 640, "lr": 8.965457585059662e-05}
{"train_loss": 0.3175607919692993, "global_step": 56985, "epoch": 640, "lr": 8.965422272892134e-05}
{"train_loss": 0.26984766125679016, "global_step": 56986, "epoch": 640, "lr": 8.965386960191502e-05}
{"train_loss": 0.346156507730484, "global_step": 56987, "epoch": 640, "lr": 8.965351646957773e-05}
{"train_loss": 0.28743109107017517, "global_step": 56988, "epoch": 640, "lr": 8.965316333190951e-05}
{"train_loss": 0.20667687058448792, "global_step": 56989, "epoch": 640, "lr": 8.965281018891041e-05}
{"train_loss": 0.3311249017715454, "global_step": 56990, "epoch": 640, "lr": 8.965245704058047e-05}
{"train_loss": 0.14174750447273254, "global_step": 56991, "epoch": 640, "lr": 8.965210388691975e-05}
{"train_loss": 0.24995945394039154, "global_step": 56992, "epoch": 640, "lr": 8.965175072792827e-05}
{"train_loss": 0.34467098116874695, "global_step": 56993, "epoch": 640, "lr": 8.96513975636061e-05}
{"train_loss": 0.22975343465805054, "global_step": 56994, "epoch": 640, "lr": 8.965104439395329e-05}
{"train_loss": 0.2536056339740753, "global_step": 56995, "epoch": 640, "lr": 8.965069121896988e-05}
{"train_loss": 0.37541741132736206, "global_step": 56996, "epoch": 640, "lr": 8.96503380386559e-05}
{"train_loss": 0.18538445234298706, "global_step": 56997, "epoch": 640, "lr": 8.964998485301143e-05}
{"train_loss": 0.2661747932434082, "global_step": 56998, "epoch": 640, "lr": 8.96496316620365e-05}
{"train_loss": 0.1810213178396225, "global_step": 56999, "epoch": 640, "lr": 8.964927846573117e-05}
{"train_loss": 0.24696871638298035, "global_step": 57000, "epoch": 640, "lr": 8.964892526409548e-05}
{"train_loss": 0.27067482471466064, "global_step": 57001, "epoch": 640, "lr": 8.964857205712946e-05}
{"train_loss": 0.24146227538585663, "global_step": 57002, "epoch": 640, "lr": 8.964821884483317e-05}
{"train_loss": 0.26497742533683777, "global_step": 57003, "epoch": 640, "lr": 8.964786562720666e-05}
{"train_loss": 0.2139243632555008, "global_step": 57004, "epoch": 640, "lr": 8.964751240424999e-05}
{"train_loss": 0.28696367144584656, "global_step": 57005, "epoch": 640, "lr": 8.964715917596319e-05}
{"train_loss": 0.15584410727024078, "global_step": 57006, "epoch": 640, "lr": 8.964680594234631e-05}
{"train_loss": 0.18674243986606598, "global_step": 57007, "epoch": 640, "lr": 8.964645270339941e-05}
{"train_loss": 0.29278621077537537, "global_step": 57008, "epoch": 640, "lr": 8.964609945912253e-05}
{"train_loss": 0.29272300004959106, "global_step": 57009, "epoch": 640, "lr": 8.964574620951569e-05}
{"train_loss": 0.22478322684764862, "global_step": 57010, "epoch": 640, "lr": 8.964539295457899e-05}
{"train_loss": 0.32855114340782166, "global_step": 57011, "epoch": 640, "lr": 8.964503969431244e-05}
{"train_loss": 0.32707756757736206, "global_step": 57012, "epoch": 640, "lr": 8.964468642871608e-05}
{"train_loss": 0.20700110495090485, "global_step": 57013, "epoch": 640, "lr": 8.964433315779e-05}
{"train_loss": 0.2040211409330368, "global_step": 57014, "epoch": 640, "lr": 8.964397988153421e-05}
{"train_loss": 0.19556142389774323, "global_step": 57015, "epoch": 640, "lr": 8.964362659994878e-05}
{"train_loss": 0.16529680788516998, "global_step": 57016, "epoch": 640, "lr": 8.964327331303375e-05}
{"train_loss": 0.2440725564956665, "global_step": 57017, "epoch": 640, "lr": 8.964292002078916e-05}
{"train_loss": 0.22148576378822327, "global_step": 57018, "epoch": 640, "lr": 8.964256672321505e-05}
{"train_loss": 0.3835490345954895, "global_step": 57019, "epoch": 640, "lr": 8.96422134203115e-05}
{"train_loss": 0.2593620717525482, "global_step": 57020, "epoch": 640, "lr": 8.964186011207852e-05}
{"train_loss": 0.2729801535606384, "global_step": 57021, "epoch": 640, "lr": 8.964150679851618e-05}
{"train_loss": 0.2533033490180969, "global_step": 57022, "epoch": 640, "lr": 8.964115347962452e-05}
{"train_loss": 0.22718293964862823, "global_step": 57023, "epoch": 640, "lr": 8.96408001554036e-05}
{"train_loss": 0.2792513072490692, "global_step": 57024, "epoch": 640, "lr": 8.964044682585345e-05}
{"train_loss": 0.31023702025413513, "global_step": 57025, "epoch": 640, "lr": 8.964009349097413e-05}
{"train_loss": 0.1877085417509079, "global_step": 57026, "epoch": 640, "lr": 8.963974015076568e-05}
{"train_loss": 0.34009870886802673, "global_step": 57027, "epoch": 640, "lr": 8.963938680522815e-05}
{"train_loss": 0.2084016501903534, "global_step": 57028, "epoch": 640, "lr": 8.963903345436158e-05}
{"train_loss": 0.18910501897335052, "global_step": 57029, "epoch": 640, "lr": 8.963868009816603e-05}
{"train_loss": 0.21802778542041779, "global_step": 57030, "epoch": 640, "lr": 8.963832673664154e-05}
{"train_loss": 0.3215242028236389, "global_step": 57031, "epoch": 640, "lr": 8.963797336978818e-05}
{"train_loss": 0.18999339640140533, "global_step": 57032, "epoch": 640, "lr": 8.963761999760596e-05}
{"train_loss": 0.28645849227905273, "global_step": 57033, "epoch": 640, "lr": 8.963726662009495e-05}
{"train_loss": 0.20128995180130005, "global_step": 57034, "epoch": 640, "lr": 8.963691323725518e-05}
{"train_loss": 0.2517182528972626, "global_step": 57035, "epoch": 640, "lr": 8.963655984908673e-05}
{"train_loss": 0.33453497290611267, "global_step": 57036, "epoch": 640, "lr": 8.963620645558962e-05}
{"train_loss": 0.22286240756511688, "global_step": 57037, "epoch": 640, "lr": 8.96358530567639e-05}
{"train_loss": 0.21700750291347504, "global_step": 57038, "epoch": 640, "lr": 8.963549965260963e-05}
{"train_loss": 0.32379150390625, "global_step": 57039, "epoch": 640, "lr": 8.963514624312684e-05}
{"train_loss": 0.40911710262298584, "global_step": 57040, "epoch": 640, "lr": 8.963479282831559e-05}
{"train_loss": 0.26569995284080505, "global_step": 57041, "epoch": 640, "lr": 8.963443940817594e-05}
{"train_loss": 0.21634674072265625, "global_step": 57042, "epoch": 640, "lr": 8.963408598270791e-05}
{"train_loss": 0.32138511538505554, "global_step": 57043, "epoch": 640, "lr": 8.963373255191156e-05}
{"train_loss": 0.18843644857406616, "global_step": 57044, "epoch": 640, "lr": 8.963337911578694e-05}
{"train_loss": 0.23922047019004822, "global_step": 57045, "epoch": 640, "lr": 8.96330256743341e-05}
{"train_loss": 0.23426195979118347, "global_step": 57046, "epoch": 640, "lr": 8.963267222755307e-05}
{"train_loss": 0.290286660194397, "global_step": 57047, "epoch": 640, "lr": 8.963231877544392e-05}
{"train_loss": 0.2523650429891736, "global_step": 57048, "epoch": 640, "lr": 8.963196531800668e-05, "val_loss": 3.2076852321624756, "train_action_mse_error": 36.152652740478516}
{"train_loss": 0.2694306969642639, "global_step": 57049, "epoch": 641, "lr": 8.963161185524142e-05}
{"train_loss": 0.19382280111312866, "global_step": 57050, "epoch": 641, "lr": 8.963125838714817e-05}
{"train_loss": 0.24402381479740143, "global_step": 57051, "epoch": 641, "lr": 8.963090491372697e-05}
{"train_loss": 0.23311787843704224, "global_step": 57052, "epoch": 641, "lr": 8.963055143497789e-05}
{"train_loss": 0.17702767252922058, "global_step": 57053, "epoch": 641, "lr": 8.963019795090094e-05}
{"train_loss": 0.4222869575023651, "global_step": 57054, "epoch": 641, "lr": 8.962984446149623e-05}
{"train_loss": 0.26134347915649414, "global_step": 57055, "epoch": 641, "lr": 8.962949096676374e-05}
{"train_loss": 0.23440048098564148, "global_step": 57056, "epoch": 641, "lr": 8.962913746670357e-05}
{"train_loss": 0.17886550724506378, "global_step": 57057, "epoch": 641, "lr": 8.962878396131574e-05}
{"train_loss": 0.22396934032440186, "global_step": 57058, "epoch": 641, "lr": 8.962843045060028e-05}
{"train_loss": 0.2510439455509186, "global_step": 57059, "epoch": 641, "lr": 8.962807693455728e-05}
{"train_loss": 0.2010997235774994, "global_step": 57060, "epoch": 641, "lr": 8.962772341318678e-05}
{"train_loss": 0.2721695005893707, "global_step": 57061, "epoch": 641, "lr": 8.962736988648879e-05}
{"train_loss": 0.23210041224956512, "global_step": 57062, "epoch": 641, "lr": 8.96270163544634e-05}
{"train_loss": 0.2555246949195862, "global_step": 57063, "epoch": 641, "lr": 8.962666281711063e-05}
{"train_loss": 0.25356531143188477, "global_step": 57064, "epoch": 641, "lr": 8.962630927443054e-05}
{"train_loss": 0.2684704065322876, "global_step": 57065, "epoch": 641, "lr": 8.962595572642318e-05}
{"train_loss": 0.25271353125572205, "global_step": 57066, "epoch": 641, "lr": 8.96256021730886e-05}
{"train_loss": 0.2920314073562622, "global_step": 57067, "epoch": 641, "lr": 8.962524861442683e-05}
{"train_loss": 0.24671734869480133, "global_step": 57068, "epoch": 641, "lr": 8.962489505043793e-05}
{"train_loss": 0.34330153465270996, "global_step": 57069, "epoch": 641, "lr": 8.962454148112195e-05}
{"train_loss": 0.26296886801719666, "global_step": 57070, "epoch": 641, "lr": 8.962418790647893e-05}
{"train_loss": 0.20418906211853027, "global_step": 57071, "epoch": 641, "lr": 8.962383432650891e-05}
{"train_loss": 0.19978396594524384, "global_step": 57072, "epoch": 641, "lr": 8.962348074121197e-05}
{"train_loss": 0.23028318583965302, "global_step": 57073, "epoch": 641, "lr": 8.962312715058813e-05}
{"train_loss": 0.21867920458316803, "global_step": 57074, "epoch": 641, "lr": 8.962277355463744e-05}
{"train_loss": 0.27862024307250977, "global_step": 57075, "epoch": 641, "lr": 8.962241995335995e-05}
{"train_loss": 0.21805083751678467, "global_step": 57076, "epoch": 641, "lr": 8.96220663467557e-05}
{"train_loss": 0.3174005448818207, "global_step": 57077, "epoch": 641, "lr": 8.962171273482476e-05}
{"train_loss": 0.2242903858423233, "global_step": 57078, "epoch": 641, "lr": 8.962135911756716e-05}
{"train_loss": 0.3144494593143463, "global_step": 57079, "epoch": 641, "lr": 8.962100549498296e-05}
{"train_loss": 0.3839491605758667, "global_step": 57080, "epoch": 641, "lr": 8.962065186707217e-05}
{"train_loss": 0.25802141427993774, "global_step": 57081, "epoch": 641, "lr": 8.96202982338349e-05}
{"train_loss": 0.22716951370239258, "global_step": 57082, "epoch": 641, "lr": 8.961994459527114e-05}
{"train_loss": 0.25904425978660583, "global_step": 57083, "epoch": 641, "lr": 8.961959095138097e-05}
{"train_loss": 0.21230429410934448, "global_step": 57084, "epoch": 641, "lr": 8.961923730216443e-05}
{"train_loss": 0.26419174671173096, "global_step": 57085, "epoch": 641, "lr": 8.961888364762157e-05}
{"train_loss": 0.2510235011577606, "global_step": 57086, "epoch": 641, "lr": 8.961852998775244e-05}
{"train_loss": 0.24264095723628998, "global_step": 57087, "epoch": 641, "lr": 8.961817632255707e-05}
{"train_loss": 0.22842782735824585, "global_step": 57088, "epoch": 641, "lr": 8.961782265203552e-05}
{"train_loss": 0.2460644692182541, "global_step": 57089, "epoch": 641, "lr": 8.961746897618783e-05}
{"train_loss": 0.3040872812271118, "global_step": 57090, "epoch": 641, "lr": 8.961711529501406e-05}
{"train_loss": 0.29017314314842224, "global_step": 57091, "epoch": 641, "lr": 8.961676160851426e-05}
{"train_loss": 0.22370226681232452, "global_step": 57092, "epoch": 641, "lr": 8.961640791668845e-05}
{"train_loss": 0.21949774026870728, "global_step": 57093, "epoch": 641, "lr": 8.961605421953671e-05}
{"train_loss": 0.21109509468078613, "global_step": 57094, "epoch": 641, "lr": 8.961570051705908e-05}
{"train_loss": 0.28838303685188293, "global_step": 57095, "epoch": 641, "lr": 8.961534680925559e-05}
{"train_loss": 0.19902129471302032, "global_step": 57096, "epoch": 641, "lr": 8.961499309612632e-05}
{"train_loss": 0.27617329359054565, "global_step": 57097, "epoch": 641, "lr": 8.961463937767128e-05}
{"train_loss": 0.3325391411781311, "global_step": 57098, "epoch": 641, "lr": 8.961428565389052e-05}
{"train_loss": 0.27690592408180237, "global_step": 57099, "epoch": 641, "lr": 8.961393192478413e-05}
{"train_loss": 0.2563437521457672, "global_step": 57100, "epoch": 641, "lr": 8.96135781903521e-05}
{"train_loss": 0.1886882334947586, "global_step": 57101, "epoch": 641, "lr": 8.961322445059454e-05}
{"train_loss": 0.18609432876110077, "global_step": 57102, "epoch": 641, "lr": 8.961287070551144e-05}
{"train_loss": 0.24711208045482635, "global_step": 57103, "epoch": 641, "lr": 8.961251695510289e-05}
{"train_loss": 0.25683334469795227, "global_step": 57104, "epoch": 641, "lr": 8.961216319936892e-05}
{"train_loss": 0.19110652804374695, "global_step": 57105, "epoch": 641, "lr": 8.961180943830956e-05}
{"train_loss": 0.21105216443538666, "global_step": 57106, "epoch": 641, "lr": 8.961145567192488e-05}
{"train_loss": 0.2592828571796417, "global_step": 57107, "epoch": 641, "lr": 8.961110190021493e-05}
{"train_loss": 0.2999267578125, "global_step": 57108, "epoch": 641, "lr": 8.961074812317976e-05}
{"train_loss": 0.19850409030914307, "global_step": 57109, "epoch": 641, "lr": 8.961039434081938e-05}
{"train_loss": 0.25133273005485535, "global_step": 57110, "epoch": 641, "lr": 8.96100405531339e-05}
{"train_loss": 0.23787716031074524, "global_step": 57111, "epoch": 641, "lr": 8.960968676012329e-05}
{"train_loss": 0.2173544466495514, "global_step": 57112, "epoch": 641, "lr": 8.960933296178767e-05}
{"train_loss": 0.2545905113220215, "global_step": 57113, "epoch": 641, "lr": 8.960897915812704e-05}
{"train_loss": 0.24724633991718292, "global_step": 57114, "epoch": 641, "lr": 8.96086253491415e-05}
{"train_loss": 0.26675355434417725, "global_step": 57115, "epoch": 641, "lr": 8.960827153483102e-05}
{"train_loss": 0.27299946546554565, "global_step": 57116, "epoch": 641, "lr": 8.960791771519572e-05}
{"train_loss": 0.2797856032848358, "global_step": 57117, "epoch": 641, "lr": 8.960756389023561e-05}
{"train_loss": 0.25847601890563965, "global_step": 57118, "epoch": 641, "lr": 8.960721005995074e-05}
{"train_loss": 0.25968316197395325, "global_step": 57119, "epoch": 641, "lr": 8.960685622434118e-05}
{"train_loss": 0.19319666922092438, "global_step": 57120, "epoch": 641, "lr": 8.960650238340694e-05}
{"train_loss": 0.15576791763305664, "global_step": 57121, "epoch": 641, "lr": 8.96061485371481e-05}
{"train_loss": 0.2563490867614746, "global_step": 57122, "epoch": 641, "lr": 8.960579468556469e-05}
{"train_loss": 0.18569642305374146, "global_step": 57123, "epoch": 641, "lr": 8.960544082865676e-05}
{"train_loss": 0.23448967933654785, "global_step": 57124, "epoch": 641, "lr": 8.960508696642438e-05}
{"train_loss": 0.29938459396362305, "global_step": 57125, "epoch": 641, "lr": 8.960473309886757e-05}
{"train_loss": 0.32546406984329224, "global_step": 57126, "epoch": 641, "lr": 8.960437922598638e-05}
{"train_loss": 0.21061579883098602, "global_step": 57127, "epoch": 641, "lr": 8.960402534778087e-05}
{"train_loss": 0.24969512224197388, "global_step": 57128, "epoch": 641, "lr": 8.960367146425108e-05}
{"train_loss": 0.30343908071517944, "global_step": 57129, "epoch": 641, "lr": 8.960331757539707e-05}
{"train_loss": 0.41397610306739807, "global_step": 57130, "epoch": 641, "lr": 8.960296368121885e-05}
{"train_loss": 0.3237246572971344, "global_step": 57131, "epoch": 641, "lr": 8.960260978171651e-05}
{"train_loss": 0.1975274235010147, "global_step": 57132, "epoch": 641, "lr": 8.960225587689008e-05}
{"train_loss": 0.226136714220047, "global_step": 57133, "epoch": 641, "lr": 8.96019019667396e-05}
{"train_loss": 0.16180643439292908, "global_step": 57134, "epoch": 641, "lr": 8.960154805126515e-05}
{"train_loss": 0.20026393234729767, "global_step": 57135, "epoch": 641, "lr": 8.960119413046673e-05}
{"train_loss": 0.1848452240228653, "global_step": 57136, "epoch": 641, "lr": 8.960084020434442e-05}
{"train_loss": 0.24998047746969074, "global_step": 57137, "epoch": 641, "lr": 8.960048627289827e-05, "val_loss": 3.4522416591644287}
{"train_loss": 0.2804073393344879, "global_step": 57138, "epoch": 642, "lr": 8.960013233612831e-05}
{"train_loss": 0.2295769453048706, "global_step": 57139, "epoch": 642, "lr": 8.959977839403459e-05}
{"train_loss": 0.2346183806657791, "global_step": 57140, "epoch": 642, "lr": 8.959942444661717e-05}
{"train_loss": 0.224497988820076, "global_step": 57141, "epoch": 642, "lr": 8.95990704938761e-05}
{"train_loss": 0.1581549197435379, "global_step": 57142, "epoch": 642, "lr": 8.95987165358114e-05}
{"train_loss": 0.30575498938560486, "global_step": 57143, "epoch": 642, "lr": 8.959836257242314e-05}
{"train_loss": 0.31950387358665466, "global_step": 57144, "epoch": 642, "lr": 8.959800860371138e-05}
{"train_loss": 0.24251899123191833, "global_step": 57145, "epoch": 642, "lr": 8.959765462967613e-05}
{"train_loss": 0.389809250831604, "global_step": 57146, "epoch": 642, "lr": 8.959730065031746e-05}
{"train_loss": 0.2691863477230072, "global_step": 57147, "epoch": 642, "lr": 8.959694666563541e-05}
{"train_loss": 0.214870885014534, "global_step": 57148, "epoch": 642, "lr": 8.959659267563004e-05}
{"train_loss": 0.27471986413002014, "global_step": 57149, "epoch": 642, "lr": 8.95962386803014e-05}
{"train_loss": 0.23760917782783508, "global_step": 57150, "epoch": 642, "lr": 8.959588467964951e-05}
{"train_loss": 0.2374325543642044, "global_step": 57151, "epoch": 642, "lr": 8.959553067367445e-05}
{"train_loss": 0.30078554153442383, "global_step": 57152, "epoch": 642, "lr": 8.959517666237624e-05}
{"train_loss": 0.21898914873600006, "global_step": 57153, "epoch": 642, "lr": 8.959482264575497e-05}
{"train_loss": 0.31434375047683716, "global_step": 57154, "epoch": 642, "lr": 8.959446862381062e-05}
{"train_loss": 0.22902952134609222, "global_step": 57155, "epoch": 642, "lr": 8.95941145965433e-05}
{"train_loss": 0.2198609709739685, "global_step": 57156, "epoch": 642, "lr": 8.959376056395304e-05}
{"train_loss": 0.16430100798606873, "global_step": 57157, "epoch": 642, "lr": 8.959340652603986e-05}
{"train_loss": 0.31156688928604126, "global_step": 57158, "epoch": 642, "lr": 8.959305248280384e-05}
{"train_loss": 0.30577611923217773, "global_step": 57159, "epoch": 642, "lr": 8.959269843424501e-05}
{"train_loss": 0.23201361298561096, "global_step": 57160, "epoch": 642, "lr": 8.959234438036344e-05}
{"train_loss": 0.3917302191257477, "global_step": 57161, "epoch": 642, "lr": 8.959199032115915e-05}
{"train_loss": 0.2008781135082245, "global_step": 57162, "epoch": 642, "lr": 8.959163625663221e-05}
{"train_loss": 0.36168336868286133, "global_step": 57163, "epoch": 642, "lr": 8.959128218678264e-05}
{"train_loss": 0.2621782422065735, "global_step": 57164, "epoch": 642, "lr": 8.959092811161052e-05}
{"train_loss": 0.26216402649879456, "global_step": 57165, "epoch": 642, "lr": 8.959057403111585e-05}
{"train_loss": 0.2660170793533325, "global_step": 57166, "epoch": 642, "lr": 8.959021994529875e-05}
{"train_loss": 0.3165935277938843, "global_step": 57167, "epoch": 642, "lr": 8.958986585415922e-05}
{"train_loss": 0.2172434777021408, "global_step": 57168, "epoch": 642, "lr": 8.958951175769731e-05}
{"train_loss": 0.2987995445728302, "global_step": 57169, "epoch": 642, "lr": 8.958915765591306e-05}
{"train_loss": 0.33307287096977234, "global_step": 57170, "epoch": 642, "lr": 8.958880354880654e-05}
{"train_loss": 0.2779935896396637, "global_step": 57171, "epoch": 642, "lr": 8.95884494363778e-05}
{"train_loss": 0.21882778406143188, "global_step": 57172, "epoch": 642, "lr": 8.958809531862686e-05}
{"train_loss": 0.3048633337020874, "global_step": 57173, "epoch": 642, "lr": 8.95877411955538e-05}
{"train_loss": 0.27533990144729614, "global_step": 57174, "epoch": 642, "lr": 8.958738706715862e-05}
{"train_loss": 0.2976221740245819, "global_step": 57175, "epoch": 642, "lr": 8.958703293344142e-05}
{"train_loss": 0.3196343779563904, "global_step": 57176, "epoch": 642, "lr": 8.958667879440222e-05}
{"train_loss": 0.37447482347488403, "global_step": 57177, "epoch": 642, "lr": 8.958632465004108e-05}
{"train_loss": 0.30188342928886414, "global_step": 57178, "epoch": 642, "lr": 8.958597050035805e-05}
{"train_loss": 0.1793503761291504, "global_step": 57179, "epoch": 642, "lr": 8.958561634535314e-05}
{"train_loss": 0.17833290994167328, "global_step": 57180, "epoch": 642, "lr": 8.958526218502645e-05}
{"train_loss": 0.34364670515060425, "global_step": 57181, "epoch": 642, "lr": 8.9584908019378e-05}
{"train_loss": 0.29806357622146606, "global_step": 57182, "epoch": 642, "lr": 8.958455384840784e-05}
{"train_loss": 0.3140343129634857, "global_step": 57183, "epoch": 642, "lr": 8.958419967211601e-05}
{"train_loss": 0.23762257397174835, "global_step": 57184, "epoch": 642, "lr": 8.958384549050257e-05}
{"train_loss": 0.19134892523288727, "global_step": 57185, "epoch": 642, "lr": 8.958349130356757e-05}
{"train_loss": 0.35831165313720703, "global_step": 57186, "epoch": 642, "lr": 8.958313711131106e-05}
{"train_loss": 0.31756001710891724, "global_step": 57187, "epoch": 642, "lr": 8.958278291373307e-05}
{"train_loss": 0.24845363199710846, "global_step": 57188, "epoch": 642, "lr": 8.958242871083365e-05}
{"train_loss": 0.21797488629817963, "global_step": 57189, "epoch": 642, "lr": 8.958207450261287e-05}
{"train_loss": 0.27887797355651855, "global_step": 57190, "epoch": 642, "lr": 8.958172028907075e-05}
{"train_loss": 0.3360527455806732, "global_step": 57191, "epoch": 642, "lr": 8.958136607020735e-05}
{"train_loss": 0.22504886984825134, "global_step": 57192, "epoch": 642, "lr": 8.958101184602272e-05}
{"train_loss": 0.19540934264659882, "global_step": 57193, "epoch": 642, "lr": 8.958065761651691e-05}
{"train_loss": 0.21568672358989716, "global_step": 57194, "epoch": 642, "lr": 8.958030338168996e-05}
{"train_loss": 0.261966347694397, "global_step": 57195, "epoch": 642, "lr": 8.957994914154194e-05}
{"train_loss": 0.230550616979599, "global_step": 57196, "epoch": 642, "lr": 8.957959489607284e-05}
{"train_loss": 0.28805485367774963, "global_step": 57197, "epoch": 642, "lr": 8.957924064528278e-05}
{"train_loss": 0.2670338451862335, "global_step": 57198, "epoch": 642, "lr": 8.957888638917175e-05}
{"train_loss": 0.2726327180862427, "global_step": 57199, "epoch": 642, "lr": 8.957853212773984e-05}
{"train_loss": 0.29467594623565674, "global_step": 57200, "epoch": 642, "lr": 8.957817786098706e-05}
{"train_loss": 0.23627571761608124, "global_step": 57201, "epoch": 642, "lr": 8.957782358891348e-05}
{"train_loss": 0.26624056696891785, "global_step": 57202, "epoch": 642, "lr": 8.957746931151916e-05}
{"train_loss": 0.33015117049217224, "global_step": 57203, "epoch": 642, "lr": 8.957711502880411e-05}
{"train_loss": 0.36007699370384216, "global_step": 57204, "epoch": 642, "lr": 8.957676074076842e-05}
{"train_loss": 0.1737283170223236, "global_step": 57205, "epoch": 642, "lr": 8.95764064474121e-05}
{"train_loss": 0.37117522954940796, "global_step": 57206, "epoch": 642, "lr": 8.957605214873524e-05}
{"train_loss": 0.2297678291797638, "global_step": 57207, "epoch": 642, "lr": 8.957569784473783e-05}
{"train_loss": 0.1962508112192154, "global_step": 57208, "epoch": 642, "lr": 8.957534353541997e-05}
{"train_loss": 0.17932535707950592, "global_step": 57209, "epoch": 642, "lr": 8.957498922078168e-05}
{"train_loss": 0.2578885853290558, "global_step": 57210, "epoch": 642, "lr": 8.9574634900823e-05}
{"train_loss": 0.24087025225162506, "global_step": 57211, "epoch": 642, "lr": 8.957428057554402e-05}
{"train_loss": 0.23307205736637115, "global_step": 57212, "epoch": 642, "lr": 8.957392624494475e-05}
{"train_loss": 0.22695761919021606, "global_step": 57213, "epoch": 642, "lr": 8.957357190902524e-05}
{"train_loss": 0.27429717779159546, "global_step": 57214, "epoch": 642, "lr": 8.957321756778556e-05}
{"train_loss": 0.23981568217277527, "global_step": 57215, "epoch": 642, "lr": 8.957286322122573e-05}
{"train_loss": 0.3126443922519684, "global_step": 57216, "epoch": 642, "lr": 8.957250886934583e-05}
{"train_loss": 0.18210753798484802, "global_step": 57217, "epoch": 642, "lr": 8.957215451214586e-05}
{"train_loss": 0.18997149169445038, "global_step": 57218, "epoch": 642, "lr": 8.957180014962592e-05}
{"train_loss": 0.27307483553886414, "global_step": 57219, "epoch": 642, "lr": 8.957144578178602e-05}
{"train_loss": 0.27919241786003113, "global_step": 57220, "epoch": 642, "lr": 8.957109140862623e-05}
{"train_loss": 0.3047168552875519, "global_step": 57221, "epoch": 642, "lr": 8.957073703014659e-05}
{"train_loss": 0.2393016666173935, "global_step": 57222, "epoch": 642, "lr": 8.957038264634714e-05}
{"train_loss": 0.20975258946418762, "global_step": 57223, "epoch": 642, "lr": 8.957002825722794e-05}
{"train_loss": 0.24109862744808197, "global_step": 57224, "epoch": 642, "lr": 8.956967386278902e-05}
{"train_loss": 0.3096609115600586, "global_step": 57225, "epoch": 642, "lr": 8.956931946303046e-05}
{"train_loss": 0.26385012737820657, "global_step": 57226, "epoch": 642, "lr": 8.956896505795227e-05, "val_loss": 3.3495965003967285}
{"train_loss": 0.29704275727272034, "global_step": 57227, "epoch": 643, "lr": 8.956861064755453e-05}
{"train_loss": 0.19708794355392456, "global_step": 57228, "epoch": 643, "lr": 8.956825623183727e-05}
{"train_loss": 0.18061451613903046, "global_step": 57229, "epoch": 643, "lr": 8.956790181080052e-05}
{"train_loss": 0.3104099631309509, "global_step": 57230, "epoch": 643, "lr": 8.956754738444436e-05}
{"train_loss": 0.25910383462905884, "global_step": 57231, "epoch": 643, "lr": 8.956719295276883e-05}
{"train_loss": 0.2242315262556076, "global_step": 57232, "epoch": 643, "lr": 8.956683851577396e-05}
{"train_loss": 0.22001637518405914, "global_step": 57233, "epoch": 643, "lr": 8.956648407345983e-05}
{"train_loss": 0.1529007852077484, "global_step": 57234, "epoch": 643, "lr": 8.956612962582647e-05}
{"train_loss": 0.24508990347385406, "global_step": 57235, "epoch": 643, "lr": 8.95657751728739e-05}
{"train_loss": 0.1591949313879013, "global_step": 57236, "epoch": 643, "lr": 8.956542071460221e-05}
{"train_loss": 0.30775123834609985, "global_step": 57237, "epoch": 643, "lr": 8.956506625101145e-05}
{"train_loss": 0.18304778635501862, "global_step": 57238, "epoch": 643, "lr": 8.956471178210161e-05}
{"train_loss": 0.23000648617744446, "global_step": 57239, "epoch": 643, "lr": 8.956435730787281e-05}
{"train_loss": 0.27328550815582275, "global_step": 57240, "epoch": 643, "lr": 8.956400282832504e-05}
{"train_loss": 0.18503205478191376, "global_step": 57241, "epoch": 643, "lr": 8.956364834345838e-05}
{"train_loss": 0.23815521597862244, "global_step": 57242, "epoch": 643, "lr": 8.956329385327287e-05}
{"train_loss": 0.2876768708229065, "global_step": 57243, "epoch": 643, "lr": 8.956293935776857e-05}
{"train_loss": 0.17110763490200043, "global_step": 57244, "epoch": 643, "lr": 8.956258485694551e-05}
{"train_loss": 0.2214919477701187, "global_step": 57245, "epoch": 643, "lr": 8.956223035080373e-05}
{"train_loss": 0.2830938994884491, "global_step": 57246, "epoch": 643, "lr": 8.95618758393433e-05}
{"train_loss": 0.27615025639533997, "global_step": 57247, "epoch": 643, "lr": 8.956152132256425e-05}
{"train_loss": 0.2876325249671936, "global_step": 57248, "epoch": 643, "lr": 8.956116680046665e-05}
{"train_loss": 0.3286045789718628, "global_step": 57249, "epoch": 643, "lr": 8.956081227305051e-05}
{"train_loss": 0.27838334441185, "global_step": 57250, "epoch": 643, "lr": 8.956045774031592e-05}
{"train_loss": 0.18117794394493103, "global_step": 57251, "epoch": 643, "lr": 8.956010320226291e-05}
{"train_loss": 0.253377228975296, "global_step": 57252, "epoch": 643, "lr": 8.955974865889151e-05}
{"train_loss": 0.2772834300994873, "global_step": 57253, "epoch": 643, "lr": 8.95593941102018e-05}
{"train_loss": 0.21588559448719025, "global_step": 57254, "epoch": 643, "lr": 8.955903955619381e-05}
{"train_loss": 0.20911112427711487, "global_step": 57255, "epoch": 643, "lr": 8.955868499686759e-05}
{"train_loss": 0.2653051018714905, "global_step": 57256, "epoch": 643, "lr": 8.955833043222318e-05}
{"train_loss": 0.28026875853538513, "global_step": 57257, "epoch": 643, "lr": 8.955797586226065e-05}
{"train_loss": 0.2609359622001648, "global_step": 57258, "epoch": 643, "lr": 8.955762128698e-05}
{"train_loss": 0.2982242703437805, "global_step": 57259, "epoch": 643, "lr": 8.955726670638134e-05}
{"train_loss": 0.29993924498558044, "global_step": 57260, "epoch": 643, "lr": 8.955691212046467e-05}
{"train_loss": 0.2427249252796173, "global_step": 57261, "epoch": 643, "lr": 8.955655752923007e-05}
{"train_loss": 0.2745649814605713, "global_step": 57262, "epoch": 643, "lr": 8.955620293267757e-05}
{"train_loss": 0.29471641778945923, "global_step": 57263, "epoch": 643, "lr": 8.955584833080722e-05}
{"train_loss": 0.3347298204898834, "global_step": 57264, "epoch": 643, "lr": 8.955549372361906e-05}
{"train_loss": 0.18985970318317413, "global_step": 57265, "epoch": 643, "lr": 8.955513911111315e-05}
{"train_loss": 0.33955118060112, "global_step": 57266, "epoch": 643, "lr": 8.955478449328953e-05}
{"train_loss": 0.2101895958185196, "global_step": 57267, "epoch": 643, "lr": 8.955442987014826e-05}
{"train_loss": 0.3247428238391876, "global_step": 57268, "epoch": 643, "lr": 8.955407524168937e-05}
{"train_loss": 0.2487904131412506, "global_step": 57269, "epoch": 643, "lr": 8.955372060791291e-05}
{"train_loss": 0.21954743564128876, "global_step": 57270, "epoch": 643, "lr": 8.955336596881896e-05}
{"train_loss": 0.31710031628608704, "global_step": 57271, "epoch": 643, "lr": 8.955301132440752e-05}
{"train_loss": 0.20104409754276276, "global_step": 57272, "epoch": 643, "lr": 8.955265667467868e-05}
{"train_loss": 0.22120249271392822, "global_step": 57273, "epoch": 643, "lr": 8.955230201963245e-05}
{"train_loss": 0.24671250581741333, "global_step": 57274, "epoch": 643, "lr": 8.955194735926888e-05}
{"train_loss": 0.32571256160736084, "global_step": 57275, "epoch": 643, "lr": 8.955159269358807e-05}
{"train_loss": 0.2714972496032715, "global_step": 57276, "epoch": 643, "lr": 8.955123802259e-05}
{"train_loss": 0.15736660361289978, "global_step": 57277, "epoch": 643, "lr": 8.955088334627477e-05}
{"train_loss": 0.22979798913002014, "global_step": 57278, "epoch": 643, "lr": 8.95505286646424e-05}
{"train_loss": 0.20066951215267181, "global_step": 57279, "epoch": 643, "lr": 8.955017397769294e-05}
{"train_loss": 0.17301617562770844, "global_step": 57280, "epoch": 643, "lr": 8.954981928542643e-05}
{"train_loss": 0.2658930718898773, "global_step": 57281, "epoch": 643, "lr": 8.954946458784295e-05}
{"train_loss": 0.299966424703598, "global_step": 57282, "epoch": 643, "lr": 8.954910988494251e-05}
{"train_loss": 0.2901113033294678, "global_step": 57283, "epoch": 643, "lr": 8.954875517672518e-05}
{"train_loss": 0.3071599006652832, "global_step": 57284, "epoch": 643, "lr": 8.9548400463191e-05}
{"train_loss": 0.26017260551452637, "global_step": 57285, "epoch": 643, "lr": 8.954804574434004e-05}
{"train_loss": 0.25811517238616943, "global_step": 57286, "epoch": 643, "lr": 8.954769102017229e-05}
{"train_loss": 0.2810955345630646, "global_step": 57287, "epoch": 643, "lr": 8.954733629068786e-05}
{"train_loss": 0.3725629150867462, "global_step": 57288, "epoch": 643, "lr": 8.954698155588677e-05}
{"train_loss": 0.23371225595474243, "global_step": 57289, "epoch": 643, "lr": 8.954662681576906e-05}
{"train_loss": 0.2699985206127167, "global_step": 57290, "epoch": 643, "lr": 8.954627207033481e-05}
{"train_loss": 0.2829059958457947, "global_step": 57291, "epoch": 643, "lr": 8.954591731958403e-05}
{"train_loss": 0.2669712007045746, "global_step": 57292, "epoch": 643, "lr": 8.954556256351678e-05}
{"train_loss": 0.2925603687763214, "global_step": 57293, "epoch": 643, "lr": 8.954520780213311e-05}
{"train_loss": 0.27265283465385437, "global_step": 57294, "epoch": 643, "lr": 8.954485303543307e-05}
{"train_loss": 0.24047651886940002, "global_step": 57295, "epoch": 643, "lr": 8.954449826341673e-05}
{"train_loss": 0.23544073104858398, "global_step": 57296, "epoch": 643, "lr": 8.95441434860841e-05}
{"train_loss": 0.2241479903459549, "global_step": 57297, "epoch": 643, "lr": 8.954378870343522e-05}
{"train_loss": 0.23144391179084778, "global_step": 57298, "epoch": 643, "lr": 8.954343391547017e-05}
{"train_loss": 0.2397150844335556, "global_step": 57299, "epoch": 643, "lr": 8.9543079122189e-05}
{"train_loss": 0.20067395269870758, "global_step": 57300, "epoch": 643, "lr": 8.954272432359174e-05}
{"train_loss": 0.2900490164756775, "global_step": 57301, "epoch": 643, "lr": 8.954236951967845e-05}
{"train_loss": 0.2588176429271698, "global_step": 57302, "epoch": 643, "lr": 8.954201471044916e-05}
{"train_loss": 0.22133362293243408, "global_step": 57303, "epoch": 643, "lr": 8.954165989590393e-05}
{"train_loss": 0.31448084115982056, "global_step": 57304, "epoch": 643, "lr": 8.95413050760428e-05}
{"train_loss": 0.26035642623901367, "global_step": 57305, "epoch": 643, "lr": 8.954095025086584e-05}
{"train_loss": 0.3131357431411743, "global_step": 57306, "epoch": 643, "lr": 8.954059542037307e-05}
{"train_loss": 0.298902302980423, "global_step": 57307, "epoch": 643, "lr": 8.954024058456455e-05}
{"train_loss": 0.29270145297050476, "global_step": 57308, "epoch": 643, "lr": 8.953988574344032e-05}
{"train_loss": 0.21181875467300415, "global_step": 57309, "epoch": 643, "lr": 8.953953089700044e-05}
{"train_loss": 0.30212119221687317, "global_step": 57310, "epoch": 643, "lr": 8.953917604524495e-05}
{"train_loss": 0.18130801618099213, "global_step": 57311, "epoch": 643, "lr": 8.953882118817391e-05}
{"train_loss": 0.30907338857650757, "global_step": 57312, "epoch": 643, "lr": 8.953846632578735e-05}
{"train_loss": 0.2897180914878845, "global_step": 57313, "epoch": 643, "lr": 8.953811145808531e-05}
{"train_loss": 0.3893972933292389, "global_step": 57314, "epoch": 643, "lr": 8.953775658506786e-05}
{"train_loss": 0.25745745756652916, "global_step": 57315, "epoch": 643, "lr": 8.953740170673506e-05, "val_loss": 3.245180368423462}
{"train_loss": 0.20879755914211273, "global_step": 57316, "epoch": 644, "lr": 8.953704682308692e-05}
{"train_loss": 0.28900885581970215, "global_step": 57317, "epoch": 644, "lr": 8.953669193412353e-05}
{"train_loss": 0.2034929245710373, "global_step": 57318, "epoch": 644, "lr": 8.953633703984489e-05}
{"train_loss": 0.227852463722229, "global_step": 57319, "epoch": 644, "lr": 8.953598214025107e-05}
{"train_loss": 0.24454337358474731, "global_step": 57320, "epoch": 644, "lr": 8.953562723534212e-05}
{"train_loss": 0.44648420810699463, "global_step": 57321, "epoch": 644, "lr": 8.95352723251181e-05}
{"train_loss": 0.2582889497280121, "global_step": 57322, "epoch": 644, "lr": 8.953491740957905e-05}
{"train_loss": 0.24861161410808563, "global_step": 57323, "epoch": 644, "lr": 8.953456248872498e-05}
{"train_loss": 0.24384792149066925, "global_step": 57324, "epoch": 644, "lr": 8.9534207562556e-05}
{"train_loss": 0.2527925372123718, "global_step": 57325, "epoch": 644, "lr": 8.953385263107212e-05}
{"train_loss": 0.36477914452552795, "global_step": 57326, "epoch": 644, "lr": 8.953349769427339e-05}
{"train_loss": 0.2044568508863449, "global_step": 57327, "epoch": 644, "lr": 8.953314275215988e-05}
{"train_loss": 0.24998047947883606, "global_step": 57328, "epoch": 644, "lr": 8.953278780473159e-05}
{"train_loss": 0.21218889951705933, "global_step": 57329, "epoch": 644, "lr": 8.953243285198861e-05}
{"train_loss": 0.34058597683906555, "global_step": 57330, "epoch": 644, "lr": 8.953207789393099e-05}
{"train_loss": 0.3288933336734772, "global_step": 57331, "epoch": 644, "lr": 8.953172293055875e-05}
{"train_loss": 0.25135737657546997, "global_step": 57332, "epoch": 644, "lr": 8.953136796187196e-05}
{"train_loss": 0.2856600284576416, "global_step": 57333, "epoch": 644, "lr": 8.953101298787066e-05}
{"train_loss": 0.326265811920166, "global_step": 57334, "epoch": 644, "lr": 8.953065800855489e-05}
{"train_loss": 0.2707740068435669, "global_step": 57335, "epoch": 644, "lr": 8.953030302392471e-05}
{"train_loss": 0.2734859883785248, "global_step": 57336, "epoch": 644, "lr": 8.952994803398017e-05}
{"train_loss": 0.2699826955795288, "global_step": 57337, "epoch": 644, "lr": 8.952959303872128e-05}
{"train_loss": 0.22658312320709229, "global_step": 57338, "epoch": 644, "lr": 8.952923803814814e-05}
{"train_loss": 0.2861618101596832, "global_step": 57339, "epoch": 644, "lr": 8.952888303226078e-05}
{"train_loss": 0.25536516308784485, "global_step": 57340, "epoch": 644, "lr": 8.952852802105925e-05}
{"train_loss": 0.2961644232273102, "global_step": 57341, "epoch": 644, "lr": 8.952817300454357e-05}
{"train_loss": 0.19125017523765564, "global_step": 57342, "epoch": 644, "lr": 8.952781798271382e-05}
{"train_loss": 0.24374280869960785, "global_step": 57343, "epoch": 644, "lr": 8.952746295557004e-05}
{"train_loss": 0.2267933487892151, "global_step": 57344, "epoch": 644, "lr": 8.952710792311226e-05}
{"train_loss": 0.27630260586738586, "global_step": 57345, "epoch": 644, "lr": 8.952675288534056e-05}
{"train_loss": 0.2800835967063904, "global_step": 57346, "epoch": 644, "lr": 8.952639784225496e-05}
{"train_loss": 0.3032618463039398, "global_step": 57347, "epoch": 644, "lr": 8.952604279385552e-05}
{"train_loss": 0.3162182569503784, "global_step": 57348, "epoch": 644, "lr": 8.952568774014227e-05}
{"train_loss": 0.21276958286762238, "global_step": 57349, "epoch": 644, "lr": 8.952533268111529e-05}
{"train_loss": 0.30345600843429565, "global_step": 57350, "epoch": 644, "lr": 8.952497761677461e-05}
{"train_loss": 0.294538676738739, "global_step": 57351, "epoch": 644, "lr": 8.952462254712028e-05}
{"train_loss": 0.16924111545085907, "global_step": 57352, "epoch": 644, "lr": 8.952426747215234e-05}
{"train_loss": 0.31749117374420166, "global_step": 57353, "epoch": 644, "lr": 8.952391239187085e-05}
{"train_loss": 0.30374613404273987, "global_step": 57354, "epoch": 644, "lr": 8.952355730627585e-05}
{"train_loss": 0.25428199768066406, "global_step": 57355, "epoch": 644, "lr": 8.952320221536738e-05}
{"train_loss": 0.33790794014930725, "global_step": 57356, "epoch": 644, "lr": 8.952284711914552e-05}
{"train_loss": 0.3001430928707123, "global_step": 57357, "epoch": 644, "lr": 8.952249201761028e-05}
{"train_loss": 0.21077954769134521, "global_step": 57358, "epoch": 644, "lr": 8.952213691076171e-05}
{"train_loss": 0.21069028973579407, "global_step": 57359, "epoch": 644, "lr": 8.95217817985999e-05}
{"train_loss": 0.24032647907733917, "global_step": 57360, "epoch": 644, "lr": 8.952142668112484e-05}
{"train_loss": 0.19769814610481262, "global_step": 57361, "epoch": 644, "lr": 8.952107155833662e-05}
{"train_loss": 0.2598525583744049, "global_step": 57362, "epoch": 644, "lr": 8.952071643023527e-05}
{"train_loss": 0.1735868602991104, "global_step": 57363, "epoch": 644, "lr": 8.952036129682084e-05}
{"train_loss": 0.2660488486289978, "global_step": 57364, "epoch": 644, "lr": 8.952000615809339e-05}
{"train_loss": 0.34441328048706055, "global_step": 57365, "epoch": 644, "lr": 8.951965101405294e-05}
{"train_loss": 0.17956000566482544, "global_step": 57366, "epoch": 644, "lr": 8.951929586469956e-05}
{"train_loss": 0.2590772211551666, "global_step": 57367, "epoch": 644, "lr": 8.95189407100333e-05}
{"train_loss": 0.21382561326026917, "global_step": 57368, "epoch": 644, "lr": 8.95185855500542e-05}
{"train_loss": 0.23021362721920013, "global_step": 57369, "epoch": 644, "lr": 8.95182303847623e-05}
{"train_loss": 0.2877466380596161, "global_step": 57370, "epoch": 644, "lr": 8.951787521415765e-05}
{"train_loss": 0.33365607261657715, "global_step": 57371, "epoch": 644, "lr": 8.951752003824033e-05}
{"train_loss": 0.2017613649368286, "global_step": 57372, "epoch": 644, "lr": 8.951716485701033e-05}
{"train_loss": 0.27121201157569885, "global_step": 57373, "epoch": 644, "lr": 8.951680967046774e-05}
{"train_loss": 0.23980587720870972, "global_step": 57374, "epoch": 644, "lr": 8.951645447861258e-05}
{"train_loss": 0.2640417814254761, "global_step": 57375, "epoch": 644, "lr": 8.951609928144495e-05}
{"train_loss": 0.23389266431331635, "global_step": 57376, "epoch": 644, "lr": 8.951574407896484e-05}
{"train_loss": 0.32416462898254395, "global_step": 57377, "epoch": 644, "lr": 8.951538887117232e-05}
{"train_loss": 0.32950955629348755, "global_step": 57378, "epoch": 644, "lr": 8.951503365806746e-05}
{"train_loss": 0.21308693289756775, "global_step": 57379, "epoch": 644, "lr": 8.951467843965027e-05}
{"train_loss": 0.2165849208831787, "global_step": 57380, "epoch": 644, "lr": 8.95143232159208e-05}
{"train_loss": 0.19868235290050507, "global_step": 57381, "epoch": 644, "lr": 8.951396798687912e-05}
{"train_loss": 0.3032080829143524, "global_step": 57382, "epoch": 644, "lr": 8.951361275252528e-05}
{"train_loss": 0.16219009459018707, "global_step": 57383, "epoch": 644, "lr": 8.95132575128593e-05}
{"train_loss": 0.2600953280925751, "global_step": 57384, "epoch": 644, "lr": 8.951290226788125e-05}
{"train_loss": 0.2717011272907257, "global_step": 57385, "epoch": 644, "lr": 8.951254701759117e-05}
{"train_loss": 0.22124408185482025, "global_step": 57386, "epoch": 644, "lr": 8.95121917619891e-05}
{"train_loss": 0.25352299213409424, "global_step": 57387, "epoch": 644, "lr": 8.951183650107512e-05}
{"train_loss": 0.30791324377059937, "global_step": 57388, "epoch": 644, "lr": 8.951148123484924e-05}
{"train_loss": 0.20493115484714508, "global_step": 57389, "epoch": 644, "lr": 8.951112596331154e-05}
{"train_loss": 0.19614510238170624, "global_step": 57390, "epoch": 644, "lr": 8.951077068646202e-05}
{"train_loss": 0.33801233768463135, "global_step": 57391, "epoch": 644, "lr": 8.951041540430079e-05}
{"train_loss": 0.2546979784965515, "global_step": 57392, "epoch": 644, "lr": 8.951006011682786e-05}
{"train_loss": 0.36932119727134705, "global_step": 57393, "epoch": 644, "lr": 8.950970482404326e-05}
{"train_loss": 0.2867090106010437, "global_step": 57394, "epoch": 644, "lr": 8.950934952594708e-05}
{"train_loss": 0.24120798707008362, "global_step": 57395, "epoch": 644, "lr": 8.950899422253935e-05}
{"train_loss": 0.22629305720329285, "global_step": 57396, "epoch": 644, "lr": 8.950863891382012e-05}
{"train_loss": 0.22364355623722076, "global_step": 57397, "epoch": 644, "lr": 8.950828359978942e-05}
{"train_loss": 0.19399040937423706, "global_step": 57398, "epoch": 644, "lr": 8.950792828044734e-05}
{"train_loss": 0.3018001616001129, "global_step": 57399, "epoch": 644, "lr": 8.950757295579389e-05}
{"train_loss": 0.25808587670326233, "global_step": 57400, "epoch": 644, "lr": 8.950721762582912e-05}
{"train_loss": 0.21490469574928284, "global_step": 57401, "epoch": 644, "lr": 8.950686229055307e-05}
{"train_loss": 0.3253280222415924, "global_step": 57402, "epoch": 644, "lr": 8.950650694996583e-05}
{"train_loss": 0.2528980076313019, "global_step": 57403, "epoch": 644, "lr": 8.950615160406741e-05}
{"train_loss": 0.2600507320982687, "global_step": 57404, "epoch": 644, "lr": 8.950579625285787e-05, "val_loss": 3.2291195392608643}
{"train_loss": 0.2977525591850281, "global_step": 57405, "epoch": 645, "lr": 8.950544089633727e-05}
{"train_loss": 0.18794091045856476, "global_step": 57406, "epoch": 645, "lr": 8.950508553450563e-05}
{"train_loss": 0.23913593590259552, "global_step": 57407, "epoch": 645, "lr": 8.950473016736303e-05}
{"train_loss": 0.2636590898036957, "global_step": 57408, "epoch": 645, "lr": 8.950437479490948e-05}
{"train_loss": 0.24536794424057007, "global_step": 57409, "epoch": 645, "lr": 8.950401941714506e-05}
{"train_loss": 0.17307248711585999, "global_step": 57410, "epoch": 645, "lr": 8.95036640340698e-05}
{"train_loss": 0.2798826992511749, "global_step": 57411, "epoch": 645, "lr": 8.950330864568376e-05}
{"train_loss": 0.2469603270292282, "global_step": 57412, "epoch": 645, "lr": 8.950295325198697e-05}
{"train_loss": 0.1361166089773178, "global_step": 57413, "epoch": 645, "lr": 8.95025978529795e-05}
{"train_loss": 0.39700835943222046, "global_step": 57414, "epoch": 645, "lr": 8.950224244866139e-05}
{"train_loss": 0.29615917801856995, "global_step": 57415, "epoch": 645, "lr": 8.950188703903269e-05}
{"train_loss": 0.2739356756210327, "global_step": 57416, "epoch": 645, "lr": 8.950153162409343e-05}
{"train_loss": 0.24263744056224823, "global_step": 57417, "epoch": 645, "lr": 8.950117620384367e-05}
{"train_loss": 0.21383383870124817, "global_step": 57418, "epoch": 645, "lr": 8.950082077828345e-05}
{"train_loss": 0.3481263518333435, "global_step": 57419, "epoch": 645, "lr": 8.950046534741285e-05}
{"train_loss": 0.15336130559444427, "global_step": 57420, "epoch": 645, "lr": 8.950010991123188e-05}
{"train_loss": 0.22236935794353485, "global_step": 57421, "epoch": 645, "lr": 8.94997544697406e-05}
{"train_loss": 0.27761325240135193, "global_step": 57422, "epoch": 645, "lr": 8.949939902293907e-05}
{"train_loss": 0.22053013741970062, "global_step": 57423, "epoch": 645, "lr": 8.949904357082732e-05}
{"train_loss": 0.2129332274198532, "global_step": 57424, "epoch": 645, "lr": 8.949868811340541e-05}
{"train_loss": 0.1818585991859436, "global_step": 57425, "epoch": 645, "lr": 8.949833265067337e-05}
{"train_loss": 0.2095024287700653, "global_step": 57426, "epoch": 645, "lr": 8.949797718263127e-05}
{"train_loss": 0.14145374298095703, "global_step": 57427, "epoch": 645, "lr": 8.949762170927914e-05}
{"train_loss": 0.26796165108680725, "global_step": 57428, "epoch": 645, "lr": 8.949726623061705e-05}
{"train_loss": 0.27496784925460815, "global_step": 57429, "epoch": 645, "lr": 8.949691074664502e-05}
{"train_loss": 0.17123639583587646, "global_step": 57430, "epoch": 645, "lr": 8.949655525736312e-05}
{"train_loss": 0.28861960768699646, "global_step": 57431, "epoch": 645, "lr": 8.94961997627714e-05}
{"train_loss": 0.19506561756134033, "global_step": 57432, "epoch": 645, "lr": 8.949584426286987e-05}
{"train_loss": 0.20384079217910767, "global_step": 57433, "epoch": 645, "lr": 8.949548875765864e-05}
{"train_loss": 0.2685432732105255, "global_step": 57434, "epoch": 645, "lr": 8.94951332471377e-05}
{"train_loss": 0.197952538728714, "global_step": 57435, "epoch": 645, "lr": 8.949477773130711e-05}
{"train_loss": 0.2588210999965668, "global_step": 57436, "epoch": 645, "lr": 8.949442221016696e-05}
{"train_loss": 0.28347471356391907, "global_step": 57437, "epoch": 645, "lr": 8.949406668371723e-05}
{"train_loss": 0.29095491766929626, "global_step": 57438, "epoch": 645, "lr": 8.949371115195803e-05}
{"train_loss": 0.2717958986759186, "global_step": 57439, "epoch": 645, "lr": 8.949335561488937e-05}
{"train_loss": 0.2911253273487091, "global_step": 57440, "epoch": 645, "lr": 8.949300007251131e-05}
{"train_loss": 0.19963796436786652, "global_step": 57441, "epoch": 645, "lr": 8.949264452482391e-05}
{"train_loss": 0.27078747749328613, "global_step": 57442, "epoch": 645, "lr": 8.94922889718272e-05}
{"train_loss": 0.32927897572517395, "global_step": 57443, "epoch": 645, "lr": 8.949193341352123e-05}
{"train_loss": 0.2570377290248871, "global_step": 57444, "epoch": 645, "lr": 8.949157784990606e-05}
{"train_loss": 0.23672093451023102, "global_step": 57445, "epoch": 645, "lr": 8.949122228098171e-05}
{"train_loss": 0.19472503662109375, "global_step": 57446, "epoch": 645, "lr": 8.949086670674827e-05}
{"train_loss": 0.2091684341430664, "global_step": 57447, "epoch": 645, "lr": 8.949051112720574e-05}
{"train_loss": 0.3163681626319885, "global_step": 57448, "epoch": 645, "lr": 8.949015554235421e-05}
{"train_loss": 0.3275780975818634, "global_step": 57449, "epoch": 645, "lr": 8.94897999521937e-05}
{"train_loss": 0.2504887580871582, "global_step": 57450, "epoch": 645, "lr": 8.948944435672426e-05}
{"train_loss": 0.3266020119190216, "global_step": 57451, "epoch": 645, "lr": 8.948908875594598e-05}
{"train_loss": 0.3100428581237793, "global_step": 57452, "epoch": 645, "lr": 8.948873314985884e-05}
{"train_loss": 0.27157583832740784, "global_step": 57453, "epoch": 645, "lr": 8.948837753846293e-05}
{"train_loss": 0.3101939260959625, "global_step": 57454, "epoch": 645, "lr": 8.94880219217583e-05}
{"train_loss": 0.2798880338668823, "global_step": 57455, "epoch": 645, "lr": 8.948766629974497e-05}
{"train_loss": 0.2625401020050049, "global_step": 57456, "epoch": 645, "lr": 8.9487310672423e-05}
{"train_loss": 0.1805553138256073, "global_step": 57457, "epoch": 645, "lr": 8.948695503979245e-05}
{"train_loss": 0.3105958104133606, "global_step": 57458, "epoch": 645, "lr": 8.948659940185338e-05}
{"train_loss": 0.29192012548446655, "global_step": 57459, "epoch": 645, "lr": 8.948624375860579e-05}
{"train_loss": 0.28892648220062256, "global_step": 57460, "epoch": 645, "lr": 8.948588811004976e-05}
{"train_loss": 0.27732735872268677, "global_step": 57461, "epoch": 645, "lr": 8.948553245618535e-05}
{"train_loss": 0.26038557291030884, "global_step": 57462, "epoch": 645, "lr": 8.948517679701258e-05}
{"train_loss": 0.2897563576698303, "global_step": 57463, "epoch": 645, "lr": 8.948482113253152e-05}
{"train_loss": 0.3741448223590851, "global_step": 57464, "epoch": 645, "lr": 8.948446546274219e-05}
{"train_loss": 0.22253943979740143, "global_step": 57465, "epoch": 645, "lr": 8.948410978764466e-05}
{"train_loss": 0.23837623000144958, "global_step": 57466, "epoch": 645, "lr": 8.948375410723899e-05}
{"train_loss": 0.30149364471435547, "global_step": 57467, "epoch": 645, "lr": 8.948339842152519e-05}
{"train_loss": 0.22402355074882507, "global_step": 57468, "epoch": 645, "lr": 8.948304273050334e-05}
{"train_loss": 0.2134476602077484, "global_step": 57469, "epoch": 645, "lr": 8.948268703417346e-05}
{"train_loss": 0.2195434272289276, "global_step": 57470, "epoch": 645, "lr": 8.948233133253564e-05}
{"train_loss": 0.3433729410171509, "global_step": 57471, "epoch": 645, "lr": 8.948197562558987e-05}
{"train_loss": 0.3647686839103699, "global_step": 57472, "epoch": 645, "lr": 8.948161991333626e-05}
{"train_loss": 0.22956885397434235, "global_step": 57473, "epoch": 645, "lr": 8.94812641957748e-05}
{"train_loss": 0.21905474364757538, "global_step": 57474, "epoch": 645, "lr": 8.948090847290558e-05}
{"train_loss": 0.1971084475517273, "global_step": 57475, "epoch": 645, "lr": 8.948055274472864e-05}
{"train_loss": 0.23785452544689178, "global_step": 57476, "epoch": 645, "lr": 8.948019701124402e-05}
{"train_loss": 0.264145165681839, "global_step": 57477, "epoch": 645, "lr": 8.947984127245174e-05}
{"train_loss": 0.22496342658996582, "global_step": 57478, "epoch": 645, "lr": 8.947948552835191e-05}
{"train_loss": 0.18789204955101013, "global_step": 57479, "epoch": 645, "lr": 8.947912977894454e-05}
{"train_loss": 0.23853957653045654, "global_step": 57480, "epoch": 645, "lr": 8.947877402422967e-05}
{"train_loss": 0.27377161383628845, "global_step": 57481, "epoch": 645, "lr": 8.947841826420737e-05}
{"train_loss": 0.2567979097366333, "global_step": 57482, "epoch": 645, "lr": 8.947806249887767e-05}
{"train_loss": 0.20679526031017303, "global_step": 57483, "epoch": 645, "lr": 8.947770672824062e-05}
{"train_loss": 0.20430943369865417, "global_step": 57484, "epoch": 645, "lr": 8.947735095229627e-05}
{"train_loss": 0.1884825974702835, "global_step": 57485, "epoch": 645, "lr": 8.947699517104471e-05}
{"train_loss": 0.26042869687080383, "global_step": 57486, "epoch": 645, "lr": 8.947663938448592e-05}
{"train_loss": 0.2536852955818176, "global_step": 57487, "epoch": 645, "lr": 8.947628359261996e-05}
{"train_loss": 0.13202601671218872, "global_step": 57488, "epoch": 645, "lr": 8.947592779544694e-05}
{"train_loss": 0.2774917483329773, "global_step": 57489, "epoch": 645, "lr": 8.947557199296682e-05}
{"train_loss": 0.33600011467933655, "global_step": 57490, "epoch": 645, "lr": 8.947521618517971e-05}
{"train_loss": 0.32261550426483154, "global_step": 57491, "epoch": 645, "lr": 8.947486037208563e-05}
{"train_loss": 0.2487131655216217, "global_step": 57492, "epoch": 645, "lr": 8.947450455368464e-05}
{"train_loss": 0.2529041601365872, "global_step": 57493, "epoch": 645, "lr": 8.94741487299768e-05, "val_loss": 3.14939284324646, "train_action_mse_error": 18.274158477783203}
{"train_loss": 0.20030975341796875, "global_step": 57494, "epoch": 646, "lr": 8.947379290096212e-05}
{"train_loss": 0.21002532541751862, "global_step": 57495, "epoch": 646, "lr": 8.947343706664068e-05}
{"train_loss": 0.3169556260108948, "global_step": 57496, "epoch": 646, "lr": 8.947308122701252e-05}
{"train_loss": 0.2211878001689911, "global_step": 57497, "epoch": 646, "lr": 8.947272538207767e-05}
{"train_loss": 0.16928012669086456, "global_step": 57498, "epoch": 646, "lr": 8.947236953183622e-05}
{"train_loss": 0.26769325137138367, "global_step": 57499, "epoch": 646, "lr": 8.947201367628817e-05}
{"train_loss": 0.2204297035932541, "global_step": 57500, "epoch": 646, "lr": 8.947165781543358e-05}
{"train_loss": 0.302320659160614, "global_step": 57501, "epoch": 646, "lr": 8.947130194927252e-05}
{"train_loss": 0.21315829455852509, "global_step": 57502, "epoch": 646, "lr": 8.947094607780503e-05}
{"train_loss": 0.23372472822666168, "global_step": 57503, "epoch": 646, "lr": 8.947059020103115e-05}
{"train_loss": 0.17619860172271729, "global_step": 57504, "epoch": 646, "lr": 8.947023431895092e-05}
{"train_loss": 0.24284884333610535, "global_step": 57505, "epoch": 646, "lr": 8.94698784315644e-05}
{"train_loss": 0.2209300547838211, "global_step": 57506, "epoch": 646, "lr": 8.946952253887164e-05}
{"train_loss": 0.26187795400619507, "global_step": 57507, "epoch": 646, "lr": 8.94691666408727e-05}
{"train_loss": 0.15727879106998444, "global_step": 57508, "epoch": 646, "lr": 8.94688107375676e-05}
{"train_loss": 0.172000452876091, "global_step": 57509, "epoch": 646, "lr": 8.946845482895638e-05}
{"train_loss": 0.22077596187591553, "global_step": 57510, "epoch": 646, "lr": 8.946809891503913e-05}
{"train_loss": 0.22934742271900177, "global_step": 57511, "epoch": 646, "lr": 8.946774299581586e-05}
{"train_loss": 0.17867298424243927, "global_step": 57512, "epoch": 646, "lr": 8.946738707128666e-05}
{"train_loss": 0.2558480501174927, "global_step": 57513, "epoch": 646, "lr": 8.946703114145152e-05}
{"train_loss": 0.18882343173027039, "global_step": 57514, "epoch": 646, "lr": 8.946667520631054e-05}
{"train_loss": 0.20890037715435028, "global_step": 57515, "epoch": 646, "lr": 8.946631926586373e-05}
{"train_loss": 0.2013024091720581, "global_step": 57516, "epoch": 646, "lr": 8.946596332011116e-05}
{"train_loss": 0.14499503374099731, "global_step": 57517, "epoch": 646, "lr": 8.946560736905288e-05}
{"train_loss": 0.201848566532135, "global_step": 57518, "epoch": 646, "lr": 8.946525141268892e-05}
{"train_loss": 0.2382134348154068, "global_step": 57519, "epoch": 646, "lr": 8.946489545101933e-05}
{"train_loss": 0.21029146015644073, "global_step": 57520, "epoch": 646, "lr": 8.946453948404418e-05}
{"train_loss": 0.11690323054790497, "global_step": 57521, "epoch": 646, "lr": 8.94641835117635e-05}
{"train_loss": 0.2188519537448883, "global_step": 57522, "epoch": 646, "lr": 8.946382753417735e-05}
{"train_loss": 0.1925521045923233, "global_step": 57523, "epoch": 646, "lr": 8.946347155128576e-05}
{"train_loss": 0.23692144453525543, "global_step": 57524, "epoch": 646, "lr": 8.94631155630888e-05}
{"train_loss": 0.2119581401348114, "global_step": 57525, "epoch": 646, "lr": 8.94627595695865e-05}
{"train_loss": 0.3137456476688385, "global_step": 57526, "epoch": 646, "lr": 8.94624035707789e-05}
{"train_loss": 0.24474506080150604, "global_step": 57527, "epoch": 646, "lr": 8.946204756666608e-05}
{"train_loss": 0.1724528968334198, "global_step": 57528, "epoch": 646, "lr": 8.946169155724803e-05}
{"train_loss": 0.1513945460319519, "global_step": 57529, "epoch": 646, "lr": 8.946133554252487e-05}
{"train_loss": 0.2247188687324524, "global_step": 57530, "epoch": 646, "lr": 8.946097952249662e-05}
{"train_loss": 0.24643433094024658, "global_step": 57531, "epoch": 646, "lr": 8.946062349716331e-05}
{"train_loss": 0.2314157336950302, "global_step": 57532, "epoch": 646, "lr": 8.9460267466525e-05}
{"train_loss": 0.2610165774822235, "global_step": 57533, "epoch": 646, "lr": 8.945991143058175e-05}
{"train_loss": 0.24631589651107788, "global_step": 57534, "epoch": 646, "lr": 8.945955538933357e-05}
{"train_loss": 0.27944493293762207, "global_step": 57535, "epoch": 646, "lr": 8.945919934278055e-05}
{"train_loss": 0.316192626953125, "global_step": 57536, "epoch": 646, "lr": 8.945884329092272e-05}
{"train_loss": 0.2573355436325073, "global_step": 57537, "epoch": 646, "lr": 8.945848723376013e-05}
{"train_loss": 0.2667296528816223, "global_step": 57538, "epoch": 646, "lr": 8.945813117129283e-05}
{"train_loss": 0.1943701058626175, "global_step": 57539, "epoch": 646, "lr": 8.945777510352084e-05}
{"train_loss": 0.23454661667346954, "global_step": 57540, "epoch": 646, "lr": 8.945741903044426e-05}
{"train_loss": 0.235224649310112, "global_step": 57541, "epoch": 646, "lr": 8.945706295206313e-05}
{"train_loss": 0.19813872873783112, "global_step": 57542, "epoch": 646, "lr": 8.945670686837744e-05}
{"train_loss": 0.21559180319309235, "global_step": 57543, "epoch": 646, "lr": 8.94563507793873e-05}
{"train_loss": 0.2490430325269699, "global_step": 57544, "epoch": 646, "lr": 8.945599468509272e-05}
{"train_loss": 0.2035761922597885, "global_step": 57545, "epoch": 646, "lr": 8.945563858549377e-05}
{"train_loss": 0.2691595256328583, "global_step": 57546, "epoch": 646, "lr": 8.945528248059049e-05}
{"train_loss": 0.19899193942546844, "global_step": 57547, "epoch": 646, "lr": 8.945492637038292e-05}
{"train_loss": 0.3176170587539673, "global_step": 57548, "epoch": 646, "lr": 8.945457025487112e-05}
{"train_loss": 0.27608081698417664, "global_step": 57549, "epoch": 646, "lr": 8.945421413405515e-05}
{"train_loss": 0.24342116713523865, "global_step": 57550, "epoch": 646, "lr": 8.945385800793502e-05}
{"train_loss": 0.21037383377552032, "global_step": 57551, "epoch": 646, "lr": 8.945350187651081e-05}
{"train_loss": 0.28261011838912964, "global_step": 57552, "epoch": 646, "lr": 8.945314573978256e-05}
{"train_loss": 0.22274869680404663, "global_step": 57553, "epoch": 646, "lr": 8.945278959775031e-05}
{"train_loss": 0.2537288963794708, "global_step": 57554, "epoch": 646, "lr": 8.945243345041411e-05}
{"train_loss": 0.18982860445976257, "global_step": 57555, "epoch": 646, "lr": 8.9452077297774e-05}
{"train_loss": 0.20751237869262695, "global_step": 57556, "epoch": 646, "lr": 8.945172113983007e-05}
{"train_loss": 0.2884116768836975, "global_step": 57557, "epoch": 646, "lr": 8.945136497658232e-05}
{"train_loss": 0.21844202280044556, "global_step": 57558, "epoch": 646, "lr": 8.945100880803081e-05}
{"train_loss": 0.27798640727996826, "global_step": 57559, "epoch": 646, "lr": 8.94506526341756e-05}
{"train_loss": 0.18174642324447632, "global_step": 57560, "epoch": 646, "lr": 8.945029645501674e-05}
{"train_loss": 0.22219885885715485, "global_step": 57561, "epoch": 646, "lr": 8.944994027055425e-05}
{"train_loss": 0.3654305636882782, "global_step": 57562, "epoch": 646, "lr": 8.94495840807882e-05}
{"train_loss": 0.24660487473011017, "global_step": 57563, "epoch": 646, "lr": 8.944922788571865e-05}
{"train_loss": 0.2066562920808792, "global_step": 57564, "epoch": 646, "lr": 8.944887168534561e-05}
{"train_loss": 0.2140020877122879, "global_step": 57565, "epoch": 646, "lr": 8.944851547966916e-05}
{"train_loss": 0.30800893902778625, "global_step": 57566, "epoch": 646, "lr": 8.944815926868934e-05}
{"train_loss": 0.2659049928188324, "global_step": 57567, "epoch": 646, "lr": 8.94478030524062e-05}
{"train_loss": 0.24941565096378326, "global_step": 57568, "epoch": 646, "lr": 8.944744683081978e-05}
{"train_loss": 0.2558552920818329, "global_step": 57569, "epoch": 646, "lr": 8.944709060393013e-05}
{"train_loss": 0.2881740927696228, "global_step": 57570, "epoch": 646, "lr": 8.94467343717373e-05}
{"train_loss": 0.22058691084384918, "global_step": 57571, "epoch": 646, "lr": 8.944637813424134e-05}
{"train_loss": 0.18209023773670197, "global_step": 57572, "epoch": 646, "lr": 8.944602189144228e-05}
{"train_loss": 0.18670763075351715, "global_step": 57573, "epoch": 646, "lr": 8.94456656433402e-05}
{"train_loss": 0.3306678831577301, "global_step": 57574, "epoch": 646, "lr": 8.944530938993513e-05}
{"train_loss": 0.15006279945373535, "global_step": 57575, "epoch": 646, "lr": 8.944495313122712e-05}
{"train_loss": 0.20107796788215637, "global_step": 57576, "epoch": 646, "lr": 8.944459686721621e-05}
{"train_loss": 0.2793482542037964, "global_step": 57577, "epoch": 646, "lr": 8.944424059790245e-05}
{"train_loss": 0.31613025069236755, "global_step": 57578, "epoch": 646, "lr": 8.94438843232859e-05}
{"train_loss": 0.23081868886947632, "global_step": 57579, "epoch": 646, "lr": 8.94435280433666e-05}
{"train_loss": 0.3334257900714874, "global_step": 57580, "epoch": 646, "lr": 8.94431717581446e-05}
{"train_loss": 0.19335059821605682, "global_step": 57581, "epoch": 646, "lr": 8.944281546761995e-05}
{"train_loss": 0.23288489392634187, "global_step": 57582, "epoch": 646, "lr": 8.944245917179269e-05, "val_loss": 3.425954580307007}
{"train_loss": 0.2899889647960663, "global_step": 57583, "epoch": 647, "lr": 8.944210287066286e-05}
{"train_loss": 0.3338359594345093, "global_step": 57584, "epoch": 647, "lr": 8.944174656423053e-05}
{"train_loss": 0.2692132294178009, "global_step": 57585, "epoch": 647, "lr": 8.944139025249574e-05}
{"train_loss": 0.2766725718975067, "global_step": 57586, "epoch": 647, "lr": 8.944103393545853e-05}
{"train_loss": 0.18830938637256622, "global_step": 57587, "epoch": 647, "lr": 8.944067761311896e-05}
{"train_loss": 0.23342356085777283, "global_step": 57588, "epoch": 647, "lr": 8.944032128547708e-05}
{"train_loss": 0.3089689612388611, "global_step": 57589, "epoch": 647, "lr": 8.943996495253291e-05}
{"train_loss": 0.2928878366947174, "global_step": 57590, "epoch": 647, "lr": 8.943960861428653e-05}
{"train_loss": 0.25916004180908203, "global_step": 57591, "epoch": 647, "lr": 8.943925227073797e-05}
{"train_loss": 0.2592487931251526, "global_step": 57592, "epoch": 647, "lr": 8.943889592188728e-05}
{"train_loss": 0.265777051448822, "global_step": 57593, "epoch": 647, "lr": 8.943853956773452e-05}
{"train_loss": 0.2868814170360565, "global_step": 57594, "epoch": 647, "lr": 8.943818320827973e-05}
{"train_loss": 0.35379043221473694, "global_step": 57595, "epoch": 647, "lr": 8.943782684352295e-05}
{"train_loss": 0.25710344314575195, "global_step": 57596, "epoch": 647, "lr": 8.943747047346424e-05}
{"train_loss": 0.2623014748096466, "global_step": 57597, "epoch": 647, "lr": 8.943711409810364e-05}
{"train_loss": 0.2232016623020172, "global_step": 57598, "epoch": 647, "lr": 8.94367577174412e-05}
{"train_loss": 0.19042053818702698, "global_step": 57599, "epoch": 647, "lr": 8.943640133147696e-05}
{"train_loss": 0.27773141860961914, "global_step": 57600, "epoch": 647, "lr": 8.943604494021099e-05}
{"train_loss": 0.20417365431785583, "global_step": 57601, "epoch": 647, "lr": 8.943568854364333e-05}
{"train_loss": 0.20888186991214752, "global_step": 57602, "epoch": 647, "lr": 8.943533214177399e-05}
{"train_loss": 0.2742389440536499, "global_step": 57603, "epoch": 647, "lr": 8.943497573460308e-05}
{"train_loss": 0.2641914486885071, "global_step": 57604, "epoch": 647, "lr": 8.943461932213062e-05}
{"train_loss": 0.29152971506118774, "global_step": 57605, "epoch": 647, "lr": 8.943426290435664e-05}
{"train_loss": 0.313641220331192, "global_step": 57606, "epoch": 647, "lr": 8.943390648128121e-05}
{"train_loss": 0.3154875636100769, "global_step": 57607, "epoch": 647, "lr": 8.943355005290437e-05}
{"train_loss": 0.2644123435020447, "global_step": 57608, "epoch": 647, "lr": 8.943319361922618e-05}
{"train_loss": 0.22063341736793518, "global_step": 57609, "epoch": 647, "lr": 8.943283718024667e-05}
{"train_loss": 0.27022647857666016, "global_step": 57610, "epoch": 647, "lr": 8.94324807359659e-05}
{"train_loss": 0.293597012758255, "global_step": 57611, "epoch": 647, "lr": 8.94321242863839e-05}
{"train_loss": 0.2451980859041214, "global_step": 57612, "epoch": 647, "lr": 8.943176783150075e-05}
{"train_loss": 0.29069945216178894, "global_step": 57613, "epoch": 647, "lr": 8.943141137131646e-05}
{"train_loss": 0.31077131628990173, "global_step": 57614, "epoch": 647, "lr": 8.943105490583111e-05}
{"train_loss": 0.28366729617118835, "global_step": 57615, "epoch": 647, "lr": 8.943069843504475e-05}
{"train_loss": 0.24074791371822357, "global_step": 57616, "epoch": 647, "lr": 8.943034195895738e-05}
{"train_loss": 0.29022979736328125, "global_step": 57617, "epoch": 647, "lr": 8.942998547756911e-05}
{"train_loss": 0.2135254144668579, "global_step": 57618, "epoch": 647, "lr": 8.942962899087994e-05}
{"train_loss": 0.2611570358276367, "global_step": 57619, "epoch": 647, "lr": 8.942927249888995e-05}
{"train_loss": 0.20672376453876495, "global_step": 57620, "epoch": 647, "lr": 8.942891600159915e-05}
{"train_loss": 0.22791387140750885, "global_step": 57621, "epoch": 647, "lr": 8.942855949900764e-05}
{"train_loss": 0.2425379604101181, "global_step": 57622, "epoch": 647, "lr": 8.942820299111542e-05}
{"train_loss": 0.3180675208568573, "global_step": 57623, "epoch": 647, "lr": 8.942784647792258e-05}
{"train_loss": 0.2593592405319214, "global_step": 57624, "epoch": 647, "lr": 8.942748995942913e-05}
{"train_loss": 0.31668537855148315, "global_step": 57625, "epoch": 647, "lr": 8.942713343563516e-05}
{"train_loss": 0.31004688143730164, "global_step": 57626, "epoch": 647, "lr": 8.942677690654067e-05}
{"train_loss": 0.2952277362346649, "global_step": 57627, "epoch": 647, "lr": 8.942642037214574e-05}
{"train_loss": 0.22806598246097565, "global_step": 57628, "epoch": 647, "lr": 8.94260638324504e-05}
{"train_loss": 0.2121037244796753, "global_step": 57629, "epoch": 647, "lr": 8.94257072874547e-05}
{"train_loss": 0.2596431374549866, "global_step": 57630, "epoch": 647, "lr": 8.942535073715871e-05}
{"train_loss": 0.30559831857681274, "global_step": 57631, "epoch": 647, "lr": 8.942499418156245e-05}
{"train_loss": 0.3250066339969635, "global_step": 57632, "epoch": 647, "lr": 8.942463762066599e-05}
{"train_loss": 0.23959165811538696, "global_step": 57633, "epoch": 647, "lr": 8.942428105446936e-05}
{"train_loss": 0.38887855410575867, "global_step": 57634, "epoch": 647, "lr": 8.942392448297263e-05}
{"train_loss": 0.3896297812461853, "global_step": 57635, "epoch": 647, "lr": 8.942356790617581e-05}
{"train_loss": 0.2201712280511856, "global_step": 57636, "epoch": 647, "lr": 8.942321132407898e-05}
{"train_loss": 0.29666033387184143, "global_step": 57637, "epoch": 647, "lr": 8.942285473668219e-05}
{"train_loss": 0.3701251745223999, "global_step": 57638, "epoch": 647, "lr": 8.942249814398546e-05}
{"train_loss": 0.184361070394516, "global_step": 57639, "epoch": 647, "lr": 8.942214154598889e-05}
{"train_loss": 0.142250195145607, "global_step": 57640, "epoch": 647, "lr": 8.942178494269247e-05}
{"train_loss": 0.23984874784946442, "global_step": 57641, "epoch": 647, "lr": 8.942142833409627e-05}
{"train_loss": 0.2246495932340622, "global_step": 57642, "epoch": 647, "lr": 8.942107172020033e-05}
{"train_loss": 0.27810636162757874, "global_step": 57643, "epoch": 647, "lr": 8.942071510100473e-05}
{"train_loss": 0.2770586609840393, "global_step": 57644, "epoch": 647, "lr": 8.94203584765095e-05}
{"train_loss": 0.21607057750225067, "global_step": 57645, "epoch": 647, "lr": 8.942000184671466e-05}
{"train_loss": 0.19920103251934052, "global_step": 57646, "epoch": 647, "lr": 8.941964521162029e-05}
{"train_loss": 0.22902750968933105, "global_step": 57647, "epoch": 647, "lr": 8.941928857122642e-05}
{"train_loss": 0.2611185908317566, "global_step": 57648, "epoch": 647, "lr": 8.941893192553313e-05}
{"train_loss": 0.29074227809906006, "global_step": 57649, "epoch": 647, "lr": 8.941857527454042e-05}
{"train_loss": 0.2932529151439667, "global_step": 57650, "epoch": 647, "lr": 8.941821861824837e-05}
{"train_loss": 0.3904326856136322, "global_step": 57651, "epoch": 647, "lr": 8.941786195665703e-05}
{"train_loss": 0.21085740625858307, "global_step": 57652, "epoch": 647, "lr": 8.941750528976644e-05}
{"train_loss": 0.38853928446769714, "global_step": 57653, "epoch": 647, "lr": 8.941714861757663e-05}
{"train_loss": 0.2004947066307068, "global_step": 57654, "epoch": 647, "lr": 8.941679194008767e-05}
{"train_loss": 0.340559720993042, "global_step": 57655, "epoch": 647, "lr": 8.941643525729962e-05}
{"train_loss": 0.2505907714366913, "global_step": 57656, "epoch": 647, "lr": 8.94160785692125e-05}
{"train_loss": 0.35224422812461853, "global_step": 57657, "epoch": 647, "lr": 8.941572187582636e-05}
{"train_loss": 0.35517075657844543, "global_step": 57658, "epoch": 647, "lr": 8.941536517714126e-05}
{"train_loss": 0.21883226931095123, "global_step": 57659, "epoch": 647, "lr": 8.941500847315726e-05}
{"train_loss": 0.26398766040802, "global_step": 57660, "epoch": 647, "lr": 8.941465176387438e-05}
{"train_loss": 0.22380635142326355, "global_step": 57661, "epoch": 647, "lr": 8.941429504929269e-05}
{"train_loss": 0.25831449031829834, "global_step": 57662, "epoch": 647, "lr": 8.941393832941221e-05}
{"train_loss": 0.19014477729797363, "global_step": 57663, "epoch": 647, "lr": 8.941358160423301e-05}
{"train_loss": 0.22491464018821716, "global_step": 57664, "epoch": 647, "lr": 8.941322487375515e-05}
{"train_loss": 0.2811744809150696, "global_step": 57665, "epoch": 647, "lr": 8.941286813797865e-05}
{"train_loss": 0.2504011392593384, "global_step": 57666, "epoch": 647, "lr": 8.941251139690357e-05}
{"train_loss": 0.26307761669158936, "global_step": 57667, "epoch": 647, "lr": 8.941215465052996e-05}
{"train_loss": 0.28700876235961914, "global_step": 57668, "epoch": 647, "lr": 8.941179789885788e-05}
{"train_loss": 0.3169407546520233, "global_step": 57669, "epoch": 647, "lr": 8.941144114188736e-05}
{"train_loss": 0.2530485987663269, "global_step": 57670, "epoch": 647, "lr": 8.941108437961842e-05}
{"train_loss": 0.268375084976132, "global_step": 57671, "epoch": 647, "lr": 8.941072761205118e-05, "val_loss": 3.1834545135498047}
{"train_loss": 0.2737790644168854, "global_step": 57672, "epoch": 648, "lr": 8.941037083918563e-05}
{"train_loss": 0.3015417456626892, "global_step": 57673, "epoch": 648, "lr": 8.941001406102182e-05}
{"train_loss": 0.19592516124248505, "global_step": 57674, "epoch": 648, "lr": 8.940965727755985e-05}
{"train_loss": 0.2444113790988922, "global_step": 57675, "epoch": 648, "lr": 8.940930048879972e-05}
{"train_loss": 0.31105536222457886, "global_step": 57676, "epoch": 648, "lr": 8.940894369474147e-05}
{"train_loss": 0.2921927869319916, "global_step": 57677, "epoch": 648, "lr": 8.940858689538518e-05}
{"train_loss": 0.30113720893859863, "global_step": 57678, "epoch": 648, "lr": 8.940823009073089e-05}
{"train_loss": 0.2848132848739624, "global_step": 57679, "epoch": 648, "lr": 8.940787328077864e-05}
{"train_loss": 0.22686313092708588, "global_step": 57680, "epoch": 648, "lr": 8.94075164655285e-05}
{"train_loss": 0.18544833362102509, "global_step": 57681, "epoch": 648, "lr": 8.940715964498048e-05}
{"train_loss": 0.22876021265983582, "global_step": 57682, "epoch": 648, "lr": 8.940680281913465e-05}
{"train_loss": 0.23937110602855682, "global_step": 57683, "epoch": 648, "lr": 8.940644598799105e-05}
{"train_loss": 0.19774311780929565, "global_step": 57684, "epoch": 648, "lr": 8.940608915154973e-05}
{"train_loss": 0.24067576229572296, "global_step": 57685, "epoch": 648, "lr": 8.940573230981075e-05}
{"train_loss": 0.27005136013031006, "global_step": 57686, "epoch": 648, "lr": 8.940537546277417e-05}
{"train_loss": 0.24819575250148773, "global_step": 57687, "epoch": 648, "lr": 8.940501861044e-05}
{"train_loss": 0.28389737010002136, "global_step": 57688, "epoch": 648, "lr": 8.94046617528083e-05}
{"train_loss": 0.20150679349899292, "global_step": 57689, "epoch": 648, "lr": 8.940430488987912e-05}
{"train_loss": 0.24688485264778137, "global_step": 57690, "epoch": 648, "lr": 8.940394802165251e-05}
{"train_loss": 0.2356770634651184, "global_step": 57691, "epoch": 648, "lr": 8.940359114812853e-05}
{"train_loss": 0.19843095541000366, "global_step": 57692, "epoch": 648, "lr": 8.940323426930723e-05}
{"train_loss": 0.2831030488014221, "global_step": 57693, "epoch": 648, "lr": 8.940287738518863e-05}
{"train_loss": 0.34596526622772217, "global_step": 57694, "epoch": 648, "lr": 8.94025204957728e-05}
{"train_loss": 0.2422514408826828, "global_step": 57695, "epoch": 648, "lr": 8.940216360105978e-05}
{"train_loss": 0.24493980407714844, "global_step": 57696, "epoch": 648, "lr": 8.940180670104961e-05}
{"train_loss": 0.2700921893119812, "global_step": 57697, "epoch": 648, "lr": 8.940144979574234e-05}
{"train_loss": 0.19957974553108215, "global_step": 57698, "epoch": 648, "lr": 8.940109288513805e-05}
{"train_loss": 0.24508076906204224, "global_step": 57699, "epoch": 648, "lr": 8.940073596923674e-05}
{"train_loss": 0.303829550743103, "global_step": 57700, "epoch": 648, "lr": 8.94003790480385e-05}
{"train_loss": 0.20319996774196625, "global_step": 57701, "epoch": 648, "lr": 8.940002212154334e-05}
{"train_loss": 0.2120114117860794, "global_step": 57702, "epoch": 648, "lr": 8.939966518975135e-05}
{"train_loss": 0.23048554360866547, "global_step": 57703, "epoch": 648, "lr": 8.939930825266253e-05}
{"train_loss": 0.2788592576980591, "global_step": 57704, "epoch": 648, "lr": 8.939895131027697e-05}
{"train_loss": 0.25976648926734924, "global_step": 57705, "epoch": 648, "lr": 8.93985943625947e-05}
{"train_loss": 0.1932007521390915, "global_step": 57706, "epoch": 648, "lr": 8.939823740961577e-05}
{"train_loss": 0.294800728559494, "global_step": 57707, "epoch": 648, "lr": 8.939788045134024e-05}
{"train_loss": 0.311034232378006, "global_step": 57708, "epoch": 648, "lr": 8.939752348776812e-05}
{"train_loss": 0.1740528792142868, "global_step": 57709, "epoch": 648, "lr": 8.939716651889948e-05}
{"train_loss": 0.273996502161026, "global_step": 57710, "epoch": 648, "lr": 8.93968095447344e-05}
{"train_loss": 0.2641695439815521, "global_step": 57711, "epoch": 648, "lr": 8.939645256527289e-05}
{"train_loss": 0.2288770228624344, "global_step": 57712, "epoch": 648, "lr": 8.939609558051499e-05}
{"train_loss": 0.28038644790649414, "global_step": 57713, "epoch": 648, "lr": 8.939573859046077e-05}
{"train_loss": 0.20039840042591095, "global_step": 57714, "epoch": 648, "lr": 8.93953815951103e-05}
{"train_loss": 0.34297090768814087, "global_step": 57715, "epoch": 648, "lr": 8.939502459446357e-05}
{"train_loss": 0.17599256336688995, "global_step": 57716, "epoch": 648, "lr": 8.939466758852067e-05}
{"train_loss": 0.2597578763961792, "global_step": 57717, "epoch": 648, "lr": 8.939431057728163e-05}
{"train_loss": 0.27312934398651123, "global_step": 57718, "epoch": 648, "lr": 8.939395356074651e-05}
{"train_loss": 0.25749531388282776, "global_step": 57719, "epoch": 648, "lr": 8.939359653891535e-05}
{"train_loss": 0.23998869955539703, "global_step": 57720, "epoch": 648, "lr": 8.939323951178821e-05}
{"train_loss": 0.26131877303123474, "global_step": 57721, "epoch": 648, "lr": 8.939288247936514e-05}
{"train_loss": 0.3160777986049652, "global_step": 57722, "epoch": 648, "lr": 8.939252544164615e-05}
{"train_loss": 0.1745937615633011, "global_step": 57723, "epoch": 648, "lr": 8.939216839863134e-05}
{"train_loss": 0.19758442044258118, "global_step": 57724, "epoch": 648, "lr": 8.939181135032071e-05}
{"train_loss": 0.218299999833107, "global_step": 57725, "epoch": 648, "lr": 8.939145429671434e-05}
{"train_loss": 0.20104925334453583, "global_step": 57726, "epoch": 648, "lr": 8.939109723781227e-05}
{"train_loss": 0.18898530304431915, "global_step": 57727, "epoch": 648, "lr": 8.939074017361455e-05}
{"train_loss": 0.2993989884853363, "global_step": 57728, "epoch": 648, "lr": 8.939038310412122e-05}
{"train_loss": 0.2677749693393707, "global_step": 57729, "epoch": 648, "lr": 8.939002602933233e-05}
{"train_loss": 0.4528827965259552, "global_step": 57730, "epoch": 648, "lr": 8.938966894924794e-05}
{"train_loss": 0.19750453531742096, "global_step": 57731, "epoch": 648, "lr": 8.93893118638681e-05}
{"train_loss": 0.21142469346523285, "global_step": 57732, "epoch": 648, "lr": 8.938895477319282e-05}
{"train_loss": 0.2991969585418701, "global_step": 57733, "epoch": 648, "lr": 8.938859767722219e-05}
{"train_loss": 0.30733588337898254, "global_step": 57734, "epoch": 648, "lr": 8.938824057595626e-05}
{"train_loss": 0.2111191749572754, "global_step": 57735, "epoch": 648, "lr": 8.938788346939503e-05}
{"train_loss": 0.2263580560684204, "global_step": 57736, "epoch": 648, "lr": 8.938752635753859e-05}
{"train_loss": 0.23670777678489685, "global_step": 57737, "epoch": 648, "lr": 8.938716924038697e-05}
{"train_loss": 0.21765798330307007, "global_step": 57738, "epoch": 648, "lr": 8.938681211794024e-05}
{"train_loss": 0.20676150918006897, "global_step": 57739, "epoch": 648, "lr": 8.938645499019842e-05}
{"train_loss": 0.2638346552848816, "global_step": 57740, "epoch": 648, "lr": 8.938609785716158e-05}
{"train_loss": 0.26775214076042175, "global_step": 57741, "epoch": 648, "lr": 8.938574071882976e-05}
{"train_loss": 0.3107069730758667, "global_step": 57742, "epoch": 648, "lr": 8.938538357520301e-05}
{"train_loss": 0.17805469036102295, "global_step": 57743, "epoch": 648, "lr": 8.938502642628137e-05}
{"train_loss": 0.18513348698616028, "global_step": 57744, "epoch": 648, "lr": 8.938466927206488e-05}
{"train_loss": 0.2523432970046997, "global_step": 57745, "epoch": 648, "lr": 8.938431211255363e-05}
{"train_loss": 0.1674383282661438, "global_step": 57746, "epoch": 648, "lr": 8.938395494774761e-05}
{"train_loss": 0.23339207470417023, "global_step": 57747, "epoch": 648, "lr": 8.938359777764692e-05}
{"train_loss": 0.2822619378566742, "global_step": 57748, "epoch": 648, "lr": 8.938324060225156e-05}
{"train_loss": 0.27992716431617737, "global_step": 57749, "epoch": 648, "lr": 8.938288342156163e-05}
{"train_loss": 0.2457696497440338, "global_step": 57750, "epoch": 648, "lr": 8.938252623557715e-05}
{"train_loss": 0.2543348968029022, "global_step": 57751, "epoch": 648, "lr": 8.938216904429815e-05}
{"train_loss": 0.28667154908180237, "global_step": 57752, "epoch": 648, "lr": 8.938181184772471e-05}
{"train_loss": 0.2171372026205063, "global_step": 57753, "epoch": 648, "lr": 8.938145464585685e-05}
{"train_loss": 0.255508691072464, "global_step": 57754, "epoch": 648, "lr": 8.938109743869465e-05}
{"train_loss": 0.2605709433555603, "global_step": 57755, "epoch": 648, "lr": 8.938074022623815e-05}
{"train_loss": 0.2116602212190628, "global_step": 57756, "epoch": 648, "lr": 8.938038300848737e-05}
{"train_loss": 0.20355816185474396, "global_step": 57757, "epoch": 648, "lr": 8.938002578544237e-05}
{"train_loss": 0.18526920676231384, "global_step": 57758, "epoch": 648, "lr": 8.937966855710323e-05}
{"train_loss": 0.22801420092582703, "global_step": 57759, "epoch": 648, "lr": 8.937931132346997e-05}
{"train_loss": 0.24733126900169286, "global_step": 57760, "epoch": 648, "lr": 8.937895408454265e-05, "val_loss": 3.4346470832824707}
{"train_loss": 0.2519528269767761, "global_step": 57761, "epoch": 649, "lr": 8.937859684032128e-05}
{"train_loss": 0.21994848549365997, "global_step": 57762, "epoch": 649, "lr": 8.937823959080596e-05}
{"train_loss": 0.18539877235889435, "global_step": 57763, "epoch": 649, "lr": 8.93778823359967e-05}
{"train_loss": 0.16154517233371735, "global_step": 57764, "epoch": 649, "lr": 8.937752507589357e-05}
{"train_loss": 0.26274168491363525, "global_step": 57765, "epoch": 649, "lr": 8.937716781049661e-05}
{"train_loss": 0.3383925259113312, "global_step": 57766, "epoch": 649, "lr": 8.937681053980588e-05}
{"train_loss": 0.39346715807914734, "global_step": 57767, "epoch": 649, "lr": 8.937645326382141e-05}
{"train_loss": 0.2627003490924835, "global_step": 57768, "epoch": 649, "lr": 8.937609598254325e-05}
{"train_loss": 0.18288205564022064, "global_step": 57769, "epoch": 649, "lr": 8.937573869597146e-05}
{"train_loss": 0.3120057284832001, "global_step": 57770, "epoch": 649, "lr": 8.937538140410608e-05}
{"train_loss": 0.17691993713378906, "global_step": 57771, "epoch": 649, "lr": 8.937502410694717e-05}
{"train_loss": 0.25602781772613525, "global_step": 57772, "epoch": 649, "lr": 8.937466680449476e-05}
{"train_loss": 0.3637395203113556, "global_step": 57773, "epoch": 649, "lr": 8.93743094967489e-05}
{"train_loss": 0.2629130184650421, "global_step": 57774, "epoch": 649, "lr": 8.937395218370966e-05}
{"train_loss": 0.3584812581539154, "global_step": 57775, "epoch": 649, "lr": 8.937359486537705e-05}
{"train_loss": 0.32879623770713806, "global_step": 57776, "epoch": 649, "lr": 8.937323754175115e-05}
{"train_loss": 0.335216760635376, "global_step": 57777, "epoch": 649, "lr": 8.937288021283201e-05}
{"train_loss": 0.30105817317962646, "global_step": 57778, "epoch": 649, "lr": 8.937252287861966e-05}
{"train_loss": 0.28534066677093506, "global_step": 57779, "epoch": 649, "lr": 8.937216553911415e-05}
{"train_loss": 0.2597571611404419, "global_step": 57780, "epoch": 649, "lr": 8.937180819431553e-05}
{"train_loss": 0.2522067725658417, "global_step": 57781, "epoch": 649, "lr": 8.937145084422385e-05}
{"train_loss": 0.29076021909713745, "global_step": 57782, "epoch": 649, "lr": 8.937109348883917e-05}
{"train_loss": 0.2562975287437439, "global_step": 57783, "epoch": 649, "lr": 8.937073612816151e-05}
{"train_loss": 0.4037988781929016, "global_step": 57784, "epoch": 649, "lr": 8.937037876219094e-05}
{"train_loss": 0.3843061625957489, "global_step": 57785, "epoch": 649, "lr": 8.93700213909275e-05}
{"train_loss": 0.27109774947166443, "global_step": 57786, "epoch": 649, "lr": 8.936966401437126e-05}
{"train_loss": 0.2049415558576584, "global_step": 57787, "epoch": 649, "lr": 8.936930663252223e-05}
{"train_loss": 0.26818549633026123, "global_step": 57788, "epoch": 649, "lr": 8.936894924538048e-05}
{"train_loss": 0.24053514003753662, "global_step": 57789, "epoch": 649, "lr": 8.936859185294606e-05}
{"train_loss": 0.18458381295204163, "global_step": 57790, "epoch": 649, "lr": 8.936823445521901e-05}
{"train_loss": 0.24800913035869598, "global_step": 57791, "epoch": 649, "lr": 8.936787705219937e-05}
{"train_loss": 0.33347439765930176, "global_step": 57792, "epoch": 649, "lr": 8.936751964388719e-05}
{"train_loss": 0.27306148409843445, "global_step": 57793, "epoch": 649, "lr": 8.936716223028255e-05}
{"train_loss": 0.2905391454696655, "global_step": 57794, "epoch": 649, "lr": 8.936680481138547e-05}
{"train_loss": 0.324289470911026, "global_step": 57795, "epoch": 649, "lr": 8.9366447387196e-05}
{"train_loss": 0.24514947831630707, "global_step": 57796, "epoch": 649, "lr": 8.936608995771421e-05}
{"train_loss": 0.25566092133522034, "global_step": 57797, "epoch": 649, "lr": 8.936573252294009e-05}
{"train_loss": 0.19482001662254333, "global_step": 57798, "epoch": 649, "lr": 8.936537508287376e-05}
{"train_loss": 0.2769443988800049, "global_step": 57799, "epoch": 649, "lr": 8.936501763751522e-05}
{"train_loss": 0.24923664331436157, "global_step": 57800, "epoch": 649, "lr": 8.936466018686453e-05}
{"train_loss": 0.24353638291358948, "global_step": 57801, "epoch": 649, "lr": 8.936430273092176e-05}
{"train_loss": 0.31085270643234253, "global_step": 57802, "epoch": 649, "lr": 8.936394526968692e-05}
{"train_loss": 0.24588224291801453, "global_step": 57803, "epoch": 649, "lr": 8.936358780316009e-05}
{"train_loss": 0.19869279861450195, "global_step": 57804, "epoch": 649, "lr": 8.93632303313413e-05}
{"train_loss": 0.14766119420528412, "global_step": 57805, "epoch": 649, "lr": 8.93628728542306e-05}
{"train_loss": 0.23066264390945435, "global_step": 57806, "epoch": 649, "lr": 8.936251537182805e-05}
{"train_loss": 0.2084411382675171, "global_step": 57807, "epoch": 649, "lr": 8.936215788413368e-05}
{"train_loss": 0.29820704460144043, "global_step": 57808, "epoch": 649, "lr": 8.936180039114755e-05}
{"train_loss": 0.30850329995155334, "global_step": 57809, "epoch": 649, "lr": 8.93614428928697e-05}
{"train_loss": 0.3441086709499359, "global_step": 57810, "epoch": 649, "lr": 8.93610853893002e-05}
{"train_loss": 0.334900438785553, "global_step": 57811, "epoch": 649, "lr": 8.936072788043908e-05}
{"train_loss": 0.21229565143585205, "global_step": 57812, "epoch": 649, "lr": 8.936037036628637e-05}
{"train_loss": 0.30145683884620667, "global_step": 57813, "epoch": 649, "lr": 8.936001284684216e-05}
{"train_loss": 0.18091560900211334, "global_step": 57814, "epoch": 649, "lr": 8.935965532210646e-05}
{"train_loss": 0.18263503909111023, "global_step": 57815, "epoch": 649, "lr": 8.935929779207935e-05}
{"train_loss": 0.21898280084133148, "global_step": 57816, "epoch": 649, "lr": 8.935894025676085e-05}
{"train_loss": 0.20249640941619873, "global_step": 57817, "epoch": 649, "lr": 8.935858271615103e-05}
{"train_loss": 0.18356925249099731, "global_step": 57818, "epoch": 649, "lr": 8.935822517024993e-05}
{"train_loss": 0.23758475482463837, "global_step": 57819, "epoch": 649, "lr": 8.935786761905758e-05}
{"train_loss": 0.39313435554504395, "global_step": 57820, "epoch": 649, "lr": 8.935751006257405e-05}
{"train_loss": 0.21363183856010437, "global_step": 57821, "epoch": 649, "lr": 8.935715250079937e-05}
{"train_loss": 0.2861705720424652, "global_step": 57822, "epoch": 649, "lr": 8.935679493373363e-05}
{"train_loss": 0.37028515338897705, "global_step": 57823, "epoch": 649, "lr": 8.935643736137683e-05}
{"train_loss": 0.17136512696743011, "global_step": 57824, "epoch": 649, "lr": 8.935607978372904e-05}
{"train_loss": 0.2961980998516083, "global_step": 57825, "epoch": 649, "lr": 8.935572220079031e-05}
{"train_loss": 0.20976798236370087, "global_step": 57826, "epoch": 649, "lr": 8.935536461256067e-05}
{"train_loss": 0.2792542278766632, "global_step": 57827, "epoch": 649, "lr": 8.93550070190402e-05}
{"train_loss": 0.23494575917720795, "global_step": 57828, "epoch": 649, "lr": 8.935464942022891e-05}
{"train_loss": 0.19795507192611694, "global_step": 57829, "epoch": 649, "lr": 8.935429181612688e-05}
{"train_loss": 0.18095318973064423, "global_step": 57830, "epoch": 649, "lr": 8.935393420673414e-05}
{"train_loss": 0.22791169583797455, "global_step": 57831, "epoch": 649, "lr": 8.935357659205075e-05}
{"train_loss": 0.2161954939365387, "global_step": 57832, "epoch": 649, "lr": 8.935321897207674e-05}
{"train_loss": 0.24245737493038177, "global_step": 57833, "epoch": 649, "lr": 8.935286134681217e-05}
{"train_loss": 0.24163897335529327, "global_step": 57834, "epoch": 649, "lr": 8.93525037162571e-05}
{"train_loss": 0.2688833773136139, "global_step": 57835, "epoch": 649, "lr": 8.935214608041156e-05}
{"train_loss": 0.1818239688873291, "global_step": 57836, "epoch": 649, "lr": 8.93517884392756e-05}
{"train_loss": 0.224564328789711, "global_step": 57837, "epoch": 649, "lr": 8.935143079284926e-05}
{"train_loss": 0.21341156959533691, "global_step": 57838, "epoch": 649, "lr": 8.935107314113263e-05}
{"train_loss": 0.2727849781513214, "global_step": 57839, "epoch": 649, "lr": 8.93507154841257e-05}
{"train_loss": 0.2659637928009033, "global_step": 57840, "epoch": 649, "lr": 8.935035782182855e-05}
{"train_loss": 0.26962408423423767, "global_step": 57841, "epoch": 649, "lr": 8.935000015424125e-05}
{"train_loss": 0.2511693239212036, "global_step": 57842, "epoch": 649, "lr": 8.93496424813638e-05}
{"train_loss": 0.2113276571035385, "global_step": 57843, "epoch": 649, "lr": 8.934928480319626e-05}
{"train_loss": 0.23375162482261658, "global_step": 57844, "epoch": 649, "lr": 8.93489271197387e-05}
{"train_loss": 0.18808823823928833, "global_step": 57845, "epoch": 649, "lr": 8.934856943099116e-05}
{"train_loss": 0.1632065623998642, "global_step": 57846, "epoch": 649, "lr": 8.934821173695368e-05}
{"train_loss": 0.272479385137558, "global_step": 57847, "epoch": 649, "lr": 8.934785403762631e-05}
{"train_loss": 0.2155895084142685, "global_step": 57848, "epoch": 649, "lr": 8.934749633300911e-05}
{"train_loss": 0.25630664524067653, "global_step": 57849, "epoch": 649, "lr": 8.934713862310212e-05, "val_loss": 3.2581193447113037}
{"train_loss": 0.23052458465099335, "global_step": 57850, "epoch": 650, "lr": 8.934678090790536e-05}
{"train_loss": 0.26453712582588196, "global_step": 57851, "epoch": 650, "lr": 8.934642318741894e-05}
{"train_loss": 0.2579101026058197, "global_step": 57852, "epoch": 650, "lr": 8.934606546164286e-05}
{"train_loss": 0.37496957182884216, "global_step": 57853, "epoch": 650, "lr": 8.934570773057718e-05}
{"train_loss": 0.22098997235298157, "global_step": 57854, "epoch": 650, "lr": 8.934534999422194e-05}
{"train_loss": 0.23654723167419434, "global_step": 57855, "epoch": 650, "lr": 8.934499225257721e-05}
{"train_loss": 0.2240360677242279, "global_step": 57856, "epoch": 650, "lr": 8.9344634505643e-05}
{"train_loss": 0.18384018540382385, "global_step": 57857, "epoch": 650, "lr": 8.934427675341942e-05}
{"train_loss": 0.10251239687204361, "global_step": 57858, "epoch": 650, "lr": 8.934391899590646e-05}
{"train_loss": 0.2523600459098816, "global_step": 57859, "epoch": 650, "lr": 8.93435612331042e-05}
{"train_loss": 0.18862709403038025, "global_step": 57860, "epoch": 650, "lr": 8.934320346501268e-05}
{"train_loss": 0.1533341109752655, "global_step": 57861, "epoch": 650, "lr": 8.934284569163194e-05}
{"train_loss": 0.2742253839969635, "global_step": 57862, "epoch": 650, "lr": 8.934248791296203e-05}
{"train_loss": 0.2425406575202942, "global_step": 57863, "epoch": 650, "lr": 8.9342130129003e-05}
{"train_loss": 0.273086279630661, "global_step": 57864, "epoch": 650, "lr": 8.93417723397549e-05}
{"train_loss": 0.36716213822364807, "global_step": 57865, "epoch": 650, "lr": 8.934141454521778e-05}
{"train_loss": 0.24619796872138977, "global_step": 57866, "epoch": 650, "lr": 8.934105674539171e-05}
{"train_loss": 0.24092385172843933, "global_step": 57867, "epoch": 650, "lr": 8.934069894027668e-05}
{"train_loss": 0.26110556721687317, "global_step": 57868, "epoch": 650, "lr": 8.93403411298728e-05}
{"train_loss": 0.27680066227912903, "global_step": 57869, "epoch": 650, "lr": 8.933998331418006e-05}
{"train_loss": 0.22580881416797638, "global_step": 57870, "epoch": 650, "lr": 8.933962549319857e-05}
{"train_loss": 0.25453171133995056, "global_step": 57871, "epoch": 650, "lr": 8.933926766692833e-05}
{"train_loss": 0.2055000513792038, "global_step": 57872, "epoch": 650, "lr": 8.93389098353694e-05}
{"train_loss": 0.23372571170330048, "global_step": 57873, "epoch": 650, "lr": 8.933855199852183e-05}
{"train_loss": 0.2425120621919632, "global_step": 57874, "epoch": 650, "lr": 8.933819415638569e-05}
{"train_loss": 0.24559429287910461, "global_step": 57875, "epoch": 650, "lr": 8.9337836308961e-05}
{"train_loss": 0.2552788555622101, "global_step": 57876, "epoch": 650, "lr": 8.933747845624782e-05}
{"train_loss": 0.3695490062236786, "global_step": 57877, "epoch": 650, "lr": 8.93371205982462e-05}
{"train_loss": 0.21844224631786346, "global_step": 57878, "epoch": 650, "lr": 8.933676273495617e-05}
{"train_loss": 0.17188282310962677, "global_step": 57879, "epoch": 650, "lr": 8.93364048663778e-05}
{"train_loss": 0.19187401235103607, "global_step": 57880, "epoch": 650, "lr": 8.933604699251112e-05}
{"train_loss": 0.2789366841316223, "global_step": 57881, "epoch": 650, "lr": 8.933568911335622e-05}
{"train_loss": 0.22083204984664917, "global_step": 57882, "epoch": 650, "lr": 8.933533122891308e-05}
{"train_loss": 0.23974674940109253, "global_step": 57883, "epoch": 650, "lr": 8.933497333918179e-05}
{"train_loss": 0.2682803273200989, "global_step": 57884, "epoch": 650, "lr": 8.933461544416241e-05}
{"train_loss": 0.31480714678764343, "global_step": 57885, "epoch": 650, "lr": 8.933425754385496e-05}
{"train_loss": 0.21516363322734833, "global_step": 57886, "epoch": 650, "lr": 8.933389963825948e-05}
{"train_loss": 0.21131455898284912, "global_step": 57887, "epoch": 650, "lr": 8.933354172737607e-05}
{"train_loss": 0.31925058364868164, "global_step": 57888, "epoch": 650, "lr": 8.933318381120472e-05}
{"train_loss": 0.35920366644859314, "global_step": 57889, "epoch": 650, "lr": 8.933282588974551e-05}
{"train_loss": 0.18364080786705017, "global_step": 57890, "epoch": 650, "lr": 8.93324679629985e-05}
{"train_loss": 0.18532419204711914, "global_step": 57891, "epoch": 650, "lr": 8.933211003096369e-05}
{"train_loss": 0.24519981443881989, "global_step": 57892, "epoch": 650, "lr": 8.933175209364118e-05}
{"train_loss": 0.28341564536094666, "global_step": 57893, "epoch": 650, "lr": 8.933139415103098e-05}
{"train_loss": 0.29716768860816956, "global_step": 57894, "epoch": 650, "lr": 8.933103620313317e-05}
{"train_loss": 0.2171211540699005, "global_step": 57895, "epoch": 650, "lr": 8.933067824994777e-05}
{"train_loss": 0.1784776747226715, "global_step": 57896, "epoch": 650, "lr": 8.933032029147483e-05}
{"train_loss": 0.12369203567504883, "global_step": 57897, "epoch": 650, "lr": 8.932996232771442e-05}
{"train_loss": 0.2832711637020111, "global_step": 57898, "epoch": 650, "lr": 8.932960435866657e-05}
{"train_loss": 0.30466869473457336, "global_step": 57899, "epoch": 650, "lr": 8.932924638433134e-05}
{"train_loss": 0.2867514193058014, "global_step": 57900, "epoch": 650, "lr": 8.932888840470876e-05}
{"train_loss": 0.14905577898025513, "global_step": 57901, "epoch": 650, "lr": 8.93285304197989e-05}
{"train_loss": 0.22446878254413605, "global_step": 57902, "epoch": 650, "lr": 8.93281724296018e-05}
{"train_loss": 0.2500261068344116, "global_step": 57903, "epoch": 650, "lr": 8.932781443411751e-05}
{"train_loss": 0.31246528029441833, "global_step": 57904, "epoch": 650, "lr": 8.932745643334607e-05}
{"train_loss": 0.21797360479831696, "global_step": 57905, "epoch": 650, "lr": 8.932709842728753e-05}
{"train_loss": 0.2540682852268219, "global_step": 57906, "epoch": 650, "lr": 8.932674041594194e-05}
{"train_loss": 0.2615003287792206, "global_step": 57907, "epoch": 650, "lr": 8.932638239930935e-05}
{"train_loss": 0.31940335035324097, "global_step": 57908, "epoch": 650, "lr": 8.932602437738981e-05}
{"train_loss": 0.31612107157707214, "global_step": 57909, "epoch": 650, "lr": 8.932566635018337e-05}
{"train_loss": 0.1681133508682251, "global_step": 57910, "epoch": 650, "lr": 8.932530831769005e-05}
{"train_loss": 0.2454526126384735, "global_step": 57911, "epoch": 650, "lr": 8.932495027990995e-05}
{"train_loss": 0.28160545229911804, "global_step": 57912, "epoch": 650, "lr": 8.932459223684309e-05}
{"train_loss": 0.25155875086784363, "global_step": 57913, "epoch": 650, "lr": 8.932423418848951e-05}
{"train_loss": 0.17143267393112183, "global_step": 57914, "epoch": 650, "lr": 8.932387613484927e-05}
{"train_loss": 0.2612556517124176, "global_step": 57915, "epoch": 650, "lr": 8.93235180759224e-05}
{"train_loss": 0.3250337541103363, "global_step": 57916, "epoch": 650, "lr": 8.932316001170897e-05}
{"train_loss": 0.1474696844816208, "global_step": 57917, "epoch": 650, "lr": 8.9322801942209e-05}
{"train_loss": 0.23397207260131836, "global_step": 57918, "epoch": 650, "lr": 8.932244386742259e-05}
{"train_loss": 0.21645373106002808, "global_step": 57919, "epoch": 650, "lr": 8.932208578734975e-05}
{"train_loss": 0.20226967334747314, "global_step": 57920, "epoch": 650, "lr": 8.932172770199054e-05}
{"train_loss": 0.30781182646751404, "global_step": 57921, "epoch": 650, "lr": 8.932136961134498e-05}
{"train_loss": 0.2957253158092499, "global_step": 57922, "epoch": 650, "lr": 8.932101151541317e-05}
{"train_loss": 0.27654388546943665, "global_step": 57923, "epoch": 650, "lr": 8.932065341419511e-05}
{"train_loss": 0.3029341995716095, "global_step": 57924, "epoch": 650, "lr": 8.932029530769086e-05}
{"train_loss": 0.3138217031955719, "global_step": 57925, "epoch": 650, "lr": 8.931993719590049e-05}
{"train_loss": 0.27838823199272156, "global_step": 57926, "epoch": 650, "lr": 8.931957907882404e-05}
{"train_loss": 0.17571525275707245, "global_step": 57927, "epoch": 650, "lr": 8.931922095646153e-05}
{"train_loss": 0.206157386302948, "global_step": 57928, "epoch": 650, "lr": 8.931886282881305e-05}
{"train_loss": 0.2808789908885956, "global_step": 57929, "epoch": 650, "lr": 8.931850469587861e-05}
{"train_loss": 0.2756926417350769, "global_step": 57930, "epoch": 650, "lr": 8.931814655765829e-05}
{"train_loss": 0.22213317453861237, "global_step": 57931, "epoch": 650, "lr": 8.931778841415213e-05}
{"train_loss": 0.34005409479141235, "global_step": 57932, "epoch": 650, "lr": 8.931743026536016e-05}
{"train_loss": 0.2198917418718338, "global_step": 57933, "epoch": 650, "lr": 8.931707211128244e-05}
{"train_loss": 0.31510135531425476, "global_step": 57934, "epoch": 650, "lr": 8.931671395191902e-05}
{"train_loss": 0.19693689048290253, "global_step": 57935, "epoch": 650, "lr": 8.931635578726996e-05}
{"train_loss": 0.29582634568214417, "global_step": 57936, "epoch": 650, "lr": 8.931599761733529e-05}
{"train_loss": 0.19538788497447968, "global_step": 57937, "epoch": 650, "lr": 8.931563944211506e-05}
{"train_loss": 0.24805673061127073, "global_step": 57938, "epoch": 650, "lr": 8.931528126160931e-05, "train/sim_max_reward_0": 0.6980197495333379, "train/sim_max_reward_1": 0.9227459788864889, "train/sim_max_reward_2": 0.6830220810353329, "train/sim_max_reward_3": 0.3514758793374822, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.7061645345943482, "test/sim_max_reward_4500000": 0.9872129647839053, "test/sim_max_reward_4500001": 0.9945899559715736, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.41853880599260146, "test/sim_max_reward_4500004": 0.6213912916217965, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.4028582246566376, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.414844239494819, "test/sim_max_reward_4500009": 0.0488633410394764, "test/sim_max_reward_4500010": 0.9513618525509712, "test/sim_max_reward_4500011": 0.94705048377395, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.48328296437442664, "test/sim_max_reward_4500014": 0.9284692471741192, "test/sim_max_reward_4500015": 0.9952504247823981, "test/sim_max_reward_4500016": 0.8702028635429316, "test/sim_max_reward_4500017": 0.5033808064602743, "test/sim_max_reward_4500018": 0.15723023654503204, "test/sim_max_reward_4500019": 0.9960973199781205, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9292565108741906, "test/sim_max_reward_4500022": 0.9556689050495504, "test/sim_max_reward_4500023": 0.530041231951769, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.5052996526981992, "test/sim_max_reward_4500026": 0.5415097976856967, "test/sim_max_reward_4500027": 0.9740472743348931, "test/sim_max_reward_4500028": 0.7216100873170933, "test/sim_max_reward_4500029": 0.264073162552818, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.4510562072475881, "test/sim_max_reward_4500032": 0.578523371183502, "test/sim_max_reward_4500033": 0.971081460431279, "test/sim_max_reward_4500034": 0.47069616709328743, "test/sim_max_reward_4500035": 1.0, "test/sim_max_reward_4500036": 0.9988559510781101, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.6521255202520901, "test/sim_max_reward_4500039": 0.9568318936429951, "test/sim_max_reward_4500040": 0.8776137864966453, "test/sim_max_reward_4500041": 0.3874259725260081, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.6834974389185091, "test/sim_max_reward_4500044": 0.20017711225407434, "test/sim_max_reward_4500045": 0.46334699835909976, "test/sim_max_reward_4500046": 0.3916528879489556, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.985268970496368, "test/sim_max_reward_4500049": 0.957422221046356, "train/mean_score": 0.7269047038978317, "test/mean_score": 0.6233541520836422, "val_loss": 3.4032931327819824, "train_action_mse_error": 36.7643928527832}
{"train_loss": 0.2586532533168793, "global_step": 57939, "epoch": 651, "lr": 8.93149230758181e-05}
{"train_loss": 0.16407883167266846, "global_step": 57940, "epoch": 651, "lr": 8.93145648847415e-05}
{"train_loss": 0.20748738944530487, "global_step": 57941, "epoch": 651, "lr": 8.931420668837952e-05}
{"train_loss": 0.24719619750976562, "global_step": 57942, "epoch": 651, "lr": 8.931384848673223e-05}
{"train_loss": 0.2739333212375641, "global_step": 57943, "epoch": 651, "lr": 8.931349027979967e-05}
{"train_loss": 0.2398298680782318, "global_step": 57944, "epoch": 651, "lr": 8.931313206758189e-05}
{"train_loss": 0.234927698969841, "global_step": 57945, "epoch": 651, "lr": 8.931277385007895e-05}
{"train_loss": 0.28468266129493713, "global_step": 57946, "epoch": 651, "lr": 8.931241562729087e-05}
{"train_loss": 0.26172035932540894, "global_step": 57947, "epoch": 651, "lr": 8.93120573992177e-05}
{"train_loss": 0.19041180610656738, "global_step": 57948, "epoch": 651, "lr": 8.931169916585953e-05}
{"train_loss": 0.2657814025878906, "global_step": 57949, "epoch": 651, "lr": 8.931134092721638e-05}
{"train_loss": 0.28012651205062866, "global_step": 57950, "epoch": 651, "lr": 8.931098268328831e-05}
{"train_loss": 0.2845500111579895, "global_step": 57951, "epoch": 651, "lr": 8.931062443407533e-05}
{"train_loss": 0.20079919695854187, "global_step": 57952, "epoch": 651, "lr": 8.931026617957752e-05}
{"train_loss": 0.2359941303730011, "global_step": 57953, "epoch": 651, "lr": 8.930990791979495e-05}
{"train_loss": 0.2367880791425705, "global_step": 57954, "epoch": 651, "lr": 8.930954965472761e-05}
{"train_loss": 0.21964997053146362, "global_step": 57955, "epoch": 651, "lr": 8.930919138437559e-05}
{"train_loss": 0.1833595186471939, "global_step": 57956, "epoch": 651, "lr": 8.930883310873893e-05}
{"train_loss": 0.29356083273887634, "global_step": 57957, "epoch": 651, "lr": 8.930847482781768e-05}
{"train_loss": 0.23033668100833893, "global_step": 57958, "epoch": 651, "lr": 8.930811654161188e-05}
{"train_loss": 0.20779837667942047, "global_step": 57959, "epoch": 651, "lr": 8.930775825012159e-05}
{"train_loss": 0.26025259494781494, "global_step": 57960, "epoch": 651, "lr": 8.930739995334685e-05}
{"train_loss": 0.28438764810562134, "global_step": 57961, "epoch": 651, "lr": 8.93070416512877e-05}
{"train_loss": 0.26746267080307007, "global_step": 57962, "epoch": 651, "lr": 8.93066833439442e-05}
{"train_loss": 0.18938618898391724, "global_step": 57963, "epoch": 651, "lr": 8.93063250313164e-05}
{"train_loss": 0.26783138513565063, "global_step": 57964, "epoch": 651, "lr": 8.930596671340434e-05}
{"train_loss": 0.32688266038894653, "global_step": 57965, "epoch": 651, "lr": 8.930560839020807e-05}
{"train_loss": 0.23596805334091187, "global_step": 57966, "epoch": 651, "lr": 8.930525006172766e-05}
{"train_loss": 0.21780595183372498, "global_step": 57967, "epoch": 651, "lr": 8.930489172796311e-05}
{"train_loss": 0.18022485077381134, "global_step": 57968, "epoch": 651, "lr": 8.930453338891451e-05}
{"train_loss": 0.13859714567661285, "global_step": 57969, "epoch": 651, "lr": 8.93041750445819e-05}
{"train_loss": 0.19671133160591125, "global_step": 57970, "epoch": 651, "lr": 8.930381669496531e-05}
{"train_loss": 0.26681092381477356, "global_step": 57971, "epoch": 651, "lr": 8.930345834006482e-05}
{"train_loss": 0.16425341367721558, "global_step": 57972, "epoch": 651, "lr": 8.930309997988045e-05}
{"train_loss": 0.22446511685848236, "global_step": 57973, "epoch": 651, "lr": 8.930274161441223e-05}
{"train_loss": 0.26311826705932617, "global_step": 57974, "epoch": 651, "lr": 8.930238324366027e-05}
{"train_loss": 0.15749047696590424, "global_step": 57975, "epoch": 651, "lr": 8.930202486762457e-05}
{"train_loss": 0.21735234558582306, "global_step": 57976, "epoch": 651, "lr": 8.93016664863052e-05}
{"train_loss": 0.17645229399204254, "global_step": 57977, "epoch": 651, "lr": 8.930130809970218e-05}
{"train_loss": 0.2997000217437744, "global_step": 57978, "epoch": 651, "lr": 8.930094970781559e-05}
{"train_loss": 0.21889182925224304, "global_step": 57979, "epoch": 651, "lr": 8.930059131064547e-05}
{"train_loss": 0.19358642399311066, "global_step": 57980, "epoch": 651, "lr": 8.930023290819187e-05}
{"train_loss": 0.2579001188278198, "global_step": 57981, "epoch": 651, "lr": 8.929987450045483e-05}
{"train_loss": 0.20145384967327118, "global_step": 57982, "epoch": 651, "lr": 8.929951608743438e-05}
{"train_loss": 0.29874682426452637, "global_step": 57983, "epoch": 651, "lr": 8.929915766913061e-05}
{"train_loss": 0.22110483050346375, "global_step": 57984, "epoch": 651, "lr": 8.929879924554354e-05}
{"train_loss": 0.25241515040397644, "global_step": 57985, "epoch": 651, "lr": 8.929844081667323e-05}
{"train_loss": 0.28631412982940674, "global_step": 57986, "epoch": 651, "lr": 8.929808238251974e-05}
{"train_loss": 0.24869033694267273, "global_step": 57987, "epoch": 651, "lr": 8.929772394308306e-05}
{"train_loss": 0.1825076937675476, "global_step": 57988, "epoch": 651, "lr": 8.929736549836331e-05}
{"train_loss": 0.26092156767845154, "global_step": 57989, "epoch": 651, "lr": 8.929700704836051e-05}
{"train_loss": 0.22386434674263, "global_step": 57990, "epoch": 651, "lr": 8.92966485930747e-05}
{"train_loss": 0.2333003580570221, "global_step": 57991, "epoch": 651, "lr": 8.929629013250593e-05}
{"train_loss": 0.23849493265151978, "global_step": 57992, "epoch": 651, "lr": 8.929593166665426e-05}
{"train_loss": 0.1603718250989914, "global_step": 57993, "epoch": 651, "lr": 8.929557319551973e-05}
{"train_loss": 0.2316443920135498, "global_step": 57994, "epoch": 651, "lr": 8.929521471910238e-05}
{"train_loss": 0.22522474825382233, "global_step": 57995, "epoch": 651, "lr": 8.929485623740227e-05}
{"train_loss": 0.2383255958557129, "global_step": 57996, "epoch": 651, "lr": 8.929449775041944e-05}
{"train_loss": 0.2964679002761841, "global_step": 57997, "epoch": 651, "lr": 8.929413925815395e-05}
{"train_loss": 0.2426663488149643, "global_step": 57998, "epoch": 651, "lr": 8.929378076060585e-05}
{"train_loss": 0.33920934796333313, "global_step": 57999, "epoch": 651, "lr": 8.929342225777517e-05}
{"train_loss": 0.2130054086446762, "global_step": 58000, "epoch": 651, "lr": 8.929306374966198e-05}
{"train_loss": 0.2464541345834732, "global_step": 58001, "epoch": 651, "lr": 8.92927052362663e-05}
{"train_loss": 0.2526899576187134, "global_step": 58002, "epoch": 651, "lr": 8.929234671758822e-05}
{"train_loss": 0.30103254318237305, "global_step": 58003, "epoch": 651, "lr": 8.929198819362774e-05}
{"train_loss": 0.25312742590904236, "global_step": 58004, "epoch": 651, "lr": 8.929162966438493e-05}
{"train_loss": 0.19807028770446777, "global_step": 58005, "epoch": 651, "lr": 8.929127112985984e-05}
{"train_loss": 0.2374248057603836, "global_step": 58006, "epoch": 651, "lr": 8.929091259005254e-05}
{"train_loss": 0.16467487812042236, "global_step": 58007, "epoch": 651, "lr": 8.929055404496303e-05}
{"train_loss": 0.340953528881073, "global_step": 58008, "epoch": 651, "lr": 8.92901954945914e-05}
{"train_loss": 0.24443265795707703, "global_step": 58009, "epoch": 651, "lr": 8.928983693893766e-05}
{"train_loss": 0.2595871388912201, "global_step": 58010, "epoch": 651, "lr": 8.928947837800192e-05}
{"train_loss": 0.20038101077079773, "global_step": 58011, "epoch": 651, "lr": 8.928911981178416e-05}
{"train_loss": 0.2838192284107208, "global_step": 58012, "epoch": 651, "lr": 8.928876124028446e-05}
{"train_loss": 0.36171939969062805, "global_step": 58013, "epoch": 651, "lr": 8.928840266350287e-05}
{"train_loss": 0.28430140018463135, "global_step": 58014, "epoch": 651, "lr": 8.928804408143943e-05}
{"train_loss": 0.31127679347991943, "global_step": 58015, "epoch": 651, "lr": 8.928768549409419e-05}
{"train_loss": 0.2781839966773987, "global_step": 58016, "epoch": 651, "lr": 8.92873269014672e-05}
{"train_loss": 0.18286320567131042, "global_step": 58017, "epoch": 651, "lr": 8.92869683035585e-05}
{"train_loss": 0.24662791192531586, "global_step": 58018, "epoch": 651, "lr": 8.928660970036815e-05}
{"train_loss": 0.20411650836467743, "global_step": 58019, "epoch": 651, "lr": 8.92862510918962e-05}
{"train_loss": 0.3132588565349579, "global_step": 58020, "epoch": 651, "lr": 8.92858924781427e-05}
{"train_loss": 0.27137455344200134, "global_step": 58021, "epoch": 651, "lr": 8.928553385910767e-05}
{"train_loss": 0.3182072639465332, "global_step": 58022, "epoch": 651, "lr": 8.928517523479119e-05}
{"train_loss": 0.21614162623882294, "global_step": 58023, "epoch": 651, "lr": 8.92848166051933e-05}
{"train_loss": 0.32606860995292664, "global_step": 58024, "epoch": 651, "lr": 8.928445797031404e-05}
{"train_loss": 0.35955649614334106, "global_step": 58025, "epoch": 651, "lr": 8.928409933015347e-05}
{"train_loss": 0.2317768931388855, "global_step": 58026, "epoch": 651, "lr": 8.928374068471162e-05}
{"train_loss": 0.2451215371322096, "global_step": 58027, "epoch": 651, "lr": 8.928338203398856e-05, "val_loss": 3.305894613265991}
{"train_loss": 0.1771746426820755, "global_step": 58028, "epoch": 652, "lr": 8.928302337798433e-05}
{"train_loss": 0.26124846935272217, "global_step": 58029, "epoch": 652, "lr": 8.928266471669898e-05}
{"train_loss": 0.1926669478416443, "global_step": 58030, "epoch": 652, "lr": 8.928230605013253e-05}
{"train_loss": 0.2394421100616455, "global_step": 58031, "epoch": 652, "lr": 8.928194737828508e-05}
{"train_loss": 0.31407302618026733, "global_step": 58032, "epoch": 652, "lr": 8.928158870115663e-05}
{"train_loss": 0.2685624957084656, "global_step": 58033, "epoch": 652, "lr": 8.928123001874727e-05}
{"train_loss": 0.35059574246406555, "global_step": 58034, "epoch": 652, "lr": 8.9280871331057e-05}
{"train_loss": 0.22099995613098145, "global_step": 58035, "epoch": 652, "lr": 8.928051263808592e-05}
{"train_loss": 0.22612200677394867, "global_step": 58036, "epoch": 652, "lr": 8.928015393983404e-05}
{"train_loss": 0.2677266299724579, "global_step": 58037, "epoch": 652, "lr": 8.927979523630143e-05}
{"train_loss": 0.18240322172641754, "global_step": 58038, "epoch": 652, "lr": 8.927943652748813e-05}
{"train_loss": 0.28977063298225403, "global_step": 58039, "epoch": 652, "lr": 8.92790778133942e-05}
{"train_loss": 0.3172391951084137, "global_step": 58040, "epoch": 652, "lr": 8.927871909401966e-05}
{"train_loss": 0.24701736867427826, "global_step": 58041, "epoch": 652, "lr": 8.927836036936457e-05}
{"train_loss": 0.2025456428527832, "global_step": 58042, "epoch": 652, "lr": 8.927800163942898e-05}
{"train_loss": 0.3018173277378082, "global_step": 58043, "epoch": 652, "lr": 8.927764290421297e-05}
{"train_loss": 0.2605676054954529, "global_step": 58044, "epoch": 652, "lr": 8.927728416371653e-05}
{"train_loss": 0.2007855772972107, "global_step": 58045, "epoch": 652, "lr": 8.927692541793975e-05}
{"train_loss": 0.28139710426330566, "global_step": 58046, "epoch": 652, "lr": 8.927656666688267e-05}
{"train_loss": 0.2424461990594864, "global_step": 58047, "epoch": 652, "lr": 8.927620791054534e-05}
{"train_loss": 0.26662567257881165, "global_step": 58048, "epoch": 652, "lr": 8.92758491489278e-05}
{"train_loss": 0.21808356046676636, "global_step": 58049, "epoch": 652, "lr": 8.927549038203008e-05}
{"train_loss": 0.33148229122161865, "global_step": 58050, "epoch": 652, "lr": 8.927513160985226e-05}
{"train_loss": 0.31082430481910706, "global_step": 58051, "epoch": 652, "lr": 8.927477283239438e-05}
{"train_loss": 0.29654768109321594, "global_step": 58052, "epoch": 652, "lr": 8.927441404965649e-05}
{"train_loss": 0.3714643120765686, "global_step": 58053, "epoch": 652, "lr": 8.927405526163863e-05}
{"train_loss": 0.2999880015850067, "global_step": 58054, "epoch": 652, "lr": 8.927369646834086e-05}
{"train_loss": 0.17320553958415985, "global_step": 58055, "epoch": 652, "lr": 8.92733376697632e-05}
{"train_loss": 0.2646902799606323, "global_step": 58056, "epoch": 652, "lr": 8.927297886590573e-05}
{"train_loss": 0.23001936078071594, "global_step": 58057, "epoch": 652, "lr": 8.927262005676848e-05}
{"train_loss": 0.2024753987789154, "global_step": 58058, "epoch": 652, "lr": 8.927226124235153e-05}
{"train_loss": 0.26334911584854126, "global_step": 58059, "epoch": 652, "lr": 8.927190242265488e-05}
{"train_loss": 0.2269468903541565, "global_step": 58060, "epoch": 652, "lr": 8.92715435976786e-05}
{"train_loss": 0.33961474895477295, "global_step": 58061, "epoch": 652, "lr": 8.927118476742275e-05}
{"train_loss": 0.24550355970859528, "global_step": 58062, "epoch": 652, "lr": 8.927082593188735e-05}
{"train_loss": 0.16165131330490112, "global_step": 58063, "epoch": 652, "lr": 8.927046709107249e-05}
{"train_loss": 0.22618156671524048, "global_step": 58064, "epoch": 652, "lr": 8.927010824497818e-05}
{"train_loss": 0.21629416942596436, "global_step": 58065, "epoch": 652, "lr": 8.926974939360448e-05}
{"train_loss": 0.23058220744132996, "global_step": 58066, "epoch": 652, "lr": 8.926939053695145e-05}
{"train_loss": 0.26868945360183716, "global_step": 58067, "epoch": 652, "lr": 8.926903167501915e-05}
{"train_loss": 0.2846674621105194, "global_step": 58068, "epoch": 652, "lr": 8.926867280780757e-05}
{"train_loss": 0.21570463478565216, "global_step": 58069, "epoch": 652, "lr": 8.92683139353168e-05}
{"train_loss": 0.2539437711238861, "global_step": 58070, "epoch": 652, "lr": 8.926795505754691e-05}
{"train_loss": 0.24306397140026093, "global_step": 58071, "epoch": 652, "lr": 8.92675961744979e-05}
{"train_loss": 0.246486097574234, "global_step": 58072, "epoch": 652, "lr": 8.926723728616984e-05}
{"train_loss": 0.2015438675880432, "global_step": 58073, "epoch": 652, "lr": 8.926687839256278e-05}
{"train_loss": 0.32863909006118774, "global_step": 58074, "epoch": 652, "lr": 8.926651949367678e-05}
{"train_loss": 0.33769485354423523, "global_step": 58075, "epoch": 652, "lr": 8.926616058951187e-05}
{"train_loss": 0.21616677939891815, "global_step": 58076, "epoch": 652, "lr": 8.92658016800681e-05}
{"train_loss": 0.3168582022190094, "global_step": 58077, "epoch": 652, "lr": 8.926544276534553e-05}
{"train_loss": 0.3621050715446472, "global_step": 58078, "epoch": 652, "lr": 8.926508384534418e-05}
{"train_loss": 0.31949302554130554, "global_step": 58079, "epoch": 652, "lr": 8.926472492006413e-05}
{"train_loss": 0.1893693208694458, "global_step": 58080, "epoch": 652, "lr": 8.92643659895054e-05}
{"train_loss": 0.17857615649700165, "global_step": 58081, "epoch": 652, "lr": 8.926400705366808e-05}
{"train_loss": 0.21401043236255646, "global_step": 58082, "epoch": 652, "lr": 8.926364811255218e-05}
{"train_loss": 0.27855184674263, "global_step": 58083, "epoch": 652, "lr": 8.926328916615776e-05}
{"train_loss": 0.30496808886528015, "global_step": 58084, "epoch": 652, "lr": 8.926293021448488e-05}
{"train_loss": 0.21293845772743225, "global_step": 58085, "epoch": 652, "lr": 8.926257125753357e-05}
{"train_loss": 0.24451591074466705, "global_step": 58086, "epoch": 652, "lr": 8.92622122953039e-05}
{"train_loss": 0.2069481760263443, "global_step": 58087, "epoch": 652, "lr": 8.926185332779587e-05}
{"train_loss": 0.21714282035827637, "global_step": 58088, "epoch": 652, "lr": 8.926149435500959e-05}
{"train_loss": 0.28944650292396545, "global_step": 58089, "epoch": 652, "lr": 8.926113537694507e-05}
{"train_loss": 0.20556427538394928, "global_step": 58090, "epoch": 652, "lr": 8.926077639360237e-05}
{"train_loss": 0.26405826210975647, "global_step": 58091, "epoch": 652, "lr": 8.926041740498155e-05}
{"train_loss": 0.23910318315029144, "global_step": 58092, "epoch": 652, "lr": 8.926005841108262e-05}
{"train_loss": 0.22286857664585114, "global_step": 58093, "epoch": 652, "lr": 8.925969941190568e-05}
{"train_loss": 0.2392895370721817, "global_step": 58094, "epoch": 652, "lr": 8.925934040745074e-05}
{"train_loss": 0.24330516159534454, "global_step": 58095, "epoch": 652, "lr": 8.925898139771786e-05}
{"train_loss": 0.19991029798984528, "global_step": 58096, "epoch": 652, "lr": 8.925862238270708e-05}
{"train_loss": 0.26691147685050964, "global_step": 58097, "epoch": 652, "lr": 8.925826336241847e-05}
{"train_loss": 0.2974461317062378, "global_step": 58098, "epoch": 652, "lr": 8.925790433685206e-05}
{"train_loss": 0.33221524953842163, "global_step": 58099, "epoch": 652, "lr": 8.925754530600791e-05}
{"train_loss": 0.33810099959373474, "global_step": 58100, "epoch": 652, "lr": 8.925718626988603e-05}
{"train_loss": 0.3478500247001648, "global_step": 58101, "epoch": 652, "lr": 8.925682722848654e-05}
{"train_loss": 0.2937283515930176, "global_step": 58102, "epoch": 652, "lr": 8.925646818180944e-05}
{"train_loss": 0.288734495639801, "global_step": 58103, "epoch": 652, "lr": 8.925610912985478e-05}
{"train_loss": 0.19506631791591644, "global_step": 58104, "epoch": 652, "lr": 8.925575007262261e-05}
{"train_loss": 0.2504723370075226, "global_step": 58105, "epoch": 652, "lr": 8.925539101011299e-05}
{"train_loss": 0.2835341989994049, "global_step": 58106, "epoch": 652, "lr": 8.925503194232597e-05}
{"train_loss": 0.24615556001663208, "global_step": 58107, "epoch": 652, "lr": 8.925467286926157e-05}
{"train_loss": 0.28937631845474243, "global_step": 58108, "epoch": 652, "lr": 8.925431379091987e-05}
{"train_loss": 0.24616335332393646, "global_step": 58109, "epoch": 652, "lr": 8.92539547073009e-05}
{"train_loss": 0.2752262055873871, "global_step": 58110, "epoch": 652, "lr": 8.925359561840469e-05}
{"train_loss": 0.18703307211399078, "global_step": 58111, "epoch": 652, "lr": 8.925323652423135e-05}
{"train_loss": 0.24326443672180176, "global_step": 58112, "epoch": 652, "lr": 8.925287742478088e-05}
{"train_loss": 0.2666091322898865, "global_step": 58113, "epoch": 652, "lr": 8.925251832005335e-05}
{"train_loss": 0.22442369163036346, "global_step": 58114, "epoch": 652, "lr": 8.925215921004877e-05}
{"train_loss": 0.14412330090999603, "global_step": 58115, "epoch": 652, "lr": 8.925180009476723e-05}
{"train_loss": 0.2553581227077527, "global_step": 58116, "epoch": 652, "lr": 8.925144097420877e-05, "val_loss": 3.301074504852295}
{"train_loss": 0.25008291006088257, "global_step": 58117, "epoch": 653, "lr": 8.925108184837343e-05}
{"train_loss": 0.1830226480960846, "global_step": 58118, "epoch": 653, "lr": 8.925072271726125e-05}
{"train_loss": 0.25908032059669495, "global_step": 58119, "epoch": 653, "lr": 8.925036358087229e-05}
{"train_loss": 0.24290421605110168, "global_step": 58120, "epoch": 653, "lr": 8.925000443920661e-05}
{"train_loss": 0.29328200221061707, "global_step": 58121, "epoch": 653, "lr": 8.924964529226424e-05}
{"train_loss": 0.27644890546798706, "global_step": 58122, "epoch": 653, "lr": 8.924928614004524e-05}
{"train_loss": 0.24043993651866913, "global_step": 58123, "epoch": 653, "lr": 8.924892698254964e-05}
{"train_loss": 0.3263714909553528, "global_step": 58124, "epoch": 653, "lr": 8.924856781977751e-05}
{"train_loss": 0.29076266288757324, "global_step": 58125, "epoch": 653, "lr": 8.924820865172887e-05}
{"train_loss": 0.23522889614105225, "global_step": 58126, "epoch": 653, "lr": 8.92478494784038e-05}
{"train_loss": 0.22110691666603088, "global_step": 58127, "epoch": 653, "lr": 8.924749029980233e-05}
{"train_loss": 0.3429088890552521, "global_step": 58128, "epoch": 653, "lr": 8.924713111592452e-05}
{"train_loss": 0.31047317385673523, "global_step": 58129, "epoch": 653, "lr": 8.92467719267704e-05}
{"train_loss": 0.23604214191436768, "global_step": 58130, "epoch": 653, "lr": 8.924641273234005e-05}
{"train_loss": 0.32977092266082764, "global_step": 58131, "epoch": 653, "lr": 8.924605353263349e-05}
{"train_loss": 0.2687474489212036, "global_step": 58132, "epoch": 653, "lr": 8.924569432765077e-05}
{"train_loss": 0.29955214262008667, "global_step": 58133, "epoch": 653, "lr": 8.924533511739194e-05}
{"train_loss": 0.17530062794685364, "global_step": 58134, "epoch": 653, "lr": 8.924497590185708e-05}
{"train_loss": 0.17132532596588135, "global_step": 58135, "epoch": 653, "lr": 8.924461668104618e-05}
{"train_loss": 0.33174052834510803, "global_step": 58136, "epoch": 653, "lr": 8.924425745495932e-05}
{"train_loss": 0.25246885418891907, "global_step": 58137, "epoch": 653, "lr": 8.924389822359657e-05}
{"train_loss": 0.2215178906917572, "global_step": 58138, "epoch": 653, "lr": 8.924353898695794e-05}
{"train_loss": 0.1733388751745224, "global_step": 58139, "epoch": 653, "lr": 8.92431797450435e-05}
{"train_loss": 0.3387833833694458, "global_step": 58140, "epoch": 653, "lr": 8.92428204978533e-05}
{"train_loss": 0.3019297122955322, "global_step": 58141, "epoch": 653, "lr": 8.924246124538738e-05}
{"train_loss": 0.20639468729496002, "global_step": 58142, "epoch": 653, "lr": 8.924210198764577e-05}
{"train_loss": 0.2603473663330078, "global_step": 58143, "epoch": 653, "lr": 8.924174272462857e-05}
{"train_loss": 0.23092254996299744, "global_step": 58144, "epoch": 653, "lr": 8.924138345633577e-05}
{"train_loss": 0.2764486074447632, "global_step": 58145, "epoch": 653, "lr": 8.924102418276745e-05}
{"train_loss": 0.2945060431957245, "global_step": 58146, "epoch": 653, "lr": 8.924066490392366e-05}
{"train_loss": 0.268062561750412, "global_step": 58147, "epoch": 653, "lr": 8.924030561980443e-05}
{"train_loss": 0.24542711675167084, "global_step": 58148, "epoch": 653, "lr": 8.923994633040982e-05}
{"train_loss": 0.20763921737670898, "global_step": 58149, "epoch": 653, "lr": 8.923958703573989e-05}
{"train_loss": 0.2497866451740265, "global_step": 58150, "epoch": 653, "lr": 8.923922773579466e-05}
{"train_loss": 0.182551309466362, "global_step": 58151, "epoch": 653, "lr": 8.923886843057421e-05}
{"train_loss": 0.2865462005138397, "global_step": 58152, "epoch": 653, "lr": 8.923850912007856e-05}
{"train_loss": 0.2504464089870453, "global_step": 58153, "epoch": 653, "lr": 8.923814980430778e-05}
{"train_loss": 0.22927699983119965, "global_step": 58154, "epoch": 653, "lr": 8.92377904832619e-05}
{"train_loss": 0.2706061601638794, "global_step": 58155, "epoch": 653, "lr": 8.923743115694097e-05}
{"train_loss": 0.2610419690608978, "global_step": 58156, "epoch": 653, "lr": 8.923707182534507e-05}
{"train_loss": 0.24358223378658295, "global_step": 58157, "epoch": 653, "lr": 8.923671248847421e-05}
{"train_loss": 0.22216397523880005, "global_step": 58158, "epoch": 653, "lr": 8.923635314632846e-05}
{"train_loss": 0.20846545696258545, "global_step": 58159, "epoch": 653, "lr": 8.923599379890784e-05}
{"train_loss": 0.195218026638031, "global_step": 58160, "epoch": 653, "lr": 8.923563444621244e-05}
{"train_loss": 0.23593860864639282, "global_step": 58161, "epoch": 653, "lr": 8.923527508824227e-05}
{"train_loss": 0.21215510368347168, "global_step": 58162, "epoch": 653, "lr": 8.923491572499741e-05}
{"train_loss": 0.21580837666988373, "global_step": 58163, "epoch": 653, "lr": 8.923455635647789e-05}
{"train_loss": 0.24391677975654602, "global_step": 58164, "epoch": 653, "lr": 8.923419698268377e-05}
{"train_loss": 0.2631595730781555, "global_step": 58165, "epoch": 653, "lr": 8.923383760361508e-05}
{"train_loss": 0.23474590480327606, "global_step": 58166, "epoch": 653, "lr": 8.923347821927187e-05}
{"train_loss": 0.2408466935157776, "global_step": 58167, "epoch": 653, "lr": 8.923311882965421e-05}
{"train_loss": 0.24850581586360931, "global_step": 58168, "epoch": 653, "lr": 8.923275943476212e-05}
{"train_loss": 0.36465543508529663, "global_step": 58169, "epoch": 653, "lr": 8.923240003459569e-05}
{"train_loss": 0.28088611364364624, "global_step": 58170, "epoch": 653, "lr": 8.923204062915492e-05}
{"train_loss": 0.24160878360271454, "global_step": 58171, "epoch": 653, "lr": 8.923168121843989e-05}
{"train_loss": 0.28814178705215454, "global_step": 58172, "epoch": 653, "lr": 8.923132180245064e-05}
{"train_loss": 0.25503042340278625, "global_step": 58173, "epoch": 653, "lr": 8.92309623811872e-05}
{"train_loss": 0.2176331877708435, "global_step": 58174, "epoch": 653, "lr": 8.923060295464967e-05}
{"train_loss": 0.2359226495027542, "global_step": 58175, "epoch": 653, "lr": 8.923024352283804e-05}
{"train_loss": 0.2352534532546997, "global_step": 58176, "epoch": 653, "lr": 8.922988408575238e-05}
{"train_loss": 0.22634029388427734, "global_step": 58177, "epoch": 653, "lr": 8.922952464339273e-05}
{"train_loss": 0.27046772837638855, "global_step": 58178, "epoch": 653, "lr": 8.922916519575918e-05}
{"train_loss": 0.2204696238040924, "global_step": 58179, "epoch": 653, "lr": 8.922880574285173e-05}
{"train_loss": 0.2601785659790039, "global_step": 58180, "epoch": 653, "lr": 8.922844628467044e-05}
{"train_loss": 0.183877632021904, "global_step": 58181, "epoch": 653, "lr": 8.922808682121537e-05}
{"train_loss": 0.27953559160232544, "global_step": 58182, "epoch": 653, "lr": 8.922772735248657e-05}
{"train_loss": 0.28707534074783325, "global_step": 58183, "epoch": 653, "lr": 8.922736787848407e-05}
{"train_loss": 0.1986227184534073, "global_step": 58184, "epoch": 653, "lr": 8.922700839920793e-05}
{"train_loss": 0.2276899516582489, "global_step": 58185, "epoch": 653, "lr": 8.922664891465819e-05}
{"train_loss": 0.20304611325263977, "global_step": 58186, "epoch": 653, "lr": 8.92262894248349e-05}
{"train_loss": 0.24777851998806, "global_step": 58187, "epoch": 653, "lr": 8.922592992973814e-05}
{"train_loss": 0.20811545848846436, "global_step": 58188, "epoch": 653, "lr": 8.922557042936793e-05}
{"train_loss": 0.25374799966812134, "global_step": 58189, "epoch": 653, "lr": 8.922521092372429e-05}
{"train_loss": 0.22218561172485352, "global_step": 58190, "epoch": 653, "lr": 8.922485141280732e-05}
{"train_loss": 0.19088341295719147, "global_step": 58191, "epoch": 653, "lr": 8.922449189661704e-05}
{"train_loss": 0.17969703674316406, "global_step": 58192, "epoch": 653, "lr": 8.922413237515351e-05}
{"train_loss": 0.19749127328395844, "global_step": 58193, "epoch": 653, "lr": 8.922377284841677e-05}
{"train_loss": 0.21951062977313995, "global_step": 58194, "epoch": 653, "lr": 8.922341331640687e-05}
{"train_loss": 0.3699023127555847, "global_step": 58195, "epoch": 653, "lr": 8.922305377912386e-05}
{"train_loss": 0.27286002039909363, "global_step": 58196, "epoch": 653, "lr": 8.922269423656779e-05}
{"train_loss": 0.25353559851646423, "global_step": 58197, "epoch": 653, "lr": 8.922233468873871e-05}
{"train_loss": 0.2781500220298767, "global_step": 58198, "epoch": 653, "lr": 8.922197513563666e-05}
{"train_loss": 0.17947474122047424, "global_step": 58199, "epoch": 653, "lr": 8.92216155772617e-05}
{"train_loss": 0.3161678910255432, "global_step": 58200, "epoch": 653, "lr": 8.922125601361388e-05}
{"train_loss": 0.14309632778167725, "global_step": 58201, "epoch": 653, "lr": 8.922089644469322e-05}
{"train_loss": 0.22811073064804077, "global_step": 58202, "epoch": 653, "lr": 8.92205368704998e-05}
{"train_loss": 0.23395328223705292, "global_step": 58203, "epoch": 653, "lr": 8.922017729103365e-05}
{"train_loss": 0.21212254464626312, "global_step": 58204, "epoch": 653, "lr": 8.921981770629483e-05}
{"train_loss": 0.24734091557813495, "global_step": 58205, "epoch": 653, "lr": 8.921945811628338e-05, "val_loss": 3.4049439430236816}
{"train_loss": 0.14719513058662415, "global_step": 58206, "epoch": 654, "lr": 8.921909852099937e-05}
{"train_loss": 0.24178610742092133, "global_step": 58207, "epoch": 654, "lr": 8.921873892044282e-05}
{"train_loss": 0.33037111163139343, "global_step": 58208, "epoch": 654, "lr": 8.921837931461376e-05}
{"train_loss": 0.3065091073513031, "global_step": 58209, "epoch": 654, "lr": 8.92180197035123e-05}
{"train_loss": 0.1899920403957367, "global_step": 58210, "epoch": 654, "lr": 8.921766008713844e-05}
{"train_loss": 0.26109588146209717, "global_step": 58211, "epoch": 654, "lr": 8.921730046549226e-05}
{"train_loss": 0.2985362410545349, "global_step": 58212, "epoch": 654, "lr": 8.921694083857376e-05}
{"train_loss": 0.25088295340538025, "global_step": 58213, "epoch": 654, "lr": 8.921658120638305e-05}
{"train_loss": 0.24413056671619415, "global_step": 58214, "epoch": 654, "lr": 8.921622156892016e-05}
{"train_loss": 0.20870116353034973, "global_step": 58215, "epoch": 654, "lr": 8.92158619261851e-05}
{"train_loss": 0.2536381185054779, "global_step": 58216, "epoch": 654, "lr": 8.921550227817794e-05}
{"train_loss": 0.17100615799427032, "global_step": 58217, "epoch": 654, "lr": 8.921514262489877e-05}
{"train_loss": 0.1741439551115036, "global_step": 58218, "epoch": 654, "lr": 8.921478296634756e-05}
{"train_loss": 0.2872476577758789, "global_step": 58219, "epoch": 654, "lr": 8.921442330252442e-05}
{"train_loss": 0.1918993443250656, "global_step": 58220, "epoch": 654, "lr": 8.921406363342938e-05}
{"train_loss": 0.2604468762874603, "global_step": 58221, "epoch": 654, "lr": 8.921370395906248e-05}
{"train_loss": 0.18848872184753418, "global_step": 58222, "epoch": 654, "lr": 8.921334427942378e-05}
{"train_loss": 0.21544775366783142, "global_step": 58223, "epoch": 654, "lr": 8.921298459451333e-05}
{"train_loss": 0.30819085240364075, "global_step": 58224, "epoch": 654, "lr": 8.921262490433116e-05}
{"train_loss": 0.14907975494861603, "global_step": 58225, "epoch": 654, "lr": 8.921226520887733e-05}
{"train_loss": 0.3238556683063507, "global_step": 58226, "epoch": 654, "lr": 8.921190550815191e-05}
{"train_loss": 0.26734158396720886, "global_step": 58227, "epoch": 654, "lr": 8.92115458021549e-05}
{"train_loss": 0.3048632740974426, "global_step": 58228, "epoch": 654, "lr": 8.921118609088639e-05}
{"train_loss": 0.20689108967781067, "global_step": 58229, "epoch": 654, "lr": 8.921082637434642e-05}
{"train_loss": 0.23278677463531494, "global_step": 58230, "epoch": 654, "lr": 8.921046665253504e-05}
{"train_loss": 0.30410999059677124, "global_step": 58231, "epoch": 654, "lr": 8.921010692545227e-05}
{"train_loss": 0.22666609287261963, "global_step": 58232, "epoch": 654, "lr": 8.920974719309817e-05}
{"train_loss": 0.29866185784339905, "global_step": 58233, "epoch": 654, "lr": 8.920938745547281e-05}
{"train_loss": 0.17447751760482788, "global_step": 58234, "epoch": 654, "lr": 8.920902771257623e-05}
{"train_loss": 0.2217782735824585, "global_step": 58235, "epoch": 654, "lr": 8.920866796440848e-05}
{"train_loss": 0.21093575656414032, "global_step": 58236, "epoch": 654, "lr": 8.92083082109696e-05}
{"train_loss": 0.2632924020290375, "global_step": 58237, "epoch": 654, "lr": 8.920794845225962e-05}
{"train_loss": 0.21321645379066467, "global_step": 58238, "epoch": 654, "lr": 8.920758868827863e-05}
{"train_loss": 0.2511354386806488, "global_step": 58239, "epoch": 654, "lr": 8.920722891902664e-05}
{"train_loss": 0.2761710286140442, "global_step": 58240, "epoch": 654, "lr": 8.920686914450374e-05}
{"train_loss": 0.2281733900308609, "global_step": 58241, "epoch": 654, "lr": 8.920650936470993e-05}
{"train_loss": 0.2695554494857788, "global_step": 58242, "epoch": 654, "lr": 8.92061495796453e-05}
{"train_loss": 0.17165696620941162, "global_step": 58243, "epoch": 654, "lr": 8.920578978930986e-05}
{"train_loss": 0.22637054324150085, "global_step": 58244, "epoch": 654, "lr": 8.920542999370371e-05}
{"train_loss": 0.27540796995162964, "global_step": 58245, "epoch": 654, "lr": 8.920507019282685e-05}
{"train_loss": 0.1920253038406372, "global_step": 58246, "epoch": 654, "lr": 8.920471038667934e-05}
{"train_loss": 0.20047126710414886, "global_step": 58247, "epoch": 654, "lr": 8.920435057526125e-05}
{"train_loss": 0.17493334412574768, "global_step": 58248, "epoch": 654, "lr": 8.920399075857262e-05}
{"train_loss": 0.2647237777709961, "global_step": 58249, "epoch": 654, "lr": 8.920363093661347e-05}
{"train_loss": 0.21773533523082733, "global_step": 58250, "epoch": 654, "lr": 8.920327110938386e-05}
{"train_loss": 0.3442653715610504, "global_step": 58251, "epoch": 654, "lr": 8.920291127688389e-05}
{"train_loss": 0.1661209762096405, "global_step": 58252, "epoch": 654, "lr": 8.920255143911353e-05}
{"train_loss": 0.3260385990142822, "global_step": 58253, "epoch": 654, "lr": 8.920219159607289e-05}
{"train_loss": 0.19539333879947662, "global_step": 58254, "epoch": 654, "lr": 8.920183174776196e-05}
{"train_loss": 0.18220984935760498, "global_step": 58255, "epoch": 654, "lr": 8.920147189418084e-05}
{"train_loss": 0.2079075425863266, "global_step": 58256, "epoch": 654, "lr": 8.920111203532957e-05}
{"train_loss": 0.1971275806427002, "global_step": 58257, "epoch": 654, "lr": 8.920075217120818e-05}
{"train_loss": 0.23182366788387299, "global_step": 58258, "epoch": 654, "lr": 8.920039230181672e-05}
{"train_loss": 0.24124771356582642, "global_step": 58259, "epoch": 654, "lr": 8.920003242715525e-05}
{"train_loss": 0.2303938865661621, "global_step": 58260, "epoch": 654, "lr": 8.919967254722383e-05}
{"train_loss": 0.31558772921562195, "global_step": 58261, "epoch": 654, "lr": 8.919931266202246e-05}
{"train_loss": 0.2534192204475403, "global_step": 58262, "epoch": 654, "lr": 8.919895277155124e-05}
{"train_loss": 0.3202064633369446, "global_step": 58263, "epoch": 654, "lr": 8.91985928758102e-05}
{"train_loss": 0.24118268489837646, "global_step": 58264, "epoch": 654, "lr": 8.919823297479937e-05}
{"train_loss": 0.15742667019367218, "global_step": 58265, "epoch": 654, "lr": 8.919787306851883e-05}
{"train_loss": 0.315033495426178, "global_step": 58266, "epoch": 654, "lr": 8.919751315696861e-05}
{"train_loss": 0.28663161396980286, "global_step": 58267, "epoch": 654, "lr": 8.919715324014877e-05}
{"train_loss": 0.21536166965961456, "global_step": 58268, "epoch": 654, "lr": 8.919679331805935e-05}
{"train_loss": 0.314569890499115, "global_step": 58269, "epoch": 654, "lr": 8.919643339070038e-05}
{"train_loss": 0.3375593423843384, "global_step": 58270, "epoch": 654, "lr": 8.919607345807195e-05}
{"train_loss": 0.22759118676185608, "global_step": 58271, "epoch": 654, "lr": 8.919571352017407e-05}
{"train_loss": 0.21025961637496948, "global_step": 58272, "epoch": 654, "lr": 8.919535357700681e-05}
{"train_loss": 0.3019762635231018, "global_step": 58273, "epoch": 654, "lr": 8.919499362857021e-05}
{"train_loss": 0.21878020465373993, "global_step": 58274, "epoch": 654, "lr": 8.919463367486433e-05}
{"train_loss": 0.3480099141597748, "global_step": 58275, "epoch": 654, "lr": 8.91942737158892e-05}
{"train_loss": 0.24325363337993622, "global_step": 58276, "epoch": 654, "lr": 8.919391375164488e-05}
{"train_loss": 0.2912158966064453, "global_step": 58277, "epoch": 654, "lr": 8.919355378213141e-05}
{"train_loss": 0.20035840570926666, "global_step": 58278, "epoch": 654, "lr": 8.919319380734886e-05}
{"train_loss": 0.24116308987140656, "global_step": 58279, "epoch": 654, "lr": 8.919283382729724e-05}
{"train_loss": 0.27263474464416504, "global_step": 58280, "epoch": 654, "lr": 8.919247384197664e-05}
{"train_loss": 0.268215537071228, "global_step": 58281, "epoch": 654, "lr": 8.919211385138708e-05}
{"train_loss": 0.2677423357963562, "global_step": 58282, "epoch": 654, "lr": 8.919175385552863e-05}
{"train_loss": 0.32599908113479614, "global_step": 58283, "epoch": 654, "lr": 8.91913938544013e-05}
{"train_loss": 0.20735636353492737, "global_step": 58284, "epoch": 654, "lr": 8.919103384800519e-05}
{"train_loss": 0.2603440582752228, "global_step": 58285, "epoch": 654, "lr": 8.919067383634033e-05}
{"train_loss": 0.2723594307899475, "global_step": 58286, "epoch": 654, "lr": 8.919031381940673e-05}
{"train_loss": 0.1900927573442459, "global_step": 58287, "epoch": 654, "lr": 8.91899537972045e-05}
{"train_loss": 0.25913211703300476, "global_step": 58288, "epoch": 654, "lr": 8.918959376973363e-05}
{"train_loss": 0.2550327479839325, "global_step": 58289, "epoch": 654, "lr": 8.918923373699421e-05}
{"train_loss": 0.27126961946487427, "global_step": 58290, "epoch": 654, "lr": 8.918887369898629e-05}
{"train_loss": 0.2502307593822479, "global_step": 58291, "epoch": 654, "lr": 8.918851365570989e-05}
{"train_loss": 0.22567029297351837, "global_step": 58292, "epoch": 654, "lr": 8.918815360716507e-05}
{"train_loss": 0.2399194836616516, "global_step": 58293, "epoch": 654, "lr": 8.918779355335188e-05}
{"train_loss": 0.245838928256142, "global_step": 58294, "epoch": 654, "lr": 8.918743349427037e-05, "val_loss": 3.4194748401641846}
{"train_loss": 0.26019230484962463, "global_step": 58295, "epoch": 655, "lr": 8.91870734299206e-05}
{"train_loss": 0.2385948747396469, "global_step": 58296, "epoch": 655, "lr": 8.91867133603026e-05}
{"train_loss": 0.2133864164352417, "global_step": 58297, "epoch": 655, "lr": 8.91863532854164e-05}
{"train_loss": 0.24978162348270416, "global_step": 58298, "epoch": 655, "lr": 8.918599320526211e-05}
{"train_loss": 0.22535471618175507, "global_step": 58299, "epoch": 655, "lr": 8.918563311983971e-05}
{"train_loss": 0.1828187108039856, "global_step": 58300, "epoch": 655, "lr": 8.918527302914929e-05}
{"train_loss": 0.2258933186531067, "global_step": 58301, "epoch": 655, "lr": 8.91849129331909e-05}
{"train_loss": 0.24972587823867798, "global_step": 58302, "epoch": 655, "lr": 8.918455283196457e-05}
{"train_loss": 0.27150100469589233, "global_step": 58303, "epoch": 655, "lr": 8.918419272547034e-05}
{"train_loss": 0.1894414722919464, "global_step": 58304, "epoch": 655, "lr": 8.918383261370829e-05}
{"train_loss": 0.2495930790901184, "global_step": 58305, "epoch": 655, "lr": 8.918347249667845e-05}
{"train_loss": 0.3445112407207489, "global_step": 58306, "epoch": 655, "lr": 8.918311237438085e-05}
{"train_loss": 0.24343857169151306, "global_step": 58307, "epoch": 655, "lr": 8.918275224681557e-05}
{"train_loss": 0.1994994729757309, "global_step": 58308, "epoch": 655, "lr": 8.918239211398266e-05}
{"train_loss": 0.1695953756570816, "global_step": 58309, "epoch": 655, "lr": 8.918203197588213e-05}
{"train_loss": 0.1789545714855194, "global_step": 58310, "epoch": 655, "lr": 8.918167183251407e-05}
{"train_loss": 0.17523080110549927, "global_step": 58311, "epoch": 655, "lr": 8.918131168387851e-05}
{"train_loss": 0.23521117866039276, "global_step": 58312, "epoch": 655, "lr": 8.91809515299755e-05}
{"train_loss": 0.22212757170200348, "global_step": 58313, "epoch": 655, "lr": 8.918059137080508e-05}
{"train_loss": 0.2815549671649933, "global_step": 58314, "epoch": 655, "lr": 8.918023120636732e-05}
{"train_loss": 0.24851752817630768, "global_step": 58315, "epoch": 655, "lr": 8.917987103666225e-05}
{"train_loss": 0.23338952660560608, "global_step": 58316, "epoch": 655, "lr": 8.917951086168991e-05}
{"train_loss": 0.22772514820098877, "global_step": 58317, "epoch": 655, "lr": 8.917915068145038e-05}
{"train_loss": 0.2836524546146393, "global_step": 58318, "epoch": 655, "lr": 8.917879049594369e-05}
{"train_loss": 0.17963536083698273, "global_step": 58319, "epoch": 655, "lr": 8.917843030516988e-05}
{"train_loss": 0.33056989312171936, "global_step": 58320, "epoch": 655, "lr": 8.917807010912903e-05}
{"train_loss": 0.18239426612854004, "global_step": 58321, "epoch": 655, "lr": 8.917770990782114e-05}
{"train_loss": 0.1883695125579834, "global_step": 58322, "epoch": 655, "lr": 8.917734970124629e-05}
{"train_loss": 0.24651609361171722, "global_step": 58323, "epoch": 655, "lr": 8.917698948940454e-05}
{"train_loss": 0.27160805463790894, "global_step": 58324, "epoch": 655, "lr": 8.91766292722959e-05}
{"train_loss": 0.31659621000289917, "global_step": 58325, "epoch": 655, "lr": 8.917626904992045e-05}
{"train_loss": 0.246124267578125, "global_step": 58326, "epoch": 655, "lr": 8.917590882227822e-05}
{"train_loss": 0.21497797966003418, "global_step": 58327, "epoch": 655, "lr": 8.917554858936927e-05}
{"train_loss": 0.22958208620548248, "global_step": 58328, "epoch": 655, "lr": 8.917518835119365e-05}
{"train_loss": 0.24858255684375763, "global_step": 58329, "epoch": 655, "lr": 8.91748281077514e-05}
{"train_loss": 0.24105583131313324, "global_step": 58330, "epoch": 655, "lr": 8.917446785904257e-05}
{"train_loss": 0.2283303588628769, "global_step": 58331, "epoch": 655, "lr": 8.917410760506721e-05}
{"train_loss": 0.2610594630241394, "global_step": 58332, "epoch": 655, "lr": 8.917374734582537e-05}
{"train_loss": 0.22815102338790894, "global_step": 58333, "epoch": 655, "lr": 8.917338708131711e-05}
{"train_loss": 0.27194008231163025, "global_step": 58334, "epoch": 655, "lr": 8.917302681154244e-05}
{"train_loss": 0.33299630880355835, "global_step": 58335, "epoch": 655, "lr": 8.917266653650145e-05}
{"train_loss": 0.2446306347846985, "global_step": 58336, "epoch": 655, "lr": 8.917230625619417e-05}
{"train_loss": 0.2781983017921448, "global_step": 58337, "epoch": 655, "lr": 8.917194597062066e-05}
{"train_loss": 0.15593387186527252, "global_step": 58338, "epoch": 655, "lr": 8.917158567978095e-05}
{"train_loss": 0.2217978835105896, "global_step": 58339, "epoch": 655, "lr": 8.917122538367509e-05}
{"train_loss": 0.22435136139392853, "global_step": 58340, "epoch": 655, "lr": 8.917086508230314e-05}
{"train_loss": 0.2556512951850891, "global_step": 58341, "epoch": 655, "lr": 8.917050477566516e-05}
{"train_loss": 0.2801932394504547, "global_step": 58342, "epoch": 655, "lr": 8.917014446376117e-05}
{"train_loss": 0.2410118579864502, "global_step": 58343, "epoch": 655, "lr": 8.916978414659125e-05}
{"train_loss": 0.21299219131469727, "global_step": 58344, "epoch": 655, "lr": 8.916942382415539e-05}
{"train_loss": 0.23738214373588562, "global_step": 58345, "epoch": 655, "lr": 8.916906349645371e-05}
{"train_loss": 0.27512702345848083, "global_step": 58346, "epoch": 655, "lr": 8.916870316348622e-05}
{"train_loss": 0.17721489071846008, "global_step": 58347, "epoch": 655, "lr": 8.916834282525297e-05}
{"train_loss": 0.24476748704910278, "global_step": 58348, "epoch": 655, "lr": 8.916798248175402e-05}
{"train_loss": 0.23855765163898468, "global_step": 58349, "epoch": 655, "lr": 8.916762213298942e-05}
{"train_loss": 0.22608493268489838, "global_step": 58350, "epoch": 655, "lr": 8.91672617789592e-05}
{"train_loss": 0.2011580616235733, "global_step": 58351, "epoch": 655, "lr": 8.916690141966342e-05}
{"train_loss": 0.22091268002986908, "global_step": 58352, "epoch": 655, "lr": 8.916654105510211e-05}
{"train_loss": 0.2989143133163452, "global_step": 58353, "epoch": 655, "lr": 8.916618068527536e-05}
{"train_loss": 0.25544875860214233, "global_step": 58354, "epoch": 655, "lr": 8.916582031018319e-05}
{"train_loss": 0.3020807206630707, "global_step": 58355, "epoch": 655, "lr": 8.916545992982565e-05}
{"train_loss": 0.1740827113389969, "global_step": 58356, "epoch": 655, "lr": 8.916509954420279e-05}
{"train_loss": 0.1961047649383545, "global_step": 58357, "epoch": 655, "lr": 8.916473915331466e-05}
{"train_loss": 0.27324265241622925, "global_step": 58358, "epoch": 655, "lr": 8.916437875716131e-05}
{"train_loss": 0.2632318139076233, "global_step": 58359, "epoch": 655, "lr": 8.916401835574279e-05}
{"train_loss": 0.22370485961437225, "global_step": 58360, "epoch": 655, "lr": 8.916365794905914e-05}
{"train_loss": 0.3781805634498596, "global_step": 58361, "epoch": 655, "lr": 8.916329753711041e-05}
{"train_loss": 0.3773382008075714, "global_step": 58362, "epoch": 655, "lr": 8.916293711989667e-05}
{"train_loss": 0.20212110877037048, "global_step": 58363, "epoch": 655, "lr": 8.916257669741793e-05}
{"train_loss": 0.15824592113494873, "global_step": 58364, "epoch": 655, "lr": 8.916221626967426e-05}
{"train_loss": 0.17608290910720825, "global_step": 58365, "epoch": 655, "lr": 8.916185583666572e-05}
{"train_loss": 0.18861329555511475, "global_step": 58366, "epoch": 655, "lr": 8.916149539839233e-05}
{"train_loss": 0.2545788586139679, "global_step": 58367, "epoch": 655, "lr": 8.916113495485416e-05}
{"train_loss": 0.31071245670318604, "global_step": 58368, "epoch": 655, "lr": 8.916077450605126e-05}
{"train_loss": 0.2898109555244446, "global_step": 58369, "epoch": 655, "lr": 8.916041405198368e-05}
{"train_loss": 0.2044627070426941, "global_step": 58370, "epoch": 655, "lr": 8.916005359265143e-05}
{"train_loss": 0.3644479513168335, "global_step": 58371, "epoch": 655, "lr": 8.915969312805462e-05}
{"train_loss": 0.20716804265975952, "global_step": 58372, "epoch": 655, "lr": 8.915933265819325e-05}
{"train_loss": 0.36809754371643066, "global_step": 58373, "epoch": 655, "lr": 8.915897218306738e-05}
{"train_loss": 0.21942217648029327, "global_step": 58374, "epoch": 655, "lr": 8.915861170267708e-05}
{"train_loss": 0.30806559324264526, "global_step": 58375, "epoch": 655, "lr": 8.915825121702237e-05}
{"train_loss": 0.2639178931713104, "global_step": 58376, "epoch": 655, "lr": 8.91578907261033e-05}
{"train_loss": 0.20510634779930115, "global_step": 58377, "epoch": 655, "lr": 8.915753022991995e-05}
{"train_loss": 0.2632567882537842, "global_step": 58378, "epoch": 655, "lr": 8.915716972847233e-05}
{"train_loss": 0.29652249813079834, "global_step": 58379, "epoch": 655, "lr": 8.915680922176051e-05}
{"train_loss": 0.18210944533348083, "global_step": 58380, "epoch": 655, "lr": 8.915644870978455e-05}
{"train_loss": 0.2228861302137375, "global_step": 58381, "epoch": 655, "lr": 8.915608819254447e-05}
{"train_loss": 0.2596721053123474, "global_step": 58382, "epoch": 655, "lr": 8.915572767004032e-05}
{"train_loss": 0.24318211215935395, "global_step": 58383, "epoch": 655, "lr": 8.915536714227217e-05, "val_loss": 3.300218105316162, "train_action_mse_error": 37.803466796875}
{"train_loss": 0.2350931316614151, "global_step": 58384, "epoch": 656, "lr": 8.915500660924006e-05}
{"train_loss": 0.21793656051158905, "global_step": 58385, "epoch": 656, "lr": 8.915464607094404e-05}
{"train_loss": 0.24796335399150848, "global_step": 58386, "epoch": 656, "lr": 8.915428552738416e-05}
{"train_loss": 0.26322004199028015, "global_step": 58387, "epoch": 656, "lr": 8.915392497856045e-05}
{"train_loss": 0.24490411579608917, "global_step": 58388, "epoch": 656, "lr": 8.915356442447299e-05}
{"train_loss": 0.21532896161079407, "global_step": 58389, "epoch": 656, "lr": 8.915320386512178e-05}
{"train_loss": 0.29551857709884644, "global_step": 58390, "epoch": 656, "lr": 8.915284330050693e-05}
{"train_loss": 0.24884769320487976, "global_step": 58391, "epoch": 656, "lr": 8.915248273062844e-05}
{"train_loss": 0.22758236527442932, "global_step": 58392, "epoch": 656, "lr": 8.915212215548639e-05}
{"train_loss": 0.2684830129146576, "global_step": 58393, "epoch": 656, "lr": 8.915176157508079e-05}
{"train_loss": 0.2986544668674469, "global_step": 58394, "epoch": 656, "lr": 8.915140098941172e-05}
{"train_loss": 0.32432207465171814, "global_step": 58395, "epoch": 656, "lr": 8.915104039847924e-05}
{"train_loss": 0.24712342023849487, "global_step": 58396, "epoch": 656, "lr": 8.915067980228337e-05}
{"train_loss": 0.2539263665676117, "global_step": 58397, "epoch": 656, "lr": 8.915031920082416e-05}
{"train_loss": 0.30995431542396545, "global_step": 58398, "epoch": 656, "lr": 8.914995859410166e-05}
{"train_loss": 0.25219762325286865, "global_step": 58399, "epoch": 656, "lr": 8.914959798211596e-05}
{"train_loss": 0.2148849070072174, "global_step": 58400, "epoch": 656, "lr": 8.914923736486704e-05}
{"train_loss": 0.21423554420471191, "global_step": 58401, "epoch": 656, "lr": 8.914887674235501e-05}
{"train_loss": 0.19779269397258759, "global_step": 58402, "epoch": 656, "lr": 8.914851611457986e-05}
{"train_loss": 0.22473004460334778, "global_step": 58403, "epoch": 656, "lr": 8.91481554815417e-05}
{"train_loss": 0.29485806822776794, "global_step": 58404, "epoch": 656, "lr": 8.914779484324052e-05}
{"train_loss": 0.23721612989902496, "global_step": 58405, "epoch": 656, "lr": 8.914743419967642e-05}
{"train_loss": 0.19180917739868164, "global_step": 58406, "epoch": 656, "lr": 8.91470735508494e-05}
{"train_loss": 0.240155890583992, "global_step": 58407, "epoch": 656, "lr": 8.914671289675956e-05}
{"train_loss": 0.21677948534488678, "global_step": 58408, "epoch": 656, "lr": 8.91463522374069e-05}
{"train_loss": 0.24987053871154785, "global_step": 58409, "epoch": 656, "lr": 8.914599157279149e-05}
{"train_loss": 0.2776373028755188, "global_step": 58410, "epoch": 656, "lr": 8.91456309029134e-05}
{"train_loss": 0.20375601947307587, "global_step": 58411, "epoch": 656, "lr": 8.914527022777264e-05}
{"train_loss": 0.24277427792549133, "global_step": 58412, "epoch": 656, "lr": 8.914490954736926e-05}
{"train_loss": 0.25473499298095703, "global_step": 58413, "epoch": 656, "lr": 8.914454886170336e-05}
{"train_loss": 0.32243219017982483, "global_step": 58414, "epoch": 656, "lr": 8.914418817077493e-05}
{"train_loss": 0.22081349790096283, "global_step": 58415, "epoch": 656, "lr": 8.914382747458405e-05}
{"train_loss": 0.24453863501548767, "global_step": 58416, "epoch": 656, "lr": 8.914346677313075e-05}
{"train_loss": 0.2521495819091797, "global_step": 58417, "epoch": 656, "lr": 8.914310606641509e-05}
{"train_loss": 0.1703016459941864, "global_step": 58418, "epoch": 656, "lr": 8.914274535443711e-05}
{"train_loss": 0.21088460087776184, "global_step": 58419, "epoch": 656, "lr": 8.914238463719688e-05}
{"train_loss": 0.2853904664516449, "global_step": 58420, "epoch": 656, "lr": 8.914202391469444e-05}
{"train_loss": 0.27921611070632935, "global_step": 58421, "epoch": 656, "lr": 8.914166318692983e-05}
{"train_loss": 0.24547918140888214, "global_step": 58422, "epoch": 656, "lr": 8.914130245390309e-05}
{"train_loss": 0.24908547103405, "global_step": 58423, "epoch": 656, "lr": 8.914094171561427e-05}
{"train_loss": 0.24815388023853302, "global_step": 58424, "epoch": 656, "lr": 8.914058097206343e-05}
{"train_loss": 0.21386504173278809, "global_step": 58425, "epoch": 656, "lr": 8.914022022325064e-05}
{"train_loss": 0.21906349062919617, "global_step": 58426, "epoch": 656, "lr": 8.91398594691759e-05}
{"train_loss": 0.17866206169128418, "global_step": 58427, "epoch": 656, "lr": 8.913949870983928e-05}
{"train_loss": 0.17098887264728546, "global_step": 58428, "epoch": 656, "lr": 8.913913794524084e-05}
{"train_loss": 0.3405466675758362, "global_step": 58429, "epoch": 656, "lr": 8.913877717538064e-05}
{"train_loss": 0.1589188575744629, "global_step": 58430, "epoch": 656, "lr": 8.913841640025869e-05}
{"train_loss": 0.17261826992034912, "global_step": 58431, "epoch": 656, "lr": 8.913805561987505e-05}
{"train_loss": 0.19096988439559937, "global_step": 58432, "epoch": 656, "lr": 8.913769483422978e-05}
{"train_loss": 0.28309836983680725, "global_step": 58433, "epoch": 656, "lr": 8.913733404332293e-05}
{"train_loss": 0.20536550879478455, "global_step": 58434, "epoch": 656, "lr": 8.913697324715455e-05}
{"train_loss": 0.16970224678516388, "global_step": 58435, "epoch": 656, "lr": 8.913661244572466e-05}
{"train_loss": 0.3287593722343445, "global_step": 58436, "epoch": 656, "lr": 8.913625163903334e-05}
{"train_loss": 0.27501004934310913, "global_step": 58437, "epoch": 656, "lr": 8.913589082708062e-05}
{"train_loss": 0.1997172236442566, "global_step": 58438, "epoch": 656, "lr": 8.913553000986658e-05}
{"train_loss": 0.23611995577812195, "global_step": 58439, "epoch": 656, "lr": 8.913516918739122e-05}
{"train_loss": 0.15089014172554016, "global_step": 58440, "epoch": 656, "lr": 8.913480835965462e-05}
{"train_loss": 0.3085387051105499, "global_step": 58441, "epoch": 656, "lr": 8.913444752665682e-05}
{"train_loss": 0.3065966069698334, "global_step": 58442, "epoch": 656, "lr": 8.913408668839789e-05}
{"train_loss": 0.19351644814014435, "global_step": 58443, "epoch": 656, "lr": 8.913372584487784e-05}
{"train_loss": 0.2503901720046997, "global_step": 58444, "epoch": 656, "lr": 8.913336499609674e-05}
{"train_loss": 0.1554918885231018, "global_step": 58445, "epoch": 656, "lr": 8.913300414205463e-05}
{"train_loss": 0.2910306453704834, "global_step": 58446, "epoch": 656, "lr": 8.913264328275158e-05}
{"train_loss": 0.340468168258667, "global_step": 58447, "epoch": 656, "lr": 8.913228241818762e-05}
{"train_loss": 0.3570590913295746, "global_step": 58448, "epoch": 656, "lr": 8.913192154836279e-05}
{"train_loss": 0.1482386291027069, "global_step": 58449, "epoch": 656, "lr": 8.913156067327717e-05}
{"train_loss": 0.2732512056827545, "global_step": 58450, "epoch": 656, "lr": 8.913119979293076e-05}
{"train_loss": 0.21300718188285828, "global_step": 58451, "epoch": 656, "lr": 8.913083890732365e-05}
{"train_loss": 0.27597492933273315, "global_step": 58452, "epoch": 656, "lr": 8.913047801645588e-05}
{"train_loss": 0.23883569240570068, "global_step": 58453, "epoch": 656, "lr": 8.913011712032751e-05}
{"train_loss": 0.1974710077047348, "global_step": 58454, "epoch": 656, "lr": 8.912975621893854e-05}
{"train_loss": 0.31601670384407043, "global_step": 58455, "epoch": 656, "lr": 8.912939531228908e-05}
{"train_loss": 0.31976884603500366, "global_step": 58456, "epoch": 656, "lr": 8.912903440037913e-05}
{"train_loss": 0.30530470609664917, "global_step": 58457, "epoch": 656, "lr": 8.912867348320877e-05}
{"train_loss": 0.2932256758213043, "global_step": 58458, "epoch": 656, "lr": 8.912831256077803e-05}
{"train_loss": 0.2517967224121094, "global_step": 58459, "epoch": 656, "lr": 8.912795163308696e-05}
{"train_loss": 0.15116341412067413, "global_step": 58460, "epoch": 656, "lr": 8.912759070013562e-05}
{"train_loss": 0.18067115545272827, "global_step": 58461, "epoch": 656, "lr": 8.912722976192406e-05}
{"train_loss": 0.28780597448349, "global_step": 58462, "epoch": 656, "lr": 8.912686881845231e-05}
{"train_loss": 0.3247188925743103, "global_step": 58463, "epoch": 656, "lr": 8.912650786972043e-05}
{"train_loss": 0.23379798233509064, "global_step": 58464, "epoch": 656, "lr": 8.912614691572847e-05}
{"train_loss": 0.20807263255119324, "global_step": 58465, "epoch": 656, "lr": 8.912578595647648e-05}
{"train_loss": 0.23640908300876617, "global_step": 58466, "epoch": 656, "lr": 8.91254249919645e-05}
{"train_loss": 0.19855248928070068, "global_step": 58467, "epoch": 656, "lr": 8.912506402219259e-05}
{"train_loss": 0.1901535540819168, "global_step": 58468, "epoch": 656, "lr": 8.912470304716078e-05}
{"train_loss": 0.26248443126678467, "global_step": 58469, "epoch": 656, "lr": 8.912434206686914e-05}
{"train_loss": 0.22816088795661926, "global_step": 58470, "epoch": 656, "lr": 8.912398108131771e-05}
{"train_loss": 0.31270071864128113, "global_step": 58471, "epoch": 656, "lr": 8.912362009050654e-05}
{"train_loss": 0.24476855306812886, "global_step": 58472, "epoch": 656, "lr": 8.912325909443566e-05, "val_loss": 3.285362482070923}
{"train_loss": 0.3004182279109955, "global_step": 58473, "epoch": 657, "lr": 8.912289809310515e-05}
{"train_loss": 0.3057194948196411, "global_step": 58474, "epoch": 657, "lr": 8.912253708651504e-05}
{"train_loss": 0.2084934115409851, "global_step": 58475, "epoch": 657, "lr": 8.912217607466538e-05}
{"train_loss": 0.25102782249450684, "global_step": 58476, "epoch": 657, "lr": 8.912181505755623e-05}
{"train_loss": 0.26728346943855286, "global_step": 58477, "epoch": 657, "lr": 8.912145403518761e-05}
{"train_loss": 0.242386132478714, "global_step": 58478, "epoch": 657, "lr": 8.91210930075596e-05}
{"train_loss": 0.32786017656326294, "global_step": 58479, "epoch": 657, "lr": 8.912073197467223e-05}
{"train_loss": 0.2717745304107666, "global_step": 58480, "epoch": 657, "lr": 8.912037093652557e-05}
{"train_loss": 0.21673746407032013, "global_step": 58481, "epoch": 657, "lr": 8.912000989311962e-05}
{"train_loss": 0.31147950887680054, "global_step": 58482, "epoch": 657, "lr": 8.91196488444545e-05}
{"train_loss": 0.3039104640483856, "global_step": 58483, "epoch": 657, "lr": 8.911928779053019e-05}
{"train_loss": 0.2117430865764618, "global_step": 58484, "epoch": 657, "lr": 8.911892673134678e-05}
{"train_loss": 0.21614742279052734, "global_step": 58485, "epoch": 657, "lr": 8.91185656669043e-05}
{"train_loss": 0.23380836844444275, "global_step": 58486, "epoch": 657, "lr": 8.911820459720282e-05}
{"train_loss": 0.3264692425727844, "global_step": 58487, "epoch": 657, "lr": 8.911784352224238e-05}
{"train_loss": 0.2927958369255066, "global_step": 58488, "epoch": 657, "lr": 8.9117482442023e-05}
{"train_loss": 0.20288477838039398, "global_step": 58489, "epoch": 657, "lr": 8.911712135654476e-05}
{"train_loss": 0.20683635771274567, "global_step": 58490, "epoch": 657, "lr": 8.911676026580771e-05}
{"train_loss": 0.18961913883686066, "global_step": 58491, "epoch": 657, "lr": 8.911639916981187e-05}
{"train_loss": 0.2148270308971405, "global_step": 58492, "epoch": 657, "lr": 8.911603806855732e-05}
{"train_loss": 0.23804111778736115, "global_step": 58493, "epoch": 657, "lr": 8.911567696204408e-05}
{"train_loss": 0.1736624836921692, "global_step": 58494, "epoch": 657, "lr": 8.911531585027224e-05}
{"train_loss": 0.253818541765213, "global_step": 58495, "epoch": 657, "lr": 8.91149547332418e-05}
{"train_loss": 0.2718733251094818, "global_step": 58496, "epoch": 657, "lr": 8.911459361095285e-05}
{"train_loss": 0.20533563196659088, "global_step": 58497, "epoch": 657, "lr": 8.91142324834054e-05}
{"train_loss": 0.1931552290916443, "global_step": 58498, "epoch": 657, "lr": 8.911387135059953e-05}
{"train_loss": 0.26681336760520935, "global_step": 58499, "epoch": 657, "lr": 8.911351021253526e-05}
{"train_loss": 0.19914740324020386, "global_step": 58500, "epoch": 657, "lr": 8.911314906921269e-05}
{"train_loss": 0.2750573754310608, "global_step": 58501, "epoch": 657, "lr": 8.91127879206318e-05}
{"train_loss": 0.22082695364952087, "global_step": 58502, "epoch": 657, "lr": 8.911242676679269e-05}
{"train_loss": 0.29683196544647217, "global_step": 58503, "epoch": 657, "lr": 8.911206560769538e-05}
{"train_loss": 0.3359821140766144, "global_step": 58504, "epoch": 657, "lr": 8.911170444333995e-05}
{"train_loss": 0.3259888291358948, "global_step": 58505, "epoch": 657, "lr": 8.911134327372641e-05}
{"train_loss": 0.2152879387140274, "global_step": 58506, "epoch": 657, "lr": 8.911098209885483e-05}
{"train_loss": 0.23567157983779907, "global_step": 58507, "epoch": 657, "lr": 8.911062091872526e-05}
{"train_loss": 0.16067883372306824, "global_step": 58508, "epoch": 657, "lr": 8.911025973333773e-05}
{"train_loss": 0.33201920986175537, "global_step": 58509, "epoch": 657, "lr": 8.910989854269232e-05}
{"train_loss": 0.3465619385242462, "global_step": 58510, "epoch": 657, "lr": 8.910953734678904e-05}
{"train_loss": 0.2691028118133545, "global_step": 58511, "epoch": 657, "lr": 8.910917614562797e-05}
{"train_loss": 0.20406678318977356, "global_step": 58512, "epoch": 657, "lr": 8.910881493920915e-05}
{"train_loss": 0.17361563444137573, "global_step": 58513, "epoch": 657, "lr": 8.910845372753262e-05}
{"train_loss": 0.2948419451713562, "global_step": 58514, "epoch": 657, "lr": 8.910809251059847e-05}
{"train_loss": 0.19706347584724426, "global_step": 58515, "epoch": 657, "lr": 8.910773128840667e-05}
{"train_loss": 0.31114500761032104, "global_step": 58516, "epoch": 657, "lr": 8.910737006095733e-05}
{"train_loss": 0.2511165142059326, "global_step": 58517, "epoch": 657, "lr": 8.910700882825047e-05}
{"train_loss": 0.30721837282180786, "global_step": 58518, "epoch": 657, "lr": 8.910664759028615e-05}
{"train_loss": 0.3073651194572449, "global_step": 58519, "epoch": 657, "lr": 8.910628634706444e-05}
{"train_loss": 0.32679182291030884, "global_step": 58520, "epoch": 657, "lr": 8.910592509858534e-05}
{"train_loss": 0.23270992934703827, "global_step": 58521, "epoch": 657, "lr": 8.910556384484894e-05}
{"train_loss": 0.3720095455646515, "global_step": 58522, "epoch": 657, "lr": 8.910520258585526e-05}
{"train_loss": 0.27353113889694214, "global_step": 58523, "epoch": 657, "lr": 8.910484132160439e-05}
{"train_loss": 0.3255155384540558, "global_step": 58524, "epoch": 657, "lr": 8.910448005209633e-05}
{"train_loss": 0.25362029671669006, "global_step": 58525, "epoch": 657, "lr": 8.910411877733114e-05}
{"train_loss": 0.2133890837430954, "global_step": 58526, "epoch": 657, "lr": 8.91037574973089e-05}
{"train_loss": 0.2056611180305481, "global_step": 58527, "epoch": 657, "lr": 8.910339621202962e-05}
{"train_loss": 0.3378988802433014, "global_step": 58528, "epoch": 657, "lr": 8.910303492149337e-05}
{"train_loss": 0.29780545830726624, "global_step": 58529, "epoch": 657, "lr": 8.91026736257002e-05}
{"train_loss": 0.26060622930526733, "global_step": 58530, "epoch": 657, "lr": 8.910231232465014e-05}
{"train_loss": 0.19819562137126923, "global_step": 58531, "epoch": 657, "lr": 8.910195101834327e-05}
{"train_loss": 0.20122894644737244, "global_step": 58532, "epoch": 657, "lr": 8.91015897067796e-05}
{"train_loss": 0.29421135783195496, "global_step": 58533, "epoch": 657, "lr": 8.91012283899592e-05}
{"train_loss": 0.22114527225494385, "global_step": 58534, "epoch": 657, "lr": 8.910086706788212e-05}
{"train_loss": 0.22709433734416962, "global_step": 58535, "epoch": 657, "lr": 8.910050574054841e-05}
{"train_loss": 0.21644651889801025, "global_step": 58536, "epoch": 657, "lr": 8.910014440795811e-05}
{"train_loss": 0.2623898386955261, "global_step": 58537, "epoch": 657, "lr": 8.909978307011127e-05}
{"train_loss": 0.21084432303905487, "global_step": 58538, "epoch": 657, "lr": 8.909942172700795e-05}
{"train_loss": 0.2719796597957611, "global_step": 58539, "epoch": 657, "lr": 8.909906037864818e-05}
{"train_loss": 0.23370495438575745, "global_step": 58540, "epoch": 657, "lr": 8.909869902503201e-05}
{"train_loss": 0.262677401304245, "global_step": 58541, "epoch": 657, "lr": 8.909833766615951e-05}
{"train_loss": 0.23552413284778595, "global_step": 58542, "epoch": 657, "lr": 8.909797630203072e-05}
{"train_loss": 0.2841207981109619, "global_step": 58543, "epoch": 657, "lr": 8.909761493264567e-05}
{"train_loss": 0.21463675796985626, "global_step": 58544, "epoch": 657, "lr": 8.909725355800442e-05}
{"train_loss": 0.2370349019765854, "global_step": 58545, "epoch": 657, "lr": 8.909689217810704e-05}
{"train_loss": 0.2723797559738159, "global_step": 58546, "epoch": 657, "lr": 8.909653079295355e-05}
{"train_loss": 0.15668447315692902, "global_step": 58547, "epoch": 657, "lr": 8.9096169402544e-05}
{"train_loss": 0.25693994760513306, "global_step": 58548, "epoch": 657, "lr": 8.909580800687845e-05}
{"train_loss": 0.224164217710495, "global_step": 58549, "epoch": 657, "lr": 8.909544660595694e-05}
{"train_loss": 0.15787412226200104, "global_step": 58550, "epoch": 657, "lr": 8.909508519977954e-05}
{"train_loss": 0.23544088006019592, "global_step": 58551, "epoch": 657, "lr": 8.909472378834627e-05}
{"train_loss": 0.30814364552497864, "global_step": 58552, "epoch": 657, "lr": 8.90943623716572e-05}
{"train_loss": 0.18664133548736572, "global_step": 58553, "epoch": 657, "lr": 8.909400094971237e-05}
{"train_loss": 0.266053169965744, "global_step": 58554, "epoch": 657, "lr": 8.90936395225118e-05}
{"train_loss": 0.2229723334312439, "global_step": 58555, "epoch": 657, "lr": 8.90932780900556e-05}
{"train_loss": 0.26278752088546753, "global_step": 58556, "epoch": 657, "lr": 8.909291665234377e-05}
{"train_loss": 0.3284904360771179, "global_step": 58557, "epoch": 657, "lr": 8.909255520937638e-05}
{"train_loss": 0.2063169628381729, "global_step": 58558, "epoch": 657, "lr": 8.909219376115348e-05}
{"train_loss": 0.1963939219713211, "global_step": 58559, "epoch": 657, "lr": 8.909183230767508e-05}
{"train_loss": 0.20814773440361023, "global_step": 58560, "epoch": 657, "lr": 8.909147084894129e-05}
{"train_loss": 0.25180450299482665, "global_step": 58561, "epoch": 657, "lr": 8.909110938495211e-05, "val_loss": 3.446070432662964}
{"train_loss": 0.2546868920326233, "global_step": 58562, "epoch": 658, "lr": 8.909074791570761e-05}
{"train_loss": 0.148235023021698, "global_step": 58563, "epoch": 658, "lr": 8.909038644120785e-05}
{"train_loss": 0.2412329614162445, "global_step": 58564, "epoch": 658, "lr": 8.909002496145286e-05}
{"train_loss": 0.2050015777349472, "global_step": 58565, "epoch": 658, "lr": 8.908966347644268e-05}
{"train_loss": 0.298855721950531, "global_step": 58566, "epoch": 658, "lr": 8.908930198617737e-05}
{"train_loss": 0.22457298636436462, "global_step": 58567, "epoch": 658, "lr": 8.908894049065699e-05}
{"train_loss": 0.2312154769897461, "global_step": 58568, "epoch": 658, "lr": 8.908857898988159e-05}
{"train_loss": 0.3461415469646454, "global_step": 58569, "epoch": 658, "lr": 8.908821748385118e-05}
{"train_loss": 0.26434335112571716, "global_step": 58570, "epoch": 658, "lr": 8.908785597256585e-05}
{"train_loss": 0.17718853056430817, "global_step": 58571, "epoch": 658, "lr": 8.908749445602564e-05}
{"train_loss": 0.21591196954250336, "global_step": 58572, "epoch": 658, "lr": 8.908713293423058e-05}
{"train_loss": 0.1865048110485077, "global_step": 58573, "epoch": 658, "lr": 8.908677140718075e-05}
{"train_loss": 0.2755142152309418, "global_step": 58574, "epoch": 658, "lr": 8.908640987487617e-05}
{"train_loss": 0.25980278849601746, "global_step": 58575, "epoch": 658, "lr": 8.908604833731689e-05}
{"train_loss": 0.15503980219364166, "global_step": 58576, "epoch": 658, "lr": 8.908568679450297e-05}
{"train_loss": 0.20197919011116028, "global_step": 58577, "epoch": 658, "lr": 8.908532524643446e-05}
{"train_loss": 0.21996288001537323, "global_step": 58578, "epoch": 658, "lr": 8.90849636931114e-05}
{"train_loss": 0.23659050464630127, "global_step": 58579, "epoch": 658, "lr": 8.908460213453385e-05}
{"train_loss": 0.2619245946407318, "global_step": 58580, "epoch": 658, "lr": 8.908424057070184e-05}
{"train_loss": 0.218510702252388, "global_step": 58581, "epoch": 658, "lr": 8.908387900161545e-05}
{"train_loss": 0.209805428981781, "global_step": 58582, "epoch": 658, "lr": 8.908351742727469e-05}
{"train_loss": 0.22981232404708862, "global_step": 58583, "epoch": 658, "lr": 8.908315584767964e-05}
{"train_loss": 0.2640798091888428, "global_step": 58584, "epoch": 658, "lr": 8.908279426283032e-05}
{"train_loss": 0.29350024461746216, "global_step": 58585, "epoch": 658, "lr": 8.908243267272681e-05}
{"train_loss": 0.22619636356830597, "global_step": 58586, "epoch": 658, "lr": 8.908207107736914e-05}
{"train_loss": 0.2566286027431488, "global_step": 58587, "epoch": 658, "lr": 8.908170947675736e-05}
{"train_loss": 0.12730488181114197, "global_step": 58588, "epoch": 658, "lr": 8.908134787089153e-05}
{"train_loss": 0.31863704323768616, "global_step": 58589, "epoch": 658, "lr": 8.908098625977167e-05}
{"train_loss": 0.18315164744853973, "global_step": 58590, "epoch": 658, "lr": 8.908062464339786e-05}
{"train_loss": 0.21888770163059235, "global_step": 58591, "epoch": 658, "lr": 8.908026302177013e-05}
{"train_loss": 0.24483110010623932, "global_step": 58592, "epoch": 658, "lr": 8.907990139488855e-05}
{"train_loss": 0.246918186545372, "global_step": 58593, "epoch": 658, "lr": 8.907953976275313e-05}
{"train_loss": 0.28289279341697693, "global_step": 58594, "epoch": 658, "lr": 8.907917812536397e-05}
{"train_loss": 0.20826758444309235, "global_step": 58595, "epoch": 658, "lr": 8.907881648272108e-05}
{"train_loss": 0.21362701058387756, "global_step": 58596, "epoch": 658, "lr": 8.90784548348245e-05}
{"train_loss": 0.2087736576795578, "global_step": 58597, "epoch": 658, "lr": 8.907809318167432e-05}
{"train_loss": 0.25722771883010864, "global_step": 58598, "epoch": 658, "lr": 8.907773152327055e-05}
{"train_loss": 0.3213382363319397, "global_step": 58599, "epoch": 658, "lr": 8.907736985961327e-05}
{"train_loss": 0.3125220537185669, "global_step": 58600, "epoch": 658, "lr": 8.90770081907025e-05}
{"train_loss": 0.24481454491615295, "global_step": 58601, "epoch": 658, "lr": 8.907664651653832e-05}
{"train_loss": 0.19321605563163757, "global_step": 58602, "epoch": 658, "lr": 8.907628483712075e-05}
{"train_loss": 0.19306133687496185, "global_step": 58603, "epoch": 658, "lr": 8.907592315244985e-05}
{"train_loss": 0.21600252389907837, "global_step": 58604, "epoch": 658, "lr": 8.907556146252567e-05}
{"train_loss": 0.23982982337474823, "global_step": 58605, "epoch": 658, "lr": 8.907519976734826e-05}
{"train_loss": 0.20917077362537384, "global_step": 58606, "epoch": 658, "lr": 8.907483806691767e-05}
{"train_loss": 0.2429889738559723, "global_step": 58607, "epoch": 658, "lr": 8.907447636123394e-05}
{"train_loss": 0.2476845681667328, "global_step": 58608, "epoch": 658, "lr": 8.907411465029712e-05}
{"train_loss": 0.2498893439769745, "global_step": 58609, "epoch": 658, "lr": 8.907375293410725e-05}
{"train_loss": 0.20959779620170593, "global_step": 58610, "epoch": 658, "lr": 8.90733912126644e-05}
{"train_loss": 0.22008724510669708, "global_step": 58611, "epoch": 658, "lr": 8.907302948596862e-05}
{"train_loss": 0.27681875228881836, "global_step": 58612, "epoch": 658, "lr": 8.907266775401994e-05}
{"train_loss": 0.22179478406906128, "global_step": 58613, "epoch": 658, "lr": 8.907230601681842e-05}
{"train_loss": 0.35746777057647705, "global_step": 58614, "epoch": 658, "lr": 8.907194427436409e-05}
{"train_loss": 0.28663554787635803, "global_step": 58615, "epoch": 658, "lr": 8.907158252665701e-05}
{"train_loss": 0.2132086306810379, "global_step": 58616, "epoch": 658, "lr": 8.907122077369724e-05}
{"train_loss": 0.26964184641838074, "global_step": 58617, "epoch": 658, "lr": 8.907085901548484e-05}
{"train_loss": 0.2470247894525528, "global_step": 58618, "epoch": 658, "lr": 8.907049725201983e-05}
{"train_loss": 0.26926738023757935, "global_step": 58619, "epoch": 658, "lr": 8.907013548330226e-05}
{"train_loss": 0.29170534014701843, "global_step": 58620, "epoch": 658, "lr": 8.906977370933218e-05}
{"train_loss": 0.21327310800552368, "global_step": 58621, "epoch": 658, "lr": 8.906941193010968e-05}
{"train_loss": 0.35271885991096497, "global_step": 58622, "epoch": 658, "lr": 8.906905014563474e-05}
{"train_loss": 0.1900126039981842, "global_step": 58623, "epoch": 658, "lr": 8.906868835590744e-05}
{"train_loss": 0.28098636865615845, "global_step": 58624, "epoch": 658, "lr": 8.906832656092784e-05}
{"train_loss": 0.19277600944042206, "global_step": 58625, "epoch": 658, "lr": 8.9067964760696e-05}
{"train_loss": 0.3085581958293915, "global_step": 58626, "epoch": 658, "lr": 8.90676029552119e-05}
{"train_loss": 0.3635556101799011, "global_step": 58627, "epoch": 658, "lr": 8.906724114447568e-05}
{"train_loss": 0.26754435896873474, "global_step": 58628, "epoch": 658, "lr": 8.906687932848733e-05}
{"train_loss": 0.3054294288158417, "global_step": 58629, "epoch": 658, "lr": 8.90665175072469e-05}
{"train_loss": 0.3642613887786865, "global_step": 58630, "epoch": 658, "lr": 8.906615568075448e-05}
{"train_loss": 0.24492090940475464, "global_step": 58631, "epoch": 658, "lr": 8.906579384901008e-05}
{"train_loss": 0.23143745958805084, "global_step": 58632, "epoch": 658, "lr": 8.906543201201376e-05}
{"train_loss": 0.2139376848936081, "global_step": 58633, "epoch": 658, "lr": 8.906507016976558e-05}
{"train_loss": 0.22076691687107086, "global_step": 58634, "epoch": 658, "lr": 8.906470832226556e-05}
{"train_loss": 0.2442675679922104, "global_step": 58635, "epoch": 658, "lr": 8.906434646951376e-05}
{"train_loss": 0.2580266296863556, "global_step": 58636, "epoch": 658, "lr": 8.906398461151026e-05}
{"train_loss": 0.4211169183254242, "global_step": 58637, "epoch": 658, "lr": 8.906362274825506e-05}
{"train_loss": 0.2399311512708664, "global_step": 58638, "epoch": 658, "lr": 8.906326087974826e-05}
{"train_loss": 0.231703519821167, "global_step": 58639, "epoch": 658, "lr": 8.906289900598986e-05}
{"train_loss": 0.3190310299396515, "global_step": 58640, "epoch": 658, "lr": 8.906253712697993e-05}
{"train_loss": 0.2435334175825119, "global_step": 58641, "epoch": 658, "lr": 8.906217524271851e-05}
{"train_loss": 0.3294965922832489, "global_step": 58642, "epoch": 658, "lr": 8.906181335320568e-05}
{"train_loss": 0.23877759277820587, "global_step": 58643, "epoch": 658, "lr": 8.906145145844145e-05}
{"train_loss": 0.3218262195587158, "global_step": 58644, "epoch": 658, "lr": 8.906108955842588e-05}
{"train_loss": 0.2554827630519867, "global_step": 58645, "epoch": 658, "lr": 8.906072765315903e-05}
{"train_loss": 0.22210128605365753, "global_step": 58646, "epoch": 658, "lr": 8.906036574264094e-05}
{"train_loss": 0.21488799154758453, "global_step": 58647, "epoch": 658, "lr": 8.906000382687166e-05}
{"train_loss": 0.2593127489089966, "global_step": 58648, "epoch": 658, "lr": 8.905964190585124e-05}
{"train_loss": 0.2551974356174469, "global_step": 58649, "epoch": 658, "lr": 8.905927997957972e-05}
{"train_loss": 0.24865451220715984, "global_step": 58650, "epoch": 658, "lr": 8.905891804805717e-05, "val_loss": 3.1378724575042725}
{"train_loss": 0.24199901521205902, "global_step": 58651, "epoch": 659, "lr": 8.905855611128362e-05}
{"train_loss": 0.24293558299541473, "global_step": 58652, "epoch": 659, "lr": 8.905819416925911e-05}
{"train_loss": 0.234109029173851, "global_step": 58653, "epoch": 659, "lr": 8.905783222198372e-05}
{"train_loss": 0.2842051684856415, "global_step": 58654, "epoch": 659, "lr": 8.905747026945746e-05}
{"train_loss": 0.2683863043785095, "global_step": 58655, "epoch": 659, "lr": 8.905710831168041e-05}
{"train_loss": 0.30303865671157837, "global_step": 58656, "epoch": 659, "lr": 8.905674634865262e-05}
{"train_loss": 0.2638919949531555, "global_step": 58657, "epoch": 659, "lr": 8.905638438037411e-05}
{"train_loss": 0.25221163034439087, "global_step": 58658, "epoch": 659, "lr": 8.905602240684496e-05}
{"train_loss": 0.28606951236724854, "global_step": 58659, "epoch": 659, "lr": 8.905566042806519e-05}
{"train_loss": 0.25076597929000854, "global_step": 58660, "epoch": 659, "lr": 8.905529844403486e-05}
{"train_loss": 0.2941267192363739, "global_step": 58661, "epoch": 659, "lr": 8.905493645475405e-05}
{"train_loss": 0.20460225641727448, "global_step": 58662, "epoch": 659, "lr": 8.905457446022274e-05}
{"train_loss": 0.17315204441547394, "global_step": 58663, "epoch": 659, "lr": 8.905421246044104e-05}
{"train_loss": 0.3077884316444397, "global_step": 58664, "epoch": 659, "lr": 8.905385045540897e-05}
{"train_loss": 0.2943767011165619, "global_step": 58665, "epoch": 659, "lr": 8.905348844512659e-05}
{"train_loss": 0.19515086710453033, "global_step": 58666, "epoch": 659, "lr": 8.905312642959394e-05}
{"train_loss": 0.30075475573539734, "global_step": 58667, "epoch": 659, "lr": 8.905276440881108e-05}
{"train_loss": 0.2601311206817627, "global_step": 58668, "epoch": 659, "lr": 8.905240238277806e-05}
{"train_loss": 0.17520350217819214, "global_step": 58669, "epoch": 659, "lr": 8.90520403514949e-05}
{"train_loss": 0.23174242675304413, "global_step": 58670, "epoch": 659, "lr": 8.905167831496168e-05}
{"train_loss": 0.1867484301328659, "global_step": 58671, "epoch": 659, "lr": 8.905131627317845e-05}
{"train_loss": 0.25200381875038147, "global_step": 58672, "epoch": 659, "lr": 8.905095422614523e-05}
{"train_loss": 0.23103469610214233, "global_step": 58673, "epoch": 659, "lr": 8.905059217386207e-05}
{"train_loss": 0.2881491482257843, "global_step": 58674, "epoch": 659, "lr": 8.905023011632906e-05}
{"train_loss": 0.23422831296920776, "global_step": 58675, "epoch": 659, "lr": 8.904986805354622e-05}
{"train_loss": 0.23114947974681854, "global_step": 58676, "epoch": 659, "lr": 8.904950598551359e-05}
{"train_loss": 0.29961958527565, "global_step": 58677, "epoch": 659, "lr": 8.904914391223123e-05}
{"train_loss": 0.24836605787277222, "global_step": 58678, "epoch": 659, "lr": 8.90487818336992e-05}
{"train_loss": 0.1584746539592743, "global_step": 58679, "epoch": 659, "lr": 8.904841974991753e-05}
{"train_loss": 0.16762395203113556, "global_step": 58680, "epoch": 659, "lr": 8.904805766088627e-05}
{"train_loss": 0.2465621531009674, "global_step": 58681, "epoch": 659, "lr": 8.904769556660548e-05}
{"train_loss": 0.19498242437839508, "global_step": 58682, "epoch": 659, "lr": 8.90473334670752e-05}
{"train_loss": 0.2964290976524353, "global_step": 58683, "epoch": 659, "lr": 8.904697136229549e-05}
{"train_loss": 0.20223785936832428, "global_step": 58684, "epoch": 659, "lr": 8.90466092522664e-05}
{"train_loss": 0.25284260511398315, "global_step": 58685, "epoch": 659, "lr": 8.904624713698794e-05}
{"train_loss": 0.2436729520559311, "global_step": 58686, "epoch": 659, "lr": 8.904588501646024e-05}
{"train_loss": 0.22920922935009003, "global_step": 58687, "epoch": 659, "lr": 8.904552289068324e-05}
{"train_loss": 0.19270533323287964, "global_step": 58688, "epoch": 659, "lr": 8.904516075965708e-05}
{"train_loss": 0.19029909372329712, "global_step": 58689, "epoch": 659, "lr": 8.904479862338175e-05}
{"train_loss": 0.2100064903497696, "global_step": 58690, "epoch": 659, "lr": 8.904443648185733e-05}
{"train_loss": 0.23497599363327026, "global_step": 58691, "epoch": 659, "lr": 8.904407433508386e-05}
{"train_loss": 0.3263062834739685, "global_step": 58692, "epoch": 659, "lr": 8.90437121830614e-05}
{"train_loss": 0.311972975730896, "global_step": 58693, "epoch": 659, "lr": 8.904335002578999e-05}
{"train_loss": 0.2724786698818207, "global_step": 58694, "epoch": 659, "lr": 8.904298786326965e-05}
{"train_loss": 0.2182033807039261, "global_step": 58695, "epoch": 659, "lr": 8.904262569550049e-05}
{"train_loss": 0.15252406895160675, "global_step": 58696, "epoch": 659, "lr": 8.904226352248251e-05}
{"train_loss": 0.2453979253768921, "global_step": 58697, "epoch": 659, "lr": 8.904190134421578e-05}
{"train_loss": 0.29411277174949646, "global_step": 58698, "epoch": 659, "lr": 8.904153916070034e-05}
{"train_loss": 0.15963180363178253, "global_step": 58699, "epoch": 659, "lr": 8.904117697193624e-05}
{"train_loss": 0.21699276566505432, "global_step": 58700, "epoch": 659, "lr": 8.904081477792352e-05}
{"train_loss": 0.19493073225021362, "global_step": 58701, "epoch": 659, "lr": 8.904045257866225e-05}
{"train_loss": 0.27121177315711975, "global_step": 58702, "epoch": 659, "lr": 8.904009037415246e-05}
{"train_loss": 0.272176593542099, "global_step": 58703, "epoch": 659, "lr": 8.90397281643942e-05}
{"train_loss": 0.24369926750659943, "global_step": 58704, "epoch": 659, "lr": 8.903936594938756e-05}
{"train_loss": 0.19820952415466309, "global_step": 58705, "epoch": 659, "lr": 8.903900372913252e-05}
{"train_loss": 0.24129344522953033, "global_step": 58706, "epoch": 659, "lr": 8.903864150362917e-05}
{"train_loss": 0.21302053332328796, "global_step": 58707, "epoch": 659, "lr": 8.903827927287754e-05}
{"train_loss": 0.2068495750427246, "global_step": 58708, "epoch": 659, "lr": 8.903791703687769e-05}
{"train_loss": 0.2653908133506775, "global_step": 58709, "epoch": 659, "lr": 8.903755479562968e-05}
{"train_loss": 0.2547856867313385, "global_step": 58710, "epoch": 659, "lr": 8.903719254913355e-05}
{"train_loss": 0.19808824360370636, "global_step": 58711, "epoch": 659, "lr": 8.903683029738933e-05}
{"train_loss": 0.2321702539920807, "global_step": 58712, "epoch": 659, "lr": 8.90364680403971e-05}
{"train_loss": 0.30435049533843994, "global_step": 58713, "epoch": 659, "lr": 8.903610577815688e-05}
{"train_loss": 0.31930142641067505, "global_step": 58714, "epoch": 659, "lr": 8.903574351066874e-05}
{"train_loss": 0.19823512434959412, "global_step": 58715, "epoch": 659, "lr": 8.903538123793272e-05}
{"train_loss": 0.17536623775959015, "global_step": 58716, "epoch": 659, "lr": 8.903501895994886e-05}
{"train_loss": 0.24198772013187408, "global_step": 58717, "epoch": 659, "lr": 8.903465667671722e-05}
{"train_loss": 0.28832417726516724, "global_step": 58718, "epoch": 659, "lr": 8.903429438823786e-05}
{"train_loss": 0.15985117852687836, "global_step": 58719, "epoch": 659, "lr": 8.903393209451079e-05}
{"train_loss": 0.2601016163825989, "global_step": 58720, "epoch": 659, "lr": 8.90335697955361e-05}
{"train_loss": 0.18935540318489075, "global_step": 58721, "epoch": 659, "lr": 8.903320749131382e-05}
{"train_loss": 0.22458681464195251, "global_step": 58722, "epoch": 659, "lr": 8.903284518184401e-05}
{"train_loss": 0.2243598997592926, "global_step": 58723, "epoch": 659, "lr": 8.903248286712669e-05}
{"train_loss": 0.2399943470954895, "global_step": 58724, "epoch": 659, "lr": 8.903212054716194e-05}
{"train_loss": 0.21449126303195953, "global_step": 58725, "epoch": 659, "lr": 8.90317582219498e-05}
{"train_loss": 0.18541273474693298, "global_step": 58726, "epoch": 659, "lr": 8.90313958914903e-05}
{"train_loss": 0.22982919216156006, "global_step": 58727, "epoch": 659, "lr": 8.903103355578353e-05}
{"train_loss": 0.21087177097797394, "global_step": 58728, "epoch": 659, "lr": 8.903067121482949e-05}
{"train_loss": 0.3380969166755676, "global_step": 58729, "epoch": 659, "lr": 8.903030886862827e-05}
{"train_loss": 0.23174133896827698, "global_step": 58730, "epoch": 659, "lr": 8.902994651717989e-05}
{"train_loss": 0.324689656496048, "global_step": 58731, "epoch": 659, "lr": 8.902958416048441e-05}
{"train_loss": 0.15864156186580658, "global_step": 58732, "epoch": 659, "lr": 8.902922179854187e-05}
{"train_loss": 0.1835726946592331, "global_step": 58733, "epoch": 659, "lr": 8.902885943135234e-05}
{"train_loss": 0.24255652725696564, "global_step": 58734, "epoch": 659, "lr": 8.902849705891585e-05}
{"train_loss": 0.23493504524230957, "global_step": 58735, "epoch": 659, "lr": 8.902813468123245e-05}
{"train_loss": 0.3130723834037781, "global_step": 58736, "epoch": 659, "lr": 8.90277722983022e-05}
{"train_loss": 0.24843214452266693, "global_step": 58737, "epoch": 659, "lr": 8.902740991012514e-05}
{"train_loss": 0.2617526054382324, "global_step": 58738, "epoch": 659, "lr": 8.902704751670132e-05}
{"train_loss": 0.23910272255372467, "global_step": 58739, "epoch": 659, "lr": 8.90266851180308e-05, "val_loss": 3.562173366546631}
{"train_loss": 0.25617894530296326, "global_step": 58740, "epoch": 660, "lr": 8.90263227141136e-05}
{"train_loss": 0.2199173867702484, "global_step": 58741, "epoch": 660, "lr": 8.90259603049498e-05}
{"train_loss": 0.28328150510787964, "global_step": 58742, "epoch": 660, "lr": 8.902559789053943e-05}
{"train_loss": 0.30428582429885864, "global_step": 58743, "epoch": 660, "lr": 8.902523547088255e-05}
{"train_loss": 0.3008556365966797, "global_step": 58744, "epoch": 660, "lr": 8.90248730459792e-05}
{"train_loss": 0.1669853925704956, "global_step": 58745, "epoch": 660, "lr": 8.902451061582941e-05}
{"train_loss": 0.29477939009666443, "global_step": 58746, "epoch": 660, "lr": 8.902414818043328e-05}
{"train_loss": 0.13130809366703033, "global_step": 58747, "epoch": 660, "lr": 8.902378573979081e-05}
{"train_loss": 0.19185779988765717, "global_step": 58748, "epoch": 660, "lr": 8.902342329390208e-05}
{"train_loss": 0.35254785418510437, "global_step": 58749, "epoch": 660, "lr": 8.902306084276711e-05}
{"train_loss": 0.2880529761314392, "global_step": 58750, "epoch": 660, "lr": 8.902269838638598e-05}
{"train_loss": 0.225869283080101, "global_step": 58751, "epoch": 660, "lr": 8.902233592475873e-05}
{"train_loss": 0.26590627431869507, "global_step": 58752, "epoch": 660, "lr": 8.902197345788539e-05}
{"train_loss": 0.27429860830307007, "global_step": 58753, "epoch": 660, "lr": 8.902161098576602e-05}
{"train_loss": 0.2929912805557251, "global_step": 58754, "epoch": 660, "lr": 8.902124850840068e-05}
{"train_loss": 0.2537774443626404, "global_step": 58755, "epoch": 660, "lr": 8.90208860257894e-05}
{"train_loss": 0.26816123723983765, "global_step": 58756, "epoch": 660, "lr": 8.902052353793224e-05}
{"train_loss": 0.1253688782453537, "global_step": 58757, "epoch": 660, "lr": 8.902016104482924e-05}
{"train_loss": 0.1632167100906372, "global_step": 58758, "epoch": 660, "lr": 8.901979854648046e-05}
{"train_loss": 0.3488176167011261, "global_step": 58759, "epoch": 660, "lr": 8.901943604288594e-05}
{"train_loss": 0.2392381876707077, "global_step": 58760, "epoch": 660, "lr": 8.901907353404575e-05}
{"train_loss": 0.29480603337287903, "global_step": 58761, "epoch": 660, "lr": 8.901871101995992e-05}
{"train_loss": 0.29451602697372437, "global_step": 58762, "epoch": 660, "lr": 8.901834850062847e-05}
{"train_loss": 0.2442415952682495, "global_step": 58763, "epoch": 660, "lr": 8.90179859760515e-05}
{"train_loss": 0.22972208261489868, "global_step": 58764, "epoch": 660, "lr": 8.901762344622904e-05}
{"train_loss": 0.25898030400276184, "global_step": 58765, "epoch": 660, "lr": 8.901726091116112e-05}
{"train_loss": 0.28182855248451233, "global_step": 58766, "epoch": 660, "lr": 8.901689837084781e-05}
{"train_loss": 0.258286714553833, "global_step": 58767, "epoch": 660, "lr": 8.901653582528917e-05}
{"train_loss": 0.3177281320095062, "global_step": 58768, "epoch": 660, "lr": 8.901617327448522e-05}
{"train_loss": 0.21299614012241364, "global_step": 58769, "epoch": 660, "lr": 8.901581071843602e-05}
{"train_loss": 0.16389067471027374, "global_step": 58770, "epoch": 660, "lr": 8.901544815714162e-05}
{"train_loss": 0.2564736008644104, "global_step": 58771, "epoch": 660, "lr": 8.901508559060207e-05}
{"train_loss": 0.2784450352191925, "global_step": 58772, "epoch": 660, "lr": 8.901472301881741e-05}
{"train_loss": 0.25800636410713196, "global_step": 58773, "epoch": 660, "lr": 8.901436044178772e-05}
{"train_loss": 0.17437870800495148, "global_step": 58774, "epoch": 660, "lr": 8.9013997859513e-05}
{"train_loss": 0.26219093799591064, "global_step": 58775, "epoch": 660, "lr": 8.901363527199333e-05}
{"train_loss": 0.21430760622024536, "global_step": 58776, "epoch": 660, "lr": 8.901327267922876e-05}
{"train_loss": 0.25320786237716675, "global_step": 58777, "epoch": 660, "lr": 8.901291008121932e-05}
{"train_loss": 0.28553178906440735, "global_step": 58778, "epoch": 660, "lr": 8.901254747796508e-05}
{"train_loss": 0.22922717034816742, "global_step": 58779, "epoch": 660, "lr": 8.901218486946607e-05}
{"train_loss": 0.28082358837127686, "global_step": 58780, "epoch": 660, "lr": 8.901182225572235e-05}
{"train_loss": 0.2304251492023468, "global_step": 58781, "epoch": 660, "lr": 8.901145963673397e-05}
{"train_loss": 0.31942081451416016, "global_step": 58782, "epoch": 660, "lr": 8.901109701250097e-05}
{"train_loss": 0.23130249977111816, "global_step": 58783, "epoch": 660, "lr": 8.90107343830234e-05}
{"train_loss": 0.35009336471557617, "global_step": 58784, "epoch": 660, "lr": 8.901037174830132e-05}
{"train_loss": 0.290734201669693, "global_step": 58785, "epoch": 660, "lr": 8.901000910833476e-05}
{"train_loss": 0.2508609890937805, "global_step": 58786, "epoch": 660, "lr": 8.90096464631238e-05}
{"train_loss": 0.2472366988658905, "global_step": 58787, "epoch": 660, "lr": 8.900928381266845e-05}
{"train_loss": 0.3348378539085388, "global_step": 58788, "epoch": 660, "lr": 8.900892115696879e-05}
{"train_loss": 0.24327050149440765, "global_step": 58789, "epoch": 660, "lr": 8.900855849602486e-05}
{"train_loss": 0.287453830242157, "global_step": 58790, "epoch": 660, "lr": 8.900819582983668e-05}
{"train_loss": 0.17684897780418396, "global_step": 58791, "epoch": 660, "lr": 8.900783315840434e-05}
{"train_loss": 0.21246397495269775, "global_step": 58792, "epoch": 660, "lr": 8.900747048172787e-05}
{"train_loss": 0.21433153748512268, "global_step": 58793, "epoch": 660, "lr": 8.900710779980733e-05}
{"train_loss": 0.21988017857074738, "global_step": 58794, "epoch": 660, "lr": 8.900674511264274e-05}
{"train_loss": 0.24821636080741882, "global_step": 58795, "epoch": 660, "lr": 8.900638242023419e-05}
{"train_loss": 0.2409093976020813, "global_step": 58796, "epoch": 660, "lr": 8.900601972258169e-05}
{"train_loss": 0.24641701579093933, "global_step": 58797, "epoch": 660, "lr": 8.900565701968531e-05}
{"train_loss": 0.2911350429058075, "global_step": 58798, "epoch": 660, "lr": 8.90052943115451e-05}
{"train_loss": 0.2126304656267166, "global_step": 58799, "epoch": 660, "lr": 8.90049315981611e-05}
{"train_loss": 0.22754700481891632, "global_step": 58800, "epoch": 660, "lr": 8.900456887953337e-05}
{"train_loss": 0.21225108206272125, "global_step": 58801, "epoch": 660, "lr": 8.900420615566194e-05}
{"train_loss": 0.2363722324371338, "global_step": 58802, "epoch": 660, "lr": 8.900384342654687e-05}
{"train_loss": 0.2599400281906128, "global_step": 58803, "epoch": 660, "lr": 8.900348069218822e-05}
{"train_loss": 0.29977700114250183, "global_step": 58804, "epoch": 660, "lr": 8.900311795258603e-05}
{"train_loss": 0.22040152549743652, "global_step": 58805, "epoch": 660, "lr": 8.900275520774032e-05}
{"train_loss": 0.28750479221343994, "global_step": 58806, "epoch": 660, "lr": 8.900239245765119e-05}
{"train_loss": 0.28823864459991455, "global_step": 58807, "epoch": 660, "lr": 8.900202970231865e-05}
{"train_loss": 0.18044814467430115, "global_step": 58808, "epoch": 660, "lr": 8.900166694174277e-05}
{"train_loss": 0.19523760676383972, "global_step": 58809, "epoch": 660, "lr": 8.900130417592358e-05}
{"train_loss": 0.24542230367660522, "global_step": 58810, "epoch": 660, "lr": 8.900094140486116e-05}
{"train_loss": 0.26354920864105225, "global_step": 58811, "epoch": 660, "lr": 8.900057862855553e-05}
{"train_loss": 0.3394722044467926, "global_step": 58812, "epoch": 660, "lr": 8.900021584700674e-05}
{"train_loss": 0.16275697946548462, "global_step": 58813, "epoch": 660, "lr": 8.899985306021485e-05}
{"train_loss": 0.3681706190109253, "global_step": 58814, "epoch": 660, "lr": 8.899949026817991e-05}
{"train_loss": 0.24469050765037537, "global_step": 58815, "epoch": 660, "lr": 8.899912747090195e-05}
{"train_loss": 0.27293381094932556, "global_step": 58816, "epoch": 660, "lr": 8.899876466838105e-05}
{"train_loss": 0.2832863926887512, "global_step": 58817, "epoch": 660, "lr": 8.899840186061723e-05}
{"train_loss": 0.2395717054605484, "global_step": 58818, "epoch": 660, "lr": 8.899803904761057e-05}
{"train_loss": 0.18172885477542877, "global_step": 58819, "epoch": 660, "lr": 8.899767622936108e-05}
{"train_loss": 0.314283549785614, "global_step": 58820, "epoch": 660, "lr": 8.899731340586883e-05}
{"train_loss": 0.28861552476882935, "global_step": 58821, "epoch": 660, "lr": 8.899695057713387e-05}
{"train_loss": 0.2789282500743866, "global_step": 58822, "epoch": 660, "lr": 8.899658774315624e-05}
{"train_loss": 0.3576766848564148, "global_step": 58823, "epoch": 660, "lr": 8.8996224903936e-05}
{"train_loss": 0.25819867849349976, "global_step": 58824, "epoch": 660, "lr": 8.899586205947319e-05}
{"train_loss": 0.22609063982963562, "global_step": 58825, "epoch": 660, "lr": 8.899549920976787e-05}
{"train_loss": 0.22164711356163025, "global_step": 58826, "epoch": 660, "lr": 8.899513635482006e-05}
{"train_loss": 0.19165991246700287, "global_step": 58827, "epoch": 660, "lr": 8.899477349462985e-05}
{"train_loss": 0.25243097058172975, "global_step": 58828, "epoch": 660, "lr": 8.899441062919725e-05, "val_loss": 3.3754379749298096, "train_action_mse_error": 46.25078582763672}
{"train_loss": 0.2904350161552429, "global_step": 58829, "epoch": 661, "lr": 8.899404775852234e-05}
{"train_loss": 0.30798035860061646, "global_step": 58830, "epoch": 661, "lr": 8.899368488260516e-05}
{"train_loss": 0.33002883195877075, "global_step": 58831, "epoch": 661, "lr": 8.899332200144575e-05}
{"train_loss": 0.2755802273750305, "global_step": 58832, "epoch": 661, "lr": 8.899295911504414e-05}
{"train_loss": 0.22280825674533844, "global_step": 58833, "epoch": 661, "lr": 8.899259622340044e-05}
{"train_loss": 0.30143752694129944, "global_step": 58834, "epoch": 661, "lr": 8.899223332651464e-05}
{"train_loss": 0.22069385647773743, "global_step": 58835, "epoch": 661, "lr": 8.89918704243868e-05}
{"train_loss": 0.347763329744339, "global_step": 58836, "epoch": 661, "lr": 8.8991507517017e-05}
{"train_loss": 0.22887861728668213, "global_step": 58837, "epoch": 661, "lr": 8.899114460440525e-05}
{"train_loss": 0.2852731943130493, "global_step": 58838, "epoch": 661, "lr": 8.899078168655163e-05}
{"train_loss": 0.32262560725212097, "global_step": 58839, "epoch": 661, "lr": 8.899041876345616e-05}
{"train_loss": 0.29078295826911926, "global_step": 58840, "epoch": 661, "lr": 8.899005583511892e-05}
{"train_loss": 0.32264775037765503, "global_step": 58841, "epoch": 661, "lr": 8.898969290153994e-05}
{"train_loss": 0.22350579500198364, "global_step": 58842, "epoch": 661, "lr": 8.898932996271924e-05}
{"train_loss": 0.2487313449382782, "global_step": 58843, "epoch": 661, "lr": 8.898896701865693e-05}
{"train_loss": 0.3156167268753052, "global_step": 58844, "epoch": 661, "lr": 8.898860406935303e-05}
{"train_loss": 0.24895144999027252, "global_step": 58845, "epoch": 661, "lr": 8.898824111480757e-05}
{"train_loss": 0.2903904914855957, "global_step": 58846, "epoch": 661, "lr": 8.898787815502063e-05}
{"train_loss": 0.23096130788326263, "global_step": 58847, "epoch": 661, "lr": 8.898751518999224e-05}
{"train_loss": 0.23511043190956116, "global_step": 58848, "epoch": 661, "lr": 8.898715221972246e-05}
{"train_loss": 0.2588614225387573, "global_step": 58849, "epoch": 661, "lr": 8.898678924421133e-05}
{"train_loss": 0.25876161456108093, "global_step": 58850, "epoch": 661, "lr": 8.89864262634589e-05}
{"train_loss": 0.2296912968158722, "global_step": 58851, "epoch": 661, "lr": 8.898606327746521e-05}
{"train_loss": 0.1845739483833313, "global_step": 58852, "epoch": 661, "lr": 8.898570028623032e-05}
{"train_loss": 0.26025938987731934, "global_step": 58853, "epoch": 661, "lr": 8.898533728975429e-05}
{"train_loss": 0.29595839977264404, "global_step": 58854, "epoch": 661, "lr": 8.898497428803716e-05}
{"train_loss": 0.2347736358642578, "global_step": 58855, "epoch": 661, "lr": 8.898461128107895e-05}
{"train_loss": 0.22783246636390686, "global_step": 58856, "epoch": 661, "lr": 8.898424826887977e-05}
{"train_loss": 0.17781300842761993, "global_step": 58857, "epoch": 661, "lr": 8.898388525143962e-05}
{"train_loss": 0.21023236215114594, "global_step": 58858, "epoch": 661, "lr": 8.898352222875854e-05}
{"train_loss": 0.24480068683624268, "global_step": 58859, "epoch": 661, "lr": 8.898315920083663e-05}
{"train_loss": 0.27022379636764526, "global_step": 58860, "epoch": 661, "lr": 8.898279616767389e-05}
{"train_loss": 0.2473592460155487, "global_step": 58861, "epoch": 661, "lr": 8.898243312927039e-05}
{"train_loss": 0.18952617049217224, "global_step": 58862, "epoch": 661, "lr": 8.898207008562618e-05}
{"train_loss": 0.23793569207191467, "global_step": 58863, "epoch": 661, "lr": 8.898170703674131e-05}
{"train_loss": 0.2817195951938629, "global_step": 58864, "epoch": 661, "lr": 8.898134398261582e-05}
{"train_loss": 0.26950860023498535, "global_step": 58865, "epoch": 661, "lr": 8.898098092324978e-05}
{"train_loss": 0.23445898294448853, "global_step": 58866, "epoch": 661, "lr": 8.89806178586432e-05}
{"train_loss": 0.2451035976409912, "global_step": 58867, "epoch": 661, "lr": 8.898025478879617e-05}
{"train_loss": 0.277786523103714, "global_step": 58868, "epoch": 661, "lr": 8.897989171370871e-05}
{"train_loss": 0.25682854652404785, "global_step": 58869, "epoch": 661, "lr": 8.897952863338086e-05}
{"train_loss": 0.2464447170495987, "global_step": 58870, "epoch": 661, "lr": 8.897916554781271e-05}
{"train_loss": 0.17152120172977448, "global_step": 58871, "epoch": 661, "lr": 8.897880245700427e-05}
{"train_loss": 0.23359833657741547, "global_step": 58872, "epoch": 661, "lr": 8.897843936095562e-05}
{"train_loss": 0.25314977765083313, "global_step": 58873, "epoch": 661, "lr": 8.897807625966678e-05}
{"train_loss": 0.3294641971588135, "global_step": 58874, "epoch": 661, "lr": 8.897771315313783e-05}
{"train_loss": 0.18053849041461945, "global_step": 58875, "epoch": 661, "lr": 8.89773500413688e-05}
{"train_loss": 0.18201035261154175, "global_step": 58876, "epoch": 661, "lr": 8.897698692435974e-05}
{"train_loss": 0.3453168570995331, "global_step": 58877, "epoch": 661, "lr": 8.89766238021107e-05}
{"train_loss": 0.1748725026845932, "global_step": 58878, "epoch": 661, "lr": 8.897626067462171e-05}
{"train_loss": 0.20945827662944794, "global_step": 58879, "epoch": 661, "lr": 8.897589754189287e-05}
{"train_loss": 0.3747895658016205, "global_step": 58880, "epoch": 661, "lr": 8.897553440392416e-05}
{"train_loss": 0.2397191971540451, "global_step": 58881, "epoch": 661, "lr": 8.897517126071567e-05}
{"train_loss": 0.2446432262659073, "global_step": 58882, "epoch": 661, "lr": 8.897480811226746e-05}
{"train_loss": 0.23739440739154816, "global_step": 58883, "epoch": 661, "lr": 8.897444495857955e-05}
{"train_loss": 0.357957124710083, "global_step": 58884, "epoch": 661, "lr": 8.8974081799652e-05}
{"train_loss": 0.21814750134944916, "global_step": 58885, "epoch": 661, "lr": 8.897371863548488e-05}
{"train_loss": 0.27625447511672974, "global_step": 58886, "epoch": 661, "lr": 8.897335546607819e-05}
{"train_loss": 0.2646697461605072, "global_step": 58887, "epoch": 661, "lr": 8.897299229143203e-05}
{"train_loss": 0.28736549615859985, "global_step": 58888, "epoch": 661, "lr": 8.897262911154641e-05}
{"train_loss": 0.3386548161506653, "global_step": 58889, "epoch": 661, "lr": 8.89722659264214e-05}
{"train_loss": 0.241301029920578, "global_step": 58890, "epoch": 661, "lr": 8.897190273605705e-05}
{"train_loss": 0.2898181974887848, "global_step": 58891, "epoch": 661, "lr": 8.89715395404534e-05}
{"train_loss": 0.2885744571685791, "global_step": 58892, "epoch": 661, "lr": 8.89711763396105e-05}
{"train_loss": 0.17875748872756958, "global_step": 58893, "epoch": 661, "lr": 8.89708131335284e-05}
{"train_loss": 0.24407494068145752, "global_step": 58894, "epoch": 661, "lr": 8.897044992220714e-05}
{"train_loss": 0.24422016739845276, "global_step": 58895, "epoch": 661, "lr": 8.89700867056468e-05}
{"train_loss": 0.272794634103775, "global_step": 58896, "epoch": 661, "lr": 8.896972348384738e-05}
{"train_loss": 0.2296525239944458, "global_step": 58897, "epoch": 661, "lr": 8.896936025680897e-05}
{"train_loss": 0.24717332422733307, "global_step": 58898, "epoch": 661, "lr": 8.896899702453161e-05}
{"train_loss": 0.279599130153656, "global_step": 58899, "epoch": 661, "lr": 8.896863378701535e-05}
{"train_loss": 0.21828250586986542, "global_step": 58900, "epoch": 661, "lr": 8.89682705442602e-05}
{"train_loss": 0.21621066331863403, "global_step": 58901, "epoch": 661, "lr": 8.896790729626628e-05}
{"train_loss": 0.19037844240665436, "global_step": 58902, "epoch": 661, "lr": 8.896754404303357e-05}
{"train_loss": 0.24766919016838074, "global_step": 58903, "epoch": 661, "lr": 8.896718078456217e-05}
{"train_loss": 0.21682056784629822, "global_step": 58904, "epoch": 661, "lr": 8.896681752085208e-05}
{"train_loss": 0.2326066493988037, "global_step": 58905, "epoch": 661, "lr": 8.89664542519034e-05}
{"train_loss": 0.19869202375411987, "global_step": 58906, "epoch": 661, "lr": 8.896609097771615e-05}
{"train_loss": 0.26657864451408386, "global_step": 58907, "epoch": 661, "lr": 8.896572769829038e-05}
{"train_loss": 0.243011012673378, "global_step": 58908, "epoch": 661, "lr": 8.896536441362615e-05}
{"train_loss": 0.2588037848472595, "global_step": 58909, "epoch": 661, "lr": 8.89650011237235e-05}
{"train_loss": 0.2579067349433899, "global_step": 58910, "epoch": 661, "lr": 8.896463782858247e-05}
{"train_loss": 0.23741444945335388, "global_step": 58911, "epoch": 661, "lr": 8.896427452820313e-05}
{"train_loss": 0.19911067187786102, "global_step": 58912, "epoch": 661, "lr": 8.89639112225855e-05}
{"train_loss": 0.3589043915271759, "global_step": 58913, "epoch": 661, "lr": 8.896354791172967e-05}
{"train_loss": 0.279285192489624, "global_step": 58914, "epoch": 661, "lr": 8.896318459563566e-05}
{"train_loss": 0.19982898235321045, "global_step": 58915, "epoch": 661, "lr": 8.896282127430352e-05}
{"train_loss": 0.2633967697620392, "global_step": 58916, "epoch": 661, "lr": 8.89624579477333e-05}
{"train_loss": 0.2534721399960893, "global_step": 58917, "epoch": 661, "lr": 8.896209461592506e-05, "val_loss": 3.3065171241760254}
{"train_loss": 0.2936827838420868, "global_step": 58918, "epoch": 662, "lr": 8.896173127887883e-05}
{"train_loss": 0.2776135504245758, "global_step": 58919, "epoch": 662, "lr": 8.896136793659469e-05}
{"train_loss": 0.16242747008800507, "global_step": 58920, "epoch": 662, "lr": 8.896100458907266e-05}
{"train_loss": 0.2044171690940857, "global_step": 58921, "epoch": 662, "lr": 8.89606412363128e-05}
{"train_loss": 0.22841274738311768, "global_step": 58922, "epoch": 662, "lr": 8.896027787831515e-05}
{"train_loss": 0.28349751234054565, "global_step": 58923, "epoch": 662, "lr": 8.895991451507977e-05}
{"train_loss": 0.20099955797195435, "global_step": 58924, "epoch": 662, "lr": 8.89595511466067e-05}
{"train_loss": 0.21913516521453857, "global_step": 58925, "epoch": 662, "lr": 8.8959187772896e-05}
{"train_loss": 0.3147954046726227, "global_step": 58926, "epoch": 662, "lr": 8.89588243939477e-05}
{"train_loss": 0.2173699289560318, "global_step": 58927, "epoch": 662, "lr": 8.895846100976187e-05}
{"train_loss": 0.2313816249370575, "global_step": 58928, "epoch": 662, "lr": 8.895809762033855e-05}
{"train_loss": 0.2742677927017212, "global_step": 58929, "epoch": 662, "lr": 8.89577342256778e-05}
{"train_loss": 0.1650725156068802, "global_step": 58930, "epoch": 662, "lr": 8.895737082577964e-05}
{"train_loss": 0.19923681020736694, "global_step": 58931, "epoch": 662, "lr": 8.895700742064415e-05}
{"train_loss": 0.21792921423912048, "global_step": 58932, "epoch": 662, "lr": 8.895664401027135e-05}
{"train_loss": 0.19950494170188904, "global_step": 58933, "epoch": 662, "lr": 8.89562805946613e-05}
{"train_loss": 0.13587845861911774, "global_step": 58934, "epoch": 662, "lr": 8.895591717381406e-05}
{"train_loss": 0.13465622067451477, "global_step": 58935, "epoch": 662, "lr": 8.895555374772968e-05}
{"train_loss": 0.24169431626796722, "global_step": 58936, "epoch": 662, "lr": 8.89551903164082e-05}
{"train_loss": 0.24662788212299347, "global_step": 58937, "epoch": 662, "lr": 8.895482687984967e-05}
{"train_loss": 0.25186145305633545, "global_step": 58938, "epoch": 662, "lr": 8.895446343805413e-05}
{"train_loss": 0.2189594805240631, "global_step": 58939, "epoch": 662, "lr": 8.895409999102165e-05}
{"train_loss": 0.14808598160743713, "global_step": 58940, "epoch": 662, "lr": 8.895373653875226e-05}
{"train_loss": 0.23502635955810547, "global_step": 58941, "epoch": 662, "lr": 8.895337308124599e-05}
{"train_loss": 0.1920776218175888, "global_step": 58942, "epoch": 662, "lr": 8.895300961850295e-05}
{"train_loss": 0.21112699806690216, "global_step": 58943, "epoch": 662, "lr": 8.895264615052313e-05}
{"train_loss": 0.2918933928012848, "global_step": 58944, "epoch": 662, "lr": 8.895228267730661e-05}
{"train_loss": 0.24629086256027222, "global_step": 58945, "epoch": 662, "lr": 8.895191919885342e-05}
{"train_loss": 0.23697243630886078, "global_step": 58946, "epoch": 662, "lr": 8.895155571516363e-05}
{"train_loss": 0.30332428216934204, "global_step": 58947, "epoch": 662, "lr": 8.895119222623728e-05}
{"train_loss": 0.18273691833019257, "global_step": 58948, "epoch": 662, "lr": 8.895082873207443e-05}
{"train_loss": 0.22248150408267975, "global_step": 58949, "epoch": 662, "lr": 8.895046523267509e-05}
{"train_loss": 0.23147660493850708, "global_step": 58950, "epoch": 662, "lr": 8.895010172803934e-05}
{"train_loss": 0.28823384642601013, "global_step": 58951, "epoch": 662, "lr": 8.894973821816724e-05}
{"train_loss": 0.22015050053596497, "global_step": 58952, "epoch": 662, "lr": 8.89493747030588e-05}
{"train_loss": 0.2402108907699585, "global_step": 58953, "epoch": 662, "lr": 8.89490111827141e-05}
{"train_loss": 0.17777509987354279, "global_step": 58954, "epoch": 662, "lr": 8.89486476571332e-05}
{"train_loss": 0.21390168368816376, "global_step": 58955, "epoch": 662, "lr": 8.894828412631609e-05}
{"train_loss": 0.11481138318777084, "global_step": 58956, "epoch": 662, "lr": 8.894792059026289e-05}
{"train_loss": 0.2777557075023651, "global_step": 58957, "epoch": 662, "lr": 8.89475570489736e-05}
{"train_loss": 0.2220550775527954, "global_step": 58958, "epoch": 662, "lr": 8.894719350244829e-05}
{"train_loss": 0.24363723397254944, "global_step": 58959, "epoch": 662, "lr": 8.894682995068702e-05}
{"train_loss": 0.1790497750043869, "global_step": 58960, "epoch": 662, "lr": 8.89464663936898e-05}
{"train_loss": 0.3894338011741638, "global_step": 58961, "epoch": 662, "lr": 8.894610283145671e-05}
{"train_loss": 0.33605054020881653, "global_step": 58962, "epoch": 662, "lr": 8.894573926398779e-05}
{"train_loss": 0.24154990911483765, "global_step": 58963, "epoch": 662, "lr": 8.894537569128309e-05}
{"train_loss": 0.2159448266029358, "global_step": 58964, "epoch": 662, "lr": 8.894501211334267e-05}
{"train_loss": 0.2091163545846939, "global_step": 58965, "epoch": 662, "lr": 8.894464853016655e-05}
{"train_loss": 0.2835138738155365, "global_step": 58966, "epoch": 662, "lr": 8.894428494175482e-05}
{"train_loss": 0.23787832260131836, "global_step": 58967, "epoch": 662, "lr": 8.894392134810749e-05}
{"train_loss": 0.35627084970474243, "global_step": 58968, "epoch": 662, "lr": 8.894355774922463e-05}
{"train_loss": 0.1469632238149643, "global_step": 58969, "epoch": 662, "lr": 8.894319414510627e-05}
{"train_loss": 0.1628030240535736, "global_step": 58970, "epoch": 662, "lr": 8.894283053575248e-05}
{"train_loss": 0.1816820502281189, "global_step": 58971, "epoch": 662, "lr": 8.89424669211633e-05}
{"train_loss": 0.14481407403945923, "global_step": 58972, "epoch": 662, "lr": 8.894210330133879e-05}
{"train_loss": 0.2851240038871765, "global_step": 58973, "epoch": 662, "lr": 8.894173967627898e-05}
{"train_loss": 0.2593110501766205, "global_step": 58974, "epoch": 662, "lr": 8.894137604598393e-05}
{"train_loss": 0.3010033965110779, "global_step": 58975, "epoch": 662, "lr": 8.894101241045369e-05}
{"train_loss": 0.18071527779102325, "global_step": 58976, "epoch": 662, "lr": 8.89406487696883e-05}
{"train_loss": 0.20492996275424957, "global_step": 58977, "epoch": 662, "lr": 8.894028512368782e-05}
{"train_loss": 0.23319178819656372, "global_step": 58978, "epoch": 662, "lr": 8.893992147245228e-05}
{"train_loss": 0.24383725225925446, "global_step": 58979, "epoch": 662, "lr": 8.893955781598176e-05}
{"train_loss": 0.20425212383270264, "global_step": 58980, "epoch": 662, "lr": 8.893919415427627e-05}
{"train_loss": 0.2747800648212433, "global_step": 58981, "epoch": 662, "lr": 8.89388304873359e-05}
{"train_loss": 0.2796219289302826, "global_step": 58982, "epoch": 662, "lr": 8.893846681516067e-05}
{"train_loss": 0.2615916132926941, "global_step": 58983, "epoch": 662, "lr": 8.893810313775064e-05}
{"train_loss": 0.13862675428390503, "global_step": 58984, "epoch": 662, "lr": 8.893773945510586e-05}
{"train_loss": 0.17662574350833893, "global_step": 58985, "epoch": 662, "lr": 8.893737576722635e-05}
{"train_loss": 0.31074097752571106, "global_step": 58986, "epoch": 662, "lr": 8.893701207411222e-05}
{"train_loss": 0.2795785069465637, "global_step": 58987, "epoch": 662, "lr": 8.893664837576347e-05}
{"train_loss": 0.23795519769191742, "global_step": 58988, "epoch": 662, "lr": 8.893628467218016e-05}
{"train_loss": 0.19545267522335052, "global_step": 58989, "epoch": 662, "lr": 8.893592096336234e-05}
{"train_loss": 0.23233675956726074, "global_step": 58990, "epoch": 662, "lr": 8.893555724931007e-05}
{"train_loss": 0.24741233885288239, "global_step": 58991, "epoch": 662, "lr": 8.893519353002337e-05}
{"train_loss": 0.26641982793807983, "global_step": 58992, "epoch": 662, "lr": 8.893482980550232e-05}
{"train_loss": 0.2567121684551239, "global_step": 58993, "epoch": 662, "lr": 8.893446607574695e-05}
{"train_loss": 0.2401132732629776, "global_step": 58994, "epoch": 662, "lr": 8.893410234075733e-05}
{"train_loss": 0.24143120646476746, "global_step": 58995, "epoch": 662, "lr": 8.893373860053349e-05}
{"train_loss": 0.3688521385192871, "global_step": 58996, "epoch": 662, "lr": 8.893337485507547e-05}
{"train_loss": 0.24306651949882507, "global_step": 58997, "epoch": 662, "lr": 8.893301110438335e-05}
{"train_loss": 0.26814690232276917, "global_step": 58998, "epoch": 662, "lr": 8.893264734845714e-05}
{"train_loss": 0.2624949812889099, "global_step": 58999, "epoch": 662, "lr": 8.893228358729692e-05}
{"train_loss": 0.29303300380706787, "global_step": 59000, "epoch": 662, "lr": 8.893191982090275e-05}
{"train_loss": 0.21974270045757294, "global_step": 59001, "epoch": 662, "lr": 8.893155604927464e-05}
{"train_loss": 0.21749034523963928, "global_step": 59002, "epoch": 662, "lr": 8.893119227241264e-05}
{"train_loss": 0.13812074065208435, "global_step": 59003, "epoch": 662, "lr": 8.893082849031684e-05}
{"train_loss": 0.2693431079387665, "global_step": 59004, "epoch": 662, "lr": 8.893046470298725e-05}
{"train_loss": 0.2932758033275604, "global_step": 59005, "epoch": 662, "lr": 8.893010091042396e-05}
{"train_loss": 0.23477196735277606, "global_step": 59006, "epoch": 662, "lr": 8.892973711262696e-05, "val_loss": 3.379756212234497}
{"train_loss": 0.34061866998672485, "global_step": 59007, "epoch": 663, "lr": 8.892937330959634e-05}
{"train_loss": 0.32273438572883606, "global_step": 59008, "epoch": 663, "lr": 8.892900950133214e-05}
{"train_loss": 0.31364184617996216, "global_step": 59009, "epoch": 663, "lr": 8.892864568783442e-05}
{"train_loss": 0.302023708820343, "global_step": 59010, "epoch": 663, "lr": 8.89282818691032e-05}
{"train_loss": 0.16180065274238586, "global_step": 59011, "epoch": 663, "lr": 8.892791804513856e-05}
{"train_loss": 0.2977535128593445, "global_step": 59012, "epoch": 663, "lr": 8.892755421594054e-05}
{"train_loss": 0.24914202094078064, "global_step": 59013, "epoch": 663, "lr": 8.892719038150918e-05}
{"train_loss": 0.24365319311618805, "global_step": 59014, "epoch": 663, "lr": 8.892682654184454e-05}
{"train_loss": 0.23114949464797974, "global_step": 59015, "epoch": 663, "lr": 8.892646269694664e-05}
{"train_loss": 0.16598066687583923, "global_step": 59016, "epoch": 663, "lr": 8.892609884681557e-05}
{"train_loss": 0.26823168992996216, "global_step": 59017, "epoch": 663, "lr": 8.892573499145136e-05}
{"train_loss": 0.13387958705425262, "global_step": 59018, "epoch": 663, "lr": 8.892537113085407e-05}
{"train_loss": 0.1844148486852646, "global_step": 59019, "epoch": 663, "lr": 8.892500726502371e-05}
{"train_loss": 0.2133161574602127, "global_step": 59020, "epoch": 663, "lr": 8.892464339396039e-05}
{"train_loss": 0.3482801914215088, "global_step": 59021, "epoch": 663, "lr": 8.892427951766409e-05}
{"train_loss": 0.15978369116783142, "global_step": 59022, "epoch": 663, "lr": 8.892391563613491e-05}
{"train_loss": 0.2875973582267761, "global_step": 59023, "epoch": 663, "lr": 8.892355174937289e-05}
{"train_loss": 0.1709970235824585, "global_step": 59024, "epoch": 663, "lr": 8.892318785737808e-05}
{"train_loss": 0.17906081676483154, "global_step": 59025, "epoch": 663, "lr": 8.89228239601505e-05}
{"train_loss": 0.24817466735839844, "global_step": 59026, "epoch": 663, "lr": 8.892246005769023e-05}
{"train_loss": 0.16250170767307281, "global_step": 59027, "epoch": 663, "lr": 8.892209614999732e-05}
{"train_loss": 0.1954791396856308, "global_step": 59028, "epoch": 663, "lr": 8.89217322370718e-05}
{"train_loss": 0.17045488953590393, "global_step": 59029, "epoch": 663, "lr": 8.892136831891374e-05}
{"train_loss": 0.3109273314476013, "global_step": 59030, "epoch": 663, "lr": 8.892100439552316e-05}
{"train_loss": 0.22329771518707275, "global_step": 59031, "epoch": 663, "lr": 8.892064046690012e-05}
{"train_loss": 0.22015969455242157, "global_step": 59032, "epoch": 663, "lr": 8.892027653304468e-05}
{"train_loss": 0.2990947365760803, "global_step": 59033, "epoch": 663, "lr": 8.891991259395689e-05}
{"train_loss": 0.20400531589984894, "global_step": 59034, "epoch": 663, "lr": 8.891954864963678e-05}
{"train_loss": 0.3342893123626709, "global_step": 59035, "epoch": 663, "lr": 8.891918470008442e-05}
{"train_loss": 0.2300013303756714, "global_step": 59036, "epoch": 663, "lr": 8.891882074529987e-05}
{"train_loss": 0.20994341373443604, "global_step": 59037, "epoch": 663, "lr": 8.891845678528313e-05}
{"train_loss": 0.3255896270275116, "global_step": 59038, "epoch": 663, "lr": 8.89180928200343e-05}
{"train_loss": 0.2880935072898865, "global_step": 59039, "epoch": 663, "lr": 8.891772884955339e-05}
{"train_loss": 0.25682374835014343, "global_step": 59040, "epoch": 663, "lr": 8.891736487384045e-05}
{"train_loss": 0.28449445962905884, "global_step": 59041, "epoch": 663, "lr": 8.891700089289558e-05}
{"train_loss": 0.2532269060611725, "global_step": 59042, "epoch": 663, "lr": 8.891663690671877e-05}
{"train_loss": 0.18883953988552094, "global_step": 59043, "epoch": 663, "lr": 8.89162729153101e-05}
{"train_loss": 0.2218397855758667, "global_step": 59044, "epoch": 663, "lr": 8.891590891866962e-05}
{"train_loss": 0.20763321220874786, "global_step": 59045, "epoch": 663, "lr": 8.891554491679737e-05}
{"train_loss": 0.22085368633270264, "global_step": 59046, "epoch": 663, "lr": 8.891518090969338e-05}
{"train_loss": 0.2972784638404846, "global_step": 59047, "epoch": 663, "lr": 8.891481689735774e-05}
{"train_loss": 0.27681609988212585, "global_step": 59048, "epoch": 663, "lr": 8.891445287979045e-05}
{"train_loss": 0.28507310152053833, "global_step": 59049, "epoch": 663, "lr": 8.891408885699161e-05}
{"train_loss": 0.2816751301288605, "global_step": 59050, "epoch": 663, "lr": 8.891372482896124e-05}
{"train_loss": 0.27096524834632874, "global_step": 59051, "epoch": 663, "lr": 8.891336079569938e-05}
{"train_loss": 0.239024356007576, "global_step": 59052, "epoch": 663, "lr": 8.891299675720612e-05}
{"train_loss": 0.20952536165714264, "global_step": 59053, "epoch": 663, "lr": 8.891263271348147e-05}
{"train_loss": 0.2682521343231201, "global_step": 59054, "epoch": 663, "lr": 8.891226866452547e-05}
{"train_loss": 0.2945067882537842, "global_step": 59055, "epoch": 663, "lr": 8.891190461033822e-05}
{"train_loss": 0.2642287015914917, "global_step": 59056, "epoch": 663, "lr": 8.891154055091973e-05}
{"train_loss": 0.24031344056129456, "global_step": 59057, "epoch": 663, "lr": 8.891117648627006e-05}
{"train_loss": 0.2007657140493393, "global_step": 59058, "epoch": 663, "lr": 8.891081241638924e-05}
{"train_loss": 0.3050413727760315, "global_step": 59059, "epoch": 663, "lr": 8.891044834127735e-05}
{"train_loss": 0.20083403587341309, "global_step": 59060, "epoch": 663, "lr": 8.891008426093441e-05}
{"train_loss": 0.23472225666046143, "global_step": 59061, "epoch": 663, "lr": 8.89097201753605e-05}
{"train_loss": 0.2420574575662613, "global_step": 59062, "epoch": 663, "lr": 8.890935608455564e-05}
{"train_loss": 0.36591869592666626, "global_step": 59063, "epoch": 663, "lr": 8.89089919885199e-05}
{"train_loss": 0.2722805440425873, "global_step": 59064, "epoch": 663, "lr": 8.890862788725331e-05}
{"train_loss": 0.3010258376598358, "global_step": 59065, "epoch": 663, "lr": 8.890826378075594e-05}
{"train_loss": 0.1891578882932663, "global_step": 59066, "epoch": 663, "lr": 8.890789966902783e-05}
{"train_loss": 0.2364371120929718, "global_step": 59067, "epoch": 663, "lr": 8.890753555206904e-05}
{"train_loss": 0.210429385304451, "global_step": 59068, "epoch": 663, "lr": 8.890717142987956e-05}
{"train_loss": 0.33237317204475403, "global_step": 59069, "epoch": 663, "lr": 8.890680730245952e-05}
{"train_loss": 0.23145803809165955, "global_step": 59070, "epoch": 663, "lr": 8.890644316980893e-05}
{"train_loss": 0.24470914900302887, "global_step": 59071, "epoch": 663, "lr": 8.890607903192784e-05}
{"train_loss": 0.17608636617660522, "global_step": 59072, "epoch": 663, "lr": 8.89057148888163e-05}
{"train_loss": 0.28700196743011475, "global_step": 59073, "epoch": 663, "lr": 8.890535074047437e-05}
{"train_loss": 0.26271119713783264, "global_step": 59074, "epoch": 663, "lr": 8.890498658690207e-05}
{"train_loss": 0.25458842515945435, "global_step": 59075, "epoch": 663, "lr": 8.890462242809949e-05}
{"train_loss": 0.23287251591682434, "global_step": 59076, "epoch": 663, "lr": 8.890425826406666e-05}
{"train_loss": 0.35024577379226685, "global_step": 59077, "epoch": 663, "lr": 8.89038940948036e-05}
{"train_loss": 0.23673133552074432, "global_step": 59078, "epoch": 663, "lr": 8.890352992031042e-05}
{"train_loss": 0.2516813576221466, "global_step": 59079, "epoch": 663, "lr": 8.89031657405871e-05}
{"train_loss": 0.2482851892709732, "global_step": 59080, "epoch": 663, "lr": 8.890280155563374e-05}
{"train_loss": 0.22314821183681488, "global_step": 59081, "epoch": 663, "lr": 8.890243736545037e-05}
{"train_loss": 0.26345348358154297, "global_step": 59082, "epoch": 663, "lr": 8.890207317003704e-05}
{"train_loss": 0.19389553368091583, "global_step": 59083, "epoch": 663, "lr": 8.890170896939381e-05}
{"train_loss": 0.21901538968086243, "global_step": 59084, "epoch": 663, "lr": 8.89013447635207e-05}
{"train_loss": 0.2210497111082077, "global_step": 59085, "epoch": 663, "lr": 8.89009805524178e-05}
{"train_loss": 0.18000636994838715, "global_step": 59086, "epoch": 663, "lr": 8.890061633608513e-05}
{"train_loss": 0.22621095180511475, "global_step": 59087, "epoch": 663, "lr": 8.890025211452276e-05}
{"train_loss": 0.262191504240036, "global_step": 59088, "epoch": 663, "lr": 8.88998878877307e-05}
{"train_loss": 0.2982306182384491, "global_step": 59089, "epoch": 663, "lr": 8.889952365570904e-05}
{"train_loss": 0.22855262458324432, "global_step": 59090, "epoch": 663, "lr": 8.88991594184578e-05}
{"train_loss": 0.18585234880447388, "global_step": 59091, "epoch": 663, "lr": 8.889879517597705e-05}
{"train_loss": 0.13914154469966888, "global_step": 59092, "epoch": 663, "lr": 8.889843092826683e-05}
{"train_loss": 0.2736823856830597, "global_step": 59093, "epoch": 663, "lr": 8.889806667532718e-05}
{"train_loss": 0.20821824669837952, "global_step": 59094, "epoch": 663, "lr": 8.889770241715818e-05}
{"train_loss": 0.24457480431942458, "global_step": 59095, "epoch": 663, "lr": 8.889733815375985e-05, "val_loss": 3.4183850288391113}
{"train_loss": 0.10994231700897217, "global_step": 59096, "epoch": 664, "lr": 8.889697388513224e-05}
{"train_loss": 0.2418980747461319, "global_step": 59097, "epoch": 664, "lr": 8.889660961127541e-05}
{"train_loss": 0.19712990522384644, "global_step": 59098, "epoch": 664, "lr": 8.889624533218942e-05}
{"train_loss": 0.27402573823928833, "global_step": 59099, "epoch": 664, "lr": 8.889588104787427e-05}
{"train_loss": 0.2135065644979477, "global_step": 59100, "epoch": 664, "lr": 8.889551675833008e-05}
{"train_loss": 0.24426792562007904, "global_step": 59101, "epoch": 664, "lr": 8.889515246355684e-05}
{"train_loss": 0.26927146315574646, "global_step": 59102, "epoch": 664, "lr": 8.889478816355462e-05}
{"train_loss": 0.25057923793792725, "global_step": 59103, "epoch": 664, "lr": 8.889442385832347e-05}
{"train_loss": 0.24732181429862976, "global_step": 59104, "epoch": 664, "lr": 8.889405954786345e-05}
{"train_loss": 0.18240571022033691, "global_step": 59105, "epoch": 664, "lr": 8.88936952321746e-05}
{"train_loss": 0.1893158107995987, "global_step": 59106, "epoch": 664, "lr": 8.889333091125695e-05}
{"train_loss": 0.2303912490606308, "global_step": 59107, "epoch": 664, "lr": 8.889296658511056e-05}
{"train_loss": 0.2099812626838684, "global_step": 59108, "epoch": 664, "lr": 8.889260225373552e-05}
{"train_loss": 0.2670859694480896, "global_step": 59109, "epoch": 664, "lr": 8.889223791713181e-05}
{"train_loss": 0.21957525610923767, "global_step": 59110, "epoch": 664, "lr": 8.889187357529953e-05}
{"train_loss": 0.27979597449302673, "global_step": 59111, "epoch": 664, "lr": 8.88915092282387e-05}
{"train_loss": 0.2329602688550949, "global_step": 59112, "epoch": 664, "lr": 8.88911448759494e-05}
{"train_loss": 0.1506553590297699, "global_step": 59113, "epoch": 664, "lr": 8.889078051843165e-05}
{"train_loss": 0.22176846861839294, "global_step": 59114, "epoch": 664, "lr": 8.88904161556855e-05}
{"train_loss": 0.20482972264289856, "global_step": 59115, "epoch": 664, "lr": 8.889005178771101e-05}
{"train_loss": 0.16543611884117126, "global_step": 59116, "epoch": 664, "lr": 8.888968741450823e-05}
{"train_loss": 0.2957597076892853, "global_step": 59117, "epoch": 664, "lr": 8.88893230360772e-05}
{"train_loss": 0.23477919399738312, "global_step": 59118, "epoch": 664, "lr": 8.888895865241798e-05}
{"train_loss": 0.18062888085842133, "global_step": 59119, "epoch": 664, "lr": 8.888859426353062e-05}
{"train_loss": 0.1576966643333435, "global_step": 59120, "epoch": 664, "lr": 8.888822986941515e-05}
{"train_loss": 0.2142462432384491, "global_step": 59121, "epoch": 664, "lr": 8.888786547007164e-05}
{"train_loss": 0.23250548541545868, "global_step": 59122, "epoch": 664, "lr": 8.888750106550012e-05}
{"train_loss": 0.2123698741197586, "global_step": 59123, "epoch": 664, "lr": 8.888713665570066e-05}
{"train_loss": 0.2709430158138275, "global_step": 59124, "epoch": 664, "lr": 8.888677224067329e-05}
{"train_loss": 0.18673746287822723, "global_step": 59125, "epoch": 664, "lr": 8.888640782041808e-05}
{"train_loss": 0.3031036853790283, "global_step": 59126, "epoch": 664, "lr": 8.888604339493505e-05}
{"train_loss": 0.1514214724302292, "global_step": 59127, "epoch": 664, "lr": 8.888567896422428e-05}
{"train_loss": 0.19979701936244965, "global_step": 59128, "epoch": 664, "lr": 8.888531452828581e-05}
{"train_loss": 0.2767811715602875, "global_step": 59129, "epoch": 664, "lr": 8.888495008711967e-05}
{"train_loss": 0.2425393909215927, "global_step": 59130, "epoch": 664, "lr": 8.888458564072592e-05}
{"train_loss": 0.2767976224422455, "global_step": 59131, "epoch": 664, "lr": 8.888422118910463e-05}
{"train_loss": 0.2853908836841583, "global_step": 59132, "epoch": 664, "lr": 8.888385673225581e-05}
{"train_loss": 0.24290364980697632, "global_step": 59133, "epoch": 664, "lr": 8.888349227017953e-05}
{"train_loss": 0.16946476697921753, "global_step": 59134, "epoch": 664, "lr": 8.888312780287585e-05}
{"train_loss": 0.17932657897472382, "global_step": 59135, "epoch": 664, "lr": 8.88827633303448e-05}
{"train_loss": 0.2715533971786499, "global_step": 59136, "epoch": 664, "lr": 8.888239885258644e-05}
{"train_loss": 0.22384026646614075, "global_step": 59137, "epoch": 664, "lr": 8.888203436960081e-05}
{"train_loss": 0.2776457369327545, "global_step": 59138, "epoch": 664, "lr": 8.888166988138797e-05}
{"train_loss": 0.1799802929162979, "global_step": 59139, "epoch": 664, "lr": 8.888130538794797e-05}
{"train_loss": 0.20996630191802979, "global_step": 59140, "epoch": 664, "lr": 8.888094088928082e-05}
{"train_loss": 0.21406714618206024, "global_step": 59141, "epoch": 664, "lr": 8.888057638538663e-05}
{"train_loss": 0.2377280294895172, "global_step": 59142, "epoch": 664, "lr": 8.888021187626542e-05}
{"train_loss": 0.13369129598140717, "global_step": 59143, "epoch": 664, "lr": 8.887984736191723e-05}
{"train_loss": 0.2665432095527649, "global_step": 59144, "epoch": 664, "lr": 8.887948284234213e-05}
{"train_loss": 0.20229464769363403, "global_step": 59145, "epoch": 664, "lr": 8.887911831754015e-05}
{"train_loss": 0.2043069452047348, "global_step": 59146, "epoch": 664, "lr": 8.887875378751134e-05}
{"train_loss": 0.28785157203674316, "global_step": 59147, "epoch": 664, "lr": 8.887838925225578e-05}
{"train_loss": 0.28417766094207764, "global_step": 59148, "epoch": 664, "lr": 8.887802471177347e-05}
{"train_loss": 0.2156202346086502, "global_step": 59149, "epoch": 664, "lr": 8.88776601660645e-05}
{"train_loss": 0.14559078216552734, "global_step": 59150, "epoch": 664, "lr": 8.887729561512888e-05}
{"train_loss": 0.266571581363678, "global_step": 59151, "epoch": 664, "lr": 8.88769310589667e-05}
{"train_loss": 0.18364259600639343, "global_step": 59152, "epoch": 664, "lr": 8.887656649757798e-05}
{"train_loss": 0.22054392099380493, "global_step": 59153, "epoch": 664, "lr": 8.887620193096278e-05}
{"train_loss": 0.14569422602653503, "global_step": 59154, "epoch": 664, "lr": 8.887583735912115e-05}
{"train_loss": 0.23854142427444458, "global_step": 59155, "epoch": 664, "lr": 8.887547278205315e-05}
{"train_loss": 0.23790811002254486, "global_step": 59156, "epoch": 664, "lr": 8.887510819975881e-05}
{"train_loss": 0.23228618502616882, "global_step": 59157, "epoch": 664, "lr": 8.887474361223818e-05}
{"train_loss": 0.26793262362480164, "global_step": 59158, "epoch": 664, "lr": 8.887437901949132e-05}
{"train_loss": 0.2149457186460495, "global_step": 59159, "epoch": 664, "lr": 8.887401442151826e-05}
{"train_loss": 0.1860094517469406, "global_step": 59160, "epoch": 664, "lr": 8.887364981831907e-05}
{"train_loss": 0.2141210287809372, "global_step": 59161, "epoch": 664, "lr": 8.88732852098938e-05}
{"train_loss": 0.32180899381637573, "global_step": 59162, "epoch": 664, "lr": 8.887292059624248e-05}
{"train_loss": 0.2768186628818512, "global_step": 59163, "epoch": 664, "lr": 8.887255597736518e-05}
{"train_loss": 0.15876153111457825, "global_step": 59164, "epoch": 664, "lr": 8.887219135326193e-05}
{"train_loss": 0.3709736466407776, "global_step": 59165, "epoch": 664, "lr": 8.88718267239328e-05}
{"train_loss": 0.32050782442092896, "global_step": 59166, "epoch": 664, "lr": 8.887146208937781e-05}
{"train_loss": 0.2651621997356415, "global_step": 59167, "epoch": 664, "lr": 8.887109744959703e-05}
{"train_loss": 0.2979763448238373, "global_step": 59168, "epoch": 664, "lr": 8.887073280459051e-05}
{"train_loss": 0.16117361187934875, "global_step": 59169, "epoch": 664, "lr": 8.88703681543583e-05}
{"train_loss": 0.36242446303367615, "global_step": 59170, "epoch": 664, "lr": 8.887000349890042e-05}
{"train_loss": 0.19289693236351013, "global_step": 59171, "epoch": 664, "lr": 8.886963883821698e-05}
{"train_loss": 0.2615777254104614, "global_step": 59172, "epoch": 664, "lr": 8.886927417230795e-05}
{"train_loss": 0.3091048002243042, "global_step": 59173, "epoch": 664, "lr": 8.886890950117343e-05}
{"train_loss": 0.1964087337255478, "global_step": 59174, "epoch": 664, "lr": 8.886854482481348e-05}
{"train_loss": 0.33577603101730347, "global_step": 59175, "epoch": 664, "lr": 8.886818014322811e-05}
{"train_loss": 0.22070984542369843, "global_step": 59176, "epoch": 664, "lr": 8.886781545641739e-05}
{"train_loss": 0.37367403507232666, "global_step": 59177, "epoch": 664, "lr": 8.886745076438136e-05}
{"train_loss": 0.20058663189411163, "global_step": 59178, "epoch": 664, "lr": 8.886708606712008e-05}
{"train_loss": 0.2755986154079437, "global_step": 59179, "epoch": 664, "lr": 8.886672136463361e-05}
{"train_loss": 0.20167100429534912, "global_step": 59180, "epoch": 664, "lr": 8.886635665692196e-05}
{"train_loss": 0.26013392210006714, "global_step": 59181, "epoch": 664, "lr": 8.886599194398521e-05}
{"train_loss": 0.14435531198978424, "global_step": 59182, "epoch": 664, "lr": 8.886562722582341e-05}
{"train_loss": 0.27032893896102905, "global_step": 59183, "epoch": 664, "lr": 8.886526250243659e-05}
{"train_loss": 0.23345750905154797, "global_step": 59184, "epoch": 664, "lr": 8.886489777382481e-05, "val_loss": 3.2894632816314697}
{"train_loss": 0.17558054625988007, "global_step": 59185, "epoch": 665, "lr": 8.886453303998812e-05}
{"train_loss": 0.23123228549957275, "global_step": 59186, "epoch": 665, "lr": 8.886416830092656e-05}
{"train_loss": 0.3469989001750946, "global_step": 59187, "epoch": 665, "lr": 8.886380355664019e-05}
{"train_loss": 0.29981300234794617, "global_step": 59188, "epoch": 665, "lr": 8.886343880712907e-05}
{"train_loss": 0.2030734419822693, "global_step": 59189, "epoch": 665, "lr": 8.886307405239321e-05}
{"train_loss": 0.24226567149162292, "global_step": 59190, "epoch": 665, "lr": 8.88627092924327e-05}
{"train_loss": 0.324998140335083, "global_step": 59191, "epoch": 665, "lr": 8.886234452724758e-05}
{"train_loss": 0.2510012984275818, "global_step": 59192, "epoch": 665, "lr": 8.886197975683788e-05}
{"train_loss": 0.1965411752462387, "global_step": 59193, "epoch": 665, "lr": 8.886161498120365e-05}
{"train_loss": 0.22689521312713623, "global_step": 59194, "epoch": 665, "lr": 8.886125020034495e-05}
{"train_loss": 0.2551558017730713, "global_step": 59195, "epoch": 665, "lr": 8.886088541426185e-05}
{"train_loss": 0.14525152742862701, "global_step": 59196, "epoch": 665, "lr": 8.886052062295436e-05}
{"train_loss": 0.2256779670715332, "global_step": 59197, "epoch": 665, "lr": 8.886015582642255e-05}
{"train_loss": 0.3062927722930908, "global_step": 59198, "epoch": 665, "lr": 8.885979102466647e-05}
{"train_loss": 0.2508825957775116, "global_step": 59199, "epoch": 665, "lr": 8.885942621768617e-05}
{"train_loss": 0.2107515037059784, "global_step": 59200, "epoch": 665, "lr": 8.885906140548168e-05}
{"train_loss": 0.19745971262454987, "global_step": 59201, "epoch": 665, "lr": 8.885869658805308e-05}
{"train_loss": 0.18232174217700958, "global_step": 59202, "epoch": 665, "lr": 8.88583317654004e-05}
{"train_loss": 0.27976465225219727, "global_step": 59203, "epoch": 665, "lr": 8.885796693752367e-05}
{"train_loss": 0.22890624403953552, "global_step": 59204, "epoch": 665, "lr": 8.885760210442298e-05}
{"train_loss": 0.3463151156902313, "global_step": 59205, "epoch": 665, "lr": 8.885723726609834e-05}
{"train_loss": 0.3125670552253723, "global_step": 59206, "epoch": 665, "lr": 8.885687242254983e-05}
{"train_loss": 0.29412218928337097, "global_step": 59207, "epoch": 665, "lr": 8.885650757377749e-05}
{"train_loss": 0.18299531936645508, "global_step": 59208, "epoch": 665, "lr": 8.885614271978135e-05}
{"train_loss": 0.3208652138710022, "global_step": 59209, "epoch": 665, "lr": 8.88557778605615e-05}
{"train_loss": 0.28680509328842163, "global_step": 59210, "epoch": 665, "lr": 8.885541299611795e-05}
{"train_loss": 0.20645561814308167, "global_step": 59211, "epoch": 665, "lr": 8.885504812645076e-05}
{"train_loss": 0.32126790285110474, "global_step": 59212, "epoch": 665, "lr": 8.885468325155998e-05}
{"train_loss": 0.20866192877292633, "global_step": 59213, "epoch": 665, "lr": 8.885431837144569e-05}
{"train_loss": 0.19400610029697418, "global_step": 59214, "epoch": 665, "lr": 8.885395348610788e-05}
{"train_loss": 0.2513967752456665, "global_step": 59215, "epoch": 665, "lr": 8.885358859554664e-05}
{"train_loss": 0.16003960371017456, "global_step": 59216, "epoch": 665, "lr": 8.8853223699762e-05}
{"train_loss": 0.23635470867156982, "global_step": 59217, "epoch": 665, "lr": 8.885285879875403e-05}
{"train_loss": 0.1867588460445404, "global_step": 59218, "epoch": 665, "lr": 8.885249389252275e-05}
{"train_loss": 0.28878194093704224, "global_step": 59219, "epoch": 665, "lr": 8.885212898106825e-05}
{"train_loss": 0.2653978765010834, "global_step": 59220, "epoch": 665, "lr": 8.885176406439054e-05}
{"train_loss": 0.26368966698646545, "global_step": 59221, "epoch": 665, "lr": 8.885139914248968e-05}
{"train_loss": 0.1760842502117157, "global_step": 59222, "epoch": 665, "lr": 8.885103421536574e-05}
{"train_loss": 0.2689535617828369, "global_step": 59223, "epoch": 665, "lr": 8.885066928301874e-05}
{"train_loss": 0.22115205228328705, "global_step": 59224, "epoch": 665, "lr": 8.885030434544874e-05}
{"train_loss": 0.2243817150592804, "global_step": 59225, "epoch": 665, "lr": 8.884993940265579e-05}
{"train_loss": 0.18592438101768494, "global_step": 59226, "epoch": 665, "lr": 8.884957445463995e-05}
{"train_loss": 0.3141116499900818, "global_step": 59227, "epoch": 665, "lr": 8.884920950140124e-05}
{"train_loss": 0.3200254738330841, "global_step": 59228, "epoch": 665, "lr": 8.884884454293973e-05}
{"train_loss": 0.2089361548423767, "global_step": 59229, "epoch": 665, "lr": 8.884847957925549e-05}
{"train_loss": 0.26038530468940735, "global_step": 59230, "epoch": 665, "lr": 8.884811461034852e-05}
{"train_loss": 0.2644960582256317, "global_step": 59231, "epoch": 665, "lr": 8.884774963621892e-05}
{"train_loss": 0.28075140714645386, "global_step": 59232, "epoch": 665, "lr": 8.884738465686669e-05}
{"train_loss": 0.22261656820774078, "global_step": 59233, "epoch": 665, "lr": 8.88470196722919e-05}
{"train_loss": 0.20790383219718933, "global_step": 59234, "epoch": 665, "lr": 8.884665468249461e-05}
{"train_loss": 0.30840831995010376, "global_step": 59235, "epoch": 665, "lr": 8.884628968747486e-05}
{"train_loss": 0.24380259215831757, "global_step": 59236, "epoch": 665, "lr": 8.88459246872327e-05}
{"train_loss": 0.258024662733078, "global_step": 59237, "epoch": 665, "lr": 8.884555968176818e-05}
{"train_loss": 0.2822894752025604, "global_step": 59238, "epoch": 665, "lr": 8.884519467108136e-05}
{"train_loss": 0.32192501425743103, "global_step": 59239, "epoch": 665, "lr": 8.884482965517227e-05}
{"train_loss": 0.18277662992477417, "global_step": 59240, "epoch": 665, "lr": 8.884446463404095e-05}
{"train_loss": 0.2486468106508255, "global_step": 59241, "epoch": 665, "lr": 8.884409960768749e-05}
{"train_loss": 0.19797958433628082, "global_step": 59242, "epoch": 665, "lr": 8.88437345761119e-05}
{"train_loss": 0.1318720281124115, "global_step": 59243, "epoch": 665, "lr": 8.884336953931423e-05}
{"train_loss": 0.2083510011434555, "global_step": 59244, "epoch": 665, "lr": 8.884300449729456e-05}
{"train_loss": 0.2371366024017334, "global_step": 59245, "epoch": 665, "lr": 8.884263945005291e-05}
{"train_loss": 0.3008171021938324, "global_step": 59246, "epoch": 665, "lr": 8.884227439758937e-05}
{"train_loss": 0.27083897590637207, "global_step": 59247, "epoch": 665, "lr": 8.884190933990393e-05}
{"train_loss": 0.21014241874217987, "global_step": 59248, "epoch": 665, "lr": 8.884154427699667e-05}
{"train_loss": 0.2803822457790375, "global_step": 59249, "epoch": 665, "lr": 8.884117920886765e-05}
{"train_loss": 0.2694859206676483, "global_step": 59250, "epoch": 665, "lr": 8.88408141355169e-05}
{"train_loss": 0.24624375998973846, "global_step": 59251, "epoch": 665, "lr": 8.884044905694448e-05}
{"train_loss": 0.2017093449831009, "global_step": 59252, "epoch": 665, "lr": 8.884008397315042e-05}
{"train_loss": 0.2823581397533417, "global_step": 59253, "epoch": 665, "lr": 8.88397188841348e-05}
{"train_loss": 0.24387653172016144, "global_step": 59254, "epoch": 665, "lr": 8.883935378989765e-05}
{"train_loss": 0.2380901575088501, "global_step": 59255, "epoch": 665, "lr": 8.883898869043903e-05}
{"train_loss": 0.20029747486114502, "global_step": 59256, "epoch": 665, "lr": 8.883862358575898e-05}
{"train_loss": 0.27125102281570435, "global_step": 59257, "epoch": 665, "lr": 8.883825847585754e-05}
{"train_loss": 0.27262917160987854, "global_step": 59258, "epoch": 665, "lr": 8.883789336073478e-05}
{"train_loss": 0.16787773370742798, "global_step": 59259, "epoch": 665, "lr": 8.883752824039074e-05}
{"train_loss": 0.30212122201919556, "global_step": 59260, "epoch": 665, "lr": 8.883716311482544e-05}
{"train_loss": 0.274148553609848, "global_step": 59261, "epoch": 665, "lr": 8.883679798403898e-05}
{"train_loss": 0.273995965719223, "global_step": 59262, "epoch": 665, "lr": 8.883643284803139e-05}
{"train_loss": 0.22232240438461304, "global_step": 59263, "epoch": 665, "lr": 8.88360677068027e-05}
{"train_loss": 0.23430435359477997, "global_step": 59264, "epoch": 665, "lr": 8.883570256035298e-05}
{"train_loss": 0.29700255393981934, "global_step": 59265, "epoch": 665, "lr": 8.883533740868228e-05}
{"train_loss": 0.18983981013298035, "global_step": 59266, "epoch": 665, "lr": 8.883497225179063e-05}
{"train_loss": 0.1705683171749115, "global_step": 59267, "epoch": 665, "lr": 8.88346070896781e-05}
{"train_loss": 0.23237869143486023, "global_step": 59268, "epoch": 665, "lr": 8.883424192234473e-05}
{"train_loss": 0.26315775513648987, "global_step": 59269, "epoch": 665, "lr": 8.883387674979057e-05}
{"train_loss": 0.2679472863674164, "global_step": 59270, "epoch": 665, "lr": 8.883351157201567e-05}
{"train_loss": 0.203520730137825, "global_step": 59271, "epoch": 665, "lr": 8.883314638902007e-05}
{"train_loss": 0.2520767152309418, "global_step": 59272, "epoch": 665, "lr": 8.883278120080383e-05}
{"train_loss": 0.2437804974197002, "global_step": 59273, "epoch": 665, "lr": 8.8832416007367e-05, "val_loss": 3.3669660091400146, "train_action_mse_error": 55.599815368652344}
{"train_loss": 0.16321174800395966, "global_step": 59274, "epoch": 666, "lr": 8.883205080870962e-05}
{"train_loss": 0.26898765563964844, "global_step": 59275, "epoch": 666, "lr": 8.883168560483175e-05}
{"train_loss": 0.18763798475265503, "global_step": 59276, "epoch": 666, "lr": 8.883132039573343e-05}
{"train_loss": 0.20133578777313232, "global_step": 59277, "epoch": 666, "lr": 8.88309551814147e-05}
{"train_loss": 0.20474964380264282, "global_step": 59278, "epoch": 666, "lr": 8.883058996187565e-05}
{"train_loss": 0.28269529342651367, "global_step": 59279, "epoch": 666, "lr": 8.883022473711627e-05}
{"train_loss": 0.17934753000736237, "global_step": 59280, "epoch": 666, "lr": 8.882985950713665e-05}
{"train_loss": 0.20741531252861023, "global_step": 59281, "epoch": 666, "lr": 8.882949427193683e-05}
{"train_loss": 0.2140878587961197, "global_step": 59282, "epoch": 666, "lr": 8.882912903151686e-05}
{"train_loss": 0.24849198758602142, "global_step": 59283, "epoch": 666, "lr": 8.882876378587679e-05}
{"train_loss": 0.28381025791168213, "global_step": 59284, "epoch": 666, "lr": 8.882839853501667e-05}
{"train_loss": 0.16418209671974182, "global_step": 59285, "epoch": 666, "lr": 8.882803327893654e-05}
{"train_loss": 0.2183743417263031, "global_step": 59286, "epoch": 666, "lr": 8.882766801763644e-05}
{"train_loss": 0.1982271820306778, "global_step": 59287, "epoch": 666, "lr": 8.882730275111646e-05}
{"train_loss": 0.2732483446598053, "global_step": 59288, "epoch": 666, "lr": 8.882693747937659e-05}
{"train_loss": 0.44407644867897034, "global_step": 59289, "epoch": 666, "lr": 8.882657220241694e-05}
{"train_loss": 0.25873979926109314, "global_step": 59290, "epoch": 666, "lr": 8.882620692023752e-05}
{"train_loss": 0.19231878221035004, "global_step": 59291, "epoch": 666, "lr": 8.882584163283838e-05}
{"train_loss": 0.16783538460731506, "global_step": 59292, "epoch": 666, "lr": 8.882547634021958e-05}
{"train_loss": 0.21847593784332275, "global_step": 59293, "epoch": 666, "lr": 8.882511104238119e-05}
{"train_loss": 0.23734834790229797, "global_step": 59294, "epoch": 666, "lr": 8.882474573932323e-05}
{"train_loss": 0.13430730998516083, "global_step": 59295, "epoch": 666, "lr": 8.882438043104573e-05}
{"train_loss": 0.2825945317745209, "global_step": 59296, "epoch": 666, "lr": 8.882401511754879e-05}
{"train_loss": 0.26059359312057495, "global_step": 59297, "epoch": 666, "lr": 8.882364979883243e-05}
{"train_loss": 0.33659544587135315, "global_step": 59298, "epoch": 666, "lr": 8.88232844748967e-05}
{"train_loss": 0.2201993465423584, "global_step": 59299, "epoch": 666, "lr": 8.882291914574165e-05}
{"train_loss": 0.20816054940223694, "global_step": 59300, "epoch": 666, "lr": 8.882255381136735e-05}
{"train_loss": 0.18956899642944336, "global_step": 59301, "epoch": 666, "lr": 8.882218847177382e-05}
{"train_loss": 0.2714783847332001, "global_step": 59302, "epoch": 666, "lr": 8.88218231269611e-05}
{"train_loss": 0.3055105209350586, "global_step": 59303, "epoch": 666, "lr": 8.882145777692927e-05}
{"train_loss": 0.34076380729675293, "global_step": 59304, "epoch": 666, "lr": 8.882109242167839e-05}
{"train_loss": 0.29449722170829773, "global_step": 59305, "epoch": 666, "lr": 8.882072706120846e-05}
{"train_loss": 0.27610260248184204, "global_step": 59306, "epoch": 666, "lr": 8.882036169551957e-05}
{"train_loss": 0.1783420890569687, "global_step": 59307, "epoch": 666, "lr": 8.881999632461175e-05}
{"train_loss": 0.2722577452659607, "global_step": 59308, "epoch": 666, "lr": 8.881963094848507e-05}
{"train_loss": 0.23637554049491882, "global_step": 59309, "epoch": 666, "lr": 8.881926556713956e-05}
{"train_loss": 0.2942407429218292, "global_step": 59310, "epoch": 666, "lr": 8.881890018057527e-05}
{"train_loss": 0.2573084235191345, "global_step": 59311, "epoch": 666, "lr": 8.881853478879223e-05}
{"train_loss": 0.1456323266029358, "global_step": 59312, "epoch": 666, "lr": 8.881816939179053e-05}
{"train_loss": 0.30597177147865295, "global_step": 59313, "epoch": 666, "lr": 8.881780398957019e-05}
{"train_loss": 0.3152444064617157, "global_step": 59314, "epoch": 666, "lr": 8.881743858213128e-05}
{"train_loss": 0.1861390322446823, "global_step": 59315, "epoch": 666, "lr": 8.881707316947383e-05}
{"train_loss": 0.2969941794872284, "global_step": 59316, "epoch": 666, "lr": 8.881670775159791e-05}
{"train_loss": 0.19906510412693024, "global_step": 59317, "epoch": 666, "lr": 8.881634232850355e-05}
{"train_loss": 0.17153117060661316, "global_step": 59318, "epoch": 666, "lr": 8.881597690019081e-05}
{"train_loss": 0.18344208598136902, "global_step": 59319, "epoch": 666, "lr": 8.881561146665973e-05}
{"train_loss": 0.299651175737381, "global_step": 59320, "epoch": 666, "lr": 8.881524602791036e-05}
{"train_loss": 0.2873322367668152, "global_step": 59321, "epoch": 666, "lr": 8.881488058394277e-05}
{"train_loss": 0.22620610892772675, "global_step": 59322, "epoch": 666, "lr": 8.881451513475699e-05}
{"train_loss": 0.18462097644805908, "global_step": 59323, "epoch": 666, "lr": 8.881414968035304e-05}
{"train_loss": 0.23816505074501038, "global_step": 59324, "epoch": 666, "lr": 8.881378422073103e-05}
{"train_loss": 0.24542737007141113, "global_step": 59325, "epoch": 666, "lr": 8.881341875589097e-05}
{"train_loss": 0.2601662874221802, "global_step": 59326, "epoch": 666, "lr": 8.881305328583293e-05}
{"train_loss": 0.2452872097492218, "global_step": 59327, "epoch": 666, "lr": 8.881268781055694e-05}
{"train_loss": 0.20594751834869385, "global_step": 59328, "epoch": 666, "lr": 8.881232233006305e-05}
{"train_loss": 0.2260955274105072, "global_step": 59329, "epoch": 666, "lr": 8.881195684435132e-05}
{"train_loss": 0.20176340639591217, "global_step": 59330, "epoch": 666, "lr": 8.881159135342181e-05}
{"train_loss": 0.24552775919437408, "global_step": 59331, "epoch": 666, "lr": 8.881122585727453e-05}
{"train_loss": 0.3028115928173065, "global_step": 59332, "epoch": 666, "lr": 8.881086035590957e-05}
{"train_loss": 0.3369113504886627, "global_step": 59333, "epoch": 666, "lr": 8.881049484932696e-05}
{"train_loss": 0.29412201046943665, "global_step": 59334, "epoch": 666, "lr": 8.881012933752674e-05}
{"train_loss": 0.2403954416513443, "global_step": 59335, "epoch": 666, "lr": 8.880976382050901e-05}
{"train_loss": 0.17056933045387268, "global_step": 59336, "epoch": 666, "lr": 8.880939829827375e-05}
{"train_loss": 0.23197400569915771, "global_step": 59337, "epoch": 666, "lr": 8.880903277082103e-05}
{"train_loss": 0.2576850354671478, "global_step": 59338, "epoch": 666, "lr": 8.880866723815093e-05}
{"train_loss": 0.26731279492378235, "global_step": 59339, "epoch": 666, "lr": 8.880830170026347e-05}
{"train_loss": 0.26311635971069336, "global_step": 59340, "epoch": 666, "lr": 8.88079361571587e-05}
{"train_loss": 0.25134357810020447, "global_step": 59341, "epoch": 666, "lr": 8.880757060883669e-05}
{"train_loss": 0.22776338458061218, "global_step": 59342, "epoch": 666, "lr": 8.880720505529745e-05}
{"train_loss": 0.19598457217216492, "global_step": 59343, "epoch": 666, "lr": 8.880683949654108e-05}
{"train_loss": 0.2022390514612198, "global_step": 59344, "epoch": 666, "lr": 8.88064739325676e-05}
{"train_loss": 0.23534457385540009, "global_step": 59345, "epoch": 666, "lr": 8.880610836337706e-05}
{"train_loss": 0.2792115807533264, "global_step": 59346, "epoch": 666, "lr": 8.880574278896951e-05}
{"train_loss": 0.19449755549430847, "global_step": 59347, "epoch": 666, "lr": 8.880537720934499e-05}
{"train_loss": 0.22814716398715973, "global_step": 59348, "epoch": 666, "lr": 8.880501162450357e-05}
{"train_loss": 0.2781097888946533, "global_step": 59349, "epoch": 666, "lr": 8.88046460344453e-05}
{"train_loss": 0.29377058148384094, "global_step": 59350, "epoch": 666, "lr": 8.880428043917019e-05}
{"train_loss": 0.15957267582416534, "global_step": 59351, "epoch": 666, "lr": 8.880391483867833e-05}
{"train_loss": 0.2577444612979889, "global_step": 59352, "epoch": 666, "lr": 8.880354923296977e-05}
{"train_loss": 0.27445122599601746, "global_step": 59353, "epoch": 666, "lr": 8.880318362204454e-05}
{"train_loss": 0.230193093419075, "global_step": 59354, "epoch": 666, "lr": 8.880281800590269e-05}
{"train_loss": 0.30943357944488525, "global_step": 59355, "epoch": 666, "lr": 8.880245238454428e-05}
{"train_loss": 0.3243839144706726, "global_step": 59356, "epoch": 666, "lr": 8.880208675796935e-05}
{"train_loss": 0.17910055816173553, "global_step": 59357, "epoch": 666, "lr": 8.880172112617796e-05}
{"train_loss": 0.2833518981933594, "global_step": 59358, "epoch": 666, "lr": 8.880135548917013e-05}
{"train_loss": 0.24654850363731384, "global_step": 59359, "epoch": 666, "lr": 8.880098984694595e-05}
{"train_loss": 0.22708499431610107, "global_step": 59360, "epoch": 666, "lr": 8.880062419950543e-05}
{"train_loss": 0.2406700998544693, "global_step": 59361, "epoch": 666, "lr": 8.880025854684866e-05}
{"train_loss": 0.2422683069880089, "global_step": 59362, "epoch": 666, "lr": 8.879989288897566e-05, "val_loss": 3.418417453765869}
{"train_loss": 0.3718591630458832, "global_step": 59363, "epoch": 667, "lr": 8.879952722588649e-05}
{"train_loss": 0.20202885568141937, "global_step": 59364, "epoch": 667, "lr": 8.87991615575812e-05}
{"train_loss": 0.2755300998687744, "global_step": 59365, "epoch": 667, "lr": 8.879879588405983e-05}
{"train_loss": 0.26168254017829895, "global_step": 59366, "epoch": 667, "lr": 8.879843020532244e-05}
{"train_loss": 0.23301763832569122, "global_step": 59367, "epoch": 667, "lr": 8.879806452136907e-05}
{"train_loss": 0.169899582862854, "global_step": 59368, "epoch": 667, "lr": 8.879769883219977e-05}
{"train_loss": 0.29230454564094543, "global_step": 59369, "epoch": 667, "lr": 8.87973331378146e-05}
{"train_loss": 0.20561730861663818, "global_step": 59370, "epoch": 667, "lr": 8.879696743821357e-05}
{"train_loss": 0.21848109364509583, "global_step": 59371, "epoch": 667, "lr": 8.87966017333968e-05}
{"train_loss": 0.22339296340942383, "global_step": 59372, "epoch": 667, "lr": 8.879623602336429e-05}
{"train_loss": 0.33415794372558594, "global_step": 59373, "epoch": 667, "lr": 8.87958703081161e-05}
{"train_loss": 0.26849403977394104, "global_step": 59374, "epoch": 667, "lr": 8.879550458765227e-05}
{"train_loss": 0.22034654021263123, "global_step": 59375, "epoch": 667, "lr": 8.879513886197286e-05}
{"train_loss": 0.2051752358675003, "global_step": 59376, "epoch": 667, "lr": 8.87947731310779e-05}
{"train_loss": 0.2572333812713623, "global_step": 59377, "epoch": 667, "lr": 8.87944073949675e-05}
{"train_loss": 0.25544747710227966, "global_step": 59378, "epoch": 667, "lr": 8.879404165364163e-05}
{"train_loss": 0.23512528836727142, "global_step": 59379, "epoch": 667, "lr": 8.879367590710036e-05}
{"train_loss": 0.1330709159374237, "global_step": 59380, "epoch": 667, "lr": 8.879331015534378e-05}
{"train_loss": 0.2374621480703354, "global_step": 59381, "epoch": 667, "lr": 8.879294439837191e-05}
{"train_loss": 0.23418883979320526, "global_step": 59382, "epoch": 667, "lr": 8.87925786361848e-05}
{"train_loss": 0.3559800684452057, "global_step": 59383, "epoch": 667, "lr": 8.87922128687825e-05}
{"train_loss": 0.2643425762653351, "global_step": 59384, "epoch": 667, "lr": 8.879184709616506e-05}
{"train_loss": 0.20749925076961517, "global_step": 59385, "epoch": 667, "lr": 8.879148131833253e-05}
{"train_loss": 0.36073657870292664, "global_step": 59386, "epoch": 667, "lr": 8.879111553528495e-05}
{"train_loss": 0.2178102731704712, "global_step": 59387, "epoch": 667, "lr": 8.879074974702238e-05}
{"train_loss": 0.2355380803346634, "global_step": 59388, "epoch": 667, "lr": 8.879038395354487e-05}
{"train_loss": 0.15757541358470917, "global_step": 59389, "epoch": 667, "lr": 8.879001815485247e-05}
{"train_loss": 0.25027891993522644, "global_step": 59390, "epoch": 667, "lr": 8.878965235094521e-05}
{"train_loss": 0.22826310992240906, "global_step": 59391, "epoch": 667, "lr": 8.878928654182317e-05}
{"train_loss": 0.27742257714271545, "global_step": 59392, "epoch": 667, "lr": 8.878892072748638e-05}
{"train_loss": 0.2181004136800766, "global_step": 59393, "epoch": 667, "lr": 8.87885549079349e-05}
{"train_loss": 0.3293657898902893, "global_step": 59394, "epoch": 667, "lr": 8.878818908316875e-05}
{"train_loss": 0.2248714566230774, "global_step": 59395, "epoch": 667, "lr": 8.878782325318802e-05}
{"train_loss": 0.24803675711154938, "global_step": 59396, "epoch": 667, "lr": 8.878745741799273e-05}
{"train_loss": 0.1689704954624176, "global_step": 59397, "epoch": 667, "lr": 8.878709157758293e-05}
{"train_loss": 0.23841102421283722, "global_step": 59398, "epoch": 667, "lr": 8.87867257319587e-05}
{"train_loss": 0.19488783180713654, "global_step": 59399, "epoch": 667, "lr": 8.878635988112005e-05}
{"train_loss": 0.21898938715457916, "global_step": 59400, "epoch": 667, "lr": 8.878599402506705e-05}
{"train_loss": 0.249521866440773, "global_step": 59401, "epoch": 667, "lr": 8.878562816379976e-05}
{"train_loss": 0.240303635597229, "global_step": 59402, "epoch": 667, "lr": 8.87852622973182e-05}
{"train_loss": 0.2601996660232544, "global_step": 59403, "epoch": 667, "lr": 8.878489642562245e-05}
{"train_loss": 0.2629513442516327, "global_step": 59404, "epoch": 667, "lr": 8.878453054871252e-05}
{"train_loss": 0.25372859835624695, "global_step": 59405, "epoch": 667, "lr": 8.878416466658849e-05}
{"train_loss": 0.21149112284183502, "global_step": 59406, "epoch": 667, "lr": 8.87837987792504e-05}
{"train_loss": 0.2811118960380554, "global_step": 59407, "epoch": 667, "lr": 8.878343288669831e-05}
{"train_loss": 0.2566215693950653, "global_step": 59408, "epoch": 667, "lr": 8.878306698893226e-05}
{"train_loss": 0.18073703348636627, "global_step": 59409, "epoch": 667, "lr": 8.87827010859523e-05}
{"train_loss": 0.2140878438949585, "global_step": 59410, "epoch": 667, "lr": 8.878233517775846e-05}
{"train_loss": 0.2635307312011719, "global_step": 59411, "epoch": 667, "lr": 8.878196926435083e-05}
{"train_loss": 0.16310718655586243, "global_step": 59412, "epoch": 667, "lr": 8.878160334572943e-05}
{"train_loss": 0.28009700775146484, "global_step": 59413, "epoch": 667, "lr": 8.878123742189432e-05}
{"train_loss": 0.22212056815624237, "global_step": 59414, "epoch": 667, "lr": 8.878087149284554e-05}
{"train_loss": 0.26761704683303833, "global_step": 59415, "epoch": 667, "lr": 8.878050555858314e-05}
{"train_loss": 0.2368909865617752, "global_step": 59416, "epoch": 667, "lr": 8.878013961910718e-05}
{"train_loss": 0.2581300139427185, "global_step": 59417, "epoch": 667, "lr": 8.87797736744177e-05}
{"train_loss": 0.2703667879104614, "global_step": 59418, "epoch": 667, "lr": 8.877940772451477e-05}
{"train_loss": 0.34012147784233093, "global_step": 59419, "epoch": 667, "lr": 8.877904176939839e-05}
{"train_loss": 0.19883902370929718, "global_step": 59420, "epoch": 667, "lr": 8.877867580906867e-05}
{"train_loss": 0.1726328432559967, "global_step": 59421, "epoch": 667, "lr": 8.877830984352562e-05}
{"train_loss": 0.16158510744571686, "global_step": 59422, "epoch": 667, "lr": 8.877794387276928e-05}
{"train_loss": 0.21954314410686493, "global_step": 59423, "epoch": 667, "lr": 8.877757789679973e-05}
{"train_loss": 0.1893889307975769, "global_step": 59424, "epoch": 667, "lr": 8.8777211915617e-05}
{"train_loss": 0.29332906007766724, "global_step": 59425, "epoch": 667, "lr": 8.877684592922118e-05}
{"train_loss": 0.28465035557746887, "global_step": 59426, "epoch": 667, "lr": 8.877647993761224e-05}
{"train_loss": 0.24189434945583344, "global_step": 59427, "epoch": 667, "lr": 8.877611394079031e-05}
{"train_loss": 0.18234997987747192, "global_step": 59428, "epoch": 667, "lr": 8.877574793875538e-05}
{"train_loss": 0.2675776183605194, "global_step": 59429, "epoch": 667, "lr": 8.877538193150754e-05}
{"train_loss": 0.27335014939308167, "global_step": 59430, "epoch": 667, "lr": 8.877501591904682e-05}
{"train_loss": 0.23460330069065094, "global_step": 59431, "epoch": 667, "lr": 8.877464990137328e-05}
{"train_loss": 0.22911867499351501, "global_step": 59432, "epoch": 667, "lr": 8.877428387848693e-05}
{"train_loss": 0.2229638546705246, "global_step": 59433, "epoch": 667, "lr": 8.877391785038788e-05}
{"train_loss": 0.3070010542869568, "global_step": 59434, "epoch": 667, "lr": 8.877355181707614e-05}
{"train_loss": 0.2708891034126282, "global_step": 59435, "epoch": 667, "lr": 8.877318577855176e-05}
{"train_loss": 0.2919173836708069, "global_step": 59436, "epoch": 667, "lr": 8.877281973481481e-05}
{"train_loss": 0.21647052466869354, "global_step": 59437, "epoch": 667, "lr": 8.877245368586533e-05}
{"train_loss": 0.2354893535375595, "global_step": 59438, "epoch": 667, "lr": 8.877208763170335e-05}
{"train_loss": 0.2990964353084564, "global_step": 59439, "epoch": 667, "lr": 8.877172157232894e-05}
{"train_loss": 0.3714154362678528, "global_step": 59440, "epoch": 667, "lr": 8.877135550774216e-05}
{"train_loss": 0.25528445839881897, "global_step": 59441, "epoch": 667, "lr": 8.877098943794303e-05}
{"train_loss": 0.25301989912986755, "global_step": 59442, "epoch": 667, "lr": 8.87706233629316e-05}
{"train_loss": 0.2418597936630249, "global_step": 59443, "epoch": 667, "lr": 8.877025728270796e-05}
{"train_loss": 0.2525199055671692, "global_step": 59444, "epoch": 667, "lr": 8.876989119727211e-05}
{"train_loss": 0.2346685230731964, "global_step": 59445, "epoch": 667, "lr": 8.876952510662413e-05}
{"train_loss": 0.18665549159049988, "global_step": 59446, "epoch": 667, "lr": 8.876915901076406e-05}
{"train_loss": 0.16701500117778778, "global_step": 59447, "epoch": 667, "lr": 8.876879290969195e-05}
{"train_loss": 0.26177647709846497, "global_step": 59448, "epoch": 667, "lr": 8.876842680340786e-05}
{"train_loss": 0.17291761934757233, "global_step": 59449, "epoch": 667, "lr": 8.876806069191181e-05}
{"train_loss": 0.21506373584270477, "global_step": 59450, "epoch": 667, "lr": 8.876769457520387e-05}
{"train_loss": 0.2418123229166095, "global_step": 59451, "epoch": 667, "lr": 8.876732845328409e-05, "val_loss": 3.2731070518493652}
{"train_loss": 0.2313905507326126, "global_step": 59452, "epoch": 668, "lr": 8.876696232615252e-05}
{"train_loss": 0.18518123030662537, "global_step": 59453, "epoch": 668, "lr": 8.876659619380918e-05}
{"train_loss": 0.2755500376224518, "global_step": 59454, "epoch": 668, "lr": 8.876623005625417e-05}
{"train_loss": 0.3004891872406006, "global_step": 59455, "epoch": 668, "lr": 8.87658639134875e-05}
{"train_loss": 0.24207668006420135, "global_step": 59456, "epoch": 668, "lr": 8.876549776550923e-05}
{"train_loss": 0.1776864379644394, "global_step": 59457, "epoch": 668, "lr": 8.876513161231941e-05}
{"train_loss": 0.18017485737800598, "global_step": 59458, "epoch": 668, "lr": 8.876476545391811e-05}
{"train_loss": 0.2635846436023712, "global_step": 59459, "epoch": 668, "lr": 8.876439929030535e-05}
{"train_loss": 0.2784498929977417, "global_step": 59460, "epoch": 668, "lr": 8.876403312148118e-05}
{"train_loss": 0.29275014996528625, "global_step": 59461, "epoch": 668, "lr": 8.876366694744567e-05}
{"train_loss": 0.1951633244752884, "global_step": 59462, "epoch": 668, "lr": 8.876330076819884e-05}
{"train_loss": 0.256174772977829, "global_step": 59463, "epoch": 668, "lr": 8.876293458374078e-05}
{"train_loss": 0.18759748339653015, "global_step": 59464, "epoch": 668, "lr": 8.87625683940715e-05}
{"train_loss": 0.25105249881744385, "global_step": 59465, "epoch": 668, "lr": 8.876220219919107e-05}
{"train_loss": 0.2270328253507614, "global_step": 59466, "epoch": 668, "lr": 8.876183599909954e-05}
{"train_loss": 0.2958292067050934, "global_step": 59467, "epoch": 668, "lr": 8.876146979379694e-05}
{"train_loss": 0.22161923348903656, "global_step": 59468, "epoch": 668, "lr": 8.876110358328333e-05}
{"train_loss": 0.26956015825271606, "global_step": 59469, "epoch": 668, "lr": 8.876073736755877e-05}
{"train_loss": 0.18049734830856323, "global_step": 59470, "epoch": 668, "lr": 8.87603711466233e-05}
{"train_loss": 0.26199403405189514, "global_step": 59471, "epoch": 668, "lr": 8.876000492047697e-05}
{"train_loss": 0.2769987881183624, "global_step": 59472, "epoch": 668, "lr": 8.875963868911986e-05}
{"train_loss": 0.24197466671466827, "global_step": 59473, "epoch": 668, "lr": 8.875927245255196e-05}
{"train_loss": 0.24749287962913513, "global_step": 59474, "epoch": 668, "lr": 8.875890621077334e-05}
{"train_loss": 0.22338803112506866, "global_step": 59475, "epoch": 668, "lr": 8.875853996378407e-05}
{"train_loss": 0.25711822509765625, "global_step": 59476, "epoch": 668, "lr": 8.875817371158419e-05}
{"train_loss": 0.1928698718547821, "global_step": 59477, "epoch": 668, "lr": 8.875780745417375e-05}
{"train_loss": 0.24319055676460266, "global_step": 59478, "epoch": 668, "lr": 8.875744119155278e-05}
{"train_loss": 0.23866379261016846, "global_step": 59479, "epoch": 668, "lr": 8.875707492372134e-05}
{"train_loss": 0.2528195381164551, "global_step": 59480, "epoch": 668, "lr": 8.87567086506795e-05}
{"train_loss": 0.28411075472831726, "global_step": 59481, "epoch": 668, "lr": 8.875634237242729e-05}
{"train_loss": 0.21556591987609863, "global_step": 59482, "epoch": 668, "lr": 8.875597608896475e-05}
{"train_loss": 0.2850378751754761, "global_step": 59483, "epoch": 668, "lr": 8.875560980029197e-05}
{"train_loss": 0.2517591118812561, "global_step": 59484, "epoch": 668, "lr": 8.875524350640894e-05}
{"train_loss": 0.3215928077697754, "global_step": 59485, "epoch": 668, "lr": 8.875487720731575e-05}
{"train_loss": 0.2550015151500702, "global_step": 59486, "epoch": 668, "lr": 8.875451090301244e-05}
{"train_loss": 0.22389031946659088, "global_step": 59487, "epoch": 668, "lr": 8.875414459349905e-05}
{"train_loss": 0.14489363133907318, "global_step": 59488, "epoch": 668, "lr": 8.875377827877565e-05}
{"train_loss": 0.16353806853294373, "global_step": 59489, "epoch": 668, "lr": 8.875341195884228e-05}
{"train_loss": 0.3144276440143585, "global_step": 59490, "epoch": 668, "lr": 8.875304563369898e-05}
{"train_loss": 0.23027291893959045, "global_step": 59491, "epoch": 668, "lr": 8.875267930334581e-05}
{"train_loss": 0.26940208673477173, "global_step": 59492, "epoch": 668, "lr": 8.87523129677828e-05}
{"train_loss": 0.266826331615448, "global_step": 59493, "epoch": 668, "lr": 8.875194662701003e-05}
{"train_loss": 0.29981833696365356, "global_step": 59494, "epoch": 668, "lr": 8.875158028102753e-05}
{"train_loss": 0.20649030804634094, "global_step": 59495, "epoch": 668, "lr": 8.875121392983534e-05}
{"train_loss": 0.23709462583065033, "global_step": 59496, "epoch": 668, "lr": 8.875084757343353e-05}
{"train_loss": 0.21617421507835388, "global_step": 59497, "epoch": 668, "lr": 8.875048121182214e-05}
{"train_loss": 0.18628886342048645, "global_step": 59498, "epoch": 668, "lr": 8.875011484500121e-05}
{"train_loss": 0.2705201208591461, "global_step": 59499, "epoch": 668, "lr": 8.874974847297082e-05}
{"train_loss": 0.2429816573858261, "global_step": 59500, "epoch": 668, "lr": 8.874938209573098e-05}
{"train_loss": 0.2885420620441437, "global_step": 59501, "epoch": 668, "lr": 8.874901571328178e-05}
{"train_loss": 0.2511245608329773, "global_step": 59502, "epoch": 668, "lr": 8.874864932562323e-05}
{"train_loss": 0.3066452443599701, "global_step": 59503, "epoch": 668, "lr": 8.87482829327554e-05}
{"train_loss": 0.26793956756591797, "global_step": 59504, "epoch": 668, "lr": 8.874791653467833e-05}
{"train_loss": 0.2573411762714386, "global_step": 59505, "epoch": 668, "lr": 8.874755013139208e-05}
{"train_loss": 0.2583359479904175, "global_step": 59506, "epoch": 668, "lr": 8.87471837228967e-05}
{"train_loss": 0.29325753450393677, "global_step": 59507, "epoch": 668, "lr": 8.874681730919223e-05}
{"train_loss": 0.3021453619003296, "global_step": 59508, "epoch": 668, "lr": 8.874645089027872e-05}
{"train_loss": 0.18661920726299286, "global_step": 59509, "epoch": 668, "lr": 8.874608446615622e-05}
{"train_loss": 0.2768460810184479, "global_step": 59510, "epoch": 668, "lr": 8.874571803682479e-05}
{"train_loss": 0.27403169870376587, "global_step": 59511, "epoch": 668, "lr": 8.874535160228447e-05}
{"train_loss": 0.19553424417972565, "global_step": 59512, "epoch": 668, "lr": 8.874498516253529e-05}
{"train_loss": 0.22969332337379456, "global_step": 59513, "epoch": 668, "lr": 8.874461871757734e-05}
{"train_loss": 0.243422269821167, "global_step": 59514, "epoch": 668, "lr": 8.874425226741064e-05}
{"train_loss": 0.23345449566841125, "global_step": 59515, "epoch": 668, "lr": 8.874388581203526e-05}
{"train_loss": 0.1732083410024643, "global_step": 59516, "epoch": 668, "lr": 8.874351935145122e-05}
{"train_loss": 0.21931639313697815, "global_step": 59517, "epoch": 668, "lr": 8.87431528856586e-05}
{"train_loss": 0.25651317834854126, "global_step": 59518, "epoch": 668, "lr": 8.874278641465742e-05}
{"train_loss": 0.24771371483802795, "global_step": 59519, "epoch": 668, "lr": 8.874241993844776e-05}
{"train_loss": 0.2708057463169098, "global_step": 59520, "epoch": 668, "lr": 8.874205345702964e-05}
{"train_loss": 0.2564798593521118, "global_step": 59521, "epoch": 668, "lr": 8.874168697040313e-05}
{"train_loss": 0.19027338922023773, "global_step": 59522, "epoch": 668, "lr": 8.874132047856827e-05}
{"train_loss": 0.20715324580669403, "global_step": 59523, "epoch": 668, "lr": 8.87409539815251e-05}
{"train_loss": 0.1923730969429016, "global_step": 59524, "epoch": 668, "lr": 8.874058747927369e-05}
{"train_loss": 0.27407628297805786, "global_step": 59525, "epoch": 668, "lr": 8.87402209718141e-05}
{"train_loss": 0.15338866412639618, "global_step": 59526, "epoch": 668, "lr": 8.873985445914635e-05}
{"train_loss": 0.28856849670410156, "global_step": 59527, "epoch": 668, "lr": 8.873948794127048e-05}
{"train_loss": 0.2139177918434143, "global_step": 59528, "epoch": 668, "lr": 8.873912141818657e-05}
{"train_loss": 0.22260431945323944, "global_step": 59529, "epoch": 668, "lr": 8.873875488989466e-05}
{"train_loss": 0.17910754680633545, "global_step": 59530, "epoch": 668, "lr": 8.873838835639479e-05}
{"train_loss": 0.2624441385269165, "global_step": 59531, "epoch": 668, "lr": 8.873802181768702e-05}
{"train_loss": 0.1998339295387268, "global_step": 59532, "epoch": 668, "lr": 8.87376552737714e-05}
{"train_loss": 0.17547915875911713, "global_step": 59533, "epoch": 668, "lr": 8.873728872464798e-05}
{"train_loss": 0.2500825524330139, "global_step": 59534, "epoch": 668, "lr": 8.873692217031678e-05}
{"train_loss": 0.22143206000328064, "global_step": 59535, "epoch": 668, "lr": 8.873655561077788e-05}
{"train_loss": 0.3306310176849365, "global_step": 59536, "epoch": 668, "lr": 8.873618904603134e-05}
{"train_loss": 0.2788800001144409, "global_step": 59537, "epoch": 668, "lr": 8.873582247607716e-05}
{"train_loss": 0.27554866671562195, "global_step": 59538, "epoch": 668, "lr": 8.873545590091546e-05}
{"train_loss": 0.26297855377197266, "global_step": 59539, "epoch": 668, "lr": 8.873508932054622e-05}
{"train_loss": 0.24122511938716587, "global_step": 59540, "epoch": 668, "lr": 8.873472273496954e-05, "val_loss": 3.3242347240448}
{"train_loss": 0.32229894399642944, "global_step": 59541, "epoch": 669, "lr": 8.873435614418544e-05}
{"train_loss": 0.18982048332691193, "global_step": 59542, "epoch": 669, "lr": 8.873398954819396e-05}
{"train_loss": 0.21277301013469696, "global_step": 59543, "epoch": 669, "lr": 8.873362294699518e-05}
{"train_loss": 0.2343011051416397, "global_step": 59544, "epoch": 669, "lr": 8.873325634058916e-05}
{"train_loss": 0.18228009343147278, "global_step": 59545, "epoch": 669, "lr": 8.87328897289759e-05}
{"train_loss": 0.18129949271678925, "global_step": 59546, "epoch": 669, "lr": 8.873252311215547e-05}
{"train_loss": 0.1800319254398346, "global_step": 59547, "epoch": 669, "lr": 8.873215649012793e-05}
{"train_loss": 0.19577722251415253, "global_step": 59548, "epoch": 669, "lr": 8.873178986289334e-05}
{"train_loss": 0.249800443649292, "global_step": 59549, "epoch": 669, "lr": 8.873142323045173e-05}
{"train_loss": 0.22558394074440002, "global_step": 59550, "epoch": 669, "lr": 8.873105659280314e-05}
{"train_loss": 0.19368712604045868, "global_step": 59551, "epoch": 669, "lr": 8.873068994994764e-05}
{"train_loss": 0.22241027653217316, "global_step": 59552, "epoch": 669, "lr": 8.873032330188526e-05}
{"train_loss": 0.28659266233444214, "global_step": 59553, "epoch": 669, "lr": 8.872995664861607e-05}
{"train_loss": 0.26768729090690613, "global_step": 59554, "epoch": 669, "lr": 8.87295899901401e-05}
{"train_loss": 0.2639889121055603, "global_step": 59555, "epoch": 669, "lr": 8.872922332645742e-05}
{"train_loss": 0.2739880084991455, "global_step": 59556, "epoch": 669, "lr": 8.872885665756807e-05}
{"train_loss": 0.16004405915737152, "global_step": 59557, "epoch": 669, "lr": 8.872848998347209e-05}
{"train_loss": 0.279550164937973, "global_step": 59558, "epoch": 669, "lr": 8.872812330416954e-05}
{"train_loss": 0.1496521383523941, "global_step": 59559, "epoch": 669, "lr": 8.872775661966044e-05}
{"train_loss": 0.24957504868507385, "global_step": 59560, "epoch": 669, "lr": 8.872738992994489e-05}
{"train_loss": 0.217937171459198, "global_step": 59561, "epoch": 669, "lr": 8.872702323502291e-05}
{"train_loss": 0.17082157731056213, "global_step": 59562, "epoch": 669, "lr": 8.872665653489456e-05}
{"train_loss": 0.3018716275691986, "global_step": 59563, "epoch": 669, "lr": 8.872628982955988e-05}
{"train_loss": 0.26845526695251465, "global_step": 59564, "epoch": 669, "lr": 8.872592311901893e-05}
{"train_loss": 0.2443118691444397, "global_step": 59565, "epoch": 669, "lr": 8.872555640327174e-05}
{"train_loss": 0.22275400161743164, "global_step": 59566, "epoch": 669, "lr": 8.872518968231838e-05}
{"train_loss": 0.18329544365406036, "global_step": 59567, "epoch": 669, "lr": 8.872482295615887e-05}
{"train_loss": 0.23235788941383362, "global_step": 59568, "epoch": 669, "lr": 8.872445622479331e-05}
{"train_loss": 0.22333544492721558, "global_step": 59569, "epoch": 669, "lr": 8.872408948822168e-05}
{"train_loss": 0.21848180890083313, "global_step": 59570, "epoch": 669, "lr": 8.87237227464441e-05}
{"train_loss": 0.21033306419849396, "global_step": 59571, "epoch": 669, "lr": 8.872335599946058e-05}
{"train_loss": 0.20960453152656555, "global_step": 59572, "epoch": 669, "lr": 8.872298924727118e-05}
{"train_loss": 0.2668014466762543, "global_step": 59573, "epoch": 669, "lr": 8.872262248987592e-05}
{"train_loss": 0.25543123483657837, "global_step": 59574, "epoch": 669, "lr": 8.872225572727491e-05}
{"train_loss": 0.28745725750923157, "global_step": 59575, "epoch": 669, "lr": 8.872188895946816e-05}
{"train_loss": 0.25344765186309814, "global_step": 59576, "epoch": 669, "lr": 8.87215221864557e-05}
{"train_loss": 0.21922576427459717, "global_step": 59577, "epoch": 669, "lr": 8.872115540823762e-05}
{"train_loss": 0.2201555371284485, "global_step": 59578, "epoch": 669, "lr": 8.872078862481395e-05}
{"train_loss": 0.18427564203739166, "global_step": 59579, "epoch": 669, "lr": 8.872042183618475e-05}
{"train_loss": 0.24981161952018738, "global_step": 59580, "epoch": 669, "lr": 8.872005504235005e-05}
{"train_loss": 0.28629451990127563, "global_step": 59581, "epoch": 669, "lr": 8.871968824330992e-05}
{"train_loss": 0.24203263223171234, "global_step": 59582, "epoch": 669, "lr": 8.871932143906439e-05}
{"train_loss": 0.22138763964176178, "global_step": 59583, "epoch": 669, "lr": 8.871895462961353e-05}
{"train_loss": 0.27971410751342773, "global_step": 59584, "epoch": 669, "lr": 8.871858781495737e-05}
{"train_loss": 0.1867114007472992, "global_step": 59585, "epoch": 669, "lr": 8.871822099509598e-05}
{"train_loss": 0.23841740190982819, "global_step": 59586, "epoch": 669, "lr": 8.871785417002937e-05}
{"train_loss": 0.2921335697174072, "global_step": 59587, "epoch": 669, "lr": 8.871748733975762e-05}
{"train_loss": 0.23715737462043762, "global_step": 59588, "epoch": 669, "lr": 8.87171205042808e-05}
{"train_loss": 0.262986421585083, "global_step": 59589, "epoch": 669, "lr": 8.871675366359893e-05}
{"train_loss": 0.3272112011909485, "global_step": 59590, "epoch": 669, "lr": 8.871638681771205e-05}
{"train_loss": 0.14337681233882904, "global_step": 59591, "epoch": 669, "lr": 8.871601996662023e-05}
{"train_loss": 0.24900871515274048, "global_step": 59592, "epoch": 669, "lr": 8.871565311032352e-05}
{"train_loss": 0.19939832389354706, "global_step": 59593, "epoch": 669, "lr": 8.871528624882195e-05}
{"train_loss": 0.23183730244636536, "global_step": 59594, "epoch": 669, "lr": 8.87149193821156e-05}
{"train_loss": 0.22966600954532623, "global_step": 59595, "epoch": 669, "lr": 8.871455251020449e-05}
{"train_loss": 0.258565217256546, "global_step": 59596, "epoch": 669, "lr": 8.871418563308866e-05}
{"train_loss": 0.28786569833755493, "global_step": 59597, "epoch": 669, "lr": 8.871381875076821e-05}
{"train_loss": 0.1934773474931717, "global_step": 59598, "epoch": 669, "lr": 8.871345186324314e-05}
{"train_loss": 0.33821073174476624, "global_step": 59599, "epoch": 669, "lr": 8.871308497051354e-05}
{"train_loss": 0.15495632588863373, "global_step": 59600, "epoch": 669, "lr": 8.871271807257943e-05}
{"train_loss": 0.2353060394525528, "global_step": 59601, "epoch": 669, "lr": 8.871235116944086e-05}
{"train_loss": 0.21771517395973206, "global_step": 59602, "epoch": 669, "lr": 8.871198426109788e-05}
{"train_loss": 0.199270099401474, "global_step": 59603, "epoch": 669, "lr": 8.871161734755056e-05}
{"train_loss": 0.21303589642047882, "global_step": 59604, "epoch": 669, "lr": 8.871125042879893e-05}
{"train_loss": 0.18962107598781586, "global_step": 59605, "epoch": 669, "lr": 8.871088350484303e-05}
{"train_loss": 0.20847558975219727, "global_step": 59606, "epoch": 669, "lr": 8.871051657568294e-05}
{"train_loss": 0.21042150259017944, "global_step": 59607, "epoch": 669, "lr": 8.871014964131869e-05}
{"train_loss": 0.3016155958175659, "global_step": 59608, "epoch": 669, "lr": 8.870978270175032e-05}
{"train_loss": 0.3108022212982178, "global_step": 59609, "epoch": 669, "lr": 8.870941575697791e-05}
{"train_loss": 0.33273017406463623, "global_step": 59610, "epoch": 669, "lr": 8.870904880700148e-05}
{"train_loss": 0.2534920871257782, "global_step": 59611, "epoch": 669, "lr": 8.870868185182109e-05}
{"train_loss": 0.2637149691581726, "global_step": 59612, "epoch": 669, "lr": 8.870831489143681e-05}
{"train_loss": 0.2217205911874771, "global_step": 59613, "epoch": 669, "lr": 8.870794792584864e-05}
{"train_loss": 0.28333723545074463, "global_step": 59614, "epoch": 669, "lr": 8.870758095505667e-05}
{"train_loss": 0.34704092144966125, "global_step": 59615, "epoch": 669, "lr": 8.870721397906094e-05}
{"train_loss": 0.1953752189874649, "global_step": 59616, "epoch": 669, "lr": 8.87068469978615e-05}
{"train_loss": 0.2501327693462372, "global_step": 59617, "epoch": 669, "lr": 8.870648001145838e-05}
{"train_loss": 0.27193698287010193, "global_step": 59618, "epoch": 669, "lr": 8.870611301985166e-05}
{"train_loss": 0.27065691351890564, "global_step": 59619, "epoch": 669, "lr": 8.870574602304136e-05}
{"train_loss": 0.18589553236961365, "global_step": 59620, "epoch": 669, "lr": 8.870537902102756e-05}
{"train_loss": 0.1630076766014099, "global_step": 59621, "epoch": 669, "lr": 8.87050120138103e-05}
{"train_loss": 0.26774993538856506, "global_step": 59622, "epoch": 669, "lr": 8.87046450013896e-05}
{"train_loss": 0.34045541286468506, "global_step": 59623, "epoch": 669, "lr": 8.870427798376555e-05}
{"train_loss": 0.28000229597091675, "global_step": 59624, "epoch": 669, "lr": 8.870391096093817e-05}
{"train_loss": 0.26355934143066406, "global_step": 59625, "epoch": 669, "lr": 8.870354393290751e-05}
{"train_loss": 0.33854788541793823, "global_step": 59626, "epoch": 669, "lr": 8.870317689967367e-05}
{"train_loss": 0.3370688855648041, "global_step": 59627, "epoch": 669, "lr": 8.870280986123662e-05}
{"train_loss": 0.28554922342300415, "global_step": 59628, "epoch": 669, "lr": 8.870244281759648e-05}
{"train_loss": 0.2408283225605997, "global_step": 59629, "epoch": 669, "lr": 8.870207576875324e-05, "val_loss": 3.338178873062134}
{"train_loss": 0.17747488617897034, "global_step": 59630, "epoch": 670, "lr": 8.8701708714707e-05}
{"train_loss": 0.2514208257198334, "global_step": 59631, "epoch": 670, "lr": 8.870134165545778e-05}
{"train_loss": 0.25203239917755127, "global_step": 59632, "epoch": 670, "lr": 8.870097459100564e-05}
{"train_loss": 0.21455144882202148, "global_step": 59633, "epoch": 670, "lr": 8.870060752135061e-05}
{"train_loss": 0.25977572798728943, "global_step": 59634, "epoch": 670, "lr": 8.870024044649277e-05}
{"train_loss": 0.17431220412254333, "global_step": 59635, "epoch": 670, "lr": 8.869987336643214e-05}
{"train_loss": 0.18580393493175507, "global_step": 59636, "epoch": 670, "lr": 8.869950628116879e-05}
{"train_loss": 0.24924373626708984, "global_step": 59637, "epoch": 670, "lr": 8.869913919070276e-05}
{"train_loss": 0.24582524597644806, "global_step": 59638, "epoch": 670, "lr": 8.869877209503412e-05}
{"train_loss": 0.18579736351966858, "global_step": 59639, "epoch": 670, "lr": 8.869840499416289e-05}
{"train_loss": 0.29602736234664917, "global_step": 59640, "epoch": 670, "lr": 8.869803788808912e-05}
{"train_loss": 0.23538453876972198, "global_step": 59641, "epoch": 670, "lr": 8.86976707768129e-05}
{"train_loss": 0.30853331089019775, "global_step": 59642, "epoch": 670, "lr": 8.869730366033422e-05}
{"train_loss": 0.2620793879032135, "global_step": 59643, "epoch": 670, "lr": 8.869693653865318e-05}
{"train_loss": 0.19154584407806396, "global_step": 59644, "epoch": 670, "lr": 8.869656941176979e-05}
{"train_loss": 0.1766510307788849, "global_step": 59645, "epoch": 670, "lr": 8.869620227968412e-05}
{"train_loss": 0.23976179957389832, "global_step": 59646, "epoch": 670, "lr": 8.869583514239623e-05}
{"train_loss": 0.2494412511587143, "global_step": 59647, "epoch": 670, "lr": 8.869546799990615e-05}
{"train_loss": 0.2342798262834549, "global_step": 59648, "epoch": 670, "lr": 8.869510085221394e-05}
{"train_loss": 0.34411898255348206, "global_step": 59649, "epoch": 670, "lr": 8.869473369931964e-05}
{"train_loss": 0.20151196420192719, "global_step": 59650, "epoch": 670, "lr": 8.869436654122332e-05}
{"train_loss": 0.26988592743873596, "global_step": 59651, "epoch": 670, "lr": 8.869399937792499e-05}
{"train_loss": 0.22368262708187103, "global_step": 59652, "epoch": 670, "lr": 8.869363220942473e-05}
{"train_loss": 0.21466316282749176, "global_step": 59653, "epoch": 670, "lr": 8.869326503572259e-05}
{"train_loss": 0.17956161499023438, "global_step": 59654, "epoch": 670, "lr": 8.869289785681861e-05}
{"train_loss": 0.2332279533147812, "global_step": 59655, "epoch": 670, "lr": 8.869253067271284e-05}
{"train_loss": 0.20533747971057892, "global_step": 59656, "epoch": 670, "lr": 8.869216348340533e-05}
{"train_loss": 0.1606978327035904, "global_step": 59657, "epoch": 670, "lr": 8.869179628889613e-05}
{"train_loss": 0.16778695583343506, "global_step": 59658, "epoch": 670, "lr": 8.86914290891853e-05}
{"train_loss": 0.21399235725402832, "global_step": 59659, "epoch": 670, "lr": 8.869106188427287e-05}
{"train_loss": 0.1772664189338684, "global_step": 59660, "epoch": 670, "lr": 8.869069467415889e-05}
{"train_loss": 0.17750221490859985, "global_step": 59661, "epoch": 670, "lr": 8.869032745884343e-05}
{"train_loss": 0.23922976851463318, "global_step": 59662, "epoch": 670, "lr": 8.868996023832652e-05}
{"train_loss": 0.26614806056022644, "global_step": 59663, "epoch": 670, "lr": 8.868959301260822e-05}
{"train_loss": 0.23342308402061462, "global_step": 59664, "epoch": 670, "lr": 8.868922578168858e-05}
{"train_loss": 0.12116249650716782, "global_step": 59665, "epoch": 670, "lr": 8.868885854556764e-05}
{"train_loss": 0.19852429628372192, "global_step": 59666, "epoch": 670, "lr": 8.868849130424546e-05}
{"train_loss": 0.32898926734924316, "global_step": 59667, "epoch": 670, "lr": 8.868812405772206e-05}
{"train_loss": 0.2657076120376587, "global_step": 59668, "epoch": 670, "lr": 8.868775680599755e-05}
{"train_loss": 0.22672145068645477, "global_step": 59669, "epoch": 670, "lr": 8.868738954907191e-05}
{"train_loss": 0.20787669718265533, "global_step": 59670, "epoch": 670, "lr": 8.868702228694525e-05}
{"train_loss": 0.29834526777267456, "global_step": 59671, "epoch": 670, "lr": 8.868665501961758e-05}
{"train_loss": 0.1898133009672165, "global_step": 59672, "epoch": 670, "lr": 8.868628774708896e-05}
{"train_loss": 0.30202844738960266, "global_step": 59673, "epoch": 670, "lr": 8.868592046935943e-05}
{"train_loss": 0.2896682322025299, "global_step": 59674, "epoch": 670, "lr": 8.868555318642906e-05}
{"train_loss": 0.2824020981788635, "global_step": 59675, "epoch": 670, "lr": 8.868518589829789e-05}
{"train_loss": 0.253137469291687, "global_step": 59676, "epoch": 670, "lr": 8.868481860496595e-05}
{"train_loss": 0.2084253579378128, "global_step": 59677, "epoch": 670, "lr": 8.868445130643334e-05}
{"train_loss": 0.16476556658744812, "global_step": 59678, "epoch": 670, "lr": 8.868408400270005e-05}
{"train_loss": 0.22633317112922668, "global_step": 59679, "epoch": 670, "lr": 8.868371669376619e-05}
{"train_loss": 0.2203207015991211, "global_step": 59680, "epoch": 670, "lr": 8.868334937963174e-05}
{"train_loss": 0.19849897921085358, "global_step": 59681, "epoch": 670, "lr": 8.868298206029679e-05}
{"train_loss": 0.2684761881828308, "global_step": 59682, "epoch": 670, "lr": 8.86826147357614e-05}
{"train_loss": 0.31166791915893555, "global_step": 59683, "epoch": 670, "lr": 8.86822474060256e-05}
{"train_loss": 0.32612144947052, "global_step": 59684, "epoch": 670, "lr": 8.868188007108944e-05}
{"train_loss": 0.206309974193573, "global_step": 59685, "epoch": 670, "lr": 8.868151273095296e-05}
{"train_loss": 0.27421748638153076, "global_step": 59686, "epoch": 670, "lr": 8.868114538561624e-05}
{"train_loss": 0.1849069446325302, "global_step": 59687, "epoch": 670, "lr": 8.86807780350793e-05}
{"train_loss": 0.18013468384742737, "global_step": 59688, "epoch": 670, "lr": 8.86804106793422e-05}
{"train_loss": 0.27183428406715393, "global_step": 59689, "epoch": 670, "lr": 8.8680043318405e-05}
{"train_loss": 0.24813076853752136, "global_step": 59690, "epoch": 670, "lr": 8.867967595226773e-05}
{"train_loss": 0.1808258444070816, "global_step": 59691, "epoch": 670, "lr": 8.867930858093047e-05}
{"train_loss": 0.27673831582069397, "global_step": 59692, "epoch": 670, "lr": 8.867894120439324e-05}
{"train_loss": 0.23273368179798126, "global_step": 59693, "epoch": 670, "lr": 8.867857382265609e-05}
{"train_loss": 0.1824667602777481, "global_step": 59694, "epoch": 670, "lr": 8.867820643571907e-05}
{"train_loss": 0.22820144891738892, "global_step": 59695, "epoch": 670, "lr": 8.867783904358224e-05}
{"train_loss": 0.2884707450866699, "global_step": 59696, "epoch": 670, "lr": 8.867747164624566e-05}
{"train_loss": 0.21201038360595703, "global_step": 59697, "epoch": 670, "lr": 8.867710424370936e-05}
{"train_loss": 0.1912236511707306, "global_step": 59698, "epoch": 670, "lr": 8.86767368359734e-05}
{"train_loss": 0.2806532382965088, "global_step": 59699, "epoch": 670, "lr": 8.86763694230378e-05}
{"train_loss": 0.2630113363265991, "global_step": 59700, "epoch": 670, "lr": 8.867600200490265e-05}
{"train_loss": 0.24750205874443054, "global_step": 59701, "epoch": 670, "lr": 8.867563458156798e-05}
{"train_loss": 0.23762522637844086, "global_step": 59702, "epoch": 670, "lr": 8.867526715303385e-05}
{"train_loss": 0.19680815935134888, "global_step": 59703, "epoch": 670, "lr": 8.86748997193003e-05}
{"train_loss": 0.17139258980751038, "global_step": 59704, "epoch": 670, "lr": 8.867453228036738e-05}
{"train_loss": 0.20503996312618256, "global_step": 59705, "epoch": 670, "lr": 8.867416483623514e-05}
{"train_loss": 0.2507571578025818, "global_step": 59706, "epoch": 670, "lr": 8.867379738690363e-05}
{"train_loss": 0.22849999368190765, "global_step": 59707, "epoch": 670, "lr": 8.867342993237289e-05}
{"train_loss": 0.17250296473503113, "global_step": 59708, "epoch": 670, "lr": 8.867306247264298e-05}
{"train_loss": 0.260321706533432, "global_step": 59709, "epoch": 670, "lr": 8.867269500771396e-05}
{"train_loss": 0.3768559396266937, "global_step": 59710, "epoch": 670, "lr": 8.867232753758585e-05}
{"train_loss": 0.26456326246261597, "global_step": 59711, "epoch": 670, "lr": 8.867196006225873e-05}
{"train_loss": 0.22256386280059814, "global_step": 59712, "epoch": 670, "lr": 8.867159258173263e-05}
{"train_loss": 0.2851060926914215, "global_step": 59713, "epoch": 670, "lr": 8.867122509600762e-05}
{"train_loss": 0.2630390524864197, "global_step": 59714, "epoch": 670, "lr": 8.867085760508373e-05}
{"train_loss": 0.19440224766731262, "global_step": 59715, "epoch": 670, "lr": 8.8670490108961e-05}
{"train_loss": 0.1554935723543167, "global_step": 59716, "epoch": 670, "lr": 8.867012260763951e-05}
{"train_loss": 0.2908182442188263, "global_step": 59717, "epoch": 670, "lr": 8.866975510111927e-05}
{"train_loss": 0.2329164483071713, "global_step": 59718, "epoch": 670, "lr": 8.866938758940038e-05, "val_loss": 3.4214470386505127, "train_action_mse_error": 27.469375610351562}
{"train_loss": 0.2756318151950836, "global_step": 59719, "epoch": 671, "lr": 8.866902007248286e-05}
{"train_loss": 0.17064310610294342, "global_step": 59720, "epoch": 671, "lr": 8.866865255036675e-05}
{"train_loss": 0.2234210968017578, "global_step": 59721, "epoch": 671, "lr": 8.866828502305212e-05}
{"train_loss": 0.30172300338745117, "global_step": 59722, "epoch": 671, "lr": 8.8667917490539e-05}
{"train_loss": 0.24519337713718414, "global_step": 59723, "epoch": 671, "lr": 8.866754995282744e-05}
{"train_loss": 0.16643685102462769, "global_step": 59724, "epoch": 671, "lr": 8.866718240991754e-05}
{"train_loss": 0.27863121032714844, "global_step": 59725, "epoch": 671, "lr": 8.866681486180928e-05}
{"train_loss": 0.24047094583511353, "global_step": 59726, "epoch": 671, "lr": 8.866644730850273e-05}
{"train_loss": 0.1765580028295517, "global_step": 59727, "epoch": 671, "lr": 8.866607974999797e-05}
{"train_loss": 0.2473645806312561, "global_step": 59728, "epoch": 671, "lr": 8.8665712186295e-05}
{"train_loss": 0.23156292736530304, "global_step": 59729, "epoch": 671, "lr": 8.866534461739391e-05}
{"train_loss": 0.2644769251346588, "global_step": 59730, "epoch": 671, "lr": 8.866497704329476e-05}
{"train_loss": 0.336238831281662, "global_step": 59731, "epoch": 671, "lr": 8.866460946399753e-05}
{"train_loss": 0.3153010606765747, "global_step": 59732, "epoch": 671, "lr": 8.866424187950233e-05}
{"train_loss": 0.24868202209472656, "global_step": 59733, "epoch": 671, "lr": 8.866387428980921e-05}
{"train_loss": 0.29321086406707764, "global_step": 59734, "epoch": 671, "lr": 8.86635066949182e-05}
{"train_loss": 0.28550177812576294, "global_step": 59735, "epoch": 671, "lr": 8.866313909482934e-05}
{"train_loss": 0.25054067373275757, "global_step": 59736, "epoch": 671, "lr": 8.86627714895427e-05}
{"train_loss": 0.2867986261844635, "global_step": 59737, "epoch": 671, "lr": 8.866240387905832e-05}
{"train_loss": 0.2243562638759613, "global_step": 59738, "epoch": 671, "lr": 8.866203626337625e-05}
{"train_loss": 0.16013866662979126, "global_step": 59739, "epoch": 671, "lr": 8.866166864249654e-05}
{"train_loss": 0.24175821244716644, "global_step": 59740, "epoch": 671, "lr": 8.866130101641925e-05}
{"train_loss": 0.24768304824829102, "global_step": 59741, "epoch": 671, "lr": 8.866093338514441e-05}
{"train_loss": 0.2321126013994217, "global_step": 59742, "epoch": 671, "lr": 8.866056574867208e-05}
{"train_loss": 0.24552763998508453, "global_step": 59743, "epoch": 671, "lr": 8.86601981070023e-05}
{"train_loss": 0.21082647144794464, "global_step": 59744, "epoch": 671, "lr": 8.865983046013514e-05}
{"train_loss": 0.2124844789505005, "global_step": 59745, "epoch": 671, "lr": 8.865946280807063e-05}
{"train_loss": 0.2383735030889511, "global_step": 59746, "epoch": 671, "lr": 8.865909515080884e-05}
{"train_loss": 0.27253830432891846, "global_step": 59747, "epoch": 671, "lr": 8.865872748834978e-05}
{"train_loss": 0.2675400376319885, "global_step": 59748, "epoch": 671, "lr": 8.865835982069355e-05}
{"train_loss": 0.14135898649692535, "global_step": 59749, "epoch": 671, "lr": 8.865799214784016e-05}
{"train_loss": 0.2602311670780182, "global_step": 59750, "epoch": 671, "lr": 8.865762446978968e-05}
{"train_loss": 0.22001412510871887, "global_step": 59751, "epoch": 671, "lr": 8.865725678654216e-05}
{"train_loss": 0.23512855172157288, "global_step": 59752, "epoch": 671, "lr": 8.865688909809762e-05}
{"train_loss": 0.22111445665359497, "global_step": 59753, "epoch": 671, "lr": 8.865652140445615e-05}
{"train_loss": 0.23971301317214966, "global_step": 59754, "epoch": 671, "lr": 8.865615370561779e-05}
{"train_loss": 0.24282148480415344, "global_step": 59755, "epoch": 671, "lr": 8.865578600158257e-05}
{"train_loss": 0.1694749891757965, "global_step": 59756, "epoch": 671, "lr": 8.865541829235054e-05}
{"train_loss": 0.2959522604942322, "global_step": 59757, "epoch": 671, "lr": 8.865505057792179e-05}
{"train_loss": 0.41858676075935364, "global_step": 59758, "epoch": 671, "lr": 8.86546828582963e-05}
{"train_loss": 0.18005278706550598, "global_step": 59759, "epoch": 671, "lr": 8.865431513347419e-05}
{"train_loss": 0.3117827773094177, "global_step": 59760, "epoch": 671, "lr": 8.865394740345546e-05}
{"train_loss": 0.24608570337295532, "global_step": 59761, "epoch": 671, "lr": 8.86535796682402e-05}
{"train_loss": 0.2962847650051117, "global_step": 59762, "epoch": 671, "lr": 8.865321192782842e-05}
{"train_loss": 0.19578365981578827, "global_step": 59763, "epoch": 671, "lr": 8.865284418222019e-05}
{"train_loss": 0.25070667266845703, "global_step": 59764, "epoch": 671, "lr": 8.865247643141555e-05}
{"train_loss": 0.2261214703321457, "global_step": 59765, "epoch": 671, "lr": 8.865210867541456e-05}
{"train_loss": 0.30277666449546814, "global_step": 59766, "epoch": 671, "lr": 8.865174091421727e-05}
{"train_loss": 0.25928837060928345, "global_step": 59767, "epoch": 671, "lr": 8.865137314782372e-05}
{"train_loss": 0.271332323551178, "global_step": 59768, "epoch": 671, "lr": 8.865100537623397e-05}
{"train_loss": 0.26213735342025757, "global_step": 59769, "epoch": 671, "lr": 8.865063759944807e-05}
{"train_loss": 0.2058289498090744, "global_step": 59770, "epoch": 671, "lr": 8.865026981746605e-05}
{"train_loss": 0.26076391339302063, "global_step": 59771, "epoch": 671, "lr": 8.864990203028797e-05}
{"train_loss": 0.19768624007701874, "global_step": 59772, "epoch": 671, "lr": 8.864953423791389e-05}
{"train_loss": 0.1503281444311142, "global_step": 59773, "epoch": 671, "lr": 8.864916644034385e-05}
{"train_loss": 0.26956018805503845, "global_step": 59774, "epoch": 671, "lr": 8.864879863757789e-05}
{"train_loss": 0.19078893959522247, "global_step": 59775, "epoch": 671, "lr": 8.864843082961608e-05}
{"train_loss": 0.2553461194038391, "global_step": 59776, "epoch": 671, "lr": 8.864806301645846e-05}
{"train_loss": 0.24094511568546295, "global_step": 59777, "epoch": 671, "lr": 8.864769519810508e-05}
{"train_loss": 0.26735782623291016, "global_step": 59778, "epoch": 671, "lr": 8.864732737455598e-05}
{"train_loss": 0.2992287576198578, "global_step": 59779, "epoch": 671, "lr": 8.864695954581122e-05}
{"train_loss": 0.23918813467025757, "global_step": 59780, "epoch": 671, "lr": 8.864659171187086e-05}
{"train_loss": 0.28076350688934326, "global_step": 59781, "epoch": 671, "lr": 8.864622387273492e-05}
{"train_loss": 0.23286473751068115, "global_step": 59782, "epoch": 671, "lr": 8.864585602840348e-05}
{"train_loss": 0.20478183031082153, "global_step": 59783, "epoch": 671, "lr": 8.864548817887657e-05}
{"train_loss": 0.28556928038597107, "global_step": 59784, "epoch": 671, "lr": 8.864512032415423e-05}
{"train_loss": 0.2240777611732483, "global_step": 59785, "epoch": 671, "lr": 8.864475246423655e-05}
{"train_loss": 0.3496801555156708, "global_step": 59786, "epoch": 671, "lr": 8.864438459912354e-05}
{"train_loss": 0.3399381935596466, "global_step": 59787, "epoch": 671, "lr": 8.864401672881526e-05}
{"train_loss": 0.3031562268733978, "global_step": 59788, "epoch": 671, "lr": 8.864364885331178e-05}
{"train_loss": 0.15900519490242004, "global_step": 59789, "epoch": 671, "lr": 8.864328097261312e-05}
{"train_loss": 0.23516610264778137, "global_step": 59790, "epoch": 671, "lr": 8.864291308671933e-05}
{"train_loss": 0.22693848609924316, "global_step": 59791, "epoch": 671, "lr": 8.86425451956305e-05}
{"train_loss": 0.20521262288093567, "global_step": 59792, "epoch": 671, "lr": 8.864217729934663e-05}
{"train_loss": 0.3236406743526459, "global_step": 59793, "epoch": 671, "lr": 8.864180939786779e-05}
{"train_loss": 0.4725121557712555, "global_step": 59794, "epoch": 671, "lr": 8.864144149119405e-05}
{"train_loss": 0.3442562222480774, "global_step": 59795, "epoch": 671, "lr": 8.864107357932542e-05}
{"train_loss": 0.16475819051265717, "global_step": 59796, "epoch": 671, "lr": 8.864070566226197e-05}
{"train_loss": 0.3177172541618347, "global_step": 59797, "epoch": 671, "lr": 8.864033774000375e-05}
{"train_loss": 0.37729379534721375, "global_step": 59798, "epoch": 671, "lr": 8.86399698125508e-05}
{"train_loss": 0.29243728518486023, "global_step": 59799, "epoch": 671, "lr": 8.86396018799032e-05}
{"train_loss": 0.2553858458995819, "global_step": 59800, "epoch": 671, "lr": 8.863923394206096e-05}
{"train_loss": 0.2549942135810852, "global_step": 59801, "epoch": 671, "lr": 8.863886599902416e-05}
{"train_loss": 0.31106388568878174, "global_step": 59802, "epoch": 671, "lr": 8.863849805079281e-05}
{"train_loss": 0.22060459852218628, "global_step": 59803, "epoch": 671, "lr": 8.863813009736702e-05}
{"train_loss": 0.31473642587661743, "global_step": 59804, "epoch": 671, "lr": 8.863776213874678e-05}
{"train_loss": 0.210377037525177, "global_step": 59805, "epoch": 671, "lr": 8.863739417493215e-05}
{"train_loss": 0.28296905755996704, "global_step": 59806, "epoch": 671, "lr": 8.863702620592321e-05}
{"train_loss": 0.25490657868010275, "global_step": 59807, "epoch": 671, "lr": 8.863665823172e-05, "val_loss": 3.338139772415161}
{"train_loss": 0.3229302763938904, "global_step": 59808, "epoch": 672, "lr": 8.863629025232255e-05}
{"train_loss": 0.2874578535556793, "global_step": 59809, "epoch": 672, "lr": 8.863592226773093e-05}
{"train_loss": 0.3417547643184662, "global_step": 59810, "epoch": 672, "lr": 8.863555427794518e-05}
{"train_loss": 0.27512648701667786, "global_step": 59811, "epoch": 672, "lr": 8.863518628296534e-05}
{"train_loss": 0.18953552842140198, "global_step": 59812, "epoch": 672, "lr": 8.86348182827915e-05}
{"train_loss": 0.2045527845621109, "global_step": 59813, "epoch": 672, "lr": 8.863445027742366e-05}
{"train_loss": 0.2869340479373932, "global_step": 59814, "epoch": 672, "lr": 8.863408226686187e-05}
{"train_loss": 0.26831698417663574, "global_step": 59815, "epoch": 672, "lr": 8.863371425110622e-05}
{"train_loss": 0.1912255436182022, "global_step": 59816, "epoch": 672, "lr": 8.863334623015673e-05}
{"train_loss": 0.2227713167667389, "global_step": 59817, "epoch": 672, "lr": 8.863297820401347e-05}
{"train_loss": 0.3800874650478363, "global_step": 59818, "epoch": 672, "lr": 8.863261017267645e-05}
{"train_loss": 0.15312498807907104, "global_step": 59819, "epoch": 672, "lr": 8.863224213614578e-05}
{"train_loss": 0.22745175659656525, "global_step": 59820, "epoch": 672, "lr": 8.863187409442146e-05}
{"train_loss": 0.2746124267578125, "global_step": 59821, "epoch": 672, "lr": 8.863150604750356e-05}
{"train_loss": 0.22556720674037933, "global_step": 59822, "epoch": 672, "lr": 8.863113799539212e-05}
{"train_loss": 0.24144352972507477, "global_step": 59823, "epoch": 672, "lr": 8.863076993808719e-05}
{"train_loss": 0.29677966237068176, "global_step": 59824, "epoch": 672, "lr": 8.863040187558883e-05}
{"train_loss": 0.2648460268974304, "global_step": 59825, "epoch": 672, "lr": 8.863003380789708e-05}
{"train_loss": 0.23263953626155853, "global_step": 59826, "epoch": 672, "lr": 8.8629665735012e-05}
{"train_loss": 0.17492182552814484, "global_step": 59827, "epoch": 672, "lr": 8.862929765693361e-05}
{"train_loss": 0.3240381181240082, "global_step": 59828, "epoch": 672, "lr": 8.862892957366202e-05}
{"train_loss": 0.13430355489253998, "global_step": 59829, "epoch": 672, "lr": 8.862856148519721e-05}
{"train_loss": 0.20366327464580536, "global_step": 59830, "epoch": 672, "lr": 8.862819339153927e-05}
{"train_loss": 0.286002516746521, "global_step": 59831, "epoch": 672, "lr": 8.862782529268825e-05}
{"train_loss": 0.26669326424598694, "global_step": 59832, "epoch": 672, "lr": 8.862745718864418e-05}
{"train_loss": 0.15525245666503906, "global_step": 59833, "epoch": 672, "lr": 8.862708907940712e-05}
{"train_loss": 0.18252046406269073, "global_step": 59834, "epoch": 672, "lr": 8.862672096497712e-05}
{"train_loss": 0.26302289962768555, "global_step": 59835, "epoch": 672, "lr": 8.862635284535423e-05}
{"train_loss": 0.2548556327819824, "global_step": 59836, "epoch": 672, "lr": 8.86259847205385e-05}
{"train_loss": 0.2969040870666504, "global_step": 59837, "epoch": 672, "lr": 8.862561659052998e-05}
{"train_loss": 0.1703532487154007, "global_step": 59838, "epoch": 672, "lr": 8.862524845532868e-05}
{"train_loss": 0.25790396332740784, "global_step": 59839, "epoch": 672, "lr": 8.862488031493473e-05}
{"train_loss": 0.2866514325141907, "global_step": 59840, "epoch": 672, "lr": 8.862451216934812e-05}
{"train_loss": 0.2306591272354126, "global_step": 59841, "epoch": 672, "lr": 8.86241440185689e-05}
{"train_loss": 0.23576857149600983, "global_step": 59842, "epoch": 672, "lr": 8.862377586259716e-05}
{"train_loss": 0.22425039112567902, "global_step": 59843, "epoch": 672, "lr": 8.862340770143291e-05}
{"train_loss": 0.3055137097835541, "global_step": 59844, "epoch": 672, "lr": 8.862303953507623e-05}
{"train_loss": 0.2371225357055664, "global_step": 59845, "epoch": 672, "lr": 8.862267136352713e-05}
{"train_loss": 0.25652140378952026, "global_step": 59846, "epoch": 672, "lr": 8.86223031867857e-05}
{"train_loss": 0.2155919373035431, "global_step": 59847, "epoch": 672, "lr": 8.862193500485196e-05}
{"train_loss": 0.23098556697368622, "global_step": 59848, "epoch": 672, "lr": 8.862156681772598e-05}
{"train_loss": 0.2925766110420227, "global_step": 59849, "epoch": 672, "lr": 8.86211986254078e-05}
{"train_loss": 0.26034754514694214, "global_step": 59850, "epoch": 672, "lr": 8.862083042789746e-05}
{"train_loss": 0.17641392350196838, "global_step": 59851, "epoch": 672, "lr": 8.862046222519504e-05}
{"train_loss": 0.2664251923561096, "global_step": 59852, "epoch": 672, "lr": 8.862009401730057e-05}
{"train_loss": 0.19825534522533417, "global_step": 59853, "epoch": 672, "lr": 8.861972580421407e-05}
{"train_loss": 0.21373599767684937, "global_step": 59854, "epoch": 672, "lr": 8.861935758593563e-05}
{"train_loss": 0.24220915138721466, "global_step": 59855, "epoch": 672, "lr": 8.861898936246531e-05}
{"train_loss": 0.19170115888118744, "global_step": 59856, "epoch": 672, "lr": 8.861862113380312e-05}
{"train_loss": 0.21144232153892517, "global_step": 59857, "epoch": 672, "lr": 8.861825289994912e-05}
{"train_loss": 0.2627739906311035, "global_step": 59858, "epoch": 672, "lr": 8.861788466090337e-05}
{"train_loss": 0.2327098250389099, "global_step": 59859, "epoch": 672, "lr": 8.861751641666594e-05}
{"train_loss": 0.30283093452453613, "global_step": 59860, "epoch": 672, "lr": 8.861714816723683e-05}
{"train_loss": 0.2777179181575775, "global_step": 59861, "epoch": 672, "lr": 8.861677991261612e-05}
{"train_loss": 0.20748819410800934, "global_step": 59862, "epoch": 672, "lr": 8.861641165280385e-05}
{"train_loss": 0.3092902898788452, "global_step": 59863, "epoch": 672, "lr": 8.861604338780009e-05}
{"train_loss": 0.2781996726989746, "global_step": 59864, "epoch": 672, "lr": 8.861567511760486e-05}
{"train_loss": 0.2750582695007324, "global_step": 59865, "epoch": 672, "lr": 8.861530684221822e-05}
{"train_loss": 0.2577502727508545, "global_step": 59866, "epoch": 672, "lr": 8.861493856164023e-05}
{"train_loss": 0.2238163948059082, "global_step": 59867, "epoch": 672, "lr": 8.861457027587095e-05}
{"train_loss": 0.28505823016166687, "global_step": 59868, "epoch": 672, "lr": 8.861420198491038e-05}
{"train_loss": 0.2548717260360718, "global_step": 59869, "epoch": 672, "lr": 8.861383368875861e-05}
{"train_loss": 0.1798534393310547, "global_step": 59870, "epoch": 672, "lr": 8.86134653874157e-05}
{"train_loss": 0.2357996702194214, "global_step": 59871, "epoch": 672, "lr": 8.861309708088166e-05}
{"train_loss": 0.18135680258274078, "global_step": 59872, "epoch": 672, "lr": 8.861272876915658e-05}
{"train_loss": 0.2246243953704834, "global_step": 59873, "epoch": 672, "lr": 8.861236045224047e-05}
{"train_loss": 0.18353013694286346, "global_step": 59874, "epoch": 672, "lr": 8.86119921301334e-05}
{"train_loss": 0.2803986668586731, "global_step": 59875, "epoch": 672, "lr": 8.861162380283542e-05}
{"train_loss": 0.2136106640100479, "global_step": 59876, "epoch": 672, "lr": 8.861125547034659e-05}
{"train_loss": 0.3259620666503906, "global_step": 59877, "epoch": 672, "lr": 8.861088713266694e-05}
{"train_loss": 0.2002229541540146, "global_step": 59878, "epoch": 672, "lr": 8.861051878979652e-05}
{"train_loss": 0.20327487587928772, "global_step": 59879, "epoch": 672, "lr": 8.861015044173538e-05}
{"train_loss": 0.2206796556711197, "global_step": 59880, "epoch": 672, "lr": 8.860978208848358e-05}
{"train_loss": 0.2940976917743683, "global_step": 59881, "epoch": 672, "lr": 8.860941373004118e-05}
{"train_loss": 0.29459425806999207, "global_step": 59882, "epoch": 672, "lr": 8.86090453664082e-05}
{"train_loss": 0.1939275711774826, "global_step": 59883, "epoch": 672, "lr": 8.860867699758472e-05}
{"train_loss": 0.2229420244693756, "global_step": 59884, "epoch": 672, "lr": 8.860830862357077e-05}
{"train_loss": 0.30231162905693054, "global_step": 59885, "epoch": 672, "lr": 8.860794024436639e-05}
{"train_loss": 0.2100009322166443, "global_step": 59886, "epoch": 672, "lr": 8.860757185997166e-05}
{"train_loss": 0.34467408061027527, "global_step": 59887, "epoch": 672, "lr": 8.86072034703866e-05}
{"train_loss": 0.30006957054138184, "global_step": 59888, "epoch": 672, "lr": 8.860683507561127e-05}
{"train_loss": 0.24099035561084747, "global_step": 59889, "epoch": 672, "lr": 8.860646667564574e-05}
{"train_loss": 0.30554884672164917, "global_step": 59890, "epoch": 672, "lr": 8.860609827049002e-05}
{"train_loss": 0.30894362926483154, "global_step": 59891, "epoch": 672, "lr": 8.860572986014419e-05}
{"train_loss": 0.22319796681404114, "global_step": 59892, "epoch": 672, "lr": 8.860536144460829e-05}
{"train_loss": 0.2840900719165802, "global_step": 59893, "epoch": 672, "lr": 8.860499302388237e-05}
{"train_loss": 0.3461286127567291, "global_step": 59894, "epoch": 672, "lr": 8.860462459796648e-05}
{"train_loss": 0.29968300461769104, "global_step": 59895, "epoch": 672, "lr": 8.860425616686067e-05}
{"train_loss": 0.24991212182500389, "global_step": 59896, "epoch": 672, "lr": 8.860388773056498e-05, "val_loss": 3.412775754928589}
{"train_loss": 0.20246487855911255, "global_step": 59897, "epoch": 673, "lr": 8.860351928907948e-05}
{"train_loss": 0.3725683093070984, "global_step": 59898, "epoch": 673, "lr": 8.860315084240419e-05}
{"train_loss": 0.16178758442401886, "global_step": 59899, "epoch": 673, "lr": 8.860278239053919e-05}
{"train_loss": 0.29877132177352905, "global_step": 59900, "epoch": 673, "lr": 8.860241393348452e-05}
{"train_loss": 0.2825808823108673, "global_step": 59901, "epoch": 673, "lr": 8.860204547124022e-05}
{"train_loss": 0.3653545677661896, "global_step": 59902, "epoch": 673, "lr": 8.860167700380634e-05}
{"train_loss": 0.25504836440086365, "global_step": 59903, "epoch": 673, "lr": 8.860130853118293e-05}
{"train_loss": 0.20913781225681305, "global_step": 59904, "epoch": 673, "lr": 8.860094005337008e-05}
{"train_loss": 0.2582763135433197, "global_step": 59905, "epoch": 673, "lr": 8.860057157036776e-05}
{"train_loss": 0.2745812237262726, "global_step": 59906, "epoch": 673, "lr": 8.860020308217609e-05}
{"train_loss": 0.24773140251636505, "global_step": 59907, "epoch": 673, "lr": 8.859983458879508e-05}
{"train_loss": 0.2589510977268219, "global_step": 59908, "epoch": 673, "lr": 8.85994660902248e-05}
{"train_loss": 0.30555951595306396, "global_step": 59909, "epoch": 673, "lr": 8.85990975864653e-05}
{"train_loss": 0.20623324811458588, "global_step": 59910, "epoch": 673, "lr": 8.85987290775166e-05}
{"train_loss": 0.26837021112442017, "global_step": 59911, "epoch": 673, "lr": 8.859836056337879e-05}
{"train_loss": 0.145282581448555, "global_step": 59912, "epoch": 673, "lr": 8.859799204405189e-05}
{"train_loss": 0.16878676414489746, "global_step": 59913, "epoch": 673, "lr": 8.859762351953598e-05}
{"train_loss": 0.27608659863471985, "global_step": 59914, "epoch": 673, "lr": 8.859725498983107e-05}
{"train_loss": 0.17336350679397583, "global_step": 59915, "epoch": 673, "lr": 8.859688645493722e-05}
{"train_loss": 0.26375842094421387, "global_step": 59916, "epoch": 673, "lr": 8.859651791485452e-05}
{"train_loss": 0.2682965397834778, "global_step": 59917, "epoch": 673, "lr": 8.859614936958296e-05}
{"train_loss": 0.20341750979423523, "global_step": 59918, "epoch": 673, "lr": 8.859578081912263e-05}
{"train_loss": 0.2425510287284851, "global_step": 59919, "epoch": 673, "lr": 8.859541226347358e-05}
{"train_loss": 0.21661286056041718, "global_step": 59920, "epoch": 673, "lr": 8.859504370263583e-05}
{"train_loss": 0.23380351066589355, "global_step": 59921, "epoch": 673, "lr": 8.859467513660945e-05}
{"train_loss": 0.2552661597728729, "global_step": 59922, "epoch": 673, "lr": 8.859430656539449e-05}
{"train_loss": 0.17195044457912445, "global_step": 59923, "epoch": 673, "lr": 8.859393798899099e-05}
{"train_loss": 0.2993881106376648, "global_step": 59924, "epoch": 673, "lr": 8.859356940739901e-05}
{"train_loss": 0.22351425886154175, "global_step": 59925, "epoch": 673, "lr": 8.85932008206186e-05}
{"train_loss": 0.20875319838523865, "global_step": 59926, "epoch": 673, "lr": 8.85928322286498e-05}
{"train_loss": 0.29251012206077576, "global_step": 59927, "epoch": 673, "lr": 8.859246363149266e-05}
{"train_loss": 0.30835336446762085, "global_step": 59928, "epoch": 673, "lr": 8.859209502914724e-05}
{"train_loss": 0.22789236903190613, "global_step": 59929, "epoch": 673, "lr": 8.859172642161357e-05}
{"train_loss": 0.26479142904281616, "global_step": 59930, "epoch": 673, "lr": 8.859135780889173e-05}
{"train_loss": 0.25761860609054565, "global_step": 59931, "epoch": 673, "lr": 8.859098919098175e-05}
{"train_loss": 0.21425841748714447, "global_step": 59932, "epoch": 673, "lr": 8.859062056788368e-05}
{"train_loss": 0.15650802850723267, "global_step": 59933, "epoch": 673, "lr": 8.859025193959755e-05}
{"train_loss": 0.16888824105262756, "global_step": 59934, "epoch": 673, "lr": 8.858988330612345e-05}
{"train_loss": 0.22311906516551971, "global_step": 59935, "epoch": 673, "lr": 8.858951466746141e-05}
{"train_loss": 0.2656146287918091, "global_step": 59936, "epoch": 673, "lr": 8.858914602361148e-05}
{"train_loss": 0.2154577672481537, "global_step": 59937, "epoch": 673, "lr": 8.858877737457372e-05}
{"train_loss": 0.2647005319595337, "global_step": 59938, "epoch": 673, "lr": 8.858840872034816e-05}
{"train_loss": 0.24241293966770172, "global_step": 59939, "epoch": 673, "lr": 8.858804006093486e-05}
{"train_loss": 0.23836539685726166, "global_step": 59940, "epoch": 673, "lr": 8.858767139633386e-05}
{"train_loss": 0.22502784430980682, "global_step": 59941, "epoch": 673, "lr": 8.858730272654523e-05}
{"train_loss": 0.2248631715774536, "global_step": 59942, "epoch": 673, "lr": 8.8586934051569e-05}
{"train_loss": 0.22798492014408112, "global_step": 59943, "epoch": 673, "lr": 8.858656537140523e-05}
{"train_loss": 0.26314806938171387, "global_step": 59944, "epoch": 673, "lr": 8.858619668605397e-05}
{"train_loss": 0.26120635867118835, "global_step": 59945, "epoch": 673, "lr": 8.858582799551524e-05}
{"train_loss": 0.23040103912353516, "global_step": 59946, "epoch": 673, "lr": 8.858545929978915e-05}
{"train_loss": 0.23396408557891846, "global_step": 59947, "epoch": 673, "lr": 8.85850905988757e-05}
{"train_loss": 0.18461617827415466, "global_step": 59948, "epoch": 673, "lr": 8.858472189277496e-05}
{"train_loss": 0.23500265181064606, "global_step": 59949, "epoch": 673, "lr": 8.858435318148696e-05}
{"train_loss": 0.24281448125839233, "global_step": 59950, "epoch": 673, "lr": 8.858398446501178e-05}
{"train_loss": 0.22380129992961884, "global_step": 59951, "epoch": 673, "lr": 8.858361574334946e-05}
{"train_loss": 0.20396383106708527, "global_step": 59952, "epoch": 673, "lr": 8.858324701650003e-05}
{"train_loss": 0.2027241438627243, "global_step": 59953, "epoch": 673, "lr": 8.858287828446355e-05}
{"train_loss": 0.2973141372203827, "global_step": 59954, "epoch": 673, "lr": 8.858250954724008e-05}
{"train_loss": 0.25002700090408325, "global_step": 59955, "epoch": 673, "lr": 8.858214080482966e-05}
{"train_loss": 0.29012972116470337, "global_step": 59956, "epoch": 673, "lr": 8.858177205723235e-05}
{"train_loss": 0.16770559549331665, "global_step": 59957, "epoch": 673, "lr": 8.858140330444817e-05}
{"train_loss": 0.3066725432872772, "global_step": 59958, "epoch": 673, "lr": 8.858103454647721e-05}
{"train_loss": 0.21487827599048615, "global_step": 59959, "epoch": 673, "lr": 8.858066578331949e-05}
{"train_loss": 0.21607860922813416, "global_step": 59960, "epoch": 673, "lr": 8.858029701497507e-05}
{"train_loss": 0.2085387408733368, "global_step": 59961, "epoch": 673, "lr": 8.857992824144399e-05}
{"train_loss": 0.22933581471443176, "global_step": 59962, "epoch": 673, "lr": 8.857955946272633e-05}
{"train_loss": 0.2568744421005249, "global_step": 59963, "epoch": 673, "lr": 8.85791906788221e-05}
{"train_loss": 0.25716927647590637, "global_step": 59964, "epoch": 673, "lr": 8.85788218897314e-05}
{"train_loss": 0.27953189611434937, "global_step": 59965, "epoch": 673, "lr": 8.85784530954542e-05}
{"train_loss": 0.23660533130168915, "global_step": 59966, "epoch": 673, "lr": 8.857808429599064e-05}
{"train_loss": 0.25178852677345276, "global_step": 59967, "epoch": 673, "lr": 8.85777154913407e-05}
{"train_loss": 0.27489393949508667, "global_step": 59968, "epoch": 673, "lr": 8.857734668150446e-05}
{"train_loss": 0.22901304066181183, "global_step": 59969, "epoch": 673, "lr": 8.857697786648198e-05}
{"train_loss": 0.24953293800354004, "global_step": 59970, "epoch": 673, "lr": 8.857660904627327e-05}
{"train_loss": 0.2425096184015274, "global_step": 59971, "epoch": 673, "lr": 8.857624022087843e-05}
{"train_loss": 0.261066734790802, "global_step": 59972, "epoch": 673, "lr": 8.857587139029748e-05}
{"train_loss": 0.23888792097568512, "global_step": 59973, "epoch": 673, "lr": 8.857550255453046e-05}
{"train_loss": 0.24096669256687164, "global_step": 59974, "epoch": 673, "lr": 8.857513371357744e-05}
{"train_loss": 0.3323485851287842, "global_step": 59975, "epoch": 673, "lr": 8.857476486743848e-05}
{"train_loss": 0.2920489013195038, "global_step": 59976, "epoch": 673, "lr": 8.857439601611359e-05}
{"train_loss": 0.14969515800476074, "global_step": 59977, "epoch": 673, "lr": 8.857402715960285e-05}
{"train_loss": 0.21197660267353058, "global_step": 59978, "epoch": 673, "lr": 8.857365829790632e-05}
{"train_loss": 0.237949937582016, "global_step": 59979, "epoch": 673, "lr": 8.857328943102401e-05}
{"train_loss": 0.27111878991127014, "global_step": 59980, "epoch": 673, "lr": 8.8572920558956e-05}
{"train_loss": 0.29999929666519165, "global_step": 59981, "epoch": 673, "lr": 8.857255168170233e-05}
{"train_loss": 0.27163147926330566, "global_step": 59982, "epoch": 673, "lr": 8.857218279926305e-05}
{"train_loss": 0.2751973569393158, "global_step": 59983, "epoch": 673, "lr": 8.85718139116382e-05}
{"train_loss": 0.2914391756057739, "global_step": 59984, "epoch": 673, "lr": 8.857144501882786e-05}
{"train_loss": 0.2432144247414021, "global_step": 59985, "epoch": 673, "lr": 8.857107612083204e-05, "val_loss": 3.505793333053589}
{"train_loss": 0.22734032571315765, "global_step": 59986, "epoch": 674, "lr": 8.857070721765082e-05}
{"train_loss": 0.2290762960910797, "global_step": 59987, "epoch": 674, "lr": 8.857033830928423e-05}
{"train_loss": 0.2438572645187378, "global_step": 59988, "epoch": 674, "lr": 8.856996939573235e-05}
{"train_loss": 0.14565740525722504, "global_step": 59989, "epoch": 674, "lr": 8.856960047699518e-05}
{"train_loss": 0.30179131031036377, "global_step": 59990, "epoch": 674, "lr": 8.856923155307282e-05}
{"train_loss": 0.25653278827667236, "global_step": 59991, "epoch": 674, "lr": 8.856886262396528e-05}
{"train_loss": 0.15081720054149628, "global_step": 59992, "epoch": 674, "lr": 8.856849368967264e-05}
{"train_loss": 0.2195633053779602, "global_step": 59993, "epoch": 674, "lr": 8.856812475019493e-05}
{"train_loss": 0.2528524100780487, "global_step": 59994, "epoch": 674, "lr": 8.856775580553221e-05}
{"train_loss": 0.3024173974990845, "global_step": 59995, "epoch": 674, "lr": 8.856738685568451e-05}
{"train_loss": 0.2246086448431015, "global_step": 59996, "epoch": 674, "lr": 8.856701790065189e-05}
{"train_loss": 0.19499929249286652, "global_step": 59997, "epoch": 674, "lr": 8.856664894043442e-05}
{"train_loss": 0.181539848446846, "global_step": 59998, "epoch": 674, "lr": 8.856627997503215e-05}
{"train_loss": 0.18839439749717712, "global_step": 59999, "epoch": 674, "lr": 8.856591100444509e-05}
{"train_loss": 0.19192373752593994, "global_step": 60000, "epoch": 674, "lr": 8.856554202867332e-05}
{"train_loss": 0.19049568474292755, "global_step": 60001, "epoch": 674, "lr": 8.856517304771689e-05}
{"train_loss": 0.27461501955986023, "global_step": 60002, "epoch": 674, "lr": 8.856480406157584e-05}
{"train_loss": 0.19288361072540283, "global_step": 60003, "epoch": 674, "lr": 8.856443507025021e-05}
{"train_loss": 0.16059064865112305, "global_step": 60004, "epoch": 674, "lr": 8.856406607374008e-05}
{"train_loss": 0.1840599924325943, "global_step": 60005, "epoch": 674, "lr": 8.856369707204546e-05}
{"train_loss": 0.2772689163684845, "global_step": 60006, "epoch": 674, "lr": 8.856332806516644e-05}
{"train_loss": 0.274578332901001, "global_step": 60007, "epoch": 674, "lr": 8.856295905310303e-05}
{"train_loss": 0.22037285566329956, "global_step": 60008, "epoch": 674, "lr": 8.856259003585532e-05}
{"train_loss": 0.2258775234222412, "global_step": 60009, "epoch": 674, "lr": 8.856222101342333e-05}
{"train_loss": 0.24944323301315308, "global_step": 60010, "epoch": 674, "lr": 8.856185198580713e-05}
{"train_loss": 0.22787518799304962, "global_step": 60011, "epoch": 674, "lr": 8.856148295300675e-05}
{"train_loss": 0.20956973731517792, "global_step": 60012, "epoch": 674, "lr": 8.856111391502224e-05}
{"train_loss": 0.2203955501317978, "global_step": 60013, "epoch": 674, "lr": 8.856074487185368e-05}
{"train_loss": 0.24688920378684998, "global_step": 60014, "epoch": 674, "lr": 8.856037582350107e-05}
{"train_loss": 0.23007842898368835, "global_step": 60015, "epoch": 674, "lr": 8.856000676996452e-05}
{"train_loss": 0.2546381652355194, "global_step": 60016, "epoch": 674, "lr": 8.855963771124403e-05}
{"train_loss": 0.2281181812286377, "global_step": 60017, "epoch": 674, "lr": 8.855926864733966e-05}
{"train_loss": 0.23068977892398834, "global_step": 60018, "epoch": 674, "lr": 8.855889957825148e-05}
{"train_loss": 0.21368129551410675, "global_step": 60019, "epoch": 674, "lr": 8.85585305039795e-05}
{"train_loss": 0.28823110461235046, "global_step": 60020, "epoch": 674, "lr": 8.855816142452383e-05}
{"train_loss": 0.3076956570148468, "global_step": 60021, "epoch": 674, "lr": 8.855779233988447e-05}
{"train_loss": 0.24509888887405396, "global_step": 60022, "epoch": 674, "lr": 8.855742325006149e-05}
{"train_loss": 0.2973267436027527, "global_step": 60023, "epoch": 674, "lr": 8.855705415505492e-05}
{"train_loss": 0.24875327944755554, "global_step": 60024, "epoch": 674, "lr": 8.855668505486483e-05}
{"train_loss": 0.18794317543506622, "global_step": 60025, "epoch": 674, "lr": 8.855631594949127e-05}
{"train_loss": 0.26966506242752075, "global_step": 60026, "epoch": 674, "lr": 8.855594683893429e-05}
{"train_loss": 0.21428893506526947, "global_step": 60027, "epoch": 674, "lr": 8.855557772319392e-05}
{"train_loss": 0.32485735416412354, "global_step": 60028, "epoch": 674, "lr": 8.855520860227022e-05}
{"train_loss": 0.2667202949523926, "global_step": 60029, "epoch": 674, "lr": 8.855483947616326e-05}
{"train_loss": 0.24522550404071808, "global_step": 60030, "epoch": 674, "lr": 8.855447034487305e-05}
{"train_loss": 0.2454700618982315, "global_step": 60031, "epoch": 674, "lr": 8.855410120839966e-05}
{"train_loss": 0.2736675441265106, "global_step": 60032, "epoch": 674, "lr": 8.855373206674315e-05}
{"train_loss": 0.2203407883644104, "global_step": 60033, "epoch": 674, "lr": 8.855336291990357e-05}
{"train_loss": 0.2002921998500824, "global_step": 60034, "epoch": 674, "lr": 8.855299376788095e-05}
{"train_loss": 0.3045012950897217, "global_step": 60035, "epoch": 674, "lr": 8.855262461067534e-05}
{"train_loss": 0.27132704854011536, "global_step": 60036, "epoch": 674, "lr": 8.855225544828682e-05}
{"train_loss": 0.3154173791408539, "global_step": 60037, "epoch": 674, "lr": 8.855188628071543e-05}
{"train_loss": 0.3020769953727722, "global_step": 60038, "epoch": 674, "lr": 8.855151710796117e-05}
{"train_loss": 0.24511928856372833, "global_step": 60039, "epoch": 674, "lr": 8.855114793002416e-05}
{"train_loss": 0.16656973958015442, "global_step": 60040, "epoch": 674, "lr": 8.85507787469044e-05}
{"train_loss": 0.2547001540660858, "global_step": 60041, "epoch": 674, "lr": 8.855040955860197e-05}
{"train_loss": 0.2556491196155548, "global_step": 60042, "epoch": 674, "lr": 8.85500403651169e-05}
{"train_loss": 0.30246397852897644, "global_step": 60043, "epoch": 674, "lr": 8.854967116644925e-05}
{"train_loss": 0.29289165139198303, "global_step": 60044, "epoch": 674, "lr": 8.854930196259906e-05}
{"train_loss": 0.19587309658527374, "global_step": 60045, "epoch": 674, "lr": 8.85489327535664e-05}
{"train_loss": 0.24554771184921265, "global_step": 60046, "epoch": 674, "lr": 8.85485635393513e-05}
{"train_loss": 0.29574525356292725, "global_step": 60047, "epoch": 674, "lr": 8.85481943199538e-05}
{"train_loss": 0.2069321572780609, "global_step": 60048, "epoch": 674, "lr": 8.854782509537399e-05}
{"train_loss": 0.3712601363658905, "global_step": 60049, "epoch": 674, "lr": 8.854745586561188e-05}
{"train_loss": 0.22336618602275848, "global_step": 60050, "epoch": 674, "lr": 8.854708663066753e-05}
{"train_loss": 0.26550453901290894, "global_step": 60051, "epoch": 674, "lr": 8.8546717390541e-05}
{"train_loss": 0.17898845672607422, "global_step": 60052, "epoch": 674, "lr": 8.854634814523233e-05}
{"train_loss": 0.22933341562747955, "global_step": 60053, "epoch": 674, "lr": 8.854597889474159e-05}
{"train_loss": 0.23874883353710175, "global_step": 60054, "epoch": 674, "lr": 8.854560963906879e-05}
{"train_loss": 0.2927398085594177, "global_step": 60055, "epoch": 674, "lr": 8.854524037821401e-05}
{"train_loss": 0.20950353145599365, "global_step": 60056, "epoch": 674, "lr": 8.85448711121773e-05}
{"train_loss": 0.21783775091171265, "global_step": 60057, "epoch": 674, "lr": 8.85445018409587e-05}
{"train_loss": 0.25845733284950256, "global_step": 60058, "epoch": 674, "lr": 8.854413256455826e-05}
{"train_loss": 0.2253434807062149, "global_step": 60059, "epoch": 674, "lr": 8.854376328297602e-05}
{"train_loss": 0.23383170366287231, "global_step": 60060, "epoch": 674, "lr": 8.854339399621205e-05}
{"train_loss": 0.23871150612831116, "global_step": 60061, "epoch": 674, "lr": 8.854302470426639e-05}
{"train_loss": 0.25595584511756897, "global_step": 60062, "epoch": 674, "lr": 8.85426554071391e-05}
{"train_loss": 0.2840765416622162, "global_step": 60063, "epoch": 674, "lr": 8.854228610483022e-05}
{"train_loss": 0.2908250093460083, "global_step": 60064, "epoch": 674, "lr": 8.854191679733979e-05}
{"train_loss": 0.22944600880146027, "global_step": 60065, "epoch": 674, "lr": 8.854154748466786e-05}
{"train_loss": 0.2096216231584549, "global_step": 60066, "epoch": 674, "lr": 8.85411781668145e-05}
{"train_loss": 0.2512342035770416, "global_step": 60067, "epoch": 674, "lr": 8.854080884377974e-05}
{"train_loss": 0.23475632071495056, "global_step": 60068, "epoch": 674, "lr": 8.854043951556365e-05}
{"train_loss": 0.19282835721969604, "global_step": 60069, "epoch": 674, "lr": 8.854007018216626e-05}
{"train_loss": 0.2223259061574936, "global_step": 60070, "epoch": 674, "lr": 8.853970084358762e-05}
{"train_loss": 0.2549949884414673, "global_step": 60071, "epoch": 674, "lr": 8.85393314998278e-05}
{"train_loss": 0.21224743127822876, "global_step": 60072, "epoch": 674, "lr": 8.853896215088683e-05}
{"train_loss": 0.28828105330467224, "global_step": 60073, "epoch": 674, "lr": 8.853859279676477e-05}
{"train_loss": 0.24110408780280124, "global_step": 60074, "epoch": 674, "lr": 8.853822343746167e-05, "val_loss": 3.4996426105499268}
{"train_loss": 0.2151445597410202, "global_step": 60075, "epoch": 675, "lr": 8.853785407297758e-05}
{"train_loss": 0.22532925009727478, "global_step": 60076, "epoch": 675, "lr": 8.853748470331253e-05}
{"train_loss": 0.2550583779811859, "global_step": 60077, "epoch": 675, "lr": 8.85371153284666e-05}
{"train_loss": 0.32706695795059204, "global_step": 60078, "epoch": 675, "lr": 8.853674594843981e-05}
{"train_loss": 0.20453013479709625, "global_step": 60079, "epoch": 675, "lr": 8.853637656323222e-05}
{"train_loss": 0.1444997638463974, "global_step": 60080, "epoch": 675, "lr": 8.853600717284391e-05}
{"train_loss": 0.39458441734313965, "global_step": 60081, "epoch": 675, "lr": 8.853563777727487e-05}
{"train_loss": 0.22079308331012726, "global_step": 60082, "epoch": 675, "lr": 8.853526837652522e-05}
{"train_loss": 0.30265045166015625, "global_step": 60083, "epoch": 675, "lr": 8.853489897059495e-05}
{"train_loss": 0.2841029763221741, "global_step": 60084, "epoch": 675, "lr": 8.853452955948413e-05}
{"train_loss": 0.24872243404388428, "global_step": 60085, "epoch": 675, "lr": 8.853416014319282e-05}
{"train_loss": 0.16975559294223785, "global_step": 60086, "epoch": 675, "lr": 8.853379072172105e-05}
{"train_loss": 0.191290482878685, "global_step": 60087, "epoch": 675, "lr": 8.85334212950689e-05}
{"train_loss": 0.23627826571464539, "global_step": 60088, "epoch": 675, "lr": 8.853305186323639e-05}
{"train_loss": 0.22535094618797302, "global_step": 60089, "epoch": 675, "lr": 8.853268242622358e-05}
{"train_loss": 0.25336015224456787, "global_step": 60090, "epoch": 675, "lr": 8.853231298403052e-05}
{"train_loss": 0.2808174192905426, "global_step": 60091, "epoch": 675, "lr": 8.853194353665726e-05}
{"train_loss": 0.1512795239686966, "global_step": 60092, "epoch": 675, "lr": 8.853157408410386e-05}
{"train_loss": 0.3696196377277374, "global_step": 60093, "epoch": 675, "lr": 8.853120462637036e-05}
{"train_loss": 0.21719442307949066, "global_step": 60094, "epoch": 675, "lr": 8.853083516345679e-05}
{"train_loss": 0.18734228610992432, "global_step": 60095, "epoch": 675, "lr": 8.853046569536323e-05}
{"train_loss": 0.20639145374298096, "global_step": 60096, "epoch": 675, "lr": 8.853009622208973e-05}
{"train_loss": 0.20837906002998352, "global_step": 60097, "epoch": 675, "lr": 8.85297267436363e-05}
{"train_loss": 0.25948554277420044, "global_step": 60098, "epoch": 675, "lr": 8.852935726000305e-05}
{"train_loss": 0.1705172061920166, "global_step": 60099, "epoch": 675, "lr": 8.852898777118999e-05}
{"train_loss": 0.25645211338996887, "global_step": 60100, "epoch": 675, "lr": 8.852861827719716e-05}
{"train_loss": 0.34590694308280945, "global_step": 60101, "epoch": 675, "lr": 8.852824877802463e-05}
{"train_loss": 0.1978338360786438, "global_step": 60102, "epoch": 675, "lr": 8.852787927367247e-05}
{"train_loss": 0.27771010994911194, "global_step": 60103, "epoch": 675, "lr": 8.852750976414068e-05}
{"train_loss": 0.30114224553108215, "global_step": 60104, "epoch": 675, "lr": 8.852714024942934e-05}
{"train_loss": 0.2300562858581543, "global_step": 60105, "epoch": 675, "lr": 8.852677072953851e-05}
{"train_loss": 0.1929810494184494, "global_step": 60106, "epoch": 675, "lr": 8.852640120446822e-05}
{"train_loss": 0.22138182818889618, "global_step": 60107, "epoch": 675, "lr": 8.852603167421852e-05}
{"train_loss": 0.13949771225452423, "global_step": 60108, "epoch": 675, "lr": 8.852566213878947e-05}
{"train_loss": 0.20662495493888855, "global_step": 60109, "epoch": 675, "lr": 8.852529259818111e-05}
{"train_loss": 0.2637719213962555, "global_step": 60110, "epoch": 675, "lr": 8.85249230523935e-05}
{"train_loss": 0.17987781763076782, "global_step": 60111, "epoch": 675, "lr": 8.852455350142668e-05}
{"train_loss": 0.19891349971294403, "global_step": 60112, "epoch": 675, "lr": 8.852418394528069e-05}
{"train_loss": 0.2665320038795471, "global_step": 60113, "epoch": 675, "lr": 8.852381438395561e-05}
{"train_loss": 0.2863895297050476, "global_step": 60114, "epoch": 675, "lr": 8.852344481745146e-05}
{"train_loss": 0.27420058846473694, "global_step": 60115, "epoch": 675, "lr": 8.852307524576831e-05}
{"train_loss": 0.27900195121765137, "global_step": 60116, "epoch": 675, "lr": 8.852270566890621e-05}
{"train_loss": 0.2778615355491638, "global_step": 60117, "epoch": 675, "lr": 8.852233608686518e-05}
{"train_loss": 0.2217138707637787, "global_step": 60118, "epoch": 675, "lr": 8.852196649964532e-05}
{"train_loss": 0.2975917160511017, "global_step": 60119, "epoch": 675, "lr": 8.852159690724663e-05}
{"train_loss": 0.3119131922721863, "global_step": 60120, "epoch": 675, "lr": 8.852122730966919e-05}
{"train_loss": 0.23021593689918518, "global_step": 60121, "epoch": 675, "lr": 8.852085770691303e-05}
{"train_loss": 0.26340651512145996, "global_step": 60122, "epoch": 675, "lr": 8.852048809897821e-05}
{"train_loss": 0.240193173289299, "global_step": 60123, "epoch": 675, "lr": 8.85201184858648e-05}
{"train_loss": 0.23091205954551697, "global_step": 60124, "epoch": 675, "lr": 8.851974886757282e-05}
{"train_loss": 0.2200755923986435, "global_step": 60125, "epoch": 675, "lr": 8.851937924410232e-05}
{"train_loss": 0.24361976981163025, "global_step": 60126, "epoch": 675, "lr": 8.851900961545337e-05}
{"train_loss": 0.16927538812160492, "global_step": 60127, "epoch": 675, "lr": 8.8518639981626e-05}
{"train_loss": 0.24937081336975098, "global_step": 60128, "epoch": 675, "lr": 8.851827034262027e-05}
{"train_loss": 0.24590596556663513, "global_step": 60129, "epoch": 675, "lr": 8.851790069843622e-05}
{"train_loss": 0.22421452403068542, "global_step": 60130, "epoch": 675, "lr": 8.851753104907392e-05}
{"train_loss": 0.1869903951883316, "global_step": 60131, "epoch": 675, "lr": 8.85171613945334e-05}
{"train_loss": 0.23345020413398743, "global_step": 60132, "epoch": 675, "lr": 8.851679173481473e-05}
{"train_loss": 0.2105967402458191, "global_step": 60133, "epoch": 675, "lr": 8.851642206991795e-05}
{"train_loss": 0.22353743016719818, "global_step": 60134, "epoch": 675, "lr": 8.851605239984308e-05}
{"train_loss": 0.2988216280937195, "global_step": 60135, "epoch": 675, "lr": 8.851568272459022e-05}
{"train_loss": 0.25931668281555176, "global_step": 60136, "epoch": 675, "lr": 8.851531304415939e-05}
{"train_loss": 0.23694542050361633, "global_step": 60137, "epoch": 675, "lr": 8.851494335855065e-05}
{"train_loss": 0.3737104535102844, "global_step": 60138, "epoch": 675, "lr": 8.851457366776404e-05}
{"train_loss": 0.34435179829597473, "global_step": 60139, "epoch": 675, "lr": 8.85142039717996e-05}
{"train_loss": 0.37738797068595886, "global_step": 60140, "epoch": 675, "lr": 8.851383427065741e-05}
{"train_loss": 0.16762453317642212, "global_step": 60141, "epoch": 675, "lr": 8.85134645643375e-05}
{"train_loss": 0.2900508642196655, "global_step": 60142, "epoch": 675, "lr": 8.851309485283992e-05}
{"train_loss": 0.38215717673301697, "global_step": 60143, "epoch": 675, "lr": 8.851272513616473e-05}
{"train_loss": 0.3131537139415741, "global_step": 60144, "epoch": 675, "lr": 8.851235541431199e-05}
{"train_loss": 0.35928890109062195, "global_step": 60145, "epoch": 675, "lr": 8.851198568728172e-05}
{"train_loss": 0.19728818535804749, "global_step": 60146, "epoch": 675, "lr": 8.851161595507398e-05}
{"train_loss": 0.2576044499874115, "global_step": 60147, "epoch": 675, "lr": 8.851124621768882e-05}
{"train_loss": 0.21000456809997559, "global_step": 60148, "epoch": 675, "lr": 8.851087647512629e-05}
{"train_loss": 0.22201845049858093, "global_step": 60149, "epoch": 675, "lr": 8.851050672738646e-05}
{"train_loss": 0.34203386306762695, "global_step": 60150, "epoch": 675, "lr": 8.851013697446934e-05}
{"train_loss": 0.3452519476413727, "global_step": 60151, "epoch": 675, "lr": 8.8509767216375e-05}
{"train_loss": 0.34854140877723694, "global_step": 60152, "epoch": 675, "lr": 8.850939745310351e-05}
{"train_loss": 0.22500601410865784, "global_step": 60153, "epoch": 675, "lr": 8.850902768465489e-05}
{"train_loss": 0.2845500409603119, "global_step": 60154, "epoch": 675, "lr": 8.85086579110292e-05}
{"train_loss": 0.2885877192020416, "global_step": 60155, "epoch": 675, "lr": 8.850828813222649e-05}
{"train_loss": 0.31913769245147705, "global_step": 60156, "epoch": 675, "lr": 8.850791834824682e-05}
{"train_loss": 0.20482097566127777, "global_step": 60157, "epoch": 675, "lr": 8.850754855909023e-05}
{"train_loss": 0.30246129631996155, "global_step": 60158, "epoch": 675, "lr": 8.850717876475677e-05}
{"train_loss": 0.19685819745063782, "global_step": 60159, "epoch": 675, "lr": 8.850680896524647e-05}
{"train_loss": 0.3892597556114197, "global_step": 60160, "epoch": 675, "lr": 8.850643916055941e-05}
{"train_loss": 0.312420517206192, "global_step": 60161, "epoch": 675, "lr": 8.850606935069563e-05}
{"train_loss": 0.3326091766357422, "global_step": 60162, "epoch": 675, "lr": 8.850569953565517e-05}
{"train_loss": 0.2555863647983315, "global_step": 60163, "epoch": 675, "lr": 8.85053297154381e-05, "val_loss": 3.346900463104248, "train_action_mse_error": 23.097469329833984}
{"train_loss": 0.3365371525287628, "global_step": 60164, "epoch": 676, "lr": 8.850495989004446e-05}
{"train_loss": 0.15566881000995636, "global_step": 60165, "epoch": 676, "lr": 8.85045900594743e-05}
{"train_loss": 0.23867373168468475, "global_step": 60166, "epoch": 676, "lr": 8.850422022372765e-05}
{"train_loss": 0.2787719666957855, "global_step": 60167, "epoch": 676, "lr": 8.850385038280458e-05}
{"train_loss": 0.27131927013397217, "global_step": 60168, "epoch": 676, "lr": 8.850348053670514e-05}
{"train_loss": 0.18987445533275604, "global_step": 60169, "epoch": 676, "lr": 8.850311068542938e-05}
{"train_loss": 0.3144168555736542, "global_step": 60170, "epoch": 676, "lr": 8.850274082897733e-05}
{"train_loss": 0.24185095727443695, "global_step": 60171, "epoch": 676, "lr": 8.850237096734908e-05}
{"train_loss": 0.2430502474308014, "global_step": 60172, "epoch": 676, "lr": 8.850200110054464e-05}
{"train_loss": 0.20311865210533142, "global_step": 60173, "epoch": 676, "lr": 8.850163122856409e-05}
{"train_loss": 0.2822737693786621, "global_step": 60174, "epoch": 676, "lr": 8.850126135140745e-05}
{"train_loss": 0.2788536548614502, "global_step": 60175, "epoch": 676, "lr": 8.850089146907478e-05}
{"train_loss": 0.24503637850284576, "global_step": 60176, "epoch": 676, "lr": 8.850052158156616e-05}
{"train_loss": 0.26103535294532776, "global_step": 60177, "epoch": 676, "lr": 8.85001516888816e-05}
{"train_loss": 0.24154949188232422, "global_step": 60178, "epoch": 676, "lr": 8.849978179102116e-05}
{"train_loss": 0.2410089671611786, "global_step": 60179, "epoch": 676, "lr": 8.84994118879849e-05}
{"train_loss": 0.19447198510169983, "global_step": 60180, "epoch": 676, "lr": 8.849904197977284e-05}
{"train_loss": 0.24622471630573273, "global_step": 60181, "epoch": 676, "lr": 8.849867206638508e-05}
{"train_loss": 0.30007463693618774, "global_step": 60182, "epoch": 676, "lr": 8.849830214782163e-05}
{"train_loss": 0.20829564332962036, "global_step": 60183, "epoch": 676, "lr": 8.849793222408256e-05}
{"train_loss": 0.1879086047410965, "global_step": 60184, "epoch": 676, "lr": 8.849756229516791e-05}
{"train_loss": 0.18689744174480438, "global_step": 60185, "epoch": 676, "lr": 8.849719236107774e-05}
{"train_loss": 0.21480749547481537, "global_step": 60186, "epoch": 676, "lr": 8.849682242181208e-05}
{"train_loss": 0.24383707344532013, "global_step": 60187, "epoch": 676, "lr": 8.849645247737099e-05}
{"train_loss": 0.2723262906074524, "global_step": 60188, "epoch": 676, "lr": 8.849608252775452e-05}
{"train_loss": 0.249733567237854, "global_step": 60189, "epoch": 676, "lr": 8.849571257296274e-05}
{"train_loss": 0.20404650270938873, "global_step": 60190, "epoch": 676, "lr": 8.849534261299566e-05}
{"train_loss": 0.290168821811676, "global_step": 60191, "epoch": 676, "lr": 8.849497264785336e-05}
{"train_loss": 0.215617835521698, "global_step": 60192, "epoch": 676, "lr": 8.849460267753588e-05}
{"train_loss": 0.19878384470939636, "global_step": 60193, "epoch": 676, "lr": 8.849423270204327e-05}
{"train_loss": 0.19197168946266174, "global_step": 60194, "epoch": 676, "lr": 8.849386272137557e-05}
{"train_loss": 0.25560709834098816, "global_step": 60195, "epoch": 676, "lr": 8.849349273553284e-05}
{"train_loss": 0.2540142238140106, "global_step": 60196, "epoch": 676, "lr": 8.849312274451515e-05}
{"train_loss": 0.3290497660636902, "global_step": 60197, "epoch": 676, "lr": 8.849275274832252e-05}
{"train_loss": 0.2404659241437912, "global_step": 60198, "epoch": 676, "lr": 8.8492382746955e-05}
{"train_loss": 0.22253191471099854, "global_step": 60199, "epoch": 676, "lr": 8.849201274041264e-05}
{"train_loss": 0.22539328038692474, "global_step": 60200, "epoch": 676, "lr": 8.84916427286955e-05}
{"train_loss": 0.25623008608818054, "global_step": 60201, "epoch": 676, "lr": 8.849127271180365e-05}
{"train_loss": 0.1742798537015915, "global_step": 60202, "epoch": 676, "lr": 8.849090268973712e-05}
{"train_loss": 0.24305087327957153, "global_step": 60203, "epoch": 676, "lr": 8.849053266249593e-05}
{"train_loss": 0.2316468358039856, "global_step": 60204, "epoch": 676, "lr": 8.849016263008017e-05}
{"train_loss": 0.24774770438671112, "global_step": 60205, "epoch": 676, "lr": 8.848979259248988e-05}
{"train_loss": 0.35966596007347107, "global_step": 60206, "epoch": 676, "lr": 8.84894225497251e-05}
{"train_loss": 0.2532453238964081, "global_step": 60207, "epoch": 676, "lr": 8.848905250178589e-05}
{"train_loss": 0.248881995677948, "global_step": 60208, "epoch": 676, "lr": 8.848868244867229e-05}
{"train_loss": 0.2755354940891266, "global_step": 60209, "epoch": 676, "lr": 8.848831239038435e-05}
{"train_loss": 0.3333287835121155, "global_step": 60210, "epoch": 676, "lr": 8.848794232692215e-05}
{"train_loss": 0.2505783140659332, "global_step": 60211, "epoch": 676, "lr": 8.84875722582857e-05}
{"train_loss": 0.24400818347930908, "global_step": 60212, "epoch": 676, "lr": 8.848720218447506e-05}
{"train_loss": 0.1910322606563568, "global_step": 60213, "epoch": 676, "lr": 8.848683210549028e-05}
{"train_loss": 0.28016072511672974, "global_step": 60214, "epoch": 676, "lr": 8.848646202133144e-05}
{"train_loss": 0.1739262193441391, "global_step": 60215, "epoch": 676, "lr": 8.848609193199855e-05}
{"train_loss": 0.2958765923976898, "global_step": 60216, "epoch": 676, "lr": 8.848572183749167e-05}
{"train_loss": 0.25998392701148987, "global_step": 60217, "epoch": 676, "lr": 8.848535173781085e-05}
{"train_loss": 0.2303193360567093, "global_step": 60218, "epoch": 676, "lr": 8.848498163295616e-05}
{"train_loss": 0.2570396363735199, "global_step": 60219, "epoch": 676, "lr": 8.848461152292762e-05}
{"train_loss": 0.21567732095718384, "global_step": 60220, "epoch": 676, "lr": 8.84842414077253e-05}
{"train_loss": 0.2965303659439087, "global_step": 60221, "epoch": 676, "lr": 8.848387128734925e-05}
{"train_loss": 0.19280709326267242, "global_step": 60222, "epoch": 676, "lr": 8.848350116179949e-05}
{"train_loss": 0.24246379733085632, "global_step": 60223, "epoch": 676, "lr": 8.84831310310761e-05}
{"train_loss": 0.2541269361972809, "global_step": 60224, "epoch": 676, "lr": 8.848276089517912e-05}
{"train_loss": 0.24620680510997772, "global_step": 60225, "epoch": 676, "lr": 8.848239075410862e-05}
{"train_loss": 0.22169512510299683, "global_step": 60226, "epoch": 676, "lr": 8.848202060786463e-05}
{"train_loss": 0.23381023108959198, "global_step": 60227, "epoch": 676, "lr": 8.848165045644719e-05}
{"train_loss": 0.22871506214141846, "global_step": 60228, "epoch": 676, "lr": 8.848128029985636e-05}
{"train_loss": 0.2295464277267456, "global_step": 60229, "epoch": 676, "lr": 8.848091013809219e-05}
{"train_loss": 0.2448209971189499, "global_step": 60230, "epoch": 676, "lr": 8.848053997115474e-05}
{"train_loss": 0.23232972621917725, "global_step": 60231, "epoch": 676, "lr": 8.848016979904404e-05}
{"train_loss": 0.2689487636089325, "global_step": 60232, "epoch": 676, "lr": 8.847979962176017e-05}
{"train_loss": 0.2550339996814728, "global_step": 60233, "epoch": 676, "lr": 8.847942943930315e-05}
{"train_loss": 0.28967875242233276, "global_step": 60234, "epoch": 676, "lr": 8.847905925167302e-05}
{"train_loss": 0.18000662326812744, "global_step": 60235, "epoch": 676, "lr": 8.847868905886987e-05}
{"train_loss": 0.31392788887023926, "global_step": 60236, "epoch": 676, "lr": 8.847831886089372e-05}
{"train_loss": 0.251444935798645, "global_step": 60237, "epoch": 676, "lr": 8.847794865774463e-05}
{"train_loss": 0.22647254168987274, "global_step": 60238, "epoch": 676, "lr": 8.847757844942265e-05}
{"train_loss": 0.29417508840560913, "global_step": 60239, "epoch": 676, "lr": 8.847720823592784e-05}
{"train_loss": 0.18992242217063904, "global_step": 60240, "epoch": 676, "lr": 8.847683801726022e-05}
{"train_loss": 0.1209322139620781, "global_step": 60241, "epoch": 676, "lr": 8.847646779341987e-05}
{"train_loss": 0.1945878118276596, "global_step": 60242, "epoch": 676, "lr": 8.847609756440683e-05}
{"train_loss": 0.1927204132080078, "global_step": 60243, "epoch": 676, "lr": 8.847572733022114e-05}
{"train_loss": 0.29598698019981384, "global_step": 60244, "epoch": 676, "lr": 8.847535709086286e-05}
{"train_loss": 0.18449081480503082, "global_step": 60245, "epoch": 676, "lr": 8.847498684633202e-05}
{"train_loss": 0.24563957750797272, "global_step": 60246, "epoch": 676, "lr": 8.847461659662872e-05}
{"train_loss": 0.22189638018608093, "global_step": 60247, "epoch": 676, "lr": 8.847424634175295e-05}
{"train_loss": 0.24567539989948273, "global_step": 60248, "epoch": 676, "lr": 8.84738760817048e-05}
{"train_loss": 0.24655739963054657, "global_step": 60249, "epoch": 676, "lr": 8.84735058164843e-05}
{"train_loss": 0.23733709752559662, "global_step": 60250, "epoch": 676, "lr": 8.847313554609151e-05}
{"train_loss": 0.27542394399642944, "global_step": 60251, "epoch": 676, "lr": 8.847276527052648e-05}
{"train_loss": 0.2426374788197239, "global_step": 60252, "epoch": 676, "lr": 8.847239498978924e-05, "val_loss": 3.3108694553375244}
{"train_loss": 0.17664337158203125, "global_step": 60253, "epoch": 677, "lr": 8.847202470387987e-05}
{"train_loss": 0.23921281099319458, "global_step": 60254, "epoch": 677, "lr": 8.847165441279839e-05}
{"train_loss": 0.20844916999340057, "global_step": 60255, "epoch": 677, "lr": 8.847128411654488e-05}
{"train_loss": 0.21666671335697174, "global_step": 60256, "epoch": 677, "lr": 8.847091381511938e-05}
{"train_loss": 0.24675500392913818, "global_step": 60257, "epoch": 677, "lr": 8.847054350852193e-05}
{"train_loss": 0.2602466344833374, "global_step": 60258, "epoch": 677, "lr": 8.847017319675257e-05}
{"train_loss": 0.22260548174381256, "global_step": 60259, "epoch": 677, "lr": 8.846980287981136e-05}
{"train_loss": 0.34111732244491577, "global_step": 60260, "epoch": 677, "lr": 8.846943255769838e-05}
{"train_loss": 0.212640643119812, "global_step": 60261, "epoch": 677, "lr": 8.846906223041365e-05}
{"train_loss": 0.2511545419692993, "global_step": 60262, "epoch": 677, "lr": 8.846869189795722e-05}
{"train_loss": 0.1742158681154251, "global_step": 60263, "epoch": 677, "lr": 8.846832156032913e-05}
{"train_loss": 0.20737089216709137, "global_step": 60264, "epoch": 677, "lr": 8.846795121752944e-05}
{"train_loss": 0.2320346236228943, "global_step": 60265, "epoch": 677, "lr": 8.846758086955821e-05}
{"train_loss": 0.33774587512016296, "global_step": 60266, "epoch": 677, "lr": 8.84672105164155e-05}
{"train_loss": 0.27426788210868835, "global_step": 60267, "epoch": 677, "lr": 8.846684015810132e-05}
{"train_loss": 0.16459155082702637, "global_step": 60268, "epoch": 677, "lr": 8.846646979461576e-05}
{"train_loss": 0.14378248155117035, "global_step": 60269, "epoch": 677, "lr": 8.846609942595884e-05}
{"train_loss": 0.2607872784137726, "global_step": 60270, "epoch": 677, "lr": 8.846572905213062e-05}
{"train_loss": 0.23618607223033905, "global_step": 60271, "epoch": 677, "lr": 8.846535867313117e-05}
{"train_loss": 0.16313491761684418, "global_step": 60272, "epoch": 677, "lr": 8.84649882889605e-05}
{"train_loss": 0.2618821859359741, "global_step": 60273, "epoch": 677, "lr": 8.846461789961869e-05}
{"train_loss": 0.1721232384443283, "global_step": 60274, "epoch": 677, "lr": 8.846424750510578e-05}
{"train_loss": 0.18900857865810394, "global_step": 60275, "epoch": 677, "lr": 8.846387710542182e-05}
{"train_loss": 0.2770566940307617, "global_step": 60276, "epoch": 677, "lr": 8.846350670056686e-05}
{"train_loss": 0.2107652723789215, "global_step": 60277, "epoch": 677, "lr": 8.846313629054094e-05}
{"train_loss": 0.25506591796875, "global_step": 60278, "epoch": 677, "lr": 8.846276587534415e-05}
{"train_loss": 0.308398962020874, "global_step": 60279, "epoch": 677, "lr": 8.846239545497649e-05}
{"train_loss": 0.23058433830738068, "global_step": 60280, "epoch": 677, "lr": 8.846202502943803e-05}
{"train_loss": 0.2134055495262146, "global_step": 60281, "epoch": 677, "lr": 8.846165459872881e-05}
{"train_loss": 0.2688189744949341, "global_step": 60282, "epoch": 677, "lr": 8.846128416284891e-05}
{"train_loss": 0.2744160592556, "global_step": 60283, "epoch": 677, "lr": 8.846091372179834e-05}
{"train_loss": 0.24259889125823975, "global_step": 60284, "epoch": 677, "lr": 8.846054327557718e-05}
{"train_loss": 0.20408543944358826, "global_step": 60285, "epoch": 677, "lr": 8.846017282418547e-05}
{"train_loss": 0.28553488850593567, "global_step": 60286, "epoch": 677, "lr": 8.845980236762326e-05}
{"train_loss": 0.2285006195306778, "global_step": 60287, "epoch": 677, "lr": 8.845943190589058e-05}
{"train_loss": 0.1956821084022522, "global_step": 60288, "epoch": 677, "lr": 8.845906143898753e-05}
{"train_loss": 0.2589099109172821, "global_step": 60289, "epoch": 677, "lr": 8.845869096691411e-05}
{"train_loss": 0.17901307344436646, "global_step": 60290, "epoch": 677, "lr": 8.845832048967038e-05}
{"train_loss": 0.1943526268005371, "global_step": 60291, "epoch": 677, "lr": 8.845795000725641e-05}
{"train_loss": 0.15330927073955536, "global_step": 60292, "epoch": 677, "lr": 8.845757951967222e-05}
{"train_loss": 0.2455609291791916, "global_step": 60293, "epoch": 677, "lr": 8.84572090269179e-05}
{"train_loss": 0.2583531439304352, "global_step": 60294, "epoch": 677, "lr": 8.845683852899347e-05}
{"train_loss": 0.23790033161640167, "global_step": 60295, "epoch": 677, "lr": 8.845646802589897e-05}
{"train_loss": 0.2680273950099945, "global_step": 60296, "epoch": 677, "lr": 8.845609751763448e-05}
{"train_loss": 0.18693336844444275, "global_step": 60297, "epoch": 677, "lr": 8.845572700420004e-05}
{"train_loss": 0.18257200717926025, "global_step": 60298, "epoch": 677, "lr": 8.845535648559569e-05}
{"train_loss": 0.28503456711769104, "global_step": 60299, "epoch": 677, "lr": 8.84549859618215e-05}
{"train_loss": 0.2058209925889969, "global_step": 60300, "epoch": 677, "lr": 8.845461543287749e-05}
{"train_loss": 0.20292304456233978, "global_step": 60301, "epoch": 677, "lr": 8.845424489876372e-05}
{"train_loss": 0.2448960840702057, "global_step": 60302, "epoch": 677, "lr": 8.845387435948026e-05}
{"train_loss": 0.20059539377689362, "global_step": 60303, "epoch": 677, "lr": 8.845350381502715e-05}
{"train_loss": 0.30391982197761536, "global_step": 60304, "epoch": 677, "lr": 8.845313326540441e-05}
{"train_loss": 0.2295350432395935, "global_step": 60305, "epoch": 677, "lr": 8.845276271061214e-05}
{"train_loss": 0.22017908096313477, "global_step": 60306, "epoch": 677, "lr": 8.845239215065036e-05}
{"train_loss": 0.20275379717350006, "global_step": 60307, "epoch": 677, "lr": 8.845202158551911e-05}
{"train_loss": 0.22685669362545013, "global_step": 60308, "epoch": 677, "lr": 8.845165101521846e-05}
{"train_loss": 0.28701040148735046, "global_step": 60309, "epoch": 677, "lr": 8.845128043974846e-05}
{"train_loss": 0.2034676969051361, "global_step": 60310, "epoch": 677, "lr": 8.845090985910915e-05}
{"train_loss": 0.2785092890262604, "global_step": 60311, "epoch": 677, "lr": 8.84505392733006e-05}
{"train_loss": 0.22609087824821472, "global_step": 60312, "epoch": 677, "lr": 8.845016868232282e-05}
{"train_loss": 0.21185798943042755, "global_step": 60313, "epoch": 677, "lr": 8.844979808617589e-05}
{"train_loss": 0.2135447859764099, "global_step": 60314, "epoch": 677, "lr": 8.844942748485987e-05}
{"train_loss": 0.2170989066362381, "global_step": 60315, "epoch": 677, "lr": 8.844905687837477e-05}
{"train_loss": 0.28001654148101807, "global_step": 60316, "epoch": 677, "lr": 8.844868626672069e-05}
{"train_loss": 0.20152193307876587, "global_step": 60317, "epoch": 677, "lr": 8.844831564989762e-05}
{"train_loss": 0.2542394697666168, "global_step": 60318, "epoch": 677, "lr": 8.844794502790566e-05}
{"train_loss": 0.2985418438911438, "global_step": 60319, "epoch": 677, "lr": 8.844757440074485e-05}
{"train_loss": 0.20301219820976257, "global_step": 60320, "epoch": 677, "lr": 8.844720376841522e-05}
{"train_loss": 0.23568521440029144, "global_step": 60321, "epoch": 677, "lr": 8.844683313091683e-05}
{"train_loss": 0.2146754264831543, "global_step": 60322, "epoch": 677, "lr": 8.844646248824972e-05}
{"train_loss": 0.211794912815094, "global_step": 60323, "epoch": 677, "lr": 8.844609184041398e-05}
{"train_loss": 0.265312522649765, "global_step": 60324, "epoch": 677, "lr": 8.844572118740962e-05}
{"train_loss": 0.2717324495315552, "global_step": 60325, "epoch": 677, "lr": 8.84453505292367e-05}
{"train_loss": 0.21044480800628662, "global_step": 60326, "epoch": 677, "lr": 8.844497986589529e-05}
{"train_loss": 0.20767177641391754, "global_step": 60327, "epoch": 677, "lr": 8.84446091973854e-05}
{"train_loss": 0.2804358899593353, "global_step": 60328, "epoch": 677, "lr": 8.844423852370709e-05}
{"train_loss": 0.23123089969158173, "global_step": 60329, "epoch": 677, "lr": 8.844386784486045e-05}
{"train_loss": 0.2562457323074341, "global_step": 60330, "epoch": 677, "lr": 8.844349716084548e-05}
{"train_loss": 0.20438092947006226, "global_step": 60331, "epoch": 677, "lr": 8.844312647166225e-05}
{"train_loss": 0.2422705739736557, "global_step": 60332, "epoch": 677, "lr": 8.844275577731082e-05}
{"train_loss": 0.19536757469177246, "global_step": 60333, "epoch": 677, "lr": 8.844238507779122e-05}
{"train_loss": 0.22076989710330963, "global_step": 60334, "epoch": 677, "lr": 8.844201437310351e-05}
{"train_loss": 0.24642464518547058, "global_step": 60335, "epoch": 677, "lr": 8.844164366324776e-05}
{"train_loss": 0.27091890573501587, "global_step": 60336, "epoch": 677, "lr": 8.844127294822399e-05}
{"train_loss": 0.28069186210632324, "global_step": 60337, "epoch": 677, "lr": 8.844090222803225e-05}
{"train_loss": 0.2066228985786438, "global_step": 60338, "epoch": 677, "lr": 8.84405315026726e-05}
{"train_loss": 0.18197473883628845, "global_step": 60339, "epoch": 677, "lr": 8.84401607721451e-05}
{"train_loss": 0.24133139848709106, "global_step": 60340, "epoch": 677, "lr": 8.843979003644978e-05}
{"train_loss": 0.231686835710922, "global_step": 60341, "epoch": 677, "lr": 8.84394192955867e-05, "val_loss": 3.44538950920105}
{"train_loss": 0.33455580472946167, "global_step": 60342, "epoch": 678, "lr": 8.843904854955591e-05}
{"train_loss": 0.19053345918655396, "global_step": 60343, "epoch": 678, "lr": 8.843867779835746e-05}
{"train_loss": 0.2157641500234604, "global_step": 60344, "epoch": 678, "lr": 8.843830704199138e-05}
{"train_loss": 0.29647892713546753, "global_step": 60345, "epoch": 678, "lr": 8.843793628045777e-05}
{"train_loss": 0.2044692039489746, "global_step": 60346, "epoch": 678, "lr": 8.843756551375664e-05}
{"train_loss": 0.3336319923400879, "global_step": 60347, "epoch": 678, "lr": 8.843719474188803e-05}
{"train_loss": 0.26343533396720886, "global_step": 60348, "epoch": 678, "lr": 8.843682396485203e-05}
{"train_loss": 0.2443421632051468, "global_step": 60349, "epoch": 678, "lr": 8.843645318264864e-05}
{"train_loss": 0.25974801182746887, "global_step": 60350, "epoch": 678, "lr": 8.843608239527796e-05}
{"train_loss": 0.26018816232681274, "global_step": 60351, "epoch": 678, "lr": 8.843571160274e-05}
{"train_loss": 0.27364903688430786, "global_step": 60352, "epoch": 678, "lr": 8.843534080503484e-05}
{"train_loss": 0.23192842304706573, "global_step": 60353, "epoch": 678, "lr": 8.843497000216251e-05}
{"train_loss": 0.1734808385372162, "global_step": 60354, "epoch": 678, "lr": 8.843459919412307e-05}
{"train_loss": 0.27343592047691345, "global_step": 60355, "epoch": 678, "lr": 8.843422838091657e-05}
{"train_loss": 0.24280911684036255, "global_step": 60356, "epoch": 678, "lr": 8.843385756254306e-05}
{"train_loss": 0.4279894530773163, "global_step": 60357, "epoch": 678, "lr": 8.843348673900258e-05}
{"train_loss": 0.20228241384029388, "global_step": 60358, "epoch": 678, "lr": 8.843311591029517e-05}
{"train_loss": 0.21941222250461578, "global_step": 60359, "epoch": 678, "lr": 8.843274507642093e-05}
{"train_loss": 0.2778724431991577, "global_step": 60360, "epoch": 678, "lr": 8.843237423737985e-05}
{"train_loss": 0.31002530455589294, "global_step": 60361, "epoch": 678, "lr": 8.843200339317201e-05}
{"train_loss": 0.2988412082195282, "global_step": 60362, "epoch": 678, "lr": 8.843163254379745e-05}
{"train_loss": 0.15832839906215668, "global_step": 60363, "epoch": 678, "lr": 8.843126168925624e-05}
{"train_loss": 0.23578718304634094, "global_step": 60364, "epoch": 678, "lr": 8.843089082954841e-05}
{"train_loss": 0.2777966260910034, "global_step": 60365, "epoch": 678, "lr": 8.843051996467402e-05}
{"train_loss": 0.20995891094207764, "global_step": 60366, "epoch": 678, "lr": 8.84301490946331e-05}
{"train_loss": 0.30054861307144165, "global_step": 60367, "epoch": 678, "lr": 8.842977821942573e-05}
{"train_loss": 0.3883321285247803, "global_step": 60368, "epoch": 678, "lr": 8.842940733905194e-05}
{"train_loss": 0.17995008826255798, "global_step": 60369, "epoch": 678, "lr": 8.842903645351178e-05}
{"train_loss": 0.2763111889362335, "global_step": 60370, "epoch": 678, "lr": 8.842866556280531e-05}
{"train_loss": 0.2225690484046936, "global_step": 60371, "epoch": 678, "lr": 8.842829466693258e-05}
{"train_loss": 0.3294243812561035, "global_step": 60372, "epoch": 678, "lr": 8.842792376589361e-05}
{"train_loss": 0.22635966539382935, "global_step": 60373, "epoch": 678, "lr": 8.842755285968852e-05}
{"train_loss": 0.23668892681598663, "global_step": 60374, "epoch": 678, "lr": 8.842718194831728e-05}
{"train_loss": 0.23546385765075684, "global_step": 60375, "epoch": 678, "lr": 8.842681103177997e-05}
{"train_loss": 0.38105255365371704, "global_step": 60376, "epoch": 678, "lr": 8.842644011007665e-05}
{"train_loss": 0.32843998074531555, "global_step": 60377, "epoch": 678, "lr": 8.842606918320736e-05}
{"train_loss": 0.2989150881767273, "global_step": 60378, "epoch": 678, "lr": 8.842569825117217e-05}
{"train_loss": 0.23240475356578827, "global_step": 60379, "epoch": 678, "lr": 8.842532731397109e-05}
{"train_loss": 0.21039098501205444, "global_step": 60380, "epoch": 678, "lr": 8.84249563716042e-05}
{"train_loss": 0.21125641465187073, "global_step": 60381, "epoch": 678, "lr": 8.842458542407154e-05}
{"train_loss": 0.254925400018692, "global_step": 60382, "epoch": 678, "lr": 8.842421447137319e-05}
{"train_loss": 0.15131735801696777, "global_step": 60383, "epoch": 678, "lr": 8.842384351350915e-05}
{"train_loss": 0.2155189961194992, "global_step": 60384, "epoch": 678, "lr": 8.84234725504795e-05}
{"train_loss": 0.22588440775871277, "global_step": 60385, "epoch": 678, "lr": 8.842310158228428e-05}
{"train_loss": 0.2764753997325897, "global_step": 60386, "epoch": 678, "lr": 8.842273060892354e-05}
{"train_loss": 0.2896232008934021, "global_step": 60387, "epoch": 678, "lr": 8.842235963039734e-05}
{"train_loss": 0.14511536061763763, "global_step": 60388, "epoch": 678, "lr": 8.84219886467057e-05}
{"train_loss": 0.28261032700538635, "global_step": 60389, "epoch": 678, "lr": 8.842161765784871e-05}
{"train_loss": 0.22414204478263855, "global_step": 60390, "epoch": 678, "lr": 8.842124666382641e-05}
{"train_loss": 0.22145989537239075, "global_step": 60391, "epoch": 678, "lr": 8.842087566463882e-05}
{"train_loss": 0.1865675449371338, "global_step": 60392, "epoch": 678, "lr": 8.842050466028602e-05}
{"train_loss": 0.28938719630241394, "global_step": 60393, "epoch": 678, "lr": 8.842013365076806e-05}
{"train_loss": 0.24413523077964783, "global_step": 60394, "epoch": 678, "lr": 8.841976263608497e-05}
{"train_loss": 0.16048279404640198, "global_step": 60395, "epoch": 678, "lr": 8.841939161623683e-05}
{"train_loss": 0.23609715700149536, "global_step": 60396, "epoch": 678, "lr": 8.841902059122364e-05}
{"train_loss": 0.22692915797233582, "global_step": 60397, "epoch": 678, "lr": 8.84186495610455e-05}
{"train_loss": 0.23803916573524475, "global_step": 60398, "epoch": 678, "lr": 8.841827852570244e-05}
{"train_loss": 0.18390107154846191, "global_step": 60399, "epoch": 678, "lr": 8.841790748519451e-05}
{"train_loss": 0.2863903343677521, "global_step": 60400, "epoch": 678, "lr": 8.841753643952175e-05}
{"train_loss": 0.19182007014751434, "global_step": 60401, "epoch": 678, "lr": 8.841716538868424e-05}
{"train_loss": 0.227565735578537, "global_step": 60402, "epoch": 678, "lr": 8.8416794332682e-05}
{"train_loss": 0.27500277757644653, "global_step": 60403, "epoch": 678, "lr": 8.841642327151509e-05}
{"train_loss": 0.28587985038757324, "global_step": 60404, "epoch": 678, "lr": 8.841605220518356e-05}
{"train_loss": 0.224100723862648, "global_step": 60405, "epoch": 678, "lr": 8.841568113368747e-05}
{"train_loss": 0.26578018069267273, "global_step": 60406, "epoch": 678, "lr": 8.841531005702685e-05}
{"train_loss": 0.2293345332145691, "global_step": 60407, "epoch": 678, "lr": 8.841493897520176e-05}
{"train_loss": 0.2806561589241028, "global_step": 60408, "epoch": 678, "lr": 8.841456788821226e-05}
{"train_loss": 0.2601853609085083, "global_step": 60409, "epoch": 678, "lr": 8.841419679605837e-05}
{"train_loss": 0.15806753933429718, "global_step": 60410, "epoch": 678, "lr": 8.841382569874017e-05}
{"train_loss": 0.22952979803085327, "global_step": 60411, "epoch": 678, "lr": 8.84134545962577e-05}
{"train_loss": 0.24476264417171478, "global_step": 60412, "epoch": 678, "lr": 8.841308348861101e-05}
{"train_loss": 0.21035315096378326, "global_step": 60413, "epoch": 678, "lr": 8.841271237580015e-05}
{"train_loss": 0.2719613015651703, "global_step": 60414, "epoch": 678, "lr": 8.841234125782516e-05}
{"train_loss": 0.19845633208751678, "global_step": 60415, "epoch": 678, "lr": 8.841197013468609e-05}
{"train_loss": 0.17453166842460632, "global_step": 60416, "epoch": 678, "lr": 8.841159900638303e-05}
{"train_loss": 0.22647714614868164, "global_step": 60417, "epoch": 678, "lr": 8.841122787291598e-05}
{"train_loss": 0.27207350730895996, "global_step": 60418, "epoch": 678, "lr": 8.841085673428503e-05}
{"train_loss": 0.2500443458557129, "global_step": 60419, "epoch": 678, "lr": 8.841048559049017e-05}
{"train_loss": 0.20412954688072205, "global_step": 60420, "epoch": 678, "lr": 8.841011444153152e-05}
{"train_loss": 0.23409594595432281, "global_step": 60421, "epoch": 678, "lr": 8.840974328740908e-05}
{"train_loss": 0.26985156536102295, "global_step": 60422, "epoch": 678, "lr": 8.840937212812293e-05}
{"train_loss": 0.17049528658390045, "global_step": 60423, "epoch": 678, "lr": 8.84090009636731e-05}
{"train_loss": 0.2788010835647583, "global_step": 60424, "epoch": 678, "lr": 8.840862979405963e-05}
{"train_loss": 0.2886638641357422, "global_step": 60425, "epoch": 678, "lr": 8.840825861928262e-05}
{"train_loss": 0.23913763463497162, "global_step": 60426, "epoch": 678, "lr": 8.840788743934207e-05}
{"train_loss": 0.21598227322101593, "global_step": 60427, "epoch": 678, "lr": 8.840751625423807e-05}
{"train_loss": 0.2609584927558899, "global_step": 60428, "epoch": 678, "lr": 8.840714506397062e-05}
{"train_loss": 0.2057226598262787, "global_step": 60429, "epoch": 678, "lr": 8.840677386853981e-05}
{"train_loss": 0.24615454422623922, "global_step": 60430, "epoch": 678, "lr": 8.840640266794568e-05, "val_loss": 3.392603635787964}
{"train_loss": 0.18091163039207458, "global_step": 60431, "epoch": 679, "lr": 8.840603146218826e-05}
{"train_loss": 0.20006981492042542, "global_step": 60432, "epoch": 679, "lr": 8.840566025126764e-05}
{"train_loss": 0.2886282801628113, "global_step": 60433, "epoch": 679, "lr": 8.840528903518384e-05}
{"train_loss": 0.15797924995422363, "global_step": 60434, "epoch": 679, "lr": 8.840491781393691e-05}
{"train_loss": 0.15320558845996857, "global_step": 60435, "epoch": 679, "lr": 8.840454658752691e-05}
{"train_loss": 0.25104355812072754, "global_step": 60436, "epoch": 679, "lr": 8.84041753559539e-05}
{"train_loss": 0.29845142364501953, "global_step": 60437, "epoch": 679, "lr": 8.84038041192179e-05}
{"train_loss": 0.26770490407943726, "global_step": 60438, "epoch": 679, "lr": 8.840343287731898e-05}
{"train_loss": 0.2994042932987213, "global_step": 60439, "epoch": 679, "lr": 8.840306163025719e-05}
{"train_loss": 0.22935204207897186, "global_step": 60440, "epoch": 679, "lr": 8.840269037803257e-05}
{"train_loss": 0.11477261036634445, "global_step": 60441, "epoch": 679, "lr": 8.840231912064518e-05}
{"train_loss": 0.3695208430290222, "global_step": 60442, "epoch": 679, "lr": 8.840194785809508e-05}
{"train_loss": 0.20633652806282043, "global_step": 60443, "epoch": 679, "lr": 8.840157659038229e-05}
{"train_loss": 0.26522767543792725, "global_step": 60444, "epoch": 679, "lr": 8.840120531750688e-05}
{"train_loss": 0.1579078733921051, "global_step": 60445, "epoch": 679, "lr": 8.840083403946888e-05}
{"train_loss": 0.30741551518440247, "global_step": 60446, "epoch": 679, "lr": 8.840046275626838e-05}
{"train_loss": 0.22927320003509521, "global_step": 60447, "epoch": 679, "lr": 8.84000914679054e-05}
{"train_loss": 0.2054491490125656, "global_step": 60448, "epoch": 679, "lr": 8.839972017437999e-05}
{"train_loss": 0.20593604445457458, "global_step": 60449, "epoch": 679, "lr": 8.839934887569222e-05}
{"train_loss": 0.24680007994174957, "global_step": 60450, "epoch": 679, "lr": 8.83989775718421e-05}
{"train_loss": 0.20394597947597504, "global_step": 60451, "epoch": 679, "lr": 8.839860626282972e-05}
{"train_loss": 0.1885613203048706, "global_step": 60452, "epoch": 679, "lr": 8.839823494865511e-05}
{"train_loss": 0.2100811004638672, "global_step": 60453, "epoch": 679, "lr": 8.839786362931834e-05}
{"train_loss": 0.23498719930648804, "global_step": 60454, "epoch": 679, "lr": 8.839749230481944e-05}
{"train_loss": 0.3138493299484253, "global_step": 60455, "epoch": 679, "lr": 8.839712097515847e-05}
{"train_loss": 0.23454196751117706, "global_step": 60456, "epoch": 679, "lr": 8.839674964033546e-05}
{"train_loss": 0.20205846428871155, "global_step": 60457, "epoch": 679, "lr": 8.839637830035048e-05}
{"train_loss": 0.296487420797348, "global_step": 60458, "epoch": 679, "lr": 8.839600695520357e-05}
{"train_loss": 0.25401395559310913, "global_step": 60459, "epoch": 679, "lr": 8.839563560489479e-05}
{"train_loss": 0.20310664176940918, "global_step": 60460, "epoch": 679, "lr": 8.839526424942419e-05}
{"train_loss": 0.2087172418832779, "global_step": 60461, "epoch": 679, "lr": 8.839489288879181e-05}
{"train_loss": 0.2215329259634018, "global_step": 60462, "epoch": 679, "lr": 8.83945215229977e-05}
{"train_loss": 0.20158597826957703, "global_step": 60463, "epoch": 679, "lr": 8.839415015204192e-05}
{"train_loss": 0.22570577263832092, "global_step": 60464, "epoch": 679, "lr": 8.839377877592452e-05}
{"train_loss": 0.28610050678253174, "global_step": 60465, "epoch": 679, "lr": 8.839340739464553e-05}
{"train_loss": 0.21705181896686554, "global_step": 60466, "epoch": 679, "lr": 8.839303600820502e-05}
{"train_loss": 0.19623662531375885, "global_step": 60467, "epoch": 679, "lr": 8.839266461660303e-05}
{"train_loss": 0.21233436465263367, "global_step": 60468, "epoch": 679, "lr": 8.839229321983963e-05}
{"train_loss": 0.31986671686172485, "global_step": 60469, "epoch": 679, "lr": 8.839192181791485e-05}
{"train_loss": 0.21127933263778687, "global_step": 60470, "epoch": 679, "lr": 8.839155041082873e-05}
{"train_loss": 0.18724283576011658, "global_step": 60471, "epoch": 679, "lr": 8.839117899858135e-05}
{"train_loss": 0.18189512193202972, "global_step": 60472, "epoch": 679, "lr": 8.839080758117272e-05}
{"train_loss": 0.20183534920215607, "global_step": 60473, "epoch": 679, "lr": 8.839043615860294e-05}
{"train_loss": 0.28563278913497925, "global_step": 60474, "epoch": 679, "lr": 8.839006473087203e-05}
{"train_loss": 0.2372399866580963, "global_step": 60475, "epoch": 679, "lr": 8.838969329798003e-05}
{"train_loss": 0.19392360746860504, "global_step": 60476, "epoch": 679, "lr": 8.838932185992701e-05}
{"train_loss": 0.2688754200935364, "global_step": 60477, "epoch": 679, "lr": 8.838895041671302e-05}
{"train_loss": 0.23984675109386444, "global_step": 60478, "epoch": 679, "lr": 8.83885789683381e-05}
{"train_loss": 0.22617757320404053, "global_step": 60479, "epoch": 679, "lr": 8.83882075148023e-05}
{"train_loss": 0.15996576845645905, "global_step": 60480, "epoch": 679, "lr": 8.838783605610567e-05}
{"train_loss": 0.20136767625808716, "global_step": 60481, "epoch": 679, "lr": 8.838746459224828e-05}
{"train_loss": 0.2659134268760681, "global_step": 60482, "epoch": 679, "lr": 8.838709312323016e-05}
{"train_loss": 0.24358870089054108, "global_step": 60483, "epoch": 679, "lr": 8.838672164905135e-05}
{"train_loss": 0.2118033766746521, "global_step": 60484, "epoch": 679, "lr": 8.838635016971192e-05}
{"train_loss": 0.285586416721344, "global_step": 60485, "epoch": 679, "lr": 8.838597868521191e-05}
{"train_loss": 0.19855573773384094, "global_step": 60486, "epoch": 679, "lr": 8.838560719555138e-05}
{"train_loss": 0.21986810863018036, "global_step": 60487, "epoch": 679, "lr": 8.838523570073037e-05}
{"train_loss": 0.2117706835269928, "global_step": 60488, "epoch": 679, "lr": 8.838486420074891e-05}
{"train_loss": 0.20602336525917053, "global_step": 60489, "epoch": 679, "lr": 8.838449269560712e-05}
{"train_loss": 0.22479034960269928, "global_step": 60490, "epoch": 679, "lr": 8.838412118530497e-05}
{"train_loss": 0.27654388546943665, "global_step": 60491, "epoch": 679, "lr": 8.838374966984254e-05}
{"train_loss": 0.29762712121009827, "global_step": 60492, "epoch": 679, "lr": 8.83833781492199e-05}
{"train_loss": 0.25467973947525024, "global_step": 60493, "epoch": 679, "lr": 8.838300662343707e-05}
{"train_loss": 0.21705275774002075, "global_step": 60494, "epoch": 679, "lr": 8.838263509249412e-05}
{"train_loss": 0.28992632031440735, "global_step": 60495, "epoch": 679, "lr": 8.838226355639108e-05}
{"train_loss": 0.28057512640953064, "global_step": 60496, "epoch": 679, "lr": 8.838189201512803e-05}
{"train_loss": 0.32480525970458984, "global_step": 60497, "epoch": 679, "lr": 8.838152046870499e-05}
{"train_loss": 0.26056718826293945, "global_step": 60498, "epoch": 679, "lr": 8.838114891712203e-05}
{"train_loss": 0.26556217670440674, "global_step": 60499, "epoch": 679, "lr": 8.83807773603792e-05}
{"train_loss": 0.18855063617229462, "global_step": 60500, "epoch": 679, "lr": 8.838040579847653e-05}
{"train_loss": 0.26260679960250854, "global_step": 60501, "epoch": 679, "lr": 8.838003423141408e-05}
{"train_loss": 0.24041518568992615, "global_step": 60502, "epoch": 679, "lr": 8.837966265919191e-05}
{"train_loss": 0.21620523929595947, "global_step": 60503, "epoch": 679, "lr": 8.837929108181004e-05}
{"train_loss": 0.31147462129592896, "global_step": 60504, "epoch": 679, "lr": 8.837891949926857e-05}
{"train_loss": 0.2366999238729477, "global_step": 60505, "epoch": 679, "lr": 8.83785479115675e-05}
{"train_loss": 0.20972034335136414, "global_step": 60506, "epoch": 679, "lr": 8.837817631870694e-05}
{"train_loss": 0.23331373929977417, "global_step": 60507, "epoch": 679, "lr": 8.837780472068687e-05}
{"train_loss": 0.2913976311683655, "global_step": 60508, "epoch": 679, "lr": 8.837743311750737e-05}
{"train_loss": 0.21803705394268036, "global_step": 60509, "epoch": 679, "lr": 8.83770615091685e-05}
{"train_loss": 0.2593803107738495, "global_step": 60510, "epoch": 679, "lr": 8.837668989567029e-05}
{"train_loss": 0.19902509450912476, "global_step": 60511, "epoch": 679, "lr": 8.837631827701283e-05}
{"train_loss": 0.3084217309951782, "global_step": 60512, "epoch": 679, "lr": 8.837594665319612e-05}
{"train_loss": 0.3616286814212799, "global_step": 60513, "epoch": 679, "lr": 8.837557502422022e-05}
{"train_loss": 0.21932470798492432, "global_step": 60514, "epoch": 679, "lr": 8.837520339008523e-05}
{"train_loss": 0.19623278081417084, "global_step": 60515, "epoch": 679, "lr": 8.837483175079113e-05}
{"train_loss": 0.26434892416000366, "global_step": 60516, "epoch": 679, "lr": 8.837446010633802e-05}
{"train_loss": 0.22221976518630981, "global_step": 60517, "epoch": 679, "lr": 8.837408845672592e-05}
{"train_loss": 0.26019957661628723, "global_step": 60518, "epoch": 679, "lr": 8.83737168019549e-05}
{"train_loss": 0.2369870939448978, "global_step": 60519, "epoch": 679, "lr": 8.8373345142025e-05, "val_loss": 3.35642409324646}
{"train_loss": 0.2788631021976471, "global_step": 60520, "epoch": 680, "lr": 8.837297347693625e-05}
{"train_loss": 0.28868842124938965, "global_step": 60521, "epoch": 680, "lr": 8.837260180668875e-05}
{"train_loss": 0.2745482623577118, "global_step": 60522, "epoch": 680, "lr": 8.83722301312825e-05}
{"train_loss": 0.2648625373840332, "global_step": 60523, "epoch": 680, "lr": 8.837185845071759e-05}
{"train_loss": 0.3472729027271271, "global_step": 60524, "epoch": 680, "lr": 8.837148676499405e-05}
{"train_loss": 0.21850134432315826, "global_step": 60525, "epoch": 680, "lr": 8.837111507411191e-05}
{"train_loss": 0.16722595691680908, "global_step": 60526, "epoch": 680, "lr": 8.837074337807128e-05}
{"train_loss": 0.2062964141368866, "global_step": 60527, "epoch": 680, "lr": 8.837037167687214e-05}
{"train_loss": 0.25373631715774536, "global_step": 60528, "epoch": 680, "lr": 8.836999997051458e-05}
{"train_loss": 0.2859169840812683, "global_step": 60529, "epoch": 680, "lr": 8.836962825899864e-05}
{"train_loss": 0.24882973730564117, "global_step": 60530, "epoch": 680, "lr": 8.836925654232436e-05}
{"train_loss": 0.17727546393871307, "global_step": 60531, "epoch": 680, "lr": 8.836888482049181e-05}
{"train_loss": 0.27413931488990784, "global_step": 60532, "epoch": 680, "lr": 8.836851309350103e-05}
{"train_loss": 0.2637802064418793, "global_step": 60533, "epoch": 680, "lr": 8.836814136135207e-05}
{"train_loss": 0.28242963552474976, "global_step": 60534, "epoch": 680, "lr": 8.8367769624045e-05}
{"train_loss": 0.16335266828536987, "global_step": 60535, "epoch": 680, "lr": 8.836739788157983e-05}
{"train_loss": 0.26276522874832153, "global_step": 60536, "epoch": 680, "lr": 8.836702613395664e-05}
{"train_loss": 0.3098861277103424, "global_step": 60537, "epoch": 680, "lr": 8.836665438117547e-05}
{"train_loss": 0.22532057762145996, "global_step": 60538, "epoch": 680, "lr": 8.836628262323637e-05}
{"train_loss": 0.23773840069770813, "global_step": 60539, "epoch": 680, "lr": 8.836591086013939e-05}
{"train_loss": 0.4341506361961365, "global_step": 60540, "epoch": 680, "lr": 8.836553909188456e-05}
{"train_loss": 0.18898969888687134, "global_step": 60541, "epoch": 680, "lr": 8.836516731847197e-05}
{"train_loss": 0.1491253525018692, "global_step": 60542, "epoch": 680, "lr": 8.836479553990166e-05}
{"train_loss": 0.27760565280914307, "global_step": 60543, "epoch": 680, "lr": 8.836442375617366e-05}
{"train_loss": 0.2385464906692505, "global_step": 60544, "epoch": 680, "lr": 8.836405196728804e-05}
{"train_loss": 0.21103250980377197, "global_step": 60545, "epoch": 680, "lr": 8.836368017324483e-05}
{"train_loss": 0.21787472069263458, "global_step": 60546, "epoch": 680, "lr": 8.836330837404411e-05}
{"train_loss": 0.20696857571601868, "global_step": 60547, "epoch": 680, "lr": 8.836293656968589e-05}
{"train_loss": 0.2107374370098114, "global_step": 60548, "epoch": 680, "lr": 8.836256476017025e-05}
{"train_loss": 0.22471128404140472, "global_step": 60549, "epoch": 680, "lr": 8.836219294549722e-05}
{"train_loss": 0.23228538036346436, "global_step": 60550, "epoch": 680, "lr": 8.836182112566687e-05}
{"train_loss": 0.18905465304851532, "global_step": 60551, "epoch": 680, "lr": 8.836144930067924e-05}
{"train_loss": 0.27516549825668335, "global_step": 60552, "epoch": 680, "lr": 8.836107747053439e-05}
{"train_loss": 0.26666751503944397, "global_step": 60553, "epoch": 680, "lr": 8.836070563523235e-05}
{"train_loss": 0.19940507411956787, "global_step": 60554, "epoch": 680, "lr": 8.836033379477318e-05}
{"train_loss": 0.24674756824970245, "global_step": 60555, "epoch": 680, "lr": 8.835996194915694e-05}
{"train_loss": 0.3002393841743469, "global_step": 60556, "epoch": 680, "lr": 8.835959009838366e-05}
{"train_loss": 0.23504257202148438, "global_step": 60557, "epoch": 680, "lr": 8.835921824245339e-05}
{"train_loss": 0.17787081003189087, "global_step": 60558, "epoch": 680, "lr": 8.83588463813662e-05}
{"train_loss": 0.32394376397132874, "global_step": 60559, "epoch": 680, "lr": 8.835847451512214e-05}
{"train_loss": 0.32693812251091003, "global_step": 60560, "epoch": 680, "lr": 8.835810264372124e-05}
{"train_loss": 0.22691628336906433, "global_step": 60561, "epoch": 680, "lr": 8.835773076716359e-05}
{"train_loss": 0.21162033081054688, "global_step": 60562, "epoch": 680, "lr": 8.835735888544917e-05}
{"train_loss": 0.23495085537433624, "global_step": 60563, "epoch": 680, "lr": 8.835698699857809e-05}
{"train_loss": 0.21995604038238525, "global_step": 60564, "epoch": 680, "lr": 8.835661510655038e-05}
{"train_loss": 0.2142695039510727, "global_step": 60565, "epoch": 680, "lr": 8.835624320936609e-05}
{"train_loss": 0.32602572441101074, "global_step": 60566, "epoch": 680, "lr": 8.835587130702527e-05}
{"train_loss": 0.24940454959869385, "global_step": 60567, "epoch": 680, "lr": 8.835549939952796e-05}
{"train_loss": 0.38003045320510864, "global_step": 60568, "epoch": 680, "lr": 8.835512748687424e-05}
{"train_loss": 0.30695101618766785, "global_step": 60569, "epoch": 680, "lr": 8.835475556906412e-05}
{"train_loss": 0.22055767476558685, "global_step": 60570, "epoch": 680, "lr": 8.835438364609768e-05}
{"train_loss": 0.37320056557655334, "global_step": 60571, "epoch": 680, "lr": 8.835401171797497e-05}
{"train_loss": 0.23676109313964844, "global_step": 60572, "epoch": 680, "lr": 8.835363978469602e-05}
{"train_loss": 0.20179088413715363, "global_step": 60573, "epoch": 680, "lr": 8.83532678462609e-05}
{"train_loss": 0.20520900189876556, "global_step": 60574, "epoch": 680, "lr": 8.835289590266965e-05}
{"train_loss": 0.20897966623306274, "global_step": 60575, "epoch": 680, "lr": 8.83525239539223e-05}
{"train_loss": 0.16663944721221924, "global_step": 60576, "epoch": 680, "lr": 8.835215200001895e-05}
{"train_loss": 0.19045299291610718, "global_step": 60577, "epoch": 680, "lr": 8.835178004095958e-05}
{"train_loss": 0.2632911801338196, "global_step": 60578, "epoch": 680, "lr": 8.83514080767443e-05}
{"train_loss": 0.296787291765213, "global_step": 60579, "epoch": 680, "lr": 8.835103610737316e-05}
{"train_loss": 0.24424727261066437, "global_step": 60580, "epoch": 680, "lr": 8.835066413284619e-05}
{"train_loss": 0.2353677749633789, "global_step": 60581, "epoch": 680, "lr": 8.835029215316342e-05}
{"train_loss": 0.19256591796875, "global_step": 60582, "epoch": 680, "lr": 8.834992016832492e-05}
{"train_loss": 0.3329918086528778, "global_step": 60583, "epoch": 680, "lr": 8.834954817833076e-05}
{"train_loss": 0.24457041919231415, "global_step": 60584, "epoch": 680, "lr": 8.834917618318096e-05}
{"train_loss": 0.21929121017456055, "global_step": 60585, "epoch": 680, "lr": 8.834880418287558e-05}
{"train_loss": 0.2987770140171051, "global_step": 60586, "epoch": 680, "lr": 8.834843217741467e-05}
{"train_loss": 0.3068351149559021, "global_step": 60587, "epoch": 680, "lr": 8.834806016679829e-05}
{"train_loss": 0.24822020530700684, "global_step": 60588, "epoch": 680, "lr": 8.834768815102646e-05}
{"train_loss": 0.25043895840644836, "global_step": 60589, "epoch": 680, "lr": 8.834731613009927e-05}
{"train_loss": 0.23084892332553864, "global_step": 60590, "epoch": 680, "lr": 8.834694410401674e-05}
{"train_loss": 0.2157147228717804, "global_step": 60591, "epoch": 680, "lr": 8.834657207277894e-05}
{"train_loss": 0.3056058883666992, "global_step": 60592, "epoch": 680, "lr": 8.834620003638589e-05}
{"train_loss": 0.2820090651512146, "global_step": 60593, "epoch": 680, "lr": 8.834582799483769e-05}
{"train_loss": 0.23659846186637878, "global_step": 60594, "epoch": 680, "lr": 8.834545594813435e-05}
{"train_loss": 0.29921212792396545, "global_step": 60595, "epoch": 680, "lr": 8.834508389627592e-05}
{"train_loss": 0.25016093254089355, "global_step": 60596, "epoch": 680, "lr": 8.834471183926247e-05}
{"train_loss": 0.28330257534980774, "global_step": 60597, "epoch": 680, "lr": 8.834433977709404e-05}
{"train_loss": 0.23221462965011597, "global_step": 60598, "epoch": 680, "lr": 8.834396770977069e-05}
{"train_loss": 0.27340012788772583, "global_step": 60599, "epoch": 680, "lr": 8.834359563729244e-05}
{"train_loss": 0.2469375878572464, "global_step": 60600, "epoch": 680, "lr": 8.834322355965938e-05}
{"train_loss": 0.2193794548511505, "global_step": 60601, "epoch": 680, "lr": 8.834285147687153e-05}
{"train_loss": 0.2625316381454468, "global_step": 60602, "epoch": 680, "lr": 8.834247938892896e-05}
{"train_loss": 0.24894081056118011, "global_step": 60603, "epoch": 680, "lr": 8.83421072958317e-05}
{"train_loss": 0.26043984293937683, "global_step": 60604, "epoch": 680, "lr": 8.834173519757982e-05}
{"train_loss": 0.18899418413639069, "global_step": 60605, "epoch": 680, "lr": 8.834136309417335e-05}
{"train_loss": 0.2732835114002228, "global_step": 60606, "epoch": 680, "lr": 8.834099098561238e-05}
{"train_loss": 0.22424153983592987, "global_step": 60607, "epoch": 680, "lr": 8.834061887189688e-05}
{"train_loss": 0.24992993909321473, "global_step": 60608, "epoch": 680, "lr": 8.834024675302699e-05, "val_loss": 3.3807029724121094, "train_action_mse_error": 16.308467864990234}
{"train_loss": 0.21950840950012207, "global_step": 60609, "epoch": 681, "lr": 8.833987462900272e-05}
{"train_loss": 0.26396825909614563, "global_step": 60610, "epoch": 681, "lr": 8.83395024998241e-05}
{"train_loss": 0.27595293521881104, "global_step": 60611, "epoch": 681, "lr": 8.833913036549122e-05}
{"train_loss": 0.23185983300209045, "global_step": 60612, "epoch": 681, "lr": 8.83387582260041e-05}
{"train_loss": 0.15932556986808777, "global_step": 60613, "epoch": 681, "lr": 8.833838608136282e-05}
{"train_loss": 0.20909371972084045, "global_step": 60614, "epoch": 681, "lr": 8.83380139315674e-05}
{"train_loss": 0.2244696021080017, "global_step": 60615, "epoch": 681, "lr": 8.83376417766179e-05}
{"train_loss": 0.2908842861652374, "global_step": 60616, "epoch": 681, "lr": 8.833726961651437e-05}
{"train_loss": 0.2513653337955475, "global_step": 60617, "epoch": 681, "lr": 8.833689745125687e-05}
{"train_loss": 0.16670583188533783, "global_step": 60618, "epoch": 681, "lr": 8.833652528084542e-05}
{"train_loss": 0.25169357657432556, "global_step": 60619, "epoch": 681, "lr": 8.83361531052801e-05}
{"train_loss": 0.21364983916282654, "global_step": 60620, "epoch": 681, "lr": 8.833578092456097e-05}
{"train_loss": 0.184508815407753, "global_step": 60621, "epoch": 681, "lr": 8.833540873868804e-05}
{"train_loss": 0.1391393095254898, "global_step": 60622, "epoch": 681, "lr": 8.833503654766138e-05}
{"train_loss": 0.2829453945159912, "global_step": 60623, "epoch": 681, "lr": 8.833466435148105e-05}
{"train_loss": 0.20893041789531708, "global_step": 60624, "epoch": 681, "lr": 8.83342921501471e-05}
{"train_loss": 0.3160635828971863, "global_step": 60625, "epoch": 681, "lr": 8.833391994365956e-05}
{"train_loss": 0.20746812224388123, "global_step": 60626, "epoch": 681, "lr": 8.83335477320185e-05}
{"train_loss": 0.35343194007873535, "global_step": 60627, "epoch": 681, "lr": 8.833317551522396e-05}
{"train_loss": 0.2178826928138733, "global_step": 60628, "epoch": 681, "lr": 8.833280329327598e-05}
{"train_loss": 0.18752142786979675, "global_step": 60629, "epoch": 681, "lr": 8.833243106617464e-05}
{"train_loss": 0.25621771812438965, "global_step": 60630, "epoch": 681, "lr": 8.833205883391995e-05}
{"train_loss": 0.18204869329929352, "global_step": 60631, "epoch": 681, "lr": 8.8331686596512e-05}
{"train_loss": 0.22182555496692657, "global_step": 60632, "epoch": 681, "lr": 8.833131435395081e-05}
{"train_loss": 0.2634255588054657, "global_step": 60633, "epoch": 681, "lr": 8.833094210623646e-05}
{"train_loss": 0.21732322871685028, "global_step": 60634, "epoch": 681, "lr": 8.833056985336896e-05}
{"train_loss": 0.19807644188404083, "global_step": 60635, "epoch": 681, "lr": 8.83301975953484e-05}
{"train_loss": 0.23940064013004303, "global_step": 60636, "epoch": 681, "lr": 8.83298253321748e-05}
{"train_loss": 0.20697200298309326, "global_step": 60637, "epoch": 681, "lr": 8.832945306384824e-05}
{"train_loss": 0.18699778616428375, "global_step": 60638, "epoch": 681, "lr": 8.832908079036873e-05}
{"train_loss": 0.2872372567653656, "global_step": 60639, "epoch": 681, "lr": 8.832870851173635e-05}
{"train_loss": 0.25566163659095764, "global_step": 60640, "epoch": 681, "lr": 8.832833622795115e-05}
{"train_loss": 0.13987529277801514, "global_step": 60641, "epoch": 681, "lr": 8.832796393901317e-05}
{"train_loss": 0.21470069885253906, "global_step": 60642, "epoch": 681, "lr": 8.832759164492246e-05}
{"train_loss": 0.22211484611034393, "global_step": 60643, "epoch": 681, "lr": 8.832721934567907e-05}
{"train_loss": 0.2089771181344986, "global_step": 60644, "epoch": 681, "lr": 8.832684704128304e-05}
{"train_loss": 0.2694189250469208, "global_step": 60645, "epoch": 681, "lr": 8.832647473173445e-05}
{"train_loss": 0.22620615363121033, "global_step": 60646, "epoch": 681, "lr": 8.832610241703333e-05}
{"train_loss": 0.2542824447154999, "global_step": 60647, "epoch": 681, "lr": 8.832573009717973e-05}
{"train_loss": 0.21140477061271667, "global_step": 60648, "epoch": 681, "lr": 8.83253577721737e-05}
{"train_loss": 0.26839226484298706, "global_step": 60649, "epoch": 681, "lr": 8.83249854420153e-05}
{"train_loss": 0.17314022779464722, "global_step": 60650, "epoch": 681, "lr": 8.832461310670457e-05}
{"train_loss": 0.29557308554649353, "global_step": 60651, "epoch": 681, "lr": 8.832424076624156e-05}
{"train_loss": 0.2723490595817566, "global_step": 60652, "epoch": 681, "lr": 8.832386842062632e-05}
{"train_loss": 0.2415660172700882, "global_step": 60653, "epoch": 681, "lr": 8.832349606985893e-05}
{"train_loss": 0.32006123661994934, "global_step": 60654, "epoch": 681, "lr": 8.832312371393938e-05}
{"train_loss": 0.21211905777454376, "global_step": 60655, "epoch": 681, "lr": 8.832275135286777e-05}
{"train_loss": 0.21365946531295776, "global_step": 60656, "epoch": 681, "lr": 8.832237898664412e-05}
{"train_loss": 0.24169322848320007, "global_step": 60657, "epoch": 681, "lr": 8.83220066152685e-05}
{"train_loss": 0.29097694158554077, "global_step": 60658, "epoch": 681, "lr": 8.832163423874096e-05}
{"train_loss": 0.1857907623052597, "global_step": 60659, "epoch": 681, "lr": 8.832126185706153e-05}
{"train_loss": 0.2056279331445694, "global_step": 60660, "epoch": 681, "lr": 8.832088947023028e-05}
{"train_loss": 0.15664592385292053, "global_step": 60661, "epoch": 681, "lr": 8.832051707824727e-05}
{"train_loss": 0.24211950600147247, "global_step": 60662, "epoch": 681, "lr": 8.832014468111252e-05}
{"train_loss": 0.18664681911468506, "global_step": 60663, "epoch": 681, "lr": 8.831977227882609e-05}
{"train_loss": 0.20837606489658356, "global_step": 60664, "epoch": 681, "lr": 8.831939987138801e-05}
{"train_loss": 0.2264440506696701, "global_step": 60665, "epoch": 681, "lr": 8.831902745879838e-05}
{"train_loss": 0.1821041703224182, "global_step": 60666, "epoch": 681, "lr": 8.831865504105723e-05}
{"train_loss": 0.2597953677177429, "global_step": 60667, "epoch": 681, "lr": 8.831828261816459e-05}
{"train_loss": 0.24494098126888275, "global_step": 60668, "epoch": 681, "lr": 8.831791019012054e-05}
{"train_loss": 0.256914347410202, "global_step": 60669, "epoch": 681, "lr": 8.83175377569251e-05}
{"train_loss": 0.38809946179389954, "global_step": 60670, "epoch": 681, "lr": 8.831716531857834e-05}
{"train_loss": 0.22590863704681396, "global_step": 60671, "epoch": 681, "lr": 8.831679287508031e-05}
{"train_loss": 0.21813379228115082, "global_step": 60672, "epoch": 681, "lr": 8.831642042643104e-05}
{"train_loss": 0.32350629568099976, "global_step": 60673, "epoch": 681, "lr": 8.83160479726306e-05}
{"train_loss": 0.23825736343860626, "global_step": 60674, "epoch": 681, "lr": 8.831567551367904e-05}
{"train_loss": 0.22814640402793884, "global_step": 60675, "epoch": 681, "lr": 8.83153030495764e-05}
{"train_loss": 0.38938355445861816, "global_step": 60676, "epoch": 681, "lr": 8.831493058032274e-05}
{"train_loss": 0.2646423280239105, "global_step": 60677, "epoch": 681, "lr": 8.831455810591808e-05}
{"train_loss": 0.21419748663902283, "global_step": 60678, "epoch": 681, "lr": 8.831418562636253e-05}
{"train_loss": 0.26294979453086853, "global_step": 60679, "epoch": 681, "lr": 8.831381314165609e-05}
{"train_loss": 0.2345091849565506, "global_step": 60680, "epoch": 681, "lr": 8.831344065179883e-05}
{"train_loss": 0.2978544235229492, "global_step": 60681, "epoch": 681, "lr": 8.831306815679078e-05}
{"train_loss": 0.25639235973358154, "global_step": 60682, "epoch": 681, "lr": 8.831269565663202e-05}
{"train_loss": 0.20304779708385468, "global_step": 60683, "epoch": 681, "lr": 8.831232315132257e-05}
{"train_loss": 0.28028246760368347, "global_step": 60684, "epoch": 681, "lr": 8.831195064086251e-05}
{"train_loss": 0.20885610580444336, "global_step": 60685, "epoch": 681, "lr": 8.831157812525186e-05}
{"train_loss": 0.22081288695335388, "global_step": 60686, "epoch": 681, "lr": 8.831120560449068e-05}
{"train_loss": 0.24870753288269043, "global_step": 60687, "epoch": 681, "lr": 8.831083307857906e-05}
{"train_loss": 0.2300730049610138, "global_step": 60688, "epoch": 681, "lr": 8.831046054751699e-05}
{"train_loss": 0.23692989349365234, "global_step": 60689, "epoch": 681, "lr": 8.831008801130456e-05}
{"train_loss": 0.24117974936962128, "global_step": 60690, "epoch": 681, "lr": 8.83097154699418e-05}
{"train_loss": 0.30898308753967285, "global_step": 60691, "epoch": 681, "lr": 8.830934292342875e-05}
{"train_loss": 0.2493714541196823, "global_step": 60692, "epoch": 681, "lr": 8.83089703717655e-05}
{"train_loss": 0.18176817893981934, "global_step": 60693, "epoch": 681, "lr": 8.830859781495204e-05}
{"train_loss": 0.22054538130760193, "global_step": 60694, "epoch": 681, "lr": 8.830822525298848e-05}
{"train_loss": 0.23516158759593964, "global_step": 60695, "epoch": 681, "lr": 8.830785268587485e-05}
{"train_loss": 0.20002338290214539, "global_step": 60696, "epoch": 681, "lr": 8.830748011361119e-05}
{"train_loss": 0.23568618297576904, "global_step": 60697, "epoch": 681, "lr": 8.830710753619756e-05, "val_loss": 3.3596889972686768}
{"train_loss": 0.18412886559963226, "global_step": 60698, "epoch": 682, "lr": 8.830673495363401e-05}
{"train_loss": 0.21608735620975494, "global_step": 60699, "epoch": 682, "lr": 8.830636236592059e-05}
{"train_loss": 0.2734055817127228, "global_step": 60700, "epoch": 682, "lr": 8.830598977305733e-05}
{"train_loss": 0.2016114592552185, "global_step": 60701, "epoch": 682, "lr": 8.83056171750443e-05}
{"train_loss": 0.2515937089920044, "global_step": 60702, "epoch": 682, "lr": 8.830524457188153e-05}
{"train_loss": 0.18357573449611664, "global_step": 60703, "epoch": 682, "lr": 8.830487196356911e-05}
{"train_loss": 0.20999954640865326, "global_step": 60704, "epoch": 682, "lr": 8.830449935010707e-05}
{"train_loss": 0.24723844230175018, "global_step": 60705, "epoch": 682, "lr": 8.830412673149546e-05}
{"train_loss": 0.26903966069221497, "global_step": 60706, "epoch": 682, "lr": 8.83037541077343e-05}
{"train_loss": 0.27756842970848083, "global_step": 60707, "epoch": 682, "lr": 8.83033814788237e-05}
{"train_loss": 0.2638014554977417, "global_step": 60708, "epoch": 682, "lr": 8.830300884476365e-05}
{"train_loss": 0.2851921319961548, "global_step": 60709, "epoch": 682, "lr": 8.830263620555424e-05}
{"train_loss": 0.2208542674779892, "global_step": 60710, "epoch": 682, "lr": 8.830226356119551e-05}
{"train_loss": 0.26204073429107666, "global_step": 60711, "epoch": 682, "lr": 8.83018909116875e-05}
{"train_loss": 0.2217123806476593, "global_step": 60712, "epoch": 682, "lr": 8.830151825703027e-05}
{"train_loss": 0.21843221783638, "global_step": 60713, "epoch": 682, "lr": 8.830114559722387e-05}
{"train_loss": 0.2514628767967224, "global_step": 60714, "epoch": 682, "lr": 8.830077293226835e-05}
{"train_loss": 0.2648611068725586, "global_step": 60715, "epoch": 682, "lr": 8.830040026216375e-05}
{"train_loss": 0.21675333380699158, "global_step": 60716, "epoch": 682, "lr": 8.830002758691014e-05}
{"train_loss": 0.23692208528518677, "global_step": 60717, "epoch": 682, "lr": 8.829965490650755e-05}
{"train_loss": 0.26924318075180054, "global_step": 60718, "epoch": 682, "lr": 8.829928222095605e-05}
{"train_loss": 0.28561103343963623, "global_step": 60719, "epoch": 682, "lr": 8.829890953025566e-05}
{"train_loss": 0.2629164755344391, "global_step": 60720, "epoch": 682, "lr": 8.829853683440645e-05}
{"train_loss": 0.19680704176425934, "global_step": 60721, "epoch": 682, "lr": 8.829816413340848e-05}
{"train_loss": 0.16605164110660553, "global_step": 60722, "epoch": 682, "lr": 8.829779142726178e-05}
{"train_loss": 0.2700052261352539, "global_step": 60723, "epoch": 682, "lr": 8.829741871596641e-05}
{"train_loss": 0.20945756137371063, "global_step": 60724, "epoch": 682, "lr": 8.829704599952243e-05}
{"train_loss": 0.13941386342048645, "global_step": 60725, "epoch": 682, "lr": 8.829667327792985e-05}
{"train_loss": 0.15740744769573212, "global_step": 60726, "epoch": 682, "lr": 8.829630055118878e-05}
{"train_loss": 0.23704801499843597, "global_step": 60727, "epoch": 682, "lr": 8.829592781929924e-05}
{"train_loss": 0.2068680077791214, "global_step": 60728, "epoch": 682, "lr": 8.829555508226126e-05}
{"train_loss": 0.254701167345047, "global_step": 60729, "epoch": 682, "lr": 8.829518234007492e-05}
{"train_loss": 0.23775634169578552, "global_step": 60730, "epoch": 682, "lr": 8.829480959274025e-05}
{"train_loss": 0.20653533935546875, "global_step": 60731, "epoch": 682, "lr": 8.82944368402573e-05}
{"train_loss": 0.29602038860321045, "global_step": 60732, "epoch": 682, "lr": 8.829406408262615e-05}
{"train_loss": 0.18402986228466034, "global_step": 60733, "epoch": 682, "lr": 8.829369131984684e-05}
{"train_loss": 0.2870645821094513, "global_step": 60734, "epoch": 682, "lr": 8.829331855191938e-05}
{"train_loss": 0.2152547389268875, "global_step": 60735, "epoch": 682, "lr": 8.829294577884387e-05}
{"train_loss": 0.23353524506092072, "global_step": 60736, "epoch": 682, "lr": 8.829257300062034e-05}
{"train_loss": 0.19492974877357483, "global_step": 60737, "epoch": 682, "lr": 8.829220021724883e-05}
{"train_loss": 0.23222962021827698, "global_step": 60738, "epoch": 682, "lr": 8.829182742872941e-05}
{"train_loss": 0.29909461736679077, "global_step": 60739, "epoch": 682, "lr": 8.829145463506213e-05}
{"train_loss": 0.22059829533100128, "global_step": 60740, "epoch": 682, "lr": 8.8291081836247e-05}
{"train_loss": 0.1853848397731781, "global_step": 60741, "epoch": 682, "lr": 8.829070903228411e-05}
{"train_loss": 0.23818686604499817, "global_step": 60742, "epoch": 682, "lr": 8.829033622317351e-05}
{"train_loss": 0.23839126527309418, "global_step": 60743, "epoch": 682, "lr": 8.828996340891525e-05}
{"train_loss": 0.16673581302165985, "global_step": 60744, "epoch": 682, "lr": 8.828959058950937e-05}
{"train_loss": 0.2060052752494812, "global_step": 60745, "epoch": 682, "lr": 8.82892177649559e-05}
{"train_loss": 0.27027881145477295, "global_step": 60746, "epoch": 682, "lr": 8.828884493525493e-05}
{"train_loss": 0.2758682072162628, "global_step": 60747, "epoch": 682, "lr": 8.828847210040648e-05}
{"train_loss": 0.30898356437683105, "global_step": 60748, "epoch": 682, "lr": 8.828809926041062e-05}
{"train_loss": 0.20954951643943787, "global_step": 60749, "epoch": 682, "lr": 8.828772641526738e-05}
{"train_loss": 0.1783110350370407, "global_step": 60750, "epoch": 682, "lr": 8.828735356497683e-05}
{"train_loss": 0.26496943831443787, "global_step": 60751, "epoch": 682, "lr": 8.828698070953901e-05}
{"train_loss": 0.17874743044376373, "global_step": 60752, "epoch": 682, "lr": 8.828660784895398e-05}
{"train_loss": 0.16946543753147125, "global_step": 60753, "epoch": 682, "lr": 8.828623498322177e-05}
{"train_loss": 0.23145917057991028, "global_step": 60754, "epoch": 682, "lr": 8.828586211234244e-05}
{"train_loss": 0.3198069930076599, "global_step": 60755, "epoch": 682, "lr": 8.828548923631606e-05}
{"train_loss": 0.2433689534664154, "global_step": 60756, "epoch": 682, "lr": 8.828511635514266e-05}
{"train_loss": 0.20858252048492432, "global_step": 60757, "epoch": 682, "lr": 8.828474346882228e-05}
{"train_loss": 0.20170056819915771, "global_step": 60758, "epoch": 682, "lr": 8.828437057735499e-05}
{"train_loss": 0.30270665884017944, "global_step": 60759, "epoch": 682, "lr": 8.828399768074083e-05}
{"train_loss": 0.21131223440170288, "global_step": 60760, "epoch": 682, "lr": 8.828362477897986e-05}
{"train_loss": 0.2919836938381195, "global_step": 60761, "epoch": 682, "lr": 8.828325187207212e-05}
{"train_loss": 0.21629366278648376, "global_step": 60762, "epoch": 682, "lr": 8.828287896001765e-05}
{"train_loss": 0.28466206789016724, "global_step": 60763, "epoch": 682, "lr": 8.828250604281654e-05}
{"train_loss": 0.1510280966758728, "global_step": 60764, "epoch": 682, "lr": 8.82821331204688e-05}
{"train_loss": 0.3142799437046051, "global_step": 60765, "epoch": 682, "lr": 8.828176019297449e-05}
{"train_loss": 0.19479943811893463, "global_step": 60766, "epoch": 682, "lr": 8.828138726033365e-05}
{"train_loss": 0.2232719361782074, "global_step": 60767, "epoch": 682, "lr": 8.828101432254637e-05}
{"train_loss": 0.1513749361038208, "global_step": 60768, "epoch": 682, "lr": 8.828064137961267e-05}
{"train_loss": 0.29743969440460205, "global_step": 60769, "epoch": 682, "lr": 8.828026843153259e-05}
{"train_loss": 0.20865462720394135, "global_step": 60770, "epoch": 682, "lr": 8.827989547830621e-05}
{"train_loss": 0.21021313965320587, "global_step": 60771, "epoch": 682, "lr": 8.827952251993357e-05}
{"train_loss": 0.23630493879318237, "global_step": 60772, "epoch": 682, "lr": 8.82791495564147e-05}
{"train_loss": 0.25001275539398193, "global_step": 60773, "epoch": 682, "lr": 8.827877658774966e-05}
{"train_loss": 0.15818776190280914, "global_step": 60774, "epoch": 682, "lr": 8.827840361393852e-05}
{"train_loss": 0.2570873498916626, "global_step": 60775, "epoch": 682, "lr": 8.82780306349813e-05}
{"train_loss": 0.19581153988838196, "global_step": 60776, "epoch": 682, "lr": 8.827765765087807e-05}
{"train_loss": 0.2842443585395813, "global_step": 60777, "epoch": 682, "lr": 8.827728466162889e-05}
{"train_loss": 0.23693053424358368, "global_step": 60778, "epoch": 682, "lr": 8.827691166723378e-05}
{"train_loss": 0.22520115971565247, "global_step": 60779, "epoch": 682, "lr": 8.827653866769281e-05}
{"train_loss": 0.2163989543914795, "global_step": 60780, "epoch": 682, "lr": 8.827616566300602e-05}
{"train_loss": 0.2509838044643402, "global_step": 60781, "epoch": 682, "lr": 8.827579265317349e-05}
{"train_loss": 0.22137939929962158, "global_step": 60782, "epoch": 682, "lr": 8.827541963819522e-05}
{"train_loss": 0.2221519649028778, "global_step": 60783, "epoch": 682, "lr": 8.827504661807129e-05}
{"train_loss": 0.22600917518138885, "global_step": 60784, "epoch": 682, "lr": 8.827467359280176e-05}
{"train_loss": 0.29065167903900146, "global_step": 60785, "epoch": 682, "lr": 8.827430056238665e-05}
{"train_loss": 0.23226789242765877, "global_step": 60786, "epoch": 682, "lr": 8.827392752682604e-05, "val_loss": 3.43186616897583}
{"train_loss": 0.26368141174316406, "global_step": 60787, "epoch": 683, "lr": 8.827355448611996e-05}
{"train_loss": 0.2853594124317169, "global_step": 60788, "epoch": 683, "lr": 8.827318144026848e-05}
{"train_loss": 0.2063809037208557, "global_step": 60789, "epoch": 683, "lr": 8.82728083892716e-05}
{"train_loss": 0.15514017641544342, "global_step": 60790, "epoch": 683, "lr": 8.827243533312946e-05}
{"train_loss": 0.15453936159610748, "global_step": 60791, "epoch": 683, "lr": 8.827206227184202e-05}
{"train_loss": 0.26994848251342773, "global_step": 60792, "epoch": 683, "lr": 8.827168920540939e-05}
{"train_loss": 0.24925312399864197, "global_step": 60793, "epoch": 683, "lr": 8.827131613383158e-05}
{"train_loss": 0.13704070448875427, "global_step": 60794, "epoch": 683, "lr": 8.827094305710866e-05}
{"train_loss": 0.19517378509044647, "global_step": 60795, "epoch": 683, "lr": 8.827056997524068e-05}
{"train_loss": 0.3033587336540222, "global_step": 60796, "epoch": 683, "lr": 8.827019688822769e-05}
{"train_loss": 0.2748931050300598, "global_step": 60797, "epoch": 683, "lr": 8.826982379606974e-05}
{"train_loss": 0.22529198229312897, "global_step": 60798, "epoch": 683, "lr": 8.826945069876689e-05}
{"train_loss": 0.297037810087204, "global_step": 60799, "epoch": 683, "lr": 8.826907759631916e-05}
{"train_loss": 0.2546538710594177, "global_step": 60800, "epoch": 683, "lr": 8.826870448872662e-05}
{"train_loss": 0.25212687253952026, "global_step": 60801, "epoch": 683, "lr": 8.826833137598933e-05}
{"train_loss": 0.27019742131233215, "global_step": 60802, "epoch": 683, "lr": 8.826795825810733e-05}
{"train_loss": 0.24552486836910248, "global_step": 60803, "epoch": 683, "lr": 8.826758513508066e-05}
{"train_loss": 0.2808469533920288, "global_step": 60804, "epoch": 683, "lr": 8.826721200690938e-05}
{"train_loss": 0.28982996940612793, "global_step": 60805, "epoch": 683, "lr": 8.826683887359355e-05}
{"train_loss": 0.23786847293376923, "global_step": 60806, "epoch": 683, "lr": 8.82664657351332e-05}
{"train_loss": 0.19055871665477753, "global_step": 60807, "epoch": 683, "lr": 8.82660925915284e-05}
{"train_loss": 0.2603098154067993, "global_step": 60808, "epoch": 683, "lr": 8.826571944277916e-05}
{"train_loss": 0.22935940325260162, "global_step": 60809, "epoch": 683, "lr": 8.826534628888561e-05}
{"train_loss": 0.19530877470970154, "global_step": 60810, "epoch": 683, "lr": 8.826497312984772e-05}
{"train_loss": 0.198715940117836, "global_step": 60811, "epoch": 683, "lr": 8.826459996566558e-05}
{"train_loss": 0.22846738994121552, "global_step": 60812, "epoch": 683, "lr": 8.826422679633922e-05}
{"train_loss": 0.2460169494152069, "global_step": 60813, "epoch": 683, "lr": 8.826385362186871e-05}
{"train_loss": 0.2290344387292862, "global_step": 60814, "epoch": 683, "lr": 8.826348044225409e-05}
{"train_loss": 0.3037883937358856, "global_step": 60815, "epoch": 683, "lr": 8.82631072574954e-05}
{"train_loss": 0.14224353432655334, "global_step": 60816, "epoch": 683, "lr": 8.826273406759271e-05}
{"train_loss": 0.2674178183078766, "global_step": 60817, "epoch": 683, "lr": 8.826236087254607e-05}
{"train_loss": 0.22591078281402588, "global_step": 60818, "epoch": 683, "lr": 8.826198767235552e-05}
{"train_loss": 0.19648070633411407, "global_step": 60819, "epoch": 683, "lr": 8.82616144670211e-05}
{"train_loss": 0.15100903809070587, "global_step": 60820, "epoch": 683, "lr": 8.826124125654289e-05}
{"train_loss": 0.22669343650341034, "global_step": 60821, "epoch": 683, "lr": 8.82608680409209e-05}
{"train_loss": 0.2734485864639282, "global_step": 60822, "epoch": 683, "lr": 8.826049482015521e-05}
{"train_loss": 0.24948635697364807, "global_step": 60823, "epoch": 683, "lr": 8.826012159424589e-05}
{"train_loss": 0.25542229413986206, "global_step": 60824, "epoch": 683, "lr": 8.825974836319293e-05}
{"train_loss": 0.1403760313987732, "global_step": 60825, "epoch": 683, "lr": 8.825937512699642e-05}
{"train_loss": 0.1987295299768448, "global_step": 60826, "epoch": 683, "lr": 8.825900188565641e-05}
{"train_loss": 0.30154332518577576, "global_step": 60827, "epoch": 683, "lr": 8.825862863917294e-05}
{"train_loss": 0.2409479022026062, "global_step": 60828, "epoch": 683, "lr": 8.825825538754607e-05}
{"train_loss": 0.17775866389274597, "global_step": 60829, "epoch": 683, "lr": 8.825788213077583e-05}
{"train_loss": 0.18122631311416626, "global_step": 60830, "epoch": 683, "lr": 8.82575088688623e-05}
{"train_loss": 0.15689262747764587, "global_step": 60831, "epoch": 683, "lr": 8.825713560180549e-05}
{"train_loss": 0.22477832436561584, "global_step": 60832, "epoch": 683, "lr": 8.825676232960549e-05}
{"train_loss": 0.22576187551021576, "global_step": 60833, "epoch": 683, "lr": 8.825638905226235e-05}
{"train_loss": 0.2555384635925293, "global_step": 60834, "epoch": 683, "lr": 8.825601576977609e-05}
{"train_loss": 0.1848863810300827, "global_step": 60835, "epoch": 683, "lr": 8.825564248214678e-05}
{"train_loss": 0.2792857587337494, "global_step": 60836, "epoch": 683, "lr": 8.825526918937446e-05}
{"train_loss": 0.24125142395496368, "global_step": 60837, "epoch": 683, "lr": 8.825489589145918e-05}
{"train_loss": 0.2363193929195404, "global_step": 60838, "epoch": 683, "lr": 8.825452258840099e-05}
{"train_loss": 0.27969086170196533, "global_step": 60839, "epoch": 683, "lr": 8.825414928019996e-05}
{"train_loss": 0.2332451194524765, "global_step": 60840, "epoch": 683, "lr": 8.825377596685613e-05}
{"train_loss": 0.2568267285823822, "global_step": 60841, "epoch": 683, "lr": 8.825340264836953e-05}
{"train_loss": 0.32210925221443176, "global_step": 60842, "epoch": 683, "lr": 8.825302932474024e-05}
{"train_loss": 0.24129335582256317, "global_step": 60843, "epoch": 683, "lr": 8.825265599596829e-05}
{"train_loss": 0.19728128612041473, "global_step": 60844, "epoch": 683, "lr": 8.825228266205374e-05}
{"train_loss": 0.21080686151981354, "global_step": 60845, "epoch": 683, "lr": 8.825190932299663e-05}
{"train_loss": 0.2916894257068634, "global_step": 60846, "epoch": 683, "lr": 8.825153597879702e-05}
{"train_loss": 0.253893107175827, "global_step": 60847, "epoch": 683, "lr": 8.825116262945497e-05}
{"train_loss": 0.2674707770347595, "global_step": 60848, "epoch": 683, "lr": 8.82507892749705e-05}
{"train_loss": 0.28855273127555847, "global_step": 60849, "epoch": 683, "lr": 8.825041591534369e-05}
{"train_loss": 0.21683822572231293, "global_step": 60850, "epoch": 683, "lr": 8.825004255057456e-05}
{"train_loss": 0.21112897992134094, "global_step": 60851, "epoch": 683, "lr": 8.82496691806632e-05}
{"train_loss": 0.267531156539917, "global_step": 60852, "epoch": 683, "lr": 8.824929580560962e-05}
{"train_loss": 0.30627456307411194, "global_step": 60853, "epoch": 683, "lr": 8.82489224254139e-05}
{"train_loss": 0.23601526021957397, "global_step": 60854, "epoch": 683, "lr": 8.824854904007607e-05}
{"train_loss": 0.2392740100622177, "global_step": 60855, "epoch": 683, "lr": 8.82481756495962e-05}
{"train_loss": 0.2238835245370865, "global_step": 60856, "epoch": 683, "lr": 8.824780225397432e-05}
{"train_loss": 0.215719535946846, "global_step": 60857, "epoch": 683, "lr": 8.82474288532105e-05}
{"train_loss": 0.2345283329486847, "global_step": 60858, "epoch": 683, "lr": 8.824705544730476e-05}
{"train_loss": 0.16970211267471313, "global_step": 60859, "epoch": 683, "lr": 8.824668203625719e-05}
{"train_loss": 0.21216075122356415, "global_step": 60860, "epoch": 683, "lr": 8.824630862006781e-05}
{"train_loss": 0.24327285587787628, "global_step": 60861, "epoch": 683, "lr": 8.824593519873667e-05}
{"train_loss": 0.2449132800102234, "global_step": 60862, "epoch": 683, "lr": 8.824556177226385e-05}
{"train_loss": 0.2768779695034027, "global_step": 60863, "epoch": 683, "lr": 8.824518834064937e-05}
{"train_loss": 0.2632181644439697, "global_step": 60864, "epoch": 683, "lr": 8.82448149038933e-05}
{"train_loss": 0.379987508058548, "global_step": 60865, "epoch": 683, "lr": 8.824444146199567e-05}
{"train_loss": 0.21126240491867065, "global_step": 60866, "epoch": 683, "lr": 8.824406801495653e-05}
{"train_loss": 0.19585998356342316, "global_step": 60867, "epoch": 683, "lr": 8.824369456277596e-05}
{"train_loss": 0.3178689777851105, "global_step": 60868, "epoch": 683, "lr": 8.824332110545398e-05}
{"train_loss": 0.20249676704406738, "global_step": 60869, "epoch": 683, "lr": 8.824294764299066e-05}
{"train_loss": 0.28165915608406067, "global_step": 60870, "epoch": 683, "lr": 8.824257417538604e-05}
{"train_loss": 0.20503970980644226, "global_step": 60871, "epoch": 683, "lr": 8.824220070264018e-05}
{"train_loss": 0.15519124269485474, "global_step": 60872, "epoch": 683, "lr": 8.824182722475312e-05}
{"train_loss": 0.19223371148109436, "global_step": 60873, "epoch": 683, "lr": 8.82414537417249e-05}
{"train_loss": 0.264857679605484, "global_step": 60874, "epoch": 683, "lr": 8.824108025355558e-05}
{"train_loss": 0.23589065101709258, "global_step": 60875, "epoch": 683, "lr": 8.824070676024521e-05, "val_loss": 3.474008798599243}
{"train_loss": 0.36397185921669006, "global_step": 60876, "epoch": 684, "lr": 8.824033326179386e-05}
{"train_loss": 0.17973127961158752, "global_step": 60877, "epoch": 684, "lr": 8.823995975820156e-05}
{"train_loss": 0.19044210016727448, "global_step": 60878, "epoch": 684, "lr": 8.823958624946837e-05}
{"train_loss": 0.2507602870464325, "global_step": 60879, "epoch": 684, "lr": 8.823921273559432e-05}
{"train_loss": 0.2384798377752304, "global_step": 60880, "epoch": 684, "lr": 8.823883921657948e-05}
{"train_loss": 0.2433634251356125, "global_step": 60881, "epoch": 684, "lr": 8.823846569242389e-05}
{"train_loss": 0.21957622468471527, "global_step": 60882, "epoch": 684, "lr": 8.823809216312761e-05}
{"train_loss": 0.2313375473022461, "global_step": 60883, "epoch": 684, "lr": 8.823771862869068e-05}
{"train_loss": 0.19684821367263794, "global_step": 60884, "epoch": 684, "lr": 8.823734508911316e-05}
{"train_loss": 0.25226038694381714, "global_step": 60885, "epoch": 684, "lr": 8.823697154439509e-05}
{"train_loss": 0.2346096634864807, "global_step": 60886, "epoch": 684, "lr": 8.823659799453653e-05}
{"train_loss": 0.28288424015045166, "global_step": 60887, "epoch": 684, "lr": 8.823622443953751e-05}
{"train_loss": 0.2427714467048645, "global_step": 60888, "epoch": 684, "lr": 8.823585087939812e-05}
{"train_loss": 0.3191347122192383, "global_step": 60889, "epoch": 684, "lr": 8.823547731411838e-05}
{"train_loss": 0.21843983232975006, "global_step": 60890, "epoch": 684, "lr": 8.823510374369834e-05}
{"train_loss": 0.3631981313228607, "global_step": 60891, "epoch": 684, "lr": 8.823473016813805e-05}
{"train_loss": 0.2688020169734955, "global_step": 60892, "epoch": 684, "lr": 8.823435658743757e-05}
{"train_loss": 0.18330861628055573, "global_step": 60893, "epoch": 684, "lr": 8.823398300159696e-05}
{"train_loss": 0.18306505680084229, "global_step": 60894, "epoch": 684, "lr": 8.823360941061623e-05}
{"train_loss": 0.25556373596191406, "global_step": 60895, "epoch": 684, "lr": 8.823323581449547e-05}
{"train_loss": 0.2934706509113312, "global_step": 60896, "epoch": 684, "lr": 8.823286221323472e-05}
{"train_loss": 0.23175038397312164, "global_step": 60897, "epoch": 684, "lr": 8.823248860683403e-05}
{"train_loss": 0.25382909178733826, "global_step": 60898, "epoch": 684, "lr": 8.823211499529346e-05}
{"train_loss": 0.40148526430130005, "global_step": 60899, "epoch": 684, "lr": 8.823174137861303e-05}
{"train_loss": 0.2166305035352707, "global_step": 60900, "epoch": 684, "lr": 8.823136775679282e-05}
{"train_loss": 0.26578953862190247, "global_step": 60901, "epoch": 684, "lr": 8.823099412983287e-05}
{"train_loss": 0.160588338971138, "global_step": 60902, "epoch": 684, "lr": 8.82306204977332e-05}
{"train_loss": 0.23946750164031982, "global_step": 60903, "epoch": 684, "lr": 8.823024686049392e-05}
{"train_loss": 0.1651100069284439, "global_step": 60904, "epoch": 684, "lr": 8.822987321811503e-05}
{"train_loss": 0.167519673705101, "global_step": 60905, "epoch": 684, "lr": 8.822949957059661e-05}
{"train_loss": 0.23883472383022308, "global_step": 60906, "epoch": 684, "lr": 8.82291259179387e-05}
{"train_loss": 0.19793835282325745, "global_step": 60907, "epoch": 684, "lr": 8.822875226014136e-05}
{"train_loss": 0.1810981035232544, "global_step": 60908, "epoch": 684, "lr": 8.822837859720461e-05}
{"train_loss": 0.2705935835838318, "global_step": 60909, "epoch": 684, "lr": 8.822800492912853e-05}
{"train_loss": 0.2505686581134796, "global_step": 60910, "epoch": 684, "lr": 8.822763125591316e-05}
{"train_loss": 0.20585383474826813, "global_step": 60911, "epoch": 684, "lr": 8.822725757755855e-05}
{"train_loss": 0.24259938299655914, "global_step": 60912, "epoch": 684, "lr": 8.822688389406474e-05}
{"train_loss": 0.1661558598279953, "global_step": 60913, "epoch": 684, "lr": 8.822651020543181e-05}
{"train_loss": 0.14630256593227386, "global_step": 60914, "epoch": 684, "lr": 8.822613651165978e-05}
{"train_loss": 0.2128584384918213, "global_step": 60915, "epoch": 684, "lr": 8.822576281274872e-05}
{"train_loss": 0.17926080524921417, "global_step": 60916, "epoch": 684, "lr": 8.822538910869867e-05}
{"train_loss": 0.26310616731643677, "global_step": 60917, "epoch": 684, "lr": 8.822501539950969e-05}
{"train_loss": 0.2713705897331238, "global_step": 60918, "epoch": 684, "lr": 8.822464168518181e-05}
{"train_loss": 0.22091564536094666, "global_step": 60919, "epoch": 684, "lr": 8.82242679657151e-05}
{"train_loss": 0.3204628825187683, "global_step": 60920, "epoch": 684, "lr": 8.82238942411096e-05}
{"train_loss": 0.19665934145450592, "global_step": 60921, "epoch": 684, "lr": 8.822352051136536e-05}
{"train_loss": 0.28669631481170654, "global_step": 60922, "epoch": 684, "lr": 8.822314677648242e-05}
{"train_loss": 0.29720550775527954, "global_step": 60923, "epoch": 684, "lr": 8.822277303646088e-05}
{"train_loss": 0.2181762009859085, "global_step": 60924, "epoch": 684, "lr": 8.822239929130073e-05}
{"train_loss": 0.2869735062122345, "global_step": 60925, "epoch": 684, "lr": 8.822202554100205e-05}
{"train_loss": 0.18440920114517212, "global_step": 60926, "epoch": 684, "lr": 8.822165178556487e-05}
{"train_loss": 0.22544969618320465, "global_step": 60927, "epoch": 684, "lr": 8.822127802498927e-05}
{"train_loss": 0.29442888498306274, "global_step": 60928, "epoch": 684, "lr": 8.822090425927529e-05}
{"train_loss": 0.21878743171691895, "global_step": 60929, "epoch": 684, "lr": 8.822053048842297e-05}
{"train_loss": 0.23117642104625702, "global_step": 60930, "epoch": 684, "lr": 8.822015671243236e-05}
{"train_loss": 0.272522896528244, "global_step": 60931, "epoch": 684, "lr": 8.821978293130352e-05}
{"train_loss": 0.16819290816783905, "global_step": 60932, "epoch": 684, "lr": 8.821940914503648e-05}
{"train_loss": 0.285364031791687, "global_step": 60933, "epoch": 684, "lr": 8.821903535363134e-05}
{"train_loss": 0.22666990756988525, "global_step": 60934, "epoch": 684, "lr": 8.821866155708808e-05}
{"train_loss": 0.22212116420269012, "global_step": 60935, "epoch": 684, "lr": 8.821828775540681e-05}
{"train_loss": 0.19160017371177673, "global_step": 60936, "epoch": 684, "lr": 8.821791394858756e-05}
{"train_loss": 0.14905232191085815, "global_step": 60937, "epoch": 684, "lr": 8.821754013663037e-05}
{"train_loss": 0.21303889155387878, "global_step": 60938, "epoch": 684, "lr": 8.821716631953529e-05}
{"train_loss": 0.27908799052238464, "global_step": 60939, "epoch": 684, "lr": 8.82167924973024e-05}
{"train_loss": 0.30715861916542053, "global_step": 60940, "epoch": 684, "lr": 8.82164186699317e-05}
{"train_loss": 0.2096264511346817, "global_step": 60941, "epoch": 684, "lr": 8.821604483742328e-05}
{"train_loss": 0.2213296890258789, "global_step": 60942, "epoch": 684, "lr": 8.82156709997772e-05}
{"train_loss": 0.29391366243362427, "global_step": 60943, "epoch": 684, "lr": 8.821529715699345e-05}
{"train_loss": 0.37881308794021606, "global_step": 60944, "epoch": 684, "lr": 8.821492330907214e-05}
{"train_loss": 0.1917577087879181, "global_step": 60945, "epoch": 684, "lr": 8.82145494560133e-05}
{"train_loss": 0.30855146050453186, "global_step": 60946, "epoch": 684, "lr": 8.821417559781697e-05}
{"train_loss": 0.20292384922504425, "global_step": 60947, "epoch": 684, "lr": 8.821380173448321e-05}
{"train_loss": 0.22234854102134705, "global_step": 60948, "epoch": 684, "lr": 8.821342786601207e-05}
{"train_loss": 0.2586001753807068, "global_step": 60949, "epoch": 684, "lr": 8.821305399240362e-05}
{"train_loss": 0.23247987031936646, "global_step": 60950, "epoch": 684, "lr": 8.821268011365787e-05}
{"train_loss": 0.22467321157455444, "global_step": 60951, "epoch": 684, "lr": 8.82123062297749e-05}
{"train_loss": 0.2952333092689514, "global_step": 60952, "epoch": 684, "lr": 8.821193234075476e-05}
{"train_loss": 0.2879989445209503, "global_step": 60953, "epoch": 684, "lr": 8.821155844659748e-05}
{"train_loss": 0.285352885723114, "global_step": 60954, "epoch": 684, "lr": 8.821118454730311e-05}
{"train_loss": 0.2314891815185547, "global_step": 60955, "epoch": 684, "lr": 8.821081064287173e-05}
{"train_loss": 0.2808387875556946, "global_step": 60956, "epoch": 684, "lr": 8.821043673330337e-05}
{"train_loss": 0.27305540442466736, "global_step": 60957, "epoch": 684, "lr": 8.821006281859807e-05}
{"train_loss": 0.24317914247512817, "global_step": 60958, "epoch": 684, "lr": 8.820968889875592e-05}
{"train_loss": 0.3183611035346985, "global_step": 60959, "epoch": 684, "lr": 8.820931497377691e-05}
{"train_loss": 0.18139582872390747, "global_step": 60960, "epoch": 684, "lr": 8.820894104366115e-05}
{"train_loss": 0.18285007774829865, "global_step": 60961, "epoch": 684, "lr": 8.820856710840865e-05}
{"train_loss": 0.21955262124538422, "global_step": 60962, "epoch": 684, "lr": 8.820819316801947e-05}
{"train_loss": 0.27606552839279175, "global_step": 60963, "epoch": 684, "lr": 8.820781922249367e-05}
{"train_loss": 0.24189233495278306, "global_step": 60964, "epoch": 684, "lr": 8.820744527183129e-05, "val_loss": 3.5174572467803955}
{"train_loss": 0.2768651843070984, "global_step": 60965, "epoch": 685, "lr": 8.82070713160324e-05}
{"train_loss": 0.3294103741645813, "global_step": 60966, "epoch": 685, "lr": 8.820669735509701e-05}
{"train_loss": 0.20464779436588287, "global_step": 60967, "epoch": 685, "lr": 8.820632338902521e-05}
{"train_loss": 0.21253079175949097, "global_step": 60968, "epoch": 685, "lr": 8.820594941781703e-05}
{"train_loss": 0.25974559783935547, "global_step": 60969, "epoch": 685, "lr": 8.820557544147255e-05}
{"train_loss": 0.170296311378479, "global_step": 60970, "epoch": 685, "lr": 8.820520145999176e-05}
{"train_loss": 0.3077344000339508, "global_step": 60971, "epoch": 685, "lr": 8.820482747337477e-05}
{"train_loss": 0.18884654343128204, "global_step": 60972, "epoch": 685, "lr": 8.82044534816216e-05}
{"train_loss": 0.2504633069038391, "global_step": 60973, "epoch": 685, "lr": 8.820407948473231e-05}
{"train_loss": 0.3594866693019867, "global_step": 60974, "epoch": 685, "lr": 8.820370548270693e-05}
{"train_loss": 0.27405306696891785, "global_step": 60975, "epoch": 685, "lr": 8.820333147554555e-05}
{"train_loss": 0.25746434926986694, "global_step": 60976, "epoch": 685, "lr": 8.820295746324818e-05}
{"train_loss": 0.24201829731464386, "global_step": 60977, "epoch": 685, "lr": 8.82025834458149e-05}
{"train_loss": 0.18708735704421997, "global_step": 60978, "epoch": 685, "lr": 8.820220942324575e-05}
{"train_loss": 0.27893519401550293, "global_step": 60979, "epoch": 685, "lr": 8.820183539554077e-05}
{"train_loss": 0.28026917576789856, "global_step": 60980, "epoch": 685, "lr": 8.820146136270002e-05}
{"train_loss": 0.3063320517539978, "global_step": 60981, "epoch": 685, "lr": 8.820108732472355e-05}
{"train_loss": 0.22830908000469208, "global_step": 60982, "epoch": 685, "lr": 8.820071328161142e-05}
{"train_loss": 0.27457869052886963, "global_step": 60983, "epoch": 685, "lr": 8.820033923336367e-05}
{"train_loss": 0.2116469442844391, "global_step": 60984, "epoch": 685, "lr": 8.819996517998033e-05}
{"train_loss": 0.28256353735923767, "global_step": 60985, "epoch": 685, "lr": 8.819959112146147e-05}
{"train_loss": 0.24686037003993988, "global_step": 60986, "epoch": 685, "lr": 8.819921705780715e-05}
{"train_loss": 0.2325039505958557, "global_step": 60987, "epoch": 685, "lr": 8.819884298901742e-05}
{"train_loss": 0.26862832903862, "global_step": 60988, "epoch": 685, "lr": 8.819846891509231e-05}
{"train_loss": 0.3318322002887726, "global_step": 60989, "epoch": 685, "lr": 8.819809483603189e-05}
{"train_loss": 0.29678553342819214, "global_step": 60990, "epoch": 685, "lr": 8.81977207518362e-05}
{"train_loss": 0.16227270662784576, "global_step": 60991, "epoch": 685, "lr": 8.819734666250527e-05}
{"train_loss": 0.27671876549720764, "global_step": 60992, "epoch": 685, "lr": 8.81969725680392e-05}
{"train_loss": 0.24395392835140228, "global_step": 60993, "epoch": 685, "lr": 8.8196598468438e-05}
{"train_loss": 0.20371687412261963, "global_step": 60994, "epoch": 685, "lr": 8.819622436370174e-05}
{"train_loss": 0.21950344741344452, "global_step": 60995, "epoch": 685, "lr": 8.819585025383046e-05}
{"train_loss": 0.24795468151569366, "global_step": 60996, "epoch": 685, "lr": 8.819547613882421e-05}
{"train_loss": 0.23557692766189575, "global_step": 60997, "epoch": 685, "lr": 8.819510201868304e-05}
{"train_loss": 0.2462397664785385, "global_step": 60998, "epoch": 685, "lr": 8.819472789340701e-05}
{"train_loss": 0.23482371866703033, "global_step": 60999, "epoch": 685, "lr": 8.819435376299617e-05}
{"train_loss": 0.26842236518859863, "global_step": 61000, "epoch": 685, "lr": 8.819397962745056e-05}
{"train_loss": 0.2557990252971649, "global_step": 61001, "epoch": 685, "lr": 8.819360548677023e-05}
{"train_loss": 0.24220696091651917, "global_step": 61002, "epoch": 685, "lr": 8.819323134095525e-05}
{"train_loss": 0.21520112454891205, "global_step": 61003, "epoch": 685, "lr": 8.819285719000565e-05}
{"train_loss": 0.2558872699737549, "global_step": 61004, "epoch": 685, "lr": 8.819248303392147e-05}
{"train_loss": 0.26625359058380127, "global_step": 61005, "epoch": 685, "lr": 8.819210887270279e-05}
{"train_loss": 0.3099154829978943, "global_step": 61006, "epoch": 685, "lr": 8.819173470634964e-05}
{"train_loss": 0.349643349647522, "global_step": 61007, "epoch": 685, "lr": 8.819136053486209e-05}
{"train_loss": 0.26713067293167114, "global_step": 61008, "epoch": 685, "lr": 8.819098635824017e-05}
{"train_loss": 0.24659456312656403, "global_step": 61009, "epoch": 685, "lr": 8.819061217648392e-05}
{"train_loss": 0.33186057209968567, "global_step": 61010, "epoch": 685, "lr": 8.819023798959343e-05}
{"train_loss": 0.27426913380622864, "global_step": 61011, "epoch": 685, "lr": 8.818986379756873e-05}
{"train_loss": 0.2838601768016815, "global_step": 61012, "epoch": 685, "lr": 8.818948960040985e-05}
{"train_loss": 0.2651013731956482, "global_step": 61013, "epoch": 685, "lr": 8.818911539811688e-05}
{"train_loss": 0.3037312924861908, "global_step": 61014, "epoch": 685, "lr": 8.818874119068983e-05}
{"train_loss": 0.28329429030418396, "global_step": 61015, "epoch": 685, "lr": 8.818836697812878e-05}
{"train_loss": 0.2592897415161133, "global_step": 61016, "epoch": 685, "lr": 8.818799276043376e-05}
{"train_loss": 0.2763170301914215, "global_step": 61017, "epoch": 685, "lr": 8.818761853760483e-05}
{"train_loss": 0.31128472089767456, "global_step": 61018, "epoch": 685, "lr": 8.818724430964204e-05}
{"train_loss": 0.3285619914531708, "global_step": 61019, "epoch": 685, "lr": 8.818687007654545e-05}
{"train_loss": 0.25072652101516724, "global_step": 61020, "epoch": 685, "lr": 8.81864958383151e-05}
{"train_loss": 0.25138142704963684, "global_step": 61021, "epoch": 685, "lr": 8.818612159495103e-05}
{"train_loss": 0.17538920044898987, "global_step": 61022, "epoch": 685, "lr": 8.81857473464533e-05}
{"train_loss": 0.2738954424858093, "global_step": 61023, "epoch": 685, "lr": 8.818537309282197e-05}
{"train_loss": 0.3042702376842499, "global_step": 61024, "epoch": 685, "lr": 8.81849988340571e-05}
{"train_loss": 0.2356535941362381, "global_step": 61025, "epoch": 685, "lr": 8.81846245701587e-05}
{"train_loss": 0.2802543044090271, "global_step": 61026, "epoch": 685, "lr": 8.818425030112685e-05}
{"train_loss": 0.23657798767089844, "global_step": 61027, "epoch": 685, "lr": 8.818387602696159e-05}
{"train_loss": 0.26458439230918884, "global_step": 61028, "epoch": 685, "lr": 8.818350174766297e-05}
{"train_loss": 0.18896333873271942, "global_step": 61029, "epoch": 685, "lr": 8.818312746323106e-05}
{"train_loss": 0.26045525074005127, "global_step": 61030, "epoch": 685, "lr": 8.818275317366587e-05}
{"train_loss": 0.17336329817771912, "global_step": 61031, "epoch": 685, "lr": 8.818237887896748e-05}
{"train_loss": 0.2461490035057068, "global_step": 61032, "epoch": 685, "lr": 8.818200457913594e-05}
{"train_loss": 0.23740138113498688, "global_step": 61033, "epoch": 685, "lr": 8.81816302741713e-05}
{"train_loss": 0.3044026494026184, "global_step": 61034, "epoch": 685, "lr": 8.81812559640736e-05}
{"train_loss": 0.17394258081912994, "global_step": 61035, "epoch": 685, "lr": 8.81808816488429e-05}
{"train_loss": 0.24785110354423523, "global_step": 61036, "epoch": 685, "lr": 8.818050732847923e-05}
{"train_loss": 0.17187821865081787, "global_step": 61037, "epoch": 685, "lr": 8.818013300298266e-05}
{"train_loss": 0.2634560763835907, "global_step": 61038, "epoch": 685, "lr": 8.817975867235326e-05}
{"train_loss": 0.21124407649040222, "global_step": 61039, "epoch": 685, "lr": 8.817938433659103e-05}
{"train_loss": 0.15439914166927338, "global_step": 61040, "epoch": 685, "lr": 8.817900999569605e-05}
{"train_loss": 0.2460729330778122, "global_step": 61041, "epoch": 685, "lr": 8.81786356496684e-05}
{"train_loss": 0.21772688627243042, "global_step": 61042, "epoch": 685, "lr": 8.817826129850805e-05}
{"train_loss": 0.3222430348396301, "global_step": 61043, "epoch": 685, "lr": 8.817788694221514e-05}
{"train_loss": 0.26283225417137146, "global_step": 61044, "epoch": 685, "lr": 8.817751258078966e-05}
{"train_loss": 0.40768560767173767, "global_step": 61045, "epoch": 685, "lr": 8.817713821423166e-05}
{"train_loss": 0.27393725514411926, "global_step": 61046, "epoch": 685, "lr": 8.817676384254124e-05}
{"train_loss": 0.2743520438671112, "global_step": 61047, "epoch": 685, "lr": 8.81763894657184e-05}
{"train_loss": 0.26997125148773193, "global_step": 61048, "epoch": 685, "lr": 8.817601508376322e-05}
{"train_loss": 0.30543822050094604, "global_step": 61049, "epoch": 685, "lr": 8.817564069667574e-05}
{"train_loss": 0.2547915279865265, "global_step": 61050, "epoch": 685, "lr": 8.8175266304456e-05}
{"train_loss": 0.27086228132247925, "global_step": 61051, "epoch": 685, "lr": 8.817489190710407e-05}
{"train_loss": 0.31194961071014404, "global_step": 61052, "epoch": 685, "lr": 8.817451750462e-05}
{"train_loss": 0.2590929160961944, "global_step": 61053, "epoch": 685, "lr": 8.817414309700381e-05, "val_loss": 3.4364013671875, "train_action_mse_error": 30.139419555664062}
{"train_loss": 0.16455145180225372, "global_step": 61054, "epoch": 686, "lr": 8.817376868425558e-05}
{"train_loss": 0.22348417341709137, "global_step": 61055, "epoch": 686, "lr": 8.817339426637536e-05}
{"train_loss": 0.20174485445022583, "global_step": 61056, "epoch": 686, "lr": 8.81730198433632e-05}
{"train_loss": 0.22629287838935852, "global_step": 61057, "epoch": 686, "lr": 8.817264541521911e-05}
{"train_loss": 0.2501521706581116, "global_step": 61058, "epoch": 686, "lr": 8.817227098194318e-05}
{"train_loss": 0.2449704259634018, "global_step": 61059, "epoch": 686, "lr": 8.817189654353548e-05}
{"train_loss": 0.21984706819057465, "global_step": 61060, "epoch": 686, "lr": 8.817152209999602e-05}
{"train_loss": 0.23115713894367218, "global_step": 61061, "epoch": 686, "lr": 8.817114765132486e-05}
{"train_loss": 0.19634081423282623, "global_step": 61062, "epoch": 686, "lr": 8.817077319752206e-05}
{"train_loss": 0.2796204686164856, "global_step": 61063, "epoch": 686, "lr": 8.817039873858766e-05}
{"train_loss": 0.2434111088514328, "global_step": 61064, "epoch": 686, "lr": 8.817002427452173e-05}
{"train_loss": 0.2355329692363739, "global_step": 61065, "epoch": 686, "lr": 8.816964980532428e-05}
{"train_loss": 0.24239704012870789, "global_step": 61066, "epoch": 686, "lr": 8.816927533099539e-05}
{"train_loss": 0.22170960903167725, "global_step": 61067, "epoch": 686, "lr": 8.816890085153513e-05}
{"train_loss": 0.17782694101333618, "global_step": 61068, "epoch": 686, "lr": 8.816852636694351e-05}
{"train_loss": 0.27601274847984314, "global_step": 61069, "epoch": 686, "lr": 8.816815187722061e-05}
{"train_loss": 0.24359266459941864, "global_step": 61070, "epoch": 686, "lr": 8.816777738236644e-05}
{"train_loss": 0.24252156913280487, "global_step": 61071, "epoch": 686, "lr": 8.81674028823811e-05}
{"train_loss": 0.21305501461029053, "global_step": 61072, "epoch": 686, "lr": 8.81670283772646e-05}
{"train_loss": 0.2760966420173645, "global_step": 61073, "epoch": 686, "lr": 8.816665386701703e-05}
{"train_loss": 0.2239219695329666, "global_step": 61074, "epoch": 686, "lr": 8.816627935163842e-05}
{"train_loss": 0.1415157914161682, "global_step": 61075, "epoch": 686, "lr": 8.816590483112879e-05}
{"train_loss": 0.287587970495224, "global_step": 61076, "epoch": 686, "lr": 8.816553030548824e-05}
{"train_loss": 0.2625757157802582, "global_step": 61077, "epoch": 686, "lr": 8.81651557747168e-05}
{"train_loss": 0.187136709690094, "global_step": 61078, "epoch": 686, "lr": 8.81647812388145e-05}
{"train_loss": 0.365403950214386, "global_step": 61079, "epoch": 686, "lr": 8.816440669778143e-05}
{"train_loss": 0.17112529277801514, "global_step": 61080, "epoch": 686, "lr": 8.816403215161762e-05}
{"train_loss": 0.19512157142162323, "global_step": 61081, "epoch": 686, "lr": 8.816365760032313e-05}
{"train_loss": 0.21606376767158508, "global_step": 61082, "epoch": 686, "lr": 8.816328304389799e-05}
{"train_loss": 0.16944648325443268, "global_step": 61083, "epoch": 686, "lr": 8.816290848234225e-05}
{"train_loss": 0.1783924400806427, "global_step": 61084, "epoch": 686, "lr": 8.816253391565599e-05}
{"train_loss": 0.27972152829170227, "global_step": 61085, "epoch": 686, "lr": 8.816215934383922e-05}
{"train_loss": 0.20207089185714722, "global_step": 61086, "epoch": 686, "lr": 8.816178476689203e-05}
{"train_loss": 0.200240358710289, "global_step": 61087, "epoch": 686, "lr": 8.816141018481446e-05}
{"train_loss": 0.249463751912117, "global_step": 61088, "epoch": 686, "lr": 8.816103559760654e-05}
{"train_loss": 0.2325809895992279, "global_step": 61089, "epoch": 686, "lr": 8.816066100526834e-05}
{"train_loss": 0.2823265790939331, "global_step": 61090, "epoch": 686, "lr": 8.816028640779991e-05}
{"train_loss": 0.22311708331108093, "global_step": 61091, "epoch": 686, "lr": 8.81599118052013e-05}
{"train_loss": 0.21835163235664368, "global_step": 61092, "epoch": 686, "lr": 8.815953719747252e-05}
{"train_loss": 0.2670636773109436, "global_step": 61093, "epoch": 686, "lr": 8.815916258461368e-05}
{"train_loss": 0.2568862736225128, "global_step": 61094, "epoch": 686, "lr": 8.81587879666248e-05}
{"train_loss": 0.19957783818244934, "global_step": 61095, "epoch": 686, "lr": 8.815841334350594e-05}
{"train_loss": 0.3240712881088257, "global_step": 61096, "epoch": 686, "lr": 8.815803871525715e-05}
{"train_loss": 0.1891774982213974, "global_step": 61097, "epoch": 686, "lr": 8.815766408187847e-05}
{"train_loss": 0.2483081817626953, "global_step": 61098, "epoch": 686, "lr": 8.815728944336996e-05}
{"train_loss": 0.2589932382106781, "global_step": 61099, "epoch": 686, "lr": 8.815691479973167e-05}
{"train_loss": 0.2653549909591675, "global_step": 61100, "epoch": 686, "lr": 8.815654015096364e-05}
{"train_loss": 0.25867870450019836, "global_step": 61101, "epoch": 686, "lr": 8.815616549706594e-05}
{"train_loss": 0.2681499123573303, "global_step": 61102, "epoch": 686, "lr": 8.81557908380386e-05}
{"train_loss": 0.2573620676994324, "global_step": 61103, "epoch": 686, "lr": 8.815541617388168e-05}
{"train_loss": 0.21189342439174652, "global_step": 61104, "epoch": 686, "lr": 8.815504150459524e-05}
{"train_loss": 0.22378036379814148, "global_step": 61105, "epoch": 686, "lr": 8.81546668301793e-05}
{"train_loss": 0.2054569125175476, "global_step": 61106, "epoch": 686, "lr": 8.815429215063395e-05}
{"train_loss": 0.18713054060935974, "global_step": 61107, "epoch": 686, "lr": 8.81539174659592e-05}
{"train_loss": 0.2980371415615082, "global_step": 61108, "epoch": 686, "lr": 8.815354277615514e-05}
{"train_loss": 0.34883368015289307, "global_step": 61109, "epoch": 686, "lr": 8.815316808122179e-05}
{"train_loss": 0.22980651259422302, "global_step": 61110, "epoch": 686, "lr": 8.815279338115921e-05}
{"train_loss": 0.24533569812774658, "global_step": 61111, "epoch": 686, "lr": 8.815241867596746e-05}
{"train_loss": 0.2843174636363983, "global_step": 61112, "epoch": 686, "lr": 8.815204396564659e-05}
{"train_loss": 0.1866084635257721, "global_step": 61113, "epoch": 686, "lr": 8.815166925019662e-05}
{"train_loss": 0.2680148184299469, "global_step": 61114, "epoch": 686, "lr": 8.815129452961763e-05}
{"train_loss": 0.282569020986557, "global_step": 61115, "epoch": 686, "lr": 8.815091980390968e-05}
{"train_loss": 0.24184013903141022, "global_step": 61116, "epoch": 686, "lr": 8.815054507307279e-05}
{"train_loss": 0.28354206681251526, "global_step": 61117, "epoch": 686, "lr": 8.815017033710702e-05}
{"train_loss": 0.29710304737091064, "global_step": 61118, "epoch": 686, "lr": 8.814979559601244e-05}
{"train_loss": 0.21261316537857056, "global_step": 61119, "epoch": 686, "lr": 8.814942084978907e-05}
{"train_loss": 0.1976982206106186, "global_step": 61120, "epoch": 686, "lr": 8.814904609843697e-05}
{"train_loss": 0.21318352222442627, "global_step": 61121, "epoch": 686, "lr": 8.814867134195621e-05}
{"train_loss": 0.268436998128891, "global_step": 61122, "epoch": 686, "lr": 8.814829658034683e-05}
{"train_loss": 0.27337443828582764, "global_step": 61123, "epoch": 686, "lr": 8.814792181360887e-05}
{"train_loss": 0.30557388067245483, "global_step": 61124, "epoch": 686, "lr": 8.814754704174237e-05}
{"train_loss": 0.27222740650177, "global_step": 61125, "epoch": 686, "lr": 8.814717226474742e-05}
{"train_loss": 0.19671544432640076, "global_step": 61126, "epoch": 686, "lr": 8.814679748262404e-05}
{"train_loss": 0.27558040618896484, "global_step": 61127, "epoch": 686, "lr": 8.81464226953723e-05}
{"train_loss": 0.1908673644065857, "global_step": 61128, "epoch": 686, "lr": 8.814604790299222e-05}
{"train_loss": 0.25612595677375793, "global_step": 61129, "epoch": 686, "lr": 8.814567310548387e-05}
{"train_loss": 0.22320637106895447, "global_step": 61130, "epoch": 686, "lr": 8.814529830284731e-05}
{"train_loss": 0.3741738200187683, "global_step": 61131, "epoch": 686, "lr": 8.814492349508257e-05}
{"train_loss": 0.323406845331192, "global_step": 61132, "epoch": 686, "lr": 8.814454868218972e-05}
{"train_loss": 0.24052301049232483, "global_step": 61133, "epoch": 686, "lr": 8.81441738641688e-05}
{"train_loss": 0.29979753494262695, "global_step": 61134, "epoch": 686, "lr": 8.814379904101986e-05}
{"train_loss": 0.19131748378276825, "global_step": 61135, "epoch": 686, "lr": 8.814342421274294e-05}
{"train_loss": 0.2422705739736557, "global_step": 61136, "epoch": 686, "lr": 8.81430493793381e-05}
{"train_loss": 0.17280711233615875, "global_step": 61137, "epoch": 686, "lr": 8.814267454080541e-05}
{"train_loss": 0.2622174322605133, "global_step": 61138, "epoch": 686, "lr": 8.814229969714489e-05}
{"train_loss": 0.25501787662506104, "global_step": 61139, "epoch": 686, "lr": 8.81419248483566e-05}
{"train_loss": 0.23977014422416687, "global_step": 61140, "epoch": 686, "lr": 8.81415499944406e-05}
{"train_loss": 0.35530734062194824, "global_step": 61141, "epoch": 686, "lr": 8.814117513539693e-05}
{"train_loss": 0.24370861639467517, "global_step": 61142, "epoch": 686, "lr": 8.814080027122565e-05, "val_loss": 3.4703614711761475}
{"train_loss": 0.2748319208621979, "global_step": 61143, "epoch": 687, "lr": 8.81404254019268e-05}
{"train_loss": 0.23891377449035645, "global_step": 61144, "epoch": 687, "lr": 8.814005052750044e-05}
{"train_loss": 0.2060978263616562, "global_step": 61145, "epoch": 687, "lr": 8.813967564794661e-05}
{"train_loss": 0.22238823771476746, "global_step": 61146, "epoch": 687, "lr": 8.813930076326536e-05}
{"train_loss": 0.20447832345962524, "global_step": 61147, "epoch": 687, "lr": 8.813892587345676e-05}
{"train_loss": 0.3062581419944763, "global_step": 61148, "epoch": 687, "lr": 8.813855097852084e-05}
{"train_loss": 0.35693883895874023, "global_step": 61149, "epoch": 687, "lr": 8.813817607845764e-05}
{"train_loss": 0.24471406638622284, "global_step": 61150, "epoch": 687, "lr": 8.813780117326724e-05}
{"train_loss": 0.37455180287361145, "global_step": 61151, "epoch": 687, "lr": 8.813742626294969e-05}
{"train_loss": 0.2028665542602539, "global_step": 61152, "epoch": 687, "lr": 8.813705134750501e-05}
{"train_loss": 0.20676732063293457, "global_step": 61153, "epoch": 687, "lr": 8.813667642693329e-05}
{"train_loss": 0.3364206850528717, "global_step": 61154, "epoch": 687, "lr": 8.813630150123453e-05}
{"train_loss": 0.23988677561283112, "global_step": 61155, "epoch": 687, "lr": 8.813592657040883e-05}
{"train_loss": 0.23569633066654205, "global_step": 61156, "epoch": 687, "lr": 8.813555163445621e-05}
{"train_loss": 0.3102186918258667, "global_step": 61157, "epoch": 687, "lr": 8.813517669337672e-05}
{"train_loss": 0.2201090008020401, "global_step": 61158, "epoch": 687, "lr": 8.813480174717044e-05}
{"train_loss": 0.2512781620025635, "global_step": 61159, "epoch": 687, "lr": 8.81344267958374e-05}
{"train_loss": 0.2506081163883209, "global_step": 61160, "epoch": 687, "lr": 8.813405183937765e-05}
{"train_loss": 0.2141861617565155, "global_step": 61161, "epoch": 687, "lr": 8.813367687779122e-05}
{"train_loss": 0.21586015820503235, "global_step": 61162, "epoch": 687, "lr": 8.81333019110782e-05}
{"train_loss": 0.26433104276657104, "global_step": 61163, "epoch": 687, "lr": 8.813292693923863e-05}
{"train_loss": 0.23572151362895966, "global_step": 61164, "epoch": 687, "lr": 8.813255196227252e-05}
{"train_loss": 0.22025510668754578, "global_step": 61165, "epoch": 687, "lr": 8.813217698018e-05}
{"train_loss": 0.23064078390598297, "global_step": 61166, "epoch": 687, "lr": 8.813180199296104e-05}
{"train_loss": 0.23018164932727814, "global_step": 61167, "epoch": 687, "lr": 8.813142700061573e-05}
{"train_loss": 0.2869397699832916, "global_step": 61168, "epoch": 687, "lr": 8.813105200314411e-05}
{"train_loss": 0.22973166406154633, "global_step": 61169, "epoch": 687, "lr": 8.813067700054625e-05}
{"train_loss": 0.2655943036079407, "global_step": 61170, "epoch": 687, "lr": 8.813030199282217e-05}
{"train_loss": 0.2304205447435379, "global_step": 61171, "epoch": 687, "lr": 8.812992697997196e-05}
{"train_loss": 0.25463178753852844, "global_step": 61172, "epoch": 687, "lr": 8.812955196199561e-05}
{"train_loss": 0.21690934896469116, "global_step": 61173, "epoch": 687, "lr": 8.812917693889325e-05}
{"train_loss": 0.2512243688106537, "global_step": 61174, "epoch": 687, "lr": 8.812880191066485e-05}
{"train_loss": 0.21540600061416626, "global_step": 61175, "epoch": 687, "lr": 8.812842687731051e-05}
{"train_loss": 0.2523970305919647, "global_step": 61176, "epoch": 687, "lr": 8.812805183883026e-05}
{"train_loss": 0.32992568612098694, "global_step": 61177, "epoch": 687, "lr": 8.812767679522418e-05}
{"train_loss": 0.19703543186187744, "global_step": 61178, "epoch": 687, "lr": 8.812730174649229e-05}
{"train_loss": 0.17203816771507263, "global_step": 61179, "epoch": 687, "lr": 8.812692669263463e-05}
{"train_loss": 0.21980321407318115, "global_step": 61180, "epoch": 687, "lr": 8.812655163365129e-05}
{"train_loss": 0.2152634710073471, "global_step": 61181, "epoch": 687, "lr": 8.81261765695423e-05}
{"train_loss": 0.21961292624473572, "global_step": 61182, "epoch": 687, "lr": 8.812580150030769e-05}
{"train_loss": 0.3350611627101898, "global_step": 61183, "epoch": 687, "lr": 8.812542642594755e-05}
{"train_loss": 0.22091273963451385, "global_step": 61184, "epoch": 687, "lr": 8.81250513464619e-05}
{"train_loss": 0.29874464869499207, "global_step": 61185, "epoch": 687, "lr": 8.812467626185081e-05}
{"train_loss": 0.2462652027606964, "global_step": 61186, "epoch": 687, "lr": 8.812430117211432e-05}
{"train_loss": 0.2925238311290741, "global_step": 61187, "epoch": 687, "lr": 8.812392607725248e-05}
{"train_loss": 0.230282261967659, "global_step": 61188, "epoch": 687, "lr": 8.812355097726534e-05}
{"train_loss": 0.2504204511642456, "global_step": 61189, "epoch": 687, "lr": 8.812317587215295e-05}
{"train_loss": 0.3096391558647156, "global_step": 61190, "epoch": 687, "lr": 8.812280076191537e-05}
{"train_loss": 0.3083723485469818, "global_step": 61191, "epoch": 687, "lr": 8.812242564655264e-05}
{"train_loss": 0.18334202468395233, "global_step": 61192, "epoch": 687, "lr": 8.812205052606484e-05}
{"train_loss": 0.23670224845409393, "global_step": 61193, "epoch": 687, "lr": 8.812167540045196e-05}
{"train_loss": 0.27990996837615967, "global_step": 61194, "epoch": 687, "lr": 8.81213002697141e-05}
{"train_loss": 0.2284868061542511, "global_step": 61195, "epoch": 687, "lr": 8.81209251338513e-05}
{"train_loss": 0.22999826073646545, "global_step": 61196, "epoch": 687, "lr": 8.812054999286359e-05}
{"train_loss": 0.18436741828918457, "global_step": 61197, "epoch": 687, "lr": 8.812017484675105e-05}
{"train_loss": 0.2578850984573364, "global_step": 61198, "epoch": 687, "lr": 8.811979969551372e-05}
{"train_loss": 0.1926085203886032, "global_step": 61199, "epoch": 687, "lr": 8.811942453915165e-05}
{"train_loss": 0.23296217620372772, "global_step": 61200, "epoch": 687, "lr": 8.811904937766488e-05}
{"train_loss": 0.20450560748577118, "global_step": 61201, "epoch": 687, "lr": 8.811867421105347e-05}
{"train_loss": 0.25347065925598145, "global_step": 61202, "epoch": 687, "lr": 8.811829903931746e-05}
{"train_loss": 0.32861196994781494, "global_step": 61203, "epoch": 687, "lr": 8.811792386245692e-05}
{"train_loss": 0.22312131524085999, "global_step": 61204, "epoch": 687, "lr": 8.811754868047189e-05}
{"train_loss": 0.1507997065782547, "global_step": 61205, "epoch": 687, "lr": 8.811717349336241e-05}
{"train_loss": 0.2324540764093399, "global_step": 61206, "epoch": 687, "lr": 8.811679830112855e-05}
{"train_loss": 0.2980406880378723, "global_step": 61207, "epoch": 687, "lr": 8.811642310377036e-05}
{"train_loss": 0.1928424835205078, "global_step": 61208, "epoch": 687, "lr": 8.811604790128787e-05}
{"train_loss": 0.30131062865257263, "global_step": 61209, "epoch": 687, "lr": 8.811567269368113e-05}
{"train_loss": 0.30197083950042725, "global_step": 61210, "epoch": 687, "lr": 8.811529748095023e-05}
{"train_loss": 0.20332218706607819, "global_step": 61211, "epoch": 687, "lr": 8.811492226309516e-05}
{"train_loss": 0.2178477793931961, "global_step": 61212, "epoch": 687, "lr": 8.811454704011602e-05}
{"train_loss": 0.19407129287719727, "global_step": 61213, "epoch": 687, "lr": 8.811417181201285e-05}
{"train_loss": 0.2331712543964386, "global_step": 61214, "epoch": 687, "lr": 8.81137965787857e-05}
{"train_loss": 0.19234363734722137, "global_step": 61215, "epoch": 687, "lr": 8.811342134043459e-05}
{"train_loss": 0.22019962966442108, "global_step": 61216, "epoch": 687, "lr": 8.811304609695961e-05}
{"train_loss": 0.22002287209033966, "global_step": 61217, "epoch": 687, "lr": 8.811267084836079e-05}
{"train_loss": 0.2676076292991638, "global_step": 61218, "epoch": 687, "lr": 8.811229559463819e-05}
{"train_loss": 0.15409252047538757, "global_step": 61219, "epoch": 687, "lr": 8.811192033579185e-05}
{"train_loss": 0.2806609570980072, "global_step": 61220, "epoch": 687, "lr": 8.811154507182185e-05}
{"train_loss": 0.23112007975578308, "global_step": 61221, "epoch": 687, "lr": 8.81111698027282e-05}
{"train_loss": 0.260591059923172, "global_step": 61222, "epoch": 687, "lr": 8.811079452851097e-05}
{"train_loss": 0.2809496521949768, "global_step": 61223, "epoch": 687, "lr": 8.811041924917019e-05}
{"train_loss": 0.2763407230377197, "global_step": 61224, "epoch": 687, "lr": 8.811004396470595e-05}
{"train_loss": 0.29079723358154297, "global_step": 61225, "epoch": 687, "lr": 8.810966867511828e-05}
{"train_loss": 0.3297097384929657, "global_step": 61226, "epoch": 687, "lr": 8.810929338040721e-05}
{"train_loss": 0.18747934699058533, "global_step": 61227, "epoch": 687, "lr": 8.810891808057282e-05}
{"train_loss": 0.2147464156150818, "global_step": 61228, "epoch": 687, "lr": 8.810854277561516e-05}
{"train_loss": 0.29168108105659485, "global_step": 61229, "epoch": 687, "lr": 8.810816746553427e-05}
{"train_loss": 0.3031556010246277, "global_step": 61230, "epoch": 687, "lr": 8.81077921503302e-05}
{"train_loss": 0.24697901107621997, "global_step": 61231, "epoch": 687, "lr": 8.810741683000298e-05, "val_loss": 3.5116934776306152}
{"train_loss": 0.27398765087127686, "global_step": 61232, "epoch": 688, "lr": 8.810704150455271e-05}
{"train_loss": 0.20526929199695587, "global_step": 61233, "epoch": 688, "lr": 8.81066661739794e-05}
{"train_loss": 0.19783130288124084, "global_step": 61234, "epoch": 688, "lr": 8.810629083828312e-05}
{"train_loss": 0.1680108904838562, "global_step": 61235, "epoch": 688, "lr": 8.810591549746391e-05}
{"train_loss": 0.26463815569877625, "global_step": 61236, "epoch": 688, "lr": 8.810554015152183e-05}
{"train_loss": 0.26747241616249084, "global_step": 61237, "epoch": 688, "lr": 8.810516480045692e-05}
{"train_loss": 0.19266614317893982, "global_step": 61238, "epoch": 688, "lr": 8.810478944426921e-05}
{"train_loss": 0.1884748339653015, "global_step": 61239, "epoch": 688, "lr": 8.810441408295881e-05}
{"train_loss": 0.21973444521427155, "global_step": 61240, "epoch": 688, "lr": 8.810403871652572e-05}
{"train_loss": 0.27847251296043396, "global_step": 61241, "epoch": 688, "lr": 8.810366334497002e-05}
{"train_loss": 0.2981278598308563, "global_step": 61242, "epoch": 688, "lr": 8.810328796829174e-05}
{"train_loss": 0.30156949162483215, "global_step": 61243, "epoch": 688, "lr": 8.810291258649093e-05}
{"train_loss": 0.2891748249530792, "global_step": 61244, "epoch": 688, "lr": 8.810253719956766e-05}
{"train_loss": 0.27758777141571045, "global_step": 61245, "epoch": 688, "lr": 8.810216180752195e-05}
{"train_loss": 0.30271288752555847, "global_step": 61246, "epoch": 688, "lr": 8.81017864103539e-05}
{"train_loss": 0.16551420092582703, "global_step": 61247, "epoch": 688, "lr": 8.81014110080635e-05}
{"train_loss": 0.2102959156036377, "global_step": 61248, "epoch": 688, "lr": 8.810103560065083e-05}
{"train_loss": 0.27050545811653137, "global_step": 61249, "epoch": 688, "lr": 8.810066018811595e-05}
{"train_loss": 0.17853513360023499, "global_step": 61250, "epoch": 688, "lr": 8.810028477045891e-05}
{"train_loss": 0.2306315302848816, "global_step": 61251, "epoch": 688, "lr": 8.809990934767973e-05}
{"train_loss": 0.28195273876190186, "global_step": 61252, "epoch": 688, "lr": 8.80995339197785e-05}
{"train_loss": 0.21883119642734528, "global_step": 61253, "epoch": 688, "lr": 8.809915848675525e-05}
{"train_loss": 0.18845482170581818, "global_step": 61254, "epoch": 688, "lr": 8.809878304861003e-05}
{"train_loss": 0.21469759941101074, "global_step": 61255, "epoch": 688, "lr": 8.809840760534288e-05}
{"train_loss": 0.2788912057876587, "global_step": 61256, "epoch": 688, "lr": 8.809803215695387e-05}
{"train_loss": 0.30504828691482544, "global_step": 61257, "epoch": 688, "lr": 8.809765670344306e-05}
{"train_loss": 0.2186892032623291, "global_step": 61258, "epoch": 688, "lr": 8.809728124481046e-05}
{"train_loss": 0.19846341013908386, "global_step": 61259, "epoch": 688, "lr": 8.809690578105616e-05}
{"train_loss": 0.22030848264694214, "global_step": 61260, "epoch": 688, "lr": 8.809653031218019e-05}
{"train_loss": 0.2475510984659195, "global_step": 61261, "epoch": 688, "lr": 8.80961548381826e-05}
{"train_loss": 0.26929035782814026, "global_step": 61262, "epoch": 688, "lr": 8.809577935906345e-05}
{"train_loss": 0.23626823723316193, "global_step": 61263, "epoch": 688, "lr": 8.80954038748228e-05}
{"train_loss": 0.28205636143684387, "global_step": 61264, "epoch": 688, "lr": 8.809502838546069e-05}
{"train_loss": 0.22392429411411285, "global_step": 61265, "epoch": 688, "lr": 8.809465289097715e-05}
{"train_loss": 0.31152456998825073, "global_step": 61266, "epoch": 688, "lr": 8.809427739137224e-05}
{"train_loss": 0.31048136949539185, "global_step": 61267, "epoch": 688, "lr": 8.809390188664604e-05}
{"train_loss": 0.2161649912595749, "global_step": 61268, "epoch": 688, "lr": 8.809352637679858e-05}
{"train_loss": 0.3049671947956085, "global_step": 61269, "epoch": 688, "lr": 8.80931508618299e-05}
{"train_loss": 0.3647831678390503, "global_step": 61270, "epoch": 688, "lr": 8.809277534174007e-05}
{"train_loss": 0.35748928785324097, "global_step": 61271, "epoch": 688, "lr": 8.809239981652911e-05}
{"train_loss": 0.21399469673633575, "global_step": 61272, "epoch": 688, "lr": 8.80920242861971e-05}
{"train_loss": 0.32436057925224304, "global_step": 61273, "epoch": 688, "lr": 8.809164875074409e-05}
{"train_loss": 0.23370341956615448, "global_step": 61274, "epoch": 688, "lr": 8.809127321017013e-05}
{"train_loss": 0.25627681612968445, "global_step": 61275, "epoch": 688, "lr": 8.809089766447524e-05}
{"train_loss": 0.25533196330070496, "global_step": 61276, "epoch": 688, "lr": 8.80905221136595e-05}
{"train_loss": 0.36235377192497253, "global_step": 61277, "epoch": 688, "lr": 8.809014655772297e-05}
{"train_loss": 0.27094778418540955, "global_step": 61278, "epoch": 688, "lr": 8.808977099666567e-05}
{"train_loss": 0.3039349615573883, "global_step": 61279, "epoch": 688, "lr": 8.808939543048766e-05}
{"train_loss": 0.2453264594078064, "global_step": 61280, "epoch": 688, "lr": 8.8089019859189e-05}
{"train_loss": 0.22965191304683685, "global_step": 61281, "epoch": 688, "lr": 8.808864428276973e-05}
{"train_loss": 0.38456761837005615, "global_step": 61282, "epoch": 688, "lr": 8.808826870122991e-05}
{"train_loss": 0.2499973177909851, "global_step": 61283, "epoch": 688, "lr": 8.80878931145696e-05}
{"train_loss": 0.28221285343170166, "global_step": 61284, "epoch": 688, "lr": 8.808751752278881e-05}
{"train_loss": 0.1865372359752655, "global_step": 61285, "epoch": 688, "lr": 8.808714192588764e-05}
{"train_loss": 0.3462277054786682, "global_step": 61286, "epoch": 688, "lr": 8.808676632386611e-05}
{"train_loss": 0.19818149507045746, "global_step": 61287, "epoch": 688, "lr": 8.808639071672427e-05}
{"train_loss": 0.2690016031265259, "global_step": 61288, "epoch": 688, "lr": 8.808601510446219e-05}
{"train_loss": 0.17962157726287842, "global_step": 61289, "epoch": 688, "lr": 8.808563948707992e-05}
{"train_loss": 0.2553982138633728, "global_step": 61290, "epoch": 688, "lr": 8.808526386457748e-05}
{"train_loss": 0.2703094184398651, "global_step": 61291, "epoch": 688, "lr": 8.808488823695495e-05}
{"train_loss": 0.2810284495353699, "global_step": 61292, "epoch": 688, "lr": 8.808451260421238e-05}
{"train_loss": 0.32259148359298706, "global_step": 61293, "epoch": 688, "lr": 8.808413696634978e-05}
{"train_loss": 0.35994985699653625, "global_step": 61294, "epoch": 688, "lr": 8.808376132336727e-05}
{"train_loss": 0.28978830575942993, "global_step": 61295, "epoch": 688, "lr": 8.808338567526484e-05}
{"train_loss": 0.289580762386322, "global_step": 61296, "epoch": 688, "lr": 8.808301002204255e-05}
{"train_loss": 0.24673067033290863, "global_step": 61297, "epoch": 688, "lr": 8.80826343637005e-05}
{"train_loss": 0.19207578897476196, "global_step": 61298, "epoch": 688, "lr": 8.808225870023867e-05}
{"train_loss": 0.22530196607112885, "global_step": 61299, "epoch": 688, "lr": 8.808188303165718e-05}
{"train_loss": 0.2023652344942093, "global_step": 61300, "epoch": 688, "lr": 8.808150735795602e-05}
{"train_loss": 0.31337231397628784, "global_step": 61301, "epoch": 688, "lr": 8.808113167913527e-05}
{"train_loss": 0.23743309080600739, "global_step": 61302, "epoch": 688, "lr": 8.808075599519498e-05}
{"train_loss": 0.28627440333366394, "global_step": 61303, "epoch": 688, "lr": 8.808038030613519e-05}
{"train_loss": 0.22848479449748993, "global_step": 61304, "epoch": 688, "lr": 8.808000461195596e-05}
{"train_loss": 0.2526835799217224, "global_step": 61305, "epoch": 688, "lr": 8.807962891265733e-05}
{"train_loss": 0.21596913039684296, "global_step": 61306, "epoch": 688, "lr": 8.807925320823936e-05}
{"train_loss": 0.2951987385749817, "global_step": 61307, "epoch": 688, "lr": 8.807887749870211e-05}
{"train_loss": 0.21628662943840027, "global_step": 61308, "epoch": 688, "lr": 8.807850178404562e-05}
{"train_loss": 0.2582366466522217, "global_step": 61309, "epoch": 688, "lr": 8.807812606426992e-05}
{"train_loss": 0.2893068492412567, "global_step": 61310, "epoch": 688, "lr": 8.807775033937511e-05}
{"train_loss": 0.21974876523017883, "global_step": 61311, "epoch": 688, "lr": 8.807737460936118e-05}
{"train_loss": 0.2988605201244354, "global_step": 61312, "epoch": 688, "lr": 8.807699887422822e-05}
{"train_loss": 0.19185495376586914, "global_step": 61313, "epoch": 688, "lr": 8.807662313397629e-05}
{"train_loss": 0.1500229388475418, "global_step": 61314, "epoch": 688, "lr": 8.80762473886054e-05}
{"train_loss": 0.22065988183021545, "global_step": 61315, "epoch": 688, "lr": 8.807587163811563e-05}
{"train_loss": 0.21126317977905273, "global_step": 61316, "epoch": 688, "lr": 8.807549588250703e-05}
{"train_loss": 0.19064714014530182, "global_step": 61317, "epoch": 688, "lr": 8.807512012177963e-05}
{"train_loss": 0.2853163182735443, "global_step": 61318, "epoch": 688, "lr": 8.80747443559335e-05}
{"train_loss": 0.27603599429130554, "global_step": 61319, "epoch": 688, "lr": 8.80743685849687e-05}
{"train_loss": 0.25508154911941355, "global_step": 61320, "epoch": 688, "lr": 8.807399280888524e-05, "val_loss": 3.4056756496429443}
{"train_loss": 0.147759810090065, "global_step": 61321, "epoch": 689, "lr": 8.807361702768321e-05}
{"train_loss": 0.30855792760849, "global_step": 61322, "epoch": 689, "lr": 8.807324124136264e-05}
{"train_loss": 0.24586106836795807, "global_step": 61323, "epoch": 689, "lr": 8.807286544992359e-05}
{"train_loss": 0.20931105315685272, "global_step": 61324, "epoch": 689, "lr": 8.80724896533661e-05}
{"train_loss": 0.19925753772258759, "global_step": 61325, "epoch": 689, "lr": 8.807211385169023e-05}
{"train_loss": 0.2646177411079407, "global_step": 61326, "epoch": 689, "lr": 8.807173804489605e-05}
{"train_loss": 0.24554653465747833, "global_step": 61327, "epoch": 689, "lr": 8.807136223298356e-05}
{"train_loss": 0.1894017457962036, "global_step": 61328, "epoch": 689, "lr": 8.807098641595285e-05}
{"train_loss": 0.20746636390686035, "global_step": 61329, "epoch": 689, "lr": 8.807061059380397e-05}
{"train_loss": 0.15533554553985596, "global_step": 61330, "epoch": 689, "lr": 8.807023476653696e-05}
{"train_loss": 0.23723946511745453, "global_step": 61331, "epoch": 689, "lr": 8.806985893415184e-05}
{"train_loss": 0.2426794022321701, "global_step": 61332, "epoch": 689, "lr": 8.806948309664873e-05}
{"train_loss": 0.21806056797504425, "global_step": 61333, "epoch": 689, "lr": 8.806910725402763e-05}
{"train_loss": 0.2664494514465332, "global_step": 61334, "epoch": 689, "lr": 8.80687314062886e-05}
{"train_loss": 0.2206554263830185, "global_step": 61335, "epoch": 689, "lr": 8.806835555343169e-05}
{"train_loss": 0.2357335090637207, "global_step": 61336, "epoch": 689, "lr": 8.806797969545694e-05}
{"train_loss": 0.17426568269729614, "global_step": 61337, "epoch": 689, "lr": 8.806760383236445e-05}
{"train_loss": 0.21240238845348358, "global_step": 61338, "epoch": 689, "lr": 8.806722796415421e-05}
{"train_loss": 0.22760826349258423, "global_step": 61339, "epoch": 689, "lr": 8.80668520908263e-05}
{"train_loss": 0.3196510672569275, "global_step": 61340, "epoch": 689, "lr": 8.806647621238076e-05}
{"train_loss": 0.2100667655467987, "global_step": 61341, "epoch": 689, "lr": 8.806610032881766e-05}
{"train_loss": 0.21064886450767517, "global_step": 61342, "epoch": 689, "lr": 8.806572444013703e-05}
{"train_loss": 0.39930102229118347, "global_step": 61343, "epoch": 689, "lr": 8.806534854633892e-05}
{"train_loss": 0.19772371649742126, "global_step": 61344, "epoch": 689, "lr": 8.80649726474234e-05}
{"train_loss": 0.20862694084644318, "global_step": 61345, "epoch": 689, "lr": 8.806459674339049e-05}
{"train_loss": 0.2610386312007904, "global_step": 61346, "epoch": 689, "lr": 8.806422083424027e-05}
{"train_loss": 0.2560926377773285, "global_step": 61347, "epoch": 689, "lr": 8.80638449199728e-05}
{"train_loss": 0.3099397122859955, "global_step": 61348, "epoch": 689, "lr": 8.806346900058808e-05}
{"train_loss": 0.23007185757160187, "global_step": 61349, "epoch": 689, "lr": 8.80630930760862e-05}
{"train_loss": 0.23418603837490082, "global_step": 61350, "epoch": 689, "lr": 8.80627171464672e-05}
{"train_loss": 0.2551121115684509, "global_step": 61351, "epoch": 689, "lr": 8.806234121173114e-05}
{"train_loss": 0.1837500035762787, "global_step": 61352, "epoch": 689, "lr": 8.806196527187805e-05}
{"train_loss": 0.26096928119659424, "global_step": 61353, "epoch": 689, "lr": 8.8061589326908e-05}
{"train_loss": 0.2671188414096832, "global_step": 61354, "epoch": 689, "lr": 8.806121337682102e-05}
{"train_loss": 0.2742561399936676, "global_step": 61355, "epoch": 689, "lr": 8.806083742161719e-05}
{"train_loss": 0.19701248407363892, "global_step": 61356, "epoch": 689, "lr": 8.806046146129654e-05}
{"train_loss": 0.16983561217784882, "global_step": 61357, "epoch": 689, "lr": 8.806008549585912e-05}
{"train_loss": 0.26494917273521423, "global_step": 61358, "epoch": 689, "lr": 8.8059709525305e-05}
{"train_loss": 0.1937875896692276, "global_step": 61359, "epoch": 689, "lr": 8.805933354963421e-05}
{"train_loss": 0.2556810677051544, "global_step": 61360, "epoch": 689, "lr": 8.80589575688468e-05}
{"train_loss": 0.22536441683769226, "global_step": 61361, "epoch": 689, "lr": 8.805858158294282e-05}
{"train_loss": 0.21197840571403503, "global_step": 61362, "epoch": 689, "lr": 8.805820559192234e-05}
{"train_loss": 0.22400900721549988, "global_step": 61363, "epoch": 689, "lr": 8.805782959578539e-05}
{"train_loss": 0.19037532806396484, "global_step": 61364, "epoch": 689, "lr": 8.805745359453204e-05}
{"train_loss": 0.19341053068637848, "global_step": 61365, "epoch": 689, "lr": 8.805707758816232e-05}
{"train_loss": 0.2779267728328705, "global_step": 61366, "epoch": 689, "lr": 8.80567015766763e-05}
{"train_loss": 0.19678914546966553, "global_step": 61367, "epoch": 689, "lr": 8.8056325560074e-05}
{"train_loss": 0.254092276096344, "global_step": 61368, "epoch": 689, "lr": 8.80559495383555e-05}
{"train_loss": 0.2145458310842514, "global_step": 61369, "epoch": 689, "lr": 8.805557351152085e-05}
{"train_loss": 0.20689180493354797, "global_step": 61370, "epoch": 689, "lr": 8.80551974795701e-05}
{"train_loss": 0.17016805708408356, "global_step": 61371, "epoch": 689, "lr": 8.805482144250327e-05}
{"train_loss": 0.2938072979450226, "global_step": 61372, "epoch": 689, "lr": 8.805444540032043e-05}
{"train_loss": 0.17696602642536163, "global_step": 61373, "epoch": 689, "lr": 8.805406935302165e-05}
{"train_loss": 0.25004705786705017, "global_step": 61374, "epoch": 689, "lr": 8.805369330060698e-05}
{"train_loss": 0.23819443583488464, "global_step": 61375, "epoch": 689, "lr": 8.805331724307643e-05}
{"train_loss": 0.26196178793907166, "global_step": 61376, "epoch": 689, "lr": 8.805294118043008e-05}
{"train_loss": 0.19318753480911255, "global_step": 61377, "epoch": 689, "lr": 8.805256511266797e-05}
{"train_loss": 0.26413607597351074, "global_step": 61378, "epoch": 689, "lr": 8.805218903979017e-05}
{"train_loss": 0.1439450979232788, "global_step": 61379, "epoch": 689, "lr": 8.805181296179669e-05}
{"train_loss": 0.24578236043453217, "global_step": 61380, "epoch": 689, "lr": 8.805143687868765e-05}
{"train_loss": 0.18305577337741852, "global_step": 61381, "epoch": 689, "lr": 8.805106079046302e-05}
{"train_loss": 0.17709709703922272, "global_step": 61382, "epoch": 689, "lr": 8.80506846971229e-05}
{"train_loss": 0.24788375198841095, "global_step": 61383, "epoch": 689, "lr": 8.805030859866734e-05}
{"train_loss": 0.22075214982032776, "global_step": 61384, "epoch": 689, "lr": 8.804993249509638e-05}
{"train_loss": 0.2070702165365219, "global_step": 61385, "epoch": 689, "lr": 8.804955638641006e-05}
{"train_loss": 0.29549914598464966, "global_step": 61386, "epoch": 689, "lr": 8.804918027260844e-05}
{"train_loss": 0.23805125057697296, "global_step": 61387, "epoch": 689, "lr": 8.804880415369159e-05}
{"train_loss": 0.17682451009750366, "global_step": 61388, "epoch": 689, "lr": 8.804842802965951e-05}
{"train_loss": 0.2702484726905823, "global_step": 61389, "epoch": 689, "lr": 8.804805190051231e-05}
{"train_loss": 0.25640761852264404, "global_step": 61390, "epoch": 689, "lr": 8.804767576625e-05}
{"train_loss": 0.2649795413017273, "global_step": 61391, "epoch": 689, "lr": 8.804729962687265e-05}
{"train_loss": 0.2792869508266449, "global_step": 61392, "epoch": 689, "lr": 8.80469234823803e-05}
{"train_loss": 0.24643202126026154, "global_step": 61393, "epoch": 689, "lr": 8.804654733277301e-05}
{"train_loss": 0.25669237971305847, "global_step": 61394, "epoch": 689, "lr": 8.804617117805082e-05}
{"train_loss": 0.23037675023078918, "global_step": 61395, "epoch": 689, "lr": 8.804579501821379e-05}
{"train_loss": 0.18924358487129211, "global_step": 61396, "epoch": 689, "lr": 8.804541885326196e-05}
{"train_loss": 0.17942240834236145, "global_step": 61397, "epoch": 689, "lr": 8.80450426831954e-05}
{"train_loss": 0.27089786529541016, "global_step": 61398, "epoch": 689, "lr": 8.804466650801413e-05}
{"train_loss": 0.20086120069026947, "global_step": 61399, "epoch": 689, "lr": 8.804429032771823e-05}
{"train_loss": 0.2785928547382355, "global_step": 61400, "epoch": 689, "lr": 8.804391414230773e-05}
{"train_loss": 0.21627910435199738, "global_step": 61401, "epoch": 689, "lr": 8.804353795178271e-05}
{"train_loss": 0.2204701155424118, "global_step": 61402, "epoch": 689, "lr": 8.804316175614317e-05}
{"train_loss": 0.11669211089611053, "global_step": 61403, "epoch": 689, "lr": 8.804278555538922e-05}
{"train_loss": 0.2127065360546112, "global_step": 61404, "epoch": 689, "lr": 8.804240934952085e-05}
{"train_loss": 0.2547130286693573, "global_step": 61405, "epoch": 689, "lr": 8.804203313853817e-05}
{"train_loss": 0.28119832277297974, "global_step": 61406, "epoch": 689, "lr": 8.80416569224412e-05}
{"train_loss": 0.26698410511016846, "global_step": 61407, "epoch": 689, "lr": 8.804128070122998e-05}
{"train_loss": 0.3059321939945221, "global_step": 61408, "epoch": 689, "lr": 8.804090447490459e-05}
{"train_loss": 0.2323103922136714, "global_step": 61409, "epoch": 689, "lr": 8.804052824346503e-05, "val_loss": 3.5944314002990723}
{"train_loss": 0.19102437794208527, "global_step": 61410, "epoch": 690, "lr": 8.804015200691142e-05}
{"train_loss": 0.161459282040596, "global_step": 61411, "epoch": 690, "lr": 8.803977576524375e-05}
{"train_loss": 0.16842292249202728, "global_step": 61412, "epoch": 690, "lr": 8.803939951846212e-05}
{"train_loss": 0.23017561435699463, "global_step": 61413, "epoch": 690, "lr": 8.803902326656655e-05}
{"train_loss": 0.1864103525876999, "global_step": 61414, "epoch": 690, "lr": 8.803864700955708e-05}
{"train_loss": 0.2511540651321411, "global_step": 61415, "epoch": 690, "lr": 8.803827074743378e-05}
{"train_loss": 0.24288833141326904, "global_step": 61416, "epoch": 690, "lr": 8.803789448019671e-05}
{"train_loss": 0.31580033898353577, "global_step": 61417, "epoch": 690, "lr": 8.803751820784592e-05}
{"train_loss": 0.3236245810985565, "global_step": 61418, "epoch": 690, "lr": 8.803714193038143e-05}
{"train_loss": 0.29471591114997864, "global_step": 61419, "epoch": 690, "lr": 8.80367656478033e-05}
{"train_loss": 0.270844429731369, "global_step": 61420, "epoch": 690, "lr": 8.80363893601116e-05}
{"train_loss": 0.21950551867485046, "global_step": 61421, "epoch": 690, "lr": 8.803601306730639e-05}
{"train_loss": 0.2560891807079315, "global_step": 61422, "epoch": 690, "lr": 8.803563676938769e-05}
{"train_loss": 0.2020605504512787, "global_step": 61423, "epoch": 690, "lr": 8.803526046635555e-05}
{"train_loss": 0.21336914598941803, "global_step": 61424, "epoch": 690, "lr": 8.803488415821006e-05}
{"train_loss": 0.20752429962158203, "global_step": 61425, "epoch": 690, "lr": 8.803450784495122e-05}
{"train_loss": 0.2758444845676422, "global_step": 61426, "epoch": 690, "lr": 8.803413152657912e-05}
{"train_loss": 0.28427475690841675, "global_step": 61427, "epoch": 690, "lr": 8.803375520309378e-05}
{"train_loss": 0.21111109852790833, "global_step": 61428, "epoch": 690, "lr": 8.803337887449527e-05}
{"train_loss": 0.3159949779510498, "global_step": 61429, "epoch": 690, "lr": 8.803300254078363e-05}
{"train_loss": 0.2376777082681656, "global_step": 61430, "epoch": 690, "lr": 8.803262620195891e-05}
{"train_loss": 0.2678624391555786, "global_step": 61431, "epoch": 690, "lr": 8.803224985802118e-05}
{"train_loss": 0.42148834466934204, "global_step": 61432, "epoch": 690, "lr": 8.803187350897049e-05}
{"train_loss": 0.15547582507133484, "global_step": 61433, "epoch": 690, "lr": 8.803149715480686e-05}
{"train_loss": 0.27796700596809387, "global_step": 61434, "epoch": 690, "lr": 8.803112079553037e-05}
{"train_loss": 0.19908057153224945, "global_step": 61435, "epoch": 690, "lr": 8.803074443114105e-05}
{"train_loss": 0.2429841011762619, "global_step": 61436, "epoch": 690, "lr": 8.803036806163896e-05}
{"train_loss": 0.23941850662231445, "global_step": 61437, "epoch": 690, "lr": 8.802999168702415e-05}
{"train_loss": 0.21529634296894073, "global_step": 61438, "epoch": 690, "lr": 8.802961530729667e-05}
{"train_loss": 0.17847004532814026, "global_step": 61439, "epoch": 690, "lr": 8.802923892245657e-05}
{"train_loss": 0.31389909982681274, "global_step": 61440, "epoch": 690, "lr": 8.802886253250392e-05}
{"train_loss": 0.3159477412700653, "global_step": 61441, "epoch": 690, "lr": 8.802848613743875e-05}
{"train_loss": 0.23105326294898987, "global_step": 61442, "epoch": 690, "lr": 8.80281097372611e-05}
{"train_loss": 0.23666001856327057, "global_step": 61443, "epoch": 690, "lr": 8.802773333197104e-05}
{"train_loss": 0.24080947041511536, "global_step": 61444, "epoch": 690, "lr": 8.802735692156861e-05}
{"train_loss": 0.2588062286376953, "global_step": 61445, "epoch": 690, "lr": 8.802698050605387e-05}
{"train_loss": 0.24554826319217682, "global_step": 61446, "epoch": 690, "lr": 8.802660408542686e-05}
{"train_loss": 0.2334459275007248, "global_step": 61447, "epoch": 690, "lr": 8.802622765968764e-05}
{"train_loss": 0.2724721133708954, "global_step": 61448, "epoch": 690, "lr": 8.802585122883626e-05}
{"train_loss": 0.2791992127895355, "global_step": 61449, "epoch": 690, "lr": 8.802547479287278e-05}
{"train_loss": 0.3101578950881958, "global_step": 61450, "epoch": 690, "lr": 8.802509835179722e-05}
{"train_loss": 0.16060923039913177, "global_step": 61451, "epoch": 690, "lr": 8.802472190560965e-05}
{"train_loss": 0.2363150268793106, "global_step": 61452, "epoch": 690, "lr": 8.802434545431013e-05}
{"train_loss": 0.3028622567653656, "global_step": 61453, "epoch": 690, "lr": 8.802396899789869e-05}
{"train_loss": 0.18251743912696838, "global_step": 61454, "epoch": 690, "lr": 8.80235925363754e-05}
{"train_loss": 0.21510949730873108, "global_step": 61455, "epoch": 690, "lr": 8.802321606974029e-05}
{"train_loss": 0.1970997154712677, "global_step": 61456, "epoch": 690, "lr": 8.802283959799342e-05}
{"train_loss": 0.27689236402511597, "global_step": 61457, "epoch": 690, "lr": 8.802246312113486e-05}
{"train_loss": 0.21299946308135986, "global_step": 61458, "epoch": 690, "lr": 8.802208663916463e-05}
{"train_loss": 0.23812995851039886, "global_step": 61459, "epoch": 690, "lr": 8.80217101520828e-05}
{"train_loss": 0.26214396953582764, "global_step": 61460, "epoch": 690, "lr": 8.802133365988942e-05}
{"train_loss": 0.3632744252681732, "global_step": 61461, "epoch": 690, "lr": 8.802095716258452e-05}
{"train_loss": 0.218353271484375, "global_step": 61462, "epoch": 690, "lr": 8.802058066016817e-05}
{"train_loss": 0.30014434456825256, "global_step": 61463, "epoch": 690, "lr": 8.802020415264043e-05}
{"train_loss": 0.1874220222234726, "global_step": 61464, "epoch": 690, "lr": 8.801982764000132e-05}
{"train_loss": 0.25900784134864807, "global_step": 61465, "epoch": 690, "lr": 8.801945112225092e-05}
{"train_loss": 0.27453190088272095, "global_step": 61466, "epoch": 690, "lr": 8.801907459938925e-05}
{"train_loss": 0.3740009069442749, "global_step": 61467, "epoch": 690, "lr": 8.80186980714164e-05}
{"train_loss": 0.22765322029590607, "global_step": 61468, "epoch": 690, "lr": 8.801832153833239e-05}
{"train_loss": 0.31911855936050415, "global_step": 61469, "epoch": 690, "lr": 8.801794500013727e-05}
{"train_loss": 0.2085646688938141, "global_step": 61470, "epoch": 690, "lr": 8.801756845683112e-05}
{"train_loss": 0.19767270982265472, "global_step": 61471, "epoch": 690, "lr": 8.801719190841396e-05}
{"train_loss": 0.20208947360515594, "global_step": 61472, "epoch": 690, "lr": 8.801681535488587e-05}
{"train_loss": 0.2459302544593811, "global_step": 61473, "epoch": 690, "lr": 8.801643879624686e-05}
{"train_loss": 0.3048427104949951, "global_step": 61474, "epoch": 690, "lr": 8.801606223249701e-05}
{"train_loss": 0.19537600874900818, "global_step": 61475, "epoch": 690, "lr": 8.801568566363636e-05}
{"train_loss": 0.2305915802717209, "global_step": 61476, "epoch": 690, "lr": 8.801530908966497e-05}
{"train_loss": 0.2496698647737503, "global_step": 61477, "epoch": 690, "lr": 8.801493251058289e-05}
{"train_loss": 0.3088080883026123, "global_step": 61478, "epoch": 690, "lr": 8.801455592639014e-05}
{"train_loss": 0.24098661541938782, "global_step": 61479, "epoch": 690, "lr": 8.801417933708682e-05}
{"train_loss": 0.23255318403244019, "global_step": 61480, "epoch": 690, "lr": 8.801380274267295e-05}
{"train_loss": 0.1994170993566513, "global_step": 61481, "epoch": 690, "lr": 8.80134261431486e-05}
{"train_loss": 0.2689008116722107, "global_step": 61482, "epoch": 690, "lr": 8.801304953851379e-05}
{"train_loss": 0.2117696851491928, "global_step": 61483, "epoch": 690, "lr": 8.80126729287686e-05}
{"train_loss": 0.23724274337291718, "global_step": 61484, "epoch": 690, "lr": 8.801229631391306e-05}
{"train_loss": 0.21773560345172882, "global_step": 61485, "epoch": 690, "lr": 8.801191969394724e-05}
{"train_loss": 0.22027255594730377, "global_step": 61486, "epoch": 690, "lr": 8.801154306887119e-05}
{"train_loss": 0.1834183782339096, "global_step": 61487, "epoch": 690, "lr": 8.801116643868492e-05}
{"train_loss": 0.20958201587200165, "global_step": 61488, "epoch": 690, "lr": 8.801078980338854e-05}
{"train_loss": 0.3019072115421295, "global_step": 61489, "epoch": 690, "lr": 8.801041316298206e-05}
{"train_loss": 0.27475374937057495, "global_step": 61490, "epoch": 690, "lr": 8.801003651746556e-05}
{"train_loss": 0.16939590871334076, "global_step": 61491, "epoch": 690, "lr": 8.800965986683904e-05}
{"train_loss": 0.22305859625339508, "global_step": 61492, "epoch": 690, "lr": 8.800928321110262e-05}
{"train_loss": 0.2390504628419876, "global_step": 61493, "epoch": 690, "lr": 8.800890655025629e-05}
{"train_loss": 0.2035723179578781, "global_step": 61494, "epoch": 690, "lr": 8.800852988430012e-05}
{"train_loss": 0.196912944316864, "global_step": 61495, "epoch": 690, "lr": 8.800815321323419e-05}
{"train_loss": 0.24855880439281464, "global_step": 61496, "epoch": 690, "lr": 8.800777653705852e-05}
{"train_loss": 0.16544106602668762, "global_step": 61497, "epoch": 690, "lr": 8.800739985577318e-05}
{"train_loss": 0.24314301737238853, "global_step": 61498, "epoch": 690, "lr": 8.800702316937819e-05, "val_loss": 3.3628334999084473, "train_action_mse_error": 24.289470672607422}
{"train_loss": 0.21534167230129242, "global_step": 61499, "epoch": 691, "lr": 8.800664647787362e-05}
{"train_loss": 0.19092042744159698, "global_step": 61500, "epoch": 691, "lr": 8.800626978125952e-05}
{"train_loss": 0.2858734130859375, "global_step": 61501, "epoch": 691, "lr": 8.800589307953595e-05}
{"train_loss": 0.19820302724838257, "global_step": 61502, "epoch": 691, "lr": 8.800551637270293e-05}
{"train_loss": 0.15905353426933289, "global_step": 61503, "epoch": 691, "lr": 8.800513966076056e-05}
{"train_loss": 0.18827509880065918, "global_step": 61504, "epoch": 691, "lr": 8.800476294370885e-05}
{"train_loss": 0.23379535973072052, "global_step": 61505, "epoch": 691, "lr": 8.800438622154785e-05}
{"train_loss": 0.2032768279314041, "global_step": 61506, "epoch": 691, "lr": 8.800400949427764e-05}
{"train_loss": 0.18296152353286743, "global_step": 61507, "epoch": 691, "lr": 8.800363276189825e-05}
{"train_loss": 0.2894708812236786, "global_step": 61508, "epoch": 691, "lr": 8.800325602440973e-05}
{"train_loss": 0.19993890821933746, "global_step": 61509, "epoch": 691, "lr": 8.800287928181214e-05}
{"train_loss": 0.20631320774555206, "global_step": 61510, "epoch": 691, "lr": 8.800250253410553e-05}
{"train_loss": 0.1796777993440628, "global_step": 61511, "epoch": 691, "lr": 8.800212578128994e-05}
{"train_loss": 0.29599279165267944, "global_step": 61512, "epoch": 691, "lr": 8.800174902336543e-05}
{"train_loss": 0.31210777163505554, "global_step": 61513, "epoch": 691, "lr": 8.800137226033204e-05}
{"train_loss": 0.17860306799411774, "global_step": 61514, "epoch": 691, "lr": 8.800099549218985e-05}
{"train_loss": 0.1662236750125885, "global_step": 61515, "epoch": 691, "lr": 8.800061871893887e-05}
{"train_loss": 0.2424980252981186, "global_step": 61516, "epoch": 691, "lr": 8.800024194057918e-05}
{"train_loss": 0.21370697021484375, "global_step": 61517, "epoch": 691, "lr": 8.799986515711081e-05}
{"train_loss": 0.1964501291513443, "global_step": 61518, "epoch": 691, "lr": 8.799948836853383e-05}
{"train_loss": 0.22896525263786316, "global_step": 61519, "epoch": 691, "lr": 8.799911157484829e-05}
{"train_loss": 0.2583185136318207, "global_step": 61520, "epoch": 691, "lr": 8.79987347760542e-05}
{"train_loss": 0.24094590544700623, "global_step": 61521, "epoch": 691, "lr": 8.799835797215168e-05}
{"train_loss": 0.2350868582725525, "global_step": 61522, "epoch": 691, "lr": 8.799798116314072e-05}
{"train_loss": 0.21333730220794678, "global_step": 61523, "epoch": 691, "lr": 8.79976043490214e-05}
{"train_loss": 0.19593989849090576, "global_step": 61524, "epoch": 691, "lr": 8.799722752979378e-05}
{"train_loss": 0.26586782932281494, "global_step": 61525, "epoch": 691, "lr": 8.799685070545788e-05}
{"train_loss": 0.19229713082313538, "global_step": 61526, "epoch": 691, "lr": 8.799647387601378e-05}
{"train_loss": 0.21648824214935303, "global_step": 61527, "epoch": 691, "lr": 8.79960970414615e-05}
{"train_loss": 0.1840796321630478, "global_step": 61528, "epoch": 691, "lr": 8.799572020180111e-05}
{"train_loss": 0.2202538549900055, "global_step": 61529, "epoch": 691, "lr": 8.799534335703267e-05}
{"train_loss": 0.20758193731307983, "global_step": 61530, "epoch": 691, "lr": 8.799496650715622e-05}
{"train_loss": 0.27871763706207275, "global_step": 61531, "epoch": 691, "lr": 8.79945896521718e-05}
{"train_loss": 0.2376253306865692, "global_step": 61532, "epoch": 691, "lr": 8.799421279207946e-05}
{"train_loss": 0.2923722565174103, "global_step": 61533, "epoch": 691, "lr": 8.799383592687928e-05}
{"train_loss": 0.13478341698646545, "global_step": 61534, "epoch": 691, "lr": 8.799345905657129e-05}
{"train_loss": 0.268380343914032, "global_step": 61535, "epoch": 691, "lr": 8.799308218115554e-05}
{"train_loss": 0.1731138378381729, "global_step": 61536, "epoch": 691, "lr": 8.799270530063209e-05}
{"train_loss": 0.23904985189437866, "global_step": 61537, "epoch": 691, "lr": 8.799232841500097e-05}
{"train_loss": 0.1678144782781601, "global_step": 61538, "epoch": 691, "lr": 8.799195152426225e-05}
{"train_loss": 0.32495492696762085, "global_step": 61539, "epoch": 691, "lr": 8.799157462841597e-05}
{"train_loss": 0.2379523515701294, "global_step": 61540, "epoch": 691, "lr": 8.79911977274622e-05}
{"train_loss": 0.2500063478946686, "global_step": 61541, "epoch": 691, "lr": 8.799082082140096e-05}
{"train_loss": 0.20759639143943787, "global_step": 61542, "epoch": 691, "lr": 8.799044391023233e-05}
{"train_loss": 0.1829993724822998, "global_step": 61543, "epoch": 691, "lr": 8.799006699395634e-05}
{"train_loss": 0.1815231591463089, "global_step": 61544, "epoch": 691, "lr": 8.798969007257306e-05}
{"train_loss": 0.17012187838554382, "global_step": 61545, "epoch": 691, "lr": 8.798931314608252e-05}
{"train_loss": 0.19525232911109924, "global_step": 61546, "epoch": 691, "lr": 8.798893621448479e-05}
{"train_loss": 0.1860569715499878, "global_step": 61547, "epoch": 691, "lr": 8.798855927777989e-05}
{"train_loss": 0.17461802065372467, "global_step": 61548, "epoch": 691, "lr": 8.798818233596791e-05}
{"train_loss": 0.22886590659618378, "global_step": 61549, "epoch": 691, "lr": 8.798780538904887e-05}
{"train_loss": 0.2604706883430481, "global_step": 61550, "epoch": 691, "lr": 8.798742843702283e-05}
{"train_loss": 0.24157345294952393, "global_step": 61551, "epoch": 691, "lr": 8.798705147988986e-05}
{"train_loss": 0.24349799752235413, "global_step": 61552, "epoch": 691, "lr": 8.798667451765e-05}
{"train_loss": 0.16246727108955383, "global_step": 61553, "epoch": 691, "lr": 8.798629755030327e-05}
{"train_loss": 0.20631143450737, "global_step": 61554, "epoch": 691, "lr": 8.798592057784974e-05}
{"train_loss": 0.4003641903400421, "global_step": 61555, "epoch": 691, "lr": 8.798554360028949e-05}
{"train_loss": 0.22370779514312744, "global_step": 61556, "epoch": 691, "lr": 8.798516661762252e-05}
{"train_loss": 0.3015596568584442, "global_step": 61557, "epoch": 691, "lr": 8.798478962984894e-05}
{"train_loss": 0.1170293390750885, "global_step": 61558, "epoch": 691, "lr": 8.798441263696875e-05}
{"train_loss": 0.14058105647563934, "global_step": 61559, "epoch": 691, "lr": 8.798403563898203e-05}
{"train_loss": 0.2273002713918686, "global_step": 61560, "epoch": 691, "lr": 8.79836586358888e-05}
{"train_loss": 0.2736590504646301, "global_step": 61561, "epoch": 691, "lr": 8.798328162768913e-05}
{"train_loss": 0.2179366946220398, "global_step": 61562, "epoch": 691, "lr": 8.79829046143831e-05}
{"train_loss": 0.2023836225271225, "global_step": 61563, "epoch": 691, "lr": 8.798252759597071e-05}
{"train_loss": 0.218455970287323, "global_step": 61564, "epoch": 691, "lr": 8.798215057245202e-05}
{"train_loss": 0.21068286895751953, "global_step": 61565, "epoch": 691, "lr": 8.798177354382713e-05}
{"train_loss": 0.31753233075141907, "global_step": 61566, "epoch": 691, "lr": 8.798139651009602e-05}
{"train_loss": 0.2511065602302551, "global_step": 61567, "epoch": 691, "lr": 8.798101947125879e-05}
{"train_loss": 0.17846372723579407, "global_step": 61568, "epoch": 691, "lr": 8.798064242731547e-05}
{"train_loss": 0.1923626959323883, "global_step": 61569, "epoch": 691, "lr": 8.798026537826611e-05}
{"train_loss": 0.26882630586624146, "global_step": 61570, "epoch": 691, "lr": 8.797988832411077e-05}
{"train_loss": 0.20834681391716003, "global_step": 61571, "epoch": 691, "lr": 8.797951126484951e-05}
{"train_loss": 0.2594359815120697, "global_step": 61572, "epoch": 691, "lr": 8.797913420048234e-05}
{"train_loss": 0.15452544391155243, "global_step": 61573, "epoch": 691, "lr": 8.797875713100935e-05}
{"train_loss": 0.24326622486114502, "global_step": 61574, "epoch": 691, "lr": 8.79783800564306e-05}
{"train_loss": 0.14732575416564941, "global_step": 61575, "epoch": 691, "lr": 8.79780029767461e-05}
{"train_loss": 0.26198482513427734, "global_step": 61576, "epoch": 691, "lr": 8.797762589195593e-05}
{"train_loss": 0.25458961725234985, "global_step": 61577, "epoch": 691, "lr": 8.797724880206013e-05}
{"train_loss": 0.22744283080101013, "global_step": 61578, "epoch": 691, "lr": 8.797687170705874e-05}
{"train_loss": 0.2991020083427429, "global_step": 61579, "epoch": 691, "lr": 8.797649460695183e-05}
{"train_loss": 0.2674125134944916, "global_step": 61580, "epoch": 691, "lr": 8.797611750173944e-05}
{"train_loss": 0.29048290848731995, "global_step": 61581, "epoch": 691, "lr": 8.797574039142162e-05}
{"train_loss": 0.20580452680587769, "global_step": 61582, "epoch": 691, "lr": 8.797536327599845e-05}
{"train_loss": 0.299701988697052, "global_step": 61583, "epoch": 691, "lr": 8.797498615546993e-05}
{"train_loss": 0.21453377604484558, "global_step": 61584, "epoch": 691, "lr": 8.797460902983615e-05}
{"train_loss": 0.2611022889614105, "global_step": 61585, "epoch": 691, "lr": 8.797423189909713e-05}
{"train_loss": 0.18637718260288239, "global_step": 61586, "epoch": 691, "lr": 8.797385476325294e-05}
{"train_loss": 0.22467761093311095, "global_step": 61587, "epoch": 691, "lr": 8.797347762230364e-05, "val_loss": 3.521260976791382}
{"train_loss": 0.3091636300086975, "global_step": 61588, "epoch": 692, "lr": 8.797310047624926e-05}
{"train_loss": 0.21077492833137512, "global_step": 61589, "epoch": 692, "lr": 8.797272332508985e-05}
{"train_loss": 0.23532642424106598, "global_step": 61590, "epoch": 692, "lr": 8.797234616882549e-05}
{"train_loss": 0.19783808290958405, "global_step": 61591, "epoch": 692, "lr": 8.797196900745621e-05}
{"train_loss": 0.27111613750457764, "global_step": 61592, "epoch": 692, "lr": 8.797159184098205e-05}
{"train_loss": 0.20633873343467712, "global_step": 61593, "epoch": 692, "lr": 8.797121466940308e-05}
{"train_loss": 0.27397167682647705, "global_step": 61594, "epoch": 692, "lr": 8.797083749271934e-05}
{"train_loss": 0.2638731300830841, "global_step": 61595, "epoch": 692, "lr": 8.797046031093088e-05}
{"train_loss": 0.22569145262241364, "global_step": 61596, "epoch": 692, "lr": 8.797008312403776e-05}
{"train_loss": 0.2890159487724304, "global_step": 61597, "epoch": 692, "lr": 8.796970593204002e-05}
{"train_loss": 0.1971612572669983, "global_step": 61598, "epoch": 692, "lr": 8.796932873493773e-05}
{"train_loss": 0.2450280487537384, "global_step": 61599, "epoch": 692, "lr": 8.79689515327309e-05}
{"train_loss": 0.24408796429634094, "global_step": 61600, "epoch": 692, "lr": 8.796857432541963e-05}
{"train_loss": 0.19776985049247742, "global_step": 61601, "epoch": 692, "lr": 8.796819711300395e-05}
{"train_loss": 0.21872112154960632, "global_step": 61602, "epoch": 692, "lr": 8.796781989548388e-05}
{"train_loss": 0.315348744392395, "global_step": 61603, "epoch": 692, "lr": 8.796744267285953e-05}
{"train_loss": 0.23073163628578186, "global_step": 61604, "epoch": 692, "lr": 8.796706544513092e-05}
{"train_loss": 0.22498607635498047, "global_step": 61605, "epoch": 692, "lr": 8.796668821229808e-05}
{"train_loss": 0.22349008917808533, "global_step": 61606, "epoch": 692, "lr": 8.79663109743611e-05}
{"train_loss": 0.21284262835979462, "global_step": 61607, "epoch": 692, "lr": 8.796593373132e-05}
{"train_loss": 0.20583191514015198, "global_step": 61608, "epoch": 692, "lr": 8.796555648317484e-05}
{"train_loss": 0.22764772176742554, "global_step": 61609, "epoch": 692, "lr": 8.796517922992569e-05}
{"train_loss": 0.30224186182022095, "global_step": 61610, "epoch": 692, "lr": 8.796480197157257e-05}
{"train_loss": 0.22067296504974365, "global_step": 61611, "epoch": 692, "lr": 8.796442470811555e-05}
{"train_loss": 0.2887280285358429, "global_step": 61612, "epoch": 692, "lr": 8.796404743955467e-05}
{"train_loss": 0.2787680923938751, "global_step": 61613, "epoch": 692, "lr": 8.796367016589e-05}
{"train_loss": 0.3709198832511902, "global_step": 61614, "epoch": 692, "lr": 8.796329288712158e-05}
{"train_loss": 0.17028184235095978, "global_step": 61615, "epoch": 692, "lr": 8.796291560324942e-05}
{"train_loss": 0.2945794463157654, "global_step": 61616, "epoch": 692, "lr": 8.796253831427366e-05}
{"train_loss": 0.2807062268257141, "global_step": 61617, "epoch": 692, "lr": 8.796216102019427e-05}
{"train_loss": 0.2223612368106842, "global_step": 61618, "epoch": 692, "lr": 8.796178372101134e-05}
{"train_loss": 0.19585686922073364, "global_step": 61619, "epoch": 692, "lr": 8.79614064167249e-05}
{"train_loss": 0.2588975131511688, "global_step": 61620, "epoch": 692, "lr": 8.796102910733501e-05}
{"train_loss": 0.2585071921348572, "global_step": 61621, "epoch": 692, "lr": 8.796065179284174e-05}
{"train_loss": 0.23605084419250488, "global_step": 61622, "epoch": 692, "lr": 8.796027447324511e-05}
{"train_loss": 0.27333593368530273, "global_step": 61623, "epoch": 692, "lr": 8.795989714854518e-05}
{"train_loss": 0.22068212926387787, "global_step": 61624, "epoch": 692, "lr": 8.795951981874202e-05}
{"train_loss": 0.13556750118732452, "global_step": 61625, "epoch": 692, "lr": 8.795914248383566e-05}
{"train_loss": 0.287687748670578, "global_step": 61626, "epoch": 692, "lr": 8.795876514382615e-05}
{"train_loss": 0.2597665786743164, "global_step": 61627, "epoch": 692, "lr": 8.795838779871355e-05}
{"train_loss": 0.23966579139232635, "global_step": 61628, "epoch": 692, "lr": 8.79580104484979e-05}
{"train_loss": 0.2742048501968384, "global_step": 61629, "epoch": 692, "lr": 8.795763309317927e-05}
{"train_loss": 0.2335643470287323, "global_step": 61630, "epoch": 692, "lr": 8.795725573275768e-05}
{"train_loss": 0.3290677070617676, "global_step": 61631, "epoch": 692, "lr": 8.795687836723323e-05}
{"train_loss": 0.23985593020915985, "global_step": 61632, "epoch": 692, "lr": 8.795650099660593e-05}
{"train_loss": 0.23313885927200317, "global_step": 61633, "epoch": 692, "lr": 8.795612362087582e-05}
{"train_loss": 0.27369222044944763, "global_step": 61634, "epoch": 692, "lr": 8.795574624004299e-05}
{"train_loss": 0.22869162261486053, "global_step": 61635, "epoch": 692, "lr": 8.795536885410746e-05}
{"train_loss": 0.21808560192584991, "global_step": 61636, "epoch": 692, "lr": 8.795499146306931e-05}
{"train_loss": 0.2684210538864136, "global_step": 61637, "epoch": 692, "lr": 8.795461406692858e-05}
{"train_loss": 0.25486716628074646, "global_step": 61638, "epoch": 692, "lr": 8.795423666568529e-05}
{"train_loss": 0.23660379648208618, "global_step": 61639, "epoch": 692, "lr": 8.795385925933952e-05}
{"train_loss": 0.3098578453063965, "global_step": 61640, "epoch": 692, "lr": 8.795348184789133e-05}
{"train_loss": 0.2005084604024887, "global_step": 61641, "epoch": 692, "lr": 8.795310443134075e-05}
{"train_loss": 0.18020197749137878, "global_step": 61642, "epoch": 692, "lr": 8.795272700968784e-05}
{"train_loss": 0.26772433519363403, "global_step": 61643, "epoch": 692, "lr": 8.795234958293264e-05}
{"train_loss": 0.21465913951396942, "global_step": 61644, "epoch": 692, "lr": 8.795197215107523e-05}
{"train_loss": 0.22478710114955902, "global_step": 61645, "epoch": 692, "lr": 8.795159471411561e-05}
{"train_loss": 0.2307933121919632, "global_step": 61646, "epoch": 692, "lr": 8.795121727205389e-05}
{"train_loss": 0.27199897170066833, "global_step": 61647, "epoch": 692, "lr": 8.795083982489007e-05}
{"train_loss": 0.30923599004745483, "global_step": 61648, "epoch": 692, "lr": 8.795046237262423e-05}
{"train_loss": 0.23449496924877167, "global_step": 61649, "epoch": 692, "lr": 8.795008491525641e-05}
{"train_loss": 0.25581806898117065, "global_step": 61650, "epoch": 692, "lr": 8.794970745278667e-05}
{"train_loss": 0.2038930356502533, "global_step": 61651, "epoch": 692, "lr": 8.794932998521505e-05}
{"train_loss": 0.2024691253900528, "global_step": 61652, "epoch": 692, "lr": 8.794895251254161e-05}
{"train_loss": 0.20974160730838776, "global_step": 61653, "epoch": 692, "lr": 8.794857503476639e-05}
{"train_loss": 0.2566392123699188, "global_step": 61654, "epoch": 692, "lr": 8.794819755188944e-05}
{"train_loss": 0.27557092905044556, "global_step": 61655, "epoch": 692, "lr": 8.794782006391082e-05}
{"train_loss": 0.25137218832969666, "global_step": 61656, "epoch": 692, "lr": 8.794744257083059e-05}
{"train_loss": 0.20605076849460602, "global_step": 61657, "epoch": 692, "lr": 8.794706507264878e-05}
{"train_loss": 0.22664079070091248, "global_step": 61658, "epoch": 692, "lr": 8.794668756936546e-05}
{"train_loss": 0.29449501633644104, "global_step": 61659, "epoch": 692, "lr": 8.794631006098065e-05}
{"train_loss": 0.2685549259185791, "global_step": 61660, "epoch": 692, "lr": 8.794593254749444e-05}
{"train_loss": 0.2705193758010864, "global_step": 61661, "epoch": 692, "lr": 8.794555502890685e-05}
{"train_loss": 0.2435302883386612, "global_step": 61662, "epoch": 692, "lr": 8.794517750521796e-05}
{"train_loss": 0.34935060143470764, "global_step": 61663, "epoch": 692, "lr": 8.794479997642779e-05}
{"train_loss": 0.16326649487018585, "global_step": 61664, "epoch": 692, "lr": 8.79444224425364e-05}
{"train_loss": 0.20758196711540222, "global_step": 61665, "epoch": 692, "lr": 8.794404490354384e-05}
{"train_loss": 0.2752673327922821, "global_step": 61666, "epoch": 692, "lr": 8.79436673594502e-05}
{"train_loss": 0.24698051810264587, "global_step": 61667, "epoch": 692, "lr": 8.794328981025546e-05}
{"train_loss": 0.25904831290245056, "global_step": 61668, "epoch": 692, "lr": 8.794291225595972e-05}
{"train_loss": 0.19464807212352753, "global_step": 61669, "epoch": 692, "lr": 8.794253469656302e-05}
{"train_loss": 0.2547728419303894, "global_step": 61670, "epoch": 692, "lr": 8.79421571320654e-05}
{"train_loss": 0.32998722791671753, "global_step": 61671, "epoch": 692, "lr": 8.794177956246692e-05}
{"train_loss": 0.25212764739990234, "global_step": 61672, "epoch": 692, "lr": 8.794140198776763e-05}
{"train_loss": 0.19676847755908966, "global_step": 61673, "epoch": 692, "lr": 8.79410244079676e-05}
{"train_loss": 0.21825136244297028, "global_step": 61674, "epoch": 692, "lr": 8.794064682306684e-05}
{"train_loss": 0.19582341611385345, "global_step": 61675, "epoch": 692, "lr": 8.794026923306544e-05}
{"train_loss": 0.2442290699214078, "global_step": 61676, "epoch": 692, "lr": 8.793989163796341e-05, "val_loss": 3.526529550552368}
{"train_loss": 0.2400796264410019, "global_step": 61677, "epoch": 693, "lr": 8.793951403776084e-05}
{"train_loss": 0.23185017704963684, "global_step": 61678, "epoch": 693, "lr": 8.793913643245777e-05}
{"train_loss": 0.2097085416316986, "global_step": 61679, "epoch": 693, "lr": 8.793875882205425e-05}
{"train_loss": 0.17308644950389862, "global_step": 61680, "epoch": 693, "lr": 8.793838120655032e-05}
{"train_loss": 0.22888541221618652, "global_step": 61681, "epoch": 693, "lr": 8.793800358594602e-05}
{"train_loss": 0.21250346302986145, "global_step": 61682, "epoch": 693, "lr": 8.793762596024143e-05}
{"train_loss": 0.19551067054271698, "global_step": 61683, "epoch": 693, "lr": 8.793724832943659e-05}
{"train_loss": 0.2803010046482086, "global_step": 61684, "epoch": 693, "lr": 8.793687069353154e-05}
{"train_loss": 0.14740605652332306, "global_step": 61685, "epoch": 693, "lr": 8.793649305252634e-05}
{"train_loss": 0.22199638187885284, "global_step": 61686, "epoch": 693, "lr": 8.793611540642107e-05}
{"train_loss": 0.16121047735214233, "global_step": 61687, "epoch": 693, "lr": 8.793573775521571e-05}
{"train_loss": 0.2326820194721222, "global_step": 61688, "epoch": 693, "lr": 8.793536009891037e-05}
{"train_loss": 0.15861333906650543, "global_step": 61689, "epoch": 693, "lr": 8.793498243750509e-05}
{"train_loss": 0.18684926629066467, "global_step": 61690, "epoch": 693, "lr": 8.793460477099992e-05}
{"train_loss": 0.325594037771225, "global_step": 61691, "epoch": 693, "lr": 8.79342270993949e-05}
{"train_loss": 0.27847281098365784, "global_step": 61692, "epoch": 693, "lr": 8.793384942269008e-05}
{"train_loss": 0.14299574494361877, "global_step": 61693, "epoch": 693, "lr": 8.79334717408855e-05}
{"train_loss": 0.23212189972400665, "global_step": 61694, "epoch": 693, "lr": 8.793309405398125e-05}
{"train_loss": 0.2600458860397339, "global_step": 61695, "epoch": 693, "lr": 8.793271636197736e-05}
{"train_loss": 0.24682709574699402, "global_step": 61696, "epoch": 693, "lr": 8.793233866487385e-05}
{"train_loss": 0.25018128752708435, "global_step": 61697, "epoch": 693, "lr": 8.793196096267083e-05}
{"train_loss": 0.258852481842041, "global_step": 61698, "epoch": 693, "lr": 8.79315832553683e-05}
{"train_loss": 0.18545563519001007, "global_step": 61699, "epoch": 693, "lr": 8.793120554296635e-05}
{"train_loss": 0.29361286759376526, "global_step": 61700, "epoch": 693, "lr": 8.7930827825465e-05}
{"train_loss": 0.1720535308122635, "global_step": 61701, "epoch": 693, "lr": 8.793045010286432e-05}
{"train_loss": 0.19228799641132355, "global_step": 61702, "epoch": 693, "lr": 8.793007237516433e-05}
{"train_loss": 0.2906843423843384, "global_step": 61703, "epoch": 693, "lr": 8.792969464236513e-05}
{"train_loss": 0.20831164717674255, "global_step": 61704, "epoch": 693, "lr": 8.792931690446673e-05}
{"train_loss": 0.19538460671901703, "global_step": 61705, "epoch": 693, "lr": 8.79289391614692e-05}
{"train_loss": 0.21480567753314972, "global_step": 61706, "epoch": 693, "lr": 8.792856141337261e-05}
{"train_loss": 0.2751156687736511, "global_step": 61707, "epoch": 693, "lr": 8.792818366017695e-05}
{"train_loss": 0.1392057240009308, "global_step": 61708, "epoch": 693, "lr": 8.792780590188232e-05}
{"train_loss": 0.323713481426239, "global_step": 61709, "epoch": 693, "lr": 8.792742813848877e-05}
{"train_loss": 0.2214522808790207, "global_step": 61710, "epoch": 693, "lr": 8.792705036999633e-05}
{"train_loss": 0.20397095382213593, "global_step": 61711, "epoch": 693, "lr": 8.792667259640506e-05}
{"train_loss": 0.20146618783473969, "global_step": 61712, "epoch": 693, "lr": 8.792629481771501e-05}
{"train_loss": 0.1949351578950882, "global_step": 61713, "epoch": 693, "lr": 8.792591703392624e-05}
{"train_loss": 0.20618836581707, "global_step": 61714, "epoch": 693, "lr": 8.79255392450388e-05}
{"train_loss": 0.19204933941364288, "global_step": 61715, "epoch": 693, "lr": 8.792516145105272e-05}
{"train_loss": 0.20359548926353455, "global_step": 61716, "epoch": 693, "lr": 8.792478365196806e-05}
{"train_loss": 0.2917172312736511, "global_step": 61717, "epoch": 693, "lr": 8.792440584778487e-05}
{"train_loss": 0.19346003234386444, "global_step": 61718, "epoch": 693, "lr": 8.792402803850322e-05}
{"train_loss": 0.24386006593704224, "global_step": 61719, "epoch": 693, "lr": 8.792365022412314e-05}
{"train_loss": 0.23574765026569366, "global_step": 61720, "epoch": 693, "lr": 8.79232724046447e-05}
{"train_loss": 0.25442391633987427, "global_step": 61721, "epoch": 693, "lr": 8.792289458006793e-05}
{"train_loss": 0.1960344761610031, "global_step": 61722, "epoch": 693, "lr": 8.792251675039289e-05}
{"train_loss": 0.26072338223457336, "global_step": 61723, "epoch": 693, "lr": 8.792213891561962e-05}
{"train_loss": 0.17722187936306, "global_step": 61724, "epoch": 693, "lr": 8.79217610757482e-05}
{"train_loss": 0.2443988174200058, "global_step": 61725, "epoch": 693, "lr": 8.792138323077865e-05}
{"train_loss": 0.16850002110004425, "global_step": 61726, "epoch": 693, "lr": 8.792100538071103e-05}
{"train_loss": 0.2599075138568878, "global_step": 61727, "epoch": 693, "lr": 8.79206275255454e-05}
{"train_loss": 0.23729883134365082, "global_step": 61728, "epoch": 693, "lr": 8.79202496652818e-05}
{"train_loss": 0.2488911747932434, "global_step": 61729, "epoch": 693, "lr": 8.791987179992027e-05}
{"train_loss": 0.1716819852590561, "global_step": 61730, "epoch": 693, "lr": 8.79194939294609e-05}
{"train_loss": 0.18901865184307098, "global_step": 61731, "epoch": 693, "lr": 8.79191160539037e-05}
{"train_loss": 0.22164960205554962, "global_step": 61732, "epoch": 693, "lr": 8.791873817324874e-05}
{"train_loss": 0.24386943876743317, "global_step": 61733, "epoch": 693, "lr": 8.791836028749607e-05}
{"train_loss": 0.29729244112968445, "global_step": 61734, "epoch": 693, "lr": 8.791798239664575e-05}
{"train_loss": 0.2703917622566223, "global_step": 61735, "epoch": 693, "lr": 8.791760450069781e-05}
{"train_loss": 0.18900127708911896, "global_step": 61736, "epoch": 693, "lr": 8.79172265996523e-05}
{"train_loss": 0.16542935371398926, "global_step": 61737, "epoch": 693, "lr": 8.791684869350929e-05}
{"train_loss": 0.2535422444343567, "global_step": 61738, "epoch": 693, "lr": 8.791647078226882e-05}
{"train_loss": 0.1427508443593979, "global_step": 61739, "epoch": 693, "lr": 8.791609286593095e-05}
{"train_loss": 0.20104563236236572, "global_step": 61740, "epoch": 693, "lr": 8.791571494449572e-05}
{"train_loss": 0.13800953328609467, "global_step": 61741, "epoch": 693, "lr": 8.791533701796318e-05}
{"train_loss": 0.2639787197113037, "global_step": 61742, "epoch": 693, "lr": 8.791495908633338e-05}
{"train_loss": 0.19587677717208862, "global_step": 61743, "epoch": 693, "lr": 8.79145811496064e-05}
{"train_loss": 0.20421290397644043, "global_step": 61744, "epoch": 693, "lr": 8.791420320778222e-05}
{"train_loss": 0.2758251130580902, "global_step": 61745, "epoch": 693, "lr": 8.791382526086099e-05}
{"train_loss": 0.20036107301712036, "global_step": 61746, "epoch": 693, "lr": 8.791344730884267e-05}
{"train_loss": 0.25098374485969543, "global_step": 61747, "epoch": 693, "lr": 8.791306935172737e-05}
{"train_loss": 0.15772871673107147, "global_step": 61748, "epoch": 693, "lr": 8.791269138951512e-05}
{"train_loss": 0.2933503985404968, "global_step": 61749, "epoch": 693, "lr": 8.791231342220596e-05}
{"train_loss": 0.22040730714797974, "global_step": 61750, "epoch": 693, "lr": 8.791193544979997e-05}
{"train_loss": 0.1715821921825409, "global_step": 61751, "epoch": 693, "lr": 8.791155747229716e-05}
{"train_loss": 0.15682761371135712, "global_step": 61752, "epoch": 693, "lr": 8.79111794896976e-05}
{"train_loss": 0.22900502383708954, "global_step": 61753, "epoch": 693, "lr": 8.791080150200137e-05}
{"train_loss": 0.2711397707462311, "global_step": 61754, "epoch": 693, "lr": 8.791042350920849e-05}
{"train_loss": 0.2733208239078522, "global_step": 61755, "epoch": 693, "lr": 8.7910045511319e-05}
{"train_loss": 0.31597524881362915, "global_step": 61756, "epoch": 693, "lr": 8.790966750833297e-05}
{"train_loss": 0.19779491424560547, "global_step": 61757, "epoch": 693, "lr": 8.790928950025045e-05}
{"train_loss": 0.2453598827123642, "global_step": 61758, "epoch": 693, "lr": 8.790891148707149e-05}
{"train_loss": 0.21635903418064117, "global_step": 61759, "epoch": 693, "lr": 8.790853346879613e-05}
{"train_loss": 0.2208184152841568, "global_step": 61760, "epoch": 693, "lr": 8.790815544542444e-05}
{"train_loss": 0.19218449294567108, "global_step": 61761, "epoch": 693, "lr": 8.790777741695646e-05}
{"train_loss": 0.2951226830482483, "global_step": 61762, "epoch": 693, "lr": 8.790739938339223e-05}
{"train_loss": 0.16371600329875946, "global_step": 61763, "epoch": 693, "lr": 8.790702134473183e-05}
{"train_loss": 0.3051835596561432, "global_step": 61764, "epoch": 693, "lr": 8.790664330097528e-05}
{"train_loss": 0.2223184004928289, "global_step": 61765, "epoch": 693, "lr": 8.790626525212264e-05, "val_loss": 3.5588700771331787}
{"train_loss": 0.25699731707572937, "global_step": 61766, "epoch": 694, "lr": 8.790588719817398e-05}
{"train_loss": 0.23003743588924408, "global_step": 61767, "epoch": 694, "lr": 8.790550913912934e-05}
{"train_loss": 0.21677733957767487, "global_step": 61768, "epoch": 694, "lr": 8.790513107498875e-05}
{"train_loss": 0.27211111783981323, "global_step": 61769, "epoch": 694, "lr": 8.790475300575227e-05}
{"train_loss": 0.16988857090473175, "global_step": 61770, "epoch": 694, "lr": 8.790437493141998e-05}
{"train_loss": 0.22244949638843536, "global_step": 61771, "epoch": 694, "lr": 8.790399685199189e-05}
{"train_loss": 0.28554636240005493, "global_step": 61772, "epoch": 694, "lr": 8.790361876746809e-05}
{"train_loss": 0.17328979074954987, "global_step": 61773, "epoch": 694, "lr": 8.790324067784858e-05}
{"train_loss": 0.22420093417167664, "global_step": 61774, "epoch": 694, "lr": 8.790286258313347e-05}
{"train_loss": 0.1982516050338745, "global_step": 61775, "epoch": 694, "lr": 8.790248448332276e-05}
{"train_loss": 0.20135733485221863, "global_step": 61776, "epoch": 694, "lr": 8.790210637841653e-05}
{"train_loss": 0.2403119057416916, "global_step": 61777, "epoch": 694, "lr": 8.790172826841485e-05}
{"train_loss": 0.2333449274301529, "global_step": 61778, "epoch": 694, "lr": 8.790135015331772e-05}
{"train_loss": 0.26857778429985046, "global_step": 61779, "epoch": 694, "lr": 8.790097203312522e-05}
{"train_loss": 0.2027094066143036, "global_step": 61780, "epoch": 694, "lr": 8.790059390783739e-05}
{"train_loss": 0.23835960030555725, "global_step": 61781, "epoch": 694, "lr": 8.790021577745429e-05}
{"train_loss": 0.321834534406662, "global_step": 61782, "epoch": 694, "lr": 8.789983764197597e-05}
{"train_loss": 0.23117545247077942, "global_step": 61783, "epoch": 694, "lr": 8.789945950140248e-05}
{"train_loss": 0.19651736319065094, "global_step": 61784, "epoch": 694, "lr": 8.789908135573386e-05}
{"train_loss": 0.24443292617797852, "global_step": 61785, "epoch": 694, "lr": 8.789870320497018e-05}
{"train_loss": 0.1849266141653061, "global_step": 61786, "epoch": 694, "lr": 8.789832504911148e-05}
{"train_loss": 0.29378682374954224, "global_step": 61787, "epoch": 694, "lr": 8.789794688815782e-05}
{"train_loss": 0.2688617706298828, "global_step": 61788, "epoch": 694, "lr": 8.789756872210923e-05}
{"train_loss": 0.26138994097709656, "global_step": 61789, "epoch": 694, "lr": 8.789719055096578e-05}
{"train_loss": 0.27510905265808105, "global_step": 61790, "epoch": 694, "lr": 8.78968123747275e-05}
{"train_loss": 0.2107928991317749, "global_step": 61791, "epoch": 694, "lr": 8.789643419339448e-05}
{"train_loss": 0.2238464504480362, "global_step": 61792, "epoch": 694, "lr": 8.789605600696674e-05}
{"train_loss": 0.21521992981433868, "global_step": 61793, "epoch": 694, "lr": 8.789567781544433e-05}
{"train_loss": 0.1586076319217682, "global_step": 61794, "epoch": 694, "lr": 8.78952996188273e-05}
{"train_loss": 0.2422802597284317, "global_step": 61795, "epoch": 694, "lr": 8.789492141711573e-05}
{"train_loss": 0.36814671754837036, "global_step": 61796, "epoch": 694, "lr": 8.789454321030963e-05}
{"train_loss": 0.3084903359413147, "global_step": 61797, "epoch": 694, "lr": 8.789416499840908e-05}
{"train_loss": 0.2482495903968811, "global_step": 61798, "epoch": 694, "lr": 8.789378678141413e-05}
{"train_loss": 0.18665079772472382, "global_step": 61799, "epoch": 694, "lr": 8.78934085593248e-05}
{"train_loss": 0.19609075784683228, "global_step": 61800, "epoch": 694, "lr": 8.789303033214118e-05}
{"train_loss": 0.19770947098731995, "global_step": 61801, "epoch": 694, "lr": 8.78926520998633e-05}
{"train_loss": 0.18621249496936798, "global_step": 61802, "epoch": 694, "lr": 8.789227386249122e-05}
{"train_loss": 0.27267664670944214, "global_step": 61803, "epoch": 694, "lr": 8.789189562002498e-05}
{"train_loss": 0.20899063348770142, "global_step": 61804, "epoch": 694, "lr": 8.789151737246464e-05}
{"train_loss": 0.1519022434949875, "global_step": 61805, "epoch": 694, "lr": 8.789113911981024e-05}
{"train_loss": 0.29781869053840637, "global_step": 61806, "epoch": 694, "lr": 8.789076086206183e-05}
{"train_loss": 0.20061208307743073, "global_step": 61807, "epoch": 694, "lr": 8.789038259921949e-05}
{"train_loss": 0.29076308012008667, "global_step": 61808, "epoch": 694, "lr": 8.789000433128324e-05}
{"train_loss": 0.2105293720960617, "global_step": 61809, "epoch": 694, "lr": 8.788962605825314e-05}
{"train_loss": 0.20388756692409515, "global_step": 61810, "epoch": 694, "lr": 8.788924778012924e-05}
{"train_loss": 0.24938668310642242, "global_step": 61811, "epoch": 694, "lr": 8.78888694969116e-05}
{"train_loss": 0.3487794101238251, "global_step": 61812, "epoch": 694, "lr": 8.788849120860025e-05}
{"train_loss": 0.2824893593788147, "global_step": 61813, "epoch": 694, "lr": 8.788811291519527e-05}
{"train_loss": 0.2602822780609131, "global_step": 61814, "epoch": 694, "lr": 8.788773461669669e-05}
{"train_loss": 0.1983659416437149, "global_step": 61815, "epoch": 694, "lr": 8.788735631310456e-05}
{"train_loss": 0.2645431160926819, "global_step": 61816, "epoch": 694, "lr": 8.788697800441895e-05}
{"train_loss": 0.3878619074821472, "global_step": 61817, "epoch": 694, "lr": 8.788659969063988e-05}
{"train_loss": 0.24842298030853271, "global_step": 61818, "epoch": 694, "lr": 8.788622137176743e-05}
{"train_loss": 0.3532946705818176, "global_step": 61819, "epoch": 694, "lr": 8.788584304780163e-05}
{"train_loss": 0.15765263140201569, "global_step": 61820, "epoch": 694, "lr": 8.788546471874255e-05}
{"train_loss": 0.24972398579120636, "global_step": 61821, "epoch": 694, "lr": 8.788508638459023e-05}
{"train_loss": 0.19187591969966888, "global_step": 61822, "epoch": 694, "lr": 8.788470804534473e-05}
{"train_loss": 0.18521149456501007, "global_step": 61823, "epoch": 694, "lr": 8.788432970100608e-05}
{"train_loss": 0.2757716476917267, "global_step": 61824, "epoch": 694, "lr": 8.788395135157434e-05}
{"train_loss": 0.22118604183197021, "global_step": 61825, "epoch": 694, "lr": 8.788357299704957e-05}
{"train_loss": 0.21261055767536163, "global_step": 61826, "epoch": 694, "lr": 8.788319463743182e-05}
{"train_loss": 0.18485425412654877, "global_step": 61827, "epoch": 694, "lr": 8.788281627272112e-05}
{"train_loss": 0.1946079134941101, "global_step": 61828, "epoch": 694, "lr": 8.788243790291755e-05}
{"train_loss": 0.28153306245803833, "global_step": 61829, "epoch": 694, "lr": 8.788205952802115e-05}
{"train_loss": 0.3245290219783783, "global_step": 61830, "epoch": 694, "lr": 8.788168114803197e-05}
{"train_loss": 0.33736392855644226, "global_step": 61831, "epoch": 694, "lr": 8.788130276295005e-05}
{"train_loss": 0.18965737521648407, "global_step": 61832, "epoch": 694, "lr": 8.788092437277546e-05}
{"train_loss": 0.338290810585022, "global_step": 61833, "epoch": 694, "lr": 8.788054597750823e-05}
{"train_loss": 0.16530951857566833, "global_step": 61834, "epoch": 694, "lr": 8.788016757714843e-05}
{"train_loss": 0.28409403562545776, "global_step": 61835, "epoch": 694, "lr": 8.787978917169611e-05}
{"train_loss": 0.19752706587314606, "global_step": 61836, "epoch": 694, "lr": 8.78794107611513e-05}
{"train_loss": 0.2422722429037094, "global_step": 61837, "epoch": 694, "lr": 8.787903234551406e-05}
{"train_loss": 0.24421367049217224, "global_step": 61838, "epoch": 694, "lr": 8.787865392478447e-05}
{"train_loss": 0.2951118052005768, "global_step": 61839, "epoch": 694, "lr": 8.787827549896255e-05}
{"train_loss": 0.2554202675819397, "global_step": 61840, "epoch": 694, "lr": 8.787789706804834e-05}
{"train_loss": 0.2673073410987854, "global_step": 61841, "epoch": 694, "lr": 8.787751863204192e-05}
{"train_loss": 0.2717494070529938, "global_step": 61842, "epoch": 694, "lr": 8.787714019094334e-05}
{"train_loss": 0.1885705292224884, "global_step": 61843, "epoch": 694, "lr": 8.787676174475264e-05}
{"train_loss": 0.17058269679546356, "global_step": 61844, "epoch": 694, "lr": 8.787638329346985e-05}
{"train_loss": 0.25901177525520325, "global_step": 61845, "epoch": 694, "lr": 8.787600483709505e-05}
{"train_loss": 0.18139849603176117, "global_step": 61846, "epoch": 694, "lr": 8.787562637562828e-05}
{"train_loss": 0.2831081748008728, "global_step": 61847, "epoch": 694, "lr": 8.78752479090696e-05}
{"train_loss": 0.22421139478683472, "global_step": 61848, "epoch": 694, "lr": 8.787486943741905e-05}
{"train_loss": 0.2835486829280853, "global_step": 61849, "epoch": 694, "lr": 8.787449096067669e-05}
{"train_loss": 0.23122870922088623, "global_step": 61850, "epoch": 694, "lr": 8.787411247884256e-05}
{"train_loss": 0.27319321036338806, "global_step": 61851, "epoch": 694, "lr": 8.787373399191673e-05}
{"train_loss": 0.2846677899360657, "global_step": 61852, "epoch": 694, "lr": 8.787335549989923e-05}
{"train_loss": 0.1864640712738037, "global_step": 61853, "epoch": 694, "lr": 8.787297700279013e-05}
{"train_loss": 0.24122209947430687, "global_step": 61854, "epoch": 694, "lr": 8.787259850058945e-05, "val_loss": 3.4864602088928223}
{"train_loss": 0.24678105115890503, "global_step": 61855, "epoch": 695, "lr": 8.787221999329726e-05}
{"train_loss": 0.2544569969177246, "global_step": 61856, "epoch": 695, "lr": 8.787184148091364e-05}
{"train_loss": 0.29584378004074097, "global_step": 61857, "epoch": 695, "lr": 8.787146296343858e-05}
{"train_loss": 0.2518276572227478, "global_step": 61858, "epoch": 695, "lr": 8.787108444087218e-05}
{"train_loss": 0.149422287940979, "global_step": 61859, "epoch": 695, "lr": 8.787070591321446e-05}
{"train_loss": 0.2167651653289795, "global_step": 61860, "epoch": 695, "lr": 8.787032738046551e-05}
{"train_loss": 0.19089798629283905, "global_step": 61861, "epoch": 695, "lr": 8.786994884262536e-05}
{"train_loss": 0.3820653259754181, "global_step": 61862, "epoch": 695, "lr": 8.786957029969402e-05}
{"train_loss": 0.21262668073177338, "global_step": 61863, "epoch": 695, "lr": 8.786919175167161e-05}
{"train_loss": 0.17667752504348755, "global_step": 61864, "epoch": 695, "lr": 8.786881319855813e-05}
{"train_loss": 0.2484475076198578, "global_step": 61865, "epoch": 695, "lr": 8.786843464035366e-05}
{"train_loss": 0.23422379791736603, "global_step": 61866, "epoch": 695, "lr": 8.786805607705823e-05}
{"train_loss": 0.20011933147907257, "global_step": 61867, "epoch": 695, "lr": 8.78676775086719e-05}
{"train_loss": 0.2123035490512848, "global_step": 61868, "epoch": 695, "lr": 8.786729893519474e-05}
{"train_loss": 0.3086131513118744, "global_step": 61869, "epoch": 695, "lr": 8.786692035662677e-05}
{"train_loss": 0.34604260325431824, "global_step": 61870, "epoch": 695, "lr": 8.786654177296806e-05}
{"train_loss": 0.20404928922653198, "global_step": 61871, "epoch": 695, "lr": 8.786616318421865e-05}
{"train_loss": 0.28035151958465576, "global_step": 61872, "epoch": 695, "lr": 8.78657845903786e-05}
{"train_loss": 0.22173868119716644, "global_step": 61873, "epoch": 695, "lr": 8.786540599144795e-05}
{"train_loss": 0.18092602491378784, "global_step": 61874, "epoch": 695, "lr": 8.786502738742677e-05}
{"train_loss": 0.19599288702011108, "global_step": 61875, "epoch": 695, "lr": 8.786464877831509e-05}
{"train_loss": 0.25332197546958923, "global_step": 61876, "epoch": 695, "lr": 8.786427016411298e-05}
{"train_loss": 0.2826298177242279, "global_step": 61877, "epoch": 695, "lr": 8.786389154482047e-05}
{"train_loss": 0.2710447609424591, "global_step": 61878, "epoch": 695, "lr": 8.786351292043762e-05}
{"train_loss": 0.21020498871803284, "global_step": 61879, "epoch": 695, "lr": 8.786313429096448e-05}
{"train_loss": 0.25239652395248413, "global_step": 61880, "epoch": 695, "lr": 8.78627556564011e-05}
{"train_loss": 0.3644433617591858, "global_step": 61881, "epoch": 695, "lr": 8.786237701674755e-05}
{"train_loss": 0.24770767986774445, "global_step": 61882, "epoch": 695, "lr": 8.786199837200386e-05}
{"train_loss": 0.246052086353302, "global_step": 61883, "epoch": 695, "lr": 8.786161972217008e-05}
{"train_loss": 0.19919845461845398, "global_step": 61884, "epoch": 695, "lr": 8.786124106724627e-05}
{"train_loss": 0.2902067303657532, "global_step": 61885, "epoch": 695, "lr": 8.786086240723248e-05}
{"train_loss": 0.2728866934776306, "global_step": 61886, "epoch": 695, "lr": 8.786048374212875e-05}
{"train_loss": 0.2290160059928894, "global_step": 61887, "epoch": 695, "lr": 8.786010507193514e-05}
{"train_loss": 0.26438599824905396, "global_step": 61888, "epoch": 695, "lr": 8.78597263966517e-05}
{"train_loss": 0.2162148654460907, "global_step": 61889, "epoch": 695, "lr": 8.78593477162785e-05}
{"train_loss": 0.21508333086967468, "global_step": 61890, "epoch": 695, "lr": 8.785896903081556e-05}
{"train_loss": 0.1781853437423706, "global_step": 61891, "epoch": 695, "lr": 8.785859034026293e-05}
{"train_loss": 0.29276931285858154, "global_step": 61892, "epoch": 695, "lr": 8.785821164462069e-05}
{"train_loss": 0.27561303973197937, "global_step": 61893, "epoch": 695, "lr": 8.785783294388888e-05}
{"train_loss": 0.42453238368034363, "global_step": 61894, "epoch": 695, "lr": 8.785745423806753e-05}
{"train_loss": 0.34546950459480286, "global_step": 61895, "epoch": 695, "lr": 8.785707552715672e-05}
{"train_loss": 0.28878435492515564, "global_step": 61896, "epoch": 695, "lr": 8.785669681115647e-05}
{"train_loss": 0.249907448887825, "global_step": 61897, "epoch": 695, "lr": 8.785631809006685e-05}
{"train_loss": 0.23517398536205292, "global_step": 61898, "epoch": 695, "lr": 8.785593936388792e-05}
{"train_loss": 0.23611938953399658, "global_step": 61899, "epoch": 695, "lr": 8.785556063261973e-05}
{"train_loss": 0.19778870046138763, "global_step": 61900, "epoch": 695, "lr": 8.785518189626231e-05}
{"train_loss": 0.21203458309173584, "global_step": 61901, "epoch": 695, "lr": 8.785480315481571e-05}
{"train_loss": 0.3701297640800476, "global_step": 61902, "epoch": 695, "lr": 8.785442440828002e-05}
{"train_loss": 0.41068753600120544, "global_step": 61903, "epoch": 695, "lr": 8.785404565665524e-05}
{"train_loss": 0.22933781147003174, "global_step": 61904, "epoch": 695, "lr": 8.785366689994145e-05}
{"train_loss": 0.20711098611354828, "global_step": 61905, "epoch": 695, "lr": 8.785328813813872e-05}
{"train_loss": 0.23980368673801422, "global_step": 61906, "epoch": 695, "lr": 8.785290937124704e-05}
{"train_loss": 0.2767108380794525, "global_step": 61907, "epoch": 695, "lr": 8.785253059926652e-05}
{"train_loss": 0.30700379610061646, "global_step": 61908, "epoch": 695, "lr": 8.785215182219719e-05}
{"train_loss": 0.23105677962303162, "global_step": 61909, "epoch": 695, "lr": 8.785177304003908e-05}
{"train_loss": 0.2307240515947342, "global_step": 61910, "epoch": 695, "lr": 8.785139425279227e-05}
{"train_loss": 0.32097750902175903, "global_step": 61911, "epoch": 695, "lr": 8.785101546045682e-05}
{"train_loss": 0.23827534914016724, "global_step": 61912, "epoch": 695, "lr": 8.785063666303275e-05}
{"train_loss": 0.2466997355222702, "global_step": 61913, "epoch": 695, "lr": 8.785025786052011e-05}
{"train_loss": 0.15518541634082794, "global_step": 61914, "epoch": 695, "lr": 8.784987905291898e-05}
{"train_loss": 0.20955346524715424, "global_step": 61915, "epoch": 695, "lr": 8.784950024022937e-05}
{"train_loss": 0.33527621626853943, "global_step": 61916, "epoch": 695, "lr": 8.784912142245139e-05}
{"train_loss": 0.2957463562488556, "global_step": 61917, "epoch": 695, "lr": 8.784874259958504e-05}
{"train_loss": 0.19425952434539795, "global_step": 61918, "epoch": 695, "lr": 8.784836377163039e-05}
{"train_loss": 0.34230920672416687, "global_step": 61919, "epoch": 695, "lr": 8.784798493858749e-05}
{"train_loss": 0.29589101672172546, "global_step": 61920, "epoch": 695, "lr": 8.784760610045639e-05}
{"train_loss": 0.21310661733150482, "global_step": 61921, "epoch": 695, "lr": 8.784722725723715e-05}
{"train_loss": 0.24485129117965698, "global_step": 61922, "epoch": 695, "lr": 8.784684840892978e-05}
{"train_loss": 0.20814277231693268, "global_step": 61923, "epoch": 695, "lr": 8.78464695555344e-05}
{"train_loss": 0.20606572926044464, "global_step": 61924, "epoch": 695, "lr": 8.784609069705101e-05}
{"train_loss": 0.223249152302742, "global_step": 61925, "epoch": 695, "lr": 8.784571183347966e-05}
{"train_loss": 0.2646395266056061, "global_step": 61926, "epoch": 695, "lr": 8.784533296482043e-05}
{"train_loss": 0.29161253571510315, "global_step": 61927, "epoch": 695, "lr": 8.784495409107335e-05}
{"train_loss": 0.2131074070930481, "global_step": 61928, "epoch": 695, "lr": 8.784457521223847e-05}
{"train_loss": 0.2030559480190277, "global_step": 61929, "epoch": 695, "lr": 8.784419632831586e-05}
{"train_loss": 0.3149314224720001, "global_step": 61930, "epoch": 695, "lr": 8.784381743930557e-05}
{"train_loss": 0.22394654154777527, "global_step": 61931, "epoch": 695, "lr": 8.784343854520763e-05}
{"train_loss": 0.20684640109539032, "global_step": 61932, "epoch": 695, "lr": 8.784305964602208e-05}
{"train_loss": 0.3011109530925751, "global_step": 61933, "epoch": 695, "lr": 8.784268074174901e-05}
{"train_loss": 0.26242297887802124, "global_step": 61934, "epoch": 695, "lr": 8.784230183238846e-05}
{"train_loss": 0.18349580466747284, "global_step": 61935, "epoch": 695, "lr": 8.784192291794046e-05}
{"train_loss": 0.1773243397474289, "global_step": 61936, "epoch": 695, "lr": 8.78415439984051e-05}
{"train_loss": 0.2502174377441406, "global_step": 61937, "epoch": 695, "lr": 8.784116507378239e-05}
{"train_loss": 0.19058586657047272, "global_step": 61938, "epoch": 695, "lr": 8.784078614407238e-05}
{"train_loss": 0.30390381813049316, "global_step": 61939, "epoch": 695, "lr": 8.784040720927516e-05}
{"train_loss": 0.2775620222091675, "global_step": 61940, "epoch": 695, "lr": 8.784002826939075e-05}
{"train_loss": 0.3899751603603363, "global_step": 61941, "epoch": 695, "lr": 8.783964932441921e-05}
{"train_loss": 0.28989070653915405, "global_step": 61942, "epoch": 695, "lr": 8.783927037436058e-05}
{"train_loss": 0.25313151033406844, "global_step": 61943, "epoch": 695, "lr": 8.783889141921493e-05, "val_loss": 3.414660930633545, "train_action_mse_error": 28.625930786132812}
{"train_loss": 0.16215981543064117, "global_step": 61944, "epoch": 696, "lr": 8.783851245898231e-05}
{"train_loss": 0.17779666185379028, "global_step": 61945, "epoch": 696, "lr": 8.783813349366275e-05}
{"train_loss": 0.25389668345451355, "global_step": 61946, "epoch": 696, "lr": 8.783775452325633e-05}
{"train_loss": 0.30468231439590454, "global_step": 61947, "epoch": 696, "lr": 8.783737554776307e-05}
{"train_loss": 0.2090909630060196, "global_step": 61948, "epoch": 696, "lr": 8.783699656718304e-05}
{"train_loss": 0.29883262515068054, "global_step": 61949, "epoch": 696, "lr": 8.783661758151629e-05}
{"train_loss": 0.2232566773891449, "global_step": 61950, "epoch": 696, "lr": 8.783623859076286e-05}
{"train_loss": 0.18558591604232788, "global_step": 61951, "epoch": 696, "lr": 8.783585959492282e-05}
{"train_loss": 0.21210907399654388, "global_step": 61952, "epoch": 696, "lr": 8.78354805939962e-05}
{"train_loss": 0.20902739465236664, "global_step": 61953, "epoch": 696, "lr": 8.783510158798308e-05}
{"train_loss": 0.32296186685562134, "global_step": 61954, "epoch": 696, "lr": 8.783472257688347e-05}
{"train_loss": 0.15396066009998322, "global_step": 61955, "epoch": 696, "lr": 8.783434356069744e-05}
{"train_loss": 0.2071290761232376, "global_step": 61956, "epoch": 696, "lr": 8.783396453942507e-05}
{"train_loss": 0.2657473087310791, "global_step": 61957, "epoch": 696, "lr": 8.783358551306635e-05}
{"train_loss": 0.31859293580055237, "global_step": 61958, "epoch": 696, "lr": 8.783320648162139e-05}
{"train_loss": 0.19572174549102783, "global_step": 61959, "epoch": 696, "lr": 8.783282744509021e-05}
{"train_loss": 0.24867510795593262, "global_step": 61960, "epoch": 696, "lr": 8.783244840347286e-05}
{"train_loss": 0.24441075325012207, "global_step": 61961, "epoch": 696, "lr": 8.783206935676941e-05}
{"train_loss": 0.18764658272266388, "global_step": 61962, "epoch": 696, "lr": 8.78316903049799e-05}
{"train_loss": 0.1949182152748108, "global_step": 61963, "epoch": 696, "lr": 8.783131124810438e-05}
{"train_loss": 0.3133063316345215, "global_step": 61964, "epoch": 696, "lr": 8.783093218614288e-05}
{"train_loss": 0.291899710893631, "global_step": 61965, "epoch": 696, "lr": 8.78305531190955e-05}
{"train_loss": 0.22384099662303925, "global_step": 61966, "epoch": 696, "lr": 8.783017404696226e-05}
{"train_loss": 0.26817962527275085, "global_step": 61967, "epoch": 696, "lr": 8.782979496974321e-05}
{"train_loss": 0.27558931708335876, "global_step": 61968, "epoch": 696, "lr": 8.782941588743841e-05}
{"train_loss": 0.18356762826442719, "global_step": 61969, "epoch": 696, "lr": 8.782903680004789e-05}
{"train_loss": 0.21673770248889923, "global_step": 61970, "epoch": 696, "lr": 8.782865770757171e-05}
{"train_loss": 0.20244468748569489, "global_step": 61971, "epoch": 696, "lr": 8.782827861000995e-05}
{"train_loss": 0.35325920581817627, "global_step": 61972, "epoch": 696, "lr": 8.782789950736264e-05}
{"train_loss": 0.2817736566066742, "global_step": 61973, "epoch": 696, "lr": 8.782752039962981e-05}
{"train_loss": 0.2906491756439209, "global_step": 61974, "epoch": 696, "lr": 8.782714128681155e-05}
{"train_loss": 0.33047008514404297, "global_step": 61975, "epoch": 696, "lr": 8.782676216890788e-05}
{"train_loss": 0.20773358643054962, "global_step": 61976, "epoch": 696, "lr": 8.782638304591889e-05}
{"train_loss": 0.2448340803384781, "global_step": 61977, "epoch": 696, "lr": 8.782600391784457e-05}
{"train_loss": 0.18877249956130981, "global_step": 61978, "epoch": 696, "lr": 8.782562478468501e-05}
{"train_loss": 0.14572949707508087, "global_step": 61979, "epoch": 696, "lr": 8.782524564644026e-05}
{"train_loss": 0.2686462700366974, "global_step": 61980, "epoch": 696, "lr": 8.782486650311037e-05}
{"train_loss": 0.21410028636455536, "global_step": 61981, "epoch": 696, "lr": 8.78244873546954e-05}
{"train_loss": 0.2631780207157135, "global_step": 61982, "epoch": 696, "lr": 8.782410820119536e-05}
{"train_loss": 0.2816324234008789, "global_step": 61983, "epoch": 696, "lr": 8.782372904261034e-05}
{"train_loss": 0.2528705596923828, "global_step": 61984, "epoch": 696, "lr": 8.78233498789404e-05}
{"train_loss": 0.2276393324136734, "global_step": 61985, "epoch": 696, "lr": 8.782297071018555e-05}
{"train_loss": 0.20401370525360107, "global_step": 61986, "epoch": 696, "lr": 8.782259153634588e-05}
{"train_loss": 0.1849401593208313, "global_step": 61987, "epoch": 696, "lr": 8.78222123574214e-05}
{"train_loss": 0.13249772787094116, "global_step": 61988, "epoch": 696, "lr": 8.782183317341222e-05}
{"train_loss": 0.3086722493171692, "global_step": 61989, "epoch": 696, "lr": 8.782145398431832e-05}
{"train_loss": 0.26918503642082214, "global_step": 61990, "epoch": 696, "lr": 8.78210747901398e-05}
{"train_loss": 0.29997313022613525, "global_step": 61991, "epoch": 696, "lr": 8.782069559087668e-05}
{"train_loss": 0.27665624022483826, "global_step": 61992, "epoch": 696, "lr": 8.782031638652906e-05}
{"train_loss": 0.19296051561832428, "global_step": 61993, "epoch": 696, "lr": 8.781993717709695e-05}
{"train_loss": 0.16874894499778748, "global_step": 61994, "epoch": 696, "lr": 8.781955796258042e-05}
{"train_loss": 0.1515168696641922, "global_step": 61995, "epoch": 696, "lr": 8.781917874297949e-05}
{"train_loss": 0.21270088851451874, "global_step": 61996, "epoch": 696, "lr": 8.781879951829424e-05}
{"train_loss": 0.2198004275560379, "global_step": 61997, "epoch": 696, "lr": 8.78184202885247e-05}
{"train_loss": 0.21504661440849304, "global_step": 61998, "epoch": 696, "lr": 8.781804105367095e-05}
{"train_loss": 0.24480989575386047, "global_step": 61999, "epoch": 696, "lr": 8.781766181373304e-05}
{"train_loss": 0.23633310198783875, "global_step": 62000, "epoch": 696, "lr": 8.781728256871097e-05}
{"train_loss": 0.17622213065624237, "global_step": 62001, "epoch": 696, "lr": 8.781690331860485e-05}
{"train_loss": 0.23590964078903198, "global_step": 62002, "epoch": 696, "lr": 8.78165240634147e-05}
{"train_loss": 0.19251449406147003, "global_step": 62003, "epoch": 696, "lr": 8.781614480314059e-05}
{"train_loss": 0.2092195451259613, "global_step": 62004, "epoch": 696, "lr": 8.781576553778255e-05}
{"train_loss": 0.2259105145931244, "global_step": 62005, "epoch": 696, "lr": 8.781538626734065e-05}
{"train_loss": 0.2824259102344513, "global_step": 62006, "epoch": 696, "lr": 8.781500699181494e-05}
{"train_loss": 0.21029694378376007, "global_step": 62007, "epoch": 696, "lr": 8.781462771120544e-05}
{"train_loss": 0.27240121364593506, "global_step": 62008, "epoch": 696, "lr": 8.781424842551224e-05}
{"train_loss": 0.28349992632865906, "global_step": 62009, "epoch": 696, "lr": 8.781386913473535e-05}
{"train_loss": 0.2662459909915924, "global_step": 62010, "epoch": 696, "lr": 8.781348983887487e-05}
{"train_loss": 0.31719911098480225, "global_step": 62011, "epoch": 696, "lr": 8.781311053793083e-05}
{"train_loss": 0.25882217288017273, "global_step": 62012, "epoch": 696, "lr": 8.781273123190328e-05}
{"train_loss": 0.23063768446445465, "global_step": 62013, "epoch": 696, "lr": 8.781235192079226e-05}
{"train_loss": 0.2883963882923126, "global_step": 62014, "epoch": 696, "lr": 8.781197260459783e-05}
{"train_loss": 0.2552434802055359, "global_step": 62015, "epoch": 696, "lr": 8.781159328332003e-05}
{"train_loss": 0.2946869134902954, "global_step": 62016, "epoch": 696, "lr": 8.781121395695895e-05}
{"train_loss": 0.2282063215970993, "global_step": 62017, "epoch": 696, "lr": 8.781083462551459e-05}
{"train_loss": 0.28413867950439453, "global_step": 62018, "epoch": 696, "lr": 8.781045528898703e-05}
{"train_loss": 0.4610660672187805, "global_step": 62019, "epoch": 696, "lr": 8.781007594737632e-05}
{"train_loss": 0.24964290857315063, "global_step": 62020, "epoch": 696, "lr": 8.78096966006825e-05}
{"train_loss": 0.19264180958271027, "global_step": 62021, "epoch": 696, "lr": 8.780931724890563e-05}
{"train_loss": 0.29017341136932373, "global_step": 62022, "epoch": 696, "lr": 8.780893789204576e-05}
{"train_loss": 0.19859884679317474, "global_step": 62023, "epoch": 696, "lr": 8.780855853010293e-05}
{"train_loss": 0.2009904831647873, "global_step": 62024, "epoch": 696, "lr": 8.780817916307722e-05}
{"train_loss": 0.20439192652702332, "global_step": 62025, "epoch": 696, "lr": 8.780779979096865e-05}
{"train_loss": 0.17929449677467346, "global_step": 62026, "epoch": 696, "lr": 8.780742041377727e-05}
{"train_loss": 0.2235158085823059, "global_step": 62027, "epoch": 696, "lr": 8.780704103150315e-05}
{"train_loss": 0.24429798126220703, "global_step": 62028, "epoch": 696, "lr": 8.780666164414635e-05}
{"train_loss": 0.2486063838005066, "global_step": 62029, "epoch": 696, "lr": 8.780628225170689e-05}
{"train_loss": 0.30406782031059265, "global_step": 62030, "epoch": 696, "lr": 8.780590285418484e-05}
{"train_loss": 0.2768454849720001, "global_step": 62031, "epoch": 696, "lr": 8.780552345158023e-05}
{"train_loss": 0.24012052744961856, "global_step": 62032, "epoch": 696, "lr": 8.780514404389315e-05, "val_loss": 3.420560836791992}
{"train_loss": 0.2561904191970825, "global_step": 62033, "epoch": 697, "lr": 8.780476463112364e-05}
{"train_loss": 0.1976722776889801, "global_step": 62034, "epoch": 697, "lr": 8.780438521327171e-05}
{"train_loss": 0.28733208775520325, "global_step": 62035, "epoch": 697, "lr": 8.780400579033745e-05}
{"train_loss": 0.18809759616851807, "global_step": 62036, "epoch": 697, "lr": 8.780362636232091e-05}
{"train_loss": 0.21886253356933594, "global_step": 62037, "epoch": 697, "lr": 8.780324692922213e-05}
{"train_loss": 0.3114506006240845, "global_step": 62038, "epoch": 697, "lr": 8.780286749104116e-05}
{"train_loss": 0.181480273604393, "global_step": 62039, "epoch": 697, "lr": 8.780248804777806e-05}
{"train_loss": 0.23865696787834167, "global_step": 62040, "epoch": 697, "lr": 8.780210859943287e-05}
{"train_loss": 0.22570592164993286, "global_step": 62041, "epoch": 697, "lr": 8.780172914600567e-05}
{"train_loss": 0.21932940185070038, "global_step": 62042, "epoch": 697, "lr": 8.780134968749647e-05}
{"train_loss": 0.18144835531711578, "global_step": 62043, "epoch": 697, "lr": 8.780097022390535e-05}
{"train_loss": 0.1825762540102005, "global_step": 62044, "epoch": 697, "lr": 8.780059075523232e-05}
{"train_loss": 0.23277778923511505, "global_step": 62045, "epoch": 697, "lr": 8.78002112814775e-05}
{"train_loss": 0.2602766752243042, "global_step": 62046, "epoch": 697, "lr": 8.779983180264088e-05}
{"train_loss": 0.2432781457901001, "global_step": 62047, "epoch": 697, "lr": 8.779945231872253e-05}
{"train_loss": 0.2381003201007843, "global_step": 62048, "epoch": 697, "lr": 8.779907282972252e-05}
{"train_loss": 0.2822153568267822, "global_step": 62049, "epoch": 697, "lr": 8.779869333564088e-05}
{"train_loss": 0.1975220888853073, "global_step": 62050, "epoch": 697, "lr": 8.779831383647765e-05}
{"train_loss": 0.15042926371097565, "global_step": 62051, "epoch": 697, "lr": 8.77979343322329e-05}
{"train_loss": 0.3066438138484955, "global_step": 62052, "epoch": 697, "lr": 8.779755482290672e-05}
{"train_loss": 0.2186015248298645, "global_step": 62053, "epoch": 697, "lr": 8.779717530849908e-05}
{"train_loss": 0.2068203091621399, "global_step": 62054, "epoch": 697, "lr": 8.779679578901007e-05}
{"train_loss": 0.24935318529605865, "global_step": 62055, "epoch": 697, "lr": 8.779641626443976e-05}
{"train_loss": 0.20908236503601074, "global_step": 62056, "epoch": 697, "lr": 8.779603673478818e-05}
{"train_loss": 0.2381342351436615, "global_step": 62057, "epoch": 697, "lr": 8.779565720005537e-05}
{"train_loss": 0.21190455555915833, "global_step": 62058, "epoch": 697, "lr": 8.77952776602414e-05}
{"train_loss": 0.25191670656204224, "global_step": 62059, "epoch": 697, "lr": 8.779489811534631e-05}
{"train_loss": 0.24379514157772064, "global_step": 62060, "epoch": 697, "lr": 8.779451856537018e-05}
{"train_loss": 0.18594586849212646, "global_step": 62061, "epoch": 697, "lr": 8.779413901031302e-05}
{"train_loss": 0.2227916121482849, "global_step": 62062, "epoch": 697, "lr": 8.77937594501749e-05}
{"train_loss": 0.2542276680469513, "global_step": 62063, "epoch": 697, "lr": 8.779337988495587e-05}
{"train_loss": 0.2640599012374878, "global_step": 62064, "epoch": 697, "lr": 8.779300031465598e-05}
{"train_loss": 0.2794882357120514, "global_step": 62065, "epoch": 697, "lr": 8.779262073927528e-05}
{"train_loss": 0.17660003900527954, "global_step": 62066, "epoch": 697, "lr": 8.779224115881383e-05}
{"train_loss": 0.31520798802375793, "global_step": 62067, "epoch": 697, "lr": 8.779186157327166e-05}
{"train_loss": 0.1862756311893463, "global_step": 62068, "epoch": 697, "lr": 8.779148198264885e-05}
{"train_loss": 0.28802135586738586, "global_step": 62069, "epoch": 697, "lr": 8.779110238694544e-05}
{"train_loss": 0.25349920988082886, "global_step": 62070, "epoch": 697, "lr": 8.779072278616145e-05}
{"train_loss": 0.4114876985549927, "global_step": 62071, "epoch": 697, "lr": 8.779034318029698e-05}
{"train_loss": 0.24039992690086365, "global_step": 62072, "epoch": 697, "lr": 8.778996356935206e-05}
{"train_loss": 0.17758160829544067, "global_step": 62073, "epoch": 697, "lr": 8.778958395332673e-05}
{"train_loss": 0.22286631166934967, "global_step": 62074, "epoch": 697, "lr": 8.778920433222105e-05}
{"train_loss": 0.26618289947509766, "global_step": 62075, "epoch": 697, "lr": 8.778882470603508e-05}
{"train_loss": 0.18036969006061554, "global_step": 62076, "epoch": 697, "lr": 8.778844507476885e-05}
{"train_loss": 0.17434576153755188, "global_step": 62077, "epoch": 697, "lr": 8.778806543842244e-05}
{"train_loss": 0.18314749002456665, "global_step": 62078, "epoch": 697, "lr": 8.778768579699588e-05}
{"train_loss": 0.23362772166728973, "global_step": 62079, "epoch": 697, "lr": 8.778730615048923e-05}
{"train_loss": 0.21807268261909485, "global_step": 62080, "epoch": 697, "lr": 8.778692649890252e-05}
{"train_loss": 0.20134520530700684, "global_step": 62081, "epoch": 697, "lr": 8.778654684223584e-05}
{"train_loss": 0.2651776969432831, "global_step": 62082, "epoch": 697, "lr": 8.778616718048921e-05}
{"train_loss": 0.22024774551391602, "global_step": 62083, "epoch": 697, "lr": 8.77857875136627e-05}
{"train_loss": 0.4023836851119995, "global_step": 62084, "epoch": 697, "lr": 8.778540784175633e-05}
{"train_loss": 0.18957316875457764, "global_step": 62085, "epoch": 697, "lr": 8.778502816477019e-05}
{"train_loss": 0.20370258390903473, "global_step": 62086, "epoch": 697, "lr": 8.77846484827043e-05}
{"train_loss": 0.1568811535835266, "global_step": 62087, "epoch": 697, "lr": 8.778426879555873e-05}
{"train_loss": 0.2990870475769043, "global_step": 62088, "epoch": 697, "lr": 8.778388910333353e-05}
{"train_loss": 0.1609661877155304, "global_step": 62089, "epoch": 697, "lr": 8.778350940602875e-05}
{"train_loss": 0.21211495995521545, "global_step": 62090, "epoch": 697, "lr": 8.778312970364443e-05}
{"train_loss": 0.24666637182235718, "global_step": 62091, "epoch": 697, "lr": 8.778274999618063e-05}
{"train_loss": 0.2350507378578186, "global_step": 62092, "epoch": 697, "lr": 8.77823702836374e-05}
{"train_loss": 0.17791800200939178, "global_step": 62093, "epoch": 697, "lr": 8.77819905660148e-05}
{"train_loss": 0.24088379740715027, "global_step": 62094, "epoch": 697, "lr": 8.778161084331285e-05}
{"train_loss": 0.21808761358261108, "global_step": 62095, "epoch": 697, "lr": 8.778123111553165e-05}
{"train_loss": 0.28936707973480225, "global_step": 62096, "epoch": 697, "lr": 8.778085138267121e-05}
{"train_loss": 0.26876604557037354, "global_step": 62097, "epoch": 697, "lr": 8.778047164473159e-05}
{"train_loss": 0.2056811898946762, "global_step": 62098, "epoch": 697, "lr": 8.778009190171285e-05}
{"train_loss": 0.2194105088710785, "global_step": 62099, "epoch": 697, "lr": 8.777971215361504e-05}
{"train_loss": 0.20407134294509888, "global_step": 62100, "epoch": 697, "lr": 8.77793324004382e-05}
{"train_loss": 0.19242365658283234, "global_step": 62101, "epoch": 697, "lr": 8.77789526421824e-05}
{"train_loss": 0.20711708068847656, "global_step": 62102, "epoch": 697, "lr": 8.777857287884767e-05}
{"train_loss": 0.24663524329662323, "global_step": 62103, "epoch": 697, "lr": 8.777819311043408e-05}
{"train_loss": 0.22619284689426422, "global_step": 62104, "epoch": 697, "lr": 8.777781333694167e-05}
{"train_loss": 0.26070189476013184, "global_step": 62105, "epoch": 697, "lr": 8.777743355837049e-05}
{"train_loss": 0.18656183779239655, "global_step": 62106, "epoch": 697, "lr": 8.777705377472061e-05}
{"train_loss": 0.21703584492206573, "global_step": 62107, "epoch": 697, "lr": 8.777667398599205e-05}
{"train_loss": 0.20719130337238312, "global_step": 62108, "epoch": 697, "lr": 8.777629419218488e-05}
{"train_loss": 0.24834328889846802, "global_step": 62109, "epoch": 697, "lr": 8.777591439329915e-05}
{"train_loss": 0.30839961767196655, "global_step": 62110, "epoch": 697, "lr": 8.77755345893349e-05}
{"train_loss": 0.1759081482887268, "global_step": 62111, "epoch": 697, "lr": 8.777515478029219e-05}
{"train_loss": 0.19247455894947052, "global_step": 62112, "epoch": 697, "lr": 8.777477496617109e-05}
{"train_loss": 0.3206152021884918, "global_step": 62113, "epoch": 697, "lr": 8.777439514697161e-05}
{"train_loss": 0.29190099239349365, "global_step": 62114, "epoch": 697, "lr": 8.777401532269384e-05}
{"train_loss": 0.30754339694976807, "global_step": 62115, "epoch": 697, "lr": 8.77736354933378e-05}
{"train_loss": 0.17598757147789001, "global_step": 62116, "epoch": 697, "lr": 8.777325565890357e-05}
{"train_loss": 0.16493277251720428, "global_step": 62117, "epoch": 697, "lr": 8.777287581939117e-05}
{"train_loss": 0.2942603826522827, "global_step": 62118, "epoch": 697, "lr": 8.777249597480067e-05}
{"train_loss": 0.3112294673919678, "global_step": 62119, "epoch": 697, "lr": 8.777211612513212e-05}
{"train_loss": 0.40733808279037476, "global_step": 62120, "epoch": 697, "lr": 8.777173627038557e-05}
{"train_loss": 0.23695136238349956, "global_step": 62121, "epoch": 697, "lr": 8.77713564105611e-05, "val_loss": 3.5690689086914062}
{"train_loss": 0.212384894490242, "global_step": 62122, "epoch": 698, "lr": 8.777097654565869e-05}
{"train_loss": 0.23729217052459717, "global_step": 62123, "epoch": 698, "lr": 8.777059667567844e-05}
{"train_loss": 0.1809753030538559, "global_step": 62124, "epoch": 698, "lr": 8.77702168006204e-05}
{"train_loss": 0.23194289207458496, "global_step": 62125, "epoch": 698, "lr": 8.776983692048462e-05}
{"train_loss": 0.2278914600610733, "global_step": 62126, "epoch": 698, "lr": 8.776945703527114e-05}
{"train_loss": 0.24406224489212036, "global_step": 62127, "epoch": 698, "lr": 8.776907714498001e-05}
{"train_loss": 0.30636855959892273, "global_step": 62128, "epoch": 698, "lr": 8.77686972496113e-05}
{"train_loss": 0.3463607132434845, "global_step": 62129, "epoch": 698, "lr": 8.776831734916504e-05}
{"train_loss": 0.1840057075023651, "global_step": 62130, "epoch": 698, "lr": 8.77679374436413e-05}
{"train_loss": 0.253458172082901, "global_step": 62131, "epoch": 698, "lr": 8.776755753304011e-05}
{"train_loss": 0.2787621021270752, "global_step": 62132, "epoch": 698, "lr": 8.776717761736155e-05}
{"train_loss": 0.28189602494239807, "global_step": 62133, "epoch": 698, "lr": 8.776679769660563e-05}
{"train_loss": 0.2609938979148865, "global_step": 62134, "epoch": 698, "lr": 8.776641777077243e-05}
{"train_loss": 0.2110452950000763, "global_step": 62135, "epoch": 698, "lr": 8.776603783986201e-05}
{"train_loss": 0.31200936436653137, "global_step": 62136, "epoch": 698, "lr": 8.77656579038744e-05}
{"train_loss": 0.18349377810955048, "global_step": 62137, "epoch": 698, "lr": 8.776527796280965e-05}
{"train_loss": 0.20119604468345642, "global_step": 62138, "epoch": 698, "lr": 8.776489801666783e-05}
{"train_loss": 0.18261466920375824, "global_step": 62139, "epoch": 698, "lr": 8.776451806544897e-05}
{"train_loss": 0.20489388704299927, "global_step": 62140, "epoch": 698, "lr": 8.776413810915314e-05}
{"train_loss": 0.2542612552642822, "global_step": 62141, "epoch": 698, "lr": 8.776375814778037e-05}
{"train_loss": 0.26068273186683655, "global_step": 62142, "epoch": 698, "lr": 8.776337818133074e-05}
{"train_loss": 0.25644466280937195, "global_step": 62143, "epoch": 698, "lr": 8.776299820980428e-05}
{"train_loss": 0.2550607919692993, "global_step": 62144, "epoch": 698, "lr": 8.776261823320102e-05}
{"train_loss": 0.2824535071849823, "global_step": 62145, "epoch": 698, "lr": 8.776223825152106e-05}
{"train_loss": 0.22585752606391907, "global_step": 62146, "epoch": 698, "lr": 8.776185826476442e-05}
{"train_loss": 0.2533109486103058, "global_step": 62147, "epoch": 698, "lr": 8.776147827293116e-05}
{"train_loss": 0.26648804545402527, "global_step": 62148, "epoch": 698, "lr": 8.776109827602133e-05}
{"train_loss": 0.27293914556503296, "global_step": 62149, "epoch": 698, "lr": 8.776071827403498e-05}
{"train_loss": 0.17935222387313843, "global_step": 62150, "epoch": 698, "lr": 8.776033826697216e-05}
{"train_loss": 0.22416076064109802, "global_step": 62151, "epoch": 698, "lr": 8.775995825483293e-05}
{"train_loss": 0.15921486914157867, "global_step": 62152, "epoch": 698, "lr": 8.775957823761733e-05}
{"train_loss": 0.21443811058998108, "global_step": 62153, "epoch": 698, "lr": 8.775919821532542e-05}
{"train_loss": 0.331208199262619, "global_step": 62154, "epoch": 698, "lr": 8.775881818795724e-05}
{"train_loss": 0.25878769159317017, "global_step": 62155, "epoch": 698, "lr": 8.775843815551283e-05}
{"train_loss": 0.18961867690086365, "global_step": 62156, "epoch": 698, "lr": 8.77580581179923e-05}
{"train_loss": 0.19250606000423431, "global_step": 62157, "epoch": 698, "lr": 8.775767807539562e-05}
{"train_loss": 0.2109050154685974, "global_step": 62158, "epoch": 698, "lr": 8.77572980277229e-05}
{"train_loss": 0.1874808520078659, "global_step": 62159, "epoch": 698, "lr": 8.775691797497416e-05}
{"train_loss": 0.2283310890197754, "global_step": 62160, "epoch": 698, "lr": 8.775653791714947e-05}
{"train_loss": 0.2939651608467102, "global_step": 62161, "epoch": 698, "lr": 8.775615785424887e-05}
{"train_loss": 0.14750780165195465, "global_step": 62162, "epoch": 698, "lr": 8.775577778627242e-05}
{"train_loss": 0.1776631474494934, "global_step": 62163, "epoch": 698, "lr": 8.775539771322016e-05}
{"train_loss": 0.18931664526462555, "global_step": 62164, "epoch": 698, "lr": 8.775501763509215e-05}
{"train_loss": 0.3146134316921234, "global_step": 62165, "epoch": 698, "lr": 8.775463755188844e-05}
{"train_loss": 0.2419634908437729, "global_step": 62166, "epoch": 698, "lr": 8.775425746360908e-05}
{"train_loss": 0.26760706305503845, "global_step": 62167, "epoch": 698, "lr": 8.775387737025411e-05}
{"train_loss": 0.2935628890991211, "global_step": 62168, "epoch": 698, "lr": 8.77534972718236e-05}
{"train_loss": 0.20740355551242828, "global_step": 62169, "epoch": 698, "lr": 8.77531171683176e-05}
{"train_loss": 0.1310281902551651, "global_step": 62170, "epoch": 698, "lr": 8.775273705973614e-05}
{"train_loss": 0.1476651281118393, "global_step": 62171, "epoch": 698, "lr": 8.775235694607928e-05}
{"train_loss": 0.17399805784225464, "global_step": 62172, "epoch": 698, "lr": 8.77519768273471e-05}
{"train_loss": 0.19054678082466125, "global_step": 62173, "epoch": 698, "lr": 8.77515967035396e-05}
{"train_loss": 0.1566108614206314, "global_step": 62174, "epoch": 698, "lr": 8.775121657465688e-05}
{"train_loss": 0.17571857571601868, "global_step": 62175, "epoch": 698, "lr": 8.775083644069896e-05}
{"train_loss": 0.16429303586483002, "global_step": 62176, "epoch": 698, "lr": 8.775045630166589e-05}
{"train_loss": 0.19359323382377625, "global_step": 62177, "epoch": 698, "lr": 8.775007615755775e-05}
{"train_loss": 0.24784952402114868, "global_step": 62178, "epoch": 698, "lr": 8.774969600837455e-05}
{"train_loss": 0.227442666888237, "global_step": 62179, "epoch": 698, "lr": 8.774931585411638e-05}
{"train_loss": 0.22420941293239594, "global_step": 62180, "epoch": 698, "lr": 8.774893569478328e-05}
{"train_loss": 0.27031952142715454, "global_step": 62181, "epoch": 698, "lr": 8.774855553037528e-05}
{"train_loss": 0.15963256359100342, "global_step": 62182, "epoch": 698, "lr": 8.774817536089246e-05}
{"train_loss": 0.21726971864700317, "global_step": 62183, "epoch": 698, "lr": 8.774779518633484e-05}
{"train_loss": 0.21125417947769165, "global_step": 62184, "epoch": 698, "lr": 8.77474150067025e-05}
{"train_loss": 0.2587212920188904, "global_step": 62185, "epoch": 698, "lr": 8.774703482199549e-05}
{"train_loss": 0.20358985662460327, "global_step": 62186, "epoch": 698, "lr": 8.774665463221384e-05}
{"train_loss": 0.32186833024024963, "global_step": 62187, "epoch": 698, "lr": 8.774627443735761e-05}
{"train_loss": 0.287907212972641, "global_step": 62188, "epoch": 698, "lr": 8.774589423742686e-05}
{"train_loss": 0.301171213388443, "global_step": 62189, "epoch": 698, "lr": 8.774551403242164e-05}
{"train_loss": 0.2387075126171112, "global_step": 62190, "epoch": 698, "lr": 8.774513382234197e-05}
{"train_loss": 0.1796659678220749, "global_step": 62191, "epoch": 698, "lr": 8.774475360718794e-05}
{"train_loss": 0.16143043339252472, "global_step": 62192, "epoch": 698, "lr": 8.77443733869596e-05}
{"train_loss": 0.3341623842716217, "global_step": 62193, "epoch": 698, "lr": 8.774399316165698e-05}
{"train_loss": 0.2793389856815338, "global_step": 62194, "epoch": 698, "lr": 8.774361293128016e-05}
{"train_loss": 0.25444430112838745, "global_step": 62195, "epoch": 698, "lr": 8.774323269582915e-05}
{"train_loss": 0.1604650318622589, "global_step": 62196, "epoch": 698, "lr": 8.774285245530402e-05}
{"train_loss": 0.2416166365146637, "global_step": 62197, "epoch": 698, "lr": 8.774247220970482e-05}
{"train_loss": 0.271253377199173, "global_step": 62198, "epoch": 698, "lr": 8.774209195903163e-05}
{"train_loss": 0.2573978304862976, "global_step": 62199, "epoch": 698, "lr": 8.774171170328444e-05}
{"train_loss": 0.2704625129699707, "global_step": 62200, "epoch": 698, "lr": 8.774133144246337e-05}
{"train_loss": 0.2545444965362549, "global_step": 62201, "epoch": 698, "lr": 8.774095117656841e-05}
{"train_loss": 0.23714953660964966, "global_step": 62202, "epoch": 698, "lr": 8.774057090559966e-05}
{"train_loss": 0.1943424940109253, "global_step": 62203, "epoch": 698, "lr": 8.774019062955714e-05}
{"train_loss": 0.2073870450258255, "global_step": 62204, "epoch": 698, "lr": 8.773981034844092e-05}
{"train_loss": 0.2182954102754593, "global_step": 62205, "epoch": 698, "lr": 8.773943006225105e-05}
{"train_loss": 0.2264760136604309, "global_step": 62206, "epoch": 698, "lr": 8.773904977098755e-05}
{"train_loss": 0.23885120451450348, "global_step": 62207, "epoch": 698, "lr": 8.77386694746505e-05}
{"train_loss": 0.18772591650485992, "global_step": 62208, "epoch": 698, "lr": 8.773828917323997e-05}
{"train_loss": 0.31500208377838135, "global_step": 62209, "epoch": 698, "lr": 8.773790886675595e-05}
{"train_loss": 0.23136799757400256, "global_step": 62210, "epoch": 698, "lr": 8.773752855519854e-05, "val_loss": 3.5927205085754395}
{"train_loss": 0.252701997756958, "global_step": 62211, "epoch": 699, "lr": 8.77371482385678e-05}
{"train_loss": 0.32990801334381104, "global_step": 62212, "epoch": 699, "lr": 8.773676791686374e-05}
{"train_loss": 0.2347428947687149, "global_step": 62213, "epoch": 699, "lr": 8.773638759008644e-05}
{"train_loss": 0.17720568180084229, "global_step": 62214, "epoch": 699, "lr": 8.773600725823594e-05}
{"train_loss": 0.23810537159442902, "global_step": 62215, "epoch": 699, "lr": 8.77356269213123e-05}
{"train_loss": 0.22786200046539307, "global_step": 62216, "epoch": 699, "lr": 8.773524657931556e-05}
{"train_loss": 0.2787730097770691, "global_step": 62217, "epoch": 699, "lr": 8.773486623224577e-05}
{"train_loss": 0.2832376956939697, "global_step": 62218, "epoch": 699, "lr": 8.7734485880103e-05}
{"train_loss": 0.30448511242866516, "global_step": 62219, "epoch": 699, "lr": 8.773410552288727e-05}
{"train_loss": 0.15130843222141266, "global_step": 62220, "epoch": 699, "lr": 8.773372516059865e-05}
{"train_loss": 0.22446401417255402, "global_step": 62221, "epoch": 699, "lr": 8.773334479323721e-05}
{"train_loss": 0.24519950151443481, "global_step": 62222, "epoch": 699, "lr": 8.773296442080298e-05}
{"train_loss": 0.2770324647426605, "global_step": 62223, "epoch": 699, "lr": 8.7732584043296e-05}
{"train_loss": 0.2685736119747162, "global_step": 62224, "epoch": 699, "lr": 8.773220366071635e-05}
{"train_loss": 0.25395116209983826, "global_step": 62225, "epoch": 699, "lr": 8.773182327306404e-05}
{"train_loss": 0.22093214094638824, "global_step": 62226, "epoch": 699, "lr": 8.773144288033916e-05}
{"train_loss": 0.20616769790649414, "global_step": 62227, "epoch": 699, "lr": 8.773106248254175e-05}
{"train_loss": 0.2122153341770172, "global_step": 62228, "epoch": 699, "lr": 8.773068207967185e-05}
{"train_loss": 0.20989683270454407, "global_step": 62229, "epoch": 699, "lr": 8.773030167172954e-05}
{"train_loss": 0.2130550593137741, "global_step": 62230, "epoch": 699, "lr": 8.772992125871484e-05}
{"train_loss": 0.2566440999507904, "global_step": 62231, "epoch": 699, "lr": 8.77295408406278e-05}
{"train_loss": 0.26382386684417725, "global_step": 62232, "epoch": 699, "lr": 8.772916041746849e-05}
{"train_loss": 0.20426374673843384, "global_step": 62233, "epoch": 699, "lr": 8.772877998923696e-05}
{"train_loss": 0.23403668403625488, "global_step": 62234, "epoch": 699, "lr": 8.772839955593325e-05}
{"train_loss": 0.17348085343837738, "global_step": 62235, "epoch": 699, "lr": 8.772801911755742e-05}
{"train_loss": 0.21156732738018036, "global_step": 62236, "epoch": 699, "lr": 8.772763867410951e-05}
{"train_loss": 0.21284201741218567, "global_step": 62237, "epoch": 699, "lr": 8.772725822558959e-05}
{"train_loss": 0.15678492188453674, "global_step": 62238, "epoch": 699, "lr": 8.772687777199769e-05}
{"train_loss": 0.17756380140781403, "global_step": 62239, "epoch": 699, "lr": 8.772649731333387e-05}
{"train_loss": 0.2966744899749756, "global_step": 62240, "epoch": 699, "lr": 8.77261168495982e-05}
{"train_loss": 0.33696064352989197, "global_step": 62241, "epoch": 699, "lr": 8.772573638079069e-05}
{"train_loss": 0.20494309067726135, "global_step": 62242, "epoch": 699, "lr": 8.772535590691142e-05}
{"train_loss": 0.21818533539772034, "global_step": 62243, "epoch": 699, "lr": 8.772497542796045e-05}
{"train_loss": 0.17701461911201477, "global_step": 62244, "epoch": 699, "lr": 8.772459494393781e-05}
{"train_loss": 0.2630351483821869, "global_step": 62245, "epoch": 699, "lr": 8.772421445484356e-05}
{"train_loss": 0.19283609092235565, "global_step": 62246, "epoch": 699, "lr": 8.772383396067774e-05}
{"train_loss": 0.1727791130542755, "global_step": 62247, "epoch": 699, "lr": 8.772345346144042e-05}
{"train_loss": 0.16696812212467194, "global_step": 62248, "epoch": 699, "lr": 8.772307295713164e-05}
{"train_loss": 0.2606165111064911, "global_step": 62249, "epoch": 699, "lr": 8.772269244775144e-05}
{"train_loss": 0.23626253008842468, "global_step": 62250, "epoch": 699, "lr": 8.77223119332999e-05}
{"train_loss": 0.3205951452255249, "global_step": 62251, "epoch": 699, "lr": 8.772193141377705e-05}
{"train_loss": 0.23794594407081604, "global_step": 62252, "epoch": 699, "lr": 8.772155088918294e-05}
{"train_loss": 0.217610165476799, "global_step": 62253, "epoch": 699, "lr": 8.772117035951764e-05}
{"train_loss": 0.18625834584236145, "global_step": 62254, "epoch": 699, "lr": 8.772078982478117e-05}
{"train_loss": 0.1719161421060562, "global_step": 62255, "epoch": 699, "lr": 8.772040928497361e-05}
{"train_loss": 0.2569368779659271, "global_step": 62256, "epoch": 699, "lr": 8.7720028740095e-05}
{"train_loss": 0.19858315587043762, "global_step": 62257, "epoch": 699, "lr": 8.77196481901454e-05}
{"train_loss": 0.26296234130859375, "global_step": 62258, "epoch": 699, "lr": 8.771926763512484e-05}
{"train_loss": 0.1946471780538559, "global_step": 62259, "epoch": 699, "lr": 8.77188870750334e-05}
{"train_loss": 0.32983750104904175, "global_step": 62260, "epoch": 699, "lr": 8.77185065098711e-05}
{"train_loss": 0.23141801357269287, "global_step": 62261, "epoch": 699, "lr": 8.771812593963801e-05}
{"train_loss": 0.21189676225185394, "global_step": 62262, "epoch": 699, "lr": 8.771774536433418e-05}
{"train_loss": 0.22752565145492554, "global_step": 62263, "epoch": 699, "lr": 8.771736478395966e-05}
{"train_loss": 0.20410314202308655, "global_step": 62264, "epoch": 699, "lr": 8.771698419851449e-05}
{"train_loss": 0.2640598714351654, "global_step": 62265, "epoch": 699, "lr": 8.771660360799873e-05}
{"train_loss": 0.23917508125305176, "global_step": 62266, "epoch": 699, "lr": 8.771622301241245e-05}
{"train_loss": 0.16839078068733215, "global_step": 62267, "epoch": 699, "lr": 8.771584241175568e-05}
{"train_loss": 0.32934093475341797, "global_step": 62268, "epoch": 699, "lr": 8.771546180602847e-05}
{"train_loss": 0.2730179727077484, "global_step": 62269, "epoch": 699, "lr": 8.771508119523087e-05}
{"train_loss": 0.3260587453842163, "global_step": 62270, "epoch": 699, "lr": 8.771470057936295e-05}
{"train_loss": 0.25804826617240906, "global_step": 62271, "epoch": 699, "lr": 8.771431995842474e-05}
{"train_loss": 0.3003254532814026, "global_step": 62272, "epoch": 699, "lr": 8.77139393324163e-05}
{"train_loss": 0.2080531120300293, "global_step": 62273, "epoch": 699, "lr": 8.771355870133768e-05}
{"train_loss": 0.19128482043743134, "global_step": 62274, "epoch": 699, "lr": 8.771317806518893e-05}
{"train_loss": 0.17520222067832947, "global_step": 62275, "epoch": 699, "lr": 8.771279742397011e-05}
{"train_loss": 0.2741667628288269, "global_step": 62276, "epoch": 699, "lr": 8.771241677768124e-05}
{"train_loss": 0.2024492621421814, "global_step": 62277, "epoch": 699, "lr": 8.771203612632242e-05}
{"train_loss": 0.2009909600019455, "global_step": 62278, "epoch": 699, "lr": 8.771165546989368e-05}
{"train_loss": 0.23363855481147766, "global_step": 62279, "epoch": 699, "lr": 8.771127480839506e-05}
{"train_loss": 0.2335871160030365, "global_step": 62280, "epoch": 699, "lr": 8.771089414182662e-05}
{"train_loss": 0.22674992680549622, "global_step": 62281, "epoch": 699, "lr": 8.77105134701884e-05}
{"train_loss": 0.3358561396598816, "global_step": 62282, "epoch": 699, "lr": 8.771013279348047e-05}
{"train_loss": 0.20585374534130096, "global_step": 62283, "epoch": 699, "lr": 8.770975211170287e-05}
{"train_loss": 0.20561136305332184, "global_step": 62284, "epoch": 699, "lr": 8.770937142485566e-05}
{"train_loss": 0.34958693385124207, "global_step": 62285, "epoch": 699, "lr": 8.770899073293888e-05}
{"train_loss": 0.277018278837204, "global_step": 62286, "epoch": 699, "lr": 8.770861003595259e-05}
{"train_loss": 0.24150030314922333, "global_step": 62287, "epoch": 699, "lr": 8.770822933389682e-05}
{"train_loss": 0.28154048323631287, "global_step": 62288, "epoch": 699, "lr": 8.770784862677164e-05}
{"train_loss": 0.22085581719875336, "global_step": 62289, "epoch": 699, "lr": 8.770746791457711e-05}
{"train_loss": 0.24112768471240997, "global_step": 62290, "epoch": 699, "lr": 8.770708719731329e-05}
{"train_loss": 0.22622616589069366, "global_step": 62291, "epoch": 699, "lr": 8.770670647498018e-05}
{"train_loss": 0.22874169051647186, "global_step": 62292, "epoch": 699, "lr": 8.770632574757787e-05}
{"train_loss": 0.2976182699203491, "global_step": 62293, "epoch": 699, "lr": 8.770594501510641e-05}
{"train_loss": 0.2430276870727539, "global_step": 62294, "epoch": 699, "lr": 8.770556427756585e-05}
{"train_loss": 0.21718229353427887, "global_step": 62295, "epoch": 699, "lr": 8.770518353495622e-05}
{"train_loss": 0.2583373188972473, "global_step": 62296, "epoch": 699, "lr": 8.77048027872776e-05}
{"train_loss": 0.18278343975543976, "global_step": 62297, "epoch": 699, "lr": 8.770442203453002e-05}
{"train_loss": 0.2646861970424652, "global_step": 62298, "epoch": 699, "lr": 8.770404127671356e-05}
{"train_loss": 0.23640297520696446, "global_step": 62299, "epoch": 699, "lr": 8.770366051382824e-05, "val_loss": 3.5100224018096924}
{"train_loss": 0.33692336082458496, "global_step": 62300, "epoch": 700, "lr": 8.770327974587411e-05}
{"train_loss": 0.24410295486450195, "global_step": 62301, "epoch": 700, "lr": 8.770289897285125e-05}
{"train_loss": 0.1902739703655243, "global_step": 62302, "epoch": 700, "lr": 8.770251819475968e-05}
{"train_loss": 0.21582160890102386, "global_step": 62303, "epoch": 700, "lr": 8.770213741159947e-05}
{"train_loss": 0.23234246671199799, "global_step": 62304, "epoch": 700, "lr": 8.770175662337068e-05}
{"train_loss": 0.21623371541500092, "global_step": 62305, "epoch": 700, "lr": 8.770137583007334e-05}
{"train_loss": 0.30114904046058655, "global_step": 62306, "epoch": 700, "lr": 8.77009950317075e-05}
{"train_loss": 0.19429074227809906, "global_step": 62307, "epoch": 700, "lr": 8.770061422827324e-05}
{"train_loss": 0.26019012928009033, "global_step": 62308, "epoch": 700, "lr": 8.770023341977058e-05}
{"train_loss": 0.2037506252527237, "global_step": 62309, "epoch": 700, "lr": 8.769985260619958e-05}
{"train_loss": 0.21378213167190552, "global_step": 62310, "epoch": 700, "lr": 8.76994717875603e-05}
{"train_loss": 0.29458749294281006, "global_step": 62311, "epoch": 700, "lr": 8.769909096385278e-05}
{"train_loss": 0.17152127623558044, "global_step": 62312, "epoch": 700, "lr": 8.769871013507708e-05}
{"train_loss": 0.26221734285354614, "global_step": 62313, "epoch": 700, "lr": 8.769832930123327e-05}
{"train_loss": 0.2684752345085144, "global_step": 62314, "epoch": 700, "lr": 8.769794846232136e-05}
{"train_loss": 0.20586630702018738, "global_step": 62315, "epoch": 700, "lr": 8.76975676183414e-05}
{"train_loss": 0.2148197740316391, "global_step": 62316, "epoch": 700, "lr": 8.769718676929348e-05}
{"train_loss": 0.1696041077375412, "global_step": 62317, "epoch": 700, "lr": 8.769680591517763e-05}
{"train_loss": 0.2058236300945282, "global_step": 62318, "epoch": 700, "lr": 8.769642505599391e-05}
{"train_loss": 0.2092631459236145, "global_step": 62319, "epoch": 700, "lr": 8.769604419174238e-05}
{"train_loss": 0.1729840189218521, "global_step": 62320, "epoch": 700, "lr": 8.769566332242305e-05}
{"train_loss": 0.1981864869594574, "global_step": 62321, "epoch": 700, "lr": 8.7695282448036e-05}
{"train_loss": 0.22246576845645905, "global_step": 62322, "epoch": 700, "lr": 8.769490156858128e-05}
{"train_loss": 0.3293609321117401, "global_step": 62323, "epoch": 700, "lr": 8.769452068405894e-05}
{"train_loss": 0.29036226868629456, "global_step": 62324, "epoch": 700, "lr": 8.769413979446906e-05}
{"train_loss": 0.24980074167251587, "global_step": 62325, "epoch": 700, "lr": 8.769375889981163e-05}
{"train_loss": 0.16302260756492615, "global_step": 62326, "epoch": 700, "lr": 8.769337800008675e-05}
{"train_loss": 0.2902105748653412, "global_step": 62327, "epoch": 700, "lr": 8.769299709529443e-05}
{"train_loss": 0.2116965353488922, "global_step": 62328, "epoch": 700, "lr": 8.769261618543476e-05}
{"train_loss": 0.27077946066856384, "global_step": 62329, "epoch": 700, "lr": 8.769223527050779e-05}
{"train_loss": 0.26451537013053894, "global_step": 62330, "epoch": 700, "lr": 8.769185435051355e-05}
{"train_loss": 0.3195542097091675, "global_step": 62331, "epoch": 700, "lr": 8.769147342545209e-05}
{"train_loss": 0.2257813811302185, "global_step": 62332, "epoch": 700, "lr": 8.769109249532348e-05}
{"train_loss": 0.3044080138206482, "global_step": 62333, "epoch": 700, "lr": 8.769071156012776e-05}
{"train_loss": 0.3445490300655365, "global_step": 62334, "epoch": 700, "lr": 8.769033061986498e-05}
{"train_loss": 0.3051100969314575, "global_step": 62335, "epoch": 700, "lr": 8.76899496745352e-05}
{"train_loss": 0.2180784046649933, "global_step": 62336, "epoch": 700, "lr": 8.768956872413846e-05}
{"train_loss": 0.2535281777381897, "global_step": 62337, "epoch": 700, "lr": 8.768918776867483e-05}
{"train_loss": 0.23866499960422516, "global_step": 62338, "epoch": 700, "lr": 8.768880680814433e-05}
{"train_loss": 0.13406552374362946, "global_step": 62339, "epoch": 700, "lr": 8.768842584254703e-05}
{"train_loss": 0.18862682580947876, "global_step": 62340, "epoch": 700, "lr": 8.7688044871883e-05}
{"train_loss": 0.3051760196685791, "global_step": 62341, "epoch": 700, "lr": 8.768766389615226e-05}
{"train_loss": 0.3727208077907562, "global_step": 62342, "epoch": 700, "lr": 8.768728291535486e-05}
{"train_loss": 0.2937650680541992, "global_step": 62343, "epoch": 700, "lr": 8.768690192949089e-05}
{"train_loss": 0.2961938679218292, "global_step": 62344, "epoch": 700, "lr": 8.768652093856036e-05}
{"train_loss": 0.23589973151683807, "global_step": 62345, "epoch": 700, "lr": 8.768613994256334e-05}
{"train_loss": 0.18314161896705627, "global_step": 62346, "epoch": 700, "lr": 8.768575894149987e-05}
{"train_loss": 0.24439986050128937, "global_step": 62347, "epoch": 700, "lr": 8.768537793537002e-05}
{"train_loss": 0.3020966053009033, "global_step": 62348, "epoch": 700, "lr": 8.768499692417382e-05}
{"train_loss": 0.23288728296756744, "global_step": 62349, "epoch": 700, "lr": 8.768461590791135e-05}
{"train_loss": 0.3059980869293213, "global_step": 62350, "epoch": 700, "lr": 8.768423488658262e-05}
{"train_loss": 0.15586687624454498, "global_step": 62351, "epoch": 700, "lr": 8.768385386018771e-05}
{"train_loss": 0.21866539120674133, "global_step": 62352, "epoch": 700, "lr": 8.768347282872667e-05}
{"train_loss": 0.23002326488494873, "global_step": 62353, "epoch": 700, "lr": 8.768309179219955e-05}
{"train_loss": 0.20096205174922943, "global_step": 62354, "epoch": 700, "lr": 8.768271075060639e-05}
{"train_loss": 0.17502300441265106, "global_step": 62355, "epoch": 700, "lr": 8.768232970394725e-05}
{"train_loss": 0.2538648545742035, "global_step": 62356, "epoch": 700, "lr": 8.76819486522222e-05}
{"train_loss": 0.21767562627792358, "global_step": 62357, "epoch": 700, "lr": 8.768156759543124e-05}
{"train_loss": 0.271553099155426, "global_step": 62358, "epoch": 700, "lr": 8.768118653357446e-05}
{"train_loss": 0.2607422173023224, "global_step": 62359, "epoch": 700, "lr": 8.768080546665191e-05}
{"train_loss": 0.24661798775196075, "global_step": 62360, "epoch": 700, "lr": 8.768042439466365e-05}
{"train_loss": 0.21775227785110474, "global_step": 62361, "epoch": 700, "lr": 8.76800433176097e-05}
{"train_loss": 0.19714894890785217, "global_step": 62362, "epoch": 700, "lr": 8.767966223549012e-05}
{"train_loss": 0.24085327982902527, "global_step": 62363, "epoch": 700, "lr": 8.767928114830497e-05}
{"train_loss": 0.25233590602874756, "global_step": 62364, "epoch": 700, "lr": 8.767890005605431e-05}
{"train_loss": 0.2566809356212616, "global_step": 62365, "epoch": 700, "lr": 8.767851895873818e-05}
{"train_loss": 0.28574684262275696, "global_step": 62366, "epoch": 700, "lr": 8.767813785635663e-05}
{"train_loss": 0.23074378073215485, "global_step": 62367, "epoch": 700, "lr": 8.767775674890972e-05}
{"train_loss": 0.20911811292171478, "global_step": 62368, "epoch": 700, "lr": 8.767737563639748e-05}
{"train_loss": 0.23435500264167786, "global_step": 62369, "epoch": 700, "lr": 8.767699451881999e-05}
{"train_loss": 0.19051635265350342, "global_step": 62370, "epoch": 700, "lr": 8.767661339617728e-05}
{"train_loss": 0.3635358214378357, "global_step": 62371, "epoch": 700, "lr": 8.767623226846941e-05}
{"train_loss": 0.24443192780017853, "global_step": 62372, "epoch": 700, "lr": 8.767585113569643e-05}
{"train_loss": 0.23278142511844635, "global_step": 62373, "epoch": 700, "lr": 8.76754699978584e-05}
{"train_loss": 0.2228401005268097, "global_step": 62374, "epoch": 700, "lr": 8.767508885495534e-05}
{"train_loss": 0.3039111793041229, "global_step": 62375, "epoch": 700, "lr": 8.767470770698735e-05}
{"train_loss": 0.3357556462287903, "global_step": 62376, "epoch": 700, "lr": 8.767432655395444e-05}
{"train_loss": 0.2918687164783478, "global_step": 62377, "epoch": 700, "lr": 8.767394539585668e-05}
{"train_loss": 0.26834413409233093, "global_step": 62378, "epoch": 700, "lr": 8.767356423269411e-05}
{"train_loss": 0.29122549295425415, "global_step": 62379, "epoch": 700, "lr": 8.76731830644668e-05}
{"train_loss": 0.2190101146697998, "global_step": 62380, "epoch": 700, "lr": 8.767280189117477e-05}
{"train_loss": 0.2612398862838745, "global_step": 62381, "epoch": 700, "lr": 8.767242071281811e-05}
{"train_loss": 0.23478353023529053, "global_step": 62382, "epoch": 700, "lr": 8.767203952939684e-05}
{"train_loss": 0.2579458951950073, "global_step": 62383, "epoch": 700, "lr": 8.767165834091105e-05}
{"train_loss": 0.23330233991146088, "global_step": 62384, "epoch": 700, "lr": 8.767127714736073e-05}
{"train_loss": 0.2690121531486511, "global_step": 62385, "epoch": 700, "lr": 8.767089594874599e-05}
{"train_loss": 0.24528656899929047, "global_step": 62386, "epoch": 700, "lr": 8.767051474506685e-05}
{"train_loss": 0.283687025308609, "global_step": 62387, "epoch": 700, "lr": 8.767013353632336e-05}
{"train_loss": 0.2466863366994965, "global_step": 62388, "epoch": 700, "lr": 8.766975232251558e-05, "train/sim_max_reward_0": 0.523629946896399, "train/sim_max_reward_1": 0.5506299673361928, "train/sim_max_reward_2": 0.4910605596137468, "train/sim_max_reward_3": 0.5493274977648938, "train/sim_max_reward_4": 0.9856202570439159, "train/sim_max_reward_5": 0.9549880656394512, "test/sim_max_reward_4500000": 0.9313110446206763, "test/sim_max_reward_4500001": 0.9066646807330663, "test/sim_max_reward_4500002": 0.9159886181924757, "test/sim_max_reward_4500003": 0.3176188255072351, "test/sim_max_reward_4500004": 0.9637447370889405, "test/sim_max_reward_4500005": 0.9651740114340744, "test/sim_max_reward_4500006": 0.3926680166732894, "test/sim_max_reward_4500007": 0.3967329337279982, "test/sim_max_reward_4500008": 0.798961094619379, "test/sim_max_reward_4500009": 0.05141983944905604, "test/sim_max_reward_4500010": 0.45558856376762374, "test/sim_max_reward_4500011": 0.9960258751839858, "test/sim_max_reward_4500012": 0.9995983430316966, "test/sim_max_reward_4500013": 0.5418091099248235, "test/sim_max_reward_4500014": 0.9730706149607669, "test/sim_max_reward_4500015": 0.9442852248963031, "test/sim_max_reward_4500016": 0.29991243421503344, "test/sim_max_reward_4500017": 0.14719613927182162, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9505856452306702, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.647286649326544, "test/sim_max_reward_4500022": 0.9294223566732004, "test/sim_max_reward_4500023": 0.5835273936143447, "test/sim_max_reward_4500024": 0.13830852784694117, "test/sim_max_reward_4500025": 0.8836464365385354, "test/sim_max_reward_4500026": 0.4796993272041067, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.39582578771046534, "test/sim_max_reward_4500029": 0.3812564856590374, "test/sim_max_reward_4500030": 0.9420341027302197, "test/sim_max_reward_4500031": 0.5898757596445752, "test/sim_max_reward_4500032": 0.8873444493859229, "test/sim_max_reward_4500033": 0.9861887236632613, "test/sim_max_reward_4500034": 0.7230840142081807, "test/sim_max_reward_4500035": 0.9640700371000576, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 1.0, "test/sim_max_reward_4500039": 0.9724934822394704, "test/sim_max_reward_4500040": 0.9612208334668371, "test/sim_max_reward_4500041": 0.1877390823683643, "test/sim_max_reward_4500042": 0.4732374547530057, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.9961876693480193, "test/sim_max_reward_4500045": 0.6802927299832009, "test/sim_max_reward_4500046": 0.4515073472003063, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.8613295351617039, "test/sim_max_reward_4500049": 0.6334420928587495, "train/mean_score": 0.6758760490490999, "test/mean_score": 0.6739475206242793, "val_loss": 3.3987319469451904, "train_action_mse_error": 25.035572052001953}
{"train_loss": 0.18853825330734253, "global_step": 62389, "epoch": 701, "lr": 8.766937110364357e-05}
{"train_loss": 0.2081831693649292, "global_step": 62390, "epoch": 701, "lr": 8.766898987970738e-05}
{"train_loss": 0.22126691043376923, "global_step": 62391, "epoch": 701, "lr": 8.766860865070703e-05}
{"train_loss": 0.25098538398742676, "global_step": 62392, "epoch": 701, "lr": 8.766822741664261e-05}
{"train_loss": 0.2283579707145691, "global_step": 62393, "epoch": 701, "lr": 8.766784617751415e-05}
{"train_loss": 0.2827127277851105, "global_step": 62394, "epoch": 701, "lr": 8.766746493332171e-05}
{"train_loss": 0.21055199205875397, "global_step": 62395, "epoch": 701, "lr": 8.766708368406535e-05}
{"train_loss": 0.16659344732761383, "global_step": 62396, "epoch": 701, "lr": 8.766670242974509e-05}
{"train_loss": 0.3082575500011444, "global_step": 62397, "epoch": 701, "lr": 8.766632117036101e-05}
{"train_loss": 0.2563527524471283, "global_step": 62398, "epoch": 701, "lr": 8.766593990591316e-05}
{"train_loss": 0.28743213415145874, "global_step": 62399, "epoch": 701, "lr": 8.766555863640155e-05}
{"train_loss": 0.2218964397907257, "global_step": 62400, "epoch": 701, "lr": 8.76651773618263e-05}
{"train_loss": 0.18691496551036835, "global_step": 62401, "epoch": 701, "lr": 8.766479608218741e-05}
{"train_loss": 0.21070516109466553, "global_step": 62402, "epoch": 701, "lr": 8.766441479748497e-05}
{"train_loss": 0.2458331286907196, "global_step": 62403, "epoch": 701, "lr": 8.766403350771899e-05}
{"train_loss": 0.1876579076051712, "global_step": 62404, "epoch": 701, "lr": 8.766365221288953e-05}
{"train_loss": 0.2284882366657257, "global_step": 62405, "epoch": 701, "lr": 8.766327091299667e-05}
{"train_loss": 0.24292230606079102, "global_step": 62406, "epoch": 701, "lr": 8.766288960804043e-05}
{"train_loss": 0.24134662747383118, "global_step": 62407, "epoch": 701, "lr": 8.766250829802088e-05}
{"train_loss": 0.2812054753303528, "global_step": 62408, "epoch": 701, "lr": 8.766212698293807e-05}
{"train_loss": 0.20745545625686646, "global_step": 62409, "epoch": 701, "lr": 8.766174566279202e-05}
{"train_loss": 0.2192280888557434, "global_step": 62410, "epoch": 701, "lr": 8.766136433758284e-05}
{"train_loss": 0.215837299823761, "global_step": 62411, "epoch": 701, "lr": 8.766098300731053e-05}
{"train_loss": 0.2020363211631775, "global_step": 62412, "epoch": 701, "lr": 8.766060167197517e-05}
{"train_loss": 0.2603774964809418, "global_step": 62413, "epoch": 701, "lr": 8.766022033157679e-05}
{"train_loss": 0.3032957911491394, "global_step": 62414, "epoch": 701, "lr": 8.765983898611547e-05}
{"train_loss": 0.26548880338668823, "global_step": 62415, "epoch": 701, "lr": 8.765945763559123e-05}
{"train_loss": 0.30523279309272766, "global_step": 62416, "epoch": 701, "lr": 8.765907628000413e-05}
{"train_loss": 0.25826773047447205, "global_step": 62417, "epoch": 701, "lr": 8.765869491935423e-05}
{"train_loss": 0.30009838938713074, "global_step": 62418, "epoch": 701, "lr": 8.765831355364158e-05}
{"train_loss": 0.25400692224502563, "global_step": 62419, "epoch": 701, "lr": 8.765793218286623e-05}
{"train_loss": 0.2660048007965088, "global_step": 62420, "epoch": 701, "lr": 8.765755080702822e-05}
{"train_loss": 0.20742221176624298, "global_step": 62421, "epoch": 701, "lr": 8.765716942612763e-05}
{"train_loss": 0.19170966744422913, "global_step": 62422, "epoch": 701, "lr": 8.765678804016448e-05}
{"train_loss": 0.21893882751464844, "global_step": 62423, "epoch": 701, "lr": 8.765640664913883e-05}
{"train_loss": 0.2490823119878769, "global_step": 62424, "epoch": 701, "lr": 8.765602525305075e-05}
{"train_loss": 0.19149579107761383, "global_step": 62425, "epoch": 701, "lr": 8.765564385190028e-05}
{"train_loss": 0.206796333193779, "global_step": 62426, "epoch": 701, "lr": 8.765526244568744e-05}
{"train_loss": 0.18428322672843933, "global_step": 62427, "epoch": 701, "lr": 8.765488103441233e-05}
{"train_loss": 0.15479280054569244, "global_step": 62428, "epoch": 701, "lr": 8.765449961807496e-05}
{"train_loss": 0.21589452028274536, "global_step": 62429, "epoch": 701, "lr": 8.765411819667543e-05}
{"train_loss": 0.18450358510017395, "global_step": 62430, "epoch": 701, "lr": 8.765373677021376e-05}
{"train_loss": 0.2834540903568268, "global_step": 62431, "epoch": 701, "lr": 8.765335533868999e-05}
{"train_loss": 0.2099151760339737, "global_step": 62432, "epoch": 701, "lr": 8.765297390210418e-05}
{"train_loss": 0.2013881951570511, "global_step": 62433, "epoch": 701, "lr": 8.76525924604564e-05}
{"train_loss": 0.266491562128067, "global_step": 62434, "epoch": 701, "lr": 8.765221101374667e-05}
{"train_loss": 0.16669206321239471, "global_step": 62435, "epoch": 701, "lr": 8.765182956197508e-05}
{"train_loss": 0.2552075982093811, "global_step": 62436, "epoch": 701, "lr": 8.765144810514165e-05}
{"train_loss": 0.28830671310424805, "global_step": 62437, "epoch": 701, "lr": 8.765106664324645e-05}
{"train_loss": 0.20325541496276855, "global_step": 62438, "epoch": 701, "lr": 8.765068517628951e-05}
{"train_loss": 0.1977243721485138, "global_step": 62439, "epoch": 701, "lr": 8.765030370427091e-05}
{"train_loss": 0.27026882767677307, "global_step": 62440, "epoch": 701, "lr": 8.764992222719068e-05}
{"train_loss": 0.19554659724235535, "global_step": 62441, "epoch": 701, "lr": 8.764954074504888e-05}
{"train_loss": 0.17609423398971558, "global_step": 62442, "epoch": 701, "lr": 8.764915925784553e-05}
{"train_loss": 0.20885393023490906, "global_step": 62443, "epoch": 701, "lr": 8.764877776558074e-05}
{"train_loss": 0.21467939019203186, "global_step": 62444, "epoch": 701, "lr": 8.764839626825452e-05}
{"train_loss": 0.1832786500453949, "global_step": 62445, "epoch": 701, "lr": 8.764801476586694e-05}
{"train_loss": 0.23683525621891022, "global_step": 62446, "epoch": 701, "lr": 8.764763325841803e-05}
{"train_loss": 0.18275028467178345, "global_step": 62447, "epoch": 701, "lr": 8.764725174590786e-05}
{"train_loss": 0.12329442799091339, "global_step": 62448, "epoch": 701, "lr": 8.764687022833649e-05}
{"train_loss": 0.1960473358631134, "global_step": 62449, "epoch": 701, "lr": 8.764648870570394e-05}
{"train_loss": 0.30360251665115356, "global_step": 62450, "epoch": 701, "lr": 8.764610717801029e-05}
{"train_loss": 0.2772551476955414, "global_step": 62451, "epoch": 701, "lr": 8.764572564525557e-05}
{"train_loss": 0.1991497427225113, "global_step": 62452, "epoch": 701, "lr": 8.764534410743984e-05}
{"train_loss": 0.16533519327640533, "global_step": 62453, "epoch": 701, "lr": 8.764496256456316e-05}
{"train_loss": 0.20127344131469727, "global_step": 62454, "epoch": 701, "lr": 8.764458101662556e-05}
{"train_loss": 0.16995830833911896, "global_step": 62455, "epoch": 701, "lr": 8.764419946362712e-05}
{"train_loss": 0.18218311667442322, "global_step": 62456, "epoch": 701, "lr": 8.764381790556787e-05}
{"train_loss": 0.24078845977783203, "global_step": 62457, "epoch": 701, "lr": 8.764343634244787e-05}
{"train_loss": 0.2750149369239807, "global_step": 62458, "epoch": 701, "lr": 8.764305477426718e-05}
{"train_loss": 0.20404542982578278, "global_step": 62459, "epoch": 701, "lr": 8.764267320102582e-05}
{"train_loss": 0.3040185868740082, "global_step": 62460, "epoch": 701, "lr": 8.764229162272387e-05}
{"train_loss": 0.18330231308937073, "global_step": 62461, "epoch": 701, "lr": 8.764191003936137e-05}
{"train_loss": 0.1970384418964386, "global_step": 62462, "epoch": 701, "lr": 8.764152845093837e-05}
{"train_loss": 0.24217785894870758, "global_step": 62463, "epoch": 701, "lr": 8.764114685745493e-05}
{"train_loss": 0.29763689637184143, "global_step": 62464, "epoch": 701, "lr": 8.76407652589111e-05}
{"train_loss": 0.23541149497032166, "global_step": 62465, "epoch": 701, "lr": 8.764038365530694e-05}
{"train_loss": 0.2166721671819687, "global_step": 62466, "epoch": 701, "lr": 8.764000204664247e-05}
{"train_loss": 0.1656835824251175, "global_step": 62467, "epoch": 701, "lr": 8.763962043291776e-05}
{"train_loss": 0.16010314226150513, "global_step": 62468, "epoch": 701, "lr": 8.763923881413286e-05}
{"train_loss": 0.1435094028711319, "global_step": 62469, "epoch": 701, "lr": 8.763885719028784e-05}
{"train_loss": 0.2274164855480194, "global_step": 62470, "epoch": 701, "lr": 8.763847556138272e-05}
{"train_loss": 0.192827969789505, "global_step": 62471, "epoch": 701, "lr": 8.763809392741758e-05}
{"train_loss": 0.28165820240974426, "global_step": 62472, "epoch": 701, "lr": 8.763771228839244e-05}
{"train_loss": 0.22738198935985565, "global_step": 62473, "epoch": 701, "lr": 8.763733064430737e-05}
{"train_loss": 0.2070034146308899, "global_step": 62474, "epoch": 701, "lr": 8.763694899516243e-05}
{"train_loss": 0.23450250923633575, "global_step": 62475, "epoch": 701, "lr": 8.763656734095766e-05}
{"train_loss": 0.23557420074939728, "global_step": 62476, "epoch": 701, "lr": 8.76361856816931e-05}
{"train_loss": 0.2248964530698369, "global_step": 62477, "epoch": 701, "lr": 8.763580401736882e-05, "val_loss": 3.5624659061431885}
{"train_loss": 0.305345356464386, "global_step": 62478, "epoch": 702, "lr": 8.763542234798487e-05}
{"train_loss": 0.12205982208251953, "global_step": 62479, "epoch": 702, "lr": 8.763504067354129e-05}
{"train_loss": 0.2521823048591614, "global_step": 62480, "epoch": 702, "lr": 8.763465899403814e-05}
{"train_loss": 0.21303781867027283, "global_step": 62481, "epoch": 702, "lr": 8.763427730947547e-05}
{"train_loss": 0.22870391607284546, "global_step": 62482, "epoch": 702, "lr": 8.763389561985334e-05}
{"train_loss": 0.24127958714962006, "global_step": 62483, "epoch": 702, "lr": 8.763351392517177e-05}
{"train_loss": 0.19376279413700104, "global_step": 62484, "epoch": 702, "lr": 8.763313222543083e-05}
{"train_loss": 0.2512750029563904, "global_step": 62485, "epoch": 702, "lr": 8.76327505206306e-05}
{"train_loss": 0.21661241352558136, "global_step": 62486, "epoch": 702, "lr": 8.76323688107711e-05}
{"train_loss": 0.36612534523010254, "global_step": 62487, "epoch": 702, "lr": 8.763198709585238e-05}
{"train_loss": 0.25849881768226624, "global_step": 62488, "epoch": 702, "lr": 8.76316053758745e-05}
{"train_loss": 0.28006941080093384, "global_step": 62489, "epoch": 702, "lr": 8.763122365083749e-05}
{"train_loss": 0.21920718252658844, "global_step": 62490, "epoch": 702, "lr": 8.763084192074144e-05}
{"train_loss": 0.18487296998500824, "global_step": 62491, "epoch": 702, "lr": 8.763046018558639e-05}
{"train_loss": 0.27683132886886597, "global_step": 62492, "epoch": 702, "lr": 8.763007844537236e-05}
{"train_loss": 0.24804149568080902, "global_step": 62493, "epoch": 702, "lr": 8.762969670009944e-05}
{"train_loss": 0.2359044849872589, "global_step": 62494, "epoch": 702, "lr": 8.762931494976767e-05}
{"train_loss": 0.20424805581569672, "global_step": 62495, "epoch": 702, "lr": 8.762893319437709e-05}
{"train_loss": 0.2605380713939667, "global_step": 62496, "epoch": 702, "lr": 8.762855143392776e-05}
{"train_loss": 0.18759159743785858, "global_step": 62497, "epoch": 702, "lr": 8.762816966841973e-05}
{"train_loss": 0.278766393661499, "global_step": 62498, "epoch": 702, "lr": 8.762778789785305e-05}
{"train_loss": 0.28079742193222046, "global_step": 62499, "epoch": 702, "lr": 8.762740612222777e-05}
{"train_loss": 0.2809735834598541, "global_step": 62500, "epoch": 702, "lr": 8.762702434154395e-05}
{"train_loss": 0.23675395548343658, "global_step": 62501, "epoch": 702, "lr": 8.762664255580163e-05}
{"train_loss": 0.2633187770843506, "global_step": 62502, "epoch": 702, "lr": 8.762626076500086e-05}
{"train_loss": 0.24149630963802338, "global_step": 62503, "epoch": 702, "lr": 8.762587896914172e-05}
{"train_loss": 0.2613937556743622, "global_step": 62504, "epoch": 702, "lr": 8.762549716822421e-05}
{"train_loss": 0.1679801493883133, "global_step": 62505, "epoch": 702, "lr": 8.762511536224843e-05}
{"train_loss": 0.3277572989463806, "global_step": 62506, "epoch": 702, "lr": 8.762473355121441e-05}
{"train_loss": 0.2679867148399353, "global_step": 62507, "epoch": 702, "lr": 8.76243517351222e-05}
{"train_loss": 0.22538766264915466, "global_step": 62508, "epoch": 702, "lr": 8.762396991397186e-05}
{"train_loss": 0.18583279848098755, "global_step": 62509, "epoch": 702, "lr": 8.762358808776343e-05}
{"train_loss": 0.21364392340183258, "global_step": 62510, "epoch": 702, "lr": 8.762320625649696e-05}
{"train_loss": 0.276724636554718, "global_step": 62511, "epoch": 702, "lr": 8.762282442017252e-05}
{"train_loss": 0.22640907764434814, "global_step": 62512, "epoch": 702, "lr": 8.762244257879015e-05}
{"train_loss": 0.1872539520263672, "global_step": 62513, "epoch": 702, "lr": 8.76220607323499e-05}
{"train_loss": 0.2150820940732956, "global_step": 62514, "epoch": 702, "lr": 8.762167888085182e-05}
{"train_loss": 0.21407441794872284, "global_step": 62515, "epoch": 702, "lr": 8.762129702429595e-05}
{"train_loss": 0.20393958687782288, "global_step": 62516, "epoch": 702, "lr": 8.762091516268238e-05}
{"train_loss": 0.20040589570999146, "global_step": 62517, "epoch": 702, "lr": 8.762053329601112e-05}
{"train_loss": 0.1551644653081894, "global_step": 62518, "epoch": 702, "lr": 8.762015142428224e-05}
{"train_loss": 0.2855512499809265, "global_step": 62519, "epoch": 702, "lr": 8.76197695474958e-05}
{"train_loss": 0.21030250191688538, "global_step": 62520, "epoch": 702, "lr": 8.761938766565183e-05}
{"train_loss": 0.17658638954162598, "global_step": 62521, "epoch": 702, "lr": 8.761900577875039e-05}
{"train_loss": 0.2129891812801361, "global_step": 62522, "epoch": 702, "lr": 8.761862388679155e-05}
{"train_loss": 0.2683033049106598, "global_step": 62523, "epoch": 702, "lr": 8.761824198977533e-05}
{"train_loss": 0.19054128229618073, "global_step": 62524, "epoch": 702, "lr": 8.761786008770181e-05}
{"train_loss": 0.34164679050445557, "global_step": 62525, "epoch": 702, "lr": 8.7617478180571e-05}
{"train_loss": 0.216373011469841, "global_step": 62526, "epoch": 702, "lr": 8.7617096268383e-05}
{"train_loss": 0.19087693095207214, "global_step": 62527, "epoch": 702, "lr": 8.761671435113784e-05}
{"train_loss": 0.22225967049598694, "global_step": 62528, "epoch": 702, "lr": 8.761633242883555e-05}
{"train_loss": 0.17187941074371338, "global_step": 62529, "epoch": 702, "lr": 8.761595050147623e-05}
{"train_loss": 0.23438790440559387, "global_step": 62530, "epoch": 702, "lr": 8.76155685690599e-05}
{"train_loss": 0.2817786633968353, "global_step": 62531, "epoch": 702, "lr": 8.76151866315866e-05}
{"train_loss": 0.26360562443733215, "global_step": 62532, "epoch": 702, "lr": 8.76148046890564e-05}
{"train_loss": 0.2067107856273651, "global_step": 62533, "epoch": 702, "lr": 8.761442274146936e-05}
{"train_loss": 0.2445215880870819, "global_step": 62534, "epoch": 702, "lr": 8.76140407888255e-05}
{"train_loss": 0.17919538915157318, "global_step": 62535, "epoch": 702, "lr": 8.761365883112491e-05}
{"train_loss": 0.26783499121665955, "global_step": 62536, "epoch": 702, "lr": 8.76132768683676e-05}
{"train_loss": 0.2232353240251541, "global_step": 62537, "epoch": 702, "lr": 8.761289490055367e-05}
{"train_loss": 0.20068293809890747, "global_step": 62538, "epoch": 702, "lr": 8.761251292768313e-05}
{"train_loss": 0.24615773558616638, "global_step": 62539, "epoch": 702, "lr": 8.761213094975604e-05}
{"train_loss": 0.22065073251724243, "global_step": 62540, "epoch": 702, "lr": 8.761174896677247e-05}
{"train_loss": 0.2294287085533142, "global_step": 62541, "epoch": 702, "lr": 8.761136697873246e-05}
{"train_loss": 0.22991639375686646, "global_step": 62542, "epoch": 702, "lr": 8.761098498563605e-05}
{"train_loss": 0.22545026242733002, "global_step": 62543, "epoch": 702, "lr": 8.76106029874833e-05}
{"train_loss": 0.20759333670139313, "global_step": 62544, "epoch": 702, "lr": 8.761022098427427e-05}
{"train_loss": 0.19263890385627747, "global_step": 62545, "epoch": 702, "lr": 8.760983897600901e-05}
{"train_loss": 0.14673729240894318, "global_step": 62546, "epoch": 702, "lr": 8.760945696268756e-05}
{"train_loss": 0.2708849012851715, "global_step": 62547, "epoch": 702, "lr": 8.760907494430996e-05}
{"train_loss": 0.28001660108566284, "global_step": 62548, "epoch": 702, "lr": 8.76086929208763e-05}
{"train_loss": 0.2844676673412323, "global_step": 62549, "epoch": 702, "lr": 8.760831089238662e-05}
{"train_loss": 0.23551100492477417, "global_step": 62550, "epoch": 702, "lr": 8.760792885884092e-05}
{"train_loss": 0.28171733021736145, "global_step": 62551, "epoch": 702, "lr": 8.760754682023933e-05}
{"train_loss": 0.3168598413467407, "global_step": 62552, "epoch": 702, "lr": 8.760716477658184e-05}
{"train_loss": 0.2214888632297516, "global_step": 62553, "epoch": 702, "lr": 8.760678272786854e-05}
{"train_loss": 0.12144012004137039, "global_step": 62554, "epoch": 702, "lr": 8.760640067409946e-05}
{"train_loss": 0.24591395258903503, "global_step": 62555, "epoch": 702, "lr": 8.760601861527466e-05}
{"train_loss": 0.23117658495903015, "global_step": 62556, "epoch": 702, "lr": 8.760563655139418e-05}
{"train_loss": 0.17041932046413422, "global_step": 62557, "epoch": 702, "lr": 8.760525448245811e-05}
{"train_loss": 0.2835107743740082, "global_step": 62558, "epoch": 702, "lr": 8.760487240846644e-05}
{"train_loss": 0.27050068974494934, "global_step": 62559, "epoch": 702, "lr": 8.760449032941926e-05}
{"train_loss": 0.15072740614414215, "global_step": 62560, "epoch": 702, "lr": 8.760410824531662e-05}
{"train_loss": 0.1874438226222992, "global_step": 62561, "epoch": 702, "lr": 8.760372615615857e-05}
{"train_loss": 0.187898650765419, "global_step": 62562, "epoch": 702, "lr": 8.760334406194515e-05}
{"train_loss": 0.20309261977672577, "global_step": 62563, "epoch": 702, "lr": 8.760296196267642e-05}
{"train_loss": 0.19382379949092865, "global_step": 62564, "epoch": 702, "lr": 8.760257985835243e-05}
{"train_loss": 0.22284245491027832, "global_step": 62565, "epoch": 702, "lr": 8.760219774897324e-05}
{"train_loss": 0.2312753534551417, "global_step": 62566, "epoch": 702, "lr": 8.760181563453889e-05, "val_loss": 3.5671918392181396}
{"train_loss": 0.29003769159317017, "global_step": 62567, "epoch": 703, "lr": 8.760143351504941e-05}
{"train_loss": 0.2811620533466339, "global_step": 62568, "epoch": 703, "lr": 8.76010513905049e-05}
{"train_loss": 0.4668251872062683, "global_step": 62569, "epoch": 703, "lr": 8.760066926090539e-05}
{"train_loss": 0.2747204303741455, "global_step": 62570, "epoch": 703, "lr": 8.760028712625092e-05}
{"train_loss": 0.23002420365810394, "global_step": 62571, "epoch": 703, "lr": 8.759990498654154e-05}
{"train_loss": 0.17750953137874603, "global_step": 62572, "epoch": 703, "lr": 8.759952284177733e-05}
{"train_loss": 0.18054401874542236, "global_step": 62573, "epoch": 703, "lr": 8.759914069195831e-05}
{"train_loss": 0.15470196306705475, "global_step": 62574, "epoch": 703, "lr": 8.759875853708454e-05}
{"train_loss": 0.20240098237991333, "global_step": 62575, "epoch": 703, "lr": 8.759837637715608e-05}
{"train_loss": 0.23236525058746338, "global_step": 62576, "epoch": 703, "lr": 8.759799421217298e-05}
{"train_loss": 0.2594216465950012, "global_step": 62577, "epoch": 703, "lr": 8.75976120421353e-05}
{"train_loss": 0.22657141089439392, "global_step": 62578, "epoch": 703, "lr": 8.759722986704307e-05}
{"train_loss": 0.3213014602661133, "global_step": 62579, "epoch": 703, "lr": 8.759684768689635e-05}
{"train_loss": 0.2570442259311676, "global_step": 62580, "epoch": 703, "lr": 8.759646550169518e-05}
{"train_loss": 0.30672335624694824, "global_step": 62581, "epoch": 703, "lr": 8.759608331143962e-05}
{"train_loss": 0.24387723207473755, "global_step": 62582, "epoch": 703, "lr": 8.759570111612974e-05}
{"train_loss": 0.1926594078540802, "global_step": 62583, "epoch": 703, "lr": 8.759531891576557e-05}
{"train_loss": 0.1796540468931198, "global_step": 62584, "epoch": 703, "lr": 8.759493671034718e-05}
{"train_loss": 0.18768224120140076, "global_step": 62585, "epoch": 703, "lr": 8.75945544998746e-05}
{"train_loss": 0.2545587420463562, "global_step": 62586, "epoch": 703, "lr": 8.75941722843479e-05}
{"train_loss": 0.43976059556007385, "global_step": 62587, "epoch": 703, "lr": 8.759379006376711e-05}
{"train_loss": 0.27983734011650085, "global_step": 62588, "epoch": 703, "lr": 8.759340783813228e-05}
{"train_loss": 0.27094635367393494, "global_step": 62589, "epoch": 703, "lr": 8.75930256074435e-05}
{"train_loss": 0.20150348544120789, "global_step": 62590, "epoch": 703, "lr": 8.759264337170079e-05}
{"train_loss": 0.2628597617149353, "global_step": 62591, "epoch": 703, "lr": 8.75922611309042e-05}
{"train_loss": 0.2657518684864044, "global_step": 62592, "epoch": 703, "lr": 8.759187888505377e-05}
{"train_loss": 0.1768815517425537, "global_step": 62593, "epoch": 703, "lr": 8.759149663414961e-05}
{"train_loss": 0.38132649660110474, "global_step": 62594, "epoch": 703, "lr": 8.759111437819171e-05}
{"train_loss": 0.3511742353439331, "global_step": 62595, "epoch": 703, "lr": 8.759073211718015e-05}
{"train_loss": 0.2487914264202118, "global_step": 62596, "epoch": 703, "lr": 8.759034985111498e-05}
{"train_loss": 0.2179948389530182, "global_step": 62597, "epoch": 703, "lr": 8.758996757999622e-05}
{"train_loss": 0.22522743046283722, "global_step": 62598, "epoch": 703, "lr": 8.758958530382397e-05}
{"train_loss": 0.2867288291454315, "global_step": 62599, "epoch": 703, "lr": 8.758920302259825e-05}
{"train_loss": 0.26863402128219604, "global_step": 62600, "epoch": 703, "lr": 8.758882073631912e-05}
{"train_loss": 0.25474002957344055, "global_step": 62601, "epoch": 703, "lr": 8.758843844498663e-05}
{"train_loss": 0.2415885329246521, "global_step": 62602, "epoch": 703, "lr": 8.758805614860084e-05}
{"train_loss": 0.26890701055526733, "global_step": 62603, "epoch": 703, "lr": 8.758767384716177e-05}
{"train_loss": 0.307208389043808, "global_step": 62604, "epoch": 703, "lr": 8.758729154066954e-05}
{"train_loss": 0.30303847789764404, "global_step": 62605, "epoch": 703, "lr": 8.758690922912413e-05}
{"train_loss": 0.22655166685581207, "global_step": 62606, "epoch": 703, "lr": 8.758652691252561e-05}
{"train_loss": 0.21965521574020386, "global_step": 62607, "epoch": 703, "lr": 8.758614459087405e-05}
{"train_loss": 0.35909339785575867, "global_step": 62608, "epoch": 703, "lr": 8.758576226416949e-05}
{"train_loss": 0.2498978227376938, "global_step": 62609, "epoch": 703, "lr": 8.758537993241198e-05}
{"train_loss": 0.24190722405910492, "global_step": 62610, "epoch": 703, "lr": 8.758499759560157e-05}
{"train_loss": 0.30098041892051697, "global_step": 62611, "epoch": 703, "lr": 8.758461525373833e-05}
{"train_loss": 0.2633203864097595, "global_step": 62612, "epoch": 703, "lr": 8.758423290682228e-05}
{"train_loss": 0.2190604954957962, "global_step": 62613, "epoch": 703, "lr": 8.75838505548535e-05}
{"train_loss": 0.2931707203388214, "global_step": 62614, "epoch": 703, "lr": 8.758346819783203e-05}
{"train_loss": 0.2495109885931015, "global_step": 62615, "epoch": 703, "lr": 8.758308583575791e-05}
{"train_loss": 0.3111021816730499, "global_step": 62616, "epoch": 703, "lr": 8.758270346863121e-05}
{"train_loss": 0.29339584708213806, "global_step": 62617, "epoch": 703, "lr": 8.758232109645197e-05}
{"train_loss": 0.22759370505809784, "global_step": 62618, "epoch": 703, "lr": 8.758193871922025e-05}
{"train_loss": 0.2608604431152344, "global_step": 62619, "epoch": 703, "lr": 8.758155633693609e-05}
{"train_loss": 0.2570274770259857, "global_step": 62620, "epoch": 703, "lr": 8.758117394959954e-05}
{"train_loss": 0.23285309970378876, "global_step": 62621, "epoch": 703, "lr": 8.758079155721069e-05}
{"train_loss": 0.22877340018749237, "global_step": 62622, "epoch": 703, "lr": 8.758040915976953e-05}
{"train_loss": 0.2855810821056366, "global_step": 62623, "epoch": 703, "lr": 8.758002675727615e-05}
{"train_loss": 0.21157075464725494, "global_step": 62624, "epoch": 703, "lr": 8.75796443497306e-05}
{"train_loss": 0.27038994431495667, "global_step": 62625, "epoch": 703, "lr": 8.757926193713292e-05}
{"train_loss": 0.23738275468349457, "global_step": 62626, "epoch": 703, "lr": 8.757887951948316e-05}
{"train_loss": 0.33169934153556824, "global_step": 62627, "epoch": 703, "lr": 8.757849709678139e-05}
{"train_loss": 0.262267142534256, "global_step": 62628, "epoch": 703, "lr": 8.757811466902764e-05}
{"train_loss": 0.2348676174879074, "global_step": 62629, "epoch": 703, "lr": 8.757773223622199e-05}
{"train_loss": 0.18789790570735931, "global_step": 62630, "epoch": 703, "lr": 8.757734979836445e-05}
{"train_loss": 0.35473036766052246, "global_step": 62631, "epoch": 703, "lr": 8.75769673554551e-05}
{"train_loss": 0.29973167181015015, "global_step": 62632, "epoch": 703, "lr": 8.757658490749398e-05}
{"train_loss": 0.35500043630599976, "global_step": 62633, "epoch": 703, "lr": 8.757620245448115e-05}
{"train_loss": 0.2531098425388336, "global_step": 62634, "epoch": 703, "lr": 8.757581999641666e-05}
{"train_loss": 0.21939563751220703, "global_step": 62635, "epoch": 703, "lr": 8.757543753330055e-05}
{"train_loss": 0.22622737288475037, "global_step": 62636, "epoch": 703, "lr": 8.757505506513289e-05}
{"train_loss": 0.22856345772743225, "global_step": 62637, "epoch": 703, "lr": 8.757467259191372e-05}
{"train_loss": 0.28657904267311096, "global_step": 62638, "epoch": 703, "lr": 8.757429011364308e-05}
{"train_loss": 0.2004420906305313, "global_step": 62639, "epoch": 703, "lr": 8.757390763032106e-05}
{"train_loss": 0.17344236373901367, "global_step": 62640, "epoch": 703, "lr": 8.757352514194768e-05}
{"train_loss": 0.25613465905189514, "global_step": 62641, "epoch": 703, "lr": 8.757314264852298e-05}
{"train_loss": 0.2877645194530487, "global_step": 62642, "epoch": 703, "lr": 8.757276015004706e-05}
{"train_loss": 0.26292744278907776, "global_step": 62643, "epoch": 703, "lr": 8.757237764651991e-05}
{"train_loss": 0.23658810555934906, "global_step": 62644, "epoch": 703, "lr": 8.75719951379416e-05}
{"train_loss": 0.225126713514328, "global_step": 62645, "epoch": 703, "lr": 8.757161262431223e-05}
{"train_loss": 0.20685596764087677, "global_step": 62646, "epoch": 703, "lr": 8.757123010563179e-05}
{"train_loss": 0.1624729335308075, "global_step": 62647, "epoch": 703, "lr": 8.757084758190037e-05}
{"train_loss": 0.3909105360507965, "global_step": 62648, "epoch": 703, "lr": 8.757046505311801e-05}
{"train_loss": 0.2541868984699249, "global_step": 62649, "epoch": 703, "lr": 8.757008251928476e-05}
{"train_loss": 0.23129260540008545, "global_step": 62650, "epoch": 703, "lr": 8.756969998040066e-05}
{"train_loss": 0.18448160588741302, "global_step": 62651, "epoch": 703, "lr": 8.756931743646577e-05}
{"train_loss": 0.20296531915664673, "global_step": 62652, "epoch": 703, "lr": 8.756893488748015e-05}
{"train_loss": 0.3245043158531189, "global_step": 62653, "epoch": 703, "lr": 8.756855233344383e-05}
{"train_loss": 0.25771015882492065, "global_step": 62654, "epoch": 703, "lr": 8.756816977435688e-05}
{"train_loss": 0.2581907964489433, "global_step": 62655, "epoch": 703, "lr": 8.756778721021937e-05, "val_loss": 3.501230239868164}
{"train_loss": 0.23308688402175903, "global_step": 62656, "epoch": 704, "lr": 8.756740464103132e-05}
{"train_loss": 0.20189060270786285, "global_step": 62657, "epoch": 704, "lr": 8.756702206679278e-05}
{"train_loss": 0.23988567292690277, "global_step": 62658, "epoch": 704, "lr": 8.756663948750381e-05}
{"train_loss": 0.237498939037323, "global_step": 62659, "epoch": 704, "lr": 8.756625690316447e-05}
{"train_loss": 0.2177533209323883, "global_step": 62660, "epoch": 704, "lr": 8.756587431377481e-05}
{"train_loss": 0.2126331478357315, "global_step": 62661, "epoch": 704, "lr": 8.756549171933487e-05}
{"train_loss": 0.20795485377311707, "global_step": 62662, "epoch": 704, "lr": 8.75651091198447e-05}
{"train_loss": 0.18962322175502777, "global_step": 62663, "epoch": 704, "lr": 8.756472651530437e-05}
{"train_loss": 0.19388391077518463, "global_step": 62664, "epoch": 704, "lr": 8.756434390571391e-05}
{"train_loss": 0.26217374205589294, "global_step": 62665, "epoch": 704, "lr": 8.756396129107341e-05}
{"train_loss": 0.2493004947900772, "global_step": 62666, "epoch": 704, "lr": 8.756357867138287e-05}
{"train_loss": 0.19343715906143188, "global_step": 62667, "epoch": 704, "lr": 8.756319604664235e-05}
{"train_loss": 0.36608660221099854, "global_step": 62668, "epoch": 704, "lr": 8.756281341685194e-05}
{"train_loss": 0.3148418366909027, "global_step": 62669, "epoch": 704, "lr": 8.756243078201166e-05}
{"train_loss": 0.2338564395904541, "global_step": 62670, "epoch": 704, "lr": 8.756204814212157e-05}
{"train_loss": 0.2749369442462921, "global_step": 62671, "epoch": 704, "lr": 8.756166549718172e-05}
{"train_loss": 0.1971375048160553, "global_step": 62672, "epoch": 704, "lr": 8.756128284719216e-05}
{"train_loss": 0.16800302267074585, "global_step": 62673, "epoch": 704, "lr": 8.756090019215295e-05}
{"train_loss": 0.21158719062805176, "global_step": 62674, "epoch": 704, "lr": 8.756051753206413e-05}
{"train_loss": 0.25020670890808105, "global_step": 62675, "epoch": 704, "lr": 8.756013486692576e-05}
{"train_loss": 0.24360352754592896, "global_step": 62676, "epoch": 704, "lr": 8.755975219673787e-05}
{"train_loss": 0.16588981449604034, "global_step": 62677, "epoch": 704, "lr": 8.755936952150055e-05}
{"train_loss": 0.18476064503192902, "global_step": 62678, "epoch": 704, "lr": 8.755898684121383e-05}
{"train_loss": 0.19704678654670715, "global_step": 62679, "epoch": 704, "lr": 8.755860415587776e-05}
{"train_loss": 0.2983333468437195, "global_step": 62680, "epoch": 704, "lr": 8.755822146549238e-05}
{"train_loss": 0.19057616591453552, "global_step": 62681, "epoch": 704, "lr": 8.755783877005777e-05}
{"train_loss": 0.22368428111076355, "global_step": 62682, "epoch": 704, "lr": 8.755745606957395e-05}
{"train_loss": 0.1462595909833908, "global_step": 62683, "epoch": 704, "lr": 8.7557073364041e-05}
{"train_loss": 0.23517031967639923, "global_step": 62684, "epoch": 704, "lr": 8.755669065345896e-05}
{"train_loss": 0.26090431213378906, "global_step": 62685, "epoch": 704, "lr": 8.755630793782788e-05}
{"train_loss": 0.27279314398765564, "global_step": 62686, "epoch": 704, "lr": 8.755592521714781e-05}
{"train_loss": 0.26778534054756165, "global_step": 62687, "epoch": 704, "lr": 8.75555424914188e-05}
{"train_loss": 0.21807712316513062, "global_step": 62688, "epoch": 704, "lr": 8.75551597606409e-05}
{"train_loss": 0.2330435812473297, "global_step": 62689, "epoch": 704, "lr": 8.755477702481418e-05}
{"train_loss": 0.20099002122879028, "global_step": 62690, "epoch": 704, "lr": 8.755439428393868e-05}
{"train_loss": 0.14168526232242584, "global_step": 62691, "epoch": 704, "lr": 8.755401153801444e-05}
{"train_loss": 0.2725549340248108, "global_step": 62692, "epoch": 704, "lr": 8.755362878704153e-05}
{"train_loss": 0.269719660282135, "global_step": 62693, "epoch": 704, "lr": 8.755324603101997e-05}
{"train_loss": 0.29183462262153625, "global_step": 62694, "epoch": 704, "lr": 8.755286326994985e-05}
{"train_loss": 0.27557918429374695, "global_step": 62695, "epoch": 704, "lr": 8.75524805038312e-05}
{"train_loss": 0.31572771072387695, "global_step": 62696, "epoch": 704, "lr": 8.755209773266409e-05}
{"train_loss": 0.24746356904506683, "global_step": 62697, "epoch": 704, "lr": 8.755171495644855e-05}
{"train_loss": 0.1897188127040863, "global_step": 62698, "epoch": 704, "lr": 8.755133217518465e-05}
{"train_loss": 0.25738921761512756, "global_step": 62699, "epoch": 704, "lr": 8.755094938887241e-05}
{"train_loss": 0.13938723504543304, "global_step": 62700, "epoch": 704, "lr": 8.755056659751191e-05}
{"train_loss": 0.22747883200645447, "global_step": 62701, "epoch": 704, "lr": 8.75501838011032e-05}
{"train_loss": 0.2555621862411499, "global_step": 62702, "epoch": 704, "lr": 8.754980099964632e-05}
{"train_loss": 0.2929913401603699, "global_step": 62703, "epoch": 704, "lr": 8.754941819314133e-05}
{"train_loss": 0.2644079029560089, "global_step": 62704, "epoch": 704, "lr": 8.754903538158827e-05}
{"train_loss": 0.28279435634613037, "global_step": 62705, "epoch": 704, "lr": 8.75486525649872e-05}
{"train_loss": 0.21166324615478516, "global_step": 62706, "epoch": 704, "lr": 8.754826974333819e-05}
{"train_loss": 0.25966939330101013, "global_step": 62707, "epoch": 704, "lr": 8.754788691664127e-05}
{"train_loss": 0.17007538676261902, "global_step": 62708, "epoch": 704, "lr": 8.754750408489647e-05}
{"train_loss": 0.2744568884372711, "global_step": 62709, "epoch": 704, "lr": 8.754712124810389e-05}
{"train_loss": 0.24978990852832794, "global_step": 62710, "epoch": 704, "lr": 8.754673840626355e-05}
{"train_loss": 0.2901706099510193, "global_step": 62711, "epoch": 704, "lr": 8.754635555937551e-05}
{"train_loss": 0.21968722343444824, "global_step": 62712, "epoch": 704, "lr": 8.754597270743981e-05}
{"train_loss": 0.2275543361902237, "global_step": 62713, "epoch": 704, "lr": 8.754558985045653e-05}
{"train_loss": 0.1447514444589615, "global_step": 62714, "epoch": 704, "lr": 8.754520698842568e-05}
{"train_loss": 0.28372958302497864, "global_step": 62715, "epoch": 704, "lr": 8.754482412134735e-05}
{"train_loss": 0.2899119555950165, "global_step": 62716, "epoch": 704, "lr": 8.754444124922157e-05}
{"train_loss": 0.16917604207992554, "global_step": 62717, "epoch": 704, "lr": 8.754405837204839e-05}
{"train_loss": 0.22506402432918549, "global_step": 62718, "epoch": 704, "lr": 8.754367548982789e-05}
{"train_loss": 0.31264886260032654, "global_step": 62719, "epoch": 704, "lr": 8.754329260256009e-05}
{"train_loss": 0.2410549372434616, "global_step": 62720, "epoch": 704, "lr": 8.754290971024506e-05}
{"train_loss": 0.26943331956863403, "global_step": 62721, "epoch": 704, "lr": 8.754252681288282e-05}
{"train_loss": 0.2694756090641022, "global_step": 62722, "epoch": 704, "lr": 8.754214391047346e-05}
{"train_loss": 0.28385451436042786, "global_step": 62723, "epoch": 704, "lr": 8.754176100301703e-05}
{"train_loss": 0.20281162858009338, "global_step": 62724, "epoch": 704, "lr": 8.754137809051355e-05}
{"train_loss": 0.24692721664905548, "global_step": 62725, "epoch": 704, "lr": 8.75409951729631e-05}
{"train_loss": 0.2378024160861969, "global_step": 62726, "epoch": 704, "lr": 8.75406122503657e-05}
{"train_loss": 0.23814688622951508, "global_step": 62727, "epoch": 704, "lr": 8.754022932272144e-05}
{"train_loss": 0.2330009788274765, "global_step": 62728, "epoch": 704, "lr": 8.753984639003034e-05}
{"train_loss": 0.22472652792930603, "global_step": 62729, "epoch": 704, "lr": 8.75394634522925e-05}
{"train_loss": 0.25338107347488403, "global_step": 62730, "epoch": 704, "lr": 8.75390805095079e-05}
{"train_loss": 0.35694417357444763, "global_step": 62731, "epoch": 704, "lr": 8.753869756167664e-05}
{"train_loss": 0.21070843935012817, "global_step": 62732, "epoch": 704, "lr": 8.753831460879876e-05}
{"train_loss": 0.20807935297489166, "global_step": 62733, "epoch": 704, "lr": 8.753793165087431e-05}
{"train_loss": 0.21232274174690247, "global_step": 62734, "epoch": 704, "lr": 8.753754868790334e-05}
{"train_loss": 0.23459799587726593, "global_step": 62735, "epoch": 704, "lr": 8.75371657198859e-05}
{"train_loss": 0.19444291293621063, "global_step": 62736, "epoch": 704, "lr": 8.753678274682205e-05}
{"train_loss": 0.2577413022518158, "global_step": 62737, "epoch": 704, "lr": 8.753639976871184e-05}
{"train_loss": 0.3276751935482025, "global_step": 62738, "epoch": 704, "lr": 8.753601678555532e-05}
{"train_loss": 0.21816854178905487, "global_step": 62739, "epoch": 704, "lr": 8.753563379735254e-05}
{"train_loss": 0.15679189562797546, "global_step": 62740, "epoch": 704, "lr": 8.753525080410357e-05}
{"train_loss": 0.21564124524593353, "global_step": 62741, "epoch": 704, "lr": 8.753486780580842e-05}
{"train_loss": 0.22703763842582703, "global_step": 62742, "epoch": 704, "lr": 8.753448480246715e-05}
{"train_loss": 0.22597740590572357, "global_step": 62743, "epoch": 704, "lr": 8.753410179407986e-05}
{"train_loss": 0.23579144477844238, "global_step": 62744, "epoch": 704, "lr": 8.753371878064654e-05, "val_loss": 3.527643918991089}
{"train_loss": 0.23154567182064056, "global_step": 62745, "epoch": 705, "lr": 8.753333576216727e-05}
{"train_loss": 0.24091270565986633, "global_step": 62746, "epoch": 705, "lr": 8.753295273864213e-05}
{"train_loss": 0.22257333993911743, "global_step": 62747, "epoch": 705, "lr": 8.753256971007111e-05}
{"train_loss": 0.25600534677505493, "global_step": 62748, "epoch": 705, "lr": 8.753218667645431e-05}
{"train_loss": 0.19288207590579987, "global_step": 62749, "epoch": 705, "lr": 8.753180363779177e-05}
{"train_loss": 0.2139575481414795, "global_step": 62750, "epoch": 705, "lr": 8.753142059408353e-05}
{"train_loss": 0.24925008416175842, "global_step": 62751, "epoch": 705, "lr": 8.753103754532962e-05}
{"train_loss": 0.19910971820354462, "global_step": 62752, "epoch": 705, "lr": 8.753065449153016e-05}
{"train_loss": 0.20491614937782288, "global_step": 62753, "epoch": 705, "lr": 8.753027143268514e-05}
{"train_loss": 0.319851279258728, "global_step": 62754, "epoch": 705, "lr": 8.752988836879463e-05}
{"train_loss": 0.22739413380622864, "global_step": 62755, "epoch": 705, "lr": 8.75295052998587e-05}
{"train_loss": 0.19177033007144928, "global_step": 62756, "epoch": 705, "lr": 8.752912222587738e-05}
{"train_loss": 0.2214360535144806, "global_step": 62757, "epoch": 705, "lr": 8.752873914685072e-05}
{"train_loss": 0.17970456182956696, "global_step": 62758, "epoch": 705, "lr": 8.752835606277878e-05}
{"train_loss": 0.1996089518070221, "global_step": 62759, "epoch": 705, "lr": 8.752797297366162e-05}
{"train_loss": 0.30413711071014404, "global_step": 62760, "epoch": 705, "lr": 8.752758987949927e-05}
{"train_loss": 0.2402890920639038, "global_step": 62761, "epoch": 705, "lr": 8.75272067802918e-05}
{"train_loss": 0.20807337760925293, "global_step": 62762, "epoch": 705, "lr": 8.752682367603926e-05}
{"train_loss": 0.28669631481170654, "global_step": 62763, "epoch": 705, "lr": 8.752644056674169e-05}
{"train_loss": 0.22028842568397522, "global_step": 62764, "epoch": 705, "lr": 8.752605745239914e-05}
{"train_loss": 0.24657370150089264, "global_step": 62765, "epoch": 705, "lr": 8.752567433301167e-05}
{"train_loss": 0.16908273100852966, "global_step": 62766, "epoch": 705, "lr": 8.752529120857933e-05}
{"train_loss": 0.15826672315597534, "global_step": 62767, "epoch": 705, "lr": 8.752490807910219e-05}
{"train_loss": 0.25722163915634155, "global_step": 62768, "epoch": 705, "lr": 8.752452494458026e-05}
{"train_loss": 0.18563538789749146, "global_step": 62769, "epoch": 705, "lr": 8.752414180501362e-05}
{"train_loss": 0.18351931869983673, "global_step": 62770, "epoch": 705, "lr": 8.752375866040232e-05}
{"train_loss": 0.2752669155597687, "global_step": 62771, "epoch": 705, "lr": 8.752337551074641e-05}
{"train_loss": 0.25100886821746826, "global_step": 62772, "epoch": 705, "lr": 8.752299235604594e-05}
{"train_loss": 0.27852940559387207, "global_step": 62773, "epoch": 705, "lr": 8.752260919630097e-05}
{"train_loss": 0.19196608662605286, "global_step": 62774, "epoch": 705, "lr": 8.752222603151152e-05}
{"train_loss": 0.25465136766433716, "global_step": 62775, "epoch": 705, "lr": 8.752184286167769e-05}
{"train_loss": 0.23982687294483185, "global_step": 62776, "epoch": 705, "lr": 8.752145968679948e-05}
{"train_loss": 0.20481210947036743, "global_step": 62777, "epoch": 705, "lr": 8.752107650687698e-05}
{"train_loss": 0.22205278277397156, "global_step": 62778, "epoch": 705, "lr": 8.752069332191023e-05}
{"train_loss": 0.24016617238521576, "global_step": 62779, "epoch": 705, "lr": 8.752031013189928e-05}
{"train_loss": 0.19399848580360413, "global_step": 62780, "epoch": 705, "lr": 8.751992693684418e-05}
{"train_loss": 0.24647869169712067, "global_step": 62781, "epoch": 705, "lr": 8.751954373674497e-05}
{"train_loss": 0.23481206595897675, "global_step": 62782, "epoch": 705, "lr": 8.751916053160174e-05}
{"train_loss": 0.1435692459344864, "global_step": 62783, "epoch": 705, "lr": 8.75187773214145e-05}
{"train_loss": 0.3236719071865082, "global_step": 62784, "epoch": 705, "lr": 8.751839410618333e-05}
{"train_loss": 0.2055288404226303, "global_step": 62785, "epoch": 705, "lr": 8.751801088590825e-05}
{"train_loss": 0.27342721819877625, "global_step": 62786, "epoch": 705, "lr": 8.751762766058935e-05}
{"train_loss": 0.29946863651275635, "global_step": 62787, "epoch": 705, "lr": 8.751724443022664e-05}
{"train_loss": 0.20018260180950165, "global_step": 62788, "epoch": 705, "lr": 8.751686119482022e-05}
{"train_loss": 0.2095676213502884, "global_step": 62789, "epoch": 705, "lr": 8.75164779543701e-05}
{"train_loss": 0.2579217255115509, "global_step": 62790, "epoch": 705, "lr": 8.751609470887634e-05}
{"train_loss": 0.2195625603199005, "global_step": 62791, "epoch": 705, "lr": 8.751571145833901e-05}
{"train_loss": 0.1545252948999405, "global_step": 62792, "epoch": 705, "lr": 8.751532820275813e-05}
{"train_loss": 0.2184654176235199, "global_step": 62793, "epoch": 705, "lr": 8.751494494213381e-05}
{"train_loss": 0.3018988072872162, "global_step": 62794, "epoch": 705, "lr": 8.751456167646603e-05}
{"train_loss": 0.2802039682865143, "global_step": 62795, "epoch": 705, "lr": 8.75141784057549e-05}
{"train_loss": 0.2423902004957199, "global_step": 62796, "epoch": 705, "lr": 8.751379513000042e-05}
{"train_loss": 0.15892021358013153, "global_step": 62797, "epoch": 705, "lr": 8.751341184920269e-05}
{"train_loss": 0.1524641215801239, "global_step": 62798, "epoch": 705, "lr": 8.751302856336172e-05}
{"train_loss": 0.21335187554359436, "global_step": 62799, "epoch": 705, "lr": 8.751264527247759e-05}
{"train_loss": 0.20010066032409668, "global_step": 62800, "epoch": 705, "lr": 8.751226197655034e-05}
{"train_loss": 0.2657192647457123, "global_step": 62801, "epoch": 705, "lr": 8.751187867558002e-05}
{"train_loss": 0.1538090854883194, "global_step": 62802, "epoch": 705, "lr": 8.751149536956669e-05}
{"train_loss": 0.26433274149894714, "global_step": 62803, "epoch": 705, "lr": 8.751111205851039e-05}
{"train_loss": 0.20239265263080597, "global_step": 62804, "epoch": 705, "lr": 8.751072874241119e-05}
{"train_loss": 0.24949610233306885, "global_step": 62805, "epoch": 705, "lr": 8.751034542126912e-05}
{"train_loss": 0.21932682394981384, "global_step": 62806, "epoch": 705, "lr": 8.750996209508427e-05}
{"train_loss": 0.28228604793548584, "global_step": 62807, "epoch": 705, "lr": 8.750957876385663e-05}
{"train_loss": 0.2348739206790924, "global_step": 62808, "epoch": 705, "lr": 8.750919542758628e-05}
{"train_loss": 0.2873934507369995, "global_step": 62809, "epoch": 705, "lr": 8.750881208627328e-05}
{"train_loss": 0.21203380823135376, "global_step": 62810, "epoch": 705, "lr": 8.750842873991769e-05}
{"train_loss": 0.22844728827476501, "global_step": 62811, "epoch": 705, "lr": 8.750804538851955e-05}
{"train_loss": 0.2046477496623993, "global_step": 62812, "epoch": 705, "lr": 8.750766203207891e-05}
{"train_loss": 0.2188764363527298, "global_step": 62813, "epoch": 705, "lr": 8.750727867059581e-05}
{"train_loss": 0.17416752874851227, "global_step": 62814, "epoch": 705, "lr": 8.750689530407033e-05}
{"train_loss": 0.22597277164459229, "global_step": 62815, "epoch": 705, "lr": 8.750651193250249e-05}
{"train_loss": 0.2055792361497879, "global_step": 62816, "epoch": 705, "lr": 8.750612855589237e-05}
{"train_loss": 0.23355591297149658, "global_step": 62817, "epoch": 705, "lr": 8.750574517423999e-05}
{"train_loss": 0.2724006474018097, "global_step": 62818, "epoch": 705, "lr": 8.750536178754542e-05}
{"train_loss": 0.17102566361427307, "global_step": 62819, "epoch": 705, "lr": 8.750497839580873e-05}
{"train_loss": 0.19133034348487854, "global_step": 62820, "epoch": 705, "lr": 8.750459499902994e-05}
{"train_loss": 0.21461725234985352, "global_step": 62821, "epoch": 705, "lr": 8.750421159720912e-05}
{"train_loss": 0.17493045330047607, "global_step": 62822, "epoch": 705, "lr": 8.750382819034633e-05}
{"train_loss": 0.21705210208892822, "global_step": 62823, "epoch": 705, "lr": 8.750344477844158e-05}
{"train_loss": 0.2951142191886902, "global_step": 62824, "epoch": 705, "lr": 8.750306136149495e-05}
{"train_loss": 0.21834813058376312, "global_step": 62825, "epoch": 705, "lr": 8.750267793950651e-05}
{"train_loss": 0.19952142238616943, "global_step": 62826, "epoch": 705, "lr": 8.750229451247629e-05}
{"train_loss": 0.24900394678115845, "global_step": 62827, "epoch": 705, "lr": 8.750191108040433e-05}
{"train_loss": 0.19325217604637146, "global_step": 62828, "epoch": 705, "lr": 8.75015276432907e-05}
{"train_loss": 0.1237349882721901, "global_step": 62829, "epoch": 705, "lr": 8.750114420113545e-05}
{"train_loss": 0.25860488414764404, "global_step": 62830, "epoch": 705, "lr": 8.750076075393863e-05}
{"train_loss": 0.23136509954929352, "global_step": 62831, "epoch": 705, "lr": 8.750037730170029e-05}
{"train_loss": 0.2685125172138214, "global_step": 62832, "epoch": 705, "lr": 8.749999384442047e-05}
{"train_loss": 0.22591556114762018, "global_step": 62833, "epoch": 705, "lr": 8.749961038209924e-05, "val_loss": 3.5021414756774902, "train_action_mse_error": 29.99502182006836}
{"train_loss": 0.23592303693294525, "global_step": 62834, "epoch": 706, "lr": 8.749922691473664e-05}
{"train_loss": 0.15170589089393616, "global_step": 62835, "epoch": 706, "lr": 8.749884344233275e-05}
{"train_loss": 0.19814497232437134, "global_step": 62836, "epoch": 706, "lr": 8.749845996488757e-05}
{"train_loss": 0.18408341705799103, "global_step": 62837, "epoch": 706, "lr": 8.749807648240118e-05}
{"train_loss": 0.2805659770965576, "global_step": 62838, "epoch": 706, "lr": 8.749769299487365e-05}
{"train_loss": 0.12612780928611755, "global_step": 62839, "epoch": 706, "lr": 8.749730950230499e-05}
{"train_loss": 0.15569019317626953, "global_step": 62840, "epoch": 706, "lr": 8.749692600469529e-05}
{"train_loss": 0.25645679235458374, "global_step": 62841, "epoch": 706, "lr": 8.749654250204458e-05}
{"train_loss": 0.18970021605491638, "global_step": 62842, "epoch": 706, "lr": 8.749615899435291e-05}
{"train_loss": 0.2261127531528473, "global_step": 62843, "epoch": 706, "lr": 8.749577548162035e-05}
{"train_loss": 0.20060934126377106, "global_step": 62844, "epoch": 706, "lr": 8.749539196384693e-05}
{"train_loss": 0.2032439261674881, "global_step": 62845, "epoch": 706, "lr": 8.749500844103272e-05}
{"train_loss": 0.2263764888048172, "global_step": 62846, "epoch": 706, "lr": 8.749462491317776e-05}
{"train_loss": 0.2107972353696823, "global_step": 62847, "epoch": 706, "lr": 8.749424138028211e-05}
{"train_loss": 0.20514316856861115, "global_step": 62848, "epoch": 706, "lr": 8.749385784234581e-05}
{"train_loss": 0.26840540766716003, "global_step": 62849, "epoch": 706, "lr": 8.749347429936892e-05}
{"train_loss": 0.2440520077943802, "global_step": 62850, "epoch": 706, "lr": 8.749309075135149e-05}
{"train_loss": 0.17870444059371948, "global_step": 62851, "epoch": 706, "lr": 8.749270719829356e-05}
{"train_loss": 0.22432945668697357, "global_step": 62852, "epoch": 706, "lr": 8.74923236401952e-05}
{"train_loss": 0.2738788425922394, "global_step": 62853, "epoch": 706, "lr": 8.749194007705646e-05}
{"train_loss": 0.36532068252563477, "global_step": 62854, "epoch": 706, "lr": 8.749155650887739e-05}
{"train_loss": 0.21088631451129913, "global_step": 62855, "epoch": 706, "lr": 8.749117293565801e-05}
{"train_loss": 0.24519619345664978, "global_step": 62856, "epoch": 706, "lr": 8.749078935739843e-05}
{"train_loss": 0.24530115723609924, "global_step": 62857, "epoch": 706, "lr": 8.749040577409867e-05}
{"train_loss": 0.34538549184799194, "global_step": 62858, "epoch": 706, "lr": 8.749002218575874e-05}
{"train_loss": 0.2829146981239319, "global_step": 62859, "epoch": 706, "lr": 8.748963859237878e-05}
{"train_loss": 0.2737506926059723, "global_step": 62860, "epoch": 706, "lr": 8.748925499395877e-05}
{"train_loss": 0.2027348130941391, "global_step": 62861, "epoch": 706, "lr": 8.748887139049879e-05}
{"train_loss": 0.2322886735200882, "global_step": 62862, "epoch": 706, "lr": 8.74884877819989e-05}
{"train_loss": 0.171838641166687, "global_step": 62863, "epoch": 706, "lr": 8.748810416845912e-05}
{"train_loss": 0.2794881761074066, "global_step": 62864, "epoch": 706, "lr": 8.748772054987955e-05}
{"train_loss": 0.2672548294067383, "global_step": 62865, "epoch": 706, "lr": 8.748733692626019e-05}
{"train_loss": 0.2622101902961731, "global_step": 62866, "epoch": 706, "lr": 8.748695329760113e-05}
{"train_loss": 0.2479977309703827, "global_step": 62867, "epoch": 706, "lr": 8.748656966390239e-05}
{"train_loss": 0.2542422413825989, "global_step": 62868, "epoch": 706, "lr": 8.748618602516405e-05}
{"train_loss": 0.197999507188797, "global_step": 62869, "epoch": 706, "lr": 8.748580238138613e-05}
{"train_loss": 0.16337835788726807, "global_step": 62870, "epoch": 706, "lr": 8.748541873256873e-05}
{"train_loss": 0.2884858250617981, "global_step": 62871, "epoch": 706, "lr": 8.748503507871185e-05}
{"train_loss": 0.29735320806503296, "global_step": 62872, "epoch": 706, "lr": 8.748465141981557e-05}
{"train_loss": 0.30342164635658264, "global_step": 62873, "epoch": 706, "lr": 8.748426775587993e-05}
{"train_loss": 0.19090433418750763, "global_step": 62874, "epoch": 706, "lr": 8.7483884086905e-05}
{"train_loss": 0.212476447224617, "global_step": 62875, "epoch": 706, "lr": 8.74835004128908e-05}
{"train_loss": 0.17784814536571503, "global_step": 62876, "epoch": 706, "lr": 8.748311673383743e-05}
{"train_loss": 0.21785713732242584, "global_step": 62877, "epoch": 706, "lr": 8.748273304974487e-05}
{"train_loss": 0.22744716703891754, "global_step": 62878, "epoch": 706, "lr": 8.748234936061326e-05}
{"train_loss": 0.3254409432411194, "global_step": 62879, "epoch": 706, "lr": 8.748196566644257e-05}
{"train_loss": 0.33576327562332153, "global_step": 62880, "epoch": 706, "lr": 8.748158196723289e-05}
{"train_loss": 0.2185339331626892, "global_step": 62881, "epoch": 706, "lr": 8.748119826298429e-05}
{"train_loss": 0.2155529409646988, "global_step": 62882, "epoch": 706, "lr": 8.748081455369679e-05}
{"train_loss": 0.2363090217113495, "global_step": 62883, "epoch": 706, "lr": 8.748043083937044e-05}
{"train_loss": 0.3425144851207733, "global_step": 62884, "epoch": 706, "lr": 8.748004712000531e-05}
{"train_loss": 0.21846862137317657, "global_step": 62885, "epoch": 706, "lr": 8.747966339560145e-05}
{"train_loss": 0.2661559581756592, "global_step": 62886, "epoch": 706, "lr": 8.74792796661589e-05}
{"train_loss": 0.21899116039276123, "global_step": 62887, "epoch": 706, "lr": 8.747889593167772e-05}
{"train_loss": 0.26574409008026123, "global_step": 62888, "epoch": 706, "lr": 8.747851219215795e-05}
{"train_loss": 0.2563636302947998, "global_step": 62889, "epoch": 706, "lr": 8.747812844759966e-05}
{"train_loss": 0.19660811126232147, "global_step": 62890, "epoch": 706, "lr": 8.747774469800289e-05}
{"train_loss": 0.25164908170700073, "global_step": 62891, "epoch": 706, "lr": 8.74773609433677e-05}
{"train_loss": 0.17690864205360413, "global_step": 62892, "epoch": 706, "lr": 8.747697718369412e-05}
{"train_loss": 0.20806556940078735, "global_step": 62893, "epoch": 706, "lr": 8.747659341898225e-05}
{"train_loss": 0.1713920384645462, "global_step": 62894, "epoch": 706, "lr": 8.747620964923207e-05}
{"train_loss": 0.22109299898147583, "global_step": 62895, "epoch": 706, "lr": 8.74758258744437e-05}
{"train_loss": 0.23006625473499298, "global_step": 62896, "epoch": 706, "lr": 8.747544209461714e-05}
{"train_loss": 0.3151913285255432, "global_step": 62897, "epoch": 706, "lr": 8.747505830975248e-05}
{"train_loss": 0.20053526759147644, "global_step": 62898, "epoch": 706, "lr": 8.747467451984975e-05}
{"train_loss": 0.22073931992053986, "global_step": 62899, "epoch": 706, "lr": 8.7474290724909e-05}
{"train_loss": 0.24767045676708221, "global_step": 62900, "epoch": 706, "lr": 8.747390692493029e-05}
{"train_loss": 0.26332342624664307, "global_step": 62901, "epoch": 706, "lr": 8.747352311991369e-05}
{"train_loss": 0.2582770884037018, "global_step": 62902, "epoch": 706, "lr": 8.747313930985922e-05}
{"train_loss": 0.228457510471344, "global_step": 62903, "epoch": 706, "lr": 8.747275549476692e-05}
{"train_loss": 0.2136135846376419, "global_step": 62904, "epoch": 706, "lr": 8.747237167463689e-05}
{"train_loss": 0.2752808630466461, "global_step": 62905, "epoch": 706, "lr": 8.747198784946915e-05}
{"train_loss": 0.19794045388698578, "global_step": 62906, "epoch": 706, "lr": 8.747160401926375e-05}
{"train_loss": 0.25532469153404236, "global_step": 62907, "epoch": 706, "lr": 8.747122018402076e-05}
{"train_loss": 0.1304538995027542, "global_step": 62908, "epoch": 706, "lr": 8.747083634374021e-05}
{"train_loss": 0.21515002846717834, "global_step": 62909, "epoch": 706, "lr": 8.747045249842217e-05}
{"train_loss": 0.3015661835670471, "global_step": 62910, "epoch": 706, "lr": 8.747006864806668e-05}
{"train_loss": 0.23975160717964172, "global_step": 62911, "epoch": 706, "lr": 8.746968479267382e-05}
{"train_loss": 0.2581698000431061, "global_step": 62912, "epoch": 706, "lr": 8.746930093224359e-05}
{"train_loss": 0.2376810908317566, "global_step": 62913, "epoch": 706, "lr": 8.746891706677608e-05}
{"train_loss": 0.2343713492155075, "global_step": 62914, "epoch": 706, "lr": 8.746853319627133e-05}
{"train_loss": 0.26266857981681824, "global_step": 62915, "epoch": 706, "lr": 8.746814932072939e-05}
{"train_loss": 0.2894521653652191, "global_step": 62916, "epoch": 706, "lr": 8.746776544015032e-05}
{"train_loss": 0.23702207207679749, "global_step": 62917, "epoch": 706, "lr": 8.746738155453415e-05}
{"train_loss": 0.2135199010372162, "global_step": 62918, "epoch": 706, "lr": 8.746699766388096e-05}
{"train_loss": 0.20515196025371552, "global_step": 62919, "epoch": 706, "lr": 8.746661376819079e-05}
{"train_loss": 0.2921772301197052, "global_step": 62920, "epoch": 706, "lr": 8.746622986746369e-05}
{"train_loss": 0.2158154994249344, "global_step": 62921, "epoch": 706, "lr": 8.74658459616997e-05}
{"train_loss": 0.23614656858229904, "global_step": 62922, "epoch": 706, "lr": 8.746546205089889e-05, "val_loss": 3.460055112838745}
{"train_loss": 0.15692782402038574, "global_step": 62923, "epoch": 707, "lr": 8.74650781350613e-05}
{"train_loss": 0.32440584897994995, "global_step": 62924, "epoch": 707, "lr": 8.7464694214187e-05}
{"train_loss": 0.26067864894866943, "global_step": 62925, "epoch": 707, "lr": 8.746431028827603e-05}
{"train_loss": 0.20186051726341248, "global_step": 62926, "epoch": 707, "lr": 8.746392635732842e-05}
{"train_loss": 0.21418333053588867, "global_step": 62927, "epoch": 707, "lr": 8.746354242134426e-05}
{"train_loss": 0.1671883910894394, "global_step": 62928, "epoch": 707, "lr": 8.746315848032358e-05}
{"train_loss": 0.20132896304130554, "global_step": 62929, "epoch": 707, "lr": 8.746277453426643e-05}
{"train_loss": 0.33122318983078003, "global_step": 62930, "epoch": 707, "lr": 8.746239058317287e-05}
{"train_loss": 0.23833705484867096, "global_step": 62931, "epoch": 707, "lr": 8.746200662704296e-05}
{"train_loss": 0.206047922372818, "global_step": 62932, "epoch": 707, "lr": 8.746162266587671e-05}
{"train_loss": 0.2558826804161072, "global_step": 62933, "epoch": 707, "lr": 8.746123869967424e-05}
{"train_loss": 0.24075400829315186, "global_step": 62934, "epoch": 707, "lr": 8.746085472843554e-05}
{"train_loss": 0.21289905905723572, "global_step": 62935, "epoch": 707, "lr": 8.746047075216069e-05}
{"train_loss": 0.20854029059410095, "global_step": 62936, "epoch": 707, "lr": 8.746008677084972e-05}
{"train_loss": 0.18245606124401093, "global_step": 62937, "epoch": 707, "lr": 8.74597027845027e-05}
{"train_loss": 0.14750009775161743, "global_step": 62938, "epoch": 707, "lr": 8.74593187931197e-05}
{"train_loss": 0.21957191824913025, "global_step": 62939, "epoch": 707, "lr": 8.745893479670076e-05}
{"train_loss": 0.13914155960083008, "global_step": 62940, "epoch": 707, "lr": 8.74585507952459e-05}
{"train_loss": 0.1883571296930313, "global_step": 62941, "epoch": 707, "lr": 8.745816678875519e-05}
{"train_loss": 0.1989128291606903, "global_step": 62942, "epoch": 707, "lr": 8.74577827772287e-05}
{"train_loss": 0.26822736859321594, "global_step": 62943, "epoch": 707, "lr": 8.745739876066645e-05}
{"train_loss": 0.24113833904266357, "global_step": 62944, "epoch": 707, "lr": 8.745701473906853e-05}
{"train_loss": 0.22688187658786774, "global_step": 62945, "epoch": 707, "lr": 8.745663071243497e-05}
{"train_loss": 0.21737121045589447, "global_step": 62946, "epoch": 707, "lr": 8.74562466807658e-05}
{"train_loss": 0.2934379577636719, "global_step": 62947, "epoch": 707, "lr": 8.745586264406112e-05}
{"train_loss": 0.2732193171977997, "global_step": 62948, "epoch": 707, "lr": 8.745547860232095e-05}
{"train_loss": 0.19523753225803375, "global_step": 62949, "epoch": 707, "lr": 8.745509455554534e-05}
{"train_loss": 0.23702698945999146, "global_step": 62950, "epoch": 707, "lr": 8.745471050373434e-05}
{"train_loss": 0.2276168316602707, "global_step": 62951, "epoch": 707, "lr": 8.745432644688803e-05}
{"train_loss": 0.2998791038990021, "global_step": 62952, "epoch": 707, "lr": 8.745394238500643e-05}
{"train_loss": 0.3371250629425049, "global_step": 62953, "epoch": 707, "lr": 8.745355831808961e-05}
{"train_loss": 0.3016907870769501, "global_step": 62954, "epoch": 707, "lr": 8.745317424613761e-05}
{"train_loss": 0.16788817942142487, "global_step": 62955, "epoch": 707, "lr": 8.74527901691505e-05}
{"train_loss": 0.19408217072486877, "global_step": 62956, "epoch": 707, "lr": 8.745240608712831e-05}
{"train_loss": 0.3274962902069092, "global_step": 62957, "epoch": 707, "lr": 8.745202200007109e-05}
{"train_loss": 0.2437296211719513, "global_step": 62958, "epoch": 707, "lr": 8.745163790797892e-05}
{"train_loss": 0.25556570291519165, "global_step": 62959, "epoch": 707, "lr": 8.745125381085184e-05}
{"train_loss": 0.2125859558582306, "global_step": 62960, "epoch": 707, "lr": 8.745086970868987e-05}
{"train_loss": 0.22665975987911224, "global_step": 62961, "epoch": 707, "lr": 8.745048560149311e-05}
{"train_loss": 0.2503627836704254, "global_step": 62962, "epoch": 707, "lr": 8.745010148926157e-05}
{"train_loss": 0.2810242474079132, "global_step": 62963, "epoch": 707, "lr": 8.744971737199533e-05}
{"train_loss": 0.23392513394355774, "global_step": 62964, "epoch": 707, "lr": 8.744933324969442e-05}
{"train_loss": 0.20730672776699066, "global_step": 62965, "epoch": 707, "lr": 8.744894912235892e-05}
{"train_loss": 0.22284001111984253, "global_step": 62966, "epoch": 707, "lr": 8.744856498998886e-05}
{"train_loss": 0.26372626423835754, "global_step": 62967, "epoch": 707, "lr": 8.74481808525843e-05}
{"train_loss": 0.20758779346942902, "global_step": 62968, "epoch": 707, "lr": 8.744779671014529e-05}
{"train_loss": 0.283449649810791, "global_step": 62969, "epoch": 707, "lr": 8.744741256267187e-05}
{"train_loss": 0.22475753724575043, "global_step": 62970, "epoch": 707, "lr": 8.74470284101641e-05}
{"train_loss": 0.2354874461889267, "global_step": 62971, "epoch": 707, "lr": 8.744664425262204e-05}
{"train_loss": 0.3231621980667114, "global_step": 62972, "epoch": 707, "lr": 8.744626009004574e-05}
{"train_loss": 0.2125369906425476, "global_step": 62973, "epoch": 707, "lr": 8.744587592243524e-05}
{"train_loss": 0.3205612301826477, "global_step": 62974, "epoch": 707, "lr": 8.74454917497906e-05}
{"train_loss": 0.3145937919616699, "global_step": 62975, "epoch": 707, "lr": 8.744510757211186e-05}
{"train_loss": 0.30214911699295044, "global_step": 62976, "epoch": 707, "lr": 8.744472338939911e-05}
{"train_loss": 0.2891536056995392, "global_step": 62977, "epoch": 707, "lr": 8.744433920165235e-05}
{"train_loss": 0.18679893016815186, "global_step": 62978, "epoch": 707, "lr": 8.744395500887166e-05}
{"train_loss": 0.22731664776802063, "global_step": 62979, "epoch": 707, "lr": 8.744357081105708e-05}
{"train_loss": 0.17946530878543854, "global_step": 62980, "epoch": 707, "lr": 8.744318660820867e-05}
{"train_loss": 0.23411592841148376, "global_step": 62981, "epoch": 707, "lr": 8.74428024003265e-05}
{"train_loss": 0.2020615041255951, "global_step": 62982, "epoch": 707, "lr": 8.744241818741058e-05}
{"train_loss": 0.3016892075538635, "global_step": 62983, "epoch": 707, "lr": 8.744203396946099e-05}
{"train_loss": 0.3072899281978607, "global_step": 62984, "epoch": 707, "lr": 8.744164974647778e-05}
{"train_loss": 0.20391249656677246, "global_step": 62985, "epoch": 707, "lr": 8.7441265518461e-05}
{"train_loss": 0.26519474387168884, "global_step": 62986, "epoch": 707, "lr": 8.744088128541067e-05}
{"train_loss": 0.2861354947090149, "global_step": 62987, "epoch": 707, "lr": 8.74404970473269e-05}
{"train_loss": 0.3094574511051178, "global_step": 62988, "epoch": 707, "lr": 8.74401128042097e-05}
{"train_loss": 0.2197248488664627, "global_step": 62989, "epoch": 707, "lr": 8.743972855605912e-05}
{"train_loss": 0.20343321561813354, "global_step": 62990, "epoch": 707, "lr": 8.743934430287523e-05}
{"train_loss": 0.21909435093402863, "global_step": 62991, "epoch": 707, "lr": 8.74389600446581e-05}
{"train_loss": 0.20178602635860443, "global_step": 62992, "epoch": 707, "lr": 8.743857578140775e-05}
{"train_loss": 0.2028178870677948, "global_step": 62993, "epoch": 707, "lr": 8.743819151312422e-05}
{"train_loss": 0.26244834065437317, "global_step": 62994, "epoch": 707, "lr": 8.74378072398076e-05}
{"train_loss": 0.2201923429965973, "global_step": 62995, "epoch": 707, "lr": 8.743742296145791e-05}
{"train_loss": 0.2622469365596771, "global_step": 62996, "epoch": 707, "lr": 8.743703867807522e-05}
{"train_loss": 0.26593056321144104, "global_step": 62997, "epoch": 707, "lr": 8.74366543896596e-05}
{"train_loss": 0.22970864176750183, "global_step": 62998, "epoch": 707, "lr": 8.743627009621104e-05}
{"train_loss": 0.22383859753608704, "global_step": 62999, "epoch": 707, "lr": 8.743588579772966e-05}
{"train_loss": 0.18073439598083496, "global_step": 63000, "epoch": 707, "lr": 8.743550149421546e-05}
{"train_loss": 0.21654993295669556, "global_step": 63001, "epoch": 707, "lr": 8.743511718566852e-05}
{"train_loss": 0.21540144085884094, "global_step": 63002, "epoch": 707, "lr": 8.743473287208889e-05}
{"train_loss": 0.2210245430469513, "global_step": 63003, "epoch": 707, "lr": 8.743434855347661e-05}
{"train_loss": 0.2690972089767456, "global_step": 63004, "epoch": 707, "lr": 8.743396422983175e-05}
{"train_loss": 0.20304624736309052, "global_step": 63005, "epoch": 707, "lr": 8.743357990115433e-05}
{"train_loss": 0.22110038995742798, "global_step": 63006, "epoch": 707, "lr": 8.743319556744444e-05}
{"train_loss": 0.2874191105365753, "global_step": 63007, "epoch": 707, "lr": 8.74328112287021e-05}
{"train_loss": 0.2503810226917267, "global_step": 63008, "epoch": 707, "lr": 8.743242688492738e-05}
{"train_loss": 0.19391344487667084, "global_step": 63009, "epoch": 707, "lr": 8.743204253612032e-05}
{"train_loss": 0.3207128942012787, "global_step": 63010, "epoch": 707, "lr": 8.743165818228099e-05}
{"train_loss": 0.23797052847535421, "global_step": 63011, "epoch": 707, "lr": 8.743127382340942e-05, "val_loss": 3.4155890941619873}
{"train_loss": 0.20418429374694824, "global_step": 63012, "epoch": 708, "lr": 8.743088945950568e-05}
{"train_loss": 0.27588263154029846, "global_step": 63013, "epoch": 708, "lr": 8.74305050905698e-05}
{"train_loss": 0.2108626663684845, "global_step": 63014, "epoch": 708, "lr": 8.743012071660186e-05}
{"train_loss": 0.22068554162979126, "global_step": 63015, "epoch": 708, "lr": 8.742973633760189e-05}
{"train_loss": 0.27679917216300964, "global_step": 63016, "epoch": 708, "lr": 8.742935195356995e-05}
{"train_loss": 0.23315925896167755, "global_step": 63017, "epoch": 708, "lr": 8.742896756450609e-05}
{"train_loss": 0.2747836709022522, "global_step": 63018, "epoch": 708, "lr": 8.742858317041036e-05}
{"train_loss": 0.23007206618785858, "global_step": 63019, "epoch": 708, "lr": 8.742819877128282e-05}
{"train_loss": 0.18667469918727875, "global_step": 63020, "epoch": 708, "lr": 8.74278143671235e-05}
{"train_loss": 0.23665344715118408, "global_step": 63021, "epoch": 708, "lr": 8.742742995793247e-05}
{"train_loss": 0.22805465757846832, "global_step": 63022, "epoch": 708, "lr": 8.742704554370977e-05}
{"train_loss": 0.26891642808914185, "global_step": 63023, "epoch": 708, "lr": 8.742666112445547e-05}
{"train_loss": 0.27196088433265686, "global_step": 63024, "epoch": 708, "lr": 8.742627670016963e-05}
{"train_loss": 0.23730595409870148, "global_step": 63025, "epoch": 708, "lr": 8.742589227085227e-05}
{"train_loss": 0.2283884584903717, "global_step": 63026, "epoch": 708, "lr": 8.742550783650345e-05}
{"train_loss": 0.1933826059103012, "global_step": 63027, "epoch": 708, "lr": 8.742512339712322e-05}
{"train_loss": 0.19825179874897003, "global_step": 63028, "epoch": 708, "lr": 8.742473895271165e-05}
{"train_loss": 0.16539965569972992, "global_step": 63029, "epoch": 708, "lr": 8.742435450326878e-05}
{"train_loss": 0.3214341104030609, "global_step": 63030, "epoch": 708, "lr": 8.742397004879466e-05}
{"train_loss": 0.1927684098482132, "global_step": 63031, "epoch": 708, "lr": 8.742358558928933e-05}
{"train_loss": 0.2782496511936188, "global_step": 63032, "epoch": 708, "lr": 8.742320112475287e-05}
{"train_loss": 0.2879081070423126, "global_step": 63033, "epoch": 708, "lr": 8.742281665518531e-05}
{"train_loss": 0.1376066356897354, "global_step": 63034, "epoch": 708, "lr": 8.742243218058672e-05}
{"train_loss": 0.29306817054748535, "global_step": 63035, "epoch": 708, "lr": 8.742204770095712e-05}
{"train_loss": 0.19828614592552185, "global_step": 63036, "epoch": 708, "lr": 8.74216632162966e-05}
{"train_loss": 0.24480068683624268, "global_step": 63037, "epoch": 708, "lr": 8.74212787266052e-05}
{"train_loss": 0.27391111850738525, "global_step": 63038, "epoch": 708, "lr": 8.742089423188295e-05}
{"train_loss": 0.17158615589141846, "global_step": 63039, "epoch": 708, "lr": 8.742050973212992e-05}
{"train_loss": 0.3479636609554291, "global_step": 63040, "epoch": 708, "lr": 8.742012522734615e-05}
{"train_loss": 0.1457955688238144, "global_step": 63041, "epoch": 708, "lr": 8.74197407175317e-05}
{"train_loss": 0.27911436557769775, "global_step": 63042, "epoch": 708, "lr": 8.741935620268665e-05}
{"train_loss": 0.1678263545036316, "global_step": 63043, "epoch": 708, "lr": 8.741897168281099e-05}
{"train_loss": 0.21456781029701233, "global_step": 63044, "epoch": 708, "lr": 8.741858715790483e-05}
{"train_loss": 0.2159154862165451, "global_step": 63045, "epoch": 708, "lr": 8.741820262796818e-05}
{"train_loss": 0.27106401324272156, "global_step": 63046, "epoch": 708, "lr": 8.741781809300112e-05}
{"train_loss": 0.1987786740064621, "global_step": 63047, "epoch": 708, "lr": 8.741743355300369e-05}
{"train_loss": 0.23088310658931732, "global_step": 63048, "epoch": 708, "lr": 8.741704900797592e-05}
{"train_loss": 0.23176021873950958, "global_step": 63049, "epoch": 708, "lr": 8.74166644579179e-05}
{"train_loss": 0.18863946199417114, "global_step": 63050, "epoch": 708, "lr": 8.741627990282968e-05}
{"train_loss": 0.20276600122451782, "global_step": 63051, "epoch": 708, "lr": 8.741589534271128e-05}
{"train_loss": 0.1500306874513626, "global_step": 63052, "epoch": 708, "lr": 8.741551077756277e-05}
{"train_loss": 0.2415672093629837, "global_step": 63053, "epoch": 708, "lr": 8.74151262073842e-05}
{"train_loss": 0.21023261547088623, "global_step": 63054, "epoch": 708, "lr": 8.741474163217562e-05}
{"train_loss": 0.16686855256557465, "global_step": 63055, "epoch": 708, "lr": 8.74143570519371e-05}
{"train_loss": 0.2529141306877136, "global_step": 63056, "epoch": 708, "lr": 8.741397246666867e-05}
{"train_loss": 0.17262953519821167, "global_step": 63057, "epoch": 708, "lr": 8.741358787637037e-05}
{"train_loss": 0.29179704189300537, "global_step": 63058, "epoch": 708, "lr": 8.741320328104229e-05}
{"train_loss": 0.25178253650665283, "global_step": 63059, "epoch": 708, "lr": 8.741281868068445e-05}
{"train_loss": 0.20530574023723602, "global_step": 63060, "epoch": 708, "lr": 8.741243407529692e-05}
{"train_loss": 0.17215533554553986, "global_step": 63061, "epoch": 708, "lr": 8.741204946487973e-05}
{"train_loss": 0.16300928592681885, "global_step": 63062, "epoch": 708, "lr": 8.741166484943296e-05}
{"train_loss": 0.16835591197013855, "global_step": 63063, "epoch": 708, "lr": 8.741128022895664e-05}
{"train_loss": 0.20296961069107056, "global_step": 63064, "epoch": 708, "lr": 8.741089560345084e-05}
{"train_loss": 0.1839045286178589, "global_step": 63065, "epoch": 708, "lr": 8.74105109729156e-05}
{"train_loss": 0.2447558343410492, "global_step": 63066, "epoch": 708, "lr": 8.741012633735095e-05}
{"train_loss": 0.2513108551502228, "global_step": 63067, "epoch": 708, "lr": 8.740974169675698e-05}
{"train_loss": 0.23335444927215576, "global_step": 63068, "epoch": 708, "lr": 8.740935705113374e-05}
{"train_loss": 0.19185316562652588, "global_step": 63069, "epoch": 708, "lr": 8.740897240048125e-05}
{"train_loss": 0.15671446919441223, "global_step": 63070, "epoch": 708, "lr": 8.740858774479958e-05}
{"train_loss": 0.15148672461509705, "global_step": 63071, "epoch": 708, "lr": 8.740820308408879e-05}
{"train_loss": 0.17467127740383148, "global_step": 63072, "epoch": 708, "lr": 8.740781841834892e-05}
{"train_loss": 0.20464691519737244, "global_step": 63073, "epoch": 708, "lr": 8.740743374758002e-05}
{"train_loss": 0.2605729401111603, "global_step": 63074, "epoch": 708, "lr": 8.740704907178216e-05}
{"train_loss": 0.15800394117832184, "global_step": 63075, "epoch": 708, "lr": 8.740666439095536e-05}
{"train_loss": 0.23332469165325165, "global_step": 63076, "epoch": 708, "lr": 8.74062797050997e-05}
{"train_loss": 0.23643143475055695, "global_step": 63077, "epoch": 708, "lr": 8.740589501421523e-05}
{"train_loss": 0.20628619194030762, "global_step": 63078, "epoch": 708, "lr": 8.740551031830197e-05}
{"train_loss": 0.17637231945991516, "global_step": 63079, "epoch": 708, "lr": 8.740512561736e-05}
{"train_loss": 0.18882358074188232, "global_step": 63080, "epoch": 708, "lr": 8.740474091138939e-05}
{"train_loss": 0.14545485377311707, "global_step": 63081, "epoch": 708, "lr": 8.740435620039015e-05}
{"train_loss": 0.18573641777038574, "global_step": 63082, "epoch": 708, "lr": 8.740397148436235e-05}
{"train_loss": 0.20234332978725433, "global_step": 63083, "epoch": 708, "lr": 8.740358676330605e-05}
{"train_loss": 0.21521201729774475, "global_step": 63084, "epoch": 708, "lr": 8.74032020372213e-05}
{"train_loss": 0.16659626364707947, "global_step": 63085, "epoch": 708, "lr": 8.740281730610811e-05}
{"train_loss": 0.2749854028224945, "global_step": 63086, "epoch": 708, "lr": 8.740243256996661e-05}
{"train_loss": 0.2537018060684204, "global_step": 63087, "epoch": 708, "lr": 8.740204782879678e-05}
{"train_loss": 0.21303918957710266, "global_step": 63088, "epoch": 708, "lr": 8.740166308259872e-05}
{"train_loss": 0.20390130579471588, "global_step": 63089, "epoch": 708, "lr": 8.740127833137243e-05}
{"train_loss": 0.1880376636981964, "global_step": 63090, "epoch": 708, "lr": 8.740089357511803e-05}
{"train_loss": 0.25867965817451477, "global_step": 63091, "epoch": 708, "lr": 8.740050881383553e-05}
{"train_loss": 0.21019801497459412, "global_step": 63092, "epoch": 708, "lr": 8.740012404752497e-05}
{"train_loss": 0.31033632159233093, "global_step": 63093, "epoch": 708, "lr": 8.739973927618643e-05}
{"train_loss": 0.2440197467803955, "global_step": 63094, "epoch": 708, "lr": 8.739935449981996e-05}
{"train_loss": 0.1433115303516388, "global_step": 63095, "epoch": 708, "lr": 8.73989697184256e-05}
{"train_loss": 0.253303200006485, "global_step": 63096, "epoch": 708, "lr": 8.739858493200339e-05}
{"train_loss": 0.24858830869197845, "global_step": 63097, "epoch": 708, "lr": 8.73982001405534e-05}
{"train_loss": 0.20616097748279572, "global_step": 63098, "epoch": 708, "lr": 8.739781534407568e-05}
{"train_loss": 0.3527350127696991, "global_step": 63099, "epoch": 708, "lr": 8.739743054257028e-05}
{"train_loss": 0.2222816234224298, "global_step": 63100, "epoch": 708, "lr": 8.739704573603726e-05, "val_loss": 3.702932357788086}
{"train_loss": 0.3540104627609253, "global_step": 63101, "epoch": 709, "lr": 8.739666092447665e-05}
{"train_loss": 0.19193071126937866, "global_step": 63102, "epoch": 709, "lr": 8.739627610788852e-05}
{"train_loss": 0.15406273305416107, "global_step": 63103, "epoch": 709, "lr": 8.739589128627292e-05}
{"train_loss": 0.20207452774047852, "global_step": 63104, "epoch": 709, "lr": 8.739550645962989e-05}
{"train_loss": 0.21003535389900208, "global_step": 63105, "epoch": 709, "lr": 8.73951216279595e-05}
{"train_loss": 0.2325379103422165, "global_step": 63106, "epoch": 709, "lr": 8.739473679126177e-05}
{"train_loss": 0.21186742186546326, "global_step": 63107, "epoch": 709, "lr": 8.73943519495368e-05}
{"train_loss": 0.2178439199924469, "global_step": 63108, "epoch": 709, "lr": 8.73939671027846e-05}
{"train_loss": 0.24300648272037506, "global_step": 63109, "epoch": 709, "lr": 8.739358225100524e-05}
{"train_loss": 0.33016437292099, "global_step": 63110, "epoch": 709, "lr": 8.739319739419877e-05}
{"train_loss": 0.2950400412082672, "global_step": 63111, "epoch": 709, "lr": 8.739281253236524e-05}
{"train_loss": 0.2525674104690552, "global_step": 63112, "epoch": 709, "lr": 8.73924276655047e-05}
{"train_loss": 0.21261711418628693, "global_step": 63113, "epoch": 709, "lr": 8.739204279361721e-05}
{"train_loss": 0.22173556685447693, "global_step": 63114, "epoch": 709, "lr": 8.739165791670282e-05}
{"train_loss": 0.26553240418434143, "global_step": 63115, "epoch": 709, "lr": 8.739127303476158e-05}
{"train_loss": 0.22771970927715302, "global_step": 63116, "epoch": 709, "lr": 8.739088814779351e-05}
{"train_loss": 0.3166755735874176, "global_step": 63117, "epoch": 709, "lr": 8.739050325579872e-05}
{"train_loss": 0.24586518108844757, "global_step": 63118, "epoch": 709, "lr": 8.739011835877722e-05}
{"train_loss": 0.23266960680484772, "global_step": 63119, "epoch": 709, "lr": 8.738973345672907e-05}
{"train_loss": 0.3034708797931671, "global_step": 63120, "epoch": 709, "lr": 8.738934854965433e-05}
{"train_loss": 0.3167065978050232, "global_step": 63121, "epoch": 709, "lr": 8.738896363755306e-05}
{"train_loss": 0.23414096236228943, "global_step": 63122, "epoch": 709, "lr": 8.73885787204253e-05}
{"train_loss": 0.2714140713214874, "global_step": 63123, "epoch": 709, "lr": 8.738819379827108e-05}
{"train_loss": 0.2352972775697708, "global_step": 63124, "epoch": 709, "lr": 8.738780887109048e-05}
{"train_loss": 0.33206114172935486, "global_step": 63125, "epoch": 709, "lr": 8.738742393888355e-05}
{"train_loss": 0.32824259996414185, "global_step": 63126, "epoch": 709, "lr": 8.738703900165033e-05}
{"train_loss": 0.33059388399124146, "global_step": 63127, "epoch": 709, "lr": 8.738665405939088e-05}
{"train_loss": 0.29622840881347656, "global_step": 63128, "epoch": 709, "lr": 8.738626911210526e-05}
{"train_loss": 0.31078338623046875, "global_step": 63129, "epoch": 709, "lr": 8.738588415979352e-05}
{"train_loss": 0.2845629155635834, "global_step": 63130, "epoch": 709, "lr": 8.738549920245568e-05}
{"train_loss": 0.237611323595047, "global_step": 63131, "epoch": 709, "lr": 8.738511424009182e-05}
{"train_loss": 0.1489977091550827, "global_step": 63132, "epoch": 709, "lr": 8.7384729272702e-05}
{"train_loss": 0.22443163394927979, "global_step": 63133, "epoch": 709, "lr": 8.738434430028624e-05}
{"train_loss": 0.2433386594057083, "global_step": 63134, "epoch": 709, "lr": 8.738395932284462e-05}
{"train_loss": 0.26512569189071655, "global_step": 63135, "epoch": 709, "lr": 8.738357434037718e-05}
{"train_loss": 0.2722458243370056, "global_step": 63136, "epoch": 709, "lr": 8.738318935288399e-05}
{"train_loss": 0.2666531503200531, "global_step": 63137, "epoch": 709, "lr": 8.738280436036507e-05}
{"train_loss": 0.20172719657421112, "global_step": 63138, "epoch": 709, "lr": 8.73824193628205e-05}
{"train_loss": 0.28855499625205994, "global_step": 63139, "epoch": 709, "lr": 8.738203436025029e-05}
{"train_loss": 0.21848580241203308, "global_step": 63140, "epoch": 709, "lr": 8.738164935265456e-05}
{"train_loss": 0.22625324130058289, "global_step": 63141, "epoch": 709, "lr": 8.73812643400333e-05}
{"train_loss": 0.2546108067035675, "global_step": 63142, "epoch": 709, "lr": 8.738087932238659e-05}
{"train_loss": 0.22613200545310974, "global_step": 63143, "epoch": 709, "lr": 8.738049429971446e-05}
{"train_loss": 0.2220095545053482, "global_step": 63144, "epoch": 709, "lr": 8.738010927201699e-05}
{"train_loss": 0.26876693964004517, "global_step": 63145, "epoch": 709, "lr": 8.737972423929423e-05}
{"train_loss": 0.2161567509174347, "global_step": 63146, "epoch": 709, "lr": 8.737933920154622e-05}
{"train_loss": 0.2139616757631302, "global_step": 63147, "epoch": 709, "lr": 8.737895415877299e-05}
{"train_loss": 0.2732025384902954, "global_step": 63148, "epoch": 709, "lr": 8.737856911097464e-05}
{"train_loss": 0.24829480051994324, "global_step": 63149, "epoch": 709, "lr": 8.73781840581512e-05}
{"train_loss": 0.1975872814655304, "global_step": 63150, "epoch": 709, "lr": 8.737779900030272e-05}
{"train_loss": 0.267719566822052, "global_step": 63151, "epoch": 709, "lr": 8.737741393742923e-05}
{"train_loss": 0.2194983959197998, "global_step": 63152, "epoch": 709, "lr": 8.73770288695308e-05}
{"train_loss": 0.2964068651199341, "global_step": 63153, "epoch": 709, "lr": 8.737664379660751e-05}
{"train_loss": 0.39036452770233154, "global_step": 63154, "epoch": 709, "lr": 8.737625871865937e-05}
{"train_loss": 0.25911733508110046, "global_step": 63155, "epoch": 709, "lr": 8.737587363568644e-05}
{"train_loss": 0.31080007553100586, "global_step": 63156, "epoch": 709, "lr": 8.73754885476888e-05}
{"train_loss": 0.18692399561405182, "global_step": 63157, "epoch": 709, "lr": 8.737510345466646e-05}
{"train_loss": 0.28011882305145264, "global_step": 63158, "epoch": 709, "lr": 8.73747183566195e-05}
{"train_loss": 0.25154173374176025, "global_step": 63159, "epoch": 709, "lr": 8.737433325354797e-05}
{"train_loss": 0.2731139659881592, "global_step": 63160, "epoch": 709, "lr": 8.737394814545193e-05}
{"train_loss": 0.2536214590072632, "global_step": 63161, "epoch": 709, "lr": 8.737356303233139e-05}
{"train_loss": 0.15403710305690765, "global_step": 63162, "epoch": 709, "lr": 8.737317791418644e-05}
{"train_loss": 0.20330549776554108, "global_step": 63163, "epoch": 709, "lr": 8.737279279101712e-05}
{"train_loss": 0.25049227476119995, "global_step": 63164, "epoch": 709, "lr": 8.737240766282349e-05}
{"train_loss": 0.18621324002742767, "global_step": 63165, "epoch": 709, "lr": 8.737202252960558e-05}
{"train_loss": 0.19927628338336945, "global_step": 63166, "epoch": 709, "lr": 8.737163739136349e-05}
{"train_loss": 0.23014600574970245, "global_step": 63167, "epoch": 709, "lr": 8.737125224809721e-05}
{"train_loss": 0.19333626329898834, "global_step": 63168, "epoch": 709, "lr": 8.737086709980683e-05}
{"train_loss": 0.29598402976989746, "global_step": 63169, "epoch": 709, "lr": 8.737048194649238e-05}
{"train_loss": 0.17074483633041382, "global_step": 63170, "epoch": 709, "lr": 8.737009678815396e-05}
{"train_loss": 0.2562146782875061, "global_step": 63171, "epoch": 709, "lr": 8.736971162479155e-05}
{"train_loss": 0.19335119426250458, "global_step": 63172, "epoch": 709, "lr": 8.736932645640524e-05}
{"train_loss": 0.17767216265201569, "global_step": 63173, "epoch": 709, "lr": 8.736894128299509e-05}
{"train_loss": 0.2386399358510971, "global_step": 63174, "epoch": 709, "lr": 8.736855610456114e-05}
{"train_loss": 0.3108392059803009, "global_step": 63175, "epoch": 709, "lr": 8.736817092110344e-05}
{"train_loss": 0.28455650806427, "global_step": 63176, "epoch": 709, "lr": 8.736778573262205e-05}
{"train_loss": 0.32405194640159607, "global_step": 63177, "epoch": 709, "lr": 8.736740053911702e-05}
{"train_loss": 0.23454353213310242, "global_step": 63178, "epoch": 709, "lr": 8.73670153405884e-05}
{"train_loss": 0.2985178530216217, "global_step": 63179, "epoch": 709, "lr": 8.736663013703624e-05}
{"train_loss": 0.2255706936120987, "global_step": 63180, "epoch": 709, "lr": 8.736624492846057e-05}
{"train_loss": 0.20757555961608887, "global_step": 63181, "epoch": 709, "lr": 8.736585971486148e-05}
{"train_loss": 0.24207663536071777, "global_step": 63182, "epoch": 709, "lr": 8.7365474496239e-05}
{"train_loss": 0.19349464774131775, "global_step": 63183, "epoch": 709, "lr": 8.73650892725932e-05}
{"train_loss": 0.3602631986141205, "global_step": 63184, "epoch": 709, "lr": 8.736470404392411e-05}
{"train_loss": 0.2762780785560608, "global_step": 63185, "epoch": 709, "lr": 8.73643188102318e-05}
{"train_loss": 0.23022542893886566, "global_step": 63186, "epoch": 709, "lr": 8.736393357151629e-05}
{"train_loss": 0.21849142014980316, "global_step": 63187, "epoch": 709, "lr": 8.736354832777767e-05}
{"train_loss": 0.21767070889472961, "global_step": 63188, "epoch": 709, "lr": 8.736316307901597e-05}
{"train_loss": 0.24857285769467943, "global_step": 63189, "epoch": 709, "lr": 8.736277782523126e-05, "val_loss": 3.5260143280029297}
{"train_loss": 0.20327559113502502, "global_step": 63190, "epoch": 710, "lr": 8.736239256642358e-05}
{"train_loss": 0.2172721028327942, "global_step": 63191, "epoch": 710, "lr": 8.736200730259296e-05}
{"train_loss": 0.20974783599376678, "global_step": 63192, "epoch": 710, "lr": 8.73616220337395e-05}
{"train_loss": 0.3043770492076874, "global_step": 63193, "epoch": 710, "lr": 8.736123675986321e-05}
{"train_loss": 0.15101158618927002, "global_step": 63194, "epoch": 710, "lr": 8.736085148096415e-05}
{"train_loss": 0.2481490969657898, "global_step": 63195, "epoch": 710, "lr": 8.736046619704241e-05}
{"train_loss": 0.17809651792049408, "global_step": 63196, "epoch": 710, "lr": 8.736008090809799e-05}
{"train_loss": 0.32577595114707947, "global_step": 63197, "epoch": 710, "lr": 8.735969561413095e-05}
{"train_loss": 0.2467816174030304, "global_step": 63198, "epoch": 710, "lr": 8.735931031514137e-05}
{"train_loss": 0.2157081663608551, "global_step": 63199, "epoch": 710, "lr": 8.735892501112928e-05}
{"train_loss": 0.2180994749069214, "global_step": 63200, "epoch": 710, "lr": 8.735853970209475e-05}
{"train_loss": 0.2518155872821808, "global_step": 63201, "epoch": 710, "lr": 8.735815438803781e-05}
{"train_loss": 0.3005543351173401, "global_step": 63202, "epoch": 710, "lr": 8.735776906895851e-05}
{"train_loss": 0.24687831103801727, "global_step": 63203, "epoch": 710, "lr": 8.735738374485693e-05}
{"train_loss": 0.264376699924469, "global_step": 63204, "epoch": 710, "lr": 8.73569984157331e-05}
{"train_loss": 0.19837337732315063, "global_step": 63205, "epoch": 710, "lr": 8.735661308158709e-05}
{"train_loss": 0.20280501246452332, "global_step": 63206, "epoch": 710, "lr": 8.735622774241893e-05}
{"train_loss": 0.3141897916793823, "global_step": 63207, "epoch": 710, "lr": 8.735584239822867e-05}
{"train_loss": 0.2843548059463501, "global_step": 63208, "epoch": 710, "lr": 8.735545704901637e-05}
{"train_loss": 0.36852237582206726, "global_step": 63209, "epoch": 710, "lr": 8.73550716947821e-05}
{"train_loss": 0.29375767707824707, "global_step": 63210, "epoch": 710, "lr": 8.73546863355259e-05}
{"train_loss": 0.2914011478424072, "global_step": 63211, "epoch": 710, "lr": 8.73543009712478e-05}
{"train_loss": 0.35106465220451355, "global_step": 63212, "epoch": 710, "lr": 8.735391560194787e-05}
{"train_loss": 0.2973220646381378, "global_step": 63213, "epoch": 710, "lr": 8.735353022762618e-05}
{"train_loss": 0.29011937975883484, "global_step": 63214, "epoch": 710, "lr": 8.735314484828274e-05}
{"train_loss": 0.18265412747859955, "global_step": 63215, "epoch": 710, "lr": 8.735275946391764e-05}
{"train_loss": 0.19575050473213196, "global_step": 63216, "epoch": 710, "lr": 8.735237407453091e-05}
{"train_loss": 0.2238897830247879, "global_step": 63217, "epoch": 710, "lr": 8.735198868012262e-05}
{"train_loss": 0.2977341115474701, "global_step": 63218, "epoch": 710, "lr": 8.73516032806928e-05}
{"train_loss": 0.23923610150814056, "global_step": 63219, "epoch": 710, "lr": 8.735121787624151e-05}
{"train_loss": 0.21933653950691223, "global_step": 63220, "epoch": 710, "lr": 8.73508324667688e-05}
{"train_loss": 0.239050030708313, "global_step": 63221, "epoch": 710, "lr": 8.735044705227475e-05}
{"train_loss": 0.20675711333751678, "global_step": 63222, "epoch": 710, "lr": 8.735006163275936e-05}
{"train_loss": 0.27483266592025757, "global_step": 63223, "epoch": 710, "lr": 8.734967620822272e-05}
{"train_loss": 0.22980433702468872, "global_step": 63224, "epoch": 710, "lr": 8.734929077866487e-05}
{"train_loss": 0.19625796377658844, "global_step": 63225, "epoch": 710, "lr": 8.734890534408588e-05}
{"train_loss": 0.23881833255290985, "global_step": 63226, "epoch": 710, "lr": 8.734851990448576e-05}
{"train_loss": 0.15629969537258148, "global_step": 63227, "epoch": 710, "lr": 8.73481344598646e-05}
{"train_loss": 0.25265681743621826, "global_step": 63228, "epoch": 710, "lr": 8.734774901022243e-05}
{"train_loss": 0.21595390141010284, "global_step": 63229, "epoch": 710, "lr": 8.734736355555933e-05}
{"train_loss": 0.2138977199792862, "global_step": 63230, "epoch": 710, "lr": 8.734697809587532e-05}
{"train_loss": 0.27964305877685547, "global_step": 63231, "epoch": 710, "lr": 8.734659263117046e-05}
{"train_loss": 0.2503668963909149, "global_step": 63232, "epoch": 710, "lr": 8.734620716144479e-05}
{"train_loss": 0.2037038654088974, "global_step": 63233, "epoch": 710, "lr": 8.73458216866984e-05}
{"train_loss": 0.2462713122367859, "global_step": 63234, "epoch": 710, "lr": 8.734543620693132e-05}
{"train_loss": 0.188236266374588, "global_step": 63235, "epoch": 710, "lr": 8.734505072214359e-05}
{"train_loss": 0.23920968174934387, "global_step": 63236, "epoch": 710, "lr": 8.734466523233529e-05}
{"train_loss": 0.3426886200904846, "global_step": 63237, "epoch": 710, "lr": 8.734427973750644e-05}
{"train_loss": 0.22470307350158691, "global_step": 63238, "epoch": 710, "lr": 8.734389423765712e-05}
{"train_loss": 0.20352508127689362, "global_step": 63239, "epoch": 710, "lr": 8.734350873278735e-05}
{"train_loss": 0.1928909420967102, "global_step": 63240, "epoch": 710, "lr": 8.734312322289721e-05}
{"train_loss": 0.2015243023633957, "global_step": 63241, "epoch": 710, "lr": 8.734273770798674e-05}
{"train_loss": 0.2447889745235443, "global_step": 63242, "epoch": 710, "lr": 8.7342352188056e-05}
{"train_loss": 0.238600954413414, "global_step": 63243, "epoch": 710, "lr": 8.734196666310503e-05}
{"train_loss": 0.2014087587594986, "global_step": 63244, "epoch": 710, "lr": 8.734158113313389e-05}
{"train_loss": 0.33928748965263367, "global_step": 63245, "epoch": 710, "lr": 8.734119559814263e-05}
{"train_loss": 0.3138139247894287, "global_step": 63246, "epoch": 710, "lr": 8.73408100581313e-05}
{"train_loss": 0.20561078190803528, "global_step": 63247, "epoch": 710, "lr": 8.734042451309995e-05}
{"train_loss": 0.23309555649757385, "global_step": 63248, "epoch": 710, "lr": 8.734003896304865e-05}
{"train_loss": 0.21075813472270966, "global_step": 63249, "epoch": 710, "lr": 8.733965340797742e-05}
{"train_loss": 0.24157372117042542, "global_step": 63250, "epoch": 710, "lr": 8.733926784788633e-05}
{"train_loss": 0.42743805050849915, "global_step": 63251, "epoch": 710, "lr": 8.733888228277544e-05}
{"train_loss": 0.2983163893222809, "global_step": 63252, "epoch": 710, "lr": 8.733849671264478e-05}
{"train_loss": 0.25763288140296936, "global_step": 63253, "epoch": 710, "lr": 8.733811113749443e-05}
{"train_loss": 0.16802629828453064, "global_step": 63254, "epoch": 710, "lr": 8.733772555732441e-05}
{"train_loss": 0.26173073053359985, "global_step": 63255, "epoch": 710, "lr": 8.73373399721348e-05}
{"train_loss": 0.2147538661956787, "global_step": 63256, "epoch": 710, "lr": 8.733695438192564e-05}
{"train_loss": 0.25308144092559814, "global_step": 63257, "epoch": 710, "lr": 8.733656878669697e-05}
{"train_loss": 0.1943780481815338, "global_step": 63258, "epoch": 710, "lr": 8.733618318644886e-05}
{"train_loss": 0.27105912566185, "global_step": 63259, "epoch": 710, "lr": 8.733579758118136e-05}
{"train_loss": 0.2304035872220993, "global_step": 63260, "epoch": 710, "lr": 8.733541197089451e-05}
{"train_loss": 0.21429124474525452, "global_step": 63261, "epoch": 710, "lr": 8.733502635558837e-05}
{"train_loss": 0.25818949937820435, "global_step": 63262, "epoch": 710, "lr": 8.733464073526298e-05}
{"train_loss": 0.20942765474319458, "global_step": 63263, "epoch": 710, "lr": 8.733425510991843e-05}
{"train_loss": 0.26332637667655945, "global_step": 63264, "epoch": 710, "lr": 8.733386947955473e-05}
{"train_loss": 0.21303679049015045, "global_step": 63265, "epoch": 710, "lr": 8.733348384417195e-05}
{"train_loss": 0.24170561134815216, "global_step": 63266, "epoch": 710, "lr": 8.733309820377012e-05}
{"train_loss": 0.2684527039527893, "global_step": 63267, "epoch": 710, "lr": 8.733271255834933e-05}
{"train_loss": 0.24820232391357422, "global_step": 63268, "epoch": 710, "lr": 8.733232690790961e-05}
{"train_loss": 0.29157981276512146, "global_step": 63269, "epoch": 710, "lr": 8.733194125245102e-05}
{"train_loss": 0.261713445186615, "global_step": 63270, "epoch": 710, "lr": 8.733155559197357e-05}
{"train_loss": 0.32425767183303833, "global_step": 63271, "epoch": 710, "lr": 8.733116992647737e-05}
{"train_loss": 0.27865785360336304, "global_step": 63272, "epoch": 710, "lr": 8.733078425596246e-05}
{"train_loss": 0.17833732068538666, "global_step": 63273, "epoch": 710, "lr": 8.733039858042887e-05}
{"train_loss": 0.21762852370738983, "global_step": 63274, "epoch": 710, "lr": 8.733001289987666e-05}
{"train_loss": 0.17209209501743317, "global_step": 63275, "epoch": 710, "lr": 8.73296272143059e-05}
{"train_loss": 0.3011860251426697, "global_step": 63276, "epoch": 710, "lr": 8.732924152371663e-05}
{"train_loss": 0.2846514582633972, "global_step": 63277, "epoch": 710, "lr": 8.732885582810886e-05}
{"train_loss": 0.24649163444390457, "global_step": 63278, "epoch": 710, "lr": 8.732847012748271e-05, "val_loss": 3.4728567600250244, "train_action_mse_error": 25.742996215820312}
{"train_loss": 0.16962724924087524, "global_step": 63279, "epoch": 711, "lr": 8.73280844218382e-05}
{"train_loss": 0.20414842665195465, "global_step": 63280, "epoch": 711, "lr": 8.732769871117537e-05}
{"train_loss": 0.26596513390541077, "global_step": 63281, "epoch": 711, "lr": 8.732731299549431e-05}
{"train_loss": 0.1750280261039734, "global_step": 63282, "epoch": 711, "lr": 8.732692727479503e-05}
{"train_loss": 0.24230754375457764, "global_step": 63283, "epoch": 711, "lr": 8.73265415490776e-05}
{"train_loss": 0.1519138216972351, "global_step": 63284, "epoch": 711, "lr": 8.732615581834209e-05}
{"train_loss": 0.20563730597496033, "global_step": 63285, "epoch": 711, "lr": 8.732577008258851e-05}
{"train_loss": 0.3037627041339874, "global_step": 63286, "epoch": 711, "lr": 8.732538434181695e-05}
{"train_loss": 0.30371296405792236, "global_step": 63287, "epoch": 711, "lr": 8.732499859602743e-05}
{"train_loss": 0.20255059003829956, "global_step": 63288, "epoch": 711, "lr": 8.732461284522004e-05}
{"train_loss": 0.2346130907535553, "global_step": 63289, "epoch": 711, "lr": 8.732422708939481e-05}
{"train_loss": 0.19609804451465607, "global_step": 63290, "epoch": 711, "lr": 8.732384132855177e-05}
{"train_loss": 0.24731242656707764, "global_step": 63291, "epoch": 711, "lr": 8.732345556269101e-05}
{"train_loss": 0.22343853116035461, "global_step": 63292, "epoch": 711, "lr": 8.732306979181258e-05}
{"train_loss": 0.2744383215904236, "global_step": 63293, "epoch": 711, "lr": 8.732268401591649e-05}
{"train_loss": 0.2380884289741516, "global_step": 63294, "epoch": 711, "lr": 8.732229823500283e-05}
{"train_loss": 0.257837176322937, "global_step": 63295, "epoch": 711, "lr": 8.732191244907163e-05}
{"train_loss": 0.1431351751089096, "global_step": 63296, "epoch": 711, "lr": 8.732152665812298e-05}
{"train_loss": 0.23474474251270294, "global_step": 63297, "epoch": 711, "lr": 8.73211408621569e-05}
{"train_loss": 0.2024243324995041, "global_step": 63298, "epoch": 711, "lr": 8.732075506117343e-05}
{"train_loss": 0.23021283745765686, "global_step": 63299, "epoch": 711, "lr": 8.732036925517266e-05}
{"train_loss": 0.2700476050376892, "global_step": 63300, "epoch": 711, "lr": 8.731998344415461e-05}
{"train_loss": 0.2910822927951813, "global_step": 63301, "epoch": 711, "lr": 8.731959762811934e-05}
{"train_loss": 0.2180081158876419, "global_step": 63302, "epoch": 711, "lr": 8.731921180706693e-05}
{"train_loss": 0.21738064289093018, "global_step": 63303, "epoch": 711, "lr": 8.731882598099738e-05}
{"train_loss": 0.21858876943588257, "global_step": 63304, "epoch": 711, "lr": 8.731844014991077e-05}
{"train_loss": 0.21543292701244354, "global_step": 63305, "epoch": 711, "lr": 8.731805431380715e-05}
{"train_loss": 0.26115137338638306, "global_step": 63306, "epoch": 711, "lr": 8.731766847268659e-05}
{"train_loss": 0.2222362607717514, "global_step": 63307, "epoch": 711, "lr": 8.731728262654911e-05}
{"train_loss": 0.16947075724601746, "global_step": 63308, "epoch": 711, "lr": 8.731689677539478e-05}
{"train_loss": 0.220133438706398, "global_step": 63309, "epoch": 711, "lr": 8.731651091922364e-05}
{"train_loss": 0.19397735595703125, "global_step": 63310, "epoch": 711, "lr": 8.731612505803576e-05}
{"train_loss": 0.24649101495742798, "global_step": 63311, "epoch": 711, "lr": 8.731573919183119e-05}
{"train_loss": 0.20977813005447388, "global_step": 63312, "epoch": 711, "lr": 8.731535332060997e-05}
{"train_loss": 0.143739715218544, "global_step": 63313, "epoch": 711, "lr": 8.731496744437213e-05}
{"train_loss": 0.13809944689273834, "global_step": 63314, "epoch": 711, "lr": 8.731458156311776e-05}
{"train_loss": 0.19182108342647552, "global_step": 63315, "epoch": 711, "lr": 8.731419567684692e-05}
{"train_loss": 0.2244635671377182, "global_step": 63316, "epoch": 711, "lr": 8.731380978555961e-05}
{"train_loss": 0.22113938629627228, "global_step": 63317, "epoch": 711, "lr": 8.731342388925593e-05}
{"train_loss": 0.1801348477602005, "global_step": 63318, "epoch": 711, "lr": 8.731303798793591e-05}
{"train_loss": 0.14980292320251465, "global_step": 63319, "epoch": 711, "lr": 8.731265208159962e-05}
{"train_loss": 0.19326506555080414, "global_step": 63320, "epoch": 711, "lr": 8.731226617024709e-05}
{"train_loss": 0.2682175636291504, "global_step": 63321, "epoch": 711, "lr": 8.731188025387838e-05}
{"train_loss": 0.21079598367214203, "global_step": 63322, "epoch": 711, "lr": 8.731149433249355e-05}
{"train_loss": 0.2223888337612152, "global_step": 63323, "epoch": 711, "lr": 8.731110840609262e-05}
{"train_loss": 0.2710784375667572, "global_step": 63324, "epoch": 711, "lr": 8.731072247467568e-05}
{"train_loss": 0.19433702528476715, "global_step": 63325, "epoch": 711, "lr": 8.731033653824279e-05}
{"train_loss": 0.18516884744167328, "global_step": 63326, "epoch": 711, "lr": 8.730995059679395e-05}
{"train_loss": 0.1469830572605133, "global_step": 63327, "epoch": 711, "lr": 8.730956465032926e-05}
{"train_loss": 0.24065636098384857, "global_step": 63328, "epoch": 711, "lr": 8.730917869884874e-05}
{"train_loss": 0.22087302803993225, "global_step": 63329, "epoch": 711, "lr": 8.730879274235246e-05}
{"train_loss": 0.24108648300170898, "global_step": 63330, "epoch": 711, "lr": 8.730840678084048e-05}
{"train_loss": 0.2240108698606491, "global_step": 63331, "epoch": 711, "lr": 8.730802081431282e-05}
{"train_loss": 0.21103407442569733, "global_step": 63332, "epoch": 711, "lr": 8.730763484276958e-05}
{"train_loss": 0.2455718219280243, "global_step": 63333, "epoch": 711, "lr": 8.730724886621076e-05}
{"train_loss": 0.18859228491783142, "global_step": 63334, "epoch": 711, "lr": 8.730686288463643e-05}
{"train_loss": 0.22647014260292053, "global_step": 63335, "epoch": 711, "lr": 8.730647689804665e-05}
{"train_loss": 0.2933961749076843, "global_step": 63336, "epoch": 711, "lr": 8.730609090644148e-05}
{"train_loss": 0.2678535580635071, "global_step": 63337, "epoch": 711, "lr": 8.730570490982095e-05}
{"train_loss": 0.1254701167345047, "global_step": 63338, "epoch": 711, "lr": 8.730531890818514e-05}
{"train_loss": 0.2022343873977661, "global_step": 63339, "epoch": 711, "lr": 8.730493290153406e-05}
{"train_loss": 0.24586331844329834, "global_step": 63340, "epoch": 711, "lr": 8.730454688986781e-05}
{"train_loss": 0.27400994300842285, "global_step": 63341, "epoch": 711, "lr": 8.73041608731864e-05}
{"train_loss": 0.258217453956604, "global_step": 63342, "epoch": 711, "lr": 8.730377485148991e-05}
{"train_loss": 0.19043360650539398, "global_step": 63343, "epoch": 711, "lr": 8.730338882477838e-05}
{"train_loss": 0.25182193517684937, "global_step": 63344, "epoch": 711, "lr": 8.730300279305187e-05}
{"train_loss": 0.1912182867527008, "global_step": 63345, "epoch": 711, "lr": 8.730261675631042e-05}
{"train_loss": 0.22326679527759552, "global_step": 63346, "epoch": 711, "lr": 8.730223071455408e-05}
{"train_loss": 0.2554924786090851, "global_step": 63347, "epoch": 711, "lr": 8.730184466778293e-05}
{"train_loss": 0.19310960173606873, "global_step": 63348, "epoch": 711, "lr": 8.7301458615997e-05}
{"train_loss": 0.28483280539512634, "global_step": 63349, "epoch": 711, "lr": 8.730107255919632e-05}
{"train_loss": 0.1826987862586975, "global_step": 63350, "epoch": 711, "lr": 8.730068649738098e-05}
{"train_loss": 0.2503463923931122, "global_step": 63351, "epoch": 711, "lr": 8.730030043055102e-05}
{"train_loss": 0.1571444869041443, "global_step": 63352, "epoch": 711, "lr": 8.729991435870649e-05}
{"train_loss": 0.2278456836938858, "global_step": 63353, "epoch": 711, "lr": 8.729952828184742e-05}
{"train_loss": 0.21624763309955597, "global_step": 63354, "epoch": 711, "lr": 8.72991421999739e-05}
{"train_loss": 0.18950580060482025, "global_step": 63355, "epoch": 711, "lr": 8.729875611308596e-05}
{"train_loss": 0.2386624813079834, "global_step": 63356, "epoch": 711, "lr": 8.729837002118366e-05}
{"train_loss": 0.2626708447933197, "global_step": 63357, "epoch": 711, "lr": 8.729798392426705e-05}
{"train_loss": 0.23431506752967834, "global_step": 63358, "epoch": 711, "lr": 8.729759782233619e-05}
{"train_loss": 0.2894902527332306, "global_step": 63359, "epoch": 711, "lr": 8.729721171539111e-05}
{"train_loss": 0.15958523750305176, "global_step": 63360, "epoch": 711, "lr": 8.729682560343189e-05}
{"train_loss": 0.2072066217660904, "global_step": 63361, "epoch": 711, "lr": 8.729643948645854e-05}
{"train_loss": 0.23361562192440033, "global_step": 63362, "epoch": 711, "lr": 8.729605336447115e-05}
{"train_loss": 0.19348494708538055, "global_step": 63363, "epoch": 711, "lr": 8.729566723746975e-05}
{"train_loss": 0.1995149552822113, "global_step": 63364, "epoch": 711, "lr": 8.729528110545444e-05}
{"train_loss": 0.24578920006752014, "global_step": 63365, "epoch": 711, "lr": 8.72948949684252e-05}
{"train_loss": 0.19857220351696014, "global_step": 63366, "epoch": 711, "lr": 8.729450882638213e-05}
{"train_loss": 0.22086200955208768, "global_step": 63367, "epoch": 711, "lr": 8.729412267932526e-05, "val_loss": 3.630156993865967}
{"train_loss": 0.17958658933639526, "global_step": 63368, "epoch": 712, "lr": 8.729373652725464e-05}
{"train_loss": 0.30210867524147034, "global_step": 63369, "epoch": 712, "lr": 8.729335037017034e-05}
{"train_loss": 0.21821661293506622, "global_step": 63370, "epoch": 712, "lr": 8.729296420807241e-05}
{"train_loss": 0.3145259916782379, "global_step": 63371, "epoch": 712, "lr": 8.72925780409609e-05}
{"train_loss": 0.19566981494426727, "global_step": 63372, "epoch": 712, "lr": 8.729219186883585e-05}
{"train_loss": 0.20008566975593567, "global_step": 63373, "epoch": 712, "lr": 8.729180569169732e-05}
{"train_loss": 0.2069854438304901, "global_step": 63374, "epoch": 712, "lr": 8.729141950954536e-05}
{"train_loss": 0.3247543275356293, "global_step": 63375, "epoch": 712, "lr": 8.729103332238003e-05}
{"train_loss": 0.19517894089221954, "global_step": 63376, "epoch": 712, "lr": 8.729064713020137e-05}
{"train_loss": 0.16916432976722717, "global_step": 63377, "epoch": 712, "lr": 8.729026093300942e-05}
{"train_loss": 0.2641798257827759, "global_step": 63378, "epoch": 712, "lr": 8.728987473080428e-05}
{"train_loss": 0.2528010606765747, "global_step": 63379, "epoch": 712, "lr": 8.728948852358595e-05}
{"train_loss": 0.15450911223888397, "global_step": 63380, "epoch": 712, "lr": 8.728910231135451e-05}
{"train_loss": 0.21469074487686157, "global_step": 63381, "epoch": 712, "lr": 8.728871609411e-05}
{"train_loss": 0.2521349787712097, "global_step": 63382, "epoch": 712, "lr": 8.728832987185248e-05}
{"train_loss": 0.28049197793006897, "global_step": 63383, "epoch": 712, "lr": 8.7287943644582e-05}
{"train_loss": 0.2126713991165161, "global_step": 63384, "epoch": 712, "lr": 8.72875574122986e-05}
{"train_loss": 0.22628472745418549, "global_step": 63385, "epoch": 712, "lr": 8.728717117500236e-05}
{"train_loss": 0.2559729814529419, "global_step": 63386, "epoch": 712, "lr": 8.728678493269331e-05}
{"train_loss": 0.2973645329475403, "global_step": 63387, "epoch": 712, "lr": 8.728639868537149e-05}
{"train_loss": 0.21139806509017944, "global_step": 63388, "epoch": 712, "lr": 8.728601243303699e-05}
{"train_loss": 0.17271161079406738, "global_step": 63389, "epoch": 712, "lr": 8.728562617568983e-05}
{"train_loss": 0.2983524799346924, "global_step": 63390, "epoch": 712, "lr": 8.728523991333006e-05}
{"train_loss": 0.3369509279727936, "global_step": 63391, "epoch": 712, "lr": 8.728485364595777e-05}
{"train_loss": 0.26207125186920166, "global_step": 63392, "epoch": 712, "lr": 8.728446737357296e-05}
{"train_loss": 0.29341059923171997, "global_step": 63393, "epoch": 712, "lr": 8.728408109617571e-05}
{"train_loss": 0.25039398670196533, "global_step": 63394, "epoch": 712, "lr": 8.728369481376609e-05}
{"train_loss": 0.22653749585151672, "global_step": 63395, "epoch": 712, "lr": 8.728330852634412e-05}
{"train_loss": 0.3519556522369385, "global_step": 63396, "epoch": 712, "lr": 8.728292223390987e-05}
{"train_loss": 0.2796699106693268, "global_step": 63397, "epoch": 712, "lr": 8.728253593646339e-05}
{"train_loss": 0.18836911022663116, "global_step": 63398, "epoch": 712, "lr": 8.72821496340047e-05}
{"train_loss": 0.21063707768917084, "global_step": 63399, "epoch": 712, "lr": 8.72817633265339e-05}
{"train_loss": 0.23014305531978607, "global_step": 63400, "epoch": 712, "lr": 8.728137701405101e-05}
{"train_loss": 0.3249181807041168, "global_step": 63401, "epoch": 712, "lr": 8.728099069655611e-05}
{"train_loss": 0.24304988980293274, "global_step": 63402, "epoch": 712, "lr": 8.728060437404924e-05}
{"train_loss": 0.20802289247512817, "global_step": 63403, "epoch": 712, "lr": 8.728021804653042e-05}
{"train_loss": 0.2516087591648102, "global_step": 63404, "epoch": 712, "lr": 8.727983171399974e-05}
{"train_loss": 0.2252204269170761, "global_step": 63405, "epoch": 712, "lr": 8.727944537645725e-05}
{"train_loss": 0.18347470462322235, "global_step": 63406, "epoch": 712, "lr": 8.727905903390299e-05}
{"train_loss": 0.3252507150173187, "global_step": 63407, "epoch": 712, "lr": 8.7278672686337e-05}
{"train_loss": 0.16871266067028046, "global_step": 63408, "epoch": 712, "lr": 8.727828633375936e-05}
{"train_loss": 0.14152328670024872, "global_step": 63409, "epoch": 712, "lr": 8.727789997617011e-05}
{"train_loss": 0.23166844248771667, "global_step": 63410, "epoch": 712, "lr": 8.72775136135693e-05}
{"train_loss": 0.24868594110012054, "global_step": 63411, "epoch": 712, "lr": 8.727712724595697e-05}
{"train_loss": 0.18136483430862427, "global_step": 63412, "epoch": 712, "lr": 8.72767408733332e-05}
{"train_loss": 0.2455415278673172, "global_step": 63413, "epoch": 712, "lr": 8.727635449569802e-05}
{"train_loss": 0.20756074786186218, "global_step": 63414, "epoch": 712, "lr": 8.72759681130515e-05}
{"train_loss": 0.22445926070213318, "global_step": 63415, "epoch": 712, "lr": 8.727558172539368e-05}
{"train_loss": 0.23253236711025238, "global_step": 63416, "epoch": 712, "lr": 8.727519533272458e-05}
{"train_loss": 0.2394002079963684, "global_step": 63417, "epoch": 712, "lr": 8.727480893504432e-05}
{"train_loss": 0.16967085003852844, "global_step": 63418, "epoch": 712, "lr": 8.72744225323529e-05}
{"train_loss": 0.15778253972530365, "global_step": 63419, "epoch": 712, "lr": 8.72740361246504e-05}
{"train_loss": 0.174141064286232, "global_step": 63420, "epoch": 712, "lr": 8.727364971193684e-05}
{"train_loss": 0.2512853443622589, "global_step": 63421, "epoch": 712, "lr": 8.727326329421231e-05}
{"train_loss": 0.2268301248550415, "global_step": 63422, "epoch": 712, "lr": 8.727287687147684e-05}
{"train_loss": 0.15960782766342163, "global_step": 63423, "epoch": 712, "lr": 8.727249044373048e-05}
{"train_loss": 0.255168616771698, "global_step": 63424, "epoch": 712, "lr": 8.72721040109733e-05}
{"train_loss": 0.17634855210781097, "global_step": 63425, "epoch": 712, "lr": 8.727171757320532e-05}
{"train_loss": 0.20124614238739014, "global_step": 63426, "epoch": 712, "lr": 8.727133113042663e-05}
{"train_loss": 0.16316461563110352, "global_step": 63427, "epoch": 712, "lr": 8.727094468263726e-05}
{"train_loss": 0.2475052773952484, "global_step": 63428, "epoch": 712, "lr": 8.727055822983726e-05}
{"train_loss": 0.2527981996536255, "global_step": 63429, "epoch": 712, "lr": 8.727017177202669e-05}
{"train_loss": 0.20635797083377838, "global_step": 63430, "epoch": 712, "lr": 8.72697853092056e-05}
{"train_loss": 0.19176509976387024, "global_step": 63431, "epoch": 712, "lr": 8.726939884137404e-05}
{"train_loss": 0.17850659787654877, "global_step": 63432, "epoch": 712, "lr": 8.726901236853207e-05}
{"train_loss": 0.19402924180030823, "global_step": 63433, "epoch": 712, "lr": 8.726862589067972e-05}
{"train_loss": 0.26211097836494446, "global_step": 63434, "epoch": 712, "lr": 8.726823940781708e-05}
{"train_loss": 0.1797151118516922, "global_step": 63435, "epoch": 712, "lr": 8.726785291994418e-05}
{"train_loss": 0.19120554625988007, "global_step": 63436, "epoch": 712, "lr": 8.726746642706105e-05}
{"train_loss": 0.254290372133255, "global_step": 63437, "epoch": 712, "lr": 8.726707992916777e-05}
{"train_loss": 0.34407854080200195, "global_step": 63438, "epoch": 712, "lr": 8.72666934262644e-05}
{"train_loss": 0.20926807820796967, "global_step": 63439, "epoch": 712, "lr": 8.726630691835096e-05}
{"train_loss": 0.11855266243219376, "global_step": 63440, "epoch": 712, "lr": 8.726592040542752e-05}
{"train_loss": 0.1906842440366745, "global_step": 63441, "epoch": 712, "lr": 8.726553388749414e-05}
{"train_loss": 0.16719189286231995, "global_step": 63442, "epoch": 712, "lr": 8.726514736455085e-05}
{"train_loss": 0.24540622532367706, "global_step": 63443, "epoch": 712, "lr": 8.726476083659774e-05}
{"train_loss": 0.26194146275520325, "global_step": 63444, "epoch": 712, "lr": 8.726437430363481e-05}
{"train_loss": 0.1969396471977234, "global_step": 63445, "epoch": 712, "lr": 8.726398776566216e-05}
{"train_loss": 0.15930764377117157, "global_step": 63446, "epoch": 712, "lr": 8.72636012226798e-05}
{"train_loss": 0.21201221644878387, "global_step": 63447, "epoch": 712, "lr": 8.726321467468781e-05}
{"train_loss": 0.26132917404174805, "global_step": 63448, "epoch": 712, "lr": 8.726282812168625e-05}
{"train_loss": 0.1440758854150772, "global_step": 63449, "epoch": 712, "lr": 8.726244156367514e-05}
{"train_loss": 0.23571984469890594, "global_step": 63450, "epoch": 712, "lr": 8.726205500065456e-05}
{"train_loss": 0.3068389892578125, "global_step": 63451, "epoch": 712, "lr": 8.726166843262455e-05}
{"train_loss": 0.24291887879371643, "global_step": 63452, "epoch": 712, "lr": 8.726128185958515e-05}
{"train_loss": 0.27284443378448486, "global_step": 63453, "epoch": 712, "lr": 8.726089528153643e-05}
{"train_loss": 0.20916013419628143, "global_step": 63454, "epoch": 712, "lr": 8.726050869847844e-05}
{"train_loss": 0.2070430964231491, "global_step": 63455, "epoch": 712, "lr": 8.726012211041123e-05}
{"train_loss": 0.22758512810040055, "global_step": 63456, "epoch": 712, "lr": 8.725973551733486e-05, "val_loss": 3.52632474899292}
{"train_loss": 0.21036945283412933, "global_step": 63457, "epoch": 713, "lr": 8.725934891924937e-05}
{"train_loss": 0.20308133959770203, "global_step": 63458, "epoch": 713, "lr": 8.72589623161548e-05}
{"train_loss": 0.1560484617948532, "global_step": 63459, "epoch": 713, "lr": 8.725857570805123e-05}
{"train_loss": 0.19494014978408813, "global_step": 63460, "epoch": 713, "lr": 8.725818909493868e-05}
{"train_loss": 0.2690449655056, "global_step": 63461, "epoch": 713, "lr": 8.725780247681722e-05}
{"train_loss": 0.26728320121765137, "global_step": 63462, "epoch": 713, "lr": 8.725741585368693e-05}
{"train_loss": 0.1717694103717804, "global_step": 63463, "epoch": 713, "lr": 8.725702922554783e-05}
{"train_loss": 0.2105857878923416, "global_step": 63464, "epoch": 713, "lr": 8.725664259239996e-05}
{"train_loss": 0.21548861265182495, "global_step": 63465, "epoch": 713, "lr": 8.72562559542434e-05}
{"train_loss": 0.22306446731090546, "global_step": 63466, "epoch": 713, "lr": 8.725586931107819e-05}
{"train_loss": 0.23920519649982452, "global_step": 63467, "epoch": 713, "lr": 8.725548266290437e-05}
{"train_loss": 0.23051150143146515, "global_step": 63468, "epoch": 713, "lr": 8.725509600972201e-05}
{"train_loss": 0.1864994764328003, "global_step": 63469, "epoch": 713, "lr": 8.725470935153117e-05}
{"train_loss": 0.30038848519325256, "global_step": 63470, "epoch": 713, "lr": 8.725432268833187e-05}
{"train_loss": 0.2114337980747223, "global_step": 63471, "epoch": 713, "lr": 8.725393602012419e-05}
{"train_loss": 0.19492366909980774, "global_step": 63472, "epoch": 713, "lr": 8.725354934690817e-05}
{"train_loss": 0.16624662280082703, "global_step": 63473, "epoch": 713, "lr": 8.725316266868386e-05}
{"train_loss": 0.2599553167819977, "global_step": 63474, "epoch": 713, "lr": 8.725277598545134e-05}
{"train_loss": 0.23150724172592163, "global_step": 63475, "epoch": 713, "lr": 8.72523892972106e-05}
{"train_loss": 0.24160870909690857, "global_step": 63476, "epoch": 713, "lr": 8.725200260396176e-05}
{"train_loss": 0.24184809625148773, "global_step": 63477, "epoch": 713, "lr": 8.725161590570484e-05}
{"train_loss": 0.2340906262397766, "global_step": 63478, "epoch": 713, "lr": 8.725122920243986e-05}
{"train_loss": 0.2466398924589157, "global_step": 63479, "epoch": 713, "lr": 8.725084249416695e-05}
{"train_loss": 0.22903074324131012, "global_step": 63480, "epoch": 713, "lr": 8.725045578088609e-05}
{"train_loss": 0.36118394136428833, "global_step": 63481, "epoch": 713, "lr": 8.725006906259738e-05}
{"train_loss": 0.2084217220544815, "global_step": 63482, "epoch": 713, "lr": 8.724968233930084e-05}
{"train_loss": 0.20419842004776, "global_step": 63483, "epoch": 713, "lr": 8.724929561099654e-05}
{"train_loss": 0.23147062957286835, "global_step": 63484, "epoch": 713, "lr": 8.724890887768451e-05}
{"train_loss": 0.22709393501281738, "global_step": 63485, "epoch": 713, "lr": 8.724852213936486e-05}
{"train_loss": 0.2705281972885132, "global_step": 63486, "epoch": 713, "lr": 8.724813539603755e-05}
{"train_loss": 0.14277413487434387, "global_step": 63487, "epoch": 713, "lr": 8.724774864770271e-05}
{"train_loss": 0.30059295892715454, "global_step": 63488, "epoch": 713, "lr": 8.724736189436036e-05}
{"train_loss": 0.17962034046649933, "global_step": 63489, "epoch": 713, "lr": 8.724697513601054e-05}
{"train_loss": 0.1527693271636963, "global_step": 63490, "epoch": 713, "lr": 8.724658837265333e-05}
{"train_loss": 0.22467593848705292, "global_step": 63491, "epoch": 713, "lr": 8.724620160428878e-05}
{"train_loss": 0.19800692796707153, "global_step": 63492, "epoch": 713, "lr": 8.724581483091694e-05}
{"train_loss": 0.28370434045791626, "global_step": 63493, "epoch": 713, "lr": 8.724542805253784e-05}
{"train_loss": 0.13345219194889069, "global_step": 63494, "epoch": 713, "lr": 8.724504126915154e-05}
{"train_loss": 0.2818733751773834, "global_step": 63495, "epoch": 713, "lr": 8.72446544807581e-05}
{"train_loss": 0.2239033430814743, "global_step": 63496, "epoch": 713, "lr": 8.724426768735758e-05}
{"train_loss": 0.24682943522930145, "global_step": 63497, "epoch": 713, "lr": 8.724388088895002e-05}
{"train_loss": 0.1899741142988205, "global_step": 63498, "epoch": 713, "lr": 8.724349408553546e-05}
{"train_loss": 0.21660761535167694, "global_step": 63499, "epoch": 713, "lr": 8.724310727711398e-05}
{"train_loss": 0.2535114288330078, "global_step": 63500, "epoch": 713, "lr": 8.724272046368561e-05}
{"train_loss": 0.3161093592643738, "global_step": 63501, "epoch": 713, "lr": 8.724233364525041e-05}
{"train_loss": 0.20468908548355103, "global_step": 63502, "epoch": 713, "lr": 8.724194682180844e-05}
{"train_loss": 0.18210017681121826, "global_step": 63503, "epoch": 713, "lr": 8.724155999335975e-05}
{"train_loss": 0.18340909481048584, "global_step": 63504, "epoch": 713, "lr": 8.724117315990437e-05}
{"train_loss": 0.17950473725795746, "global_step": 63505, "epoch": 713, "lr": 8.724078632144238e-05}
{"train_loss": 0.2051878720521927, "global_step": 63506, "epoch": 713, "lr": 8.724039947797381e-05}
{"train_loss": 0.17144611477851868, "global_step": 63507, "epoch": 713, "lr": 8.724001262949872e-05}
{"train_loss": 0.23839925229549408, "global_step": 63508, "epoch": 713, "lr": 8.723962577601718e-05}
{"train_loss": 0.24234208464622498, "global_step": 63509, "epoch": 713, "lr": 8.72392389175292e-05}
{"train_loss": 0.14957427978515625, "global_step": 63510, "epoch": 713, "lr": 8.723885205403489e-05}
{"train_loss": 0.2113598734140396, "global_step": 63511, "epoch": 713, "lr": 8.723846518553427e-05}
{"train_loss": 0.27460378408432007, "global_step": 63512, "epoch": 713, "lr": 8.723807831202738e-05}
{"train_loss": 0.24949738383293152, "global_step": 63513, "epoch": 713, "lr": 8.723769143351428e-05}
{"train_loss": 0.2283552885055542, "global_step": 63514, "epoch": 713, "lr": 8.723730454999502e-05}
{"train_loss": 0.3397033214569092, "global_step": 63515, "epoch": 713, "lr": 8.723691766146967e-05}
{"train_loss": 0.1976192742586136, "global_step": 63516, "epoch": 713, "lr": 8.723653076793826e-05}
{"train_loss": 0.16803719103336334, "global_step": 63517, "epoch": 713, "lr": 8.723614386940087e-05}
{"train_loss": 0.27712753415107727, "global_step": 63518, "epoch": 713, "lr": 8.723575696585752e-05}
{"train_loss": 0.22036337852478027, "global_step": 63519, "epoch": 713, "lr": 8.723537005730828e-05}
{"train_loss": 0.19451987743377686, "global_step": 63520, "epoch": 713, "lr": 8.72349831437532e-05}
{"train_loss": 0.20323306322097778, "global_step": 63521, "epoch": 713, "lr": 8.723459622519233e-05}
{"train_loss": 0.34617120027542114, "global_step": 63522, "epoch": 713, "lr": 8.723420930162573e-05}
{"train_loss": 0.2171558141708374, "global_step": 63523, "epoch": 713, "lr": 8.723382237305343e-05}
{"train_loss": 0.20570135116577148, "global_step": 63524, "epoch": 713, "lr": 8.72334354394755e-05}
{"train_loss": 0.23001070320606232, "global_step": 63525, "epoch": 713, "lr": 8.723304850089199e-05}
{"train_loss": 0.23933762311935425, "global_step": 63526, "epoch": 713, "lr": 8.723266155730295e-05}
{"train_loss": 0.2628650367259979, "global_step": 63527, "epoch": 713, "lr": 8.723227460870845e-05}
{"train_loss": 0.214948371052742, "global_step": 63528, "epoch": 713, "lr": 8.72318876551085e-05}
{"train_loss": 0.24966007471084595, "global_step": 63529, "epoch": 713, "lr": 8.723150069650317e-05}
{"train_loss": 0.20819160342216492, "global_step": 63530, "epoch": 713, "lr": 8.723111373289255e-05}
{"train_loss": 0.31109610199928284, "global_step": 63531, "epoch": 713, "lr": 8.723072676427664e-05}
{"train_loss": 0.2197326123714447, "global_step": 63532, "epoch": 713, "lr": 8.723033979065551e-05}
{"train_loss": 0.28505662083625793, "global_step": 63533, "epoch": 713, "lr": 8.722995281202921e-05}
{"train_loss": 0.1719455122947693, "global_step": 63534, "epoch": 713, "lr": 8.722956582839782e-05}
{"train_loss": 0.2206871509552002, "global_step": 63535, "epoch": 713, "lr": 8.722917883976136e-05}
{"train_loss": 0.2722102701663971, "global_step": 63536, "epoch": 713, "lr": 8.722879184611987e-05}
{"train_loss": 0.18469949066638947, "global_step": 63537, "epoch": 713, "lr": 8.722840484747346e-05}
{"train_loss": 0.23474404215812683, "global_step": 63538, "epoch": 713, "lr": 8.722801784382211e-05}
{"train_loss": 0.24542458355426788, "global_step": 63539, "epoch": 713, "lr": 8.722763083516592e-05}
{"train_loss": 0.23614461719989777, "global_step": 63540, "epoch": 713, "lr": 8.722724382150493e-05}
{"train_loss": 0.21503520011901855, "global_step": 63541, "epoch": 713, "lr": 8.72268568028392e-05}
{"train_loss": 0.31885844469070435, "global_step": 63542, "epoch": 713, "lr": 8.722646977916875e-05}
{"train_loss": 0.2243395447731018, "global_step": 63543, "epoch": 713, "lr": 8.722608275049368e-05}
{"train_loss": 0.2180618941783905, "global_step": 63544, "epoch": 713, "lr": 8.7225695716814e-05}
{"train_loss": 0.22736276167162348, "global_step": 63545, "epoch": 713, "lr": 8.722530867812979e-05, "val_loss": 3.600456953048706}
{"train_loss": 0.2052152454853058, "global_step": 63546, "epoch": 714, "lr": 8.722492163444108e-05}
{"train_loss": 0.26237666606903076, "global_step": 63547, "epoch": 714, "lr": 8.722453458574795e-05}
{"train_loss": 0.289387047290802, "global_step": 63548, "epoch": 714, "lr": 8.722414753205041e-05}
{"train_loss": 0.22015012800693512, "global_step": 63549, "epoch": 714, "lr": 8.722376047334855e-05}
{"train_loss": 0.17481422424316406, "global_step": 63550, "epoch": 714, "lr": 8.722337340964241e-05}
{"train_loss": 0.24411016702651978, "global_step": 63551, "epoch": 714, "lr": 8.722298634093204e-05}
{"train_loss": 0.2346256822347641, "global_step": 63552, "epoch": 714, "lr": 8.72225992672175e-05}
{"train_loss": 0.29644152522087097, "global_step": 63553, "epoch": 714, "lr": 8.722221218849881e-05}
{"train_loss": 0.3275929093360901, "global_step": 63554, "epoch": 714, "lr": 8.722182510477607e-05}
{"train_loss": 0.25916266441345215, "global_step": 63555, "epoch": 714, "lr": 8.722143801604932e-05}
{"train_loss": 0.1917247176170349, "global_step": 63556, "epoch": 714, "lr": 8.722105092231858e-05}
{"train_loss": 0.34181734919548035, "global_step": 63557, "epoch": 714, "lr": 8.722066382358393e-05}
{"train_loss": 0.28969940543174744, "global_step": 63558, "epoch": 714, "lr": 8.722027671984541e-05}
{"train_loss": 0.19892610609531403, "global_step": 63559, "epoch": 714, "lr": 8.72198896111031e-05}
{"train_loss": 0.2328799068927765, "global_step": 63560, "epoch": 714, "lr": 8.7219502497357e-05}
{"train_loss": 0.1961199790239334, "global_step": 63561, "epoch": 714, "lr": 8.721911537860721e-05}
{"train_loss": 0.25012916326522827, "global_step": 63562, "epoch": 714, "lr": 8.721872825485375e-05}
{"train_loss": 0.15482789278030396, "global_step": 63563, "epoch": 714, "lr": 8.721834112609671e-05}
{"train_loss": 0.17454494535923004, "global_step": 63564, "epoch": 714, "lr": 8.721795399233608e-05}
{"train_loss": 0.13342729210853577, "global_step": 63565, "epoch": 714, "lr": 8.721756685357199e-05}
{"train_loss": 0.3030954897403717, "global_step": 63566, "epoch": 714, "lr": 8.721717970980443e-05}
{"train_loss": 0.37038564682006836, "global_step": 63567, "epoch": 714, "lr": 8.721679256103347e-05}
{"train_loss": 0.16883954405784607, "global_step": 63568, "epoch": 714, "lr": 8.721640540725917e-05}
{"train_loss": 0.22102902829647064, "global_step": 63569, "epoch": 714, "lr": 8.721601824848158e-05}
{"train_loss": 0.16273409128189087, "global_step": 63570, "epoch": 714, "lr": 8.721563108470074e-05}
{"train_loss": 0.23096178472042084, "global_step": 63571, "epoch": 714, "lr": 8.721524391591673e-05}
{"train_loss": 0.29370012879371643, "global_step": 63572, "epoch": 714, "lr": 8.721485674212957e-05}
{"train_loss": 0.24515897035598755, "global_step": 63573, "epoch": 714, "lr": 8.721446956333934e-05}
{"train_loss": 0.1985083967447281, "global_step": 63574, "epoch": 714, "lr": 8.721408237954606e-05}
{"train_loss": 0.21934451162815094, "global_step": 63575, "epoch": 714, "lr": 8.721369519074982e-05}
{"train_loss": 0.22735652327537537, "global_step": 63576, "epoch": 714, "lr": 8.721330799695065e-05}
{"train_loss": 0.28206735849380493, "global_step": 63577, "epoch": 714, "lr": 8.721292079814858e-05}
{"train_loss": 0.18816901743412018, "global_step": 63578, "epoch": 714, "lr": 8.721253359434371e-05}
{"train_loss": 0.33172929286956787, "global_step": 63579, "epoch": 714, "lr": 8.721214638553605e-05}
{"train_loss": 0.18927250802516937, "global_step": 63580, "epoch": 714, "lr": 8.721175917172569e-05}
{"train_loss": 0.24697324633598328, "global_step": 63581, "epoch": 714, "lr": 8.721137195291266e-05}
{"train_loss": 0.2117508351802826, "global_step": 63582, "epoch": 714, "lr": 8.721098472909701e-05}
{"train_loss": 0.2743539810180664, "global_step": 63583, "epoch": 714, "lr": 8.721059750027879e-05}
{"train_loss": 0.1176866963505745, "global_step": 63584, "epoch": 714, "lr": 8.721021026645805e-05}
{"train_loss": 0.20800818502902985, "global_step": 63585, "epoch": 714, "lr": 8.720982302763488e-05}
{"train_loss": 0.3023620545864105, "global_step": 63586, "epoch": 714, "lr": 8.720943578380929e-05}
{"train_loss": 0.3041013181209564, "global_step": 63587, "epoch": 714, "lr": 8.720904853498134e-05}
{"train_loss": 0.2880878150463104, "global_step": 63588, "epoch": 714, "lr": 8.720866128115108e-05}
{"train_loss": 0.23773342370986938, "global_step": 63589, "epoch": 714, "lr": 8.72082740223186e-05}
{"train_loss": 0.1818803995847702, "global_step": 63590, "epoch": 714, "lr": 8.72078867584839e-05}
{"train_loss": 0.23382914066314697, "global_step": 63591, "epoch": 714, "lr": 8.720749948964704e-05}
{"train_loss": 0.28611448407173157, "global_step": 63592, "epoch": 714, "lr": 8.72071122158081e-05}
{"train_loss": 0.3459427058696747, "global_step": 63593, "epoch": 714, "lr": 8.720672493696711e-05}
{"train_loss": 0.18117395043373108, "global_step": 63594, "epoch": 714, "lr": 8.720633765312413e-05}
{"train_loss": 0.24156618118286133, "global_step": 63595, "epoch": 714, "lr": 8.720595036427923e-05}
{"train_loss": 0.195029154419899, "global_step": 63596, "epoch": 714, "lr": 8.720556307043243e-05}
{"train_loss": 0.22688913345336914, "global_step": 63597, "epoch": 714, "lr": 8.720517577158379e-05}
{"train_loss": 0.23709864914417267, "global_step": 63598, "epoch": 714, "lr": 8.720478846773336e-05}
{"train_loss": 0.30186909437179565, "global_step": 63599, "epoch": 714, "lr": 8.720440115888122e-05}
{"train_loss": 0.2086993008852005, "global_step": 63600, "epoch": 714, "lr": 8.720401384502739e-05}
{"train_loss": 0.2043309062719345, "global_step": 63601, "epoch": 714, "lr": 8.720362652617195e-05}
{"train_loss": 0.2287280261516571, "global_step": 63602, "epoch": 714, "lr": 8.720323920231492e-05}
{"train_loss": 0.32398712635040283, "global_step": 63603, "epoch": 714, "lr": 8.720285187345639e-05}
{"train_loss": 0.19736793637275696, "global_step": 63604, "epoch": 714, "lr": 8.720246453959635e-05}
{"train_loss": 0.19695551693439484, "global_step": 63605, "epoch": 714, "lr": 8.720207720073492e-05}
{"train_loss": 0.24278852343559265, "global_step": 63606, "epoch": 714, "lr": 8.720168985687211e-05}
{"train_loss": 0.2501133680343628, "global_step": 63607, "epoch": 714, "lr": 8.7201302508008e-05}
{"train_loss": 0.1980690360069275, "global_step": 63608, "epoch": 714, "lr": 8.720091515414263e-05}
{"train_loss": 0.31799548864364624, "global_step": 63609, "epoch": 714, "lr": 8.720052779527604e-05}
{"train_loss": 0.24535168707370758, "global_step": 63610, "epoch": 714, "lr": 8.72001404314083e-05}
{"train_loss": 0.23255838453769684, "global_step": 63611, "epoch": 714, "lr": 8.719975306253944e-05}
{"train_loss": 0.27569907903671265, "global_step": 63612, "epoch": 714, "lr": 8.719936568866953e-05}
{"train_loss": 0.26663053035736084, "global_step": 63613, "epoch": 714, "lr": 8.719897830979864e-05}
{"train_loss": 0.25920847058296204, "global_step": 63614, "epoch": 714, "lr": 8.719859092592678e-05}
{"train_loss": 0.21143873035907745, "global_step": 63615, "epoch": 714, "lr": 8.719820353705402e-05}
{"train_loss": 0.2610556483268738, "global_step": 63616, "epoch": 714, "lr": 8.719781614318043e-05}
{"train_loss": 0.20038631558418274, "global_step": 63617, "epoch": 714, "lr": 8.719742874430605e-05}
{"train_loss": 0.2068139761686325, "global_step": 63618, "epoch": 714, "lr": 8.719704134043091e-05}
{"train_loss": 0.1513882875442505, "global_step": 63619, "epoch": 714, "lr": 8.719665393155509e-05}
{"train_loss": 0.22780749201774597, "global_step": 63620, "epoch": 714, "lr": 8.719626651767864e-05}
{"train_loss": 0.22702255845069885, "global_step": 63621, "epoch": 714, "lr": 8.719587909880159e-05}
{"train_loss": 0.24913011491298676, "global_step": 63622, "epoch": 714, "lr": 8.719549167492402e-05}
{"train_loss": 0.16529029607772827, "global_step": 63623, "epoch": 714, "lr": 8.719510424604597e-05}
{"train_loss": 0.23939722776412964, "global_step": 63624, "epoch": 714, "lr": 8.719471681216749e-05}
{"train_loss": 0.19503682851791382, "global_step": 63625, "epoch": 714, "lr": 8.719432937328862e-05}
{"train_loss": 0.26105883717536926, "global_step": 63626, "epoch": 714, "lr": 8.719394192940943e-05}
{"train_loss": 0.2772364616394043, "global_step": 63627, "epoch": 714, "lr": 8.719355448052999e-05}
{"train_loss": 0.2665378451347351, "global_step": 63628, "epoch": 714, "lr": 8.719316702665031e-05}
{"train_loss": 0.25432687997817993, "global_step": 63629, "epoch": 714, "lr": 8.719277956777047e-05}
{"train_loss": 0.20081831514835358, "global_step": 63630, "epoch": 714, "lr": 8.719239210389052e-05}
{"train_loss": 0.3151484429836273, "global_step": 63631, "epoch": 714, "lr": 8.719200463501049e-05}
{"train_loss": 0.22672909498214722, "global_step": 63632, "epoch": 714, "lr": 8.719161716113045e-05}
{"train_loss": 0.23993344604969025, "global_step": 63633, "epoch": 714, "lr": 8.719122968225046e-05}
{"train_loss": 0.23760675136627776, "global_step": 63634, "epoch": 714, "lr": 8.719084219837056e-05, "val_loss": 3.480318546295166}
{"train_loss": 0.22877313196659088, "global_step": 63635, "epoch": 715, "lr": 8.719045470949081e-05}
{"train_loss": 0.29278427362442017, "global_step": 63636, "epoch": 715, "lr": 8.719006721561125e-05}
{"train_loss": 0.2428276091814041, "global_step": 63637, "epoch": 715, "lr": 8.718967971673194e-05}
{"train_loss": 0.21915581822395325, "global_step": 63638, "epoch": 715, "lr": 8.718929221285292e-05}
{"train_loss": 0.2512252628803253, "global_step": 63639, "epoch": 715, "lr": 8.718890470397426e-05}
{"train_loss": 0.26328665018081665, "global_step": 63640, "epoch": 715, "lr": 8.718851719009601e-05}
{"train_loss": 0.2965323030948639, "global_step": 63641, "epoch": 715, "lr": 8.718812967121821e-05}
{"train_loss": 0.3475581705570221, "global_step": 63642, "epoch": 715, "lr": 8.718774214734093e-05}
{"train_loss": 0.3380360007286072, "global_step": 63643, "epoch": 715, "lr": 8.71873546184642e-05}
{"train_loss": 0.22037287056446075, "global_step": 63644, "epoch": 715, "lr": 8.718696708458808e-05}
{"train_loss": 0.28464755415916443, "global_step": 63645, "epoch": 715, "lr": 8.718657954571264e-05}
{"train_loss": 0.20199692249298096, "global_step": 63646, "epoch": 715, "lr": 8.71861920018379e-05}
{"train_loss": 0.21306347846984863, "global_step": 63647, "epoch": 715, "lr": 8.718580445296396e-05}
{"train_loss": 0.18136663734912872, "global_step": 63648, "epoch": 715, "lr": 8.718541689909082e-05}
{"train_loss": 0.211966872215271, "global_step": 63649, "epoch": 715, "lr": 8.718502934021855e-05}
{"train_loss": 0.20046120882034302, "global_step": 63650, "epoch": 715, "lr": 8.718464177634722e-05}
{"train_loss": 0.19315515458583832, "global_step": 63651, "epoch": 715, "lr": 8.718425420747686e-05}
{"train_loss": 0.19242145121097565, "global_step": 63652, "epoch": 715, "lr": 8.718386663360752e-05}
{"train_loss": 0.27810972929000854, "global_step": 63653, "epoch": 715, "lr": 8.718347905473929e-05}
{"train_loss": 0.23064236342906952, "global_step": 63654, "epoch": 715, "lr": 8.718309147087219e-05}
{"train_loss": 0.3077116906642914, "global_step": 63655, "epoch": 715, "lr": 8.718270388200625e-05}
{"train_loss": 0.24848923087120056, "global_step": 63656, "epoch": 715, "lr": 8.718231628814156e-05}
{"train_loss": 0.18893033266067505, "global_step": 63657, "epoch": 715, "lr": 8.718192868927817e-05}
{"train_loss": 0.21894113719463348, "global_step": 63658, "epoch": 715, "lr": 8.718154108541611e-05}
{"train_loss": 0.2652082145214081, "global_step": 63659, "epoch": 715, "lr": 8.718115347655546e-05}
{"train_loss": 0.17773760855197906, "global_step": 63660, "epoch": 715, "lr": 8.718076586269623e-05}
{"train_loss": 0.20926763117313385, "global_step": 63661, "epoch": 715, "lr": 8.718037824383853e-05}
{"train_loss": 0.17365102469921112, "global_step": 63662, "epoch": 715, "lr": 8.717999061998237e-05}
{"train_loss": 0.18485704064369202, "global_step": 63663, "epoch": 715, "lr": 8.717960299112781e-05}
{"train_loss": 0.14383810758590698, "global_step": 63664, "epoch": 715, "lr": 8.717921535727491e-05}
{"train_loss": 0.195176362991333, "global_step": 63665, "epoch": 715, "lr": 8.717882771842372e-05}
{"train_loss": 0.24389271438121796, "global_step": 63666, "epoch": 715, "lr": 8.717844007457429e-05}
{"train_loss": 0.32319870591163635, "global_step": 63667, "epoch": 715, "lr": 8.717805242572667e-05}
{"train_loss": 0.1927073746919632, "global_step": 63668, "epoch": 715, "lr": 8.717766477188089e-05}
{"train_loss": 0.27679404616355896, "global_step": 63669, "epoch": 715, "lr": 8.717727711303706e-05}
{"train_loss": 0.18789808452129364, "global_step": 63670, "epoch": 715, "lr": 8.717688944919517e-05}
{"train_loss": 0.24216654896736145, "global_step": 63671, "epoch": 715, "lr": 8.717650178035532e-05}
{"train_loss": 0.16627062857151031, "global_step": 63672, "epoch": 715, "lr": 8.717611410651755e-05}
{"train_loss": 0.1819920539855957, "global_step": 63673, "epoch": 715, "lr": 8.71757264276819e-05}
{"train_loss": 0.19423049688339233, "global_step": 63674, "epoch": 715, "lr": 8.717533874384842e-05}
{"train_loss": 0.14915916323661804, "global_step": 63675, "epoch": 715, "lr": 8.717495105501716e-05}
{"train_loss": 0.20150810480117798, "global_step": 63676, "epoch": 715, "lr": 8.717456336118819e-05}
{"train_loss": 0.2515764534473419, "global_step": 63677, "epoch": 715, "lr": 8.717417566236156e-05}
{"train_loss": 0.31604957580566406, "global_step": 63678, "epoch": 715, "lr": 8.71737879585373e-05}
{"train_loss": 0.25243157148361206, "global_step": 63679, "epoch": 715, "lr": 8.717340024971549e-05}
{"train_loss": 0.1966300755739212, "global_step": 63680, "epoch": 715, "lr": 8.717301253589616e-05}
{"train_loss": 0.31358981132507324, "global_step": 63681, "epoch": 715, "lr": 8.717262481707937e-05}
{"train_loss": 0.21982094645500183, "global_step": 63682, "epoch": 715, "lr": 8.717223709326518e-05}
{"train_loss": 0.24074411392211914, "global_step": 63683, "epoch": 715, "lr": 8.717184936445364e-05}
{"train_loss": 0.16579394042491913, "global_step": 63684, "epoch": 715, "lr": 8.71714616306448e-05}
{"train_loss": 0.2170245200395584, "global_step": 63685, "epoch": 715, "lr": 8.71710738918387e-05}
{"train_loss": 0.2605442404747009, "global_step": 63686, "epoch": 715, "lr": 8.717068614803541e-05}
{"train_loss": 0.1506287008523941, "global_step": 63687, "epoch": 715, "lr": 8.717029839923498e-05}
{"train_loss": 0.26518455147743225, "global_step": 63688, "epoch": 715, "lr": 8.716991064543744e-05}
{"train_loss": 0.28286102414131165, "global_step": 63689, "epoch": 715, "lr": 8.716952288664286e-05}
{"train_loss": 0.22531311213970184, "global_step": 63690, "epoch": 715, "lr": 8.71691351228513e-05}
{"train_loss": 0.1762278527021408, "global_step": 63691, "epoch": 715, "lr": 8.71687473540628e-05}
{"train_loss": 0.21742981672286987, "global_step": 63692, "epoch": 715, "lr": 8.716835958027741e-05}
{"train_loss": 0.3160763680934906, "global_step": 63693, "epoch": 715, "lr": 8.716797180149518e-05}
{"train_loss": 0.2781996428966522, "global_step": 63694, "epoch": 715, "lr": 8.716758401771619e-05}
{"train_loss": 0.25202715396881104, "global_step": 63695, "epoch": 715, "lr": 8.716719622894046e-05}
{"train_loss": 0.21740038692951202, "global_step": 63696, "epoch": 715, "lr": 8.716680843516805e-05}
{"train_loss": 0.31669098138809204, "global_step": 63697, "epoch": 715, "lr": 8.716642063639903e-05}
{"train_loss": 0.1887381374835968, "global_step": 63698, "epoch": 715, "lr": 8.716603283263342e-05}
{"train_loss": 0.20309080183506012, "global_step": 63699, "epoch": 715, "lr": 8.716564502387131e-05}
{"train_loss": 0.1681661158800125, "global_step": 63700, "epoch": 715, "lr": 8.71652572101127e-05}
{"train_loss": 0.2831977903842926, "global_step": 63701, "epoch": 715, "lr": 8.71648693913577e-05}
{"train_loss": 0.2906172573566437, "global_step": 63702, "epoch": 715, "lr": 8.716448156760634e-05}
{"train_loss": 0.2619352340698242, "global_step": 63703, "epoch": 715, "lr": 8.716409373885866e-05}
{"train_loss": 0.15767116844654083, "global_step": 63704, "epoch": 715, "lr": 8.716370590511471e-05}
{"train_loss": 0.3065984845161438, "global_step": 63705, "epoch": 715, "lr": 8.716331806637457e-05}
{"train_loss": 0.2818189561367035, "global_step": 63706, "epoch": 715, "lr": 8.716293022263826e-05}
{"train_loss": 0.20655457675457, "global_step": 63707, "epoch": 715, "lr": 8.716254237390585e-05}
{"train_loss": 0.22253821790218353, "global_step": 63708, "epoch": 715, "lr": 8.716215452017739e-05}
{"train_loss": 0.1895030438899994, "global_step": 63709, "epoch": 715, "lr": 8.716176666145295e-05}
{"train_loss": 0.2717593312263489, "global_step": 63710, "epoch": 715, "lr": 8.716137879773253e-05}
{"train_loss": 0.25427302718162537, "global_step": 63711, "epoch": 715, "lr": 8.716099092901624e-05}
{"train_loss": 0.2519705295562744, "global_step": 63712, "epoch": 715, "lr": 8.716060305530411e-05}
{"train_loss": 0.24413348734378815, "global_step": 63713, "epoch": 715, "lr": 8.716021517659619e-05}
{"train_loss": 0.2276758998632431, "global_step": 63714, "epoch": 715, "lr": 8.715982729289251e-05}
{"train_loss": 0.2027781903743744, "global_step": 63715, "epoch": 715, "lr": 8.715943940419315e-05}
{"train_loss": 0.168918639421463, "global_step": 63716, "epoch": 715, "lr": 8.715905151049817e-05}
{"train_loss": 0.2130918949842453, "global_step": 63717, "epoch": 715, "lr": 8.71586636118076e-05}
{"train_loss": 0.21123695373535156, "global_step": 63718, "epoch": 715, "lr": 8.715827570812149e-05}
{"train_loss": 0.27296990156173706, "global_step": 63719, "epoch": 715, "lr": 8.715788779943991e-05}
{"train_loss": 0.2515442967414856, "global_step": 63720, "epoch": 715, "lr": 8.715749988576293e-05}
{"train_loss": 0.23110812902450562, "global_step": 63721, "epoch": 715, "lr": 8.715711196709055e-05}
{"train_loss": 0.21626679599285126, "global_step": 63722, "epoch": 715, "lr": 8.715672404342285e-05}
{"train_loss": 0.23281978740451043, "global_step": 63723, "epoch": 715, "lr": 8.71563361147599e-05, "val_loss": 3.7392096519470215, "train_action_mse_error": 28.932655334472656}
{"train_loss": 0.21867258846759796, "global_step": 63724, "epoch": 716, "lr": 8.715594818110171e-05}
{"train_loss": 0.19005531072616577, "global_step": 63725, "epoch": 716, "lr": 8.715556024244838e-05}
{"train_loss": 0.30240345001220703, "global_step": 63726, "epoch": 716, "lr": 8.715517229879993e-05}
{"train_loss": 0.1764669120311737, "global_step": 63727, "epoch": 716, "lr": 8.715478435015641e-05}
{"train_loss": 0.22354623675346375, "global_step": 63728, "epoch": 716, "lr": 8.715439639651788e-05}
{"train_loss": 0.19265447556972504, "global_step": 63729, "epoch": 716, "lr": 8.71540084378844e-05}
{"train_loss": 0.17096112668514252, "global_step": 63730, "epoch": 716, "lr": 8.715362047425601e-05}
{"train_loss": 0.13306252658367157, "global_step": 63731, "epoch": 716, "lr": 8.715323250563279e-05}
{"train_loss": 0.2365179806947708, "global_step": 63732, "epoch": 716, "lr": 8.715284453201475e-05}
{"train_loss": 0.2001689374446869, "global_step": 63733, "epoch": 716, "lr": 8.715245655340196e-05}
{"train_loss": 0.21146970987319946, "global_step": 63734, "epoch": 716, "lr": 8.71520685697945e-05}
{"train_loss": 0.17075857520103455, "global_step": 63735, "epoch": 716, "lr": 8.715168058119237e-05}
{"train_loss": 0.24389809370040894, "global_step": 63736, "epoch": 716, "lr": 8.715129258759566e-05}
{"train_loss": 0.2685345709323883, "global_step": 63737, "epoch": 716, "lr": 8.715090458900441e-05}
{"train_loss": 0.32333457469940186, "global_step": 63738, "epoch": 716, "lr": 8.715051658541868e-05}
{"train_loss": 0.16458676755428314, "global_step": 63739, "epoch": 716, "lr": 8.715012857683851e-05}
{"train_loss": 0.31096330285072327, "global_step": 63740, "epoch": 716, "lr": 8.714974056326395e-05}
{"train_loss": 0.2494926005601883, "global_step": 63741, "epoch": 716, "lr": 8.714935254469509e-05}
{"train_loss": 0.16431517899036407, "global_step": 63742, "epoch": 716, "lr": 8.714896452113192e-05}
{"train_loss": 0.1617608666419983, "global_step": 63743, "epoch": 716, "lr": 8.714857649257454e-05}
{"train_loss": 0.19533517956733704, "global_step": 63744, "epoch": 716, "lr": 8.714818845902298e-05}
{"train_loss": 0.1688382625579834, "global_step": 63745, "epoch": 716, "lr": 8.71478004204773e-05}
{"train_loss": 0.23747430741786957, "global_step": 63746, "epoch": 716, "lr": 8.714741237693755e-05}
{"train_loss": 0.25950103998184204, "global_step": 63747, "epoch": 716, "lr": 8.71470243284038e-05}
{"train_loss": 0.34192216396331787, "global_step": 63748, "epoch": 716, "lr": 8.714663627487607e-05}
{"train_loss": 0.24363888800144196, "global_step": 63749, "epoch": 716, "lr": 8.714624821635442e-05}
{"train_loss": 0.290446400642395, "global_step": 63750, "epoch": 716, "lr": 8.714586015283894e-05}
{"train_loss": 0.23978359997272491, "global_step": 63751, "epoch": 716, "lr": 8.714547208432962e-05}
{"train_loss": 0.2683714032173157, "global_step": 63752, "epoch": 716, "lr": 8.714508401082655e-05}
{"train_loss": 0.2750954329967499, "global_step": 63753, "epoch": 716, "lr": 8.71446959323298e-05}
{"train_loss": 0.2516511082649231, "global_step": 63754, "epoch": 716, "lr": 8.714430784883936e-05}
{"train_loss": 0.20228442549705505, "global_step": 63755, "epoch": 716, "lr": 8.714391976035534e-05}
{"train_loss": 0.1905880868434906, "global_step": 63756, "epoch": 716, "lr": 8.714353166687778e-05}
{"train_loss": 0.23416577279567719, "global_step": 63757, "epoch": 716, "lr": 8.714314356840672e-05}
{"train_loss": 0.20635069906711578, "global_step": 63758, "epoch": 716, "lr": 8.71427554649422e-05}
{"train_loss": 0.2015523761510849, "global_step": 63759, "epoch": 716, "lr": 8.714236735648432e-05}
{"train_loss": 0.27144482731819153, "global_step": 63760, "epoch": 716, "lr": 8.714197924303308e-05}
{"train_loss": 0.276664674282074, "global_step": 63761, "epoch": 716, "lr": 8.714159112458855e-05}
{"train_loss": 0.24389415979385376, "global_step": 63762, "epoch": 716, "lr": 8.714120300115081e-05}
{"train_loss": 0.24038463830947876, "global_step": 63763, "epoch": 716, "lr": 8.714081487271986e-05}
{"train_loss": 0.19649721682071686, "global_step": 63764, "epoch": 716, "lr": 8.714042673929579e-05}
{"train_loss": 0.18770453333854675, "global_step": 63765, "epoch": 716, "lr": 8.714003860087865e-05}
{"train_loss": 0.24470999836921692, "global_step": 63766, "epoch": 716, "lr": 8.713965045746848e-05}
{"train_loss": 0.2685290575027466, "global_step": 63767, "epoch": 716, "lr": 8.713926230906533e-05}
{"train_loss": 0.23983527719974518, "global_step": 63768, "epoch": 716, "lr": 8.713887415566927e-05}
{"train_loss": 0.3207116723060608, "global_step": 63769, "epoch": 716, "lr": 8.713848599728033e-05}
{"train_loss": 0.1565493494272232, "global_step": 63770, "epoch": 716, "lr": 8.71380978338986e-05}
{"train_loss": 0.25934460759162903, "global_step": 63771, "epoch": 716, "lr": 8.713770966552408e-05}
{"train_loss": 0.18373076617717743, "global_step": 63772, "epoch": 716, "lr": 8.713732149215685e-05}
{"train_loss": 0.19294406473636627, "global_step": 63773, "epoch": 716, "lr": 8.713693331379695e-05}
{"train_loss": 0.31854161620140076, "global_step": 63774, "epoch": 716, "lr": 8.713654513044446e-05}
{"train_loss": 0.20370395481586456, "global_step": 63775, "epoch": 716, "lr": 8.71361569420994e-05}
{"train_loss": 0.2743203341960907, "global_step": 63776, "epoch": 716, "lr": 8.713576874876184e-05}
{"train_loss": 0.20572014153003693, "global_step": 63777, "epoch": 716, "lr": 8.713538055043185e-05}
{"train_loss": 0.20788462460041046, "global_step": 63778, "epoch": 716, "lr": 8.713499234710944e-05}
{"train_loss": 0.13102103769779205, "global_step": 63779, "epoch": 716, "lr": 8.713460413879468e-05}
{"train_loss": 0.2609879672527313, "global_step": 63780, "epoch": 716, "lr": 8.713421592548762e-05}
{"train_loss": 0.22436976432800293, "global_step": 63781, "epoch": 716, "lr": 8.713382770718835e-05}
{"train_loss": 0.1568848341703415, "global_step": 63782, "epoch": 716, "lr": 8.713343948389686e-05}
{"train_loss": 0.21581393480300903, "global_step": 63783, "epoch": 716, "lr": 8.713305125561325e-05}
{"train_loss": 0.2161099761724472, "global_step": 63784, "epoch": 716, "lr": 8.713266302233753e-05}
{"train_loss": 0.25109124183654785, "global_step": 63785, "epoch": 716, "lr": 8.71322747840698e-05}
{"train_loss": 0.24856625497341156, "global_step": 63786, "epoch": 716, "lr": 8.713188654081009e-05}
{"train_loss": 0.17205095291137695, "global_step": 63787, "epoch": 716, "lr": 8.713149829255842e-05}
{"train_loss": 0.23501449823379517, "global_step": 63788, "epoch": 716, "lr": 8.713111003931489e-05}
{"train_loss": 0.21493659913539886, "global_step": 63789, "epoch": 716, "lr": 8.713072178107954e-05}
{"train_loss": 0.2087165266275406, "global_step": 63790, "epoch": 716, "lr": 8.713033351785242e-05}
{"train_loss": 0.2398356795310974, "global_step": 63791, "epoch": 716, "lr": 8.712994524963356e-05}
{"train_loss": 0.23980891704559326, "global_step": 63792, "epoch": 716, "lr": 8.712955697642304e-05}
{"train_loss": 0.2295421063899994, "global_step": 63793, "epoch": 716, "lr": 8.712916869822092e-05}
{"train_loss": 0.2714185416698456, "global_step": 63794, "epoch": 716, "lr": 8.712878041502721e-05}
{"train_loss": 0.15388645231723785, "global_step": 63795, "epoch": 716, "lr": 8.712839212684201e-05}
{"train_loss": 0.1701986938714981, "global_step": 63796, "epoch": 716, "lr": 8.712800383366534e-05}
{"train_loss": 0.30168208479881287, "global_step": 63797, "epoch": 716, "lr": 8.712761553549727e-05}
{"train_loss": 0.16855737566947937, "global_step": 63798, "epoch": 716, "lr": 8.712722723233784e-05}
{"train_loss": 0.2070564329624176, "global_step": 63799, "epoch": 716, "lr": 8.71268389241871e-05}
{"train_loss": 0.29924336075782776, "global_step": 63800, "epoch": 716, "lr": 8.712645061104512e-05}
{"train_loss": 0.2922113835811615, "global_step": 63801, "epoch": 716, "lr": 8.712606229291192e-05}
{"train_loss": 0.33906006813049316, "global_step": 63802, "epoch": 716, "lr": 8.712567396978759e-05}
{"train_loss": 0.1955896019935608, "global_step": 63803, "epoch": 716, "lr": 8.712528564167217e-05}
{"train_loss": 0.32240134477615356, "global_step": 63804, "epoch": 716, "lr": 8.71248973085657e-05}
{"train_loss": 0.22406110167503357, "global_step": 63805, "epoch": 716, "lr": 8.712450897046824e-05}
{"train_loss": 0.1617816537618637, "global_step": 63806, "epoch": 716, "lr": 8.712412062737984e-05}
{"train_loss": 0.2595776319503784, "global_step": 63807, "epoch": 716, "lr": 8.712373227930058e-05}
{"train_loss": 0.19616594910621643, "global_step": 63808, "epoch": 716, "lr": 8.712334392623044e-05}
{"train_loss": 0.21624243259429932, "global_step": 63809, "epoch": 716, "lr": 8.712295556816956e-05}
{"train_loss": 0.29911842942237854, "global_step": 63810, "epoch": 716, "lr": 8.712256720511793e-05}
{"train_loss": 0.2217303067445755, "global_step": 63811, "epoch": 716, "lr": 8.712217883707562e-05}
{"train_loss": 0.2293995049562347, "global_step": 63812, "epoch": 716, "lr": 8.712179046404271e-05, "val_loss": 3.5868585109710693}
{"train_loss": 0.27377331256866455, "global_step": 63813, "epoch": 717, "lr": 8.71214020860192e-05}
{"train_loss": 0.2528412640094757, "global_step": 63814, "epoch": 717, "lr": 8.712101370300518e-05}
{"train_loss": 0.21150141954421997, "global_step": 63815, "epoch": 717, "lr": 8.71206253150007e-05}
{"train_loss": 0.21585087478160858, "global_step": 63816, "epoch": 717, "lr": 8.712023692200581e-05}
{"train_loss": 0.34367191791534424, "global_step": 63817, "epoch": 717, "lr": 8.711984852402053e-05}
{"train_loss": 0.20058579742908478, "global_step": 63818, "epoch": 717, "lr": 8.711946012104496e-05}
{"train_loss": 0.14761844277381897, "global_step": 63819, "epoch": 717, "lr": 8.711907171307912e-05}
{"train_loss": 0.30777794122695923, "global_step": 63820, "epoch": 717, "lr": 8.711868330012308e-05}
{"train_loss": 0.20308858156204224, "global_step": 63821, "epoch": 717, "lr": 8.711829488217689e-05}
{"train_loss": 0.23522037267684937, "global_step": 63822, "epoch": 717, "lr": 8.711790645924058e-05}
{"train_loss": 0.17331238090991974, "global_step": 63823, "epoch": 717, "lr": 8.711751803131426e-05}
{"train_loss": 0.18454180657863617, "global_step": 63824, "epoch": 717, "lr": 8.711712959839789e-05}
{"train_loss": 0.2874048352241516, "global_step": 63825, "epoch": 717, "lr": 8.71167411604916e-05}
{"train_loss": 0.2064751237630844, "global_step": 63826, "epoch": 717, "lr": 8.711635271759542e-05}
{"train_loss": 0.16096380352973938, "global_step": 63827, "epoch": 717, "lr": 8.711596426970938e-05}
{"train_loss": 0.23428159952163696, "global_step": 63828, "epoch": 717, "lr": 8.711557581683357e-05}
{"train_loss": 0.2801659107208252, "global_step": 63829, "epoch": 717, "lr": 8.711518735896802e-05}
{"train_loss": 0.20396287739276886, "global_step": 63830, "epoch": 717, "lr": 8.711479889611278e-05}
{"train_loss": 0.22281846404075623, "global_step": 63831, "epoch": 717, "lr": 8.711441042826792e-05}
{"train_loss": 0.1900699883699417, "global_step": 63832, "epoch": 717, "lr": 8.711402195543345e-05}
{"train_loss": 0.19321762025356293, "global_step": 63833, "epoch": 717, "lr": 8.711363347760948e-05}
{"train_loss": 0.21883238852024078, "global_step": 63834, "epoch": 717, "lr": 8.711324499479603e-05}
{"train_loss": 0.17264869809150696, "global_step": 63835, "epoch": 717, "lr": 8.711285650699315e-05}
{"train_loss": 0.3080366253852844, "global_step": 63836, "epoch": 717, "lr": 8.71124680142009e-05}
{"train_loss": 0.1839287430047989, "global_step": 63837, "epoch": 717, "lr": 8.711207951641933e-05}
{"train_loss": 0.28383737802505493, "global_step": 63838, "epoch": 717, "lr": 8.711169101364851e-05}
{"train_loss": 0.14984966814517975, "global_step": 63839, "epoch": 717, "lr": 8.711130250588846e-05}
{"train_loss": 0.21745970845222473, "global_step": 63840, "epoch": 717, "lr": 8.711091399313924e-05}
{"train_loss": 0.21101176738739014, "global_step": 63841, "epoch": 717, "lr": 8.711052547540093e-05}
{"train_loss": 0.16756637394428253, "global_step": 63842, "epoch": 717, "lr": 8.711013695267354e-05}
{"train_loss": 0.14256598055362701, "global_step": 63843, "epoch": 717, "lr": 8.710974842495716e-05}
{"train_loss": 0.1910977065563202, "global_step": 63844, "epoch": 717, "lr": 8.710935989225182e-05}
{"train_loss": 0.20215223729610443, "global_step": 63845, "epoch": 717, "lr": 8.710897135455757e-05}
{"train_loss": 0.19691476225852966, "global_step": 63846, "epoch": 717, "lr": 8.71085828118745e-05}
{"train_loss": 0.20504862070083618, "global_step": 63847, "epoch": 717, "lr": 8.710819426420262e-05}
{"train_loss": 0.2790289521217346, "global_step": 63848, "epoch": 717, "lr": 8.710780571154198e-05}
{"train_loss": 0.18760469555854797, "global_step": 63849, "epoch": 717, "lr": 8.710741715389265e-05}
{"train_loss": 0.24865759909152985, "global_step": 63850, "epoch": 717, "lr": 8.710702859125471e-05}
{"train_loss": 0.25477004051208496, "global_step": 63851, "epoch": 717, "lr": 8.710664002362815e-05}
{"train_loss": 0.13672661781311035, "global_step": 63852, "epoch": 717, "lr": 8.710625145101306e-05}
{"train_loss": 0.3243636190891266, "global_step": 63853, "epoch": 717, "lr": 8.71058628734095e-05}
{"train_loss": 0.25822582840919495, "global_step": 63854, "epoch": 717, "lr": 8.71054742908175e-05}
{"train_loss": 0.20297373831272125, "global_step": 63855, "epoch": 717, "lr": 8.710508570323712e-05}
{"train_loss": 0.19410482048988342, "global_step": 63856, "epoch": 717, "lr": 8.710469711066842e-05}
{"train_loss": 0.17136847972869873, "global_step": 63857, "epoch": 717, "lr": 8.710430851311145e-05}
{"train_loss": 0.13847863674163818, "global_step": 63858, "epoch": 717, "lr": 8.710391991056624e-05}
{"train_loss": 0.21314717829227448, "global_step": 63859, "epoch": 717, "lr": 8.710353130303288e-05}
{"train_loss": 0.1765184849500656, "global_step": 63860, "epoch": 717, "lr": 8.710314269051138e-05}
{"train_loss": 0.1752311736345291, "global_step": 63861, "epoch": 717, "lr": 8.710275407300183e-05}
{"train_loss": 0.23583629727363586, "global_step": 63862, "epoch": 717, "lr": 8.710236545050427e-05}
{"train_loss": 0.22940757870674133, "global_step": 63863, "epoch": 717, "lr": 8.710197682301874e-05}
{"train_loss": 0.2902098298072815, "global_step": 63864, "epoch": 717, "lr": 8.71015881905453e-05}
{"train_loss": 0.17160865664482117, "global_step": 63865, "epoch": 717, "lr": 8.710119955308402e-05}
{"train_loss": 0.30774152278900146, "global_step": 63866, "epoch": 717, "lr": 8.710081091063492e-05}
{"train_loss": 0.22308140993118286, "global_step": 63867, "epoch": 717, "lr": 8.710042226319808e-05}
{"train_loss": 0.18307949602603912, "global_step": 63868, "epoch": 717, "lr": 8.710003361077352e-05}
{"train_loss": 0.20226392149925232, "global_step": 63869, "epoch": 717, "lr": 8.709964495336133e-05}
{"train_loss": 0.28948071599006653, "global_step": 63870, "epoch": 717, "lr": 8.709925629096154e-05}
{"train_loss": 0.23640726506710052, "global_step": 63871, "epoch": 717, "lr": 8.709886762357421e-05}
{"train_loss": 0.17045682668685913, "global_step": 63872, "epoch": 717, "lr": 8.709847895119938e-05}
{"train_loss": 0.2794065475463867, "global_step": 63873, "epoch": 717, "lr": 8.709809027383712e-05}
{"train_loss": 0.19202111661434174, "global_step": 63874, "epoch": 717, "lr": 8.709770159148749e-05}
{"train_loss": 0.20366500318050385, "global_step": 63875, "epoch": 717, "lr": 8.70973129041505e-05}
{"train_loss": 0.2082434743642807, "global_step": 63876, "epoch": 717, "lr": 8.709692421182623e-05}
{"train_loss": 0.2714306712150574, "global_step": 63877, "epoch": 717, "lr": 8.709653551451474e-05}
{"train_loss": 0.250189870595932, "global_step": 63878, "epoch": 717, "lr": 8.709614681221608e-05}
{"train_loss": 0.20499266684055328, "global_step": 63879, "epoch": 717, "lr": 8.709575810493028e-05}
{"train_loss": 0.2735013961791992, "global_step": 63880, "epoch": 717, "lr": 8.70953693926574e-05}
{"train_loss": 0.25010988116264343, "global_step": 63881, "epoch": 717, "lr": 8.709498067539751e-05}
{"train_loss": 0.2767760157585144, "global_step": 63882, "epoch": 717, "lr": 8.709459195315067e-05}
{"train_loss": 0.24372746050357819, "global_step": 63883, "epoch": 717, "lr": 8.70942032259169e-05}
{"train_loss": 0.3436787724494934, "global_step": 63884, "epoch": 717, "lr": 8.709381449369627e-05}
{"train_loss": 0.23855510354042053, "global_step": 63885, "epoch": 717, "lr": 8.709342575648882e-05}
{"train_loss": 0.2269827127456665, "global_step": 63886, "epoch": 717, "lr": 8.709303701429463e-05}
{"train_loss": 0.2419164627790451, "global_step": 63887, "epoch": 717, "lr": 8.709264826711371e-05}
{"train_loss": 0.2122926115989685, "global_step": 63888, "epoch": 717, "lr": 8.709225951494614e-05}
{"train_loss": 0.2866685092449188, "global_step": 63889, "epoch": 717, "lr": 8.709187075779197e-05}
{"train_loss": 0.2010478377342224, "global_step": 63890, "epoch": 717, "lr": 8.709148199565126e-05}
{"train_loss": 0.21222230792045593, "global_step": 63891, "epoch": 717, "lr": 8.709109322852405e-05}
{"train_loss": 0.2271774262189865, "global_step": 63892, "epoch": 717, "lr": 8.709070445641038e-05}
{"train_loss": 0.3152317404747009, "global_step": 63893, "epoch": 717, "lr": 8.709031567931034e-05}
{"train_loss": 0.4089638590812683, "global_step": 63894, "epoch": 717, "lr": 8.708992689722392e-05}
{"train_loss": 0.2973559498786926, "global_step": 63895, "epoch": 717, "lr": 8.708953811015125e-05}
{"train_loss": 0.20986011624336243, "global_step": 63896, "epoch": 717, "lr": 8.708914931809233e-05}
{"train_loss": 0.24045579135417938, "global_step": 63897, "epoch": 717, "lr": 8.708876052104722e-05}
{"train_loss": 0.2369643747806549, "global_step": 63898, "epoch": 717, "lr": 8.708837171901599e-05}
{"train_loss": 0.2412807196378708, "global_step": 63899, "epoch": 717, "lr": 8.708798291199867e-05}
{"train_loss": 0.22948181629180908, "global_step": 63900, "epoch": 717, "lr": 8.708759409999534e-05}
{"train_loss": 0.2287436274999983, "global_step": 63901, "epoch": 717, "lr": 8.708720528300602e-05, "val_loss": 3.607605218887329}
{"train_loss": 0.21562543511390686, "global_step": 63902, "epoch": 718, "lr": 8.708681646103078e-05}
{"train_loss": 0.20146577060222626, "global_step": 63903, "epoch": 718, "lr": 8.708642763406966e-05}
{"train_loss": 0.3404151201248169, "global_step": 63904, "epoch": 718, "lr": 8.708603880212274e-05}
{"train_loss": 0.22848618030548096, "global_step": 63905, "epoch": 718, "lr": 8.708564996519004e-05}
{"train_loss": 0.2261192500591278, "global_step": 63906, "epoch": 718, "lr": 8.708526112327163e-05}
{"train_loss": 0.2083241194486618, "global_step": 63907, "epoch": 718, "lr": 8.708487227636757e-05}
{"train_loss": 0.2326958030462265, "global_step": 63908, "epoch": 718, "lr": 8.708448342447788e-05}
{"train_loss": 0.32925945520401, "global_step": 63909, "epoch": 718, "lr": 8.708409456760265e-05}
{"train_loss": 0.18580719828605652, "global_step": 63910, "epoch": 718, "lr": 8.708370570574191e-05}
{"train_loss": 0.22130757570266724, "global_step": 63911, "epoch": 718, "lr": 8.708331683889573e-05}
{"train_loss": 0.24784299731254578, "global_step": 63912, "epoch": 718, "lr": 8.708292796706413e-05}
{"train_loss": 0.33583924174308777, "global_step": 63913, "epoch": 718, "lr": 8.708253909024719e-05}
{"train_loss": 0.20933285355567932, "global_step": 63914, "epoch": 718, "lr": 8.708215020844494e-05}
{"train_loss": 0.14906147122383118, "global_step": 63915, "epoch": 718, "lr": 8.708176132165748e-05}
{"train_loss": 0.23923319578170776, "global_step": 63916, "epoch": 718, "lr": 8.708137242988481e-05}
{"train_loss": 0.1532718986272812, "global_step": 63917, "epoch": 718, "lr": 8.7080983533127e-05}
{"train_loss": 0.18897603452205658, "global_step": 63918, "epoch": 718, "lr": 8.70805946313841e-05}
{"train_loss": 0.22441156208515167, "global_step": 63919, "epoch": 718, "lr": 8.708020572465618e-05}
{"train_loss": 0.22816434502601624, "global_step": 63920, "epoch": 718, "lr": 8.707981681294326e-05}
{"train_loss": 0.20332512259483337, "global_step": 63921, "epoch": 718, "lr": 8.707942789624544e-05}
{"train_loss": 0.28765493631362915, "global_step": 63922, "epoch": 718, "lr": 8.707903897456271e-05}
{"train_loss": 0.23692762851715088, "global_step": 63923, "epoch": 718, "lr": 8.707865004789517e-05}
{"train_loss": 0.28999581933021545, "global_step": 63924, "epoch": 718, "lr": 8.707826111624287e-05}
{"train_loss": 0.19173942506313324, "global_step": 63925, "epoch": 718, "lr": 8.707787217960582e-05}
{"train_loss": 0.21153493225574493, "global_step": 63926, "epoch": 718, "lr": 8.707748323798413e-05}
{"train_loss": 0.24761414527893066, "global_step": 63927, "epoch": 718, "lr": 8.70770942913778e-05}
{"train_loss": 0.24257893860340118, "global_step": 63928, "epoch": 718, "lr": 8.707670533978692e-05}
{"train_loss": 0.23360566794872284, "global_step": 63929, "epoch": 718, "lr": 8.707631638321153e-05}
{"train_loss": 0.2183648645877838, "global_step": 63930, "epoch": 718, "lr": 8.707592742165168e-05}
{"train_loss": 0.1837933510541916, "global_step": 63931, "epoch": 718, "lr": 8.707553845510744e-05}
{"train_loss": 0.19566909968852997, "global_step": 63932, "epoch": 718, "lr": 8.707514948357882e-05}
{"train_loss": 0.23225779831409454, "global_step": 63933, "epoch": 718, "lr": 8.707476050706589e-05}
{"train_loss": 0.3142452538013458, "global_step": 63934, "epoch": 718, "lr": 8.707437152556873e-05}
{"train_loss": 0.19710199534893036, "global_step": 63935, "epoch": 718, "lr": 8.707398253908737e-05}
{"train_loss": 0.38577505946159363, "global_step": 63936, "epoch": 718, "lr": 8.707359354762187e-05}
{"train_loss": 0.3086215853691101, "global_step": 63937, "epoch": 718, "lr": 8.707320455117228e-05}
{"train_loss": 0.17620234191417694, "global_step": 63938, "epoch": 718, "lr": 8.707281554973863e-05}
{"train_loss": 0.16737279295921326, "global_step": 63939, "epoch": 718, "lr": 8.707242654332099e-05}
{"train_loss": 0.2325248271226883, "global_step": 63940, "epoch": 718, "lr": 8.707203753191944e-05}
{"train_loss": 0.20967763662338257, "global_step": 63941, "epoch": 718, "lr": 8.707164851553399e-05}
{"train_loss": 0.17372140288352966, "global_step": 63942, "epoch": 718, "lr": 8.707125949416472e-05}
{"train_loss": 0.12989866733551025, "global_step": 63943, "epoch": 718, "lr": 8.707087046781166e-05}
{"train_loss": 0.2926398813724518, "global_step": 63944, "epoch": 718, "lr": 8.707048143647487e-05}
{"train_loss": 0.21940262615680695, "global_step": 63945, "epoch": 718, "lr": 8.707009240015441e-05}
{"train_loss": 0.1830761879682541, "global_step": 63946, "epoch": 718, "lr": 8.706970335885035e-05}
{"train_loss": 0.2807050347328186, "global_step": 63947, "epoch": 718, "lr": 8.706931431256269e-05}
{"train_loss": 0.24618752300739288, "global_step": 63948, "epoch": 718, "lr": 8.706892526129151e-05}
{"train_loss": 0.22636176645755768, "global_step": 63949, "epoch": 718, "lr": 8.706853620503687e-05}
{"train_loss": 0.18570156395435333, "global_step": 63950, "epoch": 718, "lr": 8.706814714379883e-05}
{"train_loss": 0.12657307088375092, "global_step": 63951, "epoch": 718, "lr": 8.706775807757744e-05}
{"train_loss": 0.30152735114097595, "global_step": 63952, "epoch": 718, "lr": 8.706736900637271e-05}
{"train_loss": 0.27372604608535767, "global_step": 63953, "epoch": 718, "lr": 8.706697993018475e-05}
{"train_loss": 0.2692614197731018, "global_step": 63954, "epoch": 718, "lr": 8.706659084901356e-05}
{"train_loss": 0.17450453341007233, "global_step": 63955, "epoch": 718, "lr": 8.706620176285925e-05}
{"train_loss": 0.23338453471660614, "global_step": 63956, "epoch": 718, "lr": 8.706581267172182e-05}
{"train_loss": 0.25921109318733215, "global_step": 63957, "epoch": 718, "lr": 8.706542357560135e-05}
{"train_loss": 0.19187860190868378, "global_step": 63958, "epoch": 718, "lr": 8.706503447449788e-05}
{"train_loss": 0.16057930886745453, "global_step": 63959, "epoch": 718, "lr": 8.706464536841148e-05}
{"train_loss": 0.20807835459709167, "global_step": 63960, "epoch": 718, "lr": 8.706425625734217e-05}
{"train_loss": 0.21272629499435425, "global_step": 63961, "epoch": 718, "lr": 8.706386714129005e-05}
{"train_loss": 0.21872113645076752, "global_step": 63962, "epoch": 718, "lr": 8.706347802025512e-05}
{"train_loss": 0.18022941052913666, "global_step": 63963, "epoch": 718, "lr": 8.706308889423746e-05}
{"train_loss": 0.23536436259746552, "global_step": 63964, "epoch": 718, "lr": 8.706269976323713e-05}
{"train_loss": 0.26681089401245117, "global_step": 63965, "epoch": 718, "lr": 8.706231062725417e-05}
{"train_loss": 0.1707724630832672, "global_step": 63966, "epoch": 718, "lr": 8.706192148628864e-05}
{"train_loss": 0.1661139577627182, "global_step": 63967, "epoch": 718, "lr": 8.706153234034058e-05}
{"train_loss": 0.3147837221622467, "global_step": 63968, "epoch": 718, "lr": 8.706114318941003e-05}
{"train_loss": 0.2823971211910248, "global_step": 63969, "epoch": 718, "lr": 8.70607540334971e-05}
{"train_loss": 0.2568919062614441, "global_step": 63970, "epoch": 718, "lr": 8.706036487260176e-05}
{"train_loss": 0.2788941562175751, "global_step": 63971, "epoch": 718, "lr": 8.705997570672412e-05}
{"train_loss": 0.17845147848129272, "global_step": 63972, "epoch": 718, "lr": 8.705958653586424e-05}
{"train_loss": 0.17203113436698914, "global_step": 63973, "epoch": 718, "lr": 8.705919736002213e-05}
{"train_loss": 0.3244403004646301, "global_step": 63974, "epoch": 718, "lr": 8.705880817919787e-05}
{"train_loss": 0.18653641641139984, "global_step": 63975, "epoch": 718, "lr": 8.705841899339152e-05}
{"train_loss": 0.3088046610355377, "global_step": 63976, "epoch": 718, "lr": 8.705802980260309e-05}
{"train_loss": 0.3264622390270233, "global_step": 63977, "epoch": 718, "lr": 8.705764060683265e-05}
{"train_loss": 0.16591987013816833, "global_step": 63978, "epoch": 718, "lr": 8.70572514060803e-05}
{"train_loss": 0.26251140236854553, "global_step": 63979, "epoch": 718, "lr": 8.705686220034603e-05}
{"train_loss": 0.21300239861011505, "global_step": 63980, "epoch": 718, "lr": 8.705647298962993e-05}
{"train_loss": 0.28612425923347473, "global_step": 63981, "epoch": 718, "lr": 8.705608377393202e-05}
{"train_loss": 0.19244837760925293, "global_step": 63982, "epoch": 718, "lr": 8.705569455325238e-05}
{"train_loss": 0.1929835081100464, "global_step": 63983, "epoch": 718, "lr": 8.705530532759106e-05}
{"train_loss": 0.18409790098667145, "global_step": 63984, "epoch": 718, "lr": 8.705491609694811e-05}
{"train_loss": 0.19534847140312195, "global_step": 63985, "epoch": 718, "lr": 8.705452686132355e-05}
{"train_loss": 0.19953015446662903, "global_step": 63986, "epoch": 718, "lr": 8.705413762071749e-05}
{"train_loss": 0.20355281233787537, "global_step": 63987, "epoch": 718, "lr": 8.705374837512994e-05}
{"train_loss": 0.2433120608329773, "global_step": 63988, "epoch": 718, "lr": 8.705335912456098e-05}
{"train_loss": 0.2386651635169983, "global_step": 63989, "epoch": 718, "lr": 8.705296986901063e-05}
{"train_loss": 0.2284355205431413, "global_step": 63990, "epoch": 718, "lr": 8.705258060847896e-05, "val_loss": 3.7098355293273926}
{"train_loss": 0.27882951498031616, "global_step": 63991, "epoch": 719, "lr": 8.705219134296603e-05}
{"train_loss": 0.17432758212089539, "global_step": 63992, "epoch": 719, "lr": 8.705180207247188e-05}
{"train_loss": 0.19636258482933044, "global_step": 63993, "epoch": 719, "lr": 8.705141279699658e-05}
{"train_loss": 0.28750449419021606, "global_step": 63994, "epoch": 719, "lr": 8.705102351654016e-05}
{"train_loss": 0.23317809402942657, "global_step": 63995, "epoch": 719, "lr": 8.705063423110267e-05}
{"train_loss": 0.28396397829055786, "global_step": 63996, "epoch": 719, "lr": 8.705024494068419e-05}
{"train_loss": 0.26876580715179443, "global_step": 63997, "epoch": 719, "lr": 8.704985564528474e-05}
{"train_loss": 0.24945150315761566, "global_step": 63998, "epoch": 719, "lr": 8.704946634490441e-05}
{"train_loss": 0.27267569303512573, "global_step": 63999, "epoch": 719, "lr": 8.704907703954321e-05}
{"train_loss": 0.2370380312204361, "global_step": 64000, "epoch": 719, "lr": 8.704868772920122e-05}
{"train_loss": 0.3516700863838196, "global_step": 64001, "epoch": 719, "lr": 8.70482984138785e-05}
{"train_loss": 0.2291576862335205, "global_step": 64002, "epoch": 719, "lr": 8.704790909357506e-05}
{"train_loss": 0.23536410927772522, "global_step": 64003, "epoch": 719, "lr": 8.7047519768291e-05}
{"train_loss": 0.31652072072029114, "global_step": 64004, "epoch": 719, "lr": 8.704713043802634e-05}
{"train_loss": 0.30296048521995544, "global_step": 64005, "epoch": 719, "lr": 8.704674110278116e-05}
{"train_loss": 0.2681933641433716, "global_step": 64006, "epoch": 719, "lr": 8.704635176255549e-05}
{"train_loss": 0.21069905161857605, "global_step": 64007, "epoch": 719, "lr": 8.704596241734939e-05}
{"train_loss": 0.24361953139305115, "global_step": 64008, "epoch": 719, "lr": 8.704557306716291e-05}
{"train_loss": 0.27610060572624207, "global_step": 64009, "epoch": 719, "lr": 8.70451837119961e-05}
{"train_loss": 0.21979749202728271, "global_step": 64010, "epoch": 719, "lr": 8.704479435184903e-05}
{"train_loss": 0.22180846333503723, "global_step": 64011, "epoch": 719, "lr": 8.704440498672173e-05}
{"train_loss": 0.196748748421669, "global_step": 64012, "epoch": 719, "lr": 8.704401561661426e-05}
{"train_loss": 0.2726651132106781, "global_step": 64013, "epoch": 719, "lr": 8.70436262415267e-05}
{"train_loss": 0.1820714920759201, "global_step": 64014, "epoch": 719, "lr": 8.704323686145904e-05}
{"train_loss": 0.17397426068782806, "global_step": 64015, "epoch": 719, "lr": 8.704284747641137e-05}
{"train_loss": 0.16812185943126678, "global_step": 64016, "epoch": 719, "lr": 8.704245808638376e-05}
{"train_loss": 0.32897669076919556, "global_step": 64017, "epoch": 719, "lr": 8.704206869137623e-05}
{"train_loss": 0.22981007397174835, "global_step": 64018, "epoch": 719, "lr": 8.704167929138884e-05}
{"train_loss": 0.2714378237724304, "global_step": 64019, "epoch": 719, "lr": 8.704128988642167e-05}
{"train_loss": 0.24118897318840027, "global_step": 64020, "epoch": 719, "lr": 8.704090047647472e-05}
{"train_loss": 0.3020186722278595, "global_step": 64021, "epoch": 719, "lr": 8.70405110615481e-05}
{"train_loss": 0.23303650319576263, "global_step": 64022, "epoch": 719, "lr": 8.704012164164182e-05}
{"train_loss": 0.2614538073539734, "global_step": 64023, "epoch": 719, "lr": 8.703973221675594e-05}
{"train_loss": 0.30732461810112, "global_step": 64024, "epoch": 719, "lr": 8.703934278689054e-05}
{"train_loss": 0.31501924991607666, "global_step": 64025, "epoch": 719, "lr": 8.703895335204564e-05}
{"train_loss": 0.27834564447402954, "global_step": 64026, "epoch": 719, "lr": 8.703856391222131e-05}
{"train_loss": 0.2268652319908142, "global_step": 64027, "epoch": 719, "lr": 8.70381744674176e-05}
{"train_loss": 0.16765272617340088, "global_step": 64028, "epoch": 719, "lr": 8.703778501763456e-05}
{"train_loss": 0.27439582347869873, "global_step": 64029, "epoch": 719, "lr": 8.703739556287223e-05}
{"train_loss": 0.30819055438041687, "global_step": 64030, "epoch": 719, "lr": 8.703700610313067e-05}
{"train_loss": 0.2939879894256592, "global_step": 64031, "epoch": 719, "lr": 8.703661663840994e-05}
{"train_loss": 0.24149303138256073, "global_step": 64032, "epoch": 719, "lr": 8.70362271687101e-05}
{"train_loss": 0.2592576742172241, "global_step": 64033, "epoch": 719, "lr": 8.703583769403116e-05}
{"train_loss": 0.24542826414108276, "global_step": 64034, "epoch": 719, "lr": 8.703544821437323e-05}
{"train_loss": 0.2434065043926239, "global_step": 64035, "epoch": 719, "lr": 8.703505872973633e-05}
{"train_loss": 0.22196152806282043, "global_step": 64036, "epoch": 719, "lr": 8.703466924012052e-05}
{"train_loss": 0.2979601323604584, "global_step": 64037, "epoch": 719, "lr": 8.703427974552585e-05}
{"train_loss": 0.22889447212219238, "global_step": 64038, "epoch": 719, "lr": 8.703389024595237e-05}
{"train_loss": 0.19381658732891083, "global_step": 64039, "epoch": 719, "lr": 8.703350074140015e-05}
{"train_loss": 0.27025240659713745, "global_step": 64040, "epoch": 719, "lr": 8.70331112318692e-05}
{"train_loss": 0.2785544693470001, "global_step": 64041, "epoch": 719, "lr": 8.703272171735961e-05}
{"train_loss": 0.20566941797733307, "global_step": 64042, "epoch": 719, "lr": 8.703233219787143e-05}
{"train_loss": 0.21228396892547607, "global_step": 64043, "epoch": 719, "lr": 8.703194267340468e-05}
{"train_loss": 0.2073114812374115, "global_step": 64044, "epoch": 719, "lr": 8.703155314395947e-05}
{"train_loss": 0.1705159693956375, "global_step": 64045, "epoch": 719, "lr": 8.70311636095358e-05}
{"train_loss": 0.19907613098621368, "global_step": 64046, "epoch": 719, "lr": 8.703077407013375e-05}
{"train_loss": 0.23728454113006592, "global_step": 64047, "epoch": 719, "lr": 8.703038452575336e-05}
{"train_loss": 0.20662012696266174, "global_step": 64048, "epoch": 719, "lr": 8.702999497639467e-05}
{"train_loss": 0.22087758779525757, "global_step": 64049, "epoch": 719, "lr": 8.702960542205778e-05}
{"train_loss": 0.25295931100845337, "global_step": 64050, "epoch": 719, "lr": 8.702921586274269e-05}
{"train_loss": 0.15217748284339905, "global_step": 64051, "epoch": 719, "lr": 8.702882629844948e-05}
{"train_loss": 0.2783238887786865, "global_step": 64052, "epoch": 719, "lr": 8.702843672917819e-05}
{"train_loss": 0.22142106294631958, "global_step": 64053, "epoch": 719, "lr": 8.702804715492888e-05}
{"train_loss": 0.18942676484584808, "global_step": 64054, "epoch": 719, "lr": 8.70276575757016e-05}
{"train_loss": 0.17395968735218048, "global_step": 64055, "epoch": 719, "lr": 8.70272679914964e-05}
{"train_loss": 0.29784926772117615, "global_step": 64056, "epoch": 719, "lr": 8.702687840231336e-05}
{"train_loss": 0.20121966302394867, "global_step": 64057, "epoch": 719, "lr": 8.702648880815249e-05}
{"train_loss": 0.16871489584445953, "global_step": 64058, "epoch": 719, "lr": 8.702609920901385e-05}
{"train_loss": 0.26635825634002686, "global_step": 64059, "epoch": 719, "lr": 8.702570960489751e-05}
{"train_loss": 0.23979078233242035, "global_step": 64060, "epoch": 719, "lr": 8.702531999580351e-05}
{"train_loss": 0.2257482260465622, "global_step": 64061, "epoch": 719, "lr": 8.702493038173191e-05}
{"train_loss": 0.23747088015079498, "global_step": 64062, "epoch": 719, "lr": 8.702454076268276e-05}
{"train_loss": 0.163747638463974, "global_step": 64063, "epoch": 719, "lr": 8.702415113865611e-05}
{"train_loss": 0.18501049280166626, "global_step": 64064, "epoch": 719, "lr": 8.702376150965202e-05}
{"train_loss": 0.23313991725444794, "global_step": 64065, "epoch": 719, "lr": 8.702337187567054e-05}
{"train_loss": 0.20972755551338196, "global_step": 64066, "epoch": 719, "lr": 8.70229822367117e-05}
{"train_loss": 0.2182748168706894, "global_step": 64067, "epoch": 719, "lr": 8.702259259277559e-05}
{"train_loss": 0.15703116357326508, "global_step": 64068, "epoch": 719, "lr": 8.702220294386223e-05}
{"train_loss": 0.2251586765050888, "global_step": 64069, "epoch": 719, "lr": 8.70218132899717e-05}
{"train_loss": 0.20815451443195343, "global_step": 64070, "epoch": 719, "lr": 8.702142363110402e-05}
{"train_loss": 0.2664254307746887, "global_step": 64071, "epoch": 719, "lr": 8.702103396725928e-05}
{"train_loss": 0.2181035876274109, "global_step": 64072, "epoch": 719, "lr": 8.702064429843751e-05}
{"train_loss": 0.21265098452568054, "global_step": 64073, "epoch": 719, "lr": 8.702025462463875e-05}
{"train_loss": 0.23454920947551727, "global_step": 64074, "epoch": 719, "lr": 8.701986494586309e-05}
{"train_loss": 0.2460605800151825, "global_step": 64075, "epoch": 719, "lr": 8.701947526211054e-05}
{"train_loss": 0.25196802616119385, "global_step": 64076, "epoch": 719, "lr": 8.701908557338119e-05}
{"train_loss": 0.40218034386634827, "global_step": 64077, "epoch": 719, "lr": 8.701869587967506e-05}
{"train_loss": 0.16898977756500244, "global_step": 64078, "epoch": 719, "lr": 8.701830618099222e-05}
{"train_loss": 0.2386137767789069, "global_step": 64079, "epoch": 719, "lr": 8.701791647733273e-05, "val_loss": 3.553030014038086}
{"train_loss": 0.22872783243656158, "global_step": 64080, "epoch": 720, "lr": 8.701752676869664e-05}
{"train_loss": 0.17109961807727814, "global_step": 64081, "epoch": 720, "lr": 8.701713705508397e-05}
{"train_loss": 0.1868826150894165, "global_step": 64082, "epoch": 720, "lr": 8.701674733649481e-05}
{"train_loss": 0.1717093288898468, "global_step": 64083, "epoch": 720, "lr": 8.701635761292919e-05}
{"train_loss": 0.18084940314292908, "global_step": 64084, "epoch": 720, "lr": 8.70159678843872e-05}
{"train_loss": 0.23997505009174347, "global_step": 64085, "epoch": 720, "lr": 8.701557815086882e-05}
{"train_loss": 0.14338889718055725, "global_step": 64086, "epoch": 720, "lr": 8.701518841237418e-05}
{"train_loss": 0.22512273490428925, "global_step": 64087, "epoch": 720, "lr": 8.70147986689033e-05}
{"train_loss": 0.30319684743881226, "global_step": 64088, "epoch": 720, "lr": 8.701440892045622e-05}
{"train_loss": 0.23127587139606476, "global_step": 64089, "epoch": 720, "lr": 8.7014019167033e-05}
{"train_loss": 0.23173116147518158, "global_step": 64090, "epoch": 720, "lr": 8.701362940863371e-05}
{"train_loss": 0.26729997992515564, "global_step": 64091, "epoch": 720, "lr": 8.701323964525836e-05}
{"train_loss": 0.1625765860080719, "global_step": 64092, "epoch": 720, "lr": 8.701284987690705e-05}
{"train_loss": 0.14286907017230988, "global_step": 64093, "epoch": 720, "lr": 8.701246010357983e-05}
{"train_loss": 0.28017574548721313, "global_step": 64094, "epoch": 720, "lr": 8.701207032527672e-05}
{"train_loss": 0.14176064729690552, "global_step": 64095, "epoch": 720, "lr": 8.70116805419978e-05}
{"train_loss": 0.3201357424259186, "global_step": 64096, "epoch": 720, "lr": 8.701129075374307e-05}
{"train_loss": 0.18228434026241302, "global_step": 64097, "epoch": 720, "lr": 8.701090096051268e-05}
{"train_loss": 0.23638373613357544, "global_step": 64098, "epoch": 720, "lr": 8.701051116230658e-05}
{"train_loss": 0.30036309361457825, "global_step": 64099, "epoch": 720, "lr": 8.701012135912489e-05}
{"train_loss": 0.1631777584552765, "global_step": 64100, "epoch": 720, "lr": 8.700973155096764e-05}
{"train_loss": 0.23015443980693817, "global_step": 64101, "epoch": 720, "lr": 8.700934173783488e-05}
{"train_loss": 0.14259830117225647, "global_step": 64102, "epoch": 720, "lr": 8.700895191972667e-05}
{"train_loss": 0.17446675896644592, "global_step": 64103, "epoch": 720, "lr": 8.700856209664305e-05}
{"train_loss": 0.2154838889837265, "global_step": 64104, "epoch": 720, "lr": 8.700817226858407e-05}
{"train_loss": 0.27172017097473145, "global_step": 64105, "epoch": 720, "lr": 8.700778243554981e-05}
{"train_loss": 0.3438543379306793, "global_step": 64106, "epoch": 720, "lr": 8.70073925975403e-05}
{"train_loss": 0.15303181111812592, "global_step": 64107, "epoch": 720, "lr": 8.700700275455559e-05}
{"train_loss": 0.1623271107673645, "global_step": 64108, "epoch": 720, "lr": 8.700661290659576e-05}
{"train_loss": 0.15907087922096252, "global_step": 64109, "epoch": 720, "lr": 8.700622305366081e-05}
{"train_loss": 0.2795529067516327, "global_step": 64110, "epoch": 720, "lr": 8.700583319575084e-05}
{"train_loss": 0.18021339178085327, "global_step": 64111, "epoch": 720, "lr": 8.700544333286589e-05}
{"train_loss": 0.22242501378059387, "global_step": 64112, "epoch": 720, "lr": 8.700505346500601e-05}
{"train_loss": 0.2266732007265091, "global_step": 64113, "epoch": 720, "lr": 8.700466359217125e-05}
{"train_loss": 0.34115445613861084, "global_step": 64114, "epoch": 720, "lr": 8.700427371436166e-05}
{"train_loss": 0.26583805680274963, "global_step": 64115, "epoch": 720, "lr": 8.70038838315773e-05}
{"train_loss": 0.2572348117828369, "global_step": 64116, "epoch": 720, "lr": 8.700349394381821e-05}
{"train_loss": 0.1958070993423462, "global_step": 64117, "epoch": 720, "lr": 8.700310405108447e-05}
{"train_loss": 0.19110237061977386, "global_step": 64118, "epoch": 720, "lr": 8.70027141533761e-05}
{"train_loss": 0.18628627061843872, "global_step": 64119, "epoch": 720, "lr": 8.700232425069317e-05}
{"train_loss": 0.2039663940668106, "global_step": 64120, "epoch": 720, "lr": 8.700193434303572e-05}
{"train_loss": 0.252284437417984, "global_step": 64121, "epoch": 720, "lr": 8.700154443040382e-05}
{"train_loss": 0.21715286374092102, "global_step": 64122, "epoch": 720, "lr": 8.700115451279752e-05}
{"train_loss": 0.28175032138824463, "global_step": 64123, "epoch": 720, "lr": 8.700076459021685e-05}
{"train_loss": 0.15109777450561523, "global_step": 64124, "epoch": 720, "lr": 8.700037466266188e-05}
{"train_loss": 0.2001495510339737, "global_step": 64125, "epoch": 720, "lr": 8.699998473013268e-05}
{"train_loss": 0.2172212302684784, "global_step": 64126, "epoch": 720, "lr": 8.699959479262926e-05}
{"train_loss": 0.22336778044700623, "global_step": 64127, "epoch": 720, "lr": 8.699920485015172e-05}
{"train_loss": 0.22452089190483093, "global_step": 64128, "epoch": 720, "lr": 8.699881490270007e-05}
{"train_loss": 0.24545593559741974, "global_step": 64129, "epoch": 720, "lr": 8.699842495027437e-05}
{"train_loss": 0.23695874214172363, "global_step": 64130, "epoch": 720, "lr": 8.699803499287471e-05}
{"train_loss": 0.22592219710350037, "global_step": 64131, "epoch": 720, "lr": 8.699764503050109e-05}
{"train_loss": 0.23770302534103394, "global_step": 64132, "epoch": 720, "lr": 8.69972550631536e-05}
{"train_loss": 0.217800110578537, "global_step": 64133, "epoch": 720, "lr": 8.699686509083228e-05}
{"train_loss": 0.1711941808462143, "global_step": 64134, "epoch": 720, "lr": 8.69964751135372e-05}
{"train_loss": 0.17192380130290985, "global_step": 64135, "epoch": 720, "lr": 8.699608513126836e-05}
{"train_loss": 0.21770280599594116, "global_step": 64136, "epoch": 720, "lr": 8.699569514402585e-05}
{"train_loss": 0.2382521778345108, "global_step": 64137, "epoch": 720, "lr": 8.699530515180975e-05}
{"train_loss": 0.24080125987529755, "global_step": 64138, "epoch": 720, "lr": 8.699491515462005e-05}
{"train_loss": 0.2074924111366272, "global_step": 64139, "epoch": 720, "lr": 8.699452515245685e-05}
{"train_loss": 0.16534920036792755, "global_step": 64140, "epoch": 720, "lr": 8.69941351453202e-05}
{"train_loss": 0.1696508228778839, "global_step": 64141, "epoch": 720, "lr": 8.699374513321012e-05}
{"train_loss": 0.262890487909317, "global_step": 64142, "epoch": 720, "lr": 8.69933551161267e-05}
{"train_loss": 0.20631346106529236, "global_step": 64143, "epoch": 720, "lr": 8.699296509406995e-05}
{"train_loss": 0.15060797333717346, "global_step": 64144, "epoch": 720, "lr": 8.699257506703996e-05}
{"train_loss": 0.255610853433609, "global_step": 64145, "epoch": 720, "lr": 8.699218503503677e-05}
{"train_loss": 0.21563060581684113, "global_step": 64146, "epoch": 720, "lr": 8.699179499806043e-05}
{"train_loss": 0.18120524287223816, "global_step": 64147, "epoch": 720, "lr": 8.699140495611098e-05}
{"train_loss": 0.2457147091627121, "global_step": 64148, "epoch": 720, "lr": 8.69910149091885e-05}
{"train_loss": 0.25118255615234375, "global_step": 64149, "epoch": 720, "lr": 8.699062485729303e-05}
{"train_loss": 0.1691824346780777, "global_step": 64150, "epoch": 720, "lr": 8.699023480042463e-05}
{"train_loss": 0.17224864661693573, "global_step": 64151, "epoch": 720, "lr": 8.698984473858334e-05}
{"train_loss": 0.285113662481308, "global_step": 64152, "epoch": 720, "lr": 8.69894546717692e-05}
{"train_loss": 0.25828760862350464, "global_step": 64153, "epoch": 720, "lr": 8.698906459998228e-05}
{"train_loss": 0.23240892589092255, "global_step": 64154, "epoch": 720, "lr": 8.698867452322265e-05}
{"train_loss": 0.16242718696594238, "global_step": 64155, "epoch": 720, "lr": 8.698828444149034e-05}
{"train_loss": 0.22949548065662384, "global_step": 64156, "epoch": 720, "lr": 8.698789435478539e-05}
{"train_loss": 0.24132779240608215, "global_step": 64157, "epoch": 720, "lr": 8.698750426310789e-05}
{"train_loss": 0.22787658870220184, "global_step": 64158, "epoch": 720, "lr": 8.698711416645785e-05}
{"train_loss": 0.26211273670196533, "global_step": 64159, "epoch": 720, "lr": 8.698672406483535e-05}
{"train_loss": 0.2558642625808716, "global_step": 64160, "epoch": 720, "lr": 8.698633395824044e-05}
{"train_loss": 0.18844246864318848, "global_step": 64161, "epoch": 720, "lr": 8.698594384667316e-05}
{"train_loss": 0.2386380136013031, "global_step": 64162, "epoch": 720, "lr": 8.698555373013359e-05}
{"train_loss": 0.2758360505104065, "global_step": 64163, "epoch": 720, "lr": 8.698516360862175e-05}
{"train_loss": 0.17634862661361694, "global_step": 64164, "epoch": 720, "lr": 8.698477348213769e-05}
{"train_loss": 0.27766990661621094, "global_step": 64165, "epoch": 720, "lr": 8.698438335068151e-05}
{"train_loss": 0.2377048283815384, "global_step": 64166, "epoch": 720, "lr": 8.698399321425321e-05}
{"train_loss": 0.30383357405662537, "global_step": 64167, "epoch": 720, "lr": 8.698360307285286e-05}
{"train_loss": 0.2213762735717752, "global_step": 64168, "epoch": 720, "lr": 8.698321292648053e-05, "val_loss": 3.624474287033081, "train_action_mse_error": 28.987035751342773}
{"train_loss": 0.20736198127269745, "global_step": 64169, "epoch": 721, "lr": 8.698282277513625e-05}
{"train_loss": 0.20958441495895386, "global_step": 64170, "epoch": 721, "lr": 8.698243261882007e-05}
{"train_loss": 0.1788477599620819, "global_step": 64171, "epoch": 721, "lr": 8.698204245753207e-05}
{"train_loss": 0.22882957756519318, "global_step": 64172, "epoch": 721, "lr": 8.698165229127228e-05}
{"train_loss": 0.24675457179546356, "global_step": 64173, "epoch": 721, "lr": 8.698126212004075e-05}
{"train_loss": 0.2217247635126114, "global_step": 64174, "epoch": 721, "lr": 8.698087194383755e-05}
{"train_loss": 0.2179381251335144, "global_step": 64175, "epoch": 721, "lr": 8.698048176266272e-05}
{"train_loss": 0.22470629215240479, "global_step": 64176, "epoch": 721, "lr": 8.698009157651629e-05}
{"train_loss": 0.20501668751239777, "global_step": 64177, "epoch": 721, "lr": 8.697970138539838e-05}
{"train_loss": 0.22551549971103668, "global_step": 64178, "epoch": 721, "lr": 8.697931118930898e-05}
{"train_loss": 0.30292391777038574, "global_step": 64179, "epoch": 721, "lr": 8.697892098824816e-05}
{"train_loss": 0.22209089994430542, "global_step": 64180, "epoch": 721, "lr": 8.697853078221598e-05}
{"train_loss": 0.1344956010580063, "global_step": 64181, "epoch": 721, "lr": 8.697814057121247e-05}
{"train_loss": 0.19590318202972412, "global_step": 64182, "epoch": 721, "lr": 8.697775035523772e-05}
{"train_loss": 0.19639600813388824, "global_step": 64183, "epoch": 721, "lr": 8.697736013429176e-05}
{"train_loss": 0.175008624792099, "global_step": 64184, "epoch": 721, "lr": 8.697696990837465e-05}
{"train_loss": 0.2704600989818573, "global_step": 64185, "epoch": 721, "lr": 8.697657967748642e-05}
{"train_loss": 0.31090959906578064, "global_step": 64186, "epoch": 721, "lr": 8.697618944162716e-05}
{"train_loss": 0.2164333611726761, "global_step": 64187, "epoch": 721, "lr": 8.697579920079688e-05}
{"train_loss": 0.3516363501548767, "global_step": 64188, "epoch": 721, "lr": 8.697540895499567e-05}
{"train_loss": 0.2815980315208435, "global_step": 64189, "epoch": 721, "lr": 8.697501870422355e-05}
{"train_loss": 0.19791243970394135, "global_step": 64190, "epoch": 721, "lr": 8.69746284484806e-05}
{"train_loss": 0.2209455370903015, "global_step": 64191, "epoch": 721, "lr": 8.697423818776687e-05}
{"train_loss": 0.30724936723709106, "global_step": 64192, "epoch": 721, "lr": 8.697384792208238e-05}
{"train_loss": 0.24004602432250977, "global_step": 64193, "epoch": 721, "lr": 8.697345765142723e-05}
{"train_loss": 0.3854018449783325, "global_step": 64194, "epoch": 721, "lr": 8.697306737580144e-05}
{"train_loss": 0.15910033881664276, "global_step": 64195, "epoch": 721, "lr": 8.697267709520508e-05}
{"train_loss": 0.29716160893440247, "global_step": 64196, "epoch": 721, "lr": 8.697228680963818e-05}
{"train_loss": 0.28288623690605164, "global_step": 64197, "epoch": 721, "lr": 8.697189651910082e-05}
{"train_loss": 0.23803727328777313, "global_step": 64198, "epoch": 721, "lr": 8.697150622359305e-05}
{"train_loss": 0.20813293755054474, "global_step": 64199, "epoch": 721, "lr": 8.697111592311489e-05}
{"train_loss": 0.15549826622009277, "global_step": 64200, "epoch": 721, "lr": 8.697072561766642e-05}
{"train_loss": 0.3072083294391632, "global_step": 64201, "epoch": 721, "lr": 8.697033530724767e-05}
{"train_loss": 0.1692475825548172, "global_step": 64202, "epoch": 721, "lr": 8.696994499185873e-05}
{"train_loss": 0.2383956015110016, "global_step": 64203, "epoch": 721, "lr": 8.696955467149964e-05}
{"train_loss": 0.27626708149909973, "global_step": 64204, "epoch": 721, "lr": 8.696916434617042e-05}
{"train_loss": 0.22201763093471527, "global_step": 64205, "epoch": 721, "lr": 8.696877401587116e-05}
{"train_loss": 0.23288461565971375, "global_step": 64206, "epoch": 721, "lr": 8.69683836806019e-05}
{"train_loss": 0.3046000003814697, "global_step": 64207, "epoch": 721, "lr": 8.696799334036268e-05}
{"train_loss": 0.27059245109558105, "global_step": 64208, "epoch": 721, "lr": 8.696760299515359e-05}
{"train_loss": 0.2897964119911194, "global_step": 64209, "epoch": 721, "lr": 8.696721264497462e-05}
{"train_loss": 0.13484454154968262, "global_step": 64210, "epoch": 721, "lr": 8.696682228982589e-05}
{"train_loss": 0.3634362816810608, "global_step": 64211, "epoch": 721, "lr": 8.696643192970741e-05}
{"train_loss": 0.2806491255760193, "global_step": 64212, "epoch": 721, "lr": 8.696604156461923e-05}
{"train_loss": 0.2247265875339508, "global_step": 64213, "epoch": 721, "lr": 8.696565119456144e-05}
{"train_loss": 0.21354492008686066, "global_step": 64214, "epoch": 721, "lr": 8.696526081953404e-05}
{"train_loss": 0.2468547224998474, "global_step": 64215, "epoch": 721, "lr": 8.696487043953714e-05}
{"train_loss": 0.2337556779384613, "global_step": 64216, "epoch": 721, "lr": 8.696448005457075e-05}
{"train_loss": 0.17410214245319366, "global_step": 64217, "epoch": 721, "lr": 8.696408966463495e-05}
{"train_loss": 0.18872252106666565, "global_step": 64218, "epoch": 721, "lr": 8.696369926972975e-05}
{"train_loss": 0.34113797545433044, "global_step": 64219, "epoch": 721, "lr": 8.696330886985526e-05}
{"train_loss": 0.2478039562702179, "global_step": 64220, "epoch": 721, "lr": 8.696291846501148e-05}
{"train_loss": 0.18086761236190796, "global_step": 64221, "epoch": 721, "lr": 8.69625280551985e-05}
{"train_loss": 0.1788787692785263, "global_step": 64222, "epoch": 721, "lr": 8.696213764041637e-05}
{"train_loss": 0.13962934911251068, "global_step": 64223, "epoch": 721, "lr": 8.69617472206651e-05}
{"train_loss": 0.1344466358423233, "global_step": 64224, "epoch": 721, "lr": 8.69613567959448e-05}
{"train_loss": 0.28887373208999634, "global_step": 64225, "epoch": 721, "lr": 8.696096636625548e-05}
{"train_loss": 0.2569803297519684, "global_step": 64226, "epoch": 721, "lr": 8.696057593159722e-05}
{"train_loss": 0.2148430049419403, "global_step": 64227, "epoch": 721, "lr": 8.696018549197004e-05}
{"train_loss": 0.2050597369670868, "global_step": 64228, "epoch": 721, "lr": 8.695979504737405e-05}
{"train_loss": 0.25587335228919983, "global_step": 64229, "epoch": 721, "lr": 8.695940459780924e-05}
{"train_loss": 0.3220526874065399, "global_step": 64230, "epoch": 721, "lr": 8.69590141432757e-05}
{"train_loss": 0.2527696192264557, "global_step": 64231, "epoch": 721, "lr": 8.695862368377345e-05}
{"train_loss": 0.14956305921077728, "global_step": 64232, "epoch": 721, "lr": 8.695823321930258e-05}
{"train_loss": 0.25657975673675537, "global_step": 64233, "epoch": 721, "lr": 8.695784274986314e-05}
{"train_loss": 0.221999853849411, "global_step": 64234, "epoch": 721, "lr": 8.695745227545515e-05}
{"train_loss": 0.288675993680954, "global_step": 64235, "epoch": 721, "lr": 8.695706179607867e-05}
{"train_loss": 0.22131094336509705, "global_step": 64236, "epoch": 721, "lr": 8.69566713117338e-05}
{"train_loss": 0.189117431640625, "global_step": 64237, "epoch": 721, "lr": 8.695628082242053e-05}
{"train_loss": 0.20839792490005493, "global_step": 64238, "epoch": 721, "lr": 8.695589032813893e-05}
{"train_loss": 0.23390750586986542, "global_step": 64239, "epoch": 721, "lr": 8.695549982888909e-05}
{"train_loss": 0.15752682089805603, "global_step": 64240, "epoch": 721, "lr": 8.695510932467102e-05}
{"train_loss": 0.30323970317840576, "global_step": 64241, "epoch": 721, "lr": 8.695471881548478e-05}
{"train_loss": 0.18398627638816833, "global_step": 64242, "epoch": 721, "lr": 8.695432830133042e-05}
{"train_loss": 0.18712367117404938, "global_step": 64243, "epoch": 721, "lr": 8.695393778220801e-05}
{"train_loss": 0.23361781239509583, "global_step": 64244, "epoch": 721, "lr": 8.69535472581176e-05}
{"train_loss": 0.24837762117385864, "global_step": 64245, "epoch": 721, "lr": 8.695315672905923e-05}
{"train_loss": 0.20815856754779816, "global_step": 64246, "epoch": 721, "lr": 8.695276619503296e-05}
{"train_loss": 0.17733676731586456, "global_step": 64247, "epoch": 721, "lr": 8.695237565603885e-05}
{"train_loss": 0.25836068391799927, "global_step": 64248, "epoch": 721, "lr": 8.695198511207694e-05}
{"train_loss": 0.23494185507297516, "global_step": 64249, "epoch": 721, "lr": 8.695159456314728e-05}
{"train_loss": 0.21687979996204376, "global_step": 64250, "epoch": 721, "lr": 8.695120400924992e-05}
{"train_loss": 0.2711021602153778, "global_step": 64251, "epoch": 721, "lr": 8.695081345038494e-05}
{"train_loss": 0.27382194995880127, "global_step": 64252, "epoch": 721, "lr": 8.695042288655235e-05}
{"train_loss": 0.22136607766151428, "global_step": 64253, "epoch": 721, "lr": 8.695003231775225e-05}
{"train_loss": 0.2662952244281769, "global_step": 64254, "epoch": 721, "lr": 8.694964174398465e-05}
{"train_loss": 0.14303921163082123, "global_step": 64255, "epoch": 721, "lr": 8.694925116524963e-05}
{"train_loss": 0.2267141342163086, "global_step": 64256, "epoch": 721, "lr": 8.694886058154725e-05}
{"train_loss": 0.2336999047338293, "global_step": 64257, "epoch": 721, "lr": 8.694846999287751e-05, "val_loss": 3.6057987213134766}
{"train_loss": 0.25604355335235596, "global_step": 64258, "epoch": 722, "lr": 8.694807939924053e-05}
{"train_loss": 0.20465275645256042, "global_step": 64259, "epoch": 722, "lr": 8.694768880063631e-05}
{"train_loss": 0.21298113465309143, "global_step": 64260, "epoch": 722, "lr": 8.694729819706493e-05}
{"train_loss": 0.28197500109672546, "global_step": 64261, "epoch": 722, "lr": 8.694690758852643e-05}
{"train_loss": 0.2097722291946411, "global_step": 64262, "epoch": 722, "lr": 8.69465169750209e-05}
{"train_loss": 0.22198662161827087, "global_step": 64263, "epoch": 722, "lr": 8.694612635654833e-05}
{"train_loss": 0.346842497587204, "global_step": 64264, "epoch": 722, "lr": 8.69457357331088e-05}
{"train_loss": 0.20156775414943695, "global_step": 64265, "epoch": 722, "lr": 8.694534510470239e-05}
{"train_loss": 0.18865980207920074, "global_step": 64266, "epoch": 722, "lr": 8.694495447132911e-05}
{"train_loss": 0.3053843677043915, "global_step": 64267, "epoch": 722, "lr": 8.694456383298905e-05}
{"train_loss": 0.2203071117401123, "global_step": 64268, "epoch": 722, "lr": 8.694417318968223e-05}
{"train_loss": 0.20140910148620605, "global_step": 64269, "epoch": 722, "lr": 8.694378254140872e-05}
{"train_loss": 0.3142818212509155, "global_step": 64270, "epoch": 722, "lr": 8.694339188816856e-05}
{"train_loss": 0.26692935824394226, "global_step": 64271, "epoch": 722, "lr": 8.694300122996182e-05}
{"train_loss": 0.27570417523384094, "global_step": 64272, "epoch": 722, "lr": 8.694261056678854e-05}
{"train_loss": 0.14411155879497528, "global_step": 64273, "epoch": 722, "lr": 8.694221989864876e-05}
{"train_loss": 0.14400257170200348, "global_step": 64274, "epoch": 722, "lr": 8.694182922554258e-05}
{"train_loss": 0.28748735785484314, "global_step": 64275, "epoch": 722, "lr": 8.694143854747e-05}
{"train_loss": 0.21264100074768066, "global_step": 64276, "epoch": 722, "lr": 8.694104786443111e-05}
{"train_loss": 0.1379050463438034, "global_step": 64277, "epoch": 722, "lr": 8.694065717642594e-05}
{"train_loss": 0.19972527027130127, "global_step": 64278, "epoch": 722, "lr": 8.694026648345454e-05}
{"train_loss": 0.35903865098953247, "global_step": 64279, "epoch": 722, "lr": 8.693987578551699e-05}
{"train_loss": 0.2940399944782257, "global_step": 64280, "epoch": 722, "lr": 8.693948508261331e-05}
{"train_loss": 0.2600492238998413, "global_step": 64281, "epoch": 722, "lr": 8.693909437474356e-05}
{"train_loss": 0.20330816507339478, "global_step": 64282, "epoch": 722, "lr": 8.693870366190782e-05}
{"train_loss": 0.20145970582962036, "global_step": 64283, "epoch": 722, "lr": 8.693831294410611e-05}
{"train_loss": 0.2232559770345688, "global_step": 64284, "epoch": 722, "lr": 8.693792222133849e-05}
{"train_loss": 0.17920295894145966, "global_step": 64285, "epoch": 722, "lr": 8.693753149360501e-05}
{"train_loss": 0.31650152802467346, "global_step": 64286, "epoch": 722, "lr": 8.693714076090574e-05}
{"train_loss": 0.2960890233516693, "global_step": 64287, "epoch": 722, "lr": 8.693675002324072e-05}
{"train_loss": 0.19100797176361084, "global_step": 64288, "epoch": 722, "lr": 8.693635928061e-05}
{"train_loss": 0.3488256335258484, "global_step": 64289, "epoch": 722, "lr": 8.693596853301364e-05}
{"train_loss": 0.3329302668571472, "global_step": 64290, "epoch": 722, "lr": 8.693557778045169e-05}
{"train_loss": 0.22661705315113068, "global_step": 64291, "epoch": 722, "lr": 8.69351870229242e-05}
{"train_loss": 0.19848273694515228, "global_step": 64292, "epoch": 722, "lr": 8.693479626043124e-05}
{"train_loss": 0.35571229457855225, "global_step": 64293, "epoch": 722, "lr": 8.693440549297282e-05}
{"train_loss": 0.17227651178836823, "global_step": 64294, "epoch": 722, "lr": 8.693401472054904e-05}
{"train_loss": 0.24622255563735962, "global_step": 64295, "epoch": 722, "lr": 8.693362394315993e-05}
{"train_loss": 0.32049456238746643, "global_step": 64296, "epoch": 722, "lr": 8.693323316080553e-05}
{"train_loss": 0.2475246638059616, "global_step": 64297, "epoch": 722, "lr": 8.693284237348591e-05}
{"train_loss": 0.21753619611263275, "global_step": 64298, "epoch": 722, "lr": 8.693245158120113e-05}
{"train_loss": 0.2297486662864685, "global_step": 64299, "epoch": 722, "lr": 8.693206078395122e-05}
{"train_loss": 0.3235306441783905, "global_step": 64300, "epoch": 722, "lr": 8.693166998173627e-05}
{"train_loss": 0.22507473826408386, "global_step": 64301, "epoch": 722, "lr": 8.693127917455627e-05}
{"train_loss": 0.22418954968452454, "global_step": 64302, "epoch": 722, "lr": 8.693088836241134e-05}
{"train_loss": 0.21570324897766113, "global_step": 64303, "epoch": 722, "lr": 8.693049754530149e-05}
{"train_loss": 0.19416137039661407, "global_step": 64304, "epoch": 722, "lr": 8.69301067232268e-05}
{"train_loss": 0.304197758436203, "global_step": 64305, "epoch": 722, "lr": 8.692971589618728e-05}
{"train_loss": 0.26224973797798157, "global_step": 64306, "epoch": 722, "lr": 8.692932506418302e-05}
{"train_loss": 0.21242061257362366, "global_step": 64307, "epoch": 722, "lr": 8.692893422721407e-05}
{"train_loss": 0.21407419443130493, "global_step": 64308, "epoch": 722, "lr": 8.692854338528048e-05}
{"train_loss": 0.2660447061061859, "global_step": 64309, "epoch": 722, "lr": 8.692815253838227e-05}
{"train_loss": 0.2426626831293106, "global_step": 64310, "epoch": 722, "lr": 8.692776168651955e-05}
{"train_loss": 0.2617931067943573, "global_step": 64311, "epoch": 722, "lr": 8.692737082969233e-05}
{"train_loss": 0.22383011877536774, "global_step": 64312, "epoch": 722, "lr": 8.692697996790066e-05}
{"train_loss": 0.19889602065086365, "global_step": 64313, "epoch": 722, "lr": 8.692658910114464e-05}
{"train_loss": 0.2767886519432068, "global_step": 64314, "epoch": 722, "lr": 8.692619822942426e-05}
{"train_loss": 0.2215154469013214, "global_step": 64315, "epoch": 722, "lr": 8.692580735273962e-05}
{"train_loss": 0.2162879854440689, "global_step": 64316, "epoch": 722, "lr": 8.692541647109075e-05}
{"train_loss": 0.16866469383239746, "global_step": 64317, "epoch": 722, "lr": 8.69250255844777e-05}
{"train_loss": 0.3262142241001129, "global_step": 64318, "epoch": 722, "lr": 8.692463469290055e-05}
{"train_loss": 0.2781420946121216, "global_step": 64319, "epoch": 722, "lr": 8.692424379635932e-05}
{"train_loss": 0.2259722203016281, "global_step": 64320, "epoch": 722, "lr": 8.692385289485407e-05}
{"train_loss": 0.24256333708763123, "global_step": 64321, "epoch": 722, "lr": 8.692346198838487e-05}
{"train_loss": 0.27275335788726807, "global_step": 64322, "epoch": 722, "lr": 8.692307107695176e-05}
{"train_loss": 0.3122018277645111, "global_step": 64323, "epoch": 722, "lr": 8.692268016055478e-05}
{"train_loss": 0.20125606656074524, "global_step": 64324, "epoch": 722, "lr": 8.692228923919402e-05}
{"train_loss": 0.23987014591693878, "global_step": 64325, "epoch": 722, "lr": 8.692189831286947e-05}
{"train_loss": 0.1730438470840454, "global_step": 64326, "epoch": 722, "lr": 8.692150738158125e-05}
{"train_loss": 0.2309129685163498, "global_step": 64327, "epoch": 722, "lr": 8.692111644532939e-05}
{"train_loss": 0.2100662738084793, "global_step": 64328, "epoch": 722, "lr": 8.692072550411393e-05}
{"train_loss": 0.22409333288669586, "global_step": 64329, "epoch": 722, "lr": 8.692033455793492e-05}
{"train_loss": 0.21630313992500305, "global_step": 64330, "epoch": 722, "lr": 8.69199436067924e-05}
{"train_loss": 0.2217024266719818, "global_step": 64331, "epoch": 722, "lr": 8.691955265068647e-05}
{"train_loss": 0.27973413467407227, "global_step": 64332, "epoch": 722, "lr": 8.691916168961715e-05}
{"train_loss": 0.18896503746509552, "global_step": 64333, "epoch": 722, "lr": 8.69187707235845e-05}
{"train_loss": 0.2687889039516449, "global_step": 64334, "epoch": 722, "lr": 8.691837975258858e-05}
{"train_loss": 0.21587319672107697, "global_step": 64335, "epoch": 722, "lr": 8.691798877662942e-05}
{"train_loss": 0.2703099846839905, "global_step": 64336, "epoch": 722, "lr": 8.69175977957071e-05}
{"train_loss": 0.2747599482536316, "global_step": 64337, "epoch": 722, "lr": 8.691720680982165e-05}
{"train_loss": 0.2716902196407318, "global_step": 64338, "epoch": 722, "lr": 8.691681581897313e-05}
{"train_loss": 0.21908682584762573, "global_step": 64339, "epoch": 722, "lr": 8.69164248231616e-05}
{"train_loss": 0.21661034226417542, "global_step": 64340, "epoch": 722, "lr": 8.69160338223871e-05}
{"train_loss": 0.20555052161216736, "global_step": 64341, "epoch": 722, "lr": 8.69156428166497e-05}
{"train_loss": 0.2409900426864624, "global_step": 64342, "epoch": 722, "lr": 8.691525180594943e-05}
{"train_loss": 0.18975096940994263, "global_step": 64343, "epoch": 722, "lr": 8.691486079028636e-05}
{"train_loss": 0.2104879468679428, "global_step": 64344, "epoch": 722, "lr": 8.691446976966054e-05}
{"train_loss": 0.31467825174331665, "global_step": 64345, "epoch": 722, "lr": 8.6914078744072e-05}
{"train_loss": 0.24149909381116375, "global_step": 64346, "epoch": 722, "lr": 8.691368771352084e-05, "val_loss": 3.5010972023010254}
{"train_loss": 0.1266627311706543, "global_step": 64347, "epoch": 723, "lr": 8.691329667800707e-05}
{"train_loss": 0.1939423829317093, "global_step": 64348, "epoch": 723, "lr": 8.691290563753076e-05}
{"train_loss": 0.2862081825733185, "global_step": 64349, "epoch": 723, "lr": 8.691251459209196e-05}
{"train_loss": 0.1829911470413208, "global_step": 64350, "epoch": 723, "lr": 8.691212354169071e-05}
{"train_loss": 0.18942028284072876, "global_step": 64351, "epoch": 723, "lr": 8.69117324863271e-05}
{"train_loss": 0.18720264732837677, "global_step": 64352, "epoch": 723, "lr": 8.691134142600113e-05}
{"train_loss": 0.21828673779964447, "global_step": 64353, "epoch": 723, "lr": 8.691095036071289e-05}
{"train_loss": 0.24142035841941833, "global_step": 64354, "epoch": 723, "lr": 8.691055929046243e-05}
{"train_loss": 0.2503790557384491, "global_step": 64355, "epoch": 723, "lr": 8.69101682152498e-05}
{"train_loss": 0.210357204079628, "global_step": 64356, "epoch": 723, "lr": 8.690977713507503e-05}
{"train_loss": 0.19474776089191437, "global_step": 64357, "epoch": 723, "lr": 8.69093860499382e-05}
{"train_loss": 0.192280575633049, "global_step": 64358, "epoch": 723, "lr": 8.690899495983935e-05}
{"train_loss": 0.36620035767555237, "global_step": 64359, "epoch": 723, "lr": 8.690860386477855e-05}
{"train_loss": 0.2072496861219406, "global_step": 64360, "epoch": 723, "lr": 8.690821276475581e-05}
{"train_loss": 0.23741638660430908, "global_step": 64361, "epoch": 723, "lr": 8.690782165977123e-05}
{"train_loss": 0.28032219409942627, "global_step": 64362, "epoch": 723, "lr": 8.690743054982484e-05}
{"train_loss": 0.21160176396369934, "global_step": 64363, "epoch": 723, "lr": 8.690703943491669e-05}
{"train_loss": 0.2230779230594635, "global_step": 64364, "epoch": 723, "lr": 8.690664831504684e-05}
{"train_loss": 0.23948447406291962, "global_step": 64365, "epoch": 723, "lr": 8.690625719021536e-05}
{"train_loss": 0.2427314966917038, "global_step": 64366, "epoch": 723, "lr": 8.690586606042225e-05}
{"train_loss": 0.20337392389774323, "global_step": 64367, "epoch": 723, "lr": 8.690547492566764e-05}
{"train_loss": 0.18233622610569, "global_step": 64368, "epoch": 723, "lr": 8.69050837859515e-05}
{"train_loss": 0.20407254993915558, "global_step": 64369, "epoch": 723, "lr": 8.690469264127393e-05}
{"train_loss": 0.18968918919563293, "global_step": 64370, "epoch": 723, "lr": 8.690430149163498e-05}
{"train_loss": 0.2565346956253052, "global_step": 64371, "epoch": 723, "lr": 8.69039103370347e-05}
{"train_loss": 0.1967165172100067, "global_step": 64372, "epoch": 723, "lr": 8.690351917747313e-05}
{"train_loss": 0.24977938830852509, "global_step": 64373, "epoch": 723, "lr": 8.690312801295033e-05}
{"train_loss": 0.26953020691871643, "global_step": 64374, "epoch": 723, "lr": 8.690273684346637e-05}
{"train_loss": 0.2344517707824707, "global_step": 64375, "epoch": 723, "lr": 8.690234566902127e-05}
{"train_loss": 0.23718713223934174, "global_step": 64376, "epoch": 723, "lr": 8.690195448961511e-05}
{"train_loss": 0.15713484585285187, "global_step": 64377, "epoch": 723, "lr": 8.690156330524792e-05}
{"train_loss": 0.17267924547195435, "global_step": 64378, "epoch": 723, "lr": 8.690117211591977e-05}
{"train_loss": 0.15314102172851562, "global_step": 64379, "epoch": 723, "lr": 8.690078092163072e-05}
{"train_loss": 0.19180849194526672, "global_step": 64380, "epoch": 723, "lr": 8.690038972238079e-05}
{"train_loss": 0.19567853212356567, "global_step": 64381, "epoch": 723, "lr": 8.689999851817007e-05}
{"train_loss": 0.18576586246490479, "global_step": 64382, "epoch": 723, "lr": 8.689960730899858e-05}
{"train_loss": 0.1692633032798767, "global_step": 64383, "epoch": 723, "lr": 8.68992160948664e-05}
{"train_loss": 0.2517751157283783, "global_step": 64384, "epoch": 723, "lr": 8.689882487577357e-05}
{"train_loss": 0.20815934240818024, "global_step": 64385, "epoch": 723, "lr": 8.689843365172012e-05}
{"train_loss": 0.19087283313274384, "global_step": 64386, "epoch": 723, "lr": 8.689804242270614e-05}
{"train_loss": 0.269508957862854, "global_step": 64387, "epoch": 723, "lr": 8.689765118873167e-05}
{"train_loss": 0.2815901041030884, "global_step": 64388, "epoch": 723, "lr": 8.689725994979677e-05}
{"train_loss": 0.29989197850227356, "global_step": 64389, "epoch": 723, "lr": 8.689686870590147e-05}
{"train_loss": 0.21920271217823029, "global_step": 64390, "epoch": 723, "lr": 8.689647745704584e-05}
{"train_loss": 0.250733345746994, "global_step": 64391, "epoch": 723, "lr": 8.689608620322993e-05}
{"train_loss": 0.3009018301963806, "global_step": 64392, "epoch": 723, "lr": 8.68956949444538e-05}
{"train_loss": 0.1755719929933548, "global_step": 64393, "epoch": 723, "lr": 8.689530368071748e-05}
{"train_loss": 0.1961345225572586, "global_step": 64394, "epoch": 723, "lr": 8.689491241202103e-05}
{"train_loss": 0.18821901082992554, "global_step": 64395, "epoch": 723, "lr": 8.689452113836452e-05}
{"train_loss": 0.24377784132957458, "global_step": 64396, "epoch": 723, "lr": 8.689412985974798e-05}
{"train_loss": 0.12942476570606232, "global_step": 64397, "epoch": 723, "lr": 8.689373857617149e-05}
{"train_loss": 0.12220801413059235, "global_step": 64398, "epoch": 723, "lr": 8.689334728763507e-05}
{"train_loss": 0.21959632635116577, "global_step": 64399, "epoch": 723, "lr": 8.68929559941388e-05}
{"train_loss": 0.31104499101638794, "global_step": 64400, "epoch": 723, "lr": 8.689256469568272e-05}
{"train_loss": 0.21735242009162903, "global_step": 64401, "epoch": 723, "lr": 8.68921733922669e-05}
{"train_loss": 0.25120845437049866, "global_step": 64402, "epoch": 723, "lr": 8.689178208389135e-05}
{"train_loss": 0.2998241186141968, "global_step": 64403, "epoch": 723, "lr": 8.689139077055616e-05}
{"train_loss": 0.1907539814710617, "global_step": 64404, "epoch": 723, "lr": 8.689099945226137e-05}
{"train_loss": 0.2570464611053467, "global_step": 64405, "epoch": 723, "lr": 8.689060812900705e-05}
{"train_loss": 0.19324134290218353, "global_step": 64406, "epoch": 723, "lr": 8.689021680079321e-05}
{"train_loss": 0.2588547170162201, "global_step": 64407, "epoch": 723, "lr": 8.688982546761995e-05}
{"train_loss": 0.23885130882263184, "global_step": 64408, "epoch": 723, "lr": 8.68894341294873e-05}
{"train_loss": 0.27510538697242737, "global_step": 64409, "epoch": 723, "lr": 8.688904278639532e-05}
{"train_loss": 0.21177810430526733, "global_step": 64410, "epoch": 723, "lr": 8.688865143834404e-05}
{"train_loss": 0.22356343269348145, "global_step": 64411, "epoch": 723, "lr": 8.688826008533354e-05}
{"train_loss": 0.18875227868556976, "global_step": 64412, "epoch": 723, "lr": 8.688786872736386e-05}
{"train_loss": 0.24010193347930908, "global_step": 64413, "epoch": 723, "lr": 8.688747736443506e-05}
{"train_loss": 0.20176692306995392, "global_step": 64414, "epoch": 723, "lr": 8.688708599654719e-05}
{"train_loss": 0.20471005141735077, "global_step": 64415, "epoch": 723, "lr": 8.68866946237003e-05}
{"train_loss": 0.216134175658226, "global_step": 64416, "epoch": 723, "lr": 8.688630324589444e-05}
{"train_loss": 0.20509745180606842, "global_step": 64417, "epoch": 723, "lr": 8.688591186312966e-05}
{"train_loss": 0.21731746196746826, "global_step": 64418, "epoch": 723, "lr": 8.688552047540604e-05}
{"train_loss": 0.1691269725561142, "global_step": 64419, "epoch": 723, "lr": 8.688512908272359e-05}
{"train_loss": 0.17834870517253876, "global_step": 64420, "epoch": 723, "lr": 8.68847376850824e-05}
{"train_loss": 0.31866219639778137, "global_step": 64421, "epoch": 723, "lr": 8.68843462824825e-05}
{"train_loss": 0.22788742184638977, "global_step": 64422, "epoch": 723, "lr": 8.688395487492395e-05}
{"train_loss": 0.27282580733299255, "global_step": 64423, "epoch": 723, "lr": 8.68835634624068e-05}
{"train_loss": 0.260940819978714, "global_step": 64424, "epoch": 723, "lr": 8.68831720449311e-05}
{"train_loss": 0.21949703991413116, "global_step": 64425, "epoch": 723, "lr": 8.688278062249691e-05}
{"train_loss": 0.2721558213233948, "global_step": 64426, "epoch": 723, "lr": 8.688238919510427e-05}
{"train_loss": 0.1626284420490265, "global_step": 64427, "epoch": 723, "lr": 8.688199776275326e-05}
{"train_loss": 0.4336422085762024, "global_step": 64428, "epoch": 723, "lr": 8.688160632544391e-05}
{"train_loss": 0.27566972374916077, "global_step": 64429, "epoch": 723, "lr": 8.688121488317629e-05}
{"train_loss": 0.2673969268798828, "global_step": 64430, "epoch": 723, "lr": 8.688082343595042e-05}
{"train_loss": 0.19546517729759216, "global_step": 64431, "epoch": 723, "lr": 8.688043198376637e-05}
{"train_loss": 0.23023980855941772, "global_step": 64432, "epoch": 723, "lr": 8.68800405266242e-05}
{"train_loss": 0.26207607984542847, "global_step": 64433, "epoch": 723, "lr": 8.687964906452398e-05}
{"train_loss": 0.17697127163410187, "global_step": 64434, "epoch": 723, "lr": 8.687925759746572e-05}
{"train_loss": 0.22552331461665337, "global_step": 64435, "epoch": 723, "lr": 8.687886612544948e-05, "val_loss": 3.5112099647521973}
{"train_loss": 0.1591353416442871, "global_step": 64436, "epoch": 724, "lr": 8.687847464847537e-05}
{"train_loss": 0.23299336433410645, "global_step": 64437, "epoch": 724, "lr": 8.687808316654336e-05}
{"train_loss": 0.20851199328899384, "global_step": 64438, "epoch": 724, "lr": 8.687769167965354e-05}
{"train_loss": 0.15570247173309326, "global_step": 64439, "epoch": 724, "lr": 8.687730018780597e-05}
{"train_loss": 0.2707492709159851, "global_step": 64440, "epoch": 724, "lr": 8.687690869100071e-05}
{"train_loss": 0.10499919205904007, "global_step": 64441, "epoch": 724, "lr": 8.687651718923779e-05}
{"train_loss": 0.2667149305343628, "global_step": 64442, "epoch": 724, "lr": 8.687612568251728e-05}
{"train_loss": 0.19363701343536377, "global_step": 64443, "epoch": 724, "lr": 8.68757341708392e-05}
{"train_loss": 0.29605382680892944, "global_step": 64444, "epoch": 724, "lr": 8.687534265420365e-05}
{"train_loss": 0.22823674976825714, "global_step": 64445, "epoch": 724, "lr": 8.687495113261066e-05}
{"train_loss": 0.199347585439682, "global_step": 64446, "epoch": 724, "lr": 8.687455960606028e-05}
{"train_loss": 0.2623189687728882, "global_step": 64447, "epoch": 724, "lr": 8.687416807455256e-05}
{"train_loss": 0.15963250398635864, "global_step": 64448, "epoch": 724, "lr": 8.687377653808756e-05}
{"train_loss": 0.23454071581363678, "global_step": 64449, "epoch": 724, "lr": 8.687338499666533e-05}
{"train_loss": 0.21743491291999817, "global_step": 64450, "epoch": 724, "lr": 8.687299345028592e-05}
{"train_loss": 0.2651263475418091, "global_step": 64451, "epoch": 724, "lr": 8.687260189894939e-05}
{"train_loss": 0.19588594138622284, "global_step": 64452, "epoch": 724, "lr": 8.687221034265577e-05}
{"train_loss": 0.3099893033504486, "global_step": 64453, "epoch": 724, "lr": 8.687181878140516e-05}
{"train_loss": 0.1657012701034546, "global_step": 64454, "epoch": 724, "lr": 8.687142721519756e-05}
{"train_loss": 0.22514888644218445, "global_step": 64455, "epoch": 724, "lr": 8.687103564403306e-05}
{"train_loss": 0.22542549669742584, "global_step": 64456, "epoch": 724, "lr": 8.68706440679117e-05}
{"train_loss": 0.3274973928928375, "global_step": 64457, "epoch": 724, "lr": 8.687025248683352e-05}
{"train_loss": 0.27590224146842957, "global_step": 64458, "epoch": 724, "lr": 8.686986090079859e-05}
{"train_loss": 0.23835721611976624, "global_step": 64459, "epoch": 724, "lr": 8.686946930980696e-05}
{"train_loss": 0.22626231610774994, "global_step": 64460, "epoch": 724, "lr": 8.686907771385868e-05}
{"train_loss": 0.2506038248538971, "global_step": 64461, "epoch": 724, "lr": 8.686868611295379e-05}
{"train_loss": 0.2015734761953354, "global_step": 64462, "epoch": 724, "lr": 8.686829450709235e-05}
{"train_loss": 0.2551240921020508, "global_step": 64463, "epoch": 724, "lr": 8.686790289627445e-05}
{"train_loss": 0.22999818623065948, "global_step": 64464, "epoch": 724, "lr": 8.686751128050008e-05}
{"train_loss": 0.1599399447441101, "global_step": 64465, "epoch": 724, "lr": 8.686711965976932e-05}
{"train_loss": 0.1998186558485031, "global_step": 64466, "epoch": 724, "lr": 8.686672803408224e-05}
{"train_loss": 0.19902852177619934, "global_step": 64467, "epoch": 724, "lr": 8.686633640343888e-05}
{"train_loss": 0.2566158175468445, "global_step": 64468, "epoch": 724, "lr": 8.686594476783928e-05}
{"train_loss": 0.20856718719005585, "global_step": 64469, "epoch": 724, "lr": 8.686555312728351e-05}
{"train_loss": 0.3504626154899597, "global_step": 64470, "epoch": 724, "lr": 8.686516148177161e-05}
{"train_loss": 0.2781205475330353, "global_step": 64471, "epoch": 724, "lr": 8.686476983130364e-05}
{"train_loss": 0.20415939390659332, "global_step": 64472, "epoch": 724, "lr": 8.686437817587966e-05}
{"train_loss": 0.19913740456104279, "global_step": 64473, "epoch": 724, "lr": 8.68639865154997e-05}
{"train_loss": 0.21435292065143585, "global_step": 64474, "epoch": 724, "lr": 8.686359485016383e-05}
{"train_loss": 0.13338246941566467, "global_step": 64475, "epoch": 724, "lr": 8.686320317987211e-05}
{"train_loss": 0.3060920834541321, "global_step": 64476, "epoch": 724, "lr": 8.686281150462456e-05}
{"train_loss": 0.21698996424674988, "global_step": 64477, "epoch": 724, "lr": 8.686241982442128e-05}
{"train_loss": 0.16860798001289368, "global_step": 64478, "epoch": 724, "lr": 8.686202813926228e-05}
{"train_loss": 0.20782017707824707, "global_step": 64479, "epoch": 724, "lr": 8.686163644914764e-05}
{"train_loss": 0.2275744378566742, "global_step": 64480, "epoch": 724, "lr": 8.686124475407739e-05}
{"train_loss": 0.2396204173564911, "global_step": 64481, "epoch": 724, "lr": 8.686085305405161e-05}
{"train_loss": 0.21434950828552246, "global_step": 64482, "epoch": 724, "lr": 8.686046134907031e-05}
{"train_loss": 0.19342082738876343, "global_step": 64483, "epoch": 724, "lr": 8.68600696391336e-05}
{"train_loss": 0.11926711350679398, "global_step": 64484, "epoch": 724, "lr": 8.685967792424149e-05}
{"train_loss": 0.19374117255210876, "global_step": 64485, "epoch": 724, "lr": 8.685928620439405e-05}
{"train_loss": 0.18052254617214203, "global_step": 64486, "epoch": 724, "lr": 8.685889447959132e-05}
{"train_loss": 0.2289048135280609, "global_step": 64487, "epoch": 724, "lr": 8.685850274983336e-05}
{"train_loss": 0.17168186604976654, "global_step": 64488, "epoch": 724, "lr": 8.685811101512023e-05}
{"train_loss": 0.22818493843078613, "global_step": 64489, "epoch": 724, "lr": 8.685771927545198e-05}
{"train_loss": 0.19153904914855957, "global_step": 64490, "epoch": 724, "lr": 8.685732753082866e-05}
{"train_loss": 0.13191308081150055, "global_step": 64491, "epoch": 724, "lr": 8.685693578125031e-05}
{"train_loss": 0.2539280951023102, "global_step": 64492, "epoch": 724, "lr": 8.685654402671701e-05}
{"train_loss": 0.2352706640958786, "global_step": 64493, "epoch": 724, "lr": 8.685615226722878e-05}
{"train_loss": 0.3258533179759979, "global_step": 64494, "epoch": 724, "lr": 8.68557605027857e-05}
{"train_loss": 0.16561394929885864, "global_step": 64495, "epoch": 724, "lr": 8.685536873338781e-05}
{"train_loss": 0.28449156880378723, "global_step": 64496, "epoch": 724, "lr": 8.685497695903516e-05}
{"train_loss": 0.16608424484729767, "global_step": 64497, "epoch": 724, "lr": 8.685458517972783e-05}
{"train_loss": 0.22119180858135223, "global_step": 64498, "epoch": 724, "lr": 8.685419339546581e-05}
{"train_loss": 0.30249831080436707, "global_step": 64499, "epoch": 724, "lr": 8.685380160624922e-05}
{"train_loss": 0.2652553915977478, "global_step": 64500, "epoch": 724, "lr": 8.685340981207807e-05}
{"train_loss": 0.2720828056335449, "global_step": 64501, "epoch": 724, "lr": 8.685301801295245e-05}
{"train_loss": 0.2354508489370346, "global_step": 64502, "epoch": 724, "lr": 8.685262620887238e-05}
{"train_loss": 0.21492093801498413, "global_step": 64503, "epoch": 724, "lr": 8.685223439983792e-05}
{"train_loss": 0.27681487798690796, "global_step": 64504, "epoch": 724, "lr": 8.685184258584913e-05}
{"train_loss": 0.2713426947593689, "global_step": 64505, "epoch": 724, "lr": 8.685145076690606e-05}
{"train_loss": 0.20796498656272888, "global_step": 64506, "epoch": 724, "lr": 8.685105894300876e-05}
{"train_loss": 0.28993627429008484, "global_step": 64507, "epoch": 724, "lr": 8.685066711415727e-05}
{"train_loss": 0.23600472509860992, "global_step": 64508, "epoch": 724, "lr": 8.685027528035169e-05}
{"train_loss": 0.25252023339271545, "global_step": 64509, "epoch": 724, "lr": 8.684988344159201e-05}
{"train_loss": 0.23004670441150665, "global_step": 64510, "epoch": 724, "lr": 8.684949159787833e-05}
{"train_loss": 0.27498558163642883, "global_step": 64511, "epoch": 724, "lr": 8.684909974921068e-05}
{"train_loss": 0.2581441104412079, "global_step": 64512, "epoch": 724, "lr": 8.684870789558912e-05}
{"train_loss": 0.31538155674934387, "global_step": 64513, "epoch": 724, "lr": 8.684831603701369e-05}
{"train_loss": 0.24299581348896027, "global_step": 64514, "epoch": 724, "lr": 8.684792417348446e-05}
{"train_loss": 0.18453346192836761, "global_step": 64515, "epoch": 724, "lr": 8.684753230500147e-05}
{"train_loss": 0.12300869822502136, "global_step": 64516, "epoch": 724, "lr": 8.684714043156479e-05}
{"train_loss": 0.22767458856105804, "global_step": 64517, "epoch": 724, "lr": 8.684674855317444e-05}
{"train_loss": 0.3203207552433014, "global_step": 64518, "epoch": 724, "lr": 8.684635666983053e-05}
{"train_loss": 0.3407340347766876, "global_step": 64519, "epoch": 724, "lr": 8.684596478153304e-05}
{"train_loss": 0.25136134028434753, "global_step": 64520, "epoch": 724, "lr": 8.684557288828207e-05}
{"train_loss": 0.18507753312587738, "global_step": 64521, "epoch": 724, "lr": 8.684518099007767e-05}
{"train_loss": 0.23504093289375305, "global_step": 64522, "epoch": 724, "lr": 8.684478908691988e-05}
{"train_loss": 0.25958240032196045, "global_step": 64523, "epoch": 724, "lr": 8.684439717880877e-05}
{"train_loss": 0.23024817447314103, "global_step": 64524, "epoch": 724, "lr": 8.684400526574435e-05, "val_loss": 3.5670523643493652}
{"train_loss": 0.19968493282794952, "global_step": 64525, "epoch": 725, "lr": 8.684361334772673e-05}
{"train_loss": 0.1798960268497467, "global_step": 64526, "epoch": 725, "lr": 8.684322142475591e-05}
{"train_loss": 0.20008444786071777, "global_step": 64527, "epoch": 725, "lr": 8.6842829496832e-05}
{"train_loss": 0.17575357854366302, "global_step": 64528, "epoch": 725, "lr": 8.6842437563955e-05}
{"train_loss": 0.2377282679080963, "global_step": 64529, "epoch": 725, "lr": 8.6842045626125e-05}
{"train_loss": 0.19347091019153595, "global_step": 64530, "epoch": 725, "lr": 8.684165368334201e-05}
{"train_loss": 0.23671096563339233, "global_step": 64531, "epoch": 725, "lr": 8.684126173560612e-05}
{"train_loss": 0.17426751554012299, "global_step": 64532, "epoch": 725, "lr": 8.684086978291736e-05}
{"train_loss": 0.34303733706474304, "global_step": 64533, "epoch": 725, "lr": 8.684047782527582e-05}
{"train_loss": 0.1776321977376938, "global_step": 64534, "epoch": 725, "lr": 8.684008586268151e-05}
{"train_loss": 0.22101154923439026, "global_step": 64535, "epoch": 725, "lr": 8.683969389513451e-05}
{"train_loss": 0.23410974442958832, "global_step": 64536, "epoch": 725, "lr": 8.683930192263486e-05}
{"train_loss": 0.24097909033298492, "global_step": 64537, "epoch": 725, "lr": 8.68389099451826e-05}
{"train_loss": 0.2950126528739929, "global_step": 64538, "epoch": 725, "lr": 8.683851796277781e-05}
{"train_loss": 0.1523684710264206, "global_step": 64539, "epoch": 725, "lr": 8.683812597542052e-05}
{"train_loss": 0.19564656913280487, "global_step": 64540, "epoch": 725, "lr": 8.683773398311082e-05}
{"train_loss": 0.11226813495159149, "global_step": 64541, "epoch": 725, "lr": 8.683734198584869e-05}
{"train_loss": 0.18213802576065063, "global_step": 64542, "epoch": 725, "lr": 8.683694998363427e-05}
{"train_loss": 0.19597892463207245, "global_step": 64543, "epoch": 725, "lr": 8.683655797646754e-05}
{"train_loss": 0.3204151391983032, "global_step": 64544, "epoch": 725, "lr": 8.683616596434862e-05}
{"train_loss": 0.13427305221557617, "global_step": 64545, "epoch": 725, "lr": 8.68357739472775e-05}
{"train_loss": 0.23682904243469238, "global_step": 64546, "epoch": 725, "lr": 8.683538192525427e-05}
{"train_loss": 0.20792096853256226, "global_step": 64547, "epoch": 725, "lr": 8.683498989827896e-05}
{"train_loss": 0.2194598913192749, "global_step": 64548, "epoch": 725, "lr": 8.683459786635164e-05}
{"train_loss": 0.15284587442874908, "global_step": 64549, "epoch": 725, "lr": 8.683420582947237e-05}
{"train_loss": 0.17468923330307007, "global_step": 64550, "epoch": 725, "lr": 8.683381378764117e-05}
{"train_loss": 0.22745363414287567, "global_step": 64551, "epoch": 725, "lr": 8.683342174085812e-05}
{"train_loss": 0.2281864881515503, "global_step": 64552, "epoch": 725, "lr": 8.683302968912327e-05}
{"train_loss": 0.20333237946033478, "global_step": 64553, "epoch": 725, "lr": 8.683263763243666e-05}
{"train_loss": 0.23115837574005127, "global_step": 64554, "epoch": 725, "lr": 8.683224557079837e-05}
{"train_loss": 0.18081015348434448, "global_step": 64555, "epoch": 725, "lr": 8.683185350420841e-05}
{"train_loss": 0.32058143615722656, "global_step": 64556, "epoch": 725, "lr": 8.683146143266687e-05}
{"train_loss": 0.27029159665107727, "global_step": 64557, "epoch": 725, "lr": 8.683106935617378e-05}
{"train_loss": 0.2881380617618561, "global_step": 64558, "epoch": 725, "lr": 8.68306772747292e-05}
{"train_loss": 0.23845897614955902, "global_step": 64559, "epoch": 725, "lr": 8.68302851883332e-05}
{"train_loss": 0.20045429468154907, "global_step": 64560, "epoch": 725, "lr": 8.68298930969858e-05}
{"train_loss": 0.28827208280563354, "global_step": 64561, "epoch": 725, "lr": 8.682950100068708e-05}
{"train_loss": 0.23094484210014343, "global_step": 64562, "epoch": 725, "lr": 8.682910889943708e-05}
{"train_loss": 0.1927047073841095, "global_step": 64563, "epoch": 725, "lr": 8.682871679323585e-05}
{"train_loss": 0.19817382097244263, "global_step": 64564, "epoch": 725, "lr": 8.682832468208347e-05}
{"train_loss": 0.20380443334579468, "global_step": 64565, "epoch": 725, "lr": 8.682793256597994e-05}
{"train_loss": 0.207951158285141, "global_step": 64566, "epoch": 725, "lr": 8.682754044492536e-05}
{"train_loss": 0.2526302933692932, "global_step": 64567, "epoch": 725, "lr": 8.682714831891976e-05}
{"train_loss": 0.15802861750125885, "global_step": 64568, "epoch": 725, "lr": 8.682675618796321e-05}
{"train_loss": 0.30050456523895264, "global_step": 64569, "epoch": 725, "lr": 8.682636405205573e-05}
{"train_loss": 0.24907515943050385, "global_step": 64570, "epoch": 725, "lr": 8.68259719111974e-05}
{"train_loss": 0.28876566886901855, "global_step": 64571, "epoch": 725, "lr": 8.68255797653883e-05}
{"train_loss": 0.3155442178249359, "global_step": 64572, "epoch": 725, "lr": 8.68251876146284e-05}
{"train_loss": 0.22119981050491333, "global_step": 64573, "epoch": 725, "lr": 8.682479545891785e-05}
{"train_loss": 0.17812037467956543, "global_step": 64574, "epoch": 725, "lr": 8.682440329825663e-05}
{"train_loss": 0.22529104351997375, "global_step": 64575, "epoch": 725, "lr": 8.682401113264481e-05}
{"train_loss": 0.20560471713542938, "global_step": 64576, "epoch": 725, "lr": 8.682361896208247e-05}
{"train_loss": 0.24057388305664062, "global_step": 64577, "epoch": 725, "lr": 8.682322678656962e-05}
{"train_loss": 0.20051562786102295, "global_step": 64578, "epoch": 725, "lr": 8.682283460610636e-05}
{"train_loss": 0.20842283964157104, "global_step": 64579, "epoch": 725, "lr": 8.682244242069271e-05}
{"train_loss": 0.2160218060016632, "global_step": 64580, "epoch": 725, "lr": 8.682205023032872e-05}
{"train_loss": 0.24497045576572418, "global_step": 64581, "epoch": 725, "lr": 8.682165803501446e-05}
{"train_loss": 0.26228636503219604, "global_step": 64582, "epoch": 725, "lr": 8.682126583474999e-05}
{"train_loss": 0.30231815576553345, "global_step": 64583, "epoch": 725, "lr": 8.682087362953533e-05}
{"train_loss": 0.2839515805244446, "global_step": 64584, "epoch": 725, "lr": 8.682048141937055e-05}
{"train_loss": 0.2911526560783386, "global_step": 64585, "epoch": 725, "lr": 8.682008920425573e-05}
{"train_loss": 0.21572187542915344, "global_step": 64586, "epoch": 725, "lr": 8.681969698419089e-05}
{"train_loss": 0.19675055146217346, "global_step": 64587, "epoch": 725, "lr": 8.681930475917608e-05}
{"train_loss": 0.2745172083377838, "global_step": 64588, "epoch": 725, "lr": 8.681891252921137e-05}
{"train_loss": 0.2398027628660202, "global_step": 64589, "epoch": 725, "lr": 8.68185202942968e-05}
{"train_loss": 0.2116260826587677, "global_step": 64590, "epoch": 725, "lr": 8.681812805443244e-05}
{"train_loss": 0.21029457449913025, "global_step": 64591, "epoch": 725, "lr": 8.681773580961832e-05}
{"train_loss": 0.3201752305030823, "global_step": 64592, "epoch": 725, "lr": 8.681734355985451e-05}
{"train_loss": 0.2687450349330902, "global_step": 64593, "epoch": 725, "lr": 8.681695130514106e-05}
{"train_loss": 0.1891917884349823, "global_step": 64594, "epoch": 725, "lr": 8.6816559045478e-05}
{"train_loss": 0.19563700258731842, "global_step": 64595, "epoch": 725, "lr": 8.681616678086542e-05}
{"train_loss": 0.16650046408176422, "global_step": 64596, "epoch": 725, "lr": 8.681577451130336e-05}
{"train_loss": 0.20000694692134857, "global_step": 64597, "epoch": 725, "lr": 8.681538223679186e-05}
{"train_loss": 0.20033009350299835, "global_step": 64598, "epoch": 725, "lr": 8.6814989957331e-05}
{"train_loss": 0.2380429208278656, "global_step": 64599, "epoch": 725, "lr": 8.681459767292078e-05}
{"train_loss": 0.27115440368652344, "global_step": 64600, "epoch": 725, "lr": 8.681420538356131e-05}
{"train_loss": 0.2661133408546448, "global_step": 64601, "epoch": 725, "lr": 8.681381308925261e-05}
{"train_loss": 0.2272040843963623, "global_step": 64602, "epoch": 725, "lr": 8.681342078999472e-05}
{"train_loss": 0.24643346667289734, "global_step": 64603, "epoch": 725, "lr": 8.681302848578774e-05}
{"train_loss": 0.2981586158275604, "global_step": 64604, "epoch": 725, "lr": 8.681263617663169e-05}
{"train_loss": 0.2858443260192871, "global_step": 64605, "epoch": 725, "lr": 8.681224386252664e-05}
{"train_loss": 0.23377259075641632, "global_step": 64606, "epoch": 725, "lr": 8.681185154347262e-05}
{"train_loss": 0.2322569042444229, "global_step": 64607, "epoch": 725, "lr": 8.681145921946971e-05}
{"train_loss": 0.12717494368553162, "global_step": 64608, "epoch": 725, "lr": 8.681106689051794e-05}
{"train_loss": 0.24378514289855957, "global_step": 64609, "epoch": 725, "lr": 8.681067455661735e-05}
{"train_loss": 0.2442450076341629, "global_step": 64610, "epoch": 725, "lr": 8.681028221776804e-05}
{"train_loss": 0.16529430449008942, "global_step": 64611, "epoch": 725, "lr": 8.680988987397003e-05}
{"train_loss": 0.2134426385164261, "global_step": 64612, "epoch": 725, "lr": 8.680949752522339e-05}
{"train_loss": 0.2252681041868885, "global_step": 64613, "epoch": 725, "lr": 8.680910517152813e-05, "val_loss": 3.5751113891601562, "train_action_mse_error": 28.435791015625}
{"train_loss": 0.26118743419647217, "global_step": 64614, "epoch": 726, "lr": 8.680871281288436e-05}
{"train_loss": 0.21652613580226898, "global_step": 64615, "epoch": 726, "lr": 8.68083204492921e-05}
{"train_loss": 0.27431657910346985, "global_step": 64616, "epoch": 726, "lr": 8.68079280807514e-05}
{"train_loss": 0.22643838822841644, "global_step": 64617, "epoch": 726, "lr": 8.680753570726235e-05}
{"train_loss": 0.23432765901088715, "global_step": 64618, "epoch": 726, "lr": 8.680714332882496e-05}
{"train_loss": 0.21754851937294006, "global_step": 64619, "epoch": 726, "lr": 8.680675094543929e-05}
{"train_loss": 0.25219738483428955, "global_step": 64620, "epoch": 726, "lr": 8.68063585571054e-05}
{"train_loss": 0.28395920991897583, "global_step": 64621, "epoch": 726, "lr": 8.680596616382336e-05}
{"train_loss": 0.1610196977853775, "global_step": 64622, "epoch": 726, "lr": 8.680557376559319e-05}
{"train_loss": 0.18731427192687988, "global_step": 64623, "epoch": 726, "lr": 8.680518136241496e-05}
{"train_loss": 0.21946260333061218, "global_step": 64624, "epoch": 726, "lr": 8.680478895428874e-05}
{"train_loss": 0.18901221454143524, "global_step": 64625, "epoch": 726, "lr": 8.680439654121456e-05}
{"train_loss": 0.24365295469760895, "global_step": 64626, "epoch": 726, "lr": 8.680400412319246e-05}
{"train_loss": 0.2533877193927765, "global_step": 64627, "epoch": 726, "lr": 8.680361170022253e-05}
{"train_loss": 0.2105729728937149, "global_step": 64628, "epoch": 726, "lr": 8.68032192723048e-05}
{"train_loss": 0.3131149113178253, "global_step": 64629, "epoch": 726, "lr": 8.680282683943931e-05}
{"train_loss": 0.19567811489105225, "global_step": 64630, "epoch": 726, "lr": 8.680243440162615e-05}
{"train_loss": 0.1959785521030426, "global_step": 64631, "epoch": 726, "lr": 8.680204195886532e-05}
{"train_loss": 0.17384184896945953, "global_step": 64632, "epoch": 726, "lr": 8.680164951115692e-05}
{"train_loss": 0.23546767234802246, "global_step": 64633, "epoch": 726, "lr": 8.680125705850099e-05}
{"train_loss": 0.15387804806232452, "global_step": 64634, "epoch": 726, "lr": 8.680086460089759e-05}
{"train_loss": 0.1336880922317505, "global_step": 64635, "epoch": 726, "lr": 8.680047213834675e-05}
{"train_loss": 0.2488124817609787, "global_step": 64636, "epoch": 726, "lr": 8.680007967084854e-05}
{"train_loss": 0.27744388580322266, "global_step": 64637, "epoch": 726, "lr": 8.679968719840301e-05}
{"train_loss": 0.32706037163734436, "global_step": 64638, "epoch": 726, "lr": 8.67992947210102e-05}
{"train_loss": 0.3067760467529297, "global_step": 64639, "epoch": 726, "lr": 8.679890223867019e-05}
{"train_loss": 0.21888293325901031, "global_step": 64640, "epoch": 726, "lr": 8.6798509751383e-05}
{"train_loss": 0.2561059892177582, "global_step": 64641, "epoch": 726, "lr": 8.67981172591487e-05}
{"train_loss": 0.27299925684928894, "global_step": 64642, "epoch": 726, "lr": 8.679772476196735e-05}
{"train_loss": 0.19873374700546265, "global_step": 64643, "epoch": 726, "lr": 8.679733225983899e-05}
{"train_loss": 0.27438652515411377, "global_step": 64644, "epoch": 726, "lr": 8.679693975276367e-05}
{"train_loss": 0.26440417766571045, "global_step": 64645, "epoch": 726, "lr": 8.679654724074146e-05}
{"train_loss": 0.20169267058372498, "global_step": 64646, "epoch": 726, "lr": 8.679615472377241e-05}
{"train_loss": 0.29779088497161865, "global_step": 64647, "epoch": 726, "lr": 8.679576220185655e-05}
{"train_loss": 0.19981461763381958, "global_step": 64648, "epoch": 726, "lr": 8.679536967499396e-05}
{"train_loss": 0.2996496856212616, "global_step": 64649, "epoch": 726, "lr": 8.679497714318468e-05}
{"train_loss": 0.18378973007202148, "global_step": 64650, "epoch": 726, "lr": 8.679458460642875e-05}
{"train_loss": 0.2674614489078522, "global_step": 64651, "epoch": 726, "lr": 8.679419206472625e-05}
{"train_loss": 0.25715896487236023, "global_step": 64652, "epoch": 726, "lr": 8.67937995180772e-05}
{"train_loss": 0.24696128070354462, "global_step": 64653, "epoch": 726, "lr": 8.67934069664817e-05}
{"train_loss": 0.2585545480251312, "global_step": 64654, "epoch": 726, "lr": 8.679301440993976e-05}
{"train_loss": 0.24573102593421936, "global_step": 64655, "epoch": 726, "lr": 8.679262184845145e-05}
{"train_loss": 0.2822779715061188, "global_step": 64656, "epoch": 726, "lr": 8.679222928201683e-05}
{"train_loss": 0.28320083022117615, "global_step": 64657, "epoch": 726, "lr": 8.679183671063592e-05}
{"train_loss": 0.3121550679206848, "global_step": 64658, "epoch": 726, "lr": 8.67914441343088e-05}
{"train_loss": 0.22600296139717102, "global_step": 64659, "epoch": 726, "lr": 8.679105155303553e-05}
{"train_loss": 0.2645617723464966, "global_step": 64660, "epoch": 726, "lr": 8.679065896681615e-05}
{"train_loss": 0.265312522649765, "global_step": 64661, "epoch": 726, "lr": 8.67902663756507e-05}
{"train_loss": 0.2978266477584839, "global_step": 64662, "epoch": 726, "lr": 8.678987377953927e-05}
{"train_loss": 0.2399905025959015, "global_step": 64663, "epoch": 726, "lr": 8.678948117848189e-05}
{"train_loss": 0.289995938539505, "global_step": 64664, "epoch": 726, "lr": 8.67890885724786e-05}
{"train_loss": 0.22793371975421906, "global_step": 64665, "epoch": 726, "lr": 8.678869596152946e-05}
{"train_loss": 0.16032731533050537, "global_step": 64666, "epoch": 726, "lr": 8.678830334563455e-05}
{"train_loss": 0.23133811354637146, "global_step": 64667, "epoch": 726, "lr": 8.678791072479389e-05}
{"train_loss": 0.15822339057922363, "global_step": 64668, "epoch": 726, "lr": 8.678751809900754e-05}
{"train_loss": 0.17408733069896698, "global_step": 64669, "epoch": 726, "lr": 8.678712546827555e-05}
{"train_loss": 0.2651042938232422, "global_step": 64670, "epoch": 726, "lr": 8.6786732832598e-05}
{"train_loss": 0.3339690566062927, "global_step": 64671, "epoch": 726, "lr": 8.67863401919749e-05}
{"train_loss": 0.2889590859413147, "global_step": 64672, "epoch": 726, "lr": 8.678594754640633e-05}
{"train_loss": 0.26151803135871887, "global_step": 64673, "epoch": 726, "lr": 8.678555489589235e-05}
{"train_loss": 0.1922658234834671, "global_step": 64674, "epoch": 726, "lr": 8.678516224043298e-05}
{"train_loss": 0.24870717525482178, "global_step": 64675, "epoch": 726, "lr": 8.678476958002831e-05}
{"train_loss": 0.3257758319377899, "global_step": 64676, "epoch": 726, "lr": 8.678437691467838e-05}
{"train_loss": 0.17161589860916138, "global_step": 64677, "epoch": 726, "lr": 8.678398424438323e-05}
{"train_loss": 0.18504342436790466, "global_step": 64678, "epoch": 726, "lr": 8.678359156914291e-05}
{"train_loss": 0.2144899070262909, "global_step": 64679, "epoch": 726, "lr": 8.678319888895751e-05}
{"train_loss": 0.2501864433288574, "global_step": 64680, "epoch": 726, "lr": 8.678280620382704e-05}
{"train_loss": 0.1955500990152359, "global_step": 64681, "epoch": 726, "lr": 8.678241351375157e-05}
{"train_loss": 0.26116815209388733, "global_step": 64682, "epoch": 726, "lr": 8.678202081873116e-05}
{"train_loss": 0.2834340035915375, "global_step": 64683, "epoch": 726, "lr": 8.678162811876584e-05}
{"train_loss": 0.22410613298416138, "global_step": 64684, "epoch": 726, "lr": 8.67812354138557e-05}
{"train_loss": 0.1646406352519989, "global_step": 64685, "epoch": 726, "lr": 8.678084270400077e-05}
{"train_loss": 0.22061020135879517, "global_step": 64686, "epoch": 726, "lr": 8.67804499892011e-05}
{"train_loss": 0.2609204649925232, "global_step": 64687, "epoch": 726, "lr": 8.678005726945672e-05}
{"train_loss": 0.2576307952404022, "global_step": 64688, "epoch": 726, "lr": 8.677966454476774e-05}
{"train_loss": 0.207609161734581, "global_step": 64689, "epoch": 726, "lr": 8.677927181513419e-05}
{"train_loss": 0.20230406522750854, "global_step": 64690, "epoch": 726, "lr": 8.677887908055609e-05}
{"train_loss": 0.22128163278102875, "global_step": 64691, "epoch": 726, "lr": 8.677848634103354e-05}
{"train_loss": 0.3119398057460785, "global_step": 64692, "epoch": 726, "lr": 8.677809359656655e-05}
{"train_loss": 0.15194198489189148, "global_step": 64693, "epoch": 726, "lr": 8.67777008471552e-05}
{"train_loss": 0.3605957329273224, "global_step": 64694, "epoch": 726, "lr": 8.677730809279954e-05}
{"train_loss": 0.23701749742031097, "global_step": 64695, "epoch": 726, "lr": 8.677691533349963e-05}
{"train_loss": 0.28338953852653503, "global_step": 64696, "epoch": 726, "lr": 8.67765225692555e-05}
{"train_loss": 0.271615594625473, "global_step": 64697, "epoch": 726, "lr": 8.67761298000672e-05}
{"train_loss": 0.1951904147863388, "global_step": 64698, "epoch": 726, "lr": 8.677573702593483e-05}
{"train_loss": 0.16487380862236023, "global_step": 64699, "epoch": 726, "lr": 8.677534424685838e-05}
{"train_loss": 0.23715263605117798, "global_step": 64700, "epoch": 726, "lr": 8.677495146283794e-05}
{"train_loss": 0.2193724513053894, "global_step": 64701, "epoch": 726, "lr": 8.677455867387357e-05}
{"train_loss": 0.23860953028282422, "global_step": 64702, "epoch": 726, "lr": 8.67741658799653e-05, "val_loss": 3.5375475883483887}
{"train_loss": 0.16675633192062378, "global_step": 64703, "epoch": 727, "lr": 8.677377308111318e-05}
{"train_loss": 0.22767148911952972, "global_step": 64704, "epoch": 727, "lr": 8.677338027731729e-05}
{"train_loss": 0.2029104381799698, "global_step": 64705, "epoch": 727, "lr": 8.677298746857767e-05}
{"train_loss": 0.22319871187210083, "global_step": 64706, "epoch": 727, "lr": 8.677259465489437e-05}
{"train_loss": 0.31662777066230774, "global_step": 64707, "epoch": 727, "lr": 8.677220183626744e-05}
{"train_loss": 0.21229006350040436, "global_step": 64708, "epoch": 727, "lr": 8.677180901269692e-05}
{"train_loss": 0.1995571404695511, "global_step": 64709, "epoch": 727, "lr": 8.67714161841829e-05}
{"train_loss": 0.26172325015068054, "global_step": 64710, "epoch": 727, "lr": 8.67710233507254e-05}
{"train_loss": 0.20285402238368988, "global_step": 64711, "epoch": 727, "lr": 8.677063051232448e-05}
{"train_loss": 0.21358056366443634, "global_step": 64712, "epoch": 727, "lr": 8.67702376689802e-05}
{"train_loss": 0.21812716126441956, "global_step": 64713, "epoch": 727, "lr": 8.676984482069262e-05}
{"train_loss": 0.23590651154518127, "global_step": 64714, "epoch": 727, "lr": 8.676945196746178e-05}
{"train_loss": 0.1512160748243332, "global_step": 64715, "epoch": 727, "lr": 8.676905910928773e-05}
{"train_loss": 0.24293728172779083, "global_step": 64716, "epoch": 727, "lr": 8.676866624617053e-05}
{"train_loss": 0.2642245888710022, "global_step": 64717, "epoch": 727, "lr": 8.676827337811022e-05}
{"train_loss": 0.30066537857055664, "global_step": 64718, "epoch": 727, "lr": 8.676788050510687e-05}
{"train_loss": 0.2023593634366989, "global_step": 64719, "epoch": 727, "lr": 8.676748762716054e-05}
{"train_loss": 0.23053820431232452, "global_step": 64720, "epoch": 727, "lr": 8.676709474427126e-05}
{"train_loss": 0.25893434882164, "global_step": 64721, "epoch": 727, "lr": 8.676670185643909e-05}
{"train_loss": 0.24190451204776764, "global_step": 64722, "epoch": 727, "lr": 8.676630896366409e-05}
{"train_loss": 0.14849430322647095, "global_step": 64723, "epoch": 727, "lr": 8.676591606594631e-05}
{"train_loss": 0.20008134841918945, "global_step": 64724, "epoch": 727, "lr": 8.676552316328579e-05}
{"train_loss": 0.26050281524658203, "global_step": 64725, "epoch": 727, "lr": 8.67651302556826e-05}
{"train_loss": 0.17401374876499176, "global_step": 64726, "epoch": 727, "lr": 8.67647373431368e-05}
{"train_loss": 0.22608686983585358, "global_step": 64727, "epoch": 727, "lr": 8.67643444256484e-05}
{"train_loss": 0.16036444902420044, "global_step": 64728, "epoch": 727, "lr": 8.67639515032175e-05}
{"train_loss": 0.24241666495800018, "global_step": 64729, "epoch": 727, "lr": 8.676355857584413e-05}
{"train_loss": 0.1813802868127823, "global_step": 64730, "epoch": 727, "lr": 8.676316564352835e-05}
{"train_loss": 0.25792986154556274, "global_step": 64731, "epoch": 727, "lr": 8.67627727062702e-05}
{"train_loss": 0.22976265847682953, "global_step": 64732, "epoch": 727, "lr": 8.676237976406975e-05}
{"train_loss": 0.2461291402578354, "global_step": 64733, "epoch": 727, "lr": 8.676198681692706e-05}
{"train_loss": 0.3087276518344879, "global_step": 64734, "epoch": 727, "lr": 8.676159386484217e-05}
{"train_loss": 0.22911477088928223, "global_step": 64735, "epoch": 727, "lr": 8.676120090781512e-05}
{"train_loss": 0.29745200276374817, "global_step": 64736, "epoch": 727, "lr": 8.676080794584596e-05}
{"train_loss": 0.1569298356771469, "global_step": 64737, "epoch": 727, "lr": 8.676041497893476e-05}
{"train_loss": 0.30703264474868774, "global_step": 64738, "epoch": 727, "lr": 8.67600220070816e-05}
{"train_loss": 0.268574059009552, "global_step": 64739, "epoch": 727, "lr": 8.675962903028648e-05}
{"train_loss": 0.23630835115909576, "global_step": 64740, "epoch": 727, "lr": 8.675923604854947e-05}
{"train_loss": 0.2089228481054306, "global_step": 64741, "epoch": 727, "lr": 8.675884306187065e-05}
{"train_loss": 0.2768036127090454, "global_step": 64742, "epoch": 727, "lr": 8.675845007025006e-05}
{"train_loss": 0.18695205450057983, "global_step": 64743, "epoch": 727, "lr": 8.675805707368772e-05}
{"train_loss": 0.1746358871459961, "global_step": 64744, "epoch": 727, "lr": 8.675766407218371e-05}
{"train_loss": 0.24047662317752838, "global_step": 64745, "epoch": 727, "lr": 8.675727106573808e-05}
{"train_loss": 0.17326343059539795, "global_step": 64746, "epoch": 727, "lr": 8.675687805435089e-05}
{"train_loss": 0.12773582339286804, "global_step": 64747, "epoch": 727, "lr": 8.675648503802218e-05}
{"train_loss": 0.25611019134521484, "global_step": 64748, "epoch": 727, "lr": 8.675609201675201e-05}
{"train_loss": 0.19365213811397552, "global_step": 64749, "epoch": 727, "lr": 8.675569899054042e-05}
{"train_loss": 0.1519768089056015, "global_step": 64750, "epoch": 727, "lr": 8.675530595938748e-05}
{"train_loss": 0.24569272994995117, "global_step": 64751, "epoch": 727, "lr": 8.675491292329325e-05}
{"train_loss": 0.18054457008838654, "global_step": 64752, "epoch": 727, "lr": 8.675451988225777e-05}
{"train_loss": 0.31118977069854736, "global_step": 64753, "epoch": 727, "lr": 8.675412683628108e-05}
{"train_loss": 0.16640600562095642, "global_step": 64754, "epoch": 727, "lr": 8.675373378536326e-05}
{"train_loss": 0.22976824641227722, "global_step": 64755, "epoch": 727, "lr": 8.675334072950435e-05}
{"train_loss": 0.20547892153263092, "global_step": 64756, "epoch": 727, "lr": 8.675294766870437e-05}
{"train_loss": 0.3006618320941925, "global_step": 64757, "epoch": 727, "lr": 8.675255460296343e-05}
{"train_loss": 0.26968562602996826, "global_step": 64758, "epoch": 727, "lr": 8.675216153228155e-05}
{"train_loss": 0.26664742827415466, "global_step": 64759, "epoch": 727, "lr": 8.67517684566588e-05}
{"train_loss": 0.29702088236808777, "global_step": 64760, "epoch": 727, "lr": 8.675137537609521e-05}
{"train_loss": 0.17988993227481842, "global_step": 64761, "epoch": 727, "lr": 8.675098229059085e-05}
{"train_loss": 0.2618298828601837, "global_step": 64762, "epoch": 727, "lr": 8.675058920014577e-05}
{"train_loss": 0.1710786074399948, "global_step": 64763, "epoch": 727, "lr": 8.675019610476002e-05}
{"train_loss": 0.20874252915382385, "global_step": 64764, "epoch": 727, "lr": 8.674980300443365e-05}
{"train_loss": 0.1684294044971466, "global_step": 64765, "epoch": 727, "lr": 8.674940989916673e-05}
{"train_loss": 0.20150195062160492, "global_step": 64766, "epoch": 727, "lr": 8.674901678895928e-05}
{"train_loss": 0.2274545133113861, "global_step": 64767, "epoch": 727, "lr": 8.674862367381138e-05}
{"train_loss": 0.21598519384860992, "global_step": 64768, "epoch": 727, "lr": 8.674823055372309e-05}
{"train_loss": 0.26774895191192627, "global_step": 64769, "epoch": 727, "lr": 8.674783742869444e-05}
{"train_loss": 0.20043379068374634, "global_step": 64770, "epoch": 727, "lr": 8.674744429872548e-05}
{"train_loss": 0.329574853181839, "global_step": 64771, "epoch": 727, "lr": 8.674705116381628e-05}
{"train_loss": 0.3014678657054901, "global_step": 64772, "epoch": 727, "lr": 8.674665802396688e-05}
{"train_loss": 0.247712641954422, "global_step": 64773, "epoch": 727, "lr": 8.674626487917736e-05}
{"train_loss": 0.20708979666233063, "global_step": 64774, "epoch": 727, "lr": 8.674587172944775e-05}
{"train_loss": 0.21507404744625092, "global_step": 64775, "epoch": 727, "lr": 8.674547857477808e-05}
{"train_loss": 0.1606360673904419, "global_step": 64776, "epoch": 727, "lr": 8.674508541516847e-05}
{"train_loss": 0.239283487200737, "global_step": 64777, "epoch": 727, "lr": 8.674469225061888e-05}
{"train_loss": 0.19269150495529175, "global_step": 64778, "epoch": 727, "lr": 8.674429908112945e-05}
{"train_loss": 0.21959419548511505, "global_step": 64779, "epoch": 727, "lr": 8.674390590670018e-05}
{"train_loss": 0.20422588288784027, "global_step": 64780, "epoch": 727, "lr": 8.674351272733115e-05}
{"train_loss": 0.27464959025382996, "global_step": 64781, "epoch": 727, "lr": 8.67431195430224e-05}
{"train_loss": 0.2812109589576721, "global_step": 64782, "epoch": 727, "lr": 8.674272635377398e-05}
{"train_loss": 0.11800726503133774, "global_step": 64783, "epoch": 727, "lr": 8.674233315958596e-05}
{"train_loss": 0.29733842611312866, "global_step": 64784, "epoch": 727, "lr": 8.674193996045837e-05}
{"train_loss": 0.15357494354248047, "global_step": 64785, "epoch": 727, "lr": 8.674154675639128e-05}
{"train_loss": 0.1710982471704483, "global_step": 64786, "epoch": 727, "lr": 8.674115354738473e-05}
{"train_loss": 0.22943691909313202, "global_step": 64787, "epoch": 727, "lr": 8.674076033343879e-05}
{"train_loss": 0.24494948983192444, "global_step": 64788, "epoch": 727, "lr": 8.674036711455349e-05}
{"train_loss": 0.21760988235473633, "global_step": 64789, "epoch": 727, "lr": 8.673997389072892e-05}
{"train_loss": 0.26660090684890747, "global_step": 64790, "epoch": 727, "lr": 8.673958066196508e-05}
{"train_loss": 0.22544429267055532, "global_step": 64791, "epoch": 727, "lr": 8.673918742826206e-05, "val_loss": 3.6223416328430176}
{"train_loss": 0.23409566283226013, "global_step": 64792, "epoch": 728, "lr": 8.67387941896199e-05}
{"train_loss": 0.25179943442344666, "global_step": 64793, "epoch": 728, "lr": 8.673840094603867e-05}
{"train_loss": 0.2279801219701767, "global_step": 64794, "epoch": 728, "lr": 8.673800769751841e-05}
{"train_loss": 0.24150696396827698, "global_step": 64795, "epoch": 728, "lr": 8.673761444405916e-05}
{"train_loss": 0.23412460088729858, "global_step": 64796, "epoch": 728, "lr": 8.6737221185661e-05}
{"train_loss": 0.22304777801036835, "global_step": 64797, "epoch": 728, "lr": 8.673682792232395e-05}
{"train_loss": 0.25299620628356934, "global_step": 64798, "epoch": 728, "lr": 8.673643465404808e-05}
{"train_loss": 0.24661557376384735, "global_step": 64799, "epoch": 728, "lr": 8.673604138083347e-05}
{"train_loss": 0.2917870283126831, "global_step": 64800, "epoch": 728, "lr": 8.673564810268013e-05}
{"train_loss": 0.1487760990858078, "global_step": 64801, "epoch": 728, "lr": 8.673525481958813e-05}
{"train_loss": 0.18151018023490906, "global_step": 64802, "epoch": 728, "lr": 8.673486153155752e-05}
{"train_loss": 0.1697484403848648, "global_step": 64803, "epoch": 728, "lr": 8.673446823858836e-05}
{"train_loss": 0.25575530529022217, "global_step": 64804, "epoch": 728, "lr": 8.67340749406807e-05}
{"train_loss": 0.20918838679790497, "global_step": 64805, "epoch": 728, "lr": 8.67336816378346e-05}
{"train_loss": 0.32847556471824646, "global_step": 64806, "epoch": 728, "lr": 8.673328833005007e-05}
{"train_loss": 0.13363485038280487, "global_step": 64807, "epoch": 728, "lr": 8.673289501732723e-05}
{"train_loss": 0.24539750814437866, "global_step": 64808, "epoch": 728, "lr": 8.673250169966609e-05}
{"train_loss": 0.22475123405456543, "global_step": 64809, "epoch": 728, "lr": 8.673210837706673e-05}
{"train_loss": 0.2360028177499771, "global_step": 64810, "epoch": 728, "lr": 8.673171504952916e-05}
{"train_loss": 0.20736676454544067, "global_step": 64811, "epoch": 728, "lr": 8.673132171705348e-05}
{"train_loss": 0.24046780169010162, "global_step": 64812, "epoch": 728, "lr": 8.67309283796397e-05}
{"train_loss": 0.23496879637241364, "global_step": 64813, "epoch": 728, "lr": 8.67305350372879e-05}
{"train_loss": 0.20475441217422485, "global_step": 64814, "epoch": 728, "lr": 8.673014168999813e-05}
{"train_loss": 0.2207770049571991, "global_step": 64815, "epoch": 728, "lr": 8.672974833777044e-05}
{"train_loss": 0.21216470003128052, "global_step": 64816, "epoch": 728, "lr": 8.672935498060488e-05}
{"train_loss": 0.2471727877855301, "global_step": 64817, "epoch": 728, "lr": 8.672896161850152e-05}
{"train_loss": 0.1852254420518875, "global_step": 64818, "epoch": 728, "lr": 8.672856825146039e-05}
{"train_loss": 0.17761653661727905, "global_step": 64819, "epoch": 728, "lr": 8.672817487948156e-05}
{"train_loss": 0.20605921745300293, "global_step": 64820, "epoch": 728, "lr": 8.672778150256505e-05}
{"train_loss": 0.27204522490501404, "global_step": 64821, "epoch": 728, "lr": 8.672738812071096e-05}
{"train_loss": 0.20566657185554504, "global_step": 64822, "epoch": 728, "lr": 8.672699473391931e-05}
{"train_loss": 0.22098807990550995, "global_step": 64823, "epoch": 728, "lr": 8.672660134219016e-05}
{"train_loss": 0.26195642352104187, "global_step": 64824, "epoch": 728, "lr": 8.672620794552358e-05}
{"train_loss": 0.12153501808643341, "global_step": 64825, "epoch": 728, "lr": 8.67258145439196e-05}
{"train_loss": 0.32246118783950806, "global_step": 64826, "epoch": 728, "lr": 8.672542113737829e-05}
{"train_loss": 0.25355085730552673, "global_step": 64827, "epoch": 728, "lr": 8.672502772589968e-05}
{"train_loss": 0.23043540120124817, "global_step": 64828, "epoch": 728, "lr": 8.672463430948385e-05}
{"train_loss": 0.15911471843719482, "global_step": 64829, "epoch": 728, "lr": 8.672424088813083e-05}
{"train_loss": 0.1836729794740677, "global_step": 64830, "epoch": 728, "lr": 8.67238474618407e-05}
{"train_loss": 0.23899120092391968, "global_step": 64831, "epoch": 728, "lr": 8.672345403061348e-05}
{"train_loss": 0.2091578096151352, "global_step": 64832, "epoch": 728, "lr": 8.672306059444923e-05}
{"train_loss": 0.19832240045070648, "global_step": 64833, "epoch": 728, "lr": 8.672266715334804e-05}
{"train_loss": 0.24807240068912506, "global_step": 64834, "epoch": 728, "lr": 8.672227370730992e-05}
{"train_loss": 0.28790482878685, "global_step": 64835, "epoch": 728, "lr": 8.672188025633494e-05}
{"train_loss": 0.2630070149898529, "global_step": 64836, "epoch": 728, "lr": 8.672148680042314e-05}
{"train_loss": 0.21812349557876587, "global_step": 64837, "epoch": 728, "lr": 8.672109333957461e-05}
{"train_loss": 0.3098526895046234, "global_step": 64838, "epoch": 728, "lr": 8.672069987378936e-05}
{"train_loss": 0.28162845969200134, "global_step": 64839, "epoch": 728, "lr": 8.672030640306745e-05}
{"train_loss": 0.1840735524892807, "global_step": 64840, "epoch": 728, "lr": 8.671991292740896e-05}
{"train_loss": 0.206630676984787, "global_step": 64841, "epoch": 728, "lr": 8.671951944681391e-05}
{"train_loss": 0.2500503361225128, "global_step": 64842, "epoch": 728, "lr": 8.671912596128237e-05}
{"train_loss": 0.2045338898897171, "global_step": 64843, "epoch": 728, "lr": 8.671873247081441e-05}
{"train_loss": 0.15882721543312073, "global_step": 64844, "epoch": 728, "lr": 8.671833897541006e-05}
{"train_loss": 0.2694963812828064, "global_step": 64845, "epoch": 728, "lr": 8.671794547506935e-05}
{"train_loss": 0.21235167980194092, "global_step": 64846, "epoch": 728, "lr": 8.671755196979239e-05}
{"train_loss": 0.15879902243614197, "global_step": 64847, "epoch": 728, "lr": 8.67171584595792e-05}
{"train_loss": 0.23681609332561493, "global_step": 64848, "epoch": 728, "lr": 8.671676494442982e-05}
{"train_loss": 0.2629108428955078, "global_step": 64849, "epoch": 728, "lr": 8.671637142434433e-05}
{"train_loss": 0.24185515940189362, "global_step": 64850, "epoch": 728, "lr": 8.671597789932276e-05}
{"train_loss": 0.2159917652606964, "global_step": 64851, "epoch": 728, "lr": 8.671558436936518e-05}
{"train_loss": 0.27314624190330505, "global_step": 64852, "epoch": 728, "lr": 8.671519083447164e-05}
{"train_loss": 0.2285354882478714, "global_step": 64853, "epoch": 728, "lr": 8.671479729464218e-05}
{"train_loss": 0.21230627596378326, "global_step": 64854, "epoch": 728, "lr": 8.671440374987687e-05}
{"train_loss": 0.29938840866088867, "global_step": 64855, "epoch": 728, "lr": 8.671401020017576e-05}
{"train_loss": 0.30787694454193115, "global_step": 64856, "epoch": 728, "lr": 8.671361664553889e-05}
{"train_loss": 0.20122677087783813, "global_step": 64857, "epoch": 728, "lr": 8.671322308596632e-05}
{"train_loss": 0.17086493968963623, "global_step": 64858, "epoch": 728, "lr": 8.671282952145812e-05}
{"train_loss": 0.19729070365428925, "global_step": 64859, "epoch": 728, "lr": 8.671243595201431e-05}
{"train_loss": 0.1550758183002472, "global_step": 64860, "epoch": 728, "lr": 8.671204237763498e-05}
{"train_loss": 0.22357533872127533, "global_step": 64861, "epoch": 728, "lr": 8.671164879832014e-05}
{"train_loss": 0.15899062156677246, "global_step": 64862, "epoch": 728, "lr": 8.671125521406989e-05}
{"train_loss": 0.2665630877017975, "global_step": 64863, "epoch": 728, "lr": 8.671086162488425e-05}
{"train_loss": 0.2758982181549072, "global_step": 64864, "epoch": 728, "lr": 8.671046803076328e-05}
{"train_loss": 0.23274379968643188, "global_step": 64865, "epoch": 728, "lr": 8.671007443170703e-05}
{"train_loss": 0.2606467306613922, "global_step": 64866, "epoch": 728, "lr": 8.670968082771556e-05}
{"train_loss": 0.22863450646400452, "global_step": 64867, "epoch": 728, "lr": 8.670928721878893e-05}
{"train_loss": 0.15724220871925354, "global_step": 64868, "epoch": 728, "lr": 8.670889360492718e-05}
{"train_loss": 0.2372843325138092, "global_step": 64869, "epoch": 728, "lr": 8.670849998613035e-05}
{"train_loss": 0.2753491699695587, "global_step": 64870, "epoch": 728, "lr": 8.670810636239853e-05}
{"train_loss": 0.2227386087179184, "global_step": 64871, "epoch": 728, "lr": 8.670771273373176e-05}
{"train_loss": 0.2881694436073303, "global_step": 64872, "epoch": 728, "lr": 8.670731910013006e-05}
{"train_loss": 0.3113248944282532, "global_step": 64873, "epoch": 728, "lr": 8.670692546159353e-05}
{"train_loss": 0.18002502620220184, "global_step": 64874, "epoch": 728, "lr": 8.670653181812218e-05}
{"train_loss": 0.17698955535888672, "global_step": 64875, "epoch": 728, "lr": 8.67061381697161e-05}
{"train_loss": 0.295853853225708, "global_step": 64876, "epoch": 728, "lr": 8.670574451637533e-05}
{"train_loss": 0.20595790445804596, "global_step": 64877, "epoch": 728, "lr": 8.670535085809992e-05}
{"train_loss": 0.2582671344280243, "global_step": 64878, "epoch": 728, "lr": 8.670495719488991e-05}
{"train_loss": 0.2093413919210434, "global_step": 64879, "epoch": 728, "lr": 8.670456352674537e-05}
{"train_loss": 0.2276011189717925, "global_step": 64880, "epoch": 728, "lr": 8.670416985366634e-05, "val_loss": 3.545846462249756}
{"train_loss": 0.21893256902694702, "global_step": 64881, "epoch": 729, "lr": 8.670377617565291e-05}
{"train_loss": 0.16304224729537964, "global_step": 64882, "epoch": 729, "lr": 8.670338249270507e-05}
{"train_loss": 0.16494590044021606, "global_step": 64883, "epoch": 729, "lr": 8.670298880482293e-05}
{"train_loss": 0.1804409921169281, "global_step": 64884, "epoch": 729, "lr": 8.670259511200652e-05}
{"train_loss": 0.20159515738487244, "global_step": 64885, "epoch": 729, "lr": 8.670220141425589e-05}
{"train_loss": 0.21107137203216553, "global_step": 64886, "epoch": 729, "lr": 8.670180771157108e-05}
{"train_loss": 0.173095241189003, "global_step": 64887, "epoch": 729, "lr": 8.670141400395217e-05}
{"train_loss": 0.22152650356292725, "global_step": 64888, "epoch": 729, "lr": 8.670102029139921e-05}
{"train_loss": 0.2817544937133789, "global_step": 64889, "epoch": 729, "lr": 8.670062657391224e-05}
{"train_loss": 0.2233584225177765, "global_step": 64890, "epoch": 729, "lr": 8.670023285149134e-05}
{"train_loss": 0.1885843575000763, "global_step": 64891, "epoch": 729, "lr": 8.66998391241365e-05}
{"train_loss": 0.2680180072784424, "global_step": 64892, "epoch": 729, "lr": 8.669944539184784e-05}
{"train_loss": 0.24432578682899475, "global_step": 64893, "epoch": 729, "lr": 8.669905165462538e-05}
{"train_loss": 0.2821562588214874, "global_step": 64894, "epoch": 729, "lr": 8.669865791246919e-05}
{"train_loss": 0.17426730692386627, "global_step": 64895, "epoch": 729, "lr": 8.66982641653793e-05}
{"train_loss": 0.26522618532180786, "global_step": 64896, "epoch": 729, "lr": 8.669787041335577e-05}
{"train_loss": 0.2237323373556137, "global_step": 64897, "epoch": 729, "lr": 8.669747665639868e-05}
{"train_loss": 0.12469203770160675, "global_step": 64898, "epoch": 729, "lr": 8.669708289450804e-05}
{"train_loss": 0.17969529330730438, "global_step": 64899, "epoch": 729, "lr": 8.669668912768394e-05}
{"train_loss": 0.26656949520111084, "global_step": 64900, "epoch": 729, "lr": 8.669629535592641e-05}
{"train_loss": 0.2090378701686859, "global_step": 64901, "epoch": 729, "lr": 8.669590157923551e-05}
{"train_loss": 0.23303170502185822, "global_step": 64902, "epoch": 729, "lr": 8.66955077976113e-05}
{"train_loss": 0.21492022275924683, "global_step": 64903, "epoch": 729, "lr": 8.669511401105382e-05}
{"train_loss": 0.3238113522529602, "global_step": 64904, "epoch": 729, "lr": 8.669472021956313e-05}
{"train_loss": 0.12192703783512115, "global_step": 64905, "epoch": 729, "lr": 8.669432642313929e-05}
{"train_loss": 0.26390647888183594, "global_step": 64906, "epoch": 729, "lr": 8.669393262178234e-05}
{"train_loss": 0.1758514642715454, "global_step": 64907, "epoch": 729, "lr": 8.669353881549234e-05}
{"train_loss": 0.13892792165279388, "global_step": 64908, "epoch": 729, "lr": 8.669314500426933e-05}
{"train_loss": 0.29187121987342834, "global_step": 64909, "epoch": 729, "lr": 8.669275118811337e-05}
{"train_loss": 0.2811439335346222, "global_step": 64910, "epoch": 729, "lr": 8.669235736702454e-05}
{"train_loss": 0.24242094159126282, "global_step": 64911, "epoch": 729, "lr": 8.669196354100285e-05}
{"train_loss": 0.18356028199195862, "global_step": 64912, "epoch": 729, "lr": 8.66915697100484e-05}
{"train_loss": 0.2557944357395172, "global_step": 64913, "epoch": 729, "lr": 8.669117587416118e-05}
{"train_loss": 0.18642401695251465, "global_step": 64914, "epoch": 729, "lr": 8.66907820333413e-05}
{"train_loss": 0.15533259510993958, "global_step": 64915, "epoch": 729, "lr": 8.669038818758877e-05}
{"train_loss": 0.28798288106918335, "global_step": 64916, "epoch": 729, "lr": 8.668999433690369e-05}
{"train_loss": 0.16505390405654907, "global_step": 64917, "epoch": 729, "lr": 8.668960048128607e-05}
{"train_loss": 0.22312262654304504, "global_step": 64918, "epoch": 729, "lr": 8.668920662073599e-05}
{"train_loss": 0.18727929890155792, "global_step": 64919, "epoch": 729, "lr": 8.668881275525349e-05}
{"train_loss": 0.1770375370979309, "global_step": 64920, "epoch": 729, "lr": 8.668841888483863e-05}
{"train_loss": 0.26219990849494934, "global_step": 64921, "epoch": 729, "lr": 8.668802500949146e-05}
{"train_loss": 0.23017162084579468, "global_step": 64922, "epoch": 729, "lr": 8.668763112921203e-05}
{"train_loss": 0.19014684855937958, "global_step": 64923, "epoch": 729, "lr": 8.66872372440004e-05}
{"train_loss": 0.23084808886051178, "global_step": 64924, "epoch": 729, "lr": 8.668684335385662e-05}
{"train_loss": 0.22505807876586914, "global_step": 64925, "epoch": 729, "lr": 8.668644945878073e-05}
{"train_loss": 0.23217551410198212, "global_step": 64926, "epoch": 729, "lr": 8.668605555877278e-05}
{"train_loss": 0.30616721510887146, "global_step": 64927, "epoch": 729, "lr": 8.668566165383287e-05}
{"train_loss": 0.24990157783031464, "global_step": 64928, "epoch": 729, "lr": 8.6685267743961e-05}
{"train_loss": 0.22226877510547638, "global_step": 64929, "epoch": 729, "lr": 8.668487382915724e-05}
{"train_loss": 0.30710074305534363, "global_step": 64930, "epoch": 729, "lr": 8.668447990942166e-05}
{"train_loss": 0.29466885328292847, "global_step": 64931, "epoch": 729, "lr": 8.66840859847543e-05}
{"train_loss": 0.18339768052101135, "global_step": 64932, "epoch": 729, "lr": 8.668369205515519e-05}
{"train_loss": 0.21560420095920563, "global_step": 64933, "epoch": 729, "lr": 8.668329812062443e-05}
{"train_loss": 0.2913014590740204, "global_step": 64934, "epoch": 729, "lr": 8.668290418116203e-05}
{"train_loss": 0.21020081639289856, "global_step": 64935, "epoch": 729, "lr": 8.668251023676806e-05}
{"train_loss": 0.2788195013999939, "global_step": 64936, "epoch": 729, "lr": 8.668211628744259e-05}
{"train_loss": 0.23208941519260406, "global_step": 64937, "epoch": 729, "lr": 8.668172233318565e-05}
{"train_loss": 0.26652687788009644, "global_step": 64938, "epoch": 729, "lr": 8.66813283739973e-05}
{"train_loss": 0.3800627887248993, "global_step": 64939, "epoch": 729, "lr": 8.668093440987757e-05}
{"train_loss": 0.1639447957277298, "global_step": 64940, "epoch": 729, "lr": 8.668054044082656e-05}
{"train_loss": 0.3128737211227417, "global_step": 64941, "epoch": 729, "lr": 8.66801464668443e-05}
{"train_loss": 0.3253799080848694, "global_step": 64942, "epoch": 729, "lr": 8.667975248793083e-05}
{"train_loss": 0.26055434346199036, "global_step": 64943, "epoch": 729, "lr": 8.667935850408621e-05}
{"train_loss": 0.20347589254379272, "global_step": 64944, "epoch": 729, "lr": 8.667896451531053e-05}
{"train_loss": 0.3344065845012665, "global_step": 64945, "epoch": 729, "lr": 8.667857052160377e-05}
{"train_loss": 0.40273213386535645, "global_step": 64946, "epoch": 729, "lr": 8.667817652296604e-05}
{"train_loss": 0.24233320355415344, "global_step": 64947, "epoch": 729, "lr": 8.667778251939738e-05}
{"train_loss": 0.22250713407993317, "global_step": 64948, "epoch": 729, "lr": 8.667738851089783e-05}
{"train_loss": 0.2630012035369873, "global_step": 64949, "epoch": 729, "lr": 8.667699449746746e-05}
{"train_loss": 0.24268785119056702, "global_step": 64950, "epoch": 729, "lr": 8.667660047910631e-05}
{"train_loss": 0.19868367910385132, "global_step": 64951, "epoch": 729, "lr": 8.667620645581445e-05}
{"train_loss": 0.2709370255470276, "global_step": 64952, "epoch": 729, "lr": 8.667581242759191e-05}
{"train_loss": 0.21350258588790894, "global_step": 64953, "epoch": 729, "lr": 8.667541839443874e-05}
{"train_loss": 0.21456684172153473, "global_step": 64954, "epoch": 729, "lr": 8.667502435635503e-05}
{"train_loss": 0.18720456957817078, "global_step": 64955, "epoch": 729, "lr": 8.66746303133408e-05}
{"train_loss": 0.26784828305244446, "global_step": 64956, "epoch": 729, "lr": 8.667423626539613e-05}
{"train_loss": 0.30627259612083435, "global_step": 64957, "epoch": 729, "lr": 8.667384221252103e-05}
{"train_loss": 0.20364956557750702, "global_step": 64958, "epoch": 729, "lr": 8.667344815471558e-05}
{"train_loss": 0.30030593276023865, "global_step": 64959, "epoch": 729, "lr": 8.667305409197984e-05}
{"train_loss": 0.2658716142177582, "global_step": 64960, "epoch": 729, "lr": 8.667266002431386e-05}
{"train_loss": 0.1873668134212494, "global_step": 64961, "epoch": 729, "lr": 8.667226595171769e-05}
{"train_loss": 0.18162311613559723, "global_step": 64962, "epoch": 729, "lr": 8.667187187419139e-05}
{"train_loss": 0.11518162488937378, "global_step": 64963, "epoch": 729, "lr": 8.667147779173498e-05}
{"train_loss": 0.24654994904994965, "global_step": 64964, "epoch": 729, "lr": 8.667108370434854e-05}
{"train_loss": 0.330685019493103, "global_step": 64965, "epoch": 729, "lr": 8.667068961203213e-05}
{"train_loss": 0.2939363420009613, "global_step": 64966, "epoch": 729, "lr": 8.667029551478579e-05}
{"train_loss": 0.1383063793182373, "global_step": 64967, "epoch": 729, "lr": 8.666990141260956e-05}
{"train_loss": 0.2938958406448364, "global_step": 64968, "epoch": 729, "lr": 8.666950730550353e-05}
{"train_loss": 0.2340835848551118, "global_step": 64969, "epoch": 729, "lr": 8.666911319346772e-05, "val_loss": 3.489577531814575}
{"train_loss": 0.2444109171628952, "global_step": 64970, "epoch": 730, "lr": 8.66687190765022e-05}
{"train_loss": 0.2146887332201004, "global_step": 64971, "epoch": 730, "lr": 8.666832495460701e-05}
{"train_loss": 0.2802388370037079, "global_step": 64972, "epoch": 730, "lr": 8.666793082778223e-05}
{"train_loss": 0.3015623688697815, "global_step": 64973, "epoch": 730, "lr": 8.666753669602789e-05}
{"train_loss": 0.19787821173667908, "global_step": 64974, "epoch": 730, "lr": 8.666714255934402e-05}
{"train_loss": 0.26832520961761475, "global_step": 64975, "epoch": 730, "lr": 8.666674841773073e-05}
{"train_loss": 0.3016071021556854, "global_step": 64976, "epoch": 730, "lr": 8.666635427118801e-05}
{"train_loss": 0.23148386180400848, "global_step": 64977, "epoch": 730, "lr": 8.666596011971597e-05}
{"train_loss": 0.19808948040008545, "global_step": 64978, "epoch": 730, "lr": 8.666556596331464e-05}
{"train_loss": 0.24159234762191772, "global_step": 64979, "epoch": 730, "lr": 8.666517180198407e-05}
{"train_loss": 0.2428659349679947, "global_step": 64980, "epoch": 730, "lr": 8.666477763572431e-05}
{"train_loss": 0.22290174663066864, "global_step": 64981, "epoch": 730, "lr": 8.666438346453543e-05}
{"train_loss": 0.16299082338809967, "global_step": 64982, "epoch": 730, "lr": 8.666398928841744e-05}
{"train_loss": 0.23943355679512024, "global_step": 64983, "epoch": 730, "lr": 8.666359510737045e-05}
{"train_loss": 0.2322472184896469, "global_step": 64984, "epoch": 730, "lr": 8.666320092139448e-05}
{"train_loss": 0.25626587867736816, "global_step": 64985, "epoch": 730, "lr": 8.666280673048959e-05}
{"train_loss": 0.26354557275772095, "global_step": 64986, "epoch": 730, "lr": 8.666241253465583e-05}
{"train_loss": 0.2618952989578247, "global_step": 64987, "epoch": 730, "lr": 8.666201833389325e-05}
{"train_loss": 0.16585379838943481, "global_step": 64988, "epoch": 730, "lr": 8.666162412820191e-05}
{"train_loss": 0.17793087661266327, "global_step": 64989, "epoch": 730, "lr": 8.666122991758187e-05}
{"train_loss": 0.20707082748413086, "global_step": 64990, "epoch": 730, "lr": 8.666083570203317e-05}
{"train_loss": 0.23953653872013092, "global_step": 64991, "epoch": 730, "lr": 8.666044148155588e-05}
{"train_loss": 0.19487738609313965, "global_step": 64992, "epoch": 730, "lr": 8.666004725615003e-05}
{"train_loss": 0.19294819235801697, "global_step": 64993, "epoch": 730, "lr": 8.665965302581567e-05}
{"train_loss": 0.18344569206237793, "global_step": 64994, "epoch": 730, "lr": 8.665925879055288e-05}
{"train_loss": 0.30771884322166443, "global_step": 64995, "epoch": 730, "lr": 8.665886455036169e-05}
{"train_loss": 0.24256925284862518, "global_step": 64996, "epoch": 730, "lr": 8.665847030524216e-05}
{"train_loss": 0.17836816608905792, "global_step": 64997, "epoch": 730, "lr": 8.665807605519436e-05}
{"train_loss": 0.2541791796684265, "global_step": 64998, "epoch": 730, "lr": 8.665768180021833e-05}
{"train_loss": 0.23427081108093262, "global_step": 64999, "epoch": 730, "lr": 8.665728754031413e-05}
{"train_loss": 0.18476083874702454, "global_step": 65000, "epoch": 730, "lr": 8.665689327548177e-05}
{"train_loss": 0.1708904504776001, "global_step": 65001, "epoch": 730, "lr": 8.665649900572137e-05}
{"train_loss": 0.2916995584964752, "global_step": 65002, "epoch": 730, "lr": 8.665610473103293e-05}
{"train_loss": 0.2605322301387787, "global_step": 65003, "epoch": 730, "lr": 8.665571045141654e-05}
{"train_loss": 0.1920442432165146, "global_step": 65004, "epoch": 730, "lr": 8.665531616687221e-05}
{"train_loss": 0.24858073890209198, "global_step": 65005, "epoch": 730, "lr": 8.665492187740005e-05}
{"train_loss": 0.12153572589159012, "global_step": 65006, "epoch": 730, "lr": 8.665452758300006e-05}
{"train_loss": 0.20549292862415314, "global_step": 65007, "epoch": 730, "lr": 8.665413328367233e-05}
{"train_loss": 0.2786702513694763, "global_step": 65008, "epoch": 730, "lr": 8.665373897941688e-05}
{"train_loss": 0.21679480373859406, "global_step": 65009, "epoch": 730, "lr": 8.665334467023381e-05}
{"train_loss": 0.24333319067955017, "global_step": 65010, "epoch": 730, "lr": 8.665295035612313e-05}
{"train_loss": 0.26923802495002747, "global_step": 65011, "epoch": 730, "lr": 8.66525560370849e-05}
{"train_loss": 0.26343217492103577, "global_step": 65012, "epoch": 730, "lr": 8.66521617131192e-05}
{"train_loss": 0.17059431970119476, "global_step": 65013, "epoch": 730, "lr": 8.665176738422605e-05}
{"train_loss": 0.1654144674539566, "global_step": 65014, "epoch": 730, "lr": 8.665137305040552e-05}
{"train_loss": 0.2450283169746399, "global_step": 65015, "epoch": 730, "lr": 8.665097871165766e-05}
{"train_loss": 0.19362470507621765, "global_step": 65016, "epoch": 730, "lr": 8.665058436798252e-05}
{"train_loss": 0.2645788788795471, "global_step": 65017, "epoch": 730, "lr": 8.665019001938015e-05}
{"train_loss": 0.27623945474624634, "global_step": 65018, "epoch": 730, "lr": 8.664979566585062e-05}
{"train_loss": 0.2169143557548523, "global_step": 65019, "epoch": 730, "lr": 8.664940130739396e-05}
{"train_loss": 0.2406221330165863, "global_step": 65020, "epoch": 730, "lr": 8.664900694401024e-05}
{"train_loss": 0.3683358132839203, "global_step": 65021, "epoch": 730, "lr": 8.664861257569952e-05}
{"train_loss": 0.31175246834754944, "global_step": 65022, "epoch": 730, "lr": 8.664821820246182e-05}
{"train_loss": 0.20064978301525116, "global_step": 65023, "epoch": 730, "lr": 8.664782382429723e-05}
{"train_loss": 0.33006367087364197, "global_step": 65024, "epoch": 730, "lr": 8.664742944120578e-05}
{"train_loss": 0.26564842462539673, "global_step": 65025, "epoch": 730, "lr": 8.664703505318752e-05}
{"train_loss": 0.20466841757297516, "global_step": 65026, "epoch": 730, "lr": 8.664664066024253e-05}
{"train_loss": 0.2958846092224121, "global_step": 65027, "epoch": 730, "lr": 8.664624626237083e-05}
{"train_loss": 0.16940978169441223, "global_step": 65028, "epoch": 730, "lr": 8.66458518595725e-05}
{"train_loss": 0.22637712955474854, "global_step": 65029, "epoch": 730, "lr": 8.664545745184758e-05}
{"train_loss": 0.22592180967330933, "global_step": 65030, "epoch": 730, "lr": 8.664506303919613e-05}
{"train_loss": 0.2571902573108673, "global_step": 65031, "epoch": 730, "lr": 8.664466862161819e-05}
{"train_loss": 0.22431370615959167, "global_step": 65032, "epoch": 730, "lr": 8.664427419911383e-05}
{"train_loss": 0.26234716176986694, "global_step": 65033, "epoch": 730, "lr": 8.664387977168308e-05}
{"train_loss": 0.24312175810337067, "global_step": 65034, "epoch": 730, "lr": 8.664348533932601e-05}
{"train_loss": 0.20729279518127441, "global_step": 65035, "epoch": 730, "lr": 8.664309090204268e-05}
{"train_loss": 0.23609603941440582, "global_step": 65036, "epoch": 730, "lr": 8.664269645983311e-05}
{"train_loss": 0.22937846183776855, "global_step": 65037, "epoch": 730, "lr": 8.66423020126974e-05}
{"train_loss": 0.3119860291481018, "global_step": 65038, "epoch": 730, "lr": 8.664190756063558e-05}
{"train_loss": 0.22064024209976196, "global_step": 65039, "epoch": 730, "lr": 8.66415131036477e-05}
{"train_loss": 0.27400240302085876, "global_step": 65040, "epoch": 730, "lr": 8.66411186417338e-05}
{"train_loss": 0.22824548184871674, "global_step": 65041, "epoch": 730, "lr": 8.664072417489396e-05}
{"train_loss": 0.22030502557754517, "global_step": 65042, "epoch": 730, "lr": 8.664032970312823e-05}
{"train_loss": 0.2064174860715866, "global_step": 65043, "epoch": 730, "lr": 8.663993522643663e-05}
{"train_loss": 0.2849605083465576, "global_step": 65044, "epoch": 730, "lr": 8.663954074481926e-05}
{"train_loss": 0.22573554515838623, "global_step": 65045, "epoch": 730, "lr": 8.663914625827614e-05}
{"train_loss": 0.23580148816108704, "global_step": 65046, "epoch": 730, "lr": 8.663875176680733e-05}
{"train_loss": 0.17665940523147583, "global_step": 65047, "epoch": 730, "lr": 8.66383572704129e-05}
{"train_loss": 0.22987595200538635, "global_step": 65048, "epoch": 730, "lr": 8.663796276909286e-05}
{"train_loss": 0.2277640998363495, "global_step": 65049, "epoch": 730, "lr": 8.663756826284732e-05}
{"train_loss": 0.22320136427879333, "global_step": 65050, "epoch": 730, "lr": 8.66371737516763e-05}
{"train_loss": 0.33840906620025635, "global_step": 65051, "epoch": 730, "lr": 8.663677923557985e-05}
{"train_loss": 0.20414918661117554, "global_step": 65052, "epoch": 730, "lr": 8.663638471455804e-05}
{"train_loss": 0.2005014419555664, "global_step": 65053, "epoch": 730, "lr": 8.663599018861091e-05}
{"train_loss": 0.28888845443725586, "global_step": 65054, "epoch": 730, "lr": 8.663559565773852e-05}
{"train_loss": 0.2463294267654419, "global_step": 65055, "epoch": 730, "lr": 8.663520112194092e-05}
{"train_loss": 0.29193925857543945, "global_step": 65056, "epoch": 730, "lr": 8.663480658121817e-05}
{"train_loss": 0.19434691965579987, "global_step": 65057, "epoch": 730, "lr": 8.66344120355703e-05}
{"train_loss": 0.23516408370786837, "global_step": 65058, "epoch": 730, "lr": 8.66340174849974e-05, "val_loss": 3.5282957553863525, "train_action_mse_error": 18.702457427978516}
{"train_loss": 0.20818205177783966, "global_step": 65059, "epoch": 731, "lr": 8.663362292949951e-05}
{"train_loss": 0.21369683742523193, "global_step": 65060, "epoch": 731, "lr": 8.663322836907664e-05}
{"train_loss": 0.2618357837200165, "global_step": 65061, "epoch": 731, "lr": 8.663283380372891e-05}
{"train_loss": 0.16249263286590576, "global_step": 65062, "epoch": 731, "lr": 8.663243923345632e-05}
{"train_loss": 0.2788741886615753, "global_step": 65063, "epoch": 731, "lr": 8.663204465825896e-05}
{"train_loss": 0.1868860274553299, "global_step": 65064, "epoch": 731, "lr": 8.663165007813687e-05}
{"train_loss": 0.3555075228214264, "global_step": 65065, "epoch": 731, "lr": 8.66312554930901e-05}
{"train_loss": 0.19375595450401306, "global_step": 65066, "epoch": 731, "lr": 8.66308609031187e-05}
{"train_loss": 0.19247740507125854, "global_step": 65067, "epoch": 731, "lr": 8.663046630822271e-05}
{"train_loss": 0.1851917803287506, "global_step": 65068, "epoch": 731, "lr": 8.663007170840223e-05}
{"train_loss": 0.22820241749286652, "global_step": 65069, "epoch": 731, "lr": 8.662967710365725e-05}
{"train_loss": 0.2084791362285614, "global_step": 65070, "epoch": 731, "lr": 8.662928249398789e-05}
{"train_loss": 0.16124196350574493, "global_step": 65071, "epoch": 731, "lr": 8.662888787939416e-05}
{"train_loss": 0.24258506298065186, "global_step": 65072, "epoch": 731, "lr": 8.662849325987613e-05}
{"train_loss": 0.28048160672187805, "global_step": 65073, "epoch": 731, "lr": 8.662809863543382e-05}
{"train_loss": 0.18363572657108307, "global_step": 65074, "epoch": 731, "lr": 8.662770400606732e-05}
{"train_loss": 0.26684796810150146, "global_step": 65075, "epoch": 731, "lr": 8.662730937177667e-05}
{"train_loss": 0.1939842849969864, "global_step": 65076, "epoch": 731, "lr": 8.662691473256192e-05}
{"train_loss": 0.21613134443759918, "global_step": 65077, "epoch": 731, "lr": 8.662652008842314e-05}
{"train_loss": 0.20213966071605682, "global_step": 65078, "epoch": 731, "lr": 8.662612543936038e-05}
{"train_loss": 0.2469712793827057, "global_step": 65079, "epoch": 731, "lr": 8.662573078537365e-05}
{"train_loss": 0.20593467354774475, "global_step": 65080, "epoch": 731, "lr": 8.662533612646308e-05}
{"train_loss": 0.2627881169319153, "global_step": 65081, "epoch": 731, "lr": 8.662494146262864e-05}
{"train_loss": 0.22454169392585754, "global_step": 65082, "epoch": 731, "lr": 8.662454679387045e-05}
{"train_loss": 0.2800077795982361, "global_step": 65083, "epoch": 731, "lr": 8.662415212018853e-05}
{"train_loss": 0.22994685173034668, "global_step": 65084, "epoch": 731, "lr": 8.662375744158293e-05}
{"train_loss": 0.2109212726354599, "global_step": 65085, "epoch": 731, "lr": 8.662336275805373e-05}
{"train_loss": 0.24996605515480042, "global_step": 65086, "epoch": 731, "lr": 8.662296806960095e-05}
{"train_loss": 0.20722343027591705, "global_step": 65087, "epoch": 731, "lr": 8.662257337622466e-05}
{"train_loss": 0.22646749019622803, "global_step": 65088, "epoch": 731, "lr": 8.662217867792491e-05}
{"train_loss": 0.252662718296051, "global_step": 65089, "epoch": 731, "lr": 8.662178397470176e-05}
{"train_loss": 0.1960308700799942, "global_step": 65090, "epoch": 731, "lr": 8.662138926655525e-05}
{"train_loss": 0.21556343138217926, "global_step": 65091, "epoch": 731, "lr": 8.662099455348545e-05}
{"train_loss": 0.19334173202514648, "global_step": 65092, "epoch": 731, "lr": 8.66205998354924e-05}
{"train_loss": 0.17479825019836426, "global_step": 65093, "epoch": 731, "lr": 8.662020511257616e-05}
{"train_loss": 0.17382535338401794, "global_step": 65094, "epoch": 731, "lr": 8.661981038473678e-05}
{"train_loss": 0.23355399072170258, "global_step": 65095, "epoch": 731, "lr": 8.661941565197431e-05}
{"train_loss": 0.26333996653556824, "global_step": 65096, "epoch": 731, "lr": 8.661902091428879e-05}
{"train_loss": 0.19747714698314667, "global_step": 65097, "epoch": 731, "lr": 8.66186261716803e-05}
{"train_loss": 0.23047292232513428, "global_step": 65098, "epoch": 731, "lr": 8.66182314241489e-05}
{"train_loss": 0.2456243336200714, "global_step": 65099, "epoch": 731, "lr": 8.661783667169461e-05}
{"train_loss": 0.2935914695262909, "global_step": 65100, "epoch": 731, "lr": 8.66174419143175e-05}
{"train_loss": 0.244389608502388, "global_step": 65101, "epoch": 731, "lr": 8.661704715201762e-05}
{"train_loss": 0.13108383119106293, "global_step": 65102, "epoch": 731, "lr": 8.661665238479502e-05}
{"train_loss": 0.3190743923187256, "global_step": 65103, "epoch": 731, "lr": 8.661625761264977e-05}
{"train_loss": 0.2542821168899536, "global_step": 65104, "epoch": 731, "lr": 8.66158628355819e-05}
{"train_loss": 0.23632490634918213, "global_step": 65105, "epoch": 731, "lr": 8.661546805359146e-05}
{"train_loss": 0.1960076540708542, "global_step": 65106, "epoch": 731, "lr": 8.661507326667854e-05}
{"train_loss": 0.1311071664094925, "global_step": 65107, "epoch": 731, "lr": 8.661467847484317e-05}
{"train_loss": 0.2568713426589966, "global_step": 65108, "epoch": 731, "lr": 8.66142836780854e-05}
{"train_loss": 0.17256619036197662, "global_step": 65109, "epoch": 731, "lr": 8.661388887640527e-05}
{"train_loss": 0.12507371604442596, "global_step": 65110, "epoch": 731, "lr": 8.661349406980286e-05}
{"train_loss": 0.2341783493757248, "global_step": 65111, "epoch": 731, "lr": 8.661309925827821e-05}
{"train_loss": 0.12849034368991852, "global_step": 65112, "epoch": 731, "lr": 8.661270444183138e-05}
{"train_loss": 0.2651674449443817, "global_step": 65113, "epoch": 731, "lr": 8.661230962046241e-05}
{"train_loss": 0.19074274599552155, "global_step": 65114, "epoch": 731, "lr": 8.661191479417136e-05}
{"train_loss": 0.19105379283428192, "global_step": 65115, "epoch": 731, "lr": 8.66115199629583e-05}
{"train_loss": 0.15390056371688843, "global_step": 65116, "epoch": 731, "lr": 8.661112512682325e-05}
{"train_loss": 0.19286103546619415, "global_step": 65117, "epoch": 731, "lr": 8.66107302857663e-05}
{"train_loss": 0.31305283308029175, "global_step": 65118, "epoch": 731, "lr": 8.661033543978748e-05}
{"train_loss": 0.1743784099817276, "global_step": 65119, "epoch": 731, "lr": 8.660994058888683e-05}
{"train_loss": 0.16167855262756348, "global_step": 65120, "epoch": 731, "lr": 8.660954573306442e-05}
{"train_loss": 0.14195604622364044, "global_step": 65121, "epoch": 731, "lr": 8.660915087232032e-05}
{"train_loss": 0.18891143798828125, "global_step": 65122, "epoch": 731, "lr": 8.660875600665454e-05}
{"train_loss": 0.23252509534358978, "global_step": 65123, "epoch": 731, "lr": 8.660836113606719e-05}
{"train_loss": 0.18173441290855408, "global_step": 65124, "epoch": 731, "lr": 8.660796626055827e-05}
{"train_loss": 0.13202376663684845, "global_step": 65125, "epoch": 731, "lr": 8.660757138012787e-05}
{"train_loss": 0.1887046843767166, "global_step": 65126, "epoch": 731, "lr": 8.660717649477601e-05}
{"train_loss": 0.19149360060691833, "global_step": 65127, "epoch": 731, "lr": 8.660678160450278e-05}
{"train_loss": 0.24128466844558716, "global_step": 65128, "epoch": 731, "lr": 8.66063867093082e-05}
{"train_loss": 0.37259048223495483, "global_step": 65129, "epoch": 731, "lr": 8.660599180919235e-05}
{"train_loss": 0.19374679028987885, "global_step": 65130, "epoch": 731, "lr": 8.660559690415526e-05}
{"train_loss": 0.20524102449417114, "global_step": 65131, "epoch": 731, "lr": 8.660520199419699e-05}
{"train_loss": 0.18132300674915314, "global_step": 65132, "epoch": 731, "lr": 8.660480707931762e-05}
{"train_loss": 0.16459085047245026, "global_step": 65133, "epoch": 731, "lr": 8.660441215951715e-05}
{"train_loss": 0.2684670686721802, "global_step": 65134, "epoch": 731, "lr": 8.66040172347957e-05}
{"train_loss": 0.28686556220054626, "global_step": 65135, "epoch": 731, "lr": 8.660362230515325e-05}
{"train_loss": 0.2237323522567749, "global_step": 65136, "epoch": 731, "lr": 8.66032273705899e-05}
{"train_loss": 0.19717231392860413, "global_step": 65137, "epoch": 731, "lr": 8.660283243110569e-05}
{"train_loss": 0.2076304405927658, "global_step": 65138, "epoch": 731, "lr": 8.66024374867007e-05}
{"train_loss": 0.20024526119232178, "global_step": 65139, "epoch": 731, "lr": 8.660204253737493e-05}
{"train_loss": 0.22158366441726685, "global_step": 65140, "epoch": 731, "lr": 8.660164758312847e-05}
{"train_loss": 0.1650114506483078, "global_step": 65141, "epoch": 731, "lr": 8.660125262396137e-05}
{"train_loss": 0.2384016364812851, "global_step": 65142, "epoch": 731, "lr": 8.660085765987367e-05}
{"train_loss": 0.2240774929523468, "global_step": 65143, "epoch": 731, "lr": 8.660046269086543e-05}
{"train_loss": 0.14346253871917725, "global_step": 65144, "epoch": 731, "lr": 8.660006771693672e-05}
{"train_loss": 0.3310824930667877, "global_step": 65145, "epoch": 731, "lr": 8.659967273808756e-05}
{"train_loss": 0.327735036611557, "global_step": 65146, "epoch": 731, "lr": 8.659927775431801e-05}
{"train_loss": 0.21883433320549098, "global_step": 65147, "epoch": 731, "lr": 8.659888276562815e-05, "val_loss": 3.6356985569000244}
{"train_loss": 0.23904411494731903, "global_step": 65148, "epoch": 732, "lr": 8.659848777201802e-05}
{"train_loss": 0.18488776683807373, "global_step": 65149, "epoch": 732, "lr": 8.659809277348766e-05}
{"train_loss": 0.3331020176410675, "global_step": 65150, "epoch": 732, "lr": 8.659769777003713e-05}
{"train_loss": 0.2383168786764145, "global_step": 65151, "epoch": 732, "lr": 8.65973027616665e-05}
{"train_loss": 0.2920167148113251, "global_step": 65152, "epoch": 732, "lr": 8.65969077483758e-05}
{"train_loss": 0.26474592089653015, "global_step": 65153, "epoch": 732, "lr": 8.65965127301651e-05}
{"train_loss": 0.18775177001953125, "global_step": 65154, "epoch": 732, "lr": 8.659611770703444e-05}
{"train_loss": 0.1789172738790512, "global_step": 65155, "epoch": 732, "lr": 8.659572267898386e-05}
{"train_loss": 0.17524094879627228, "global_step": 65156, "epoch": 732, "lr": 8.659532764601346e-05}
{"train_loss": 0.23153266310691833, "global_step": 65157, "epoch": 732, "lr": 8.659493260812323e-05}
{"train_loss": 0.19494710862636566, "global_step": 65158, "epoch": 732, "lr": 8.659453756531329e-05}
{"train_loss": 0.1997195929288864, "global_step": 65159, "epoch": 732, "lr": 8.659414251758365e-05}
{"train_loss": 0.2645362615585327, "global_step": 65160, "epoch": 732, "lr": 8.659374746493436e-05}
{"train_loss": 0.2593320906162262, "global_step": 65161, "epoch": 732, "lr": 8.65933524073655e-05}
{"train_loss": 0.3241938650608063, "global_step": 65162, "epoch": 732, "lr": 8.65929573448771e-05}
{"train_loss": 0.2025182843208313, "global_step": 65163, "epoch": 732, "lr": 8.659256227746924e-05}
{"train_loss": 0.22238710522651672, "global_step": 65164, "epoch": 732, "lr": 8.659216720514194e-05}
{"train_loss": 0.23933616280555725, "global_step": 65165, "epoch": 732, "lr": 8.659177212789528e-05}
{"train_loss": 0.2840498685836792, "global_step": 65166, "epoch": 732, "lr": 8.65913770457293e-05}
{"train_loss": 0.2251604050397873, "global_step": 65167, "epoch": 732, "lr": 8.659098195864406e-05}
{"train_loss": 0.23765689134597778, "global_step": 65168, "epoch": 732, "lr": 8.65905868666396e-05}
{"train_loss": 0.21578966081142426, "global_step": 65169, "epoch": 732, "lr": 8.659019176971599e-05}
{"train_loss": 0.1936943084001541, "global_step": 65170, "epoch": 732, "lr": 8.658979666787325e-05}
{"train_loss": 0.18866120278835297, "global_step": 65171, "epoch": 732, "lr": 8.658940156111148e-05}
{"train_loss": 0.22663818299770355, "global_step": 65172, "epoch": 732, "lr": 8.65890064494307e-05}
{"train_loss": 0.23089614510536194, "global_step": 65173, "epoch": 732, "lr": 8.6588611332831e-05}
{"train_loss": 0.23339001834392548, "global_step": 65174, "epoch": 732, "lr": 8.65882162113124e-05}
{"train_loss": 0.19258351624011993, "global_step": 65175, "epoch": 732, "lr": 8.658782108487493e-05}
{"train_loss": 0.2767642140388489, "global_step": 65176, "epoch": 732, "lr": 8.65874259535187e-05}
{"train_loss": 0.23262791335582733, "global_step": 65177, "epoch": 732, "lr": 8.658703081724373e-05}
{"train_loss": 0.20811323821544647, "global_step": 65178, "epoch": 732, "lr": 8.658663567605007e-05}
{"train_loss": 0.29382842779159546, "global_step": 65179, "epoch": 732, "lr": 8.658624052993779e-05}
{"train_loss": 0.32476097345352173, "global_step": 65180, "epoch": 732, "lr": 8.658584537890693e-05}
{"train_loss": 0.19831956923007965, "global_step": 65181, "epoch": 732, "lr": 8.658545022295756e-05}
{"train_loss": 0.27258947491645813, "global_step": 65182, "epoch": 732, "lr": 8.658505506208972e-05}
{"train_loss": 0.19822385907173157, "global_step": 65183, "epoch": 732, "lr": 8.658465989630346e-05}
{"train_loss": 0.2406814843416214, "global_step": 65184, "epoch": 732, "lr": 8.658426472559885e-05}
{"train_loss": 0.2609384059906006, "global_step": 65185, "epoch": 732, "lr": 8.658386954997591e-05}
{"train_loss": 0.20945608615875244, "global_step": 65186, "epoch": 732, "lr": 8.658347436943473e-05}
{"train_loss": 0.273836612701416, "global_step": 65187, "epoch": 732, "lr": 8.658307918397534e-05}
{"train_loss": 0.24743372201919556, "global_step": 65188, "epoch": 732, "lr": 8.658268399359781e-05}
{"train_loss": 0.237858846783638, "global_step": 65189, "epoch": 732, "lr": 8.658228879830219e-05}
{"train_loss": 0.2506555914878845, "global_step": 65190, "epoch": 732, "lr": 8.658189359808851e-05}
{"train_loss": 0.23123611509799957, "global_step": 65191, "epoch": 732, "lr": 8.658149839295686e-05}
{"train_loss": 0.1941101849079132, "global_step": 65192, "epoch": 732, "lr": 8.658110318290726e-05}
{"train_loss": 0.19947391748428345, "global_step": 65193, "epoch": 732, "lr": 8.658070796793977e-05}
{"train_loss": 0.20529644191265106, "global_step": 65194, "epoch": 732, "lr": 8.658031274805447e-05}
{"train_loss": 0.21007587015628815, "global_step": 65195, "epoch": 732, "lr": 8.657991752325136e-05}
{"train_loss": 0.1822022646665573, "global_step": 65196, "epoch": 732, "lr": 8.657952229353054e-05}
{"train_loss": 0.1679610162973404, "global_step": 65197, "epoch": 732, "lr": 8.657912705889207e-05}
{"train_loss": 0.2811219394207001, "global_step": 65198, "epoch": 732, "lr": 8.657873181933596e-05}
{"train_loss": 0.17647628486156464, "global_step": 65199, "epoch": 732, "lr": 8.65783365748623e-05}
{"train_loss": 0.18063826858997345, "global_step": 65200, "epoch": 732, "lr": 8.657794132547112e-05}
{"train_loss": 0.2596386969089508, "global_step": 65201, "epoch": 732, "lr": 8.657754607116249e-05}
{"train_loss": 0.22770214080810547, "global_step": 65202, "epoch": 732, "lr": 8.657715081193645e-05}
{"train_loss": 0.2391068935394287, "global_step": 65203, "epoch": 732, "lr": 8.657675554779304e-05}
{"train_loss": 0.2460010051727295, "global_step": 65204, "epoch": 732, "lr": 8.657636027873235e-05}
{"train_loss": 0.16291095316410065, "global_step": 65205, "epoch": 732, "lr": 8.657596500475441e-05}
{"train_loss": 0.2169542908668518, "global_step": 65206, "epoch": 732, "lr": 8.657556972585927e-05}
{"train_loss": 0.20678876340389252, "global_step": 65207, "epoch": 732, "lr": 8.657517444204702e-05}
{"train_loss": 0.2356824278831482, "global_step": 65208, "epoch": 732, "lr": 8.657477915331764e-05}
{"train_loss": 0.1546797752380371, "global_step": 65209, "epoch": 732, "lr": 8.657438385967126e-05}
{"train_loss": 0.24562682211399078, "global_step": 65210, "epoch": 732, "lr": 8.657398856110789e-05}
{"train_loss": 0.20286428928375244, "global_step": 65211, "epoch": 732, "lr": 8.657359325762759e-05}
{"train_loss": 0.15880899131298065, "global_step": 65212, "epoch": 732, "lr": 8.657319794923043e-05}
{"train_loss": 0.2561020255088806, "global_step": 65213, "epoch": 732, "lr": 8.657280263591643e-05}
{"train_loss": 0.21827682852745056, "global_step": 65214, "epoch": 732, "lr": 8.657240731768565e-05}
{"train_loss": 0.25102749466896057, "global_step": 65215, "epoch": 732, "lr": 8.65720119945382e-05}
{"train_loss": 0.2588014304637909, "global_step": 65216, "epoch": 732, "lr": 8.657161666647404e-05}
{"train_loss": 0.27638763189315796, "global_step": 65217, "epoch": 732, "lr": 8.657122133349332e-05}
{"train_loss": 0.2484194040298462, "global_step": 65218, "epoch": 732, "lr": 8.657082599559601e-05}
{"train_loss": 0.21283814311027527, "global_step": 65219, "epoch": 732, "lr": 8.657043065278221e-05}
{"train_loss": 0.354872465133667, "global_step": 65220, "epoch": 732, "lr": 8.657003530505197e-05}
{"train_loss": 0.1748371571302414, "global_step": 65221, "epoch": 732, "lr": 8.656963995240531e-05}
{"train_loss": 0.19265015423297882, "global_step": 65222, "epoch": 732, "lr": 8.656924459484232e-05}
{"train_loss": 0.1759345829486847, "global_step": 65223, "epoch": 732, "lr": 8.656884923236305e-05}
{"train_loss": 0.2154470682144165, "global_step": 65224, "epoch": 732, "lr": 8.656845386496753e-05}
{"train_loss": 0.21085099875926971, "global_step": 65225, "epoch": 732, "lr": 8.656805849265583e-05}
{"train_loss": 0.20498749613761902, "global_step": 65226, "epoch": 732, "lr": 8.6567663115428e-05}
{"train_loss": 0.2213582694530487, "global_step": 65227, "epoch": 732, "lr": 8.656726773328411e-05}
{"train_loss": 0.2950598895549774, "global_step": 65228, "epoch": 732, "lr": 8.656687234622418e-05}
{"train_loss": 0.2661019265651703, "global_step": 65229, "epoch": 732, "lr": 8.656647695424828e-05}
{"train_loss": 0.21556055545806885, "global_step": 65230, "epoch": 732, "lr": 8.656608155735647e-05}
{"train_loss": 0.20937573909759521, "global_step": 65231, "epoch": 732, "lr": 8.65656861555488e-05}
{"train_loss": 0.19725626707077026, "global_step": 65232, "epoch": 732, "lr": 8.65652907488253e-05}
{"train_loss": 0.2620406150817871, "global_step": 65233, "epoch": 732, "lr": 8.656489533718606e-05}
{"train_loss": 0.22295571863651276, "global_step": 65234, "epoch": 732, "lr": 8.656449992063112e-05}
{"train_loss": 0.21261171996593475, "global_step": 65235, "epoch": 732, "lr": 8.65641044991605e-05}
{"train_loss": 0.22843658170673284, "global_step": 65236, "epoch": 732, "lr": 8.656370907277431e-05, "val_loss": 3.6233136653900146}
{"train_loss": 0.24806788563728333, "global_step": 65237, "epoch": 733, "lr": 8.656331364147257e-05}
{"train_loss": 0.21384228765964508, "global_step": 65238, "epoch": 733, "lr": 8.656291820525532e-05}
{"train_loss": 0.3050130307674408, "global_step": 65239, "epoch": 733, "lr": 8.656252276412266e-05}
{"train_loss": 0.14688025414943695, "global_step": 65240, "epoch": 733, "lr": 8.65621273180746e-05}
{"train_loss": 0.17326921224594116, "global_step": 65241, "epoch": 733, "lr": 8.656173186711121e-05}
{"train_loss": 0.32465696334838867, "global_step": 65242, "epoch": 733, "lr": 8.656133641123253e-05}
{"train_loss": 0.23521564900875092, "global_step": 65243, "epoch": 733, "lr": 8.656094095043863e-05}
{"train_loss": 0.2585705816745758, "global_step": 65244, "epoch": 733, "lr": 8.656054548472955e-05}
{"train_loss": 0.17639948427677155, "global_step": 65245, "epoch": 733, "lr": 8.656015001410537e-05}
{"train_loss": 0.2818012237548828, "global_step": 65246, "epoch": 733, "lr": 8.65597545385661e-05}
{"train_loss": 0.22641165554523468, "global_step": 65247, "epoch": 733, "lr": 8.655935905811184e-05}
{"train_loss": 0.12452829629182816, "global_step": 65248, "epoch": 733, "lr": 8.65589635727426e-05}
{"train_loss": 0.19113041460514069, "global_step": 65249, "epoch": 733, "lr": 8.655856808245847e-05}
{"train_loss": 0.2943251430988312, "global_step": 65250, "epoch": 733, "lr": 8.655817258725947e-05}
{"train_loss": 0.23253177106380463, "global_step": 65251, "epoch": 733, "lr": 8.655777708714568e-05}
{"train_loss": 0.2976697087287903, "global_step": 65252, "epoch": 733, "lr": 8.655738158211715e-05}
{"train_loss": 0.18381193280220032, "global_step": 65253, "epoch": 733, "lr": 8.655698607217391e-05}
{"train_loss": 0.29437559843063354, "global_step": 65254, "epoch": 733, "lr": 8.655659055731604e-05}
{"train_loss": 0.27274951338768005, "global_step": 65255, "epoch": 733, "lr": 8.655619503754356e-05}
{"train_loss": 0.23709462583065033, "global_step": 65256, "epoch": 733, "lr": 8.655579951285657e-05}
{"train_loss": 0.18224811553955078, "global_step": 65257, "epoch": 733, "lr": 8.655540398325508e-05}
{"train_loss": 0.181280255317688, "global_step": 65258, "epoch": 733, "lr": 8.655500844873919e-05}
{"train_loss": 0.2994700074195862, "global_step": 65259, "epoch": 733, "lr": 8.655461290930889e-05}
{"train_loss": 0.22466573119163513, "global_step": 65260, "epoch": 733, "lr": 8.655421736496428e-05}
{"train_loss": 0.31919705867767334, "global_step": 65261, "epoch": 733, "lr": 8.655382181570542e-05}
{"train_loss": 0.16920648515224457, "global_step": 65262, "epoch": 733, "lr": 8.65534262615323e-05}
{"train_loss": 0.2111910879611969, "global_step": 65263, "epoch": 733, "lr": 8.655303070244507e-05}
{"train_loss": 0.2607230544090271, "global_step": 65264, "epoch": 733, "lr": 8.655263513844369e-05}
{"train_loss": 0.27967244386672974, "global_step": 65265, "epoch": 733, "lr": 8.655223956952827e-05}
{"train_loss": 0.20494824647903442, "global_step": 65266, "epoch": 733, "lr": 8.655184399569886e-05}
{"train_loss": 0.1856633424758911, "global_step": 65267, "epoch": 733, "lr": 8.655144841695548e-05}
{"train_loss": 0.1932842880487442, "global_step": 65268, "epoch": 733, "lr": 8.655105283329821e-05}
{"train_loss": 0.21708688139915466, "global_step": 65269, "epoch": 733, "lr": 8.655065724472711e-05}
{"train_loss": 0.2317638397216797, "global_step": 65270, "epoch": 733, "lr": 8.65502616512422e-05}
{"train_loss": 0.24080675840377808, "global_step": 65271, "epoch": 733, "lr": 8.654986605284355e-05}
{"train_loss": 0.19083908200263977, "global_step": 65272, "epoch": 733, "lr": 8.654947044953123e-05}
{"train_loss": 0.35114797949790955, "global_step": 65273, "epoch": 733, "lr": 8.654907484130527e-05}
{"train_loss": 0.3085552752017975, "global_step": 65274, "epoch": 733, "lr": 8.654867922816574e-05}
{"train_loss": 0.2797209918498993, "global_step": 65275, "epoch": 733, "lr": 8.654828361011269e-05}
{"train_loss": 0.32348188757896423, "global_step": 65276, "epoch": 733, "lr": 8.654788798714615e-05}
{"train_loss": 0.30714958906173706, "global_step": 65277, "epoch": 733, "lr": 8.65474923592662e-05}
{"train_loss": 0.24747022986412048, "global_step": 65278, "epoch": 733, "lr": 8.65470967264729e-05}
{"train_loss": 0.28920120000839233, "global_step": 65279, "epoch": 733, "lr": 8.654670108876627e-05}
{"train_loss": 0.27482354640960693, "global_step": 65280, "epoch": 733, "lr": 8.65463054461464e-05}
{"train_loss": 0.2729687988758087, "global_step": 65281, "epoch": 733, "lr": 8.65459097986133e-05}
{"train_loss": 0.21426710486412048, "global_step": 65282, "epoch": 733, "lr": 8.654551414616707e-05}
{"train_loss": 0.239937424659729, "global_step": 65283, "epoch": 733, "lr": 8.654511848880772e-05}
{"train_loss": 0.2696574926376343, "global_step": 65284, "epoch": 733, "lr": 8.654472282653534e-05}
{"train_loss": 0.24877619743347168, "global_step": 65285, "epoch": 733, "lr": 8.654432715934996e-05}
{"train_loss": 0.26273113489151, "global_step": 65286, "epoch": 733, "lr": 8.654393148725165e-05}
{"train_loss": 0.2009160816669464, "global_step": 65287, "epoch": 733, "lr": 8.654353581024044e-05}
{"train_loss": 0.3609553575515747, "global_step": 65288, "epoch": 733, "lr": 8.654314012831641e-05}
{"train_loss": 0.20000039041042328, "global_step": 65289, "epoch": 733, "lr": 8.654274444147958e-05}
{"train_loss": 0.2073557823896408, "global_step": 65290, "epoch": 733, "lr": 8.654234874973004e-05}
{"train_loss": 0.1936291605234146, "global_step": 65291, "epoch": 733, "lr": 8.654195305306783e-05}
{"train_loss": 0.24841798841953278, "global_step": 65292, "epoch": 733, "lr": 8.6541557351493e-05}
{"train_loss": 0.2070242464542389, "global_step": 65293, "epoch": 733, "lr": 8.654116164500559e-05}
{"train_loss": 0.2632603943347931, "global_step": 65294, "epoch": 733, "lr": 8.654076593360568e-05}
{"train_loss": 0.19149905443191528, "global_step": 65295, "epoch": 733, "lr": 8.65403702172933e-05}
{"train_loss": 0.28162872791290283, "global_step": 65296, "epoch": 733, "lr": 8.653997449606852e-05}
{"train_loss": 0.20996685326099396, "global_step": 65297, "epoch": 733, "lr": 8.653957876993137e-05}
{"train_loss": 0.24122367799282074, "global_step": 65298, "epoch": 733, "lr": 8.653918303888194e-05}
{"train_loss": 0.18809141218662262, "global_step": 65299, "epoch": 733, "lr": 8.653878730292025e-05}
{"train_loss": 0.25356054306030273, "global_step": 65300, "epoch": 733, "lr": 8.653839156204637e-05}
{"train_loss": 0.249529629945755, "global_step": 65301, "epoch": 733, "lr": 8.653799581626035e-05}
{"train_loss": 0.2410251796245575, "global_step": 65302, "epoch": 733, "lr": 8.653760006556222e-05}
{"train_loss": 0.23692576587200165, "global_step": 65303, "epoch": 733, "lr": 8.65372043099521e-05}
{"train_loss": 0.3396632969379425, "global_step": 65304, "epoch": 733, "lr": 8.653680854942996e-05}
{"train_loss": 0.27918747067451477, "global_step": 65305, "epoch": 733, "lr": 8.653641278399591e-05}
{"train_loss": 0.32039904594421387, "global_step": 65306, "epoch": 733, "lr": 8.653601701364997e-05}
{"train_loss": 0.19687744975090027, "global_step": 65307, "epoch": 733, "lr": 8.65356212383922e-05}
{"train_loss": 0.16274680197238922, "global_step": 65308, "epoch": 733, "lr": 8.653522545822269e-05}
{"train_loss": 0.26808202266693115, "global_step": 65309, "epoch": 733, "lr": 8.653482967314145e-05}
{"train_loss": 0.24246668815612793, "global_step": 65310, "epoch": 733, "lr": 8.653443388314855e-05}
{"train_loss": 0.3392680287361145, "global_step": 65311, "epoch": 733, "lr": 8.653403808824403e-05}
{"train_loss": 0.30117687582969666, "global_step": 65312, "epoch": 733, "lr": 8.653364228842796e-05}
{"train_loss": 0.21626685559749603, "global_step": 65313, "epoch": 733, "lr": 8.65332464837004e-05}
{"train_loss": 0.27508068084716797, "global_step": 65314, "epoch": 733, "lr": 8.653285067406139e-05}
{"train_loss": 0.20136991143226624, "global_step": 65315, "epoch": 733, "lr": 8.653245485951095e-05}
{"train_loss": 0.3044232428073883, "global_step": 65316, "epoch": 733, "lr": 8.65320590400492e-05}
{"train_loss": 0.21407148241996765, "global_step": 65317, "epoch": 733, "lr": 8.653166321567615e-05}
{"train_loss": 0.2318507730960846, "global_step": 65318, "epoch": 733, "lr": 8.653126738639187e-05}
{"train_loss": 0.28034496307373047, "global_step": 65319, "epoch": 733, "lr": 8.65308715521964e-05}
{"train_loss": 0.22553479671478271, "global_step": 65320, "epoch": 733, "lr": 8.653047571308978e-05}
{"train_loss": 0.3296484649181366, "global_step": 65321, "epoch": 733, "lr": 8.653007986907212e-05}
{"train_loss": 0.2008587121963501, "global_step": 65322, "epoch": 733, "lr": 8.652968402014341e-05}
{"train_loss": 0.2676839232444763, "global_step": 65323, "epoch": 733, "lr": 8.652928816630375e-05}
{"train_loss": 0.2923894226551056, "global_step": 65324, "epoch": 733, "lr": 8.652889230755315e-05}
{"train_loss": 0.24571354226784758, "global_step": 65325, "epoch": 733, "lr": 8.652849644389169e-05, "val_loss": 3.443904399871826}
{"train_loss": 0.2779916524887085, "global_step": 65326, "epoch": 734, "lr": 8.652810057531943e-05}
{"train_loss": 0.23950523138046265, "global_step": 65327, "epoch": 734, "lr": 8.65277047018364e-05}
{"train_loss": 0.22644974291324615, "global_step": 65328, "epoch": 734, "lr": 8.652730882344268e-05}
{"train_loss": 0.26104313135147095, "global_step": 65329, "epoch": 734, "lr": 8.65269129401383e-05}
{"train_loss": 0.25856468081474304, "global_step": 65330, "epoch": 734, "lr": 8.652651705192332e-05}
{"train_loss": 0.232065349817276, "global_step": 65331, "epoch": 734, "lr": 8.652612115879779e-05}
{"train_loss": 0.16884325444698334, "global_step": 65332, "epoch": 734, "lr": 8.652572526076178e-05}
{"train_loss": 0.22896501421928406, "global_step": 65333, "epoch": 734, "lr": 8.652532935781532e-05}
{"train_loss": 0.23161816596984863, "global_step": 65334, "epoch": 734, "lr": 8.652493344995848e-05}
{"train_loss": 0.20746460556983948, "global_step": 65335, "epoch": 734, "lr": 8.652453753719131e-05}
{"train_loss": 0.29186949133872986, "global_step": 65336, "epoch": 734, "lr": 8.652414161951386e-05}
{"train_loss": 0.24440689384937286, "global_step": 65337, "epoch": 734, "lr": 8.652374569692619e-05}
{"train_loss": 0.22657500207424164, "global_step": 65338, "epoch": 734, "lr": 8.652334976942836e-05}
{"train_loss": 0.23489220440387726, "global_step": 65339, "epoch": 734, "lr": 8.652295383702038e-05}
{"train_loss": 0.1985534280538559, "global_step": 65340, "epoch": 734, "lr": 8.652255789970234e-05}
{"train_loss": 0.2576846480369568, "global_step": 65341, "epoch": 734, "lr": 8.65221619574743e-05}
{"train_loss": 0.20403358340263367, "global_step": 65342, "epoch": 734, "lr": 8.652176601033629e-05}
{"train_loss": 0.2330055683851242, "global_step": 65343, "epoch": 734, "lr": 8.652137005828838e-05}
{"train_loss": 0.18459971249103546, "global_step": 65344, "epoch": 734, "lr": 8.652097410133062e-05}
{"train_loss": 0.21827837824821472, "global_step": 65345, "epoch": 734, "lr": 8.652057813946305e-05}
{"train_loss": 0.31705281138420105, "global_step": 65346, "epoch": 734, "lr": 8.652018217268574e-05}
{"train_loss": 0.20967797935009003, "global_step": 65347, "epoch": 734, "lr": 8.651978620099873e-05}
{"train_loss": 0.1904069483280182, "global_step": 65348, "epoch": 734, "lr": 8.651939022440208e-05}
{"train_loss": 0.15314197540283203, "global_step": 65349, "epoch": 734, "lr": 8.651899424289585e-05}
{"train_loss": 0.23427939414978027, "global_step": 65350, "epoch": 734, "lr": 8.651859825648008e-05}
{"train_loss": 0.2715276777744293, "global_step": 65351, "epoch": 734, "lr": 8.651820226515483e-05}
{"train_loss": 0.19675090909004211, "global_step": 65352, "epoch": 734, "lr": 8.651780626892017e-05}
{"train_loss": 0.20036564767360687, "global_step": 65353, "epoch": 734, "lr": 8.651741026777613e-05}
{"train_loss": 0.18982529640197754, "global_step": 65354, "epoch": 734, "lr": 8.651701426172276e-05}
{"train_loss": 0.24463389813899994, "global_step": 65355, "epoch": 734, "lr": 8.651661825076011e-05}
{"train_loss": 0.280484139919281, "global_step": 65356, "epoch": 734, "lr": 8.651622223488827e-05}
{"train_loss": 0.24099920690059662, "global_step": 65357, "epoch": 734, "lr": 8.651582621410726e-05}
{"train_loss": 0.1736443042755127, "global_step": 65358, "epoch": 734, "lr": 8.651543018841715e-05}
{"train_loss": 0.24476750195026398, "global_step": 65359, "epoch": 734, "lr": 8.651503415781797e-05}
{"train_loss": 0.20003551244735718, "global_step": 65360, "epoch": 734, "lr": 8.65146381223098e-05}
{"train_loss": 0.22346539795398712, "global_step": 65361, "epoch": 734, "lr": 8.651424208189268e-05}
{"train_loss": 0.22949379682540894, "global_step": 65362, "epoch": 734, "lr": 8.651384603656665e-05}
{"train_loss": 0.21154117584228516, "global_step": 65363, "epoch": 734, "lr": 8.65134499863318e-05}
{"train_loss": 0.16862091422080994, "global_step": 65364, "epoch": 734, "lr": 8.651305393118816e-05}
{"train_loss": 0.15056787431240082, "global_step": 65365, "epoch": 734, "lr": 8.651265787113577e-05}
{"train_loss": 0.27223068475723267, "global_step": 65366, "epoch": 734, "lr": 8.651226180617471e-05}
{"train_loss": 0.254730761051178, "global_step": 65367, "epoch": 734, "lr": 8.6511865736305e-05}
{"train_loss": 0.27768248319625854, "global_step": 65368, "epoch": 734, "lr": 8.651146966152675e-05}
{"train_loss": 0.21508267521858215, "global_step": 65369, "epoch": 734, "lr": 8.651107358183995e-05}
{"train_loss": 0.2015640139579773, "global_step": 65370, "epoch": 734, "lr": 8.651067749724469e-05}
{"train_loss": 0.21562063694000244, "global_step": 65371, "epoch": 734, "lr": 8.651028140774102e-05}
{"train_loss": 0.22563490271568298, "global_step": 65372, "epoch": 734, "lr": 8.650988531332899e-05}
{"train_loss": 0.24560923874378204, "global_step": 65373, "epoch": 734, "lr": 8.650948921400864e-05}
{"train_loss": 0.30815207958221436, "global_step": 65374, "epoch": 734, "lr": 8.650909310978004e-05}
{"train_loss": 0.2894101142883301, "global_step": 65375, "epoch": 734, "lr": 8.650869700064323e-05}
{"train_loss": 0.27554261684417725, "global_step": 65376, "epoch": 734, "lr": 8.650830088659828e-05}
{"train_loss": 0.20601744949817657, "global_step": 65377, "epoch": 734, "lr": 8.650790476764523e-05}
{"train_loss": 0.18443511426448822, "global_step": 65378, "epoch": 734, "lr": 8.650750864378415e-05}
{"train_loss": 0.1931421458721161, "global_step": 65379, "epoch": 734, "lr": 8.650711251501505e-05}
{"train_loss": 0.19353243708610535, "global_step": 65380, "epoch": 734, "lr": 8.650671638133804e-05}
{"train_loss": 0.38251322507858276, "global_step": 65381, "epoch": 734, "lr": 8.650632024275313e-05}
{"train_loss": 0.23604750633239746, "global_step": 65382, "epoch": 734, "lr": 8.650592409926041e-05}
{"train_loss": 0.22616267204284668, "global_step": 65383, "epoch": 734, "lr": 8.65055279508599e-05}
{"train_loss": 0.293343186378479, "global_step": 65384, "epoch": 734, "lr": 8.650513179755166e-05}
{"train_loss": 0.1732826828956604, "global_step": 65385, "epoch": 734, "lr": 8.650473563933577e-05}
{"train_loss": 0.27529996633529663, "global_step": 65386, "epoch": 734, "lr": 8.650433947621225e-05}
{"train_loss": 0.2296077162027359, "global_step": 65387, "epoch": 734, "lr": 8.650394330818117e-05}
{"train_loss": 0.2525462806224823, "global_step": 65388, "epoch": 734, "lr": 8.650354713524258e-05}
{"train_loss": 0.28361472487449646, "global_step": 65389, "epoch": 734, "lr": 8.650315095739653e-05}
{"train_loss": 0.2149994969367981, "global_step": 65390, "epoch": 734, "lr": 8.650275477464307e-05}
{"train_loss": 0.30198991298675537, "global_step": 65391, "epoch": 734, "lr": 8.650235858698228e-05}
{"train_loss": 0.23584511876106262, "global_step": 65392, "epoch": 734, "lr": 8.650196239441417e-05}
{"train_loss": 0.2153952717781067, "global_step": 65393, "epoch": 734, "lr": 8.650156619693884e-05}
{"train_loss": 0.33051568269729614, "global_step": 65394, "epoch": 734, "lr": 8.650116999455629e-05}
{"train_loss": 0.22001080214977264, "global_step": 65395, "epoch": 734, "lr": 8.650077378726662e-05}
{"train_loss": 0.32943177223205566, "global_step": 65396, "epoch": 734, "lr": 8.650037757506987e-05}
{"train_loss": 0.20258216559886932, "global_step": 65397, "epoch": 734, "lr": 8.649998135796608e-05}
{"train_loss": 0.22067111730575562, "global_step": 65398, "epoch": 734, "lr": 8.649958513595531e-05}
{"train_loss": 0.3222343325614929, "global_step": 65399, "epoch": 734, "lr": 8.649918890903764e-05}
{"train_loss": 0.1817064881324768, "global_step": 65400, "epoch": 734, "lr": 8.649879267721306e-05}
{"train_loss": 0.16001597046852112, "global_step": 65401, "epoch": 734, "lr": 8.64983964404817e-05}
{"train_loss": 0.27497631311416626, "global_step": 65402, "epoch": 734, "lr": 8.649800019884354e-05}
{"train_loss": 0.1880262792110443, "global_step": 65403, "epoch": 734, "lr": 8.649760395229869e-05}
{"train_loss": 0.17528755962848663, "global_step": 65404, "epoch": 734, "lr": 8.649720770084717e-05}
{"train_loss": 0.23060569167137146, "global_step": 65405, "epoch": 734, "lr": 8.649681144448907e-05}
{"train_loss": 0.3394791781902313, "global_step": 65406, "epoch": 734, "lr": 8.64964151832244e-05}
{"train_loss": 0.19090381264686584, "global_step": 65407, "epoch": 734, "lr": 8.649601891705323e-05}
{"train_loss": 0.2155740112066269, "global_step": 65408, "epoch": 734, "lr": 8.649562264597562e-05}
{"train_loss": 0.1984528750181198, "global_step": 65409, "epoch": 734, "lr": 8.649522636999162e-05}
{"train_loss": 0.25614941120147705, "global_step": 65410, "epoch": 734, "lr": 8.649483008910129e-05}
{"train_loss": 0.24874967336654663, "global_step": 65411, "epoch": 734, "lr": 8.649443380330464e-05}
{"train_loss": 0.20936790108680725, "global_step": 65412, "epoch": 734, "lr": 8.649403751260179e-05}
{"train_loss": 0.2761448323726654, "global_step": 65413, "epoch": 734, "lr": 8.649364121699275e-05}
{"train_loss": 0.2337129743916265, "global_step": 65414, "epoch": 734, "lr": 8.649324491647759e-05, "val_loss": 3.592466115951538}
{"train_loss": 0.21216998994350433, "global_step": 65415, "epoch": 735, "lr": 8.649284861105636e-05}
{"train_loss": 0.2780185639858246, "global_step": 65416, "epoch": 735, "lr": 8.649245230072911e-05}
{"train_loss": 0.2130114734172821, "global_step": 65417, "epoch": 735, "lr": 8.649205598549589e-05}
{"train_loss": 0.2335212379693985, "global_step": 65418, "epoch": 735, "lr": 8.649165966535676e-05}
{"train_loss": 0.20363177359104156, "global_step": 65419, "epoch": 735, "lr": 8.649126334031177e-05}
{"train_loss": 0.2008586972951889, "global_step": 65420, "epoch": 735, "lr": 8.649086701036097e-05}
{"train_loss": 0.20555347204208374, "global_step": 65421, "epoch": 735, "lr": 8.649047067550442e-05}
{"train_loss": 0.19941671192646027, "global_step": 65422, "epoch": 735, "lr": 8.649007433574219e-05}
{"train_loss": 0.22077059745788574, "global_step": 65423, "epoch": 735, "lr": 8.64896779910743e-05}
{"train_loss": 0.09198257327079773, "global_step": 65424, "epoch": 735, "lr": 8.64892816415008e-05}
{"train_loss": 0.22412416338920593, "global_step": 65425, "epoch": 735, "lr": 8.648888528702178e-05}
{"train_loss": 0.18048657476902008, "global_step": 65426, "epoch": 735, "lr": 8.648848892763728e-05}
{"train_loss": 0.18822944164276123, "global_step": 65427, "epoch": 735, "lr": 8.648809256334733e-05}
{"train_loss": 0.12887299060821533, "global_step": 65428, "epoch": 735, "lr": 8.648769619415201e-05}
{"train_loss": 0.21378761529922485, "global_step": 65429, "epoch": 735, "lr": 8.648729982005137e-05}
{"train_loss": 0.2521228790283203, "global_step": 65430, "epoch": 735, "lr": 8.648690344104544e-05}
{"train_loss": 0.23387354612350464, "global_step": 65431, "epoch": 735, "lr": 8.64865070571343e-05}
{"train_loss": 0.25260353088378906, "global_step": 65432, "epoch": 735, "lr": 8.6486110668318e-05}
{"train_loss": 0.2472773641347885, "global_step": 65433, "epoch": 735, "lr": 8.648571427459658e-05}
{"train_loss": 0.27731019258499146, "global_step": 65434, "epoch": 735, "lr": 8.64853178759701e-05}
{"train_loss": 0.27291691303253174, "global_step": 65435, "epoch": 735, "lr": 8.648492147243862e-05}
{"train_loss": 0.23956775665283203, "global_step": 65436, "epoch": 735, "lr": 8.648452506400217e-05}
{"train_loss": 0.23326712846755981, "global_step": 65437, "epoch": 735, "lr": 8.648412865066085e-05}
{"train_loss": 0.262016236782074, "global_step": 65438, "epoch": 735, "lr": 8.648373223241465e-05}
{"train_loss": 0.1876598596572876, "global_step": 65439, "epoch": 735, "lr": 8.648333580926369e-05}
{"train_loss": 0.17562559247016907, "global_step": 65440, "epoch": 735, "lr": 8.648293938120797e-05}
{"train_loss": 0.2993006110191345, "global_step": 65441, "epoch": 735, "lr": 8.648254294824756e-05}
{"train_loss": 0.3692031502723694, "global_step": 65442, "epoch": 735, "lr": 8.648214651038254e-05}
{"train_loss": 0.1851390302181244, "global_step": 65443, "epoch": 735, "lr": 8.648175006761292e-05}
{"train_loss": 0.26005232334136963, "global_step": 65444, "epoch": 735, "lr": 8.648135361993877e-05}
{"train_loss": 0.20587077736854553, "global_step": 65445, "epoch": 735, "lr": 8.648095716736015e-05}
{"train_loss": 0.2089017927646637, "global_step": 65446, "epoch": 735, "lr": 8.648056070987713e-05}
{"train_loss": 0.23461145162582397, "global_step": 65447, "epoch": 735, "lr": 8.648016424748973e-05}
{"train_loss": 0.1913439929485321, "global_step": 65448, "epoch": 735, "lr": 8.647976778019799e-05}
{"train_loss": 0.2359176129102707, "global_step": 65449, "epoch": 735, "lr": 8.647937130800203e-05}
{"train_loss": 0.22989831864833832, "global_step": 65450, "epoch": 735, "lr": 8.647897483090184e-05}
{"train_loss": 0.18737474083900452, "global_step": 65451, "epoch": 735, "lr": 8.64785783488975e-05}
{"train_loss": 0.26984328031539917, "global_step": 65452, "epoch": 735, "lr": 8.647818186198907e-05}
{"train_loss": 0.19795967638492584, "global_step": 65453, "epoch": 735, "lr": 8.647778537017659e-05}
{"train_loss": 0.19448484480381012, "global_step": 65454, "epoch": 735, "lr": 8.64773888734601e-05}
{"train_loss": 0.2523970305919647, "global_step": 65455, "epoch": 735, "lr": 8.647699237183968e-05}
{"train_loss": 0.3016562759876251, "global_step": 65456, "epoch": 735, "lr": 8.647659586531539e-05}
{"train_loss": 0.23303404450416565, "global_step": 65457, "epoch": 735, "lr": 8.647619935388724e-05}
{"train_loss": 0.29772764444351196, "global_step": 65458, "epoch": 735, "lr": 8.647580283755532e-05}
{"train_loss": 0.21824458241462708, "global_step": 65459, "epoch": 735, "lr": 8.647540631631968e-05}
{"train_loss": 0.2796004116535187, "global_step": 65460, "epoch": 735, "lr": 8.647500979018036e-05}
{"train_loss": 0.19304955005645752, "global_step": 65461, "epoch": 735, "lr": 8.647461325913741e-05}
{"train_loss": 0.21623378992080688, "global_step": 65462, "epoch": 735, "lr": 8.647421672319092e-05}
{"train_loss": 0.25272491574287415, "global_step": 65463, "epoch": 735, "lr": 8.647382018234089e-05}
{"train_loss": 0.2565065324306488, "global_step": 65464, "epoch": 735, "lr": 8.647342363658741e-05}
{"train_loss": 0.26428931951522827, "global_step": 65465, "epoch": 735, "lr": 8.647302708593053e-05}
{"train_loss": 0.2780779302120209, "global_step": 65466, "epoch": 735, "lr": 8.647263053037027e-05}
{"train_loss": 0.2291242480278015, "global_step": 65467, "epoch": 735, "lr": 8.647223396990673e-05}
{"train_loss": 0.2401147186756134, "global_step": 65468, "epoch": 735, "lr": 8.647183740453995e-05}
{"train_loss": 0.21215230226516724, "global_step": 65469, "epoch": 735, "lr": 8.647144083426996e-05}
{"train_loss": 0.3160223960876465, "global_step": 65470, "epoch": 735, "lr": 8.647104425909683e-05}
{"train_loss": 0.28052985668182373, "global_step": 65471, "epoch": 735, "lr": 8.647064767902062e-05}
{"train_loss": 0.2820441424846649, "global_step": 65472, "epoch": 735, "lr": 8.647025109404137e-05}
{"train_loss": 0.20414167642593384, "global_step": 65473, "epoch": 735, "lr": 8.646985450415914e-05}
{"train_loss": 0.23465867340564728, "global_step": 65474, "epoch": 735, "lr": 8.646945790937398e-05}
{"train_loss": 0.17476370930671692, "global_step": 65475, "epoch": 735, "lr": 8.646906130968595e-05}
{"train_loss": 0.17942842841148376, "global_step": 65476, "epoch": 735, "lr": 8.646866470509511e-05}
{"train_loss": 0.1844213902950287, "global_step": 65477, "epoch": 735, "lr": 8.646826809560149e-05}
{"train_loss": 0.24929919838905334, "global_step": 65478, "epoch": 735, "lr": 8.646787148120515e-05}
{"train_loss": 0.221698597073555, "global_step": 65479, "epoch": 735, "lr": 8.646747486190616e-05}
{"train_loss": 0.27088892459869385, "global_step": 65480, "epoch": 735, "lr": 8.646707823770456e-05}
{"train_loss": 0.2560214698314667, "global_step": 65481, "epoch": 735, "lr": 8.646668160860041e-05}
{"train_loss": 0.12022656202316284, "global_step": 65482, "epoch": 735, "lr": 8.646628497459375e-05}
{"train_loss": 0.21458609402179718, "global_step": 65483, "epoch": 735, "lr": 8.646588833568465e-05}
{"train_loss": 0.23773853480815887, "global_step": 65484, "epoch": 735, "lr": 8.646549169187315e-05}
{"train_loss": 0.20605765283107758, "global_step": 65485, "epoch": 735, "lr": 8.646509504315932e-05}
{"train_loss": 0.2558497488498688, "global_step": 65486, "epoch": 735, "lr": 8.646469838954318e-05}
{"train_loss": 0.22847294807434082, "global_step": 65487, "epoch": 735, "lr": 8.646430173102483e-05}
{"train_loss": 0.2099839299917221, "global_step": 65488, "epoch": 735, "lr": 8.646390506760429e-05}
{"train_loss": 0.21484249830245972, "global_step": 65489, "epoch": 735, "lr": 8.64635083992816e-05}
{"train_loss": 0.19192999601364136, "global_step": 65490, "epoch": 735, "lr": 8.646311172605688e-05}
{"train_loss": 0.22425659000873566, "global_step": 65491, "epoch": 735, "lr": 8.646271504793011e-05}
{"train_loss": 0.17418338358402252, "global_step": 65492, "epoch": 735, "lr": 8.646231836490138e-05}
{"train_loss": 0.30431318283081055, "global_step": 65493, "epoch": 735, "lr": 8.646192167697073e-05}
{"train_loss": 0.31294843554496765, "global_step": 65494, "epoch": 735, "lr": 8.646152498413822e-05}
{"train_loss": 0.2583944499492645, "global_step": 65495, "epoch": 735, "lr": 8.646112828640392e-05}
{"train_loss": 0.23482288420200348, "global_step": 65496, "epoch": 735, "lr": 8.646073158376785e-05}
{"train_loss": 0.19834424555301666, "global_step": 65497, "epoch": 735, "lr": 8.646033487623007e-05}
{"train_loss": 0.14728811383247375, "global_step": 65498, "epoch": 735, "lr": 8.645993816379065e-05}
{"train_loss": 0.1639774888753891, "global_step": 65499, "epoch": 735, "lr": 8.645954144644964e-05}
{"train_loss": 0.3487511873245239, "global_step": 65500, "epoch": 735, "lr": 8.645914472420709e-05}
{"train_loss": 0.3235189914703369, "global_step": 65501, "epoch": 735, "lr": 8.645874799706306e-05}
{"train_loss": 0.24428586661815643, "global_step": 65502, "epoch": 735, "lr": 8.645835126501758e-05}
{"train_loss": 0.22987183878261053, "global_step": 65503, "epoch": 735, "lr": 8.645795452807072e-05, "val_loss": 3.575404405593872, "train_action_mse_error": 29.391965866088867}
{"train_loss": 0.24377693235874176, "global_step": 65504, "epoch": 736, "lr": 8.645755778622253e-05}
{"train_loss": 0.23303572833538055, "global_step": 65505, "epoch": 736, "lr": 8.645716103947307e-05}
{"train_loss": 0.20747050642967224, "global_step": 65506, "epoch": 736, "lr": 8.64567642878224e-05}
{"train_loss": 0.242141455411911, "global_step": 65507, "epoch": 736, "lr": 8.645636753127055e-05}
{"train_loss": 0.24792338907718658, "global_step": 65508, "epoch": 736, "lr": 8.64559707698176e-05}
{"train_loss": 0.2782401740550995, "global_step": 65509, "epoch": 736, "lr": 8.645557400346357e-05}
{"train_loss": 0.19292719662189484, "global_step": 65510, "epoch": 736, "lr": 8.645517723220855e-05}
{"train_loss": 0.2494143694639206, "global_step": 65511, "epoch": 736, "lr": 8.645478045605255e-05}
{"train_loss": 0.16185367107391357, "global_step": 65512, "epoch": 736, "lr": 8.645438367499567e-05}
{"train_loss": 0.2021728903055191, "global_step": 65513, "epoch": 736, "lr": 8.645398688903795e-05}
{"train_loss": 0.18542765080928802, "global_step": 65514, "epoch": 736, "lr": 8.645359009817942e-05}
{"train_loss": 0.21943065524101257, "global_step": 65515, "epoch": 736, "lr": 8.645319330242015e-05}
{"train_loss": 0.23823970556259155, "global_step": 65516, "epoch": 736, "lr": 8.64527965017602e-05}
{"train_loss": 0.343161016702652, "global_step": 65517, "epoch": 736, "lr": 8.64523996961996e-05}
{"train_loss": 0.1677490621805191, "global_step": 65518, "epoch": 736, "lr": 8.645200288573845e-05}
{"train_loss": 0.19289523363113403, "global_step": 65519, "epoch": 736, "lr": 8.645160607037675e-05}
{"train_loss": 0.24546422064304352, "global_step": 65520, "epoch": 736, "lr": 8.645120925011457e-05}
{"train_loss": 0.1898665726184845, "global_step": 65521, "epoch": 736, "lr": 8.645081242495198e-05}
{"train_loss": 0.2021491974592209, "global_step": 65522, "epoch": 736, "lr": 8.645041559488903e-05}
{"train_loss": 0.2230544090270996, "global_step": 65523, "epoch": 736, "lr": 8.645001875992576e-05}
{"train_loss": 0.17513567209243774, "global_step": 65524, "epoch": 736, "lr": 8.644962192006221e-05}
{"train_loss": 0.1705780327320099, "global_step": 65525, "epoch": 736, "lr": 8.644922507529847e-05}
{"train_loss": 0.18428997695446014, "global_step": 65526, "epoch": 736, "lr": 8.644882822563458e-05}
{"train_loss": 0.27060461044311523, "global_step": 65527, "epoch": 736, "lr": 8.644843137107059e-05}
{"train_loss": 0.16037626564502716, "global_step": 65528, "epoch": 736, "lr": 8.644803451160654e-05}
{"train_loss": 0.13253553211688995, "global_step": 65529, "epoch": 736, "lr": 8.64476376472425e-05}
{"train_loss": 0.2794388234615326, "global_step": 65530, "epoch": 736, "lr": 8.644724077797852e-05}
{"train_loss": 0.21072453260421753, "global_step": 65531, "epoch": 736, "lr": 8.644684390381466e-05}
{"train_loss": 0.25303542613983154, "global_step": 65532, "epoch": 736, "lr": 8.644644702475095e-05}
{"train_loss": 0.33218273520469666, "global_step": 65533, "epoch": 736, "lr": 8.644605014078748e-05}
{"train_loss": 0.20091210305690765, "global_step": 65534, "epoch": 736, "lr": 8.644565325192426e-05}
{"train_loss": 0.21433956921100616, "global_step": 65535, "epoch": 736, "lr": 8.644525635816139e-05}
{"train_loss": 0.1847267597913742, "global_step": 65536, "epoch": 736, "lr": 8.644485945949889e-05}
{"train_loss": 0.23835298418998718, "global_step": 65537, "epoch": 736, "lr": 8.644446255593682e-05}
{"train_loss": 0.19964995980262756, "global_step": 65538, "epoch": 736, "lr": 8.644406564747524e-05}
{"train_loss": 0.2429230511188507, "global_step": 65539, "epoch": 736, "lr": 8.644366873411419e-05}
{"train_loss": 0.22206591069698334, "global_step": 65540, "epoch": 736, "lr": 8.644327181585375e-05}
{"train_loss": 0.1774357557296753, "global_step": 65541, "epoch": 736, "lr": 8.644287489269393e-05}
{"train_loss": 0.2555136978626251, "global_step": 65542, "epoch": 736, "lr": 8.644247796463483e-05}
{"train_loss": 0.15813802182674408, "global_step": 65543, "epoch": 736, "lr": 8.644208103167648e-05}
{"train_loss": 0.1351664960384369, "global_step": 65544, "epoch": 736, "lr": 8.644168409381894e-05}
{"train_loss": 0.22892612218856812, "global_step": 65545, "epoch": 736, "lr": 8.644128715106225e-05}
{"train_loss": 0.16994331777095795, "global_step": 65546, "epoch": 736, "lr": 8.644089020340649e-05}
{"train_loss": 0.23181778192520142, "global_step": 65547, "epoch": 736, "lr": 8.644049325085167e-05}
{"train_loss": 0.25429585576057434, "global_step": 65548, "epoch": 736, "lr": 8.644009629339789e-05}
{"train_loss": 0.16064172983169556, "global_step": 65549, "epoch": 736, "lr": 8.643969933104517e-05}
{"train_loss": 0.15189939737319946, "global_step": 65550, "epoch": 736, "lr": 8.643930236379358e-05}
{"train_loss": 0.20978090167045593, "global_step": 65551, "epoch": 736, "lr": 8.643890539164318e-05}
{"train_loss": 0.14263509213924408, "global_step": 65552, "epoch": 736, "lr": 8.643850841459402e-05}
{"train_loss": 0.2251860648393631, "global_step": 65553, "epoch": 736, "lr": 8.643811143264613e-05}
{"train_loss": 0.27314841747283936, "global_step": 65554, "epoch": 736, "lr": 8.643771444579957e-05}
{"train_loss": 0.19725391268730164, "global_step": 65555, "epoch": 736, "lr": 8.643731745405443e-05}
{"train_loss": 0.24401038885116577, "global_step": 65556, "epoch": 736, "lr": 8.643692045741071e-05}
{"train_loss": 0.18861724436283112, "global_step": 65557, "epoch": 736, "lr": 8.643652345586852e-05}
{"train_loss": 0.18176046013832092, "global_step": 65558, "epoch": 736, "lr": 8.643612644942787e-05}
{"train_loss": 0.18175004422664642, "global_step": 65559, "epoch": 736, "lr": 8.643572943808882e-05}
{"train_loss": 0.17461924254894257, "global_step": 65560, "epoch": 736, "lr": 8.643533242185143e-05}
{"train_loss": 0.1674962192773819, "global_step": 65561, "epoch": 736, "lr": 8.643493540071577e-05}
{"train_loss": 0.14935286343097687, "global_step": 65562, "epoch": 736, "lr": 8.643453837468186e-05}
{"train_loss": 0.17786620557308197, "global_step": 65563, "epoch": 736, "lr": 8.643414134374978e-05}
{"train_loss": 0.24534490704536438, "global_step": 65564, "epoch": 736, "lr": 8.643374430791957e-05}
{"train_loss": 0.19004887342453003, "global_step": 65565, "epoch": 736, "lr": 8.643334726719128e-05}
{"train_loss": 0.2122184783220291, "global_step": 65566, "epoch": 736, "lr": 8.643295022156498e-05}
{"train_loss": 0.1732967644929886, "global_step": 65567, "epoch": 736, "lr": 8.643255317104071e-05}
{"train_loss": 0.2534131407737732, "global_step": 65568, "epoch": 736, "lr": 8.643215611561852e-05}
{"train_loss": 0.2582120895385742, "global_step": 65569, "epoch": 736, "lr": 8.643175905529849e-05}
{"train_loss": 0.28093093633651733, "global_step": 65570, "epoch": 736, "lr": 8.643136199008064e-05}
{"train_loss": 0.27110445499420166, "global_step": 65571, "epoch": 736, "lr": 8.643096491996505e-05}
{"train_loss": 0.2933145761489868, "global_step": 65572, "epoch": 736, "lr": 8.643056784495175e-05}
{"train_loss": 0.28835973143577576, "global_step": 65573, "epoch": 736, "lr": 8.64301707650408e-05}
{"train_loss": 0.22657343745231628, "global_step": 65574, "epoch": 736, "lr": 8.642977368023226e-05}
{"train_loss": 0.1764446347951889, "global_step": 65575, "epoch": 736, "lr": 8.642937659052619e-05}
{"train_loss": 0.1860610842704773, "global_step": 65576, "epoch": 736, "lr": 8.642897949592264e-05}
{"train_loss": 0.18529953062534332, "global_step": 65577, "epoch": 736, "lr": 8.642858239642164e-05}
{"train_loss": 0.18767611682415009, "global_step": 65578, "epoch": 736, "lr": 8.642818529202325e-05}
{"train_loss": 0.16189679503440857, "global_step": 65579, "epoch": 736, "lr": 8.642778818272756e-05}
{"train_loss": 0.27249467372894287, "global_step": 65580, "epoch": 736, "lr": 8.642739106853459e-05}
{"train_loss": 0.2665460407733917, "global_step": 65581, "epoch": 736, "lr": 8.64269939494444e-05}
{"train_loss": 0.21058382093906403, "global_step": 65582, "epoch": 736, "lr": 8.642659682545706e-05}
{"train_loss": 0.1858399212360382, "global_step": 65583, "epoch": 736, "lr": 8.64261996965726e-05}
{"train_loss": 0.16881148517131805, "global_step": 65584, "epoch": 736, "lr": 8.642580256279106e-05}
{"train_loss": 0.2961036264896393, "global_step": 65585, "epoch": 736, "lr": 8.642540542411252e-05}
{"train_loss": 0.19833731651306152, "global_step": 65586, "epoch": 736, "lr": 8.642500828053705e-05}
{"train_loss": 0.13986743986606598, "global_step": 65587, "epoch": 736, "lr": 8.642461113206467e-05}
{"train_loss": 0.2518124580383301, "global_step": 65588, "epoch": 736, "lr": 8.642421397869545e-05}
{"train_loss": 0.30051711201667786, "global_step": 65589, "epoch": 736, "lr": 8.642381682042943e-05}
{"train_loss": 0.2798094153404236, "global_step": 65590, "epoch": 736, "lr": 8.642341965726667e-05}
{"train_loss": 0.1835487186908722, "global_step": 65591, "epoch": 736, "lr": 8.642302248920724e-05}
{"train_loss": 0.21557963882269485, "global_step": 65592, "epoch": 736, "lr": 8.642262531625117e-05, "val_loss": 3.8021187782287598}
{"train_loss": 0.23977769911289215, "global_step": 65593, "epoch": 737, "lr": 8.64222281383985e-05}
{"train_loss": 0.2623095214366913, "global_step": 65594, "epoch": 737, "lr": 8.642183095564933e-05}
{"train_loss": 0.2000395953655243, "global_step": 65595, "epoch": 737, "lr": 8.642143376800368e-05}
{"train_loss": 0.259977251291275, "global_step": 65596, "epoch": 737, "lr": 8.642103657546163e-05}
{"train_loss": 0.19666002690792084, "global_step": 65597, "epoch": 737, "lr": 8.64206393780232e-05}
{"train_loss": 0.16395848989486694, "global_step": 65598, "epoch": 737, "lr": 8.642024217568845e-05}
{"train_loss": 0.14402401447296143, "global_step": 65599, "epoch": 737, "lr": 8.641984496845747e-05}
{"train_loss": 0.2582612931728363, "global_step": 65600, "epoch": 737, "lr": 8.641944775633026e-05}
{"train_loss": 0.3126065731048584, "global_step": 65601, "epoch": 737, "lr": 8.641905053930693e-05}
{"train_loss": 0.13953951001167297, "global_step": 65602, "epoch": 737, "lr": 8.641865331738746e-05}
{"train_loss": 0.2984146475791931, "global_step": 65603, "epoch": 737, "lr": 8.641825609057196e-05}
{"train_loss": 0.37543806433677673, "global_step": 65604, "epoch": 737, "lr": 8.641785885886049e-05}
{"train_loss": 0.22785088419914246, "global_step": 65605, "epoch": 737, "lr": 8.641746162225306e-05}
{"train_loss": 0.2221396267414093, "global_step": 65606, "epoch": 737, "lr": 8.641706438074976e-05}
{"train_loss": 0.2942473888397217, "global_step": 65607, "epoch": 737, "lr": 8.641666713435063e-05}
{"train_loss": 0.3178041875362396, "global_step": 65608, "epoch": 737, "lr": 8.641626988305572e-05}
{"train_loss": 0.18817266821861267, "global_step": 65609, "epoch": 737, "lr": 8.641587262686507e-05}
{"train_loss": 0.23182620108127594, "global_step": 65610, "epoch": 737, "lr": 8.641547536577877e-05}
{"train_loss": 0.2534193694591522, "global_step": 65611, "epoch": 737, "lr": 8.641507809979686e-05}
{"train_loss": 0.27426329255104065, "global_step": 65612, "epoch": 737, "lr": 8.641468082891937e-05}
{"train_loss": 0.2577071785926819, "global_step": 65613, "epoch": 737, "lr": 8.641428355314637e-05}
{"train_loss": 0.17607012391090393, "global_step": 65614, "epoch": 737, "lr": 8.641388627247792e-05}
{"train_loss": 0.2821884751319885, "global_step": 65615, "epoch": 737, "lr": 8.641348898691406e-05}
{"train_loss": 0.22312678396701813, "global_step": 65616, "epoch": 737, "lr": 8.641309169645486e-05}
{"train_loss": 0.1900978684425354, "global_step": 65617, "epoch": 737, "lr": 8.641269440110035e-05}
{"train_loss": 0.2157040387392044, "global_step": 65618, "epoch": 737, "lr": 8.641229710085062e-05}
{"train_loss": 0.27119314670562744, "global_step": 65619, "epoch": 737, "lr": 8.641189979570567e-05}
{"train_loss": 0.17984354496002197, "global_step": 65620, "epoch": 737, "lr": 8.641150248566561e-05}
{"train_loss": 0.37124377489089966, "global_step": 65621, "epoch": 737, "lr": 8.641110517073045e-05}
{"train_loss": 0.31261342763900757, "global_step": 65622, "epoch": 737, "lr": 8.641070785090026e-05}
{"train_loss": 0.24795211851596832, "global_step": 65623, "epoch": 737, "lr": 8.641031052617511e-05}
{"train_loss": 0.21249407529830933, "global_step": 65624, "epoch": 737, "lr": 8.640991319655503e-05}
{"train_loss": 0.2918393313884735, "global_step": 65625, "epoch": 737, "lr": 8.640951586204007e-05}
{"train_loss": 0.33565670251846313, "global_step": 65626, "epoch": 737, "lr": 8.640911852263031e-05}
{"train_loss": 0.2606585621833801, "global_step": 65627, "epoch": 737, "lr": 8.640872117832577e-05}
{"train_loss": 0.250407874584198, "global_step": 65628, "epoch": 737, "lr": 8.640832382912654e-05}
{"train_loss": 0.22847917675971985, "global_step": 65629, "epoch": 737, "lr": 8.640792647503265e-05}
{"train_loss": 0.26065143942832947, "global_step": 65630, "epoch": 737, "lr": 8.640752911604415e-05}
{"train_loss": 0.25250715017318726, "global_step": 65631, "epoch": 737, "lr": 8.64071317521611e-05}
{"train_loss": 0.2094716578722, "global_step": 65632, "epoch": 737, "lr": 8.640673438338359e-05}
{"train_loss": 0.2667136490345001, "global_step": 65633, "epoch": 737, "lr": 8.640633700971159e-05}
{"train_loss": 0.18966920673847198, "global_step": 65634, "epoch": 737, "lr": 8.640593963114523e-05}
{"train_loss": 0.2393103539943695, "global_step": 65635, "epoch": 737, "lr": 8.640554224768454e-05}
{"train_loss": 0.30008500814437866, "global_step": 65636, "epoch": 737, "lr": 8.640514485932954e-05}
{"train_loss": 0.264834463596344, "global_step": 65637, "epoch": 737, "lr": 8.640474746608033e-05}
{"train_loss": 0.21092386543750763, "global_step": 65638, "epoch": 737, "lr": 8.640435006793694e-05}
{"train_loss": 0.2151440978050232, "global_step": 65639, "epoch": 737, "lr": 8.640395266489943e-05}
{"train_loss": 0.20693807303905487, "global_step": 65640, "epoch": 737, "lr": 8.640355525696787e-05}
{"train_loss": 0.1675037443637848, "global_step": 65641, "epoch": 737, "lr": 8.640315784414226e-05}
{"train_loss": 0.2032967507839203, "global_step": 65642, "epoch": 737, "lr": 8.640276042642272e-05}
{"train_loss": 0.2237628549337387, "global_step": 65643, "epoch": 737, "lr": 8.640236300380927e-05}
{"train_loss": 0.2649729251861572, "global_step": 65644, "epoch": 737, "lr": 8.640196557630194e-05}
{"train_loss": 0.2353687584400177, "global_step": 65645, "epoch": 737, "lr": 8.640156814390084e-05}
{"train_loss": 0.242187961935997, "global_step": 65646, "epoch": 737, "lr": 8.640117070660598e-05}
{"train_loss": 0.25185421109199524, "global_step": 65647, "epoch": 737, "lr": 8.640077326441742e-05}
{"train_loss": 0.2472188025712967, "global_step": 65648, "epoch": 737, "lr": 8.640037581733522e-05}
{"train_loss": 0.2689146399497986, "global_step": 65649, "epoch": 737, "lr": 8.639997836535944e-05}
{"train_loss": 0.27161744236946106, "global_step": 65650, "epoch": 737, "lr": 8.639958090849013e-05}
{"train_loss": 0.12713323533535004, "global_step": 65651, "epoch": 737, "lr": 8.639918344672734e-05}
{"train_loss": 0.2246570587158203, "global_step": 65652, "epoch": 737, "lr": 8.639878598007112e-05}
{"train_loss": 0.1887127161026001, "global_step": 65653, "epoch": 737, "lr": 8.639838850852152e-05}
{"train_loss": 0.20813339948654175, "global_step": 65654, "epoch": 737, "lr": 8.63979910320786e-05}
{"train_loss": 0.33852678537368774, "global_step": 65655, "epoch": 737, "lr": 8.639759355074243e-05}
{"train_loss": 0.19029724597930908, "global_step": 65656, "epoch": 737, "lr": 8.639719606451305e-05}
{"train_loss": 0.17147061228752136, "global_step": 65657, "epoch": 737, "lr": 8.63967985733905e-05}
{"train_loss": 0.2143361121416092, "global_step": 65658, "epoch": 737, "lr": 8.639640107737483e-05}
{"train_loss": 0.1955600529909134, "global_step": 65659, "epoch": 737, "lr": 8.639600357646611e-05}
{"train_loss": 0.1732686460018158, "global_step": 65660, "epoch": 737, "lr": 8.63956060706644e-05}
{"train_loss": 0.28725266456604004, "global_step": 65661, "epoch": 737, "lr": 8.639520855996974e-05}
{"train_loss": 0.19800415635108948, "global_step": 65662, "epoch": 737, "lr": 8.639481104438221e-05}
{"train_loss": 0.21660403907299042, "global_step": 65663, "epoch": 737, "lr": 8.639441352390181e-05}
{"train_loss": 0.263381689786911, "global_step": 65664, "epoch": 737, "lr": 8.639401599852865e-05}
{"train_loss": 0.20984308421611786, "global_step": 65665, "epoch": 737, "lr": 8.639361846826274e-05}
{"train_loss": 0.2583593428134918, "global_step": 65666, "epoch": 737, "lr": 8.639322093310415e-05}
{"train_loss": 0.2698652148246765, "global_step": 65667, "epoch": 737, "lr": 8.639282339305294e-05}
{"train_loss": 0.20610693097114563, "global_step": 65668, "epoch": 737, "lr": 8.639242584810917e-05}
{"train_loss": 0.16036160290241241, "global_step": 65669, "epoch": 737, "lr": 8.639202829827287e-05}
{"train_loss": 0.19147878885269165, "global_step": 65670, "epoch": 737, "lr": 8.639163074354411e-05}
{"train_loss": 0.16977010667324066, "global_step": 65671, "epoch": 737, "lr": 8.639123318392294e-05}
{"train_loss": 0.2552511692047119, "global_step": 65672, "epoch": 737, "lr": 8.639083561940942e-05}
{"train_loss": 0.27155300974845886, "global_step": 65673, "epoch": 737, "lr": 8.639043805000357e-05}
{"train_loss": 0.1766662299633026, "global_step": 65674, "epoch": 737, "lr": 8.639004047570549e-05}
{"train_loss": 0.26891228556632996, "global_step": 65675, "epoch": 737, "lr": 8.638964289651521e-05}
{"train_loss": 0.2606927454471588, "global_step": 65676, "epoch": 737, "lr": 8.638924531243278e-05}
{"train_loss": 0.20051588118076324, "global_step": 65677, "epoch": 737, "lr": 8.638884772345827e-05}
{"train_loss": 0.2768624424934387, "global_step": 65678, "epoch": 737, "lr": 8.638845012959171e-05}
{"train_loss": 0.20587053894996643, "global_step": 65679, "epoch": 737, "lr": 8.638805253083317e-05}
{"train_loss": 0.2628399133682251, "global_step": 65680, "epoch": 737, "lr": 8.63876549271827e-05}
{"train_loss": 0.2360494660862376, "global_step": 65681, "epoch": 737, "lr": 8.638725731864035e-05, "val_loss": 3.498567581176758}
{"train_loss": 0.34023594856262207, "global_step": 65682, "epoch": 738, "lr": 8.638685970520619e-05}
{"train_loss": 0.17904698848724365, "global_step": 65683, "epoch": 738, "lr": 8.638646208688024e-05}
{"train_loss": 0.20612302422523499, "global_step": 65684, "epoch": 738, "lr": 8.63860644636626e-05}
{"train_loss": 0.17030391097068787, "global_step": 65685, "epoch": 738, "lr": 8.638566683555328e-05}
{"train_loss": 0.29132965207099915, "global_step": 65686, "epoch": 738, "lr": 8.638526920255236e-05}
{"train_loss": 0.20628228783607483, "global_step": 65687, "epoch": 738, "lr": 8.638487156465987e-05}
{"train_loss": 0.1921897679567337, "global_step": 65688, "epoch": 738, "lr": 8.638447392187588e-05}
{"train_loss": 0.21155564486980438, "global_step": 65689, "epoch": 738, "lr": 8.638407627420045e-05}
{"train_loss": 0.15381522476673126, "global_step": 65690, "epoch": 738, "lr": 8.638367862163362e-05}
{"train_loss": 0.26803797483444214, "global_step": 65691, "epoch": 738, "lr": 8.638328096417544e-05}
{"train_loss": 0.19323313236236572, "global_step": 65692, "epoch": 738, "lr": 8.638288330182598e-05}
{"train_loss": 0.2334561049938202, "global_step": 65693, "epoch": 738, "lr": 8.638248563458528e-05}
{"train_loss": 0.18700815737247467, "global_step": 65694, "epoch": 738, "lr": 8.638208796245342e-05}
{"train_loss": 0.16960303485393524, "global_step": 65695, "epoch": 738, "lr": 8.638169028543042e-05}
{"train_loss": 0.2467411458492279, "global_step": 65696, "epoch": 738, "lr": 8.638129260351633e-05}
{"train_loss": 0.21583914756774902, "global_step": 65697, "epoch": 738, "lr": 8.638089491671123e-05}
{"train_loss": 0.21561338007450104, "global_step": 65698, "epoch": 738, "lr": 8.638049722501516e-05}
{"train_loss": 0.23501944541931152, "global_step": 65699, "epoch": 738, "lr": 8.638009952842818e-05}
{"train_loss": 0.21952085196971893, "global_step": 65700, "epoch": 738, "lr": 8.637970182695034e-05}
{"train_loss": 0.19597195088863373, "global_step": 65701, "epoch": 738, "lr": 8.637930412058167e-05}
{"train_loss": 0.2441168874502182, "global_step": 65702, "epoch": 738, "lr": 8.637890640932227e-05}
{"train_loss": 0.24485184252262115, "global_step": 65703, "epoch": 738, "lr": 8.637850869317217e-05}
{"train_loss": 0.11197137832641602, "global_step": 65704, "epoch": 738, "lr": 8.63781109721314e-05}
{"train_loss": 0.32589074969291687, "global_step": 65705, "epoch": 738, "lr": 8.637771324620006e-05}
{"train_loss": 0.1884518563747406, "global_step": 65706, "epoch": 738, "lr": 8.637731551537818e-05}
{"train_loss": 0.18355348706245422, "global_step": 65707, "epoch": 738, "lr": 8.63769177796658e-05}
{"train_loss": 0.20548227429389954, "global_step": 65708, "epoch": 738, "lr": 8.637652003906299e-05}
{"train_loss": 0.16535580158233643, "global_step": 65709, "epoch": 738, "lr": 8.637612229356979e-05}
{"train_loss": 0.19536836445331573, "global_step": 65710, "epoch": 738, "lr": 8.637572454318628e-05}
{"train_loss": 0.1686844378709793, "global_step": 65711, "epoch": 738, "lr": 8.63753267879125e-05}
{"train_loss": 0.21583004295825958, "global_step": 65712, "epoch": 738, "lr": 8.637492902774849e-05}
{"train_loss": 0.16911117732524872, "global_step": 65713, "epoch": 738, "lr": 8.637453126269433e-05}
{"train_loss": 0.2308589220046997, "global_step": 65714, "epoch": 738, "lr": 8.637413349275003e-05}
{"train_loss": 0.23265300691127777, "global_step": 65715, "epoch": 738, "lr": 8.637373571791569e-05}
{"train_loss": 0.300881952047348, "global_step": 65716, "epoch": 738, "lr": 8.637333793819134e-05}
{"train_loss": 0.2313903272151947, "global_step": 65717, "epoch": 738, "lr": 8.637294015357704e-05}
{"train_loss": 0.19313517212867737, "global_step": 65718, "epoch": 738, "lr": 8.637254236407283e-05}
{"train_loss": 0.33907756209373474, "global_step": 65719, "epoch": 738, "lr": 8.63721445696788e-05}
{"train_loss": 0.19338640570640564, "global_step": 65720, "epoch": 738, "lr": 8.637174677039497e-05}
{"train_loss": 0.2674989402294159, "global_step": 65721, "epoch": 738, "lr": 8.637134896622139e-05}
{"train_loss": 0.27433300018310547, "global_step": 65722, "epoch": 738, "lr": 8.637095115715814e-05}
{"train_loss": 0.18635988235473633, "global_step": 65723, "epoch": 738, "lr": 8.637055334320524e-05}
{"train_loss": 0.20028606057167053, "global_step": 65724, "epoch": 738, "lr": 8.637015552436279e-05}
{"train_loss": 0.2930041253566742, "global_step": 65725, "epoch": 738, "lr": 8.636975770063078e-05}
{"train_loss": 0.189987912774086, "global_step": 65726, "epoch": 738, "lr": 8.636935987200932e-05}
{"train_loss": 0.2929690480232239, "global_step": 65727, "epoch": 738, "lr": 8.636896203849845e-05}
{"train_loss": 0.2085888683795929, "global_step": 65728, "epoch": 738, "lr": 8.63685642000982e-05}
{"train_loss": 0.2150702029466629, "global_step": 65729, "epoch": 738, "lr": 8.636816635680864e-05}
{"train_loss": 0.22324208915233612, "global_step": 65730, "epoch": 738, "lr": 8.636776850862983e-05}
{"train_loss": 0.25142842531204224, "global_step": 65731, "epoch": 738, "lr": 8.636737065556182e-05}
{"train_loss": 0.29969221353530884, "global_step": 65732, "epoch": 738, "lr": 8.636697279760466e-05}
{"train_loss": 0.24303282797336578, "global_step": 65733, "epoch": 738, "lr": 8.63665749347584e-05}
{"train_loss": 0.20129959285259247, "global_step": 65734, "epoch": 738, "lr": 8.63661770670231e-05}
{"train_loss": 0.326913058757782, "global_step": 65735, "epoch": 738, "lr": 8.636577919439881e-05}
{"train_loss": 0.18758268654346466, "global_step": 65736, "epoch": 738, "lr": 8.636538131688558e-05}
{"train_loss": 0.19330112636089325, "global_step": 65737, "epoch": 738, "lr": 8.636498343448347e-05}
{"train_loss": 0.18961705267429352, "global_step": 65738, "epoch": 738, "lr": 8.636458554719254e-05}
{"train_loss": 0.16667255759239197, "global_step": 65739, "epoch": 738, "lr": 8.636418765501281e-05}
{"train_loss": 0.2351037710905075, "global_step": 65740, "epoch": 738, "lr": 8.636378975794438e-05}
{"train_loss": 0.19264566898345947, "global_step": 65741, "epoch": 738, "lr": 8.636339185598729e-05}
{"train_loss": 0.2768923342227936, "global_step": 65742, "epoch": 738, "lr": 8.636299394914156e-05}
{"train_loss": 0.10797400772571564, "global_step": 65743, "epoch": 738, "lr": 8.636259603740728e-05}
{"train_loss": 0.2484194040298462, "global_step": 65744, "epoch": 738, "lr": 8.63621981207845e-05}
{"train_loss": 0.23339709639549255, "global_step": 65745, "epoch": 738, "lr": 8.636180019927327e-05}
{"train_loss": 0.305198073387146, "global_step": 65746, "epoch": 738, "lr": 8.636140227287362e-05}
{"train_loss": 0.2191227525472641, "global_step": 65747, "epoch": 738, "lr": 8.636100434158563e-05}
{"train_loss": 0.21890553832054138, "global_step": 65748, "epoch": 738, "lr": 8.636060640540935e-05}
{"train_loss": 0.2851351797580719, "global_step": 65749, "epoch": 738, "lr": 8.636020846434482e-05}
{"train_loss": 0.27083879709243774, "global_step": 65750, "epoch": 738, "lr": 8.63598105183921e-05}
{"train_loss": 0.24452915787696838, "global_step": 65751, "epoch": 738, "lr": 8.635941256755127e-05}
{"train_loss": 0.23505070805549622, "global_step": 65752, "epoch": 738, "lr": 8.635901461182234e-05}
{"train_loss": 0.28381893038749695, "global_step": 65753, "epoch": 738, "lr": 8.635861665120539e-05}
{"train_loss": 0.2407156229019165, "global_step": 65754, "epoch": 738, "lr": 8.635821868570046e-05}
{"train_loss": 0.2575930655002594, "global_step": 65755, "epoch": 738, "lr": 8.635782071530763e-05}
{"train_loss": 0.27315500378608704, "global_step": 65756, "epoch": 738, "lr": 8.635742274002692e-05}
{"train_loss": 0.23594456911087036, "global_step": 65757, "epoch": 738, "lr": 8.63570247598584e-05}
{"train_loss": 0.22197102010250092, "global_step": 65758, "epoch": 738, "lr": 8.635662677480211e-05}
{"train_loss": 0.3184134066104889, "global_step": 65759, "epoch": 738, "lr": 8.635622878485814e-05}
{"train_loss": 0.2806747853755951, "global_step": 65760, "epoch": 738, "lr": 8.635583079002649e-05}
{"train_loss": 0.200835719704628, "global_step": 65761, "epoch": 738, "lr": 8.635543279030725e-05}
{"train_loss": 0.30352678894996643, "global_step": 65762, "epoch": 738, "lr": 8.635503478570048e-05}
{"train_loss": 0.20827341079711914, "global_step": 65763, "epoch": 738, "lr": 8.635463677620621e-05}
{"train_loss": 0.14261767268180847, "global_step": 65764, "epoch": 738, "lr": 8.63542387618245e-05}
{"train_loss": 0.19536764919757843, "global_step": 65765, "epoch": 738, "lr": 8.63538407425554e-05}
{"train_loss": 0.19969233870506287, "global_step": 65766, "epoch": 738, "lr": 8.635344271839897e-05}
{"train_loss": 0.19950394332408905, "global_step": 65767, "epoch": 738, "lr": 8.635304468935528e-05}
{"train_loss": 0.2824244499206543, "global_step": 65768, "epoch": 738, "lr": 8.635264665542435e-05}
{"train_loss": 0.20897947251796722, "global_step": 65769, "epoch": 738, "lr": 8.635224861660625e-05}
{"train_loss": 0.2276300789600008, "global_step": 65770, "epoch": 738, "lr": 8.635185057290105e-05, "val_loss": 3.618920087814331}
{"train_loss": 0.2645467519760132, "global_step": 65771, "epoch": 739, "lr": 8.635145252430876e-05}
{"train_loss": 0.19852349162101746, "global_step": 65772, "epoch": 739, "lr": 8.63510544708295e-05}
{"train_loss": 0.23949900269508362, "global_step": 65773, "epoch": 739, "lr": 8.635065641246326e-05}
{"train_loss": 0.2412296086549759, "global_step": 65774, "epoch": 739, "lr": 8.63502583492101e-05}
{"train_loss": 0.15663285553455353, "global_step": 65775, "epoch": 739, "lr": 8.63498602810701e-05}
{"train_loss": 0.25866007804870605, "global_step": 65776, "epoch": 739, "lr": 8.634946220804332e-05}
{"train_loss": 0.19308751821517944, "global_step": 65777, "epoch": 739, "lr": 8.634906413012979e-05}
{"train_loss": 0.29097694158554077, "global_step": 65778, "epoch": 739, "lr": 8.634866604732956e-05}
{"train_loss": 0.2679929733276367, "global_step": 65779, "epoch": 739, "lr": 8.634826795964271e-05}
{"train_loss": 0.18749241530895233, "global_step": 65780, "epoch": 739, "lr": 8.634786986706928e-05}
{"train_loss": 0.19255580008029938, "global_step": 65781, "epoch": 739, "lr": 8.634747176960932e-05}
{"train_loss": 0.30101239681243896, "global_step": 65782, "epoch": 739, "lr": 8.634707366726287e-05}
{"train_loss": 0.21967270970344543, "global_step": 65783, "epoch": 739, "lr": 8.634667556003001e-05}
{"train_loss": 0.22902528941631317, "global_step": 65784, "epoch": 739, "lr": 8.634627744791079e-05}
{"train_loss": 0.21605609357357025, "global_step": 65785, "epoch": 739, "lr": 8.634587933090524e-05}
{"train_loss": 0.20386843383312225, "global_step": 65786, "epoch": 739, "lr": 8.634548120901345e-05}
{"train_loss": 0.1817711889743805, "global_step": 65787, "epoch": 739, "lr": 8.634508308223543e-05}
{"train_loss": 0.19276608526706696, "global_step": 65788, "epoch": 739, "lr": 8.634468495057126e-05}
{"train_loss": 0.22391247749328613, "global_step": 65789, "epoch": 739, "lr": 8.6344286814021e-05}
{"train_loss": 0.22241947054862976, "global_step": 65790, "epoch": 739, "lr": 8.63438886725847e-05}
{"train_loss": 0.2069990336894989, "global_step": 65791, "epoch": 739, "lr": 8.63434905262624e-05}
{"train_loss": 0.16564308106899261, "global_step": 65792, "epoch": 739, "lr": 8.634309237505416e-05}
{"train_loss": 0.37068411707878113, "global_step": 65793, "epoch": 739, "lr": 8.634269421896002e-05}
{"train_loss": 0.19989824295043945, "global_step": 65794, "epoch": 739, "lr": 8.634229605798006e-05}
{"train_loss": 0.23499684035778046, "global_step": 65795, "epoch": 739, "lr": 8.634189789211433e-05}
{"train_loss": 0.3026718497276306, "global_step": 65796, "epoch": 739, "lr": 8.634149972136286e-05}
{"train_loss": 0.24028562009334564, "global_step": 65797, "epoch": 739, "lr": 8.634110154572572e-05}
{"train_loss": 0.18367666006088257, "global_step": 65798, "epoch": 739, "lr": 8.634070336520297e-05}
{"train_loss": 0.19336459040641785, "global_step": 65799, "epoch": 739, "lr": 8.634030517979466e-05}
{"train_loss": 0.21024037897586823, "global_step": 65800, "epoch": 739, "lr": 8.633990698950083e-05}
{"train_loss": 0.33702096343040466, "global_step": 65801, "epoch": 739, "lr": 8.633950879432155e-05}
{"train_loss": 0.24176017940044403, "global_step": 65802, "epoch": 739, "lr": 8.633911059425686e-05}
{"train_loss": 0.23428742587566376, "global_step": 65803, "epoch": 739, "lr": 8.633871238930682e-05}
{"train_loss": 0.21134130656719208, "global_step": 65804, "epoch": 739, "lr": 8.633831417947147e-05}
{"train_loss": 0.19239380955696106, "global_step": 65805, "epoch": 739, "lr": 8.633791596475089e-05}
{"train_loss": 0.2476978600025177, "global_step": 65806, "epoch": 739, "lr": 8.633751774514511e-05}
{"train_loss": 0.18403424322605133, "global_step": 65807, "epoch": 739, "lr": 8.633711952065422e-05}
{"train_loss": 0.2462819516658783, "global_step": 65808, "epoch": 739, "lr": 8.633672129127822e-05}
{"train_loss": 0.15471790730953217, "global_step": 65809, "epoch": 739, "lr": 8.63363230570172e-05}
{"train_loss": 0.2517487406730652, "global_step": 65810, "epoch": 739, "lr": 8.63359248178712e-05}
{"train_loss": 0.2133939564228058, "global_step": 65811, "epoch": 739, "lr": 8.63355265738403e-05}
{"train_loss": 0.311989963054657, "global_step": 65812, "epoch": 739, "lr": 8.633512832492449e-05}
{"train_loss": 0.2697470784187317, "global_step": 65813, "epoch": 739, "lr": 8.633473007112391e-05}
{"train_loss": 0.21741050481796265, "global_step": 65814, "epoch": 739, "lr": 8.633433181243855e-05}
{"train_loss": 0.2899255156517029, "global_step": 65815, "epoch": 739, "lr": 8.633393354886847e-05}
{"train_loss": 0.24646374583244324, "global_step": 65816, "epoch": 739, "lr": 8.633353528041375e-05}
{"train_loss": 0.27989673614501953, "global_step": 65817, "epoch": 739, "lr": 8.633313700707442e-05}
{"train_loss": 0.1902485489845276, "global_step": 65818, "epoch": 739, "lr": 8.633273872885055e-05}
{"train_loss": 0.22002387046813965, "global_step": 65819, "epoch": 739, "lr": 8.633234044574217e-05}
{"train_loss": 0.2365742027759552, "global_step": 65820, "epoch": 739, "lr": 8.633194215774936e-05}
{"train_loss": 0.17628835141658783, "global_step": 65821, "epoch": 739, "lr": 8.633154386487216e-05}
{"train_loss": 0.258213073015213, "global_step": 65822, "epoch": 739, "lr": 8.633114556711064e-05}
{"train_loss": 0.174428328871727, "global_step": 65823, "epoch": 739, "lr": 8.633074726446483e-05}
{"train_loss": 0.21361202001571655, "global_step": 65824, "epoch": 739, "lr": 8.63303489569348e-05}
{"train_loss": 0.24281415343284607, "global_step": 65825, "epoch": 739, "lr": 8.632995064452059e-05}
{"train_loss": 0.25390762090682983, "global_step": 65826, "epoch": 739, "lr": 8.632955232722227e-05}
{"train_loss": 0.26197198033332825, "global_step": 65827, "epoch": 739, "lr": 8.632915400503987e-05}
{"train_loss": 0.2194419652223587, "global_step": 65828, "epoch": 739, "lr": 8.632875567797347e-05}
{"train_loss": 0.1659661829471588, "global_step": 65829, "epoch": 739, "lr": 8.632835734602311e-05}
{"train_loss": 0.21722178161144257, "global_step": 65830, "epoch": 739, "lr": 8.632795900918884e-05}
{"train_loss": 0.2174285352230072, "global_step": 65831, "epoch": 739, "lr": 8.632756066747072e-05}
{"train_loss": 0.1688779890537262, "global_step": 65832, "epoch": 739, "lr": 8.632716232086881e-05}
{"train_loss": 0.22975368797779083, "global_step": 65833, "epoch": 739, "lr": 8.632676396938316e-05}
{"train_loss": 0.24601112306118011, "global_step": 65834, "epoch": 739, "lr": 8.63263656130138e-05}
{"train_loss": 0.1793370246887207, "global_step": 65835, "epoch": 739, "lr": 8.632596725176081e-05}
{"train_loss": 0.3160775899887085, "global_step": 65836, "epoch": 739, "lr": 8.632556888562425e-05}
{"train_loss": 0.2626291811466217, "global_step": 65837, "epoch": 739, "lr": 8.632517051460414e-05}
{"train_loss": 0.24292130768299103, "global_step": 65838, "epoch": 739, "lr": 8.632477213870056e-05}
{"train_loss": 0.2097032517194748, "global_step": 65839, "epoch": 739, "lr": 8.632437375791356e-05}
{"train_loss": 0.28575262427330017, "global_step": 65840, "epoch": 739, "lr": 8.63239753722432e-05}
{"train_loss": 0.17393918335437775, "global_step": 65841, "epoch": 739, "lr": 8.63235769816895e-05}
{"train_loss": 0.20092101395130157, "global_step": 65842, "epoch": 739, "lr": 8.632317858625256e-05}
{"train_loss": 0.12662330269813538, "global_step": 65843, "epoch": 739, "lr": 8.632278018593241e-05}
{"train_loss": 0.14951550960540771, "global_step": 65844, "epoch": 739, "lr": 8.632238178072909e-05}
{"train_loss": 0.22518309950828552, "global_step": 65845, "epoch": 739, "lr": 8.632198337064268e-05}
{"train_loss": 0.20773319900035858, "global_step": 65846, "epoch": 739, "lr": 8.632158495567322e-05}
{"train_loss": 0.27283912897109985, "global_step": 65847, "epoch": 739, "lr": 8.632118653582075e-05}
{"train_loss": 0.17696307599544525, "global_step": 65848, "epoch": 739, "lr": 8.632078811108536e-05}
{"train_loss": 0.18245352804660797, "global_step": 65849, "epoch": 739, "lr": 8.632038968146707e-05}
{"train_loss": 0.23225180804729462, "global_step": 65850, "epoch": 739, "lr": 8.631999124696596e-05}
{"train_loss": 0.26237377524375916, "global_step": 65851, "epoch": 739, "lr": 8.631959280758206e-05}
{"train_loss": 0.20663438737392426, "global_step": 65852, "epoch": 739, "lr": 8.631919436331544e-05}
{"train_loss": 0.27776336669921875, "global_step": 65853, "epoch": 739, "lr": 8.631879591416614e-05}
{"train_loss": 0.22726017236709595, "global_step": 65854, "epoch": 739, "lr": 8.631839746013423e-05}
{"train_loss": 0.2691386044025421, "global_step": 65855, "epoch": 739, "lr": 8.631799900121974e-05}
{"train_loss": 0.2477172464132309, "global_step": 65856, "epoch": 739, "lr": 8.631760053742275e-05}
{"train_loss": 0.19329555332660675, "global_step": 65857, "epoch": 739, "lr": 8.63172020687433e-05}
{"train_loss": 0.17036351561546326, "global_step": 65858, "epoch": 739, "lr": 8.631680359518144e-05}
{"train_loss": 0.2274559023005239, "global_step": 65859, "epoch": 739, "lr": 8.631640511673723e-05, "val_loss": 3.5487563610076904}
{"train_loss": 0.3151828944683075, "global_step": 65860, "epoch": 740, "lr": 8.631600663341072e-05}
{"train_loss": 0.14999692142009735, "global_step": 65861, "epoch": 740, "lr": 8.631560814520196e-05}
{"train_loss": 0.2842859923839569, "global_step": 65862, "epoch": 740, "lr": 8.631520965211102e-05}
{"train_loss": 0.21263974905014038, "global_step": 65863, "epoch": 740, "lr": 8.631481115413795e-05}
{"train_loss": 0.21767066419124603, "global_step": 65864, "epoch": 740, "lr": 8.631441265128277e-05}
{"train_loss": 0.2576393187046051, "global_step": 65865, "epoch": 740, "lr": 8.631401414354557e-05}
{"train_loss": 0.12909221649169922, "global_step": 65866, "epoch": 740, "lr": 8.631361563092642e-05}
{"train_loss": 0.28508514165878296, "global_step": 65867, "epoch": 740, "lr": 8.631321711342531e-05}
{"train_loss": 0.15164270997047424, "global_step": 65868, "epoch": 740, "lr": 8.631281859104235e-05}
{"train_loss": 0.2562030851840973, "global_step": 65869, "epoch": 740, "lr": 8.631242006377758e-05}
{"train_loss": 0.2244650423526764, "global_step": 65870, "epoch": 740, "lr": 8.631202153163102e-05}
{"train_loss": 0.18515317142009735, "global_step": 65871, "epoch": 740, "lr": 8.631162299460279e-05}
{"train_loss": 0.22535371780395508, "global_step": 65872, "epoch": 740, "lr": 8.631122445269288e-05}
{"train_loss": 0.14693023264408112, "global_step": 65873, "epoch": 740, "lr": 8.631082590590136e-05}
{"train_loss": 0.23212505877017975, "global_step": 65874, "epoch": 740, "lr": 8.63104273542283e-05}
{"train_loss": 0.20768111944198608, "global_step": 65875, "epoch": 740, "lr": 8.631002879767376e-05}
{"train_loss": 0.290934294462204, "global_step": 65876, "epoch": 740, "lr": 8.630963023623777e-05}
{"train_loss": 0.18507742881774902, "global_step": 65877, "epoch": 740, "lr": 8.630923166992039e-05}
{"train_loss": 0.3001842796802521, "global_step": 65878, "epoch": 740, "lr": 8.630883309872168e-05}
{"train_loss": 0.21650737524032593, "global_step": 65879, "epoch": 740, "lr": 8.630843452264169e-05}
{"train_loss": 0.2337644100189209, "global_step": 65880, "epoch": 740, "lr": 8.630803594168046e-05}
{"train_loss": 0.27001824975013733, "global_step": 65881, "epoch": 740, "lr": 8.630763735583808e-05}
{"train_loss": 0.22380122542381287, "global_step": 65882, "epoch": 740, "lr": 8.630723876511457e-05}
{"train_loss": 0.19647717475891113, "global_step": 65883, "epoch": 740, "lr": 8.630684016950998e-05}
{"train_loss": 0.23576678335666656, "global_step": 65884, "epoch": 740, "lr": 8.63064415690244e-05}
{"train_loss": 0.25508031249046326, "global_step": 65885, "epoch": 740, "lr": 8.630604296365786e-05}
{"train_loss": 0.23427878320217133, "global_step": 65886, "epoch": 740, "lr": 8.63056443534104e-05}
{"train_loss": 0.1959439516067505, "global_step": 65887, "epoch": 740, "lr": 8.630524573828211e-05}
{"train_loss": 0.23175190389156342, "global_step": 65888, "epoch": 740, "lr": 8.630484711827301e-05}
{"train_loss": 0.24148482084274292, "global_step": 65889, "epoch": 740, "lr": 8.630444849338314e-05}
{"train_loss": 0.20797792077064514, "global_step": 65890, "epoch": 740, "lr": 8.630404986361262e-05}
{"train_loss": 0.22744454443454742, "global_step": 65891, "epoch": 740, "lr": 8.630365122896144e-05}
{"train_loss": 0.2068764865398407, "global_step": 65892, "epoch": 740, "lr": 8.630325258942968e-05}
{"train_loss": 0.1649639904499054, "global_step": 65893, "epoch": 740, "lr": 8.63028539450174e-05}
{"train_loss": 0.1799076348543167, "global_step": 65894, "epoch": 740, "lr": 8.630245529572464e-05}
{"train_loss": 0.25171715021133423, "global_step": 65895, "epoch": 740, "lr": 8.630205664155146e-05}
{"train_loss": 0.24483080208301544, "global_step": 65896, "epoch": 740, "lr": 8.630165798249791e-05}
{"train_loss": 0.1944274604320526, "global_step": 65897, "epoch": 740, "lr": 8.630125931856403e-05}
{"train_loss": 0.15043628215789795, "global_step": 65898, "epoch": 740, "lr": 8.630086064974991e-05}
{"train_loss": 0.13185441493988037, "global_step": 65899, "epoch": 740, "lr": 8.630046197605557e-05}
{"train_loss": 0.17928734421730042, "global_step": 65900, "epoch": 740, "lr": 8.630006329748108e-05}
{"train_loss": 0.2156139761209488, "global_step": 65901, "epoch": 740, "lr": 8.629966461402647e-05}
{"train_loss": 0.20399020612239838, "global_step": 65902, "epoch": 740, "lr": 8.629926592569184e-05}
{"train_loss": 0.18460392951965332, "global_step": 65903, "epoch": 740, "lr": 8.62988672324772e-05}
{"train_loss": 0.16147880256175995, "global_step": 65904, "epoch": 740, "lr": 8.629846853438262e-05}
{"train_loss": 0.2997390627861023, "global_step": 65905, "epoch": 740, "lr": 8.629806983140816e-05}
{"train_loss": 0.22243021428585052, "global_step": 65906, "epoch": 740, "lr": 8.629767112355386e-05}
{"train_loss": 0.23657117784023285, "global_step": 65907, "epoch": 740, "lr": 8.629727241081979e-05}
{"train_loss": 0.26442766189575195, "global_step": 65908, "epoch": 740, "lr": 8.629687369320599e-05}
{"train_loss": 0.20947274565696716, "global_step": 65909, "epoch": 740, "lr": 8.629647497071252e-05}
{"train_loss": 0.22017091512680054, "global_step": 65910, "epoch": 740, "lr": 8.629607624333942e-05}
{"train_loss": 0.20282714068889618, "global_step": 65911, "epoch": 740, "lr": 8.629567751108676e-05}
{"train_loss": 0.2041546106338501, "global_step": 65912, "epoch": 740, "lr": 8.629527877395458e-05}
{"train_loss": 0.15550190210342407, "global_step": 65913, "epoch": 740, "lr": 8.629488003194295e-05}
{"train_loss": 0.3895149230957031, "global_step": 65914, "epoch": 740, "lr": 8.629448128505193e-05}
{"train_loss": 0.21727943420410156, "global_step": 65915, "epoch": 740, "lr": 8.629408253328154e-05}
{"train_loss": 0.24112457036972046, "global_step": 65916, "epoch": 740, "lr": 8.629368377663187e-05}
{"train_loss": 0.21593411266803741, "global_step": 65917, "epoch": 740, "lr": 8.629328501510294e-05}
{"train_loss": 0.18798910081386566, "global_step": 65918, "epoch": 740, "lr": 8.629288624869483e-05}
{"train_loss": 0.20722998678684235, "global_step": 65919, "epoch": 740, "lr": 8.629248747740758e-05}
{"train_loss": 0.18296076357364655, "global_step": 65920, "epoch": 740, "lr": 8.629208870124125e-05}
{"train_loss": 0.23334528505802155, "global_step": 65921, "epoch": 740, "lr": 8.62916899201959e-05}
{"train_loss": 0.18574751913547516, "global_step": 65922, "epoch": 740, "lr": 8.629129113427155e-05}
{"train_loss": 0.21363034844398499, "global_step": 65923, "epoch": 740, "lr": 8.62908923434683e-05}
{"train_loss": 0.18297630548477173, "global_step": 65924, "epoch": 740, "lr": 8.629049354778618e-05}
{"train_loss": 0.29043135046958923, "global_step": 65925, "epoch": 740, "lr": 8.629009474722523e-05}
{"train_loss": 0.17622439563274384, "global_step": 65926, "epoch": 740, "lr": 8.628969594178554e-05}
{"train_loss": 0.17488734424114227, "global_step": 65927, "epoch": 740, "lr": 8.628929713146712e-05}
{"train_loss": 0.20070427656173706, "global_step": 65928, "epoch": 740, "lr": 8.628889831627006e-05}
{"train_loss": 0.25728195905685425, "global_step": 65929, "epoch": 740, "lr": 8.62884994961944e-05}
{"train_loss": 0.24681639671325684, "global_step": 65930, "epoch": 740, "lr": 8.628810067124018e-05}
{"train_loss": 0.22774234414100647, "global_step": 65931, "epoch": 740, "lr": 8.628770184140748e-05}
{"train_loss": 0.2826404571533203, "global_step": 65932, "epoch": 740, "lr": 8.628730300669634e-05}
{"train_loss": 0.1975252628326416, "global_step": 65933, "epoch": 740, "lr": 8.628690416710682e-05}
{"train_loss": 0.24979330599308014, "global_step": 65934, "epoch": 740, "lr": 8.628650532263896e-05}
{"train_loss": 0.2564972937107086, "global_step": 65935, "epoch": 740, "lr": 8.628610647329282e-05}
{"train_loss": 0.2633165121078491, "global_step": 65936, "epoch": 740, "lr": 8.628570761906846e-05}
{"train_loss": 0.2740011513233185, "global_step": 65937, "epoch": 740, "lr": 8.628530875996594e-05}
{"train_loss": 0.21780405938625336, "global_step": 65938, "epoch": 740, "lr": 8.628490989598527e-05}
{"train_loss": 0.19806230068206787, "global_step": 65939, "epoch": 740, "lr": 8.628451102712656e-05}
{"train_loss": 0.18123580515384674, "global_step": 65940, "epoch": 740, "lr": 8.628411215338984e-05}
{"train_loss": 0.19673414528369904, "global_step": 65941, "epoch": 740, "lr": 8.628371327477515e-05}
{"train_loss": 0.2648133635520935, "global_step": 65942, "epoch": 740, "lr": 8.628331439128258e-05}
{"train_loss": 0.2355286031961441, "global_step": 65943, "epoch": 740, "lr": 8.628291550291214e-05}
{"train_loss": 0.2741157114505768, "global_step": 65944, "epoch": 740, "lr": 8.628251660966393e-05}
{"train_loss": 0.29934343695640564, "global_step": 65945, "epoch": 740, "lr": 8.628211771153794e-05}
{"train_loss": 0.21659938991069794, "global_step": 65946, "epoch": 740, "lr": 8.628171880853428e-05}
{"train_loss": 0.27967968583106995, "global_step": 65947, "epoch": 740, "lr": 8.628131990065299e-05}
{"train_loss": 0.22367664953965818, "global_step": 65948, "epoch": 740, "lr": 8.628092098789411e-05, "val_loss": 3.5901598930358887, "train_action_mse_error": 20.340660095214844}
{"train_loss": 0.2014937847852707, "global_step": 65949, "epoch": 741, "lr": 8.628052207025772e-05}
{"train_loss": 0.20906716585159302, "global_step": 65950, "epoch": 741, "lr": 8.628012314774383e-05}
{"train_loss": 0.24674104154109955, "global_step": 65951, "epoch": 741, "lr": 8.627972422035253e-05}
{"train_loss": 0.32025161385536194, "global_step": 65952, "epoch": 741, "lr": 8.627932528808387e-05}
{"train_loss": 0.20528945326805115, "global_step": 65953, "epoch": 741, "lr": 8.627892635093791e-05}
{"train_loss": 0.19547341763973236, "global_step": 65954, "epoch": 741, "lr": 8.627852740891467e-05}
{"train_loss": 0.2772720158100128, "global_step": 65955, "epoch": 741, "lr": 8.627812846201422e-05}
{"train_loss": 0.23404070734977722, "global_step": 65956, "epoch": 741, "lr": 8.627772951023663e-05}
{"train_loss": 0.20210108160972595, "global_step": 65957, "epoch": 741, "lr": 8.627733055358193e-05}
{"train_loss": 0.2352064549922943, "global_step": 65958, "epoch": 741, "lr": 8.627693159205019e-05}
{"train_loss": 0.19573883712291718, "global_step": 65959, "epoch": 741, "lr": 8.627653262564147e-05}
{"train_loss": 0.15150253474712372, "global_step": 65960, "epoch": 741, "lr": 8.627613365435578e-05}
{"train_loss": 0.2309916615486145, "global_step": 65961, "epoch": 741, "lr": 8.627573467819324e-05}
{"train_loss": 0.19309811294078827, "global_step": 65962, "epoch": 741, "lr": 8.627533569715385e-05}
{"train_loss": 0.22585956752300262, "global_step": 65963, "epoch": 741, "lr": 8.627493671123769e-05}
{"train_loss": 0.19745320081710815, "global_step": 65964, "epoch": 741, "lr": 8.627453772044483e-05}
{"train_loss": 0.25902771949768066, "global_step": 65965, "epoch": 741, "lr": 8.627413872477527e-05}
{"train_loss": 0.22748899459838867, "global_step": 65966, "epoch": 741, "lr": 8.62737397242291e-05}
{"train_loss": 0.1838897317647934, "global_step": 65967, "epoch": 741, "lr": 8.627334071880638e-05}
{"train_loss": 0.3097565770149231, "global_step": 65968, "epoch": 741, "lr": 8.627294170850713e-05}
{"train_loss": 0.20621882379055023, "global_step": 65969, "epoch": 741, "lr": 8.627254269333144e-05}
{"train_loss": 0.15635405480861664, "global_step": 65970, "epoch": 741, "lr": 8.627214367327933e-05}
{"train_loss": 0.1819811463356018, "global_step": 65971, "epoch": 741, "lr": 8.62717446483509e-05}
{"train_loss": 0.20719832181930542, "global_step": 65972, "epoch": 741, "lr": 8.627134561854617e-05}
{"train_loss": 0.18344388902187347, "global_step": 65973, "epoch": 741, "lr": 8.627094658386519e-05}
{"train_loss": 0.2234153002500534, "global_step": 65974, "epoch": 741, "lr": 8.627054754430803e-05}
{"train_loss": 0.1853892058134079, "global_step": 65975, "epoch": 741, "lr": 8.627014849987474e-05}
{"train_loss": 0.22367729246616364, "global_step": 65976, "epoch": 741, "lr": 8.626974945056537e-05}
{"train_loss": 0.18621037900447845, "global_step": 65977, "epoch": 741, "lr": 8.626935039637996e-05}
{"train_loss": 0.22218523919582367, "global_step": 65978, "epoch": 741, "lr": 8.626895133731858e-05}
{"train_loss": 0.2714255154132843, "global_step": 65979, "epoch": 741, "lr": 8.626855227338129e-05}
{"train_loss": 0.279788076877594, "global_step": 65980, "epoch": 741, "lr": 8.626815320456814e-05}
{"train_loss": 0.2172282189130783, "global_step": 65981, "epoch": 741, "lr": 8.626775413087917e-05}
{"train_loss": 0.22539013624191284, "global_step": 65982, "epoch": 741, "lr": 8.626735505231445e-05}
{"train_loss": 0.24597200751304626, "global_step": 65983, "epoch": 741, "lr": 8.626695596887402e-05}
{"train_loss": 0.1392519325017929, "global_step": 65984, "epoch": 741, "lr": 8.626655688055793e-05}
{"train_loss": 0.24145226180553436, "global_step": 65985, "epoch": 741, "lr": 8.626615778736627e-05}
{"train_loss": 0.28070178627967834, "global_step": 65986, "epoch": 741, "lr": 8.626575868929905e-05}
{"train_loss": 0.17015311121940613, "global_step": 65987, "epoch": 741, "lr": 8.626535958635633e-05}
{"train_loss": 0.2987655997276306, "global_step": 65988, "epoch": 741, "lr": 8.626496047853819e-05}
{"train_loss": 0.23204576969146729, "global_step": 65989, "epoch": 741, "lr": 8.626456136584466e-05}
{"train_loss": 0.23835480213165283, "global_step": 65990, "epoch": 741, "lr": 8.626416224827581e-05}
{"train_loss": 0.16825829446315765, "global_step": 65991, "epoch": 741, "lr": 8.626376312583169e-05}
{"train_loss": 0.18711629509925842, "global_step": 65992, "epoch": 741, "lr": 8.626336399851233e-05}
{"train_loss": 0.21871936321258545, "global_step": 65993, "epoch": 741, "lr": 8.626296486631782e-05}
{"train_loss": 0.2701418399810791, "global_step": 65994, "epoch": 741, "lr": 8.626256572924819e-05}
{"train_loss": 0.19169123470783234, "global_step": 65995, "epoch": 741, "lr": 8.62621665873035e-05}
{"train_loss": 0.21048250794410706, "global_step": 65996, "epoch": 741, "lr": 8.62617674404838e-05}
{"train_loss": 0.22364681959152222, "global_step": 65997, "epoch": 741, "lr": 8.626136828878916e-05}
{"train_loss": 0.26301395893096924, "global_step": 65998, "epoch": 741, "lr": 8.626096913221962e-05}
{"train_loss": 0.19486168026924133, "global_step": 65999, "epoch": 741, "lr": 8.626056997077521e-05}
{"train_loss": 0.21880437433719635, "global_step": 66000, "epoch": 741, "lr": 8.626017080445603e-05}
{"train_loss": 0.18735934793949127, "global_step": 66001, "epoch": 741, "lr": 8.62597716332621e-05}
{"train_loss": 0.15300320088863373, "global_step": 66002, "epoch": 741, "lr": 8.62593724571935e-05}
{"train_loss": 0.22942644357681274, "global_step": 66003, "epoch": 741, "lr": 8.625897327625026e-05}
{"train_loss": 0.23911598324775696, "global_step": 66004, "epoch": 741, "lr": 8.625857409043246e-05}
{"train_loss": 0.15788860619068146, "global_step": 66005, "epoch": 741, "lr": 8.62581748997401e-05}
{"train_loss": 0.24599261581897736, "global_step": 66006, "epoch": 741, "lr": 8.625777570417331e-05}
{"train_loss": 0.1434061974287033, "global_step": 66007, "epoch": 741, "lr": 8.625737650373208e-05}
{"train_loss": 0.20163017511367798, "global_step": 66008, "epoch": 741, "lr": 8.625697729841649e-05}
{"train_loss": 0.1905040442943573, "global_step": 66009, "epoch": 741, "lr": 8.62565780882266e-05}
{"train_loss": 0.16935911774635315, "global_step": 66010, "epoch": 741, "lr": 8.625617887316244e-05}
{"train_loss": 0.2087075263261795, "global_step": 66011, "epoch": 741, "lr": 8.62557796532241e-05}
{"train_loss": 0.24643459916114807, "global_step": 66012, "epoch": 741, "lr": 8.62553804284116e-05}
{"train_loss": 0.1741621196269989, "global_step": 66013, "epoch": 741, "lr": 8.6254981198725e-05}
{"train_loss": 0.2445884495973587, "global_step": 66014, "epoch": 741, "lr": 8.625458196416438e-05}
{"train_loss": 0.19945402443408966, "global_step": 66015, "epoch": 741, "lr": 8.625418272472975e-05}
{"train_loss": 0.18948021531105042, "global_step": 66016, "epoch": 741, "lr": 8.62537834804212e-05}
{"train_loss": 0.164859339594841, "global_step": 66017, "epoch": 741, "lr": 8.625338423123876e-05}
{"train_loss": 0.22191771864891052, "global_step": 66018, "epoch": 741, "lr": 8.62529849771825e-05}
{"train_loss": 0.19250167906284332, "global_step": 66019, "epoch": 741, "lr": 8.625258571825248e-05}
{"train_loss": 0.13492079079151154, "global_step": 66020, "epoch": 741, "lr": 8.625218645444872e-05}
{"train_loss": 0.22057411074638367, "global_step": 66021, "epoch": 741, "lr": 8.625178718577132e-05}
{"train_loss": 0.21939797699451447, "global_step": 66022, "epoch": 741, "lr": 8.625138791222029e-05}
{"train_loss": 0.18273170292377472, "global_step": 66023, "epoch": 741, "lr": 8.62509886337957e-05}
{"train_loss": 0.27697303891181946, "global_step": 66024, "epoch": 741, "lr": 8.625058935049763e-05}
{"train_loss": 0.22104647755622864, "global_step": 66025, "epoch": 741, "lr": 8.625019006232609e-05}
{"train_loss": 0.2902939021587372, "global_step": 66026, "epoch": 741, "lr": 8.624979076928116e-05}
{"train_loss": 0.2014637291431427, "global_step": 66027, "epoch": 741, "lr": 8.624939147136289e-05}
{"train_loss": 0.190631702542305, "global_step": 66028, "epoch": 741, "lr": 8.624899216857132e-05}
{"train_loss": 0.24775612354278564, "global_step": 66029, "epoch": 741, "lr": 8.624859286090652e-05}
{"train_loss": 0.25281524658203125, "global_step": 66030, "epoch": 741, "lr": 8.624819354836855e-05}
{"train_loss": 0.20455197989940643, "global_step": 66031, "epoch": 741, "lr": 8.624779423095744e-05}
{"train_loss": 0.18714287877082825, "global_step": 66032, "epoch": 741, "lr": 8.624739490867326e-05}
{"train_loss": 0.40718770027160645, "global_step": 66033, "epoch": 741, "lr": 8.624699558151606e-05}
{"train_loss": 0.2519223392009735, "global_step": 66034, "epoch": 741, "lr": 8.624659624948589e-05}
{"train_loss": 0.16782909631729126, "global_step": 66035, "epoch": 741, "lr": 8.624619691258281e-05}
{"train_loss": 0.1800667643547058, "global_step": 66036, "epoch": 741, "lr": 8.624579757080687e-05}
{"train_loss": 0.21671021319507214, "global_step": 66037, "epoch": 741, "lr": 8.624539822415812e-05, "val_loss": 3.7778775691986084}
{"train_loss": 0.24968896806240082, "global_step": 66038, "epoch": 742, "lr": 8.624499887263662e-05}
{"train_loss": 0.25224781036376953, "global_step": 66039, "epoch": 742, "lr": 8.624459951624243e-05}
{"train_loss": 0.2198927104473114, "global_step": 66040, "epoch": 742, "lr": 8.624420015497558e-05}
{"train_loss": 0.23257900774478912, "global_step": 66041, "epoch": 742, "lr": 8.624380078883616e-05}
{"train_loss": 0.2777442932128906, "global_step": 66042, "epoch": 742, "lr": 8.624340141782419e-05}
{"train_loss": 0.2043953239917755, "global_step": 66043, "epoch": 742, "lr": 8.624300204193973e-05}
{"train_loss": 0.3426746726036072, "global_step": 66044, "epoch": 742, "lr": 8.624260266118284e-05}
{"train_loss": 0.17358751595020294, "global_step": 66045, "epoch": 742, "lr": 8.62422032755536e-05}
{"train_loss": 0.17749133706092834, "global_step": 66046, "epoch": 742, "lr": 8.624180388505201e-05}
{"train_loss": 0.1851801723241806, "global_step": 66047, "epoch": 742, "lr": 8.624140448967815e-05}
{"train_loss": 0.2534501850605011, "global_step": 66048, "epoch": 742, "lr": 8.62410050894321e-05}
{"train_loss": 0.22531689703464508, "global_step": 66049, "epoch": 742, "lr": 8.624060568431388e-05}
{"train_loss": 0.18896688520908356, "global_step": 66050, "epoch": 742, "lr": 8.624020627432354e-05}
{"train_loss": 0.15277758240699768, "global_step": 66051, "epoch": 742, "lr": 8.623980685946113e-05}
{"train_loss": 0.2613133192062378, "global_step": 66052, "epoch": 742, "lr": 8.623940743972676e-05}
{"train_loss": 0.21680143475532532, "global_step": 66053, "epoch": 742, "lr": 8.623900801512041e-05}
{"train_loss": 0.17505480349063873, "global_step": 66054, "epoch": 742, "lr": 8.623860858564219e-05}
{"train_loss": 0.17973196506500244, "global_step": 66055, "epoch": 742, "lr": 8.623820915129212e-05}
{"train_loss": 0.2211582362651825, "global_step": 66056, "epoch": 742, "lr": 8.623780971207026e-05}
{"train_loss": 0.21582819521427155, "global_step": 66057, "epoch": 742, "lr": 8.623741026797668e-05}
{"train_loss": 0.2556844651699066, "global_step": 66058, "epoch": 742, "lr": 8.623701081901143e-05}
{"train_loss": 0.19477514922618866, "global_step": 66059, "epoch": 742, "lr": 8.623661136517454e-05}
{"train_loss": 0.19358588755130768, "global_step": 66060, "epoch": 742, "lr": 8.623621190646607e-05}
{"train_loss": 0.36423954367637634, "global_step": 66061, "epoch": 742, "lr": 8.62358124428861e-05}
{"train_loss": 0.2559888958930969, "global_step": 66062, "epoch": 742, "lr": 8.623541297443466e-05}
{"train_loss": 0.20191222429275513, "global_step": 66063, "epoch": 742, "lr": 8.62350135011118e-05}
{"train_loss": 0.17624516785144806, "global_step": 66064, "epoch": 742, "lr": 8.623461402291761e-05}
{"train_loss": 0.2872365117073059, "global_step": 66065, "epoch": 742, "lr": 8.62342145398521e-05}
{"train_loss": 0.1974838823080063, "global_step": 66066, "epoch": 742, "lr": 8.623381505191535e-05}
{"train_loss": 0.2674359977245331, "global_step": 66067, "epoch": 742, "lr": 8.62334155591074e-05}
{"train_loss": 0.3051820397377014, "global_step": 66068, "epoch": 742, "lr": 8.623301606142831e-05}
{"train_loss": 0.3095201551914215, "global_step": 66069, "epoch": 742, "lr": 8.623261655887813e-05}
{"train_loss": 0.24870243668556213, "global_step": 66070, "epoch": 742, "lr": 8.623221705145691e-05}
{"train_loss": 0.3505900800228119, "global_step": 66071, "epoch": 742, "lr": 8.623181753916471e-05}
{"train_loss": 0.25484320521354675, "global_step": 66072, "epoch": 742, "lr": 8.623141802200161e-05}
{"train_loss": 0.13629043102264404, "global_step": 66073, "epoch": 742, "lr": 8.623101849996761e-05}
{"train_loss": 0.1717296540737152, "global_step": 66074, "epoch": 742, "lr": 8.623061897306279e-05}
{"train_loss": 0.1193358302116394, "global_step": 66075, "epoch": 742, "lr": 8.623021944128722e-05}
{"train_loss": 0.25304892659187317, "global_step": 66076, "epoch": 742, "lr": 8.622981990464094e-05}
{"train_loss": 0.27105414867401123, "global_step": 66077, "epoch": 742, "lr": 8.6229420363124e-05}
{"train_loss": 0.2274823784828186, "global_step": 66078, "epoch": 742, "lr": 8.622902081673645e-05}
{"train_loss": 0.2100774347782135, "global_step": 66079, "epoch": 742, "lr": 8.622862126547835e-05}
{"train_loss": 0.17246198654174805, "global_step": 66080, "epoch": 742, "lr": 8.622822170934976e-05}
{"train_loss": 0.23205673694610596, "global_step": 66081, "epoch": 742, "lr": 8.622782214835073e-05}
{"train_loss": 0.18634584546089172, "global_step": 66082, "epoch": 742, "lr": 8.622742258248128e-05}
{"train_loss": 0.23865894973278046, "global_step": 66083, "epoch": 742, "lr": 8.622702301174152e-05}
{"train_loss": 0.2062297910451889, "global_step": 66084, "epoch": 742, "lr": 8.622662343613148e-05}
{"train_loss": 0.2545478940010071, "global_step": 66085, "epoch": 742, "lr": 8.62262238556512e-05}
{"train_loss": 0.3194253742694855, "global_step": 66086, "epoch": 742, "lr": 8.622582427030077e-05}
{"train_loss": 0.16798971593379974, "global_step": 66087, "epoch": 742, "lr": 8.622542468008019e-05}
{"train_loss": 0.20630960166454315, "global_step": 66088, "epoch": 742, "lr": 8.622502508498957e-05}
{"train_loss": 0.21941342949867249, "global_step": 66089, "epoch": 742, "lr": 8.622462548502892e-05}
{"train_loss": 0.19688040018081665, "global_step": 66090, "epoch": 742, "lr": 8.622422588019831e-05}
{"train_loss": 0.17554643750190735, "global_step": 66091, "epoch": 742, "lr": 8.62238262704978e-05}
{"train_loss": 0.22393310070037842, "global_step": 66092, "epoch": 742, "lr": 8.622342665592743e-05}
{"train_loss": 0.24231603741645813, "global_step": 66093, "epoch": 742, "lr": 8.622302703648727e-05}
{"train_loss": 0.31673508882522583, "global_step": 66094, "epoch": 742, "lr": 8.622262741217738e-05}
{"train_loss": 0.27450841665267944, "global_step": 66095, "epoch": 742, "lr": 8.622222778299776e-05}
{"train_loss": 0.22152839601039886, "global_step": 66096, "epoch": 742, "lr": 8.622182814894853e-05}
{"train_loss": 0.2526199519634247, "global_step": 66097, "epoch": 742, "lr": 8.622142851002972e-05}
{"train_loss": 0.22196902334690094, "global_step": 66098, "epoch": 742, "lr": 8.622102886624137e-05}
{"train_loss": 0.2729889452457428, "global_step": 66099, "epoch": 742, "lr": 8.622062921758355e-05}
{"train_loss": 0.27329012751579285, "global_step": 66100, "epoch": 742, "lr": 8.62202295640563e-05}
{"train_loss": 0.21739448606967926, "global_step": 66101, "epoch": 742, "lr": 8.621982990565969e-05}
{"train_loss": 0.2603912651538849, "global_step": 66102, "epoch": 742, "lr": 8.621943024239376e-05}
{"train_loss": 0.2778519093990326, "global_step": 66103, "epoch": 742, "lr": 8.621903057425857e-05}
{"train_loss": 0.28942325711250305, "global_step": 66104, "epoch": 742, "lr": 8.621863090125417e-05}
{"train_loss": 0.2973940074443817, "global_step": 66105, "epoch": 742, "lr": 8.621823122338062e-05}
{"train_loss": 0.15422242879867554, "global_step": 66106, "epoch": 742, "lr": 8.621783154063799e-05}
{"train_loss": 0.1798219084739685, "global_step": 66107, "epoch": 742, "lr": 8.621743185302629e-05}
{"train_loss": 0.205213725566864, "global_step": 66108, "epoch": 742, "lr": 8.62170321605456e-05}
{"train_loss": 0.18360784649848938, "global_step": 66109, "epoch": 742, "lr": 8.621663246319597e-05}
{"train_loss": 0.24328134953975677, "global_step": 66110, "epoch": 742, "lr": 8.621623276097746e-05}
{"train_loss": 0.27033868432044983, "global_step": 66111, "epoch": 742, "lr": 8.621583305389011e-05}
{"train_loss": 0.15110228955745697, "global_step": 66112, "epoch": 742, "lr": 8.621543334193399e-05}
{"train_loss": 0.19033324718475342, "global_step": 66113, "epoch": 742, "lr": 8.621503362510914e-05}
{"train_loss": 0.3101457357406616, "global_step": 66114, "epoch": 742, "lr": 8.621463390341564e-05}
{"train_loss": 0.20601138472557068, "global_step": 66115, "epoch": 742, "lr": 8.62142341768535e-05}
{"train_loss": 0.2765127122402191, "global_step": 66116, "epoch": 742, "lr": 8.621383444542282e-05}
{"train_loss": 0.25613686442375183, "global_step": 66117, "epoch": 742, "lr": 8.621343470912362e-05}
{"train_loss": 0.27994704246520996, "global_step": 66118, "epoch": 742, "lr": 8.621303496795596e-05}
{"train_loss": 0.2131326049566269, "global_step": 66119, "epoch": 742, "lr": 8.62126352219199e-05}
{"train_loss": 0.2494347244501114, "global_step": 66120, "epoch": 742, "lr": 8.62122354710155e-05}
{"train_loss": 0.22406655550003052, "global_step": 66121, "epoch": 742, "lr": 8.621183571524281e-05}
{"train_loss": 0.2001943588256836, "global_step": 66122, "epoch": 742, "lr": 8.621143595460186e-05}
{"train_loss": 0.1614334136247635, "global_step": 66123, "epoch": 742, "lr": 8.621103618909274e-05}
{"train_loss": 0.24133706092834473, "global_step": 66124, "epoch": 742, "lr": 8.621063641871549e-05}
{"train_loss": 0.15613679587841034, "global_step": 66125, "epoch": 742, "lr": 8.621023664347015e-05}
{"train_loss": 0.23050179022751496, "global_step": 66126, "epoch": 742, "lr": 8.62098368633568e-05, "val_loss": 3.4905896186828613}
{"train_loss": 0.2659166753292084, "global_step": 66127, "epoch": 743, "lr": 8.620943707837546e-05}
{"train_loss": 0.19766288995742798, "global_step": 66128, "epoch": 743, "lr": 8.62090372885262e-05}
{"train_loss": 0.22275948524475098, "global_step": 66129, "epoch": 743, "lr": 8.62086374938091e-05}
{"train_loss": 0.2220519483089447, "global_step": 66130, "epoch": 743, "lr": 8.620823769422417e-05}
{"train_loss": 0.18658266961574554, "global_step": 66131, "epoch": 743, "lr": 8.620783788977149e-05}
{"train_loss": 0.20291374623775482, "global_step": 66132, "epoch": 743, "lr": 8.62074380804511e-05}
{"train_loss": 0.2267664521932602, "global_step": 66133, "epoch": 743, "lr": 8.620703826626308e-05}
{"train_loss": 0.22639767825603485, "global_step": 66134, "epoch": 743, "lr": 8.620663844720746e-05}
{"train_loss": 0.1770823895931244, "global_step": 66135, "epoch": 743, "lr": 8.62062386232843e-05}
{"train_loss": 0.19980935752391815, "global_step": 66136, "epoch": 743, "lr": 8.620583879449364e-05}
{"train_loss": 0.20174016058444977, "global_step": 66137, "epoch": 743, "lr": 8.620543896083555e-05}
{"train_loss": 0.23692139983177185, "global_step": 66138, "epoch": 743, "lr": 8.620503912231008e-05}
{"train_loss": 0.2488071769475937, "global_step": 66139, "epoch": 743, "lr": 8.620463927891728e-05}
{"train_loss": 0.259194940328598, "global_step": 66140, "epoch": 743, "lr": 8.620423943065723e-05}
{"train_loss": 0.2184649109840393, "global_step": 66141, "epoch": 743, "lr": 8.620383957752995e-05}
{"train_loss": 0.2206198126077652, "global_step": 66142, "epoch": 743, "lr": 8.62034397195355e-05}
{"train_loss": 0.2656345069408417, "global_step": 66143, "epoch": 743, "lr": 8.620303985667392e-05}
{"train_loss": 0.20246094465255737, "global_step": 66144, "epoch": 743, "lr": 8.620263998894532e-05}
{"train_loss": 0.1951870322227478, "global_step": 66145, "epoch": 743, "lr": 8.62022401163497e-05}
{"train_loss": 0.21822386980056763, "global_step": 66146, "epoch": 743, "lr": 8.620184023888712e-05}
{"train_loss": 0.20278222858905792, "global_step": 66147, "epoch": 743, "lr": 8.620144035655765e-05}
{"train_loss": 0.26150277256965637, "global_step": 66148, "epoch": 743, "lr": 8.620104046936135e-05}
{"train_loss": 0.23697464168071747, "global_step": 66149, "epoch": 743, "lr": 8.620064057729824e-05}
{"train_loss": 0.25307875871658325, "global_step": 66150, "epoch": 743, "lr": 8.620024068036841e-05}
{"train_loss": 0.29062017798423767, "global_step": 66151, "epoch": 743, "lr": 8.619984077857189e-05}
{"train_loss": 0.1924942433834076, "global_step": 66152, "epoch": 743, "lr": 8.619944087190874e-05}
{"train_loss": 0.2949705719947815, "global_step": 66153, "epoch": 743, "lr": 8.619904096037903e-05}
{"train_loss": 0.2280576527118683, "global_step": 66154, "epoch": 743, "lr": 8.61986410439828e-05}
{"train_loss": 0.2581641674041748, "global_step": 66155, "epoch": 743, "lr": 8.61982411227201e-05}
{"train_loss": 0.21328948438167572, "global_step": 66156, "epoch": 743, "lr": 8.619784119659098e-05}
{"train_loss": 0.2064184993505478, "global_step": 66157, "epoch": 743, "lr": 8.61974412655955e-05}
{"train_loss": 0.14700576663017273, "global_step": 66158, "epoch": 743, "lr": 8.619704132973373e-05}
{"train_loss": 0.3218134343624115, "global_step": 66159, "epoch": 743, "lr": 8.61966413890057e-05}
{"train_loss": 0.26570016145706177, "global_step": 66160, "epoch": 743, "lr": 8.619624144341147e-05}
{"train_loss": 0.20292310416698456, "global_step": 66161, "epoch": 743, "lr": 8.61958414929511e-05}
{"train_loss": 0.2026924043893814, "global_step": 66162, "epoch": 743, "lr": 8.619544153762464e-05}
{"train_loss": 0.2165956348180771, "global_step": 66163, "epoch": 743, "lr": 8.619504157743215e-05}
{"train_loss": 0.2825770974159241, "global_step": 66164, "epoch": 743, "lr": 8.619464161237367e-05}
{"train_loss": 0.17532026767730713, "global_step": 66165, "epoch": 743, "lr": 8.619424164244926e-05}
{"train_loss": 0.1526205837726593, "global_step": 66166, "epoch": 743, "lr": 8.619384166765898e-05}
{"train_loss": 0.19240710139274597, "global_step": 66167, "epoch": 743, "lr": 8.619344168800288e-05}
{"train_loss": 0.20149099826812744, "global_step": 66168, "epoch": 743, "lr": 8.619304170348102e-05}
{"train_loss": 0.16438429057598114, "global_step": 66169, "epoch": 743, "lr": 8.619264171409343e-05}
{"train_loss": 0.18291965126991272, "global_step": 66170, "epoch": 743, "lr": 8.619224171984019e-05}
{"train_loss": 0.26911067962646484, "global_step": 66171, "epoch": 743, "lr": 8.619184172072135e-05}
{"train_loss": 0.24496816098690033, "global_step": 66172, "epoch": 743, "lr": 8.619144171673695e-05}
{"train_loss": 0.22893677651882172, "global_step": 66173, "epoch": 743, "lr": 8.619104170788705e-05}
{"train_loss": 0.20859935879707336, "global_step": 66174, "epoch": 743, "lr": 8.61906416941717e-05}
{"train_loss": 0.22836166620254517, "global_step": 66175, "epoch": 743, "lr": 8.619024167559097e-05}
{"train_loss": 0.27693110704421997, "global_step": 66176, "epoch": 743, "lr": 8.618984165214491e-05}
{"train_loss": 0.21101941168308258, "global_step": 66177, "epoch": 743, "lr": 8.618944162383356e-05}
{"train_loss": 0.2814693748950958, "global_step": 66178, "epoch": 743, "lr": 8.618904159065697e-05}
{"train_loss": 0.2686674892902374, "global_step": 66179, "epoch": 743, "lr": 8.618864155261521e-05}
{"train_loss": 0.17929232120513916, "global_step": 66180, "epoch": 743, "lr": 8.618824150970833e-05}
{"train_loss": 0.23071026802062988, "global_step": 66181, "epoch": 743, "lr": 8.618784146193639e-05}
{"train_loss": 0.270954966545105, "global_step": 66182, "epoch": 743, "lr": 8.618744140929943e-05}
{"train_loss": 0.1845822036266327, "global_step": 66183, "epoch": 743, "lr": 8.618704135179751e-05}
{"train_loss": 0.2433614432811737, "global_step": 66184, "epoch": 743, "lr": 8.618664128943067e-05}
{"train_loss": 0.2704208791255951, "global_step": 66185, "epoch": 743, "lr": 8.6186241222199e-05}
{"train_loss": 0.21086078882217407, "global_step": 66186, "epoch": 743, "lr": 8.618584115010251e-05}
{"train_loss": 0.37318283319473267, "global_step": 66187, "epoch": 743, "lr": 8.618544107314128e-05}
{"train_loss": 0.361323744058609, "global_step": 66188, "epoch": 743, "lr": 8.618504099131537e-05}
{"train_loss": 0.3209148943424225, "global_step": 66189, "epoch": 743, "lr": 8.618464090462481e-05}
{"train_loss": 0.234413743019104, "global_step": 66190, "epoch": 743, "lr": 8.618424081306968e-05}
{"train_loss": 0.2827983498573303, "global_step": 66191, "epoch": 743, "lr": 8.618384071665001e-05}
{"train_loss": 0.257230281829834, "global_step": 66192, "epoch": 743, "lr": 8.618344061536586e-05}
{"train_loss": 0.22628770768642426, "global_step": 66193, "epoch": 743, "lr": 8.618304050921729e-05}
{"train_loss": 0.3068554401397705, "global_step": 66194, "epoch": 743, "lr": 8.618264039820436e-05}
{"train_loss": 0.31679755449295044, "global_step": 66195, "epoch": 743, "lr": 8.618224028232711e-05}
{"train_loss": 0.24917201697826385, "global_step": 66196, "epoch": 743, "lr": 8.618184016158559e-05}
{"train_loss": 0.178523987531662, "global_step": 66197, "epoch": 743, "lr": 8.618144003597988e-05}
{"train_loss": 0.18646062910556793, "global_step": 66198, "epoch": 743, "lr": 8.618103990551e-05}
{"train_loss": 0.18311046063899994, "global_step": 66199, "epoch": 743, "lr": 8.618063977017604e-05}
{"train_loss": 0.2651064693927765, "global_step": 66200, "epoch": 743, "lr": 8.618023962997802e-05}
{"train_loss": 0.21065348386764526, "global_step": 66201, "epoch": 743, "lr": 8.617983948491599e-05}
{"train_loss": 0.3775327801704407, "global_step": 66202, "epoch": 743, "lr": 8.617943933499005e-05}
{"train_loss": 0.2532711327075958, "global_step": 66203, "epoch": 743, "lr": 8.617903918020021e-05}
{"train_loss": 0.3207221031188965, "global_step": 66204, "epoch": 743, "lr": 8.617863902054655e-05}
{"train_loss": 0.3416491150856018, "global_step": 66205, "epoch": 743, "lr": 8.617823885602909e-05}
{"train_loss": 0.38806790113449097, "global_step": 66206, "epoch": 743, "lr": 8.617783868664793e-05}
{"train_loss": 0.21668876707553864, "global_step": 66207, "epoch": 743, "lr": 8.617743851240309e-05}
{"train_loss": 0.29253509640693665, "global_step": 66208, "epoch": 743, "lr": 8.617703833329464e-05}
{"train_loss": 0.15439802408218384, "global_step": 66209, "epoch": 743, "lr": 8.617663814932263e-05}
{"train_loss": 0.28729546070098877, "global_step": 66210, "epoch": 743, "lr": 8.61762379604871e-05}
{"train_loss": 0.2772604823112488, "global_step": 66211, "epoch": 743, "lr": 8.617583776678813e-05}
{"train_loss": 0.24907782673835754, "global_step": 66212, "epoch": 743, "lr": 8.617543756822576e-05}
{"train_loss": 0.20513659715652466, "global_step": 66213, "epoch": 743, "lr": 8.617503736480004e-05}
{"train_loss": 0.36812594532966614, "global_step": 66214, "epoch": 743, "lr": 8.617463715651102e-05}
{"train_loss": 0.2411517370282934, "global_step": 66215, "epoch": 743, "lr": 8.617423694335876e-05, "val_loss": 3.5772945880889893}
{"train_loss": 0.24595652520656586, "global_step": 66216, "epoch": 744, "lr": 8.617383672534333e-05}
{"train_loss": 0.3250507712364197, "global_step": 66217, "epoch": 744, "lr": 8.617343650246475e-05}
{"train_loss": 0.20506344735622406, "global_step": 66218, "epoch": 744, "lr": 8.617303627472311e-05}
{"train_loss": 0.29692599177360535, "global_step": 66219, "epoch": 744, "lr": 8.617263604211843e-05}
{"train_loss": 0.2824229300022125, "global_step": 66220, "epoch": 744, "lr": 8.617223580465079e-05}
{"train_loss": 0.35596829652786255, "global_step": 66221, "epoch": 744, "lr": 8.617183556232022e-05}
{"train_loss": 0.26582780480384827, "global_step": 66222, "epoch": 744, "lr": 8.617143531512681e-05}
{"train_loss": 0.311390221118927, "global_step": 66223, "epoch": 744, "lr": 8.617103506307058e-05}
{"train_loss": 0.13956817984580994, "global_step": 66224, "epoch": 744, "lr": 8.61706348061516e-05}
{"train_loss": 0.1903044581413269, "global_step": 66225, "epoch": 744, "lr": 8.61702345443699e-05}
{"train_loss": 0.241546168923378, "global_step": 66226, "epoch": 744, "lr": 8.616983427772557e-05}
{"train_loss": 0.24537909030914307, "global_step": 66227, "epoch": 744, "lr": 8.616943400621864e-05}
{"train_loss": 0.30323171615600586, "global_step": 66228, "epoch": 744, "lr": 8.616903372984918e-05}
{"train_loss": 0.19198766350746155, "global_step": 66229, "epoch": 744, "lr": 8.616863344861723e-05}
{"train_loss": 0.22895821928977966, "global_step": 66230, "epoch": 744, "lr": 8.616823316252285e-05}
{"train_loss": 0.305622398853302, "global_step": 66231, "epoch": 744, "lr": 8.616783287156608e-05}
{"train_loss": 0.1919003129005432, "global_step": 66232, "epoch": 744, "lr": 8.6167432575747e-05}
{"train_loss": 0.18370932340621948, "global_step": 66233, "epoch": 744, "lr": 8.616703227506564e-05}
{"train_loss": 0.2337135374546051, "global_step": 66234, "epoch": 744, "lr": 8.616663196952206e-05}
{"train_loss": 0.21723195910453796, "global_step": 66235, "epoch": 744, "lr": 8.616623165911633e-05}
{"train_loss": 0.15781383216381073, "global_step": 66236, "epoch": 744, "lr": 8.616583134384848e-05}
{"train_loss": 0.2356986105442047, "global_step": 66237, "epoch": 744, "lr": 8.616543102371857e-05}
{"train_loss": 0.24267058074474335, "global_step": 66238, "epoch": 744, "lr": 8.616503069872668e-05}
{"train_loss": 0.2088329941034317, "global_step": 66239, "epoch": 744, "lr": 8.616463036887283e-05}
{"train_loss": 0.17356307804584503, "global_step": 66240, "epoch": 744, "lr": 8.616423003415708e-05}
{"train_loss": 0.2318074256181717, "global_step": 66241, "epoch": 744, "lr": 8.61638296945795e-05}
{"train_loss": 0.30706819891929626, "global_step": 66242, "epoch": 744, "lr": 8.616342935014014e-05}
{"train_loss": 0.22225496172904968, "global_step": 66243, "epoch": 744, "lr": 8.616302900083904e-05}
{"train_loss": 0.24586530029773712, "global_step": 66244, "epoch": 744, "lr": 8.616262864667625e-05}
{"train_loss": 0.22555868327617645, "global_step": 66245, "epoch": 744, "lr": 8.616222828765185e-05}
{"train_loss": 0.2017308622598648, "global_step": 66246, "epoch": 744, "lr": 8.616182792376587e-05}
{"train_loss": 0.24840910732746124, "global_step": 66247, "epoch": 744, "lr": 8.616142755501837e-05}
{"train_loss": 0.28284215927124023, "global_step": 66248, "epoch": 744, "lr": 8.616102718140943e-05}
{"train_loss": 0.2378140389919281, "global_step": 66249, "epoch": 744, "lr": 8.616062680293906e-05}
{"train_loss": 0.17258033156394958, "global_step": 66250, "epoch": 744, "lr": 8.616022641960734e-05}
{"train_loss": 0.3146957755088806, "global_step": 66251, "epoch": 744, "lr": 8.615982603141432e-05}
{"train_loss": 0.34506070613861084, "global_step": 66252, "epoch": 744, "lr": 8.615942563836005e-05}
{"train_loss": 0.18188193440437317, "global_step": 66253, "epoch": 744, "lr": 8.615902524044458e-05}
{"train_loss": 0.16116368770599365, "global_step": 66254, "epoch": 744, "lr": 8.615862483766797e-05}
{"train_loss": 0.2370167374610901, "global_step": 66255, "epoch": 744, "lr": 8.615822443003028e-05}
{"train_loss": 0.2222987413406372, "global_step": 66256, "epoch": 744, "lr": 8.615782401753156e-05}
{"train_loss": 0.19587652385234833, "global_step": 66257, "epoch": 744, "lr": 8.615742360017186e-05}
{"train_loss": 0.24704045057296753, "global_step": 66258, "epoch": 744, "lr": 8.615702317795123e-05}
{"train_loss": 0.23012447357177734, "global_step": 66259, "epoch": 744, "lr": 8.615662275086973e-05}
{"train_loss": 0.2515460252761841, "global_step": 66260, "epoch": 744, "lr": 8.615622231892742e-05}
{"train_loss": 0.20102094113826752, "global_step": 66261, "epoch": 744, "lr": 8.615582188212435e-05}
{"train_loss": 0.21676501631736755, "global_step": 66262, "epoch": 744, "lr": 8.615542144046055e-05}
{"train_loss": 0.2698479890823364, "global_step": 66263, "epoch": 744, "lr": 8.615502099393611e-05}
{"train_loss": 0.133640319108963, "global_step": 66264, "epoch": 744, "lr": 8.615462054255106e-05}
{"train_loss": 0.2572856545448303, "global_step": 66265, "epoch": 744, "lr": 8.615422008630546e-05}
{"train_loss": 0.1550516039133072, "global_step": 66266, "epoch": 744, "lr": 8.615381962519938e-05}
{"train_loss": 0.24745868146419525, "global_step": 66267, "epoch": 744, "lr": 8.615341915923285e-05}
{"train_loss": 0.17001892626285553, "global_step": 66268, "epoch": 744, "lr": 8.615301868840593e-05}
{"train_loss": 0.32252997159957886, "global_step": 66269, "epoch": 744, "lr": 8.615261821271869e-05}
{"train_loss": 0.18747279047966003, "global_step": 66270, "epoch": 744, "lr": 8.615221773217116e-05}
{"train_loss": 0.26562005281448364, "global_step": 66271, "epoch": 744, "lr": 8.61518172467634e-05}
{"train_loss": 0.3088696002960205, "global_step": 66272, "epoch": 744, "lr": 8.615141675649548e-05}
{"train_loss": 0.2328527271747589, "global_step": 66273, "epoch": 744, "lr": 8.615101626136744e-05}
{"train_loss": 0.1758461594581604, "global_step": 66274, "epoch": 744, "lr": 8.615061576137932e-05}
{"train_loss": 0.25747236609458923, "global_step": 66275, "epoch": 744, "lr": 8.615021525653122e-05}
{"train_loss": 0.20932430028915405, "global_step": 66276, "epoch": 744, "lr": 8.614981474682315e-05}
{"train_loss": 0.27868321537971497, "global_step": 66277, "epoch": 744, "lr": 8.614941423225518e-05}
{"train_loss": 0.18711642920970917, "global_step": 66278, "epoch": 744, "lr": 8.614901371282736e-05}
{"train_loss": 0.2213825285434723, "global_step": 66279, "epoch": 744, "lr": 8.614861318853976e-05}
{"train_loss": 0.26349878311157227, "global_step": 66280, "epoch": 744, "lr": 8.614821265939239e-05}
{"train_loss": 0.15495245158672333, "global_step": 66281, "epoch": 744, "lr": 8.614781212538535e-05}
{"train_loss": 0.21927642822265625, "global_step": 66282, "epoch": 744, "lr": 8.614741158651867e-05}
{"train_loss": 0.3573222756385803, "global_step": 66283, "epoch": 744, "lr": 8.614701104279241e-05}
{"train_loss": 0.2603297233581543, "global_step": 66284, "epoch": 744, "lr": 8.614661049420665e-05}
{"train_loss": 0.26062464714050293, "global_step": 66285, "epoch": 744, "lr": 8.614620994076139e-05}
{"train_loss": 0.28379765152931213, "global_step": 66286, "epoch": 744, "lr": 8.614580938245674e-05}
{"train_loss": 0.27213215827941895, "global_step": 66287, "epoch": 744, "lr": 8.614540881929269e-05}
{"train_loss": 0.38678881525993347, "global_step": 66288, "epoch": 744, "lr": 8.614500825126936e-05}
{"train_loss": 0.31479042768478394, "global_step": 66289, "epoch": 744, "lr": 8.614460767838677e-05}
{"train_loss": 0.22541172802448273, "global_step": 66290, "epoch": 744, "lr": 8.614420710064496e-05}
{"train_loss": 0.24100157618522644, "global_step": 66291, "epoch": 744, "lr": 8.614380651804402e-05}
{"train_loss": 0.20087549090385437, "global_step": 66292, "epoch": 744, "lr": 8.614340593058398e-05}
{"train_loss": 0.30785417556762695, "global_step": 66293, "epoch": 744, "lr": 8.61430053382649e-05}
{"train_loss": 0.31878602504730225, "global_step": 66294, "epoch": 744, "lr": 8.614260474108683e-05}
{"train_loss": 0.2379804402589798, "global_step": 66295, "epoch": 744, "lr": 8.614220413904984e-05}
{"train_loss": 0.21007537841796875, "global_step": 66296, "epoch": 744, "lr": 8.614180353215396e-05}
{"train_loss": 0.26342707872390747, "global_step": 66297, "epoch": 744, "lr": 8.614140292039925e-05}
{"train_loss": 0.20202670991420746, "global_step": 66298, "epoch": 744, "lr": 8.614100230378578e-05}
{"train_loss": 0.31555819511413574, "global_step": 66299, "epoch": 744, "lr": 8.614060168231359e-05}
{"train_loss": 0.37804555892944336, "global_step": 66300, "epoch": 744, "lr": 8.614020105598273e-05}
{"train_loss": 0.2119705080986023, "global_step": 66301, "epoch": 744, "lr": 8.613980042479326e-05}
{"train_loss": 0.21481548249721527, "global_step": 66302, "epoch": 744, "lr": 8.613939978874525e-05}
{"train_loss": 0.2551606297492981, "global_step": 66303, "epoch": 744, "lr": 8.613899914783872e-05}
{"train_loss": 0.2428752759869179, "global_step": 66304, "epoch": 744, "lr": 8.613859850207375e-05, "val_loss": 3.578279972076416}
{"train_loss": 0.24249805510044098, "global_step": 66305, "epoch": 745, "lr": 8.613819785145039e-05}
{"train_loss": 0.1723567098379135, "global_step": 66306, "epoch": 745, "lr": 8.613779719596868e-05}
{"train_loss": 0.2149287313222885, "global_step": 66307, "epoch": 745, "lr": 8.613739653562869e-05}
{"train_loss": 0.28154993057250977, "global_step": 66308, "epoch": 745, "lr": 8.613699587043048e-05}
{"train_loss": 0.18419861793518066, "global_step": 66309, "epoch": 745, "lr": 8.613659520037407e-05}
{"train_loss": 0.23559816181659698, "global_step": 66310, "epoch": 745, "lr": 8.613619452545953e-05}
{"train_loss": 0.3333711624145508, "global_step": 66311, "epoch": 745, "lr": 8.613579384568694e-05}
{"train_loss": 0.13721510767936707, "global_step": 66312, "epoch": 745, "lr": 8.613539316105632e-05}
{"train_loss": 0.2421373724937439, "global_step": 66313, "epoch": 745, "lr": 8.613499247156775e-05}
{"train_loss": 0.1595872938632965, "global_step": 66314, "epoch": 745, "lr": 8.613459177722126e-05}
{"train_loss": 0.2672540545463562, "global_step": 66315, "epoch": 745, "lr": 8.613419107801692e-05}
{"train_loss": 0.23693393170833588, "global_step": 66316, "epoch": 745, "lr": 8.613379037395478e-05}
{"train_loss": 0.2400001883506775, "global_step": 66317, "epoch": 745, "lr": 8.613338966503488e-05}
{"train_loss": 0.2078947126865387, "global_step": 66318, "epoch": 745, "lr": 8.613298895125729e-05}
{"train_loss": 0.19150304794311523, "global_step": 66319, "epoch": 745, "lr": 8.613258823262207e-05}
{"train_loss": 0.15245619416236877, "global_step": 66320, "epoch": 745, "lr": 8.613218750912926e-05}
{"train_loss": 0.21455040574073792, "global_step": 66321, "epoch": 745, "lr": 8.613178678077891e-05}
{"train_loss": 0.26668837666511536, "global_step": 66322, "epoch": 745, "lr": 8.61313860475711e-05}
{"train_loss": 0.20448774099349976, "global_step": 66323, "epoch": 745, "lr": 8.613098530950585e-05}
{"train_loss": 0.2671223282814026, "global_step": 66324, "epoch": 745, "lr": 8.613058456658323e-05}
{"train_loss": 0.31863343715667725, "global_step": 66325, "epoch": 745, "lr": 8.61301838188033e-05}
{"train_loss": 0.22028270363807678, "global_step": 66326, "epoch": 745, "lr": 8.61297830661661e-05}
{"train_loss": 0.1815127730369568, "global_step": 66327, "epoch": 745, "lr": 8.612938230867169e-05}
{"train_loss": 0.21372227370738983, "global_step": 66328, "epoch": 745, "lr": 8.612898154632013e-05}
{"train_loss": 0.24871543049812317, "global_step": 66329, "epoch": 745, "lr": 8.612858077911146e-05}
{"train_loss": 0.25810912251472473, "global_step": 66330, "epoch": 745, "lr": 8.612818000704577e-05}
{"train_loss": 0.13754838705062866, "global_step": 66331, "epoch": 745, "lr": 8.612777923012305e-05}
{"train_loss": 0.30998942255973816, "global_step": 66332, "epoch": 745, "lr": 8.612737844834342e-05}
{"train_loss": 0.23659171164035797, "global_step": 66333, "epoch": 745, "lr": 8.612697766170689e-05}
{"train_loss": 0.18296240270137787, "global_step": 66334, "epoch": 745, "lr": 8.612657687021353e-05}
{"train_loss": 0.19132666289806366, "global_step": 66335, "epoch": 745, "lr": 8.61261760738634e-05}
{"train_loss": 0.19297631084918976, "global_step": 66336, "epoch": 745, "lr": 8.612577527265653e-05}
{"train_loss": 0.15873831510543823, "global_step": 66337, "epoch": 745, "lr": 8.6125374466593e-05}
{"train_loss": 0.20174671709537506, "global_step": 66338, "epoch": 745, "lr": 8.612497365567284e-05}
{"train_loss": 0.18012289702892303, "global_step": 66339, "epoch": 745, "lr": 8.612457283989613e-05}
{"train_loss": 0.22066326439380646, "global_step": 66340, "epoch": 745, "lr": 8.61241720192629e-05}
{"train_loss": 0.2686920762062073, "global_step": 66341, "epoch": 745, "lr": 8.612377119377324e-05}
{"train_loss": 0.32756468653678894, "global_step": 66342, "epoch": 745, "lr": 8.612337036342717e-05}
{"train_loss": 0.16503198444843292, "global_step": 66343, "epoch": 745, "lr": 8.612296952822474e-05}
{"train_loss": 0.1682644933462143, "global_step": 66344, "epoch": 745, "lr": 8.612256868816603e-05}
{"train_loss": 0.1783689558506012, "global_step": 66345, "epoch": 745, "lr": 8.612216784325108e-05}
{"train_loss": 0.16536997258663177, "global_step": 66346, "epoch": 745, "lr": 8.612176699347993e-05}
{"train_loss": 0.2951802611351013, "global_step": 66347, "epoch": 745, "lr": 8.612136613885266e-05}
{"train_loss": 0.14965170621871948, "global_step": 66348, "epoch": 745, "lr": 8.612096527936932e-05}
{"train_loss": 0.24696825444698334, "global_step": 66349, "epoch": 745, "lr": 8.612056441502994e-05}
{"train_loss": 0.24272525310516357, "global_step": 66350, "epoch": 745, "lr": 8.612016354583461e-05}
{"train_loss": 0.30375346541404724, "global_step": 66351, "epoch": 745, "lr": 8.611976267178335e-05}
{"train_loss": 0.20208729803562164, "global_step": 66352, "epoch": 745, "lr": 8.611936179287624e-05}
{"train_loss": 0.24889366328716278, "global_step": 66353, "epoch": 745, "lr": 8.61189609091133e-05}
{"train_loss": 0.1617778092622757, "global_step": 66354, "epoch": 745, "lr": 8.611856002049464e-05}
{"train_loss": 0.26491981744766235, "global_step": 66355, "epoch": 745, "lr": 8.611815912702024e-05}
{"train_loss": 0.2836149334907532, "global_step": 66356, "epoch": 745, "lr": 8.611775822869022e-05}
{"train_loss": 0.1792394071817398, "global_step": 66357, "epoch": 745, "lr": 8.611735732550458e-05}
{"train_loss": 0.24007943272590637, "global_step": 66358, "epoch": 745, "lr": 8.611695641746343e-05}
{"train_loss": 0.21222946047782898, "global_step": 66359, "epoch": 745, "lr": 8.611655550456678e-05}
{"train_loss": 0.20210857689380646, "global_step": 66360, "epoch": 745, "lr": 8.611615458681472e-05}
{"train_loss": 0.28662118315696716, "global_step": 66361, "epoch": 745, "lr": 8.611575366420727e-05}
{"train_loss": 0.19637510180473328, "global_step": 66362, "epoch": 745, "lr": 8.611535273674449e-05}
{"train_loss": 0.2009599506855011, "global_step": 66363, "epoch": 745, "lr": 8.611495180442646e-05}
{"train_loss": 0.25827690958976746, "global_step": 66364, "epoch": 745, "lr": 8.61145508672532e-05}
{"train_loss": 0.19336694478988647, "global_step": 66365, "epoch": 745, "lr": 8.611414992522478e-05}
{"train_loss": 0.2036987841129303, "global_step": 66366, "epoch": 745, "lr": 8.611374897834125e-05}
{"train_loss": 0.21763445436954498, "global_step": 66367, "epoch": 745, "lr": 8.611334802660267e-05}
{"train_loss": 0.20251035690307617, "global_step": 66368, "epoch": 745, "lr": 8.611294707000909e-05}
{"train_loss": 0.21570050716400146, "global_step": 66369, "epoch": 745, "lr": 8.611254610856056e-05}
{"train_loss": 0.21120403707027435, "global_step": 66370, "epoch": 745, "lr": 8.611214514225716e-05}
{"train_loss": 0.23388609290122986, "global_step": 66371, "epoch": 745, "lr": 8.61117441710989e-05}
{"train_loss": 0.2576921284198761, "global_step": 66372, "epoch": 745, "lr": 8.611134319508586e-05}
{"train_loss": 0.22149191796779633, "global_step": 66373, "epoch": 745, "lr": 8.611094221421809e-05}
{"train_loss": 0.1436854898929596, "global_step": 66374, "epoch": 745, "lr": 8.611054122849566e-05}
{"train_loss": 0.21892209351062775, "global_step": 66375, "epoch": 745, "lr": 8.61101402379186e-05}
{"train_loss": 0.24567465484142303, "global_step": 66376, "epoch": 745, "lr": 8.610973924248695e-05}
{"train_loss": 0.25363290309906006, "global_step": 66377, "epoch": 745, "lr": 8.610933824220081e-05}
{"train_loss": 0.12876665592193604, "global_step": 66378, "epoch": 745, "lr": 8.610893723706019e-05}
{"train_loss": 0.17387655377388, "global_step": 66379, "epoch": 745, "lr": 8.610853622706519e-05}
{"train_loss": 0.22278745472431183, "global_step": 66380, "epoch": 745, "lr": 8.610813521221581e-05}
{"train_loss": 0.20000573992729187, "global_step": 66381, "epoch": 745, "lr": 8.610773419251215e-05}
{"train_loss": 0.27973031997680664, "global_step": 66382, "epoch": 745, "lr": 8.610733316795424e-05}
{"train_loss": 0.2519082725048065, "global_step": 66383, "epoch": 745, "lr": 8.610693213854214e-05}
{"train_loss": 0.24801409244537354, "global_step": 66384, "epoch": 745, "lr": 8.61065311042759e-05}
{"train_loss": 0.17419970035552979, "global_step": 66385, "epoch": 745, "lr": 8.610613006515558e-05}
{"train_loss": 0.2056107372045517, "global_step": 66386, "epoch": 745, "lr": 8.610572902118122e-05}
{"train_loss": 0.21154096722602844, "global_step": 66387, "epoch": 745, "lr": 8.61053279723529e-05}
{"train_loss": 0.21863581240177155, "global_step": 66388, "epoch": 745, "lr": 8.610492691867064e-05}
{"train_loss": 0.19809149205684662, "global_step": 66389, "epoch": 745, "lr": 8.610452586013453e-05}
{"train_loss": 0.2387741506099701, "global_step": 66390, "epoch": 745, "lr": 8.610412479674461e-05}
{"train_loss": 0.16815975308418274, "global_step": 66391, "epoch": 745, "lr": 8.610372372850093e-05}
{"train_loss": 0.18981632590293884, "global_step": 66392, "epoch": 745, "lr": 8.610332265540353e-05}
{"train_loss": 0.21992488908633756, "global_step": 66393, "epoch": 745, "lr": 8.610292157745248e-05, "val_loss": 3.594883441925049, "train_action_mse_error": 26.26935386657715}
{"train_loss": 0.16829806566238403, "global_step": 66394, "epoch": 746, "lr": 8.610252049464784e-05}
{"train_loss": 0.168617844581604, "global_step": 66395, "epoch": 746, "lr": 8.610211940698965e-05}
{"train_loss": 0.14522337913513184, "global_step": 66396, "epoch": 746, "lr": 8.610171831447798e-05}
{"train_loss": 0.15427140891551971, "global_step": 66397, "epoch": 746, "lr": 8.610131721711287e-05}
{"train_loss": 0.24459554255008698, "global_step": 66398, "epoch": 746, "lr": 8.610091611489437e-05}
{"train_loss": 0.19060668349266052, "global_step": 66399, "epoch": 746, "lr": 8.610051500782254e-05}
{"train_loss": 0.18651604652404785, "global_step": 66400, "epoch": 746, "lr": 8.610011389589745e-05}
{"train_loss": 0.18205581605434418, "global_step": 66401, "epoch": 746, "lr": 8.609971277911914e-05}
{"train_loss": 0.19730833172798157, "global_step": 66402, "epoch": 746, "lr": 8.609931165748766e-05}
{"train_loss": 0.1880759745836258, "global_step": 66403, "epoch": 746, "lr": 8.609891053100305e-05}
{"train_loss": 0.14013636112213135, "global_step": 66404, "epoch": 746, "lr": 8.60985093996654e-05}
{"train_loss": 0.2584094703197479, "global_step": 66405, "epoch": 746, "lr": 8.609810826347474e-05}
{"train_loss": 0.2518027126789093, "global_step": 66406, "epoch": 746, "lr": 8.609770712243112e-05}
{"train_loss": 0.20940764248371124, "global_step": 66407, "epoch": 746, "lr": 8.609730597653462e-05}
{"train_loss": 0.3318166434764862, "global_step": 66408, "epoch": 746, "lr": 8.609690482578527e-05}
{"train_loss": 0.23034146428108215, "global_step": 66409, "epoch": 746, "lr": 8.609650367018314e-05}
{"train_loss": 0.23504821956157684, "global_step": 66410, "epoch": 746, "lr": 8.609610250972826e-05}
{"train_loss": 0.17674580216407776, "global_step": 66411, "epoch": 746, "lr": 8.609570134442071e-05}
{"train_loss": 0.22067546844482422, "global_step": 66412, "epoch": 746, "lr": 8.609530017426052e-05}
{"train_loss": 0.2521018981933594, "global_step": 66413, "epoch": 746, "lr": 8.609489899924776e-05}
{"train_loss": 0.25518691539764404, "global_step": 66414, "epoch": 746, "lr": 8.60944978193825e-05}
{"train_loss": 0.2101103961467743, "global_step": 66415, "epoch": 746, "lr": 8.609409663466475e-05}
{"train_loss": 0.2672765552997589, "global_step": 66416, "epoch": 746, "lr": 8.60936954450946e-05}
{"train_loss": 0.15372927486896515, "global_step": 66417, "epoch": 746, "lr": 8.609329425067209e-05}
{"train_loss": 0.21107636392116547, "global_step": 66418, "epoch": 746, "lr": 8.609289305139728e-05}
{"train_loss": 0.22644665837287903, "global_step": 66419, "epoch": 746, "lr": 8.609249184727022e-05}
{"train_loss": 0.2541087567806244, "global_step": 66420, "epoch": 746, "lr": 8.609209063829097e-05}
{"train_loss": 0.2368832379579544, "global_step": 66421, "epoch": 746, "lr": 8.609168942445958e-05}
{"train_loss": 0.25370872020721436, "global_step": 66422, "epoch": 746, "lr": 8.609128820577607e-05}
{"train_loss": 0.15724807977676392, "global_step": 66423, "epoch": 746, "lr": 8.609088698224057e-05}
{"train_loss": 0.22005364298820496, "global_step": 66424, "epoch": 746, "lr": 8.609048575385306e-05}
{"train_loss": 0.22504094243049622, "global_step": 66425, "epoch": 746, "lr": 8.609008452061365e-05}
{"train_loss": 0.16172166168689728, "global_step": 66426, "epoch": 746, "lr": 8.608968328252234e-05}
{"train_loss": 0.2512381374835968, "global_step": 66427, "epoch": 746, "lr": 8.608928203957923e-05}
{"train_loss": 0.2354603111743927, "global_step": 66428, "epoch": 746, "lr": 8.608888079178435e-05}
{"train_loss": 0.1988784223794937, "global_step": 66429, "epoch": 746, "lr": 8.608847953913776e-05}
{"train_loss": 0.17010638117790222, "global_step": 66430, "epoch": 746, "lr": 8.60880782816395e-05}
{"train_loss": 0.15890097618103027, "global_step": 66431, "epoch": 746, "lr": 8.608767701928966e-05}
{"train_loss": 0.18325670063495636, "global_step": 66432, "epoch": 746, "lr": 8.608727575208825e-05}
{"train_loss": 0.1402696818113327, "global_step": 66433, "epoch": 746, "lr": 8.608687448003537e-05}
{"train_loss": 0.2455979734659195, "global_step": 66434, "epoch": 746, "lr": 8.608647320313103e-05}
{"train_loss": 0.14806273579597473, "global_step": 66435, "epoch": 746, "lr": 8.608607192137532e-05}
{"train_loss": 0.22275520861148834, "global_step": 66436, "epoch": 746, "lr": 8.608567063476827e-05}
{"train_loss": 0.30279669165611267, "global_step": 66437, "epoch": 746, "lr": 8.608526934330994e-05}
{"train_loss": 0.2110370248556137, "global_step": 66438, "epoch": 746, "lr": 8.608486804700038e-05}
{"train_loss": 0.15786166489124298, "global_step": 66439, "epoch": 746, "lr": 8.608446674583966e-05}
{"train_loss": 0.2864592373371124, "global_step": 66440, "epoch": 746, "lr": 8.608406543982782e-05}
{"train_loss": 0.3031694293022156, "global_step": 66441, "epoch": 746, "lr": 8.608366412896491e-05}
{"train_loss": 0.325502872467041, "global_step": 66442, "epoch": 746, "lr": 8.6083262813251e-05}
{"train_loss": 0.23062729835510254, "global_step": 66443, "epoch": 746, "lr": 8.608286149268614e-05}
{"train_loss": 0.23087482154369354, "global_step": 66444, "epoch": 746, "lr": 8.608246016727037e-05}
{"train_loss": 0.18615418672561646, "global_step": 66445, "epoch": 746, "lr": 8.608205883700376e-05}
{"train_loss": 0.26794785261154175, "global_step": 66446, "epoch": 746, "lr": 8.608165750188634e-05}
{"train_loss": 0.15738897025585175, "global_step": 66447, "epoch": 746, "lr": 8.608125616191819e-05}
{"train_loss": 0.15476907789707184, "global_step": 66448, "epoch": 746, "lr": 8.608085481709935e-05}
{"train_loss": 0.29281362891197205, "global_step": 66449, "epoch": 746, "lr": 8.60804534674299e-05}
{"train_loss": 0.22430634498596191, "global_step": 66450, "epoch": 746, "lr": 8.608005211290986e-05}
{"train_loss": 0.22594070434570312, "global_step": 66451, "epoch": 746, "lr": 8.607965075353929e-05}
{"train_loss": 0.23842717707157135, "global_step": 66452, "epoch": 746, "lr": 8.607924938931827e-05}
{"train_loss": 0.29406803846359253, "global_step": 66453, "epoch": 746, "lr": 8.607884802024681e-05}
{"train_loss": 0.23133549094200134, "global_step": 66454, "epoch": 746, "lr": 8.607844664632499e-05}
{"train_loss": 0.21535810828208923, "global_step": 66455, "epoch": 746, "lr": 8.607804526755287e-05}
{"train_loss": 0.25726479291915894, "global_step": 66456, "epoch": 746, "lr": 8.60776438839305e-05}
{"train_loss": 0.3168424069881439, "global_step": 66457, "epoch": 746, "lr": 8.607724249545793e-05}
{"train_loss": 0.2402389645576477, "global_step": 66458, "epoch": 746, "lr": 8.607684110213521e-05}
{"train_loss": 0.21095891296863556, "global_step": 66459, "epoch": 746, "lr": 8.60764397039624e-05}
{"train_loss": 0.2133404165506363, "global_step": 66460, "epoch": 746, "lr": 8.607603830093956e-05}
{"train_loss": 0.17519396543502808, "global_step": 66461, "epoch": 746, "lr": 8.607563689306673e-05}
{"train_loss": 0.1226566806435585, "global_step": 66462, "epoch": 746, "lr": 8.607523548034397e-05}
{"train_loss": 0.24816539883613586, "global_step": 66463, "epoch": 746, "lr": 8.607483406277135e-05}
{"train_loss": 0.3539423644542694, "global_step": 66464, "epoch": 746, "lr": 8.607443264034888e-05}
{"train_loss": 0.21928465366363525, "global_step": 66465, "epoch": 746, "lr": 8.607403121307665e-05}
{"train_loss": 0.14499914646148682, "global_step": 66466, "epoch": 746, "lr": 8.607362978095472e-05}
{"train_loss": 0.1739930957555771, "global_step": 66467, "epoch": 746, "lr": 8.607322834398313e-05}
{"train_loss": 0.19732992351055145, "global_step": 66468, "epoch": 746, "lr": 8.607282690216192e-05}
{"train_loss": 0.26566463708877563, "global_step": 66469, "epoch": 746, "lr": 8.607242545549116e-05}
{"train_loss": 0.275010883808136, "global_step": 66470, "epoch": 746, "lr": 8.60720240039709e-05}
{"train_loss": 0.2745252847671509, "global_step": 66471, "epoch": 746, "lr": 8.607162254760122e-05}
{"train_loss": 0.22541536390781403, "global_step": 66472, "epoch": 746, "lr": 8.607122108638212e-05}
{"train_loss": 0.2430916279554367, "global_step": 66473, "epoch": 746, "lr": 8.60708196203137e-05}
{"train_loss": 0.18411293625831604, "global_step": 66474, "epoch": 746, "lr": 8.607041814939598e-05}
{"train_loss": 0.2220921367406845, "global_step": 66475, "epoch": 746, "lr": 8.607001667362907e-05}
{"train_loss": 0.20414277911186218, "global_step": 66476, "epoch": 746, "lr": 8.606961519301295e-05}
{"train_loss": 0.26319336891174316, "global_step": 66477, "epoch": 746, "lr": 8.606921370754773e-05}
{"train_loss": 0.22076460719108582, "global_step": 66478, "epoch": 746, "lr": 8.606881221723343e-05}
{"train_loss": 0.2670760154724121, "global_step": 66479, "epoch": 746, "lr": 8.606841072207013e-05}
{"train_loss": 0.17375245690345764, "global_step": 66480, "epoch": 746, "lr": 8.606800922205785e-05}
{"train_loss": 0.1866670548915863, "global_step": 66481, "epoch": 746, "lr": 8.606760771719669e-05}
{"train_loss": 0.21978951429717997, "global_step": 66482, "epoch": 746, "lr": 8.606720620748667e-05, "val_loss": 3.6299617290496826}
{"train_loss": 0.27409663796424866, "global_step": 66483, "epoch": 747, "lr": 8.606680469292785e-05}
{"train_loss": 0.24071642756462097, "global_step": 66484, "epoch": 747, "lr": 8.60664031735203e-05}
{"train_loss": 0.19216695427894592, "global_step": 66485, "epoch": 747, "lr": 8.606600164926404e-05}
{"train_loss": 0.21384398639202118, "global_step": 66486, "epoch": 747, "lr": 8.606560012015916e-05}
{"train_loss": 0.25491565465927124, "global_step": 66487, "epoch": 747, "lr": 8.60651985862057e-05}
{"train_loss": 0.16376347839832306, "global_step": 66488, "epoch": 747, "lr": 8.606479704740373e-05}
{"train_loss": 0.20174027979373932, "global_step": 66489, "epoch": 747, "lr": 8.606439550375326e-05}
{"train_loss": 0.19479714334011078, "global_step": 66490, "epoch": 747, "lr": 8.606399395525439e-05}
{"train_loss": 0.2016698569059372, "global_step": 66491, "epoch": 747, "lr": 8.606359240190713e-05}
{"train_loss": 0.22853419184684753, "global_step": 66492, "epoch": 747, "lr": 8.606319084371158e-05}
{"train_loss": 0.25499552488327026, "global_step": 66493, "epoch": 747, "lr": 8.606278928066778e-05}
{"train_loss": 0.3463115692138672, "global_step": 66494, "epoch": 747, "lr": 8.606238771277576e-05}
{"train_loss": 0.22435608506202698, "global_step": 66495, "epoch": 747, "lr": 8.60619861400356e-05}
{"train_loss": 0.24241766333580017, "global_step": 66496, "epoch": 747, "lr": 8.606158456244735e-05}
{"train_loss": 0.22420354187488556, "global_step": 66497, "epoch": 747, "lr": 8.606118298001105e-05}
{"train_loss": 0.1182071641087532, "global_step": 66498, "epoch": 747, "lr": 8.606078139272678e-05}
{"train_loss": 0.2408233880996704, "global_step": 66499, "epoch": 747, "lr": 8.606037980059457e-05}
{"train_loss": 0.17527160048484802, "global_step": 66500, "epoch": 747, "lr": 8.605997820361446e-05}
{"train_loss": 0.14525945484638214, "global_step": 66501, "epoch": 747, "lr": 8.605957660178654e-05}
{"train_loss": 0.1320815086364746, "global_step": 66502, "epoch": 747, "lr": 8.605917499511086e-05}
{"train_loss": 0.2514820098876953, "global_step": 66503, "epoch": 747, "lr": 8.605877338358746e-05}
{"train_loss": 0.18042531609535217, "global_step": 66504, "epoch": 747, "lr": 8.605837176721639e-05}
{"train_loss": 0.13866400718688965, "global_step": 66505, "epoch": 747, "lr": 8.60579701459977e-05}
{"train_loss": 0.16537226736545563, "global_step": 66506, "epoch": 747, "lr": 8.605756851993147e-05}
{"train_loss": 0.2671407163143158, "global_step": 66507, "epoch": 747, "lr": 8.605716688901775e-05}
{"train_loss": 0.23052676022052765, "global_step": 66508, "epoch": 747, "lr": 8.605676525325656e-05}
{"train_loss": 0.1901766061782837, "global_step": 66509, "epoch": 747, "lr": 8.6056363612648e-05}
{"train_loss": 0.2786771059036255, "global_step": 66510, "epoch": 747, "lr": 8.605596196719209e-05}
{"train_loss": 0.17230866849422455, "global_step": 66511, "epoch": 747, "lr": 8.605556031688889e-05}
{"train_loss": 0.2810940146446228, "global_step": 66512, "epoch": 747, "lr": 8.605515866173848e-05}
{"train_loss": 0.21988871693611145, "global_step": 66513, "epoch": 747, "lr": 8.605475700174087e-05}
{"train_loss": 0.1850111335515976, "global_step": 66514, "epoch": 747, "lr": 8.605435533689614e-05}
{"train_loss": 0.2569175958633423, "global_step": 66515, "epoch": 747, "lr": 8.605395366720435e-05}
{"train_loss": 0.2524144649505615, "global_step": 66516, "epoch": 747, "lr": 8.605355199266556e-05}
{"train_loss": 0.23298735916614532, "global_step": 66517, "epoch": 747, "lr": 8.605315031327978e-05}
{"train_loss": 0.21983930468559265, "global_step": 66518, "epoch": 747, "lr": 8.605274862904711e-05}
{"train_loss": 0.24639742076396942, "global_step": 66519, "epoch": 747, "lr": 8.605234693996758e-05}
{"train_loss": 0.30485138297080994, "global_step": 66520, "epoch": 747, "lr": 8.605194524604125e-05}
{"train_loss": 0.1471400111913681, "global_step": 66521, "epoch": 747, "lr": 8.605154354726818e-05}
{"train_loss": 0.3188193142414093, "global_step": 66522, "epoch": 747, "lr": 8.605114184364842e-05}
{"train_loss": 0.19216026365756989, "global_step": 66523, "epoch": 747, "lr": 8.605074013518203e-05}
{"train_loss": 0.24542252719402313, "global_step": 66524, "epoch": 747, "lr": 8.605033842186905e-05}
{"train_loss": 0.23819133639335632, "global_step": 66525, "epoch": 747, "lr": 8.604993670370953e-05}
{"train_loss": 0.16119171679019928, "global_step": 66526, "epoch": 747, "lr": 8.604953498070357e-05}
{"train_loss": 0.1929507851600647, "global_step": 66527, "epoch": 747, "lr": 8.604913325285115e-05}
{"train_loss": 0.1784302294254303, "global_step": 66528, "epoch": 747, "lr": 8.604873152015238e-05}
{"train_loss": 0.34518834948539734, "global_step": 66529, "epoch": 747, "lr": 8.604832978260731e-05}
{"train_loss": 0.28472739458084106, "global_step": 66530, "epoch": 747, "lr": 8.604792804021595e-05}
{"train_loss": 0.23792290687561035, "global_step": 66531, "epoch": 747, "lr": 8.604752629297841e-05}
{"train_loss": 0.22522027790546417, "global_step": 66532, "epoch": 747, "lr": 8.604712454089473e-05}
{"train_loss": 0.37310492992401123, "global_step": 66533, "epoch": 747, "lr": 8.604672278396493e-05}
{"train_loss": 0.22817392647266388, "global_step": 66534, "epoch": 747, "lr": 8.604632102218909e-05}
{"train_loss": 0.21860668063163757, "global_step": 66535, "epoch": 747, "lr": 8.604591925556726e-05}
{"train_loss": 0.17258916795253754, "global_step": 66536, "epoch": 747, "lr": 8.60455174840995e-05}
{"train_loss": 0.2212395966053009, "global_step": 66537, "epoch": 747, "lr": 8.604511570778586e-05}
{"train_loss": 0.20766007900238037, "global_step": 66538, "epoch": 747, "lr": 8.604471392662641e-05}
{"train_loss": 0.22887980937957764, "global_step": 66539, "epoch": 747, "lr": 8.604431214062116e-05}
{"train_loss": 0.1965821087360382, "global_step": 66540, "epoch": 747, "lr": 8.60439103497702e-05}
{"train_loss": 0.2869539260864258, "global_step": 66541, "epoch": 747, "lr": 8.60435085540736e-05}
{"train_loss": 0.18922428786754608, "global_step": 66542, "epoch": 747, "lr": 8.604310675353135e-05}
{"train_loss": 0.2169298231601715, "global_step": 66543, "epoch": 747, "lr": 8.604270494814357e-05}
{"train_loss": 0.16303525865077972, "global_step": 66544, "epoch": 747, "lr": 8.604230313791028e-05}
{"train_loss": 0.2664623260498047, "global_step": 66545, "epoch": 747, "lr": 8.604190132283154e-05}
{"train_loss": 0.22183746099472046, "global_step": 66546, "epoch": 747, "lr": 8.60414995029074e-05}
{"train_loss": 0.14509449899196625, "global_step": 66547, "epoch": 747, "lr": 8.604109767813793e-05}
{"train_loss": 0.2856554388999939, "global_step": 66548, "epoch": 747, "lr": 8.604069584852317e-05}
{"train_loss": 0.20221573114395142, "global_step": 66549, "epoch": 747, "lr": 8.604029401406319e-05}
{"train_loss": 0.2252221554517746, "global_step": 66550, "epoch": 747, "lr": 8.6039892174758e-05}
{"train_loss": 0.2666000723838806, "global_step": 66551, "epoch": 747, "lr": 8.603949033060772e-05}
{"train_loss": 0.2816784679889679, "global_step": 66552, "epoch": 747, "lr": 8.603908848161236e-05}
{"train_loss": 0.17264549434185028, "global_step": 66553, "epoch": 747, "lr": 8.603868662777197e-05}
{"train_loss": 0.16365428268909454, "global_step": 66554, "epoch": 747, "lr": 8.603828476908663e-05}
{"train_loss": 0.27884289622306824, "global_step": 66555, "epoch": 747, "lr": 8.60378829055564e-05}
{"train_loss": 0.1588314324617386, "global_step": 66556, "epoch": 747, "lr": 8.603748103718128e-05}
{"train_loss": 0.2529434561729431, "global_step": 66557, "epoch": 747, "lr": 8.603707916396137e-05}
{"train_loss": 0.18982639908790588, "global_step": 66558, "epoch": 747, "lr": 8.603667728589672e-05}
{"train_loss": 0.2782738506793976, "global_step": 66559, "epoch": 747, "lr": 8.603627540298738e-05}
{"train_loss": 0.23517218232154846, "global_step": 66560, "epoch": 747, "lr": 8.603587351523338e-05}
{"train_loss": 0.176473006606102, "global_step": 66561, "epoch": 747, "lr": 8.603547162263484e-05}
{"train_loss": 0.23482009768486023, "global_step": 66562, "epoch": 747, "lr": 8.603506972519172e-05}
{"train_loss": 0.18364179134368896, "global_step": 66563, "epoch": 747, "lr": 8.603466782290416e-05}
{"train_loss": 0.2269127070903778, "global_step": 66564, "epoch": 747, "lr": 8.603426591577217e-05}
{"train_loss": 0.21115821599960327, "global_step": 66565, "epoch": 747, "lr": 8.60338640037958e-05}
{"train_loss": 0.2134505957365036, "global_step": 66566, "epoch": 747, "lr": 8.603346208697514e-05}
{"train_loss": 0.1953694075345993, "global_step": 66567, "epoch": 747, "lr": 8.603306016531021e-05}
{"train_loss": 0.20848539471626282, "global_step": 66568, "epoch": 747, "lr": 8.603265823880106e-05}
{"train_loss": 0.24144938588142395, "global_step": 66569, "epoch": 747, "lr": 8.603225630744776e-05}
{"train_loss": 0.2771134376525879, "global_step": 66570, "epoch": 747, "lr": 8.603185437125039e-05}
{"train_loss": 0.22308382684929987, "global_step": 66571, "epoch": 747, "lr": 8.603145243020894e-05, "val_loss": 3.7170486450195312}
{"train_loss": 0.2792358100414276, "global_step": 66572, "epoch": 748, "lr": 8.603105048432351e-05}
{"train_loss": 0.20922863483428955, "global_step": 66573, "epoch": 748, "lr": 8.603064853359417e-05}
{"train_loss": 0.17082802951335907, "global_step": 66574, "epoch": 748, "lr": 8.603024657802093e-05}
{"train_loss": 0.26059412956237793, "global_step": 66575, "epoch": 748, "lr": 8.602984461760386e-05}
{"train_loss": 0.16609779000282288, "global_step": 66576, "epoch": 748, "lr": 8.602944265234302e-05}
{"train_loss": 0.31066828966140747, "global_step": 66577, "epoch": 748, "lr": 8.602904068223845e-05}
{"train_loss": 0.14873981475830078, "global_step": 66578, "epoch": 748, "lr": 8.602863870729023e-05}
{"train_loss": 0.24379286170005798, "global_step": 66579, "epoch": 748, "lr": 8.60282367274984e-05}
{"train_loss": 0.21019017696380615, "global_step": 66580, "epoch": 748, "lr": 8.6027834742863e-05}
{"train_loss": 0.26496535539627075, "global_step": 66581, "epoch": 748, "lr": 8.60274327533841e-05}
{"train_loss": 0.1727561354637146, "global_step": 66582, "epoch": 748, "lr": 8.602703075906176e-05}
{"train_loss": 0.21419350802898407, "global_step": 66583, "epoch": 748, "lr": 8.602662875989602e-05}
{"train_loss": 0.20874662697315216, "global_step": 66584, "epoch": 748, "lr": 8.602622675588693e-05}
{"train_loss": 0.201494038105011, "global_step": 66585, "epoch": 748, "lr": 8.602582474703456e-05}
{"train_loss": 0.23832102119922638, "global_step": 66586, "epoch": 748, "lr": 8.602542273333895e-05}
{"train_loss": 0.17804941534996033, "global_step": 66587, "epoch": 748, "lr": 8.602502071480017e-05}
{"train_loss": 0.20726004242897034, "global_step": 66588, "epoch": 748, "lr": 8.602461869141826e-05}
{"train_loss": 0.2528282701969147, "global_step": 66589, "epoch": 748, "lr": 8.602421666319329e-05}
{"train_loss": 0.257208913564682, "global_step": 66590, "epoch": 748, "lr": 8.60238146301253e-05}
{"train_loss": 0.2249053418636322, "global_step": 66591, "epoch": 748, "lr": 8.602341259221434e-05}
{"train_loss": 0.30937981605529785, "global_step": 66592, "epoch": 748, "lr": 8.602301054946047e-05}
{"train_loss": 0.2086259424686432, "global_step": 66593, "epoch": 748, "lr": 8.602260850186375e-05}
{"train_loss": 0.2177383303642273, "global_step": 66594, "epoch": 748, "lr": 8.602220644942423e-05}
{"train_loss": 0.1649661362171173, "global_step": 66595, "epoch": 748, "lr": 8.602180439214196e-05}
{"train_loss": 0.21552030742168427, "global_step": 66596, "epoch": 748, "lr": 8.6021402330017e-05}
{"train_loss": 0.2067660391330719, "global_step": 66597, "epoch": 748, "lr": 8.60210002630494e-05}
{"train_loss": 0.25619378685951233, "global_step": 66598, "epoch": 748, "lr": 8.602059819123921e-05}
{"train_loss": 0.24618777632713318, "global_step": 66599, "epoch": 748, "lr": 8.602019611458649e-05}
{"train_loss": 0.2314690202474594, "global_step": 66600, "epoch": 748, "lr": 8.60197940330913e-05}
{"train_loss": 0.18882575631141663, "global_step": 66601, "epoch": 748, "lr": 8.60193919467537e-05}
{"train_loss": 0.21433953940868378, "global_step": 66602, "epoch": 748, "lr": 8.60189898555737e-05}
{"train_loss": 0.24742108583450317, "global_step": 66603, "epoch": 748, "lr": 8.601858775955141e-05}
{"train_loss": 0.26151442527770996, "global_step": 66604, "epoch": 748, "lr": 8.601818565868687e-05}
{"train_loss": 0.1760282814502716, "global_step": 66605, "epoch": 748, "lr": 8.601778355298011e-05}
{"train_loss": 0.19728310406208038, "global_step": 66606, "epoch": 748, "lr": 8.601738144243119e-05}
{"train_loss": 0.17662152647972107, "global_step": 66607, "epoch": 748, "lr": 8.601697932704018e-05}
{"train_loss": 0.32507219910621643, "global_step": 66608, "epoch": 748, "lr": 8.601657720680712e-05}
{"train_loss": 0.21989645063877106, "global_step": 66609, "epoch": 748, "lr": 8.601617508173207e-05}
{"train_loss": 0.17974159121513367, "global_step": 66610, "epoch": 748, "lr": 8.601577295181508e-05}
{"train_loss": 0.3065597712993622, "global_step": 66611, "epoch": 748, "lr": 8.601537081705623e-05}
{"train_loss": 0.19127888977527618, "global_step": 66612, "epoch": 748, "lr": 8.601496867745553e-05}
{"train_loss": 0.2128298580646515, "global_step": 66613, "epoch": 748, "lr": 8.601456653301307e-05}
{"train_loss": 0.19616341590881348, "global_step": 66614, "epoch": 748, "lr": 8.601416438372889e-05}
{"train_loss": 0.22651199996471405, "global_step": 66615, "epoch": 748, "lr": 8.601376222960305e-05}
{"train_loss": 0.1932518631219864, "global_step": 66616, "epoch": 748, "lr": 8.601336007063558e-05}
{"train_loss": 0.12746086716651917, "global_step": 66617, "epoch": 748, "lr": 8.601295790682656e-05}
{"train_loss": 0.14004933834075928, "global_step": 66618, "epoch": 748, "lr": 8.601255573817603e-05}
{"train_loss": 0.21195478737354279, "global_step": 66619, "epoch": 748, "lr": 8.601215356468407e-05}
{"train_loss": 0.2682156264781952, "global_step": 66620, "epoch": 748, "lr": 8.601175138635071e-05}
{"train_loss": 0.17646639049053192, "global_step": 66621, "epoch": 748, "lr": 8.601134920317599e-05}
{"train_loss": 0.20211760699748993, "global_step": 66622, "epoch": 748, "lr": 8.601094701516001e-05}
{"train_loss": 0.2634255588054657, "global_step": 66623, "epoch": 748, "lr": 8.601054482230279e-05}
{"train_loss": 0.18361076712608337, "global_step": 66624, "epoch": 748, "lr": 8.601014262460439e-05}
{"train_loss": 0.1810445487499237, "global_step": 66625, "epoch": 748, "lr": 8.600974042206484e-05}
{"train_loss": 0.21784266829490662, "global_step": 66626, "epoch": 748, "lr": 8.600933821468424e-05}
{"train_loss": 0.208601713180542, "global_step": 66627, "epoch": 748, "lr": 8.600893600246262e-05}
{"train_loss": 0.25112488865852356, "global_step": 66628, "epoch": 748, "lr": 8.600853378540004e-05}
{"train_loss": 0.22081510722637177, "global_step": 66629, "epoch": 748, "lr": 8.600813156349657e-05}
{"train_loss": 0.21821613609790802, "global_step": 66630, "epoch": 748, "lr": 8.600772933675222e-05}
{"train_loss": 0.22545035183429718, "global_step": 66631, "epoch": 748, "lr": 8.600732710516708e-05}
{"train_loss": 0.1660136729478836, "global_step": 66632, "epoch": 748, "lr": 8.600692486874119e-05}
{"train_loss": 0.2250857949256897, "global_step": 66633, "epoch": 748, "lr": 8.600652262747461e-05}
{"train_loss": 0.19791942834854126, "global_step": 66634, "epoch": 748, "lr": 8.60061203813674e-05}
{"train_loss": 0.23337103426456451, "global_step": 66635, "epoch": 748, "lr": 8.600571813041958e-05}
{"train_loss": 0.18827125430107117, "global_step": 66636, "epoch": 748, "lr": 8.600531587463125e-05}
{"train_loss": 0.2616141736507416, "global_step": 66637, "epoch": 748, "lr": 8.600491361400243e-05}
{"train_loss": 0.2690860629081726, "global_step": 66638, "epoch": 748, "lr": 8.60045113485332e-05}
{"train_loss": 0.17297852039337158, "global_step": 66639, "epoch": 748, "lr": 8.600410907822361e-05}
{"train_loss": 0.260348916053772, "global_step": 66640, "epoch": 748, "lr": 8.600370680307371e-05}
{"train_loss": 0.21849621832370758, "global_step": 66641, "epoch": 748, "lr": 8.600330452308353e-05}
{"train_loss": 0.18463826179504395, "global_step": 66642, "epoch": 748, "lr": 8.600290223825315e-05}
{"train_loss": 0.2796929180622101, "global_step": 66643, "epoch": 748, "lr": 8.600249994858262e-05}
{"train_loss": 0.22365862131118774, "global_step": 66644, "epoch": 748, "lr": 8.600209765407198e-05}
{"train_loss": 0.20579256117343903, "global_step": 66645, "epoch": 748, "lr": 8.600169535472132e-05}
{"train_loss": 0.32431289553642273, "global_step": 66646, "epoch": 748, "lr": 8.600129305053067e-05}
{"train_loss": 0.20768387615680695, "global_step": 66647, "epoch": 748, "lr": 8.600089074150007e-05}
{"train_loss": 0.21362176537513733, "global_step": 66648, "epoch": 748, "lr": 8.60004884276296e-05}
{"train_loss": 0.20444954931735992, "global_step": 66649, "epoch": 748, "lr": 8.60000861089193e-05}
{"train_loss": 0.19926363229751587, "global_step": 66650, "epoch": 748, "lr": 8.599968378536923e-05}
{"train_loss": 0.2218572348356247, "global_step": 66651, "epoch": 748, "lr": 8.599928145697943e-05}
{"train_loss": 0.18570104241371155, "global_step": 66652, "epoch": 748, "lr": 8.599887912374998e-05}
{"train_loss": 0.22274117171764374, "global_step": 66653, "epoch": 748, "lr": 8.599847678568091e-05}
{"train_loss": 0.24455583095550537, "global_step": 66654, "epoch": 748, "lr": 8.599807444277229e-05}
{"train_loss": 0.27589330077171326, "global_step": 66655, "epoch": 748, "lr": 8.599767209502414e-05}
{"train_loss": 0.1520078480243683, "global_step": 66656, "epoch": 748, "lr": 8.599726974243657e-05}
{"train_loss": 0.26218754053115845, "global_step": 66657, "epoch": 748, "lr": 8.599686738500961e-05}
{"train_loss": 0.22233085334300995, "global_step": 66658, "epoch": 748, "lr": 8.59964650227433e-05}
{"train_loss": 0.23519788682460785, "global_step": 66659, "epoch": 748, "lr": 8.59960626556377e-05}
{"train_loss": 0.22127758016747034, "global_step": 66660, "epoch": 748, "lr": 8.599566028369288e-05, "val_loss": 3.667858362197876}
{"train_loss": 0.24150384962558746, "global_step": 66661, "epoch": 749, "lr": 8.599525790690886e-05}
{"train_loss": 0.2135312408208847, "global_step": 66662, "epoch": 749, "lr": 8.599485552528575e-05}
{"train_loss": 0.19804491102695465, "global_step": 66663, "epoch": 749, "lr": 8.599445313882353e-05}
{"train_loss": 0.2244933843612671, "global_step": 66664, "epoch": 749, "lr": 8.599405074752233e-05}
{"train_loss": 0.16853326559066772, "global_step": 66665, "epoch": 749, "lr": 8.599364835138215e-05}
{"train_loss": 0.20348863303661346, "global_step": 66666, "epoch": 749, "lr": 8.599324595040307e-05}
{"train_loss": 0.25425490736961365, "global_step": 66667, "epoch": 749, "lr": 8.599284354458512e-05}
{"train_loss": 0.2987872362136841, "global_step": 66668, "epoch": 749, "lr": 8.599244113392839e-05}
{"train_loss": 0.2901938259601593, "global_step": 66669, "epoch": 749, "lr": 8.59920387184329e-05}
{"train_loss": 0.22327271103858948, "global_step": 66670, "epoch": 749, "lr": 8.599163629809875e-05}
{"train_loss": 0.29478123784065247, "global_step": 66671, "epoch": 749, "lr": 8.599123387292593e-05}
{"train_loss": 0.2734037935733795, "global_step": 66672, "epoch": 749, "lr": 8.599083144291454e-05}
{"train_loss": 0.25143963098526, "global_step": 66673, "epoch": 749, "lr": 8.599042900806462e-05}
{"train_loss": 0.23592066764831543, "global_step": 66674, "epoch": 749, "lr": 8.599002656837622e-05}
{"train_loss": 0.20246420800685883, "global_step": 66675, "epoch": 749, "lr": 8.598962412384941e-05}
{"train_loss": 0.21807317435741425, "global_step": 66676, "epoch": 749, "lr": 8.598922167448425e-05}
{"train_loss": 0.227998748421669, "global_step": 66677, "epoch": 749, "lr": 8.598881922028074e-05}
{"train_loss": 0.20496194064617157, "global_step": 66678, "epoch": 749, "lr": 8.5988416761239e-05}
{"train_loss": 0.24564619362354279, "global_step": 66679, "epoch": 749, "lr": 8.598801429735904e-05}
{"train_loss": 0.21078872680664062, "global_step": 66680, "epoch": 749, "lr": 8.598761182864095e-05}
{"train_loss": 0.19619177281856537, "global_step": 66681, "epoch": 749, "lr": 8.598720935508473e-05}
{"train_loss": 0.17687077820301056, "global_step": 66682, "epoch": 749, "lr": 8.598680687669049e-05}
{"train_loss": 0.19508448243141174, "global_step": 66683, "epoch": 749, "lr": 8.598640439345827e-05}
{"train_loss": 0.23850637674331665, "global_step": 66684, "epoch": 749, "lr": 8.59860019053881e-05}
{"train_loss": 0.21099187433719635, "global_step": 66685, "epoch": 749, "lr": 8.598559941248006e-05}
{"train_loss": 0.30503711104393005, "global_step": 66686, "epoch": 749, "lr": 8.598519691473419e-05}
{"train_loss": 0.1913987547159195, "global_step": 66687, "epoch": 749, "lr": 8.598479441215055e-05}
{"train_loss": 0.11412230134010315, "global_step": 66688, "epoch": 749, "lr": 8.59843919047292e-05}
{"train_loss": 0.2495669424533844, "global_step": 66689, "epoch": 749, "lr": 8.598398939247018e-05}
{"train_loss": 0.29121536016464233, "global_step": 66690, "epoch": 749, "lr": 8.598358687537354e-05}
{"train_loss": 0.22342264652252197, "global_step": 66691, "epoch": 749, "lr": 8.598318435343937e-05}
{"train_loss": 0.2291734367609024, "global_step": 66692, "epoch": 749, "lr": 8.598278182666767e-05}
{"train_loss": 0.17375129461288452, "global_step": 66693, "epoch": 749, "lr": 8.598237929505854e-05}
{"train_loss": 0.20278123021125793, "global_step": 66694, "epoch": 749, "lr": 8.598197675861204e-05}
{"train_loss": 0.14775292575359344, "global_step": 66695, "epoch": 749, "lr": 8.598157421732818e-05}
{"train_loss": 0.26311540603637695, "global_step": 66696, "epoch": 749, "lr": 8.598117167120702e-05}
{"train_loss": 0.2850966453552246, "global_step": 66697, "epoch": 749, "lr": 8.598076912024866e-05}
{"train_loss": 0.2977294623851776, "global_step": 66698, "epoch": 749, "lr": 8.598036656445311e-05}
{"train_loss": 0.22962932288646698, "global_step": 66699, "epoch": 749, "lr": 8.597996400382044e-05}
{"train_loss": 0.21483170986175537, "global_step": 66700, "epoch": 749, "lr": 8.597956143835068e-05}
{"train_loss": 0.259743869304657, "global_step": 66701, "epoch": 749, "lr": 8.597915886804394e-05}
{"train_loss": 0.19356706738471985, "global_step": 66702, "epoch": 749, "lr": 8.597875629290024e-05}
{"train_loss": 0.15327756106853485, "global_step": 66703, "epoch": 749, "lr": 8.597835371291961e-05}
{"train_loss": 0.16882599890232086, "global_step": 66704, "epoch": 749, "lr": 8.597795112810215e-05}
{"train_loss": 0.2508343756198883, "global_step": 66705, "epoch": 749, "lr": 8.597754853844788e-05}
{"train_loss": 0.25401031970977783, "global_step": 66706, "epoch": 749, "lr": 8.597714594395688e-05}
{"train_loss": 0.17546512186527252, "global_step": 66707, "epoch": 749, "lr": 8.597674334462917e-05}
{"train_loss": 0.2049662172794342, "global_step": 66708, "epoch": 749, "lr": 8.597634074046483e-05}
{"train_loss": 0.16082540154457092, "global_step": 66709, "epoch": 749, "lr": 8.597593813146393e-05}
{"train_loss": 0.2756763994693756, "global_step": 66710, "epoch": 749, "lr": 8.597553551762649e-05}
{"train_loss": 0.28468945622444153, "global_step": 66711, "epoch": 749, "lr": 8.597513289895259e-05}
{"train_loss": 0.26441559195518494, "global_step": 66712, "epoch": 749, "lr": 8.597473027544225e-05}
{"train_loss": 0.2582719922065735, "global_step": 66713, "epoch": 749, "lr": 8.597432764709556e-05}
{"train_loss": 0.18364766240119934, "global_step": 66714, "epoch": 749, "lr": 8.597392501391255e-05}
{"train_loss": 0.15459881722927094, "global_step": 66715, "epoch": 749, "lr": 8.597352237589329e-05}
{"train_loss": 0.23419028520584106, "global_step": 66716, "epoch": 749, "lr": 8.597311973303783e-05}
{"train_loss": 0.26471641659736633, "global_step": 66717, "epoch": 749, "lr": 8.597271708534623e-05}
{"train_loss": 0.14613546431064606, "global_step": 66718, "epoch": 749, "lr": 8.597231443281852e-05}
{"train_loss": 0.21056024730205536, "global_step": 66719, "epoch": 749, "lr": 8.597191177545479e-05}
{"train_loss": 0.23762263357639313, "global_step": 66720, "epoch": 749, "lr": 8.597150911325506e-05}
{"train_loss": 0.27711722254753113, "global_step": 66721, "epoch": 749, "lr": 8.59711064462194e-05}
{"train_loss": 0.26146095991134644, "global_step": 66722, "epoch": 749, "lr": 8.597070377434787e-05}
{"train_loss": 0.2455780953168869, "global_step": 66723, "epoch": 749, "lr": 8.597030109764052e-05}
{"train_loss": 0.18510225415229797, "global_step": 66724, "epoch": 749, "lr": 8.596989841609739e-05}
{"train_loss": 0.22006307542324066, "global_step": 66725, "epoch": 749, "lr": 8.596949572971855e-05}
{"train_loss": 0.2831580340862274, "global_step": 66726, "epoch": 749, "lr": 8.596909303850406e-05}
{"train_loss": 0.2602306604385376, "global_step": 66727, "epoch": 749, "lr": 8.596869034245394e-05}
{"train_loss": 0.18267159163951874, "global_step": 66728, "epoch": 749, "lr": 8.59682876415683e-05}
{"train_loss": 0.18993999063968658, "global_step": 66729, "epoch": 749, "lr": 8.596788493584713e-05}
{"train_loss": 0.2762947976589203, "global_step": 66730, "epoch": 749, "lr": 8.596748222529054e-05}
{"train_loss": 0.2595451772212982, "global_step": 66731, "epoch": 749, "lr": 8.596707950989856e-05}
{"train_loss": 0.1691783368587494, "global_step": 66732, "epoch": 749, "lr": 8.596667678967122e-05}
{"train_loss": 0.1847655177116394, "global_step": 66733, "epoch": 749, "lr": 8.596627406460862e-05}
{"train_loss": 0.1920916736125946, "global_step": 66734, "epoch": 749, "lr": 8.596587133471079e-05}
{"train_loss": 0.2014552354812622, "global_step": 66735, "epoch": 749, "lr": 8.596546859997778e-05}
{"train_loss": 0.2060147523880005, "global_step": 66736, "epoch": 749, "lr": 8.596506586040965e-05}
{"train_loss": 0.2910252809524536, "global_step": 66737, "epoch": 749, "lr": 8.596466311600646e-05}
{"train_loss": 0.2064334899187088, "global_step": 66738, "epoch": 749, "lr": 8.596426036676826e-05}
{"train_loss": 0.19687283039093018, "global_step": 66739, "epoch": 749, "lr": 8.596385761269511e-05}
{"train_loss": 0.24418869614601135, "global_step": 66740, "epoch": 749, "lr": 8.596345485378704e-05}
{"train_loss": 0.26414015889167786, "global_step": 66741, "epoch": 749, "lr": 8.596305209004414e-05}
{"train_loss": 0.2037280797958374, "global_step": 66742, "epoch": 749, "lr": 8.596264932146644e-05}
{"train_loss": 0.2813464403152466, "global_step": 66743, "epoch": 749, "lr": 8.5962246548054e-05}
{"train_loss": 0.15992887318134308, "global_step": 66744, "epoch": 749, "lr": 8.596184376980687e-05}
{"train_loss": 0.18805834650993347, "global_step": 66745, "epoch": 749, "lr": 8.596144098672512e-05}
{"train_loss": 0.19681115448474884, "global_step": 66746, "epoch": 749, "lr": 8.596103819880878e-05}
{"train_loss": 0.10317794233560562, "global_step": 66747, "epoch": 749, "lr": 8.596063540605793e-05}
{"train_loss": 0.24368442595005035, "global_step": 66748, "epoch": 749, "lr": 8.596023260847259e-05}
{"train_loss": 0.2229417377475942, "global_step": 66749, "epoch": 749, "lr": 8.595982980605285e-05, "val_loss": 3.6001458168029785}
{"train_loss": 0.2654570937156677, "global_step": 66750, "epoch": 750, "lr": 8.595942699879874e-05}
{"train_loss": 0.29012030363082886, "global_step": 66751, "epoch": 750, "lr": 8.595902418671034e-05}
{"train_loss": 0.1660122573375702, "global_step": 66752, "epoch": 750, "lr": 8.595862136978767e-05}
{"train_loss": 0.1895955502986908, "global_step": 66753, "epoch": 750, "lr": 8.595821854803081e-05}
{"train_loss": 0.13474617898464203, "global_step": 66754, "epoch": 750, "lr": 8.595781572143981e-05}
{"train_loss": 0.1862817257642746, "global_step": 66755, "epoch": 750, "lr": 8.595741289001472e-05}
{"train_loss": 0.19014336168766022, "global_step": 66756, "epoch": 750, "lr": 8.595701005375559e-05}
{"train_loss": 0.19520144164562225, "global_step": 66757, "epoch": 750, "lr": 8.595660721266246e-05}
{"train_loss": 0.18014352023601532, "global_step": 66758, "epoch": 750, "lr": 8.595620436673541e-05}
{"train_loss": 0.20805400609970093, "global_step": 66759, "epoch": 750, "lr": 8.595580151597451e-05}
{"train_loss": 0.191123828291893, "global_step": 66760, "epoch": 750, "lr": 8.595539866037979e-05}
{"train_loss": 0.18083494901657104, "global_step": 66761, "epoch": 750, "lr": 8.595499579995127e-05}
{"train_loss": 0.21944737434387207, "global_step": 66762, "epoch": 750, "lr": 8.595459293468906e-05}
{"train_loss": 0.23579394817352295, "global_step": 66763, "epoch": 750, "lr": 8.59541900645932e-05}
{"train_loss": 0.21678832173347473, "global_step": 66764, "epoch": 750, "lr": 8.595378718966374e-05}
{"train_loss": 0.23024603724479675, "global_step": 66765, "epoch": 750, "lr": 8.595338430990072e-05}
{"train_loss": 0.23069298267364502, "global_step": 66766, "epoch": 750, "lr": 8.59529814253042e-05}
{"train_loss": 0.27154937386512756, "global_step": 66767, "epoch": 750, "lr": 8.595257853587426e-05}
{"train_loss": 0.29179251194000244, "global_step": 66768, "epoch": 750, "lr": 8.595217564161091e-05}
{"train_loss": 0.19233153760433197, "global_step": 66769, "epoch": 750, "lr": 8.595177274251424e-05}
{"train_loss": 0.3227984607219696, "global_step": 66770, "epoch": 750, "lr": 8.59513698385843e-05}
{"train_loss": 0.2524617910385132, "global_step": 66771, "epoch": 750, "lr": 8.595096692982113e-05}
{"train_loss": 0.14905422925949097, "global_step": 66772, "epoch": 750, "lr": 8.595056401622479e-05}
{"train_loss": 0.20059432089328766, "global_step": 66773, "epoch": 750, "lr": 8.595016109779533e-05}
{"train_loss": 0.28478339314460754, "global_step": 66774, "epoch": 750, "lr": 8.594975817453281e-05}
{"train_loss": 0.2392614632844925, "global_step": 66775, "epoch": 750, "lr": 8.594935524643727e-05}
{"train_loss": 0.20948240160942078, "global_step": 66776, "epoch": 750, "lr": 8.594895231350881e-05}
{"train_loss": 0.18603336811065674, "global_step": 66777, "epoch": 750, "lr": 8.594854937574743e-05}
{"train_loss": 0.20107245445251465, "global_step": 66778, "epoch": 750, "lr": 8.594814643315323e-05}
{"train_loss": 0.233725443482399, "global_step": 66779, "epoch": 750, "lr": 8.59477434857262e-05}
{"train_loss": 0.24752408266067505, "global_step": 66780, "epoch": 750, "lr": 8.594734053346647e-05}
{"train_loss": 0.19156503677368164, "global_step": 66781, "epoch": 750, "lr": 8.594693757637403e-05}
{"train_loss": 0.20958609879016876, "global_step": 66782, "epoch": 750, "lr": 8.594653461444898e-05}
{"train_loss": 0.23627236485481262, "global_step": 66783, "epoch": 750, "lr": 8.594613164769136e-05}
{"train_loss": 0.19287501275539398, "global_step": 66784, "epoch": 750, "lr": 8.59457286761012e-05}
{"train_loss": 0.2250523567199707, "global_step": 66785, "epoch": 750, "lr": 8.594532569967859e-05}
{"train_loss": 0.22722874581813812, "global_step": 66786, "epoch": 750, "lr": 8.594492271842358e-05}
{"train_loss": 0.1807258278131485, "global_step": 66787, "epoch": 750, "lr": 8.594451973233619e-05}
{"train_loss": 0.13643881678581238, "global_step": 66788, "epoch": 750, "lr": 8.594411674141652e-05}
{"train_loss": 0.26245996356010437, "global_step": 66789, "epoch": 750, "lr": 8.594371374566458e-05}
{"train_loss": 0.24183149635791779, "global_step": 66790, "epoch": 750, "lr": 8.594331074508045e-05}
{"train_loss": 0.20069262385368347, "global_step": 66791, "epoch": 750, "lr": 8.59429077396642e-05}
{"train_loss": 0.16933731734752655, "global_step": 66792, "epoch": 750, "lr": 8.594250472941584e-05}
{"train_loss": 0.1851486712694168, "global_step": 66793, "epoch": 750, "lr": 8.594210171433545e-05}
{"train_loss": 0.24014469981193542, "global_step": 66794, "epoch": 750, "lr": 8.59416986944231e-05}
{"train_loss": 0.19014281034469604, "global_step": 66795, "epoch": 750, "lr": 8.594129566967881e-05}
{"train_loss": 0.15192918479442596, "global_step": 66796, "epoch": 750, "lr": 8.594089264010266e-05}
{"train_loss": 0.16519129276275635, "global_step": 66797, "epoch": 750, "lr": 8.59404896056947e-05}
{"train_loss": 0.17567066848278046, "global_step": 66798, "epoch": 750, "lr": 8.594008656645495e-05}
{"train_loss": 0.29720211029052734, "global_step": 66799, "epoch": 750, "lr": 8.593968352238353e-05}
{"train_loss": 0.35350459814071655, "global_step": 66800, "epoch": 750, "lr": 8.593928047348044e-05}
{"train_loss": 0.21853472292423248, "global_step": 66801, "epoch": 750, "lr": 8.593887741974575e-05}
{"train_loss": 0.15553811192512512, "global_step": 66802, "epoch": 750, "lr": 8.593847436117952e-05}
{"train_loss": 0.1401214748620987, "global_step": 66803, "epoch": 750, "lr": 8.593807129778181e-05}
{"train_loss": 0.3309905529022217, "global_step": 66804, "epoch": 750, "lr": 8.593766822955265e-05}
{"train_loss": 0.1890837699174881, "global_step": 66805, "epoch": 750, "lr": 8.593726515649211e-05}
{"train_loss": 0.168862983584404, "global_step": 66806, "epoch": 750, "lr": 8.593686207860025e-05}
{"train_loss": 0.21746715903282166, "global_step": 66807, "epoch": 750, "lr": 8.593645899587713e-05}
{"train_loss": 0.22749461233615875, "global_step": 66808, "epoch": 750, "lr": 8.593605590832276e-05}
{"train_loss": 0.2612784504890442, "global_step": 66809, "epoch": 750, "lr": 8.593565281593722e-05}
{"train_loss": 0.2019757628440857, "global_step": 66810, "epoch": 750, "lr": 8.59352497187206e-05}
{"train_loss": 0.28215330839157104, "global_step": 66811, "epoch": 750, "lr": 8.593484661667292e-05}
{"train_loss": 0.264402836561203, "global_step": 66812, "epoch": 750, "lr": 8.593444350979423e-05}
{"train_loss": 0.23233316838741302, "global_step": 66813, "epoch": 750, "lr": 8.59340403980846e-05}
{"train_loss": 0.17460261285305023, "global_step": 66814, "epoch": 750, "lr": 8.593363728154407e-05}
{"train_loss": 0.1924513727426529, "global_step": 66815, "epoch": 750, "lr": 8.593323416017269e-05}
{"train_loss": 0.2910615801811218, "global_step": 66816, "epoch": 750, "lr": 8.593283103397054e-05}
{"train_loss": 0.2749212086200714, "global_step": 66817, "epoch": 750, "lr": 8.593242790293766e-05}
{"train_loss": 0.2612075209617615, "global_step": 66818, "epoch": 750, "lr": 8.593202476707408e-05}
{"train_loss": 0.20219019055366516, "global_step": 66819, "epoch": 750, "lr": 8.593162162637989e-05}
{"train_loss": 0.21563591063022614, "global_step": 66820, "epoch": 750, "lr": 8.593121848085515e-05}
{"train_loss": 0.22718313336372375, "global_step": 66821, "epoch": 750, "lr": 8.593081533049986e-05}
{"train_loss": 0.19751518964767456, "global_step": 66822, "epoch": 750, "lr": 8.593041217531414e-05}
{"train_loss": 0.1541653722524643, "global_step": 66823, "epoch": 750, "lr": 8.5930009015298e-05}
{"train_loss": 0.12246189266443253, "global_step": 66824, "epoch": 750, "lr": 8.592960585045151e-05}
{"train_loss": 0.22782056033611298, "global_step": 66825, "epoch": 750, "lr": 8.592920268077473e-05}
{"train_loss": 0.24123218655586243, "global_step": 66826, "epoch": 750, "lr": 8.59287995062677e-05}
{"train_loss": 0.2638014256954193, "global_step": 66827, "epoch": 750, "lr": 8.592839632693048e-05}
{"train_loss": 0.30711475014686584, "global_step": 66828, "epoch": 750, "lr": 8.592799314276313e-05}
{"train_loss": 0.21992525458335876, "global_step": 66829, "epoch": 750, "lr": 8.59275899537657e-05}
{"train_loss": 0.28421586751937866, "global_step": 66830, "epoch": 750, "lr": 8.592718675993824e-05}
{"train_loss": 0.2048916220664978, "global_step": 66831, "epoch": 750, "lr": 8.59267835612808e-05}
{"train_loss": 0.1893724799156189, "global_step": 66832, "epoch": 750, "lr": 8.592638035779346e-05}
{"train_loss": 0.25834089517593384, "global_step": 66833, "epoch": 750, "lr": 8.592597714947623e-05}
{"train_loss": 0.2650225758552551, "global_step": 66834, "epoch": 750, "lr": 8.592557393632921e-05}
{"train_loss": 0.2511536180973053, "global_step": 66835, "epoch": 750, "lr": 8.592517071835243e-05}
{"train_loss": 0.19596388936042786, "global_step": 66836, "epoch": 750, "lr": 8.592476749554594e-05}
{"train_loss": 0.25675657391548157, "global_step": 66837, "epoch": 750, "lr": 8.592436426790982e-05}
{"train_loss": 0.22052733675482566, "global_step": 66838, "epoch": 750, "lr": 8.59239610354441e-05, "train/sim_max_reward_0": 0.20505817527326042, "train/sim_max_reward_1": 0.3750657153603619, "train/sim_max_reward_2": 0.1585066426252127, "train/sim_max_reward_3": 0.4868063909365216, "train/sim_max_reward_4": 0.9733408433458819, "train/sim_max_reward_5": 0.3015598726204188, "test/sim_max_reward_4500000": 0.4481222555603562, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9707970410315091, "test/sim_max_reward_4500003": 0.5523269931558948, "test/sim_max_reward_4500004": 1.0, "test/sim_max_reward_4500005": 0.9943623910088102, "test/sim_max_reward_4500006": 0.9747691934190181, "test/sim_max_reward_4500007": 0.47750527071537646, "test/sim_max_reward_4500008": 0.4750988855467416, "test/sim_max_reward_4500009": 0.04760180391912235, "test/sim_max_reward_4500010": 1.0, "test/sim_max_reward_4500011": 0.9712928267238446, "test/sim_max_reward_4500012": 0.9358466521608377, "test/sim_max_reward_4500013": 0.43388809707573417, "test/sim_max_reward_4500014": 0.9875681383748456, "test/sim_max_reward_4500015": 0.9994896334423963, "test/sim_max_reward_4500016": 0.5049024534847042, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 0.9574347839844326, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.8920679440821776, "test/sim_max_reward_4500022": 0.9963380159192345, "test/sim_max_reward_4500023": 0.659803069142099, "test/sim_max_reward_4500024": 0.5565540151758536, "test/sim_max_reward_4500025": 0.0, "test/sim_max_reward_4500026": 0.39797760020430384, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.5197698681263595, "test/sim_max_reward_4500029": 0.04649655337544686, "test/sim_max_reward_4500030": 0.9951169461677719, "test/sim_max_reward_4500031": 0.8376575201513692, "test/sim_max_reward_4500032": 0.44390153362202467, "test/sim_max_reward_4500033": 0.6196026717515573, "test/sim_max_reward_4500034": 0.9295964634387173, "test/sim_max_reward_4500035": 0.4994776694875188, "test/sim_max_reward_4500036": 0.989615857874309, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.16233278745004792, "test/sim_max_reward_4500039": 0.9792623891878138, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.9769538665803977, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9988602479784219, "test/sim_max_reward_4500044": 0.7846080539793835, "test/sim_max_reward_4500045": 0.592726932553022, "test/sim_max_reward_4500046": 0.35671028997227533, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9491720845892274, "test/sim_max_reward_4500049": 0.9769206876992682, "train/mean_score": 0.4167229400269428, "test/mean_score": 0.6778505897622444, "val_loss": 3.581033229827881, "train_action_mse_error": 24.210124969482422}
{"train_loss": 0.2365553230047226, "global_step": 66839, "epoch": 751, "lr": 8.592355779814884e-05}
{"train_loss": 0.17196202278137207, "global_step": 66840, "epoch": 751, "lr": 8.59231545560241e-05}
{"train_loss": 0.1741543859243393, "global_step": 66841, "epoch": 751, "lr": 8.592275130906992e-05}
{"train_loss": 0.26545071601867676, "global_step": 66842, "epoch": 751, "lr": 8.592234805728637e-05}
{"train_loss": 0.22082053124904633, "global_step": 66843, "epoch": 751, "lr": 8.592194480067351e-05}
{"train_loss": 0.25184065103530884, "global_step": 66844, "epoch": 751, "lr": 8.592154153923135e-05}
{"train_loss": 0.26829037070274353, "global_step": 66845, "epoch": 751, "lr": 8.592113827296e-05}
{"train_loss": 0.22542566061019897, "global_step": 66846, "epoch": 751, "lr": 8.592073500185948e-05}
{"train_loss": 0.2075071781873703, "global_step": 66847, "epoch": 751, "lr": 8.592033172592986e-05}
{"train_loss": 0.24534958600997925, "global_step": 66848, "epoch": 751, "lr": 8.591992844517121e-05}
{"train_loss": 0.23458966612815857, "global_step": 66849, "epoch": 751, "lr": 8.591952515958353e-05}
{"train_loss": 0.29300326108932495, "global_step": 66850, "epoch": 751, "lr": 8.591912186916693e-05}
{"train_loss": 0.23573142290115356, "global_step": 66851, "epoch": 751, "lr": 8.591871857392143e-05}
{"train_loss": 0.2556664049625397, "global_step": 66852, "epoch": 751, "lr": 8.591831527384711e-05}
{"train_loss": 0.3312833607196808, "global_step": 66853, "epoch": 751, "lr": 8.591791196894399e-05}
{"train_loss": 0.23566921055316925, "global_step": 66854, "epoch": 751, "lr": 8.591750865921216e-05}
{"train_loss": 0.24981361627578735, "global_step": 66855, "epoch": 751, "lr": 8.591710534465165e-05}
{"train_loss": 0.18593499064445496, "global_step": 66856, "epoch": 751, "lr": 8.591670202526252e-05}
{"train_loss": 0.23460397124290466, "global_step": 66857, "epoch": 751, "lr": 8.591629870104482e-05}
{"train_loss": 0.2088659256696701, "global_step": 66858, "epoch": 751, "lr": 8.591589537199865e-05}
{"train_loss": 0.2591659724712372, "global_step": 66859, "epoch": 751, "lr": 8.5915492038124e-05}
{"train_loss": 0.18190014362335205, "global_step": 66860, "epoch": 751, "lr": 8.591508869942094e-05}
{"train_loss": 0.20992785692214966, "global_step": 66861, "epoch": 751, "lr": 8.591468535588956e-05}
{"train_loss": 0.3037760853767395, "global_step": 66862, "epoch": 751, "lr": 8.591428200752987e-05}
{"train_loss": 0.1802048236131668, "global_step": 66863, "epoch": 751, "lr": 8.591387865434193e-05}
{"train_loss": 0.2297217845916748, "global_step": 66864, "epoch": 751, "lr": 8.591347529632583e-05}
{"train_loss": 0.23724791407585144, "global_step": 66865, "epoch": 751, "lr": 8.59130719334816e-05}
{"train_loss": 0.21859732270240784, "global_step": 66866, "epoch": 751, "lr": 8.591266856580927e-05}
{"train_loss": 0.25377029180526733, "global_step": 66867, "epoch": 751, "lr": 8.591226519330895e-05}
{"train_loss": 0.25460222363471985, "global_step": 66868, "epoch": 751, "lr": 8.591186181598064e-05}
{"train_loss": 0.1961515098810196, "global_step": 66869, "epoch": 751, "lr": 8.591145843382444e-05}
{"train_loss": 0.21902869641780853, "global_step": 66870, "epoch": 751, "lr": 8.591105504684036e-05}
{"train_loss": 0.1828441470861435, "global_step": 66871, "epoch": 751, "lr": 8.591065165502849e-05}
{"train_loss": 0.18056446313858032, "global_step": 66872, "epoch": 751, "lr": 8.591024825838887e-05}
{"train_loss": 0.3109920918941498, "global_step": 66873, "epoch": 751, "lr": 8.590984485692154e-05}
{"train_loss": 0.15689241886138916, "global_step": 66874, "epoch": 751, "lr": 8.590944145062659e-05}
{"train_loss": 0.16531750559806824, "global_step": 66875, "epoch": 751, "lr": 8.590903803950405e-05}
{"train_loss": 0.21152672171592712, "global_step": 66876, "epoch": 751, "lr": 8.590863462355398e-05}
{"train_loss": 0.156039759516716, "global_step": 66877, "epoch": 751, "lr": 8.590823120277642e-05}
{"train_loss": 0.25995537638664246, "global_step": 66878, "epoch": 751, "lr": 8.590782777717144e-05}
{"train_loss": 0.11585621535778046, "global_step": 66879, "epoch": 751, "lr": 8.590742434673908e-05}
{"train_loss": 0.18382865190505981, "global_step": 66880, "epoch": 751, "lr": 8.590702091147943e-05}
{"train_loss": 0.20841242372989655, "global_step": 66881, "epoch": 751, "lr": 8.59066174713925e-05}
{"train_loss": 0.236399307847023, "global_step": 66882, "epoch": 751, "lr": 8.590621402647836e-05}
{"train_loss": 0.1931304782629013, "global_step": 66883, "epoch": 751, "lr": 8.590581057673708e-05}
{"train_loss": 0.17640405893325806, "global_step": 66884, "epoch": 751, "lr": 8.59054071221687e-05}
{"train_loss": 0.19158048927783966, "global_step": 66885, "epoch": 751, "lr": 8.590500366277326e-05}
{"train_loss": 0.27648282051086426, "global_step": 66886, "epoch": 751, "lr": 8.590460019855085e-05}
{"train_loss": 0.2931019365787506, "global_step": 66887, "epoch": 751, "lr": 8.59041967295015e-05}
{"train_loss": 0.26369354128837585, "global_step": 66888, "epoch": 751, "lr": 8.590379325562526e-05}
{"train_loss": 0.16067776083946228, "global_step": 66889, "epoch": 751, "lr": 8.59033897769222e-05}
{"train_loss": 0.1776672750711441, "global_step": 66890, "epoch": 751, "lr": 8.590298629339235e-05}
{"train_loss": 0.24598993360996246, "global_step": 66891, "epoch": 751, "lr": 8.59025828050358e-05}
{"train_loss": 0.2219800353050232, "global_step": 66892, "epoch": 751, "lr": 8.590217931185258e-05}
{"train_loss": 0.19806146621704102, "global_step": 66893, "epoch": 751, "lr": 8.590177581384275e-05}
{"train_loss": 0.21441498398780823, "global_step": 66894, "epoch": 751, "lr": 8.590137231100636e-05}
{"train_loss": 0.21418821811676025, "global_step": 66895, "epoch": 751, "lr": 8.590096880334346e-05}
{"train_loss": 0.22960209846496582, "global_step": 66896, "epoch": 751, "lr": 8.590056529085413e-05}
{"train_loss": 0.2257826179265976, "global_step": 66897, "epoch": 751, "lr": 8.59001617735384e-05}
{"train_loss": 0.1397966742515564, "global_step": 66898, "epoch": 751, "lr": 8.589975825139634e-05}
{"train_loss": 0.15911683440208435, "global_step": 66899, "epoch": 751, "lr": 8.589935472442798e-05}
{"train_loss": 0.2054648995399475, "global_step": 66900, "epoch": 751, "lr": 8.589895119263338e-05}
{"train_loss": 0.21461644768714905, "global_step": 66901, "epoch": 751, "lr": 8.589854765601263e-05}
{"train_loss": 0.15208613872528076, "global_step": 66902, "epoch": 751, "lr": 8.589814411456574e-05}
{"train_loss": 0.30881011486053467, "global_step": 66903, "epoch": 751, "lr": 8.589774056829279e-05}
{"train_loss": 0.23822885751724243, "global_step": 66904, "epoch": 751, "lr": 8.589733701719382e-05}
{"train_loss": 0.23675668239593506, "global_step": 66905, "epoch": 751, "lr": 8.58969334612689e-05}
{"train_loss": 0.1713801920413971, "global_step": 66906, "epoch": 751, "lr": 8.589652990051807e-05}
{"train_loss": 0.18090172111988068, "global_step": 66907, "epoch": 751, "lr": 8.58961263349414e-05}
{"train_loss": 0.14924506843090057, "global_step": 66908, "epoch": 751, "lr": 8.589572276453891e-05}
{"train_loss": 0.264344722032547, "global_step": 66909, "epoch": 751, "lr": 8.589531918931069e-05}
{"train_loss": 0.13007639348506927, "global_step": 66910, "epoch": 751, "lr": 8.589491560925678e-05}
{"train_loss": 0.16465647518634796, "global_step": 66911, "epoch": 751, "lr": 8.589451202437724e-05}
{"train_loss": 0.17894762754440308, "global_step": 66912, "epoch": 751, "lr": 8.589410843467212e-05}
{"train_loss": 0.21978510916233063, "global_step": 66913, "epoch": 751, "lr": 8.589370484014148e-05}
{"train_loss": 0.17454010248184204, "global_step": 66914, "epoch": 751, "lr": 8.589330124078535e-05}
{"train_loss": 0.17077288031578064, "global_step": 66915, "epoch": 751, "lr": 8.589289763660382e-05}
{"train_loss": 0.24611032009124756, "global_step": 66916, "epoch": 751, "lr": 8.589249402759692e-05}
{"train_loss": 0.22857144474983215, "global_step": 66917, "epoch": 751, "lr": 8.589209041376472e-05}
{"train_loss": 0.14388491213321686, "global_step": 66918, "epoch": 751, "lr": 8.589168679510726e-05}
{"train_loss": 0.23804166913032532, "global_step": 66919, "epoch": 751, "lr": 8.58912831716246e-05}
{"train_loss": 0.187526673078537, "global_step": 66920, "epoch": 751, "lr": 8.589087954331678e-05}
{"train_loss": 0.2317250370979309, "global_step": 66921, "epoch": 751, "lr": 8.58904759101839e-05}
{"train_loss": 0.28574633598327637, "global_step": 66922, "epoch": 751, "lr": 8.589007227222596e-05}
{"train_loss": 0.2960605323314667, "global_step": 66923, "epoch": 751, "lr": 8.588966862944304e-05}
{"train_loss": 0.16766493022441864, "global_step": 66924, "epoch": 751, "lr": 8.58892649818352e-05}
{"train_loss": 0.3087944984436035, "global_step": 66925, "epoch": 751, "lr": 8.588886132940248e-05}
{"train_loss": 0.35685300827026367, "global_step": 66926, "epoch": 751, "lr": 8.588845767214493e-05}
{"train_loss": 0.21980598535430565, "global_step": 66927, "epoch": 751, "lr": 8.588805401006263e-05, "val_loss": 3.579744338989258}
{"train_loss": 0.29789143800735474, "global_step": 66928, "epoch": 752, "lr": 8.588765034315562e-05}
{"train_loss": 0.2260885238647461, "global_step": 66929, "epoch": 752, "lr": 8.588724667142394e-05}
{"train_loss": 0.31693345308303833, "global_step": 66930, "epoch": 752, "lr": 8.588684299486766e-05}
{"train_loss": 0.26102691888809204, "global_step": 66931, "epoch": 752, "lr": 8.588643931348683e-05}
{"train_loss": 0.2610231637954712, "global_step": 66932, "epoch": 752, "lr": 8.588603562728152e-05}
{"train_loss": 0.2781800925731659, "global_step": 66933, "epoch": 752, "lr": 8.588563193625175e-05}
{"train_loss": 0.28902941942214966, "global_step": 66934, "epoch": 752, "lr": 8.58852282403976e-05}
{"train_loss": 0.1950400471687317, "global_step": 66935, "epoch": 752, "lr": 8.588482453971913e-05}
{"train_loss": 0.16170616447925568, "global_step": 66936, "epoch": 752, "lr": 8.588442083421637e-05}
{"train_loss": 0.17646795511245728, "global_step": 66937, "epoch": 752, "lr": 8.58840171238894e-05}
{"train_loss": 0.24867713451385498, "global_step": 66938, "epoch": 752, "lr": 8.588361340873825e-05}
{"train_loss": 0.16242843866348267, "global_step": 66939, "epoch": 752, "lr": 8.5883209688763e-05}
{"train_loss": 0.26448044180870056, "global_step": 66940, "epoch": 752, "lr": 8.588280596396367e-05}
{"train_loss": 0.14917437732219696, "global_step": 66941, "epoch": 752, "lr": 8.588240223434035e-05}
{"train_loss": 0.18112415075302124, "global_step": 66942, "epoch": 752, "lr": 8.588199849989307e-05}
{"train_loss": 0.3226332664489746, "global_step": 66943, "epoch": 752, "lr": 8.58815947606219e-05}
{"train_loss": 0.19951122999191284, "global_step": 66944, "epoch": 752, "lr": 8.588119101652687e-05}
{"train_loss": 0.15667013823986053, "global_step": 66945, "epoch": 752, "lr": 8.588078726760806e-05}
{"train_loss": 0.2150656282901764, "global_step": 66946, "epoch": 752, "lr": 8.588038351386552e-05}
{"train_loss": 0.34893500804901123, "global_step": 66947, "epoch": 752, "lr": 8.58799797552993e-05}
{"train_loss": 0.20428501069545746, "global_step": 66948, "epoch": 752, "lr": 8.587957599190945e-05}
{"train_loss": 0.20352032780647278, "global_step": 66949, "epoch": 752, "lr": 8.587917222369603e-05}
{"train_loss": 0.23883923888206482, "global_step": 66950, "epoch": 752, "lr": 8.58787684506591e-05}
{"train_loss": 0.19832023978233337, "global_step": 66951, "epoch": 752, "lr": 8.587836467279868e-05}
{"train_loss": 0.1958744376897812, "global_step": 66952, "epoch": 752, "lr": 8.587796089011489e-05}
{"train_loss": 0.24788345396518707, "global_step": 66953, "epoch": 752, "lr": 8.587755710260772e-05}
{"train_loss": 0.1784244030714035, "global_step": 66954, "epoch": 752, "lr": 8.587715331027725e-05}
{"train_loss": 0.2895382344722748, "global_step": 66955, "epoch": 752, "lr": 8.587674951312354e-05}
{"train_loss": 0.29022306203842163, "global_step": 66956, "epoch": 752, "lr": 8.587634571114664e-05}
{"train_loss": 0.20696184039115906, "global_step": 66957, "epoch": 752, "lr": 8.58759419043466e-05}
{"train_loss": 0.1878674328327179, "global_step": 66958, "epoch": 752, "lr": 8.587553809272348e-05}
{"train_loss": 0.2177845984697342, "global_step": 66959, "epoch": 752, "lr": 8.587513427627733e-05}
{"train_loss": 0.3177972137928009, "global_step": 66960, "epoch": 752, "lr": 8.58747304550082e-05}
{"train_loss": 0.14380694925785065, "global_step": 66961, "epoch": 752, "lr": 8.587432662891615e-05}
{"train_loss": 0.17706520855426788, "global_step": 66962, "epoch": 752, "lr": 8.587392279800124e-05}
{"train_loss": 0.22850301861763, "global_step": 66963, "epoch": 752, "lr": 8.587351896226353e-05}
{"train_loss": 0.2983393669128418, "global_step": 66964, "epoch": 752, "lr": 8.587311512170305e-05}
{"train_loss": 0.30488213896751404, "global_step": 66965, "epoch": 752, "lr": 8.587271127631985e-05}
{"train_loss": 0.16384153068065643, "global_step": 66966, "epoch": 752, "lr": 8.587230742611402e-05}
{"train_loss": 0.3408096134662628, "global_step": 66967, "epoch": 752, "lr": 8.58719035710856e-05}
{"train_loss": 0.29917484521865845, "global_step": 66968, "epoch": 752, "lr": 8.587149971123462e-05}
{"train_loss": 0.17257054150104523, "global_step": 66969, "epoch": 752, "lr": 8.587109584656116e-05}
{"train_loss": 0.26922181248664856, "global_step": 66970, "epoch": 752, "lr": 8.587069197706528e-05}
{"train_loss": 0.22865381836891174, "global_step": 66971, "epoch": 752, "lr": 8.587028810274702e-05}
{"train_loss": 0.21289797127246857, "global_step": 66972, "epoch": 752, "lr": 8.586988422360643e-05}
{"train_loss": 0.19546103477478027, "global_step": 66973, "epoch": 752, "lr": 8.586948033964358e-05}
{"train_loss": 0.21932737529277802, "global_step": 66974, "epoch": 752, "lr": 8.586907645085852e-05}
{"train_loss": 0.2243269979953766, "global_step": 66975, "epoch": 752, "lr": 8.586867255725129e-05}
{"train_loss": 0.22099345922470093, "global_step": 66976, "epoch": 752, "lr": 8.586826865882194e-05}
{"train_loss": 0.25711485743522644, "global_step": 66977, "epoch": 752, "lr": 8.586786475557056e-05}
{"train_loss": 0.21924038231372833, "global_step": 66978, "epoch": 752, "lr": 8.586746084749718e-05}
{"train_loss": 0.19133906066417694, "global_step": 66979, "epoch": 752, "lr": 8.586705693460185e-05}
{"train_loss": 0.22059375047683716, "global_step": 66980, "epoch": 752, "lr": 8.586665301688463e-05}
{"train_loss": 0.2562516927719116, "global_step": 66981, "epoch": 752, "lr": 8.586624909434558e-05}
{"train_loss": 0.22844944894313812, "global_step": 66982, "epoch": 752, "lr": 8.586584516698474e-05}
{"train_loss": 0.4010716676712036, "global_step": 66983, "epoch": 752, "lr": 8.58654412348022e-05}
{"train_loss": 0.13959643244743347, "global_step": 66984, "epoch": 752, "lr": 8.586503729779797e-05}
{"train_loss": 0.20433109998703003, "global_step": 66985, "epoch": 752, "lr": 8.586463335597212e-05}
{"train_loss": 0.19288136065006256, "global_step": 66986, "epoch": 752, "lr": 8.586422940932472e-05}
{"train_loss": 0.2514457106590271, "global_step": 66987, "epoch": 752, "lr": 8.586382545785581e-05}
{"train_loss": 0.20347043871879578, "global_step": 66988, "epoch": 752, "lr": 8.586342150156544e-05}
{"train_loss": 0.24629081785678864, "global_step": 66989, "epoch": 752, "lr": 8.586301754045367e-05}
{"train_loss": 0.16017025709152222, "global_step": 66990, "epoch": 752, "lr": 8.586261357452055e-05}
{"train_loss": 0.295793354511261, "global_step": 66991, "epoch": 752, "lr": 8.586220960376615e-05}
{"train_loss": 0.24953804910182953, "global_step": 66992, "epoch": 752, "lr": 8.586180562819051e-05}
{"train_loss": 0.18709716200828552, "global_step": 66993, "epoch": 752, "lr": 8.586140164779367e-05}
{"train_loss": 0.17417185008525848, "global_step": 66994, "epoch": 752, "lr": 8.586099766257572e-05}
{"train_loss": 0.32220861315727234, "global_step": 66995, "epoch": 752, "lr": 8.586059367253669e-05}
{"train_loss": 0.1841541826725006, "global_step": 66996, "epoch": 752, "lr": 8.586018967767664e-05}
{"train_loss": 0.1756477952003479, "global_step": 66997, "epoch": 752, "lr": 8.585978567799562e-05}
{"train_loss": 0.20584802329540253, "global_step": 66998, "epoch": 752, "lr": 8.58593816734937e-05}
{"train_loss": 0.199575275182724, "global_step": 66999, "epoch": 752, "lr": 8.585897766417091e-05}
{"train_loss": 0.24162735044956207, "global_step": 67000, "epoch": 752, "lr": 8.585857365002735e-05}
{"train_loss": 0.2486381083726883, "global_step": 67001, "epoch": 752, "lr": 8.585816963106301e-05}
{"train_loss": 0.28938037157058716, "global_step": 67002, "epoch": 752, "lr": 8.585776560727799e-05}
{"train_loss": 0.2080344557762146, "global_step": 67003, "epoch": 752, "lr": 8.585736157867232e-05}
{"train_loss": 0.2567991614341736, "global_step": 67004, "epoch": 752, "lr": 8.585695754524607e-05}
{"train_loss": 0.14038167893886566, "global_step": 67005, "epoch": 752, "lr": 8.585655350699928e-05}
{"train_loss": 0.1752379685640335, "global_step": 67006, "epoch": 752, "lr": 8.585614946393202e-05}
{"train_loss": 0.22815734148025513, "global_step": 67007, "epoch": 752, "lr": 8.585574541604434e-05}
{"train_loss": 0.14631682634353638, "global_step": 67008, "epoch": 752, "lr": 8.585534136333628e-05}
{"train_loss": 0.25466740131378174, "global_step": 67009, "epoch": 752, "lr": 8.585493730580793e-05}
{"train_loss": 0.17554226517677307, "global_step": 67010, "epoch": 752, "lr": 8.585453324345931e-05}
{"train_loss": 0.1819777935743332, "global_step": 67011, "epoch": 752, "lr": 8.585412917629048e-05}
{"train_loss": 0.15722167491912842, "global_step": 67012, "epoch": 752, "lr": 8.585372510430149e-05}
{"train_loss": 0.2332453429698944, "global_step": 67013, "epoch": 752, "lr": 8.585332102749242e-05}
{"train_loss": 0.24748194217681885, "global_step": 67014, "epoch": 752, "lr": 8.585291694586332e-05}
{"train_loss": 0.2389143854379654, "global_step": 67015, "epoch": 752, "lr": 8.585251285941422e-05}
{"train_loss": 0.22709628470828025, "global_step": 67016, "epoch": 752, "lr": 8.585210876814518e-05, "val_loss": 3.734868288040161}
{"train_loss": 0.2760031819343567, "global_step": 67017, "epoch": 753, "lr": 8.585170467205625e-05}
{"train_loss": 0.25303369760513306, "global_step": 67018, "epoch": 753, "lr": 8.585130057114752e-05}
{"train_loss": 0.18995724618434906, "global_step": 67019, "epoch": 753, "lr": 8.5850896465419e-05}
{"train_loss": 0.20856870710849762, "global_step": 67020, "epoch": 753, "lr": 8.585049235487077e-05}
{"train_loss": 0.22086402773857117, "global_step": 67021, "epoch": 753, "lr": 8.585008823950289e-05}
{"train_loss": 0.25604018568992615, "global_step": 67022, "epoch": 753, "lr": 8.58496841193154e-05}
{"train_loss": 0.21658745408058167, "global_step": 67023, "epoch": 753, "lr": 8.584927999430834e-05}
{"train_loss": 0.22377724945545197, "global_step": 67024, "epoch": 753, "lr": 8.584887586448181e-05}
{"train_loss": 0.2112618386745453, "global_step": 67025, "epoch": 753, "lr": 8.584847172983581e-05}
{"train_loss": 0.2473081797361374, "global_step": 67026, "epoch": 753, "lr": 8.584806759037042e-05}
{"train_loss": 0.3010728657245636, "global_step": 67027, "epoch": 753, "lr": 8.58476634460857e-05}
{"train_loss": 0.21960961818695068, "global_step": 67028, "epoch": 753, "lr": 8.584725929698169e-05}
{"train_loss": 0.18445837497711182, "global_step": 67029, "epoch": 753, "lr": 8.584685514305847e-05}
{"train_loss": 0.3011496067047119, "global_step": 67030, "epoch": 753, "lr": 8.584645098431607e-05}
{"train_loss": 0.21731404960155487, "global_step": 67031, "epoch": 753, "lr": 8.584604682075457e-05}
{"train_loss": 0.21995602548122406, "global_step": 67032, "epoch": 753, "lr": 8.584564265237399e-05}
{"train_loss": 0.21430304646492004, "global_step": 67033, "epoch": 753, "lr": 8.584523847917439e-05}
{"train_loss": 0.1887386292219162, "global_step": 67034, "epoch": 753, "lr": 8.584483430115584e-05}
{"train_loss": 0.16189876198768616, "global_step": 67035, "epoch": 753, "lr": 8.584443011831839e-05}
{"train_loss": 0.16034182906150818, "global_step": 67036, "epoch": 753, "lr": 8.584402593066209e-05}
{"train_loss": 0.22406846284866333, "global_step": 67037, "epoch": 753, "lr": 8.5843621738187e-05}
{"train_loss": 0.12665487825870514, "global_step": 67038, "epoch": 753, "lr": 8.584321754089317e-05}
{"train_loss": 0.1814134418964386, "global_step": 67039, "epoch": 753, "lr": 8.584281333878067e-05}
{"train_loss": 0.21016216278076172, "global_step": 67040, "epoch": 753, "lr": 8.584240913184953e-05}
{"train_loss": 0.1554824411869049, "global_step": 67041, "epoch": 753, "lr": 8.584200492009982e-05}
{"train_loss": 0.3407444655895233, "global_step": 67042, "epoch": 753, "lr": 8.584160070353159e-05}
{"train_loss": 0.22317129373550415, "global_step": 67043, "epoch": 753, "lr": 8.584119648214488e-05}
{"train_loss": 0.20622585713863373, "global_step": 67044, "epoch": 753, "lr": 8.584079225593977e-05}
{"train_loss": 0.1606122851371765, "global_step": 67045, "epoch": 753, "lr": 8.58403880249163e-05}
{"train_loss": 0.2709329426288605, "global_step": 67046, "epoch": 753, "lr": 8.583998378907451e-05}
{"train_loss": 0.1423865109682083, "global_step": 67047, "epoch": 753, "lr": 8.583957954841451e-05}
{"train_loss": 0.26366040110588074, "global_step": 67048, "epoch": 753, "lr": 8.583917530293629e-05}
{"train_loss": 0.24515095353126526, "global_step": 67049, "epoch": 753, "lr": 8.583877105263994e-05}
{"train_loss": 0.28720635175704956, "global_step": 67050, "epoch": 753, "lr": 8.583836679752549e-05}
{"train_loss": 0.23208467662334442, "global_step": 67051, "epoch": 753, "lr": 8.583796253759302e-05}
{"train_loss": 0.23187638819217682, "global_step": 67052, "epoch": 753, "lr": 8.583755827284257e-05}
{"train_loss": 0.2803317606449127, "global_step": 67053, "epoch": 753, "lr": 8.583715400327419e-05}
{"train_loss": 0.23745590448379517, "global_step": 67054, "epoch": 753, "lr": 8.583674972888794e-05}
{"train_loss": 0.1660279631614685, "global_step": 67055, "epoch": 753, "lr": 8.583634544968388e-05}
{"train_loss": 0.2889147698879242, "global_step": 67056, "epoch": 753, "lr": 8.583594116566205e-05}
{"train_loss": 0.21068666875362396, "global_step": 67057, "epoch": 753, "lr": 8.583553687682253e-05}
{"train_loss": 0.1763792186975479, "global_step": 67058, "epoch": 753, "lr": 8.583513258316536e-05}
{"train_loss": 0.21273162961006165, "global_step": 67059, "epoch": 753, "lr": 8.583472828469059e-05}
{"train_loss": 0.2251790463924408, "global_step": 67060, "epoch": 753, "lr": 8.583432398139827e-05}
{"train_loss": 0.1779554784297943, "global_step": 67061, "epoch": 753, "lr": 8.583391967328848e-05}
{"train_loss": 0.2048664391040802, "global_step": 67062, "epoch": 753, "lr": 8.583351536036124e-05}
{"train_loss": 0.18653318285942078, "global_step": 67063, "epoch": 753, "lr": 8.583311104261662e-05}
{"train_loss": 0.2727608382701874, "global_step": 67064, "epoch": 753, "lr": 8.583270672005468e-05}
{"train_loss": 0.20713293552398682, "global_step": 67065, "epoch": 753, "lr": 8.583230239267548e-05}
{"train_loss": 0.26031455397605896, "global_step": 67066, "epoch": 753, "lr": 8.583189806047905e-05}
{"train_loss": 0.20051775872707367, "global_step": 67067, "epoch": 753, "lr": 8.583149372346546e-05}
{"train_loss": 0.27431920170783997, "global_step": 67068, "epoch": 753, "lr": 8.583108938163476e-05}
{"train_loss": 0.2054438292980194, "global_step": 67069, "epoch": 753, "lr": 8.583068503498701e-05}
{"train_loss": 0.2952924966812134, "global_step": 67070, "epoch": 753, "lr": 8.583028068352225e-05}
{"train_loss": 0.2614263594150543, "global_step": 67071, "epoch": 753, "lr": 8.582987632724057e-05}
{"train_loss": 0.19878637790679932, "global_step": 67072, "epoch": 753, "lr": 8.582947196614198e-05}
{"train_loss": 0.23860976099967957, "global_step": 67073, "epoch": 753, "lr": 8.582906760022656e-05}
{"train_loss": 0.1868942528963089, "global_step": 67074, "epoch": 753, "lr": 8.582866322949436e-05}
{"train_loss": 0.37665480375289917, "global_step": 67075, "epoch": 753, "lr": 8.582825885394543e-05}
{"train_loss": 0.3042345345020294, "global_step": 67076, "epoch": 753, "lr": 8.582785447357984e-05}
{"train_loss": 0.19791090488433838, "global_step": 67077, "epoch": 753, "lr": 8.582745008839761e-05}
{"train_loss": 0.1994771659374237, "global_step": 67078, "epoch": 753, "lr": 8.582704569839883e-05}
{"train_loss": 0.19464710354804993, "global_step": 67079, "epoch": 753, "lr": 8.582664130358355e-05}
{"train_loss": 0.16434475779533386, "global_step": 67080, "epoch": 753, "lr": 8.582623690395179e-05}
{"train_loss": 0.201110377907753, "global_step": 67081, "epoch": 753, "lr": 8.582583249950365e-05}
{"train_loss": 0.23167824745178223, "global_step": 67082, "epoch": 753, "lr": 8.582542809023915e-05}
{"train_loss": 0.2317901849746704, "global_step": 67083, "epoch": 753, "lr": 8.582502367615835e-05}
{"train_loss": 0.23956012725830078, "global_step": 67084, "epoch": 753, "lr": 8.582461925726132e-05}
{"train_loss": 0.22590014338493347, "global_step": 67085, "epoch": 753, "lr": 8.582421483354812e-05}
{"train_loss": 0.19298376142978668, "global_step": 67086, "epoch": 753, "lr": 8.582381040501879e-05}
{"train_loss": 0.2598934769630432, "global_step": 67087, "epoch": 753, "lr": 8.582340597167337e-05}
{"train_loss": 0.24809952080249786, "global_step": 67088, "epoch": 753, "lr": 8.582300153351195e-05}
{"train_loss": 0.18752455711364746, "global_step": 67089, "epoch": 753, "lr": 8.582259709053455e-05}
{"train_loss": 0.2399948537349701, "global_step": 67090, "epoch": 753, "lr": 8.582219264274123e-05}
{"train_loss": 0.20859797298908234, "global_step": 67091, "epoch": 753, "lr": 8.582178819013206e-05}
{"train_loss": 0.21455372869968414, "global_step": 67092, "epoch": 753, "lr": 8.58213837327071e-05}
{"train_loss": 0.28384339809417725, "global_step": 67093, "epoch": 753, "lr": 8.582097927046637e-05}
{"train_loss": 0.303914874792099, "global_step": 67094, "epoch": 753, "lr": 8.582057480340998e-05}
{"train_loss": 0.208660289645195, "global_step": 67095, "epoch": 753, "lr": 8.582017033153791e-05}
{"train_loss": 0.2498590648174286, "global_step": 67096, "epoch": 753, "lr": 8.581976585485028e-05}
{"train_loss": 0.2504165470600128, "global_step": 67097, "epoch": 753, "lr": 8.58193613733471e-05}
{"train_loss": 0.2754160761833191, "global_step": 67098, "epoch": 753, "lr": 8.581895688702846e-05}
{"train_loss": 0.25301986932754517, "global_step": 67099, "epoch": 753, "lr": 8.581855239589439e-05}
{"train_loss": 0.12829838693141937, "global_step": 67100, "epoch": 753, "lr": 8.581814789994496e-05}
{"train_loss": 0.23271481692790985, "global_step": 67101, "epoch": 753, "lr": 8.581774339918021e-05}
{"train_loss": 0.27480778098106384, "global_step": 67102, "epoch": 753, "lr": 8.58173388936002e-05}
{"train_loss": 0.25691550970077515, "global_step": 67103, "epoch": 753, "lr": 8.5816934383205e-05}
{"train_loss": 0.15979641675949097, "global_step": 67104, "epoch": 753, "lr": 8.581652986799462e-05}
{"train_loss": 0.22630058228969574, "global_step": 67105, "epoch": 753, "lr": 8.581612534796917e-05, "val_loss": 3.659726619720459}
{"train_loss": 0.25100621581077576, "global_step": 67106, "epoch": 754, "lr": 8.581572082312866e-05}
{"train_loss": 0.14727024734020233, "global_step": 67107, "epoch": 754, "lr": 8.581531629347317e-05}
{"train_loss": 0.165286585688591, "global_step": 67108, "epoch": 754, "lr": 8.581491175900276e-05}
{"train_loss": 0.2892696261405945, "global_step": 67109, "epoch": 754, "lr": 8.581450721971747e-05}
{"train_loss": 0.4522576630115509, "global_step": 67110, "epoch": 754, "lr": 8.581410267561733e-05}
{"train_loss": 0.20230109989643097, "global_step": 67111, "epoch": 754, "lr": 8.581369812670244e-05}
{"train_loss": 0.17801937460899353, "global_step": 67112, "epoch": 754, "lr": 8.581329357297283e-05}
{"train_loss": 0.17245417833328247, "global_step": 67113, "epoch": 754, "lr": 8.581288901442857e-05}
{"train_loss": 0.1762591302394867, "global_step": 67114, "epoch": 754, "lr": 8.58124844510697e-05}
{"train_loss": 0.20405595004558563, "global_step": 67115, "epoch": 754, "lr": 8.581207988289626e-05}
{"train_loss": 0.32135409116744995, "global_step": 67116, "epoch": 754, "lr": 8.581167530990834e-05}
{"train_loss": 0.1698758602142334, "global_step": 67117, "epoch": 754, "lr": 8.581127073210597e-05}
{"train_loss": 0.2274290770292282, "global_step": 67118, "epoch": 754, "lr": 8.581086614948922e-05}
{"train_loss": 0.24355795979499817, "global_step": 67119, "epoch": 754, "lr": 8.581046156205813e-05}
{"train_loss": 0.3134867548942566, "global_step": 67120, "epoch": 754, "lr": 8.581005696981276e-05}
{"train_loss": 0.14081305265426636, "global_step": 67121, "epoch": 754, "lr": 8.580965237275315e-05}
{"train_loss": 0.22366389632225037, "global_step": 67122, "epoch": 754, "lr": 8.58092477708794e-05}
{"train_loss": 0.28791365027427673, "global_step": 67123, "epoch": 754, "lr": 8.58088431641915e-05}
{"train_loss": 0.19986967742443085, "global_step": 67124, "epoch": 754, "lr": 8.580843855268957e-05}
{"train_loss": 0.3107462227344513, "global_step": 67125, "epoch": 754, "lr": 8.580803393637361e-05}
{"train_loss": 0.27100056409835815, "global_step": 67126, "epoch": 754, "lr": 8.58076293152437e-05}
{"train_loss": 0.21036340296268463, "global_step": 67127, "epoch": 754, "lr": 8.58072246892999e-05}
{"train_loss": 0.2668018937110901, "global_step": 67128, "epoch": 754, "lr": 8.580682005854224e-05}
{"train_loss": 0.21741367876529694, "global_step": 67129, "epoch": 754, "lr": 8.580641542297079e-05}
{"train_loss": 0.36455589532852173, "global_step": 67130, "epoch": 754, "lr": 8.580601078258561e-05}
{"train_loss": 0.24688319861888885, "global_step": 67131, "epoch": 754, "lr": 8.580560613738675e-05}
{"train_loss": 0.30331775546073914, "global_step": 67132, "epoch": 754, "lr": 8.580520148737427e-05}
{"train_loss": 0.18612967431545258, "global_step": 67133, "epoch": 754, "lr": 8.580479683254821e-05}
{"train_loss": 0.20113462209701538, "global_step": 67134, "epoch": 754, "lr": 8.580439217290862e-05}
{"train_loss": 0.2847011983394623, "global_step": 67135, "epoch": 754, "lr": 8.580398750845557e-05}
{"train_loss": 0.23494014143943787, "global_step": 67136, "epoch": 754, "lr": 8.580358283918912e-05}
{"train_loss": 0.1567426472902298, "global_step": 67137, "epoch": 754, "lr": 8.580317816510931e-05}
{"train_loss": 0.2972692847251892, "global_step": 67138, "epoch": 754, "lr": 8.58027734862162e-05}
{"train_loss": 0.1700228750705719, "global_step": 67139, "epoch": 754, "lr": 8.580236880250984e-05}
{"train_loss": 0.1969740241765976, "global_step": 67140, "epoch": 754, "lr": 8.580196411399029e-05}
{"train_loss": 0.13451573252677917, "global_step": 67141, "epoch": 754, "lr": 8.580155942065761e-05}
{"train_loss": 0.24215006828308105, "global_step": 67142, "epoch": 754, "lr": 8.580115472251185e-05}
{"train_loss": 0.21905876696109772, "global_step": 67143, "epoch": 754, "lr": 8.580075001955303e-05}
{"train_loss": 0.24110522866249084, "global_step": 67144, "epoch": 754, "lr": 8.580034531178127e-05}
{"train_loss": 0.21179227530956268, "global_step": 67145, "epoch": 754, "lr": 8.579994059919658e-05}
{"train_loss": 0.24155230820178986, "global_step": 67146, "epoch": 754, "lr": 8.579953588179901e-05}
{"train_loss": 0.20538511872291565, "global_step": 67147, "epoch": 754, "lr": 8.579913115958864e-05}
{"train_loss": 0.1722884476184845, "global_step": 67148, "epoch": 754, "lr": 8.579872643256552e-05}
{"train_loss": 0.29625818133354187, "global_step": 67149, "epoch": 754, "lr": 8.579832170072968e-05}
{"train_loss": 0.23663932085037231, "global_step": 67150, "epoch": 754, "lr": 8.579791696408121e-05}
{"train_loss": 0.2804507613182068, "global_step": 67151, "epoch": 754, "lr": 8.579751222262013e-05}
{"train_loss": 0.22095972299575806, "global_step": 67152, "epoch": 754, "lr": 8.579710747634653e-05}
{"train_loss": 0.1891953945159912, "global_step": 67153, "epoch": 754, "lr": 8.579670272526043e-05}
{"train_loss": 0.14733558893203735, "global_step": 67154, "epoch": 754, "lr": 8.57962979693619e-05}
{"train_loss": 0.27770501375198364, "global_step": 67155, "epoch": 754, "lr": 8.5795893208651e-05}
{"train_loss": 0.1936226338148117, "global_step": 67156, "epoch": 754, "lr": 8.579548844312777e-05}
{"train_loss": 0.19888877868652344, "global_step": 67157, "epoch": 754, "lr": 8.579508367279229e-05}
{"train_loss": 0.21318168938159943, "global_step": 67158, "epoch": 754, "lr": 8.579467889764459e-05}
{"train_loss": 0.22079645097255707, "global_step": 67159, "epoch": 754, "lr": 8.579427411768473e-05}
{"train_loss": 0.2154553085565567, "global_step": 67160, "epoch": 754, "lr": 8.579386933291275e-05}
{"train_loss": 0.25757864117622375, "global_step": 67161, "epoch": 754, "lr": 8.579346454332874e-05}
{"train_loss": 0.22660647332668304, "global_step": 67162, "epoch": 754, "lr": 8.579305974893273e-05}
{"train_loss": 0.19881120324134827, "global_step": 67163, "epoch": 754, "lr": 8.579265494972477e-05}
{"train_loss": 0.20248085260391235, "global_step": 67164, "epoch": 754, "lr": 8.579225014570494e-05}
{"train_loss": 0.20810946822166443, "global_step": 67165, "epoch": 754, "lr": 8.579184533687327e-05}
{"train_loss": 0.21033738553524017, "global_step": 67166, "epoch": 754, "lr": 8.579144052322983e-05}
{"train_loss": 0.15812942385673523, "global_step": 67167, "epoch": 754, "lr": 8.579103570477466e-05}
{"train_loss": 0.16804660856723785, "global_step": 67168, "epoch": 754, "lr": 8.579063088150783e-05}
{"train_loss": 0.22441722452640533, "global_step": 67169, "epoch": 754, "lr": 8.579022605342937e-05}
{"train_loss": 0.24511176347732544, "global_step": 67170, "epoch": 754, "lr": 8.578982122053935e-05}
{"train_loss": 0.22826644778251648, "global_step": 67171, "epoch": 754, "lr": 8.578941638283784e-05}
{"train_loss": 0.34623202681541443, "global_step": 67172, "epoch": 754, "lr": 8.578901154032488e-05}
{"train_loss": 0.265186071395874, "global_step": 67173, "epoch": 754, "lr": 8.578860669300051e-05}
{"train_loss": 0.2034820020198822, "global_step": 67174, "epoch": 754, "lr": 8.57882018408648e-05}
{"train_loss": 0.24239292740821838, "global_step": 67175, "epoch": 754, "lr": 8.578779698391781e-05}
{"train_loss": 0.19752448797225952, "global_step": 67176, "epoch": 754, "lr": 8.578739212215958e-05}
{"train_loss": 0.19014495611190796, "global_step": 67177, "epoch": 754, "lr": 8.578698725559018e-05}
{"train_loss": 0.19643718004226685, "global_step": 67178, "epoch": 754, "lr": 8.578658238420965e-05}
{"train_loss": 0.2435007393360138, "global_step": 67179, "epoch": 754, "lr": 8.578617750801804e-05}
{"train_loss": 0.21316851675510406, "global_step": 67180, "epoch": 754, "lr": 8.578577262701543e-05}
{"train_loss": 0.2523641586303711, "global_step": 67181, "epoch": 754, "lr": 8.578536774120185e-05}
{"train_loss": 0.17611217498779297, "global_step": 67182, "epoch": 754, "lr": 8.578496285057738e-05}
{"train_loss": 0.275469571352005, "global_step": 67183, "epoch": 754, "lr": 8.578455795514204e-05}
{"train_loss": 0.2762434482574463, "global_step": 67184, "epoch": 754, "lr": 8.578415305489591e-05}
{"train_loss": 0.2620890140533447, "global_step": 67185, "epoch": 754, "lr": 8.578374814983904e-05}
{"train_loss": 0.24264661967754364, "global_step": 67186, "epoch": 754, "lr": 8.578334323997148e-05}
{"train_loss": 0.22524087131023407, "global_step": 67187, "epoch": 754, "lr": 8.578293832529328e-05}
{"train_loss": 0.2770754098892212, "global_step": 67188, "epoch": 754, "lr": 8.57825334058045e-05}
{"train_loss": 0.21869350969791412, "global_step": 67189, "epoch": 754, "lr": 8.578212848150519e-05}
{"train_loss": 0.16870102286338806, "global_step": 67190, "epoch": 754, "lr": 8.578172355239542e-05}
{"train_loss": 0.2609911561012268, "global_step": 67191, "epoch": 754, "lr": 8.578131861847523e-05}
{"train_loss": 0.2454679012298584, "global_step": 67192, "epoch": 754, "lr": 8.578091367974468e-05}
{"train_loss": 0.17298130691051483, "global_step": 67193, "epoch": 754, "lr": 8.578050873620381e-05}
{"train_loss": 0.2282146058055792, "global_step": 67194, "epoch": 754, "lr": 8.578010378785271e-05, "val_loss": 3.606029987335205}
{"train_loss": 0.16131675243377686, "global_step": 67195, "epoch": 755, "lr": 8.577969883469139e-05}
{"train_loss": 0.1952965259552002, "global_step": 67196, "epoch": 755, "lr": 8.577929387671992e-05}
{"train_loss": 0.19842199981212616, "global_step": 67197, "epoch": 755, "lr": 8.577888891393838e-05}
{"train_loss": 0.3060121536254883, "global_step": 67198, "epoch": 755, "lr": 8.57784839463468e-05}
{"train_loss": 0.22142140567302704, "global_step": 67199, "epoch": 755, "lr": 8.577807897394523e-05}
{"train_loss": 0.2981395721435547, "global_step": 67200, "epoch": 755, "lr": 8.577767399673372e-05}
{"train_loss": 0.2826659679412842, "global_step": 67201, "epoch": 755, "lr": 8.577726901471236e-05}
{"train_loss": 0.17849045991897583, "global_step": 67202, "epoch": 755, "lr": 8.57768640278812e-05}
{"train_loss": 0.2533857524394989, "global_step": 67203, "epoch": 755, "lr": 8.577645903624025e-05}
{"train_loss": 0.19030612707138062, "global_step": 67204, "epoch": 755, "lr": 8.57760540397896e-05}
{"train_loss": 0.2179788500070572, "global_step": 67205, "epoch": 755, "lr": 8.57756490385293e-05}
{"train_loss": 0.15308474004268646, "global_step": 67206, "epoch": 755, "lr": 8.577524403245939e-05}
{"train_loss": 0.23596696555614471, "global_step": 67207, "epoch": 755, "lr": 8.577483902157992e-05}
{"train_loss": 0.1523781418800354, "global_step": 67208, "epoch": 755, "lr": 8.577443400589098e-05}
{"train_loss": 0.19799114763736725, "global_step": 67209, "epoch": 755, "lr": 8.57740289853926e-05}
{"train_loss": 0.3416294455528259, "global_step": 67210, "epoch": 755, "lr": 8.577362396008483e-05}
{"train_loss": 0.26478034257888794, "global_step": 67211, "epoch": 755, "lr": 8.577321892996775e-05}
{"train_loss": 0.26368120312690735, "global_step": 67212, "epoch": 755, "lr": 8.577281389504139e-05}
{"train_loss": 0.16736458241939545, "global_step": 67213, "epoch": 755, "lr": 8.57724088553058e-05}
{"train_loss": 0.2281181663274765, "global_step": 67214, "epoch": 755, "lr": 8.577200381076106e-05}
{"train_loss": 0.2773269712924957, "global_step": 67215, "epoch": 755, "lr": 8.57715987614072e-05}
{"train_loss": 0.27032211422920227, "global_step": 67216, "epoch": 755, "lr": 8.577119370724428e-05}
{"train_loss": 0.23468098044395447, "global_step": 67217, "epoch": 755, "lr": 8.577078864827237e-05}
{"train_loss": 0.3033904731273651, "global_step": 67218, "epoch": 755, "lr": 8.577038358449152e-05}
{"train_loss": 0.3105168640613556, "global_step": 67219, "epoch": 755, "lr": 8.576997851590175e-05}
{"train_loss": 0.26249954104423523, "global_step": 67220, "epoch": 755, "lr": 8.576957344250317e-05}
{"train_loss": 0.22136615216732025, "global_step": 67221, "epoch": 755, "lr": 8.576916836429581e-05}
{"train_loss": 0.16512326896190643, "global_step": 67222, "epoch": 755, "lr": 8.57687632812797e-05}
{"train_loss": 0.2305581420660019, "global_step": 67223, "epoch": 755, "lr": 8.576835819345493e-05}
{"train_loss": 0.22798393666744232, "global_step": 67224, "epoch": 755, "lr": 8.576795310082153e-05}
{"train_loss": 0.1981605738401413, "global_step": 67225, "epoch": 755, "lr": 8.576754800337957e-05}
{"train_loss": 0.27699190378189087, "global_step": 67226, "epoch": 755, "lr": 8.576714290112909e-05}
{"train_loss": 0.21845977008342743, "global_step": 67227, "epoch": 755, "lr": 8.576673779407016e-05}
{"train_loss": 0.2531678378582001, "global_step": 67228, "epoch": 755, "lr": 8.576633268220284e-05}
{"train_loss": 0.27276989817619324, "global_step": 67229, "epoch": 755, "lr": 8.576592756552714e-05}
{"train_loss": 0.15974979102611542, "global_step": 67230, "epoch": 755, "lr": 8.576552244404319e-05}
{"train_loss": 0.19995719194412231, "global_step": 67231, "epoch": 755, "lr": 8.576511731775099e-05}
{"train_loss": 0.24579687416553497, "global_step": 67232, "epoch": 755, "lr": 8.576471218665058e-05}
{"train_loss": 0.2482958883047104, "global_step": 67233, "epoch": 755, "lr": 8.576430705074206e-05}
{"train_loss": 0.2331608384847641, "global_step": 67234, "epoch": 755, "lr": 8.576390191002546e-05}
{"train_loss": 0.22975319623947144, "global_step": 67235, "epoch": 755, "lr": 8.576349676450084e-05}
{"train_loss": 0.18199074268341064, "global_step": 67236, "epoch": 755, "lr": 8.576309161416825e-05}
{"train_loss": 0.30114540457725525, "global_step": 67237, "epoch": 755, "lr": 8.576268645902776e-05}
{"train_loss": 0.2518272399902344, "global_step": 67238, "epoch": 755, "lr": 8.57622812990794e-05}
{"train_loss": 0.20331647992134094, "global_step": 67239, "epoch": 755, "lr": 8.576187613432325e-05}
{"train_loss": 0.24657978117465973, "global_step": 67240, "epoch": 755, "lr": 8.576147096475935e-05}
{"train_loss": 0.16647660732269287, "global_step": 67241, "epoch": 755, "lr": 8.576106579038774e-05}
{"train_loss": 0.19558236002922058, "global_step": 67242, "epoch": 755, "lr": 8.576066061120851e-05}
{"train_loss": 0.20301567018032074, "global_step": 67243, "epoch": 755, "lr": 8.576025542722168e-05}
{"train_loss": 0.21675686538219452, "global_step": 67244, "epoch": 755, "lr": 8.575985023842731e-05}
{"train_loss": 0.27422896027565, "global_step": 67245, "epoch": 755, "lr": 8.57594450448255e-05}
{"train_loss": 0.24714647233486176, "global_step": 67246, "epoch": 755, "lr": 8.575903984641625e-05}
{"train_loss": 0.17555035650730133, "global_step": 67247, "epoch": 755, "lr": 8.575863464319962e-05}
{"train_loss": 0.19039148092269897, "global_step": 67248, "epoch": 755, "lr": 8.575822943517568e-05}
{"train_loss": 0.17711670696735382, "global_step": 67249, "epoch": 755, "lr": 8.575782422234451e-05}
{"train_loss": 0.23907604813575745, "global_step": 67250, "epoch": 755, "lr": 8.575741900470611e-05}
{"train_loss": 0.2040785402059555, "global_step": 67251, "epoch": 755, "lr": 8.575701378226057e-05}
{"train_loss": 0.23084111511707306, "global_step": 67252, "epoch": 755, "lr": 8.575660855500794e-05}
{"train_loss": 0.2747679054737091, "global_step": 67253, "epoch": 755, "lr": 8.575620332294825e-05}
{"train_loss": 0.25256747007369995, "global_step": 67254, "epoch": 755, "lr": 8.575579808608159e-05}
{"train_loss": 0.1741405576467514, "global_step": 67255, "epoch": 755, "lr": 8.575539284440799e-05}
{"train_loss": 0.28367966413497925, "global_step": 67256, "epoch": 755, "lr": 8.575498759792753e-05}
{"train_loss": 0.267415851354599, "global_step": 67257, "epoch": 755, "lr": 8.575458234664024e-05}
{"train_loss": 0.2124294489622116, "global_step": 67258, "epoch": 755, "lr": 8.575417709054617e-05}
{"train_loss": 0.2618445158004761, "global_step": 67259, "epoch": 755, "lr": 8.575377182964539e-05}
{"train_loss": 0.1637868881225586, "global_step": 67260, "epoch": 755, "lr": 8.575336656393797e-05}
{"train_loss": 0.19634424149990082, "global_step": 67261, "epoch": 755, "lr": 8.575296129342393e-05}
{"train_loss": 0.22272445261478424, "global_step": 67262, "epoch": 755, "lr": 8.575255601810333e-05}
{"train_loss": 0.21568657457828522, "global_step": 67263, "epoch": 755, "lr": 8.575215073797626e-05}
{"train_loss": 0.21985208988189697, "global_step": 67264, "epoch": 755, "lr": 8.575174545304273e-05}
{"train_loss": 0.219330832362175, "global_step": 67265, "epoch": 755, "lr": 8.575134016330281e-05}
{"train_loss": 0.15366099774837494, "global_step": 67266, "epoch": 755, "lr": 8.575093486875658e-05}
{"train_loss": 0.2075793594121933, "global_step": 67267, "epoch": 755, "lr": 8.575052956940406e-05}
{"train_loss": 0.22196723520755768, "global_step": 67268, "epoch": 755, "lr": 8.575012426524531e-05}
{"train_loss": 0.27358299493789673, "global_step": 67269, "epoch": 755, "lr": 8.574971895628041e-05}
{"train_loss": 0.15519854426383972, "global_step": 67270, "epoch": 755, "lr": 8.574931364250938e-05}
{"train_loss": 0.25233182311058044, "global_step": 67271, "epoch": 755, "lr": 8.57489083239323e-05}
{"train_loss": 0.2268250584602356, "global_step": 67272, "epoch": 755, "lr": 8.574850300054919e-05}
{"train_loss": 0.20120789110660553, "global_step": 67273, "epoch": 755, "lr": 8.574809767236016e-05}
{"train_loss": 0.19370339810848236, "global_step": 67274, "epoch": 755, "lr": 8.574769233936523e-05}
{"train_loss": 0.23316732048988342, "global_step": 67275, "epoch": 755, "lr": 8.574728700156445e-05}
{"train_loss": 0.213189959526062, "global_step": 67276, "epoch": 755, "lr": 8.574688165895788e-05}
{"train_loss": 0.3327949047088623, "global_step": 67277, "epoch": 755, "lr": 8.574647631154559e-05}
{"train_loss": 0.18470127880573273, "global_step": 67278, "epoch": 755, "lr": 8.574607095932761e-05}
{"train_loss": 0.23612073063850403, "global_step": 67279, "epoch": 755, "lr": 8.574566560230401e-05}
{"train_loss": 0.17182621359825134, "global_step": 67280, "epoch": 755, "lr": 8.574526024047484e-05}
{"train_loss": 0.20172002911567688, "global_step": 67281, "epoch": 755, "lr": 8.574485487384016e-05}
{"train_loss": 0.19781416654586792, "global_step": 67282, "epoch": 755, "lr": 8.574444950240002e-05}
{"train_loss": 0.22605784639213863, "global_step": 67283, "epoch": 755, "lr": 8.574404412615447e-05, "val_loss": 3.548006296157837, "train_action_mse_error": 32.58163070678711}
{"train_loss": 0.37521877884864807, "global_step": 67284, "epoch": 756, "lr": 8.574363874510357e-05}
{"train_loss": 0.19869954884052277, "global_step": 67285, "epoch": 756, "lr": 8.574323335924738e-05}
{"train_loss": 0.23396271467208862, "global_step": 67286, "epoch": 756, "lr": 8.574282796858594e-05}
{"train_loss": 0.3091796934604645, "global_step": 67287, "epoch": 756, "lr": 8.574242257311931e-05}
{"train_loss": 0.18389663100242615, "global_step": 67288, "epoch": 756, "lr": 8.574201717284756e-05}
{"train_loss": 0.20363788306713104, "global_step": 67289, "epoch": 756, "lr": 8.574161176777071e-05}
{"train_loss": 0.17856524884700775, "global_step": 67290, "epoch": 756, "lr": 8.574120635788886e-05}
{"train_loss": 0.21421398222446442, "global_step": 67291, "epoch": 756, "lr": 8.574080094320202e-05}
{"train_loss": 0.1458633989095688, "global_step": 67292, "epoch": 756, "lr": 8.574039552371028e-05}
{"train_loss": 0.19397372007369995, "global_step": 67293, "epoch": 756, "lr": 8.573999009941367e-05}
{"train_loss": 0.3485437035560608, "global_step": 67294, "epoch": 756, "lr": 8.573958467031227e-05}
{"train_loss": 0.32341060042381287, "global_step": 67295, "epoch": 756, "lr": 8.57391792364061e-05}
{"train_loss": 0.22874446213245392, "global_step": 67296, "epoch": 756, "lr": 8.573877379769524e-05}
{"train_loss": 0.2544393539428711, "global_step": 67297, "epoch": 756, "lr": 8.573836835417973e-05}
{"train_loss": 0.23233069479465485, "global_step": 67298, "epoch": 756, "lr": 8.573796290585964e-05}
{"train_loss": 0.14845439791679382, "global_step": 67299, "epoch": 756, "lr": 8.573755745273502e-05}
{"train_loss": 0.21187926828861237, "global_step": 67300, "epoch": 756, "lr": 8.573715199480591e-05}
{"train_loss": 0.22290518879890442, "global_step": 67301, "epoch": 756, "lr": 8.573674653207237e-05}
{"train_loss": 0.13874463737010956, "global_step": 67302, "epoch": 756, "lr": 8.573634106453448e-05}
{"train_loss": 0.15808817744255066, "global_step": 67303, "epoch": 756, "lr": 8.573593559219227e-05}
{"train_loss": 0.1953626126050949, "global_step": 67304, "epoch": 756, "lr": 8.573553011504579e-05}
{"train_loss": 0.23394714295864105, "global_step": 67305, "epoch": 756, "lr": 8.573512463309512e-05}
{"train_loss": 0.22505804896354675, "global_step": 67306, "epoch": 756, "lr": 8.573471914634029e-05}
{"train_loss": 0.24757011234760284, "global_step": 67307, "epoch": 756, "lr": 8.573431365478135e-05}
{"train_loss": 0.17948713898658752, "global_step": 67308, "epoch": 756, "lr": 8.573390815841838e-05}
{"train_loss": 0.2052249014377594, "global_step": 67309, "epoch": 756, "lr": 8.573350265725142e-05}
{"train_loss": 0.18332092463970184, "global_step": 67310, "epoch": 756, "lr": 8.573309715128051e-05}
{"train_loss": 0.21131382882595062, "global_step": 67311, "epoch": 756, "lr": 8.573269164050574e-05}
{"train_loss": 0.2573077082633972, "global_step": 67312, "epoch": 756, "lr": 8.573228612492714e-05}
{"train_loss": 0.2575191855430603, "global_step": 67313, "epoch": 756, "lr": 8.573188060454477e-05}
{"train_loss": 0.1875487118959427, "global_step": 67314, "epoch": 756, "lr": 8.573147507935868e-05}
{"train_loss": 0.29441118240356445, "global_step": 67315, "epoch": 756, "lr": 8.573106954936893e-05}
{"train_loss": 0.23269008100032806, "global_step": 67316, "epoch": 756, "lr": 8.573066401457558e-05}
{"train_loss": 0.23484019935131073, "global_step": 67317, "epoch": 756, "lr": 8.573025847497866e-05}
{"train_loss": 0.1838306188583374, "global_step": 67318, "epoch": 756, "lr": 8.572985293057828e-05}
{"train_loss": 0.18146641552448273, "global_step": 67319, "epoch": 756, "lr": 8.572944738137441e-05}
{"train_loss": 0.25094813108444214, "global_step": 67320, "epoch": 756, "lr": 8.572904182736719e-05}
{"train_loss": 0.2218959778547287, "global_step": 67321, "epoch": 756, "lr": 8.57286362685566e-05}
{"train_loss": 0.1686675250530243, "global_step": 67322, "epoch": 756, "lr": 8.572823070494276e-05}
{"train_loss": 0.17315703630447388, "global_step": 67323, "epoch": 756, "lr": 8.572782513652569e-05}
{"train_loss": 0.19526827335357666, "global_step": 67324, "epoch": 756, "lr": 8.572741956330543e-05}
{"train_loss": 0.24646274745464325, "global_step": 67325, "epoch": 756, "lr": 8.572701398528206e-05}
{"train_loss": 0.26425784826278687, "global_step": 67326, "epoch": 756, "lr": 8.572660840245565e-05}
{"train_loss": 0.25879722833633423, "global_step": 67327, "epoch": 756, "lr": 8.572620281482621e-05}
{"train_loss": 0.2262144833803177, "global_step": 67328, "epoch": 756, "lr": 8.572579722239384e-05}
{"train_loss": 0.20602019131183624, "global_step": 67329, "epoch": 756, "lr": 8.572539162515855e-05}
{"train_loss": 0.19224554300308228, "global_step": 67330, "epoch": 756, "lr": 8.572498602312042e-05}
{"train_loss": 0.3548833131790161, "global_step": 67331, "epoch": 756, "lr": 8.572458041627951e-05}
{"train_loss": 0.2929019033908844, "global_step": 67332, "epoch": 756, "lr": 8.572417480463586e-05}
{"train_loss": 0.24966180324554443, "global_step": 67333, "epoch": 756, "lr": 8.572376918818953e-05}
{"train_loss": 0.20437391102313995, "global_step": 67334, "epoch": 756, "lr": 8.572336356694057e-05}
{"train_loss": 0.2522246241569519, "global_step": 67335, "epoch": 756, "lr": 8.572295794088906e-05}
{"train_loss": 0.17475645244121552, "global_step": 67336, "epoch": 756, "lr": 8.5722552310035e-05}
{"train_loss": 0.13881118595600128, "global_step": 67337, "epoch": 756, "lr": 8.572214667437851e-05}
{"train_loss": 0.23719632625579834, "global_step": 67338, "epoch": 756, "lr": 8.572174103391961e-05}
{"train_loss": 0.1536494344472885, "global_step": 67339, "epoch": 756, "lr": 8.572133538865833e-05}
{"train_loss": 0.27485260367393494, "global_step": 67340, "epoch": 756, "lr": 8.572092973859477e-05}
{"train_loss": 0.2000700831413269, "global_step": 67341, "epoch": 756, "lr": 8.572052408372897e-05}
{"train_loss": 0.20816145837306976, "global_step": 67342, "epoch": 756, "lr": 8.572011842406098e-05}
{"train_loss": 0.22031095623970032, "global_step": 67343, "epoch": 756, "lr": 8.571971275959085e-05}
{"train_loss": 0.21336102485656738, "global_step": 67344, "epoch": 756, "lr": 8.571930709031865e-05}
{"train_loss": 0.32253649830818176, "global_step": 67345, "epoch": 756, "lr": 8.571890141624442e-05}
{"train_loss": 0.16819410026073456, "global_step": 67346, "epoch": 756, "lr": 8.571849573736821e-05}
{"train_loss": 0.2491701990365982, "global_step": 67347, "epoch": 756, "lr": 8.571809005369009e-05}
{"train_loss": 0.2436685413122177, "global_step": 67348, "epoch": 756, "lr": 8.571768436521011e-05}
{"train_loss": 0.15096528828144073, "global_step": 67349, "epoch": 756, "lr": 8.571727867192831e-05}
{"train_loss": 0.2467796355485916, "global_step": 67350, "epoch": 756, "lr": 8.571687297384476e-05}
{"train_loss": 0.2635323703289032, "global_step": 67351, "epoch": 756, "lr": 8.571646727095953e-05}
{"train_loss": 0.33633217215538025, "global_step": 67352, "epoch": 756, "lr": 8.571606156327265e-05}
{"train_loss": 0.3422594964504242, "global_step": 67353, "epoch": 756, "lr": 8.571565585078416e-05}
{"train_loss": 0.24635504186153412, "global_step": 67354, "epoch": 756, "lr": 8.571525013349416e-05}
{"train_loss": 0.171199768781662, "global_step": 67355, "epoch": 756, "lr": 8.571484441140267e-05}
{"train_loss": 0.23303847014904022, "global_step": 67356, "epoch": 756, "lr": 8.571443868450975e-05}
{"train_loss": 0.20136083662509918, "global_step": 67357, "epoch": 756, "lr": 8.571403295281548e-05}
{"train_loss": 0.13307425379753113, "global_step": 67358, "epoch": 756, "lr": 8.571362721631988e-05}
{"train_loss": 0.17768412828445435, "global_step": 67359, "epoch": 756, "lr": 8.571322147502301e-05}
{"train_loss": 0.22554649412631989, "global_step": 67360, "epoch": 756, "lr": 8.571281572892493e-05}
{"train_loss": 0.195653036236763, "global_step": 67361, "epoch": 756, "lr": 8.571240997802573e-05}
{"train_loss": 0.20320086181163788, "global_step": 67362, "epoch": 756, "lr": 8.571200422232539e-05}
{"train_loss": 0.2889390289783478, "global_step": 67363, "epoch": 756, "lr": 8.571159846182402e-05}
{"train_loss": 0.26941585540771484, "global_step": 67364, "epoch": 756, "lr": 8.571119269652167e-05}
{"train_loss": 0.23376686871051788, "global_step": 67365, "epoch": 756, "lr": 8.571078692641839e-05}
{"train_loss": 0.2767612636089325, "global_step": 67366, "epoch": 756, "lr": 8.571038115151422e-05}
{"train_loss": 0.22327764332294464, "global_step": 67367, "epoch": 756, "lr": 8.570997537180923e-05}
{"train_loss": 0.25702959299087524, "global_step": 67368, "epoch": 756, "lr": 8.570956958730345e-05}
{"train_loss": 0.23099254071712494, "global_step": 67369, "epoch": 756, "lr": 8.570916379799698e-05}
{"train_loss": 0.29215583205223083, "global_step": 67370, "epoch": 756, "lr": 8.570875800388983e-05}
{"train_loss": 0.2745693325996399, "global_step": 67371, "epoch": 756, "lr": 8.570835220498208e-05}
{"train_loss": 0.2296756561217683, "global_step": 67372, "epoch": 756, "lr": 8.570794640127377e-05, "val_loss": 3.4241065979003906}
{"train_loss": 0.22600914537906647, "global_step": 67373, "epoch": 757, "lr": 8.570754059276497e-05}
{"train_loss": 0.19861005246639252, "global_step": 67374, "epoch": 757, "lr": 8.570713477945573e-05}
{"train_loss": 0.2139519900083542, "global_step": 67375, "epoch": 757, "lr": 8.570672896134609e-05}
{"train_loss": 0.3125872015953064, "global_step": 67376, "epoch": 757, "lr": 8.570632313843612e-05}
{"train_loss": 0.24350209534168243, "global_step": 67377, "epoch": 757, "lr": 8.570591731072587e-05}
{"train_loss": 0.18969428539276123, "global_step": 67378, "epoch": 757, "lr": 8.570551147821539e-05}
{"train_loss": 0.17960911989212036, "global_step": 67379, "epoch": 757, "lr": 8.570510564090475e-05}
{"train_loss": 0.2708819508552551, "global_step": 67380, "epoch": 757, "lr": 8.570469979879399e-05}
{"train_loss": 0.15412072837352753, "global_step": 67381, "epoch": 757, "lr": 8.570429395188316e-05}
{"train_loss": 0.3218107521533966, "global_step": 67382, "epoch": 757, "lr": 8.570388810017233e-05}
{"train_loss": 0.24720735847949982, "global_step": 67383, "epoch": 757, "lr": 8.570348224366155e-05}
{"train_loss": 0.21907587349414825, "global_step": 67384, "epoch": 757, "lr": 8.570307638235087e-05}
{"train_loss": 0.2564452886581421, "global_step": 67385, "epoch": 757, "lr": 8.570267051624035e-05}
{"train_loss": 0.21275600790977478, "global_step": 67386, "epoch": 757, "lr": 8.570226464533002e-05}
{"train_loss": 0.15614880621433258, "global_step": 67387, "epoch": 757, "lr": 8.570185876961998e-05}
{"train_loss": 0.19307994842529297, "global_step": 67388, "epoch": 757, "lr": 8.570145288911025e-05}
{"train_loss": 0.25887230038642883, "global_step": 67389, "epoch": 757, "lr": 8.570104700380089e-05}
{"train_loss": 0.1819046288728714, "global_step": 67390, "epoch": 757, "lr": 8.570064111369197e-05}
{"train_loss": 0.12466845661401749, "global_step": 67391, "epoch": 757, "lr": 8.570023521878353e-05}
{"train_loss": 0.26927649974823, "global_step": 67392, "epoch": 757, "lr": 8.569982931907563e-05}
{"train_loss": 0.18325984477996826, "global_step": 67393, "epoch": 757, "lr": 8.569942341456832e-05}
{"train_loss": 0.23091872036457062, "global_step": 67394, "epoch": 757, "lr": 8.569901750526165e-05}
{"train_loss": 0.2260311245918274, "global_step": 67395, "epoch": 757, "lr": 8.569861159115571e-05}
{"train_loss": 0.29192548990249634, "global_step": 67396, "epoch": 757, "lr": 8.56982056722505e-05}
{"train_loss": 0.21535348892211914, "global_step": 67397, "epoch": 757, "lr": 8.569779974854612e-05}
{"train_loss": 0.32448872923851013, "global_step": 67398, "epoch": 757, "lr": 8.569739382004258e-05}
{"train_loss": 0.1421513706445694, "global_step": 67399, "epoch": 757, "lr": 8.569698788673998e-05}
{"train_loss": 0.19507355988025665, "global_step": 67400, "epoch": 757, "lr": 8.569658194863835e-05}
{"train_loss": 0.24363896250724792, "global_step": 67401, "epoch": 757, "lr": 8.569617600573776e-05}
{"train_loss": 0.19234754145145416, "global_step": 67402, "epoch": 757, "lr": 8.569577005803824e-05}
{"train_loss": 0.18187305331230164, "global_step": 67403, "epoch": 757, "lr": 8.569536410553987e-05}
{"train_loss": 0.27590838074684143, "global_step": 67404, "epoch": 757, "lr": 8.569495814824268e-05}
{"train_loss": 0.34073200821876526, "global_step": 67405, "epoch": 757, "lr": 8.569455218614675e-05}
{"train_loss": 0.3248911499977112, "global_step": 67406, "epoch": 757, "lr": 8.569414621925212e-05}
{"train_loss": 0.3014478087425232, "global_step": 67407, "epoch": 757, "lr": 8.569374024755884e-05}
{"train_loss": 0.17233727872371674, "global_step": 67408, "epoch": 757, "lr": 8.569333427106698e-05}
{"train_loss": 0.17758901417255402, "global_step": 67409, "epoch": 757, "lr": 8.569292828977657e-05}
{"train_loss": 0.21139995753765106, "global_step": 67410, "epoch": 757, "lr": 8.569252230368769e-05}
{"train_loss": 0.15944522619247437, "global_step": 67411, "epoch": 757, "lr": 8.569211631280041e-05}
{"train_loss": 0.13575638830661774, "global_step": 67412, "epoch": 757, "lr": 8.569171031711474e-05}
{"train_loss": 0.2109721302986145, "global_step": 67413, "epoch": 757, "lr": 8.569130431663075e-05}
{"train_loss": 0.19577737152576447, "global_step": 67414, "epoch": 757, "lr": 8.56908983113485e-05}
{"train_loss": 0.227477565407753, "global_step": 67415, "epoch": 757, "lr": 8.569049230126804e-05}
{"train_loss": 0.17758862674236298, "global_step": 67416, "epoch": 757, "lr": 8.569008628638944e-05}
{"train_loss": 0.2145994007587433, "global_step": 67417, "epoch": 757, "lr": 8.568968026671274e-05}
{"train_loss": 0.20911657810211182, "global_step": 67418, "epoch": 757, "lr": 8.568927424223798e-05}
{"train_loss": 0.27649879455566406, "global_step": 67419, "epoch": 757, "lr": 8.568886821296525e-05}
{"train_loss": 0.29357627034187317, "global_step": 67420, "epoch": 757, "lr": 8.568846217889459e-05}
{"train_loss": 0.22901037335395813, "global_step": 67421, "epoch": 757, "lr": 8.568805614002604e-05}
{"train_loss": 0.21167270839214325, "global_step": 67422, "epoch": 757, "lr": 8.568765009635969e-05}
{"train_loss": 0.15619520843029022, "global_step": 67423, "epoch": 757, "lr": 8.568724404789555e-05}
{"train_loss": 0.19170773029327393, "global_step": 67424, "epoch": 757, "lr": 8.568683799463368e-05}
{"train_loss": 0.3025137186050415, "global_step": 67425, "epoch": 757, "lr": 8.568643193657417e-05}
{"train_loss": 0.12084930390119553, "global_step": 67426, "epoch": 757, "lr": 8.568602587371706e-05}
{"train_loss": 0.23378242552280426, "global_step": 67427, "epoch": 757, "lr": 8.568561980606238e-05}
{"train_loss": 0.15566261112689972, "global_step": 67428, "epoch": 757, "lr": 8.56852137336102e-05}
{"train_loss": 0.22186294198036194, "global_step": 67429, "epoch": 757, "lr": 8.56848076563606e-05}
{"train_loss": 0.16489920020103455, "global_step": 67430, "epoch": 757, "lr": 8.568440157431361e-05}
{"train_loss": 0.1370653361082077, "global_step": 67431, "epoch": 757, "lr": 8.568399548746929e-05}
{"train_loss": 0.2169734388589859, "global_step": 67432, "epoch": 757, "lr": 8.568358939582766e-05}
{"train_loss": 0.22312207520008087, "global_step": 67433, "epoch": 757, "lr": 8.568318329938884e-05}
{"train_loss": 0.25817641615867615, "global_step": 67434, "epoch": 757, "lr": 8.568277719815282e-05}
{"train_loss": 0.2379360795021057, "global_step": 67435, "epoch": 757, "lr": 8.568237109211972e-05}
{"train_loss": 0.2049221396446228, "global_step": 67436, "epoch": 757, "lr": 8.568196498128955e-05}
{"train_loss": 0.296039342880249, "global_step": 67437, "epoch": 757, "lr": 8.568155886566236e-05}
{"train_loss": 0.1585172712802887, "global_step": 67438, "epoch": 757, "lr": 8.568115274523822e-05}
{"train_loss": 0.2671571969985962, "global_step": 67439, "epoch": 757, "lr": 8.568074662001719e-05}
{"train_loss": 0.2377980351448059, "global_step": 67440, "epoch": 757, "lr": 8.568034048999933e-05}
{"train_loss": 0.19253383576869965, "global_step": 67441, "epoch": 757, "lr": 8.567993435518468e-05}
{"train_loss": 0.13643059134483337, "global_step": 67442, "epoch": 757, "lr": 8.56795282155733e-05}
{"train_loss": 0.2514542043209076, "global_step": 67443, "epoch": 757, "lr": 8.567912207116522e-05}
{"train_loss": 0.20150764286518097, "global_step": 67444, "epoch": 757, "lr": 8.567871592196052e-05}
{"train_loss": 0.20245569944381714, "global_step": 67445, "epoch": 757, "lr": 8.567830976795928e-05}
{"train_loss": 0.18723808228969574, "global_step": 67446, "epoch": 757, "lr": 8.567790360916151e-05}
{"train_loss": 0.20928622782230377, "global_step": 67447, "epoch": 757, "lr": 8.567749744556728e-05}
{"train_loss": 0.13293632864952087, "global_step": 67448, "epoch": 757, "lr": 8.567709127717664e-05}
{"train_loss": 0.17855966091156006, "global_step": 67449, "epoch": 757, "lr": 8.567668510398965e-05}
{"train_loss": 0.20675040781497955, "global_step": 67450, "epoch": 757, "lr": 8.567627892600637e-05}
{"train_loss": 0.2662385404109955, "global_step": 67451, "epoch": 757, "lr": 8.567587274322686e-05}
{"train_loss": 0.22334805130958557, "global_step": 67452, "epoch": 757, "lr": 8.567546655565114e-05}
{"train_loss": 0.24326086044311523, "global_step": 67453, "epoch": 757, "lr": 8.56750603632793e-05}
{"train_loss": 0.17861507833003998, "global_step": 67454, "epoch": 757, "lr": 8.567465416611138e-05}
{"train_loss": 0.15428246557712555, "global_step": 67455, "epoch": 757, "lr": 8.567424796414745e-05}
{"train_loss": 0.31059321761131287, "global_step": 67456, "epoch": 757, "lr": 8.567384175738753e-05}
{"train_loss": 0.22925609350204468, "global_step": 67457, "epoch": 757, "lr": 8.567343554583172e-05}
{"train_loss": 0.15258432924747467, "global_step": 67458, "epoch": 757, "lr": 8.567302932948004e-05}
{"train_loss": 0.2459169328212738, "global_step": 67459, "epoch": 757, "lr": 8.567262310833255e-05}
{"train_loss": 0.27391061186790466, "global_step": 67460, "epoch": 757, "lr": 8.567221688238933e-05}
{"train_loss": 0.21728912431202577, "global_step": 67461, "epoch": 757, "lr": 8.56718106516504e-05, "val_loss": 3.6704447269439697}
{"train_loss": 0.15710175037384033, "global_step": 67462, "epoch": 758, "lr": 8.567140441611583e-05}
{"train_loss": 0.24159377813339233, "global_step": 67463, "epoch": 758, "lr": 8.567099817578568e-05}
{"train_loss": 0.13365162909030914, "global_step": 67464, "epoch": 758, "lr": 8.567059193065999e-05}
{"train_loss": 0.16770514845848083, "global_step": 67465, "epoch": 758, "lr": 8.567018568073884e-05}
{"train_loss": 0.14231672883033752, "global_step": 67466, "epoch": 758, "lr": 8.566977942602225e-05}
{"train_loss": 0.19646401703357697, "global_step": 67467, "epoch": 758, "lr": 8.56693731665103e-05}
{"train_loss": 0.2352531999349594, "global_step": 67468, "epoch": 758, "lr": 8.566896690220305e-05}
{"train_loss": 0.18332625925540924, "global_step": 67469, "epoch": 758, "lr": 8.566856063310051e-05}
{"train_loss": 0.21058262884616852, "global_step": 67470, "epoch": 758, "lr": 8.56681543592028e-05}
{"train_loss": 0.22497375309467316, "global_step": 67471, "epoch": 758, "lr": 8.566774808050992e-05}
{"train_loss": 0.22481505572795868, "global_step": 67472, "epoch": 758, "lr": 8.566734179702196e-05}
{"train_loss": 0.22869069874286652, "global_step": 67473, "epoch": 758, "lr": 8.566693550873896e-05}
{"train_loss": 0.24607008695602417, "global_step": 67474, "epoch": 758, "lr": 8.566652921566096e-05}
{"train_loss": 0.16754429042339325, "global_step": 67475, "epoch": 758, "lr": 8.566612291778805e-05}
{"train_loss": 0.20168474316596985, "global_step": 67476, "epoch": 758, "lr": 8.566571661512026e-05}
{"train_loss": 0.22615621984004974, "global_step": 67477, "epoch": 758, "lr": 8.566531030765764e-05}
{"train_loss": 0.1719307005405426, "global_step": 67478, "epoch": 758, "lr": 8.566490399540026e-05}
{"train_loss": 0.1498357206583023, "global_step": 67479, "epoch": 758, "lr": 8.566449767834817e-05}
{"train_loss": 0.19905126094818115, "global_step": 67480, "epoch": 758, "lr": 8.566409135650142e-05}
{"train_loss": 0.17463910579681396, "global_step": 67481, "epoch": 758, "lr": 8.566368502986006e-05}
{"train_loss": 0.17429091036319733, "global_step": 67482, "epoch": 758, "lr": 8.566327869842417e-05}
{"train_loss": 0.17030000686645508, "global_step": 67483, "epoch": 758, "lr": 8.566287236219377e-05}
{"train_loss": 0.19510306417942047, "global_step": 67484, "epoch": 758, "lr": 8.566246602116893e-05}
{"train_loss": 0.26334497332572937, "global_step": 67485, "epoch": 758, "lr": 8.566205967534971e-05}
{"train_loss": 0.2533591389656067, "global_step": 67486, "epoch": 758, "lr": 8.566165332473617e-05}
{"train_loss": 0.37039434909820557, "global_step": 67487, "epoch": 758, "lr": 8.566124696932836e-05}
{"train_loss": 0.14337745308876038, "global_step": 67488, "epoch": 758, "lr": 8.566084060912631e-05}
{"train_loss": 0.3136690855026245, "global_step": 67489, "epoch": 758, "lr": 8.56604342441301e-05}
{"train_loss": 0.2206067442893982, "global_step": 67490, "epoch": 758, "lr": 8.566002787433979e-05}
{"train_loss": 0.2135821133852005, "global_step": 67491, "epoch": 758, "lr": 8.565962149975542e-05}
{"train_loss": 0.28281575441360474, "global_step": 67492, "epoch": 758, "lr": 8.565921512037705e-05}
{"train_loss": 0.21110105514526367, "global_step": 67493, "epoch": 758, "lr": 8.565880873620473e-05}
{"train_loss": 0.3078159689903259, "global_step": 67494, "epoch": 758, "lr": 8.56584023472385e-05}
{"train_loss": 0.2852417528629303, "global_step": 67495, "epoch": 758, "lr": 8.565799595347846e-05}
{"train_loss": 0.1655811369419098, "global_step": 67496, "epoch": 758, "lr": 8.565758955492462e-05}
{"train_loss": 0.22367945313453674, "global_step": 67497, "epoch": 758, "lr": 8.565718315157706e-05}
{"train_loss": 0.19521352648735046, "global_step": 67498, "epoch": 758, "lr": 8.565677674343581e-05}
{"train_loss": 0.3625360429286957, "global_step": 67499, "epoch": 758, "lr": 8.565637033050097e-05}
{"train_loss": 0.22915750741958618, "global_step": 67500, "epoch": 758, "lr": 8.565596391277255e-05}
{"train_loss": 0.28606098890304565, "global_step": 67501, "epoch": 758, "lr": 8.565555749025062e-05}
{"train_loss": 0.25701266527175903, "global_step": 67502, "epoch": 758, "lr": 8.565515106293525e-05}
{"train_loss": 0.2867518663406372, "global_step": 67503, "epoch": 758, "lr": 8.565474463082646e-05}
{"train_loss": 0.22237633168697357, "global_step": 67504, "epoch": 758, "lr": 8.565433819392433e-05}
{"train_loss": 0.16371391713619232, "global_step": 67505, "epoch": 758, "lr": 8.56539317522289e-05}
{"train_loss": 0.18659383058547974, "global_step": 67506, "epoch": 758, "lr": 8.565352530574025e-05}
{"train_loss": 0.29325321316719055, "global_step": 67507, "epoch": 758, "lr": 8.565311885445841e-05}
{"train_loss": 0.2558286190032959, "global_step": 67508, "epoch": 758, "lr": 8.565271239838343e-05}
{"train_loss": 0.17315243184566498, "global_step": 67509, "epoch": 758, "lr": 8.565230593751539e-05}
{"train_loss": 0.17980526387691498, "global_step": 67510, "epoch": 758, "lr": 8.565189947185435e-05}
{"train_loss": 0.28114965558052063, "global_step": 67511, "epoch": 758, "lr": 8.565149300140032e-05}
{"train_loss": 0.1849910318851471, "global_step": 67512, "epoch": 758, "lr": 8.565108652615339e-05}
{"train_loss": 0.22455061972141266, "global_step": 67513, "epoch": 758, "lr": 8.565068004611361e-05}
{"train_loss": 0.14235776662826538, "global_step": 67514, "epoch": 758, "lr": 8.565027356128102e-05}
{"train_loss": 0.21698534488677979, "global_step": 67515, "epoch": 758, "lr": 8.564986707165569e-05}
{"train_loss": 0.203873410820961, "global_step": 67516, "epoch": 758, "lr": 8.564946057723768e-05}
{"train_loss": 0.25121939182281494, "global_step": 67517, "epoch": 758, "lr": 8.564905407802702e-05}
{"train_loss": 0.3506636917591095, "global_step": 67518, "epoch": 758, "lr": 8.564864757402379e-05}
{"train_loss": 0.14719156920909882, "global_step": 67519, "epoch": 758, "lr": 8.564824106522802e-05}
{"train_loss": 0.2312667965888977, "global_step": 67520, "epoch": 758, "lr": 8.564783455163979e-05}
{"train_loss": 0.2300994098186493, "global_step": 67521, "epoch": 758, "lr": 8.564742803325913e-05}
{"train_loss": 0.17569680511951447, "global_step": 67522, "epoch": 758, "lr": 8.564702151008614e-05}
{"train_loss": 0.22479678690433502, "global_step": 67523, "epoch": 758, "lr": 8.564661498212081e-05}
{"train_loss": 0.1717355102300644, "global_step": 67524, "epoch": 758, "lr": 8.564620844936324e-05}
{"train_loss": 0.22406849265098572, "global_step": 67525, "epoch": 758, "lr": 8.564580191181347e-05}
{"train_loss": 0.20600025355815887, "global_step": 67526, "epoch": 758, "lr": 8.564539536947156e-05}
{"train_loss": 0.2170787900686264, "global_step": 67527, "epoch": 758, "lr": 8.564498882233756e-05}
{"train_loss": 0.18506449460983276, "global_step": 67528, "epoch": 758, "lr": 8.564458227041153e-05}
{"train_loss": 0.2549905478954315, "global_step": 67529, "epoch": 758, "lr": 8.56441757136935e-05}
{"train_loss": 0.26165440678596497, "global_step": 67530, "epoch": 758, "lr": 8.564376915218356e-05}
{"train_loss": 0.14153696596622467, "global_step": 67531, "epoch": 758, "lr": 8.564336258588175e-05}
{"train_loss": 0.23017363250255585, "global_step": 67532, "epoch": 758, "lr": 8.564295601478812e-05}
{"train_loss": 0.1744898557662964, "global_step": 67533, "epoch": 758, "lr": 8.564254943890273e-05}
{"train_loss": 0.26094019412994385, "global_step": 67534, "epoch": 758, "lr": 8.564214285822563e-05}
{"train_loss": 0.24209409952163696, "global_step": 67535, "epoch": 758, "lr": 8.564173627275689e-05}
{"train_loss": 0.1820329874753952, "global_step": 67536, "epoch": 758, "lr": 8.564132968249654e-05}
{"train_loss": 0.13368169963359833, "global_step": 67537, "epoch": 758, "lr": 8.564092308744467e-05}
{"train_loss": 0.2605566084384918, "global_step": 67538, "epoch": 758, "lr": 8.564051648760129e-05}
{"train_loss": 0.19098304212093353, "global_step": 67539, "epoch": 758, "lr": 8.564010988296648e-05}
{"train_loss": 0.14620164036750793, "global_step": 67540, "epoch": 758, "lr": 8.56397032735403e-05}
{"train_loss": 0.23697537183761597, "global_step": 67541, "epoch": 758, "lr": 8.563929665932278e-05}
{"train_loss": 0.30498403310775757, "global_step": 67542, "epoch": 758, "lr": 8.5638890040314e-05}
{"train_loss": 0.23749595880508423, "global_step": 67543, "epoch": 758, "lr": 8.563848341651401e-05}
{"train_loss": 0.2654244899749756, "global_step": 67544, "epoch": 758, "lr": 8.563807678792286e-05}
{"train_loss": 0.1727108210325241, "global_step": 67545, "epoch": 758, "lr": 8.56376701545406e-05}
{"train_loss": 0.2389710545539856, "global_step": 67546, "epoch": 758, "lr": 8.563726351636729e-05}
{"train_loss": 0.2554413974285126, "global_step": 67547, "epoch": 758, "lr": 8.5636856873403e-05}
{"train_loss": 0.21255619823932648, "global_step": 67548, "epoch": 758, "lr": 8.563645022564774e-05}
{"train_loss": 0.185080423951149, "global_step": 67549, "epoch": 758, "lr": 8.563604357310162e-05}
{"train_loss": 0.21808232099152683, "global_step": 67550, "epoch": 758, "lr": 8.563563691576467e-05, "val_loss": 3.651700735092163}
{"train_loss": 0.20055662095546722, "global_step": 67551, "epoch": 759, "lr": 8.563523025363692e-05}
{"train_loss": 0.2054852545261383, "global_step": 67552, "epoch": 759, "lr": 8.563482358671846e-05}
{"train_loss": 0.2750145196914673, "global_step": 67553, "epoch": 759, "lr": 8.563441691500933e-05}
{"train_loss": 0.23135356605052948, "global_step": 67554, "epoch": 759, "lr": 8.563401023850958e-05}
{"train_loss": 0.21397556364536285, "global_step": 67555, "epoch": 759, "lr": 8.563360355721928e-05}
{"train_loss": 0.1760248839855194, "global_step": 67556, "epoch": 759, "lr": 8.563319687113849e-05}
{"train_loss": 0.25891387462615967, "global_step": 67557, "epoch": 759, "lr": 8.563279018026723e-05}
{"train_loss": 0.27007216215133667, "global_step": 67558, "epoch": 759, "lr": 8.563238348460558e-05}
{"train_loss": 0.2496010661125183, "global_step": 67559, "epoch": 759, "lr": 8.56319767841536e-05}
{"train_loss": 0.16111324727535248, "global_step": 67560, "epoch": 759, "lr": 8.563157007891132e-05}
{"train_loss": 0.1397985816001892, "global_step": 67561, "epoch": 759, "lr": 8.563116336887882e-05}
{"train_loss": 0.19277258217334747, "global_step": 67562, "epoch": 759, "lr": 8.563075665405613e-05}
{"train_loss": 0.23015841841697693, "global_step": 67563, "epoch": 759, "lr": 8.563034993444334e-05}
{"train_loss": 0.14625221490859985, "global_step": 67564, "epoch": 759, "lr": 8.562994321004047e-05}
{"train_loss": 0.19823342561721802, "global_step": 67565, "epoch": 759, "lr": 8.56295364808476e-05}
{"train_loss": 0.2672395408153534, "global_step": 67566, "epoch": 759, "lr": 8.562912974686477e-05}
{"train_loss": 0.21550866961479187, "global_step": 67567, "epoch": 759, "lr": 8.562872300809202e-05}
{"train_loss": 0.19923464953899384, "global_step": 67568, "epoch": 759, "lr": 8.562831626452945e-05}
{"train_loss": 0.19433631002902985, "global_step": 67569, "epoch": 759, "lr": 8.562790951617707e-05}
{"train_loss": 0.1823337972164154, "global_step": 67570, "epoch": 759, "lr": 8.562750276303495e-05}
{"train_loss": 0.24134258925914764, "global_step": 67571, "epoch": 759, "lr": 8.562709600510314e-05}
{"train_loss": 0.19691146910190582, "global_step": 67572, "epoch": 759, "lr": 8.562668924238172e-05}
{"train_loss": 0.1276092529296875, "global_step": 67573, "epoch": 759, "lr": 8.562628247487072e-05}
{"train_loss": 0.2704767882823944, "global_step": 67574, "epoch": 759, "lr": 8.56258757025702e-05}
{"train_loss": 0.1458563208580017, "global_step": 67575, "epoch": 759, "lr": 8.562546892548023e-05}
{"train_loss": 0.21077534556388855, "global_step": 67576, "epoch": 759, "lr": 8.562506214360082e-05}
{"train_loss": 0.24003273248672485, "global_step": 67577, "epoch": 759, "lr": 8.562465535693207e-05}
{"train_loss": 0.1849871277809143, "global_step": 67578, "epoch": 759, "lr": 8.562424856547402e-05}
{"train_loss": 0.18172238767147064, "global_step": 67579, "epoch": 759, "lr": 8.562384176922671e-05}
{"train_loss": 0.30704545974731445, "global_step": 67580, "epoch": 759, "lr": 8.562343496819023e-05}
{"train_loss": 0.25131893157958984, "global_step": 67581, "epoch": 759, "lr": 8.56230281623646e-05}
{"train_loss": 0.16761277616024017, "global_step": 67582, "epoch": 759, "lr": 8.562262135174989e-05}
{"train_loss": 0.2050556093454361, "global_step": 67583, "epoch": 759, "lr": 8.562221453634615e-05}
{"train_loss": 0.24171333014965057, "global_step": 67584, "epoch": 759, "lr": 8.562180771615344e-05}
{"train_loss": 0.2111959159374237, "global_step": 67585, "epoch": 759, "lr": 8.562140089117183e-05}
{"train_loss": 0.19726422429084778, "global_step": 67586, "epoch": 759, "lr": 8.562099406140133e-05}
{"train_loss": 0.20862601697444916, "global_step": 67587, "epoch": 759, "lr": 8.562058722684205e-05}
{"train_loss": 0.24149134755134583, "global_step": 67588, "epoch": 759, "lr": 8.562018038749399e-05}
{"train_loss": 0.2188149392604828, "global_step": 67589, "epoch": 759, "lr": 8.561977354335724e-05}
{"train_loss": 0.15662319958209991, "global_step": 67590, "epoch": 759, "lr": 8.561936669443184e-05}
{"train_loss": 0.2206718474626541, "global_step": 67591, "epoch": 759, "lr": 8.561895984071786e-05}
{"train_loss": 0.27964070439338684, "global_step": 67592, "epoch": 759, "lr": 8.561855298221534e-05}
{"train_loss": 0.1984938532114029, "global_step": 67593, "epoch": 759, "lr": 8.561814611892434e-05}
{"train_loss": 0.20622795820236206, "global_step": 67594, "epoch": 759, "lr": 8.561773925084492e-05}
{"train_loss": 0.2559252977371216, "global_step": 67595, "epoch": 759, "lr": 8.561733237797713e-05}
{"train_loss": 0.26542338728904724, "global_step": 67596, "epoch": 759, "lr": 8.561692550032101e-05}
{"train_loss": 0.22307313978672028, "global_step": 67597, "epoch": 759, "lr": 8.561651861787663e-05}
{"train_loss": 0.15964144468307495, "global_step": 67598, "epoch": 759, "lr": 8.561611173064405e-05}
{"train_loss": 0.1843891739845276, "global_step": 67599, "epoch": 759, "lr": 8.561570483862331e-05}
{"train_loss": 0.19421784579753876, "global_step": 67600, "epoch": 759, "lr": 8.561529794181449e-05}
{"train_loss": 0.20970462262630463, "global_step": 67601, "epoch": 759, "lr": 8.561489104021761e-05}
{"train_loss": 0.17647862434387207, "global_step": 67602, "epoch": 759, "lr": 8.561448413383275e-05}
{"train_loss": 0.1629999577999115, "global_step": 67603, "epoch": 759, "lr": 8.561407722265996e-05}
{"train_loss": 0.18171121180057526, "global_step": 67604, "epoch": 759, "lr": 8.561367030669929e-05}
{"train_loss": 0.2795574963092804, "global_step": 67605, "epoch": 759, "lr": 8.56132633859508e-05}
{"train_loss": 0.29391995072364807, "global_step": 67606, "epoch": 759, "lr": 8.561285646041453e-05}
{"train_loss": 0.1364665925502777, "global_step": 67607, "epoch": 759, "lr": 8.561244953009054e-05}
{"train_loss": 0.18931226432323456, "global_step": 67608, "epoch": 759, "lr": 8.56120425949789e-05}
{"train_loss": 0.19688479602336884, "global_step": 67609, "epoch": 759, "lr": 8.561163565507966e-05}
{"train_loss": 0.2741057574748993, "global_step": 67610, "epoch": 759, "lr": 8.561122871039286e-05}
{"train_loss": 0.26940855383872986, "global_step": 67611, "epoch": 759, "lr": 8.561082176091856e-05}
{"train_loss": 0.2019423246383667, "global_step": 67612, "epoch": 759, "lr": 8.561041480665684e-05}
{"train_loss": 0.2055261880159378, "global_step": 67613, "epoch": 759, "lr": 8.561000784760772e-05}
{"train_loss": 0.2847777009010315, "global_step": 67614, "epoch": 759, "lr": 8.560960088377127e-05}
{"train_loss": 0.18922221660614014, "global_step": 67615, "epoch": 759, "lr": 8.560919391514754e-05}
{"train_loss": 0.17416132986545563, "global_step": 67616, "epoch": 759, "lr": 8.56087869417366e-05}
{"train_loss": 0.25081971287727356, "global_step": 67617, "epoch": 759, "lr": 8.560837996353848e-05}
{"train_loss": 0.17994724214076996, "global_step": 67618, "epoch": 759, "lr": 8.560797298055326e-05}
{"train_loss": 0.2517958879470825, "global_step": 67619, "epoch": 759, "lr": 8.560756599278096e-05}
{"train_loss": 0.20872558653354645, "global_step": 67620, "epoch": 759, "lr": 8.560715900022168e-05}
{"train_loss": 0.14372876286506653, "global_step": 67621, "epoch": 759, "lr": 8.560675200287542e-05}
{"train_loss": 0.2013692855834961, "global_step": 67622, "epoch": 759, "lr": 8.56063450007423e-05}
{"train_loss": 0.17217721045017242, "global_step": 67623, "epoch": 759, "lr": 8.560593799382232e-05}
{"train_loss": 0.24079695343971252, "global_step": 67624, "epoch": 759, "lr": 8.560553098211557e-05}
{"train_loss": 0.1852622777223587, "global_step": 67625, "epoch": 759, "lr": 8.560512396562207e-05}
{"train_loss": 0.15055663883686066, "global_step": 67626, "epoch": 759, "lr": 8.560471694434192e-05}
{"train_loss": 0.15201765298843384, "global_step": 67627, "epoch": 759, "lr": 8.560430991827514e-05}
{"train_loss": 0.2674287259578705, "global_step": 67628, "epoch": 759, "lr": 8.56039028874218e-05}
{"train_loss": 0.22884193062782288, "global_step": 67629, "epoch": 759, "lr": 8.560349585178193e-05}
{"train_loss": 0.17661991715431213, "global_step": 67630, "epoch": 759, "lr": 8.560308881135562e-05}
{"train_loss": 0.25690892338752747, "global_step": 67631, "epoch": 759, "lr": 8.560268176614291e-05}
{"train_loss": 0.17787550389766693, "global_step": 67632, "epoch": 759, "lr": 8.560227471614385e-05}
{"train_loss": 0.16087500751018524, "global_step": 67633, "epoch": 759, "lr": 8.560186766135849e-05}
{"train_loss": 0.23843099176883698, "global_step": 67634, "epoch": 759, "lr": 8.56014606017869e-05}
{"train_loss": 0.20456722378730774, "global_step": 67635, "epoch": 759, "lr": 8.560105353742914e-05}
{"train_loss": 0.2677062451839447, "global_step": 67636, "epoch": 759, "lr": 8.560064646828523e-05}
{"train_loss": 0.20113469660282135, "global_step": 67637, "epoch": 759, "lr": 8.560023939435526e-05}
{"train_loss": 0.22210010886192322, "global_step": 67638, "epoch": 759, "lr": 8.559983231563927e-05}
{"train_loss": 0.21152945482329036, "global_step": 67639, "epoch": 759, "lr": 8.559942523213731e-05, "val_loss": 3.6629481315612793}
{"train_loss": 0.2089107781648636, "global_step": 67640, "epoch": 760, "lr": 8.559901814384945e-05}
{"train_loss": 0.24333663284778595, "global_step": 67641, "epoch": 760, "lr": 8.559861105077572e-05}
{"train_loss": 0.17463761568069458, "global_step": 67642, "epoch": 760, "lr": 8.559820395291622e-05}
{"train_loss": 0.24303977191448212, "global_step": 67643, "epoch": 760, "lr": 8.559779685027094e-05}
{"train_loss": 0.24828660488128662, "global_step": 67644, "epoch": 760, "lr": 8.559738974284e-05}
{"train_loss": 0.19362236559391022, "global_step": 67645, "epoch": 760, "lr": 8.55969826306234e-05}
{"train_loss": 0.16573522984981537, "global_step": 67646, "epoch": 760, "lr": 8.559657551362124e-05}
{"train_loss": 0.24055399000644684, "global_step": 67647, "epoch": 760, "lr": 8.559616839183354e-05}
{"train_loss": 0.26234230399131775, "global_step": 67648, "epoch": 760, "lr": 8.559576126526038e-05}
{"train_loss": 0.3049476444721222, "global_step": 67649, "epoch": 760, "lr": 8.55953541339018e-05}
{"train_loss": 0.1327873319387436, "global_step": 67650, "epoch": 760, "lr": 8.559494699775785e-05}
{"train_loss": 0.23401778936386108, "global_step": 67651, "epoch": 760, "lr": 8.559453985682861e-05}
{"train_loss": 0.23065721988677979, "global_step": 67652, "epoch": 760, "lr": 8.55941327111141e-05}
{"train_loss": 0.1721200942993164, "global_step": 67653, "epoch": 760, "lr": 8.55937255606144e-05}
{"train_loss": 0.1278974711894989, "global_step": 67654, "epoch": 760, "lr": 8.559331840532956e-05}
{"train_loss": 0.190001979470253, "global_step": 67655, "epoch": 760, "lr": 8.559291124525963e-05}
{"train_loss": 0.16354885697364807, "global_step": 67656, "epoch": 760, "lr": 8.559250408040467e-05}
{"train_loss": 0.21279200911521912, "global_step": 67657, "epoch": 760, "lr": 8.559209691076474e-05}
{"train_loss": 0.18125979602336884, "global_step": 67658, "epoch": 760, "lr": 8.559168973633986e-05}
{"train_loss": 0.128227099776268, "global_step": 67659, "epoch": 760, "lr": 8.559128255713013e-05}
{"train_loss": 0.22084733843803406, "global_step": 67660, "epoch": 760, "lr": 8.559087537313559e-05}
{"train_loss": 0.16441820561885834, "global_step": 67661, "epoch": 760, "lr": 8.559046818435627e-05}
{"train_loss": 0.28199899196624756, "global_step": 67662, "epoch": 760, "lr": 8.559006099079228e-05}
{"train_loss": 0.1725883036851883, "global_step": 67663, "epoch": 760, "lr": 8.558965379244361e-05}
{"train_loss": 0.1868390589952469, "global_step": 67664, "epoch": 760, "lr": 8.558924658931036e-05}
{"train_loss": 0.17858122289180756, "global_step": 67665, "epoch": 760, "lr": 8.558883938139255e-05}
{"train_loss": 0.25122693181037903, "global_step": 67666, "epoch": 760, "lr": 8.558843216869027e-05}
{"train_loss": 0.27197423577308655, "global_step": 67667, "epoch": 760, "lr": 8.558802495120356e-05}
{"train_loss": 0.21375112235546112, "global_step": 67668, "epoch": 760, "lr": 8.558761772893246e-05}
{"train_loss": 0.2022099792957306, "global_step": 67669, "epoch": 760, "lr": 8.558721050187706e-05}
{"train_loss": 0.1898787021636963, "global_step": 67670, "epoch": 760, "lr": 8.558680327003739e-05}
{"train_loss": 0.1829906404018402, "global_step": 67671, "epoch": 760, "lr": 8.558639603341349e-05}
{"train_loss": 0.18529640138149261, "global_step": 67672, "epoch": 760, "lr": 8.558598879200544e-05}
{"train_loss": 0.2255878895521164, "global_step": 67673, "epoch": 760, "lr": 8.55855815458133e-05}
{"train_loss": 0.2776336073875427, "global_step": 67674, "epoch": 760, "lr": 8.55851742948371e-05}
{"train_loss": 0.2237444519996643, "global_step": 67675, "epoch": 760, "lr": 8.558476703907691e-05}
{"train_loss": 0.1733759492635727, "global_step": 67676, "epoch": 760, "lr": 8.558435977853278e-05}
{"train_loss": 0.19107389450073242, "global_step": 67677, "epoch": 760, "lr": 8.558395251320477e-05}
{"train_loss": 0.23395243287086487, "global_step": 67678, "epoch": 760, "lr": 8.558354524309293e-05}
{"train_loss": 0.24794608354568481, "global_step": 67679, "epoch": 760, "lr": 8.558313796819732e-05}
{"train_loss": 0.23848719894886017, "global_step": 67680, "epoch": 760, "lr": 8.558273068851799e-05}
{"train_loss": 0.14321279525756836, "global_step": 67681, "epoch": 760, "lr": 8.558232340405498e-05}
{"train_loss": 0.1937428116798401, "global_step": 67682, "epoch": 760, "lr": 8.558191611480837e-05}
{"train_loss": 0.23787781596183777, "global_step": 67683, "epoch": 760, "lr": 8.558150882077821e-05}
{"train_loss": 0.2706514298915863, "global_step": 67684, "epoch": 760, "lr": 8.558110152196455e-05}
{"train_loss": 0.2949611246585846, "global_step": 67685, "epoch": 760, "lr": 8.558069421836745e-05}
{"train_loss": 0.14826294779777527, "global_step": 67686, "epoch": 760, "lr": 8.558028690998694e-05}
{"train_loss": 0.24017490446567535, "global_step": 67687, "epoch": 760, "lr": 8.557987959682311e-05}
{"train_loss": 0.23427020013332367, "global_step": 67688, "epoch": 760, "lr": 8.557947227887599e-05}
{"train_loss": 0.2815733551979065, "global_step": 67689, "epoch": 760, "lr": 8.557906495614565e-05}
{"train_loss": 0.23491711914539337, "global_step": 67690, "epoch": 760, "lr": 8.557865762863213e-05}
{"train_loss": 0.27701371908187866, "global_step": 67691, "epoch": 760, "lr": 8.557825029633551e-05}
{"train_loss": 0.3321468234062195, "global_step": 67692, "epoch": 760, "lr": 8.557784295925582e-05}
{"train_loss": 0.23391881585121155, "global_step": 67693, "epoch": 760, "lr": 8.557743561739311e-05}
{"train_loss": 0.19984781742095947, "global_step": 67694, "epoch": 760, "lr": 8.557702827074746e-05}
{"train_loss": 0.24812401831150055, "global_step": 67695, "epoch": 760, "lr": 8.55766209193189e-05}
{"train_loss": 0.18883278965950012, "global_step": 67696, "epoch": 760, "lr": 8.557621356310751e-05}
{"train_loss": 0.17661666870117188, "global_step": 67697, "epoch": 760, "lr": 8.557580620211333e-05}
{"train_loss": 0.23624655604362488, "global_step": 67698, "epoch": 760, "lr": 8.557539883633641e-05}
{"train_loss": 0.2745141088962555, "global_step": 67699, "epoch": 760, "lr": 8.557499146577682e-05}
{"train_loss": 0.2250375747680664, "global_step": 67700, "epoch": 760, "lr": 8.557458409043459e-05}
{"train_loss": 0.19572418928146362, "global_step": 67701, "epoch": 760, "lr": 8.557417671030981e-05}
{"train_loss": 0.21657542884349823, "global_step": 67702, "epoch": 760, "lr": 8.557376932540251e-05}
{"train_loss": 0.20847348868846893, "global_step": 67703, "epoch": 760, "lr": 8.557336193571275e-05}
{"train_loss": 0.202745258808136, "global_step": 67704, "epoch": 760, "lr": 8.557295454124059e-05}
{"train_loss": 0.2279449850320816, "global_step": 67705, "epoch": 760, "lr": 8.557254714198607e-05}
{"train_loss": 0.2732764482498169, "global_step": 67706, "epoch": 760, "lr": 8.557213973794925e-05}
{"train_loss": 0.1697501242160797, "global_step": 67707, "epoch": 760, "lr": 8.557173232913021e-05}
{"train_loss": 0.15023210644721985, "global_step": 67708, "epoch": 760, "lr": 8.557132491552897e-05}
{"train_loss": 0.2502095699310303, "global_step": 67709, "epoch": 760, "lr": 8.557091749714558e-05}
{"train_loss": 0.2548390030860901, "global_step": 67710, "epoch": 760, "lr": 8.557051007398015e-05}
{"train_loss": 0.23913106322288513, "global_step": 67711, "epoch": 760, "lr": 8.557010264603267e-05}
{"train_loss": 0.2738756239414215, "global_step": 67712, "epoch": 760, "lr": 8.556969521330325e-05}
{"train_loss": 0.1908746361732483, "global_step": 67713, "epoch": 760, "lr": 8.556928777579189e-05}
{"train_loss": 0.30999672412872314, "global_step": 67714, "epoch": 760, "lr": 8.55688803334987e-05}
{"train_loss": 0.24073919653892517, "global_step": 67715, "epoch": 760, "lr": 8.55684728864237e-05}
{"train_loss": 0.2705766558647156, "global_step": 67716, "epoch": 760, "lr": 8.556806543456694e-05}
{"train_loss": 0.23317280411720276, "global_step": 67717, "epoch": 760, "lr": 8.556765797792849e-05}
{"train_loss": 0.21182861924171448, "global_step": 67718, "epoch": 760, "lr": 8.55672505165084e-05}
{"train_loss": 0.29554712772369385, "global_step": 67719, "epoch": 760, "lr": 8.556684305030673e-05}
{"train_loss": 0.2359876185655594, "global_step": 67720, "epoch": 760, "lr": 8.556643557932353e-05}
{"train_loss": 0.24484196305274963, "global_step": 67721, "epoch": 760, "lr": 8.556602810355887e-05}
{"train_loss": 0.28196680545806885, "global_step": 67722, "epoch": 760, "lr": 8.556562062301278e-05}
{"train_loss": 0.2154913991689682, "global_step": 67723, "epoch": 760, "lr": 8.556521313768533e-05}
{"train_loss": 0.23072029650211334, "global_step": 67724, "epoch": 760, "lr": 8.556480564757656e-05}
{"train_loss": 0.27255865931510925, "global_step": 67725, "epoch": 760, "lr": 8.556439815268654e-05}
{"train_loss": 0.24440617859363556, "global_step": 67726, "epoch": 760, "lr": 8.556399065301533e-05}
{"train_loss": 0.31688275933265686, "global_step": 67727, "epoch": 760, "lr": 8.556358314856296e-05}
{"train_loss": 0.2235029542379165, "global_step": 67728, "epoch": 760, "lr": 8.556317563932952e-05, "val_loss": 3.555842876434326, "train_action_mse_error": 24.671146392822266}
{"train_loss": 0.2568259537220001, "global_step": 67729, "epoch": 761, "lr": 8.556276812531503e-05}
{"train_loss": 0.2588878273963928, "global_step": 67730, "epoch": 761, "lr": 8.556236060651956e-05}
{"train_loss": 0.21817143261432648, "global_step": 67731, "epoch": 761, "lr": 8.556195308294316e-05}
{"train_loss": 0.30914637446403503, "global_step": 67732, "epoch": 761, "lr": 8.556154555458588e-05}
{"train_loss": 0.28791627287864685, "global_step": 67733, "epoch": 761, "lr": 8.55611380214478e-05}
{"train_loss": 0.2482212483882904, "global_step": 67734, "epoch": 761, "lr": 8.556073048352896e-05}
{"train_loss": 0.18326249718666077, "global_step": 67735, "epoch": 761, "lr": 8.55603229408294e-05}
{"train_loss": 0.2287273406982422, "global_step": 67736, "epoch": 761, "lr": 8.555991539334921e-05}
{"train_loss": 0.23507185280323029, "global_step": 67737, "epoch": 761, "lr": 8.555950784108842e-05}
{"train_loss": 0.18003696203231812, "global_step": 67738, "epoch": 761, "lr": 8.555910028404707e-05}
{"train_loss": 0.22620819509029388, "global_step": 67739, "epoch": 761, "lr": 8.555869272222523e-05}
{"train_loss": 0.3279612362384796, "global_step": 67740, "epoch": 761, "lr": 8.555828515562297e-05}
{"train_loss": 0.187771275639534, "global_step": 67741, "epoch": 761, "lr": 8.555787758424032e-05}
{"train_loss": 0.2446710169315338, "global_step": 67742, "epoch": 761, "lr": 8.555747000807737e-05}
{"train_loss": 0.338647723197937, "global_step": 67743, "epoch": 761, "lr": 8.555706242713412e-05}
{"train_loss": 0.2139286994934082, "global_step": 67744, "epoch": 761, "lr": 8.555665484141069e-05}
{"train_loss": 0.18409667909145355, "global_step": 67745, "epoch": 761, "lr": 8.555624725090709e-05}
{"train_loss": 0.2756388485431671, "global_step": 67746, "epoch": 761, "lr": 8.555583965562337e-05}
{"train_loss": 0.29755985736846924, "global_step": 67747, "epoch": 761, "lr": 8.555543205555961e-05}
{"train_loss": 0.21610459685325623, "global_step": 67748, "epoch": 761, "lr": 8.555502445071586e-05}
{"train_loss": 0.2394467145204544, "global_step": 67749, "epoch": 761, "lr": 8.555461684109218e-05}
{"train_loss": 0.2811276614665985, "global_step": 67750, "epoch": 761, "lr": 8.55542092266886e-05}
{"train_loss": 0.14830651879310608, "global_step": 67751, "epoch": 761, "lr": 8.555380160750519e-05}
{"train_loss": 0.16725122928619385, "global_step": 67752, "epoch": 761, "lr": 8.5553393983542e-05}
{"train_loss": 0.3541189134120941, "global_step": 67753, "epoch": 761, "lr": 8.555298635479909e-05}
{"train_loss": 0.2082900106906891, "global_step": 67754, "epoch": 761, "lr": 8.555257872127653e-05}
{"train_loss": 0.1917339265346527, "global_step": 67755, "epoch": 761, "lr": 8.555217108297435e-05}
{"train_loss": 0.2693360447883606, "global_step": 67756, "epoch": 761, "lr": 8.555176343989262e-05}
{"train_loss": 0.20031388103961945, "global_step": 67757, "epoch": 761, "lr": 8.555135579203137e-05}
{"train_loss": 0.25515344738960266, "global_step": 67758, "epoch": 761, "lr": 8.555094813939068e-05}
{"train_loss": 0.1676841378211975, "global_step": 67759, "epoch": 761, "lr": 8.555054048197063e-05}
{"train_loss": 0.19919759035110474, "global_step": 67760, "epoch": 761, "lr": 8.55501328197712e-05}
{"train_loss": 0.13894273340702057, "global_step": 67761, "epoch": 761, "lr": 8.55497251527925e-05}
{"train_loss": 0.28999534249305725, "global_step": 67762, "epoch": 761, "lr": 8.554931748103459e-05}
{"train_loss": 0.18670780956745148, "global_step": 67763, "epoch": 761, "lr": 8.554890980449747e-05}
{"train_loss": 0.20336058735847473, "global_step": 67764, "epoch": 761, "lr": 8.554850212318128e-05}
{"train_loss": 0.14563366770744324, "global_step": 67765, "epoch": 761, "lr": 8.5548094437086e-05}
{"train_loss": 0.20748275518417358, "global_step": 67766, "epoch": 761, "lr": 8.554768674621172e-05}
{"train_loss": 0.21825793385505676, "global_step": 67767, "epoch": 761, "lr": 8.554727905055848e-05}
{"train_loss": 0.2984031140804291, "global_step": 67768, "epoch": 761, "lr": 8.554687135012635e-05}
{"train_loss": 0.2330792397260666, "global_step": 67769, "epoch": 761, "lr": 8.554646364491536e-05}
{"train_loss": 0.26414981484413147, "global_step": 67770, "epoch": 761, "lr": 8.554605593492559e-05}
{"train_loss": 0.2004433572292328, "global_step": 67771, "epoch": 761, "lr": 8.55456482201571e-05}
{"train_loss": 0.17572544515132904, "global_step": 67772, "epoch": 761, "lr": 8.554524050060993e-05}
{"train_loss": 0.2145284116268158, "global_step": 67773, "epoch": 761, "lr": 8.554483277628411e-05}
{"train_loss": 0.14417661726474762, "global_step": 67774, "epoch": 761, "lr": 8.554442504717974e-05}
{"train_loss": 0.286068320274353, "global_step": 67775, "epoch": 761, "lr": 8.554401731329687e-05}
{"train_loss": 0.3231775760650635, "global_step": 67776, "epoch": 761, "lr": 8.554360957463551e-05}
{"train_loss": 0.33053043484687805, "global_step": 67777, "epoch": 761, "lr": 8.554320183119575e-05}
{"train_loss": 0.18544353544712067, "global_step": 67778, "epoch": 761, "lr": 8.554279408297765e-05}
{"train_loss": 0.18681780993938446, "global_step": 67779, "epoch": 761, "lr": 8.554238632998125e-05}
{"train_loss": 0.2294715940952301, "global_step": 67780, "epoch": 761, "lr": 8.554197857220661e-05}
{"train_loss": 0.20711271464824677, "global_step": 67781, "epoch": 761, "lr": 8.554157080965378e-05}
{"train_loss": 0.161407008767128, "global_step": 67782, "epoch": 761, "lr": 8.554116304232282e-05}
{"train_loss": 0.22686031460762024, "global_step": 67783, "epoch": 761, "lr": 8.554075527021379e-05}
{"train_loss": 0.23384864628314972, "global_step": 67784, "epoch": 761, "lr": 8.554034749332674e-05}
{"train_loss": 0.18387028574943542, "global_step": 67785, "epoch": 761, "lr": 8.553993971166172e-05}
{"train_loss": 0.24455220997333527, "global_step": 67786, "epoch": 761, "lr": 8.553953192521879e-05}
{"train_loss": 0.16908825933933258, "global_step": 67787, "epoch": 761, "lr": 8.553912413399801e-05}
{"train_loss": 0.29328683018684387, "global_step": 67788, "epoch": 761, "lr": 8.553871633799941e-05}
{"train_loss": 0.19294849038124084, "global_step": 67789, "epoch": 761, "lr": 8.553830853722307e-05}
{"train_loss": 0.24352282285690308, "global_step": 67790, "epoch": 761, "lr": 8.553790073166902e-05}
{"train_loss": 0.20872046053409576, "global_step": 67791, "epoch": 761, "lr": 8.553749292133736e-05}
{"train_loss": 0.18398703634738922, "global_step": 67792, "epoch": 761, "lr": 8.553708510622811e-05}
{"train_loss": 0.31128552556037903, "global_step": 67793, "epoch": 761, "lr": 8.553667728634134e-05}
{"train_loss": 0.24841634929180145, "global_step": 67794, "epoch": 761, "lr": 8.553626946167708e-05}
{"train_loss": 0.22937631607055664, "global_step": 67795, "epoch": 761, "lr": 8.553586163223541e-05}
{"train_loss": 0.27147147059440613, "global_step": 67796, "epoch": 761, "lr": 8.553545379801638e-05}
{"train_loss": 0.24828848242759705, "global_step": 67797, "epoch": 761, "lr": 8.553504595902004e-05}
{"train_loss": 0.1848250776529312, "global_step": 67798, "epoch": 761, "lr": 8.553463811524643e-05}
{"train_loss": 0.2590647339820862, "global_step": 67799, "epoch": 761, "lr": 8.553423026669564e-05}
{"train_loss": 0.17299632728099823, "global_step": 67800, "epoch": 761, "lr": 8.55338224133677e-05}
{"train_loss": 0.1501081883907318, "global_step": 67801, "epoch": 761, "lr": 8.553341455526267e-05}
{"train_loss": 0.2459263950586319, "global_step": 67802, "epoch": 761, "lr": 8.55330066923806e-05}
{"train_loss": 0.22542914748191833, "global_step": 67803, "epoch": 761, "lr": 8.553259882472156e-05}
{"train_loss": 0.24933284521102905, "global_step": 67804, "epoch": 761, "lr": 8.55321909522856e-05}
{"train_loss": 0.15363186597824097, "global_step": 67805, "epoch": 761, "lr": 8.553178307507275e-05}
{"train_loss": 0.3200114667415619, "global_step": 67806, "epoch": 761, "lr": 8.55313751930831e-05}
{"train_loss": 0.14230264723300934, "global_step": 67807, "epoch": 761, "lr": 8.553096730631669e-05}
{"train_loss": 0.23344731330871582, "global_step": 67808, "epoch": 761, "lr": 8.553055941477357e-05}
{"train_loss": 0.15136046707630157, "global_step": 67809, "epoch": 761, "lr": 8.55301515184538e-05}
{"train_loss": 0.2237187623977661, "global_step": 67810, "epoch": 761, "lr": 8.552974361735744e-05}
{"train_loss": 0.2215278446674347, "global_step": 67811, "epoch": 761, "lr": 8.552933571148453e-05}
{"train_loss": 0.2429933398962021, "global_step": 67812, "epoch": 761, "lr": 8.552892780083513e-05}
{"train_loss": 0.18811461329460144, "global_step": 67813, "epoch": 761, "lr": 8.552851988540931e-05}
{"train_loss": 0.23944954574108124, "global_step": 67814, "epoch": 761, "lr": 8.552811196520711e-05}
{"train_loss": 0.24979397654533386, "global_step": 67815, "epoch": 761, "lr": 8.552770404022858e-05}
{"train_loss": 0.2668972313404083, "global_step": 67816, "epoch": 761, "lr": 8.552729611047378e-05}
{"train_loss": 0.2271096957533547, "global_step": 67817, "epoch": 761, "lr": 8.55268881759428e-05, "val_loss": 3.6659016609191895}
{"train_loss": 0.1587463766336441, "global_step": 67818, "epoch": 762, "lr": 8.552648023663564e-05}
{"train_loss": 0.18339483439922333, "global_step": 67819, "epoch": 762, "lr": 8.552607229255238e-05}
{"train_loss": 0.3364643454551697, "global_step": 67820, "epoch": 762, "lr": 8.552566434369307e-05}
{"train_loss": 0.17113730311393738, "global_step": 67821, "epoch": 762, "lr": 8.552525639005778e-05}
{"train_loss": 0.1681528091430664, "global_step": 67822, "epoch": 762, "lr": 8.552484843164652e-05}
{"train_loss": 0.20213697850704193, "global_step": 67823, "epoch": 762, "lr": 8.552444046845941e-05}
{"train_loss": 0.14790016412734985, "global_step": 67824, "epoch": 762, "lr": 8.552403250049645e-05}
{"train_loss": 0.18505167961120605, "global_step": 67825, "epoch": 762, "lr": 8.552362452775773e-05}
{"train_loss": 0.19777420163154602, "global_step": 67826, "epoch": 762, "lr": 8.552321655024329e-05}
{"train_loss": 0.24190863966941833, "global_step": 67827, "epoch": 762, "lr": 8.552280856795318e-05}
{"train_loss": 0.23639461398124695, "global_step": 67828, "epoch": 762, "lr": 8.552240058088747e-05}
{"train_loss": 0.26687633991241455, "global_step": 67829, "epoch": 762, "lr": 8.55219925890462e-05}
{"train_loss": 0.2196691483259201, "global_step": 67830, "epoch": 762, "lr": 8.552158459242945e-05}
{"train_loss": 0.1670038402080536, "global_step": 67831, "epoch": 762, "lr": 8.552117659103723e-05}
{"train_loss": 0.20001040399074554, "global_step": 67832, "epoch": 762, "lr": 8.552076858486964e-05}
{"train_loss": 0.22668927907943726, "global_step": 67833, "epoch": 762, "lr": 8.55203605739267e-05}
{"train_loss": 0.2228531688451767, "global_step": 67834, "epoch": 762, "lr": 8.551995255820848e-05}
{"train_loss": 0.17488248646259308, "global_step": 67835, "epoch": 762, "lr": 8.551954453771504e-05}
{"train_loss": 0.30658507347106934, "global_step": 67836, "epoch": 762, "lr": 8.551913651244645e-05}
{"train_loss": 0.20390275120735168, "global_step": 67837, "epoch": 762, "lr": 8.551872848240273e-05}
{"train_loss": 0.27945488691329956, "global_step": 67838, "epoch": 762, "lr": 8.551832044758394e-05}
{"train_loss": 0.34733229875564575, "global_step": 67839, "epoch": 762, "lr": 8.551791240799016e-05}
{"train_loss": 0.20857113599777222, "global_step": 67840, "epoch": 762, "lr": 8.551750436362141e-05}
{"train_loss": 0.2204560488462448, "global_step": 67841, "epoch": 762, "lr": 8.551709631447777e-05}
{"train_loss": 0.21323639154434204, "global_step": 67842, "epoch": 762, "lr": 8.55166882605593e-05}
{"train_loss": 0.23213805258274078, "global_step": 67843, "epoch": 762, "lr": 8.551628020186604e-05}
{"train_loss": 0.21749648451805115, "global_step": 67844, "epoch": 762, "lr": 8.551587213839805e-05}
{"train_loss": 0.21546556055545807, "global_step": 67845, "epoch": 762, "lr": 8.551546407015539e-05}
{"train_loss": 0.210560604929924, "global_step": 67846, "epoch": 762, "lr": 8.551505599713809e-05}
{"train_loss": 0.24571792781352997, "global_step": 67847, "epoch": 762, "lr": 8.551464791934625e-05}
{"train_loss": 0.1735759824514389, "global_step": 67848, "epoch": 762, "lr": 8.551423983677988e-05}
{"train_loss": 0.21978691220283508, "global_step": 67849, "epoch": 762, "lr": 8.551383174943905e-05}
{"train_loss": 0.20281946659088135, "global_step": 67850, "epoch": 762, "lr": 8.551342365732381e-05}
{"train_loss": 0.2434471696615219, "global_step": 67851, "epoch": 762, "lr": 8.551301556043425e-05}
{"train_loss": 0.19231092929840088, "global_step": 67852, "epoch": 762, "lr": 8.551260745877038e-05}
{"train_loss": 0.17532938718795776, "global_step": 67853, "epoch": 762, "lr": 8.55121993523323e-05}
{"train_loss": 0.2234940230846405, "global_step": 67854, "epoch": 762, "lr": 8.551179124112002e-05}
{"train_loss": 0.3150096833705902, "global_step": 67855, "epoch": 762, "lr": 8.551138312513359e-05}
{"train_loss": 0.2015521377325058, "global_step": 67856, "epoch": 762, "lr": 8.551097500437311e-05}
{"train_loss": 0.25993672013282776, "global_step": 67857, "epoch": 762, "lr": 8.55105668788386e-05}
{"train_loss": 0.22820380330085754, "global_step": 67858, "epoch": 762, "lr": 8.551015874853015e-05}
{"train_loss": 0.260201096534729, "global_step": 67859, "epoch": 762, "lr": 8.550975061344776e-05}
{"train_loss": 0.2230944037437439, "global_step": 67860, "epoch": 762, "lr": 8.550934247359154e-05}
{"train_loss": 0.23750735819339752, "global_step": 67861, "epoch": 762, "lr": 8.550893432896153e-05}
{"train_loss": 0.13599568605422974, "global_step": 67862, "epoch": 762, "lr": 8.550852617955775e-05}
{"train_loss": 0.1881384253501892, "global_step": 67863, "epoch": 762, "lr": 8.55081180253803e-05}
{"train_loss": 0.2612074017524719, "global_step": 67864, "epoch": 762, "lr": 8.55077098664292e-05}
{"train_loss": 0.18003028631210327, "global_step": 67865, "epoch": 762, "lr": 8.550730170270453e-05}
{"train_loss": 0.2080511599779129, "global_step": 67866, "epoch": 762, "lr": 8.550689353420635e-05}
{"train_loss": 0.2551312744617462, "global_step": 67867, "epoch": 762, "lr": 8.550648536093469e-05}
{"train_loss": 0.231881245970726, "global_step": 67868, "epoch": 762, "lr": 8.550607718288961e-05}
{"train_loss": 0.2488429844379425, "global_step": 67869, "epoch": 762, "lr": 8.550566900007118e-05}
{"train_loss": 0.22932086884975433, "global_step": 67870, "epoch": 762, "lr": 8.550526081247943e-05}
{"train_loss": 0.24345934391021729, "global_step": 67871, "epoch": 762, "lr": 8.550485262011444e-05}
{"train_loss": 0.17361435294151306, "global_step": 67872, "epoch": 762, "lr": 8.550444442297626e-05}
{"train_loss": 0.23906779289245605, "global_step": 67873, "epoch": 762, "lr": 8.550403622106492e-05}
{"train_loss": 0.1436089724302292, "global_step": 67874, "epoch": 762, "lr": 8.550362801438053e-05}
{"train_loss": 0.23479607701301575, "global_step": 67875, "epoch": 762, "lr": 8.550321980292308e-05}
{"train_loss": 0.2736652195453644, "global_step": 67876, "epoch": 762, "lr": 8.550281158669266e-05}
{"train_loss": 0.29544973373413086, "global_step": 67877, "epoch": 762, "lr": 8.550240336568933e-05}
{"train_loss": 0.17546480894088745, "global_step": 67878, "epoch": 762, "lr": 8.550199513991313e-05}
{"train_loss": 0.1883665919303894, "global_step": 67879, "epoch": 762, "lr": 8.550158690936413e-05}
{"train_loss": 0.2812751531600952, "global_step": 67880, "epoch": 762, "lr": 8.550117867404237e-05}
{"train_loss": 0.20020660758018494, "global_step": 67881, "epoch": 762, "lr": 8.55007704339479e-05}
{"train_loss": 0.19889378547668457, "global_step": 67882, "epoch": 762, "lr": 8.550036218908079e-05}
{"train_loss": 0.21516460180282593, "global_step": 67883, "epoch": 762, "lr": 8.549995393944107e-05}
{"train_loss": 0.2729357182979584, "global_step": 67884, "epoch": 762, "lr": 8.549954568502885e-05}
{"train_loss": 0.18726205825805664, "global_step": 67885, "epoch": 762, "lr": 8.549913742584412e-05}
{"train_loss": 0.21816645562648773, "global_step": 67886, "epoch": 762, "lr": 8.549872916188697e-05}
{"train_loss": 0.19392988085746765, "global_step": 67887, "epoch": 762, "lr": 8.549832089315745e-05}
{"train_loss": 0.2175837755203247, "global_step": 67888, "epoch": 762, "lr": 8.549791261965561e-05}
{"train_loss": 0.17971113324165344, "global_step": 67889, "epoch": 762, "lr": 8.549750434138154e-05}
{"train_loss": 0.19002865254878998, "global_step": 67890, "epoch": 762, "lr": 8.549709605833522e-05}
{"train_loss": 0.24493259191513062, "global_step": 67891, "epoch": 762, "lr": 8.549668777051676e-05}
{"train_loss": 0.1992868185043335, "global_step": 67892, "epoch": 762, "lr": 8.549627947792622e-05}
{"train_loss": 0.1644897609949112, "global_step": 67893, "epoch": 762, "lr": 8.549587118056361e-05}
{"train_loss": 0.2026960700750351, "global_step": 67894, "epoch": 762, "lr": 8.549546287842903e-05}
{"train_loss": 0.21310468018054962, "global_step": 67895, "epoch": 762, "lr": 8.549505457152252e-05}
{"train_loss": 0.24093391001224518, "global_step": 67896, "epoch": 762, "lr": 8.549464625984412e-05}
{"train_loss": 0.1996578872203827, "global_step": 67897, "epoch": 762, "lr": 8.54942379433939e-05}
{"train_loss": 0.2144014984369278, "global_step": 67898, "epoch": 762, "lr": 8.549382962217193e-05}
{"train_loss": 0.16435480117797852, "global_step": 67899, "epoch": 762, "lr": 8.549342129617823e-05}
{"train_loss": 0.18542742729187012, "global_step": 67900, "epoch": 762, "lr": 8.549301296541287e-05}
{"train_loss": 0.2400788515806198, "global_step": 67901, "epoch": 762, "lr": 8.549260462987593e-05}
{"train_loss": 0.279175728559494, "global_step": 67902, "epoch": 762, "lr": 8.549219628956742e-05}
{"train_loss": 0.24553780257701874, "global_step": 67903, "epoch": 762, "lr": 8.549178794448742e-05}
{"train_loss": 0.22222399711608887, "global_step": 67904, "epoch": 762, "lr": 8.5491379594636e-05}
{"train_loss": 0.31937843561172485, "global_step": 67905, "epoch": 762, "lr": 8.549097124001317e-05}
{"train_loss": 0.22127495471681102, "global_step": 67906, "epoch": 762, "lr": 8.549056288061902e-05, "val_loss": 3.6808319091796875}
{"train_loss": 0.3406623601913452, "global_step": 67907, "epoch": 763, "lr": 8.54901545164536e-05}
{"train_loss": 0.17240183055400848, "global_step": 67908, "epoch": 763, "lr": 8.548974614751696e-05}
{"train_loss": 0.22736482322216034, "global_step": 67909, "epoch": 763, "lr": 8.548933777380916e-05}
{"train_loss": 0.2562396824359894, "global_step": 67910, "epoch": 763, "lr": 8.548892939533025e-05}
{"train_loss": 0.18231913447380066, "global_step": 67911, "epoch": 763, "lr": 8.548852101208029e-05}
{"train_loss": 0.30297034978866577, "global_step": 67912, "epoch": 763, "lr": 8.548811262405931e-05}
{"train_loss": 0.19355422258377075, "global_step": 67913, "epoch": 763, "lr": 8.54877042312674e-05}
{"train_loss": 0.2311689257621765, "global_step": 67914, "epoch": 763, "lr": 8.548729583370461e-05}
{"train_loss": 0.24794995784759521, "global_step": 67915, "epoch": 763, "lr": 8.548688743137097e-05}
{"train_loss": 0.2078298032283783, "global_step": 67916, "epoch": 763, "lr": 8.548647902426656e-05}
{"train_loss": 0.3164268136024475, "global_step": 67917, "epoch": 763, "lr": 8.548607061239142e-05}
{"train_loss": 0.19616356492042542, "global_step": 67918, "epoch": 763, "lr": 8.548566219574561e-05}
{"train_loss": 0.2112504541873932, "global_step": 67919, "epoch": 763, "lr": 8.548525377432919e-05}
{"train_loss": 0.18846502900123596, "global_step": 67920, "epoch": 763, "lr": 8.54848453481422e-05}
{"train_loss": 0.2992914319038391, "global_step": 67921, "epoch": 763, "lr": 8.548443691718473e-05}
{"train_loss": 0.25642380118370056, "global_step": 67922, "epoch": 763, "lr": 8.548402848145677e-05}
{"train_loss": 0.15316404402256012, "global_step": 67923, "epoch": 763, "lr": 8.548362004095846e-05}
{"train_loss": 0.15662333369255066, "global_step": 67924, "epoch": 763, "lr": 8.548321159568978e-05}
{"train_loss": 0.23263543844223022, "global_step": 67925, "epoch": 763, "lr": 8.548280314565082e-05}
{"train_loss": 0.1859291046857834, "global_step": 67926, "epoch": 763, "lr": 8.548239469084163e-05}
{"train_loss": 0.19118528068065643, "global_step": 67927, "epoch": 763, "lr": 8.548198623126227e-05}
{"train_loss": 0.1964889019727707, "global_step": 67928, "epoch": 763, "lr": 8.54815777669128e-05}
{"train_loss": 0.1115712970495224, "global_step": 67929, "epoch": 763, "lr": 8.548116929779324e-05}
{"train_loss": 0.17116744816303253, "global_step": 67930, "epoch": 763, "lr": 8.548076082390368e-05}
{"train_loss": 0.2322588860988617, "global_step": 67931, "epoch": 763, "lr": 8.548035234524417e-05}
{"train_loss": 0.2536744177341461, "global_step": 67932, "epoch": 763, "lr": 8.547994386181476e-05}
{"train_loss": 0.2180095911026001, "global_step": 67933, "epoch": 763, "lr": 8.547953537361549e-05}
{"train_loss": 0.23856575787067413, "global_step": 67934, "epoch": 763, "lr": 8.547912688064644e-05}
{"train_loss": 0.24367041885852814, "global_step": 67935, "epoch": 763, "lr": 8.547871838290764e-05}
{"train_loss": 0.33054593205451965, "global_step": 67936, "epoch": 763, "lr": 8.547830988039917e-05}
{"train_loss": 0.15655042231082916, "global_step": 67937, "epoch": 763, "lr": 8.547790137312106e-05}
{"train_loss": 0.2096388190984726, "global_step": 67938, "epoch": 763, "lr": 8.54774928610734e-05}
{"train_loss": 0.2355605661869049, "global_step": 67939, "epoch": 763, "lr": 8.547708434425621e-05}
{"train_loss": 0.1640331745147705, "global_step": 67940, "epoch": 763, "lr": 8.547667582266956e-05}
{"train_loss": 0.2305586338043213, "global_step": 67941, "epoch": 763, "lr": 8.547626729631351e-05}
{"train_loss": 0.21217042207717896, "global_step": 67942, "epoch": 763, "lr": 8.54758587651881e-05}
{"train_loss": 0.2736690044403076, "global_step": 67943, "epoch": 763, "lr": 8.54754502292934e-05}
{"train_loss": 0.20195788145065308, "global_step": 67944, "epoch": 763, "lr": 8.547504168862945e-05}
{"train_loss": 0.21819542348384857, "global_step": 67945, "epoch": 763, "lr": 8.547463314319632e-05}
{"train_loss": 0.18882879614830017, "global_step": 67946, "epoch": 763, "lr": 8.547422459299405e-05}
{"train_loss": 0.26678791642189026, "global_step": 67947, "epoch": 763, "lr": 8.547381603802272e-05}
{"train_loss": 0.138260155916214, "global_step": 67948, "epoch": 763, "lr": 8.547340747828236e-05}
{"train_loss": 0.2359592318534851, "global_step": 67949, "epoch": 763, "lr": 8.547299891377303e-05}
{"train_loss": 0.19194470345973969, "global_step": 67950, "epoch": 763, "lr": 8.547259034449479e-05}
{"train_loss": 0.2834359407424927, "global_step": 67951, "epoch": 763, "lr": 8.547218177044768e-05}
{"train_loss": 0.29702648520469666, "global_step": 67952, "epoch": 763, "lr": 8.547177319163178e-05}
{"train_loss": 0.17293411493301392, "global_step": 67953, "epoch": 763, "lr": 8.547136460804712e-05}
{"train_loss": 0.2177877575159073, "global_step": 67954, "epoch": 763, "lr": 8.54709560196938e-05}
{"train_loss": 0.22797228395938873, "global_step": 67955, "epoch": 763, "lr": 8.547054742657181e-05}
{"train_loss": 0.16513018310070038, "global_step": 67956, "epoch": 763, "lr": 8.547013882868125e-05}
{"train_loss": 0.16986674070358276, "global_step": 67957, "epoch": 763, "lr": 8.546973022602216e-05}
{"train_loss": 0.17436738312244415, "global_step": 67958, "epoch": 763, "lr": 8.54693216185946e-05}
{"train_loss": 0.21043512225151062, "global_step": 67959, "epoch": 763, "lr": 8.546891300639861e-05}
{"train_loss": 0.14974437654018402, "global_step": 67960, "epoch": 763, "lr": 8.546850438943426e-05}
{"train_loss": 0.24199561774730682, "global_step": 67961, "epoch": 763, "lr": 8.54680957677016e-05}
{"train_loss": 0.2537482976913452, "global_step": 67962, "epoch": 763, "lr": 8.54676871412007e-05}
{"train_loss": 0.19473358988761902, "global_step": 67963, "epoch": 763, "lr": 8.546727850993159e-05}
{"train_loss": 0.13275431096553802, "global_step": 67964, "epoch": 763, "lr": 8.546686987389434e-05}
{"train_loss": 0.20918972790241241, "global_step": 67965, "epoch": 763, "lr": 8.546646123308902e-05}
{"train_loss": 0.26000964641571045, "global_step": 67966, "epoch": 763, "lr": 8.546605258751564e-05}
{"train_loss": 0.2893257439136505, "global_step": 67967, "epoch": 763, "lr": 8.546564393717429e-05}
{"train_loss": 0.22837218642234802, "global_step": 67968, "epoch": 763, "lr": 8.546523528206503e-05}
{"train_loss": 0.2744547724723816, "global_step": 67969, "epoch": 763, "lr": 8.546482662218789e-05}
{"train_loss": 0.1870894879102707, "global_step": 67970, "epoch": 763, "lr": 8.546441795754292e-05}
{"train_loss": 0.2780131995677948, "global_step": 67971, "epoch": 763, "lr": 8.546400928813021e-05}
{"train_loss": 0.17860658466815948, "global_step": 67972, "epoch": 763, "lr": 8.54636006139498e-05}
{"train_loss": 0.3188821077346802, "global_step": 67973, "epoch": 763, "lr": 8.546319193500172e-05}
{"train_loss": 0.2573883831501007, "global_step": 67974, "epoch": 763, "lr": 8.546278325128607e-05}
{"train_loss": 0.1886099874973297, "global_step": 67975, "epoch": 763, "lr": 8.546237456280288e-05}
{"train_loss": 0.2229529768228531, "global_step": 67976, "epoch": 763, "lr": 8.54619658695522e-05}
{"train_loss": 0.26243337988853455, "global_step": 67977, "epoch": 763, "lr": 8.546155717153408e-05}
{"train_loss": 0.24041177332401276, "global_step": 67978, "epoch": 763, "lr": 8.54611484687486e-05}
{"train_loss": 0.20794351398944855, "global_step": 67979, "epoch": 763, "lr": 8.54607397611958e-05}
{"train_loss": 0.21116454899311066, "global_step": 67980, "epoch": 763, "lr": 8.546033104887572e-05}
{"train_loss": 0.19757673144340515, "global_step": 67981, "epoch": 763, "lr": 8.545992233178843e-05}
{"train_loss": 0.2212737649679184, "global_step": 67982, "epoch": 763, "lr": 8.545951360993401e-05}
{"train_loss": 0.18290230631828308, "global_step": 67983, "epoch": 763, "lr": 8.545910488331248e-05}
{"train_loss": 0.2566317915916443, "global_step": 67984, "epoch": 763, "lr": 8.54586961519239e-05}
{"train_loss": 0.20085622370243073, "global_step": 67985, "epoch": 763, "lr": 8.545828741576833e-05}
{"train_loss": 0.19703422486782074, "global_step": 67986, "epoch": 763, "lr": 8.545787867484583e-05}
{"train_loss": 0.2644672691822052, "global_step": 67987, "epoch": 763, "lr": 8.545746992915645e-05}
{"train_loss": 0.20767726004123688, "global_step": 67988, "epoch": 763, "lr": 8.545706117870025e-05}
{"train_loss": 0.2173367440700531, "global_step": 67989, "epoch": 763, "lr": 8.545665242347727e-05}
{"train_loss": 0.23305413126945496, "global_step": 67990, "epoch": 763, "lr": 8.545624366348758e-05}
{"train_loss": 0.19681985676288605, "global_step": 67991, "epoch": 763, "lr": 8.545583489873123e-05}
{"train_loss": 0.279153048992157, "global_step": 67992, "epoch": 763, "lr": 8.545542612920827e-05}
{"train_loss": 0.23126526176929474, "global_step": 67993, "epoch": 763, "lr": 8.545501735491877e-05}
{"train_loss": 0.2793917953968048, "global_step": 67994, "epoch": 763, "lr": 8.545460857586278e-05}
{"train_loss": 0.221716043319595, "global_step": 67995, "epoch": 763, "lr": 8.545419979204035e-05, "val_loss": 3.6286561489105225}
{"train_loss": 0.2586996257305145, "global_step": 67996, "epoch": 764, "lr": 8.545379100345152e-05}
{"train_loss": 0.17879503965377808, "global_step": 67997, "epoch": 764, "lr": 8.545338221009636e-05}
{"train_loss": 0.29286107420921326, "global_step": 67998, "epoch": 764, "lr": 8.545297341197493e-05}
{"train_loss": 0.2900914251804352, "global_step": 67999, "epoch": 764, "lr": 8.545256460908729e-05}
{"train_loss": 0.29343006014823914, "global_step": 68000, "epoch": 764, "lr": 8.545215580143346e-05}
{"train_loss": 0.19096238911151886, "global_step": 68001, "epoch": 764, "lr": 8.545174698901354e-05}
{"train_loss": 0.16569483280181885, "global_step": 68002, "epoch": 764, "lr": 8.545133817182756e-05}
{"train_loss": 0.2249777615070343, "global_step": 68003, "epoch": 764, "lr": 8.545092934987557e-05}
{"train_loss": 0.2792736291885376, "global_step": 68004, "epoch": 764, "lr": 8.545052052315765e-05}
{"train_loss": 0.2581471800804138, "global_step": 68005, "epoch": 764, "lr": 8.545011169167384e-05}
{"train_loss": 0.269748717546463, "global_step": 68006, "epoch": 764, "lr": 8.544970285542419e-05}
{"train_loss": 0.3269132077693939, "global_step": 68007, "epoch": 764, "lr": 8.544929401440875e-05}
{"train_loss": 0.19957853853702545, "global_step": 68008, "epoch": 764, "lr": 8.544888516862759e-05}
{"train_loss": 0.2611066401004791, "global_step": 68009, "epoch": 764, "lr": 8.544847631808077e-05}
{"train_loss": 0.2502369284629822, "global_step": 68010, "epoch": 764, "lr": 8.544806746276832e-05}
{"train_loss": 0.18425381183624268, "global_step": 68011, "epoch": 764, "lr": 8.544765860269031e-05}
{"train_loss": 0.22879773378372192, "global_step": 68012, "epoch": 764, "lr": 8.544724973784681e-05}
{"train_loss": 0.2691356837749481, "global_step": 68013, "epoch": 764, "lr": 8.544684086823784e-05}
{"train_loss": 0.23699280619621277, "global_step": 68014, "epoch": 764, "lr": 8.54464319938635e-05}
{"train_loss": 0.2860119342803955, "global_step": 68015, "epoch": 764, "lr": 8.54460231147238e-05}
{"train_loss": 0.2354963719844818, "global_step": 68016, "epoch": 764, "lr": 8.54456142308188e-05}
{"train_loss": 0.14856353402137756, "global_step": 68017, "epoch": 764, "lr": 8.544520534214859e-05}
{"train_loss": 0.20135051012039185, "global_step": 68018, "epoch": 764, "lr": 8.54447964487132e-05}
{"train_loss": 0.2536444067955017, "global_step": 68019, "epoch": 764, "lr": 8.54443875505127e-05}
{"train_loss": 0.2643217444419861, "global_step": 68020, "epoch": 764, "lr": 8.544397864754712e-05}
{"train_loss": 0.34941431879997253, "global_step": 68021, "epoch": 764, "lr": 8.544356973981652e-05}
{"train_loss": 0.1828773319721222, "global_step": 68022, "epoch": 764, "lr": 8.544316082732099e-05}
{"train_loss": 0.1571323722600937, "global_step": 68023, "epoch": 764, "lr": 8.544275191006055e-05}
{"train_loss": 0.19621214270591736, "global_step": 68024, "epoch": 764, "lr": 8.544234298803526e-05}
{"train_loss": 0.2660900354385376, "global_step": 68025, "epoch": 764, "lr": 8.544193406124516e-05}
{"train_loss": 0.2175595462322235, "global_step": 68026, "epoch": 764, "lr": 8.544152512969035e-05}
{"train_loss": 0.17364828288555145, "global_step": 68027, "epoch": 764, "lr": 8.544111619337085e-05}
{"train_loss": 0.19481949508190155, "global_step": 68028, "epoch": 764, "lr": 8.544070725228673e-05}
{"train_loss": 0.20494388043880463, "global_step": 68029, "epoch": 764, "lr": 8.544029830643803e-05}
{"train_loss": 0.17821472883224487, "global_step": 68030, "epoch": 764, "lr": 8.543988935582483e-05}
{"train_loss": 0.21409054100513458, "global_step": 68031, "epoch": 764, "lr": 8.543948040044716e-05}
{"train_loss": 0.16570152342319489, "global_step": 68032, "epoch": 764, "lr": 8.543907144030508e-05}
{"train_loss": 0.24303700029850006, "global_step": 68033, "epoch": 764, "lr": 8.543866247539864e-05}
{"train_loss": 0.20290710031986237, "global_step": 68034, "epoch": 764, "lr": 8.543825350572792e-05}
{"train_loss": 0.20251165330410004, "global_step": 68035, "epoch": 764, "lr": 8.543784453129295e-05}
{"train_loss": 0.1982690840959549, "global_step": 68036, "epoch": 764, "lr": 8.54374355520938e-05}
{"train_loss": 0.25467467308044434, "global_step": 68037, "epoch": 764, "lr": 8.543702656813051e-05}
{"train_loss": 0.14960891008377075, "global_step": 68038, "epoch": 764, "lr": 8.543661757940314e-05}
{"train_loss": 0.22217823565006256, "global_step": 68039, "epoch": 764, "lr": 8.543620858591177e-05}
{"train_loss": 0.23889553546905518, "global_step": 68040, "epoch": 764, "lr": 8.543579958765642e-05}
{"train_loss": 0.16425754129886627, "global_step": 68041, "epoch": 764, "lr": 8.543539058463715e-05}
{"train_loss": 0.2808881402015686, "global_step": 68042, "epoch": 764, "lr": 8.543498157685404e-05}
{"train_loss": 0.1918621063232422, "global_step": 68043, "epoch": 764, "lr": 8.543457256430712e-05}
{"train_loss": 0.27316737174987793, "global_step": 68044, "epoch": 764, "lr": 8.543416354699644e-05}
{"train_loss": 0.1328020989894867, "global_step": 68045, "epoch": 764, "lr": 8.54337545249221e-05}
{"train_loss": 0.1856818050146103, "global_step": 68046, "epoch": 764, "lr": 8.54333454980841e-05}
{"train_loss": 0.24843573570251465, "global_step": 68047, "epoch": 764, "lr": 8.543293646648253e-05}
{"train_loss": 0.12603455781936646, "global_step": 68048, "epoch": 764, "lr": 8.543252743011742e-05}
{"train_loss": 0.20924989879131317, "global_step": 68049, "epoch": 764, "lr": 8.543211838898886e-05}
{"train_loss": 0.25270816683769226, "global_step": 68050, "epoch": 764, "lr": 8.543170934309686e-05}
{"train_loss": 0.1296686828136444, "global_step": 68051, "epoch": 764, "lr": 8.543130029244152e-05}
{"train_loss": 0.18199346959590912, "global_step": 68052, "epoch": 764, "lr": 8.543089123702286e-05}
{"train_loss": 0.26702380180358887, "global_step": 68053, "epoch": 764, "lr": 8.543048217684095e-05}
{"train_loss": 0.2781846225261688, "global_step": 68054, "epoch": 764, "lr": 8.543007311189585e-05}
{"train_loss": 0.19972969591617584, "global_step": 68055, "epoch": 764, "lr": 8.54296640421876e-05}
{"train_loss": 0.2236003279685974, "global_step": 68056, "epoch": 764, "lr": 8.542925496771628e-05}
{"train_loss": 0.15162061154842377, "global_step": 68057, "epoch": 764, "lr": 8.54288458884819e-05}
{"train_loss": 0.1646435409784317, "global_step": 68058, "epoch": 764, "lr": 8.542843680448456e-05}
{"train_loss": 0.21698012948036194, "global_step": 68059, "epoch": 764, "lr": 8.54280277157243e-05}
{"train_loss": 0.18858662247657776, "global_step": 68060, "epoch": 764, "lr": 8.542761862220116e-05}
{"train_loss": 0.21334317326545715, "global_step": 68061, "epoch": 764, "lr": 8.542720952391523e-05}
{"train_loss": 0.24014781415462494, "global_step": 68062, "epoch": 764, "lr": 8.542680042086654e-05}
{"train_loss": 0.18940189480781555, "global_step": 68063, "epoch": 764, "lr": 8.542639131305514e-05}
{"train_loss": 0.22695423662662506, "global_step": 68064, "epoch": 764, "lr": 8.54259822004811e-05}
{"train_loss": 0.22563906013965607, "global_step": 68065, "epoch": 764, "lr": 8.542557308314445e-05}
{"train_loss": 0.25146400928497314, "global_step": 68066, "epoch": 764, "lr": 8.542516396104528e-05}
{"train_loss": 0.26187843084335327, "global_step": 68067, "epoch": 764, "lr": 8.542475483418363e-05}
{"train_loss": 0.2825750708580017, "global_step": 68068, "epoch": 764, "lr": 8.542434570255954e-05}
{"train_loss": 0.3270851671695709, "global_step": 68069, "epoch": 764, "lr": 8.542393656617308e-05}
{"train_loss": 0.1872958391904831, "global_step": 68070, "epoch": 764, "lr": 8.542352742502433e-05}
{"train_loss": 0.16602344810962677, "global_step": 68071, "epoch": 764, "lr": 8.54231182791133e-05}
{"train_loss": 0.19011829793453217, "global_step": 68072, "epoch": 764, "lr": 8.542270912844006e-05}
{"train_loss": 0.2540922164916992, "global_step": 68073, "epoch": 764, "lr": 8.542229997300467e-05}
{"train_loss": 0.2668594419956207, "global_step": 68074, "epoch": 764, "lr": 8.542189081280718e-05}
{"train_loss": 0.20177015662193298, "global_step": 68075, "epoch": 764, "lr": 8.542148164784767e-05}
{"train_loss": 0.3329108953475952, "global_step": 68076, "epoch": 764, "lr": 8.542107247812616e-05}
{"train_loss": 0.24358931183815002, "global_step": 68077, "epoch": 764, "lr": 8.542066330364272e-05}
{"train_loss": 0.17461156845092773, "global_step": 68078, "epoch": 764, "lr": 8.542025412439738e-05}
{"train_loss": 0.18958790600299835, "global_step": 68079, "epoch": 764, "lr": 8.541984494039024e-05}
{"train_loss": 0.2970876693725586, "global_step": 68080, "epoch": 764, "lr": 8.541943575162134e-05}
{"train_loss": 0.17615167796611786, "global_step": 68081, "epoch": 764, "lr": 8.54190265580907e-05}
{"train_loss": 0.34517866373062134, "global_step": 68082, "epoch": 764, "lr": 8.541861735979844e-05}
{"train_loss": 0.2478024810552597, "global_step": 68083, "epoch": 764, "lr": 8.541820815674455e-05}
{"train_loss": 0.22532999549019203, "global_step": 68084, "epoch": 764, "lr": 8.541779894892914e-05, "val_loss": 3.663677453994751}
{"train_loss": 0.167942613363266, "global_step": 68085, "epoch": 765, "lr": 8.541738973635223e-05}
{"train_loss": 0.2686610817909241, "global_step": 68086, "epoch": 765, "lr": 8.541698051901387e-05}
{"train_loss": 0.2673453986644745, "global_step": 68087, "epoch": 765, "lr": 8.541657129691412e-05}
{"train_loss": 0.24083448946475983, "global_step": 68088, "epoch": 765, "lr": 8.541616207005308e-05}
{"train_loss": 0.21916869282722473, "global_step": 68089, "epoch": 765, "lr": 8.541575283843074e-05}
{"train_loss": 0.1778237223625183, "global_step": 68090, "epoch": 765, "lr": 8.54153436020472e-05}
{"train_loss": 0.18294045329093933, "global_step": 68091, "epoch": 765, "lr": 8.541493436090248e-05}
{"train_loss": 0.20756042003631592, "global_step": 68092, "epoch": 765, "lr": 8.541452511499668e-05}
{"train_loss": 0.20476184785366058, "global_step": 68093, "epoch": 765, "lr": 8.541411586432981e-05}
{"train_loss": 0.23540948331356049, "global_step": 68094, "epoch": 765, "lr": 8.541370660890194e-05}
{"train_loss": 0.14690326154232025, "global_step": 68095, "epoch": 765, "lr": 8.541329734871313e-05}
{"train_loss": 0.1881462186574936, "global_step": 68096, "epoch": 765, "lr": 8.541288808376344e-05}
{"train_loss": 0.1970186084508896, "global_step": 68097, "epoch": 765, "lr": 8.541247881405292e-05}
{"train_loss": 0.2150222659111023, "global_step": 68098, "epoch": 765, "lr": 8.541206953958163e-05}
{"train_loss": 0.20787157118320465, "global_step": 68099, "epoch": 765, "lr": 8.541166026034961e-05}
{"train_loss": 0.20587997138500214, "global_step": 68100, "epoch": 765, "lr": 8.541125097635692e-05}
{"train_loss": 0.2453431934118271, "global_step": 68101, "epoch": 765, "lr": 8.541084168760363e-05}
{"train_loss": 0.38006606698036194, "global_step": 68102, "epoch": 765, "lr": 8.54104323940898e-05}
{"train_loss": 0.34612029790878296, "global_step": 68103, "epoch": 765, "lr": 8.541002309581543e-05}
{"train_loss": 0.21037836372852325, "global_step": 68104, "epoch": 765, "lr": 8.540961379278063e-05}
{"train_loss": 0.20958463847637177, "global_step": 68105, "epoch": 765, "lr": 8.540920448498544e-05}
{"train_loss": 0.30440405011177063, "global_step": 68106, "epoch": 765, "lr": 8.540879517242992e-05}
{"train_loss": 0.2103503942489624, "global_step": 68107, "epoch": 765, "lr": 8.54083858551141e-05}
{"train_loss": 0.15651263296604156, "global_step": 68108, "epoch": 765, "lr": 8.540797653303807e-05}
{"train_loss": 0.2429044246673584, "global_step": 68109, "epoch": 765, "lr": 8.540756720620187e-05}
{"train_loss": 0.22911572456359863, "global_step": 68110, "epoch": 765, "lr": 8.540715787460556e-05}
{"train_loss": 0.3213787376880646, "global_step": 68111, "epoch": 765, "lr": 8.540674853824917e-05}
{"train_loss": 0.18066814541816711, "global_step": 68112, "epoch": 765, "lr": 8.540633919713279e-05}
{"train_loss": 0.22306039929389954, "global_step": 68113, "epoch": 765, "lr": 8.540592985125644e-05}
{"train_loss": 0.22329740226268768, "global_step": 68114, "epoch": 765, "lr": 8.540552050062021e-05}
{"train_loss": 0.24430465698242188, "global_step": 68115, "epoch": 765, "lr": 8.540511114522414e-05}
{"train_loss": 0.21711553633213043, "global_step": 68116, "epoch": 765, "lr": 8.540470178506827e-05}
{"train_loss": 0.22895655035972595, "global_step": 68117, "epoch": 765, "lr": 8.540429242015269e-05}
{"train_loss": 0.19610831141471863, "global_step": 68118, "epoch": 765, "lr": 8.540388305047744e-05}
{"train_loss": 0.12211829423904419, "global_step": 68119, "epoch": 765, "lr": 8.540347367604254e-05}
{"train_loss": 0.15025363862514496, "global_step": 68120, "epoch": 765, "lr": 8.54030642968481e-05}
{"train_loss": 0.21842005848884583, "global_step": 68121, "epoch": 765, "lr": 8.540265491289415e-05}
{"train_loss": 0.1862742006778717, "global_step": 68122, "epoch": 765, "lr": 8.540224552418072e-05}
{"train_loss": 0.22533251345157623, "global_step": 68123, "epoch": 765, "lr": 8.54018361307079e-05}
{"train_loss": 0.25208452343940735, "global_step": 68124, "epoch": 765, "lr": 8.540142673247574e-05}
{"train_loss": 0.21300511062145233, "global_step": 68125, "epoch": 765, "lr": 8.540101732948427e-05}
{"train_loss": 0.20608969032764435, "global_step": 68126, "epoch": 765, "lr": 8.540060792173359e-05}
{"train_loss": 0.23839496076107025, "global_step": 68127, "epoch": 765, "lr": 8.540019850922372e-05}
{"train_loss": 0.25853052735328674, "global_step": 68128, "epoch": 765, "lr": 8.539978909195474e-05}
{"train_loss": 0.2730283737182617, "global_step": 68129, "epoch": 765, "lr": 8.539937966992667e-05}
{"train_loss": 0.2346036285161972, "global_step": 68130, "epoch": 765, "lr": 8.539897024313959e-05}
{"train_loss": 0.23688079416751862, "global_step": 68131, "epoch": 765, "lr": 8.539856081159356e-05}
{"train_loss": 0.28731781244277954, "global_step": 68132, "epoch": 765, "lr": 8.53981513752886e-05}
{"train_loss": 0.21211102604866028, "global_step": 68133, "epoch": 765, "lr": 8.53977419342248e-05}
{"train_loss": 0.1968960165977478, "global_step": 68134, "epoch": 765, "lr": 8.539733248840223e-05}
{"train_loss": 0.16046543419361115, "global_step": 68135, "epoch": 765, "lr": 8.53969230378209e-05}
{"train_loss": 0.1837339550256729, "global_step": 68136, "epoch": 765, "lr": 8.539651358248087e-05}
{"train_loss": 0.22217431664466858, "global_step": 68137, "epoch": 765, "lr": 8.539610412238223e-05}
{"train_loss": 0.26969340443611145, "global_step": 68138, "epoch": 765, "lr": 8.539569465752502e-05}
{"train_loss": 0.32150503993034363, "global_step": 68139, "epoch": 765, "lr": 8.539528518790928e-05}
{"train_loss": 0.170828714966774, "global_step": 68140, "epoch": 765, "lr": 8.539487571353508e-05}
{"train_loss": 0.26588964462280273, "global_step": 68141, "epoch": 765, "lr": 8.539446623440246e-05}
{"train_loss": 0.28399327397346497, "global_step": 68142, "epoch": 765, "lr": 8.539405675051149e-05}
{"train_loss": 0.33730077743530273, "global_step": 68143, "epoch": 765, "lr": 8.539364726186223e-05}
{"train_loss": 0.3102628290653229, "global_step": 68144, "epoch": 765, "lr": 8.539323776845472e-05}
{"train_loss": 0.22921395301818848, "global_step": 68145, "epoch": 765, "lr": 8.539282827028901e-05}
{"train_loss": 0.315470814704895, "global_step": 68146, "epoch": 765, "lr": 8.539241876736518e-05}
{"train_loss": 0.2812880873680115, "global_step": 68147, "epoch": 765, "lr": 8.539200925968327e-05}
{"train_loss": 0.3275691270828247, "global_step": 68148, "epoch": 765, "lr": 8.539159974724333e-05}
{"train_loss": 0.21346059441566467, "global_step": 68149, "epoch": 765, "lr": 8.539119023004543e-05}
{"train_loss": 0.2514086067676544, "global_step": 68150, "epoch": 765, "lr": 8.539078070808959e-05}
{"train_loss": 0.2153082937002182, "global_step": 68151, "epoch": 765, "lr": 8.539037118137591e-05}
{"train_loss": 0.21482035517692566, "global_step": 68152, "epoch": 765, "lr": 8.538996164990442e-05}
{"train_loss": 0.21389959752559662, "global_step": 68153, "epoch": 765, "lr": 8.538955211367519e-05}
{"train_loss": 0.3580170273780823, "global_step": 68154, "epoch": 765, "lr": 8.538914257268827e-05}
{"train_loss": 0.28259679675102234, "global_step": 68155, "epoch": 765, "lr": 8.53887330269437e-05}
{"train_loss": 0.15371279418468475, "global_step": 68156, "epoch": 765, "lr": 8.538832347644155e-05}
{"train_loss": 0.26878124475479126, "global_step": 68157, "epoch": 765, "lr": 8.538791392118186e-05}
{"train_loss": 0.2554589807987213, "global_step": 68158, "epoch": 765, "lr": 8.538750436116472e-05}
{"train_loss": 0.23789896070957184, "global_step": 68159, "epoch": 765, "lr": 8.538709479639014e-05}
{"train_loss": 0.2277614325284958, "global_step": 68160, "epoch": 765, "lr": 8.53866852268582e-05}
{"train_loss": 0.19706007838249207, "global_step": 68161, "epoch": 765, "lr": 8.538627565256895e-05}
{"train_loss": 0.18110373616218567, "global_step": 68162, "epoch": 765, "lr": 8.538586607352245e-05}
{"train_loss": 0.1885073184967041, "global_step": 68163, "epoch": 765, "lr": 8.538545648971876e-05}
{"train_loss": 0.22292464971542358, "global_step": 68164, "epoch": 765, "lr": 8.538504690115792e-05}
{"train_loss": 0.24553228914737701, "global_step": 68165, "epoch": 765, "lr": 8.538463730784e-05}
{"train_loss": 0.24766454100608826, "global_step": 68166, "epoch": 765, "lr": 8.538422770976505e-05}
{"train_loss": 0.26767656207084656, "global_step": 68167, "epoch": 765, "lr": 8.53838181069331e-05}
{"train_loss": 0.21310599148273468, "global_step": 68168, "epoch": 765, "lr": 8.538340849934425e-05}
{"train_loss": 0.244826540350914, "global_step": 68169, "epoch": 765, "lr": 8.53829988869985e-05}
{"train_loss": 0.28047314286231995, "global_step": 68170, "epoch": 765, "lr": 8.538258926989596e-05}
{"train_loss": 0.21313157677650452, "global_step": 68171, "epoch": 765, "lr": 8.538217964803667e-05}
{"train_loss": 0.22982807457447052, "global_step": 68172, "epoch": 765, "lr": 8.538177002142066e-05}
{"train_loss": 0.23232628204179614, "global_step": 68173, "epoch": 765, "lr": 8.5381360390048e-05, "val_loss": 3.6227259635925293, "train_action_mse_error": 27.008329391479492}
{"train_loss": 0.29007330536842346, "global_step": 68174, "epoch": 766, "lr": 8.538095075391876e-05}
{"train_loss": 0.1857805699110031, "global_step": 68175, "epoch": 766, "lr": 8.538054111303299e-05}
{"train_loss": 0.18840499222278595, "global_step": 68176, "epoch": 766, "lr": 8.538013146739072e-05}
{"train_loss": 0.1658407300710678, "global_step": 68177, "epoch": 766, "lr": 8.537972181699203e-05}
{"train_loss": 0.23570966720581055, "global_step": 68178, "epoch": 766, "lr": 8.537931216183697e-05}
{"train_loss": 0.3122658133506775, "global_step": 68179, "epoch": 766, "lr": 8.537890250192558e-05}
{"train_loss": 0.25152432918548584, "global_step": 68180, "epoch": 766, "lr": 8.537849283725794e-05}
{"train_loss": 0.2028147429227829, "global_step": 68181, "epoch": 766, "lr": 8.537808316783409e-05}
{"train_loss": 0.2387743592262268, "global_step": 68182, "epoch": 766, "lr": 8.537767349365408e-05}
{"train_loss": 0.21020257472991943, "global_step": 68183, "epoch": 766, "lr": 8.537726381471797e-05}
{"train_loss": 0.24124349653720856, "global_step": 68184, "epoch": 766, "lr": 8.537685413102584e-05}
{"train_loss": 0.26561081409454346, "global_step": 68185, "epoch": 766, "lr": 8.53764444425777e-05}
{"train_loss": 0.23371721804141998, "global_step": 68186, "epoch": 766, "lr": 8.537603474937366e-05}
{"train_loss": 0.2795542776584625, "global_step": 68187, "epoch": 766, "lr": 8.53756250514137e-05}
{"train_loss": 0.2035950869321823, "global_step": 68188, "epoch": 766, "lr": 8.537521534869795e-05}
{"train_loss": 0.23851017653942108, "global_step": 68189, "epoch": 766, "lr": 8.537480564122643e-05}
{"train_loss": 0.26930585503578186, "global_step": 68190, "epoch": 766, "lr": 8.537439592899918e-05}
{"train_loss": 0.18342112004756927, "global_step": 68191, "epoch": 766, "lr": 8.53739862120163e-05}
{"train_loss": 0.2022395133972168, "global_step": 68192, "epoch": 766, "lr": 8.537357649027781e-05}
{"train_loss": 0.2545323073863983, "global_step": 68193, "epoch": 766, "lr": 8.537316676378377e-05}
{"train_loss": 0.161988765001297, "global_step": 68194, "epoch": 766, "lr": 8.537275703253423e-05}
{"train_loss": 0.17516107857227325, "global_step": 68195, "epoch": 766, "lr": 8.537234729652927e-05}
{"train_loss": 0.14726820588111877, "global_step": 68196, "epoch": 766, "lr": 8.537193755576891e-05}
{"train_loss": 0.29586783051490784, "global_step": 68197, "epoch": 766, "lr": 8.537152781025323e-05}
{"train_loss": 0.23107077181339264, "global_step": 68198, "epoch": 766, "lr": 8.53711180599823e-05}
{"train_loss": 0.24271747469902039, "global_step": 68199, "epoch": 766, "lr": 8.537070830495613e-05}
{"train_loss": 0.27520880103111267, "global_step": 68200, "epoch": 766, "lr": 8.53702985451748e-05}
{"train_loss": 0.3214295506477356, "global_step": 68201, "epoch": 766, "lr": 8.536988878063839e-05}
{"train_loss": 0.19637572765350342, "global_step": 68202, "epoch": 766, "lr": 8.53694790113469e-05}
{"train_loss": 0.1916392743587494, "global_step": 68203, "epoch": 766, "lr": 8.536906923730044e-05}
{"train_loss": 0.13273602724075317, "global_step": 68204, "epoch": 766, "lr": 8.536865945849903e-05}
{"train_loss": 0.319551020860672, "global_step": 68205, "epoch": 766, "lr": 8.536824967494272e-05}
{"train_loss": 0.2657107412815094, "global_step": 68206, "epoch": 766, "lr": 8.53678398866316e-05}
{"train_loss": 0.21555624902248383, "global_step": 68207, "epoch": 766, "lr": 8.536743009356569e-05}
{"train_loss": 0.20248939096927643, "global_step": 68208, "epoch": 766, "lr": 8.536702029574506e-05}
{"train_loss": 0.21723681688308716, "global_step": 68209, "epoch": 766, "lr": 8.536661049316978e-05}
{"train_loss": 0.23475727438926697, "global_step": 68210, "epoch": 766, "lr": 8.536620068583988e-05}
{"train_loss": 0.1979946345090866, "global_step": 68211, "epoch": 766, "lr": 8.536579087375542e-05}
{"train_loss": 0.19450269639492035, "global_step": 68212, "epoch": 766, "lr": 8.536538105691647e-05}
{"train_loss": 0.22665810585021973, "global_step": 68213, "epoch": 766, "lr": 8.536497123532307e-05}
{"train_loss": 0.22438736259937286, "global_step": 68214, "epoch": 766, "lr": 8.536456140897529e-05}
{"train_loss": 0.247516468167305, "global_step": 68215, "epoch": 766, "lr": 8.536415157787317e-05}
{"train_loss": 0.21739791333675385, "global_step": 68216, "epoch": 766, "lr": 8.536374174201677e-05}
{"train_loss": 0.25371837615966797, "global_step": 68217, "epoch": 766, "lr": 8.536333190140615e-05}
{"train_loss": 0.2754063010215759, "global_step": 68218, "epoch": 766, "lr": 8.536292205604136e-05}
{"train_loss": 0.22525718808174133, "global_step": 68219, "epoch": 766, "lr": 8.536251220592244e-05}
{"train_loss": 0.19382353127002716, "global_step": 68220, "epoch": 766, "lr": 8.536210235104948e-05}
{"train_loss": 0.2647663950920105, "global_step": 68221, "epoch": 766, "lr": 8.536169249142252e-05}
{"train_loss": 0.19846531748771667, "global_step": 68222, "epoch": 766, "lr": 8.53612826270416e-05}
{"train_loss": 0.192947655916214, "global_step": 68223, "epoch": 766, "lr": 8.536087275790678e-05}
{"train_loss": 0.1691582202911377, "global_step": 68224, "epoch": 766, "lr": 8.536046288401814e-05}
{"train_loss": 0.30366286635398865, "global_step": 68225, "epoch": 766, "lr": 8.53600530053757e-05}
{"train_loss": 0.21971893310546875, "global_step": 68226, "epoch": 766, "lr": 8.535964312197955e-05}
{"train_loss": 0.1916716992855072, "global_step": 68227, "epoch": 766, "lr": 8.53592332338297e-05}
{"train_loss": 0.184929758310318, "global_step": 68228, "epoch": 766, "lr": 8.535882334092626e-05}
{"train_loss": 0.1581985056400299, "global_step": 68229, "epoch": 766, "lr": 8.535841344326925e-05}
{"train_loss": 0.14959879219532013, "global_step": 68230, "epoch": 766, "lr": 8.535800354085872e-05}
{"train_loss": 0.219216451048851, "global_step": 68231, "epoch": 766, "lr": 8.535759363369475e-05}
{"train_loss": 0.26342442631721497, "global_step": 68232, "epoch": 766, "lr": 8.535718372177736e-05}
{"train_loss": 0.23733605444431305, "global_step": 68233, "epoch": 766, "lr": 8.535677380510666e-05}
{"train_loss": 0.28054526448249817, "global_step": 68234, "epoch": 766, "lr": 8.535636388368265e-05}
{"train_loss": 0.29473942518234253, "global_step": 68235, "epoch": 766, "lr": 8.535595395750542e-05}
{"train_loss": 0.20934666693210602, "global_step": 68236, "epoch": 766, "lr": 8.535554402657502e-05}
{"train_loss": 0.26567813754081726, "global_step": 68237, "epoch": 766, "lr": 8.535513409089148e-05}
{"train_loss": 0.22497305274009705, "global_step": 68238, "epoch": 766, "lr": 8.535472415045487e-05}
{"train_loss": 0.2181408703327179, "global_step": 68239, "epoch": 766, "lr": 8.535431420526526e-05}
{"train_loss": 0.2366647720336914, "global_step": 68240, "epoch": 766, "lr": 8.535390425532269e-05}
{"train_loss": 0.2838507890701294, "global_step": 68241, "epoch": 766, "lr": 8.535349430062722e-05}
{"train_loss": 0.2018831968307495, "global_step": 68242, "epoch": 766, "lr": 8.535308434117892e-05}
{"train_loss": 0.19476374983787537, "global_step": 68243, "epoch": 766, "lr": 8.535267437697781e-05}
{"train_loss": 0.18980807065963745, "global_step": 68244, "epoch": 766, "lr": 8.535226440802395e-05}
{"train_loss": 0.16635163128376007, "global_step": 68245, "epoch": 766, "lr": 8.535185443431744e-05}
{"train_loss": 0.21718573570251465, "global_step": 68246, "epoch": 766, "lr": 8.535144445585827e-05}
{"train_loss": 0.20013946294784546, "global_step": 68247, "epoch": 766, "lr": 8.535103447264657e-05}
{"train_loss": 0.1729157716035843, "global_step": 68248, "epoch": 766, "lr": 8.535062448468232e-05}
{"train_loss": 0.20027193427085876, "global_step": 68249, "epoch": 766, "lr": 8.535021449196563e-05}
{"train_loss": 0.26960039138793945, "global_step": 68250, "epoch": 766, "lr": 8.534980449449652e-05}
{"train_loss": 0.13325099647045135, "global_step": 68251, "epoch": 766, "lr": 8.534939449227507e-05}
{"train_loss": 0.2025642842054367, "global_step": 68252, "epoch": 766, "lr": 8.534898448530132e-05}
{"train_loss": 0.20749348402023315, "global_step": 68253, "epoch": 766, "lr": 8.534857447357534e-05}
{"train_loss": 0.19681338965892792, "global_step": 68254, "epoch": 766, "lr": 8.534816445709717e-05}
{"train_loss": 0.21952024102210999, "global_step": 68255, "epoch": 766, "lr": 8.534775443586684e-05}
{"train_loss": 0.20649859309196472, "global_step": 68256, "epoch": 766, "lr": 8.534734440988446e-05}
{"train_loss": 0.2277066856622696, "global_step": 68257, "epoch": 766, "lr": 8.534693437915007e-05}
{"train_loss": 0.20107603073120117, "global_step": 68258, "epoch": 766, "lr": 8.534652434366371e-05}
{"train_loss": 0.17723770439624786, "global_step": 68259, "epoch": 766, "lr": 8.534611430342543e-05}
{"train_loss": 0.18417638540267944, "global_step": 68260, "epoch": 766, "lr": 8.53457042584353e-05}
{"train_loss": 0.22530010342597961, "global_step": 68261, "epoch": 766, "lr": 8.53452942086934e-05}
{"train_loss": 0.22132007961862543, "global_step": 68262, "epoch": 766, "lr": 8.534488415419973e-05, "val_loss": 3.741931915283203}
{"train_loss": 0.18173295259475708, "global_step": 68263, "epoch": 767, "lr": 8.534447409495436e-05}
{"train_loss": 0.18426330387592316, "global_step": 68264, "epoch": 767, "lr": 8.534406403095736e-05}
{"train_loss": 0.215862438082695, "global_step": 68265, "epoch": 767, "lr": 8.534365396220881e-05}
{"train_loss": 0.18836461007595062, "global_step": 68266, "epoch": 767, "lr": 8.534324388870871e-05}
{"train_loss": 0.16320830583572388, "global_step": 68267, "epoch": 767, "lr": 8.534283381045713e-05}
{"train_loss": 0.192612886428833, "global_step": 68268, "epoch": 767, "lr": 8.534242372745416e-05}
{"train_loss": 0.2118070125579834, "global_step": 68269, "epoch": 767, "lr": 8.534201363969983e-05}
{"train_loss": 0.2179979383945465, "global_step": 68270, "epoch": 767, "lr": 8.534160354719418e-05}
{"train_loss": 0.2557806074619293, "global_step": 68271, "epoch": 767, "lr": 8.534119344993731e-05}
{"train_loss": 0.2328188121318817, "global_step": 68272, "epoch": 767, "lr": 8.534078334792923e-05}
{"train_loss": 0.20342575013637543, "global_step": 68273, "epoch": 767, "lr": 8.534037324117002e-05}
{"train_loss": 0.24258197844028473, "global_step": 68274, "epoch": 767, "lr": 8.53399631296597e-05}
{"train_loss": 0.19158007204532623, "global_step": 68275, "epoch": 767, "lr": 8.533955301339837e-05}
{"train_loss": 0.33051925897598267, "global_step": 68276, "epoch": 767, "lr": 8.533914289238609e-05}
{"train_loss": 0.29113268852233887, "global_step": 68277, "epoch": 767, "lr": 8.533873276662288e-05}
{"train_loss": 0.23897986114025116, "global_step": 68278, "epoch": 767, "lr": 8.533832263610879e-05}
{"train_loss": 0.2597876787185669, "global_step": 68279, "epoch": 767, "lr": 8.533791250084391e-05}
{"train_loss": 0.18049907684326172, "global_step": 68280, "epoch": 767, "lr": 8.533750236082828e-05}
{"train_loss": 0.21334603428840637, "global_step": 68281, "epoch": 767, "lr": 8.533709221606195e-05}
{"train_loss": 0.19468379020690918, "global_step": 68282, "epoch": 767, "lr": 8.533668206654498e-05}
{"train_loss": 0.2097187042236328, "global_step": 68283, "epoch": 767, "lr": 8.533627191227743e-05}
{"train_loss": 0.27648505568504333, "global_step": 68284, "epoch": 767, "lr": 8.533586175325934e-05}
{"train_loss": 0.15502889454364777, "global_step": 68285, "epoch": 767, "lr": 8.533545158949077e-05}
{"train_loss": 0.21419891715049744, "global_step": 68286, "epoch": 767, "lr": 8.533504142097178e-05}
{"train_loss": 0.2597714066505432, "global_step": 68287, "epoch": 767, "lr": 8.533463124770242e-05}
{"train_loss": 0.21689370274543762, "global_step": 68288, "epoch": 767, "lr": 8.533422106968276e-05}
{"train_loss": 0.25838562846183777, "global_step": 68289, "epoch": 767, "lr": 8.533381088691285e-05}
{"train_loss": 0.16136005520820618, "global_step": 68290, "epoch": 767, "lr": 8.533340069939272e-05}
{"train_loss": 0.15908463299274445, "global_step": 68291, "epoch": 767, "lr": 8.533299050712246e-05}
{"train_loss": 0.18567223846912384, "global_step": 68292, "epoch": 767, "lr": 8.533258031010211e-05}
{"train_loss": 0.23776842653751373, "global_step": 68293, "epoch": 767, "lr": 8.533217010833172e-05}
{"train_loss": 0.25150156021118164, "global_step": 68294, "epoch": 767, "lr": 8.533175990181134e-05}
{"train_loss": 0.24125026166439056, "global_step": 68295, "epoch": 767, "lr": 8.533134969054106e-05}
{"train_loss": 0.2175247073173523, "global_step": 68296, "epoch": 767, "lr": 8.53309394745209e-05}
{"train_loss": 0.2828502058982849, "global_step": 68297, "epoch": 767, "lr": 8.533052925375093e-05}
{"train_loss": 0.23100163042545319, "global_step": 68298, "epoch": 767, "lr": 8.533011902823118e-05}
{"train_loss": 0.23944413661956787, "global_step": 68299, "epoch": 767, "lr": 8.532970879796177e-05}
{"train_loss": 0.27331554889678955, "global_step": 68300, "epoch": 767, "lr": 8.532929856294268e-05}
{"train_loss": 0.2495145946741104, "global_step": 68301, "epoch": 767, "lr": 8.532888832317398e-05}
{"train_loss": 0.2349592000246048, "global_step": 68302, "epoch": 767, "lr": 8.532847807865576e-05}
{"train_loss": 0.2480676770210266, "global_step": 68303, "epoch": 767, "lr": 8.532806782938807e-05}
{"train_loss": 0.23548217117786407, "global_step": 68304, "epoch": 767, "lr": 8.532765757537094e-05}
{"train_loss": 0.23349198698997498, "global_step": 68305, "epoch": 767, "lr": 8.532724731660443e-05}
{"train_loss": 0.23748786747455597, "global_step": 68306, "epoch": 767, "lr": 8.532683705308862e-05}
{"train_loss": 0.22502917051315308, "global_step": 68307, "epoch": 767, "lr": 8.532642678482352e-05}
{"train_loss": 0.2097640186548233, "global_step": 68308, "epoch": 767, "lr": 8.532601651180923e-05}
{"train_loss": 0.2939808666706085, "global_step": 68309, "epoch": 767, "lr": 8.532560623404578e-05}
{"train_loss": 0.15265829861164093, "global_step": 68310, "epoch": 767, "lr": 8.532519595153324e-05}
{"train_loss": 0.20291727781295776, "global_step": 68311, "epoch": 767, "lr": 8.532478566427166e-05}
{"train_loss": 0.30066192150115967, "global_step": 68312, "epoch": 767, "lr": 8.532437537226109e-05}
{"train_loss": 0.1918838620185852, "global_step": 68313, "epoch": 767, "lr": 8.532396507550159e-05}
{"train_loss": 0.3209332823753357, "global_step": 68314, "epoch": 767, "lr": 8.53235547739932e-05}
{"train_loss": 0.2229633778333664, "global_step": 68315, "epoch": 767, "lr": 8.532314446773602e-05}
{"train_loss": 0.1709435135126114, "global_step": 68316, "epoch": 767, "lr": 8.532273415673005e-05}
{"train_loss": 0.2448996901512146, "global_step": 68317, "epoch": 767, "lr": 8.532232384097536e-05}
{"train_loss": 0.2470228224992752, "global_step": 68318, "epoch": 767, "lr": 8.532191352047203e-05}
{"train_loss": 0.18697324395179749, "global_step": 68319, "epoch": 767, "lr": 8.53215031952201e-05}
{"train_loss": 0.19721925258636475, "global_step": 68320, "epoch": 767, "lr": 8.532109286521962e-05}
{"train_loss": 0.18802501261234283, "global_step": 68321, "epoch": 767, "lr": 8.532068253047065e-05}
{"train_loss": 0.22153085470199585, "global_step": 68322, "epoch": 767, "lr": 8.532027219097325e-05}
{"train_loss": 0.3205863833427429, "global_step": 68323, "epoch": 767, "lr": 8.531986184672746e-05}
{"train_loss": 0.28347763419151306, "global_step": 68324, "epoch": 767, "lr": 8.531945149773335e-05}
{"train_loss": 0.24035613238811493, "global_step": 68325, "epoch": 767, "lr": 8.531904114399098e-05}
{"train_loss": 0.3195364773273468, "global_step": 68326, "epoch": 767, "lr": 8.531863078550037e-05}
{"train_loss": 0.23242689669132233, "global_step": 68327, "epoch": 767, "lr": 8.531822042226163e-05}
{"train_loss": 0.2551811933517456, "global_step": 68328, "epoch": 767, "lr": 8.531781005427477e-05}
{"train_loss": 0.3192983567714691, "global_step": 68329, "epoch": 767, "lr": 8.531739968153985e-05}
{"train_loss": 0.178678959608078, "global_step": 68330, "epoch": 767, "lr": 8.531698930405696e-05}
{"train_loss": 0.19306263327598572, "global_step": 68331, "epoch": 767, "lr": 8.531657892182612e-05}
{"train_loss": 0.2573401927947998, "global_step": 68332, "epoch": 767, "lr": 8.531616853484738e-05}
{"train_loss": 0.22933700680732727, "global_step": 68333, "epoch": 767, "lr": 8.531575814312084e-05}
{"train_loss": 0.22938498854637146, "global_step": 68334, "epoch": 767, "lr": 8.531534774664651e-05}
{"train_loss": 0.27825167775154114, "global_step": 68335, "epoch": 767, "lr": 8.531493734542447e-05}
{"train_loss": 0.27192339301109314, "global_step": 68336, "epoch": 767, "lr": 8.531452693945476e-05}
{"train_loss": 0.3105275332927704, "global_step": 68337, "epoch": 767, "lr": 8.531411652873745e-05}
{"train_loss": 0.19885365664958954, "global_step": 68338, "epoch": 767, "lr": 8.531370611327259e-05}
{"train_loss": 0.36651554703712463, "global_step": 68339, "epoch": 767, "lr": 8.531329569306021e-05}
{"train_loss": 0.1626543402671814, "global_step": 68340, "epoch": 767, "lr": 8.531288526810042e-05}
{"train_loss": 0.2301599234342575, "global_step": 68341, "epoch": 767, "lr": 8.531247483839321e-05}
{"train_loss": 0.2897595167160034, "global_step": 68342, "epoch": 767, "lr": 8.53120644039387e-05}
{"train_loss": 0.17561982572078705, "global_step": 68343, "epoch": 767, "lr": 8.531165396473689e-05}
{"train_loss": 0.17520040273666382, "global_step": 68344, "epoch": 767, "lr": 8.531124352078788e-05}
{"train_loss": 0.222892165184021, "global_step": 68345, "epoch": 767, "lr": 8.531083307209169e-05}
{"train_loss": 0.24183036386966705, "global_step": 68346, "epoch": 767, "lr": 8.531042261864838e-05}
{"train_loss": 0.1628551483154297, "global_step": 68347, "epoch": 767, "lr": 8.531001216045802e-05}
{"train_loss": 0.2032385766506195, "global_step": 68348, "epoch": 767, "lr": 8.530960169752067e-05}
{"train_loss": 0.1916288435459137, "global_step": 68349, "epoch": 767, "lr": 8.530919122983636e-05}
{"train_loss": 0.23710526525974274, "global_step": 68350, "epoch": 767, "lr": 8.530878075740516e-05}
{"train_loss": 0.23072533928946162, "global_step": 68351, "epoch": 767, "lr": 8.530837028022714e-05, "val_loss": 3.678212881088257}
{"train_loss": 0.22016355395317078, "global_step": 68352, "epoch": 768, "lr": 8.530795979830233e-05}
{"train_loss": 0.23926153779029846, "global_step": 68353, "epoch": 768, "lr": 8.53075493116308e-05}
{"train_loss": 0.1663609743118286, "global_step": 68354, "epoch": 768, "lr": 8.53071388202126e-05}
{"train_loss": 0.22056034207344055, "global_step": 68355, "epoch": 768, "lr": 8.530672832404779e-05}
{"train_loss": 0.18745143711566925, "global_step": 68356, "epoch": 768, "lr": 8.53063178231364e-05}
{"train_loss": 0.23958039283752441, "global_step": 68357, "epoch": 768, "lr": 8.530590731747851e-05}
{"train_loss": 0.13159780204296112, "global_step": 68358, "epoch": 768, "lr": 8.53054968070742e-05}
{"train_loss": 0.25137802958488464, "global_step": 68359, "epoch": 768, "lr": 8.530508629192346e-05}
{"train_loss": 0.22215108573436737, "global_step": 68360, "epoch": 768, "lr": 8.530467577202641e-05}
{"train_loss": 0.23977380990982056, "global_step": 68361, "epoch": 768, "lr": 8.530426524738307e-05}
{"train_loss": 0.1586143672466278, "global_step": 68362, "epoch": 768, "lr": 8.530385471799349e-05}
{"train_loss": 0.16988742351531982, "global_step": 68363, "epoch": 768, "lr": 8.530344418385774e-05}
{"train_loss": 0.21293511986732483, "global_step": 68364, "epoch": 768, "lr": 8.530303364497588e-05}
{"train_loss": 0.1979781836271286, "global_step": 68365, "epoch": 768, "lr": 8.530262310134795e-05}
{"train_loss": 0.19850243628025055, "global_step": 68366, "epoch": 768, "lr": 8.530221255297402e-05}
{"train_loss": 0.22543984651565552, "global_step": 68367, "epoch": 768, "lr": 8.530180199985413e-05}
{"train_loss": 0.19010135531425476, "global_step": 68368, "epoch": 768, "lr": 8.530139144198835e-05}
{"train_loss": 0.23001834750175476, "global_step": 68369, "epoch": 768, "lr": 8.530098087937672e-05}
{"train_loss": 0.27488648891448975, "global_step": 68370, "epoch": 768, "lr": 8.53005703120193e-05}
{"train_loss": 0.18740196526050568, "global_step": 68371, "epoch": 768, "lr": 8.530015973991616e-05}
{"train_loss": 0.17893856763839722, "global_step": 68372, "epoch": 768, "lr": 8.529974916306733e-05}
{"train_loss": 0.1458863765001297, "global_step": 68373, "epoch": 768, "lr": 8.52993385814729e-05}
{"train_loss": 0.28067290782928467, "global_step": 68374, "epoch": 768, "lr": 8.52989279951329e-05}
{"train_loss": 0.21637780964374542, "global_step": 68375, "epoch": 768, "lr": 8.529851740404738e-05}
{"train_loss": 0.17999593913555145, "global_step": 68376, "epoch": 768, "lr": 8.52981068082164e-05}
{"train_loss": 0.2545669972896576, "global_step": 68377, "epoch": 768, "lr": 8.529769620764003e-05}
{"train_loss": 0.28129443526268005, "global_step": 68378, "epoch": 768, "lr": 8.52972856023183e-05}
{"train_loss": 0.2776166498661041, "global_step": 68379, "epoch": 768, "lr": 8.529687499225131e-05}
{"train_loss": 0.2897570729255676, "global_step": 68380, "epoch": 768, "lr": 8.529646437743907e-05}
{"train_loss": 0.2633770704269409, "global_step": 68381, "epoch": 768, "lr": 8.529605375788164e-05}
{"train_loss": 0.1658506542444229, "global_step": 68382, "epoch": 768, "lr": 8.52956431335791e-05}
{"train_loss": 0.1263773888349533, "global_step": 68383, "epoch": 768, "lr": 8.529523250453149e-05}
{"train_loss": 0.1874285638332367, "global_step": 68384, "epoch": 768, "lr": 8.529482187073885e-05}
{"train_loss": 0.2537364661693573, "global_step": 68385, "epoch": 768, "lr": 8.529441123220126e-05}
{"train_loss": 0.2317034751176834, "global_step": 68386, "epoch": 768, "lr": 8.529400058891877e-05}
{"train_loss": 0.18424274027347565, "global_step": 68387, "epoch": 768, "lr": 8.529358994089142e-05}
{"train_loss": 0.18732738494873047, "global_step": 68388, "epoch": 768, "lr": 8.52931792881193e-05}
{"train_loss": 0.21500861644744873, "global_step": 68389, "epoch": 768, "lr": 8.529276863060243e-05}
{"train_loss": 0.23311583697795868, "global_step": 68390, "epoch": 768, "lr": 8.529235796834088e-05}
{"train_loss": 0.1754748821258545, "global_step": 68391, "epoch": 768, "lr": 8.52919473013347e-05}
{"train_loss": 0.2551272511482239, "global_step": 68392, "epoch": 768, "lr": 8.529153662958394e-05}
{"train_loss": 0.14054161310195923, "global_step": 68393, "epoch": 768, "lr": 8.529112595308868e-05}
{"train_loss": 0.19301998615264893, "global_step": 68394, "epoch": 768, "lr": 8.529071527184896e-05}
{"train_loss": 0.3331572413444519, "global_step": 68395, "epoch": 768, "lr": 8.529030458586481e-05}
{"train_loss": 0.20638886094093323, "global_step": 68396, "epoch": 768, "lr": 8.528989389513633e-05}
{"train_loss": 0.20457758009433746, "global_step": 68397, "epoch": 768, "lr": 8.528948319966354e-05}
{"train_loss": 0.3638211786746979, "global_step": 68398, "epoch": 768, "lr": 8.528907249944651e-05}
{"train_loss": 0.12298377603292465, "global_step": 68399, "epoch": 768, "lr": 8.52886617944853e-05}
{"train_loss": 0.17190824449062347, "global_step": 68400, "epoch": 768, "lr": 8.528825108477996e-05}
{"train_loss": 0.15925884246826172, "global_step": 68401, "epoch": 768, "lr": 8.528784037033053e-05}
{"train_loss": 0.21894915401935577, "global_step": 68402, "epoch": 768, "lr": 8.52874296511371e-05}
{"train_loss": 0.20547989010810852, "global_step": 68403, "epoch": 768, "lr": 8.52870189271997e-05}
{"train_loss": 0.19265854358673096, "global_step": 68404, "epoch": 768, "lr": 8.528660819851839e-05}
{"train_loss": 0.14534859359264374, "global_step": 68405, "epoch": 768, "lr": 8.528619746509324e-05}
{"train_loss": 0.2862898111343384, "global_step": 68406, "epoch": 768, "lr": 8.528578672692427e-05}
{"train_loss": 0.2552822530269623, "global_step": 68407, "epoch": 768, "lr": 8.528537598401155e-05}
{"train_loss": 0.20192687213420868, "global_step": 68408, "epoch": 768, "lr": 8.528496523635515e-05}
{"train_loss": 0.2716006934642792, "global_step": 68409, "epoch": 768, "lr": 8.528455448395512e-05}
{"train_loss": 0.23229533433914185, "global_step": 68410, "epoch": 768, "lr": 8.52841437268115e-05}
{"train_loss": 0.19334407150745392, "global_step": 68411, "epoch": 768, "lr": 8.528373296492438e-05}
{"train_loss": 0.19699923694133759, "global_step": 68412, "epoch": 768, "lr": 8.528332219829378e-05}
{"train_loss": 0.2340320646762848, "global_step": 68413, "epoch": 768, "lr": 8.528291142691976e-05}
{"train_loss": 0.25002890825271606, "global_step": 68414, "epoch": 768, "lr": 8.528250065080239e-05}
{"train_loss": 0.22477281093597412, "global_step": 68415, "epoch": 768, "lr": 8.528208986994173e-05}
{"train_loss": 0.35636940598487854, "global_step": 68416, "epoch": 768, "lr": 8.52816790843378e-05}
{"train_loss": 0.17207324504852295, "global_step": 68417, "epoch": 768, "lr": 8.52812682939907e-05}
{"train_loss": 0.17828041315078735, "global_step": 68418, "epoch": 768, "lr": 8.528085749890044e-05}
{"train_loss": 0.21094700694084167, "global_step": 68419, "epoch": 768, "lr": 8.528044669906712e-05}
{"train_loss": 0.18287624418735504, "global_step": 68420, "epoch": 768, "lr": 8.528003589449076e-05}
{"train_loss": 0.23150862753391266, "global_step": 68421, "epoch": 768, "lr": 8.527962508517142e-05}
{"train_loss": 0.22492529451847076, "global_step": 68422, "epoch": 768, "lr": 8.527921427110919e-05}
{"train_loss": 0.21580125391483307, "global_step": 68423, "epoch": 768, "lr": 8.52788034523041e-05}
{"train_loss": 0.28597164154052734, "global_step": 68424, "epoch": 768, "lr": 8.527839262875619e-05}
{"train_loss": 0.1575896292924881, "global_step": 68425, "epoch": 768, "lr": 8.527798180046553e-05}
{"train_loss": 0.1693722903728485, "global_step": 68426, "epoch": 768, "lr": 8.527757096743218e-05}
{"train_loss": 0.21925553679466248, "global_step": 68427, "epoch": 768, "lr": 8.527716012965619e-05}
{"train_loss": 0.1624225676059723, "global_step": 68428, "epoch": 768, "lr": 8.52767492871376e-05}
{"train_loss": 0.20311136543750763, "global_step": 68429, "epoch": 768, "lr": 8.527633843987651e-05}
{"train_loss": 0.2199687361717224, "global_step": 68430, "epoch": 768, "lr": 8.527592758787293e-05}
{"train_loss": 0.21205756068229675, "global_step": 68431, "epoch": 768, "lr": 8.527551673112692e-05}
{"train_loss": 0.29959726333618164, "global_step": 68432, "epoch": 768, "lr": 8.527510586963856e-05}
{"train_loss": 0.21724393963813782, "global_step": 68433, "epoch": 768, "lr": 8.52746950034079e-05}
{"train_loss": 0.18052639067173004, "global_step": 68434, "epoch": 768, "lr": 8.527428413243498e-05}
{"train_loss": 0.25879064202308655, "global_step": 68435, "epoch": 768, "lr": 8.527387325671987e-05}
{"train_loss": 0.2367897629737854, "global_step": 68436, "epoch": 768, "lr": 8.52734623762626e-05}
{"train_loss": 0.20469218492507935, "global_step": 68437, "epoch": 768, "lr": 8.527305149106325e-05}
{"train_loss": 0.2603956162929535, "global_step": 68438, "epoch": 768, "lr": 8.527264060112188e-05}
{"train_loss": 0.1818067580461502, "global_step": 68439, "epoch": 768, "lr": 8.527222970643851e-05}
{"train_loss": 0.2162167090880737, "global_step": 68440, "epoch": 768, "lr": 8.527181880701323e-05, "val_loss": 3.6252315044403076}
{"train_loss": 0.18048802018165588, "global_step": 68441, "epoch": 769, "lr": 8.52714079028461e-05}
{"train_loss": 0.16848649084568024, "global_step": 68442, "epoch": 769, "lr": 8.527099699393714e-05}
{"train_loss": 0.14110416173934937, "global_step": 68443, "epoch": 769, "lr": 8.527058608028644e-05}
{"train_loss": 0.2515750527381897, "global_step": 68444, "epoch": 769, "lr": 8.527017516189401e-05}
{"train_loss": 0.2084864377975464, "global_step": 68445, "epoch": 769, "lr": 8.526976423875996e-05}
{"train_loss": 0.23136433959007263, "global_step": 68446, "epoch": 769, "lr": 8.526935331088432e-05}
{"train_loss": 0.26711419224739075, "global_step": 68447, "epoch": 769, "lr": 8.526894237826715e-05}
{"train_loss": 0.23871365189552307, "global_step": 68448, "epoch": 769, "lr": 8.526853144090848e-05}
{"train_loss": 0.25805461406707764, "global_step": 68449, "epoch": 769, "lr": 8.526812049880838e-05}
{"train_loss": 0.24052080512046814, "global_step": 68450, "epoch": 769, "lr": 8.526770955196694e-05}
{"train_loss": 0.15212823450565338, "global_step": 68451, "epoch": 769, "lr": 8.526729860038419e-05}
{"train_loss": 0.21556471288204193, "global_step": 68452, "epoch": 769, "lr": 8.526688764406015e-05}
{"train_loss": 0.17956861853599548, "global_step": 68453, "epoch": 769, "lr": 8.526647668299492e-05}
{"train_loss": 0.21221177279949188, "global_step": 68454, "epoch": 769, "lr": 8.526606571718855e-05}
{"train_loss": 0.1296100914478302, "global_step": 68455, "epoch": 769, "lr": 8.526565474664108e-05}
{"train_loss": 0.20447920262813568, "global_step": 68456, "epoch": 769, "lr": 8.526524377135257e-05}
{"train_loss": 0.28209003806114197, "global_step": 68457, "epoch": 769, "lr": 8.526483279132307e-05}
{"train_loss": 0.216741144657135, "global_step": 68458, "epoch": 769, "lr": 8.526442180655266e-05}
{"train_loss": 0.24115459620952606, "global_step": 68459, "epoch": 769, "lr": 8.526401081704137e-05}
{"train_loss": 0.22543053328990936, "global_step": 68460, "epoch": 769, "lr": 8.526359982278927e-05}
{"train_loss": 0.2570951282978058, "global_step": 68461, "epoch": 769, "lr": 8.52631888237964e-05}
{"train_loss": 0.19745226204395294, "global_step": 68462, "epoch": 769, "lr": 8.526277782006285e-05}
{"train_loss": 0.18404550850391388, "global_step": 68463, "epoch": 769, "lr": 8.526236681158862e-05}
{"train_loss": 0.23013076186180115, "global_step": 68464, "epoch": 769, "lr": 8.526195579837379e-05}
{"train_loss": 0.23421600461006165, "global_step": 68465, "epoch": 769, "lr": 8.526154478041844e-05}
{"train_loss": 0.16398374736309052, "global_step": 68466, "epoch": 769, "lr": 8.52611337577226e-05}
{"train_loss": 0.28691086173057556, "global_step": 68467, "epoch": 769, "lr": 8.526072273028634e-05}
{"train_loss": 0.17846055328845978, "global_step": 68468, "epoch": 769, "lr": 8.526031169810969e-05}
{"train_loss": 0.1486607939004898, "global_step": 68469, "epoch": 769, "lr": 8.525990066119271e-05}
{"train_loss": 0.21212105453014374, "global_step": 68470, "epoch": 769, "lr": 8.52594896195355e-05}
{"train_loss": 0.1890326887369156, "global_step": 68471, "epoch": 769, "lr": 8.525907857313805e-05}
{"train_loss": 0.26085612177848816, "global_step": 68472, "epoch": 769, "lr": 8.525866752200047e-05}
{"train_loss": 0.20421630144119263, "global_step": 68473, "epoch": 769, "lr": 8.525825646612277e-05}
{"train_loss": 0.1965828835964203, "global_step": 68474, "epoch": 769, "lr": 8.525784540550503e-05}
{"train_loss": 0.21457324922084808, "global_step": 68475, "epoch": 769, "lr": 8.525743434014732e-05}
{"train_loss": 0.20592787861824036, "global_step": 68476, "epoch": 769, "lr": 8.525702327004967e-05}
{"train_loss": 0.26822757720947266, "global_step": 68477, "epoch": 769, "lr": 8.525661219521214e-05}
{"train_loss": 0.20740774273872375, "global_step": 68478, "epoch": 769, "lr": 8.525620111563479e-05}
{"train_loss": 0.23221395909786224, "global_step": 68479, "epoch": 769, "lr": 8.525579003131768e-05}
{"train_loss": 0.2577737867832184, "global_step": 68480, "epoch": 769, "lr": 8.525537894226084e-05}
{"train_loss": 0.3009410798549652, "global_step": 68481, "epoch": 769, "lr": 8.525496784846436e-05}
{"train_loss": 0.35406196117401123, "global_step": 68482, "epoch": 769, "lr": 8.525455674992827e-05}
{"train_loss": 0.20344427227973938, "global_step": 68483, "epoch": 769, "lr": 8.525414564665264e-05}
{"train_loss": 0.39372143149375916, "global_step": 68484, "epoch": 769, "lr": 8.525373453863752e-05}
{"train_loss": 0.24917718768119812, "global_step": 68485, "epoch": 769, "lr": 8.525332342588296e-05}
{"train_loss": 0.21123650670051575, "global_step": 68486, "epoch": 769, "lr": 8.525291230838903e-05}
{"train_loss": 0.21072076261043549, "global_step": 68487, "epoch": 769, "lr": 8.525250118615577e-05}
{"train_loss": 0.29255226254463196, "global_step": 68488, "epoch": 769, "lr": 8.525209005918324e-05}
{"train_loss": 0.2305137664079666, "global_step": 68489, "epoch": 769, "lr": 8.525167892747151e-05}
{"train_loss": 0.15846551954746246, "global_step": 68490, "epoch": 769, "lr": 8.52512677910206e-05}
{"train_loss": 0.1905454695224762, "global_step": 68491, "epoch": 769, "lr": 8.525085664983059e-05}
{"train_loss": 0.1773659586906433, "global_step": 68492, "epoch": 769, "lr": 8.525044550390154e-05}
{"train_loss": 0.23179158568382263, "global_step": 68493, "epoch": 769, "lr": 8.525003435323349e-05}
{"train_loss": 0.2759740352630615, "global_step": 68494, "epoch": 769, "lr": 8.52496231978265e-05}
{"train_loss": 0.2560306191444397, "global_step": 68495, "epoch": 769, "lr": 8.524921203768064e-05}
{"train_loss": 0.2728514075279236, "global_step": 68496, "epoch": 769, "lr": 8.524880087279595e-05}
{"train_loss": 0.16370733082294464, "global_step": 68497, "epoch": 769, "lr": 8.524838970317248e-05}
{"train_loss": 0.22394973039627075, "global_step": 68498, "epoch": 769, "lr": 8.52479785288103e-05}
{"train_loss": 0.24290601909160614, "global_step": 68499, "epoch": 769, "lr": 8.524756734970947e-05}
{"train_loss": 0.15251952409744263, "global_step": 68500, "epoch": 769, "lr": 8.524715616587002e-05}
{"train_loss": 0.2480650693178177, "global_step": 68501, "epoch": 769, "lr": 8.524674497729202e-05}
{"train_loss": 0.23896092176437378, "global_step": 68502, "epoch": 769, "lr": 8.524633378397553e-05}
{"train_loss": 0.2174852192401886, "global_step": 68503, "epoch": 769, "lr": 8.524592258592058e-05}
{"train_loss": 0.2116978019475937, "global_step": 68504, "epoch": 769, "lr": 8.524551138312729e-05}
{"train_loss": 0.21479709446430206, "global_step": 68505, "epoch": 769, "lr": 8.524510017559564e-05}
{"train_loss": 0.22312334179878235, "global_step": 68506, "epoch": 769, "lr": 8.524468896332572e-05}
{"train_loss": 0.27652159333229065, "global_step": 68507, "epoch": 769, "lr": 8.524427774631758e-05}
{"train_loss": 0.3722430467605591, "global_step": 68508, "epoch": 769, "lr": 8.524386652457126e-05}
{"train_loss": 0.22928451001644135, "global_step": 68509, "epoch": 769, "lr": 8.524345529808685e-05}
{"train_loss": 0.22128432989120483, "global_step": 68510, "epoch": 769, "lr": 8.524304406686439e-05}
{"train_loss": 0.22451342642307281, "global_step": 68511, "epoch": 769, "lr": 8.524263283090393e-05}
{"train_loss": 0.1686336100101471, "global_step": 68512, "epoch": 769, "lr": 8.524222159020552e-05}
{"train_loss": 0.2898680567741394, "global_step": 68513, "epoch": 769, "lr": 8.524181034476922e-05}
{"train_loss": 0.20788224041461945, "global_step": 68514, "epoch": 769, "lr": 8.524139909459509e-05}
{"train_loss": 0.27368542551994324, "global_step": 68515, "epoch": 769, "lr": 8.52409878396832e-05}
{"train_loss": 0.2081812471151352, "global_step": 68516, "epoch": 769, "lr": 8.524057658003357e-05}
{"train_loss": 0.23085929453372955, "global_step": 68517, "epoch": 769, "lr": 8.52401653156463e-05}
{"train_loss": 0.2078305184841156, "global_step": 68518, "epoch": 769, "lr": 8.52397540465214e-05}
{"train_loss": 0.35092684626579285, "global_step": 68519, "epoch": 769, "lr": 8.523934277265895e-05}
{"train_loss": 0.1641867756843567, "global_step": 68520, "epoch": 769, "lr": 8.5238931494059e-05}
{"train_loss": 0.23952053487300873, "global_step": 68521, "epoch": 769, "lr": 8.523852021072159e-05}
{"train_loss": 0.3043094873428345, "global_step": 68522, "epoch": 769, "lr": 8.52381089226468e-05}
{"train_loss": 0.22655853629112244, "global_step": 68523, "epoch": 769, "lr": 8.523769762983468e-05}
{"train_loss": 0.2294120490550995, "global_step": 68524, "epoch": 769, "lr": 8.523728633228527e-05}
{"train_loss": 0.1836167424917221, "global_step": 68525, "epoch": 769, "lr": 8.523687502999864e-05}
{"train_loss": 0.22653727233409882, "global_step": 68526, "epoch": 769, "lr": 8.523646372297486e-05}
{"train_loss": 0.19650094211101532, "global_step": 68527, "epoch": 769, "lr": 8.523605241121395e-05}
{"train_loss": 0.12661206722259521, "global_step": 68528, "epoch": 769, "lr": 8.523564109471598e-05}
{"train_loss": 0.22573881119154812, "global_step": 68529, "epoch": 769, "lr": 8.523522977348102e-05, "val_loss": 3.608940601348877}
{"train_loss": 0.19956514239311218, "global_step": 68530, "epoch": 770, "lr": 8.52348184475091e-05}
{"train_loss": 0.2733939588069916, "global_step": 68531, "epoch": 770, "lr": 8.523440711680029e-05}
{"train_loss": 0.2298395037651062, "global_step": 68532, "epoch": 770, "lr": 8.523399578135465e-05}
{"train_loss": 0.23031966388225555, "global_step": 68533, "epoch": 770, "lr": 8.523358444117222e-05}
{"train_loss": 0.27029865980148315, "global_step": 68534, "epoch": 770, "lr": 8.523317309625307e-05}
{"train_loss": 0.1925506740808487, "global_step": 68535, "epoch": 770, "lr": 8.523276174659725e-05}
{"train_loss": 0.2536013126373291, "global_step": 68536, "epoch": 770, "lr": 8.523235039220482e-05}
{"train_loss": 0.2886592745780945, "global_step": 68537, "epoch": 770, "lr": 8.523193903307582e-05}
{"train_loss": 0.26363039016723633, "global_step": 68538, "epoch": 770, "lr": 8.523152766921033e-05}
{"train_loss": 0.23049741983413696, "global_step": 68539, "epoch": 770, "lr": 8.523111630060837e-05}
{"train_loss": 0.22629863023757935, "global_step": 68540, "epoch": 770, "lr": 8.523070492727004e-05}
{"train_loss": 0.37028858065605164, "global_step": 68541, "epoch": 770, "lr": 8.523029354919534e-05}
{"train_loss": 0.29746827483177185, "global_step": 68542, "epoch": 770, "lr": 8.522988216638438e-05}
{"train_loss": 0.2347339540719986, "global_step": 68543, "epoch": 770, "lr": 8.522947077883719e-05}
{"train_loss": 0.17009125649929047, "global_step": 68544, "epoch": 770, "lr": 8.522905938655381e-05}
{"train_loss": 0.22408680617809296, "global_step": 68545, "epoch": 770, "lr": 8.522864798953432e-05}
{"train_loss": 0.24705296754837036, "global_step": 68546, "epoch": 770, "lr": 8.522823658777878e-05}
{"train_loss": 0.20641234517097473, "global_step": 68547, "epoch": 770, "lr": 8.522782518128722e-05}
{"train_loss": 0.22256191074848175, "global_step": 68548, "epoch": 770, "lr": 8.522741377005972e-05}
{"train_loss": 0.24953758716583252, "global_step": 68549, "epoch": 770, "lr": 8.52270023540963e-05}
{"train_loss": 0.2058100700378418, "global_step": 68550, "epoch": 770, "lr": 8.522659093339706e-05}
{"train_loss": 0.23977729678153992, "global_step": 68551, "epoch": 770, "lr": 8.522617950796203e-05}
{"train_loss": 0.23897376656532288, "global_step": 68552, "epoch": 770, "lr": 8.522576807779128e-05}
{"train_loss": 0.19487561285495758, "global_step": 68553, "epoch": 770, "lr": 8.522535664288483e-05}
{"train_loss": 0.1644374430179596, "global_step": 68554, "epoch": 770, "lr": 8.522494520324278e-05}
{"train_loss": 0.28548532724380493, "global_step": 68555, "epoch": 770, "lr": 8.522453375886516e-05}
{"train_loss": 0.28906726837158203, "global_step": 68556, "epoch": 770, "lr": 8.522412230975203e-05}
{"train_loss": 0.25440946221351624, "global_step": 68557, "epoch": 770, "lr": 8.522371085590344e-05}
{"train_loss": 0.22925345599651337, "global_step": 68558, "epoch": 770, "lr": 8.522329939731947e-05}
{"train_loss": 0.20881450176239014, "global_step": 68559, "epoch": 770, "lr": 8.522288793400014e-05}
{"train_loss": 0.2308024764060974, "global_step": 68560, "epoch": 770, "lr": 8.522247646594552e-05}
{"train_loss": 0.26264747977256775, "global_step": 68561, "epoch": 770, "lr": 8.522206499315568e-05}
{"train_loss": 0.2647179961204529, "global_step": 68562, "epoch": 770, "lr": 8.522165351563065e-05}
{"train_loss": 0.2552541494369507, "global_step": 68563, "epoch": 770, "lr": 8.522124203337052e-05}
{"train_loss": 0.18189197778701782, "global_step": 68564, "epoch": 770, "lr": 8.52208305463753e-05}
{"train_loss": 0.18410581350326538, "global_step": 68565, "epoch": 770, "lr": 8.522041905464507e-05}
{"train_loss": 0.16080257296562195, "global_step": 68566, "epoch": 770, "lr": 8.52200075581799e-05}
{"train_loss": 0.23120781779289246, "global_step": 68567, "epoch": 770, "lr": 8.521959605697981e-05}
{"train_loss": 0.30126407742500305, "global_step": 68568, "epoch": 770, "lr": 8.521918455104488e-05}
{"train_loss": 0.2309149205684662, "global_step": 68569, "epoch": 770, "lr": 8.521877304037517e-05}
{"train_loss": 0.19925551116466522, "global_step": 68570, "epoch": 770, "lr": 8.521836152497071e-05}
{"train_loss": 0.26240670680999756, "global_step": 68571, "epoch": 770, "lr": 8.521795000483159e-05}
{"train_loss": 0.24695917963981628, "global_step": 68572, "epoch": 770, "lr": 8.521753847995784e-05}
{"train_loss": 0.30050742626190186, "global_step": 68573, "epoch": 770, "lr": 8.521712695034952e-05}
{"train_loss": 0.2876514494419098, "global_step": 68574, "epoch": 770, "lr": 8.521671541600667e-05}
{"train_loss": 0.19427350163459778, "global_step": 68575, "epoch": 770, "lr": 8.521630387692938e-05}
{"train_loss": 0.18508280813694, "global_step": 68576, "epoch": 770, "lr": 8.521589233311769e-05}
{"train_loss": 0.2699018716812134, "global_step": 68577, "epoch": 770, "lr": 8.521548078457165e-05}
{"train_loss": 0.20299270749092102, "global_step": 68578, "epoch": 770, "lr": 8.52150692312913e-05}
{"train_loss": 0.3370254933834076, "global_step": 68579, "epoch": 770, "lr": 8.521465767327672e-05}
{"train_loss": 0.28170451521873474, "global_step": 68580, "epoch": 770, "lr": 8.521424611052797e-05}
{"train_loss": 0.16592645645141602, "global_step": 68581, "epoch": 770, "lr": 8.521383454304509e-05}
{"train_loss": 0.1928427517414093, "global_step": 68582, "epoch": 770, "lr": 8.521342297082813e-05}
{"train_loss": 0.2674787938594818, "global_step": 68583, "epoch": 770, "lr": 8.521301139387716e-05}
{"train_loss": 0.24875804781913757, "global_step": 68584, "epoch": 770, "lr": 8.521259981219223e-05}
{"train_loss": 0.21214701235294342, "global_step": 68585, "epoch": 770, "lr": 8.52121882257734e-05}
{"train_loss": 0.15124616026878357, "global_step": 68586, "epoch": 770, "lr": 8.521177663462072e-05}
{"train_loss": 0.17794258892536163, "global_step": 68587, "epoch": 770, "lr": 8.521136503873423e-05}
{"train_loss": 0.24883665144443512, "global_step": 68588, "epoch": 770, "lr": 8.521095343811403e-05}
{"train_loss": 0.2095964252948761, "global_step": 68589, "epoch": 770, "lr": 8.521054183276013e-05}
{"train_loss": 0.2438557744026184, "global_step": 68590, "epoch": 770, "lr": 8.52101302226726e-05}
{"train_loss": 0.14959101378917694, "global_step": 68591, "epoch": 770, "lr": 8.520971860785149e-05}
{"train_loss": 0.21993619203567505, "global_step": 68592, "epoch": 770, "lr": 8.520930698829687e-05}
{"train_loss": 0.26273518800735474, "global_step": 68593, "epoch": 770, "lr": 8.520889536400881e-05}
{"train_loss": 0.28909018635749817, "global_step": 68594, "epoch": 770, "lr": 8.520848373498732e-05}
{"train_loss": 0.2257688343524933, "global_step": 68595, "epoch": 770, "lr": 8.520807210123247e-05}
{"train_loss": 0.1847449541091919, "global_step": 68596, "epoch": 770, "lr": 8.520766046274435e-05}
{"train_loss": 0.22303420305252075, "global_step": 68597, "epoch": 770, "lr": 8.520724881952297e-05}
{"train_loss": 0.18667636811733246, "global_step": 68598, "epoch": 770, "lr": 8.520683717156841e-05}
{"train_loss": 0.16949576139450073, "global_step": 68599, "epoch": 770, "lr": 8.520642551888072e-05}
{"train_loss": 0.21889594197273254, "global_step": 68600, "epoch": 770, "lr": 8.520601386145994e-05}
{"train_loss": 0.2687017321586609, "global_step": 68601, "epoch": 770, "lr": 8.520560219930615e-05}
{"train_loss": 0.13559304177761078, "global_step": 68602, "epoch": 770, "lr": 8.520519053241941e-05}
{"train_loss": 0.18324726819992065, "global_step": 68603, "epoch": 770, "lr": 8.520477886079975e-05}
{"train_loss": 0.19252760708332062, "global_step": 68604, "epoch": 770, "lr": 8.520436718444724e-05}
{"train_loss": 0.20943321287631989, "global_step": 68605, "epoch": 770, "lr": 8.520395550336193e-05}
{"train_loss": 0.1764640212059021, "global_step": 68606, "epoch": 770, "lr": 8.520354381754387e-05}
{"train_loss": 0.2338918000459671, "global_step": 68607, "epoch": 770, "lr": 8.520313212699313e-05}
{"train_loss": 0.18338613212108612, "global_step": 68608, "epoch": 770, "lr": 8.520272043170976e-05}
{"train_loss": 0.24690911173820496, "global_step": 68609, "epoch": 770, "lr": 8.520230873169382e-05}
{"train_loss": 0.22377391159534454, "global_step": 68610, "epoch": 770, "lr": 8.520189702694534e-05}
{"train_loss": 0.23638470470905304, "global_step": 68611, "epoch": 770, "lr": 8.520148531746441e-05}
{"train_loss": 0.21087615191936493, "global_step": 68612, "epoch": 770, "lr": 8.520107360325107e-05}
{"train_loss": 0.17866557836532593, "global_step": 68613, "epoch": 770, "lr": 8.520066188430535e-05}
{"train_loss": 0.21068331599235535, "global_step": 68614, "epoch": 770, "lr": 8.520025016062736e-05}
{"train_loss": 0.19480200111865997, "global_step": 68615, "epoch": 770, "lr": 8.519983843221712e-05}
{"train_loss": 0.24424633383750916, "global_step": 68616, "epoch": 770, "lr": 8.519942669907468e-05}
{"train_loss": 0.17859862744808197, "global_step": 68617, "epoch": 770, "lr": 8.519901496120011e-05}
{"train_loss": 0.2272880497608292, "global_step": 68618, "epoch": 770, "lr": 8.519860321859345e-05, "val_loss": 3.750525951385498, "train_action_mse_error": 14.383045196533203}
{"train_loss": 0.2184462547302246, "global_step": 68619, "epoch": 771, "lr": 8.519819147125478e-05}
{"train_loss": 0.20706254243850708, "global_step": 68620, "epoch": 771, "lr": 8.519777971918415e-05}
{"train_loss": 0.30676957964897156, "global_step": 68621, "epoch": 771, "lr": 8.51973679623816e-05}
{"train_loss": 0.16416999697685242, "global_step": 68622, "epoch": 771, "lr": 8.519695620084718e-05}
{"train_loss": 0.22917208075523376, "global_step": 68623, "epoch": 771, "lr": 8.519654443458098e-05}
{"train_loss": 0.16989320516586304, "global_step": 68624, "epoch": 771, "lr": 8.519613266358302e-05}
{"train_loss": 0.13598930835723877, "global_step": 68625, "epoch": 771, "lr": 8.519572088785337e-05}
{"train_loss": 0.2564011514186859, "global_step": 68626, "epoch": 771, "lr": 8.519530910739208e-05}
{"train_loss": 0.1452319622039795, "global_step": 68627, "epoch": 771, "lr": 8.519489732219924e-05}
{"train_loss": 0.2056250274181366, "global_step": 68628, "epoch": 771, "lr": 8.519448553227482e-05}
{"train_loss": 0.16045880317687988, "global_step": 68629, "epoch": 771, "lr": 8.519407373761897e-05}
{"train_loss": 0.16760899126529694, "global_step": 68630, "epoch": 771, "lr": 8.519366193823171e-05}
{"train_loss": 0.17175108194351196, "global_step": 68631, "epoch": 771, "lr": 8.519325013411307e-05}
{"train_loss": 0.15408679842948914, "global_step": 68632, "epoch": 771, "lr": 8.519283832526314e-05}
{"train_loss": 0.19805529713630676, "global_step": 68633, "epoch": 771, "lr": 8.519242651168195e-05}
{"train_loss": 0.14025118947029114, "global_step": 68634, "epoch": 771, "lr": 8.519201469336957e-05}
{"train_loss": 0.18600493669509888, "global_step": 68635, "epoch": 771, "lr": 8.519160287032605e-05}
{"train_loss": 0.17942096292972565, "global_step": 68636, "epoch": 771, "lr": 8.519119104255147e-05}
{"train_loss": 0.177202045917511, "global_step": 68637, "epoch": 771, "lr": 8.519077921004584e-05}
{"train_loss": 0.1849452406167984, "global_step": 68638, "epoch": 771, "lr": 8.519036737280925e-05}
{"train_loss": 0.21612034738063812, "global_step": 68639, "epoch": 771, "lr": 8.518995553084172e-05}
{"train_loss": 0.17144303023815155, "global_step": 68640, "epoch": 771, "lr": 8.518954368414336e-05}
{"train_loss": 0.19883517920970917, "global_step": 68641, "epoch": 771, "lr": 8.518913183271419e-05}
{"train_loss": 0.157505601644516, "global_step": 68642, "epoch": 771, "lr": 8.518871997655427e-05}
{"train_loss": 0.2223745882511139, "global_step": 68643, "epoch": 771, "lr": 8.518830811566365e-05}
{"train_loss": 0.15326111018657684, "global_step": 68644, "epoch": 771, "lr": 8.518789625004237e-05}
{"train_loss": 0.17435888946056366, "global_step": 68645, "epoch": 771, "lr": 8.518748437969054e-05}
{"train_loss": 0.2466680258512497, "global_step": 68646, "epoch": 771, "lr": 8.518707250460816e-05}
{"train_loss": 0.2602601945400238, "global_step": 68647, "epoch": 771, "lr": 8.518666062479533e-05}
{"train_loss": 0.15420451760292053, "global_step": 68648, "epoch": 771, "lr": 8.518624874025206e-05}
{"train_loss": 0.19701772928237915, "global_step": 68649, "epoch": 771, "lr": 8.518583685097843e-05}
{"train_loss": 0.17374387383460999, "global_step": 68650, "epoch": 771, "lr": 8.518542495697451e-05}
{"train_loss": 0.1409129798412323, "global_step": 68651, "epoch": 771, "lr": 8.518501305824031e-05}
{"train_loss": 0.26005828380584717, "global_step": 68652, "epoch": 771, "lr": 8.518460115477595e-05}
{"train_loss": 0.1847478598356247, "global_step": 68653, "epoch": 771, "lr": 8.518418924658142e-05}
{"train_loss": 0.13947583734989166, "global_step": 68654, "epoch": 771, "lr": 8.518377733365681e-05}
{"train_loss": 0.20568662881851196, "global_step": 68655, "epoch": 771, "lr": 8.518336541600219e-05}
{"train_loss": 0.3516652584075928, "global_step": 68656, "epoch": 771, "lr": 8.518295349361757e-05}
{"train_loss": 0.24745702743530273, "global_step": 68657, "epoch": 771, "lr": 8.518254156650304e-05}
{"train_loss": 0.24533025920391083, "global_step": 68658, "epoch": 771, "lr": 8.518212963465865e-05}
{"train_loss": 0.2675746977329254, "global_step": 68659, "epoch": 771, "lr": 8.518171769808444e-05}
{"train_loss": 0.3482261300086975, "global_step": 68660, "epoch": 771, "lr": 8.51813057567805e-05}
{"train_loss": 0.14872682094573975, "global_step": 68661, "epoch": 771, "lr": 8.518089381074683e-05}
{"train_loss": 0.16016776859760284, "global_step": 68662, "epoch": 771, "lr": 8.518048185998355e-05}
{"train_loss": 0.2543354332447052, "global_step": 68663, "epoch": 771, "lr": 8.518006990449065e-05}
{"train_loss": 0.27823513746261597, "global_step": 68664, "epoch": 771, "lr": 8.517965794426826e-05}
{"train_loss": 0.18538625538349152, "global_step": 68665, "epoch": 771, "lr": 8.517924597931636e-05}
{"train_loss": 0.10249096900224686, "global_step": 68666, "epoch": 771, "lr": 8.517883400963504e-05}
{"train_loss": 0.20903854072093964, "global_step": 68667, "epoch": 771, "lr": 8.517842203522437e-05}
{"train_loss": 0.19011694192886353, "global_step": 68668, "epoch": 771, "lr": 8.517801005608439e-05}
{"train_loss": 0.16097892820835114, "global_step": 68669, "epoch": 771, "lr": 8.517759807221515e-05}
{"train_loss": 0.2027300000190735, "global_step": 68670, "epoch": 771, "lr": 8.51771860836167e-05}
{"train_loss": 0.2205234169960022, "global_step": 68671, "epoch": 771, "lr": 8.517677409028913e-05}
{"train_loss": 0.2845273017883301, "global_step": 68672, "epoch": 771, "lr": 8.517636209223244e-05}
{"train_loss": 0.14501163363456726, "global_step": 68673, "epoch": 771, "lr": 8.517595008944674e-05}
{"train_loss": 0.1765650361776352, "global_step": 68674, "epoch": 771, "lr": 8.517553808193206e-05}
{"train_loss": 0.2216319888830185, "global_step": 68675, "epoch": 771, "lr": 8.517512606968844e-05}
{"train_loss": 0.228524848818779, "global_step": 68676, "epoch": 771, "lr": 8.517471405271597e-05}
{"train_loss": 0.2272760421037674, "global_step": 68677, "epoch": 771, "lr": 8.517430203101468e-05}
{"train_loss": 0.14655275642871857, "global_step": 68678, "epoch": 771, "lr": 8.517389000458464e-05}
{"train_loss": 0.19823598861694336, "global_step": 68679, "epoch": 771, "lr": 8.517347797342591e-05}
{"train_loss": 0.1780012547969818, "global_step": 68680, "epoch": 771, "lr": 8.517306593753852e-05}
{"train_loss": 0.2962035834789276, "global_step": 68681, "epoch": 771, "lr": 8.517265389692255e-05}
{"train_loss": 0.17531348764896393, "global_step": 68682, "epoch": 771, "lr": 8.517224185157803e-05}
{"train_loss": 0.23949012160301208, "global_step": 68683, "epoch": 771, "lr": 8.517182980150504e-05}
{"train_loss": 0.23074650764465332, "global_step": 68684, "epoch": 771, "lr": 8.517141774670363e-05}
{"train_loss": 0.19403180480003357, "global_step": 68685, "epoch": 771, "lr": 8.517100568717384e-05}
{"train_loss": 0.16723953187465668, "global_step": 68686, "epoch": 771, "lr": 8.517059362291576e-05}
{"train_loss": 0.17391039431095123, "global_step": 68687, "epoch": 771, "lr": 8.51701815539294e-05}
{"train_loss": 0.14404746890068054, "global_step": 68688, "epoch": 771, "lr": 8.516976948021484e-05}
{"train_loss": 0.24376772344112396, "global_step": 68689, "epoch": 771, "lr": 8.516935740177214e-05}
{"train_loss": 0.18496423959732056, "global_step": 68690, "epoch": 771, "lr": 8.516894531860133e-05}
{"train_loss": 0.24501514434814453, "global_step": 68691, "epoch": 771, "lr": 8.516853323070251e-05}
{"train_loss": 0.21312808990478516, "global_step": 68692, "epoch": 771, "lr": 8.51681211380757e-05}
{"train_loss": 0.19560211896896362, "global_step": 68693, "epoch": 771, "lr": 8.516770904072096e-05}
{"train_loss": 0.16793997585773468, "global_step": 68694, "epoch": 771, "lr": 8.516729693863836e-05}
{"train_loss": 0.21700236201286316, "global_step": 68695, "epoch": 771, "lr": 8.516688483182793e-05}
{"train_loss": 0.22743922472000122, "global_step": 68696, "epoch": 771, "lr": 8.516647272028974e-05}
{"train_loss": 0.1698368638753891, "global_step": 68697, "epoch": 771, "lr": 8.516606060402387e-05}
{"train_loss": 0.4140893518924713, "global_step": 68698, "epoch": 771, "lr": 8.516564848303034e-05}
{"train_loss": 0.1929905116558075, "global_step": 68699, "epoch": 771, "lr": 8.516523635730921e-05}
{"train_loss": 0.21879445016384125, "global_step": 68700, "epoch": 771, "lr": 8.516482422686056e-05}
{"train_loss": 0.16120728850364685, "global_step": 68701, "epoch": 771, "lr": 8.516441209168443e-05}
{"train_loss": 0.16213880479335785, "global_step": 68702, "epoch": 771, "lr": 8.516399995178084e-05}
{"train_loss": 0.21148519217967987, "global_step": 68703, "epoch": 771, "lr": 8.51635878071499e-05}
{"train_loss": 0.21005725860595703, "global_step": 68704, "epoch": 771, "lr": 8.516317565779164e-05}
{"train_loss": 0.28831321001052856, "global_step": 68705, "epoch": 771, "lr": 8.516276350370612e-05}
{"train_loss": 0.1493707299232483, "global_step": 68706, "epoch": 771, "lr": 8.516235134489341e-05}
{"train_loss": 0.20373723789882126, "global_step": 68707, "epoch": 771, "lr": 8.516193918135353e-05, "val_loss": 3.7120492458343506}
{"train_loss": 0.24241653084754944, "global_step": 68708, "epoch": 772, "lr": 8.516152701308657e-05}
{"train_loss": 0.1853952407836914, "global_step": 68709, "epoch": 772, "lr": 8.516111484009255e-05}
{"train_loss": 0.2668222486972809, "global_step": 68710, "epoch": 772, "lr": 8.516070266237157e-05}
{"train_loss": 0.21259267628192902, "global_step": 68711, "epoch": 772, "lr": 8.516029047992365e-05}
{"train_loss": 0.17499427497386932, "global_step": 68712, "epoch": 772, "lr": 8.515987829274885e-05}
{"train_loss": 0.18987992405891418, "global_step": 68713, "epoch": 772, "lr": 8.515946610084725e-05}
{"train_loss": 0.20402206480503082, "global_step": 68714, "epoch": 772, "lr": 8.515905390421889e-05}
{"train_loss": 0.1728784441947937, "global_step": 68715, "epoch": 772, "lr": 8.51586417028638e-05}
{"train_loss": 0.276145875453949, "global_step": 68716, "epoch": 772, "lr": 8.515822949678209e-05}
{"train_loss": 0.18546247482299805, "global_step": 68717, "epoch": 772, "lr": 8.515781728597376e-05}
{"train_loss": 0.2647419273853302, "global_step": 68718, "epoch": 772, "lr": 8.51574050704389e-05}
{"train_loss": 0.318788081407547, "global_step": 68719, "epoch": 772, "lr": 8.515699285017757e-05}
{"train_loss": 0.29470527172088623, "global_step": 68720, "epoch": 772, "lr": 8.515658062518978e-05}
{"train_loss": 0.2031830996274948, "global_step": 68721, "epoch": 772, "lr": 8.515616839547565e-05}
{"train_loss": 0.22069799900054932, "global_step": 68722, "epoch": 772, "lr": 8.515575616103517e-05}
{"train_loss": 0.20535148680210114, "global_step": 68723, "epoch": 772, "lr": 8.515534392186845e-05}
{"train_loss": 0.2775699496269226, "global_step": 68724, "epoch": 772, "lr": 8.51549316779755e-05}
{"train_loss": 0.2593548595905304, "global_step": 68725, "epoch": 772, "lr": 8.515451942935641e-05}
{"train_loss": 0.2611401677131653, "global_step": 68726, "epoch": 772, "lr": 8.515410717601125e-05}
{"train_loss": 0.28690487146377563, "global_step": 68727, "epoch": 772, "lr": 8.515369491794001e-05}
{"train_loss": 0.23332609236240387, "global_step": 68728, "epoch": 772, "lr": 8.51532826551428e-05}
{"train_loss": 0.27118119597435, "global_step": 68729, "epoch": 772, "lr": 8.515287038761966e-05}
{"train_loss": 0.20437031984329224, "global_step": 68730, "epoch": 772, "lr": 8.515245811537063e-05}
{"train_loss": 0.1846894919872284, "global_step": 68731, "epoch": 772, "lr": 8.51520458383958e-05}
{"train_loss": 0.20259881019592285, "global_step": 68732, "epoch": 772, "lr": 8.51516335566952e-05}
{"train_loss": 0.2784259617328644, "global_step": 68733, "epoch": 772, "lr": 8.515122127026888e-05}
{"train_loss": 0.32171642780303955, "global_step": 68734, "epoch": 772, "lr": 8.515080897911691e-05}
{"train_loss": 0.2262691706418991, "global_step": 68735, "epoch": 772, "lr": 8.515039668323937e-05}
{"train_loss": 0.2418820858001709, "global_step": 68736, "epoch": 772, "lr": 8.514998438263626e-05}
{"train_loss": 0.21906867623329163, "global_step": 68737, "epoch": 772, "lr": 8.514957207730767e-05}
{"train_loss": 0.28244855999946594, "global_step": 68738, "epoch": 772, "lr": 8.514915976725363e-05}
{"train_loss": 0.21665795147418976, "global_step": 68739, "epoch": 772, "lr": 8.514874745247426e-05}
{"train_loss": 0.26506078243255615, "global_step": 68740, "epoch": 772, "lr": 8.514833513296953e-05}
{"train_loss": 0.150636687874794, "global_step": 68741, "epoch": 772, "lr": 8.514792280873954e-05}
{"train_loss": 0.20211616158485413, "global_step": 68742, "epoch": 772, "lr": 8.514751047978436e-05}
{"train_loss": 0.28877580165863037, "global_step": 68743, "epoch": 772, "lr": 8.5147098146104e-05}
{"train_loss": 0.23245422542095184, "global_step": 68744, "epoch": 772, "lr": 8.514668580769855e-05}
{"train_loss": 0.14492309093475342, "global_step": 68745, "epoch": 772, "lr": 8.514627346456805e-05}
{"train_loss": 0.19742654263973236, "global_step": 68746, "epoch": 772, "lr": 8.514586111671258e-05}
{"train_loss": 0.2529980540275574, "global_step": 68747, "epoch": 772, "lr": 8.514544876413218e-05}
{"train_loss": 0.2181328982114792, "global_step": 68748, "epoch": 772, "lr": 8.514503640682687e-05}
{"train_loss": 0.2707068622112274, "global_step": 68749, "epoch": 772, "lr": 8.514462404479678e-05}
{"train_loss": 0.21194325387477875, "global_step": 68750, "epoch": 772, "lr": 8.51442116780419e-05}
{"train_loss": 0.21387313306331635, "global_step": 68751, "epoch": 772, "lr": 8.51437993065623e-05}
{"train_loss": 0.16383671760559082, "global_step": 68752, "epoch": 772, "lr": 8.514338693035806e-05}
{"train_loss": 0.285433828830719, "global_step": 68753, "epoch": 772, "lr": 8.514297454942921e-05}
{"train_loss": 0.11753495782613754, "global_step": 68754, "epoch": 772, "lr": 8.514256216377584e-05}
{"train_loss": 0.18525397777557373, "global_step": 68755, "epoch": 772, "lr": 8.514214977339795e-05}
{"train_loss": 0.18556542694568634, "global_step": 68756, "epoch": 772, "lr": 8.514173737829563e-05}
{"train_loss": 0.2327805459499359, "global_step": 68757, "epoch": 772, "lr": 8.514132497846894e-05}
{"train_loss": 0.2357354462146759, "global_step": 68758, "epoch": 772, "lr": 8.514091257391794e-05}
{"train_loss": 0.22680895030498505, "global_step": 68759, "epoch": 772, "lr": 8.514050016464266e-05}
{"train_loss": 0.18933597207069397, "global_step": 68760, "epoch": 772, "lr": 8.514008775064317e-05}
{"train_loss": 0.1822909116744995, "global_step": 68761, "epoch": 772, "lr": 8.513967533191952e-05}
{"train_loss": 0.15478193759918213, "global_step": 68762, "epoch": 772, "lr": 8.513926290847177e-05}
{"train_loss": 0.2132687270641327, "global_step": 68763, "epoch": 772, "lr": 8.513885048029997e-05}
{"train_loss": 0.3071556091308594, "global_step": 68764, "epoch": 772, "lr": 8.513843804740417e-05}
{"train_loss": 0.2529160678386688, "global_step": 68765, "epoch": 772, "lr": 8.513802560978447e-05}
{"train_loss": 0.18649569153785706, "global_step": 68766, "epoch": 772, "lr": 8.513761316744086e-05}
{"train_loss": 0.22354619204998016, "global_step": 68767, "epoch": 772, "lr": 8.513720072037345e-05}
{"train_loss": 0.24432337284088135, "global_step": 68768, "epoch": 772, "lr": 8.513678826858226e-05}
{"train_loss": 0.23421336710453033, "global_step": 68769, "epoch": 772, "lr": 8.513637581206734e-05}
{"train_loss": 0.2109612673521042, "global_step": 68770, "epoch": 772, "lr": 8.513596335082879e-05}
{"train_loss": 0.24824519455432892, "global_step": 68771, "epoch": 772, "lr": 8.513555088486662e-05}
{"train_loss": 0.26035645604133606, "global_step": 68772, "epoch": 772, "lr": 8.513513841418092e-05}
{"train_loss": 0.20779305696487427, "global_step": 68773, "epoch": 772, "lr": 8.513472593877172e-05}
{"train_loss": 0.24059435725212097, "global_step": 68774, "epoch": 772, "lr": 8.513431345863907e-05}
{"train_loss": 0.15719635784626007, "global_step": 68775, "epoch": 772, "lr": 8.513390097378306e-05}
{"train_loss": 0.18030409514904022, "global_step": 68776, "epoch": 772, "lr": 8.513348848420372e-05}
{"train_loss": 0.20322130620479584, "global_step": 68777, "epoch": 772, "lr": 8.513307598990111e-05}
{"train_loss": 0.324168860912323, "global_step": 68778, "epoch": 772, "lr": 8.513266349087529e-05}
{"train_loss": 0.20715360343456268, "global_step": 68779, "epoch": 772, "lr": 8.51322509871263e-05}
{"train_loss": 0.21628035604953766, "global_step": 68780, "epoch": 772, "lr": 8.513183847865422e-05}
{"train_loss": 0.15027762949466705, "global_step": 68781, "epoch": 772, "lr": 8.513142596545908e-05}
{"train_loss": 0.26620280742645264, "global_step": 68782, "epoch": 772, "lr": 8.513101344754095e-05}
{"train_loss": 0.22721225023269653, "global_step": 68783, "epoch": 772, "lr": 8.513060092489988e-05}
{"train_loss": 0.26116225123405457, "global_step": 68784, "epoch": 772, "lr": 8.513018839753595e-05}
{"train_loss": 0.25460144877433777, "global_step": 68785, "epoch": 772, "lr": 8.512977586544918e-05}
{"train_loss": 0.20388992130756378, "global_step": 68786, "epoch": 772, "lr": 8.512936332863964e-05}
{"train_loss": 0.16450664401054382, "global_step": 68787, "epoch": 772, "lr": 8.512895078710739e-05}
{"train_loss": 0.12556986510753632, "global_step": 68788, "epoch": 772, "lr": 8.512853824085247e-05}
{"train_loss": 0.20905299484729767, "global_step": 68789, "epoch": 772, "lr": 8.512812568987496e-05}
{"train_loss": 0.2370055615901947, "global_step": 68790, "epoch": 772, "lr": 8.51277131341749e-05}
{"train_loss": 0.22639904916286469, "global_step": 68791, "epoch": 772, "lr": 8.512730057375233e-05}
{"train_loss": 0.26795753836631775, "global_step": 68792, "epoch": 772, "lr": 8.512688800860733e-05}
{"train_loss": 0.23942025005817413, "global_step": 68793, "epoch": 772, "lr": 8.512647543873994e-05}
{"train_loss": 0.211472749710083, "global_step": 68794, "epoch": 772, "lr": 8.512606286415026e-05}
{"train_loss": 0.19696210324764252, "global_step": 68795, "epoch": 772, "lr": 8.512565028483827e-05}
{"train_loss": 0.22397388609942426, "global_step": 68796, "epoch": 772, "lr": 8.512523770080409e-05, "val_loss": 3.807591199874878}
{"train_loss": 0.20427505671977997, "global_step": 68797, "epoch": 773, "lr": 8.512482511204773e-05}
{"train_loss": 0.19625923037528992, "global_step": 68798, "epoch": 773, "lr": 8.512441251856929e-05}
{"train_loss": 0.2583808898925781, "global_step": 68799, "epoch": 773, "lr": 8.512399992036877e-05}
{"train_loss": 0.1573648303747177, "global_step": 68800, "epoch": 773, "lr": 8.512358731744627e-05}
{"train_loss": 0.23738257586956024, "global_step": 68801, "epoch": 773, "lr": 8.512317470980183e-05}
{"train_loss": 0.221580371260643, "global_step": 68802, "epoch": 773, "lr": 8.512276209743552e-05}
{"train_loss": 0.23664121329784393, "global_step": 68803, "epoch": 773, "lr": 8.512234948034737e-05}
{"train_loss": 0.16921408474445343, "global_step": 68804, "epoch": 773, "lr": 8.512193685853746e-05}
{"train_loss": 0.29434382915496826, "global_step": 68805, "epoch": 773, "lr": 8.512152423200581e-05}
{"train_loss": 0.2015119045972824, "global_step": 68806, "epoch": 773, "lr": 8.512111160075252e-05}
{"train_loss": 0.18347273766994476, "global_step": 68807, "epoch": 773, "lr": 8.512069896477763e-05}
{"train_loss": 0.2975400686264038, "global_step": 68808, "epoch": 773, "lr": 8.512028632408118e-05}
{"train_loss": 0.1933821737766266, "global_step": 68809, "epoch": 773, "lr": 8.511987367866323e-05}
{"train_loss": 0.26080214977264404, "global_step": 68810, "epoch": 773, "lr": 8.511946102852384e-05}
{"train_loss": 0.22313444316387177, "global_step": 68811, "epoch": 773, "lr": 8.511904837366306e-05}
{"train_loss": 0.23558779060840607, "global_step": 68812, "epoch": 773, "lr": 8.511863571408098e-05}
{"train_loss": 0.17267945408821106, "global_step": 68813, "epoch": 773, "lr": 8.511822304977762e-05}
{"train_loss": 0.17403769493103027, "global_step": 68814, "epoch": 773, "lr": 8.511781038075301e-05}
{"train_loss": 0.17193879187107086, "global_step": 68815, "epoch": 773, "lr": 8.511739770700728e-05}
{"train_loss": 0.14694148302078247, "global_step": 68816, "epoch": 773, "lr": 8.511698502854043e-05}
{"train_loss": 0.2598051428794861, "global_step": 68817, "epoch": 773, "lr": 8.511657234535252e-05}
{"train_loss": 0.2526518702507019, "global_step": 68818, "epoch": 773, "lr": 8.511615965744361e-05}
{"train_loss": 0.17733150720596313, "global_step": 68819, "epoch": 773, "lr": 8.511574696481377e-05}
{"train_loss": 0.21086475253105164, "global_step": 68820, "epoch": 773, "lr": 8.511533426746305e-05}
{"train_loss": 0.22186297178268433, "global_step": 68821, "epoch": 773, "lr": 8.51149215653915e-05}
{"train_loss": 0.2830164134502411, "global_step": 68822, "epoch": 773, "lr": 8.511450885859918e-05}
{"train_loss": 0.1802157312631607, "global_step": 68823, "epoch": 773, "lr": 8.511409614708614e-05}
{"train_loss": 0.17535164952278137, "global_step": 68824, "epoch": 773, "lr": 8.511368343085245e-05}
{"train_loss": 0.18585091829299927, "global_step": 68825, "epoch": 773, "lr": 8.511327070989813e-05}
{"train_loss": 0.15179508924484253, "global_step": 68826, "epoch": 773, "lr": 8.511285798422325e-05}
{"train_loss": 0.15628114342689514, "global_step": 68827, "epoch": 773, "lr": 8.511244525382792e-05}
{"train_loss": 0.19783973693847656, "global_step": 68828, "epoch": 773, "lr": 8.511203251871211e-05}
{"train_loss": 0.16143764555454254, "global_step": 68829, "epoch": 773, "lr": 8.511161977887592e-05}
{"train_loss": 0.15425463020801544, "global_step": 68830, "epoch": 773, "lr": 8.511120703431941e-05}
{"train_loss": 0.1948055475950241, "global_step": 68831, "epoch": 773, "lr": 8.511079428504262e-05}
{"train_loss": 0.2920701503753662, "global_step": 68832, "epoch": 773, "lr": 8.511038153104562e-05}
{"train_loss": 0.2700527310371399, "global_step": 68833, "epoch": 773, "lr": 8.510996877232845e-05}
{"train_loss": 0.26836276054382324, "global_step": 68834, "epoch": 773, "lr": 8.510955600889116e-05}
{"train_loss": 0.22446998953819275, "global_step": 68835, "epoch": 773, "lr": 8.510914324073385e-05}
{"train_loss": 0.22409123182296753, "global_step": 68836, "epoch": 773, "lr": 8.510873046785653e-05}
{"train_loss": 0.18732434511184692, "global_step": 68837, "epoch": 773, "lr": 8.510831769025926e-05}
{"train_loss": 0.20978060364723206, "global_step": 68838, "epoch": 773, "lr": 8.51079049079421e-05}
{"train_loss": 0.22021524608135223, "global_step": 68839, "epoch": 773, "lr": 8.510749212090511e-05}
{"train_loss": 0.16382603347301483, "global_step": 68840, "epoch": 773, "lr": 8.510707932914836e-05}
{"train_loss": 0.20998911559581757, "global_step": 68841, "epoch": 773, "lr": 8.510666653267189e-05}
{"train_loss": 0.23292581737041473, "global_step": 68842, "epoch": 773, "lr": 8.510625373147577e-05}
{"train_loss": 0.17524750530719757, "global_step": 68843, "epoch": 773, "lr": 8.510584092556003e-05}
{"train_loss": 0.24547706544399261, "global_step": 68844, "epoch": 773, "lr": 8.510542811492473e-05}
{"train_loss": 0.2708388864994049, "global_step": 68845, "epoch": 773, "lr": 8.510501529956994e-05}
{"train_loss": 0.1611216962337494, "global_step": 68846, "epoch": 773, "lr": 8.51046024794957e-05}
{"train_loss": 0.2549702227115631, "global_step": 68847, "epoch": 773, "lr": 8.510418965470208e-05}
{"train_loss": 0.11346805840730667, "global_step": 68848, "epoch": 773, "lr": 8.510377682518912e-05}
{"train_loss": 0.2771751582622528, "global_step": 68849, "epoch": 773, "lr": 8.510336399095688e-05}
{"train_loss": 0.19061800837516785, "global_step": 68850, "epoch": 773, "lr": 8.510295115200544e-05}
{"train_loss": 0.24821346998214722, "global_step": 68851, "epoch": 773, "lr": 8.510253830833484e-05}
{"train_loss": 0.19111746549606323, "global_step": 68852, "epoch": 773, "lr": 8.510212545994511e-05}
{"train_loss": 0.19698330760002136, "global_step": 68853, "epoch": 773, "lr": 8.510171260683634e-05}
{"train_loss": 0.17697271704673767, "global_step": 68854, "epoch": 773, "lr": 8.510129974900858e-05}
{"train_loss": 0.22256772220134735, "global_step": 68855, "epoch": 773, "lr": 8.510088688646186e-05}
{"train_loss": 0.1341247707605362, "global_step": 68856, "epoch": 773, "lr": 8.510047401919627e-05}
{"train_loss": 0.18792983889579773, "global_step": 68857, "epoch": 773, "lr": 8.510006114721184e-05}
{"train_loss": 0.244966521859169, "global_step": 68858, "epoch": 773, "lr": 8.509964827050864e-05}
{"train_loss": 0.20782694220542908, "global_step": 68859, "epoch": 773, "lr": 8.50992353890867e-05}
{"train_loss": 0.24926163256168365, "global_step": 68860, "epoch": 773, "lr": 8.509882250294611e-05}
{"train_loss": 0.21548983454704285, "global_step": 68861, "epoch": 773, "lr": 8.509840961208693e-05}
{"train_loss": 0.19242386519908905, "global_step": 68862, "epoch": 773, "lr": 8.509799671650917e-05}
{"train_loss": 0.270987331867218, "global_step": 68863, "epoch": 773, "lr": 8.509758381621293e-05}
{"train_loss": 0.21781645715236664, "global_step": 68864, "epoch": 773, "lr": 8.509717091119824e-05}
{"train_loss": 0.28909632563591003, "global_step": 68865, "epoch": 773, "lr": 8.509675800146517e-05}
{"train_loss": 0.26294371485710144, "global_step": 68866, "epoch": 773, "lr": 8.509634508701376e-05}
{"train_loss": 0.2298824042081833, "global_step": 68867, "epoch": 773, "lr": 8.509593216784408e-05}
{"train_loss": 0.18903735280036926, "global_step": 68868, "epoch": 773, "lr": 8.509551924395617e-05}
{"train_loss": 0.22320812940597534, "global_step": 68869, "epoch": 773, "lr": 8.509510631535009e-05}
{"train_loss": 0.2263999581336975, "global_step": 68870, "epoch": 773, "lr": 8.509469338202592e-05}
{"train_loss": 0.19609035551548004, "global_step": 68871, "epoch": 773, "lr": 8.509428044398369e-05}
{"train_loss": 0.22196350991725922, "global_step": 68872, "epoch": 773, "lr": 8.509386750122345e-05}
{"train_loss": 0.25934308767318726, "global_step": 68873, "epoch": 773, "lr": 8.509345455374528e-05}
{"train_loss": 0.24110794067382812, "global_step": 68874, "epoch": 773, "lr": 8.509304160154921e-05}
{"train_loss": 0.24896538257598877, "global_step": 68875, "epoch": 773, "lr": 8.509262864463532e-05}
{"train_loss": 0.1304120570421219, "global_step": 68876, "epoch": 773, "lr": 8.509221568300365e-05}
{"train_loss": 0.2617221474647522, "global_step": 68877, "epoch": 773, "lr": 8.509180271665427e-05}
{"train_loss": 0.3049488961696625, "global_step": 68878, "epoch": 773, "lr": 8.509138974558722e-05}
{"train_loss": 0.24988067150115967, "global_step": 68879, "epoch": 773, "lr": 8.509097676980256e-05}
{"train_loss": 0.24543310701847076, "global_step": 68880, "epoch": 773, "lr": 8.509056378930033e-05}
{"train_loss": 0.17390699684619904, "global_step": 68881, "epoch": 773, "lr": 8.509015080408061e-05}
{"train_loss": 0.2834194004535675, "global_step": 68882, "epoch": 773, "lr": 8.508973781414345e-05}
{"train_loss": 0.278095006942749, "global_step": 68883, "epoch": 773, "lr": 8.50893248194889e-05}
{"train_loss": 0.20449504256248474, "global_step": 68884, "epoch": 773, "lr": 8.508891182011702e-05}
{"train_loss": 0.2171327072415459, "global_step": 68885, "epoch": 773, "lr": 8.508849881602787e-05, "val_loss": 3.811880350112915}
{"train_loss": 0.19119437038898468, "global_step": 68886, "epoch": 774, "lr": 8.50880858072215e-05}
{"train_loss": 0.1596938520669937, "global_step": 68887, "epoch": 774, "lr": 8.508767279369796e-05}
{"train_loss": 0.19798657298088074, "global_step": 68888, "epoch": 774, "lr": 8.508725977545731e-05}
{"train_loss": 0.10192025452852249, "global_step": 68889, "epoch": 774, "lr": 8.50868467524996e-05}
{"train_loss": 0.20811665058135986, "global_step": 68890, "epoch": 774, "lr": 8.508643372482487e-05}
{"train_loss": 0.18318405747413635, "global_step": 68891, "epoch": 774, "lr": 8.508602069243323e-05}
{"train_loss": 0.23823577165603638, "global_step": 68892, "epoch": 774, "lr": 8.50856076553247e-05}
{"train_loss": 0.2846314013004303, "global_step": 68893, "epoch": 774, "lr": 8.508519461349934e-05}
{"train_loss": 0.23642654716968536, "global_step": 68894, "epoch": 774, "lr": 8.508478156695718e-05}
{"train_loss": 0.2636799216270447, "global_step": 68895, "epoch": 774, "lr": 8.508436851569832e-05}
{"train_loss": 0.1855219006538391, "global_step": 68896, "epoch": 774, "lr": 8.508395545972279e-05}
{"train_loss": 0.1903429925441742, "global_step": 68897, "epoch": 774, "lr": 8.508354239903064e-05}
{"train_loss": 0.15560123324394226, "global_step": 68898, "epoch": 774, "lr": 8.508312933362195e-05}
{"train_loss": 0.2835715413093567, "global_step": 68899, "epoch": 774, "lr": 8.508271626349674e-05}
{"train_loss": 0.20382855832576752, "global_step": 68900, "epoch": 774, "lr": 8.508230318865511e-05}
{"train_loss": 0.17325283586978912, "global_step": 68901, "epoch": 774, "lr": 8.508189010909709e-05}
{"train_loss": 0.2933472990989685, "global_step": 68902, "epoch": 774, "lr": 8.508147702482273e-05}
{"train_loss": 0.16526412963867188, "global_step": 68903, "epoch": 774, "lr": 8.508106393583209e-05}
{"train_loss": 0.2271062582731247, "global_step": 68904, "epoch": 774, "lr": 8.508065084212522e-05}
{"train_loss": 0.22182518243789673, "global_step": 68905, "epoch": 774, "lr": 8.50802377437022e-05}
{"train_loss": 0.24565637111663818, "global_step": 68906, "epoch": 774, "lr": 8.507982464056306e-05}
{"train_loss": 0.34819576144218445, "global_step": 68907, "epoch": 774, "lr": 8.507941153270787e-05}
{"train_loss": 0.21069806814193726, "global_step": 68908, "epoch": 774, "lr": 8.507899842013667e-05}
{"train_loss": 0.2547966539859772, "global_step": 68909, "epoch": 774, "lr": 8.507858530284954e-05}
{"train_loss": 0.32333534955978394, "global_step": 68910, "epoch": 774, "lr": 8.50781721808465e-05}
{"train_loss": 0.149454265832901, "global_step": 68911, "epoch": 774, "lr": 8.507775905412766e-05}
{"train_loss": 0.19505318999290466, "global_step": 68912, "epoch": 774, "lr": 8.507734592269304e-05}
{"train_loss": 0.2081354260444641, "global_step": 68913, "epoch": 774, "lr": 8.507693278654268e-05}
{"train_loss": 0.3122105002403259, "global_step": 68914, "epoch": 774, "lr": 8.507651964567665e-05}
{"train_loss": 0.2590996325016022, "global_step": 68915, "epoch": 774, "lr": 8.507610650009502e-05}
{"train_loss": 0.1999501734972, "global_step": 68916, "epoch": 774, "lr": 8.507569334979783e-05}
{"train_loss": 0.26555532217025757, "global_step": 68917, "epoch": 774, "lr": 8.507528019478515e-05}
{"train_loss": 0.22690214216709137, "global_step": 68918, "epoch": 774, "lr": 8.507486703505702e-05}
{"train_loss": 0.24064992368221283, "global_step": 68919, "epoch": 774, "lr": 8.50744538706135e-05}
{"train_loss": 0.2750570774078369, "global_step": 68920, "epoch": 774, "lr": 8.507404070145466e-05}
{"train_loss": 0.20937761664390564, "global_step": 68921, "epoch": 774, "lr": 8.507362752758052e-05}
{"train_loss": 0.19803869724273682, "global_step": 68922, "epoch": 774, "lr": 8.507321434899118e-05}
{"train_loss": 0.21947818994522095, "global_step": 68923, "epoch": 774, "lr": 8.507280116568666e-05}
{"train_loss": 0.20070117712020874, "global_step": 68924, "epoch": 774, "lr": 8.507238797766703e-05}
{"train_loss": 0.2439269721508026, "global_step": 68925, "epoch": 774, "lr": 8.507197478493236e-05}
{"train_loss": 0.16874364018440247, "global_step": 68926, "epoch": 774, "lr": 8.507156158748268e-05}
{"train_loss": 0.21796928346157074, "global_step": 68927, "epoch": 774, "lr": 8.507114838531804e-05}
{"train_loss": 0.1908486783504486, "global_step": 68928, "epoch": 774, "lr": 8.507073517843854e-05}
{"train_loss": 0.2593156099319458, "global_step": 68929, "epoch": 774, "lr": 8.507032196684419e-05}
{"train_loss": 0.25152164697647095, "global_step": 68930, "epoch": 774, "lr": 8.506990875053506e-05}
{"train_loss": 0.22361727058887482, "global_step": 68931, "epoch": 774, "lr": 8.506949552951122e-05}
{"train_loss": 0.24199119210243225, "global_step": 68932, "epoch": 774, "lr": 8.50690823037727e-05}
{"train_loss": 0.387652188539505, "global_step": 68933, "epoch": 774, "lr": 8.506866907331958e-05}
{"train_loss": 0.3141949474811554, "global_step": 68934, "epoch": 774, "lr": 8.506825583815191e-05}
{"train_loss": 0.1968543827533722, "global_step": 68935, "epoch": 774, "lr": 8.506784259826974e-05}
{"train_loss": 0.18288984894752502, "global_step": 68936, "epoch": 774, "lr": 8.506742935367312e-05}
{"train_loss": 0.21223479509353638, "global_step": 68937, "epoch": 774, "lr": 8.506701610436211e-05}
{"train_loss": 0.279738187789917, "global_step": 68938, "epoch": 774, "lr": 8.506660285033676e-05}
{"train_loss": 0.17720571160316467, "global_step": 68939, "epoch": 774, "lr": 8.506618959159716e-05}
{"train_loss": 0.29685503244400024, "global_step": 68940, "epoch": 774, "lr": 8.50657763281433e-05}
{"train_loss": 0.23723754286766052, "global_step": 68941, "epoch": 774, "lr": 8.50653630599753e-05}
{"train_loss": 0.26097550988197327, "global_step": 68942, "epoch": 774, "lr": 8.506494978709318e-05}
{"train_loss": 0.2089482843875885, "global_step": 68943, "epoch": 774, "lr": 8.506453650949701e-05}
{"train_loss": 0.14933855831623077, "global_step": 68944, "epoch": 774, "lr": 8.506412322718685e-05}
{"train_loss": 0.24958018958568573, "global_step": 68945, "epoch": 774, "lr": 8.506370994016271e-05}
{"train_loss": 0.2335038185119629, "global_step": 68946, "epoch": 774, "lr": 8.50632966484247e-05}
{"train_loss": 0.2821076512336731, "global_step": 68947, "epoch": 774, "lr": 8.506288335197287e-05}
{"train_loss": 0.19911594688892365, "global_step": 68948, "epoch": 774, "lr": 8.506247005080726e-05}
{"train_loss": 0.1906118243932724, "global_step": 68949, "epoch": 774, "lr": 8.506205674492792e-05}
{"train_loss": 0.28889939188957214, "global_step": 68950, "epoch": 774, "lr": 8.506164343433491e-05}
{"train_loss": 0.18840600550174713, "global_step": 68951, "epoch": 774, "lr": 8.50612301190283e-05}
{"train_loss": 0.13460376858711243, "global_step": 68952, "epoch": 774, "lr": 8.506081679900812e-05}
{"train_loss": 0.2364058643579483, "global_step": 68953, "epoch": 774, "lr": 8.506040347427446e-05}
{"train_loss": 0.2476254552602768, "global_step": 68954, "epoch": 774, "lr": 8.505999014482735e-05}
{"train_loss": 0.12337194383144379, "global_step": 68955, "epoch": 774, "lr": 8.505957681066686e-05}
{"train_loss": 0.18225736916065216, "global_step": 68956, "epoch": 774, "lr": 8.505916347179301e-05}
{"train_loss": 0.21540668606758118, "global_step": 68957, "epoch": 774, "lr": 8.50587501282059e-05}
{"train_loss": 0.2739771902561188, "global_step": 68958, "epoch": 774, "lr": 8.505833677990558e-05}
{"train_loss": 0.20648209750652313, "global_step": 68959, "epoch": 774, "lr": 8.505792342689208e-05}
{"train_loss": 0.2502394914627075, "global_step": 68960, "epoch": 774, "lr": 8.505751006916548e-05}
{"train_loss": 0.17579986155033112, "global_step": 68961, "epoch": 774, "lr": 8.505709670672582e-05}
{"train_loss": 0.22684037685394287, "global_step": 68962, "epoch": 774, "lr": 8.505668333957314e-05}
{"train_loss": 0.24561063945293427, "global_step": 68963, "epoch": 774, "lr": 8.505626996770752e-05}
{"train_loss": 0.18297462165355682, "global_step": 68964, "epoch": 774, "lr": 8.505585659112902e-05}
{"train_loss": 0.23983517289161682, "global_step": 68965, "epoch": 774, "lr": 8.505544320983772e-05}
{"train_loss": 0.25245580077171326, "global_step": 68966, "epoch": 774, "lr": 8.505502982383359e-05}
{"train_loss": 0.24698401987552643, "global_step": 68967, "epoch": 774, "lr": 8.505461643311678e-05}
{"train_loss": 0.1754937618970871, "global_step": 68968, "epoch": 774, "lr": 8.505420303768729e-05}
{"train_loss": 0.21520152688026428, "global_step": 68969, "epoch": 774, "lr": 8.505378963754518e-05}
{"train_loss": 0.23639987409114838, "global_step": 68970, "epoch": 774, "lr": 8.505337623269052e-05}
{"train_loss": 0.19065210223197937, "global_step": 68971, "epoch": 774, "lr": 8.505296282312337e-05}
{"train_loss": 0.2522965967655182, "global_step": 68972, "epoch": 774, "lr": 8.505254940884378e-05}
{"train_loss": 0.1343042552471161, "global_step": 68973, "epoch": 774, "lr": 8.505213598985178e-05}
{"train_loss": 0.22378672450111153, "global_step": 68974, "epoch": 774, "lr": 8.505172256614747e-05, "val_loss": 3.6756558418273926}
{"train_loss": 0.2247549295425415, "global_step": 68975, "epoch": 775, "lr": 8.505130913773086e-05}
{"train_loss": 0.21292312443256378, "global_step": 68976, "epoch": 775, "lr": 8.505089570460204e-05}
{"train_loss": 0.23663857579231262, "global_step": 68977, "epoch": 775, "lr": 8.505048226676107e-05}
{"train_loss": 0.15631799399852753, "global_step": 68978, "epoch": 775, "lr": 8.505006882420796e-05}
{"train_loss": 0.23518283665180206, "global_step": 68979, "epoch": 775, "lr": 8.504965537694281e-05}
{"train_loss": 0.22538545727729797, "global_step": 68980, "epoch": 775, "lr": 8.504924192496566e-05}
{"train_loss": 0.1444580852985382, "global_step": 68981, "epoch": 775, "lr": 8.504882846827657e-05}
{"train_loss": 0.25611400604248047, "global_step": 68982, "epoch": 775, "lr": 8.504841500687559e-05}
{"train_loss": 0.1713910698890686, "global_step": 68983, "epoch": 775, "lr": 8.504800154076277e-05}
{"train_loss": 0.2078031599521637, "global_step": 68984, "epoch": 775, "lr": 8.504758806993819e-05}
{"train_loss": 0.2038462609052658, "global_step": 68985, "epoch": 775, "lr": 8.504717459440188e-05}
{"train_loss": 0.24980035424232483, "global_step": 68986, "epoch": 775, "lr": 8.50467611141539e-05}
{"train_loss": 0.23373591899871826, "global_step": 68987, "epoch": 775, "lr": 8.504634762919432e-05}
{"train_loss": 0.2217106968164444, "global_step": 68988, "epoch": 775, "lr": 8.504593413952318e-05}
{"train_loss": 0.23613999783992767, "global_step": 68989, "epoch": 775, "lr": 8.504552064514054e-05}
{"train_loss": 0.20131319761276245, "global_step": 68990, "epoch": 775, "lr": 8.504510714604647e-05}
{"train_loss": 0.28341835737228394, "global_step": 68991, "epoch": 775, "lr": 8.5044693642241e-05}
{"train_loss": 0.2868206202983856, "global_step": 68992, "epoch": 775, "lr": 8.504428013372418e-05}
{"train_loss": 0.22626735270023346, "global_step": 68993, "epoch": 775, "lr": 8.504386662049611e-05}
{"train_loss": 0.20133674144744873, "global_step": 68994, "epoch": 775, "lr": 8.504345310255681e-05}
{"train_loss": 0.3181305229663849, "global_step": 68995, "epoch": 775, "lr": 8.504303957990636e-05}
{"train_loss": 0.22518181800842285, "global_step": 68996, "epoch": 775, "lr": 8.504262605254478e-05}
{"train_loss": 0.24178063869476318, "global_step": 68997, "epoch": 775, "lr": 8.504221252047217e-05}
{"train_loss": 0.17178462445735931, "global_step": 68998, "epoch": 775, "lr": 8.504179898368854e-05}
{"train_loss": 0.22859938442707062, "global_step": 68999, "epoch": 775, "lr": 8.504138544219395e-05}
{"train_loss": 0.26474496722221375, "global_step": 69000, "epoch": 775, "lr": 8.50409718959885e-05}
{"train_loss": 0.2850300371646881, "global_step": 69001, "epoch": 775, "lr": 8.504055834507223e-05}
{"train_loss": 0.2883622646331787, "global_step": 69002, "epoch": 775, "lr": 8.504014478944515e-05}
{"train_loss": 0.21057885885238647, "global_step": 69003, "epoch": 775, "lr": 8.503973122910737e-05}
{"train_loss": 0.2670246958732605, "global_step": 69004, "epoch": 775, "lr": 8.503931766405894e-05}
{"train_loss": 0.19814659655094147, "global_step": 69005, "epoch": 775, "lr": 8.503890409429988e-05}
{"train_loss": 0.24174582958221436, "global_step": 69006, "epoch": 775, "lr": 8.503849051983026e-05}
{"train_loss": 0.2310859113931656, "global_step": 69007, "epoch": 775, "lr": 8.503807694065016e-05}
{"train_loss": 0.25367820262908936, "global_step": 69008, "epoch": 775, "lr": 8.503766335675961e-05}
{"train_loss": 0.22879399359226227, "global_step": 69009, "epoch": 775, "lr": 8.503724976815868e-05}
{"train_loss": 0.35241734981536865, "global_step": 69010, "epoch": 775, "lr": 8.503683617484741e-05}
{"train_loss": 0.28338587284088135, "global_step": 69011, "epoch": 775, "lr": 8.503642257682586e-05}
{"train_loss": 0.2593882083892822, "global_step": 69012, "epoch": 775, "lr": 8.50360089740941e-05}
{"train_loss": 0.31584781408309937, "global_step": 69013, "epoch": 775, "lr": 8.503559536665217e-05}
{"train_loss": 0.20495030283927917, "global_step": 69014, "epoch": 775, "lr": 8.503518175450014e-05}
{"train_loss": 0.2475753277540207, "global_step": 69015, "epoch": 775, "lr": 8.503476813763805e-05}
{"train_loss": 0.15493686497211456, "global_step": 69016, "epoch": 775, "lr": 8.503435451606596e-05}
{"train_loss": 0.17434027791023254, "global_step": 69017, "epoch": 775, "lr": 8.503394088978391e-05}
{"train_loss": 0.319712370634079, "global_step": 69018, "epoch": 775, "lr": 8.5033527258792e-05}
{"train_loss": 0.23217011988162994, "global_step": 69019, "epoch": 775, "lr": 8.503311362309025e-05}
{"train_loss": 0.2910715937614441, "global_step": 69020, "epoch": 775, "lr": 8.503269998267874e-05}
{"train_loss": 0.1919897049665451, "global_step": 69021, "epoch": 775, "lr": 8.503228633755749e-05}
{"train_loss": 0.2365625500679016, "global_step": 69022, "epoch": 775, "lr": 8.503187268772659e-05}
{"train_loss": 0.25923487544059753, "global_step": 69023, "epoch": 775, "lr": 8.503145903318608e-05}
{"train_loss": 0.1909647136926651, "global_step": 69024, "epoch": 775, "lr": 8.503104537393602e-05}
{"train_loss": 0.16346396505832672, "global_step": 69025, "epoch": 775, "lr": 8.503063170997645e-05}
{"train_loss": 0.26907792687416077, "global_step": 69026, "epoch": 775, "lr": 8.503021804130744e-05}
{"train_loss": 0.2883657217025757, "global_step": 69027, "epoch": 775, "lr": 8.502980436792905e-05}
{"train_loss": 0.24601921439170837, "global_step": 69028, "epoch": 775, "lr": 8.502939068984132e-05}
{"train_loss": 0.20040813088417053, "global_step": 69029, "epoch": 775, "lr": 8.502897700704433e-05}
{"train_loss": 0.20764568448066711, "global_step": 69030, "epoch": 775, "lr": 8.502856331953812e-05}
{"train_loss": 0.22592492401599884, "global_step": 69031, "epoch": 775, "lr": 8.502814962732274e-05}
{"train_loss": 0.2658126950263977, "global_step": 69032, "epoch": 775, "lr": 8.502773593039826e-05}
{"train_loss": 0.19925767183303833, "global_step": 69033, "epoch": 775, "lr": 8.502732222876472e-05}
{"train_loss": 0.23203715682029724, "global_step": 69034, "epoch": 775, "lr": 8.502690852242219e-05}
{"train_loss": 0.15748661756515503, "global_step": 69035, "epoch": 775, "lr": 8.50264948113707e-05}
{"train_loss": 0.14419037103652954, "global_step": 69036, "epoch": 775, "lr": 8.502608109561034e-05}
{"train_loss": 0.26275867223739624, "global_step": 69037, "epoch": 775, "lr": 8.502566737514115e-05}
{"train_loss": 0.2755145728588104, "global_step": 69038, "epoch": 775, "lr": 8.502525364996318e-05}
{"train_loss": 0.1679748296737671, "global_step": 69039, "epoch": 775, "lr": 8.50248399200765e-05}
{"train_loss": 0.2694219946861267, "global_step": 69040, "epoch": 775, "lr": 8.502442618548117e-05}
{"train_loss": 0.21484418213367462, "global_step": 69041, "epoch": 775, "lr": 8.50240124461772e-05}
{"train_loss": 0.2678392827510834, "global_step": 69042, "epoch": 775, "lr": 8.50235987021647e-05}
{"train_loss": 0.32621148228645325, "global_step": 69043, "epoch": 775, "lr": 8.50231849534437e-05}
{"train_loss": 0.2075171321630478, "global_step": 69044, "epoch": 775, "lr": 8.502277120001426e-05}
{"train_loss": 0.19801488518714905, "global_step": 69045, "epoch": 775, "lr": 8.502235744187643e-05}
{"train_loss": 0.2907869517803192, "global_step": 69046, "epoch": 775, "lr": 8.502194367903028e-05}
{"train_loss": 0.18353042006492615, "global_step": 69047, "epoch": 775, "lr": 8.502152991147585e-05}
{"train_loss": 0.2357814908027649, "global_step": 69048, "epoch": 775, "lr": 8.502111613921321e-05}
{"train_loss": 0.24926510453224182, "global_step": 69049, "epoch": 775, "lr": 8.50207023622424e-05}
{"train_loss": 0.18756349384784698, "global_step": 69050, "epoch": 775, "lr": 8.502028858056348e-05}
{"train_loss": 0.25024545192718506, "global_step": 69051, "epoch": 775, "lr": 8.501987479417653e-05}
{"train_loss": 0.2719455063343048, "global_step": 69052, "epoch": 775, "lr": 8.501946100308155e-05}
{"train_loss": 0.24915501475334167, "global_step": 69053, "epoch": 775, "lr": 8.501904720727865e-05}
{"train_loss": 0.2056405246257782, "global_step": 69054, "epoch": 775, "lr": 8.501863340676786e-05}
{"train_loss": 0.3150705397129059, "global_step": 69055, "epoch": 775, "lr": 8.501821960154926e-05}
{"train_loss": 0.20833922922611237, "global_step": 69056, "epoch": 775, "lr": 8.501780579162287e-05}
{"train_loss": 0.2817549705505371, "global_step": 69057, "epoch": 775, "lr": 8.501739197698877e-05}
{"train_loss": 0.2739477753639221, "global_step": 69058, "epoch": 775, "lr": 8.5016978157647e-05}
{"train_loss": 0.250977098941803, "global_step": 69059, "epoch": 775, "lr": 8.501656433359762e-05}
{"train_loss": 0.23639419674873352, "global_step": 69060, "epoch": 775, "lr": 8.501615050484071e-05}
{"train_loss": 0.3187803626060486, "global_step": 69061, "epoch": 775, "lr": 8.501573667137628e-05}
{"train_loss": 0.2579917013645172, "global_step": 69062, "epoch": 775, "lr": 8.501532283320444e-05}
{"train_loss": 0.2369537040423811, "global_step": 69063, "epoch": 775, "lr": 8.50149089903252e-05, "val_loss": 3.581761598587036, "train_action_mse_error": 21.563411712646484}
{"train_loss": 0.20746013522148132, "global_step": 69064, "epoch": 776, "lr": 8.501449514273864e-05}
{"train_loss": 0.3221054971218109, "global_step": 69065, "epoch": 776, "lr": 8.501408129044479e-05}
{"train_loss": 0.20568162202835083, "global_step": 69066, "epoch": 776, "lr": 8.501366743344374e-05}
{"train_loss": 0.1573306918144226, "global_step": 69067, "epoch": 776, "lr": 8.501325357173553e-05}
{"train_loss": 0.17718198895454407, "global_step": 69068, "epoch": 776, "lr": 8.501283970532022e-05}
{"train_loss": 0.17458876967430115, "global_step": 69069, "epoch": 776, "lr": 8.501242583419784e-05}
{"train_loss": 0.23161248862743378, "global_step": 69070, "epoch": 776, "lr": 8.501201195836848e-05}
{"train_loss": 0.15334907174110413, "global_step": 69071, "epoch": 776, "lr": 8.501159807783216e-05}
{"train_loss": 0.18288150429725647, "global_step": 69072, "epoch": 776, "lr": 8.501118419258897e-05}
{"train_loss": 0.2553403675556183, "global_step": 69073, "epoch": 776, "lr": 8.501077030263898e-05}
{"train_loss": 0.1799672394990921, "global_step": 69074, "epoch": 776, "lr": 8.501035640798219e-05}
{"train_loss": 0.18414048850536346, "global_step": 69075, "epoch": 776, "lr": 8.500994250861869e-05}
{"train_loss": 0.214918315410614, "global_step": 69076, "epoch": 776, "lr": 8.500952860454853e-05}
{"train_loss": 0.20393633842468262, "global_step": 69077, "epoch": 776, "lr": 8.500911469577176e-05}
{"train_loss": 0.22070500254631042, "global_step": 69078, "epoch": 776, "lr": 8.500870078228845e-05}
{"train_loss": 0.21239601075649261, "global_step": 69079, "epoch": 776, "lr": 8.500828686409865e-05}
{"train_loss": 0.26222652196884155, "global_step": 69080, "epoch": 776, "lr": 8.50078729412024e-05}
{"train_loss": 0.20842353999614716, "global_step": 69081, "epoch": 776, "lr": 8.500745901359979e-05}
{"train_loss": 0.3066446781158447, "global_step": 69082, "epoch": 776, "lr": 8.500704508129083e-05}
{"train_loss": 0.17500652372837067, "global_step": 69083, "epoch": 776, "lr": 8.50066311442756e-05}
{"train_loss": 0.2330581545829773, "global_step": 69084, "epoch": 776, "lr": 8.500621720255417e-05}
{"train_loss": 0.1964396834373474, "global_step": 69085, "epoch": 776, "lr": 8.500580325612657e-05}
{"train_loss": 0.13041329383850098, "global_step": 69086, "epoch": 776, "lr": 8.500538930499288e-05}
{"train_loss": 0.205570787191391, "global_step": 69087, "epoch": 776, "lr": 8.500497534915313e-05}
{"train_loss": 0.20341980457305908, "global_step": 69088, "epoch": 776, "lr": 8.50045613886074e-05}
{"train_loss": 0.18421563506126404, "global_step": 69089, "epoch": 776, "lr": 8.500414742335572e-05}
{"train_loss": 0.23421968519687653, "global_step": 69090, "epoch": 776, "lr": 8.500373345339815e-05}
{"train_loss": 0.17658373713493347, "global_step": 69091, "epoch": 776, "lr": 8.500331947873478e-05}
{"train_loss": 0.1618572622537613, "global_step": 69092, "epoch": 776, "lr": 8.500290549936563e-05}
{"train_loss": 0.2037135660648346, "global_step": 69093, "epoch": 776, "lr": 8.500249151529076e-05}
{"train_loss": 0.31498482823371887, "global_step": 69094, "epoch": 776, "lr": 8.500207752651023e-05}
{"train_loss": 0.21903568506240845, "global_step": 69095, "epoch": 776, "lr": 8.500166353302411e-05}
{"train_loss": 0.27424728870391846, "global_step": 69096, "epoch": 776, "lr": 8.500124953483243e-05}
{"train_loss": 0.20806413888931274, "global_step": 69097, "epoch": 776, "lr": 8.500083553193527e-05}
{"train_loss": 0.2607939541339874, "global_step": 69098, "epoch": 776, "lr": 8.500042152433266e-05}
{"train_loss": 0.25301218032836914, "global_step": 69099, "epoch": 776, "lr": 8.500000751202469e-05}
{"train_loss": 0.12112927436828613, "global_step": 69100, "epoch": 776, "lr": 8.499959349501137e-05}
{"train_loss": 0.22143027186393738, "global_step": 69101, "epoch": 776, "lr": 8.499917947329281e-05}
{"train_loss": 0.40934792160987854, "global_step": 69102, "epoch": 776, "lr": 8.499876544686902e-05}
{"train_loss": 0.2818683683872223, "global_step": 69103, "epoch": 776, "lr": 8.499835141574007e-05}
{"train_loss": 0.19138173758983612, "global_step": 69104, "epoch": 776, "lr": 8.499793737990602e-05}
{"train_loss": 0.21383486688137054, "global_step": 69105, "epoch": 776, "lr": 8.499752333936693e-05}
{"train_loss": 0.19005051255226135, "global_step": 69106, "epoch": 776, "lr": 8.499710929412285e-05}
{"train_loss": 0.2506365180015564, "global_step": 69107, "epoch": 776, "lr": 8.499669524417385e-05}
{"train_loss": 0.24723336100578308, "global_step": 69108, "epoch": 776, "lr": 8.499628118951994e-05}
{"train_loss": 0.2496584951877594, "global_step": 69109, "epoch": 776, "lr": 8.499586713016121e-05}
{"train_loss": 0.26154568791389465, "global_step": 69110, "epoch": 776, "lr": 8.499545306609774e-05}
{"train_loss": 0.21964040398597717, "global_step": 69111, "epoch": 776, "lr": 8.499503899732952e-05}
{"train_loss": 0.24940691888332367, "global_step": 69112, "epoch": 776, "lr": 8.499462492385666e-05}
{"train_loss": 0.22511982917785645, "global_step": 69113, "epoch": 776, "lr": 8.49942108456792e-05}
{"train_loss": 0.1778515726327896, "global_step": 69114, "epoch": 776, "lr": 8.499379676279721e-05}
{"train_loss": 0.22794823348522186, "global_step": 69115, "epoch": 776, "lr": 8.499338267521071e-05}
{"train_loss": 0.3009227514266968, "global_step": 69116, "epoch": 776, "lr": 8.499296858291979e-05}
{"train_loss": 0.3308876156806946, "global_step": 69117, "epoch": 776, "lr": 8.499255448592448e-05}
{"train_loss": 0.2754358947277069, "global_step": 69118, "epoch": 776, "lr": 8.499214038422485e-05}
{"train_loss": 0.2461770623922348, "global_step": 69119, "epoch": 776, "lr": 8.499172627782096e-05}
{"train_loss": 0.24695731699466705, "global_step": 69120, "epoch": 776, "lr": 8.499131216671286e-05}
{"train_loss": 0.18028949201107025, "global_step": 69121, "epoch": 776, "lr": 8.49908980509006e-05}
{"train_loss": 0.21317972242832184, "global_step": 69122, "epoch": 776, "lr": 8.499048393038425e-05}
{"train_loss": 0.2404894232749939, "global_step": 69123, "epoch": 776, "lr": 8.499006980516384e-05}
{"train_loss": 0.2798171043395996, "global_step": 69124, "epoch": 776, "lr": 8.498965567523944e-05}
{"train_loss": 0.19946976006031036, "global_step": 69125, "epoch": 776, "lr": 8.498924154061112e-05}
{"train_loss": 0.1864730417728424, "global_step": 69126, "epoch": 776, "lr": 8.498882740127891e-05}
{"train_loss": 0.22891345620155334, "global_step": 69127, "epoch": 776, "lr": 8.49884132572429e-05}
{"train_loss": 0.282701700925827, "global_step": 69128, "epoch": 776, "lr": 8.49879991085031e-05}
{"train_loss": 0.30306750535964966, "global_step": 69129, "epoch": 776, "lr": 8.49875849550596e-05}
{"train_loss": 0.1756688505411148, "global_step": 69130, "epoch": 776, "lr": 8.498717079691245e-05}
{"train_loss": 0.16377681493759155, "global_step": 69131, "epoch": 776, "lr": 8.49867566340617e-05}
{"train_loss": 0.25830313563346863, "global_step": 69132, "epoch": 776, "lr": 8.49863424665074e-05}
{"train_loss": 0.21206070482730865, "global_step": 69133, "epoch": 776, "lr": 8.498592829424962e-05}
{"train_loss": 0.1864227056503296, "global_step": 69134, "epoch": 776, "lr": 8.49855141172884e-05}
{"train_loss": 0.16011254489421844, "global_step": 69135, "epoch": 776, "lr": 8.49850999356238e-05}
{"train_loss": 0.20534782111644745, "global_step": 69136, "epoch": 776, "lr": 8.49846857492559e-05}
{"train_loss": 0.1544346660375595, "global_step": 69137, "epoch": 776, "lr": 8.498427155818472e-05}
{"train_loss": 0.1953396499156952, "global_step": 69138, "epoch": 776, "lr": 8.498385736241035e-05}
{"train_loss": 0.18018220365047455, "global_step": 69139, "epoch": 776, "lr": 8.49834431619328e-05}
{"train_loss": 0.17918793857097626, "global_step": 69140, "epoch": 776, "lr": 8.498302895675217e-05}
{"train_loss": 0.20302842557430267, "global_step": 69141, "epoch": 776, "lr": 8.498261474686849e-05}
{"train_loss": 0.27651992440223694, "global_step": 69142, "epoch": 776, "lr": 8.498220053228183e-05}
{"train_loss": 0.26733699440956116, "global_step": 69143, "epoch": 776, "lr": 8.498178631299224e-05}
{"train_loss": 0.2094571441411972, "global_step": 69144, "epoch": 776, "lr": 8.498137208899977e-05}
{"train_loss": 0.16707125306129456, "global_step": 69145, "epoch": 776, "lr": 8.498095786030446e-05}
{"train_loss": 0.1894734650850296, "global_step": 69146, "epoch": 776, "lr": 8.498054362690643e-05}
{"train_loss": 0.2057141810655594, "global_step": 69147, "epoch": 776, "lr": 8.498012938880566e-05}
{"train_loss": 0.21728266775608063, "global_step": 69148, "epoch": 776, "lr": 8.497971514600226e-05}
{"train_loss": 0.20661814510822296, "global_step": 69149, "epoch": 776, "lr": 8.497930089849625e-05}
{"train_loss": 0.17531119287014008, "global_step": 69150, "epoch": 776, "lr": 8.49788866462877e-05}
{"train_loss": 0.21176008880138397, "global_step": 69151, "epoch": 776, "lr": 8.497847238937668e-05}
{"train_loss": 0.22033409472931637, "global_step": 69152, "epoch": 776, "lr": 8.497805812776321e-05, "val_loss": 3.6129746437072754}
{"train_loss": 0.20852506160736084, "global_step": 69153, "epoch": 777, "lr": 8.497764386144738e-05}
{"train_loss": 0.16757002472877502, "global_step": 69154, "epoch": 777, "lr": 8.497722959042922e-05}
{"train_loss": 0.20505951344966888, "global_step": 69155, "epoch": 777, "lr": 8.497681531470881e-05}
{"train_loss": 0.2114787995815277, "global_step": 69156, "epoch": 777, "lr": 8.497640103428619e-05}
{"train_loss": 0.23718883097171783, "global_step": 69157, "epoch": 777, "lr": 8.49759867491614e-05}
{"train_loss": 0.21842041611671448, "global_step": 69158, "epoch": 777, "lr": 8.497557245933454e-05}
{"train_loss": 0.28700509667396545, "global_step": 69159, "epoch": 777, "lr": 8.497515816480564e-05}
{"train_loss": 0.2836487591266632, "global_step": 69160, "epoch": 777, "lr": 8.497474386557474e-05}
{"train_loss": 0.2004258781671524, "global_step": 69161, "epoch": 777, "lr": 8.497432956164194e-05}
{"train_loss": 0.1505763977766037, "global_step": 69162, "epoch": 777, "lr": 8.497391525300723e-05}
{"train_loss": 0.19732381403446198, "global_step": 69163, "epoch": 777, "lr": 8.497350093967071e-05}
{"train_loss": 0.13132432103157043, "global_step": 69164, "epoch": 777, "lr": 8.497308662163246e-05}
{"train_loss": 0.24571342766284943, "global_step": 69165, "epoch": 777, "lr": 8.497267229889247e-05}
{"train_loss": 0.3935201168060303, "global_step": 69166, "epoch": 777, "lr": 8.497225797145085e-05}
{"train_loss": 0.31108394265174866, "global_step": 69167, "epoch": 777, "lr": 8.497184363930763e-05}
{"train_loss": 0.2508653998374939, "global_step": 69168, "epoch": 777, "lr": 8.497142930246289e-05}
{"train_loss": 0.19865873456001282, "global_step": 69169, "epoch": 777, "lr": 8.497101496091664e-05}
{"train_loss": 0.21094869077205658, "global_step": 69170, "epoch": 777, "lr": 8.497060061466898e-05}
{"train_loss": 0.25086909532546997, "global_step": 69171, "epoch": 777, "lr": 8.497018626371994e-05}
{"train_loss": 0.16313035786151886, "global_step": 69172, "epoch": 777, "lr": 8.49697719080696e-05}
{"train_loss": 0.25791358947753906, "global_step": 69173, "epoch": 777, "lr": 8.496935754771797e-05}
{"train_loss": 0.24306361377239227, "global_step": 69174, "epoch": 777, "lr": 8.496894318266518e-05}
{"train_loss": 0.2271772027015686, "global_step": 69175, "epoch": 777, "lr": 8.49685288129112e-05}
{"train_loss": 0.18806886672973633, "global_step": 69176, "epoch": 777, "lr": 8.496811443845615e-05}
{"train_loss": 0.19792762398719788, "global_step": 69177, "epoch": 777, "lr": 8.496770005930005e-05}
{"train_loss": 0.2471165955066681, "global_step": 69178, "epoch": 777, "lr": 8.496728567544299e-05}
{"train_loss": 0.21420818567276, "global_step": 69179, "epoch": 777, "lr": 8.496687128688499e-05}
{"train_loss": 0.18141447007656097, "global_step": 69180, "epoch": 777, "lr": 8.496645689362611e-05}
{"train_loss": 0.1629796177148819, "global_step": 69181, "epoch": 777, "lr": 8.496604249566644e-05}
{"train_loss": 0.17272736132144928, "global_step": 69182, "epoch": 777, "lr": 8.496562809300599e-05}
{"train_loss": 0.1671769618988037, "global_step": 69183, "epoch": 777, "lr": 8.496521368564485e-05}
{"train_loss": 0.1714540272951126, "global_step": 69184, "epoch": 777, "lr": 8.496479927358305e-05}
{"train_loss": 0.2805135250091553, "global_step": 69185, "epoch": 777, "lr": 8.496438485682068e-05}
{"train_loss": 0.1787678599357605, "global_step": 69186, "epoch": 777, "lr": 8.496397043535777e-05}
{"train_loss": 0.19716307520866394, "global_step": 69187, "epoch": 777, "lr": 8.496355600919437e-05}
{"train_loss": 0.21435098350048065, "global_step": 69188, "epoch": 777, "lr": 8.496314157833054e-05}
{"train_loss": 0.26712316274642944, "global_step": 69189, "epoch": 777, "lr": 8.496272714276636e-05}
{"train_loss": 0.14856809377670288, "global_step": 69190, "epoch": 777, "lr": 8.496231270250185e-05}
{"train_loss": 0.2314063459634781, "global_step": 69191, "epoch": 777, "lr": 8.496189825753709e-05}
{"train_loss": 0.18999452888965607, "global_step": 69192, "epoch": 777, "lr": 8.496148380787214e-05}
{"train_loss": 0.18678180873394012, "global_step": 69193, "epoch": 777, "lr": 8.496106935350704e-05}
{"train_loss": 0.16027437150478363, "global_step": 69194, "epoch": 777, "lr": 8.496065489444184e-05}
{"train_loss": 0.20250563323497772, "global_step": 69195, "epoch": 777, "lr": 8.49602404306766e-05}
{"train_loss": 0.22177384793758392, "global_step": 69196, "epoch": 777, "lr": 8.495982596221141e-05}
{"train_loss": 0.1441710740327835, "global_step": 69197, "epoch": 777, "lr": 8.49594114890463e-05}
{"train_loss": 0.21622885763645172, "global_step": 69198, "epoch": 777, "lr": 8.495899701118129e-05}
{"train_loss": 0.26654520630836487, "global_step": 69199, "epoch": 777, "lr": 8.495858252861649e-05}
{"train_loss": 0.24972014129161835, "global_step": 69200, "epoch": 777, "lr": 8.495816804135192e-05}
{"train_loss": 0.216892272233963, "global_step": 69201, "epoch": 777, "lr": 8.495775354938765e-05}
{"train_loss": 0.25222283601760864, "global_step": 69202, "epoch": 777, "lr": 8.495733905272374e-05}
{"train_loss": 0.22649255394935608, "global_step": 69203, "epoch": 777, "lr": 8.495692455136027e-05}
{"train_loss": 0.2031688541173935, "global_step": 69204, "epoch": 777, "lr": 8.495651004529723e-05}
{"train_loss": 0.26739639043807983, "global_step": 69205, "epoch": 777, "lr": 8.495609553453473e-05}
{"train_loss": 0.19699600338935852, "global_step": 69206, "epoch": 777, "lr": 8.49556810190728e-05}
{"train_loss": 0.2488514482975006, "global_step": 69207, "epoch": 777, "lr": 8.495526649891152e-05}
{"train_loss": 0.26480263471603394, "global_step": 69208, "epoch": 777, "lr": 8.495485197405092e-05}
{"train_loss": 0.22382725775241852, "global_step": 69209, "epoch": 777, "lr": 8.495443744449107e-05}
{"train_loss": 0.29232358932495117, "global_step": 69210, "epoch": 777, "lr": 8.495402291023202e-05}
{"train_loss": 0.29202210903167725, "global_step": 69211, "epoch": 777, "lr": 8.495360837127382e-05}
{"train_loss": 0.21230269968509674, "global_step": 69212, "epoch": 777, "lr": 8.495319382761655e-05}
{"train_loss": 0.1839660257101059, "global_step": 69213, "epoch": 777, "lr": 8.495277927926024e-05}
{"train_loss": 0.2548350393772125, "global_step": 69214, "epoch": 777, "lr": 8.495236472620496e-05}
{"train_loss": 0.1410990059375763, "global_step": 69215, "epoch": 777, "lr": 8.495195016845075e-05}
{"train_loss": 0.291056364774704, "global_step": 69216, "epoch": 777, "lr": 8.495153560599768e-05}
{"train_loss": 0.2931838035583496, "global_step": 69217, "epoch": 777, "lr": 8.49511210388458e-05}
{"train_loss": 0.2463652640581131, "global_step": 69218, "epoch": 777, "lr": 8.495070646699517e-05}
{"train_loss": 0.28027239441871643, "global_step": 69219, "epoch": 777, "lr": 8.495029189044585e-05}
{"train_loss": 0.2219880372285843, "global_step": 69220, "epoch": 777, "lr": 8.494987730919788e-05}
{"train_loss": 0.23352459073066711, "global_step": 69221, "epoch": 777, "lr": 8.494946272325133e-05}
{"train_loss": 0.14109937846660614, "global_step": 69222, "epoch": 777, "lr": 8.494904813260625e-05}
{"train_loss": 0.32177215814590454, "global_step": 69223, "epoch": 777, "lr": 8.49486335372627e-05}
{"train_loss": 0.24496515095233917, "global_step": 69224, "epoch": 777, "lr": 8.494821893722073e-05}
{"train_loss": 0.23847182095050812, "global_step": 69225, "epoch": 777, "lr": 8.494780433248039e-05}
{"train_loss": 0.22820888459682465, "global_step": 69226, "epoch": 777, "lr": 8.494738972304175e-05}
{"train_loss": 0.2926844656467438, "global_step": 69227, "epoch": 777, "lr": 8.494697510890486e-05}
{"train_loss": 0.28055697679519653, "global_step": 69228, "epoch": 777, "lr": 8.494656049006977e-05}
{"train_loss": 0.19580785930156708, "global_step": 69229, "epoch": 777, "lr": 8.494614586653655e-05}
{"train_loss": 0.19843026995658875, "global_step": 69230, "epoch": 777, "lr": 8.494573123830525e-05}
{"train_loss": 0.273275226354599, "global_step": 69231, "epoch": 777, "lr": 8.49453166053759e-05}
{"train_loss": 0.23230698704719543, "global_step": 69232, "epoch": 777, "lr": 8.49449019677486e-05}
{"train_loss": 0.23505520820617676, "global_step": 69233, "epoch": 777, "lr": 8.494448732542337e-05}
{"train_loss": 0.23197269439697266, "global_step": 69234, "epoch": 777, "lr": 8.49440726784003e-05}
{"train_loss": 0.2849180996417999, "global_step": 69235, "epoch": 777, "lr": 8.494365802667941e-05}
{"train_loss": 0.2720847725868225, "global_step": 69236, "epoch": 777, "lr": 8.494324337026076e-05}
{"train_loss": 0.23983170092105865, "global_step": 69237, "epoch": 777, "lr": 8.494282870914443e-05}
{"train_loss": 0.16983206570148468, "global_step": 69238, "epoch": 777, "lr": 8.494241404333047e-05}
{"train_loss": 0.22197705507278442, "global_step": 69239, "epoch": 777, "lr": 8.494199937281893e-05}
{"train_loss": 0.24263140559196472, "global_step": 69240, "epoch": 777, "lr": 8.494158469760984e-05}
{"train_loss": 0.22619272048553724, "global_step": 69241, "epoch": 777, "lr": 8.49411700177033e-05, "val_loss": 3.6135165691375732}
{"train_loss": 0.23301053047180176, "global_step": 69242, "epoch": 778, "lr": 8.494075533309933e-05}
{"train_loss": 0.1885421723127365, "global_step": 69243, "epoch": 778, "lr": 8.494034064379802e-05}
{"train_loss": 0.4073677957057953, "global_step": 69244, "epoch": 778, "lr": 8.49399259497994e-05}
{"train_loss": 0.2941662073135376, "global_step": 69245, "epoch": 778, "lr": 8.493951125110353e-05}
{"train_loss": 0.2068466991186142, "global_step": 69246, "epoch": 778, "lr": 8.493909654771047e-05}
{"train_loss": 0.24446739256381989, "global_step": 69247, "epoch": 778, "lr": 8.493868183962027e-05}
{"train_loss": 0.33498990535736084, "global_step": 69248, "epoch": 778, "lr": 8.493826712683299e-05}
{"train_loss": 0.3059532046318054, "global_step": 69249, "epoch": 778, "lr": 8.493785240934869e-05}
{"train_loss": 0.14408347010612488, "global_step": 69250, "epoch": 778, "lr": 8.493743768716742e-05}
{"train_loss": 0.23085135221481323, "global_step": 69251, "epoch": 778, "lr": 8.493702296028925e-05}
{"train_loss": 0.2627902328968048, "global_step": 69252, "epoch": 778, "lr": 8.493660822871421e-05}
{"train_loss": 0.2622351050376892, "global_step": 69253, "epoch": 778, "lr": 8.493619349244238e-05}
{"train_loss": 0.29260748624801636, "global_step": 69254, "epoch": 778, "lr": 8.493577875147379e-05}
{"train_loss": 0.23133744299411774, "global_step": 69255, "epoch": 778, "lr": 8.49353640058085e-05}
{"train_loss": 0.28286731243133545, "global_step": 69256, "epoch": 778, "lr": 8.493494925544661e-05}
{"train_loss": 0.1416148841381073, "global_step": 69257, "epoch": 778, "lr": 8.49345345003881e-05}
{"train_loss": 0.2814777195453644, "global_step": 69258, "epoch": 778, "lr": 8.49341197406331e-05}
{"train_loss": 0.2121279537677765, "global_step": 69259, "epoch": 778, "lr": 8.493370497618163e-05}
{"train_loss": 0.2760325074195862, "global_step": 69260, "epoch": 778, "lr": 8.493329020703373e-05}
{"train_loss": 0.27938205003738403, "global_step": 69261, "epoch": 778, "lr": 8.49328754331895e-05}
{"train_loss": 0.2348211705684662, "global_step": 69262, "epoch": 778, "lr": 8.493246065464895e-05}
{"train_loss": 0.22135180234909058, "global_step": 69263, "epoch": 778, "lr": 8.493204587141217e-05}
{"train_loss": 0.20984424650669098, "global_step": 69264, "epoch": 778, "lr": 8.49316310834792e-05}
{"train_loss": 0.20754334330558777, "global_step": 69265, "epoch": 778, "lr": 8.49312162908501e-05}
{"train_loss": 0.21109986305236816, "global_step": 69266, "epoch": 778, "lr": 8.493080149352493e-05}
{"train_loss": 0.27389752864837646, "global_step": 69267, "epoch": 778, "lr": 8.493038669150373e-05}
{"train_loss": 0.14460809528827667, "global_step": 69268, "epoch": 778, "lr": 8.492997188478656e-05}
{"train_loss": 0.2135331630706787, "global_step": 69269, "epoch": 778, "lr": 8.492955707337347e-05}
{"train_loss": 0.2501586675643921, "global_step": 69270, "epoch": 778, "lr": 8.492914225726455e-05}
{"train_loss": 0.22291532158851624, "global_step": 69271, "epoch": 778, "lr": 8.492872743645982e-05}
{"train_loss": 0.251491516828537, "global_step": 69272, "epoch": 778, "lr": 8.492831261095935e-05}
{"train_loss": 0.19724440574645996, "global_step": 69273, "epoch": 778, "lr": 8.49278977807632e-05}
{"train_loss": 0.13753850758075714, "global_step": 69274, "epoch": 778, "lr": 8.492748294587141e-05}
{"train_loss": 0.23156721889972687, "global_step": 69275, "epoch": 778, "lr": 8.492706810628406e-05}
{"train_loss": 0.15752887725830078, "global_step": 69276, "epoch": 778, "lr": 8.492665326200118e-05}
{"train_loss": 0.24795114994049072, "global_step": 69277, "epoch": 778, "lr": 8.492623841302285e-05}
{"train_loss": 0.17517727613449097, "global_step": 69278, "epoch": 778, "lr": 8.49258235593491e-05}
{"train_loss": 0.25790226459503174, "global_step": 69279, "epoch": 778, "lr": 8.492540870098e-05}
{"train_loss": 0.25762122869491577, "global_step": 69280, "epoch": 778, "lr": 8.492499383791561e-05}
{"train_loss": 0.2006845474243164, "global_step": 69281, "epoch": 778, "lr": 8.492457897015597e-05}
{"train_loss": 0.25254514813423157, "global_step": 69282, "epoch": 778, "lr": 8.492416409770117e-05}
{"train_loss": 0.3276577889919281, "global_step": 69283, "epoch": 778, "lr": 8.492374922055121e-05}
{"train_loss": 0.17236587405204773, "global_step": 69284, "epoch": 778, "lr": 8.49233343387062e-05}
{"train_loss": 0.23065102100372314, "global_step": 69285, "epoch": 778, "lr": 8.492291945216618e-05}
{"train_loss": 0.2041238248348236, "global_step": 69286, "epoch": 778, "lr": 8.492250456093119e-05}
{"train_loss": 0.21355296671390533, "global_step": 69287, "epoch": 778, "lr": 8.492208966500128e-05}
{"train_loss": 0.18909338116645813, "global_step": 69288, "epoch": 778, "lr": 8.492167476437653e-05}
{"train_loss": 0.2744239866733551, "global_step": 69289, "epoch": 778, "lr": 8.492125985905698e-05}
{"train_loss": 0.18720297515392303, "global_step": 69290, "epoch": 778, "lr": 8.492084494904269e-05}
{"train_loss": 0.2194674015045166, "global_step": 69291, "epoch": 778, "lr": 8.492043003433374e-05}
{"train_loss": 0.12682467699050903, "global_step": 69292, "epoch": 778, "lr": 8.492001511493015e-05}
{"train_loss": 0.1561623513698578, "global_step": 69293, "epoch": 778, "lr": 8.491960019083199e-05}
{"train_loss": 0.244197279214859, "global_step": 69294, "epoch": 778, "lr": 8.491918526203932e-05}
{"train_loss": 0.13840746879577637, "global_step": 69295, "epoch": 778, "lr": 8.491877032855218e-05}
{"train_loss": 0.19168546795845032, "global_step": 69296, "epoch": 778, "lr": 8.491835539037066e-05}
{"train_loss": 0.21524953842163086, "global_step": 69297, "epoch": 778, "lr": 8.491794044749477e-05}
{"train_loss": 0.22906330227851868, "global_step": 69298, "epoch": 778, "lr": 8.49175254999246e-05}
{"train_loss": 0.20148614048957825, "global_step": 69299, "epoch": 778, "lr": 8.491711054766018e-05}
{"train_loss": 0.19783924520015717, "global_step": 69300, "epoch": 778, "lr": 8.491669559070159e-05}
{"train_loss": 0.22677753865718842, "global_step": 69301, "epoch": 778, "lr": 8.491628062904889e-05}
{"train_loss": 0.244329035282135, "global_step": 69302, "epoch": 778, "lr": 8.49158656627021e-05}
{"train_loss": 0.1567993313074112, "global_step": 69303, "epoch": 778, "lr": 8.49154506916613e-05}
{"train_loss": 0.23090100288391113, "global_step": 69304, "epoch": 778, "lr": 8.491503571592655e-05}
{"train_loss": 0.21989499032497406, "global_step": 69305, "epoch": 778, "lr": 8.49146207354979e-05}
{"train_loss": 0.18289169669151306, "global_step": 69306, "epoch": 778, "lr": 8.49142057503754e-05}
{"train_loss": 0.26474159955978394, "global_step": 69307, "epoch": 778, "lr": 8.49137907605591e-05}
{"train_loss": 0.2211582213640213, "global_step": 69308, "epoch": 778, "lr": 8.491337576604909e-05}
{"train_loss": 0.21033334732055664, "global_step": 69309, "epoch": 778, "lr": 8.491296076684538e-05}
{"train_loss": 0.20479129254817963, "global_step": 69310, "epoch": 778, "lr": 8.491254576294808e-05}
{"train_loss": 0.20290404558181763, "global_step": 69311, "epoch": 778, "lr": 8.491213075435717e-05}
{"train_loss": 0.19966267049312592, "global_step": 69312, "epoch": 778, "lr": 8.491171574107279e-05}
{"train_loss": 0.25427868962287903, "global_step": 69313, "epoch": 778, "lr": 8.491130072309492e-05}
{"train_loss": 0.19323129951953888, "global_step": 69314, "epoch": 778, "lr": 8.491088570042369e-05}
{"train_loss": 0.1827944666147232, "global_step": 69315, "epoch": 778, "lr": 8.491047067305908e-05}
{"train_loss": 0.21381217241287231, "global_step": 69316, "epoch": 778, "lr": 8.49100556410012e-05}
{"train_loss": 0.1498045027256012, "global_step": 69317, "epoch": 778, "lr": 8.490964060425009e-05}
{"train_loss": 0.22462312877178192, "global_step": 69318, "epoch": 778, "lr": 8.49092255628058e-05}
{"train_loss": 0.16846339404582977, "global_step": 69319, "epoch": 778, "lr": 8.49088105166684e-05}
{"train_loss": 0.23780545592308044, "global_step": 69320, "epoch": 778, "lr": 8.490839546583793e-05}
{"train_loss": 0.25631895661354065, "global_step": 69321, "epoch": 778, "lr": 8.490798041031445e-05}
{"train_loss": 0.2089475840330124, "global_step": 69322, "epoch": 778, "lr": 8.4907565350098e-05}
{"train_loss": 0.2595329284667969, "global_step": 69323, "epoch": 778, "lr": 8.490715028518868e-05}
{"train_loss": 0.3156917095184326, "global_step": 69324, "epoch": 778, "lr": 8.490673521558651e-05}
{"train_loss": 0.2475685030221939, "global_step": 69325, "epoch": 778, "lr": 8.490632014129155e-05}
{"train_loss": 0.2903496026992798, "global_step": 69326, "epoch": 778, "lr": 8.490590506230386e-05}
{"train_loss": 0.33145907521247864, "global_step": 69327, "epoch": 778, "lr": 8.490548997862351e-05}
{"train_loss": 0.2504752278327942, "global_step": 69328, "epoch": 778, "lr": 8.490507489025052e-05}
{"train_loss": 0.17112348973751068, "global_step": 69329, "epoch": 778, "lr": 8.490465979718499e-05}
{"train_loss": 0.22597073319922673, "global_step": 69330, "epoch": 778, "lr": 8.490424469942695e-05, "val_loss": 3.710623025894165}
{"train_loss": 0.120256207883358, "global_step": 69331, "epoch": 779, "lr": 8.490382959697645e-05}
{"train_loss": 0.2651289701461792, "global_step": 69332, "epoch": 779, "lr": 8.490341448983357e-05}
{"train_loss": 0.18288764357566833, "global_step": 69333, "epoch": 779, "lr": 8.490299937799833e-05}
{"train_loss": 0.2089432030916214, "global_step": 69334, "epoch": 779, "lr": 8.490258426147082e-05}
{"train_loss": 0.2363850325345993, "global_step": 69335, "epoch": 779, "lr": 8.490216914025109e-05}
{"train_loss": 0.20462460815906525, "global_step": 69336, "epoch": 779, "lr": 8.490175401433917e-05}
{"train_loss": 0.18088188767433167, "global_step": 69337, "epoch": 779, "lr": 8.490133888373515e-05}
{"train_loss": 0.20849768817424774, "global_step": 69338, "epoch": 779, "lr": 8.490092374843906e-05}
{"train_loss": 0.15455262362957, "global_step": 69339, "epoch": 779, "lr": 8.490050860845097e-05}
{"train_loss": 0.34963324666023254, "global_step": 69340, "epoch": 779, "lr": 8.490009346377095e-05}
{"train_loss": 0.2702934741973877, "global_step": 69341, "epoch": 779, "lr": 8.489967831439902e-05}
{"train_loss": 0.29332584142684937, "global_step": 69342, "epoch": 779, "lr": 8.489926316033525e-05}
{"train_loss": 0.27180078625679016, "global_step": 69343, "epoch": 779, "lr": 8.48988480015797e-05}
{"train_loss": 0.2073967605829239, "global_step": 69344, "epoch": 779, "lr": 8.489843283813243e-05}
{"train_loss": 0.18499605357646942, "global_step": 69345, "epoch": 779, "lr": 8.489801766999348e-05}
{"train_loss": 0.2243683785200119, "global_step": 69346, "epoch": 779, "lr": 8.489760249716294e-05}
{"train_loss": 0.17354066669940948, "global_step": 69347, "epoch": 779, "lr": 8.489718731964082e-05}
{"train_loss": 0.17098115384578705, "global_step": 69348, "epoch": 779, "lr": 8.489677213742722e-05}
{"train_loss": 0.2195957601070404, "global_step": 69349, "epoch": 779, "lr": 8.489635695052215e-05}
{"train_loss": 0.3331756293773651, "global_step": 69350, "epoch": 779, "lr": 8.489594175892572e-05}
{"train_loss": 0.2582018971443176, "global_step": 69351, "epoch": 779, "lr": 8.489552656263793e-05}
{"train_loss": 0.16568930447101593, "global_step": 69352, "epoch": 779, "lr": 8.489511136165887e-05}
{"train_loss": 0.25328585505485535, "global_step": 69353, "epoch": 779, "lr": 8.48946961559886e-05}
{"train_loss": 0.21626396477222443, "global_step": 69354, "epoch": 779, "lr": 8.489428094562715e-05}
{"train_loss": 0.21745216846466064, "global_step": 69355, "epoch": 779, "lr": 8.489386573057458e-05}
{"train_loss": 0.16526839137077332, "global_step": 69356, "epoch": 779, "lr": 8.489345051083098e-05}
{"train_loss": 0.20084792375564575, "global_step": 69357, "epoch": 779, "lr": 8.489303528639637e-05}
{"train_loss": 0.18794140219688416, "global_step": 69358, "epoch": 779, "lr": 8.489262005727081e-05}
{"train_loss": 0.2641712427139282, "global_step": 69359, "epoch": 779, "lr": 8.489220482345437e-05}
{"train_loss": 0.20278656482696533, "global_step": 69360, "epoch": 779, "lr": 8.48917895849471e-05}
{"train_loss": 0.33230310678482056, "global_step": 69361, "epoch": 779, "lr": 8.489137434174904e-05}
{"train_loss": 0.1916794627904892, "global_step": 69362, "epoch": 779, "lr": 8.489095909386026e-05}
{"train_loss": 0.21404361724853516, "global_step": 69363, "epoch": 779, "lr": 8.489054384128083e-05}
{"train_loss": 0.1586829125881195, "global_step": 69364, "epoch": 779, "lr": 8.489012858401078e-05}
{"train_loss": 0.2239825427532196, "global_step": 69365, "epoch": 779, "lr": 8.48897133220502e-05}
{"train_loss": 0.2174609899520874, "global_step": 69366, "epoch": 779, "lr": 8.488929805539911e-05}
{"train_loss": 0.18052953481674194, "global_step": 69367, "epoch": 779, "lr": 8.488888278405757e-05}
{"train_loss": 0.17605480551719666, "global_step": 69368, "epoch": 779, "lr": 8.488846750802565e-05}
{"train_loss": 0.20902784168720245, "global_step": 69369, "epoch": 779, "lr": 8.48880522273034e-05}
{"train_loss": 0.2492780238389969, "global_step": 69370, "epoch": 779, "lr": 8.488763694189087e-05}
{"train_loss": 0.19464322924613953, "global_step": 69371, "epoch": 779, "lr": 8.488722165178814e-05}
{"train_loss": 0.3412324786186218, "global_step": 69372, "epoch": 779, "lr": 8.488680635699524e-05}
{"train_loss": 0.1893690824508667, "global_step": 69373, "epoch": 779, "lr": 8.488639105751223e-05}
{"train_loss": 0.20313474535942078, "global_step": 69374, "epoch": 779, "lr": 8.488597575333917e-05}
{"train_loss": 0.2460440695285797, "global_step": 69375, "epoch": 779, "lr": 8.488556044447613e-05}
{"train_loss": 0.1823083609342575, "global_step": 69376, "epoch": 779, "lr": 8.488514513092313e-05}
{"train_loss": 0.2135189175605774, "global_step": 69377, "epoch": 779, "lr": 8.488472981268027e-05}
{"train_loss": 0.1414533406496048, "global_step": 69378, "epoch": 779, "lr": 8.488431448974757e-05}
{"train_loss": 0.22937121987342834, "global_step": 69379, "epoch": 779, "lr": 8.488389916212511e-05}
{"train_loss": 0.1729165017604828, "global_step": 69380, "epoch": 779, "lr": 8.48834838298129e-05}
{"train_loss": 0.22605539858341217, "global_step": 69381, "epoch": 779, "lr": 8.488306849281107e-05}
{"train_loss": 0.2846906781196594, "global_step": 69382, "epoch": 779, "lr": 8.488265315111962e-05}
{"train_loss": 0.19170372188091278, "global_step": 69383, "epoch": 779, "lr": 8.488223780473863e-05}
{"train_loss": 0.19229646027088165, "global_step": 69384, "epoch": 779, "lr": 8.488182245366813e-05}
{"train_loss": 0.17297445237636566, "global_step": 69385, "epoch": 779, "lr": 8.488140709790822e-05}
{"train_loss": 0.23911024630069733, "global_step": 69386, "epoch": 779, "lr": 8.488099173745892e-05}
{"train_loss": 0.31762421131134033, "global_step": 69387, "epoch": 779, "lr": 8.48805763723203e-05}
{"train_loss": 0.2553463578224182, "global_step": 69388, "epoch": 779, "lr": 8.48801610024924e-05}
{"train_loss": 0.20872272551059723, "global_step": 69389, "epoch": 779, "lr": 8.48797456279753e-05}
{"train_loss": 0.19532525539398193, "global_step": 69390, "epoch": 779, "lr": 8.487933024876902e-05}
{"train_loss": 0.19923195242881775, "global_step": 69391, "epoch": 779, "lr": 8.487891486487366e-05}
{"train_loss": 0.2680363059043884, "global_step": 69392, "epoch": 779, "lr": 8.487849947628925e-05}
{"train_loss": 0.18583057820796967, "global_step": 69393, "epoch": 779, "lr": 8.487808408301586e-05}
{"train_loss": 0.2531275749206543, "global_step": 69394, "epoch": 779, "lr": 8.487766868505352e-05}
{"train_loss": 0.19880929589271545, "global_step": 69395, "epoch": 779, "lr": 8.48772532824023e-05}
{"train_loss": 0.2592216432094574, "global_step": 69396, "epoch": 779, "lr": 8.487683787506228e-05}
{"train_loss": 0.18497541546821594, "global_step": 69397, "epoch": 779, "lr": 8.487642246303347e-05}
{"train_loss": 0.20431728661060333, "global_step": 69398, "epoch": 779, "lr": 8.487600704631597e-05}
{"train_loss": 0.21559444069862366, "global_step": 69399, "epoch": 779, "lr": 8.48755916249098e-05}
{"train_loss": 0.18088464438915253, "global_step": 69400, "epoch": 779, "lr": 8.487517619881505e-05}
{"train_loss": 0.24008715152740479, "global_step": 69401, "epoch": 779, "lr": 8.487476076803174e-05}
{"train_loss": 0.2409421056509018, "global_step": 69402, "epoch": 779, "lr": 8.487434533255995e-05}
{"train_loss": 0.18453864753246307, "global_step": 69403, "epoch": 779, "lr": 8.487392989239973e-05}
{"train_loss": 0.13681070506572723, "global_step": 69404, "epoch": 779, "lr": 8.487351444755114e-05}
{"train_loss": 0.16892798244953156, "global_step": 69405, "epoch": 779, "lr": 8.487309899801424e-05}
{"train_loss": 0.179952934384346, "global_step": 69406, "epoch": 779, "lr": 8.487268354378906e-05}
{"train_loss": 0.1925438642501831, "global_step": 69407, "epoch": 779, "lr": 8.487226808487568e-05}
{"train_loss": 0.17516601085662842, "global_step": 69408, "epoch": 779, "lr": 8.487185262127416e-05}
{"train_loss": 0.19231675565242767, "global_step": 69409, "epoch": 779, "lr": 8.487143715298451e-05}
{"train_loss": 0.1567312777042389, "global_step": 69410, "epoch": 779, "lr": 8.487102168000685e-05}
{"train_loss": 0.22419437766075134, "global_step": 69411, "epoch": 779, "lr": 8.487060620234119e-05}
{"train_loss": 0.20675507187843323, "global_step": 69412, "epoch": 779, "lr": 8.48701907199876e-05}
{"train_loss": 0.14174087345600128, "global_step": 69413, "epoch": 779, "lr": 8.486977523294615e-05}
{"train_loss": 0.21584148705005646, "global_step": 69414, "epoch": 779, "lr": 8.486935974121688e-05}
{"train_loss": 0.17145830392837524, "global_step": 69415, "epoch": 779, "lr": 8.486894424479985e-05}
{"train_loss": 0.16242679953575134, "global_step": 69416, "epoch": 779, "lr": 8.486852874369511e-05}
{"train_loss": 0.21800671517848969, "global_step": 69417, "epoch": 779, "lr": 8.486811323790274e-05}
{"train_loss": 0.29321029782295227, "global_step": 69418, "epoch": 779, "lr": 8.486769772742275e-05}
{"train_loss": 0.21441925752363847, "global_step": 69419, "epoch": 779, "lr": 8.486728221225525e-05, "val_loss": 3.712629795074463}
{"train_loss": 0.20038165152072906, "global_step": 69420, "epoch": 780, "lr": 8.486686669240024e-05}
{"train_loss": 0.21877029538154602, "global_step": 69421, "epoch": 780, "lr": 8.486645116785783e-05}
{"train_loss": 0.1648065745830536, "global_step": 69422, "epoch": 780, "lr": 8.486603563862803e-05}
{"train_loss": 0.12567032873630524, "global_step": 69423, "epoch": 780, "lr": 8.486562010471093e-05}
{"train_loss": 0.27007001638412476, "global_step": 69424, "epoch": 780, "lr": 8.486520456610658e-05}
{"train_loss": 0.14028911292552948, "global_step": 69425, "epoch": 780, "lr": 8.486478902281501e-05}
{"train_loss": 0.18922913074493408, "global_step": 69426, "epoch": 780, "lr": 8.486437347483631e-05}
{"train_loss": 0.18569707870483398, "global_step": 69427, "epoch": 780, "lr": 8.486395792217052e-05}
{"train_loss": 0.18671101331710815, "global_step": 69428, "epoch": 780, "lr": 8.486354236481768e-05}
{"train_loss": 0.1900126338005066, "global_step": 69429, "epoch": 780, "lr": 8.486312680277786e-05}
{"train_loss": 0.21983404457569122, "global_step": 69430, "epoch": 780, "lr": 8.486271123605113e-05}
{"train_loss": 0.2322934865951538, "global_step": 69431, "epoch": 780, "lr": 8.486229566463752e-05}
{"train_loss": 0.17979981005191803, "global_step": 69432, "epoch": 780, "lr": 8.486188008853712e-05}
{"train_loss": 0.21674518287181854, "global_step": 69433, "epoch": 780, "lr": 8.486146450774996e-05}
{"train_loss": 0.17904752492904663, "global_step": 69434, "epoch": 780, "lr": 8.486104892227609e-05}
{"train_loss": 0.2216392457485199, "global_step": 69435, "epoch": 780, "lr": 8.486063333211558e-05}
{"train_loss": 0.21719887852668762, "global_step": 69436, "epoch": 780, "lr": 8.486021773726849e-05}
{"train_loss": 0.26009246706962585, "global_step": 69437, "epoch": 780, "lr": 8.485980213773485e-05}
{"train_loss": 0.24548718333244324, "global_step": 69438, "epoch": 780, "lr": 8.485938653351474e-05}
{"train_loss": 0.17576190829277039, "global_step": 69439, "epoch": 780, "lr": 8.485897092460824e-05}
{"train_loss": 0.22570312023162842, "global_step": 69440, "epoch": 780, "lr": 8.485855531101536e-05}
{"train_loss": 0.2755733132362366, "global_step": 69441, "epoch": 780, "lr": 8.485813969273616e-05}
{"train_loss": 0.24919523298740387, "global_step": 69442, "epoch": 780, "lr": 8.485772406977071e-05}
{"train_loss": 0.1888469010591507, "global_step": 69443, "epoch": 780, "lr": 8.485730844211907e-05}
{"train_loss": 0.3136449456214905, "global_step": 69444, "epoch": 780, "lr": 8.48568928097813e-05}
{"train_loss": 0.16070972383022308, "global_step": 69445, "epoch": 780, "lr": 8.485647717275742e-05}
{"train_loss": 0.19322334229946136, "global_step": 69446, "epoch": 780, "lr": 8.485606153104753e-05}
{"train_loss": 0.21783198416233063, "global_step": 69447, "epoch": 780, "lr": 8.485564588465167e-05}
{"train_loss": 0.2252030074596405, "global_step": 69448, "epoch": 780, "lr": 8.485523023356988e-05}
{"train_loss": 0.2657375633716583, "global_step": 69449, "epoch": 780, "lr": 8.485481457780225e-05}
{"train_loss": 0.18807464838027954, "global_step": 69450, "epoch": 780, "lr": 8.485439891734879e-05}
{"train_loss": 0.1380450427532196, "global_step": 69451, "epoch": 780, "lr": 8.485398325220959e-05}
{"train_loss": 0.18830005824565887, "global_step": 69452, "epoch": 780, "lr": 8.485356758238469e-05}
{"train_loss": 0.2628357708454132, "global_step": 69453, "epoch": 780, "lr": 8.485315190787416e-05}
{"train_loss": 0.2779270112514496, "global_step": 69454, "epoch": 780, "lr": 8.485273622867806e-05}
{"train_loss": 0.2559417486190796, "global_step": 69455, "epoch": 780, "lr": 8.485232054479641e-05}
{"train_loss": 0.2863641083240509, "global_step": 69456, "epoch": 780, "lr": 8.485190485622931e-05}
{"train_loss": 0.1859227865934372, "global_step": 69457, "epoch": 780, "lr": 8.485148916297678e-05}
{"train_loss": 0.14573703706264496, "global_step": 69458, "epoch": 780, "lr": 8.485107346503891e-05}
{"train_loss": 0.26214683055877686, "global_step": 69459, "epoch": 780, "lr": 8.485065776241573e-05}
{"train_loss": 0.14409904181957245, "global_step": 69460, "epoch": 780, "lr": 8.48502420551073e-05}
{"train_loss": 0.1733771711587906, "global_step": 69461, "epoch": 780, "lr": 8.484982634311369e-05}
{"train_loss": 0.23994386196136475, "global_step": 69462, "epoch": 780, "lr": 8.484941062643494e-05}
{"train_loss": 0.1670922040939331, "global_step": 69463, "epoch": 780, "lr": 8.484899490507111e-05}
{"train_loss": 0.24758560955524445, "global_step": 69464, "epoch": 780, "lr": 8.484857917902226e-05}
{"train_loss": 0.23794785141944885, "global_step": 69465, "epoch": 780, "lr": 8.484816344828844e-05}
{"train_loss": 0.2995043396949768, "global_step": 69466, "epoch": 780, "lr": 8.484774771286972e-05}
{"train_loss": 0.20477168262004852, "global_step": 69467, "epoch": 780, "lr": 8.484733197276613e-05}
{"train_loss": 0.23200006783008575, "global_step": 69468, "epoch": 780, "lr": 8.484691622797774e-05}
{"train_loss": 0.16451825201511383, "global_step": 69469, "epoch": 780, "lr": 8.484650047850462e-05}
{"train_loss": 0.17765173316001892, "global_step": 69470, "epoch": 780, "lr": 8.484608472434681e-05}
{"train_loss": 0.3580266535282135, "global_step": 69471, "epoch": 780, "lr": 8.484566896550436e-05}
{"train_loss": 0.21917016804218292, "global_step": 69472, "epoch": 780, "lr": 8.484525320197733e-05}
{"train_loss": 0.17836984992027283, "global_step": 69473, "epoch": 780, "lr": 8.48448374337658e-05}
{"train_loss": 0.226905956864357, "global_step": 69474, "epoch": 780, "lr": 8.484442166086978e-05}
{"train_loss": 0.19383884966373444, "global_step": 69475, "epoch": 780, "lr": 8.484400588328938e-05}
{"train_loss": 0.223795086145401, "global_step": 69476, "epoch": 780, "lr": 8.48435901010246e-05}
{"train_loss": 0.20344354212284088, "global_step": 69477, "epoch": 780, "lr": 8.484317431407556e-05}
{"train_loss": 0.18805745244026184, "global_step": 69478, "epoch": 780, "lr": 8.484275852244226e-05}
{"train_loss": 0.17378202080726624, "global_step": 69479, "epoch": 780, "lr": 8.484234272612477e-05}
{"train_loss": 0.2877607047557831, "global_step": 69480, "epoch": 780, "lr": 8.484192692512316e-05}
{"train_loss": 0.1954820603132248, "global_step": 69481, "epoch": 780, "lr": 8.484151111943747e-05}
{"train_loss": 0.1915811151266098, "global_step": 69482, "epoch": 780, "lr": 8.484109530906778e-05}
{"train_loss": 0.2702539563179016, "global_step": 69483, "epoch": 780, "lr": 8.484067949401411e-05}
{"train_loss": 0.15606209635734558, "global_step": 69484, "epoch": 780, "lr": 8.484026367427654e-05}
{"train_loss": 0.23423688113689423, "global_step": 69485, "epoch": 780, "lr": 8.483984784985512e-05}
{"train_loss": 0.23672761023044586, "global_step": 69486, "epoch": 780, "lr": 8.483943202074992e-05}
{"train_loss": 0.260591059923172, "global_step": 69487, "epoch": 780, "lr": 8.483901618696097e-05}
{"train_loss": 0.26274174451828003, "global_step": 69488, "epoch": 780, "lr": 8.483860034848835e-05}
{"train_loss": 0.1845712512731552, "global_step": 69489, "epoch": 780, "lr": 8.483818450533209e-05}
{"train_loss": 0.25172197818756104, "global_step": 69490, "epoch": 780, "lr": 8.483776865749227e-05}
{"train_loss": 0.1459479033946991, "global_step": 69491, "epoch": 780, "lr": 8.483735280496894e-05}
{"train_loss": 0.24334031343460083, "global_step": 69492, "epoch": 780, "lr": 8.483693694776215e-05}
{"train_loss": 0.1893996298313141, "global_step": 69493, "epoch": 780, "lr": 8.483652108587195e-05}
{"train_loss": 0.22412711381912231, "global_step": 69494, "epoch": 780, "lr": 8.483610521929842e-05}
{"train_loss": 0.17256984114646912, "global_step": 69495, "epoch": 780, "lr": 8.483568934804158e-05}
{"train_loss": 0.20829153060913086, "global_step": 69496, "epoch": 780, "lr": 8.483527347210151e-05}
{"train_loss": 0.2873685359954834, "global_step": 69497, "epoch": 780, "lr": 8.48348575914783e-05}
{"train_loss": 0.1469404250383377, "global_step": 69498, "epoch": 780, "lr": 8.483444170617192e-05}
{"train_loss": 0.2211066484451294, "global_step": 69499, "epoch": 780, "lr": 8.483402581618249e-05}
{"train_loss": 0.24660055339336395, "global_step": 69500, "epoch": 780, "lr": 8.483360992151005e-05}
{"train_loss": 0.2507152855396271, "global_step": 69501, "epoch": 780, "lr": 8.483319402215466e-05}
{"train_loss": 0.2161802351474762, "global_step": 69502, "epoch": 780, "lr": 8.483277811811636e-05}
{"train_loss": 0.15462249517440796, "global_step": 69503, "epoch": 780, "lr": 8.483236220939524e-05}
{"train_loss": 0.15148183703422546, "global_step": 69504, "epoch": 780, "lr": 8.483194629599132e-05}
{"train_loss": 0.2519087493419647, "global_step": 69505, "epoch": 780, "lr": 8.483153037790467e-05}
{"train_loss": 0.22993475198745728, "global_step": 69506, "epoch": 780, "lr": 8.483111445513534e-05}
{"train_loss": 0.24764181673526764, "global_step": 69507, "epoch": 780, "lr": 8.483069852768338e-05}
{"train_loss": 0.21387788653373718, "global_step": 69508, "epoch": 780, "lr": 8.483028259554888e-05, "val_loss": 3.7465741634368896, "train_action_mse_error": 28.238868713378906}
{"train_loss": 0.25660043954849243, "global_step": 69509, "epoch": 781, "lr": 8.482986665873188e-05}
{"train_loss": 0.22802329063415527, "global_step": 69510, "epoch": 781, "lr": 8.48294507172324e-05}
{"train_loss": 0.184358611702919, "global_step": 69511, "epoch": 781, "lr": 8.482903477105054e-05}
{"train_loss": 0.21386738121509552, "global_step": 69512, "epoch": 781, "lr": 8.482861882018633e-05}
{"train_loss": 0.21218989789485931, "global_step": 69513, "epoch": 781, "lr": 8.482820286463986e-05}
{"train_loss": 0.28883275389671326, "global_step": 69514, "epoch": 781, "lr": 8.482778690441115e-05}
{"train_loss": 0.20217356085777283, "global_step": 69515, "epoch": 781, "lr": 8.482737093950026e-05}
{"train_loss": 0.25320711731910706, "global_step": 69516, "epoch": 781, "lr": 8.482695496990727e-05}
{"train_loss": 0.15038065612316132, "global_step": 69517, "epoch": 781, "lr": 8.482653899563222e-05}
{"train_loss": 0.23635566234588623, "global_step": 69518, "epoch": 781, "lr": 8.482612301667514e-05}
{"train_loss": 0.2508697211742401, "global_step": 69519, "epoch": 781, "lr": 8.482570703303614e-05}
{"train_loss": 0.28180548548698425, "global_step": 69520, "epoch": 781, "lr": 8.482529104471524e-05}
{"train_loss": 0.287352591753006, "global_step": 69521, "epoch": 781, "lr": 8.48248750517125e-05}
{"train_loss": 0.24134580790996552, "global_step": 69522, "epoch": 781, "lr": 8.482445905402798e-05}
{"train_loss": 0.2206682711839676, "global_step": 69523, "epoch": 781, "lr": 8.482404305166175e-05}
{"train_loss": 0.2570194900035858, "global_step": 69524, "epoch": 781, "lr": 8.482362704461383e-05}
{"train_loss": 0.32643139362335205, "global_step": 69525, "epoch": 781, "lr": 8.482321103288431e-05}
{"train_loss": 0.189711332321167, "global_step": 69526, "epoch": 781, "lr": 8.482279501647324e-05}
{"train_loss": 0.21616409718990326, "global_step": 69527, "epoch": 781, "lr": 8.482237899538066e-05}
{"train_loss": 0.21166136860847473, "global_step": 69528, "epoch": 781, "lr": 8.482196296960664e-05}
{"train_loss": 0.17185437679290771, "global_step": 69529, "epoch": 781, "lr": 8.482154693915123e-05}
{"train_loss": 0.20830532908439636, "global_step": 69530, "epoch": 781, "lr": 8.48211309040145e-05}
{"train_loss": 0.293763130903244, "global_step": 69531, "epoch": 781, "lr": 8.482071486419647e-05}
{"train_loss": 0.25308454036712646, "global_step": 69532, "epoch": 781, "lr": 8.482029881969722e-05}
{"train_loss": 0.23791912198066711, "global_step": 69533, "epoch": 781, "lr": 8.481988277051682e-05}
{"train_loss": 0.21154391765594482, "global_step": 69534, "epoch": 781, "lr": 8.48194667166553e-05}
{"train_loss": 0.1794138103723526, "global_step": 69535, "epoch": 781, "lr": 8.481905065811273e-05}
{"train_loss": 0.18370480835437775, "global_step": 69536, "epoch": 781, "lr": 8.481863459488917e-05}
{"train_loss": 0.2619503438472748, "global_step": 69537, "epoch": 781, "lr": 8.481821852698466e-05}
{"train_loss": 0.1907375603914261, "global_step": 69538, "epoch": 781, "lr": 8.481780245439927e-05}
{"train_loss": 0.2849500775337219, "global_step": 69539, "epoch": 781, "lr": 8.481738637713306e-05}
{"train_loss": 0.1944548487663269, "global_step": 69540, "epoch": 781, "lr": 8.481697029518607e-05}
{"train_loss": 0.159676194190979, "global_step": 69541, "epoch": 781, "lr": 8.481655420855837e-05}
{"train_loss": 0.2048376351594925, "global_step": 69542, "epoch": 781, "lr": 8.481613811724999e-05}
{"train_loss": 0.1873987764120102, "global_step": 69543, "epoch": 781, "lr": 8.4815722021261e-05}
{"train_loss": 0.24899661540985107, "global_step": 69544, "epoch": 781, "lr": 8.481530592059148e-05}
{"train_loss": 0.2674146592617035, "global_step": 69545, "epoch": 781, "lr": 8.481488981524146e-05}
{"train_loss": 0.19574517011642456, "global_step": 69546, "epoch": 781, "lr": 8.481447370521101e-05}
{"train_loss": 0.2061898559331894, "global_step": 69547, "epoch": 781, "lr": 8.481405759050016e-05}
{"train_loss": 0.2867673337459564, "global_step": 69548, "epoch": 781, "lr": 8.4813641471109e-05}
{"train_loss": 0.2739027738571167, "global_step": 69549, "epoch": 781, "lr": 8.481322534703757e-05}
{"train_loss": 0.2944166362285614, "global_step": 69550, "epoch": 781, "lr": 8.481280921828591e-05}
{"train_loss": 0.31456631422042847, "global_step": 69551, "epoch": 781, "lr": 8.481239308485411e-05}
{"train_loss": 0.19319579005241394, "global_step": 69552, "epoch": 781, "lr": 8.481197694674221e-05}
{"train_loss": 0.23889486491680145, "global_step": 69553, "epoch": 781, "lr": 8.481156080395025e-05}
{"train_loss": 0.31292933225631714, "global_step": 69554, "epoch": 781, "lr": 8.481114465647831e-05}
{"train_loss": 0.16589780151844025, "global_step": 69555, "epoch": 781, "lr": 8.481072850432643e-05}
{"train_loss": 0.22552117705345154, "global_step": 69556, "epoch": 781, "lr": 8.481031234749467e-05}
{"train_loss": 0.31038352847099304, "global_step": 69557, "epoch": 781, "lr": 8.48098961859831e-05}
{"train_loss": 0.27174487709999084, "global_step": 69558, "epoch": 781, "lr": 8.480948001979175e-05}
{"train_loss": 0.2569970488548279, "global_step": 69559, "epoch": 781, "lr": 8.480906384892069e-05}
{"train_loss": 0.20072022080421448, "global_step": 69560, "epoch": 781, "lr": 8.480864767336998e-05}
{"train_loss": 0.18417435884475708, "global_step": 69561, "epoch": 781, "lr": 8.480823149313967e-05}
{"train_loss": 0.2590748071670532, "global_step": 69562, "epoch": 781, "lr": 8.480781530822983e-05}
{"train_loss": 0.26566120982170105, "global_step": 69563, "epoch": 781, "lr": 8.480739911864048e-05}
{"train_loss": 0.17020699381828308, "global_step": 69564, "epoch": 781, "lr": 8.480698292437171e-05}
{"train_loss": 0.18485459685325623, "global_step": 69565, "epoch": 781, "lr": 8.480656672542358e-05}
{"train_loss": 0.19661931693553925, "global_step": 69566, "epoch": 781, "lr": 8.480615052179612e-05}
{"train_loss": 0.22952152788639069, "global_step": 69567, "epoch": 781, "lr": 8.48057343134894e-05}
{"train_loss": 0.20618005096912384, "global_step": 69568, "epoch": 781, "lr": 8.480531810050346e-05}
{"train_loss": 0.16115586459636688, "global_step": 69569, "epoch": 781, "lr": 8.480490188283839e-05}
{"train_loss": 0.22439202666282654, "global_step": 69570, "epoch": 781, "lr": 8.48044856604942e-05}
{"train_loss": 0.19978845119476318, "global_step": 69571, "epoch": 781, "lr": 8.480406943347099e-05}
{"train_loss": 0.17197942733764648, "global_step": 69572, "epoch": 781, "lr": 8.480365320176879e-05}
{"train_loss": 0.17658579349517822, "global_step": 69573, "epoch": 781, "lr": 8.480323696538767e-05}
{"train_loss": 0.30279460549354553, "global_step": 69574, "epoch": 781, "lr": 8.480282072432767e-05}
{"train_loss": 0.19477862119674683, "global_step": 69575, "epoch": 781, "lr": 8.480240447858887e-05}
{"train_loss": 0.2485916167497635, "global_step": 69576, "epoch": 781, "lr": 8.48019882281713e-05}
{"train_loss": 0.16164614260196686, "global_step": 69577, "epoch": 781, "lr": 8.480157197307501e-05}
{"train_loss": 0.38802340626716614, "global_step": 69578, "epoch": 781, "lr": 8.48011557133001e-05}
{"train_loss": 0.195220947265625, "global_step": 69579, "epoch": 781, "lr": 8.480073944884658e-05}
{"train_loss": 0.20575425028800964, "global_step": 69580, "epoch": 781, "lr": 8.480032317971453e-05}
{"train_loss": 0.18447290360927582, "global_step": 69581, "epoch": 781, "lr": 8.4799906905904e-05}
{"train_loss": 0.13938315212726593, "global_step": 69582, "epoch": 781, "lr": 8.479949062741505e-05}
{"train_loss": 0.1734728366136551, "global_step": 69583, "epoch": 781, "lr": 8.479907434424774e-05}
{"train_loss": 0.2091313749551773, "global_step": 69584, "epoch": 781, "lr": 8.479865805640212e-05}
{"train_loss": 0.18534062802791595, "global_step": 69585, "epoch": 781, "lr": 8.479824176387823e-05}
{"train_loss": 0.27041152119636536, "global_step": 69586, "epoch": 781, "lr": 8.479782546667615e-05}
{"train_loss": 0.1835731714963913, "global_step": 69587, "epoch": 781, "lr": 8.479740916479593e-05}
{"train_loss": 0.25925126671791077, "global_step": 69588, "epoch": 781, "lr": 8.479699285823761e-05}
{"train_loss": 0.24959057569503784, "global_step": 69589, "epoch": 781, "lr": 8.479657654700125e-05}
{"train_loss": 0.23889616131782532, "global_step": 69590, "epoch": 781, "lr": 8.479616023108694e-05}
{"train_loss": 0.22004935145378113, "global_step": 69591, "epoch": 781, "lr": 8.479574391049469e-05}
{"train_loss": 0.20968769490718842, "global_step": 69592, "epoch": 781, "lr": 8.479532758522459e-05}
{"train_loss": 0.2425520420074463, "global_step": 69593, "epoch": 781, "lr": 8.479491125527668e-05}
{"train_loss": 0.1952524036169052, "global_step": 69594, "epoch": 781, "lr": 8.479449492065101e-05}
{"train_loss": 0.17728538811206818, "global_step": 69595, "epoch": 781, "lr": 8.479407858134766e-05}
{"train_loss": 0.21896058320999146, "global_step": 69596, "epoch": 781, "lr": 8.479366223736667e-05}
{"train_loss": 0.22577684113148894, "global_step": 69597, "epoch": 781, "lr": 8.479324588870808e-05, "val_loss": 3.7333595752716064}
{"train_loss": 0.3153012692928314, "global_step": 69598, "epoch": 782, "lr": 8.479282953537196e-05}
{"train_loss": 0.2545282542705536, "global_step": 69599, "epoch": 782, "lr": 8.479241317735837e-05}
{"train_loss": 0.21575386822223663, "global_step": 69600, "epoch": 782, "lr": 8.479199681466738e-05}
{"train_loss": 0.1706738919019699, "global_step": 69601, "epoch": 782, "lr": 8.479158044729903e-05}
{"train_loss": 0.27544137835502625, "global_step": 69602, "epoch": 782, "lr": 8.479116407525337e-05}
{"train_loss": 0.22402004897594452, "global_step": 69603, "epoch": 782, "lr": 8.479074769853044e-05}
{"train_loss": 0.2374207079410553, "global_step": 69604, "epoch": 782, "lr": 8.479033131713034e-05}
{"train_loss": 0.20574043691158295, "global_step": 69605, "epoch": 782, "lr": 8.478991493105311e-05}
{"train_loss": 0.20991627871990204, "global_step": 69606, "epoch": 782, "lr": 8.478949854029878e-05}
{"train_loss": 0.22169405221939087, "global_step": 69607, "epoch": 782, "lr": 8.478908214486745e-05}
{"train_loss": 0.1772555708885193, "global_step": 69608, "epoch": 782, "lr": 8.478866574475914e-05}
{"train_loss": 0.17226170003414154, "global_step": 69609, "epoch": 782, "lr": 8.478824933997391e-05}
{"train_loss": 0.17837630212306976, "global_step": 69610, "epoch": 782, "lr": 8.478783293051183e-05}
{"train_loss": 0.27132347226142883, "global_step": 69611, "epoch": 782, "lr": 8.478741651637295e-05}
{"train_loss": 0.23765698075294495, "global_step": 69612, "epoch": 782, "lr": 8.478700009755734e-05}
{"train_loss": 0.24290695786476135, "global_step": 69613, "epoch": 782, "lr": 8.478658367406502e-05}
{"train_loss": 0.21746928989887238, "global_step": 69614, "epoch": 782, "lr": 8.478616724589607e-05}
{"train_loss": 0.18232442438602448, "global_step": 69615, "epoch": 782, "lr": 8.478575081305056e-05}
{"train_loss": 0.253140926361084, "global_step": 69616, "epoch": 782, "lr": 8.478533437552851e-05}
{"train_loss": 0.2321701943874359, "global_step": 69617, "epoch": 782, "lr": 8.478491793333001e-05}
{"train_loss": 0.18243663012981415, "global_step": 69618, "epoch": 782, "lr": 8.47845014864551e-05}
{"train_loss": 0.31641075015068054, "global_step": 69619, "epoch": 782, "lr": 8.478408503490382e-05}
{"train_loss": 0.21032270789146423, "global_step": 69620, "epoch": 782, "lr": 8.478366857867627e-05}
{"train_loss": 0.25379088521003723, "global_step": 69621, "epoch": 782, "lr": 8.478325211777248e-05}
{"train_loss": 0.2773035764694214, "global_step": 69622, "epoch": 782, "lr": 8.478283565219248e-05}
{"train_loss": 0.24903500080108643, "global_step": 69623, "epoch": 782, "lr": 8.478241918193637e-05}
{"train_loss": 0.22252234816551208, "global_step": 69624, "epoch": 782, "lr": 8.478200270700418e-05}
{"train_loss": 0.15198954939842224, "global_step": 69625, "epoch": 782, "lr": 8.478158622739599e-05}
{"train_loss": 0.24920859932899475, "global_step": 69626, "epoch": 782, "lr": 8.47811697431118e-05}
{"train_loss": 0.27552124857902527, "global_step": 69627, "epoch": 782, "lr": 8.478075325415176e-05}
{"train_loss": 0.1899135708808899, "global_step": 69628, "epoch": 782, "lr": 8.478033676051583e-05}
{"train_loss": 0.18611638247966766, "global_step": 69629, "epoch": 782, "lr": 8.477992026220412e-05}
{"train_loss": 0.253481388092041, "global_step": 69630, "epoch": 782, "lr": 8.477950375921668e-05}
{"train_loss": 0.2199087142944336, "global_step": 69631, "epoch": 782, "lr": 8.477908725155355e-05}
{"train_loss": 0.26583805680274963, "global_step": 69632, "epoch": 782, "lr": 8.47786707392148e-05}
{"train_loss": 0.16837415099143982, "global_step": 69633, "epoch": 782, "lr": 8.47782542222005e-05}
{"train_loss": 0.1999790072441101, "global_step": 69634, "epoch": 782, "lr": 8.477783770051065e-05}
{"train_loss": 0.19071514904499054, "global_step": 69635, "epoch": 782, "lr": 8.477742117414537e-05}
{"train_loss": 0.17815783619880676, "global_step": 69636, "epoch": 782, "lr": 8.477700464310467e-05}
{"train_loss": 0.13744744658470154, "global_step": 69637, "epoch": 782, "lr": 8.477658810738864e-05}
{"train_loss": 0.13966932892799377, "global_step": 69638, "epoch": 782, "lr": 8.477617156699733e-05}
{"train_loss": 0.1893848478794098, "global_step": 69639, "epoch": 782, "lr": 8.477575502193079e-05}
{"train_loss": 0.19172628223896027, "global_step": 69640, "epoch": 782, "lr": 8.477533847218905e-05}
{"train_loss": 0.2887098789215088, "global_step": 69641, "epoch": 782, "lr": 8.47749219177722e-05}
{"train_loss": 0.18045327067375183, "global_step": 69642, "epoch": 782, "lr": 8.477450535868029e-05}
{"train_loss": 0.18932488560676575, "global_step": 69643, "epoch": 782, "lr": 8.477408879491337e-05}
{"train_loss": 0.18664701282978058, "global_step": 69644, "epoch": 782, "lr": 8.47736722264715e-05}
{"train_loss": 0.20247705280780792, "global_step": 69645, "epoch": 782, "lr": 8.477325565335472e-05}
{"train_loss": 0.24921810626983643, "global_step": 69646, "epoch": 782, "lr": 8.477283907556312e-05}
{"train_loss": 0.24350522458553314, "global_step": 69647, "epoch": 782, "lr": 8.477242249309671e-05}
{"train_loss": 0.2317713052034378, "global_step": 69648, "epoch": 782, "lr": 8.47720059059556e-05}
{"train_loss": 0.24213847517967224, "global_step": 69649, "epoch": 782, "lr": 8.47715893141398e-05}
{"train_loss": 0.22917670011520386, "global_step": 69650, "epoch": 782, "lr": 8.477117271764938e-05}
{"train_loss": 0.22489048540592194, "global_step": 69651, "epoch": 782, "lr": 8.477075611648442e-05}
{"train_loss": 0.2645445466041565, "global_step": 69652, "epoch": 782, "lr": 8.477033951064493e-05}
{"train_loss": 0.20682810246944427, "global_step": 69653, "epoch": 782, "lr": 8.4769922900131e-05}
{"train_loss": 0.21875929832458496, "global_step": 69654, "epoch": 782, "lr": 8.47695062849427e-05}
{"train_loss": 0.21710534393787384, "global_step": 69655, "epoch": 782, "lr": 8.476908966508004e-05}
{"train_loss": 0.21335509419441223, "global_step": 69656, "epoch": 782, "lr": 8.47686730405431e-05}
{"train_loss": 0.1723695993423462, "global_step": 69657, "epoch": 782, "lr": 8.476825641133194e-05}
{"train_loss": 0.14986105263233185, "global_step": 69658, "epoch": 782, "lr": 8.476783977744661e-05}
{"train_loss": 0.2133936882019043, "global_step": 69659, "epoch": 782, "lr": 8.476742313888717e-05}
{"train_loss": 0.21305780112743378, "global_step": 69660, "epoch": 782, "lr": 8.476700649565368e-05}
{"train_loss": 0.20394255220890045, "global_step": 69661, "epoch": 782, "lr": 8.476658984774618e-05}
{"train_loss": 0.1716548353433609, "global_step": 69662, "epoch": 782, "lr": 8.476617319516475e-05}
{"train_loss": 0.2021857500076294, "global_step": 69663, "epoch": 782, "lr": 8.476575653790942e-05}
{"train_loss": 0.21640314161777496, "global_step": 69664, "epoch": 782, "lr": 8.476533987598026e-05}
{"train_loss": 0.4002677798271179, "global_step": 69665, "epoch": 782, "lr": 8.476492320937732e-05}
{"train_loss": 0.2319110631942749, "global_step": 69666, "epoch": 782, "lr": 8.476450653810067e-05}
{"train_loss": 0.17602556943893433, "global_step": 69667, "epoch": 782, "lr": 8.476408986215034e-05}
{"train_loss": 0.17401006817817688, "global_step": 69668, "epoch": 782, "lr": 8.476367318152641e-05}
{"train_loss": 0.26455509662628174, "global_step": 69669, "epoch": 782, "lr": 8.476325649622892e-05}
{"train_loss": 0.16618642210960388, "global_step": 69670, "epoch": 782, "lr": 8.476283980625794e-05}
{"train_loss": 0.18881483376026154, "global_step": 69671, "epoch": 782, "lr": 8.476242311161353e-05}
{"train_loss": 0.14007170498371124, "global_step": 69672, "epoch": 782, "lr": 8.476200641229573e-05}
{"train_loss": 0.2784309685230255, "global_step": 69673, "epoch": 782, "lr": 8.476158970830459e-05}
{"train_loss": 0.2715800106525421, "global_step": 69674, "epoch": 782, "lr": 8.476117299964019e-05}
{"train_loss": 0.20869871973991394, "global_step": 69675, "epoch": 782, "lr": 8.476075628630258e-05}
{"train_loss": 0.26938724517822266, "global_step": 69676, "epoch": 782, "lr": 8.476033956829179e-05}
{"train_loss": 0.23536579310894012, "global_step": 69677, "epoch": 782, "lr": 8.475992284560792e-05}
{"train_loss": 0.21904443204402924, "global_step": 69678, "epoch": 782, "lr": 8.475950611825098e-05}
{"train_loss": 0.21380065381526947, "global_step": 69679, "epoch": 782, "lr": 8.475908938622106e-05}
{"train_loss": 0.2583944499492645, "global_step": 69680, "epoch": 782, "lr": 8.47586726495182e-05}
{"train_loss": 0.2785918414592743, "global_step": 69681, "epoch": 782, "lr": 8.475825590814247e-05}
{"train_loss": 0.18610960245132446, "global_step": 69682, "epoch": 782, "lr": 8.47578391620939e-05}
{"train_loss": 0.22356168925762177, "global_step": 69683, "epoch": 782, "lr": 8.475742241137258e-05}
{"train_loss": 0.2200489193201065, "global_step": 69684, "epoch": 782, "lr": 8.475700565597854e-05}
{"train_loss": 0.12744681537151337, "global_step": 69685, "epoch": 782, "lr": 8.475658889591185e-05}
{"train_loss": 0.2206205687161242, "global_step": 69686, "epoch": 782, "lr": 8.475617213117256e-05, "val_loss": 3.7039127349853516}
{"train_loss": 0.21953023970127106, "global_step": 69687, "epoch": 783, "lr": 8.47557553617607e-05}
{"train_loss": 0.19983112812042236, "global_step": 69688, "epoch": 783, "lr": 8.475533858767637e-05}
{"train_loss": 0.23167774081230164, "global_step": 69689, "epoch": 783, "lr": 8.475492180891961e-05}
{"train_loss": 0.30248644948005676, "global_step": 69690, "epoch": 783, "lr": 8.475450502549047e-05}
{"train_loss": 0.24900120496749878, "global_step": 69691, "epoch": 783, "lr": 8.475408823738902e-05}
{"train_loss": 0.22327351570129395, "global_step": 69692, "epoch": 783, "lr": 8.47536714446153e-05}
{"train_loss": 0.17164340615272522, "global_step": 69693, "epoch": 783, "lr": 8.475325464716937e-05}
{"train_loss": 0.2620623707771301, "global_step": 69694, "epoch": 783, "lr": 8.47528378450513e-05}
{"train_loss": 0.18769308924674988, "global_step": 69695, "epoch": 783, "lr": 8.475242103826112e-05}
{"train_loss": 0.22498631477355957, "global_step": 69696, "epoch": 783, "lr": 8.475200422679891e-05}
{"train_loss": 0.2903749644756317, "global_step": 69697, "epoch": 783, "lr": 8.475158741066471e-05}
{"train_loss": 0.23619458079338074, "global_step": 69698, "epoch": 783, "lr": 8.47511705898586e-05}
{"train_loss": 0.2725297808647156, "global_step": 69699, "epoch": 783, "lr": 8.47507537643806e-05}
{"train_loss": 0.28650161623954773, "global_step": 69700, "epoch": 783, "lr": 8.475033693423077e-05}
{"train_loss": 0.2947743833065033, "global_step": 69701, "epoch": 783, "lr": 8.47499200994092e-05}
{"train_loss": 0.2828376889228821, "global_step": 69702, "epoch": 783, "lr": 8.474950325991596e-05}
{"train_loss": 0.2788394093513489, "global_step": 69703, "epoch": 783, "lr": 8.474908641575103e-05}
{"train_loss": 0.17497919499874115, "global_step": 69704, "epoch": 783, "lr": 8.474866956691452e-05}
{"train_loss": 0.23235878348350525, "global_step": 69705, "epoch": 783, "lr": 8.474825271340647e-05}
{"train_loss": 0.2801821827888489, "global_step": 69706, "epoch": 783, "lr": 8.474783585522695e-05}
{"train_loss": 0.2648339867591858, "global_step": 69707, "epoch": 783, "lr": 8.4747418992376e-05}
{"train_loss": 0.3385888636112213, "global_step": 69708, "epoch": 783, "lr": 8.474700212485367e-05}
{"train_loss": 0.21838782727718353, "global_step": 69709, "epoch": 783, "lr": 8.474658525266005e-05}
{"train_loss": 0.18205507099628448, "global_step": 69710, "epoch": 783, "lr": 8.474616837579516e-05}
{"train_loss": 0.20430001616477966, "global_step": 69711, "epoch": 783, "lr": 8.474575149425909e-05}
{"train_loss": 0.1749388575553894, "global_step": 69712, "epoch": 783, "lr": 8.474533460805187e-05}
{"train_loss": 0.20513159036636353, "global_step": 69713, "epoch": 783, "lr": 8.474491771717356e-05}
{"train_loss": 0.26774948835372925, "global_step": 69714, "epoch": 783, "lr": 8.474450082162422e-05}
{"train_loss": 0.21001990139484406, "global_step": 69715, "epoch": 783, "lr": 8.474408392140391e-05}
{"train_loss": 0.2301849126815796, "global_step": 69716, "epoch": 783, "lr": 8.474366701651266e-05}
{"train_loss": 0.201083242893219, "global_step": 69717, "epoch": 783, "lr": 8.474325010695057e-05}
{"train_loss": 0.26214829087257385, "global_step": 69718, "epoch": 783, "lr": 8.474283319271767e-05}
{"train_loss": 0.18084615468978882, "global_step": 69719, "epoch": 783, "lr": 8.474241627381401e-05}
{"train_loss": 0.2609191834926605, "global_step": 69720, "epoch": 783, "lr": 8.474199935023968e-05}
{"train_loss": 0.21984101831912994, "global_step": 69721, "epoch": 783, "lr": 8.474158242199469e-05}
{"train_loss": 0.15777133405208588, "global_step": 69722, "epoch": 783, "lr": 8.474116548907913e-05}
{"train_loss": 0.20087957382202148, "global_step": 69723, "epoch": 783, "lr": 8.474074855149303e-05}
{"train_loss": 0.24020345509052277, "global_step": 69724, "epoch": 783, "lr": 8.474033160923647e-05}
{"train_loss": 0.29312726855278015, "global_step": 69725, "epoch": 783, "lr": 8.473991466230948e-05}
{"train_loss": 0.22246740758419037, "global_step": 69726, "epoch": 783, "lr": 8.473949771071215e-05}
{"train_loss": 0.1842682808637619, "global_step": 69727, "epoch": 783, "lr": 8.47390807544445e-05}
{"train_loss": 0.3772874176502228, "global_step": 69728, "epoch": 783, "lr": 8.473866379350663e-05}
{"train_loss": 0.22294394671916962, "global_step": 69729, "epoch": 783, "lr": 8.473824682789855e-05}
{"train_loss": 0.1658315509557724, "global_step": 69730, "epoch": 783, "lr": 8.473782985762034e-05}
{"train_loss": 0.18283624947071075, "global_step": 69731, "epoch": 783, "lr": 8.473741288267205e-05}
{"train_loss": 0.1921709179878235, "global_step": 69732, "epoch": 783, "lr": 8.473699590305375e-05}
{"train_loss": 0.16438302397727966, "global_step": 69733, "epoch": 783, "lr": 8.473657891876547e-05}
{"train_loss": 0.1873340755701065, "global_step": 69734, "epoch": 783, "lr": 8.473616192980729e-05}
{"train_loss": 0.22826391458511353, "global_step": 69735, "epoch": 783, "lr": 8.473574493617924e-05}
{"train_loss": 0.39451006054878235, "global_step": 69736, "epoch": 783, "lr": 8.473532793788142e-05}
{"train_loss": 0.2288888692855835, "global_step": 69737, "epoch": 783, "lr": 8.473491093491384e-05}
{"train_loss": 0.2744672894477844, "global_step": 69738, "epoch": 783, "lr": 8.473449392727657e-05}
{"train_loss": 0.2392306625843048, "global_step": 69739, "epoch": 783, "lr": 8.473407691496968e-05}
{"train_loss": 0.3242531418800354, "global_step": 69740, "epoch": 783, "lr": 8.473365989799323e-05}
{"train_loss": 0.259052574634552, "global_step": 69741, "epoch": 783, "lr": 8.473324287634724e-05}
{"train_loss": 0.2212260514497757, "global_step": 69742, "epoch": 783, "lr": 8.47328258500318e-05}
{"train_loss": 0.24267515540122986, "global_step": 69743, "epoch": 783, "lr": 8.473240881904694e-05}
{"train_loss": 0.1959424614906311, "global_step": 69744, "epoch": 783, "lr": 8.473199178339276e-05}
{"train_loss": 0.2164941430091858, "global_step": 69745, "epoch": 783, "lr": 8.473157474306926e-05}
{"train_loss": 0.2570759057998657, "global_step": 69746, "epoch": 783, "lr": 8.473115769807653e-05}
{"train_loss": 0.2136632800102234, "global_step": 69747, "epoch": 783, "lr": 8.473074064841463e-05}
{"train_loss": 0.172671839594841, "global_step": 69748, "epoch": 783, "lr": 8.47303235940836e-05}
{"train_loss": 0.20833204686641693, "global_step": 69749, "epoch": 783, "lr": 8.472990653508349e-05}
{"train_loss": 0.17870396375656128, "global_step": 69750, "epoch": 783, "lr": 8.472948947141438e-05}
{"train_loss": 0.15128515660762787, "global_step": 69751, "epoch": 783, "lr": 8.47290724030763e-05}
{"train_loss": 0.22085466980934143, "global_step": 69752, "epoch": 783, "lr": 8.472865533006934e-05}
{"train_loss": 0.24108940362930298, "global_step": 69753, "epoch": 783, "lr": 8.472823825239353e-05}
{"train_loss": 0.25486254692077637, "global_step": 69754, "epoch": 783, "lr": 8.472782117004891e-05}
{"train_loss": 0.2523677349090576, "global_step": 69755, "epoch": 783, "lr": 8.472740408303559e-05}
{"train_loss": 0.3023252785205841, "global_step": 69756, "epoch": 783, "lr": 8.472698699135357e-05}
{"train_loss": 0.20995168387889862, "global_step": 69757, "epoch": 783, "lr": 8.472656989500294e-05}
{"train_loss": 0.18566776812076569, "global_step": 69758, "epoch": 783, "lr": 8.472615279398375e-05}
{"train_loss": 0.23248259723186493, "global_step": 69759, "epoch": 783, "lr": 8.472573568829605e-05}
{"train_loss": 0.21053099632263184, "global_step": 69760, "epoch": 783, "lr": 8.472531857793988e-05}
{"train_loss": 0.1480456441640854, "global_step": 69761, "epoch": 783, "lr": 8.472490146291533e-05}
{"train_loss": 0.1751869022846222, "global_step": 69762, "epoch": 783, "lr": 8.472448434322244e-05}
{"train_loss": 0.2317127138376236, "global_step": 69763, "epoch": 783, "lr": 8.472406721886125e-05}
{"train_loss": 0.21245147287845612, "global_step": 69764, "epoch": 783, "lr": 8.472365008983186e-05}
{"train_loss": 0.1947532594203949, "global_step": 69765, "epoch": 783, "lr": 8.472323295613429e-05}
{"train_loss": 0.1996786743402481, "global_step": 69766, "epoch": 783, "lr": 8.47228158177686e-05}
{"train_loss": 0.19379779696464539, "global_step": 69767, "epoch": 783, "lr": 8.472239867473485e-05}
{"train_loss": 0.16169440746307373, "global_step": 69768, "epoch": 783, "lr": 8.47219815270331e-05}
{"train_loss": 0.1939529925584793, "global_step": 69769, "epoch": 783, "lr": 8.47215643746634e-05}
{"train_loss": 0.2015618085861206, "global_step": 69770, "epoch": 783, "lr": 8.47211472176258e-05}
{"train_loss": 0.27389344573020935, "global_step": 69771, "epoch": 783, "lr": 8.472073005592038e-05}
{"train_loss": 0.23457397520542145, "global_step": 69772, "epoch": 783, "lr": 8.472031288954719e-05}
{"train_loss": 0.22105172276496887, "global_step": 69773, "epoch": 783, "lr": 8.471989571850624e-05}
{"train_loss": 0.20490935444831848, "global_step": 69774, "epoch": 783, "lr": 8.471947854279765e-05}
{"train_loss": 0.2283956738335363, "global_step": 69775, "epoch": 783, "lr": 8.471906136242145e-05, "val_loss": 3.558067560195923}
{"train_loss": 0.31028756499290466, "global_step": 69776, "epoch": 784, "lr": 8.47186441773777e-05}
{"train_loss": 0.21535898745059967, "global_step": 69777, "epoch": 784, "lr": 8.471822698766643e-05}
{"train_loss": 0.20041188597679138, "global_step": 69778, "epoch": 784, "lr": 8.471780979328773e-05}
{"train_loss": 0.17547759413719177, "global_step": 69779, "epoch": 784, "lr": 8.471739259424163e-05}
{"train_loss": 0.31215178966522217, "global_step": 69780, "epoch": 784, "lr": 8.471697539052821e-05}
{"train_loss": 0.1618947833776474, "global_step": 69781, "epoch": 784, "lr": 8.471655818214752e-05}
{"train_loss": 0.2751542329788208, "global_step": 69782, "epoch": 784, "lr": 8.471614096909961e-05}
{"train_loss": 0.21801388263702393, "global_step": 69783, "epoch": 784, "lr": 8.471572375138455e-05}
{"train_loss": 0.2559741735458374, "global_step": 69784, "epoch": 784, "lr": 8.471530652900237e-05}
{"train_loss": 0.23478800058364868, "global_step": 69785, "epoch": 784, "lr": 8.471488930195314e-05}
{"train_loss": 0.2131875455379486, "global_step": 69786, "epoch": 784, "lr": 8.471447207023692e-05}
{"train_loss": 0.14938220381736755, "global_step": 69787, "epoch": 784, "lr": 8.471405483385377e-05}
{"train_loss": 0.18970976769924164, "global_step": 69788, "epoch": 784, "lr": 8.471363759280374e-05}
{"train_loss": 0.3131343126296997, "global_step": 69789, "epoch": 784, "lr": 8.471322034708686e-05}
{"train_loss": 0.2705387771129608, "global_step": 69790, "epoch": 784, "lr": 8.471280309670322e-05}
{"train_loss": 0.15856796503067017, "global_step": 69791, "epoch": 784, "lr": 8.471238584165287e-05}
{"train_loss": 0.22268004715442657, "global_step": 69792, "epoch": 784, "lr": 8.471196858193586e-05}
{"train_loss": 0.17076492309570312, "global_step": 69793, "epoch": 784, "lr": 8.471155131755226e-05}
{"train_loss": 0.20747201144695282, "global_step": 69794, "epoch": 784, "lr": 8.471113404850211e-05}
{"train_loss": 0.22714100778102875, "global_step": 69795, "epoch": 784, "lr": 8.471071677478546e-05}
{"train_loss": 0.171921506524086, "global_step": 69796, "epoch": 784, "lr": 8.471029949640239e-05}
{"train_loss": 0.2293250560760498, "global_step": 69797, "epoch": 784, "lr": 8.470988221335294e-05}
{"train_loss": 0.26935961842536926, "global_step": 69798, "epoch": 784, "lr": 8.470946492563717e-05}
{"train_loss": 0.2946454882621765, "global_step": 69799, "epoch": 784, "lr": 8.470904763325513e-05}
{"train_loss": 0.22035983204841614, "global_step": 69800, "epoch": 784, "lr": 8.470863033620689e-05}
{"train_loss": 0.2820054590702057, "global_step": 69801, "epoch": 784, "lr": 8.470821303449249e-05}
{"train_loss": 0.18845166265964508, "global_step": 69802, "epoch": 784, "lr": 8.4707795728112e-05}
{"train_loss": 0.18704883754253387, "global_step": 69803, "epoch": 784, "lr": 8.470737841706547e-05}
{"train_loss": 0.22322362661361694, "global_step": 69804, "epoch": 784, "lr": 8.470696110135295e-05}
{"train_loss": 0.17524513602256775, "global_step": 69805, "epoch": 784, "lr": 8.47065437809745e-05}
{"train_loss": 0.21795713901519775, "global_step": 69806, "epoch": 784, "lr": 8.470612645593018e-05}
{"train_loss": 0.16567489504814148, "global_step": 69807, "epoch": 784, "lr": 8.470570912622006e-05}
{"train_loss": 0.2197747528553009, "global_step": 69808, "epoch": 784, "lr": 8.470529179184417e-05}
{"train_loss": 0.172842338681221, "global_step": 69809, "epoch": 784, "lr": 8.470487445280258e-05}
{"train_loss": 0.24124523997306824, "global_step": 69810, "epoch": 784, "lr": 8.470445710909532e-05}
{"train_loss": 0.2824677526950836, "global_step": 69811, "epoch": 784, "lr": 8.47040397607225e-05}
{"train_loss": 0.20974871516227722, "global_step": 69812, "epoch": 784, "lr": 8.470362240768413e-05}
{"train_loss": 0.13194309175014496, "global_step": 69813, "epoch": 784, "lr": 8.470320504998028e-05}
{"train_loss": 0.1345289945602417, "global_step": 69814, "epoch": 784, "lr": 8.4702787687611e-05}
{"train_loss": 0.20397402346134186, "global_step": 69815, "epoch": 784, "lr": 8.470237032057636e-05}
{"train_loss": 0.22341127693653107, "global_step": 69816, "epoch": 784, "lr": 8.470195294887641e-05}
{"train_loss": 0.21006236970424652, "global_step": 69817, "epoch": 784, "lr": 8.470153557251121e-05}
{"train_loss": 0.19640369713306427, "global_step": 69818, "epoch": 784, "lr": 8.470111819148081e-05}
{"train_loss": 0.1538456380367279, "global_step": 69819, "epoch": 784, "lr": 8.470070080578527e-05}
{"train_loss": 0.27118492126464844, "global_step": 69820, "epoch": 784, "lr": 8.470028341542465e-05}
{"train_loss": 0.13458310067653656, "global_step": 69821, "epoch": 784, "lr": 8.469986602039897e-05}
{"train_loss": 0.1376926749944687, "global_step": 69822, "epoch": 784, "lr": 8.469944862070833e-05}
{"train_loss": 0.19567681849002838, "global_step": 69823, "epoch": 784, "lr": 8.469903121635278e-05}
{"train_loss": 0.20373910665512085, "global_step": 69824, "epoch": 784, "lr": 8.469861380733238e-05}
{"train_loss": 0.19247864186763763, "global_step": 69825, "epoch": 784, "lr": 8.469819639364716e-05}
{"train_loss": 0.23317113518714905, "global_step": 69826, "epoch": 784, "lr": 8.469777897529718e-05}
{"train_loss": 0.23804540932178497, "global_step": 69827, "epoch": 784, "lr": 8.469736155228252e-05}
{"train_loss": 0.24742159247398376, "global_step": 69828, "epoch": 784, "lr": 8.469694412460322e-05}
{"train_loss": 0.2746754586696625, "global_step": 69829, "epoch": 784, "lr": 8.469652669225934e-05}
{"train_loss": 0.19719967246055603, "global_step": 69830, "epoch": 784, "lr": 8.469610925525093e-05}
{"train_loss": 0.21453341841697693, "global_step": 69831, "epoch": 784, "lr": 8.469569181357806e-05}
{"train_loss": 0.21400359272956848, "global_step": 69832, "epoch": 784, "lr": 8.469527436724076e-05}
{"train_loss": 0.21396538615226746, "global_step": 69833, "epoch": 784, "lr": 8.469485691623912e-05}
{"train_loss": 0.17196190357208252, "global_step": 69834, "epoch": 784, "lr": 8.469443946057317e-05}
{"train_loss": 0.2329172044992447, "global_step": 69835, "epoch": 784, "lr": 8.469402200024299e-05}
{"train_loss": 0.18700367212295532, "global_step": 69836, "epoch": 784, "lr": 8.46936045352486e-05}
{"train_loss": 0.2679777145385742, "global_step": 69837, "epoch": 784, "lr": 8.469318706559008e-05}
{"train_loss": 0.2161804735660553, "global_step": 69838, "epoch": 784, "lr": 8.469276959126751e-05}
{"train_loss": 0.2228935956954956, "global_step": 69839, "epoch": 784, "lr": 8.469235211228091e-05}
{"train_loss": 0.3036032021045685, "global_step": 69840, "epoch": 784, "lr": 8.469193462863031e-05}
{"train_loss": 0.18095813691616058, "global_step": 69841, "epoch": 784, "lr": 8.469151714031585e-05}
{"train_loss": 0.17389315366744995, "global_step": 69842, "epoch": 784, "lr": 8.469109964733751e-05}
{"train_loss": 0.2591000199317932, "global_step": 69843, "epoch": 784, "lr": 8.469068214969539e-05}
{"train_loss": 0.23259691894054413, "global_step": 69844, "epoch": 784, "lr": 8.469026464738951e-05}
{"train_loss": 0.22190125286579132, "global_step": 69845, "epoch": 784, "lr": 8.468984714041998e-05}
{"train_loss": 0.2163023203611374, "global_step": 69846, "epoch": 784, "lr": 8.468942962878681e-05}
{"train_loss": 0.16606971621513367, "global_step": 69847, "epoch": 784, "lr": 8.468901211249005e-05}
{"train_loss": 0.138467475771904, "global_step": 69848, "epoch": 784, "lr": 8.46885945915298e-05}
{"train_loss": 0.22912286221981049, "global_step": 69849, "epoch": 784, "lr": 8.468817706590607e-05}
{"train_loss": 0.14510773122310638, "global_step": 69850, "epoch": 784, "lr": 8.468775953561895e-05}
{"train_loss": 0.2025972604751587, "global_step": 69851, "epoch": 784, "lr": 8.468734200066848e-05}
{"train_loss": 0.21586297452449799, "global_step": 69852, "epoch": 784, "lr": 8.46869244610547e-05}
{"train_loss": 0.12352554500102997, "global_step": 69853, "epoch": 784, "lr": 8.468650691677771e-05}
{"train_loss": 0.20611202716827393, "global_step": 69854, "epoch": 784, "lr": 8.468608936783755e-05}
{"train_loss": 0.18818560242652893, "global_step": 69855, "epoch": 784, "lr": 8.468567181423425e-05}
{"train_loss": 0.24654442071914673, "global_step": 69856, "epoch": 784, "lr": 8.468525425596789e-05}
{"train_loss": 0.22937551140785217, "global_step": 69857, "epoch": 784, "lr": 8.468483669303853e-05}
{"train_loss": 0.17816010117530823, "global_step": 69858, "epoch": 784, "lr": 8.46844191254462e-05}
{"train_loss": 0.20403295755386353, "global_step": 69859, "epoch": 784, "lr": 8.468400155319098e-05}
{"train_loss": 0.16280928254127502, "global_step": 69860, "epoch": 784, "lr": 8.468358397627291e-05}
{"train_loss": 0.22254669666290283, "global_step": 69861, "epoch": 784, "lr": 8.468316639469207e-05}
{"train_loss": 0.24462451040744781, "global_step": 69862, "epoch": 784, "lr": 8.468274880844849e-05}
{"train_loss": 0.20341378450393677, "global_step": 69863, "epoch": 784, "lr": 8.468233121754224e-05}
{"train_loss": 0.21205372321471738, "global_step": 69864, "epoch": 784, "lr": 8.468191362197337e-05, "val_loss": 3.7369790077209473}
{"train_loss": 0.21346597373485565, "global_step": 69865, "epoch": 785, "lr": 8.468149602174196e-05}
{"train_loss": 0.23166505992412567, "global_step": 69866, "epoch": 785, "lr": 8.468107841684803e-05}
{"train_loss": 0.20172929763793945, "global_step": 69867, "epoch": 785, "lr": 8.468066080729164e-05}
{"train_loss": 0.2552925944328308, "global_step": 69868, "epoch": 785, "lr": 8.468024319307287e-05}
{"train_loss": 0.2441674917936325, "global_step": 69869, "epoch": 785, "lr": 8.467982557419177e-05}
{"train_loss": 0.16795627772808075, "global_step": 69870, "epoch": 785, "lr": 8.467940795064837e-05}
{"train_loss": 0.12835223972797394, "global_step": 69871, "epoch": 785, "lr": 8.467899032244276e-05}
{"train_loss": 0.18348656594753265, "global_step": 69872, "epoch": 785, "lr": 8.467857268957498e-05}
{"train_loss": 0.26805201172828674, "global_step": 69873, "epoch": 785, "lr": 8.467815505204509e-05}
{"train_loss": 0.292609304189682, "global_step": 69874, "epoch": 785, "lr": 8.467773740985314e-05}
{"train_loss": 0.29489845037460327, "global_step": 69875, "epoch": 785, "lr": 8.467731976299919e-05}
{"train_loss": 0.16017432510852814, "global_step": 69876, "epoch": 785, "lr": 8.467690211148332e-05}
{"train_loss": 0.20345118641853333, "global_step": 69877, "epoch": 785, "lr": 8.467648445530554e-05}
{"train_loss": 0.2537747323513031, "global_step": 69878, "epoch": 785, "lr": 8.467606679446594e-05}
{"train_loss": 0.20839743316173553, "global_step": 69879, "epoch": 785, "lr": 8.467564912896455e-05}
{"train_loss": 0.15097910165786743, "global_step": 69880, "epoch": 785, "lr": 8.467523145880145e-05}
{"train_loss": 0.21758507192134857, "global_step": 69881, "epoch": 785, "lr": 8.46748137839767e-05}
{"train_loss": 0.19808818399906158, "global_step": 69882, "epoch": 785, "lr": 8.467439610449032e-05}
{"train_loss": 0.23531073331832886, "global_step": 69883, "epoch": 785, "lr": 8.467397842034242e-05}
{"train_loss": 0.22712735831737518, "global_step": 69884, "epoch": 785, "lr": 8.4673560731533e-05}
{"train_loss": 0.20761612057685852, "global_step": 69885, "epoch": 785, "lr": 8.467314303806217e-05}
{"train_loss": 0.10750687122344971, "global_step": 69886, "epoch": 785, "lr": 8.467272533992994e-05}
{"train_loss": 0.21712899208068848, "global_step": 69887, "epoch": 785, "lr": 8.467230763713638e-05}
{"train_loss": 0.28633949160575867, "global_step": 69888, "epoch": 785, "lr": 8.467188992968154e-05}
{"train_loss": 0.19753557443618774, "global_step": 69889, "epoch": 785, "lr": 8.467147221756551e-05}
{"train_loss": 0.26999154686927795, "global_step": 69890, "epoch": 785, "lr": 8.467105450078831e-05}
{"train_loss": 0.3074045777320862, "global_step": 69891, "epoch": 785, "lr": 8.467063677935003e-05}
{"train_loss": 0.2211228609085083, "global_step": 69892, "epoch": 785, "lr": 8.467021905325069e-05}
{"train_loss": 0.2223109006881714, "global_step": 69893, "epoch": 785, "lr": 8.466980132249036e-05}
{"train_loss": 0.203678160905838, "global_step": 69894, "epoch": 785, "lr": 8.46693835870691e-05}
{"train_loss": 0.1905737966299057, "global_step": 69895, "epoch": 785, "lr": 8.466896584698697e-05}
{"train_loss": 0.13679417967796326, "global_step": 69896, "epoch": 785, "lr": 8.466854810224401e-05}
{"train_loss": 0.17203998565673828, "global_step": 69897, "epoch": 785, "lr": 8.46681303528403e-05}
{"train_loss": 0.2819826304912567, "global_step": 69898, "epoch": 785, "lr": 8.466771259877587e-05}
{"train_loss": 0.20631654560565948, "global_step": 69899, "epoch": 785, "lr": 8.466729484005079e-05}
{"train_loss": 0.2737194895744324, "global_step": 69900, "epoch": 785, "lr": 8.466687707666511e-05}
{"train_loss": 0.28148525953292847, "global_step": 69901, "epoch": 785, "lr": 8.466645930861891e-05}
{"train_loss": 0.22334028780460358, "global_step": 69902, "epoch": 785, "lr": 8.466604153591221e-05}
{"train_loss": 0.2143601030111313, "global_step": 69903, "epoch": 785, "lr": 8.46656237585451e-05}
{"train_loss": 0.23832891881465912, "global_step": 69904, "epoch": 785, "lr": 8.46652059765176e-05}
{"train_loss": 0.23992106318473816, "global_step": 69905, "epoch": 785, "lr": 8.46647881898298e-05}
{"train_loss": 0.22139433026313782, "global_step": 69906, "epoch": 785, "lr": 8.466437039848172e-05}
{"train_loss": 0.21419525146484375, "global_step": 69907, "epoch": 785, "lr": 8.466395260247345e-05}
{"train_loss": 0.4033733308315277, "global_step": 69908, "epoch": 785, "lr": 8.466353480180504e-05}
{"train_loss": 0.2347947657108307, "global_step": 69909, "epoch": 785, "lr": 8.466311699647654e-05}
{"train_loss": 0.27360522747039795, "global_step": 69910, "epoch": 785, "lr": 8.4662699186488e-05}
{"train_loss": 0.16505183279514313, "global_step": 69911, "epoch": 785, "lr": 8.46622813718395e-05}
{"train_loss": 0.20749199390411377, "global_step": 69912, "epoch": 785, "lr": 8.466186355253105e-05}
{"train_loss": 0.19125749170780182, "global_step": 69913, "epoch": 785, "lr": 8.466144572856276e-05}
{"train_loss": 0.2085488736629486, "global_step": 69914, "epoch": 785, "lr": 8.466102789993464e-05}
{"train_loss": 0.2412959337234497, "global_step": 69915, "epoch": 785, "lr": 8.466061006664678e-05}
{"train_loss": 0.2769112288951874, "global_step": 69916, "epoch": 785, "lr": 8.466019222869923e-05}
{"train_loss": 0.23080551624298096, "global_step": 69917, "epoch": 785, "lr": 8.465977438609202e-05}
{"train_loss": 0.19748929142951965, "global_step": 69918, "epoch": 785, "lr": 8.465935653882524e-05}
{"train_loss": 0.2345755398273468, "global_step": 69919, "epoch": 785, "lr": 8.465893868689894e-05}
{"train_loss": 0.2460530400276184, "global_step": 69920, "epoch": 785, "lr": 8.465852083031315e-05}
{"train_loss": 0.14462488889694214, "global_step": 69921, "epoch": 785, "lr": 8.465810296906795e-05}
{"train_loss": 0.202749565243721, "global_step": 69922, "epoch": 785, "lr": 8.465768510316338e-05}
{"train_loss": 0.21558547019958496, "global_step": 69923, "epoch": 785, "lr": 8.465726723259952e-05}
{"train_loss": 0.2351377308368683, "global_step": 69924, "epoch": 785, "lr": 8.465684935737641e-05}
{"train_loss": 0.2740740478038788, "global_step": 69925, "epoch": 785, "lr": 8.465643147749411e-05}
{"train_loss": 0.34104689955711365, "global_step": 69926, "epoch": 785, "lr": 8.465601359295267e-05}
{"train_loss": 0.33176204562187195, "global_step": 69927, "epoch": 785, "lr": 8.465559570375216e-05}
{"train_loss": 0.24865680932998657, "global_step": 69928, "epoch": 785, "lr": 8.465517780989263e-05}
{"train_loss": 0.24247309565544128, "global_step": 69929, "epoch": 785, "lr": 8.465475991137412e-05}
{"train_loss": 0.21865162253379822, "global_step": 69930, "epoch": 785, "lr": 8.465434200819671e-05}
{"train_loss": 0.27099794149398804, "global_step": 69931, "epoch": 785, "lr": 8.465392410036046e-05}
{"train_loss": 0.20836323499679565, "global_step": 69932, "epoch": 785, "lr": 8.465350618786538e-05}
{"train_loss": 0.26726725697517395, "global_step": 69933, "epoch": 785, "lr": 8.465308827071159e-05}
{"train_loss": 0.2768717408180237, "global_step": 69934, "epoch": 785, "lr": 8.46526703488991e-05}
{"train_loss": 0.2505412995815277, "global_step": 69935, "epoch": 785, "lr": 8.465225242242799e-05}
{"train_loss": 0.2008105218410492, "global_step": 69936, "epoch": 785, "lr": 8.465183449129827e-05}
{"train_loss": 0.14342744648456573, "global_step": 69937, "epoch": 785, "lr": 8.465141655551008e-05}
{"train_loss": 0.2611033618450165, "global_step": 69938, "epoch": 785, "lr": 8.46509986150634e-05}
{"train_loss": 0.30292317271232605, "global_step": 69939, "epoch": 785, "lr": 8.465058066995832e-05}
{"train_loss": 0.25598540902137756, "global_step": 69940, "epoch": 785, "lr": 8.46501627201949e-05}
{"train_loss": 0.15139923989772797, "global_step": 69941, "epoch": 785, "lr": 8.464974476577319e-05}
{"train_loss": 0.20771750807762146, "global_step": 69942, "epoch": 785, "lr": 8.464932680669325e-05}
{"train_loss": 0.24376070499420166, "global_step": 69943, "epoch": 785, "lr": 8.464890884295511e-05}
{"train_loss": 0.23287084698677063, "global_step": 69944, "epoch": 785, "lr": 8.464849087455885e-05}
{"train_loss": 0.21301500499248505, "global_step": 69945, "epoch": 785, "lr": 8.464807290150453e-05}
{"train_loss": 0.1541374772787094, "global_step": 69946, "epoch": 785, "lr": 8.464765492379219e-05}
{"train_loss": 0.2551429569721222, "global_step": 69947, "epoch": 785, "lr": 8.464723694142192e-05}
{"train_loss": 0.19385644793510437, "global_step": 69948, "epoch": 785, "lr": 8.464681895439372e-05}
{"train_loss": 0.16433294117450714, "global_step": 69949, "epoch": 785, "lr": 8.46464009627077e-05}
{"train_loss": 0.23421670496463776, "global_step": 69950, "epoch": 785, "lr": 8.464598296636388e-05}
{"train_loss": 0.26137155294418335, "global_step": 69951, "epoch": 785, "lr": 8.464556496536233e-05}
{"train_loss": 0.24833384156227112, "global_step": 69952, "epoch": 785, "lr": 8.464514695970311e-05}
{"train_loss": 0.2279011124975226, "global_step": 69953, "epoch": 785, "lr": 8.464472894938628e-05, "val_loss": 3.594531536102295, "train_action_mse_error": 32.88595199584961}
{"train_loss": 0.23494181036949158, "global_step": 69954, "epoch": 786, "lr": 8.464431093441188e-05}
{"train_loss": 0.3271964192390442, "global_step": 69955, "epoch": 786, "lr": 8.464389291478e-05}
{"train_loss": 0.2155575305223465, "global_step": 69956, "epoch": 786, "lr": 8.464347489049063e-05}
{"train_loss": 0.2821333110332489, "global_step": 69957, "epoch": 786, "lr": 8.464305686154388e-05}
{"train_loss": 0.2918139398097992, "global_step": 69958, "epoch": 786, "lr": 8.46426388279398e-05}
{"train_loss": 0.21288861334323883, "global_step": 69959, "epoch": 786, "lr": 8.464222078967844e-05}
{"train_loss": 0.22919149696826935, "global_step": 69960, "epoch": 786, "lr": 8.464180274675986e-05}
{"train_loss": 0.21451453864574432, "global_step": 69961, "epoch": 786, "lr": 8.46413846991841e-05}
{"train_loss": 0.23303468525409698, "global_step": 69962, "epoch": 786, "lr": 8.464096664695122e-05}
{"train_loss": 0.20848986506462097, "global_step": 69963, "epoch": 786, "lr": 8.464054859006129e-05}
{"train_loss": 0.16914090514183044, "global_step": 69964, "epoch": 786, "lr": 8.464013052851438e-05}
{"train_loss": 0.18174603581428528, "global_step": 69965, "epoch": 786, "lr": 8.46397124623105e-05}
{"train_loss": 0.23049139976501465, "global_step": 69966, "epoch": 786, "lr": 8.463929439144975e-05}
{"train_loss": 0.2660978436470032, "global_step": 69967, "epoch": 786, "lr": 8.463887631593217e-05}
{"train_loss": 0.17043763399124146, "global_step": 69968, "epoch": 786, "lr": 8.463845823575782e-05}
{"train_loss": 0.2020578682422638, "global_step": 69969, "epoch": 786, "lr": 8.463804015092674e-05}
{"train_loss": 0.27143606543540955, "global_step": 69970, "epoch": 786, "lr": 8.4637622061439e-05}
{"train_loss": 0.2302536964416504, "global_step": 69971, "epoch": 786, "lr": 8.463720396729467e-05}
{"train_loss": 0.1941339671611786, "global_step": 69972, "epoch": 786, "lr": 8.463678586849377e-05}
{"train_loss": 0.27264416217803955, "global_step": 69973, "epoch": 786, "lr": 8.463636776503637e-05}
{"train_loss": 0.21179243922233582, "global_step": 69974, "epoch": 786, "lr": 8.463594965692254e-05}
{"train_loss": 0.31799763441085815, "global_step": 69975, "epoch": 786, "lr": 8.463553154415234e-05}
{"train_loss": 0.2525394856929779, "global_step": 69976, "epoch": 786, "lr": 8.463511342672581e-05}
{"train_loss": 0.1893490105867386, "global_step": 69977, "epoch": 786, "lr": 8.463469530464301e-05}
{"train_loss": 0.15884876251220703, "global_step": 69978, "epoch": 786, "lr": 8.463427717790399e-05}
{"train_loss": 0.2508509159088135, "global_step": 69979, "epoch": 786, "lr": 8.463385904650881e-05}
{"train_loss": 0.22518014907836914, "global_step": 69980, "epoch": 786, "lr": 8.463344091045755e-05}
{"train_loss": 0.22202444076538086, "global_step": 69981, "epoch": 786, "lr": 8.463302276975022e-05}
{"train_loss": 0.24848094582557678, "global_step": 69982, "epoch": 786, "lr": 8.463260462438692e-05}
{"train_loss": 0.18423064053058624, "global_step": 69983, "epoch": 786, "lr": 8.463218647436768e-05}
{"train_loss": 0.17573226988315582, "global_step": 69984, "epoch": 786, "lr": 8.463176831969257e-05}
{"train_loss": 0.17157872021198273, "global_step": 69985, "epoch": 786, "lr": 8.463135016036164e-05}
{"train_loss": 0.2425960898399353, "global_step": 69986, "epoch": 786, "lr": 8.463093199637495e-05}
{"train_loss": 0.2551601231098175, "global_step": 69987, "epoch": 786, "lr": 8.463051382773255e-05}
{"train_loss": 0.30548095703125, "global_step": 69988, "epoch": 786, "lr": 8.463009565443452e-05}
{"train_loss": 0.32184579968452454, "global_step": 69989, "epoch": 786, "lr": 8.462967747648087e-05}
{"train_loss": 0.1591660976409912, "global_step": 69990, "epoch": 786, "lr": 8.462925929387169e-05}
{"train_loss": 0.19671288132667542, "global_step": 69991, "epoch": 786, "lr": 8.462884110660703e-05}
{"train_loss": 0.22162877023220062, "global_step": 69992, "epoch": 786, "lr": 8.462842291468693e-05}
{"train_loss": 0.1835208684206009, "global_step": 69993, "epoch": 786, "lr": 8.462800471811147e-05}
{"train_loss": 0.1894228458404541, "global_step": 69994, "epoch": 786, "lr": 8.462758651688069e-05}
{"train_loss": 0.22289986908435822, "global_step": 69995, "epoch": 786, "lr": 8.462716831099466e-05}
{"train_loss": 0.23285599052906036, "global_step": 69996, "epoch": 786, "lr": 8.462675010045344e-05}
{"train_loss": 0.2479533702135086, "global_step": 69997, "epoch": 786, "lr": 8.462633188525706e-05}
{"train_loss": 0.3242957592010498, "global_step": 69998, "epoch": 786, "lr": 8.46259136654056e-05}
{"train_loss": 0.2107347548007965, "global_step": 69999, "epoch": 786, "lr": 8.46254954408991e-05}
{"train_loss": 0.2404457926750183, "global_step": 70000, "epoch": 786, "lr": 8.462507721173764e-05}
{"train_loss": 0.2002391219139099, "global_step": 70001, "epoch": 786, "lr": 8.462465897792124e-05}
{"train_loss": 0.23414498567581177, "global_step": 70002, "epoch": 786, "lr": 8.462424073944999e-05}
{"train_loss": 0.2808183431625366, "global_step": 70003, "epoch": 786, "lr": 8.462382249632393e-05}
{"train_loss": 0.2527437210083008, "global_step": 70004, "epoch": 786, "lr": 8.462340424854312e-05}
{"train_loss": 0.2830113470554352, "global_step": 70005, "epoch": 786, "lr": 8.462298599610761e-05}
{"train_loss": 0.18955263495445251, "global_step": 70006, "epoch": 786, "lr": 8.462256773901748e-05}
{"train_loss": 0.23146462440490723, "global_step": 70007, "epoch": 786, "lr": 8.462214947727275e-05}
{"train_loss": 0.27329009771347046, "global_step": 70008, "epoch": 786, "lr": 8.46217312108735e-05}
{"train_loss": 0.21489214897155762, "global_step": 70009, "epoch": 786, "lr": 8.462131293981977e-05}
{"train_loss": 0.08778324723243713, "global_step": 70010, "epoch": 786, "lr": 8.462089466411163e-05}
{"train_loss": 0.27388110756874084, "global_step": 70011, "epoch": 786, "lr": 8.462047638374915e-05}
{"train_loss": 0.2743105888366699, "global_step": 70012, "epoch": 786, "lr": 8.462005809873234e-05}
{"train_loss": 0.23583130538463593, "global_step": 70013, "epoch": 786, "lr": 8.461963980906132e-05}
{"train_loss": 0.34542110562324524, "global_step": 70014, "epoch": 786, "lr": 8.461922151473609e-05}
{"train_loss": 0.3003552556037903, "global_step": 70015, "epoch": 786, "lr": 8.461880321575673e-05}
{"train_loss": 0.2882140874862671, "global_step": 70016, "epoch": 786, "lr": 8.461838491212329e-05}
{"train_loss": 0.20918630063533783, "global_step": 70017, "epoch": 786, "lr": 8.461796660383582e-05}
{"train_loss": 0.1535143107175827, "global_step": 70018, "epoch": 786, "lr": 8.461754829089441e-05}
{"train_loss": 0.19032268226146698, "global_step": 70019, "epoch": 786, "lr": 8.461712997329908e-05}
{"train_loss": 0.2551300823688507, "global_step": 70020, "epoch": 786, "lr": 8.46167116510499e-05}
{"train_loss": 0.31872257590293884, "global_step": 70021, "epoch": 786, "lr": 8.461629332414693e-05}
{"train_loss": 0.2028605043888092, "global_step": 70022, "epoch": 786, "lr": 8.461587499259021e-05}
{"train_loss": 0.22330094873905182, "global_step": 70023, "epoch": 786, "lr": 8.461545665637983e-05}
{"train_loss": 0.24896441400051117, "global_step": 70024, "epoch": 786, "lr": 8.46150383155158e-05}
{"train_loss": 0.21253623068332672, "global_step": 70025, "epoch": 786, "lr": 8.46146199699982e-05}
{"train_loss": 0.2598036825656891, "global_step": 70026, "epoch": 786, "lr": 8.46142016198271e-05}
{"train_loss": 0.24064680933952332, "global_step": 70027, "epoch": 786, "lr": 8.461378326500254e-05}
{"train_loss": 0.2164125293493271, "global_step": 70028, "epoch": 786, "lr": 8.461336490552458e-05}
{"train_loss": 0.2204281985759735, "global_step": 70029, "epoch": 786, "lr": 8.461294654139326e-05}
{"train_loss": 0.19818565249443054, "global_step": 70030, "epoch": 786, "lr": 8.461252817260867e-05}
{"train_loss": 0.26395827531814575, "global_step": 70031, "epoch": 786, "lr": 8.461210979917085e-05}
{"train_loss": 0.25004321336746216, "global_step": 70032, "epoch": 786, "lr": 8.461169142107983e-05}
{"train_loss": 0.32154011726379395, "global_step": 70033, "epoch": 786, "lr": 8.461127303833571e-05}
{"train_loss": 0.15942725539207458, "global_step": 70034, "epoch": 786, "lr": 8.461085465093853e-05}
{"train_loss": 0.18716496229171753, "global_step": 70035, "epoch": 786, "lr": 8.461043625888831e-05}
{"train_loss": 0.31009232997894287, "global_step": 70036, "epoch": 786, "lr": 8.461001786218517e-05}
{"train_loss": 0.15487472712993622, "global_step": 70037, "epoch": 786, "lr": 8.460959946082912e-05}
{"train_loss": 0.2128005027770996, "global_step": 70038, "epoch": 786, "lr": 8.460918105482024e-05}
{"train_loss": 0.2756216526031494, "global_step": 70039, "epoch": 786, "lr": 8.460876264415857e-05}
{"train_loss": 0.23127448558807373, "global_step": 70040, "epoch": 786, "lr": 8.460834422884419e-05}
{"train_loss": 0.17004434764385223, "global_step": 70041, "epoch": 786, "lr": 8.460792580887714e-05}
{"train_loss": 0.23252377215396153, "global_step": 70042, "epoch": 786, "lr": 8.460750738425745e-05, "val_loss": 3.6636288166046143}
{"train_loss": 0.253398597240448, "global_step": 70043, "epoch": 787, "lr": 8.460708895498521e-05}
{"train_loss": 0.16980315744876862, "global_step": 70044, "epoch": 787, "lr": 8.460667052106048e-05}
{"train_loss": 0.14246323704719543, "global_step": 70045, "epoch": 787, "lr": 8.46062520824833e-05}
{"train_loss": 0.21191567182540894, "global_step": 70046, "epoch": 787, "lr": 8.460583363925375e-05}
{"train_loss": 0.18362057209014893, "global_step": 70047, "epoch": 787, "lr": 8.460541519137184e-05}
{"train_loss": 0.23115235567092896, "global_step": 70048, "epoch": 787, "lr": 8.460499673883766e-05}
{"train_loss": 0.15034407377243042, "global_step": 70049, "epoch": 787, "lr": 8.460457828165126e-05}
{"train_loss": 0.2787746489048004, "global_step": 70050, "epoch": 787, "lr": 8.46041598198127e-05}
{"train_loss": 0.1894157975912094, "global_step": 70051, "epoch": 787, "lr": 8.460374135332202e-05}
{"train_loss": 0.22567647695541382, "global_step": 70052, "epoch": 787, "lr": 8.46033228821793e-05}
{"train_loss": 0.18730874359607697, "global_step": 70053, "epoch": 787, "lr": 8.460290440638458e-05}
{"train_loss": 0.15261079370975494, "global_step": 70054, "epoch": 787, "lr": 8.460248592593793e-05}
{"train_loss": 0.20144568383693695, "global_step": 70055, "epoch": 787, "lr": 8.46020674408394e-05}
{"train_loss": 0.24001581966876984, "global_step": 70056, "epoch": 787, "lr": 8.460164895108905e-05}
{"train_loss": 0.20200732350349426, "global_step": 70057, "epoch": 787, "lr": 8.460123045668691e-05}
{"train_loss": 0.2603968381881714, "global_step": 70058, "epoch": 787, "lr": 8.460081195763306e-05}
{"train_loss": 0.19347979128360748, "global_step": 70059, "epoch": 787, "lr": 8.460039345392756e-05}
{"train_loss": 0.19408155977725983, "global_step": 70060, "epoch": 787, "lr": 8.459997494557046e-05}
{"train_loss": 0.2100507766008377, "global_step": 70061, "epoch": 787, "lr": 8.45995564325618e-05}
{"train_loss": 0.21977391839027405, "global_step": 70062, "epoch": 787, "lr": 8.459913791490166e-05}
{"train_loss": 0.16571764647960663, "global_step": 70063, "epoch": 787, "lr": 8.45987193925901e-05}
{"train_loss": 0.19242781400680542, "global_step": 70064, "epoch": 787, "lr": 8.459830086562715e-05}
{"train_loss": 0.2524016499519348, "global_step": 70065, "epoch": 787, "lr": 8.459788233401289e-05}
{"train_loss": 0.24265408515930176, "global_step": 70066, "epoch": 787, "lr": 8.459746379774736e-05}
{"train_loss": 0.24449017643928528, "global_step": 70067, "epoch": 787, "lr": 8.459704525683061e-05}
{"train_loss": 0.2375241369009018, "global_step": 70068, "epoch": 787, "lr": 8.459662671126272e-05}
{"train_loss": 0.16946831345558167, "global_step": 70069, "epoch": 787, "lr": 8.459620816104374e-05}
{"train_loss": 0.14933238923549652, "global_step": 70070, "epoch": 787, "lr": 8.459578960617371e-05}
{"train_loss": 0.25766149163246155, "global_step": 70071, "epoch": 787, "lr": 8.459537104665271e-05}
{"train_loss": 0.2888874113559723, "global_step": 70072, "epoch": 787, "lr": 8.459495248248078e-05}
{"train_loss": 0.3061324656009674, "global_step": 70073, "epoch": 787, "lr": 8.459453391365799e-05}
{"train_loss": 0.15557102859020233, "global_step": 70074, "epoch": 787, "lr": 8.459411534018438e-05}
{"train_loss": 0.13254688680171967, "global_step": 70075, "epoch": 787, "lr": 8.459369676205999e-05}
{"train_loss": 0.19216789305210114, "global_step": 70076, "epoch": 787, "lr": 8.459327817928493e-05}
{"train_loss": 0.21692508459091187, "global_step": 70077, "epoch": 787, "lr": 8.45928595918592e-05}
{"train_loss": 0.1739025115966797, "global_step": 70078, "epoch": 787, "lr": 8.459244099978291e-05}
{"train_loss": 0.27906662225723267, "global_step": 70079, "epoch": 787, "lr": 8.459202240305609e-05}
{"train_loss": 0.1637677699327469, "global_step": 70080, "epoch": 787, "lr": 8.459160380167877e-05}
{"train_loss": 0.16205903887748718, "global_step": 70081, "epoch": 787, "lr": 8.459118519565104e-05}
{"train_loss": 0.30304476618766785, "global_step": 70082, "epoch": 787, "lr": 8.459076658497294e-05}
{"train_loss": 0.22877120971679688, "global_step": 70083, "epoch": 787, "lr": 8.459034796964453e-05}
{"train_loss": 0.21360160410404205, "global_step": 70084, "epoch": 787, "lr": 8.458992934966588e-05}
{"train_loss": 0.20846211910247803, "global_step": 70085, "epoch": 787, "lr": 8.458951072503704e-05}
{"train_loss": 0.1429707109928131, "global_step": 70086, "epoch": 787, "lr": 8.458909209575805e-05}
{"train_loss": 0.28772562742233276, "global_step": 70087, "epoch": 787, "lr": 8.4588673461829e-05}
{"train_loss": 0.21998102962970734, "global_step": 70088, "epoch": 787, "lr": 8.458825482324987e-05}
{"train_loss": 0.2179989516735077, "global_step": 70089, "epoch": 787, "lr": 8.458783618002082e-05}
{"train_loss": 0.17534157633781433, "global_step": 70090, "epoch": 787, "lr": 8.458741753214186e-05}
{"train_loss": 0.21380744874477386, "global_step": 70091, "epoch": 787, "lr": 8.458699887961302e-05}
{"train_loss": 0.22484323382377625, "global_step": 70092, "epoch": 787, "lr": 8.458658022243438e-05}
{"train_loss": 0.14080822467803955, "global_step": 70093, "epoch": 787, "lr": 8.458616156060602e-05}
{"train_loss": 0.25053393840789795, "global_step": 70094, "epoch": 787, "lr": 8.458574289412795e-05}
{"train_loss": 0.21489179134368896, "global_step": 70095, "epoch": 787, "lr": 8.458532422300024e-05}
{"train_loss": 0.28373774886131287, "global_step": 70096, "epoch": 787, "lr": 8.458490554722298e-05}
{"train_loss": 0.23059189319610596, "global_step": 70097, "epoch": 787, "lr": 8.458448686679619e-05}
{"train_loss": 0.19773465394973755, "global_step": 70098, "epoch": 787, "lr": 8.458406818171993e-05}
{"train_loss": 0.1919568032026291, "global_step": 70099, "epoch": 787, "lr": 8.458364949199428e-05}
{"train_loss": 0.19951243698596954, "global_step": 70100, "epoch": 787, "lr": 8.458323079761927e-05}
{"train_loss": 0.3122333288192749, "global_step": 70101, "epoch": 787, "lr": 8.458281209859497e-05}
{"train_loss": 0.21307623386383057, "global_step": 70102, "epoch": 787, "lr": 8.458239339492143e-05}
{"train_loss": 0.1789441555738449, "global_step": 70103, "epoch": 787, "lr": 8.45819746865987e-05}
{"train_loss": 0.2915956676006317, "global_step": 70104, "epoch": 787, "lr": 8.458155597362687e-05}
{"train_loss": 0.20235759019851685, "global_step": 70105, "epoch": 787, "lr": 8.458113725600595e-05}
{"train_loss": 0.20788484811782837, "global_step": 70106, "epoch": 787, "lr": 8.458071853373602e-05}
{"train_loss": 0.13902316987514496, "global_step": 70107, "epoch": 787, "lr": 8.458029980681712e-05}
{"train_loss": 0.22999832034111023, "global_step": 70108, "epoch": 787, "lr": 8.457988107524934e-05}
{"train_loss": 0.17064990103244781, "global_step": 70109, "epoch": 787, "lr": 8.457946233903272e-05}
{"train_loss": 0.21673475205898285, "global_step": 70110, "epoch": 787, "lr": 8.45790435981673e-05}
{"train_loss": 0.1278705894947052, "global_step": 70111, "epoch": 787, "lr": 8.457862485265316e-05}
{"train_loss": 0.2136673927307129, "global_step": 70112, "epoch": 787, "lr": 8.457820610249034e-05}
{"train_loss": 0.19058440625667572, "global_step": 70113, "epoch": 787, "lr": 8.457778734767889e-05}
{"train_loss": 0.22161348164081573, "global_step": 70114, "epoch": 787, "lr": 8.457736858821889e-05}
{"train_loss": 0.18090255558490753, "global_step": 70115, "epoch": 787, "lr": 8.457694982411039e-05}
{"train_loss": 0.13875454664230347, "global_step": 70116, "epoch": 787, "lr": 8.457653105535343e-05}
{"train_loss": 0.16957826912403107, "global_step": 70117, "epoch": 787, "lr": 8.457611228194809e-05}
{"train_loss": 0.16490204632282257, "global_step": 70118, "epoch": 787, "lr": 8.45756935038944e-05}
{"train_loss": 0.2040724754333496, "global_step": 70119, "epoch": 787, "lr": 8.457527472119243e-05}
{"train_loss": 0.24836423993110657, "global_step": 70120, "epoch": 787, "lr": 8.457485593384224e-05}
{"train_loss": 0.2051318734884262, "global_step": 70121, "epoch": 787, "lr": 8.457443714184388e-05}
{"train_loss": 0.1953088641166687, "global_step": 70122, "epoch": 787, "lr": 8.457401834519741e-05}
{"train_loss": 0.18655885756015778, "global_step": 70123, "epoch": 787, "lr": 8.457359954390289e-05}
{"train_loss": 0.1653623729944229, "global_step": 70124, "epoch": 787, "lr": 8.457318073796036e-05}
{"train_loss": 0.26478469371795654, "global_step": 70125, "epoch": 787, "lr": 8.457276192736989e-05}
{"train_loss": 0.22297924757003784, "global_step": 70126, "epoch": 787, "lr": 8.457234311213154e-05}
{"train_loss": 0.2832087278366089, "global_step": 70127, "epoch": 787, "lr": 8.457192429224534e-05}
{"train_loss": 0.21961237490177155, "global_step": 70128, "epoch": 787, "lr": 8.45715054677114e-05}
{"train_loss": 0.22908829152584076, "global_step": 70129, "epoch": 787, "lr": 8.457108663852971e-05}
{"train_loss": 0.14428389072418213, "global_step": 70130, "epoch": 787, "lr": 8.457066780470037e-05}
{"train_loss": 0.20855987390105643, "global_step": 70131, "epoch": 787, "lr": 8.457024896622344e-05, "val_loss": 3.733053684234619}
{"train_loss": 0.24930477142333984, "global_step": 70132, "epoch": 788, "lr": 8.456983012309895e-05}
{"train_loss": 0.19201289117336273, "global_step": 70133, "epoch": 788, "lr": 8.456941127532696e-05}
{"train_loss": 0.21517980098724365, "global_step": 70134, "epoch": 788, "lr": 8.456899242290755e-05}
{"train_loss": 0.20950791239738464, "global_step": 70135, "epoch": 788, "lr": 8.456857356584074e-05}
{"train_loss": 0.1404053419828415, "global_step": 70136, "epoch": 788, "lr": 8.456815470412662e-05}
{"train_loss": 0.1602100282907486, "global_step": 70137, "epoch": 788, "lr": 8.456773583776524e-05}
{"train_loss": 0.2244526594877243, "global_step": 70138, "epoch": 788, "lr": 8.456731696675664e-05}
{"train_loss": 0.21207131445407867, "global_step": 70139, "epoch": 788, "lr": 8.456689809110088e-05}
{"train_loss": 0.1955382227897644, "global_step": 70140, "epoch": 788, "lr": 8.456647921079803e-05}
{"train_loss": 0.25548839569091797, "global_step": 70141, "epoch": 788, "lr": 8.456606032584813e-05}
{"train_loss": 0.2176368534564972, "global_step": 70142, "epoch": 788, "lr": 8.456564143625126e-05}
{"train_loss": 0.2849630117416382, "global_step": 70143, "epoch": 788, "lr": 8.456522254200745e-05}
{"train_loss": 0.259470671415329, "global_step": 70144, "epoch": 788, "lr": 8.456480364311676e-05}
{"train_loss": 0.28296780586242676, "global_step": 70145, "epoch": 788, "lr": 8.456438473957926e-05}
{"train_loss": 0.19911493360996246, "global_step": 70146, "epoch": 788, "lr": 8.456396583139501e-05}
{"train_loss": 0.28485384583473206, "global_step": 70147, "epoch": 788, "lr": 8.456354691856404e-05}
{"train_loss": 0.21294356882572174, "global_step": 70148, "epoch": 788, "lr": 8.456312800108644e-05}
{"train_loss": 0.23260316252708435, "global_step": 70149, "epoch": 788, "lr": 8.456270907896223e-05}
{"train_loss": 0.23928740620613098, "global_step": 70150, "epoch": 788, "lr": 8.45622901521915e-05}
{"train_loss": 0.1667761355638504, "global_step": 70151, "epoch": 788, "lr": 8.456187122077428e-05}
{"train_loss": 0.19814035296440125, "global_step": 70152, "epoch": 788, "lr": 8.456145228471064e-05}
{"train_loss": 0.32158762216567993, "global_step": 70153, "epoch": 788, "lr": 8.456103334400063e-05}
{"train_loss": 0.1908947080373764, "global_step": 70154, "epoch": 788, "lr": 8.456061439864434e-05}
{"train_loss": 0.15207864344120026, "global_step": 70155, "epoch": 788, "lr": 8.456019544864176e-05}
{"train_loss": 0.19668225944042206, "global_step": 70156, "epoch": 788, "lr": 8.455977649399301e-05}
{"train_loss": 0.16370747983455658, "global_step": 70157, "epoch": 788, "lr": 8.45593575346981e-05}
{"train_loss": 0.20773620903491974, "global_step": 70158, "epoch": 788, "lr": 8.455893857075711e-05}
{"train_loss": 0.31796571612358093, "global_step": 70159, "epoch": 788, "lr": 8.45585196021701e-05}
{"train_loss": 0.24210956692695618, "global_step": 70160, "epoch": 788, "lr": 8.45581006289371e-05}
{"train_loss": 0.2225046306848526, "global_step": 70161, "epoch": 788, "lr": 8.45576816510582e-05}
{"train_loss": 0.2680390775203705, "global_step": 70162, "epoch": 788, "lr": 8.455726266853343e-05}
{"train_loss": 0.1964932382106781, "global_step": 70163, "epoch": 788, "lr": 8.455684368136286e-05}
{"train_loss": 0.23789232969284058, "global_step": 70164, "epoch": 788, "lr": 8.455642468954655e-05}
{"train_loss": 0.3434933125972748, "global_step": 70165, "epoch": 788, "lr": 8.455600569308455e-05}
{"train_loss": 0.22063593566417694, "global_step": 70166, "epoch": 788, "lr": 8.455558669197692e-05}
{"train_loss": 0.2547706961631775, "global_step": 70167, "epoch": 788, "lr": 8.455516768622369e-05}
{"train_loss": 0.17465440928936005, "global_step": 70168, "epoch": 788, "lr": 8.455474867582497e-05}
{"train_loss": 0.1886403113603592, "global_step": 70169, "epoch": 788, "lr": 8.455432966078076e-05}
{"train_loss": 0.18307428061962128, "global_step": 70170, "epoch": 788, "lr": 8.455391064109115e-05}
{"train_loss": 0.23486191034317017, "global_step": 70171, "epoch": 788, "lr": 8.45534916167562e-05}
{"train_loss": 0.239476278424263, "global_step": 70172, "epoch": 788, "lr": 8.455307258777593e-05}
{"train_loss": 0.29940369725227356, "global_step": 70173, "epoch": 788, "lr": 8.455265355415045e-05}
{"train_loss": 0.3725678026676178, "global_step": 70174, "epoch": 788, "lr": 8.455223451587976e-05}
{"train_loss": 0.14417865872383118, "global_step": 70175, "epoch": 788, "lr": 8.455181547296396e-05}
{"train_loss": 0.1982111930847168, "global_step": 70176, "epoch": 788, "lr": 8.455139642540308e-05}
{"train_loss": 0.24505381286144257, "global_step": 70177, "epoch": 788, "lr": 8.45509773731972e-05}
{"train_loss": 0.2517029047012329, "global_step": 70178, "epoch": 788, "lr": 8.455055831634636e-05}
{"train_loss": 0.173444464802742, "global_step": 70179, "epoch": 788, "lr": 8.45501392548506e-05}
{"train_loss": 0.14753390848636627, "global_step": 70180, "epoch": 788, "lr": 8.454972018871e-05}
{"train_loss": 0.16685700416564941, "global_step": 70181, "epoch": 788, "lr": 8.454930111792461e-05}
{"train_loss": 0.22948458790779114, "global_step": 70182, "epoch": 788, "lr": 8.454888204249451e-05}
{"train_loss": 0.2138594686985016, "global_step": 70183, "epoch": 788, "lr": 8.454846296241972e-05}
{"train_loss": 0.16907300055027008, "global_step": 70184, "epoch": 788, "lr": 8.45480438777003e-05}
{"train_loss": 0.18075107038021088, "global_step": 70185, "epoch": 788, "lr": 8.454762478833633e-05}
{"train_loss": 0.1743917465209961, "global_step": 70186, "epoch": 788, "lr": 8.454720569432784e-05}
{"train_loss": 0.207464799284935, "global_step": 70187, "epoch": 788, "lr": 8.454678659567491e-05}
{"train_loss": 0.22283029556274414, "global_step": 70188, "epoch": 788, "lr": 8.454636749237759e-05}
{"train_loss": 0.25585776567459106, "global_step": 70189, "epoch": 788, "lr": 8.454594838443592e-05}
{"train_loss": 0.15551143884658813, "global_step": 70190, "epoch": 788, "lr": 8.454552927184998e-05}
{"train_loss": 0.2567923367023468, "global_step": 70191, "epoch": 788, "lr": 8.454511015461981e-05}
{"train_loss": 0.16802014410495758, "global_step": 70192, "epoch": 788, "lr": 8.454469103274547e-05}
{"train_loss": 0.2143905609846115, "global_step": 70193, "epoch": 788, "lr": 8.454427190622701e-05}
{"train_loss": 0.30396130681037903, "global_step": 70194, "epoch": 788, "lr": 8.454385277506449e-05}
{"train_loss": 0.2441011369228363, "global_step": 70195, "epoch": 788, "lr": 8.454343363925799e-05}
{"train_loss": 0.14267268776893616, "global_step": 70196, "epoch": 788, "lr": 8.454301449880753e-05}
{"train_loss": 0.2179054319858551, "global_step": 70197, "epoch": 788, "lr": 8.454259535371318e-05}
{"train_loss": 0.1935843825340271, "global_step": 70198, "epoch": 788, "lr": 8.454217620397503e-05}
{"train_loss": 0.28786516189575195, "global_step": 70199, "epoch": 788, "lr": 8.454175704959307e-05}
{"train_loss": 0.23743076622486115, "global_step": 70200, "epoch": 788, "lr": 8.45413378905674e-05}
{"train_loss": 0.30705708265304565, "global_step": 70201, "epoch": 788, "lr": 8.454091872689808e-05}
{"train_loss": 0.21072392165660858, "global_step": 70202, "epoch": 788, "lr": 8.454049955858515e-05}
{"train_loss": 0.15776784718036652, "global_step": 70203, "epoch": 788, "lr": 8.454008038562866e-05}
{"train_loss": 0.26337799429893494, "global_step": 70204, "epoch": 788, "lr": 8.45396612080287e-05}
{"train_loss": 0.26155146956443787, "global_step": 70205, "epoch": 788, "lr": 8.453924202578528e-05}
{"train_loss": 0.2986558675765991, "global_step": 70206, "epoch": 788, "lr": 8.45388228388985e-05}
{"train_loss": 0.14992861449718475, "global_step": 70207, "epoch": 788, "lr": 8.453840364736837e-05}
{"train_loss": 0.2085709273815155, "global_step": 70208, "epoch": 788, "lr": 8.4537984451195e-05}
{"train_loss": 0.1547665148973465, "global_step": 70209, "epoch": 788, "lr": 8.45375652503784e-05}
{"train_loss": 0.30983883142471313, "global_step": 70210, "epoch": 788, "lr": 8.453714604491865e-05}
{"train_loss": 0.18596751987934113, "global_step": 70211, "epoch": 788, "lr": 8.453672683481581e-05}
{"train_loss": 0.18281632661819458, "global_step": 70212, "epoch": 788, "lr": 8.453630762006993e-05}
{"train_loss": 0.1887727975845337, "global_step": 70213, "epoch": 788, "lr": 8.453588840068105e-05}
{"train_loss": 0.19873563945293427, "global_step": 70214, "epoch": 788, "lr": 8.453546917664925e-05}
{"train_loss": 0.20351159572601318, "global_step": 70215, "epoch": 788, "lr": 8.453504994797457e-05}
{"train_loss": 0.21745197474956512, "global_step": 70216, "epoch": 788, "lr": 8.453463071465708e-05}
{"train_loss": 0.29139000177383423, "global_step": 70217, "epoch": 788, "lr": 8.453421147669682e-05}
{"train_loss": 0.25125062465667725, "global_step": 70218, "epoch": 788, "lr": 8.453379223409386e-05}
{"train_loss": 0.2643236219882965, "global_step": 70219, "epoch": 788, "lr": 8.453337298684827e-05}
{"train_loss": 0.2212065305100398, "global_step": 70220, "epoch": 788, "lr": 8.453295373496006e-05, "val_loss": 3.716531991958618}
{"train_loss": 0.25061658024787903, "global_step": 70221, "epoch": 789, "lr": 8.453253447842934e-05}
{"train_loss": 0.1795373558998108, "global_step": 70222, "epoch": 789, "lr": 8.453211521725613e-05}
{"train_loss": 0.1551659256219864, "global_step": 70223, "epoch": 789, "lr": 8.45316959514405e-05}
{"train_loss": 0.19930852949619293, "global_step": 70224, "epoch": 789, "lr": 8.453127668098251e-05}
{"train_loss": 0.18347114324569702, "global_step": 70225, "epoch": 789, "lr": 8.45308574058822e-05}
{"train_loss": 0.18315710127353668, "global_step": 70226, "epoch": 789, "lr": 8.453043812613965e-05}
{"train_loss": 0.23424899578094482, "global_step": 70227, "epoch": 789, "lr": 8.453001884175489e-05}
{"train_loss": 0.237296462059021, "global_step": 70228, "epoch": 789, "lr": 8.452959955272798e-05}
{"train_loss": 0.2145526260137558, "global_step": 70229, "epoch": 789, "lr": 8.452918025905903e-05}
{"train_loss": 0.20477788150310516, "global_step": 70230, "epoch": 789, "lr": 8.452876096074801e-05}
{"train_loss": 0.18856921792030334, "global_step": 70231, "epoch": 789, "lr": 8.452834165779505e-05}
{"train_loss": 0.21667945384979248, "global_step": 70232, "epoch": 789, "lr": 8.452792235020015e-05}
{"train_loss": 0.24876338243484497, "global_step": 70233, "epoch": 789, "lr": 8.45275030379634e-05}
{"train_loss": 0.2422514110803604, "global_step": 70234, "epoch": 789, "lr": 8.452708372108485e-05}
{"train_loss": 0.19274009764194489, "global_step": 70235, "epoch": 789, "lr": 8.452666439956454e-05}
{"train_loss": 0.28511297702789307, "global_step": 70236, "epoch": 789, "lr": 8.452624507340255e-05}
{"train_loss": 0.17841744422912598, "global_step": 70237, "epoch": 789, "lr": 8.452582574259893e-05}
{"train_loss": 0.3000795245170593, "global_step": 70238, "epoch": 789, "lr": 8.452540640715373e-05}
{"train_loss": 0.1769067943096161, "global_step": 70239, "epoch": 789, "lr": 8.452498706706702e-05}
{"train_loss": 0.2704600393772125, "global_step": 70240, "epoch": 789, "lr": 8.452456772233885e-05}
{"train_loss": 0.19293545186519623, "global_step": 70241, "epoch": 789, "lr": 8.452414837296926e-05}
{"train_loss": 0.1859537661075592, "global_step": 70242, "epoch": 789, "lr": 8.452372901895832e-05}
{"train_loss": 0.18162886798381805, "global_step": 70243, "epoch": 789, "lr": 8.452330966030608e-05}
{"train_loss": 0.18771657347679138, "global_step": 70244, "epoch": 789, "lr": 8.452289029701261e-05}
{"train_loss": 0.16345185041427612, "global_step": 70245, "epoch": 789, "lr": 8.452247092907796e-05}
{"train_loss": 0.2591959834098816, "global_step": 70246, "epoch": 789, "lr": 8.452205155650219e-05}
{"train_loss": 0.3102840781211853, "global_step": 70247, "epoch": 789, "lr": 8.452163217928535e-05}
{"train_loss": 0.2204294204711914, "global_step": 70248, "epoch": 789, "lr": 8.452121279742748e-05}
{"train_loss": 0.21359260380268097, "global_step": 70249, "epoch": 789, "lr": 8.452079341092867e-05}
{"train_loss": 0.23379965126514435, "global_step": 70250, "epoch": 789, "lr": 8.452037401978895e-05}
{"train_loss": 0.21842534840106964, "global_step": 70251, "epoch": 789, "lr": 8.451995462400839e-05}
{"train_loss": 0.15226410329341888, "global_step": 70252, "epoch": 789, "lr": 8.451953522358704e-05}
{"train_loss": 0.1550096720457077, "global_step": 70253, "epoch": 789, "lr": 8.451911581852496e-05}
{"train_loss": 0.22388318181037903, "global_step": 70254, "epoch": 789, "lr": 8.451869640882222e-05}
{"train_loss": 0.2872224748134613, "global_step": 70255, "epoch": 789, "lr": 8.451827699447883e-05}
{"train_loss": 0.23674878478050232, "global_step": 70256, "epoch": 789, "lr": 8.45178575754949e-05}
{"train_loss": 0.16890975832939148, "global_step": 70257, "epoch": 789, "lr": 8.451743815187048e-05}
{"train_loss": 0.13805563747882843, "global_step": 70258, "epoch": 789, "lr": 8.451701872360558e-05}
{"train_loss": 0.21452824771404266, "global_step": 70259, "epoch": 789, "lr": 8.45165992907003e-05}
{"train_loss": 0.25961384177207947, "global_step": 70260, "epoch": 789, "lr": 8.45161798531547e-05}
{"train_loss": 0.22308938205242157, "global_step": 70261, "epoch": 789, "lr": 8.45157604109688e-05}
{"train_loss": 0.19854769110679626, "global_step": 70262, "epoch": 789, "lr": 8.451534096414269e-05}
{"train_loss": 0.1823522448539734, "global_step": 70263, "epoch": 789, "lr": 8.45149215126764e-05}
{"train_loss": 0.23398397862911224, "global_step": 70264, "epoch": 789, "lr": 8.451450205657001e-05}
{"train_loss": 0.221382275223732, "global_step": 70265, "epoch": 789, "lr": 8.451408259582356e-05}
{"train_loss": 0.29996544122695923, "global_step": 70266, "epoch": 789, "lr": 8.451366313043711e-05}
{"train_loss": 0.2889047861099243, "global_step": 70267, "epoch": 789, "lr": 8.451324366041072e-05}
{"train_loss": 0.25797176361083984, "global_step": 70268, "epoch": 789, "lr": 8.451282418574445e-05}
{"train_loss": 0.2048666775226593, "global_step": 70269, "epoch": 789, "lr": 8.451240470643834e-05}
{"train_loss": 0.19571007788181305, "global_step": 70270, "epoch": 789, "lr": 8.451198522249247e-05}
{"train_loss": 0.1780676543712616, "global_step": 70271, "epoch": 789, "lr": 8.451156573390688e-05}
{"train_loss": 0.1848684847354889, "global_step": 70272, "epoch": 789, "lr": 8.451114624068164e-05}
{"train_loss": 0.21580272912979126, "global_step": 70273, "epoch": 789, "lr": 8.45107267428168e-05}
{"train_loss": 0.261959433555603, "global_step": 70274, "epoch": 789, "lr": 8.451030724031239e-05}
{"train_loss": 0.15651561319828033, "global_step": 70275, "epoch": 789, "lr": 8.45098877331685e-05}
{"train_loss": 0.22732344269752502, "global_step": 70276, "epoch": 789, "lr": 8.450946822138518e-05}
{"train_loss": 0.24917849898338318, "global_step": 70277, "epoch": 789, "lr": 8.450904870496248e-05}
{"train_loss": 0.2244914025068283, "global_step": 70278, "epoch": 789, "lr": 8.450862918390045e-05}
{"train_loss": 0.2709181308746338, "global_step": 70279, "epoch": 789, "lr": 8.450820965819917e-05}
{"train_loss": 0.22029156982898712, "global_step": 70280, "epoch": 789, "lr": 8.450779012785867e-05}
{"train_loss": 0.1953209936618805, "global_step": 70281, "epoch": 789, "lr": 8.450737059287903e-05}
{"train_loss": 0.2445436418056488, "global_step": 70282, "epoch": 789, "lr": 8.45069510532603e-05}
{"train_loss": 0.22001111507415771, "global_step": 70283, "epoch": 789, "lr": 8.450653150900252e-05}
{"train_loss": 0.08795474469661713, "global_step": 70284, "epoch": 789, "lr": 8.450611196010576e-05}
{"train_loss": 0.18004490435123444, "global_step": 70285, "epoch": 789, "lr": 8.450569240657007e-05}
{"train_loss": 0.23146983981132507, "global_step": 70286, "epoch": 789, "lr": 8.450527284839551e-05}
{"train_loss": 0.15079446136951447, "global_step": 70287, "epoch": 789, "lr": 8.450485328558213e-05}
{"train_loss": 0.26089635491371155, "global_step": 70288, "epoch": 789, "lr": 8.450443371812999e-05}
{"train_loss": 0.2154606729745865, "global_step": 70289, "epoch": 789, "lr": 8.450401414603917e-05}
{"train_loss": 0.27559030055999756, "global_step": 70290, "epoch": 789, "lr": 8.45035945693097e-05}
{"train_loss": 0.1845446079969406, "global_step": 70291, "epoch": 789, "lr": 8.450317498794164e-05}
{"train_loss": 0.2263839840888977, "global_step": 70292, "epoch": 789, "lr": 8.450275540193505e-05}
{"train_loss": 0.21346749365329742, "global_step": 70293, "epoch": 789, "lr": 8.450233581128999e-05}
{"train_loss": 0.23896123468875885, "global_step": 70294, "epoch": 789, "lr": 8.450191621600649e-05}
{"train_loss": 0.23146089911460876, "global_step": 70295, "epoch": 789, "lr": 8.450149661608465e-05}
{"train_loss": 0.24446164071559906, "global_step": 70296, "epoch": 789, "lr": 8.45010770115245e-05}
{"train_loss": 0.2081841081380844, "global_step": 70297, "epoch": 789, "lr": 8.450065740232609e-05}
{"train_loss": 0.212400883436203, "global_step": 70298, "epoch": 789, "lr": 8.45002377884895e-05}
{"train_loss": 0.21769475936889648, "global_step": 70299, "epoch": 789, "lr": 8.449981817001476e-05}
{"train_loss": 0.2499084174633026, "global_step": 70300, "epoch": 789, "lr": 8.449939854690195e-05}
{"train_loss": 0.1769619733095169, "global_step": 70301, "epoch": 789, "lr": 8.449897891915112e-05}
{"train_loss": 0.20971745252609253, "global_step": 70302, "epoch": 789, "lr": 8.449855928676232e-05}
{"train_loss": 0.1875329315662384, "global_step": 70303, "epoch": 789, "lr": 8.449813964973561e-05}
{"train_loss": 0.2203005701303482, "global_step": 70304, "epoch": 789, "lr": 8.449772000807103e-05}
{"train_loss": 0.21837766468524933, "global_step": 70305, "epoch": 789, "lr": 8.449730036176866e-05}
{"train_loss": 0.15924125909805298, "global_step": 70306, "epoch": 789, "lr": 8.449688071082855e-05}
{"train_loss": 0.21441544592380524, "global_step": 70307, "epoch": 789, "lr": 8.449646105525077e-05}
{"train_loss": 0.20250624418258667, "global_step": 70308, "epoch": 789, "lr": 8.449604139503535e-05}
{"train_loss": 0.21510827206493763, "global_step": 70309, "epoch": 789, "lr": 8.449562173018235e-05, "val_loss": 3.80778169631958}
{"train_loss": 0.16410857439041138, "global_step": 70310, "epoch": 790, "lr": 8.449520206069184e-05}
{"train_loss": 0.1565595269203186, "global_step": 70311, "epoch": 790, "lr": 8.449478238656387e-05}
{"train_loss": 0.172227680683136, "global_step": 70312, "epoch": 790, "lr": 8.44943627077985e-05}
{"train_loss": 0.1987045705318451, "global_step": 70313, "epoch": 790, "lr": 8.449394302439577e-05}
{"train_loss": 0.19915056228637695, "global_step": 70314, "epoch": 790, "lr": 8.449352333635576e-05}
{"train_loss": 0.23928751051425934, "global_step": 70315, "epoch": 790, "lr": 8.449310364367852e-05}
{"train_loss": 0.1515333354473114, "global_step": 70316, "epoch": 790, "lr": 8.44926839463641e-05}
{"train_loss": 0.2871643602848053, "global_step": 70317, "epoch": 790, "lr": 8.449226424441257e-05}
{"train_loss": 0.17529475688934326, "global_step": 70318, "epoch": 790, "lr": 8.449184453782397e-05}
{"train_loss": 0.17078691720962524, "global_step": 70319, "epoch": 790, "lr": 8.449142482659835e-05}
{"train_loss": 0.21914497017860413, "global_step": 70320, "epoch": 790, "lr": 8.449100511073578e-05}
{"train_loss": 0.137185737490654, "global_step": 70321, "epoch": 790, "lr": 8.449058539023632e-05}
{"train_loss": 0.21857596933841705, "global_step": 70322, "epoch": 790, "lr": 8.449016566510002e-05}
{"train_loss": 0.2578997015953064, "global_step": 70323, "epoch": 790, "lr": 8.448974593532693e-05}
{"train_loss": 0.12058853358030319, "global_step": 70324, "epoch": 790, "lr": 8.448932620091712e-05}
{"train_loss": 0.21913902461528778, "global_step": 70325, "epoch": 790, "lr": 8.448890646187065e-05}
{"train_loss": 0.2635093629360199, "global_step": 70326, "epoch": 790, "lr": 8.448848671818756e-05}
{"train_loss": 0.18336938321590424, "global_step": 70327, "epoch": 790, "lr": 8.44880669698679e-05}
{"train_loss": 0.1877528578042984, "global_step": 70328, "epoch": 790, "lr": 8.448764721691176e-05}
{"train_loss": 0.1753164827823639, "global_step": 70329, "epoch": 790, "lr": 8.448722745931917e-05}
{"train_loss": 0.1677115112543106, "global_step": 70330, "epoch": 790, "lr": 8.448680769709019e-05}
{"train_loss": 0.26357245445251465, "global_step": 70331, "epoch": 790, "lr": 8.448638793022488e-05}
{"train_loss": 0.20604467391967773, "global_step": 70332, "epoch": 790, "lr": 8.448596815872329e-05}
{"train_loss": 0.2651808261871338, "global_step": 70333, "epoch": 790, "lr": 8.448554838258549e-05}
{"train_loss": 0.16149786114692688, "global_step": 70334, "epoch": 790, "lr": 8.448512860181152e-05}
{"train_loss": 0.24682992696762085, "global_step": 70335, "epoch": 790, "lr": 8.448470881640147e-05}
{"train_loss": 0.2619938254356384, "global_step": 70336, "epoch": 790, "lr": 8.448428902635534e-05}
{"train_loss": 0.20915940403938293, "global_step": 70337, "epoch": 790, "lr": 8.448386923167324e-05}
{"train_loss": 0.3404944837093353, "global_step": 70338, "epoch": 790, "lr": 8.44834494323552e-05}
{"train_loss": 0.2310589998960495, "global_step": 70339, "epoch": 790, "lr": 8.448302962840127e-05}
{"train_loss": 0.163192018866539, "global_step": 70340, "epoch": 790, "lr": 8.44826098198115e-05}
{"train_loss": 0.20425792038440704, "global_step": 70341, "epoch": 790, "lr": 8.4482190006586e-05}
{"train_loss": 0.1636449247598648, "global_step": 70342, "epoch": 790, "lr": 8.448177018872476e-05}
{"train_loss": 0.16372916102409363, "global_step": 70343, "epoch": 790, "lr": 8.448135036622789e-05}
{"train_loss": 0.2648484408855438, "global_step": 70344, "epoch": 790, "lr": 8.448093053909541e-05}
{"train_loss": 0.19982761144638062, "global_step": 70345, "epoch": 790, "lr": 8.448051070732738e-05}
{"train_loss": 0.1518068015575409, "global_step": 70346, "epoch": 790, "lr": 8.44800908709239e-05}
{"train_loss": 0.26422053575515747, "global_step": 70347, "epoch": 790, "lr": 8.447967102988497e-05}
{"train_loss": 0.21384918689727783, "global_step": 70348, "epoch": 790, "lr": 8.447925118421067e-05}
{"train_loss": 0.19507500529289246, "global_step": 70349, "epoch": 790, "lr": 8.447883133390104e-05}
{"train_loss": 0.242720365524292, "global_step": 70350, "epoch": 790, "lr": 8.447841147895616e-05}
{"train_loss": 0.3784922957420349, "global_step": 70351, "epoch": 790, "lr": 8.44779916193761e-05}
{"train_loss": 0.17325037717819214, "global_step": 70352, "epoch": 790, "lr": 8.447757175516087e-05}
{"train_loss": 0.21399861574172974, "global_step": 70353, "epoch": 790, "lr": 8.447715188631056e-05}
{"train_loss": 0.16062720119953156, "global_step": 70354, "epoch": 790, "lr": 8.447673201282522e-05}
{"train_loss": 0.14062686264514923, "global_step": 70355, "epoch": 790, "lr": 8.447631213470489e-05}
{"train_loss": 0.23180291056632996, "global_step": 70356, "epoch": 790, "lr": 8.447589225194965e-05}
{"train_loss": 0.22940126061439514, "global_step": 70357, "epoch": 790, "lr": 8.447547236455954e-05}
{"train_loss": 0.20442168414592743, "global_step": 70358, "epoch": 790, "lr": 8.447505247253464e-05}
{"train_loss": 0.2468770444393158, "global_step": 70359, "epoch": 790, "lr": 8.447463257587497e-05}
{"train_loss": 0.2598258852958679, "global_step": 70360, "epoch": 790, "lr": 8.447421267458061e-05}
{"train_loss": 0.20472271740436554, "global_step": 70361, "epoch": 790, "lr": 8.447379276865162e-05}
{"train_loss": 0.2703799605369568, "global_step": 70362, "epoch": 790, "lr": 8.447337285808804e-05}
{"train_loss": 0.256707102060318, "global_step": 70363, "epoch": 790, "lr": 8.447295294288994e-05}
{"train_loss": 0.2562505602836609, "global_step": 70364, "epoch": 790, "lr": 8.447253302305736e-05}
{"train_loss": 0.15894745290279388, "global_step": 70365, "epoch": 790, "lr": 8.447211309859039e-05}
{"train_loss": 0.21488045156002045, "global_step": 70366, "epoch": 790, "lr": 8.447169316948904e-05}
{"train_loss": 0.215470090508461, "global_step": 70367, "epoch": 790, "lr": 8.447127323575341e-05}
{"train_loss": 0.219618558883667, "global_step": 70368, "epoch": 790, "lr": 8.447085329738354e-05}
{"train_loss": 0.20208320021629333, "global_step": 70369, "epoch": 790, "lr": 8.447043335437948e-05}
{"train_loss": 0.19613847136497498, "global_step": 70370, "epoch": 790, "lr": 8.447001340674128e-05}
{"train_loss": 0.2009349912405014, "global_step": 70371, "epoch": 790, "lr": 8.4469593454469e-05}
{"train_loss": 0.24635522067546844, "global_step": 70372, "epoch": 790, "lr": 8.446917349756273e-05}
{"train_loss": 0.21962155401706696, "global_step": 70373, "epoch": 790, "lr": 8.446875353602248e-05}
{"train_loss": 0.22399723529815674, "global_step": 70374, "epoch": 790, "lr": 8.446833356984833e-05}
{"train_loss": 0.16854874789714813, "global_step": 70375, "epoch": 790, "lr": 8.446791359904034e-05}
{"train_loss": 0.2270929217338562, "global_step": 70376, "epoch": 790, "lr": 8.446749362359856e-05}
{"train_loss": 0.1812087446451187, "global_step": 70377, "epoch": 790, "lr": 8.446707364352305e-05}
{"train_loss": 0.2875822186470032, "global_step": 70378, "epoch": 790, "lr": 8.446665365881383e-05}
{"train_loss": 0.23796549439430237, "global_step": 70379, "epoch": 790, "lr": 8.446623366947102e-05}
{"train_loss": 0.18639791011810303, "global_step": 70380, "epoch": 790, "lr": 8.446581367549463e-05}
{"train_loss": 0.21282050013542175, "global_step": 70381, "epoch": 790, "lr": 8.446539367688472e-05}
{"train_loss": 0.2527598738670349, "global_step": 70382, "epoch": 790, "lr": 8.446497367364138e-05}
{"train_loss": 0.2648959457874298, "global_step": 70383, "epoch": 790, "lr": 8.446455366576464e-05}
{"train_loss": 0.13099248707294464, "global_step": 70384, "epoch": 790, "lr": 8.446413365325456e-05}
{"train_loss": 0.1494695544242859, "global_step": 70385, "epoch": 790, "lr": 8.44637136361112e-05}
{"train_loss": 0.2286817878484726, "global_step": 70386, "epoch": 790, "lr": 8.446329361433462e-05}
{"train_loss": 0.23486419022083282, "global_step": 70387, "epoch": 790, "lr": 8.446287358792486e-05}
{"train_loss": 0.17104001343250275, "global_step": 70388, "epoch": 790, "lr": 8.4462453556882e-05}
{"train_loss": 0.10219981521368027, "global_step": 70389, "epoch": 790, "lr": 8.446203352120607e-05}
{"train_loss": 0.2174147069454193, "global_step": 70390, "epoch": 790, "lr": 8.446161348089714e-05}
{"train_loss": 0.24066679179668427, "global_step": 70391, "epoch": 790, "lr": 8.446119343595527e-05}
{"train_loss": 0.16451551020145416, "global_step": 70392, "epoch": 790, "lr": 8.44607733863805e-05}
{"train_loss": 0.20182040333747864, "global_step": 70393, "epoch": 790, "lr": 8.44603533321729e-05}
{"train_loss": 0.2100508213043213, "global_step": 70394, "epoch": 790, "lr": 8.445993327333254e-05}
{"train_loss": 0.2638884484767914, "global_step": 70395, "epoch": 790, "lr": 8.445951320985946e-05}
{"train_loss": 0.22247573733329773, "global_step": 70396, "epoch": 790, "lr": 8.445909314175371e-05}
{"train_loss": 0.2261936217546463, "global_step": 70397, "epoch": 790, "lr": 8.445867306901535e-05}
{"train_loss": 0.2109086982319864, "global_step": 70398, "epoch": 790, "lr": 8.445825299164445e-05, "val_loss": 3.893998622894287, "train_action_mse_error": 18.95532989501953}
{"train_loss": 0.24967104196548462, "global_step": 70399, "epoch": 791, "lr": 8.445783290964106e-05}
{"train_loss": 0.18824532628059387, "global_step": 70400, "epoch": 791, "lr": 8.445741282300523e-05}
{"train_loss": 0.2511175870895386, "global_step": 70401, "epoch": 791, "lr": 8.445699273173703e-05}
{"train_loss": 0.18418385088443756, "global_step": 70402, "epoch": 791, "lr": 8.44565726358365e-05}
{"train_loss": 0.1606888473033905, "global_step": 70403, "epoch": 791, "lr": 8.44561525353037e-05}
{"train_loss": 0.26398536562919617, "global_step": 70404, "epoch": 791, "lr": 8.445573243013869e-05}
{"train_loss": 0.23365141451358795, "global_step": 70405, "epoch": 791, "lr": 8.445531232034154e-05}
{"train_loss": 0.18306943774223328, "global_step": 70406, "epoch": 791, "lr": 8.445489220591229e-05}
{"train_loss": 0.17549195885658264, "global_step": 70407, "epoch": 791, "lr": 8.4454472086851e-05}
{"train_loss": 0.20227058231830597, "global_step": 70408, "epoch": 791, "lr": 8.445405196315771e-05}
{"train_loss": 0.14981728792190552, "global_step": 70409, "epoch": 791, "lr": 8.44536318348325e-05}
{"train_loss": 0.12222687155008316, "global_step": 70410, "epoch": 791, "lr": 8.445321170187543e-05}
{"train_loss": 0.20108796656131744, "global_step": 70411, "epoch": 791, "lr": 8.445279156428654e-05}
{"train_loss": 0.23051360249519348, "global_step": 70412, "epoch": 791, "lr": 8.445237142206588e-05}
{"train_loss": 0.260977566242218, "global_step": 70413, "epoch": 791, "lr": 8.445195127521353e-05}
{"train_loss": 0.23116616904735565, "global_step": 70414, "epoch": 791, "lr": 8.445153112372953e-05}
{"train_loss": 0.230373814702034, "global_step": 70415, "epoch": 791, "lr": 8.445111096761396e-05}
{"train_loss": 0.2084352821111679, "global_step": 70416, "epoch": 791, "lr": 8.445069080686682e-05}
{"train_loss": 0.1740407943725586, "global_step": 70417, "epoch": 791, "lr": 8.445027064148824e-05}
{"train_loss": 0.1797708123922348, "global_step": 70418, "epoch": 791, "lr": 8.444985047147823e-05}
{"train_loss": 0.22724217176437378, "global_step": 70419, "epoch": 791, "lr": 8.444943029683685e-05}
{"train_loss": 0.1905868947505951, "global_step": 70420, "epoch": 791, "lr": 8.444901011756418e-05}
{"train_loss": 0.24713857471942902, "global_step": 70421, "epoch": 791, "lr": 8.444858993366025e-05}
{"train_loss": 0.18792793154716492, "global_step": 70422, "epoch": 791, "lr": 8.444816974512513e-05}
{"train_loss": 0.17139247059822083, "global_step": 70423, "epoch": 791, "lr": 8.444774955195887e-05}
{"train_loss": 0.243504598736763, "global_step": 70424, "epoch": 791, "lr": 8.444732935416152e-05}
{"train_loss": 0.15610218048095703, "global_step": 70425, "epoch": 791, "lr": 8.444690915173317e-05}
{"train_loss": 0.17676442861557007, "global_step": 70426, "epoch": 791, "lr": 8.444648894467384e-05}
{"train_loss": 0.1690950244665146, "global_step": 70427, "epoch": 791, "lr": 8.444606873298359e-05}
{"train_loss": 0.25150999426841736, "global_step": 70428, "epoch": 791, "lr": 8.44456485166625e-05}
{"train_loss": 0.2888810932636261, "global_step": 70429, "epoch": 791, "lr": 8.44452282957106e-05}
{"train_loss": 0.2225121706724167, "global_step": 70430, "epoch": 791, "lr": 8.444480807012798e-05}
{"train_loss": 0.19051095843315125, "global_step": 70431, "epoch": 791, "lr": 8.444438783991466e-05}
{"train_loss": 0.27051079273223877, "global_step": 70432, "epoch": 791, "lr": 8.444396760507071e-05}
{"train_loss": 0.21722233295440674, "global_step": 70433, "epoch": 791, "lr": 8.444354736559619e-05}
{"train_loss": 0.18303163349628448, "global_step": 70434, "epoch": 791, "lr": 8.444312712149117e-05}
{"train_loss": 0.2958559989929199, "global_step": 70435, "epoch": 791, "lr": 8.444270687275566e-05}
{"train_loss": 0.1800885647535324, "global_step": 70436, "epoch": 791, "lr": 8.444228661938978e-05}
{"train_loss": 0.2105378955602646, "global_step": 70437, "epoch": 791, "lr": 8.444186636139353e-05}
{"train_loss": 0.20447269082069397, "global_step": 70438, "epoch": 791, "lr": 8.4441446098767e-05}
{"train_loss": 0.22789989411830902, "global_step": 70439, "epoch": 791, "lr": 8.444102583151024e-05}
{"train_loss": 0.15194623172283173, "global_step": 70440, "epoch": 791, "lr": 8.44406055596233e-05}
{"train_loss": 0.24949748814105988, "global_step": 70441, "epoch": 791, "lr": 8.444018528310625e-05}
{"train_loss": 0.20912377536296844, "global_step": 70442, "epoch": 791, "lr": 8.443976500195914e-05}
{"train_loss": 0.20377495884895325, "global_step": 70443, "epoch": 791, "lr": 8.4439344716182e-05}
{"train_loss": 0.20728950202465057, "global_step": 70444, "epoch": 791, "lr": 8.443892442577492e-05}
{"train_loss": 0.31145817041397095, "global_step": 70445, "epoch": 791, "lr": 8.443850413073796e-05}
{"train_loss": 0.22525997459888458, "global_step": 70446, "epoch": 791, "lr": 8.443808383107116e-05}
{"train_loss": 0.2263864278793335, "global_step": 70447, "epoch": 791, "lr": 8.443766352677457e-05}
{"train_loss": 0.36045920848846436, "global_step": 70448, "epoch": 791, "lr": 8.443724321784825e-05}
{"train_loss": 0.23580653965473175, "global_step": 70449, "epoch": 791, "lr": 8.443682290429228e-05}
{"train_loss": 0.3279963731765747, "global_step": 70450, "epoch": 791, "lr": 8.443640258610668e-05}
{"train_loss": 0.27987468242645264, "global_step": 70451, "epoch": 791, "lr": 8.443598226329154e-05}
{"train_loss": 0.16607064008712769, "global_step": 70452, "epoch": 791, "lr": 8.44355619358469e-05}
{"train_loss": 0.2677120864391327, "global_step": 70453, "epoch": 791, "lr": 8.443514160377282e-05}
{"train_loss": 0.26798468828201294, "global_step": 70454, "epoch": 791, "lr": 8.443472126706934e-05}
{"train_loss": 0.19374284148216248, "global_step": 70455, "epoch": 791, "lr": 8.443430092573654e-05}
{"train_loss": 0.2311219722032547, "global_step": 70456, "epoch": 791, "lr": 8.443388057977446e-05}
{"train_loss": 0.33200421929359436, "global_step": 70457, "epoch": 791, "lr": 8.443346022918318e-05}
{"train_loss": 0.288131982088089, "global_step": 70458, "epoch": 791, "lr": 8.443303987396274e-05}
{"train_loss": 0.3144141137599945, "global_step": 70459, "epoch": 791, "lr": 8.443261951411318e-05}
{"train_loss": 0.28860941529273987, "global_step": 70460, "epoch": 791, "lr": 8.443219914963459e-05}
{"train_loss": 0.23284906148910522, "global_step": 70461, "epoch": 791, "lr": 8.443177878052701e-05}
{"train_loss": 0.27379387617111206, "global_step": 70462, "epoch": 791, "lr": 8.44313584067905e-05}
{"train_loss": 0.23236149549484253, "global_step": 70463, "epoch": 791, "lr": 8.443093802842509e-05}
{"train_loss": 0.1796298623085022, "global_step": 70464, "epoch": 791, "lr": 8.443051764543089e-05}
{"train_loss": 0.268319696187973, "global_step": 70465, "epoch": 791, "lr": 8.443009725780792e-05}
{"train_loss": 0.23929300904273987, "global_step": 70466, "epoch": 791, "lr": 8.442967686555623e-05}
{"train_loss": 0.17663899064064026, "global_step": 70467, "epoch": 791, "lr": 8.44292564686759e-05}
{"train_loss": 0.1900951862335205, "global_step": 70468, "epoch": 791, "lr": 8.442883606716697e-05}
{"train_loss": 0.14682529866695404, "global_step": 70469, "epoch": 791, "lr": 8.442841566102951e-05}
{"train_loss": 0.16565605998039246, "global_step": 70470, "epoch": 791, "lr": 8.442799525026355e-05}
{"train_loss": 0.21148540079593658, "global_step": 70471, "epoch": 791, "lr": 8.442757483486918e-05}
{"train_loss": 0.25446826219558716, "global_step": 70472, "epoch": 791, "lr": 8.442715441484645e-05}
{"train_loss": 0.23895008862018585, "global_step": 70473, "epoch": 791, "lr": 8.44267339901954e-05}
{"train_loss": 0.22100751101970673, "global_step": 70474, "epoch": 791, "lr": 8.442631356091609e-05}
{"train_loss": 0.23278728127479553, "global_step": 70475, "epoch": 791, "lr": 8.44258931270086e-05}
{"train_loss": 0.24834397435188293, "global_step": 70476, "epoch": 791, "lr": 8.442547268847295e-05}
{"train_loss": 0.24543872475624084, "global_step": 70477, "epoch": 791, "lr": 8.442505224530923e-05}
{"train_loss": 0.20346969366073608, "global_step": 70478, "epoch": 791, "lr": 8.442463179751746e-05}
{"train_loss": 0.29745587706565857, "global_step": 70479, "epoch": 791, "lr": 8.442421134509773e-05}
{"train_loss": 0.2949015498161316, "global_step": 70480, "epoch": 791, "lr": 8.442379088805008e-05}
{"train_loss": 0.2454807311296463, "global_step": 70481, "epoch": 791, "lr": 8.442337042637458e-05}
{"train_loss": 0.1898680180311203, "global_step": 70482, "epoch": 791, "lr": 8.442294996007127e-05}
{"train_loss": 0.2466718852519989, "global_step": 70483, "epoch": 791, "lr": 8.442252948914022e-05}
{"train_loss": 0.2025107741355896, "global_step": 70484, "epoch": 791, "lr": 8.442210901358147e-05}
{"train_loss": 0.24553139507770538, "global_step": 70485, "epoch": 791, "lr": 8.44216885333951e-05}
{"train_loss": 0.2545247972011566, "global_step": 70486, "epoch": 791, "lr": 8.442126804858114e-05}
{"train_loss": 0.225703426375148, "global_step": 70487, "epoch": 791, "lr": 8.442084755913967e-05, "val_loss": 3.6745128631591797}
{"train_loss": 0.1732245236635208, "global_step": 70488, "epoch": 792, "lr": 8.442042706507073e-05}
{"train_loss": 0.19018402695655823, "global_step": 70489, "epoch": 792, "lr": 8.442000656637439e-05}
{"train_loss": 0.2628771662712097, "global_step": 70490, "epoch": 792, "lr": 8.441958606305069e-05}
{"train_loss": 0.23568782210350037, "global_step": 70491, "epoch": 792, "lr": 8.44191655550997e-05}
{"train_loss": 0.21085256338119507, "global_step": 70492, "epoch": 792, "lr": 8.44187450425215e-05}
{"train_loss": 0.277560830116272, "global_step": 70493, "epoch": 792, "lr": 8.441832452531609e-05}
{"train_loss": 0.14974375069141388, "global_step": 70494, "epoch": 792, "lr": 8.441790400348355e-05}
{"train_loss": 0.24691519141197205, "global_step": 70495, "epoch": 792, "lr": 8.441748347702396e-05}
{"train_loss": 0.18141110241413116, "global_step": 70496, "epoch": 792, "lr": 8.441706294593735e-05}
{"train_loss": 0.26163485646247864, "global_step": 70497, "epoch": 792, "lr": 8.441664241022378e-05}
{"train_loss": 0.1297006756067276, "global_step": 70498, "epoch": 792, "lr": 8.441622186988333e-05}
{"train_loss": 0.2037988305091858, "global_step": 70499, "epoch": 792, "lr": 8.441580132491602e-05}
{"train_loss": 0.1521422564983368, "global_step": 70500, "epoch": 792, "lr": 8.441538077532194e-05}
{"train_loss": 0.23684559762477875, "global_step": 70501, "epoch": 792, "lr": 8.441496022110111e-05}
{"train_loss": 0.21046674251556396, "global_step": 70502, "epoch": 792, "lr": 8.441453966225363e-05}
{"train_loss": 0.1415872573852539, "global_step": 70503, "epoch": 792, "lr": 8.441411909877953e-05}
{"train_loss": 0.20514559745788574, "global_step": 70504, "epoch": 792, "lr": 8.441369853067886e-05}
{"train_loss": 0.21942256391048431, "global_step": 70505, "epoch": 792, "lr": 8.441327795795169e-05}
{"train_loss": 0.1936475932598114, "global_step": 70506, "epoch": 792, "lr": 8.441285738059808e-05}
{"train_loss": 0.1639033704996109, "global_step": 70507, "epoch": 792, "lr": 8.441243679861809e-05}
{"train_loss": 0.15217067301273346, "global_step": 70508, "epoch": 792, "lr": 8.441201621201176e-05}
{"train_loss": 0.3141512870788574, "global_step": 70509, "epoch": 792, "lr": 8.441159562077915e-05}
{"train_loss": 0.27105459570884705, "global_step": 70510, "epoch": 792, "lr": 8.441117502492031e-05}
{"train_loss": 0.2731640338897705, "global_step": 70511, "epoch": 792, "lr": 8.441075442443533e-05}
{"train_loss": 0.19818061590194702, "global_step": 70512, "epoch": 792, "lr": 8.441033381932423e-05}
{"train_loss": 0.14161188900470734, "global_step": 70513, "epoch": 792, "lr": 8.440991320958709e-05}
{"train_loss": 0.16885942220687866, "global_step": 70514, "epoch": 792, "lr": 8.440949259522395e-05}
{"train_loss": 0.33359628915786743, "global_step": 70515, "epoch": 792, "lr": 8.440907197623487e-05}
{"train_loss": 0.19414275884628296, "global_step": 70516, "epoch": 792, "lr": 8.440865135261991e-05}
{"train_loss": 0.14152516424655914, "global_step": 70517, "epoch": 792, "lr": 8.440823072437913e-05}
{"train_loss": 0.2452775537967682, "global_step": 70518, "epoch": 792, "lr": 8.440781009151257e-05}
{"train_loss": 0.17197071015834808, "global_step": 70519, "epoch": 792, "lr": 8.440738945402031e-05}
{"train_loss": 0.192476287484169, "global_step": 70520, "epoch": 792, "lr": 8.44069688119024e-05}
{"train_loss": 0.14722216129302979, "global_step": 70521, "epoch": 792, "lr": 8.44065481651589e-05}
{"train_loss": 0.24904581904411316, "global_step": 70522, "epoch": 792, "lr": 8.440612751378984e-05}
{"train_loss": 0.2147539108991623, "global_step": 70523, "epoch": 792, "lr": 8.44057068577953e-05}
{"train_loss": 0.21791592240333557, "global_step": 70524, "epoch": 792, "lr": 8.440528619717534e-05}
{"train_loss": 0.16722115874290466, "global_step": 70525, "epoch": 792, "lr": 8.440486553193e-05}
{"train_loss": 0.17456060647964478, "global_step": 70526, "epoch": 792, "lr": 8.440444486205934e-05}
{"train_loss": 0.23150435090065002, "global_step": 70527, "epoch": 792, "lr": 8.440402418756344e-05}
{"train_loss": 0.25424087047576904, "global_step": 70528, "epoch": 792, "lr": 8.440360350844232e-05}
{"train_loss": 0.1925291270017624, "global_step": 70529, "epoch": 792, "lr": 8.440318282469607e-05}
{"train_loss": 0.18924914300441742, "global_step": 70530, "epoch": 792, "lr": 8.440276213632472e-05}
{"train_loss": 0.14987854659557343, "global_step": 70531, "epoch": 792, "lr": 8.440234144332835e-05}
{"train_loss": 0.15766289830207825, "global_step": 70532, "epoch": 792, "lr": 8.4401920745707e-05}
{"train_loss": 0.16689640283584595, "global_step": 70533, "epoch": 792, "lr": 8.440150004346072e-05}
{"train_loss": 0.22445975244045258, "global_step": 70534, "epoch": 792, "lr": 8.440107933658958e-05}
{"train_loss": 0.17204752564430237, "global_step": 70535, "epoch": 792, "lr": 8.440065862509365e-05}
{"train_loss": 0.24522359669208527, "global_step": 70536, "epoch": 792, "lr": 8.440023790897296e-05}
{"train_loss": 0.19119161367416382, "global_step": 70537, "epoch": 792, "lr": 8.439981718822758e-05}
{"train_loss": 0.2205188125371933, "global_step": 70538, "epoch": 792, "lr": 8.439939646285756e-05}
{"train_loss": 0.22585974633693695, "global_step": 70539, "epoch": 792, "lr": 8.439897573286297e-05}
{"train_loss": 0.21817229688167572, "global_step": 70540, "epoch": 792, "lr": 8.439855499824386e-05}
{"train_loss": 0.19070447981357574, "global_step": 70541, "epoch": 792, "lr": 8.439813425900025e-05}
{"train_loss": 0.2375386357307434, "global_step": 70542, "epoch": 792, "lr": 8.439771351513226e-05}
{"train_loss": 0.21853995323181152, "global_step": 70543, "epoch": 792, "lr": 8.43972927666399e-05}
{"train_loss": 0.2314884513616562, "global_step": 70544, "epoch": 792, "lr": 8.439687201352326e-05}
{"train_loss": 0.23539312183856964, "global_step": 70545, "epoch": 792, "lr": 8.439645125578237e-05}
{"train_loss": 0.24979715049266815, "global_step": 70546, "epoch": 792, "lr": 8.43960304934173e-05}
{"train_loss": 0.2273872345685959, "global_step": 70547, "epoch": 792, "lr": 8.43956097264281e-05}
{"train_loss": 0.16216464340686798, "global_step": 70548, "epoch": 792, "lr": 8.439518895481484e-05}
{"train_loss": 0.2170846313238144, "global_step": 70549, "epoch": 792, "lr": 8.439476817857754e-05}
{"train_loss": 0.20592093467712402, "global_step": 70550, "epoch": 792, "lr": 8.439434739771633e-05}
{"train_loss": 0.17924559116363525, "global_step": 70551, "epoch": 792, "lr": 8.439392661223118e-05}
{"train_loss": 0.1975087672472, "global_step": 70552, "epoch": 792, "lr": 8.439350582212219e-05}
{"train_loss": 0.1692848801612854, "global_step": 70553, "epoch": 792, "lr": 8.439308502738942e-05}
{"train_loss": 0.20415498316287994, "global_step": 70554, "epoch": 792, "lr": 8.439266422803292e-05}
{"train_loss": 0.2235737144947052, "global_step": 70555, "epoch": 792, "lr": 8.439224342405274e-05}
{"train_loss": 0.23621819913387299, "global_step": 70556, "epoch": 792, "lr": 8.439182261544893e-05}
{"train_loss": 0.20867101848125458, "global_step": 70557, "epoch": 792, "lr": 8.439140180222158e-05}
{"train_loss": 0.19484347105026245, "global_step": 70558, "epoch": 792, "lr": 8.439098098437073e-05}
{"train_loss": 0.2370007038116455, "global_step": 70559, "epoch": 792, "lr": 8.43905601618964e-05}
{"train_loss": 0.22907398641109467, "global_step": 70560, "epoch": 792, "lr": 8.43901393347987e-05}
{"train_loss": 0.16793005168437958, "global_step": 70561, "epoch": 792, "lr": 8.438971850307767e-05}
{"train_loss": 0.17448388040065765, "global_step": 70562, "epoch": 792, "lr": 8.438929766673335e-05}
{"train_loss": 0.2724320888519287, "global_step": 70563, "epoch": 792, "lr": 8.438887682576582e-05}
{"train_loss": 0.17396152019500732, "global_step": 70564, "epoch": 792, "lr": 8.43884559801751e-05}
{"train_loss": 0.1557568907737732, "global_step": 70565, "epoch": 792, "lr": 8.43880351299613e-05}
{"train_loss": 0.22384755313396454, "global_step": 70566, "epoch": 792, "lr": 8.438761427512442e-05}
{"train_loss": 0.15898330509662628, "global_step": 70567, "epoch": 792, "lr": 8.438719341566456e-05}
{"train_loss": 0.18548035621643066, "global_step": 70568, "epoch": 792, "lr": 8.438677255158178e-05}
{"train_loss": 0.20251679420471191, "global_step": 70569, "epoch": 792, "lr": 8.43863516828761e-05}
{"train_loss": 0.2553337514400482, "global_step": 70570, "epoch": 792, "lr": 8.43859308095476e-05}
{"train_loss": 0.17927920818328857, "global_step": 70571, "epoch": 792, "lr": 8.438550993159631e-05}
{"train_loss": 0.22482776641845703, "global_step": 70572, "epoch": 792, "lr": 8.438508904902232e-05}
{"train_loss": 0.11022830754518509, "global_step": 70573, "epoch": 792, "lr": 8.438466816182568e-05}
{"train_loss": 0.17255637049674988, "global_step": 70574, "epoch": 792, "lr": 8.438424727000644e-05}
{"train_loss": 0.218293696641922, "global_step": 70575, "epoch": 792, "lr": 8.438382637356465e-05}
{"train_loss": 0.20407238209180617, "global_step": 70576, "epoch": 792, "lr": 8.438340547250038e-05, "val_loss": 3.841261386871338}
{"train_loss": 0.1957249492406845, "global_step": 70577, "epoch": 793, "lr": 8.438298456681366e-05}
{"train_loss": 0.19667887687683105, "global_step": 70578, "epoch": 793, "lr": 8.438256365650459e-05}
{"train_loss": 0.09460821747779846, "global_step": 70579, "epoch": 793, "lr": 8.438214274157319e-05}
{"train_loss": 0.19595371186733246, "global_step": 70580, "epoch": 793, "lr": 8.438172182201955e-05}
{"train_loss": 0.177954763174057, "global_step": 70581, "epoch": 793, "lr": 8.438130089784369e-05}
{"train_loss": 0.15220096707344055, "global_step": 70582, "epoch": 793, "lr": 8.438087996904568e-05}
{"train_loss": 0.17655742168426514, "global_step": 70583, "epoch": 793, "lr": 8.43804590356256e-05}
{"train_loss": 0.18081077933311462, "global_step": 70584, "epoch": 793, "lr": 8.438003809758345e-05}
{"train_loss": 0.20292408764362335, "global_step": 70585, "epoch": 793, "lr": 8.437961715491936e-05}
{"train_loss": 0.323551744222641, "global_step": 70586, "epoch": 793, "lr": 8.437919620763332e-05}
{"train_loss": 0.1738819181919098, "global_step": 70587, "epoch": 793, "lr": 8.437877525572544e-05}
{"train_loss": 0.24573113024234772, "global_step": 70588, "epoch": 793, "lr": 8.437835429919573e-05}
{"train_loss": 0.23716230690479279, "global_step": 70589, "epoch": 793, "lr": 8.43779333380443e-05}
{"train_loss": 0.16362035274505615, "global_step": 70590, "epoch": 793, "lr": 8.437751237227117e-05}
{"train_loss": 0.21770650148391724, "global_step": 70591, "epoch": 793, "lr": 8.437709140187638e-05}
{"train_loss": 0.21937060356140137, "global_step": 70592, "epoch": 793, "lr": 8.437667042686002e-05}
{"train_loss": 0.17459428310394287, "global_step": 70593, "epoch": 793, "lr": 8.437624944722214e-05}
{"train_loss": 0.23412606120109558, "global_step": 70594, "epoch": 793, "lr": 8.437582846296279e-05}
{"train_loss": 0.21310363709926605, "global_step": 70595, "epoch": 793, "lr": 8.437540747408201e-05}
{"train_loss": 0.1769985556602478, "global_step": 70596, "epoch": 793, "lr": 8.43749864805799e-05}
{"train_loss": 0.2900315225124359, "global_step": 70597, "epoch": 793, "lr": 8.437456548245648e-05}
{"train_loss": 0.19500061869621277, "global_step": 70598, "epoch": 793, "lr": 8.437414447971182e-05}
{"train_loss": 0.15598487854003906, "global_step": 70599, "epoch": 793, "lr": 8.437372347234598e-05}
{"train_loss": 0.18100042641162872, "global_step": 70600, "epoch": 793, "lr": 8.437330246035902e-05}
{"train_loss": 0.2570672035217285, "global_step": 70601, "epoch": 793, "lr": 8.437288144375097e-05}
{"train_loss": 0.1943725049495697, "global_step": 70602, "epoch": 793, "lr": 8.437246042252191e-05}
{"train_loss": 0.17381319403648376, "global_step": 70603, "epoch": 793, "lr": 8.43720393966719e-05}
{"train_loss": 0.2831490635871887, "global_step": 70604, "epoch": 793, "lr": 8.437161836620097e-05}
{"train_loss": 0.2667233645915985, "global_step": 70605, "epoch": 793, "lr": 8.437119733110923e-05}
{"train_loss": 0.18903949856758118, "global_step": 70606, "epoch": 793, "lr": 8.437077629139666e-05}
{"train_loss": 0.18236926198005676, "global_step": 70607, "epoch": 793, "lr": 8.43703552470634e-05}
{"train_loss": 0.2567777633666992, "global_step": 70608, "epoch": 793, "lr": 8.436993419810943e-05}
{"train_loss": 0.315061092376709, "global_step": 70609, "epoch": 793, "lr": 8.436951314453486e-05}
{"train_loss": 0.17059603333473206, "global_step": 70610, "epoch": 793, "lr": 8.436909208633972e-05}
{"train_loss": 0.18696755170822144, "global_step": 70611, "epoch": 793, "lr": 8.436867102352409e-05}
{"train_loss": 0.16817694902420044, "global_step": 70612, "epoch": 793, "lr": 8.436824995608799e-05}
{"train_loss": 0.18540513515472412, "global_step": 70613, "epoch": 793, "lr": 8.43678288840315e-05}
{"train_loss": 0.2941901683807373, "global_step": 70614, "epoch": 793, "lr": 8.436740780735467e-05}
{"train_loss": 0.1454138308763504, "global_step": 70615, "epoch": 793, "lr": 8.436698672605758e-05}
{"train_loss": 0.23322691023349762, "global_step": 70616, "epoch": 793, "lr": 8.436656564014025e-05}
{"train_loss": 0.20316347479820251, "global_step": 70617, "epoch": 793, "lr": 8.436614454960277e-05}
{"train_loss": 0.20382066071033478, "global_step": 70618, "epoch": 793, "lr": 8.436572345444516e-05}
{"train_loss": 0.18480616807937622, "global_step": 70619, "epoch": 793, "lr": 8.43653023546675e-05}
{"train_loss": 0.2050093561410904, "global_step": 70620, "epoch": 793, "lr": 8.436488125026987e-05}
{"train_loss": 0.15421229600906372, "global_step": 70621, "epoch": 793, "lr": 8.436446014125228e-05}
{"train_loss": 0.27212899923324585, "global_step": 70622, "epoch": 793, "lr": 8.43640390276148e-05}
{"train_loss": 0.2519201338291168, "global_step": 70623, "epoch": 793, "lr": 8.436361790935752e-05}
{"train_loss": 0.1975553333759308, "global_step": 70624, "epoch": 793, "lr": 8.436319678648045e-05}
{"train_loss": 0.17443612217903137, "global_step": 70625, "epoch": 793, "lr": 8.436277565898367e-05}
{"train_loss": 0.20276634395122528, "global_step": 70626, "epoch": 793, "lr": 8.436235452686724e-05}
{"train_loss": 0.22607631981372833, "global_step": 70627, "epoch": 793, "lr": 8.43619333901312e-05}
{"train_loss": 0.19240814447402954, "global_step": 70628, "epoch": 793, "lr": 8.436151224877564e-05}
{"train_loss": 0.16603955626487732, "global_step": 70629, "epoch": 793, "lr": 8.436109110280057e-05}
{"train_loss": 0.2592073082923889, "global_step": 70630, "epoch": 793, "lr": 8.436066995220608e-05}
{"train_loss": 0.1705145239830017, "global_step": 70631, "epoch": 793, "lr": 8.436024879699222e-05}
{"train_loss": 0.20354196429252625, "global_step": 70632, "epoch": 793, "lr": 8.435982763715901e-05}
{"train_loss": 0.2368101179599762, "global_step": 70633, "epoch": 793, "lr": 8.435940647270658e-05}
{"train_loss": 0.26007047295570374, "global_step": 70634, "epoch": 793, "lr": 8.435898530363493e-05}
{"train_loss": 0.22068548202514648, "global_step": 70635, "epoch": 793, "lr": 8.435856412994414e-05}
{"train_loss": 0.14906291663646698, "global_step": 70636, "epoch": 793, "lr": 8.435814295163426e-05}
{"train_loss": 0.31851038336753845, "global_step": 70637, "epoch": 793, "lr": 8.435772176870534e-05}
{"train_loss": 0.22901217639446259, "global_step": 70638, "epoch": 793, "lr": 8.435730058115745e-05}
{"train_loss": 0.21930524706840515, "global_step": 70639, "epoch": 793, "lr": 8.435687938899064e-05}
{"train_loss": 0.20939812064170837, "global_step": 70640, "epoch": 793, "lr": 8.435645819220495e-05}
{"train_loss": 0.2143058180809021, "global_step": 70641, "epoch": 793, "lr": 8.435603699080047e-05}
{"train_loss": 0.2579263150691986, "global_step": 70642, "epoch": 793, "lr": 8.435561578477724e-05}
{"train_loss": 0.3039942979812622, "global_step": 70643, "epoch": 793, "lr": 8.43551945741353e-05}
{"train_loss": 0.2410055249929428, "global_step": 70644, "epoch": 793, "lr": 8.435477335887474e-05}
{"train_loss": 0.22094540297985077, "global_step": 70645, "epoch": 793, "lr": 8.435435213899559e-05}
{"train_loss": 0.20029649138450623, "global_step": 70646, "epoch": 793, "lr": 8.435393091449792e-05}
{"train_loss": 0.19731193780899048, "global_step": 70647, "epoch": 793, "lr": 8.435350968538179e-05}
{"train_loss": 0.227955162525177, "global_step": 70648, "epoch": 793, "lr": 8.435308845164725e-05}
{"train_loss": 0.2499961405992508, "global_step": 70649, "epoch": 793, "lr": 8.435266721329433e-05}
{"train_loss": 0.22587813436985016, "global_step": 70650, "epoch": 793, "lr": 8.435224597032312e-05}
{"train_loss": 0.17660494148731232, "global_step": 70651, "epoch": 793, "lr": 8.435182472273369e-05}
{"train_loss": 0.23211362957954407, "global_step": 70652, "epoch": 793, "lr": 8.435140347052605e-05}
{"train_loss": 0.24902985990047455, "global_step": 70653, "epoch": 793, "lr": 8.43509822137003e-05}
{"train_loss": 0.26883918046951294, "global_step": 70654, "epoch": 793, "lr": 8.435056095225649e-05}
{"train_loss": 0.27431508898735046, "global_step": 70655, "epoch": 793, "lr": 8.435013968619465e-05}
{"train_loss": 0.23309150338172913, "global_step": 70656, "epoch": 793, "lr": 8.434971841551485e-05}
{"train_loss": 0.2463485598564148, "global_step": 70657, "epoch": 793, "lr": 8.434929714021715e-05}
{"train_loss": 0.17649514973163605, "global_step": 70658, "epoch": 793, "lr": 8.434887586030163e-05}
{"train_loss": 0.2268519401550293, "global_step": 70659, "epoch": 793, "lr": 8.434845457576829e-05}
{"train_loss": 0.2930651605129242, "global_step": 70660, "epoch": 793, "lr": 8.434803328661722e-05}
{"train_loss": 0.2747703194618225, "global_step": 70661, "epoch": 793, "lr": 8.43476119928485e-05}
{"train_loss": 0.2862788438796997, "global_step": 70662, "epoch": 793, "lr": 8.434719069446214e-05}
{"train_loss": 0.21736840903759003, "global_step": 70663, "epoch": 793, "lr": 8.434676939145823e-05}
{"train_loss": 0.2459680587053299, "global_step": 70664, "epoch": 793, "lr": 8.434634808383681e-05}
{"train_loss": 0.21679659606365675, "global_step": 70665, "epoch": 793, "lr": 8.434592677159795e-05, "val_loss": 3.7428390979766846}
{"train_loss": 0.25414901971817017, "global_step": 70666, "epoch": 794, "lr": 8.43455054547417e-05}
{"train_loss": 0.24470579624176025, "global_step": 70667, "epoch": 794, "lr": 8.434508413326811e-05}
{"train_loss": 0.17751987278461456, "global_step": 70668, "epoch": 794, "lr": 8.434466280717723e-05}
{"train_loss": 0.2127874195575714, "global_step": 70669, "epoch": 794, "lr": 8.434424147646913e-05}
{"train_loss": 0.24547497928142548, "global_step": 70670, "epoch": 794, "lr": 8.434382014114388e-05}
{"train_loss": 0.21787945926189423, "global_step": 70671, "epoch": 794, "lr": 8.434339880120152e-05}
{"train_loss": 0.22302046418190002, "global_step": 70672, "epoch": 794, "lr": 8.43429774566421e-05}
{"train_loss": 0.25799474120140076, "global_step": 70673, "epoch": 794, "lr": 8.43425561074657e-05}
{"train_loss": 0.2533497214317322, "global_step": 70674, "epoch": 794, "lr": 8.434213475367236e-05}
{"train_loss": 0.17125901579856873, "global_step": 70675, "epoch": 794, "lr": 8.434171339526212e-05}
{"train_loss": 0.24204584956169128, "global_step": 70676, "epoch": 794, "lr": 8.434129203223506e-05}
{"train_loss": 0.17687194049358368, "global_step": 70677, "epoch": 794, "lr": 8.434087066459124e-05}
{"train_loss": 0.17853222787380219, "global_step": 70678, "epoch": 794, "lr": 8.434044929233071e-05}
{"train_loss": 0.2861468195915222, "global_step": 70679, "epoch": 794, "lr": 8.434002791545353e-05}
{"train_loss": 0.27152228355407715, "global_step": 70680, "epoch": 794, "lr": 8.433960653395974e-05}
{"train_loss": 0.3142458200454712, "global_step": 70681, "epoch": 794, "lr": 8.433918514784941e-05}
{"train_loss": 0.23041152954101562, "global_step": 70682, "epoch": 794, "lr": 8.43387637571226e-05}
{"train_loss": 0.23595724999904633, "global_step": 70683, "epoch": 794, "lr": 8.433834236177935e-05}
{"train_loss": 0.205066978931427, "global_step": 70684, "epoch": 794, "lr": 8.433792096181973e-05}
{"train_loss": 0.24508576095104218, "global_step": 70685, "epoch": 794, "lr": 8.433749955724381e-05}
{"train_loss": 0.27625903487205505, "global_step": 70686, "epoch": 794, "lr": 8.433707814805163e-05}
{"train_loss": 0.1590307354927063, "global_step": 70687, "epoch": 794, "lr": 8.433665673424324e-05}
{"train_loss": 0.20602965354919434, "global_step": 70688, "epoch": 794, "lr": 8.433623531581871e-05}
{"train_loss": 0.16170074045658112, "global_step": 70689, "epoch": 794, "lr": 8.43358138927781e-05}
{"train_loss": 0.14609938859939575, "global_step": 70690, "epoch": 794, "lr": 8.433539246512143e-05}
{"train_loss": 0.18383097648620605, "global_step": 70691, "epoch": 794, "lr": 8.43349710328488e-05}
{"train_loss": 0.21194303035736084, "global_step": 70692, "epoch": 794, "lr": 8.433454959596025e-05}
{"train_loss": 0.16453838348388672, "global_step": 70693, "epoch": 794, "lr": 8.433412815445585e-05}
{"train_loss": 0.17738652229309082, "global_step": 70694, "epoch": 794, "lr": 8.433370670833564e-05}
{"train_loss": 0.1691693216562271, "global_step": 70695, "epoch": 794, "lr": 8.433328525759968e-05}
{"train_loss": 0.23980966210365295, "global_step": 70696, "epoch": 794, "lr": 8.433286380224802e-05}
{"train_loss": 0.21286408603191376, "global_step": 70697, "epoch": 794, "lr": 8.433244234228074e-05}
{"train_loss": 0.23306556046009064, "global_step": 70698, "epoch": 794, "lr": 8.433202087769786e-05}
{"train_loss": 0.20532427728176117, "global_step": 70699, "epoch": 794, "lr": 8.433159940849948e-05}
{"train_loss": 0.11047381907701492, "global_step": 70700, "epoch": 794, "lr": 8.433117793468564e-05}
{"train_loss": 0.26119890809059143, "global_step": 70701, "epoch": 794, "lr": 8.433075645625636e-05}
{"train_loss": 0.21176207065582275, "global_step": 70702, "epoch": 794, "lr": 8.433033497321175e-05}
{"train_loss": 0.22476816177368164, "global_step": 70703, "epoch": 794, "lr": 8.432991348555185e-05}
{"train_loss": 0.18954253196716309, "global_step": 70704, "epoch": 794, "lr": 8.43294919932767e-05}
{"train_loss": 0.24660620093345642, "global_step": 70705, "epoch": 794, "lr": 8.432907049638637e-05}
{"train_loss": 0.24528512358665466, "global_step": 70706, "epoch": 794, "lr": 8.43286489948809e-05}
{"train_loss": 0.22271782159805298, "global_step": 70707, "epoch": 794, "lr": 8.43282274887604e-05}
{"train_loss": 0.190491184592247, "global_step": 70708, "epoch": 794, "lr": 8.432780597802486e-05}
{"train_loss": 0.2037094086408615, "global_step": 70709, "epoch": 794, "lr": 8.432738446267436e-05}
{"train_loss": 0.22619670629501343, "global_step": 70710, "epoch": 794, "lr": 8.432696294270899e-05}
{"train_loss": 0.22948728501796722, "global_step": 70711, "epoch": 794, "lr": 8.432654141812875e-05}
{"train_loss": 0.16262289881706238, "global_step": 70712, "epoch": 794, "lr": 8.432611988893375e-05}
{"train_loss": 0.3059178590774536, "global_step": 70713, "epoch": 794, "lr": 8.4325698355124e-05}
{"train_loss": 0.18542008101940155, "global_step": 70714, "epoch": 794, "lr": 8.432527681669958e-05}
{"train_loss": 0.16847150027751923, "global_step": 70715, "epoch": 794, "lr": 8.432485527366056e-05}
{"train_loss": 0.14415527880191803, "global_step": 70716, "epoch": 794, "lr": 8.432443372600697e-05}
{"train_loss": 0.3442152738571167, "global_step": 70717, "epoch": 794, "lr": 8.432401217373888e-05}
{"train_loss": 0.22484318912029266, "global_step": 70718, "epoch": 794, "lr": 8.432359061685634e-05}
{"train_loss": 0.1591498851776123, "global_step": 70719, "epoch": 794, "lr": 8.432316905535942e-05}
{"train_loss": 0.19663500785827637, "global_step": 70720, "epoch": 794, "lr": 8.432274748924816e-05}
{"train_loss": 0.24296239018440247, "global_step": 70721, "epoch": 794, "lr": 8.432232591852262e-05}
{"train_loss": 0.23491258919239044, "global_step": 70722, "epoch": 794, "lr": 8.432190434318288e-05}
{"train_loss": 0.28276488184928894, "global_step": 70723, "epoch": 794, "lr": 8.432148276322896e-05}
{"train_loss": 0.27774810791015625, "global_step": 70724, "epoch": 794, "lr": 8.432106117866095e-05}
{"train_loss": 0.15668262541294098, "global_step": 70725, "epoch": 794, "lr": 8.432063958947889e-05}
{"train_loss": 0.20994028449058533, "global_step": 70726, "epoch": 794, "lr": 8.432021799568283e-05}
{"train_loss": 0.24324952065944672, "global_step": 70727, "epoch": 794, "lr": 8.431979639727284e-05}
{"train_loss": 0.18314822018146515, "global_step": 70728, "epoch": 794, "lr": 8.431937479424897e-05}
{"train_loss": 0.22090794146060944, "global_step": 70729, "epoch": 794, "lr": 8.431895318661128e-05}
{"train_loss": 0.20847921073436737, "global_step": 70730, "epoch": 794, "lr": 8.431853157435982e-05}
{"train_loss": 0.20250916481018066, "global_step": 70731, "epoch": 794, "lr": 8.431810995749468e-05}
{"train_loss": 0.19925475120544434, "global_step": 70732, "epoch": 794, "lr": 8.431768833601586e-05}
{"train_loss": 0.18303067982196808, "global_step": 70733, "epoch": 794, "lr": 8.431726670992343e-05}
{"train_loss": 0.16538041830062866, "global_step": 70734, "epoch": 794, "lr": 8.431684507921748e-05}
{"train_loss": 0.17251868546009064, "global_step": 70735, "epoch": 794, "lr": 8.431642344389807e-05}
{"train_loss": 0.3096570372581482, "global_step": 70736, "epoch": 794, "lr": 8.43160018039652e-05}
{"train_loss": 0.16664227843284607, "global_step": 70737, "epoch": 794, "lr": 8.431558015941898e-05}
{"train_loss": 0.1630973368883133, "global_step": 70738, "epoch": 794, "lr": 8.431515851025945e-05}
{"train_loss": 0.2544044554233551, "global_step": 70739, "epoch": 794, "lr": 8.431473685648665e-05}
{"train_loss": 0.19366878271102905, "global_step": 70740, "epoch": 794, "lr": 8.431431519810067e-05}
{"train_loss": 0.2093125283718109, "global_step": 70741, "epoch": 794, "lr": 8.431389353510154e-05}
{"train_loss": 0.24833127856254578, "global_step": 70742, "epoch": 794, "lr": 8.431347186748933e-05}
{"train_loss": 0.15479432046413422, "global_step": 70743, "epoch": 794, "lr": 8.431305019526407e-05}
{"train_loss": 0.27214714884757996, "global_step": 70744, "epoch": 794, "lr": 8.431262851842588e-05}
{"train_loss": 0.22531327605247498, "global_step": 70745, "epoch": 794, "lr": 8.431220683697474e-05}
{"train_loss": 0.26910194754600525, "global_step": 70746, "epoch": 794, "lr": 8.431178515091076e-05}
{"train_loss": 0.26235976815223694, "global_step": 70747, "epoch": 794, "lr": 8.431136346023396e-05}
{"train_loss": 0.19973091781139374, "global_step": 70748, "epoch": 794, "lr": 8.431094176494443e-05}
{"train_loss": 0.14802157878875732, "global_step": 70749, "epoch": 794, "lr": 8.431052006504222e-05}
{"train_loss": 0.22709254920482635, "global_step": 70750, "epoch": 794, "lr": 8.431009836052736e-05}
{"train_loss": 0.11984074115753174, "global_step": 70751, "epoch": 794, "lr": 8.430967665139995e-05}
{"train_loss": 0.16642454266548157, "global_step": 70752, "epoch": 794, "lr": 8.430925493766e-05}
{"train_loss": 0.22300460934638977, "global_step": 70753, "epoch": 794, "lr": 8.43088332193076e-05}
{"train_loss": 0.2138272360636947, "global_step": 70754, "epoch": 794, "lr": 8.430841149634279e-05, "val_loss": 3.9383299350738525}
{"train_loss": 0.3670315444469452, "global_step": 70755, "epoch": 795, "lr": 8.430798976876564e-05}
{"train_loss": 0.301476389169693, "global_step": 70756, "epoch": 795, "lr": 8.430756803657618e-05}
{"train_loss": 0.26651501655578613, "global_step": 70757, "epoch": 795, "lr": 8.430714629977449e-05}
{"train_loss": 0.32489973306655884, "global_step": 70758, "epoch": 795, "lr": 8.430672455836065e-05}
{"train_loss": 0.21905961632728577, "global_step": 70759, "epoch": 795, "lr": 8.430630281233464e-05}
{"train_loss": 0.24148143827915192, "global_step": 70760, "epoch": 795, "lr": 8.430588106169661e-05}
{"train_loss": 0.2881540358066559, "global_step": 70761, "epoch": 795, "lr": 8.430545930644655e-05}
{"train_loss": 0.2097163051366806, "global_step": 70762, "epoch": 795, "lr": 8.430503754658455e-05}
{"train_loss": 0.18199990689754486, "global_step": 70763, "epoch": 795, "lr": 8.430461578211066e-05}
{"train_loss": 0.30813610553741455, "global_step": 70764, "epoch": 795, "lr": 8.430419401302492e-05}
{"train_loss": 0.27190715074539185, "global_step": 70765, "epoch": 795, "lr": 8.430377223932741e-05}
{"train_loss": 0.1784851998090744, "global_step": 70766, "epoch": 795, "lr": 8.430335046101817e-05}
{"train_loss": 0.2027820348739624, "global_step": 70767, "epoch": 795, "lr": 8.430292867809725e-05}
{"train_loss": 0.26019495725631714, "global_step": 70768, "epoch": 795, "lr": 8.430250689056474e-05}
{"train_loss": 0.1802579015493393, "global_step": 70769, "epoch": 795, "lr": 8.430208509842065e-05}
{"train_loss": 0.12929075956344604, "global_step": 70770, "epoch": 795, "lr": 8.43016633016651e-05}
{"train_loss": 0.1962965577840805, "global_step": 70771, "epoch": 795, "lr": 8.430124150029809e-05}
{"train_loss": 0.1817748099565506, "global_step": 70772, "epoch": 795, "lr": 8.430081969431969e-05}
{"train_loss": 0.2899579405784607, "global_step": 70773, "epoch": 795, "lr": 8.430039788372998e-05}
{"train_loss": 0.25565797090530396, "global_step": 70774, "epoch": 795, "lr": 8.429997606852898e-05}
{"train_loss": 0.22994686663150787, "global_step": 70775, "epoch": 795, "lr": 8.429955424871679e-05}
{"train_loss": 0.22024230659008026, "global_step": 70776, "epoch": 795, "lr": 8.429913242429341e-05}
{"train_loss": 0.19267885386943817, "global_step": 70777, "epoch": 795, "lr": 8.429871059525896e-05}
{"train_loss": 0.2983431816101074, "global_step": 70778, "epoch": 795, "lr": 8.429828876161344e-05}
{"train_loss": 0.3271580934524536, "global_step": 70779, "epoch": 795, "lr": 8.429786692335695e-05}
{"train_loss": 0.16873478889465332, "global_step": 70780, "epoch": 795, "lr": 8.429744508048954e-05}
{"train_loss": 0.31803154945373535, "global_step": 70781, "epoch": 795, "lr": 8.429702323301123e-05}
{"train_loss": 0.38413363695144653, "global_step": 70782, "epoch": 795, "lr": 8.429660138092212e-05}
{"train_loss": 0.17914614081382751, "global_step": 70783, "epoch": 795, "lr": 8.429617952422224e-05}
{"train_loss": 0.1605520248413086, "global_step": 70784, "epoch": 795, "lr": 8.429575766291166e-05}
{"train_loss": 0.22846712172031403, "global_step": 70785, "epoch": 795, "lr": 8.429533579699045e-05}
{"train_loss": 0.14895598590373993, "global_step": 70786, "epoch": 795, "lr": 8.429491392645863e-05}
{"train_loss": 0.12406574934720993, "global_step": 70787, "epoch": 795, "lr": 8.42944920513163e-05}
{"train_loss": 0.23537865281105042, "global_step": 70788, "epoch": 795, "lr": 8.429407017156345e-05}
{"train_loss": 0.2304658442735672, "global_step": 70789, "epoch": 795, "lr": 8.429364828720022e-05}
{"train_loss": 0.20987780392169952, "global_step": 70790, "epoch": 795, "lr": 8.429322639822661e-05}
{"train_loss": 0.26069942116737366, "global_step": 70791, "epoch": 795, "lr": 8.429280450464271e-05}
{"train_loss": 0.1970815658569336, "global_step": 70792, "epoch": 795, "lr": 8.429238260644854e-05}
{"train_loss": 0.1366722285747528, "global_step": 70793, "epoch": 795, "lr": 8.42919607036442e-05}
{"train_loss": 0.22190339863300323, "global_step": 70794, "epoch": 795, "lr": 8.42915387962297e-05}
{"train_loss": 0.265904039144516, "global_step": 70795, "epoch": 795, "lr": 8.429111688420512e-05}
{"train_loss": 0.2150879204273224, "global_step": 70796, "epoch": 795, "lr": 8.429069496757053e-05}
{"train_loss": 0.22052457928657532, "global_step": 70797, "epoch": 795, "lr": 8.429027304632597e-05}
{"train_loss": 0.13514021039009094, "global_step": 70798, "epoch": 795, "lr": 8.428985112047151e-05}
{"train_loss": 0.21103422343730927, "global_step": 70799, "epoch": 795, "lr": 8.428942919000719e-05}
{"train_loss": 0.23285317420959473, "global_step": 70800, "epoch": 795, "lr": 8.428900725493307e-05}
{"train_loss": 0.19749268889427185, "global_step": 70801, "epoch": 795, "lr": 8.428858531524921e-05}
{"train_loss": 0.232594296336174, "global_step": 70802, "epoch": 795, "lr": 8.428816337095567e-05}
{"train_loss": 0.22492870688438416, "global_step": 70803, "epoch": 795, "lr": 8.428774142205251e-05}
{"train_loss": 0.20268529653549194, "global_step": 70804, "epoch": 795, "lr": 8.428731946853979e-05}
{"train_loss": 0.3234386444091797, "global_step": 70805, "epoch": 795, "lr": 8.428689751041753e-05}
{"train_loss": 0.14530715346336365, "global_step": 70806, "epoch": 795, "lr": 8.428647554768582e-05}
{"train_loss": 0.2915758490562439, "global_step": 70807, "epoch": 795, "lr": 8.428605358034474e-05}
{"train_loss": 0.2032628059387207, "global_step": 70808, "epoch": 795, "lr": 8.42856316083943e-05}
{"train_loss": 0.23456089198589325, "global_step": 70809, "epoch": 795, "lr": 8.428520963183457e-05}
{"train_loss": 0.20847977697849274, "global_step": 70810, "epoch": 795, "lr": 8.428478765066562e-05}
{"train_loss": 0.2065197229385376, "global_step": 70811, "epoch": 795, "lr": 8.428436566488748e-05}
{"train_loss": 0.16530846059322357, "global_step": 70812, "epoch": 795, "lr": 8.428394367450026e-05}
{"train_loss": 0.22593750059604645, "global_step": 70813, "epoch": 795, "lr": 8.428352167950394e-05}
{"train_loss": 0.21147236227989197, "global_step": 70814, "epoch": 795, "lr": 8.428309967989865e-05}
{"train_loss": 0.14210295677185059, "global_step": 70815, "epoch": 795, "lr": 8.42826776756844e-05}
{"train_loss": 0.2638336420059204, "global_step": 70816, "epoch": 795, "lr": 8.428225566686126e-05}
{"train_loss": 0.19410915672779083, "global_step": 70817, "epoch": 795, "lr": 8.42818336534293e-05}
{"train_loss": 0.14777347445487976, "global_step": 70818, "epoch": 795, "lr": 8.428141163538856e-05}
{"train_loss": 0.18594439327716827, "global_step": 70819, "epoch": 795, "lr": 8.428098961273909e-05}
{"train_loss": 0.18013139069080353, "global_step": 70820, "epoch": 795, "lr": 8.428056758548097e-05}
{"train_loss": 0.1931028962135315, "global_step": 70821, "epoch": 795, "lr": 8.428014555361426e-05}
{"train_loss": 0.2691877484321594, "global_step": 70822, "epoch": 795, "lr": 8.427972351713898e-05}
{"train_loss": 0.2522312104701996, "global_step": 70823, "epoch": 795, "lr": 8.427930147605522e-05}
{"train_loss": 0.2592927813529968, "global_step": 70824, "epoch": 795, "lr": 8.427887943036303e-05}
{"train_loss": 0.1788628101348877, "global_step": 70825, "epoch": 795, "lr": 8.427845738006244e-05}
{"train_loss": 0.15091203153133392, "global_step": 70826, "epoch": 795, "lr": 8.427803532515355e-05}
{"train_loss": 0.1806473582983017, "global_step": 70827, "epoch": 795, "lr": 8.427761326563639e-05}
{"train_loss": 0.21875713765621185, "global_step": 70828, "epoch": 795, "lr": 8.427719120151102e-05}
{"train_loss": 0.2268139123916626, "global_step": 70829, "epoch": 795, "lr": 8.42767691327775e-05}
{"train_loss": 0.16042721271514893, "global_step": 70830, "epoch": 795, "lr": 8.42763470594359e-05}
{"train_loss": 0.19668076932430267, "global_step": 70831, "epoch": 795, "lr": 8.427592498148624e-05}
{"train_loss": 0.12468086183071136, "global_step": 70832, "epoch": 795, "lr": 8.42755028989286e-05}
{"train_loss": 0.29676297307014465, "global_step": 70833, "epoch": 795, "lr": 8.427508081176306e-05}
{"train_loss": 0.2206142097711563, "global_step": 70834, "epoch": 795, "lr": 8.427465871998963e-05}
{"train_loss": 0.2581004500389099, "global_step": 70835, "epoch": 795, "lr": 8.42742366236084e-05}
{"train_loss": 0.28085795044898987, "global_step": 70836, "epoch": 795, "lr": 8.427381452261943e-05}
{"train_loss": 0.2874268591403961, "global_step": 70837, "epoch": 795, "lr": 8.427339241702274e-05}
{"train_loss": 0.20930975675582886, "global_step": 70838, "epoch": 795, "lr": 8.427297030681842e-05}
{"train_loss": 0.24670922756195068, "global_step": 70839, "epoch": 795, "lr": 8.427254819200653e-05}
{"train_loss": 0.18735651671886444, "global_step": 70840, "epoch": 795, "lr": 8.42721260725871e-05}
{"train_loss": 0.22960618138313293, "global_step": 70841, "epoch": 795, "lr": 8.42717039485602e-05}
{"train_loss": 0.175512433052063, "global_step": 70842, "epoch": 795, "lr": 8.42712818199259e-05}
{"train_loss": 0.22181249694542937, "global_step": 70843, "epoch": 795, "lr": 8.427085968668422e-05, "val_loss": 3.7477986812591553, "train_action_mse_error": 17.392724990844727}
{"train_loss": 0.2338458001613617, "global_step": 70844, "epoch": 796, "lr": 8.427043754883526e-05}
{"train_loss": 0.21678315103054047, "global_step": 70845, "epoch": 796, "lr": 8.427001540637904e-05}
{"train_loss": 0.2600609362125397, "global_step": 70846, "epoch": 796, "lr": 8.426959325931564e-05}
{"train_loss": 0.22028562426567078, "global_step": 70847, "epoch": 796, "lr": 8.426917110764513e-05}
{"train_loss": 0.17317311465740204, "global_step": 70848, "epoch": 796, "lr": 8.426874895136753e-05}
{"train_loss": 0.2274358868598938, "global_step": 70849, "epoch": 796, "lr": 8.426832679048291e-05}
{"train_loss": 0.36168569326400757, "global_step": 70850, "epoch": 796, "lr": 8.426790462499134e-05}
{"train_loss": 0.12755464017391205, "global_step": 70851, "epoch": 796, "lr": 8.426748245489286e-05}
{"train_loss": 0.15158188343048096, "global_step": 70852, "epoch": 796, "lr": 8.426706028018754e-05}
{"train_loss": 0.20540031790733337, "global_step": 70853, "epoch": 796, "lr": 8.426663810087544e-05}
{"train_loss": 0.21336320042610168, "global_step": 70854, "epoch": 796, "lr": 8.42662159169566e-05}
{"train_loss": 0.20311740040779114, "global_step": 70855, "epoch": 796, "lr": 8.426579372843109e-05}
{"train_loss": 0.284999817609787, "global_step": 70856, "epoch": 796, "lr": 8.426537153529895e-05}
{"train_loss": 0.35488471388816833, "global_step": 70857, "epoch": 796, "lr": 8.426494933756026e-05}
{"train_loss": 0.2834630012512207, "global_step": 70858, "epoch": 796, "lr": 8.426452713521507e-05}
{"train_loss": 0.20329028367996216, "global_step": 70859, "epoch": 796, "lr": 8.426410492826342e-05}
{"train_loss": 0.17688888311386108, "global_step": 70860, "epoch": 796, "lr": 8.426368271670537e-05}
{"train_loss": 0.17082884907722473, "global_step": 70861, "epoch": 796, "lr": 8.426326050054101e-05}
{"train_loss": 0.15119972825050354, "global_step": 70862, "epoch": 796, "lr": 8.426283827977035e-05}
{"train_loss": 0.19341535866260529, "global_step": 70863, "epoch": 796, "lr": 8.426241605439348e-05}
{"train_loss": 0.20339354872703552, "global_step": 70864, "epoch": 796, "lr": 8.426199382441045e-05}
{"train_loss": 0.19914211332798004, "global_step": 70865, "epoch": 796, "lr": 8.42615715898213e-05}
{"train_loss": 0.31649115681648254, "global_step": 70866, "epoch": 796, "lr": 8.42611493506261e-05}
{"train_loss": 0.22549982368946075, "global_step": 70867, "epoch": 796, "lr": 8.426072710682493e-05}
{"train_loss": 0.2056661695241928, "global_step": 70868, "epoch": 796, "lr": 8.426030485841778e-05}
{"train_loss": 0.19813020527362823, "global_step": 70869, "epoch": 796, "lr": 8.425988260540478e-05}
{"train_loss": 0.23052114248275757, "global_step": 70870, "epoch": 796, "lr": 8.425946034778595e-05}
{"train_loss": 0.2708714008331299, "global_step": 70871, "epoch": 796, "lr": 8.425903808556135e-05}
{"train_loss": 0.3109050989151001, "global_step": 70872, "epoch": 796, "lr": 8.425861581873104e-05}
{"train_loss": 0.24605758488178253, "global_step": 70873, "epoch": 796, "lr": 8.42581935472951e-05}
{"train_loss": 0.2612401247024536, "global_step": 70874, "epoch": 796, "lr": 8.425777127125353e-05}
{"train_loss": 0.16748064756393433, "global_step": 70875, "epoch": 796, "lr": 8.425734899060645e-05}
{"train_loss": 0.2153243124485016, "global_step": 70876, "epoch": 796, "lr": 8.425692670535385e-05}
{"train_loss": 0.21934419870376587, "global_step": 70877, "epoch": 796, "lr": 8.425650441549586e-05}
{"train_loss": 0.1638902872800827, "global_step": 70878, "epoch": 796, "lr": 8.425608212103248e-05}
{"train_loss": 0.2331245094537735, "global_step": 70879, "epoch": 796, "lr": 8.425565982196379e-05}
{"train_loss": 0.17776647210121155, "global_step": 70880, "epoch": 796, "lr": 8.425523751828985e-05}
{"train_loss": 0.19766494631767273, "global_step": 70881, "epoch": 796, "lr": 8.425481521001071e-05}
{"train_loss": 0.2520836293697357, "global_step": 70882, "epoch": 796, "lr": 8.425439289712642e-05}
{"train_loss": 0.18737578392028809, "global_step": 70883, "epoch": 796, "lr": 8.425397057963704e-05}
{"train_loss": 0.20653985440731049, "global_step": 70884, "epoch": 796, "lr": 8.425354825754266e-05}
{"train_loss": 0.23545219004154205, "global_step": 70885, "epoch": 796, "lr": 8.425312593084328e-05}
{"train_loss": 0.21194395422935486, "global_step": 70886, "epoch": 796, "lr": 8.425270359953899e-05}
{"train_loss": 0.2212805300951004, "global_step": 70887, "epoch": 796, "lr": 8.425228126362984e-05}
{"train_loss": 0.2058250904083252, "global_step": 70888, "epoch": 796, "lr": 8.425185892311589e-05}
{"train_loss": 0.1767645925283432, "global_step": 70889, "epoch": 796, "lr": 8.42514365779972e-05}
{"train_loss": 0.31927409768104553, "global_step": 70890, "epoch": 796, "lr": 8.425101422827382e-05}
{"train_loss": 0.12778134644031525, "global_step": 70891, "epoch": 796, "lr": 8.425059187394581e-05}
{"train_loss": 0.16279254853725433, "global_step": 70892, "epoch": 796, "lr": 8.425016951501322e-05}
{"train_loss": 0.19554510712623596, "global_step": 70893, "epoch": 796, "lr": 8.424974715147612e-05}
{"train_loss": 0.20031017065048218, "global_step": 70894, "epoch": 796, "lr": 8.424932478333456e-05}
{"train_loss": 0.17440824210643768, "global_step": 70895, "epoch": 796, "lr": 8.424890241058858e-05}
{"train_loss": 0.2462560534477234, "global_step": 70896, "epoch": 796, "lr": 8.424848003323827e-05}
{"train_loss": 0.21454569697380066, "global_step": 70897, "epoch": 796, "lr": 8.424805765128366e-05}
{"train_loss": 0.23687000572681427, "global_step": 70898, "epoch": 796, "lr": 8.424763526472483e-05}
{"train_loss": 0.27188342809677124, "global_step": 70899, "epoch": 796, "lr": 8.42472128735618e-05}
{"train_loss": 0.27994513511657715, "global_step": 70900, "epoch": 796, "lr": 8.424679047779466e-05}
{"train_loss": 0.17188005149364471, "global_step": 70901, "epoch": 796, "lr": 8.424636807742346e-05}
{"train_loss": 0.20463748276233673, "global_step": 70902, "epoch": 796, "lr": 8.424594567244826e-05}
{"train_loss": 0.2307928055524826, "global_step": 70903, "epoch": 796, "lr": 8.42455232628691e-05}
{"train_loss": 0.18503405153751373, "global_step": 70904, "epoch": 796, "lr": 8.424510084868603e-05}
{"train_loss": 0.20097976922988892, "global_step": 70905, "epoch": 796, "lr": 8.424467842989913e-05}
{"train_loss": 0.16210903227329254, "global_step": 70906, "epoch": 796, "lr": 8.424425600650847e-05}
{"train_loss": 0.2458583414554596, "global_step": 70907, "epoch": 796, "lr": 8.424383357851409e-05}
{"train_loss": 0.2590576112270355, "global_step": 70908, "epoch": 796, "lr": 8.424341114591603e-05}
{"train_loss": 0.17998342216014862, "global_step": 70909, "epoch": 796, "lr": 8.424298870871437e-05}
{"train_loss": 0.20226159691810608, "global_step": 70910, "epoch": 796, "lr": 8.424256626690913e-05}
{"train_loss": 0.15407870709896088, "global_step": 70911, "epoch": 796, "lr": 8.424214382050041e-05}
{"train_loss": 0.23412348330020905, "global_step": 70912, "epoch": 796, "lr": 8.424172136948827e-05}
{"train_loss": 0.18679828941822052, "global_step": 70913, "epoch": 796, "lr": 8.424129891387273e-05}
{"train_loss": 0.15766313672065735, "global_step": 70914, "epoch": 796, "lr": 8.424087645365388e-05}
{"train_loss": 0.25179043412208557, "global_step": 70915, "epoch": 796, "lr": 8.424045398883175e-05}
{"train_loss": 0.23885607719421387, "global_step": 70916, "epoch": 796, "lr": 8.424003151940641e-05}
{"train_loss": 0.19312922656536102, "global_step": 70917, "epoch": 796, "lr": 8.423960904537791e-05}
{"train_loss": 0.15035121142864227, "global_step": 70918, "epoch": 796, "lr": 8.423918656674632e-05}
{"train_loss": 0.22109706699848175, "global_step": 70919, "epoch": 796, "lr": 8.423876408351167e-05}
{"train_loss": 0.24246147274971008, "global_step": 70920, "epoch": 796, "lr": 8.423834159567406e-05}
{"train_loss": 0.27551335096359253, "global_step": 70921, "epoch": 796, "lr": 8.42379191032335e-05}
{"train_loss": 0.26933005452156067, "global_step": 70922, "epoch": 796, "lr": 8.423749660619009e-05}
{"train_loss": 0.17342624068260193, "global_step": 70923, "epoch": 796, "lr": 8.423707410454384e-05}
{"train_loss": 0.21484191715717316, "global_step": 70924, "epoch": 796, "lr": 8.423665159829486e-05}
{"train_loss": 0.20021013915538788, "global_step": 70925, "epoch": 796, "lr": 8.423622908744317e-05}
{"train_loss": 0.1672828197479248, "global_step": 70926, "epoch": 796, "lr": 8.423580657198884e-05}
{"train_loss": 0.1735846847295761, "global_step": 70927, "epoch": 796, "lr": 8.423538405193192e-05}
{"train_loss": 0.17236201465129852, "global_step": 70928, "epoch": 796, "lr": 8.423496152727246e-05}
{"train_loss": 0.20172469317913055, "global_step": 70929, "epoch": 796, "lr": 8.423453899801054e-05}
{"train_loss": 0.28747326135635376, "global_step": 70930, "epoch": 796, "lr": 8.42341164641462e-05}
{"train_loss": 0.3336338400840759, "global_step": 70931, "epoch": 796, "lr": 8.42336939256795e-05}
{"train_loss": 0.21667483683382527, "global_step": 70932, "epoch": 796, "lr": 8.423327138261051e-05, "val_loss": 3.84236741065979}
{"train_loss": 0.18068915605545044, "global_step": 70933, "epoch": 797, "lr": 8.423284883493924e-05}
{"train_loss": 0.23596997559070587, "global_step": 70934, "epoch": 797, "lr": 8.423242628266581e-05}
{"train_loss": 0.21407848596572876, "global_step": 70935, "epoch": 797, "lr": 8.423200372579023e-05}
{"train_loss": 0.2093721330165863, "global_step": 70936, "epoch": 797, "lr": 8.423158116431258e-05}
{"train_loss": 0.13914431631565094, "global_step": 70937, "epoch": 797, "lr": 8.423115859823292e-05}
{"train_loss": 0.17872163653373718, "global_step": 70938, "epoch": 797, "lr": 8.42307360275513e-05}
{"train_loss": 0.19587960839271545, "global_step": 70939, "epoch": 797, "lr": 8.423031345226775e-05}
{"train_loss": 0.1707654893398285, "global_step": 70940, "epoch": 797, "lr": 8.422989087238237e-05}
{"train_loss": 0.17006179690361023, "global_step": 70941, "epoch": 797, "lr": 8.422946828789519e-05}
{"train_loss": 0.14502188563346863, "global_step": 70942, "epoch": 797, "lr": 8.422904569880628e-05}
{"train_loss": 0.270138144493103, "global_step": 70943, "epoch": 797, "lr": 8.42286231051157e-05}
{"train_loss": 0.16794615983963013, "global_step": 70944, "epoch": 797, "lr": 8.42282005068235e-05}
{"train_loss": 0.12276899814605713, "global_step": 70945, "epoch": 797, "lr": 8.422777790392971e-05}
{"train_loss": 0.22831839323043823, "global_step": 70946, "epoch": 797, "lr": 8.422735529643444e-05}
{"train_loss": 0.24153798818588257, "global_step": 70947, "epoch": 797, "lr": 8.42269326843377e-05}
{"train_loss": 0.26680517196655273, "global_step": 70948, "epoch": 797, "lr": 8.422651006763958e-05}
{"train_loss": 0.24060650169849396, "global_step": 70949, "epoch": 797, "lr": 8.422608744634012e-05}
{"train_loss": 0.1870410293340683, "global_step": 70950, "epoch": 797, "lr": 8.422566482043938e-05}
{"train_loss": 0.26092827320098877, "global_step": 70951, "epoch": 797, "lr": 8.422524218993742e-05}
{"train_loss": 0.14817984402179718, "global_step": 70952, "epoch": 797, "lr": 8.422481955483427e-05}
{"train_loss": 0.19503149390220642, "global_step": 70953, "epoch": 797, "lr": 8.422439691513005e-05}
{"train_loss": 0.27000313997268677, "global_step": 70954, "epoch": 797, "lr": 8.422397427082475e-05}
{"train_loss": 0.27554354071617126, "global_step": 70955, "epoch": 797, "lr": 8.422355162191847e-05}
{"train_loss": 0.19357852637767792, "global_step": 70956, "epoch": 797, "lr": 8.422312896841122e-05}
{"train_loss": 0.25311410427093506, "global_step": 70957, "epoch": 797, "lr": 8.422270631030311e-05}
{"train_loss": 0.2070980817079544, "global_step": 70958, "epoch": 797, "lr": 8.422228364759416e-05}
{"train_loss": 0.14070852100849152, "global_step": 70959, "epoch": 797, "lr": 8.422186098028446e-05}
{"train_loss": 0.20720697939395905, "global_step": 70960, "epoch": 797, "lr": 8.422143830837403e-05}
{"train_loss": 0.133804053068161, "global_step": 70961, "epoch": 797, "lr": 8.422101563186297e-05}
{"train_loss": 0.2162308245897293, "global_step": 70962, "epoch": 797, "lr": 8.422059295075129e-05}
{"train_loss": 0.22055836021900177, "global_step": 70963, "epoch": 797, "lr": 8.422017026503908e-05}
{"train_loss": 0.19266265630722046, "global_step": 70964, "epoch": 797, "lr": 8.421974757472637e-05}
{"train_loss": 0.22221237421035767, "global_step": 70965, "epoch": 797, "lr": 8.421932487981325e-05}
{"train_loss": 0.2063361257314682, "global_step": 70966, "epoch": 797, "lr": 8.421890218029976e-05}
{"train_loss": 0.22771099209785461, "global_step": 70967, "epoch": 797, "lr": 8.421847947618593e-05}
{"train_loss": 0.22045016288757324, "global_step": 70968, "epoch": 797, "lr": 8.421805676747187e-05}
{"train_loss": 0.18306873738765717, "global_step": 70969, "epoch": 797, "lr": 8.421763405415759e-05}
{"train_loss": 0.24335452914237976, "global_step": 70970, "epoch": 797, "lr": 8.421721133624318e-05}
{"train_loss": 0.26769930124282837, "global_step": 70971, "epoch": 797, "lr": 8.421678861372868e-05}
{"train_loss": 0.2657277584075928, "global_step": 70972, "epoch": 797, "lr": 8.421636588661415e-05}
{"train_loss": 0.1455165147781372, "global_step": 70973, "epoch": 797, "lr": 8.421594315489964e-05}
{"train_loss": 0.16193245351314545, "global_step": 70974, "epoch": 797, "lr": 8.421552041858521e-05}
{"train_loss": 0.14886870980262756, "global_step": 70975, "epoch": 797, "lr": 8.421509767767093e-05}
{"train_loss": 0.2852027416229248, "global_step": 70976, "epoch": 797, "lr": 8.421467493215684e-05}
{"train_loss": 0.24302169680595398, "global_step": 70977, "epoch": 797, "lr": 8.421425218204302e-05}
{"train_loss": 0.2765096426010132, "global_step": 70978, "epoch": 797, "lr": 8.42138294273295e-05}
{"train_loss": 0.21406154334545135, "global_step": 70979, "epoch": 797, "lr": 8.421340666801633e-05}
{"train_loss": 0.18061400949954987, "global_step": 70980, "epoch": 797, "lr": 8.421298390410362e-05}
{"train_loss": 0.21832038462162018, "global_step": 70981, "epoch": 797, "lr": 8.421256113559136e-05}
{"train_loss": 0.1578608900308609, "global_step": 70982, "epoch": 797, "lr": 8.421213836247965e-05}
{"train_loss": 0.23914098739624023, "global_step": 70983, "epoch": 797, "lr": 8.421171558476855e-05}
{"train_loss": 0.28184112906455994, "global_step": 70984, "epoch": 797, "lr": 8.421129280245807e-05}
{"train_loss": 0.2850850522518158, "global_step": 70985, "epoch": 797, "lr": 8.421087001554833e-05}
{"train_loss": 0.15221667289733887, "global_step": 70986, "epoch": 797, "lr": 8.421044722403933e-05}
{"train_loss": 0.15349657833576202, "global_step": 70987, "epoch": 797, "lr": 8.421002442793118e-05}
{"train_loss": 0.2253795713186264, "global_step": 70988, "epoch": 797, "lr": 8.420960162722389e-05}
{"train_loss": 0.1972874402999878, "global_step": 70989, "epoch": 797, "lr": 8.420917882191754e-05}
{"train_loss": 0.17899323999881744, "global_step": 70990, "epoch": 797, "lr": 8.420875601201218e-05}
{"train_loss": 0.19652517139911652, "global_step": 70991, "epoch": 797, "lr": 8.420833319750788e-05}
{"train_loss": 0.15694068372249603, "global_step": 70992, "epoch": 797, "lr": 8.420791037840468e-05}
{"train_loss": 0.19641512632369995, "global_step": 70993, "epoch": 797, "lr": 8.420748755470264e-05}
{"train_loss": 0.24315430223941803, "global_step": 70994, "epoch": 797, "lr": 8.420706472640181e-05}
{"train_loss": 0.21507230401039124, "global_step": 70995, "epoch": 797, "lr": 8.420664189350227e-05}
{"train_loss": 0.3063909411430359, "global_step": 70996, "epoch": 797, "lr": 8.420621905600405e-05}
{"train_loss": 0.2934439480304718, "global_step": 70997, "epoch": 797, "lr": 8.420579621390725e-05}
{"train_loss": 0.19569391012191772, "global_step": 70998, "epoch": 797, "lr": 8.420537336721189e-05}
{"train_loss": 0.17378462851047516, "global_step": 70999, "epoch": 797, "lr": 8.420495051591802e-05}
{"train_loss": 0.222142294049263, "global_step": 71000, "epoch": 797, "lr": 8.420452766002571e-05}
{"train_loss": 0.27657902240753174, "global_step": 71001, "epoch": 797, "lr": 8.420410479953503e-05}
{"train_loss": 0.2327696532011032, "global_step": 71002, "epoch": 797, "lr": 8.420368193444601e-05}
{"train_loss": 0.17500421404838562, "global_step": 71003, "epoch": 797, "lr": 8.420325906475875e-05}
{"train_loss": 0.15002645552158356, "global_step": 71004, "epoch": 797, "lr": 8.420283619047325e-05}
{"train_loss": 0.2976611852645874, "global_step": 71005, "epoch": 797, "lr": 8.42024133115896e-05}
{"train_loss": 0.22507968544960022, "global_step": 71006, "epoch": 797, "lr": 8.420199042810787e-05}
{"train_loss": 0.21320953965187073, "global_step": 71007, "epoch": 797, "lr": 8.420156754002808e-05}
{"train_loss": 0.23647883534431458, "global_step": 71008, "epoch": 797, "lr": 8.420114464735032e-05}
{"train_loss": 0.12797969579696655, "global_step": 71009, "epoch": 797, "lr": 8.420072175007462e-05}
{"train_loss": 0.25787508487701416, "global_step": 71010, "epoch": 797, "lr": 8.420029884820107e-05}
{"train_loss": 0.18097183108329773, "global_step": 71011, "epoch": 797, "lr": 8.419987594172968e-05}
{"train_loss": 0.22176285088062286, "global_step": 71012, "epoch": 797, "lr": 8.419945303066056e-05}
{"train_loss": 0.2678852081298828, "global_step": 71013, "epoch": 797, "lr": 8.419903011499372e-05}
{"train_loss": 0.2748532295227051, "global_step": 71014, "epoch": 797, "lr": 8.419860719472925e-05}
{"train_loss": 0.16387461125850677, "global_step": 71015, "epoch": 797, "lr": 8.419818426986718e-05}
{"train_loss": 0.24771928787231445, "global_step": 71016, "epoch": 797, "lr": 8.41977613404076e-05}
{"train_loss": 0.22626207768917084, "global_step": 71017, "epoch": 797, "lr": 8.419733840635053e-05}
{"train_loss": 0.142863929271698, "global_step": 71018, "epoch": 797, "lr": 8.419691546769605e-05}
{"train_loss": 0.1755031943321228, "global_step": 71019, "epoch": 797, "lr": 8.419649252444422e-05}
{"train_loss": 0.19137607514858246, "global_step": 71020, "epoch": 797, "lr": 8.419606957659509e-05}
{"train_loss": 0.21181097492743073, "global_step": 71021, "epoch": 797, "lr": 8.419564662414869e-05, "val_loss": 3.767958402633667}
{"train_loss": 0.16955401003360748, "global_step": 71022, "epoch": 798, "lr": 8.419522366710513e-05}
{"train_loss": 0.17847511172294617, "global_step": 71023, "epoch": 798, "lr": 8.419480070546443e-05}
{"train_loss": 0.2733208239078522, "global_step": 71024, "epoch": 798, "lr": 8.419437773922664e-05}
{"train_loss": 0.16149520874023438, "global_step": 71025, "epoch": 798, "lr": 8.419395476839186e-05}
{"train_loss": 0.20272710919380188, "global_step": 71026, "epoch": 798, "lr": 8.41935317929601e-05}
{"train_loss": 0.2389272302389145, "global_step": 71027, "epoch": 798, "lr": 8.419310881293146e-05}
{"train_loss": 0.2149723768234253, "global_step": 71028, "epoch": 798, "lr": 8.419268582830595e-05}
{"train_loss": 0.2074318528175354, "global_step": 71029, "epoch": 798, "lr": 8.419226283908365e-05}
{"train_loss": 0.22225749492645264, "global_step": 71030, "epoch": 798, "lr": 8.419183984526463e-05}
{"train_loss": 0.27975112199783325, "global_step": 71031, "epoch": 798, "lr": 8.419141684684893e-05}
{"train_loss": 0.16117043793201447, "global_step": 71032, "epoch": 798, "lr": 8.419099384383659e-05}
{"train_loss": 0.2421121448278427, "global_step": 71033, "epoch": 798, "lr": 8.419057083622771e-05}
{"train_loss": 0.20610100030899048, "global_step": 71034, "epoch": 798, "lr": 8.419014782402233e-05}
{"train_loss": 0.2277810126543045, "global_step": 71035, "epoch": 798, "lr": 8.41897248072205e-05}
{"train_loss": 0.21398741006851196, "global_step": 71036, "epoch": 798, "lr": 8.418930178582226e-05}
{"train_loss": 0.18342511355876923, "global_step": 71037, "epoch": 798, "lr": 8.41888787598277e-05}
{"train_loss": 0.22367912530899048, "global_step": 71038, "epoch": 798, "lr": 8.418845572923685e-05}
{"train_loss": 0.1999415010213852, "global_step": 71039, "epoch": 798, "lr": 8.418803269404978e-05}
{"train_loss": 0.17960777878761292, "global_step": 71040, "epoch": 798, "lr": 8.418760965426655e-05}
{"train_loss": 0.20131725072860718, "global_step": 71041, "epoch": 798, "lr": 8.418718660988722e-05}
{"train_loss": 0.2283017635345459, "global_step": 71042, "epoch": 798, "lr": 8.418676356091184e-05}
{"train_loss": 0.262762188911438, "global_step": 71043, "epoch": 798, "lr": 8.418634050734046e-05}
{"train_loss": 0.14127355813980103, "global_step": 71044, "epoch": 798, "lr": 8.418591744917313e-05}
{"train_loss": 0.15760774910449982, "global_step": 71045, "epoch": 798, "lr": 8.418549438640995e-05}
{"train_loss": 0.17656321823596954, "global_step": 71046, "epoch": 798, "lr": 8.41850713190509e-05}
{"train_loss": 0.17491509020328522, "global_step": 71047, "epoch": 798, "lr": 8.418464824709613e-05}
{"train_loss": 0.24013245105743408, "global_step": 71048, "epoch": 798, "lr": 8.418422517054565e-05}
{"train_loss": 0.24750177562236786, "global_step": 71049, "epoch": 798, "lr": 8.418380208939949e-05}
{"train_loss": 0.16181927919387817, "global_step": 71050, "epoch": 798, "lr": 8.418337900365774e-05}
{"train_loss": 0.2171454131603241, "global_step": 71051, "epoch": 798, "lr": 8.418295591332047e-05}
{"train_loss": 0.2013651430606842, "global_step": 71052, "epoch": 798, "lr": 8.418253281838769e-05}
{"train_loss": 0.15278801321983337, "global_step": 71053, "epoch": 798, "lr": 8.418210971885951e-05}
{"train_loss": 0.32773569226264954, "global_step": 71054, "epoch": 798, "lr": 8.418168661473596e-05}
{"train_loss": 0.15810690820217133, "global_step": 71055, "epoch": 798, "lr": 8.418126350601708e-05}
{"train_loss": 0.2749606966972351, "global_step": 71056, "epoch": 798, "lr": 8.418084039270296e-05}
{"train_loss": 0.19945600628852844, "global_step": 71057, "epoch": 798, "lr": 8.418041727479364e-05}
{"train_loss": 0.18219657242298126, "global_step": 71058, "epoch": 798, "lr": 8.417999415228919e-05}
{"train_loss": 0.18019205331802368, "global_step": 71059, "epoch": 798, "lr": 8.417957102518964e-05}
{"train_loss": 0.17766201496124268, "global_step": 71060, "epoch": 798, "lr": 8.417914789349507e-05}
{"train_loss": 0.16259132325649261, "global_step": 71061, "epoch": 798, "lr": 8.417872475720552e-05}
{"train_loss": 0.23995518684387207, "global_step": 71062, "epoch": 798, "lr": 8.417830161632108e-05}
{"train_loss": 0.2685403525829315, "global_step": 71063, "epoch": 798, "lr": 8.417787847084177e-05}
{"train_loss": 0.1529318392276764, "global_step": 71064, "epoch": 798, "lr": 8.417745532076767e-05}
{"train_loss": 0.2630191445350647, "global_step": 71065, "epoch": 798, "lr": 8.41770321660988e-05}
{"train_loss": 0.20420794188976288, "global_step": 71066, "epoch": 798, "lr": 8.417660900683527e-05}
{"train_loss": 0.1331411898136139, "global_step": 71067, "epoch": 798, "lr": 8.41761858429771e-05}
{"train_loss": 0.15869663655757904, "global_step": 71068, "epoch": 798, "lr": 8.417576267452436e-05}
{"train_loss": 0.17762638628482819, "global_step": 71069, "epoch": 798, "lr": 8.417533950147712e-05}
{"train_loss": 0.20465776324272156, "global_step": 71070, "epoch": 798, "lr": 8.417491632383541e-05}
{"train_loss": 0.2426266074180603, "global_step": 71071, "epoch": 798, "lr": 8.417449314159929e-05}
{"train_loss": 0.19201909005641937, "global_step": 71072, "epoch": 798, "lr": 8.417406995476884e-05}
{"train_loss": 0.22652912139892578, "global_step": 71073, "epoch": 798, "lr": 8.417364676334409e-05}
{"train_loss": 0.308028906583786, "global_step": 71074, "epoch": 798, "lr": 8.417322356732513e-05}
{"train_loss": 0.21318461000919342, "global_step": 71075, "epoch": 798, "lr": 8.417280036671197e-05}
{"train_loss": 0.2322944700717926, "global_step": 71076, "epoch": 798, "lr": 8.417237716150473e-05}
{"train_loss": 0.21097682416439056, "global_step": 71077, "epoch": 798, "lr": 8.41719539517034e-05}
{"train_loss": 0.16473694145679474, "global_step": 71078, "epoch": 798, "lr": 8.417153073730808e-05}
{"train_loss": 0.25154730677604675, "global_step": 71079, "epoch": 798, "lr": 8.41711075183188e-05}
{"train_loss": 0.329410195350647, "global_step": 71080, "epoch": 798, "lr": 8.417068429473565e-05}
{"train_loss": 0.23589184880256653, "global_step": 71081, "epoch": 798, "lr": 8.417026106655866e-05}
{"train_loss": 0.2199847400188446, "global_step": 71082, "epoch": 798, "lr": 8.416983783378789e-05}
{"train_loss": 0.21391870081424713, "global_step": 71083, "epoch": 798, "lr": 8.41694145964234e-05}
{"train_loss": 0.18941932916641235, "global_step": 71084, "epoch": 798, "lr": 8.416899135446526e-05}
{"train_loss": 0.20489946007728577, "global_step": 71085, "epoch": 798, "lr": 8.416856810791352e-05}
{"train_loss": 0.20865175127983093, "global_step": 71086, "epoch": 798, "lr": 8.416814485676821e-05}
{"train_loss": 0.3132968246936798, "global_step": 71087, "epoch": 798, "lr": 8.416772160102941e-05}
{"train_loss": 0.28879672288894653, "global_step": 71088, "epoch": 798, "lr": 8.41672983406972e-05}
{"train_loss": 0.22332462668418884, "global_step": 71089, "epoch": 798, "lr": 8.41668750757716e-05}
{"train_loss": 0.1808972954750061, "global_step": 71090, "epoch": 798, "lr": 8.416645180625268e-05}
{"train_loss": 0.20658789575099945, "global_step": 71091, "epoch": 798, "lr": 8.416602853214051e-05}
{"train_loss": 0.21457795798778534, "global_step": 71092, "epoch": 798, "lr": 8.41656052534351e-05}
{"train_loss": 0.22558769583702087, "global_step": 71093, "epoch": 798, "lr": 8.416518197013658e-05}
{"train_loss": 0.19855770468711853, "global_step": 71094, "epoch": 798, "lr": 8.416475868224496e-05}
{"train_loss": 0.19620977342128754, "global_step": 71095, "epoch": 798, "lr": 8.416433538976028e-05}
{"train_loss": 0.25842639803886414, "global_step": 71096, "epoch": 798, "lr": 8.416391209268265e-05}
{"train_loss": 0.2336377650499344, "global_step": 71097, "epoch": 798, "lr": 8.416348879101207e-05}
{"train_loss": 0.21116916835308075, "global_step": 71098, "epoch": 798, "lr": 8.416306548474865e-05}
{"train_loss": 0.17407307028770447, "global_step": 71099, "epoch": 798, "lr": 8.416264217389242e-05}
{"train_loss": 0.197811558842659, "global_step": 71100, "epoch": 798, "lr": 8.416221885844342e-05}
{"train_loss": 0.1858501136302948, "global_step": 71101, "epoch": 798, "lr": 8.416179553840174e-05}
{"train_loss": 0.18469905853271484, "global_step": 71102, "epoch": 798, "lr": 8.416137221376743e-05}
{"train_loss": 0.18263599276542664, "global_step": 71103, "epoch": 798, "lr": 8.416094888454053e-05}
{"train_loss": 0.29274412989616394, "global_step": 71104, "epoch": 798, "lr": 8.416052555072109e-05}
{"train_loss": 0.18596017360687256, "global_step": 71105, "epoch": 798, "lr": 8.416010221230921e-05}
{"train_loss": 0.3635271489620209, "global_step": 71106, "epoch": 798, "lr": 8.415967886930492e-05}
{"train_loss": 0.2291475236415863, "global_step": 71107, "epoch": 798, "lr": 8.415925552170826e-05}
{"train_loss": 0.26427948474884033, "global_step": 71108, "epoch": 798, "lr": 8.415883216951932e-05}
{"train_loss": 0.20619025826454163, "global_step": 71109, "epoch": 798, "lr": 8.415840881273813e-05}
{"train_loss": 0.21481829195210103, "global_step": 71110, "epoch": 798, "lr": 8.415798545136477e-05, "val_loss": 3.833662271499634}
{"train_loss": 0.2525928020477295, "global_step": 71111, "epoch": 799, "lr": 8.415756208539928e-05}
{"train_loss": 0.25393375754356384, "global_step": 71112, "epoch": 799, "lr": 8.415713871484172e-05}
{"train_loss": 0.26832228899002075, "global_step": 71113, "epoch": 799, "lr": 8.415671533969214e-05}
{"train_loss": 0.2279880791902542, "global_step": 71114, "epoch": 799, "lr": 8.415629195995063e-05}
{"train_loss": 0.18529820442199707, "global_step": 71115, "epoch": 799, "lr": 8.415586857561722e-05}
{"train_loss": 0.2381841391324997, "global_step": 71116, "epoch": 799, "lr": 8.415544518669194e-05}
{"train_loss": 0.18282075226306915, "global_step": 71117, "epoch": 799, "lr": 8.415502179317489e-05}
{"train_loss": 0.23549482226371765, "global_step": 71118, "epoch": 799, "lr": 8.415459839506613e-05}
{"train_loss": 0.1759946346282959, "global_step": 71119, "epoch": 799, "lr": 8.415417499236568e-05}
{"train_loss": 0.21351458132266998, "global_step": 71120, "epoch": 799, "lr": 8.415375158507364e-05}
{"train_loss": 0.24598602950572968, "global_step": 71121, "epoch": 799, "lr": 8.415332817319003e-05}
{"train_loss": 0.3077271580696106, "global_step": 71122, "epoch": 799, "lr": 8.415290475671492e-05}
{"train_loss": 0.1817583590745926, "global_step": 71123, "epoch": 799, "lr": 8.415248133564837e-05}
{"train_loss": 0.18258942663669586, "global_step": 71124, "epoch": 799, "lr": 8.415205790999043e-05}
{"train_loss": 0.21908335387706757, "global_step": 71125, "epoch": 799, "lr": 8.415163447974117e-05}
{"train_loss": 0.23592643439769745, "global_step": 71126, "epoch": 799, "lr": 8.415121104490063e-05}
{"train_loss": 0.18602168560028076, "global_step": 71127, "epoch": 799, "lr": 8.415078760546888e-05}
{"train_loss": 0.27229994535446167, "global_step": 71128, "epoch": 799, "lr": 8.415036416144598e-05}
{"train_loss": 0.1900712549686432, "global_step": 71129, "epoch": 799, "lr": 8.414994071283198e-05}
{"train_loss": 0.20226313173770905, "global_step": 71130, "epoch": 799, "lr": 8.414951725962693e-05}
{"train_loss": 0.30422523617744446, "global_step": 71131, "epoch": 799, "lr": 8.41490938018309e-05}
{"train_loss": 0.17160166800022125, "global_step": 71132, "epoch": 799, "lr": 8.414867033944394e-05}
{"train_loss": 0.21042500436306, "global_step": 71133, "epoch": 799, "lr": 8.41482468724661e-05}
{"train_loss": 0.13814376294612885, "global_step": 71134, "epoch": 799, "lr": 8.414782340089745e-05}
{"train_loss": 0.21615779399871826, "global_step": 71135, "epoch": 799, "lr": 8.414739992473804e-05}
{"train_loss": 0.243376687169075, "global_step": 71136, "epoch": 799, "lr": 8.414697644398793e-05}
{"train_loss": 0.13458576798439026, "global_step": 71137, "epoch": 799, "lr": 8.414655295864718e-05}
{"train_loss": 0.1965562403202057, "global_step": 71138, "epoch": 799, "lr": 8.414612946871582e-05}
{"train_loss": 0.1709774136543274, "global_step": 71139, "epoch": 799, "lr": 8.414570597419395e-05}
{"train_loss": 0.3230212926864624, "global_step": 71140, "epoch": 799, "lr": 8.41452824750816e-05}
{"train_loss": 0.1753380000591278, "global_step": 71141, "epoch": 799, "lr": 8.414485897137882e-05}
{"train_loss": 0.22539184987545013, "global_step": 71142, "epoch": 799, "lr": 8.41444354630857e-05}
{"train_loss": 0.2623121440410614, "global_step": 71143, "epoch": 799, "lr": 8.414401195020227e-05}
{"train_loss": 0.16868457198143005, "global_step": 71144, "epoch": 799, "lr": 8.41435884327286e-05}
{"train_loss": 0.16165082156658173, "global_step": 71145, "epoch": 799, "lr": 8.414316491066474e-05}
{"train_loss": 0.2284208983182907, "global_step": 71146, "epoch": 799, "lr": 8.414274138401074e-05}
{"train_loss": 0.2906886637210846, "global_step": 71147, "epoch": 799, "lr": 8.414231785276666e-05}
{"train_loss": 0.16073617339134216, "global_step": 71148, "epoch": 799, "lr": 8.414189431693257e-05}
{"train_loss": 0.2817314565181732, "global_step": 71149, "epoch": 799, "lr": 8.414147077650851e-05}
{"train_loss": 0.22807249426841736, "global_step": 71150, "epoch": 799, "lr": 8.414104723149456e-05}
{"train_loss": 0.21687327325344086, "global_step": 71151, "epoch": 799, "lr": 8.414062368189075e-05}
{"train_loss": 0.2009998857975006, "global_step": 71152, "epoch": 799, "lr": 8.414020012769716e-05}
{"train_loss": 0.2025456577539444, "global_step": 71153, "epoch": 799, "lr": 8.413977656891383e-05}
{"train_loss": 0.2510285973548889, "global_step": 71154, "epoch": 799, "lr": 8.413935300554082e-05}
{"train_loss": 0.2614809572696686, "global_step": 71155, "epoch": 799, "lr": 8.41389294375782e-05}
{"train_loss": 0.19616475701332092, "global_step": 71156, "epoch": 799, "lr": 8.4138505865026e-05}
{"train_loss": 0.20856589078903198, "global_step": 71157, "epoch": 799, "lr": 8.413808228788431e-05}
{"train_loss": 0.19066250324249268, "global_step": 71158, "epoch": 799, "lr": 8.413765870615315e-05}
{"train_loss": 0.2598290741443634, "global_step": 71159, "epoch": 799, "lr": 8.413723511983263e-05}
{"train_loss": 0.2380269467830658, "global_step": 71160, "epoch": 799, "lr": 8.413681152892274e-05}
{"train_loss": 0.22581151127815247, "global_step": 71161, "epoch": 799, "lr": 8.41363879334236e-05}
{"train_loss": 0.1467880755662918, "global_step": 71162, "epoch": 799, "lr": 8.413596433333522e-05}
{"train_loss": 0.22884993255138397, "global_step": 71163, "epoch": 799, "lr": 8.41355407286577e-05}
{"train_loss": 0.25960400700569153, "global_step": 71164, "epoch": 799, "lr": 8.413511711939106e-05}
{"train_loss": 0.21302294731140137, "global_step": 71165, "epoch": 799, "lr": 8.413469350553535e-05}
{"train_loss": 0.16413366794586182, "global_step": 71166, "epoch": 799, "lr": 8.413426988709067e-05}
{"train_loss": 0.18403193354606628, "global_step": 71167, "epoch": 799, "lr": 8.413384626405704e-05}
{"train_loss": 0.19847837090492249, "global_step": 71168, "epoch": 799, "lr": 8.413342263643453e-05}
{"train_loss": 0.20182694494724274, "global_step": 71169, "epoch": 799, "lr": 8.413299900422322e-05}
{"train_loss": 0.2636633813381195, "global_step": 71170, "epoch": 799, "lr": 8.413257536742311e-05}
{"train_loss": 0.1798297017812729, "global_step": 71171, "epoch": 799, "lr": 8.413215172603432e-05}
{"train_loss": 0.19653451442718506, "global_step": 71172, "epoch": 799, "lr": 8.413172808005687e-05}
{"train_loss": 0.1966097205877304, "global_step": 71173, "epoch": 799, "lr": 8.413130442949082e-05}
{"train_loss": 0.15483470261096954, "global_step": 71174, "epoch": 799, "lr": 8.413088077433625e-05}
{"train_loss": 0.12614309787750244, "global_step": 71175, "epoch": 799, "lr": 8.413045711459318e-05}
{"train_loss": 0.15769866108894348, "global_step": 71176, "epoch": 799, "lr": 8.413003345026167e-05}
{"train_loss": 0.2917739748954773, "global_step": 71177, "epoch": 799, "lr": 8.412960978134181e-05}
{"train_loss": 0.23098085820674896, "global_step": 71178, "epoch": 799, "lr": 8.412918610783365e-05}
{"train_loss": 0.13519664108753204, "global_step": 71179, "epoch": 799, "lr": 8.412876242973723e-05}
{"train_loss": 0.23072601854801178, "global_step": 71180, "epoch": 799, "lr": 8.412833874705262e-05}
{"train_loss": 0.21493355929851532, "global_step": 71181, "epoch": 799, "lr": 8.412791505977986e-05}
{"train_loss": 0.24960528314113617, "global_step": 71182, "epoch": 799, "lr": 8.412749136791903e-05}
{"train_loss": 0.23674359917640686, "global_step": 71183, "epoch": 799, "lr": 8.412706767147016e-05}
{"train_loss": 0.19315123558044434, "global_step": 71184, "epoch": 799, "lr": 8.412664397043332e-05}
{"train_loss": 0.138761967420578, "global_step": 71185, "epoch": 799, "lr": 8.412622026480859e-05}
{"train_loss": 0.2572389245033264, "global_step": 71186, "epoch": 799, "lr": 8.412579655459598e-05}
{"train_loss": 0.26084455847740173, "global_step": 71187, "epoch": 799, "lr": 8.412537283979559e-05}
{"train_loss": 0.16757524013519287, "global_step": 71188, "epoch": 799, "lr": 8.412494912040745e-05}
{"train_loss": 0.18415766954421997, "global_step": 71189, "epoch": 799, "lr": 8.412452539643163e-05}
{"train_loss": 0.1766289621591568, "global_step": 71190, "epoch": 799, "lr": 8.41241016678682e-05}
{"train_loss": 0.185444638133049, "global_step": 71191, "epoch": 799, "lr": 8.412367793471719e-05}
{"train_loss": 0.2135917842388153, "global_step": 71192, "epoch": 799, "lr": 8.412325419697866e-05}
{"train_loss": 0.177358940243721, "global_step": 71193, "epoch": 799, "lr": 8.412283045465268e-05}
{"train_loss": 0.26100194454193115, "global_step": 71194, "epoch": 799, "lr": 8.412240670773928e-05}
{"train_loss": 0.15895976126194, "global_step": 71195, "epoch": 799, "lr": 8.412198295623857e-05}
{"train_loss": 0.2766919434070587, "global_step": 71196, "epoch": 799, "lr": 8.412155920015054e-05}
{"train_loss": 0.18060986697673798, "global_step": 71197, "epoch": 799, "lr": 8.412113543947531e-05}
{"train_loss": 0.32293304800987244, "global_step": 71198, "epoch": 799, "lr": 8.412071167421292e-05}
{"train_loss": 0.213020804892765, "global_step": 71199, "epoch": 799, "lr": 8.41202879043634e-05, "val_loss": 3.792602300643921}
{"train_loss": 0.21905112266540527, "global_step": 71200, "epoch": 800, "lr": 8.411986412992682e-05}
{"train_loss": 0.11520537734031677, "global_step": 71201, "epoch": 800, "lr": 8.411944035090324e-05}
{"train_loss": 0.26652729511260986, "global_step": 71202, "epoch": 800, "lr": 8.411901656729272e-05}
{"train_loss": 0.2043500691652298, "global_step": 71203, "epoch": 800, "lr": 8.411859277909531e-05}
{"train_loss": 0.19259844720363617, "global_step": 71204, "epoch": 800, "lr": 8.411816898631108e-05}
{"train_loss": 0.21375814080238342, "global_step": 71205, "epoch": 800, "lr": 8.411774518894006e-05}
{"train_loss": 0.17543752491474152, "global_step": 71206, "epoch": 800, "lr": 8.411732138698236e-05}
{"train_loss": 0.16307204961776733, "global_step": 71207, "epoch": 800, "lr": 8.411689758043797e-05}
{"train_loss": 0.18017055094242096, "global_step": 71208, "epoch": 800, "lr": 8.4116473769307e-05}
{"train_loss": 0.14128103852272034, "global_step": 71209, "epoch": 800, "lr": 8.411604995358947e-05}
{"train_loss": 0.18496902287006378, "global_step": 71210, "epoch": 800, "lr": 8.411562613328545e-05}
{"train_loss": 0.22513511776924133, "global_step": 71211, "epoch": 800, "lr": 8.411520230839502e-05}
{"train_loss": 0.1731947362422943, "global_step": 71212, "epoch": 800, "lr": 8.41147784789182e-05}
{"train_loss": 0.23817148804664612, "global_step": 71213, "epoch": 800, "lr": 8.411435464485507e-05}
{"train_loss": 0.2002626359462738, "global_step": 71214, "epoch": 800, "lr": 8.411393080620567e-05}
{"train_loss": 0.26112815737724304, "global_step": 71215, "epoch": 800, "lr": 8.411350696297008e-05}
{"train_loss": 0.22131860256195068, "global_step": 71216, "epoch": 800, "lr": 8.411308311514834e-05}
{"train_loss": 0.1676981896162033, "global_step": 71217, "epoch": 800, "lr": 8.411265926274052e-05}
{"train_loss": 0.16703879833221436, "global_step": 71218, "epoch": 800, "lr": 8.411223540574666e-05}
{"train_loss": 0.16581083834171295, "global_step": 71219, "epoch": 800, "lr": 8.411181154416682e-05}
{"train_loss": 0.17636479437351227, "global_step": 71220, "epoch": 800, "lr": 8.411138767800109e-05}
{"train_loss": 0.19161060452461243, "global_step": 71221, "epoch": 800, "lr": 8.411096380724947e-05}
{"train_loss": 0.20838776230812073, "global_step": 71222, "epoch": 800, "lr": 8.411053993191205e-05}
{"train_loss": 0.1137586236000061, "global_step": 71223, "epoch": 800, "lr": 8.41101160519889e-05}
{"train_loss": 0.3127681314945221, "global_step": 71224, "epoch": 800, "lr": 8.410969216748005e-05}
{"train_loss": 0.21087472140789032, "global_step": 71225, "epoch": 800, "lr": 8.410926827838557e-05}
{"train_loss": 0.2847902774810791, "global_step": 71226, "epoch": 800, "lr": 8.410884438470551e-05}
{"train_loss": 0.16485100984573364, "global_step": 71227, "epoch": 800, "lr": 8.410842048643994e-05}
{"train_loss": 0.16144438087940216, "global_step": 71228, "epoch": 800, "lr": 8.41079965835889e-05}
{"train_loss": 0.3002660274505615, "global_step": 71229, "epoch": 800, "lr": 8.410757267615246e-05}
{"train_loss": 0.3160525858402252, "global_step": 71230, "epoch": 800, "lr": 8.410714876413068e-05}
{"train_loss": 0.21861141920089722, "global_step": 71231, "epoch": 800, "lr": 8.41067248475236e-05}
{"train_loss": 0.12979742884635925, "global_step": 71232, "epoch": 800, "lr": 8.410630092633127e-05}
{"train_loss": 0.231961190700531, "global_step": 71233, "epoch": 800, "lr": 8.410587700055378e-05}
{"train_loss": 0.1777287721633911, "global_step": 71234, "epoch": 800, "lr": 8.410545307019118e-05}
{"train_loss": 0.1617257446050644, "global_step": 71235, "epoch": 800, "lr": 8.410502913524352e-05}
{"train_loss": 0.15392926335334778, "global_step": 71236, "epoch": 800, "lr": 8.410460519571084e-05}
{"train_loss": 0.20268747210502625, "global_step": 71237, "epoch": 800, "lr": 8.410418125159322e-05}
{"train_loss": 0.17663025856018066, "global_step": 71238, "epoch": 800, "lr": 8.41037573028907e-05}
{"train_loss": 0.28624242544174194, "global_step": 71239, "epoch": 800, "lr": 8.410333334960335e-05}
{"train_loss": 0.18446028232574463, "global_step": 71240, "epoch": 800, "lr": 8.410290939173122e-05}
{"train_loss": 0.17085516452789307, "global_step": 71241, "epoch": 800, "lr": 8.410248542927438e-05}
{"train_loss": 0.16158640384674072, "global_step": 71242, "epoch": 800, "lr": 8.410206146223286e-05}
{"train_loss": 0.186044842004776, "global_step": 71243, "epoch": 800, "lr": 8.410163749060674e-05}
{"train_loss": 0.24774716794490814, "global_step": 71244, "epoch": 800, "lr": 8.410121351439608e-05}
{"train_loss": 0.30757251381874084, "global_step": 71245, "epoch": 800, "lr": 8.410078953360092e-05}
{"train_loss": 0.1570531576871872, "global_step": 71246, "epoch": 800, "lr": 8.410036554822132e-05}
{"train_loss": 0.2323479950428009, "global_step": 71247, "epoch": 800, "lr": 8.409994155825734e-05}
{"train_loss": 0.12970194220542908, "global_step": 71248, "epoch": 800, "lr": 8.409951756370906e-05}
{"train_loss": 0.22783306241035461, "global_step": 71249, "epoch": 800, "lr": 8.409909356457648e-05}
{"train_loss": 0.248700350522995, "global_step": 71250, "epoch": 800, "lr": 8.409866956085973e-05}
{"train_loss": 0.20553195476531982, "global_step": 71251, "epoch": 800, "lr": 8.40982455525588e-05}
{"train_loss": 0.1793835163116455, "global_step": 71252, "epoch": 800, "lr": 8.409782153967378e-05}
{"train_loss": 0.19579720497131348, "global_step": 71253, "epoch": 800, "lr": 8.409739752220473e-05}
{"train_loss": 0.18907982110977173, "global_step": 71254, "epoch": 800, "lr": 8.409697350015171e-05}
{"train_loss": 0.21355019509792328, "global_step": 71255, "epoch": 800, "lr": 8.409654947351477e-05}
{"train_loss": 0.20052935183048248, "global_step": 71256, "epoch": 800, "lr": 8.409612544229395e-05}
{"train_loss": 0.22332073748111725, "global_step": 71257, "epoch": 800, "lr": 8.409570140648933e-05}
{"train_loss": 0.17611980438232422, "global_step": 71258, "epoch": 800, "lr": 8.409527736610094e-05}
{"train_loss": 0.1604539155960083, "global_step": 71259, "epoch": 800, "lr": 8.409485332112887e-05}
{"train_loss": 0.25309106707572937, "global_step": 71260, "epoch": 800, "lr": 8.409442927157317e-05}
{"train_loss": 0.22577893733978271, "global_step": 71261, "epoch": 800, "lr": 8.409400521743389e-05}
{"train_loss": 0.29965531826019287, "global_step": 71262, "epoch": 800, "lr": 8.409358115871106e-05}
{"train_loss": 0.2555239796638489, "global_step": 71263, "epoch": 800, "lr": 8.40931570954048e-05}
{"train_loss": 0.188520610332489, "global_step": 71264, "epoch": 800, "lr": 8.40927330275151e-05}
{"train_loss": 0.1604616492986679, "global_step": 71265, "epoch": 800, "lr": 8.409230895504206e-05}
{"train_loss": 0.28059592843055725, "global_step": 71266, "epoch": 800, "lr": 8.409188487798574e-05}
{"train_loss": 0.2517279386520386, "global_step": 71267, "epoch": 800, "lr": 8.409146079634615e-05}
{"train_loss": 0.21188019216060638, "global_step": 71268, "epoch": 800, "lr": 8.409103671012339e-05}
{"train_loss": 0.21878187358379364, "global_step": 71269, "epoch": 800, "lr": 8.409061261931751e-05}
{"train_loss": 0.33140870928764343, "global_step": 71270, "epoch": 800, "lr": 8.409018852392857e-05}
{"train_loss": 0.26896750926971436, "global_step": 71271, "epoch": 800, "lr": 8.408976442395662e-05}
{"train_loss": 0.2092536985874176, "global_step": 71272, "epoch": 800, "lr": 8.40893403194017e-05}
{"train_loss": 0.15424244105815887, "global_step": 71273, "epoch": 800, "lr": 8.408891621026389e-05}
{"train_loss": 0.23873424530029297, "global_step": 71274, "epoch": 800, "lr": 8.408849209654325e-05}
{"train_loss": 0.14253340661525726, "global_step": 71275, "epoch": 800, "lr": 8.408806797823983e-05}
{"train_loss": 0.25426965951919556, "global_step": 71276, "epoch": 800, "lr": 8.408764385535367e-05}
{"train_loss": 0.2128644585609436, "global_step": 71277, "epoch": 800, "lr": 8.408721972788485e-05}
{"train_loss": 0.24842670559883118, "global_step": 71278, "epoch": 800, "lr": 8.408679559583342e-05}
{"train_loss": 0.14157912135124207, "global_step": 71279, "epoch": 800, "lr": 8.408637145919943e-05}
{"train_loss": 0.14679816365242004, "global_step": 71280, "epoch": 800, "lr": 8.408594731798293e-05}
{"train_loss": 0.1953810751438141, "global_step": 71281, "epoch": 800, "lr": 8.4085523172184e-05}
{"train_loss": 0.24405577778816223, "global_step": 71282, "epoch": 800, "lr": 8.40850990218027e-05}
{"train_loss": 0.21968014538288116, "global_step": 71283, "epoch": 800, "lr": 8.408467486683908e-05}
{"train_loss": 0.2779729664325714, "global_step": 71284, "epoch": 800, "lr": 8.408425070729317e-05}
{"train_loss": 0.26854729652404785, "global_step": 71285, "epoch": 800, "lr": 8.408382654316505e-05}
{"train_loss": 0.22817112505435944, "global_step": 71286, "epoch": 800, "lr": 8.408340237445477e-05}
{"train_loss": 0.2378167062997818, "global_step": 71287, "epoch": 800, "lr": 8.408297820116242e-05}
{"train_loss": 0.20973081786311074, "global_step": 71288, "epoch": 800, "lr": 8.4082554023288e-05, "train/sim_max_reward_0": 0.4577240125809739, "train/sim_max_reward_1": 0.7151969948192453, "train/sim_max_reward_2": 0.6140971047646289, "train/sim_max_reward_3": 0.47042566713658823, "train/sim_max_reward_4": 0.9602320394872832, "train/sim_max_reward_5": 0.5719741849939927, "test/sim_max_reward_4500000": 0.4019865118243266, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9643937554028531, "test/sim_max_reward_4500003": 0.4613871944364437, "test/sim_max_reward_4500004": 0.7072307924285651, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.7992453914554841, "test/sim_max_reward_4500007": 0.9674053985492193, "test/sim_max_reward_4500008": 0.44783299415284844, "test/sim_max_reward_4500009": 0.5120519143779424, "test/sim_max_reward_4500010": 0.31857806831013413, "test/sim_max_reward_4500011": 0.9765895190151683, "test/sim_max_reward_4500012": 0.9225137383737599, "test/sim_max_reward_4500013": 0.6070194002920946, "test/sim_max_reward_4500014": 0.9310783432096457, "test/sim_max_reward_4500015": 0.9929904860178318, "test/sim_max_reward_4500016": 0.9825964728802336, "test/sim_max_reward_4500017": 1.0, "test/sim_max_reward_4500018": 0.9627988227653943, "test/sim_max_reward_4500019": 0.5339285812543996, "test/sim_max_reward_4500020": 1.0, "test/sim_max_reward_4500021": 0.9961612871646867, "test/sim_max_reward_4500022": 0.998608096262625, "test/sim_max_reward_4500023": 0.6211700706508797, "test/sim_max_reward_4500024": 0.682768011456719, "test/sim_max_reward_4500025": 0.0, "test/sim_max_reward_4500026": 0.9708471117581253, "test/sim_max_reward_4500027": 0.550815730837667, "test/sim_max_reward_4500028": 0.9942229691792162, "test/sim_max_reward_4500029": 0.48717895327377264, "test/sim_max_reward_4500030": 0.6325359932341622, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.1894542709910032, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.9361240928160369, "test/sim_max_reward_4500036": 0.4424024947523827, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.6321905697631114, "test/sim_max_reward_4500039": 0.9781558027729669, "test/sim_max_reward_4500040": 0.9513089397537114, "test/sim_max_reward_4500041": 0.5151512714136409, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.8813344652857041, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.7035727042075736, "test/sim_max_reward_4500046": 0.34540720308541634, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9508064364542868, "test/sim_max_reward_4500049": 0.7625242875548232, "train/mean_score": 0.6316083339637854, "test/mean_score": 0.6942473629482973, "val_loss": 3.792006254196167, "train_action_mse_error": 22.41189956665039}
{"train_loss": 0.17977149784564972, "global_step": 71289, "epoch": 801, "lr": 8.408212984083161e-05}
{"train_loss": 0.37907907366752625, "global_step": 71290, "epoch": 801, "lr": 8.40817056537933e-05}
{"train_loss": 0.18385446071624756, "global_step": 71291, "epoch": 801, "lr": 8.40812814621731e-05}
{"train_loss": 0.1851470023393631, "global_step": 71292, "epoch": 801, "lr": 8.408085726597111e-05}
{"train_loss": 0.21872879564762115, "global_step": 71293, "epoch": 801, "lr": 8.408043306518733e-05}
{"train_loss": 0.18031291663646698, "global_step": 71294, "epoch": 801, "lr": 8.408000885982187e-05}
{"train_loss": 0.24629618227481842, "global_step": 71295, "epoch": 801, "lr": 8.407958464987478e-05}
{"train_loss": 0.2332915961742401, "global_step": 71296, "epoch": 801, "lr": 8.40791604353461e-05}
{"train_loss": 0.21215882897377014, "global_step": 71297, "epoch": 801, "lr": 8.407873621623589e-05}
{"train_loss": 0.1952962577342987, "global_step": 71298, "epoch": 801, "lr": 8.407831199254422e-05}
{"train_loss": 0.1718641221523285, "global_step": 71299, "epoch": 801, "lr": 8.40778877642711e-05}
{"train_loss": 0.19039367139339447, "global_step": 71300, "epoch": 801, "lr": 8.407746353141666e-05}
{"train_loss": 0.10834138840436935, "global_step": 71301, "epoch": 801, "lr": 8.40770392939809e-05}
{"train_loss": 0.25271010398864746, "global_step": 71302, "epoch": 801, "lr": 8.407661505196391e-05}
{"train_loss": 0.19796626269817352, "global_step": 71303, "epoch": 801, "lr": 8.407619080536573e-05}
{"train_loss": 0.27823761105537415, "global_step": 71304, "epoch": 801, "lr": 8.40757665541864e-05}
{"train_loss": 0.19832836091518402, "global_step": 71305, "epoch": 801, "lr": 8.407534229842602e-05}
{"train_loss": 0.11769761890172958, "global_step": 71306, "epoch": 801, "lr": 8.407491803808463e-05}
{"train_loss": 0.2036830335855484, "global_step": 71307, "epoch": 801, "lr": 8.407449377316227e-05}
{"train_loss": 0.2358379364013672, "global_step": 71308, "epoch": 801, "lr": 8.407406950365904e-05}
{"train_loss": 0.29880809783935547, "global_step": 71309, "epoch": 801, "lr": 8.407364522957492e-05}
{"train_loss": 0.318354070186615, "global_step": 71310, "epoch": 801, "lr": 8.407322095091005e-05}
{"train_loss": 0.22540071606636047, "global_step": 71311, "epoch": 801, "lr": 8.407279666766443e-05}
{"train_loss": 0.21354256570339203, "global_step": 71312, "epoch": 801, "lr": 8.407237237983816e-05}
{"train_loss": 0.2619573175907135, "global_step": 71313, "epoch": 801, "lr": 8.407194808743126e-05}
{"train_loss": 0.20828215777873993, "global_step": 71314, "epoch": 801, "lr": 8.40715237904438e-05}
{"train_loss": 0.2968308925628662, "global_step": 71315, "epoch": 801, "lr": 8.407109948887584e-05}
{"train_loss": 0.20918966829776764, "global_step": 71316, "epoch": 801, "lr": 8.407067518272745e-05}
{"train_loss": 0.24025501310825348, "global_step": 71317, "epoch": 801, "lr": 8.407025087199864e-05}
{"train_loss": 0.23965811729431152, "global_step": 71318, "epoch": 801, "lr": 8.406982655668953e-05}
{"train_loss": 0.23191183805465698, "global_step": 71319, "epoch": 801, "lr": 8.406940223680013e-05}
{"train_loss": 0.26984769105911255, "global_step": 71320, "epoch": 801, "lr": 8.406897791233052e-05}
{"train_loss": 0.19855602085590363, "global_step": 71321, "epoch": 801, "lr": 8.406855358328076e-05}
{"train_loss": 0.14720238745212555, "global_step": 71322, "epoch": 801, "lr": 8.406812924965088e-05}
{"train_loss": 0.16732266545295715, "global_step": 71323, "epoch": 801, "lr": 8.406770491144097e-05}
{"train_loss": 0.20767158269882202, "global_step": 71324, "epoch": 801, "lr": 8.406728056865107e-05}
{"train_loss": 0.1564628779888153, "global_step": 71325, "epoch": 801, "lr": 8.406685622128123e-05}
{"train_loss": 0.28272661566734314, "global_step": 71326, "epoch": 801, "lr": 8.406643186933153e-05}
{"train_loss": 0.20516502857208252, "global_step": 71327, "epoch": 801, "lr": 8.4066007512802e-05}
{"train_loss": 0.23488979041576385, "global_step": 71328, "epoch": 801, "lr": 8.406558315169272e-05}
{"train_loss": 0.16233202815055847, "global_step": 71329, "epoch": 801, "lr": 8.406515878600374e-05}
{"train_loss": 0.1754695475101471, "global_step": 71330, "epoch": 801, "lr": 8.406473441573511e-05}
{"train_loss": 0.13673123717308044, "global_step": 71331, "epoch": 801, "lr": 8.40643100408869e-05}
{"train_loss": 0.22391217947006226, "global_step": 71332, "epoch": 801, "lr": 8.406388566145915e-05}
{"train_loss": 0.19926071166992188, "global_step": 71333, "epoch": 801, "lr": 8.406346127745193e-05}
{"train_loss": 0.28486117720603943, "global_step": 71334, "epoch": 801, "lr": 8.406303688886528e-05}
{"train_loss": 0.2006629854440689, "global_step": 71335, "epoch": 801, "lr": 8.406261249569928e-05}
{"train_loss": 0.15651294589042664, "global_step": 71336, "epoch": 801, "lr": 8.406218809795397e-05}
{"train_loss": 0.16359351575374603, "global_step": 71337, "epoch": 801, "lr": 8.406176369562942e-05}
{"train_loss": 0.19397644698619843, "global_step": 71338, "epoch": 801, "lr": 8.406133928872569e-05}
{"train_loss": 0.24267330765724182, "global_step": 71339, "epoch": 801, "lr": 8.406091487724282e-05}
{"train_loss": 0.2371203899383545, "global_step": 71340, "epoch": 801, "lr": 8.406049046118087e-05}
{"train_loss": 0.2187812179327011, "global_step": 71341, "epoch": 801, "lr": 8.40600660405399e-05}
{"train_loss": 0.20928728580474854, "global_step": 71342, "epoch": 801, "lr": 8.405964161531999e-05}
{"train_loss": 0.16638006269931793, "global_step": 71343, "epoch": 801, "lr": 8.405921718552114e-05}
{"train_loss": 0.19830137491226196, "global_step": 71344, "epoch": 801, "lr": 8.405879275114347e-05}
{"train_loss": 0.1584915965795517, "global_step": 71345, "epoch": 801, "lr": 8.405836831218701e-05}
{"train_loss": 0.26093170046806335, "global_step": 71346, "epoch": 801, "lr": 8.405794386865181e-05}
{"train_loss": 0.13777095079421997, "global_step": 71347, "epoch": 801, "lr": 8.405751942053794e-05}
{"train_loss": 0.15866483747959137, "global_step": 71348, "epoch": 801, "lr": 8.405709496784545e-05}
{"train_loss": 0.1717434525489807, "global_step": 71349, "epoch": 801, "lr": 8.40566705105744e-05}
{"train_loss": 0.2398722767829895, "global_step": 71350, "epoch": 801, "lr": 8.405624604872483e-05}
{"train_loss": 0.18250568211078644, "global_step": 71351, "epoch": 801, "lr": 8.405582158229683e-05}
{"train_loss": 0.2255019098520279, "global_step": 71352, "epoch": 801, "lr": 8.405539711129045e-05}
{"train_loss": 0.1812288761138916, "global_step": 71353, "epoch": 801, "lr": 8.405497263570572e-05}
{"train_loss": 0.1946614682674408, "global_step": 71354, "epoch": 801, "lr": 8.405454815554272e-05}
{"train_loss": 0.23104576766490936, "global_step": 71355, "epoch": 801, "lr": 8.40541236708015e-05}
{"train_loss": 0.20382823050022125, "global_step": 71356, "epoch": 801, "lr": 8.405369918148211e-05}
{"train_loss": 0.22039636969566345, "global_step": 71357, "epoch": 801, "lr": 8.405327468758463e-05}
{"train_loss": 0.20703408122062683, "global_step": 71358, "epoch": 801, "lr": 8.405285018910908e-05}
{"train_loss": 0.2423378974199295, "global_step": 71359, "epoch": 801, "lr": 8.405242568605555e-05}
{"train_loss": 0.17722545564174652, "global_step": 71360, "epoch": 801, "lr": 8.405200117842409e-05}
{"train_loss": 0.23892292380332947, "global_step": 71361, "epoch": 801, "lr": 8.405157666621477e-05}
{"train_loss": 0.21903106570243835, "global_step": 71362, "epoch": 801, "lr": 8.40511521494276e-05}
{"train_loss": 0.2234177589416504, "global_step": 71363, "epoch": 801, "lr": 8.405072762806268e-05}
{"train_loss": 0.20228257775306702, "global_step": 71364, "epoch": 801, "lr": 8.405030310212006e-05}
{"train_loss": 0.2455977350473404, "global_step": 71365, "epoch": 801, "lr": 8.404987857159978e-05}
{"train_loss": 0.26050570607185364, "global_step": 71366, "epoch": 801, "lr": 8.404945403650193e-05}
{"train_loss": 0.22071106731891632, "global_step": 71367, "epoch": 801, "lr": 8.404902949682653e-05}
{"train_loss": 0.16790585219860077, "global_step": 71368, "epoch": 801, "lr": 8.404860495257365e-05}
{"train_loss": 0.2444176971912384, "global_step": 71369, "epoch": 801, "lr": 8.404818040374335e-05}
{"train_loss": 0.2163325548171997, "global_step": 71370, "epoch": 801, "lr": 8.404775585033571e-05}
{"train_loss": 0.25274237990379333, "global_step": 71371, "epoch": 801, "lr": 8.404733129235073e-05}
{"train_loss": 0.3027532696723938, "global_step": 71372, "epoch": 801, "lr": 8.404690672978852e-05}
{"train_loss": 0.2768288850784302, "global_step": 71373, "epoch": 801, "lr": 8.404648216264912e-05}
{"train_loss": 0.24777279794216156, "global_step": 71374, "epoch": 801, "lr": 8.404605759093258e-05}
{"train_loss": 0.2501537799835205, "global_step": 71375, "epoch": 801, "lr": 8.404563301463897e-05}
{"train_loss": 0.17613264918327332, "global_step": 71376, "epoch": 801, "lr": 8.404520843376833e-05}
{"train_loss": 0.21527602967251552, "global_step": 71377, "epoch": 801, "lr": 8.404478384832073e-05, "val_loss": 3.8327324390411377}
{"train_loss": 0.27735841274261475, "global_step": 71378, "epoch": 802, "lr": 8.404435925829623e-05}
{"train_loss": 0.24054113030433655, "global_step": 71379, "epoch": 802, "lr": 8.404393466369486e-05}
{"train_loss": 0.18338099122047424, "global_step": 71380, "epoch": 802, "lr": 8.404351006451673e-05}
{"train_loss": 0.19784051179885864, "global_step": 71381, "epoch": 802, "lr": 8.404308546076185e-05}
{"train_loss": 0.1480768918991089, "global_step": 71382, "epoch": 802, "lr": 8.40426608524303e-05}
{"train_loss": 0.1853446364402771, "global_step": 71383, "epoch": 802, "lr": 8.40422362395221e-05}
{"train_loss": 0.1887611746788025, "global_step": 71384, "epoch": 802, "lr": 8.404181162203736e-05}
{"train_loss": 0.21046505868434906, "global_step": 71385, "epoch": 802, "lr": 8.40413869999761e-05}
{"train_loss": 0.17830981314182281, "global_step": 71386, "epoch": 802, "lr": 8.404096237333841e-05}
{"train_loss": 0.13245312869548798, "global_step": 71387, "epoch": 802, "lr": 8.404053774212431e-05}
{"train_loss": 0.28129148483276367, "global_step": 71388, "epoch": 802, "lr": 8.404011310633388e-05}
{"train_loss": 0.19957832992076874, "global_step": 71389, "epoch": 802, "lr": 8.403968846596718e-05}
{"train_loss": 0.16661500930786133, "global_step": 71390, "epoch": 802, "lr": 8.403926382102424e-05}
{"train_loss": 0.21526628732681274, "global_step": 71391, "epoch": 802, "lr": 8.403883917150516e-05}
{"train_loss": 0.2723725736141205, "global_step": 71392, "epoch": 802, "lr": 8.403841451740996e-05}
{"train_loss": 0.2455398440361023, "global_step": 71393, "epoch": 802, "lr": 8.403798985873871e-05}
{"train_loss": 0.23958192765712738, "global_step": 71394, "epoch": 802, "lr": 8.403756519549146e-05}
{"train_loss": 0.17392507195472717, "global_step": 71395, "epoch": 802, "lr": 8.403714052766829e-05}
{"train_loss": 0.22185583412647247, "global_step": 71396, "epoch": 802, "lr": 8.403671585526923e-05}
{"train_loss": 0.14315913617610931, "global_step": 71397, "epoch": 802, "lr": 8.403629117829435e-05}
{"train_loss": 0.3315392732620239, "global_step": 71398, "epoch": 802, "lr": 8.40358664967437e-05}
{"train_loss": 0.22126556932926178, "global_step": 71399, "epoch": 802, "lr": 8.403544181061737e-05}
{"train_loss": 0.1455581784248352, "global_step": 71400, "epoch": 802, "lr": 8.403501711991537e-05}
{"train_loss": 0.2661082148551941, "global_step": 71401, "epoch": 802, "lr": 8.403459242463776e-05}
{"train_loss": 0.1769341081380844, "global_step": 71402, "epoch": 802, "lr": 8.403416772478464e-05}
{"train_loss": 0.19804906845092773, "global_step": 71403, "epoch": 802, "lr": 8.403374302035603e-05}
{"train_loss": 0.2107124626636505, "global_step": 71404, "epoch": 802, "lr": 8.403331831135201e-05}
{"train_loss": 0.2509172558784485, "global_step": 71405, "epoch": 802, "lr": 8.403289359777261e-05}
{"train_loss": 0.24564936757087708, "global_step": 71406, "epoch": 802, "lr": 8.403246887961791e-05}
{"train_loss": 0.17874731123447418, "global_step": 71407, "epoch": 802, "lr": 8.403204415688794e-05}
{"train_loss": 0.2463262677192688, "global_step": 71408, "epoch": 802, "lr": 8.40316194295828e-05}
{"train_loss": 0.21508702635765076, "global_step": 71409, "epoch": 802, "lr": 8.403119469770252e-05}
{"train_loss": 0.2774713337421417, "global_step": 71410, "epoch": 802, "lr": 8.403076996124717e-05}
{"train_loss": 0.1435817927122116, "global_step": 71411, "epoch": 802, "lr": 8.403034522021677e-05}
{"train_loss": 0.33820533752441406, "global_step": 71412, "epoch": 802, "lr": 8.402992047461143e-05}
{"train_loss": 0.21577675640583038, "global_step": 71413, "epoch": 802, "lr": 8.402949572443115e-05}
{"train_loss": 0.23622579872608185, "global_step": 71414, "epoch": 802, "lr": 8.402907096967605e-05}
{"train_loss": 0.1452459692955017, "global_step": 71415, "epoch": 802, "lr": 8.402864621034614e-05}
{"train_loss": 0.14094975590705872, "global_step": 71416, "epoch": 802, "lr": 8.402822144644151e-05}
{"train_loss": 0.18173639476299286, "global_step": 71417, "epoch": 802, "lr": 8.40277966779622e-05}
{"train_loss": 0.20336183905601501, "global_step": 71418, "epoch": 802, "lr": 8.402737190490825e-05}
{"train_loss": 0.1767617166042328, "global_step": 71419, "epoch": 802, "lr": 8.402694712727975e-05}
{"train_loss": 0.15461036562919617, "global_step": 71420, "epoch": 802, "lr": 8.402652234507674e-05}
{"train_loss": 0.2844896912574768, "global_step": 71421, "epoch": 802, "lr": 8.402609755829926e-05}
{"train_loss": 0.20920152962207794, "global_step": 71422, "epoch": 802, "lr": 8.40256727669474e-05}
{"train_loss": 0.21559953689575195, "global_step": 71423, "epoch": 802, "lr": 8.402524797102121e-05}
{"train_loss": 0.24137455224990845, "global_step": 71424, "epoch": 802, "lr": 8.402482317052071e-05}
{"train_loss": 0.1977366954088211, "global_step": 71425, "epoch": 802, "lr": 8.402439836544603e-05}
{"train_loss": 0.26487061381340027, "global_step": 71426, "epoch": 802, "lr": 8.402397355579715e-05}
{"train_loss": 0.1625140905380249, "global_step": 71427, "epoch": 802, "lr": 8.402354874157417e-05}
{"train_loss": 0.21873094141483307, "global_step": 71428, "epoch": 802, "lr": 8.402312392277716e-05}
{"train_loss": 0.2632991671562195, "global_step": 71429, "epoch": 802, "lr": 8.402269909940613e-05}
{"train_loss": 0.1737062633037567, "global_step": 71430, "epoch": 802, "lr": 8.402227427146116e-05}
{"train_loss": 0.19555208086967468, "global_step": 71431, "epoch": 802, "lr": 8.402184943894234e-05}
{"train_loss": 0.15985874831676483, "global_step": 71432, "epoch": 802, "lr": 8.402142460184968e-05}
{"train_loss": 0.18732605874538422, "global_step": 71433, "epoch": 802, "lr": 8.402099976018324e-05}
{"train_loss": 0.19874416291713715, "global_step": 71434, "epoch": 802, "lr": 8.402057491394311e-05}
{"train_loss": 0.1673278510570526, "global_step": 71435, "epoch": 802, "lr": 8.402015006312931e-05}
{"train_loss": 0.2233753502368927, "global_step": 71436, "epoch": 802, "lr": 8.401972520774195e-05}
{"train_loss": 0.19007204473018646, "global_step": 71437, "epoch": 802, "lr": 8.401930034778103e-05}
{"train_loss": 0.20093348622322083, "global_step": 71438, "epoch": 802, "lr": 8.401887548324663e-05}
{"train_loss": 0.17499326169490814, "global_step": 71439, "epoch": 802, "lr": 8.40184506141388e-05}
{"train_loss": 0.1800735741853714, "global_step": 71440, "epoch": 802, "lr": 8.401802574045761e-05}
{"train_loss": 0.21552392840385437, "global_step": 71441, "epoch": 802, "lr": 8.401760086220311e-05}
{"train_loss": 0.21684257686138153, "global_step": 71442, "epoch": 802, "lr": 8.401717597937536e-05}
{"train_loss": 0.21391677856445312, "global_step": 71443, "epoch": 802, "lr": 8.401675109197443e-05}
{"train_loss": 0.2085038423538208, "global_step": 71444, "epoch": 802, "lr": 8.401632620000032e-05}
{"train_loss": 0.18920086324214935, "global_step": 71445, "epoch": 802, "lr": 8.401590130345316e-05}
{"train_loss": 0.23038600385189056, "global_step": 71446, "epoch": 802, "lr": 8.401547640233298e-05}
{"train_loss": 0.16029158234596252, "global_step": 71447, "epoch": 802, "lr": 8.401505149663984e-05}
{"train_loss": 0.2209695279598236, "global_step": 71448, "epoch": 802, "lr": 8.401462658637377e-05}
{"train_loss": 0.2073521614074707, "global_step": 71449, "epoch": 802, "lr": 8.401420167153486e-05}
{"train_loss": 0.3652604818344116, "global_step": 71450, "epoch": 802, "lr": 8.401377675212315e-05}
{"train_loss": 0.2643323242664337, "global_step": 71451, "epoch": 802, "lr": 8.40133518281387e-05}
{"train_loss": 0.32363542914390564, "global_step": 71452, "epoch": 802, "lr": 8.401292689958158e-05}
{"train_loss": 0.13829892873764038, "global_step": 71453, "epoch": 802, "lr": 8.401250196645182e-05}
{"train_loss": 0.2707204520702362, "global_step": 71454, "epoch": 802, "lr": 8.401207702874952e-05}
{"train_loss": 0.19614142179489136, "global_step": 71455, "epoch": 802, "lr": 8.401165208647468e-05}
{"train_loss": 0.2076980620622635, "global_step": 71456, "epoch": 802, "lr": 8.401122713962741e-05}
{"train_loss": 0.2112494856119156, "global_step": 71457, "epoch": 802, "lr": 8.401080218820774e-05}
{"train_loss": 0.2537488043308258, "global_step": 71458, "epoch": 802, "lr": 8.401037723221574e-05}
{"train_loss": 0.21343860030174255, "global_step": 71459, "epoch": 802, "lr": 8.400995227165145e-05}
{"train_loss": 0.22980163991451263, "global_step": 71460, "epoch": 802, "lr": 8.400952730651493e-05}
{"train_loss": 0.2580985128879547, "global_step": 71461, "epoch": 802, "lr": 8.400910233680624e-05}
{"train_loss": 0.2510221004486084, "global_step": 71462, "epoch": 802, "lr": 8.400867736252546e-05}
{"train_loss": 0.21463972330093384, "global_step": 71463, "epoch": 802, "lr": 8.400825238367261e-05}
{"train_loss": 0.18919481337070465, "global_step": 71464, "epoch": 802, "lr": 8.400782740024777e-05}
{"train_loss": 0.16267886757850647, "global_step": 71465, "epoch": 802, "lr": 8.400740241225101e-05}
{"train_loss": 0.21326010515180865, "global_step": 71466, "epoch": 802, "lr": 8.400697741968234e-05, "val_loss": 3.715923547744751}
{"train_loss": 0.208324134349823, "global_step": 71467, "epoch": 803, "lr": 8.400655242254186e-05}
{"train_loss": 0.2672262191772461, "global_step": 71468, "epoch": 803, "lr": 8.400612742082962e-05}
{"train_loss": 0.2325817495584488, "global_step": 71469, "epoch": 803, "lr": 8.400570241454566e-05}
{"train_loss": 0.20764857530593872, "global_step": 71470, "epoch": 803, "lr": 8.400527740369006e-05}
{"train_loss": 0.2248300462961197, "global_step": 71471, "epoch": 803, "lr": 8.400485238826286e-05}
{"train_loss": 0.2243947982788086, "global_step": 71472, "epoch": 803, "lr": 8.40044273682641e-05}
{"train_loss": 0.2397845983505249, "global_step": 71473, "epoch": 803, "lr": 8.400400234369389e-05}
{"train_loss": 0.21529819071292877, "global_step": 71474, "epoch": 803, "lr": 8.400357731455224e-05}
{"train_loss": 0.24193629622459412, "global_step": 71475, "epoch": 803, "lr": 8.400315228083924e-05}
{"train_loss": 0.17606937885284424, "global_step": 71476, "epoch": 803, "lr": 8.400272724255491e-05}
{"train_loss": 0.1720489263534546, "global_step": 71477, "epoch": 803, "lr": 8.400230219969935e-05}
{"train_loss": 0.18514852225780487, "global_step": 71478, "epoch": 803, "lr": 8.400187715227257e-05}
{"train_loss": 0.2848134934902191, "global_step": 71479, "epoch": 803, "lr": 8.400145210027467e-05}
{"train_loss": 0.23254869878292084, "global_step": 71480, "epoch": 803, "lr": 8.400102704370568e-05}
{"train_loss": 0.23175133764743805, "global_step": 71481, "epoch": 803, "lr": 8.400060198256568e-05}
{"train_loss": 0.23929287493228912, "global_step": 71482, "epoch": 803, "lr": 8.400017691685468e-05}
{"train_loss": 0.17817145586013794, "global_step": 71483, "epoch": 803, "lr": 8.39997518465728e-05}
{"train_loss": 0.1937226504087448, "global_step": 71484, "epoch": 803, "lr": 8.399932677172007e-05}
{"train_loss": 0.21092718839645386, "global_step": 71485, "epoch": 803, "lr": 8.399890169229653e-05}
{"train_loss": 0.18647794425487518, "global_step": 71486, "epoch": 803, "lr": 8.399847660830225e-05}
{"train_loss": 0.32742196321487427, "global_step": 71487, "epoch": 803, "lr": 8.399805151973729e-05}
{"train_loss": 0.24774613976478577, "global_step": 71488, "epoch": 803, "lr": 8.399762642660173e-05}
{"train_loss": 0.18843573331832886, "global_step": 71489, "epoch": 803, "lr": 8.399720132889559e-05}
{"train_loss": 0.19976550340652466, "global_step": 71490, "epoch": 803, "lr": 8.399677622661894e-05}
{"train_loss": 0.2046518474817276, "global_step": 71491, "epoch": 803, "lr": 8.399635111977183e-05}
{"train_loss": 0.1944994330406189, "global_step": 71492, "epoch": 803, "lr": 8.399592600835433e-05}
{"train_loss": 0.1745615303516388, "global_step": 71493, "epoch": 803, "lr": 8.39955008923665e-05}
{"train_loss": 0.2711346447467804, "global_step": 71494, "epoch": 803, "lr": 8.399507577180838e-05}
{"train_loss": 0.2255602478981018, "global_step": 71495, "epoch": 803, "lr": 8.399465064668004e-05}
{"train_loss": 0.12160858511924744, "global_step": 71496, "epoch": 803, "lr": 8.399422551698152e-05}
{"train_loss": 0.24554187059402466, "global_step": 71497, "epoch": 803, "lr": 8.399380038271291e-05}
{"train_loss": 0.1992541253566742, "global_step": 71498, "epoch": 803, "lr": 8.399337524387426e-05}
{"train_loss": 0.27095112204551697, "global_step": 71499, "epoch": 803, "lr": 8.399295010046558e-05}
{"train_loss": 0.17995785176753998, "global_step": 71500, "epoch": 803, "lr": 8.399252495248698e-05}
{"train_loss": 0.173428475856781, "global_step": 71501, "epoch": 803, "lr": 8.39920997999385e-05}
{"train_loss": 0.1780431866645813, "global_step": 71502, "epoch": 803, "lr": 8.399167464282018e-05}
{"train_loss": 0.25468844175338745, "global_step": 71503, "epoch": 803, "lr": 8.399124948113211e-05}
{"train_loss": 0.26882508397102356, "global_step": 71504, "epoch": 803, "lr": 8.399082431487432e-05}
{"train_loss": 0.23279649019241333, "global_step": 71505, "epoch": 803, "lr": 8.399039914404689e-05}
{"train_loss": 0.18152447044849396, "global_step": 71506, "epoch": 803, "lr": 8.398997396864987e-05}
{"train_loss": 0.22506260871887207, "global_step": 71507, "epoch": 803, "lr": 8.39895487886833e-05}
{"train_loss": 0.2168707400560379, "global_step": 71508, "epoch": 803, "lr": 8.398912360414726e-05}
{"train_loss": 0.14562839269638062, "global_step": 71509, "epoch": 803, "lr": 8.398869841504178e-05}
{"train_loss": 0.19440904259681702, "global_step": 71510, "epoch": 803, "lr": 8.398827322136694e-05}
{"train_loss": 0.15911981463432312, "global_step": 71511, "epoch": 803, "lr": 8.39878480231228e-05}
{"train_loss": 0.14659838378429413, "global_step": 71512, "epoch": 803, "lr": 8.39874228203094e-05}
{"train_loss": 0.21480844914913177, "global_step": 71513, "epoch": 803, "lr": 8.398699761292682e-05}
{"train_loss": 0.1880645900964737, "global_step": 71514, "epoch": 803, "lr": 8.398657240097507e-05}
{"train_loss": 0.22734110057353973, "global_step": 71515, "epoch": 803, "lr": 8.398614718445427e-05}
{"train_loss": 0.18219289183616638, "global_step": 71516, "epoch": 803, "lr": 8.398572196336443e-05}
{"train_loss": 0.18578390777111053, "global_step": 71517, "epoch": 803, "lr": 8.398529673770563e-05}
{"train_loss": 0.1737416386604309, "global_step": 71518, "epoch": 803, "lr": 8.398487150747792e-05}
{"train_loss": 0.241875559091568, "global_step": 71519, "epoch": 803, "lr": 8.398444627268137e-05}
{"train_loss": 0.17407295107841492, "global_step": 71520, "epoch": 803, "lr": 8.398402103331601e-05}
{"train_loss": 0.2267107516527176, "global_step": 71521, "epoch": 803, "lr": 8.398359578938191e-05}
{"train_loss": 0.2602613568305969, "global_step": 71522, "epoch": 803, "lr": 8.398317054087915e-05}
{"train_loss": 0.2633127272129059, "global_step": 71523, "epoch": 803, "lr": 8.398274528780774e-05}
{"train_loss": 0.148007333278656, "global_step": 71524, "epoch": 803, "lr": 8.398232003016778e-05}
{"train_loss": 0.16964682936668396, "global_step": 71525, "epoch": 803, "lr": 8.398189476795932e-05}
{"train_loss": 0.2926672399044037, "global_step": 71526, "epoch": 803, "lr": 8.398146950118239e-05}
{"train_loss": 0.20354938507080078, "global_step": 71527, "epoch": 803, "lr": 8.398104422983709e-05}
{"train_loss": 0.2237391471862793, "global_step": 71528, "epoch": 803, "lr": 8.398061895392341e-05}
{"train_loss": 0.23505078256130219, "global_step": 71529, "epoch": 803, "lr": 8.39801936734415e-05}
{"train_loss": 0.32917770743370056, "global_step": 71530, "epoch": 803, "lr": 8.397976838839132e-05}
{"train_loss": 0.26284968852996826, "global_step": 71531, "epoch": 803, "lr": 8.3979343098773e-05}
{"train_loss": 0.18954704701900482, "global_step": 71532, "epoch": 803, "lr": 8.397891780458658e-05}
{"train_loss": 0.12828496098518372, "global_step": 71533, "epoch": 803, "lr": 8.39784925058321e-05}
{"train_loss": 0.21468956768512726, "global_step": 71534, "epoch": 803, "lr": 8.397806720250963e-05}
{"train_loss": 0.3146662414073944, "global_step": 71535, "epoch": 803, "lr": 8.39776418946192e-05}
{"train_loss": 0.1435251533985138, "global_step": 71536, "epoch": 803, "lr": 8.397721658216091e-05}
{"train_loss": 0.18644866347312927, "global_step": 71537, "epoch": 803, "lr": 8.39767912651348e-05}
{"train_loss": 0.2791973948478699, "global_step": 71538, "epoch": 803, "lr": 8.397636594354092e-05}
{"train_loss": 0.1505109667778015, "global_step": 71539, "epoch": 803, "lr": 8.397594061737932e-05}
{"train_loss": 0.24161630868911743, "global_step": 71540, "epoch": 803, "lr": 8.397551528665008e-05}
{"train_loss": 0.3242507576942444, "global_step": 71541, "epoch": 803, "lr": 8.397508995135325e-05}
{"train_loss": 0.2826298475265503, "global_step": 71542, "epoch": 803, "lr": 8.397466461148887e-05}
{"train_loss": 0.2124427855014801, "global_step": 71543, "epoch": 803, "lr": 8.397423926705703e-05}
{"train_loss": 0.2252623736858368, "global_step": 71544, "epoch": 803, "lr": 8.397381391805774e-05}
{"train_loss": 0.2809985280036926, "global_step": 71545, "epoch": 803, "lr": 8.39733885644911e-05}
{"train_loss": 0.1691933572292328, "global_step": 71546, "epoch": 803, "lr": 8.397296320635716e-05}
{"train_loss": 0.17337195575237274, "global_step": 71547, "epoch": 803, "lr": 8.397253784365598e-05}
{"train_loss": 0.2552829682826996, "global_step": 71548, "epoch": 803, "lr": 8.397211247638758e-05}
{"train_loss": 0.26336368918418884, "global_step": 71549, "epoch": 803, "lr": 8.397168710455206e-05}
{"train_loss": 0.24891695380210876, "global_step": 71550, "epoch": 803, "lr": 8.397126172814945e-05}
{"train_loss": 0.2291756421327591, "global_step": 71551, "epoch": 803, "lr": 8.397083634717981e-05}
{"train_loss": 0.22567680478096008, "global_step": 71552, "epoch": 803, "lr": 8.397041096164323e-05}
{"train_loss": 0.11691760271787643, "global_step": 71553, "epoch": 803, "lr": 8.396998557153973e-05}
{"train_loss": 0.20704735815525055, "global_step": 71554, "epoch": 803, "lr": 8.396956017686936e-05}
{"train_loss": 0.21722208524353048, "global_step": 71555, "epoch": 803, "lr": 8.396913477763222e-05, "val_loss": 3.849677324295044}
{"train_loss": 0.2208491712808609, "global_step": 71556, "epoch": 804, "lr": 8.396870937382835e-05}
{"train_loss": 0.2050325721502304, "global_step": 71557, "epoch": 804, "lr": 8.396828396545778e-05}
{"train_loss": 0.41563042998313904, "global_step": 71558, "epoch": 804, "lr": 8.396785855252059e-05}
{"train_loss": 0.20217330753803253, "global_step": 71559, "epoch": 804, "lr": 8.396743313501684e-05}
{"train_loss": 0.1704733669757843, "global_step": 71560, "epoch": 804, "lr": 8.396700771294659e-05}
{"train_loss": 0.1532738208770752, "global_step": 71561, "epoch": 804, "lr": 8.39665822863099e-05}
{"train_loss": 0.19113346934318542, "global_step": 71562, "epoch": 804, "lr": 8.396615685510679e-05}
{"train_loss": 0.29949983954429626, "global_step": 71563, "epoch": 804, "lr": 8.396573141933735e-05}
{"train_loss": 0.347548246383667, "global_step": 71564, "epoch": 804, "lr": 8.396530597900166e-05}
{"train_loss": 0.20567171275615692, "global_step": 71565, "epoch": 804, "lr": 8.39648805340997e-05}
{"train_loss": 0.3052241802215576, "global_step": 71566, "epoch": 804, "lr": 8.396445508463162e-05}
{"train_loss": 0.21099351346492767, "global_step": 71567, "epoch": 804, "lr": 8.396402963059741e-05}
{"train_loss": 0.19168706238269806, "global_step": 71568, "epoch": 804, "lr": 8.396360417199716e-05}
{"train_loss": 0.12966389954090118, "global_step": 71569, "epoch": 804, "lr": 8.39631787088309e-05}
{"train_loss": 0.14445056021213531, "global_step": 71570, "epoch": 804, "lr": 8.396275324109872e-05}
{"train_loss": 0.31876957416534424, "global_step": 71571, "epoch": 804, "lr": 8.396232776880066e-05}
{"train_loss": 0.2569572329521179, "global_step": 71572, "epoch": 804, "lr": 8.39619022919368e-05}
{"train_loss": 0.2276591658592224, "global_step": 71573, "epoch": 804, "lr": 8.396147681050714e-05}
{"train_loss": 0.22430917620658875, "global_step": 71574, "epoch": 804, "lr": 8.39610513245118e-05}
{"train_loss": 0.2321484386920929, "global_step": 71575, "epoch": 804, "lr": 8.396062583395077e-05}
{"train_loss": 0.1903543323278427, "global_step": 71576, "epoch": 804, "lr": 8.396020033882419e-05}
{"train_loss": 0.22447504103183746, "global_step": 71577, "epoch": 804, "lr": 8.395977483913207e-05}
{"train_loss": 0.24507242441177368, "global_step": 71578, "epoch": 804, "lr": 8.395934933487446e-05}
{"train_loss": 0.16774815320968628, "global_step": 71579, "epoch": 804, "lr": 8.395892382605142e-05}
{"train_loss": 0.18672716617584229, "global_step": 71580, "epoch": 804, "lr": 8.395849831266304e-05}
{"train_loss": 0.24742397665977478, "global_step": 71581, "epoch": 804, "lr": 8.395807279470935e-05}
{"train_loss": 0.22912980616092682, "global_step": 71582, "epoch": 804, "lr": 8.39576472721904e-05}
{"train_loss": 0.17549386620521545, "global_step": 71583, "epoch": 804, "lr": 8.395722174510625e-05}
{"train_loss": 0.2537779211997986, "global_step": 71584, "epoch": 804, "lr": 8.395679621345698e-05}
{"train_loss": 0.2395884394645691, "global_step": 71585, "epoch": 804, "lr": 8.395637067724264e-05}
{"train_loss": 0.1922370195388794, "global_step": 71586, "epoch": 804, "lr": 8.395594513646327e-05}
{"train_loss": 0.2253599613904953, "global_step": 71587, "epoch": 804, "lr": 8.395551959111894e-05}
{"train_loss": 0.14890022575855255, "global_step": 71588, "epoch": 804, "lr": 8.39550940412097e-05}
{"train_loss": 0.22752654552459717, "global_step": 71589, "epoch": 804, "lr": 8.395466848673561e-05}
{"train_loss": 0.27031493186950684, "global_step": 71590, "epoch": 804, "lr": 8.395424292769673e-05}
{"train_loss": 0.2610832750797272, "global_step": 71591, "epoch": 804, "lr": 8.395381736409313e-05}
{"train_loss": 0.1576501578092575, "global_step": 71592, "epoch": 804, "lr": 8.395339179592483e-05}
{"train_loss": 0.2530004680156708, "global_step": 71593, "epoch": 804, "lr": 8.395296622319191e-05}
{"train_loss": 0.25631263852119446, "global_step": 71594, "epoch": 804, "lr": 8.395254064589444e-05}
{"train_loss": 0.14000369608402252, "global_step": 71595, "epoch": 804, "lr": 8.395211506403247e-05}
{"train_loss": 0.20562127232551575, "global_step": 71596, "epoch": 804, "lr": 8.395168947760604e-05}
{"train_loss": 0.17354759573936462, "global_step": 71597, "epoch": 804, "lr": 8.395126388661523e-05}
{"train_loss": 0.19065846502780914, "global_step": 71598, "epoch": 804, "lr": 8.395083829106007e-05}
{"train_loss": 0.16355955600738525, "global_step": 71599, "epoch": 804, "lr": 8.395041269094066e-05}
{"train_loss": 0.18979668617248535, "global_step": 71600, "epoch": 804, "lr": 8.3949987086257e-05}
{"train_loss": 0.1875048130750656, "global_step": 71601, "epoch": 804, "lr": 8.39495614770092e-05}
{"train_loss": 0.14029093086719513, "global_step": 71602, "epoch": 804, "lr": 8.39491358631973e-05}
{"train_loss": 0.21999159455299377, "global_step": 71603, "epoch": 804, "lr": 8.394871024482133e-05}
{"train_loss": 0.19291381537914276, "global_step": 71604, "epoch": 804, "lr": 8.394828462188138e-05}
{"train_loss": 0.1727941632270813, "global_step": 71605, "epoch": 804, "lr": 8.394785899437749e-05}
{"train_loss": 0.20312021672725677, "global_step": 71606, "epoch": 804, "lr": 8.394743336230974e-05}
{"train_loss": 0.18226492404937744, "global_step": 71607, "epoch": 804, "lr": 8.394700772567816e-05}
{"train_loss": 0.21892917156219482, "global_step": 71608, "epoch": 804, "lr": 8.394658208448283e-05}
{"train_loss": 0.2722766101360321, "global_step": 71609, "epoch": 804, "lr": 8.394615643872379e-05}
{"train_loss": 0.2783430814743042, "global_step": 71610, "epoch": 804, "lr": 8.39457307884011e-05}
{"train_loss": 0.20016610622406006, "global_step": 71611, "epoch": 804, "lr": 8.394530513351482e-05}
{"train_loss": 0.1814330518245697, "global_step": 71612, "epoch": 804, "lr": 8.3944879474065e-05}
{"train_loss": 0.25838935375213623, "global_step": 71613, "epoch": 804, "lr": 8.394445381005172e-05}
{"train_loss": 0.2056511640548706, "global_step": 71614, "epoch": 804, "lr": 8.394402814147501e-05}
{"train_loss": 0.22568485140800476, "global_step": 71615, "epoch": 804, "lr": 8.394360246833495e-05}
{"train_loss": 0.25694021582603455, "global_step": 71616, "epoch": 804, "lr": 8.394317679063158e-05}
{"train_loss": 0.2586652636528015, "global_step": 71617, "epoch": 804, "lr": 8.394275110836496e-05}
{"train_loss": 0.15281197428703308, "global_step": 71618, "epoch": 804, "lr": 8.394232542153515e-05}
{"train_loss": 0.2313871532678604, "global_step": 71619, "epoch": 804, "lr": 8.394189973014221e-05}
{"train_loss": 0.23596908152103424, "global_step": 71620, "epoch": 804, "lr": 8.39414740341862e-05}
{"train_loss": 0.19316285848617554, "global_step": 71621, "epoch": 804, "lr": 8.394104833366718e-05}
{"train_loss": 0.26829734444618225, "global_step": 71622, "epoch": 804, "lr": 8.394062262858518e-05}
{"train_loss": 0.21318663656711578, "global_step": 71623, "epoch": 804, "lr": 8.394019691894029e-05}
{"train_loss": 0.16787207126617432, "global_step": 71624, "epoch": 804, "lr": 8.393977120473256e-05}
{"train_loss": 0.29383721947669983, "global_step": 71625, "epoch": 804, "lr": 8.393934548596203e-05}
{"train_loss": 0.1646755486726761, "global_step": 71626, "epoch": 804, "lr": 8.393891976262877e-05}
{"train_loss": 0.23814603686332703, "global_step": 71627, "epoch": 804, "lr": 8.393849403473284e-05}
{"train_loss": 0.22617575526237488, "global_step": 71628, "epoch": 804, "lr": 8.393806830227429e-05}
{"train_loss": 0.18945762515068054, "global_step": 71629, "epoch": 804, "lr": 8.393764256525317e-05}
{"train_loss": 0.31639862060546875, "global_step": 71630, "epoch": 804, "lr": 8.393721682366955e-05}
{"train_loss": 0.18052871525287628, "global_step": 71631, "epoch": 804, "lr": 8.39367910775235e-05}
{"train_loss": 0.3239451050758362, "global_step": 71632, "epoch": 804, "lr": 8.393636532681505e-05}
{"train_loss": 0.2127620130777359, "global_step": 71633, "epoch": 804, "lr": 8.393593957154428e-05}
{"train_loss": 0.20136815309524536, "global_step": 71634, "epoch": 804, "lr": 8.393551381171122e-05}
{"train_loss": 0.26639610528945923, "global_step": 71635, "epoch": 804, "lr": 8.393508804731597e-05}
{"train_loss": 0.24894803762435913, "global_step": 71636, "epoch": 804, "lr": 8.393466227835855e-05}
{"train_loss": 0.24037204682826996, "global_step": 71637, "epoch": 804, "lr": 8.393423650483903e-05}
{"train_loss": 0.26105669140815735, "global_step": 71638, "epoch": 804, "lr": 8.393381072675745e-05}
{"train_loss": 0.28655585646629333, "global_step": 71639, "epoch": 804, "lr": 8.39333849441139e-05}
{"train_loss": 0.24067358672618866, "global_step": 71640, "epoch": 804, "lr": 8.39329591569084e-05}
{"train_loss": 0.26183852553367615, "global_step": 71641, "epoch": 804, "lr": 8.393253336514104e-05}
{"train_loss": 0.23965726792812347, "global_step": 71642, "epoch": 804, "lr": 8.393210756881187e-05}
{"train_loss": 0.107076495885849, "global_step": 71643, "epoch": 804, "lr": 8.393168176792093e-05}
{"train_loss": 0.22078131557850356, "global_step": 71644, "epoch": 804, "lr": 8.39312559624683e-05, "val_loss": 3.8239550590515137}
{"train_loss": 0.22323693335056305, "global_step": 71645, "epoch": 805, "lr": 8.393083015245403e-05}
{"train_loss": 0.21500539779663086, "global_step": 71646, "epoch": 805, "lr": 8.393040433787815e-05}
{"train_loss": 0.1890019029378891, "global_step": 71647, "epoch": 805, "lr": 8.392997851874076e-05}
{"train_loss": 0.24076375365257263, "global_step": 71648, "epoch": 805, "lr": 8.392955269504189e-05}
{"train_loss": 0.31583648920059204, "global_step": 71649, "epoch": 805, "lr": 8.392912686678161e-05}
{"train_loss": 0.26413804292678833, "global_step": 71650, "epoch": 805, "lr": 8.392870103395997e-05}
{"train_loss": 0.1717434823513031, "global_step": 71651, "epoch": 805, "lr": 8.392827519657702e-05}
{"train_loss": 0.19893209636211395, "global_step": 71652, "epoch": 805, "lr": 8.392784935463285e-05}
{"train_loss": 0.32770466804504395, "global_step": 71653, "epoch": 805, "lr": 8.392742350812748e-05}
{"train_loss": 0.17358845472335815, "global_step": 71654, "epoch": 805, "lr": 8.392699765706098e-05}
{"train_loss": 0.1695140153169632, "global_step": 71655, "epoch": 805, "lr": 8.392657180143341e-05}
{"train_loss": 0.20211905241012573, "global_step": 71656, "epoch": 805, "lr": 8.392614594124484e-05}
{"train_loss": 0.19698511064052582, "global_step": 71657, "epoch": 805, "lr": 8.392572007649529e-05}
{"train_loss": 0.20411823689937592, "global_step": 71658, "epoch": 805, "lr": 8.392529420718484e-05}
{"train_loss": 0.26902398467063904, "global_step": 71659, "epoch": 805, "lr": 8.392486833331357e-05}
{"train_loss": 0.15522269904613495, "global_step": 71660, "epoch": 805, "lr": 8.392444245488151e-05}
{"train_loss": 0.29726776480674744, "global_step": 71661, "epoch": 805, "lr": 8.392401657188871e-05}
{"train_loss": 0.16286440193653107, "global_step": 71662, "epoch": 805, "lr": 8.392359068433524e-05}
{"train_loss": 0.18670524656772614, "global_step": 71663, "epoch": 805, "lr": 8.392316479222117e-05}
{"train_loss": 0.12984633445739746, "global_step": 71664, "epoch": 805, "lr": 8.392273889554654e-05}
{"train_loss": 0.18806156516075134, "global_step": 71665, "epoch": 805, "lr": 8.39223129943114e-05}
{"train_loss": 0.28467848896980286, "global_step": 71666, "epoch": 805, "lr": 8.392188708851583e-05}
{"train_loss": 0.2062969207763672, "global_step": 71667, "epoch": 805, "lr": 8.392146117815987e-05}
{"train_loss": 0.19055528938770294, "global_step": 71668, "epoch": 805, "lr": 8.392103526324358e-05}
{"train_loss": 0.2274833619594574, "global_step": 71669, "epoch": 805, "lr": 8.392060934376702e-05}
{"train_loss": 0.17684637010097504, "global_step": 71670, "epoch": 805, "lr": 8.392018341973026e-05}
{"train_loss": 0.23980414867401123, "global_step": 71671, "epoch": 805, "lr": 8.391975749113333e-05}
{"train_loss": 0.19262701272964478, "global_step": 71672, "epoch": 805, "lr": 8.391933155797631e-05}
{"train_loss": 0.2786804735660553, "global_step": 71673, "epoch": 805, "lr": 8.391890562025926e-05}
{"train_loss": 0.15362876653671265, "global_step": 71674, "epoch": 805, "lr": 8.391847967798221e-05}
{"train_loss": 0.2626558840274811, "global_step": 71675, "epoch": 805, "lr": 8.391805373114523e-05}
{"train_loss": 0.2138347625732422, "global_step": 71676, "epoch": 805, "lr": 8.39176277797484e-05}
{"train_loss": 0.19537201523780823, "global_step": 71677, "epoch": 805, "lr": 8.391720182379175e-05}
{"train_loss": 0.2112208902835846, "global_step": 71678, "epoch": 805, "lr": 8.391677586327533e-05}
{"train_loss": 0.304708331823349, "global_step": 71679, "epoch": 805, "lr": 8.391634989819923e-05}
{"train_loss": 0.15159957110881805, "global_step": 71680, "epoch": 805, "lr": 8.391592392856348e-05}
{"train_loss": 0.24230387806892395, "global_step": 71681, "epoch": 805, "lr": 8.391549795436817e-05}
{"train_loss": 0.20785632729530334, "global_step": 71682, "epoch": 805, "lr": 8.391507197561331e-05}
{"train_loss": 0.20395401120185852, "global_step": 71683, "epoch": 805, "lr": 8.391464599229898e-05}
{"train_loss": 0.20991960167884827, "global_step": 71684, "epoch": 805, "lr": 8.391422000442526e-05}
{"train_loss": 0.2319600135087967, "global_step": 71685, "epoch": 805, "lr": 8.391379401199218e-05}
{"train_loss": 0.209425151348114, "global_step": 71686, "epoch": 805, "lr": 8.39133680149998e-05}
{"train_loss": 0.16841046512126923, "global_step": 71687, "epoch": 805, "lr": 8.391294201344817e-05}
{"train_loss": 0.20312266051769257, "global_step": 71688, "epoch": 805, "lr": 8.391251600733738e-05}
{"train_loss": 0.1788375824689865, "global_step": 71689, "epoch": 805, "lr": 8.391208999666746e-05}
{"train_loss": 0.24128960072994232, "global_step": 71690, "epoch": 805, "lr": 8.391166398143847e-05}
{"train_loss": 0.24899375438690186, "global_step": 71691, "epoch": 805, "lr": 8.391123796165046e-05}
{"train_loss": 0.2366190254688263, "global_step": 71692, "epoch": 805, "lr": 8.39108119373035e-05}
{"train_loss": 0.18492752313613892, "global_step": 71693, "epoch": 805, "lr": 8.391038590839766e-05}
{"train_loss": 0.2699195146560669, "global_step": 71694, "epoch": 805, "lr": 8.390995987493297e-05}
{"train_loss": 0.2203323096036911, "global_step": 71695, "epoch": 805, "lr": 8.390953383690949e-05}
{"train_loss": 0.19473513960838318, "global_step": 71696, "epoch": 805, "lr": 8.39091077943273e-05}
{"train_loss": 0.21718311309814453, "global_step": 71697, "epoch": 805, "lr": 8.390868174718645e-05}
{"train_loss": 0.174377903342247, "global_step": 71698, "epoch": 805, "lr": 8.390825569548698e-05}
{"train_loss": 0.21305888891220093, "global_step": 71699, "epoch": 805, "lr": 8.390782963922896e-05}
{"train_loss": 0.22288772463798523, "global_step": 71700, "epoch": 805, "lr": 8.390740357841245e-05}
{"train_loss": 0.17314468324184418, "global_step": 71701, "epoch": 805, "lr": 8.39069775130375e-05}
{"train_loss": 0.20419177412986755, "global_step": 71702, "epoch": 805, "lr": 8.390655144310416e-05}
{"train_loss": 0.15689441561698914, "global_step": 71703, "epoch": 805, "lr": 8.390612536861251e-05}
{"train_loss": 0.19225968420505524, "global_step": 71704, "epoch": 805, "lr": 8.390569928956259e-05}
{"train_loss": 0.19103579223155975, "global_step": 71705, "epoch": 805, "lr": 8.390527320595447e-05}
{"train_loss": 0.17313694953918457, "global_step": 71706, "epoch": 805, "lr": 8.390484711778819e-05}
{"train_loss": 0.17343828082084656, "global_step": 71707, "epoch": 805, "lr": 8.390442102506381e-05}
{"train_loss": 0.23397740721702576, "global_step": 71708, "epoch": 805, "lr": 8.390399492778141e-05}
{"train_loss": 0.16601823270320892, "global_step": 71709, "epoch": 805, "lr": 8.390356882594102e-05}
{"train_loss": 0.1742451786994934, "global_step": 71710, "epoch": 805, "lr": 8.390314271954271e-05}
{"train_loss": 0.1684592217206955, "global_step": 71711, "epoch": 805, "lr": 8.390271660858654e-05}
{"train_loss": 0.26977789402008057, "global_step": 71712, "epoch": 805, "lr": 8.390229049307256e-05}
{"train_loss": 0.20425042510032654, "global_step": 71713, "epoch": 805, "lr": 8.390186437300082e-05}
{"train_loss": 0.20703615248203278, "global_step": 71714, "epoch": 805, "lr": 8.390143824837143e-05}
{"train_loss": 0.22404691576957703, "global_step": 71715, "epoch": 805, "lr": 8.390101211918437e-05}
{"train_loss": 0.15198546648025513, "global_step": 71716, "epoch": 805, "lr": 8.390058598543973e-05}
{"train_loss": 0.1639658659696579, "global_step": 71717, "epoch": 805, "lr": 8.390015984713757e-05}
{"train_loss": 0.1950608640909195, "global_step": 71718, "epoch": 805, "lr": 8.389973370427795e-05}
{"train_loss": 0.22137777507305145, "global_step": 71719, "epoch": 805, "lr": 8.389930755686092e-05}
{"train_loss": 0.24734313786029816, "global_step": 71720, "epoch": 805, "lr": 8.389888140488655e-05}
{"train_loss": 0.2313971370458603, "global_step": 71721, "epoch": 805, "lr": 8.389845524835488e-05}
{"train_loss": 0.25129392743110657, "global_step": 71722, "epoch": 805, "lr": 8.389802908726598e-05}
{"train_loss": 0.15149565041065216, "global_step": 71723, "epoch": 805, "lr": 8.389760292161992e-05}
{"train_loss": 0.26147839426994324, "global_step": 71724, "epoch": 805, "lr": 8.389717675141673e-05}
{"train_loss": 0.19900794327259064, "global_step": 71725, "epoch": 805, "lr": 8.389675057665646e-05}
{"train_loss": 0.17114496231079102, "global_step": 71726, "epoch": 805, "lr": 8.389632439733922e-05}
{"train_loss": 0.17783719301223755, "global_step": 71727, "epoch": 805, "lr": 8.3895898213465e-05}
{"train_loss": 0.2800051271915436, "global_step": 71728, "epoch": 805, "lr": 8.38954720250339e-05}
{"train_loss": 0.2036401629447937, "global_step": 71729, "epoch": 805, "lr": 8.389504583204596e-05}
{"train_loss": 0.1271972358226776, "global_step": 71730, "epoch": 805, "lr": 8.389461963450125e-05}
{"train_loss": 0.1986968219280243, "global_step": 71731, "epoch": 805, "lr": 8.389419343239981e-05}
{"train_loss": 0.12583772838115692, "global_step": 71732, "epoch": 805, "lr": 8.389376722574174e-05}
{"train_loss": 0.20894233915912971, "global_step": 71733, "epoch": 805, "lr": 8.389334101452703e-05, "val_loss": 3.677129030227661, "train_action_mse_error": 20.029983520507812}
{"train_loss": 0.16338585317134857, "global_step": 71734, "epoch": 806, "lr": 8.38929147987558e-05}
{"train_loss": 0.20131544768810272, "global_step": 71735, "epoch": 806, "lr": 8.389248857842808e-05}
{"train_loss": 0.18504714965820312, "global_step": 71736, "epoch": 806, "lr": 8.389206235354392e-05}
{"train_loss": 0.22571872174739838, "global_step": 71737, "epoch": 806, "lr": 8.389163612410338e-05}
{"train_loss": 0.12323267757892609, "global_step": 71738, "epoch": 806, "lr": 8.389120989010653e-05}
{"train_loss": 0.17191165685653687, "global_step": 71739, "epoch": 806, "lr": 8.389078365155341e-05}
{"train_loss": 0.17846353352069855, "global_step": 71740, "epoch": 806, "lr": 8.389035740844409e-05}
{"train_loss": 0.17610982060432434, "global_step": 71741, "epoch": 806, "lr": 8.388993116077863e-05}
{"train_loss": 0.21161341667175293, "global_step": 71742, "epoch": 806, "lr": 8.38895049085571e-05}
{"train_loss": 0.15229324996471405, "global_step": 71743, "epoch": 806, "lr": 8.38890786517795e-05}
{"train_loss": 0.22071707248687744, "global_step": 71744, "epoch": 806, "lr": 8.388865239044595e-05}
{"train_loss": 0.20656180381774902, "global_step": 71745, "epoch": 806, "lr": 8.388822612455649e-05}
{"train_loss": 0.12478631734848022, "global_step": 71746, "epoch": 806, "lr": 8.388779985411114e-05}
{"train_loss": 0.22270286083221436, "global_step": 71747, "epoch": 806, "lr": 8.388737357911003e-05}
{"train_loss": 0.23807096481323242, "global_step": 71748, "epoch": 806, "lr": 8.388694729955315e-05}
{"train_loss": 0.30307844281196594, "global_step": 71749, "epoch": 806, "lr": 8.388652101544059e-05}
{"train_loss": 0.30162930488586426, "global_step": 71750, "epoch": 806, "lr": 8.38860947267724e-05}
{"train_loss": 0.20634707808494568, "global_step": 71751, "epoch": 806, "lr": 8.388566843354864e-05}
{"train_loss": 0.22163818776607513, "global_step": 71752, "epoch": 806, "lr": 8.388524213576936e-05}
{"train_loss": 0.19774115085601807, "global_step": 71753, "epoch": 806, "lr": 8.388481583343462e-05}
{"train_loss": 0.15298619866371155, "global_step": 71754, "epoch": 806, "lr": 8.388438952654448e-05}
{"train_loss": 0.15215149521827698, "global_step": 71755, "epoch": 806, "lr": 8.3883963215099e-05}
{"train_loss": 0.27623453736305237, "global_step": 71756, "epoch": 806, "lr": 8.388353689909824e-05}
{"train_loss": 0.1292334794998169, "global_step": 71757, "epoch": 806, "lr": 8.388311057854224e-05}
{"train_loss": 0.14355432987213135, "global_step": 71758, "epoch": 806, "lr": 8.388268425343106e-05}
{"train_loss": 0.16050401329994202, "global_step": 71759, "epoch": 806, "lr": 8.388225792376478e-05}
{"train_loss": 0.2537383735179901, "global_step": 71760, "epoch": 806, "lr": 8.388183158954346e-05}
{"train_loss": 0.1563146710395813, "global_step": 71761, "epoch": 806, "lr": 8.388140525076713e-05}
{"train_loss": 0.21556328237056732, "global_step": 71762, "epoch": 806, "lr": 8.388097890743584e-05}
{"train_loss": 0.13664610683918, "global_step": 71763, "epoch": 806, "lr": 8.388055255954967e-05}
{"train_loss": 0.1425398588180542, "global_step": 71764, "epoch": 806, "lr": 8.388012620710869e-05}
{"train_loss": 0.2396848052740097, "global_step": 71765, "epoch": 806, "lr": 8.387969985011292e-05}
{"train_loss": 0.22493065893650055, "global_step": 71766, "epoch": 806, "lr": 8.387927348856246e-05}
{"train_loss": 0.1922857165336609, "global_step": 71767, "epoch": 806, "lr": 8.387884712245733e-05}
{"train_loss": 0.14342010021209717, "global_step": 71768, "epoch": 806, "lr": 8.387842075179759e-05}
{"train_loss": 0.20147095620632172, "global_step": 71769, "epoch": 806, "lr": 8.387799437658332e-05}
{"train_loss": 0.1575394570827484, "global_step": 71770, "epoch": 806, "lr": 8.387756799681459e-05}
{"train_loss": 0.14977507293224335, "global_step": 71771, "epoch": 806, "lr": 8.387714161249141e-05}
{"train_loss": 0.13335220515727997, "global_step": 71772, "epoch": 806, "lr": 8.387671522361386e-05}
{"train_loss": 0.22267410159111023, "global_step": 71773, "epoch": 806, "lr": 8.387628883018202e-05}
{"train_loss": 0.17420607805252075, "global_step": 71774, "epoch": 806, "lr": 8.38758624321959e-05}
{"train_loss": 0.24201075732707977, "global_step": 71775, "epoch": 806, "lr": 8.387543602965557e-05}
{"train_loss": 0.18133653700351715, "global_step": 71776, "epoch": 806, "lr": 8.387500962256114e-05}
{"train_loss": 0.21071912348270416, "global_step": 71777, "epoch": 806, "lr": 8.38745832109126e-05}
{"train_loss": 0.16095328330993652, "global_step": 71778, "epoch": 806, "lr": 8.387415679471005e-05}
{"train_loss": 0.19955961406230927, "global_step": 71779, "epoch": 806, "lr": 8.387373037395352e-05}
{"train_loss": 0.2358538657426834, "global_step": 71780, "epoch": 806, "lr": 8.38733039486431e-05}
{"train_loss": 0.2501440644264221, "global_step": 71781, "epoch": 806, "lr": 8.387287751877882e-05}
{"train_loss": 0.26398995518684387, "global_step": 71782, "epoch": 806, "lr": 8.387245108436074e-05}
{"train_loss": 0.16862569749355316, "global_step": 71783, "epoch": 806, "lr": 8.387202464538892e-05}
{"train_loss": 0.1551203429698944, "global_step": 71784, "epoch": 806, "lr": 8.387159820186341e-05}
{"train_loss": 0.13584846258163452, "global_step": 71785, "epoch": 806, "lr": 8.387117175378429e-05}
{"train_loss": 0.1671704202890396, "global_step": 71786, "epoch": 806, "lr": 8.38707453011516e-05}
{"train_loss": 0.17912667989730835, "global_step": 71787, "epoch": 806, "lr": 8.387031884396541e-05}
{"train_loss": 0.2288385033607483, "global_step": 71788, "epoch": 806, "lr": 8.386989238222575e-05}
{"train_loss": 0.1884041130542755, "global_step": 71789, "epoch": 806, "lr": 8.386946591593273e-05}
{"train_loss": 0.23492874205112457, "global_step": 71790, "epoch": 806, "lr": 8.386903944508632e-05}
{"train_loss": 0.21210519969463348, "global_step": 71791, "epoch": 806, "lr": 8.386861296968666e-05}
{"train_loss": 0.18832622468471527, "global_step": 71792, "epoch": 806, "lr": 8.38681864897338e-05}
{"train_loss": 0.1661795824766159, "global_step": 71793, "epoch": 806, "lr": 8.386776000522774e-05}
{"train_loss": 0.1707286536693573, "global_step": 71794, "epoch": 806, "lr": 8.386733351616858e-05}
{"train_loss": 0.23682886362075806, "global_step": 71795, "epoch": 806, "lr": 8.386690702255637e-05}
{"train_loss": 0.18165072798728943, "global_step": 71796, "epoch": 806, "lr": 8.386648052439119e-05}
{"train_loss": 0.20756329596042633, "global_step": 71797, "epoch": 806, "lr": 8.386605402167304e-05}
{"train_loss": 0.16911664605140686, "global_step": 71798, "epoch": 806, "lr": 8.386562751440204e-05}
{"train_loss": 0.23956629633903503, "global_step": 71799, "epoch": 806, "lr": 8.386520100257821e-05}
{"train_loss": 0.21199242770671844, "global_step": 71800, "epoch": 806, "lr": 8.386477448620161e-05}
{"train_loss": 0.231244757771492, "global_step": 71801, "epoch": 806, "lr": 8.386434796527231e-05}
{"train_loss": 0.13928885757923126, "global_step": 71802, "epoch": 806, "lr": 8.386392143979036e-05}
{"train_loss": 0.17185260355472565, "global_step": 71803, "epoch": 806, "lr": 8.386349490975582e-05}
{"train_loss": 0.1797671914100647, "global_step": 71804, "epoch": 806, "lr": 8.386306837516875e-05}
{"train_loss": 0.26061713695526123, "global_step": 71805, "epoch": 806, "lr": 8.38626418360292e-05}
{"train_loss": 0.17339737713336945, "global_step": 71806, "epoch": 806, "lr": 8.386221529233721e-05}
{"train_loss": 0.24993932247161865, "global_step": 71807, "epoch": 806, "lr": 8.386178874409287e-05}
{"train_loss": 0.1587284356355667, "global_step": 71808, "epoch": 806, "lr": 8.386136219129623e-05}
{"train_loss": 0.2808799147605896, "global_step": 71809, "epoch": 806, "lr": 8.386093563394734e-05}
{"train_loss": 0.2076045572757721, "global_step": 71810, "epoch": 806, "lr": 8.386050907204628e-05}
{"train_loss": 0.16646012663841248, "global_step": 71811, "epoch": 806, "lr": 8.386008250559306e-05}
{"train_loss": 0.29445770382881165, "global_step": 71812, "epoch": 806, "lr": 8.385965593458777e-05}
{"train_loss": 0.3155083656311035, "global_step": 71813, "epoch": 806, "lr": 8.385922935903046e-05}
{"train_loss": 0.30310946702957153, "global_step": 71814, "epoch": 806, "lr": 8.38588027789212e-05}
{"train_loss": 0.13163147866725922, "global_step": 71815, "epoch": 806, "lr": 8.385837619426004e-05}
{"train_loss": 0.17137177288532257, "global_step": 71816, "epoch": 806, "lr": 8.385794960504703e-05}
{"train_loss": 0.21457266807556152, "global_step": 71817, "epoch": 806, "lr": 8.385752301128223e-05}
{"train_loss": 0.19186463952064514, "global_step": 71818, "epoch": 806, "lr": 8.385709641296569e-05}
{"train_loss": 0.2048758864402771, "global_step": 71819, "epoch": 806, "lr": 8.385666981009747e-05}
{"train_loss": 0.24395135045051575, "global_step": 71820, "epoch": 806, "lr": 8.385624320267766e-05}
{"train_loss": 0.27551111578941345, "global_step": 71821, "epoch": 806, "lr": 8.385581659070627e-05}
{"train_loss": 0.20061415232969135, "global_step": 71822, "epoch": 806, "lr": 8.385538997418339e-05, "val_loss": 3.877138137817383}
{"train_loss": 0.2237856090068817, "global_step": 71823, "epoch": 807, "lr": 8.385496335310906e-05}
{"train_loss": 0.17212483286857605, "global_step": 71824, "epoch": 807, "lr": 8.385453672748334e-05}
{"train_loss": 0.1970280557870865, "global_step": 71825, "epoch": 807, "lr": 8.38541100973063e-05}
{"train_loss": 0.15716330707073212, "global_step": 71826, "epoch": 807, "lr": 8.385368346257798e-05}
{"train_loss": 0.20079755783081055, "global_step": 71827, "epoch": 807, "lr": 8.385325682329844e-05}
{"train_loss": 0.17400908470153809, "global_step": 71828, "epoch": 807, "lr": 8.385283017946774e-05}
{"train_loss": 0.16841787099838257, "global_step": 71829, "epoch": 807, "lr": 8.385240353108596e-05}
{"train_loss": 0.1667751520872116, "global_step": 71830, "epoch": 807, "lr": 8.385197687815311e-05}
{"train_loss": 0.22386756539344788, "global_step": 71831, "epoch": 807, "lr": 8.38515502206693e-05}
{"train_loss": 0.21932955086231232, "global_step": 71832, "epoch": 807, "lr": 8.385112355863455e-05}
{"train_loss": 0.23128370940685272, "global_step": 71833, "epoch": 807, "lr": 8.385069689204893e-05}
{"train_loss": 0.15374930202960968, "global_step": 71834, "epoch": 807, "lr": 8.385027022091249e-05}
{"train_loss": 0.16042250394821167, "global_step": 71835, "epoch": 807, "lr": 8.38498435452253e-05}
{"train_loss": 0.13602182269096375, "global_step": 71836, "epoch": 807, "lr": 8.384941686498742e-05}
{"train_loss": 0.169686421751976, "global_step": 71837, "epoch": 807, "lr": 8.384899018019889e-05}
{"train_loss": 0.24236921966075897, "global_step": 71838, "epoch": 807, "lr": 8.384856349085977e-05}
{"train_loss": 0.26789766550064087, "global_step": 71839, "epoch": 807, "lr": 8.384813679697013e-05}
{"train_loss": 0.24756063520908356, "global_step": 71840, "epoch": 807, "lr": 8.384771009853002e-05}
{"train_loss": 0.2820914089679718, "global_step": 71841, "epoch": 807, "lr": 8.384728339553948e-05}
{"train_loss": 0.22234241664409637, "global_step": 71842, "epoch": 807, "lr": 8.384685668799861e-05}
{"train_loss": 0.2128051519393921, "global_step": 71843, "epoch": 807, "lr": 8.384642997590744e-05}
{"train_loss": 0.20998640358448029, "global_step": 71844, "epoch": 807, "lr": 8.384600325926602e-05}
{"train_loss": 0.2734855115413666, "global_step": 71845, "epoch": 807, "lr": 8.384557653807444e-05}
{"train_loss": 0.1928592026233673, "global_step": 71846, "epoch": 807, "lr": 8.38451498123327e-05}
{"train_loss": 0.19760023057460785, "global_step": 71847, "epoch": 807, "lr": 8.38447230820409e-05}
{"train_loss": 0.17280787229537964, "global_step": 71848, "epoch": 807, "lr": 8.38442963471991e-05}
{"train_loss": 0.13743682205677032, "global_step": 71849, "epoch": 807, "lr": 8.384386960780734e-05}
{"train_loss": 0.15188996493816376, "global_step": 71850, "epoch": 807, "lr": 8.38434428638657e-05}
{"train_loss": 0.20880697667598724, "global_step": 71851, "epoch": 807, "lr": 8.384301611537422e-05}
{"train_loss": 0.15121400356292725, "global_step": 71852, "epoch": 807, "lr": 8.384258936233293e-05}
{"train_loss": 0.3057231605052948, "global_step": 71853, "epoch": 807, "lr": 8.384216260474194e-05}
{"train_loss": 0.15203680098056793, "global_step": 71854, "epoch": 807, "lr": 8.384173584260129e-05}
{"train_loss": 0.2316620647907257, "global_step": 71855, "epoch": 807, "lr": 8.384130907591101e-05}
{"train_loss": 0.17827390134334564, "global_step": 71856, "epoch": 807, "lr": 8.38408823046712e-05}
{"train_loss": 0.2339787632226944, "global_step": 71857, "epoch": 807, "lr": 8.384045552888189e-05}
{"train_loss": 0.2518792450428009, "global_step": 71858, "epoch": 807, "lr": 8.384002874854313e-05}
{"train_loss": 0.23129448294639587, "global_step": 71859, "epoch": 807, "lr": 8.383960196365501e-05}
{"train_loss": 0.2885168492794037, "global_step": 71860, "epoch": 807, "lr": 8.383917517421756e-05}
{"train_loss": 0.20708994567394257, "global_step": 71861, "epoch": 807, "lr": 8.383874838023084e-05}
{"train_loss": 0.18477977812290192, "global_step": 71862, "epoch": 807, "lr": 8.383832158169491e-05}
{"train_loss": 0.1526431441307068, "global_step": 71863, "epoch": 807, "lr": 8.383789477860984e-05}
{"train_loss": 0.1571197658777237, "global_step": 71864, "epoch": 807, "lr": 8.383746797097567e-05}
{"train_loss": 0.19542056322097778, "global_step": 71865, "epoch": 807, "lr": 8.383704115879247e-05}
{"train_loss": 0.17376869916915894, "global_step": 71866, "epoch": 807, "lr": 8.383661434206028e-05}
{"train_loss": 0.2021065652370453, "global_step": 71867, "epoch": 807, "lr": 8.383618752077918e-05}
{"train_loss": 0.17732331156730652, "global_step": 71868, "epoch": 807, "lr": 8.383576069494922e-05}
{"train_loss": 0.19438929855823517, "global_step": 71869, "epoch": 807, "lr": 8.383533386457046e-05}
{"train_loss": 0.20217250287532806, "global_step": 71870, "epoch": 807, "lr": 8.383490702964293e-05}
{"train_loss": 0.2607501447200775, "global_step": 71871, "epoch": 807, "lr": 8.383448019016672e-05}
{"train_loss": 0.16013726592063904, "global_step": 71872, "epoch": 807, "lr": 8.383405334614189e-05}
{"train_loss": 0.2041149139404297, "global_step": 71873, "epoch": 807, "lr": 8.383362649756846e-05}
{"train_loss": 0.2260534018278122, "global_step": 71874, "epoch": 807, "lr": 8.383319964444653e-05}
{"train_loss": 0.2839297652244568, "global_step": 71875, "epoch": 807, "lr": 8.383277278677612e-05}
{"train_loss": 0.21933124959468842, "global_step": 71876, "epoch": 807, "lr": 8.383234592455733e-05}
{"train_loss": 0.2127794772386551, "global_step": 71877, "epoch": 807, "lr": 8.383191905779018e-05}
{"train_loss": 0.16268305480480194, "global_step": 71878, "epoch": 807, "lr": 8.383149218647474e-05}
{"train_loss": 0.15899457037448883, "global_step": 71879, "epoch": 807, "lr": 8.383106531061106e-05}
{"train_loss": 0.2981847822666168, "global_step": 71880, "epoch": 807, "lr": 8.383063843019923e-05}
{"train_loss": 0.3547613322734833, "global_step": 71881, "epoch": 807, "lr": 8.383021154523927e-05}
{"train_loss": 0.14749857783317566, "global_step": 71882, "epoch": 807, "lr": 8.382978465573126e-05}
{"train_loss": 0.15793263912200928, "global_step": 71883, "epoch": 807, "lr": 8.382935776167522e-05}
{"train_loss": 0.20941105484962463, "global_step": 71884, "epoch": 807, "lr": 8.382893086307126e-05}
{"train_loss": 0.2065691500902176, "global_step": 71885, "epoch": 807, "lr": 8.382850395991941e-05}
{"train_loss": 0.2681225836277008, "global_step": 71886, "epoch": 807, "lr": 8.38280770522197e-05}
{"train_loss": 0.20081433653831482, "global_step": 71887, "epoch": 807, "lr": 8.382765013997226e-05}
{"train_loss": 0.18560533225536346, "global_step": 71888, "epoch": 807, "lr": 8.38272232231771e-05}
{"train_loss": 0.26887786388397217, "global_step": 71889, "epoch": 807, "lr": 8.382679630183425e-05}
{"train_loss": 0.24592097103595734, "global_step": 71890, "epoch": 807, "lr": 8.382636937594382e-05}
{"train_loss": 0.2542370855808258, "global_step": 71891, "epoch": 807, "lr": 8.382594244550585e-05}
{"train_loss": 0.23638297617435455, "global_step": 71892, "epoch": 807, "lr": 8.38255155105204e-05}
{"train_loss": 0.2432553619146347, "global_step": 71893, "epoch": 807, "lr": 8.38250885709875e-05}
{"train_loss": 0.1670827716588974, "global_step": 71894, "epoch": 807, "lr": 8.382466162690724e-05}
{"train_loss": 0.19238944351673126, "global_step": 71895, "epoch": 807, "lr": 8.382423467827967e-05}
{"train_loss": 0.22703249752521515, "global_step": 71896, "epoch": 807, "lr": 8.382380772510485e-05}
{"train_loss": 0.26340460777282715, "global_step": 71897, "epoch": 807, "lr": 8.382338076738282e-05}
{"train_loss": 0.10644073784351349, "global_step": 71898, "epoch": 807, "lr": 8.382295380511365e-05}
{"train_loss": 0.19012348353862762, "global_step": 71899, "epoch": 807, "lr": 8.38225268382974e-05}
{"train_loss": 0.24781399965286255, "global_step": 71900, "epoch": 807, "lr": 8.382209986693412e-05}
{"train_loss": 0.22282883524894714, "global_step": 71901, "epoch": 807, "lr": 8.382167289102388e-05}
{"train_loss": 0.16605690121650696, "global_step": 71902, "epoch": 807, "lr": 8.38212459105667e-05}
{"train_loss": 0.22007408738136292, "global_step": 71903, "epoch": 807, "lr": 8.382081892556269e-05}
{"train_loss": 0.15106281638145447, "global_step": 71904, "epoch": 807, "lr": 8.382039193601188e-05}
{"train_loss": 0.12620505690574646, "global_step": 71905, "epoch": 807, "lr": 8.381996494191433e-05}
{"train_loss": 0.19899031519889832, "global_step": 71906, "epoch": 807, "lr": 8.38195379432701e-05}
{"train_loss": 0.2303735464811325, "global_step": 71907, "epoch": 807, "lr": 8.381911094007923e-05}
{"train_loss": 0.15266509354114532, "global_step": 71908, "epoch": 807, "lr": 8.381868393234182e-05}
{"train_loss": 0.18448279798030853, "global_step": 71909, "epoch": 807, "lr": 8.381825692005788e-05}
{"train_loss": 0.1751755177974701, "global_step": 71910, "epoch": 807, "lr": 8.381782990322749e-05}
{"train_loss": 0.2049525815449404, "global_step": 71911, "epoch": 807, "lr": 8.381740288185071e-05, "val_loss": 3.9066905975341797}
{"train_loss": 0.24900706112384796, "global_step": 71912, "epoch": 808, "lr": 8.38169758559276e-05}
{"train_loss": 0.15810810029506683, "global_step": 71913, "epoch": 808, "lr": 8.38165488254582e-05}
{"train_loss": 0.19685620069503784, "global_step": 71914, "epoch": 808, "lr": 8.381612179044257e-05}
{"train_loss": 0.18856586515903473, "global_step": 71915, "epoch": 808, "lr": 8.381569475088079e-05}
{"train_loss": 0.20824149250984192, "global_step": 71916, "epoch": 808, "lr": 8.38152677067729e-05}
{"train_loss": 0.24885135889053345, "global_step": 71917, "epoch": 808, "lr": 8.381484065811896e-05}
{"train_loss": 0.20522211492061615, "global_step": 71918, "epoch": 808, "lr": 8.381441360491901e-05}
{"train_loss": 0.20666112005710602, "global_step": 71919, "epoch": 808, "lr": 8.381398654717313e-05}
{"train_loss": 0.213234543800354, "global_step": 71920, "epoch": 808, "lr": 8.381355948488136e-05}
{"train_loss": 0.1602189838886261, "global_step": 71921, "epoch": 808, "lr": 8.38131324180438e-05}
{"train_loss": 0.2451450377702713, "global_step": 71922, "epoch": 808, "lr": 8.381270534666046e-05}
{"train_loss": 0.21476687490940094, "global_step": 71923, "epoch": 808, "lr": 8.381227827073141e-05}
{"train_loss": 0.1383112221956253, "global_step": 71924, "epoch": 808, "lr": 8.381185119025672e-05}
{"train_loss": 0.25899738073349, "global_step": 71925, "epoch": 808, "lr": 8.381142410523643e-05}
{"train_loss": 0.1488248109817505, "global_step": 71926, "epoch": 808, "lr": 8.381099701567061e-05}
{"train_loss": 0.22386425733566284, "global_step": 71927, "epoch": 808, "lr": 8.381056992155933e-05}
{"train_loss": 0.17710332572460175, "global_step": 71928, "epoch": 808, "lr": 8.38101428229026e-05}
{"train_loss": 0.14981332421302795, "global_step": 71929, "epoch": 808, "lr": 8.380971571970053e-05}
{"train_loss": 0.17028771340847015, "global_step": 71930, "epoch": 808, "lr": 8.380928861195317e-05}
{"train_loss": 0.26609766483306885, "global_step": 71931, "epoch": 808, "lr": 8.380886149966053e-05}
{"train_loss": 0.17652197182178497, "global_step": 71932, "epoch": 808, "lr": 8.380843438282272e-05}
{"train_loss": 0.25025561451911926, "global_step": 71933, "epoch": 808, "lr": 8.380800726143976e-05}
{"train_loss": 0.2162439227104187, "global_step": 71934, "epoch": 808, "lr": 8.380758013551175e-05}
{"train_loss": 0.24722911417484283, "global_step": 71935, "epoch": 808, "lr": 8.380715300503871e-05}
{"train_loss": 0.24123211205005646, "global_step": 71936, "epoch": 808, "lr": 8.380672587002071e-05}
{"train_loss": 0.15891115367412567, "global_step": 71937, "epoch": 808, "lr": 8.380629873045781e-05}
{"train_loss": 0.2098304182291031, "global_step": 71938, "epoch": 808, "lr": 8.380587158635005e-05}
{"train_loss": 0.23441578447818756, "global_step": 71939, "epoch": 808, "lr": 8.380544443769753e-05}
{"train_loss": 0.18657296895980835, "global_step": 71940, "epoch": 808, "lr": 8.380501728450025e-05}
{"train_loss": 0.25776106119155884, "global_step": 71941, "epoch": 808, "lr": 8.380459012675832e-05}
{"train_loss": 0.2651934027671814, "global_step": 71942, "epoch": 808, "lr": 8.380416296447178e-05}
{"train_loss": 0.2930876612663269, "global_step": 71943, "epoch": 808, "lr": 8.380373579764067e-05}
{"train_loss": 0.19192862510681152, "global_step": 71944, "epoch": 808, "lr": 8.380330862626507e-05}
{"train_loss": 0.2403460144996643, "global_step": 71945, "epoch": 808, "lr": 8.3802881450345e-05}
{"train_loss": 0.2935311794281006, "global_step": 71946, "epoch": 808, "lr": 8.380245426988056e-05}
{"train_loss": 0.13245385885238647, "global_step": 71947, "epoch": 808, "lr": 8.38020270848718e-05}
{"train_loss": 0.1863025277853012, "global_step": 71948, "epoch": 808, "lr": 8.380159989531877e-05}
{"train_loss": 0.17857927083969116, "global_step": 71949, "epoch": 808, "lr": 8.380117270122152e-05}
{"train_loss": 0.1677740216255188, "global_step": 71950, "epoch": 808, "lr": 8.380074550258011e-05}
{"train_loss": 0.17309732735157013, "global_step": 71951, "epoch": 808, "lr": 8.38003182993946e-05}
{"train_loss": 0.20613400638103485, "global_step": 71952, "epoch": 808, "lr": 8.379989109166506e-05}
{"train_loss": 0.2245630919933319, "global_step": 71953, "epoch": 808, "lr": 8.379946387939155e-05}
{"train_loss": 0.14807915687561035, "global_step": 71954, "epoch": 808, "lr": 8.379903666257409e-05}
{"train_loss": 0.15482044219970703, "global_step": 71955, "epoch": 808, "lr": 8.379860944121277e-05}
{"train_loss": 0.35644853115081787, "global_step": 71956, "epoch": 808, "lr": 8.379818221530763e-05}
{"train_loss": 0.2674621343612671, "global_step": 71957, "epoch": 808, "lr": 8.379775498485874e-05}
{"train_loss": 0.23066601157188416, "global_step": 71958, "epoch": 808, "lr": 8.379732774986616e-05}
{"train_loss": 0.20175686478614807, "global_step": 71959, "epoch": 808, "lr": 8.379690051032993e-05}
{"train_loss": 0.2626090347766876, "global_step": 71960, "epoch": 808, "lr": 8.379647326625015e-05}
{"train_loss": 0.30748340487480164, "global_step": 71961, "epoch": 808, "lr": 8.379604601762681e-05}
{"train_loss": 0.17118141055107117, "global_step": 71962, "epoch": 808, "lr": 8.379561876446003e-05}
{"train_loss": 0.219431534409523, "global_step": 71963, "epoch": 808, "lr": 8.379519150674981e-05}
{"train_loss": 0.17743295431137085, "global_step": 71964, "epoch": 808, "lr": 8.379476424449626e-05}
{"train_loss": 0.255104124546051, "global_step": 71965, "epoch": 808, "lr": 8.379433697769941e-05}
{"train_loss": 0.14988143742084503, "global_step": 71966, "epoch": 808, "lr": 8.379390970635933e-05}
{"train_loss": 0.1456487476825714, "global_step": 71967, "epoch": 808, "lr": 8.379348243047606e-05}
{"train_loss": 0.15755248069763184, "global_step": 71968, "epoch": 808, "lr": 8.379305515004968e-05}
{"train_loss": 0.20799313485622406, "global_step": 71969, "epoch": 808, "lr": 8.379262786508023e-05}
{"train_loss": 0.17598414421081543, "global_step": 71970, "epoch": 808, "lr": 8.379220057556778e-05}
{"train_loss": 0.23926465213298798, "global_step": 71971, "epoch": 808, "lr": 8.379177328151237e-05}
{"train_loss": 0.2816907465457916, "global_step": 71972, "epoch": 808, "lr": 8.379134598291407e-05}
{"train_loss": 0.19673867523670197, "global_step": 71973, "epoch": 808, "lr": 8.379091867977293e-05}
{"train_loss": 0.24548861384391785, "global_step": 71974, "epoch": 808, "lr": 8.379049137208904e-05}
{"train_loss": 0.2607315182685852, "global_step": 71975, "epoch": 808, "lr": 8.379006405986239e-05}
{"train_loss": 0.20337487757205963, "global_step": 71976, "epoch": 808, "lr": 8.37896367430931e-05}
{"train_loss": 0.20531748235225677, "global_step": 71977, "epoch": 808, "lr": 8.37892094217812e-05}
{"train_loss": 0.2590779960155487, "global_step": 71978, "epoch": 808, "lr": 8.378878209592675e-05}
{"train_loss": 0.19305023550987244, "global_step": 71979, "epoch": 808, "lr": 8.378835476552982e-05}
{"train_loss": 0.1639500856399536, "global_step": 71980, "epoch": 808, "lr": 8.378792743059046e-05}
{"train_loss": 0.24389439821243286, "global_step": 71981, "epoch": 808, "lr": 8.378750009110872e-05}
{"train_loss": 0.2542964816093445, "global_step": 71982, "epoch": 808, "lr": 8.378707274708466e-05}
{"train_loss": 0.2521677613258362, "global_step": 71983, "epoch": 808, "lr": 8.378664539851835e-05}
{"train_loss": 0.24377569556236267, "global_step": 71984, "epoch": 808, "lr": 8.378621804540983e-05}
{"train_loss": 0.24388715624809265, "global_step": 71985, "epoch": 808, "lr": 8.378579068775916e-05}
{"train_loss": 0.21474707126617432, "global_step": 71986, "epoch": 808, "lr": 8.37853633255664e-05}
{"train_loss": 0.22216002643108368, "global_step": 71987, "epoch": 808, "lr": 8.37849359588316e-05}
{"train_loss": 0.26040640473365784, "global_step": 71988, "epoch": 808, "lr": 8.378450858755484e-05}
{"train_loss": 0.24029570817947388, "global_step": 71989, "epoch": 808, "lr": 8.378408121173617e-05}
{"train_loss": 0.1899522840976715, "global_step": 71990, "epoch": 808, "lr": 8.378365383137566e-05}
{"train_loss": 0.1621323674917221, "global_step": 71991, "epoch": 808, "lr": 8.378322644647332e-05}
{"train_loss": 0.17504757642745972, "global_step": 71992, "epoch": 808, "lr": 8.378279905702923e-05}
{"train_loss": 0.20153629779815674, "global_step": 71993, "epoch": 808, "lr": 8.378237166304349e-05}
{"train_loss": 0.2574639320373535, "global_step": 71994, "epoch": 808, "lr": 8.378194426451608e-05}
{"train_loss": 0.2026517242193222, "global_step": 71995, "epoch": 808, "lr": 8.378151686144713e-05}
{"train_loss": 0.21575689315795898, "global_step": 71996, "epoch": 808, "lr": 8.378108945383665e-05}
{"train_loss": 0.21791650354862213, "global_step": 71997, "epoch": 808, "lr": 8.378066204168473e-05}
{"train_loss": 0.2827225625514984, "global_step": 71998, "epoch": 808, "lr": 8.37802346249914e-05}
{"train_loss": 0.24340038001537323, "global_step": 71999, "epoch": 808, "lr": 8.377980720375673e-05}
{"train_loss": 0.21497202187441708, "global_step": 72000, "epoch": 808, "lr": 8.377937977798078e-05, "val_loss": 3.714470386505127}
{"train_loss": 0.1894683986902237, "global_step": 72001, "epoch": 809, "lr": 8.37789523476636e-05}
{"train_loss": 0.24316443502902985, "global_step": 72002, "epoch": 809, "lr": 8.377852491280526e-05}
{"train_loss": 0.14469389617443085, "global_step": 72003, "epoch": 809, "lr": 8.377809747340581e-05}
{"train_loss": 0.21186955273151398, "global_step": 72004, "epoch": 809, "lr": 8.37776700294653e-05}
{"train_loss": 0.2579843997955322, "global_step": 72005, "epoch": 809, "lr": 8.377724258098378e-05}
{"train_loss": 0.2310323417186737, "global_step": 72006, "epoch": 809, "lr": 8.377681512796134e-05}
{"train_loss": 0.16666851937770844, "global_step": 72007, "epoch": 809, "lr": 8.377638767039801e-05}
{"train_loss": 0.20176854729652405, "global_step": 72008, "epoch": 809, "lr": 8.377596020829387e-05}
{"train_loss": 0.17105546593666077, "global_step": 72009, "epoch": 809, "lr": 8.377553274164895e-05}
{"train_loss": 0.19734223186969757, "global_step": 72010, "epoch": 809, "lr": 8.377510527046333e-05}
{"train_loss": 0.16517233848571777, "global_step": 72011, "epoch": 809, "lr": 8.377467779473704e-05}
{"train_loss": 0.2295074760913849, "global_step": 72012, "epoch": 809, "lr": 8.377425031447018e-05}
{"train_loss": 0.168736070394516, "global_step": 72013, "epoch": 809, "lr": 8.377382282966278e-05}
{"train_loss": 0.1977185308933258, "global_step": 72014, "epoch": 809, "lr": 8.37733953403149e-05}
{"train_loss": 0.16544906795024872, "global_step": 72015, "epoch": 809, "lr": 8.377296784642659e-05}
{"train_loss": 0.1895279884338379, "global_step": 72016, "epoch": 809, "lr": 8.377254034799791e-05}
{"train_loss": 0.3098340630531311, "global_step": 72017, "epoch": 809, "lr": 8.377211284502893e-05}
{"train_loss": 0.1737336665391922, "global_step": 72018, "epoch": 809, "lr": 8.377168533751971e-05}
{"train_loss": 0.14433324337005615, "global_step": 72019, "epoch": 809, "lr": 8.37712578254703e-05}
{"train_loss": 0.22122804820537567, "global_step": 72020, "epoch": 809, "lr": 8.377083030888075e-05}
{"train_loss": 0.19751529395580292, "global_step": 72021, "epoch": 809, "lr": 8.377040278775113e-05}
{"train_loss": 0.16312164068222046, "global_step": 72022, "epoch": 809, "lr": 8.376997526208148e-05}
{"train_loss": 0.27243804931640625, "global_step": 72023, "epoch": 809, "lr": 8.376954773187187e-05}
{"train_loss": 0.19526605308055878, "global_step": 72024, "epoch": 809, "lr": 8.376912019712235e-05}
{"train_loss": 0.27278265357017517, "global_step": 72025, "epoch": 809, "lr": 8.3768692657833e-05}
{"train_loss": 0.22580014169216156, "global_step": 72026, "epoch": 809, "lr": 8.376826511400386e-05}
{"train_loss": 0.16765698790550232, "global_step": 72027, "epoch": 809, "lr": 8.376783756563497e-05}
{"train_loss": 0.18190094828605652, "global_step": 72028, "epoch": 809, "lr": 8.376741001272642e-05}
{"train_loss": 0.16028472781181335, "global_step": 72029, "epoch": 809, "lr": 8.376698245527825e-05}
{"train_loss": 0.2474641650915146, "global_step": 72030, "epoch": 809, "lr": 8.376655489329052e-05}
{"train_loss": 0.22864854335784912, "global_step": 72031, "epoch": 809, "lr": 8.376612732676329e-05}
{"train_loss": 0.22285082936286926, "global_step": 72032, "epoch": 809, "lr": 8.376569975569662e-05}
{"train_loss": 0.1777721643447876, "global_step": 72033, "epoch": 809, "lr": 8.376527218009055e-05}
{"train_loss": 0.22473210096359253, "global_step": 72034, "epoch": 809, "lr": 8.376484459994516e-05}
{"train_loss": 0.1575912982225418, "global_step": 72035, "epoch": 809, "lr": 8.37644170152605e-05}
{"train_loss": 0.23040258884429932, "global_step": 72036, "epoch": 809, "lr": 8.376398942603662e-05}
{"train_loss": 0.1991574466228485, "global_step": 72037, "epoch": 809, "lr": 8.376356183227358e-05}
{"train_loss": 0.18742500245571136, "global_step": 72038, "epoch": 809, "lr": 8.376313423397145e-05}
{"train_loss": 0.2166733741760254, "global_step": 72039, "epoch": 809, "lr": 8.376270663113028e-05}
{"train_loss": 0.25412601232528687, "global_step": 72040, "epoch": 809, "lr": 8.376227902375011e-05}
{"train_loss": 0.1441466510295868, "global_step": 72041, "epoch": 809, "lr": 8.376185141183104e-05}
{"train_loss": 0.1918773502111435, "global_step": 72042, "epoch": 809, "lr": 8.376142379537308e-05}
{"train_loss": 0.18737785518169403, "global_step": 72043, "epoch": 809, "lr": 8.37609961743763e-05}
{"train_loss": 0.21999096870422363, "global_step": 72044, "epoch": 809, "lr": 8.376056854884078e-05}
{"train_loss": 0.18755756318569183, "global_step": 72045, "epoch": 809, "lr": 8.376014091876656e-05}
{"train_loss": 0.22274649143218994, "global_step": 72046, "epoch": 809, "lr": 8.37597132841537e-05}
{"train_loss": 0.23832988739013672, "global_step": 72047, "epoch": 809, "lr": 8.375928564500226e-05}
{"train_loss": 0.18847978115081787, "global_step": 72048, "epoch": 809, "lr": 8.375885800131228e-05}
{"train_loss": 0.19742892682552338, "global_step": 72049, "epoch": 809, "lr": 8.375843035308386e-05}
{"train_loss": 0.22650755941867828, "global_step": 72050, "epoch": 809, "lr": 8.375800270031701e-05}
{"train_loss": 0.15302816033363342, "global_step": 72051, "epoch": 809, "lr": 8.375757504301183e-05}
{"train_loss": 0.19601917266845703, "global_step": 72052, "epoch": 809, "lr": 8.375714738116834e-05}
{"train_loss": 0.1887604147195816, "global_step": 72053, "epoch": 809, "lr": 8.37567197147866e-05}
{"train_loss": 0.17543278634548187, "global_step": 72054, "epoch": 809, "lr": 8.37562920438667e-05}
{"train_loss": 0.23076047003269196, "global_step": 72055, "epoch": 809, "lr": 8.375586436840868e-05}
{"train_loss": 0.18948806822299957, "global_step": 72056, "epoch": 809, "lr": 8.375543668841259e-05}
{"train_loss": 0.1863182932138443, "global_step": 72057, "epoch": 809, "lr": 8.37550090038785e-05}
{"train_loss": 0.17827792465686798, "global_step": 72058, "epoch": 809, "lr": 8.375458131480644e-05}
{"train_loss": 0.18478521704673767, "global_step": 72059, "epoch": 809, "lr": 8.375415362119649e-05}
{"train_loss": 0.16808342933654785, "global_step": 72060, "epoch": 809, "lr": 8.375372592304873e-05}
{"train_loss": 0.1900641918182373, "global_step": 72061, "epoch": 809, "lr": 8.375329822036317e-05}
{"train_loss": 0.13405068218708038, "global_step": 72062, "epoch": 809, "lr": 8.37528705131399e-05}
{"train_loss": 0.13699816167354584, "global_step": 72063, "epoch": 809, "lr": 8.375244280137897e-05}
{"train_loss": 0.20379619300365448, "global_step": 72064, "epoch": 809, "lr": 8.375201508508044e-05}
{"train_loss": 0.17750941216945648, "global_step": 72065, "epoch": 809, "lr": 8.375158736424434e-05}
{"train_loss": 0.22645969688892365, "global_step": 72066, "epoch": 809, "lr": 8.375115963887077e-05}
{"train_loss": 0.24665114283561707, "global_step": 72067, "epoch": 809, "lr": 8.375073190895977e-05}
{"train_loss": 0.16727687418460846, "global_step": 72068, "epoch": 809, "lr": 8.375030417451137e-05}
{"train_loss": 0.25911030173301697, "global_step": 72069, "epoch": 809, "lr": 8.374987643552568e-05}
{"train_loss": 0.16186752915382385, "global_step": 72070, "epoch": 809, "lr": 8.374944869200272e-05}
{"train_loss": 0.15376397967338562, "global_step": 72071, "epoch": 809, "lr": 8.374902094394257e-05}
{"train_loss": 0.25318726897239685, "global_step": 72072, "epoch": 809, "lr": 8.374859319134526e-05}
{"train_loss": 0.2517841160297394, "global_step": 72073, "epoch": 809, "lr": 8.374816543421087e-05}
{"train_loss": 0.1645650565624237, "global_step": 72074, "epoch": 809, "lr": 8.374773767253945e-05}
{"train_loss": 0.14585119485855103, "global_step": 72075, "epoch": 809, "lr": 8.374730990633106e-05}
{"train_loss": 0.1924467533826828, "global_step": 72076, "epoch": 809, "lr": 8.374688213558575e-05}
{"train_loss": 0.34232407808303833, "global_step": 72077, "epoch": 809, "lr": 8.37464543603036e-05}
{"train_loss": 0.15312761068344116, "global_step": 72078, "epoch": 809, "lr": 8.374602658048462e-05}
{"train_loss": 0.2960761785507202, "global_step": 72079, "epoch": 809, "lr": 8.374559879612891e-05}
{"train_loss": 0.27720391750335693, "global_step": 72080, "epoch": 809, "lr": 8.374517100723653e-05}
{"train_loss": 0.2345517873764038, "global_step": 72081, "epoch": 809, "lr": 8.37447432138075e-05}
{"train_loss": 0.16826851665973663, "global_step": 72082, "epoch": 809, "lr": 8.374431541584193e-05}
{"train_loss": 0.14590921998023987, "global_step": 72083, "epoch": 809, "lr": 8.374388761333982e-05}
{"train_loss": 0.23141735792160034, "global_step": 72084, "epoch": 809, "lr": 8.374345980630127e-05}
{"train_loss": 0.22565637528896332, "global_step": 72085, "epoch": 809, "lr": 8.374303199472631e-05}
{"train_loss": 0.24011659622192383, "global_step": 72086, "epoch": 809, "lr": 8.374260417861503e-05}
{"train_loss": 0.17767219245433807, "global_step": 72087, "epoch": 809, "lr": 8.374217635796747e-05}
{"train_loss": 0.22641271352767944, "global_step": 72088, "epoch": 809, "lr": 8.374174853278366e-05}
{"train_loss": 0.20334608032462304, "global_step": 72089, "epoch": 809, "lr": 8.37413207030637e-05, "val_loss": 3.824307918548584}
{"train_loss": 0.3302263021469116, "global_step": 72090, "epoch": 810, "lr": 8.374089286880762e-05}
{"train_loss": 0.2590618431568146, "global_step": 72091, "epoch": 810, "lr": 8.374046503001551e-05}
{"train_loss": 0.2535196840763092, "global_step": 72092, "epoch": 810, "lr": 8.37400371866874e-05}
{"train_loss": 0.21431821584701538, "global_step": 72093, "epoch": 810, "lr": 8.373960933882333e-05}
{"train_loss": 0.28883224725723267, "global_step": 72094, "epoch": 810, "lr": 8.373918148642339e-05}
{"train_loss": 0.3263342082500458, "global_step": 72095, "epoch": 810, "lr": 8.373875362948765e-05}
{"train_loss": 0.2391413152217865, "global_step": 72096, "epoch": 810, "lr": 8.373832576801612e-05}
{"train_loss": 0.2693171203136444, "global_step": 72097, "epoch": 810, "lr": 8.37378979020089e-05}
{"train_loss": 0.20744389295578003, "global_step": 72098, "epoch": 810, "lr": 8.373747003146602e-05}
{"train_loss": 0.2389340102672577, "global_step": 72099, "epoch": 810, "lr": 8.373704215638755e-05}
{"train_loss": 0.2942737936973572, "global_step": 72100, "epoch": 810, "lr": 8.373661427677356e-05}
{"train_loss": 0.2569482922554016, "global_step": 72101, "epoch": 810, "lr": 8.373618639262406e-05}
{"train_loss": 0.23264428973197937, "global_step": 72102, "epoch": 810, "lr": 8.373575850393918e-05}
{"train_loss": 0.15677505731582642, "global_step": 72103, "epoch": 810, "lr": 8.373533061071892e-05}
{"train_loss": 0.18574953079223633, "global_step": 72104, "epoch": 810, "lr": 8.373490271296336e-05}
{"train_loss": 0.24209311604499817, "global_step": 72105, "epoch": 810, "lr": 8.373447481067255e-05}
{"train_loss": 0.19332651793956757, "global_step": 72106, "epoch": 810, "lr": 8.373404690384654e-05}
{"train_loss": 0.1901174634695053, "global_step": 72107, "epoch": 810, "lr": 8.373361899248542e-05}
{"train_loss": 0.1708788275718689, "global_step": 72108, "epoch": 810, "lr": 8.373319107658921e-05}
{"train_loss": 0.18410241603851318, "global_step": 72109, "epoch": 810, "lr": 8.3732763156158e-05}
{"train_loss": 0.23053424060344696, "global_step": 72110, "epoch": 810, "lr": 8.373233523119182e-05}
{"train_loss": 0.26986512541770935, "global_step": 72111, "epoch": 810, "lr": 8.373190730169074e-05}
{"train_loss": 0.3082520067691803, "global_step": 72112, "epoch": 810, "lr": 8.373147936765483e-05}
{"train_loss": 0.19694633781909943, "global_step": 72113, "epoch": 810, "lr": 8.373105142908411e-05}
{"train_loss": 0.24214573204517365, "global_step": 72114, "epoch": 810, "lr": 8.373062348597867e-05}
{"train_loss": 0.1906130015850067, "global_step": 72115, "epoch": 810, "lr": 8.373019553833856e-05}
{"train_loss": 0.19556941092014313, "global_step": 72116, "epoch": 810, "lr": 8.372976758616383e-05}
{"train_loss": 0.213155597448349, "global_step": 72117, "epoch": 810, "lr": 8.372933962945455e-05}
{"train_loss": 0.23899127542972565, "global_step": 72118, "epoch": 810, "lr": 8.372891166821078e-05}
{"train_loss": 0.24975420534610748, "global_step": 72119, "epoch": 810, "lr": 8.372848370243255e-05}
{"train_loss": 0.15567654371261597, "global_step": 72120, "epoch": 810, "lr": 8.372805573211995e-05}
{"train_loss": 0.21773424744606018, "global_step": 72121, "epoch": 810, "lr": 8.372762775727301e-05}
{"train_loss": 0.2667383849620819, "global_step": 72122, "epoch": 810, "lr": 8.372719977789181e-05}
{"train_loss": 0.22587081789970398, "global_step": 72123, "epoch": 810, "lr": 8.372677179397642e-05}
{"train_loss": 0.2240278571844101, "global_step": 72124, "epoch": 810, "lr": 8.372634380552686e-05}
{"train_loss": 0.22253841161727905, "global_step": 72125, "epoch": 810, "lr": 8.37259158125432e-05}
{"train_loss": 0.22303470969200134, "global_step": 72126, "epoch": 810, "lr": 8.372548781502549e-05}
{"train_loss": 0.1628970503807068, "global_step": 72127, "epoch": 810, "lr": 8.372505981297383e-05}
{"train_loss": 0.18858399987220764, "global_step": 72128, "epoch": 810, "lr": 8.372463180638823e-05}
{"train_loss": 0.3261692225933075, "global_step": 72129, "epoch": 810, "lr": 8.372420379526875e-05}
{"train_loss": 0.28830602765083313, "global_step": 72130, "epoch": 810, "lr": 8.372377577961547e-05}
{"train_loss": 0.2989729940891266, "global_step": 72131, "epoch": 810, "lr": 8.372334775942844e-05}
{"train_loss": 0.3388669490814209, "global_step": 72132, "epoch": 810, "lr": 8.372291973470774e-05}
{"train_loss": 0.26594269275665283, "global_step": 72133, "epoch": 810, "lr": 8.372249170545338e-05}
{"train_loss": 0.23529332876205444, "global_step": 72134, "epoch": 810, "lr": 8.372206367166546e-05}
{"train_loss": 0.196486696600914, "global_step": 72135, "epoch": 810, "lr": 8.372163563334401e-05}
{"train_loss": 0.24059870839118958, "global_step": 72136, "epoch": 810, "lr": 8.372120759048909e-05}
{"train_loss": 0.17689016461372375, "global_step": 72137, "epoch": 810, "lr": 8.372077954310077e-05}
{"train_loss": 0.2224460393190384, "global_step": 72138, "epoch": 810, "lr": 8.372035149117909e-05}
{"train_loss": 0.186701700091362, "global_step": 72139, "epoch": 810, "lr": 8.371992343472413e-05}
{"train_loss": 0.22829020023345947, "global_step": 72140, "epoch": 810, "lr": 8.371949537373594e-05}
{"train_loss": 0.29605138301849365, "global_step": 72141, "epoch": 810, "lr": 8.371906730821459e-05}
{"train_loss": 0.1250278353691101, "global_step": 72142, "epoch": 810, "lr": 8.37186392381601e-05}
{"train_loss": 0.19455502927303314, "global_step": 72143, "epoch": 810, "lr": 8.371821116357256e-05}
{"train_loss": 0.22001731395721436, "global_step": 72144, "epoch": 810, "lr": 8.371778308445202e-05}
{"train_loss": 0.22315150499343872, "global_step": 72145, "epoch": 810, "lr": 8.371735500079852e-05}
{"train_loss": 0.22510799765586853, "global_step": 72146, "epoch": 810, "lr": 8.371692691261214e-05}
{"train_loss": 0.18962602317333221, "global_step": 72147, "epoch": 810, "lr": 8.371649881989294e-05}
{"train_loss": 0.24482697248458862, "global_step": 72148, "epoch": 810, "lr": 8.371607072264096e-05}
{"train_loss": 0.24368149042129517, "global_step": 72149, "epoch": 810, "lr": 8.371564262085625e-05}
{"train_loss": 0.20104853808879852, "global_step": 72150, "epoch": 810, "lr": 8.37152145145389e-05}
{"train_loss": 0.2546945810317993, "global_step": 72151, "epoch": 810, "lr": 8.371478640368895e-05}
{"train_loss": 0.16949276626110077, "global_step": 72152, "epoch": 810, "lr": 8.371435828830646e-05}
{"train_loss": 0.22357599437236786, "global_step": 72153, "epoch": 810, "lr": 8.371393016839149e-05}
{"train_loss": 0.15343593060970306, "global_step": 72154, "epoch": 810, "lr": 8.371350204394409e-05}
{"train_loss": 0.2133244276046753, "global_step": 72155, "epoch": 810, "lr": 8.371307391496431e-05}
{"train_loss": 0.2770604193210602, "global_step": 72156, "epoch": 810, "lr": 8.371264578145221e-05}
{"train_loss": 0.2590898275375366, "global_step": 72157, "epoch": 810, "lr": 8.371221764340787e-05}
{"train_loss": 0.18529567122459412, "global_step": 72158, "epoch": 810, "lr": 8.371178950083133e-05}
{"train_loss": 0.1517374962568283, "global_step": 72159, "epoch": 810, "lr": 8.371136135372265e-05}
{"train_loss": 0.19350366294384003, "global_step": 72160, "epoch": 810, "lr": 8.371093320208189e-05}
{"train_loss": 0.2555510103702545, "global_step": 72161, "epoch": 810, "lr": 8.371050504590912e-05}
{"train_loss": 0.3999929428100586, "global_step": 72162, "epoch": 810, "lr": 8.371007688520436e-05}
{"train_loss": 0.2107100635766983, "global_step": 72163, "epoch": 810, "lr": 8.37096487199677e-05}
{"train_loss": 0.26011916995048523, "global_step": 72164, "epoch": 810, "lr": 8.370922055019919e-05}
{"train_loss": 0.20822009444236755, "global_step": 72165, "epoch": 810, "lr": 8.370879237589888e-05}
{"train_loss": 0.232464000582695, "global_step": 72166, "epoch": 810, "lr": 8.370836419706685e-05}
{"train_loss": 0.23015855252742767, "global_step": 72167, "epoch": 810, "lr": 8.370793601370313e-05}
{"train_loss": 0.21566542983055115, "global_step": 72168, "epoch": 810, "lr": 8.370750782580779e-05}
{"train_loss": 0.18541130423545837, "global_step": 72169, "epoch": 810, "lr": 8.37070796333809e-05}
{"train_loss": 0.2938939332962036, "global_step": 72170, "epoch": 810, "lr": 8.370665143642247e-05}
{"train_loss": 0.1966204196214676, "global_step": 72171, "epoch": 810, "lr": 8.370622323493263e-05}
{"train_loss": 0.12941494584083557, "global_step": 72172, "epoch": 810, "lr": 8.370579502891136e-05}
{"train_loss": 0.24353435635566711, "global_step": 72173, "epoch": 810, "lr": 8.370536681835878e-05}
{"train_loss": 0.14717631042003632, "global_step": 72174, "epoch": 810, "lr": 8.370493860327493e-05}
{"train_loss": 0.2676606774330139, "global_step": 72175, "epoch": 810, "lr": 8.370451038365985e-05}
{"train_loss": 0.22875268757343292, "global_step": 72176, "epoch": 810, "lr": 8.37040821595136e-05}
{"train_loss": 0.28989216685295105, "global_step": 72177, "epoch": 810, "lr": 8.370365393083627e-05}
{"train_loss": 0.22885593068733645, "global_step": 72178, "epoch": 810, "lr": 8.370322569762789e-05, "val_loss": 3.7542855739593506, "train_action_mse_error": 26.681608200073242}
{"train_loss": 0.2757198214530945, "global_step": 72179, "epoch": 811, "lr": 8.370279745988852e-05}
{"train_loss": 0.18209071457386017, "global_step": 72180, "epoch": 811, "lr": 8.370236921761821e-05}
{"train_loss": 0.20433424413204193, "global_step": 72181, "epoch": 811, "lr": 8.370194097081703e-05}
{"train_loss": 0.1660960614681244, "global_step": 72182, "epoch": 811, "lr": 8.370151271948504e-05}
{"train_loss": 0.28961363434791565, "global_step": 72183, "epoch": 811, "lr": 8.37010844636223e-05}
{"train_loss": 0.20249032974243164, "global_step": 72184, "epoch": 811, "lr": 8.370065620322884e-05}
{"train_loss": 0.43530768156051636, "global_step": 72185, "epoch": 811, "lr": 8.370022793830475e-05}
{"train_loss": 0.161522775888443, "global_step": 72186, "epoch": 811, "lr": 8.369979966885007e-05}
{"train_loss": 0.23615500330924988, "global_step": 72187, "epoch": 811, "lr": 8.369937139486487e-05}
{"train_loss": 0.19010816514492035, "global_step": 72188, "epoch": 811, "lr": 8.369894311634919e-05}
{"train_loss": 0.25329500436782837, "global_step": 72189, "epoch": 811, "lr": 8.369851483330312e-05}
{"train_loss": 0.23855336010456085, "global_step": 72190, "epoch": 811, "lr": 8.369808654572666e-05}
{"train_loss": 0.2300565242767334, "global_step": 72191, "epoch": 811, "lr": 8.369765825361993e-05}
{"train_loss": 0.15839990973472595, "global_step": 72192, "epoch": 811, "lr": 8.369722995698295e-05}
{"train_loss": 0.22637812793254852, "global_step": 72193, "epoch": 811, "lr": 8.369680165581579e-05}
{"train_loss": 0.24943841993808746, "global_step": 72194, "epoch": 811, "lr": 8.36963733501185e-05}
{"train_loss": 0.24129754304885864, "global_step": 72195, "epoch": 811, "lr": 8.369594503989115e-05}
{"train_loss": 0.23860867321491241, "global_step": 72196, "epoch": 811, "lr": 8.369551672513381e-05}
{"train_loss": 0.2474001944065094, "global_step": 72197, "epoch": 811, "lr": 8.369508840584649e-05}
{"train_loss": 0.27339789271354675, "global_step": 72198, "epoch": 811, "lr": 8.369466008202928e-05}
{"train_loss": 0.17941856384277344, "global_step": 72199, "epoch": 811, "lr": 8.369423175368225e-05}
{"train_loss": 0.26898160576820374, "global_step": 72200, "epoch": 811, "lr": 8.369380342080543e-05}
{"train_loss": 0.22241902351379395, "global_step": 72201, "epoch": 811, "lr": 8.369337508339888e-05}
{"train_loss": 0.21376065909862518, "global_step": 72202, "epoch": 811, "lr": 8.369294674146269e-05}
{"train_loss": 0.19333583116531372, "global_step": 72203, "epoch": 811, "lr": 8.369251839499688e-05}
{"train_loss": 0.22518455982208252, "global_step": 72204, "epoch": 811, "lr": 8.369209004400151e-05}
{"train_loss": 0.16197100281715393, "global_step": 72205, "epoch": 811, "lr": 8.369166168847666e-05}
{"train_loss": 0.23587000370025635, "global_step": 72206, "epoch": 811, "lr": 8.369123332842237e-05}
{"train_loss": 0.20465527474880219, "global_step": 72207, "epoch": 811, "lr": 8.369080496383872e-05}
{"train_loss": 0.25956636667251587, "global_step": 72208, "epoch": 811, "lr": 8.369037659472573e-05}
{"train_loss": 0.30480101704597473, "global_step": 72209, "epoch": 811, "lr": 8.36899482210835e-05}
{"train_loss": 0.18380391597747803, "global_step": 72210, "epoch": 811, "lr": 8.368951984291206e-05}
{"train_loss": 0.13197708129882812, "global_step": 72211, "epoch": 811, "lr": 8.368909146021147e-05}
{"train_loss": 0.24496662616729736, "global_step": 72212, "epoch": 811, "lr": 8.36886630729818e-05}
{"train_loss": 0.2082245796918869, "global_step": 72213, "epoch": 811, "lr": 8.368823468122309e-05}
{"train_loss": 0.23505839705467224, "global_step": 72214, "epoch": 811, "lr": 8.368780628493542e-05}
{"train_loss": 0.2480715811252594, "global_step": 72215, "epoch": 811, "lr": 8.368737788411883e-05}
{"train_loss": 0.26214686036109924, "global_step": 72216, "epoch": 811, "lr": 8.368694947877338e-05}
{"train_loss": 0.14275115728378296, "global_step": 72217, "epoch": 811, "lr": 8.368652106889914e-05}
{"train_loss": 0.2381761372089386, "global_step": 72218, "epoch": 811, "lr": 8.368609265449613e-05}
{"train_loss": 0.1353222131729126, "global_step": 72219, "epoch": 811, "lr": 8.368566423556447e-05}
{"train_loss": 0.13231953978538513, "global_step": 72220, "epoch": 811, "lr": 8.368523581210417e-05}
{"train_loss": 0.2154419869184494, "global_step": 72221, "epoch": 811, "lr": 8.36848073841153e-05}
{"train_loss": 0.1862843632698059, "global_step": 72222, "epoch": 811, "lr": 8.36843789515979e-05}
{"train_loss": 0.2945476770401001, "global_step": 72223, "epoch": 811, "lr": 8.368395051455208e-05}
{"train_loss": 0.17973609268665314, "global_step": 72224, "epoch": 811, "lr": 8.368352207297784e-05}
{"train_loss": 0.1738085299730301, "global_step": 72225, "epoch": 811, "lr": 8.368309362687527e-05}
{"train_loss": 0.2311556041240692, "global_step": 72226, "epoch": 811, "lr": 8.368266517624441e-05}
{"train_loss": 0.1502085030078888, "global_step": 72227, "epoch": 811, "lr": 8.368223672108534e-05}
{"train_loss": 0.3131789565086365, "global_step": 72228, "epoch": 811, "lr": 8.368180826139809e-05}
{"train_loss": 0.2190060168504715, "global_step": 72229, "epoch": 811, "lr": 8.368137979718273e-05}
{"train_loss": 0.21937188506126404, "global_step": 72230, "epoch": 811, "lr": 8.368095132843934e-05}
{"train_loss": 0.19543911516666412, "global_step": 72231, "epoch": 811, "lr": 8.368052285516795e-05}
{"train_loss": 0.18879275023937225, "global_step": 72232, "epoch": 811, "lr": 8.36800943773686e-05}
{"train_loss": 0.17299313843250275, "global_step": 72233, "epoch": 811, "lr": 8.367966589504139e-05}
{"train_loss": 0.2214449942111969, "global_step": 72234, "epoch": 811, "lr": 8.367923740818637e-05}
{"train_loss": 0.19127850234508514, "global_step": 72235, "epoch": 811, "lr": 8.367880891680357e-05}
{"train_loss": 0.180962473154068, "global_step": 72236, "epoch": 811, "lr": 8.367838042089306e-05}
{"train_loss": 0.18719910085201263, "global_step": 72237, "epoch": 811, "lr": 8.367795192045492e-05}
{"train_loss": 0.21388840675354004, "global_step": 72238, "epoch": 811, "lr": 8.367752341548918e-05}
{"train_loss": 0.19803200662136078, "global_step": 72239, "epoch": 811, "lr": 8.367709490599591e-05}
{"train_loss": 0.23210996389389038, "global_step": 72240, "epoch": 811, "lr": 8.367666639197516e-05}
{"train_loss": 0.22316139936447144, "global_step": 72241, "epoch": 811, "lr": 8.3676237873427e-05}
{"train_loss": 0.2957897186279297, "global_step": 72242, "epoch": 811, "lr": 8.367580935035148e-05}
{"train_loss": 0.17615103721618652, "global_step": 72243, "epoch": 811, "lr": 8.367538082274866e-05}
{"train_loss": 0.14309175312519073, "global_step": 72244, "epoch": 811, "lr": 8.367495229061859e-05}
{"train_loss": 0.28678423166275024, "global_step": 72245, "epoch": 811, "lr": 8.367452375396133e-05}
{"train_loss": 0.26567766070365906, "global_step": 72246, "epoch": 811, "lr": 8.367409521277695e-05}
{"train_loss": 0.3321726620197296, "global_step": 72247, "epoch": 811, "lr": 8.367366666706549e-05}
{"train_loss": 0.2884277403354645, "global_step": 72248, "epoch": 811, "lr": 8.3673238116827e-05}
{"train_loss": 0.1540939211845398, "global_step": 72249, "epoch": 811, "lr": 8.367280956206159e-05}
{"train_loss": 0.14571644365787506, "global_step": 72250, "epoch": 811, "lr": 8.367238100276926e-05}
{"train_loss": 0.22737987339496613, "global_step": 72251, "epoch": 811, "lr": 8.36719524389501e-05}
{"train_loss": 0.1840091496706009, "global_step": 72252, "epoch": 811, "lr": 8.367152387060414e-05}
{"train_loss": 0.2983778715133667, "global_step": 72253, "epoch": 811, "lr": 8.367109529773146e-05}
{"train_loss": 0.2814924418926239, "global_step": 72254, "epoch": 811, "lr": 8.367066672033213e-05}
{"train_loss": 0.2650473713874817, "global_step": 72255, "epoch": 811, "lr": 8.367023813840616e-05}
{"train_loss": 0.36029139161109924, "global_step": 72256, "epoch": 811, "lr": 8.366980955195365e-05}
{"train_loss": 0.25477421283721924, "global_step": 72257, "epoch": 811, "lr": 8.366938096097465e-05}
{"train_loss": 0.17111192643642426, "global_step": 72258, "epoch": 811, "lr": 8.366895236546921e-05}
{"train_loss": 0.19860102236270905, "global_step": 72259, "epoch": 811, "lr": 8.366852376543739e-05}
{"train_loss": 0.23135462403297424, "global_step": 72260, "epoch": 811, "lr": 8.366809516087923e-05}
{"train_loss": 0.2813979685306549, "global_step": 72261, "epoch": 811, "lr": 8.366766655179484e-05}
{"train_loss": 0.3285200893878937, "global_step": 72262, "epoch": 811, "lr": 8.366723793818421e-05}
{"train_loss": 0.24534651637077332, "global_step": 72263, "epoch": 811, "lr": 8.366680932004744e-05}
{"train_loss": 0.28995227813720703, "global_step": 72264, "epoch": 811, "lr": 8.366638069738457e-05}
{"train_loss": 0.1998334378004074, "global_step": 72265, "epoch": 811, "lr": 8.366595207019568e-05}
{"train_loss": 0.20338362455368042, "global_step": 72266, "epoch": 811, "lr": 8.366552343848081e-05}
{"train_loss": 0.22443399184875273, "global_step": 72267, "epoch": 811, "lr": 8.366509480224001e-05, "val_loss": 3.7909111976623535}
{"train_loss": 0.2999202311038971, "global_step": 72268, "epoch": 812, "lr": 8.366466616147337e-05}
{"train_loss": 0.2896592319011688, "global_step": 72269, "epoch": 812, "lr": 8.366423751618091e-05}
{"train_loss": 0.23137697577476501, "global_step": 72270, "epoch": 812, "lr": 8.36638088663627e-05}
{"train_loss": 0.34946513175964355, "global_step": 72271, "epoch": 812, "lr": 8.366338021201881e-05}
{"train_loss": 0.2544578015804291, "global_step": 72272, "epoch": 812, "lr": 8.366295155314929e-05}
{"train_loss": 0.3238341510295868, "global_step": 72273, "epoch": 812, "lr": 8.366252288975419e-05}
{"train_loss": 0.19145777821540833, "global_step": 72274, "epoch": 812, "lr": 8.366209422183358e-05}
{"train_loss": 0.11755582690238953, "global_step": 72275, "epoch": 812, "lr": 8.36616655493875e-05}
{"train_loss": 0.2660253047943115, "global_step": 72276, "epoch": 812, "lr": 8.366123687241603e-05}
{"train_loss": 0.30333393812179565, "global_step": 72277, "epoch": 812, "lr": 8.366080819091923e-05}
{"train_loss": 0.23765027523040771, "global_step": 72278, "epoch": 812, "lr": 8.366037950489713e-05}
{"train_loss": 0.2522275745868683, "global_step": 72279, "epoch": 812, "lr": 8.365995081434981e-05}
{"train_loss": 0.42315346002578735, "global_step": 72280, "epoch": 812, "lr": 8.36595221192773e-05}
{"train_loss": 0.17502868175506592, "global_step": 72281, "epoch": 812, "lr": 8.365909341967969e-05}
{"train_loss": 0.19120800495147705, "global_step": 72282, "epoch": 812, "lr": 8.365866471555703e-05}
{"train_loss": 0.15518708527088165, "global_step": 72283, "epoch": 812, "lr": 8.365823600690936e-05}
{"train_loss": 0.2217235267162323, "global_step": 72284, "epoch": 812, "lr": 8.365780729373676e-05}
{"train_loss": 0.23741085827350616, "global_step": 72285, "epoch": 812, "lr": 8.365737857603928e-05}
{"train_loss": 0.27802950143814087, "global_step": 72286, "epoch": 812, "lr": 8.365694985381695e-05}
{"train_loss": 0.25745058059692383, "global_step": 72287, "epoch": 812, "lr": 8.36565211270699e-05}
{"train_loss": 0.21574313938617706, "global_step": 72288, "epoch": 812, "lr": 8.36560923957981e-05}
{"train_loss": 0.23500265181064606, "global_step": 72289, "epoch": 812, "lr": 8.365566366000167e-05}
{"train_loss": 0.15600043535232544, "global_step": 72290, "epoch": 812, "lr": 8.365523491968063e-05}
{"train_loss": 0.19193750619888306, "global_step": 72291, "epoch": 812, "lr": 8.365480617483508e-05}
{"train_loss": 0.19494101405143738, "global_step": 72292, "epoch": 812, "lr": 8.365437742546502e-05}
{"train_loss": 0.17287856340408325, "global_step": 72293, "epoch": 812, "lr": 8.365394867157057e-05}
{"train_loss": 0.2782355546951294, "global_step": 72294, "epoch": 812, "lr": 8.365351991315174e-05}
{"train_loss": 0.1684320569038391, "global_step": 72295, "epoch": 812, "lr": 8.36530911502086e-05}
{"train_loss": 0.14871278405189514, "global_step": 72296, "epoch": 812, "lr": 8.365266238274123e-05}
{"train_loss": 0.25474604964256287, "global_step": 72297, "epoch": 812, "lr": 8.365223361074966e-05}
{"train_loss": 0.2998654842376709, "global_step": 72298, "epoch": 812, "lr": 8.365180483423394e-05}
{"train_loss": 0.2746939957141876, "global_step": 72299, "epoch": 812, "lr": 8.365137605319418e-05}
{"train_loss": 0.19260922074317932, "global_step": 72300, "epoch": 812, "lr": 8.365094726763037e-05}
{"train_loss": 0.26365455985069275, "global_step": 72301, "epoch": 812, "lr": 8.365051847754261e-05}
{"train_loss": 0.21136152744293213, "global_step": 72302, "epoch": 812, "lr": 8.365008968293095e-05}
{"train_loss": 0.17803385853767395, "global_step": 72303, "epoch": 812, "lr": 8.364966088379545e-05}
{"train_loss": 0.1903405636548996, "global_step": 72304, "epoch": 812, "lr": 8.364923208013617e-05}
{"train_loss": 0.25121626257896423, "global_step": 72305, "epoch": 812, "lr": 8.364880327195314e-05}
{"train_loss": 0.1914457082748413, "global_step": 72306, "epoch": 812, "lr": 8.364837445924644e-05}
{"train_loss": 0.19817695021629333, "global_step": 72307, "epoch": 812, "lr": 8.364794564201614e-05}
{"train_loss": 0.2488306164741516, "global_step": 72308, "epoch": 812, "lr": 8.364751682026228e-05}
{"train_loss": 0.23843643069267273, "global_step": 72309, "epoch": 812, "lr": 8.364708799398494e-05}
{"train_loss": 0.2060144692659378, "global_step": 72310, "epoch": 812, "lr": 8.364665916318412e-05}
{"train_loss": 0.1312018185853958, "global_step": 72311, "epoch": 812, "lr": 8.364623032785994e-05}
{"train_loss": 0.27945220470428467, "global_step": 72312, "epoch": 812, "lr": 8.364580148801245e-05}
{"train_loss": 0.17110668122768402, "global_step": 72313, "epoch": 812, "lr": 8.364537264364167e-05}
{"train_loss": 0.2202320098876953, "global_step": 72314, "epoch": 812, "lr": 8.364494379474768e-05}
{"train_loss": 0.19076736271381378, "global_step": 72315, "epoch": 812, "lr": 8.364451494133054e-05}
{"train_loss": 0.22338387370109558, "global_step": 72316, "epoch": 812, "lr": 8.364408608339031e-05}
{"train_loss": 0.22197450697422028, "global_step": 72317, "epoch": 812, "lr": 8.364365722092704e-05}
{"train_loss": 0.1827882081270218, "global_step": 72318, "epoch": 812, "lr": 8.364322835394079e-05}
{"train_loss": 0.22201991081237793, "global_step": 72319, "epoch": 812, "lr": 8.36427994824316e-05}
{"train_loss": 0.2517780065536499, "global_step": 72320, "epoch": 812, "lr": 8.364237060639957e-05}
{"train_loss": 0.25833648443222046, "global_step": 72321, "epoch": 812, "lr": 8.364194172584472e-05}
{"train_loss": 0.16593033075332642, "global_step": 72322, "epoch": 812, "lr": 8.364151284076713e-05}
{"train_loss": 0.21152296662330627, "global_step": 72323, "epoch": 812, "lr": 8.364108395116684e-05}
{"train_loss": 0.16872210800647736, "global_step": 72324, "epoch": 812, "lr": 8.364065505704393e-05}
{"train_loss": 0.21199965476989746, "global_step": 72325, "epoch": 812, "lr": 8.364022615839844e-05}
{"train_loss": 0.20534880459308624, "global_step": 72326, "epoch": 812, "lr": 8.363979725523041e-05}
{"train_loss": 0.15908676385879517, "global_step": 72327, "epoch": 812, "lr": 8.363936834753993e-05}
{"train_loss": 0.2467840611934662, "global_step": 72328, "epoch": 812, "lr": 8.363893943532707e-05}
{"train_loss": 0.24064640700817108, "global_step": 72329, "epoch": 812, "lr": 8.363851051859184e-05}
{"train_loss": 0.26565271615982056, "global_step": 72330, "epoch": 812, "lr": 8.363808159733435e-05}
{"train_loss": 0.20976075530052185, "global_step": 72331, "epoch": 812, "lr": 8.363765267155459e-05}
{"train_loss": 0.1972336769104004, "global_step": 72332, "epoch": 812, "lr": 8.36372237412527e-05}
{"train_loss": 0.15433262288570404, "global_step": 72333, "epoch": 812, "lr": 8.363679480642865e-05}
{"train_loss": 0.16795654594898224, "global_step": 72334, "epoch": 812, "lr": 8.363636586708257e-05}
{"train_loss": 0.1992625743150711, "global_step": 72335, "epoch": 812, "lr": 8.36359369232145e-05}
{"train_loss": 0.23892246186733246, "global_step": 72336, "epoch": 812, "lr": 8.363550797482447e-05}
{"train_loss": 0.19268475472927094, "global_step": 72337, "epoch": 812, "lr": 8.363507902191254e-05}
{"train_loss": 0.2394813895225525, "global_step": 72338, "epoch": 812, "lr": 8.363465006447882e-05}
{"train_loss": 0.15532401204109192, "global_step": 72339, "epoch": 812, "lr": 8.36342211025233e-05}
{"train_loss": 0.23435671627521515, "global_step": 72340, "epoch": 812, "lr": 8.363379213604609e-05}
{"train_loss": 0.1980447769165039, "global_step": 72341, "epoch": 812, "lr": 8.363336316504722e-05}
{"train_loss": 0.2210756242275238, "global_step": 72342, "epoch": 812, "lr": 8.363293418952676e-05}
{"train_loss": 0.19573357701301575, "global_step": 72343, "epoch": 812, "lr": 8.363250520948476e-05}
{"train_loss": 0.32243672013282776, "global_step": 72344, "epoch": 812, "lr": 8.363207622492129e-05}
{"train_loss": 0.25782403349876404, "global_step": 72345, "epoch": 812, "lr": 8.363164723583636e-05}
{"train_loss": 0.16874121129512787, "global_step": 72346, "epoch": 812, "lr": 8.36312182422301e-05}
{"train_loss": 0.23981764912605286, "global_step": 72347, "epoch": 812, "lr": 8.363078924410252e-05}
{"train_loss": 0.26037561893463135, "global_step": 72348, "epoch": 812, "lr": 8.36303602414537e-05}
{"train_loss": 0.18069703876972198, "global_step": 72349, "epoch": 812, "lr": 8.362993123428366e-05}
{"train_loss": 0.15753789246082306, "global_step": 72350, "epoch": 812, "lr": 8.362950222259251e-05}
{"train_loss": 0.20033548772335052, "global_step": 72351, "epoch": 812, "lr": 8.36290732063803e-05}
{"train_loss": 0.18421684205532074, "global_step": 72352, "epoch": 812, "lr": 8.362864418564704e-05}
{"train_loss": 0.19565936923027039, "global_step": 72353, "epoch": 812, "lr": 8.362821516039283e-05}
{"train_loss": 0.20574292540550232, "global_step": 72354, "epoch": 812, "lr": 8.362778613061772e-05}
{"train_loss": 0.19116361439228058, "global_step": 72355, "epoch": 812, "lr": 8.362735709632174e-05}
{"train_loss": 0.2215213045645296, "global_step": 72356, "epoch": 812, "lr": 8.362692805750499e-05, "val_loss": 3.7492127418518066}
{"train_loss": 0.2403619885444641, "global_step": 72357, "epoch": 813, "lr": 8.362649901416751e-05}
{"train_loss": 0.20018815994262695, "global_step": 72358, "epoch": 813, "lr": 8.362606996630936e-05}
{"train_loss": 0.17300821840763092, "global_step": 72359, "epoch": 813, "lr": 8.36256409139306e-05}
{"train_loss": 0.19275212287902832, "global_step": 72360, "epoch": 813, "lr": 8.362521185703127e-05}
{"train_loss": 0.09885197877883911, "global_step": 72361, "epoch": 813, "lr": 8.362478279561145e-05}
{"train_loss": 0.20313331484794617, "global_step": 72362, "epoch": 813, "lr": 8.362435372967118e-05}
{"train_loss": 0.179072767496109, "global_step": 72363, "epoch": 813, "lr": 8.362392465921052e-05}
{"train_loss": 0.11276016384363174, "global_step": 72364, "epoch": 813, "lr": 8.362349558422955e-05}
{"train_loss": 0.18942588567733765, "global_step": 72365, "epoch": 813, "lr": 8.36230665047283e-05}
{"train_loss": 0.18946880102157593, "global_step": 72366, "epoch": 813, "lr": 8.362263742070684e-05}
{"train_loss": 0.23404547572135925, "global_step": 72367, "epoch": 813, "lr": 8.362220833216522e-05}
{"train_loss": 0.13269300758838654, "global_step": 72368, "epoch": 813, "lr": 8.36217792391035e-05}
{"train_loss": 0.19726690649986267, "global_step": 72369, "epoch": 813, "lr": 8.362135014152177e-05}
{"train_loss": 0.2165323942899704, "global_step": 72370, "epoch": 813, "lr": 8.362092103942003e-05}
{"train_loss": 0.17211851477622986, "global_step": 72371, "epoch": 813, "lr": 8.362049193279839e-05}
{"train_loss": 0.191335067152977, "global_step": 72372, "epoch": 813, "lr": 8.362006282165687e-05}
{"train_loss": 0.1931540071964264, "global_step": 72373, "epoch": 813, "lr": 8.361963370599554e-05}
{"train_loss": 0.31933170557022095, "global_step": 72374, "epoch": 813, "lr": 8.361920458581446e-05}
{"train_loss": 0.20627310872077942, "global_step": 72375, "epoch": 813, "lr": 8.36187754611137e-05}
{"train_loss": 0.2115703821182251, "global_step": 72376, "epoch": 813, "lr": 8.36183463318933e-05}
{"train_loss": 0.12520159780979156, "global_step": 72377, "epoch": 813, "lr": 8.36179171981533e-05}
{"train_loss": 0.16194899380207062, "global_step": 72378, "epoch": 813, "lr": 8.36174880598938e-05}
{"train_loss": 0.1958315372467041, "global_step": 72379, "epoch": 813, "lr": 8.361705891711485e-05}
{"train_loss": 0.16794756054878235, "global_step": 72380, "epoch": 813, "lr": 8.361662976981648e-05}
{"train_loss": 0.19122953712940216, "global_step": 72381, "epoch": 813, "lr": 8.361620061799877e-05}
{"train_loss": 0.22409789264202118, "global_step": 72382, "epoch": 813, "lr": 8.361577146166178e-05}
{"train_loss": 0.14122305810451508, "global_step": 72383, "epoch": 813, "lr": 8.361534230080556e-05}
{"train_loss": 0.2171575278043747, "global_step": 72384, "epoch": 813, "lr": 8.361491313543015e-05}
{"train_loss": 0.265133261680603, "global_step": 72385, "epoch": 813, "lr": 8.361448396553565e-05}
{"train_loss": 0.24948789179325104, "global_step": 72386, "epoch": 813, "lr": 8.361405479112205e-05}
{"train_loss": 0.22065366804599762, "global_step": 72387, "epoch": 813, "lr": 8.361362561218948e-05}
{"train_loss": 0.1894746869802475, "global_step": 72388, "epoch": 813, "lr": 8.361319642873797e-05}
{"train_loss": 0.18587033450603485, "global_step": 72389, "epoch": 813, "lr": 8.361276724076757e-05}
{"train_loss": 0.17095308005809784, "global_step": 72390, "epoch": 813, "lr": 8.361233804827833e-05}
{"train_loss": 0.21578308939933777, "global_step": 72391, "epoch": 813, "lr": 8.361190885127033e-05}
{"train_loss": 0.22971990704536438, "global_step": 72392, "epoch": 813, "lr": 8.361147964974363e-05}
{"train_loss": 0.19194164872169495, "global_step": 72393, "epoch": 813, "lr": 8.361105044369826e-05}
{"train_loss": 0.16169337928295135, "global_step": 72394, "epoch": 813, "lr": 8.36106212331343e-05}
{"train_loss": 0.24975359439849854, "global_step": 72395, "epoch": 813, "lr": 8.36101920180518e-05}
{"train_loss": 0.2679482698440552, "global_step": 72396, "epoch": 813, "lr": 8.360976279845083e-05}
{"train_loss": 0.14741966128349304, "global_step": 72397, "epoch": 813, "lr": 8.360933357433142e-05}
{"train_loss": 0.2510317862033844, "global_step": 72398, "epoch": 813, "lr": 8.360890434569366e-05}
{"train_loss": 0.2568291127681732, "global_step": 72399, "epoch": 813, "lr": 8.360847511253757e-05}
{"train_loss": 0.14257566630840302, "global_step": 72400, "epoch": 813, "lr": 8.360804587486326e-05}
{"train_loss": 0.2385942041873932, "global_step": 72401, "epoch": 813, "lr": 8.360761663267075e-05}
{"train_loss": 0.16742248833179474, "global_step": 72402, "epoch": 813, "lr": 8.360718738596009e-05}
{"train_loss": 0.12990373373031616, "global_step": 72403, "epoch": 813, "lr": 8.360675813473137e-05}
{"train_loss": 0.2218616008758545, "global_step": 72404, "epoch": 813, "lr": 8.360632887898463e-05}
{"train_loss": 0.2016509622335434, "global_step": 72405, "epoch": 813, "lr": 8.360589961871991e-05}
{"train_loss": 0.20169584453105927, "global_step": 72406, "epoch": 813, "lr": 8.360547035393731e-05}
{"train_loss": 0.25235527753829956, "global_step": 72407, "epoch": 813, "lr": 8.360504108463684e-05}
{"train_loss": 0.28204742074012756, "global_step": 72408, "epoch": 813, "lr": 8.36046118108186e-05}
{"train_loss": 0.21678484976291656, "global_step": 72409, "epoch": 813, "lr": 8.360418253248264e-05}
{"train_loss": 0.20540036261081696, "global_step": 72410, "epoch": 813, "lr": 8.360375324962896e-05}
{"train_loss": 0.19611401855945587, "global_step": 72411, "epoch": 813, "lr": 8.36033239622577e-05}
{"train_loss": 0.17595846951007843, "global_step": 72412, "epoch": 813, "lr": 8.360289467036888e-05}
{"train_loss": 0.21060408651828766, "global_step": 72413, "epoch": 813, "lr": 8.360246537396257e-05}
{"train_loss": 0.200958251953125, "global_step": 72414, "epoch": 813, "lr": 8.36020360730388e-05}
{"train_loss": 0.18010163307189941, "global_step": 72415, "epoch": 813, "lr": 8.360160676759765e-05}
{"train_loss": 0.21013957262039185, "global_step": 72416, "epoch": 813, "lr": 8.360117745763917e-05}
{"train_loss": 0.11273025721311569, "global_step": 72417, "epoch": 813, "lr": 8.360074814316344e-05}
{"train_loss": 0.33971208333969116, "global_step": 72418, "epoch": 813, "lr": 8.360031882417049e-05}
{"train_loss": 0.2651110291481018, "global_step": 72419, "epoch": 813, "lr": 8.35998895006604e-05}
{"train_loss": 0.2607437074184418, "global_step": 72420, "epoch": 813, "lr": 8.359946017263318e-05}
{"train_loss": 0.15147665143013, "global_step": 72421, "epoch": 813, "lr": 8.359903084008894e-05}
{"train_loss": 0.2564975619316101, "global_step": 72422, "epoch": 813, "lr": 8.359860150302772e-05}
{"train_loss": 0.2302348017692566, "global_step": 72423, "epoch": 813, "lr": 8.359817216144959e-05}
{"train_loss": 0.22305481135845184, "global_step": 72424, "epoch": 813, "lr": 8.359774281535457e-05}
{"train_loss": 0.21095171570777893, "global_step": 72425, "epoch": 813, "lr": 8.359731346474275e-05}
{"train_loss": 0.21101005375385284, "global_step": 72426, "epoch": 813, "lr": 8.359688410961419e-05}
{"train_loss": 0.10598528385162354, "global_step": 72427, "epoch": 813, "lr": 8.359645474996894e-05}
{"train_loss": 0.20791774988174438, "global_step": 72428, "epoch": 813, "lr": 8.359602538580704e-05}
{"train_loss": 0.29292890429496765, "global_step": 72429, "epoch": 813, "lr": 8.359559601712858e-05}
{"train_loss": 0.26832425594329834, "global_step": 72430, "epoch": 813, "lr": 8.359516664393361e-05}
{"train_loss": 0.2862173616886139, "global_step": 72431, "epoch": 813, "lr": 8.359473726622215e-05}
{"train_loss": 0.18826743960380554, "global_step": 72432, "epoch": 813, "lr": 8.359430788399431e-05}
{"train_loss": 0.19350092113018036, "global_step": 72433, "epoch": 813, "lr": 8.359387849725011e-05}
{"train_loss": 0.1868852823972702, "global_step": 72434, "epoch": 813, "lr": 8.359344910598963e-05}
{"train_loss": 0.17506909370422363, "global_step": 72435, "epoch": 813, "lr": 8.359301971021292e-05}
{"train_loss": 0.25423523783683777, "global_step": 72436, "epoch": 813, "lr": 8.359259030992004e-05}
{"train_loss": 0.25194570422172546, "global_step": 72437, "epoch": 813, "lr": 8.359216090511105e-05}
{"train_loss": 0.21776524186134338, "global_step": 72438, "epoch": 813, "lr": 8.3591731495786e-05}
{"train_loss": 0.23227761685848236, "global_step": 72439, "epoch": 813, "lr": 8.359130208194494e-05}
{"train_loss": 0.2464168518781662, "global_step": 72440, "epoch": 813, "lr": 8.359087266358796e-05}
{"train_loss": 0.28950178623199463, "global_step": 72441, "epoch": 813, "lr": 8.359044324071508e-05}
{"train_loss": 0.2311473786830902, "global_step": 72442, "epoch": 813, "lr": 8.359001381332638e-05}
{"train_loss": 0.20005282759666443, "global_step": 72443, "epoch": 813, "lr": 8.35895843814219e-05}
{"train_loss": 0.15868976712226868, "global_step": 72444, "epoch": 813, "lr": 8.358915494500172e-05}
{"train_loss": 0.2075375866019324, "global_step": 72445, "epoch": 813, "lr": 8.35887255040659e-05, "val_loss": 3.8392586708068848}
{"train_loss": 0.25233739614486694, "global_step": 72446, "epoch": 814, "lr": 8.358829605861447e-05}
{"train_loss": 0.25735899806022644, "global_step": 72447, "epoch": 814, "lr": 8.35878666086475e-05}
{"train_loss": 0.162612184882164, "global_step": 72448, "epoch": 814, "lr": 8.358743715416506e-05}
{"train_loss": 0.18646760284900665, "global_step": 72449, "epoch": 814, "lr": 8.358700769516719e-05}
{"train_loss": 0.19063426554203033, "global_step": 72450, "epoch": 814, "lr": 8.358657823165397e-05}
{"train_loss": 0.2262018322944641, "global_step": 72451, "epoch": 814, "lr": 8.358614876362542e-05}
{"train_loss": 0.1966598629951477, "global_step": 72452, "epoch": 814, "lr": 8.358571929108165e-05}
{"train_loss": 0.26266542077064514, "global_step": 72453, "epoch": 814, "lr": 8.358528981402267e-05}
{"train_loss": 0.27010658383369446, "global_step": 72454, "epoch": 814, "lr": 8.358486033244857e-05}
{"train_loss": 0.1897190809249878, "global_step": 72455, "epoch": 814, "lr": 8.358443084635938e-05}
{"train_loss": 0.17325790226459503, "global_step": 72456, "epoch": 814, "lr": 8.358400135575518e-05}
{"train_loss": 0.2429731786251068, "global_step": 72457, "epoch": 814, "lr": 8.358357186063602e-05}
{"train_loss": 0.16539359092712402, "global_step": 72458, "epoch": 814, "lr": 8.358314236100196e-05}
{"train_loss": 0.14633648097515106, "global_step": 72459, "epoch": 814, "lr": 8.358271285685307e-05}
{"train_loss": 0.18790563941001892, "global_step": 72460, "epoch": 814, "lr": 8.358228334818937e-05}
{"train_loss": 0.20801757276058197, "global_step": 72461, "epoch": 814, "lr": 8.358185383501095e-05}
{"train_loss": 0.2138235867023468, "global_step": 72462, "epoch": 814, "lr": 8.358142431731785e-05}
{"train_loss": 0.19893839955329895, "global_step": 72463, "epoch": 814, "lr": 8.358099479511015e-05}
{"train_loss": 0.1469469964504242, "global_step": 72464, "epoch": 814, "lr": 8.35805652683879e-05}
{"train_loss": 0.1953262984752655, "global_step": 72465, "epoch": 814, "lr": 8.358013573715113e-05}
{"train_loss": 0.2723270356655121, "global_step": 72466, "epoch": 814, "lr": 8.357970620139994e-05}
{"train_loss": 0.26183369755744934, "global_step": 72467, "epoch": 814, "lr": 8.357927666113436e-05}
{"train_loss": 0.2408917248249054, "global_step": 72468, "epoch": 814, "lr": 8.357884711635446e-05}
{"train_loss": 0.11011181026697159, "global_step": 72469, "epoch": 814, "lr": 8.357841756706028e-05}
{"train_loss": 0.341395765542984, "global_step": 72470, "epoch": 814, "lr": 8.357798801325189e-05}
{"train_loss": 0.2272430658340454, "global_step": 72471, "epoch": 814, "lr": 8.357755845492936e-05}
{"train_loss": 0.27595946192741394, "global_step": 72472, "epoch": 814, "lr": 8.357712889209275e-05}
{"train_loss": 0.28757449984550476, "global_step": 72473, "epoch": 814, "lr": 8.357669932474208e-05}
{"train_loss": 0.1921573281288147, "global_step": 72474, "epoch": 814, "lr": 8.357626975287744e-05}
{"train_loss": 0.2800366282463074, "global_step": 72475, "epoch": 814, "lr": 8.357584017649889e-05}
{"train_loss": 0.15778061747550964, "global_step": 72476, "epoch": 814, "lr": 8.357541059560648e-05}
{"train_loss": 0.17822863161563873, "global_step": 72477, "epoch": 814, "lr": 8.357498101020025e-05}
{"train_loss": 0.1908777505159378, "global_step": 72478, "epoch": 814, "lr": 8.357455142028028e-05}
{"train_loss": 0.2813315689563751, "global_step": 72479, "epoch": 814, "lr": 8.357412182584661e-05}
{"train_loss": 0.166462704539299, "global_step": 72480, "epoch": 814, "lr": 8.357369222689932e-05}
{"train_loss": 0.1735512912273407, "global_step": 72481, "epoch": 814, "lr": 8.357326262343845e-05}
{"train_loss": 0.20797273516654968, "global_step": 72482, "epoch": 814, "lr": 8.357283301546405e-05}
{"train_loss": 0.2969410717487335, "global_step": 72483, "epoch": 814, "lr": 8.357240340297622e-05}
{"train_loss": 0.2481914460659027, "global_step": 72484, "epoch": 814, "lr": 8.357197378597498e-05}
{"train_loss": 0.21738779544830322, "global_step": 72485, "epoch": 814, "lr": 8.357154416446039e-05}
{"train_loss": 0.26190871000289917, "global_step": 72486, "epoch": 814, "lr": 8.357111453843252e-05}
{"train_loss": 0.17042948305606842, "global_step": 72487, "epoch": 814, "lr": 8.357068490789143e-05}
{"train_loss": 0.2535066306591034, "global_step": 72488, "epoch": 814, "lr": 8.357025527283716e-05}
{"train_loss": 0.14434751868247986, "global_step": 72489, "epoch": 814, "lr": 8.356982563326979e-05}
{"train_loss": 0.15819396078586578, "global_step": 72490, "epoch": 814, "lr": 8.356939598918935e-05}
{"train_loss": 0.15030278265476227, "global_step": 72491, "epoch": 814, "lr": 8.356896634059592e-05}
{"train_loss": 0.18266570568084717, "global_step": 72492, "epoch": 814, "lr": 8.356853668748956e-05}
{"train_loss": 0.2866211533546448, "global_step": 72493, "epoch": 814, "lr": 8.356810702987031e-05}
{"train_loss": 0.19583265483379364, "global_step": 72494, "epoch": 814, "lr": 8.356767736773826e-05}
{"train_loss": 0.23866590857505798, "global_step": 72495, "epoch": 814, "lr": 8.356724770109341e-05}
{"train_loss": 0.2790793776512146, "global_step": 72496, "epoch": 814, "lr": 8.356681802993587e-05}
{"train_loss": 0.2679683566093445, "global_step": 72497, "epoch": 814, "lr": 8.356638835426568e-05}
{"train_loss": 0.1802607774734497, "global_step": 72498, "epoch": 814, "lr": 8.35659586740829e-05}
{"train_loss": 0.17875604331493378, "global_step": 72499, "epoch": 814, "lr": 8.356552898938758e-05}
{"train_loss": 0.20499084889888763, "global_step": 72500, "epoch": 814, "lr": 8.356509930017978e-05}
{"train_loss": 0.16484485566616058, "global_step": 72501, "epoch": 814, "lr": 8.356466960645957e-05}
{"train_loss": 0.15774351358413696, "global_step": 72502, "epoch": 814, "lr": 8.3564239908227e-05}
{"train_loss": 0.17889629304409027, "global_step": 72503, "epoch": 814, "lr": 8.356381020548212e-05}
{"train_loss": 0.23529335856437683, "global_step": 72504, "epoch": 814, "lr": 8.3563380498225e-05}
{"train_loss": 0.18312515318393707, "global_step": 72505, "epoch": 814, "lr": 8.356295078645568e-05}
{"train_loss": 0.23125731945037842, "global_step": 72506, "epoch": 814, "lr": 8.356252107017426e-05}
{"train_loss": 0.2923870086669922, "global_step": 72507, "epoch": 814, "lr": 8.356209134938075e-05}
{"train_loss": 0.15311509370803833, "global_step": 72508, "epoch": 814, "lr": 8.356166162407521e-05}
{"train_loss": 0.2168523073196411, "global_step": 72509, "epoch": 814, "lr": 8.356123189425774e-05}
{"train_loss": 0.22160235047340393, "global_step": 72510, "epoch": 814, "lr": 8.356080215992836e-05}
{"train_loss": 0.1896534264087677, "global_step": 72511, "epoch": 814, "lr": 8.356037242108714e-05}
{"train_loss": 0.30650776624679565, "global_step": 72512, "epoch": 814, "lr": 8.355994267773414e-05}
{"train_loss": 0.17647333443164825, "global_step": 72513, "epoch": 814, "lr": 8.355951292986941e-05}
{"train_loss": 0.2453346699476242, "global_step": 72514, "epoch": 814, "lr": 8.355908317749301e-05}
{"train_loss": 0.18212099373340607, "global_step": 72515, "epoch": 814, "lr": 8.3558653420605e-05}
{"train_loss": 0.31347939372062683, "global_step": 72516, "epoch": 814, "lr": 8.355822365920544e-05}
{"train_loss": 0.26184701919555664, "global_step": 72517, "epoch": 814, "lr": 8.35577938932944e-05}
{"train_loss": 0.23132218420505524, "global_step": 72518, "epoch": 814, "lr": 8.355736412287192e-05}
{"train_loss": 0.23990190029144287, "global_step": 72519, "epoch": 814, "lr": 8.355693434793805e-05}
{"train_loss": 0.21490031480789185, "global_step": 72520, "epoch": 814, "lr": 8.355650456849286e-05}
{"train_loss": 0.19667062163352966, "global_step": 72521, "epoch": 814, "lr": 8.35560747845364e-05}
{"train_loss": 0.28849437832832336, "global_step": 72522, "epoch": 814, "lr": 8.355564499606874e-05}
{"train_loss": 0.21222785115242004, "global_step": 72523, "epoch": 814, "lr": 8.355521520308995e-05}
{"train_loss": 0.1869909167289734, "global_step": 72524, "epoch": 814, "lr": 8.355478540560006e-05}
{"train_loss": 0.23395615816116333, "global_step": 72525, "epoch": 814, "lr": 8.355435560359913e-05}
{"train_loss": 0.24044592678546906, "global_step": 72526, "epoch": 814, "lr": 8.355392579708723e-05}
{"train_loss": 0.24529211223125458, "global_step": 72527, "epoch": 814, "lr": 8.355349598606441e-05}
{"train_loss": 0.23901082575321198, "global_step": 72528, "epoch": 814, "lr": 8.355306617053074e-05}
{"train_loss": 0.2888055741786957, "global_step": 72529, "epoch": 814, "lr": 8.355263635048626e-05}
{"train_loss": 0.2558627128601074, "global_step": 72530, "epoch": 814, "lr": 8.355220652593106e-05}
{"train_loss": 0.3071392774581909, "global_step": 72531, "epoch": 814, "lr": 8.355177669686516e-05}
{"train_loss": 0.22497589886188507, "global_step": 72532, "epoch": 814, "lr": 8.355134686328862e-05}
{"train_loss": 0.23594416677951813, "global_step": 72533, "epoch": 814, "lr": 8.355091702520153e-05}
{"train_loss": 0.22041668608952103, "global_step": 72534, "epoch": 814, "lr": 8.355048718260391e-05, "val_loss": 3.663883924484253}
{"train_loss": 0.21775798499584198, "global_step": 72535, "epoch": 815, "lr": 8.355005733549585e-05}
{"train_loss": 0.21624620258808136, "global_step": 72536, "epoch": 815, "lr": 8.354962748387738e-05}
{"train_loss": 0.23959629237651825, "global_step": 72537, "epoch": 815, "lr": 8.354919762774858e-05}
{"train_loss": 0.1938987821340561, "global_step": 72538, "epoch": 815, "lr": 8.35487677671095e-05}
{"train_loss": 0.20279118418693542, "global_step": 72539, "epoch": 815, "lr": 8.354833790196019e-05}
{"train_loss": 0.2179276943206787, "global_step": 72540, "epoch": 815, "lr": 8.354790803230072e-05}
{"train_loss": 0.22694475948810577, "global_step": 72541, "epoch": 815, "lr": 8.354747815813115e-05}
{"train_loss": 0.22706475853919983, "global_step": 72542, "epoch": 815, "lr": 8.354704827945153e-05}
{"train_loss": 0.23170170187950134, "global_step": 72543, "epoch": 815, "lr": 8.35466183962619e-05}
{"train_loss": 0.19172678887844086, "global_step": 72544, "epoch": 815, "lr": 8.354618850856235e-05}
{"train_loss": 0.21939358115196228, "global_step": 72545, "epoch": 815, "lr": 8.354575861635292e-05}
{"train_loss": 0.16225120425224304, "global_step": 72546, "epoch": 815, "lr": 8.354532871963368e-05}
{"train_loss": 0.24123935401439667, "global_step": 72547, "epoch": 815, "lr": 8.354489881840467e-05}
{"train_loss": 0.2379743903875351, "global_step": 72548, "epoch": 815, "lr": 8.354446891266596e-05}
{"train_loss": 0.22293035686016083, "global_step": 72549, "epoch": 815, "lr": 8.354403900241761e-05}
{"train_loss": 0.1892653852701187, "global_step": 72550, "epoch": 815, "lr": 8.354360908765966e-05}
{"train_loss": 0.207842156291008, "global_step": 72551, "epoch": 815, "lr": 8.35431791683922e-05}
{"train_loss": 0.14456871151924133, "global_step": 72552, "epoch": 815, "lr": 8.354274924461525e-05}
{"train_loss": 0.16363508999347687, "global_step": 72553, "epoch": 815, "lr": 8.354231931632889e-05}
{"train_loss": 0.20509079098701477, "global_step": 72554, "epoch": 815, "lr": 8.354188938353317e-05}
{"train_loss": 0.17900483310222626, "global_step": 72555, "epoch": 815, "lr": 8.354145944622817e-05}
{"train_loss": 0.216844841837883, "global_step": 72556, "epoch": 815, "lr": 8.354102950441392e-05}
{"train_loss": 0.2152811586856842, "global_step": 72557, "epoch": 815, "lr": 8.354059955809049e-05}
{"train_loss": 0.17263369262218475, "global_step": 72558, "epoch": 815, "lr": 8.354016960725793e-05}
{"train_loss": 0.2009945660829544, "global_step": 72559, "epoch": 815, "lr": 8.353973965191631e-05}
{"train_loss": 0.16878026723861694, "global_step": 72560, "epoch": 815, "lr": 8.353930969206565e-05}
{"train_loss": 0.20360811054706573, "global_step": 72561, "epoch": 815, "lr": 8.353887972770607e-05}
{"train_loss": 0.16414125263690948, "global_step": 72562, "epoch": 815, "lr": 8.35384497588376e-05}
{"train_loss": 0.13518579304218292, "global_step": 72563, "epoch": 815, "lr": 8.353801978546027e-05}
{"train_loss": 0.24270422756671906, "global_step": 72564, "epoch": 815, "lr": 8.35375898075742e-05}
{"train_loss": 0.19520558416843414, "global_step": 72565, "epoch": 815, "lr": 8.353715982517937e-05}
{"train_loss": 0.1501525491476059, "global_step": 72566, "epoch": 815, "lr": 8.35367298382759e-05}
{"train_loss": 0.17507845163345337, "global_step": 72567, "epoch": 815, "lr": 8.353629984686382e-05}
{"train_loss": 0.25435230135917664, "global_step": 72568, "epoch": 815, "lr": 8.353586985094319e-05}
{"train_loss": 0.16218946874141693, "global_step": 72569, "epoch": 815, "lr": 8.353543985051408e-05}
{"train_loss": 0.21329301595687866, "global_step": 72570, "epoch": 815, "lr": 8.353500984557652e-05}
{"train_loss": 0.195574089884758, "global_step": 72571, "epoch": 815, "lr": 8.35345798361306e-05}
{"train_loss": 0.1909380406141281, "global_step": 72572, "epoch": 815, "lr": 8.353414982217637e-05}
{"train_loss": 0.1525854617357254, "global_step": 72573, "epoch": 815, "lr": 8.353371980371387e-05}
{"train_loss": 0.1857243925333023, "global_step": 72574, "epoch": 815, "lr": 8.353328978074317e-05}
{"train_loss": 0.20735691487789154, "global_step": 72575, "epoch": 815, "lr": 8.353285975326434e-05}
{"train_loss": 0.17668305337429047, "global_step": 72576, "epoch": 815, "lr": 8.353242972127741e-05}
{"train_loss": 0.21906419098377228, "global_step": 72577, "epoch": 815, "lr": 8.353199968478247e-05}
{"train_loss": 0.24423067271709442, "global_step": 72578, "epoch": 815, "lr": 8.353156964377954e-05}
{"train_loss": 0.22224649786949158, "global_step": 72579, "epoch": 815, "lr": 8.353113959826873e-05}
{"train_loss": 0.22340355813503265, "global_step": 72580, "epoch": 815, "lr": 8.353070954825004e-05}
{"train_loss": 0.1428866982460022, "global_step": 72581, "epoch": 815, "lr": 8.353027949372358e-05}
{"train_loss": 0.1887032836675644, "global_step": 72582, "epoch": 815, "lr": 8.352984943468937e-05}
{"train_loss": 0.18379491567611694, "global_step": 72583, "epoch": 815, "lr": 8.352941937114746e-05}
{"train_loss": 0.1568666249513626, "global_step": 72584, "epoch": 815, "lr": 8.352898930309796e-05}
{"train_loss": 0.17448721826076508, "global_step": 72585, "epoch": 815, "lr": 8.352855923054088e-05}
{"train_loss": 0.23727840185165405, "global_step": 72586, "epoch": 815, "lr": 8.35281291534763e-05}
{"train_loss": 0.25025510787963867, "global_step": 72587, "epoch": 815, "lr": 8.352769907190426e-05}
{"train_loss": 0.1884026825428009, "global_step": 72588, "epoch": 815, "lr": 8.352726898582484e-05}
{"train_loss": 0.21930694580078125, "global_step": 72589, "epoch": 815, "lr": 8.352683889523809e-05}
{"train_loss": 0.31373918056488037, "global_step": 72590, "epoch": 815, "lr": 8.352640880014406e-05}
{"train_loss": 0.1853155493736267, "global_step": 72591, "epoch": 815, "lr": 8.35259787005428e-05}
{"train_loss": 0.18826919794082642, "global_step": 72592, "epoch": 815, "lr": 8.35255485964344e-05}
{"train_loss": 0.13858269155025482, "global_step": 72593, "epoch": 815, "lr": 8.352511848781889e-05}
{"train_loss": 0.20738543570041656, "global_step": 72594, "epoch": 815, "lr": 8.352468837469634e-05}
{"train_loss": 0.20865783095359802, "global_step": 72595, "epoch": 815, "lr": 8.35242582570668e-05}
{"train_loss": 0.21288780868053436, "global_step": 72596, "epoch": 815, "lr": 8.352382813493033e-05}
{"train_loss": 0.22067634761333466, "global_step": 72597, "epoch": 815, "lr": 8.3523398008287e-05}
{"train_loss": 0.16123482584953308, "global_step": 72598, "epoch": 815, "lr": 8.352296787713685e-05}
{"train_loss": 0.20136329531669617, "global_step": 72599, "epoch": 815, "lr": 8.352253774147996e-05}
{"train_loss": 0.1918071061372757, "global_step": 72600, "epoch": 815, "lr": 8.352210760131636e-05}
{"train_loss": 0.2227177768945694, "global_step": 72601, "epoch": 815, "lr": 8.352167745664612e-05}
{"train_loss": 0.13768304884433746, "global_step": 72602, "epoch": 815, "lr": 8.35212473074693e-05}
{"train_loss": 0.23191609978675842, "global_step": 72603, "epoch": 815, "lr": 8.352081715378596e-05}
{"train_loss": 0.20961830019950867, "global_step": 72604, "epoch": 815, "lr": 8.352038699559616e-05}
{"train_loss": 0.17960308492183685, "global_step": 72605, "epoch": 815, "lr": 8.351995683289995e-05}
{"train_loss": 0.15144340693950653, "global_step": 72606, "epoch": 815, "lr": 8.351952666569738e-05}
{"train_loss": 0.19314990937709808, "global_step": 72607, "epoch": 815, "lr": 8.351909649398852e-05}
{"train_loss": 0.12534227967262268, "global_step": 72608, "epoch": 815, "lr": 8.351866631777343e-05}
{"train_loss": 0.14954319596290588, "global_step": 72609, "epoch": 815, "lr": 8.351823613705217e-05}
{"train_loss": 0.11112203449010849, "global_step": 72610, "epoch": 815, "lr": 8.351780595182478e-05}
{"train_loss": 0.20360971987247467, "global_step": 72611, "epoch": 815, "lr": 8.351737576209133e-05}
{"train_loss": 0.17965957522392273, "global_step": 72612, "epoch": 815, "lr": 8.351694556785189e-05}
{"train_loss": 0.151381254196167, "global_step": 72613, "epoch": 815, "lr": 8.35165153691065e-05}
{"train_loss": 0.18659919500350952, "global_step": 72614, "epoch": 815, "lr": 8.351608516585522e-05}
{"train_loss": 0.19483071565628052, "global_step": 72615, "epoch": 815, "lr": 8.351565495809811e-05}
{"train_loss": 0.20996439456939697, "global_step": 72616, "epoch": 815, "lr": 8.351522474583523e-05}
{"train_loss": 0.1920730173587799, "global_step": 72617, "epoch": 815, "lr": 8.351479452906663e-05}
{"train_loss": 0.27517491579055786, "global_step": 72618, "epoch": 815, "lr": 8.351436430779239e-05}
{"train_loss": 0.21217463910579681, "global_step": 72619, "epoch": 815, "lr": 8.351393408201254e-05}
{"train_loss": 0.17064054310321808, "global_step": 72620, "epoch": 815, "lr": 8.351350385172716e-05}
{"train_loss": 0.2070300579071045, "global_step": 72621, "epoch": 815, "lr": 8.35130736169363e-05}
{"train_loss": 0.2195395678281784, "global_step": 72622, "epoch": 815, "lr": 8.351264337764e-05}
{"train_loss": 0.19723481668180295, "global_step": 72623, "epoch": 815, "lr": 8.351221313383835e-05, "val_loss": 3.8914551734924316, "train_action_mse_error": 39.535308837890625}
{"train_loss": 0.18353350460529327, "global_step": 72624, "epoch": 816, "lr": 8.351178288553138e-05}
{"train_loss": 0.17938658595085144, "global_step": 72625, "epoch": 816, "lr": 8.351135263271915e-05}
{"train_loss": 0.12423056364059448, "global_step": 72626, "epoch": 816, "lr": 8.351092237540174e-05}
{"train_loss": 0.16428117454051971, "global_step": 72627, "epoch": 816, "lr": 8.35104921135792e-05}
{"train_loss": 0.2125002145767212, "global_step": 72628, "epoch": 816, "lr": 8.351006184725157e-05}
{"train_loss": 0.26077741384506226, "global_step": 72629, "epoch": 816, "lr": 8.350963157641893e-05}
{"train_loss": 0.312611848115921, "global_step": 72630, "epoch": 816, "lr": 8.350920130108132e-05}
{"train_loss": 0.2128850817680359, "global_step": 72631, "epoch": 816, "lr": 8.350877102123881e-05}
{"train_loss": 0.2444458305835724, "global_step": 72632, "epoch": 816, "lr": 8.350834073689146e-05}
{"train_loss": 0.20335599780082703, "global_step": 72633, "epoch": 816, "lr": 8.350791044803932e-05}
{"train_loss": 0.24213038384914398, "global_step": 72634, "epoch": 816, "lr": 8.350748015468244e-05}
{"train_loss": 0.13478165864944458, "global_step": 72635, "epoch": 816, "lr": 8.35070498568209e-05}
{"train_loss": 0.17876020073890686, "global_step": 72636, "epoch": 816, "lr": 8.350661955445473e-05}
{"train_loss": 0.26597848534584045, "global_step": 72637, "epoch": 816, "lr": 8.350618924758402e-05}
{"train_loss": 0.15836820006370544, "global_step": 72638, "epoch": 816, "lr": 8.350575893620881e-05}
{"train_loss": 0.22866696119308472, "global_step": 72639, "epoch": 816, "lr": 8.350532862032916e-05}
{"train_loss": 0.22058936953544617, "global_step": 72640, "epoch": 816, "lr": 8.350489829994512e-05}
{"train_loss": 0.23444820940494537, "global_step": 72641, "epoch": 816, "lr": 8.350446797505676e-05}
{"train_loss": 0.1467868685722351, "global_step": 72642, "epoch": 816, "lr": 8.350403764566413e-05}
{"train_loss": 0.21635448932647705, "global_step": 72643, "epoch": 816, "lr": 8.350360731176729e-05}
{"train_loss": 0.21852019429206848, "global_step": 72644, "epoch": 816, "lr": 8.350317697336629e-05}
{"train_loss": 0.2115069180727005, "global_step": 72645, "epoch": 816, "lr": 8.350274663046121e-05}
{"train_loss": 0.2082340270280838, "global_step": 72646, "epoch": 816, "lr": 8.350231628305209e-05}
{"train_loss": 0.21068161725997925, "global_step": 72647, "epoch": 816, "lr": 8.3501885931139e-05}
{"train_loss": 0.24093830585479736, "global_step": 72648, "epoch": 816, "lr": 8.350145557472195e-05}
{"train_loss": 0.27090054750442505, "global_step": 72649, "epoch": 816, "lr": 8.350102521380107e-05}
{"train_loss": 0.19809593260288239, "global_step": 72650, "epoch": 816, "lr": 8.350059484837639e-05}
{"train_loss": 0.2003791779279709, "global_step": 72651, "epoch": 816, "lr": 8.350016447844795e-05}
{"train_loss": 0.1923818439245224, "global_step": 72652, "epoch": 816, "lr": 8.349973410401584e-05}
{"train_loss": 0.17630864679813385, "global_step": 72653, "epoch": 816, "lr": 8.349930372508007e-05}
{"train_loss": 0.320425420999527, "global_step": 72654, "epoch": 816, "lr": 8.349887334164075e-05}
{"train_loss": 0.19019830226898193, "global_step": 72655, "epoch": 816, "lr": 8.349844295369791e-05}
{"train_loss": 0.17230871319770813, "global_step": 72656, "epoch": 816, "lr": 8.349801256125162e-05}
{"train_loss": 0.3184433877468109, "global_step": 72657, "epoch": 816, "lr": 8.349758216430192e-05}
{"train_loss": 0.20808017253875732, "global_step": 72658, "epoch": 816, "lr": 8.349715176284887e-05}
{"train_loss": 0.19029459357261658, "global_step": 72659, "epoch": 816, "lr": 8.349672135689256e-05}
{"train_loss": 0.1430249810218811, "global_step": 72660, "epoch": 816, "lr": 8.3496290946433e-05}
{"train_loss": 0.23598939180374146, "global_step": 72661, "epoch": 816, "lr": 8.34958605314703e-05}
{"train_loss": 0.15470804274082184, "global_step": 72662, "epoch": 816, "lr": 8.349543011200446e-05}
{"train_loss": 0.18694166839122772, "global_step": 72663, "epoch": 816, "lr": 8.349499968803558e-05}
{"train_loss": 0.2697520852088928, "global_step": 72664, "epoch": 816, "lr": 8.349456925956369e-05}
{"train_loss": 0.27273446321487427, "global_step": 72665, "epoch": 816, "lr": 8.349413882658888e-05}
{"train_loss": 0.16569769382476807, "global_step": 72666, "epoch": 816, "lr": 8.349370838911119e-05}
{"train_loss": 0.17096903920173645, "global_step": 72667, "epoch": 816, "lr": 8.349327794713069e-05}
{"train_loss": 0.2138528823852539, "global_step": 72668, "epoch": 816, "lr": 8.349284750064741e-05}
{"train_loss": 0.22767114639282227, "global_step": 72669, "epoch": 816, "lr": 8.349241704966143e-05}
{"train_loss": 0.22763584554195404, "global_step": 72670, "epoch": 816, "lr": 8.349198659417282e-05}
{"train_loss": 0.2881792485713959, "global_step": 72671, "epoch": 816, "lr": 8.34915561341816e-05}
{"train_loss": 0.319240540266037, "global_step": 72672, "epoch": 816, "lr": 8.349112566968784e-05}
{"train_loss": 0.2962401509284973, "global_step": 72673, "epoch": 816, "lr": 8.349069520069161e-05}
{"train_loss": 0.18977466225624084, "global_step": 72674, "epoch": 816, "lr": 8.3490264727193e-05}
{"train_loss": 0.20843371748924255, "global_step": 72675, "epoch": 816, "lr": 8.348983424919198e-05}
{"train_loss": 0.22958683967590332, "global_step": 72676, "epoch": 816, "lr": 8.34894037666887e-05}
{"train_loss": 0.2103479504585266, "global_step": 72677, "epoch": 816, "lr": 8.348897327968316e-05}
{"train_loss": 0.23571915924549103, "global_step": 72678, "epoch": 816, "lr": 8.348854278817543e-05}
{"train_loss": 0.18694396317005157, "global_step": 72679, "epoch": 816, "lr": 8.34881122921656e-05}
{"train_loss": 0.21855227649211884, "global_step": 72680, "epoch": 816, "lr": 8.348768179165367e-05}
{"train_loss": 0.22969576716423035, "global_step": 72681, "epoch": 816, "lr": 8.348725128663974e-05}
{"train_loss": 0.26927319169044495, "global_step": 72682, "epoch": 816, "lr": 8.348682077712386e-05}
{"train_loss": 0.23654761910438538, "global_step": 72683, "epoch": 816, "lr": 8.348639026310609e-05}
{"train_loss": 0.1624789834022522, "global_step": 72684, "epoch": 816, "lr": 8.348595974458649e-05}
{"train_loss": 0.2024184763431549, "global_step": 72685, "epoch": 816, "lr": 8.34855292215651e-05}
{"train_loss": 0.17518411576747894, "global_step": 72686, "epoch": 816, "lr": 8.348509869404197e-05}
{"train_loss": 0.2557004988193512, "global_step": 72687, "epoch": 816, "lr": 8.34846681620172e-05}
{"train_loss": 0.154376819729805, "global_step": 72688, "epoch": 816, "lr": 8.348423762549082e-05}
{"train_loss": 0.18851706385612488, "global_step": 72689, "epoch": 816, "lr": 8.348380708446289e-05}
{"train_loss": 0.21416009962558746, "global_step": 72690, "epoch": 816, "lr": 8.348337653893348e-05}
{"train_loss": 0.1419154405593872, "global_step": 72691, "epoch": 816, "lr": 8.348294598890263e-05}
{"train_loss": 0.19137464463710785, "global_step": 72692, "epoch": 816, "lr": 8.34825154343704e-05}
{"train_loss": 0.16906815767288208, "global_step": 72693, "epoch": 816, "lr": 8.348208487533686e-05}
{"train_loss": 0.2699337899684906, "global_step": 72694, "epoch": 816, "lr": 8.348165431180206e-05}
{"train_loss": 0.24421383440494537, "global_step": 72695, "epoch": 816, "lr": 8.348122374376605e-05}
{"train_loss": 0.22502589225769043, "global_step": 72696, "epoch": 816, "lr": 8.348079317122893e-05}
{"train_loss": 0.2587383985519409, "global_step": 72697, "epoch": 816, "lr": 8.34803625941907e-05}
{"train_loss": 0.3177749812602997, "global_step": 72698, "epoch": 816, "lr": 8.347993201265145e-05}
{"train_loss": 0.19457435607910156, "global_step": 72699, "epoch": 816, "lr": 8.347950142661123e-05}
{"train_loss": 0.24793940782546997, "global_step": 72700, "epoch": 816, "lr": 8.347907083607009e-05}
{"train_loss": 0.17032703757286072, "global_step": 72701, "epoch": 816, "lr": 8.34786402410281e-05}
{"train_loss": 0.28276634216308594, "global_step": 72702, "epoch": 816, "lr": 8.347820964148531e-05}
{"train_loss": 0.279704213142395, "global_step": 72703, "epoch": 816, "lr": 8.34777790374418e-05}
{"train_loss": 0.1658995896577835, "global_step": 72704, "epoch": 816, "lr": 8.34773484288976e-05}
{"train_loss": 0.25500816106796265, "global_step": 72705, "epoch": 816, "lr": 8.347691781585279e-05}
{"train_loss": 0.22466152906417847, "global_step": 72706, "epoch": 816, "lr": 8.347648719830741e-05}
{"train_loss": 0.18898923695087433, "global_step": 72707, "epoch": 816, "lr": 8.347605657626152e-05}
{"train_loss": 0.22039420902729034, "global_step": 72708, "epoch": 816, "lr": 8.347562594971518e-05}
{"train_loss": 0.22745642066001892, "global_step": 72709, "epoch": 816, "lr": 8.347519531866845e-05}
{"train_loss": 0.3028732240200043, "global_step": 72710, "epoch": 816, "lr": 8.34747646831214e-05}
{"train_loss": 0.19026346504688263, "global_step": 72711, "epoch": 816, "lr": 8.347433404307408e-05}
{"train_loss": 0.21810623149523575, "global_step": 72712, "epoch": 816, "lr": 8.347390339852652e-05, "val_loss": 3.685579538345337}
{"train_loss": 0.2344864457845688, "global_step": 72713, "epoch": 817, "lr": 8.347347274947881e-05}
{"train_loss": 0.2109280526638031, "global_step": 72714, "epoch": 817, "lr": 8.347304209593101e-05}
{"train_loss": 0.23577377200126648, "global_step": 72715, "epoch": 817, "lr": 8.347261143788317e-05}
{"train_loss": 0.27471962571144104, "global_step": 72716, "epoch": 817, "lr": 8.347218077533532e-05}
{"train_loss": 0.22138731181621552, "global_step": 72717, "epoch": 817, "lr": 8.347175010828758e-05}
{"train_loss": 0.19850923120975494, "global_step": 72718, "epoch": 817, "lr": 8.347131943673993e-05}
{"train_loss": 0.21913281083106995, "global_step": 72719, "epoch": 817, "lr": 8.34708887606925e-05}
{"train_loss": 0.16015098989009857, "global_step": 72720, "epoch": 817, "lr": 8.34704580801453e-05}
{"train_loss": 0.1481631100177765, "global_step": 72721, "epoch": 817, "lr": 8.347002739509841e-05}
{"train_loss": 0.15542806684970856, "global_step": 72722, "epoch": 817, "lr": 8.346959670555188e-05}
{"train_loss": 0.20216220617294312, "global_step": 72723, "epoch": 817, "lr": 8.346916601150578e-05}
{"train_loss": 0.2005363404750824, "global_step": 72724, "epoch": 817, "lr": 8.346873531296016e-05}
{"train_loss": 0.18724873661994934, "global_step": 72725, "epoch": 817, "lr": 8.346830460991506e-05}
{"train_loss": 0.2580440938472748, "global_step": 72726, "epoch": 817, "lr": 8.346787390237056e-05}
{"train_loss": 0.28811919689178467, "global_step": 72727, "epoch": 817, "lr": 8.346744319032673e-05}
{"train_loss": 0.12586179375648499, "global_step": 72728, "epoch": 817, "lr": 8.34670124737836e-05}
{"train_loss": 0.1501888483762741, "global_step": 72729, "epoch": 817, "lr": 8.346658175274123e-05}
{"train_loss": 0.1937071681022644, "global_step": 72730, "epoch": 817, "lr": 8.346615102719969e-05}
{"train_loss": 0.19528786838054657, "global_step": 72731, "epoch": 817, "lr": 8.346572029715904e-05}
{"train_loss": 0.16522115468978882, "global_step": 72732, "epoch": 817, "lr": 8.346528956261933e-05}
{"train_loss": 0.11487323045730591, "global_step": 72733, "epoch": 817, "lr": 8.346485882358061e-05}
{"train_loss": 0.16730159521102905, "global_step": 72734, "epoch": 817, "lr": 8.346442808004296e-05}
{"train_loss": 0.15643367171287537, "global_step": 72735, "epoch": 817, "lr": 8.34639973320064e-05}
{"train_loss": 0.13899172842502594, "global_step": 72736, "epoch": 817, "lr": 8.346356657947105e-05}
{"train_loss": 0.2679064869880676, "global_step": 72737, "epoch": 817, "lr": 8.34631358224369e-05}
{"train_loss": 0.13978716731071472, "global_step": 72738, "epoch": 817, "lr": 8.346270506090405e-05}
{"train_loss": 0.23536740243434906, "global_step": 72739, "epoch": 817, "lr": 8.346227429487256e-05}
{"train_loss": 0.19340810179710388, "global_step": 72740, "epoch": 817, "lr": 8.346184352434246e-05}
{"train_loss": 0.20328333973884583, "global_step": 72741, "epoch": 817, "lr": 8.346141274931383e-05}
{"train_loss": 0.1781589835882187, "global_step": 72742, "epoch": 817, "lr": 8.346098196978672e-05}
{"train_loss": 0.13861671090126038, "global_step": 72743, "epoch": 817, "lr": 8.34605511857612e-05}
{"train_loss": 0.227134108543396, "global_step": 72744, "epoch": 817, "lr": 8.34601203972373e-05}
{"train_loss": 0.23511506617069244, "global_step": 72745, "epoch": 817, "lr": 8.345968960421511e-05}
{"train_loss": 0.2660384178161621, "global_step": 72746, "epoch": 817, "lr": 8.345925880669465e-05}
{"train_loss": 0.2760624587535858, "global_step": 72747, "epoch": 817, "lr": 8.345882800467601e-05}
{"train_loss": 0.1996401846408844, "global_step": 72748, "epoch": 817, "lr": 8.345839719815925e-05}
{"train_loss": 0.20066112279891968, "global_step": 72749, "epoch": 817, "lr": 8.34579663871444e-05}
{"train_loss": 0.13238832354545593, "global_step": 72750, "epoch": 817, "lr": 8.345753557163154e-05}
{"train_loss": 0.20931224524974823, "global_step": 72751, "epoch": 817, "lr": 8.345710475162073e-05}
{"train_loss": 0.21089808642864227, "global_step": 72752, "epoch": 817, "lr": 8.3456673927112e-05}
{"train_loss": 0.24730341136455536, "global_step": 72753, "epoch": 817, "lr": 8.345624309810544e-05}
{"train_loss": 0.22282321751117706, "global_step": 72754, "epoch": 817, "lr": 8.345581226460111e-05}
{"train_loss": 0.1761912703514099, "global_step": 72755, "epoch": 817, "lr": 8.345538142659904e-05}
{"train_loss": 0.19053688645362854, "global_step": 72756, "epoch": 817, "lr": 8.34549505840993e-05}
{"train_loss": 0.18399113416671753, "global_step": 72757, "epoch": 817, "lr": 8.345451973710195e-05}
{"train_loss": 0.22223231196403503, "global_step": 72758, "epoch": 817, "lr": 8.345408888560706e-05}
{"train_loss": 0.21990682184696198, "global_step": 72759, "epoch": 817, "lr": 8.345365802961467e-05}
{"train_loss": 0.20012180507183075, "global_step": 72760, "epoch": 817, "lr": 8.345322716912482e-05}
{"train_loss": 0.22964055836200714, "global_step": 72761, "epoch": 817, "lr": 8.345279630413762e-05}
{"train_loss": 0.29580026865005493, "global_step": 72762, "epoch": 817, "lr": 8.345236543465309e-05}
{"train_loss": 0.1888747215270996, "global_step": 72763, "epoch": 817, "lr": 8.345193456067131e-05}
{"train_loss": 0.1532595157623291, "global_step": 72764, "epoch": 817, "lr": 8.34515036821923e-05}
{"train_loss": 0.24967452883720398, "global_step": 72765, "epoch": 817, "lr": 8.345107279921616e-05}
{"train_loss": 0.25906720757484436, "global_step": 72766, "epoch": 817, "lr": 8.345064191174293e-05}
{"train_loss": 0.168426051735878, "global_step": 72767, "epoch": 817, "lr": 8.345021101977266e-05}
{"train_loss": 0.20484155416488647, "global_step": 72768, "epoch": 817, "lr": 8.344978012330543e-05}
{"train_loss": 0.1701064109802246, "global_step": 72769, "epoch": 817, "lr": 8.344934922234126e-05}
{"train_loss": 0.22596794366836548, "global_step": 72770, "epoch": 817, "lr": 8.344891831688027e-05}
{"train_loss": 0.20624017715454102, "global_step": 72771, "epoch": 817, "lr": 8.344848740692244e-05}
{"train_loss": 0.17143790423870087, "global_step": 72772, "epoch": 817, "lr": 8.34480564924679e-05}
{"train_loss": 0.24095192551612854, "global_step": 72773, "epoch": 817, "lr": 8.344762557351667e-05}
{"train_loss": 0.2762623131275177, "global_step": 72774, "epoch": 817, "lr": 8.344719465006881e-05}
{"train_loss": 0.21522629261016846, "global_step": 72775, "epoch": 817, "lr": 8.344676372212436e-05}
{"train_loss": 0.2699389159679413, "global_step": 72776, "epoch": 817, "lr": 8.344633278968344e-05}
{"train_loss": 0.31566640734672546, "global_step": 72777, "epoch": 817, "lr": 8.344590185274605e-05}
{"train_loss": 0.22308887541294098, "global_step": 72778, "epoch": 817, "lr": 8.344547091131226e-05}
{"train_loss": 0.25225764513015747, "global_step": 72779, "epoch": 817, "lr": 8.344503996538214e-05}
{"train_loss": 0.25675371289253235, "global_step": 72780, "epoch": 817, "lr": 8.344460901495574e-05}
{"train_loss": 0.16065528988838196, "global_step": 72781, "epoch": 817, "lr": 8.344417806003313e-05}
{"train_loss": 0.17437319457530975, "global_step": 72782, "epoch": 817, "lr": 8.344374710061436e-05}
{"train_loss": 0.24672634899616241, "global_step": 72783, "epoch": 817, "lr": 8.344331613669947e-05}
{"train_loss": 0.2803407907485962, "global_step": 72784, "epoch": 817, "lr": 8.344288516828853e-05}
{"train_loss": 0.19011293351650238, "global_step": 72785, "epoch": 817, "lr": 8.344245419538161e-05}
{"train_loss": 0.24772807955741882, "global_step": 72786, "epoch": 817, "lr": 8.344202321797877e-05}
{"train_loss": 0.20462463796138763, "global_step": 72787, "epoch": 817, "lr": 8.344159223608006e-05}
{"train_loss": 0.2809719443321228, "global_step": 72788, "epoch": 817, "lr": 8.344116124968551e-05}
{"train_loss": 0.1649712771177292, "global_step": 72789, "epoch": 817, "lr": 8.344073025879522e-05}
{"train_loss": 0.2505534291267395, "global_step": 72790, "epoch": 817, "lr": 8.344029926340923e-05}
{"train_loss": 0.2014051228761673, "global_step": 72791, "epoch": 817, "lr": 8.343986826352759e-05}
{"train_loss": 0.2976531982421875, "global_step": 72792, "epoch": 817, "lr": 8.343943725915038e-05}
{"train_loss": 0.19039562344551086, "global_step": 72793, "epoch": 817, "lr": 8.343900625027764e-05}
{"train_loss": 0.203522726893425, "global_step": 72794, "epoch": 817, "lr": 8.343857523690944e-05}
{"train_loss": 0.23375417292118073, "global_step": 72795, "epoch": 817, "lr": 8.343814421904583e-05}
{"train_loss": 0.19395102560520172, "global_step": 72796, "epoch": 817, "lr": 8.343771319668686e-05}
{"train_loss": 0.2689162790775299, "global_step": 72797, "epoch": 817, "lr": 8.343728216983259e-05}
{"train_loss": 0.19695381820201874, "global_step": 72798, "epoch": 817, "lr": 8.34368511384831e-05}
{"train_loss": 0.328971266746521, "global_step": 72799, "epoch": 817, "lr": 8.343642010263843e-05}
{"train_loss": 0.1626308113336563, "global_step": 72800, "epoch": 817, "lr": 8.343598906229865e-05}
{"train_loss": 0.21247387736031179, "global_step": 72801, "epoch": 817, "lr": 8.34355580174638e-05, "val_loss": 3.5878102779388428}
{"train_loss": 0.18015511333942413, "global_step": 72802, "epoch": 818, "lr": 8.343512696813394e-05}
{"train_loss": 0.19916880130767822, "global_step": 72803, "epoch": 818, "lr": 8.343469591430914e-05}
{"train_loss": 0.2356598824262619, "global_step": 72804, "epoch": 818, "lr": 8.343426485598946e-05}
{"train_loss": 0.2648162245750427, "global_step": 72805, "epoch": 818, "lr": 8.343383379317494e-05}
{"train_loss": 0.15612342953681946, "global_step": 72806, "epoch": 818, "lr": 8.343340272586566e-05}
{"train_loss": 0.2027149200439453, "global_step": 72807, "epoch": 818, "lr": 8.343297165406165e-05}
{"train_loss": 0.22810301184654236, "global_step": 72808, "epoch": 818, "lr": 8.3432540577763e-05}
{"train_loss": 0.1761111617088318, "global_step": 72809, "epoch": 818, "lr": 8.343210949696973e-05}
{"train_loss": 0.192683607339859, "global_step": 72810, "epoch": 818, "lr": 8.343167841168193e-05}
{"train_loss": 0.12243691086769104, "global_step": 72811, "epoch": 818, "lr": 8.343124732189966e-05}
{"train_loss": 0.14421002566814423, "global_step": 72812, "epoch": 818, "lr": 8.343081622762297e-05}
{"train_loss": 0.2141290158033371, "global_step": 72813, "epoch": 818, "lr": 8.343038512885191e-05}
{"train_loss": 0.2614848017692566, "global_step": 72814, "epoch": 818, "lr": 8.342995402558653e-05}
{"train_loss": 0.20889832079410553, "global_step": 72815, "epoch": 818, "lr": 8.342952291782691e-05}
{"train_loss": 0.16849014163017273, "global_step": 72816, "epoch": 818, "lr": 8.342909180557309e-05}
{"train_loss": 0.17698299884796143, "global_step": 72817, "epoch": 818, "lr": 8.342866068882514e-05}
{"train_loss": 0.21038517355918884, "global_step": 72818, "epoch": 818, "lr": 8.342822956758311e-05}
{"train_loss": 0.20431503653526306, "global_step": 72819, "epoch": 818, "lr": 8.342779844184708e-05}
{"train_loss": 0.2000536322593689, "global_step": 72820, "epoch": 818, "lr": 8.342736731161708e-05}
{"train_loss": 0.33104684948921204, "global_step": 72821, "epoch": 818, "lr": 8.342693617689317e-05}
{"train_loss": 0.1847814917564392, "global_step": 72822, "epoch": 818, "lr": 8.342650503767541e-05}
{"train_loss": 0.2616705298423767, "global_step": 72823, "epoch": 818, "lr": 8.342607389396389e-05}
{"train_loss": 0.2594853639602661, "global_step": 72824, "epoch": 818, "lr": 8.342564274575863e-05}
{"train_loss": 0.14221630990505219, "global_step": 72825, "epoch": 818, "lr": 8.342521159305969e-05}
{"train_loss": 0.17646081745624542, "global_step": 72826, "epoch": 818, "lr": 8.342478043586714e-05}
{"train_loss": 0.2407664805650711, "global_step": 72827, "epoch": 818, "lr": 8.342434927418104e-05}
{"train_loss": 0.24975387752056122, "global_step": 72828, "epoch": 818, "lr": 8.342391810800144e-05}
{"train_loss": 0.3438374400138855, "global_step": 72829, "epoch": 818, "lr": 8.342348693732841e-05}
{"train_loss": 0.24456541240215302, "global_step": 72830, "epoch": 818, "lr": 8.3423055762162e-05}
{"train_loss": 0.25262653827667236, "global_step": 72831, "epoch": 818, "lr": 8.342262458250226e-05}
{"train_loss": 0.24067676067352295, "global_step": 72832, "epoch": 818, "lr": 8.342219339834924e-05}
{"train_loss": 0.25246143341064453, "global_step": 72833, "epoch": 818, "lr": 8.342176220970304e-05}
{"train_loss": 0.17254775762557983, "global_step": 72834, "epoch": 818, "lr": 8.342133101656369e-05}
{"train_loss": 0.1661057472229004, "global_step": 72835, "epoch": 818, "lr": 8.342089981893124e-05}
{"train_loss": 0.171385258436203, "global_step": 72836, "epoch": 818, "lr": 8.342046861680574e-05}
{"train_loss": 0.1382901519536972, "global_step": 72837, "epoch": 818, "lr": 8.34200374101873e-05}
{"train_loss": 0.16518059372901917, "global_step": 72838, "epoch": 818, "lr": 8.341960619907593e-05}
{"train_loss": 0.22306428849697113, "global_step": 72839, "epoch": 818, "lr": 8.341917498347169e-05}
{"train_loss": 0.2373208999633789, "global_step": 72840, "epoch": 818, "lr": 8.341874376337467e-05}
{"train_loss": 0.2252529412508011, "global_step": 72841, "epoch": 818, "lr": 8.34183125387849e-05}
{"train_loss": 0.2112281620502472, "global_step": 72842, "epoch": 818, "lr": 8.341788130970243e-05}
{"train_loss": 0.23419640958309174, "global_step": 72843, "epoch": 818, "lr": 8.341745007612735e-05}
{"train_loss": 0.179458349943161, "global_step": 72844, "epoch": 818, "lr": 8.34170188380597e-05}
{"train_loss": 0.23601393401622772, "global_step": 72845, "epoch": 818, "lr": 8.341658759549952e-05}
{"train_loss": 0.19264426827430725, "global_step": 72846, "epoch": 818, "lr": 8.34161563484469e-05}
{"train_loss": 0.196372389793396, "global_step": 72847, "epoch": 818, "lr": 8.34157250969019e-05}
{"train_loss": 0.15998229384422302, "global_step": 72848, "epoch": 818, "lr": 8.341529384086454e-05}
{"train_loss": 0.1762957125902176, "global_step": 72849, "epoch": 818, "lr": 8.34148625803349e-05}
{"train_loss": 0.15596923232078552, "global_step": 72850, "epoch": 818, "lr": 8.341443131531305e-05}
{"train_loss": 0.16146603226661682, "global_step": 72851, "epoch": 818, "lr": 8.341400004579904e-05}
{"train_loss": 0.21440938115119934, "global_step": 72852, "epoch": 818, "lr": 8.341356877179292e-05}
{"train_loss": 0.16780388355255127, "global_step": 72853, "epoch": 818, "lr": 8.341313749329475e-05}
{"train_loss": 0.26086172461509705, "global_step": 72854, "epoch": 818, "lr": 8.34127062103046e-05}
{"train_loss": 0.13027748465538025, "global_step": 72855, "epoch": 818, "lr": 8.341227492282252e-05}
{"train_loss": 0.18420667946338654, "global_step": 72856, "epoch": 818, "lr": 8.341184363084855e-05}
{"train_loss": 0.22647802531719208, "global_step": 72857, "epoch": 818, "lr": 8.341141233438279e-05}
{"train_loss": 0.172928586602211, "global_step": 72858, "epoch": 818, "lr": 8.341098103342526e-05}
{"train_loss": 0.18007056415081024, "global_step": 72859, "epoch": 818, "lr": 8.341054972797603e-05}
{"train_loss": 0.19065642356872559, "global_step": 72860, "epoch": 818, "lr": 8.341011841803516e-05}
{"train_loss": 0.1971992701292038, "global_step": 72861, "epoch": 818, "lr": 8.340968710360271e-05}
{"train_loss": 0.12523239850997925, "global_step": 72862, "epoch": 818, "lr": 8.340925578467872e-05}
{"train_loss": 0.11343112587928772, "global_step": 72863, "epoch": 818, "lr": 8.340882446126328e-05}
{"train_loss": 0.2091415375471115, "global_step": 72864, "epoch": 818, "lr": 8.340839313335639e-05}
{"train_loss": 0.09653336554765701, "global_step": 72865, "epoch": 818, "lr": 8.34079618009582e-05}
{"train_loss": 0.22630272805690765, "global_step": 72866, "epoch": 818, "lr": 8.34075304640687e-05}
{"train_loss": 0.20519323647022247, "global_step": 72867, "epoch": 818, "lr": 8.340709912268796e-05}
{"train_loss": 0.20072318613529205, "global_step": 72868, "epoch": 818, "lr": 8.340666777681606e-05}
{"train_loss": 0.1916779726743698, "global_step": 72869, "epoch": 818, "lr": 8.340623642645302e-05}
{"train_loss": 0.20337563753128052, "global_step": 72870, "epoch": 818, "lr": 8.340580507159892e-05}
{"train_loss": 0.24197226762771606, "global_step": 72871, "epoch": 818, "lr": 8.340537371225383e-05}
{"train_loss": 0.19899077713489532, "global_step": 72872, "epoch": 818, "lr": 8.340494234841778e-05}
{"train_loss": 0.19407251477241516, "global_step": 72873, "epoch": 818, "lr": 8.340451098009086e-05}
{"train_loss": 0.27395811676979065, "global_step": 72874, "epoch": 818, "lr": 8.340407960727309e-05}
{"train_loss": 0.2992286682128906, "global_step": 72875, "epoch": 818, "lr": 8.340364822996455e-05}
{"train_loss": 0.24567748606204987, "global_step": 72876, "epoch": 818, "lr": 8.340321684816532e-05}
{"train_loss": 0.19576744735240936, "global_step": 72877, "epoch": 818, "lr": 8.34027854618754e-05}
{"train_loss": 0.26348650455474854, "global_step": 72878, "epoch": 818, "lr": 8.340235407109492e-05}
{"train_loss": 0.1173040121793747, "global_step": 72879, "epoch": 818, "lr": 8.340192267582388e-05}
{"train_loss": 0.19699738919734955, "global_step": 72880, "epoch": 818, "lr": 8.340149127606236e-05}
{"train_loss": 0.11235937476158142, "global_step": 72881, "epoch": 818, "lr": 8.340105987181041e-05}
{"train_loss": 0.2786267399787903, "global_step": 72882, "epoch": 818, "lr": 8.340062846306811e-05}
{"train_loss": 0.1444614827632904, "global_step": 72883, "epoch": 818, "lr": 8.34001970498355e-05}
{"train_loss": 0.2193223536014557, "global_step": 72884, "epoch": 818, "lr": 8.339976563211262e-05}
{"train_loss": 0.16663409769535065, "global_step": 72885, "epoch": 818, "lr": 8.339933420989959e-05}
{"train_loss": 0.18587695062160492, "global_step": 72886, "epoch": 818, "lr": 8.33989027831964e-05}
{"train_loss": 0.20049364864826202, "global_step": 72887, "epoch": 818, "lr": 8.339847135200313e-05}
{"train_loss": 0.15005019307136536, "global_step": 72888, "epoch": 818, "lr": 8.339803991631985e-05}
{"train_loss": 0.22351211309432983, "global_step": 72889, "epoch": 818, "lr": 8.33976084761466e-05}
{"train_loss": 0.20087842115860308, "global_step": 72890, "epoch": 818, "lr": 8.339717703148347e-05, "val_loss": 3.9641268253326416}
{"train_loss": 0.149877667427063, "global_step": 72891, "epoch": 819, "lr": 8.339674558233048e-05}
{"train_loss": 0.19968006014823914, "global_step": 72892, "epoch": 819, "lr": 8.339631412868771e-05}
{"train_loss": 0.21624988317489624, "global_step": 72893, "epoch": 819, "lr": 8.33958826705552e-05}
{"train_loss": 0.2605826258659363, "global_step": 72894, "epoch": 819, "lr": 8.339545120793303e-05}
{"train_loss": 0.21281571686267853, "global_step": 72895, "epoch": 819, "lr": 8.339501974082127e-05}
{"train_loss": 0.14929667115211487, "global_step": 72896, "epoch": 819, "lr": 8.339458826921993e-05}
{"train_loss": 0.24432027339935303, "global_step": 72897, "epoch": 819, "lr": 8.33941567931291e-05}
{"train_loss": 0.18383103609085083, "global_step": 72898, "epoch": 819, "lr": 8.339372531254884e-05}
{"train_loss": 0.2578306198120117, "global_step": 72899, "epoch": 819, "lr": 8.339329382747919e-05}
{"train_loss": 0.21649067103862762, "global_step": 72900, "epoch": 819, "lr": 8.339286233792022e-05}
{"train_loss": 0.15945182740688324, "global_step": 72901, "epoch": 819, "lr": 8.339243084387198e-05}
{"train_loss": 0.20566049218177795, "global_step": 72902, "epoch": 819, "lr": 8.339199934533455e-05}
{"train_loss": 0.24210470914840698, "global_step": 72903, "epoch": 819, "lr": 8.339156784230797e-05}
{"train_loss": 0.1914098709821701, "global_step": 72904, "epoch": 819, "lr": 8.33911363347923e-05}
{"train_loss": 0.21582382917404175, "global_step": 72905, "epoch": 819, "lr": 8.33907048227876e-05}
{"train_loss": 0.14940446615219116, "global_step": 72906, "epoch": 819, "lr": 8.339027330629392e-05}
{"train_loss": 0.18000705540180206, "global_step": 72907, "epoch": 819, "lr": 8.338984178531133e-05}
{"train_loss": 0.1529184728860855, "global_step": 72908, "epoch": 819, "lr": 8.338941025983988e-05}
{"train_loss": 0.16437509655952454, "global_step": 72909, "epoch": 819, "lr": 8.338897872987963e-05}
{"train_loss": 0.23764680325984955, "global_step": 72910, "epoch": 819, "lr": 8.338854719543066e-05}
{"train_loss": 0.21302282810211182, "global_step": 72911, "epoch": 819, "lr": 8.338811565649298e-05}
{"train_loss": 0.2635081708431244, "global_step": 72912, "epoch": 819, "lr": 8.338768411306667e-05}
{"train_loss": 0.20889101922512054, "global_step": 72913, "epoch": 819, "lr": 8.338725256515182e-05}
{"train_loss": 0.19001556932926178, "global_step": 72914, "epoch": 819, "lr": 8.338682101274845e-05}
{"train_loss": 0.24782805144786835, "global_step": 72915, "epoch": 819, "lr": 8.338638945585662e-05}
{"train_loss": 0.2857811152935028, "global_step": 72916, "epoch": 819, "lr": 8.33859578944764e-05}
{"train_loss": 0.17578724026679993, "global_step": 72917, "epoch": 819, "lr": 8.338552632860786e-05}
{"train_loss": 0.1260623335838318, "global_step": 72918, "epoch": 819, "lr": 8.338509475825101e-05}
{"train_loss": 0.2423158437013626, "global_step": 72919, "epoch": 819, "lr": 8.338466318340596e-05}
{"train_loss": 0.2660475969314575, "global_step": 72920, "epoch": 819, "lr": 8.338423160407277e-05}
{"train_loss": 0.17891155183315277, "global_step": 72921, "epoch": 819, "lr": 8.338380002025143e-05}
{"train_loss": 0.21205095946788788, "global_step": 72922, "epoch": 819, "lr": 8.338336843194207e-05}
{"train_loss": 0.16236749291419983, "global_step": 72923, "epoch": 819, "lr": 8.338293683914474e-05}
{"train_loss": 0.19058164954185486, "global_step": 72924, "epoch": 819, "lr": 8.338250524185947e-05}
{"train_loss": 0.18193037807941437, "global_step": 72925, "epoch": 819, "lr": 8.338207364008632e-05}
{"train_loss": 0.1988651305437088, "global_step": 72926, "epoch": 819, "lr": 8.338164203382536e-05}
{"train_loss": 0.20215821266174316, "global_step": 72927, "epoch": 819, "lr": 8.338121042307665e-05}
{"train_loss": 0.21374887228012085, "global_step": 72928, "epoch": 819, "lr": 8.338077880784023e-05}
{"train_loss": 0.18610848486423492, "global_step": 72929, "epoch": 819, "lr": 8.338034718811618e-05}
{"train_loss": 0.18670745193958282, "global_step": 72930, "epoch": 819, "lr": 8.337991556390456e-05}
{"train_loss": 0.20716360211372375, "global_step": 72931, "epoch": 819, "lr": 8.33794839352054e-05}
{"train_loss": 0.25587648153305054, "global_step": 72932, "epoch": 819, "lr": 8.337905230201878e-05}
{"train_loss": 0.2660866975784302, "global_step": 72933, "epoch": 819, "lr": 8.337862066434476e-05}
{"train_loss": 0.22668828070163727, "global_step": 72934, "epoch": 819, "lr": 8.337818902218339e-05}
{"train_loss": 0.2669023275375366, "global_step": 72935, "epoch": 819, "lr": 8.337775737553472e-05}
{"train_loss": 0.21861371397972107, "global_step": 72936, "epoch": 819, "lr": 8.337732572439883e-05}
{"train_loss": 0.2864765226840973, "global_step": 72937, "epoch": 819, "lr": 8.337689406877576e-05}
{"train_loss": 0.20739607512950897, "global_step": 72938, "epoch": 819, "lr": 8.337646240866557e-05}
{"train_loss": 0.29017844796180725, "global_step": 72939, "epoch": 819, "lr": 8.337603074406832e-05}
{"train_loss": 0.28188833594322205, "global_step": 72940, "epoch": 819, "lr": 8.337559907498409e-05}
{"train_loss": 0.2007974535226822, "global_step": 72941, "epoch": 819, "lr": 8.337516740141289e-05}
{"train_loss": 0.19589564204216003, "global_step": 72942, "epoch": 819, "lr": 8.337473572335482e-05}
{"train_loss": 0.30943232774734497, "global_step": 72943, "epoch": 819, "lr": 8.337430404080991e-05}
{"train_loss": 0.23886531591415405, "global_step": 72944, "epoch": 819, "lr": 8.337387235377825e-05}
{"train_loss": 0.20917902886867523, "global_step": 72945, "epoch": 819, "lr": 8.337344066225987e-05}
{"train_loss": 0.2166706919670105, "global_step": 72946, "epoch": 819, "lr": 8.337300896625484e-05}
{"train_loss": 0.2153773456811905, "global_step": 72947, "epoch": 819, "lr": 8.337257726576321e-05}
{"train_loss": 0.26430103182792664, "global_step": 72948, "epoch": 819, "lr": 8.337214556078505e-05}
{"train_loss": 0.22930370271205902, "global_step": 72949, "epoch": 819, "lr": 8.33717138513204e-05}
{"train_loss": 0.24637627601623535, "global_step": 72950, "epoch": 819, "lr": 8.337128213736936e-05}
{"train_loss": 0.21364836394786835, "global_step": 72951, "epoch": 819, "lr": 8.337085041893194e-05}
{"train_loss": 0.21776613593101501, "global_step": 72952, "epoch": 819, "lr": 8.337041869600821e-05}
{"train_loss": 0.2683199942111969, "global_step": 72953, "epoch": 819, "lr": 8.336998696859823e-05}
{"train_loss": 0.13605725765228271, "global_step": 72954, "epoch": 819, "lr": 8.336955523670207e-05}
{"train_loss": 0.19931228458881378, "global_step": 72955, "epoch": 819, "lr": 8.336912350031977e-05}
{"train_loss": 0.21040141582489014, "global_step": 72956, "epoch": 819, "lr": 8.336869175945142e-05}
{"train_loss": 0.14452753961086273, "global_step": 72957, "epoch": 819, "lr": 8.336826001409704e-05}
{"train_loss": 0.2516181468963623, "global_step": 72958, "epoch": 819, "lr": 8.33678282642567e-05}
{"train_loss": 0.17274604737758636, "global_step": 72959, "epoch": 819, "lr": 8.336739650993048e-05}
{"train_loss": 0.21558645367622375, "global_step": 72960, "epoch": 819, "lr": 8.33669647511184e-05}
{"train_loss": 0.18009375035762787, "global_step": 72961, "epoch": 819, "lr": 8.336653298782056e-05}
{"train_loss": 0.21180111169815063, "global_step": 72962, "epoch": 819, "lr": 8.336610122003699e-05}
{"train_loss": 0.17912904918193817, "global_step": 72963, "epoch": 819, "lr": 8.336566944776774e-05}
{"train_loss": 0.17152173817157745, "global_step": 72964, "epoch": 819, "lr": 8.33652376710129e-05}
{"train_loss": 0.29565805196762085, "global_step": 72965, "epoch": 819, "lr": 8.33648058897725e-05}
{"train_loss": 0.1772584319114685, "global_step": 72966, "epoch": 819, "lr": 8.33643741040466e-05}
{"train_loss": 0.20000341534614563, "global_step": 72967, "epoch": 819, "lr": 8.336394231383529e-05}
{"train_loss": 0.2440205216407776, "global_step": 72968, "epoch": 819, "lr": 8.336351051913858e-05}
{"train_loss": 0.1431845724582672, "global_step": 72969, "epoch": 819, "lr": 8.336307871995657e-05}
{"train_loss": 0.27296456694602966, "global_step": 72970, "epoch": 819, "lr": 8.33626469162893e-05}
{"train_loss": 0.20580846071243286, "global_step": 72971, "epoch": 819, "lr": 8.336221510813682e-05}
{"train_loss": 0.23736749589443207, "global_step": 72972, "epoch": 819, "lr": 8.33617832954992e-05}
{"train_loss": 0.204668790102005, "global_step": 72973, "epoch": 819, "lr": 8.336135147837649e-05}
{"train_loss": 0.33364593982696533, "global_step": 72974, "epoch": 819, "lr": 8.336091965676876e-05}
{"train_loss": 0.18425528705120087, "global_step": 72975, "epoch": 819, "lr": 8.336048783067607e-05}
{"train_loss": 0.22508102655410767, "global_step": 72976, "epoch": 819, "lr": 8.336005600009845e-05}
{"train_loss": 0.20983220636844635, "global_step": 72977, "epoch": 819, "lr": 8.3359624165036e-05}
{"train_loss": 0.26512008905410767, "global_step": 72978, "epoch": 819, "lr": 8.335919232548872e-05}
{"train_loss": 0.21590038380596074, "global_step": 72979, "epoch": 819, "lr": 8.335876048145674e-05, "val_loss": 3.686573028564453}
{"train_loss": 0.2396525740623474, "global_step": 72980, "epoch": 820, "lr": 8.335832863294006e-05}
{"train_loss": 0.15822792053222656, "global_step": 72981, "epoch": 820, "lr": 8.335789677993876e-05}
{"train_loss": 0.1843036413192749, "global_step": 72982, "epoch": 820, "lr": 8.335746492245291e-05}
{"train_loss": 0.31362104415893555, "global_step": 72983, "epoch": 820, "lr": 8.335703306048254e-05}
{"train_loss": 0.2697073519229889, "global_step": 72984, "epoch": 820, "lr": 8.335660119402773e-05}
{"train_loss": 0.21170084178447723, "global_step": 72985, "epoch": 820, "lr": 8.335616932308854e-05}
{"train_loss": 0.3014328181743622, "global_step": 72986, "epoch": 820, "lr": 8.3355737447665e-05}
{"train_loss": 0.15299397706985474, "global_step": 72987, "epoch": 820, "lr": 8.335530556775722e-05}
{"train_loss": 0.17752644419670105, "global_step": 72988, "epoch": 820, "lr": 8.335487368336519e-05}
{"train_loss": 0.23963704705238342, "global_step": 72989, "epoch": 820, "lr": 8.335444179448903e-05}
{"train_loss": 0.25259894132614136, "global_step": 72990, "epoch": 820, "lr": 8.335400990112876e-05}
{"train_loss": 0.15172912180423737, "global_step": 72991, "epoch": 820, "lr": 8.335357800328444e-05}
{"train_loss": 0.23443278670310974, "global_step": 72992, "epoch": 820, "lr": 8.335314610095615e-05}
{"train_loss": 0.21130219101905823, "global_step": 72993, "epoch": 820, "lr": 8.335271419414393e-05}
{"train_loss": 0.26040536165237427, "global_step": 72994, "epoch": 820, "lr": 8.335228228284786e-05}
{"train_loss": 0.20199042558670044, "global_step": 72995, "epoch": 820, "lr": 8.335185036706796e-05}
{"train_loss": 0.21560560166835785, "global_step": 72996, "epoch": 820, "lr": 8.335141844680432e-05}
{"train_loss": 0.2844744026660919, "global_step": 72997, "epoch": 820, "lr": 8.335098652205699e-05}
{"train_loss": 0.24263238906860352, "global_step": 72998, "epoch": 820, "lr": 8.335055459282603e-05}
{"train_loss": 0.1792508065700531, "global_step": 72999, "epoch": 820, "lr": 8.335012265911149e-05}
{"train_loss": 0.24182374775409698, "global_step": 73000, "epoch": 820, "lr": 8.334969072091343e-05}
{"train_loss": 0.18038183450698853, "global_step": 73001, "epoch": 820, "lr": 8.334925877823192e-05}
{"train_loss": 0.1927824765443802, "global_step": 73002, "epoch": 820, "lr": 8.3348826831067e-05}
{"train_loss": 0.186073899269104, "global_step": 73003, "epoch": 820, "lr": 8.334839487941874e-05}
{"train_loss": 0.18077266216278076, "global_step": 73004, "epoch": 820, "lr": 8.33479629232872e-05}
{"train_loss": 0.23738466203212738, "global_step": 73005, "epoch": 820, "lr": 8.334753096267243e-05}
{"train_loss": 0.20997551083564758, "global_step": 73006, "epoch": 820, "lr": 8.334709899757447e-05}
{"train_loss": 0.27260783314704895, "global_step": 73007, "epoch": 820, "lr": 8.334666702799343e-05}
{"train_loss": 0.14297884702682495, "global_step": 73008, "epoch": 820, "lr": 8.334623505392931e-05}
{"train_loss": 0.25094693899154663, "global_step": 73009, "epoch": 820, "lr": 8.334580307538222e-05}
{"train_loss": 0.18829554319381714, "global_step": 73010, "epoch": 820, "lr": 8.334537109235217e-05}
{"train_loss": 0.19364884495735168, "global_step": 73011, "epoch": 820, "lr": 8.334493910483924e-05}
{"train_loss": 0.17337049543857574, "global_step": 73012, "epoch": 820, "lr": 8.334450711284352e-05}
{"train_loss": 0.13998368382453918, "global_step": 73013, "epoch": 820, "lr": 8.334407511636501e-05}
{"train_loss": 0.15110233426094055, "global_step": 73014, "epoch": 820, "lr": 8.334364311540382e-05}
{"train_loss": 0.22090765833854675, "global_step": 73015, "epoch": 820, "lr": 8.334321110995996e-05}
{"train_loss": 0.20767267048358917, "global_step": 73016, "epoch": 820, "lr": 8.334277910003351e-05}
{"train_loss": 0.2343566119670868, "global_step": 73017, "epoch": 820, "lr": 8.334234708562455e-05}
{"train_loss": 0.28859564661979675, "global_step": 73018, "epoch": 820, "lr": 8.33419150667331e-05}
{"train_loss": 0.1898382306098938, "global_step": 73019, "epoch": 820, "lr": 8.334148304335925e-05}
{"train_loss": 0.17237411439418793, "global_step": 73020, "epoch": 820, "lr": 8.334105101550305e-05}
{"train_loss": 0.18785549700260162, "global_step": 73021, "epoch": 820, "lr": 8.334061898316454e-05}
{"train_loss": 0.1950455754995346, "global_step": 73022, "epoch": 820, "lr": 8.334018694634378e-05}
{"train_loss": 0.14044664800167084, "global_step": 73023, "epoch": 820, "lr": 8.333975490504085e-05}
{"train_loss": 0.21961353719234467, "global_step": 73024, "epoch": 820, "lr": 8.333932285925579e-05}
{"train_loss": 0.28471240401268005, "global_step": 73025, "epoch": 820, "lr": 8.333889080898867e-05}
{"train_loss": 0.2520810067653656, "global_step": 73026, "epoch": 820, "lr": 8.333845875423955e-05}
{"train_loss": 0.28735482692718506, "global_step": 73027, "epoch": 820, "lr": 8.333802669500845e-05}
{"train_loss": 0.16244426369667053, "global_step": 73028, "epoch": 820, "lr": 8.333759463129549e-05}
{"train_loss": 0.21225064992904663, "global_step": 73029, "epoch": 820, "lr": 8.33371625631007e-05}
{"train_loss": 0.19136501848697662, "global_step": 73030, "epoch": 820, "lr": 8.333673049042412e-05}
{"train_loss": 0.14629201591014862, "global_step": 73031, "epoch": 820, "lr": 8.333629841326582e-05}
{"train_loss": 0.15830327570438385, "global_step": 73032, "epoch": 820, "lr": 8.333586633162586e-05}
{"train_loss": 0.15647710859775543, "global_step": 73033, "epoch": 820, "lr": 8.333543424550432e-05}
{"train_loss": 0.2076907455921173, "global_step": 73034, "epoch": 820, "lr": 8.333500215490122e-05}
{"train_loss": 0.2230040282011032, "global_step": 73035, "epoch": 820, "lr": 8.333457005981662e-05}
{"train_loss": 0.20206932723522186, "global_step": 73036, "epoch": 820, "lr": 8.333413796025063e-05}
{"train_loss": 0.20072513818740845, "global_step": 73037, "epoch": 820, "lr": 8.333370585620324e-05}
{"train_loss": 0.25519707798957825, "global_step": 73038, "epoch": 820, "lr": 8.333327374767456e-05}
{"train_loss": 0.24575434625148773, "global_step": 73039, "epoch": 820, "lr": 8.333284163466461e-05}
{"train_loss": 0.16182422637939453, "global_step": 73040, "epoch": 820, "lr": 8.333240951717348e-05}
{"train_loss": 0.25191691517829895, "global_step": 73041, "epoch": 820, "lr": 8.333197739520121e-05}
{"train_loss": 0.19531427323818207, "global_step": 73042, "epoch": 820, "lr": 8.333154526874787e-05}
{"train_loss": 0.2574770152568817, "global_step": 73043, "epoch": 820, "lr": 8.33311131378135e-05}
{"train_loss": 0.23461738228797913, "global_step": 73044, "epoch": 820, "lr": 8.333068100239815e-05}
{"train_loss": 0.24885623157024384, "global_step": 73045, "epoch": 820, "lr": 8.333024886250191e-05}
{"train_loss": 0.26679614186286926, "global_step": 73046, "epoch": 820, "lr": 8.332981671812484e-05}
{"train_loss": 0.1815192699432373, "global_step": 73047, "epoch": 820, "lr": 8.332938456926698e-05}
{"train_loss": 0.18253841996192932, "global_step": 73048, "epoch": 820, "lr": 8.332895241592838e-05}
{"train_loss": 0.18541255593299866, "global_step": 73049, "epoch": 820, "lr": 8.332852025810911e-05}
{"train_loss": 0.17271050810813904, "global_step": 73050, "epoch": 820, "lr": 8.332808809580922e-05}
{"train_loss": 0.1941409856081009, "global_step": 73051, "epoch": 820, "lr": 8.332765592902878e-05}
{"train_loss": 0.23343981802463531, "global_step": 73052, "epoch": 820, "lr": 8.332722375776785e-05}
{"train_loss": 0.21103334426879883, "global_step": 73053, "epoch": 820, "lr": 8.332679158202647e-05}
{"train_loss": 0.1911553293466568, "global_step": 73054, "epoch": 820, "lr": 8.332635940180473e-05}
{"train_loss": 0.2281896471977234, "global_step": 73055, "epoch": 820, "lr": 8.332592721710264e-05}
{"train_loss": 0.195944681763649, "global_step": 73056, "epoch": 820, "lr": 8.332549502792031e-05}
{"train_loss": 0.18735244870185852, "global_step": 73057, "epoch": 820, "lr": 8.332506283425775e-05}
{"train_loss": 0.21475568413734436, "global_step": 73058, "epoch": 820, "lr": 8.332463063611507e-05}
{"train_loss": 0.21528248488903046, "global_step": 73059, "epoch": 820, "lr": 8.332419843349226e-05}
{"train_loss": 0.2238151878118515, "global_step": 73060, "epoch": 820, "lr": 8.332376622638945e-05}
{"train_loss": 0.16234643757343292, "global_step": 73061, "epoch": 820, "lr": 8.332333401480666e-05}
{"train_loss": 0.3097502291202545, "global_step": 73062, "epoch": 820, "lr": 8.332290179874394e-05}
{"train_loss": 0.17843399941921234, "global_step": 73063, "epoch": 820, "lr": 8.332246957820138e-05}
{"train_loss": 0.15376433730125427, "global_step": 73064, "epoch": 820, "lr": 8.332203735317901e-05}
{"train_loss": 0.19812187552452087, "global_step": 73065, "epoch": 820, "lr": 8.33216051236769e-05}
{"train_loss": 0.16468824446201324, "global_step": 73066, "epoch": 820, "lr": 8.33211728896951e-05}
{"train_loss": 0.28555023670196533, "global_step": 73067, "epoch": 820, "lr": 8.332074065123368e-05}
{"train_loss": 0.2110552935118086, "global_step": 73068, "epoch": 820, "lr": 8.33203084082927e-05, "val_loss": 3.6615121364593506, "train_action_mse_error": 28.857765197753906}
{"train_loss": 0.19900618493556976, "global_step": 73069, "epoch": 821, "lr": 8.33198761608722e-05}
{"train_loss": 0.17323973774909973, "global_step": 73070, "epoch": 821, "lr": 8.331944390897226e-05}
{"train_loss": 0.18887268006801605, "global_step": 73071, "epoch": 821, "lr": 8.33190116525929e-05}
{"train_loss": 0.16319414973258972, "global_step": 73072, "epoch": 821, "lr": 8.331857939173425e-05}
{"train_loss": 0.13176368176937103, "global_step": 73073, "epoch": 821, "lr": 8.331814712639629e-05}
{"train_loss": 0.2078060358762741, "global_step": 73074, "epoch": 821, "lr": 8.331771485657911e-05}
{"train_loss": 0.1616506427526474, "global_step": 73075, "epoch": 821, "lr": 8.331728258228277e-05}
{"train_loss": 0.3017338812351227, "global_step": 73076, "epoch": 821, "lr": 8.331685030350735e-05}
{"train_loss": 0.1983279287815094, "global_step": 73077, "epoch": 821, "lr": 8.331641802025288e-05}
{"train_loss": 0.1511961966753006, "global_step": 73078, "epoch": 821, "lr": 8.33159857325194e-05}
{"train_loss": 0.186368927359581, "global_step": 73079, "epoch": 821, "lr": 8.331555344030701e-05}
{"train_loss": 0.23006874322891235, "global_step": 73080, "epoch": 821, "lr": 8.331512114361575e-05}
{"train_loss": 0.23465444147586823, "global_step": 73081, "epoch": 821, "lr": 8.331468884244567e-05}
{"train_loss": 0.21194209158420563, "global_step": 73082, "epoch": 821, "lr": 8.331425653679683e-05}
{"train_loss": 0.1726282685995102, "global_step": 73083, "epoch": 821, "lr": 8.331382422666932e-05}
{"train_loss": 0.15998055040836334, "global_step": 73084, "epoch": 821, "lr": 8.331339191206315e-05}
{"train_loss": 0.16944947838783264, "global_step": 73085, "epoch": 821, "lr": 8.33129595929784e-05}
{"train_loss": 0.23972643911838531, "global_step": 73086, "epoch": 821, "lr": 8.331252726941515e-05}
{"train_loss": 0.202736958861351, "global_step": 73087, "epoch": 821, "lr": 8.331209494137343e-05}
{"train_loss": 0.19503790140151978, "global_step": 73088, "epoch": 821, "lr": 8.331166260885329e-05}
{"train_loss": 0.15523512661457062, "global_step": 73089, "epoch": 821, "lr": 8.331123027185481e-05}
{"train_loss": 0.16910718381404877, "global_step": 73090, "epoch": 821, "lr": 8.331079793037804e-05}
{"train_loss": 0.2114659696817398, "global_step": 73091, "epoch": 821, "lr": 8.331036558442305e-05}
{"train_loss": 0.1591150313615799, "global_step": 73092, "epoch": 821, "lr": 8.330993323398987e-05}
{"train_loss": 0.1932504028081894, "global_step": 73093, "epoch": 821, "lr": 8.330950087907858e-05}
{"train_loss": 0.19063007831573486, "global_step": 73094, "epoch": 821, "lr": 8.330906851968924e-05}
{"train_loss": 0.23397895693778992, "global_step": 73095, "epoch": 821, "lr": 8.330863615582189e-05}
{"train_loss": 0.21635392308235168, "global_step": 73096, "epoch": 821, "lr": 8.330820378747661e-05}
{"train_loss": 0.15511703491210938, "global_step": 73097, "epoch": 821, "lr": 8.330777141465345e-05}
{"train_loss": 0.19018614292144775, "global_step": 73098, "epoch": 821, "lr": 8.330733903735247e-05}
{"train_loss": 0.15946432948112488, "global_step": 73099, "epoch": 821, "lr": 8.33069066555737e-05}
{"train_loss": 0.1906542032957077, "global_step": 73100, "epoch": 821, "lr": 8.330647426931725e-05}
{"train_loss": 0.2521263360977173, "global_step": 73101, "epoch": 821, "lr": 8.330604187858314e-05}
{"train_loss": 0.2019723355770111, "global_step": 73102, "epoch": 821, "lr": 8.330560948337144e-05}
{"train_loss": 0.21904203295707703, "global_step": 73103, "epoch": 821, "lr": 8.33051770836822e-05}
{"train_loss": 0.20002835988998413, "global_step": 73104, "epoch": 821, "lr": 8.33047446795155e-05}
{"train_loss": 0.19112245738506317, "global_step": 73105, "epoch": 821, "lr": 8.330431227087137e-05}
{"train_loss": 0.15159229934215546, "global_step": 73106, "epoch": 821, "lr": 8.330387985774988e-05}
{"train_loss": 0.25747251510620117, "global_step": 73107, "epoch": 821, "lr": 8.33034474401511e-05}
{"train_loss": 0.26591071486473083, "global_step": 73108, "epoch": 821, "lr": 8.330301501807507e-05}
{"train_loss": 0.21561209857463837, "global_step": 73109, "epoch": 821, "lr": 8.330258259152186e-05}
{"train_loss": 0.17727798223495483, "global_step": 73110, "epoch": 821, "lr": 8.330215016049153e-05}
{"train_loss": 0.16647423803806305, "global_step": 73111, "epoch": 821, "lr": 8.330171772498411e-05}
{"train_loss": 0.19051383435726166, "global_step": 73112, "epoch": 821, "lr": 8.330128528499969e-05}
{"train_loss": 0.22776839137077332, "global_step": 73113, "epoch": 821, "lr": 8.330085284053832e-05}
{"train_loss": 0.2235277146100998, "global_step": 73114, "epoch": 821, "lr": 8.330042039160006e-05}
{"train_loss": 0.1894131302833557, "global_step": 73115, "epoch": 821, "lr": 8.329998793818497e-05}
{"train_loss": 0.18117080628871918, "global_step": 73116, "epoch": 821, "lr": 8.329955548029309e-05}
{"train_loss": 0.2445470094680786, "global_step": 73117, "epoch": 821, "lr": 8.329912301792449e-05}
{"train_loss": 0.14865286648273468, "global_step": 73118, "epoch": 821, "lr": 8.329869055107921e-05}
{"train_loss": 0.15930293500423431, "global_step": 73119, "epoch": 821, "lr": 8.329825807975737e-05}
{"train_loss": 0.2682313919067383, "global_step": 73120, "epoch": 821, "lr": 8.329782560395896e-05}
{"train_loss": 0.20066237449645996, "global_step": 73121, "epoch": 821, "lr": 8.329739312368406e-05}
{"train_loss": 0.13549171388149261, "global_step": 73122, "epoch": 821, "lr": 8.329696063893275e-05}
{"train_loss": 0.20374970138072968, "global_step": 73123, "epoch": 821, "lr": 8.329652814970505e-05}
{"train_loss": 0.11654739081859589, "global_step": 73124, "epoch": 821, "lr": 8.329609565600105e-05}
{"train_loss": 0.1663254201412201, "global_step": 73125, "epoch": 821, "lr": 8.32956631578208e-05}
{"train_loss": 0.21671225130558014, "global_step": 73126, "epoch": 821, "lr": 8.329523065516434e-05}
{"train_loss": 0.16846205294132233, "global_step": 73127, "epoch": 821, "lr": 8.329479814803175e-05}
{"train_loss": 0.2761470675468445, "global_step": 73128, "epoch": 821, "lr": 8.329436563642308e-05}
{"train_loss": 0.18364238739013672, "global_step": 73129, "epoch": 821, "lr": 8.329393312033838e-05}
{"train_loss": 0.17791226506233215, "global_step": 73130, "epoch": 821, "lr": 8.329350059977773e-05}
{"train_loss": 0.11912135779857635, "global_step": 73131, "epoch": 821, "lr": 8.329306807474116e-05}
{"train_loss": 0.2031075805425644, "global_step": 73132, "epoch": 821, "lr": 8.329263554522876e-05}
{"train_loss": 0.23857781291007996, "global_step": 73133, "epoch": 821, "lr": 8.329220301124056e-05}
{"train_loss": 0.20868909358978271, "global_step": 73134, "epoch": 821, "lr": 8.329177047277663e-05}
{"train_loss": 0.1487596333026886, "global_step": 73135, "epoch": 821, "lr": 8.329133792983703e-05}
{"train_loss": 0.22070135176181793, "global_step": 73136, "epoch": 821, "lr": 8.329090538242181e-05}
{"train_loss": 0.20381838083267212, "global_step": 73137, "epoch": 821, "lr": 8.329047283053103e-05}
{"train_loss": 0.15795089304447174, "global_step": 73138, "epoch": 821, "lr": 8.329004027416476e-05}
{"train_loss": 0.21998462080955505, "global_step": 73139, "epoch": 821, "lr": 8.328960771332305e-05}
{"train_loss": 0.2438863217830658, "global_step": 73140, "epoch": 821, "lr": 8.328917514800596e-05}
{"train_loss": 0.1814855933189392, "global_step": 73141, "epoch": 821, "lr": 8.328874257821353e-05}
{"train_loss": 0.20379520952701569, "global_step": 73142, "epoch": 821, "lr": 8.328831000394586e-05}
{"train_loss": 0.21071581542491913, "global_step": 73143, "epoch": 821, "lr": 8.328787742520296e-05}
{"train_loss": 0.17768605053424835, "global_step": 73144, "epoch": 821, "lr": 8.328744484198492e-05}
{"train_loss": 0.14503975212574005, "global_step": 73145, "epoch": 821, "lr": 8.32870122542918e-05}
{"train_loss": 0.20890700817108154, "global_step": 73146, "epoch": 821, "lr": 8.328657966212363e-05}
{"train_loss": 0.29557281732559204, "global_step": 73147, "epoch": 821, "lr": 8.32861470654805e-05}
{"train_loss": 0.25604939460754395, "global_step": 73148, "epoch": 821, "lr": 8.328571446436244e-05}
{"train_loss": 0.1848834604024887, "global_step": 73149, "epoch": 821, "lr": 8.328528185876952e-05}
{"train_loss": 0.20521067082881927, "global_step": 73150, "epoch": 821, "lr": 8.32848492487018e-05}
{"train_loss": 0.21547509729862213, "global_step": 73151, "epoch": 821, "lr": 8.328441663415934e-05}
{"train_loss": 0.1842014193534851, "global_step": 73152, "epoch": 821, "lr": 8.32839840151422e-05}
{"train_loss": 0.13902486860752106, "global_step": 73153, "epoch": 821, "lr": 8.328355139165043e-05}
{"train_loss": 0.1710931360721588, "global_step": 73154, "epoch": 821, "lr": 8.328311876368409e-05}
{"train_loss": 0.15353645384311676, "global_step": 73155, "epoch": 821, "lr": 8.328268613124325e-05}
{"train_loss": 0.13847887516021729, "global_step": 73156, "epoch": 821, "lr": 8.328225349432794e-05}
{"train_loss": 0.1947971349686719, "global_step": 73157, "epoch": 821, "lr": 8.328182085293827e-05, "val_loss": 4.054081439971924}
{"train_loss": 0.2200857549905777, "global_step": 73158, "epoch": 822, "lr": 8.328138820707422e-05}
{"train_loss": 0.1533224880695343, "global_step": 73159, "epoch": 822, "lr": 8.328095555673594e-05}
{"train_loss": 0.13551433384418488, "global_step": 73160, "epoch": 822, "lr": 8.328052290192341e-05}
{"train_loss": 0.17865754663944244, "global_step": 73161, "epoch": 822, "lr": 8.328009024263673e-05}
{"train_loss": 0.18748416006565094, "global_step": 73162, "epoch": 822, "lr": 8.327965757887594e-05}
{"train_loss": 0.14158684015274048, "global_step": 73163, "epoch": 822, "lr": 8.327922491064112e-05}
{"train_loss": 0.284688800573349, "global_step": 73164, "epoch": 822, "lr": 8.327879223793232e-05}
{"train_loss": 0.2310255467891693, "global_step": 73165, "epoch": 822, "lr": 8.327835956074957e-05}
{"train_loss": 0.235811248421669, "global_step": 73166, "epoch": 822, "lr": 8.327792687909295e-05}
{"train_loss": 0.19124676287174225, "global_step": 73167, "epoch": 822, "lr": 8.327749419296252e-05}
{"train_loss": 0.17064891755580902, "global_step": 73168, "epoch": 822, "lr": 8.327706150235835e-05}
{"train_loss": 0.16715826094150543, "global_step": 73169, "epoch": 822, "lr": 8.327662880728048e-05}
{"train_loss": 0.23622174561023712, "global_step": 73170, "epoch": 822, "lr": 8.327619610772897e-05}
{"train_loss": 0.1565573662519455, "global_step": 73171, "epoch": 822, "lr": 8.327576340370388e-05}
{"train_loss": 0.17382679879665375, "global_step": 73172, "epoch": 822, "lr": 8.327533069520526e-05}
{"train_loss": 0.22860808670520782, "global_step": 73173, "epoch": 822, "lr": 8.327489798223318e-05}
{"train_loss": 0.19723322987556458, "global_step": 73174, "epoch": 822, "lr": 8.32744652647877e-05}
{"train_loss": 0.22760829329490662, "global_step": 73175, "epoch": 822, "lr": 8.327403254286889e-05}
{"train_loss": 0.16560977697372437, "global_step": 73176, "epoch": 822, "lr": 8.327359981647677e-05}
{"train_loss": 0.15510840713977814, "global_step": 73177, "epoch": 822, "lr": 8.327316708561143e-05}
{"train_loss": 0.21407507359981537, "global_step": 73178, "epoch": 822, "lr": 8.327273435027292e-05}
{"train_loss": 0.23494373261928558, "global_step": 73179, "epoch": 822, "lr": 8.32723016104613e-05}
{"train_loss": 0.20037518441677094, "global_step": 73180, "epoch": 822, "lr": 8.327186886617662e-05}
{"train_loss": 0.20888803899288177, "global_step": 73181, "epoch": 822, "lr": 8.327143611741893e-05}
{"train_loss": 0.12098193168640137, "global_step": 73182, "epoch": 822, "lr": 8.327100336418832e-05}
{"train_loss": 0.16291338205337524, "global_step": 73183, "epoch": 822, "lr": 8.327057060648483e-05}
{"train_loss": 0.3349722623825073, "global_step": 73184, "epoch": 822, "lr": 8.32701378443085e-05}
{"train_loss": 0.15336163341999054, "global_step": 73185, "epoch": 822, "lr": 8.326970507765942e-05}
{"train_loss": 0.212480828166008, "global_step": 73186, "epoch": 822, "lr": 8.326927230653762e-05}
{"train_loss": 0.16224031150341034, "global_step": 73187, "epoch": 822, "lr": 8.326883953094318e-05}
{"train_loss": 0.2200133204460144, "global_step": 73188, "epoch": 822, "lr": 8.326840675087615e-05}
{"train_loss": 0.21778085827827454, "global_step": 73189, "epoch": 822, "lr": 8.326797396633658e-05}
{"train_loss": 0.22079819440841675, "global_step": 73190, "epoch": 822, "lr": 8.326754117732454e-05}
{"train_loss": 0.29802805185317993, "global_step": 73191, "epoch": 822, "lr": 8.326710838384008e-05}
{"train_loss": 0.16245803236961365, "global_step": 73192, "epoch": 822, "lr": 8.326667558588328e-05}
{"train_loss": 0.15698948502540588, "global_step": 73193, "epoch": 822, "lr": 8.326624278345416e-05}
{"train_loss": 0.1873876005411148, "global_step": 73194, "epoch": 822, "lr": 8.326580997655283e-05}
{"train_loss": 0.171644926071167, "global_step": 73195, "epoch": 822, "lr": 8.326537716517928e-05}
{"train_loss": 0.19477249681949615, "global_step": 73196, "epoch": 822, "lr": 8.326494434933363e-05}
{"train_loss": 0.12308277189731598, "global_step": 73197, "epoch": 822, "lr": 8.32645115290159e-05}
{"train_loss": 0.17349117994308472, "global_step": 73198, "epoch": 822, "lr": 8.326407870422617e-05}
{"train_loss": 0.17232085764408112, "global_step": 73199, "epoch": 822, "lr": 8.326364587496449e-05}
{"train_loss": 0.2866691052913666, "global_step": 73200, "epoch": 822, "lr": 8.326321304123093e-05}
{"train_loss": 0.1325923204421997, "global_step": 73201, "epoch": 822, "lr": 8.326278020302552e-05}
{"train_loss": 0.20082929730415344, "global_step": 73202, "epoch": 822, "lr": 8.326234736034833e-05}
{"train_loss": 0.2525257468223572, "global_step": 73203, "epoch": 822, "lr": 8.326191451319944e-05}
{"train_loss": 0.16772601008415222, "global_step": 73204, "epoch": 822, "lr": 8.326148166157888e-05}
{"train_loss": 0.178431898355484, "global_step": 73205, "epoch": 822, "lr": 8.326104880548671e-05}
{"train_loss": 0.23847565054893494, "global_step": 73206, "epoch": 822, "lr": 8.326061594492302e-05}
{"train_loss": 0.22689905762672424, "global_step": 73207, "epoch": 822, "lr": 8.326018307988784e-05}
{"train_loss": 0.13763341307640076, "global_step": 73208, "epoch": 822, "lr": 8.325975021038121e-05}
{"train_loss": 0.2851923704147339, "global_step": 73209, "epoch": 822, "lr": 8.325931733640325e-05}
{"train_loss": 0.2566962242126465, "global_step": 73210, "epoch": 822, "lr": 8.325888445795395e-05}
{"train_loss": 0.26277920603752136, "global_step": 73211, "epoch": 822, "lr": 8.32584515750334e-05}
{"train_loss": 0.13931480050086975, "global_step": 73212, "epoch": 822, "lr": 8.325801868764167e-05}
{"train_loss": 0.20602509379386902, "global_step": 73213, "epoch": 822, "lr": 8.325758579577881e-05}
{"train_loss": 0.16806580126285553, "global_step": 73214, "epoch": 822, "lr": 8.325715289944485e-05}
{"train_loss": 0.16474922001361847, "global_step": 73215, "epoch": 822, "lr": 8.325671999863988e-05}
{"train_loss": 0.22179408371448517, "global_step": 73216, "epoch": 822, "lr": 8.325628709336396e-05}
{"train_loss": 0.23714466392993927, "global_step": 73217, "epoch": 822, "lr": 8.325585418361714e-05}
{"train_loss": 0.21222051978111267, "global_step": 73218, "epoch": 822, "lr": 8.325542126939945e-05}
{"train_loss": 0.18910454213619232, "global_step": 73219, "epoch": 822, "lr": 8.325498835071099e-05}
{"train_loss": 0.19973096251487732, "global_step": 73220, "epoch": 822, "lr": 8.325455542755181e-05}
{"train_loss": 0.23190532624721527, "global_step": 73221, "epoch": 822, "lr": 8.325412249992195e-05}
{"train_loss": 0.20423123240470886, "global_step": 73222, "epoch": 822, "lr": 8.325368956782149e-05}
{"train_loss": 0.14312127232551575, "global_step": 73223, "epoch": 822, "lr": 8.325325663125047e-05}
{"train_loss": 0.23594528436660767, "global_step": 73224, "epoch": 822, "lr": 8.325282369020894e-05}
{"train_loss": 0.17769867181777954, "global_step": 73225, "epoch": 822, "lr": 8.3252390744697e-05}
{"train_loss": 0.186050683259964, "global_step": 73226, "epoch": 822, "lr": 8.325195779471465e-05}
{"train_loss": 0.20378637313842773, "global_step": 73227, "epoch": 822, "lr": 8.3251524840262e-05}
{"train_loss": 0.16729797422885895, "global_step": 73228, "epoch": 822, "lr": 8.325109188133908e-05}
{"train_loss": 0.18729367852210999, "global_step": 73229, "epoch": 822, "lr": 8.325065891794597e-05}
{"train_loss": 0.20373737812042236, "global_step": 73230, "epoch": 822, "lr": 8.32502259500827e-05}
{"train_loss": 0.22375178337097168, "global_step": 73231, "epoch": 822, "lr": 8.324979297774936e-05}
{"train_loss": 0.19648906588554382, "global_step": 73232, "epoch": 822, "lr": 8.324936000094597e-05}
{"train_loss": 0.25971826910972595, "global_step": 73233, "epoch": 822, "lr": 8.324892701967261e-05}
{"train_loss": 0.24727654457092285, "global_step": 73234, "epoch": 822, "lr": 8.324849403392935e-05}
{"train_loss": 0.21941739320755005, "global_step": 73235, "epoch": 822, "lr": 8.324806104371622e-05}
{"train_loss": 0.18725305795669556, "global_step": 73236, "epoch": 822, "lr": 8.324762804903331e-05}
{"train_loss": 0.24350182712078094, "global_step": 73237, "epoch": 822, "lr": 8.324719504988065e-05}
{"train_loss": 0.17276714742183685, "global_step": 73238, "epoch": 822, "lr": 8.324676204625833e-05}
{"train_loss": 0.18735681474208832, "global_step": 73239, "epoch": 822, "lr": 8.324632903816636e-05}
{"train_loss": 0.20568454265594482, "global_step": 73240, "epoch": 822, "lr": 8.324589602560485e-05}
{"train_loss": 0.20060522854328156, "global_step": 73241, "epoch": 822, "lr": 8.324546300857382e-05}
{"train_loss": 0.24309247732162476, "global_step": 73242, "epoch": 822, "lr": 8.324502998707334e-05}
{"train_loss": 0.17700885236263275, "global_step": 73243, "epoch": 822, "lr": 8.324459696110348e-05}
{"train_loss": 0.17873546481132507, "global_step": 73244, "epoch": 822, "lr": 8.324416393066428e-05}
{"train_loss": 0.24405184388160706, "global_step": 73245, "epoch": 822, "lr": 8.324373089575581e-05}
{"train_loss": 0.19931794167234657, "global_step": 73246, "epoch": 822, "lr": 8.324329785637813e-05, "val_loss": 3.9507157802581787}
{"train_loss": 0.19409114122390747, "global_step": 73247, "epoch": 823, "lr": 8.32428648125313e-05}
{"train_loss": 0.2280123233795166, "global_step": 73248, "epoch": 823, "lr": 8.324243176421535e-05}
{"train_loss": 0.2270072102546692, "global_step": 73249, "epoch": 823, "lr": 8.32419987114304e-05}
{"train_loss": 0.16296321153640747, "global_step": 73250, "epoch": 823, "lr": 8.324156565417642e-05}
{"train_loss": 0.23576460778713226, "global_step": 73251, "epoch": 823, "lr": 8.324113259245353e-05}
{"train_loss": 0.183852419257164, "global_step": 73252, "epoch": 823, "lr": 8.32406995262618e-05}
{"train_loss": 0.22073175013065338, "global_step": 73253, "epoch": 823, "lr": 8.324026645560125e-05}
{"train_loss": 0.23112016916275024, "global_step": 73254, "epoch": 823, "lr": 8.323983338047194e-05}
{"train_loss": 0.23238220810890198, "global_step": 73255, "epoch": 823, "lr": 8.323940030087396e-05}
{"train_loss": 0.17471374571323395, "global_step": 73256, "epoch": 823, "lr": 8.323896721680733e-05}
{"train_loss": 0.2099662572145462, "global_step": 73257, "epoch": 823, "lr": 8.323853412827213e-05}
{"train_loss": 0.22845888137817383, "global_step": 73258, "epoch": 823, "lr": 8.323810103526841e-05}
{"train_loss": 0.1503591537475586, "global_step": 73259, "epoch": 823, "lr": 8.323766793779625e-05}
{"train_loss": 0.12775540351867676, "global_step": 73260, "epoch": 823, "lr": 8.323723483585568e-05}
{"train_loss": 0.27485498785972595, "global_step": 73261, "epoch": 823, "lr": 8.323680172944676e-05}
{"train_loss": 0.3113665282726288, "global_step": 73262, "epoch": 823, "lr": 8.323636861856959e-05}
{"train_loss": 0.24738584458827972, "global_step": 73263, "epoch": 823, "lr": 8.323593550322415e-05}
{"train_loss": 0.18887759745121002, "global_step": 73264, "epoch": 823, "lr": 8.323550238341056e-05}
{"train_loss": 0.21385277807712555, "global_step": 73265, "epoch": 823, "lr": 8.323506925912887e-05}
{"train_loss": 0.2177029848098755, "global_step": 73266, "epoch": 823, "lr": 8.323463613037912e-05}
{"train_loss": 0.36908555030822754, "global_step": 73267, "epoch": 823, "lr": 8.32342029971614e-05}
{"train_loss": 0.18582656979560852, "global_step": 73268, "epoch": 823, "lr": 8.323376985947572e-05}
{"train_loss": 0.2625196576118469, "global_step": 73269, "epoch": 823, "lr": 8.323333671732219e-05}
{"train_loss": 0.2130906730890274, "global_step": 73270, "epoch": 823, "lr": 8.323290357070081e-05}
{"train_loss": 0.19013261795043945, "global_step": 73271, "epoch": 823, "lr": 8.323247041961169e-05}
{"train_loss": 0.30638760328292847, "global_step": 73272, "epoch": 823, "lr": 8.323203726405487e-05}
{"train_loss": 0.31494614481925964, "global_step": 73273, "epoch": 823, "lr": 8.323160410403042e-05}
{"train_loss": 0.2436431348323822, "global_step": 73274, "epoch": 823, "lr": 8.323117093953837e-05}
{"train_loss": 0.1849275827407837, "global_step": 73275, "epoch": 823, "lr": 8.323073777057879e-05}
{"train_loss": 0.1732168048620224, "global_step": 73276, "epoch": 823, "lr": 8.323030459715176e-05}
{"train_loss": 0.1708226203918457, "global_step": 73277, "epoch": 823, "lr": 8.32298714192573e-05}
{"train_loss": 0.202304869890213, "global_step": 73278, "epoch": 823, "lr": 8.32294382368955e-05}
{"train_loss": 0.30007708072662354, "global_step": 73279, "epoch": 823, "lr": 8.322900505006642e-05}
{"train_loss": 0.23614534735679626, "global_step": 73280, "epoch": 823, "lr": 8.322857185877008e-05}
{"train_loss": 0.2824786305427551, "global_step": 73281, "epoch": 823, "lr": 8.322813866300656e-05}
{"train_loss": 0.1335107535123825, "global_step": 73282, "epoch": 823, "lr": 8.322770546277595e-05}
{"train_loss": 0.20770108699798584, "global_step": 73283, "epoch": 823, "lr": 8.322727225807825e-05}
{"train_loss": 0.19129015505313873, "global_step": 73284, "epoch": 823, "lr": 8.322683904891358e-05}
{"train_loss": 0.16533036530017853, "global_step": 73285, "epoch": 823, "lr": 8.322640583528194e-05}
{"train_loss": 0.21702709794044495, "global_step": 73286, "epoch": 823, "lr": 8.322597261718344e-05}
{"train_loss": 0.27173981070518494, "global_step": 73287, "epoch": 823, "lr": 8.322553939461809e-05}
{"train_loss": 0.23607835173606873, "global_step": 73288, "epoch": 823, "lr": 8.322510616758597e-05}
{"train_loss": 0.15422512590885162, "global_step": 73289, "epoch": 823, "lr": 8.322467293608716e-05}
{"train_loss": 0.1761191189289093, "global_step": 73290, "epoch": 823, "lr": 8.322423970012168e-05}
{"train_loss": 0.17628087103366852, "global_step": 73291, "epoch": 823, "lr": 8.322380645968961e-05}
{"train_loss": 0.3010497987270355, "global_step": 73292, "epoch": 823, "lr": 8.322337321479101e-05}
{"train_loss": 0.2660244107246399, "global_step": 73293, "epoch": 823, "lr": 8.322293996542594e-05}
{"train_loss": 0.22132225334644318, "global_step": 73294, "epoch": 823, "lr": 8.322250671159443e-05}
{"train_loss": 0.2661827504634857, "global_step": 73295, "epoch": 823, "lr": 8.322207345329657e-05}
{"train_loss": 0.2038789689540863, "global_step": 73296, "epoch": 823, "lr": 8.32216401905324e-05}
{"train_loss": 0.1514536440372467, "global_step": 73297, "epoch": 823, "lr": 8.3221206923302e-05}
{"train_loss": 0.20869505405426025, "global_step": 73298, "epoch": 823, "lr": 8.32207736516054e-05}
{"train_loss": 0.1709480583667755, "global_step": 73299, "epoch": 823, "lr": 8.322034037544267e-05}
{"train_loss": 0.1625278890132904, "global_step": 73300, "epoch": 823, "lr": 8.321990709481388e-05}
{"train_loss": 0.18573898077011108, "global_step": 73301, "epoch": 823, "lr": 8.321947380971908e-05}
{"train_loss": 0.20118509232997894, "global_step": 73302, "epoch": 823, "lr": 8.321904052015832e-05}
{"train_loss": 0.2543393671512604, "global_step": 73303, "epoch": 823, "lr": 8.321860722613166e-05}
{"train_loss": 0.28475746512413025, "global_step": 73304, "epoch": 823, "lr": 8.321817392763918e-05}
{"train_loss": 0.1859285980463028, "global_step": 73305, "epoch": 823, "lr": 8.32177406246809e-05}
{"train_loss": 0.17095091938972473, "global_step": 73306, "epoch": 823, "lr": 8.321730731725691e-05}
{"train_loss": 0.2542581856250763, "global_step": 73307, "epoch": 823, "lr": 8.321687400536726e-05}
{"train_loss": 0.18250243365764618, "global_step": 73308, "epoch": 823, "lr": 8.321644068901203e-05}
{"train_loss": 0.23166802525520325, "global_step": 73309, "epoch": 823, "lr": 8.321600736819122e-05}
{"train_loss": 0.1762566715478897, "global_step": 73310, "epoch": 823, "lr": 8.321557404290492e-05}
{"train_loss": 0.1653595268726349, "global_step": 73311, "epoch": 823, "lr": 8.32151407131532e-05}
{"train_loss": 0.19259823858737946, "global_step": 73312, "epoch": 823, "lr": 8.321470737893611e-05}
{"train_loss": 0.24022871255874634, "global_step": 73313, "epoch": 823, "lr": 8.321427404025373e-05}
{"train_loss": 0.20092739164829254, "global_step": 73314, "epoch": 823, "lr": 8.321384069710606e-05}
{"train_loss": 0.2613298296928406, "global_step": 73315, "epoch": 823, "lr": 8.321340734949321e-05}
{"train_loss": 0.22328190505504608, "global_step": 73316, "epoch": 823, "lr": 8.321297399741522e-05}
{"train_loss": 0.25339093804359436, "global_step": 73317, "epoch": 823, "lr": 8.321254064087216e-05}
{"train_loss": 0.1397305727005005, "global_step": 73318, "epoch": 823, "lr": 8.321210727986406e-05}
{"train_loss": 0.2582913041114807, "global_step": 73319, "epoch": 823, "lr": 8.321167391439101e-05}
{"train_loss": 0.29720184206962585, "global_step": 73320, "epoch": 823, "lr": 8.321124054445305e-05}
{"train_loss": 0.29498109221458435, "global_step": 73321, "epoch": 823, "lr": 8.321080717005025e-05}
{"train_loss": 0.13356567919254303, "global_step": 73322, "epoch": 823, "lr": 8.321037379118266e-05}
{"train_loss": 0.19441045820713043, "global_step": 73323, "epoch": 823, "lr": 8.320994040785033e-05}
{"train_loss": 0.16445189714431763, "global_step": 73324, "epoch": 823, "lr": 8.320950702005335e-05}
{"train_loss": 0.23890085518360138, "global_step": 73325, "epoch": 823, "lr": 8.320907362779174e-05}
{"train_loss": 0.2501866817474365, "global_step": 73326, "epoch": 823, "lr": 8.320864023106558e-05}
{"train_loss": 0.2700362801551819, "global_step": 73327, "epoch": 823, "lr": 8.320820682987492e-05}
{"train_loss": 0.21433094143867493, "global_step": 73328, "epoch": 823, "lr": 8.320777342421982e-05}
{"train_loss": 0.2587752640247345, "global_step": 73329, "epoch": 823, "lr": 8.320734001410034e-05}
{"train_loss": 0.22688931226730347, "global_step": 73330, "epoch": 823, "lr": 8.320690659951653e-05}
{"train_loss": 0.26848098635673523, "global_step": 73331, "epoch": 823, "lr": 8.320647318046848e-05}
{"train_loss": 0.2002088576555252, "global_step": 73332, "epoch": 823, "lr": 8.32060397569562e-05}
{"train_loss": 0.2294653207063675, "global_step": 73333, "epoch": 823, "lr": 8.320560632897978e-05}
{"train_loss": 0.165650874376297, "global_step": 73334, "epoch": 823, "lr": 8.320517289653929e-05}
{"train_loss": 0.21878360211849213, "global_step": 73335, "epoch": 823, "lr": 8.320473945963474e-05, "val_loss": 3.6858134269714355}
{"train_loss": 0.20110616087913513, "global_step": 73336, "epoch": 824, "lr": 8.320430601826623e-05}
{"train_loss": 0.3028257191181183, "global_step": 73337, "epoch": 824, "lr": 8.32038725724338e-05}
{"train_loss": 0.11804260313510895, "global_step": 73338, "epoch": 824, "lr": 8.320343912213753e-05}
{"train_loss": 0.22970782220363617, "global_step": 73339, "epoch": 824, "lr": 8.320300566737744e-05}
{"train_loss": 0.18515610694885254, "global_step": 73340, "epoch": 824, "lr": 8.320257220815364e-05}
{"train_loss": 0.14452823996543884, "global_step": 73341, "epoch": 824, "lr": 8.320213874446615e-05}
{"train_loss": 0.1981954574584961, "global_step": 73342, "epoch": 824, "lr": 8.320170527631502e-05}
{"train_loss": 0.1279301792383194, "global_step": 73343, "epoch": 824, "lr": 8.320127180370032e-05}
{"train_loss": 0.23664383590221405, "global_step": 73344, "epoch": 824, "lr": 8.320083832662213e-05}
{"train_loss": 0.21889930963516235, "global_step": 73345, "epoch": 824, "lr": 8.320040484508049e-05}
{"train_loss": 0.21897847950458527, "global_step": 73346, "epoch": 824, "lr": 8.319997135907546e-05}
{"train_loss": 0.2545010447502136, "global_step": 73347, "epoch": 824, "lr": 8.31995378686071e-05}
{"train_loss": 0.21338540315628052, "global_step": 73348, "epoch": 824, "lr": 8.319910437367547e-05}
{"train_loss": 0.2524853050708771, "global_step": 73349, "epoch": 824, "lr": 8.319867087428063e-05}
{"train_loss": 0.12816055119037628, "global_step": 73350, "epoch": 824, "lr": 8.319823737042262e-05}
{"train_loss": 0.1828262209892273, "global_step": 73351, "epoch": 824, "lr": 8.319780386210151e-05}
{"train_loss": 0.18586818873882294, "global_step": 73352, "epoch": 824, "lr": 8.319737034931737e-05}
{"train_loss": 0.2083573043346405, "global_step": 73353, "epoch": 824, "lr": 8.319693683207023e-05}
{"train_loss": 0.2346506267786026, "global_step": 73354, "epoch": 824, "lr": 8.31965033103602e-05}
{"train_loss": 0.1570373773574829, "global_step": 73355, "epoch": 824, "lr": 8.319606978418728e-05}
{"train_loss": 0.1988634169101715, "global_step": 73356, "epoch": 824, "lr": 8.319563625355156e-05}
{"train_loss": 0.14375804364681244, "global_step": 73357, "epoch": 824, "lr": 8.319520271845309e-05}
{"train_loss": 0.24736616015434265, "global_step": 73358, "epoch": 824, "lr": 8.319476917889192e-05}
{"train_loss": 0.17004677653312683, "global_step": 73359, "epoch": 824, "lr": 8.319433563486813e-05}
{"train_loss": 0.2248290479183197, "global_step": 73360, "epoch": 824, "lr": 8.319390208638176e-05}
{"train_loss": 0.22711072862148285, "global_step": 73361, "epoch": 824, "lr": 8.319346853343288e-05}
{"train_loss": 0.1471029818058014, "global_step": 73362, "epoch": 824, "lr": 8.319303497602154e-05}
{"train_loss": 0.1799495965242386, "global_step": 73363, "epoch": 824, "lr": 8.31926014141478e-05}
{"train_loss": 0.22827398777008057, "global_step": 73364, "epoch": 824, "lr": 8.319216784781173e-05}
{"train_loss": 0.15768957138061523, "global_step": 73365, "epoch": 824, "lr": 8.319173427701336e-05}
{"train_loss": 0.18401168286800385, "global_step": 73366, "epoch": 824, "lr": 8.319130070175277e-05}
{"train_loss": 0.18048682808876038, "global_step": 73367, "epoch": 824, "lr": 8.319086712203002e-05}
{"train_loss": 0.16639350354671478, "global_step": 73368, "epoch": 824, "lr": 8.319043353784516e-05}
{"train_loss": 0.25357601046562195, "global_step": 73369, "epoch": 824, "lr": 8.318999994919825e-05}
{"train_loss": 0.18229232728481293, "global_step": 73370, "epoch": 824, "lr": 8.318956635608935e-05}
{"train_loss": 0.14554208517074585, "global_step": 73371, "epoch": 824, "lr": 8.318913275851851e-05}
{"train_loss": 0.2578747570514679, "global_step": 73372, "epoch": 824, "lr": 8.318869915648582e-05}
{"train_loss": 0.20757371187210083, "global_step": 73373, "epoch": 824, "lr": 8.318826554999129e-05}
{"train_loss": 0.13682915270328522, "global_step": 73374, "epoch": 824, "lr": 8.318783193903501e-05}
{"train_loss": 0.18906240165233612, "global_step": 73375, "epoch": 824, "lr": 8.318739832361701e-05}
{"train_loss": 0.1894882619380951, "global_step": 73376, "epoch": 824, "lr": 8.31869647037374e-05}
{"train_loss": 0.3485288918018341, "global_step": 73377, "epoch": 824, "lr": 8.31865310793962e-05}
{"train_loss": 0.2497541755437851, "global_step": 73378, "epoch": 824, "lr": 8.318609745059347e-05}
{"train_loss": 0.2775387167930603, "global_step": 73379, "epoch": 824, "lr": 8.318566381732927e-05}
{"train_loss": 0.21367888152599335, "global_step": 73380, "epoch": 824, "lr": 8.318523017960367e-05}
{"train_loss": 0.27244970202445984, "global_step": 73381, "epoch": 824, "lr": 8.318479653741671e-05}
{"train_loss": 0.29410940408706665, "global_step": 73382, "epoch": 824, "lr": 8.318436289076847e-05}
{"train_loss": 0.20108678936958313, "global_step": 73383, "epoch": 824, "lr": 8.3183929239659e-05}
{"train_loss": 0.21497154235839844, "global_step": 73384, "epoch": 824, "lr": 8.318349558408834e-05}
{"train_loss": 0.18413224816322327, "global_step": 73385, "epoch": 824, "lr": 8.318306192405659e-05}
{"train_loss": 0.1684010922908783, "global_step": 73386, "epoch": 824, "lr": 8.318262825956375e-05}
{"train_loss": 0.19400934875011444, "global_step": 73387, "epoch": 824, "lr": 8.318219459060992e-05}
{"train_loss": 0.2269883155822754, "global_step": 73388, "epoch": 824, "lr": 8.318176091719517e-05}
{"train_loss": 0.19918417930603027, "global_step": 73389, "epoch": 824, "lr": 8.31813272393195e-05}
{"train_loss": 0.20101836323738098, "global_step": 73390, "epoch": 824, "lr": 8.318089355698302e-05}
{"train_loss": 0.3095777630805969, "global_step": 73391, "epoch": 824, "lr": 8.318045987018578e-05}
{"train_loss": 0.17344671487808228, "global_step": 73392, "epoch": 824, "lr": 8.318002617892781e-05}
{"train_loss": 0.17663328349590302, "global_step": 73393, "epoch": 824, "lr": 8.317959248320922e-05}
{"train_loss": 0.20862460136413574, "global_step": 73394, "epoch": 824, "lr": 8.317915878303002e-05}
{"train_loss": 0.21690154075622559, "global_step": 73395, "epoch": 824, "lr": 8.317872507839031e-05}
{"train_loss": 0.2365507036447525, "global_step": 73396, "epoch": 824, "lr": 8.31782913692901e-05}
{"train_loss": 0.2973695397377014, "global_step": 73397, "epoch": 824, "lr": 8.317785765572947e-05}
{"train_loss": 0.13967671990394592, "global_step": 73398, "epoch": 824, "lr": 8.31774239377085e-05}
{"train_loss": 0.23065045475959778, "global_step": 73399, "epoch": 824, "lr": 8.317699021522723e-05}
{"train_loss": 0.3194105327129364, "global_step": 73400, "epoch": 824, "lr": 8.31765564882857e-05}
{"train_loss": 0.13978615403175354, "global_step": 73401, "epoch": 824, "lr": 8.3176122756884e-05}
{"train_loss": 0.17045946419239044, "global_step": 73402, "epoch": 824, "lr": 8.317568902102217e-05}
{"train_loss": 0.1615927666425705, "global_step": 73403, "epoch": 824, "lr": 8.317525528070027e-05}
{"train_loss": 0.2387341856956482, "global_step": 73404, "epoch": 824, "lr": 8.317482153591837e-05}
{"train_loss": 0.22024710476398468, "global_step": 73405, "epoch": 824, "lr": 8.317438778667651e-05}
{"train_loss": 0.15828374028205872, "global_step": 73406, "epoch": 824, "lr": 8.317395403297477e-05}
{"train_loss": 0.18336547911167145, "global_step": 73407, "epoch": 824, "lr": 8.317352027481318e-05}
{"train_loss": 0.24100643396377563, "global_step": 73408, "epoch": 824, "lr": 8.317308651219182e-05}
{"train_loss": 0.265624076128006, "global_step": 73409, "epoch": 824, "lr": 8.317265274511075e-05}
{"train_loss": 0.2566565275192261, "global_step": 73410, "epoch": 824, "lr": 8.317221897357001e-05}
{"train_loss": 0.2515306770801544, "global_step": 73411, "epoch": 824, "lr": 8.317178519756968e-05}
{"train_loss": 0.2102419137954712, "global_step": 73412, "epoch": 824, "lr": 8.317135141710982e-05}
{"train_loss": 0.1981455534696579, "global_step": 73413, "epoch": 824, "lr": 8.317091763219045e-05}
{"train_loss": 0.1812857687473297, "global_step": 73414, "epoch": 824, "lr": 8.317048384281165e-05}
{"train_loss": 0.1725071668624878, "global_step": 73415, "epoch": 824, "lr": 8.31700500489735e-05}
{"train_loss": 0.16571249067783356, "global_step": 73416, "epoch": 824, "lr": 8.316961625067605e-05}
{"train_loss": 0.18490397930145264, "global_step": 73417, "epoch": 824, "lr": 8.316918244791934e-05}
{"train_loss": 0.26094889640808105, "global_step": 73418, "epoch": 824, "lr": 8.316874864070343e-05}
{"train_loss": 0.21693795919418335, "global_step": 73419, "epoch": 824, "lr": 8.316831482902838e-05}
{"train_loss": 0.21294914186000824, "global_step": 73420, "epoch": 824, "lr": 8.316788101289428e-05}
{"train_loss": 0.2422480285167694, "global_step": 73421, "epoch": 824, "lr": 8.316744719230115e-05}
{"train_loss": 0.2672041654586792, "global_step": 73422, "epoch": 824, "lr": 8.316701336724905e-05}
{"train_loss": 0.1943437159061432, "global_step": 73423, "epoch": 824, "lr": 8.316657953773807e-05}
{"train_loss": 0.20832490653134464, "global_step": 73424, "epoch": 824, "lr": 8.316614570376823e-05, "val_loss": 3.8373847007751465}
{"train_loss": 0.18230922520160675, "global_step": 73425, "epoch": 825, "lr": 8.316571186533962e-05}
{"train_loss": 0.1509547084569931, "global_step": 73426, "epoch": 825, "lr": 8.316527802245226e-05}
{"train_loss": 0.19511912763118744, "global_step": 73427, "epoch": 825, "lr": 8.316484417510625e-05}
{"train_loss": 0.22376559674739838, "global_step": 73428, "epoch": 825, "lr": 8.316441032330163e-05}
{"train_loss": 0.21344688534736633, "global_step": 73429, "epoch": 825, "lr": 8.316397646703846e-05}
{"train_loss": 0.19776372611522675, "global_step": 73430, "epoch": 825, "lr": 8.31635426063168e-05}
{"train_loss": 0.17590981721878052, "global_step": 73431, "epoch": 825, "lr": 8.31631087411367e-05}
{"train_loss": 0.2031891644001007, "global_step": 73432, "epoch": 825, "lr": 8.316267487149823e-05}
{"train_loss": 0.1498517096042633, "global_step": 73433, "epoch": 825, "lr": 8.316224099740144e-05}
{"train_loss": 0.2903600335121155, "global_step": 73434, "epoch": 825, "lr": 8.316180711884639e-05}
{"train_loss": 0.14469541609287262, "global_step": 73435, "epoch": 825, "lr": 8.316137323583313e-05}
{"train_loss": 0.1754504293203354, "global_step": 73436, "epoch": 825, "lr": 8.316093934836176e-05}
{"train_loss": 0.13592854142189026, "global_step": 73437, "epoch": 825, "lr": 8.316050545643227e-05}
{"train_loss": 0.28095531463623047, "global_step": 73438, "epoch": 825, "lr": 8.316007156004478e-05}
{"train_loss": 0.1745588779449463, "global_step": 73439, "epoch": 825, "lr": 8.31596376591993e-05}
{"train_loss": 0.13881540298461914, "global_step": 73440, "epoch": 825, "lr": 8.315920375389592e-05}
{"train_loss": 0.253464013338089, "global_step": 73441, "epoch": 825, "lr": 8.315876984413468e-05}
{"train_loss": 0.31863605976104736, "global_step": 73442, "epoch": 825, "lr": 8.315833592991566e-05}
{"train_loss": 0.17996680736541748, "global_step": 73443, "epoch": 825, "lr": 8.315790201123893e-05}
{"train_loss": 0.14047516882419586, "global_step": 73444, "epoch": 825, "lr": 8.315746808810447e-05}
{"train_loss": 0.21023547649383545, "global_step": 73445, "epoch": 825, "lr": 8.315703416051242e-05}
{"train_loss": 0.19211046397686005, "global_step": 73446, "epoch": 825, "lr": 8.315660022846281e-05}
{"train_loss": 0.2819267511367798, "global_step": 73447, "epoch": 825, "lr": 8.315616629195571e-05}
{"train_loss": 0.23501475155353546, "global_step": 73448, "epoch": 825, "lr": 8.315573235099116e-05}
{"train_loss": 0.24688276648521423, "global_step": 73449, "epoch": 825, "lr": 8.315529840556922e-05}
{"train_loss": 0.17477042973041534, "global_step": 73450, "epoch": 825, "lr": 8.315486445568996e-05}
{"train_loss": 0.18356239795684814, "global_step": 73451, "epoch": 825, "lr": 8.315443050135343e-05}
{"train_loss": 0.2509356737136841, "global_step": 73452, "epoch": 825, "lr": 8.315399654255969e-05}
{"train_loss": 0.26152896881103516, "global_step": 73453, "epoch": 825, "lr": 8.31535625793088e-05}
{"train_loss": 0.24293969571590424, "global_step": 73454, "epoch": 825, "lr": 8.315312861160083e-05}
{"train_loss": 0.26760369539260864, "global_step": 73455, "epoch": 825, "lr": 8.315269463943581e-05}
{"train_loss": 0.24560803174972534, "global_step": 73456, "epoch": 825, "lr": 8.315226066281381e-05}
{"train_loss": 0.20319148898124695, "global_step": 73457, "epoch": 825, "lr": 8.315182668173491e-05}
{"train_loss": 0.19835367798805237, "global_step": 73458, "epoch": 825, "lr": 8.315139269619914e-05}
{"train_loss": 0.14809191226959229, "global_step": 73459, "epoch": 825, "lr": 8.315095870620657e-05}
{"train_loss": 0.22800801694393158, "global_step": 73460, "epoch": 825, "lr": 8.315052471175725e-05}
{"train_loss": 0.26166456937789917, "global_step": 73461, "epoch": 825, "lr": 8.315009071285127e-05}
{"train_loss": 0.22078977525234222, "global_step": 73462, "epoch": 825, "lr": 8.314965670948864e-05}
{"train_loss": 0.2045692652463913, "global_step": 73463, "epoch": 825, "lr": 8.314922270166944e-05}
{"train_loss": 0.19017347693443298, "global_step": 73464, "epoch": 825, "lr": 8.314878868939375e-05}
{"train_loss": 0.22378620505332947, "global_step": 73465, "epoch": 825, "lr": 8.31483546726616e-05}
{"train_loss": 0.22790101170539856, "global_step": 73466, "epoch": 825, "lr": 8.314792065147306e-05}
{"train_loss": 0.18588203191757202, "global_step": 73467, "epoch": 825, "lr": 8.314748662582818e-05}
{"train_loss": 0.19523684680461884, "global_step": 73468, "epoch": 825, "lr": 8.314705259572704e-05}
{"train_loss": 0.17357280850410461, "global_step": 73469, "epoch": 825, "lr": 8.314661856116967e-05}
{"train_loss": 0.19530943036079407, "global_step": 73470, "epoch": 825, "lr": 8.314618452215616e-05}
{"train_loss": 0.1897081881761551, "global_step": 73471, "epoch": 825, "lr": 8.314575047868654e-05}
{"train_loss": 0.15507788956165314, "global_step": 73472, "epoch": 825, "lr": 8.314531643076087e-05}
{"train_loss": 0.3105293810367584, "global_step": 73473, "epoch": 825, "lr": 8.31448823783792e-05}
{"train_loss": 0.18435117602348328, "global_step": 73474, "epoch": 825, "lr": 8.314444832154164e-05}
{"train_loss": 0.22997350990772247, "global_step": 73475, "epoch": 825, "lr": 8.314401426024818e-05}
{"train_loss": 0.16106481850147247, "global_step": 73476, "epoch": 825, "lr": 8.314358019449894e-05}
{"train_loss": 0.15202824771404266, "global_step": 73477, "epoch": 825, "lr": 8.314314612429394e-05}
{"train_loss": 0.16198168694972992, "global_step": 73478, "epoch": 825, "lr": 8.314271204963324e-05}
{"train_loss": 0.21237064898014069, "global_step": 73479, "epoch": 825, "lr": 8.31422779705169e-05}
{"train_loss": 0.23899386823177338, "global_step": 73480, "epoch": 825, "lr": 8.314184388694499e-05}
{"train_loss": 0.20032034814357758, "global_step": 73481, "epoch": 825, "lr": 8.314140979891757e-05}
{"train_loss": 0.19724737107753754, "global_step": 73482, "epoch": 825, "lr": 8.31409757064347e-05}
{"train_loss": 0.2829517126083374, "global_step": 73483, "epoch": 825, "lr": 8.314054160949642e-05}
{"train_loss": 0.23702998459339142, "global_step": 73484, "epoch": 825, "lr": 8.31401075081028e-05}
{"train_loss": 0.30188146233558655, "global_step": 73485, "epoch": 825, "lr": 8.31396734022539e-05}
{"train_loss": 0.2384856641292572, "global_step": 73486, "epoch": 825, "lr": 8.313923929194976e-05}
{"train_loss": 0.17067170143127441, "global_step": 73487, "epoch": 825, "lr": 8.313880517719045e-05}
{"train_loss": 0.1915244162082672, "global_step": 73488, "epoch": 825, "lr": 8.313837105797606e-05}
{"train_loss": 0.19971421360969543, "global_step": 73489, "epoch": 825, "lr": 8.31379369343066e-05}
{"train_loss": 0.22443245351314545, "global_step": 73490, "epoch": 825, "lr": 8.313750280618216e-05}
{"train_loss": 0.17651627957820892, "global_step": 73491, "epoch": 825, "lr": 8.313706867360277e-05}
{"train_loss": 0.15480796992778778, "global_step": 73492, "epoch": 825, "lr": 8.313663453656852e-05}
{"train_loss": 0.1598084419965744, "global_step": 73493, "epoch": 825, "lr": 8.313620039507945e-05}
{"train_loss": 0.2709801197052002, "global_step": 73494, "epoch": 825, "lr": 8.313576624913562e-05}
{"train_loss": 0.22857683897018433, "global_step": 73495, "epoch": 825, "lr": 8.313533209873708e-05}
{"train_loss": 0.20023033022880554, "global_step": 73496, "epoch": 825, "lr": 8.313489794388392e-05}
{"train_loss": 0.2598053216934204, "global_step": 73497, "epoch": 825, "lr": 8.313446378457615e-05}
{"train_loss": 0.12497969716787338, "global_step": 73498, "epoch": 825, "lr": 8.313402962081387e-05}
{"train_loss": 0.2218017727136612, "global_step": 73499, "epoch": 825, "lr": 8.313359545259712e-05}
{"train_loss": 0.2515564560890198, "global_step": 73500, "epoch": 825, "lr": 8.313316127992597e-05}
{"train_loss": 0.2440527081489563, "global_step": 73501, "epoch": 825, "lr": 8.313272710280047e-05}
{"train_loss": 0.2155887931585312, "global_step": 73502, "epoch": 825, "lr": 8.313229292122066e-05}
{"train_loss": 0.2677422761917114, "global_step": 73503, "epoch": 825, "lr": 8.313185873518663e-05}
{"train_loss": 0.24963954091072083, "global_step": 73504, "epoch": 825, "lr": 8.313142454469843e-05}
{"train_loss": 0.2003936916589737, "global_step": 73505, "epoch": 825, "lr": 8.313099034975611e-05}
{"train_loss": 0.20680275559425354, "global_step": 73506, "epoch": 825, "lr": 8.313055615035975e-05}
{"train_loss": 0.17269499599933624, "global_step": 73507, "epoch": 825, "lr": 8.313012194650935e-05}
{"train_loss": 0.2723638117313385, "global_step": 73508, "epoch": 825, "lr": 8.312968773820505e-05}
{"train_loss": 0.23794890940189362, "global_step": 73509, "epoch": 825, "lr": 8.312925352544685e-05}
{"train_loss": 0.1645335704088211, "global_step": 73510, "epoch": 825, "lr": 8.312881930823482e-05}
{"train_loss": 0.16855524480342865, "global_step": 73511, "epoch": 825, "lr": 8.312838508656903e-05}
{"train_loss": 0.22050148248672485, "global_step": 73512, "epoch": 825, "lr": 8.312795086044953e-05}
{"train_loss": 0.20956841337212015, "global_step": 73513, "epoch": 825, "lr": 8.312751662987638e-05, "val_loss": 3.7575266361236572, "train_action_mse_error": 30.79026985168457}
{"train_loss": 0.18996396660804749, "global_step": 73514, "epoch": 826, "lr": 8.312708239484963e-05}
{"train_loss": 0.16203005611896515, "global_step": 73515, "epoch": 826, "lr": 8.312664815536937e-05}
{"train_loss": 0.1666388213634491, "global_step": 73516, "epoch": 826, "lr": 8.312621391143562e-05}
{"train_loss": 0.16749823093414307, "global_step": 73517, "epoch": 826, "lr": 8.312577966304847e-05}
{"train_loss": 0.18064725399017334, "global_step": 73518, "epoch": 826, "lr": 8.312534541020794e-05}
{"train_loss": 0.20349229872226715, "global_step": 73519, "epoch": 826, "lr": 8.312491115291412e-05}
{"train_loss": 0.20404475927352905, "global_step": 73520, "epoch": 826, "lr": 8.312447689116707e-05}
{"train_loss": 0.27534765005111694, "global_step": 73521, "epoch": 826, "lr": 8.312404262496683e-05}
{"train_loss": 0.22037142515182495, "global_step": 73522, "epoch": 826, "lr": 8.312360835431345e-05}
{"train_loss": 0.2134387195110321, "global_step": 73523, "epoch": 826, "lr": 8.312317407920702e-05}
{"train_loss": 0.20113933086395264, "global_step": 73524, "epoch": 826, "lr": 8.31227397996476e-05}
{"train_loss": 0.309535413980484, "global_step": 73525, "epoch": 826, "lr": 8.31223055156352e-05}
{"train_loss": 0.20670157670974731, "global_step": 73526, "epoch": 826, "lr": 8.312187122716994e-05}
{"train_loss": 0.17985142767429352, "global_step": 73527, "epoch": 826, "lr": 8.312143693425183e-05}
{"train_loss": 0.1947767436504364, "global_step": 73528, "epoch": 826, "lr": 8.312100263688094e-05}
{"train_loss": 0.17110614478588104, "global_step": 73529, "epoch": 826, "lr": 8.312056833505735e-05}
{"train_loss": 0.2052120864391327, "global_step": 73530, "epoch": 826, "lr": 8.312013402878109e-05}
{"train_loss": 0.22265759110450745, "global_step": 73531, "epoch": 826, "lr": 8.311969971805225e-05}
{"train_loss": 0.22100208699703217, "global_step": 73532, "epoch": 826, "lr": 8.311926540287085e-05}
{"train_loss": 0.1616242527961731, "global_step": 73533, "epoch": 826, "lr": 8.311883108323698e-05}
{"train_loss": 0.19818544387817383, "global_step": 73534, "epoch": 826, "lr": 8.311839675915068e-05}
{"train_loss": 0.13398997485637665, "global_step": 73535, "epoch": 826, "lr": 8.311796243061202e-05}
{"train_loss": 0.1902264952659607, "global_step": 73536, "epoch": 826, "lr": 8.311752809762105e-05}
{"train_loss": 0.22970886528491974, "global_step": 73537, "epoch": 826, "lr": 8.311709376017784e-05}
{"train_loss": 0.16717013716697693, "global_step": 73538, "epoch": 826, "lr": 8.311665941828244e-05}
{"train_loss": 0.2032368779182434, "global_step": 73539, "epoch": 826, "lr": 8.311622507193491e-05}
{"train_loss": 0.14992423355579376, "global_step": 73540, "epoch": 826, "lr": 8.311579072113529e-05}
{"train_loss": 0.20469139516353607, "global_step": 73541, "epoch": 826, "lr": 8.311535636588367e-05}
{"train_loss": 0.20154495537281036, "global_step": 73542, "epoch": 826, "lr": 8.31149220061801e-05}
{"train_loss": 0.11730977147817612, "global_step": 73543, "epoch": 826, "lr": 8.311448764202461e-05}
{"train_loss": 0.17769446969032288, "global_step": 73544, "epoch": 826, "lr": 8.31140532734173e-05}
{"train_loss": 0.2992597222328186, "global_step": 73545, "epoch": 826, "lr": 8.31136189003582e-05}
{"train_loss": 0.22005729377269745, "global_step": 73546, "epoch": 826, "lr": 8.311318452284738e-05}
{"train_loss": 0.15275217592716217, "global_step": 73547, "epoch": 826, "lr": 8.31127501408849e-05}
{"train_loss": 0.24543076753616333, "global_step": 73548, "epoch": 826, "lr": 8.31123157544708e-05}
{"train_loss": 0.1811327189207077, "global_step": 73549, "epoch": 826, "lr": 8.311188136360517e-05}
{"train_loss": 0.21033412218093872, "global_step": 73550, "epoch": 826, "lr": 8.311144696828805e-05}
{"train_loss": 0.20445872843265533, "global_step": 73551, "epoch": 826, "lr": 8.311101256851949e-05}
{"train_loss": 0.18924765288829803, "global_step": 73552, "epoch": 826, "lr": 8.311057816429957e-05}
{"train_loss": 0.16699746251106262, "global_step": 73553, "epoch": 826, "lr": 8.311014375562833e-05}
{"train_loss": 0.18658363819122314, "global_step": 73554, "epoch": 826, "lr": 8.310970934250583e-05}
{"train_loss": 0.23315836489200592, "global_step": 73555, "epoch": 826, "lr": 8.310927492493212e-05}
{"train_loss": 0.13598793745040894, "global_step": 73556, "epoch": 826, "lr": 8.310884050290731e-05}
{"train_loss": 0.2154170423746109, "global_step": 73557, "epoch": 826, "lr": 8.310840607643138e-05}
{"train_loss": 0.241921067237854, "global_step": 73558, "epoch": 826, "lr": 8.310797164550445e-05}
{"train_loss": 0.20269015431404114, "global_step": 73559, "epoch": 826, "lr": 8.310753721012656e-05}
{"train_loss": 0.16014479100704193, "global_step": 73560, "epoch": 826, "lr": 8.310710277029775e-05}
{"train_loss": 0.2031337320804596, "global_step": 73561, "epoch": 826, "lr": 8.31066683260181e-05}
{"train_loss": 0.23904797434806824, "global_step": 73562, "epoch": 826, "lr": 8.310623387728767e-05}
{"train_loss": 0.2849339246749878, "global_step": 73563, "epoch": 826, "lr": 8.31057994241065e-05}
{"train_loss": 0.22092309594154358, "global_step": 73564, "epoch": 826, "lr": 8.310536496647467e-05}
{"train_loss": 0.2025824934244156, "global_step": 73565, "epoch": 826, "lr": 8.310493050439222e-05}
{"train_loss": 0.24120496213436127, "global_step": 73566, "epoch": 826, "lr": 8.310449603785921e-05}
{"train_loss": 0.2039252370595932, "global_step": 73567, "epoch": 826, "lr": 8.310406156687571e-05}
{"train_loss": 0.1828281730413437, "global_step": 73568, "epoch": 826, "lr": 8.310362709144177e-05}
{"train_loss": 0.16916386783123016, "global_step": 73569, "epoch": 826, "lr": 8.310319261155746e-05}
{"train_loss": 0.15162423253059387, "global_step": 73570, "epoch": 826, "lr": 8.310275812722282e-05}
{"train_loss": 0.17362096905708313, "global_step": 73571, "epoch": 826, "lr": 8.310232363843791e-05}
{"train_loss": 0.1635255366563797, "global_step": 73572, "epoch": 826, "lr": 8.310188914520281e-05}
{"train_loss": 0.20271463692188263, "global_step": 73573, "epoch": 826, "lr": 8.310145464751757e-05}
{"train_loss": 0.19430559873580933, "global_step": 73574, "epoch": 826, "lr": 8.310102014538222e-05}
{"train_loss": 0.21248160302639008, "global_step": 73575, "epoch": 826, "lr": 8.310058563879686e-05}
{"train_loss": 0.18232278525829315, "global_step": 73576, "epoch": 826, "lr": 8.310015112776152e-05}
{"train_loss": 0.23081371188163757, "global_step": 73577, "epoch": 826, "lr": 8.309971661227627e-05}
{"train_loss": 0.17129647731781006, "global_step": 73578, "epoch": 826, "lr": 8.309928209234117e-05}
{"train_loss": 0.24467046558856964, "global_step": 73579, "epoch": 826, "lr": 8.309884756795627e-05}
{"train_loss": 0.18719734251499176, "global_step": 73580, "epoch": 826, "lr": 8.309841303912165e-05}
{"train_loss": 0.232990100979805, "global_step": 73581, "epoch": 826, "lr": 8.309797850583733e-05}
{"train_loss": 0.2353418469429016, "global_step": 73582, "epoch": 826, "lr": 8.309754396810341e-05}
{"train_loss": 0.17497524619102478, "global_step": 73583, "epoch": 826, "lr": 8.309710942591991e-05}
{"train_loss": 0.12812823057174683, "global_step": 73584, "epoch": 826, "lr": 8.309667487928691e-05}
{"train_loss": 0.13919277489185333, "global_step": 73585, "epoch": 826, "lr": 8.309624032820447e-05}
{"train_loss": 0.22127394378185272, "global_step": 73586, "epoch": 826, "lr": 8.309580577267266e-05}
{"train_loss": 0.24930283427238464, "global_step": 73587, "epoch": 826, "lr": 8.30953712126915e-05}
{"train_loss": 0.166007399559021, "global_step": 73588, "epoch": 826, "lr": 8.309493664826107e-05}
{"train_loss": 0.17358127236366272, "global_step": 73589, "epoch": 826, "lr": 8.309450207938144e-05}
{"train_loss": 0.23260000348091125, "global_step": 73590, "epoch": 826, "lr": 8.309406750605264e-05}
{"train_loss": 0.16870366036891937, "global_step": 73591, "epoch": 826, "lr": 8.309363292827477e-05}
{"train_loss": 0.18903474509716034, "global_step": 73592, "epoch": 826, "lr": 8.309319834604785e-05}
{"train_loss": 0.22167102992534637, "global_step": 73593, "epoch": 826, "lr": 8.309276375937196e-05}
{"train_loss": 0.19331161677837372, "global_step": 73594, "epoch": 826, "lr": 8.309232916824714e-05}
{"train_loss": 0.27640217542648315, "global_step": 73595, "epoch": 826, "lr": 8.309189457267347e-05}
{"train_loss": 0.165142223238945, "global_step": 73596, "epoch": 826, "lr": 8.3091459972651e-05}
{"train_loss": 0.14615197479724884, "global_step": 73597, "epoch": 826, "lr": 8.309102536817979e-05}
{"train_loss": 0.1783393919467926, "global_step": 73598, "epoch": 826, "lr": 8.309059075925988e-05}
{"train_loss": 0.24289754033088684, "global_step": 73599, "epoch": 826, "lr": 8.309015614589135e-05}
{"train_loss": 0.16145531833171844, "global_step": 73600, "epoch": 826, "lr": 8.308972152807426e-05}
{"train_loss": 0.21890321373939514, "global_step": 73601, "epoch": 826, "lr": 8.308928690580864e-05}
{"train_loss": 0.1981353166230609, "global_step": 73602, "epoch": 826, "lr": 8.308885227909459e-05, "val_loss": 3.9778289794921875}
{"train_loss": 0.17895066738128662, "global_step": 73603, "epoch": 827, "lr": 8.308841764793214e-05}
{"train_loss": 0.23006461560726166, "global_step": 73604, "epoch": 827, "lr": 8.308798301232136e-05}
{"train_loss": 0.19898606836795807, "global_step": 73605, "epoch": 827, "lr": 8.30875483722623e-05}
{"train_loss": 0.22834782302379608, "global_step": 73606, "epoch": 827, "lr": 8.308711372775502e-05}
{"train_loss": 0.1481170952320099, "global_step": 73607, "epoch": 827, "lr": 8.308667907879956e-05}
{"train_loss": 0.13397517800331116, "global_step": 73608, "epoch": 827, "lr": 8.308624442539603e-05}
{"train_loss": 0.20378167927265167, "global_step": 73609, "epoch": 827, "lr": 8.308580976754445e-05}
{"train_loss": 0.23176519572734833, "global_step": 73610, "epoch": 827, "lr": 8.308537510524487e-05}
{"train_loss": 0.17992061376571655, "global_step": 73611, "epoch": 827, "lr": 8.30849404384974e-05}
{"train_loss": 0.21295993030071259, "global_step": 73612, "epoch": 827, "lr": 8.308450576730203e-05}
{"train_loss": 0.2028350681066513, "global_step": 73613, "epoch": 827, "lr": 8.308407109165887e-05}
{"train_loss": 0.21679076552391052, "global_step": 73614, "epoch": 827, "lr": 8.308363641156795e-05}
{"train_loss": 0.21843740344047546, "global_step": 73615, "epoch": 827, "lr": 8.308320172702934e-05}
{"train_loss": 0.28576338291168213, "global_step": 73616, "epoch": 827, "lr": 8.30827670380431e-05}
{"train_loss": 0.2721920311450958, "global_step": 73617, "epoch": 827, "lr": 8.308233234460929e-05}
{"train_loss": 0.18548144400119781, "global_step": 73618, "epoch": 827, "lr": 8.308189764672796e-05}
{"train_loss": 0.2836188077926636, "global_step": 73619, "epoch": 827, "lr": 8.308146294439916e-05}
{"train_loss": 0.21017853915691376, "global_step": 73620, "epoch": 827, "lr": 8.308102823762297e-05}
{"train_loss": 0.21476562321186066, "global_step": 73621, "epoch": 827, "lr": 8.308059352639944e-05}
{"train_loss": 0.1499255895614624, "global_step": 73622, "epoch": 827, "lr": 8.308015881072863e-05}
{"train_loss": 0.1830490231513977, "global_step": 73623, "epoch": 827, "lr": 8.30797240906106e-05}
{"train_loss": 0.24240903556346893, "global_step": 73624, "epoch": 827, "lr": 8.307928936604539e-05}
{"train_loss": 0.19933107495307922, "global_step": 73625, "epoch": 827, "lr": 8.307885463703308e-05}
{"train_loss": 0.2101743072271347, "global_step": 73626, "epoch": 827, "lr": 8.307841990357371e-05}
{"train_loss": 0.36623096466064453, "global_step": 73627, "epoch": 827, "lr": 8.307798516566736e-05}
{"train_loss": 0.23871272802352905, "global_step": 73628, "epoch": 827, "lr": 8.307755042331408e-05}
{"train_loss": 0.2575961649417877, "global_step": 73629, "epoch": 827, "lr": 8.307711567651393e-05}
{"train_loss": 0.22572585940361023, "global_step": 73630, "epoch": 827, "lr": 8.307668092526695e-05}
{"train_loss": 0.20700237154960632, "global_step": 73631, "epoch": 827, "lr": 8.307624616957324e-05}
{"train_loss": 0.24075353145599365, "global_step": 73632, "epoch": 827, "lr": 8.307581140943281e-05}
{"train_loss": 0.21241308748722076, "global_step": 73633, "epoch": 827, "lr": 8.307537664484574e-05}
{"train_loss": 0.22554560005664825, "global_step": 73634, "epoch": 827, "lr": 8.307494187581208e-05}
{"train_loss": 0.21863628923892975, "global_step": 73635, "epoch": 827, "lr": 8.307450710233193e-05}
{"train_loss": 0.28116297721862793, "global_step": 73636, "epoch": 827, "lr": 8.307407232440529e-05}
{"train_loss": 0.3696178197860718, "global_step": 73637, "epoch": 827, "lr": 8.307363754203225e-05}
{"train_loss": 0.2084798514842987, "global_step": 73638, "epoch": 827, "lr": 8.307320275521287e-05}
{"train_loss": 0.19735880196094513, "global_step": 73639, "epoch": 827, "lr": 8.307276796394719e-05}
{"train_loss": 0.18509432673454285, "global_step": 73640, "epoch": 827, "lr": 8.307233316823529e-05}
{"train_loss": 0.18195582926273346, "global_step": 73641, "epoch": 827, "lr": 8.30718983680772e-05}
{"train_loss": 0.2035202980041504, "global_step": 73642, "epoch": 827, "lr": 8.307146356347301e-05}
{"train_loss": 0.21906979382038116, "global_step": 73643, "epoch": 827, "lr": 8.307102875442277e-05}
{"train_loss": 0.17152273654937744, "global_step": 73644, "epoch": 827, "lr": 8.307059394092651e-05}
{"train_loss": 0.21249504387378693, "global_step": 73645, "epoch": 827, "lr": 8.307015912298432e-05}
{"train_loss": 0.1971796602010727, "global_step": 73646, "epoch": 827, "lr": 8.306972430059628e-05}
{"train_loss": 0.26685377955436707, "global_step": 73647, "epoch": 827, "lr": 8.306928947376238e-05}
{"train_loss": 0.31744328141212463, "global_step": 73648, "epoch": 827, "lr": 8.306885464248273e-05}
{"train_loss": 0.16741319000720978, "global_step": 73649, "epoch": 827, "lr": 8.306841980675739e-05}
{"train_loss": 0.14799559116363525, "global_step": 73650, "epoch": 827, "lr": 8.306798496658638e-05}
{"train_loss": 0.1702074110507965, "global_step": 73651, "epoch": 827, "lr": 8.306755012196981e-05}
{"train_loss": 0.1360778957605362, "global_step": 73652, "epoch": 827, "lr": 8.30671152729077e-05}
{"train_loss": 0.19443382322788239, "global_step": 73653, "epoch": 827, "lr": 8.306668041940011e-05}
{"train_loss": 0.17306950688362122, "global_step": 73654, "epoch": 827, "lr": 8.30662455614471e-05}
{"train_loss": 0.12424661964178085, "global_step": 73655, "epoch": 827, "lr": 8.306581069904876e-05}
{"train_loss": 0.1542307734489441, "global_step": 73656, "epoch": 827, "lr": 8.306537583220511e-05}
{"train_loss": 0.19199146330356598, "global_step": 73657, "epoch": 827, "lr": 8.306494096091624e-05}
{"train_loss": 0.22157737612724304, "global_step": 73658, "epoch": 827, "lr": 8.306450608518217e-05}
{"train_loss": 0.2568604648113251, "global_step": 73659, "epoch": 827, "lr": 8.3064071205003e-05}
{"train_loss": 0.18738611042499542, "global_step": 73660, "epoch": 827, "lr": 8.306363632037875e-05}
{"train_loss": 0.19152617454528809, "global_step": 73661, "epoch": 827, "lr": 8.306320143130951e-05}
{"train_loss": 0.17633669078350067, "global_step": 73662, "epoch": 827, "lr": 8.306276653779533e-05}
{"train_loss": 0.1456412672996521, "global_step": 73663, "epoch": 827, "lr": 8.306233163983624e-05}
{"train_loss": 0.1733427494764328, "global_step": 73664, "epoch": 827, "lr": 8.306189673743234e-05}
{"train_loss": 0.22910349071025848, "global_step": 73665, "epoch": 827, "lr": 8.306146183058367e-05}
{"train_loss": 0.17560705542564392, "global_step": 73666, "epoch": 827, "lr": 8.306102691929028e-05}
{"train_loss": 0.19029518961906433, "global_step": 73667, "epoch": 827, "lr": 8.306059200355225e-05}
{"train_loss": 0.19054622948169708, "global_step": 73668, "epoch": 827, "lr": 8.306015708336962e-05}
{"train_loss": 0.25185826420783997, "global_step": 73669, "epoch": 827, "lr": 8.305972215874247e-05}
{"train_loss": 0.20652584731578827, "global_step": 73670, "epoch": 827, "lr": 8.305928722967083e-05}
{"train_loss": 0.16287454962730408, "global_step": 73671, "epoch": 827, "lr": 8.305885229615477e-05}
{"train_loss": 0.16265752911567688, "global_step": 73672, "epoch": 827, "lr": 8.305841735819435e-05}
{"train_loss": 0.1678108423948288, "global_step": 73673, "epoch": 827, "lr": 8.305798241578964e-05}
{"train_loss": 0.29308924078941345, "global_step": 73674, "epoch": 827, "lr": 8.305754746894067e-05}
{"train_loss": 0.1758424937725067, "global_step": 73675, "epoch": 827, "lr": 8.305711251764752e-05}
{"train_loss": 0.12020321935415268, "global_step": 73676, "epoch": 827, "lr": 8.305667756191025e-05}
{"train_loss": 0.21163208782672882, "global_step": 73677, "epoch": 827, "lr": 8.30562426017289e-05}
{"train_loss": 0.29536935687065125, "global_step": 73678, "epoch": 827, "lr": 8.305580763710357e-05}
{"train_loss": 0.17519110441207886, "global_step": 73679, "epoch": 827, "lr": 8.305537266803427e-05}
{"train_loss": 0.17539389431476593, "global_step": 73680, "epoch": 827, "lr": 8.305493769452106e-05}
{"train_loss": 0.22879251837730408, "global_step": 73681, "epoch": 827, "lr": 8.305450271656404e-05}
{"train_loss": 0.1533961445093155, "global_step": 73682, "epoch": 827, "lr": 8.305406773416324e-05}
{"train_loss": 0.18051934242248535, "global_step": 73683, "epoch": 827, "lr": 8.305363274731873e-05}
{"train_loss": 0.18944832682609558, "global_step": 73684, "epoch": 827, "lr": 8.305319775603055e-05}
{"train_loss": 0.3118554949760437, "global_step": 73685, "epoch": 827, "lr": 8.305276276029879e-05}
{"train_loss": 0.22848740220069885, "global_step": 73686, "epoch": 827, "lr": 8.305232776012346e-05}
{"train_loss": 0.16519363224506378, "global_step": 73687, "epoch": 827, "lr": 8.305189275550467e-05}
{"train_loss": 0.23713262379169464, "global_step": 73688, "epoch": 827, "lr": 8.305145774644245e-05}
{"train_loss": 0.17128081619739532, "global_step": 73689, "epoch": 827, "lr": 8.305102273293687e-05}
{"train_loss": 0.2339809685945511, "global_step": 73690, "epoch": 827, "lr": 8.305058771498798e-05}
{"train_loss": 0.20940545050615675, "global_step": 73691, "epoch": 827, "lr": 8.305015269259583e-05, "val_loss": 3.8588836193084717}
{"train_loss": 0.16779783368110657, "global_step": 73692, "epoch": 828, "lr": 8.30497176657605e-05}
{"train_loss": 0.23267458379268646, "global_step": 73693, "epoch": 828, "lr": 8.304928263448203e-05}
{"train_loss": 0.18358486890792847, "global_step": 73694, "epoch": 828, "lr": 8.304884759876049e-05}
{"train_loss": 0.23006673157215118, "global_step": 73695, "epoch": 828, "lr": 8.304841255859595e-05}
{"train_loss": 0.20091673731803894, "global_step": 73696, "epoch": 828, "lr": 8.304797751398843e-05}
{"train_loss": 0.24967385828495026, "global_step": 73697, "epoch": 828, "lr": 8.304754246493802e-05}
{"train_loss": 0.19847799837589264, "global_step": 73698, "epoch": 828, "lr": 8.304710741144476e-05}
{"train_loss": 0.21159657835960388, "global_step": 73699, "epoch": 828, "lr": 8.304667235350874e-05}
{"train_loss": 0.18834088742733002, "global_step": 73700, "epoch": 828, "lr": 8.304623729112999e-05}
{"train_loss": 0.20330189168453217, "global_step": 73701, "epoch": 828, "lr": 8.304580222430858e-05}
{"train_loss": 0.2547353208065033, "global_step": 73702, "epoch": 828, "lr": 8.304536715304457e-05}
{"train_loss": 0.2625039219856262, "global_step": 73703, "epoch": 828, "lr": 8.3044932077338e-05}
{"train_loss": 0.16384294629096985, "global_step": 73704, "epoch": 828, "lr": 8.304449699718894e-05}
{"train_loss": 0.23364010453224182, "global_step": 73705, "epoch": 828, "lr": 8.304406191259746e-05}
{"train_loss": 0.22927725315093994, "global_step": 73706, "epoch": 828, "lr": 8.304362682356362e-05}
{"train_loss": 0.22295236587524414, "global_step": 73707, "epoch": 828, "lr": 8.304319173008744e-05}
{"train_loss": 0.17895860970020294, "global_step": 73708, "epoch": 828, "lr": 8.304275663216902e-05}
{"train_loss": 0.3066038489341736, "global_step": 73709, "epoch": 828, "lr": 8.30423215298084e-05}
{"train_loss": 0.2885315418243408, "global_step": 73710, "epoch": 828, "lr": 8.304188642300566e-05}
{"train_loss": 0.2284369170665741, "global_step": 73711, "epoch": 828, "lr": 8.304145131176082e-05}
{"train_loss": 0.23679637908935547, "global_step": 73712, "epoch": 828, "lr": 8.304101619607396e-05}
{"train_loss": 0.1974785327911377, "global_step": 73713, "epoch": 828, "lr": 8.304058107594515e-05}
{"train_loss": 0.1838502734899521, "global_step": 73714, "epoch": 828, "lr": 8.304014595137443e-05}
{"train_loss": 0.18602050840854645, "global_step": 73715, "epoch": 828, "lr": 8.303971082236186e-05}
{"train_loss": 0.19598643481731415, "global_step": 73716, "epoch": 828, "lr": 8.30392756889075e-05}
{"train_loss": 0.24099519848823547, "global_step": 73717, "epoch": 828, "lr": 8.303884055101143e-05}
{"train_loss": 0.20031175017356873, "global_step": 73718, "epoch": 828, "lr": 8.303840540867367e-05}
{"train_loss": 0.15031975507736206, "global_step": 73719, "epoch": 828, "lr": 8.303797026189432e-05}
{"train_loss": 0.29629579186439514, "global_step": 73720, "epoch": 828, "lr": 8.303753511067341e-05}
{"train_loss": 0.17058828473091125, "global_step": 73721, "epoch": 828, "lr": 8.3037099955011e-05}
{"train_loss": 0.23075354099273682, "global_step": 73722, "epoch": 828, "lr": 8.303666479490716e-05}
{"train_loss": 0.1810273379087448, "global_step": 73723, "epoch": 828, "lr": 8.303622963036194e-05}
{"train_loss": 0.12728315591812134, "global_step": 73724, "epoch": 828, "lr": 8.303579446137541e-05}
{"train_loss": 0.18612930178642273, "global_step": 73725, "epoch": 828, "lr": 8.303535928794761e-05}
{"train_loss": 0.1825706511735916, "global_step": 73726, "epoch": 828, "lr": 8.30349241100786e-05}
{"train_loss": 0.17695997655391693, "global_step": 73727, "epoch": 828, "lr": 8.303448892776848e-05}
{"train_loss": 0.28932973742485046, "global_step": 73728, "epoch": 828, "lr": 8.303405374101724e-05}
{"train_loss": 0.25719550251960754, "global_step": 73729, "epoch": 828, "lr": 8.303361854982498e-05}
{"train_loss": 0.1730620563030243, "global_step": 73730, "epoch": 828, "lr": 8.303318335419178e-05}
{"train_loss": 0.18838079273700714, "global_step": 73731, "epoch": 828, "lr": 8.303274815411765e-05}
{"train_loss": 0.25232192873954773, "global_step": 73732, "epoch": 828, "lr": 8.303231294960267e-05}
{"train_loss": 0.23913036286830902, "global_step": 73733, "epoch": 828, "lr": 8.30318777406469e-05}
{"train_loss": 0.1988736391067505, "global_step": 73734, "epoch": 828, "lr": 8.30314425272504e-05}
{"train_loss": 0.1942470520734787, "global_step": 73735, "epoch": 828, "lr": 8.303100730941322e-05}
{"train_loss": 0.2913362383842468, "global_step": 73736, "epoch": 828, "lr": 8.303057208713543e-05}
{"train_loss": 0.30412760376930237, "global_step": 73737, "epoch": 828, "lr": 8.303013686041708e-05}
{"train_loss": 0.3585233986377716, "global_step": 73738, "epoch": 828, "lr": 8.302970162925821e-05}
{"train_loss": 0.1369369775056839, "global_step": 73739, "epoch": 828, "lr": 8.302926639365893e-05}
{"train_loss": 0.15511927008628845, "global_step": 73740, "epoch": 828, "lr": 8.302883115361926e-05}
{"train_loss": 0.2385212630033493, "global_step": 73741, "epoch": 828, "lr": 8.302839590913925e-05}
{"train_loss": 0.2890666425228119, "global_step": 73742, "epoch": 828, "lr": 8.302796066021898e-05}
{"train_loss": 0.15349401533603668, "global_step": 73743, "epoch": 828, "lr": 8.302752540685851e-05}
{"train_loss": 0.3347569406032562, "global_step": 73744, "epoch": 828, "lr": 8.30270901490579e-05}
{"train_loss": 0.26357540488243103, "global_step": 73745, "epoch": 828, "lr": 8.30266548868172e-05}
{"train_loss": 0.26715537905693054, "global_step": 73746, "epoch": 828, "lr": 8.302621962013646e-05}
{"train_loss": 0.20142750442028046, "global_step": 73747, "epoch": 828, "lr": 8.302578434901575e-05}
{"train_loss": 0.1944117248058319, "global_step": 73748, "epoch": 828, "lr": 8.302534907345512e-05}
{"train_loss": 0.21520783007144928, "global_step": 73749, "epoch": 828, "lr": 8.302491379345462e-05}
{"train_loss": 0.20116068422794342, "global_step": 73750, "epoch": 828, "lr": 8.302447850901436e-05}
{"train_loss": 0.19981178641319275, "global_step": 73751, "epoch": 828, "lr": 8.302404322013434e-05}
{"train_loss": 0.2139936238527298, "global_step": 73752, "epoch": 828, "lr": 8.302360792681464e-05}
{"train_loss": 0.19117818772792816, "global_step": 73753, "epoch": 828, "lr": 8.302317262905532e-05}
{"train_loss": 0.2579679787158966, "global_step": 73754, "epoch": 828, "lr": 8.302273732685644e-05}
{"train_loss": 0.1981719732284546, "global_step": 73755, "epoch": 828, "lr": 8.302230202021804e-05}
{"train_loss": 0.1922195702791214, "global_step": 73756, "epoch": 828, "lr": 8.302186670914021e-05}
{"train_loss": 0.2582103908061981, "global_step": 73757, "epoch": 828, "lr": 8.302143139362299e-05}
{"train_loss": 0.16568434238433838, "global_step": 73758, "epoch": 828, "lr": 8.302099607366645e-05}
{"train_loss": 0.23365114629268646, "global_step": 73759, "epoch": 828, "lr": 8.302056074927063e-05}
{"train_loss": 0.25996220111846924, "global_step": 73760, "epoch": 828, "lr": 8.30201254204356e-05}
{"train_loss": 0.30866608023643494, "global_step": 73761, "epoch": 828, "lr": 8.30196900871614e-05}
{"train_loss": 0.21736198663711548, "global_step": 73762, "epoch": 828, "lr": 8.301925474944813e-05}
{"train_loss": 0.2652786076068878, "global_step": 73763, "epoch": 828, "lr": 8.301881940729581e-05}
{"train_loss": 0.278910368680954, "global_step": 73764, "epoch": 828, "lr": 8.301838406070451e-05}
{"train_loss": 0.20776131749153137, "global_step": 73765, "epoch": 828, "lr": 8.301794870967429e-05}
{"train_loss": 0.17950741946697235, "global_step": 73766, "epoch": 828, "lr": 8.301751335420522e-05}
{"train_loss": 0.18096591532230377, "global_step": 73767, "epoch": 828, "lr": 8.301707799429735e-05}
{"train_loss": 0.2057604044675827, "global_step": 73768, "epoch": 828, "lr": 8.301664262995072e-05}
{"train_loss": 0.14930009841918945, "global_step": 73769, "epoch": 828, "lr": 8.301620726116543e-05}
{"train_loss": 0.3476255238056183, "global_step": 73770, "epoch": 828, "lr": 8.30157718879415e-05}
{"train_loss": 0.2725011110305786, "global_step": 73771, "epoch": 828, "lr": 8.301533651027898e-05}
{"train_loss": 0.18398034572601318, "global_step": 73772, "epoch": 828, "lr": 8.301490112817798e-05}
{"train_loss": 0.24274995923042297, "global_step": 73773, "epoch": 828, "lr": 8.301446574163851e-05}
{"train_loss": 0.2833617925643921, "global_step": 73774, "epoch": 828, "lr": 8.301403035066067e-05}
{"train_loss": 0.27599596977233887, "global_step": 73775, "epoch": 828, "lr": 8.301359495524448e-05}
{"train_loss": 0.2169751226902008, "global_step": 73776, "epoch": 828, "lr": 8.301315955539002e-05}
{"train_loss": 0.2362220734357834, "global_step": 73777, "epoch": 828, "lr": 8.301272415109733e-05}
{"train_loss": 0.19549891352653503, "global_step": 73778, "epoch": 828, "lr": 8.30122887423665e-05}
{"train_loss": 0.22015152871608734, "global_step": 73779, "epoch": 828, "lr": 8.301185332919757e-05}
{"train_loss": 0.22302820307485174, "global_step": 73780, "epoch": 828, "lr": 8.301141791159057e-05, "val_loss": 3.7761802673339844}
{"train_loss": 0.2837292551994324, "global_step": 73781, "epoch": 829, "lr": 8.301098248954562e-05}
{"train_loss": 0.27757978439331055, "global_step": 73782, "epoch": 829, "lr": 8.301054706306273e-05}
{"train_loss": 0.23747099936008453, "global_step": 73783, "epoch": 829, "lr": 8.301011163214197e-05}
{"train_loss": 0.2134718894958496, "global_step": 73784, "epoch": 829, "lr": 8.300967619678341e-05}
{"train_loss": 0.20232975482940674, "global_step": 73785, "epoch": 829, "lr": 8.30092407569871e-05}
{"train_loss": 0.22556625306606293, "global_step": 73786, "epoch": 829, "lr": 8.300880531275312e-05}
{"train_loss": 0.1685074269771576, "global_step": 73787, "epoch": 829, "lr": 8.300836986408149e-05}
{"train_loss": 0.22260782122612, "global_step": 73788, "epoch": 829, "lr": 8.300793441097228e-05}
{"train_loss": 0.2382926493883133, "global_step": 73789, "epoch": 829, "lr": 8.300749895342556e-05}
{"train_loss": 0.16488872468471527, "global_step": 73790, "epoch": 829, "lr": 8.300706349144138e-05}
{"train_loss": 0.35016342997550964, "global_step": 73791, "epoch": 829, "lr": 8.300662802501982e-05}
{"train_loss": 0.20487284660339355, "global_step": 73792, "epoch": 829, "lr": 8.300619255416091e-05}
{"train_loss": 0.2006642371416092, "global_step": 73793, "epoch": 829, "lr": 8.300575707886471e-05}
{"train_loss": 0.26865363121032715, "global_step": 73794, "epoch": 829, "lr": 8.30053215991313e-05}
{"train_loss": 0.24744626879692078, "global_step": 73795, "epoch": 829, "lr": 8.300488611496074e-05}
{"train_loss": 0.1904076784849167, "global_step": 73796, "epoch": 829, "lr": 8.300445062635305e-05}
{"train_loss": 0.20677967369556427, "global_step": 73797, "epoch": 829, "lr": 8.300401513330834e-05}
{"train_loss": 0.34934258460998535, "global_step": 73798, "epoch": 829, "lr": 8.30035796358266e-05}
{"train_loss": 0.24920421838760376, "global_step": 73799, "epoch": 829, "lr": 8.300314413390798e-05}
{"train_loss": 0.1732119917869568, "global_step": 73800, "epoch": 829, "lr": 8.300270862755246e-05}
{"train_loss": 0.24484005570411682, "global_step": 73801, "epoch": 829, "lr": 8.300227311676013e-05}
{"train_loss": 0.24514487385749817, "global_step": 73802, "epoch": 829, "lr": 8.300183760153106e-05}
{"train_loss": 0.2266865074634552, "global_step": 73803, "epoch": 829, "lr": 8.300140208186528e-05}
{"train_loss": 0.18429745733737946, "global_step": 73804, "epoch": 829, "lr": 8.300096655776289e-05}
{"train_loss": 0.18587681651115417, "global_step": 73805, "epoch": 829, "lr": 8.300053102922389e-05}
{"train_loss": 0.207171231508255, "global_step": 73806, "epoch": 829, "lr": 8.300009549624839e-05}
{"train_loss": 0.123774953186512, "global_step": 73807, "epoch": 829, "lr": 8.299965995883641e-05}
{"train_loss": 0.21123559772968292, "global_step": 73808, "epoch": 829, "lr": 8.299922441698804e-05}
{"train_loss": 0.17832277715206146, "global_step": 73809, "epoch": 829, "lr": 8.299878887070333e-05}
{"train_loss": 0.1997709572315216, "global_step": 73810, "epoch": 829, "lr": 8.299835331998234e-05}
{"train_loss": 0.1818545162677765, "global_step": 73811, "epoch": 829, "lr": 8.29979177648251e-05}
{"train_loss": 0.23846879601478577, "global_step": 73812, "epoch": 829, "lr": 8.299748220523171e-05}
{"train_loss": 0.28658872842788696, "global_step": 73813, "epoch": 829, "lr": 8.299704664120223e-05}
{"train_loss": 0.24525704979896545, "global_step": 73814, "epoch": 829, "lr": 8.299661107273666e-05}
{"train_loss": 0.17633138597011566, "global_step": 73815, "epoch": 829, "lr": 8.299617549983513e-05}
{"train_loss": 0.3084081709384918, "global_step": 73816, "epoch": 829, "lr": 8.299573992249765e-05}
{"train_loss": 0.20797429978847504, "global_step": 73817, "epoch": 829, "lr": 8.29953043407243e-05}
{"train_loss": 0.1992780715227127, "global_step": 73818, "epoch": 829, "lr": 8.299486875451512e-05}
{"train_loss": 0.16476191580295563, "global_step": 73819, "epoch": 829, "lr": 8.29944331638702e-05}
{"train_loss": 0.1874184012413025, "global_step": 73820, "epoch": 829, "lr": 8.299399756878957e-05}
{"train_loss": 0.14432255923748016, "global_step": 73821, "epoch": 829, "lr": 8.29935619692733e-05}
{"train_loss": 0.1691380888223648, "global_step": 73822, "epoch": 829, "lr": 8.299312636532146e-05}
{"train_loss": 0.24231722950935364, "global_step": 73823, "epoch": 829, "lr": 8.29926907569341e-05}
{"train_loss": 0.3003191649913788, "global_step": 73824, "epoch": 829, "lr": 8.299225514411125e-05}
{"train_loss": 0.22613385319709778, "global_step": 73825, "epoch": 829, "lr": 8.2991819526853e-05}
{"train_loss": 0.29912930727005005, "global_step": 73826, "epoch": 829, "lr": 8.29913839051594e-05}
{"train_loss": 0.23430699110031128, "global_step": 73827, "epoch": 829, "lr": 8.299094827903053e-05}
{"train_loss": 0.1882929652929306, "global_step": 73828, "epoch": 829, "lr": 8.29905126484664e-05}
{"train_loss": 0.21729789674282074, "global_step": 73829, "epoch": 829, "lr": 8.299007701346713e-05}
{"train_loss": 0.21931833028793335, "global_step": 73830, "epoch": 829, "lr": 8.298964137403273e-05}
{"train_loss": 0.16669583320617676, "global_step": 73831, "epoch": 829, "lr": 8.298920573016328e-05}
{"train_loss": 0.1462133824825287, "global_step": 73832, "epoch": 829, "lr": 8.298877008185882e-05}
{"train_loss": 0.25587400794029236, "global_step": 73833, "epoch": 829, "lr": 8.298833442911945e-05}
{"train_loss": 0.21328510344028473, "global_step": 73834, "epoch": 829, "lr": 8.298789877194517e-05}
{"train_loss": 0.2526404559612274, "global_step": 73835, "epoch": 829, "lr": 8.298746311033607e-05}
{"train_loss": 0.15701529383659363, "global_step": 73836, "epoch": 829, "lr": 8.298702744429224e-05}
{"train_loss": 0.2036818563938141, "global_step": 73837, "epoch": 829, "lr": 8.298659177381368e-05}
{"train_loss": 0.16073139011859894, "global_step": 73838, "epoch": 829, "lr": 8.298615609890047e-05}
{"train_loss": 0.25005707144737244, "global_step": 73839, "epoch": 829, "lr": 8.298572041955269e-05}
{"train_loss": 0.15470242500305176, "global_step": 73840, "epoch": 829, "lr": 8.298528473577037e-05}
{"train_loss": 0.2609032094478607, "global_step": 73841, "epoch": 829, "lr": 8.298484904755358e-05}
{"train_loss": 0.29494360089302063, "global_step": 73842, "epoch": 829, "lr": 8.29844133549024e-05}
{"train_loss": 0.23097941279411316, "global_step": 73843, "epoch": 829, "lr": 8.298397765781682e-05}
{"train_loss": 0.23860685527324677, "global_step": 73844, "epoch": 829, "lr": 8.298354195629698e-05}
{"train_loss": 0.2007102519273758, "global_step": 73845, "epoch": 829, "lr": 8.298310625034292e-05}
{"train_loss": 0.18377842009067535, "global_step": 73846, "epoch": 829, "lr": 8.298267053995466e-05}
{"train_loss": 0.20068497955799103, "global_step": 73847, "epoch": 829, "lr": 8.298223482513229e-05}
{"train_loss": 0.2647342085838318, "global_step": 73848, "epoch": 829, "lr": 8.298179910587586e-05}
{"train_loss": 0.2985207736492157, "global_step": 73849, "epoch": 829, "lr": 8.298136338218542e-05}
{"train_loss": 0.1358330398797989, "global_step": 73850, "epoch": 829, "lr": 8.298092765406105e-05}
{"train_loss": 0.18958348035812378, "global_step": 73851, "epoch": 829, "lr": 8.29804919215028e-05}
{"train_loss": 0.21805687248706818, "global_step": 73852, "epoch": 829, "lr": 8.29800561845107e-05}
{"train_loss": 0.18007004261016846, "global_step": 73853, "epoch": 829, "lr": 8.297962044308485e-05}
{"train_loss": 0.2019226998090744, "global_step": 73854, "epoch": 829, "lr": 8.29791846972253e-05}
{"train_loss": 0.17365014553070068, "global_step": 73855, "epoch": 829, "lr": 8.297874894693209e-05}
{"train_loss": 0.15023992955684662, "global_step": 73856, "epoch": 829, "lr": 8.29783131922053e-05}
{"train_loss": 0.19119368493556976, "global_step": 73857, "epoch": 829, "lr": 8.297787743304497e-05}
{"train_loss": 0.18918082118034363, "global_step": 73858, "epoch": 829, "lr": 8.297744166945116e-05}
{"train_loss": 0.15506723523139954, "global_step": 73859, "epoch": 829, "lr": 8.297700590142394e-05}
{"train_loss": 0.22260655462741852, "global_step": 73860, "epoch": 829, "lr": 8.297657012896335e-05}
{"train_loss": 0.1860761195421219, "global_step": 73861, "epoch": 829, "lr": 8.29761343520695e-05}
{"train_loss": 0.2759549617767334, "global_step": 73862, "epoch": 829, "lr": 8.297569857074237e-05}
{"train_loss": 0.1758519560098648, "global_step": 73863, "epoch": 829, "lr": 8.297526278498209e-05}
{"train_loss": 0.2171202301979065, "global_step": 73864, "epoch": 829, "lr": 8.297482699478867e-05}
{"train_loss": 0.2197292447090149, "global_step": 73865, "epoch": 829, "lr": 8.297439120016218e-05}
{"train_loss": 0.2436169683933258, "global_step": 73866, "epoch": 829, "lr": 8.29739554011027e-05}
{"train_loss": 0.30250611901283264, "global_step": 73867, "epoch": 829, "lr": 8.297351959761027e-05}
{"train_loss": 0.24240736663341522, "global_step": 73868, "epoch": 829, "lr": 8.297308378968495e-05}
{"train_loss": 0.2171026531062769, "global_step": 73869, "epoch": 829, "lr": 8.29726479773268e-05, "val_loss": 3.8856523036956787}
{"train_loss": 0.12946587800979614, "global_step": 73870, "epoch": 830, "lr": 8.297221216053589e-05}
{"train_loss": 0.17707709968090057, "global_step": 73871, "epoch": 830, "lr": 8.297177633931226e-05}
{"train_loss": 0.25332707166671753, "global_step": 73872, "epoch": 830, "lr": 8.297134051365597e-05}
{"train_loss": 0.24183538556098938, "global_step": 73873, "epoch": 830, "lr": 8.29709046835671e-05}
{"train_loss": 0.17152352631092072, "global_step": 73874, "epoch": 830, "lr": 8.297046884904567e-05}
{"train_loss": 0.25186583399772644, "global_step": 73875, "epoch": 830, "lr": 8.297003301009178e-05}
{"train_loss": 0.2612776458263397, "global_step": 73876, "epoch": 830, "lr": 8.296959716670547e-05}
{"train_loss": 0.19800391793251038, "global_step": 73877, "epoch": 830, "lr": 8.29691613188868e-05}
{"train_loss": 0.2997276484966278, "global_step": 73878, "epoch": 830, "lr": 8.296872546663583e-05}
{"train_loss": 0.1969987154006958, "global_step": 73879, "epoch": 830, "lr": 8.29682896099526e-05}
{"train_loss": 0.18079407513141632, "global_step": 73880, "epoch": 830, "lr": 8.29678537488372e-05}
{"train_loss": 0.3022901117801666, "global_step": 73881, "epoch": 830, "lr": 8.296741788328968e-05}
{"train_loss": 0.2283143252134323, "global_step": 73882, "epoch": 830, "lr": 8.296698201331008e-05}
{"train_loss": 0.1877146065235138, "global_step": 73883, "epoch": 830, "lr": 8.296654613889847e-05}
{"train_loss": 0.23542916774749756, "global_step": 73884, "epoch": 830, "lr": 8.296611026005492e-05}
{"train_loss": 0.17990708351135254, "global_step": 73885, "epoch": 830, "lr": 8.296567437677947e-05}
{"train_loss": 0.1834372878074646, "global_step": 73886, "epoch": 830, "lr": 8.296523848907219e-05}
{"train_loss": 0.2700417637825012, "global_step": 73887, "epoch": 830, "lr": 8.296480259693313e-05}
{"train_loss": 0.21104319393634796, "global_step": 73888, "epoch": 830, "lr": 8.296436670036236e-05}
{"train_loss": 0.3345836400985718, "global_step": 73889, "epoch": 830, "lr": 8.296393079935994e-05}
{"train_loss": 0.19599060714244843, "global_step": 73890, "epoch": 830, "lr": 8.296349489392591e-05}
{"train_loss": 0.1994716227054596, "global_step": 73891, "epoch": 830, "lr": 8.296305898406034e-05}
{"train_loss": 0.259611576795578, "global_step": 73892, "epoch": 830, "lr": 8.29626230697633e-05}
{"train_loss": 0.15422460436820984, "global_step": 73893, "epoch": 830, "lr": 8.296218715103482e-05}
{"train_loss": 0.2255556583404541, "global_step": 73894, "epoch": 830, "lr": 8.2961751227875e-05}
{"train_loss": 0.280306339263916, "global_step": 73895, "epoch": 830, "lr": 8.296131530028385e-05}
{"train_loss": 0.1280607283115387, "global_step": 73896, "epoch": 830, "lr": 8.296087936826145e-05}
{"train_loss": 0.17067617177963257, "global_step": 73897, "epoch": 830, "lr": 8.296044343180789e-05}
{"train_loss": 0.23930317163467407, "global_step": 73898, "epoch": 830, "lr": 8.296000749092317e-05}
{"train_loss": 0.11639528721570969, "global_step": 73899, "epoch": 830, "lr": 8.295957154560738e-05}
{"train_loss": 0.16300499439239502, "global_step": 73900, "epoch": 830, "lr": 8.295913559586059e-05}
{"train_loss": 0.2520030736923218, "global_step": 73901, "epoch": 830, "lr": 8.295869964168286e-05}
{"train_loss": 0.20705214142799377, "global_step": 73902, "epoch": 830, "lr": 8.295826368307422e-05}
{"train_loss": 0.16339775919914246, "global_step": 73903, "epoch": 830, "lr": 8.295782772003474e-05}
{"train_loss": 0.17862790822982788, "global_step": 73904, "epoch": 830, "lr": 8.295739175256448e-05}
{"train_loss": 0.13189750909805298, "global_step": 73905, "epoch": 830, "lr": 8.295695578066351e-05}
{"train_loss": 0.2316485494375229, "global_step": 73906, "epoch": 830, "lr": 8.295651980433187e-05}
{"train_loss": 0.19052205979824066, "global_step": 73907, "epoch": 830, "lr": 8.295608382356964e-05}
{"train_loss": 0.23829308152198792, "global_step": 73908, "epoch": 830, "lr": 8.295564783837686e-05}
{"train_loss": 0.28424128890037537, "global_step": 73909, "epoch": 830, "lr": 8.29552118487536e-05}
{"train_loss": 0.22132906317710876, "global_step": 73910, "epoch": 830, "lr": 8.295477585469988e-05}
{"train_loss": 0.1626463532447815, "global_step": 73911, "epoch": 830, "lr": 8.295433985621583e-05}
{"train_loss": 0.20766755938529968, "global_step": 73912, "epoch": 830, "lr": 8.295390385330147e-05}
{"train_loss": 0.16717280447483063, "global_step": 73913, "epoch": 830, "lr": 8.295346784595683e-05}
{"train_loss": 0.2870998978614807, "global_step": 73914, "epoch": 830, "lr": 8.295303183418201e-05}
{"train_loss": 0.15754196047782898, "global_step": 73915, "epoch": 830, "lr": 8.295259581797707e-05}
{"train_loss": 0.2611830532550812, "global_step": 73916, "epoch": 830, "lr": 8.295215979734205e-05}
{"train_loss": 0.25693416595458984, "global_step": 73917, "epoch": 830, "lr": 8.2951723772277e-05}
{"train_loss": 0.13269758224487305, "global_step": 73918, "epoch": 830, "lr": 8.2951287742782e-05}
{"train_loss": 0.2598327100276947, "global_step": 73919, "epoch": 830, "lr": 8.295085170885712e-05}
{"train_loss": 0.18638692796230316, "global_step": 73920, "epoch": 830, "lr": 8.295041567050238e-05}
{"train_loss": 0.20438532531261444, "global_step": 73921, "epoch": 830, "lr": 8.294997962771786e-05}
{"train_loss": 0.15811295807361603, "global_step": 73922, "epoch": 830, "lr": 8.294954358050362e-05}
{"train_loss": 0.17760439217090607, "global_step": 73923, "epoch": 830, "lr": 8.294910752885971e-05}
{"train_loss": 0.14461052417755127, "global_step": 73924, "epoch": 830, "lr": 8.294867147278619e-05}
{"train_loss": 0.20648998022079468, "global_step": 73925, "epoch": 830, "lr": 8.294823541228315e-05}
{"train_loss": 0.18676258623600006, "global_step": 73926, "epoch": 830, "lr": 8.29477993473506e-05}
{"train_loss": 0.211778923869133, "global_step": 73927, "epoch": 830, "lr": 8.294736327798863e-05}
{"train_loss": 0.22524049878120422, "global_step": 73928, "epoch": 830, "lr": 8.294692720419728e-05}
{"train_loss": 0.16271954774856567, "global_step": 73929, "epoch": 830, "lr": 8.294649112597661e-05}
{"train_loss": 0.18069447576999664, "global_step": 73930, "epoch": 830, "lr": 8.294605504332671e-05}
{"train_loss": 0.20199237763881683, "global_step": 73931, "epoch": 830, "lr": 8.29456189562476e-05}
{"train_loss": 0.18353769183158875, "global_step": 73932, "epoch": 830, "lr": 8.294518286473933e-05}
{"train_loss": 0.21368089318275452, "global_step": 73933, "epoch": 830, "lr": 8.294474676880201e-05}
{"train_loss": 0.24989651143550873, "global_step": 73934, "epoch": 830, "lr": 8.294431066843566e-05}
{"train_loss": 0.1922721266746521, "global_step": 73935, "epoch": 830, "lr": 8.294387456364036e-05}
{"train_loss": 0.15680451691150665, "global_step": 73936, "epoch": 830, "lr": 8.294343845441615e-05}
{"train_loss": 0.22352007031440735, "global_step": 73937, "epoch": 830, "lr": 8.29430023407631e-05}
{"train_loss": 0.18563278019428253, "global_step": 73938, "epoch": 830, "lr": 8.294256622268126e-05}
{"train_loss": 0.212273508310318, "global_step": 73939, "epoch": 830, "lr": 8.294213010017069e-05}
{"train_loss": 0.13995561003684998, "global_step": 73940, "epoch": 830, "lr": 8.294169397323146e-05}
{"train_loss": 0.20434889197349548, "global_step": 73941, "epoch": 830, "lr": 8.294125784186361e-05}
{"train_loss": 0.18199384212493896, "global_step": 73942, "epoch": 830, "lr": 8.294082170606723e-05}
{"train_loss": 0.222927987575531, "global_step": 73943, "epoch": 830, "lr": 8.294038556584234e-05}
{"train_loss": 0.1889059990644455, "global_step": 73944, "epoch": 830, "lr": 8.293994942118902e-05}
{"train_loss": 0.27539581060409546, "global_step": 73945, "epoch": 830, "lr": 8.293951327210733e-05}
{"train_loss": 0.1741112619638443, "global_step": 73946, "epoch": 830, "lr": 8.293907711859732e-05}
{"train_loss": 0.31395891308784485, "global_step": 73947, "epoch": 830, "lr": 8.293864096065906e-05}
{"train_loss": 0.19448480010032654, "global_step": 73948, "epoch": 830, "lr": 8.29382047982926e-05}
{"train_loss": 0.2134886384010315, "global_step": 73949, "epoch": 830, "lr": 8.293776863149799e-05}
{"train_loss": 0.259129136800766, "global_step": 73950, "epoch": 830, "lr": 8.29373324602753e-05}
{"train_loss": 0.28928983211517334, "global_step": 73951, "epoch": 830, "lr": 8.29368962846246e-05}
{"train_loss": 0.18203355371952057, "global_step": 73952, "epoch": 830, "lr": 8.293646010454591e-05}
{"train_loss": 0.16885484755039215, "global_step": 73953, "epoch": 830, "lr": 8.293602392003934e-05}
{"train_loss": 0.260524183511734, "global_step": 73954, "epoch": 830, "lr": 8.29355877311049e-05}
{"train_loss": 0.2290613055229187, "global_step": 73955, "epoch": 830, "lr": 8.29351515377427e-05}
{"train_loss": 0.17689453065395355, "global_step": 73956, "epoch": 830, "lr": 8.293471533995275e-05}
{"train_loss": 0.19621050357818604, "global_step": 73957, "epoch": 830, "lr": 8.293427913773513e-05}
{"train_loss": 0.20838439874769596, "global_step": 73958, "epoch": 830, "lr": 8.29338429310899e-05, "val_loss": 3.847585916519165, "train_action_mse_error": 36.3152961730957}
{"train_loss": 0.22809353470802307, "global_step": 73959, "epoch": 831, "lr": 8.293340672001712e-05}
{"train_loss": 0.17524370551109314, "global_step": 73960, "epoch": 831, "lr": 8.293297050451685e-05}
{"train_loss": 0.24447248876094818, "global_step": 73961, "epoch": 831, "lr": 8.293253428458913e-05}
{"train_loss": 0.20972537994384766, "global_step": 73962, "epoch": 831, "lr": 8.293209806023404e-05}
{"train_loss": 0.15433856844902039, "global_step": 73963, "epoch": 831, "lr": 8.293166183145162e-05}
{"train_loss": 0.23679901659488678, "global_step": 73964, "epoch": 831, "lr": 8.293122559824196e-05}
{"train_loss": 0.22764185070991516, "global_step": 73965, "epoch": 831, "lr": 8.29307893606051e-05}
{"train_loss": 0.22771680355072021, "global_step": 73966, "epoch": 831, "lr": 8.293035311854107e-05}
{"train_loss": 0.19731846451759338, "global_step": 73967, "epoch": 831, "lr": 8.292991687204997e-05}
{"train_loss": 0.24978084862232208, "global_step": 73968, "epoch": 831, "lr": 8.292948062113185e-05}
{"train_loss": 0.3188343048095703, "global_step": 73969, "epoch": 831, "lr": 8.292904436578675e-05}
{"train_loss": 0.2010490447282791, "global_step": 73970, "epoch": 831, "lr": 8.292860810601475e-05}
{"train_loss": 0.28213629126548767, "global_step": 73971, "epoch": 831, "lr": 8.29281718418159e-05}
{"train_loss": 0.18333160877227783, "global_step": 73972, "epoch": 831, "lr": 8.292773557319025e-05}
{"train_loss": 0.21476233005523682, "global_step": 73973, "epoch": 831, "lr": 8.292729930013787e-05}
{"train_loss": 0.2046683132648468, "global_step": 73974, "epoch": 831, "lr": 8.292686302265882e-05}
{"train_loss": 0.19033780694007874, "global_step": 73975, "epoch": 831, "lr": 8.292642674075315e-05}
{"train_loss": 0.18221959471702576, "global_step": 73976, "epoch": 831, "lr": 8.292599045442094e-05}
{"train_loss": 0.3088172972202301, "global_step": 73977, "epoch": 831, "lr": 8.292555416366221e-05}
{"train_loss": 0.18944475054740906, "global_step": 73978, "epoch": 831, "lr": 8.292511786847706e-05}
{"train_loss": 0.2213325798511505, "global_step": 73979, "epoch": 831, "lr": 8.292468156886551e-05}
{"train_loss": 0.21219809353351593, "global_step": 73980, "epoch": 831, "lr": 8.292424526482765e-05}
{"train_loss": 0.1982228010892868, "global_step": 73981, "epoch": 831, "lr": 8.292380895636352e-05}
{"train_loss": 0.2045825719833374, "global_step": 73982, "epoch": 831, "lr": 8.292337264347319e-05}
{"train_loss": 0.2337103635072708, "global_step": 73983, "epoch": 831, "lr": 8.292293632615671e-05}
{"train_loss": 0.15778978168964386, "global_step": 73984, "epoch": 831, "lr": 8.292250000441415e-05}
{"train_loss": 0.18942992389202118, "global_step": 73985, "epoch": 831, "lr": 8.292206367824555e-05}
{"train_loss": 0.1596868485212326, "global_step": 73986, "epoch": 831, "lr": 8.292162734765099e-05}
{"train_loss": 0.2539573609828949, "global_step": 73987, "epoch": 831, "lr": 8.292119101263052e-05}
{"train_loss": 0.25707924365997314, "global_step": 73988, "epoch": 831, "lr": 8.292075467318417e-05}
{"train_loss": 0.20980454981327057, "global_step": 73989, "epoch": 831, "lr": 8.292031832931205e-05}
{"train_loss": 0.15051545202732086, "global_step": 73990, "epoch": 831, "lr": 8.291988198101419e-05}
{"train_loss": 0.21555517613887787, "global_step": 73991, "epoch": 831, "lr": 8.291944562829065e-05}
{"train_loss": 0.15332573652267456, "global_step": 73992, "epoch": 831, "lr": 8.291900927114151e-05}
{"train_loss": 0.30461522936820984, "global_step": 73993, "epoch": 831, "lr": 8.29185729095668e-05}
{"train_loss": 0.23225834965705872, "global_step": 73994, "epoch": 831, "lr": 8.291813654356657e-05}
{"train_loss": 0.2683189809322357, "global_step": 73995, "epoch": 831, "lr": 8.291770017314092e-05}
{"train_loss": 0.2310740351676941, "global_step": 73996, "epoch": 831, "lr": 8.291726379828989e-05}
{"train_loss": 0.16302739083766937, "global_step": 73997, "epoch": 831, "lr": 8.291682741901352e-05}
{"train_loss": 0.16128703951835632, "global_step": 73998, "epoch": 831, "lr": 8.291639103531188e-05}
{"train_loss": 0.22622449696063995, "global_step": 73999, "epoch": 831, "lr": 8.291595464718505e-05}
{"train_loss": 0.2074614316225052, "global_step": 74000, "epoch": 831, "lr": 8.291551825463306e-05}
{"train_loss": 0.2248573750257492, "global_step": 74001, "epoch": 831, "lr": 8.291508185765598e-05}
{"train_loss": 0.2251589298248291, "global_step": 74002, "epoch": 831, "lr": 8.291464545625387e-05}
{"train_loss": 0.1872422993183136, "global_step": 74003, "epoch": 831, "lr": 8.291420905042679e-05}
{"train_loss": 0.2723333537578583, "global_step": 74004, "epoch": 831, "lr": 8.291377264017479e-05}
{"train_loss": 0.14763879776000977, "global_step": 74005, "epoch": 831, "lr": 8.291333622549793e-05}
{"train_loss": 0.18297405540943146, "global_step": 74006, "epoch": 831, "lr": 8.29128998063963e-05}
{"train_loss": 0.18166686594486237, "global_step": 74007, "epoch": 831, "lr": 8.29124633828699e-05}
{"train_loss": 0.2455100566148758, "global_step": 74008, "epoch": 831, "lr": 8.291202695491883e-05}
{"train_loss": 0.17662076652050018, "global_step": 74009, "epoch": 831, "lr": 8.291159052254314e-05}
{"train_loss": 0.17558002471923828, "global_step": 74010, "epoch": 831, "lr": 8.29111540857429e-05}
{"train_loss": 0.16270846128463745, "global_step": 74011, "epoch": 831, "lr": 8.291071764451814e-05}
{"train_loss": 0.1646917462348938, "global_step": 74012, "epoch": 831, "lr": 8.291028119886895e-05}
{"train_loss": 0.21953770518302917, "global_step": 74013, "epoch": 831, "lr": 8.290984474879536e-05}
{"train_loss": 0.21623444557189941, "global_step": 74014, "epoch": 831, "lr": 8.290940829429745e-05}
{"train_loss": 0.17631709575653076, "global_step": 74015, "epoch": 831, "lr": 8.290897183537528e-05}
{"train_loss": 0.18832148611545563, "global_step": 74016, "epoch": 831, "lr": 8.290853537202887e-05}
{"train_loss": 0.22032979130744934, "global_step": 74017, "epoch": 831, "lr": 8.290809890425832e-05}
{"train_loss": 0.22699680924415588, "global_step": 74018, "epoch": 831, "lr": 8.29076624320637e-05}
{"train_loss": 0.14841768145561218, "global_step": 74019, "epoch": 831, "lr": 8.290722595544503e-05}
{"train_loss": 0.23640336096286774, "global_step": 74020, "epoch": 831, "lr": 8.290678947440238e-05}
{"train_loss": 0.21170836687088013, "global_step": 74021, "epoch": 831, "lr": 8.290635298893582e-05}
{"train_loss": 0.2565747797489166, "global_step": 74022, "epoch": 831, "lr": 8.29059164990454e-05}
{"train_loss": 0.2123788595199585, "global_step": 74023, "epoch": 831, "lr": 8.290548000473119e-05}
{"train_loss": 0.2451166957616806, "global_step": 74024, "epoch": 831, "lr": 8.290504350599322e-05}
{"train_loss": 0.31271350383758545, "global_step": 74025, "epoch": 831, "lr": 8.290460700283158e-05}
{"train_loss": 0.16688290238380432, "global_step": 74026, "epoch": 831, "lr": 8.29041704952463e-05}
{"train_loss": 0.13219968974590302, "global_step": 74027, "epoch": 831, "lr": 8.290373398323747e-05}
{"train_loss": 0.153024360537529, "global_step": 74028, "epoch": 831, "lr": 8.290329746680514e-05}
{"train_loss": 0.1805759072303772, "global_step": 74029, "epoch": 831, "lr": 8.290286094594935e-05}
{"train_loss": 0.14729365706443787, "global_step": 74030, "epoch": 831, "lr": 8.290242442067019e-05}
{"train_loss": 0.26802465319633484, "global_step": 74031, "epoch": 831, "lr": 8.290198789096768e-05}
{"train_loss": 0.20503593981266022, "global_step": 74032, "epoch": 831, "lr": 8.290155135684191e-05}
{"train_loss": 0.23140396177768707, "global_step": 74033, "epoch": 831, "lr": 8.290111481829292e-05}
{"train_loss": 0.1852240115404129, "global_step": 74034, "epoch": 831, "lr": 8.290067827532079e-05}
{"train_loss": 0.2272549420595169, "global_step": 74035, "epoch": 831, "lr": 8.290024172792555e-05}
{"train_loss": 0.16899263858795166, "global_step": 74036, "epoch": 831, "lr": 8.289980517610729e-05}
{"train_loss": 0.17182539403438568, "global_step": 74037, "epoch": 831, "lr": 8.289936861986603e-05}
{"train_loss": 0.2834267020225525, "global_step": 74038, "epoch": 831, "lr": 8.289893205920186e-05}
{"train_loss": 0.22102725505828857, "global_step": 74039, "epoch": 831, "lr": 8.289849549411483e-05}
{"train_loss": 0.17579473555088043, "global_step": 74040, "epoch": 831, "lr": 8.2898058924605e-05}
{"train_loss": 0.2159375697374344, "global_step": 74041, "epoch": 831, "lr": 8.289762235067242e-05}
{"train_loss": 0.19105859100818634, "global_step": 74042, "epoch": 831, "lr": 8.289718577231718e-05}
{"train_loss": 0.19281502068042755, "global_step": 74043, "epoch": 831, "lr": 8.289674918953929e-05}
{"train_loss": 0.2544190287590027, "global_step": 74044, "epoch": 831, "lr": 8.289631260233883e-05}
{"train_loss": 0.18897607922554016, "global_step": 74045, "epoch": 831, "lr": 8.289587601071587e-05}
{"train_loss": 0.18358412384986877, "global_step": 74046, "epoch": 831, "lr": 8.289543941467046e-05}
{"train_loss": 0.20945301903097818, "global_step": 74047, "epoch": 831, "lr": 8.289500281420266e-05, "val_loss": 3.8178906440734863}
{"train_loss": 0.2555837035179138, "global_step": 74048, "epoch": 832, "lr": 8.289456620931253e-05}
{"train_loss": 0.1796007603406906, "global_step": 74049, "epoch": 832, "lr": 8.289412960000013e-05}
{"train_loss": 0.2131415754556656, "global_step": 74050, "epoch": 832, "lr": 8.289369298626552e-05}
{"train_loss": 0.2092113196849823, "global_step": 74051, "epoch": 832, "lr": 8.289325636810874e-05}
{"train_loss": 0.16325737535953522, "global_step": 74052, "epoch": 832, "lr": 8.289281974552987e-05}
{"train_loss": 0.1791873574256897, "global_step": 74053, "epoch": 832, "lr": 8.289238311852895e-05}
{"train_loss": 0.2563379108905792, "global_step": 74054, "epoch": 832, "lr": 8.289194648710606e-05}
{"train_loss": 0.19202011823654175, "global_step": 74055, "epoch": 832, "lr": 8.289150985126124e-05}
{"train_loss": 0.2635253071784973, "global_step": 74056, "epoch": 832, "lr": 8.289107321099456e-05}
{"train_loss": 0.08766764402389526, "global_step": 74057, "epoch": 832, "lr": 8.28906365663061e-05}
{"train_loss": 0.19619978964328766, "global_step": 74058, "epoch": 832, "lr": 8.289019991719586e-05}
{"train_loss": 0.23954349756240845, "global_step": 74059, "epoch": 832, "lr": 8.288976326366395e-05}
{"train_loss": 0.17821554839611053, "global_step": 74060, "epoch": 832, "lr": 8.288932660571041e-05}
{"train_loss": 0.18338367342948914, "global_step": 74061, "epoch": 832, "lr": 8.28888899433353e-05}
{"train_loss": 0.18907508254051208, "global_step": 74062, "epoch": 832, "lr": 8.288845327653868e-05}
{"train_loss": 0.1503390371799469, "global_step": 74063, "epoch": 832, "lr": 8.288801660532062e-05}
{"train_loss": 0.1850341409444809, "global_step": 74064, "epoch": 832, "lr": 8.288757992968115e-05}
{"train_loss": 0.25585490465164185, "global_step": 74065, "epoch": 832, "lr": 8.288714324962034e-05}
{"train_loss": 0.2743058502674103, "global_step": 74066, "epoch": 832, "lr": 8.288670656513829e-05}
{"train_loss": 0.10973662883043289, "global_step": 74067, "epoch": 832, "lr": 8.288626987623497e-05}
{"train_loss": 0.14581306278705597, "global_step": 74068, "epoch": 832, "lr": 8.288583318291054e-05}
{"train_loss": 0.2080763280391693, "global_step": 74069, "epoch": 832, "lr": 8.2885396485165e-05}
{"train_loss": 0.11048585921525955, "global_step": 74070, "epoch": 832, "lr": 8.28849597829984e-05}
{"train_loss": 0.1518915295600891, "global_step": 74071, "epoch": 832, "lr": 8.288452307641084e-05}
{"train_loss": 0.15535898506641388, "global_step": 74072, "epoch": 832, "lr": 8.288408636540235e-05}
{"train_loss": 0.22589097917079926, "global_step": 74073, "epoch": 832, "lr": 8.288364964997299e-05}
{"train_loss": 0.14826902747154236, "global_step": 74074, "epoch": 832, "lr": 8.288321293012282e-05}
{"train_loss": 0.18794968724250793, "global_step": 74075, "epoch": 832, "lr": 8.28827762058519e-05}
{"train_loss": 0.25278350710868835, "global_step": 74076, "epoch": 832, "lr": 8.288233947716031e-05}
{"train_loss": 0.24294760823249817, "global_step": 74077, "epoch": 832, "lr": 8.288190274404807e-05}
{"train_loss": 0.14912937581539154, "global_step": 74078, "epoch": 832, "lr": 8.288146600651527e-05}
{"train_loss": 0.2491014152765274, "global_step": 74079, "epoch": 832, "lr": 8.288102926456196e-05}
{"train_loss": 0.24805495142936707, "global_step": 74080, "epoch": 832, "lr": 8.288059251818818e-05}
{"train_loss": 0.19395123422145844, "global_step": 74081, "epoch": 832, "lr": 8.288015576739403e-05}
{"train_loss": 0.16616621613502502, "global_step": 74082, "epoch": 832, "lr": 8.287971901217952e-05}
{"train_loss": 0.19353264570236206, "global_step": 74083, "epoch": 832, "lr": 8.287928225254476e-05}
{"train_loss": 0.16563400626182556, "global_step": 74084, "epoch": 832, "lr": 8.287884548848977e-05}
{"train_loss": 0.19806994497776031, "global_step": 74085, "epoch": 832, "lr": 8.28784087200146e-05}
{"train_loss": 0.19360759854316711, "global_step": 74086, "epoch": 832, "lr": 8.287797194711935e-05}
{"train_loss": 0.197307288646698, "global_step": 74087, "epoch": 832, "lr": 8.287753516980406e-05}
{"train_loss": 0.12803050875663757, "global_step": 74088, "epoch": 832, "lr": 8.287709838806877e-05}
{"train_loss": 0.25553950667381287, "global_step": 74089, "epoch": 832, "lr": 8.287666160191356e-05}
{"train_loss": 0.20049647986888885, "global_step": 74090, "epoch": 832, "lr": 8.28762248113385e-05}
{"train_loss": 0.2559356093406677, "global_step": 74091, "epoch": 832, "lr": 8.287578801634362e-05}
{"train_loss": 0.3317979574203491, "global_step": 74092, "epoch": 832, "lr": 8.287535121692898e-05}
{"train_loss": 0.22504763305187225, "global_step": 74093, "epoch": 832, "lr": 8.287491441309465e-05}
{"train_loss": 0.23379382491111755, "global_step": 74094, "epoch": 832, "lr": 8.28744776048407e-05}
{"train_loss": 0.2630915641784668, "global_step": 74095, "epoch": 832, "lr": 8.287404079216719e-05}
{"train_loss": 0.10230913013219833, "global_step": 74096, "epoch": 832, "lr": 8.287360397507414e-05}
{"train_loss": 0.2059989720582962, "global_step": 74097, "epoch": 832, "lr": 8.287316715356164e-05}
{"train_loss": 0.19002756476402283, "global_step": 74098, "epoch": 832, "lr": 8.287273032762974e-05}
{"train_loss": 0.19560344517230988, "global_step": 74099, "epoch": 832, "lr": 8.287229349727853e-05}
{"train_loss": 0.3268139064311981, "global_step": 74100, "epoch": 832, "lr": 8.287185666250801e-05}
{"train_loss": 0.171600803732872, "global_step": 74101, "epoch": 832, "lr": 8.287141982331828e-05}
{"train_loss": 0.28765758872032166, "global_step": 74102, "epoch": 832, "lr": 8.287098297970939e-05}
{"train_loss": 0.15966330468654633, "global_step": 74103, "epoch": 832, "lr": 8.287054613168141e-05}
{"train_loss": 0.21359258890151978, "global_step": 74104, "epoch": 832, "lr": 8.287010927923437e-05}
{"train_loss": 0.22386839985847473, "global_step": 74105, "epoch": 832, "lr": 8.286967242236834e-05}
{"train_loss": 0.2712024450302124, "global_step": 74106, "epoch": 832, "lr": 8.28692355610834e-05}
{"train_loss": 0.1865035891532898, "global_step": 74107, "epoch": 832, "lr": 8.286879869537956e-05}
{"train_loss": 0.2012653797864914, "global_step": 74108, "epoch": 832, "lr": 8.286836182525695e-05}
{"train_loss": 0.20203450322151184, "global_step": 74109, "epoch": 832, "lr": 8.286792495071557e-05}
{"train_loss": 0.20912036299705505, "global_step": 74110, "epoch": 832, "lr": 8.28674880717555e-05}
{"train_loss": 0.2128869891166687, "global_step": 74111, "epoch": 832, "lr": 8.286705118837678e-05}
{"train_loss": 0.18350283801555634, "global_step": 74112, "epoch": 832, "lr": 8.286661430057951e-05}
{"train_loss": 0.20378345251083374, "global_step": 74113, "epoch": 832, "lr": 8.286617740836372e-05}
{"train_loss": 0.2364031970500946, "global_step": 74114, "epoch": 832, "lr": 8.286574051172948e-05}
{"train_loss": 0.23368597030639648, "global_step": 74115, "epoch": 832, "lr": 8.286530361067682e-05}
{"train_loss": 0.23950332403182983, "global_step": 74116, "epoch": 832, "lr": 8.286486670520584e-05}
{"train_loss": 0.2275221347808838, "global_step": 74117, "epoch": 832, "lr": 8.286442979531656e-05}
{"train_loss": 0.19542567431926727, "global_step": 74118, "epoch": 832, "lr": 8.286399288100908e-05}
{"train_loss": 0.2131611406803131, "global_step": 74119, "epoch": 832, "lr": 8.286355596228341e-05}
{"train_loss": 0.20890319347381592, "global_step": 74120, "epoch": 832, "lr": 8.286311903913965e-05}
{"train_loss": 0.20266282558441162, "global_step": 74121, "epoch": 832, "lr": 8.286268211157785e-05}
{"train_loss": 0.14284977316856384, "global_step": 74122, "epoch": 832, "lr": 8.286224517959807e-05}
{"train_loss": 0.17665210366249084, "global_step": 74123, "epoch": 832, "lr": 8.286180824320033e-05}
{"train_loss": 0.2034132033586502, "global_step": 74124, "epoch": 832, "lr": 8.286137130238474e-05}
{"train_loss": 0.18682889640331268, "global_step": 74125, "epoch": 832, "lr": 8.286093435715135e-05}
{"train_loss": 0.12473662942647934, "global_step": 74126, "epoch": 832, "lr": 8.286049740750019e-05}
{"train_loss": 0.1725885272026062, "global_step": 74127, "epoch": 832, "lr": 8.286006045343135e-05}
{"train_loss": 0.2028743475675583, "global_step": 74128, "epoch": 832, "lr": 8.285962349494487e-05}
{"train_loss": 0.26487720012664795, "global_step": 74129, "epoch": 832, "lr": 8.285918653204082e-05}
{"train_loss": 0.18136046826839447, "global_step": 74130, "epoch": 832, "lr": 8.285874956471924e-05}
{"train_loss": 0.17238856852054596, "global_step": 74131, "epoch": 832, "lr": 8.28583125929802e-05}
{"train_loss": 0.16699323058128357, "global_step": 74132, "epoch": 832, "lr": 8.285787561682378e-05}
{"train_loss": 0.2440769225358963, "global_step": 74133, "epoch": 832, "lr": 8.285743863625001e-05}
{"train_loss": 0.16517634689807892, "global_step": 74134, "epoch": 832, "lr": 8.285700165125895e-05}
{"train_loss": 0.183720201253891, "global_step": 74135, "epoch": 832, "lr": 8.285656466185068e-05}
{"train_loss": 0.20168863940105008, "global_step": 74136, "epoch": 832, "lr": 8.285612766802522e-05, "val_loss": 3.9767508506774902}
{"train_loss": 0.16767099499702454, "global_step": 74137, "epoch": 833, "lr": 8.285569066978267e-05}
{"train_loss": 0.16329827904701233, "global_step": 74138, "epoch": 833, "lr": 8.285525366712308e-05}
{"train_loss": 0.1921261101961136, "global_step": 74139, "epoch": 833, "lr": 8.28548166600465e-05}
{"train_loss": 0.12690900266170502, "global_step": 74140, "epoch": 833, "lr": 8.285437964855298e-05}
{"train_loss": 0.21272657811641693, "global_step": 74141, "epoch": 833, "lr": 8.28539426326426e-05}
{"train_loss": 0.2522655129432678, "global_step": 74142, "epoch": 833, "lr": 8.285350561231539e-05}
{"train_loss": 0.27973473072052, "global_step": 74143, "epoch": 833, "lr": 8.285306858757144e-05}
{"train_loss": 0.21218636631965637, "global_step": 74144, "epoch": 833, "lr": 8.28526315584108e-05}
{"train_loss": 0.2368655502796173, "global_step": 74145, "epoch": 833, "lr": 8.285219452483352e-05}
{"train_loss": 0.1744818240404129, "global_step": 74146, "epoch": 833, "lr": 8.285175748683966e-05}
{"train_loss": 0.27686601877212524, "global_step": 74147, "epoch": 833, "lr": 8.285132044442929e-05}
{"train_loss": 0.2683233320713043, "global_step": 74148, "epoch": 833, "lr": 8.285088339760245e-05}
{"train_loss": 0.2514551281929016, "global_step": 74149, "epoch": 833, "lr": 8.285044634635923e-05}
{"train_loss": 0.1990775465965271, "global_step": 74150, "epoch": 833, "lr": 8.285000929069963e-05}
{"train_loss": 0.21482229232788086, "global_step": 74151, "epoch": 833, "lr": 8.284957223062377e-05}
{"train_loss": 0.18963101506233215, "global_step": 74152, "epoch": 833, "lr": 8.284913516613168e-05}
{"train_loss": 0.11450592428445816, "global_step": 74153, "epoch": 833, "lr": 8.284869809722344e-05}
{"train_loss": 0.18986357748508453, "global_step": 74154, "epoch": 833, "lr": 8.284826102389907e-05}
{"train_loss": 0.12827858328819275, "global_step": 74155, "epoch": 833, "lr": 8.284782394615867e-05}
{"train_loss": 0.23487690091133118, "global_step": 74156, "epoch": 833, "lr": 8.284738686400226e-05}
{"train_loss": 0.2556558847427368, "global_step": 74157, "epoch": 833, "lr": 8.284694977742993e-05}
{"train_loss": 0.22181375324726105, "global_step": 74158, "epoch": 833, "lr": 8.284651268644175e-05}
{"train_loss": 0.19948211312294006, "global_step": 74159, "epoch": 833, "lr": 8.284607559103771e-05}
{"train_loss": 0.23352999985218048, "global_step": 74160, "epoch": 833, "lr": 8.284563849121796e-05}
{"train_loss": 0.1952342540025711, "global_step": 74161, "epoch": 833, "lr": 8.284520138698249e-05}
{"train_loss": 0.13106293976306915, "global_step": 74162, "epoch": 833, "lr": 8.284476427833139e-05}
{"train_loss": 0.20198599994182587, "global_step": 74163, "epoch": 833, "lr": 8.284432716526471e-05}
{"train_loss": 0.25291645526885986, "global_step": 74164, "epoch": 833, "lr": 8.284389004778251e-05}
{"train_loss": 0.20982272922992706, "global_step": 74165, "epoch": 833, "lr": 8.284345292588485e-05}
{"train_loss": 0.26349198818206787, "global_step": 74166, "epoch": 833, "lr": 8.28430157995718e-05}
{"train_loss": 0.17705965042114258, "global_step": 74167, "epoch": 833, "lr": 8.28425786688434e-05}
{"train_loss": 0.19445034861564636, "global_step": 74168, "epoch": 833, "lr": 8.284214153369971e-05}
{"train_loss": 0.1911773830652237, "global_step": 74169, "epoch": 833, "lr": 8.284170439414078e-05}
{"train_loss": 0.1757882535457611, "global_step": 74170, "epoch": 833, "lr": 8.28412672501667e-05}
{"train_loss": 0.23651330173015594, "global_step": 74171, "epoch": 833, "lr": 8.28408301017775e-05}
{"train_loss": 0.24244752526283264, "global_step": 74172, "epoch": 833, "lr": 8.284039294897327e-05}
{"train_loss": 0.207055002450943, "global_step": 74173, "epoch": 833, "lr": 8.283995579175404e-05}
{"train_loss": 0.18535302579402924, "global_step": 74174, "epoch": 833, "lr": 8.283951863011987e-05}
{"train_loss": 0.23156854510307312, "global_step": 74175, "epoch": 833, "lr": 8.283908146407084e-05}
{"train_loss": 0.2544775903224945, "global_step": 74176, "epoch": 833, "lr": 8.283864429360699e-05}
{"train_loss": 0.30778130888938904, "global_step": 74177, "epoch": 833, "lr": 8.283820711872839e-05}
{"train_loss": 0.3323468267917633, "global_step": 74178, "epoch": 833, "lr": 8.283776993943508e-05}
{"train_loss": 0.2050822377204895, "global_step": 74179, "epoch": 833, "lr": 8.283733275572715e-05}
{"train_loss": 0.1962447464466095, "global_step": 74180, "epoch": 833, "lr": 8.283689556760463e-05}
{"train_loss": 0.14416244626045227, "global_step": 74181, "epoch": 833, "lr": 8.283645837506759e-05}
{"train_loss": 0.20943930745124817, "global_step": 74182, "epoch": 833, "lr": 8.283602117811608e-05}
{"train_loss": 0.16067485511302948, "global_step": 74183, "epoch": 833, "lr": 8.283558397675019e-05}
{"train_loss": 0.2960330545902252, "global_step": 74184, "epoch": 833, "lr": 8.283514677096995e-05}
{"train_loss": 0.24687983095645905, "global_step": 74185, "epoch": 833, "lr": 8.283470956077541e-05}
{"train_loss": 0.17990927398204803, "global_step": 74186, "epoch": 833, "lr": 8.283427234616666e-05}
{"train_loss": 0.18546529114246368, "global_step": 74187, "epoch": 833, "lr": 8.283383512714373e-05}
{"train_loss": 0.23240654170513153, "global_step": 74188, "epoch": 833, "lr": 8.283339790370667e-05}
{"train_loss": 0.21645762026309967, "global_step": 74189, "epoch": 833, "lr": 8.28329606758556e-05}
{"train_loss": 0.2858302891254425, "global_step": 74190, "epoch": 833, "lr": 8.283252344359051e-05}
{"train_loss": 0.1203189417719841, "global_step": 74191, "epoch": 833, "lr": 8.283208620691152e-05}
{"train_loss": 0.15267705917358398, "global_step": 74192, "epoch": 833, "lr": 8.283164896581863e-05}
{"train_loss": 0.2631785273551941, "global_step": 74193, "epoch": 833, "lr": 8.283121172031194e-05}
{"train_loss": 0.2249608188867569, "global_step": 74194, "epoch": 833, "lr": 8.283077447039148e-05}
{"train_loss": 0.20342622697353363, "global_step": 74195, "epoch": 833, "lr": 8.283033721605733e-05}
{"train_loss": 0.1614881008863449, "global_step": 74196, "epoch": 833, "lr": 8.282989995730954e-05}
{"train_loss": 0.25080493092536926, "global_step": 74197, "epoch": 833, "lr": 8.282946269414816e-05}
{"train_loss": 0.2020217776298523, "global_step": 74198, "epoch": 833, "lr": 8.282902542657327e-05}
{"train_loss": 0.17692430317401886, "global_step": 74199, "epoch": 833, "lr": 8.282858815458492e-05}
{"train_loss": 0.17686693370342255, "global_step": 74200, "epoch": 833, "lr": 8.282815087818316e-05}
{"train_loss": 0.2504984438419342, "global_step": 74201, "epoch": 833, "lr": 8.282771359736806e-05}
{"train_loss": 0.25861626863479614, "global_step": 74202, "epoch": 833, "lr": 8.282727631213967e-05}
{"train_loss": 0.22462059557437897, "global_step": 74203, "epoch": 833, "lr": 8.282683902249805e-05}
{"train_loss": 0.17282074689865112, "global_step": 74204, "epoch": 833, "lr": 8.282640172844327e-05}
{"train_loss": 0.23235216736793518, "global_step": 74205, "epoch": 833, "lr": 8.282596442997538e-05}
{"train_loss": 0.2843463122844696, "global_step": 74206, "epoch": 833, "lr": 8.282552712709442e-05}
{"train_loss": 0.24706558883190155, "global_step": 74207, "epoch": 833, "lr": 8.282508981980049e-05}
{"train_loss": 0.16180811822414398, "global_step": 74208, "epoch": 833, "lr": 8.282465250809361e-05}
{"train_loss": 0.34406542778015137, "global_step": 74209, "epoch": 833, "lr": 8.282421519197388e-05}
{"train_loss": 0.1860194355249405, "global_step": 74210, "epoch": 833, "lr": 8.282377787144131e-05}
{"train_loss": 0.2563955783843994, "global_step": 74211, "epoch": 833, "lr": 8.2823340546496e-05}
{"train_loss": 0.17798693478107452, "global_step": 74212, "epoch": 833, "lr": 8.282290321713799e-05}
{"train_loss": 0.2105141431093216, "global_step": 74213, "epoch": 833, "lr": 8.282246588336733e-05}
{"train_loss": 0.24497808516025543, "global_step": 74214, "epoch": 833, "lr": 8.282202854518411e-05}
{"train_loss": 0.1669503152370453, "global_step": 74215, "epoch": 833, "lr": 8.282159120258835e-05}
{"train_loss": 0.230563685297966, "global_step": 74216, "epoch": 833, "lr": 8.282115385558013e-05}
{"train_loss": 0.20710046589374542, "global_step": 74217, "epoch": 833, "lr": 8.282071650415951e-05}
{"train_loss": 0.13631759583950043, "global_step": 74218, "epoch": 833, "lr": 8.282027914832653e-05}
{"train_loss": 0.1701364517211914, "global_step": 74219, "epoch": 833, "lr": 8.281984178808128e-05}
{"train_loss": 0.2217516303062439, "global_step": 74220, "epoch": 833, "lr": 8.28194044234238e-05}
{"train_loss": 0.2063475251197815, "global_step": 74221, "epoch": 833, "lr": 8.281896705435415e-05}
{"train_loss": 0.2434338927268982, "global_step": 74222, "epoch": 833, "lr": 8.281852968087238e-05}
{"train_loss": 0.3432936370372772, "global_step": 74223, "epoch": 833, "lr": 8.281809230297857e-05}
{"train_loss": 0.3301139175891876, "global_step": 74224, "epoch": 833, "lr": 8.281765492067275e-05}
{"train_loss": 0.21630457127362154, "global_step": 74225, "epoch": 833, "lr": 8.281721753395502e-05, "val_loss": 3.77254056930542}
{"train_loss": 0.1886053830385208, "global_step": 74226, "epoch": 834, "lr": 8.28167801428254e-05}
{"train_loss": 0.16412527859210968, "global_step": 74227, "epoch": 834, "lr": 8.281634274728397e-05}
{"train_loss": 0.17648467421531677, "global_step": 74228, "epoch": 834, "lr": 8.281590534733078e-05}
{"train_loss": 0.1974547803401947, "global_step": 74229, "epoch": 834, "lr": 8.28154679429659e-05}
{"train_loss": 0.15702232718467712, "global_step": 74230, "epoch": 834, "lr": 8.281503053418938e-05}
{"train_loss": 0.16172991693019867, "global_step": 74231, "epoch": 834, "lr": 8.281459312100126e-05}
{"train_loss": 0.22733864188194275, "global_step": 74232, "epoch": 834, "lr": 8.281415570340163e-05}
{"train_loss": 0.25054600834846497, "global_step": 74233, "epoch": 834, "lr": 8.281371828139055e-05}
{"train_loss": 0.22997306287288666, "global_step": 74234, "epoch": 834, "lr": 8.281328085496805e-05}
{"train_loss": 0.25534242391586304, "global_step": 74235, "epoch": 834, "lr": 8.28128434241342e-05}
{"train_loss": 0.16787472367286682, "global_step": 74236, "epoch": 834, "lr": 8.281240598888908e-05}
{"train_loss": 0.1500105857849121, "global_step": 74237, "epoch": 834, "lr": 8.281196854923272e-05}
{"train_loss": 0.165876567363739, "global_step": 74238, "epoch": 834, "lr": 8.28115311051652e-05}
{"train_loss": 0.252514123916626, "global_step": 74239, "epoch": 834, "lr": 8.281109365668655e-05}
{"train_loss": 0.21266303956508636, "global_step": 74240, "epoch": 834, "lr": 8.281065620379687e-05}
{"train_loss": 0.27046066522598267, "global_step": 74241, "epoch": 834, "lr": 8.281021874649618e-05}
{"train_loss": 0.14201027154922485, "global_step": 74242, "epoch": 834, "lr": 8.280978128478456e-05}
{"train_loss": 0.32276761531829834, "global_step": 74243, "epoch": 834, "lr": 8.280934381866207e-05}
{"train_loss": 0.2652243971824646, "global_step": 74244, "epoch": 834, "lr": 8.280890634812876e-05}
{"train_loss": 0.21800850331783295, "global_step": 74245, "epoch": 834, "lr": 8.28084688731847e-05}
{"train_loss": 0.1543615162372589, "global_step": 74246, "epoch": 834, "lr": 8.280803139382993e-05}
{"train_loss": 0.16547363996505737, "global_step": 74247, "epoch": 834, "lr": 8.280759391006454e-05}
{"train_loss": 0.15903300046920776, "global_step": 74248, "epoch": 834, "lr": 8.280715642188854e-05}
{"train_loss": 0.2086716890335083, "global_step": 74249, "epoch": 834, "lr": 8.280671892930204e-05}
{"train_loss": 0.1997256726026535, "global_step": 74250, "epoch": 834, "lr": 8.280628143230508e-05}
{"train_loss": 0.17207537591457367, "global_step": 74251, "epoch": 834, "lr": 8.28058439308977e-05}
{"train_loss": 0.21504688262939453, "global_step": 74252, "epoch": 834, "lr": 8.280540642507998e-05}
{"train_loss": 0.1568126082420349, "global_step": 74253, "epoch": 834, "lr": 8.280496891485197e-05}
{"train_loss": 0.18189558386802673, "global_step": 74254, "epoch": 834, "lr": 8.280453140021374e-05}
{"train_loss": 0.2713856101036072, "global_step": 74255, "epoch": 834, "lr": 8.280409388116532e-05}
{"train_loss": 0.14797914028167725, "global_step": 74256, "epoch": 834, "lr": 8.280365635770682e-05}
{"train_loss": 0.1700393706560135, "global_step": 74257, "epoch": 834, "lr": 8.280321882983824e-05}
{"train_loss": 0.12300223112106323, "global_step": 74258, "epoch": 834, "lr": 8.280278129755969e-05}
{"train_loss": 0.25117188692092896, "global_step": 74259, "epoch": 834, "lr": 8.28023437608712e-05}
{"train_loss": 0.13705769181251526, "global_step": 74260, "epoch": 834, "lr": 8.280190621977282e-05}
{"train_loss": 0.3101534843444824, "global_step": 74261, "epoch": 834, "lr": 8.280146867426462e-05}
{"train_loss": 0.3070851266384125, "global_step": 74262, "epoch": 834, "lr": 8.280103112434668e-05}
{"train_loss": 0.1943444162607193, "global_step": 74263, "epoch": 834, "lr": 8.280059357001903e-05}
{"train_loss": 0.1776162087917328, "global_step": 74264, "epoch": 834, "lr": 8.280015601128175e-05}
{"train_loss": 0.17607362568378448, "global_step": 74265, "epoch": 834, "lr": 8.27997184481349e-05}
{"train_loss": 0.2988291382789612, "global_step": 74266, "epoch": 834, "lr": 8.279928088057851e-05}
{"train_loss": 0.2114965170621872, "global_step": 74267, "epoch": 834, "lr": 8.279884330861265e-05}
{"train_loss": 0.19440023601055145, "global_step": 74268, "epoch": 834, "lr": 8.27984057322374e-05}
{"train_loss": 0.1841774731874466, "global_step": 74269, "epoch": 834, "lr": 8.279796815145279e-05}
{"train_loss": 0.23727187514305115, "global_step": 74270, "epoch": 834, "lr": 8.279753056625892e-05}
{"train_loss": 0.24498522281646729, "global_step": 74271, "epoch": 834, "lr": 8.27970929766558e-05}
{"train_loss": 0.12809525430202484, "global_step": 74272, "epoch": 834, "lr": 8.279665538264351e-05}
{"train_loss": 0.1739058792591095, "global_step": 74273, "epoch": 834, "lr": 8.27962177842221e-05}
{"train_loss": 0.23074255883693695, "global_step": 74274, "epoch": 834, "lr": 8.279578018139167e-05}
{"train_loss": 0.2062321901321411, "global_step": 74275, "epoch": 834, "lr": 8.279534257415223e-05}
{"train_loss": 0.26745444536209106, "global_step": 74276, "epoch": 834, "lr": 8.279490496250385e-05}
{"train_loss": 0.29559528827667236, "global_step": 74277, "epoch": 834, "lr": 8.27944673464466e-05}
{"train_loss": 0.19211265444755554, "global_step": 74278, "epoch": 834, "lr": 8.279402972598054e-05}
{"train_loss": 0.2893403172492981, "global_step": 74279, "epoch": 834, "lr": 8.279359210110571e-05}
{"train_loss": 0.18200412392616272, "global_step": 74280, "epoch": 834, "lr": 8.279315447182219e-05}
{"train_loss": 0.20598922669887543, "global_step": 74281, "epoch": 834, "lr": 8.279271683813001e-05}
{"train_loss": 0.16229282319545746, "global_step": 74282, "epoch": 834, "lr": 8.279227920002928e-05}
{"train_loss": 0.1933242380619049, "global_step": 74283, "epoch": 834, "lr": 8.279184155752e-05}
{"train_loss": 0.3004738390445709, "global_step": 74284, "epoch": 834, "lr": 8.279140391060227e-05}
{"train_loss": 0.3123265206813812, "global_step": 74285, "epoch": 834, "lr": 8.279096625927614e-05}
{"train_loss": 0.1691799908876419, "global_step": 74286, "epoch": 834, "lr": 8.279052860354166e-05}
{"train_loss": 0.14417195320129395, "global_step": 74287, "epoch": 834, "lr": 8.279009094339889e-05}
{"train_loss": 0.1279139220714569, "global_step": 74288, "epoch": 834, "lr": 8.27896532788479e-05}
{"train_loss": 0.19551055133342743, "global_step": 74289, "epoch": 834, "lr": 8.278921560988873e-05}
{"train_loss": 0.23789098858833313, "global_step": 74290, "epoch": 834, "lr": 8.278877793652147e-05}
{"train_loss": 0.20936281979084015, "global_step": 74291, "epoch": 834, "lr": 8.278834025874613e-05}
{"train_loss": 0.18040113151073456, "global_step": 74292, "epoch": 834, "lr": 8.278790257656281e-05}
{"train_loss": 0.3244323134422302, "global_step": 74293, "epoch": 834, "lr": 8.278746488997156e-05}
{"train_loss": 0.3133651614189148, "global_step": 74294, "epoch": 834, "lr": 8.278702719897245e-05}
{"train_loss": 0.23278483748435974, "global_step": 74295, "epoch": 834, "lr": 8.278658950356549e-05}
{"train_loss": 0.16146259009838104, "global_step": 74296, "epoch": 834, "lr": 8.278615180375079e-05}
{"train_loss": 0.1284046769142151, "global_step": 74297, "epoch": 834, "lr": 8.278571409952838e-05}
{"train_loss": 0.2398141324520111, "global_step": 74298, "epoch": 834, "lr": 8.278527639089834e-05}
{"train_loss": 0.14626137912273407, "global_step": 74299, "epoch": 834, "lr": 8.278483867786073e-05}
{"train_loss": 0.154981791973114, "global_step": 74300, "epoch": 834, "lr": 8.278440096041557e-05}
{"train_loss": 0.2132059633731842, "global_step": 74301, "epoch": 834, "lr": 8.278396323856298e-05}
{"train_loss": 0.2421845644712448, "global_step": 74302, "epoch": 834, "lr": 8.278352551230297e-05}
{"train_loss": 0.29674017429351807, "global_step": 74303, "epoch": 834, "lr": 8.278308778163562e-05}
{"train_loss": 0.2489931583404541, "global_step": 74304, "epoch": 834, "lr": 8.278265004656097e-05}
{"train_loss": 0.2714237570762634, "global_step": 74305, "epoch": 834, "lr": 8.27822123070791e-05}
{"train_loss": 0.15454287827014923, "global_step": 74306, "epoch": 834, "lr": 8.278177456319007e-05}
{"train_loss": 0.14296379685401917, "global_step": 74307, "epoch": 834, "lr": 8.27813368148939e-05}
{"train_loss": 0.23558717966079712, "global_step": 74308, "epoch": 834, "lr": 8.278089906219071e-05}
{"train_loss": 0.26299434900283813, "global_step": 74309, "epoch": 834, "lr": 8.278046130508052e-05}
{"train_loss": 0.15552255511283875, "global_step": 74310, "epoch": 834, "lr": 8.27800235435634e-05}
{"train_loss": 0.21479135751724243, "global_step": 74311, "epoch": 834, "lr": 8.27795857776394e-05}
{"train_loss": 0.16565166413784027, "global_step": 74312, "epoch": 834, "lr": 8.277914800730859e-05}
{"train_loss": 0.29006636142730713, "global_step": 74313, "epoch": 834, "lr": 8.277871023257101e-05}
{"train_loss": 0.20883154885822466, "global_step": 74314, "epoch": 834, "lr": 8.277827245342674e-05, "val_loss": 3.669149398803711}
{"train_loss": 0.22464026510715485, "global_step": 74315, "epoch": 835, "lr": 8.277783466987583e-05}
{"train_loss": 0.2089535892009735, "global_step": 74316, "epoch": 835, "lr": 8.277739688191833e-05}
{"train_loss": 0.2340693175792694, "global_step": 74317, "epoch": 835, "lr": 8.277695908955431e-05}
{"train_loss": 0.15156446397304535, "global_step": 74318, "epoch": 835, "lr": 8.277652129278385e-05}
{"train_loss": 0.17790386080741882, "global_step": 74319, "epoch": 835, "lr": 8.277608349160697e-05}
{"train_loss": 0.256832480430603, "global_step": 74320, "epoch": 835, "lr": 8.277564568602374e-05}
{"train_loss": 0.1805942803621292, "global_step": 74321, "epoch": 835, "lr": 8.277520787603424e-05}
{"train_loss": 0.19567625224590302, "global_step": 74322, "epoch": 835, "lr": 8.27747700616385e-05}
{"train_loss": 0.164537712931633, "global_step": 74323, "epoch": 835, "lr": 8.27743322428366e-05}
{"train_loss": 0.18528379499912262, "global_step": 74324, "epoch": 835, "lr": 8.277389441962857e-05}
{"train_loss": 0.21636737883090973, "global_step": 74325, "epoch": 835, "lr": 8.277345659201451e-05}
{"train_loss": 0.24829576909542084, "global_step": 74326, "epoch": 835, "lr": 8.277301875999445e-05}
{"train_loss": 0.16259144246578217, "global_step": 74327, "epoch": 835, "lr": 8.277258092356847e-05}
{"train_loss": 0.15703219175338745, "global_step": 74328, "epoch": 835, "lr": 8.27721430827366e-05}
{"train_loss": 0.24564200639724731, "global_step": 74329, "epoch": 835, "lr": 8.277170523749893e-05}
{"train_loss": 0.24502140283584595, "global_step": 74330, "epoch": 835, "lr": 8.27712673878555e-05}
{"train_loss": 0.22233018279075623, "global_step": 74331, "epoch": 835, "lr": 8.277082953380635e-05}
{"train_loss": 0.17238882184028625, "global_step": 74332, "epoch": 835, "lr": 8.277039167535158e-05}
{"train_loss": 0.14728762209415436, "global_step": 74333, "epoch": 835, "lr": 8.276995381249122e-05}
{"train_loss": 0.16439132392406464, "global_step": 74334, "epoch": 835, "lr": 8.276951594522535e-05}
{"train_loss": 0.16750498116016388, "global_step": 74335, "epoch": 835, "lr": 8.276907807355403e-05}
{"train_loss": 0.20706969499588013, "global_step": 74336, "epoch": 835, "lr": 8.276864019747727e-05}
{"train_loss": 0.2405782788991928, "global_step": 74337, "epoch": 835, "lr": 8.27682023169952e-05}
{"train_loss": 0.19255441427230835, "global_step": 74338, "epoch": 835, "lr": 8.276776443210783e-05}
{"train_loss": 0.2015235871076584, "global_step": 74339, "epoch": 835, "lr": 8.276732654281524e-05}
{"train_loss": 0.15755809843540192, "global_step": 74340, "epoch": 835, "lr": 8.276688864911748e-05}
{"train_loss": 0.19556620717048645, "global_step": 74341, "epoch": 835, "lr": 8.276645075101461e-05}
{"train_loss": 0.2983414828777313, "global_step": 74342, "epoch": 835, "lr": 8.27660128485067e-05}
{"train_loss": 0.2217828631401062, "global_step": 74343, "epoch": 835, "lr": 8.276557494159378e-05}
{"train_loss": 0.1725005805492401, "global_step": 74344, "epoch": 835, "lr": 8.276513703027593e-05}
{"train_loss": 0.17540127038955688, "global_step": 74345, "epoch": 835, "lr": 8.276469911455322e-05}
{"train_loss": 0.1514282077550888, "global_step": 74346, "epoch": 835, "lr": 8.276426119442569e-05}
{"train_loss": 0.17460401356220245, "global_step": 74347, "epoch": 835, "lr": 8.27638232698934e-05}
{"train_loss": 0.17802844941616058, "global_step": 74348, "epoch": 835, "lr": 8.276338534095642e-05}
{"train_loss": 0.09631455689668655, "global_step": 74349, "epoch": 835, "lr": 8.27629474076148e-05}
{"train_loss": 0.1801481395959854, "global_step": 74350, "epoch": 835, "lr": 8.27625094698686e-05}
{"train_loss": 0.1870391070842743, "global_step": 74351, "epoch": 835, "lr": 8.276207152771786e-05}
{"train_loss": 0.1552685648202896, "global_step": 74352, "epoch": 835, "lr": 8.276163358116269e-05}
{"train_loss": 0.21659798920154572, "global_step": 74353, "epoch": 835, "lr": 8.276119563020311e-05}
{"train_loss": 0.16335529088974, "global_step": 74354, "epoch": 835, "lr": 8.276075767483917e-05}
{"train_loss": 0.16297027468681335, "global_step": 74355, "epoch": 835, "lr": 8.276031971507096e-05}
{"train_loss": 0.2042289823293686, "global_step": 74356, "epoch": 835, "lr": 8.275988175089853e-05}
{"train_loss": 0.218099907040596, "global_step": 74357, "epoch": 835, "lr": 8.275944378232192e-05}
{"train_loss": 0.2447606325149536, "global_step": 74358, "epoch": 835, "lr": 8.275900580934122e-05}
{"train_loss": 0.18300527334213257, "global_step": 74359, "epoch": 835, "lr": 8.275856783195646e-05}
{"train_loss": 0.23855549097061157, "global_step": 74360, "epoch": 835, "lr": 8.27581298501677e-05}
{"train_loss": 0.17761355638504028, "global_step": 74361, "epoch": 835, "lr": 8.275769186397501e-05}
{"train_loss": 0.14399608969688416, "global_step": 74362, "epoch": 835, "lr": 8.275725387337846e-05}
{"train_loss": 0.28530481457710266, "global_step": 74363, "epoch": 835, "lr": 8.275681587837809e-05}
{"train_loss": 0.2435227781534195, "global_step": 74364, "epoch": 835, "lr": 8.275637787897397e-05}
{"train_loss": 0.22618356347084045, "global_step": 74365, "epoch": 835, "lr": 8.275593987516614e-05}
{"train_loss": 0.19297567009925842, "global_step": 74366, "epoch": 835, "lr": 8.27555018669547e-05}
{"train_loss": 0.13744938373565674, "global_step": 74367, "epoch": 835, "lr": 8.275506385433967e-05}
{"train_loss": 0.21402446925640106, "global_step": 74368, "epoch": 835, "lr": 8.275462583732111e-05}
{"train_loss": 0.23404502868652344, "global_step": 74369, "epoch": 835, "lr": 8.275418781589909e-05}
{"train_loss": 0.19887949526309967, "global_step": 74370, "epoch": 835, "lr": 8.275374979007368e-05}
{"train_loss": 0.14695227146148682, "global_step": 74371, "epoch": 835, "lr": 8.275331175984491e-05}
{"train_loss": 0.13640297949314117, "global_step": 74372, "epoch": 835, "lr": 8.275287372521288e-05}
{"train_loss": 0.23497658967971802, "global_step": 74373, "epoch": 835, "lr": 8.275243568617762e-05}
{"train_loss": 0.2180027812719345, "global_step": 74374, "epoch": 835, "lr": 8.275199764273918e-05}
{"train_loss": 0.25553759932518005, "global_step": 74375, "epoch": 835, "lr": 8.275155959489765e-05}
{"train_loss": 0.1909441202878952, "global_step": 74376, "epoch": 835, "lr": 8.275112154265305e-05}
{"train_loss": 0.1329321265220642, "global_step": 74377, "epoch": 835, "lr": 8.275068348600547e-05}
{"train_loss": 0.25715172290802, "global_step": 74378, "epoch": 835, "lr": 8.275024542495498e-05}
{"train_loss": 0.18668201565742493, "global_step": 74379, "epoch": 835, "lr": 8.274980735950159e-05}
{"train_loss": 0.1809430867433548, "global_step": 74380, "epoch": 835, "lr": 8.27493692896454e-05}
{"train_loss": 0.25409868359565735, "global_step": 74381, "epoch": 835, "lr": 8.274893121538647e-05}
{"train_loss": 0.17778536677360535, "global_step": 74382, "epoch": 835, "lr": 8.274849313672484e-05}
{"train_loss": 0.1697293370962143, "global_step": 74383, "epoch": 835, "lr": 8.274805505366055e-05}
{"train_loss": 0.1252090036869049, "global_step": 74384, "epoch": 835, "lr": 8.274761696619371e-05}
{"train_loss": 0.15330998599529266, "global_step": 74385, "epoch": 835, "lr": 8.274717887432434e-05}
{"train_loss": 0.2588267922401428, "global_step": 74386, "epoch": 835, "lr": 8.274674077805252e-05}
{"train_loss": 0.3125018775463104, "global_step": 74387, "epoch": 835, "lr": 8.274630267737829e-05}
{"train_loss": 0.23915284872055054, "global_step": 74388, "epoch": 835, "lr": 8.274586457230171e-05}
{"train_loss": 0.265770822763443, "global_step": 74389, "epoch": 835, "lr": 8.274542646282287e-05}
{"train_loss": 0.24547316133975983, "global_step": 74390, "epoch": 835, "lr": 8.274498834894178e-05}
{"train_loss": 0.23317421972751617, "global_step": 74391, "epoch": 835, "lr": 8.274455023065854e-05}
{"train_loss": 0.24627022445201874, "global_step": 74392, "epoch": 835, "lr": 8.27441121079732e-05}
{"train_loss": 0.21914999186992645, "global_step": 74393, "epoch": 835, "lr": 8.274367398088582e-05}
{"train_loss": 0.1970432847738266, "global_step": 74394, "epoch": 835, "lr": 8.274323584939644e-05}
{"train_loss": 0.22274145483970642, "global_step": 74395, "epoch": 835, "lr": 8.274279771350513e-05}
{"train_loss": 0.2514773905277252, "global_step": 74396, "epoch": 835, "lr": 8.274235957321195e-05}
{"train_loss": 0.22333082556724548, "global_step": 74397, "epoch": 835, "lr": 8.274192142851696e-05}
{"train_loss": 0.1555267870426178, "global_step": 74398, "epoch": 835, "lr": 8.274148327942021e-05}
{"train_loss": 0.2541176378726959, "global_step": 74399, "epoch": 835, "lr": 8.274104512592178e-05}
{"train_loss": 0.20135368406772614, "global_step": 74400, "epoch": 835, "lr": 8.274060696802171e-05}
{"train_loss": 0.21998386085033417, "global_step": 74401, "epoch": 835, "lr": 8.274016880572006e-05}
{"train_loss": 0.20473745465278625, "global_step": 74402, "epoch": 835, "lr": 8.27397306390169e-05}
{"train_loss": 0.20125917060656495, "global_step": 74403, "epoch": 835, "lr": 8.273929246791228e-05, "val_loss": 3.9878571033477783, "train_action_mse_error": 26.087215423583984}
{"train_loss": 0.25358837842941284, "global_step": 74404, "epoch": 836, "lr": 8.273885429240626e-05}
{"train_loss": 0.19337023794651031, "global_step": 74405, "epoch": 836, "lr": 8.27384161124989e-05}
{"train_loss": 0.3157860338687897, "global_step": 74406, "epoch": 836, "lr": 8.273797792819027e-05}
{"train_loss": 0.1749892234802246, "global_step": 74407, "epoch": 836, "lr": 8.273753973948041e-05}
{"train_loss": 0.22497372329235077, "global_step": 74408, "epoch": 836, "lr": 8.273710154636938e-05}
{"train_loss": 0.15069496631622314, "global_step": 74409, "epoch": 836, "lr": 8.273666334885726e-05}
{"train_loss": 0.23315876722335815, "global_step": 74410, "epoch": 836, "lr": 8.273622514694409e-05}
{"train_loss": 0.19906653463840485, "global_step": 74411, "epoch": 836, "lr": 8.273578694062992e-05}
{"train_loss": 0.19179992377758026, "global_step": 74412, "epoch": 836, "lr": 8.273534872991483e-05}
{"train_loss": 0.17490118741989136, "global_step": 74413, "epoch": 836, "lr": 8.273491051479886e-05}
{"train_loss": 0.1702919602394104, "global_step": 74414, "epoch": 836, "lr": 8.27344722952821e-05}
{"train_loss": 0.14476999640464783, "global_step": 74415, "epoch": 836, "lr": 8.27340340713646e-05}
{"train_loss": 0.1633944809436798, "global_step": 74416, "epoch": 836, "lr": 8.273359584304638e-05}
{"train_loss": 0.1931779682636261, "global_step": 74417, "epoch": 836, "lr": 8.273315761032754e-05}
{"train_loss": 0.20068565011024475, "global_step": 74418, "epoch": 836, "lr": 8.273271937320812e-05}
{"train_loss": 0.22481368482112885, "global_step": 74419, "epoch": 836, "lr": 8.273228113168818e-05}
{"train_loss": 0.21672053635120392, "global_step": 74420, "epoch": 836, "lr": 8.27318428857678e-05}
{"train_loss": 0.22110286355018616, "global_step": 74421, "epoch": 836, "lr": 8.273140463544702e-05}
{"train_loss": 0.24597325921058655, "global_step": 74422, "epoch": 836, "lr": 8.273096638072589e-05}
{"train_loss": 0.25085097551345825, "global_step": 74423, "epoch": 836, "lr": 8.273052812160449e-05}
{"train_loss": 0.24504601955413818, "global_step": 74424, "epoch": 836, "lr": 8.273008985808286e-05}
{"train_loss": 0.1982748955488205, "global_step": 74425, "epoch": 836, "lr": 8.272965159016109e-05}
{"train_loss": 0.17542323470115662, "global_step": 74426, "epoch": 836, "lr": 8.272921331783918e-05}
{"train_loss": 0.18675732612609863, "global_step": 74427, "epoch": 836, "lr": 8.272877504111726e-05}
{"train_loss": 0.15206953883171082, "global_step": 74428, "epoch": 836, "lr": 8.272833675999534e-05}
{"train_loss": 0.11377860605716705, "global_step": 74429, "epoch": 836, "lr": 8.27278984744735e-05}
{"train_loss": 0.15999194979667664, "global_step": 74430, "epoch": 836, "lr": 8.272746018455178e-05}
{"train_loss": 0.16311080753803253, "global_step": 74431, "epoch": 836, "lr": 8.272702189023027e-05}
{"train_loss": 0.22179438173770905, "global_step": 74432, "epoch": 836, "lr": 8.2726583591509e-05}
{"train_loss": 0.15548075735569, "global_step": 74433, "epoch": 836, "lr": 8.272614528838804e-05}
{"train_loss": 0.18779495358467102, "global_step": 74434, "epoch": 836, "lr": 8.272570698086746e-05}
{"train_loss": 0.1580497920513153, "global_step": 74435, "epoch": 836, "lr": 8.27252686689473e-05}
{"train_loss": 0.19450119137763977, "global_step": 74436, "epoch": 836, "lr": 8.272483035262763e-05}
{"train_loss": 0.14079660177230835, "global_step": 74437, "epoch": 836, "lr": 8.27243920319085e-05}
{"train_loss": 0.18186964094638824, "global_step": 74438, "epoch": 836, "lr": 8.272395370678997e-05}
{"train_loss": 0.2845704257488251, "global_step": 74439, "epoch": 836, "lr": 8.27235153772721e-05}
{"train_loss": 0.217824786901474, "global_step": 74440, "epoch": 836, "lr": 8.272307704335497e-05}
{"train_loss": 0.2334575355052948, "global_step": 74441, "epoch": 836, "lr": 8.27226387050386e-05}
{"train_loss": 0.2476060539484024, "global_step": 74442, "epoch": 836, "lr": 8.272220036232307e-05}
{"train_loss": 0.23369258642196655, "global_step": 74443, "epoch": 836, "lr": 8.272176201520847e-05}
{"train_loss": 0.14364422857761383, "global_step": 74444, "epoch": 836, "lr": 8.27213236636948e-05}
{"train_loss": 0.2643201947212219, "global_step": 74445, "epoch": 836, "lr": 8.272088530778216e-05}
{"train_loss": 0.23597146570682526, "global_step": 74446, "epoch": 836, "lr": 8.272044694747059e-05}
{"train_loss": 0.18398380279541016, "global_step": 74447, "epoch": 836, "lr": 8.272000858276016e-05}
{"train_loss": 0.14765383303165436, "global_step": 74448, "epoch": 836, "lr": 8.27195702136509e-05}
{"train_loss": 0.1896398961544037, "global_step": 74449, "epoch": 836, "lr": 8.271913184014292e-05}
{"train_loss": 0.18608811497688293, "global_step": 74450, "epoch": 836, "lr": 8.271869346223625e-05}
{"train_loss": 0.2099512368440628, "global_step": 74451, "epoch": 836, "lr": 8.271825507993095e-05}
{"train_loss": 0.31329911947250366, "global_step": 74452, "epoch": 836, "lr": 8.271781669322707e-05}
{"train_loss": 0.24190539121627808, "global_step": 74453, "epoch": 836, "lr": 8.271737830212468e-05}
{"train_loss": 0.16362504661083221, "global_step": 74454, "epoch": 836, "lr": 8.271693990662383e-05}
{"train_loss": 0.1680421680212021, "global_step": 74455, "epoch": 836, "lr": 8.27165015067246e-05}
{"train_loss": 0.26250606775283813, "global_step": 74456, "epoch": 836, "lr": 8.271606310242704e-05}
{"train_loss": 0.1981787532567978, "global_step": 74457, "epoch": 836, "lr": 8.27156246937312e-05}
{"train_loss": 0.20200158655643463, "global_step": 74458, "epoch": 836, "lr": 8.271518628063713e-05}
{"train_loss": 0.17014721035957336, "global_step": 74459, "epoch": 836, "lr": 8.27147478631449e-05}
{"train_loss": 0.24696075916290283, "global_step": 74460, "epoch": 836, "lr": 8.271430944125458e-05}
{"train_loss": 0.20311203598976135, "global_step": 74461, "epoch": 836, "lr": 8.271387101496623e-05}
{"train_loss": 0.22559785842895508, "global_step": 74462, "epoch": 836, "lr": 8.271343258427989e-05}
{"train_loss": 0.22937357425689697, "global_step": 74463, "epoch": 836, "lr": 8.271299414919563e-05}
{"train_loss": 0.19078531861305237, "global_step": 74464, "epoch": 836, "lr": 8.27125557097135e-05}
{"train_loss": 0.20629435777664185, "global_step": 74465, "epoch": 836, "lr": 8.271211726583356e-05}
{"train_loss": 0.23620888590812683, "global_step": 74466, "epoch": 836, "lr": 8.271167881755588e-05}
{"train_loss": 0.2004351168870926, "global_step": 74467, "epoch": 836, "lr": 8.271124036488052e-05}
{"train_loss": 0.16345739364624023, "global_step": 74468, "epoch": 836, "lr": 8.271080190780753e-05}
{"train_loss": 0.2567848563194275, "global_step": 74469, "epoch": 836, "lr": 8.271036344633698e-05}
{"train_loss": 0.3250771164894104, "global_step": 74470, "epoch": 836, "lr": 8.27099249804689e-05}
{"train_loss": 0.19343584775924683, "global_step": 74471, "epoch": 836, "lr": 8.270948651020338e-05}
{"train_loss": 0.10136403143405914, "global_step": 74472, "epoch": 836, "lr": 8.270904803554047e-05}
{"train_loss": 0.3043873906135559, "global_step": 74473, "epoch": 836, "lr": 8.270860955648024e-05}
{"train_loss": 0.11602557450532913, "global_step": 74474, "epoch": 836, "lr": 8.270817107302271e-05}
{"train_loss": 0.1931770294904709, "global_step": 74475, "epoch": 836, "lr": 8.270773258516798e-05}
{"train_loss": 0.21697884798049927, "global_step": 74476, "epoch": 836, "lr": 8.270729409291608e-05}
{"train_loss": 0.20910198986530304, "global_step": 74477, "epoch": 836, "lr": 8.27068555962671e-05}
{"train_loss": 0.18270693719387054, "global_step": 74478, "epoch": 836, "lr": 8.270641709522108e-05}
{"train_loss": 0.17968228459358215, "global_step": 74479, "epoch": 836, "lr": 8.270597858977806e-05}
{"train_loss": 0.2323407381772995, "global_step": 74480, "epoch": 836, "lr": 8.270554007993814e-05}
{"train_loss": 0.1434951275587082, "global_step": 74481, "epoch": 836, "lr": 8.270510156570137e-05}
{"train_loss": 0.19449730217456818, "global_step": 74482, "epoch": 836, "lr": 8.270466304706777e-05}
{"train_loss": 0.1709490716457367, "global_step": 74483, "epoch": 836, "lr": 8.270422452403744e-05}
{"train_loss": 0.2595449984073639, "global_step": 74484, "epoch": 836, "lr": 8.270378599661043e-05}
{"train_loss": 0.14607389271259308, "global_step": 74485, "epoch": 836, "lr": 8.270334746478679e-05}
{"train_loss": 0.2985556125640869, "global_step": 74486, "epoch": 836, "lr": 8.270290892856658e-05}
{"train_loss": 0.20767155289649963, "global_step": 74487, "epoch": 836, "lr": 8.270247038794987e-05}
{"train_loss": 0.16456402838230133, "global_step": 74488, "epoch": 836, "lr": 8.27020318429367e-05}
{"train_loss": 0.196406289935112, "global_step": 74489, "epoch": 836, "lr": 8.270159329352716e-05}
{"train_loss": 0.27001726627349854, "global_step": 74490, "epoch": 836, "lr": 8.270115473972127e-05}
{"train_loss": 0.16721364855766296, "global_step": 74491, "epoch": 836, "lr": 8.270071618151911e-05}
{"train_loss": 0.20341441612899974, "global_step": 74492, "epoch": 836, "lr": 8.270027761892075e-05, "val_loss": 3.8537089824676514}
{"train_loss": 0.17932313680648804, "global_step": 74493, "epoch": 837, "lr": 8.269983905192623e-05}
{"train_loss": 0.2209038883447647, "global_step": 74494, "epoch": 837, "lr": 8.269940048053562e-05}
{"train_loss": 0.1848149448633194, "global_step": 74495, "epoch": 837, "lr": 8.269896190474897e-05}
{"train_loss": 0.2683563530445099, "global_step": 74496, "epoch": 837, "lr": 8.269852332456633e-05}
{"train_loss": 0.19092673063278198, "global_step": 74497, "epoch": 837, "lr": 8.269808473998779e-05}
{"train_loss": 0.20377211272716522, "global_step": 74498, "epoch": 837, "lr": 8.269764615101338e-05}
{"train_loss": 0.23323187232017517, "global_step": 74499, "epoch": 837, "lr": 8.269720755764316e-05}
{"train_loss": 0.2646864950656891, "global_step": 74500, "epoch": 837, "lr": 8.269676895987723e-05}
{"train_loss": 0.21686449646949768, "global_step": 74501, "epoch": 837, "lr": 8.26963303577156e-05}
{"train_loss": 0.2730424404144287, "global_step": 74502, "epoch": 837, "lr": 8.269589175115836e-05}
{"train_loss": 0.22433649003505707, "global_step": 74503, "epoch": 837, "lr": 8.269545314020553e-05}
{"train_loss": 0.19020836055278778, "global_step": 74504, "epoch": 837, "lr": 8.269501452485721e-05}
{"train_loss": 0.25969958305358887, "global_step": 74505, "epoch": 837, "lr": 8.269457590511344e-05}
{"train_loss": 0.21355243027210236, "global_step": 74506, "epoch": 837, "lr": 8.269413728097429e-05}
{"train_loss": 0.19151008129119873, "global_step": 74507, "epoch": 837, "lr": 8.26936986524398e-05}
{"train_loss": 0.19123858213424683, "global_step": 74508, "epoch": 837, "lr": 8.269326001951005e-05}
{"train_loss": 0.23112481832504272, "global_step": 74509, "epoch": 837, "lr": 8.269282138218509e-05}
{"train_loss": 0.18014292418956757, "global_step": 74510, "epoch": 837, "lr": 8.269238274046499e-05}
{"train_loss": 0.17401182651519775, "global_step": 74511, "epoch": 837, "lr": 8.269194409434977e-05}
{"train_loss": 0.12878136336803436, "global_step": 74512, "epoch": 837, "lr": 8.269150544383952e-05}
{"train_loss": 0.16563045978546143, "global_step": 74513, "epoch": 837, "lr": 8.269106678893431e-05}
{"train_loss": 0.211255744099617, "global_step": 74514, "epoch": 837, "lr": 8.269062812963419e-05}
{"train_loss": 0.2120094895362854, "global_step": 74515, "epoch": 837, "lr": 8.269018946593918e-05}
{"train_loss": 0.13685032725334167, "global_step": 74516, "epoch": 837, "lr": 8.268975079784941e-05}
{"train_loss": 0.1330474615097046, "global_step": 74517, "epoch": 837, "lr": 8.268931212536488e-05}
{"train_loss": 0.2192520797252655, "global_step": 74518, "epoch": 837, "lr": 8.268887344848568e-05}
{"train_loss": 0.25549232959747314, "global_step": 74519, "epoch": 837, "lr": 8.268843476721186e-05}
{"train_loss": 0.2192242443561554, "global_step": 74520, "epoch": 837, "lr": 8.268799608154348e-05}
{"train_loss": 0.2312413603067398, "global_step": 74521, "epoch": 837, "lr": 8.26875573914806e-05}
{"train_loss": 0.2796385884284973, "global_step": 74522, "epoch": 837, "lr": 8.268711869702326e-05}
{"train_loss": 0.205349862575531, "global_step": 74523, "epoch": 837, "lr": 8.268667999817154e-05}
{"train_loss": 0.17629757523536682, "global_step": 74524, "epoch": 837, "lr": 8.26862412949255e-05}
{"train_loss": 0.18505357205867767, "global_step": 74525, "epoch": 837, "lr": 8.26858025872852e-05}
{"train_loss": 0.13188183307647705, "global_step": 74526, "epoch": 837, "lr": 8.268536387525068e-05}
{"train_loss": 0.1576603651046753, "global_step": 74527, "epoch": 837, "lr": 8.268492515882203e-05}
{"train_loss": 0.15992721915245056, "global_step": 74528, "epoch": 837, "lr": 8.268448643799927e-05}
{"train_loss": 0.1907246708869934, "global_step": 74529, "epoch": 837, "lr": 8.268404771278249e-05}
{"train_loss": 0.28858017921447754, "global_step": 74530, "epoch": 837, "lr": 8.268360898317172e-05}
{"train_loss": 0.15592476725578308, "global_step": 74531, "epoch": 837, "lr": 8.268317024916705e-05}
{"train_loss": 0.14900757372379303, "global_step": 74532, "epoch": 837, "lr": 8.268273151076854e-05}
{"train_loss": 0.20138749480247498, "global_step": 74533, "epoch": 837, "lr": 8.268229276797621e-05}
{"train_loss": 0.19120898842811584, "global_step": 74534, "epoch": 837, "lr": 8.268185402079015e-05}
{"train_loss": 0.19739456474781036, "global_step": 74535, "epoch": 837, "lr": 8.268141526921043e-05}
{"train_loss": 0.13380534946918488, "global_step": 74536, "epoch": 837, "lr": 8.268097651323707e-05}
{"train_loss": 0.235057070851326, "global_step": 74537, "epoch": 837, "lr": 8.268053775287018e-05}
{"train_loss": 0.17897096276283264, "global_step": 74538, "epoch": 837, "lr": 8.268009898810976e-05}
{"train_loss": 0.21287141740322113, "global_step": 74539, "epoch": 837, "lr": 8.267966021895591e-05}
{"train_loss": 0.15191280841827393, "global_step": 74540, "epoch": 837, "lr": 8.267922144540867e-05}
{"train_loss": 0.3345920741558075, "global_step": 74541, "epoch": 837, "lr": 8.267878266746812e-05}
{"train_loss": 0.1803450584411621, "global_step": 74542, "epoch": 837, "lr": 8.26783438851343e-05}
{"train_loss": 0.2065865397453308, "global_step": 74543, "epoch": 837, "lr": 8.267790509840728e-05}
{"train_loss": 0.2007957249879837, "global_step": 74544, "epoch": 837, "lr": 8.267746630728712e-05}
{"train_loss": 0.2519175410270691, "global_step": 74545, "epoch": 837, "lr": 8.267702751177385e-05}
{"train_loss": 0.16253246366977692, "global_step": 74546, "epoch": 837, "lr": 8.267658871186756e-05}
{"train_loss": 0.3194211721420288, "global_step": 74547, "epoch": 837, "lr": 8.267614990756831e-05}
{"train_loss": 0.22418934106826782, "global_step": 74548, "epoch": 837, "lr": 8.267571109887616e-05}
{"train_loss": 0.2320345640182495, "global_step": 74549, "epoch": 837, "lr": 8.267527228579115e-05}
{"train_loss": 0.2748921811580658, "global_step": 74550, "epoch": 837, "lr": 8.267483346831333e-05}
{"train_loss": 0.16426819562911987, "global_step": 74551, "epoch": 837, "lr": 8.267439464644279e-05}
{"train_loss": 0.19888468086719513, "global_step": 74552, "epoch": 837, "lr": 8.267395582017957e-05}
{"train_loss": 0.2800562381744385, "global_step": 74553, "epoch": 837, "lr": 8.267351698952374e-05}
{"train_loss": 0.20993471145629883, "global_step": 74554, "epoch": 837, "lr": 8.267307815447536e-05}
{"train_loss": 0.23337194323539734, "global_step": 74555, "epoch": 837, "lr": 8.267263931503449e-05}
{"train_loss": 0.18140505254268646, "global_step": 74556, "epoch": 837, "lr": 8.267220047120116e-05}
{"train_loss": 0.21590717136859894, "global_step": 74557, "epoch": 837, "lr": 8.267176162297546e-05}
{"train_loss": 0.196509450674057, "global_step": 74558, "epoch": 837, "lr": 8.267132277035743e-05}
{"train_loss": 0.24662119150161743, "global_step": 74559, "epoch": 837, "lr": 8.267088391334716e-05}
{"train_loss": 0.18322868645191193, "global_step": 74560, "epoch": 837, "lr": 8.267044505194467e-05}
{"train_loss": 0.16022029519081116, "global_step": 74561, "epoch": 837, "lr": 8.267000618615006e-05}
{"train_loss": 0.1898946464061737, "global_step": 74562, "epoch": 837, "lr": 8.266956731596334e-05}
{"train_loss": 0.1321409046649933, "global_step": 74563, "epoch": 837, "lr": 8.266912844138461e-05}
{"train_loss": 0.19308093190193176, "global_step": 74564, "epoch": 837, "lr": 8.26686895624139e-05}
{"train_loss": 0.12998205423355103, "global_step": 74565, "epoch": 837, "lr": 8.266825067905129e-05}
{"train_loss": 0.1391889452934265, "global_step": 74566, "epoch": 837, "lr": 8.266781179129684e-05}
{"train_loss": 0.17900508642196655, "global_step": 74567, "epoch": 837, "lr": 8.266737289915059e-05}
{"train_loss": 0.19407886266708374, "global_step": 74568, "epoch": 837, "lr": 8.266693400261262e-05}
{"train_loss": 0.2332904189825058, "global_step": 74569, "epoch": 837, "lr": 8.266649510168296e-05}
{"train_loss": 0.20533387362957, "global_step": 74570, "epoch": 837, "lr": 8.26660561963617e-05}
{"train_loss": 0.2522384822368622, "global_step": 74571, "epoch": 837, "lr": 8.266561728664889e-05}
{"train_loss": 0.19987349212169647, "global_step": 74572, "epoch": 837, "lr": 8.266517837254458e-05}
{"train_loss": 0.19124218821525574, "global_step": 74573, "epoch": 837, "lr": 8.266473945404885e-05}
{"train_loss": 0.14433129131793976, "global_step": 74574, "epoch": 837, "lr": 8.266430053116172e-05}
{"train_loss": 0.15607157349586487, "global_step": 74575, "epoch": 837, "lr": 8.266386160388328e-05}
{"train_loss": 0.2023668885231018, "global_step": 74576, "epoch": 837, "lr": 8.26634226722136e-05}
{"train_loss": 0.2622718811035156, "global_step": 74577, "epoch": 837, "lr": 8.266298373615271e-05}
{"train_loss": 0.15828514099121094, "global_step": 74578, "epoch": 837, "lr": 8.266254479570067e-05}
{"train_loss": 0.12222500145435333, "global_step": 74579, "epoch": 837, "lr": 8.266210585085757e-05}
{"train_loss": 0.1845145970582962, "global_step": 74580, "epoch": 837, "lr": 8.266166690162343e-05}
{"train_loss": 0.2019030689523461, "global_step": 74581, "epoch": 837, "lr": 8.266122794799833e-05, "val_loss": 3.7987864017486572}
{"train_loss": 0.2636415958404541, "global_step": 74582, "epoch": 838, "lr": 8.266078898998233e-05}
{"train_loss": 0.22735080122947693, "global_step": 74583, "epoch": 838, "lr": 8.266035002757549e-05}
{"train_loss": 0.1844252496957779, "global_step": 74584, "epoch": 838, "lr": 8.265991106077784e-05}
{"train_loss": 0.22923719882965088, "global_step": 74585, "epoch": 838, "lr": 8.265947208958949e-05}
{"train_loss": 0.2141556590795517, "global_step": 74586, "epoch": 838, "lr": 8.265903311401046e-05}
{"train_loss": 0.25769519805908203, "global_step": 74587, "epoch": 838, "lr": 8.265859413404081e-05}
{"train_loss": 0.21167026460170746, "global_step": 74588, "epoch": 838, "lr": 8.265815514968063e-05}
{"train_loss": 0.16684383153915405, "global_step": 74589, "epoch": 838, "lr": 8.265771616092995e-05}
{"train_loss": 0.27702596783638, "global_step": 74590, "epoch": 838, "lr": 8.265727716778883e-05}
{"train_loss": 0.23058275878429413, "global_step": 74591, "epoch": 838, "lr": 8.265683817025735e-05}
{"train_loss": 0.19078481197357178, "global_step": 74592, "epoch": 838, "lr": 8.265639916833557e-05}
{"train_loss": 0.1417563259601593, "global_step": 74593, "epoch": 838, "lr": 8.26559601620235e-05}
{"train_loss": 0.19605699181556702, "global_step": 74594, "epoch": 838, "lr": 8.265552115132126e-05}
{"train_loss": 0.155244380235672, "global_step": 74595, "epoch": 838, "lr": 8.265508213622888e-05}
{"train_loss": 0.3107254207134247, "global_step": 74596, "epoch": 838, "lr": 8.265464311674642e-05}
{"train_loss": 0.17803189158439636, "global_step": 74597, "epoch": 838, "lr": 8.265420409287395e-05}
{"train_loss": 0.15657839179039001, "global_step": 74598, "epoch": 838, "lr": 8.26537650646115e-05}
{"train_loss": 0.14118435978889465, "global_step": 74599, "epoch": 838, "lr": 8.265332603195916e-05}
{"train_loss": 0.18588289618492126, "global_step": 74600, "epoch": 838, "lr": 8.265288699491698e-05}
{"train_loss": 0.21777406334877014, "global_step": 74601, "epoch": 838, "lr": 8.265244795348501e-05}
{"train_loss": 0.13057225942611694, "global_step": 74602, "epoch": 838, "lr": 8.265200890766331e-05}
{"train_loss": 0.16060827672481537, "global_step": 74603, "epoch": 838, "lr": 8.265156985745198e-05}
{"train_loss": 0.19446884095668793, "global_step": 74604, "epoch": 838, "lr": 8.2651130802851e-05}
{"train_loss": 0.24877960979938507, "global_step": 74605, "epoch": 838, "lr": 8.265069174386049e-05}
{"train_loss": 0.12737590074539185, "global_step": 74606, "epoch": 838, "lr": 8.26502526804805e-05}
{"train_loss": 0.227291077375412, "global_step": 74607, "epoch": 838, "lr": 8.264981361271109e-05}
{"train_loss": 0.24732527136802673, "global_step": 74608, "epoch": 838, "lr": 8.264937454055229e-05}
{"train_loss": 0.28436627984046936, "global_step": 74609, "epoch": 838, "lr": 8.264893546400419e-05}
{"train_loss": 0.1678604781627655, "global_step": 74610, "epoch": 838, "lr": 8.264849638306684e-05}
{"train_loss": 0.25328564643859863, "global_step": 74611, "epoch": 838, "lr": 8.26480572977403e-05}
{"train_loss": 0.18512222170829773, "global_step": 74612, "epoch": 838, "lr": 8.26476182080246e-05}
{"train_loss": 0.12901157140731812, "global_step": 74613, "epoch": 838, "lr": 8.264717911391986e-05}
{"train_loss": 0.1806378811597824, "global_step": 74614, "epoch": 838, "lr": 8.264674001542609e-05}
{"train_loss": 0.22817981243133545, "global_step": 74615, "epoch": 838, "lr": 8.264630091254336e-05}
{"train_loss": 0.15440446138381958, "global_step": 74616, "epoch": 838, "lr": 8.264586180527174e-05}
{"train_loss": 0.1884419023990631, "global_step": 74617, "epoch": 838, "lr": 8.264542269361128e-05}
{"train_loss": 0.1778452843427658, "global_step": 74618, "epoch": 838, "lr": 8.264498357756204e-05}
{"train_loss": 0.18860411643981934, "global_step": 74619, "epoch": 838, "lr": 8.264454445712408e-05}
{"train_loss": 0.21088172495365143, "global_step": 74620, "epoch": 838, "lr": 8.264410533229747e-05}
{"train_loss": 0.2722645103931427, "global_step": 74621, "epoch": 838, "lr": 8.264366620308223e-05}
{"train_loss": 0.21075484156608582, "global_step": 74622, "epoch": 838, "lr": 8.264322706947847e-05}
{"train_loss": 0.2572064995765686, "global_step": 74623, "epoch": 838, "lr": 8.264278793148621e-05}
{"train_loss": 0.25686222314834595, "global_step": 74624, "epoch": 838, "lr": 8.264234878910553e-05}
{"train_loss": 0.12793396413326263, "global_step": 74625, "epoch": 838, "lr": 8.26419096423365e-05}
{"train_loss": 0.27626752853393555, "global_step": 74626, "epoch": 838, "lr": 8.264147049117915e-05}
{"train_loss": 0.2188950777053833, "global_step": 74627, "epoch": 838, "lr": 8.264103133563356e-05}
{"train_loss": 0.24434438347816467, "global_step": 74628, "epoch": 838, "lr": 8.264059217569977e-05}
{"train_loss": 0.16858409345149994, "global_step": 74629, "epoch": 838, "lr": 8.264015301137787e-05}
{"train_loss": 0.15452544391155243, "global_step": 74630, "epoch": 838, "lr": 8.263971384266787e-05}
{"train_loss": 0.14887496829032898, "global_step": 74631, "epoch": 838, "lr": 8.263927466956986e-05}
{"train_loss": 0.22498130798339844, "global_step": 74632, "epoch": 838, "lr": 8.263883549208393e-05}
{"train_loss": 0.23933959007263184, "global_step": 74633, "epoch": 838, "lr": 8.263839631021009e-05}
{"train_loss": 0.21854564547538757, "global_step": 74634, "epoch": 838, "lr": 8.26379571239484e-05}
{"train_loss": 0.26599544286727905, "global_step": 74635, "epoch": 838, "lr": 8.263751793329895e-05}
{"train_loss": 0.2005605399608612, "global_step": 74636, "epoch": 838, "lr": 8.263707873826179e-05}
{"train_loss": 0.182773157954216, "global_step": 74637, "epoch": 838, "lr": 8.263663953883695e-05}
{"train_loss": 0.17264650762081146, "global_step": 74638, "epoch": 838, "lr": 8.263620033502453e-05}
{"train_loss": 0.22702044248580933, "global_step": 74639, "epoch": 838, "lr": 8.263576112682457e-05}
{"train_loss": 0.16655848920345306, "global_step": 74640, "epoch": 838, "lr": 8.263532191423711e-05}
{"train_loss": 0.16402657330036163, "global_step": 74641, "epoch": 838, "lr": 8.263488269726226e-05}
{"train_loss": 0.2545839548110962, "global_step": 74642, "epoch": 838, "lr": 8.263444347590001e-05}
{"train_loss": 0.19063623249530792, "global_step": 74643, "epoch": 838, "lr": 8.263400425015049e-05}
{"train_loss": 0.18504509329795837, "global_step": 74644, "epoch": 838, "lr": 8.263356502001372e-05}
{"train_loss": 0.22528518736362457, "global_step": 74645, "epoch": 838, "lr": 8.263312578548975e-05}
{"train_loss": 0.17494015395641327, "global_step": 74646, "epoch": 838, "lr": 8.263268654657867e-05}
{"train_loss": 0.16949982941150665, "global_step": 74647, "epoch": 838, "lr": 8.26322473032805e-05}
{"train_loss": 0.19390133023262024, "global_step": 74648, "epoch": 838, "lr": 8.263180805559534e-05}
{"train_loss": 0.16506801545619965, "global_step": 74649, "epoch": 838, "lr": 8.263136880352323e-05}
{"train_loss": 0.17758958041667938, "global_step": 74650, "epoch": 838, "lr": 8.263092954706424e-05}
{"train_loss": 0.21346886456012726, "global_step": 74651, "epoch": 838, "lr": 8.26304902862184e-05}
{"train_loss": 0.2046535760164261, "global_step": 74652, "epoch": 838, "lr": 8.26300510209858e-05}
{"train_loss": 0.26418009400367737, "global_step": 74653, "epoch": 838, "lr": 8.262961175136648e-05}
{"train_loss": 0.20854206383228302, "global_step": 74654, "epoch": 838, "lr": 8.262917247736052e-05}
{"train_loss": 0.3051651120185852, "global_step": 74655, "epoch": 838, "lr": 8.262873319896795e-05}
{"train_loss": 0.10224398225545883, "global_step": 74656, "epoch": 838, "lr": 8.262829391618885e-05}
{"train_loss": 0.15595471858978271, "global_step": 74657, "epoch": 838, "lr": 8.262785462902326e-05}
{"train_loss": 0.29104241728782654, "global_step": 74658, "epoch": 838, "lr": 8.262741533747129e-05}
{"train_loss": 0.32300710678100586, "global_step": 74659, "epoch": 838, "lr": 8.262697604153293e-05}
{"train_loss": 0.24565888941287994, "global_step": 74660, "epoch": 838, "lr": 8.262653674120829e-05}
{"train_loss": 0.1636110544204712, "global_step": 74661, "epoch": 838, "lr": 8.262609743649738e-05}
{"train_loss": 0.24615024030208588, "global_step": 74662, "epoch": 838, "lr": 8.26256581274003e-05}
{"train_loss": 0.1915900558233261, "global_step": 74663, "epoch": 838, "lr": 8.262521881391712e-05}
{"train_loss": 0.1498660445213318, "global_step": 74664, "epoch": 838, "lr": 8.262477949604786e-05}
{"train_loss": 0.23903681337833405, "global_step": 74665, "epoch": 838, "lr": 8.26243401737926e-05}
{"train_loss": 0.23885388672351837, "global_step": 74666, "epoch": 838, "lr": 8.262390084715139e-05}
{"train_loss": 0.15398617088794708, "global_step": 74667, "epoch": 838, "lr": 8.26234615161243e-05}
{"train_loss": 0.1776222437620163, "global_step": 74668, "epoch": 838, "lr": 8.262302218071138e-05}
{"train_loss": 0.24519777297973633, "global_step": 74669, "epoch": 838, "lr": 8.26225828409127e-05}
{"train_loss": 0.20483569742253657, "global_step": 74670, "epoch": 838, "lr": 8.26221434967283e-05, "val_loss": 3.866060495376587}
{"train_loss": 0.23123495280742645, "global_step": 74671, "epoch": 839, "lr": 8.262170414815826e-05}
{"train_loss": 0.15222790837287903, "global_step": 74672, "epoch": 839, "lr": 8.262126479520262e-05}
{"train_loss": 0.22225435078144073, "global_step": 74673, "epoch": 839, "lr": 8.262082543786146e-05}
{"train_loss": 0.20311427116394043, "global_step": 74674, "epoch": 839, "lr": 8.262038607613483e-05}
{"train_loss": 0.23881877958774567, "global_step": 74675, "epoch": 839, "lr": 8.261994671002277e-05}
{"train_loss": 0.16079261898994446, "global_step": 74676, "epoch": 839, "lr": 8.261950733952536e-05}
{"train_loss": 0.20896318554878235, "global_step": 74677, "epoch": 839, "lr": 8.261906796464265e-05}
{"train_loss": 0.14120899140834808, "global_step": 74678, "epoch": 839, "lr": 8.261862858537471e-05}
{"train_loss": 0.17150281369686127, "global_step": 74679, "epoch": 839, "lr": 8.261818920172159e-05}
{"train_loss": 0.2187572717666626, "global_step": 74680, "epoch": 839, "lr": 8.261774981368336e-05}
{"train_loss": 0.2236616164445877, "global_step": 74681, "epoch": 839, "lr": 8.261731042126008e-05}
{"train_loss": 0.20513060688972473, "global_step": 74682, "epoch": 839, "lr": 8.261687102445178e-05}
{"train_loss": 0.2525811791419983, "global_step": 74683, "epoch": 839, "lr": 8.261643162325855e-05}
{"train_loss": 0.1381983757019043, "global_step": 74684, "epoch": 839, "lr": 8.261599221768043e-05}
{"train_loss": 0.18220938742160797, "global_step": 74685, "epoch": 839, "lr": 8.26155528077175e-05}
{"train_loss": 0.2280416488647461, "global_step": 74686, "epoch": 839, "lr": 8.26151133933698e-05}
{"train_loss": 0.20390717685222626, "global_step": 74687, "epoch": 839, "lr": 8.261467397463739e-05}
{"train_loss": 0.20011773705482483, "global_step": 74688, "epoch": 839, "lr": 8.261423455152034e-05}
{"train_loss": 0.13797639310359955, "global_step": 74689, "epoch": 839, "lr": 8.26137951240187e-05}
{"train_loss": 0.2861197888851166, "global_step": 74690, "epoch": 839, "lr": 8.261335569213254e-05}
{"train_loss": 0.23302631080150604, "global_step": 74691, "epoch": 839, "lr": 8.261291625586191e-05}
{"train_loss": 0.13423870503902435, "global_step": 74692, "epoch": 839, "lr": 8.261247681520688e-05}
{"train_loss": 0.16978806257247925, "global_step": 74693, "epoch": 839, "lr": 8.261203737016748e-05}
{"train_loss": 0.14618416130542755, "global_step": 74694, "epoch": 839, "lr": 8.26115979207438e-05}
{"train_loss": 0.21951939165592194, "global_step": 74695, "epoch": 839, "lr": 8.261115846693588e-05}
{"train_loss": 0.2563655972480774, "global_step": 74696, "epoch": 839, "lr": 8.26107190087438e-05}
{"train_loss": 0.19553235173225403, "global_step": 74697, "epoch": 839, "lr": 8.26102795461676e-05}
{"train_loss": 0.19294613599777222, "global_step": 74698, "epoch": 839, "lr": 8.260984007920734e-05}
{"train_loss": 0.24121002852916718, "global_step": 74699, "epoch": 839, "lr": 8.260940060786308e-05}
{"train_loss": 0.24374696612358093, "global_step": 74700, "epoch": 839, "lr": 8.26089611321349e-05}
{"train_loss": 0.24050481617450714, "global_step": 74701, "epoch": 839, "lr": 8.260852165202284e-05}
{"train_loss": 0.15399092435836792, "global_step": 74702, "epoch": 839, "lr": 8.260808216752695e-05}
{"train_loss": 0.191504567861557, "global_step": 74703, "epoch": 839, "lr": 8.260764267864731e-05}
{"train_loss": 0.1615244448184967, "global_step": 74704, "epoch": 839, "lr": 8.260720318538398e-05}
{"train_loss": 0.3061181902885437, "global_step": 74705, "epoch": 839, "lr": 8.260676368773699e-05}
{"train_loss": 0.2394348680973053, "global_step": 74706, "epoch": 839, "lr": 8.260632418570643e-05}
{"train_loss": 0.15796954929828644, "global_step": 74707, "epoch": 839, "lr": 8.260588467929233e-05}
{"train_loss": 0.18756961822509766, "global_step": 74708, "epoch": 839, "lr": 8.260544516849477e-05}
{"train_loss": 0.26878291368484497, "global_step": 74709, "epoch": 839, "lr": 8.260500565331382e-05}
{"train_loss": 0.15930552780628204, "global_step": 74710, "epoch": 839, "lr": 8.260456613374952e-05}
{"train_loss": 0.20925697684288025, "global_step": 74711, "epoch": 839, "lr": 8.260412660980193e-05}
{"train_loss": 0.19114276766777039, "global_step": 74712, "epoch": 839, "lr": 8.260368708147113e-05}
{"train_loss": 0.19912011921405792, "global_step": 74713, "epoch": 839, "lr": 8.260324754875714e-05}
{"train_loss": 0.11943161487579346, "global_step": 74714, "epoch": 839, "lr": 8.260280801166003e-05}
{"train_loss": 0.2227705419063568, "global_step": 74715, "epoch": 839, "lr": 8.26023684701799e-05}
{"train_loss": 0.22103692591190338, "global_step": 74716, "epoch": 839, "lr": 8.260192892431676e-05}
{"train_loss": 0.2897742688655853, "global_step": 74717, "epoch": 839, "lr": 8.260148937407069e-05}
{"train_loss": 0.19099247455596924, "global_step": 74718, "epoch": 839, "lr": 8.260104981944176e-05}
{"train_loss": 0.3299964368343353, "global_step": 74719, "epoch": 839, "lr": 8.260061026043e-05}
{"train_loss": 0.2256803959608078, "global_step": 74720, "epoch": 839, "lr": 8.26001706970355e-05}
{"train_loss": 0.26928025484085083, "global_step": 74721, "epoch": 839, "lr": 8.259973112925829e-05}
{"train_loss": 0.2462449073791504, "global_step": 74722, "epoch": 839, "lr": 8.259929155709845e-05}
{"train_loss": 0.19950979948043823, "global_step": 74723, "epoch": 839, "lr": 8.259885198055604e-05}
{"train_loss": 0.23992380499839783, "global_step": 74724, "epoch": 839, "lr": 8.259841239963108e-05}
{"train_loss": 0.2298423796892166, "global_step": 74725, "epoch": 839, "lr": 8.25979728143237e-05}
{"train_loss": 0.24873247742652893, "global_step": 74726, "epoch": 839, "lr": 8.25975332246339e-05}
{"train_loss": 0.20865145325660706, "global_step": 74727, "epoch": 839, "lr": 8.259709363056177e-05}
{"train_loss": 0.2003231942653656, "global_step": 74728, "epoch": 839, "lr": 8.259665403210734e-05}
{"train_loss": 0.18770785629749298, "global_step": 74729, "epoch": 839, "lr": 8.25962144292707e-05}
{"train_loss": 0.2712199091911316, "global_step": 74730, "epoch": 839, "lr": 8.25957748220519e-05}
{"train_loss": 0.25846022367477417, "global_step": 74731, "epoch": 839, "lr": 8.2595335210451e-05}
{"train_loss": 0.26418885588645935, "global_step": 74732, "epoch": 839, "lr": 8.259489559446803e-05}
{"train_loss": 0.20913848280906677, "global_step": 74733, "epoch": 839, "lr": 8.25944559741031e-05}
{"train_loss": 0.18836694955825806, "global_step": 74734, "epoch": 839, "lr": 8.259401634935622e-05}
{"train_loss": 0.13390612602233887, "global_step": 74735, "epoch": 839, "lr": 8.259357672022748e-05}
{"train_loss": 0.2507147192955017, "global_step": 74736, "epoch": 839, "lr": 8.259313708671694e-05}
{"train_loss": 0.21989263594150543, "global_step": 74737, "epoch": 839, "lr": 8.259269744882466e-05}
{"train_loss": 0.2081560641527176, "global_step": 74738, "epoch": 839, "lr": 8.259225780655067e-05}
{"train_loss": 0.26285383105278015, "global_step": 74739, "epoch": 839, "lr": 8.259181815989504e-05}
{"train_loss": 0.16113990545272827, "global_step": 74740, "epoch": 839, "lr": 8.259137850885785e-05}
{"train_loss": 0.18624812364578247, "global_step": 74741, "epoch": 839, "lr": 8.259093885343913e-05}
{"train_loss": 0.23538707196712494, "global_step": 74742, "epoch": 839, "lr": 8.259049919363899e-05}
{"train_loss": 0.2082599401473999, "global_step": 74743, "epoch": 839, "lr": 8.259005952945742e-05}
{"train_loss": 0.3003539443016052, "global_step": 74744, "epoch": 839, "lr": 8.258961986089452e-05}
{"train_loss": 0.24272958934307098, "global_step": 74745, "epoch": 839, "lr": 8.258918018795037e-05}
{"train_loss": 0.2409266084432602, "global_step": 74746, "epoch": 839, "lr": 8.258874051062498e-05}
{"train_loss": 0.2767472267150879, "global_step": 74747, "epoch": 839, "lr": 8.258830082891843e-05}
{"train_loss": 0.21043236553668976, "global_step": 74748, "epoch": 839, "lr": 8.258786114283079e-05}
{"train_loss": 0.19138330221176147, "global_step": 74749, "epoch": 839, "lr": 8.258742145236212e-05}
{"train_loss": 0.2060280591249466, "global_step": 74750, "epoch": 839, "lr": 8.258698175751243e-05}
{"train_loss": 0.17325511574745178, "global_step": 74751, "epoch": 839, "lr": 8.258654205828183e-05}
{"train_loss": 0.18221230804920197, "global_step": 74752, "epoch": 839, "lr": 8.258610235467039e-05}
{"train_loss": 0.24845880270004272, "global_step": 74753, "epoch": 839, "lr": 8.258566264667813e-05}
{"train_loss": 0.09856351464986801, "global_step": 74754, "epoch": 839, "lr": 8.258522293430513e-05}
{"train_loss": 0.21742267906665802, "global_step": 74755, "epoch": 839, "lr": 8.258478321755144e-05}
{"train_loss": 0.1542893648147583, "global_step": 74756, "epoch": 839, "lr": 8.258434349641712e-05}
{"train_loss": 0.20173834264278412, "global_step": 74757, "epoch": 839, "lr": 8.258390377090224e-05}
{"train_loss": 0.2333613485097885, "global_step": 74758, "epoch": 839, "lr": 8.258346404100684e-05}
{"train_loss": 0.21059974282979965, "global_step": 74759, "epoch": 839, "lr": 8.258302430673099e-05, "val_loss": 3.761683940887451}
{"train_loss": 0.24919790029525757, "global_step": 74760, "epoch": 840, "lr": 8.258258456807478e-05}
{"train_loss": 0.31689396500587463, "global_step": 74761, "epoch": 840, "lr": 8.258214482503821e-05}
{"train_loss": 0.1798388659954071, "global_step": 74762, "epoch": 840, "lr": 8.258170507762136e-05}
{"train_loss": 0.18208517134189606, "global_step": 74763, "epoch": 840, "lr": 8.258126532582431e-05}
{"train_loss": 0.07347389310598373, "global_step": 74764, "epoch": 840, "lr": 8.258082556964711e-05}
{"train_loss": 0.22868221998214722, "global_step": 74765, "epoch": 840, "lr": 8.258038580908981e-05}
{"train_loss": 0.2397986352443695, "global_step": 74766, "epoch": 840, "lr": 8.257994604415248e-05}
{"train_loss": 0.13952134549617767, "global_step": 74767, "epoch": 840, "lr": 8.257950627483518e-05}
{"train_loss": 0.25134438276290894, "global_step": 74768, "epoch": 840, "lr": 8.257906650113794e-05}
{"train_loss": 0.21722042560577393, "global_step": 74769, "epoch": 840, "lr": 8.257862672306085e-05}
{"train_loss": 0.23912456631660461, "global_step": 74770, "epoch": 840, "lr": 8.257818694060396e-05}
{"train_loss": 0.1554839313030243, "global_step": 74771, "epoch": 840, "lr": 8.257774715376734e-05}
{"train_loss": 0.20168407261371613, "global_step": 74772, "epoch": 840, "lr": 8.257730736255103e-05}
{"train_loss": 0.2025114893913269, "global_step": 74773, "epoch": 840, "lr": 8.25768675669551e-05}
{"train_loss": 0.18399712443351746, "global_step": 74774, "epoch": 840, "lr": 8.25764277669796e-05}
{"train_loss": 0.23054087162017822, "global_step": 74775, "epoch": 840, "lr": 8.257598796262461e-05}
{"train_loss": 0.2693181037902832, "global_step": 74776, "epoch": 840, "lr": 8.257554815389017e-05}
{"train_loss": 0.2267981469631195, "global_step": 74777, "epoch": 840, "lr": 8.257510834077635e-05}
{"train_loss": 0.22547009587287903, "global_step": 74778, "epoch": 840, "lr": 8.25746685232832e-05}
{"train_loss": 0.23929932713508606, "global_step": 74779, "epoch": 840, "lr": 8.257422870141077e-05}
{"train_loss": 0.18543973565101624, "global_step": 74780, "epoch": 840, "lr": 8.257378887515914e-05}
{"train_loss": 0.24719172716140747, "global_step": 74781, "epoch": 840, "lr": 8.257334904452837e-05}
{"train_loss": 0.15548379719257355, "global_step": 74782, "epoch": 840, "lr": 8.25729092095185e-05}
{"train_loss": 0.2077321708202362, "global_step": 74783, "epoch": 840, "lr": 8.257246937012962e-05}
{"train_loss": 0.21384888887405396, "global_step": 74784, "epoch": 840, "lr": 8.257202952636174e-05}
{"train_loss": 0.15600518882274628, "global_step": 74785, "epoch": 840, "lr": 8.257158967821497e-05}
{"train_loss": 0.17707543075084686, "global_step": 74786, "epoch": 840, "lr": 8.257114982568932e-05}
{"train_loss": 0.2063547968864441, "global_step": 74787, "epoch": 840, "lr": 8.257070996878488e-05}
{"train_loss": 0.2591913640499115, "global_step": 74788, "epoch": 840, "lr": 8.257027010750173e-05}
{"train_loss": 0.3496972918510437, "global_step": 74789, "epoch": 840, "lr": 8.256983024183988e-05}
{"train_loss": 0.20264804363250732, "global_step": 74790, "epoch": 840, "lr": 8.256939037179944e-05}
{"train_loss": 0.25971394777297974, "global_step": 74791, "epoch": 840, "lr": 8.256895049738043e-05}
{"train_loss": 0.21632492542266846, "global_step": 74792, "epoch": 840, "lr": 8.256851061858291e-05}
{"train_loss": 0.20333102345466614, "global_step": 74793, "epoch": 840, "lr": 8.256807073540695e-05}
{"train_loss": 0.23505234718322754, "global_step": 74794, "epoch": 840, "lr": 8.256763084785265e-05}
{"train_loss": 0.20933149755001068, "global_step": 74795, "epoch": 840, "lr": 8.256719095591998e-05}
{"train_loss": 0.2684837281703949, "global_step": 74796, "epoch": 840, "lr": 8.256675105960907e-05}
{"train_loss": 0.20396862924098969, "global_step": 74797, "epoch": 840, "lr": 8.256631115891995e-05}
{"train_loss": 0.2768442630767822, "global_step": 74798, "epoch": 840, "lr": 8.25658712538527e-05}
{"train_loss": 0.15653151273727417, "global_step": 74799, "epoch": 840, "lr": 8.256543134440736e-05}
{"train_loss": 0.211835578083992, "global_step": 74800, "epoch": 840, "lr": 8.256499143058398e-05}
{"train_loss": 0.22452834248542786, "global_step": 74801, "epoch": 840, "lr": 8.256455151238264e-05}
{"train_loss": 0.11116664856672287, "global_step": 74802, "epoch": 840, "lr": 8.256411158980339e-05}
{"train_loss": 0.1498645395040512, "global_step": 74803, "epoch": 840, "lr": 8.256367166284631e-05}
{"train_loss": 0.25618061423301697, "global_step": 74804, "epoch": 840, "lr": 8.256323173151142e-05}
{"train_loss": 0.24709463119506836, "global_step": 74805, "epoch": 840, "lr": 8.256279179579881e-05}
{"train_loss": 0.1892985701560974, "global_step": 74806, "epoch": 840, "lr": 8.256235185570854e-05}
{"train_loss": 0.10526826232671738, "global_step": 74807, "epoch": 840, "lr": 8.256191191124064e-05}
{"train_loss": 0.13117322325706482, "global_step": 74808, "epoch": 840, "lr": 8.25614719623952e-05}
{"train_loss": 0.22681325674057007, "global_step": 74809, "epoch": 840, "lr": 8.256103200917227e-05}
{"train_loss": 0.19496944546699524, "global_step": 74810, "epoch": 840, "lr": 8.25605920515719e-05}
{"train_loss": 0.2148936688899994, "global_step": 74811, "epoch": 840, "lr": 8.256015208959414e-05}
{"train_loss": 0.21188849210739136, "global_step": 74812, "epoch": 840, "lr": 8.255971212323909e-05}
{"train_loss": 0.24050413072109222, "global_step": 74813, "epoch": 840, "lr": 8.255927215250677e-05}
{"train_loss": 0.2692234218120575, "global_step": 74814, "epoch": 840, "lr": 8.255883217739725e-05}
{"train_loss": 0.2436317503452301, "global_step": 74815, "epoch": 840, "lr": 8.25583921979106e-05}
{"train_loss": 0.240095853805542, "global_step": 74816, "epoch": 840, "lr": 8.255795221404687e-05}
{"train_loss": 0.17470428347587585, "global_step": 74817, "epoch": 840, "lr": 8.255751222580613e-05}
{"train_loss": 0.21672850847244263, "global_step": 74818, "epoch": 840, "lr": 8.25570722331884e-05}
{"train_loss": 0.1893407553434372, "global_step": 74819, "epoch": 840, "lr": 8.25566322361938e-05}
{"train_loss": 0.1943831443786621, "global_step": 74820, "epoch": 840, "lr": 8.255619223482234e-05}
{"train_loss": 0.18707625567913055, "global_step": 74821, "epoch": 840, "lr": 8.25557522290741e-05}
{"train_loss": 0.1987796425819397, "global_step": 74822, "epoch": 840, "lr": 8.255531221894914e-05}
{"train_loss": 0.23481832444667816, "global_step": 74823, "epoch": 840, "lr": 8.255487220444751e-05}
{"train_loss": 0.2230900377035141, "global_step": 74824, "epoch": 840, "lr": 8.255443218556927e-05}
{"train_loss": 0.26632577180862427, "global_step": 74825, "epoch": 840, "lr": 8.25539921623145e-05}
{"train_loss": 0.22962866723537445, "global_step": 74826, "epoch": 840, "lr": 8.255355213468322e-05}
{"train_loss": 0.22417519986629486, "global_step": 74827, "epoch": 840, "lr": 8.255311210267552e-05}
{"train_loss": 0.22046153247356415, "global_step": 74828, "epoch": 840, "lr": 8.255267206629146e-05}
{"train_loss": 0.2658504247665405, "global_step": 74829, "epoch": 840, "lr": 8.255223202553109e-05}
{"train_loss": 0.24050042033195496, "global_step": 74830, "epoch": 840, "lr": 8.255179198039445e-05}
{"train_loss": 0.24058905243873596, "global_step": 74831, "epoch": 840, "lr": 8.255135193088164e-05}
{"train_loss": 0.18442733585834503, "global_step": 74832, "epoch": 840, "lr": 8.255091187699268e-05}
{"train_loss": 0.16133473813533783, "global_step": 74833, "epoch": 840, "lr": 8.255047181872767e-05}
{"train_loss": 0.2969791889190674, "global_step": 74834, "epoch": 840, "lr": 8.255003175608663e-05}
{"train_loss": 0.13191021978855133, "global_step": 74835, "epoch": 840, "lr": 8.254959168906963e-05}
{"train_loss": 0.2420198619365692, "global_step": 74836, "epoch": 840, "lr": 8.254915161767675e-05}
{"train_loss": 0.18187037110328674, "global_step": 74837, "epoch": 840, "lr": 8.254871154190802e-05}
{"train_loss": 0.214975506067276, "global_step": 74838, "epoch": 840, "lr": 8.254827146176352e-05}
{"train_loss": 0.26332971453666687, "global_step": 74839, "epoch": 840, "lr": 8.254783137724329e-05}
{"train_loss": 0.2660379111766815, "global_step": 74840, "epoch": 840, "lr": 8.254739128834741e-05}
{"train_loss": 0.1921689659357071, "global_step": 74841, "epoch": 840, "lr": 8.254695119507593e-05}
{"train_loss": 0.226589173078537, "global_step": 74842, "epoch": 840, "lr": 8.25465110974289e-05}
{"train_loss": 0.21112850308418274, "global_step": 74843, "epoch": 840, "lr": 8.25460709954064e-05}
{"train_loss": 0.20592764019966125, "global_step": 74844, "epoch": 840, "lr": 8.254563088900848e-05}
{"train_loss": 0.22048288583755493, "global_step": 74845, "epoch": 840, "lr": 8.25451907782352e-05}
{"train_loss": 0.16098064184188843, "global_step": 74846, "epoch": 840, "lr": 8.25447506630866e-05}
{"train_loss": 0.1289798617362976, "global_step": 74847, "epoch": 840, "lr": 8.254431054356275e-05}
{"train_loss": 0.2115670366401083, "global_step": 74848, "epoch": 840, "lr": 8.254387041966373e-05, "val_loss": 3.874337673187256, "train_action_mse_error": 49.18566131591797}
{"train_loss": 0.1539062112569809, "global_step": 74849, "epoch": 841, "lr": 8.254343029138958e-05}
{"train_loss": 0.20089100301265717, "global_step": 74850, "epoch": 841, "lr": 8.254299015874037e-05}
{"train_loss": 0.20339465141296387, "global_step": 74851, "epoch": 841, "lr": 8.254255002171613e-05}
{"train_loss": 0.2299019992351532, "global_step": 74852, "epoch": 841, "lr": 8.254210988031696e-05}
{"train_loss": 0.12544752657413483, "global_step": 74853, "epoch": 841, "lr": 8.25416697345429e-05}
{"train_loss": 0.18993569910526276, "global_step": 74854, "epoch": 841, "lr": 8.254122958439398e-05}
{"train_loss": 0.2165486216545105, "global_step": 74855, "epoch": 841, "lr": 8.254078942987033e-05}
{"train_loss": 0.1620413064956665, "global_step": 74856, "epoch": 841, "lr": 8.254034927097195e-05}
{"train_loss": 0.2480773627758026, "global_step": 74857, "epoch": 841, "lr": 8.253990910769891e-05}
{"train_loss": 0.1701212227344513, "global_step": 74858, "epoch": 841, "lr": 8.253946894005128e-05}
{"train_loss": 0.22224247455596924, "global_step": 74859, "epoch": 841, "lr": 8.25390287680291e-05}
{"train_loss": 0.1394391506910324, "global_step": 74860, "epoch": 841, "lr": 8.253858859163247e-05}
{"train_loss": 0.34565454721450806, "global_step": 74861, "epoch": 841, "lr": 8.25381484108614e-05}
{"train_loss": 0.2038780152797699, "global_step": 74862, "epoch": 841, "lr": 8.2537708225716e-05}
{"train_loss": 0.17965269088745117, "global_step": 74863, "epoch": 841, "lr": 8.253726803619628e-05}
{"train_loss": 0.16139258444309235, "global_step": 74864, "epoch": 841, "lr": 8.253682784230233e-05}
{"train_loss": 0.24078965187072754, "global_step": 74865, "epoch": 841, "lr": 8.253638764403419e-05}
{"train_loss": 0.25729942321777344, "global_step": 74866, "epoch": 841, "lr": 8.253594744139195e-05}
{"train_loss": 0.20300856232643127, "global_step": 74867, "epoch": 841, "lr": 8.253550723437563e-05}
{"train_loss": 0.23470330238342285, "global_step": 74868, "epoch": 841, "lr": 8.253506702298533e-05}
{"train_loss": 0.1762222796678543, "global_step": 74869, "epoch": 841, "lr": 8.253462680722106e-05}
{"train_loss": 0.16347314417362213, "global_step": 74870, "epoch": 841, "lr": 8.253418658708292e-05}
{"train_loss": 0.13905754685401917, "global_step": 74871, "epoch": 841, "lr": 8.253374636257095e-05}
{"train_loss": 0.1640169322490692, "global_step": 74872, "epoch": 841, "lr": 8.253330613368522e-05}
{"train_loss": 0.16229262948036194, "global_step": 74873, "epoch": 841, "lr": 8.253286590042577e-05}
{"train_loss": 0.2287517488002777, "global_step": 74874, "epoch": 841, "lr": 8.25324256627927e-05}
{"train_loss": 0.2054796814918518, "global_step": 74875, "epoch": 841, "lr": 8.253198542078601e-05}
{"train_loss": 0.18089145421981812, "global_step": 74876, "epoch": 841, "lr": 8.25315451744058e-05}
{"train_loss": 0.20090708136558533, "global_step": 74877, "epoch": 841, "lr": 8.253110492365212e-05}
{"train_loss": 0.29149535298347473, "global_step": 74878, "epoch": 841, "lr": 8.253066466852504e-05}
{"train_loss": 0.1730985939502716, "global_step": 74879, "epoch": 841, "lr": 8.25302244090246e-05}
{"train_loss": 0.14114703238010406, "global_step": 74880, "epoch": 841, "lr": 8.252978414515087e-05}
{"train_loss": 0.28595560789108276, "global_step": 74881, "epoch": 841, "lr": 8.252934387690392e-05}
{"train_loss": 0.1885831356048584, "global_step": 74882, "epoch": 841, "lr": 8.252890360428378e-05}
{"train_loss": 0.176078200340271, "global_step": 74883, "epoch": 841, "lr": 8.252846332729052e-05}
{"train_loss": 0.17161522805690765, "global_step": 74884, "epoch": 841, "lr": 8.252802304592423e-05}
{"train_loss": 0.2741290032863617, "global_step": 74885, "epoch": 841, "lr": 8.252758276018492e-05}
{"train_loss": 0.15671752393245697, "global_step": 74886, "epoch": 841, "lr": 8.252714247007268e-05}
{"train_loss": 0.17799773812294006, "global_step": 74887, "epoch": 841, "lr": 8.252670217558756e-05}
{"train_loss": 0.21224957704544067, "global_step": 74888, "epoch": 841, "lr": 8.252626187672963e-05}
{"train_loss": 0.21137458086013794, "global_step": 74889, "epoch": 841, "lr": 8.252582157349893e-05}
{"train_loss": 0.16957367956638336, "global_step": 74890, "epoch": 841, "lr": 8.252538126589552e-05}
{"train_loss": 0.195992112159729, "global_step": 74891, "epoch": 841, "lr": 8.25249409539195e-05}
{"train_loss": 0.3005184829235077, "global_step": 74892, "epoch": 841, "lr": 8.25245006375709e-05}
{"train_loss": 0.18558017909526825, "global_step": 74893, "epoch": 841, "lr": 8.252406031684974e-05}
{"train_loss": 0.12663151323795319, "global_step": 74894, "epoch": 841, "lr": 8.252361999175615e-05}
{"train_loss": 0.13141481578350067, "global_step": 74895, "epoch": 841, "lr": 8.252317966229013e-05}
{"train_loss": 0.1808655709028244, "global_step": 74896, "epoch": 841, "lr": 8.252273932845177e-05}
{"train_loss": 0.2289799153804779, "global_step": 74897, "epoch": 841, "lr": 8.252229899024113e-05}
{"train_loss": 0.13907577097415924, "global_step": 74898, "epoch": 841, "lr": 8.252185864765827e-05}
{"train_loss": 0.14867815375328064, "global_step": 74899, "epoch": 841, "lr": 8.252141830070323e-05}
{"train_loss": 0.17717668414115906, "global_step": 74900, "epoch": 841, "lr": 8.25209779493761e-05}
{"train_loss": 0.1849287748336792, "global_step": 74901, "epoch": 841, "lr": 8.252053759367691e-05}
{"train_loss": 0.24724966287612915, "global_step": 74902, "epoch": 841, "lr": 8.252009723360572e-05}
{"train_loss": 0.23790554702281952, "global_step": 74903, "epoch": 841, "lr": 8.25196568691626e-05}
{"train_loss": 0.26690617203712463, "global_step": 74904, "epoch": 841, "lr": 8.251921650034761e-05}
{"train_loss": 0.17958542704582214, "global_step": 74905, "epoch": 841, "lr": 8.251877612716083e-05}
{"train_loss": 0.20013345777988434, "global_step": 74906, "epoch": 841, "lr": 8.251833574960229e-05}
{"train_loss": 0.2654334008693695, "global_step": 74907, "epoch": 841, "lr": 8.251789536767202e-05}
{"train_loss": 0.2042994648218155, "global_step": 74908, "epoch": 841, "lr": 8.251745498137016e-05}
{"train_loss": 0.15784792602062225, "global_step": 74909, "epoch": 841, "lr": 8.25170145906967e-05}
{"train_loss": 0.28711017966270447, "global_step": 74910, "epoch": 841, "lr": 8.251657419565174e-05}
{"train_loss": 0.21544338762760162, "global_step": 74911, "epoch": 841, "lr": 8.25161337962353e-05}
{"train_loss": 0.2363968789577484, "global_step": 74912, "epoch": 841, "lr": 8.25156933924475e-05}
{"train_loss": 0.1909235417842865, "global_step": 74913, "epoch": 841, "lr": 8.251525298428832e-05}
{"train_loss": 0.20966747403144836, "global_step": 74914, "epoch": 841, "lr": 8.251481257175788e-05}
{"train_loss": 0.15533147752285004, "global_step": 74915, "epoch": 841, "lr": 8.251437215485621e-05}
{"train_loss": 0.1705193817615509, "global_step": 74916, "epoch": 841, "lr": 8.25139317335834e-05}
{"train_loss": 0.15599533915519714, "global_step": 74917, "epoch": 841, "lr": 8.251349130793946e-05}
{"train_loss": 0.23913206160068512, "global_step": 74918, "epoch": 841, "lr": 8.251305087792449e-05}
{"train_loss": 0.17705431580543518, "global_step": 74919, "epoch": 841, "lr": 8.251261044353856e-05}
{"train_loss": 0.23065805435180664, "global_step": 74920, "epoch": 841, "lr": 8.251217000478167e-05}
{"train_loss": 0.15009309351444244, "global_step": 74921, "epoch": 841, "lr": 8.251172956165392e-05}
{"train_loss": 0.20095448195934296, "global_step": 74922, "epoch": 841, "lr": 8.251128911415538e-05}
{"train_loss": 0.16969376802444458, "global_step": 74923, "epoch": 841, "lr": 8.251084866228608e-05}
{"train_loss": 0.1690857857465744, "global_step": 74924, "epoch": 841, "lr": 8.251040820604611e-05}
{"train_loss": 0.2775733768939972, "global_step": 74925, "epoch": 841, "lr": 8.25099677454355e-05}
{"train_loss": 0.18249204754829407, "global_step": 74926, "epoch": 841, "lr": 8.250952728045433e-05}
{"train_loss": 0.2343984842300415, "global_step": 74927, "epoch": 841, "lr": 8.250908681110264e-05}
{"train_loss": 0.415208637714386, "global_step": 74928, "epoch": 841, "lr": 8.25086463373805e-05}
{"train_loss": 0.23070783913135529, "global_step": 74929, "epoch": 841, "lr": 8.250820585928798e-05}
{"train_loss": 0.1438283622264862, "global_step": 74930, "epoch": 841, "lr": 8.250776537682512e-05}
{"train_loss": 0.1736563742160797, "global_step": 74931, "epoch": 841, "lr": 8.250732488999197e-05}
{"train_loss": 0.19043703377246857, "global_step": 74932, "epoch": 841, "lr": 8.250688439878862e-05}
{"train_loss": 0.17680415511131287, "global_step": 74933, "epoch": 841, "lr": 8.250644390321512e-05}
{"train_loss": 0.1825396567583084, "global_step": 74934, "epoch": 841, "lr": 8.250600340327152e-05}
{"train_loss": 0.2549995481967926, "global_step": 74935, "epoch": 841, "lr": 8.250556289895788e-05}
{"train_loss": 0.21266499161720276, "global_step": 74936, "epoch": 841, "lr": 8.250512239027428e-05}
{"train_loss": 0.20170528835125184, "global_step": 74937, "epoch": 841, "lr": 8.250468187722075e-05, "val_loss": 4.017757892608643}
{"train_loss": 0.2518489956855774, "global_step": 74938, "epoch": 842, "lr": 8.250424135979738e-05}
{"train_loss": 0.18885669112205505, "global_step": 74939, "epoch": 842, "lr": 8.250380083800418e-05}
{"train_loss": 0.13546186685562134, "global_step": 74940, "epoch": 842, "lr": 8.250336031184127e-05}
{"train_loss": 0.2078026980161667, "global_step": 74941, "epoch": 842, "lr": 8.250291978130866e-05}
{"train_loss": 0.15484990179538727, "global_step": 74942, "epoch": 842, "lr": 8.250247924640644e-05}
{"train_loss": 0.20626047253608704, "global_step": 74943, "epoch": 842, "lr": 8.250203870713464e-05}
{"train_loss": 0.19661317765712738, "global_step": 74944, "epoch": 842, "lr": 8.250159816349335e-05}
{"train_loss": 0.26677149534225464, "global_step": 74945, "epoch": 842, "lr": 8.250115761548261e-05}
{"train_loss": 0.19189982116222382, "global_step": 74946, "epoch": 842, "lr": 8.25007170631025e-05}
{"train_loss": 0.18558502197265625, "global_step": 74947, "epoch": 842, "lr": 8.250027650635305e-05}
{"train_loss": 0.15464897453784943, "global_step": 74948, "epoch": 842, "lr": 8.249983594523434e-05}
{"train_loss": 0.21976278722286224, "global_step": 74949, "epoch": 842, "lr": 8.249939537974642e-05}
{"train_loss": 0.1578991562128067, "global_step": 74950, "epoch": 842, "lr": 8.249895480988935e-05}
{"train_loss": 0.19262202084064484, "global_step": 74951, "epoch": 842, "lr": 8.249851423566319e-05}
{"train_loss": 0.17999187111854553, "global_step": 74952, "epoch": 842, "lr": 8.2498073657068e-05}
{"train_loss": 0.20737028121948242, "global_step": 74953, "epoch": 842, "lr": 8.249763307410385e-05}
{"train_loss": 0.21792511641979218, "global_step": 74954, "epoch": 842, "lr": 8.249719248677079e-05}
{"train_loss": 0.1399383544921875, "global_step": 74955, "epoch": 842, "lr": 8.249675189506888e-05}
{"train_loss": 0.15228421986103058, "global_step": 74956, "epoch": 842, "lr": 8.249631129899818e-05}
{"train_loss": 0.20821276307106018, "global_step": 74957, "epoch": 842, "lr": 8.249587069855873e-05}
{"train_loss": 0.27866408228874207, "global_step": 74958, "epoch": 842, "lr": 8.249543009375063e-05}
{"train_loss": 0.21565084159374237, "global_step": 74959, "epoch": 842, "lr": 8.24949894845739e-05}
{"train_loss": 0.15193819999694824, "global_step": 74960, "epoch": 842, "lr": 8.249454887102861e-05}
{"train_loss": 0.24469538033008575, "global_step": 74961, "epoch": 842, "lr": 8.249410825311483e-05}
{"train_loss": 0.21496282517910004, "global_step": 74962, "epoch": 842, "lr": 8.249366763083261e-05}
{"train_loss": 0.18580077588558197, "global_step": 74963, "epoch": 842, "lr": 8.249322700418201e-05}
{"train_loss": 0.2644405663013458, "global_step": 74964, "epoch": 842, "lr": 8.249278637316312e-05}
{"train_loss": 0.2828436493873596, "global_step": 74965, "epoch": 842, "lr": 8.249234573777593e-05}
{"train_loss": 0.35436591506004333, "global_step": 74966, "epoch": 842, "lr": 8.249190509802056e-05}
{"train_loss": 0.22783908247947693, "global_step": 74967, "epoch": 842, "lr": 8.249146445389705e-05}
{"train_loss": 0.2281913310289383, "global_step": 74968, "epoch": 842, "lr": 8.249102380540545e-05}
{"train_loss": 0.20468534529209137, "global_step": 74969, "epoch": 842, "lr": 8.249058315254583e-05}
{"train_loss": 0.23542071878910065, "global_step": 74970, "epoch": 842, "lr": 8.249014249531826e-05}
{"train_loss": 0.13959667086601257, "global_step": 74971, "epoch": 842, "lr": 8.248970183372277e-05}
{"train_loss": 0.25274282693862915, "global_step": 74972, "epoch": 842, "lr": 8.248926116775945e-05}
{"train_loss": 0.1725577563047409, "global_step": 74973, "epoch": 842, "lr": 8.248882049742834e-05}
{"train_loss": 0.1825660765171051, "global_step": 74974, "epoch": 842, "lr": 8.24883798227295e-05}
{"train_loss": 0.20864233374595642, "global_step": 74975, "epoch": 842, "lr": 8.2487939143663e-05}
{"train_loss": 0.19978736340999603, "global_step": 74976, "epoch": 842, "lr": 8.248749846022889e-05}
{"train_loss": 0.18900996446609497, "global_step": 74977, "epoch": 842, "lr": 8.248705777242722e-05}
{"train_loss": 0.29626861214637756, "global_step": 74978, "epoch": 842, "lr": 8.248661708025808e-05}
{"train_loss": 0.26676610112190247, "global_step": 74979, "epoch": 842, "lr": 8.24861763837215e-05}
{"train_loss": 0.17796145379543304, "global_step": 74980, "epoch": 842, "lr": 8.248573568281754e-05}
{"train_loss": 0.22086776793003082, "global_step": 74981, "epoch": 842, "lr": 8.248529497754629e-05}
{"train_loss": 0.13059714436531067, "global_step": 74982, "epoch": 842, "lr": 8.248485426790777e-05}
{"train_loss": 0.22217683494091034, "global_step": 74983, "epoch": 842, "lr": 8.248441355390207e-05}
{"train_loss": 0.16586259007453918, "global_step": 74984, "epoch": 842, "lr": 8.248397283552923e-05}
{"train_loss": 0.286807656288147, "global_step": 74985, "epoch": 842, "lr": 8.248353211278931e-05}
{"train_loss": 0.1909303367137909, "global_step": 74986, "epoch": 842, "lr": 8.248309138568239e-05}
{"train_loss": 0.22569586336612701, "global_step": 74987, "epoch": 842, "lr": 8.24826506542085e-05}
{"train_loss": 0.2706294357776642, "global_step": 74988, "epoch": 842, "lr": 8.248220991836772e-05}
{"train_loss": 0.23142793774604797, "global_step": 74989, "epoch": 842, "lr": 8.24817691781601e-05}
{"train_loss": 0.09964974969625473, "global_step": 74990, "epoch": 842, "lr": 8.248132843358572e-05}
{"train_loss": 0.17012223601341248, "global_step": 74991, "epoch": 842, "lr": 8.248088768464459e-05}
{"train_loss": 0.20331554114818573, "global_step": 74992, "epoch": 842, "lr": 8.248044693133683e-05}
{"train_loss": 0.2421390861272812, "global_step": 74993, "epoch": 842, "lr": 8.248000617366245e-05}
{"train_loss": 0.16634492576122284, "global_step": 74994, "epoch": 842, "lr": 8.247956541162154e-05}
{"train_loss": 0.19420529901981354, "global_step": 74995, "epoch": 842, "lr": 8.247912464521415e-05}
{"train_loss": 0.1745942384004593, "global_step": 74996, "epoch": 842, "lr": 8.247868387444032e-05}
{"train_loss": 0.19762763381004333, "global_step": 74997, "epoch": 842, "lr": 8.247824309930015e-05}
{"train_loss": 0.2552352547645569, "global_step": 74998, "epoch": 842, "lr": 8.247780231979365e-05}
{"train_loss": 0.3325590193271637, "global_step": 74999, "epoch": 842, "lr": 8.247736153592091e-05}
{"train_loss": 0.15194359421730042, "global_step": 75000, "epoch": 842, "lr": 8.247692074768201e-05}
{"train_loss": 0.20644856989383698, "global_step": 75001, "epoch": 842, "lr": 8.247647995507697e-05}
{"train_loss": 0.24647805094718933, "global_step": 75002, "epoch": 842, "lr": 8.247603915810585e-05}
{"train_loss": 0.16737420856952667, "global_step": 75003, "epoch": 842, "lr": 8.247559835676874e-05}
{"train_loss": 0.2319563627243042, "global_step": 75004, "epoch": 842, "lr": 8.247515755106567e-05}
{"train_loss": 0.16629336774349213, "global_step": 75005, "epoch": 842, "lr": 8.247471674099671e-05}
{"train_loss": 0.20634400844573975, "global_step": 75006, "epoch": 842, "lr": 8.247427592656192e-05}
{"train_loss": 0.1702306568622589, "global_step": 75007, "epoch": 842, "lr": 8.247383510776138e-05}
{"train_loss": 0.20314377546310425, "global_step": 75008, "epoch": 842, "lr": 8.247339428459511e-05}
{"train_loss": 0.2715418040752411, "global_step": 75009, "epoch": 842, "lr": 8.247295345706319e-05}
{"train_loss": 0.26169735193252563, "global_step": 75010, "epoch": 842, "lr": 8.247251262516568e-05}
{"train_loss": 0.19900672137737274, "global_step": 75011, "epoch": 842, "lr": 8.247207178890264e-05}
{"train_loss": 0.20791012048721313, "global_step": 75012, "epoch": 842, "lr": 8.247163094827413e-05}
{"train_loss": 0.2250548154115677, "global_step": 75013, "epoch": 842, "lr": 8.247119010328018e-05}
{"train_loss": 0.2695530354976654, "global_step": 75014, "epoch": 842, "lr": 8.24707492539209e-05}
{"train_loss": 0.3398239314556122, "global_step": 75015, "epoch": 842, "lr": 8.247030840019632e-05}
{"train_loss": 0.19270265102386475, "global_step": 75016, "epoch": 842, "lr": 8.246986754210648e-05}
{"train_loss": 0.21947257220745087, "global_step": 75017, "epoch": 842, "lr": 8.246942667965149e-05}
{"train_loss": 0.23237738013267517, "global_step": 75018, "epoch": 842, "lr": 8.246898581283136e-05}
{"train_loss": 0.19991014897823334, "global_step": 75019, "epoch": 842, "lr": 8.24685449416462e-05}
{"train_loss": 0.15288375318050385, "global_step": 75020, "epoch": 842, "lr": 8.2468104066096e-05}
{"train_loss": 0.1815323531627655, "global_step": 75021, "epoch": 842, "lr": 8.246766318618089e-05}
{"train_loss": 0.1860496550798416, "global_step": 75022, "epoch": 842, "lr": 8.24672223019009e-05}
{"train_loss": 0.21529562771320343, "global_step": 75023, "epoch": 842, "lr": 8.246678141325606e-05}
{"train_loss": 0.1982010155916214, "global_step": 75024, "epoch": 842, "lr": 8.246634052024646e-05}
{"train_loss": 0.1969609260559082, "global_step": 75025, "epoch": 842, "lr": 8.246589962287219e-05}
{"train_loss": 0.21006794298967618, "global_step": 75026, "epoch": 842, "lr": 8.246545872113325e-05, "val_loss": 3.9639182090759277}
{"train_loss": 0.27870190143585205, "global_step": 75027, "epoch": 843, "lr": 8.246501781502973e-05}
{"train_loss": 0.2484852522611618, "global_step": 75028, "epoch": 843, "lr": 8.246457690456168e-05}
{"train_loss": 0.2934584617614746, "global_step": 75029, "epoch": 843, "lr": 8.246413598972918e-05}
{"train_loss": 0.12755927443504333, "global_step": 75030, "epoch": 843, "lr": 8.246369507053224e-05}
{"train_loss": 0.250266432762146, "global_step": 75031, "epoch": 843, "lr": 8.246325414697098e-05}
{"train_loss": 0.2866043150424957, "global_step": 75032, "epoch": 843, "lr": 8.246281321904542e-05}
{"train_loss": 0.20458148419857025, "global_step": 75033, "epoch": 843, "lr": 8.246237228675563e-05}
{"train_loss": 0.22081637382507324, "global_step": 75034, "epoch": 843, "lr": 8.246193135010168e-05}
{"train_loss": 0.12697762250900269, "global_step": 75035, "epoch": 843, "lr": 8.24614904090836e-05}
{"train_loss": 0.14050571620464325, "global_step": 75036, "epoch": 843, "lr": 8.24610494637015e-05}
{"train_loss": 0.2595464885234833, "global_step": 75037, "epoch": 843, "lr": 8.246060851395539e-05}
{"train_loss": 0.20861618220806122, "global_step": 75038, "epoch": 843, "lr": 8.246016755984533e-05}
{"train_loss": 0.20706816017627716, "global_step": 75039, "epoch": 843, "lr": 8.245972660137141e-05}
{"train_loss": 0.21131236851215363, "global_step": 75040, "epoch": 843, "lr": 8.245928563853368e-05}
{"train_loss": 0.3283790051937103, "global_step": 75041, "epoch": 843, "lr": 8.245884467133218e-05}
{"train_loss": 0.19542451202869415, "global_step": 75042, "epoch": 843, "lr": 8.245840369976699e-05}
{"train_loss": 0.23254843056201935, "global_step": 75043, "epoch": 843, "lr": 8.245796272383816e-05}
{"train_loss": 0.20703890919685364, "global_step": 75044, "epoch": 843, "lr": 8.245752174354577e-05}
{"train_loss": 0.18946783244609833, "global_step": 75045, "epoch": 843, "lr": 8.245708075888986e-05}
{"train_loss": 0.21195240318775177, "global_step": 75046, "epoch": 843, "lr": 8.245663976987047e-05}
{"train_loss": 0.25558385252952576, "global_step": 75047, "epoch": 843, "lr": 8.24561987764877e-05}
{"train_loss": 0.24522243440151215, "global_step": 75048, "epoch": 843, "lr": 8.24557577787416e-05}
{"train_loss": 0.20704923570156097, "global_step": 75049, "epoch": 843, "lr": 8.245531677663218e-05}
{"train_loss": 0.18996253609657288, "global_step": 75050, "epoch": 843, "lr": 8.245487577015956e-05}
{"train_loss": 0.17553015053272247, "global_step": 75051, "epoch": 843, "lr": 8.245443475932379e-05}
{"train_loss": 0.29792940616607666, "global_step": 75052, "epoch": 843, "lr": 8.24539937441249e-05}
{"train_loss": 0.21557210385799408, "global_step": 75053, "epoch": 843, "lr": 8.245355272456298e-05}
{"train_loss": 0.21060366928577423, "global_step": 75054, "epoch": 843, "lr": 8.245311170063806e-05}
{"train_loss": 0.2612055242061615, "global_step": 75055, "epoch": 843, "lr": 8.245267067235023e-05}
{"train_loss": 0.1896982491016388, "global_step": 75056, "epoch": 843, "lr": 8.245222963969952e-05}
{"train_loss": 0.22109679877758026, "global_step": 75057, "epoch": 843, "lr": 8.245178860268602e-05}
{"train_loss": 0.23453165590763092, "global_step": 75058, "epoch": 843, "lr": 8.245134756130974e-05}
{"train_loss": 0.19455137848854065, "global_step": 75059, "epoch": 843, "lr": 8.245090651557079e-05}
{"train_loss": 0.21646669507026672, "global_step": 75060, "epoch": 843, "lr": 8.245046546546921e-05}
{"train_loss": 0.24737535417079926, "global_step": 75061, "epoch": 843, "lr": 8.245002441100506e-05}
{"train_loss": 0.2921077311038971, "global_step": 75062, "epoch": 843, "lr": 8.244958335217841e-05}
{"train_loss": 0.13851310312747955, "global_step": 75063, "epoch": 843, "lr": 8.24491422889893e-05}
{"train_loss": 0.1647636592388153, "global_step": 75064, "epoch": 843, "lr": 8.24487012214378e-05}
{"train_loss": 0.23923474550247192, "global_step": 75065, "epoch": 843, "lr": 8.244826014952399e-05}
{"train_loss": 0.18044260144233704, "global_step": 75066, "epoch": 843, "lr": 8.244781907324788e-05}
{"train_loss": 0.09500092267990112, "global_step": 75067, "epoch": 843, "lr": 8.244737799260955e-05}
{"train_loss": 0.27182161808013916, "global_step": 75068, "epoch": 843, "lr": 8.244693690760907e-05}
{"train_loss": 0.17200174927711487, "global_step": 75069, "epoch": 843, "lr": 8.24464958182465e-05}
{"train_loss": 0.1565742790699005, "global_step": 75070, "epoch": 843, "lr": 8.244605472452191e-05}
{"train_loss": 0.29082563519477844, "global_step": 75071, "epoch": 843, "lr": 8.244561362643532e-05}
{"train_loss": 0.20446659624576569, "global_step": 75072, "epoch": 843, "lr": 8.244517252398684e-05}
{"train_loss": 0.2234889417886734, "global_step": 75073, "epoch": 843, "lr": 8.244473141717648e-05}
{"train_loss": 0.26330992579460144, "global_step": 75074, "epoch": 843, "lr": 8.244429030600432e-05}
{"train_loss": 0.2010340392589569, "global_step": 75075, "epoch": 843, "lr": 8.244384919047043e-05}
{"train_loss": 0.2447718381881714, "global_step": 75076, "epoch": 843, "lr": 8.244340807057486e-05}
{"train_loss": 0.21544045209884644, "global_step": 75077, "epoch": 843, "lr": 8.244296694631766e-05}
{"train_loss": 0.20430536568164825, "global_step": 75078, "epoch": 843, "lr": 8.244252581769889e-05}
{"train_loss": 0.2163325846195221, "global_step": 75079, "epoch": 843, "lr": 8.244208468471864e-05}
{"train_loss": 0.23518502712249756, "global_step": 75080, "epoch": 843, "lr": 8.244164354737693e-05}
{"train_loss": 0.22119960188865662, "global_step": 75081, "epoch": 843, "lr": 8.244120240567385e-05}
{"train_loss": 0.1546296924352646, "global_step": 75082, "epoch": 843, "lr": 8.244076125960944e-05}
{"train_loss": 0.20613926649093628, "global_step": 75083, "epoch": 843, "lr": 8.244032010918377e-05}
{"train_loss": 0.2159223109483719, "global_step": 75084, "epoch": 843, "lr": 8.243987895439688e-05}
{"train_loss": 0.24876169860363007, "global_step": 75085, "epoch": 843, "lr": 8.243943779524887e-05}
{"train_loss": 0.13646380603313446, "global_step": 75086, "epoch": 843, "lr": 8.243899663173975e-05}
{"train_loss": 0.2858697175979614, "global_step": 75087, "epoch": 843, "lr": 8.24385554638696e-05}
{"train_loss": 0.20947472751140594, "global_step": 75088, "epoch": 843, "lr": 8.24381142916385e-05}
{"train_loss": 0.24275535345077515, "global_step": 75089, "epoch": 843, "lr": 8.243767311504649e-05}
{"train_loss": 0.17609618604183197, "global_step": 75090, "epoch": 843, "lr": 8.243723193409362e-05}
{"train_loss": 0.2119438797235489, "global_step": 75091, "epoch": 843, "lr": 8.243679074877997e-05}
{"train_loss": 0.2133336216211319, "global_step": 75092, "epoch": 843, "lr": 8.243634955910557e-05}
{"train_loss": 0.18802866339683533, "global_step": 75093, "epoch": 843, "lr": 8.243590836507052e-05}
{"train_loss": 0.20474204421043396, "global_step": 75094, "epoch": 843, "lr": 8.243546716667483e-05}
{"train_loss": 0.25187021493911743, "global_step": 75095, "epoch": 843, "lr": 8.24350259639186e-05}
{"train_loss": 0.20905500650405884, "global_step": 75096, "epoch": 843, "lr": 8.243458475680188e-05}
{"train_loss": 0.1455698013305664, "global_step": 75097, "epoch": 843, "lr": 8.243414354532473e-05}
{"train_loss": 0.18384061753749847, "global_step": 75098, "epoch": 843, "lr": 8.24337023294872e-05}
{"train_loss": 0.250937283039093, "global_step": 75099, "epoch": 843, "lr": 8.243326110928936e-05}
{"train_loss": 0.12540803849697113, "global_step": 75100, "epoch": 843, "lr": 8.243281988473123e-05}
{"train_loss": 0.2945488393306732, "global_step": 75101, "epoch": 843, "lr": 8.243237865581295e-05}
{"train_loss": 0.13315242528915405, "global_step": 75102, "epoch": 843, "lr": 8.24319374225345e-05}
{"train_loss": 0.26203078031539917, "global_step": 75103, "epoch": 843, "lr": 8.243149618489598e-05}
{"train_loss": 0.16093002259731293, "global_step": 75104, "epoch": 843, "lr": 8.243105494289745e-05}
{"train_loss": 0.23154205083847046, "global_step": 75105, "epoch": 843, "lr": 8.243061369653895e-05}
{"train_loss": 0.13416798412799835, "global_step": 75106, "epoch": 843, "lr": 8.243017244582056e-05}
{"train_loss": 0.14513461291790009, "global_step": 75107, "epoch": 843, "lr": 8.24297311907423e-05}
{"train_loss": 0.15274189412593842, "global_step": 75108, "epoch": 843, "lr": 8.24292899313043e-05}
{"train_loss": 0.21786443889141083, "global_step": 75109, "epoch": 843, "lr": 8.242884866750655e-05}
{"train_loss": 0.1628810614347458, "global_step": 75110, "epoch": 843, "lr": 8.242840739934914e-05}
{"train_loss": 0.22665482759475708, "global_step": 75111, "epoch": 843, "lr": 8.242796612683213e-05}
{"train_loss": 0.18048137426376343, "global_step": 75112, "epoch": 843, "lr": 8.242752484995557e-05}
{"train_loss": 0.20842856168746948, "global_step": 75113, "epoch": 843, "lr": 8.242708356871953e-05}
{"train_loss": 0.15142706036567688, "global_step": 75114, "epoch": 843, "lr": 8.242664228312405e-05}
{"train_loss": 0.21138443203454607, "global_step": 75115, "epoch": 843, "lr": 8.242620099316922e-05, "val_loss": 4.1378278732299805}
{"train_loss": 0.24691292643547058, "global_step": 75116, "epoch": 844, "lr": 8.242575969885509e-05}
{"train_loss": 0.15415385365486145, "global_step": 75117, "epoch": 844, "lr": 8.242531840018169e-05}
{"train_loss": 0.17895177006721497, "global_step": 75118, "epoch": 844, "lr": 8.242487709714911e-05}
{"train_loss": 0.18508478999137878, "global_step": 75119, "epoch": 844, "lr": 8.24244357897574e-05}
{"train_loss": 0.22305597364902496, "global_step": 75120, "epoch": 844, "lr": 8.242399447800662e-05}
{"train_loss": 0.09549195319414139, "global_step": 75121, "epoch": 844, "lr": 8.242355316189683e-05}
{"train_loss": 0.12996366620063782, "global_step": 75122, "epoch": 844, "lr": 8.24231118414281e-05}
{"train_loss": 0.1487826406955719, "global_step": 75123, "epoch": 844, "lr": 8.242267051660046e-05}
{"train_loss": 0.17803435027599335, "global_step": 75124, "epoch": 844, "lr": 8.242222918741399e-05}
{"train_loss": 0.24966755509376526, "global_step": 75125, "epoch": 844, "lr": 8.242178785386876e-05}
{"train_loss": 0.2373918890953064, "global_step": 75126, "epoch": 844, "lr": 8.242134651596481e-05}
{"train_loss": 0.18358366191387177, "global_step": 75127, "epoch": 844, "lr": 8.242090517370219e-05}
{"train_loss": 0.15029537677764893, "global_step": 75128, "epoch": 844, "lr": 8.242046382708099e-05}
{"train_loss": 0.22277089953422546, "global_step": 75129, "epoch": 844, "lr": 8.242002247610123e-05}
{"train_loss": 0.25093016028404236, "global_step": 75130, "epoch": 844, "lr": 8.241958112076303e-05}
{"train_loss": 0.2091352790594101, "global_step": 75131, "epoch": 844, "lr": 8.241913976106639e-05}
{"train_loss": 0.160322904586792, "global_step": 75132, "epoch": 844, "lr": 8.24186983970114e-05}
{"train_loss": 0.16336382925510406, "global_step": 75133, "epoch": 844, "lr": 8.24182570285981e-05}
{"train_loss": 0.20093868672847748, "global_step": 75134, "epoch": 844, "lr": 8.241781565582655e-05}
{"train_loss": 0.23879672586917877, "global_step": 75135, "epoch": 844, "lr": 8.241737427869684e-05}
{"train_loss": 0.27184391021728516, "global_step": 75136, "epoch": 844, "lr": 8.241693289720902e-05}
{"train_loss": 0.24450838565826416, "global_step": 75137, "epoch": 844, "lr": 8.241649151136312e-05}
{"train_loss": 0.30399343371391296, "global_step": 75138, "epoch": 844, "lr": 8.241605012115922e-05}
{"train_loss": 0.2891727089881897, "global_step": 75139, "epoch": 844, "lr": 8.24156087265974e-05}
{"train_loss": 0.28951701521873474, "global_step": 75140, "epoch": 844, "lr": 8.241516732767766e-05}
{"train_loss": 0.1859637349843979, "global_step": 75141, "epoch": 844, "lr": 8.241472592440011e-05}
{"train_loss": 0.20630143582820892, "global_step": 75142, "epoch": 844, "lr": 8.24142845167648e-05}
{"train_loss": 0.2824109196662903, "global_step": 75143, "epoch": 844, "lr": 8.241384310477179e-05}
{"train_loss": 0.23545493185520172, "global_step": 75144, "epoch": 844, "lr": 8.241340168842113e-05}
{"train_loss": 0.3119361400604248, "global_step": 75145, "epoch": 844, "lr": 8.241296026771288e-05}
{"train_loss": 0.2072264850139618, "global_step": 75146, "epoch": 844, "lr": 8.241251884264709e-05}
{"train_loss": 0.25539612770080566, "global_step": 75147, "epoch": 844, "lr": 8.241207741322386e-05}
{"train_loss": 0.22967977821826935, "global_step": 75148, "epoch": 844, "lr": 8.241163597944319e-05}
{"train_loss": 0.24524065852165222, "global_step": 75149, "epoch": 844, "lr": 8.241119454130519e-05}
{"train_loss": 0.17671459913253784, "global_step": 75150, "epoch": 844, "lr": 8.24107530988099e-05}
{"train_loss": 0.11610806733369827, "global_step": 75151, "epoch": 844, "lr": 8.241031165195737e-05}
{"train_loss": 0.2373497486114502, "global_step": 75152, "epoch": 844, "lr": 8.240987020074766e-05}
{"train_loss": 0.1648777425289154, "global_step": 75153, "epoch": 844, "lr": 8.240942874518085e-05}
{"train_loss": 0.17433686554431915, "global_step": 75154, "epoch": 844, "lr": 8.2408987285257e-05}
{"train_loss": 0.2133726328611374, "global_step": 75155, "epoch": 844, "lr": 8.240854582097615e-05}
{"train_loss": 0.1404980719089508, "global_step": 75156, "epoch": 844, "lr": 8.240810435233836e-05}
{"train_loss": 0.25323963165283203, "global_step": 75157, "epoch": 844, "lr": 8.24076628793437e-05}
{"train_loss": 0.26251745223999023, "global_step": 75158, "epoch": 844, "lr": 8.240722140199222e-05}
{"train_loss": 0.13941225409507751, "global_step": 75159, "epoch": 844, "lr": 8.240677992028398e-05}
{"train_loss": 0.21988935768604279, "global_step": 75160, "epoch": 844, "lr": 8.240633843421905e-05}
{"train_loss": 0.2494819313287735, "global_step": 75161, "epoch": 844, "lr": 8.240589694379748e-05}
{"train_loss": 0.1899007260799408, "global_step": 75162, "epoch": 844, "lr": 8.240545544901934e-05}
{"train_loss": 0.2001819908618927, "global_step": 75163, "epoch": 844, "lr": 8.240501394988467e-05}
{"train_loss": 0.2449975311756134, "global_step": 75164, "epoch": 844, "lr": 8.240457244639354e-05}
{"train_loss": 0.2056027501821518, "global_step": 75165, "epoch": 844, "lr": 8.240413093854602e-05}
{"train_loss": 0.20062156021595, "global_step": 75166, "epoch": 844, "lr": 8.240368942634215e-05}
{"train_loss": 0.19858236610889435, "global_step": 75167, "epoch": 844, "lr": 8.240324790978202e-05}
{"train_loss": 0.1776195615530014, "global_step": 75168, "epoch": 844, "lr": 8.240280638886564e-05}
{"train_loss": 0.23030927777290344, "global_step": 75169, "epoch": 844, "lr": 8.240236486359312e-05}
{"train_loss": 0.20894542336463928, "global_step": 75170, "epoch": 844, "lr": 8.240192333396448e-05}
{"train_loss": 0.2989189922809601, "global_step": 75171, "epoch": 844, "lr": 8.240148179997981e-05}
{"train_loss": 0.2898368835449219, "global_step": 75172, "epoch": 844, "lr": 8.240104026163916e-05}
{"train_loss": 0.2536225914955139, "global_step": 75173, "epoch": 844, "lr": 8.240059871894257e-05}
{"train_loss": 0.2369094341993332, "global_step": 75174, "epoch": 844, "lr": 8.240015717189011e-05}
{"train_loss": 0.17198342084884644, "global_step": 75175, "epoch": 844, "lr": 8.239971562048185e-05}
{"train_loss": 0.22122269868850708, "global_step": 75176, "epoch": 844, "lr": 8.239927406471786e-05}
{"train_loss": 0.18860040605068207, "global_step": 75177, "epoch": 844, "lr": 8.239883250459817e-05}
{"train_loss": 0.20185832679271698, "global_step": 75178, "epoch": 844, "lr": 8.239839094012285e-05}
{"train_loss": 0.1614609658718109, "global_step": 75179, "epoch": 844, "lr": 8.239794937129195e-05}
{"train_loss": 0.13745784759521484, "global_step": 75180, "epoch": 844, "lr": 8.239750779810556e-05}
{"train_loss": 0.16061392426490784, "global_step": 75181, "epoch": 844, "lr": 8.239706622056372e-05}
{"train_loss": 0.24757640063762665, "global_step": 75182, "epoch": 844, "lr": 8.239662463866646e-05}
{"train_loss": 0.19616009294986725, "global_step": 75183, "epoch": 844, "lr": 8.23961830524139e-05}
{"train_loss": 0.2062242478132248, "global_step": 75184, "epoch": 844, "lr": 8.239574146180607e-05}
{"train_loss": 0.3527602255344391, "global_step": 75185, "epoch": 844, "lr": 8.2395299866843e-05}
{"train_loss": 0.2206353396177292, "global_step": 75186, "epoch": 844, "lr": 8.239485826752478e-05}
{"train_loss": 0.2746610641479492, "global_step": 75187, "epoch": 844, "lr": 8.239441666385148e-05}
{"train_loss": 0.25028613209724426, "global_step": 75188, "epoch": 844, "lr": 8.239397505582316e-05}
{"train_loss": 0.16914069652557373, "global_step": 75189, "epoch": 844, "lr": 8.239353344343985e-05}
{"train_loss": 0.24369245767593384, "global_step": 75190, "epoch": 844, "lr": 8.23930918267016e-05}
{"train_loss": 0.23432962596416473, "global_step": 75191, "epoch": 844, "lr": 8.239265020560853e-05}
{"train_loss": 0.19752554595470428, "global_step": 75192, "epoch": 844, "lr": 8.239220858016065e-05}
{"train_loss": 0.2273232638835907, "global_step": 75193, "epoch": 844, "lr": 8.239176695035804e-05}
{"train_loss": 0.23756778240203857, "global_step": 75194, "epoch": 844, "lr": 8.239132531620073e-05}
{"train_loss": 0.16776669025421143, "global_step": 75195, "epoch": 844, "lr": 8.239088367768882e-05}
{"train_loss": 0.21272404491901398, "global_step": 75196, "epoch": 844, "lr": 8.239044203482236e-05}
{"train_loss": 0.13283772766590118, "global_step": 75197, "epoch": 844, "lr": 8.239000038760137e-05}
{"train_loss": 0.15866611897945404, "global_step": 75198, "epoch": 844, "lr": 8.238955873602597e-05}
{"train_loss": 0.2707442343235016, "global_step": 75199, "epoch": 844, "lr": 8.238911708009616e-05}
{"train_loss": 0.24140766263008118, "global_step": 75200, "epoch": 844, "lr": 8.238867541981205e-05}
{"train_loss": 0.24236787855625153, "global_step": 75201, "epoch": 844, "lr": 8.238823375517366e-05}
{"train_loss": 0.23589368164539337, "global_step": 75202, "epoch": 844, "lr": 8.238779208618108e-05}
{"train_loss": 0.26040393114089966, "global_step": 75203, "epoch": 844, "lr": 8.238735041283434e-05}
{"train_loss": 0.214594655157475, "global_step": 75204, "epoch": 844, "lr": 8.238690873513352e-05, "val_loss": 3.9886088371276855}
{"train_loss": 0.16990332305431366, "global_step": 75205, "epoch": 845, "lr": 8.238646705307869e-05}
{"train_loss": 0.1854528784751892, "global_step": 75206, "epoch": 845, "lr": 8.238602536666989e-05}
{"train_loss": 0.22054968774318695, "global_step": 75207, "epoch": 845, "lr": 8.238558367590719e-05}
{"train_loss": 0.1521466076374054, "global_step": 75208, "epoch": 845, "lr": 8.238514198079061e-05}
{"train_loss": 0.23536749184131622, "global_step": 75209, "epoch": 845, "lr": 8.238470028132028e-05}
{"train_loss": 0.2546822726726532, "global_step": 75210, "epoch": 845, "lr": 8.238425857749621e-05}
{"train_loss": 0.3212527930736542, "global_step": 75211, "epoch": 845, "lr": 8.238381686931846e-05}
{"train_loss": 0.1907683163881302, "global_step": 75212, "epoch": 845, "lr": 8.23833751567871e-05}
{"train_loss": 0.25364935398101807, "global_step": 75213, "epoch": 845, "lr": 8.23829334399022e-05}
{"train_loss": 0.18654124438762665, "global_step": 75214, "epoch": 845, "lr": 8.238249171866382e-05}
{"train_loss": 0.1791534572839737, "global_step": 75215, "epoch": 845, "lr": 8.238204999307199e-05}
{"train_loss": 0.30839982628822327, "global_step": 75216, "epoch": 845, "lr": 8.238160826312679e-05}
{"train_loss": 0.2558128833770752, "global_step": 75217, "epoch": 845, "lr": 8.23811665288283e-05}
{"train_loss": 0.2744004726409912, "global_step": 75218, "epoch": 845, "lr": 8.238072479017653e-05}
{"train_loss": 0.24988558888435364, "global_step": 75219, "epoch": 845, "lr": 8.238028304717158e-05}
{"train_loss": 0.1993451565504074, "global_step": 75220, "epoch": 845, "lr": 8.237984129981348e-05}
{"train_loss": 0.1876945197582245, "global_step": 75221, "epoch": 845, "lr": 8.237939954810232e-05}
{"train_loss": 0.1539095938205719, "global_step": 75222, "epoch": 845, "lr": 8.237895779203813e-05}
{"train_loss": 0.17535392940044403, "global_step": 75223, "epoch": 845, "lr": 8.237851603162099e-05}
{"train_loss": 0.10739514231681824, "global_step": 75224, "epoch": 845, "lr": 8.237807426685096e-05}
{"train_loss": 0.17042918503284454, "global_step": 75225, "epoch": 845, "lr": 8.23776324977281e-05}
{"train_loss": 0.2928570806980133, "global_step": 75226, "epoch": 845, "lr": 8.237719072425245e-05}
{"train_loss": 0.19436687231063843, "global_step": 75227, "epoch": 845, "lr": 8.237674894642407e-05}
{"train_loss": 0.30877479910850525, "global_step": 75228, "epoch": 845, "lr": 8.237630716424305e-05}
{"train_loss": 0.22598102688789368, "global_step": 75229, "epoch": 845, "lr": 8.237586537770943e-05}
{"train_loss": 0.2034498155117035, "global_step": 75230, "epoch": 845, "lr": 8.237542358682325e-05}
{"train_loss": 0.14027458429336548, "global_step": 75231, "epoch": 845, "lr": 8.23749817915846e-05}
{"train_loss": 0.17154428362846375, "global_step": 75232, "epoch": 845, "lr": 8.237453999199354e-05}
{"train_loss": 0.17520807683467865, "global_step": 75233, "epoch": 845, "lr": 8.23740981880501e-05}
{"train_loss": 0.12789073586463928, "global_step": 75234, "epoch": 845, "lr": 8.237365637975437e-05}
{"train_loss": 0.2925892770290375, "global_step": 75235, "epoch": 845, "lr": 8.23732145671064e-05}
{"train_loss": 0.3222919702529907, "global_step": 75236, "epoch": 845, "lr": 8.237277275010623e-05}
{"train_loss": 0.12721695005893707, "global_step": 75237, "epoch": 845, "lr": 8.237233092875394e-05}
{"train_loss": 0.15299789607524872, "global_step": 75238, "epoch": 845, "lr": 8.23718891030496e-05}
{"train_loss": 0.16200266778469086, "global_step": 75239, "epoch": 845, "lr": 8.237144727299323e-05}
{"train_loss": 0.26951777935028076, "global_step": 75240, "epoch": 845, "lr": 8.237100543858491e-05}
{"train_loss": 0.2281104177236557, "global_step": 75241, "epoch": 845, "lr": 8.237056359982474e-05}
{"train_loss": 0.26844921708106995, "global_step": 75242, "epoch": 845, "lr": 8.23701217567127e-05}
{"train_loss": 0.21110227704048157, "global_step": 75243, "epoch": 845, "lr": 8.236967990924892e-05}
{"train_loss": 0.19669698178768158, "global_step": 75244, "epoch": 845, "lr": 8.236923805743342e-05}
{"train_loss": 0.14177004992961884, "global_step": 75245, "epoch": 845, "lr": 8.236879620126628e-05}
{"train_loss": 0.2235521525144577, "global_step": 75246, "epoch": 845, "lr": 8.236835434074755e-05}
{"train_loss": 0.1326656937599182, "global_step": 75247, "epoch": 845, "lr": 8.236791247587728e-05}
{"train_loss": 0.3064891993999481, "global_step": 75248, "epoch": 845, "lr": 8.236747060665553e-05}
{"train_loss": 0.2636997103691101, "global_step": 75249, "epoch": 845, "lr": 8.236702873308238e-05}
{"train_loss": 0.18665270507335663, "global_step": 75250, "epoch": 845, "lr": 8.236658685515788e-05}
{"train_loss": 0.2437339574098587, "global_step": 75251, "epoch": 845, "lr": 8.236614497288209e-05}
{"train_loss": 0.19928182661533356, "global_step": 75252, "epoch": 845, "lr": 8.236570308625504e-05}
{"train_loss": 0.19552618265151978, "global_step": 75253, "epoch": 845, "lr": 8.236526119527685e-05}
{"train_loss": 0.19709432125091553, "global_step": 75254, "epoch": 845, "lr": 8.236481929994753e-05}
{"train_loss": 0.19424618780612946, "global_step": 75255, "epoch": 845, "lr": 8.236437740026715e-05}
{"train_loss": 0.17870566248893738, "global_step": 75256, "epoch": 845, "lr": 8.236393549623577e-05}
{"train_loss": 0.10896699130535126, "global_step": 75257, "epoch": 845, "lr": 8.236349358785348e-05}
{"train_loss": 0.24360866844654083, "global_step": 75258, "epoch": 845, "lr": 8.236305167512029e-05}
{"train_loss": 0.20990866422653198, "global_step": 75259, "epoch": 845, "lr": 8.236260975803627e-05}
{"train_loss": 0.1430969089269638, "global_step": 75260, "epoch": 845, "lr": 8.23621678366015e-05}
{"train_loss": 0.21470443904399872, "global_step": 75261, "epoch": 845, "lr": 8.236172591081606e-05}
{"train_loss": 0.164383664727211, "global_step": 75262, "epoch": 845, "lr": 8.236128398067996e-05}
{"train_loss": 0.22740286588668823, "global_step": 75263, "epoch": 845, "lr": 8.236084204619327e-05}
{"train_loss": 0.14072193205356598, "global_step": 75264, "epoch": 845, "lr": 8.236040010735608e-05}
{"train_loss": 0.17592817544937134, "global_step": 75265, "epoch": 845, "lr": 8.235995816416841e-05}
{"train_loss": 0.19498474895954132, "global_step": 75266, "epoch": 845, "lr": 8.235951621663034e-05}
{"train_loss": 0.197636678814888, "global_step": 75267, "epoch": 845, "lr": 8.235907426474194e-05}
{"train_loss": 0.26790478825569153, "global_step": 75268, "epoch": 845, "lr": 8.235863230850325e-05}
{"train_loss": 0.1930399090051651, "global_step": 75269, "epoch": 845, "lr": 8.235819034791434e-05}
{"train_loss": 0.24148117005825043, "global_step": 75270, "epoch": 845, "lr": 8.235774838297526e-05}
{"train_loss": 0.1970844864845276, "global_step": 75271, "epoch": 845, "lr": 8.235730641368608e-05}
{"train_loss": 0.18268954753875732, "global_step": 75272, "epoch": 845, "lr": 8.235686444004686e-05}
{"train_loss": 0.1259969174861908, "global_step": 75273, "epoch": 845, "lr": 8.235642246205764e-05}
{"train_loss": 0.25298669934272766, "global_step": 75274, "epoch": 845, "lr": 8.23559804797185e-05}
{"train_loss": 0.19504934549331665, "global_step": 75275, "epoch": 845, "lr": 8.235553849302948e-05}
{"train_loss": 0.15147428214550018, "global_step": 75276, "epoch": 845, "lr": 8.235509650199066e-05}
{"train_loss": 0.1637781858444214, "global_step": 75277, "epoch": 845, "lr": 8.235465450660209e-05}
{"train_loss": 0.1558883786201477, "global_step": 75278, "epoch": 845, "lr": 8.235421250686385e-05}
{"train_loss": 0.1928098201751709, "global_step": 75279, "epoch": 845, "lr": 8.235377050277598e-05}
{"train_loss": 0.2913905680179596, "global_step": 75280, "epoch": 845, "lr": 8.235332849433852e-05}
{"train_loss": 0.2154652625322342, "global_step": 75281, "epoch": 845, "lr": 8.235288648155155e-05}
{"train_loss": 0.25735220313072205, "global_step": 75282, "epoch": 845, "lr": 8.235244446441513e-05}
{"train_loss": 0.1881026327610016, "global_step": 75283, "epoch": 845, "lr": 8.235200244292934e-05}
{"train_loss": 0.20386981964111328, "global_step": 75284, "epoch": 845, "lr": 8.235156041709419e-05}
{"train_loss": 0.18712592124938965, "global_step": 75285, "epoch": 845, "lr": 8.235111838690977e-05}
{"train_loss": 0.1589634269475937, "global_step": 75286, "epoch": 845, "lr": 8.235067635237615e-05}
{"train_loss": 0.15553171932697296, "global_step": 75287, "epoch": 845, "lr": 8.235023431349337e-05}
{"train_loss": 0.1809324324131012, "global_step": 75288, "epoch": 845, "lr": 8.234979227026149e-05}
{"train_loss": 0.14413267374038696, "global_step": 75289, "epoch": 845, "lr": 8.234935022268058e-05}
{"train_loss": 0.11753003299236298, "global_step": 75290, "epoch": 845, "lr": 8.23489081707507e-05}
{"train_loss": 0.18768471479415894, "global_step": 75291, "epoch": 845, "lr": 8.23484661144719e-05}
{"train_loss": 0.24748308956623077, "global_step": 75292, "epoch": 845, "lr": 8.234802405384424e-05}
{"train_loss": 0.2022778909192996, "global_step": 75293, "epoch": 845, "lr": 8.234758198886779e-05, "val_loss": 4.030359745025635, "train_action_mse_error": 30.932371139526367}
{"train_loss": 0.24638991057872772, "global_step": 75294, "epoch": 846, "lr": 8.234713991954259e-05}
{"train_loss": 0.13799194991588593, "global_step": 75295, "epoch": 846, "lr": 8.23466978458687e-05}
{"train_loss": 0.1993435174226761, "global_step": 75296, "epoch": 846, "lr": 8.234625576784622e-05}
{"train_loss": 0.13807767629623413, "global_step": 75297, "epoch": 846, "lr": 8.234581368547516e-05}
{"train_loss": 0.2339007705450058, "global_step": 75298, "epoch": 846, "lr": 8.234537159875562e-05}
{"train_loss": 0.20635293424129486, "global_step": 75299, "epoch": 846, "lr": 8.234492950768761e-05}
{"train_loss": 0.20980998873710632, "global_step": 75300, "epoch": 846, "lr": 8.234448741227124e-05}
{"train_loss": 0.18751230835914612, "global_step": 75301, "epoch": 846, "lr": 8.234404531250655e-05}
{"train_loss": 0.21000680327415466, "global_step": 75302, "epoch": 846, "lr": 8.234360320839359e-05}
{"train_loss": 0.22045879065990448, "global_step": 75303, "epoch": 846, "lr": 8.234316109993242e-05}
{"train_loss": 0.290226548910141, "global_step": 75304, "epoch": 846, "lr": 8.234271898712311e-05}
{"train_loss": 0.22403059899806976, "global_step": 75305, "epoch": 846, "lr": 8.234227686996572e-05}
{"train_loss": 0.23314815759658813, "global_step": 75306, "epoch": 846, "lr": 8.234183474846031e-05}
{"train_loss": 0.15092767775058746, "global_step": 75307, "epoch": 846, "lr": 8.234139262260691e-05}
{"train_loss": 0.23376524448394775, "global_step": 75308, "epoch": 846, "lr": 8.234095049240564e-05}
{"train_loss": 0.21133141219615936, "global_step": 75309, "epoch": 846, "lr": 8.234050835785649e-05}
{"train_loss": 0.23580466210842133, "global_step": 75310, "epoch": 846, "lr": 8.234006621895958e-05}
{"train_loss": 0.1819687932729721, "global_step": 75311, "epoch": 846, "lr": 8.233962407571493e-05}
{"train_loss": 0.13505856692790985, "global_step": 75312, "epoch": 846, "lr": 8.233918192812262e-05}
{"train_loss": 0.20191597938537598, "global_step": 75313, "epoch": 846, "lr": 8.233873977618268e-05}
{"train_loss": 0.18404443562030792, "global_step": 75314, "epoch": 846, "lr": 8.23382976198952e-05}
{"train_loss": 0.2541326582431793, "global_step": 75315, "epoch": 846, "lr": 8.233785545926023e-05}
{"train_loss": 0.17591412365436554, "global_step": 75316, "epoch": 846, "lr": 8.233741329427784e-05}
{"train_loss": 0.19899919629096985, "global_step": 75317, "epoch": 846, "lr": 8.233697112494807e-05}
{"train_loss": 0.174537792801857, "global_step": 75318, "epoch": 846, "lr": 8.2336528951271e-05}
{"train_loss": 0.28360122442245483, "global_step": 75319, "epoch": 846, "lr": 8.233608677324666e-05}
{"train_loss": 0.27912744879722595, "global_step": 75320, "epoch": 846, "lr": 8.233564459087513e-05}
{"train_loss": 0.23427489399909973, "global_step": 75321, "epoch": 846, "lr": 8.233520240415648e-05}
{"train_loss": 0.18587246537208557, "global_step": 75322, "epoch": 846, "lr": 8.233476021309075e-05}
{"train_loss": 0.23001940548419952, "global_step": 75323, "epoch": 846, "lr": 8.233431801767799e-05}
{"train_loss": 0.25431329011917114, "global_step": 75324, "epoch": 846, "lr": 8.23338758179183e-05}
{"train_loss": 0.26406654715538025, "global_step": 75325, "epoch": 846, "lr": 8.23334336138117e-05}
{"train_loss": 0.16218216717243195, "global_step": 75326, "epoch": 846, "lr": 8.233299140535826e-05}
{"train_loss": 0.24868233501911163, "global_step": 75327, "epoch": 846, "lr": 8.233254919255806e-05}
{"train_loss": 0.16706214845180511, "global_step": 75328, "epoch": 846, "lr": 8.233210697541114e-05}
{"train_loss": 0.24020622670650482, "global_step": 75329, "epoch": 846, "lr": 8.233166475391754e-05}
{"train_loss": 0.2235819399356842, "global_step": 75330, "epoch": 846, "lr": 8.233122252807737e-05}
{"train_loss": 0.18625488877296448, "global_step": 75331, "epoch": 846, "lr": 8.233078029789064e-05}
{"train_loss": 0.1912558376789093, "global_step": 75332, "epoch": 846, "lr": 8.233033806335744e-05}
{"train_loss": 0.21676301956176758, "global_step": 75333, "epoch": 846, "lr": 8.232989582447782e-05}
{"train_loss": 0.1756918579339981, "global_step": 75334, "epoch": 846, "lr": 8.232945358125184e-05}
{"train_loss": 0.25295916199684143, "global_step": 75335, "epoch": 846, "lr": 8.232901133367956e-05}
{"train_loss": 0.15725742280483246, "global_step": 75336, "epoch": 846, "lr": 8.232856908176103e-05}
{"train_loss": 0.2765439450740814, "global_step": 75337, "epoch": 846, "lr": 8.232812682549633e-05}
{"train_loss": 0.22385069727897644, "global_step": 75338, "epoch": 846, "lr": 8.23276845648855e-05}
{"train_loss": 0.17536446452140808, "global_step": 75339, "epoch": 846, "lr": 8.232724229992862e-05}
{"train_loss": 0.14672352373600006, "global_step": 75340, "epoch": 846, "lr": 8.232680003062572e-05}
{"train_loss": 0.188775897026062, "global_step": 75341, "epoch": 846, "lr": 8.232635775697689e-05}
{"train_loss": 0.15197691321372986, "global_step": 75342, "epoch": 846, "lr": 8.232591547898215e-05}
{"train_loss": 0.19936800003051758, "global_step": 75343, "epoch": 846, "lr": 8.232547319664162e-05}
{"train_loss": 0.18930858373641968, "global_step": 75344, "epoch": 846, "lr": 8.23250309099553e-05}
{"train_loss": 0.18172213435173035, "global_step": 75345, "epoch": 846, "lr": 8.232458861892328e-05}
{"train_loss": 0.17337274551391602, "global_step": 75346, "epoch": 846, "lr": 8.23241463235456e-05}
{"train_loss": 0.2882877588272095, "global_step": 75347, "epoch": 846, "lr": 8.232370402382234e-05}
{"train_loss": 0.23548538982868195, "global_step": 75348, "epoch": 846, "lr": 8.232326171975355e-05}
{"train_loss": 0.21154452860355377, "global_step": 75349, "epoch": 846, "lr": 8.232281941133931e-05}
{"train_loss": 0.20653900504112244, "global_step": 75350, "epoch": 846, "lr": 8.232237709857965e-05}
{"train_loss": 0.13957422971725464, "global_step": 75351, "epoch": 846, "lr": 8.232193478147462e-05}
{"train_loss": 0.20656560361385345, "global_step": 75352, "epoch": 846, "lr": 8.232149246002432e-05}
{"train_loss": 0.19757652282714844, "global_step": 75353, "epoch": 846, "lr": 8.23210501342288e-05}
{"train_loss": 0.1877492070198059, "global_step": 75354, "epoch": 846, "lr": 8.232060780408809e-05}
{"train_loss": 0.21024540066719055, "global_step": 75355, "epoch": 846, "lr": 8.232016546960228e-05}
{"train_loss": 0.19057202339172363, "global_step": 75356, "epoch": 846, "lr": 8.23197231307714e-05}
{"train_loss": 0.20489799976348877, "global_step": 75357, "epoch": 846, "lr": 8.231928078759555e-05}
{"train_loss": 0.20893749594688416, "global_step": 75358, "epoch": 846, "lr": 8.231883844007475e-05}
{"train_loss": 0.19413121044635773, "global_step": 75359, "epoch": 846, "lr": 8.231839608820909e-05}
{"train_loss": 0.13464702665805817, "global_step": 75360, "epoch": 846, "lr": 8.23179537319986e-05}
{"train_loss": 0.21903395652770996, "global_step": 75361, "epoch": 846, "lr": 8.231751137144336e-05}
{"train_loss": 0.2365807443857193, "global_step": 75362, "epoch": 846, "lr": 8.231706900654343e-05}
{"train_loss": 0.2712790071964264, "global_step": 75363, "epoch": 846, "lr": 8.231662663729887e-05}
{"train_loss": 0.16025955975055695, "global_step": 75364, "epoch": 846, "lr": 8.231618426370972e-05}
{"train_loss": 0.2883286476135254, "global_step": 75365, "epoch": 846, "lr": 8.231574188577605e-05}
{"train_loss": 0.1381317377090454, "global_step": 75366, "epoch": 846, "lr": 8.231529950349793e-05}
{"train_loss": 0.21083036065101624, "global_step": 75367, "epoch": 846, "lr": 8.231485711687541e-05}
{"train_loss": 0.15684889256954193, "global_step": 75368, "epoch": 846, "lr": 8.231441472590855e-05}
{"train_loss": 0.13110952079296112, "global_step": 75369, "epoch": 846, "lr": 8.231397233059741e-05}
{"train_loss": 0.22093044221401215, "global_step": 75370, "epoch": 846, "lr": 8.231352993094206e-05}
{"train_loss": 0.1516542285680771, "global_step": 75371, "epoch": 846, "lr": 8.231308752694254e-05}
{"train_loss": 0.21277804672718048, "global_step": 75372, "epoch": 846, "lr": 8.231264511859893e-05}
{"train_loss": 0.26813241839408875, "global_step": 75373, "epoch": 846, "lr": 8.231220270591127e-05}
{"train_loss": 0.2092372626066208, "global_step": 75374, "epoch": 846, "lr": 8.231176028887964e-05}
{"train_loss": 0.2199408859014511, "global_step": 75375, "epoch": 846, "lr": 8.231131786750407e-05}
{"train_loss": 0.15731194615364075, "global_step": 75376, "epoch": 846, "lr": 8.231087544178465e-05}
{"train_loss": 0.2227519154548645, "global_step": 75377, "epoch": 846, "lr": 8.231043301172141e-05}
{"train_loss": 0.23095189034938812, "global_step": 75378, "epoch": 846, "lr": 8.230999057731444e-05}
{"train_loss": 0.19429226219654083, "global_step": 75379, "epoch": 846, "lr": 8.23095481385638e-05}
{"train_loss": 0.1792302280664444, "global_step": 75380, "epoch": 846, "lr": 8.230910569546951e-05}
{"train_loss": 0.1639336347579956, "global_step": 75381, "epoch": 846, "lr": 8.230866324803166e-05}
{"train_loss": 0.20586088031865238, "global_step": 75382, "epoch": 846, "lr": 8.230822079625031e-05, "val_loss": 3.9068195819854736}
{"train_loss": 0.34776571393013, "global_step": 75383, "epoch": 847, "lr": 8.23077783401255e-05}
{"train_loss": 0.22229556739330292, "global_step": 75384, "epoch": 847, "lr": 8.230733587965731e-05}
{"train_loss": 0.20824988186359406, "global_step": 75385, "epoch": 847, "lr": 8.23068934148458e-05}
{"train_loss": 0.30241015553474426, "global_step": 75386, "epoch": 847, "lr": 8.230645094569102e-05}
{"train_loss": 0.19031670689582825, "global_step": 75387, "epoch": 847, "lr": 8.230600847219303e-05}
{"train_loss": 0.20105627179145813, "global_step": 75388, "epoch": 847, "lr": 8.230556599435188e-05}
{"train_loss": 0.1930091828107834, "global_step": 75389, "epoch": 847, "lr": 8.230512351216766e-05}
{"train_loss": 0.24221022427082062, "global_step": 75390, "epoch": 847, "lr": 8.23046810256404e-05}
{"train_loss": 0.22126881778240204, "global_step": 75391, "epoch": 847, "lr": 8.230423853477016e-05}
{"train_loss": 0.23354528844356537, "global_step": 75392, "epoch": 847, "lr": 8.230379603955703e-05}
{"train_loss": 0.21142670512199402, "global_step": 75393, "epoch": 847, "lr": 8.230335354000103e-05}
{"train_loss": 0.22191005945205688, "global_step": 75394, "epoch": 847, "lr": 8.230291103610224e-05}
{"train_loss": 0.21357132494449615, "global_step": 75395, "epoch": 847, "lr": 8.230246852786073e-05}
{"train_loss": 0.2047576606273651, "global_step": 75396, "epoch": 847, "lr": 8.230202601527655e-05}
{"train_loss": 0.20341990888118744, "global_step": 75397, "epoch": 847, "lr": 8.230158349834973e-05}
{"train_loss": 0.21413345634937286, "global_step": 75398, "epoch": 847, "lr": 8.230114097708038e-05}
{"train_loss": 0.20993107557296753, "global_step": 75399, "epoch": 847, "lr": 8.230069845146853e-05}
{"train_loss": 0.184651181101799, "global_step": 75400, "epoch": 847, "lr": 8.230025592151423e-05}
{"train_loss": 0.28532764315605164, "global_step": 75401, "epoch": 847, "lr": 8.229981338721757e-05}
{"train_loss": 0.17475442588329315, "global_step": 75402, "epoch": 847, "lr": 8.229937084857858e-05}
{"train_loss": 0.22107790410518646, "global_step": 75403, "epoch": 847, "lr": 8.229892830559734e-05}
{"train_loss": 0.21416731178760529, "global_step": 75404, "epoch": 847, "lr": 8.22984857582739e-05}
{"train_loss": 0.20270240306854248, "global_step": 75405, "epoch": 847, "lr": 8.229804320660834e-05}
{"train_loss": 0.16426829993724823, "global_step": 75406, "epoch": 847, "lr": 8.229760065060068e-05}
{"train_loss": 0.17418624460697174, "global_step": 75407, "epoch": 847, "lr": 8.229715809025102e-05}
{"train_loss": 0.20499782264232635, "global_step": 75408, "epoch": 847, "lr": 8.229671552555938e-05}
{"train_loss": 0.20799201726913452, "global_step": 75409, "epoch": 847, "lr": 8.229627295652586e-05}
{"train_loss": 0.16589021682739258, "global_step": 75410, "epoch": 847, "lr": 8.229583038315049e-05}
{"train_loss": 0.17098316550254822, "global_step": 75411, "epoch": 847, "lr": 8.229538780543334e-05}
{"train_loss": 0.2030085027217865, "global_step": 75412, "epoch": 847, "lr": 8.229494522337446e-05}
{"train_loss": 0.20997042953968048, "global_step": 75413, "epoch": 847, "lr": 8.229450263697393e-05}
{"train_loss": 0.21885943412780762, "global_step": 75414, "epoch": 847, "lr": 8.229406004623179e-05}
{"train_loss": 0.3199826180934906, "global_step": 75415, "epoch": 847, "lr": 8.229361745114812e-05}
{"train_loss": 0.1790638566017151, "global_step": 75416, "epoch": 847, "lr": 8.229317485172296e-05}
{"train_loss": 0.2152177095413208, "global_step": 75417, "epoch": 847, "lr": 8.229273224795637e-05}
{"train_loss": 0.19200152158737183, "global_step": 75418, "epoch": 847, "lr": 8.229228963984843e-05}
{"train_loss": 0.21653887629508972, "global_step": 75419, "epoch": 847, "lr": 8.229184702739918e-05}
{"train_loss": 0.2574114501476288, "global_step": 75420, "epoch": 847, "lr": 8.229140441060867e-05}
{"train_loss": 0.20613163709640503, "global_step": 75421, "epoch": 847, "lr": 8.2290961789477e-05}
{"train_loss": 0.22198176383972168, "global_step": 75422, "epoch": 847, "lr": 8.229051916400416e-05}
{"train_loss": 0.24605253338813782, "global_step": 75423, "epoch": 847, "lr": 8.22900765341903e-05}
{"train_loss": 0.1566651463508606, "global_step": 75424, "epoch": 847, "lr": 8.22896339000354e-05}
{"train_loss": 0.1948101818561554, "global_step": 75425, "epoch": 847, "lr": 8.228919126153958e-05}
{"train_loss": 0.14122390747070312, "global_step": 75426, "epoch": 847, "lr": 8.228874861870286e-05}
{"train_loss": 0.18597574532032013, "global_step": 75427, "epoch": 847, "lr": 8.22883059715253e-05}
{"train_loss": 0.29363304376602173, "global_step": 75428, "epoch": 847, "lr": 8.228786332000699e-05}
{"train_loss": 0.1478816419839859, "global_step": 75429, "epoch": 847, "lr": 8.228742066414796e-05}
{"train_loss": 0.18674176931381226, "global_step": 75430, "epoch": 847, "lr": 8.228697800394828e-05}
{"train_loss": 0.24286377429962158, "global_step": 75431, "epoch": 847, "lr": 8.2286535339408e-05}
{"train_loss": 0.22524140775203705, "global_step": 75432, "epoch": 847, "lr": 8.228609267052721e-05}
{"train_loss": 0.2025178074836731, "global_step": 75433, "epoch": 847, "lr": 8.228564999730594e-05}
{"train_loss": 0.16367821395397186, "global_step": 75434, "epoch": 847, "lr": 8.228520731974425e-05}
{"train_loss": 0.2282937467098236, "global_step": 75435, "epoch": 847, "lr": 8.228476463784222e-05}
{"train_loss": 0.32240501046180725, "global_step": 75436, "epoch": 847, "lr": 8.228432195159989e-05}
{"train_loss": 0.2271871566772461, "global_step": 75437, "epoch": 847, "lr": 8.228387926101733e-05}
{"train_loss": 0.16154134273529053, "global_step": 75438, "epoch": 847, "lr": 8.228343656609458e-05}
{"train_loss": 0.2710464596748352, "global_step": 75439, "epoch": 847, "lr": 8.228299386683174e-05}
{"train_loss": 0.22852347791194916, "global_step": 75440, "epoch": 847, "lr": 8.228255116322882e-05}
{"train_loss": 0.2074439972639084, "global_step": 75441, "epoch": 847, "lr": 8.228210845528593e-05}
{"train_loss": 0.2053622305393219, "global_step": 75442, "epoch": 847, "lr": 8.228166574300308e-05}
{"train_loss": 0.1580355167388916, "global_step": 75443, "epoch": 847, "lr": 8.228122302638036e-05}
{"train_loss": 0.22274409234523773, "global_step": 75444, "epoch": 847, "lr": 8.228078030541782e-05}
{"train_loss": 0.22257697582244873, "global_step": 75445, "epoch": 847, "lr": 8.228033758011555e-05}
{"train_loss": 0.20795580744743347, "global_step": 75446, "epoch": 847, "lr": 8.227989485047355e-05}
{"train_loss": 0.15222366154193878, "global_step": 75447, "epoch": 847, "lr": 8.227945211649193e-05}
{"train_loss": 0.18945741653442383, "global_step": 75448, "epoch": 847, "lr": 8.227900937817073e-05}
{"train_loss": 0.19295471906661987, "global_step": 75449, "epoch": 847, "lr": 8.227856663550998e-05}
{"train_loss": 0.22819094359874725, "global_step": 75450, "epoch": 847, "lr": 8.227812388850982e-05}
{"train_loss": 0.19460079073905945, "global_step": 75451, "epoch": 847, "lr": 8.227768113717022e-05}
{"train_loss": 0.20460879802703857, "global_step": 75452, "epoch": 847, "lr": 8.22772383814913e-05}
{"train_loss": 0.1854313164949417, "global_step": 75453, "epoch": 847, "lr": 8.22767956214731e-05}
{"train_loss": 0.275115966796875, "global_step": 75454, "epoch": 847, "lr": 8.227635285711567e-05}
{"train_loss": 0.20509107410907745, "global_step": 75455, "epoch": 847, "lr": 8.227591008841908e-05}
{"train_loss": 0.2802667021751404, "global_step": 75456, "epoch": 847, "lr": 8.227546731538338e-05}
{"train_loss": 0.1329362690448761, "global_step": 75457, "epoch": 847, "lr": 8.227502453800864e-05}
{"train_loss": 0.17119862139225006, "global_step": 75458, "epoch": 847, "lr": 8.227458175629492e-05}
{"train_loss": 0.28531980514526367, "global_step": 75459, "epoch": 847, "lr": 8.227413897024227e-05}
{"train_loss": 0.2581798732280731, "global_step": 75460, "epoch": 847, "lr": 8.227369617985076e-05}
{"train_loss": 0.2636074721813202, "global_step": 75461, "epoch": 847, "lr": 8.227325338512045e-05}
{"train_loss": 0.19978754222393036, "global_step": 75462, "epoch": 847, "lr": 8.227281058605138e-05}
{"train_loss": 0.13427719473838806, "global_step": 75463, "epoch": 847, "lr": 8.227236778264364e-05}
{"train_loss": 0.20172631740570068, "global_step": 75464, "epoch": 847, "lr": 8.227192497489727e-05}
{"train_loss": 0.21349124610424042, "global_step": 75465, "epoch": 847, "lr": 8.227148216281233e-05}
{"train_loss": 0.2540115714073181, "global_step": 75466, "epoch": 847, "lr": 8.227103934638888e-05}
{"train_loss": 0.20875529944896698, "global_step": 75467, "epoch": 847, "lr": 8.227059652562699e-05}
{"train_loss": 0.23487845063209534, "global_step": 75468, "epoch": 847, "lr": 8.227015370052671e-05}
{"train_loss": 0.19191612303256989, "global_step": 75469, "epoch": 847, "lr": 8.22697108710881e-05}
{"train_loss": 0.16688643395900726, "global_step": 75470, "epoch": 847, "lr": 8.226926803731121e-05}
{"train_loss": 0.2128840122330055, "global_step": 75471, "epoch": 847, "lr": 8.226882519919612e-05, "val_loss": 3.882884979248047}
{"train_loss": 0.1926887482404709, "global_step": 75472, "epoch": 848, "lr": 8.226838235674289e-05}
{"train_loss": 0.19394677877426147, "global_step": 75473, "epoch": 848, "lr": 8.226793950995156e-05}
{"train_loss": 0.23724707961082458, "global_step": 75474, "epoch": 848, "lr": 8.226749665882219e-05}
{"train_loss": 0.20347487926483154, "global_step": 75475, "epoch": 848, "lr": 8.226705380335486e-05}
{"train_loss": 0.13150730729103088, "global_step": 75476, "epoch": 848, "lr": 8.226661094354959e-05}
{"train_loss": 0.15412452816963196, "global_step": 75477, "epoch": 848, "lr": 8.22661680794065e-05}
{"train_loss": 0.20743045210838318, "global_step": 75478, "epoch": 848, "lr": 8.22657252109256e-05}
{"train_loss": 0.17482008039951324, "global_step": 75479, "epoch": 848, "lr": 8.226528233810698e-05}
{"train_loss": 0.18797509372234344, "global_step": 75480, "epoch": 848, "lr": 8.226483946095069e-05}
{"train_loss": 0.16777130961418152, "global_step": 75481, "epoch": 848, "lr": 8.226439657945677e-05}
{"train_loss": 0.23662126064300537, "global_step": 75482, "epoch": 848, "lr": 8.22639536936253e-05}
{"train_loss": 0.21071386337280273, "global_step": 75483, "epoch": 848, "lr": 8.226351080345633e-05}
{"train_loss": 0.19717304408550262, "global_step": 75484, "epoch": 848, "lr": 8.226306790894993e-05}
{"train_loss": 0.13200899958610535, "global_step": 75485, "epoch": 848, "lr": 8.226262501010615e-05}
{"train_loss": 0.19785816967487335, "global_step": 75486, "epoch": 848, "lr": 8.226218210692507e-05}
{"train_loss": 0.23376934230327606, "global_step": 75487, "epoch": 848, "lr": 8.226173919940671e-05}
{"train_loss": 0.2240324169397354, "global_step": 75488, "epoch": 848, "lr": 8.226129628755117e-05}
{"train_loss": 0.2048773467540741, "global_step": 75489, "epoch": 848, "lr": 8.226085337135847e-05}
{"train_loss": 0.16577400267124176, "global_step": 75490, "epoch": 848, "lr": 8.226041045082871e-05}
{"train_loss": 0.2099437713623047, "global_step": 75491, "epoch": 848, "lr": 8.225996752596192e-05}
{"train_loss": 0.2248881310224533, "global_step": 75492, "epoch": 848, "lr": 8.225952459675819e-05}
{"train_loss": 0.2452521026134491, "global_step": 75493, "epoch": 848, "lr": 8.225908166321754e-05}
{"train_loss": 0.2223253697156906, "global_step": 75494, "epoch": 848, "lr": 8.225863872534006e-05}
{"train_loss": 0.1323738992214203, "global_step": 75495, "epoch": 848, "lr": 8.225819578312579e-05}
{"train_loss": 0.19905807077884674, "global_step": 75496, "epoch": 848, "lr": 8.22577528365748e-05}
{"train_loss": 0.16735723614692688, "global_step": 75497, "epoch": 848, "lr": 8.225730988568716e-05}
{"train_loss": 0.13620629906654358, "global_step": 75498, "epoch": 848, "lr": 8.22568669304629e-05}
{"train_loss": 0.1426607370376587, "global_step": 75499, "epoch": 848, "lr": 8.225642397090212e-05}
{"train_loss": 0.28826820850372314, "global_step": 75500, "epoch": 848, "lr": 8.225598100700484e-05}
{"train_loss": 0.29516464471817017, "global_step": 75501, "epoch": 848, "lr": 8.225553803877115e-05}
{"train_loss": 0.19342729449272156, "global_step": 75502, "epoch": 848, "lr": 8.225509506620108e-05}
{"train_loss": 0.18205927312374115, "global_step": 75503, "epoch": 848, "lr": 8.225465208929471e-05}
{"train_loss": 0.17099808156490326, "global_step": 75504, "epoch": 848, "lr": 8.225420910805211e-05}
{"train_loss": 0.13598540425300598, "global_step": 75505, "epoch": 848, "lr": 8.225376612247331e-05}
{"train_loss": 0.18693111836910248, "global_step": 75506, "epoch": 848, "lr": 8.225332313255838e-05}
{"train_loss": 0.30006590485572815, "global_step": 75507, "epoch": 848, "lr": 8.225288013830738e-05}
{"train_loss": 0.181091770529747, "global_step": 75508, "epoch": 848, "lr": 8.225243713972039e-05}
{"train_loss": 0.22085075080394745, "global_step": 75509, "epoch": 848, "lr": 8.225199413679746e-05}
{"train_loss": 0.1791747659444809, "global_step": 75510, "epoch": 848, "lr": 8.225155112953862e-05}
{"train_loss": 0.2857319116592407, "global_step": 75511, "epoch": 848, "lr": 8.225110811794397e-05}
{"train_loss": 0.19617825746536255, "global_step": 75512, "epoch": 848, "lr": 8.225066510201354e-05}
{"train_loss": 0.2356135994195938, "global_step": 75513, "epoch": 848, "lr": 8.22502220817474e-05}
{"train_loss": 0.1918685883283615, "global_step": 75514, "epoch": 848, "lr": 8.224977905714561e-05}
{"train_loss": 0.18688805401325226, "global_step": 75515, "epoch": 848, "lr": 8.224933602820823e-05}
{"train_loss": 0.20085598528385162, "global_step": 75516, "epoch": 848, "lr": 8.224889299493534e-05}
{"train_loss": 0.1663639098405838, "global_step": 75517, "epoch": 848, "lr": 8.224844995732696e-05}
{"train_loss": 0.2745170593261719, "global_step": 75518, "epoch": 848, "lr": 8.224800691538317e-05}
{"train_loss": 0.1717115044593811, "global_step": 75519, "epoch": 848, "lr": 8.224756386910404e-05}
{"train_loss": 0.20212364196777344, "global_step": 75520, "epoch": 848, "lr": 8.224712081848961e-05}
{"train_loss": 0.3475548326969147, "global_step": 75521, "epoch": 848, "lr": 8.224667776353995e-05}
{"train_loss": 0.2253691852092743, "global_step": 75522, "epoch": 848, "lr": 8.224623470425511e-05}
{"train_loss": 0.21436549723148346, "global_step": 75523, "epoch": 848, "lr": 8.224579164063516e-05}
{"train_loss": 0.13659599423408508, "global_step": 75524, "epoch": 848, "lr": 8.224534857268016e-05}
{"train_loss": 0.37833839654922485, "global_step": 75525, "epoch": 848, "lr": 8.224490550039017e-05}
{"train_loss": 0.23419617116451263, "global_step": 75526, "epoch": 848, "lr": 8.224446242376523e-05}
{"train_loss": 0.28205469250679016, "global_step": 75527, "epoch": 848, "lr": 8.224401934280543e-05}
{"train_loss": 0.18683411180973053, "global_step": 75528, "epoch": 848, "lr": 8.22435762575108e-05}
{"train_loss": 0.17025813460350037, "global_step": 75529, "epoch": 848, "lr": 8.224313316788143e-05}
{"train_loss": 0.1899820864200592, "global_step": 75530, "epoch": 848, "lr": 8.224269007391736e-05}
{"train_loss": 0.23757793009281158, "global_step": 75531, "epoch": 848, "lr": 8.224224697561864e-05}
{"train_loss": 0.23164288699626923, "global_step": 75532, "epoch": 848, "lr": 8.224180387298536e-05}
{"train_loss": 0.17163658142089844, "global_step": 75533, "epoch": 848, "lr": 8.224136076601757e-05}
{"train_loss": 0.20665311813354492, "global_step": 75534, "epoch": 848, "lr": 8.224091765471529e-05}
{"train_loss": 0.19650998711585999, "global_step": 75535, "epoch": 848, "lr": 8.224047453907864e-05}
{"train_loss": 0.2085079848766327, "global_step": 75536, "epoch": 848, "lr": 8.224003141910763e-05}
{"train_loss": 0.3495779037475586, "global_step": 75537, "epoch": 848, "lr": 8.223958829480235e-05}
{"train_loss": 0.30108949542045593, "global_step": 75538, "epoch": 848, "lr": 8.223914516616286e-05}
{"train_loss": 0.17291633784770966, "global_step": 75539, "epoch": 848, "lr": 8.22387020331892e-05}
{"train_loss": 0.2833569645881653, "global_step": 75540, "epoch": 848, "lr": 8.223825889588144e-05}
{"train_loss": 0.1975565403699875, "global_step": 75541, "epoch": 848, "lr": 8.223781575423964e-05}
{"train_loss": 0.25370296835899353, "global_step": 75542, "epoch": 848, "lr": 8.223737260826386e-05}
{"train_loss": 0.21151801943778992, "global_step": 75543, "epoch": 848, "lr": 8.223692945795415e-05}
{"train_loss": 0.2390766441822052, "global_step": 75544, "epoch": 848, "lr": 8.223648630331059e-05}
{"train_loss": 0.23076362907886505, "global_step": 75545, "epoch": 848, "lr": 8.223604314433322e-05}
{"train_loss": 0.15374045073986053, "global_step": 75546, "epoch": 848, "lr": 8.22355999810221e-05}
{"train_loss": 0.23917971551418304, "global_step": 75547, "epoch": 848, "lr": 8.223515681337732e-05}
{"train_loss": 0.25733682513237, "global_step": 75548, "epoch": 848, "lr": 8.22347136413989e-05}
{"train_loss": 0.18446451425552368, "global_step": 75549, "epoch": 848, "lr": 8.223427046508692e-05}
{"train_loss": 0.13740606606006622, "global_step": 75550, "epoch": 848, "lr": 8.223382728444143e-05}
{"train_loss": 0.19933681190013885, "global_step": 75551, "epoch": 848, "lr": 8.223338409946248e-05}
{"train_loss": 0.168821319937706, "global_step": 75552, "epoch": 848, "lr": 8.223294091015016e-05}
{"train_loss": 0.25044891238212585, "global_step": 75553, "epoch": 848, "lr": 8.223249771650452e-05}
{"train_loss": 0.17335781455039978, "global_step": 75554, "epoch": 848, "lr": 8.223205451852562e-05}
{"train_loss": 0.23735810816287994, "global_step": 75555, "epoch": 848, "lr": 8.223161131621349e-05}
{"train_loss": 0.20214329659938812, "global_step": 75556, "epoch": 848, "lr": 8.223116810956824e-05}
{"train_loss": 0.2955937087535858, "global_step": 75557, "epoch": 848, "lr": 8.223072489858988e-05}
{"train_loss": 0.19329780340194702, "global_step": 75558, "epoch": 848, "lr": 8.22302816832785e-05}
{"train_loss": 0.2402215152978897, "global_step": 75559, "epoch": 848, "lr": 8.222983846363415e-05}
{"train_loss": 0.21069579529628324, "global_step": 75560, "epoch": 848, "lr": 8.22293952396569e-05, "val_loss": 3.9413058757781982}
{"train_loss": 0.15214619040489197, "global_step": 75561, "epoch": 849, "lr": 8.222895201134679e-05}
{"train_loss": 0.21791015565395355, "global_step": 75562, "epoch": 849, "lr": 8.222850877870389e-05}
{"train_loss": 0.23241527378559113, "global_step": 75563, "epoch": 849, "lr": 8.222806554172827e-05}
{"train_loss": 0.195623978972435, "global_step": 75564, "epoch": 849, "lr": 8.222762230041996e-05}
{"train_loss": 0.19341588020324707, "global_step": 75565, "epoch": 849, "lr": 8.222717905477904e-05}
{"train_loss": 0.2784985899925232, "global_step": 75566, "epoch": 849, "lr": 8.222673580480558e-05}
{"train_loss": 0.31554487347602844, "global_step": 75567, "epoch": 849, "lr": 8.222629255049962e-05}
{"train_loss": 0.18894247710704803, "global_step": 75568, "epoch": 849, "lr": 8.222584929186124e-05}
{"train_loss": 0.18520814180374146, "global_step": 75569, "epoch": 849, "lr": 8.222540602889049e-05}
{"train_loss": 0.18090061843395233, "global_step": 75570, "epoch": 849, "lr": 8.222496276158742e-05}
{"train_loss": 0.2341262847185135, "global_step": 75571, "epoch": 849, "lr": 8.222451948995209e-05}
{"train_loss": 0.19910499453544617, "global_step": 75572, "epoch": 849, "lr": 8.222407621398456e-05}
{"train_loss": 0.16721026599407196, "global_step": 75573, "epoch": 849, "lr": 8.222363293368492e-05}
{"train_loss": 0.24273060262203217, "global_step": 75574, "epoch": 849, "lr": 8.222318964905319e-05}
{"train_loss": 0.23840978741645813, "global_step": 75575, "epoch": 849, "lr": 8.222274636008944e-05}
{"train_loss": 0.3093084990978241, "global_step": 75576, "epoch": 849, "lr": 8.222230306679375e-05}
{"train_loss": 0.1836998015642166, "global_step": 75577, "epoch": 849, "lr": 8.222185976916616e-05}
{"train_loss": 0.15736819803714752, "global_step": 75578, "epoch": 849, "lr": 8.222141646720673e-05}
{"train_loss": 0.13376857340335846, "global_step": 75579, "epoch": 849, "lr": 8.222097316091552e-05}
{"train_loss": 0.20744316279888153, "global_step": 75580, "epoch": 849, "lr": 8.222052985029261e-05}
{"train_loss": 0.22016172111034393, "global_step": 75581, "epoch": 849, "lr": 8.222008653533801e-05}
{"train_loss": 0.21365654468536377, "global_step": 75582, "epoch": 849, "lr": 8.221964321605184e-05}
{"train_loss": 0.1605810523033142, "global_step": 75583, "epoch": 849, "lr": 8.221919989243411e-05}
{"train_loss": 0.21677754819393158, "global_step": 75584, "epoch": 849, "lr": 8.221875656448492e-05}
{"train_loss": 0.2099730670452118, "global_step": 75585, "epoch": 849, "lr": 8.22183132322043e-05}
{"train_loss": 0.21840216219425201, "global_step": 75586, "epoch": 849, "lr": 8.221786989559233e-05}
{"train_loss": 0.13440370559692383, "global_step": 75587, "epoch": 849, "lr": 8.221742655464905e-05}
{"train_loss": 0.1546057164669037, "global_step": 75588, "epoch": 849, "lr": 8.221698320937454e-05}
{"train_loss": 0.25007399916648865, "global_step": 75589, "epoch": 849, "lr": 8.221653985976884e-05}
{"train_loss": 0.17267629504203796, "global_step": 75590, "epoch": 849, "lr": 8.221609650583201e-05}
{"train_loss": 0.17396984994411469, "global_step": 75591, "epoch": 849, "lr": 8.221565314756415e-05}
{"train_loss": 0.18923157453536987, "global_step": 75592, "epoch": 849, "lr": 8.221520978496526e-05}
{"train_loss": 0.22663205862045288, "global_step": 75593, "epoch": 849, "lr": 8.221476641803544e-05}
{"train_loss": 0.14978544414043427, "global_step": 75594, "epoch": 849, "lr": 8.221432304677474e-05}
{"train_loss": 0.28670769929885864, "global_step": 75595, "epoch": 849, "lr": 8.221387967118321e-05}
{"train_loss": 0.32384800910949707, "global_step": 75596, "epoch": 849, "lr": 8.221343629126091e-05}
{"train_loss": 0.25549980998039246, "global_step": 75597, "epoch": 849, "lr": 8.221299290700792e-05}
{"train_loss": 0.13680915534496307, "global_step": 75598, "epoch": 849, "lr": 8.221254951842427e-05}
{"train_loss": 0.18042269349098206, "global_step": 75599, "epoch": 849, "lr": 8.221210612551005e-05}
{"train_loss": 0.16859516501426697, "global_step": 75600, "epoch": 849, "lr": 8.221166272826531e-05}
{"train_loss": 0.20829594135284424, "global_step": 75601, "epoch": 849, "lr": 8.221121932669009e-05}
{"train_loss": 0.199208065867424, "global_step": 75602, "epoch": 849, "lr": 8.221077592078447e-05}
{"train_loss": 0.17689523100852966, "global_step": 75603, "epoch": 849, "lr": 8.221033251054849e-05}
{"train_loss": 0.1858951449394226, "global_step": 75604, "epoch": 849, "lr": 8.220988909598223e-05}
{"train_loss": 0.22262535989284515, "global_step": 75605, "epoch": 849, "lr": 8.220944567708576e-05}
{"train_loss": 0.20557191967964172, "global_step": 75606, "epoch": 849, "lr": 8.22090022538591e-05}
{"train_loss": 0.16888655722141266, "global_step": 75607, "epoch": 849, "lr": 8.220855882630235e-05}
{"train_loss": 0.15848273038864136, "global_step": 75608, "epoch": 849, "lr": 8.220811539441555e-05}
{"train_loss": 0.2080085426568985, "global_step": 75609, "epoch": 849, "lr": 8.220767195819877e-05}
{"train_loss": 0.1740819364786148, "global_step": 75610, "epoch": 849, "lr": 8.220722851765203e-05}
{"train_loss": 0.1770586520433426, "global_step": 75611, "epoch": 849, "lr": 8.220678507277544e-05}
{"train_loss": 0.2545333504676819, "global_step": 75612, "epoch": 849, "lr": 8.220634162356904e-05}
{"train_loss": 0.21231025457382202, "global_step": 75613, "epoch": 849, "lr": 8.220589817003289e-05}
{"train_loss": 0.14526551961898804, "global_step": 75614, "epoch": 849, "lr": 8.220545471216705e-05}
{"train_loss": 0.14618036150932312, "global_step": 75615, "epoch": 849, "lr": 8.220501124997158e-05}
{"train_loss": 0.09893552958965302, "global_step": 75616, "epoch": 849, "lr": 8.220456778344653e-05}
{"train_loss": 0.23909872770309448, "global_step": 75617, "epoch": 849, "lr": 8.220412431259197e-05}
{"train_loss": 0.23933248221874237, "global_step": 75618, "epoch": 849, "lr": 8.220368083740798e-05}
{"train_loss": 0.209974005818367, "global_step": 75619, "epoch": 849, "lr": 8.220323735789457e-05}
{"train_loss": 0.18531076610088348, "global_step": 75620, "epoch": 849, "lr": 8.220279387405183e-05}
{"train_loss": 0.16390445828437805, "global_step": 75621, "epoch": 849, "lr": 8.220235038587983e-05}
{"train_loss": 0.21053048968315125, "global_step": 75622, "epoch": 849, "lr": 8.220190689337861e-05}
{"train_loss": 0.12683641910552979, "global_step": 75623, "epoch": 849, "lr": 8.220146339654823e-05}
{"train_loss": 0.15658843517303467, "global_step": 75624, "epoch": 849, "lr": 8.220101989538877e-05}
{"train_loss": 0.20819100737571716, "global_step": 75625, "epoch": 849, "lr": 8.220057638990026e-05}
{"train_loss": 0.22263751924037933, "global_step": 75626, "epoch": 849, "lr": 8.22001328800828e-05}
{"train_loss": 0.19642947614192963, "global_step": 75627, "epoch": 849, "lr": 8.21996893659364e-05}
{"train_loss": 0.20464637875556946, "global_step": 75628, "epoch": 849, "lr": 8.219924584746115e-05}
{"train_loss": 0.2930203974246979, "global_step": 75629, "epoch": 849, "lr": 8.219880232465711e-05}
{"train_loss": 0.21715272963047028, "global_step": 75630, "epoch": 849, "lr": 8.219835879752433e-05}
{"train_loss": 0.17643241584300995, "global_step": 75631, "epoch": 849, "lr": 8.219791526606286e-05}
{"train_loss": 0.24723803997039795, "global_step": 75632, "epoch": 849, "lr": 8.219747173027279e-05}
{"train_loss": 0.1753571629524231, "global_step": 75633, "epoch": 849, "lr": 8.219702819015416e-05}
{"train_loss": 0.20871670544147491, "global_step": 75634, "epoch": 849, "lr": 8.219658464570703e-05}
{"train_loss": 0.17320409417152405, "global_step": 75635, "epoch": 849, "lr": 8.219614109693146e-05}
{"train_loss": 0.1620902270078659, "global_step": 75636, "epoch": 849, "lr": 8.21956975438275e-05}
{"train_loss": 0.35597676038742065, "global_step": 75637, "epoch": 849, "lr": 8.219525398639525e-05}
{"train_loss": 0.16773302853107452, "global_step": 75638, "epoch": 849, "lr": 8.219481042463471e-05}
{"train_loss": 0.23145203292369843, "global_step": 75639, "epoch": 849, "lr": 8.219436685854598e-05}
{"train_loss": 0.17890602350234985, "global_step": 75640, "epoch": 849, "lr": 8.219392328812911e-05}
{"train_loss": 0.20589759945869446, "global_step": 75641, "epoch": 849, "lr": 8.219347971338417e-05}
{"train_loss": 0.2193118780851364, "global_step": 75642, "epoch": 849, "lr": 8.21930361343112e-05}
{"train_loss": 0.15515564382076263, "global_step": 75643, "epoch": 849, "lr": 8.219259255091028e-05}
{"train_loss": 0.18535053730010986, "global_step": 75644, "epoch": 849, "lr": 8.219214896318143e-05}
{"train_loss": 0.20159119367599487, "global_step": 75645, "epoch": 849, "lr": 8.219170537112478e-05}
{"train_loss": 0.22411125898361206, "global_step": 75646, "epoch": 849, "lr": 8.219126177474031e-05}
{"train_loss": 0.2182430922985077, "global_step": 75647, "epoch": 849, "lr": 8.219081817402815e-05}
{"train_loss": 0.19225944578647614, "global_step": 75648, "epoch": 849, "lr": 8.219037456898829e-05}
{"train_loss": 0.20328733911005298, "global_step": 75649, "epoch": 849, "lr": 8.218993095962084e-05, "val_loss": 4.122375011444092}
{"train_loss": 0.22819411754608154, "global_step": 75650, "epoch": 850, "lr": 8.218948734592586e-05}
{"train_loss": 0.13629262149333954, "global_step": 75651, "epoch": 850, "lr": 8.218904372790339e-05}
{"train_loss": 0.1588633507490158, "global_step": 75652, "epoch": 850, "lr": 8.218860010555349e-05}
{"train_loss": 0.14522109925746918, "global_step": 75653, "epoch": 850, "lr": 8.218815647887624e-05}
{"train_loss": 0.25327661633491516, "global_step": 75654, "epoch": 850, "lr": 8.218771284787165e-05}
{"train_loss": 0.18893490731716156, "global_step": 75655, "epoch": 850, "lr": 8.218726921253985e-05}
{"train_loss": 0.16263066232204437, "global_step": 75656, "epoch": 850, "lr": 8.218682557288084e-05}
{"train_loss": 0.18591800332069397, "global_step": 75657, "epoch": 850, "lr": 8.218638192889472e-05}
{"train_loss": 0.20673494040966034, "global_step": 75658, "epoch": 850, "lr": 8.218593828058153e-05}
{"train_loss": 0.15872973203659058, "global_step": 75659, "epoch": 850, "lr": 8.218549462794132e-05}
{"train_loss": 0.25802114605903625, "global_step": 75660, "epoch": 850, "lr": 8.218505097097417e-05}
{"train_loss": 0.1896219402551651, "global_step": 75661, "epoch": 850, "lr": 8.218460730968014e-05}
{"train_loss": 0.21043024957180023, "global_step": 75662, "epoch": 850, "lr": 8.218416364405928e-05}
{"train_loss": 0.14166134595870972, "global_step": 75663, "epoch": 850, "lr": 8.218371997411166e-05}
{"train_loss": 0.16709738969802856, "global_step": 75664, "epoch": 850, "lr": 8.218327629983731e-05}
{"train_loss": 0.27343496680259705, "global_step": 75665, "epoch": 850, "lr": 8.218283262123631e-05}
{"train_loss": 0.17084094882011414, "global_step": 75666, "epoch": 850, "lr": 8.218238893830874e-05}
{"train_loss": 0.1675797998905182, "global_step": 75667, "epoch": 850, "lr": 8.218194525105463e-05}
{"train_loss": 0.3292340934276581, "global_step": 75668, "epoch": 850, "lr": 8.218150155947404e-05}
{"train_loss": 0.2136741578578949, "global_step": 75669, "epoch": 850, "lr": 8.218105786356705e-05}
{"train_loss": 0.27798283100128174, "global_step": 75670, "epoch": 850, "lr": 8.218061416333372e-05}
{"train_loss": 0.2093178629875183, "global_step": 75671, "epoch": 850, "lr": 8.218017045877409e-05}
{"train_loss": 0.11320289969444275, "global_step": 75672, "epoch": 850, "lr": 8.21797267498882e-05}
{"train_loss": 0.22702866792678833, "global_step": 75673, "epoch": 850, "lr": 8.217928303667618e-05}
{"train_loss": 0.17382675409317017, "global_step": 75674, "epoch": 850, "lr": 8.217883931913803e-05}
{"train_loss": 0.20372554659843445, "global_step": 75675, "epoch": 850, "lr": 8.217839559727382e-05}
{"train_loss": 0.2054571807384491, "global_step": 75676, "epoch": 850, "lr": 8.217795187108362e-05}
{"train_loss": 0.16032318770885468, "global_step": 75677, "epoch": 850, "lr": 8.217750814056749e-05}
{"train_loss": 0.24951206147670746, "global_step": 75678, "epoch": 850, "lr": 8.21770644057255e-05}
{"train_loss": 0.2647457718849182, "global_step": 75679, "epoch": 850, "lr": 8.217662066655768e-05}
{"train_loss": 0.2876482903957367, "global_step": 75680, "epoch": 850, "lr": 8.217617692306412e-05}
{"train_loss": 0.16390672326087952, "global_step": 75681, "epoch": 850, "lr": 8.217573317524485e-05}
{"train_loss": 0.24032872915267944, "global_step": 75682, "epoch": 850, "lr": 8.217528942309994e-05}
{"train_loss": 0.2800600230693817, "global_step": 75683, "epoch": 850, "lr": 8.217484566662947e-05}
{"train_loss": 0.28607988357543945, "global_step": 75684, "epoch": 850, "lr": 8.217440190583347e-05}
{"train_loss": 0.27612829208374023, "global_step": 75685, "epoch": 850, "lr": 8.217395814071202e-05}
{"train_loss": 0.2703891098499298, "global_step": 75686, "epoch": 850, "lr": 8.217351437126519e-05}
{"train_loss": 0.25032806396484375, "global_step": 75687, "epoch": 850, "lr": 8.2173070597493e-05}
{"train_loss": 0.24160520732402802, "global_step": 75688, "epoch": 850, "lr": 8.217262681939555e-05}
{"train_loss": 0.2455497831106186, "global_step": 75689, "epoch": 850, "lr": 8.217218303697287e-05}
{"train_loss": 0.22152367234230042, "global_step": 75690, "epoch": 850, "lr": 8.217173925022504e-05}
{"train_loss": 0.2194007933139801, "global_step": 75691, "epoch": 850, "lr": 8.21712954591521e-05}
{"train_loss": 0.2220325618982315, "global_step": 75692, "epoch": 850, "lr": 8.217085166375414e-05}
{"train_loss": 0.18796996772289276, "global_step": 75693, "epoch": 850, "lr": 8.217040786403119e-05}
{"train_loss": 0.2620275020599365, "global_step": 75694, "epoch": 850, "lr": 8.21699640599833e-05}
{"train_loss": 0.17072990536689758, "global_step": 75695, "epoch": 850, "lr": 8.216952025161058e-05}
{"train_loss": 0.24840547144412994, "global_step": 75696, "epoch": 850, "lr": 8.216907643891305e-05}
{"train_loss": 0.1301971673965454, "global_step": 75697, "epoch": 850, "lr": 8.216863262189078e-05}
{"train_loss": 0.18589676916599274, "global_step": 75698, "epoch": 850, "lr": 8.216818880054382e-05}
{"train_loss": 0.19715918600559235, "global_step": 75699, "epoch": 850, "lr": 8.216774497487225e-05}
{"train_loss": 0.2276402860879898, "global_step": 75700, "epoch": 850, "lr": 8.216730114487613e-05}
{"train_loss": 0.2124069631099701, "global_step": 75701, "epoch": 850, "lr": 8.216685731055549e-05}
{"train_loss": 0.21934598684310913, "global_step": 75702, "epoch": 850, "lr": 8.216641347191042e-05}
{"train_loss": 0.1435258835554123, "global_step": 75703, "epoch": 850, "lr": 8.216596962894096e-05}
{"train_loss": 0.32297924160957336, "global_step": 75704, "epoch": 850, "lr": 8.216552578164718e-05}
{"train_loss": 0.24153250455856323, "global_step": 75705, "epoch": 850, "lr": 8.216508193002914e-05}
{"train_loss": 0.19889724254608154, "global_step": 75706, "epoch": 850, "lr": 8.21646380740869e-05}
{"train_loss": 0.16581490635871887, "global_step": 75707, "epoch": 850, "lr": 8.21641942138205e-05}
{"train_loss": 0.14079873263835907, "global_step": 75708, "epoch": 850, "lr": 8.216375034923003e-05}
{"train_loss": 0.14338898658752441, "global_step": 75709, "epoch": 850, "lr": 8.216330648031554e-05}
{"train_loss": 0.21413548290729523, "global_step": 75710, "epoch": 850, "lr": 8.216286260707707e-05}
{"train_loss": 0.1769370287656784, "global_step": 75711, "epoch": 850, "lr": 8.21624187295147e-05}
{"train_loss": 0.18530502915382385, "global_step": 75712, "epoch": 850, "lr": 8.216197484762849e-05}
{"train_loss": 0.24011757969856262, "global_step": 75713, "epoch": 850, "lr": 8.216153096141848e-05}
{"train_loss": 0.20700225234031677, "global_step": 75714, "epoch": 850, "lr": 8.216108707088476e-05}
{"train_loss": 0.15259678661823273, "global_step": 75715, "epoch": 850, "lr": 8.216064317602737e-05}
{"train_loss": 0.17139418423175812, "global_step": 75716, "epoch": 850, "lr": 8.216019927684637e-05}
{"train_loss": 0.2681256830692291, "global_step": 75717, "epoch": 850, "lr": 8.215975537334182e-05}
{"train_loss": 0.23474986851215363, "global_step": 75718, "epoch": 850, "lr": 8.21593114655138e-05}
{"train_loss": 0.1674567610025406, "global_step": 75719, "epoch": 850, "lr": 8.215886755336236e-05}
{"train_loss": 0.20197664201259613, "global_step": 75720, "epoch": 850, "lr": 8.215842363688753e-05}
{"train_loss": 0.3513718843460083, "global_step": 75721, "epoch": 850, "lr": 8.215797971608939e-05}
{"train_loss": 0.2182260900735855, "global_step": 75722, "epoch": 850, "lr": 8.215753579096801e-05}
{"train_loss": 0.27946949005126953, "global_step": 75723, "epoch": 850, "lr": 8.215709186152343e-05}
{"train_loss": 0.22488737106323242, "global_step": 75724, "epoch": 850, "lr": 8.215664792775574e-05}
{"train_loss": 0.1743164211511612, "global_step": 75725, "epoch": 850, "lr": 8.215620398966497e-05}
{"train_loss": 0.18671832978725433, "global_step": 75726, "epoch": 850, "lr": 8.215576004725119e-05}
{"train_loss": 0.26747432351112366, "global_step": 75727, "epoch": 850, "lr": 8.215531610051446e-05}
{"train_loss": 0.16994711756706238, "global_step": 75728, "epoch": 850, "lr": 8.215487214945483e-05}
{"train_loss": 0.16570918262004852, "global_step": 75729, "epoch": 850, "lr": 8.215442819407239e-05}
{"train_loss": 0.2550642192363739, "global_step": 75730, "epoch": 850, "lr": 8.215398423436715e-05}
{"train_loss": 0.18044829368591309, "global_step": 75731, "epoch": 850, "lr": 8.215354027033923e-05}
{"train_loss": 0.21967297792434692, "global_step": 75732, "epoch": 850, "lr": 8.215309630198864e-05}
{"train_loss": 0.1453336477279663, "global_step": 75733, "epoch": 850, "lr": 8.215265232931547e-05}
{"train_loss": 0.2489016056060791, "global_step": 75734, "epoch": 850, "lr": 8.215220835231975e-05}
{"train_loss": 0.22007259726524353, "global_step": 75735, "epoch": 850, "lr": 8.215176437100159e-05}
{"train_loss": 0.19943027198314667, "global_step": 75736, "epoch": 850, "lr": 8.215132038536099e-05}
{"train_loss": 0.16921447217464447, "global_step": 75737, "epoch": 850, "lr": 8.215087639539804e-05}
{"train_loss": 0.21079511722821867, "global_step": 75738, "epoch": 850, "lr": 8.21504324011128e-05, "train/sim_max_reward_0": 0.4878067164842611, "train/sim_max_reward_1": 0.5950332757594355, "train/sim_max_reward_2": 0.49049978208534, "train/sim_max_reward_3": 0.916154184169826, "train/sim_max_reward_4": 0.9152089476152167, "train/sim_max_reward_5": 0.7279768434478125, "test/sim_max_reward_4500000": 0.9974821391956415, "test/sim_max_reward_4500001": 0.9876480920187687, "test/sim_max_reward_4500002": 0.2358746388293339, "test/sim_max_reward_4500003": 0.4985002813889487, "test/sim_max_reward_4500004": 0.567822185059979, "test/sim_max_reward_4500005": 0.9700037842774818, "test/sim_max_reward_4500006": 0.9995016463984479, "test/sim_max_reward_4500007": 0.9460520107450456, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.05072878768834122, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9929920158791536, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.6573068506781995, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.45227119852733505, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 0.9800783319464983, "test/sim_max_reward_4500019": 0.39496762743265906, "test/sim_max_reward_4500020": 1.0, "test/sim_max_reward_4500021": 0.9679891461238165, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6836565064606628, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9911266567196051, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6076329804174866, "test/sim_max_reward_4500029": 0.46772099574166176, "test/sim_max_reward_4500030": 0.2459906034143236, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.8202308148646711, "test/sim_max_reward_4500033": 0.6642927004251022, "test/sim_max_reward_4500034": 0.9723992880793604, "test/sim_max_reward_4500035": 0.6682126369463066, "test/sim_max_reward_4500036": 0.1525203373075606, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.3419965067625457, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9290670754793541, "test/sim_max_reward_4500041": 0.0, "test/sim_max_reward_4500042": 0.715000245442268, "test/sim_max_reward_4500043": 0.6986451142821228, "test/sim_max_reward_4500044": 0.4718276020081674, "test/sim_max_reward_4500045": 0.65934400807563, "test/sim_max_reward_4500046": 0.4749874924714128, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9984788544242356, "test/sim_max_reward_4500049": 0.5470539513622397, "train/mean_score": 0.6887799582603154, "test/mean_score": 0.6161880621374873, "val_loss": 3.8834993839263916, "train_action_mse_error": 33.023338317871094}
{"train_loss": 0.16706261038780212, "global_step": 75739, "epoch": 851, "lr": 8.214998840250533e-05}
{"train_loss": 0.16773022711277008, "global_step": 75740, "epoch": 851, "lr": 8.214954439957569e-05}
{"train_loss": 0.13256947696208954, "global_step": 75741, "epoch": 851, "lr": 8.214910039232392e-05}
{"train_loss": 0.20458155870437622, "global_step": 75742, "epoch": 851, "lr": 8.214865638075011e-05}
{"train_loss": 0.22561833262443542, "global_step": 75743, "epoch": 851, "lr": 8.21482123648543e-05}
{"train_loss": 0.30808165669441223, "global_step": 75744, "epoch": 851, "lr": 8.214776834463654e-05}
{"train_loss": 0.19806382060050964, "global_step": 75745, "epoch": 851, "lr": 8.214732432009691e-05}
{"train_loss": 0.18556883931159973, "global_step": 75746, "epoch": 851, "lr": 8.214688029123549e-05}
{"train_loss": 0.14315521717071533, "global_step": 75747, "epoch": 851, "lr": 8.214643625805229e-05}
{"train_loss": 0.14234937727451324, "global_step": 75748, "epoch": 851, "lr": 8.214599222054739e-05}
{"train_loss": 0.14715130627155304, "global_step": 75749, "epoch": 851, "lr": 8.214554817872086e-05}
{"train_loss": 0.26566407084465027, "global_step": 75750, "epoch": 851, "lr": 8.214510413257276e-05}
{"train_loss": 0.23697660863399506, "global_step": 75751, "epoch": 851, "lr": 8.214466008210313e-05}
{"train_loss": 0.20988130569458008, "global_step": 75752, "epoch": 851, "lr": 8.214421602731206e-05}
{"train_loss": 0.2602740228176117, "global_step": 75753, "epoch": 851, "lr": 8.214377196819957e-05}
{"train_loss": 0.19106945395469666, "global_step": 75754, "epoch": 851, "lr": 8.214332790476575e-05}
{"train_loss": 0.24201074242591858, "global_step": 75755, "epoch": 851, "lr": 8.214288383701066e-05}
{"train_loss": 0.17828528583049774, "global_step": 75756, "epoch": 851, "lr": 8.214243976493435e-05}
{"train_loss": 0.15921436250209808, "global_step": 75757, "epoch": 851, "lr": 8.214199568853687e-05}
{"train_loss": 0.2109934538602829, "global_step": 75758, "epoch": 851, "lr": 8.21415516078183e-05}
{"train_loss": 0.1864156275987625, "global_step": 75759, "epoch": 851, "lr": 8.214110752277868e-05}
{"train_loss": 0.22726936638355255, "global_step": 75760, "epoch": 851, "lr": 8.214066343341809e-05}
{"train_loss": 0.193475142121315, "global_step": 75761, "epoch": 851, "lr": 8.214021933973658e-05}
{"train_loss": 0.2128966897726059, "global_step": 75762, "epoch": 851, "lr": 8.213977524173419e-05}
{"train_loss": 0.1681828796863556, "global_step": 75763, "epoch": 851, "lr": 8.213933113941104e-05}
{"train_loss": 0.15243718028068542, "global_step": 75764, "epoch": 851, "lr": 8.213888703276711e-05}
{"train_loss": 0.24218489229679108, "global_step": 75765, "epoch": 851, "lr": 8.213844292180251e-05}
{"train_loss": 0.13087549805641174, "global_step": 75766, "epoch": 851, "lr": 8.213799880651729e-05}
{"train_loss": 0.21752427518367767, "global_step": 75767, "epoch": 851, "lr": 8.21375546869115e-05}
{"train_loss": 0.16093771159648895, "global_step": 75768, "epoch": 851, "lr": 8.21371105629852e-05}
{"train_loss": 0.16400359570980072, "global_step": 75769, "epoch": 851, "lr": 8.213666643473849e-05}
{"train_loss": 0.2404058277606964, "global_step": 75770, "epoch": 851, "lr": 8.213622230217137e-05}
{"train_loss": 0.18462297320365906, "global_step": 75771, "epoch": 851, "lr": 8.213577816528394e-05}
{"train_loss": 0.14503350853919983, "global_step": 75772, "epoch": 851, "lr": 8.213533402407625e-05}
{"train_loss": 0.2399245649576187, "global_step": 75773, "epoch": 851, "lr": 8.213488987854834e-05}
{"train_loss": 0.2285589873790741, "global_step": 75774, "epoch": 851, "lr": 8.21344457287003e-05}
{"train_loss": 0.3015476167201996, "global_step": 75775, "epoch": 851, "lr": 8.213400157453215e-05}
{"train_loss": 0.12857626378536224, "global_step": 75776, "epoch": 851, "lr": 8.213355741604401e-05}
{"train_loss": 0.1263817697763443, "global_step": 75777, "epoch": 851, "lr": 8.21331132532359e-05}
{"train_loss": 0.1448710560798645, "global_step": 75778, "epoch": 851, "lr": 8.213266908610785e-05}
{"train_loss": 0.17417262494564056, "global_step": 75779, "epoch": 851, "lr": 8.213222491466e-05}
{"train_loss": 0.19405516982078552, "global_step": 75780, "epoch": 851, "lr": 8.213178073889232e-05}
{"train_loss": 0.14775241911411285, "global_step": 75781, "epoch": 851, "lr": 8.213133655880494e-05}
{"train_loss": 0.15695546567440033, "global_step": 75782, "epoch": 851, "lr": 8.213089237439789e-05}
{"train_loss": 0.23936408758163452, "global_step": 75783, "epoch": 851, "lr": 8.213044818567124e-05}
{"train_loss": 0.31787487864494324, "global_step": 75784, "epoch": 851, "lr": 8.213000399262503e-05}
{"train_loss": 0.151122584939003, "global_step": 75785, "epoch": 851, "lr": 8.212955979525934e-05}
{"train_loss": 0.18204645812511444, "global_step": 75786, "epoch": 851, "lr": 8.212911559357421e-05}
{"train_loss": 0.178190216422081, "global_step": 75787, "epoch": 851, "lr": 8.212867138756972e-05}
{"train_loss": 0.17124198377132416, "global_step": 75788, "epoch": 851, "lr": 8.212822717724592e-05}
{"train_loss": 0.23403295874595642, "global_step": 75789, "epoch": 851, "lr": 8.212778296260288e-05}
{"train_loss": 0.22072537243366241, "global_step": 75790, "epoch": 851, "lr": 8.212733874364065e-05}
{"train_loss": 0.19069433212280273, "global_step": 75791, "epoch": 851, "lr": 8.212689452035928e-05}
{"train_loss": 0.2431420236825943, "global_step": 75792, "epoch": 851, "lr": 8.212645029275886e-05}
{"train_loss": 0.2197970300912857, "global_step": 75793, "epoch": 851, "lr": 8.21260060608394e-05}
{"train_loss": 0.23651954531669617, "global_step": 75794, "epoch": 851, "lr": 8.2125561824601e-05}
{"train_loss": 0.19272896647453308, "global_step": 75795, "epoch": 851, "lr": 8.212511758404373e-05}
{"train_loss": 0.2269129902124405, "global_step": 75796, "epoch": 851, "lr": 8.212467333916763e-05}
{"train_loss": 0.1974453330039978, "global_step": 75797, "epoch": 851, "lr": 8.212422908997274e-05}
{"train_loss": 0.21737302839756012, "global_step": 75798, "epoch": 851, "lr": 8.212378483645914e-05}
{"train_loss": 0.17874585092067719, "global_step": 75799, "epoch": 851, "lr": 8.212334057862689e-05}
{"train_loss": 0.29283443093299866, "global_step": 75800, "epoch": 851, "lr": 8.212289631647605e-05}
{"train_loss": 0.1646045744419098, "global_step": 75801, "epoch": 851, "lr": 8.212245205000668e-05}
{"train_loss": 0.24368172883987427, "global_step": 75802, "epoch": 851, "lr": 8.212200777921884e-05}
{"train_loss": 0.27956172823905945, "global_step": 75803, "epoch": 851, "lr": 8.212156350411259e-05}
{"train_loss": 0.25044184923171997, "global_step": 75804, "epoch": 851, "lr": 8.212111922468797e-05}
{"train_loss": 0.22490264475345612, "global_step": 75805, "epoch": 851, "lr": 8.212067494094507e-05}
{"train_loss": 0.17256727814674377, "global_step": 75806, "epoch": 851, "lr": 8.212023065288394e-05}
{"train_loss": 0.2774563133716583, "global_step": 75807, "epoch": 851, "lr": 8.211978636050463e-05}
{"train_loss": 0.3508256673812866, "global_step": 75808, "epoch": 851, "lr": 8.211934206380722e-05}
{"train_loss": 0.2872679829597473, "global_step": 75809, "epoch": 851, "lr": 8.211889776279174e-05}
{"train_loss": 0.13906748592853546, "global_step": 75810, "epoch": 851, "lr": 8.211845345745827e-05}
{"train_loss": 0.17701689898967743, "global_step": 75811, "epoch": 851, "lr": 8.211800914780688e-05}
{"train_loss": 0.23198194801807404, "global_step": 75812, "epoch": 851, "lr": 8.211756483383759e-05}
{"train_loss": 0.25457122921943665, "global_step": 75813, "epoch": 851, "lr": 8.211712051555049e-05}
{"train_loss": 0.18756629526615143, "global_step": 75814, "epoch": 851, "lr": 8.211667619294566e-05}
{"train_loss": 0.19467224180698395, "global_step": 75815, "epoch": 851, "lr": 8.21162318660231e-05}
{"train_loss": 0.20230762660503387, "global_step": 75816, "epoch": 851, "lr": 8.211578753478293e-05}
{"train_loss": 0.2409513294696808, "global_step": 75817, "epoch": 851, "lr": 8.211534319922517e-05}
{"train_loss": 0.17521129548549652, "global_step": 75818, "epoch": 851, "lr": 8.211489885934989e-05}
{"train_loss": 0.15028417110443115, "global_step": 75819, "epoch": 851, "lr": 8.211445451515715e-05}
{"train_loss": 0.16730627417564392, "global_step": 75820, "epoch": 851, "lr": 8.211401016664703e-05}
{"train_loss": 0.21577486395835876, "global_step": 75821, "epoch": 851, "lr": 8.211356581381956e-05}
{"train_loss": 0.18277660012245178, "global_step": 75822, "epoch": 851, "lr": 8.211312145667483e-05}
{"train_loss": 0.19682671129703522, "global_step": 75823, "epoch": 851, "lr": 8.211267709521287e-05}
{"train_loss": 0.18084128201007843, "global_step": 75824, "epoch": 851, "lr": 8.211223272943375e-05}
{"train_loss": 0.18671731650829315, "global_step": 75825, "epoch": 851, "lr": 8.211178835933753e-05}
{"train_loss": 0.17690126597881317, "global_step": 75826, "epoch": 851, "lr": 8.211134398492429e-05}
{"train_loss": 0.20264841783582493, "global_step": 75827, "epoch": 851, "lr": 8.211089960619405e-05, "val_loss": 3.800236225128174}
{"train_loss": 0.19485123455524445, "global_step": 75828, "epoch": 852, "lr": 8.211045522314688e-05}
{"train_loss": 0.2581018805503845, "global_step": 75829, "epoch": 852, "lr": 8.211001083578288e-05}
{"train_loss": 0.1754630208015442, "global_step": 75830, "epoch": 852, "lr": 8.210956644410206e-05}
{"train_loss": 0.18919780850410461, "global_step": 75831, "epoch": 852, "lr": 8.210912204810452e-05}
{"train_loss": 0.19577892124652863, "global_step": 75832, "epoch": 852, "lr": 8.210867764779029e-05}
{"train_loss": 0.19950930774211884, "global_step": 75833, "epoch": 852, "lr": 8.210823324315944e-05}
{"train_loss": 0.18627066910266876, "global_step": 75834, "epoch": 852, "lr": 8.210778883421201e-05}
{"train_loss": 0.22464177012443542, "global_step": 75835, "epoch": 852, "lr": 8.210734442094812e-05}
{"train_loss": 0.160567969083786, "global_step": 75836, "epoch": 852, "lr": 8.210690000336776e-05}
{"train_loss": 0.14701765775680542, "global_step": 75837, "epoch": 852, "lr": 8.210645558147102e-05}
{"train_loss": 0.27804476022720337, "global_step": 75838, "epoch": 852, "lr": 8.210601115525795e-05}
{"train_loss": 0.2595132887363434, "global_step": 75839, "epoch": 852, "lr": 8.210556672472863e-05}
{"train_loss": 0.263237863779068, "global_step": 75840, "epoch": 852, "lr": 8.210512228988312e-05}
{"train_loss": 0.16354094445705414, "global_step": 75841, "epoch": 852, "lr": 8.210467785072145e-05}
{"train_loss": 0.2236141413450241, "global_step": 75842, "epoch": 852, "lr": 8.210423340724371e-05}
{"train_loss": 0.23848271369934082, "global_step": 75843, "epoch": 852, "lr": 8.210378895944995e-05}
{"train_loss": 0.24298670887947083, "global_step": 75844, "epoch": 852, "lr": 8.210334450734021e-05}
{"train_loss": 0.22101131081581116, "global_step": 75845, "epoch": 852, "lr": 8.210290005091457e-05}
{"train_loss": 0.1538977175951004, "global_step": 75846, "epoch": 852, "lr": 8.210245559017308e-05}
{"train_loss": 0.28430670499801636, "global_step": 75847, "epoch": 852, "lr": 8.210201112511582e-05}
{"train_loss": 0.19653892517089844, "global_step": 75848, "epoch": 852, "lr": 8.210156665574284e-05}
{"train_loss": 0.14981456100940704, "global_step": 75849, "epoch": 852, "lr": 8.21011221820542e-05}
{"train_loss": 0.19390979409217834, "global_step": 75850, "epoch": 852, "lr": 8.210067770404993e-05}
{"train_loss": 0.15965832769870758, "global_step": 75851, "epoch": 852, "lr": 8.210023322173013e-05}
{"train_loss": 0.13717536628246307, "global_step": 75852, "epoch": 852, "lr": 8.209978873509484e-05}
{"train_loss": 0.2431030422449112, "global_step": 75853, "epoch": 852, "lr": 8.209934424414414e-05}
{"train_loss": 0.25308704376220703, "global_step": 75854, "epoch": 852, "lr": 8.209889974887805e-05}
{"train_loss": 0.19944028556346893, "global_step": 75855, "epoch": 852, "lr": 8.209845524929667e-05}
{"train_loss": 0.21226848661899567, "global_step": 75856, "epoch": 852, "lr": 8.209801074540004e-05}
{"train_loss": 0.1753464639186859, "global_step": 75857, "epoch": 852, "lr": 8.209756623718822e-05}
{"train_loss": 0.18671338260173798, "global_step": 75858, "epoch": 852, "lr": 8.209712172466129e-05}
{"train_loss": 0.2158694565296173, "global_step": 75859, "epoch": 852, "lr": 8.209667720781926e-05}
{"train_loss": 0.20307758450508118, "global_step": 75860, "epoch": 852, "lr": 8.209623268666225e-05}
{"train_loss": 0.23113533854484558, "global_step": 75861, "epoch": 852, "lr": 8.20957881611903e-05}
{"train_loss": 0.1805921196937561, "global_step": 75862, "epoch": 852, "lr": 8.209534363140345e-05}
{"train_loss": 0.2945709824562073, "global_step": 75863, "epoch": 852, "lr": 8.209489909730177e-05}
{"train_loss": 0.24457195401191711, "global_step": 75864, "epoch": 852, "lr": 8.209445455888531e-05}
{"train_loss": 0.3089969754219055, "global_step": 75865, "epoch": 852, "lr": 8.209401001615417e-05}
{"train_loss": 0.14074841141700745, "global_step": 75866, "epoch": 852, "lr": 8.209356546910836e-05}
{"train_loss": 0.19506630301475525, "global_step": 75867, "epoch": 852, "lr": 8.209312091774797e-05}
{"train_loss": 0.2139338254928589, "global_step": 75868, "epoch": 852, "lr": 8.209267636207305e-05}
{"train_loss": 0.2696804106235504, "global_step": 75869, "epoch": 852, "lr": 8.209223180208366e-05}
{"train_loss": 0.1706097424030304, "global_step": 75870, "epoch": 852, "lr": 8.209178723777986e-05}
{"train_loss": 0.3181339204311371, "global_step": 75871, "epoch": 852, "lr": 8.20913426691617e-05}
{"train_loss": 0.29295048117637634, "global_step": 75872, "epoch": 852, "lr": 8.209089809622927e-05}
{"train_loss": 0.2009533792734146, "global_step": 75873, "epoch": 852, "lr": 8.20904535189826e-05}
{"train_loss": 0.22234371304512024, "global_step": 75874, "epoch": 852, "lr": 8.209000893742176e-05}
{"train_loss": 0.2290162444114685, "global_step": 75875, "epoch": 852, "lr": 8.20895643515468e-05}
{"train_loss": 0.16977722942829132, "global_step": 75876, "epoch": 852, "lr": 8.20891197613578e-05}
{"train_loss": 0.2795087397098541, "global_step": 75877, "epoch": 852, "lr": 8.208867516685482e-05}
{"train_loss": 0.1816757321357727, "global_step": 75878, "epoch": 852, "lr": 8.208823056803789e-05}
{"train_loss": 0.15515899658203125, "global_step": 75879, "epoch": 852, "lr": 8.208778596490709e-05}
{"train_loss": 0.2310742437839508, "global_step": 75880, "epoch": 852, "lr": 8.20873413574625e-05}
{"train_loss": 0.22696660459041595, "global_step": 75881, "epoch": 852, "lr": 8.208689674570413e-05}
{"train_loss": 0.2632301151752472, "global_step": 75882, "epoch": 852, "lr": 8.208645212963209e-05}
{"train_loss": 0.18302419781684875, "global_step": 75883, "epoch": 852, "lr": 8.20860075092464e-05}
{"train_loss": 0.18266825377941132, "global_step": 75884, "epoch": 852, "lr": 8.208556288454714e-05}
{"train_loss": 0.20873166620731354, "global_step": 75885, "epoch": 852, "lr": 8.208511825553438e-05}
{"train_loss": 0.17165572941303253, "global_step": 75886, "epoch": 852, "lr": 8.208467362220815e-05}
{"train_loss": 0.17422765493392944, "global_step": 75887, "epoch": 852, "lr": 8.208422898456854e-05}
{"train_loss": 0.24772220849990845, "global_step": 75888, "epoch": 852, "lr": 8.208378434261559e-05}
{"train_loss": 0.22405636310577393, "global_step": 75889, "epoch": 852, "lr": 8.208333969634936e-05}
{"train_loss": 0.14958886802196503, "global_step": 75890, "epoch": 852, "lr": 8.208289504576994e-05}
{"train_loss": 0.12924998998641968, "global_step": 75891, "epoch": 852, "lr": 8.208245039087734e-05}
{"train_loss": 0.23788811266422272, "global_step": 75892, "epoch": 852, "lr": 8.208200573167167e-05}
{"train_loss": 0.20163296163082123, "global_step": 75893, "epoch": 852, "lr": 8.208156106815293e-05}
{"train_loss": 0.3003735840320587, "global_step": 75894, "epoch": 852, "lr": 8.208111640032123e-05}
{"train_loss": 0.1898307353258133, "global_step": 75895, "epoch": 852, "lr": 8.208067172817662e-05}
{"train_loss": 0.28635725378990173, "global_step": 75896, "epoch": 852, "lr": 8.208022705171916e-05}
{"train_loss": 0.18460504710674286, "global_step": 75897, "epoch": 852, "lr": 8.207978237094891e-05}
{"train_loss": 0.2546241283416748, "global_step": 75898, "epoch": 852, "lr": 8.207933768586591e-05}
{"train_loss": 0.11772850155830383, "global_step": 75899, "epoch": 852, "lr": 8.207889299647024e-05}
{"train_loss": 0.20079545676708221, "global_step": 75900, "epoch": 852, "lr": 8.207844830276195e-05}
{"train_loss": 0.18864183127880096, "global_step": 75901, "epoch": 852, "lr": 8.207800360474111e-05}
{"train_loss": 0.16116398572921753, "global_step": 75902, "epoch": 852, "lr": 8.207755890240776e-05}
{"train_loss": 0.30623117089271545, "global_step": 75903, "epoch": 852, "lr": 8.207711419576198e-05}
{"train_loss": 0.232884481549263, "global_step": 75904, "epoch": 852, "lr": 8.207666948480384e-05}
{"train_loss": 0.17246301472187042, "global_step": 75905, "epoch": 852, "lr": 8.207622476953337e-05}
{"train_loss": 0.16799363493919373, "global_step": 75906, "epoch": 852, "lr": 8.207578004995062e-05}
{"train_loss": 0.20144613087177277, "global_step": 75907, "epoch": 852, "lr": 8.20753353260557e-05}
{"train_loss": 0.16797198355197906, "global_step": 75908, "epoch": 852, "lr": 8.207489059784863e-05}
{"train_loss": 0.16387584805488586, "global_step": 75909, "epoch": 852, "lr": 8.207444586532948e-05}
{"train_loss": 0.17103509604930878, "global_step": 75910, "epoch": 852, "lr": 8.207400112849832e-05}
{"train_loss": 0.21435454487800598, "global_step": 75911, "epoch": 852, "lr": 8.207355638735521e-05}
{"train_loss": 0.19777211546897888, "global_step": 75912, "epoch": 852, "lr": 8.207311164190019e-05}
{"train_loss": 0.22500015795230865, "global_step": 75913, "epoch": 852, "lr": 8.207266689213334e-05}
{"train_loss": 0.2525603473186493, "global_step": 75914, "epoch": 852, "lr": 8.20722221380547e-05}
{"train_loss": 0.25835728645324707, "global_step": 75915, "epoch": 852, "lr": 8.207177737966434e-05}
{"train_loss": 0.21044571329368633, "global_step": 75916, "epoch": 852, "lr": 8.207133261696231e-05, "val_loss": 3.7699344158172607}
{"train_loss": 0.11130330711603165, "global_step": 75917, "epoch": 853, "lr": 8.207088784994869e-05}
{"train_loss": 0.12313593924045563, "global_step": 75918, "epoch": 853, "lr": 8.207044307862355e-05}
{"train_loss": 0.1439685970544815, "global_step": 75919, "epoch": 853, "lr": 8.206999830298691e-05}
{"train_loss": 0.12032986432313919, "global_step": 75920, "epoch": 853, "lr": 8.206955352303886e-05}
{"train_loss": 0.1340879201889038, "global_step": 75921, "epoch": 853, "lr": 8.206910873877944e-05}
{"train_loss": 0.2179618924856186, "global_step": 75922, "epoch": 853, "lr": 8.206866395020872e-05}
{"train_loss": 0.2203761637210846, "global_step": 75923, "epoch": 853, "lr": 8.206821915732677e-05}
{"train_loss": 0.1895488053560257, "global_step": 75924, "epoch": 853, "lr": 8.206777436013362e-05}
{"train_loss": 0.21102631092071533, "global_step": 75925, "epoch": 853, "lr": 8.206732955862937e-05}
{"train_loss": 0.31474974751472473, "global_step": 75926, "epoch": 853, "lr": 8.206688475281405e-05}
{"train_loss": 0.21184252202510834, "global_step": 75927, "epoch": 853, "lr": 8.206643994268772e-05}
{"train_loss": 0.17129650712013245, "global_step": 75928, "epoch": 853, "lr": 8.206599512825045e-05}
{"train_loss": 0.20224010944366455, "global_step": 75929, "epoch": 853, "lr": 8.206555030950231e-05}
{"train_loss": 0.23591135442256927, "global_step": 75930, "epoch": 853, "lr": 8.206510548644333e-05}
{"train_loss": 0.23525168001651764, "global_step": 75931, "epoch": 853, "lr": 8.206466065907358e-05}
{"train_loss": 0.327576607465744, "global_step": 75932, "epoch": 853, "lr": 8.206421582739315e-05}
{"train_loss": 0.21653781831264496, "global_step": 75933, "epoch": 853, "lr": 8.206377099140207e-05}
{"train_loss": 0.30240577459335327, "global_step": 75934, "epoch": 853, "lr": 8.206332615110042e-05}
{"train_loss": 0.17124059796333313, "global_step": 75935, "epoch": 853, "lr": 8.206288130648822e-05}
{"train_loss": 0.33221864700317383, "global_step": 75936, "epoch": 853, "lr": 8.206243645756557e-05}
{"train_loss": 0.1926632821559906, "global_step": 75937, "epoch": 853, "lr": 8.206199160433253e-05}
{"train_loss": 0.148025244474411, "global_step": 75938, "epoch": 853, "lr": 8.206154674678912e-05}
{"train_loss": 0.20278729498386383, "global_step": 75939, "epoch": 853, "lr": 8.206110188493544e-05}
{"train_loss": 0.20571012794971466, "global_step": 75940, "epoch": 853, "lr": 8.206065701877154e-05}
{"train_loss": 0.18873754143714905, "global_step": 75941, "epoch": 853, "lr": 8.206021214829746e-05}
{"train_loss": 0.2984934449195862, "global_step": 75942, "epoch": 853, "lr": 8.20597672735133e-05}
{"train_loss": 0.21831385791301727, "global_step": 75943, "epoch": 853, "lr": 8.205932239441907e-05}
{"train_loss": 0.25635942816734314, "global_step": 75944, "epoch": 853, "lr": 8.205887751101487e-05}
{"train_loss": 0.22811521589756012, "global_step": 75945, "epoch": 853, "lr": 8.205843262330073e-05}
{"train_loss": 0.26139330863952637, "global_step": 75946, "epoch": 853, "lr": 8.205798773127674e-05}
{"train_loss": 0.21640877425670624, "global_step": 75947, "epoch": 853, "lr": 8.205754283494293e-05}
{"train_loss": 0.24255169928073883, "global_step": 75948, "epoch": 853, "lr": 8.205709793429939e-05}
{"train_loss": 0.2797873020172119, "global_step": 75949, "epoch": 853, "lr": 8.205665302934615e-05}
{"train_loss": 0.27433884143829346, "global_step": 75950, "epoch": 853, "lr": 8.205620812008329e-05}
{"train_loss": 0.25376325845718384, "global_step": 75951, "epoch": 853, "lr": 8.205576320651086e-05}
{"train_loss": 0.2369355857372284, "global_step": 75952, "epoch": 853, "lr": 8.205531828862892e-05}
{"train_loss": 0.21057917177677155, "global_step": 75953, "epoch": 853, "lr": 8.205487336643754e-05}
{"train_loss": 0.18855683505535126, "global_step": 75954, "epoch": 853, "lr": 8.205442843993678e-05}
{"train_loss": 0.2621595859527588, "global_step": 75955, "epoch": 853, "lr": 8.205398350912667e-05}
{"train_loss": 0.12935739755630493, "global_step": 75956, "epoch": 853, "lr": 8.205353857400733e-05}
{"train_loss": 0.2599025070667267, "global_step": 75957, "epoch": 853, "lr": 8.205309363457875e-05}
{"train_loss": 0.19534580409526825, "global_step": 75958, "epoch": 853, "lr": 8.205264869084104e-05}
{"train_loss": 0.3019966185092926, "global_step": 75959, "epoch": 853, "lr": 8.205220374279424e-05}
{"train_loss": 0.21085228025913239, "global_step": 75960, "epoch": 853, "lr": 8.205175879043839e-05}
{"train_loss": 0.23325474560260773, "global_step": 75961, "epoch": 853, "lr": 8.20513138337736e-05}
{"train_loss": 0.20489342510700226, "global_step": 75962, "epoch": 853, "lr": 8.20508688727999e-05}
{"train_loss": 0.11837948113679886, "global_step": 75963, "epoch": 853, "lr": 8.205042390751733e-05}
{"train_loss": 0.14420875906944275, "global_step": 75964, "epoch": 853, "lr": 8.204997893792597e-05}
{"train_loss": 0.22981785237789154, "global_step": 75965, "epoch": 853, "lr": 8.204953396402591e-05}
{"train_loss": 0.33737656474113464, "global_step": 75966, "epoch": 853, "lr": 8.204908898581714e-05}
{"train_loss": 0.23044393956661224, "global_step": 75967, "epoch": 853, "lr": 8.20486440032998e-05}
{"train_loss": 0.08418425917625427, "global_step": 75968, "epoch": 853, "lr": 8.20481990164739e-05}
{"train_loss": 0.23641252517700195, "global_step": 75969, "epoch": 853, "lr": 8.204775402533949e-05}
{"train_loss": 0.15875035524368286, "global_step": 75970, "epoch": 853, "lr": 8.204730902989667e-05}
{"train_loss": 0.1637667715549469, "global_step": 75971, "epoch": 853, "lr": 8.204686403014548e-05}
{"train_loss": 0.17010338604450226, "global_step": 75972, "epoch": 853, "lr": 8.204641902608596e-05}
{"train_loss": 0.1267252117395401, "global_step": 75973, "epoch": 853, "lr": 8.204597401771821e-05}
{"train_loss": 0.34487420320510864, "global_step": 75974, "epoch": 853, "lr": 8.204552900504226e-05}
{"train_loss": 0.27370837330818176, "global_step": 75975, "epoch": 853, "lr": 8.204508398805819e-05}
{"train_loss": 0.2807677090167999, "global_step": 75976, "epoch": 853, "lr": 8.204463896676604e-05}
{"train_loss": 0.22214782238006592, "global_step": 75977, "epoch": 853, "lr": 8.204419394116588e-05}
{"train_loss": 0.10158635675907135, "global_step": 75978, "epoch": 853, "lr": 8.204374891125777e-05}
{"train_loss": 0.18768970668315887, "global_step": 75979, "epoch": 853, "lr": 8.204330387704178e-05}
{"train_loss": 0.29677584767341614, "global_step": 75980, "epoch": 853, "lr": 8.204285883851793e-05}
{"train_loss": 0.2282751351594925, "global_step": 75981, "epoch": 853, "lr": 8.204241379568634e-05}
{"train_loss": 0.21727359294891357, "global_step": 75982, "epoch": 853, "lr": 8.204196874854702e-05}
{"train_loss": 0.23248723149299622, "global_step": 75983, "epoch": 853, "lr": 8.204152369710005e-05}
{"train_loss": 0.16802062094211578, "global_step": 75984, "epoch": 853, "lr": 8.204107864134547e-05}
{"train_loss": 0.2777114510536194, "global_step": 75985, "epoch": 853, "lr": 8.204063358128338e-05}
{"train_loss": 0.2510969638824463, "global_step": 75986, "epoch": 853, "lr": 8.204018851691382e-05}
{"train_loss": 0.25443601608276367, "global_step": 75987, "epoch": 853, "lr": 8.203974344823684e-05}
{"train_loss": 0.3333081007003784, "global_step": 75988, "epoch": 853, "lr": 8.20392983752525e-05}
{"train_loss": 0.2601873576641083, "global_step": 75989, "epoch": 853, "lr": 8.203885329796088e-05}
{"train_loss": 0.17929597198963165, "global_step": 75990, "epoch": 853, "lr": 8.203840821636202e-05}
{"train_loss": 0.22996746003627777, "global_step": 75991, "epoch": 853, "lr": 8.203796313045598e-05}
{"train_loss": 0.18585212528705597, "global_step": 75992, "epoch": 853, "lr": 8.203751804024284e-05}
{"train_loss": 0.2155168056488037, "global_step": 75993, "epoch": 853, "lr": 8.203707294572262e-05}
{"train_loss": 0.18727046251296997, "global_step": 75994, "epoch": 853, "lr": 8.203662784689544e-05}
{"train_loss": 0.2244575172662735, "global_step": 75995, "epoch": 853, "lr": 8.20361827437613e-05}
{"train_loss": 0.18769362568855286, "global_step": 75996, "epoch": 853, "lr": 8.203573763632031e-05}
{"train_loss": 0.20454008877277374, "global_step": 75997, "epoch": 853, "lr": 8.20352925245725e-05}
{"train_loss": 0.16357757151126862, "global_step": 75998, "epoch": 853, "lr": 8.203484740851792e-05}
{"train_loss": 0.25060126185417175, "global_step": 75999, "epoch": 853, "lr": 8.203440228815665e-05}
{"train_loss": 0.15394069254398346, "global_step": 76000, "epoch": 853, "lr": 8.203395716348876e-05}
{"train_loss": 0.10698287934064865, "global_step": 76001, "epoch": 853, "lr": 8.203351203451428e-05}
{"train_loss": 0.18864160776138306, "global_step": 76002, "epoch": 853, "lr": 8.203306690123327e-05}
{"train_loss": 0.2810938358306885, "global_step": 76003, "epoch": 853, "lr": 8.203262176364584e-05}
{"train_loss": 0.21553216874599457, "global_step": 76004, "epoch": 853, "lr": 8.203217662175199e-05}
{"train_loss": 0.21594224568833126, "global_step": 76005, "epoch": 853, "lr": 8.203173147555182e-05, "val_loss": 3.7961556911468506}
{"train_loss": 0.22457817196846008, "global_step": 76006, "epoch": 854, "lr": 8.203128632504535e-05}
{"train_loss": 0.1760849803686142, "global_step": 76007, "epoch": 854, "lr": 8.203084117023268e-05}
{"train_loss": 0.18134711682796478, "global_step": 76008, "epoch": 854, "lr": 8.203039601111386e-05}
{"train_loss": 0.21586011350154877, "global_step": 76009, "epoch": 854, "lr": 8.202995084768894e-05}
{"train_loss": 0.17992711067199707, "global_step": 76010, "epoch": 854, "lr": 8.202950567995797e-05}
{"train_loss": 0.2323080152273178, "global_step": 76011, "epoch": 854, "lr": 8.202906050792104e-05}
{"train_loss": 0.16070416569709778, "global_step": 76012, "epoch": 854, "lr": 8.202861533157818e-05}
{"train_loss": 0.20734290778636932, "global_step": 76013, "epoch": 854, "lr": 8.202817015092947e-05}
{"train_loss": 0.17275510728359222, "global_step": 76014, "epoch": 854, "lr": 8.202772496597497e-05}
{"train_loss": 0.18931715190410614, "global_step": 76015, "epoch": 854, "lr": 8.202727977671471e-05}
{"train_loss": 0.19413498044013977, "global_step": 76016, "epoch": 854, "lr": 8.202683458314881e-05}
{"train_loss": 0.2593633234500885, "global_step": 76017, "epoch": 854, "lr": 8.202638938527727e-05}
{"train_loss": 0.23102843761444092, "global_step": 76018, "epoch": 854, "lr": 8.202594418310017e-05}
{"train_loss": 0.276820570230484, "global_step": 76019, "epoch": 854, "lr": 8.202549897661758e-05}
{"train_loss": 0.1596485823392868, "global_step": 76020, "epoch": 854, "lr": 8.202505376582953e-05}
{"train_loss": 0.1563173532485962, "global_step": 76021, "epoch": 854, "lr": 8.202460855073613e-05}
{"train_loss": 0.28837960958480835, "global_step": 76022, "epoch": 854, "lr": 8.202416333133739e-05}
{"train_loss": 0.12423276901245117, "global_step": 76023, "epoch": 854, "lr": 8.202371810763341e-05}
{"train_loss": 0.17163297533988953, "global_step": 76024, "epoch": 854, "lr": 8.202327287962421e-05}
{"train_loss": 0.13915309309959412, "global_step": 76025, "epoch": 854, "lr": 8.20228276473099e-05}
{"train_loss": 0.2337457686662674, "global_step": 76026, "epoch": 854, "lr": 8.202238241069048e-05}
{"train_loss": 0.16907621920108795, "global_step": 76027, "epoch": 854, "lr": 8.202193716976607e-05}
{"train_loss": 0.2095128893852234, "global_step": 76028, "epoch": 854, "lr": 8.202149192453669e-05}
{"train_loss": 0.2513182759284973, "global_step": 76029, "epoch": 854, "lr": 8.20210466750024e-05}
{"train_loss": 0.22641132771968842, "global_step": 76030, "epoch": 854, "lr": 8.202060142116328e-05}
{"train_loss": 0.18673057854175568, "global_step": 76031, "epoch": 854, "lr": 8.202015616301938e-05}
{"train_loss": 0.18806758522987366, "global_step": 76032, "epoch": 854, "lr": 8.201971090057077e-05}
{"train_loss": 0.1344982385635376, "global_step": 76033, "epoch": 854, "lr": 8.201926563381748e-05}
{"train_loss": 0.194830521941185, "global_step": 76034, "epoch": 854, "lr": 8.20188203627596e-05}
{"train_loss": 0.22044771909713745, "global_step": 76035, "epoch": 854, "lr": 8.201837508739718e-05}
{"train_loss": 0.18791982531547546, "global_step": 76036, "epoch": 854, "lr": 8.201792980773029e-05}
{"train_loss": 0.16734778881072998, "global_step": 76037, "epoch": 854, "lr": 8.201748452375896e-05}
{"train_loss": 0.17404799163341522, "global_step": 76038, "epoch": 854, "lr": 8.201703923548327e-05}
{"train_loss": 0.23241306841373444, "global_step": 76039, "epoch": 854, "lr": 8.201659394290329e-05}
{"train_loss": 0.23846405744552612, "global_step": 76040, "epoch": 854, "lr": 8.201614864601906e-05}
{"train_loss": 0.2016540914773941, "global_step": 76041, "epoch": 854, "lr": 8.201570334483067e-05}
{"train_loss": 0.2666190266609192, "global_step": 76042, "epoch": 854, "lr": 8.201525803933814e-05}
{"train_loss": 0.20737744867801666, "global_step": 76043, "epoch": 854, "lr": 8.201481272954155e-05}
{"train_loss": 0.2016810029745102, "global_step": 76044, "epoch": 854, "lr": 8.201436741544096e-05}
{"train_loss": 0.1819274127483368, "global_step": 76045, "epoch": 854, "lr": 8.201392209703644e-05}
{"train_loss": 0.2137199491262436, "global_step": 76046, "epoch": 854, "lr": 8.201347677432803e-05}
{"train_loss": 0.24099260568618774, "global_step": 76047, "epoch": 854, "lr": 8.20130314473158e-05}
{"train_loss": 0.18591183423995972, "global_step": 76048, "epoch": 854, "lr": 8.20125861159998e-05}
{"train_loss": 0.34297674894332886, "global_step": 76049, "epoch": 854, "lr": 8.20121407803801e-05}
{"train_loss": 0.1878681182861328, "global_step": 76050, "epoch": 854, "lr": 8.201169544045677e-05}
{"train_loss": 0.19425325095653534, "global_step": 76051, "epoch": 854, "lr": 8.201125009622984e-05}
{"train_loss": 0.17988358438014984, "global_step": 76052, "epoch": 854, "lr": 8.20108047476994e-05}
{"train_loss": 0.24695825576782227, "global_step": 76053, "epoch": 854, "lr": 8.20103593948655e-05}
{"train_loss": 0.18403016030788422, "global_step": 76054, "epoch": 854, "lr": 8.200991403772818e-05}
{"train_loss": 0.25409266352653503, "global_step": 76055, "epoch": 854, "lr": 8.200946867628754e-05}
{"train_loss": 0.19523797929286957, "global_step": 76056, "epoch": 854, "lr": 8.200902331054359e-05}
{"train_loss": 0.21723894774913788, "global_step": 76057, "epoch": 854, "lr": 8.200857794049644e-05}
{"train_loss": 0.12325804680585861, "global_step": 76058, "epoch": 854, "lr": 8.200813256614611e-05}
{"train_loss": 0.2427242249250412, "global_step": 76059, "epoch": 854, "lr": 8.200768718749269e-05}
{"train_loss": 0.15429545938968658, "global_step": 76060, "epoch": 854, "lr": 8.200724180453622e-05}
{"train_loss": 0.15529316663742065, "global_step": 76061, "epoch": 854, "lr": 8.200679641727677e-05}
{"train_loss": 0.15866844356060028, "global_step": 76062, "epoch": 854, "lr": 8.20063510257144e-05}
{"train_loss": 0.24372383952140808, "global_step": 76063, "epoch": 854, "lr": 8.200590562984916e-05}
{"train_loss": 0.18216383457183838, "global_step": 76064, "epoch": 854, "lr": 8.200546022968112e-05}
{"train_loss": 0.21364238858222961, "global_step": 76065, "epoch": 854, "lr": 8.200501482521033e-05}
{"train_loss": 0.23493178188800812, "global_step": 76066, "epoch": 854, "lr": 8.200456941643686e-05}
{"train_loss": 0.2080477476119995, "global_step": 76067, "epoch": 854, "lr": 8.200412400336077e-05}
{"train_loss": 0.18319912254810333, "global_step": 76068, "epoch": 854, "lr": 8.20036785859821e-05}
{"train_loss": 0.17957933247089386, "global_step": 76069, "epoch": 854, "lr": 8.200323316430094e-05}
{"train_loss": 0.23161883652210236, "global_step": 76070, "epoch": 854, "lr": 8.200278773831733e-05}
{"train_loss": 0.15426810085773468, "global_step": 76071, "epoch": 854, "lr": 8.200234230803135e-05}
{"train_loss": 0.23643416166305542, "global_step": 76072, "epoch": 854, "lr": 8.200189687344302e-05}
{"train_loss": 0.2603008449077606, "global_step": 76073, "epoch": 854, "lr": 8.200145143455245e-05}
{"train_loss": 0.18841546773910522, "global_step": 76074, "epoch": 854, "lr": 8.200100599135966e-05}
{"train_loss": 0.19557346403598785, "global_step": 76075, "epoch": 854, "lr": 8.200056054386472e-05}
{"train_loss": 0.21323078870773315, "global_step": 76076, "epoch": 854, "lr": 8.20001150920677e-05}
{"train_loss": 0.23868328332901, "global_step": 76077, "epoch": 854, "lr": 8.199966963596866e-05}
{"train_loss": 0.23667992651462555, "global_step": 76078, "epoch": 854, "lr": 8.199922417556765e-05}
{"train_loss": 0.14295947551727295, "global_step": 76079, "epoch": 854, "lr": 8.199877871086472e-05}
{"train_loss": 0.22271962463855743, "global_step": 76080, "epoch": 854, "lr": 8.199833324185997e-05}
{"train_loss": 0.19513636827468872, "global_step": 76081, "epoch": 854, "lr": 8.199788776855342e-05}
{"train_loss": 0.2262016236782074, "global_step": 76082, "epoch": 854, "lr": 8.199744229094514e-05}
{"train_loss": 0.1756696105003357, "global_step": 76083, "epoch": 854, "lr": 8.199699680903521e-05}
{"train_loss": 0.14877234399318695, "global_step": 76084, "epoch": 854, "lr": 8.199655132282366e-05}
{"train_loss": 0.14088134467601776, "global_step": 76085, "epoch": 854, "lr": 8.199610583231057e-05}
{"train_loss": 0.16908971965312958, "global_step": 76086, "epoch": 854, "lr": 8.199566033749599e-05}
{"train_loss": 0.22634872794151306, "global_step": 76087, "epoch": 854, "lr": 8.199521483838e-05}
{"train_loss": 0.16603516042232513, "global_step": 76088, "epoch": 854, "lr": 8.199476933496261e-05}
{"train_loss": 0.28177979588508606, "global_step": 76089, "epoch": 854, "lr": 8.199432382724395e-05}
{"train_loss": 0.20361582934856415, "global_step": 76090, "epoch": 854, "lr": 8.199387831522402e-05}
{"train_loss": 0.16897346079349518, "global_step": 76091, "epoch": 854, "lr": 8.19934327989029e-05}
{"train_loss": 0.1885046660900116, "global_step": 76092, "epoch": 854, "lr": 8.199298727828066e-05}
{"train_loss": 0.21918781101703644, "global_step": 76093, "epoch": 854, "lr": 8.199254175335735e-05}
{"train_loss": 0.20191767853632403, "global_step": 76094, "epoch": 854, "lr": 8.199209622413304e-05, "val_loss": 4.091236591339111}
{"train_loss": 0.18766124546527863, "global_step": 76095, "epoch": 855, "lr": 8.199165069060778e-05}
{"train_loss": 0.24282145500183105, "global_step": 76096, "epoch": 855, "lr": 8.199120515278163e-05}
{"train_loss": 0.15366533398628235, "global_step": 76097, "epoch": 855, "lr": 8.199075961065466e-05}
{"train_loss": 0.22427065670490265, "global_step": 76098, "epoch": 855, "lr": 8.19903140642269e-05}
{"train_loss": 0.13496193289756775, "global_step": 76099, "epoch": 855, "lr": 8.198986851349846e-05}
{"train_loss": 0.1834847778081894, "global_step": 76100, "epoch": 855, "lr": 8.198942295846936e-05}
{"train_loss": 0.169390007853508, "global_step": 76101, "epoch": 855, "lr": 8.198897739913968e-05}
{"train_loss": 0.15992547571659088, "global_step": 76102, "epoch": 855, "lr": 8.198853183550946e-05}
{"train_loss": 0.2425735592842102, "global_step": 76103, "epoch": 855, "lr": 8.198808626757876e-05}
{"train_loss": 0.18268020451068878, "global_step": 76104, "epoch": 855, "lr": 8.198764069534767e-05}
{"train_loss": 0.21189431846141815, "global_step": 76105, "epoch": 855, "lr": 8.198719511881625e-05}
{"train_loss": 0.24445682764053345, "global_step": 76106, "epoch": 855, "lr": 8.19867495379845e-05}
{"train_loss": 0.21395625174045563, "global_step": 76107, "epoch": 855, "lr": 8.198630395285256e-05}
{"train_loss": 0.20495176315307617, "global_step": 76108, "epoch": 855, "lr": 8.198585836342043e-05}
{"train_loss": 0.13875547051429749, "global_step": 76109, "epoch": 855, "lr": 8.198541276968819e-05}
{"train_loss": 0.1887119710445404, "global_step": 76110, "epoch": 855, "lr": 8.198496717165591e-05}
{"train_loss": 0.2821645140647888, "global_step": 76111, "epoch": 855, "lr": 8.198452156932364e-05}
{"train_loss": 0.19247420132160187, "global_step": 76112, "epoch": 855, "lr": 8.198407596269143e-05}
{"train_loss": 0.2016887366771698, "global_step": 76113, "epoch": 855, "lr": 8.198363035175936e-05}
{"train_loss": 0.15468302369117737, "global_step": 76114, "epoch": 855, "lr": 8.198318473652747e-05}
{"train_loss": 0.26148441433906555, "global_step": 76115, "epoch": 855, "lr": 8.198273911699585e-05}
{"train_loss": 0.2002681940793991, "global_step": 76116, "epoch": 855, "lr": 8.198229349316453e-05}
{"train_loss": 0.11653157323598862, "global_step": 76117, "epoch": 855, "lr": 8.198184786503358e-05}
{"train_loss": 0.19339853525161743, "global_step": 76118, "epoch": 855, "lr": 8.198140223260306e-05}
{"train_loss": 0.13950248062610626, "global_step": 76119, "epoch": 855, "lr": 8.198095659587304e-05}
{"train_loss": 0.11919283866882324, "global_step": 76120, "epoch": 855, "lr": 8.198051095484356e-05}
{"train_loss": 0.1782372146844864, "global_step": 76121, "epoch": 855, "lr": 8.198006530951468e-05}
{"train_loss": 0.23855550587177277, "global_step": 76122, "epoch": 855, "lr": 8.19796196598865e-05}
{"train_loss": 0.28317561745643616, "global_step": 76123, "epoch": 855, "lr": 8.197917400595902e-05}
{"train_loss": 0.18264275789260864, "global_step": 76124, "epoch": 855, "lr": 8.197872834773235e-05}
{"train_loss": 0.1692601591348648, "global_step": 76125, "epoch": 855, "lr": 8.197828268520651e-05}
{"train_loss": 0.24774105846881866, "global_step": 76126, "epoch": 855, "lr": 8.19778370183816e-05}
{"train_loss": 0.13253483176231384, "global_step": 76127, "epoch": 855, "lr": 8.197739134725764e-05}
{"train_loss": 0.21780525147914886, "global_step": 76128, "epoch": 855, "lr": 8.197694567183472e-05}
{"train_loss": 0.150410458445549, "global_step": 76129, "epoch": 855, "lr": 8.197649999211289e-05}
{"train_loss": 0.24377672374248505, "global_step": 76130, "epoch": 855, "lr": 8.19760543080922e-05}
{"train_loss": 0.23726505041122437, "global_step": 76131, "epoch": 855, "lr": 8.197560861977273e-05}
{"train_loss": 0.16851605474948883, "global_step": 76132, "epoch": 855, "lr": 8.197516292715453e-05}
{"train_loss": 0.15681013464927673, "global_step": 76133, "epoch": 855, "lr": 8.197471723023767e-05}
{"train_loss": 0.16308604180812836, "global_step": 76134, "epoch": 855, "lr": 8.197427152902218e-05}
{"train_loss": 0.2309918999671936, "global_step": 76135, "epoch": 855, "lr": 8.197382582350814e-05}
{"train_loss": 0.16674651205539703, "global_step": 76136, "epoch": 855, "lr": 8.19733801136956e-05}
{"train_loss": 0.21116025745868683, "global_step": 76137, "epoch": 855, "lr": 8.197293439958465e-05}
{"train_loss": 0.21235808730125427, "global_step": 76138, "epoch": 855, "lr": 8.197248868117533e-05}
{"train_loss": 0.25980105996131897, "global_step": 76139, "epoch": 855, "lr": 8.197204295846768e-05}
{"train_loss": 0.1339111030101776, "global_step": 76140, "epoch": 855, "lr": 8.19715972314618e-05}
{"train_loss": 0.1498921811580658, "global_step": 76141, "epoch": 855, "lr": 8.197115150015771e-05}
{"train_loss": 0.2500435709953308, "global_step": 76142, "epoch": 855, "lr": 8.19707057645555e-05}
{"train_loss": 0.14384858310222626, "global_step": 76143, "epoch": 855, "lr": 8.19702600246552e-05}
{"train_loss": 0.23337595164775848, "global_step": 76144, "epoch": 855, "lr": 8.19698142804569e-05}
{"train_loss": 0.18998955190181732, "global_step": 76145, "epoch": 855, "lr": 8.196936853196066e-05}
{"train_loss": 0.18510735034942627, "global_step": 76146, "epoch": 855, "lr": 8.196892277916651e-05}
{"train_loss": 0.20701229572296143, "global_step": 76147, "epoch": 855, "lr": 8.196847702207454e-05}
{"train_loss": 0.2702115774154663, "global_step": 76148, "epoch": 855, "lr": 8.196803126068478e-05}
{"train_loss": 0.16800935566425323, "global_step": 76149, "epoch": 855, "lr": 8.196758549499733e-05}
{"train_loss": 0.272876113653183, "global_step": 76150, "epoch": 855, "lr": 8.196713972501221e-05}
{"train_loss": 0.16769763827323914, "global_step": 76151, "epoch": 855, "lr": 8.196669395072952e-05}
{"train_loss": 0.17224517464637756, "global_step": 76152, "epoch": 855, "lr": 8.196624817214927e-05}
{"train_loss": 0.1930559277534485, "global_step": 76153, "epoch": 855, "lr": 8.196580238927157e-05}
{"train_loss": 0.1925467699766159, "global_step": 76154, "epoch": 855, "lr": 8.196535660209644e-05}
{"train_loss": 0.22430357336997986, "global_step": 76155, "epoch": 855, "lr": 8.196491081062396e-05}
{"train_loss": 0.19201387465000153, "global_step": 76156, "epoch": 855, "lr": 8.19644650148542e-05}
{"train_loss": 0.2249741405248642, "global_step": 76157, "epoch": 855, "lr": 8.196401921478721e-05}
{"train_loss": 0.21157337725162506, "global_step": 76158, "epoch": 855, "lr": 8.196357341042303e-05}
{"train_loss": 0.21260467171669006, "global_step": 76159, "epoch": 855, "lr": 8.196312760176175e-05}
{"train_loss": 0.2627445161342621, "global_step": 76160, "epoch": 855, "lr": 8.19626817888034e-05}
{"train_loss": 0.24179776012897491, "global_step": 76161, "epoch": 855, "lr": 8.196223597154809e-05}
{"train_loss": 0.2911843955516815, "global_step": 76162, "epoch": 855, "lr": 8.19617901499958e-05}
{"train_loss": 0.13904578983783722, "global_step": 76163, "epoch": 855, "lr": 8.196134432414668e-05}
{"train_loss": 0.18021248281002045, "global_step": 76164, "epoch": 855, "lr": 8.196089849400073e-05}
{"train_loss": 0.15845312178134918, "global_step": 76165, "epoch": 855, "lr": 8.196045265955802e-05}
{"train_loss": 0.19047264754772186, "global_step": 76166, "epoch": 855, "lr": 8.196000682081862e-05}
{"train_loss": 0.16848565638065338, "global_step": 76167, "epoch": 855, "lr": 8.195956097778258e-05}
{"train_loss": 0.25868499279022217, "global_step": 76168, "epoch": 855, "lr": 8.195911513044998e-05}
{"train_loss": 0.193208709359169, "global_step": 76169, "epoch": 855, "lr": 8.195866927882087e-05}
{"train_loss": 0.2316538691520691, "global_step": 76170, "epoch": 855, "lr": 8.19582234228953e-05}
{"train_loss": 0.16808123886585236, "global_step": 76171, "epoch": 855, "lr": 8.195777756267333e-05}
{"train_loss": 0.18036125600337982, "global_step": 76172, "epoch": 855, "lr": 8.195733169815504e-05}
{"train_loss": 0.16809159517288208, "global_step": 76173, "epoch": 855, "lr": 8.195688582934046e-05}
{"train_loss": 0.17441019415855408, "global_step": 76174, "epoch": 855, "lr": 8.195643995622968e-05}
{"train_loss": 0.18026749789714813, "global_step": 76175, "epoch": 855, "lr": 8.195599407882274e-05}
{"train_loss": 0.21176867187023163, "global_step": 76176, "epoch": 855, "lr": 8.195554819711971e-05}
{"train_loss": 0.16106967628002167, "global_step": 76177, "epoch": 855, "lr": 8.195510231112063e-05}
{"train_loss": 0.25580018758773804, "global_step": 76178, "epoch": 855, "lr": 8.195465642082561e-05}
{"train_loss": 0.19419054687023163, "global_step": 76179, "epoch": 855, "lr": 8.195421052623464e-05}
{"train_loss": 0.22624404728412628, "global_step": 76180, "epoch": 855, "lr": 8.195376462734783e-05}
{"train_loss": 0.22516217827796936, "global_step": 76181, "epoch": 855, "lr": 8.195331872416523e-05}
{"train_loss": 0.2583431303501129, "global_step": 76182, "epoch": 855, "lr": 8.195287281668689e-05}
{"train_loss": 0.19912961301174056, "global_step": 76183, "epoch": 855, "lr": 8.195242690491289e-05, "val_loss": 3.882371664047241, "train_action_mse_error": 42.499332427978516}
{"train_loss": 0.18864944577217102, "global_step": 76184, "epoch": 856, "lr": 8.195198098884325e-05}
{"train_loss": 0.2541217505931854, "global_step": 76185, "epoch": 856, "lr": 8.195153506847807e-05}
{"train_loss": 0.19190391898155212, "global_step": 76186, "epoch": 856, "lr": 8.19510891438174e-05}
{"train_loss": 0.1488518863916397, "global_step": 76187, "epoch": 856, "lr": 8.195064321486128e-05}
{"train_loss": 0.18594396114349365, "global_step": 76188, "epoch": 856, "lr": 8.19501972816098e-05}
{"train_loss": 0.27711474895477295, "global_step": 76189, "epoch": 856, "lr": 8.1949751344063e-05}
{"train_loss": 0.1681196689605713, "global_step": 76190, "epoch": 856, "lr": 8.194930540222094e-05}
{"train_loss": 0.220962256193161, "global_step": 76191, "epoch": 856, "lr": 8.194885945608369e-05}
{"train_loss": 0.20742885768413544, "global_step": 76192, "epoch": 856, "lr": 8.19484135056513e-05}
{"train_loss": 0.24145902693271637, "global_step": 76193, "epoch": 856, "lr": 8.194796755092385e-05}
{"train_loss": 0.23027224838733673, "global_step": 76194, "epoch": 856, "lr": 8.194752159190138e-05}
{"train_loss": 0.20797182619571686, "global_step": 76195, "epoch": 856, "lr": 8.194707562858394e-05}
{"train_loss": 0.3093020021915436, "global_step": 76196, "epoch": 856, "lr": 8.194662966097163e-05}
{"train_loss": 0.2176656275987625, "global_step": 76197, "epoch": 856, "lr": 8.194618368906448e-05}
{"train_loss": 0.1895909458398819, "global_step": 76198, "epoch": 856, "lr": 8.194573771286253e-05}
{"train_loss": 0.24541813135147095, "global_step": 76199, "epoch": 856, "lr": 8.194529173236589e-05}
{"train_loss": 0.2544251084327698, "global_step": 76200, "epoch": 856, "lr": 8.19448457475746e-05}
{"train_loss": 0.2172834873199463, "global_step": 76201, "epoch": 856, "lr": 8.194439975848868e-05}
{"train_loss": 0.1329914629459381, "global_step": 76202, "epoch": 856, "lr": 8.194395376510824e-05}
{"train_loss": 0.19450972974300385, "global_step": 76203, "epoch": 856, "lr": 8.194350776743334e-05}
{"train_loss": 0.182000994682312, "global_step": 76204, "epoch": 856, "lr": 8.194306176546402e-05}
{"train_loss": 0.1617426723241806, "global_step": 76205, "epoch": 856, "lr": 8.194261575920035e-05}
{"train_loss": 0.1963236778974533, "global_step": 76206, "epoch": 856, "lr": 8.194216974864237e-05}
{"train_loss": 0.21093939244747162, "global_step": 76207, "epoch": 856, "lr": 8.194172373379016e-05}
{"train_loss": 0.1862632930278778, "global_step": 76208, "epoch": 856, "lr": 8.194127771464377e-05}
{"train_loss": 0.1604796200990677, "global_step": 76209, "epoch": 856, "lr": 8.194083169120328e-05}
{"train_loss": 0.1728898286819458, "global_step": 76210, "epoch": 856, "lr": 8.194038566346871e-05}
{"train_loss": 0.22349026799201965, "global_step": 76211, "epoch": 856, "lr": 8.193993963144016e-05}
{"train_loss": 0.1707911491394043, "global_step": 76212, "epoch": 856, "lr": 8.193949359511768e-05}
{"train_loss": 0.284275084733963, "global_step": 76213, "epoch": 856, "lr": 8.193904755450131e-05}
{"train_loss": 0.13038040697574615, "global_step": 76214, "epoch": 856, "lr": 8.193860150959115e-05}
{"train_loss": 0.28088030219078064, "global_step": 76215, "epoch": 856, "lr": 8.19381554603872e-05}
{"train_loss": 0.21458721160888672, "global_step": 76216, "epoch": 856, "lr": 8.193770940688957e-05}
{"train_loss": 0.19055432081222534, "global_step": 76217, "epoch": 856, "lr": 8.19372633490983e-05}
{"train_loss": 0.15534654259681702, "global_step": 76218, "epoch": 856, "lr": 8.193681728701345e-05}
{"train_loss": 0.22568455338478088, "global_step": 76219, "epoch": 856, "lr": 8.19363712206351e-05}
{"train_loss": 0.27616387605667114, "global_step": 76220, "epoch": 856, "lr": 8.193592514996329e-05}
{"train_loss": 0.14253802597522736, "global_step": 76221, "epoch": 856, "lr": 8.193547907499808e-05}
{"train_loss": 0.23086993396282196, "global_step": 76222, "epoch": 856, "lr": 8.193503299573953e-05}
{"train_loss": 0.160028874874115, "global_step": 76223, "epoch": 856, "lr": 8.19345869121877e-05}
{"train_loss": 0.18929852545261383, "global_step": 76224, "epoch": 856, "lr": 8.193414082434267e-05}
{"train_loss": 0.3210345208644867, "global_step": 76225, "epoch": 856, "lr": 8.193369473220448e-05}
{"train_loss": 0.20149250328540802, "global_step": 76226, "epoch": 856, "lr": 8.193324863577319e-05}
{"train_loss": 0.23190483450889587, "global_step": 76227, "epoch": 856, "lr": 8.193280253504887e-05}
{"train_loss": 0.16871428489685059, "global_step": 76228, "epoch": 856, "lr": 8.193235643003157e-05}
{"train_loss": 0.2049168050289154, "global_step": 76229, "epoch": 856, "lr": 8.193191032072134e-05}
{"train_loss": 0.21992342174053192, "global_step": 76230, "epoch": 856, "lr": 8.193146420711829e-05}
{"train_loss": 0.24399566650390625, "global_step": 76231, "epoch": 856, "lr": 8.19310180892224e-05}
{"train_loss": 0.17493389546871185, "global_step": 76232, "epoch": 856, "lr": 8.19305719670338e-05}
{"train_loss": 0.14609843492507935, "global_step": 76233, "epoch": 856, "lr": 8.19301258405525e-05}
{"train_loss": 0.2291782647371292, "global_step": 76234, "epoch": 856, "lr": 8.192967970977862e-05}
{"train_loss": 0.1665426641702652, "global_step": 76235, "epoch": 856, "lr": 8.192923357471216e-05}
{"train_loss": 0.25060757994651794, "global_step": 76236, "epoch": 856, "lr": 8.192878743535322e-05}
{"train_loss": 0.23961573839187622, "global_step": 76237, "epoch": 856, "lr": 8.192834129170183e-05}
{"train_loss": 0.1675679236650467, "global_step": 76238, "epoch": 856, "lr": 8.192789514375804e-05}
{"train_loss": 0.12002598494291306, "global_step": 76239, "epoch": 856, "lr": 8.192744899152197e-05}
{"train_loss": 0.1455548256635666, "global_step": 76240, "epoch": 856, "lr": 8.192700283499364e-05}
{"train_loss": 0.19351336359977722, "global_step": 76241, "epoch": 856, "lr": 8.192655667417311e-05}
{"train_loss": 0.27358174324035645, "global_step": 76242, "epoch": 856, "lr": 8.192611050906043e-05}
{"train_loss": 0.08603715896606445, "global_step": 76243, "epoch": 856, "lr": 8.19256643396557e-05}
{"train_loss": 0.1707751452922821, "global_step": 76244, "epoch": 856, "lr": 8.192521816595893e-05}
{"train_loss": 0.11994989961385727, "global_step": 76245, "epoch": 856, "lr": 8.192477198797022e-05}
{"train_loss": 0.13330888748168945, "global_step": 76246, "epoch": 856, "lr": 8.19243258056896e-05}
{"train_loss": 0.17813342809677124, "global_step": 76247, "epoch": 856, "lr": 8.192387961911714e-05}
{"train_loss": 0.2637541592121124, "global_step": 76248, "epoch": 856, "lr": 8.192343342825292e-05}
{"train_loss": 0.22484835982322693, "global_step": 76249, "epoch": 856, "lr": 8.192298723309697e-05}
{"train_loss": 0.25255322456359863, "global_step": 76250, "epoch": 856, "lr": 8.192254103364937e-05}
{"train_loss": 0.19148673117160797, "global_step": 76251, "epoch": 856, "lr": 8.192209482991018e-05}
{"train_loss": 0.19097477197647095, "global_step": 76252, "epoch": 856, "lr": 8.192164862187944e-05}
{"train_loss": 0.354332834482193, "global_step": 76253, "epoch": 856, "lr": 8.192120240955723e-05}
{"train_loss": 0.2222457379102707, "global_step": 76254, "epoch": 856, "lr": 8.192075619294359e-05}
{"train_loss": 0.24818232655525208, "global_step": 76255, "epoch": 856, "lr": 8.192030997203861e-05}
{"train_loss": 0.2005355805158615, "global_step": 76256, "epoch": 856, "lr": 8.191986374684233e-05}
{"train_loss": 0.13541066646575928, "global_step": 76257, "epoch": 856, "lr": 8.191941751735481e-05}
{"train_loss": 0.2811182141304016, "global_step": 76258, "epoch": 856, "lr": 8.191897128357612e-05}
{"train_loss": 0.16625210642814636, "global_step": 76259, "epoch": 856, "lr": 8.19185250455063e-05}
{"train_loss": 0.15655018389225006, "global_step": 76260, "epoch": 856, "lr": 8.191807880314544e-05}
{"train_loss": 0.1728312224149704, "global_step": 76261, "epoch": 856, "lr": 8.191763255649358e-05}
{"train_loss": 0.22635005414485931, "global_step": 76262, "epoch": 856, "lr": 8.191718630555078e-05}
{"train_loss": 0.14068524539470673, "global_step": 76263, "epoch": 856, "lr": 8.19167400503171e-05}
{"train_loss": 0.1864432990550995, "global_step": 76264, "epoch": 856, "lr": 8.19162937907926e-05}
{"train_loss": 0.18155749142169952, "global_step": 76265, "epoch": 856, "lr": 8.191584752697735e-05}
{"train_loss": 0.23832806944847107, "global_step": 76266, "epoch": 856, "lr": 8.19154012588714e-05}
{"train_loss": 0.21414510905742645, "global_step": 76267, "epoch": 856, "lr": 8.191495498647483e-05}
{"train_loss": 0.2220762073993683, "global_step": 76268, "epoch": 856, "lr": 8.191450870978766e-05}
{"train_loss": 0.15482187271118164, "global_step": 76269, "epoch": 856, "lr": 8.191406242881e-05}
{"train_loss": 0.2420826405286789, "global_step": 76270, "epoch": 856, "lr": 8.191361614354185e-05}
{"train_loss": 0.2558273375034332, "global_step": 76271, "epoch": 856, "lr": 8.191316985398332e-05}
{"train_loss": 0.2043327329198966, "global_step": 76272, "epoch": 856, "lr": 8.191272356013445e-05, "val_loss": 3.9821221828460693}
{"train_loss": 0.14749078452587128, "global_step": 76273, "epoch": 857, "lr": 8.19122772619953e-05}
{"train_loss": 0.24461977183818817, "global_step": 76274, "epoch": 857, "lr": 8.191183095956594e-05}
{"train_loss": 0.24748829007148743, "global_step": 76275, "epoch": 857, "lr": 8.191138465284642e-05}
{"train_loss": 0.23055127263069153, "global_step": 76276, "epoch": 857, "lr": 8.19109383418368e-05}
{"train_loss": 0.14843857288360596, "global_step": 76277, "epoch": 857, "lr": 8.191049202653716e-05}
{"train_loss": 0.2113274335861206, "global_step": 76278, "epoch": 857, "lr": 8.191004570694752e-05}
{"train_loss": 0.19068844616413116, "global_step": 76279, "epoch": 857, "lr": 8.190959938306798e-05}
{"train_loss": 0.1951625645160675, "global_step": 76280, "epoch": 857, "lr": 8.190915305489858e-05}
{"train_loss": 0.09059373289346695, "global_step": 76281, "epoch": 857, "lr": 8.190870672243937e-05}
{"train_loss": 0.1547544151544571, "global_step": 76282, "epoch": 857, "lr": 8.190826038569046e-05}
{"train_loss": 0.2498767077922821, "global_step": 76283, "epoch": 857, "lr": 8.190781404465184e-05}
{"train_loss": 0.18715815246105194, "global_step": 76284, "epoch": 857, "lr": 8.190736769932361e-05}
{"train_loss": 0.24793896079063416, "global_step": 76285, "epoch": 857, "lr": 8.190692134970582e-05}
{"train_loss": 0.25114163756370544, "global_step": 76286, "epoch": 857, "lr": 8.190647499579854e-05}
{"train_loss": 0.2567731738090515, "global_step": 76287, "epoch": 857, "lr": 8.190602863760182e-05}
{"train_loss": 0.2603934407234192, "global_step": 76288, "epoch": 857, "lr": 8.190558227511573e-05}
{"train_loss": 0.1176934540271759, "global_step": 76289, "epoch": 857, "lr": 8.190513590834032e-05}
{"train_loss": 0.18392755091190338, "global_step": 76290, "epoch": 857, "lr": 8.190468953727565e-05}
{"train_loss": 0.15972977876663208, "global_step": 76291, "epoch": 857, "lr": 8.190424316192178e-05}
{"train_loss": 0.23827292025089264, "global_step": 76292, "epoch": 857, "lr": 8.190379678227878e-05}
{"train_loss": 0.1892266869544983, "global_step": 76293, "epoch": 857, "lr": 8.19033503983467e-05}
{"train_loss": 0.1907234787940979, "global_step": 76294, "epoch": 857, "lr": 8.19029040101256e-05}
{"train_loss": 0.21224327385425568, "global_step": 76295, "epoch": 857, "lr": 8.190245761761554e-05}
{"train_loss": 0.18985937535762787, "global_step": 76296, "epoch": 857, "lr": 8.19020112208166e-05}
{"train_loss": 0.18133194744586945, "global_step": 76297, "epoch": 857, "lr": 8.190156481972882e-05}
{"train_loss": 0.30283355712890625, "global_step": 76298, "epoch": 857, "lr": 8.190111841435226e-05}
{"train_loss": 0.1657886505126953, "global_step": 76299, "epoch": 857, "lr": 8.190067200468699e-05}
{"train_loss": 0.12849144637584686, "global_step": 76300, "epoch": 857, "lr": 8.190022559073305e-05}
{"train_loss": 0.17001089453697205, "global_step": 76301, "epoch": 857, "lr": 8.189977917249052e-05}
{"train_loss": 0.1444270759820938, "global_step": 76302, "epoch": 857, "lr": 8.189933274995945e-05}
{"train_loss": 0.14699877798557281, "global_step": 76303, "epoch": 857, "lr": 8.189888632313992e-05}
{"train_loss": 0.23183846473693848, "global_step": 76304, "epoch": 857, "lr": 8.189843989203195e-05}
{"train_loss": 0.1724502444267273, "global_step": 76305, "epoch": 857, "lr": 8.189799345663562e-05}
{"train_loss": 0.08049885183572769, "global_step": 76306, "epoch": 857, "lr": 8.189754701695101e-05}
{"train_loss": 0.2377782166004181, "global_step": 76307, "epoch": 857, "lr": 8.189710057297817e-05}
{"train_loss": 0.22034405171871185, "global_step": 76308, "epoch": 857, "lr": 8.189665412471714e-05}
{"train_loss": 0.2073577642440796, "global_step": 76309, "epoch": 857, "lr": 8.189620767216799e-05}
{"train_loss": 0.18761351704597473, "global_step": 76310, "epoch": 857, "lr": 8.18957612153308e-05}
{"train_loss": 0.24259401857852936, "global_step": 76311, "epoch": 857, "lr": 8.18953147542056e-05}
{"train_loss": 0.17192895710468292, "global_step": 76312, "epoch": 857, "lr": 8.189486828879247e-05}
{"train_loss": 0.21232211589813232, "global_step": 76313, "epoch": 857, "lr": 8.189442181909146e-05}
{"train_loss": 0.22514905035495758, "global_step": 76314, "epoch": 857, "lr": 8.189397534510264e-05}
{"train_loss": 0.20892533659934998, "global_step": 76315, "epoch": 857, "lr": 8.189352886682606e-05}
{"train_loss": 0.20614615082740784, "global_step": 76316, "epoch": 857, "lr": 8.189308238426178e-05}
{"train_loss": 0.19485771656036377, "global_step": 76317, "epoch": 857, "lr": 8.189263589740986e-05}
{"train_loss": 0.2255655825138092, "global_step": 76318, "epoch": 857, "lr": 8.189218940627037e-05}
{"train_loss": 0.1980236917734146, "global_step": 76319, "epoch": 857, "lr": 8.189174291084338e-05}
{"train_loss": 0.15996316075325012, "global_step": 76320, "epoch": 857, "lr": 8.189129641112891e-05}
{"train_loss": 0.18024824559688568, "global_step": 76321, "epoch": 857, "lr": 8.189084990712705e-05}
{"train_loss": 0.13338258862495422, "global_step": 76322, "epoch": 857, "lr": 8.189040339883787e-05}
{"train_loss": 0.15696324408054352, "global_step": 76323, "epoch": 857, "lr": 8.18899568862614e-05}
{"train_loss": 0.2142987698316574, "global_step": 76324, "epoch": 857, "lr": 8.188951036939772e-05}
{"train_loss": 0.1872248351573944, "global_step": 76325, "epoch": 857, "lr": 8.188906384824685e-05}
{"train_loss": 0.11341157555580139, "global_step": 76326, "epoch": 857, "lr": 8.188861732280892e-05}
{"train_loss": 0.2116246372461319, "global_step": 76327, "epoch": 857, "lr": 8.188817079308395e-05}
{"train_loss": 0.19893954694271088, "global_step": 76328, "epoch": 857, "lr": 8.188772425907201e-05}
{"train_loss": 0.20707479119300842, "global_step": 76329, "epoch": 857, "lr": 8.188727772077313e-05}
{"train_loss": 0.18397657573223114, "global_step": 76330, "epoch": 857, "lr": 8.188683117818742e-05}
{"train_loss": 0.17491033673286438, "global_step": 76331, "epoch": 857, "lr": 8.188638463131491e-05}
{"train_loss": 0.20589575171470642, "global_step": 76332, "epoch": 857, "lr": 8.188593808015565e-05}
{"train_loss": 0.2695472538471222, "global_step": 76333, "epoch": 857, "lr": 8.188549152470972e-05}
{"train_loss": 0.21445779502391815, "global_step": 76334, "epoch": 857, "lr": 8.188504496497717e-05}
{"train_loss": 0.1475161612033844, "global_step": 76335, "epoch": 857, "lr": 8.188459840095808e-05}
{"train_loss": 0.13839253783226013, "global_step": 76336, "epoch": 857, "lr": 8.188415183265248e-05}
{"train_loss": 0.09462438523769379, "global_step": 76337, "epoch": 857, "lr": 8.188370526006045e-05}
{"train_loss": 0.2246340662240982, "global_step": 76338, "epoch": 857, "lr": 8.188325868318204e-05}
{"train_loss": 0.15626409649848938, "global_step": 76339, "epoch": 857, "lr": 8.188281210201732e-05}
{"train_loss": 0.2165803462266922, "global_step": 76340, "epoch": 857, "lr": 8.188236551656634e-05}
{"train_loss": 0.18333080410957336, "global_step": 76341, "epoch": 857, "lr": 8.188191892682917e-05}
{"train_loss": 0.23042204976081848, "global_step": 76342, "epoch": 857, "lr": 8.188147233280587e-05}
{"train_loss": 0.2486032098531723, "global_step": 76343, "epoch": 857, "lr": 8.188102573449648e-05}
{"train_loss": 0.3124161660671234, "global_step": 76344, "epoch": 857, "lr": 8.188057913190108e-05}
{"train_loss": 0.1583869904279709, "global_step": 76345, "epoch": 857, "lr": 8.188013252501972e-05}
{"train_loss": 0.22640466690063477, "global_step": 76346, "epoch": 857, "lr": 8.187968591385248e-05}
{"train_loss": 0.16208425164222717, "global_step": 76347, "epoch": 857, "lr": 8.18792392983994e-05}
{"train_loss": 0.19177432358264923, "global_step": 76348, "epoch": 857, "lr": 8.187879267866052e-05}
{"train_loss": 0.21208931505680084, "global_step": 76349, "epoch": 857, "lr": 8.187834605463596e-05}
{"train_loss": 0.23163892328739166, "global_step": 76350, "epoch": 857, "lr": 8.187789942632572e-05}
{"train_loss": 0.17338581383228302, "global_step": 76351, "epoch": 857, "lr": 8.18774527937299e-05}
{"train_loss": 0.2600802779197693, "global_step": 76352, "epoch": 857, "lr": 8.187700615684855e-05}
{"train_loss": 0.17688164114952087, "global_step": 76353, "epoch": 857, "lr": 8.18765595156817e-05}
{"train_loss": 0.28934425115585327, "global_step": 76354, "epoch": 857, "lr": 8.187611287022945e-05}
{"train_loss": 0.253436803817749, "global_step": 76355, "epoch": 857, "lr": 8.187566622049186e-05}
{"train_loss": 0.17741526663303375, "global_step": 76356, "epoch": 857, "lr": 8.187521956646895e-05}
{"train_loss": 0.3167646527290344, "global_step": 76357, "epoch": 857, "lr": 8.187477290816081e-05}
{"train_loss": 0.2287912666797638, "global_step": 76358, "epoch": 857, "lr": 8.18743262455675e-05}
{"train_loss": 0.18104790151119232, "global_step": 76359, "epoch": 857, "lr": 8.187387957868908e-05}
{"train_loss": 0.20756413042545319, "global_step": 76360, "epoch": 857, "lr": 8.18734329075256e-05}
{"train_loss": 0.19797007873486938, "global_step": 76361, "epoch": 857, "lr": 8.187298623207713e-05, "val_loss": 3.940140962600708}
{"train_loss": 0.342295378446579, "global_step": 76362, "epoch": 858, "lr": 8.187253955234373e-05}
{"train_loss": 0.23997095227241516, "global_step": 76363, "epoch": 858, "lr": 8.187209286832545e-05}
{"train_loss": 0.20165549218654633, "global_step": 76364, "epoch": 858, "lr": 8.187164618002235e-05}
{"train_loss": 0.18182602524757385, "global_step": 76365, "epoch": 858, "lr": 8.18711994874345e-05}
{"train_loss": 0.17649218440055847, "global_step": 76366, "epoch": 858, "lr": 8.187075279056194e-05}
{"train_loss": 0.19892051815986633, "global_step": 76367, "epoch": 858, "lr": 8.187030608940476e-05}
{"train_loss": 0.16455134749412537, "global_step": 76368, "epoch": 858, "lr": 8.186985938396301e-05}
{"train_loss": 0.2643750011920929, "global_step": 76369, "epoch": 858, "lr": 8.186941267423674e-05}
{"train_loss": 0.21355120837688446, "global_step": 76370, "epoch": 858, "lr": 8.186896596022601e-05}
{"train_loss": 0.14733269810676575, "global_step": 76371, "epoch": 858, "lr": 8.18685192419309e-05}
{"train_loss": 0.27103620767593384, "global_step": 76372, "epoch": 858, "lr": 8.186807251935145e-05}
{"train_loss": 0.22267913818359375, "global_step": 76373, "epoch": 858, "lr": 8.186762579248773e-05}
{"train_loss": 0.23288904130458832, "global_step": 76374, "epoch": 858, "lr": 8.186717906133979e-05}
{"train_loss": 0.15191614627838135, "global_step": 76375, "epoch": 858, "lr": 8.18667323259077e-05}
{"train_loss": 0.24462224543094635, "global_step": 76376, "epoch": 858, "lr": 8.186628558619151e-05}
{"train_loss": 0.20376813411712646, "global_step": 76377, "epoch": 858, "lr": 8.18658388421913e-05}
{"train_loss": 0.19655954837799072, "global_step": 76378, "epoch": 858, "lr": 8.186539209390711e-05}
{"train_loss": 0.20287618041038513, "global_step": 76379, "epoch": 858, "lr": 8.1864945341339e-05}
{"train_loss": 0.1696583330631256, "global_step": 76380, "epoch": 858, "lr": 8.186449858448704e-05}
{"train_loss": 0.15850259363651276, "global_step": 76381, "epoch": 858, "lr": 8.186405182335129e-05}
{"train_loss": 0.307657390832901, "global_step": 76382, "epoch": 858, "lr": 8.18636050579318e-05}
{"train_loss": 0.19293861091136932, "global_step": 76383, "epoch": 858, "lr": 8.186315828822863e-05}
{"train_loss": 0.21037742495536804, "global_step": 76384, "epoch": 858, "lr": 8.186271151424185e-05}
{"train_loss": 0.2275395542383194, "global_step": 76385, "epoch": 858, "lr": 8.186226473597154e-05}
{"train_loss": 0.19847917556762695, "global_step": 76386, "epoch": 858, "lr": 8.186181795341772e-05}
{"train_loss": 0.193595290184021, "global_step": 76387, "epoch": 858, "lr": 8.186137116658046e-05}
{"train_loss": 0.23440784215927124, "global_step": 76388, "epoch": 858, "lr": 8.186092437545985e-05}
{"train_loss": 0.23550379276275635, "global_step": 76389, "epoch": 858, "lr": 8.186047758005591e-05}
{"train_loss": 0.3056226670742035, "global_step": 76390, "epoch": 858, "lr": 8.18600307803687e-05}
{"train_loss": 0.16507655382156372, "global_step": 76391, "epoch": 858, "lr": 8.185958397639833e-05}
{"train_loss": 0.3161464035511017, "global_step": 76392, "epoch": 858, "lr": 8.18591371681448e-05}
{"train_loss": 0.27798187732696533, "global_step": 76393, "epoch": 858, "lr": 8.185869035560822e-05}
{"train_loss": 0.11587728559970856, "global_step": 76394, "epoch": 858, "lr": 8.185824353878861e-05}
{"train_loss": 0.16347001492977142, "global_step": 76395, "epoch": 858, "lr": 8.185779671768606e-05}
{"train_loss": 0.22314512729644775, "global_step": 76396, "epoch": 858, "lr": 8.185734989230061e-05}
{"train_loss": 0.18556854128837585, "global_step": 76397, "epoch": 858, "lr": 8.185690306263233e-05}
{"train_loss": 0.16955845057964325, "global_step": 76398, "epoch": 858, "lr": 8.185645622868128e-05}
{"train_loss": 0.2277996838092804, "global_step": 76399, "epoch": 858, "lr": 8.185600939044752e-05}
{"train_loss": 0.2682982385158539, "global_step": 76400, "epoch": 858, "lr": 8.18555625479311e-05}
{"train_loss": 0.23340219259262085, "global_step": 76401, "epoch": 858, "lr": 8.185511570113209e-05}
{"train_loss": 0.20429624617099762, "global_step": 76402, "epoch": 858, "lr": 8.185466885005055e-05}
{"train_loss": 0.20017004013061523, "global_step": 76403, "epoch": 858, "lr": 8.185422199468653e-05}
{"train_loss": 0.18750189244747162, "global_step": 76404, "epoch": 858, "lr": 8.185377513504012e-05}
{"train_loss": 0.19046194851398468, "global_step": 76405, "epoch": 858, "lr": 8.185332827111133e-05}
{"train_loss": 0.28601911664009094, "global_step": 76406, "epoch": 858, "lr": 8.185288140290027e-05}
{"train_loss": 0.2493952214717865, "global_step": 76407, "epoch": 858, "lr": 8.185243453040697e-05}
{"train_loss": 0.22861231863498688, "global_step": 76408, "epoch": 858, "lr": 8.185198765363149e-05}
{"train_loss": 0.2817385792732239, "global_step": 76409, "epoch": 858, "lr": 8.185154077257391e-05}
{"train_loss": 0.27546554803848267, "global_step": 76410, "epoch": 858, "lr": 8.185109388723428e-05}
{"train_loss": 0.27617713809013367, "global_step": 76411, "epoch": 858, "lr": 8.185064699761264e-05}
{"train_loss": 0.2117384672164917, "global_step": 76412, "epoch": 858, "lr": 8.18502001037091e-05}
{"train_loss": 0.17840594053268433, "global_step": 76413, "epoch": 858, "lr": 8.184975320552366e-05}
{"train_loss": 0.19078385829925537, "global_step": 76414, "epoch": 858, "lr": 8.184930630305643e-05}
{"train_loss": 0.22741225361824036, "global_step": 76415, "epoch": 858, "lr": 8.184885939630745e-05}
{"train_loss": 0.1582222729921341, "global_step": 76416, "epoch": 858, "lr": 8.184841248527675e-05}
{"train_loss": 0.2652926743030548, "global_step": 76417, "epoch": 858, "lr": 8.184796556996444e-05}
{"train_loss": 0.20313173532485962, "global_step": 76418, "epoch": 858, "lr": 8.184751865037057e-05}
{"train_loss": 0.24710242450237274, "global_step": 76419, "epoch": 858, "lr": 8.184707172649518e-05}
{"train_loss": 0.19865021109580994, "global_step": 76420, "epoch": 858, "lr": 8.184662479833831e-05}
{"train_loss": 0.12053903192281723, "global_step": 76421, "epoch": 858, "lr": 8.184617786590007e-05}
{"train_loss": 0.175207257270813, "global_step": 76422, "epoch": 858, "lr": 8.18457309291805e-05}
{"train_loss": 0.16356448829174042, "global_step": 76423, "epoch": 858, "lr": 8.184528398817968e-05}
{"train_loss": 0.15725940465927124, "global_step": 76424, "epoch": 858, "lr": 8.184483704289764e-05}
{"train_loss": 0.18456736207008362, "global_step": 76425, "epoch": 858, "lr": 8.184439009333445e-05}
{"train_loss": 0.22563472390174866, "global_step": 76426, "epoch": 858, "lr": 8.184394313949015e-05}
{"train_loss": 0.18791081011295319, "global_step": 76427, "epoch": 858, "lr": 8.184349618136484e-05}
{"train_loss": 0.14593985676765442, "global_step": 76428, "epoch": 858, "lr": 8.184304921895856e-05}
{"train_loss": 0.1719527244567871, "global_step": 76429, "epoch": 858, "lr": 8.184260225227136e-05}
{"train_loss": 0.21631576120853424, "global_step": 76430, "epoch": 858, "lr": 8.18421552813033e-05}
{"train_loss": 0.20239800214767456, "global_step": 76431, "epoch": 858, "lr": 8.184170830605447e-05}
{"train_loss": 0.19595646858215332, "global_step": 76432, "epoch": 858, "lr": 8.18412613265249e-05}
{"train_loss": 0.26148998737335205, "global_step": 76433, "epoch": 858, "lr": 8.184081434271467e-05}
{"train_loss": 0.14912955462932587, "global_step": 76434, "epoch": 858, "lr": 8.184036735462383e-05}
{"train_loss": 0.09203783422708511, "global_step": 76435, "epoch": 858, "lr": 8.183992036225243e-05}
{"train_loss": 0.22439923882484436, "global_step": 76436, "epoch": 858, "lr": 8.183947336560056e-05}
{"train_loss": 0.22738921642303467, "global_step": 76437, "epoch": 858, "lr": 8.183902636466825e-05}
{"train_loss": 0.25569388270378113, "global_step": 76438, "epoch": 858, "lr": 8.183857935945556e-05}
{"train_loss": 0.2008238136768341, "global_step": 76439, "epoch": 858, "lr": 8.183813234996256e-05}
{"train_loss": 0.1751444935798645, "global_step": 76440, "epoch": 858, "lr": 8.183768533618932e-05}
{"train_loss": 0.31258875131607056, "global_step": 76441, "epoch": 858, "lr": 8.18372383181359e-05}
{"train_loss": 0.19364973902702332, "global_step": 76442, "epoch": 858, "lr": 8.183679129580234e-05}
{"train_loss": 0.3730626404285431, "global_step": 76443, "epoch": 858, "lr": 8.183634426918871e-05}
{"train_loss": 0.24439312517642975, "global_step": 76444, "epoch": 858, "lr": 8.183589723829507e-05}
{"train_loss": 0.17890943586826324, "global_step": 76445, "epoch": 858, "lr": 8.18354502031215e-05}
{"train_loss": 0.253009170293808, "global_step": 76446, "epoch": 858, "lr": 8.183500316366804e-05}
{"train_loss": 0.15451040863990784, "global_step": 76447, "epoch": 858, "lr": 8.183455611993473e-05}
{"train_loss": 0.23818311095237732, "global_step": 76448, "epoch": 858, "lr": 8.183410907192168e-05}
{"train_loss": 0.21864989399909973, "global_step": 76449, "epoch": 858, "lr": 8.18336620196289e-05}
{"train_loss": 0.21317258108867687, "global_step": 76450, "epoch": 858, "lr": 8.183321496305647e-05, "val_loss": 4.0920090675354}
{"train_loss": 0.20805664360523224, "global_step": 76451, "epoch": 859, "lr": 8.183276790220447e-05}
{"train_loss": 0.2163054496049881, "global_step": 76452, "epoch": 859, "lr": 8.183232083707293e-05}
{"train_loss": 0.27956175804138184, "global_step": 76453, "epoch": 859, "lr": 8.183187376766193e-05}
{"train_loss": 0.24663394689559937, "global_step": 76454, "epoch": 859, "lr": 8.183142669397151e-05}
{"train_loss": 0.21564635634422302, "global_step": 76455, "epoch": 859, "lr": 8.183097961600174e-05}
{"train_loss": 0.1437317430973053, "global_step": 76456, "epoch": 859, "lr": 8.18305325337527e-05}
{"train_loss": 0.17554277181625366, "global_step": 76457, "epoch": 859, "lr": 8.183008544722443e-05}
{"train_loss": 0.19095979630947113, "global_step": 76458, "epoch": 859, "lr": 8.182963835641698e-05}
{"train_loss": 0.16080209612846375, "global_step": 76459, "epoch": 859, "lr": 8.182919126133044e-05}
{"train_loss": 0.18636731803417206, "global_step": 76460, "epoch": 859, "lr": 8.182874416196484e-05}
{"train_loss": 0.2536424696445465, "global_step": 76461, "epoch": 859, "lr": 8.182829705832026e-05}
{"train_loss": 0.20087115466594696, "global_step": 76462, "epoch": 859, "lr": 8.182784995039675e-05}
{"train_loss": 0.1800319254398346, "global_step": 76463, "epoch": 859, "lr": 8.182740283819439e-05}
{"train_loss": 0.2881070077419281, "global_step": 76464, "epoch": 859, "lr": 8.18269557217132e-05}
{"train_loss": 0.24548450112342834, "global_step": 76465, "epoch": 859, "lr": 8.182650860095326e-05}
{"train_loss": 0.31746983528137207, "global_step": 76466, "epoch": 859, "lr": 8.182606147591466e-05}
{"train_loss": 0.17712508141994476, "global_step": 76467, "epoch": 859, "lr": 8.182561434659741e-05}
{"train_loss": 0.1369250863790512, "global_step": 76468, "epoch": 859, "lr": 8.182516721300162e-05}
{"train_loss": 0.228229820728302, "global_step": 76469, "epoch": 859, "lr": 8.18247200751273e-05}
{"train_loss": 0.3076103627681732, "global_step": 76470, "epoch": 859, "lr": 8.182427293297454e-05}
{"train_loss": 0.17270015180110931, "global_step": 76471, "epoch": 859, "lr": 8.182382578654341e-05}
{"train_loss": 0.14615964889526367, "global_step": 76472, "epoch": 859, "lr": 8.182337863583394e-05}
{"train_loss": 0.25643858313560486, "global_step": 76473, "epoch": 859, "lr": 8.182293148084621e-05}
{"train_loss": 0.177794948220253, "global_step": 76474, "epoch": 859, "lr": 8.182248432158027e-05}
{"train_loss": 0.23929812014102936, "global_step": 76475, "epoch": 859, "lr": 8.182203715803619e-05}
{"train_loss": 0.15854519605636597, "global_step": 76476, "epoch": 859, "lr": 8.182158999021401e-05}
{"train_loss": 0.17222563922405243, "global_step": 76477, "epoch": 859, "lr": 8.182114281811382e-05}
{"train_loss": 0.1641586273908615, "global_step": 76478, "epoch": 859, "lr": 8.182069564173565e-05}
{"train_loss": 0.23870113492012024, "global_step": 76479, "epoch": 859, "lr": 8.182024846107959e-05}
{"train_loss": 0.19916893541812897, "global_step": 76480, "epoch": 859, "lr": 8.181980127614568e-05}
{"train_loss": 0.1315714418888092, "global_step": 76481, "epoch": 859, "lr": 8.181935408693399e-05}
{"train_loss": 0.24070116877555847, "global_step": 76482, "epoch": 859, "lr": 8.181890689344459e-05}
{"train_loss": 0.365064412355423, "global_step": 76483, "epoch": 859, "lr": 8.18184596956775e-05}
{"train_loss": 0.19746418297290802, "global_step": 76484, "epoch": 859, "lr": 8.181801249363282e-05}
{"train_loss": 0.29549887776374817, "global_step": 76485, "epoch": 859, "lr": 8.181756528731058e-05}
{"train_loss": 0.2298392653465271, "global_step": 76486, "epoch": 859, "lr": 8.181711807671086e-05}
{"train_loss": 0.18259915709495544, "global_step": 76487, "epoch": 859, "lr": 8.181667086183373e-05}
{"train_loss": 0.2289014458656311, "global_step": 76488, "epoch": 859, "lr": 8.181622364267922e-05}
{"train_loss": 0.25125396251678467, "global_step": 76489, "epoch": 859, "lr": 8.181577641924742e-05}
{"train_loss": 0.22096572816371918, "global_step": 76490, "epoch": 859, "lr": 8.181532919153838e-05}
{"train_loss": 0.22178351879119873, "global_step": 76491, "epoch": 859, "lr": 8.181488195955216e-05}
{"train_loss": 0.24345090985298157, "global_step": 76492, "epoch": 859, "lr": 8.181443472328879e-05}
{"train_loss": 0.26324471831321716, "global_step": 76493, "epoch": 859, "lr": 8.181398748274837e-05}
{"train_loss": 0.22663727402687073, "global_step": 76494, "epoch": 859, "lr": 8.181354023793095e-05}
{"train_loss": 0.17634320259094238, "global_step": 76495, "epoch": 859, "lr": 8.18130929888366e-05}
{"train_loss": 0.21526473760604858, "global_step": 76496, "epoch": 859, "lr": 8.181264573546534e-05}
{"train_loss": 0.2266852706670761, "global_step": 76497, "epoch": 859, "lr": 8.181219847781728e-05}
{"train_loss": 0.163677379488945, "global_step": 76498, "epoch": 859, "lr": 8.181175121589245e-05}
{"train_loss": 0.1757676601409912, "global_step": 76499, "epoch": 859, "lr": 8.181130394969092e-05}
{"train_loss": 0.25459495186805725, "global_step": 76500, "epoch": 859, "lr": 8.181085667921276e-05}
{"train_loss": 0.1958485096693039, "global_step": 76501, "epoch": 859, "lr": 8.181040940445798e-05}
{"train_loss": 0.24705958366394043, "global_step": 76502, "epoch": 859, "lr": 8.180996212542672e-05}
{"train_loss": 0.23978328704833984, "global_step": 76503, "epoch": 859, "lr": 8.180951484211899e-05}
{"train_loss": 0.18297430872917175, "global_step": 76504, "epoch": 859, "lr": 8.180906755453483e-05}
{"train_loss": 0.1969071328639984, "global_step": 76505, "epoch": 859, "lr": 8.180862026267435e-05}
{"train_loss": 0.19983594119548798, "global_step": 76506, "epoch": 859, "lr": 8.18081729665376e-05}
{"train_loss": 0.2111639827489853, "global_step": 76507, "epoch": 859, "lr": 8.18077256661246e-05}
{"train_loss": 0.25103384256362915, "global_step": 76508, "epoch": 859, "lr": 8.180727836143546e-05}
{"train_loss": 0.17427417635917664, "global_step": 76509, "epoch": 859, "lr": 8.180683105247022e-05}
{"train_loss": 0.16476809978485107, "global_step": 76510, "epoch": 859, "lr": 8.180638373922893e-05}
{"train_loss": 0.20789018273353577, "global_step": 76511, "epoch": 859, "lr": 8.180593642171166e-05}
{"train_loss": 0.17158223688602448, "global_step": 76512, "epoch": 859, "lr": 8.180548909991848e-05}
{"train_loss": 0.19159051775932312, "global_step": 76513, "epoch": 859, "lr": 8.180504177384943e-05}
{"train_loss": 0.21805061399936676, "global_step": 76514, "epoch": 859, "lr": 8.180459444350457e-05}
{"train_loss": 0.18164698779582977, "global_step": 76515, "epoch": 859, "lr": 8.180414710888397e-05}
{"train_loss": 0.1966635137796402, "global_step": 76516, "epoch": 859, "lr": 8.180369976998771e-05}
{"train_loss": 0.25010305643081665, "global_step": 76517, "epoch": 859, "lr": 8.180325242681582e-05}
{"train_loss": 0.15616077184677124, "global_step": 76518, "epoch": 859, "lr": 8.180280507936837e-05}
{"train_loss": 0.2854948043823242, "global_step": 76519, "epoch": 859, "lr": 8.180235772764541e-05}
{"train_loss": 0.17879082262516022, "global_step": 76520, "epoch": 859, "lr": 8.180191037164704e-05}
{"train_loss": 0.1886398047208786, "global_step": 76521, "epoch": 859, "lr": 8.180146301137329e-05}
{"train_loss": 0.28267422318458557, "global_step": 76522, "epoch": 859, "lr": 8.180101564682418e-05}
{"train_loss": 0.21066772937774658, "global_step": 76523, "epoch": 859, "lr": 8.180056827799985e-05}
{"train_loss": 0.13525734841823578, "global_step": 76524, "epoch": 859, "lr": 8.18001209049003e-05}
{"train_loss": 0.16177509725093842, "global_step": 76525, "epoch": 859, "lr": 8.179967352752562e-05}
{"train_loss": 0.2662833333015442, "global_step": 76526, "epoch": 859, "lr": 8.179922614587585e-05}
{"train_loss": 0.21166253089904785, "global_step": 76527, "epoch": 859, "lr": 8.179877875995108e-05}
{"train_loss": 0.2664336562156677, "global_step": 76528, "epoch": 859, "lr": 8.179833136975133e-05}
{"train_loss": 0.15555262565612793, "global_step": 76529, "epoch": 859, "lr": 8.17978839752767e-05}
{"train_loss": 0.2661619484424591, "global_step": 76530, "epoch": 859, "lr": 8.179743657652722e-05}
{"train_loss": 0.149696484208107, "global_step": 76531, "epoch": 859, "lr": 8.179698917350298e-05}
{"train_loss": 0.1488039791584015, "global_step": 76532, "epoch": 859, "lr": 8.179654176620402e-05}
{"train_loss": 0.267179012298584, "global_step": 76533, "epoch": 859, "lr": 8.179609435463038e-05}
{"train_loss": 0.12038936465978622, "global_step": 76534, "epoch": 859, "lr": 8.179564693878217e-05}
{"train_loss": 0.1726229041814804, "global_step": 76535, "epoch": 859, "lr": 8.179519951865939e-05}
{"train_loss": 0.2378159463405609, "global_step": 76536, "epoch": 859, "lr": 8.179475209426216e-05}
{"train_loss": 0.18081630766391754, "global_step": 76537, "epoch": 859, "lr": 8.179430466559051e-05}
{"train_loss": 0.1825772374868393, "global_step": 76538, "epoch": 859, "lr": 8.17938572326445e-05}
{"train_loss": 0.20982313549585557, "global_step": 76539, "epoch": 859, "lr": 8.179340979542418e-05, "val_loss": 3.9037063121795654}
{"train_loss": 0.204359769821167, "global_step": 76540, "epoch": 860, "lr": 8.179296235392965e-05}
{"train_loss": 0.2043568342924118, "global_step": 76541, "epoch": 860, "lr": 8.179251490816093e-05}
{"train_loss": 0.24168060719966888, "global_step": 76542, "epoch": 860, "lr": 8.179206745811809e-05}
{"train_loss": 0.17318277060985565, "global_step": 76543, "epoch": 860, "lr": 8.179162000380119e-05}
{"train_loss": 0.14488759636878967, "global_step": 76544, "epoch": 860, "lr": 8.17911725452103e-05}
{"train_loss": 0.1602628380060196, "global_step": 76545, "epoch": 860, "lr": 8.179072508234548e-05}
{"train_loss": 0.1328425258398056, "global_step": 76546, "epoch": 860, "lr": 8.179027761520678e-05}
{"train_loss": 0.19017308950424194, "global_step": 76547, "epoch": 860, "lr": 8.178983014379427e-05}
{"train_loss": 0.2236638069152832, "global_step": 76548, "epoch": 860, "lr": 8.178938266810799e-05}
{"train_loss": 0.19981683790683746, "global_step": 76549, "epoch": 860, "lr": 8.178893518814803e-05}
{"train_loss": 0.18612556159496307, "global_step": 76550, "epoch": 860, "lr": 8.178848770391443e-05}
{"train_loss": 0.2020653486251831, "global_step": 76551, "epoch": 860, "lr": 8.178804021540724e-05}
{"train_loss": 0.233071967959404, "global_step": 76552, "epoch": 860, "lr": 8.178759272262656e-05}
{"train_loss": 0.2213917374610901, "global_step": 76553, "epoch": 860, "lr": 8.178714522557242e-05}
{"train_loss": 0.18996286392211914, "global_step": 76554, "epoch": 860, "lr": 8.178669772424488e-05}
{"train_loss": 0.15701936185359955, "global_step": 76555, "epoch": 860, "lr": 8.178625021864401e-05}
{"train_loss": 0.16765613853931427, "global_step": 76556, "epoch": 860, "lr": 8.178580270876986e-05}
{"train_loss": 0.30636143684387207, "global_step": 76557, "epoch": 860, "lr": 8.17853551946225e-05}
{"train_loss": 0.14339180290699005, "global_step": 76558, "epoch": 860, "lr": 8.178490767620198e-05}
{"train_loss": 0.15220434963703156, "global_step": 76559, "epoch": 860, "lr": 8.178446015350838e-05}
{"train_loss": 0.16664375364780426, "global_step": 76560, "epoch": 860, "lr": 8.178401262654174e-05}
{"train_loss": 0.16689029335975647, "global_step": 76561, "epoch": 860, "lr": 8.178356509530212e-05}
{"train_loss": 0.18546339869499207, "global_step": 76562, "epoch": 860, "lr": 8.178311755978961e-05}
{"train_loss": 0.25107961893081665, "global_step": 76563, "epoch": 860, "lr": 8.178267002000421e-05}
{"train_loss": 0.28015920519828796, "global_step": 76564, "epoch": 860, "lr": 8.178222247594604e-05}
{"train_loss": 0.15204885601997375, "global_step": 76565, "epoch": 860, "lr": 8.178177492761514e-05}
{"train_loss": 0.22816318273544312, "global_step": 76566, "epoch": 860, "lr": 8.178132737501156e-05}
{"train_loss": 0.23508432507514954, "global_step": 76567, "epoch": 860, "lr": 8.178087981813539e-05}
{"train_loss": 0.20585790276527405, "global_step": 76568, "epoch": 860, "lr": 8.178043225698664e-05}
{"train_loss": 0.2800666391849518, "global_step": 76569, "epoch": 860, "lr": 8.17799846915654e-05}
{"train_loss": 0.23596179485321045, "global_step": 76570, "epoch": 860, "lr": 8.177953712187174e-05}
{"train_loss": 0.17222093045711517, "global_step": 76571, "epoch": 860, "lr": 8.17790895479057e-05}
{"train_loss": 0.17869116365909576, "global_step": 76572, "epoch": 860, "lr": 8.177864196966736e-05}
{"train_loss": 0.20514540374279022, "global_step": 76573, "epoch": 860, "lr": 8.177819438715676e-05}
{"train_loss": 0.20258796215057373, "global_step": 76574, "epoch": 860, "lr": 8.177774680037397e-05}
{"train_loss": 0.21423734724521637, "global_step": 76575, "epoch": 860, "lr": 8.177729920931905e-05}
{"train_loss": 0.2568519115447998, "global_step": 76576, "epoch": 860, "lr": 8.177685161399206e-05}
{"train_loss": 0.28493741154670715, "global_step": 76577, "epoch": 860, "lr": 8.177640401439306e-05}
{"train_loss": 0.24808864295482635, "global_step": 76578, "epoch": 860, "lr": 8.177595641052212e-05}
{"train_loss": 0.18488532304763794, "global_step": 76579, "epoch": 860, "lr": 8.177550880237928e-05}
{"train_loss": 0.19514711201190948, "global_step": 76580, "epoch": 860, "lr": 8.17750611899646e-05}
{"train_loss": 0.23765164613723755, "global_step": 76581, "epoch": 860, "lr": 8.177461357327816e-05}
{"train_loss": 0.18834786117076874, "global_step": 76582, "epoch": 860, "lr": 8.177416595232001e-05}
{"train_loss": 0.2791646718978882, "global_step": 76583, "epoch": 860, "lr": 8.17737183270902e-05}
{"train_loss": 0.19749896228313446, "global_step": 76584, "epoch": 860, "lr": 8.177327069758882e-05}
{"train_loss": 0.19832049310207367, "global_step": 76585, "epoch": 860, "lr": 8.177282306381589e-05}
{"train_loss": 0.20814646780490875, "global_step": 76586, "epoch": 860, "lr": 8.177237542577151e-05}
{"train_loss": 0.2642955183982849, "global_step": 76587, "epoch": 860, "lr": 8.177192778345572e-05}
{"train_loss": 0.18681509792804718, "global_step": 76588, "epoch": 860, "lr": 8.177148013686858e-05}
{"train_loss": 0.16526691615581512, "global_step": 76589, "epoch": 860, "lr": 8.177103248601015e-05}
{"train_loss": 0.19902248680591583, "global_step": 76590, "epoch": 860, "lr": 8.17705848308805e-05}
{"train_loss": 0.20962944626808167, "global_step": 76591, "epoch": 860, "lr": 8.177013717147966e-05}
{"train_loss": 0.21716313064098358, "global_step": 76592, "epoch": 860, "lr": 8.176968950780772e-05}
{"train_loss": 0.1471843123435974, "global_step": 76593, "epoch": 860, "lr": 8.176924183986474e-05}
{"train_loss": 0.2223384976387024, "global_step": 76594, "epoch": 860, "lr": 8.176879416765077e-05}
{"train_loss": 0.14579351246356964, "global_step": 76595, "epoch": 860, "lr": 8.176834649116587e-05}
{"train_loss": 0.24138754606246948, "global_step": 76596, "epoch": 860, "lr": 8.176789881041011e-05}
{"train_loss": 0.2318309247493744, "global_step": 76597, "epoch": 860, "lr": 8.176745112538353e-05}
{"train_loss": 0.19068937003612518, "global_step": 76598, "epoch": 860, "lr": 8.176700343608623e-05}
{"train_loss": 0.08647533506155014, "global_step": 76599, "epoch": 860, "lr": 8.176655574251822e-05}
{"train_loss": 0.28200507164001465, "global_step": 76600, "epoch": 860, "lr": 8.176610804467958e-05}
{"train_loss": 0.20637927949428558, "global_step": 76601, "epoch": 860, "lr": 8.176566034257038e-05}
{"train_loss": 0.2159128487110138, "global_step": 76602, "epoch": 860, "lr": 8.176521263619068e-05}
{"train_loss": 0.20390750467777252, "global_step": 76603, "epoch": 860, "lr": 8.176476492554054e-05}
{"train_loss": 0.12737926840782166, "global_step": 76604, "epoch": 860, "lr": 8.176431721062e-05}
{"train_loss": 0.20684902369976044, "global_step": 76605, "epoch": 860, "lr": 8.176386949142916e-05}
{"train_loss": 0.21683269739151, "global_step": 76606, "epoch": 860, "lr": 8.176342176796803e-05}
{"train_loss": 0.22404739260673523, "global_step": 76607, "epoch": 860, "lr": 8.17629740402367e-05}
{"train_loss": 0.10757698118686676, "global_step": 76608, "epoch": 860, "lr": 8.176252630823521e-05}
{"train_loss": 0.2562025785446167, "global_step": 76609, "epoch": 860, "lr": 8.176207857196367e-05}
{"train_loss": 0.227872833609581, "global_step": 76610, "epoch": 860, "lr": 8.176163083142208e-05}
{"train_loss": 0.24468454718589783, "global_step": 76611, "epoch": 860, "lr": 8.176118308661053e-05}
{"train_loss": 0.2414158284664154, "global_step": 76612, "epoch": 860, "lr": 8.176073533752908e-05}
{"train_loss": 0.2485179454088211, "global_step": 76613, "epoch": 860, "lr": 8.17602875841778e-05}
{"train_loss": 0.19652681052684784, "global_step": 76614, "epoch": 860, "lr": 8.175983982655671e-05}
{"train_loss": 0.20119169354438782, "global_step": 76615, "epoch": 860, "lr": 8.175939206466592e-05}
{"train_loss": 0.16423368453979492, "global_step": 76616, "epoch": 860, "lr": 8.175894429850546e-05}
{"train_loss": 0.21843117475509644, "global_step": 76617, "epoch": 860, "lr": 8.17584965280754e-05}
{"train_loss": 0.18774107098579407, "global_step": 76618, "epoch": 860, "lr": 8.175804875337578e-05}
{"train_loss": 0.12220150232315063, "global_step": 76619, "epoch": 860, "lr": 8.175760097440668e-05}
{"train_loss": 0.17187811434268951, "global_step": 76620, "epoch": 860, "lr": 8.175715319116816e-05}
{"train_loss": 0.18605968356132507, "global_step": 76621, "epoch": 860, "lr": 8.175670540366028e-05}
{"train_loss": 0.21899676322937012, "global_step": 76622, "epoch": 860, "lr": 8.17562576118831e-05}
{"train_loss": 0.18758966028690338, "global_step": 76623, "epoch": 860, "lr": 8.175580981583668e-05}
{"train_loss": 0.1851717084646225, "global_step": 76624, "epoch": 860, "lr": 8.175536201552108e-05}
{"train_loss": 0.19610214233398438, "global_step": 76625, "epoch": 860, "lr": 8.175491421093637e-05}
{"train_loss": 0.21784821152687073, "global_step": 76626, "epoch": 860, "lr": 8.175446640208258e-05}
{"train_loss": 0.1896645426750183, "global_step": 76627, "epoch": 860, "lr": 8.175401858895979e-05}
{"train_loss": 0.20178233589349168, "global_step": 76628, "epoch": 860, "lr": 8.175357077156807e-05, "val_loss": 4.044458866119385, "train_action_mse_error": 28.517337799072266}
{"train_loss": 0.18989096581935883, "global_step": 76629, "epoch": 861, "lr": 8.175312294990746e-05}
{"train_loss": 0.23781320452690125, "global_step": 76630, "epoch": 861, "lr": 8.175267512397803e-05}
{"train_loss": 0.17826081812381744, "global_step": 76631, "epoch": 861, "lr": 8.175222729377985e-05}
{"train_loss": 0.23945261538028717, "global_step": 76632, "epoch": 861, "lr": 8.175177945931295e-05}
{"train_loss": 0.20164978504180908, "global_step": 76633, "epoch": 861, "lr": 8.175133162057742e-05}
{"train_loss": 0.1829671710729599, "global_step": 76634, "epoch": 861, "lr": 8.175088377757332e-05}
{"train_loss": 0.2164822816848755, "global_step": 76635, "epoch": 861, "lr": 8.17504359303007e-05}
{"train_loss": 0.1690913587808609, "global_step": 76636, "epoch": 861, "lr": 8.17499880787596e-05}
{"train_loss": 0.15270404517650604, "global_step": 76637, "epoch": 861, "lr": 8.174954022295012e-05}
{"train_loss": 0.2097405344247818, "global_step": 76638, "epoch": 861, "lr": 8.17490923628723e-05}
{"train_loss": 0.1822478473186493, "global_step": 76639, "epoch": 861, "lr": 8.17486444985262e-05}
{"train_loss": 0.1576106995344162, "global_step": 76640, "epoch": 861, "lr": 8.174819662991187e-05}
{"train_loss": 0.18967494368553162, "global_step": 76641, "epoch": 861, "lr": 8.17477487570294e-05}
{"train_loss": 0.122569739818573, "global_step": 76642, "epoch": 861, "lr": 8.174730087987883e-05}
{"train_loss": 0.16326120495796204, "global_step": 76643, "epoch": 861, "lr": 8.174685299846021e-05}
{"train_loss": 0.21827936172485352, "global_step": 76644, "epoch": 861, "lr": 8.174640511277362e-05}
{"train_loss": 0.12887249886989594, "global_step": 76645, "epoch": 861, "lr": 8.174595722281911e-05}
{"train_loss": 0.23764516413211823, "global_step": 76646, "epoch": 861, "lr": 8.174550932859675e-05}
{"train_loss": 0.16059435904026031, "global_step": 76647, "epoch": 861, "lr": 8.174506143010659e-05}
{"train_loss": 0.19388353824615479, "global_step": 76648, "epoch": 861, "lr": 8.174461352734868e-05}
{"train_loss": 0.22609026730060577, "global_step": 76649, "epoch": 861, "lr": 8.174416562032312e-05}
{"train_loss": 0.2707970440387726, "global_step": 76650, "epoch": 861, "lr": 8.174371770902992e-05}
{"train_loss": 0.2171112596988678, "global_step": 76651, "epoch": 861, "lr": 8.174326979346917e-05}
{"train_loss": 0.19950836896896362, "global_step": 76652, "epoch": 861, "lr": 8.174282187364093e-05}
{"train_loss": 0.2529524564743042, "global_step": 76653, "epoch": 861, "lr": 8.174237394954524e-05}
{"train_loss": 0.28482726216316223, "global_step": 76654, "epoch": 861, "lr": 8.174192602118221e-05}
{"train_loss": 0.1565549522638321, "global_step": 76655, "epoch": 861, "lr": 8.174147808855182e-05}
{"train_loss": 0.19572187960147858, "global_step": 76656, "epoch": 861, "lr": 8.174103015165419e-05}
{"train_loss": 0.1798885464668274, "global_step": 76657, "epoch": 861, "lr": 8.174058221048939e-05}
{"train_loss": 0.16794025897979736, "global_step": 76658, "epoch": 861, "lr": 8.174013426505743e-05}
{"train_loss": 0.20395809412002563, "global_step": 76659, "epoch": 861, "lr": 8.17396863153584e-05}
{"train_loss": 0.2287234365940094, "global_step": 76660, "epoch": 861, "lr": 8.173923836139236e-05}
{"train_loss": 0.17954330146312714, "global_step": 76661, "epoch": 861, "lr": 8.173879040315937e-05}
{"train_loss": 0.16854465007781982, "global_step": 76662, "epoch": 861, "lr": 8.173834244065948e-05}
{"train_loss": 0.20995257794857025, "global_step": 76663, "epoch": 861, "lr": 8.173789447389276e-05}
{"train_loss": 0.19075246155261993, "global_step": 76664, "epoch": 861, "lr": 8.173744650285927e-05}
{"train_loss": 0.141346275806427, "global_step": 76665, "epoch": 861, "lr": 8.173699852755906e-05}
{"train_loss": 0.16975869238376617, "global_step": 76666, "epoch": 861, "lr": 8.173655054799221e-05}
{"train_loss": 0.20093506574630737, "global_step": 76667, "epoch": 861, "lr": 8.173610256415875e-05}
{"train_loss": 0.1992526650428772, "global_step": 76668, "epoch": 861, "lr": 8.173565457605876e-05}
{"train_loss": 0.22952474653720856, "global_step": 76669, "epoch": 861, "lr": 8.173520658369231e-05}
{"train_loss": 0.218556746840477, "global_step": 76670, "epoch": 861, "lr": 8.173475858705943e-05}
{"train_loss": 0.24234624207019806, "global_step": 76671, "epoch": 861, "lr": 8.173431058616023e-05}
{"train_loss": 0.2627198398113251, "global_step": 76672, "epoch": 861, "lr": 8.173386258099471e-05}
{"train_loss": 0.1396474540233612, "global_step": 76673, "epoch": 861, "lr": 8.173341457156296e-05}
{"train_loss": 0.24431143701076508, "global_step": 76674, "epoch": 861, "lr": 8.173296655786506e-05}
{"train_loss": 0.3344745337963104, "global_step": 76675, "epoch": 861, "lr": 8.173251853990102e-05}
{"train_loss": 0.27200525999069214, "global_step": 76676, "epoch": 861, "lr": 8.173207051767095e-05}
{"train_loss": 0.22429342567920685, "global_step": 76677, "epoch": 861, "lr": 8.173162249117489e-05}
{"train_loss": 0.2214987576007843, "global_step": 76678, "epoch": 861, "lr": 8.17311744604129e-05}
{"train_loss": 0.21888431906700134, "global_step": 76679, "epoch": 861, "lr": 8.173072642538505e-05}
{"train_loss": 0.19423158466815948, "global_step": 76680, "epoch": 861, "lr": 8.173027838609135e-05}
{"train_loss": 0.2103791981935501, "global_step": 76681, "epoch": 861, "lr": 8.172983034253195e-05}
{"train_loss": 0.23651856184005737, "global_step": 76682, "epoch": 861, "lr": 8.172938229470684e-05}
{"train_loss": 0.2711499333381653, "global_step": 76683, "epoch": 861, "lr": 8.172893424261609e-05}
{"train_loss": 0.1992156058549881, "global_step": 76684, "epoch": 861, "lr": 8.172848618625979e-05}
{"train_loss": 0.22665128111839294, "global_step": 76685, "epoch": 861, "lr": 8.172803812563796e-05}
{"train_loss": 0.18029826879501343, "global_step": 76686, "epoch": 861, "lr": 8.17275900607507e-05}
{"train_loss": 0.28750428557395935, "global_step": 76687, "epoch": 861, "lr": 8.172714199159806e-05}
{"train_loss": 0.15380451083183289, "global_step": 76688, "epoch": 861, "lr": 8.172669391818006e-05}
{"train_loss": 0.22915202379226685, "global_step": 76689, "epoch": 861, "lr": 8.172624584049681e-05}
{"train_loss": 0.22316008806228638, "global_step": 76690, "epoch": 861, "lr": 8.172579775854836e-05}
{"train_loss": 0.2577843964099884, "global_step": 76691, "epoch": 861, "lr": 8.172534967233476e-05}
{"train_loss": 0.1951856017112732, "global_step": 76692, "epoch": 861, "lr": 8.172490158185608e-05}
{"train_loss": 0.23603036999702454, "global_step": 76693, "epoch": 861, "lr": 8.172445348711236e-05}
{"train_loss": 0.24727565050125122, "global_step": 76694, "epoch": 861, "lr": 8.172400538810368e-05}
{"train_loss": 0.22228021919727325, "global_step": 76695, "epoch": 861, "lr": 8.17235572848301e-05}
{"train_loss": 0.18242157995700836, "global_step": 76696, "epoch": 861, "lr": 8.172310917729167e-05}
{"train_loss": 0.2516341805458069, "global_step": 76697, "epoch": 861, "lr": 8.172266106548844e-05}
{"train_loss": 0.2282269150018692, "global_step": 76698, "epoch": 861, "lr": 8.172221294942051e-05}
{"train_loss": 0.20805533230304718, "global_step": 76699, "epoch": 861, "lr": 8.172176482908789e-05}
{"train_loss": 0.1573762744665146, "global_step": 76700, "epoch": 861, "lr": 8.172131670449068e-05}
{"train_loss": 0.23675991594791412, "global_step": 76701, "epoch": 861, "lr": 8.172086857562893e-05}
{"train_loss": 0.1625717282295227, "global_step": 76702, "epoch": 861, "lr": 8.17204204425027e-05}
{"train_loss": 0.18026889860630035, "global_step": 76703, "epoch": 861, "lr": 8.171997230511204e-05}
{"train_loss": 0.18335182964801788, "global_step": 76704, "epoch": 861, "lr": 8.1719524163457e-05}
{"train_loss": 0.21625030040740967, "global_step": 76705, "epoch": 861, "lr": 8.171907601753768e-05}
{"train_loss": 0.19283543527126312, "global_step": 76706, "epoch": 861, "lr": 8.171862786735409e-05}
{"train_loss": 0.13437117636203766, "global_step": 76707, "epoch": 861, "lr": 8.171817971290635e-05}
{"train_loss": 0.2122003436088562, "global_step": 76708, "epoch": 861, "lr": 8.171773155419446e-05}
{"train_loss": 0.15808089077472687, "global_step": 76709, "epoch": 861, "lr": 8.171728339121851e-05}
{"train_loss": 0.24648703634738922, "global_step": 76710, "epoch": 861, "lr": 8.171683522397858e-05}
{"train_loss": 0.23713064193725586, "global_step": 76711, "epoch": 861, "lr": 8.171638705247468e-05}
{"train_loss": 0.296409547328949, "global_step": 76712, "epoch": 861, "lr": 8.171593887670692e-05}
{"train_loss": 0.19055122137069702, "global_step": 76713, "epoch": 861, "lr": 8.171549069667532e-05}
{"train_loss": 0.28300240635871887, "global_step": 76714, "epoch": 861, "lr": 8.171504251237998e-05}
{"train_loss": 0.1469177007675171, "global_step": 76715, "epoch": 861, "lr": 8.171459432382092e-05}
{"train_loss": 0.22378255426883698, "global_step": 76716, "epoch": 861, "lr": 8.171414613099824e-05}
{"train_loss": 0.20770215670044503, "global_step": 76717, "epoch": 861, "lr": 8.171369793391195e-05, "val_loss": 3.878535032272339}
{"train_loss": 0.17189854383468628, "global_step": 76718, "epoch": 862, "lr": 8.171324973256217e-05}
{"train_loss": 0.22426536679267883, "global_step": 76719, "epoch": 862, "lr": 8.171280152694892e-05}
{"train_loss": 0.21504072844982147, "global_step": 76720, "epoch": 862, "lr": 8.171235331707226e-05}
{"train_loss": 0.2833257019519806, "global_step": 76721, "epoch": 862, "lr": 8.171190510293228e-05}
{"train_loss": 0.19924090802669525, "global_step": 76722, "epoch": 862, "lr": 8.171145688452901e-05}
{"train_loss": 0.23143422603607178, "global_step": 76723, "epoch": 862, "lr": 8.171100866186252e-05}
{"train_loss": 0.23397551476955414, "global_step": 76724, "epoch": 862, "lr": 8.171056043493287e-05}
{"train_loss": 0.11904539167881012, "global_step": 76725, "epoch": 862, "lr": 8.171011220374012e-05}
{"train_loss": 0.17413492500782013, "global_step": 76726, "epoch": 862, "lr": 8.170966396828434e-05}
{"train_loss": 0.18359646201133728, "global_step": 76727, "epoch": 862, "lr": 8.170921572856557e-05}
{"train_loss": 0.21023935079574585, "global_step": 76728, "epoch": 862, "lr": 8.170876748458389e-05}
{"train_loss": 0.1907404512166977, "global_step": 76729, "epoch": 862, "lr": 8.170831923633935e-05}
{"train_loss": 0.1305096447467804, "global_step": 76730, "epoch": 862, "lr": 8.170787098383201e-05}
{"train_loss": 0.16837038099765778, "global_step": 76731, "epoch": 862, "lr": 8.170742272706194e-05}
{"train_loss": 0.23680926859378815, "global_step": 76732, "epoch": 862, "lr": 8.17069744660292e-05}
{"train_loss": 0.18048982322216034, "global_step": 76733, "epoch": 862, "lr": 8.170652620073382e-05}
{"train_loss": 0.19498324394226074, "global_step": 76734, "epoch": 862, "lr": 8.170607793117591e-05}
{"train_loss": 0.1274450570344925, "global_step": 76735, "epoch": 862, "lr": 8.17056296573555e-05}
{"train_loss": 0.10891743749380112, "global_step": 76736, "epoch": 862, "lr": 8.170518137927265e-05}
{"train_loss": 0.23742255568504333, "global_step": 76737, "epoch": 862, "lr": 8.17047330969274e-05}
{"train_loss": 0.19519497454166412, "global_step": 76738, "epoch": 862, "lr": 8.170428481031988e-05}
{"train_loss": 0.20057499408721924, "global_step": 76739, "epoch": 862, "lr": 8.170383651945008e-05}
{"train_loss": 0.17587290704250336, "global_step": 76740, "epoch": 862, "lr": 8.170338822431809e-05}
{"train_loss": 0.19466833770275116, "global_step": 76741, "epoch": 862, "lr": 8.170293992492397e-05}
{"train_loss": 0.1533721387386322, "global_step": 76742, "epoch": 862, "lr": 8.170249162126777e-05}
{"train_loss": 0.2547839283943176, "global_step": 76743, "epoch": 862, "lr": 8.170204331334957e-05}
{"train_loss": 0.18088489770889282, "global_step": 76744, "epoch": 862, "lr": 8.17015950011694e-05}
{"train_loss": 0.16197378933429718, "global_step": 76745, "epoch": 862, "lr": 8.170114668472734e-05}
{"train_loss": 0.24689677357673645, "global_step": 76746, "epoch": 862, "lr": 8.170069836402344e-05}
{"train_loss": 0.2771933078765869, "global_step": 76747, "epoch": 862, "lr": 8.170025003905777e-05}
{"train_loss": 0.2577393352985382, "global_step": 76748, "epoch": 862, "lr": 8.16998017098304e-05}
{"train_loss": 0.19681841135025024, "global_step": 76749, "epoch": 862, "lr": 8.169935337634136e-05}
{"train_loss": 0.21118071675300598, "global_step": 76750, "epoch": 862, "lr": 8.169890503859074e-05}
{"train_loss": 0.15314067900180817, "global_step": 76751, "epoch": 862, "lr": 8.16984566965786e-05}
{"train_loss": 0.1996750384569168, "global_step": 76752, "epoch": 862, "lr": 8.169800835030497e-05}
{"train_loss": 0.18581759929656982, "global_step": 76753, "epoch": 862, "lr": 8.169755999976994e-05}
{"train_loss": 0.21258848905563354, "global_step": 76754, "epoch": 862, "lr": 8.169711164497355e-05}
{"train_loss": 0.14275118708610535, "global_step": 76755, "epoch": 862, "lr": 8.169666328591588e-05}
{"train_loss": 0.18880906701087952, "global_step": 76756, "epoch": 862, "lr": 8.169621492259698e-05}
{"train_loss": 0.19639143347740173, "global_step": 76757, "epoch": 862, "lr": 8.169576655501691e-05}
{"train_loss": 0.20587487518787384, "global_step": 76758, "epoch": 862, "lr": 8.169531818317573e-05}
{"train_loss": 0.2230575978755951, "global_step": 76759, "epoch": 862, "lr": 8.169486980707349e-05}
{"train_loss": 0.19641456007957458, "global_step": 76760, "epoch": 862, "lr": 8.169442142671028e-05}
{"train_loss": 0.22470247745513916, "global_step": 76761, "epoch": 862, "lr": 8.169397304208612e-05}
{"train_loss": 0.22824035584926605, "global_step": 76762, "epoch": 862, "lr": 8.16935246532011e-05}
{"train_loss": 0.22793139517307281, "global_step": 76763, "epoch": 862, "lr": 8.169307626005528e-05}
{"train_loss": 0.20353877544403076, "global_step": 76764, "epoch": 862, "lr": 8.16926278626487e-05}
{"train_loss": 0.19419750571250916, "global_step": 76765, "epoch": 862, "lr": 8.169217946098144e-05}
{"train_loss": 0.1509135514497757, "global_step": 76766, "epoch": 862, "lr": 8.169173105505354e-05}
{"train_loss": 0.17004907131195068, "global_step": 76767, "epoch": 862, "lr": 8.16912826448651e-05}
{"train_loss": 0.15196970105171204, "global_step": 76768, "epoch": 862, "lr": 8.169083423041612e-05}
{"train_loss": 0.2052556574344635, "global_step": 76769, "epoch": 862, "lr": 8.169038581170672e-05}
{"train_loss": 0.21531321108341217, "global_step": 76770, "epoch": 862, "lr": 8.168993738873693e-05}
{"train_loss": 0.2608824670314789, "global_step": 76771, "epoch": 862, "lr": 8.16894889615068e-05}
{"train_loss": 0.13110558688640594, "global_step": 76772, "epoch": 862, "lr": 8.168904053001642e-05}
{"train_loss": 0.1684068888425827, "global_step": 76773, "epoch": 862, "lr": 8.168859209426582e-05}
{"train_loss": 0.16429811716079712, "global_step": 76774, "epoch": 862, "lr": 8.168814365425508e-05}
{"train_loss": 0.16911305487155914, "global_step": 76775, "epoch": 862, "lr": 8.168769520998427e-05}
{"train_loss": 0.18649572134017944, "global_step": 76776, "epoch": 862, "lr": 8.168724676145341e-05}
{"train_loss": 0.214629665017128, "global_step": 76777, "epoch": 862, "lr": 8.168679830866261e-05}
{"train_loss": 0.2099374681711197, "global_step": 76778, "epoch": 862, "lr": 8.16863498516119e-05}
{"train_loss": 0.13131579756736755, "global_step": 76779, "epoch": 862, "lr": 8.168590139030135e-05}
{"train_loss": 0.17781278491020203, "global_step": 76780, "epoch": 862, "lr": 8.168545292473099e-05}
{"train_loss": 0.14191928505897522, "global_step": 76781, "epoch": 862, "lr": 8.168500445490094e-05}
{"train_loss": 0.22970245778560638, "global_step": 76782, "epoch": 862, "lr": 8.16845559808112e-05}
{"train_loss": 0.23151901364326477, "global_step": 76783, "epoch": 862, "lr": 8.168410750246187e-05}
{"train_loss": 0.13533826172351837, "global_step": 76784, "epoch": 862, "lr": 8.1683659019853e-05}
{"train_loss": 0.2110067903995514, "global_step": 76785, "epoch": 862, "lr": 8.168321053298464e-05}
{"train_loss": 0.14248038828372955, "global_step": 76786, "epoch": 862, "lr": 8.168276204185687e-05}
{"train_loss": 0.1713702529668808, "global_step": 76787, "epoch": 862, "lr": 8.168231354646973e-05}
{"train_loss": 0.16984543204307556, "global_step": 76788, "epoch": 862, "lr": 8.168186504682331e-05}
{"train_loss": 0.2993358373641968, "global_step": 76789, "epoch": 862, "lr": 8.168141654291764e-05}
{"train_loss": 0.1565648317337036, "global_step": 76790, "epoch": 862, "lr": 8.168096803475277e-05}
{"train_loss": 0.23652924597263336, "global_step": 76791, "epoch": 862, "lr": 8.16805195223288e-05}
{"train_loss": 0.2613827586174011, "global_step": 76792, "epoch": 862, "lr": 8.168007100564576e-05}
{"train_loss": 0.20899584889411926, "global_step": 76793, "epoch": 862, "lr": 8.167962248470372e-05}
{"train_loss": 0.23670074343681335, "global_step": 76794, "epoch": 862, "lr": 8.167917395950276e-05}
{"train_loss": 0.17376281321048737, "global_step": 76795, "epoch": 862, "lr": 8.167872543004289e-05}
{"train_loss": 0.25051069259643555, "global_step": 76796, "epoch": 862, "lr": 8.167827689632421e-05}
{"train_loss": 0.23913678526878357, "global_step": 76797, "epoch": 862, "lr": 8.167782835834678e-05}
{"train_loss": 0.17539308965206146, "global_step": 76798, "epoch": 862, "lr": 8.167737981611065e-05}
{"train_loss": 0.18358758091926575, "global_step": 76799, "epoch": 862, "lr": 8.16769312696159e-05}
{"train_loss": 0.24255065619945526, "global_step": 76800, "epoch": 862, "lr": 8.167648271886254e-05}
{"train_loss": 0.24627095460891724, "global_step": 76801, "epoch": 862, "lr": 8.167603416385069e-05}
{"train_loss": 0.2339959740638733, "global_step": 76802, "epoch": 862, "lr": 8.167558560458036e-05}
{"train_loss": 0.2187635749578476, "global_step": 76803, "epoch": 862, "lr": 8.167513704105165e-05}
{"train_loss": 0.1964954286813736, "global_step": 76804, "epoch": 862, "lr": 8.167468847326461e-05}
{"train_loss": 0.20417430996894836, "global_step": 76805, "epoch": 862, "lr": 8.167423990121927e-05}
{"train_loss": 0.19769590046633495, "global_step": 76806, "epoch": 862, "lr": 8.167379132491574e-05, "val_loss": 4.016860485076904}
{"train_loss": 0.20994625985622406, "global_step": 76807, "epoch": 863, "lr": 8.167334274435403e-05}
{"train_loss": 0.2772577702999115, "global_step": 76808, "epoch": 863, "lr": 8.167289415953426e-05}
{"train_loss": 0.290080189704895, "global_step": 76809, "epoch": 863, "lr": 8.167244557045641e-05}
{"train_loss": 0.25275325775146484, "global_step": 76810, "epoch": 863, "lr": 8.167199697712062e-05}
{"train_loss": 0.21244165301322937, "global_step": 76811, "epoch": 863, "lr": 8.167154837952691e-05}
{"train_loss": 0.16813889145851135, "global_step": 76812, "epoch": 863, "lr": 8.167109977767534e-05}
{"train_loss": 0.17694836854934692, "global_step": 76813, "epoch": 863, "lr": 8.167065117156598e-05}
{"train_loss": 0.1672641932964325, "global_step": 76814, "epoch": 863, "lr": 8.167020256119889e-05}
{"train_loss": 0.23393039405345917, "global_step": 76815, "epoch": 863, "lr": 8.166975394657411e-05}
{"train_loss": 0.24480445683002472, "global_step": 76816, "epoch": 863, "lr": 8.166930532769174e-05}
{"train_loss": 0.20573364198207855, "global_step": 76817, "epoch": 863, "lr": 8.16688567045518e-05}
{"train_loss": 0.14924950897693634, "global_step": 76818, "epoch": 863, "lr": 8.166840807715439e-05}
{"train_loss": 0.2071646898984909, "global_step": 76819, "epoch": 863, "lr": 8.166795944549952e-05}
{"train_loss": 0.17653773725032806, "global_step": 76820, "epoch": 863, "lr": 8.16675108095873e-05}
{"train_loss": 0.23504137992858887, "global_step": 76821, "epoch": 863, "lr": 8.166706216941776e-05}
{"train_loss": 0.1866830438375473, "global_step": 76822, "epoch": 863, "lr": 8.166661352499097e-05}
{"train_loss": 0.25038257241249084, "global_step": 76823, "epoch": 863, "lr": 8.166616487630698e-05}
{"train_loss": 0.26272380352020264, "global_step": 76824, "epoch": 863, "lr": 8.166571622336588e-05}
{"train_loss": 0.20141638815402985, "global_step": 76825, "epoch": 863, "lr": 8.16652675661677e-05}
{"train_loss": 0.13011272251605988, "global_step": 76826, "epoch": 863, "lr": 8.166481890471251e-05}
{"train_loss": 0.1623448133468628, "global_step": 76827, "epoch": 863, "lr": 8.166437023900038e-05}
{"train_loss": 0.20794129371643066, "global_step": 76828, "epoch": 863, "lr": 8.166392156903137e-05}
{"train_loss": 0.22008532285690308, "global_step": 76829, "epoch": 863, "lr": 8.166347289480552e-05}
{"train_loss": 0.2451838254928589, "global_step": 76830, "epoch": 863, "lr": 8.16630242163229e-05}
{"train_loss": 0.18075378239154816, "global_step": 76831, "epoch": 863, "lr": 8.166257553358356e-05}
{"train_loss": 0.17310041189193726, "global_step": 76832, "epoch": 863, "lr": 8.166212684658759e-05}
{"train_loss": 0.2449665665626526, "global_step": 76833, "epoch": 863, "lr": 8.166167815533503e-05}
{"train_loss": 0.24410408735275269, "global_step": 76834, "epoch": 863, "lr": 8.166122945982593e-05}
{"train_loss": 0.1762445718050003, "global_step": 76835, "epoch": 863, "lr": 8.166078076006039e-05}
{"train_loss": 0.2420325130224228, "global_step": 76836, "epoch": 863, "lr": 8.166033205603841e-05}
{"train_loss": 0.1441757082939148, "global_step": 76837, "epoch": 863, "lr": 8.16598833477601e-05}
{"train_loss": 0.10371735692024231, "global_step": 76838, "epoch": 863, "lr": 8.165943463522551e-05}
{"train_loss": 0.2982059717178345, "global_step": 76839, "epoch": 863, "lr": 8.16589859184347e-05}
{"train_loss": 0.22992412745952606, "global_step": 76840, "epoch": 863, "lr": 8.165853719738771e-05}
{"train_loss": 0.2518058121204376, "global_step": 76841, "epoch": 863, "lr": 8.165808847208462e-05}
{"train_loss": 0.25723665952682495, "global_step": 76842, "epoch": 863, "lr": 8.16576397425255e-05}
{"train_loss": 0.21783097088336945, "global_step": 76843, "epoch": 863, "lr": 8.165719100871038e-05}
{"train_loss": 0.20243436098098755, "global_step": 76844, "epoch": 863, "lr": 8.165674227063933e-05}
{"train_loss": 0.20267988741397858, "global_step": 76845, "epoch": 863, "lr": 8.165629352831243e-05}
{"train_loss": 0.23333525657653809, "global_step": 76846, "epoch": 863, "lr": 8.165584478172973e-05}
{"train_loss": 0.22581587731838226, "global_step": 76847, "epoch": 863, "lr": 8.165539603089128e-05}
{"train_loss": 0.1788063496351242, "global_step": 76848, "epoch": 863, "lr": 8.165494727579715e-05}
{"train_loss": 0.1980190873146057, "global_step": 76849, "epoch": 863, "lr": 8.16544985164474e-05}
{"train_loss": 0.33538952469825745, "global_step": 76850, "epoch": 863, "lr": 8.165404975284209e-05}
{"train_loss": 0.1815015822649002, "global_step": 76851, "epoch": 863, "lr": 8.165360098498128e-05}
{"train_loss": 0.1824863702058792, "global_step": 76852, "epoch": 863, "lr": 8.165315221286502e-05}
{"train_loss": 0.19571620225906372, "global_step": 76853, "epoch": 863, "lr": 8.165270343649338e-05}
{"train_loss": 0.2199869602918625, "global_step": 76854, "epoch": 863, "lr": 8.165225465586643e-05}
{"train_loss": 0.13347014784812927, "global_step": 76855, "epoch": 863, "lr": 8.165180587098421e-05}
{"train_loss": 0.15392565727233887, "global_step": 76856, "epoch": 863, "lr": 8.165135708184679e-05}
{"train_loss": 0.24597017467021942, "global_step": 76857, "epoch": 863, "lr": 8.165090828845425e-05}
{"train_loss": 0.2529069781303406, "global_step": 76858, "epoch": 863, "lr": 8.16504594908066e-05}
{"train_loss": 0.1894560158252716, "global_step": 76859, "epoch": 863, "lr": 8.165001068890395e-05}
{"train_loss": 0.11925037950277328, "global_step": 76860, "epoch": 863, "lr": 8.164956188274635e-05}
{"train_loss": 0.2411074936389923, "global_step": 76861, "epoch": 863, "lr": 8.164911307233383e-05}
{"train_loss": 0.1746523380279541, "global_step": 76862, "epoch": 863, "lr": 8.164866425766649e-05}
{"train_loss": 0.23196618258953094, "global_step": 76863, "epoch": 863, "lr": 8.164821543874437e-05}
{"train_loss": 0.21328195929527283, "global_step": 76864, "epoch": 863, "lr": 8.164776661556755e-05}
{"train_loss": 0.1698278784751892, "global_step": 76865, "epoch": 863, "lr": 8.164731778813603e-05}
{"train_loss": 0.28465092182159424, "global_step": 76866, "epoch": 863, "lr": 8.164686895644995e-05}
{"train_loss": 0.1941601037979126, "global_step": 76867, "epoch": 863, "lr": 8.164642012050931e-05}
{"train_loss": 0.2404036968946457, "global_step": 76868, "epoch": 863, "lr": 8.164597128031422e-05}
{"train_loss": 0.21535733342170715, "global_step": 76869, "epoch": 863, "lr": 8.164552243586469e-05}
{"train_loss": 0.25245416164398193, "global_step": 76870, "epoch": 863, "lr": 8.164507358716083e-05}
{"train_loss": 0.1458798348903656, "global_step": 76871, "epoch": 863, "lr": 8.164462473420266e-05}
{"train_loss": 0.24080213904380798, "global_step": 76872, "epoch": 863, "lr": 8.164417587699026e-05}
{"train_loss": 0.1593339592218399, "global_step": 76873, "epoch": 863, "lr": 8.164372701552367e-05}
{"train_loss": 0.11699581146240234, "global_step": 76874, "epoch": 863, "lr": 8.1643278149803e-05}
{"train_loss": 0.3346579074859619, "global_step": 76875, "epoch": 863, "lr": 8.164282927982825e-05}
{"train_loss": 0.2722117304801941, "global_step": 76876, "epoch": 863, "lr": 8.164238040559952e-05}
{"train_loss": 0.23887580633163452, "global_step": 76877, "epoch": 863, "lr": 8.164193152711685e-05}
{"train_loss": 0.2505621016025543, "global_step": 76878, "epoch": 863, "lr": 8.164148264438031e-05}
{"train_loss": 0.22729790210723877, "global_step": 76879, "epoch": 863, "lr": 8.164103375738996e-05}
{"train_loss": 0.20307272672653198, "global_step": 76880, "epoch": 863, "lr": 8.164058486614585e-05}
{"train_loss": 0.1789197474718094, "global_step": 76881, "epoch": 863, "lr": 8.164013597064805e-05}
{"train_loss": 0.3003731966018677, "global_step": 76882, "epoch": 863, "lr": 8.163968707089665e-05}
{"train_loss": 0.2150542438030243, "global_step": 76883, "epoch": 863, "lr": 8.163923816689165e-05}
{"train_loss": 0.1535942703485489, "global_step": 76884, "epoch": 863, "lr": 8.163878925863315e-05}
{"train_loss": 0.2709481120109558, "global_step": 76885, "epoch": 863, "lr": 8.163834034612119e-05}
{"train_loss": 0.19637998938560486, "global_step": 76886, "epoch": 863, "lr": 8.163789142935585e-05}
{"train_loss": 0.21217450499534607, "global_step": 76887, "epoch": 863, "lr": 8.163744250833717e-05}
{"train_loss": 0.23881228268146515, "global_step": 76888, "epoch": 863, "lr": 8.163699358306524e-05}
{"train_loss": 0.12450738251209259, "global_step": 76889, "epoch": 863, "lr": 8.16365446535401e-05}
{"train_loss": 0.14066217839717865, "global_step": 76890, "epoch": 863, "lr": 8.163609571976179e-05}
{"train_loss": 0.15779337286949158, "global_step": 76891, "epoch": 863, "lr": 8.163564678173041e-05}
{"train_loss": 0.22047077119350433, "global_step": 76892, "epoch": 863, "lr": 8.1635197839446e-05}
{"train_loss": 0.23670819401741028, "global_step": 76893, "epoch": 863, "lr": 8.163474889290862e-05}
{"train_loss": 0.1743367463350296, "global_step": 76894, "epoch": 863, "lr": 8.163429994211833e-05}
{"train_loss": 0.20985203638170544, "global_step": 76895, "epoch": 863, "lr": 8.163385098707521e-05, "val_loss": 4.137510299682617}
{"train_loss": 0.20890003442764282, "global_step": 76896, "epoch": 864, "lr": 8.16334020277793e-05}
{"train_loss": 0.22515162825584412, "global_step": 76897, "epoch": 864, "lr": 8.163295306423064e-05}
{"train_loss": 0.12745271623134613, "global_step": 76898, "epoch": 864, "lr": 8.163250409642935e-05}
{"train_loss": 0.24768541753292084, "global_step": 76899, "epoch": 864, "lr": 8.163205512437544e-05}
{"train_loss": 0.1675739586353302, "global_step": 76900, "epoch": 864, "lr": 8.163160614806897e-05}
{"train_loss": 0.20162077248096466, "global_step": 76901, "epoch": 864, "lr": 8.163115716751004e-05}
{"train_loss": 0.1837097704410553, "global_step": 76902, "epoch": 864, "lr": 8.163070818269868e-05}
{"train_loss": 0.28142833709716797, "global_step": 76903, "epoch": 864, "lr": 8.163025919363496e-05}
{"train_loss": 0.29191854596138, "global_step": 76904, "epoch": 864, "lr": 8.162981020031892e-05}
{"train_loss": 0.19709542393684387, "global_step": 76905, "epoch": 864, "lr": 8.162936120275067e-05}
{"train_loss": 0.26486754417419434, "global_step": 76906, "epoch": 864, "lr": 8.16289122009302e-05}
{"train_loss": 0.21315772831439972, "global_step": 76907, "epoch": 864, "lr": 8.162846319485763e-05}
{"train_loss": 0.18762977421283722, "global_step": 76908, "epoch": 864, "lr": 8.162801418453298e-05}
{"train_loss": 0.25449511408805847, "global_step": 76909, "epoch": 864, "lr": 8.162756516995635e-05}
{"train_loss": 0.22664372622966766, "global_step": 76910, "epoch": 864, "lr": 8.162711615112777e-05}
{"train_loss": 0.20614390075206757, "global_step": 76911, "epoch": 864, "lr": 8.162666712804732e-05}
{"train_loss": 0.16871017217636108, "global_step": 76912, "epoch": 864, "lr": 8.162621810071504e-05}
{"train_loss": 0.21354004740715027, "global_step": 76913, "epoch": 864, "lr": 8.1625769069131e-05}
{"train_loss": 0.20094671845436096, "global_step": 76914, "epoch": 864, "lr": 8.162532003329527e-05}
{"train_loss": 0.2583785653114319, "global_step": 76915, "epoch": 864, "lr": 8.162487099320789e-05}
{"train_loss": 0.18226510286331177, "global_step": 76916, "epoch": 864, "lr": 8.162442194886893e-05}
{"train_loss": 0.2329072803258896, "global_step": 76917, "epoch": 864, "lr": 8.162397290027846e-05}
{"train_loss": 0.17313891649246216, "global_step": 76918, "epoch": 864, "lr": 8.162352384743654e-05}
{"train_loss": 0.1970105767250061, "global_step": 76919, "epoch": 864, "lr": 8.162307479034322e-05}
{"train_loss": 0.14847856760025024, "global_step": 76920, "epoch": 864, "lr": 8.162262572899855e-05}
{"train_loss": 0.2943515181541443, "global_step": 76921, "epoch": 864, "lr": 8.162217666340259e-05}
{"train_loss": 0.24698849022388458, "global_step": 76922, "epoch": 864, "lr": 8.162172759355546e-05}
{"train_loss": 0.23702165484428406, "global_step": 76923, "epoch": 864, "lr": 8.162127851945713e-05}
{"train_loss": 0.23095031082630157, "global_step": 76924, "epoch": 864, "lr": 8.162082944110772e-05}
{"train_loss": 0.1731489896774292, "global_step": 76925, "epoch": 864, "lr": 8.162038035850729e-05}
{"train_loss": 0.2732226550579071, "global_step": 76926, "epoch": 864, "lr": 8.161993127165587e-05}
{"train_loss": 0.14699500799179077, "global_step": 76927, "epoch": 864, "lr": 8.161948218055353e-05}
{"train_loss": 0.20870673656463623, "global_step": 76928, "epoch": 864, "lr": 8.161903308520033e-05}
{"train_loss": 0.21012485027313232, "global_step": 76929, "epoch": 864, "lr": 8.161858398559636e-05}
{"train_loss": 0.19165940582752228, "global_step": 76930, "epoch": 864, "lr": 8.161813488174166e-05}
{"train_loss": 0.23230095207691193, "global_step": 76931, "epoch": 864, "lr": 8.161768577363627e-05}
{"train_loss": 0.23046474158763885, "global_step": 76932, "epoch": 864, "lr": 8.161723666128027e-05}
{"train_loss": 0.18898643553256989, "global_step": 76933, "epoch": 864, "lr": 8.161678754467372e-05}
{"train_loss": 0.1418178379535675, "global_step": 76934, "epoch": 864, "lr": 8.161633842381668e-05}
{"train_loss": 0.18052732944488525, "global_step": 76935, "epoch": 864, "lr": 8.16158892987092e-05}
{"train_loss": 0.27413469552993774, "global_step": 76936, "epoch": 864, "lr": 8.161544016935135e-05}
{"train_loss": 0.21438288688659668, "global_step": 76937, "epoch": 864, "lr": 8.16149910357432e-05}
{"train_loss": 0.24109627306461334, "global_step": 76938, "epoch": 864, "lr": 8.16145418978848e-05}
{"train_loss": 0.18580199778079987, "global_step": 76939, "epoch": 864, "lr": 8.16140927557762e-05}
{"train_loss": 0.2533800005912781, "global_step": 76940, "epoch": 864, "lr": 8.161364360941748e-05}
{"train_loss": 0.30056536197662354, "global_step": 76941, "epoch": 864, "lr": 8.161319445880868e-05}
{"train_loss": 0.25681260228157043, "global_step": 76942, "epoch": 864, "lr": 8.16127453039499e-05}
{"train_loss": 0.1321306973695755, "global_step": 76943, "epoch": 864, "lr": 8.161229614484114e-05}
{"train_loss": 0.1601588875055313, "global_step": 76944, "epoch": 864, "lr": 8.16118469814825e-05}
{"train_loss": 0.22024349868297577, "global_step": 76945, "epoch": 864, "lr": 8.161139781387402e-05}
{"train_loss": 0.2305787354707718, "global_step": 76946, "epoch": 864, "lr": 8.161094864201578e-05}
{"train_loss": 0.22102519869804382, "global_step": 76947, "epoch": 864, "lr": 8.161049946590784e-05}
{"train_loss": 0.27998486161231995, "global_step": 76948, "epoch": 864, "lr": 8.161005028555026e-05}
{"train_loss": 0.17826607823371887, "global_step": 76949, "epoch": 864, "lr": 8.160960110094309e-05}
{"train_loss": 0.2767249345779419, "global_step": 76950, "epoch": 864, "lr": 8.16091519120864e-05}
{"train_loss": 0.21808983385562897, "global_step": 76951, "epoch": 864, "lr": 8.160870271898022e-05}
{"train_loss": 0.24104847013950348, "global_step": 76952, "epoch": 864, "lr": 8.160825352162465e-05}
{"train_loss": 0.22458012402057648, "global_step": 76953, "epoch": 864, "lr": 8.160780432001975e-05}
{"train_loss": 0.1711486130952835, "global_step": 76954, "epoch": 864, "lr": 8.160735511416555e-05}
{"train_loss": 0.12398757040500641, "global_step": 76955, "epoch": 864, "lr": 8.160690590406212e-05}
{"train_loss": 0.17136609554290771, "global_step": 76956, "epoch": 864, "lr": 8.160645668970955e-05}
{"train_loss": 0.1569831669330597, "global_step": 76957, "epoch": 864, "lr": 8.160600747110786e-05}
{"train_loss": 0.20096330344676971, "global_step": 76958, "epoch": 864, "lr": 8.160555824825713e-05}
{"train_loss": 0.22088342905044556, "global_step": 76959, "epoch": 864, "lr": 8.16051090211574e-05}
{"train_loss": 0.23338153958320618, "global_step": 76960, "epoch": 864, "lr": 8.160465978980878e-05}
{"train_loss": 0.11372032761573792, "global_step": 76961, "epoch": 864, "lr": 8.16042105542113e-05}
{"train_loss": 0.2056782990694046, "global_step": 76962, "epoch": 864, "lr": 8.1603761314365e-05}
{"train_loss": 0.24491339921951294, "global_step": 76963, "epoch": 864, "lr": 8.160331207026996e-05}
{"train_loss": 0.17296378314495087, "global_step": 76964, "epoch": 864, "lr": 8.160286282192623e-05}
{"train_loss": 0.15840189158916473, "global_step": 76965, "epoch": 864, "lr": 8.160241356933391e-05}
{"train_loss": 0.2366151511669159, "global_step": 76966, "epoch": 864, "lr": 8.160196431249301e-05}
{"train_loss": 0.24866971373558044, "global_step": 76967, "epoch": 864, "lr": 8.16015150514036e-05}
{"train_loss": 0.18082019686698914, "global_step": 76968, "epoch": 864, "lr": 8.160106578606578e-05}
{"train_loss": 0.21331314742565155, "global_step": 76969, "epoch": 864, "lr": 8.160061651647958e-05}
{"train_loss": 0.19319045543670654, "global_step": 76970, "epoch": 864, "lr": 8.160016724264506e-05}
{"train_loss": 0.2445117086172104, "global_step": 76971, "epoch": 864, "lr": 8.159971796456228e-05}
{"train_loss": 0.20257657766342163, "global_step": 76972, "epoch": 864, "lr": 8.159926868223129e-05}
{"train_loss": 0.2538454830646515, "global_step": 76973, "epoch": 864, "lr": 8.159881939565217e-05}
{"train_loss": 0.18077833950519562, "global_step": 76974, "epoch": 864, "lr": 8.159837010482496e-05}
{"train_loss": 0.26034823060035706, "global_step": 76975, "epoch": 864, "lr": 8.159792080974975e-05}
{"train_loss": 0.21538695693016052, "global_step": 76976, "epoch": 864, "lr": 8.15974715104266e-05}
{"train_loss": 0.24800361692905426, "global_step": 76977, "epoch": 864, "lr": 8.159702220685553e-05}
{"train_loss": 0.22226671874523163, "global_step": 76978, "epoch": 864, "lr": 8.159657289903664e-05}
{"train_loss": 0.21738798916339874, "global_step": 76979, "epoch": 864, "lr": 8.159612358696996e-05}
{"train_loss": 0.24122005701065063, "global_step": 76980, "epoch": 864, "lr": 8.159567427065559e-05}
{"train_loss": 0.3256855905056, "global_step": 76981, "epoch": 864, "lr": 8.159522495009355e-05}
{"train_loss": 0.17113935947418213, "global_step": 76982, "epoch": 864, "lr": 8.159477562528392e-05}
{"train_loss": 0.13276873528957367, "global_step": 76983, "epoch": 864, "lr": 8.159432629622675e-05}
{"train_loss": 0.21198904564541377, "global_step": 76984, "epoch": 864, "lr": 8.159387696292211e-05, "val_loss": 3.7353157997131348}
{"train_loss": 0.23808816075325012, "global_step": 76985, "epoch": 865, "lr": 8.159342762537008e-05}
{"train_loss": 0.1444007307291031, "global_step": 76986, "epoch": 865, "lr": 8.159297828357068e-05}
{"train_loss": 0.14308319985866547, "global_step": 76987, "epoch": 865, "lr": 8.159252893752398e-05}
{"train_loss": 0.18622490763664246, "global_step": 76988, "epoch": 865, "lr": 8.159207958723008e-05}
{"train_loss": 0.17435221374034882, "global_step": 76989, "epoch": 865, "lr": 8.159163023268899e-05}
{"train_loss": 0.17215535044670105, "global_step": 76990, "epoch": 865, "lr": 8.159118087390079e-05}
{"train_loss": 0.23478157818317413, "global_step": 76991, "epoch": 865, "lr": 8.159073151086554e-05}
{"train_loss": 0.1562393307685852, "global_step": 76992, "epoch": 865, "lr": 8.159028214358331e-05}
{"train_loss": 0.1975427120923996, "global_step": 76993, "epoch": 865, "lr": 8.158983277205414e-05}
{"train_loss": 0.2540767192840576, "global_step": 76994, "epoch": 865, "lr": 8.158938339627811e-05}
{"train_loss": 0.202439546585083, "global_step": 76995, "epoch": 865, "lr": 8.158893401625527e-05}
{"train_loss": 0.2377164363861084, "global_step": 76996, "epoch": 865, "lr": 8.158848463198567e-05}
{"train_loss": 0.22073927521705627, "global_step": 76997, "epoch": 865, "lr": 8.15880352434694e-05}
{"train_loss": 0.22159267961978912, "global_step": 76998, "epoch": 865, "lr": 8.15875858507065e-05}
{"train_loss": 0.18925705552101135, "global_step": 76999, "epoch": 865, "lr": 8.158713645369705e-05}
{"train_loss": 0.12623324990272522, "global_step": 77000, "epoch": 865, "lr": 8.158668705244106e-05}
{"train_loss": 0.21369850635528564, "global_step": 77001, "epoch": 865, "lr": 8.158623764693864e-05}
{"train_loss": 0.16762441396713257, "global_step": 77002, "epoch": 865, "lr": 8.158578823718985e-05}
{"train_loss": 0.21016770601272583, "global_step": 77003, "epoch": 865, "lr": 8.158533882319472e-05}
{"train_loss": 0.2149006575345993, "global_step": 77004, "epoch": 865, "lr": 8.158488940495333e-05}
{"train_loss": 0.23121698200702667, "global_step": 77005, "epoch": 865, "lr": 8.158443998246573e-05}
{"train_loss": 0.1699233502149582, "global_step": 77006, "epoch": 865, "lr": 8.1583990555732e-05}
{"train_loss": 0.1393696516752243, "global_step": 77007, "epoch": 865, "lr": 8.158354112475218e-05}
{"train_loss": 0.31618732213974, "global_step": 77008, "epoch": 865, "lr": 8.158309168952634e-05}
{"train_loss": 0.19159626960754395, "global_step": 77009, "epoch": 865, "lr": 8.158264225005454e-05}
{"train_loss": 0.09324750304222107, "global_step": 77010, "epoch": 865, "lr": 8.158219280633683e-05}
{"train_loss": 0.25054043531417847, "global_step": 77011, "epoch": 865, "lr": 8.15817433583733e-05}
{"train_loss": 0.14007341861724854, "global_step": 77012, "epoch": 865, "lr": 8.158129390616396e-05}
{"train_loss": 0.23426195979118347, "global_step": 77013, "epoch": 865, "lr": 8.158084444970893e-05}
{"train_loss": 0.10839210450649261, "global_step": 77014, "epoch": 865, "lr": 8.158039498900822e-05}
{"train_loss": 0.17528584599494934, "global_step": 77015, "epoch": 865, "lr": 8.157994552406192e-05}
{"train_loss": 0.22377875447273254, "global_step": 77016, "epoch": 865, "lr": 8.157949605487007e-05}
{"train_loss": 0.20263193547725677, "global_step": 77017, "epoch": 865, "lr": 8.157904658143275e-05}
{"train_loss": 0.21269142627716064, "global_step": 77018, "epoch": 865, "lr": 8.157859710375e-05}
{"train_loss": 0.19525118172168732, "global_step": 77019, "epoch": 865, "lr": 8.157814762182191e-05}
{"train_loss": 0.16396164894104004, "global_step": 77020, "epoch": 865, "lr": 8.157769813564852e-05}
{"train_loss": 0.19639556109905243, "global_step": 77021, "epoch": 865, "lr": 8.157724864522989e-05}
{"train_loss": 0.2010548710823059, "global_step": 77022, "epoch": 865, "lr": 8.157679915056608e-05}
{"train_loss": 0.22060026228427887, "global_step": 77023, "epoch": 865, "lr": 8.157634965165716e-05}
{"train_loss": 0.23197424411773682, "global_step": 77024, "epoch": 865, "lr": 8.157590014850318e-05}
{"train_loss": 0.20585750043392181, "global_step": 77025, "epoch": 865, "lr": 8.157545064110422e-05}
{"train_loss": 0.255141943693161, "global_step": 77026, "epoch": 865, "lr": 8.15750011294603e-05}
{"train_loss": 0.18364813923835754, "global_step": 77027, "epoch": 865, "lr": 8.157455161357153e-05}
{"train_loss": 0.24872905015945435, "global_step": 77028, "epoch": 865, "lr": 8.157410209343793e-05}
{"train_loss": 0.24907127022743225, "global_step": 77029, "epoch": 865, "lr": 8.157365256905958e-05}
{"train_loss": 0.213664710521698, "global_step": 77030, "epoch": 865, "lr": 8.157320304043654e-05}
{"train_loss": 0.1623033732175827, "global_step": 77031, "epoch": 865, "lr": 8.157275350756886e-05}
{"train_loss": 0.17918190360069275, "global_step": 77032, "epoch": 865, "lr": 8.157230397045663e-05}
{"train_loss": 0.14308488368988037, "global_step": 77033, "epoch": 865, "lr": 8.157185442909987e-05}
{"train_loss": 0.23045231401920319, "global_step": 77034, "epoch": 865, "lr": 8.157140488349866e-05}
{"train_loss": 0.14498385787010193, "global_step": 77035, "epoch": 865, "lr": 8.157095533365307e-05}
{"train_loss": 0.2534542679786682, "global_step": 77036, "epoch": 865, "lr": 8.157050577956314e-05}
{"train_loss": 0.13580484688282013, "global_step": 77037, "epoch": 865, "lr": 8.157005622122893e-05}
{"train_loss": 0.20948617160320282, "global_step": 77038, "epoch": 865, "lr": 8.156960665865052e-05}
{"train_loss": 0.25335636734962463, "global_step": 77039, "epoch": 865, "lr": 8.156915709182798e-05}
{"train_loss": 0.30229848623275757, "global_step": 77040, "epoch": 865, "lr": 8.156870752076133e-05}
{"train_loss": 0.31521663069725037, "global_step": 77041, "epoch": 865, "lr": 8.156825794545066e-05}
{"train_loss": 0.2668363153934479, "global_step": 77042, "epoch": 865, "lr": 8.156780836589602e-05}
{"train_loss": 0.1978497952222824, "global_step": 77043, "epoch": 865, "lr": 8.156735878209747e-05}
{"train_loss": 0.18683861196041107, "global_step": 77044, "epoch": 865, "lr": 8.156690919405507e-05}
{"train_loss": 0.1636626422405243, "global_step": 77045, "epoch": 865, "lr": 8.15664596017689e-05}
{"train_loss": 0.22133441269397736, "global_step": 77046, "epoch": 865, "lr": 8.156601000523899e-05}
{"train_loss": 0.23362374305725098, "global_step": 77047, "epoch": 865, "lr": 8.156556040446542e-05}
{"train_loss": 0.15632405877113342, "global_step": 77048, "epoch": 865, "lr": 8.156511079944824e-05}
{"train_loss": 0.20649662613868713, "global_step": 77049, "epoch": 865, "lr": 8.156466119018753e-05}
{"train_loss": 0.22686322033405304, "global_step": 77050, "epoch": 865, "lr": 8.156421157668332e-05}
{"train_loss": 0.23644772171974182, "global_step": 77051, "epoch": 865, "lr": 8.156376195893569e-05}
{"train_loss": 0.2017747014760971, "global_step": 77052, "epoch": 865, "lr": 8.15633123369447e-05}
{"train_loss": 0.24376007914543152, "global_step": 77053, "epoch": 865, "lr": 8.156286271071041e-05}
{"train_loss": 0.22685092687606812, "global_step": 77054, "epoch": 865, "lr": 8.156241308023286e-05}
{"train_loss": 0.16409793496131897, "global_step": 77055, "epoch": 865, "lr": 8.156196344551215e-05}
{"train_loss": 0.20694224536418915, "global_step": 77056, "epoch": 865, "lr": 8.156151380654832e-05}
{"train_loss": 0.15735815465450287, "global_step": 77057, "epoch": 865, "lr": 8.156106416334142e-05}
{"train_loss": 0.2407299280166626, "global_step": 77058, "epoch": 865, "lr": 8.156061451589153e-05}
{"train_loss": 0.214711993932724, "global_step": 77059, "epoch": 865, "lr": 8.156016486419868e-05}
{"train_loss": 0.18255601823329926, "global_step": 77060, "epoch": 865, "lr": 8.155971520826296e-05}
{"train_loss": 0.21760369837284088, "global_step": 77061, "epoch": 865, "lr": 8.155926554808441e-05}
{"train_loss": 0.22403331100940704, "global_step": 77062, "epoch": 865, "lr": 8.155881588366311e-05}
{"train_loss": 0.1889885812997818, "global_step": 77063, "epoch": 865, "lr": 8.155836621499913e-05}
{"train_loss": 0.1717824935913086, "global_step": 77064, "epoch": 865, "lr": 8.15579165420925e-05}
{"train_loss": 0.22289271652698517, "global_step": 77065, "epoch": 865, "lr": 8.155746686494329e-05}
{"train_loss": 0.14876310527324677, "global_step": 77066, "epoch": 865, "lr": 8.155701718355157e-05}
{"train_loss": 0.1612013876438141, "global_step": 77067, "epoch": 865, "lr": 8.155656749791738e-05}
{"train_loss": 0.21214579045772552, "global_step": 77068, "epoch": 865, "lr": 8.15561178080408e-05}
{"train_loss": 0.1998312622308731, "global_step": 77069, "epoch": 865, "lr": 8.155566811392188e-05}
{"train_loss": 0.11949575692415237, "global_step": 77070, "epoch": 865, "lr": 8.155521841556069e-05}
{"train_loss": 0.2366216480731964, "global_step": 77071, "epoch": 865, "lr": 8.15547687129573e-05}
{"train_loss": 0.16862066090106964, "global_step": 77072, "epoch": 865, "lr": 8.155431900611174e-05}
{"train_loss": 0.20143000455049986, "global_step": 77073, "epoch": 865, "lr": 8.155386929502409e-05, "val_loss": 3.9557018280029297, "train_action_mse_error": 22.571325302124023}
{"train_loss": 0.20377296209335327, "global_step": 77074, "epoch": 866, "lr": 8.15534195796944e-05}
{"train_loss": 0.3319382071495056, "global_step": 77075, "epoch": 866, "lr": 8.155296986012274e-05}
{"train_loss": 0.19800050556659698, "global_step": 77076, "epoch": 866, "lr": 8.155252013630917e-05}
{"train_loss": 0.2625789940357208, "global_step": 77077, "epoch": 866, "lr": 8.155207040825374e-05}
{"train_loss": 0.21070978045463562, "global_step": 77078, "epoch": 866, "lr": 8.155162067595655e-05}
{"train_loss": 0.21749426424503326, "global_step": 77079, "epoch": 866, "lr": 8.15511709394176e-05}
{"train_loss": 0.20213191211223602, "global_step": 77080, "epoch": 866, "lr": 8.155072119863699e-05}
{"train_loss": 0.20032751560211182, "global_step": 77081, "epoch": 866, "lr": 8.155027145361476e-05}
{"train_loss": 0.17110486328601837, "global_step": 77082, "epoch": 866, "lr": 8.154982170435098e-05}
{"train_loss": 0.17818263173103333, "global_step": 77083, "epoch": 866, "lr": 8.154937195084572e-05}
{"train_loss": 0.09467165917158127, "global_step": 77084, "epoch": 866, "lr": 8.154892219309904e-05}
{"train_loss": 0.1716088354587555, "global_step": 77085, "epoch": 866, "lr": 8.154847243111097e-05}
{"train_loss": 0.1967690885066986, "global_step": 77086, "epoch": 866, "lr": 8.154802266488161e-05}
{"train_loss": 0.18016317486763, "global_step": 77087, "epoch": 866, "lr": 8.1547572894411e-05}
{"train_loss": 0.13245737552642822, "global_step": 77088, "epoch": 866, "lr": 8.154712311969919e-05}
{"train_loss": 0.1809227168560028, "global_step": 77089, "epoch": 866, "lr": 8.154667334074627e-05}
{"train_loss": 0.2648754417896271, "global_step": 77090, "epoch": 866, "lr": 8.154622355755228e-05}
{"train_loss": 0.1398477703332901, "global_step": 77091, "epoch": 866, "lr": 8.154577377011727e-05}
{"train_loss": 0.23445528745651245, "global_step": 77092, "epoch": 866, "lr": 8.154532397844133e-05}
{"train_loss": 0.11954692006111145, "global_step": 77093, "epoch": 866, "lr": 8.15448741825245e-05}
{"train_loss": 0.24817906320095062, "global_step": 77094, "epoch": 866, "lr": 8.154442438236686e-05}
{"train_loss": 0.24072574079036713, "global_step": 77095, "epoch": 866, "lr": 8.154397457796845e-05}
{"train_loss": 0.11571426689624786, "global_step": 77096, "epoch": 866, "lr": 8.154352476932932e-05}
{"train_loss": 0.24570690095424652, "global_step": 77097, "epoch": 866, "lr": 8.154307495644958e-05}
{"train_loss": 0.14650849997997284, "global_step": 77098, "epoch": 866, "lr": 8.154262513932922e-05}
{"train_loss": 0.13653214275836945, "global_step": 77099, "epoch": 866, "lr": 8.154217531796837e-05}
{"train_loss": 0.1895531266927719, "global_step": 77100, "epoch": 866, "lr": 8.154172549236704e-05}
{"train_loss": 0.13680534064769745, "global_step": 77101, "epoch": 866, "lr": 8.154127566252532e-05}
{"train_loss": 0.17577409744262695, "global_step": 77102, "epoch": 866, "lr": 8.154082582844325e-05}
{"train_loss": 0.1694653332233429, "global_step": 77103, "epoch": 866, "lr": 8.154037599012092e-05}
{"train_loss": 0.21450045704841614, "global_step": 77104, "epoch": 866, "lr": 8.153992614755835e-05}
{"train_loss": 0.1700124442577362, "global_step": 77105, "epoch": 866, "lr": 8.153947630075562e-05}
{"train_loss": 0.2669808268547058, "global_step": 77106, "epoch": 866, "lr": 8.15390264497128e-05}
{"train_loss": 0.18449974060058594, "global_step": 77107, "epoch": 866, "lr": 8.153857659442995e-05}
{"train_loss": 0.26208052039146423, "global_step": 77108, "epoch": 866, "lr": 8.153812673490712e-05}
{"train_loss": 0.18024666607379913, "global_step": 77109, "epoch": 866, "lr": 8.153767687114435e-05}
{"train_loss": 0.16846804320812225, "global_step": 77110, "epoch": 866, "lr": 8.153722700314175e-05}
{"train_loss": 0.16244009137153625, "global_step": 77111, "epoch": 866, "lr": 8.153677713089936e-05}
{"train_loss": 0.17989371716976166, "global_step": 77112, "epoch": 866, "lr": 8.153632725441721e-05}
{"train_loss": 0.1571630984544754, "global_step": 77113, "epoch": 866, "lr": 8.15358773736954e-05}
{"train_loss": 0.2417692095041275, "global_step": 77114, "epoch": 866, "lr": 8.153542748873397e-05}
{"train_loss": 0.164811909198761, "global_step": 77115, "epoch": 866, "lr": 8.153497759953299e-05}
{"train_loss": 0.22666865587234497, "global_step": 77116, "epoch": 866, "lr": 8.153452770609251e-05}
{"train_loss": 0.1835804432630539, "global_step": 77117, "epoch": 866, "lr": 8.15340778084126e-05}
{"train_loss": 0.23646770417690277, "global_step": 77118, "epoch": 866, "lr": 8.153362790649333e-05}
{"train_loss": 0.13419294357299805, "global_step": 77119, "epoch": 866, "lr": 8.153317800033473e-05}
{"train_loss": 0.2215002030134201, "global_step": 77120, "epoch": 866, "lr": 8.153272808993688e-05}
{"train_loss": 0.1434251368045807, "global_step": 77121, "epoch": 866, "lr": 8.153227817529984e-05}
{"train_loss": 0.14437660574913025, "global_step": 77122, "epoch": 866, "lr": 8.15318282564237e-05}
{"train_loss": 0.272898405790329, "global_step": 77123, "epoch": 866, "lr": 8.153137833330846e-05}
{"train_loss": 0.20069891214370728, "global_step": 77124, "epoch": 866, "lr": 8.153092840595422e-05}
{"train_loss": 0.25382280349731445, "global_step": 77125, "epoch": 866, "lr": 8.153047847436102e-05}
{"train_loss": 0.12965407967567444, "global_step": 77126, "epoch": 866, "lr": 8.153002853852894e-05}
{"train_loss": 0.11668067425489426, "global_step": 77127, "epoch": 866, "lr": 8.152957859845803e-05}
{"train_loss": 0.1715743988752365, "global_step": 77128, "epoch": 866, "lr": 8.152912865414834e-05}
{"train_loss": 0.1925356388092041, "global_step": 77129, "epoch": 866, "lr": 8.152867870559996e-05}
{"train_loss": 0.11145366728305817, "global_step": 77130, "epoch": 866, "lr": 8.152822875281293e-05}
{"train_loss": 0.1909167319536209, "global_step": 77131, "epoch": 866, "lr": 8.152777879578731e-05}
{"train_loss": 0.19745229184627533, "global_step": 77132, "epoch": 866, "lr": 8.152732883452316e-05}
{"train_loss": 0.1448049247264862, "global_step": 77133, "epoch": 866, "lr": 8.152687886902055e-05}
{"train_loss": 0.24453558027744293, "global_step": 77134, "epoch": 866, "lr": 8.152642889927955e-05}
{"train_loss": 0.21786776185035706, "global_step": 77135, "epoch": 866, "lr": 8.152597892530019e-05}
{"train_loss": 0.11569399386644363, "global_step": 77136, "epoch": 866, "lr": 8.152552894708253e-05}
{"train_loss": 0.2642834484577179, "global_step": 77137, "epoch": 866, "lr": 8.152507896462668e-05}
{"train_loss": 0.25812914967536926, "global_step": 77138, "epoch": 866, "lr": 8.152462897793266e-05}
{"train_loss": 0.14088821411132812, "global_step": 77139, "epoch": 866, "lr": 8.152417898700053e-05}
{"train_loss": 0.21050456166267395, "global_step": 77140, "epoch": 866, "lr": 8.152372899183036e-05}
{"train_loss": 0.1786501109600067, "global_step": 77141, "epoch": 866, "lr": 8.152327899242221e-05}
{"train_loss": 0.22084835171699524, "global_step": 77142, "epoch": 866, "lr": 8.152282898877614e-05}
{"train_loss": 0.21836748719215393, "global_step": 77143, "epoch": 866, "lr": 8.152237898089221e-05}
{"train_loss": 0.2161957323551178, "global_step": 77144, "epoch": 866, "lr": 8.152192896877047e-05}
{"train_loss": 0.16865213215351105, "global_step": 77145, "epoch": 866, "lr": 8.1521478952411e-05}
{"train_loss": 0.22007372975349426, "global_step": 77146, "epoch": 866, "lr": 8.152102893181385e-05}
{"train_loss": 0.1453126221895218, "global_step": 77147, "epoch": 866, "lr": 8.152057890697908e-05}
{"train_loss": 0.2443658709526062, "global_step": 77148, "epoch": 866, "lr": 8.152012887790675e-05}
{"train_loss": 0.21516625583171844, "global_step": 77149, "epoch": 866, "lr": 8.151967884459694e-05}
{"train_loss": 0.27551838755607605, "global_step": 77150, "epoch": 866, "lr": 8.151922880704967e-05}
{"train_loss": 0.11980130523443222, "global_step": 77151, "epoch": 866, "lr": 8.151877876526503e-05}
{"train_loss": 0.24919229745864868, "global_step": 77152, "epoch": 866, "lr": 8.15183287192431e-05}
{"train_loss": 0.18319302797317505, "global_step": 77153, "epoch": 866, "lr": 8.151787866898388e-05}
{"train_loss": 0.17589102685451508, "global_step": 77154, "epoch": 866, "lr": 8.151742861448747e-05}
{"train_loss": 0.21482156217098236, "global_step": 77155, "epoch": 866, "lr": 8.151697855575395e-05}
{"train_loss": 0.23194698989391327, "global_step": 77156, "epoch": 866, "lr": 8.151652849278335e-05}
{"train_loss": 0.2197355479001999, "global_step": 77157, "epoch": 866, "lr": 8.151607842557571e-05}
{"train_loss": 0.17677021026611328, "global_step": 77158, "epoch": 866, "lr": 8.151562835413115e-05}
{"train_loss": 0.16101977229118347, "global_step": 77159, "epoch": 866, "lr": 8.151517827844968e-05}
{"train_loss": 0.15090592205524445, "global_step": 77160, "epoch": 866, "lr": 8.151472819853139e-05}
{"train_loss": 0.11256802827119827, "global_step": 77161, "epoch": 866, "lr": 8.151427811437632e-05}
{"train_loss": 0.19194004852115437, "global_step": 77162, "epoch": 866, "lr": 8.151382802598453e-05, "val_loss": 3.9898717403411865}
{"train_loss": 0.19759178161621094, "global_step": 77163, "epoch": 867, "lr": 8.15133779333561e-05}
{"train_loss": 0.13677124679088593, "global_step": 77164, "epoch": 867, "lr": 8.151292783649108e-05}
{"train_loss": 0.2488418072462082, "global_step": 77165, "epoch": 867, "lr": 8.151247773538953e-05}
{"train_loss": 0.2212066501379013, "global_step": 77166, "epoch": 867, "lr": 8.151202763005151e-05}
{"train_loss": 0.11966077983379364, "global_step": 77167, "epoch": 867, "lr": 8.151157752047709e-05}
{"train_loss": 0.22511076927185059, "global_step": 77168, "epoch": 867, "lr": 8.151112740666633e-05}
{"train_loss": 0.2509528696537018, "global_step": 77169, "epoch": 867, "lr": 8.151067728861927e-05}
{"train_loss": 0.13891731202602386, "global_step": 77170, "epoch": 867, "lr": 8.151022716633598e-05}
{"train_loss": 0.14497636258602142, "global_step": 77171, "epoch": 867, "lr": 8.150977703981653e-05}
{"train_loss": 0.18250788748264313, "global_step": 77172, "epoch": 867, "lr": 8.150932690906097e-05}
{"train_loss": 0.23849956691265106, "global_step": 77173, "epoch": 867, "lr": 8.150887677406937e-05}
{"train_loss": 0.2509806156158447, "global_step": 77174, "epoch": 867, "lr": 8.150842663484177e-05}
{"train_loss": 0.17855125665664673, "global_step": 77175, "epoch": 867, "lr": 8.150797649137827e-05}
{"train_loss": 0.27020034193992615, "global_step": 77176, "epoch": 867, "lr": 8.150752634367889e-05}
{"train_loss": 0.18621596693992615, "global_step": 77177, "epoch": 867, "lr": 8.150707619174371e-05}
{"train_loss": 0.17072990536689758, "global_step": 77178, "epoch": 867, "lr": 8.15066260355728e-05}
{"train_loss": 0.19133716821670532, "global_step": 77179, "epoch": 867, "lr": 8.15061758751662e-05}
{"train_loss": 0.19841310381889343, "global_step": 77180, "epoch": 867, "lr": 8.150572571052397e-05}
{"train_loss": 0.2362063080072403, "global_step": 77181, "epoch": 867, "lr": 8.15052755416462e-05}
{"train_loss": 0.17094306647777557, "global_step": 77182, "epoch": 867, "lr": 8.15048253685329e-05}
{"train_loss": 0.28673994541168213, "global_step": 77183, "epoch": 867, "lr": 8.150437519118418e-05}
{"train_loss": 0.2584238350391388, "global_step": 77184, "epoch": 867, "lr": 8.150392500960007e-05}
{"train_loss": 0.17720460891723633, "global_step": 77185, "epoch": 867, "lr": 8.150347482378066e-05}
{"train_loss": 0.18097732961177826, "global_step": 77186, "epoch": 867, "lr": 8.150302463372596e-05}
{"train_loss": 0.28875821828842163, "global_step": 77187, "epoch": 867, "lr": 8.15025744394361e-05}
{"train_loss": 0.20141106843948364, "global_step": 77188, "epoch": 867, "lr": 8.150212424091108e-05}
{"train_loss": 0.21414992213249207, "global_step": 77189, "epoch": 867, "lr": 8.150167403815098e-05}
{"train_loss": 0.29552918672561646, "global_step": 77190, "epoch": 867, "lr": 8.150122383115589e-05}
{"train_loss": 0.14894838631153107, "global_step": 77191, "epoch": 867, "lr": 8.150077361992583e-05}
{"train_loss": 0.19637128710746765, "global_step": 77192, "epoch": 867, "lr": 8.150032340446086e-05}
{"train_loss": 0.21788394451141357, "global_step": 77193, "epoch": 867, "lr": 8.149987318476107e-05}
{"train_loss": 0.14447109401226044, "global_step": 77194, "epoch": 867, "lr": 8.14994229608265e-05}
{"train_loss": 0.17539909482002258, "global_step": 77195, "epoch": 867, "lr": 8.149897273265722e-05}
{"train_loss": 0.17497338354587555, "global_step": 77196, "epoch": 867, "lr": 8.14985225002533e-05}
{"train_loss": 0.17256085574626923, "global_step": 77197, "epoch": 867, "lr": 8.149807226361477e-05}
{"train_loss": 0.1694408506155014, "global_step": 77198, "epoch": 867, "lr": 8.149762202274171e-05}
{"train_loss": 0.24198412895202637, "global_step": 77199, "epoch": 867, "lr": 8.149717177763419e-05}
{"train_loss": 0.2667909264564514, "global_step": 77200, "epoch": 867, "lr": 8.149672152829224e-05}
{"train_loss": 0.19011753797531128, "global_step": 77201, "epoch": 867, "lr": 8.149627127471595e-05}
{"train_loss": 0.16997087001800537, "global_step": 77202, "epoch": 867, "lr": 8.149582101690537e-05}
{"train_loss": 0.2613731026649475, "global_step": 77203, "epoch": 867, "lr": 8.149537075486056e-05}
{"train_loss": 0.17899958789348602, "global_step": 77204, "epoch": 867, "lr": 8.149492048858159e-05}
{"train_loss": 0.2288469821214676, "global_step": 77205, "epoch": 867, "lr": 8.149447021806849e-05}
{"train_loss": 0.20572113990783691, "global_step": 77206, "epoch": 867, "lr": 8.149401994332137e-05}
{"train_loss": 0.15059438347816467, "global_step": 77207, "epoch": 867, "lr": 8.149356966434025e-05}
{"train_loss": 0.10580245405435562, "global_step": 77208, "epoch": 867, "lr": 8.14931193811252e-05}
{"train_loss": 0.2664696276187897, "global_step": 77209, "epoch": 867, "lr": 8.149266909367628e-05}
{"train_loss": 0.14600111544132233, "global_step": 77210, "epoch": 867, "lr": 8.149221880199357e-05}
{"train_loss": 0.16420914232730865, "global_step": 77211, "epoch": 867, "lr": 8.14917685060771e-05}
{"train_loss": 0.20721258223056793, "global_step": 77212, "epoch": 867, "lr": 8.149131820592695e-05}
{"train_loss": 0.18776758015155792, "global_step": 77213, "epoch": 867, "lr": 8.149086790154319e-05}
{"train_loss": 0.17343267798423767, "global_step": 77214, "epoch": 867, "lr": 8.149041759292585e-05}
{"train_loss": 0.31627726554870605, "global_step": 77215, "epoch": 867, "lr": 8.148996728007501e-05}
{"train_loss": 0.3449538052082062, "global_step": 77216, "epoch": 867, "lr": 8.148951696299073e-05}
{"train_loss": 0.21912749111652374, "global_step": 77217, "epoch": 867, "lr": 8.148906664167306e-05}
{"train_loss": 0.18538889288902283, "global_step": 77218, "epoch": 867, "lr": 8.148861631612208e-05}
{"train_loss": 0.2348247468471527, "global_step": 77219, "epoch": 867, "lr": 8.148816598633784e-05}
{"train_loss": 0.19236066937446594, "global_step": 77220, "epoch": 867, "lr": 8.148771565232038e-05}
{"train_loss": 0.1557857245206833, "global_step": 77221, "epoch": 867, "lr": 8.14872653140698e-05}
{"train_loss": 0.2259577214717865, "global_step": 77222, "epoch": 867, "lr": 8.148681497158615e-05}
{"train_loss": 0.20371757447719574, "global_step": 77223, "epoch": 867, "lr": 8.148636462486945e-05}
{"train_loss": 0.22270923852920532, "global_step": 77224, "epoch": 867, "lr": 8.148591427391983e-05}
{"train_loss": 0.15288418531417847, "global_step": 77225, "epoch": 867, "lr": 8.148546391873728e-05}
{"train_loss": 0.2519432008266449, "global_step": 77226, "epoch": 867, "lr": 8.148501355932192e-05}
{"train_loss": 0.2124270498752594, "global_step": 77227, "epoch": 867, "lr": 8.148456319567377e-05}
{"train_loss": 0.25721466541290283, "global_step": 77228, "epoch": 867, "lr": 8.148411282779289e-05}
{"train_loss": 0.1299954205751419, "global_step": 77229, "epoch": 867, "lr": 8.148366245567937e-05}
{"train_loss": 0.2614890933036804, "global_step": 77230, "epoch": 867, "lr": 8.148321207933325e-05}
{"train_loss": 0.1831175982952118, "global_step": 77231, "epoch": 867, "lr": 8.148276169875461e-05}
{"train_loss": 0.2822234332561493, "global_step": 77232, "epoch": 867, "lr": 8.148231131394349e-05}
{"train_loss": 0.16924433410167694, "global_step": 77233, "epoch": 867, "lr": 8.148186092489996e-05}
{"train_loss": 0.2518612742424011, "global_step": 77234, "epoch": 867, "lr": 8.148141053162406e-05}
{"train_loss": 0.1947251856327057, "global_step": 77235, "epoch": 867, "lr": 8.148096013411588e-05}
{"train_loss": 0.25549980998039246, "global_step": 77236, "epoch": 867, "lr": 8.148050973237546e-05}
{"train_loss": 0.22173334658145905, "global_step": 77237, "epoch": 867, "lr": 8.148005932640288e-05}
{"train_loss": 0.19573462009429932, "global_step": 77238, "epoch": 867, "lr": 8.14796089161982e-05}
{"train_loss": 0.22230520844459534, "global_step": 77239, "epoch": 867, "lr": 8.147915850176145e-05}
{"train_loss": 0.17586344480514526, "global_step": 77240, "epoch": 867, "lr": 8.147870808309271e-05}
{"train_loss": 0.18220730125904083, "global_step": 77241, "epoch": 867, "lr": 8.147825766019206e-05}
{"train_loss": 0.18172934651374817, "global_step": 77242, "epoch": 867, "lr": 8.147780723305952e-05}
{"train_loss": 0.28522294759750366, "global_step": 77243, "epoch": 867, "lr": 8.14773568016952e-05}
{"train_loss": 0.14351539313793182, "global_step": 77244, "epoch": 867, "lr": 8.147690636609911e-05}
{"train_loss": 0.15374286472797394, "global_step": 77245, "epoch": 867, "lr": 8.147645592627134e-05}
{"train_loss": 0.23256343603134155, "global_step": 77246, "epoch": 867, "lr": 8.147600548221195e-05}
{"train_loss": 0.18642272055149078, "global_step": 77247, "epoch": 867, "lr": 8.147555503392098e-05}
{"train_loss": 0.2203950136899948, "global_step": 77248, "epoch": 867, "lr": 8.147510458139852e-05}
{"train_loss": 0.19065746665000916, "global_step": 77249, "epoch": 867, "lr": 8.147465412464461e-05}
{"train_loss": 0.1763228476047516, "global_step": 77250, "epoch": 867, "lr": 8.147420366365931e-05}
{"train_loss": 0.20695309483268287, "global_step": 77251, "epoch": 867, "lr": 8.147375319844269e-05, "val_loss": 3.9762282371520996}
{"train_loss": 0.1704888492822647, "global_step": 77252, "epoch": 868, "lr": 8.147330272899481e-05}
{"train_loss": 0.19150345027446747, "global_step": 77253, "epoch": 868, "lr": 8.147285225531574e-05}
{"train_loss": 0.18233101069927216, "global_step": 77254, "epoch": 868, "lr": 8.14724017774055e-05}
{"train_loss": 0.14634688198566437, "global_step": 77255, "epoch": 868, "lr": 8.147195129526421e-05}
{"train_loss": 0.29733970761299133, "global_step": 77256, "epoch": 868, "lr": 8.147150080889188e-05}
{"train_loss": 0.2068745344877243, "global_step": 77257, "epoch": 868, "lr": 8.147105031828859e-05}
{"train_loss": 0.21412988007068634, "global_step": 77258, "epoch": 868, "lr": 8.147059982345442e-05}
{"train_loss": 0.2052212953567505, "global_step": 77259, "epoch": 868, "lr": 8.147014932438939e-05}
{"train_loss": 0.17584678530693054, "global_step": 77260, "epoch": 868, "lr": 8.146969882109359e-05}
{"train_loss": 0.09421022236347198, "global_step": 77261, "epoch": 868, "lr": 8.146924831356708e-05}
{"train_loss": 0.12160065770149231, "global_step": 77262, "epoch": 868, "lr": 8.14687978018099e-05}
{"train_loss": 0.24336197972297668, "global_step": 77263, "epoch": 868, "lr": 8.146834728582213e-05}
{"train_loss": 0.18418951332569122, "global_step": 77264, "epoch": 868, "lr": 8.146789676560383e-05}
{"train_loss": 0.32911795377731323, "global_step": 77265, "epoch": 868, "lr": 8.146744624115506e-05}
{"train_loss": 0.28183650970458984, "global_step": 77266, "epoch": 868, "lr": 8.146699571247585e-05}
{"train_loss": 0.2517423927783966, "global_step": 77267, "epoch": 868, "lr": 8.146654517956631e-05}
{"train_loss": 0.17503605782985687, "global_step": 77268, "epoch": 868, "lr": 8.146609464242646e-05}
{"train_loss": 0.18516260385513306, "global_step": 77269, "epoch": 868, "lr": 8.146564410105639e-05}
{"train_loss": 0.22242581844329834, "global_step": 77270, "epoch": 868, "lr": 8.146519355545615e-05}
{"train_loss": 0.2221297323703766, "global_step": 77271, "epoch": 868, "lr": 8.146474300562578e-05}
{"train_loss": 0.22092904150485992, "global_step": 77272, "epoch": 868, "lr": 8.146429245156536e-05}
{"train_loss": 0.299879789352417, "global_step": 77273, "epoch": 868, "lr": 8.146384189327496e-05}
{"train_loss": 0.15289956331253052, "global_step": 77274, "epoch": 868, "lr": 8.146339133075463e-05}
{"train_loss": 0.29485777020454407, "global_step": 77275, "epoch": 868, "lr": 8.146294076400443e-05}
{"train_loss": 0.21199138462543488, "global_step": 77276, "epoch": 868, "lr": 8.146249019302441e-05}
{"train_loss": 0.22480128705501556, "global_step": 77277, "epoch": 868, "lr": 8.146203961781467e-05}
{"train_loss": 0.24953389167785645, "global_step": 77278, "epoch": 868, "lr": 8.146158903837522e-05}
{"train_loss": 0.24113109707832336, "global_step": 77279, "epoch": 868, "lr": 8.146113845470613e-05}
{"train_loss": 0.17454646527767181, "global_step": 77280, "epoch": 868, "lr": 8.14606878668075e-05}
{"train_loss": 0.24257992208003998, "global_step": 77281, "epoch": 868, "lr": 8.146023727467935e-05}
{"train_loss": 0.18027178943157196, "global_step": 77282, "epoch": 868, "lr": 8.145978667832178e-05}
{"train_loss": 0.17925973236560822, "global_step": 77283, "epoch": 868, "lr": 8.14593360777348e-05}
{"train_loss": 0.2087225466966629, "global_step": 77284, "epoch": 868, "lr": 8.145888547291849e-05}
{"train_loss": 0.2544391453266144, "global_step": 77285, "epoch": 868, "lr": 8.145843486387292e-05}
{"train_loss": 0.1647476702928543, "global_step": 77286, "epoch": 868, "lr": 8.145798425059817e-05}
{"train_loss": 0.23337779939174652, "global_step": 77287, "epoch": 868, "lr": 8.145753363309425e-05}
{"train_loss": 0.19235719740390778, "global_step": 77288, "epoch": 868, "lr": 8.145708301136126e-05}
{"train_loss": 0.16572719812393188, "global_step": 77289, "epoch": 868, "lr": 8.145663238539924e-05}
{"train_loss": 0.18682672083377838, "global_step": 77290, "epoch": 868, "lr": 8.145618175520826e-05}
{"train_loss": 0.16586868464946747, "global_step": 77291, "epoch": 868, "lr": 8.14557311207884e-05}
{"train_loss": 0.1821911334991455, "global_step": 77292, "epoch": 868, "lr": 8.145528048213967e-05}
{"train_loss": 0.2241235226392746, "global_step": 77293, "epoch": 868, "lr": 8.145482983926218e-05}
{"train_loss": 0.15103818476200104, "global_step": 77294, "epoch": 868, "lr": 8.145437919215597e-05}
{"train_loss": 0.32891845703125, "global_step": 77295, "epoch": 868, "lr": 8.14539285408211e-05}
{"train_loss": 0.26512035727500916, "global_step": 77296, "epoch": 868, "lr": 8.145347788525763e-05}
{"train_loss": 0.21452631056308746, "global_step": 77297, "epoch": 868, "lr": 8.145302722546562e-05}
{"train_loss": 0.2753257155418396, "global_step": 77298, "epoch": 868, "lr": 8.145257656144514e-05}
{"train_loss": 0.25347232818603516, "global_step": 77299, "epoch": 868, "lr": 8.145212589319625e-05}
{"train_loss": 0.15724463760852814, "global_step": 77300, "epoch": 868, "lr": 8.145167522071901e-05}
{"train_loss": 0.23330943286418915, "global_step": 77301, "epoch": 868, "lr": 8.145122454401346e-05}
{"train_loss": 0.28330230712890625, "global_step": 77302, "epoch": 868, "lr": 8.145077386307968e-05}
{"train_loss": 0.24731996655464172, "global_step": 77303, "epoch": 868, "lr": 8.145032317791772e-05}
{"train_loss": 0.3648921251296997, "global_step": 77304, "epoch": 868, "lr": 8.144987248852767e-05}
{"train_loss": 0.18236111104488373, "global_step": 77305, "epoch": 868, "lr": 8.144942179490956e-05}
{"train_loss": 0.1984328031539917, "global_step": 77306, "epoch": 868, "lr": 8.144897109706344e-05}
{"train_loss": 0.14718587696552277, "global_step": 77307, "epoch": 868, "lr": 8.14485203949894e-05}
{"train_loss": 0.2293020486831665, "global_step": 77308, "epoch": 868, "lr": 8.14480696886875e-05}
{"train_loss": 0.15512357652187347, "global_step": 77309, "epoch": 868, "lr": 8.144761897815777e-05}
{"train_loss": 0.20679816603660583, "global_step": 77310, "epoch": 868, "lr": 8.14471682634003e-05}
{"train_loss": 0.19486241042613983, "global_step": 77311, "epoch": 868, "lr": 8.144671754441515e-05}
{"train_loss": 0.19286185503005981, "global_step": 77312, "epoch": 868, "lr": 8.144626682120237e-05}
{"train_loss": 0.22324711084365845, "global_step": 77313, "epoch": 868, "lr": 8.1445816093762e-05}
{"train_loss": 0.23042209446430206, "global_step": 77314, "epoch": 868, "lr": 8.144536536209415e-05}
{"train_loss": 0.328670859336853, "global_step": 77315, "epoch": 868, "lr": 8.144491462619885e-05}
{"train_loss": 0.15862908959388733, "global_step": 77316, "epoch": 868, "lr": 8.144446388607616e-05}
{"train_loss": 0.2611013948917389, "global_step": 77317, "epoch": 868, "lr": 8.144401314172614e-05}
{"train_loss": 0.20198708772659302, "global_step": 77318, "epoch": 868, "lr": 8.144356239314887e-05}
{"train_loss": 0.1667332649230957, "global_step": 77319, "epoch": 868, "lr": 8.144311164034439e-05}
{"train_loss": 0.10157523304224014, "global_step": 77320, "epoch": 868, "lr": 8.144266088331277e-05}
{"train_loss": 0.1660618782043457, "global_step": 77321, "epoch": 868, "lr": 8.144221012205404e-05}
{"train_loss": 0.1756506711244583, "global_step": 77322, "epoch": 868, "lr": 8.144175935656833e-05}
{"train_loss": 0.13474854826927185, "global_step": 77323, "epoch": 868, "lr": 8.144130858685563e-05}
{"train_loss": 0.1404576152563095, "global_step": 77324, "epoch": 868, "lr": 8.144085781291604e-05}
{"train_loss": 0.18354396522045135, "global_step": 77325, "epoch": 868, "lr": 8.14404070347496e-05}
{"train_loss": 0.1254003793001175, "global_step": 77326, "epoch": 868, "lr": 8.143995625235638e-05}
{"train_loss": 0.26432475447654724, "global_step": 77327, "epoch": 868, "lr": 8.143950546573645e-05}
{"train_loss": 0.21566201746463776, "global_step": 77328, "epoch": 868, "lr": 8.143905467488986e-05}
{"train_loss": 0.25872763991355896, "global_step": 77329, "epoch": 868, "lr": 8.143860387981668e-05}
{"train_loss": 0.19684812426567078, "global_step": 77330, "epoch": 868, "lr": 8.143815308051696e-05}
{"train_loss": 0.20865315198898315, "global_step": 77331, "epoch": 868, "lr": 8.143770227699075e-05}
{"train_loss": 0.16058477759361267, "global_step": 77332, "epoch": 868, "lr": 8.143725146923815e-05}
{"train_loss": 0.2696695923805237, "global_step": 77333, "epoch": 868, "lr": 8.143680065725917e-05}
{"train_loss": 0.32453984022140503, "global_step": 77334, "epoch": 868, "lr": 8.143634984105391e-05}
{"train_loss": 0.21655184030532837, "global_step": 77335, "epoch": 868, "lr": 8.143589902062241e-05}
{"train_loss": 0.19510780274868011, "global_step": 77336, "epoch": 868, "lr": 8.143544819596474e-05}
{"train_loss": 0.18749573826789856, "global_step": 77337, "epoch": 868, "lr": 8.143499736708094e-05}
{"train_loss": 0.161219522356987, "global_step": 77338, "epoch": 868, "lr": 8.14345465339711e-05}
{"train_loss": 0.15393537282943726, "global_step": 77339, "epoch": 868, "lr": 8.143409569663527e-05}
{"train_loss": 0.2091557230339961, "global_step": 77340, "epoch": 868, "lr": 8.14336448550735e-05, "val_loss": 4.037373065948486}
{"train_loss": 0.1789998561143875, "global_step": 77341, "epoch": 869, "lr": 8.143319400928588e-05}
{"train_loss": 0.24502287805080414, "global_step": 77342, "epoch": 869, "lr": 8.143274315927244e-05}
{"train_loss": 0.20957867801189423, "global_step": 77343, "epoch": 869, "lr": 8.143229230503325e-05}
{"train_loss": 0.22372709214687347, "global_step": 77344, "epoch": 869, "lr": 8.143184144656837e-05}
{"train_loss": 0.1268499344587326, "global_step": 77345, "epoch": 869, "lr": 8.143139058387785e-05}
{"train_loss": 0.1813553422689438, "global_step": 77346, "epoch": 869, "lr": 8.143093971696178e-05}
{"train_loss": 0.2774227261543274, "global_step": 77347, "epoch": 869, "lr": 8.14304888458202e-05}
{"train_loss": 0.17785173654556274, "global_step": 77348, "epoch": 869, "lr": 8.143003797045317e-05}
{"train_loss": 0.17709891498088837, "global_step": 77349, "epoch": 869, "lr": 8.142958709086076e-05}
{"train_loss": 0.24148978292942047, "global_step": 77350, "epoch": 869, "lr": 8.142913620704302e-05}
{"train_loss": 0.18329423666000366, "global_step": 77351, "epoch": 869, "lr": 8.142868531900002e-05}
{"train_loss": 0.21860910952091217, "global_step": 77352, "epoch": 869, "lr": 8.14282344267318e-05}
{"train_loss": 0.1846044361591339, "global_step": 77353, "epoch": 869, "lr": 8.142778353023846e-05}
{"train_loss": 0.27634677290916443, "global_step": 77354, "epoch": 869, "lr": 8.142733262952003e-05}
{"train_loss": 0.1621152013540268, "global_step": 77355, "epoch": 869, "lr": 8.142688172457657e-05}
{"train_loss": 0.2063506692647934, "global_step": 77356, "epoch": 869, "lr": 8.142643081540816e-05}
{"train_loss": 0.19072961807250977, "global_step": 77357, "epoch": 869, "lr": 8.142597990201484e-05}
{"train_loss": 0.19408348202705383, "global_step": 77358, "epoch": 869, "lr": 8.142552898439668e-05}
{"train_loss": 0.21277637779712677, "global_step": 77359, "epoch": 869, "lr": 8.142507806255375e-05}
{"train_loss": 0.1264144778251648, "global_step": 77360, "epoch": 869, "lr": 8.142462713648609e-05}
{"train_loss": 0.143677219748497, "global_step": 77361, "epoch": 869, "lr": 8.142417620619379e-05}
{"train_loss": 0.09961217641830444, "global_step": 77362, "epoch": 869, "lr": 8.142372527167687e-05}
{"train_loss": 0.17829641699790955, "global_step": 77363, "epoch": 869, "lr": 8.142327433293543e-05}
{"train_loss": 0.1840820014476776, "global_step": 77364, "epoch": 869, "lr": 8.142282338996951e-05}
{"train_loss": 0.13749560713768005, "global_step": 77365, "epoch": 869, "lr": 8.142237244277917e-05}
{"train_loss": 0.16163639724254608, "global_step": 77366, "epoch": 869, "lr": 8.142192149136447e-05}
{"train_loss": 0.19960545003414154, "global_step": 77367, "epoch": 869, "lr": 8.142147053572549e-05}
{"train_loss": 0.212851881980896, "global_step": 77368, "epoch": 869, "lr": 8.142101957586227e-05}
{"train_loss": 0.09476358443498611, "global_step": 77369, "epoch": 869, "lr": 8.142056861177487e-05}
{"train_loss": 0.16321814060211182, "global_step": 77370, "epoch": 869, "lr": 8.142011764346336e-05}
{"train_loss": 0.1977119743824005, "global_step": 77371, "epoch": 869, "lr": 8.141966667092781e-05}
{"train_loss": 0.20343361794948578, "global_step": 77372, "epoch": 869, "lr": 8.141921569416826e-05}
{"train_loss": 0.13169831037521362, "global_step": 77373, "epoch": 869, "lr": 8.141876471318478e-05}
{"train_loss": 0.1872267872095108, "global_step": 77374, "epoch": 869, "lr": 8.141831372797743e-05}
{"train_loss": 0.17557622492313385, "global_step": 77375, "epoch": 869, "lr": 8.141786273854626e-05}
{"train_loss": 0.17913195490837097, "global_step": 77376, "epoch": 869, "lr": 8.141741174489137e-05}
{"train_loss": 0.2002696543931961, "global_step": 77377, "epoch": 869, "lr": 8.141696074701277e-05}
{"train_loss": 0.28695055842399597, "global_step": 77378, "epoch": 869, "lr": 8.141650974491054e-05}
{"train_loss": 0.301443487405777, "global_step": 77379, "epoch": 869, "lr": 8.141605873858475e-05}
{"train_loss": 0.15662606060504913, "global_step": 77380, "epoch": 869, "lr": 8.141560772803545e-05}
{"train_loss": 0.15796242654323578, "global_step": 77381, "epoch": 869, "lr": 8.14151567132627e-05}
{"train_loss": 0.16945166885852814, "global_step": 77382, "epoch": 869, "lr": 8.141470569426658e-05}
{"train_loss": 0.1921645849943161, "global_step": 77383, "epoch": 869, "lr": 8.141425467104713e-05}
{"train_loss": 0.16092927753925323, "global_step": 77384, "epoch": 869, "lr": 8.141380364360441e-05}
{"train_loss": 0.17482608556747437, "global_step": 77385, "epoch": 869, "lr": 8.14133526119385e-05}
{"train_loss": 0.33736151456832886, "global_step": 77386, "epoch": 869, "lr": 8.141290157604944e-05}
{"train_loss": 0.17460642755031586, "global_step": 77387, "epoch": 869, "lr": 8.141245053593729e-05}
{"train_loss": 0.20454828441143036, "global_step": 77388, "epoch": 869, "lr": 8.141199949160214e-05}
{"train_loss": 0.2722448706626892, "global_step": 77389, "epoch": 869, "lr": 8.1411548443044e-05}
{"train_loss": 0.16756363213062286, "global_step": 77390, "epoch": 869, "lr": 8.141109739026298e-05}
{"train_loss": 0.21029455959796906, "global_step": 77391, "epoch": 869, "lr": 8.141064633325912e-05}
{"train_loss": 0.14892031252384186, "global_step": 77392, "epoch": 869, "lr": 8.141019527203248e-05}
{"train_loss": 0.21388636529445648, "global_step": 77393, "epoch": 869, "lr": 8.140974420658311e-05}
{"train_loss": 0.11790410429239273, "global_step": 77394, "epoch": 869, "lr": 8.14092931369111e-05}
{"train_loss": 0.21884791553020477, "global_step": 77395, "epoch": 869, "lr": 8.140884206301649e-05}
{"train_loss": 0.11464286595582962, "global_step": 77396, "epoch": 869, "lr": 8.140839098489934e-05}
{"train_loss": 0.2750798463821411, "global_step": 77397, "epoch": 869, "lr": 8.14079399025597e-05}
{"train_loss": 0.1747310906648636, "global_step": 77398, "epoch": 869, "lr": 8.140748881599766e-05}
{"train_loss": 0.27668216824531555, "global_step": 77399, "epoch": 869, "lr": 8.140703772521328e-05}
{"train_loss": 0.1965734362602234, "global_step": 77400, "epoch": 869, "lr": 8.14065866302066e-05}
{"train_loss": 0.18852294981479645, "global_step": 77401, "epoch": 869, "lr": 8.140613553097766e-05}
{"train_loss": 0.2167590856552124, "global_step": 77402, "epoch": 869, "lr": 8.140568442752657e-05}
{"train_loss": 0.2010125368833542, "global_step": 77403, "epoch": 869, "lr": 8.140523331985337e-05}
{"train_loss": 0.1378641128540039, "global_step": 77404, "epoch": 869, "lr": 8.140478220795812e-05}
{"train_loss": 0.2636057138442993, "global_step": 77405, "epoch": 869, "lr": 8.140433109184087e-05}
{"train_loss": 0.16717743873596191, "global_step": 77406, "epoch": 869, "lr": 8.140387997150169e-05}
{"train_loss": 0.17917683720588684, "global_step": 77407, "epoch": 869, "lr": 8.140342884694066e-05}
{"train_loss": 0.1723276674747467, "global_step": 77408, "epoch": 869, "lr": 8.140297771815781e-05}
{"train_loss": 0.16190363466739655, "global_step": 77409, "epoch": 869, "lr": 8.140252658515321e-05}
{"train_loss": 0.19885799288749695, "global_step": 77410, "epoch": 869, "lr": 8.140207544792692e-05}
{"train_loss": 0.14786891639232635, "global_step": 77411, "epoch": 869, "lr": 8.140162430647902e-05}
{"train_loss": 0.17370006442070007, "global_step": 77412, "epoch": 869, "lr": 8.140117316080953e-05}
{"train_loss": 0.24372543394565582, "global_step": 77413, "epoch": 869, "lr": 8.140072201091855e-05}
{"train_loss": 0.20508411526679993, "global_step": 77414, "epoch": 869, "lr": 8.140027085680613e-05}
{"train_loss": 0.2770701050758362, "global_step": 77415, "epoch": 869, "lr": 8.139981969847232e-05}
{"train_loss": 0.09992248564958572, "global_step": 77416, "epoch": 869, "lr": 8.139936853591718e-05}
{"train_loss": 0.292269766330719, "global_step": 77417, "epoch": 869, "lr": 8.139891736914078e-05}
{"train_loss": 0.15362899005413055, "global_step": 77418, "epoch": 869, "lr": 8.139846619814317e-05}
{"train_loss": 0.29025524854660034, "global_step": 77419, "epoch": 869, "lr": 8.139801502292443e-05}
{"train_loss": 0.2609618604183197, "global_step": 77420, "epoch": 869, "lr": 8.139756384348463e-05}
{"train_loss": 0.2456810474395752, "global_step": 77421, "epoch": 869, "lr": 8.139711265982378e-05}
{"train_loss": 0.26497411727905273, "global_step": 77422, "epoch": 869, "lr": 8.139666147194198e-05}
{"train_loss": 0.1910129338502884, "global_step": 77423, "epoch": 869, "lr": 8.139621027983928e-05}
{"train_loss": 0.193132683634758, "global_step": 77424, "epoch": 869, "lr": 8.139575908351576e-05}
{"train_loss": 0.2523520886898041, "global_step": 77425, "epoch": 869, "lr": 8.139530788297144e-05}
{"train_loss": 0.17091576755046844, "global_step": 77426, "epoch": 869, "lr": 8.139485667820641e-05}
{"train_loss": 0.1712544560432434, "global_step": 77427, "epoch": 869, "lr": 8.139440546922073e-05}
{"train_loss": 0.22358350455760956, "global_step": 77428, "epoch": 869, "lr": 8.139395425601446e-05}
{"train_loss": 0.19654423902543744, "global_step": 77429, "epoch": 869, "lr": 8.139350303858764e-05, "val_loss": 4.018808841705322}
{"train_loss": 0.28446269035339355, "global_step": 77430, "epoch": 870, "lr": 8.139305181694034e-05}
{"train_loss": 0.21540340781211853, "global_step": 77431, "epoch": 870, "lr": 8.139260059107265e-05}
{"train_loss": 0.2150944620370865, "global_step": 77432, "epoch": 870, "lr": 8.139214936098461e-05}
{"train_loss": 0.2386780083179474, "global_step": 77433, "epoch": 870, "lr": 8.139169812667625e-05}
{"train_loss": 0.2579067647457123, "global_step": 77434, "epoch": 870, "lr": 8.139124688814768e-05}
{"train_loss": 0.23249897360801697, "global_step": 77435, "epoch": 870, "lr": 8.139079564539894e-05}
{"train_loss": 0.223672017455101, "global_step": 77436, "epoch": 870, "lr": 8.139034439843007e-05}
{"train_loss": 0.12011069059371948, "global_step": 77437, "epoch": 870, "lr": 8.138989314724116e-05}
{"train_loss": 0.16796398162841797, "global_step": 77438, "epoch": 870, "lr": 8.138944189183226e-05}
{"train_loss": 0.20215058326721191, "global_step": 77439, "epoch": 870, "lr": 8.138899063220344e-05}
{"train_loss": 0.17990192770957947, "global_step": 77440, "epoch": 870, "lr": 8.138853936835475e-05}
{"train_loss": 0.22338762879371643, "global_step": 77441, "epoch": 870, "lr": 8.138808810028626e-05}
{"train_loss": 0.1831524521112442, "global_step": 77442, "epoch": 870, "lr": 8.138763682799802e-05}
{"train_loss": 0.20878919959068298, "global_step": 77443, "epoch": 870, "lr": 8.138718555149009e-05}
{"train_loss": 0.16288138926029205, "global_step": 77444, "epoch": 870, "lr": 8.138673427076253e-05}
{"train_loss": 0.24018220603466034, "global_step": 77445, "epoch": 870, "lr": 8.138628298581541e-05}
{"train_loss": 0.22790125012397766, "global_step": 77446, "epoch": 870, "lr": 8.13858316966488e-05}
{"train_loss": 0.16749440133571625, "global_step": 77447, "epoch": 870, "lr": 8.138538040326274e-05}
{"train_loss": 0.26326701045036316, "global_step": 77448, "epoch": 870, "lr": 8.13849291056573e-05}
{"train_loss": 0.19541126489639282, "global_step": 77449, "epoch": 870, "lr": 8.138447780383254e-05}
{"train_loss": 0.27777454257011414, "global_step": 77450, "epoch": 870, "lr": 8.13840264977885e-05}
{"train_loss": 0.2312050312757492, "global_step": 77451, "epoch": 870, "lr": 8.138357518752528e-05}
{"train_loss": 0.33899831771850586, "global_step": 77452, "epoch": 870, "lr": 8.138312387304292e-05}
{"train_loss": 0.2452528178691864, "global_step": 77453, "epoch": 870, "lr": 8.138267255434147e-05}
{"train_loss": 0.1950344741344452, "global_step": 77454, "epoch": 870, "lr": 8.1382221231421e-05}
{"train_loss": 0.20045292377471924, "global_step": 77455, "epoch": 870, "lr": 8.138176990428159e-05}
{"train_loss": 0.16595914959907532, "global_step": 77456, "epoch": 870, "lr": 8.138131857292328e-05}
{"train_loss": 0.20236240327358246, "global_step": 77457, "epoch": 870, "lr": 8.138086723734613e-05}
{"train_loss": 0.21493834257125854, "global_step": 77458, "epoch": 870, "lr": 8.138041589755021e-05}
{"train_loss": 0.19048038125038147, "global_step": 77459, "epoch": 870, "lr": 8.137996455353558e-05}
{"train_loss": 0.11533484607934952, "global_step": 77460, "epoch": 870, "lr": 8.137951320530228e-05}
{"train_loss": 0.18917740881443024, "global_step": 77461, "epoch": 870, "lr": 8.137906185285039e-05}
{"train_loss": 0.2532128393650055, "global_step": 77462, "epoch": 870, "lr": 8.137861049617997e-05}
{"train_loss": 0.19609075784683228, "global_step": 77463, "epoch": 870, "lr": 8.137815913529108e-05}
{"train_loss": 0.2246999740600586, "global_step": 77464, "epoch": 870, "lr": 8.137770777018379e-05}
{"train_loss": 0.3074705898761749, "global_step": 77465, "epoch": 870, "lr": 8.137725640085814e-05}
{"train_loss": 0.29715991020202637, "global_step": 77466, "epoch": 870, "lr": 8.13768050273142e-05}
{"train_loss": 0.17328450083732605, "global_step": 77467, "epoch": 870, "lr": 8.137635364955201e-05}
{"train_loss": 0.26815783977508545, "global_step": 77468, "epoch": 870, "lr": 8.137590226757168e-05}
{"train_loss": 0.1681298464536667, "global_step": 77469, "epoch": 870, "lr": 8.137545088137324e-05}
{"train_loss": 0.2017897665500641, "global_step": 77470, "epoch": 870, "lr": 8.137499949095674e-05}
{"train_loss": 0.21415331959724426, "global_step": 77471, "epoch": 870, "lr": 8.137454809632226e-05}
{"train_loss": 0.2421654313802719, "global_step": 77472, "epoch": 870, "lr": 8.137409669746985e-05}
{"train_loss": 0.18487654626369476, "global_step": 77473, "epoch": 870, "lr": 8.137364529439957e-05}
{"train_loss": 0.17708370089530945, "global_step": 77474, "epoch": 870, "lr": 8.137319388711151e-05}
{"train_loss": 0.21605126559734344, "global_step": 77475, "epoch": 870, "lr": 8.137274247560569e-05}
{"train_loss": 0.26847535371780396, "global_step": 77476, "epoch": 870, "lr": 8.137229105988219e-05}
{"train_loss": 0.19482631981372833, "global_step": 77477, "epoch": 870, "lr": 8.137183963994106e-05}
{"train_loss": 0.3063935935497284, "global_step": 77478, "epoch": 870, "lr": 8.137138821578239e-05}
{"train_loss": 0.1697729229927063, "global_step": 77479, "epoch": 870, "lr": 8.13709367874062e-05}
{"train_loss": 0.2733166813850403, "global_step": 77480, "epoch": 870, "lr": 8.137048535481258e-05}
{"train_loss": 0.1635504812002182, "global_step": 77481, "epoch": 870, "lr": 8.137003391800156e-05}
{"train_loss": 0.2011709362268448, "global_step": 77482, "epoch": 870, "lr": 8.136958247697324e-05}
{"train_loss": 0.16335102915763855, "global_step": 77483, "epoch": 870, "lr": 8.136913103172765e-05}
{"train_loss": 0.1364802122116089, "global_step": 77484, "epoch": 870, "lr": 8.136867958226489e-05}
{"train_loss": 0.15888722240924835, "global_step": 77485, "epoch": 870, "lr": 8.136822812858496e-05}
{"train_loss": 0.21014375984668732, "global_step": 77486, "epoch": 870, "lr": 8.136777667068798e-05}
{"train_loss": 0.19903582334518433, "global_step": 77487, "epoch": 870, "lr": 8.136732520857397e-05}
{"train_loss": 0.265763521194458, "global_step": 77488, "epoch": 870, "lr": 8.1366873742243e-05}
{"train_loss": 0.09886724501848221, "global_step": 77489, "epoch": 870, "lr": 8.136642227169516e-05}
{"train_loss": 0.23227526247501373, "global_step": 77490, "epoch": 870, "lr": 8.136597079693045e-05}
{"train_loss": 0.25137022137641907, "global_step": 77491, "epoch": 870, "lr": 8.1365519317949e-05}
{"train_loss": 0.14873287081718445, "global_step": 77492, "epoch": 870, "lr": 8.136506783475083e-05}
{"train_loss": 0.18278425931930542, "global_step": 77493, "epoch": 870, "lr": 8.1364616347336e-05}
{"train_loss": 0.19315189123153687, "global_step": 77494, "epoch": 870, "lr": 8.13641648557046e-05}
{"train_loss": 0.24817141890525818, "global_step": 77495, "epoch": 870, "lr": 8.136371335985665e-05}
{"train_loss": 0.16683797538280487, "global_step": 77496, "epoch": 870, "lr": 8.136326185979225e-05}
{"train_loss": 0.25010326504707336, "global_step": 77497, "epoch": 870, "lr": 8.136281035551141e-05}
{"train_loss": 0.19267471134662628, "global_step": 77498, "epoch": 870, "lr": 8.136235884701424e-05}
{"train_loss": 0.24382944405078888, "global_step": 77499, "epoch": 870, "lr": 8.136190733430079e-05}
{"train_loss": 0.15695741772651672, "global_step": 77500, "epoch": 870, "lr": 8.136145581737111e-05}
{"train_loss": 0.25522762537002563, "global_step": 77501, "epoch": 870, "lr": 8.136100429622526e-05}
{"train_loss": 0.2993747293949127, "global_step": 77502, "epoch": 870, "lr": 8.136055277086331e-05}
{"train_loss": 0.12506632506847382, "global_step": 77503, "epoch": 870, "lr": 8.136010124128531e-05}
{"train_loss": 0.18257713317871094, "global_step": 77504, "epoch": 870, "lr": 8.135964970749133e-05}
{"train_loss": 0.2169884890317917, "global_step": 77505, "epoch": 870, "lr": 8.135919816948143e-05}
{"train_loss": 0.233218714594841, "global_step": 77506, "epoch": 870, "lr": 8.135874662725567e-05}
{"train_loss": 0.21217450499534607, "global_step": 77507, "epoch": 870, "lr": 8.13582950808141e-05}
{"train_loss": 0.21212884783744812, "global_step": 77508, "epoch": 870, "lr": 8.135784353015679e-05}
{"train_loss": 0.18621009588241577, "global_step": 77509, "epoch": 870, "lr": 8.13573919752838e-05}
{"train_loss": 0.20100727677345276, "global_step": 77510, "epoch": 870, "lr": 8.13569404161952e-05}
{"train_loss": 0.22789734601974487, "global_step": 77511, "epoch": 870, "lr": 8.135648885289104e-05}
{"train_loss": 0.1803009808063507, "global_step": 77512, "epoch": 870, "lr": 8.135603728537138e-05}
{"train_loss": 0.1892682909965515, "global_step": 77513, "epoch": 870, "lr": 8.135558571363628e-05}
{"train_loss": 0.18192453682422638, "global_step": 77514, "epoch": 870, "lr": 8.13551341376858e-05}
{"train_loss": 0.18025383353233337, "global_step": 77515, "epoch": 870, "lr": 8.135468255752e-05}
{"train_loss": 0.1930147409439087, "global_step": 77516, "epoch": 870, "lr": 8.135423097313896e-05}
{"train_loss": 0.14687538146972656, "global_step": 77517, "epoch": 870, "lr": 8.135377938454274e-05}
{"train_loss": 0.20853235580947962, "global_step": 77518, "epoch": 870, "lr": 8.135332779173136e-05, "val_loss": 4.0705485343933105, "train_action_mse_error": 21.292274475097656}
{"train_loss": 0.14342361688613892, "global_step": 77519, "epoch": 871, "lr": 8.135287619470492e-05}
{"train_loss": 0.17436808347702026, "global_step": 77520, "epoch": 871, "lr": 8.135242459346346e-05}
{"train_loss": 0.24922145903110504, "global_step": 77521, "epoch": 871, "lr": 8.135197298800705e-05}
{"train_loss": 0.19100426137447357, "global_step": 77522, "epoch": 871, "lr": 8.135152137833574e-05}
{"train_loss": 0.1521931290626526, "global_step": 77523, "epoch": 871, "lr": 8.135106976444963e-05}
{"train_loss": 0.4528028666973114, "global_step": 77524, "epoch": 871, "lr": 8.135061814634872e-05}
{"train_loss": 0.22428737580776215, "global_step": 77525, "epoch": 871, "lr": 8.135016652403313e-05}
{"train_loss": 0.21026700735092163, "global_step": 77526, "epoch": 871, "lr": 8.134971489750285e-05}
{"train_loss": 0.2513023316860199, "global_step": 77527, "epoch": 871, "lr": 8.134926326675801e-05}
{"train_loss": 0.19092410802841187, "global_step": 77528, "epoch": 871, "lr": 8.134881163179864e-05}
{"train_loss": 0.20964112877845764, "global_step": 77529, "epoch": 871, "lr": 8.134835999262482e-05}
{"train_loss": 0.18365900218486786, "global_step": 77530, "epoch": 871, "lr": 8.134790834923657e-05}
{"train_loss": 0.23668769001960754, "global_step": 77531, "epoch": 871, "lr": 8.1347456701634e-05}
{"train_loss": 0.18360337615013123, "global_step": 77532, "epoch": 871, "lr": 8.134700504981712e-05}
{"train_loss": 0.12953872978687286, "global_step": 77533, "epoch": 871, "lr": 8.134655339378604e-05}
{"train_loss": 0.1904749870300293, "global_step": 77534, "epoch": 871, "lr": 8.134610173354078e-05}
{"train_loss": 0.2030651867389679, "global_step": 77535, "epoch": 871, "lr": 8.134565006908144e-05}
{"train_loss": 0.23579813539981842, "global_step": 77536, "epoch": 871, "lr": 8.134519840040804e-05}
{"train_loss": 0.24514822661876678, "global_step": 77537, "epoch": 871, "lr": 8.134474672752066e-05}
{"train_loss": 0.2249615639448166, "global_step": 77538, "epoch": 871, "lr": 8.134429505041938e-05}
{"train_loss": 0.17860683798789978, "global_step": 77539, "epoch": 871, "lr": 8.134384336910422e-05}
{"train_loss": 0.1713469922542572, "global_step": 77540, "epoch": 871, "lr": 8.134339168357527e-05}
{"train_loss": 0.23856782913208008, "global_step": 77541, "epoch": 871, "lr": 8.134293999383259e-05}
{"train_loss": 0.2705940306186676, "global_step": 77542, "epoch": 871, "lr": 8.134248829987624e-05}
{"train_loss": 0.22870613634586334, "global_step": 77543, "epoch": 871, "lr": 8.134203660170625e-05}
{"train_loss": 0.22110405564308167, "global_step": 77544, "epoch": 871, "lr": 8.134158489932271e-05}
{"train_loss": 0.19886252284049988, "global_step": 77545, "epoch": 871, "lr": 8.13411331927257e-05}
{"train_loss": 0.16966083645820618, "global_step": 77546, "epoch": 871, "lr": 8.134068148191523e-05}
{"train_loss": 0.21495060622692108, "global_step": 77547, "epoch": 871, "lr": 8.134022976689139e-05}
{"train_loss": 0.2609262466430664, "global_step": 77548, "epoch": 871, "lr": 8.133977804765425e-05}
{"train_loss": 0.12839408218860626, "global_step": 77549, "epoch": 871, "lr": 8.133932632420385e-05}
{"train_loss": 0.19752514362335205, "global_step": 77550, "epoch": 871, "lr": 8.133887459654028e-05}
{"train_loss": 0.188138946890831, "global_step": 77551, "epoch": 871, "lr": 8.133842286466356e-05}
{"train_loss": 0.20831738412380219, "global_step": 77552, "epoch": 871, "lr": 8.133797112857378e-05}
{"train_loss": 0.23851338028907776, "global_step": 77553, "epoch": 871, "lr": 8.133751938827098e-05}
{"train_loss": 0.2118614912033081, "global_step": 77554, "epoch": 871, "lr": 8.133706764375523e-05}
{"train_loss": 0.2732282280921936, "global_step": 77555, "epoch": 871, "lr": 8.133661589502661e-05}
{"train_loss": 0.18299809098243713, "global_step": 77556, "epoch": 871, "lr": 8.133616414208516e-05}
{"train_loss": 0.17672479152679443, "global_step": 77557, "epoch": 871, "lr": 8.133571238493095e-05}
{"train_loss": 0.23185114562511444, "global_step": 77558, "epoch": 871, "lr": 8.133526062356402e-05}
{"train_loss": 0.2482903003692627, "global_step": 77559, "epoch": 871, "lr": 8.133480885798446e-05}
{"train_loss": 0.22640462219715118, "global_step": 77560, "epoch": 871, "lr": 8.13343570881923e-05}
{"train_loss": 0.20097589492797852, "global_step": 77561, "epoch": 871, "lr": 8.133390531418763e-05}
{"train_loss": 0.18293671309947968, "global_step": 77562, "epoch": 871, "lr": 8.13334535359705e-05}
{"train_loss": 0.14861942827701569, "global_step": 77563, "epoch": 871, "lr": 8.133300175354096e-05}
{"train_loss": 0.16238296031951904, "global_step": 77564, "epoch": 871, "lr": 8.133254996689909e-05}
{"train_loss": 0.14745476841926575, "global_step": 77565, "epoch": 871, "lr": 8.133209817604493e-05}
{"train_loss": 0.21333464980125427, "global_step": 77566, "epoch": 871, "lr": 8.133164638097855e-05}
{"train_loss": 0.16209284961223602, "global_step": 77567, "epoch": 871, "lr": 8.133119458170002e-05}
{"train_loss": 0.2299586832523346, "global_step": 77568, "epoch": 871, "lr": 8.13307427782094e-05}
{"train_loss": 0.18155163526535034, "global_step": 77569, "epoch": 871, "lr": 8.133029097050673e-05}
{"train_loss": 0.15207260847091675, "global_step": 77570, "epoch": 871, "lr": 8.13298391585921e-05}
{"train_loss": 0.1701006144285202, "global_step": 77571, "epoch": 871, "lr": 8.132938734246554e-05}
{"train_loss": 0.37198692560195923, "global_step": 77572, "epoch": 871, "lr": 8.132893552212712e-05}
{"train_loss": 0.13689415156841278, "global_step": 77573, "epoch": 871, "lr": 8.132848369757692e-05}
{"train_loss": 0.18469280004501343, "global_step": 77574, "epoch": 871, "lr": 8.132803186881499e-05}
{"train_loss": 0.10382425040006638, "global_step": 77575, "epoch": 871, "lr": 8.132758003584138e-05}
{"train_loss": 0.246375173330307, "global_step": 77576, "epoch": 871, "lr": 8.132712819865616e-05}
{"train_loss": 0.17724204063415527, "global_step": 77577, "epoch": 871, "lr": 8.132667635725938e-05}
{"train_loss": 0.2596747875213623, "global_step": 77578, "epoch": 871, "lr": 8.132622451165113e-05}
{"train_loss": 0.23868396878242493, "global_step": 77579, "epoch": 871, "lr": 8.132577266183143e-05}
{"train_loss": 0.18840721249580383, "global_step": 77580, "epoch": 871, "lr": 8.132532080780038e-05}
{"train_loss": 0.17672695219516754, "global_step": 77581, "epoch": 871, "lr": 8.1324868949558e-05}
{"train_loss": 0.15376995503902435, "global_step": 77582, "epoch": 871, "lr": 8.132441708710439e-05}
{"train_loss": 0.1808863878250122, "global_step": 77583, "epoch": 871, "lr": 8.13239652204396e-05}
{"train_loss": 0.1436903029680252, "global_step": 77584, "epoch": 871, "lr": 8.132351334956367e-05}
{"train_loss": 0.20435571670532227, "global_step": 77585, "epoch": 871, "lr": 8.13230614744767e-05}
{"train_loss": 0.11744808405637741, "global_step": 77586, "epoch": 871, "lr": 8.13226095951787e-05}
{"train_loss": 0.11888831108808517, "global_step": 77587, "epoch": 871, "lr": 8.132215771166975e-05}
{"train_loss": 0.15632301568984985, "global_step": 77588, "epoch": 871, "lr": 8.132170582394995e-05}
{"train_loss": 0.14238588511943817, "global_step": 77589, "epoch": 871, "lr": 8.132125393201931e-05}
{"train_loss": 0.19065316021442413, "global_step": 77590, "epoch": 871, "lr": 8.132080203587791e-05}
{"train_loss": 0.21425175666809082, "global_step": 77591, "epoch": 871, "lr": 8.132035013552581e-05}
{"train_loss": 0.23925718665122986, "global_step": 77592, "epoch": 871, "lr": 8.131989823096306e-05}
{"train_loss": 0.21779736876487732, "global_step": 77593, "epoch": 871, "lr": 8.131944632218975e-05}
{"train_loss": 0.1605808585882187, "global_step": 77594, "epoch": 871, "lr": 8.131899440920592e-05}
{"train_loss": 0.2860173285007477, "global_step": 77595, "epoch": 871, "lr": 8.131854249201162e-05}
{"train_loss": 0.21330468356609344, "global_step": 77596, "epoch": 871, "lr": 8.131809057060693e-05}
{"train_loss": 0.2280162274837494, "global_step": 77597, "epoch": 871, "lr": 8.13176386449919e-05}
{"train_loss": 0.15787485241889954, "global_step": 77598, "epoch": 871, "lr": 8.131718671516661e-05}
{"train_loss": 0.20003244280815125, "global_step": 77599, "epoch": 871, "lr": 8.131673478113109e-05}
{"train_loss": 0.1839333176612854, "global_step": 77600, "epoch": 871, "lr": 8.131628284288543e-05}
{"train_loss": 0.1349087506532669, "global_step": 77601, "epoch": 871, "lr": 8.131583090042967e-05}
{"train_loss": 0.16719624400138855, "global_step": 77602, "epoch": 871, "lr": 8.131537895376388e-05}
{"train_loss": 0.23404750227928162, "global_step": 77603, "epoch": 871, "lr": 8.131492700288812e-05}
{"train_loss": 0.18018889427185059, "global_step": 77604, "epoch": 871, "lr": 8.131447504780245e-05}
{"train_loss": 0.1830013543367386, "global_step": 77605, "epoch": 871, "lr": 8.131402308850692e-05}
{"train_loss": 0.15242929756641388, "global_step": 77606, "epoch": 871, "lr": 8.13135711250016e-05}
{"train_loss": 0.19876940238676713, "global_step": 77607, "epoch": 871, "lr": 8.131311915728656e-05, "val_loss": 4.042072772979736}
{"train_loss": 0.17786891758441925, "global_step": 77608, "epoch": 872, "lr": 8.131266718536187e-05}
{"train_loss": 0.16637691855430603, "global_step": 77609, "epoch": 872, "lr": 8.131221520922756e-05}
{"train_loss": 0.25858837366104126, "global_step": 77610, "epoch": 872, "lr": 8.131176322888369e-05}
{"train_loss": 0.16171523928642273, "global_step": 77611, "epoch": 872, "lr": 8.131131124433035e-05}
{"train_loss": 0.12082009762525558, "global_step": 77612, "epoch": 872, "lr": 8.131085925556758e-05}
{"train_loss": 0.15049859881401062, "global_step": 77613, "epoch": 872, "lr": 8.131040726259547e-05}
{"train_loss": 0.1924280822277069, "global_step": 77614, "epoch": 872, "lr": 8.130995526541403e-05}
{"train_loss": 0.1284327507019043, "global_step": 77615, "epoch": 872, "lr": 8.130950326402335e-05}
{"train_loss": 0.2655321955680847, "global_step": 77616, "epoch": 872, "lr": 8.13090512584235e-05}
{"train_loss": 0.19129687547683716, "global_step": 77617, "epoch": 872, "lr": 8.130859924861452e-05}
{"train_loss": 0.16515672206878662, "global_step": 77618, "epoch": 872, "lr": 8.130814723459648e-05}
{"train_loss": 0.18498601019382477, "global_step": 77619, "epoch": 872, "lr": 8.130769521636944e-05}
{"train_loss": 0.1538887917995453, "global_step": 77620, "epoch": 872, "lr": 8.130724319393348e-05}
{"train_loss": 0.24926212430000305, "global_step": 77621, "epoch": 872, "lr": 8.130679116728862e-05}
{"train_loss": 0.2430075705051422, "global_step": 77622, "epoch": 872, "lr": 8.130633913643496e-05}
{"train_loss": 0.15824808180332184, "global_step": 77623, "epoch": 872, "lr": 8.130588710137254e-05}
{"train_loss": 0.17832158505916595, "global_step": 77624, "epoch": 872, "lr": 8.130543506210142e-05}
{"train_loss": 0.14662271738052368, "global_step": 77625, "epoch": 872, "lr": 8.130498301862166e-05}
{"train_loss": 0.17962117493152618, "global_step": 77626, "epoch": 872, "lr": 8.130453097093334e-05}
{"train_loss": 0.23260392248630524, "global_step": 77627, "epoch": 872, "lr": 8.130407891903651e-05}
{"train_loss": 0.15407003462314606, "global_step": 77628, "epoch": 872, "lr": 8.130362686293122e-05}
{"train_loss": 0.23909428715705872, "global_step": 77629, "epoch": 872, "lr": 8.130317480261753e-05}
{"train_loss": 0.15810877084732056, "global_step": 77630, "epoch": 872, "lr": 8.130272273809553e-05}
{"train_loss": 0.20345337688922882, "global_step": 77631, "epoch": 872, "lr": 8.130227066936525e-05}
{"train_loss": 0.21165813505649567, "global_step": 77632, "epoch": 872, "lr": 8.130181859642675e-05}
{"train_loss": 0.13414266705513, "global_step": 77633, "epoch": 872, "lr": 8.130136651928012e-05}
{"train_loss": 0.13860075175762177, "global_step": 77634, "epoch": 872, "lr": 8.13009144379254e-05}
{"train_loss": 0.20212067663669586, "global_step": 77635, "epoch": 872, "lr": 8.130046235236264e-05}
{"train_loss": 0.13888511061668396, "global_step": 77636, "epoch": 872, "lr": 8.130001026259193e-05}
{"train_loss": 0.166005477309227, "global_step": 77637, "epoch": 872, "lr": 8.12995581686133e-05}
{"train_loss": 0.17188140749931335, "global_step": 77638, "epoch": 872, "lr": 8.129910607042684e-05}
{"train_loss": 0.2875838875770569, "global_step": 77639, "epoch": 872, "lr": 8.129865396803259e-05}
{"train_loss": 0.18951091170310974, "global_step": 77640, "epoch": 872, "lr": 8.129820186143063e-05}
{"train_loss": 0.11512643843889236, "global_step": 77641, "epoch": 872, "lr": 8.1297749750621e-05}
{"train_loss": 0.17098106443881989, "global_step": 77642, "epoch": 872, "lr": 8.129729763560375e-05}
{"train_loss": 0.25941577553749084, "global_step": 77643, "epoch": 872, "lr": 8.129684551637897e-05}
{"train_loss": 0.18670222163200378, "global_step": 77644, "epoch": 872, "lr": 8.129639339294672e-05}
{"train_loss": 0.12212593108415604, "global_step": 77645, "epoch": 872, "lr": 8.129594126530706e-05}
{"train_loss": 0.239806666970253, "global_step": 77646, "epoch": 872, "lr": 8.129548913346003e-05}
{"train_loss": 0.19122177362442017, "global_step": 77647, "epoch": 872, "lr": 8.129503699740571e-05}
{"train_loss": 0.2383648008108139, "global_step": 77648, "epoch": 872, "lr": 8.129458485714414e-05}
{"train_loss": 0.21074634790420532, "global_step": 77649, "epoch": 872, "lr": 8.12941327126754e-05}
{"train_loss": 0.17035961151123047, "global_step": 77650, "epoch": 872, "lr": 8.129368056399955e-05}
{"train_loss": 0.1540842205286026, "global_step": 77651, "epoch": 872, "lr": 8.129322841111664e-05}
{"train_loss": 0.20101884007453918, "global_step": 77652, "epoch": 872, "lr": 8.129277625402674e-05}
{"train_loss": 0.24528837203979492, "global_step": 77653, "epoch": 872, "lr": 8.129232409272989e-05}
{"train_loss": 0.16574764251708984, "global_step": 77654, "epoch": 872, "lr": 8.12918719272262e-05}
{"train_loss": 0.12732107937335968, "global_step": 77655, "epoch": 872, "lr": 8.129141975751567e-05}
{"train_loss": 0.2333325743675232, "global_step": 77656, "epoch": 872, "lr": 8.129096758359841e-05}
{"train_loss": 0.21748781204223633, "global_step": 77657, "epoch": 872, "lr": 8.129051540547446e-05}
{"train_loss": 0.18684422969818115, "global_step": 77658, "epoch": 872, "lr": 8.129006322314387e-05}
{"train_loss": 0.18722979724407196, "global_step": 77659, "epoch": 872, "lr": 8.128961103660673e-05}
{"train_loss": 0.20530229806900024, "global_step": 77660, "epoch": 872, "lr": 8.128915884586307e-05}
{"train_loss": 0.1599864363670349, "global_step": 77661, "epoch": 872, "lr": 8.128870665091295e-05}
{"train_loss": 0.22317907214164734, "global_step": 77662, "epoch": 872, "lr": 8.128825445175647e-05}
{"train_loss": 0.20624952018260956, "global_step": 77663, "epoch": 872, "lr": 8.128780224839365e-05}
{"train_loss": 0.2370578944683075, "global_step": 77664, "epoch": 872, "lr": 8.128735004082458e-05}
{"train_loss": 0.1813463270664215, "global_step": 77665, "epoch": 872, "lr": 8.12868978290493e-05}
{"train_loss": 0.15842877328395844, "global_step": 77666, "epoch": 872, "lr": 8.128644561306788e-05}
{"train_loss": 0.16903237998485565, "global_step": 77667, "epoch": 872, "lr": 8.128599339288038e-05}
{"train_loss": 0.13435964286327362, "global_step": 77668, "epoch": 872, "lr": 8.128554116848686e-05}
{"train_loss": 0.14018993079662323, "global_step": 77669, "epoch": 872, "lr": 8.128508893988737e-05}
{"train_loss": 0.2453048825263977, "global_step": 77670, "epoch": 872, "lr": 8.1284636707082e-05}
{"train_loss": 0.1368420124053955, "global_step": 77671, "epoch": 872, "lr": 8.128418447007076e-05}
{"train_loss": 0.24208813905715942, "global_step": 77672, "epoch": 872, "lr": 8.128373222885378e-05}
{"train_loss": 0.1879768669605255, "global_step": 77673, "epoch": 872, "lr": 8.128327998343105e-05}
{"train_loss": 0.15380138158798218, "global_step": 77674, "epoch": 872, "lr": 8.12828277338027e-05}
{"train_loss": 0.1407468169927597, "global_step": 77675, "epoch": 872, "lr": 8.128237547996872e-05}
{"train_loss": 0.21454308927059174, "global_step": 77676, "epoch": 872, "lr": 8.128192322192923e-05}
{"train_loss": 0.2772538661956787, "global_step": 77677, "epoch": 872, "lr": 8.128147095968425e-05}
{"train_loss": 0.15706439316272736, "global_step": 77678, "epoch": 872, "lr": 8.128101869323386e-05}
{"train_loss": 0.19664201140403748, "global_step": 77679, "epoch": 872, "lr": 8.128056642257813e-05}
{"train_loss": 0.16665863990783691, "global_step": 77680, "epoch": 872, "lr": 8.12801141477171e-05}
{"train_loss": 0.14276160299777985, "global_step": 77681, "epoch": 872, "lr": 8.127966186865083e-05}
{"train_loss": 0.17922112345695496, "global_step": 77682, "epoch": 872, "lr": 8.127920958537943e-05}
{"train_loss": 0.19512231647968292, "global_step": 77683, "epoch": 872, "lr": 8.127875729790288e-05}
{"train_loss": 0.1853775829076767, "global_step": 77684, "epoch": 872, "lr": 8.12783050062213e-05}
{"train_loss": 0.2804391086101532, "global_step": 77685, "epoch": 872, "lr": 8.127785271033473e-05}
{"train_loss": 0.1909448653459549, "global_step": 77686, "epoch": 872, "lr": 8.127740041024323e-05}
{"train_loss": 0.15267698466777802, "global_step": 77687, "epoch": 872, "lr": 8.127694810594689e-05}
{"train_loss": 0.25019511580467224, "global_step": 77688, "epoch": 872, "lr": 8.127649579744571e-05}
{"train_loss": 0.2690054178237915, "global_step": 77689, "epoch": 872, "lr": 8.12760434847398e-05}
{"train_loss": 0.3336849510669708, "global_step": 77690, "epoch": 872, "lr": 8.12755911678292e-05}
{"train_loss": 0.16412100195884705, "global_step": 77691, "epoch": 872, "lr": 8.127513884671399e-05}
{"train_loss": 0.2099771499633789, "global_step": 77692, "epoch": 872, "lr": 8.127468652139421e-05}
{"train_loss": 0.1420402228832245, "global_step": 77693, "epoch": 872, "lr": 8.127423419186995e-05}
{"train_loss": 0.11733358353376389, "global_step": 77694, "epoch": 872, "lr": 8.127378185814122e-05}
{"train_loss": 0.18954016268253326, "global_step": 77695, "epoch": 872, "lr": 8.127332952020814e-05}
{"train_loss": 0.18917695692415987, "global_step": 77696, "epoch": 872, "lr": 8.127287717807072e-05, "val_loss": 4.069252014160156}
{"train_loss": 0.20065733790397644, "global_step": 77697, "epoch": 873, "lr": 8.127242483172906e-05}
{"train_loss": 0.25092294812202454, "global_step": 77698, "epoch": 873, "lr": 8.127197248118319e-05}
{"train_loss": 0.19809633493423462, "global_step": 77699, "epoch": 873, "lr": 8.127152012643318e-05}
{"train_loss": 0.18904314935207367, "global_step": 77700, "epoch": 873, "lr": 8.12710677674791e-05}
{"train_loss": 0.17110048234462738, "global_step": 77701, "epoch": 873, "lr": 8.127061540432102e-05}
{"train_loss": 0.22491340339183807, "global_step": 77702, "epoch": 873, "lr": 8.127016303695897e-05}
{"train_loss": 0.245180144906044, "global_step": 77703, "epoch": 873, "lr": 8.126971066539304e-05}
{"train_loss": 0.3043456971645355, "global_step": 77704, "epoch": 873, "lr": 8.126925828962328e-05}
{"train_loss": 0.16265277564525604, "global_step": 77705, "epoch": 873, "lr": 8.126880590964973e-05}
{"train_loss": 0.2532068192958832, "global_step": 77706, "epoch": 873, "lr": 8.126835352547248e-05}
{"train_loss": 0.1803337037563324, "global_step": 77707, "epoch": 873, "lr": 8.126790113709159e-05}
{"train_loss": 0.2560383081436157, "global_step": 77708, "epoch": 873, "lr": 8.12674487445071e-05}
{"train_loss": 0.1778469681739807, "global_step": 77709, "epoch": 873, "lr": 8.126699634771908e-05}
{"train_loss": 0.15219806134700775, "global_step": 77710, "epoch": 873, "lr": 8.126654394672762e-05}
{"train_loss": 0.20980699360370636, "global_step": 77711, "epoch": 873, "lr": 8.126609154153273e-05}
{"train_loss": 0.19194822013378143, "global_step": 77712, "epoch": 873, "lr": 8.12656391321345e-05}
{"train_loss": 0.22964489459991455, "global_step": 77713, "epoch": 873, "lr": 8.126518671853298e-05}
{"train_loss": 0.24569430947303772, "global_step": 77714, "epoch": 873, "lr": 8.126473430072825e-05}
{"train_loss": 0.2085857391357422, "global_step": 77715, "epoch": 873, "lr": 8.126428187872034e-05}
{"train_loss": 0.2143515646457672, "global_step": 77716, "epoch": 873, "lr": 8.126382945250936e-05}
{"train_loss": 0.262149840593338, "global_step": 77717, "epoch": 873, "lr": 8.12633770220953e-05}
{"train_loss": 0.25315600633621216, "global_step": 77718, "epoch": 873, "lr": 8.126292458747828e-05}
{"train_loss": 0.24282126128673553, "global_step": 77719, "epoch": 873, "lr": 8.126247214865834e-05}
{"train_loss": 0.28990456461906433, "global_step": 77720, "epoch": 873, "lr": 8.126201970563556e-05}
{"train_loss": 0.21880730986595154, "global_step": 77721, "epoch": 873, "lr": 8.126156725840996e-05}
{"train_loss": 0.2904490828514099, "global_step": 77722, "epoch": 873, "lr": 8.126111480698162e-05}
{"train_loss": 0.2457350343465805, "global_step": 77723, "epoch": 873, "lr": 8.126066235135061e-05}
{"train_loss": 0.20346537232398987, "global_step": 77724, "epoch": 873, "lr": 8.126020989151699e-05}
{"train_loss": 0.19149209558963776, "global_step": 77725, "epoch": 873, "lr": 8.125975742748082e-05}
{"train_loss": 0.22386786341667175, "global_step": 77726, "epoch": 873, "lr": 8.125930495924213e-05}
{"train_loss": 0.25924456119537354, "global_step": 77727, "epoch": 873, "lr": 8.125885248680103e-05}
{"train_loss": 0.16630308330059052, "global_step": 77728, "epoch": 873, "lr": 8.125840001015755e-05}
{"train_loss": 0.2695542871952057, "global_step": 77729, "epoch": 873, "lr": 8.125794752931178e-05}
{"train_loss": 0.1647689938545227, "global_step": 77730, "epoch": 873, "lr": 8.125749504426373e-05}
{"train_loss": 0.23098142445087433, "global_step": 77731, "epoch": 873, "lr": 8.125704255501351e-05}
{"train_loss": 0.2159179449081421, "global_step": 77732, "epoch": 873, "lr": 8.125659006156116e-05}
{"train_loss": 0.2084571123123169, "global_step": 77733, "epoch": 873, "lr": 8.125613756390673e-05}
{"train_loss": 0.18192286789417267, "global_step": 77734, "epoch": 873, "lr": 8.125568506205031e-05}
{"train_loss": 0.2178293764591217, "global_step": 77735, "epoch": 873, "lr": 8.125523255599193e-05}
{"train_loss": 0.24272751808166504, "global_step": 77736, "epoch": 873, "lr": 8.125478004573167e-05}
{"train_loss": 0.20846176147460938, "global_step": 77737, "epoch": 873, "lr": 8.125432753126958e-05}
{"train_loss": 0.17424458265304565, "global_step": 77738, "epoch": 873, "lr": 8.125387501260572e-05}
{"train_loss": 0.30134737491607666, "global_step": 77739, "epoch": 873, "lr": 8.125342248974019e-05}
{"train_loss": 0.2765079438686371, "global_step": 77740, "epoch": 873, "lr": 8.1252969962673e-05}
{"train_loss": 0.19571059942245483, "global_step": 77741, "epoch": 873, "lr": 8.125251743140423e-05}
{"train_loss": 0.17868119478225708, "global_step": 77742, "epoch": 873, "lr": 8.125206489593394e-05}
{"train_loss": 0.25908151268959045, "global_step": 77743, "epoch": 873, "lr": 8.125161235626217e-05}
{"train_loss": 0.14359840750694275, "global_step": 77744, "epoch": 873, "lr": 8.125115981238902e-05}
{"train_loss": 0.18265275657176971, "global_step": 77745, "epoch": 873, "lr": 8.125070726431453e-05}
{"train_loss": 0.19931507110595703, "global_step": 77746, "epoch": 873, "lr": 8.125025471203877e-05}
{"train_loss": 0.19535937905311584, "global_step": 77747, "epoch": 873, "lr": 8.124980215556179e-05}
{"train_loss": 0.2762283384799957, "global_step": 77748, "epoch": 873, "lr": 8.124934959488366e-05}
{"train_loss": 0.3042551577091217, "global_step": 77749, "epoch": 873, "lr": 8.124889703000443e-05}
{"train_loss": 0.23634260892868042, "global_step": 77750, "epoch": 873, "lr": 8.124844446092417e-05}
{"train_loss": 0.22350722551345825, "global_step": 77751, "epoch": 873, "lr": 8.124799188764293e-05}
{"train_loss": 0.20801061391830444, "global_step": 77752, "epoch": 873, "lr": 8.124753931016078e-05}
{"train_loss": 0.17742103338241577, "global_step": 77753, "epoch": 873, "lr": 8.124708672847778e-05}
{"train_loss": 0.28886449337005615, "global_step": 77754, "epoch": 873, "lr": 8.124663414259399e-05}
{"train_loss": 0.2483421117067337, "global_step": 77755, "epoch": 873, "lr": 8.124618155250948e-05}
{"train_loss": 0.17497718334197998, "global_step": 77756, "epoch": 873, "lr": 8.12457289582243e-05}
{"train_loss": 0.2698821723461151, "global_step": 77757, "epoch": 873, "lr": 8.12452763597385e-05}
{"train_loss": 0.181434765458107, "global_step": 77758, "epoch": 873, "lr": 8.124482375705216e-05}
{"train_loss": 0.20158039033412933, "global_step": 77759, "epoch": 873, "lr": 8.124437115016534e-05}
{"train_loss": 0.17783240973949432, "global_step": 77760, "epoch": 873, "lr": 8.124391853907809e-05}
{"train_loss": 0.18296001851558685, "global_step": 77761, "epoch": 873, "lr": 8.124346592379047e-05}
{"train_loss": 0.237361878156662, "global_step": 77762, "epoch": 873, "lr": 8.124301330430255e-05}
{"train_loss": 0.11407626420259476, "global_step": 77763, "epoch": 873, "lr": 8.12425606806144e-05}
{"train_loss": 0.2006722241640091, "global_step": 77764, "epoch": 873, "lr": 8.124210805272605e-05}
{"train_loss": 0.25752687454223633, "global_step": 77765, "epoch": 873, "lr": 8.12416554206376e-05}
{"train_loss": 0.17503373324871063, "global_step": 77766, "epoch": 873, "lr": 8.124120278434906e-05}
{"train_loss": 0.22099213302135468, "global_step": 77767, "epoch": 873, "lr": 8.124075014386055e-05}
{"train_loss": 0.16717864573001862, "global_step": 77768, "epoch": 873, "lr": 8.12402974991721e-05}
{"train_loss": 0.2139192372560501, "global_step": 77769, "epoch": 873, "lr": 8.123984485028375e-05}
{"train_loss": 0.22764801979064941, "global_step": 77770, "epoch": 873, "lr": 8.12393921971956e-05}
{"train_loss": 0.19777880609035492, "global_step": 77771, "epoch": 873, "lr": 8.12389395399077e-05}
{"train_loss": 0.26824769377708435, "global_step": 77772, "epoch": 873, "lr": 8.12384868784201e-05}
{"train_loss": 0.21793079376220703, "global_step": 77773, "epoch": 873, "lr": 8.123803421273286e-05}
{"train_loss": 0.23053008317947388, "global_step": 77774, "epoch": 873, "lr": 8.123758154284606e-05}
{"train_loss": 0.17958630621433258, "global_step": 77775, "epoch": 873, "lr": 8.123712886875973e-05}
{"train_loss": 0.2790532410144806, "global_step": 77776, "epoch": 873, "lr": 8.123667619047395e-05}
{"train_loss": 0.2603023052215576, "global_step": 77777, "epoch": 873, "lr": 8.12362235079888e-05}
{"train_loss": 0.24250616133213043, "global_step": 77778, "epoch": 873, "lr": 8.123577082130432e-05}
{"train_loss": 0.22373057901859283, "global_step": 77779, "epoch": 873, "lr": 8.123531813042054e-05}
{"train_loss": 0.19818441569805145, "global_step": 77780, "epoch": 873, "lr": 8.123486543533757e-05}
{"train_loss": 0.2899748384952545, "global_step": 77781, "epoch": 873, "lr": 8.123441273605546e-05}
{"train_loss": 0.16968809068202972, "global_step": 77782, "epoch": 873, "lr": 8.123396003257426e-05}
{"train_loss": 0.1751035451889038, "global_step": 77783, "epoch": 873, "lr": 8.123350732489404e-05}
{"train_loss": 0.19917452335357666, "global_step": 77784, "epoch": 873, "lr": 8.123305461301483e-05}
{"train_loss": 0.21861743751201737, "global_step": 77785, "epoch": 873, "lr": 8.123260189693675e-05, "val_loss": 4.048933506011963}
{"train_loss": 0.23446401953697205, "global_step": 77786, "epoch": 874, "lr": 8.12321491766598e-05}
{"train_loss": 0.17009764909744263, "global_step": 77787, "epoch": 874, "lr": 8.123169645218409e-05}
{"train_loss": 0.22186943888664246, "global_step": 77788, "epoch": 874, "lr": 8.123124372350964e-05}
{"train_loss": 0.22302404046058655, "global_step": 77789, "epoch": 874, "lr": 8.123079099063655e-05}
{"train_loss": 0.1564512699842453, "global_step": 77790, "epoch": 874, "lr": 8.123033825356485e-05}
{"train_loss": 0.2480870634317398, "global_step": 77791, "epoch": 874, "lr": 8.122988551229461e-05}
{"train_loss": 0.2555420696735382, "global_step": 77792, "epoch": 874, "lr": 8.122943276682589e-05}
{"train_loss": 0.22886335849761963, "global_step": 77793, "epoch": 874, "lr": 8.122898001715876e-05}
{"train_loss": 0.1323663294315338, "global_step": 77794, "epoch": 874, "lr": 8.122852726329328e-05}
{"train_loss": 0.24627594649791718, "global_step": 77795, "epoch": 874, "lr": 8.122807450522949e-05}
{"train_loss": 0.2220558524131775, "global_step": 77796, "epoch": 874, "lr": 8.122762174296748e-05}
{"train_loss": 0.22202838957309723, "global_step": 77797, "epoch": 874, "lr": 8.122716897650728e-05}
{"train_loss": 0.1228337362408638, "global_step": 77798, "epoch": 874, "lr": 8.122671620584898e-05}
{"train_loss": 0.17622365057468414, "global_step": 77799, "epoch": 874, "lr": 8.122626343099262e-05}
{"train_loss": 0.15908892452716827, "global_step": 77800, "epoch": 874, "lr": 8.122581065193829e-05}
{"train_loss": 0.2727694809436798, "global_step": 77801, "epoch": 874, "lr": 8.122535786868602e-05}
{"train_loss": 0.2723037600517273, "global_step": 77802, "epoch": 874, "lr": 8.122490508123586e-05}
{"train_loss": 0.2619919180870056, "global_step": 77803, "epoch": 874, "lr": 8.122445228958792e-05}
{"train_loss": 0.3038487732410431, "global_step": 77804, "epoch": 874, "lr": 8.122399949374222e-05}
{"train_loss": 0.2109166979789734, "global_step": 77805, "epoch": 874, "lr": 8.122354669369883e-05}
{"train_loss": 0.2073257565498352, "global_step": 77806, "epoch": 874, "lr": 8.122309388945783e-05}
{"train_loss": 0.17017753422260284, "global_step": 77807, "epoch": 874, "lr": 8.122264108101926e-05}
{"train_loss": 0.15299035608768463, "global_step": 77808, "epoch": 874, "lr": 8.122218826838319e-05}
{"train_loss": 0.22107689082622528, "global_step": 77809, "epoch": 874, "lr": 8.122173545154966e-05}
{"train_loss": 0.24121378362178802, "global_step": 77810, "epoch": 874, "lr": 8.122128263051876e-05}
{"train_loss": 0.29710739850997925, "global_step": 77811, "epoch": 874, "lr": 8.122082980529055e-05}
{"train_loss": 0.3132959008216858, "global_step": 77812, "epoch": 874, "lr": 8.122037697586506e-05}
{"train_loss": 0.2944871485233307, "global_step": 77813, "epoch": 874, "lr": 8.121992414224238e-05}
{"train_loss": 0.20995311439037323, "global_step": 77814, "epoch": 874, "lr": 8.121947130442255e-05}
{"train_loss": 0.23631444573402405, "global_step": 77815, "epoch": 874, "lr": 8.121901846240566e-05}
{"train_loss": 0.24462835490703583, "global_step": 77816, "epoch": 874, "lr": 8.121856561619175e-05}
{"train_loss": 0.21025682985782623, "global_step": 77817, "epoch": 874, "lr": 8.121811276578088e-05}
{"train_loss": 0.2207225114107132, "global_step": 77818, "epoch": 874, "lr": 8.121765991117312e-05}
{"train_loss": 0.25431209802627563, "global_step": 77819, "epoch": 874, "lr": 8.121720705236851e-05}
{"train_loss": 0.17738953232765198, "global_step": 77820, "epoch": 874, "lr": 8.121675418936714e-05}
{"train_loss": 0.14759905636310577, "global_step": 77821, "epoch": 874, "lr": 8.121630132216907e-05}
{"train_loss": 0.15458932518959045, "global_step": 77822, "epoch": 874, "lr": 8.121584845077432e-05}
{"train_loss": 0.192188560962677, "global_step": 77823, "epoch": 874, "lr": 8.1215395575183e-05}
{"train_loss": 0.2571397125720978, "global_step": 77824, "epoch": 874, "lr": 8.121494269539515e-05}
{"train_loss": 0.2136637419462204, "global_step": 77825, "epoch": 874, "lr": 8.121448981141082e-05}
{"train_loss": 0.14813242852687836, "global_step": 77826, "epoch": 874, "lr": 8.12140369232301e-05}
{"train_loss": 0.18450765311717987, "global_step": 77827, "epoch": 874, "lr": 8.121358403085301e-05}
{"train_loss": 0.2133089303970337, "global_step": 77828, "epoch": 874, "lr": 8.121313113427966e-05}
{"train_loss": 0.20349715650081635, "global_step": 77829, "epoch": 874, "lr": 8.121267823351007e-05}
{"train_loss": 0.2415834218263626, "global_step": 77830, "epoch": 874, "lr": 8.121222532854431e-05}
{"train_loss": 0.304751992225647, "global_step": 77831, "epoch": 874, "lr": 8.121177241938248e-05}
{"train_loss": 0.19230863451957703, "global_step": 77832, "epoch": 874, "lr": 8.121131950602456e-05}
{"train_loss": 0.2089359611272812, "global_step": 77833, "epoch": 874, "lr": 8.121086658847068e-05}
{"train_loss": 0.13266684114933014, "global_step": 77834, "epoch": 874, "lr": 8.121041366672089e-05}
{"train_loss": 0.27112865447998047, "global_step": 77835, "epoch": 874, "lr": 8.120996074077523e-05}
{"train_loss": 0.1809118092060089, "global_step": 77836, "epoch": 874, "lr": 8.120950781063377e-05}
{"train_loss": 0.25763580203056335, "global_step": 77837, "epoch": 874, "lr": 8.120905487629657e-05}
{"train_loss": 0.21828272938728333, "global_step": 77838, "epoch": 874, "lr": 8.12086019377637e-05}
{"train_loss": 0.25273531675338745, "global_step": 77839, "epoch": 874, "lr": 8.120814899503521e-05}
{"train_loss": 0.18585331737995148, "global_step": 77840, "epoch": 874, "lr": 8.120769604811115e-05}
{"train_loss": 0.28922414779663086, "global_step": 77841, "epoch": 874, "lr": 8.120724309699162e-05}
{"train_loss": 0.22273920476436615, "global_step": 77842, "epoch": 874, "lr": 8.120679014167665e-05}
{"train_loss": 0.1466628462076187, "global_step": 77843, "epoch": 874, "lr": 8.12063371821663e-05}
{"train_loss": 0.1732684224843979, "global_step": 77844, "epoch": 874, "lr": 8.120588421846065e-05}
{"train_loss": 0.23710061609745026, "global_step": 77845, "epoch": 874, "lr": 8.120543125055974e-05}
{"train_loss": 0.1825728714466095, "global_step": 77846, "epoch": 874, "lr": 8.120497827846364e-05}
{"train_loss": 0.1559496968984604, "global_step": 77847, "epoch": 874, "lr": 8.120452530217242e-05}
{"train_loss": 0.09906799346208572, "global_step": 77848, "epoch": 874, "lr": 8.120407232168612e-05}
{"train_loss": 0.25243517756462097, "global_step": 77849, "epoch": 874, "lr": 8.120361933700482e-05}
{"train_loss": 0.16870011389255524, "global_step": 77850, "epoch": 874, "lr": 8.120316634812857e-05}
{"train_loss": 0.23193129897117615, "global_step": 77851, "epoch": 874, "lr": 8.120271335505745e-05}
{"train_loss": 0.2302187830209732, "global_step": 77852, "epoch": 874, "lr": 8.120226035779148e-05}
{"train_loss": 0.20574046671390533, "global_step": 77853, "epoch": 874, "lr": 8.120180735633075e-05}
{"train_loss": 0.1978362500667572, "global_step": 77854, "epoch": 874, "lr": 8.120135435067533e-05}
{"train_loss": 0.164110004901886, "global_step": 77855, "epoch": 874, "lr": 8.120090134082527e-05}
{"train_loss": 0.1636553704738617, "global_step": 77856, "epoch": 874, "lr": 8.120044832678062e-05}
{"train_loss": 0.24448414146900177, "global_step": 77857, "epoch": 874, "lr": 8.119999530854146e-05}
{"train_loss": 0.1980833262205124, "global_step": 77858, "epoch": 874, "lr": 8.119954228610782e-05}
{"train_loss": 0.32534098625183105, "global_step": 77859, "epoch": 874, "lr": 8.11990892594798e-05}
{"train_loss": 0.2735695540904999, "global_step": 77860, "epoch": 874, "lr": 8.119863622865745e-05}
{"train_loss": 0.311302125453949, "global_step": 77861, "epoch": 874, "lr": 8.11981831936408e-05}
{"train_loss": 0.1745513528585434, "global_step": 77862, "epoch": 874, "lr": 8.119773015442995e-05}
{"train_loss": 0.1610983908176422, "global_step": 77863, "epoch": 874, "lr": 8.119727711102494e-05}
{"train_loss": 0.18210633099079132, "global_step": 77864, "epoch": 874, "lr": 8.119682406342584e-05}
{"train_loss": 0.15793709456920624, "global_step": 77865, "epoch": 874, "lr": 8.119637101163271e-05}
{"train_loss": 0.20068421959877014, "global_step": 77866, "epoch": 874, "lr": 8.11959179556456e-05}
{"train_loss": 0.1711542159318924, "global_step": 77867, "epoch": 874, "lr": 8.119546489546458e-05}
{"train_loss": 0.20803165435791016, "global_step": 77868, "epoch": 874, "lr": 8.119501183108972e-05}
{"train_loss": 0.19974684715270996, "global_step": 77869, "epoch": 874, "lr": 8.119455876252107e-05}
{"train_loss": 0.27701327204704285, "global_step": 77870, "epoch": 874, "lr": 8.119410568975867e-05}
{"train_loss": 0.19432292878627777, "global_step": 77871, "epoch": 874, "lr": 8.11936526128026e-05}
{"train_loss": 0.16700898110866547, "global_step": 77872, "epoch": 874, "lr": 8.119319953165295e-05}
{"train_loss": 0.12512576580047607, "global_step": 77873, "epoch": 874, "lr": 8.119274644630976e-05}
{"train_loss": 0.21112294260705455, "global_step": 77874, "epoch": 874, "lr": 8.119229335677307e-05, "val_loss": 4.072935581207275}
{"train_loss": 0.1614779531955719, "global_step": 77875, "epoch": 875, "lr": 8.119184026304296e-05}
{"train_loss": 0.17360982298851013, "global_step": 77876, "epoch": 875, "lr": 8.119138716511948e-05}
{"train_loss": 0.24540331959724426, "global_step": 77877, "epoch": 875, "lr": 8.119093406300271e-05}
{"train_loss": 0.2042534202337265, "global_step": 77878, "epoch": 875, "lr": 8.119048095669269e-05}
{"train_loss": 0.2362179011106491, "global_step": 77879, "epoch": 875, "lr": 8.119002784618949e-05}
{"train_loss": 0.13112181425094604, "global_step": 77880, "epoch": 875, "lr": 8.118957473149318e-05}
{"train_loss": 0.15013441443443298, "global_step": 77881, "epoch": 875, "lr": 8.118912161260379e-05}
{"train_loss": 0.16165374219417572, "global_step": 77882, "epoch": 875, "lr": 8.118866848952142e-05}
{"train_loss": 0.23998823761940002, "global_step": 77883, "epoch": 875, "lr": 8.118821536224613e-05}
{"train_loss": 0.21926169097423553, "global_step": 77884, "epoch": 875, "lr": 8.118776223077793e-05}
{"train_loss": 0.18295003473758698, "global_step": 77885, "epoch": 875, "lr": 8.118730909511694e-05}
{"train_loss": 0.18778522312641144, "global_step": 77886, "epoch": 875, "lr": 8.118685595526319e-05}
{"train_loss": 0.21253179013729095, "global_step": 77887, "epoch": 875, "lr": 8.118640281121675e-05}
{"train_loss": 0.23819781839847565, "global_step": 77888, "epoch": 875, "lr": 8.118594966297767e-05}
{"train_loss": 0.1339665949344635, "global_step": 77889, "epoch": 875, "lr": 8.1185496510546e-05}
{"train_loss": 0.20247389376163483, "global_step": 77890, "epoch": 875, "lr": 8.118504335392186e-05}
{"train_loss": 0.25065168738365173, "global_step": 77891, "epoch": 875, "lr": 8.118459019310524e-05}
{"train_loss": 0.10227007418870926, "global_step": 77892, "epoch": 875, "lr": 8.118413702809625e-05}
{"train_loss": 0.2114240676164627, "global_step": 77893, "epoch": 875, "lr": 8.118368385889493e-05}
{"train_loss": 0.20263671875, "global_step": 77894, "epoch": 875, "lr": 8.118323068550134e-05}
{"train_loss": 0.24912549555301666, "global_step": 77895, "epoch": 875, "lr": 8.118277750791555e-05}
{"train_loss": 0.21170365810394287, "global_step": 77896, "epoch": 875, "lr": 8.118232432613761e-05}
{"train_loss": 0.13409653306007385, "global_step": 77897, "epoch": 875, "lr": 8.118187114016759e-05}
{"train_loss": 0.2526420056819916, "global_step": 77898, "epoch": 875, "lr": 8.118141795000556e-05}
{"train_loss": 0.19071289896965027, "global_step": 77899, "epoch": 875, "lr": 8.118096475565154e-05}
{"train_loss": 0.22156919538974762, "global_step": 77900, "epoch": 875, "lr": 8.118051155710564e-05}
{"train_loss": 0.2176225781440735, "global_step": 77901, "epoch": 875, "lr": 8.11800583543679e-05}
{"train_loss": 0.30065929889678955, "global_step": 77902, "epoch": 875, "lr": 8.117960514743836e-05}
{"train_loss": 0.19709715247154236, "global_step": 77903, "epoch": 875, "lr": 8.117915193631713e-05}
{"train_loss": 0.11094334721565247, "global_step": 77904, "epoch": 875, "lr": 8.117869872100423e-05}
{"train_loss": 0.22143971920013428, "global_step": 77905, "epoch": 875, "lr": 8.117824550149975e-05}
{"train_loss": 0.24605095386505127, "global_step": 77906, "epoch": 875, "lr": 8.11777922778037e-05}
{"train_loss": 0.24734719097614288, "global_step": 77907, "epoch": 875, "lr": 8.11773390499162e-05}
{"train_loss": 0.19959349930286407, "global_step": 77908, "epoch": 875, "lr": 8.117688581783728e-05}
{"train_loss": 0.23383748531341553, "global_step": 77909, "epoch": 875, "lr": 8.117643258156701e-05}
{"train_loss": 0.223564013838768, "global_step": 77910, "epoch": 875, "lr": 8.117597934110545e-05}
{"train_loss": 0.14839939773082733, "global_step": 77911, "epoch": 875, "lr": 8.117552609645266e-05}
{"train_loss": 0.17805172502994537, "global_step": 77912, "epoch": 875, "lr": 8.11750728476087e-05}
{"train_loss": 0.22145354747772217, "global_step": 77913, "epoch": 875, "lr": 8.117461959457362e-05}
{"train_loss": 0.21821671724319458, "global_step": 77914, "epoch": 875, "lr": 8.11741663373475e-05}
{"train_loss": 0.17814475297927856, "global_step": 77915, "epoch": 875, "lr": 8.117371307593042e-05}
{"train_loss": 0.3411865830421448, "global_step": 77916, "epoch": 875, "lr": 8.117325981032238e-05}
{"train_loss": 0.16826830804347992, "global_step": 77917, "epoch": 875, "lr": 8.117280654052348e-05}
{"train_loss": 0.2208217829465866, "global_step": 77918, "epoch": 875, "lr": 8.117235326653378e-05}
{"train_loss": 0.17804276943206787, "global_step": 77919, "epoch": 875, "lr": 8.117189998835335e-05}
{"train_loss": 0.228993222117424, "global_step": 77920, "epoch": 875, "lr": 8.117144670598222e-05}
{"train_loss": 0.25877225399017334, "global_step": 77921, "epoch": 875, "lr": 8.117099341942047e-05}
{"train_loss": 0.1869063824415207, "global_step": 77922, "epoch": 875, "lr": 8.117054012866819e-05}
{"train_loss": 0.22574742138385773, "global_step": 77923, "epoch": 875, "lr": 8.117008683372537e-05}
{"train_loss": 0.24602250754833221, "global_step": 77924, "epoch": 875, "lr": 8.116963353459213e-05}
{"train_loss": 0.21159948408603668, "global_step": 77925, "epoch": 875, "lr": 8.11691802312685e-05}
{"train_loss": 0.19656424224376678, "global_step": 77926, "epoch": 875, "lr": 8.116872692375457e-05}
{"train_loss": 0.23701533675193787, "global_step": 77927, "epoch": 875, "lr": 8.116827361205037e-05}
{"train_loss": 0.23313309252262115, "global_step": 77928, "epoch": 875, "lr": 8.1167820296156e-05}
{"train_loss": 0.20178382098674774, "global_step": 77929, "epoch": 875, "lr": 8.116736697607148e-05}
{"train_loss": 0.32979393005371094, "global_step": 77930, "epoch": 875, "lr": 8.116691365179687e-05}
{"train_loss": 0.2412947118282318, "global_step": 77931, "epoch": 875, "lr": 8.116646032333228e-05}
{"train_loss": 0.2191285490989685, "global_step": 77932, "epoch": 875, "lr": 8.116600699067773e-05}
{"train_loss": 0.21708941459655762, "global_step": 77933, "epoch": 875, "lr": 8.116555365383328e-05}
{"train_loss": 0.3694993555545807, "global_step": 77934, "epoch": 875, "lr": 8.1165100312799e-05}
{"train_loss": 0.18304763734340668, "global_step": 77935, "epoch": 875, "lr": 8.116464696757497e-05}
{"train_loss": 0.16733647882938385, "global_step": 77936, "epoch": 875, "lr": 8.116419361816123e-05}
{"train_loss": 0.15530182421207428, "global_step": 77937, "epoch": 875, "lr": 8.116374026455782e-05}
{"train_loss": 0.2752498984336853, "global_step": 77938, "epoch": 875, "lr": 8.116328690676485e-05}
{"train_loss": 0.31606847047805786, "global_step": 77939, "epoch": 875, "lr": 8.116283354478234e-05}
{"train_loss": 0.20490874350070953, "global_step": 77940, "epoch": 875, "lr": 8.116238017861037e-05}
{"train_loss": 0.20499800145626068, "global_step": 77941, "epoch": 875, "lr": 8.116192680824901e-05}
{"train_loss": 0.2677248418331146, "global_step": 77942, "epoch": 875, "lr": 8.11614734336983e-05}
{"train_loss": 0.19004212319850922, "global_step": 77943, "epoch": 875, "lr": 8.116102005495831e-05}
{"train_loss": 0.21114841103553772, "global_step": 77944, "epoch": 875, "lr": 8.116056667202909e-05}
{"train_loss": 0.3123345375061035, "global_step": 77945, "epoch": 875, "lr": 8.116011328491074e-05}
{"train_loss": 0.11940845847129822, "global_step": 77946, "epoch": 875, "lr": 8.115965989360326e-05}
{"train_loss": 0.16313977539539337, "global_step": 77947, "epoch": 875, "lr": 8.115920649810677e-05}
{"train_loss": 0.20536482334136963, "global_step": 77948, "epoch": 875, "lr": 8.11587530984213e-05}
{"train_loss": 0.22303986549377441, "global_step": 77949, "epoch": 875, "lr": 8.115829969454689e-05}
{"train_loss": 0.2010711133480072, "global_step": 77950, "epoch": 875, "lr": 8.115784628648365e-05}
{"train_loss": 0.17687609791755676, "global_step": 77951, "epoch": 875, "lr": 8.115739287423161e-05}
{"train_loss": 0.19584347307682037, "global_step": 77952, "epoch": 875, "lr": 8.115693945779085e-05}
{"train_loss": 0.20054763555526733, "global_step": 77953, "epoch": 875, "lr": 8.115648603716141e-05}
{"train_loss": 0.2642097473144531, "global_step": 77954, "epoch": 875, "lr": 8.115603261234336e-05}
{"train_loss": 0.2505809962749481, "global_step": 77955, "epoch": 875, "lr": 8.115557918333677e-05}
{"train_loss": 0.23329266905784607, "global_step": 77956, "epoch": 875, "lr": 8.115512575014167e-05}
{"train_loss": 0.1742696464061737, "global_step": 77957, "epoch": 875, "lr": 8.115467231275815e-05}
{"train_loss": 0.25222569704055786, "global_step": 77958, "epoch": 875, "lr": 8.115421887118629e-05}
{"train_loss": 0.2105707973241806, "global_step": 77959, "epoch": 875, "lr": 8.11537654254261e-05}
{"train_loss": 0.17613300681114197, "global_step": 77960, "epoch": 875, "lr": 8.115331197547766e-05}
{"train_loss": 0.1687978059053421, "global_step": 77961, "epoch": 875, "lr": 8.115285852134105e-05}
{"train_loss": 0.17353056371212006, "global_step": 77962, "epoch": 875, "lr": 8.115240506301632e-05}
{"train_loss": 0.21220224266976453, "global_step": 77963, "epoch": 875, "lr": 8.115195160050351e-05, "val_loss": 4.172711372375488, "train_action_mse_error": 31.038070678710938}
{"train_loss": 0.2814055383205414, "global_step": 77964, "epoch": 876, "lr": 8.115149813380272e-05}
{"train_loss": 0.3162016272544861, "global_step": 77965, "epoch": 876, "lr": 8.115104466291399e-05}
{"train_loss": 0.2399229109287262, "global_step": 77966, "epoch": 876, "lr": 8.115059118783737e-05}
{"train_loss": 0.21010646224021912, "global_step": 77967, "epoch": 876, "lr": 8.115013770857293e-05}
{"train_loss": 0.2599808871746063, "global_step": 77968, "epoch": 876, "lr": 8.114968422512075e-05}
{"train_loss": 0.2289513200521469, "global_step": 77969, "epoch": 876, "lr": 8.114923073748087e-05}
{"train_loss": 0.19827499985694885, "global_step": 77970, "epoch": 876, "lr": 8.114877724565337e-05}
{"train_loss": 0.18921416997909546, "global_step": 77971, "epoch": 876, "lr": 8.114832374963828e-05}
{"train_loss": 0.19686561822891235, "global_step": 77972, "epoch": 876, "lr": 8.114787024943567e-05}
{"train_loss": 0.1973060667514801, "global_step": 77973, "epoch": 876, "lr": 8.114741674504562e-05}
{"train_loss": 0.21091538667678833, "global_step": 77974, "epoch": 876, "lr": 8.114696323646818e-05}
{"train_loss": 0.15687887370586395, "global_step": 77975, "epoch": 876, "lr": 8.114650972370339e-05}
{"train_loss": 0.1727638840675354, "global_step": 77976, "epoch": 876, "lr": 8.114605620675136e-05}
{"train_loss": 0.21055012941360474, "global_step": 77977, "epoch": 876, "lr": 8.11456026856121e-05}
{"train_loss": 0.1597425639629364, "global_step": 77978, "epoch": 876, "lr": 8.114514916028571e-05}
{"train_loss": 0.16864094138145447, "global_step": 77979, "epoch": 876, "lr": 8.114469563077223e-05}
{"train_loss": 0.24424618482589722, "global_step": 77980, "epoch": 876, "lr": 8.114424209707173e-05}
{"train_loss": 0.14331713318824768, "global_step": 77981, "epoch": 876, "lr": 8.114378855918424e-05}
{"train_loss": 0.1847383975982666, "global_step": 77982, "epoch": 876, "lr": 8.114333501710987e-05}
{"train_loss": 0.23533032834529877, "global_step": 77983, "epoch": 876, "lr": 8.114288147084865e-05}
{"train_loss": 0.23800751566886902, "global_step": 77984, "epoch": 876, "lr": 8.114242792040066e-05}
{"train_loss": 0.23808592557907104, "global_step": 77985, "epoch": 876, "lr": 8.114197436576594e-05}
{"train_loss": 0.1984269767999649, "global_step": 77986, "epoch": 876, "lr": 8.114152080694457e-05}
{"train_loss": 0.1966216117143631, "global_step": 77987, "epoch": 876, "lr": 8.11410672439366e-05}
{"train_loss": 0.1782371997833252, "global_step": 77988, "epoch": 876, "lr": 8.114061367674209e-05}
{"train_loss": 0.23147964477539062, "global_step": 77989, "epoch": 876, "lr": 8.11401601053611e-05}
{"train_loss": 0.26649922132492065, "global_step": 77990, "epoch": 876, "lr": 8.11397065297937e-05}
{"train_loss": 0.21407753229141235, "global_step": 77991, "epoch": 876, "lr": 8.113925295003994e-05}
{"train_loss": 0.20142781734466553, "global_step": 77992, "epoch": 876, "lr": 8.113879936609989e-05}
{"train_loss": 0.1421208381652832, "global_step": 77993, "epoch": 876, "lr": 8.113834577797362e-05}
{"train_loss": 0.2072363942861557, "global_step": 77994, "epoch": 876, "lr": 8.113789218566117e-05}
{"train_loss": 0.25403034687042236, "global_step": 77995, "epoch": 876, "lr": 8.113743858916261e-05}
{"train_loss": 0.13794532418251038, "global_step": 77996, "epoch": 876, "lr": 8.113698498847799e-05}
{"train_loss": 0.0816701278090477, "global_step": 77997, "epoch": 876, "lr": 8.11365313836074e-05}
{"train_loss": 0.15428853034973145, "global_step": 77998, "epoch": 876, "lr": 8.113607777455086e-05}
{"train_loss": 0.168123260140419, "global_step": 77999, "epoch": 876, "lr": 8.113562416130847e-05}
{"train_loss": 0.22016243636608124, "global_step": 78000, "epoch": 876, "lr": 8.113517054388026e-05}
{"train_loss": 0.218867227435112, "global_step": 78001, "epoch": 876, "lr": 8.113471692226633e-05}
{"train_loss": 0.19368362426757812, "global_step": 78002, "epoch": 876, "lr": 8.11342632964667e-05}
{"train_loss": 0.1897205263376236, "global_step": 78003, "epoch": 876, "lr": 8.113380966648147e-05}
{"train_loss": 0.16606451570987701, "global_step": 78004, "epoch": 876, "lr": 8.113335603231065e-05}
{"train_loss": 0.18970803916454315, "global_step": 78005, "epoch": 876, "lr": 8.113290239395434e-05}
{"train_loss": 0.19196303188800812, "global_step": 78006, "epoch": 876, "lr": 8.11324487514126e-05}
{"train_loss": 0.2115698754787445, "global_step": 78007, "epoch": 876, "lr": 8.113199510468548e-05}
{"train_loss": 0.22554121911525726, "global_step": 78008, "epoch": 876, "lr": 8.113154145377303e-05}
{"train_loss": 0.14079777896404266, "global_step": 78009, "epoch": 876, "lr": 8.113108779867533e-05}
{"train_loss": 0.2284804731607437, "global_step": 78010, "epoch": 876, "lr": 8.113063413939244e-05}
{"train_loss": 0.1468113660812378, "global_step": 78011, "epoch": 876, "lr": 8.113018047592441e-05}
{"train_loss": 0.19520853459835052, "global_step": 78012, "epoch": 876, "lr": 8.112972680827131e-05}
{"train_loss": 0.17363473773002625, "global_step": 78013, "epoch": 876, "lr": 8.112927313643319e-05}
{"train_loss": 0.20211844146251678, "global_step": 78014, "epoch": 876, "lr": 8.112881946041013e-05}
{"train_loss": 0.16327981650829315, "global_step": 78015, "epoch": 876, "lr": 8.112836578020217e-05}
{"train_loss": 0.12533289194107056, "global_step": 78016, "epoch": 876, "lr": 8.112791209580939e-05}
{"train_loss": 0.19409441947937012, "global_step": 78017, "epoch": 876, "lr": 8.112745840723182e-05}
{"train_loss": 0.2556041479110718, "global_step": 78018, "epoch": 876, "lr": 8.112700471446956e-05}
{"train_loss": 0.20871664583683014, "global_step": 78019, "epoch": 876, "lr": 8.112655101752266e-05}
{"train_loss": 0.19449302554130554, "global_step": 78020, "epoch": 876, "lr": 8.112609731639116e-05}
{"train_loss": 0.19114358723163605, "global_step": 78021, "epoch": 876, "lr": 8.112564361107515e-05}
{"train_loss": 0.2591087222099304, "global_step": 78022, "epoch": 876, "lr": 8.112518990157467e-05}
{"train_loss": 0.12983137369155884, "global_step": 78023, "epoch": 876, "lr": 8.112473618788979e-05}
{"train_loss": 0.1165563091635704, "global_step": 78024, "epoch": 876, "lr": 8.112428247002055e-05}
{"train_loss": 0.17147570848464966, "global_step": 78025, "epoch": 876, "lr": 8.112382874796705e-05}
{"train_loss": 0.11652159690856934, "global_step": 78026, "epoch": 876, "lr": 8.112337502172933e-05}
{"train_loss": 0.17199939489364624, "global_step": 78027, "epoch": 876, "lr": 8.112292129130744e-05}
{"train_loss": 0.189670130610466, "global_step": 78028, "epoch": 876, "lr": 8.112246755670146e-05}
{"train_loss": 0.24800577759742737, "global_step": 78029, "epoch": 876, "lr": 8.112201381791144e-05}
{"train_loss": 0.21151089668273926, "global_step": 78030, "epoch": 876, "lr": 8.112156007493745e-05}
{"train_loss": 0.17262375354766846, "global_step": 78031, "epoch": 876, "lr": 8.112110632777955e-05}
{"train_loss": 0.18354636430740356, "global_step": 78032, "epoch": 876, "lr": 8.112065257643779e-05}
{"train_loss": 0.185048907995224, "global_step": 78033, "epoch": 876, "lr": 8.112019882091224e-05}
{"train_loss": 0.21533936262130737, "global_step": 78034, "epoch": 876, "lr": 8.111974506120294e-05}
{"train_loss": 0.1867774873971939, "global_step": 78035, "epoch": 876, "lr": 8.111929129731e-05}
{"train_loss": 0.1382230520248413, "global_step": 78036, "epoch": 876, "lr": 8.111883752923343e-05}
{"train_loss": 0.23303554952144623, "global_step": 78037, "epoch": 876, "lr": 8.111838375697332e-05}
{"train_loss": 0.14117632806301117, "global_step": 78038, "epoch": 876, "lr": 8.111792998052972e-05}
{"train_loss": 0.2493087500333786, "global_step": 78039, "epoch": 876, "lr": 8.111747619990269e-05}
{"train_loss": 0.23922398686408997, "global_step": 78040, "epoch": 876, "lr": 8.111702241509229e-05}
{"train_loss": 0.18261733651161194, "global_step": 78041, "epoch": 876, "lr": 8.11165686260986e-05}
{"train_loss": 0.20243190228939056, "global_step": 78042, "epoch": 876, "lr": 8.111611483292167e-05}
{"train_loss": 0.15735870599746704, "global_step": 78043, "epoch": 876, "lr": 8.111566103556155e-05}
{"train_loss": 0.1729317456483841, "global_step": 78044, "epoch": 876, "lr": 8.11152072340183e-05}
{"train_loss": 0.25821399688720703, "global_step": 78045, "epoch": 876, "lr": 8.1114753428292e-05}
{"train_loss": 0.1642557829618454, "global_step": 78046, "epoch": 876, "lr": 8.111429961838269e-05}
{"train_loss": 0.16917327046394348, "global_step": 78047, "epoch": 876, "lr": 8.111384580429045e-05}
{"train_loss": 0.18866482377052307, "global_step": 78048, "epoch": 876, "lr": 8.111339198601534e-05}
{"train_loss": 0.15261811017990112, "global_step": 78049, "epoch": 876, "lr": 8.111293816355741e-05}
{"train_loss": 0.24013185501098633, "global_step": 78050, "epoch": 876, "lr": 8.111248433691672e-05}
{"train_loss": 0.19440893828868866, "global_step": 78051, "epoch": 876, "lr": 8.111203050609333e-05}
{"train_loss": 0.19414978792493262, "global_step": 78052, "epoch": 876, "lr": 8.11115766710873e-05, "val_loss": 4.090846061706543}
{"train_loss": 0.2443246990442276, "global_step": 78053, "epoch": 877, "lr": 8.11111228318987e-05}
{"train_loss": 0.16508811712265015, "global_step": 78054, "epoch": 877, "lr": 8.111066898852761e-05}
{"train_loss": 0.23256215453147888, "global_step": 78055, "epoch": 877, "lr": 8.111021514097406e-05}
{"train_loss": 0.1765999048948288, "global_step": 78056, "epoch": 877, "lr": 8.110976128923812e-05}
{"train_loss": 0.22024495899677277, "global_step": 78057, "epoch": 877, "lr": 8.110930743331985e-05}
{"train_loss": 0.15296384692192078, "global_step": 78058, "epoch": 877, "lr": 8.110885357321932e-05}
{"train_loss": 0.1474834531545639, "global_step": 78059, "epoch": 877, "lr": 8.110839970893657e-05}
{"train_loss": 0.1402011662721634, "global_step": 78060, "epoch": 877, "lr": 8.110794584047168e-05}
{"train_loss": 0.26716122031211853, "global_step": 78061, "epoch": 877, "lr": 8.11074919678247e-05}
{"train_loss": 0.24448265135288239, "global_step": 78062, "epoch": 877, "lr": 8.110703809099571e-05}
{"train_loss": 0.22902056574821472, "global_step": 78063, "epoch": 877, "lr": 8.110658420998476e-05}
{"train_loss": 0.13135862350463867, "global_step": 78064, "epoch": 877, "lr": 8.11061303247919e-05}
{"train_loss": 0.17977336049079895, "global_step": 78065, "epoch": 877, "lr": 8.11056764354172e-05}
{"train_loss": 0.16329744458198547, "global_step": 78066, "epoch": 877, "lr": 8.110522254186073e-05}
{"train_loss": 0.16969133913516998, "global_step": 78067, "epoch": 877, "lr": 8.110476864412253e-05}
{"train_loss": 0.16479283571243286, "global_step": 78068, "epoch": 877, "lr": 8.110431474220268e-05}
{"train_loss": 0.14069946110248566, "global_step": 78069, "epoch": 877, "lr": 8.110386083610123e-05}
{"train_loss": 0.21690843999385834, "global_step": 78070, "epoch": 877, "lr": 8.110340692581824e-05}
{"train_loss": 0.21396900713443756, "global_step": 78071, "epoch": 877, "lr": 8.110295301135378e-05}
{"train_loss": 0.2130746990442276, "global_step": 78072, "epoch": 877, "lr": 8.11024990927079e-05}
{"train_loss": 0.17045877873897552, "global_step": 78073, "epoch": 877, "lr": 8.110204516988068e-05}
{"train_loss": 0.2237996608018875, "global_step": 78074, "epoch": 877, "lr": 8.110159124287217e-05}
{"train_loss": 0.2645004987716675, "global_step": 78075, "epoch": 877, "lr": 8.110113731168243e-05}
{"train_loss": 0.23339729011058807, "global_step": 78076, "epoch": 877, "lr": 8.110068337631152e-05}
{"train_loss": 0.17889797687530518, "global_step": 78077, "epoch": 877, "lr": 8.11002294367595e-05}
{"train_loss": 0.09503840655088425, "global_step": 78078, "epoch": 877, "lr": 8.109977549302643e-05}
{"train_loss": 0.19975128769874573, "global_step": 78079, "epoch": 877, "lr": 8.109932154511237e-05}
{"train_loss": 0.20326173305511475, "global_step": 78080, "epoch": 877, "lr": 8.109886759301741e-05}
{"train_loss": 0.2367478311061859, "global_step": 78081, "epoch": 877, "lr": 8.109841363674156e-05}
{"train_loss": 0.2532159984111786, "global_step": 78082, "epoch": 877, "lr": 8.109795967628492e-05}
{"train_loss": 0.22368451952934265, "global_step": 78083, "epoch": 877, "lr": 8.109750571164754e-05}
{"train_loss": 0.21408605575561523, "global_step": 78084, "epoch": 877, "lr": 8.109705174282946e-05}
{"train_loss": 0.22886569797992706, "global_step": 78085, "epoch": 877, "lr": 8.109659776983078e-05}
{"train_loss": 0.15696246922016144, "global_step": 78086, "epoch": 877, "lr": 8.109614379265153e-05}
{"train_loss": 0.26090407371520996, "global_step": 78087, "epoch": 877, "lr": 8.10956898112918e-05}
{"train_loss": 0.2268635779619217, "global_step": 78088, "epoch": 877, "lr": 8.109523582575162e-05}
{"train_loss": 0.2532271444797516, "global_step": 78089, "epoch": 877, "lr": 8.109478183603105e-05}
{"train_loss": 0.20163056254386902, "global_step": 78090, "epoch": 877, "lr": 8.109432784213019e-05}
{"train_loss": 0.17820873856544495, "global_step": 78091, "epoch": 877, "lr": 8.109387384404907e-05}
{"train_loss": 0.17489449679851532, "global_step": 78092, "epoch": 877, "lr": 8.109341984178775e-05}
{"train_loss": 0.17396293580532074, "global_step": 78093, "epoch": 877, "lr": 8.109296583534631e-05}
{"train_loss": 0.21595549583435059, "global_step": 78094, "epoch": 877, "lr": 8.109251182472479e-05}
{"train_loss": 0.21999314427375793, "global_step": 78095, "epoch": 877, "lr": 8.109205780992327e-05}
{"train_loss": 0.19338704645633698, "global_step": 78096, "epoch": 877, "lr": 8.10916037909418e-05}
{"train_loss": 0.1732761710882187, "global_step": 78097, "epoch": 877, "lr": 8.109114976778045e-05}
{"train_loss": 0.1967087686061859, "global_step": 78098, "epoch": 877, "lr": 8.109069574043926e-05}
{"train_loss": 0.2227471023797989, "global_step": 78099, "epoch": 877, "lr": 8.10902417089183e-05}
{"train_loss": 0.23595906794071198, "global_step": 78100, "epoch": 877, "lr": 8.108978767321765e-05}
{"train_loss": 0.28411343693733215, "global_step": 78101, "epoch": 877, "lr": 8.108933363333736e-05}
{"train_loss": 0.19207920134067535, "global_step": 78102, "epoch": 877, "lr": 8.108887958927747e-05}
{"train_loss": 0.158858522772789, "global_step": 78103, "epoch": 877, "lr": 8.108842554103807e-05}
{"train_loss": 0.20974965393543243, "global_step": 78104, "epoch": 877, "lr": 8.108797148861921e-05}
{"train_loss": 0.2411888837814331, "global_step": 78105, "epoch": 877, "lr": 8.108751743202094e-05}
{"train_loss": 0.19773663580417633, "global_step": 78106, "epoch": 877, "lr": 8.108706337124336e-05}
{"train_loss": 0.1063699871301651, "global_step": 78107, "epoch": 877, "lr": 8.108660930628647e-05}
{"train_loss": 0.1562841236591339, "global_step": 78108, "epoch": 877, "lr": 8.108615523715038e-05}
{"train_loss": 0.15297959744930267, "global_step": 78109, "epoch": 877, "lr": 8.108570116383514e-05}
{"train_loss": 0.20467768609523773, "global_step": 78110, "epoch": 877, "lr": 8.10852470863408e-05}
{"train_loss": 0.30148640275001526, "global_step": 78111, "epoch": 877, "lr": 8.108479300466742e-05}
{"train_loss": 0.2664939761161804, "global_step": 78112, "epoch": 877, "lr": 8.108433891881509e-05}
{"train_loss": 0.22610874474048615, "global_step": 78113, "epoch": 877, "lr": 8.108388482878385e-05}
{"train_loss": 0.18668334186077118, "global_step": 78114, "epoch": 877, "lr": 8.108343073457374e-05}
{"train_loss": 0.17094826698303223, "global_step": 78115, "epoch": 877, "lr": 8.108297663618485e-05}
{"train_loss": 0.26386478543281555, "global_step": 78116, "epoch": 877, "lr": 8.108252253361723e-05}
{"train_loss": 0.16358813643455505, "global_step": 78117, "epoch": 877, "lr": 8.108206842687093e-05}
{"train_loss": 0.30491921305656433, "global_step": 78118, "epoch": 877, "lr": 8.108161431594606e-05}
{"train_loss": 0.2923547029495239, "global_step": 78119, "epoch": 877, "lr": 8.108116020084261e-05}
{"train_loss": 0.3202454447746277, "global_step": 78120, "epoch": 877, "lr": 8.10807060815607e-05}
{"train_loss": 0.22019082307815552, "global_step": 78121, "epoch": 877, "lr": 8.108025195810036e-05}
{"train_loss": 0.21526341140270233, "global_step": 78122, "epoch": 877, "lr": 8.107979783046166e-05}
{"train_loss": 0.22087761759757996, "global_step": 78123, "epoch": 877, "lr": 8.107934369864465e-05}
{"train_loss": 0.12920692563056946, "global_step": 78124, "epoch": 877, "lr": 8.107888956264943e-05}
{"train_loss": 0.29700133204460144, "global_step": 78125, "epoch": 877, "lr": 8.1078435422476e-05}
{"train_loss": 0.18009908497333527, "global_step": 78126, "epoch": 877, "lr": 8.107798127812445e-05}
{"train_loss": 0.175862655043602, "global_step": 78127, "epoch": 877, "lr": 8.107752712959487e-05}
{"train_loss": 0.20278102159500122, "global_step": 78128, "epoch": 877, "lr": 8.107707297688729e-05}
{"train_loss": 0.2247535139322281, "global_step": 78129, "epoch": 877, "lr": 8.107661882000177e-05}
{"train_loss": 0.26895612478256226, "global_step": 78130, "epoch": 877, "lr": 8.107616465893837e-05}
{"train_loss": 0.19496147334575653, "global_step": 78131, "epoch": 877, "lr": 8.107571049369718e-05}
{"train_loss": 0.24981343746185303, "global_step": 78132, "epoch": 877, "lr": 8.107525632427821e-05}
{"train_loss": 0.22358891367912292, "global_step": 78133, "epoch": 877, "lr": 8.107480215068157e-05}
{"train_loss": 0.2042810469865799, "global_step": 78134, "epoch": 877, "lr": 8.107434797290731e-05}
{"train_loss": 0.19262813031673431, "global_step": 78135, "epoch": 877, "lr": 8.107389379095547e-05}
{"train_loss": 0.244984969496727, "global_step": 78136, "epoch": 877, "lr": 8.107343960482612e-05}
{"train_loss": 0.20301692187786102, "global_step": 78137, "epoch": 877, "lr": 8.107298541451934e-05}
{"train_loss": 0.15797767043113708, "global_step": 78138, "epoch": 877, "lr": 8.107253122003516e-05}
{"train_loss": 0.281364768743515, "global_step": 78139, "epoch": 877, "lr": 8.107207702137367e-05}
{"train_loss": 0.1614173948764801, "global_step": 78140, "epoch": 877, "lr": 8.107162281853492e-05}
{"train_loss": 0.20745990877405981, "global_step": 78141, "epoch": 877, "lr": 8.107116861151895e-05, "val_loss": 4.068841457366943}
{"train_loss": 0.23961973190307617, "global_step": 78142, "epoch": 878, "lr": 8.107071440032584e-05}
{"train_loss": 0.2040974199771881, "global_step": 78143, "epoch": 878, "lr": 8.107026018495566e-05}
{"train_loss": 0.2390974462032318, "global_step": 78144, "epoch": 878, "lr": 8.106980596540846e-05}
{"train_loss": 0.20325961709022522, "global_step": 78145, "epoch": 878, "lr": 8.10693517416843e-05}
{"train_loss": 0.16522856056690216, "global_step": 78146, "epoch": 878, "lr": 8.106889751378325e-05}
{"train_loss": 0.2033395618200302, "global_step": 78147, "epoch": 878, "lr": 8.106844328170537e-05}
{"train_loss": 0.1650698184967041, "global_step": 78148, "epoch": 878, "lr": 8.106798904545072e-05}
{"train_loss": 0.20462244749069214, "global_step": 78149, "epoch": 878, "lr": 8.106753480501935e-05}
{"train_loss": 0.22665807604789734, "global_step": 78150, "epoch": 878, "lr": 8.106708056041133e-05}
{"train_loss": 0.19715651869773865, "global_step": 78151, "epoch": 878, "lr": 8.106662631162672e-05}
{"train_loss": 0.20090477168560028, "global_step": 78152, "epoch": 878, "lr": 8.106617205866557e-05}
{"train_loss": 0.17224591970443726, "global_step": 78153, "epoch": 878, "lr": 8.106571780152797e-05}
{"train_loss": 0.15573905408382416, "global_step": 78154, "epoch": 878, "lr": 8.106526354021395e-05}
{"train_loss": 0.28180941939353943, "global_step": 78155, "epoch": 878, "lr": 8.106480927472359e-05}
{"train_loss": 0.23632250726222992, "global_step": 78156, "epoch": 878, "lr": 8.106435500505695e-05}
{"train_loss": 0.24973182380199432, "global_step": 78157, "epoch": 878, "lr": 8.106390073121408e-05}
{"train_loss": 0.20387466251850128, "global_step": 78158, "epoch": 878, "lr": 8.106344645319503e-05}
{"train_loss": 0.18609008193016052, "global_step": 78159, "epoch": 878, "lr": 8.10629921709999e-05}
{"train_loss": 0.2765476703643799, "global_step": 78160, "epoch": 878, "lr": 8.106253788462873e-05}
{"train_loss": 0.1826113611459732, "global_step": 78161, "epoch": 878, "lr": 8.106208359408157e-05}
{"train_loss": 0.2163063883781433, "global_step": 78162, "epoch": 878, "lr": 8.10616292993585e-05}
{"train_loss": 0.17171800136566162, "global_step": 78163, "epoch": 878, "lr": 8.106117500045956e-05}
{"train_loss": 0.18549354374408722, "global_step": 78164, "epoch": 878, "lr": 8.106072069738484e-05}
{"train_loss": 0.2056475132703781, "global_step": 78165, "epoch": 878, "lr": 8.106026639013437e-05}
{"train_loss": 0.12918820977210999, "global_step": 78166, "epoch": 878, "lr": 8.105981207870823e-05}
{"train_loss": 0.21641632914543152, "global_step": 78167, "epoch": 878, "lr": 8.105935776310648e-05}
{"train_loss": 0.1811063140630722, "global_step": 78168, "epoch": 878, "lr": 8.105890344332918e-05}
{"train_loss": 0.19600418210029602, "global_step": 78169, "epoch": 878, "lr": 8.10584491193764e-05}
{"train_loss": 0.19603359699249268, "global_step": 78170, "epoch": 878, "lr": 8.105799479124816e-05}
{"train_loss": 0.24359586834907532, "global_step": 78171, "epoch": 878, "lr": 8.105754045894459e-05}
{"train_loss": 0.15928590297698975, "global_step": 78172, "epoch": 878, "lr": 8.105708612246568e-05}
{"train_loss": 0.11265666782855988, "global_step": 78173, "epoch": 878, "lr": 8.105663178181154e-05}
{"train_loss": 0.22385947406291962, "global_step": 78174, "epoch": 878, "lr": 8.10561774369822e-05}
{"train_loss": 0.1890491396188736, "global_step": 78175, "epoch": 878, "lr": 8.105572308797773e-05}
{"train_loss": 0.18049171566963196, "global_step": 78176, "epoch": 878, "lr": 8.105526873479821e-05}
{"train_loss": 0.1784496009349823, "global_step": 78177, "epoch": 878, "lr": 8.105481437744368e-05}
{"train_loss": 0.1739821583032608, "global_step": 78178, "epoch": 878, "lr": 8.105436001591422e-05}
{"train_loss": 0.21663983166217804, "global_step": 78179, "epoch": 878, "lr": 8.105390565020987e-05}
{"train_loss": 0.12964504957199097, "global_step": 78180, "epoch": 878, "lr": 8.10534512803307e-05}
{"train_loss": 0.13689760863780975, "global_step": 78181, "epoch": 878, "lr": 8.105299690627679e-05}
{"train_loss": 0.260505735874176, "global_step": 78182, "epoch": 878, "lr": 8.105254252804817e-05}
{"train_loss": 0.13729450106620789, "global_step": 78183, "epoch": 878, "lr": 8.105208814564491e-05}
{"train_loss": 0.1782110631465912, "global_step": 78184, "epoch": 878, "lr": 8.105163375906708e-05}
{"train_loss": 0.1379893571138382, "global_step": 78185, "epoch": 878, "lr": 8.105117936831475e-05}
{"train_loss": 0.1923622339963913, "global_step": 78186, "epoch": 878, "lr": 8.105072497338795e-05}
{"train_loss": 0.1588369607925415, "global_step": 78187, "epoch": 878, "lr": 8.105027057428676e-05}
{"train_loss": 0.22620171308517456, "global_step": 78188, "epoch": 878, "lr": 8.104981617101124e-05}
{"train_loss": 0.2057095319032669, "global_step": 78189, "epoch": 878, "lr": 8.104936176356147e-05}
{"train_loss": 0.15632429718971252, "global_step": 78190, "epoch": 878, "lr": 8.104890735193747e-05}
{"train_loss": 0.15375927090644836, "global_step": 78191, "epoch": 878, "lr": 8.104845293613932e-05}
{"train_loss": 0.2035266011953354, "global_step": 78192, "epoch": 878, "lr": 8.104799851616709e-05}
{"train_loss": 0.1371106058359146, "global_step": 78193, "epoch": 878, "lr": 8.104754409202084e-05}
{"train_loss": 0.19299130141735077, "global_step": 78194, "epoch": 878, "lr": 8.104708966370063e-05}
{"train_loss": 0.20035140216350555, "global_step": 78195, "epoch": 878, "lr": 8.10466352312065e-05}
{"train_loss": 0.2471914291381836, "global_step": 78196, "epoch": 878, "lr": 8.104618079453854e-05}
{"train_loss": 0.1855241060256958, "global_step": 78197, "epoch": 878, "lr": 8.104572635369679e-05}
{"train_loss": 0.19227375090122223, "global_step": 78198, "epoch": 878, "lr": 8.104527190868133e-05}
{"train_loss": 0.21740852296352386, "global_step": 78199, "epoch": 878, "lr": 8.10448174594922e-05}
{"train_loss": 0.16187375783920288, "global_step": 78200, "epoch": 878, "lr": 8.10443630061295e-05}
{"train_loss": 0.2637273371219635, "global_step": 78201, "epoch": 878, "lr": 8.104390854859323e-05}
{"train_loss": 0.20619440078735352, "global_step": 78202, "epoch": 878, "lr": 8.104345408688351e-05}
{"train_loss": 0.18164479732513428, "global_step": 78203, "epoch": 878, "lr": 8.104299962100037e-05}
{"train_loss": 0.1759389191865921, "global_step": 78204, "epoch": 878, "lr": 8.104254515094386e-05}
{"train_loss": 0.18335755169391632, "global_step": 78205, "epoch": 878, "lr": 8.104209067671408e-05}
{"train_loss": 0.26260805130004883, "global_step": 78206, "epoch": 878, "lr": 8.104163619831107e-05}
{"train_loss": 0.16998706758022308, "global_step": 78207, "epoch": 878, "lr": 8.104118171573487e-05}
{"train_loss": 0.19288888573646545, "global_step": 78208, "epoch": 878, "lr": 8.104072722898556e-05}
{"train_loss": 0.14781813323497772, "global_step": 78209, "epoch": 878, "lr": 8.104027273806322e-05}
{"train_loss": 0.14543220400810242, "global_step": 78210, "epoch": 878, "lr": 8.10398182429679e-05}
{"train_loss": 0.17452123761177063, "global_step": 78211, "epoch": 878, "lr": 8.103936374369965e-05}
{"train_loss": 0.17084930837154388, "global_step": 78212, "epoch": 878, "lr": 8.103890924025852e-05}
{"train_loss": 0.23414269089698792, "global_step": 78213, "epoch": 878, "lr": 8.10384547326446e-05}
{"train_loss": 0.16918295621871948, "global_step": 78214, "epoch": 878, "lr": 8.103800022085795e-05}
{"train_loss": 0.23401543498039246, "global_step": 78215, "epoch": 878, "lr": 8.103754570489859e-05}
{"train_loss": 0.22893017530441284, "global_step": 78216, "epoch": 878, "lr": 8.103709118476662e-05}
{"train_loss": 0.22117508947849274, "global_step": 78217, "epoch": 878, "lr": 8.103663666046209e-05}
{"train_loss": 0.138834148645401, "global_step": 78218, "epoch": 878, "lr": 8.103618213198507e-05}
{"train_loss": 0.1421986222267151, "global_step": 78219, "epoch": 878, "lr": 8.103572759933563e-05}
{"train_loss": 0.15097546577453613, "global_step": 78220, "epoch": 878, "lr": 8.103527306251379e-05}
{"train_loss": 0.13607941567897797, "global_step": 78221, "epoch": 878, "lr": 8.103481852151964e-05}
{"train_loss": 0.2085619866847992, "global_step": 78222, "epoch": 878, "lr": 8.103436397635323e-05}
{"train_loss": 0.22933551669120789, "global_step": 78223, "epoch": 878, "lr": 8.103390942701465e-05}
{"train_loss": 0.20589356124401093, "global_step": 78224, "epoch": 878, "lr": 8.103345487350392e-05}
{"train_loss": 0.19002023339271545, "global_step": 78225, "epoch": 878, "lr": 8.103300031582112e-05}
{"train_loss": 0.1904776394367218, "global_step": 78226, "epoch": 878, "lr": 8.103254575396633e-05}
{"train_loss": 0.1798545867204666, "global_step": 78227, "epoch": 878, "lr": 8.103209118793957e-05}
{"train_loss": 0.24599634110927582, "global_step": 78228, "epoch": 878, "lr": 8.103163661774094e-05}
{"train_loss": 0.19211556017398834, "global_step": 78229, "epoch": 878, "lr": 8.103118204337048e-05}
{"train_loss": 0.19306913319598423, "global_step": 78230, "epoch": 878, "lr": 8.103072746482825e-05, "val_loss": 4.0245537757873535}
{"train_loss": 0.14826463162899017, "global_step": 78231, "epoch": 879, "lr": 8.103027288211433e-05}
{"train_loss": 0.1269175261259079, "global_step": 78232, "epoch": 879, "lr": 8.102981829522875e-05}
{"train_loss": 0.28261181712150574, "global_step": 78233, "epoch": 879, "lr": 8.10293637041716e-05}
{"train_loss": 0.2287183403968811, "global_step": 78234, "epoch": 879, "lr": 8.102890910894292e-05}
{"train_loss": 0.14865031838417053, "global_step": 78235, "epoch": 879, "lr": 8.102845450954279e-05}
{"train_loss": 0.2719520330429077, "global_step": 78236, "epoch": 879, "lr": 8.102799990597128e-05}
{"train_loss": 0.1627102792263031, "global_step": 78237, "epoch": 879, "lr": 8.10275452982284e-05}
{"train_loss": 0.13148285448551178, "global_step": 78238, "epoch": 879, "lr": 8.102709068631426e-05}
{"train_loss": 0.19320327043533325, "global_step": 78239, "epoch": 879, "lr": 8.102663607022891e-05}
{"train_loss": 0.14832429587841034, "global_step": 78240, "epoch": 879, "lr": 8.10261814499724e-05}
{"train_loss": 0.2189832329750061, "global_step": 78241, "epoch": 879, "lr": 8.102572682554481e-05}
{"train_loss": 0.25704649090766907, "global_step": 78242, "epoch": 879, "lr": 8.102527219694618e-05}
{"train_loss": 0.25692903995513916, "global_step": 78243, "epoch": 879, "lr": 8.102481756417658e-05}
{"train_loss": 0.15988531708717346, "global_step": 78244, "epoch": 879, "lr": 8.102436292723608e-05}
{"train_loss": 0.17138509452342987, "global_step": 78245, "epoch": 879, "lr": 8.102390828612471e-05}
{"train_loss": 0.15258970856666565, "global_step": 78246, "epoch": 879, "lr": 8.102345364084258e-05}
{"train_loss": 0.14060509204864502, "global_step": 78247, "epoch": 879, "lr": 8.10229989913897e-05}
{"train_loss": 0.20922045409679413, "global_step": 78248, "epoch": 879, "lr": 8.102254433776618e-05}
{"train_loss": 0.25026437640190125, "global_step": 78249, "epoch": 879, "lr": 8.102208967997205e-05}
{"train_loss": 0.2269689440727234, "global_step": 78250, "epoch": 879, "lr": 8.102163501800737e-05}
{"train_loss": 0.2504703104496002, "global_step": 78251, "epoch": 879, "lr": 8.102118035187221e-05}
{"train_loss": 0.18841251730918884, "global_step": 78252, "epoch": 879, "lr": 8.102072568156663e-05}
{"train_loss": 0.18265631794929504, "global_step": 78253, "epoch": 879, "lr": 8.10202710070907e-05}
{"train_loss": 0.22997170686721802, "global_step": 78254, "epoch": 879, "lr": 8.101981632844446e-05}
{"train_loss": 0.22231362760066986, "global_step": 78255, "epoch": 879, "lr": 8.1019361645628e-05}
{"train_loss": 0.17931824922561646, "global_step": 78256, "epoch": 879, "lr": 8.101890695864136e-05}
{"train_loss": 0.1784244179725647, "global_step": 78257, "epoch": 879, "lr": 8.101845226748461e-05}
{"train_loss": 0.2102847695350647, "global_step": 78258, "epoch": 879, "lr": 8.10179975721578e-05}
{"train_loss": 0.15088696777820587, "global_step": 78259, "epoch": 879, "lr": 8.1017542872661e-05}
{"train_loss": 0.266154021024704, "global_step": 78260, "epoch": 879, "lr": 8.101708816899425e-05}
{"train_loss": 0.12486271560192108, "global_step": 78261, "epoch": 879, "lr": 8.101663346115766e-05}
{"train_loss": 0.15979032218456268, "global_step": 78262, "epoch": 879, "lr": 8.101617874915125e-05}
{"train_loss": 0.1465463489294052, "global_step": 78263, "epoch": 879, "lr": 8.101572403297509e-05}
{"train_loss": 0.14114314317703247, "global_step": 78264, "epoch": 879, "lr": 8.101526931262925e-05}
{"train_loss": 0.2213582694530487, "global_step": 78265, "epoch": 879, "lr": 8.101481458811378e-05}
{"train_loss": 0.20714643597602844, "global_step": 78266, "epoch": 879, "lr": 8.101435985942876e-05}
{"train_loss": 0.23124006390571594, "global_step": 78267, "epoch": 879, "lr": 8.101390512657421e-05}
{"train_loss": 0.14655162394046783, "global_step": 78268, "epoch": 879, "lr": 8.101345038955025e-05}
{"train_loss": 0.21840231120586395, "global_step": 78269, "epoch": 879, "lr": 8.10129956483569e-05}
{"train_loss": 0.1672562062740326, "global_step": 78270, "epoch": 879, "lr": 8.10125409029942e-05}
{"train_loss": 0.23727622628211975, "global_step": 78271, "epoch": 879, "lr": 8.101208615346227e-05}
{"train_loss": 0.19525352120399475, "global_step": 78272, "epoch": 879, "lr": 8.101163139976113e-05}
{"train_loss": 0.15881849825382233, "global_step": 78273, "epoch": 879, "lr": 8.101117664189088e-05}
{"train_loss": 0.18673138320446014, "global_step": 78274, "epoch": 879, "lr": 8.101072187985153e-05}
{"train_loss": 0.20670382678508759, "global_step": 78275, "epoch": 879, "lr": 8.101026711364317e-05}
{"train_loss": 0.1974407136440277, "global_step": 78276, "epoch": 879, "lr": 8.100981234326588e-05}
{"train_loss": 0.2552410960197449, "global_step": 78277, "epoch": 879, "lr": 8.100935756871968e-05}
{"train_loss": 0.1854739785194397, "global_step": 78278, "epoch": 879, "lr": 8.100890279000464e-05}
{"train_loss": 0.24516211450099945, "global_step": 78279, "epoch": 879, "lr": 8.100844800712084e-05}
{"train_loss": 0.12964823842048645, "global_step": 78280, "epoch": 879, "lr": 8.100799322006833e-05}
{"train_loss": 0.17693005502223969, "global_step": 78281, "epoch": 879, "lr": 8.100753842884718e-05}
{"train_loss": 0.18058551847934723, "global_step": 78282, "epoch": 879, "lr": 8.100708363345745e-05}
{"train_loss": 0.15903276205062866, "global_step": 78283, "epoch": 879, "lr": 8.10066288338992e-05}
{"train_loss": 0.19958284497261047, "global_step": 78284, "epoch": 879, "lr": 8.100617403017245e-05}
{"train_loss": 0.12991386651992798, "global_step": 78285, "epoch": 879, "lr": 8.100571922227733e-05}
{"train_loss": 0.118479423224926, "global_step": 78286, "epoch": 879, "lr": 8.100526441021387e-05}
{"train_loss": 0.18844138085842133, "global_step": 78287, "epoch": 879, "lr": 8.100480959398212e-05}
{"train_loss": 0.16175556182861328, "global_step": 78288, "epoch": 879, "lr": 8.100435477358215e-05}
{"train_loss": 0.19410192966461182, "global_step": 78289, "epoch": 879, "lr": 8.100389994901402e-05}
{"train_loss": 0.22922830283641815, "global_step": 78290, "epoch": 879, "lr": 8.10034451202778e-05}
{"train_loss": 0.14993564784526825, "global_step": 78291, "epoch": 879, "lr": 8.100299028737354e-05}
{"train_loss": 0.13171549141407013, "global_step": 78292, "epoch": 879, "lr": 8.10025354503013e-05}
{"train_loss": 0.22247900068759918, "global_step": 78293, "epoch": 879, "lr": 8.100208060906116e-05}
{"train_loss": 0.16654571890830994, "global_step": 78294, "epoch": 879, "lr": 8.100162576365318e-05}
{"train_loss": 0.13167348504066467, "global_step": 78295, "epoch": 879, "lr": 8.100117091407738e-05}
{"train_loss": 0.1488097459077835, "global_step": 78296, "epoch": 879, "lr": 8.100071606033386e-05}
{"train_loss": 0.17367267608642578, "global_step": 78297, "epoch": 879, "lr": 8.100026120242268e-05}
{"train_loss": 0.20577171444892883, "global_step": 78298, "epoch": 879, "lr": 8.099980634034389e-05}
{"train_loss": 0.19739040732383728, "global_step": 78299, "epoch": 879, "lr": 8.099935147409755e-05}
{"train_loss": 0.15630139410495758, "global_step": 78300, "epoch": 879, "lr": 8.099889660368373e-05}
{"train_loss": 0.17662958800792694, "global_step": 78301, "epoch": 879, "lr": 8.099844172910249e-05}
{"train_loss": 0.1845913678407669, "global_step": 78302, "epoch": 879, "lr": 8.099798685035387e-05}
{"train_loss": 0.1072452962398529, "global_step": 78303, "epoch": 879, "lr": 8.099753196743795e-05}
{"train_loss": 0.19219176471233368, "global_step": 78304, "epoch": 879, "lr": 8.09970770803548e-05}
{"train_loss": 0.17287151515483856, "global_step": 78305, "epoch": 879, "lr": 8.099662218910447e-05}
{"train_loss": 0.15780453383922577, "global_step": 78306, "epoch": 879, "lr": 8.099616729368702e-05}
{"train_loss": 0.17749042809009552, "global_step": 78307, "epoch": 879, "lr": 8.099571239410251e-05}
{"train_loss": 0.26647883653640747, "global_step": 78308, "epoch": 879, "lr": 8.099525749035101e-05}
{"train_loss": 0.2379622906446457, "global_step": 78309, "epoch": 879, "lr": 8.099480258243258e-05}
{"train_loss": 0.14613498747348785, "global_step": 78310, "epoch": 879, "lr": 8.099434767034725e-05}
{"train_loss": 0.1895606517791748, "global_step": 78311, "epoch": 879, "lr": 8.099389275409515e-05}
{"train_loss": 0.22938165068626404, "global_step": 78312, "epoch": 879, "lr": 8.099343783367627e-05}
{"train_loss": 0.16024711728096008, "global_step": 78313, "epoch": 879, "lr": 8.099298290909072e-05}
{"train_loss": 0.20481325685977936, "global_step": 78314, "epoch": 879, "lr": 8.099252798033851e-05}
{"train_loss": 0.211538627743721, "global_step": 78315, "epoch": 879, "lr": 8.099207304741976e-05}
{"train_loss": 0.16684062778949738, "global_step": 78316, "epoch": 879, "lr": 8.099161811033449e-05}
{"train_loss": 0.20943288505077362, "global_step": 78317, "epoch": 879, "lr": 8.099116316908279e-05}
{"train_loss": 0.1873856782913208, "global_step": 78318, "epoch": 879, "lr": 8.099070822366469e-05}
{"train_loss": 0.18709441040004238, "global_step": 78319, "epoch": 879, "lr": 8.099025327408028e-05, "val_loss": 4.1072797775268555}
{"train_loss": 0.2349046766757965, "global_step": 78320, "epoch": 880, "lr": 8.09897983203296e-05}
{"train_loss": 0.11865870654582977, "global_step": 78321, "epoch": 880, "lr": 8.09893433624127e-05}
{"train_loss": 0.21099485456943512, "global_step": 78322, "epoch": 880, "lr": 8.098888840032969e-05}
{"train_loss": 0.2466975450515747, "global_step": 78323, "epoch": 880, "lr": 8.098843343408058e-05}
{"train_loss": 0.17514482140541077, "global_step": 78324, "epoch": 880, "lr": 8.098797846366547e-05}
{"train_loss": 0.1523183435201645, "global_step": 78325, "epoch": 880, "lr": 8.09875234890844e-05}
{"train_loss": 0.1618645191192627, "global_step": 78326, "epoch": 880, "lr": 8.098706851033743e-05}
{"train_loss": 0.16660921275615692, "global_step": 78327, "epoch": 880, "lr": 8.098661352742462e-05}
{"train_loss": 0.23881293833255768, "global_step": 78328, "epoch": 880, "lr": 8.098615854034605e-05}
{"train_loss": 0.25209808349609375, "global_step": 78329, "epoch": 880, "lr": 8.098570354910177e-05}
{"train_loss": 0.32285788655281067, "global_step": 78330, "epoch": 880, "lr": 8.098524855369183e-05}
{"train_loss": 0.28614017367362976, "global_step": 78331, "epoch": 880, "lr": 8.098479355411631e-05}
{"train_loss": 0.15068140625953674, "global_step": 78332, "epoch": 880, "lr": 8.098433855037526e-05}
{"train_loss": 0.23380164802074432, "global_step": 78333, "epoch": 880, "lr": 8.098388354246876e-05}
{"train_loss": 0.18411539494991302, "global_step": 78334, "epoch": 880, "lr": 8.098342853039682e-05}
{"train_loss": 0.1367574781179428, "global_step": 78335, "epoch": 880, "lr": 8.098297351415955e-05}
{"train_loss": 0.16510961949825287, "global_step": 78336, "epoch": 880, "lr": 8.0982518493757e-05}
{"train_loss": 0.25336483120918274, "global_step": 78337, "epoch": 880, "lr": 8.098206346918923e-05}
{"train_loss": 0.18430984020233154, "global_step": 78338, "epoch": 880, "lr": 8.09816084404563e-05}
{"train_loss": 0.24630369246006012, "global_step": 78339, "epoch": 880, "lr": 8.098115340755827e-05}
{"train_loss": 0.1834491491317749, "global_step": 78340, "epoch": 880, "lr": 8.09806983704952e-05}
{"train_loss": 0.20127563178539276, "global_step": 78341, "epoch": 880, "lr": 8.098024332926715e-05}
{"train_loss": 0.20318937301635742, "global_step": 78342, "epoch": 880, "lr": 8.097978828387418e-05}
{"train_loss": 0.27186962962150574, "global_step": 78343, "epoch": 880, "lr": 8.097933323431635e-05}
{"train_loss": 0.17760342359542847, "global_step": 78344, "epoch": 880, "lr": 8.097887818059374e-05}
{"train_loss": 0.28890299797058105, "global_step": 78345, "epoch": 880, "lr": 8.097842312270638e-05}
{"train_loss": 0.22199565172195435, "global_step": 78346, "epoch": 880, "lr": 8.097796806065436e-05}
{"train_loss": 0.19901683926582336, "global_step": 78347, "epoch": 880, "lr": 8.097751299443775e-05}
{"train_loss": 0.26331913471221924, "global_step": 78348, "epoch": 880, "lr": 8.097705792405656e-05}
{"train_loss": 0.2193523347377777, "global_step": 78349, "epoch": 880, "lr": 8.097660284951089e-05}
{"train_loss": 0.15906693041324615, "global_step": 78350, "epoch": 880, "lr": 8.097614777080081e-05}
{"train_loss": 0.23893168568611145, "global_step": 78351, "epoch": 880, "lr": 8.097569268792635e-05}
{"train_loss": 0.22471566498279572, "global_step": 78352, "epoch": 880, "lr": 8.097523760088759e-05}
{"train_loss": 0.1596119999885559, "global_step": 78353, "epoch": 880, "lr": 8.097478250968459e-05}
{"train_loss": 0.14902207255363464, "global_step": 78354, "epoch": 880, "lr": 8.097432741431739e-05}
{"train_loss": 0.16568545997142792, "global_step": 78355, "epoch": 880, "lr": 8.097387231478609e-05}
{"train_loss": 0.1858806610107422, "global_step": 78356, "epoch": 880, "lr": 8.097341721109073e-05}
{"train_loss": 0.19969460368156433, "global_step": 78357, "epoch": 880, "lr": 8.097296210323135e-05}
{"train_loss": 0.21471787989139557, "global_step": 78358, "epoch": 880, "lr": 8.097250699120806e-05}
{"train_loss": 0.19764825701713562, "global_step": 78359, "epoch": 880, "lr": 8.097205187502088e-05}
{"train_loss": 0.18670424818992615, "global_step": 78360, "epoch": 880, "lr": 8.097159675466987e-05}
{"train_loss": 0.17631879448890686, "global_step": 78361, "epoch": 880, "lr": 8.097114163015514e-05}
{"train_loss": 0.15297681093215942, "global_step": 78362, "epoch": 880, "lr": 8.097068650147668e-05}
{"train_loss": 0.1318110227584839, "global_step": 78363, "epoch": 880, "lr": 8.097023136863462e-05}
{"train_loss": 0.22543424367904663, "global_step": 78364, "epoch": 880, "lr": 8.096977623162899e-05}
{"train_loss": 0.15709565579891205, "global_step": 78365, "epoch": 880, "lr": 8.096932109045982e-05}
{"train_loss": 0.1515820026397705, "global_step": 78366, "epoch": 880, "lr": 8.096886594512722e-05}
{"train_loss": 0.16535860300064087, "global_step": 78367, "epoch": 880, "lr": 8.096841079563125e-05}
{"train_loss": 0.15490153431892395, "global_step": 78368, "epoch": 880, "lr": 8.096795564197193e-05}
{"train_loss": 0.21427474915981293, "global_step": 78369, "epoch": 880, "lr": 8.096750048414936e-05}
{"train_loss": 0.1815507411956787, "global_step": 78370, "epoch": 880, "lr": 8.096704532216357e-05}
{"train_loss": 0.24916473031044006, "global_step": 78371, "epoch": 880, "lr": 8.096659015601467e-05}
{"train_loss": 0.15686750411987305, "global_step": 78372, "epoch": 880, "lr": 8.096613498570267e-05}
{"train_loss": 0.1573868840932846, "global_step": 78373, "epoch": 880, "lr": 8.096567981122765e-05}
{"train_loss": 0.16491422057151794, "global_step": 78374, "epoch": 880, "lr": 8.096522463258966e-05}
{"train_loss": 0.1494615226984024, "global_step": 78375, "epoch": 880, "lr": 8.096476944978879e-05}
{"train_loss": 0.2808946669101715, "global_step": 78376, "epoch": 880, "lr": 8.096431426282508e-05}
{"train_loss": 0.1828870177268982, "global_step": 78377, "epoch": 880, "lr": 8.096385907169859e-05}
{"train_loss": 0.19639822840690613, "global_step": 78378, "epoch": 880, "lr": 8.09634038764094e-05}
{"train_loss": 0.21024952828884125, "global_step": 78379, "epoch": 880, "lr": 8.096294867695754e-05}
{"train_loss": 0.1986926794052124, "global_step": 78380, "epoch": 880, "lr": 8.096249347334311e-05}
{"train_loss": 0.2000475823879242, "global_step": 78381, "epoch": 880, "lr": 8.096203826556612e-05}
{"train_loss": 0.33488643169403076, "global_step": 78382, "epoch": 880, "lr": 8.096158305362669e-05}
{"train_loss": 0.26066580414772034, "global_step": 78383, "epoch": 880, "lr": 8.096112783752483e-05}
{"train_loss": 0.16173714399337769, "global_step": 78384, "epoch": 880, "lr": 8.096067261726063e-05}
{"train_loss": 0.14393027126789093, "global_step": 78385, "epoch": 880, "lr": 8.096021739283417e-05}
{"train_loss": 0.188238725066185, "global_step": 78386, "epoch": 880, "lr": 8.095976216424546e-05}
{"train_loss": 0.20962540805339813, "global_step": 78387, "epoch": 880, "lr": 8.095930693149459e-05}
{"train_loss": 0.22500109672546387, "global_step": 78388, "epoch": 880, "lr": 8.095885169458163e-05}
{"train_loss": 0.22082196176052094, "global_step": 78389, "epoch": 880, "lr": 8.095839645350663e-05}
{"train_loss": 0.2639237344264984, "global_step": 78390, "epoch": 880, "lr": 8.095794120826962e-05}
{"train_loss": 0.2474130094051361, "global_step": 78391, "epoch": 880, "lr": 8.095748595887074e-05}
{"train_loss": 0.30398932099342346, "global_step": 78392, "epoch": 880, "lr": 8.095703070530998e-05}
{"train_loss": 0.149137943983078, "global_step": 78393, "epoch": 880, "lr": 8.095657544758741e-05}
{"train_loss": 0.15844634175300598, "global_step": 78394, "epoch": 880, "lr": 8.095612018570312e-05}
{"train_loss": 0.17830520868301392, "global_step": 78395, "epoch": 880, "lr": 8.095566491965716e-05}
{"train_loss": 0.17440567910671234, "global_step": 78396, "epoch": 880, "lr": 8.095520964944959e-05}
{"train_loss": 0.2638739049434662, "global_step": 78397, "epoch": 880, "lr": 8.095475437508044e-05}
{"train_loss": 0.1657906472682953, "global_step": 78398, "epoch": 880, "lr": 8.095429909654984e-05}
{"train_loss": 0.196690633893013, "global_step": 78399, "epoch": 880, "lr": 8.095384381385779e-05}
{"train_loss": 0.19240164756774902, "global_step": 78400, "epoch": 880, "lr": 8.095338852700439e-05}
{"train_loss": 0.20683445036411285, "global_step": 78401, "epoch": 880, "lr": 8.095293323598967e-05}
{"train_loss": 0.13799509406089783, "global_step": 78402, "epoch": 880, "lr": 8.09524779408137e-05}
{"train_loss": 0.16240207850933075, "global_step": 78403, "epoch": 880, "lr": 8.095202264147655e-05}
{"train_loss": 0.22508737444877625, "global_step": 78404, "epoch": 880, "lr": 8.095156733797829e-05}
{"train_loss": 0.1699555516242981, "global_step": 78405, "epoch": 880, "lr": 8.095111203031896e-05}
{"train_loss": 0.26477861404418945, "global_step": 78406, "epoch": 880, "lr": 8.095065671849864e-05}
{"train_loss": 0.14793884754180908, "global_step": 78407, "epoch": 880, "lr": 8.095020140251737e-05}
{"train_loss": 0.20022687650798413, "global_step": 78408, "epoch": 880, "lr": 8.094974608237522e-05, "val_loss": 4.156840801239014, "train_action_mse_error": 18.186996459960938}
{"train_loss": 0.17848925292491913, "global_step": 78409, "epoch": 881, "lr": 8.094929075807226e-05}
{"train_loss": 0.2541990578174591, "global_step": 78410, "epoch": 881, "lr": 8.094883542960855e-05}
{"train_loss": 0.25915542244911194, "global_step": 78411, "epoch": 881, "lr": 8.094838009698413e-05}
{"train_loss": 0.25583863258361816, "global_step": 78412, "epoch": 881, "lr": 8.09479247601991e-05}
{"train_loss": 0.24642245471477509, "global_step": 78413, "epoch": 881, "lr": 8.094746941925349e-05}
{"train_loss": 0.1491377204656601, "global_step": 78414, "epoch": 881, "lr": 8.094701407414736e-05}
{"train_loss": 0.19330543279647827, "global_step": 78415, "epoch": 881, "lr": 8.094655872488078e-05}
{"train_loss": 0.21601304411888123, "global_step": 78416, "epoch": 881, "lr": 8.094610337145383e-05}
{"train_loss": 0.20480382442474365, "global_step": 78417, "epoch": 881, "lr": 8.094564801386654e-05}
{"train_loss": 0.10838188230991364, "global_step": 78418, "epoch": 881, "lr": 8.0945192652119e-05}
{"train_loss": 0.1643691211938858, "global_step": 78419, "epoch": 881, "lr": 8.094473728621124e-05}
{"train_loss": 0.236201673746109, "global_step": 78420, "epoch": 881, "lr": 8.094428191614333e-05}
{"train_loss": 0.15022461116313934, "global_step": 78421, "epoch": 881, "lr": 8.094382654191537e-05}
{"train_loss": 0.14614811539649963, "global_step": 78422, "epoch": 881, "lr": 8.094337116352736e-05}
{"train_loss": 0.12719926238059998, "global_step": 78423, "epoch": 881, "lr": 8.094291578097941e-05}
{"train_loss": 0.17141595482826233, "global_step": 78424, "epoch": 881, "lr": 8.094246039427155e-05}
{"train_loss": 0.15642589330673218, "global_step": 78425, "epoch": 881, "lr": 8.094200500340386e-05}
{"train_loss": 0.14967289566993713, "global_step": 78426, "epoch": 881, "lr": 8.094154960837638e-05}
{"train_loss": 0.25062617659568787, "global_step": 78427, "epoch": 881, "lr": 8.09410942091892e-05}
{"train_loss": 0.17501243948936462, "global_step": 78428, "epoch": 881, "lr": 8.094063880584238e-05}
{"train_loss": 0.202920064330101, "global_step": 78429, "epoch": 881, "lr": 8.094018339833594e-05}
{"train_loss": 0.22241847217082977, "global_step": 78430, "epoch": 881, "lr": 8.093972798666999e-05}
{"train_loss": 0.16602742671966553, "global_step": 78431, "epoch": 881, "lr": 8.093927257084456e-05}
{"train_loss": 0.15852542221546173, "global_step": 78432, "epoch": 881, "lr": 8.093881715085973e-05}
{"train_loss": 0.19778674840927124, "global_step": 78433, "epoch": 881, "lr": 8.093836172671553e-05}
{"train_loss": 0.2612565755844116, "global_step": 78434, "epoch": 881, "lr": 8.093790629841206e-05}
{"train_loss": 0.19522464275360107, "global_step": 78435, "epoch": 881, "lr": 8.093745086594937e-05}
{"train_loss": 0.19593818485736847, "global_step": 78436, "epoch": 881, "lr": 8.093699542932752e-05}
{"train_loss": 0.2992372512817383, "global_step": 78437, "epoch": 881, "lr": 8.093653998854655e-05}
{"train_loss": 0.15252377092838287, "global_step": 78438, "epoch": 881, "lr": 8.093608454360657e-05}
{"train_loss": 0.1774939000606537, "global_step": 78439, "epoch": 881, "lr": 8.093562909450758e-05}
{"train_loss": 0.10208652913570404, "global_step": 78440, "epoch": 881, "lr": 8.093517364124969e-05}
{"train_loss": 0.19495230913162231, "global_step": 78441, "epoch": 881, "lr": 8.093471818383294e-05}
{"train_loss": 0.25518137216567993, "global_step": 78442, "epoch": 881, "lr": 8.09342627222574e-05}
{"train_loss": 0.19801831245422363, "global_step": 78443, "epoch": 881, "lr": 8.093380725652311e-05}
{"train_loss": 0.1894681453704834, "global_step": 78444, "epoch": 881, "lr": 8.093335178663015e-05}
{"train_loss": 0.15810266137123108, "global_step": 78445, "epoch": 881, "lr": 8.093289631257859e-05}
{"train_loss": 0.24188615381717682, "global_step": 78446, "epoch": 881, "lr": 8.093244083436846e-05}
{"train_loss": 0.2100721299648285, "global_step": 78447, "epoch": 881, "lr": 8.093198535199985e-05}
{"train_loss": 0.21659034490585327, "global_step": 78448, "epoch": 881, "lr": 8.093152986547282e-05}
{"train_loss": 0.32066038250923157, "global_step": 78449, "epoch": 881, "lr": 8.093107437478742e-05}
{"train_loss": 0.22010372579097748, "global_step": 78450, "epoch": 881, "lr": 8.09306188799437e-05}
{"train_loss": 0.14464318752288818, "global_step": 78451, "epoch": 881, "lr": 8.093016338094175e-05}
{"train_loss": 0.32083839178085327, "global_step": 78452, "epoch": 881, "lr": 8.09297078777816e-05}
{"train_loss": 0.177435964345932, "global_step": 78453, "epoch": 881, "lr": 8.092925237046334e-05}
{"train_loss": 0.2683384120464325, "global_step": 78454, "epoch": 881, "lr": 8.092879685898701e-05}
{"train_loss": 0.2558155953884125, "global_step": 78455, "epoch": 881, "lr": 8.09283413433527e-05}
{"train_loss": 0.29366010427474976, "global_step": 78456, "epoch": 881, "lr": 8.092788582356044e-05}
{"train_loss": 0.2034275233745575, "global_step": 78457, "epoch": 881, "lr": 8.09274302996103e-05}
{"train_loss": 0.1655391901731491, "global_step": 78458, "epoch": 881, "lr": 8.092697477150234e-05}
{"train_loss": 0.17646901309490204, "global_step": 78459, "epoch": 881, "lr": 8.092651923923663e-05}
{"train_loss": 0.17782194912433624, "global_step": 78460, "epoch": 881, "lr": 8.092606370281324e-05}
{"train_loss": 0.18683886528015137, "global_step": 78461, "epoch": 881, "lr": 8.09256081622322e-05}
{"train_loss": 0.15690630674362183, "global_step": 78462, "epoch": 881, "lr": 8.092515261749358e-05}
{"train_loss": 0.21132619678974152, "global_step": 78463, "epoch": 881, "lr": 8.092469706859746e-05}
{"train_loss": 0.22136105597019196, "global_step": 78464, "epoch": 881, "lr": 8.09242415155439e-05}
{"train_loss": 0.2062569111585617, "global_step": 78465, "epoch": 881, "lr": 8.092378595833296e-05}
{"train_loss": 0.23419398069381714, "global_step": 78466, "epoch": 881, "lr": 8.092333039696468e-05}
{"train_loss": 0.263811856508255, "global_step": 78467, "epoch": 881, "lr": 8.092287483143914e-05}
{"train_loss": 0.2762911319732666, "global_step": 78468, "epoch": 881, "lr": 8.092241926175638e-05}
{"train_loss": 0.2062007337808609, "global_step": 78469, "epoch": 881, "lr": 8.092196368791651e-05}
{"train_loss": 0.26140204071998596, "global_step": 78470, "epoch": 881, "lr": 8.092150810991952e-05}
{"train_loss": 0.21033266186714172, "global_step": 78471, "epoch": 881, "lr": 8.092105252776554e-05}
{"train_loss": 0.15790492296218872, "global_step": 78472, "epoch": 881, "lr": 8.092059694145459e-05}
{"train_loss": 0.2153061181306839, "global_step": 78473, "epoch": 881, "lr": 8.092014135098675e-05}
{"train_loss": 0.23473097383975983, "global_step": 78474, "epoch": 881, "lr": 8.091968575636206e-05}
{"train_loss": 0.18580202758312225, "global_step": 78475, "epoch": 881, "lr": 8.09192301575806e-05}
{"train_loss": 0.2122488021850586, "global_step": 78476, "epoch": 881, "lr": 8.091877455464244e-05}
{"train_loss": 0.23032237589359283, "global_step": 78477, "epoch": 881, "lr": 8.091831894754763e-05}
{"train_loss": 0.1748867779970169, "global_step": 78478, "epoch": 881, "lr": 8.091786333629621e-05}
{"train_loss": 0.3215296268463135, "global_step": 78479, "epoch": 881, "lr": 8.091740772088825e-05}
{"train_loss": 0.14858275651931763, "global_step": 78480, "epoch": 881, "lr": 8.091695210132384e-05}
{"train_loss": 0.25534766912460327, "global_step": 78481, "epoch": 881, "lr": 8.091649647760302e-05}
{"train_loss": 0.1888435184955597, "global_step": 78482, "epoch": 881, "lr": 8.091604084972586e-05}
{"train_loss": 0.14071567356586456, "global_step": 78483, "epoch": 881, "lr": 8.09155852176924e-05}
{"train_loss": 0.211197629570961, "global_step": 78484, "epoch": 881, "lr": 8.091512958150273e-05}
{"train_loss": 0.12892046570777893, "global_step": 78485, "epoch": 881, "lr": 8.091467394115689e-05}
{"train_loss": 0.20699523389339447, "global_step": 78486, "epoch": 881, "lr": 8.091421829665495e-05}
{"train_loss": 0.2561037838459015, "global_step": 78487, "epoch": 881, "lr": 8.091376264799698e-05}
{"train_loss": 0.18442761898040771, "global_step": 78488, "epoch": 881, "lr": 8.091330699518302e-05}
{"train_loss": 0.21000292897224426, "global_step": 78489, "epoch": 881, "lr": 8.091285133821314e-05}
{"train_loss": 0.24324600398540497, "global_step": 78490, "epoch": 881, "lr": 8.091239567708742e-05}
{"train_loss": 0.23440693318843842, "global_step": 78491, "epoch": 881, "lr": 8.091194001180589e-05}
{"train_loss": 0.22501830756664276, "global_step": 78492, "epoch": 881, "lr": 8.091148434236863e-05}
{"train_loss": 0.15470947325229645, "global_step": 78493, "epoch": 881, "lr": 8.09110286687757e-05}
{"train_loss": 0.1374451220035553, "global_step": 78494, "epoch": 881, "lr": 8.091057299102715e-05}
{"train_loss": 0.2876807749271393, "global_step": 78495, "epoch": 881, "lr": 8.091011730912308e-05}
{"train_loss": 0.14986754953861237, "global_step": 78496, "epoch": 881, "lr": 8.09096616230635e-05}
{"train_loss": 0.20505933865402523, "global_step": 78497, "epoch": 881, "lr": 8.090920593284849e-05, "val_loss": 3.9381229877471924}
{"train_loss": 0.15613284707069397, "global_step": 78498, "epoch": 882, "lr": 8.09087502384781e-05}
{"train_loss": 0.23210053145885468, "global_step": 78499, "epoch": 882, "lr": 8.090829453995244e-05}
{"train_loss": 0.16300025582313538, "global_step": 78500, "epoch": 882, "lr": 8.090783883727151e-05}
{"train_loss": 0.2367403209209442, "global_step": 78501, "epoch": 882, "lr": 8.090738313043541e-05}
{"train_loss": 0.24304386973381042, "global_step": 78502, "epoch": 882, "lr": 8.09069274194442e-05}
{"train_loss": 0.2126375436782837, "global_step": 78503, "epoch": 882, "lr": 8.090647170429791e-05}
{"train_loss": 0.2007250189781189, "global_step": 78504, "epoch": 882, "lr": 8.090601598499663e-05}
{"train_loss": 0.22352540493011475, "global_step": 78505, "epoch": 882, "lr": 8.09055602615404e-05}
{"train_loss": 0.1784515529870987, "global_step": 78506, "epoch": 882, "lr": 8.09051045339293e-05}
{"train_loss": 0.22146956622600555, "global_step": 78507, "epoch": 882, "lr": 8.090464880216339e-05}
{"train_loss": 0.24867667257785797, "global_step": 78508, "epoch": 882, "lr": 8.090419306624273e-05}
{"train_loss": 0.3144347071647644, "global_step": 78509, "epoch": 882, "lr": 8.090373732616736e-05}
{"train_loss": 0.2048477977514267, "global_step": 78510, "epoch": 882, "lr": 8.090328158193738e-05}
{"train_loss": 0.17676712572574615, "global_step": 78511, "epoch": 882, "lr": 8.090282583355282e-05}
{"train_loss": 0.13180069625377655, "global_step": 78512, "epoch": 882, "lr": 8.090237008101376e-05}
{"train_loss": 0.12897931039333344, "global_step": 78513, "epoch": 882, "lr": 8.090191432432025e-05}
{"train_loss": 0.17090438306331635, "global_step": 78514, "epoch": 882, "lr": 8.090145856347235e-05}
{"train_loss": 0.16830673813819885, "global_step": 78515, "epoch": 882, "lr": 8.090100279847013e-05}
{"train_loss": 0.49492186307907104, "global_step": 78516, "epoch": 882, "lr": 8.090054702931364e-05}
{"train_loss": 0.19495104253292084, "global_step": 78517, "epoch": 882, "lr": 8.090009125600296e-05}
{"train_loss": 0.16166694462299347, "global_step": 78518, "epoch": 882, "lr": 8.089963547853814e-05}
{"train_loss": 0.19966278970241547, "global_step": 78519, "epoch": 882, "lr": 8.089917969691922e-05}
{"train_loss": 0.14456963539123535, "global_step": 78520, "epoch": 882, "lr": 8.08987239111463e-05}
{"train_loss": 0.2667452096939087, "global_step": 78521, "epoch": 882, "lr": 8.089826812121941e-05}
{"train_loss": 0.218592569231987, "global_step": 78522, "epoch": 882, "lr": 8.089781232713864e-05}
{"train_loss": 0.1952442228794098, "global_step": 78523, "epoch": 882, "lr": 8.089735652890402e-05}
{"train_loss": 0.19553004205226898, "global_step": 78524, "epoch": 882, "lr": 8.089690072651564e-05}
{"train_loss": 0.20521539449691772, "global_step": 78525, "epoch": 882, "lr": 8.089644491997355e-05}
{"train_loss": 0.23235084116458893, "global_step": 78526, "epoch": 882, "lr": 8.08959891092778e-05}
{"train_loss": 0.15508846938610077, "global_step": 78527, "epoch": 882, "lr": 8.089553329442846e-05}
{"train_loss": 0.18839403986930847, "global_step": 78528, "epoch": 882, "lr": 8.08950774754256e-05}
{"train_loss": 0.20679768919944763, "global_step": 78529, "epoch": 882, "lr": 8.089462165226927e-05}
{"train_loss": 0.16008515655994415, "global_step": 78530, "epoch": 882, "lr": 8.089416582495953e-05}
{"train_loss": 0.2505114674568176, "global_step": 78531, "epoch": 882, "lr": 8.089370999349646e-05}
{"train_loss": 0.26424452662467957, "global_step": 78532, "epoch": 882, "lr": 8.08932541578801e-05}
{"train_loss": 0.19913853704929352, "global_step": 78533, "epoch": 882, "lr": 8.089279831811051e-05}
{"train_loss": 0.21289148926734924, "global_step": 78534, "epoch": 882, "lr": 8.089234247418776e-05}
{"train_loss": 0.21752311289310455, "global_step": 78535, "epoch": 882, "lr": 8.089188662611192e-05}
{"train_loss": 0.19235122203826904, "global_step": 78536, "epoch": 882, "lr": 8.089143077388303e-05}
{"train_loss": 0.16726379096508026, "global_step": 78537, "epoch": 882, "lr": 8.089097491750119e-05}
{"train_loss": 0.19917650520801544, "global_step": 78538, "epoch": 882, "lr": 8.08905190569664e-05}
{"train_loss": 0.22929345071315765, "global_step": 78539, "epoch": 882, "lr": 8.089006319227878e-05}
{"train_loss": 0.23595605790615082, "global_step": 78540, "epoch": 882, "lr": 8.088960732343837e-05}
{"train_loss": 0.18134266138076782, "global_step": 78541, "epoch": 882, "lr": 8.088915145044522e-05}
{"train_loss": 0.23229189217090607, "global_step": 78542, "epoch": 882, "lr": 8.088869557329939e-05}
{"train_loss": 0.17778050899505615, "global_step": 78543, "epoch": 882, "lr": 8.088823969200096e-05}
{"train_loss": 0.1574312150478363, "global_step": 78544, "epoch": 882, "lr": 8.088778380654999e-05}
{"train_loss": 0.22328625619411469, "global_step": 78545, "epoch": 882, "lr": 8.088732791694653e-05}
{"train_loss": 0.23322694003582, "global_step": 78546, "epoch": 882, "lr": 8.088687202319065e-05}
{"train_loss": 0.18967750668525696, "global_step": 78547, "epoch": 882, "lr": 8.088641612528238e-05}
{"train_loss": 0.2502954602241516, "global_step": 78548, "epoch": 882, "lr": 8.088596022322184e-05}
{"train_loss": 0.12085013836622238, "global_step": 78549, "epoch": 882, "lr": 8.088550431700905e-05}
{"train_loss": 0.21648108959197998, "global_step": 78550, "epoch": 882, "lr": 8.088504840664407e-05}
{"train_loss": 0.1471356898546219, "global_step": 78551, "epoch": 882, "lr": 8.088459249212698e-05}
{"train_loss": 0.1722700446844101, "global_step": 78552, "epoch": 882, "lr": 8.088413657345783e-05}
{"train_loss": 0.18492214381694794, "global_step": 78553, "epoch": 882, "lr": 8.08836806506367e-05}
{"train_loss": 0.23714737594127655, "global_step": 78554, "epoch": 882, "lr": 8.08832247236636e-05}
{"train_loss": 0.15388622879981995, "global_step": 78555, "epoch": 882, "lr": 8.088276879253867e-05}
{"train_loss": 0.22481070458889008, "global_step": 78556, "epoch": 882, "lr": 8.08823128572619e-05}
{"train_loss": 0.16034245491027832, "global_step": 78557, "epoch": 882, "lr": 8.088185691783339e-05}
{"train_loss": 0.3454121947288513, "global_step": 78558, "epoch": 882, "lr": 8.088140097425319e-05}
{"train_loss": 0.23296178877353668, "global_step": 78559, "epoch": 882, "lr": 8.088094502652135e-05}
{"train_loss": 0.2399291843175888, "global_step": 78560, "epoch": 882, "lr": 8.088048907463795e-05}
{"train_loss": 0.19120103120803833, "global_step": 78561, "epoch": 882, "lr": 8.088003311860303e-05}
{"train_loss": 0.19567029178142548, "global_step": 78562, "epoch": 882, "lr": 8.08795771584167e-05}
{"train_loss": 0.18341270089149475, "global_step": 78563, "epoch": 882, "lr": 8.087912119407897e-05}
{"train_loss": 0.2028852105140686, "global_step": 78564, "epoch": 882, "lr": 8.087866522558991e-05}
{"train_loss": 0.20344944298267365, "global_step": 78565, "epoch": 882, "lr": 8.08782092529496e-05}
{"train_loss": 0.1212461069226265, "global_step": 78566, "epoch": 882, "lr": 8.08777532761581e-05}
{"train_loss": 0.20620974898338318, "global_step": 78567, "epoch": 882, "lr": 8.087729729521544e-05}
{"train_loss": 0.2098897248506546, "global_step": 78568, "epoch": 882, "lr": 8.087684131012172e-05}
{"train_loss": 0.3045841455459595, "global_step": 78569, "epoch": 882, "lr": 8.087638532087697e-05}
{"train_loss": 0.2167913019657135, "global_step": 78570, "epoch": 882, "lr": 8.087592932748129e-05}
{"train_loss": 0.16969989240169525, "global_step": 78571, "epoch": 882, "lr": 8.087547332993471e-05}
{"train_loss": 0.23781852424144745, "global_step": 78572, "epoch": 882, "lr": 8.087501732823729e-05}
{"train_loss": 0.16266681253910065, "global_step": 78573, "epoch": 882, "lr": 8.087456132238911e-05}
{"train_loss": 0.13545256853103638, "global_step": 78574, "epoch": 882, "lr": 8.087410531239021e-05}
{"train_loss": 0.17861834168434143, "global_step": 78575, "epoch": 882, "lr": 8.087364929824066e-05}
{"train_loss": 0.2316834032535553, "global_step": 78576, "epoch": 882, "lr": 8.087319327994054e-05}
{"train_loss": 0.17554239928722382, "global_step": 78577, "epoch": 882, "lr": 8.087273725748989e-05}
{"train_loss": 0.13666900992393494, "global_step": 78578, "epoch": 882, "lr": 8.087228123088876e-05}
{"train_loss": 0.15186594426631927, "global_step": 78579, "epoch": 882, "lr": 8.087182520013726e-05}
{"train_loss": 0.11266189068555832, "global_step": 78580, "epoch": 882, "lr": 8.087136916523539e-05}
{"train_loss": 0.1977190375328064, "global_step": 78581, "epoch": 882, "lr": 8.087091312618326e-05}
{"train_loss": 0.17544569075107574, "global_step": 78582, "epoch": 882, "lr": 8.08704570829809e-05}
{"train_loss": 0.10321532189846039, "global_step": 78583, "epoch": 882, "lr": 8.087000103562839e-05}
{"train_loss": 0.12560661137104034, "global_step": 78584, "epoch": 882, "lr": 8.086954498412578e-05}
{"train_loss": 0.2119302898645401, "global_step": 78585, "epoch": 882, "lr": 8.086908892847313e-05}
{"train_loss": 0.20023761179005162, "global_step": 78586, "epoch": 882, "lr": 8.086863286867053e-05, "val_loss": 4.030807018280029}
{"train_loss": 0.1958286613225937, "global_step": 78587, "epoch": 883, "lr": 8.0868176804718e-05}
{"train_loss": 0.17217256128787994, "global_step": 78588, "epoch": 883, "lr": 8.086772073661561e-05}
{"train_loss": 0.10376136749982834, "global_step": 78589, "epoch": 883, "lr": 8.086726466436345e-05}
{"train_loss": 0.20100979506969452, "global_step": 78590, "epoch": 883, "lr": 8.086680858796156e-05}
{"train_loss": 0.21099092066287994, "global_step": 78591, "epoch": 883, "lr": 8.086635250741002e-05}
{"train_loss": 0.19670912623405457, "global_step": 78592, "epoch": 883, "lr": 8.086589642270885e-05}
{"train_loss": 0.2625764310359955, "global_step": 78593, "epoch": 883, "lr": 8.086544033385814e-05}
{"train_loss": 0.20424431562423706, "global_step": 78594, "epoch": 883, "lr": 8.086498424085795e-05}
{"train_loss": 0.11847612261772156, "global_step": 78595, "epoch": 883, "lr": 8.086452814370834e-05}
{"train_loss": 0.1209753006696701, "global_step": 78596, "epoch": 883, "lr": 8.086407204240938e-05}
{"train_loss": 0.19201919436454773, "global_step": 78597, "epoch": 883, "lr": 8.08636159369611e-05}
{"train_loss": 0.19032925367355347, "global_step": 78598, "epoch": 883, "lr": 8.086315982736358e-05}
{"train_loss": 0.16438263654708862, "global_step": 78599, "epoch": 883, "lr": 8.08627037136169e-05}
{"train_loss": 0.21525397896766663, "global_step": 78600, "epoch": 883, "lr": 8.086224759572112e-05}
{"train_loss": 0.17195436358451843, "global_step": 78601, "epoch": 883, "lr": 8.086179147367627e-05}
{"train_loss": 0.1437804400920868, "global_step": 78602, "epoch": 883, "lr": 8.086133534748243e-05}
{"train_loss": 0.15127679705619812, "global_step": 78603, "epoch": 883, "lr": 8.086087921713965e-05}
{"train_loss": 0.14738598465919495, "global_step": 78604, "epoch": 883, "lr": 8.086042308264801e-05}
{"train_loss": 0.2287587821483612, "global_step": 78605, "epoch": 883, "lr": 8.085996694400758e-05}
{"train_loss": 0.18216674029827118, "global_step": 78606, "epoch": 883, "lr": 8.085951080121839e-05}
{"train_loss": 0.14374887943267822, "global_step": 78607, "epoch": 883, "lr": 8.08590546542805e-05}
{"train_loss": 0.17596015334129333, "global_step": 78608, "epoch": 883, "lr": 8.085859850319399e-05}
{"train_loss": 0.2252930998802185, "global_step": 78609, "epoch": 883, "lr": 8.085814234795893e-05}
{"train_loss": 0.26777219772338867, "global_step": 78610, "epoch": 883, "lr": 8.085768618857536e-05}
{"train_loss": 0.17617136240005493, "global_step": 78611, "epoch": 883, "lr": 8.085723002504337e-05}
{"train_loss": 0.21210643649101257, "global_step": 78612, "epoch": 883, "lr": 8.085677385736298e-05}
{"train_loss": 0.21642358601093292, "global_step": 78613, "epoch": 883, "lr": 8.085631768553428e-05}
{"train_loss": 0.2751104235649109, "global_step": 78614, "epoch": 883, "lr": 8.085586150955733e-05}
{"train_loss": 0.13388700783252716, "global_step": 78615, "epoch": 883, "lr": 8.085540532943217e-05}
{"train_loss": 0.18908827006816864, "global_step": 78616, "epoch": 883, "lr": 8.085494914515888e-05}
{"train_loss": 0.1749897301197052, "global_step": 78617, "epoch": 883, "lr": 8.085449295673753e-05}
{"train_loss": 0.1957838535308838, "global_step": 78618, "epoch": 883, "lr": 8.085403676416816e-05}
{"train_loss": 0.13813816010951996, "global_step": 78619, "epoch": 883, "lr": 8.085358056745085e-05}
{"train_loss": 0.1959960013628006, "global_step": 78620, "epoch": 883, "lr": 8.085312436658563e-05}
{"train_loss": 0.19964253902435303, "global_step": 78621, "epoch": 883, "lr": 8.085266816157261e-05}
{"train_loss": 0.2677452266216278, "global_step": 78622, "epoch": 883, "lr": 8.085221195241182e-05}
{"train_loss": 0.1576269268989563, "global_step": 78623, "epoch": 883, "lr": 8.085175573910331e-05}
{"train_loss": 0.29745328426361084, "global_step": 78624, "epoch": 883, "lr": 8.085129952164717e-05}
{"train_loss": 0.14955265820026398, "global_step": 78625, "epoch": 883, "lr": 8.085084330004344e-05}
{"train_loss": 0.2361895889043808, "global_step": 78626, "epoch": 883, "lr": 8.08503870742922e-05}
{"train_loss": 0.33965981006622314, "global_step": 78627, "epoch": 883, "lr": 8.084993084439349e-05}
{"train_loss": 0.19905680418014526, "global_step": 78628, "epoch": 883, "lr": 8.08494746103474e-05}
{"train_loss": 0.16594021022319794, "global_step": 78629, "epoch": 883, "lr": 8.084901837215398e-05}
{"train_loss": 0.1789376586675644, "global_step": 78630, "epoch": 883, "lr": 8.084856212981326e-05}
{"train_loss": 0.20004932582378387, "global_step": 78631, "epoch": 883, "lr": 8.084810588332534e-05}
{"train_loss": 0.20847062766551971, "global_step": 78632, "epoch": 883, "lr": 8.084764963269027e-05}
{"train_loss": 0.19512954354286194, "global_step": 78633, "epoch": 883, "lr": 8.08471933779081e-05}
{"train_loss": 0.2103009819984436, "global_step": 78634, "epoch": 883, "lr": 8.08467371189789e-05}
{"train_loss": 0.28541260957717896, "global_step": 78635, "epoch": 883, "lr": 8.084628085590273e-05}
{"train_loss": 0.212986022233963, "global_step": 78636, "epoch": 883, "lr": 8.084582458867966e-05}
{"train_loss": 0.20096421241760254, "global_step": 78637, "epoch": 883, "lr": 8.084536831730976e-05}
{"train_loss": 0.2530733048915863, "global_step": 78638, "epoch": 883, "lr": 8.084491204179305e-05}
{"train_loss": 0.15715065598487854, "global_step": 78639, "epoch": 883, "lr": 8.084445576212964e-05}
{"train_loss": 0.21991485357284546, "global_step": 78640, "epoch": 883, "lr": 8.084399947831955e-05}
{"train_loss": 0.17140384018421173, "global_step": 78641, "epoch": 883, "lr": 8.084354319036287e-05}
{"train_loss": 0.19805914163589478, "global_step": 78642, "epoch": 883, "lr": 8.084308689825965e-05}
{"train_loss": 0.321896493434906, "global_step": 78643, "epoch": 883, "lr": 8.084263060200996e-05}
{"train_loss": 0.18737167119979858, "global_step": 78644, "epoch": 883, "lr": 8.084217430161386e-05}
{"train_loss": 0.19143392145633698, "global_step": 78645, "epoch": 883, "lr": 8.084171799707139e-05}
{"train_loss": 0.2084970921278, "global_step": 78646, "epoch": 883, "lr": 8.084126168838261e-05}
{"train_loss": 0.17937713861465454, "global_step": 78647, "epoch": 883, "lr": 8.084080537554763e-05}
{"train_loss": 0.234027698636055, "global_step": 78648, "epoch": 883, "lr": 8.084034905856647e-05}
{"train_loss": 0.2122860997915268, "global_step": 78649, "epoch": 883, "lr": 8.083989273743919e-05}
{"train_loss": 0.21906797587871552, "global_step": 78650, "epoch": 883, "lr": 8.083943641216588e-05}
{"train_loss": 0.22310976684093475, "global_step": 78651, "epoch": 883, "lr": 8.083898008274657e-05}
{"train_loss": 0.2185852974653244, "global_step": 78652, "epoch": 883, "lr": 8.083852374918134e-05}
{"train_loss": 0.23153804242610931, "global_step": 78653, "epoch": 883, "lr": 8.083806741147025e-05}
{"train_loss": 0.22450007498264313, "global_step": 78654, "epoch": 883, "lr": 8.083761106961335e-05}
{"train_loss": 0.19554860889911652, "global_step": 78655, "epoch": 883, "lr": 8.083715472361071e-05}
{"train_loss": 0.24166010320186615, "global_step": 78656, "epoch": 883, "lr": 8.08366983734624e-05}
{"train_loss": 0.2085040658712387, "global_step": 78657, "epoch": 883, "lr": 8.083624201916846e-05}
{"train_loss": 0.23244811594486237, "global_step": 78658, "epoch": 883, "lr": 8.083578566072897e-05}
{"train_loss": 0.22842663526535034, "global_step": 78659, "epoch": 883, "lr": 8.083532929814399e-05}
{"train_loss": 0.18604673445224762, "global_step": 78660, "epoch": 883, "lr": 8.083487293141356e-05}
{"train_loss": 0.24759846925735474, "global_step": 78661, "epoch": 883, "lr": 8.083441656053776e-05}
{"train_loss": 0.26785722374916077, "global_step": 78662, "epoch": 883, "lr": 8.083396018551667e-05}
{"train_loss": 0.24620185792446136, "global_step": 78663, "epoch": 883, "lr": 8.08335038063503e-05}
{"train_loss": 0.14090898633003235, "global_step": 78664, "epoch": 883, "lr": 8.083304742303877e-05}
{"train_loss": 0.20646142959594727, "global_step": 78665, "epoch": 883, "lr": 8.083259103558209e-05}
{"train_loss": 0.23235945403575897, "global_step": 78666, "epoch": 883, "lr": 8.083213464398036e-05}
{"train_loss": 0.19013935327529907, "global_step": 78667, "epoch": 883, "lr": 8.083167824823362e-05}
{"train_loss": 0.2721787095069885, "global_step": 78668, "epoch": 883, "lr": 8.083122184834193e-05}
{"train_loss": 0.24228838086128235, "global_step": 78669, "epoch": 883, "lr": 8.083076544430535e-05}
{"train_loss": 0.22359435260295868, "global_step": 78670, "epoch": 883, "lr": 8.083030903612396e-05}
{"train_loss": 0.16862760484218597, "global_step": 78671, "epoch": 883, "lr": 8.082985262379781e-05}
{"train_loss": 0.15785127878189087, "global_step": 78672, "epoch": 883, "lr": 8.082939620732697e-05}
{"train_loss": 0.17270997166633606, "global_step": 78673, "epoch": 883, "lr": 8.082893978671148e-05}
{"train_loss": 0.20637549459934235, "global_step": 78674, "epoch": 883, "lr": 8.082848336195144e-05}
{"train_loss": 0.2036827736523714, "global_step": 78675, "epoch": 883, "lr": 8.082802693304686e-05, "val_loss": 3.8319075107574463}
{"train_loss": 0.21893158555030823, "global_step": 78676, "epoch": 884, "lr": 8.082757049999785e-05}
{"train_loss": 0.16019032895565033, "global_step": 78677, "epoch": 884, "lr": 8.082711406280443e-05}
{"train_loss": 0.17567525804042816, "global_step": 78678, "epoch": 884, "lr": 8.082665762146668e-05}
{"train_loss": 0.1835334748029709, "global_step": 78679, "epoch": 884, "lr": 8.082620117598466e-05}
{"train_loss": 0.1774463653564453, "global_step": 78680, "epoch": 884, "lr": 8.082574472635845e-05}
{"train_loss": 0.1867460161447525, "global_step": 78681, "epoch": 884, "lr": 8.082528827258808e-05}
{"train_loss": 0.3080233037471771, "global_step": 78682, "epoch": 884, "lr": 8.082483181467363e-05}
{"train_loss": 0.17448697984218597, "global_step": 78683, "epoch": 884, "lr": 8.082437535261516e-05}
{"train_loss": 0.21635392308235168, "global_step": 78684, "epoch": 884, "lr": 8.082391888641271e-05}
{"train_loss": 0.2668927013874054, "global_step": 78685, "epoch": 884, "lr": 8.082346241606637e-05}
{"train_loss": 0.30426010489463806, "global_step": 78686, "epoch": 884, "lr": 8.082300594157622e-05}
{"train_loss": 0.22388923168182373, "global_step": 78687, "epoch": 884, "lr": 8.082254946294226e-05}
{"train_loss": 0.18886011838912964, "global_step": 78688, "epoch": 884, "lr": 8.082209298016459e-05}
{"train_loss": 0.16892774403095245, "global_step": 78689, "epoch": 884, "lr": 8.082163649324326e-05}
{"train_loss": 0.20204900205135345, "global_step": 78690, "epoch": 884, "lr": 8.082118000217836e-05}
{"train_loss": 0.2613503336906433, "global_step": 78691, "epoch": 884, "lr": 8.08207235069699e-05}
{"train_loss": 0.2435399740934372, "global_step": 78692, "epoch": 884, "lr": 8.082026700761798e-05}
{"train_loss": 0.22002755105495453, "global_step": 78693, "epoch": 884, "lr": 8.081981050412267e-05}
{"train_loss": 0.2278536707162857, "global_step": 78694, "epoch": 884, "lr": 8.081935399648398e-05}
{"train_loss": 0.272604763507843, "global_step": 78695, "epoch": 884, "lr": 8.081889748470202e-05}
{"train_loss": 0.16422918438911438, "global_step": 78696, "epoch": 884, "lr": 8.081844096877685e-05}
{"train_loss": 0.13795962929725647, "global_step": 78697, "epoch": 884, "lr": 8.081798444870849e-05}
{"train_loss": 0.24005743861198425, "global_step": 78698, "epoch": 884, "lr": 8.081752792449704e-05}
{"train_loss": 0.21833287179470062, "global_step": 78699, "epoch": 884, "lr": 8.081707139614255e-05}
{"train_loss": 0.1506737917661667, "global_step": 78700, "epoch": 884, "lr": 8.081661486364507e-05}
{"train_loss": 0.22042256593704224, "global_step": 78701, "epoch": 884, "lr": 8.081615832700467e-05}
{"train_loss": 0.1832612156867981, "global_step": 78702, "epoch": 884, "lr": 8.081570178622143e-05}
{"train_loss": 0.1798652857542038, "global_step": 78703, "epoch": 884, "lr": 8.081524524129537e-05}
{"train_loss": 0.24965037405490875, "global_step": 78704, "epoch": 884, "lr": 8.08147886922266e-05}
{"train_loss": 0.2078624814748764, "global_step": 78705, "epoch": 884, "lr": 8.081433213901515e-05}
{"train_loss": 0.17583727836608887, "global_step": 78706, "epoch": 884, "lr": 8.081387558166108e-05}
{"train_loss": 0.15498553216457367, "global_step": 78707, "epoch": 884, "lr": 8.081341902016446e-05}
{"train_loss": 0.31923380494117737, "global_step": 78708, "epoch": 884, "lr": 8.081296245452536e-05}
{"train_loss": 0.13614146411418915, "global_step": 78709, "epoch": 884, "lr": 8.081250588474384e-05}
{"train_loss": 0.213373601436615, "global_step": 78710, "epoch": 884, "lr": 8.081204931081994e-05}
{"train_loss": 0.19900177419185638, "global_step": 78711, "epoch": 884, "lr": 8.081159273275374e-05}
{"train_loss": 0.1363239288330078, "global_step": 78712, "epoch": 884, "lr": 8.08111361505453e-05}
{"train_loss": 0.2109854519367218, "global_step": 78713, "epoch": 884, "lr": 8.081067956419468e-05}
{"train_loss": 0.278184711933136, "global_step": 78714, "epoch": 884, "lr": 8.081022297370193e-05}
{"train_loss": 0.29858896136283875, "global_step": 78715, "epoch": 884, "lr": 8.080976637906715e-05}
{"train_loss": 0.18449540436267853, "global_step": 78716, "epoch": 884, "lr": 8.080930978029033e-05}
{"train_loss": 0.1538982093334198, "global_step": 78717, "epoch": 884, "lr": 8.08088531773716e-05}
{"train_loss": 0.22516009211540222, "global_step": 78718, "epoch": 884, "lr": 8.080839657031099e-05}
{"train_loss": 0.23161891102790833, "global_step": 78719, "epoch": 884, "lr": 8.080793995910855e-05}
{"train_loss": 0.21200761198997498, "global_step": 78720, "epoch": 884, "lr": 8.080748334376437e-05}
{"train_loss": 0.1444726586341858, "global_step": 78721, "epoch": 884, "lr": 8.080702672427851e-05}
{"train_loss": 0.1197333112359047, "global_step": 78722, "epoch": 884, "lr": 8.080657010065103e-05}
{"train_loss": 0.2086833268404007, "global_step": 78723, "epoch": 884, "lr": 8.080611347288195e-05}
{"train_loss": 0.20035721361637115, "global_step": 78724, "epoch": 884, "lr": 8.080565684097138e-05}
{"train_loss": 0.1627614051103592, "global_step": 78725, "epoch": 884, "lr": 8.080520020491936e-05}
{"train_loss": 0.1810102015733719, "global_step": 78726, "epoch": 884, "lr": 8.080474356472597e-05}
{"train_loss": 0.22287431359291077, "global_step": 78727, "epoch": 884, "lr": 8.080428692039124e-05}
{"train_loss": 0.16446992754936218, "global_step": 78728, "epoch": 884, "lr": 8.080383027191526e-05}
{"train_loss": 0.19598113000392914, "global_step": 78729, "epoch": 884, "lr": 8.080337361929808e-05}
{"train_loss": 0.15164782106876373, "global_step": 78730, "epoch": 884, "lr": 8.080291696253976e-05}
{"train_loss": 0.22414487600326538, "global_step": 78731, "epoch": 884, "lr": 8.080246030164035e-05}
{"train_loss": 0.3133961260318756, "global_step": 78732, "epoch": 884, "lr": 8.080200363659995e-05}
{"train_loss": 0.17326825857162476, "global_step": 78733, "epoch": 884, "lr": 8.080154696741857e-05}
{"train_loss": 0.2130136787891388, "global_step": 78734, "epoch": 884, "lr": 8.080109029409632e-05}
{"train_loss": 0.2634396553039551, "global_step": 78735, "epoch": 884, "lr": 8.080063361663324e-05}
{"train_loss": 0.16914497315883636, "global_step": 78736, "epoch": 884, "lr": 8.080017693502937e-05}
{"train_loss": 0.21323645114898682, "global_step": 78737, "epoch": 884, "lr": 8.07997202492848e-05}
{"train_loss": 0.27776196599006653, "global_step": 78738, "epoch": 884, "lr": 8.079926355939959e-05}
{"train_loss": 0.2082243710756302, "global_step": 78739, "epoch": 884, "lr": 8.079880686537377e-05}
{"train_loss": 0.2024022340774536, "global_step": 78740, "epoch": 884, "lr": 8.079835016720745e-05}
{"train_loss": 0.2153327614068985, "global_step": 78741, "epoch": 884, "lr": 8.079789346490066e-05}
{"train_loss": 0.21016980707645416, "global_step": 78742, "epoch": 884, "lr": 8.079743675845347e-05}
{"train_loss": 0.16066017746925354, "global_step": 78743, "epoch": 884, "lr": 8.079698004786594e-05}
{"train_loss": 0.207182377576828, "global_step": 78744, "epoch": 884, "lr": 8.079652333313814e-05}
{"train_loss": 0.1730150580406189, "global_step": 78745, "epoch": 884, "lr": 8.079606661427011e-05}
{"train_loss": 0.18883171677589417, "global_step": 78746, "epoch": 884, "lr": 8.079560989126193e-05}
{"train_loss": 0.2180320769548416, "global_step": 78747, "epoch": 884, "lr": 8.079515316411366e-05}
{"train_loss": 0.2580660283565521, "global_step": 78748, "epoch": 884, "lr": 8.079469643282534e-05}
{"train_loss": 0.13419072329998016, "global_step": 78749, "epoch": 884, "lr": 8.079423969739706e-05}
{"train_loss": 0.17320363223552704, "global_step": 78750, "epoch": 884, "lr": 8.079378295782887e-05}
{"train_loss": 0.14997176826000214, "global_step": 78751, "epoch": 884, "lr": 8.079332621412083e-05}
{"train_loss": 0.16661500930786133, "global_step": 78752, "epoch": 884, "lr": 8.079286946627301e-05}
{"train_loss": 0.18484950065612793, "global_step": 78753, "epoch": 884, "lr": 8.079241271428546e-05}
{"train_loss": 0.19896727800369263, "global_step": 78754, "epoch": 884, "lr": 8.079195595815823e-05}
{"train_loss": 0.19982972741127014, "global_step": 78755, "epoch": 884, "lr": 8.079149919789142e-05}
{"train_loss": 0.23092526197433472, "global_step": 78756, "epoch": 884, "lr": 8.079104243348505e-05}
{"train_loss": 0.3045552372932434, "global_step": 78757, "epoch": 884, "lr": 8.079058566493921e-05}
{"train_loss": 0.23614248633384705, "global_step": 78758, "epoch": 884, "lr": 8.079012889225394e-05}
{"train_loss": 0.12249846011400223, "global_step": 78759, "epoch": 884, "lr": 8.078967211542933e-05}
{"train_loss": 0.13771186769008636, "global_step": 78760, "epoch": 884, "lr": 8.07892153344654e-05}
{"train_loss": 0.22027772665023804, "global_step": 78761, "epoch": 884, "lr": 8.078875854936225e-05}
{"train_loss": 0.14548714458942413, "global_step": 78762, "epoch": 884, "lr": 8.078830176011993e-05}
{"train_loss": 0.19881822168827057, "global_step": 78763, "epoch": 884, "lr": 8.07878449667385e-05}
{"train_loss": 0.20344872036007014, "global_step": 78764, "epoch": 884, "lr": 8.078738816921801e-05, "val_loss": 3.906843900680542}
{"train_loss": 0.2691017985343933, "global_step": 78765, "epoch": 885, "lr": 8.078693136755854e-05}
{"train_loss": 0.20411743223667145, "global_step": 78766, "epoch": 885, "lr": 8.078647456176014e-05}
{"train_loss": 0.1974707543849945, "global_step": 78767, "epoch": 885, "lr": 8.078601775182286e-05}
{"train_loss": 0.16465136408805847, "global_step": 78768, "epoch": 885, "lr": 8.078556093774678e-05}
{"train_loss": 0.21455806493759155, "global_step": 78769, "epoch": 885, "lr": 8.078510411953197e-05}
{"train_loss": 0.19888180494308472, "global_step": 78770, "epoch": 885, "lr": 8.078464729717847e-05}
{"train_loss": 0.22830061614513397, "global_step": 78771, "epoch": 885, "lr": 8.078419047068635e-05}
{"train_loss": 0.1681186556816101, "global_step": 78772, "epoch": 885, "lr": 8.078373364005568e-05}
{"train_loss": 0.2145698517560959, "global_step": 78773, "epoch": 885, "lr": 8.07832768052865e-05}
{"train_loss": 0.2243984043598175, "global_step": 78774, "epoch": 885, "lr": 8.078281996637888e-05}
{"train_loss": 0.2719833254814148, "global_step": 78775, "epoch": 885, "lr": 8.078236312333289e-05}
{"train_loss": 0.1278264969587326, "global_step": 78776, "epoch": 885, "lr": 8.07819062761486e-05}
{"train_loss": 0.21657750010490417, "global_step": 78777, "epoch": 885, "lr": 8.078144942482603e-05}
{"train_loss": 0.17964491248130798, "global_step": 78778, "epoch": 885, "lr": 8.078099256936529e-05}
{"train_loss": 0.22010521590709686, "global_step": 78779, "epoch": 885, "lr": 8.078053570976642e-05}
{"train_loss": 0.2141665518283844, "global_step": 78780, "epoch": 885, "lr": 8.078007884602947e-05}
{"train_loss": 0.28556540608406067, "global_step": 78781, "epoch": 885, "lr": 8.077962197815453e-05}
{"train_loss": 0.2844129204750061, "global_step": 78782, "epoch": 885, "lr": 8.077916510614164e-05}
{"train_loss": 0.19035397469997406, "global_step": 78783, "epoch": 885, "lr": 8.077870822999087e-05}
{"train_loss": 0.23114648461341858, "global_step": 78784, "epoch": 885, "lr": 8.077825134970226e-05}
{"train_loss": 0.10485467314720154, "global_step": 78785, "epoch": 885, "lr": 8.07777944652759e-05}
{"train_loss": 0.21027080714702606, "global_step": 78786, "epoch": 885, "lr": 8.077733757671184e-05}
{"train_loss": 0.13377611339092255, "global_step": 78787, "epoch": 885, "lr": 8.077688068401015e-05}
{"train_loss": 0.2020667940378189, "global_step": 78788, "epoch": 885, "lr": 8.077642378717087e-05}
{"train_loss": 0.23060274124145508, "global_step": 78789, "epoch": 885, "lr": 8.077596688619409e-05}
{"train_loss": 0.14977002143859863, "global_step": 78790, "epoch": 885, "lr": 8.077550998107984e-05}
{"train_loss": 0.1741051822900772, "global_step": 78791, "epoch": 885, "lr": 8.07750530718282e-05}
{"train_loss": 0.17960704863071442, "global_step": 78792, "epoch": 885, "lr": 8.077459615843925e-05}
{"train_loss": 0.16562427580356598, "global_step": 78793, "epoch": 885, "lr": 8.077413924091301e-05}
{"train_loss": 0.19725646078586578, "global_step": 78794, "epoch": 885, "lr": 8.077368231924957e-05}
{"train_loss": 0.347123384475708, "global_step": 78795, "epoch": 885, "lr": 8.077322539344898e-05}
{"train_loss": 0.22905753552913666, "global_step": 78796, "epoch": 885, "lr": 8.07727684635113e-05}
{"train_loss": 0.14093706011772156, "global_step": 78797, "epoch": 885, "lr": 8.077231152943661e-05}
{"train_loss": 0.2860320508480072, "global_step": 78798, "epoch": 885, "lr": 8.077185459122494e-05}
{"train_loss": 0.12826307117938995, "global_step": 78799, "epoch": 885, "lr": 8.077139764887638e-05}
{"train_loss": 0.20803530514240265, "global_step": 78800, "epoch": 885, "lr": 8.077094070239097e-05}
{"train_loss": 0.182313472032547, "global_step": 78801, "epoch": 885, "lr": 8.077048375176881e-05}
{"train_loss": 0.1470903754234314, "global_step": 78802, "epoch": 885, "lr": 8.077002679700991e-05}
{"train_loss": 0.1990647315979004, "global_step": 78803, "epoch": 885, "lr": 8.076956983811436e-05}
{"train_loss": 0.2107580602169037, "global_step": 78804, "epoch": 885, "lr": 8.076911287508222e-05}
{"train_loss": 0.23304806649684906, "global_step": 78805, "epoch": 885, "lr": 8.076865590791354e-05}
{"train_loss": 0.1947745680809021, "global_step": 78806, "epoch": 885, "lr": 8.076819893660838e-05}
{"train_loss": 0.19786560535430908, "global_step": 78807, "epoch": 885, "lr": 8.076774196116683e-05}
{"train_loss": 0.24991163611412048, "global_step": 78808, "epoch": 885, "lr": 8.076728498158894e-05}
{"train_loss": 0.18590950965881348, "global_step": 78809, "epoch": 885, "lr": 8.076682799787474e-05}
{"train_loss": 0.12093674391508102, "global_step": 78810, "epoch": 885, "lr": 8.076637101002433e-05}
{"train_loss": 0.1323009729385376, "global_step": 78811, "epoch": 885, "lr": 8.076591401803775e-05}
{"train_loss": 0.1300097554922104, "global_step": 78812, "epoch": 885, "lr": 8.076545702191507e-05}
{"train_loss": 0.24089570343494415, "global_step": 78813, "epoch": 885, "lr": 8.076500002165635e-05}
{"train_loss": 0.1668715626001358, "global_step": 78814, "epoch": 885, "lr": 8.076454301726164e-05}
{"train_loss": 0.1512637436389923, "global_step": 78815, "epoch": 885, "lr": 8.076408600873102e-05}
{"train_loss": 0.2048768699169159, "global_step": 78816, "epoch": 885, "lr": 8.076362899606456e-05}
{"train_loss": 0.22045697271823883, "global_step": 78817, "epoch": 885, "lr": 8.076317197926229e-05}
{"train_loss": 0.19427433609962463, "global_step": 78818, "epoch": 885, "lr": 8.076271495832428e-05}
{"train_loss": 0.2201957255601883, "global_step": 78819, "epoch": 885, "lr": 8.07622579332506e-05}
{"train_loss": 0.12437477707862854, "global_step": 78820, "epoch": 885, "lr": 8.076180090404132e-05}
{"train_loss": 0.13193057477474213, "global_step": 78821, "epoch": 885, "lr": 8.076134387069648e-05}
{"train_loss": 0.25603315234184265, "global_step": 78822, "epoch": 885, "lr": 8.076088683321616e-05}
{"train_loss": 0.1871311366558075, "global_step": 78823, "epoch": 885, "lr": 8.076042979160041e-05}
{"train_loss": 0.18513409793376923, "global_step": 78824, "epoch": 885, "lr": 8.07599727458493e-05}
{"train_loss": 0.1636156588792801, "global_step": 78825, "epoch": 885, "lr": 8.075951569596288e-05}
{"train_loss": 0.19213569164276123, "global_step": 78826, "epoch": 885, "lr": 8.075905864194121e-05}
{"train_loss": 0.12286356091499329, "global_step": 78827, "epoch": 885, "lr": 8.075860158378438e-05}
{"train_loss": 0.17144857347011566, "global_step": 78828, "epoch": 885, "lr": 8.075814452149242e-05}
{"train_loss": 0.16666288673877716, "global_step": 78829, "epoch": 885, "lr": 8.075768745506541e-05}
{"train_loss": 0.18347962200641632, "global_step": 78830, "epoch": 885, "lr": 8.075723038450338e-05}
{"train_loss": 0.23088157176971436, "global_step": 78831, "epoch": 885, "lr": 8.075677330980645e-05}
{"train_loss": 0.1908298134803772, "global_step": 78832, "epoch": 885, "lr": 8.075631623097462e-05}
{"train_loss": 0.1932620406150818, "global_step": 78833, "epoch": 885, "lr": 8.075585914800798e-05}
{"train_loss": 0.1483970433473587, "global_step": 78834, "epoch": 885, "lr": 8.075540206090659e-05}
{"train_loss": 0.12415937334299088, "global_step": 78835, "epoch": 885, "lr": 8.075494496967053e-05}
{"train_loss": 0.20975804328918457, "global_step": 78836, "epoch": 885, "lr": 8.075448787429982e-05}
{"train_loss": 0.18240682780742645, "global_step": 78837, "epoch": 885, "lr": 8.075403077479456e-05}
{"train_loss": 0.1956486701965332, "global_step": 78838, "epoch": 885, "lr": 8.075357367115479e-05}
{"train_loss": 0.1523141860961914, "global_step": 78839, "epoch": 885, "lr": 8.075311656338059e-05}
{"train_loss": 0.18301017582416534, "global_step": 78840, "epoch": 885, "lr": 8.075265945147197e-05}
{"train_loss": 0.13874951004981995, "global_step": 78841, "epoch": 885, "lr": 8.075220233542907e-05}
{"train_loss": 0.1971627026796341, "global_step": 78842, "epoch": 885, "lr": 8.075174521525188e-05}
{"train_loss": 0.18673110008239746, "global_step": 78843, "epoch": 885, "lr": 8.075128809094052e-05}
{"train_loss": 0.15523821115493774, "global_step": 78844, "epoch": 885, "lr": 8.0750830962495e-05}
{"train_loss": 0.111006960272789, "global_step": 78845, "epoch": 885, "lr": 8.075037382991543e-05}
{"train_loss": 0.24623703956604004, "global_step": 78846, "epoch": 885, "lr": 8.074991669320184e-05}
{"train_loss": 0.26227083802223206, "global_step": 78847, "epoch": 885, "lr": 8.07494595523543e-05}
{"train_loss": 0.18031255900859833, "global_step": 78848, "epoch": 885, "lr": 8.074900240737284e-05}
{"train_loss": 0.2789059579372406, "global_step": 78849, "epoch": 885, "lr": 8.074854525825759e-05}
{"train_loss": 0.24691131711006165, "global_step": 78850, "epoch": 885, "lr": 8.074808810500855e-05}
{"train_loss": 0.21148313581943512, "global_step": 78851, "epoch": 885, "lr": 8.07476309476258e-05}
{"train_loss": 0.15365736186504364, "global_step": 78852, "epoch": 885, "lr": 8.074717378610944e-05}
{"train_loss": 0.1939482203360354, "global_step": 78853, "epoch": 885, "lr": 8.074671662045947e-05, "val_loss": 4.213813304901123, "train_action_mse_error": 21.193939208984375}
{"train_loss": 0.1561126559972763, "global_step": 78854, "epoch": 886, "lr": 8.074625945067598e-05}
{"train_loss": 0.1552501618862152, "global_step": 78855, "epoch": 886, "lr": 8.074580227675902e-05}
{"train_loss": 0.14687298238277435, "global_step": 78856, "epoch": 886, "lr": 8.074534509870868e-05}
{"train_loss": 0.20156455039978027, "global_step": 78857, "epoch": 886, "lr": 8.074488791652501e-05}
{"train_loss": 0.15701398253440857, "global_step": 78858, "epoch": 886, "lr": 8.074443073020805e-05}
{"train_loss": 0.1847803294658661, "global_step": 78859, "epoch": 886, "lr": 8.074397353975788e-05}
{"train_loss": 0.22149820625782013, "global_step": 78860, "epoch": 886, "lr": 8.074351634517457e-05}
{"train_loss": 0.2426861971616745, "global_step": 78861, "epoch": 886, "lr": 8.074305914645814e-05}
{"train_loss": 0.28894728422164917, "global_step": 78862, "epoch": 886, "lr": 8.07426019436087e-05}
{"train_loss": 0.21387599408626556, "global_step": 78863, "epoch": 886, "lr": 8.07421447366263e-05}
{"train_loss": 0.11667250096797943, "global_step": 78864, "epoch": 886, "lr": 8.074168752551096e-05}
{"train_loss": 0.13875818252563477, "global_step": 78865, "epoch": 886, "lr": 8.07412303102628e-05}
{"train_loss": 0.16498614847660065, "global_step": 78866, "epoch": 886, "lr": 8.074077309088186e-05}
{"train_loss": 0.2929770052433014, "global_step": 78867, "epoch": 886, "lr": 8.07403158673682e-05}
{"train_loss": 0.23799078166484833, "global_step": 78868, "epoch": 886, "lr": 8.073985863972185e-05}
{"train_loss": 0.17169822752475739, "global_step": 78869, "epoch": 886, "lr": 8.073940140794291e-05}
{"train_loss": 0.17333918809890747, "global_step": 78870, "epoch": 886, "lr": 8.073894417203145e-05}
{"train_loss": 0.1949823796749115, "global_step": 78871, "epoch": 886, "lr": 8.073848693198751e-05}
{"train_loss": 0.2505902051925659, "global_step": 78872, "epoch": 886, "lr": 8.073802968781114e-05}
{"train_loss": 0.20100879669189453, "global_step": 78873, "epoch": 886, "lr": 8.073757243950244e-05}
{"train_loss": 0.2667417526245117, "global_step": 78874, "epoch": 886, "lr": 8.073711518706142e-05}
{"train_loss": 0.2291242629289627, "global_step": 78875, "epoch": 886, "lr": 8.073665793048819e-05}
{"train_loss": 0.18955521285533905, "global_step": 78876, "epoch": 886, "lr": 8.073620066978279e-05}
{"train_loss": 0.2597784698009491, "global_step": 78877, "epoch": 886, "lr": 8.073574340494527e-05}
{"train_loss": 0.18895772099494934, "global_step": 78878, "epoch": 886, "lr": 8.073528613597572e-05}
{"train_loss": 0.2279282957315445, "global_step": 78879, "epoch": 886, "lr": 8.073482886287417e-05}
{"train_loss": 0.20726917684078217, "global_step": 78880, "epoch": 886, "lr": 8.073437158564072e-05}
{"train_loss": 0.19231554865837097, "global_step": 78881, "epoch": 886, "lr": 8.07339143042754e-05}
{"train_loss": 0.19211718440055847, "global_step": 78882, "epoch": 886, "lr": 8.073345701877826e-05}
{"train_loss": 0.2509578466415405, "global_step": 78883, "epoch": 886, "lr": 8.07329997291494e-05}
{"train_loss": 0.2419685274362564, "global_step": 78884, "epoch": 886, "lr": 8.073254243538884e-05}
{"train_loss": 0.17468783259391785, "global_step": 78885, "epoch": 886, "lr": 8.073208513749669e-05}
{"train_loss": 0.22739681601524353, "global_step": 78886, "epoch": 886, "lr": 8.073162783547298e-05}
{"train_loss": 0.14422371983528137, "global_step": 78887, "epoch": 886, "lr": 8.073117052931777e-05}
{"train_loss": 0.1477600634098053, "global_step": 78888, "epoch": 886, "lr": 8.073071321903112e-05}
{"train_loss": 0.2798188626766205, "global_step": 78889, "epoch": 886, "lr": 8.073025590461311e-05}
{"train_loss": 0.20253495872020721, "global_step": 78890, "epoch": 886, "lr": 8.072979858606379e-05}
{"train_loss": 0.16884370148181915, "global_step": 78891, "epoch": 886, "lr": 8.072934126338324e-05}
{"train_loss": 0.19906798005104065, "global_step": 78892, "epoch": 886, "lr": 8.07288839365715e-05}
{"train_loss": 0.20179621875286102, "global_step": 78893, "epoch": 886, "lr": 8.072842660562862e-05}
{"train_loss": 0.1722862869501114, "global_step": 78894, "epoch": 886, "lr": 8.072796927055468e-05}
{"train_loss": 0.17658308148384094, "global_step": 78895, "epoch": 886, "lr": 8.072751193134976e-05}
{"train_loss": 0.21119362115859985, "global_step": 78896, "epoch": 886, "lr": 8.072705458801388e-05}
{"train_loss": 0.21976633369922638, "global_step": 78897, "epoch": 886, "lr": 8.072659724054712e-05}
{"train_loss": 0.18634049594402313, "global_step": 78898, "epoch": 886, "lr": 8.072613988894956e-05}
{"train_loss": 0.26229968667030334, "global_step": 78899, "epoch": 886, "lr": 8.072568253322124e-05}
{"train_loss": 0.3383566439151764, "global_step": 78900, "epoch": 886, "lr": 8.072522517336221e-05}
{"train_loss": 0.18642613291740417, "global_step": 78901, "epoch": 886, "lr": 8.072476780937257e-05}
{"train_loss": 0.16103529930114746, "global_step": 78902, "epoch": 886, "lr": 8.072431044125234e-05}
{"train_loss": 0.21437808871269226, "global_step": 78903, "epoch": 886, "lr": 8.072385306900161e-05}
{"train_loss": 0.14024688303470612, "global_step": 78904, "epoch": 886, "lr": 8.072339569262044e-05}
{"train_loss": 0.1606058031320572, "global_step": 78905, "epoch": 886, "lr": 8.072293831210887e-05}
{"train_loss": 0.23497533798217773, "global_step": 78906, "epoch": 886, "lr": 8.072248092746698e-05}
{"train_loss": 0.17994479835033417, "global_step": 78907, "epoch": 886, "lr": 8.072202353869483e-05}
{"train_loss": 0.19823825359344482, "global_step": 78908, "epoch": 886, "lr": 8.072156614579247e-05}
{"train_loss": 0.2446175217628479, "global_step": 78909, "epoch": 886, "lr": 8.072110874875999e-05}
{"train_loss": 0.1282852292060852, "global_step": 78910, "epoch": 886, "lr": 8.072065134759741e-05}
{"train_loss": 0.19829218089580536, "global_step": 78911, "epoch": 886, "lr": 8.072019394230483e-05}
{"train_loss": 0.27389979362487793, "global_step": 78912, "epoch": 886, "lr": 8.071973653288229e-05}
{"train_loss": 0.16209131479263306, "global_step": 78913, "epoch": 886, "lr": 8.071927911932983e-05}
{"train_loss": 0.2717532813549042, "global_step": 78914, "epoch": 886, "lr": 8.071882170164757e-05}
{"train_loss": 0.24166326224803925, "global_step": 78915, "epoch": 886, "lr": 8.071836427983554e-05}
{"train_loss": 0.18739932775497437, "global_step": 78916, "epoch": 886, "lr": 8.071790685389377e-05}
{"train_loss": 0.20701192319393158, "global_step": 78917, "epoch": 886, "lr": 8.071744942382238e-05}
{"train_loss": 0.20484870672225952, "global_step": 78918, "epoch": 886, "lr": 8.071699198962139e-05}
{"train_loss": 0.21263928711414337, "global_step": 78919, "epoch": 886, "lr": 8.071653455129089e-05}
{"train_loss": 0.25034230947494507, "global_step": 78920, "epoch": 886, "lr": 8.071607710883091e-05}
{"train_loss": 0.2154734879732132, "global_step": 78921, "epoch": 886, "lr": 8.071561966224152e-05}
{"train_loss": 0.18268853425979614, "global_step": 78922, "epoch": 886, "lr": 8.071516221152282e-05}
{"train_loss": 0.25482502579689026, "global_step": 78923, "epoch": 886, "lr": 8.071470475667483e-05}
{"train_loss": 0.1122569739818573, "global_step": 78924, "epoch": 886, "lr": 8.071424729769759e-05}
{"train_loss": 0.28458163142204285, "global_step": 78925, "epoch": 886, "lr": 8.071378983459123e-05}
{"train_loss": 0.2599897086620331, "global_step": 78926, "epoch": 886, "lr": 8.071333236735576e-05}
{"train_loss": 0.2358596920967102, "global_step": 78927, "epoch": 886, "lr": 8.071287489599126e-05}
{"train_loss": 0.24620506167411804, "global_step": 78928, "epoch": 886, "lr": 8.07124174204978e-05}
{"train_loss": 0.24736876785755157, "global_step": 78929, "epoch": 886, "lr": 8.071195994087541e-05}
{"train_loss": 0.20185747742652893, "global_step": 78930, "epoch": 886, "lr": 8.071150245712417e-05}
{"train_loss": 0.23422794044017792, "global_step": 78931, "epoch": 886, "lr": 8.071104496924416e-05}
{"train_loss": 0.18030138313770294, "global_step": 78932, "epoch": 886, "lr": 8.071058747723542e-05}
{"train_loss": 0.22627981007099152, "global_step": 78933, "epoch": 886, "lr": 8.071012998109802e-05}
{"train_loss": 0.19411759078502655, "global_step": 78934, "epoch": 886, "lr": 8.070967248083201e-05}
{"train_loss": 0.22420138120651245, "global_step": 78935, "epoch": 886, "lr": 8.070921497643745e-05}
{"train_loss": 0.20666001737117767, "global_step": 78936, "epoch": 886, "lr": 8.070875746791441e-05}
{"train_loss": 0.23081199824810028, "global_step": 78937, "epoch": 886, "lr": 8.070829995526297e-05}
{"train_loss": 0.22092115879058838, "global_step": 78938, "epoch": 886, "lr": 8.070784243848317e-05}
{"train_loss": 0.1893801987171173, "global_step": 78939, "epoch": 886, "lr": 8.070738491757507e-05}
{"train_loss": 0.20739443600177765, "global_step": 78940, "epoch": 886, "lr": 8.070692739253874e-05}
{"train_loss": 0.19333069026470184, "global_step": 78941, "epoch": 886, "lr": 8.070646986337422e-05}
{"train_loss": 0.2066003033954106, "global_step": 78942, "epoch": 886, "lr": 8.070601233008161e-05, "val_loss": 3.8891656398773193}
{"train_loss": 0.20015612244606018, "global_step": 78943, "epoch": 887, "lr": 8.070555479266094e-05}
{"train_loss": 0.11993835121393204, "global_step": 78944, "epoch": 887, "lr": 8.070509725111228e-05}
{"train_loss": 0.21343828737735748, "global_step": 78945, "epoch": 887, "lr": 8.070463970543571e-05}
{"train_loss": 0.22943991422653198, "global_step": 78946, "epoch": 887, "lr": 8.070418215563127e-05}
{"train_loss": 0.10824424773454666, "global_step": 78947, "epoch": 887, "lr": 8.070372460169901e-05}
{"train_loss": 0.1875656098127365, "global_step": 78948, "epoch": 887, "lr": 8.070326704363904e-05}
{"train_loss": 0.16306601464748383, "global_step": 78949, "epoch": 887, "lr": 8.070280948145135e-05}
{"train_loss": 0.2519094944000244, "global_step": 78950, "epoch": 887, "lr": 8.070235191513607e-05}
{"train_loss": 0.2041768580675125, "global_step": 78951, "epoch": 887, "lr": 8.070189434469323e-05}
{"train_loss": 0.1725870668888092, "global_step": 78952, "epoch": 887, "lr": 8.070143677012289e-05}
{"train_loss": 0.1729983389377594, "global_step": 78953, "epoch": 887, "lr": 8.070097919142512e-05}
{"train_loss": 0.18349917232990265, "global_step": 78954, "epoch": 887, "lr": 8.070052160859998e-05}
{"train_loss": 0.18880410492420197, "global_step": 78955, "epoch": 887, "lr": 8.070006402164752e-05}
{"train_loss": 0.12321116030216217, "global_step": 78956, "epoch": 887, "lr": 8.069960643056781e-05}
{"train_loss": 0.1370336264371872, "global_step": 78957, "epoch": 887, "lr": 8.069914883536091e-05}
{"train_loss": 0.18215279281139374, "global_step": 78958, "epoch": 887, "lr": 8.069869123602689e-05}
{"train_loss": 0.19828450679779053, "global_step": 78959, "epoch": 887, "lr": 8.069823363256581e-05}
{"train_loss": 0.21436810493469238, "global_step": 78960, "epoch": 887, "lr": 8.069777602497771e-05}
{"train_loss": 0.16100428998470306, "global_step": 78961, "epoch": 887, "lr": 8.069731841326269e-05}
{"train_loss": 0.18055348098278046, "global_step": 78962, "epoch": 887, "lr": 8.069686079742077e-05}
{"train_loss": 0.2084241807460785, "global_step": 78963, "epoch": 887, "lr": 8.069640317745204e-05}
{"train_loss": 0.1986762285232544, "global_step": 78964, "epoch": 887, "lr": 8.069594555335656e-05}
{"train_loss": 0.161937415599823, "global_step": 78965, "epoch": 887, "lr": 8.069548792513438e-05}
{"train_loss": 0.1761394739151001, "global_step": 78966, "epoch": 887, "lr": 8.069503029278556e-05}
{"train_loss": 0.1907275915145874, "global_step": 78967, "epoch": 887, "lr": 8.069457265631016e-05}
{"train_loss": 0.24107694625854492, "global_step": 78968, "epoch": 887, "lr": 8.069411501570826e-05}
{"train_loss": 0.2022276669740677, "global_step": 78969, "epoch": 887, "lr": 8.069365737097991e-05}
{"train_loss": 0.13034214079380035, "global_step": 78970, "epoch": 887, "lr": 8.069319972212517e-05}
{"train_loss": 0.24264539778232574, "global_step": 78971, "epoch": 887, "lr": 8.06927420691441e-05}
{"train_loss": 0.19518814980983734, "global_step": 78972, "epoch": 887, "lr": 8.069228441203677e-05}
{"train_loss": 0.15476179122924805, "global_step": 78973, "epoch": 887, "lr": 8.069182675080325e-05}
{"train_loss": 0.18395327031612396, "global_step": 78974, "epoch": 887, "lr": 8.069136908544357e-05}
{"train_loss": 0.25321725010871887, "global_step": 78975, "epoch": 887, "lr": 8.069091141595782e-05}
{"train_loss": 0.2517443001270294, "global_step": 78976, "epoch": 887, "lr": 8.069045374234604e-05}
{"train_loss": 0.23711609840393066, "global_step": 78977, "epoch": 887, "lr": 8.06899960646083e-05}
{"train_loss": 0.32863545417785645, "global_step": 78978, "epoch": 887, "lr": 8.068953838274466e-05}
{"train_loss": 0.1961415857076645, "global_step": 78979, "epoch": 887, "lr": 8.06890806967552e-05}
{"train_loss": 0.16931751370429993, "global_step": 78980, "epoch": 887, "lr": 8.068862300663997e-05}
{"train_loss": 0.18240414559841156, "global_step": 78981, "epoch": 887, "lr": 8.068816531239902e-05}
{"train_loss": 0.22014129161834717, "global_step": 78982, "epoch": 887, "lr": 8.068770761403243e-05}
{"train_loss": 0.209889754652977, "global_step": 78983, "epoch": 887, "lr": 8.068724991154024e-05}
{"train_loss": 0.17700070142745972, "global_step": 78984, "epoch": 887, "lr": 8.068679220492252e-05}
{"train_loss": 0.21821050345897675, "global_step": 78985, "epoch": 887, "lr": 8.068633449417935e-05}
{"train_loss": 0.26981279253959656, "global_step": 78986, "epoch": 887, "lr": 8.068587677931077e-05}
{"train_loss": 0.15313416719436646, "global_step": 78987, "epoch": 887, "lr": 8.068541906031685e-05}
{"train_loss": 0.18052689731121063, "global_step": 78988, "epoch": 887, "lr": 8.068496133719763e-05}
{"train_loss": 0.20365756750106812, "global_step": 78989, "epoch": 887, "lr": 8.068450360995322e-05}
{"train_loss": 0.15116283297538757, "global_step": 78990, "epoch": 887, "lr": 8.068404587858364e-05}
{"train_loss": 0.23226958513259888, "global_step": 78991, "epoch": 887, "lr": 8.068358814308896e-05}
{"train_loss": 0.26376548409461975, "global_step": 78992, "epoch": 887, "lr": 8.068313040346926e-05}
{"train_loss": 0.2581215798854828, "global_step": 78993, "epoch": 887, "lr": 8.068267265972458e-05}
{"train_loss": 0.27460649609565735, "global_step": 78994, "epoch": 887, "lr": 8.068221491185497e-05}
{"train_loss": 0.2105371356010437, "global_step": 78995, "epoch": 887, "lr": 8.068175715986052e-05}
{"train_loss": 0.24811404943466187, "global_step": 78996, "epoch": 887, "lr": 8.06812994037413e-05}
{"train_loss": 0.23534157872200012, "global_step": 78997, "epoch": 887, "lr": 8.068084164349733e-05}
{"train_loss": 0.2309601902961731, "global_step": 78998, "epoch": 887, "lr": 8.06803838791287e-05}
{"train_loss": 0.2222348004579544, "global_step": 78999, "epoch": 887, "lr": 8.067992611063549e-05}
{"train_loss": 0.17511996626853943, "global_step": 79000, "epoch": 887, "lr": 8.06794683380177e-05}
{"train_loss": 0.21991975605487823, "global_step": 79001, "epoch": 887, "lr": 8.067901056127545e-05}
{"train_loss": 0.23431500792503357, "global_step": 79002, "epoch": 887, "lr": 8.067855278040878e-05}
{"train_loss": 0.3784063756465912, "global_step": 79003, "epoch": 887, "lr": 8.067809499541776e-05}
{"train_loss": 0.21536549925804138, "global_step": 79004, "epoch": 887, "lr": 8.067763720630244e-05}
{"train_loss": 0.22091035544872284, "global_step": 79005, "epoch": 887, "lr": 8.067717941306287e-05}
{"train_loss": 0.20136648416519165, "global_step": 79006, "epoch": 887, "lr": 8.067672161569915e-05}
{"train_loss": 0.14539381861686707, "global_step": 79007, "epoch": 887, "lr": 8.067626381421129e-05}
{"train_loss": 0.16295133531093597, "global_step": 79008, "epoch": 887, "lr": 8.06758060085994e-05}
{"train_loss": 0.20876355469226837, "global_step": 79009, "epoch": 887, "lr": 8.067534819886354e-05}
{"train_loss": 0.14419960975646973, "global_step": 79010, "epoch": 887, "lr": 8.067489038500372e-05}
{"train_loss": 0.22032631933689117, "global_step": 79011, "epoch": 887, "lr": 8.067443256702005e-05}
{"train_loss": 0.18534083664417267, "global_step": 79012, "epoch": 887, "lr": 8.067397474491257e-05}
{"train_loss": 0.21869726479053497, "global_step": 79013, "epoch": 887, "lr": 8.067351691868135e-05}
{"train_loss": 0.1688697338104248, "global_step": 79014, "epoch": 887, "lr": 8.067305908832645e-05}
{"train_loss": 0.16975286602973938, "global_step": 79015, "epoch": 887, "lr": 8.067260125384793e-05}
{"train_loss": 0.19265210628509521, "global_step": 79016, "epoch": 887, "lr": 8.067214341524585e-05}
{"train_loss": 0.1718563288450241, "global_step": 79017, "epoch": 887, "lr": 8.067168557252029e-05}
{"train_loss": 0.25415506958961487, "global_step": 79018, "epoch": 887, "lr": 8.067122772567129e-05}
{"train_loss": 0.2540134787559509, "global_step": 79019, "epoch": 887, "lr": 8.06707698746989e-05}
{"train_loss": 0.28182992339134216, "global_step": 79020, "epoch": 887, "lr": 8.067031201960323e-05}
{"train_loss": 0.17658179998397827, "global_step": 79021, "epoch": 887, "lr": 8.066985416038428e-05}
{"train_loss": 0.19489823281764984, "global_step": 79022, "epoch": 887, "lr": 8.066939629704215e-05}
{"train_loss": 0.16954994201660156, "global_step": 79023, "epoch": 887, "lr": 8.06689384295769e-05}
{"train_loss": 0.14226429164409637, "global_step": 79024, "epoch": 887, "lr": 8.066848055798858e-05}
{"train_loss": 0.16694337129592896, "global_step": 79025, "epoch": 887, "lr": 8.066802268227726e-05}
{"train_loss": 0.18697094917297363, "global_step": 79026, "epoch": 887, "lr": 8.066756480244299e-05}
{"train_loss": 0.23522096872329712, "global_step": 79027, "epoch": 887, "lr": 8.066710691848584e-05}
{"train_loss": 0.21978788077831268, "global_step": 79028, "epoch": 887, "lr": 8.066664903040588e-05}
{"train_loss": 0.1539735645055771, "global_step": 79029, "epoch": 887, "lr": 8.066619113820317e-05}
{"train_loss": 0.16148759424686432, "global_step": 79030, "epoch": 887, "lr": 8.066573324187775e-05}
{"train_loss": 0.2022268557816409, "global_step": 79031, "epoch": 887, "lr": 8.066527534142971e-05, "val_loss": 3.9029643535614014}
{"train_loss": 0.1715395152568817, "global_step": 79032, "epoch": 888, "lr": 8.066481743685907e-05}
{"train_loss": 0.2606983780860901, "global_step": 79033, "epoch": 888, "lr": 8.066435952816593e-05}
{"train_loss": 0.18202190101146698, "global_step": 79034, "epoch": 888, "lr": 8.066390161535036e-05}
{"train_loss": 0.2639505863189697, "global_step": 79035, "epoch": 888, "lr": 8.066344369841239e-05}
{"train_loss": 0.1560993790626526, "global_step": 79036, "epoch": 888, "lr": 8.066298577735209e-05}
{"train_loss": 0.30204206705093384, "global_step": 79037, "epoch": 888, "lr": 8.066252785216953e-05}
{"train_loss": 0.20088940858840942, "global_step": 79038, "epoch": 888, "lr": 8.066206992286477e-05}
{"train_loss": 0.15413306653499603, "global_step": 79039, "epoch": 888, "lr": 8.066161198943786e-05}
{"train_loss": 0.22666938602924347, "global_step": 79040, "epoch": 888, "lr": 8.066115405188887e-05}
{"train_loss": 0.1851048469543457, "global_step": 79041, "epoch": 888, "lr": 8.066069611021787e-05}
{"train_loss": 0.1680404096841812, "global_step": 79042, "epoch": 888, "lr": 8.066023816442491e-05}
{"train_loss": 0.23146957159042358, "global_step": 79043, "epoch": 888, "lr": 8.065978021451007e-05}
{"train_loss": 0.2270025759935379, "global_step": 79044, "epoch": 888, "lr": 8.065932226047337e-05}
{"train_loss": 0.29810917377471924, "global_step": 79045, "epoch": 888, "lr": 8.065886430231491e-05}
{"train_loss": 0.17787399888038635, "global_step": 79046, "epoch": 888, "lr": 8.065840634003473e-05}
{"train_loss": 0.16153816878795624, "global_step": 79047, "epoch": 888, "lr": 8.06579483736329e-05}
{"train_loss": 0.1640930026769638, "global_step": 79048, "epoch": 888, "lr": 8.06574904031095e-05}
{"train_loss": 0.1413087099790573, "global_step": 79049, "epoch": 888, "lr": 8.065703242846456e-05}
{"train_loss": 0.30008503794670105, "global_step": 79050, "epoch": 888, "lr": 8.065657444969816e-05}
{"train_loss": 0.1842169314622879, "global_step": 79051, "epoch": 888, "lr": 8.065611646681035e-05}
{"train_loss": 0.22166642546653748, "global_step": 79052, "epoch": 888, "lr": 8.065565847980121e-05}
{"train_loss": 0.22064192593097687, "global_step": 79053, "epoch": 888, "lr": 8.06552004886708e-05}
{"train_loss": 0.1586453765630722, "global_step": 79054, "epoch": 888, "lr": 8.065474249341914e-05}
{"train_loss": 0.28493639826774597, "global_step": 79055, "epoch": 888, "lr": 8.065428449404635e-05}
{"train_loss": 0.13908681273460388, "global_step": 79056, "epoch": 888, "lr": 8.065382649055246e-05}
{"train_loss": 0.1536695510149002, "global_step": 79057, "epoch": 888, "lr": 8.065336848293752e-05}
{"train_loss": 0.20316284894943237, "global_step": 79058, "epoch": 888, "lr": 8.065291047120163e-05}
{"train_loss": 0.14852097630500793, "global_step": 79059, "epoch": 888, "lr": 8.065245245534482e-05}
{"train_loss": 0.16330762207508087, "global_step": 79060, "epoch": 888, "lr": 8.065199443536718e-05}
{"train_loss": 0.25983235239982605, "global_step": 79061, "epoch": 888, "lr": 8.065153641126873e-05}
{"train_loss": 0.16339972615242004, "global_step": 79062, "epoch": 888, "lr": 8.065107838304955e-05}
{"train_loss": 0.14649595320224762, "global_step": 79063, "epoch": 888, "lr": 8.065062035070972e-05}
{"train_loss": 0.26125890016555786, "global_step": 79064, "epoch": 888, "lr": 8.065016231424929e-05}
{"train_loss": 0.20289988815784454, "global_step": 79065, "epoch": 888, "lr": 8.064970427366831e-05}
{"train_loss": 0.15541434288024902, "global_step": 79066, "epoch": 888, "lr": 8.064924622896686e-05}
{"train_loss": 0.23054127395153046, "global_step": 79067, "epoch": 888, "lr": 8.064878818014499e-05}
{"train_loss": 0.21854186058044434, "global_step": 79068, "epoch": 888, "lr": 8.064833012720275e-05}
{"train_loss": 0.25327950716018677, "global_step": 79069, "epoch": 888, "lr": 8.064787207014022e-05}
{"train_loss": 0.2262827306985855, "global_step": 79070, "epoch": 888, "lr": 8.064741400895745e-05}
{"train_loss": 0.21689948439598083, "global_step": 79071, "epoch": 888, "lr": 8.064695594365453e-05}
{"train_loss": 0.1326044499874115, "global_step": 79072, "epoch": 888, "lr": 8.06464978742315e-05}
{"train_loss": 0.2241033911705017, "global_step": 79073, "epoch": 888, "lr": 8.064603980068841e-05}
{"train_loss": 0.16363927721977234, "global_step": 79074, "epoch": 888, "lr": 8.064558172302534e-05}
{"train_loss": 0.2202630490064621, "global_step": 79075, "epoch": 888, "lr": 8.064512364124234e-05}
{"train_loss": 0.21989814937114716, "global_step": 79076, "epoch": 888, "lr": 8.064466555533949e-05}
{"train_loss": 0.12008742243051529, "global_step": 79077, "epoch": 888, "lr": 8.064420746531683e-05}
{"train_loss": 0.21876367926597595, "global_step": 79078, "epoch": 888, "lr": 8.064374937117443e-05}
{"train_loss": 0.22637513279914856, "global_step": 79079, "epoch": 888, "lr": 8.064329127291234e-05}
{"train_loss": 0.1850862205028534, "global_step": 79080, "epoch": 888, "lr": 8.064283317053065e-05}
{"train_loss": 0.34761300683021545, "global_step": 79081, "epoch": 888, "lr": 8.064237506402939e-05}
{"train_loss": 0.2149992436170578, "global_step": 79082, "epoch": 888, "lr": 8.064191695340863e-05}
{"train_loss": 0.2057514786720276, "global_step": 79083, "epoch": 888, "lr": 8.064145883866847e-05}
{"train_loss": 0.18814481794834137, "global_step": 79084, "epoch": 888, "lr": 8.064100071980891e-05}
{"train_loss": 0.16799645125865936, "global_step": 79085, "epoch": 888, "lr": 8.064054259683004e-05}
{"train_loss": 0.2443036437034607, "global_step": 79086, "epoch": 888, "lr": 8.064008446973195e-05}
{"train_loss": 0.16220583021640778, "global_step": 79087, "epoch": 888, "lr": 8.063962633851465e-05}
{"train_loss": 0.2501128315925598, "global_step": 79088, "epoch": 888, "lr": 8.063916820317823e-05}
{"train_loss": 0.16945978999137878, "global_step": 79089, "epoch": 888, "lr": 8.063871006372274e-05}
{"train_loss": 0.18785811960697174, "global_step": 79090, "epoch": 888, "lr": 8.063825192014827e-05}
{"train_loss": 0.26584377884864807, "global_step": 79091, "epoch": 888, "lr": 8.063779377245484e-05}
{"train_loss": 0.2598523795604706, "global_step": 79092, "epoch": 888, "lr": 8.063733562064255e-05}
{"train_loss": 0.2002882957458496, "global_step": 79093, "epoch": 888, "lr": 8.063687746471142e-05}
{"train_loss": 0.18389447033405304, "global_step": 79094, "epoch": 888, "lr": 8.063641930466157e-05}
{"train_loss": 0.19542011618614197, "global_step": 79095, "epoch": 888, "lr": 8.0635961140493e-05}
{"train_loss": 0.19665347039699554, "global_step": 79096, "epoch": 888, "lr": 8.063550297220581e-05}
{"train_loss": 0.15214766561985016, "global_step": 79097, "epoch": 888, "lr": 8.063504479980004e-05}
{"train_loss": 0.24875609576702118, "global_step": 79098, "epoch": 888, "lr": 8.063458662327577e-05}
{"train_loss": 0.11879774928092957, "global_step": 79099, "epoch": 888, "lr": 8.063412844263305e-05}
{"train_loss": 0.24539917707443237, "global_step": 79100, "epoch": 888, "lr": 8.063367025787194e-05}
{"train_loss": 0.16298896074295044, "global_step": 79101, "epoch": 888, "lr": 8.063321206899252e-05}
{"train_loss": 0.1645781695842743, "global_step": 79102, "epoch": 888, "lr": 8.063275387599483e-05}
{"train_loss": 0.16382122039794922, "global_step": 79103, "epoch": 888, "lr": 8.063229567887893e-05}
{"train_loss": 0.1613483875989914, "global_step": 79104, "epoch": 888, "lr": 8.063183747764491e-05}
{"train_loss": 0.22432033717632294, "global_step": 79105, "epoch": 888, "lr": 8.063137927229279e-05}
{"train_loss": 0.21091338992118835, "global_step": 79106, "epoch": 888, "lr": 8.063092106282267e-05}
{"train_loss": 0.1613249033689499, "global_step": 79107, "epoch": 888, "lr": 8.063046284923459e-05}
{"train_loss": 0.2445315271615982, "global_step": 79108, "epoch": 888, "lr": 8.063000463152863e-05}
{"train_loss": 0.16684912145137787, "global_step": 79109, "epoch": 888, "lr": 8.062954640970484e-05}
{"train_loss": 0.2385897934436798, "global_step": 79110, "epoch": 888, "lr": 8.062908818376326e-05}
{"train_loss": 0.23065201938152313, "global_step": 79111, "epoch": 888, "lr": 8.062862995370399e-05}
{"train_loss": 0.1393749862909317, "global_step": 79112, "epoch": 888, "lr": 8.062817171952705e-05}
{"train_loss": 0.15793603658676147, "global_step": 79113, "epoch": 888, "lr": 8.062771348123255e-05}
{"train_loss": 0.19230128824710846, "global_step": 79114, "epoch": 888, "lr": 8.062725523882053e-05}
{"train_loss": 0.24249443411827087, "global_step": 79115, "epoch": 888, "lr": 8.062679699229103e-05}
{"train_loss": 0.22771504521369934, "global_step": 79116, "epoch": 888, "lr": 8.062633874164415e-05}
{"train_loss": 0.18639343976974487, "global_step": 79117, "epoch": 888, "lr": 8.062588048687991e-05}
{"train_loss": 0.19247226417064667, "global_step": 79118, "epoch": 888, "lr": 8.062542222799841e-05}
{"train_loss": 0.1750725507736206, "global_step": 79119, "epoch": 888, "lr": 8.06249639649997e-05}
{"train_loss": 0.20171526568324377, "global_step": 79120, "epoch": 888, "lr": 8.062450569788383e-05, "val_loss": 3.8572962284088135}
{"train_loss": 0.182805135846138, "global_step": 79121, "epoch": 889, "lr": 8.062404742665086e-05}
{"train_loss": 0.09066052734851837, "global_step": 79122, "epoch": 889, "lr": 8.062358915130086e-05}
{"train_loss": 0.18132060766220093, "global_step": 79123, "epoch": 889, "lr": 8.06231308718339e-05}
{"train_loss": 0.20750612020492554, "global_step": 79124, "epoch": 889, "lr": 8.062267258825003e-05}
{"train_loss": 0.2090381681919098, "global_step": 79125, "epoch": 889, "lr": 8.062221430054931e-05}
{"train_loss": 0.24653632938861847, "global_step": 79126, "epoch": 889, "lr": 8.062175600873181e-05}
{"train_loss": 0.19076019525527954, "global_step": 79127, "epoch": 889, "lr": 8.062129771279761e-05}
{"train_loss": 0.27721965312957764, "global_step": 79128, "epoch": 889, "lr": 8.062083941274672e-05}
{"train_loss": 0.1829710304737091, "global_step": 79129, "epoch": 889, "lr": 8.062038110857924e-05}
{"train_loss": 0.16005316376686096, "global_step": 79130, "epoch": 889, "lr": 8.061992280029522e-05}
{"train_loss": 0.1309310495853424, "global_step": 79131, "epoch": 889, "lr": 8.061946448789473e-05}
{"train_loss": 0.15027101337909698, "global_step": 79132, "epoch": 889, "lr": 8.061900617137783e-05}
{"train_loss": 0.17165222764015198, "global_step": 79133, "epoch": 889, "lr": 8.061854785074458e-05}
{"train_loss": 0.22842788696289062, "global_step": 79134, "epoch": 889, "lr": 8.061808952599504e-05}
{"train_loss": 0.22549739480018616, "global_step": 79135, "epoch": 889, "lr": 8.061763119712925e-05}
{"train_loss": 0.2385455071926117, "global_step": 79136, "epoch": 889, "lr": 8.061717286414732e-05}
{"train_loss": 0.1606779545545578, "global_step": 79137, "epoch": 889, "lr": 8.061671452704927e-05}
{"train_loss": 0.2725414037704468, "global_step": 79138, "epoch": 889, "lr": 8.061625618583518e-05}
{"train_loss": 0.22941821813583374, "global_step": 79139, "epoch": 889, "lr": 8.061579784050509e-05}
{"train_loss": 0.21294210851192474, "global_step": 79140, "epoch": 889, "lr": 8.06153394910591e-05}
{"train_loss": 0.1732863038778305, "global_step": 79141, "epoch": 889, "lr": 8.061488113749722e-05}
{"train_loss": 0.1954692155122757, "global_step": 79142, "epoch": 889, "lr": 8.061442277981959e-05}
{"train_loss": 0.21365880966186523, "global_step": 79143, "epoch": 889, "lr": 8.061396441802619e-05}
{"train_loss": 0.16101981699466705, "global_step": 79144, "epoch": 889, "lr": 8.061350605211713e-05}
{"train_loss": 0.16926279664039612, "global_step": 79145, "epoch": 889, "lr": 8.061304768209245e-05}
{"train_loss": 0.23948155343532562, "global_step": 79146, "epoch": 889, "lr": 8.061258930795222e-05}
{"train_loss": 0.2399015873670578, "global_step": 79147, "epoch": 889, "lr": 8.061213092969651e-05}
{"train_loss": 0.18816213309764862, "global_step": 79148, "epoch": 889, "lr": 8.061167254732536e-05}
{"train_loss": 0.2682777941226959, "global_step": 79149, "epoch": 889, "lr": 8.061121416083886e-05}
{"train_loss": 0.2173667699098587, "global_step": 79150, "epoch": 889, "lr": 8.061075577023702e-05}
{"train_loss": 0.16931426525115967, "global_step": 79151, "epoch": 889, "lr": 8.061029737551998e-05}
{"train_loss": 0.1920858919620514, "global_step": 79152, "epoch": 889, "lr": 8.060983897668772e-05}
{"train_loss": 0.19657504558563232, "global_step": 79153, "epoch": 889, "lr": 8.060938057374037e-05}
{"train_loss": 0.1781833916902542, "global_step": 79154, "epoch": 889, "lr": 8.060892216667796e-05}
{"train_loss": 0.24542613327503204, "global_step": 79155, "epoch": 889, "lr": 8.060846375550054e-05}
{"train_loss": 0.14388616383075714, "global_step": 79156, "epoch": 889, "lr": 8.060800534020819e-05}
{"train_loss": 0.15802595019340515, "global_step": 79157, "epoch": 889, "lr": 8.060754692080097e-05}
{"train_loss": 0.1943962723016739, "global_step": 79158, "epoch": 889, "lr": 8.060708849727893e-05}
{"train_loss": 0.17101512849330902, "global_step": 79159, "epoch": 889, "lr": 8.060663006964212e-05}
{"train_loss": 0.23878134787082672, "global_step": 79160, "epoch": 889, "lr": 8.060617163789066e-05}
{"train_loss": 0.27109289169311523, "global_step": 79161, "epoch": 889, "lr": 8.060571320202457e-05}
{"train_loss": 0.2330959290266037, "global_step": 79162, "epoch": 889, "lr": 8.06052547620439e-05}
{"train_loss": 0.23281249403953552, "global_step": 79163, "epoch": 889, "lr": 8.060479631794875e-05}
{"train_loss": 0.2565534710884094, "global_step": 79164, "epoch": 889, "lr": 8.060433786973913e-05}
{"train_loss": 0.21481862664222717, "global_step": 79165, "epoch": 889, "lr": 8.060387941741514e-05}
{"train_loss": 0.16020604968070984, "global_step": 79166, "epoch": 889, "lr": 8.060342096097683e-05}
{"train_loss": 0.22172771394252777, "global_step": 79167, "epoch": 889, "lr": 8.060296250042426e-05}
{"train_loss": 0.18471847474575043, "global_step": 79168, "epoch": 889, "lr": 8.060250403575749e-05}
{"train_loss": 0.28933125734329224, "global_step": 79169, "epoch": 889, "lr": 8.060204556697661e-05}
{"train_loss": 0.13447827100753784, "global_step": 79170, "epoch": 889, "lr": 8.060158709408163e-05}
{"train_loss": 0.165145605802536, "global_step": 79171, "epoch": 889, "lr": 8.060112861707265e-05}
{"train_loss": 0.27103322744369507, "global_step": 79172, "epoch": 889, "lr": 8.060067013594972e-05}
{"train_loss": 0.20477475225925446, "global_step": 79173, "epoch": 889, "lr": 8.060021165071291e-05}
{"train_loss": 0.16905635595321655, "global_step": 79174, "epoch": 889, "lr": 8.059975316136228e-05}
{"train_loss": 0.1736312061548233, "global_step": 79175, "epoch": 889, "lr": 8.059929466789789e-05}
{"train_loss": 0.30021822452545166, "global_step": 79176, "epoch": 889, "lr": 8.059883617031978e-05}
{"train_loss": 0.27212873101234436, "global_step": 79177, "epoch": 889, "lr": 8.059837766862802e-05}
{"train_loss": 0.2111271470785141, "global_step": 79178, "epoch": 889, "lr": 8.059791916282269e-05}
{"train_loss": 0.18571406602859497, "global_step": 79179, "epoch": 889, "lr": 8.059746065290386e-05}
{"train_loss": 0.2006531059741974, "global_step": 79180, "epoch": 889, "lr": 8.059700213887156e-05}
{"train_loss": 0.22616960108280182, "global_step": 79181, "epoch": 889, "lr": 8.059654362072587e-05}
{"train_loss": 0.1535804718732834, "global_step": 79182, "epoch": 889, "lr": 8.059608509846685e-05}
{"train_loss": 0.2278176099061966, "global_step": 79183, "epoch": 889, "lr": 8.059562657209455e-05}
{"train_loss": 0.27157071232795715, "global_step": 79184, "epoch": 889, "lr": 8.059516804160905e-05}
{"train_loss": 0.23669348657131195, "global_step": 79185, "epoch": 889, "lr": 8.05947095070104e-05}
{"train_loss": 0.1484495997428894, "global_step": 79186, "epoch": 889, "lr": 8.059425096829865e-05}
{"train_loss": 0.24133679270744324, "global_step": 79187, "epoch": 889, "lr": 8.059379242547389e-05}
{"train_loss": 0.1937982738018036, "global_step": 79188, "epoch": 889, "lr": 8.059333387853618e-05}
{"train_loss": 0.23454219102859497, "global_step": 79189, "epoch": 889, "lr": 8.059287532748555e-05}
{"train_loss": 0.2560771107673645, "global_step": 79190, "epoch": 889, "lr": 8.059241677232209e-05}
{"train_loss": 0.32868492603302, "global_step": 79191, "epoch": 889, "lr": 8.059195821304583e-05}
{"train_loss": 0.21686150133609772, "global_step": 79192, "epoch": 889, "lr": 8.059149964965688e-05}
{"train_loss": 0.2919691205024719, "global_step": 79193, "epoch": 889, "lr": 8.059104108215528e-05}
{"train_loss": 0.1737229824066162, "global_step": 79194, "epoch": 889, "lr": 8.059058251054108e-05}
{"train_loss": 0.1495155692100525, "global_step": 79195, "epoch": 889, "lr": 8.059012393481433e-05}
{"train_loss": 0.1357986479997635, "global_step": 79196, "epoch": 889, "lr": 8.058966535497513e-05}
{"train_loss": 0.19391301274299622, "global_step": 79197, "epoch": 889, "lr": 8.058920677102352e-05}
{"train_loss": 0.2059561014175415, "global_step": 79198, "epoch": 889, "lr": 8.058874818295956e-05}
{"train_loss": 0.28715792298316956, "global_step": 79199, "epoch": 889, "lr": 8.058828959078331e-05}
{"train_loss": 0.12072906643152237, "global_step": 79200, "epoch": 889, "lr": 8.058783099449485e-05}
{"train_loss": 0.20073957741260529, "global_step": 79201, "epoch": 889, "lr": 8.058737239409423e-05}
{"train_loss": 0.10259402543306351, "global_step": 79202, "epoch": 889, "lr": 8.05869137895815e-05}
{"train_loss": 0.24293242394924164, "global_step": 79203, "epoch": 889, "lr": 8.058645518095673e-05}
{"train_loss": 0.17595012485980988, "global_step": 79204, "epoch": 889, "lr": 8.058599656822e-05}
{"train_loss": 0.20357216894626617, "global_step": 79205, "epoch": 889, "lr": 8.058553795137134e-05}
{"train_loss": 0.18325886130332947, "global_step": 79206, "epoch": 889, "lr": 8.058507933041084e-05}
{"train_loss": 0.29382309317588806, "global_step": 79207, "epoch": 889, "lr": 8.058462070533854e-05}
{"train_loss": 0.26210352778434753, "global_step": 79208, "epoch": 889, "lr": 8.058416207615451e-05}
{"train_loss": 0.2072324101509673, "global_step": 79209, "epoch": 889, "lr": 8.058370344285881e-05, "val_loss": 3.7219293117523193}
{"train_loss": 0.19975261390209198, "global_step": 79210, "epoch": 890, "lr": 8.058324480545152e-05}
{"train_loss": 0.24093469977378845, "global_step": 79211, "epoch": 890, "lr": 8.058278616393267e-05}
{"train_loss": 0.1672433763742447, "global_step": 79212, "epoch": 890, "lr": 8.058232751830235e-05}
{"train_loss": 0.232547789812088, "global_step": 79213, "epoch": 890, "lr": 8.05818688685606e-05}
{"train_loss": 0.3045176863670349, "global_step": 79214, "epoch": 890, "lr": 8.058141021470749e-05}
{"train_loss": 0.182301327586174, "global_step": 79215, "epoch": 890, "lr": 8.058095155674307e-05}
{"train_loss": 0.15400665998458862, "global_step": 79216, "epoch": 890, "lr": 8.058049289466743e-05}
{"train_loss": 0.20798464119434357, "global_step": 79217, "epoch": 890, "lr": 8.058003422848062e-05}
{"train_loss": 0.30811816453933716, "global_step": 79218, "epoch": 890, "lr": 8.057957555818267e-05}
{"train_loss": 0.17054179310798645, "global_step": 79219, "epoch": 890, "lr": 8.057911688377371e-05}
{"train_loss": 0.23478244245052338, "global_step": 79220, "epoch": 890, "lr": 8.057865820525373e-05}
{"train_loss": 0.29041916131973267, "global_step": 79221, "epoch": 890, "lr": 8.057819952262283e-05}
{"train_loss": 0.24337589740753174, "global_step": 79222, "epoch": 890, "lr": 8.057774083588107e-05}
{"train_loss": 0.30158814787864685, "global_step": 79223, "epoch": 890, "lr": 8.057728214502849e-05}
{"train_loss": 0.17701952159404755, "global_step": 79224, "epoch": 890, "lr": 8.057682345006518e-05}
{"train_loss": 0.19042181968688965, "global_step": 79225, "epoch": 890, "lr": 8.057636475099119e-05}
{"train_loss": 0.1597362756729126, "global_step": 79226, "epoch": 890, "lr": 8.057590604780657e-05}
{"train_loss": 0.21025757491588593, "global_step": 79227, "epoch": 890, "lr": 8.05754473405114e-05}
{"train_loss": 0.21326419711112976, "global_step": 79228, "epoch": 890, "lr": 8.057498862910574e-05}
{"train_loss": 0.15388008952140808, "global_step": 79229, "epoch": 890, "lr": 8.057452991358963e-05}
{"train_loss": 0.20183414220809937, "global_step": 79230, "epoch": 890, "lr": 8.057407119396315e-05}
{"train_loss": 0.14451171457767487, "global_step": 79231, "epoch": 890, "lr": 8.057361247022637e-05}
{"train_loss": 0.17725397646427155, "global_step": 79232, "epoch": 890, "lr": 8.057315374237933e-05}
{"train_loss": 0.13476713001728058, "global_step": 79233, "epoch": 890, "lr": 8.057269501042211e-05}
{"train_loss": 0.1181870847940445, "global_step": 79234, "epoch": 890, "lr": 8.057223627435475e-05}
{"train_loss": 0.18337482213974, "global_step": 79235, "epoch": 890, "lr": 8.057177753417734e-05}
{"train_loss": 0.16642311215400696, "global_step": 79236, "epoch": 890, "lr": 8.057131878988992e-05}
{"train_loss": 0.20394864678382874, "global_step": 79237, "epoch": 890, "lr": 8.057086004149255e-05}
{"train_loss": 0.1691199541091919, "global_step": 79238, "epoch": 890, "lr": 8.057040128898531e-05}
{"train_loss": 0.16222433745861053, "global_step": 79239, "epoch": 890, "lr": 8.056994253236827e-05}
{"train_loss": 0.1685076355934143, "global_step": 79240, "epoch": 890, "lr": 8.056948377164146e-05}
{"train_loss": 0.17398282885551453, "global_step": 79241, "epoch": 890, "lr": 8.056902500680494e-05}
{"train_loss": 0.2412053346633911, "global_step": 79242, "epoch": 890, "lr": 8.056856623785881e-05}
{"train_loss": 0.1821291595697403, "global_step": 79243, "epoch": 890, "lr": 8.05681074648031e-05}
{"train_loss": 0.16291595995426178, "global_step": 79244, "epoch": 890, "lr": 8.056764868763787e-05}
{"train_loss": 0.17564240097999573, "global_step": 79245, "epoch": 890, "lr": 8.056718990636321e-05}
{"train_loss": 0.18013684451580048, "global_step": 79246, "epoch": 890, "lr": 8.056673112097916e-05}
{"train_loss": 0.20353694260120392, "global_step": 79247, "epoch": 890, "lr": 8.056627233148577e-05}
{"train_loss": 0.22548726201057434, "global_step": 79248, "epoch": 890, "lr": 8.056581353788312e-05}
{"train_loss": 0.14892248809337616, "global_step": 79249, "epoch": 890, "lr": 8.056535474017128e-05}
{"train_loss": 0.2014358639717102, "global_step": 79250, "epoch": 890, "lr": 8.056489593835031e-05}
{"train_loss": 0.19277553260326385, "global_step": 79251, "epoch": 890, "lr": 8.056443713242025e-05}
{"train_loss": 0.14641426503658295, "global_step": 79252, "epoch": 890, "lr": 8.056397832238116e-05}
{"train_loss": 0.1958482563495636, "global_step": 79253, "epoch": 890, "lr": 8.056351950823313e-05}
{"train_loss": 0.1429705023765564, "global_step": 79254, "epoch": 890, "lr": 8.05630606899762e-05}
{"train_loss": 0.10481381416320801, "global_step": 79255, "epoch": 890, "lr": 8.056260186761047e-05}
{"train_loss": 0.16139434278011322, "global_step": 79256, "epoch": 890, "lr": 8.056214304113593e-05}
{"train_loss": 0.2528351843357086, "global_step": 79257, "epoch": 890, "lr": 8.05616842105527e-05}
{"train_loss": 0.3671077489852905, "global_step": 79258, "epoch": 890, "lr": 8.056122537586084e-05}
{"train_loss": 0.20703186094760895, "global_step": 79259, "epoch": 890, "lr": 8.056076653706038e-05}
{"train_loss": 0.18136990070343018, "global_step": 79260, "epoch": 890, "lr": 8.05603076941514e-05}
{"train_loss": 0.15140599012374878, "global_step": 79261, "epoch": 890, "lr": 8.055984884713394e-05}
{"train_loss": 0.24215911328792572, "global_step": 79262, "epoch": 890, "lr": 8.055938999600809e-05}
{"train_loss": 0.28386643528938293, "global_step": 79263, "epoch": 890, "lr": 8.055893114077393e-05}
{"train_loss": 0.18293826282024384, "global_step": 79264, "epoch": 890, "lr": 8.055847228143146e-05}
{"train_loss": 0.17290246486663818, "global_step": 79265, "epoch": 890, "lr": 8.05580134179808e-05}
{"train_loss": 0.23616747558116913, "global_step": 79266, "epoch": 890, "lr": 8.055755455042196e-05}
{"train_loss": 0.2676999270915985, "global_step": 79267, "epoch": 890, "lr": 8.055709567875505e-05}
{"train_loss": 0.27208212018013, "global_step": 79268, "epoch": 890, "lr": 8.05566368029801e-05}
{"train_loss": 0.21135519444942474, "global_step": 79269, "epoch": 890, "lr": 8.05561779230972e-05}
{"train_loss": 0.26648688316345215, "global_step": 79270, "epoch": 890, "lr": 8.055571903910636e-05}
{"train_loss": 0.31697043776512146, "global_step": 79271, "epoch": 890, "lr": 8.055526015100771e-05}
{"train_loss": 0.18730908632278442, "global_step": 79272, "epoch": 890, "lr": 8.055480125880127e-05}
{"train_loss": 0.16676479578018188, "global_step": 79273, "epoch": 890, "lr": 8.055434236248712e-05}
{"train_loss": 0.21601562201976776, "global_step": 79274, "epoch": 890, "lr": 8.055388346206528e-05}
{"train_loss": 0.15136735141277313, "global_step": 79275, "epoch": 890, "lr": 8.055342455753587e-05}
{"train_loss": 0.2286684364080429, "global_step": 79276, "epoch": 890, "lr": 8.055296564889892e-05}
{"train_loss": 0.21483412384986877, "global_step": 79277, "epoch": 890, "lr": 8.055250673615449e-05}
{"train_loss": 0.16397206485271454, "global_step": 79278, "epoch": 890, "lr": 8.055204781930265e-05}
{"train_loss": 0.18432694673538208, "global_step": 79279, "epoch": 890, "lr": 8.055158889834344e-05}
{"train_loss": 0.229966402053833, "global_step": 79280, "epoch": 890, "lr": 8.055112997327697e-05}
{"train_loss": 0.12091364711523056, "global_step": 79281, "epoch": 890, "lr": 8.055067104410327e-05}
{"train_loss": 0.16006499528884888, "global_step": 79282, "epoch": 890, "lr": 8.055021211082239e-05}
{"train_loss": 0.16779248416423798, "global_step": 79283, "epoch": 890, "lr": 8.05497531734344e-05}
{"train_loss": 0.16390007734298706, "global_step": 79284, "epoch": 890, "lr": 8.05492942319394e-05}
{"train_loss": 0.18999038636684418, "global_step": 79285, "epoch": 890, "lr": 8.054883528633739e-05}
{"train_loss": 0.15140096843242645, "global_step": 79286, "epoch": 890, "lr": 8.054837633662847e-05}
{"train_loss": 0.17959876358509064, "global_step": 79287, "epoch": 890, "lr": 8.05479173828127e-05}
{"train_loss": 0.16030491888523102, "global_step": 79288, "epoch": 890, "lr": 8.054745842489013e-05}
{"train_loss": 0.20190441608428955, "global_step": 79289, "epoch": 890, "lr": 8.054699946286083e-05}
{"train_loss": 0.19428235292434692, "global_step": 79290, "epoch": 890, "lr": 8.054654049672485e-05}
{"train_loss": 0.12942221760749817, "global_step": 79291, "epoch": 890, "lr": 8.054608152648226e-05}
{"train_loss": 0.26250919699668884, "global_step": 79292, "epoch": 890, "lr": 8.054562255213313e-05}
{"train_loss": 0.20906628668308258, "global_step": 79293, "epoch": 890, "lr": 8.054516357367752e-05}
{"train_loss": 0.18008068203926086, "global_step": 79294, "epoch": 890, "lr": 8.054470459111547e-05}
{"train_loss": 0.22628945112228394, "global_step": 79295, "epoch": 890, "lr": 8.054424560444707e-05}
{"train_loss": 0.21455058455467224, "global_step": 79296, "epoch": 890, "lr": 8.054378661367236e-05}
{"train_loss": 0.1789821833372116, "global_step": 79297, "epoch": 890, "lr": 8.054332761879141e-05}
{"train_loss": 0.19803166146693604, "global_step": 79298, "epoch": 890, "lr": 8.054286861980427e-05, "val_loss": 3.991076946258545, "train_action_mse_error": 19.037857055664062}
{"train_loss": 0.2109704315662384, "global_step": 79299, "epoch": 891, "lr": 8.054240961671104e-05}
{"train_loss": 0.16007909178733826, "global_step": 79300, "epoch": 891, "lr": 8.054195060951173e-05}
{"train_loss": 0.21668030321598053, "global_step": 79301, "epoch": 891, "lr": 8.054149159820644e-05}
{"train_loss": 0.19918613135814667, "global_step": 79302, "epoch": 891, "lr": 8.054103258279521e-05}
{"train_loss": 0.12504842877388, "global_step": 79303, "epoch": 891, "lr": 8.054057356327813e-05}
{"train_loss": 0.22623199224472046, "global_step": 79304, "epoch": 891, "lr": 8.054011453965522e-05}
{"train_loss": 0.21472035348415375, "global_step": 79305, "epoch": 891, "lr": 8.053965551192657e-05}
{"train_loss": 0.26727530360221863, "global_step": 79306, "epoch": 891, "lr": 8.053919648009226e-05}
{"train_loss": 0.1732783168554306, "global_step": 79307, "epoch": 891, "lr": 8.05387374441523e-05}
{"train_loss": 0.22308702766895294, "global_step": 79308, "epoch": 891, "lr": 8.053827840410677e-05}
{"train_loss": 0.20250360667705536, "global_step": 79309, "epoch": 891, "lr": 8.053781935995577e-05}
{"train_loss": 0.19476915895938873, "global_step": 79310, "epoch": 891, "lr": 8.053736031169932e-05}
{"train_loss": 0.1519259363412857, "global_step": 79311, "epoch": 891, "lr": 8.053690125933748e-05}
{"train_loss": 0.16987550258636475, "global_step": 79312, "epoch": 891, "lr": 8.053644220287034e-05}
{"train_loss": 0.17924967408180237, "global_step": 79313, "epoch": 891, "lr": 8.053598314229796e-05}
{"train_loss": 0.20618152618408203, "global_step": 79314, "epoch": 891, "lr": 8.053552407762036e-05}
{"train_loss": 0.27625998854637146, "global_step": 79315, "epoch": 891, "lr": 8.053506500883766e-05}
{"train_loss": 0.19562184810638428, "global_step": 79316, "epoch": 891, "lr": 8.053460593594988e-05}
{"train_loss": 0.18019171059131622, "global_step": 79317, "epoch": 891, "lr": 8.053414685895708e-05}
{"train_loss": 0.13673487305641174, "global_step": 79318, "epoch": 891, "lr": 8.053368777785936e-05}
{"train_loss": 0.18471072614192963, "global_step": 79319, "epoch": 891, "lr": 8.053322869265674e-05}
{"train_loss": 0.17677925527095795, "global_step": 79320, "epoch": 891, "lr": 8.05327696033493e-05}
{"train_loss": 0.16757138073444366, "global_step": 79321, "epoch": 891, "lr": 8.053231050993712e-05}
{"train_loss": 0.19362933933734894, "global_step": 79322, "epoch": 891, "lr": 8.053185141242021e-05}
{"train_loss": 0.22110092639923096, "global_step": 79323, "epoch": 891, "lr": 8.05313923107987e-05}
{"train_loss": 0.27962857484817505, "global_step": 79324, "epoch": 891, "lr": 8.05309332050726e-05}
{"train_loss": 0.19921860098838806, "global_step": 79325, "epoch": 891, "lr": 8.053047409524199e-05}
{"train_loss": 0.1449236422777176, "global_step": 79326, "epoch": 891, "lr": 8.053001498130693e-05}
{"train_loss": 0.24941690266132355, "global_step": 79327, "epoch": 891, "lr": 8.052955586326748e-05}
{"train_loss": 0.19787989556789398, "global_step": 79328, "epoch": 891, "lr": 8.05290967411237e-05}
{"train_loss": 0.22069409489631653, "global_step": 79329, "epoch": 891, "lr": 8.052863761487565e-05}
{"train_loss": 0.2520143389701843, "global_step": 79330, "epoch": 891, "lr": 8.052817848452342e-05}
{"train_loss": 0.19720645248889923, "global_step": 79331, "epoch": 891, "lr": 8.052771935006705e-05}
{"train_loss": 0.09846076369285583, "global_step": 79332, "epoch": 891, "lr": 8.052726021150657e-05}
{"train_loss": 0.19141988456249237, "global_step": 79333, "epoch": 891, "lr": 8.05268010688421e-05}
{"train_loss": 0.18466219305992126, "global_step": 79334, "epoch": 891, "lr": 8.052634192207364e-05}
{"train_loss": 0.17840632796287537, "global_step": 79335, "epoch": 891, "lr": 8.052588277120132e-05}
{"train_loss": 0.14778293669223785, "global_step": 79336, "epoch": 891, "lr": 8.052542361622515e-05}
{"train_loss": 0.13478177785873413, "global_step": 79337, "epoch": 891, "lr": 8.052496445714521e-05}
{"train_loss": 0.21854513883590698, "global_step": 79338, "epoch": 891, "lr": 8.052450529396157e-05}
{"train_loss": 0.17673176527023315, "global_step": 79339, "epoch": 891, "lr": 8.052404612667427e-05}
{"train_loss": 0.3002063035964966, "global_step": 79340, "epoch": 891, "lr": 8.05235869552834e-05}
{"train_loss": 0.1231447383761406, "global_step": 79341, "epoch": 891, "lr": 8.052312777978899e-05}
{"train_loss": 0.25518175959587097, "global_step": 79342, "epoch": 891, "lr": 8.052266860019111e-05}
{"train_loss": 0.2765067517757416, "global_step": 79343, "epoch": 891, "lr": 8.052220941648984e-05}
{"train_loss": 0.1843530535697937, "global_step": 79344, "epoch": 891, "lr": 8.052175022868524e-05}
{"train_loss": 0.23437723517417908, "global_step": 79345, "epoch": 891, "lr": 8.052129103677736e-05}
{"train_loss": 0.20011012256145477, "global_step": 79346, "epoch": 891, "lr": 8.052083184076627e-05}
{"train_loss": 0.25214508175849915, "global_step": 79347, "epoch": 891, "lr": 8.0520372640652e-05}
{"train_loss": 0.18439675867557526, "global_step": 79348, "epoch": 891, "lr": 8.051991343643465e-05}
{"train_loss": 0.2313193380832672, "global_step": 79349, "epoch": 891, "lr": 8.051945422811427e-05}
{"train_loss": 0.19084954261779785, "global_step": 79350, "epoch": 891, "lr": 8.051899501569094e-05}
{"train_loss": 0.1971064954996109, "global_step": 79351, "epoch": 891, "lr": 8.051853579916468e-05}
{"train_loss": 0.22107744216918945, "global_step": 79352, "epoch": 891, "lr": 8.051807657853559e-05}
{"train_loss": 0.15879672765731812, "global_step": 79353, "epoch": 891, "lr": 8.05176173538037e-05}
{"train_loss": 0.2179042547941208, "global_step": 79354, "epoch": 891, "lr": 8.051715812496911e-05}
{"train_loss": 0.21472878754138947, "global_step": 79355, "epoch": 891, "lr": 8.051669889203184e-05}
{"train_loss": 0.25801876187324524, "global_step": 79356, "epoch": 891, "lr": 8.051623965499198e-05}
{"train_loss": 0.2629302740097046, "global_step": 79357, "epoch": 891, "lr": 8.051578041384959e-05}
{"train_loss": 0.29067638516426086, "global_step": 79358, "epoch": 891, "lr": 8.051532116860472e-05}
{"train_loss": 0.1408596932888031, "global_step": 79359, "epoch": 891, "lr": 8.051486191925742e-05}
{"train_loss": 0.2494950145483017, "global_step": 79360, "epoch": 891, "lr": 8.051440266580778e-05}
{"train_loss": 0.2207167148590088, "global_step": 79361, "epoch": 891, "lr": 8.051394340825586e-05}
{"train_loss": 0.19764643907546997, "global_step": 79362, "epoch": 891, "lr": 8.051348414660171e-05}
{"train_loss": 0.2113574892282486, "global_step": 79363, "epoch": 891, "lr": 8.051302488084538e-05}
{"train_loss": 0.18556785583496094, "global_step": 79364, "epoch": 891, "lr": 8.051256561098697e-05}
{"train_loss": 0.16776730120182037, "global_step": 79365, "epoch": 891, "lr": 8.051210633702651e-05}
{"train_loss": 0.272708922624588, "global_step": 79366, "epoch": 891, "lr": 8.051164705896406e-05}
{"train_loss": 0.3088512718677521, "global_step": 79367, "epoch": 891, "lr": 8.051118777679969e-05}
{"train_loss": 0.18754050135612488, "global_step": 79368, "epoch": 891, "lr": 8.051072849053347e-05}
{"train_loss": 0.21811339259147644, "global_step": 79369, "epoch": 891, "lr": 8.051026920016544e-05}
{"train_loss": 0.1782759130001068, "global_step": 79370, "epoch": 891, "lr": 8.05098099056957e-05}
{"train_loss": 0.1766270399093628, "global_step": 79371, "epoch": 891, "lr": 8.050935060712426e-05}
{"train_loss": 0.2550032436847687, "global_step": 79372, "epoch": 891, "lr": 8.050889130445123e-05}
{"train_loss": 0.15209455788135529, "global_step": 79373, "epoch": 891, "lr": 8.050843199767665e-05}
{"train_loss": 0.17169496417045593, "global_step": 79374, "epoch": 891, "lr": 8.050797268680058e-05}
{"train_loss": 0.24651403725147247, "global_step": 79375, "epoch": 891, "lr": 8.050751337182309e-05}
{"train_loss": 0.29210349917411804, "global_step": 79376, "epoch": 891, "lr": 8.050705405274423e-05}
{"train_loss": 0.20845626294612885, "global_step": 79377, "epoch": 891, "lr": 8.050659472956405e-05}
{"train_loss": 0.15459156036376953, "global_step": 79378, "epoch": 891, "lr": 8.050613540228265e-05}
{"train_loss": 0.1868910938501358, "global_step": 79379, "epoch": 891, "lr": 8.050567607090007e-05}
{"train_loss": 0.1936369687318802, "global_step": 79380, "epoch": 891, "lr": 8.050521673541637e-05}
{"train_loss": 0.22475795447826385, "global_step": 79381, "epoch": 891, "lr": 8.050475739583163e-05}
{"train_loss": 0.1788276582956314, "global_step": 79382, "epoch": 891, "lr": 8.05042980521459e-05}
{"train_loss": 0.2728702425956726, "global_step": 79383, "epoch": 891, "lr": 8.050383870435923e-05}
{"train_loss": 0.18456561863422394, "global_step": 79384, "epoch": 891, "lr": 8.050337935247167e-05}
{"train_loss": 0.2013409584760666, "global_step": 79385, "epoch": 891, "lr": 8.050291999648333e-05}
{"train_loss": 0.18647634983062744, "global_step": 79386, "epoch": 891, "lr": 8.050246063639425e-05}
{"train_loss": 0.20405411728647319, "global_step": 79387, "epoch": 891, "lr": 8.050200127220445e-05, "val_loss": 3.9759979248046875}
{"train_loss": 0.21630463004112244, "global_step": 79388, "epoch": 892, "lr": 8.050154190391406e-05}
{"train_loss": 0.2234589010477066, "global_step": 79389, "epoch": 892, "lr": 8.050108253152309e-05}
{"train_loss": 0.22304874658584595, "global_step": 79390, "epoch": 892, "lr": 8.050062315503162e-05}
{"train_loss": 0.17956916987895966, "global_step": 79391, "epoch": 892, "lr": 8.050016377443973e-05}
{"train_loss": 0.18106965720653534, "global_step": 79392, "epoch": 892, "lr": 8.049970438974746e-05}
{"train_loss": 0.2613098919391632, "global_step": 79393, "epoch": 892, "lr": 8.049924500095487e-05}
{"train_loss": 0.2145598977804184, "global_step": 79394, "epoch": 892, "lr": 8.049878560806203e-05}
{"train_loss": 0.2953118681907654, "global_step": 79395, "epoch": 892, "lr": 8.049832621106898e-05}
{"train_loss": 0.16312943398952484, "global_step": 79396, "epoch": 892, "lr": 8.049786680997585e-05}
{"train_loss": 0.23031257092952728, "global_step": 79397, "epoch": 892, "lr": 8.049740740478261e-05}
{"train_loss": 0.1412748098373413, "global_step": 79398, "epoch": 892, "lr": 8.049694799548938e-05}
{"train_loss": 0.18893899023532867, "global_step": 79399, "epoch": 892, "lr": 8.04964885820962e-05}
{"train_loss": 0.2044716775417328, "global_step": 79400, "epoch": 892, "lr": 8.049602916460316e-05}
{"train_loss": 0.24240469932556152, "global_step": 79401, "epoch": 892, "lr": 8.049556974301029e-05}
{"train_loss": 0.20038725435733795, "global_step": 79402, "epoch": 892, "lr": 8.049511031731765e-05}
{"train_loss": 0.2261488139629364, "global_step": 79403, "epoch": 892, "lr": 8.049465088752532e-05}
{"train_loss": 0.16723278164863586, "global_step": 79404, "epoch": 892, "lr": 8.049419145363338e-05}
{"train_loss": 0.21950048208236694, "global_step": 79405, "epoch": 892, "lr": 8.049373201564183e-05}
{"train_loss": 0.14853984117507935, "global_step": 79406, "epoch": 892, "lr": 8.04932725735508e-05}
{"train_loss": 0.1896480768918991, "global_step": 79407, "epoch": 892, "lr": 8.04928131273603e-05}
{"train_loss": 0.16900280117988586, "global_step": 79408, "epoch": 892, "lr": 8.049235367707042e-05}
{"train_loss": 0.2498544156551361, "global_step": 79409, "epoch": 892, "lr": 8.049189422268121e-05}
{"train_loss": 0.1616068035364151, "global_step": 79410, "epoch": 892, "lr": 8.049143476419276e-05}
{"train_loss": 0.18505005538463593, "global_step": 79411, "epoch": 892, "lr": 8.049097530160507e-05}
{"train_loss": 0.16509419679641724, "global_step": 79412, "epoch": 892, "lr": 8.049051583491827e-05}
{"train_loss": 0.19331207871437073, "global_step": 79413, "epoch": 892, "lr": 8.049005636413238e-05}
{"train_loss": 0.2040010392665863, "global_step": 79414, "epoch": 892, "lr": 8.048959688924748e-05}
{"train_loss": 0.19434130191802979, "global_step": 79415, "epoch": 892, "lr": 8.048913741026362e-05}
{"train_loss": 0.21065446734428406, "global_step": 79416, "epoch": 892, "lr": 8.048867792718087e-05}
{"train_loss": 0.17607182264328003, "global_step": 79417, "epoch": 892, "lr": 8.048821843999929e-05}
{"train_loss": 0.1354535073041916, "global_step": 79418, "epoch": 892, "lr": 8.048775894871894e-05}
{"train_loss": 0.2562824785709381, "global_step": 79419, "epoch": 892, "lr": 8.048729945333988e-05}
{"train_loss": 0.19502362608909607, "global_step": 79420, "epoch": 892, "lr": 8.048683995386218e-05}
{"train_loss": 0.25171875953674316, "global_step": 79421, "epoch": 892, "lr": 8.048638045028589e-05}
{"train_loss": 0.21873942017555237, "global_step": 79422, "epoch": 892, "lr": 8.048592094261107e-05}
{"train_loss": 0.21909339725971222, "global_step": 79423, "epoch": 892, "lr": 8.048546143083779e-05}
{"train_loss": 0.21411806344985962, "global_step": 79424, "epoch": 892, "lr": 8.048500191496612e-05}
{"train_loss": 0.23113474249839783, "global_step": 79425, "epoch": 892, "lr": 8.048454239499612e-05}
{"train_loss": 0.1980828493833542, "global_step": 79426, "epoch": 892, "lr": 8.048408287092784e-05}
{"train_loss": 0.14504283666610718, "global_step": 79427, "epoch": 892, "lr": 8.048362334276134e-05}
{"train_loss": 0.29502949118614197, "global_step": 79428, "epoch": 892, "lr": 8.048316381049668e-05}
{"train_loss": 0.2067783921957016, "global_step": 79429, "epoch": 892, "lr": 8.048270427413394e-05}
{"train_loss": 0.17652317881584167, "global_step": 79430, "epoch": 892, "lr": 8.048224473367318e-05}
{"train_loss": 0.18968665599822998, "global_step": 79431, "epoch": 892, "lr": 8.048178518911444e-05}
{"train_loss": 0.19176249206066132, "global_step": 79432, "epoch": 892, "lr": 8.04813256404578e-05}
{"train_loss": 0.23036527633666992, "global_step": 79433, "epoch": 892, "lr": 8.048086608770333e-05}
{"train_loss": 0.1699775606393814, "global_step": 79434, "epoch": 892, "lr": 8.048040653085107e-05}
{"train_loss": 0.29743850231170654, "global_step": 79435, "epoch": 892, "lr": 8.04799469699011e-05}
{"train_loss": 0.24038733541965485, "global_step": 79436, "epoch": 892, "lr": 8.047948740485344e-05}
{"train_loss": 0.1416119486093521, "global_step": 79437, "epoch": 892, "lr": 8.047902783570823e-05}
{"train_loss": 0.13515636324882507, "global_step": 79438, "epoch": 892, "lr": 8.047856826246544e-05}
{"train_loss": 0.2176583856344223, "global_step": 79439, "epoch": 892, "lr": 8.047810868512521e-05}
{"train_loss": 0.23166011273860931, "global_step": 79440, "epoch": 892, "lr": 8.047764910368757e-05}
{"train_loss": 0.21303220093250275, "global_step": 79441, "epoch": 892, "lr": 8.047718951815258e-05}
{"train_loss": 0.12873579561710358, "global_step": 79442, "epoch": 892, "lr": 8.047672992852029e-05}
{"train_loss": 0.23974382877349854, "global_step": 79443, "epoch": 892, "lr": 8.047627033479077e-05}
{"train_loss": 0.17187699675559998, "global_step": 79444, "epoch": 892, "lr": 8.047581073696411e-05}
{"train_loss": 0.21390500664710999, "global_step": 79445, "epoch": 892, "lr": 8.047535113504033e-05}
{"train_loss": 0.10702229291200638, "global_step": 79446, "epoch": 892, "lr": 8.047489152901953e-05}
{"train_loss": 0.16524341702461243, "global_step": 79447, "epoch": 892, "lr": 8.047443191890173e-05}
{"train_loss": 0.17651301622390747, "global_step": 79448, "epoch": 892, "lr": 8.047397230468703e-05}
{"train_loss": 0.20330891013145447, "global_step": 79449, "epoch": 892, "lr": 8.047351268637546e-05}
{"train_loss": 0.17053952813148499, "global_step": 79450, "epoch": 892, "lr": 8.04730530639671e-05}
{"train_loss": 0.17194753885269165, "global_step": 79451, "epoch": 892, "lr": 8.047259343746202e-05}
{"train_loss": 0.20079299807548523, "global_step": 79452, "epoch": 892, "lr": 8.047213380686027e-05}
{"train_loss": 0.18771614134311676, "global_step": 79453, "epoch": 892, "lr": 8.04716741721619e-05}
{"train_loss": 0.1566028743982315, "global_step": 79454, "epoch": 892, "lr": 8.0471214533367e-05}
{"train_loss": 0.1752406507730484, "global_step": 79455, "epoch": 892, "lr": 8.047075489047562e-05}
{"train_loss": 0.16996414959430695, "global_step": 79456, "epoch": 892, "lr": 8.04702952434878e-05}
{"train_loss": 0.14684364199638367, "global_step": 79457, "epoch": 892, "lr": 8.046983559240364e-05}
{"train_loss": 0.2061605453491211, "global_step": 79458, "epoch": 892, "lr": 8.046937593722317e-05}
{"train_loss": 0.18208174407482147, "global_step": 79459, "epoch": 892, "lr": 8.046891627794648e-05}
{"train_loss": 0.20561739802360535, "global_step": 79460, "epoch": 892, "lr": 8.046845661457359e-05}
{"train_loss": 0.15735292434692383, "global_step": 79461, "epoch": 892, "lr": 8.046799694710461e-05}
{"train_loss": 0.1175023764371872, "global_step": 79462, "epoch": 892, "lr": 8.046753727553956e-05}
{"train_loss": 0.1627817451953888, "global_step": 79463, "epoch": 892, "lr": 8.046707759987852e-05}
{"train_loss": 0.19699418544769287, "global_step": 79464, "epoch": 892, "lr": 8.046661792012156e-05}
{"train_loss": 0.11588796228170395, "global_step": 79465, "epoch": 892, "lr": 8.046615823626875e-05}
{"train_loss": 0.26345381140708923, "global_step": 79466, "epoch": 892, "lr": 8.046569854832011e-05}
{"train_loss": 0.34763103723526, "global_step": 79467, "epoch": 892, "lr": 8.046523885627575e-05}
{"train_loss": 0.24240761995315552, "global_step": 79468, "epoch": 892, "lr": 8.046477916013571e-05}
{"train_loss": 0.17827820777893066, "global_step": 79469, "epoch": 892, "lr": 8.046431945990003e-05}
{"train_loss": 0.10442076623439789, "global_step": 79470, "epoch": 892, "lr": 8.04638597555688e-05}
{"train_loss": 0.1778029501438141, "global_step": 79471, "epoch": 892, "lr": 8.046340004714209e-05}
{"train_loss": 0.1883794069290161, "global_step": 79472, "epoch": 892, "lr": 8.046294033461995e-05}
{"train_loss": 0.16264204680919647, "global_step": 79473, "epoch": 892, "lr": 8.046248061800242e-05}
{"train_loss": 0.22539901733398438, "global_step": 79474, "epoch": 892, "lr": 8.046202089728958e-05}
{"train_loss": 0.2258913666009903, "global_step": 79475, "epoch": 892, "lr": 8.046156117248152e-05}
{"train_loss": 0.19564972683954773, "global_step": 79476, "epoch": 892, "lr": 8.046110144357823e-05, "val_loss": 3.977632761001587}
{"train_loss": 0.14008569717407227, "global_step": 79477, "epoch": 893, "lr": 8.046064171057986e-05}
{"train_loss": 0.1936563402414322, "global_step": 79478, "epoch": 893, "lr": 8.046018197348641e-05}
{"train_loss": 0.19798970222473145, "global_step": 79479, "epoch": 893, "lr": 8.045972223229796e-05}
{"train_loss": 0.15824657678604126, "global_step": 79480, "epoch": 893, "lr": 8.045926248701457e-05}
{"train_loss": 0.20498111844062805, "global_step": 79481, "epoch": 893, "lr": 8.045880273763631e-05}
{"train_loss": 0.24836459755897522, "global_step": 79482, "epoch": 893, "lr": 8.045834298416321e-05}
{"train_loss": 0.19358383119106293, "global_step": 79483, "epoch": 893, "lr": 8.045788322659538e-05}
{"train_loss": 0.1546621322631836, "global_step": 79484, "epoch": 893, "lr": 8.045742346493287e-05}
{"train_loss": 0.22338499128818512, "global_step": 79485, "epoch": 893, "lr": 8.045696369917571e-05}
{"train_loss": 0.24884893000125885, "global_step": 79486, "epoch": 893, "lr": 8.045650392932397e-05}
{"train_loss": 0.22029997408390045, "global_step": 79487, "epoch": 893, "lr": 8.045604415537775e-05}
{"train_loss": 0.27130380272865295, "global_step": 79488, "epoch": 893, "lr": 8.045558437733709e-05}
{"train_loss": 0.1738818883895874, "global_step": 79489, "epoch": 893, "lr": 8.045512459520204e-05}
{"train_loss": 0.2606525421142578, "global_step": 79490, "epoch": 893, "lr": 8.045466480897266e-05}
{"train_loss": 0.21988044679164886, "global_step": 79491, "epoch": 893, "lr": 8.045420501864903e-05}
{"train_loss": 0.32725831866264343, "global_step": 79492, "epoch": 893, "lr": 8.04537452242312e-05}
{"train_loss": 0.2351829707622528, "global_step": 79493, "epoch": 893, "lr": 8.045328542571924e-05}
{"train_loss": 0.1309276521205902, "global_step": 79494, "epoch": 893, "lr": 8.04528256231132e-05}
{"train_loss": 0.16819056868553162, "global_step": 79495, "epoch": 893, "lr": 8.045236581641315e-05}
{"train_loss": 0.2913967967033386, "global_step": 79496, "epoch": 893, "lr": 8.045190600561915e-05}
{"train_loss": 0.22066494822502136, "global_step": 79497, "epoch": 893, "lr": 8.045144619073129e-05}
{"train_loss": 0.12788015604019165, "global_step": 79498, "epoch": 893, "lr": 8.045098637174957e-05}
{"train_loss": 0.2760608196258545, "global_step": 79499, "epoch": 893, "lr": 8.045052654867409e-05}
{"train_loss": 0.1774212270975113, "global_step": 79500, "epoch": 893, "lr": 8.045006672150493e-05}
{"train_loss": 0.21754923462867737, "global_step": 79501, "epoch": 893, "lr": 8.04496068902421e-05}
{"train_loss": 0.1996498852968216, "global_step": 79502, "epoch": 893, "lr": 8.044914705488571e-05}
{"train_loss": 0.20505239069461823, "global_step": 79503, "epoch": 893, "lr": 8.044868721543579e-05}
{"train_loss": 0.22388547658920288, "global_step": 79504, "epoch": 893, "lr": 8.044822737189242e-05}
{"train_loss": 0.18796709179878235, "global_step": 79505, "epoch": 893, "lr": 8.044776752425566e-05}
{"train_loss": 0.23131516575813293, "global_step": 79506, "epoch": 893, "lr": 8.044730767252557e-05}
{"train_loss": 0.191266730427742, "global_step": 79507, "epoch": 893, "lr": 8.044684781670221e-05}
{"train_loss": 0.22296763956546783, "global_step": 79508, "epoch": 893, "lr": 8.044638795678564e-05}
{"train_loss": 0.3236960172653198, "global_step": 79509, "epoch": 893, "lr": 8.044592809277592e-05}
{"train_loss": 0.21180446445941925, "global_step": 79510, "epoch": 893, "lr": 8.044546822467312e-05}
{"train_loss": 0.16321729123592377, "global_step": 79511, "epoch": 893, "lr": 8.04450083524773e-05}
{"train_loss": 0.25552576780319214, "global_step": 79512, "epoch": 893, "lr": 8.044454847618852e-05}
{"train_loss": 0.2611925005912781, "global_step": 79513, "epoch": 893, "lr": 8.044408859580686e-05}
{"train_loss": 0.20401671528816223, "global_step": 79514, "epoch": 893, "lr": 8.044362871133233e-05}
{"train_loss": 0.29202383756637573, "global_step": 79515, "epoch": 893, "lr": 8.044316882276504e-05}
{"train_loss": 0.17121128737926483, "global_step": 79516, "epoch": 893, "lr": 8.044270893010504e-05}
{"train_loss": 0.18618987500667572, "global_step": 79517, "epoch": 893, "lr": 8.044224903335238e-05}
{"train_loss": 0.2090766876935959, "global_step": 79518, "epoch": 893, "lr": 8.044178913250715e-05}
{"train_loss": 0.24749495089054108, "global_step": 79519, "epoch": 893, "lr": 8.044132922756937e-05}
{"train_loss": 0.20337195694446564, "global_step": 79520, "epoch": 893, "lr": 8.044086931853913e-05}
{"train_loss": 0.145460307598114, "global_step": 79521, "epoch": 893, "lr": 8.044040940541649e-05}
{"train_loss": 0.3185133635997772, "global_step": 79522, "epoch": 893, "lr": 8.04399494882015e-05}
{"train_loss": 0.22656813263893127, "global_step": 79523, "epoch": 893, "lr": 8.043948956689425e-05}
{"train_loss": 0.2214375138282776, "global_step": 79524, "epoch": 893, "lr": 8.043902964149476e-05}
{"train_loss": 0.14154016971588135, "global_step": 79525, "epoch": 893, "lr": 8.043856971200312e-05}
{"train_loss": 0.17997027933597565, "global_step": 79526, "epoch": 893, "lr": 8.043810977841939e-05}
{"train_loss": 0.2054017186164856, "global_step": 79527, "epoch": 893, "lr": 8.043764984074362e-05}
{"train_loss": 0.2442256510257721, "global_step": 79528, "epoch": 893, "lr": 8.043718989897589e-05}
{"train_loss": 0.3027906119823456, "global_step": 79529, "epoch": 893, "lr": 8.043672995311624e-05}
{"train_loss": 0.2731230556964874, "global_step": 79530, "epoch": 893, "lr": 8.043627000316475e-05}
{"train_loss": 0.2764216661453247, "global_step": 79531, "epoch": 893, "lr": 8.043581004912146e-05}
{"train_loss": 0.1816420555114746, "global_step": 79532, "epoch": 893, "lr": 8.043535009098646e-05}
{"train_loss": 0.25389280915260315, "global_step": 79533, "epoch": 893, "lr": 8.04348901287598e-05}
{"train_loss": 0.2086367905139923, "global_step": 79534, "epoch": 893, "lr": 8.043443016244154e-05}
{"train_loss": 0.2119588851928711, "global_step": 79535, "epoch": 893, "lr": 8.043397019203174e-05}
{"train_loss": 0.21047291159629822, "global_step": 79536, "epoch": 893, "lr": 8.043351021753046e-05}
{"train_loss": 0.20485685765743256, "global_step": 79537, "epoch": 893, "lr": 8.043305023893777e-05}
{"train_loss": 0.23031240701675415, "global_step": 79538, "epoch": 893, "lr": 8.043259025625371e-05}
{"train_loss": 0.15836918354034424, "global_step": 79539, "epoch": 893, "lr": 8.043213026947838e-05}
{"train_loss": 0.2525685131549835, "global_step": 79540, "epoch": 893, "lr": 8.04316702786118e-05}
{"train_loss": 0.19367615878582, "global_step": 79541, "epoch": 893, "lr": 8.043121028365409e-05}
{"train_loss": 0.17454399168491364, "global_step": 79542, "epoch": 893, "lr": 8.043075028460524e-05}
{"train_loss": 0.20383429527282715, "global_step": 79543, "epoch": 893, "lr": 8.043029028146536e-05}
{"train_loss": 0.350109338760376, "global_step": 79544, "epoch": 893, "lr": 8.04298302742345e-05}
{"train_loss": 0.18768763542175293, "global_step": 79545, "epoch": 893, "lr": 8.042937026291271e-05}
{"train_loss": 0.193099245429039, "global_step": 79546, "epoch": 893, "lr": 8.042891024750008e-05}
{"train_loss": 0.1816747784614563, "global_step": 79547, "epoch": 893, "lr": 8.042845022799663e-05}
{"train_loss": 0.22494998574256897, "global_step": 79548, "epoch": 893, "lr": 8.042799020440246e-05}
{"train_loss": 0.15828606486320496, "global_step": 79549, "epoch": 893, "lr": 8.042753017671763e-05}
{"train_loss": 0.22209739685058594, "global_step": 79550, "epoch": 893, "lr": 8.042707014494218e-05}
{"train_loss": 0.19280950725078583, "global_step": 79551, "epoch": 893, "lr": 8.042661010907617e-05}
{"train_loss": 0.1926339566707611, "global_step": 79552, "epoch": 893, "lr": 8.042615006911967e-05}
{"train_loss": 0.22269603610038757, "global_step": 79553, "epoch": 893, "lr": 8.042569002507277e-05}
{"train_loss": 0.23974111676216125, "global_step": 79554, "epoch": 893, "lr": 8.042522997693547e-05}
{"train_loss": 0.14916077256202698, "global_step": 79555, "epoch": 893, "lr": 8.04247699247079e-05}
{"train_loss": 0.1995856761932373, "global_step": 79556, "epoch": 893, "lr": 8.042430986839009e-05}
{"train_loss": 0.20077577233314514, "global_step": 79557, "epoch": 893, "lr": 8.04238498079821e-05}
{"train_loss": 0.3092668652534485, "global_step": 79558, "epoch": 893, "lr": 8.042338974348396e-05}
{"train_loss": 0.18830421566963196, "global_step": 79559, "epoch": 893, "lr": 8.042292967489579e-05}
{"train_loss": 0.272760272026062, "global_step": 79560, "epoch": 893, "lr": 8.042246960221764e-05}
{"train_loss": 0.24969647824764252, "global_step": 79561, "epoch": 893, "lr": 8.042200952544954e-05}
{"train_loss": 0.190831258893013, "global_step": 79562, "epoch": 893, "lr": 8.04215494445916e-05}
{"train_loss": 0.2191677689552307, "global_step": 79563, "epoch": 893, "lr": 8.042108935964382e-05}
{"train_loss": 0.17401280999183655, "global_step": 79564, "epoch": 893, "lr": 8.042062927060632e-05}
{"train_loss": 0.21607827771915478, "global_step": 79565, "epoch": 893, "lr": 8.042016917747913e-05, "val_loss": 3.738956928253174}
{"train_loss": 0.21363766491413116, "global_step": 79566, "epoch": 894, "lr": 8.041970908026231e-05}
{"train_loss": 0.21662189066410065, "global_step": 79567, "epoch": 894, "lr": 8.041924897895595e-05}
{"train_loss": 0.19177378714084625, "global_step": 79568, "epoch": 894, "lr": 8.041878887356008e-05}
{"train_loss": 0.2312055230140686, "global_step": 79569, "epoch": 894, "lr": 8.041832876407478e-05}
{"train_loss": 0.2258826494216919, "global_step": 79570, "epoch": 894, "lr": 8.04178686505001e-05}
{"train_loss": 0.24749748408794403, "global_step": 79571, "epoch": 894, "lr": 8.04174085328361e-05}
{"train_loss": 0.1932954341173172, "global_step": 79572, "epoch": 894, "lr": 8.041694841108287e-05}
{"train_loss": 0.17611530423164368, "global_step": 79573, "epoch": 894, "lr": 8.041648828524044e-05}
{"train_loss": 0.1872996836900711, "global_step": 79574, "epoch": 894, "lr": 8.041602815530889e-05}
{"train_loss": 0.19163362681865692, "global_step": 79575, "epoch": 894, "lr": 8.041556802128827e-05}
{"train_loss": 0.24180705845355988, "global_step": 79576, "epoch": 894, "lr": 8.041510788317865e-05}
{"train_loss": 0.25885307788848877, "global_step": 79577, "epoch": 894, "lr": 8.041464774098008e-05}
{"train_loss": 0.1926511824131012, "global_step": 79578, "epoch": 894, "lr": 8.041418759469265e-05}
{"train_loss": 0.16333238780498505, "global_step": 79579, "epoch": 894, "lr": 8.041372744431639e-05}
{"train_loss": 0.22965681552886963, "global_step": 79580, "epoch": 894, "lr": 8.041326728985138e-05}
{"train_loss": 0.1515107899904251, "global_step": 79581, "epoch": 894, "lr": 8.041280713129769e-05}
{"train_loss": 0.15919914841651917, "global_step": 79582, "epoch": 894, "lr": 8.041234696865533e-05}
{"train_loss": 0.20632782578468323, "global_step": 79583, "epoch": 894, "lr": 8.041188680192444e-05}
{"train_loss": 0.1819106489419937, "global_step": 79584, "epoch": 894, "lr": 8.041142663110502e-05}
{"train_loss": 0.16639098525047302, "global_step": 79585, "epoch": 894, "lr": 8.041096645619717e-05}
{"train_loss": 0.19933068752288818, "global_step": 79586, "epoch": 894, "lr": 8.041050627720092e-05}
{"train_loss": 0.14012670516967773, "global_step": 79587, "epoch": 894, "lr": 8.041004609411636e-05}
{"train_loss": 0.16665621101856232, "global_step": 79588, "epoch": 894, "lr": 8.040958590694354e-05}
{"train_loss": 0.17120392620563507, "global_step": 79589, "epoch": 894, "lr": 8.040912571568252e-05}
{"train_loss": 0.2363310009241104, "global_step": 79590, "epoch": 894, "lr": 8.040866552033336e-05}
{"train_loss": 0.19532336294651031, "global_step": 79591, "epoch": 894, "lr": 8.040820532089613e-05}
{"train_loss": 0.32218343019485474, "global_step": 79592, "epoch": 894, "lr": 8.040774511737088e-05}
{"train_loss": 0.2212618887424469, "global_step": 79593, "epoch": 894, "lr": 8.040728490975768e-05}
{"train_loss": 0.20355640351772308, "global_step": 79594, "epoch": 894, "lr": 8.04068246980566e-05}
{"train_loss": 0.17461040616035461, "global_step": 79595, "epoch": 894, "lr": 8.040636448226768e-05}
{"train_loss": 0.2370634824037552, "global_step": 79596, "epoch": 894, "lr": 8.040590426239101e-05}
{"train_loss": 0.24300611019134521, "global_step": 79597, "epoch": 894, "lr": 8.040544403842662e-05}
{"train_loss": 0.15245094895362854, "global_step": 79598, "epoch": 894, "lr": 8.04049838103746e-05}
{"train_loss": 0.1412428915500641, "global_step": 79599, "epoch": 894, "lr": 8.0404523578235e-05}
{"train_loss": 0.24783457815647125, "global_step": 79600, "epoch": 894, "lr": 8.040406334200788e-05}
{"train_loss": 0.16408765316009521, "global_step": 79601, "epoch": 894, "lr": 8.040360310169331e-05}
{"train_loss": 0.16487380862236023, "global_step": 79602, "epoch": 894, "lr": 8.040314285729134e-05}
{"train_loss": 0.23703640699386597, "global_step": 79603, "epoch": 894, "lr": 8.040268260880204e-05}
{"train_loss": 0.3131810426712036, "global_step": 79604, "epoch": 894, "lr": 8.040222235622548e-05}
{"train_loss": 0.19032880663871765, "global_step": 79605, "epoch": 894, "lr": 8.04017620995617e-05}
{"train_loss": 0.3704582750797272, "global_step": 79606, "epoch": 894, "lr": 8.040130183881077e-05}
{"train_loss": 0.34242773056030273, "global_step": 79607, "epoch": 894, "lr": 8.040084157397276e-05}
{"train_loss": 0.22359342873096466, "global_step": 79608, "epoch": 894, "lr": 8.040038130504772e-05}
{"train_loss": 0.17594636976718903, "global_step": 79609, "epoch": 894, "lr": 8.039992103203573e-05}
{"train_loss": 0.28597551584243774, "global_step": 79610, "epoch": 894, "lr": 8.039946075493685e-05}
{"train_loss": 0.10375799238681793, "global_step": 79611, "epoch": 894, "lr": 8.039900047375111e-05}
{"train_loss": 0.24756716191768646, "global_step": 79612, "epoch": 894, "lr": 8.039854018847861e-05}
{"train_loss": 0.13988113403320312, "global_step": 79613, "epoch": 894, "lr": 8.03980798991194e-05}
{"train_loss": 0.16552948951721191, "global_step": 79614, "epoch": 894, "lr": 8.039761960567353e-05}
{"train_loss": 0.20310541987419128, "global_step": 79615, "epoch": 894, "lr": 8.039715930814107e-05}
{"train_loss": 0.21271999180316925, "global_step": 79616, "epoch": 894, "lr": 8.039669900652208e-05}
{"train_loss": 0.16141976416110992, "global_step": 79617, "epoch": 894, "lr": 8.039623870081663e-05}
{"train_loss": 0.2706576883792877, "global_step": 79618, "epoch": 894, "lr": 8.039577839102476e-05}
{"train_loss": 0.23298832774162292, "global_step": 79619, "epoch": 894, "lr": 8.039531807714658e-05}
{"train_loss": 0.18912549316883087, "global_step": 79620, "epoch": 894, "lr": 8.039485775918211e-05}
{"train_loss": 0.18883709609508514, "global_step": 79621, "epoch": 894, "lr": 8.039439743713141e-05}
{"train_loss": 0.15196852385997772, "global_step": 79622, "epoch": 894, "lr": 8.039393711099457e-05}
{"train_loss": 0.1919795572757721, "global_step": 79623, "epoch": 894, "lr": 8.039347678077162e-05}
{"train_loss": 0.22599606215953827, "global_step": 79624, "epoch": 894, "lr": 8.039301644646263e-05}
{"train_loss": 0.18598505854606628, "global_step": 79625, "epoch": 894, "lr": 8.039255610806769e-05}
{"train_loss": 0.21350915729999542, "global_step": 79626, "epoch": 894, "lr": 8.039209576558682e-05}
{"train_loss": 0.22184385359287262, "global_step": 79627, "epoch": 894, "lr": 8.039163541902013e-05}
{"train_loss": 0.31413060426712036, "global_step": 79628, "epoch": 894, "lr": 8.039117506836764e-05}
{"train_loss": 0.23184967041015625, "global_step": 79629, "epoch": 894, "lr": 8.039071471362942e-05}
{"train_loss": 0.18100661039352417, "global_step": 79630, "epoch": 894, "lr": 8.039025435480555e-05}
{"train_loss": 0.25174465775489807, "global_step": 79631, "epoch": 894, "lr": 8.038979399189608e-05}
{"train_loss": 0.1849367618560791, "global_step": 79632, "epoch": 894, "lr": 8.038933362490107e-05}
{"train_loss": 0.23752759397029877, "global_step": 79633, "epoch": 894, "lr": 8.038887325382058e-05}
{"train_loss": 0.2272750586271286, "global_step": 79634, "epoch": 894, "lr": 8.038841287865468e-05}
{"train_loss": 0.2061639428138733, "global_step": 79635, "epoch": 894, "lr": 8.038795249940343e-05}
{"train_loss": 0.2779165506362915, "global_step": 79636, "epoch": 894, "lr": 8.038749211606689e-05}
{"train_loss": 0.2094484567642212, "global_step": 79637, "epoch": 894, "lr": 8.038703172864512e-05}
{"train_loss": 0.17869865894317627, "global_step": 79638, "epoch": 894, "lr": 8.038657133713819e-05}
{"train_loss": 0.23251605033874512, "global_step": 79639, "epoch": 894, "lr": 8.038611094154615e-05}
{"train_loss": 0.28412505984306335, "global_step": 79640, "epoch": 894, "lr": 8.038565054186907e-05}
{"train_loss": 0.20796242356300354, "global_step": 79641, "epoch": 894, "lr": 8.0385190138107e-05}
{"train_loss": 0.18009081482887268, "global_step": 79642, "epoch": 894, "lr": 8.038472973026002e-05}
{"train_loss": 0.3390404284000397, "global_step": 79643, "epoch": 894, "lr": 8.038426931832819e-05}
{"train_loss": 0.20844818651676178, "global_step": 79644, "epoch": 894, "lr": 8.038380890231155e-05}
{"train_loss": 0.17989887297153473, "global_step": 79645, "epoch": 894, "lr": 8.038334848221019e-05}
{"train_loss": 0.2630636990070343, "global_step": 79646, "epoch": 894, "lr": 8.038288805802415e-05}
{"train_loss": 0.19942834973335266, "global_step": 79647, "epoch": 894, "lr": 8.038242762975351e-05}
{"train_loss": 0.2564697861671448, "global_step": 79648, "epoch": 894, "lr": 8.038196719739832e-05}
{"train_loss": 0.15614186227321625, "global_step": 79649, "epoch": 894, "lr": 8.038150676095864e-05}
{"train_loss": 0.2546730041503906, "global_step": 79650, "epoch": 894, "lr": 8.038104632043454e-05}
{"train_loss": 0.1294860690832138, "global_step": 79651, "epoch": 894, "lr": 8.038058587582607e-05}
{"train_loss": 0.19945143163204193, "global_step": 79652, "epoch": 894, "lr": 8.038012542713332e-05}
{"train_loss": 0.204610675573349, "global_step": 79653, "epoch": 894, "lr": 8.037966497435631e-05}
{"train_loss": 0.21112629871689872, "global_step": 79654, "epoch": 894, "lr": 8.037920451749514e-05, "val_loss": 3.988929271697998}
{"train_loss": 0.2575867772102356, "global_step": 79655, "epoch": 895, "lr": 8.037874405654986e-05}
{"train_loss": 0.23408854007720947, "global_step": 79656, "epoch": 895, "lr": 8.037828359152051e-05}
{"train_loss": 0.2633349895477295, "global_step": 79657, "epoch": 895, "lr": 8.037782312240718e-05}
{"train_loss": 0.2629786431789398, "global_step": 79658, "epoch": 895, "lr": 8.037736264920993e-05}
{"train_loss": 0.21197140216827393, "global_step": 79659, "epoch": 895, "lr": 8.037690217192881e-05}
{"train_loss": 0.27117952704429626, "global_step": 79660, "epoch": 895, "lr": 8.037644169056389e-05}
{"train_loss": 0.18758809566497803, "global_step": 79661, "epoch": 895, "lr": 8.037598120511521e-05}
{"train_loss": 0.29070574045181274, "global_step": 79662, "epoch": 895, "lr": 8.037552071558285e-05}
{"train_loss": 0.15038159489631653, "global_step": 79663, "epoch": 895, "lr": 8.037506022196689e-05}
{"train_loss": 0.18481406569480896, "global_step": 79664, "epoch": 895, "lr": 8.037459972426736e-05}
{"train_loss": 0.13929876685142517, "global_step": 79665, "epoch": 895, "lr": 8.037413922248434e-05}
{"train_loss": 0.1709579974412918, "global_step": 79666, "epoch": 895, "lr": 8.03736787166179e-05}
{"train_loss": 0.30301764607429504, "global_step": 79667, "epoch": 895, "lr": 8.037321820666805e-05}
{"train_loss": 0.23567186295986176, "global_step": 79668, "epoch": 895, "lr": 8.037275769263492e-05}
{"train_loss": 0.2958240509033203, "global_step": 79669, "epoch": 895, "lr": 8.037229717451855e-05}
{"train_loss": 0.2232174426317215, "global_step": 79670, "epoch": 895, "lr": 8.037183665231898e-05}
{"train_loss": 0.2217702567577362, "global_step": 79671, "epoch": 895, "lr": 8.037137612603629e-05}
{"train_loss": 0.1712655872106552, "global_step": 79672, "epoch": 895, "lr": 8.037091559567053e-05}
{"train_loss": 0.16737662255764008, "global_step": 79673, "epoch": 895, "lr": 8.03704550612218e-05}
{"train_loss": 0.21185514330863953, "global_step": 79674, "epoch": 895, "lr": 8.03699945226901e-05}
{"train_loss": 0.19953404366970062, "global_step": 79675, "epoch": 895, "lr": 8.036953398007555e-05}
{"train_loss": 0.18229548633098602, "global_step": 79676, "epoch": 895, "lr": 8.036907343337816e-05}
{"train_loss": 0.12016131728887558, "global_step": 79677, "epoch": 895, "lr": 8.036861288259804e-05}
{"train_loss": 0.17732617259025574, "global_step": 79678, "epoch": 895, "lr": 8.036815232773522e-05}
{"train_loss": 0.27965328097343445, "global_step": 79679, "epoch": 895, "lr": 8.036769176878978e-05}
{"train_loss": 0.10380421578884125, "global_step": 79680, "epoch": 895, "lr": 8.036723120576177e-05}
{"train_loss": 0.21390986442565918, "global_step": 79681, "epoch": 895, "lr": 8.036677063865125e-05}
{"train_loss": 0.1956031173467636, "global_step": 79682, "epoch": 895, "lr": 8.036631006745829e-05}
{"train_loss": 0.16702182590961456, "global_step": 79683, "epoch": 895, "lr": 8.036584949218296e-05}
{"train_loss": 0.277221143245697, "global_step": 79684, "epoch": 895, "lr": 8.03653889128253e-05}
{"train_loss": 0.20070059597492218, "global_step": 79685, "epoch": 895, "lr": 8.036492832938537e-05}
{"train_loss": 0.1946457028388977, "global_step": 79686, "epoch": 895, "lr": 8.036446774186327e-05}
{"train_loss": 0.2575528919696808, "global_step": 79687, "epoch": 895, "lr": 8.036400715025903e-05}
{"train_loss": 0.19450071454048157, "global_step": 79688, "epoch": 895, "lr": 8.036354655457272e-05}
{"train_loss": 0.19625155627727509, "global_step": 79689, "epoch": 895, "lr": 8.03630859548044e-05}
{"train_loss": 0.16545535624027252, "global_step": 79690, "epoch": 895, "lr": 8.036262535095414e-05}
{"train_loss": 0.1581200659275055, "global_step": 79691, "epoch": 895, "lr": 8.036216474302199e-05}
{"train_loss": 0.12729701399803162, "global_step": 79692, "epoch": 895, "lr": 8.036170413100802e-05}
{"train_loss": 0.18421313166618347, "global_step": 79693, "epoch": 895, "lr": 8.036124351491228e-05}
{"train_loss": 0.25082626938819885, "global_step": 79694, "epoch": 895, "lr": 8.036078289473484e-05}
{"train_loss": 0.20155051350593567, "global_step": 79695, "epoch": 895, "lr": 8.036032227047576e-05}
{"train_loss": 0.2950323224067688, "global_step": 79696, "epoch": 895, "lr": 8.035986164213512e-05}
{"train_loss": 0.16073143482208252, "global_step": 79697, "epoch": 895, "lr": 8.035940100971295e-05}
{"train_loss": 0.15062463283538818, "global_step": 79698, "epoch": 895, "lr": 8.035894037320935e-05}
{"train_loss": 0.2126161903142929, "global_step": 79699, "epoch": 895, "lr": 8.035847973262434e-05}
{"train_loss": 0.24902015924453735, "global_step": 79700, "epoch": 895, "lr": 8.035801908795801e-05}
{"train_loss": 0.1787036955356598, "global_step": 79701, "epoch": 895, "lr": 8.035755843921042e-05}
{"train_loss": 0.12997478246688843, "global_step": 79702, "epoch": 895, "lr": 8.035709778638163e-05}
{"train_loss": 0.19570253789424896, "global_step": 79703, "epoch": 895, "lr": 8.035663712947167e-05}
{"train_loss": 0.18531040847301483, "global_step": 79704, "epoch": 895, "lr": 8.035617646848066e-05}
{"train_loss": 0.20177747309207916, "global_step": 79705, "epoch": 895, "lr": 8.035571580340861e-05}
{"train_loss": 0.239969864487648, "global_step": 79706, "epoch": 895, "lr": 8.035525513425563e-05}
{"train_loss": 0.285896360874176, "global_step": 79707, "epoch": 895, "lr": 8.035479446102175e-05}
{"train_loss": 0.24040062725543976, "global_step": 79708, "epoch": 895, "lr": 8.035433378370701e-05}
{"train_loss": 0.16018152236938477, "global_step": 79709, "epoch": 895, "lr": 8.035387310231153e-05}
{"train_loss": 0.23391534388065338, "global_step": 79710, "epoch": 895, "lr": 8.035341241683532e-05}
{"train_loss": 0.16475315392017365, "global_step": 79711, "epoch": 895, "lr": 8.035295172727848e-05}
{"train_loss": 0.16428588330745697, "global_step": 79712, "epoch": 895, "lr": 8.035249103364105e-05}
{"train_loss": 0.12558040022850037, "global_step": 79713, "epoch": 895, "lr": 8.03520303359231e-05}
{"train_loss": 0.20631930232048035, "global_step": 79714, "epoch": 895, "lr": 8.035156963412469e-05}
{"train_loss": 0.15381403267383575, "global_step": 79715, "epoch": 895, "lr": 8.035110892824588e-05}
{"train_loss": 0.14818152785301208, "global_step": 79716, "epoch": 895, "lr": 8.035064821828673e-05}
{"train_loss": 0.28610944747924805, "global_step": 79717, "epoch": 895, "lr": 8.03501875042473e-05}
{"train_loss": 0.24768976867198944, "global_step": 79718, "epoch": 895, "lr": 8.034972678612767e-05}
{"train_loss": 0.16353905200958252, "global_step": 79719, "epoch": 895, "lr": 8.034926606392789e-05}
{"train_loss": 0.22965072095394135, "global_step": 79720, "epoch": 895, "lr": 8.034880533764802e-05}
{"train_loss": 0.2602449655532837, "global_step": 79721, "epoch": 895, "lr": 8.034834460728811e-05}
{"train_loss": 0.1346384584903717, "global_step": 79722, "epoch": 895, "lr": 8.034788387284824e-05}
{"train_loss": 0.1899067759513855, "global_step": 79723, "epoch": 895, "lr": 8.034742313432846e-05}
{"train_loss": 0.24813427031040192, "global_step": 79724, "epoch": 895, "lr": 8.034696239172886e-05}
{"train_loss": 0.20335371792316437, "global_step": 79725, "epoch": 895, "lr": 8.034650164504947e-05}
{"train_loss": 0.22422724962234497, "global_step": 79726, "epoch": 895, "lr": 8.034604089429036e-05}
{"train_loss": 0.23952442407608032, "global_step": 79727, "epoch": 895, "lr": 8.034558013945159e-05}
{"train_loss": 0.16981928050518036, "global_step": 79728, "epoch": 895, "lr": 8.034511938053324e-05}
{"train_loss": 0.30432555079460144, "global_step": 79729, "epoch": 895, "lr": 8.034465861753535e-05}
{"train_loss": 0.2558993697166443, "global_step": 79730, "epoch": 895, "lr": 8.0344197850458e-05}
{"train_loss": 0.257080614566803, "global_step": 79731, "epoch": 895, "lr": 8.034373707930123e-05}
{"train_loss": 0.16945096850395203, "global_step": 79732, "epoch": 895, "lr": 8.03432763040651e-05}
{"train_loss": 0.23969289660453796, "global_step": 79733, "epoch": 895, "lr": 8.034281552474971e-05}
{"train_loss": 0.2210398018360138, "global_step": 79734, "epoch": 895, "lr": 8.03423547413551e-05}
{"train_loss": 0.11915791034698486, "global_step": 79735, "epoch": 895, "lr": 8.034189395388133e-05}
{"train_loss": 0.163235142827034, "global_step": 79736, "epoch": 895, "lr": 8.034143316232846e-05}
{"train_loss": 0.1812465637922287, "global_step": 79737, "epoch": 895, "lr": 8.034097236669655e-05}
{"train_loss": 0.1780558079481125, "global_step": 79738, "epoch": 895, "lr": 8.034051156698567e-05}
{"train_loss": 0.16231325268745422, "global_step": 79739, "epoch": 895, "lr": 8.034005076319586e-05}
{"train_loss": 0.24302634596824646, "global_step": 79740, "epoch": 895, "lr": 8.033958995532722e-05}
{"train_loss": 0.18558871746063232, "global_step": 79741, "epoch": 895, "lr": 8.033912914337978e-05}
{"train_loss": 0.21868591010570526, "global_step": 79742, "epoch": 895, "lr": 8.033866832735361e-05}
{"train_loss": 0.20528954778159603, "global_step": 79743, "epoch": 895, "lr": 8.033820750724881e-05, "val_loss": 3.9816906452178955, "train_action_mse_error": 21.171321868896484}
{"train_loss": 0.20161819458007812, "global_step": 79744, "epoch": 896, "lr": 8.033774668306537e-05}
{"train_loss": 0.15982724726200104, "global_step": 79745, "epoch": 896, "lr": 8.03372858548034e-05}
{"train_loss": 0.17343826591968536, "global_step": 79746, "epoch": 896, "lr": 8.033682502246296e-05}
{"train_loss": 0.18419158458709717, "global_step": 79747, "epoch": 896, "lr": 8.033636418604411e-05}
{"train_loss": 0.11391916126012802, "global_step": 79748, "epoch": 896, "lr": 8.033590334554688e-05}
{"train_loss": 0.1920226365327835, "global_step": 79749, "epoch": 896, "lr": 8.033544250097138e-05}
{"train_loss": 0.156423419713974, "global_step": 79750, "epoch": 896, "lr": 8.033498165231764e-05}
{"train_loss": 0.16062861680984497, "global_step": 79751, "epoch": 896, "lr": 8.033452079958573e-05}
{"train_loss": 0.15960747003555298, "global_step": 79752, "epoch": 896, "lr": 8.033405994277573e-05}
{"train_loss": 0.18783165514469147, "global_step": 79753, "epoch": 896, "lr": 8.033359908188767e-05}
{"train_loss": 0.16767339408397675, "global_step": 79754, "epoch": 896, "lr": 8.033313821692162e-05}
{"train_loss": 0.18416965007781982, "global_step": 79755, "epoch": 896, "lr": 8.033267734787766e-05}
{"train_loss": 0.11047802865505219, "global_step": 79756, "epoch": 896, "lr": 8.033221647475583e-05}
{"train_loss": 0.16106614470481873, "global_step": 79757, "epoch": 896, "lr": 8.033175559755623e-05}
{"train_loss": 0.21053308248519897, "global_step": 79758, "epoch": 896, "lr": 8.033129471627886e-05}
{"train_loss": 0.18407388031482697, "global_step": 79759, "epoch": 896, "lr": 8.033083383092383e-05}
{"train_loss": 0.18465515971183777, "global_step": 79760, "epoch": 896, "lr": 8.033037294149119e-05}
{"train_loss": 0.17774683237075806, "global_step": 79761, "epoch": 896, "lr": 8.032991204798101e-05}
{"train_loss": 0.23943965137004852, "global_step": 79762, "epoch": 896, "lr": 8.032945115039334e-05}
{"train_loss": 0.2227523773908615, "global_step": 79763, "epoch": 896, "lr": 8.032899024872825e-05}
{"train_loss": 0.15766064822673798, "global_step": 79764, "epoch": 896, "lr": 8.032852934298579e-05}
{"train_loss": 0.16079016029834747, "global_step": 79765, "epoch": 896, "lr": 8.032806843316603e-05}
{"train_loss": 0.14692671597003937, "global_step": 79766, "epoch": 896, "lr": 8.032760751926903e-05}
{"train_loss": 0.2225760519504547, "global_step": 79767, "epoch": 896, "lr": 8.032714660129486e-05}
{"train_loss": 0.15155990421772003, "global_step": 79768, "epoch": 896, "lr": 8.032668567924356e-05}
{"train_loss": 0.12944181263446808, "global_step": 79769, "epoch": 896, "lr": 8.032622475311522e-05}
{"train_loss": 0.1567307859659195, "global_step": 79770, "epoch": 896, "lr": 8.032576382290988e-05}
{"train_loss": 0.18177969753742218, "global_step": 79771, "epoch": 896, "lr": 8.03253028886276e-05}
{"train_loss": 0.18065504729747772, "global_step": 79772, "epoch": 896, "lr": 8.032484195026848e-05}
{"train_loss": 0.16043579578399658, "global_step": 79773, "epoch": 896, "lr": 8.032438100783253e-05}
{"train_loss": 0.2461656630039215, "global_step": 79774, "epoch": 896, "lr": 8.032392006131985e-05}
{"train_loss": 0.1996348351240158, "global_step": 79775, "epoch": 896, "lr": 8.03234591107305e-05}
{"train_loss": 0.16217324137687683, "global_step": 79776, "epoch": 896, "lr": 8.03229981560645e-05}
{"train_loss": 0.2834596335887909, "global_step": 79777, "epoch": 896, "lr": 8.032253719732195e-05}
{"train_loss": 0.20328651368618011, "global_step": 79778, "epoch": 896, "lr": 8.032207623450292e-05}
{"train_loss": 0.2102433145046234, "global_step": 79779, "epoch": 896, "lr": 8.032161526760746e-05}
{"train_loss": 0.14560763537883759, "global_step": 79780, "epoch": 896, "lr": 8.03211542966356e-05}
{"train_loss": 0.2009577453136444, "global_step": 79781, "epoch": 896, "lr": 8.032069332158744e-05}
{"train_loss": 0.16121116280555725, "global_step": 79782, "epoch": 896, "lr": 8.032023234246306e-05}
{"train_loss": 0.1550835371017456, "global_step": 79783, "epoch": 896, "lr": 8.031977135926248e-05}
{"train_loss": 0.17203228175640106, "global_step": 79784, "epoch": 896, "lr": 8.031931037198576e-05}
{"train_loss": 0.14249621331691742, "global_step": 79785, "epoch": 896, "lr": 8.031884938063299e-05}
{"train_loss": 0.17648667097091675, "global_step": 79786, "epoch": 896, "lr": 8.031838838520422e-05}
{"train_loss": 0.23853901028633118, "global_step": 79787, "epoch": 896, "lr": 8.031792738569952e-05}
{"train_loss": 0.2399783730506897, "global_step": 79788, "epoch": 896, "lr": 8.031746638211893e-05}
{"train_loss": 0.2342473268508911, "global_step": 79789, "epoch": 896, "lr": 8.031700537446254e-05}
{"train_loss": 0.27476218342781067, "global_step": 79790, "epoch": 896, "lr": 8.03165443627304e-05}
{"train_loss": 0.2043677270412445, "global_step": 79791, "epoch": 896, "lr": 8.031608334692255e-05}
{"train_loss": 0.24016594886779785, "global_step": 79792, "epoch": 896, "lr": 8.031562232703908e-05}
{"train_loss": 0.16846439242362976, "global_step": 79793, "epoch": 896, "lr": 8.031516130308004e-05}
{"train_loss": 0.268317848443985, "global_step": 79794, "epoch": 896, "lr": 8.03147002750455e-05}
{"train_loss": 0.16842105984687805, "global_step": 79795, "epoch": 896, "lr": 8.031423924293552e-05}
{"train_loss": 0.1739470213651657, "global_step": 79796, "epoch": 896, "lr": 8.031377820675017e-05}
{"train_loss": 0.18176811933517456, "global_step": 79797, "epoch": 896, "lr": 8.031331716648948e-05}
{"train_loss": 0.19662928581237793, "global_step": 79798, "epoch": 896, "lr": 8.031285612215356e-05}
{"train_loss": 0.24164582788944244, "global_step": 79799, "epoch": 896, "lr": 8.031239507374242e-05}
{"train_loss": 0.1335650086402893, "global_step": 79800, "epoch": 896, "lr": 8.031193402125617e-05}
{"train_loss": 0.20639364421367645, "global_step": 79801, "epoch": 896, "lr": 8.031147296469484e-05}
{"train_loss": 0.22744326293468475, "global_step": 79802, "epoch": 896, "lr": 8.03110119040585e-05}
{"train_loss": 0.23479335010051727, "global_step": 79803, "epoch": 896, "lr": 8.031055083934722e-05}
{"train_loss": 0.2351285219192505, "global_step": 79804, "epoch": 896, "lr": 8.031008977056104e-05}
{"train_loss": 0.18995454907417297, "global_step": 79805, "epoch": 896, "lr": 8.030962869770007e-05}
{"train_loss": 0.2775273323059082, "global_step": 79806, "epoch": 896, "lr": 8.030916762076431e-05}
{"train_loss": 0.17181174457073212, "global_step": 79807, "epoch": 896, "lr": 8.030870653975386e-05}
{"train_loss": 0.17237749695777893, "global_step": 79808, "epoch": 896, "lr": 8.030824545466879e-05}
{"train_loss": 0.18603797256946564, "global_step": 79809, "epoch": 896, "lr": 8.030778436550912e-05}
{"train_loss": 0.2364124059677124, "global_step": 79810, "epoch": 896, "lr": 8.030732327227495e-05}
{"train_loss": 0.22855493426322937, "global_step": 79811, "epoch": 896, "lr": 8.030686217496632e-05}
{"train_loss": 0.24723953008651733, "global_step": 79812, "epoch": 896, "lr": 8.030640107358332e-05}
{"train_loss": 0.16457587480545044, "global_step": 79813, "epoch": 896, "lr": 8.030593996812598e-05}
{"train_loss": 0.21272578835487366, "global_step": 79814, "epoch": 896, "lr": 8.030547885859439e-05}
{"train_loss": 0.19448231160640717, "global_step": 79815, "epoch": 896, "lr": 8.030501774498858e-05}
{"train_loss": 0.19594979286193848, "global_step": 79816, "epoch": 896, "lr": 8.030455662730865e-05}
{"train_loss": 0.2696061432361603, "global_step": 79817, "epoch": 896, "lr": 8.030409550555462e-05}
{"train_loss": 0.19559647142887115, "global_step": 79818, "epoch": 896, "lr": 8.030363437972659e-05}
{"train_loss": 0.23320041596889496, "global_step": 79819, "epoch": 896, "lr": 8.03031732498246e-05}
{"train_loss": 0.19737060368061066, "global_step": 79820, "epoch": 896, "lr": 8.030271211584872e-05}
{"train_loss": 0.25201618671417236, "global_step": 79821, "epoch": 896, "lr": 8.0302250977799e-05}
{"train_loss": 0.17280778288841248, "global_step": 79822, "epoch": 896, "lr": 8.030178983567552e-05}
{"train_loss": 0.19878189265727997, "global_step": 79823, "epoch": 896, "lr": 8.030132868947833e-05}
{"train_loss": 0.1694527566432953, "global_step": 79824, "epoch": 896, "lr": 8.030086753920749e-05}
{"train_loss": 0.25382328033447266, "global_step": 79825, "epoch": 896, "lr": 8.030040638486308e-05}
{"train_loss": 0.16539603471755981, "global_step": 79826, "epoch": 896, "lr": 8.029994522644515e-05}
{"train_loss": 0.2534257173538208, "global_step": 79827, "epoch": 896, "lr": 8.029948406395375e-05}
{"train_loss": 0.18708956241607666, "global_step": 79828, "epoch": 896, "lr": 8.029902289738895e-05}
{"train_loss": 0.24065543711185455, "global_step": 79829, "epoch": 896, "lr": 8.029856172675083e-05}
{"train_loss": 0.22501689195632935, "global_step": 79830, "epoch": 896, "lr": 8.029810055203943e-05}
{"train_loss": 0.2953626811504364, "global_step": 79831, "epoch": 896, "lr": 8.029763937325483e-05}
{"train_loss": 0.19674490199665004, "global_step": 79832, "epoch": 896, "lr": 8.029717819039707e-05, "val_loss": 3.9196174144744873}
{"train_loss": 0.11460717022418976, "global_step": 79833, "epoch": 897, "lr": 8.029671700346623e-05}
{"train_loss": 0.23395872116088867, "global_step": 79834, "epoch": 897, "lr": 8.029625581246236e-05}
{"train_loss": 0.22053250670433044, "global_step": 79835, "epoch": 897, "lr": 8.029579461738554e-05}
{"train_loss": 0.20339234173297882, "global_step": 79836, "epoch": 897, "lr": 8.029533341823581e-05}
{"train_loss": 0.20902255177497864, "global_step": 79837, "epoch": 897, "lr": 8.029487221501324e-05}
{"train_loss": 0.24793672561645508, "global_step": 79838, "epoch": 897, "lr": 8.029441100771789e-05}
{"train_loss": 0.22851958870887756, "global_step": 79839, "epoch": 897, "lr": 8.029394979634983e-05}
{"train_loss": 0.18164004385471344, "global_step": 79840, "epoch": 897, "lr": 8.029348858090913e-05}
{"train_loss": 0.15088626742362976, "global_step": 79841, "epoch": 897, "lr": 8.029302736139584e-05}
{"train_loss": 0.19123265147209167, "global_step": 79842, "epoch": 897, "lr": 8.029256613781e-05}
{"train_loss": 0.19166482985019684, "global_step": 79843, "epoch": 897, "lr": 8.029210491015172e-05}
{"train_loss": 0.18972893059253693, "global_step": 79844, "epoch": 897, "lr": 8.029164367842101e-05}
{"train_loss": 0.1591150015592575, "global_step": 79845, "epoch": 897, "lr": 8.029118244261797e-05}
{"train_loss": 0.15227878093719482, "global_step": 79846, "epoch": 897, "lr": 8.029072120274265e-05}
{"train_loss": 0.19032025337219238, "global_step": 79847, "epoch": 897, "lr": 8.029025995879511e-05}
{"train_loss": 0.25415143370628357, "global_step": 79848, "epoch": 897, "lr": 8.02897987107754e-05}
{"train_loss": 0.1431267112493515, "global_step": 79849, "epoch": 897, "lr": 8.028933745868361e-05}
{"train_loss": 0.20507419109344482, "global_step": 79850, "epoch": 897, "lr": 8.028887620251979e-05}
{"train_loss": 0.19348789751529694, "global_step": 79851, "epoch": 897, "lr": 8.028841494228401e-05}
{"train_loss": 0.17188510298728943, "global_step": 79852, "epoch": 897, "lr": 8.02879536779763e-05}
{"train_loss": 0.22834137082099915, "global_step": 79853, "epoch": 897, "lr": 8.028749240959674e-05}
{"train_loss": 0.26721516251564026, "global_step": 79854, "epoch": 897, "lr": 8.028703113714542e-05}
{"train_loss": 0.24537794291973114, "global_step": 79855, "epoch": 897, "lr": 8.028656986062236e-05}
{"train_loss": 0.24223099648952484, "global_step": 79856, "epoch": 897, "lr": 8.028610858002766e-05}
{"train_loss": 0.1775640994310379, "global_step": 79857, "epoch": 897, "lr": 8.028564729536135e-05}
{"train_loss": 0.30039891600608826, "global_step": 79858, "epoch": 897, "lr": 8.02851860066235e-05}
{"train_loss": 0.1780747026205063, "global_step": 79859, "epoch": 897, "lr": 8.028472471381418e-05}
{"train_loss": 0.21348996460437775, "global_step": 79860, "epoch": 897, "lr": 8.028426341693344e-05}
{"train_loss": 0.19182348251342773, "global_step": 79861, "epoch": 897, "lr": 8.028380211598136e-05}
{"train_loss": 0.20918375253677368, "global_step": 79862, "epoch": 897, "lr": 8.028334081095799e-05}
{"train_loss": 0.1571727693080902, "global_step": 79863, "epoch": 897, "lr": 8.028287950186337e-05}
{"train_loss": 0.17381508648395538, "global_step": 79864, "epoch": 897, "lr": 8.028241818869762e-05}
{"train_loss": 0.1946619749069214, "global_step": 79865, "epoch": 897, "lr": 8.028195687146076e-05}
{"train_loss": 0.23884621262550354, "global_step": 79866, "epoch": 897, "lr": 8.028149555015285e-05}
{"train_loss": 0.16483250260353088, "global_step": 79867, "epoch": 897, "lr": 8.028103422477396e-05}
{"train_loss": 0.22197552025318146, "global_step": 79868, "epoch": 897, "lr": 8.028057289532415e-05}
{"train_loss": 0.18829573690891266, "global_step": 79869, "epoch": 897, "lr": 8.028011156180351e-05}
{"train_loss": 0.22204181551933289, "global_step": 79870, "epoch": 897, "lr": 8.027965022421206e-05}
{"train_loss": 0.18649421632289886, "global_step": 79871, "epoch": 897, "lr": 8.027918888254988e-05}
{"train_loss": 0.17271536588668823, "global_step": 79872, "epoch": 897, "lr": 8.027872753681704e-05}
{"train_loss": 0.22060656547546387, "global_step": 79873, "epoch": 897, "lr": 8.02782661870136e-05}
{"train_loss": 0.17281898856163025, "global_step": 79874, "epoch": 897, "lr": 8.02778048331396e-05}
{"train_loss": 0.2615021765232086, "global_step": 79875, "epoch": 897, "lr": 8.027734347519514e-05}
{"train_loss": 0.2317725419998169, "global_step": 79876, "epoch": 897, "lr": 8.027688211318023e-05}
{"train_loss": 0.1760215014219284, "global_step": 79877, "epoch": 897, "lr": 8.027642074709498e-05}
{"train_loss": 0.15867005288600922, "global_step": 79878, "epoch": 897, "lr": 8.027595937693944e-05}
{"train_loss": 0.22937554121017456, "global_step": 79879, "epoch": 897, "lr": 8.027549800271365e-05}
{"train_loss": 0.27223503589630127, "global_step": 79880, "epoch": 897, "lr": 8.02750366244177e-05}
{"train_loss": 0.21963709592819214, "global_step": 79881, "epoch": 897, "lr": 8.027457524205164e-05}
{"train_loss": 0.2046988159418106, "global_step": 79882, "epoch": 897, "lr": 8.027411385561554e-05}
{"train_loss": 0.20622937381267548, "global_step": 79883, "epoch": 897, "lr": 8.027365246510944e-05}
{"train_loss": 0.23557530343532562, "global_step": 79884, "epoch": 897, "lr": 8.027319107053341e-05}
{"train_loss": 0.28315213322639465, "global_step": 79885, "epoch": 897, "lr": 8.027272967188752e-05}
{"train_loss": 0.2093532681465149, "global_step": 79886, "epoch": 897, "lr": 8.027226826917184e-05}
{"train_loss": 0.1931031048297882, "global_step": 79887, "epoch": 897, "lr": 8.027180686238644e-05}
{"train_loss": 0.22732794284820557, "global_step": 79888, "epoch": 897, "lr": 8.027134545153134e-05}
{"train_loss": 0.21509292721748352, "global_step": 79889, "epoch": 897, "lr": 8.027088403660663e-05}
{"train_loss": 0.2191477119922638, "global_step": 79890, "epoch": 897, "lr": 8.027042261761238e-05}
{"train_loss": 0.2042182832956314, "global_step": 79891, "epoch": 897, "lr": 8.026996119454862e-05}
{"train_loss": 0.15943725407123566, "global_step": 79892, "epoch": 897, "lr": 8.026949976741544e-05}
{"train_loss": 0.1626731902360916, "global_step": 79893, "epoch": 897, "lr": 8.026903833621291e-05}
{"train_loss": 0.1605115830898285, "global_step": 79894, "epoch": 897, "lr": 8.026857690094108e-05}
{"train_loss": 0.13175596296787262, "global_step": 79895, "epoch": 897, "lr": 8.026811546159998e-05}
{"train_loss": 0.19748269021511078, "global_step": 79896, "epoch": 897, "lr": 8.026765401818973e-05}
{"train_loss": 0.17811501026153564, "global_step": 79897, "epoch": 897, "lr": 8.026719257071034e-05}
{"train_loss": 0.22153429687023163, "global_step": 79898, "epoch": 897, "lr": 8.02667311191619e-05}
{"train_loss": 0.2517056167125702, "global_step": 79899, "epoch": 897, "lr": 8.026626966354447e-05}
{"train_loss": 0.22897225618362427, "global_step": 79900, "epoch": 897, "lr": 8.026580820385812e-05}
{"train_loss": 0.16841065883636475, "global_step": 79901, "epoch": 897, "lr": 8.026534674010289e-05}
{"train_loss": 0.21353770792484283, "global_step": 79902, "epoch": 897, "lr": 8.026488527227884e-05}
{"train_loss": 0.16285818815231323, "global_step": 79903, "epoch": 897, "lr": 8.026442380038606e-05}
{"train_loss": 0.3009840250015259, "global_step": 79904, "epoch": 897, "lr": 8.026396232442459e-05}
{"train_loss": 0.21876594424247742, "global_step": 79905, "epoch": 897, "lr": 8.026350084439451e-05}
{"train_loss": 0.23163218796253204, "global_step": 79906, "epoch": 897, "lr": 8.026303936029587e-05}
{"train_loss": 0.17504964768886566, "global_step": 79907, "epoch": 897, "lr": 8.026257787212873e-05}
{"train_loss": 0.2360800951719284, "global_step": 79908, "epoch": 897, "lr": 8.026211637989316e-05}
{"train_loss": 0.2286710888147354, "global_step": 79909, "epoch": 897, "lr": 8.026165488358921e-05}
{"train_loss": 0.24006593227386475, "global_step": 79910, "epoch": 897, "lr": 8.026119338321693e-05}
{"train_loss": 0.20565836131572723, "global_step": 79911, "epoch": 897, "lr": 8.026073187877644e-05}
{"train_loss": 0.20198467373847961, "global_step": 79912, "epoch": 897, "lr": 8.026027037026774e-05}
{"train_loss": 0.27706655859947205, "global_step": 79913, "epoch": 897, "lr": 8.025980885769093e-05}
{"train_loss": 0.16360865533351898, "global_step": 79914, "epoch": 897, "lr": 8.025934734104604e-05}
{"train_loss": 0.17377446591854095, "global_step": 79915, "epoch": 897, "lr": 8.025888582033316e-05}
{"train_loss": 0.2152751237154007, "global_step": 79916, "epoch": 897, "lr": 8.025842429555235e-05}
{"train_loss": 0.17778533697128296, "global_step": 79917, "epoch": 897, "lr": 8.025796276670366e-05}
{"train_loss": 0.11290979385375977, "global_step": 79918, "epoch": 897, "lr": 8.025750123378713e-05}
{"train_loss": 0.1997665911912918, "global_step": 79919, "epoch": 897, "lr": 8.025703969680287e-05}
{"train_loss": 0.13670852780342102, "global_step": 79920, "epoch": 897, "lr": 8.025657815575092e-05}
{"train_loss": 0.2034321442413866, "global_step": 79921, "epoch": 897, "lr": 8.025611661063133e-05, "val_loss": 3.9589555263519287}
{"train_loss": 0.15829433500766754, "global_step": 79922, "epoch": 898, "lr": 8.025565506144418e-05}
{"train_loss": 0.10167653858661652, "global_step": 79923, "epoch": 898, "lr": 8.025519350818953e-05}
{"train_loss": 0.2233210653066635, "global_step": 79924, "epoch": 898, "lr": 8.025473195086744e-05}
{"train_loss": 0.1871701031923294, "global_step": 79925, "epoch": 898, "lr": 8.025427038947797e-05}
{"train_loss": 0.236393541097641, "global_step": 79926, "epoch": 898, "lr": 8.025380882402118e-05}
{"train_loss": 0.27214571833610535, "global_step": 79927, "epoch": 898, "lr": 8.025334725449714e-05}
{"train_loss": 0.2139519453048706, "global_step": 79928, "epoch": 898, "lr": 8.025288568090588e-05}
{"train_loss": 0.1260938197374344, "global_step": 79929, "epoch": 898, "lr": 8.025242410324752e-05}
{"train_loss": 0.1655658632516861, "global_step": 79930, "epoch": 898, "lr": 8.025196252152207e-05}
{"train_loss": 0.2141728550195694, "global_step": 79931, "epoch": 898, "lr": 8.025150093572961e-05}
{"train_loss": 0.20914947986602783, "global_step": 79932, "epoch": 898, "lr": 8.025103934587023e-05}
{"train_loss": 0.20375056564807892, "global_step": 79933, "epoch": 898, "lr": 8.025057775194396e-05}
{"train_loss": 0.2092006951570511, "global_step": 79934, "epoch": 898, "lr": 8.025011615395085e-05}
{"train_loss": 0.25758931040763855, "global_step": 79935, "epoch": 898, "lr": 8.024965455189098e-05}
{"train_loss": 0.21896593272686005, "global_step": 79936, "epoch": 898, "lr": 8.024919294576442e-05}
{"train_loss": 0.2177393138408661, "global_step": 79937, "epoch": 898, "lr": 8.024873133557125e-05}
{"train_loss": 0.20168070495128632, "global_step": 79938, "epoch": 898, "lr": 8.024826972131147e-05}
{"train_loss": 0.229502871632576, "global_step": 79939, "epoch": 898, "lr": 8.024780810298518e-05}
{"train_loss": 0.18987786769866943, "global_step": 79940, "epoch": 898, "lr": 8.024734648059247e-05}
{"train_loss": 0.21094726026058197, "global_step": 79941, "epoch": 898, "lr": 8.024688485413334e-05}
{"train_loss": 0.16107216477394104, "global_step": 79942, "epoch": 898, "lr": 8.024642322360791e-05}
{"train_loss": 0.18002940714359283, "global_step": 79943, "epoch": 898, "lr": 8.02459615890162e-05}
{"train_loss": 0.17038340866565704, "global_step": 79944, "epoch": 898, "lr": 8.024549995035831e-05}
{"train_loss": 0.1640750616788864, "global_step": 79945, "epoch": 898, "lr": 8.024503830763427e-05}
{"train_loss": 0.1736612617969513, "global_step": 79946, "epoch": 898, "lr": 8.024457666084416e-05}
{"train_loss": 0.1588497906923294, "global_step": 79947, "epoch": 898, "lr": 8.024411500998802e-05}
{"train_loss": 0.18788762390613556, "global_step": 79948, "epoch": 898, "lr": 8.024365335506594e-05}
{"train_loss": 0.18866518139839172, "global_step": 79949, "epoch": 898, "lr": 8.024319169607795e-05}
{"train_loss": 0.24677321314811707, "global_step": 79950, "epoch": 898, "lr": 8.024273003302417e-05}
{"train_loss": 0.17203325033187866, "global_step": 79951, "epoch": 898, "lr": 8.024226836590458e-05}
{"train_loss": 0.14193178713321686, "global_step": 79952, "epoch": 898, "lr": 8.02418066947193e-05}
{"train_loss": 0.22373495995998383, "global_step": 79953, "epoch": 898, "lr": 8.024134501946838e-05}
{"train_loss": 0.19955314695835114, "global_step": 79954, "epoch": 898, "lr": 8.02408833401519e-05}
{"train_loss": 0.19291098415851593, "global_step": 79955, "epoch": 898, "lr": 8.024042165676989e-05}
{"train_loss": 0.25305408239364624, "global_step": 79956, "epoch": 898, "lr": 8.02399599693224e-05}
{"train_loss": 0.15245087444782257, "global_step": 79957, "epoch": 898, "lr": 8.023949827780953e-05}
{"train_loss": 0.13462772965431213, "global_step": 79958, "epoch": 898, "lr": 8.023903658223133e-05}
{"train_loss": 0.18874841928482056, "global_step": 79959, "epoch": 898, "lr": 8.023857488258786e-05}
{"train_loss": 0.24349209666252136, "global_step": 79960, "epoch": 898, "lr": 8.023811317887919e-05}
{"train_loss": 0.16402007639408112, "global_step": 79961, "epoch": 898, "lr": 8.023765147110537e-05}
{"train_loss": 0.18217772245407104, "global_step": 79962, "epoch": 898, "lr": 8.023718975926646e-05}
{"train_loss": 0.2918964922428131, "global_step": 79963, "epoch": 898, "lr": 8.023672804336254e-05}
{"train_loss": 0.17044243216514587, "global_step": 79964, "epoch": 898, "lr": 8.023626632339365e-05}
{"train_loss": 0.17961986362934113, "global_step": 79965, "epoch": 898, "lr": 8.023580459935987e-05}
{"train_loss": 0.1580333709716797, "global_step": 79966, "epoch": 898, "lr": 8.023534287126126e-05}
{"train_loss": 0.22978493571281433, "global_step": 79967, "epoch": 898, "lr": 8.023488113909786e-05}
{"train_loss": 0.19540823996067047, "global_step": 79968, "epoch": 898, "lr": 8.023441940286975e-05}
{"train_loss": 0.13383327424526215, "global_step": 79969, "epoch": 898, "lr": 8.0233957662577e-05}
{"train_loss": 0.15436789393424988, "global_step": 79970, "epoch": 898, "lr": 8.023349591821967e-05}
{"train_loss": 0.12010317295789719, "global_step": 79971, "epoch": 898, "lr": 8.02330341697978e-05}
{"train_loss": 0.12533371150493622, "global_step": 79972, "epoch": 898, "lr": 8.023257241731147e-05}
{"train_loss": 0.22838887572288513, "global_step": 79973, "epoch": 898, "lr": 8.023211066076073e-05}
{"train_loss": 0.1762954592704773, "global_step": 79974, "epoch": 898, "lr": 8.023164890014568e-05}
{"train_loss": 0.11997686326503754, "global_step": 79975, "epoch": 898, "lr": 8.023118713546634e-05}
{"train_loss": 0.1539742797613144, "global_step": 79976, "epoch": 898, "lr": 8.023072536672276e-05}
{"train_loss": 0.17505782842636108, "global_step": 79977, "epoch": 898, "lr": 8.023026359391506e-05}
{"train_loss": 0.1591741293668747, "global_step": 79978, "epoch": 898, "lr": 8.022980181704324e-05}
{"train_loss": 0.14184781908988953, "global_step": 79979, "epoch": 898, "lr": 8.022934003610742e-05}
{"train_loss": 0.182635098695755, "global_step": 79980, "epoch": 898, "lr": 8.022887825110761e-05}
{"train_loss": 0.22946898639202118, "global_step": 79981, "epoch": 898, "lr": 8.022841646204392e-05}
{"train_loss": 0.2052355408668518, "global_step": 79982, "epoch": 898, "lr": 8.022795466891638e-05}
{"train_loss": 0.2582378685474396, "global_step": 79983, "epoch": 898, "lr": 8.022749287172506e-05}
{"train_loss": 0.19148659706115723, "global_step": 79984, "epoch": 898, "lr": 8.022703107047001e-05}
{"train_loss": 0.1409345418214798, "global_step": 79985, "epoch": 898, "lr": 8.022656926515132e-05}
{"train_loss": 0.1514727920293808, "global_step": 79986, "epoch": 898, "lr": 8.022610745576903e-05}
{"train_loss": 0.19665350019931793, "global_step": 79987, "epoch": 898, "lr": 8.02256456423232e-05}
{"train_loss": 0.2780406177043915, "global_step": 79988, "epoch": 898, "lr": 8.02251838248139e-05}
{"train_loss": 0.1685008704662323, "global_step": 79989, "epoch": 898, "lr": 8.022472200324121e-05}
{"train_loss": 0.16888470947742462, "global_step": 79990, "epoch": 898, "lr": 8.022426017760516e-05}
{"train_loss": 0.10875369608402252, "global_step": 79991, "epoch": 898, "lr": 8.022379834790583e-05}
{"train_loss": 0.19915267825126648, "global_step": 79992, "epoch": 898, "lr": 8.022333651414329e-05}
{"train_loss": 0.20520660281181335, "global_step": 79993, "epoch": 898, "lr": 8.022287467631757e-05}
{"train_loss": 0.22338059544563293, "global_step": 79994, "epoch": 898, "lr": 8.022241283442877e-05}
{"train_loss": 0.2455609142780304, "global_step": 79995, "epoch": 898, "lr": 8.022195098847692e-05}
{"train_loss": 0.2695774435997009, "global_step": 79996, "epoch": 898, "lr": 8.02214891384621e-05}
{"train_loss": 0.17212457954883575, "global_step": 79997, "epoch": 898, "lr": 8.022102728438439e-05}
{"train_loss": 0.1731335073709488, "global_step": 79998, "epoch": 898, "lr": 8.02205654262438e-05}
{"train_loss": 0.24318832159042358, "global_step": 79999, "epoch": 898, "lr": 8.022010356404046e-05}
{"train_loss": 0.1762923151254654, "global_step": 80000, "epoch": 898, "lr": 8.021964169777436e-05}
{"train_loss": 0.1621709167957306, "global_step": 80001, "epoch": 898, "lr": 8.021917982744562e-05}
{"train_loss": 0.09307616949081421, "global_step": 80002, "epoch": 898, "lr": 8.021871795305426e-05}
{"train_loss": 0.21041372418403625, "global_step": 80003, "epoch": 898, "lr": 8.021825607460039e-05}
{"train_loss": 0.21115642786026, "global_step": 80004, "epoch": 898, "lr": 8.021779419208401e-05}
{"train_loss": 0.221849262714386, "global_step": 80005, "epoch": 898, "lr": 8.021733230550524e-05}
{"train_loss": 0.14712952077388763, "global_step": 80006, "epoch": 898, "lr": 8.021687041486412e-05}
{"train_loss": 0.3034709692001343, "global_step": 80007, "epoch": 898, "lr": 8.021640852016069e-05}
{"train_loss": 0.2327549010515213, "global_step": 80008, "epoch": 898, "lr": 8.021594662139504e-05}
{"train_loss": 0.18806768953800201, "global_step": 80009, "epoch": 898, "lr": 8.021548471856724e-05}
{"train_loss": 0.19046205407782887, "global_step": 80010, "epoch": 898, "lr": 8.021502281167731e-05, "val_loss": 4.038058757781982}
{"train_loss": 0.1717517077922821, "global_step": 80011, "epoch": 899, "lr": 8.021456090072536e-05}
{"train_loss": 0.18714217841625214, "global_step": 80012, "epoch": 899, "lr": 8.021409898571142e-05}
{"train_loss": 0.1403389275074005, "global_step": 80013, "epoch": 899, "lr": 8.021363706663556e-05}
{"train_loss": 0.18503715097904205, "global_step": 80014, "epoch": 899, "lr": 8.021317514349785e-05}
{"train_loss": 0.21174073219299316, "global_step": 80015, "epoch": 899, "lr": 8.021271321629835e-05}
{"train_loss": 0.1814211905002594, "global_step": 80016, "epoch": 899, "lr": 8.021225128503711e-05}
{"train_loss": 0.382554292678833, "global_step": 80017, "epoch": 899, "lr": 8.021178934971421e-05}
{"train_loss": 0.18241256475448608, "global_step": 80018, "epoch": 899, "lr": 8.02113274103297e-05}
{"train_loss": 0.3136844336986542, "global_step": 80019, "epoch": 899, "lr": 8.021086546688364e-05}
{"train_loss": 0.13045333325862885, "global_step": 80020, "epoch": 899, "lr": 8.021040351937611e-05}
{"train_loss": 0.1389082670211792, "global_step": 80021, "epoch": 899, "lr": 8.020994156780715e-05}
{"train_loss": 0.10402103513479233, "global_step": 80022, "epoch": 899, "lr": 8.020947961217684e-05}
{"train_loss": 0.14756229519844055, "global_step": 80023, "epoch": 899, "lr": 8.020901765248523e-05}
{"train_loss": 0.16468562185764313, "global_step": 80024, "epoch": 899, "lr": 8.020855568873238e-05}
{"train_loss": 0.19837678968906403, "global_step": 80025, "epoch": 899, "lr": 8.020809372091836e-05}
{"train_loss": 0.23609226942062378, "global_step": 80026, "epoch": 899, "lr": 8.020763174904323e-05}
{"train_loss": 0.24679316580295563, "global_step": 80027, "epoch": 899, "lr": 8.020716977310704e-05}
{"train_loss": 0.11834275722503662, "global_step": 80028, "epoch": 899, "lr": 8.020670779310989e-05}
{"train_loss": 0.19030572474002838, "global_step": 80029, "epoch": 899, "lr": 8.020624580905181e-05}
{"train_loss": 0.22988685965538025, "global_step": 80030, "epoch": 899, "lr": 8.020578382093285e-05}
{"train_loss": 0.2639843821525574, "global_step": 80031, "epoch": 899, "lr": 8.020532182875311e-05}
{"train_loss": 0.18288731575012207, "global_step": 80032, "epoch": 899, "lr": 8.020485983251262e-05}
{"train_loss": 0.26075202226638794, "global_step": 80033, "epoch": 899, "lr": 8.020439783221145e-05}
{"train_loss": 0.25420936942100525, "global_step": 80034, "epoch": 899, "lr": 8.020393582784968e-05}
{"train_loss": 0.1894969344139099, "global_step": 80035, "epoch": 899, "lr": 8.020347381942737e-05}
{"train_loss": 0.18489548563957214, "global_step": 80036, "epoch": 899, "lr": 8.020301180694455e-05}
{"train_loss": 0.13346204161643982, "global_step": 80037, "epoch": 899, "lr": 8.020254979040132e-05}
{"train_loss": 0.18937982618808746, "global_step": 80038, "epoch": 899, "lr": 8.02020877697977e-05}
{"train_loss": 0.20414455235004425, "global_step": 80039, "epoch": 899, "lr": 8.02016257451338e-05}
{"train_loss": 0.156940296292305, "global_step": 80040, "epoch": 899, "lr": 8.020116371640964e-05}
{"train_loss": 0.1992672234773636, "global_step": 80041, "epoch": 899, "lr": 8.020070168362532e-05}
{"train_loss": 0.15244656801223755, "global_step": 80042, "epoch": 899, "lr": 8.020023964678088e-05}
{"train_loss": 0.1867154985666275, "global_step": 80043, "epoch": 899, "lr": 8.019977760587639e-05}
{"train_loss": 0.19130930304527283, "global_step": 80044, "epoch": 899, "lr": 8.01993155609119e-05}
{"train_loss": 0.18587861955165863, "global_step": 80045, "epoch": 899, "lr": 8.019885351188748e-05}
{"train_loss": 0.2450246512889862, "global_step": 80046, "epoch": 899, "lr": 8.019839145880319e-05}
{"train_loss": 0.18329012393951416, "global_step": 80047, "epoch": 899, "lr": 8.01979294016591e-05}
{"train_loss": 0.2903475761413574, "global_step": 80048, "epoch": 899, "lr": 8.019746734045525e-05}
{"train_loss": 0.184122234582901, "global_step": 80049, "epoch": 899, "lr": 8.019700527519173e-05}
{"train_loss": 0.1290350705385208, "global_step": 80050, "epoch": 899, "lr": 8.019654320586861e-05}
{"train_loss": 0.1946326345205307, "global_step": 80051, "epoch": 899, "lr": 8.019608113248592e-05}
{"train_loss": 0.2888220548629761, "global_step": 80052, "epoch": 899, "lr": 8.019561905504372e-05}
{"train_loss": 0.1756443977355957, "global_step": 80053, "epoch": 899, "lr": 8.01951569735421e-05}
{"train_loss": 0.12044183164834976, "global_step": 80054, "epoch": 899, "lr": 8.01946948879811e-05}
{"train_loss": 0.13034950196743011, "global_step": 80055, "epoch": 899, "lr": 8.01942327983608e-05}
{"train_loss": 0.2166530191898346, "global_step": 80056, "epoch": 899, "lr": 8.019377070468126e-05}
{"train_loss": 0.20953336358070374, "global_step": 80057, "epoch": 899, "lr": 8.019330860694253e-05}
{"train_loss": 0.16110120713710785, "global_step": 80058, "epoch": 899, "lr": 8.019284650514467e-05}
{"train_loss": 0.18318277597427368, "global_step": 80059, "epoch": 899, "lr": 8.019238439928776e-05}
{"train_loss": 0.19399376213550568, "global_step": 80060, "epoch": 899, "lr": 8.019192228937185e-05}
{"train_loss": 0.1617848426103592, "global_step": 80061, "epoch": 899, "lr": 8.0191460175397e-05}
{"train_loss": 0.22102206945419312, "global_step": 80062, "epoch": 899, "lr": 8.019099805736328e-05}
{"train_loss": 0.19451938569545746, "global_step": 80063, "epoch": 899, "lr": 8.019053593527075e-05}
{"train_loss": 0.22744107246398926, "global_step": 80064, "epoch": 899, "lr": 8.019007380911946e-05}
{"train_loss": 0.2369139939546585, "global_step": 80065, "epoch": 899, "lr": 8.01896116789095e-05}
{"train_loss": 0.22284217178821564, "global_step": 80066, "epoch": 899, "lr": 8.018914954464089e-05}
{"train_loss": 0.19194971024990082, "global_step": 80067, "epoch": 899, "lr": 8.018868740631374e-05}
{"train_loss": 0.2084542065858841, "global_step": 80068, "epoch": 899, "lr": 8.018822526392806e-05}
{"train_loss": 0.19684989750385284, "global_step": 80069, "epoch": 899, "lr": 8.018776311748397e-05}
{"train_loss": 0.1529708206653595, "global_step": 80070, "epoch": 899, "lr": 8.018730096698149e-05}
{"train_loss": 0.2121025025844574, "global_step": 80071, "epoch": 899, "lr": 8.018683881242071e-05}
{"train_loss": 0.18335872888565063, "global_step": 80072, "epoch": 899, "lr": 8.018637665380166e-05}
{"train_loss": 0.19522574543952942, "global_step": 80073, "epoch": 899, "lr": 8.018591449112442e-05}
{"train_loss": 0.2671506404876709, "global_step": 80074, "epoch": 899, "lr": 8.018545232438905e-05}
{"train_loss": 0.19235476851463318, "global_step": 80075, "epoch": 899, "lr": 8.018499015359565e-05}
{"train_loss": 0.16600927710533142, "global_step": 80076, "epoch": 899, "lr": 8.018452797874421e-05}
{"train_loss": 0.12592393159866333, "global_step": 80077, "epoch": 899, "lr": 8.018406579983482e-05}
{"train_loss": 0.222745880484581, "global_step": 80078, "epoch": 899, "lr": 8.018360361686756e-05}
{"train_loss": 0.18419387936592102, "global_step": 80079, "epoch": 899, "lr": 8.018314142984252e-05}
{"train_loss": 0.24299855530261993, "global_step": 80080, "epoch": 899, "lr": 8.018267923875968e-05}
{"train_loss": 0.25412717461586, "global_step": 80081, "epoch": 899, "lr": 8.018221704361916e-05}
{"train_loss": 0.2306312471628189, "global_step": 80082, "epoch": 899, "lr": 8.0181754844421e-05}
{"train_loss": 0.1355869472026825, "global_step": 80083, "epoch": 899, "lr": 8.018129264116528e-05}
{"train_loss": 0.20530463755130768, "global_step": 80084, "epoch": 899, "lr": 8.018083043385205e-05}
{"train_loss": 0.1777932494878769, "global_step": 80085, "epoch": 899, "lr": 8.018036822248138e-05}
{"train_loss": 0.21708618104457855, "global_step": 80086, "epoch": 899, "lr": 8.017990600705332e-05}
{"train_loss": 0.18784144520759583, "global_step": 80087, "epoch": 899, "lr": 8.017944378756794e-05}
{"train_loss": 0.22771725058555603, "global_step": 80088, "epoch": 899, "lr": 8.017898156402531e-05}
{"train_loss": 0.21060147881507874, "global_step": 80089, "epoch": 899, "lr": 8.017851933642548e-05}
{"train_loss": 0.16340413689613342, "global_step": 80090, "epoch": 899, "lr": 8.017805710476851e-05}
{"train_loss": 0.1644136756658554, "global_step": 80091, "epoch": 899, "lr": 8.017759486905448e-05}
{"train_loss": 0.19273978471755981, "global_step": 80092, "epoch": 899, "lr": 8.017713262928344e-05}
{"train_loss": 0.19302958250045776, "global_step": 80093, "epoch": 899, "lr": 8.017667038545544e-05}
{"train_loss": 0.2320535033941269, "global_step": 80094, "epoch": 899, "lr": 8.017620813757056e-05}
{"train_loss": 0.25741279125213623, "global_step": 80095, "epoch": 899, "lr": 8.017574588562887e-05}
{"train_loss": 0.18961569666862488, "global_step": 80096, "epoch": 899, "lr": 8.017528362963039e-05}
{"train_loss": 0.1456361711025238, "global_step": 80097, "epoch": 899, "lr": 8.017482136957523e-05}
{"train_loss": 0.17752885818481445, "global_step": 80098, "epoch": 899, "lr": 8.017435910546343e-05}
{"train_loss": 0.1962669573807984, "global_step": 80099, "epoch": 899, "lr": 8.017389683729505e-05, "val_loss": 4.013890743255615}
{"train_loss": 0.1650073528289795, "global_step": 80100, "epoch": 900, "lr": 8.017343456507016e-05}
{"train_loss": 0.20752336084842682, "global_step": 80101, "epoch": 900, "lr": 8.017297228878883e-05}
{"train_loss": 0.15880846977233887, "global_step": 80102, "epoch": 900, "lr": 8.01725100084511e-05}
{"train_loss": 0.1955931931734085, "global_step": 80103, "epoch": 900, "lr": 8.017204772405704e-05}
{"train_loss": 0.24518738687038422, "global_step": 80104, "epoch": 900, "lr": 8.017158543560672e-05}
{"train_loss": 0.2634952962398529, "global_step": 80105, "epoch": 900, "lr": 8.01711231431002e-05}
{"train_loss": 0.1313166469335556, "global_step": 80106, "epoch": 900, "lr": 8.017066084653755e-05}
{"train_loss": 0.22465336322784424, "global_step": 80107, "epoch": 900, "lr": 8.01701985459188e-05}
{"train_loss": 0.17132407426834106, "global_step": 80108, "epoch": 900, "lr": 8.016973624124406e-05}
{"train_loss": 0.14864206314086914, "global_step": 80109, "epoch": 900, "lr": 8.016927393251336e-05}
{"train_loss": 0.21714580059051514, "global_step": 80110, "epoch": 900, "lr": 8.016881161972676e-05}
{"train_loss": 0.11847296357154846, "global_step": 80111, "epoch": 900, "lr": 8.016834930288433e-05}
{"train_loss": 0.1799362748861313, "global_step": 80112, "epoch": 900, "lr": 8.016788698198614e-05}
{"train_loss": 0.1512686014175415, "global_step": 80113, "epoch": 900, "lr": 8.016742465703224e-05}
{"train_loss": 0.22828184068202972, "global_step": 80114, "epoch": 900, "lr": 8.01669623280227e-05}
{"train_loss": 0.257199227809906, "global_step": 80115, "epoch": 900, "lr": 8.016649999495758e-05}
{"train_loss": 0.17679129540920258, "global_step": 80116, "epoch": 900, "lr": 8.016603765783695e-05}
{"train_loss": 0.18933239579200745, "global_step": 80117, "epoch": 900, "lr": 8.016557531666085e-05}
{"train_loss": 0.22340230643749237, "global_step": 80118, "epoch": 900, "lr": 8.016511297142937e-05}
{"train_loss": 0.1487531214952469, "global_step": 80119, "epoch": 900, "lr": 8.016465062214256e-05}
{"train_loss": 0.24358312785625458, "global_step": 80120, "epoch": 900, "lr": 8.016418826880045e-05}
{"train_loss": 0.16484637558460236, "global_step": 80121, "epoch": 900, "lr": 8.016372591140316e-05}
{"train_loss": 0.18230192363262177, "global_step": 80122, "epoch": 900, "lr": 8.016326354995071e-05}
{"train_loss": 0.16366685926914215, "global_step": 80123, "epoch": 900, "lr": 8.01628011844432e-05}
{"train_loss": 0.17521516978740692, "global_step": 80124, "epoch": 900, "lr": 8.016233881488065e-05}
{"train_loss": 0.24524450302124023, "global_step": 80125, "epoch": 900, "lr": 8.016187644126315e-05}
{"train_loss": 0.22511687874794006, "global_step": 80126, "epoch": 900, "lr": 8.016141406359074e-05}
{"train_loss": 0.24401620030403137, "global_step": 80127, "epoch": 900, "lr": 8.01609516818635e-05}
{"train_loss": 0.16321223974227905, "global_step": 80128, "epoch": 900, "lr": 8.016048929608148e-05}
{"train_loss": 0.22281424701213837, "global_step": 80129, "epoch": 900, "lr": 8.016002690624477e-05}
{"train_loss": 0.16916270554065704, "global_step": 80130, "epoch": 900, "lr": 8.01595645123534e-05}
{"train_loss": 0.16201899945735931, "global_step": 80131, "epoch": 900, "lr": 8.015910211440744e-05}
{"train_loss": 0.2749450206756592, "global_step": 80132, "epoch": 900, "lr": 8.015863971240698e-05}
{"train_loss": 0.22313931584358215, "global_step": 80133, "epoch": 900, "lr": 8.015817730635204e-05}
{"train_loss": 0.25099238753318787, "global_step": 80134, "epoch": 900, "lr": 8.015771489624271e-05}
{"train_loss": 0.22694645822048187, "global_step": 80135, "epoch": 900, "lr": 8.015725248207902e-05}
{"train_loss": 0.18806657195091248, "global_step": 80136, "epoch": 900, "lr": 8.015679006386108e-05}
{"train_loss": 0.21993139386177063, "global_step": 80137, "epoch": 900, "lr": 8.015632764158892e-05}
{"train_loss": 0.19348889589309692, "global_step": 80138, "epoch": 900, "lr": 8.01558652152626e-05}
{"train_loss": 0.19889956712722778, "global_step": 80139, "epoch": 900, "lr": 8.01554027848822e-05}
{"train_loss": 0.28152650594711304, "global_step": 80140, "epoch": 900, "lr": 8.015494035044777e-05}
{"train_loss": 0.16506023705005646, "global_step": 80141, "epoch": 900, "lr": 8.015447791195937e-05}
{"train_loss": 0.2451019436120987, "global_step": 80142, "epoch": 900, "lr": 8.015401546941708e-05}
{"train_loss": 0.1192539855837822, "global_step": 80143, "epoch": 900, "lr": 8.015355302282095e-05}
{"train_loss": 0.2125868797302246, "global_step": 80144, "epoch": 900, "lr": 8.015309057217104e-05}
{"train_loss": 0.11725295335054398, "global_step": 80145, "epoch": 900, "lr": 8.015262811746741e-05}
{"train_loss": 0.17141258716583252, "global_step": 80146, "epoch": 900, "lr": 8.015216565871013e-05}
{"train_loss": 0.16054734587669373, "global_step": 80147, "epoch": 900, "lr": 8.015170319589927e-05}
{"train_loss": 0.19186320900917053, "global_step": 80148, "epoch": 900, "lr": 8.015124072903486e-05}
{"train_loss": 0.10091409087181091, "global_step": 80149, "epoch": 900, "lr": 8.015077825811699e-05}
{"train_loss": 0.25075095891952515, "global_step": 80150, "epoch": 900, "lr": 8.015031578314571e-05}
{"train_loss": 0.17360280454158783, "global_step": 80151, "epoch": 900, "lr": 8.014985330412111e-05}
{"train_loss": 0.2691637873649597, "global_step": 80152, "epoch": 900, "lr": 8.01493908210432e-05}
{"train_loss": 0.09241977334022522, "global_step": 80153, "epoch": 900, "lr": 8.01489283339121e-05}
{"train_loss": 0.21894051134586334, "global_step": 80154, "epoch": 900, "lr": 8.014846584272782e-05}
{"train_loss": 0.17320123314857483, "global_step": 80155, "epoch": 900, "lr": 8.014800334749046e-05}
{"train_loss": 0.25323817133903503, "global_step": 80156, "epoch": 900, "lr": 8.014754084820006e-05}
{"train_loss": 0.2555482089519501, "global_step": 80157, "epoch": 900, "lr": 8.01470783448567e-05}
{"train_loss": 0.1759449988603592, "global_step": 80158, "epoch": 900, "lr": 8.014661583746043e-05}
{"train_loss": 0.2185564786195755, "global_step": 80159, "epoch": 900, "lr": 8.014615332601131e-05}
{"train_loss": 0.12002253532409668, "global_step": 80160, "epoch": 900, "lr": 8.014569081050941e-05}
{"train_loss": 0.18797731399536133, "global_step": 80161, "epoch": 900, "lr": 8.014522829095478e-05}
{"train_loss": 0.1259564608335495, "global_step": 80162, "epoch": 900, "lr": 8.01447657673475e-05}
{"train_loss": 0.19056160748004913, "global_step": 80163, "epoch": 900, "lr": 8.014430323968764e-05}
{"train_loss": 0.25091031193733215, "global_step": 80164, "epoch": 900, "lr": 8.014384070797522e-05}
{"train_loss": 0.24048638343811035, "global_step": 80165, "epoch": 900, "lr": 8.014337817221037e-05}
{"train_loss": 0.22270525991916656, "global_step": 80166, "epoch": 900, "lr": 8.014291563239307e-05}
{"train_loss": 0.18693633377552032, "global_step": 80167, "epoch": 900, "lr": 8.014245308852341e-05}
{"train_loss": 0.1952618807554245, "global_step": 80168, "epoch": 900, "lr": 8.014199054060149e-05}
{"train_loss": 0.15457873046398163, "global_step": 80169, "epoch": 900, "lr": 8.014152798862735e-05}
{"train_loss": 0.24218334257602692, "global_step": 80170, "epoch": 900, "lr": 8.014106543260105e-05}
{"train_loss": 0.2156100571155548, "global_step": 80171, "epoch": 900, "lr": 8.014060287252263e-05}
{"train_loss": 0.19525668025016785, "global_step": 80172, "epoch": 900, "lr": 8.014014030839219e-05}
{"train_loss": 0.24739831686019897, "global_step": 80173, "epoch": 900, "lr": 8.013967774020978e-05}
{"train_loss": 0.24039818346500397, "global_step": 80174, "epoch": 900, "lr": 8.013921516797545e-05}
{"train_loss": 0.21702179312705994, "global_step": 80175, "epoch": 900, "lr": 8.013875259168926e-05}
{"train_loss": 0.22037722170352936, "global_step": 80176, "epoch": 900, "lr": 8.013829001135129e-05}
{"train_loss": 0.17311570048332214, "global_step": 80177, "epoch": 900, "lr": 8.013782742696159e-05}
{"train_loss": 0.22153343260288239, "global_step": 80178, "epoch": 900, "lr": 8.013736483852023e-05}
{"train_loss": 0.16871649026870728, "global_step": 80179, "epoch": 900, "lr": 8.013690224602727e-05}
{"train_loss": 0.31599336862564087, "global_step": 80180, "epoch": 900, "lr": 8.013643964948279e-05}
{"train_loss": 0.1915910542011261, "global_step": 80181, "epoch": 900, "lr": 8.01359770488868e-05}
{"train_loss": 0.2334020584821701, "global_step": 80182, "epoch": 900, "lr": 8.013551444423941e-05}
{"train_loss": 0.24865183234214783, "global_step": 80183, "epoch": 900, "lr": 8.013505183554068e-05}
{"train_loss": 0.16613370180130005, "global_step": 80184, "epoch": 900, "lr": 8.013458922279065e-05}
{"train_loss": 0.20220881700515747, "global_step": 80185, "epoch": 900, "lr": 8.013412660598938e-05}
{"train_loss": 0.18483951687812805, "global_step": 80186, "epoch": 900, "lr": 8.013366398513695e-05}
{"train_loss": 0.21714730560779572, "global_step": 80187, "epoch": 900, "lr": 8.013320136023342e-05}
{"train_loss": 0.20010120446762342, "global_step": 80188, "epoch": 900, "lr": 8.013273873127885e-05, "train/sim_max_reward_0": 0.6848206145468647, "train/sim_max_reward_1": 0.9252417780896841, "train/sim_max_reward_2": 0.7049098525547249, "train/sim_max_reward_3": 0.3773306909654529, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.4249631602621082, "test/sim_max_reward_4500001": 0.9960344058197893, "test/sim_max_reward_4500002": 0.925054029418974, "test/sim_max_reward_4500003": 0.7166247282140504, "test/sim_max_reward_4500004": 0.8737860793336285, "test/sim_max_reward_4500005": 0.9797773451347971, "test/sim_max_reward_4500006": 0.0, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.4051831901619493, "test/sim_max_reward_4500009": 0.04957530328459842, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9847584474263436, "test/sim_max_reward_4500012": 1.0, "test/sim_max_reward_4500013": 0.9763512267295997, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.017544158064586216, "test/sim_max_reward_4500017": 0.6943098382125666, "test/sim_max_reward_4500018": 0.518711413592045, "test/sim_max_reward_4500019": 0.491597851513101, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9037847011125548, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.5396749557801503, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9682823211335531, "test/sim_max_reward_4500026": 0.4608400642227773, "test/sim_max_reward_4500027": 0.8978977092348577, "test/sim_max_reward_4500028": 0.48013238453033014, "test/sim_max_reward_4500029": 0.2531087317564509, "test/sim_max_reward_4500030": 0.6939289515448528, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6825576691168942, "test/sim_max_reward_4500033": 0.8415770401960214, "test/sim_max_reward_4500034": 0.6606166779911273, "test/sim_max_reward_4500035": 0.9887107956519079, "test/sim_max_reward_4500036": 0.9287997846054978, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 1.0, "test/sim_max_reward_4500039": 0.9936734761450724, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.616427639291841, "test/sim_max_reward_4500042": 0.1318873321702221, "test/sim_max_reward_4500043": 0.9989667999815831, "test/sim_max_reward_4500044": 0.6591671556032229, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 1.0, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9256983075137853, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7820504893594545, "test/mean_score": 0.6736000734950168, "val_loss": 3.893076181411743, "train_action_mse_error": 29.62058448791504}
{"train_loss": 0.2279745191335678, "global_step": 80189, "epoch": 901, "lr": 8.01322760982733e-05}
{"train_loss": 0.20100824534893036, "global_step": 80190, "epoch": 901, "lr": 8.013181346121683e-05}
{"train_loss": 0.191793292760849, "global_step": 80191, "epoch": 901, "lr": 8.013135082010949e-05}
{"train_loss": 0.23984606564044952, "global_step": 80192, "epoch": 901, "lr": 8.013088817495139e-05}
{"train_loss": 0.24972930550575256, "global_step": 80193, "epoch": 901, "lr": 8.013042552574252e-05}
{"train_loss": 0.22333116829395294, "global_step": 80194, "epoch": 901, "lr": 8.012996287248302e-05}
{"train_loss": 0.20656920969486237, "global_step": 80195, "epoch": 901, "lr": 8.012950021517289e-05}
{"train_loss": 0.2781200408935547, "global_step": 80196, "epoch": 901, "lr": 8.012903755381222e-05}
{"train_loss": 0.15729930996894836, "global_step": 80197, "epoch": 901, "lr": 8.012857488840107e-05}
{"train_loss": 0.17313173413276672, "global_step": 80198, "epoch": 901, "lr": 8.012811221893949e-05}
{"train_loss": 0.22955559194087982, "global_step": 80199, "epoch": 901, "lr": 8.012764954542758e-05}
{"train_loss": 0.1822156012058258, "global_step": 80200, "epoch": 901, "lr": 8.012718686786535e-05}
{"train_loss": 0.18056005239486694, "global_step": 80201, "epoch": 901, "lr": 8.012672418625291e-05}
{"train_loss": 0.16767273843288422, "global_step": 80202, "epoch": 901, "lr": 8.012626150059028e-05}
{"train_loss": 0.31461769342422485, "global_step": 80203, "epoch": 901, "lr": 8.012579881087755e-05}
{"train_loss": 0.22291454672813416, "global_step": 80204, "epoch": 901, "lr": 8.012533611711477e-05}
{"train_loss": 0.1847285032272339, "global_step": 80205, "epoch": 901, "lr": 8.0124873419302e-05}
{"train_loss": 0.21352523565292358, "global_step": 80206, "epoch": 901, "lr": 8.012441071743931e-05}
{"train_loss": 0.22647927701473236, "global_step": 80207, "epoch": 901, "lr": 8.012394801152676e-05}
{"train_loss": 0.18201719224452972, "global_step": 80208, "epoch": 901, "lr": 8.012348530156443e-05}
{"train_loss": 0.23376433551311493, "global_step": 80209, "epoch": 901, "lr": 8.012302258755234e-05}
{"train_loss": 0.15293675661087036, "global_step": 80210, "epoch": 901, "lr": 8.01225598694906e-05}
{"train_loss": 0.2530166506767273, "global_step": 80211, "epoch": 901, "lr": 8.012209714737921e-05}
{"train_loss": 0.23846223950386047, "global_step": 80212, "epoch": 901, "lr": 8.012163442121832e-05}
{"train_loss": 0.20400674641132355, "global_step": 80213, "epoch": 901, "lr": 8.012117169100791e-05}
{"train_loss": 0.23053815960884094, "global_step": 80214, "epoch": 901, "lr": 8.012070895674808e-05}
{"train_loss": 0.1622828245162964, "global_step": 80215, "epoch": 901, "lr": 8.01202462184389e-05}
{"train_loss": 0.17191937565803528, "global_step": 80216, "epoch": 901, "lr": 8.011978347608041e-05}
{"train_loss": 0.10328610986471176, "global_step": 80217, "epoch": 901, "lr": 8.01193207296727e-05}
{"train_loss": 0.20046906173229218, "global_step": 80218, "epoch": 901, "lr": 8.011885797921578e-05}
{"train_loss": 0.22695884108543396, "global_step": 80219, "epoch": 901, "lr": 8.011839522470978e-05}
{"train_loss": 0.11141765117645264, "global_step": 80220, "epoch": 901, "lr": 8.011793246615471e-05}
{"train_loss": 0.18899084627628326, "global_step": 80221, "epoch": 901, "lr": 8.011746970355066e-05}
{"train_loss": 0.23155726492404938, "global_step": 80222, "epoch": 901, "lr": 8.011700693689769e-05}
{"train_loss": 0.31557849049568176, "global_step": 80223, "epoch": 901, "lr": 8.011654416619584e-05}
{"train_loss": 0.1589612513780594, "global_step": 80224, "epoch": 901, "lr": 8.01160813914452e-05}
{"train_loss": 0.13615921139717102, "global_step": 80225, "epoch": 901, "lr": 8.011561861264581e-05}
{"train_loss": 0.14563213288784027, "global_step": 80226, "epoch": 901, "lr": 8.011515582979774e-05}
{"train_loss": 0.19172590970993042, "global_step": 80227, "epoch": 901, "lr": 8.011469304290106e-05}
{"train_loss": 0.2131689339876175, "global_step": 80228, "epoch": 901, "lr": 8.011423025195583e-05}
{"train_loss": 0.21884161233901978, "global_step": 80229, "epoch": 901, "lr": 8.011376745696211e-05}
{"train_loss": 0.24331849813461304, "global_step": 80230, "epoch": 901, "lr": 8.011330465791996e-05}
{"train_loss": 0.15430355072021484, "global_step": 80231, "epoch": 901, "lr": 8.011284185482943e-05}
{"train_loss": 0.1024598777294159, "global_step": 80232, "epoch": 901, "lr": 8.01123790476906e-05}
{"train_loss": 0.19577135145664215, "global_step": 80233, "epoch": 901, "lr": 8.011191623650354e-05}
{"train_loss": 0.12783251702785492, "global_step": 80234, "epoch": 901, "lr": 8.011145342126828e-05}
{"train_loss": 0.18385545909404755, "global_step": 80235, "epoch": 901, "lr": 8.011099060198492e-05}
{"train_loss": 0.22913171350955963, "global_step": 80236, "epoch": 901, "lr": 8.01105277786535e-05}
{"train_loss": 0.20398005843162537, "global_step": 80237, "epoch": 901, "lr": 8.011006495127409e-05}
{"train_loss": 0.18699587881565094, "global_step": 80238, "epoch": 901, "lr": 8.010960211984674e-05}
{"train_loss": 0.14621981978416443, "global_step": 80239, "epoch": 901, "lr": 8.010913928437151e-05}
{"train_loss": 0.1308438926935196, "global_step": 80240, "epoch": 901, "lr": 8.01086764448485e-05}
{"train_loss": 0.16274523735046387, "global_step": 80241, "epoch": 901, "lr": 8.010821360127774e-05}
{"train_loss": 0.12260434031486511, "global_step": 80242, "epoch": 901, "lr": 8.010775075365928e-05}
{"train_loss": 0.27474740147590637, "global_step": 80243, "epoch": 901, "lr": 8.010728790199322e-05}
{"train_loss": 0.18619608879089355, "global_step": 80244, "epoch": 901, "lr": 8.01068250462796e-05}
{"train_loss": 0.23528312146663666, "global_step": 80245, "epoch": 901, "lr": 8.010636218651848e-05}
{"train_loss": 0.15981990098953247, "global_step": 80246, "epoch": 901, "lr": 8.010589932270991e-05}
{"train_loss": 0.175661101937294, "global_step": 80247, "epoch": 901, "lr": 8.0105436454854e-05}
{"train_loss": 0.17757362127304077, "global_step": 80248, "epoch": 901, "lr": 8.010497358295075e-05}
{"train_loss": 0.22060807049274445, "global_step": 80249, "epoch": 901, "lr": 8.010451070700027e-05}
{"train_loss": 0.1790732443332672, "global_step": 80250, "epoch": 901, "lr": 8.01040478270026e-05}
{"train_loss": 0.16552650928497314, "global_step": 80251, "epoch": 901, "lr": 8.010358494295782e-05}
{"train_loss": 0.19378256797790527, "global_step": 80252, "epoch": 901, "lr": 8.010312205486596e-05}
{"train_loss": 0.28133147954940796, "global_step": 80253, "epoch": 901, "lr": 8.010265916272711e-05}
{"train_loss": 0.1468038707971573, "global_step": 80254, "epoch": 901, "lr": 8.010219626654134e-05}
{"train_loss": 0.2123444825410843, "global_step": 80255, "epoch": 901, "lr": 8.010173336630868e-05}
{"train_loss": 0.18898512423038483, "global_step": 80256, "epoch": 901, "lr": 8.01012704620292e-05}
{"train_loss": 0.1959959864616394, "global_step": 80257, "epoch": 901, "lr": 8.0100807553703e-05}
{"train_loss": 0.2380395531654358, "global_step": 80258, "epoch": 901, "lr": 8.010034464133008e-05}
{"train_loss": 0.22263425588607788, "global_step": 80259, "epoch": 901, "lr": 8.009988172491057e-05}
{"train_loss": 0.16367827355861664, "global_step": 80260, "epoch": 901, "lr": 8.009941880444447e-05}
{"train_loss": 0.30753710865974426, "global_step": 80261, "epoch": 901, "lr": 8.009895587993188e-05}
{"train_loss": 0.16202466189861298, "global_step": 80262, "epoch": 901, "lr": 8.009849295137285e-05}
{"train_loss": 0.23718979954719543, "global_step": 80263, "epoch": 901, "lr": 8.009803001876745e-05}
{"train_loss": 0.15082238614559174, "global_step": 80264, "epoch": 901, "lr": 8.009756708211574e-05}
{"train_loss": 0.2031608372926712, "global_step": 80265, "epoch": 901, "lr": 8.009710414141777e-05}
{"train_loss": 0.16972824931144714, "global_step": 80266, "epoch": 901, "lr": 8.00966411966736e-05}
{"train_loss": 0.2229786515235901, "global_step": 80267, "epoch": 901, "lr": 8.009617824788332e-05}
{"train_loss": 0.1822054386138916, "global_step": 80268, "epoch": 901, "lr": 8.009571529504697e-05}
{"train_loss": 0.21066485345363617, "global_step": 80269, "epoch": 901, "lr": 8.009525233816462e-05}
{"train_loss": 0.1103607788681984, "global_step": 80270, "epoch": 901, "lr": 8.009478937723633e-05}
{"train_loss": 0.16640020906925201, "global_step": 80271, "epoch": 901, "lr": 8.009432641226216e-05}
{"train_loss": 0.13970011472702026, "global_step": 80272, "epoch": 901, "lr": 8.009386344324219e-05}
{"train_loss": 0.2309003323316574, "global_step": 80273, "epoch": 901, "lr": 8.009340047017645e-05}
{"train_loss": 0.18574993312358856, "global_step": 80274, "epoch": 901, "lr": 8.009293749306502e-05}
{"train_loss": 0.17797893285751343, "global_step": 80275, "epoch": 901, "lr": 8.009247451190796e-05}
{"train_loss": 0.2275409698486328, "global_step": 80276, "epoch": 901, "lr": 8.009201152670535e-05}
{"train_loss": 0.1952480474549733, "global_step": 80277, "epoch": 901, "lr": 8.009154853745721e-05, "val_loss": 4.089827060699463}
{"train_loss": 0.24103295803070068, "global_step": 80278, "epoch": 902, "lr": 8.009108554416365e-05}
{"train_loss": 0.19302864372730255, "global_step": 80279, "epoch": 902, "lr": 8.00906225468247e-05}
{"train_loss": 0.22035342454910278, "global_step": 80280, "epoch": 902, "lr": 8.009015954544044e-05}
{"train_loss": 0.16948720812797546, "global_step": 80281, "epoch": 902, "lr": 8.008969654001093e-05}
{"train_loss": 0.2706330716609955, "global_step": 80282, "epoch": 902, "lr": 8.008923353053622e-05}
{"train_loss": 0.2294498085975647, "global_step": 80283, "epoch": 902, "lr": 8.008877051701637e-05}
{"train_loss": 0.24737325310707092, "global_step": 80284, "epoch": 902, "lr": 8.008830749945145e-05}
{"train_loss": 0.17252209782600403, "global_step": 80285, "epoch": 902, "lr": 8.008784447784152e-05}
{"train_loss": 0.1832488477230072, "global_step": 80286, "epoch": 902, "lr": 8.008738145218666e-05}
{"train_loss": 0.21941174566745758, "global_step": 80287, "epoch": 902, "lr": 8.008691842248693e-05}
{"train_loss": 0.23529179394245148, "global_step": 80288, "epoch": 902, "lr": 8.008645538874236e-05}
{"train_loss": 0.1526763141155243, "global_step": 80289, "epoch": 902, "lr": 8.008599235095304e-05}
{"train_loss": 0.29042941331863403, "global_step": 80290, "epoch": 902, "lr": 8.008552930911904e-05}
{"train_loss": 0.1903652846813202, "global_step": 80291, "epoch": 902, "lr": 8.008506626324037e-05}
{"train_loss": 0.20570166409015656, "global_step": 80292, "epoch": 902, "lr": 8.008460321331716e-05}
{"train_loss": 0.13456562161445618, "global_step": 80293, "epoch": 902, "lr": 8.008414015934943e-05}
{"train_loss": 0.1624656319618225, "global_step": 80294, "epoch": 902, "lr": 8.008367710133726e-05}
{"train_loss": 0.1524333953857422, "global_step": 80295, "epoch": 902, "lr": 8.00832140392807e-05}
{"train_loss": 0.14780592918395996, "global_step": 80296, "epoch": 902, "lr": 8.008275097317984e-05}
{"train_loss": 0.2564961910247803, "global_step": 80297, "epoch": 902, "lr": 8.008228790303469e-05}
{"train_loss": 0.19894352555274963, "global_step": 80298, "epoch": 902, "lr": 8.008182482884536e-05}
{"train_loss": 0.2122592329978943, "global_step": 80299, "epoch": 902, "lr": 8.00813617506119e-05}
{"train_loss": 0.20558896660804749, "global_step": 80300, "epoch": 902, "lr": 8.008089866833435e-05}
{"train_loss": 0.20887990295886993, "global_step": 80301, "epoch": 902, "lr": 8.008043558201279e-05}
{"train_loss": 0.1645592302083969, "global_step": 80302, "epoch": 902, "lr": 8.007997249164729e-05}
{"train_loss": 0.1898464858531952, "global_step": 80303, "epoch": 902, "lr": 8.00795093972379e-05}
{"train_loss": 0.19673964381217957, "global_step": 80304, "epoch": 902, "lr": 8.007904629878469e-05}
{"train_loss": 0.29168736934661865, "global_step": 80305, "epoch": 902, "lr": 8.007858319628772e-05}
{"train_loss": 0.23774105310440063, "global_step": 80306, "epoch": 902, "lr": 8.007812008974705e-05}
{"train_loss": 0.19909344613552094, "global_step": 80307, "epoch": 902, "lr": 8.007765697916275e-05}
{"train_loss": 0.1867244839668274, "global_step": 80308, "epoch": 902, "lr": 8.007719386453487e-05}
{"train_loss": 0.13967402279376984, "global_step": 80309, "epoch": 902, "lr": 8.007673074586349e-05}
{"train_loss": 0.17684301733970642, "global_step": 80310, "epoch": 902, "lr": 8.007626762314864e-05}
{"train_loss": 0.18251143395900726, "global_step": 80311, "epoch": 902, "lr": 8.00758044963904e-05}
{"train_loss": 0.2062719166278839, "global_step": 80312, "epoch": 902, "lr": 8.007534136558885e-05}
{"train_loss": 0.2881850600242615, "global_step": 80313, "epoch": 902, "lr": 8.007487823074403e-05}
{"train_loss": 0.208856999874115, "global_step": 80314, "epoch": 902, "lr": 8.007441509185601e-05}
{"train_loss": 0.13020960986614227, "global_step": 80315, "epoch": 902, "lr": 8.007395194892487e-05}
{"train_loss": 0.2258543223142624, "global_step": 80316, "epoch": 902, "lr": 8.007348880195062e-05}
{"train_loss": 0.16029328107833862, "global_step": 80317, "epoch": 902, "lr": 8.007302565093338e-05}
{"train_loss": 0.1905972808599472, "global_step": 80318, "epoch": 902, "lr": 8.007256249587318e-05}
{"train_loss": 0.2315542995929718, "global_step": 80319, "epoch": 902, "lr": 8.00720993367701e-05}
{"train_loss": 0.2549406588077545, "global_step": 80320, "epoch": 902, "lr": 8.007163617362417e-05}
{"train_loss": 0.15429866313934326, "global_step": 80321, "epoch": 902, "lr": 8.007117300643548e-05}
{"train_loss": 0.2005862295627594, "global_step": 80322, "epoch": 902, "lr": 8.007070983520411e-05}
{"train_loss": 0.19701524078845978, "global_step": 80323, "epoch": 902, "lr": 8.007024665993008e-05}
{"train_loss": 0.20760157704353333, "global_step": 80324, "epoch": 902, "lr": 8.006978348061347e-05}
{"train_loss": 0.2046496421098709, "global_step": 80325, "epoch": 902, "lr": 8.006932029725436e-05}
{"train_loss": 0.21301817893981934, "global_step": 80326, "epoch": 902, "lr": 8.006885710985278e-05}
{"train_loss": 0.17751581966876984, "global_step": 80327, "epoch": 902, "lr": 8.006839391840883e-05}
{"train_loss": 0.16813240945339203, "global_step": 80328, "epoch": 902, "lr": 8.006793072292254e-05}
{"train_loss": 0.19254547357559204, "global_step": 80329, "epoch": 902, "lr": 8.006746752339399e-05}
{"train_loss": 0.22963495552539825, "global_step": 80330, "epoch": 902, "lr": 8.006700431982322e-05}
{"train_loss": 0.2627218961715698, "global_step": 80331, "epoch": 902, "lr": 8.00665411122103e-05}
{"train_loss": 0.2589385509490967, "global_step": 80332, "epoch": 902, "lr": 8.006607790055531e-05}
{"train_loss": 0.16195949912071228, "global_step": 80333, "epoch": 902, "lr": 8.006561468485831e-05}
{"train_loss": 0.18246476352214813, "global_step": 80334, "epoch": 902, "lr": 8.006515146511936e-05}
{"train_loss": 0.1293780356645584, "global_step": 80335, "epoch": 902, "lr": 8.00646882413385e-05}
{"train_loss": 0.27539801597595215, "global_step": 80336, "epoch": 902, "lr": 8.006422501351582e-05}
{"train_loss": 0.28440558910369873, "global_step": 80337, "epoch": 902, "lr": 8.006376178165137e-05}
{"train_loss": 0.23852874338626862, "global_step": 80338, "epoch": 902, "lr": 8.006329854574521e-05}
{"train_loss": 0.1641898900270462, "global_step": 80339, "epoch": 902, "lr": 8.006283530579741e-05}
{"train_loss": 0.2137381136417389, "global_step": 80340, "epoch": 902, "lr": 8.006237206180802e-05}
{"train_loss": 0.16583311557769775, "global_step": 80341, "epoch": 902, "lr": 8.006190881377711e-05}
{"train_loss": 0.25871747732162476, "global_step": 80342, "epoch": 902, "lr": 8.006144556170476e-05}
{"train_loss": 0.2660823464393616, "global_step": 80343, "epoch": 902, "lr": 8.0060982305591e-05}
{"train_loss": 0.11700521409511566, "global_step": 80344, "epoch": 902, "lr": 8.00605190454359e-05}
{"train_loss": 0.1626908928155899, "global_step": 80345, "epoch": 902, "lr": 8.006005578123956e-05}
{"train_loss": 0.18819837272167206, "global_step": 80346, "epoch": 902, "lr": 8.005959251300199e-05}
{"train_loss": 0.30683672428131104, "global_step": 80347, "epoch": 902, "lr": 8.005912924072328e-05}
{"train_loss": 0.19925302267074585, "global_step": 80348, "epoch": 902, "lr": 8.005866596440348e-05}
{"train_loss": 0.17026662826538086, "global_step": 80349, "epoch": 902, "lr": 8.005820268404267e-05}
{"train_loss": 0.1516285389661789, "global_step": 80350, "epoch": 902, "lr": 8.005773939964089e-05}
{"train_loss": 0.20569372177124023, "global_step": 80351, "epoch": 902, "lr": 8.005727611119822e-05}
{"train_loss": 0.2736589312553406, "global_step": 80352, "epoch": 902, "lr": 8.005681281871471e-05}
{"train_loss": 0.24597831070423126, "global_step": 80353, "epoch": 902, "lr": 8.005634952219043e-05}
{"train_loss": 0.1754748672246933, "global_step": 80354, "epoch": 902, "lr": 8.005588622162543e-05}
{"train_loss": 0.23390281200408936, "global_step": 80355, "epoch": 902, "lr": 8.00554229170198e-05}
{"train_loss": 0.18026773631572723, "global_step": 80356, "epoch": 902, "lr": 8.005495960837357e-05}
{"train_loss": 0.20269599556922913, "global_step": 80357, "epoch": 902, "lr": 8.005449629568681e-05}
{"train_loss": 0.21560032665729523, "global_step": 80358, "epoch": 902, "lr": 8.005403297895961e-05}
{"train_loss": 0.25687694549560547, "global_step": 80359, "epoch": 902, "lr": 8.005356965819202e-05}
{"train_loss": 0.21969221532344818, "global_step": 80360, "epoch": 902, "lr": 8.005310633338407e-05}
{"train_loss": 0.19680897891521454, "global_step": 80361, "epoch": 902, "lr": 8.005264300453586e-05}
{"train_loss": 0.20279057323932648, "global_step": 80362, "epoch": 902, "lr": 8.005217967164743e-05}
{"train_loss": 0.2381618469953537, "global_step": 80363, "epoch": 902, "lr": 8.005171633471886e-05}
{"train_loss": 0.18033142387866974, "global_step": 80364, "epoch": 902, "lr": 8.00512529937502e-05}
{"train_loss": 0.2230212539434433, "global_step": 80365, "epoch": 902, "lr": 8.00507896487415e-05}
{"train_loss": 0.20555126466108173, "global_step": 80366, "epoch": 902, "lr": 8.005032629969287e-05, "val_loss": 3.880519390106201}
{"train_loss": 0.27272769808769226, "global_step": 80367, "epoch": 903, "lr": 8.004986294660432e-05}
{"train_loss": 0.21048970520496368, "global_step": 80368, "epoch": 903, "lr": 8.004939958947592e-05}
{"train_loss": 0.19118347764015198, "global_step": 80369, "epoch": 903, "lr": 8.004893622830776e-05}
{"train_loss": 0.16234026849269867, "global_step": 80370, "epoch": 903, "lr": 8.004847286309989e-05}
{"train_loss": 0.2855682969093323, "global_step": 80371, "epoch": 903, "lr": 8.004800949385236e-05}
{"train_loss": 0.18620805442333221, "global_step": 80372, "epoch": 903, "lr": 8.004754612056525e-05}
{"train_loss": 0.1361210197210312, "global_step": 80373, "epoch": 903, "lr": 8.00470827432386e-05}
{"train_loss": 0.12457243353128433, "global_step": 80374, "epoch": 903, "lr": 8.00466193618725e-05}
{"train_loss": 0.10596568137407303, "global_step": 80375, "epoch": 903, "lr": 8.004615597646698e-05}
{"train_loss": 0.228740394115448, "global_step": 80376, "epoch": 903, "lr": 8.004569258702214e-05}
{"train_loss": 0.21970398724079132, "global_step": 80377, "epoch": 903, "lr": 8.0045229193538e-05}
{"train_loss": 0.22208663821220398, "global_step": 80378, "epoch": 903, "lr": 8.004476579601468e-05}
{"train_loss": 0.22077526152133942, "global_step": 80379, "epoch": 903, "lr": 8.004430239445218e-05}
{"train_loss": 0.22724612057209015, "global_step": 80380, "epoch": 903, "lr": 8.00438389888506e-05}
{"train_loss": 0.18933036923408508, "global_step": 80381, "epoch": 903, "lr": 8.004337557920998e-05}
{"train_loss": 0.15190330147743225, "global_step": 80382, "epoch": 903, "lr": 8.004291216553041e-05}
{"train_loss": 0.24065080285072327, "global_step": 80383, "epoch": 903, "lr": 8.004244874781194e-05}
{"train_loss": 0.1872064471244812, "global_step": 80384, "epoch": 903, "lr": 8.004198532605461e-05}
{"train_loss": 0.23197810351848602, "global_step": 80385, "epoch": 903, "lr": 8.004152190025851e-05}
{"train_loss": 0.18334560096263885, "global_step": 80386, "epoch": 903, "lr": 8.004105847042371e-05}
{"train_loss": 0.16643062233924866, "global_step": 80387, "epoch": 903, "lr": 8.004059503655024e-05}
{"train_loss": 0.18069632351398468, "global_step": 80388, "epoch": 903, "lr": 8.004013159863817e-05}
{"train_loss": 0.18074607849121094, "global_step": 80389, "epoch": 903, "lr": 8.003966815668758e-05}
{"train_loss": 0.17649878561496735, "global_step": 80390, "epoch": 903, "lr": 8.003920471069852e-05}
{"train_loss": 0.23042519390583038, "global_step": 80391, "epoch": 903, "lr": 8.003874126067106e-05}
{"train_loss": 0.28222307562828064, "global_step": 80392, "epoch": 903, "lr": 8.003827780660524e-05}
{"train_loss": 0.24015139043331146, "global_step": 80393, "epoch": 903, "lr": 8.003781434850116e-05}
{"train_loss": 0.15307466685771942, "global_step": 80394, "epoch": 903, "lr": 8.003735088635885e-05}
{"train_loss": 0.22200456261634827, "global_step": 80395, "epoch": 903, "lr": 8.003688742017839e-05}
{"train_loss": 0.2619965076446533, "global_step": 80396, "epoch": 903, "lr": 8.003642394995983e-05}
{"train_loss": 0.1780284196138382, "global_step": 80397, "epoch": 903, "lr": 8.003596047570326e-05}
{"train_loss": 0.16234321892261505, "global_step": 80398, "epoch": 903, "lr": 8.003549699740871e-05}
{"train_loss": 0.15660998225212097, "global_step": 80399, "epoch": 903, "lr": 8.003503351507624e-05}
{"train_loss": 0.2577923536300659, "global_step": 80400, "epoch": 903, "lr": 8.003457002870593e-05}
{"train_loss": 0.15389564633369446, "global_step": 80401, "epoch": 903, "lr": 8.003410653829786e-05}
{"train_loss": 0.30361467599868774, "global_step": 80402, "epoch": 903, "lr": 8.003364304385205e-05}
{"train_loss": 0.1365719586610794, "global_step": 80403, "epoch": 903, "lr": 8.003317954536859e-05}
{"train_loss": 0.23735272884368896, "global_step": 80404, "epoch": 903, "lr": 8.003271604284753e-05}
{"train_loss": 0.1881999671459198, "global_step": 80405, "epoch": 903, "lr": 8.003225253628894e-05}
{"train_loss": 0.27689409255981445, "global_step": 80406, "epoch": 903, "lr": 8.003178902569288e-05}
{"train_loss": 0.2491285502910614, "global_step": 80407, "epoch": 903, "lr": 8.003132551105941e-05}
{"train_loss": 0.18997907638549805, "global_step": 80408, "epoch": 903, "lr": 8.003086199238859e-05}
{"train_loss": 0.21903175115585327, "global_step": 80409, "epoch": 903, "lr": 8.00303984696805e-05}
{"train_loss": 0.1674189418554306, "global_step": 80410, "epoch": 903, "lr": 8.002993494293518e-05}
{"train_loss": 0.2317957878112793, "global_step": 80411, "epoch": 903, "lr": 8.002947141215268e-05}
{"train_loss": 0.24351555109024048, "global_step": 80412, "epoch": 903, "lr": 8.002900787733312e-05}
{"train_loss": 0.26148316264152527, "global_step": 80413, "epoch": 903, "lr": 8.00285443384765e-05}
{"train_loss": 0.19215397536754608, "global_step": 80414, "epoch": 903, "lr": 8.002808079558292e-05}
{"train_loss": 0.2553650438785553, "global_step": 80415, "epoch": 903, "lr": 8.002761724865244e-05}
{"train_loss": 0.20370638370513916, "global_step": 80416, "epoch": 903, "lr": 8.002715369768511e-05}
{"train_loss": 0.15537025034427643, "global_step": 80417, "epoch": 903, "lr": 8.002669014268098e-05}
{"train_loss": 0.23453842103481293, "global_step": 80418, "epoch": 903, "lr": 8.002622658364014e-05}
{"train_loss": 0.248984694480896, "global_step": 80419, "epoch": 903, "lr": 8.002576302056263e-05}
{"train_loss": 0.285078763961792, "global_step": 80420, "epoch": 903, "lr": 8.002529945344855e-05}
{"train_loss": 0.15780793130397797, "global_step": 80421, "epoch": 903, "lr": 8.00248358822979e-05}
{"train_loss": 0.19116169214248657, "global_step": 80422, "epoch": 903, "lr": 8.002437230711079e-05}
{"train_loss": 0.15461039543151855, "global_step": 80423, "epoch": 903, "lr": 8.002390872788728e-05}
{"train_loss": 0.2269746959209442, "global_step": 80424, "epoch": 903, "lr": 8.00234451446274e-05}
{"train_loss": 0.11851292103528976, "global_step": 80425, "epoch": 903, "lr": 8.002298155733125e-05}
{"train_loss": 0.1547953486442566, "global_step": 80426, "epoch": 903, "lr": 8.002251796599887e-05}
{"train_loss": 0.14225052297115326, "global_step": 80427, "epoch": 903, "lr": 8.002205437063033e-05}
{"train_loss": 0.26164883375167847, "global_step": 80428, "epoch": 903, "lr": 8.002159077122568e-05}
{"train_loss": 0.16070528328418732, "global_step": 80429, "epoch": 903, "lr": 8.002112716778501e-05}
{"train_loss": 0.2150583416223526, "global_step": 80430, "epoch": 903, "lr": 8.002066356030835e-05}
{"train_loss": 0.2156010866165161, "global_step": 80431, "epoch": 903, "lr": 8.002019994879578e-05}
{"train_loss": 0.13739508390426636, "global_step": 80432, "epoch": 903, "lr": 8.001973633324737e-05}
{"train_loss": 0.1833891123533249, "global_step": 80433, "epoch": 903, "lr": 8.001927271366316e-05}
{"train_loss": 0.18654684722423553, "global_step": 80434, "epoch": 903, "lr": 8.001880909004323e-05}
{"train_loss": 0.18913793563842773, "global_step": 80435, "epoch": 903, "lr": 8.001834546238765e-05}
{"train_loss": 0.17599304020404816, "global_step": 80436, "epoch": 903, "lr": 8.001788183069645e-05}
{"train_loss": 0.28728610277175903, "global_step": 80437, "epoch": 903, "lr": 8.001741819496973e-05}
{"train_loss": 0.20698681473731995, "global_step": 80438, "epoch": 903, "lr": 8.001695455520752e-05}
{"train_loss": 0.2047313004732132, "global_step": 80439, "epoch": 903, "lr": 8.00164909114099e-05}
{"train_loss": 0.13377878069877625, "global_step": 80440, "epoch": 903, "lr": 8.001602726357694e-05}
{"train_loss": 0.21578936278820038, "global_step": 80441, "epoch": 903, "lr": 8.001556361170867e-05}
{"train_loss": 0.24934878945350647, "global_step": 80442, "epoch": 903, "lr": 8.001509995580519e-05}
{"train_loss": 0.18111936748027802, "global_step": 80443, "epoch": 903, "lr": 8.001463629586652e-05}
{"train_loss": 0.2188793569803238, "global_step": 80444, "epoch": 903, "lr": 8.001417263189277e-05}
{"train_loss": 0.21803565323352814, "global_step": 80445, "epoch": 903, "lr": 8.001370896388399e-05}
{"train_loss": 0.25597822666168213, "global_step": 80446, "epoch": 903, "lr": 8.00132452918402e-05}
{"train_loss": 0.22355547547340393, "global_step": 80447, "epoch": 903, "lr": 8.001278161576151e-05}
{"train_loss": 0.17752055823802948, "global_step": 80448, "epoch": 903, "lr": 8.001231793564799e-05}
{"train_loss": 0.22344480454921722, "global_step": 80449, "epoch": 903, "lr": 8.001185425149966e-05}
{"train_loss": 0.2493041753768921, "global_step": 80450, "epoch": 903, "lr": 8.00113905633166e-05}
{"train_loss": 0.19891448318958282, "global_step": 80451, "epoch": 903, "lr": 8.001092687109888e-05}
{"train_loss": 0.14419391751289368, "global_step": 80452, "epoch": 903, "lr": 8.001046317484655e-05}
{"train_loss": 0.18336661159992218, "global_step": 80453, "epoch": 903, "lr": 8.000999947455969e-05}
{"train_loss": 0.21008843183517456, "global_step": 80454, "epoch": 903, "lr": 8.000953577023834e-05}
{"train_loss": 0.20301592207691643, "global_step": 80455, "epoch": 903, "lr": 8.000907206188257e-05, "val_loss": 4.047123908996582}
{"train_loss": 0.2321418672800064, "global_step": 80456, "epoch": 904, "lr": 8.000860834949247e-05}
{"train_loss": 0.20112936198711395, "global_step": 80457, "epoch": 904, "lr": 8.000814463306805e-05}
{"train_loss": 0.13139702379703522, "global_step": 80458, "epoch": 904, "lr": 8.000768091260941e-05}
{"train_loss": 0.15941515564918518, "global_step": 80459, "epoch": 904, "lr": 8.000721718811661e-05}
{"train_loss": 0.12852303683757782, "global_step": 80460, "epoch": 904, "lr": 8.00067534595897e-05}
{"train_loss": 0.17554587125778198, "global_step": 80461, "epoch": 904, "lr": 8.000628972702876e-05}
{"train_loss": 0.2311742901802063, "global_step": 80462, "epoch": 904, "lr": 8.000582599043382e-05}
{"train_loss": 0.19375592470169067, "global_step": 80463, "epoch": 904, "lr": 8.000536224980499e-05}
{"train_loss": 0.1983904391527176, "global_step": 80464, "epoch": 904, "lr": 8.000489850514228e-05}
{"train_loss": 0.13768208026885986, "global_step": 80465, "epoch": 904, "lr": 8.000443475644578e-05}
{"train_loss": 0.19052530825138092, "global_step": 80466, "epoch": 904, "lr": 8.000397100371556e-05}
{"train_loss": 0.20679046213626862, "global_step": 80467, "epoch": 904, "lr": 8.000350724695165e-05}
{"train_loss": 0.17561598122119904, "global_step": 80468, "epoch": 904, "lr": 8.000304348615416e-05}
{"train_loss": 0.15195129811763763, "global_step": 80469, "epoch": 904, "lr": 8.000257972132312e-05}
{"train_loss": 0.07087226212024689, "global_step": 80470, "epoch": 904, "lr": 8.000211595245858e-05}
{"train_loss": 0.17062240839004517, "global_step": 80471, "epoch": 904, "lr": 8.000165217956065e-05}
{"train_loss": 0.1485855132341385, "global_step": 80472, "epoch": 904, "lr": 8.000118840262935e-05}
{"train_loss": 0.18345534801483154, "global_step": 80473, "epoch": 904, "lr": 8.000072462166475e-05}
{"train_loss": 0.1835082322359085, "global_step": 80474, "epoch": 904, "lr": 8.000026083666692e-05}
{"train_loss": 0.15463967621326447, "global_step": 80475, "epoch": 904, "lr": 7.999979704763593e-05}
{"train_loss": 0.2305907905101776, "global_step": 80476, "epoch": 904, "lr": 7.999933325457182e-05}
{"train_loss": 0.25186824798583984, "global_step": 80477, "epoch": 904, "lr": 7.999886945747467e-05}
{"train_loss": 0.18611110746860504, "global_step": 80478, "epoch": 904, "lr": 7.999840565634455e-05}
{"train_loss": 0.14927299320697784, "global_step": 80479, "epoch": 904, "lr": 7.99979418511815e-05}
{"train_loss": 0.24164614081382751, "global_step": 80480, "epoch": 904, "lr": 7.99974780419856e-05}
{"train_loss": 0.21912014484405518, "global_step": 80481, "epoch": 904, "lr": 7.99970142287569e-05}
{"train_loss": 0.1932116001844406, "global_step": 80482, "epoch": 904, "lr": 7.999655041149546e-05}
{"train_loss": 0.17765690386295319, "global_step": 80483, "epoch": 904, "lr": 7.999608659020134e-05}
{"train_loss": 0.2495315670967102, "global_step": 80484, "epoch": 904, "lr": 7.999562276487462e-05}
{"train_loss": 0.2166031450033188, "global_step": 80485, "epoch": 904, "lr": 7.999515893551537e-05}
{"train_loss": 0.17519983649253845, "global_step": 80486, "epoch": 904, "lr": 7.999469510212363e-05}
{"train_loss": 0.14865925908088684, "global_step": 80487, "epoch": 904, "lr": 7.999423126469945e-05}
{"train_loss": 0.2417641282081604, "global_step": 80488, "epoch": 904, "lr": 7.999376742324293e-05}
{"train_loss": 0.1274949014186859, "global_step": 80489, "epoch": 904, "lr": 7.999330357775413e-05}
{"train_loss": 0.23555436730384827, "global_step": 80490, "epoch": 904, "lr": 7.999283972823306e-05}
{"train_loss": 0.2261706292629242, "global_step": 80491, "epoch": 904, "lr": 7.999237587467984e-05}
{"train_loss": 0.16218148171901703, "global_step": 80492, "epoch": 904, "lr": 7.999191201709451e-05}
{"train_loss": 0.2338864505290985, "global_step": 80493, "epoch": 904, "lr": 7.999144815547713e-05}
{"train_loss": 0.18765269219875336, "global_step": 80494, "epoch": 904, "lr": 7.999098428982776e-05}
{"train_loss": 0.19898495078086853, "global_step": 80495, "epoch": 904, "lr": 7.999052042014646e-05}
{"train_loss": 0.1727844476699829, "global_step": 80496, "epoch": 904, "lr": 7.99900565464333e-05}
{"train_loss": 0.11748436093330383, "global_step": 80497, "epoch": 904, "lr": 7.998959266868836e-05}
{"train_loss": 0.24727575480937958, "global_step": 80498, "epoch": 904, "lr": 7.998912878691169e-05}
{"train_loss": 0.14491412043571472, "global_step": 80499, "epoch": 904, "lr": 7.998866490110333e-05}
{"train_loss": 0.19490033388137817, "global_step": 80500, "epoch": 904, "lr": 7.998820101126336e-05}
{"train_loss": 0.1696304827928543, "global_step": 80501, "epoch": 904, "lr": 7.998773711739184e-05}
{"train_loss": 0.20040325820446014, "global_step": 80502, "epoch": 904, "lr": 7.998727321948884e-05}
{"train_loss": 0.14850808680057526, "global_step": 80503, "epoch": 904, "lr": 7.998680931755442e-05}
{"train_loss": 0.18648792803287506, "global_step": 80504, "epoch": 904, "lr": 7.998634541158863e-05}
{"train_loss": 0.23961111903190613, "global_step": 80505, "epoch": 904, "lr": 7.998588150159156e-05}
{"train_loss": 0.18416379392147064, "global_step": 80506, "epoch": 904, "lr": 7.998541758756324e-05}
{"train_loss": 0.19808340072631836, "global_step": 80507, "epoch": 904, "lr": 7.998495366950373e-05}
{"train_loss": 0.16677820682525635, "global_step": 80508, "epoch": 904, "lr": 7.998448974741314e-05}
{"train_loss": 0.1478239893913269, "global_step": 80509, "epoch": 904, "lr": 7.998402582129149e-05}
{"train_loss": 0.2142164707183838, "global_step": 80510, "epoch": 904, "lr": 7.998356189113885e-05}
{"train_loss": 0.12015940994024277, "global_step": 80511, "epoch": 904, "lr": 7.998309795695529e-05}
{"train_loss": 0.16479027271270752, "global_step": 80512, "epoch": 904, "lr": 7.998263401874085e-05}
{"train_loss": 0.20650053024291992, "global_step": 80513, "epoch": 904, "lr": 7.998217007649562e-05}
{"train_loss": 0.3036670684814453, "global_step": 80514, "epoch": 904, "lr": 7.998170613021966e-05}
{"train_loss": 0.25110435485839844, "global_step": 80515, "epoch": 904, "lr": 7.998124217991302e-05}
{"train_loss": 0.3051111400127411, "global_step": 80516, "epoch": 904, "lr": 7.998077822557577e-05}
{"train_loss": 0.18203097581863403, "global_step": 80517, "epoch": 904, "lr": 7.998031426720796e-05}
{"train_loss": 0.2768591046333313, "global_step": 80518, "epoch": 904, "lr": 7.997985030480968e-05}
{"train_loss": 0.27015426754951477, "global_step": 80519, "epoch": 904, "lr": 7.997938633838097e-05}
{"train_loss": 0.20969882607460022, "global_step": 80520, "epoch": 904, "lr": 7.997892236792188e-05}
{"train_loss": 0.1563732773065567, "global_step": 80521, "epoch": 904, "lr": 7.99784583934325e-05}
{"train_loss": 0.17484824359416962, "global_step": 80522, "epoch": 904, "lr": 7.997799441491289e-05}
{"train_loss": 0.13198250532150269, "global_step": 80523, "epoch": 904, "lr": 7.99775304323631e-05}
{"train_loss": 0.2951534390449524, "global_step": 80524, "epoch": 904, "lr": 7.997706644578318e-05}
{"train_loss": 0.1758386492729187, "global_step": 80525, "epoch": 904, "lr": 7.997660245517323e-05}
{"train_loss": 0.21171139180660248, "global_step": 80526, "epoch": 904, "lr": 7.997613846053329e-05}
{"train_loss": 0.24006891250610352, "global_step": 80527, "epoch": 904, "lr": 7.997567446186341e-05}
{"train_loss": 0.17539143562316895, "global_step": 80528, "epoch": 904, "lr": 7.997521045916368e-05}
{"train_loss": 0.12561635673046112, "global_step": 80529, "epoch": 904, "lr": 7.997474645243413e-05}
{"train_loss": 0.19945740699768066, "global_step": 80530, "epoch": 904, "lr": 7.997428244167487e-05}
{"train_loss": 0.24501532316207886, "global_step": 80531, "epoch": 904, "lr": 7.997381842688589e-05}
{"train_loss": 0.22156697511672974, "global_step": 80532, "epoch": 904, "lr": 7.997335440806732e-05}
{"train_loss": 0.21141915023326874, "global_step": 80533, "epoch": 904, "lr": 7.997289038521921e-05}
{"train_loss": 0.2815500497817993, "global_step": 80534, "epoch": 904, "lr": 7.997242635834159e-05}
{"train_loss": 0.21130067110061646, "global_step": 80535, "epoch": 904, "lr": 7.997196232743456e-05}
{"train_loss": 0.23800264298915863, "global_step": 80536, "epoch": 904, "lr": 7.997149829249815e-05}
{"train_loss": 0.14456330239772797, "global_step": 80537, "epoch": 904, "lr": 7.997103425353245e-05}
{"train_loss": 0.14538143575191498, "global_step": 80538, "epoch": 904, "lr": 7.997057021053749e-05}
{"train_loss": 0.16589093208312988, "global_step": 80539, "epoch": 904, "lr": 7.997010616351337e-05}
{"train_loss": 0.16981707513332367, "global_step": 80540, "epoch": 904, "lr": 7.996964211246013e-05}
{"train_loss": 0.1766846776008606, "global_step": 80541, "epoch": 904, "lr": 7.996917805737784e-05}
{"train_loss": 0.1829860806465149, "global_step": 80542, "epoch": 904, "lr": 7.996871399826655e-05}
{"train_loss": 0.2498984932899475, "global_step": 80543, "epoch": 904, "lr": 7.996824993512634e-05}
{"train_loss": 0.19301971567145895, "global_step": 80544, "epoch": 904, "lr": 7.996778586795726e-05, "val_loss": 3.950951099395752}
{"train_loss": 0.1826486736536026, "global_step": 80545, "epoch": 905, "lr": 7.996732179675938e-05}
{"train_loss": 0.177180677652359, "global_step": 80546, "epoch": 905, "lr": 7.996685772153276e-05}
{"train_loss": 0.20319630205631256, "global_step": 80547, "epoch": 905, "lr": 7.996639364227745e-05}
{"train_loss": 0.17552413046360016, "global_step": 80548, "epoch": 905, "lr": 7.996592955899354e-05}
{"train_loss": 0.22058285772800446, "global_step": 80549, "epoch": 905, "lr": 7.996546547168107e-05}
{"train_loss": 0.21588939428329468, "global_step": 80550, "epoch": 905, "lr": 7.99650013803401e-05}
{"train_loss": 0.2695034444332123, "global_step": 80551, "epoch": 905, "lr": 7.996453728497071e-05}
{"train_loss": 0.18393269181251526, "global_step": 80552, "epoch": 905, "lr": 7.996407318557295e-05}
{"train_loss": 0.15466654300689697, "global_step": 80553, "epoch": 905, "lr": 7.996360908214688e-05}
{"train_loss": 0.2079823613166809, "global_step": 80554, "epoch": 905, "lr": 7.996314497469258e-05}
{"train_loss": 0.10929803550243378, "global_step": 80555, "epoch": 905, "lr": 7.996268086321008e-05}
{"train_loss": 0.12641078233718872, "global_step": 80556, "epoch": 905, "lr": 7.996221674769948e-05}
{"train_loss": 0.22906945645809174, "global_step": 80557, "epoch": 905, "lr": 7.996175262816082e-05}
{"train_loss": 0.2931542694568634, "global_step": 80558, "epoch": 905, "lr": 7.996128850459417e-05}
{"train_loss": 0.1875842660665512, "global_step": 80559, "epoch": 905, "lr": 7.99608243769996e-05}
{"train_loss": 0.18193183839321136, "global_step": 80560, "epoch": 905, "lr": 7.996036024537714e-05}
{"train_loss": 0.19752734899520874, "global_step": 80561, "epoch": 905, "lr": 7.99598961097269e-05}
{"train_loss": 0.16994301974773407, "global_step": 80562, "epoch": 905, "lr": 7.99594319700489e-05}
{"train_loss": 0.24311286211013794, "global_step": 80563, "epoch": 905, "lr": 7.99589678263432e-05}
{"train_loss": 0.17568501830101013, "global_step": 80564, "epoch": 905, "lr": 7.995850367860992e-05}
{"train_loss": 0.13834595680236816, "global_step": 80565, "epoch": 905, "lr": 7.995803952684907e-05}
{"train_loss": 0.18803954124450684, "global_step": 80566, "epoch": 905, "lr": 7.995757537106073e-05}
{"train_loss": 0.22557693719863892, "global_step": 80567, "epoch": 905, "lr": 7.995711121124495e-05}
{"train_loss": 0.1526380330324173, "global_step": 80568, "epoch": 905, "lr": 7.99566470474018e-05}
{"train_loss": 0.22553201019763947, "global_step": 80569, "epoch": 905, "lr": 7.995618287953136e-05}
{"train_loss": 0.22170040011405945, "global_step": 80570, "epoch": 905, "lr": 7.995571870763367e-05}
{"train_loss": 0.28002041578292847, "global_step": 80571, "epoch": 905, "lr": 7.99552545317088e-05}
{"train_loss": 0.25678709149360657, "global_step": 80572, "epoch": 905, "lr": 7.995479035175679e-05}
{"train_loss": 0.27866971492767334, "global_step": 80573, "epoch": 905, "lr": 7.995432616777775e-05}
{"train_loss": 0.14256668090820312, "global_step": 80574, "epoch": 905, "lr": 7.995386197977171e-05}
{"train_loss": 0.21702659130096436, "global_step": 80575, "epoch": 905, "lr": 7.995339778773871e-05}
{"train_loss": 0.14285708963871002, "global_step": 80576, "epoch": 905, "lr": 7.995293359167888e-05}
{"train_loss": 0.17163297533988953, "global_step": 80577, "epoch": 905, "lr": 7.995246939159223e-05}
{"train_loss": 0.1818418949842453, "global_step": 80578, "epoch": 905, "lr": 7.995200518747883e-05}
{"train_loss": 0.18652094900608063, "global_step": 80579, "epoch": 905, "lr": 7.995154097933874e-05}
{"train_loss": 0.2313120812177658, "global_step": 80580, "epoch": 905, "lr": 7.995107676717205e-05}
{"train_loss": 0.12288721650838852, "global_step": 80581, "epoch": 905, "lr": 7.99506125509788e-05}
{"train_loss": 0.21567845344543457, "global_step": 80582, "epoch": 905, "lr": 7.995014833075904e-05}
{"train_loss": 0.297932505607605, "global_step": 80583, "epoch": 905, "lr": 7.994968410651287e-05}
{"train_loss": 0.1696864366531372, "global_step": 80584, "epoch": 905, "lr": 7.99492198782403e-05}
{"train_loss": 0.1282993108034134, "global_step": 80585, "epoch": 905, "lr": 7.994875564594146e-05}
{"train_loss": 0.08884704858064651, "global_step": 80586, "epoch": 905, "lr": 7.994829140961635e-05}
{"train_loss": 0.20090001821517944, "global_step": 80587, "epoch": 905, "lr": 7.994782716926506e-05}
{"train_loss": 0.20798847079277039, "global_step": 80588, "epoch": 905, "lr": 7.994736292488765e-05}
{"train_loss": 0.2196112424135208, "global_step": 80589, "epoch": 905, "lr": 7.994689867648419e-05}
{"train_loss": 0.18160995841026306, "global_step": 80590, "epoch": 905, "lr": 7.994643442405472e-05}
{"train_loss": 0.18180933594703674, "global_step": 80591, "epoch": 905, "lr": 7.994597016759933e-05}
{"train_loss": 0.2170790135860443, "global_step": 80592, "epoch": 905, "lr": 7.994550590711807e-05}
{"train_loss": 0.16519032418727875, "global_step": 80593, "epoch": 905, "lr": 7.994504164261098e-05}
{"train_loss": 0.194912388920784, "global_step": 80594, "epoch": 905, "lr": 7.994457737407817e-05}
{"train_loss": 0.23569756746292114, "global_step": 80595, "epoch": 905, "lr": 7.994411310151966e-05}
{"train_loss": 0.14745594561100006, "global_step": 80596, "epoch": 905, "lr": 7.994364882493554e-05}
{"train_loss": 0.19765006005764008, "global_step": 80597, "epoch": 905, "lr": 7.994318454432586e-05}
{"train_loss": 0.2050759494304657, "global_step": 80598, "epoch": 905, "lr": 7.994272025969068e-05}
{"train_loss": 0.1194453090429306, "global_step": 80599, "epoch": 905, "lr": 7.994225597103007e-05}
{"train_loss": 0.25155600905418396, "global_step": 80600, "epoch": 905, "lr": 7.994179167834408e-05}
{"train_loss": 0.1551174372434616, "global_step": 80601, "epoch": 905, "lr": 7.994132738163279e-05}
{"train_loss": 0.18601958453655243, "global_step": 80602, "epoch": 905, "lr": 7.994086308089626e-05}
{"train_loss": 0.17024029791355133, "global_step": 80603, "epoch": 905, "lr": 7.994039877613453e-05}
{"train_loss": 0.1258278638124466, "global_step": 80604, "epoch": 905, "lr": 7.99399344673477e-05}
{"train_loss": 0.2103796899318695, "global_step": 80605, "epoch": 905, "lr": 7.993947015453578e-05}
{"train_loss": 0.10360463708639145, "global_step": 80606, "epoch": 905, "lr": 7.993900583769889e-05}
{"train_loss": 0.14886613190174103, "global_step": 80607, "epoch": 905, "lr": 7.993854151683703e-05}
{"train_loss": 0.13837769627571106, "global_step": 80608, "epoch": 905, "lr": 7.993807719195033e-05}
{"train_loss": 0.21572858095169067, "global_step": 80609, "epoch": 905, "lr": 7.993761286303882e-05}
{"train_loss": 0.30780309438705444, "global_step": 80610, "epoch": 905, "lr": 7.993714853010254e-05}
{"train_loss": 0.19473187625408173, "global_step": 80611, "epoch": 905, "lr": 7.99366841931416e-05}
{"train_loss": 0.14914309978485107, "global_step": 80612, "epoch": 905, "lr": 7.993621985215603e-05}
{"train_loss": 0.2220616638660431, "global_step": 80613, "epoch": 905, "lr": 7.99357555071459e-05}
{"train_loss": 0.20078326761722565, "global_step": 80614, "epoch": 905, "lr": 7.993529115811127e-05}
{"train_loss": 0.21827226877212524, "global_step": 80615, "epoch": 905, "lr": 7.99348268050522e-05}
{"train_loss": 0.21123507618904114, "global_step": 80616, "epoch": 905, "lr": 7.993436244796877e-05}
{"train_loss": 0.23097729682922363, "global_step": 80617, "epoch": 905, "lr": 7.993389808686103e-05}
{"train_loss": 0.15273815393447876, "global_step": 80618, "epoch": 905, "lr": 7.993343372172902e-05}
{"train_loss": 0.19705694913864136, "global_step": 80619, "epoch": 905, "lr": 7.993296935257284e-05}
{"train_loss": 0.2666444182395935, "global_step": 80620, "epoch": 905, "lr": 7.993250497939254e-05}
{"train_loss": 0.19892571866512299, "global_step": 80621, "epoch": 905, "lr": 7.993204060218818e-05}
{"train_loss": 0.2305937260389328, "global_step": 80622, "epoch": 905, "lr": 7.993157622095981e-05}
{"train_loss": 0.2024044245481491, "global_step": 80623, "epoch": 905, "lr": 7.993111183570751e-05}
{"train_loss": 0.13121922314167023, "global_step": 80624, "epoch": 905, "lr": 7.993064744643135e-05}
{"train_loss": 0.0928860530257225, "global_step": 80625, "epoch": 905, "lr": 7.993018305313138e-05}
{"train_loss": 0.19348977506160736, "global_step": 80626, "epoch": 905, "lr": 7.992971865580765e-05}
{"train_loss": 0.17807944118976593, "global_step": 80627, "epoch": 905, "lr": 7.992925425446023e-05}
{"train_loss": 0.13995015621185303, "global_step": 80628, "epoch": 905, "lr": 7.99287898490892e-05}
{"train_loss": 0.18636135756969452, "global_step": 80629, "epoch": 905, "lr": 7.992832543969459e-05}
{"train_loss": 0.19615650177001953, "global_step": 80630, "epoch": 905, "lr": 7.992786102627649e-05}
{"train_loss": 0.21036089956760406, "global_step": 80631, "epoch": 905, "lr": 7.992739660883495e-05}
{"train_loss": 0.11646275222301483, "global_step": 80632, "epoch": 905, "lr": 7.992693218737004e-05}
{"train_loss": 0.19082251894340085, "global_step": 80633, "epoch": 905, "lr": 7.992646776188182e-05, "val_loss": 4.177732467651367, "train_action_mse_error": 20.712989807128906}
{"train_loss": 0.22035224735736847, "global_step": 80634, "epoch": 906, "lr": 7.992600333237036e-05}
{"train_loss": 0.24775341153144836, "global_step": 80635, "epoch": 906, "lr": 7.992553889883569e-05}
{"train_loss": 0.24432532489299774, "global_step": 80636, "epoch": 906, "lr": 7.992507446127793e-05}
{"train_loss": 0.21179738640785217, "global_step": 80637, "epoch": 906, "lr": 7.992461001969708e-05}
{"train_loss": 0.1815401315689087, "global_step": 80638, "epoch": 906, "lr": 7.992414557409324e-05}
{"train_loss": 0.16456355154514313, "global_step": 80639, "epoch": 906, "lr": 7.992368112446647e-05}
{"train_loss": 0.26305538415908813, "global_step": 80640, "epoch": 906, "lr": 7.992321667081681e-05}
{"train_loss": 0.3336447477340698, "global_step": 80641, "epoch": 906, "lr": 7.992275221314436e-05}
{"train_loss": 0.1077364832162857, "global_step": 80642, "epoch": 906, "lr": 7.992228775144914e-05}
{"train_loss": 0.2751866281032562, "global_step": 80643, "epoch": 906, "lr": 7.992182328573125e-05}
{"train_loss": 0.18276597559452057, "global_step": 80644, "epoch": 906, "lr": 7.992135881599074e-05}
{"train_loss": 0.23265565931797028, "global_step": 80645, "epoch": 906, "lr": 7.992089434222764e-05}
{"train_loss": 0.17243333160877228, "global_step": 80646, "epoch": 906, "lr": 7.992042986444206e-05}
{"train_loss": 0.23182958364486694, "global_step": 80647, "epoch": 906, "lr": 7.991996538263403e-05}
{"train_loss": 0.22543463110923767, "global_step": 80648, "epoch": 906, "lr": 7.991950089680364e-05}
{"train_loss": 0.21336223185062408, "global_step": 80649, "epoch": 906, "lr": 7.991903640695093e-05}
{"train_loss": 0.20570465922355652, "global_step": 80650, "epoch": 906, "lr": 7.991857191307597e-05}
{"train_loss": 0.22392764687538147, "global_step": 80651, "epoch": 906, "lr": 7.991810741517882e-05}
{"train_loss": 0.2114589363336563, "global_step": 80652, "epoch": 906, "lr": 7.991764291325956e-05}
{"train_loss": 0.25353482365608215, "global_step": 80653, "epoch": 906, "lr": 7.991717840731823e-05}
{"train_loss": 0.2351437360048294, "global_step": 80654, "epoch": 906, "lr": 7.991671389735489e-05}
{"train_loss": 0.16200768947601318, "global_step": 80655, "epoch": 906, "lr": 7.991624938336963e-05}
{"train_loss": 0.21977564692497253, "global_step": 80656, "epoch": 906, "lr": 7.991578486536249e-05}
{"train_loss": 0.2062649130821228, "global_step": 80657, "epoch": 906, "lr": 7.991532034333352e-05}
{"train_loss": 0.14979973435401917, "global_step": 80658, "epoch": 906, "lr": 7.991485581728281e-05}
{"train_loss": 0.20837543904781342, "global_step": 80659, "epoch": 906, "lr": 7.99143912872104e-05}
{"train_loss": 0.2106752097606659, "global_step": 80660, "epoch": 906, "lr": 7.991392675311639e-05}
{"train_loss": 0.1105627566576004, "global_step": 80661, "epoch": 906, "lr": 7.991346221500082e-05}
{"train_loss": 0.1510976403951645, "global_step": 80662, "epoch": 906, "lr": 7.991299767286373e-05}
{"train_loss": 0.2340487539768219, "global_step": 80663, "epoch": 906, "lr": 7.991253312670521e-05}
{"train_loss": 0.2492261677980423, "global_step": 80664, "epoch": 906, "lr": 7.991206857652531e-05}
{"train_loss": 0.10019710659980774, "global_step": 80665, "epoch": 906, "lr": 7.991160402232409e-05}
{"train_loss": 0.24156463146209717, "global_step": 80666, "epoch": 906, "lr": 7.991113946410163e-05}
{"train_loss": 0.28280094265937805, "global_step": 80667, "epoch": 906, "lr": 7.991067490185799e-05}
{"train_loss": 0.14704398810863495, "global_step": 80668, "epoch": 906, "lr": 7.991021033559321e-05}
{"train_loss": 0.1925019919872284, "global_step": 80669, "epoch": 906, "lr": 7.990974576530737e-05}
{"train_loss": 0.23824642598628998, "global_step": 80670, "epoch": 906, "lr": 7.990928119100053e-05}
{"train_loss": 0.2252250760793686, "global_step": 80671, "epoch": 906, "lr": 7.990881661267275e-05}
{"train_loss": 0.26088425517082214, "global_step": 80672, "epoch": 906, "lr": 7.99083520303241e-05}
{"train_loss": 0.16858744621276855, "global_step": 80673, "epoch": 906, "lr": 7.990788744395463e-05}
{"train_loss": 0.2813383638858795, "global_step": 80674, "epoch": 906, "lr": 7.990742285356441e-05}
{"train_loss": 0.19326408207416534, "global_step": 80675, "epoch": 906, "lr": 7.990695825915351e-05}
{"train_loss": 0.23220103979110718, "global_step": 80676, "epoch": 906, "lr": 7.990649366072198e-05}
{"train_loss": 0.12626785039901733, "global_step": 80677, "epoch": 906, "lr": 7.990602905826989e-05}
{"train_loss": 0.22500239312648773, "global_step": 80678, "epoch": 906, "lr": 7.990556445179729e-05}
{"train_loss": 0.2515724003314972, "global_step": 80679, "epoch": 906, "lr": 7.990509984130425e-05}
{"train_loss": 0.24149495363235474, "global_step": 80680, "epoch": 906, "lr": 7.990463522679083e-05}
{"train_loss": 0.25575610995292664, "global_step": 80681, "epoch": 906, "lr": 7.990417060825712e-05}
{"train_loss": 0.28197023272514343, "global_step": 80682, "epoch": 906, "lr": 7.990370598570314e-05}
{"train_loss": 0.2131078690290451, "global_step": 80683, "epoch": 906, "lr": 7.990324135912897e-05}
{"train_loss": 0.16073867678642273, "global_step": 80684, "epoch": 906, "lr": 7.990277672853468e-05}
{"train_loss": 0.19226673245429993, "global_step": 80685, "epoch": 906, "lr": 7.990231209392032e-05}
{"train_loss": 0.17570984363555908, "global_step": 80686, "epoch": 906, "lr": 7.990184745528597e-05}
{"train_loss": 0.23035398125648499, "global_step": 80687, "epoch": 906, "lr": 7.990138281263167e-05}
{"train_loss": 0.19499903917312622, "global_step": 80688, "epoch": 906, "lr": 7.990091816595751e-05}
{"train_loss": 0.1639469414949417, "global_step": 80689, "epoch": 906, "lr": 7.990045351526351e-05}
{"train_loss": 0.15697349607944489, "global_step": 80690, "epoch": 906, "lr": 7.989998886054977e-05}
{"train_loss": 0.1746884286403656, "global_step": 80691, "epoch": 906, "lr": 7.989952420181634e-05}
{"train_loss": 0.1891091763973236, "global_step": 80692, "epoch": 906, "lr": 7.98990595390633e-05}
{"train_loss": 0.15090985596179962, "global_step": 80693, "epoch": 906, "lr": 7.989859487229068e-05}
{"train_loss": 0.26240670680999756, "global_step": 80694, "epoch": 906, "lr": 7.989813020149855e-05}
{"train_loss": 0.15355907380580902, "global_step": 80695, "epoch": 906, "lr": 7.989766552668701e-05}
{"train_loss": 0.1926247626543045, "global_step": 80696, "epoch": 906, "lr": 7.989720084785606e-05}
{"train_loss": 0.23427361249923706, "global_step": 80697, "epoch": 906, "lr": 7.989673616500581e-05}
{"train_loss": 0.10365643352270126, "global_step": 80698, "epoch": 906, "lr": 7.989627147813633e-05}
{"train_loss": 0.2103830724954605, "global_step": 80699, "epoch": 906, "lr": 7.989580678724761e-05}
{"train_loss": 0.19235335290431976, "global_step": 80700, "epoch": 906, "lr": 7.989534209233981e-05}
{"train_loss": 0.23826077580451965, "global_step": 80701, "epoch": 906, "lr": 7.989487739341292e-05}
{"train_loss": 0.1900491863489151, "global_step": 80702, "epoch": 906, "lr": 7.989441269046704e-05}
{"train_loss": 0.1356721818447113, "global_step": 80703, "epoch": 906, "lr": 7.98939479835022e-05}
{"train_loss": 0.19095417857170105, "global_step": 80704, "epoch": 906, "lr": 7.989348327251851e-05}
{"train_loss": 0.17340481281280518, "global_step": 80705, "epoch": 906, "lr": 7.989301855751599e-05}
{"train_loss": 0.21794722974300385, "global_step": 80706, "epoch": 906, "lr": 7.989255383849472e-05}
{"train_loss": 0.21309146285057068, "global_step": 80707, "epoch": 906, "lr": 7.989208911545479e-05}
{"train_loss": 0.22140298783779144, "global_step": 80708, "epoch": 906, "lr": 7.98916243883962e-05}
{"train_loss": 0.2750908434391022, "global_step": 80709, "epoch": 906, "lr": 7.989115965731903e-05}
{"train_loss": 0.15770140290260315, "global_step": 80710, "epoch": 906, "lr": 7.98906949222234e-05}
{"train_loss": 0.11806860566139221, "global_step": 80711, "epoch": 906, "lr": 7.98902301831093e-05}
{"train_loss": 0.21730937063694, "global_step": 80712, "epoch": 906, "lr": 7.988976543997684e-05}
{"train_loss": 0.21139506995677948, "global_step": 80713, "epoch": 906, "lr": 7.988930069282605e-05}
{"train_loss": 0.18191716074943542, "global_step": 80714, "epoch": 906, "lr": 7.988883594165702e-05}
{"train_loss": 0.1901290863752365, "global_step": 80715, "epoch": 906, "lr": 7.98883711864698e-05}
{"train_loss": 0.2774907052516937, "global_step": 80716, "epoch": 906, "lr": 7.988790642726445e-05}
{"train_loss": 0.15228626132011414, "global_step": 80717, "epoch": 906, "lr": 7.988744166404104e-05}
{"train_loss": 0.18012067675590515, "global_step": 80718, "epoch": 906, "lr": 7.988697689679962e-05}
{"train_loss": 0.1566293090581894, "global_step": 80719, "epoch": 906, "lr": 7.988651212554026e-05}
{"train_loss": 0.2739008963108063, "global_step": 80720, "epoch": 906, "lr": 7.988604735026302e-05}
{"train_loss": 0.3235104978084564, "global_step": 80721, "epoch": 906, "lr": 7.988558257096796e-05}
{"train_loss": 0.2053100559651182, "global_step": 80722, "epoch": 906, "lr": 7.988511778765517e-05, "val_loss": 4.121598720550537}
{"train_loss": 0.233916774392128, "global_step": 80723, "epoch": 907, "lr": 7.988465300032468e-05}
{"train_loss": 0.2415088415145874, "global_step": 80724, "epoch": 907, "lr": 7.988418820897656e-05}
{"train_loss": 0.22621554136276245, "global_step": 80725, "epoch": 907, "lr": 7.988372341361088e-05}
{"train_loss": 0.23826999962329865, "global_step": 80726, "epoch": 907, "lr": 7.98832586142277e-05}
{"train_loss": 0.14574898779392242, "global_step": 80727, "epoch": 907, "lr": 7.988279381082707e-05}
{"train_loss": 0.2577948570251465, "global_step": 80728, "epoch": 907, "lr": 7.988232900340906e-05}
{"train_loss": 0.25828152894973755, "global_step": 80729, "epoch": 907, "lr": 7.988186419197375e-05}
{"train_loss": 0.19422858953475952, "global_step": 80730, "epoch": 907, "lr": 7.988139937652117e-05}
{"train_loss": 0.17720729112625122, "global_step": 80731, "epoch": 907, "lr": 7.988093455705142e-05}
{"train_loss": 0.17259885370731354, "global_step": 80732, "epoch": 907, "lr": 7.988046973356453e-05}
{"train_loss": 0.14477747678756714, "global_step": 80733, "epoch": 907, "lr": 7.988000490606058e-05}
{"train_loss": 0.25189459323883057, "global_step": 80734, "epoch": 907, "lr": 7.987954007453961e-05}
{"train_loss": 0.26764753460884094, "global_step": 80735, "epoch": 907, "lr": 7.987907523900173e-05}
{"train_loss": 0.21617579460144043, "global_step": 80736, "epoch": 907, "lr": 7.987861039944695e-05}
{"train_loss": 0.1699095219373703, "global_step": 80737, "epoch": 907, "lr": 7.987814555587536e-05}
{"train_loss": 0.21403929591178894, "global_step": 80738, "epoch": 907, "lr": 7.987768070828701e-05}
{"train_loss": 0.23200421035289764, "global_step": 80739, "epoch": 907, "lr": 7.987721585668198e-05}
{"train_loss": 0.2283696085214615, "global_step": 80740, "epoch": 907, "lr": 7.987675100106034e-05}
{"train_loss": 0.11192700266838074, "global_step": 80741, "epoch": 907, "lr": 7.987628614142211e-05}
{"train_loss": 0.22582028806209564, "global_step": 80742, "epoch": 907, "lr": 7.987582127776738e-05}
{"train_loss": 0.24705278873443604, "global_step": 80743, "epoch": 907, "lr": 7.987535641009621e-05}
{"train_loss": 0.17140483856201172, "global_step": 80744, "epoch": 907, "lr": 7.987489153840866e-05}
{"train_loss": 0.18013790249824524, "global_step": 80745, "epoch": 907, "lr": 7.987442666270482e-05}
{"train_loss": 0.2016759067773819, "global_step": 80746, "epoch": 907, "lr": 7.98739617829847e-05}
{"train_loss": 0.26071351766586304, "global_step": 80747, "epoch": 907, "lr": 7.98734968992484e-05}
{"train_loss": 0.21806566417217255, "global_step": 80748, "epoch": 907, "lr": 7.987303201149598e-05}
{"train_loss": 0.22678713500499725, "global_step": 80749, "epoch": 907, "lr": 7.987256711972748e-05}
{"train_loss": 0.17427895963191986, "global_step": 80750, "epoch": 907, "lr": 7.987210222394298e-05}
{"train_loss": 0.19716867804527283, "global_step": 80751, "epoch": 907, "lr": 7.987163732414255e-05}
{"train_loss": 0.1313439905643463, "global_step": 80752, "epoch": 907, "lr": 7.987117242032623e-05}
{"train_loss": 0.15660551190376282, "global_step": 80753, "epoch": 907, "lr": 7.98707075124941e-05}
{"train_loss": 0.20472116768360138, "global_step": 80754, "epoch": 907, "lr": 7.987024260064623e-05}
{"train_loss": 0.18256710469722748, "global_step": 80755, "epoch": 907, "lr": 7.986977768478265e-05}
{"train_loss": 0.25615018606185913, "global_step": 80756, "epoch": 907, "lr": 7.986931276490346e-05}
{"train_loss": 0.19213303923606873, "global_step": 80757, "epoch": 907, "lr": 7.986884784100869e-05}
{"train_loss": 0.19214072823524475, "global_step": 80758, "epoch": 907, "lr": 7.986838291309842e-05}
{"train_loss": 0.1821407675743103, "global_step": 80759, "epoch": 907, "lr": 7.986791798117271e-05}
{"train_loss": 0.2091301530599594, "global_step": 80760, "epoch": 907, "lr": 7.986745304523162e-05}
{"train_loss": 0.19743740558624268, "global_step": 80761, "epoch": 907, "lr": 7.986698810527522e-05}
{"train_loss": 0.1695982813835144, "global_step": 80762, "epoch": 907, "lr": 7.986652316130358e-05}
{"train_loss": 0.1567995548248291, "global_step": 80763, "epoch": 907, "lr": 7.986605821331675e-05}
{"train_loss": 0.11772812157869339, "global_step": 80764, "epoch": 907, "lr": 7.986559326131477e-05}
{"train_loss": 0.20402418076992035, "global_step": 80765, "epoch": 907, "lr": 7.986512830529774e-05}
{"train_loss": 0.20904511213302612, "global_step": 80766, "epoch": 907, "lr": 7.98646633452657e-05}
{"train_loss": 0.20856541395187378, "global_step": 80767, "epoch": 907, "lr": 7.986419838121873e-05}
{"train_loss": 0.1739674061536789, "global_step": 80768, "epoch": 907, "lr": 7.986373341315687e-05}
{"train_loss": 0.1343534141778946, "global_step": 80769, "epoch": 907, "lr": 7.986326844108021e-05}
{"train_loss": 0.13019748032093048, "global_step": 80770, "epoch": 907, "lr": 7.986280346498878e-05}
{"train_loss": 0.16962333023548126, "global_step": 80771, "epoch": 907, "lr": 7.986233848488268e-05}
{"train_loss": 0.13498178124427795, "global_step": 80772, "epoch": 907, "lr": 7.986187350076193e-05}
{"train_loss": 0.14870291948318481, "global_step": 80773, "epoch": 907, "lr": 7.986140851262663e-05}
{"train_loss": 0.2363090217113495, "global_step": 80774, "epoch": 907, "lr": 7.986094352047682e-05}
{"train_loss": 0.1481437087059021, "global_step": 80775, "epoch": 907, "lr": 7.986047852431258e-05}
{"train_loss": 0.17398495972156525, "global_step": 80776, "epoch": 907, "lr": 7.986001352413395e-05}
{"train_loss": 0.19867950677871704, "global_step": 80777, "epoch": 907, "lr": 7.985954851994101e-05}
{"train_loss": 0.13858133554458618, "global_step": 80778, "epoch": 907, "lr": 7.98590835117338e-05}
{"train_loss": 0.17286783456802368, "global_step": 80779, "epoch": 907, "lr": 7.985861849951244e-05}
{"train_loss": 0.22900211811065674, "global_step": 80780, "epoch": 907, "lr": 7.985815348327692e-05}
{"train_loss": 0.12047523260116577, "global_step": 80781, "epoch": 907, "lr": 7.985768846302736e-05}
{"train_loss": 0.12239786237478256, "global_step": 80782, "epoch": 907, "lr": 7.985722343876378e-05}
{"train_loss": 0.10436834394931793, "global_step": 80783, "epoch": 907, "lr": 7.985675841048627e-05}
{"train_loss": 0.22787608206272125, "global_step": 80784, "epoch": 907, "lr": 7.985629337819488e-05}
{"train_loss": 0.17073029279708862, "global_step": 80785, "epoch": 907, "lr": 7.985582834188967e-05}
{"train_loss": 0.19875411689281464, "global_step": 80786, "epoch": 907, "lr": 7.985536330157071e-05}
{"train_loss": 0.1454838216304779, "global_step": 80787, "epoch": 907, "lr": 7.985489825723807e-05}
{"train_loss": 0.15717895328998566, "global_step": 80788, "epoch": 907, "lr": 7.985443320889179e-05}
{"train_loss": 0.23735299706459045, "global_step": 80789, "epoch": 907, "lr": 7.985396815653195e-05}
{"train_loss": 0.14266246557235718, "global_step": 80790, "epoch": 907, "lr": 7.98535031001586e-05}
{"train_loss": 0.22219039499759674, "global_step": 80791, "epoch": 907, "lr": 7.985303803977182e-05}
{"train_loss": 0.21954652667045593, "global_step": 80792, "epoch": 907, "lr": 7.985257297537165e-05}
{"train_loss": 0.13866600394248962, "global_step": 80793, "epoch": 907, "lr": 7.985210790695818e-05}
{"train_loss": 0.22023853659629822, "global_step": 80794, "epoch": 907, "lr": 7.985164283453146e-05}
{"train_loss": 0.20745308697223663, "global_step": 80795, "epoch": 907, "lr": 7.985117775809155e-05}
{"train_loss": 0.23690316081047058, "global_step": 80796, "epoch": 907, "lr": 7.98507126776385e-05}
{"train_loss": 0.17892009019851685, "global_step": 80797, "epoch": 907, "lr": 7.98502475931724e-05}
{"train_loss": 0.2621435225009918, "global_step": 80798, "epoch": 907, "lr": 7.984978250469329e-05}
{"train_loss": 0.22712792456150055, "global_step": 80799, "epoch": 907, "lr": 7.984931741220123e-05}
{"train_loss": 0.2318883240222931, "global_step": 80800, "epoch": 907, "lr": 7.98488523156963e-05}
{"train_loss": 0.11327344924211502, "global_step": 80801, "epoch": 907, "lr": 7.984838721517857e-05}
{"train_loss": 0.21905452013015747, "global_step": 80802, "epoch": 907, "lr": 7.984792211064808e-05}
{"train_loss": 0.18969857692718506, "global_step": 80803, "epoch": 907, "lr": 7.98474570021049e-05}
{"train_loss": 0.15484528243541718, "global_step": 80804, "epoch": 907, "lr": 7.98469918895491e-05}
{"train_loss": 0.17380233108997345, "global_step": 80805, "epoch": 907, "lr": 7.984652677298073e-05}
{"train_loss": 0.16329325735569, "global_step": 80806, "epoch": 907, "lr": 7.984606165239985e-05}
{"train_loss": 0.11253829300403595, "global_step": 80807, "epoch": 907, "lr": 7.984559652780653e-05}
{"train_loss": 0.21188956499099731, "global_step": 80808, "epoch": 907, "lr": 7.984513139920085e-05}
{"train_loss": 0.12113207578659058, "global_step": 80809, "epoch": 907, "lr": 7.984466626658284e-05}
{"train_loss": 0.19370846450328827, "global_step": 80810, "epoch": 907, "lr": 7.984420112995257e-05}
{"train_loss": 0.18883405969049152, "global_step": 80811, "epoch": 907, "lr": 7.984373598931015e-05, "val_loss": 4.081393241882324}
{"train_loss": 0.19216381013393402, "global_step": 80812, "epoch": 908, "lr": 7.984327084465557e-05}
{"train_loss": 0.18401899933815002, "global_step": 80813, "epoch": 908, "lr": 7.984280569598892e-05}
{"train_loss": 0.11467689275741577, "global_step": 80814, "epoch": 908, "lr": 7.984234054331029e-05}
{"train_loss": 0.143959641456604, "global_step": 80815, "epoch": 908, "lr": 7.984187538661972e-05}
{"train_loss": 0.19482748210430145, "global_step": 80816, "epoch": 908, "lr": 7.984141022591726e-05}
{"train_loss": 0.17941860854625702, "global_step": 80817, "epoch": 908, "lr": 7.9840945061203e-05}
{"train_loss": 0.17254309356212616, "global_step": 80818, "epoch": 908, "lr": 7.984047989247699e-05}
{"train_loss": 0.17958641052246094, "global_step": 80819, "epoch": 908, "lr": 7.984001471973927e-05}
{"train_loss": 0.2525680363178253, "global_step": 80820, "epoch": 908, "lr": 7.983954954298994e-05}
{"train_loss": 0.14243458211421967, "global_step": 80821, "epoch": 908, "lr": 7.983908436222904e-05}
{"train_loss": 0.17850956320762634, "global_step": 80822, "epoch": 908, "lr": 7.983861917745665e-05}
{"train_loss": 0.16003644466400146, "global_step": 80823, "epoch": 908, "lr": 7.983815398867281e-05}
{"train_loss": 0.18345907330513, "global_step": 80824, "epoch": 908, "lr": 7.983768879587761e-05}
{"train_loss": 0.22482706606388092, "global_step": 80825, "epoch": 908, "lr": 7.983722359907108e-05}
{"train_loss": 0.15019716322422028, "global_step": 80826, "epoch": 908, "lr": 7.983675839825331e-05}
{"train_loss": 0.18625380098819733, "global_step": 80827, "epoch": 908, "lr": 7.983629319342435e-05}
{"train_loss": 0.18982771039009094, "global_step": 80828, "epoch": 908, "lr": 7.983582798458423e-05}
{"train_loss": 0.24866800010204315, "global_step": 80829, "epoch": 908, "lr": 7.983536277173307e-05}
{"train_loss": 0.250903844833374, "global_step": 80830, "epoch": 908, "lr": 7.983489755487093e-05}
{"train_loss": 0.11815428733825684, "global_step": 80831, "epoch": 908, "lr": 7.983443233399783e-05}
{"train_loss": 0.17750780284404755, "global_step": 80832, "epoch": 908, "lr": 7.983396710911387e-05}
{"train_loss": 0.24500654637813568, "global_step": 80833, "epoch": 908, "lr": 7.983350188021907e-05}
{"train_loss": 0.21223501861095428, "global_step": 80834, "epoch": 908, "lr": 7.983303664731354e-05}
{"train_loss": 0.1879366934299469, "global_step": 80835, "epoch": 908, "lr": 7.983257141039733e-05}
{"train_loss": 0.20492781698703766, "global_step": 80836, "epoch": 908, "lr": 7.983210616947048e-05}
{"train_loss": 0.2643378674983978, "global_step": 80837, "epoch": 908, "lr": 7.983164092453307e-05}
{"train_loss": 0.2092137634754181, "global_step": 80838, "epoch": 908, "lr": 7.983117567558515e-05}
{"train_loss": 0.1804685890674591, "global_step": 80839, "epoch": 908, "lr": 7.983071042262681e-05}
{"train_loss": 0.1630997359752655, "global_step": 80840, "epoch": 908, "lr": 7.983024516565809e-05}
{"train_loss": 0.21109163761138916, "global_step": 80841, "epoch": 908, "lr": 7.982977990467904e-05}
{"train_loss": 0.23273758590221405, "global_step": 80842, "epoch": 908, "lr": 7.982931463968976e-05}
{"train_loss": 0.14949752390384674, "global_step": 80843, "epoch": 908, "lr": 7.982884937069029e-05}
{"train_loss": 0.2766318619251251, "global_step": 80844, "epoch": 908, "lr": 7.982838409768069e-05}
{"train_loss": 0.24115869402885437, "global_step": 80845, "epoch": 908, "lr": 7.982791882066103e-05}
{"train_loss": 0.2983011305332184, "global_step": 80846, "epoch": 908, "lr": 7.982745353963134e-05}
{"train_loss": 0.13657769560813904, "global_step": 80847, "epoch": 908, "lr": 7.982698825459174e-05}
{"train_loss": 0.12641766667366028, "global_step": 80848, "epoch": 908, "lr": 7.982652296554227e-05}
{"train_loss": 0.1606072336435318, "global_step": 80849, "epoch": 908, "lr": 7.982605767248299e-05}
{"train_loss": 0.19968517124652863, "global_step": 80850, "epoch": 908, "lr": 7.982559237541396e-05}
{"train_loss": 0.17078636586666107, "global_step": 80851, "epoch": 908, "lr": 7.982512707433523e-05}
{"train_loss": 0.16025105118751526, "global_step": 80852, "epoch": 908, "lr": 7.982466176924688e-05}
{"train_loss": 0.1701897382736206, "global_step": 80853, "epoch": 908, "lr": 7.982419646014896e-05}
{"train_loss": 0.14912930130958557, "global_step": 80854, "epoch": 908, "lr": 7.982373114704155e-05}
{"train_loss": 0.20072907209396362, "global_step": 80855, "epoch": 908, "lr": 7.98232658299247e-05}
{"train_loss": 0.33164942264556885, "global_step": 80856, "epoch": 908, "lr": 7.982280050879847e-05}
{"train_loss": 0.17207296192646027, "global_step": 80857, "epoch": 908, "lr": 7.982233518366294e-05}
{"train_loss": 0.2247220128774643, "global_step": 80858, "epoch": 908, "lr": 7.982186985451815e-05}
{"train_loss": 0.2592407166957855, "global_step": 80859, "epoch": 908, "lr": 7.982140452136418e-05}
{"train_loss": 0.18235436081886292, "global_step": 80860, "epoch": 908, "lr": 7.982093918420108e-05}
{"train_loss": 0.1898934543132782, "global_step": 80861, "epoch": 908, "lr": 7.982047384302892e-05}
{"train_loss": 0.19030483067035675, "global_step": 80862, "epoch": 908, "lr": 7.982000849784776e-05}
{"train_loss": 0.14502976834774017, "global_step": 80863, "epoch": 908, "lr": 7.981954314865767e-05}
{"train_loss": 0.10603295266628265, "global_step": 80864, "epoch": 908, "lr": 7.981907779545869e-05}
{"train_loss": 0.17959733307361603, "global_step": 80865, "epoch": 908, "lr": 7.981861243825091e-05}
{"train_loss": 0.24648959934711456, "global_step": 80866, "epoch": 908, "lr": 7.981814707703439e-05}
{"train_loss": 0.17604050040245056, "global_step": 80867, "epoch": 908, "lr": 7.981768171180917e-05}
{"train_loss": 0.18417033553123474, "global_step": 80868, "epoch": 908, "lr": 7.981721634257534e-05}
{"train_loss": 0.1797073781490326, "global_step": 80869, "epoch": 908, "lr": 7.981675096933293e-05}
{"train_loss": 0.2347065806388855, "global_step": 80870, "epoch": 908, "lr": 7.981628559208203e-05}
{"train_loss": 0.13657017052173615, "global_step": 80871, "epoch": 908, "lr": 7.98158202108227e-05}
{"train_loss": 0.17604990303516388, "global_step": 80872, "epoch": 908, "lr": 7.981535482555497e-05}
{"train_loss": 0.15351609885692596, "global_step": 80873, "epoch": 908, "lr": 7.981488943627896e-05}
{"train_loss": 0.112912118434906, "global_step": 80874, "epoch": 908, "lr": 7.981442404299468e-05}
{"train_loss": 0.18880057334899902, "global_step": 80875, "epoch": 908, "lr": 7.981395864570224e-05}
{"train_loss": 0.14659170806407928, "global_step": 80876, "epoch": 908, "lr": 7.981349324440167e-05}
{"train_loss": 0.14729449152946472, "global_step": 80877, "epoch": 908, "lr": 7.981302783909301e-05}
{"train_loss": 0.1338445246219635, "global_step": 80878, "epoch": 908, "lr": 7.981256242977638e-05}
{"train_loss": 0.1747758835554123, "global_step": 80879, "epoch": 908, "lr": 7.98120970164518e-05}
{"train_loss": 0.18327714502811432, "global_step": 80880, "epoch": 908, "lr": 7.981163159911936e-05}
{"train_loss": 0.20966574549674988, "global_step": 80881, "epoch": 908, "lr": 7.981116617777911e-05}
{"train_loss": 0.28625428676605225, "global_step": 80882, "epoch": 908, "lr": 7.98107007524311e-05}
{"train_loss": 0.14584225416183472, "global_step": 80883, "epoch": 908, "lr": 7.981023532307541e-05}
{"train_loss": 0.16397157311439514, "global_step": 80884, "epoch": 908, "lr": 7.980976988971209e-05}
{"train_loss": 0.20507720112800598, "global_step": 80885, "epoch": 908, "lr": 7.980930445234122e-05}
{"train_loss": 0.19190125167369843, "global_step": 80886, "epoch": 908, "lr": 7.980883901096284e-05}
{"train_loss": 0.1402939110994339, "global_step": 80887, "epoch": 908, "lr": 7.980837356557704e-05}
{"train_loss": 0.22087419033050537, "global_step": 80888, "epoch": 908, "lr": 7.980790811618386e-05}
{"train_loss": 0.190394788980484, "global_step": 80889, "epoch": 908, "lr": 7.980744266278337e-05}
{"train_loss": 0.16105888783931732, "global_step": 80890, "epoch": 908, "lr": 7.980697720537565e-05}
{"train_loss": 0.16454154253005981, "global_step": 80891, "epoch": 908, "lr": 7.980651174396073e-05}
{"train_loss": 0.22855642437934875, "global_step": 80892, "epoch": 908, "lr": 7.980604627853867e-05}
{"train_loss": 0.15530791878700256, "global_step": 80893, "epoch": 908, "lr": 7.980558080910959e-05}
{"train_loss": 0.2295299619436264, "global_step": 80894, "epoch": 908, "lr": 7.980511533567348e-05}
{"train_loss": 0.1529374122619629, "global_step": 80895, "epoch": 908, "lr": 7.980464985823044e-05}
{"train_loss": 0.24289779365062714, "global_step": 80896, "epoch": 908, "lr": 7.980418437678054e-05}
{"train_loss": 0.2752954959869385, "global_step": 80897, "epoch": 908, "lr": 7.980371889132382e-05}
{"train_loss": 0.230120450258255, "global_step": 80898, "epoch": 908, "lr": 7.980325340186036e-05}
{"train_loss": 0.19084645807743073, "global_step": 80899, "epoch": 908, "lr": 7.98027879083902e-05}
{"train_loss": 0.18920094536596468, "global_step": 80900, "epoch": 908, "lr": 7.980232241091344e-05, "val_loss": 4.067673206329346}
{"train_loss": 0.18044380843639374, "global_step": 80901, "epoch": 909, "lr": 7.980185690943012e-05}
{"train_loss": 0.2782427966594696, "global_step": 80902, "epoch": 909, "lr": 7.980139140394028e-05}
{"train_loss": 0.20967647433280945, "global_step": 80903, "epoch": 909, "lr": 7.980092589444403e-05}
{"train_loss": 0.2412651777267456, "global_step": 80904, "epoch": 909, "lr": 7.98004603809414e-05}
{"train_loss": 0.1397431343793869, "global_step": 80905, "epoch": 909, "lr": 7.979999486343245e-05}
{"train_loss": 0.20976585149765015, "global_step": 80906, "epoch": 909, "lr": 7.979952934191728e-05}
{"train_loss": 0.22164128720760345, "global_step": 80907, "epoch": 909, "lr": 7.979906381639592e-05}
{"train_loss": 0.20545066893100739, "global_step": 80908, "epoch": 909, "lr": 7.979859828686844e-05}
{"train_loss": 0.2774156928062439, "global_step": 80909, "epoch": 909, "lr": 7.979813275333489e-05}
{"train_loss": 0.18762828409671783, "global_step": 80910, "epoch": 909, "lr": 7.979766721579533e-05}
{"train_loss": 0.2124331295490265, "global_step": 80911, "epoch": 909, "lr": 7.979720167424987e-05}
{"train_loss": 0.2617129981517792, "global_step": 80912, "epoch": 909, "lr": 7.979673612869852e-05}
{"train_loss": 0.19157437980175018, "global_step": 80913, "epoch": 909, "lr": 7.979627057914136e-05}
{"train_loss": 0.19536055624485016, "global_step": 80914, "epoch": 909, "lr": 7.979580502557846e-05}
{"train_loss": 0.18570734560489655, "global_step": 80915, "epoch": 909, "lr": 7.979533946800989e-05}
{"train_loss": 0.13670466840267181, "global_step": 80916, "epoch": 909, "lr": 7.979487390643569e-05}
{"train_loss": 0.16180995106697083, "global_step": 80917, "epoch": 909, "lr": 7.979440834085592e-05}
{"train_loss": 0.1600039303302765, "global_step": 80918, "epoch": 909, "lr": 7.979394277127066e-05}
{"train_loss": 0.2734808921813965, "global_step": 80919, "epoch": 909, "lr": 7.979347719767997e-05}
{"train_loss": 0.22104743123054504, "global_step": 80920, "epoch": 909, "lr": 7.979301162008392e-05}
{"train_loss": 0.17264796793460846, "global_step": 80921, "epoch": 909, "lr": 7.979254603848256e-05}
{"train_loss": 0.23108088970184326, "global_step": 80922, "epoch": 909, "lr": 7.979208045287595e-05}
{"train_loss": 0.1596612185239792, "global_step": 80923, "epoch": 909, "lr": 7.979161486326417e-05}
{"train_loss": 0.1979672908782959, "global_step": 80924, "epoch": 909, "lr": 7.979114926964724e-05}
{"train_loss": 0.2111046314239502, "global_step": 80925, "epoch": 909, "lr": 7.979068367202528e-05}
{"train_loss": 0.1994011104106903, "global_step": 80926, "epoch": 909, "lr": 7.979021807039832e-05}
{"train_loss": 0.2900453209877014, "global_step": 80927, "epoch": 909, "lr": 7.978975246476643e-05}
{"train_loss": 0.23712338507175446, "global_step": 80928, "epoch": 909, "lr": 7.978928685512966e-05}
{"train_loss": 0.23059329390525818, "global_step": 80929, "epoch": 909, "lr": 7.978882124148809e-05}
{"train_loss": 0.2682720124721527, "global_step": 80930, "epoch": 909, "lr": 7.978835562384178e-05}
{"train_loss": 0.13326986134052277, "global_step": 80931, "epoch": 909, "lr": 7.978789000219077e-05}
{"train_loss": 0.18758225440979004, "global_step": 80932, "epoch": 909, "lr": 7.978742437653517e-05}
{"train_loss": 0.2341785430908203, "global_step": 80933, "epoch": 909, "lr": 7.9786958746875e-05}
{"train_loss": 0.17746378481388092, "global_step": 80934, "epoch": 909, "lr": 7.978649311321034e-05}
{"train_loss": 0.1618904173374176, "global_step": 80935, "epoch": 909, "lr": 7.978602747554125e-05}
{"train_loss": 0.14094005525112152, "global_step": 80936, "epoch": 909, "lr": 7.978556183386778e-05}
{"train_loss": 0.2469128966331482, "global_step": 80937, "epoch": 909, "lr": 7.978509618819002e-05}
{"train_loss": 0.2653474807739258, "global_step": 80938, "epoch": 909, "lr": 7.9784630538508e-05}
{"train_loss": 0.19039596617221832, "global_step": 80939, "epoch": 909, "lr": 7.978416488482181e-05}
{"train_loss": 0.19024312496185303, "global_step": 80940, "epoch": 909, "lr": 7.97836992271315e-05}
{"train_loss": 0.2335827648639679, "global_step": 80941, "epoch": 909, "lr": 7.978323356543714e-05}
{"train_loss": 0.08554188162088394, "global_step": 80942, "epoch": 909, "lr": 7.978276789973878e-05}
{"train_loss": 0.13605549931526184, "global_step": 80943, "epoch": 909, "lr": 7.978230223003649e-05}
{"train_loss": 0.21368779242038727, "global_step": 80944, "epoch": 909, "lr": 7.978183655633034e-05}
{"train_loss": 0.13480548560619354, "global_step": 80945, "epoch": 909, "lr": 7.978137087862036e-05}
{"train_loss": 0.19848202168941498, "global_step": 80946, "epoch": 909, "lr": 7.978090519690667e-05}
{"train_loss": 0.1901555061340332, "global_step": 80947, "epoch": 909, "lr": 7.97804395111893e-05}
{"train_loss": 0.28347066044807434, "global_step": 80948, "epoch": 909, "lr": 7.977997382146828e-05}
{"train_loss": 0.25596538186073303, "global_step": 80949, "epoch": 909, "lr": 7.977950812774374e-05}
{"train_loss": 0.1726359874010086, "global_step": 80950, "epoch": 909, "lr": 7.977904243001569e-05}
{"train_loss": 0.27494174242019653, "global_step": 80951, "epoch": 909, "lr": 7.977857672828421e-05}
{"train_loss": 0.2207564413547516, "global_step": 80952, "epoch": 909, "lr": 7.977811102254937e-05}
{"train_loss": 0.1850871741771698, "global_step": 80953, "epoch": 909, "lr": 7.977764531281122e-05}
{"train_loss": 0.3156275451183319, "global_step": 80954, "epoch": 909, "lr": 7.977717959906983e-05}
{"train_loss": 0.22130842506885529, "global_step": 80955, "epoch": 909, "lr": 7.977671388132527e-05}
{"train_loss": 0.13502264022827148, "global_step": 80956, "epoch": 909, "lr": 7.977624815957758e-05}
{"train_loss": 0.17941135168075562, "global_step": 80957, "epoch": 909, "lr": 7.977578243382685e-05}
{"train_loss": 0.2069491595029831, "global_step": 80958, "epoch": 909, "lr": 7.977531670407312e-05}
{"train_loss": 0.1710609793663025, "global_step": 80959, "epoch": 909, "lr": 7.977485097031647e-05}
{"train_loss": 0.21305401623249054, "global_step": 80960, "epoch": 909, "lr": 7.977438523255695e-05}
{"train_loss": 0.17920568585395813, "global_step": 80961, "epoch": 909, "lr": 7.977391949079462e-05}
{"train_loss": 0.1675703078508377, "global_step": 80962, "epoch": 909, "lr": 7.977345374502956e-05}
{"train_loss": 0.23580440878868103, "global_step": 80963, "epoch": 909, "lr": 7.977298799526181e-05}
{"train_loss": 0.13750244677066803, "global_step": 80964, "epoch": 909, "lr": 7.977252224149146e-05}
{"train_loss": 0.16028030216693878, "global_step": 80965, "epoch": 909, "lr": 7.977205648371855e-05}
{"train_loss": 0.16221883893013, "global_step": 80966, "epoch": 909, "lr": 7.977159072194314e-05}
{"train_loss": 0.21024253964424133, "global_step": 80967, "epoch": 909, "lr": 7.977112495616532e-05}
{"train_loss": 0.2012682408094406, "global_step": 80968, "epoch": 909, "lr": 7.977065918638513e-05}
{"train_loss": 0.29428836703300476, "global_step": 80969, "epoch": 909, "lr": 7.977019341260263e-05}
{"train_loss": 0.23456089198589325, "global_step": 80970, "epoch": 909, "lr": 7.976972763481789e-05}
{"train_loss": 0.22528333961963654, "global_step": 80971, "epoch": 909, "lr": 7.976926185303097e-05}
{"train_loss": 0.18052610754966736, "global_step": 80972, "epoch": 909, "lr": 7.976879606724195e-05}
{"train_loss": 0.2055298089981079, "global_step": 80973, "epoch": 909, "lr": 7.976833027745088e-05}
{"train_loss": 0.1906132996082306, "global_step": 80974, "epoch": 909, "lr": 7.976786448365779e-05}
{"train_loss": 0.2599950134754181, "global_step": 80975, "epoch": 909, "lr": 7.97673986858628e-05}
{"train_loss": 0.17314639687538147, "global_step": 80976, "epoch": 909, "lr": 7.976693288406593e-05}
{"train_loss": 0.17172615230083466, "global_step": 80977, "epoch": 909, "lr": 7.976646707826728e-05}
{"train_loss": 0.19285644590854645, "global_step": 80978, "epoch": 909, "lr": 7.976600126846688e-05}
{"train_loss": 0.2281772941350937, "global_step": 80979, "epoch": 909, "lr": 7.976553545466481e-05}
{"train_loss": 0.22074049711227417, "global_step": 80980, "epoch": 909, "lr": 7.976506963686112e-05}
{"train_loss": 0.23051786422729492, "global_step": 80981, "epoch": 909, "lr": 7.976460381505589e-05}
{"train_loss": 0.16319087147712708, "global_step": 80982, "epoch": 909, "lr": 7.976413798924915e-05}
{"train_loss": 0.26348161697387695, "global_step": 80983, "epoch": 909, "lr": 7.976367215944101e-05}
{"train_loss": 0.2305360734462738, "global_step": 80984, "epoch": 909, "lr": 7.976320632563148e-05}
{"train_loss": 0.14594481885433197, "global_step": 80985, "epoch": 909, "lr": 7.976274048782066e-05}
{"train_loss": 0.16280440986156464, "global_step": 80986, "epoch": 909, "lr": 7.976227464600861e-05}
{"train_loss": 0.21425694227218628, "global_step": 80987, "epoch": 909, "lr": 7.97618088001954e-05}
{"train_loss": 0.16698919236660004, "global_step": 80988, "epoch": 909, "lr": 7.976134295038105e-05}
{"train_loss": 0.20491105644555574, "global_step": 80989, "epoch": 909, "lr": 7.976087709656566e-05, "val_loss": 4.0734429359436035}
{"train_loss": 0.1873156726360321, "global_step": 80990, "epoch": 910, "lr": 7.976041123874928e-05}
{"train_loss": 0.1556018441915512, "global_step": 80991, "epoch": 910, "lr": 7.975994537693196e-05}
{"train_loss": 0.2615535855293274, "global_step": 80992, "epoch": 910, "lr": 7.97594795111138e-05}
{"train_loss": 0.21151293814182281, "global_step": 80993, "epoch": 910, "lr": 7.975901364129483e-05}
{"train_loss": 0.18935464322566986, "global_step": 80994, "epoch": 910, "lr": 7.975854776747513e-05}
{"train_loss": 0.22443045675754547, "global_step": 80995, "epoch": 910, "lr": 7.975808188965477e-05}
{"train_loss": 0.25900188088417053, "global_step": 80996, "epoch": 910, "lr": 7.975761600783377e-05}
{"train_loss": 0.17678247392177582, "global_step": 80997, "epoch": 910, "lr": 7.975715012201224e-05}
{"train_loss": 0.17786085605621338, "global_step": 80998, "epoch": 910, "lr": 7.975668423219022e-05}
{"train_loss": 0.23563583195209503, "global_step": 80999, "epoch": 910, "lr": 7.975621833836777e-05}
{"train_loss": 0.21566052734851837, "global_step": 81000, "epoch": 910, "lr": 7.975575244054497e-05}
{"train_loss": 0.20826265215873718, "global_step": 81001, "epoch": 910, "lr": 7.975528653872187e-05}
{"train_loss": 0.15914587676525116, "global_step": 81002, "epoch": 910, "lr": 7.975482063289852e-05}
{"train_loss": 0.22409901022911072, "global_step": 81003, "epoch": 910, "lr": 7.975435472307502e-05}
{"train_loss": 0.16151343286037445, "global_step": 81004, "epoch": 910, "lr": 7.97538888092514e-05}
{"train_loss": 0.09792931377887726, "global_step": 81005, "epoch": 910, "lr": 7.975342289142773e-05}
{"train_loss": 0.17833641171455383, "global_step": 81006, "epoch": 910, "lr": 7.975295696960407e-05}
{"train_loss": 0.18533313274383545, "global_step": 81007, "epoch": 910, "lr": 7.975249104378048e-05}
{"train_loss": 0.24260260164737701, "global_step": 81008, "epoch": 910, "lr": 7.975202511395705e-05}
{"train_loss": 0.17267347872257233, "global_step": 81009, "epoch": 910, "lr": 7.975155918013381e-05}
{"train_loss": 0.2694191634654999, "global_step": 81010, "epoch": 910, "lr": 7.975109324231083e-05}
{"train_loss": 0.1789701282978058, "global_step": 81011, "epoch": 910, "lr": 7.97506273004882e-05}
{"train_loss": 0.24108053743839264, "global_step": 81012, "epoch": 910, "lr": 7.975016135466596e-05}
{"train_loss": 0.17332477867603302, "global_step": 81013, "epoch": 910, "lr": 7.974969540484415e-05}
{"train_loss": 0.1824410855770111, "global_step": 81014, "epoch": 910, "lr": 7.974922945102288e-05}
{"train_loss": 0.16552206873893738, "global_step": 81015, "epoch": 910, "lr": 7.974876349320217e-05}
{"train_loss": 0.13459068536758423, "global_step": 81016, "epoch": 910, "lr": 7.974829753138211e-05}
{"train_loss": 0.20934151113033295, "global_step": 81017, "epoch": 910, "lr": 7.974783156556275e-05}
{"train_loss": 0.21658845245838165, "global_step": 81018, "epoch": 910, "lr": 7.974736559574416e-05}
{"train_loss": 0.23638898134231567, "global_step": 81019, "epoch": 910, "lr": 7.974689962192641e-05}
{"train_loss": 0.2182149440050125, "global_step": 81020, "epoch": 910, "lr": 7.974643364410953e-05}
{"train_loss": 0.11653199046850204, "global_step": 81021, "epoch": 910, "lr": 7.974596766229361e-05}
{"train_loss": 0.18194538354873657, "global_step": 81022, "epoch": 910, "lr": 7.974550167647872e-05}
{"train_loss": 0.18912768363952637, "global_step": 81023, "epoch": 910, "lr": 7.97450356866649e-05}
{"train_loss": 0.23876100778579712, "global_step": 81024, "epoch": 910, "lr": 7.974456969285222e-05}
{"train_loss": 0.11739541590213776, "global_step": 81025, "epoch": 910, "lr": 7.974410369504077e-05}
{"train_loss": 0.17723090946674347, "global_step": 81026, "epoch": 910, "lr": 7.974363769323055e-05}
{"train_loss": 0.18310429155826569, "global_step": 81027, "epoch": 910, "lr": 7.97431716874217e-05}
{"train_loss": 0.20822292566299438, "global_step": 81028, "epoch": 910, "lr": 7.97427056776142e-05}
{"train_loss": 0.22699500620365143, "global_step": 81029, "epoch": 910, "lr": 7.974223966380818e-05}
{"train_loss": 0.21325154602527618, "global_step": 81030, "epoch": 910, "lr": 7.974177364600367e-05}
{"train_loss": 0.2814442813396454, "global_step": 81031, "epoch": 910, "lr": 7.974130762420075e-05}
{"train_loss": 0.14650213718414307, "global_step": 81032, "epoch": 910, "lr": 7.974084159839947e-05}
{"train_loss": 0.14415420591831207, "global_step": 81033, "epoch": 910, "lr": 7.974037556859991e-05}
{"train_loss": 0.24287131428718567, "global_step": 81034, "epoch": 910, "lr": 7.973990953480209e-05}
{"train_loss": 0.11944379657506943, "global_step": 81035, "epoch": 910, "lr": 7.973944349700612e-05}
{"train_loss": 0.1981898695230484, "global_step": 81036, "epoch": 910, "lr": 7.973897745521204e-05}
{"train_loss": 0.15238189697265625, "global_step": 81037, "epoch": 910, "lr": 7.973851140941993e-05}
{"train_loss": 0.17307396233081818, "global_step": 81038, "epoch": 910, "lr": 7.97380453596298e-05}
{"train_loss": 0.12259819358587265, "global_step": 81039, "epoch": 910, "lr": 7.973757930584181e-05}
{"train_loss": 0.21623849868774414, "global_step": 81040, "epoch": 910, "lr": 7.973711324805592e-05}
{"train_loss": 0.27970483899116516, "global_step": 81041, "epoch": 910, "lr": 7.973664718627225e-05}
{"train_loss": 0.18305084109306335, "global_step": 81042, "epoch": 910, "lr": 7.973618112049086e-05}
{"train_loss": 0.24214237928390503, "global_step": 81043, "epoch": 910, "lr": 7.97357150507118e-05}
{"train_loss": 0.15987376868724823, "global_step": 81044, "epoch": 910, "lr": 7.973524897693513e-05}
{"train_loss": 0.19568724930286407, "global_step": 81045, "epoch": 910, "lr": 7.973478289916093e-05}
{"train_loss": 0.1803867369890213, "global_step": 81046, "epoch": 910, "lr": 7.973431681738921e-05}
{"train_loss": 0.26584747433662415, "global_step": 81047, "epoch": 910, "lr": 7.973385073162011e-05}
{"train_loss": 0.13940536975860596, "global_step": 81048, "epoch": 910, "lr": 7.973338464185365e-05}
{"train_loss": 0.16515468060970306, "global_step": 81049, "epoch": 910, "lr": 7.97329185480899e-05}
{"train_loss": 0.256686806678772, "global_step": 81050, "epoch": 910, "lr": 7.973245245032893e-05}
{"train_loss": 0.1969374269247055, "global_step": 81051, "epoch": 910, "lr": 7.973198634857077e-05}
{"train_loss": 0.14535844326019287, "global_step": 81052, "epoch": 910, "lr": 7.973152024281553e-05}
{"train_loss": 0.14761020243167877, "global_step": 81053, "epoch": 910, "lr": 7.973105413306324e-05}
{"train_loss": 0.2536865770816803, "global_step": 81054, "epoch": 910, "lr": 7.973058801931397e-05}
{"train_loss": 0.17143799364566803, "global_step": 81055, "epoch": 910, "lr": 7.973012190156779e-05}
{"train_loss": 0.3123217225074768, "global_step": 81056, "epoch": 910, "lr": 7.972965577982477e-05}
{"train_loss": 0.2711246609687805, "global_step": 81057, "epoch": 910, "lr": 7.972918965408494e-05}
{"train_loss": 0.1909644901752472, "global_step": 81058, "epoch": 910, "lr": 7.972872352434838e-05}
{"train_loss": 0.23046182096004486, "global_step": 81059, "epoch": 910, "lr": 7.972825739061516e-05}
{"train_loss": 0.2692853510379791, "global_step": 81060, "epoch": 910, "lr": 7.972779125288535e-05}
{"train_loss": 0.23261758685112, "global_step": 81061, "epoch": 910, "lr": 7.972732511115899e-05}
{"train_loss": 0.16781902313232422, "global_step": 81062, "epoch": 910, "lr": 7.972685896543616e-05}
{"train_loss": 0.18079271912574768, "global_step": 81063, "epoch": 910, "lr": 7.972639281571691e-05}
{"train_loss": 0.18628279864788055, "global_step": 81064, "epoch": 910, "lr": 7.97259266620013e-05}
{"train_loss": 0.2615603804588318, "global_step": 81065, "epoch": 910, "lr": 7.972546050428942e-05}
{"train_loss": 0.2359641194343567, "global_step": 81066, "epoch": 910, "lr": 7.972499434258129e-05}
{"train_loss": 0.2329992651939392, "global_step": 81067, "epoch": 910, "lr": 7.972452817687703e-05}
{"train_loss": 0.25275686383247375, "global_step": 81068, "epoch": 910, "lr": 7.972406200717665e-05}
{"train_loss": 0.25259900093078613, "global_step": 81069, "epoch": 910, "lr": 7.972359583348023e-05}
{"train_loss": 0.1931026577949524, "global_step": 81070, "epoch": 910, "lr": 7.972312965578784e-05}
{"train_loss": 0.1322242170572281, "global_step": 81071, "epoch": 910, "lr": 7.972266347409953e-05}
{"train_loss": 0.17368759214878082, "global_step": 81072, "epoch": 910, "lr": 7.972219728841539e-05}
{"train_loss": 0.1864311844110489, "global_step": 81073, "epoch": 910, "lr": 7.972173109873544e-05}
{"train_loss": 0.1614362895488739, "global_step": 81074, "epoch": 910, "lr": 7.972126490505978e-05}
{"train_loss": 0.16446255147457123, "global_step": 81075, "epoch": 910, "lr": 7.972079870738845e-05}
{"train_loss": 0.2609301805496216, "global_step": 81076, "epoch": 910, "lr": 7.972033250572153e-05}
{"train_loss": 0.2083054631948471, "global_step": 81077, "epoch": 910, "lr": 7.971986630005907e-05}
{"train_loss": 0.19978538733185008, "global_step": 81078, "epoch": 910, "lr": 7.971940009040113e-05, "val_loss": 4.142617225646973, "train_action_mse_error": 23.17456817626953}
{"train_loss": 0.30142563581466675, "global_step": 81079, "epoch": 911, "lr": 7.971893387674778e-05}
{"train_loss": 0.2395184189081192, "global_step": 81080, "epoch": 911, "lr": 7.97184676590991e-05}
{"train_loss": 0.2929108142852783, "global_step": 81081, "epoch": 911, "lr": 7.971800143745512e-05}
{"train_loss": 0.25184640288352966, "global_step": 81082, "epoch": 911, "lr": 7.97175352118159e-05}
{"train_loss": 0.13868431746959686, "global_step": 81083, "epoch": 911, "lr": 7.971706898218155e-05}
{"train_loss": 0.2390483021736145, "global_step": 81084, "epoch": 911, "lr": 7.971660274855209e-05}
{"train_loss": 0.19424620270729065, "global_step": 81085, "epoch": 911, "lr": 7.971613651092757e-05}
{"train_loss": 0.23753878474235535, "global_step": 81086, "epoch": 911, "lr": 7.97156702693081e-05}
{"train_loss": 0.20559439063072205, "global_step": 81087, "epoch": 911, "lr": 7.971520402369374e-05}
{"train_loss": 0.29627636075019836, "global_step": 81088, "epoch": 911, "lr": 7.971473777408452e-05}
{"train_loss": 0.24226164817810059, "global_step": 81089, "epoch": 911, "lr": 7.97142715204805e-05}
{"train_loss": 0.22369679808616638, "global_step": 81090, "epoch": 911, "lr": 7.971380526288177e-05}
{"train_loss": 0.18072208762168884, "global_step": 81091, "epoch": 911, "lr": 7.971333900128837e-05}
{"train_loss": 0.1465459018945694, "global_step": 81092, "epoch": 911, "lr": 7.971287273570038e-05}
{"train_loss": 0.18821243941783905, "global_step": 81093, "epoch": 911, "lr": 7.971240646611786e-05}
{"train_loss": 0.16952790319919586, "global_step": 81094, "epoch": 911, "lr": 7.971194019254087e-05}
{"train_loss": 0.13845476508140564, "global_step": 81095, "epoch": 911, "lr": 7.971147391496947e-05}
{"train_loss": 0.3781563639640808, "global_step": 81096, "epoch": 911, "lr": 7.971100763340374e-05}
{"train_loss": 0.24976807832717896, "global_step": 81097, "epoch": 911, "lr": 7.97105413478437e-05}
{"train_loss": 0.19447283446788788, "global_step": 81098, "epoch": 911, "lr": 7.971007505828945e-05}
{"train_loss": 0.1788184493780136, "global_step": 81099, "epoch": 911, "lr": 7.970960876474106e-05}
{"train_loss": 0.12716498970985413, "global_step": 81100, "epoch": 911, "lr": 7.970914246719854e-05}
{"train_loss": 0.1893954873085022, "global_step": 81101, "epoch": 911, "lr": 7.970867616566201e-05}
{"train_loss": 0.21831220388412476, "global_step": 81102, "epoch": 911, "lr": 7.97082098601315e-05}
{"train_loss": 0.16346527636051178, "global_step": 81103, "epoch": 911, "lr": 7.97077435506071e-05}
{"train_loss": 0.23756293952465057, "global_step": 81104, "epoch": 911, "lr": 7.970727723708884e-05}
{"train_loss": 0.16618992388248444, "global_step": 81105, "epoch": 911, "lr": 7.970681091957681e-05}
{"train_loss": 0.18559630215168, "global_step": 81106, "epoch": 911, "lr": 7.970634459807106e-05}
{"train_loss": 0.1932947039604187, "global_step": 81107, "epoch": 911, "lr": 7.970587827257165e-05}
{"train_loss": 0.24438132345676422, "global_step": 81108, "epoch": 911, "lr": 7.970541194307866e-05}
{"train_loss": 0.25237882137298584, "global_step": 81109, "epoch": 911, "lr": 7.970494560959212e-05}
{"train_loss": 0.24354632198810577, "global_step": 81110, "epoch": 911, "lr": 7.97044792721121e-05}
{"train_loss": 0.21821807324886322, "global_step": 81111, "epoch": 911, "lr": 7.970401293063872e-05}
{"train_loss": 0.19311968982219696, "global_step": 81112, "epoch": 911, "lr": 7.970354658517196e-05}
{"train_loss": 0.19856828451156616, "global_step": 81113, "epoch": 911, "lr": 7.970308023571193e-05}
{"train_loss": 0.1591009497642517, "global_step": 81114, "epoch": 911, "lr": 7.970261388225868e-05}
{"train_loss": 0.20388391613960266, "global_step": 81115, "epoch": 911, "lr": 7.970214752481229e-05}
{"train_loss": 0.13988102972507477, "global_step": 81116, "epoch": 911, "lr": 7.97016811633728e-05}
{"train_loss": 0.14929847419261932, "global_step": 81117, "epoch": 911, "lr": 7.970121479794029e-05}
{"train_loss": 0.1843123733997345, "global_step": 81118, "epoch": 911, "lr": 7.970074842851479e-05}
{"train_loss": 0.20854099094867706, "global_step": 81119, "epoch": 911, "lr": 7.970028205509639e-05}
{"train_loss": 0.16621719300746918, "global_step": 81120, "epoch": 911, "lr": 7.969981567768517e-05}
{"train_loss": 0.1842256635427475, "global_step": 81121, "epoch": 911, "lr": 7.969934929628116e-05}
{"train_loss": 0.21966354548931122, "global_step": 81122, "epoch": 911, "lr": 7.969888291088443e-05}
{"train_loss": 0.15310066938400269, "global_step": 81123, "epoch": 911, "lr": 7.969841652149505e-05}
{"train_loss": 0.19215968251228333, "global_step": 81124, "epoch": 911, "lr": 7.969795012811307e-05}
{"train_loss": 0.13747668266296387, "global_step": 81125, "epoch": 911, "lr": 7.969748373073859e-05}
{"train_loss": 0.25560420751571655, "global_step": 81126, "epoch": 911, "lr": 7.969701732937163e-05}
{"train_loss": 0.23149263858795166, "global_step": 81127, "epoch": 911, "lr": 7.969655092401227e-05}
{"train_loss": 0.20429475605487823, "global_step": 81128, "epoch": 911, "lr": 7.969608451466056e-05}
{"train_loss": 0.15304362773895264, "global_step": 81129, "epoch": 911, "lr": 7.969561810131658e-05}
{"train_loss": 0.1745193600654602, "global_step": 81130, "epoch": 911, "lr": 7.96951516839804e-05}
{"train_loss": 0.18874593079090118, "global_step": 81131, "epoch": 911, "lr": 7.969468526265204e-05}
{"train_loss": 0.1819424033164978, "global_step": 81132, "epoch": 911, "lr": 7.969421883733161e-05}
{"train_loss": 0.24817107617855072, "global_step": 81133, "epoch": 911, "lr": 7.969375240801916e-05}
{"train_loss": 0.21149341762065887, "global_step": 81134, "epoch": 911, "lr": 7.969328597471474e-05}
{"train_loss": 0.2381843775510788, "global_step": 81135, "epoch": 911, "lr": 7.969281953741842e-05}
{"train_loss": 0.18207725882530212, "global_step": 81136, "epoch": 911, "lr": 7.969235309613025e-05}
{"train_loss": 0.22526386380195618, "global_step": 81137, "epoch": 911, "lr": 7.969188665085032e-05}
{"train_loss": 0.19814646244049072, "global_step": 81138, "epoch": 911, "lr": 7.969142020157867e-05}
{"train_loss": 0.1651252955198288, "global_step": 81139, "epoch": 911, "lr": 7.96909537483154e-05}
{"train_loss": 0.21049198508262634, "global_step": 81140, "epoch": 911, "lr": 7.969048729106051e-05}
{"train_loss": 0.19568879902362823, "global_step": 81141, "epoch": 911, "lr": 7.969002082981409e-05}
{"train_loss": 0.1861596703529358, "global_step": 81142, "epoch": 911, "lr": 7.968955436457622e-05}
{"train_loss": 0.21846315264701843, "global_step": 81143, "epoch": 911, "lr": 7.968908789534695e-05}
{"train_loss": 0.14214956760406494, "global_step": 81144, "epoch": 911, "lr": 7.968862142212635e-05}
{"train_loss": 0.21474529802799225, "global_step": 81145, "epoch": 911, "lr": 7.968815494491448e-05}
{"train_loss": 0.23145045340061188, "global_step": 81146, "epoch": 911, "lr": 7.96876884637114e-05}
{"train_loss": 0.20933797955513, "global_step": 81147, "epoch": 911, "lr": 7.968722197851716e-05}
{"train_loss": 0.17917364835739136, "global_step": 81148, "epoch": 911, "lr": 7.968675548933184e-05}
{"train_loss": 0.20869030058383942, "global_step": 81149, "epoch": 911, "lr": 7.968628899615551e-05}
{"train_loss": 0.1748599261045456, "global_step": 81150, "epoch": 911, "lr": 7.96858224989882e-05}
{"train_loss": 0.17841576039791107, "global_step": 81151, "epoch": 911, "lr": 7.968535599783e-05}
{"train_loss": 0.2443002611398697, "global_step": 81152, "epoch": 911, "lr": 7.968488949268096e-05}
{"train_loss": 0.1800697296857834, "global_step": 81153, "epoch": 911, "lr": 7.968442298354116e-05}
{"train_loss": 0.2407960742712021, "global_step": 81154, "epoch": 911, "lr": 7.968395647041065e-05}
{"train_loss": 0.17671333253383636, "global_step": 81155, "epoch": 911, "lr": 7.968348995328948e-05}
{"train_loss": 0.17687363922595978, "global_step": 81156, "epoch": 911, "lr": 7.968302343217774e-05}
{"train_loss": 0.2099805325269699, "global_step": 81157, "epoch": 911, "lr": 7.968255690707547e-05}
{"train_loss": 0.2299501746892929, "global_step": 81158, "epoch": 911, "lr": 7.968209037798275e-05}
{"train_loss": 0.22851777076721191, "global_step": 81159, "epoch": 911, "lr": 7.968162384489963e-05}
{"train_loss": 0.15953245759010315, "global_step": 81160, "epoch": 911, "lr": 7.968115730782619e-05}
{"train_loss": 0.21477645635604858, "global_step": 81161, "epoch": 911, "lr": 7.968069076676247e-05}
{"train_loss": 0.2555140256881714, "global_step": 81162, "epoch": 911, "lr": 7.968022422170854e-05}
{"train_loss": 0.16420705616474152, "global_step": 81163, "epoch": 911, "lr": 7.967975767266447e-05}
{"train_loss": 0.20466549694538116, "global_step": 81164, "epoch": 911, "lr": 7.967929111963032e-05}
{"train_loss": 0.3138405978679657, "global_step": 81165, "epoch": 911, "lr": 7.967882456260613e-05}
{"train_loss": 0.1927735060453415, "global_step": 81166, "epoch": 911, "lr": 7.967835800159203e-05}
{"train_loss": 0.205101348543435, "global_step": 81167, "epoch": 911, "lr": 7.967789143658799e-05, "val_loss": 4.081916809082031}
{"train_loss": 0.1282680332660675, "global_step": 81168, "epoch": 912, "lr": 7.967742486759415e-05}
{"train_loss": 0.16932065784931183, "global_step": 81169, "epoch": 912, "lr": 7.967695829461052e-05}
{"train_loss": 0.1544712632894516, "global_step": 81170, "epoch": 912, "lr": 7.96764917176372e-05}
{"train_loss": 0.18392519652843475, "global_step": 81171, "epoch": 912, "lr": 7.967602513667422e-05}
{"train_loss": 0.17000342905521393, "global_step": 81172, "epoch": 912, "lr": 7.967555855172169e-05}
{"train_loss": 0.16411033272743225, "global_step": 81173, "epoch": 912, "lr": 7.967509196277961e-05}
{"train_loss": 0.19595415890216827, "global_step": 81174, "epoch": 912, "lr": 7.967462536984809e-05}
{"train_loss": 0.2064172923564911, "global_step": 81175, "epoch": 912, "lr": 7.967415877292719e-05}
{"train_loss": 0.19388210773468018, "global_step": 81176, "epoch": 912, "lr": 7.967369217201695e-05}
{"train_loss": 0.19314634799957275, "global_step": 81177, "epoch": 912, "lr": 7.967322556711744e-05}
{"train_loss": 0.28498780727386475, "global_step": 81178, "epoch": 912, "lr": 7.967275895822872e-05}
{"train_loss": 0.13468880951404572, "global_step": 81179, "epoch": 912, "lr": 7.967229234535087e-05}
{"train_loss": 0.24397757649421692, "global_step": 81180, "epoch": 912, "lr": 7.967182572848395e-05}
{"train_loss": 0.18589681386947632, "global_step": 81181, "epoch": 912, "lr": 7.9671359107628e-05}
{"train_loss": 0.2687249183654785, "global_step": 81182, "epoch": 912, "lr": 7.96708924827831e-05}
{"train_loss": 0.19629865884780884, "global_step": 81183, "epoch": 912, "lr": 7.967042585394931e-05}
{"train_loss": 0.21392206847667694, "global_step": 81184, "epoch": 912, "lr": 7.96699592211267e-05}
{"train_loss": 0.28869467973709106, "global_step": 81185, "epoch": 912, "lr": 7.966949258431534e-05}
{"train_loss": 0.18535803258419037, "global_step": 81186, "epoch": 912, "lr": 7.966902594351524e-05}
{"train_loss": 0.15898530185222626, "global_step": 81187, "epoch": 912, "lr": 7.966855929872652e-05}
{"train_loss": 0.13808637857437134, "global_step": 81188, "epoch": 912, "lr": 7.966809264994923e-05}
{"train_loss": 0.2194487750530243, "global_step": 81189, "epoch": 912, "lr": 7.966762599718342e-05}
{"train_loss": 0.19343991577625275, "global_step": 81190, "epoch": 912, "lr": 7.966715934042916e-05}
{"train_loss": 0.31562110781669617, "global_step": 81191, "epoch": 912, "lr": 7.966669267968651e-05}
{"train_loss": 0.1691076159477234, "global_step": 81192, "epoch": 912, "lr": 7.966622601495555e-05}
{"train_loss": 0.17648740112781525, "global_step": 81193, "epoch": 912, "lr": 7.96657593462363e-05}
{"train_loss": 0.13009797036647797, "global_step": 81194, "epoch": 912, "lr": 7.966529267352886e-05}
{"train_loss": 0.1802833080291748, "global_step": 81195, "epoch": 912, "lr": 7.96648259968333e-05}
{"train_loss": 0.13988901674747467, "global_step": 81196, "epoch": 912, "lr": 7.966435931614966e-05}
{"train_loss": 0.15317408740520477, "global_step": 81197, "epoch": 912, "lr": 7.9663892631478e-05}
{"train_loss": 0.2633424401283264, "global_step": 81198, "epoch": 912, "lr": 7.96634259428184e-05}
{"train_loss": 0.21278584003448486, "global_step": 81199, "epoch": 912, "lr": 7.96629592501709e-05}
{"train_loss": 0.12245301157236099, "global_step": 81200, "epoch": 912, "lr": 7.966249255353558e-05}
{"train_loss": 0.15776154398918152, "global_step": 81201, "epoch": 912, "lr": 7.966202585291251e-05}
{"train_loss": 0.18790753185749054, "global_step": 81202, "epoch": 912, "lr": 7.966155914830174e-05}
{"train_loss": 0.2525765299797058, "global_step": 81203, "epoch": 912, "lr": 7.966109243970332e-05}
{"train_loss": 0.21742738783359528, "global_step": 81204, "epoch": 912, "lr": 7.966062572711734e-05}
{"train_loss": 0.2007909119129181, "global_step": 81205, "epoch": 912, "lr": 7.966015901054386e-05}
{"train_loss": 0.26256316900253296, "global_step": 81206, "epoch": 912, "lr": 7.965969228998291e-05}
{"train_loss": 0.2102135866880417, "global_step": 81207, "epoch": 912, "lr": 7.96592255654346e-05}
{"train_loss": 0.21954749524593353, "global_step": 81208, "epoch": 912, "lr": 7.965875883689895e-05}
{"train_loss": 0.24958235025405884, "global_step": 81209, "epoch": 912, "lr": 7.965829210437604e-05}
{"train_loss": 0.15424099564552307, "global_step": 81210, "epoch": 912, "lr": 7.965782536786594e-05}
{"train_loss": 0.23328202962875366, "global_step": 81211, "epoch": 912, "lr": 7.965735862736871e-05}
{"train_loss": 0.1615697592496872, "global_step": 81212, "epoch": 912, "lr": 7.965689188288441e-05}
{"train_loss": 0.25506991147994995, "global_step": 81213, "epoch": 912, "lr": 7.96564251344131e-05}
{"train_loss": 0.1762210577726364, "global_step": 81214, "epoch": 912, "lr": 7.965595838195486e-05}
{"train_loss": 0.19488681852817535, "global_step": 81215, "epoch": 912, "lr": 7.965549162550972e-05}
{"train_loss": 0.20960600674152374, "global_step": 81216, "epoch": 912, "lr": 7.965502486507775e-05}
{"train_loss": 0.2592576742172241, "global_step": 81217, "epoch": 912, "lr": 7.965455810065905e-05}
{"train_loss": 0.1714611053466797, "global_step": 81218, "epoch": 912, "lr": 7.965409133225363e-05}
{"train_loss": 0.18051575124263763, "global_step": 81219, "epoch": 912, "lr": 7.965362455986161e-05}
{"train_loss": 0.10686946660280228, "global_step": 81220, "epoch": 912, "lr": 7.9653157783483e-05}
{"train_loss": 0.2055242508649826, "global_step": 81221, "epoch": 912, "lr": 7.965269100311787e-05}
{"train_loss": 0.19038057327270508, "global_step": 81222, "epoch": 912, "lr": 7.965222421876632e-05}
{"train_loss": 0.23090465366840363, "global_step": 81223, "epoch": 912, "lr": 7.965175743042838e-05}
{"train_loss": 0.22946062684059143, "global_step": 81224, "epoch": 912, "lr": 7.965129063810415e-05}
{"train_loss": 0.15526460111141205, "global_step": 81225, "epoch": 912, "lr": 7.965082384179364e-05}
{"train_loss": 0.14082828164100647, "global_step": 81226, "epoch": 912, "lr": 7.965035704149694e-05}
{"train_loss": 0.1908135563135147, "global_step": 81227, "epoch": 912, "lr": 7.964989023721412e-05}
{"train_loss": 0.18806278705596924, "global_step": 81228, "epoch": 912, "lr": 7.964942342894521e-05}
{"train_loss": 0.1853870451450348, "global_step": 81229, "epoch": 912, "lr": 7.964895661669032e-05}
{"train_loss": 0.3816033899784088, "global_step": 81230, "epoch": 912, "lr": 7.964848980044947e-05}
{"train_loss": 0.13721980154514313, "global_step": 81231, "epoch": 912, "lr": 7.964802298022275e-05}
{"train_loss": 0.2459791600704193, "global_step": 81232, "epoch": 912, "lr": 7.964755615601023e-05}
{"train_loss": 0.2230318933725357, "global_step": 81233, "epoch": 912, "lr": 7.964708932781196e-05}
{"train_loss": 0.18613815307617188, "global_step": 81234, "epoch": 912, "lr": 7.964662249562798e-05}
{"train_loss": 0.09069080650806427, "global_step": 81235, "epoch": 912, "lr": 7.964615565945836e-05}
{"train_loss": 0.21877798438072205, "global_step": 81236, "epoch": 912, "lr": 7.96456888193032e-05}
{"train_loss": 0.2538509964942932, "global_step": 81237, "epoch": 912, "lr": 7.964522197516254e-05}
{"train_loss": 0.2709306478500366, "global_step": 81238, "epoch": 912, "lr": 7.964475512703643e-05}
{"train_loss": 0.17580994963645935, "global_step": 81239, "epoch": 912, "lr": 7.964428827492495e-05}
{"train_loss": 0.25550177693367004, "global_step": 81240, "epoch": 912, "lr": 7.964382141882816e-05}
{"train_loss": 0.2354040890932083, "global_step": 81241, "epoch": 912, "lr": 7.96433545587461e-05}
{"train_loss": 0.2200312614440918, "global_step": 81242, "epoch": 912, "lr": 7.964288769467887e-05}
{"train_loss": 0.20583637058734894, "global_step": 81243, "epoch": 912, "lr": 7.964242082662652e-05}
{"train_loss": 0.21365849673748016, "global_step": 81244, "epoch": 912, "lr": 7.96419539545891e-05}
{"train_loss": 0.16696150600910187, "global_step": 81245, "epoch": 912, "lr": 7.964148707856668e-05}
{"train_loss": 0.20093129575252533, "global_step": 81246, "epoch": 912, "lr": 7.964102019855932e-05}
{"train_loss": 0.17455843091011047, "global_step": 81247, "epoch": 912, "lr": 7.964055331456707e-05}
{"train_loss": 0.16324326395988464, "global_step": 81248, "epoch": 912, "lr": 7.964008642659005e-05}
{"train_loss": 0.17860816419124603, "global_step": 81249, "epoch": 912, "lr": 7.963961953462826e-05}
{"train_loss": 0.20781736075878143, "global_step": 81250, "epoch": 912, "lr": 7.963915263868178e-05}
{"train_loss": 0.15747526288032532, "global_step": 81251, "epoch": 912, "lr": 7.963868573875068e-05}
{"train_loss": 0.24285368621349335, "global_step": 81252, "epoch": 912, "lr": 7.963821883483502e-05}
{"train_loss": 0.14767685532569885, "global_step": 81253, "epoch": 912, "lr": 7.963775192693487e-05}
{"train_loss": 0.1655847728252411, "global_step": 81254, "epoch": 912, "lr": 7.963728501505028e-05}
{"train_loss": 0.24489721655845642, "global_step": 81255, "epoch": 912, "lr": 7.963681809918132e-05}
{"train_loss": 0.19840954645965878, "global_step": 81256, "epoch": 912, "lr": 7.963635117932804e-05, "val_loss": 4.024491786956787}
{"train_loss": 0.18157079815864563, "global_step": 81257, "epoch": 913, "lr": 7.963588425549053e-05}
{"train_loss": 0.18272966146469116, "global_step": 81258, "epoch": 913, "lr": 7.963541732766882e-05}
{"train_loss": 0.22344690561294556, "global_step": 81259, "epoch": 913, "lr": 7.9634950395863e-05}
{"train_loss": 0.19980663061141968, "global_step": 81260, "epoch": 913, "lr": 7.963448346007313e-05}
{"train_loss": 0.158179372549057, "global_step": 81261, "epoch": 913, "lr": 7.963401652029925e-05}
{"train_loss": 0.2331620454788208, "global_step": 81262, "epoch": 913, "lr": 7.963354957654145e-05}
{"train_loss": 0.12980681657791138, "global_step": 81263, "epoch": 913, "lr": 7.963308262879979e-05}
{"train_loss": 0.18900378048419952, "global_step": 81264, "epoch": 913, "lr": 7.963261567707429e-05}
{"train_loss": 0.1559276580810547, "global_step": 81265, "epoch": 913, "lr": 7.963214872136506e-05}
{"train_loss": 0.2619471251964569, "global_step": 81266, "epoch": 913, "lr": 7.963168176167215e-05}
{"train_loss": 0.22287477552890778, "global_step": 81267, "epoch": 913, "lr": 7.963121479799562e-05}
{"train_loss": 0.19117915630340576, "global_step": 81268, "epoch": 913, "lr": 7.963074783033553e-05}
{"train_loss": 0.20612181723117828, "global_step": 81269, "epoch": 913, "lr": 7.963028085869196e-05}
{"train_loss": 0.2811163365840912, "global_step": 81270, "epoch": 913, "lr": 7.962981388306497e-05}
{"train_loss": 0.21514137089252472, "global_step": 81271, "epoch": 913, "lr": 7.962934690345459e-05}
{"train_loss": 0.17304693162441254, "global_step": 81272, "epoch": 913, "lr": 7.962887991986092e-05}
{"train_loss": 0.2457122802734375, "global_step": 81273, "epoch": 913, "lr": 7.9628412932284e-05}
{"train_loss": 0.19293254613876343, "global_step": 81274, "epoch": 913, "lr": 7.962794594072389e-05}
{"train_loss": 0.1744508594274521, "global_step": 81275, "epoch": 913, "lr": 7.962747894518068e-05}
{"train_loss": 0.1650535762310028, "global_step": 81276, "epoch": 913, "lr": 7.962701194565441e-05}
{"train_loss": 0.17590464651584625, "global_step": 81277, "epoch": 913, "lr": 7.962654494214516e-05}
{"train_loss": 0.2828412652015686, "global_step": 81278, "epoch": 913, "lr": 7.962607793465298e-05}
{"train_loss": 0.2241668403148651, "global_step": 81279, "epoch": 913, "lr": 7.962561092317792e-05}
{"train_loss": 0.18301013112068176, "global_step": 81280, "epoch": 913, "lr": 7.962514390772006e-05}
{"train_loss": 0.14309851825237274, "global_step": 81281, "epoch": 913, "lr": 7.962467688827947e-05}
{"train_loss": 0.12576141953468323, "global_step": 81282, "epoch": 913, "lr": 7.962420986485619e-05}
{"train_loss": 0.33027905225753784, "global_step": 81283, "epoch": 913, "lr": 7.96237428374503e-05}
{"train_loss": 0.1523951292037964, "global_step": 81284, "epoch": 913, "lr": 7.962327580606189e-05}
{"train_loss": 0.17024017870426178, "global_step": 81285, "epoch": 913, "lr": 7.962280877069095e-05}
{"train_loss": 0.2603113055229187, "global_step": 81286, "epoch": 913, "lr": 7.962234173133761e-05}
{"train_loss": 0.18148493766784668, "global_step": 81287, "epoch": 913, "lr": 7.96218746880019e-05}
{"train_loss": 0.19210094213485718, "global_step": 81288, "epoch": 913, "lr": 7.962140764068389e-05}
{"train_loss": 0.22667540609836578, "global_step": 81289, "epoch": 913, "lr": 7.962094058938364e-05}
{"train_loss": 0.15665723383426666, "global_step": 81290, "epoch": 913, "lr": 7.962047353410123e-05}
{"train_loss": 0.27802228927612305, "global_step": 81291, "epoch": 913, "lr": 7.962000647483669e-05}
{"train_loss": 0.15353181958198547, "global_step": 81292, "epoch": 913, "lr": 7.96195394115901e-05}
{"train_loss": 0.10493012517690659, "global_step": 81293, "epoch": 913, "lr": 7.961907234436152e-05}
{"train_loss": 0.24138115346431732, "global_step": 81294, "epoch": 913, "lr": 7.961860527315104e-05}
{"train_loss": 0.20392701029777527, "global_step": 81295, "epoch": 913, "lr": 7.961813819795867e-05}
{"train_loss": 0.1950027197599411, "global_step": 81296, "epoch": 913, "lr": 7.961767111878454e-05}
{"train_loss": 0.22449469566345215, "global_step": 81297, "epoch": 913, "lr": 7.961720403562865e-05}
{"train_loss": 0.1977311372756958, "global_step": 81298, "epoch": 913, "lr": 7.961673694849109e-05}
{"train_loss": 0.20429746806621552, "global_step": 81299, "epoch": 913, "lr": 7.961626985737191e-05}
{"train_loss": 0.25203126668930054, "global_step": 81300, "epoch": 913, "lr": 7.961580276227121e-05}
{"train_loss": 0.21059128642082214, "global_step": 81301, "epoch": 913, "lr": 7.961533566318899e-05}
{"train_loss": 0.13087575137615204, "global_step": 81302, "epoch": 913, "lr": 7.961486856012539e-05}
{"train_loss": 0.15554212033748627, "global_step": 81303, "epoch": 913, "lr": 7.961440145308041e-05}
{"train_loss": 0.24510861933231354, "global_step": 81304, "epoch": 913, "lr": 7.961393434205414e-05}
{"train_loss": 0.10616477578878403, "global_step": 81305, "epoch": 913, "lr": 7.961346722704663e-05}
{"train_loss": 0.13231003284454346, "global_step": 81306, "epoch": 913, "lr": 7.961300010805796e-05}
{"train_loss": 0.19086892902851105, "global_step": 81307, "epoch": 913, "lr": 7.961253298508818e-05}
{"train_loss": 0.124383345246315, "global_step": 81308, "epoch": 913, "lr": 7.961206585813735e-05}
{"train_loss": 0.19813774526119232, "global_step": 81309, "epoch": 913, "lr": 7.961159872720554e-05}
{"train_loss": 0.20365358889102936, "global_step": 81310, "epoch": 913, "lr": 7.961113159229282e-05}
{"train_loss": 0.2365557700395584, "global_step": 81311, "epoch": 913, "lr": 7.961066445339924e-05}
{"train_loss": 0.09182647615671158, "global_step": 81312, "epoch": 913, "lr": 7.961019731052486e-05}
{"train_loss": 0.15027886629104614, "global_step": 81313, "epoch": 913, "lr": 7.960973016366976e-05}
{"train_loss": 0.11878655105829239, "global_step": 81314, "epoch": 913, "lr": 7.9609263012834e-05}
{"train_loss": 0.1700262725353241, "global_step": 81315, "epoch": 913, "lr": 7.960879585801762e-05}
{"train_loss": 0.20576445758342743, "global_step": 81316, "epoch": 913, "lr": 7.96083286992207e-05}
{"train_loss": 0.21935635805130005, "global_step": 81317, "epoch": 913, "lr": 7.960786153644331e-05}
{"train_loss": 0.14563901722431183, "global_step": 81318, "epoch": 913, "lr": 7.960739436968548e-05}
{"train_loss": 0.2072860598564148, "global_step": 81319, "epoch": 913, "lr": 7.960692719894731e-05}
{"train_loss": 0.18364693224430084, "global_step": 81320, "epoch": 913, "lr": 7.960646002422886e-05}
{"train_loss": 0.18937554955482483, "global_step": 81321, "epoch": 913, "lr": 7.960599284553018e-05}
{"train_loss": 0.14418822526931763, "global_step": 81322, "epoch": 913, "lr": 7.960552566285134e-05}
{"train_loss": 0.22055473923683167, "global_step": 81323, "epoch": 913, "lr": 7.960505847619237e-05}
{"train_loss": 0.16619658470153809, "global_step": 81324, "epoch": 913, "lr": 7.960459128555338e-05}
{"train_loss": 0.17864757776260376, "global_step": 81325, "epoch": 913, "lr": 7.960412409093442e-05}
{"train_loss": 0.21761927008628845, "global_step": 81326, "epoch": 913, "lr": 7.960365689233553e-05}
{"train_loss": 0.28574371337890625, "global_step": 81327, "epoch": 913, "lr": 7.96031896897568e-05}
{"train_loss": 0.1435815840959549, "global_step": 81328, "epoch": 913, "lr": 7.960272248319828e-05}
{"train_loss": 0.16209837794303894, "global_step": 81329, "epoch": 913, "lr": 7.960225527266004e-05}
{"train_loss": 0.18733695149421692, "global_step": 81330, "epoch": 913, "lr": 7.960178805814214e-05}
{"train_loss": 0.2423679083585739, "global_step": 81331, "epoch": 913, "lr": 7.960132083964463e-05}
{"train_loss": 0.31952646374702454, "global_step": 81332, "epoch": 913, "lr": 7.960085361716758e-05}
{"train_loss": 0.11937128007411957, "global_step": 81333, "epoch": 913, "lr": 7.960038639071106e-05}
{"train_loss": 0.19061602652072906, "global_step": 81334, "epoch": 913, "lr": 7.959991916027514e-05}
{"train_loss": 0.15909862518310547, "global_step": 81335, "epoch": 913, "lr": 7.959945192585987e-05}
{"train_loss": 0.13962896168231964, "global_step": 81336, "epoch": 913, "lr": 7.95989846874653e-05}
{"train_loss": 0.15853539109230042, "global_step": 81337, "epoch": 913, "lr": 7.95985174450915e-05}
{"train_loss": 0.23857681453227997, "global_step": 81338, "epoch": 913, "lr": 7.959805019873856e-05}
{"train_loss": 0.21888001263141632, "global_step": 81339, "epoch": 913, "lr": 7.959758294840652e-05}
{"train_loss": 0.220465287566185, "global_step": 81340, "epoch": 913, "lr": 7.959711569409545e-05}
{"train_loss": 0.2884748578071594, "global_step": 81341, "epoch": 913, "lr": 7.95966484358054e-05}
{"train_loss": 0.1882975548505783, "global_step": 81342, "epoch": 913, "lr": 7.959618117353645e-05}
{"train_loss": 0.16650478541851044, "global_step": 81343, "epoch": 913, "lr": 7.959571390728864e-05}
{"train_loss": 0.18861253559589386, "global_step": 81344, "epoch": 913, "lr": 7.959524663706206e-05}
{"train_loss": 0.19332000529498197, "global_step": 81345, "epoch": 913, "lr": 7.959477936285676e-05, "val_loss": 4.101680755615234}
{"train_loss": 0.1525624692440033, "global_step": 81346, "epoch": 914, "lr": 7.959431208467279e-05}
{"train_loss": 0.24847069382667542, "global_step": 81347, "epoch": 914, "lr": 7.959384480251024e-05}
{"train_loss": 0.18743467330932617, "global_step": 81348, "epoch": 914, "lr": 7.959337751636914e-05}
{"train_loss": 0.12792056798934937, "global_step": 81349, "epoch": 914, "lr": 7.959291022624959e-05}
{"train_loss": 0.21852989494800568, "global_step": 81350, "epoch": 914, "lr": 7.959244293215163e-05}
{"train_loss": 0.15855447947978973, "global_step": 81351, "epoch": 914, "lr": 7.959197563407531e-05}
{"train_loss": 0.219551220536232, "global_step": 81352, "epoch": 914, "lr": 7.959150833202071e-05}
{"train_loss": 0.21193799376487732, "global_step": 81353, "epoch": 914, "lr": 7.959104102598791e-05}
{"train_loss": 0.20908136665821075, "global_step": 81354, "epoch": 914, "lr": 7.959057371597696e-05}
{"train_loss": 0.20640651881694794, "global_step": 81355, "epoch": 914, "lr": 7.959010640198789e-05}
{"train_loss": 0.18983148038387299, "global_step": 81356, "epoch": 914, "lr": 7.958963908402081e-05}
{"train_loss": 0.22631409764289856, "global_step": 81357, "epoch": 914, "lr": 7.958917176207577e-05}
{"train_loss": 0.19306978583335876, "global_step": 81358, "epoch": 914, "lr": 7.958870443615281e-05}
{"train_loss": 0.2271677404642105, "global_step": 81359, "epoch": 914, "lr": 7.958823710625203e-05}
{"train_loss": 0.14386072754859924, "global_step": 81360, "epoch": 914, "lr": 7.958776977237346e-05}
{"train_loss": 0.1384233832359314, "global_step": 81361, "epoch": 914, "lr": 7.958730243451718e-05}
{"train_loss": 0.13021261990070343, "global_step": 81362, "epoch": 914, "lr": 7.958683509268323e-05}
{"train_loss": 0.1853620857000351, "global_step": 81363, "epoch": 914, "lr": 7.95863677468717e-05}
{"train_loss": 0.17799729108810425, "global_step": 81364, "epoch": 914, "lr": 7.958590039708265e-05}
{"train_loss": 0.1253262758255005, "global_step": 81365, "epoch": 914, "lr": 7.958543304331614e-05}
{"train_loss": 0.17966556549072266, "global_step": 81366, "epoch": 914, "lr": 7.958496568557223e-05}
{"train_loss": 0.27908584475517273, "global_step": 81367, "epoch": 914, "lr": 7.958449832385096e-05}
{"train_loss": 0.1979546844959259, "global_step": 81368, "epoch": 914, "lr": 7.958403095815244e-05}
{"train_loss": 0.24415642023086548, "global_step": 81369, "epoch": 914, "lr": 7.95835635884767e-05}
{"train_loss": 0.20664775371551514, "global_step": 81370, "epoch": 914, "lr": 7.95830962148238e-05}
{"train_loss": 0.20590917766094208, "global_step": 81371, "epoch": 914, "lr": 7.958262883719382e-05}
{"train_loss": 0.1738296002149582, "global_step": 81372, "epoch": 914, "lr": 7.958216145558681e-05}
{"train_loss": 0.21250542998313904, "global_step": 81373, "epoch": 914, "lr": 7.958169407000284e-05}
{"train_loss": 0.20470118522644043, "global_step": 81374, "epoch": 914, "lr": 7.958122668044199e-05}
{"train_loss": 0.17540127038955688, "global_step": 81375, "epoch": 914, "lr": 7.958075928690429e-05}
{"train_loss": 0.16811257600784302, "global_step": 81376, "epoch": 914, "lr": 7.958029188938981e-05}
{"train_loss": 0.19906269013881683, "global_step": 81377, "epoch": 914, "lr": 7.957982448789863e-05}
{"train_loss": 0.15305261313915253, "global_step": 81378, "epoch": 914, "lr": 7.957935708243081e-05}
{"train_loss": 0.22732651233673096, "global_step": 81379, "epoch": 914, "lr": 7.95788896729864e-05}
{"train_loss": 0.22420398890972137, "global_step": 81380, "epoch": 914, "lr": 7.957842225956546e-05}
{"train_loss": 0.239764004945755, "global_step": 81381, "epoch": 914, "lr": 7.957795484216806e-05}
{"train_loss": 0.21596933901309967, "global_step": 81382, "epoch": 914, "lr": 7.957748742079428e-05}
{"train_loss": 0.12230360507965088, "global_step": 81383, "epoch": 914, "lr": 7.957701999544416e-05}
{"train_loss": 0.30698856711387634, "global_step": 81384, "epoch": 914, "lr": 7.957655256611778e-05}
{"train_loss": 0.2025662511587143, "global_step": 81385, "epoch": 914, "lr": 7.957608513281518e-05}
{"train_loss": 0.23556625843048096, "global_step": 81386, "epoch": 914, "lr": 7.957561769553644e-05}
{"train_loss": 0.17217527329921722, "global_step": 81387, "epoch": 914, "lr": 7.957515025428161e-05}
{"train_loss": 0.12504374980926514, "global_step": 81388, "epoch": 914, "lr": 7.957468280905078e-05}
{"train_loss": 0.18069571256637573, "global_step": 81389, "epoch": 914, "lr": 7.957421535984398e-05}
{"train_loss": 0.22800230979919434, "global_step": 81390, "epoch": 914, "lr": 7.95737479066613e-05}
{"train_loss": 0.19556009769439697, "global_step": 81391, "epoch": 914, "lr": 7.957328044950276e-05}
{"train_loss": 0.180422842502594, "global_step": 81392, "epoch": 914, "lr": 7.957281298836848e-05}
{"train_loss": 0.18150407075881958, "global_step": 81393, "epoch": 914, "lr": 7.95723455232585e-05}
{"train_loss": 0.14237859845161438, "global_step": 81394, "epoch": 914, "lr": 7.957187805417287e-05}
{"train_loss": 0.2600070834159851, "global_step": 81395, "epoch": 914, "lr": 7.957141058111166e-05}
{"train_loss": 0.1668711155653, "global_step": 81396, "epoch": 914, "lr": 7.957094310407494e-05}
{"train_loss": 0.19016920030117035, "global_step": 81397, "epoch": 914, "lr": 7.957047562306274e-05}
{"train_loss": 0.27475714683532715, "global_step": 81398, "epoch": 914, "lr": 7.957000813807519e-05}
{"train_loss": 0.17194890975952148, "global_step": 81399, "epoch": 914, "lr": 7.956954064911228e-05}
{"train_loss": 0.1443963348865509, "global_step": 81400, "epoch": 914, "lr": 7.956907315617412e-05}
{"train_loss": 0.21501219272613525, "global_step": 81401, "epoch": 914, "lr": 7.956860565926076e-05}
{"train_loss": 0.21547923982143402, "global_step": 81402, "epoch": 914, "lr": 7.956813815837227e-05}
{"train_loss": 0.3014931082725525, "global_step": 81403, "epoch": 914, "lr": 7.95676706535087e-05}
{"train_loss": 0.17001837491989136, "global_step": 81404, "epoch": 914, "lr": 7.95672031446701e-05}
{"train_loss": 0.17124320566654205, "global_step": 81405, "epoch": 914, "lr": 7.956673563185657e-05}
{"train_loss": 0.1504262387752533, "global_step": 81406, "epoch": 914, "lr": 7.956626811506814e-05}
{"train_loss": 0.1871717870235443, "global_step": 81407, "epoch": 914, "lr": 7.95658005943049e-05}
{"train_loss": 0.17550398409366608, "global_step": 81408, "epoch": 914, "lr": 7.956533306956688e-05}
{"train_loss": 0.23205332458019257, "global_step": 81409, "epoch": 914, "lr": 7.956486554085416e-05}
{"train_loss": 0.2269631326198578, "global_step": 81410, "epoch": 914, "lr": 7.956439800816682e-05}
{"train_loss": 0.1793685108423233, "global_step": 81411, "epoch": 914, "lr": 7.95639304715049e-05}
{"train_loss": 0.1901552677154541, "global_step": 81412, "epoch": 914, "lr": 7.956346293086847e-05}
{"train_loss": 0.20820879936218262, "global_step": 81413, "epoch": 914, "lr": 7.956299538625758e-05}
{"train_loss": 0.30806872248649597, "global_step": 81414, "epoch": 914, "lr": 7.956252783767233e-05}
{"train_loss": 0.17158013582229614, "global_step": 81415, "epoch": 914, "lr": 7.956206028511275e-05}
{"train_loss": 0.21673886477947235, "global_step": 81416, "epoch": 914, "lr": 7.95615927285789e-05}
{"train_loss": 0.2536543011665344, "global_step": 81417, "epoch": 914, "lr": 7.956112516807086e-05}
{"train_loss": 0.16151957213878632, "global_step": 81418, "epoch": 914, "lr": 7.956065760358869e-05}
{"train_loss": 0.20578275620937347, "global_step": 81419, "epoch": 914, "lr": 7.956019003513245e-05}
{"train_loss": 0.23213087022304535, "global_step": 81420, "epoch": 914, "lr": 7.95597224627022e-05}
{"train_loss": 0.212628573179245, "global_step": 81421, "epoch": 914, "lr": 7.955925488629802e-05}
{"train_loss": 0.21097451448440552, "global_step": 81422, "epoch": 914, "lr": 7.955878730591995e-05}
{"train_loss": 0.18049190938472748, "global_step": 81423, "epoch": 914, "lr": 7.955831972156804e-05}
{"train_loss": 0.20376290380954742, "global_step": 81424, "epoch": 914, "lr": 7.955785213324241e-05}
{"train_loss": 0.19551058113574982, "global_step": 81425, "epoch": 914, "lr": 7.955738454094306e-05}
{"train_loss": 0.11347878724336624, "global_step": 81426, "epoch": 914, "lr": 7.955691694467009e-05}
{"train_loss": 0.3211447596549988, "global_step": 81427, "epoch": 914, "lr": 7.955644934442355e-05}
{"train_loss": 0.21120554208755493, "global_step": 81428, "epoch": 914, "lr": 7.95559817402035e-05}
{"train_loss": 0.22073152661323547, "global_step": 81429, "epoch": 914, "lr": 7.955551413201002e-05}
{"train_loss": 0.23723866045475006, "global_step": 81430, "epoch": 914, "lr": 7.955504651984316e-05}
{"train_loss": 0.1814553141593933, "global_step": 81431, "epoch": 914, "lr": 7.955457890370297e-05}
{"train_loss": 0.1801048368215561, "global_step": 81432, "epoch": 914, "lr": 7.955411128358955e-05}
{"train_loss": 0.14413972198963165, "global_step": 81433, "epoch": 914, "lr": 7.955364365950292e-05}
{"train_loss": 0.19885326076424523, "global_step": 81434, "epoch": 914, "lr": 7.955317603144318e-05, "val_loss": 4.034419059753418}
{"train_loss": 0.19447684288024902, "global_step": 81435, "epoch": 915, "lr": 7.955270839941035e-05}
{"train_loss": 0.12772519886493683, "global_step": 81436, "epoch": 915, "lr": 7.955224076340455e-05}
{"train_loss": 0.18460848927497864, "global_step": 81437, "epoch": 915, "lr": 7.955177312342579e-05}
{"train_loss": 0.22115746140480042, "global_step": 81438, "epoch": 915, "lr": 7.955130547947417e-05}
{"train_loss": 0.18505752086639404, "global_step": 81439, "epoch": 915, "lr": 7.955083783154973e-05}
{"train_loss": 0.2375425100326538, "global_step": 81440, "epoch": 915, "lr": 7.955037017965254e-05}
{"train_loss": 0.20381149649620056, "global_step": 81441, "epoch": 915, "lr": 7.954990252378264e-05}
{"train_loss": 0.23176078498363495, "global_step": 81442, "epoch": 915, "lr": 7.954943486394015e-05}
{"train_loss": 0.1417953372001648, "global_step": 81443, "epoch": 915, "lr": 7.954896720012508e-05}
{"train_loss": 0.2257893681526184, "global_step": 81444, "epoch": 915, "lr": 7.954849953233752e-05}
{"train_loss": 0.15616412460803986, "global_step": 81445, "epoch": 915, "lr": 7.954803186057752e-05}
{"train_loss": 0.1408584862947464, "global_step": 81446, "epoch": 915, "lr": 7.954756418484515e-05}
{"train_loss": 0.13442562520503998, "global_step": 81447, "epoch": 915, "lr": 7.954709650514047e-05}
{"train_loss": 0.23693625628948212, "global_step": 81448, "epoch": 915, "lr": 7.954662882146353e-05}
{"train_loss": 0.17977777123451233, "global_step": 81449, "epoch": 915, "lr": 7.954616113381444e-05}
{"train_loss": 0.2708050310611725, "global_step": 81450, "epoch": 915, "lr": 7.954569344219319e-05}
{"train_loss": 0.24877937138080597, "global_step": 81451, "epoch": 915, "lr": 7.95452257465999e-05}
{"train_loss": 0.17069555819034576, "global_step": 81452, "epoch": 915, "lr": 7.954475804703462e-05}
{"train_loss": 0.19784942269325256, "global_step": 81453, "epoch": 915, "lr": 7.95442903434974e-05}
{"train_loss": 0.1906449943780899, "global_step": 81454, "epoch": 915, "lr": 7.954382263598831e-05}
{"train_loss": 0.23323629796504974, "global_step": 81455, "epoch": 915, "lr": 7.954335492450742e-05}
{"train_loss": 0.28283587098121643, "global_step": 81456, "epoch": 915, "lr": 7.954288720905478e-05}
{"train_loss": 0.12090512365102768, "global_step": 81457, "epoch": 915, "lr": 7.954241948963044e-05}
{"train_loss": 0.18911735713481903, "global_step": 81458, "epoch": 915, "lr": 7.954195176623451e-05}
{"train_loss": 0.1662501096725464, "global_step": 81459, "epoch": 915, "lr": 7.954148403886702e-05}
{"train_loss": 0.15193282067775726, "global_step": 81460, "epoch": 915, "lr": 7.954101630752803e-05}
{"train_loss": 0.1735994517803192, "global_step": 81461, "epoch": 915, "lr": 7.954054857221761e-05}
{"train_loss": 0.19398532807826996, "global_step": 81462, "epoch": 915, "lr": 7.954008083293583e-05}
{"train_loss": 0.17700858414173126, "global_step": 81463, "epoch": 915, "lr": 7.953961308968273e-05}
{"train_loss": 0.2075529545545578, "global_step": 81464, "epoch": 915, "lr": 7.95391453424584e-05}
{"train_loss": 0.23553979396820068, "global_step": 81465, "epoch": 915, "lr": 7.95386775912629e-05}
{"train_loss": 0.16079352796077728, "global_step": 81466, "epoch": 915, "lr": 7.953820983609628e-05}
{"train_loss": 0.20902127027511597, "global_step": 81467, "epoch": 915, "lr": 7.95377420769586e-05}
{"train_loss": 0.1931772232055664, "global_step": 81468, "epoch": 915, "lr": 7.953727431384994e-05}
{"train_loss": 0.17345382273197174, "global_step": 81469, "epoch": 915, "lr": 7.953680654677034e-05}
{"train_loss": 0.24402156472206116, "global_step": 81470, "epoch": 915, "lr": 7.953633877571988e-05}
{"train_loss": 0.17685677111148834, "global_step": 81471, "epoch": 915, "lr": 7.953587100069864e-05}
{"train_loss": 0.26655957102775574, "global_step": 81472, "epoch": 915, "lr": 7.953540322170665e-05}
{"train_loss": 0.09716686606407166, "global_step": 81473, "epoch": 915, "lr": 7.953493543874398e-05}
{"train_loss": 0.16144125163555145, "global_step": 81474, "epoch": 915, "lr": 7.95344676518107e-05}
{"train_loss": 0.2329939603805542, "global_step": 81475, "epoch": 915, "lr": 7.953399986090688e-05}
{"train_loss": 0.19530516862869263, "global_step": 81476, "epoch": 915, "lr": 7.953353206603256e-05}
{"train_loss": 0.3004503846168518, "global_step": 81477, "epoch": 915, "lr": 7.953306426718782e-05}
{"train_loss": 0.21792662143707275, "global_step": 81478, "epoch": 915, "lr": 7.953259646437272e-05}
{"train_loss": 0.185064896941185, "global_step": 81479, "epoch": 915, "lr": 7.953212865758732e-05}
{"train_loss": 0.272671103477478, "global_step": 81480, "epoch": 915, "lr": 7.953166084683169e-05}
{"train_loss": 0.15517646074295044, "global_step": 81481, "epoch": 915, "lr": 7.953119303210588e-05}
{"train_loss": 0.1782543659210205, "global_step": 81482, "epoch": 915, "lr": 7.953072521340996e-05}
{"train_loss": 0.22883623838424683, "global_step": 81483, "epoch": 915, "lr": 7.9530257390744e-05}
{"train_loss": 0.23466290533542633, "global_step": 81484, "epoch": 915, "lr": 7.952978956410807e-05}
{"train_loss": 0.23431634902954102, "global_step": 81485, "epoch": 915, "lr": 7.95293217335022e-05}
{"train_loss": 0.2582874000072479, "global_step": 81486, "epoch": 915, "lr": 7.952885389892648e-05}
{"train_loss": 0.18887297809123993, "global_step": 81487, "epoch": 915, "lr": 7.952838606038097e-05}
{"train_loss": 0.18553131818771362, "global_step": 81488, "epoch": 915, "lr": 7.952791821786573e-05}
{"train_loss": 0.20064803957939148, "global_step": 81489, "epoch": 915, "lr": 7.952745037138081e-05}
{"train_loss": 0.2086571753025055, "global_step": 81490, "epoch": 915, "lr": 7.95269825209263e-05}
{"train_loss": 0.1337752789258957, "global_step": 81491, "epoch": 915, "lr": 7.952651466650223e-05}
{"train_loss": 0.21723628044128418, "global_step": 81492, "epoch": 915, "lr": 7.952604680810868e-05}
{"train_loss": 0.15612848103046417, "global_step": 81493, "epoch": 915, "lr": 7.952557894574572e-05}
{"train_loss": 0.17910221219062805, "global_step": 81494, "epoch": 915, "lr": 7.952511107941342e-05}
{"train_loss": 0.1435585767030716, "global_step": 81495, "epoch": 915, "lr": 7.952464320911182e-05}
{"train_loss": 0.2151147574186325, "global_step": 81496, "epoch": 915, "lr": 7.952417533484097e-05}
{"train_loss": 0.1236562505364418, "global_step": 81497, "epoch": 915, "lr": 7.952370745660098e-05}
{"train_loss": 0.1807652860879898, "global_step": 81498, "epoch": 915, "lr": 7.952323957439188e-05}
{"train_loss": 0.20533066987991333, "global_step": 81499, "epoch": 915, "lr": 7.952277168821374e-05}
{"train_loss": 0.1930999457836151, "global_step": 81500, "epoch": 915, "lr": 7.952230379806661e-05}
{"train_loss": 0.13198551535606384, "global_step": 81501, "epoch": 915, "lr": 7.952183590395058e-05}
{"train_loss": 0.10974735021591187, "global_step": 81502, "epoch": 915, "lr": 7.952136800586569e-05}
{"train_loss": 0.1928279995918274, "global_step": 81503, "epoch": 915, "lr": 7.952090010381203e-05}
{"train_loss": 0.19395768642425537, "global_step": 81504, "epoch": 915, "lr": 7.952043219778961e-05}
{"train_loss": 0.22486257553100586, "global_step": 81505, "epoch": 915, "lr": 7.951996428779856e-05}
{"train_loss": 0.21205905079841614, "global_step": 81506, "epoch": 915, "lr": 7.95194963738389e-05}
{"train_loss": 0.24914121627807617, "global_step": 81507, "epoch": 915, "lr": 7.951902845591072e-05}
{"train_loss": 0.18137089908123016, "global_step": 81508, "epoch": 915, "lr": 7.951856053401405e-05}
{"train_loss": 0.3016865849494934, "global_step": 81509, "epoch": 915, "lr": 7.951809260814898e-05}
{"train_loss": 0.18680992722511292, "global_step": 81510, "epoch": 915, "lr": 7.951762467831554e-05}
{"train_loss": 0.12516175210475922, "global_step": 81511, "epoch": 915, "lr": 7.951715674451383e-05}
{"train_loss": 0.15450046956539154, "global_step": 81512, "epoch": 915, "lr": 7.95166888067439e-05}
{"train_loss": 0.1439901739358902, "global_step": 81513, "epoch": 915, "lr": 7.951622086500581e-05}
{"train_loss": 0.17854012548923492, "global_step": 81514, "epoch": 915, "lr": 7.951575291929961e-05}
{"train_loss": 0.16539451479911804, "global_step": 81515, "epoch": 915, "lr": 7.95152849696254e-05}
{"train_loss": 0.17550423741340637, "global_step": 81516, "epoch": 915, "lr": 7.95148170159832e-05}
{"train_loss": 0.21928554773330688, "global_step": 81517, "epoch": 915, "lr": 7.951434905837309e-05}
{"train_loss": 0.2483874261379242, "global_step": 81518, "epoch": 915, "lr": 7.951388109679516e-05}
{"train_loss": 0.15483057498931885, "global_step": 81519, "epoch": 915, "lr": 7.951341313124942e-05}
{"train_loss": 0.19883178174495697, "global_step": 81520, "epoch": 915, "lr": 7.951294516173598e-05}
{"train_loss": 0.20564617216587067, "global_step": 81521, "epoch": 915, "lr": 7.951247718825487e-05}
{"train_loss": 0.18929342925548553, "global_step": 81522, "epoch": 915, "lr": 7.951200921080618e-05}
{"train_loss": 0.1935039053807098, "global_step": 81523, "epoch": 915, "lr": 7.951154122938995e-05, "val_loss": 4.1466593742370605, "train_action_mse_error": 20.825477600097656}
{"train_loss": 0.23668918013572693, "global_step": 81524, "epoch": 916, "lr": 7.951107324400626e-05}
{"train_loss": 0.17185908555984497, "global_step": 81525, "epoch": 916, "lr": 7.951060525465517e-05}
{"train_loss": 0.1919698864221573, "global_step": 81526, "epoch": 916, "lr": 7.951013726133674e-05}
{"train_loss": 0.2193954437971115, "global_step": 81527, "epoch": 916, "lr": 7.950966926405103e-05}
{"train_loss": 0.1817077398300171, "global_step": 81528, "epoch": 916, "lr": 7.950920126279811e-05}
{"train_loss": 0.20954002439975739, "global_step": 81529, "epoch": 916, "lr": 7.950873325757801e-05}
{"train_loss": 0.22401940822601318, "global_step": 81530, "epoch": 916, "lr": 7.950826524839085e-05}
{"train_loss": 0.178028404712677, "global_step": 81531, "epoch": 916, "lr": 7.950779723523664e-05}
{"train_loss": 0.22510141134262085, "global_step": 81532, "epoch": 916, "lr": 7.950732921811548e-05}
{"train_loss": 0.16966480016708374, "global_step": 81533, "epoch": 916, "lr": 7.950686119702742e-05}
{"train_loss": 0.13698357343673706, "global_step": 81534, "epoch": 916, "lr": 7.950639317197253e-05}
{"train_loss": 0.12454423308372498, "global_step": 81535, "epoch": 916, "lr": 7.950592514295086e-05}
{"train_loss": 0.15973134338855743, "global_step": 81536, "epoch": 916, "lr": 7.950545710996245e-05}
{"train_loss": 0.2837097942829132, "global_step": 81537, "epoch": 916, "lr": 7.950498907300742e-05}
{"train_loss": 0.2191067487001419, "global_step": 81538, "epoch": 916, "lr": 7.950452103208579e-05}
{"train_loss": 0.11921732872724533, "global_step": 81539, "epoch": 916, "lr": 7.950405298719765e-05}
{"train_loss": 0.1357899159193039, "global_step": 81540, "epoch": 916, "lr": 7.950358493834304e-05}
{"train_loss": 0.1347028762102127, "global_step": 81541, "epoch": 916, "lr": 7.950311688552204e-05}
{"train_loss": 0.15349605679512024, "global_step": 81542, "epoch": 916, "lr": 7.95026488287347e-05}
{"train_loss": 0.2449805587530136, "global_step": 81543, "epoch": 916, "lr": 7.95021807679811e-05}
{"train_loss": 0.1788826733827591, "global_step": 81544, "epoch": 916, "lr": 7.950171270326127e-05}
{"train_loss": 0.17589710652828217, "global_step": 81545, "epoch": 916, "lr": 7.950124463457532e-05}
{"train_loss": 0.2414647489786148, "global_step": 81546, "epoch": 916, "lr": 7.950077656192327e-05}
{"train_loss": 0.18763212859630585, "global_step": 81547, "epoch": 916, "lr": 7.950030848530519e-05}
{"train_loss": 0.2685263752937317, "global_step": 81548, "epoch": 916, "lr": 7.949984040472119e-05}
{"train_loss": 0.16708099842071533, "global_step": 81549, "epoch": 916, "lr": 7.949937232017127e-05}
{"train_loss": 0.14421308040618896, "global_step": 81550, "epoch": 916, "lr": 7.949890423165551e-05}
{"train_loss": 0.2062271684408188, "global_step": 81551, "epoch": 916, "lr": 7.9498436139174e-05}
{"train_loss": 0.22387689352035522, "global_step": 81552, "epoch": 916, "lr": 7.949796804272678e-05}
{"train_loss": 0.12380502372980118, "global_step": 81553, "epoch": 916, "lr": 7.949749994231391e-05}
{"train_loss": 0.27785950899124146, "global_step": 81554, "epoch": 916, "lr": 7.949703183793546e-05}
{"train_loss": 0.17742052674293518, "global_step": 81555, "epoch": 916, "lr": 7.949656372959152e-05}
{"train_loss": 0.17465700209140778, "global_step": 81556, "epoch": 916, "lr": 7.949609561728209e-05}
{"train_loss": 0.11987944692373276, "global_step": 81557, "epoch": 916, "lr": 7.949562750100729e-05}
{"train_loss": 0.20063473284244537, "global_step": 81558, "epoch": 916, "lr": 7.949515938076715e-05}
{"train_loss": 0.1680605411529541, "global_step": 81559, "epoch": 916, "lr": 7.949469125656177e-05}
{"train_loss": 0.2366628646850586, "global_step": 81560, "epoch": 916, "lr": 7.949422312839117e-05}
{"train_loss": 0.20373564958572388, "global_step": 81561, "epoch": 916, "lr": 7.949375499625542e-05}
{"train_loss": 0.2085009664297104, "global_step": 81562, "epoch": 916, "lr": 7.949328686015461e-05}
{"train_loss": 0.21681034564971924, "global_step": 81563, "epoch": 916, "lr": 7.949281872008879e-05}
{"train_loss": 0.1304951161146164, "global_step": 81564, "epoch": 916, "lr": 7.949235057605802e-05}
{"train_loss": 0.1492512971162796, "global_step": 81565, "epoch": 916, "lr": 7.949188242806234e-05}
{"train_loss": 0.14359337091445923, "global_step": 81566, "epoch": 916, "lr": 7.949141427610186e-05}
{"train_loss": 0.18835178017616272, "global_step": 81567, "epoch": 916, "lr": 7.949094612017663e-05}
{"train_loss": 0.14916107058525085, "global_step": 81568, "epoch": 916, "lr": 7.949047796028666e-05}
{"train_loss": 0.24020794034004211, "global_step": 81569, "epoch": 916, "lr": 7.949000979643208e-05}
{"train_loss": 0.19383548200130463, "global_step": 81570, "epoch": 916, "lr": 7.948954162861293e-05}
{"train_loss": 0.17594748735427856, "global_step": 81571, "epoch": 916, "lr": 7.948907345682928e-05}
{"train_loss": 0.17529183626174927, "global_step": 81572, "epoch": 916, "lr": 7.948860528108115e-05}
{"train_loss": 0.1766585409641266, "global_step": 81573, "epoch": 916, "lr": 7.948813710136866e-05}
{"train_loss": 0.15408529341220856, "global_step": 81574, "epoch": 916, "lr": 7.948766891769185e-05}
{"train_loss": 0.18002483248710632, "global_step": 81575, "epoch": 916, "lr": 7.948720073005078e-05}
{"train_loss": 0.2883991003036499, "global_step": 81576, "epoch": 916, "lr": 7.948673253844552e-05}
{"train_loss": 0.18019741773605347, "global_step": 81577, "epoch": 916, "lr": 7.948626434287611e-05}
{"train_loss": 0.12139230221509933, "global_step": 81578, "epoch": 916, "lr": 7.948579614334263e-05}
{"train_loss": 0.1523963063955307, "global_step": 81579, "epoch": 916, "lr": 7.948532793984518e-05}
{"train_loss": 0.14183089137077332, "global_step": 81580, "epoch": 916, "lr": 7.948485973238375e-05}
{"train_loss": 0.20834191143512726, "global_step": 81581, "epoch": 916, "lr": 7.948439152095846e-05}
{"train_loss": 0.23630768060684204, "global_step": 81582, "epoch": 916, "lr": 7.948392330556935e-05}
{"train_loss": 0.1923460215330124, "global_step": 81583, "epoch": 916, "lr": 7.948345508621648e-05}
{"train_loss": 0.23924636840820312, "global_step": 81584, "epoch": 916, "lr": 7.948298686289992e-05}
{"train_loss": 0.13981856405735016, "global_step": 81585, "epoch": 916, "lr": 7.948251863561973e-05}
{"train_loss": 0.20027416944503784, "global_step": 81586, "epoch": 916, "lr": 7.9482050404376e-05}
{"train_loss": 0.18436147272586823, "global_step": 81587, "epoch": 916, "lr": 7.948158216916873e-05}
{"train_loss": 0.2074550837278366, "global_step": 81588, "epoch": 916, "lr": 7.948111392999804e-05}
{"train_loss": 0.1581629365682602, "global_step": 81589, "epoch": 916, "lr": 7.948064568686398e-05}
{"train_loss": 0.20549438893795013, "global_step": 81590, "epoch": 916, "lr": 7.948017743976658e-05}
{"train_loss": 0.16583505272865295, "global_step": 81591, "epoch": 916, "lr": 7.947970918870595e-05}
{"train_loss": 0.15554556250572205, "global_step": 81592, "epoch": 916, "lr": 7.947924093368212e-05}
{"train_loss": 0.1591218113899231, "global_step": 81593, "epoch": 916, "lr": 7.947877267469518e-05}
{"train_loss": 0.24356791377067566, "global_step": 81594, "epoch": 916, "lr": 7.947830441174517e-05}
{"train_loss": 0.2222314327955246, "global_step": 81595, "epoch": 916, "lr": 7.947783614483216e-05}
{"train_loss": 0.24787642061710358, "global_step": 81596, "epoch": 916, "lr": 7.947736787395622e-05}
{"train_loss": 0.14582177996635437, "global_step": 81597, "epoch": 916, "lr": 7.94768995991174e-05}
{"train_loss": 0.17151597142219543, "global_step": 81598, "epoch": 916, "lr": 7.947643132031578e-05}
{"train_loss": 0.16442398726940155, "global_step": 81599, "epoch": 916, "lr": 7.94759630375514e-05}
{"train_loss": 0.19112065434455872, "global_step": 81600, "epoch": 916, "lr": 7.947549475082436e-05}
{"train_loss": 0.18688389658927917, "global_step": 81601, "epoch": 916, "lr": 7.947502646013467e-05}
{"train_loss": 0.1671481728553772, "global_step": 81602, "epoch": 916, "lr": 7.947455816548244e-05}
{"train_loss": 0.17866939306259155, "global_step": 81603, "epoch": 916, "lr": 7.947408986686771e-05}
{"train_loss": 0.22071640193462372, "global_step": 81604, "epoch": 916, "lr": 7.947362156429055e-05}
{"train_loss": 0.18811559677124023, "global_step": 81605, "epoch": 916, "lr": 7.947315325775102e-05}
{"train_loss": 0.2223128080368042, "global_step": 81606, "epoch": 916, "lr": 7.947268494724917e-05}
{"train_loss": 0.1775578260421753, "global_step": 81607, "epoch": 916, "lr": 7.94722166327851e-05}
{"train_loss": 0.1947549730539322, "global_step": 81608, "epoch": 916, "lr": 7.947174831435884e-05}
{"train_loss": 0.12049072980880737, "global_step": 81609, "epoch": 916, "lr": 7.947127999197046e-05}
{"train_loss": 0.14714884757995605, "global_step": 81610, "epoch": 916, "lr": 7.947081166562002e-05}
{"train_loss": 0.14054469764232635, "global_step": 81611, "epoch": 916, "lr": 7.94703433353076e-05}
{"train_loss": 0.18599475484885528, "global_step": 81612, "epoch": 916, "lr": 7.946987500103324e-05, "val_loss": 4.0858154296875}
{"train_loss": 0.2844894230365753, "global_step": 81613, "epoch": 917, "lr": 7.946940666279703e-05}
{"train_loss": 0.20728586614131927, "global_step": 81614, "epoch": 917, "lr": 7.9468938320599e-05}
{"train_loss": 0.12186725437641144, "global_step": 81615, "epoch": 917, "lr": 7.946846997443924e-05}
{"train_loss": 0.22869127988815308, "global_step": 81616, "epoch": 917, "lr": 7.94680016243178e-05}
{"train_loss": 0.12081494927406311, "global_step": 81617, "epoch": 917, "lr": 7.946753327023475e-05}
{"train_loss": 0.21171678602695465, "global_step": 81618, "epoch": 917, "lr": 7.946706491219016e-05}
{"train_loss": 0.2234383523464203, "global_step": 81619, "epoch": 917, "lr": 7.946659655018408e-05}
{"train_loss": 0.1689220517873764, "global_step": 81620, "epoch": 917, "lr": 7.946612818421655e-05}
{"train_loss": 0.27438613772392273, "global_step": 81621, "epoch": 917, "lr": 7.946565981428768e-05}
{"train_loss": 0.18837197124958038, "global_step": 81622, "epoch": 917, "lr": 7.94651914403975e-05}
{"train_loss": 0.13755638897418976, "global_step": 81623, "epoch": 917, "lr": 7.946472306254609e-05}
{"train_loss": 0.18555967509746552, "global_step": 81624, "epoch": 917, "lr": 7.94642546807335e-05}
{"train_loss": 0.1938760131597519, "global_step": 81625, "epoch": 917, "lr": 7.946378629495981e-05}
{"train_loss": 0.2197309136390686, "global_step": 81626, "epoch": 917, "lr": 7.946331790522507e-05}
{"train_loss": 0.24715416133403778, "global_step": 81627, "epoch": 917, "lr": 7.946284951152934e-05}
{"train_loss": 0.17517220973968506, "global_step": 81628, "epoch": 917, "lr": 7.94623811138727e-05}
{"train_loss": 0.11975458264350891, "global_step": 81629, "epoch": 917, "lr": 7.94619127122552e-05}
{"train_loss": 0.18512485921382904, "global_step": 81630, "epoch": 917, "lr": 7.94614443066769e-05}
{"train_loss": 0.15752549469470978, "global_step": 81631, "epoch": 917, "lr": 7.946097589713786e-05}
{"train_loss": 0.1523650735616684, "global_step": 81632, "epoch": 917, "lr": 7.946050748363816e-05}
{"train_loss": 0.17627961933612823, "global_step": 81633, "epoch": 917, "lr": 7.946003906617787e-05}
{"train_loss": 0.1129755824804306, "global_step": 81634, "epoch": 917, "lr": 7.945957064475702e-05}
{"train_loss": 0.22151008248329163, "global_step": 81635, "epoch": 917, "lr": 7.945910221937569e-05}
{"train_loss": 0.2432943731546402, "global_step": 81636, "epoch": 917, "lr": 7.945863379003393e-05}
{"train_loss": 0.17561151087284088, "global_step": 81637, "epoch": 917, "lr": 7.945816535673184e-05}
{"train_loss": 0.17592290043830872, "global_step": 81638, "epoch": 917, "lr": 7.945769691946945e-05}
{"train_loss": 0.11764583736658096, "global_step": 81639, "epoch": 917, "lr": 7.945722847824683e-05}
{"train_loss": 0.19679994881153107, "global_step": 81640, "epoch": 917, "lr": 7.945676003306404e-05}
{"train_loss": 0.13808004558086395, "global_step": 81641, "epoch": 917, "lr": 7.945629158392114e-05}
{"train_loss": 0.1723431795835495, "global_step": 81642, "epoch": 917, "lr": 7.945582313081823e-05}
{"train_loss": 0.15769535303115845, "global_step": 81643, "epoch": 917, "lr": 7.945535467375531e-05}
{"train_loss": 0.11890634149312973, "global_step": 81644, "epoch": 917, "lr": 7.94548862127325e-05}
{"train_loss": 0.22492322325706482, "global_step": 81645, "epoch": 917, "lr": 7.945441774774982e-05}
{"train_loss": 0.18805371224880219, "global_step": 81646, "epoch": 917, "lr": 7.945394927880738e-05}
{"train_loss": 0.23559872806072235, "global_step": 81647, "epoch": 917, "lr": 7.94534808059052e-05}
{"train_loss": 0.1335468739271164, "global_step": 81648, "epoch": 917, "lr": 7.945301232904336e-05}
{"train_loss": 0.15340840816497803, "global_step": 81649, "epoch": 917, "lr": 7.945254384822192e-05}
{"train_loss": 0.15944866836071014, "global_step": 81650, "epoch": 917, "lr": 7.945207536344094e-05}
{"train_loss": 0.16019393503665924, "global_step": 81651, "epoch": 917, "lr": 7.94516068747005e-05}
{"train_loss": 0.16037698090076447, "global_step": 81652, "epoch": 917, "lr": 7.945113838200064e-05}
{"train_loss": 0.22332915663719177, "global_step": 81653, "epoch": 917, "lr": 7.945066988534143e-05}
{"train_loss": 0.2611823081970215, "global_step": 81654, "epoch": 917, "lr": 7.945020138472295e-05}
{"train_loss": 0.17763735353946686, "global_step": 81655, "epoch": 917, "lr": 7.944973288014523e-05}
{"train_loss": 0.18426628410816193, "global_step": 81656, "epoch": 917, "lr": 7.944926437160838e-05}
{"train_loss": 0.3130447566509247, "global_step": 81657, "epoch": 917, "lr": 7.94487958591124e-05}
{"train_loss": 0.22339168190956116, "global_step": 81658, "epoch": 917, "lr": 7.944832734265742e-05}
{"train_loss": 0.1822083592414856, "global_step": 81659, "epoch": 917, "lr": 7.944785882224346e-05}
{"train_loss": 0.13623136281967163, "global_step": 81660, "epoch": 917, "lr": 7.944739029787059e-05}
{"train_loss": 0.14037778973579407, "global_step": 81661, "epoch": 917, "lr": 7.944692176953887e-05}
{"train_loss": 0.21444091200828552, "global_step": 81662, "epoch": 917, "lr": 7.94464532372484e-05}
{"train_loss": 0.22911323606967926, "global_step": 81663, "epoch": 917, "lr": 7.944598470099918e-05}
{"train_loss": 0.10388750582933426, "global_step": 81664, "epoch": 917, "lr": 7.944551616079132e-05}
{"train_loss": 0.11189398914575577, "global_step": 81665, "epoch": 917, "lr": 7.944504761662487e-05}
{"train_loss": 0.20847350358963013, "global_step": 81666, "epoch": 917, "lr": 7.944457906849989e-05}
{"train_loss": 0.12389557808637619, "global_step": 81667, "epoch": 917, "lr": 7.944411051641645e-05}
{"train_loss": 0.1430535614490509, "global_step": 81668, "epoch": 917, "lr": 7.944364196037461e-05}
{"train_loss": 0.2608993649482727, "global_step": 81669, "epoch": 917, "lr": 7.944317340037441e-05}
{"train_loss": 0.1919451802968979, "global_step": 81670, "epoch": 917, "lr": 7.944270483641596e-05}
{"train_loss": 0.24269135296344757, "global_step": 81671, "epoch": 917, "lr": 7.944223626849929e-05}
{"train_loss": 0.17520950734615326, "global_step": 81672, "epoch": 917, "lr": 7.944176769662448e-05}
{"train_loss": 0.23160137236118317, "global_step": 81673, "epoch": 917, "lr": 7.944129912079157e-05}
{"train_loss": 0.1993100643157959, "global_step": 81674, "epoch": 917, "lr": 7.944083054100064e-05}
{"train_loss": 0.15706589818000793, "global_step": 81675, "epoch": 917, "lr": 7.944036195725173e-05}
{"train_loss": 0.22949689626693726, "global_step": 81676, "epoch": 917, "lr": 7.943989336954495e-05}
{"train_loss": 0.19659166038036346, "global_step": 81677, "epoch": 917, "lr": 7.943942477788032e-05}
{"train_loss": 0.22357644140720367, "global_step": 81678, "epoch": 917, "lr": 7.943895618225792e-05}
{"train_loss": 0.18632712960243225, "global_step": 81679, "epoch": 917, "lr": 7.943848758267783e-05}
{"train_loss": 0.16091641783714294, "global_step": 81680, "epoch": 917, "lr": 7.943801897914006e-05}
{"train_loss": 0.19855639338493347, "global_step": 81681, "epoch": 917, "lr": 7.943755037164473e-05}
{"train_loss": 0.2537648379802704, "global_step": 81682, "epoch": 917, "lr": 7.943708176019187e-05}
{"train_loss": 0.23300427198410034, "global_step": 81683, "epoch": 917, "lr": 7.943661314478156e-05}
{"train_loss": 0.13635817170143127, "global_step": 81684, "epoch": 917, "lr": 7.943614452541384e-05}
{"train_loss": 0.2095824033021927, "global_step": 81685, "epoch": 917, "lr": 7.943567590208882e-05}
{"train_loss": 0.16514666378498077, "global_step": 81686, "epoch": 917, "lr": 7.94352072748065e-05}
{"train_loss": 0.127128005027771, "global_step": 81687, "epoch": 917, "lr": 7.943473864356699e-05}
{"train_loss": 0.12147767841815948, "global_step": 81688, "epoch": 917, "lr": 7.943427000837034e-05}
{"train_loss": 0.1508350372314453, "global_step": 81689, "epoch": 917, "lr": 7.943380136921661e-05}
{"train_loss": 0.18872332572937012, "global_step": 81690, "epoch": 917, "lr": 7.943333272610587e-05}
{"train_loss": 0.16690261662006378, "global_step": 81691, "epoch": 917, "lr": 7.943286407903819e-05}
{"train_loss": 0.1632855236530304, "global_step": 81692, "epoch": 917, "lr": 7.943239542801358e-05}
{"train_loss": 0.21768394112586975, "global_step": 81693, "epoch": 917, "lr": 7.943192677303217e-05}
{"train_loss": 0.19395966827869415, "global_step": 81694, "epoch": 917, "lr": 7.943145811409399e-05}
{"train_loss": 0.21246910095214844, "global_step": 81695, "epoch": 917, "lr": 7.94309894511991e-05}
{"train_loss": 0.1447383463382721, "global_step": 81696, "epoch": 917, "lr": 7.943052078434759e-05}
{"train_loss": 0.12233719229698181, "global_step": 81697, "epoch": 917, "lr": 7.94300521135395e-05}
{"train_loss": 0.25039130449295044, "global_step": 81698, "epoch": 917, "lr": 7.942958343877489e-05}
{"train_loss": 0.13875922560691833, "global_step": 81699, "epoch": 917, "lr": 7.942911476005384e-05}
{"train_loss": 0.2329806685447693, "global_step": 81700, "epoch": 917, "lr": 7.942864607737638e-05}
{"train_loss": 0.18519180636392552, "global_step": 81701, "epoch": 917, "lr": 7.942817739074264e-05, "val_loss": 4.047610282897949}
{"train_loss": 0.1883753389120102, "global_step": 81702, "epoch": 918, "lr": 7.942770870015261e-05}
{"train_loss": 0.2724255621433258, "global_step": 81703, "epoch": 918, "lr": 7.94272400056064e-05}
{"train_loss": 0.2093627154827118, "global_step": 81704, "epoch": 918, "lr": 7.942677130710404e-05}
{"train_loss": 0.23605236411094666, "global_step": 81705, "epoch": 918, "lr": 7.942630260464561e-05}
{"train_loss": 0.2066185027360916, "global_step": 81706, "epoch": 918, "lr": 7.94258338982312e-05}
{"train_loss": 0.18833693861961365, "global_step": 81707, "epoch": 918, "lr": 7.94253651878608e-05}
{"train_loss": 0.15935780107975006, "global_step": 81708, "epoch": 918, "lr": 7.942489647353454e-05}
{"train_loss": 0.12463224679231644, "global_step": 81709, "epoch": 918, "lr": 7.942442775525249e-05}
{"train_loss": 0.20379149913787842, "global_step": 81710, "epoch": 918, "lr": 7.942395903301465e-05}
{"train_loss": 0.16270627081394196, "global_step": 81711, "epoch": 918, "lr": 7.942349030682112e-05}
{"train_loss": 0.24323245882987976, "global_step": 81712, "epoch": 918, "lr": 7.942302157667197e-05}
{"train_loss": 0.1746956706047058, "global_step": 81713, "epoch": 918, "lr": 7.942255284256726e-05}
{"train_loss": 0.33701252937316895, "global_step": 81714, "epoch": 918, "lr": 7.942208410450702e-05}
{"train_loss": 0.19186364114284515, "global_step": 81715, "epoch": 918, "lr": 7.942161536249136e-05}
{"train_loss": 0.19472548365592957, "global_step": 81716, "epoch": 918, "lr": 7.942114661652031e-05}
{"train_loss": 0.17333362996578217, "global_step": 81717, "epoch": 918, "lr": 7.942067786659396e-05}
{"train_loss": 0.23723846673965454, "global_step": 81718, "epoch": 918, "lr": 7.942020911271236e-05}
{"train_loss": 0.14141467213630676, "global_step": 81719, "epoch": 918, "lr": 7.941974035487556e-05}
{"train_loss": 0.16002747416496277, "global_step": 81720, "epoch": 918, "lr": 7.941927159308363e-05}
{"train_loss": 0.1965368241071701, "global_step": 81721, "epoch": 918, "lr": 7.941880282733665e-05}
{"train_loss": 0.2089499682188034, "global_step": 81722, "epoch": 918, "lr": 7.941833405763465e-05}
{"train_loss": 0.20038196444511414, "global_step": 81723, "epoch": 918, "lr": 7.941786528397774e-05}
{"train_loss": 0.18025486171245575, "global_step": 81724, "epoch": 918, "lr": 7.941739650636595e-05}
{"train_loss": 0.22481276094913483, "global_step": 81725, "epoch": 918, "lr": 7.941692772479934e-05}
{"train_loss": 0.1818203330039978, "global_step": 81726, "epoch": 918, "lr": 7.941645893927798e-05}
{"train_loss": 0.20286720991134644, "global_step": 81727, "epoch": 918, "lr": 7.941599014980195e-05}
{"train_loss": 0.1870374232530594, "global_step": 81728, "epoch": 918, "lr": 7.94155213563713e-05}
{"train_loss": 0.15306998789310455, "global_step": 81729, "epoch": 918, "lr": 7.941505255898608e-05}
{"train_loss": 0.14835940301418304, "global_step": 81730, "epoch": 918, "lr": 7.941458375764637e-05}
{"train_loss": 0.25919413566589355, "global_step": 81731, "epoch": 918, "lr": 7.941411495235223e-05}
{"train_loss": 0.19493630528450012, "global_step": 81732, "epoch": 918, "lr": 7.941364614310371e-05}
{"train_loss": 0.1581425815820694, "global_step": 81733, "epoch": 918, "lr": 7.94131773299009e-05}
{"train_loss": 0.2289421558380127, "global_step": 81734, "epoch": 918, "lr": 7.941270851274383e-05}
{"train_loss": 0.19946736097335815, "global_step": 81735, "epoch": 918, "lr": 7.941223969163258e-05}
{"train_loss": 0.21892976760864258, "global_step": 81736, "epoch": 918, "lr": 7.941177086656722e-05}
{"train_loss": 0.19133926928043365, "global_step": 81737, "epoch": 918, "lr": 7.941130203754781e-05}
{"train_loss": 0.17735964059829712, "global_step": 81738, "epoch": 918, "lr": 7.941083320457441e-05}
{"train_loss": 0.22940272092819214, "global_step": 81739, "epoch": 918, "lr": 7.941036436764707e-05}
{"train_loss": 0.21594886481761932, "global_step": 81740, "epoch": 918, "lr": 7.940989552676587e-05}
{"train_loss": 0.16296599805355072, "global_step": 81741, "epoch": 918, "lr": 7.940942668193087e-05}
{"train_loss": 0.18914389610290527, "global_step": 81742, "epoch": 918, "lr": 7.940895783314213e-05}
{"train_loss": 0.2216249257326126, "global_step": 81743, "epoch": 918, "lr": 7.940848898039972e-05}
{"train_loss": 0.12408740818500519, "global_step": 81744, "epoch": 918, "lr": 7.94080201237037e-05}
{"train_loss": 0.19378574192523956, "global_step": 81745, "epoch": 918, "lr": 7.940755126305412e-05}
{"train_loss": 0.17459097504615784, "global_step": 81746, "epoch": 918, "lr": 7.940708239845105e-05}
{"train_loss": 0.1506941318511963, "global_step": 81747, "epoch": 918, "lr": 7.940661352989456e-05}
{"train_loss": 0.17776472866535187, "global_step": 81748, "epoch": 918, "lr": 7.940614465738472e-05}
{"train_loss": 0.22578565776348114, "global_step": 81749, "epoch": 918, "lr": 7.940567578092157e-05}
{"train_loss": 0.17953617870807648, "global_step": 81750, "epoch": 918, "lr": 7.94052069005052e-05}
{"train_loss": 0.2072344422340393, "global_step": 81751, "epoch": 918, "lr": 7.940473801613563e-05}
{"train_loss": 0.17543119192123413, "global_step": 81752, "epoch": 918, "lr": 7.940426912781299e-05}
{"train_loss": 0.22998671233654022, "global_step": 81753, "epoch": 918, "lr": 7.940380023553726e-05}
{"train_loss": 0.31249910593032837, "global_step": 81754, "epoch": 918, "lr": 7.940333133930857e-05}
{"train_loss": 0.15908528864383698, "global_step": 81755, "epoch": 918, "lr": 7.940286243912696e-05}
{"train_loss": 0.19829295575618744, "global_step": 81756, "epoch": 918, "lr": 7.94023935349925e-05}
{"train_loss": 0.14173628389835358, "global_step": 81757, "epoch": 918, "lr": 7.940192462690523e-05}
{"train_loss": 0.19053389132022858, "global_step": 81758, "epoch": 918, "lr": 7.940145571486525e-05}
{"train_loss": 0.21145617961883545, "global_step": 81759, "epoch": 918, "lr": 7.940098679887259e-05}
{"train_loss": 0.2192782163619995, "global_step": 81760, "epoch": 918, "lr": 7.940051787892733e-05}
{"train_loss": 0.13736991584300995, "global_step": 81761, "epoch": 918, "lr": 7.940004895502953e-05}
{"train_loss": 0.18997585773468018, "global_step": 81762, "epoch": 918, "lr": 7.939958002717924e-05}
{"train_loss": 0.23965990543365479, "global_step": 81763, "epoch": 918, "lr": 7.939911109537654e-05}
{"train_loss": 0.203405499458313, "global_step": 81764, "epoch": 918, "lr": 7.93986421596215e-05}
{"train_loss": 0.19707193970680237, "global_step": 81765, "epoch": 918, "lr": 7.939817321991416e-05}
{"train_loss": 0.16518913209438324, "global_step": 81766, "epoch": 918, "lr": 7.93977042762546e-05}
{"train_loss": 0.2252659797668457, "global_step": 81767, "epoch": 918, "lr": 7.939723532864287e-05}
{"train_loss": 0.11673508584499359, "global_step": 81768, "epoch": 918, "lr": 7.939676637707905e-05}
{"train_loss": 0.18363039195537567, "global_step": 81769, "epoch": 918, "lr": 7.939629742156318e-05}
{"train_loss": 0.21736577153205872, "global_step": 81770, "epoch": 918, "lr": 7.939582846209535e-05}
{"train_loss": 0.16996558010578156, "global_step": 81771, "epoch": 918, "lr": 7.93953594986756e-05}
{"train_loss": 0.2544000744819641, "global_step": 81772, "epoch": 918, "lr": 7.939489053130399e-05}
{"train_loss": 0.15836286544799805, "global_step": 81773, "epoch": 918, "lr": 7.93944215599806e-05}
{"train_loss": 0.15423880517482758, "global_step": 81774, "epoch": 918, "lr": 7.939395258470551e-05}
{"train_loss": 0.16201002895832062, "global_step": 81775, "epoch": 918, "lr": 7.939348360547876e-05}
{"train_loss": 0.16278564929962158, "global_step": 81776, "epoch": 918, "lr": 7.93930146223004e-05}
{"train_loss": 0.12111221998929977, "global_step": 81777, "epoch": 918, "lr": 7.939254563517051e-05}
{"train_loss": 0.15390485525131226, "global_step": 81778, "epoch": 918, "lr": 7.939207664408915e-05}
{"train_loss": 0.19534221291542053, "global_step": 81779, "epoch": 918, "lr": 7.939160764905637e-05}
{"train_loss": 0.21432176232337952, "global_step": 81780, "epoch": 918, "lr": 7.939113865007227e-05}
{"train_loss": 0.12515392899513245, "global_step": 81781, "epoch": 918, "lr": 7.939066964713688e-05}
{"train_loss": 0.1372295767068863, "global_step": 81782, "epoch": 918, "lr": 7.939020064025026e-05}
{"train_loss": 0.16137272119522095, "global_step": 81783, "epoch": 918, "lr": 7.938973162941251e-05}
{"train_loss": 0.17837969958782196, "global_step": 81784, "epoch": 918, "lr": 7.938926261462366e-05}
{"train_loss": 0.32570499181747437, "global_step": 81785, "epoch": 918, "lr": 7.938879359588378e-05}
{"train_loss": 0.22272738814353943, "global_step": 81786, "epoch": 918, "lr": 7.938832457319294e-05}
{"train_loss": 0.2428615242242813, "global_step": 81787, "epoch": 918, "lr": 7.93878555465512e-05}
{"train_loss": 0.1416778564453125, "global_step": 81788, "epoch": 918, "lr": 7.938738651595861e-05}
{"train_loss": 0.2030014991760254, "global_step": 81789, "epoch": 918, "lr": 7.938691748141525e-05}
{"train_loss": 0.19261546349257566, "global_step": 81790, "epoch": 918, "lr": 7.938644844292116e-05, "val_loss": 4.1816606521606445}
{"train_loss": 0.14440087974071503, "global_step": 81791, "epoch": 919, "lr": 7.938597940047642e-05}
{"train_loss": 0.16357052326202393, "global_step": 81792, "epoch": 919, "lr": 7.938551035408113e-05}
{"train_loss": 0.16004809737205505, "global_step": 81793, "epoch": 919, "lr": 7.938504130373528e-05}
{"train_loss": 0.17359516024589539, "global_step": 81794, "epoch": 919, "lr": 7.9384572249439e-05}
{"train_loss": 0.19912369549274445, "global_step": 81795, "epoch": 919, "lr": 7.93841031911923e-05}
{"train_loss": 0.15188366174697876, "global_step": 81796, "epoch": 919, "lr": 7.938363412899528e-05}
{"train_loss": 0.12349187582731247, "global_step": 81797, "epoch": 919, "lr": 7.938316506284797e-05}
{"train_loss": 0.13342095911502838, "global_step": 81798, "epoch": 919, "lr": 7.938269599275047e-05}
{"train_loss": 0.10260400176048279, "global_step": 81799, "epoch": 919, "lr": 7.93822269187028e-05}
{"train_loss": 0.1853078007698059, "global_step": 81800, "epoch": 919, "lr": 7.938175784070507e-05}
{"train_loss": 0.1693156659603119, "global_step": 81801, "epoch": 919, "lr": 7.938128875875732e-05}
{"train_loss": 0.16824911534786224, "global_step": 81802, "epoch": 919, "lr": 7.93808196728596e-05}
{"train_loss": 0.26637786626815796, "global_step": 81803, "epoch": 919, "lr": 7.9380350583012e-05}
{"train_loss": 0.22102409601211548, "global_step": 81804, "epoch": 919, "lr": 7.937988148921458e-05}
{"train_loss": 0.2880173325538635, "global_step": 81805, "epoch": 919, "lr": 7.937941239146738e-05}
{"train_loss": 0.1806049346923828, "global_step": 81806, "epoch": 919, "lr": 7.937894328977047e-05}
{"train_loss": 0.16036993265151978, "global_step": 81807, "epoch": 919, "lr": 7.937847418412392e-05}
{"train_loss": 0.11539509147405624, "global_step": 81808, "epoch": 919, "lr": 7.937800507452779e-05}
{"train_loss": 0.2349415272474289, "global_step": 81809, "epoch": 919, "lr": 7.937753596098215e-05}
{"train_loss": 0.1267346739768982, "global_step": 81810, "epoch": 919, "lr": 7.937706684348707e-05}
{"train_loss": 0.18051587045192719, "global_step": 81811, "epoch": 919, "lr": 7.937659772204258e-05}
{"train_loss": 0.25505247712135315, "global_step": 81812, "epoch": 919, "lr": 7.937612859664879e-05}
{"train_loss": 0.1682412028312683, "global_step": 81813, "epoch": 919, "lr": 7.937565946730572e-05}
{"train_loss": 0.11664063483476639, "global_step": 81814, "epoch": 919, "lr": 7.937519033401345e-05}
{"train_loss": 0.1416158378124237, "global_step": 81815, "epoch": 919, "lr": 7.937472119677205e-05}
{"train_loss": 0.19640155136585236, "global_step": 81816, "epoch": 919, "lr": 7.937425205558158e-05}
{"train_loss": 0.12752950191497803, "global_step": 81817, "epoch": 919, "lr": 7.937378291044208e-05}
{"train_loss": 0.12489314377307892, "global_step": 81818, "epoch": 919, "lr": 7.937331376135365e-05}
{"train_loss": 0.23263981938362122, "global_step": 81819, "epoch": 919, "lr": 7.937284460831634e-05}
{"train_loss": 0.2276826649904251, "global_step": 81820, "epoch": 919, "lr": 7.937237545133021e-05}
{"train_loss": 0.21804159879684448, "global_step": 81821, "epoch": 919, "lr": 7.93719062903953e-05}
{"train_loss": 0.2181173413991928, "global_step": 81822, "epoch": 919, "lr": 7.937143712551171e-05}
{"train_loss": 0.2837342917919159, "global_step": 81823, "epoch": 919, "lr": 7.937096795667948e-05}
{"train_loss": 0.16324642300605774, "global_step": 81824, "epoch": 919, "lr": 7.93704987838987e-05}
{"train_loss": 0.1369500756263733, "global_step": 81825, "epoch": 919, "lr": 7.93700296071694e-05}
{"train_loss": 0.12767614424228668, "global_step": 81826, "epoch": 919, "lr": 7.936956042649166e-05}
{"train_loss": 0.18213114142417908, "global_step": 81827, "epoch": 919, "lr": 7.936909124186554e-05}
{"train_loss": 0.2092002034187317, "global_step": 81828, "epoch": 919, "lr": 7.93686220532911e-05}
{"train_loss": 0.17148999869823456, "global_step": 81829, "epoch": 919, "lr": 7.936815286076843e-05}
{"train_loss": 0.16546393930912018, "global_step": 81830, "epoch": 919, "lr": 7.936768366429754e-05}
{"train_loss": 0.15538766980171204, "global_step": 81831, "epoch": 919, "lr": 7.936721446387854e-05}
{"train_loss": 0.17677690088748932, "global_step": 81832, "epoch": 919, "lr": 7.936674525951146e-05}
{"train_loss": 0.20387877523899078, "global_step": 81833, "epoch": 919, "lr": 7.93662760511964e-05}
{"train_loss": 0.1921447366476059, "global_step": 81834, "epoch": 919, "lr": 7.936580683893339e-05}
{"train_loss": 0.20897260308265686, "global_step": 81835, "epoch": 919, "lr": 7.93653376227225e-05}
{"train_loss": 0.21522153913974762, "global_step": 81836, "epoch": 919, "lr": 7.93648684025638e-05}
{"train_loss": 0.20595455169677734, "global_step": 81837, "epoch": 919, "lr": 7.936439917845737e-05}
{"train_loss": 0.1692531257867813, "global_step": 81838, "epoch": 919, "lr": 7.936392995040324e-05}
{"train_loss": 0.08724001795053482, "global_step": 81839, "epoch": 919, "lr": 7.936346071840148e-05}
{"train_loss": 0.1682678461074829, "global_step": 81840, "epoch": 919, "lr": 7.936299148245217e-05}
{"train_loss": 0.17750360071659088, "global_step": 81841, "epoch": 919, "lr": 7.936252224255536e-05}
{"train_loss": 0.28162890672683716, "global_step": 81842, "epoch": 919, "lr": 7.936205299871114e-05}
{"train_loss": 0.17529618740081787, "global_step": 81843, "epoch": 919, "lr": 7.936158375091951e-05}
{"train_loss": 0.20694243907928467, "global_step": 81844, "epoch": 919, "lr": 7.93611144991806e-05}
{"train_loss": 0.1931150108575821, "global_step": 81845, "epoch": 919, "lr": 7.936064524349443e-05}
{"train_loss": 0.13673466444015503, "global_step": 81846, "epoch": 919, "lr": 7.93601759838611e-05}
{"train_loss": 0.2167019248008728, "global_step": 81847, "epoch": 919, "lr": 7.935970672028064e-05}
{"train_loss": 0.1756090521812439, "global_step": 81848, "epoch": 919, "lr": 7.935923745275311e-05}
{"train_loss": 0.1972913146018982, "global_step": 81849, "epoch": 919, "lr": 7.93587681812786e-05}
{"train_loss": 0.11743425577878952, "global_step": 81850, "epoch": 919, "lr": 7.935829890585717e-05}
{"train_loss": 0.17189183831214905, "global_step": 81851, "epoch": 919, "lr": 7.935782962648886e-05}
{"train_loss": 0.22072207927703857, "global_step": 81852, "epoch": 919, "lr": 7.935736034317378e-05}
{"train_loss": 0.15158121287822723, "global_step": 81853, "epoch": 919, "lr": 7.935689105591193e-05}
{"train_loss": 0.2328522801399231, "global_step": 81854, "epoch": 919, "lr": 7.935642176470342e-05}
{"train_loss": 0.3109675943851471, "global_step": 81855, "epoch": 919, "lr": 7.935595246954828e-05}
{"train_loss": 0.15014240145683289, "global_step": 81856, "epoch": 919, "lr": 7.935548317044659e-05}
{"train_loss": 0.1194685623049736, "global_step": 81857, "epoch": 919, "lr": 7.935501386739843e-05}
{"train_loss": 0.28158804774284363, "global_step": 81858, "epoch": 919, "lr": 7.935454456040384e-05}
{"train_loss": 0.1931755691766739, "global_step": 81859, "epoch": 919, "lr": 7.935407524946288e-05}
{"train_loss": 0.08397981524467468, "global_step": 81860, "epoch": 919, "lr": 7.935360593457564e-05}
{"train_loss": 0.1820555180311203, "global_step": 81861, "epoch": 919, "lr": 7.935313661574215e-05}
{"train_loss": 0.23244817554950714, "global_step": 81862, "epoch": 919, "lr": 7.935266729296249e-05}
{"train_loss": 0.20535196363925934, "global_step": 81863, "epoch": 919, "lr": 7.935219796623671e-05}
{"train_loss": 0.19056034088134766, "global_step": 81864, "epoch": 919, "lr": 7.935172863556491e-05}
{"train_loss": 0.20133653283119202, "global_step": 81865, "epoch": 919, "lr": 7.935125930094711e-05}
{"train_loss": 0.22548049688339233, "global_step": 81866, "epoch": 919, "lr": 7.935078996238341e-05}
{"train_loss": 0.2550175189971924, "global_step": 81867, "epoch": 919, "lr": 7.935032061987383e-05}
{"train_loss": 0.23386970162391663, "global_step": 81868, "epoch": 919, "lr": 7.934985127341847e-05}
{"train_loss": 0.25400131940841675, "global_step": 81869, "epoch": 919, "lr": 7.934938192301738e-05}
{"train_loss": 0.12455693632364273, "global_step": 81870, "epoch": 919, "lr": 7.934891256867063e-05}
{"train_loss": 0.19766274094581604, "global_step": 81871, "epoch": 919, "lr": 7.934844321037827e-05}
{"train_loss": 0.24381564557552338, "global_step": 81872, "epoch": 919, "lr": 7.934797384814036e-05}
{"train_loss": 0.21730826795101166, "global_step": 81873, "epoch": 919, "lr": 7.934750448195699e-05}
{"train_loss": 0.18498758971691132, "global_step": 81874, "epoch": 919, "lr": 7.93470351118282e-05}
{"train_loss": 0.11924879252910614, "global_step": 81875, "epoch": 919, "lr": 7.934656573775406e-05}
{"train_loss": 0.16788463294506073, "global_step": 81876, "epoch": 919, "lr": 7.934609635973463e-05}
{"train_loss": 0.2091040164232254, "global_step": 81877, "epoch": 919, "lr": 7.934562697776997e-05}
{"train_loss": 0.2327224314212799, "global_step": 81878, "epoch": 919, "lr": 7.934515759186014e-05}
{"train_loss": 0.18551136595144702, "global_step": 81879, "epoch": 919, "lr": 7.934468820200522e-05, "val_loss": 3.9832632541656494}
{"train_loss": 0.24755950272083282, "global_step": 81880, "epoch": 920, "lr": 7.934421880820527e-05}
{"train_loss": 0.3099112808704376, "global_step": 81881, "epoch": 920, "lr": 7.934374941046034e-05}
{"train_loss": 0.13007256388664246, "global_step": 81882, "epoch": 920, "lr": 7.93432800087705e-05}
{"train_loss": 0.22953912615776062, "global_step": 81883, "epoch": 920, "lr": 7.934281060313582e-05}
{"train_loss": 0.17555975914001465, "global_step": 81884, "epoch": 920, "lr": 7.934234119355636e-05}
{"train_loss": 0.20834961533546448, "global_step": 81885, "epoch": 920, "lr": 7.934187178003217e-05}
{"train_loss": 0.18384529650211334, "global_step": 81886, "epoch": 920, "lr": 7.934140236256334e-05}
{"train_loss": 0.24573630094528198, "global_step": 81887, "epoch": 920, "lr": 7.934093294114991e-05}
{"train_loss": 0.15460222959518433, "global_step": 81888, "epoch": 920, "lr": 7.934046351579193e-05}
{"train_loss": 0.170684814453125, "global_step": 81889, "epoch": 920, "lr": 7.933999408648948e-05}
{"train_loss": 0.2418084591627121, "global_step": 81890, "epoch": 920, "lr": 7.933952465324267e-05}
{"train_loss": 0.2043934315443039, "global_step": 81891, "epoch": 920, "lr": 7.933905521605148e-05}
{"train_loss": 0.1471775472164154, "global_step": 81892, "epoch": 920, "lr": 7.933858577491602e-05}
{"train_loss": 0.2164139747619629, "global_step": 81893, "epoch": 920, "lr": 7.933811632983634e-05}
{"train_loss": 0.18045753240585327, "global_step": 81894, "epoch": 920, "lr": 7.933764688081251e-05}
{"train_loss": 0.17468278110027313, "global_step": 81895, "epoch": 920, "lr": 7.933717742784461e-05}
{"train_loss": 0.27734705805778503, "global_step": 81896, "epoch": 920, "lr": 7.933670797093267e-05}
{"train_loss": 0.16566835343837738, "global_step": 81897, "epoch": 920, "lr": 7.933623851007676e-05}
{"train_loss": 0.19758827984333038, "global_step": 81898, "epoch": 920, "lr": 7.933576904527695e-05}
{"train_loss": 0.19502268731594086, "global_step": 81899, "epoch": 920, "lr": 7.933529957653331e-05}
{"train_loss": 0.20944838225841522, "global_step": 81900, "epoch": 920, "lr": 7.933483010384589e-05}
{"train_loss": 0.17666201293468475, "global_step": 81901, "epoch": 920, "lr": 7.933436062721477e-05}
{"train_loss": 0.14607326686382294, "global_step": 81902, "epoch": 920, "lr": 7.933389114664001e-05}
{"train_loss": 0.17184530198574066, "global_step": 81903, "epoch": 920, "lr": 7.933342166212165e-05}
{"train_loss": 0.19557012617588043, "global_step": 81904, "epoch": 920, "lr": 7.933295217365978e-05}
{"train_loss": 0.32246559858322144, "global_step": 81905, "epoch": 920, "lr": 7.933248268125443e-05}
{"train_loss": 0.25779062509536743, "global_step": 81906, "epoch": 920, "lr": 7.93320131849057e-05}
{"train_loss": 0.2556400001049042, "global_step": 81907, "epoch": 920, "lr": 7.933154368461364e-05}
{"train_loss": 0.2052166759967804, "global_step": 81908, "epoch": 920, "lr": 7.93310741803783e-05}
{"train_loss": 0.3019992709159851, "global_step": 81909, "epoch": 920, "lr": 7.933060467219978e-05}
{"train_loss": 0.32677245140075684, "global_step": 81910, "epoch": 920, "lr": 7.93301351600781e-05}
{"train_loss": 0.23961786925792694, "global_step": 81911, "epoch": 920, "lr": 7.932966564401334e-05}
{"train_loss": 0.19221360981464386, "global_step": 81912, "epoch": 920, "lr": 7.932919612400558e-05}
{"train_loss": 0.30224794149398804, "global_step": 81913, "epoch": 920, "lr": 7.932872660005483e-05}
{"train_loss": 0.26965248584747314, "global_step": 81914, "epoch": 920, "lr": 7.932825707216123e-05}
{"train_loss": 0.1978921890258789, "global_step": 81915, "epoch": 920, "lr": 7.932778754032479e-05}
{"train_loss": 0.2905869781970978, "global_step": 81916, "epoch": 920, "lr": 7.932731800454558e-05}
{"train_loss": 0.23019365966320038, "global_step": 81917, "epoch": 920, "lr": 7.932684846482367e-05}
{"train_loss": 0.2239682823419571, "global_step": 81918, "epoch": 920, "lr": 7.932637892115913e-05}
{"train_loss": 0.2064933180809021, "global_step": 81919, "epoch": 920, "lr": 7.932590937355201e-05}
{"train_loss": 0.13956698775291443, "global_step": 81920, "epoch": 920, "lr": 7.932543982200238e-05}
{"train_loss": 0.22090891003608704, "global_step": 81921, "epoch": 920, "lr": 7.93249702665103e-05}
{"train_loss": 0.23022261261940002, "global_step": 81922, "epoch": 920, "lr": 7.932450070707585e-05}
{"train_loss": 0.21143345534801483, "global_step": 81923, "epoch": 920, "lr": 7.932403114369906e-05}
{"train_loss": 0.1910650134086609, "global_step": 81924, "epoch": 920, "lr": 7.932356157638002e-05}
{"train_loss": 0.228369802236557, "global_step": 81925, "epoch": 920, "lr": 7.932309200511879e-05}
{"train_loss": 0.1445891559123993, "global_step": 81926, "epoch": 920, "lr": 7.93226224299154e-05}
{"train_loss": 0.18026776611804962, "global_step": 81927, "epoch": 920, "lr": 7.932215285076997e-05}
{"train_loss": 0.2046775370836258, "global_step": 81928, "epoch": 920, "lr": 7.932168326768253e-05}
{"train_loss": 0.19588512182235718, "global_step": 81929, "epoch": 920, "lr": 7.932121368065314e-05}
{"train_loss": 0.2518550455570221, "global_step": 81930, "epoch": 920, "lr": 7.932074408968186e-05}
{"train_loss": 0.15626107156276703, "global_step": 81931, "epoch": 920, "lr": 7.932027449476879e-05}
{"train_loss": 0.15703924000263214, "global_step": 81932, "epoch": 920, "lr": 7.931980489591395e-05}
{"train_loss": 0.21783053874969482, "global_step": 81933, "epoch": 920, "lr": 7.931933529311741e-05}
{"train_loss": 0.18022607266902924, "global_step": 81934, "epoch": 920, "lr": 7.931886568637927e-05}
{"train_loss": 0.2195882797241211, "global_step": 81935, "epoch": 920, "lr": 7.931839607569953e-05}
{"train_loss": 0.13898824155330658, "global_step": 81936, "epoch": 920, "lr": 7.931792646107832e-05}
{"train_loss": 0.2187482863664627, "global_step": 81937, "epoch": 920, "lr": 7.931745684251565e-05}
{"train_loss": 0.2177698016166687, "global_step": 81938, "epoch": 920, "lr": 7.931698722001163e-05}
{"train_loss": 0.14680998027324677, "global_step": 81939, "epoch": 920, "lr": 7.931651759356628e-05}
{"train_loss": 0.2361978441476822, "global_step": 81940, "epoch": 920, "lr": 7.931604796317969e-05}
{"train_loss": 0.1785883754491806, "global_step": 81941, "epoch": 920, "lr": 7.93155783288519e-05}
{"train_loss": 0.20727847516536713, "global_step": 81942, "epoch": 920, "lr": 7.9315108690583e-05}
{"train_loss": 0.26129069924354553, "global_step": 81943, "epoch": 920, "lr": 7.931463904837305e-05}
{"train_loss": 0.19167792797088623, "global_step": 81944, "epoch": 920, "lr": 7.931416940222209e-05}
{"train_loss": 0.18153958022594452, "global_step": 81945, "epoch": 920, "lr": 7.931369975213021e-05}
{"train_loss": 0.13354438543319702, "global_step": 81946, "epoch": 920, "lr": 7.931323009809744e-05}
{"train_loss": 0.22291983664035797, "global_step": 81947, "epoch": 920, "lr": 7.931276044012388e-05}
{"train_loss": 0.17242878675460815, "global_step": 81948, "epoch": 920, "lr": 7.931229077820958e-05}
{"train_loss": 0.14367859065532684, "global_step": 81949, "epoch": 920, "lr": 7.931182111235458e-05}
{"train_loss": 0.16517284512519836, "global_step": 81950, "epoch": 920, "lr": 7.931135144255899e-05}
{"train_loss": 0.1607491821050644, "global_step": 81951, "epoch": 920, "lr": 7.931088176882282e-05}
{"train_loss": 0.24256542325019836, "global_step": 81952, "epoch": 920, "lr": 7.931041209114618e-05}
{"train_loss": 0.15189220011234283, "global_step": 81953, "epoch": 920, "lr": 7.930994240952909e-05}
{"train_loss": 0.23659737408161163, "global_step": 81954, "epoch": 920, "lr": 7.930947272397164e-05}
{"train_loss": 0.21918347477912903, "global_step": 81955, "epoch": 920, "lr": 7.93090030344739e-05}
{"train_loss": 0.13948126137256622, "global_step": 81956, "epoch": 920, "lr": 7.930853334103592e-05}
{"train_loss": 0.17918755114078522, "global_step": 81957, "epoch": 920, "lr": 7.930806364365777e-05}
{"train_loss": 0.18358759582042694, "global_step": 81958, "epoch": 920, "lr": 7.930759394233949e-05}
{"train_loss": 0.1458464413881302, "global_step": 81959, "epoch": 920, "lr": 7.930712423708117e-05}
{"train_loss": 0.231268510222435, "global_step": 81960, "epoch": 920, "lr": 7.930665452788288e-05}
{"train_loss": 0.18597157299518585, "global_step": 81961, "epoch": 920, "lr": 7.930618481474465e-05}
{"train_loss": 0.15469573438167572, "global_step": 81962, "epoch": 920, "lr": 7.930571509766656e-05}
{"train_loss": 0.15892983973026276, "global_step": 81963, "epoch": 920, "lr": 7.930524537664867e-05}
{"train_loss": 0.13703098893165588, "global_step": 81964, "epoch": 920, "lr": 7.930477565169107e-05}
{"train_loss": 0.22309303283691406, "global_step": 81965, "epoch": 920, "lr": 7.930430592279378e-05}
{"train_loss": 0.17608408629894257, "global_step": 81966, "epoch": 920, "lr": 7.93038361899569e-05}
{"train_loss": 0.2061009556055069, "global_step": 81967, "epoch": 920, "lr": 7.930336645318045e-05}
{"train_loss": 0.20295298317175234, "global_step": 81968, "epoch": 920, "lr": 7.930289671246453e-05, "val_loss": 4.085367679595947, "train_action_mse_error": 29.89128875732422}
{"train_loss": 0.18598008155822754, "global_step": 81969, "epoch": 921, "lr": 7.93024269678092e-05}
{"train_loss": 0.19062526524066925, "global_step": 81970, "epoch": 921, "lr": 7.930195721921452e-05}
{"train_loss": 0.2050139605998993, "global_step": 81971, "epoch": 921, "lr": 7.930148746668054e-05}
{"train_loss": 0.10351699590682983, "global_step": 81972, "epoch": 921, "lr": 7.930101771020733e-05}
{"train_loss": 0.2534225881099701, "global_step": 81973, "epoch": 921, "lr": 7.930054794979496e-05}
{"train_loss": 0.18303634226322174, "global_step": 81974, "epoch": 921, "lr": 7.930007818544349e-05}
{"train_loss": 0.20329245924949646, "global_step": 81975, "epoch": 921, "lr": 7.929960841715298e-05}
{"train_loss": 0.16691738367080688, "global_step": 81976, "epoch": 921, "lr": 7.929913864492349e-05}
{"train_loss": 0.23276387155056, "global_step": 81977, "epoch": 921, "lr": 7.92986688687551e-05}
{"train_loss": 0.15223820507526398, "global_step": 81978, "epoch": 921, "lr": 7.929819908864786e-05}
{"train_loss": 0.2592256963253021, "global_step": 81979, "epoch": 921, "lr": 7.929772930460181e-05}
{"train_loss": 0.2407689392566681, "global_step": 81980, "epoch": 921, "lr": 7.929725951661707e-05}
{"train_loss": 0.15567660331726074, "global_step": 81981, "epoch": 921, "lr": 7.929678972469366e-05}
{"train_loss": 0.1483439803123474, "global_step": 81982, "epoch": 921, "lr": 7.929631992883164e-05}
{"train_loss": 0.16980671882629395, "global_step": 81983, "epoch": 921, "lr": 7.929585012903109e-05}
{"train_loss": 0.20136286318302155, "global_step": 81984, "epoch": 921, "lr": 7.929538032529209e-05}
{"train_loss": 0.18906494975090027, "global_step": 81985, "epoch": 921, "lr": 7.929491051761466e-05}
{"train_loss": 0.2733723819255829, "global_step": 81986, "epoch": 921, "lr": 7.929444070599889e-05}
{"train_loss": 0.152801513671875, "global_step": 81987, "epoch": 921, "lr": 7.929397089044485e-05}
{"train_loss": 0.25060102343559265, "global_step": 81988, "epoch": 921, "lr": 7.92935010709526e-05}
{"train_loss": 0.16926288604736328, "global_step": 81989, "epoch": 921, "lr": 7.929303124752216e-05}
{"train_loss": 0.15316560864448547, "global_step": 81990, "epoch": 921, "lr": 7.929256142015365e-05}
{"train_loss": 0.2571287155151367, "global_step": 81991, "epoch": 921, "lr": 7.929209158884711e-05}
{"train_loss": 0.11457384377717972, "global_step": 81992, "epoch": 921, "lr": 7.929162175360262e-05}
{"train_loss": 0.21426264941692352, "global_step": 81993, "epoch": 921, "lr": 7.929115191442022e-05}
{"train_loss": 0.22990331053733826, "global_step": 81994, "epoch": 921, "lr": 7.929068207129999e-05}
{"train_loss": 0.1870717704296112, "global_step": 81995, "epoch": 921, "lr": 7.929021222424197e-05}
{"train_loss": 0.19468560814857483, "global_step": 81996, "epoch": 921, "lr": 7.928974237324624e-05}
{"train_loss": 0.1032637506723404, "global_step": 81997, "epoch": 921, "lr": 7.928927251831287e-05}
{"train_loss": 0.1937636435031891, "global_step": 81998, "epoch": 921, "lr": 7.92888026594419e-05}
{"train_loss": 0.20207622647285461, "global_step": 81999, "epoch": 921, "lr": 7.928833279663341e-05}
{"train_loss": 0.2213318943977356, "global_step": 82000, "epoch": 921, "lr": 7.928786292988747e-05}
{"train_loss": 0.21094483137130737, "global_step": 82001, "epoch": 921, "lr": 7.928739305920414e-05}
{"train_loss": 0.21079903841018677, "global_step": 82002, "epoch": 921, "lr": 7.928692318458347e-05}
{"train_loss": 0.18905313313007355, "global_step": 82003, "epoch": 921, "lr": 7.928645330602551e-05}
{"train_loss": 0.20288974046707153, "global_step": 82004, "epoch": 921, "lr": 7.928598342353038e-05}
{"train_loss": 0.17305654287338257, "global_step": 82005, "epoch": 921, "lr": 7.928551353709808e-05}
{"train_loss": 0.15609212219715118, "global_step": 82006, "epoch": 921, "lr": 7.928504364672871e-05}
{"train_loss": 0.17747223377227783, "global_step": 82007, "epoch": 921, "lr": 7.92845737524223e-05}
{"train_loss": 0.19027236104011536, "global_step": 82008, "epoch": 921, "lr": 7.928410385417895e-05}
{"train_loss": 0.17807629704475403, "global_step": 82009, "epoch": 921, "lr": 7.928363395199873e-05}
{"train_loss": 0.2612308859825134, "global_step": 82010, "epoch": 921, "lr": 7.928316404588166e-05}
{"train_loss": 0.14274567365646362, "global_step": 82011, "epoch": 921, "lr": 7.928269413582783e-05}
{"train_loss": 0.19577954709529877, "global_step": 82012, "epoch": 921, "lr": 7.928222422183731e-05}
{"train_loss": 0.24089637398719788, "global_step": 82013, "epoch": 921, "lr": 7.928175430391013e-05}
{"train_loss": 0.1917419135570526, "global_step": 82014, "epoch": 921, "lr": 7.928128438204641e-05}
{"train_loss": 0.18300321698188782, "global_step": 82015, "epoch": 921, "lr": 7.928081445624615e-05}
{"train_loss": 0.194930762052536, "global_step": 82016, "epoch": 921, "lr": 7.928034452650944e-05}
{"train_loss": 0.1620808243751526, "global_step": 82017, "epoch": 921, "lr": 7.927987459283636e-05}
{"train_loss": 0.34966549277305603, "global_step": 82018, "epoch": 921, "lr": 7.927940465522694e-05}
{"train_loss": 0.22882415354251862, "global_step": 82019, "epoch": 921, "lr": 7.927893471368127e-05}
{"train_loss": 0.18057958781719208, "global_step": 82020, "epoch": 921, "lr": 7.927846476819941e-05}
{"train_loss": 0.1551763117313385, "global_step": 82021, "epoch": 921, "lr": 7.927799481878142e-05}
{"train_loss": 0.09274929761886597, "global_step": 82022, "epoch": 921, "lr": 7.927752486542735e-05}
{"train_loss": 0.18712575733661652, "global_step": 82023, "epoch": 921, "lr": 7.927705490813728e-05}
{"train_loss": 0.1582227349281311, "global_step": 82024, "epoch": 921, "lr": 7.927658494691125e-05}
{"train_loss": 0.17271068692207336, "global_step": 82025, "epoch": 921, "lr": 7.927611498174936e-05}
{"train_loss": 0.09951242059469223, "global_step": 82026, "epoch": 921, "lr": 7.927564501265164e-05}
{"train_loss": 0.1587374359369278, "global_step": 82027, "epoch": 921, "lr": 7.927517503961816e-05}
{"train_loss": 0.2639997899532318, "global_step": 82028, "epoch": 921, "lr": 7.927470506264901e-05}
{"train_loss": 0.15616874396800995, "global_step": 82029, "epoch": 921, "lr": 7.927423508174423e-05}
{"train_loss": 0.21276311576366425, "global_step": 82030, "epoch": 921, "lr": 7.927376509690388e-05}
{"train_loss": 0.20462454855442047, "global_step": 82031, "epoch": 921, "lr": 7.927329510812802e-05}
{"train_loss": 0.16755235195159912, "global_step": 82032, "epoch": 921, "lr": 7.927282511541674e-05}
{"train_loss": 0.18211247026920319, "global_step": 82033, "epoch": 921, "lr": 7.927235511877008e-05}
{"train_loss": 0.21499964594841003, "global_step": 82034, "epoch": 921, "lr": 7.92718851181881e-05}
{"train_loss": 0.16144230961799622, "global_step": 82035, "epoch": 921, "lr": 7.927141511367089e-05}
{"train_loss": 0.2827112674713135, "global_step": 82036, "epoch": 921, "lr": 7.927094510521848e-05}
{"train_loss": 0.18789774179458618, "global_step": 82037, "epoch": 921, "lr": 7.927047509283094e-05}
{"train_loss": 0.3143749237060547, "global_step": 82038, "epoch": 921, "lr": 7.927000507650836e-05}
{"train_loss": 0.2620449662208557, "global_step": 82039, "epoch": 921, "lr": 7.926953505625077e-05}
{"train_loss": 0.17717047035694122, "global_step": 82040, "epoch": 921, "lr": 7.926906503205826e-05}
{"train_loss": 0.20156876742839813, "global_step": 82041, "epoch": 921, "lr": 7.926859500393088e-05}
{"train_loss": 0.16215074062347412, "global_step": 82042, "epoch": 921, "lr": 7.926812497186868e-05}
{"train_loss": 0.19261521100997925, "global_step": 82043, "epoch": 921, "lr": 7.926765493587176e-05}
{"train_loss": 0.19247101247310638, "global_step": 82044, "epoch": 921, "lr": 7.926718489594014e-05}
{"train_loss": 0.15917626023292542, "global_step": 82045, "epoch": 921, "lr": 7.926671485207391e-05}
{"train_loss": 0.1542961299419403, "global_step": 82046, "epoch": 921, "lr": 7.926624480427312e-05}
{"train_loss": 0.26698964834213257, "global_step": 82047, "epoch": 921, "lr": 7.926577475253786e-05}
{"train_loss": 0.20984287559986115, "global_step": 82048, "epoch": 921, "lr": 7.926530469686815e-05}
{"train_loss": 0.13266484439373016, "global_step": 82049, "epoch": 921, "lr": 7.926483463726409e-05}
{"train_loss": 0.14966420829296112, "global_step": 82050, "epoch": 921, "lr": 7.926436457372573e-05}
{"train_loss": 0.19211050868034363, "global_step": 82051, "epoch": 921, "lr": 7.926389450625313e-05}
{"train_loss": 0.1881769597530365, "global_step": 82052, "epoch": 921, "lr": 7.926342443484636e-05}
{"train_loss": 0.1330775022506714, "global_step": 82053, "epoch": 921, "lr": 7.926295435950549e-05}
{"train_loss": 0.13667748868465424, "global_step": 82054, "epoch": 921, "lr": 7.926248428023056e-05}
{"train_loss": 0.18087969720363617, "global_step": 82055, "epoch": 921, "lr": 7.926201419702164e-05}
{"train_loss": 0.16745729744434357, "global_step": 82056, "epoch": 921, "lr": 7.92615441098788e-05}
{"train_loss": 0.19091477839464552, "global_step": 82057, "epoch": 921, "lr": 7.92610740188021e-05, "val_loss": 4.177279949188232}
{"train_loss": 0.1519525647163391, "global_step": 82058, "epoch": 922, "lr": 7.926060392379161e-05}
{"train_loss": 0.14601904153823853, "global_step": 82059, "epoch": 922, "lr": 7.92601338248474e-05}
{"train_loss": 0.16558392345905304, "global_step": 82060, "epoch": 922, "lr": 7.92596637219695e-05}
{"train_loss": 0.20657682418823242, "global_step": 82061, "epoch": 922, "lr": 7.925919361515801e-05}
{"train_loss": 0.2530229091644287, "global_step": 82062, "epoch": 922, "lr": 7.925872350441296e-05}
{"train_loss": 0.20844469964504242, "global_step": 82063, "epoch": 922, "lr": 7.925825338973445e-05}
{"train_loss": 0.19086408615112305, "global_step": 82064, "epoch": 922, "lr": 7.925778327112253e-05}
{"train_loss": 0.1433507204055786, "global_step": 82065, "epoch": 922, "lr": 7.925731314857722e-05}
{"train_loss": 0.19037090241909027, "global_step": 82066, "epoch": 922, "lr": 7.925684302209866e-05}
{"train_loss": 0.17004752159118652, "global_step": 82067, "epoch": 922, "lr": 7.925637289168686e-05}
{"train_loss": 0.2673344016075134, "global_step": 82068, "epoch": 922, "lr": 7.92559027573419e-05}
{"train_loss": 0.16042599081993103, "global_step": 82069, "epoch": 922, "lr": 7.925543261906384e-05}
{"train_loss": 0.22375546395778656, "global_step": 82070, "epoch": 922, "lr": 7.925496247685274e-05}
{"train_loss": 0.21149219572544098, "global_step": 82071, "epoch": 922, "lr": 7.925449233070867e-05}
{"train_loss": 0.17704397439956665, "global_step": 82072, "epoch": 922, "lr": 7.925402218063169e-05}
{"train_loss": 0.19264639914035797, "global_step": 82073, "epoch": 922, "lr": 7.925355202662186e-05}
{"train_loss": 0.2157425880432129, "global_step": 82074, "epoch": 922, "lr": 7.925308186867926e-05}
{"train_loss": 0.1567307561635971, "global_step": 82075, "epoch": 922, "lr": 7.925261170680394e-05}
{"train_loss": 0.1944536715745926, "global_step": 82076, "epoch": 922, "lr": 7.925214154099594e-05}
{"train_loss": 0.17908312380313873, "global_step": 82077, "epoch": 922, "lr": 7.925167137125536e-05}
{"train_loss": 0.22192682325839996, "global_step": 82078, "epoch": 922, "lr": 7.925120119758225e-05}
{"train_loss": 0.14130234718322754, "global_step": 82079, "epoch": 922, "lr": 7.925073101997666e-05}
{"train_loss": 0.19279618561267853, "global_step": 82080, "epoch": 922, "lr": 7.925026083843867e-05}
{"train_loss": 0.14207085967063904, "global_step": 82081, "epoch": 922, "lr": 7.924979065296836e-05}
{"train_loss": 0.1614665538072586, "global_step": 82082, "epoch": 922, "lr": 7.924932046356574e-05}
{"train_loss": 0.2438519150018692, "global_step": 82083, "epoch": 922, "lr": 7.924885027023093e-05}
{"train_loss": 0.2281811684370041, "global_step": 82084, "epoch": 922, "lr": 7.924838007296396e-05}
{"train_loss": 0.1866316795349121, "global_step": 82085, "epoch": 922, "lr": 7.924790987176491e-05}
{"train_loss": 0.1772328019142151, "global_step": 82086, "epoch": 922, "lr": 7.924743966663381e-05}
{"train_loss": 0.1808585673570633, "global_step": 82087, "epoch": 922, "lr": 7.924696945757077e-05}
{"train_loss": 0.2609137296676636, "global_step": 82088, "epoch": 922, "lr": 7.924649924457584e-05}
{"train_loss": 0.16381925344467163, "global_step": 82089, "epoch": 922, "lr": 7.924602902764907e-05}
{"train_loss": 0.17939144372940063, "global_step": 82090, "epoch": 922, "lr": 7.92455588067905e-05}
{"train_loss": 0.15361741185188293, "global_step": 82091, "epoch": 922, "lr": 7.924508858200025e-05}
{"train_loss": 0.28039413690567017, "global_step": 82092, "epoch": 922, "lr": 7.924461835327835e-05}
{"train_loss": 0.2566815912723541, "global_step": 82093, "epoch": 922, "lr": 7.924414812062486e-05}
{"train_loss": 0.2641662061214447, "global_step": 82094, "epoch": 922, "lr": 7.924367788403986e-05}
{"train_loss": 0.19453436136245728, "global_step": 82095, "epoch": 922, "lr": 7.924320764352339e-05}
{"train_loss": 0.13719606399536133, "global_step": 82096, "epoch": 922, "lr": 7.924273739907553e-05}
{"train_loss": 0.1669255644083023, "global_step": 82097, "epoch": 922, "lr": 7.924226715069635e-05}
{"train_loss": 0.13836319744586945, "global_step": 82098, "epoch": 922, "lr": 7.924179689838588e-05}
{"train_loss": 0.14393074810504913, "global_step": 82099, "epoch": 922, "lr": 7.924132664214424e-05}
{"train_loss": 0.19790972769260406, "global_step": 82100, "epoch": 922, "lr": 7.924085638197143e-05}
{"train_loss": 0.14415477216243744, "global_step": 82101, "epoch": 922, "lr": 7.924038611786757e-05}
{"train_loss": 0.24694447219371796, "global_step": 82102, "epoch": 922, "lr": 7.923991584983268e-05}
{"train_loss": 0.20368751883506775, "global_step": 82103, "epoch": 922, "lr": 7.923944557786684e-05}
{"train_loss": 0.1904715746641159, "global_step": 82104, "epoch": 922, "lr": 7.923897530197013e-05}
{"train_loss": 0.25479069352149963, "global_step": 82105, "epoch": 922, "lr": 7.923850502214258e-05}
{"train_loss": 0.2465313971042633, "global_step": 82106, "epoch": 922, "lr": 7.923803473838426e-05}
{"train_loss": 0.17771896719932556, "global_step": 82107, "epoch": 922, "lr": 7.923756445069526e-05}
{"train_loss": 0.2444688230752945, "global_step": 82108, "epoch": 922, "lr": 7.923709415907562e-05}
{"train_loss": 0.1759359985589981, "global_step": 82109, "epoch": 922, "lr": 7.923662386352541e-05}
{"train_loss": 0.2228037565946579, "global_step": 82110, "epoch": 922, "lr": 7.92361535640447e-05}
{"train_loss": 0.19891250133514404, "global_step": 82111, "epoch": 922, "lr": 7.923568326063355e-05}
{"train_loss": 0.15182964503765106, "global_step": 82112, "epoch": 922, "lr": 7.9235212953292e-05}
{"train_loss": 0.23194772005081177, "global_step": 82113, "epoch": 922, "lr": 7.923474264202013e-05}
{"train_loss": 0.18924596905708313, "global_step": 82114, "epoch": 922, "lr": 7.923427232681803e-05}
{"train_loss": 0.16273567080497742, "global_step": 82115, "epoch": 922, "lr": 7.92338020076857e-05}
{"train_loss": 0.13868778944015503, "global_step": 82116, "epoch": 922, "lr": 7.923333168462328e-05}
{"train_loss": 0.265838623046875, "global_step": 82117, "epoch": 922, "lr": 7.923286135763077e-05}
{"train_loss": 0.23168811202049255, "global_step": 82118, "epoch": 922, "lr": 7.923239102670828e-05}
{"train_loss": 0.18567734956741333, "global_step": 82119, "epoch": 922, "lr": 7.923192069185583e-05}
{"train_loss": 0.13468320667743683, "global_step": 82120, "epoch": 922, "lr": 7.92314503530735e-05}
{"train_loss": 0.1175922155380249, "global_step": 82121, "epoch": 922, "lr": 7.923098001036138e-05}
{"train_loss": 0.1657196432352066, "global_step": 82122, "epoch": 922, "lr": 7.92305096637195e-05}
{"train_loss": 0.2832950949668884, "global_step": 82123, "epoch": 922, "lr": 7.923003931314794e-05}
{"train_loss": 0.19011835753917694, "global_step": 82124, "epoch": 922, "lr": 7.922956895864675e-05}
{"train_loss": 0.2523908019065857, "global_step": 82125, "epoch": 922, "lr": 7.922909860021601e-05}
{"train_loss": 0.20764745771884918, "global_step": 82126, "epoch": 922, "lr": 7.922862823785577e-05}
{"train_loss": 0.1968686431646347, "global_step": 82127, "epoch": 922, "lr": 7.92281578715661e-05}
{"train_loss": 0.18152323365211487, "global_step": 82128, "epoch": 922, "lr": 7.922768750134705e-05}
{"train_loss": 0.1753266304731369, "global_step": 82129, "epoch": 922, "lr": 7.92272171271987e-05}
{"train_loss": 0.2349785715341568, "global_step": 82130, "epoch": 922, "lr": 7.92267467491211e-05}
{"train_loss": 0.16283947229385376, "global_step": 82131, "epoch": 922, "lr": 7.922627636711432e-05}
{"train_loss": 0.23354209959506989, "global_step": 82132, "epoch": 922, "lr": 7.922580598117843e-05}
{"train_loss": 0.14723113179206848, "global_step": 82133, "epoch": 922, "lr": 7.92253355913135e-05}
{"train_loss": 0.22558428347110748, "global_step": 82134, "epoch": 922, "lr": 7.922486519751957e-05}
{"train_loss": 0.24128198623657227, "global_step": 82135, "epoch": 922, "lr": 7.92243947997967e-05}
{"train_loss": 0.14282691478729248, "global_step": 82136, "epoch": 922, "lr": 7.922392439814497e-05}
{"train_loss": 0.26219886541366577, "global_step": 82137, "epoch": 922, "lr": 7.922345399256446e-05}
{"train_loss": 0.21617794036865234, "global_step": 82138, "epoch": 922, "lr": 7.92229835830552e-05}
{"train_loss": 0.19182248413562775, "global_step": 82139, "epoch": 922, "lr": 7.922251316961725e-05}
{"train_loss": 0.19884106516838074, "global_step": 82140, "epoch": 922, "lr": 7.922204275225071e-05}
{"train_loss": 0.16486941277980804, "global_step": 82141, "epoch": 922, "lr": 7.922157233095562e-05}
{"train_loss": 0.21513508260250092, "global_step": 82142, "epoch": 922, "lr": 7.922110190573203e-05}
{"train_loss": 0.17279988527297974, "global_step": 82143, "epoch": 922, "lr": 7.922063147658003e-05}
{"train_loss": 0.1810522824525833, "global_step": 82144, "epoch": 922, "lr": 7.922016104349968e-05}
{"train_loss": 0.1860254555940628, "global_step": 82145, "epoch": 922, "lr": 7.921969060649102e-05}
{"train_loss": 0.19469682487209192, "global_step": 82146, "epoch": 922, "lr": 7.921922016555414e-05, "val_loss": 4.069121837615967}
{"train_loss": 0.22377285361289978, "global_step": 82147, "epoch": 923, "lr": 7.92187497206891e-05}
{"train_loss": 0.1947016716003418, "global_step": 82148, "epoch": 923, "lr": 7.921827927189593e-05}
{"train_loss": 0.13115018606185913, "global_step": 82149, "epoch": 923, "lr": 7.921780881917473e-05}
{"train_loss": 0.19755125045776367, "global_step": 82150, "epoch": 923, "lr": 7.921733836252556e-05}
{"train_loss": 0.18025510013103485, "global_step": 82151, "epoch": 923, "lr": 7.921686790194847e-05}
{"train_loss": 0.16305045783519745, "global_step": 82152, "epoch": 923, "lr": 7.921639743744353e-05}
{"train_loss": 0.1714419573545456, "global_step": 82153, "epoch": 923, "lr": 7.92159269690108e-05}
{"train_loss": 0.1753879189491272, "global_step": 82154, "epoch": 923, "lr": 7.921545649665034e-05}
{"train_loss": 0.1623963564634323, "global_step": 82155, "epoch": 923, "lr": 7.921498602036222e-05}
{"train_loss": 0.1669841855764389, "global_step": 82156, "epoch": 923, "lr": 7.92145155401465e-05}
{"train_loss": 0.20384781062602997, "global_step": 82157, "epoch": 923, "lr": 7.921404505600324e-05}
{"train_loss": 0.23013009130954742, "global_step": 82158, "epoch": 923, "lr": 7.921357456793253e-05}
{"train_loss": 0.17614299058914185, "global_step": 82159, "epoch": 923, "lr": 7.921310407593438e-05}
{"train_loss": 0.21052691340446472, "global_step": 82160, "epoch": 923, "lr": 7.92126335800089e-05}
{"train_loss": 0.17861855030059814, "global_step": 82161, "epoch": 923, "lr": 7.921216308015615e-05}
{"train_loss": 0.1417854279279709, "global_step": 82162, "epoch": 923, "lr": 7.921169257637617e-05}
{"train_loss": 0.1189926341176033, "global_step": 82163, "epoch": 923, "lr": 7.921122206866902e-05}
{"train_loss": 0.1907912641763687, "global_step": 82164, "epoch": 923, "lr": 7.92107515570348e-05}
{"train_loss": 0.13025003671646118, "global_step": 82165, "epoch": 923, "lr": 7.921028104147355e-05}
{"train_loss": 0.22655634582042694, "global_step": 82166, "epoch": 923, "lr": 7.920981052198531e-05}
{"train_loss": 0.17062191665172577, "global_step": 82167, "epoch": 923, "lr": 7.920933999857018e-05}
{"train_loss": 0.1789465844631195, "global_step": 82168, "epoch": 923, "lr": 7.920886947122821e-05}
{"train_loss": 0.24669522047042847, "global_step": 82169, "epoch": 923, "lr": 7.920839893995947e-05}
{"train_loss": 0.1943880170583725, "global_step": 82170, "epoch": 923, "lr": 7.920792840476401e-05}
{"train_loss": 0.2229093760251999, "global_step": 82171, "epoch": 923, "lr": 7.92074578656419e-05}
{"train_loss": 0.22449488937854767, "global_step": 82172, "epoch": 923, "lr": 7.920698732259319e-05}
{"train_loss": 0.24814540147781372, "global_step": 82173, "epoch": 923, "lr": 7.920651677561799e-05}
{"train_loss": 0.24542321264743805, "global_step": 82174, "epoch": 923, "lr": 7.92060462247163e-05}
{"train_loss": 0.22718514502048492, "global_step": 82175, "epoch": 923, "lr": 7.920557566988824e-05}
{"train_loss": 0.23145876824855804, "global_step": 82176, "epoch": 923, "lr": 7.920510511113381e-05}
{"train_loss": 0.20245328545570374, "global_step": 82177, "epoch": 923, "lr": 7.920463454845315e-05}
{"train_loss": 0.12562531232833862, "global_step": 82178, "epoch": 923, "lr": 7.920416398184626e-05}
{"train_loss": 0.23075075447559357, "global_step": 82179, "epoch": 923, "lr": 7.920369341131321e-05}
{"train_loss": 0.16179022192955017, "global_step": 82180, "epoch": 923, "lr": 7.92032228368541e-05}
{"train_loss": 0.17752902209758759, "global_step": 82181, "epoch": 923, "lr": 7.920275225846898e-05}
{"train_loss": 0.16628113389015198, "global_step": 82182, "epoch": 923, "lr": 7.920228167615791e-05}
{"train_loss": 0.21247465908527374, "global_step": 82183, "epoch": 923, "lr": 7.920181108992092e-05}
{"train_loss": 0.2184487283229828, "global_step": 82184, "epoch": 923, "lr": 7.920134049975813e-05}
{"train_loss": 0.19899685680866241, "global_step": 82185, "epoch": 923, "lr": 7.920086990566955e-05}
{"train_loss": 0.3400523364543915, "global_step": 82186, "epoch": 923, "lr": 7.920039930765529e-05}
{"train_loss": 0.26356157660484314, "global_step": 82187, "epoch": 923, "lr": 7.919992870571538e-05}
{"train_loss": 0.20149776339530945, "global_step": 82188, "epoch": 923, "lr": 7.919945809984991e-05}
{"train_loss": 0.25513508915901184, "global_step": 82189, "epoch": 923, "lr": 7.919898749005892e-05}
{"train_loss": 0.19354155659675598, "global_step": 82190, "epoch": 923, "lr": 7.919851687634248e-05}
{"train_loss": 0.17392580211162567, "global_step": 82191, "epoch": 923, "lr": 7.919804625870064e-05}
{"train_loss": 0.22536306083202362, "global_step": 82192, "epoch": 923, "lr": 7.919757563713351e-05}
{"train_loss": 0.1598769575357437, "global_step": 82193, "epoch": 923, "lr": 7.91971050116411e-05}
{"train_loss": 0.16518771648406982, "global_step": 82194, "epoch": 923, "lr": 7.919663438222352e-05}
{"train_loss": 0.2662743031978607, "global_step": 82195, "epoch": 923, "lr": 7.919616374888078e-05}
{"train_loss": 0.18681032955646515, "global_step": 82196, "epoch": 923, "lr": 7.919569311161297e-05}
{"train_loss": 0.1942986398935318, "global_step": 82197, "epoch": 923, "lr": 7.919522247042017e-05}
{"train_loss": 0.23821675777435303, "global_step": 82198, "epoch": 923, "lr": 7.919475182530243e-05}
{"train_loss": 0.18856695294380188, "global_step": 82199, "epoch": 923, "lr": 7.919428117625981e-05}
{"train_loss": 0.2096344530582428, "global_step": 82200, "epoch": 923, "lr": 7.919381052329238e-05}
{"train_loss": 0.12122945487499237, "global_step": 82201, "epoch": 923, "lr": 7.919333986640019e-05}
{"train_loss": 0.2031717598438263, "global_step": 82202, "epoch": 923, "lr": 7.91928692055833e-05}
{"train_loss": 0.2782827317714691, "global_step": 82203, "epoch": 923, "lr": 7.919239854084179e-05}
{"train_loss": 0.176841601729393, "global_step": 82204, "epoch": 923, "lr": 7.919192787217573e-05}
{"train_loss": 0.11803609132766724, "global_step": 82205, "epoch": 923, "lr": 7.919145719958517e-05}
{"train_loss": 0.24375750124454498, "global_step": 82206, "epoch": 923, "lr": 7.919098652307015e-05}
{"train_loss": 0.20329442620277405, "global_step": 82207, "epoch": 923, "lr": 7.919051584263078e-05}
{"train_loss": 0.2251916378736496, "global_step": 82208, "epoch": 923, "lr": 7.91900451582671e-05}
{"train_loss": 0.32801496982574463, "global_step": 82209, "epoch": 923, "lr": 7.918957446997917e-05}
{"train_loss": 0.1619294285774231, "global_step": 82210, "epoch": 923, "lr": 7.918910377776706e-05}
{"train_loss": 0.17824862897396088, "global_step": 82211, "epoch": 923, "lr": 7.918863308163082e-05}
{"train_loss": 0.19724467396736145, "global_step": 82212, "epoch": 923, "lr": 7.918816238157054e-05}
{"train_loss": 0.21101757884025574, "global_step": 82213, "epoch": 923, "lr": 7.918769167758626e-05}
{"train_loss": 0.1268644630908966, "global_step": 82214, "epoch": 923, "lr": 7.918722096967806e-05}
{"train_loss": 0.2888769507408142, "global_step": 82215, "epoch": 923, "lr": 7.918675025784599e-05}
{"train_loss": 0.13776543736457825, "global_step": 82216, "epoch": 923, "lr": 7.91862795420901e-05}
{"train_loss": 0.2306302934885025, "global_step": 82217, "epoch": 923, "lr": 7.918580882241049e-05}
{"train_loss": 0.11744657158851624, "global_step": 82218, "epoch": 923, "lr": 7.918533809880719e-05}
{"train_loss": 0.10629214346408844, "global_step": 82219, "epoch": 923, "lr": 7.918486737128029e-05}
{"train_loss": 0.2306160181760788, "global_step": 82220, "epoch": 923, "lr": 7.918439663982983e-05}
{"train_loss": 0.3180406093597412, "global_step": 82221, "epoch": 923, "lr": 7.91839259044559e-05}
{"train_loss": 0.32312923669815063, "global_step": 82222, "epoch": 923, "lr": 7.918345516515852e-05}
{"train_loss": 0.21567696332931519, "global_step": 82223, "epoch": 923, "lr": 7.91829844219378e-05}
{"train_loss": 0.13908521831035614, "global_step": 82224, "epoch": 923, "lr": 7.918251367479378e-05}
{"train_loss": 0.11409612745046616, "global_step": 82225, "epoch": 923, "lr": 7.918204292372652e-05}
{"train_loss": 0.19360053539276123, "global_step": 82226, "epoch": 923, "lr": 7.91815721687361e-05}
{"train_loss": 0.17661181092262268, "global_step": 82227, "epoch": 923, "lr": 7.918110140982256e-05}
{"train_loss": 0.17199234664440155, "global_step": 82228, "epoch": 923, "lr": 7.918063064698599e-05}
{"train_loss": 0.2162446826696396, "global_step": 82229, "epoch": 923, "lr": 7.918015988022643e-05}
{"train_loss": 0.19412660598754883, "global_step": 82230, "epoch": 923, "lr": 7.917968910954396e-05}
{"train_loss": 0.23703452944755554, "global_step": 82231, "epoch": 923, "lr": 7.917921833493864e-05}
{"train_loss": 0.24930930137634277, "global_step": 82232, "epoch": 923, "lr": 7.917874755641053e-05}
{"train_loss": 0.18945668637752533, "global_step": 82233, "epoch": 923, "lr": 7.917827677395969e-05}
{"train_loss": 0.22974133491516113, "global_step": 82234, "epoch": 923, "lr": 7.917780598758618e-05}
{"train_loss": 0.19886750401405806, "global_step": 82235, "epoch": 923, "lr": 7.917733519729007e-05, "val_loss": 4.08517599105835}
{"train_loss": 0.19476886093616486, "global_step": 82236, "epoch": 924, "lr": 7.917686440307142e-05}
{"train_loss": 0.18100568652153015, "global_step": 82237, "epoch": 924, "lr": 7.91763936049303e-05}
{"train_loss": 0.13679425418376923, "global_step": 82238, "epoch": 924, "lr": 7.917592280286678e-05}
{"train_loss": 0.19559349119663239, "global_step": 82239, "epoch": 924, "lr": 7.91754519968809e-05}
{"train_loss": 0.17484644055366516, "global_step": 82240, "epoch": 924, "lr": 7.917498118697275e-05}
{"train_loss": 0.2112734615802765, "global_step": 82241, "epoch": 924, "lr": 7.917451037314236e-05}
{"train_loss": 0.11135062575340271, "global_step": 82242, "epoch": 924, "lr": 7.917403955538982e-05}
{"train_loss": 0.25844550132751465, "global_step": 82243, "epoch": 924, "lr": 7.91735687337152e-05}
{"train_loss": 0.17225654423236847, "global_step": 82244, "epoch": 924, "lr": 7.917309790811853e-05}
{"train_loss": 0.19944582879543304, "global_step": 82245, "epoch": 924, "lr": 7.91726270785999e-05}
{"train_loss": 0.12031101435422897, "global_step": 82246, "epoch": 924, "lr": 7.917215624515936e-05}
{"train_loss": 0.19720804691314697, "global_step": 82247, "epoch": 924, "lr": 7.917168540779698e-05}
{"train_loss": 0.1477511078119278, "global_step": 82248, "epoch": 924, "lr": 7.917121456651284e-05}
{"train_loss": 0.20375774800777435, "global_step": 82249, "epoch": 924, "lr": 7.917074372130697e-05}
{"train_loss": 0.17482973635196686, "global_step": 82250, "epoch": 924, "lr": 7.917027287217944e-05}
{"train_loss": 0.19924291968345642, "global_step": 82251, "epoch": 924, "lr": 7.916980201913033e-05}
{"train_loss": 0.21874137222766876, "global_step": 82252, "epoch": 924, "lr": 7.916933116215971e-05}
{"train_loss": 0.16035650670528412, "global_step": 82253, "epoch": 924, "lr": 7.916886030126763e-05}
{"train_loss": 0.15933898091316223, "global_step": 82254, "epoch": 924, "lr": 7.916838943645414e-05}
{"train_loss": 0.15011157095432281, "global_step": 82255, "epoch": 924, "lr": 7.916791856771931e-05}
{"train_loss": 0.21737106144428253, "global_step": 82256, "epoch": 924, "lr": 7.916744769506321e-05}
{"train_loss": 0.09555660933256149, "global_step": 82257, "epoch": 924, "lr": 7.916697681848591e-05}
{"train_loss": 0.21391747891902924, "global_step": 82258, "epoch": 924, "lr": 7.916650593798746e-05}
{"train_loss": 0.14772194623947144, "global_step": 82259, "epoch": 924, "lr": 7.916603505356794e-05}
{"train_loss": 0.2070026397705078, "global_step": 82260, "epoch": 924, "lr": 7.91655641652274e-05}
{"train_loss": 0.1938425749540329, "global_step": 82261, "epoch": 924, "lr": 7.916509327296591e-05}
{"train_loss": 0.11697433888912201, "global_step": 82262, "epoch": 924, "lr": 7.916462237678351e-05}
{"train_loss": 0.18622450530529022, "global_step": 82263, "epoch": 924, "lr": 7.916415147668028e-05}
{"train_loss": 0.17107701301574707, "global_step": 82264, "epoch": 924, "lr": 7.916368057265631e-05}
{"train_loss": 0.16944371163845062, "global_step": 82265, "epoch": 924, "lr": 7.916320966471162e-05}
{"train_loss": 0.22839146852493286, "global_step": 82266, "epoch": 924, "lr": 7.916273875284629e-05}
{"train_loss": 0.17102453112602234, "global_step": 82267, "epoch": 924, "lr": 7.916226783706039e-05}
{"train_loss": 0.1587386429309845, "global_step": 82268, "epoch": 924, "lr": 7.916179691735398e-05}
{"train_loss": 0.2574770450592041, "global_step": 82269, "epoch": 924, "lr": 7.916132599372713e-05}
{"train_loss": 0.19694554805755615, "global_step": 82270, "epoch": 924, "lr": 7.916085506617987e-05}
{"train_loss": 0.13548079133033752, "global_step": 82271, "epoch": 924, "lr": 7.916038413471231e-05}
{"train_loss": 0.20641577243804932, "global_step": 82272, "epoch": 924, "lr": 7.91599131993245e-05}
{"train_loss": 0.18000155687332153, "global_step": 82273, "epoch": 924, "lr": 7.915944226001648e-05}
{"train_loss": 0.2271987795829773, "global_step": 82274, "epoch": 924, "lr": 7.915897131678833e-05}
{"train_loss": 0.15256240963935852, "global_step": 82275, "epoch": 924, "lr": 7.915850036964012e-05}
{"train_loss": 0.194864422082901, "global_step": 82276, "epoch": 924, "lr": 7.915802941857188e-05}
{"train_loss": 0.20444585382938385, "global_step": 82277, "epoch": 924, "lr": 7.915755846358373e-05}
{"train_loss": 0.1019289568066597, "global_step": 82278, "epoch": 924, "lr": 7.915708750467568e-05}
{"train_loss": 0.12494725733995438, "global_step": 82279, "epoch": 924, "lr": 7.915661654184783e-05}
{"train_loss": 0.15118512511253357, "global_step": 82280, "epoch": 924, "lr": 7.91561455751002e-05}
{"train_loss": 0.31217220425605774, "global_step": 82281, "epoch": 924, "lr": 7.915567460443291e-05}
{"train_loss": 0.22338919341564178, "global_step": 82282, "epoch": 924, "lr": 7.915520362984599e-05}
{"train_loss": 0.23153555393218994, "global_step": 82283, "epoch": 924, "lr": 7.915473265133951e-05}
{"train_loss": 0.14310035109519958, "global_step": 82284, "epoch": 924, "lr": 7.915426166891353e-05}
{"train_loss": 0.16408304870128632, "global_step": 82285, "epoch": 924, "lr": 7.91537906825681e-05}
{"train_loss": 0.23143890500068665, "global_step": 82286, "epoch": 924, "lr": 7.91533196923033e-05}
{"train_loss": 0.1572621762752533, "global_step": 82287, "epoch": 924, "lr": 7.91528486981192e-05}
{"train_loss": 0.14864793419837952, "global_step": 82288, "epoch": 924, "lr": 7.915237770001587e-05}
{"train_loss": 0.21205464005470276, "global_step": 82289, "epoch": 924, "lr": 7.915190669799335e-05}
{"train_loss": 0.1390029489994049, "global_step": 82290, "epoch": 924, "lr": 7.91514356920517e-05}
{"train_loss": 0.16729606688022614, "global_step": 82291, "epoch": 924, "lr": 7.915096468219101e-05}
{"train_loss": 0.24053144454956055, "global_step": 82292, "epoch": 924, "lr": 7.915049366841133e-05}
{"train_loss": 0.17873451113700867, "global_step": 82293, "epoch": 924, "lr": 7.91500226507127e-05}
{"train_loss": 0.14198929071426392, "global_step": 82294, "epoch": 924, "lr": 7.914955162909521e-05}
{"train_loss": 0.13883289694786072, "global_step": 82295, "epoch": 924, "lr": 7.914908060355893e-05}
{"train_loss": 0.1875470131635666, "global_step": 82296, "epoch": 924, "lr": 7.91486095741039e-05}
{"train_loss": 0.23676052689552307, "global_step": 82297, "epoch": 924, "lr": 7.914813854073019e-05}
{"train_loss": 0.1316966563463211, "global_step": 82298, "epoch": 924, "lr": 7.914766750343788e-05}
{"train_loss": 0.11106181889772415, "global_step": 82299, "epoch": 924, "lr": 7.914719646222701e-05}
{"train_loss": 0.23196563124656677, "global_step": 82300, "epoch": 924, "lr": 7.914672541709767e-05}
{"train_loss": 0.2207557111978531, "global_step": 82301, "epoch": 924, "lr": 7.91462543680499e-05}
{"train_loss": 0.18091675639152527, "global_step": 82302, "epoch": 924, "lr": 7.914578331508378e-05}
{"train_loss": 0.15771323442459106, "global_step": 82303, "epoch": 924, "lr": 7.914531225819936e-05}
{"train_loss": 0.14779362082481384, "global_step": 82304, "epoch": 924, "lr": 7.914484119739669e-05}
{"train_loss": 0.23643657565116882, "global_step": 82305, "epoch": 924, "lr": 7.914437013267588e-05}
{"train_loss": 0.18156622350215912, "global_step": 82306, "epoch": 924, "lr": 7.914389906403695e-05}
{"train_loss": 0.20921608805656433, "global_step": 82307, "epoch": 924, "lr": 7.914342799147998e-05}
{"train_loss": 0.20105955004692078, "global_step": 82308, "epoch": 924, "lr": 7.914295691500503e-05}
{"train_loss": 0.2207939773797989, "global_step": 82309, "epoch": 924, "lr": 7.914248583461217e-05}
{"train_loss": 0.26154133677482605, "global_step": 82310, "epoch": 924, "lr": 7.914201475030145e-05}
{"train_loss": 0.18861712515354156, "global_step": 82311, "epoch": 924, "lr": 7.914154366207294e-05}
{"train_loss": 0.15749375522136688, "global_step": 82312, "epoch": 924, "lr": 7.91410725699267e-05}
{"train_loss": 0.1510021835565567, "global_step": 82313, "epoch": 924, "lr": 7.914060147386281e-05}
{"train_loss": 0.21020719408988953, "global_step": 82314, "epoch": 924, "lr": 7.914013037388133e-05}
{"train_loss": 0.21577896177768707, "global_step": 82315, "epoch": 924, "lr": 7.91396592699823e-05}
{"train_loss": 0.17973168194293976, "global_step": 82316, "epoch": 924, "lr": 7.91391881621658e-05}
{"train_loss": 0.25418850779533386, "global_step": 82317, "epoch": 924, "lr": 7.913871705043188e-05}
{"train_loss": 0.19680863618850708, "global_step": 82318, "epoch": 924, "lr": 7.913824593478062e-05}
{"train_loss": 0.17153479158878326, "global_step": 82319, "epoch": 924, "lr": 7.91377748152121e-05}
{"train_loss": 0.13362351059913635, "global_step": 82320, "epoch": 924, "lr": 7.913730369172633e-05}
{"train_loss": 0.12782689929008484, "global_step": 82321, "epoch": 924, "lr": 7.913683256432344e-05}
{"train_loss": 0.20225374400615692, "global_step": 82322, "epoch": 924, "lr": 7.913636143300344e-05}
{"train_loss": 0.21890604496002197, "global_step": 82323, "epoch": 924, "lr": 7.91358902977664e-05}
{"train_loss": 0.18395682608478525, "global_step": 82324, "epoch": 924, "lr": 7.913541915861241e-05, "val_loss": 4.260340690612793}
{"train_loss": 0.20658420026302338, "global_step": 82325, "epoch": 925, "lr": 7.913494801554151e-05}
{"train_loss": 0.1641807109117508, "global_step": 82326, "epoch": 925, "lr": 7.913447686855378e-05}
{"train_loss": 0.1669212430715561, "global_step": 82327, "epoch": 925, "lr": 7.913400571764926e-05}
{"train_loss": 0.21999838948249817, "global_step": 82328, "epoch": 925, "lr": 7.913353456282803e-05}
{"train_loss": 0.21315954625606537, "global_step": 82329, "epoch": 925, "lr": 7.913306340409017e-05}
{"train_loss": 0.1729237139225006, "global_step": 82330, "epoch": 925, "lr": 7.913259224143572e-05}
{"train_loss": 0.1875939965248108, "global_step": 82331, "epoch": 925, "lr": 7.913212107486473e-05}
{"train_loss": 0.17955847084522247, "global_step": 82332, "epoch": 925, "lr": 7.91316499043773e-05}
{"train_loss": 0.19576463103294373, "global_step": 82333, "epoch": 925, "lr": 7.913117872997346e-05}
{"train_loss": 0.18822577595710754, "global_step": 82334, "epoch": 925, "lr": 7.91307075516533e-05}
{"train_loss": 0.15834471583366394, "global_step": 82335, "epoch": 925, "lr": 7.913023636941686e-05}
{"train_loss": 0.21262045204639435, "global_step": 82336, "epoch": 925, "lr": 7.912976518326422e-05}
{"train_loss": 0.18759644031524658, "global_step": 82337, "epoch": 925, "lr": 7.912929399319544e-05}
{"train_loss": 0.1358633041381836, "global_step": 82338, "epoch": 925, "lr": 7.912882279921059e-05}
{"train_loss": 0.19012002646923065, "global_step": 82339, "epoch": 925, "lr": 7.912835160130972e-05}
{"train_loss": 0.16900251805782318, "global_step": 82340, "epoch": 925, "lr": 7.912788039949288e-05}
{"train_loss": 0.2576600909233093, "global_step": 82341, "epoch": 925, "lr": 7.912740919376018e-05}
{"train_loss": 0.1905001997947693, "global_step": 82342, "epoch": 925, "lr": 7.912693798411163e-05}
{"train_loss": 0.26150214672088623, "global_step": 82343, "epoch": 925, "lr": 7.912646677054733e-05}
{"train_loss": 0.10572928190231323, "global_step": 82344, "epoch": 925, "lr": 7.912599555306734e-05}
{"train_loss": 0.18182174861431122, "global_step": 82345, "epoch": 925, "lr": 7.91255243316717e-05}
{"train_loss": 0.19061434268951416, "global_step": 82346, "epoch": 925, "lr": 7.91250531063605e-05}
{"train_loss": 0.1241210475564003, "global_step": 82347, "epoch": 925, "lr": 7.912458187713378e-05}
{"train_loss": 0.17618609964847565, "global_step": 82348, "epoch": 925, "lr": 7.91241106439916e-05}
{"train_loss": 0.15919390320777893, "global_step": 82349, "epoch": 925, "lr": 7.912363940693408e-05}
{"train_loss": 0.2710166275501251, "global_step": 82350, "epoch": 925, "lr": 7.912316816596121e-05}
{"train_loss": 0.17324206233024597, "global_step": 82351, "epoch": 925, "lr": 7.91226969210731e-05}
{"train_loss": 0.14841321110725403, "global_step": 82352, "epoch": 925, "lr": 7.912222567226979e-05}
{"train_loss": 0.20481982827186584, "global_step": 82353, "epoch": 925, "lr": 7.912175441955136e-05}
{"train_loss": 0.14339953660964966, "global_step": 82354, "epoch": 925, "lr": 7.912128316291785e-05}
{"train_loss": 0.23125071823596954, "global_step": 82355, "epoch": 925, "lr": 7.912081190236935e-05}
{"train_loss": 0.24378404021263123, "global_step": 82356, "epoch": 925, "lr": 7.91203406379059e-05}
{"train_loss": 0.29183900356292725, "global_step": 82357, "epoch": 925, "lr": 7.911986936952759e-05}
{"train_loss": 0.21353264153003693, "global_step": 82358, "epoch": 925, "lr": 7.911939809723447e-05}
{"train_loss": 0.1999385505914688, "global_step": 82359, "epoch": 925, "lr": 7.911892682102659e-05}
{"train_loss": 0.13626796007156372, "global_step": 82360, "epoch": 925, "lr": 7.911845554090404e-05}
{"train_loss": 0.30423223972320557, "global_step": 82361, "epoch": 925, "lr": 7.911798425686685e-05}
{"train_loss": 0.20058085024356842, "global_step": 82362, "epoch": 925, "lr": 7.911751296891512e-05}
{"train_loss": 0.2874763011932373, "global_step": 82363, "epoch": 925, "lr": 7.911704167704889e-05}
{"train_loss": 0.19838225841522217, "global_step": 82364, "epoch": 925, "lr": 7.911657038126822e-05}
{"train_loss": 0.15538965165615082, "global_step": 82365, "epoch": 925, "lr": 7.911609908157318e-05}
{"train_loss": 0.2210480272769928, "global_step": 82366, "epoch": 925, "lr": 7.911562777796385e-05}
{"train_loss": 0.13001741468906403, "global_step": 82367, "epoch": 925, "lr": 7.911515647044027e-05}
{"train_loss": 0.18190252780914307, "global_step": 82368, "epoch": 925, "lr": 7.91146851590025e-05}
{"train_loss": 0.234344482421875, "global_step": 82369, "epoch": 925, "lr": 7.911421384365065e-05}
{"train_loss": 0.17763395607471466, "global_step": 82370, "epoch": 925, "lr": 7.911374252438472e-05}
{"train_loss": 0.19671140611171722, "global_step": 82371, "epoch": 925, "lr": 7.91132712012048e-05}
{"train_loss": 0.1625763177871704, "global_step": 82372, "epoch": 925, "lr": 7.911279987411098e-05}
{"train_loss": 0.19214977324008942, "global_step": 82373, "epoch": 925, "lr": 7.911232854310327e-05}
{"train_loss": 0.16380183398723602, "global_step": 82374, "epoch": 925, "lr": 7.911185720818178e-05}
{"train_loss": 0.24207620322704315, "global_step": 82375, "epoch": 925, "lr": 7.911138586934655e-05}
{"train_loss": 0.25760841369628906, "global_step": 82376, "epoch": 925, "lr": 7.911091452659766e-05}
{"train_loss": 0.22869513928890228, "global_step": 82377, "epoch": 925, "lr": 7.911044317993514e-05}
{"train_loss": 0.1832345724105835, "global_step": 82378, "epoch": 925, "lr": 7.91099718293591e-05}
{"train_loss": 0.31137576699256897, "global_step": 82379, "epoch": 925, "lr": 7.910950047486957e-05}
{"train_loss": 0.17225190997123718, "global_step": 82380, "epoch": 925, "lr": 7.910902911646662e-05}
{"train_loss": 0.2525239586830139, "global_step": 82381, "epoch": 925, "lr": 7.910855775415031e-05}
{"train_loss": 0.1675766110420227, "global_step": 82382, "epoch": 925, "lr": 7.910808638792072e-05}
{"train_loss": 0.14699503779411316, "global_step": 82383, "epoch": 925, "lr": 7.91076150177779e-05}
{"train_loss": 0.21960140764713287, "global_step": 82384, "epoch": 925, "lr": 7.910714364372192e-05}
{"train_loss": 0.3179605305194855, "global_step": 82385, "epoch": 925, "lr": 7.910667226575283e-05}
{"train_loss": 0.18308301270008087, "global_step": 82386, "epoch": 925, "lr": 7.91062008838707e-05}
{"train_loss": 0.2719019651412964, "global_step": 82387, "epoch": 925, "lr": 7.91057294980756e-05}
{"train_loss": 0.21517717838287354, "global_step": 82388, "epoch": 925, "lr": 7.91052581083676e-05}
{"train_loss": 0.17792168259620667, "global_step": 82389, "epoch": 925, "lr": 7.910478671474673e-05}
{"train_loss": 0.23572221398353577, "global_step": 82390, "epoch": 925, "lr": 7.91043153172131e-05}
{"train_loss": 0.19119898974895477, "global_step": 82391, "epoch": 925, "lr": 7.910384391576673e-05}
{"train_loss": 0.12227839231491089, "global_step": 82392, "epoch": 925, "lr": 7.910337251040771e-05}
{"train_loss": 0.16618406772613525, "global_step": 82393, "epoch": 925, "lr": 7.910290110113611e-05}
{"train_loss": 0.20523878931999207, "global_step": 82394, "epoch": 925, "lr": 7.910242968795196e-05}
{"train_loss": 0.214860737323761, "global_step": 82395, "epoch": 925, "lr": 7.910195827085535e-05}
{"train_loss": 0.13872122764587402, "global_step": 82396, "epoch": 925, "lr": 7.910148684984635e-05}
{"train_loss": 0.2836169898509979, "global_step": 82397, "epoch": 925, "lr": 7.9101015424925e-05}
{"train_loss": 0.10179515182971954, "global_step": 82398, "epoch": 925, "lr": 7.910054399609138e-05}
{"train_loss": 0.2001970261335373, "global_step": 82399, "epoch": 925, "lr": 7.910007256334554e-05}
{"train_loss": 0.13057932257652283, "global_step": 82400, "epoch": 925, "lr": 7.909960112668753e-05}
{"train_loss": 0.18315190076828003, "global_step": 82401, "epoch": 925, "lr": 7.909912968611746e-05}
{"train_loss": 0.1985885202884674, "global_step": 82402, "epoch": 925, "lr": 7.909865824163536e-05}
{"train_loss": 0.2183939516544342, "global_step": 82403, "epoch": 925, "lr": 7.90981867932413e-05}
{"train_loss": 0.2039342075586319, "global_step": 82404, "epoch": 925, "lr": 7.909771534093534e-05}
{"train_loss": 0.10576427727937698, "global_step": 82405, "epoch": 925, "lr": 7.909724388471753e-05}
{"train_loss": 0.1911601722240448, "global_step": 82406, "epoch": 925, "lr": 7.909677242458797e-05}
{"train_loss": 0.19136004149913788, "global_step": 82407, "epoch": 925, "lr": 7.90963009605467e-05}
{"train_loss": 0.15301740169525146, "global_step": 82408, "epoch": 925, "lr": 7.90958294925938e-05}
{"train_loss": 0.1969877928495407, "global_step": 82409, "epoch": 925, "lr": 7.909535802072929e-05}
{"train_loss": 0.17311376333236694, "global_step": 82410, "epoch": 925, "lr": 7.909488654495329e-05}
{"train_loss": 0.18267516791820526, "global_step": 82411, "epoch": 925, "lr": 7.909441506526583e-05}
{"train_loss": 0.20455874502658844, "global_step": 82412, "epoch": 925, "lr": 7.909394358166698e-05}
{"train_loss": 0.19505482426520143, "global_step": 82413, "epoch": 925, "lr": 7.90934720941568e-05, "val_loss": 4.0651326179504395, "train_action_mse_error": 26.82746124267578}
{"train_loss": 0.11016777157783508, "global_step": 82414, "epoch": 926, "lr": 7.909300060273537e-05}
{"train_loss": 0.16322867572307587, "global_step": 82415, "epoch": 926, "lr": 7.909252910740273e-05}
{"train_loss": 0.1827702820301056, "global_step": 82416, "epoch": 926, "lr": 7.909205760815896e-05}
{"train_loss": 0.16582025587558746, "global_step": 82417, "epoch": 926, "lr": 7.909158610500411e-05}
{"train_loss": 0.2075609415769577, "global_step": 82418, "epoch": 926, "lr": 7.909111459793825e-05}
{"train_loss": 0.18175825476646423, "global_step": 82419, "epoch": 926, "lr": 7.909064308696145e-05}
{"train_loss": 0.17605027556419373, "global_step": 82420, "epoch": 926, "lr": 7.909017157207379e-05}
{"train_loss": 0.28705644607543945, "global_step": 82421, "epoch": 926, "lr": 7.908970005327528e-05}
{"train_loss": 0.22379429638385773, "global_step": 82422, "epoch": 926, "lr": 7.908922853056601e-05}
{"train_loss": 0.2167477011680603, "global_step": 82423, "epoch": 926, "lr": 7.908875700394607e-05}
{"train_loss": 0.21090379357337952, "global_step": 82424, "epoch": 926, "lr": 7.90882854734155e-05}
{"train_loss": 0.26065611839294434, "global_step": 82425, "epoch": 926, "lr": 7.908781393897434e-05}
{"train_loss": 0.1696496158838272, "global_step": 82426, "epoch": 926, "lr": 7.908734240062269e-05}
{"train_loss": 0.13010066747665405, "global_step": 82427, "epoch": 926, "lr": 7.908687085836061e-05}
{"train_loss": 0.15738984942436218, "global_step": 82428, "epoch": 926, "lr": 7.908639931218814e-05}
{"train_loss": 0.17889241874217987, "global_step": 82429, "epoch": 926, "lr": 7.908592776210537e-05}
{"train_loss": 0.18271629512310028, "global_step": 82430, "epoch": 926, "lr": 7.908545620811235e-05}
{"train_loss": 0.26398879289627075, "global_step": 82431, "epoch": 926, "lr": 7.908498465020915e-05}
{"train_loss": 0.2968400716781616, "global_step": 82432, "epoch": 926, "lr": 7.908451308839582e-05}
{"train_loss": 0.17049776017665863, "global_step": 82433, "epoch": 926, "lr": 7.908404152267243e-05}
{"train_loss": 0.1334112286567688, "global_step": 82434, "epoch": 926, "lr": 7.908356995303904e-05}
{"train_loss": 0.1620977222919464, "global_step": 82435, "epoch": 926, "lr": 7.908309837949574e-05}
{"train_loss": 0.17885635793209076, "global_step": 82436, "epoch": 926, "lr": 7.908262680204256e-05}
{"train_loss": 0.17131757736206055, "global_step": 82437, "epoch": 926, "lr": 7.908215522067957e-05}
{"train_loss": 0.17412540316581726, "global_step": 82438, "epoch": 926, "lr": 7.908168363540684e-05}
{"train_loss": 0.2091701179742813, "global_step": 82439, "epoch": 926, "lr": 7.908121204622445e-05}
{"train_loss": 0.1970767229795456, "global_step": 82440, "epoch": 926, "lr": 7.908074045313242e-05}
{"train_loss": 0.17891249060630798, "global_step": 82441, "epoch": 926, "lr": 7.908026885613086e-05}
{"train_loss": 0.1419796645641327, "global_step": 82442, "epoch": 926, "lr": 7.907979725521979e-05}
{"train_loss": 0.17561231553554535, "global_step": 82443, "epoch": 926, "lr": 7.90793256503993e-05}
{"train_loss": 0.1852140724658966, "global_step": 82444, "epoch": 926, "lr": 7.907885404166946e-05}
{"train_loss": 0.15804611146450043, "global_step": 82445, "epoch": 926, "lr": 7.907838242903031e-05}
{"train_loss": 0.16291265189647675, "global_step": 82446, "epoch": 926, "lr": 7.907791081248194e-05}
{"train_loss": 0.11989565938711166, "global_step": 82447, "epoch": 926, "lr": 7.907743919202441e-05}
{"train_loss": 0.15300247073173523, "global_step": 82448, "epoch": 926, "lr": 7.907696756765775e-05}
{"train_loss": 0.21249555051326752, "global_step": 82449, "epoch": 926, "lr": 7.907649593938205e-05}
{"train_loss": 0.21119891107082367, "global_step": 82450, "epoch": 926, "lr": 7.907602430719738e-05}
{"train_loss": 0.23788978159427643, "global_step": 82451, "epoch": 926, "lr": 7.907555267110378e-05}
{"train_loss": 0.24951037764549255, "global_step": 82452, "epoch": 926, "lr": 7.907508103110134e-05}
{"train_loss": 0.23823624849319458, "global_step": 82453, "epoch": 926, "lr": 7.90746093871901e-05}
{"train_loss": 0.17247165739536285, "global_step": 82454, "epoch": 926, "lr": 7.907413773937013e-05}
{"train_loss": 0.16450811922550201, "global_step": 82455, "epoch": 926, "lr": 7.907366608764151e-05}
{"train_loss": 0.19840267300605774, "global_step": 82456, "epoch": 926, "lr": 7.907319443200428e-05}
{"train_loss": 0.18605156242847443, "global_step": 82457, "epoch": 926, "lr": 7.907272277245852e-05}
{"train_loss": 0.19273406267166138, "global_step": 82458, "epoch": 926, "lr": 7.907225110900428e-05}
{"train_loss": 0.12503409385681152, "global_step": 82459, "epoch": 926, "lr": 7.907177944164163e-05}
{"train_loss": 0.15376141667366028, "global_step": 82460, "epoch": 926, "lr": 7.907130777037063e-05}
{"train_loss": 0.12997359037399292, "global_step": 82461, "epoch": 926, "lr": 7.907083609519135e-05}
{"train_loss": 0.14867696166038513, "global_step": 82462, "epoch": 926, "lr": 7.907036441610385e-05}
{"train_loss": 0.169963538646698, "global_step": 82463, "epoch": 926, "lr": 7.90698927331082e-05}
{"train_loss": 0.22777684032917023, "global_step": 82464, "epoch": 926, "lr": 7.906942104620445e-05}
{"train_loss": 0.35103610157966614, "global_step": 82465, "epoch": 926, "lr": 7.906894935539269e-05}
{"train_loss": 0.18208886682987213, "global_step": 82466, "epoch": 926, "lr": 7.906847766067293e-05}
{"train_loss": 0.20703119039535522, "global_step": 82467, "epoch": 926, "lr": 7.906800596204529e-05}
{"train_loss": 0.21170414984226227, "global_step": 82468, "epoch": 926, "lr": 7.906753425950981e-05}
{"train_loss": 0.2676753103733063, "global_step": 82469, "epoch": 926, "lr": 7.906706255306655e-05}
{"train_loss": 0.1756332516670227, "global_step": 82470, "epoch": 926, "lr": 7.906659084271558e-05}
{"train_loss": 0.2577582001686096, "global_step": 82471, "epoch": 926, "lr": 7.906611912845695e-05}
{"train_loss": 0.14013177156448364, "global_step": 82472, "epoch": 926, "lr": 7.906564741029076e-05}
{"train_loss": 0.16628369688987732, "global_step": 82473, "epoch": 926, "lr": 7.906517568821703e-05}
{"train_loss": 0.12096668779850006, "global_step": 82474, "epoch": 926, "lr": 7.906470396223584e-05}
{"train_loss": 0.17343981564044952, "global_step": 82475, "epoch": 926, "lr": 7.906423223234727e-05}
{"train_loss": 0.20821386575698853, "global_step": 82476, "epoch": 926, "lr": 7.906376049855135e-05}
{"train_loss": 0.17972490191459656, "global_step": 82477, "epoch": 926, "lr": 7.906328876084817e-05}
{"train_loss": 0.17353196442127228, "global_step": 82478, "epoch": 926, "lr": 7.906281701923778e-05}
{"train_loss": 0.20760615170001984, "global_step": 82479, "epoch": 926, "lr": 7.906234527372026e-05}
{"train_loss": 0.23576423525810242, "global_step": 82480, "epoch": 926, "lr": 7.906187352429564e-05}
{"train_loss": 0.20553888380527496, "global_step": 82481, "epoch": 926, "lr": 7.906140177096403e-05}
{"train_loss": 0.1759408563375473, "global_step": 82482, "epoch": 926, "lr": 7.906093001372546e-05}
{"train_loss": 0.18183393776416779, "global_step": 82483, "epoch": 926, "lr": 7.906045825258e-05}
{"train_loss": 0.1510709971189499, "global_step": 82484, "epoch": 926, "lr": 7.905998648752771e-05}
{"train_loss": 0.26907041668891907, "global_step": 82485, "epoch": 926, "lr": 7.905951471856867e-05}
{"train_loss": 0.10168043524026871, "global_step": 82486, "epoch": 926, "lr": 7.905904294570292e-05}
{"train_loss": 0.1521887183189392, "global_step": 82487, "epoch": 926, "lr": 7.905857116893054e-05}
{"train_loss": 0.1329098343849182, "global_step": 82488, "epoch": 926, "lr": 7.905809938825159e-05}
{"train_loss": 0.2918241322040558, "global_step": 82489, "epoch": 926, "lr": 7.905762760366615e-05}
{"train_loss": 0.15461789071559906, "global_step": 82490, "epoch": 926, "lr": 7.905715581517424e-05}
{"train_loss": 0.2274719625711441, "global_step": 82491, "epoch": 926, "lr": 7.905668402277597e-05}
{"train_loss": 0.15917730331420898, "global_step": 82492, "epoch": 926, "lr": 7.905621222647137e-05}
{"train_loss": 0.18437334895133972, "global_step": 82493, "epoch": 926, "lr": 7.905574042626053e-05}
{"train_loss": 0.20315571129322052, "global_step": 82494, "epoch": 926, "lr": 7.905526862214348e-05}
{"train_loss": 0.2785308063030243, "global_step": 82495, "epoch": 926, "lr": 7.905479681412032e-05}
{"train_loss": 0.18608660995960236, "global_step": 82496, "epoch": 926, "lr": 7.905432500219109e-05}
{"train_loss": 0.13241878151893616, "global_step": 82497, "epoch": 926, "lr": 7.905385318635585e-05}
{"train_loss": 0.16088730096817017, "global_step": 82498, "epoch": 926, "lr": 7.905338136661468e-05}
{"train_loss": 0.20804791152477264, "global_step": 82499, "epoch": 926, "lr": 7.905290954296765e-05}
{"train_loss": 0.21992939710617065, "global_step": 82500, "epoch": 926, "lr": 7.905243771541479e-05}
{"train_loss": 0.15412160754203796, "global_step": 82501, "epoch": 926, "lr": 7.90519658839562e-05}
{"train_loss": 0.18998569015706523, "global_step": 82502, "epoch": 926, "lr": 7.905149404859194e-05, "val_loss": 4.024653911590576}
{"train_loss": 0.15154887735843658, "global_step": 82503, "epoch": 927, "lr": 7.905102220932204e-05}
{"train_loss": 0.17535601556301117, "global_step": 82504, "epoch": 927, "lr": 7.905055036614657e-05}
{"train_loss": 0.14354656636714935, "global_step": 82505, "epoch": 927, "lr": 7.905007851906562e-05}
{"train_loss": 0.1692592203617096, "global_step": 82506, "epoch": 927, "lr": 7.904960666807925e-05}
{"train_loss": 0.25066229701042175, "global_step": 82507, "epoch": 927, "lr": 7.904913481318752e-05}
{"train_loss": 0.21610957384109497, "global_step": 82508, "epoch": 927, "lr": 7.904866295439049e-05}
{"train_loss": 0.13367561995983124, "global_step": 82509, "epoch": 927, "lr": 7.90481910916882e-05}
{"train_loss": 0.22518818080425262, "global_step": 82510, "epoch": 927, "lr": 7.904771922508074e-05}
{"train_loss": 0.17041729390621185, "global_step": 82511, "epoch": 927, "lr": 7.904724735456817e-05}
{"train_loss": 0.25221002101898193, "global_step": 82512, "epoch": 927, "lr": 7.904677548015056e-05}
{"train_loss": 0.16887891292572021, "global_step": 82513, "epoch": 927, "lr": 7.904630360182797e-05}
{"train_loss": 0.15737010538578033, "global_step": 82514, "epoch": 927, "lr": 7.904583171960044e-05}
{"train_loss": 0.21898750960826874, "global_step": 82515, "epoch": 927, "lr": 7.904535983346806e-05}
{"train_loss": 0.18639801442623138, "global_step": 82516, "epoch": 927, "lr": 7.904488794343088e-05}
{"train_loss": 0.16644823551177979, "global_step": 82517, "epoch": 927, "lr": 7.904441604948897e-05}
{"train_loss": 0.25392913818359375, "global_step": 82518, "epoch": 927, "lr": 7.90439441516424e-05}
{"train_loss": 0.2791779637336731, "global_step": 82519, "epoch": 927, "lr": 7.904347224989122e-05}
{"train_loss": 0.16178177297115326, "global_step": 82520, "epoch": 927, "lr": 7.904300034423551e-05}
{"train_loss": 0.1895071268081665, "global_step": 82521, "epoch": 927, "lr": 7.90425284346753e-05}
{"train_loss": 0.15150436758995056, "global_step": 82522, "epoch": 927, "lr": 7.904205652121069e-05}
{"train_loss": 0.21264243125915527, "global_step": 82523, "epoch": 927, "lr": 7.904158460384174e-05}
{"train_loss": 0.13667811453342438, "global_step": 82524, "epoch": 927, "lr": 7.904111268256848e-05}
{"train_loss": 0.21217955648899078, "global_step": 82525, "epoch": 927, "lr": 7.904064075739101e-05}
{"train_loss": 0.17461977899074554, "global_step": 82526, "epoch": 927, "lr": 7.904016882830938e-05}
{"train_loss": 0.21474242210388184, "global_step": 82527, "epoch": 927, "lr": 7.903969689532365e-05}
{"train_loss": 0.12768886983394623, "global_step": 82528, "epoch": 927, "lr": 7.903922495843389e-05}
{"train_loss": 0.14173322916030884, "global_step": 82529, "epoch": 927, "lr": 7.903875301764016e-05}
{"train_loss": 0.19443024694919586, "global_step": 82530, "epoch": 927, "lr": 7.90382810729425e-05}
{"train_loss": 0.271488219499588, "global_step": 82531, "epoch": 927, "lr": 7.903780912434105e-05}
{"train_loss": 0.211568683385849, "global_step": 82532, "epoch": 927, "lr": 7.903733717183578e-05}
{"train_loss": 0.29384422302246094, "global_step": 82533, "epoch": 927, "lr": 7.90368652154268e-05}
{"train_loss": 0.17043910920619965, "global_step": 82534, "epoch": 927, "lr": 7.903639325511416e-05}
{"train_loss": 0.16446131467819214, "global_step": 82535, "epoch": 927, "lr": 7.903592129089795e-05}
{"train_loss": 0.16490092873573303, "global_step": 82536, "epoch": 927, "lr": 7.90354493227782e-05}
{"train_loss": 0.23409923911094666, "global_step": 82537, "epoch": 927, "lr": 7.9034977350755e-05}
{"train_loss": 0.18677031993865967, "global_step": 82538, "epoch": 927, "lr": 7.90345053748284e-05}
{"train_loss": 0.24425926804542542, "global_step": 82539, "epoch": 927, "lr": 7.903403339499845e-05}
{"train_loss": 0.18145078420639038, "global_step": 82540, "epoch": 927, "lr": 7.903356141126523e-05}
{"train_loss": 0.22232550382614136, "global_step": 82541, "epoch": 927, "lr": 7.903308942362881e-05}
{"train_loss": 0.11945924907922745, "global_step": 82542, "epoch": 927, "lr": 7.903261743208925e-05}
{"train_loss": 0.1511738896369934, "global_step": 82543, "epoch": 927, "lr": 7.903214543664661e-05}
{"train_loss": 0.11706320941448212, "global_step": 82544, "epoch": 927, "lr": 7.903167343730093e-05}
{"train_loss": 0.1966453194618225, "global_step": 82545, "epoch": 927, "lr": 7.903120143405231e-05}
{"train_loss": 0.2027851641178131, "global_step": 82546, "epoch": 927, "lr": 7.90307294269008e-05}
{"train_loss": 0.17124976217746735, "global_step": 82547, "epoch": 927, "lr": 7.903025741584646e-05}
{"train_loss": 0.2136136293411255, "global_step": 82548, "epoch": 927, "lr": 7.902978540088936e-05}
{"train_loss": 0.18319447338581085, "global_step": 82549, "epoch": 927, "lr": 7.902931338202957e-05}
{"train_loss": 0.13859513401985168, "global_step": 82550, "epoch": 927, "lr": 7.902884135926713e-05}
{"train_loss": 0.18928119540214539, "global_step": 82551, "epoch": 927, "lr": 7.902836933260212e-05}
{"train_loss": 0.15400001406669617, "global_step": 82552, "epoch": 927, "lr": 7.902789730203457e-05}
{"train_loss": 0.25484591722488403, "global_step": 82553, "epoch": 927, "lr": 7.902742526756459e-05}
{"train_loss": 0.19222392141819, "global_step": 82554, "epoch": 927, "lr": 7.902695322919225e-05}
{"train_loss": 0.19391897320747375, "global_step": 82555, "epoch": 927, "lr": 7.902648118691757e-05}
{"train_loss": 0.18304887413978577, "global_step": 82556, "epoch": 927, "lr": 7.902600914074063e-05}
{"train_loss": 0.18673965334892273, "global_step": 82557, "epoch": 927, "lr": 7.902553709066152e-05}
{"train_loss": 0.14608530700206757, "global_step": 82558, "epoch": 927, "lr": 7.902506503668026e-05}
{"train_loss": 0.16775411367416382, "global_step": 82559, "epoch": 927, "lr": 7.902459297879693e-05}
{"train_loss": 0.144943505525589, "global_step": 82560, "epoch": 927, "lr": 7.902412091701161e-05}
{"train_loss": 0.12746383249759674, "global_step": 82561, "epoch": 927, "lr": 7.902364885132433e-05}
{"train_loss": 0.23777252435684204, "global_step": 82562, "epoch": 927, "lr": 7.90231767817352e-05}
{"train_loss": 0.1578141152858734, "global_step": 82563, "epoch": 927, "lr": 7.902270470824425e-05}
{"train_loss": 0.2141241878271103, "global_step": 82564, "epoch": 927, "lr": 7.902223263085154e-05}
{"train_loss": 0.19151152670383453, "global_step": 82565, "epoch": 927, "lr": 7.902176054955716e-05}
{"train_loss": 0.14466293156147003, "global_step": 82566, "epoch": 927, "lr": 7.902128846436114e-05}
{"train_loss": 0.1338089555501938, "global_step": 82567, "epoch": 927, "lr": 7.902081637526357e-05}
{"train_loss": 0.2197396606206894, "global_step": 82568, "epoch": 927, "lr": 7.902034428226452e-05}
{"train_loss": 0.22503261268138885, "global_step": 82569, "epoch": 927, "lr": 7.901987218536404e-05}
{"train_loss": 0.205050989985466, "global_step": 82570, "epoch": 927, "lr": 7.901940008456216e-05}
{"train_loss": 0.18635165691375732, "global_step": 82571, "epoch": 927, "lr": 7.901892797985899e-05}
{"train_loss": 0.16230152547359467, "global_step": 82572, "epoch": 927, "lr": 7.901845587125459e-05}
{"train_loss": 0.1617366075515747, "global_step": 82573, "epoch": 927, "lr": 7.9017983758749e-05}
{"train_loss": 0.21764524281024933, "global_step": 82574, "epoch": 927, "lr": 7.90175116423423e-05}
{"train_loss": 0.14449436962604523, "global_step": 82575, "epoch": 927, "lr": 7.901703952203454e-05}
{"train_loss": 0.12504534423351288, "global_step": 82576, "epoch": 927, "lr": 7.901656739782582e-05}
{"train_loss": 0.18492324650287628, "global_step": 82577, "epoch": 927, "lr": 7.901609526971615e-05}
{"train_loss": 0.21610701084136963, "global_step": 82578, "epoch": 927, "lr": 7.901562313770564e-05}
{"train_loss": 0.23211108148097992, "global_step": 82579, "epoch": 927, "lr": 7.901515100179431e-05}
{"train_loss": 0.18089254200458527, "global_step": 82580, "epoch": 927, "lr": 7.901467886198227e-05}
{"train_loss": 0.1394817978143692, "global_step": 82581, "epoch": 927, "lr": 7.901420671826954e-05}
{"train_loss": 0.26064568758010864, "global_step": 82582, "epoch": 927, "lr": 7.90137345706562e-05}
{"train_loss": 0.2086520940065384, "global_step": 82583, "epoch": 927, "lr": 7.901326241914234e-05}
{"train_loss": 0.28806179761886597, "global_step": 82584, "epoch": 927, "lr": 7.901279026372797e-05}
{"train_loss": 0.27233803272247314, "global_step": 82585, "epoch": 927, "lr": 7.901231810441321e-05}
{"train_loss": 0.15257225930690765, "global_step": 82586, "epoch": 927, "lr": 7.901184594119809e-05}
{"train_loss": 0.22292055189609528, "global_step": 82587, "epoch": 927, "lr": 7.901137377408268e-05}
{"train_loss": 0.24333040416240692, "global_step": 82588, "epoch": 927, "lr": 7.901090160306704e-05}
{"train_loss": 0.16903512179851532, "global_step": 82589, "epoch": 927, "lr": 7.901042942815125e-05}
{"train_loss": 0.18272551894187927, "global_step": 82590, "epoch": 927, "lr": 7.900995724933536e-05}
{"train_loss": 0.1902964725420716, "global_step": 82591, "epoch": 927, "lr": 7.900948506661943e-05, "val_loss": 4.098161697387695}
{"train_loss": 0.2038096934556961, "global_step": 82592, "epoch": 928, "lr": 7.90090128800035e-05}
{"train_loss": 0.23423926532268524, "global_step": 82593, "epoch": 928, "lr": 7.90085406894877e-05}
{"train_loss": 0.19545875489711761, "global_step": 82594, "epoch": 928, "lr": 7.900806849507205e-05}
{"train_loss": 0.16006626188755035, "global_step": 82595, "epoch": 928, "lr": 7.900759629675661e-05}
{"train_loss": 0.1292472630739212, "global_step": 82596, "epoch": 928, "lr": 7.900712409454146e-05}
{"train_loss": 0.2127479910850525, "global_step": 82597, "epoch": 928, "lr": 7.900665188842666e-05}
{"train_loss": 0.18748214840888977, "global_step": 82598, "epoch": 928, "lr": 7.900617967841227e-05}
{"train_loss": 0.2609473168849945, "global_step": 82599, "epoch": 928, "lr": 7.900570746449834e-05}
{"train_loss": 0.14287729561328888, "global_step": 82600, "epoch": 928, "lr": 7.900523524668495e-05}
{"train_loss": 0.2037370204925537, "global_step": 82601, "epoch": 928, "lr": 7.900476302497215e-05}
{"train_loss": 0.17471356689929962, "global_step": 82602, "epoch": 928, "lr": 7.900429079936002e-05}
{"train_loss": 0.1719391942024231, "global_step": 82603, "epoch": 928, "lr": 7.900381856984863e-05}
{"train_loss": 0.21910309791564941, "global_step": 82604, "epoch": 928, "lr": 7.9003346336438e-05}
{"train_loss": 0.23764893412590027, "global_step": 82605, "epoch": 928, "lr": 7.900287409912826e-05}
{"train_loss": 0.213921919465065, "global_step": 82606, "epoch": 928, "lr": 7.90024018579194e-05}
{"train_loss": 0.11942796409130096, "global_step": 82607, "epoch": 928, "lr": 7.900192961281155e-05}
{"train_loss": 0.15897777676582336, "global_step": 82608, "epoch": 928, "lr": 7.900145736380473e-05}
{"train_loss": 0.2675042748451233, "global_step": 82609, "epoch": 928, "lr": 7.900098511089902e-05}
{"train_loss": 0.20543642342090607, "global_step": 82610, "epoch": 928, "lr": 7.900051285409449e-05}
{"train_loss": 0.12794940173625946, "global_step": 82611, "epoch": 928, "lr": 7.900004059339118e-05}
{"train_loss": 0.27076876163482666, "global_step": 82612, "epoch": 928, "lr": 7.899956832878916e-05}
{"train_loss": 0.18835556507110596, "global_step": 82613, "epoch": 928, "lr": 7.899909606028853e-05}
{"train_loss": 0.15049128234386444, "global_step": 82614, "epoch": 928, "lr": 7.899862378788931e-05}
{"train_loss": 0.2001965045928955, "global_step": 82615, "epoch": 928, "lr": 7.899815151159158e-05}
{"train_loss": 0.1796964555978775, "global_step": 82616, "epoch": 928, "lr": 7.89976792313954e-05}
{"train_loss": 0.2786466181278229, "global_step": 82617, "epoch": 928, "lr": 7.899720694730084e-05}
{"train_loss": 0.21886274218559265, "global_step": 82618, "epoch": 928, "lr": 7.899673465930796e-05}
{"train_loss": 0.23073247075080872, "global_step": 82619, "epoch": 928, "lr": 7.899626236741681e-05}
{"train_loss": 0.2234383374452591, "global_step": 82620, "epoch": 928, "lr": 7.899579007162749e-05}
{"train_loss": 0.2805542051792145, "global_step": 82621, "epoch": 928, "lr": 7.899531777194003e-05}
{"train_loss": 0.21393316984176636, "global_step": 82622, "epoch": 928, "lr": 7.899484546835449e-05}
{"train_loss": 0.2761199474334717, "global_step": 82623, "epoch": 928, "lr": 7.899437316087095e-05}
{"train_loss": 0.22145061194896698, "global_step": 82624, "epoch": 928, "lr": 7.899390084948948e-05}
{"train_loss": 0.20859204232692719, "global_step": 82625, "epoch": 928, "lr": 7.899342853421012e-05}
{"train_loss": 0.22322945296764374, "global_step": 82626, "epoch": 928, "lr": 7.899295621503295e-05}
{"train_loss": 0.15758511424064636, "global_step": 82627, "epoch": 928, "lr": 7.899248389195803e-05}
{"train_loss": 0.17570962011814117, "global_step": 82628, "epoch": 928, "lr": 7.899201156498544e-05}
{"train_loss": 0.19086985290050507, "global_step": 82629, "epoch": 928, "lr": 7.899153923411521e-05}
{"train_loss": 0.23038241267204285, "global_step": 82630, "epoch": 928, "lr": 7.899106689934742e-05}
{"train_loss": 0.18831029534339905, "global_step": 82631, "epoch": 928, "lr": 7.899059456068214e-05}
{"train_loss": 0.23351411521434784, "global_step": 82632, "epoch": 928, "lr": 7.899012221811943e-05}
{"train_loss": 0.20022732019424438, "global_step": 82633, "epoch": 928, "lr": 7.898964987165936e-05}
{"train_loss": 0.18159694969654083, "global_step": 82634, "epoch": 928, "lr": 7.898917752130196e-05}
{"train_loss": 0.1758326143026352, "global_step": 82635, "epoch": 928, "lr": 7.898870516704735e-05}
{"train_loss": 0.1942896842956543, "global_step": 82636, "epoch": 928, "lr": 7.898823280889553e-05}
{"train_loss": 0.16791978478431702, "global_step": 82637, "epoch": 928, "lr": 7.898776044684661e-05}
{"train_loss": 0.22817271947860718, "global_step": 82638, "epoch": 928, "lr": 7.898728808090065e-05}
{"train_loss": 0.11387976258993149, "global_step": 82639, "epoch": 928, "lr": 7.898681571105769e-05}
{"train_loss": 0.24552050232887268, "global_step": 82640, "epoch": 928, "lr": 7.898634333731781e-05}
{"train_loss": 0.2214149683713913, "global_step": 82641, "epoch": 928, "lr": 7.898587095968106e-05}
{"train_loss": 0.260301411151886, "global_step": 82642, "epoch": 928, "lr": 7.898539857814753e-05}
{"train_loss": 0.1287369430065155, "global_step": 82643, "epoch": 928, "lr": 7.898492619271725e-05}
{"train_loss": 0.19442377984523773, "global_step": 82644, "epoch": 928, "lr": 7.898445380339031e-05}
{"train_loss": 0.16564953327178955, "global_step": 82645, "epoch": 928, "lr": 7.898398141016677e-05}
{"train_loss": 0.23917338252067566, "global_step": 82646, "epoch": 928, "lr": 7.898350901304668e-05}
{"train_loss": 0.145623579621315, "global_step": 82647, "epoch": 928, "lr": 7.89830366120301e-05}
{"train_loss": 0.23390108346939087, "global_step": 82648, "epoch": 928, "lr": 7.898256420711711e-05}
{"train_loss": 0.17739702761173248, "global_step": 82649, "epoch": 928, "lr": 7.898209179830776e-05}
{"train_loss": 0.17339444160461426, "global_step": 82650, "epoch": 928, "lr": 7.898161938560213e-05}
{"train_loss": 0.14517946541309357, "global_step": 82651, "epoch": 928, "lr": 7.898114696900028e-05}
{"train_loss": 0.19695702195167542, "global_step": 82652, "epoch": 928, "lr": 7.898067454850227e-05}
{"train_loss": 0.23406332731246948, "global_step": 82653, "epoch": 928, "lr": 7.898020212410815e-05}
{"train_loss": 0.16504904627799988, "global_step": 82654, "epoch": 928, "lr": 7.8979729695818e-05}
{"train_loss": 0.17399689555168152, "global_step": 82655, "epoch": 928, "lr": 7.897925726363188e-05}
{"train_loss": 0.12747687101364136, "global_step": 82656, "epoch": 928, "lr": 7.897878482754986e-05}
{"train_loss": 0.24774929881095886, "global_step": 82657, "epoch": 928, "lr": 7.897831238757199e-05}
{"train_loss": 0.1635916829109192, "global_step": 82658, "epoch": 928, "lr": 7.897783994369832e-05}
{"train_loss": 0.19975943863391876, "global_step": 82659, "epoch": 928, "lr": 7.897736749592895e-05}
{"train_loss": 0.13117504119873047, "global_step": 82660, "epoch": 928, "lr": 7.897689504426392e-05}
{"train_loss": 0.20039866864681244, "global_step": 82661, "epoch": 928, "lr": 7.89764225887033e-05}
{"train_loss": 0.10059425234794617, "global_step": 82662, "epoch": 928, "lr": 7.897595012924715e-05}
{"train_loss": 0.18407610058784485, "global_step": 82663, "epoch": 928, "lr": 7.897547766589555e-05}
{"train_loss": 0.15296967327594757, "global_step": 82664, "epoch": 928, "lr": 7.897500519864854e-05}
{"train_loss": 0.1469588726758957, "global_step": 82665, "epoch": 928, "lr": 7.89745327275062e-05}
{"train_loss": 0.16964851319789886, "global_step": 82666, "epoch": 928, "lr": 7.897406025246859e-05}
{"train_loss": 0.2777014374732971, "global_step": 82667, "epoch": 928, "lr": 7.897358777353577e-05}
{"train_loss": 0.15602053701877594, "global_step": 82668, "epoch": 928, "lr": 7.897311529070778e-05}
{"train_loss": 0.14909860491752625, "global_step": 82669, "epoch": 928, "lr": 7.897264280398473e-05}
{"train_loss": 0.17970970273017883, "global_step": 82670, "epoch": 928, "lr": 7.897217031336666e-05}
{"train_loss": 0.19124789535999298, "global_step": 82671, "epoch": 928, "lr": 7.897169781885363e-05}
{"train_loss": 0.21705645322799683, "global_step": 82672, "epoch": 928, "lr": 7.897122532044572e-05}
{"train_loss": 0.1799619197845459, "global_step": 82673, "epoch": 928, "lr": 7.897075281814297e-05}
{"train_loss": 0.21777230501174927, "global_step": 82674, "epoch": 928, "lr": 7.897028031194546e-05}
{"train_loss": 0.18046006560325623, "global_step": 82675, "epoch": 928, "lr": 7.896980780185326e-05}
{"train_loss": 0.13600994646549225, "global_step": 82676, "epoch": 928, "lr": 7.89693352878664e-05}
{"train_loss": 0.2424205094575882, "global_step": 82677, "epoch": 928, "lr": 7.896886276998498e-05}
{"train_loss": 0.22046758234500885, "global_step": 82678, "epoch": 928, "lr": 7.896839024820904e-05}
{"train_loss": 0.20292013883590698, "global_step": 82679, "epoch": 928, "lr": 7.896791772253865e-05}
{"train_loss": 0.19376117384500718, "global_step": 82680, "epoch": 928, "lr": 7.89674451929739e-05, "val_loss": 4.068215847015381}
{"train_loss": 0.22960789501667023, "global_step": 82681, "epoch": 929, "lr": 7.896697265951481e-05}
{"train_loss": 0.20527608692646027, "global_step": 82682, "epoch": 929, "lr": 7.896650012216147e-05}
{"train_loss": 0.15926888585090637, "global_step": 82683, "epoch": 929, "lr": 7.896602758091393e-05}
{"train_loss": 0.12928976118564606, "global_step": 82684, "epoch": 929, "lr": 7.896555503577227e-05}
{"train_loss": 0.16532383859157562, "global_step": 82685, "epoch": 929, "lr": 7.896508248673653e-05}
{"train_loss": 0.08911221474409103, "global_step": 82686, "epoch": 929, "lr": 7.896460993380681e-05}
{"train_loss": 0.2500867247581482, "global_step": 82687, "epoch": 929, "lr": 7.896413737698313e-05}
{"train_loss": 0.1618039757013321, "global_step": 82688, "epoch": 929, "lr": 7.896366481626558e-05}
{"train_loss": 0.1345469206571579, "global_step": 82689, "epoch": 929, "lr": 7.896319225165422e-05}
{"train_loss": 0.2638157308101654, "global_step": 82690, "epoch": 929, "lr": 7.896271968314913e-05}
{"train_loss": 0.2566494941711426, "global_step": 82691, "epoch": 929, "lr": 7.896224711075033e-05}
{"train_loss": 0.1745338886976242, "global_step": 82692, "epoch": 929, "lr": 7.896177453445792e-05}
{"train_loss": 0.17606613039970398, "global_step": 82693, "epoch": 929, "lr": 7.896130195427196e-05}
{"train_loss": 0.2078731507062912, "global_step": 82694, "epoch": 929, "lr": 7.89608293701925e-05}
{"train_loss": 0.12799033522605896, "global_step": 82695, "epoch": 929, "lr": 7.89603567822196e-05}
{"train_loss": 0.14585267007350922, "global_step": 82696, "epoch": 929, "lr": 7.895988419035334e-05}
{"train_loss": 0.26388323307037354, "global_step": 82697, "epoch": 929, "lr": 7.895941159459376e-05}
{"train_loss": 0.15649695694446564, "global_step": 82698, "epoch": 929, "lr": 7.895893899494095e-05}
{"train_loss": 0.22696152329444885, "global_step": 82699, "epoch": 929, "lr": 7.895846639139499e-05}
{"train_loss": 0.2164410501718521, "global_step": 82700, "epoch": 929, "lr": 7.89579937839559e-05}
{"train_loss": 0.23657767474651337, "global_step": 82701, "epoch": 929, "lr": 7.895752117262374e-05}
{"train_loss": 0.19681884348392487, "global_step": 82702, "epoch": 929, "lr": 7.895704855739862e-05}
{"train_loss": 0.1767953485250473, "global_step": 82703, "epoch": 929, "lr": 7.895657593828057e-05}
{"train_loss": 0.13253386318683624, "global_step": 82704, "epoch": 929, "lr": 7.895610331526966e-05}
{"train_loss": 0.15767113864421844, "global_step": 82705, "epoch": 929, "lr": 7.895563068836596e-05}
{"train_loss": 0.35595381259918213, "global_step": 82706, "epoch": 929, "lr": 7.895515805756952e-05}
{"train_loss": 0.19226238131523132, "global_step": 82707, "epoch": 929, "lr": 7.895468542288042e-05}
{"train_loss": 0.198635995388031, "global_step": 82708, "epoch": 929, "lr": 7.895421278429872e-05}
{"train_loss": 0.1358463168144226, "global_step": 82709, "epoch": 929, "lr": 7.895374014182446e-05}
{"train_loss": 0.259164422750473, "global_step": 82710, "epoch": 929, "lr": 7.895326749545772e-05}
{"train_loss": 0.21409766376018524, "global_step": 82711, "epoch": 929, "lr": 7.895279484519858e-05}
{"train_loss": 0.21105381846427917, "global_step": 82712, "epoch": 929, "lr": 7.895232219104709e-05}
{"train_loss": 0.17077311873435974, "global_step": 82713, "epoch": 929, "lr": 7.895184953300332e-05}
{"train_loss": 0.1963474005460739, "global_step": 82714, "epoch": 929, "lr": 7.895137687106732e-05}
{"train_loss": 0.12464925646781921, "global_step": 82715, "epoch": 929, "lr": 7.895090420523913e-05}
{"train_loss": 0.19850988686084747, "global_step": 82716, "epoch": 929, "lr": 7.895043153551889e-05}
{"train_loss": 0.197702094912529, "global_step": 82717, "epoch": 929, "lr": 7.894995886190659e-05}
{"train_loss": 0.15941211581230164, "global_step": 82718, "epoch": 929, "lr": 7.894948618440234e-05}
{"train_loss": 0.1815984547138214, "global_step": 82719, "epoch": 929, "lr": 7.894901350300619e-05}
{"train_loss": 0.1808074712753296, "global_step": 82720, "epoch": 929, "lr": 7.894854081771817e-05}
{"train_loss": 0.18235528469085693, "global_step": 82721, "epoch": 929, "lr": 7.894806812853839e-05}
{"train_loss": 0.19615809619426727, "global_step": 82722, "epoch": 929, "lr": 7.894759543546688e-05}
{"train_loss": 0.20114146173000336, "global_step": 82723, "epoch": 929, "lr": 7.894712273850373e-05}
{"train_loss": 0.1661870926618576, "global_step": 82724, "epoch": 929, "lr": 7.894665003764899e-05}
{"train_loss": 0.20316407084465027, "global_step": 82725, "epoch": 929, "lr": 7.894617733290273e-05}
{"train_loss": 0.17811375856399536, "global_step": 82726, "epoch": 929, "lr": 7.8945704624265e-05}
{"train_loss": 0.14363513886928558, "global_step": 82727, "epoch": 929, "lr": 7.894523191173588e-05}
{"train_loss": 0.2236829251050949, "global_step": 82728, "epoch": 929, "lr": 7.894475919531543e-05}
{"train_loss": 0.23272503912448883, "global_step": 82729, "epoch": 929, "lr": 7.894428647500371e-05}
{"train_loss": 0.14863403141498566, "global_step": 82730, "epoch": 929, "lr": 7.894381375080077e-05}
{"train_loss": 0.1232907697558403, "global_step": 82731, "epoch": 929, "lr": 7.89433410227067e-05}
{"train_loss": 0.18248678743839264, "global_step": 82732, "epoch": 929, "lr": 7.894286829072155e-05}
{"train_loss": 0.17786996066570282, "global_step": 82733, "epoch": 929, "lr": 7.894239555484537e-05}
{"train_loss": 0.2220683991909027, "global_step": 82734, "epoch": 929, "lr": 7.894192281507823e-05}
{"train_loss": 0.1737101525068283, "global_step": 82735, "epoch": 929, "lr": 7.894145007142023e-05}
{"train_loss": 0.14301477372646332, "global_step": 82736, "epoch": 929, "lr": 7.89409773238714e-05}
{"train_loss": 0.20377510786056519, "global_step": 82737, "epoch": 929, "lr": 7.89405045724318e-05}
{"train_loss": 0.21806597709655762, "global_step": 82738, "epoch": 929, "lr": 7.89400318171015e-05}
{"train_loss": 0.20596715807914734, "global_step": 82739, "epoch": 929, "lr": 7.893955905788058e-05}
{"train_loss": 0.14529617130756378, "global_step": 82740, "epoch": 929, "lr": 7.893908629476907e-05}
{"train_loss": 0.26575204730033875, "global_step": 82741, "epoch": 929, "lr": 7.893861352776705e-05}
{"train_loss": 0.11871135979890823, "global_step": 82742, "epoch": 929, "lr": 7.89381407568746e-05}
{"train_loss": 0.12027006596326828, "global_step": 82743, "epoch": 929, "lr": 7.893766798209177e-05}
{"train_loss": 0.25014662742614746, "global_step": 82744, "epoch": 929, "lr": 7.893719520341862e-05}
{"train_loss": 0.19724439084529877, "global_step": 82745, "epoch": 929, "lr": 7.893672242085521e-05}
{"train_loss": 0.1857953518629074, "global_step": 82746, "epoch": 929, "lr": 7.893624963440163e-05}
{"train_loss": 0.14699657261371613, "global_step": 82747, "epoch": 929, "lr": 7.89357768440579e-05}
{"train_loss": 0.29088082909584045, "global_step": 82748, "epoch": 929, "lr": 7.893530404982413e-05}
{"train_loss": 0.20054644346237183, "global_step": 82749, "epoch": 929, "lr": 7.893483125170034e-05}
{"train_loss": 0.23444387316703796, "global_step": 82750, "epoch": 929, "lr": 7.893435844968661e-05}
{"train_loss": 0.18987196683883667, "global_step": 82751, "epoch": 929, "lr": 7.893388564378304e-05}
{"train_loss": 0.14466266334056854, "global_step": 82752, "epoch": 929, "lr": 7.893341283398963e-05}
{"train_loss": 0.2314252406358719, "global_step": 82753, "epoch": 929, "lr": 7.893294002030649e-05}
{"train_loss": 0.1736968606710434, "global_step": 82754, "epoch": 929, "lr": 7.893246720273368e-05}
{"train_loss": 0.27613580226898193, "global_step": 82755, "epoch": 929, "lr": 7.893199438127123e-05}
{"train_loss": 0.27003493905067444, "global_step": 82756, "epoch": 929, "lr": 7.893152155591924e-05}
{"train_loss": 0.23172831535339355, "global_step": 82757, "epoch": 929, "lr": 7.893104872667776e-05}
{"train_loss": 0.18155406415462494, "global_step": 82758, "epoch": 929, "lr": 7.893057589354684e-05}
{"train_loss": 0.16207754611968994, "global_step": 82759, "epoch": 929, "lr": 7.893010305652657e-05}
{"train_loss": 0.14505596458911896, "global_step": 82760, "epoch": 929, "lr": 7.8929630215617e-05}
{"train_loss": 0.17680872976779938, "global_step": 82761, "epoch": 929, "lr": 7.892915737081819e-05}
{"train_loss": 0.25545987486839294, "global_step": 82762, "epoch": 929, "lr": 7.892868452213021e-05}
{"train_loss": 0.22677141427993774, "global_step": 82763, "epoch": 929, "lr": 7.892821166955312e-05}
{"train_loss": 0.19526034593582153, "global_step": 82764, "epoch": 929, "lr": 7.892773881308697e-05}
{"train_loss": 0.1887703239917755, "global_step": 82765, "epoch": 929, "lr": 7.892726595273185e-05}
{"train_loss": 0.21052464842796326, "global_step": 82766, "epoch": 929, "lr": 7.892679308848782e-05}
{"train_loss": 0.1969679296016693, "global_step": 82767, "epoch": 929, "lr": 7.892632022035494e-05}
{"train_loss": 0.2030480057001114, "global_step": 82768, "epoch": 929, "lr": 7.892584734833323e-05}
{"train_loss": 0.19228756930051225, "global_step": 82769, "epoch": 929, "lr": 7.892537447242283e-05, "val_loss": 4.197653293609619}
{"train_loss": 0.14475920796394348, "global_step": 82770, "epoch": 930, "lr": 7.892490159262375e-05}
{"train_loss": 0.19454537332057953, "global_step": 82771, "epoch": 930, "lr": 7.892442870893609e-05}
{"train_loss": 0.2560563385486603, "global_step": 82772, "epoch": 930, "lr": 7.892395582135986e-05}
{"train_loss": 0.208877831697464, "global_step": 82773, "epoch": 930, "lr": 7.892348292989517e-05}
{"train_loss": 0.22099599242210388, "global_step": 82774, "epoch": 930, "lr": 7.892301003454208e-05}
{"train_loss": 0.10712188482284546, "global_step": 82775, "epoch": 930, "lr": 7.892253713530065e-05}
{"train_loss": 0.17177797853946686, "global_step": 82776, "epoch": 930, "lr": 7.892206423217091e-05}
{"train_loss": 0.24844425916671753, "global_step": 82777, "epoch": 930, "lr": 7.892159132515297e-05}
{"train_loss": 0.1528656929731369, "global_step": 82778, "epoch": 930, "lr": 7.892111841424687e-05}
{"train_loss": 0.17245468497276306, "global_step": 82779, "epoch": 930, "lr": 7.892064549945269e-05}
{"train_loss": 0.2174462378025055, "global_step": 82780, "epoch": 930, "lr": 7.892017258077047e-05}
{"train_loss": 0.2179710865020752, "global_step": 82781, "epoch": 930, "lr": 7.891969965820028e-05}
{"train_loss": 0.15319189429283142, "global_step": 82782, "epoch": 930, "lr": 7.89192267317422e-05}
{"train_loss": 0.12333700060844421, "global_step": 82783, "epoch": 930, "lr": 7.891875380139627e-05}
{"train_loss": 0.17946767807006836, "global_step": 82784, "epoch": 930, "lr": 7.891828086716258e-05}
{"train_loss": 0.14371636509895325, "global_step": 82785, "epoch": 930, "lr": 7.891780792904118e-05}
{"train_loss": 0.2475537657737732, "global_step": 82786, "epoch": 930, "lr": 7.891733498703211e-05}
{"train_loss": 0.15910503268241882, "global_step": 82787, "epoch": 930, "lr": 7.891686204113547e-05}
{"train_loss": 0.24843543767929077, "global_step": 82788, "epoch": 930, "lr": 7.891638909135132e-05}
{"train_loss": 0.14766685664653778, "global_step": 82789, "epoch": 930, "lr": 7.89159161376797e-05}
{"train_loss": 0.25981950759887695, "global_step": 82790, "epoch": 930, "lr": 7.89154431801207e-05}
{"train_loss": 0.17559607326984406, "global_step": 82791, "epoch": 930, "lr": 7.891497021867437e-05}
{"train_loss": 0.161013662815094, "global_step": 82792, "epoch": 930, "lr": 7.891449725334077e-05}
{"train_loss": 0.13577228784561157, "global_step": 82793, "epoch": 930, "lr": 7.891402428411996e-05}
{"train_loss": 0.259487122297287, "global_step": 82794, "epoch": 930, "lr": 7.891355131101203e-05}
{"train_loss": 0.2626710832118988, "global_step": 82795, "epoch": 930, "lr": 7.891307833401702e-05}
{"train_loss": 0.18035583198070526, "global_step": 82796, "epoch": 930, "lr": 7.891260535313499e-05}
{"train_loss": 0.15519048273563385, "global_step": 82797, "epoch": 930, "lr": 7.8912132368366e-05}
{"train_loss": 0.23371900618076324, "global_step": 82798, "epoch": 930, "lr": 7.891165937971017e-05}
{"train_loss": 0.11483314633369446, "global_step": 82799, "epoch": 930, "lr": 7.891118638716748e-05}
{"train_loss": 0.1626218557357788, "global_step": 82800, "epoch": 930, "lr": 7.891071339073806e-05}
{"train_loss": 0.19858558475971222, "global_step": 82801, "epoch": 930, "lr": 7.891024039042193e-05}
{"train_loss": 0.1894543617963791, "global_step": 82802, "epoch": 930, "lr": 7.890976738621917e-05}
{"train_loss": 0.12457197904586792, "global_step": 82803, "epoch": 930, "lr": 7.890929437812985e-05}
{"train_loss": 0.25828200578689575, "global_step": 82804, "epoch": 930, "lr": 7.890882136615403e-05}
{"train_loss": 0.2392680048942566, "global_step": 82805, "epoch": 930, "lr": 7.890834835029177e-05}
{"train_loss": 0.22020718455314636, "global_step": 82806, "epoch": 930, "lr": 7.890787533054311e-05}
{"train_loss": 0.2466982752084732, "global_step": 82807, "epoch": 930, "lr": 7.890740230690816e-05}
{"train_loss": 0.2296028435230255, "global_step": 82808, "epoch": 930, "lr": 7.890692927938698e-05}
{"train_loss": 0.24271813035011292, "global_step": 82809, "epoch": 930, "lr": 7.890645624797959e-05}
{"train_loss": 0.1939263492822647, "global_step": 82810, "epoch": 930, "lr": 7.890598321268609e-05}
{"train_loss": 0.2285807877779007, "global_step": 82811, "epoch": 930, "lr": 7.890551017350654e-05}
{"train_loss": 0.2592768371105194, "global_step": 82812, "epoch": 930, "lr": 7.890503713044098e-05}
{"train_loss": 0.1296621412038803, "global_step": 82813, "epoch": 930, "lr": 7.89045640834895e-05}
{"train_loss": 0.2712726294994354, "global_step": 82814, "epoch": 930, "lr": 7.890409103265215e-05}
{"train_loss": 0.20867721736431122, "global_step": 82815, "epoch": 930, "lr": 7.8903617977929e-05}
{"train_loss": 0.11808159947395325, "global_step": 82816, "epoch": 930, "lr": 7.890314491932012e-05}
{"train_loss": 0.12432421743869781, "global_step": 82817, "epoch": 930, "lr": 7.890267185682554e-05}
{"train_loss": 0.22219866514205933, "global_step": 82818, "epoch": 930, "lr": 7.890219879044535e-05}
{"train_loss": 0.1816365122795105, "global_step": 82819, "epoch": 930, "lr": 7.890172572017964e-05}
{"train_loss": 0.2199888676404953, "global_step": 82820, "epoch": 930, "lr": 7.890125264602843e-05}
{"train_loss": 0.14407514035701752, "global_step": 82821, "epoch": 930, "lr": 7.89007795679918e-05}
{"train_loss": 0.17273053526878357, "global_step": 82822, "epoch": 930, "lr": 7.89003064860698e-05}
{"train_loss": 0.18481238186359406, "global_step": 82823, "epoch": 930, "lr": 7.889983340026253e-05}
{"train_loss": 0.20999915897846222, "global_step": 82824, "epoch": 930, "lr": 7.889936031057e-05}
{"train_loss": 0.24471427500247955, "global_step": 82825, "epoch": 930, "lr": 7.889888721699233e-05}
{"train_loss": 0.1406150609254837, "global_step": 82826, "epoch": 930, "lr": 7.889841411952954e-05}
{"train_loss": 0.2147812694311142, "global_step": 82827, "epoch": 930, "lr": 7.889794101818171e-05}
{"train_loss": 0.16823463141918182, "global_step": 82828, "epoch": 930, "lr": 7.889746791294892e-05}
{"train_loss": 0.2336197942495346, "global_step": 82829, "epoch": 930, "lr": 7.889699480383119e-05}
{"train_loss": 0.2330739051103592, "global_step": 82830, "epoch": 930, "lr": 7.889652169082864e-05}
{"train_loss": 0.19752618670463562, "global_step": 82831, "epoch": 930, "lr": 7.88960485739413e-05}
{"train_loss": 0.21137087047100067, "global_step": 82832, "epoch": 930, "lr": 7.889557545316923e-05}
{"train_loss": 0.13875852525234222, "global_step": 82833, "epoch": 930, "lr": 7.88951023285125e-05}
{"train_loss": 0.16001826524734497, "global_step": 82834, "epoch": 930, "lr": 7.889462919997117e-05}
{"train_loss": 0.12009746581315994, "global_step": 82835, "epoch": 930, "lr": 7.889415606754533e-05}
{"train_loss": 0.25969865918159485, "global_step": 82836, "epoch": 930, "lr": 7.889368293123502e-05}
{"train_loss": 0.1886516958475113, "global_step": 82837, "epoch": 930, "lr": 7.889320979104028e-05}
{"train_loss": 0.1543751358985901, "global_step": 82838, "epoch": 930, "lr": 7.889273664696125e-05}
{"train_loss": 0.19400514662265778, "global_step": 82839, "epoch": 930, "lr": 7.889226349899789e-05}
{"train_loss": 0.14995412528514862, "global_step": 82840, "epoch": 930, "lr": 7.889179034715036e-05}
{"train_loss": 0.19762353599071503, "global_step": 82841, "epoch": 930, "lr": 7.889131719141864e-05}
{"train_loss": 0.23997290432453156, "global_step": 82842, "epoch": 930, "lr": 7.889084403180286e-05}
{"train_loss": 0.2658958435058594, "global_step": 82843, "epoch": 930, "lr": 7.889037086830304e-05}
{"train_loss": 0.1983514130115509, "global_step": 82844, "epoch": 930, "lr": 7.888989770091928e-05}
{"train_loss": 0.2672884166240692, "global_step": 82845, "epoch": 930, "lr": 7.888942452965162e-05}
{"train_loss": 0.2631932199001312, "global_step": 82846, "epoch": 930, "lr": 7.888895135450014e-05}
{"train_loss": 0.23677866160869598, "global_step": 82847, "epoch": 930, "lr": 7.888847817546488e-05}
{"train_loss": 0.1869218796491623, "global_step": 82848, "epoch": 930, "lr": 7.888800499254592e-05}
{"train_loss": 0.20602379739284515, "global_step": 82849, "epoch": 930, "lr": 7.888753180574331e-05}
{"train_loss": 0.30166178941726685, "global_step": 82850, "epoch": 930, "lr": 7.888705861505714e-05}
{"train_loss": 0.1672164648771286, "global_step": 82851, "epoch": 930, "lr": 7.888658542048744e-05}
{"train_loss": 0.19303397834300995, "global_step": 82852, "epoch": 930, "lr": 7.888611222203431e-05}
{"train_loss": 0.11187098175287247, "global_step": 82853, "epoch": 930, "lr": 7.888563901969778e-05}
{"train_loss": 0.20454946160316467, "global_step": 82854, "epoch": 930, "lr": 7.888516581347794e-05}
{"train_loss": 0.15185290575027466, "global_step": 82855, "epoch": 930, "lr": 7.888469260337484e-05}
{"train_loss": 0.11839471012353897, "global_step": 82856, "epoch": 930, "lr": 7.888421938938853e-05}
{"train_loss": 0.17476113140583038, "global_step": 82857, "epoch": 930, "lr": 7.888374617151909e-05}
{"train_loss": 0.19480373308564838, "global_step": 82858, "epoch": 930, "lr": 7.888327294976659e-05, "val_loss": 4.148010730743408, "train_action_mse_error": 22.34446907043457}
{"train_loss": 0.20520713925361633, "global_step": 82859, "epoch": 931, "lr": 7.888279972413108e-05}
{"train_loss": 0.17907913029193878, "global_step": 82860, "epoch": 931, "lr": 7.888232649461263e-05}
{"train_loss": 0.18446791172027588, "global_step": 82861, "epoch": 931, "lr": 7.888185326121131e-05}
{"train_loss": 0.23245450854301453, "global_step": 82862, "epoch": 931, "lr": 7.888138002392718e-05}
{"train_loss": 0.2326555997133255, "global_step": 82863, "epoch": 931, "lr": 7.88809067827603e-05}
{"train_loss": 0.17663119733333588, "global_step": 82864, "epoch": 931, "lr": 7.888043353771073e-05}
{"train_loss": 0.16456793248653412, "global_step": 82865, "epoch": 931, "lr": 7.887996028877854e-05}
{"train_loss": 0.24151711165905, "global_step": 82866, "epoch": 931, "lr": 7.887948703596378e-05}
{"train_loss": 0.18256814777851105, "global_step": 82867, "epoch": 931, "lr": 7.887901377926655e-05}
{"train_loss": 0.2217286080121994, "global_step": 82868, "epoch": 931, "lr": 7.887854051868686e-05}
{"train_loss": 0.22847290337085724, "global_step": 82869, "epoch": 931, "lr": 7.887806725422481e-05}
{"train_loss": 0.13618426024913788, "global_step": 82870, "epoch": 931, "lr": 7.887759398588046e-05}
{"train_loss": 0.31182047724723816, "global_step": 82871, "epoch": 931, "lr": 7.887712071365387e-05}
{"train_loss": 0.15409457683563232, "global_step": 82872, "epoch": 931, "lr": 7.88766474375451e-05}
{"train_loss": 0.1887439489364624, "global_step": 82873, "epoch": 931, "lr": 7.887617415755422e-05}
{"train_loss": 0.1849241405725479, "global_step": 82874, "epoch": 931, "lr": 7.88757008736813e-05}
{"train_loss": 0.1721508502960205, "global_step": 82875, "epoch": 931, "lr": 7.887522758592637e-05}
{"train_loss": 0.20998147130012512, "global_step": 82876, "epoch": 931, "lr": 7.887475429428955e-05}
{"train_loss": 0.1834089308977127, "global_step": 82877, "epoch": 931, "lr": 7.887428099877085e-05}
{"train_loss": 0.1842554211616516, "global_step": 82878, "epoch": 931, "lr": 7.887380769937036e-05}
{"train_loss": 0.21427558362483978, "global_step": 82879, "epoch": 931, "lr": 7.887333439608813e-05}
{"train_loss": 0.23906467854976654, "global_step": 82880, "epoch": 931, "lr": 7.887286108892422e-05}
{"train_loss": 0.29174110293388367, "global_step": 82881, "epoch": 931, "lr": 7.887238777787874e-05}
{"train_loss": 0.4253620207309723, "global_step": 82882, "epoch": 931, "lr": 7.88719144629517e-05}
{"train_loss": 0.217854306101799, "global_step": 82883, "epoch": 931, "lr": 7.887144114414319e-05}
{"train_loss": 0.17569677531719208, "global_step": 82884, "epoch": 931, "lr": 7.887096782145325e-05}
{"train_loss": 0.2230420857667923, "global_step": 82885, "epoch": 931, "lr": 7.887049449488198e-05}
{"train_loss": 0.18736831843852997, "global_step": 82886, "epoch": 931, "lr": 7.887002116442942e-05}
{"train_loss": 0.18430769443511963, "global_step": 82887, "epoch": 931, "lr": 7.886954783009564e-05}
{"train_loss": 0.24097886681556702, "global_step": 82888, "epoch": 931, "lr": 7.886907449188069e-05}
{"train_loss": 0.23223403096199036, "global_step": 82889, "epoch": 931, "lr": 7.886860114978465e-05}
{"train_loss": 0.2001471221446991, "global_step": 82890, "epoch": 931, "lr": 7.886812780380758e-05}
{"train_loss": 0.1391676813364029, "global_step": 82891, "epoch": 931, "lr": 7.886765445394955e-05}
{"train_loss": 0.20997124910354614, "global_step": 82892, "epoch": 931, "lr": 7.88671811002106e-05}
{"train_loss": 0.22339387238025665, "global_step": 82893, "epoch": 931, "lr": 7.886670774259081e-05}
{"train_loss": 0.12327427417039871, "global_step": 82894, "epoch": 931, "lr": 7.886623438109026e-05}
{"train_loss": 0.22382467985153198, "global_step": 82895, "epoch": 931, "lr": 7.886576101570897e-05}
{"train_loss": 0.2188928872346878, "global_step": 82896, "epoch": 931, "lr": 7.886528764644705e-05}
{"train_loss": 0.15171779692173004, "global_step": 82897, "epoch": 931, "lr": 7.886481427330454e-05}
{"train_loss": 0.23001185059547424, "global_step": 82898, "epoch": 931, "lr": 7.88643408962815e-05}
{"train_loss": 0.1430976688861847, "global_step": 82899, "epoch": 931, "lr": 7.886386751537802e-05}
{"train_loss": 0.22718220949172974, "global_step": 82900, "epoch": 931, "lr": 7.886339413059414e-05}
{"train_loss": 0.21185748279094696, "global_step": 82901, "epoch": 931, "lr": 7.886292074192992e-05}
{"train_loss": 0.19965246319770813, "global_step": 82902, "epoch": 931, "lr": 7.886244734938543e-05}
{"train_loss": 0.22218850255012512, "global_step": 82903, "epoch": 931, "lr": 7.886197395296074e-05}
{"train_loss": 0.1700335443019867, "global_step": 82904, "epoch": 931, "lr": 7.886150055265591e-05}
{"train_loss": 0.2398153394460678, "global_step": 82905, "epoch": 931, "lr": 7.886102714847099e-05}
{"train_loss": 0.24081702530384064, "global_step": 82906, "epoch": 931, "lr": 7.886055374040607e-05}
{"train_loss": 0.18117763102054596, "global_step": 82907, "epoch": 931, "lr": 7.886008032846121e-05}
{"train_loss": 0.197774276137352, "global_step": 82908, "epoch": 931, "lr": 7.885960691263645e-05}
{"train_loss": 0.16575154662132263, "global_step": 82909, "epoch": 931, "lr": 7.885913349293186e-05}
{"train_loss": 0.25055140256881714, "global_step": 82910, "epoch": 931, "lr": 7.885866006934752e-05}
{"train_loss": 0.26197394728660583, "global_step": 82911, "epoch": 931, "lr": 7.885818664188349e-05}
{"train_loss": 0.2227366715669632, "global_step": 82912, "epoch": 931, "lr": 7.885771321053983e-05}
{"train_loss": 0.19409294426441193, "global_step": 82913, "epoch": 931, "lr": 7.885723977531661e-05}
{"train_loss": 0.19030363857746124, "global_step": 82914, "epoch": 931, "lr": 7.885676633621387e-05}
{"train_loss": 0.19035375118255615, "global_step": 82915, "epoch": 931, "lr": 7.885629289323167e-05}
{"train_loss": 0.17807428538799286, "global_step": 82916, "epoch": 931, "lr": 7.885581944637012e-05}
{"train_loss": 0.20699137449264526, "global_step": 82917, "epoch": 931, "lr": 7.885534599562926e-05}
{"train_loss": 0.18095019459724426, "global_step": 82918, "epoch": 931, "lr": 7.885487254100915e-05}
{"train_loss": 0.21198715269565582, "global_step": 82919, "epoch": 931, "lr": 7.885439908250985e-05}
{"train_loss": 0.19575130939483643, "global_step": 82920, "epoch": 931, "lr": 7.88539256201314e-05}
{"train_loss": 0.20431387424468994, "global_step": 82921, "epoch": 931, "lr": 7.885345215387392e-05}
{"train_loss": 0.2719913423061371, "global_step": 82922, "epoch": 931, "lr": 7.885297868373746e-05}
{"train_loss": 0.17354440689086914, "global_step": 82923, "epoch": 931, "lr": 7.885250520972205e-05}
{"train_loss": 0.1711096465587616, "global_step": 82924, "epoch": 931, "lr": 7.885203173182776e-05}
{"train_loss": 0.11734988540410995, "global_step": 82925, "epoch": 931, "lr": 7.885155825005468e-05}
{"train_loss": 0.18977707624435425, "global_step": 82926, "epoch": 931, "lr": 7.885108476440286e-05}
{"train_loss": 0.22936762869358063, "global_step": 82927, "epoch": 931, "lr": 7.885061127487236e-05}
{"train_loss": 0.09950604289770126, "global_step": 82928, "epoch": 931, "lr": 7.885013778146325e-05}
{"train_loss": 0.10933378338813782, "global_step": 82929, "epoch": 931, "lr": 7.884966428417559e-05}
{"train_loss": 0.15807360410690308, "global_step": 82930, "epoch": 931, "lr": 7.884919078300944e-05}
{"train_loss": 0.18635565042495728, "global_step": 82931, "epoch": 931, "lr": 7.884871727796488e-05}
{"train_loss": 0.16541163623332977, "global_step": 82932, "epoch": 931, "lr": 7.884824376904195e-05}
{"train_loss": 0.18237963318824768, "global_step": 82933, "epoch": 931, "lr": 7.884777025624073e-05}
{"train_loss": 0.14452067017555237, "global_step": 82934, "epoch": 931, "lr": 7.884729673956127e-05}
{"train_loss": 0.1833994835615158, "global_step": 82935, "epoch": 931, "lr": 7.884682321900365e-05}
{"train_loss": 0.23423218727111816, "global_step": 82936, "epoch": 931, "lr": 7.884634969456792e-05}
{"train_loss": 0.13330362737178802, "global_step": 82937, "epoch": 931, "lr": 7.884587616625417e-05}
{"train_loss": 0.21744675934314728, "global_step": 82938, "epoch": 931, "lr": 7.884540263406241e-05}
{"train_loss": 0.17891865968704224, "global_step": 82939, "epoch": 931, "lr": 7.884492909799277e-05}
{"train_loss": 0.18080423772335052, "global_step": 82940, "epoch": 931, "lr": 7.884445555804526e-05}
{"train_loss": 0.14737875759601593, "global_step": 82941, "epoch": 931, "lr": 7.884398201421997e-05}
{"train_loss": 0.14380358159542084, "global_step": 82942, "epoch": 931, "lr": 7.884350846651696e-05}
{"train_loss": 0.1703733503818512, "global_step": 82943, "epoch": 931, "lr": 7.88430349149363e-05}
{"train_loss": 0.20047476887702942, "global_step": 82944, "epoch": 931, "lr": 7.884256135947803e-05}
{"train_loss": 0.15180791914463043, "global_step": 82945, "epoch": 931, "lr": 7.884208780014224e-05}
{"train_loss": 0.10001624375581741, "global_step": 82946, "epoch": 931, "lr": 7.884161423692898e-05}
{"train_loss": 0.1962125316429674, "global_step": 82947, "epoch": 931, "lr": 7.884114066983831e-05, "val_loss": 4.2286906242370605}
{"train_loss": 0.25559139251708984, "global_step": 82948, "epoch": 932, "lr": 7.884066709887031e-05}
{"train_loss": 0.23362207412719727, "global_step": 82949, "epoch": 932, "lr": 7.884019352402504e-05}
{"train_loss": 0.16924938559532166, "global_step": 82950, "epoch": 932, "lr": 7.883971994530252e-05}
{"train_loss": 0.1365600824356079, "global_step": 82951, "epoch": 932, "lr": 7.883924636270288e-05}
{"train_loss": 0.203258216381073, "global_step": 82952, "epoch": 932, "lr": 7.883877277622615e-05}
{"train_loss": 0.2213350534439087, "global_step": 82953, "epoch": 932, "lr": 7.883829918587241e-05}
{"train_loss": 0.25744470953941345, "global_step": 82954, "epoch": 932, "lr": 7.88378255916417e-05}
{"train_loss": 0.13677003979682922, "global_step": 82955, "epoch": 932, "lr": 7.88373519935341e-05}
{"train_loss": 0.1385229527950287, "global_step": 82956, "epoch": 932, "lr": 7.883687839154967e-05}
{"train_loss": 0.1814713329076767, "global_step": 82957, "epoch": 932, "lr": 7.883640478568848e-05}
{"train_loss": 0.16109800338745117, "global_step": 82958, "epoch": 932, "lr": 7.883593117595058e-05}
{"train_loss": 0.2060556411743164, "global_step": 82959, "epoch": 932, "lr": 7.883545756233603e-05}
{"train_loss": 0.2953752875328064, "global_step": 82960, "epoch": 932, "lr": 7.883498394484492e-05}
{"train_loss": 0.17082034051418304, "global_step": 82961, "epoch": 932, "lr": 7.883451032347728e-05}
{"train_loss": 0.27114808559417725, "global_step": 82962, "epoch": 932, "lr": 7.883403669823322e-05}
{"train_loss": 0.1869741976261139, "global_step": 82963, "epoch": 932, "lr": 7.883356306911276e-05}
{"train_loss": 0.22492900490760803, "global_step": 82964, "epoch": 932, "lr": 7.883308943611597e-05}
{"train_loss": 0.21050499379634857, "global_step": 82965, "epoch": 932, "lr": 7.883261579924293e-05}
{"train_loss": 0.14227113127708435, "global_step": 82966, "epoch": 932, "lr": 7.88321421584937e-05}
{"train_loss": 0.15633060038089752, "global_step": 82967, "epoch": 932, "lr": 7.883166851386833e-05}
{"train_loss": 0.17205332219600677, "global_step": 82968, "epoch": 932, "lr": 7.88311948653669e-05}
{"train_loss": 0.15969276428222656, "global_step": 82969, "epoch": 932, "lr": 7.883072121298947e-05}
{"train_loss": 0.15471871197223663, "global_step": 82970, "epoch": 932, "lr": 7.883024755673609e-05}
{"train_loss": 0.13310785591602325, "global_step": 82971, "epoch": 932, "lr": 7.882977389660685e-05}
{"train_loss": 0.20054005086421967, "global_step": 82972, "epoch": 932, "lr": 7.882930023260177e-05}
{"train_loss": 0.18632462620735168, "global_step": 82973, "epoch": 932, "lr": 7.882882656472098e-05}
{"train_loss": 0.1545591950416565, "global_step": 82974, "epoch": 932, "lr": 7.882835289296448e-05}
{"train_loss": 0.15419389307498932, "global_step": 82975, "epoch": 932, "lr": 7.882787921733237e-05}
{"train_loss": 0.24505652487277985, "global_step": 82976, "epoch": 932, "lr": 7.882740553782471e-05}
{"train_loss": 0.15718670189380646, "global_step": 82977, "epoch": 932, "lr": 7.882693185444155e-05}
{"train_loss": 0.25095248222351074, "global_step": 82978, "epoch": 932, "lr": 7.882645816718295e-05}
{"train_loss": 0.1437469869852066, "global_step": 82979, "epoch": 932, "lr": 7.882598447604899e-05}
{"train_loss": 0.15462008118629456, "global_step": 82980, "epoch": 932, "lr": 7.882551078103973e-05}
{"train_loss": 0.19722150266170502, "global_step": 82981, "epoch": 932, "lr": 7.882503708215522e-05}
{"train_loss": 0.14413991570472717, "global_step": 82982, "epoch": 932, "lr": 7.882456337939556e-05}
{"train_loss": 0.20033541321754456, "global_step": 82983, "epoch": 932, "lr": 7.882408967276077e-05}
{"train_loss": 0.13277168571949005, "global_step": 82984, "epoch": 932, "lr": 7.882361596225095e-05}
{"train_loss": 0.25701770186424255, "global_step": 82985, "epoch": 932, "lr": 7.882314224786612e-05}
{"train_loss": 0.1879749298095703, "global_step": 82986, "epoch": 932, "lr": 7.882266852960639e-05}
{"train_loss": 0.15012316405773163, "global_step": 82987, "epoch": 932, "lr": 7.882219480747179e-05}
{"train_loss": 0.1798248142004013, "global_step": 82988, "epoch": 932, "lr": 7.88217210814624e-05}
{"train_loss": 0.22545017302036285, "global_step": 82989, "epoch": 932, "lr": 7.882124735157828e-05}
{"train_loss": 0.22440850734710693, "global_step": 82990, "epoch": 932, "lr": 7.88207736178195e-05}
{"train_loss": 0.22067739069461823, "global_step": 82991, "epoch": 932, "lr": 7.882029988018611e-05}
{"train_loss": 0.2478361874818802, "global_step": 82992, "epoch": 932, "lr": 7.88198261386782e-05}
{"train_loss": 0.15693503618240356, "global_step": 82993, "epoch": 932, "lr": 7.881935239329581e-05}
{"train_loss": 0.2383500188589096, "global_step": 82994, "epoch": 932, "lr": 7.8818878644039e-05}
{"train_loss": 0.21320393681526184, "global_step": 82995, "epoch": 932, "lr": 7.881840489090783e-05}
{"train_loss": 0.1602727472782135, "global_step": 82996, "epoch": 932, "lr": 7.88179311339024e-05}
{"train_loss": 0.1858820617198944, "global_step": 82997, "epoch": 932, "lr": 7.881745737302276e-05}
{"train_loss": 0.13113339245319366, "global_step": 82998, "epoch": 932, "lr": 7.881698360826894e-05}
{"train_loss": 0.10974814742803574, "global_step": 82999, "epoch": 932, "lr": 7.881650983964104e-05}
{"train_loss": 0.1310848444700241, "global_step": 83000, "epoch": 932, "lr": 7.88160360671391e-05}
{"train_loss": 0.23522143065929413, "global_step": 83001, "epoch": 932, "lr": 7.881556229076321e-05}
{"train_loss": 0.24311119318008423, "global_step": 83002, "epoch": 932, "lr": 7.881508851051341e-05}
{"train_loss": 0.22981232404708862, "global_step": 83003, "epoch": 932, "lr": 7.881461472638978e-05}
{"train_loss": 0.19319815933704376, "global_step": 83004, "epoch": 932, "lr": 7.881414093839237e-05}
{"train_loss": 0.16543950140476227, "global_step": 83005, "epoch": 932, "lr": 7.881366714652125e-05}
{"train_loss": 0.1820208579301834, "global_step": 83006, "epoch": 932, "lr": 7.881319335077648e-05}
{"train_loss": 0.14280255138874054, "global_step": 83007, "epoch": 932, "lr": 7.881271955115812e-05}
{"train_loss": 0.21522630751132965, "global_step": 83008, "epoch": 932, "lr": 7.881224574766625e-05}
{"train_loss": 0.2635912597179413, "global_step": 83009, "epoch": 932, "lr": 7.881177194030093e-05}
{"train_loss": 0.28156912326812744, "global_step": 83010, "epoch": 932, "lr": 7.881129812906222e-05}
{"train_loss": 0.1827239990234375, "global_step": 83011, "epoch": 932, "lr": 7.881082431395019e-05}
{"train_loss": 0.19312767684459686, "global_step": 83012, "epoch": 932, "lr": 7.881035049496488e-05}
{"train_loss": 0.17154155671596527, "global_step": 83013, "epoch": 932, "lr": 7.880987667210637e-05}
{"train_loss": 0.16073036193847656, "global_step": 83014, "epoch": 932, "lr": 7.880940284537473e-05}
{"train_loss": 0.15236780047416687, "global_step": 83015, "epoch": 932, "lr": 7.880892901477e-05}
{"train_loss": 0.2106241136789322, "global_step": 83016, "epoch": 932, "lr": 7.880845518029228e-05}
{"train_loss": 0.26656657457351685, "global_step": 83017, "epoch": 932, "lr": 7.88079813419416e-05}
{"train_loss": 0.0979369655251503, "global_step": 83018, "epoch": 932, "lr": 7.880750749971806e-05}
{"train_loss": 0.22111085057258606, "global_step": 83019, "epoch": 932, "lr": 7.880703365362168e-05}
{"train_loss": 0.2523159086704254, "global_step": 83020, "epoch": 932, "lr": 7.880655980365256e-05}
{"train_loss": 0.30252233147621155, "global_step": 83021, "epoch": 932, "lr": 7.880608594981074e-05}
{"train_loss": 0.1877814531326294, "global_step": 83022, "epoch": 932, "lr": 7.880561209209629e-05}
{"train_loss": 0.13492867350578308, "global_step": 83023, "epoch": 932, "lr": 7.880513823050929e-05}
{"train_loss": 0.1562180370092392, "global_step": 83024, "epoch": 932, "lr": 7.880466436504979e-05}
{"train_loss": 0.13945604860782623, "global_step": 83025, "epoch": 932, "lr": 7.880419049571783e-05}
{"train_loss": 0.19819524884223938, "global_step": 83026, "epoch": 932, "lr": 7.880371662251351e-05}
{"train_loss": 0.23146063089370728, "global_step": 83027, "epoch": 932, "lr": 7.88032427454369e-05}
{"train_loss": 0.2672670781612396, "global_step": 83028, "epoch": 932, "lr": 7.880276886448804e-05}
{"train_loss": 0.17075613141059875, "global_step": 83029, "epoch": 932, "lr": 7.880229497966698e-05}
{"train_loss": 0.18924970924854279, "global_step": 83030, "epoch": 932, "lr": 7.880182109097381e-05}
{"train_loss": 0.2345963418483734, "global_step": 83031, "epoch": 932, "lr": 7.880134719840858e-05}
{"train_loss": 0.11071807891130447, "global_step": 83032, "epoch": 932, "lr": 7.880087330197136e-05}
{"train_loss": 0.21400563418865204, "global_step": 83033, "epoch": 932, "lr": 7.880039940166223e-05}
{"train_loss": 0.17672951519489288, "global_step": 83034, "epoch": 932, "lr": 7.879992549748123e-05}
{"train_loss": 0.14256490767002106, "global_step": 83035, "epoch": 932, "lr": 7.879945158942842e-05}
{"train_loss": 0.1924897495113062, "global_step": 83036, "epoch": 932, "lr": 7.879897767750391e-05, "val_loss": 4.103598594665527}
{"train_loss": 0.13365589082241058, "global_step": 83037, "epoch": 933, "lr": 7.879850376170768e-05}
{"train_loss": 0.266854852437973, "global_step": 83038, "epoch": 933, "lr": 7.879802984203988e-05}
{"train_loss": 0.1107778325676918, "global_step": 83039, "epoch": 933, "lr": 7.879755591850051e-05}
{"train_loss": 0.20847360789775848, "global_step": 83040, "epoch": 933, "lr": 7.879708199108968e-05}
{"train_loss": 0.16878275573253632, "global_step": 83041, "epoch": 933, "lr": 7.879660805980742e-05}
{"train_loss": 0.16656741499900818, "global_step": 83042, "epoch": 933, "lr": 7.87961341246538e-05}
{"train_loss": 0.18806838989257812, "global_step": 83043, "epoch": 933, "lr": 7.879566018562891e-05}
{"train_loss": 0.1579691469669342, "global_step": 83044, "epoch": 933, "lr": 7.879518624273279e-05}
{"train_loss": 0.2018250972032547, "global_step": 83045, "epoch": 933, "lr": 7.879471229596549e-05}
{"train_loss": 0.1644163578748703, "global_step": 83046, "epoch": 933, "lr": 7.879423834532711e-05}
{"train_loss": 0.12393947690725327, "global_step": 83047, "epoch": 933, "lr": 7.87937643908177e-05}
{"train_loss": 0.17484033107757568, "global_step": 83048, "epoch": 933, "lr": 7.879329043243731e-05}
{"train_loss": 0.20518596470355988, "global_step": 83049, "epoch": 933, "lr": 7.879281647018602e-05}
{"train_loss": 0.17251244187355042, "global_step": 83050, "epoch": 933, "lr": 7.879234250406386e-05}
{"train_loss": 0.18781238794326782, "global_step": 83051, "epoch": 933, "lr": 7.879186853407094e-05}
{"train_loss": 0.3027602434158325, "global_step": 83052, "epoch": 933, "lr": 7.879139456020733e-05}
{"train_loss": 0.2564656734466553, "global_step": 83053, "epoch": 933, "lr": 7.879092058247304e-05}
{"train_loss": 0.18739020824432373, "global_step": 83054, "epoch": 933, "lr": 7.879044660086817e-05}
{"train_loss": 0.21518857777118683, "global_step": 83055, "epoch": 933, "lr": 7.878997261539276e-05}
{"train_loss": 0.13046783208847046, "global_step": 83056, "epoch": 933, "lr": 7.87894986260469e-05}
{"train_loss": 0.2120937556028366, "global_step": 83057, "epoch": 933, "lr": 7.878902463283064e-05}
{"train_loss": 0.1522803157567978, "global_step": 83058, "epoch": 933, "lr": 7.878855063574405e-05}
{"train_loss": 0.17981880903244019, "global_step": 83059, "epoch": 933, "lr": 7.878807663478719e-05}
{"train_loss": 0.20299839973449707, "global_step": 83060, "epoch": 933, "lr": 7.878760262996012e-05}
{"train_loss": 0.2096029371023178, "global_step": 83061, "epoch": 933, "lr": 7.878712862126291e-05}
{"train_loss": 0.20211827754974365, "global_step": 83062, "epoch": 933, "lr": 7.878665460869562e-05}
{"train_loss": 0.15046164393424988, "global_step": 83063, "epoch": 933, "lr": 7.878618059225833e-05}
{"train_loss": 0.16717296838760376, "global_step": 83064, "epoch": 933, "lr": 7.878570657195105e-05}
{"train_loss": 0.17881819605827332, "global_step": 83065, "epoch": 933, "lr": 7.878523254777392e-05}
{"train_loss": 0.14076536893844604, "global_step": 83066, "epoch": 933, "lr": 7.878475851972695e-05}
{"train_loss": 0.1447867900133133, "global_step": 83067, "epoch": 933, "lr": 7.878428448781023e-05}
{"train_loss": 0.13702881336212158, "global_step": 83068, "epoch": 933, "lr": 7.87838104520238e-05}
{"train_loss": 0.15170210599899292, "global_step": 83069, "epoch": 933, "lr": 7.878333641236777e-05}
{"train_loss": 0.13396212458610535, "global_step": 83070, "epoch": 933, "lr": 7.878286236884214e-05}
{"train_loss": 0.2056824266910553, "global_step": 83071, "epoch": 933, "lr": 7.878238832144701e-05}
{"train_loss": 0.19473673403263092, "global_step": 83072, "epoch": 933, "lr": 7.878191427018245e-05}
{"train_loss": 0.22237884998321533, "global_step": 83073, "epoch": 933, "lr": 7.87814402150485e-05}
{"train_loss": 0.16001924872398376, "global_step": 83074, "epoch": 933, "lr": 7.878096615604524e-05}
{"train_loss": 0.1540142297744751, "global_step": 83075, "epoch": 933, "lr": 7.878049209317273e-05}
{"train_loss": 0.14864225685596466, "global_step": 83076, "epoch": 933, "lr": 7.878001802643103e-05}
{"train_loss": 0.2022312730550766, "global_step": 83077, "epoch": 933, "lr": 7.87795439558202e-05}
{"train_loss": 0.17353636026382446, "global_step": 83078, "epoch": 933, "lr": 7.877906988134032e-05}
{"train_loss": 0.27233976125717163, "global_step": 83079, "epoch": 933, "lr": 7.877859580299145e-05}
{"train_loss": 0.2700996398925781, "global_step": 83080, "epoch": 933, "lr": 7.877812172077365e-05}
{"train_loss": 0.18527406454086304, "global_step": 83081, "epoch": 933, "lr": 7.877764763468698e-05}
{"train_loss": 0.1388094574213028, "global_step": 83082, "epoch": 933, "lr": 7.87771735447315e-05}
{"train_loss": 0.13464081287384033, "global_step": 83083, "epoch": 933, "lr": 7.877669945090729e-05}
{"train_loss": 0.18521003425121307, "global_step": 83084, "epoch": 933, "lr": 7.87762253532144e-05}
{"train_loss": 0.21861495077610016, "global_step": 83085, "epoch": 933, "lr": 7.877575125165289e-05}
{"train_loss": 0.1926206648349762, "global_step": 83086, "epoch": 933, "lr": 7.877527714622283e-05}
{"train_loss": 0.21839097142219543, "global_step": 83087, "epoch": 933, "lr": 7.87748030369243e-05}
{"train_loss": 0.19630616903305054, "global_step": 83088, "epoch": 933, "lr": 7.877432892375733e-05}
{"train_loss": 0.14069928228855133, "global_step": 83089, "epoch": 933, "lr": 7.877385480672201e-05}
{"train_loss": 0.20411992073059082, "global_step": 83090, "epoch": 933, "lr": 7.87733806858184e-05}
{"train_loss": 0.1812560260295868, "global_step": 83091, "epoch": 933, "lr": 7.877290656104655e-05}
{"train_loss": 0.15970298647880554, "global_step": 83092, "epoch": 933, "lr": 7.877243243240654e-05}
{"train_loss": 0.16663271188735962, "global_step": 83093, "epoch": 933, "lr": 7.877195829989843e-05}
{"train_loss": 0.14912551641464233, "global_step": 83094, "epoch": 933, "lr": 7.877148416352228e-05}
{"train_loss": 0.1694093644618988, "global_step": 83095, "epoch": 933, "lr": 7.877101002327816e-05}
{"train_loss": 0.19560855627059937, "global_step": 83096, "epoch": 933, "lr": 7.87705358791661e-05}
{"train_loss": 0.16848158836364746, "global_step": 83097, "epoch": 933, "lr": 7.877006173118622e-05}
{"train_loss": 0.19599725306034088, "global_step": 83098, "epoch": 933, "lr": 7.876958757933854e-05}
{"train_loss": 0.20052097737789154, "global_step": 83099, "epoch": 933, "lr": 7.876911342362316e-05}
{"train_loss": 0.18334905803203583, "global_step": 83100, "epoch": 933, "lr": 7.87686392640401e-05}
{"train_loss": 0.167031392455101, "global_step": 83101, "epoch": 933, "lr": 7.876816510058947e-05}
{"train_loss": 0.2608547806739807, "global_step": 83102, "epoch": 933, "lr": 7.87676909332713e-05}
{"train_loss": 0.2485790103673935, "global_step": 83103, "epoch": 933, "lr": 7.876721676208566e-05}
{"train_loss": 0.2436075061559677, "global_step": 83104, "epoch": 933, "lr": 7.876674258703262e-05}
{"train_loss": 0.15615154802799225, "global_step": 83105, "epoch": 933, "lr": 7.876626840811225e-05}
{"train_loss": 0.15238776803016663, "global_step": 83106, "epoch": 933, "lr": 7.87657942253246e-05}
{"train_loss": 0.20301453769207, "global_step": 83107, "epoch": 933, "lr": 7.876532003866974e-05}
{"train_loss": 0.19849735498428345, "global_step": 83108, "epoch": 933, "lr": 7.876484584814773e-05}
{"train_loss": 0.18410654366016388, "global_step": 83109, "epoch": 933, "lr": 7.876437165375864e-05}
{"train_loss": 0.13357101380825043, "global_step": 83110, "epoch": 933, "lr": 7.876389745550254e-05}
{"train_loss": 0.20184975862503052, "global_step": 83111, "epoch": 933, "lr": 7.876342325337947e-05}
{"train_loss": 0.23444165289402008, "global_step": 83112, "epoch": 933, "lr": 7.87629490473895e-05}
{"train_loss": 0.2008402943611145, "global_step": 83113, "epoch": 933, "lr": 7.876247483753273e-05}
{"train_loss": 0.2406606525182724, "global_step": 83114, "epoch": 933, "lr": 7.876200062380918e-05}
{"train_loss": 0.19745799899101257, "global_step": 83115, "epoch": 933, "lr": 7.876152640621892e-05}
{"train_loss": 0.22554267942905426, "global_step": 83116, "epoch": 933, "lr": 7.876105218476205e-05}
{"train_loss": 0.2362966686487198, "global_step": 83117, "epoch": 933, "lr": 7.876057795943858e-05}
{"train_loss": 0.19275127351284027, "global_step": 83118, "epoch": 933, "lr": 7.876010373024862e-05}
{"train_loss": 0.1810884028673172, "global_step": 83119, "epoch": 933, "lr": 7.87596294971922e-05}
{"train_loss": 0.25413352251052856, "global_step": 83120, "epoch": 933, "lr": 7.87591552602694e-05}
{"train_loss": 0.148549884557724, "global_step": 83121, "epoch": 933, "lr": 7.875868101948029e-05}
{"train_loss": 0.1260157823562622, "global_step": 83122, "epoch": 933, "lr": 7.875820677482493e-05}
{"train_loss": 0.20616841316223145, "global_step": 83123, "epoch": 933, "lr": 7.875773252630337e-05}
{"train_loss": 0.19563926756381989, "global_step": 83124, "epoch": 933, "lr": 7.875725827391569e-05}
{"train_loss": 0.18682906081837214, "global_step": 83125, "epoch": 933, "lr": 7.875678401766194e-05, "val_loss": 4.208810329437256}
{"train_loss": 0.1614781618118286, "global_step": 83126, "epoch": 934, "lr": 7.87563097575422e-05}
{"train_loss": 0.16144834458827972, "global_step": 83127, "epoch": 934, "lr": 7.875583549355651e-05}
{"train_loss": 0.17661012709140778, "global_step": 83128, "epoch": 934, "lr": 7.875536122570497e-05}
{"train_loss": 0.21633487939834595, "global_step": 83129, "epoch": 934, "lr": 7.87548869539876e-05}
{"train_loss": 0.18826165795326233, "global_step": 83130, "epoch": 934, "lr": 7.87544126784045e-05}
{"train_loss": 0.10983134806156158, "global_step": 83131, "epoch": 934, "lr": 7.87539383989557e-05}
{"train_loss": 0.2437131702899933, "global_step": 83132, "epoch": 934, "lr": 7.875346411564131e-05}
{"train_loss": 0.1781860738992691, "global_step": 83133, "epoch": 934, "lr": 7.875298982846135e-05}
{"train_loss": 0.15964928269386292, "global_step": 83134, "epoch": 934, "lr": 7.875251553741592e-05}
{"train_loss": 0.1877867877483368, "global_step": 83135, "epoch": 934, "lr": 7.875204124250504e-05}
{"train_loss": 0.2596995532512665, "global_step": 83136, "epoch": 934, "lr": 7.87515669437288e-05}
{"train_loss": 0.20494161546230316, "global_step": 83137, "epoch": 934, "lr": 7.875109264108728e-05}
{"train_loss": 0.21739016473293304, "global_step": 83138, "epoch": 934, "lr": 7.875061833458052e-05}
{"train_loss": 0.19984722137451172, "global_step": 83139, "epoch": 934, "lr": 7.875014402420859e-05}
{"train_loss": 0.27834516763687134, "global_step": 83140, "epoch": 934, "lr": 7.874966970997155e-05}
{"train_loss": 0.1716342270374298, "global_step": 83141, "epoch": 934, "lr": 7.874919539186947e-05}
{"train_loss": 0.1727783977985382, "global_step": 83142, "epoch": 934, "lr": 7.874872106990241e-05}
{"train_loss": 0.17358312010765076, "global_step": 83143, "epoch": 934, "lr": 7.874824674407045e-05}
{"train_loss": 0.2345525473356247, "global_step": 83144, "epoch": 934, "lr": 7.874777241437362e-05}
{"train_loss": 0.11977911740541458, "global_step": 83145, "epoch": 934, "lr": 7.874729808081201e-05}
{"train_loss": 0.20649667084217072, "global_step": 83146, "epoch": 934, "lr": 7.874682374338567e-05}
{"train_loss": 0.19868116080760956, "global_step": 83147, "epoch": 934, "lr": 7.874634940209467e-05}
{"train_loss": 0.16076628863811493, "global_step": 83148, "epoch": 934, "lr": 7.874587505693907e-05}
{"train_loss": 0.1467374861240387, "global_step": 83149, "epoch": 934, "lr": 7.874540070791893e-05}
{"train_loss": 0.17065046727657318, "global_step": 83150, "epoch": 934, "lr": 7.874492635503434e-05}
{"train_loss": 0.3019374907016754, "global_step": 83151, "epoch": 934, "lr": 7.874445199828533e-05}
{"train_loss": 0.19895491003990173, "global_step": 83152, "epoch": 934, "lr": 7.8743977637672e-05}
{"train_loss": 0.20303405821323395, "global_step": 83153, "epoch": 934, "lr": 7.874350327319437e-05}
{"train_loss": 0.31540948152542114, "global_step": 83154, "epoch": 934, "lr": 7.874302890485254e-05}
{"train_loss": 0.16536694765090942, "global_step": 83155, "epoch": 934, "lr": 7.874255453264657e-05}
{"train_loss": 0.2114153951406479, "global_step": 83156, "epoch": 934, "lr": 7.87420801565765e-05}
{"train_loss": 0.2219950407743454, "global_step": 83157, "epoch": 934, "lr": 7.87416057766424e-05}
{"train_loss": 0.21335935592651367, "global_step": 83158, "epoch": 934, "lr": 7.874113139284436e-05}
{"train_loss": 0.1740080565214157, "global_step": 83159, "epoch": 934, "lr": 7.874065700518242e-05}
{"train_loss": 0.14473432302474976, "global_step": 83160, "epoch": 934, "lr": 7.874018261365663e-05}
{"train_loss": 0.16847053170204163, "global_step": 83161, "epoch": 934, "lr": 7.873970821826709e-05}
{"train_loss": 0.260749876499176, "global_step": 83162, "epoch": 934, "lr": 7.873923381901385e-05}
{"train_loss": 0.19892671704292297, "global_step": 83163, "epoch": 934, "lr": 7.873875941589697e-05}
{"train_loss": 0.16160699725151062, "global_step": 83164, "epoch": 934, "lr": 7.87382850089165e-05}
{"train_loss": 0.1971234381198883, "global_step": 83165, "epoch": 934, "lr": 7.873781059807253e-05}
{"train_loss": 0.22422872483730316, "global_step": 83166, "epoch": 934, "lr": 7.87373361833651e-05}
{"train_loss": 0.21087990701198578, "global_step": 83167, "epoch": 934, "lr": 7.873686176479429e-05}
{"train_loss": 0.16188253462314606, "global_step": 83168, "epoch": 934, "lr": 7.873638734236017e-05}
{"train_loss": 0.22742734849452972, "global_step": 83169, "epoch": 934, "lr": 7.873591291606277e-05}
{"train_loss": 0.2368440479040146, "global_step": 83170, "epoch": 934, "lr": 7.873543848590221e-05}
{"train_loss": 0.21288061141967773, "global_step": 83171, "epoch": 934, "lr": 7.87349640518785e-05}
{"train_loss": 0.1128142774105072, "global_step": 83172, "epoch": 934, "lr": 7.873448961399172e-05}
{"train_loss": 0.2060759961605072, "global_step": 83173, "epoch": 934, "lr": 7.873401517224194e-05}
{"train_loss": 0.15787829458713531, "global_step": 83174, "epoch": 934, "lr": 7.873354072662922e-05}
{"train_loss": 0.21094192564487457, "global_step": 83175, "epoch": 934, "lr": 7.873306627715364e-05}
{"train_loss": 0.17833265662193298, "global_step": 83176, "epoch": 934, "lr": 7.873259182381525e-05}
{"train_loss": 0.23336704075336456, "global_step": 83177, "epoch": 934, "lr": 7.87321173666141e-05}
{"train_loss": 0.18105168640613556, "global_step": 83178, "epoch": 934, "lr": 7.873164290555028e-05}
{"train_loss": 0.18769505620002747, "global_step": 83179, "epoch": 934, "lr": 7.873116844062382e-05}
{"train_loss": 0.19786851108074188, "global_step": 83180, "epoch": 934, "lr": 7.873069397183482e-05}
{"train_loss": 0.2670232653617859, "global_step": 83181, "epoch": 934, "lr": 7.873021949918333e-05}
{"train_loss": 0.19137191772460938, "global_step": 83182, "epoch": 934, "lr": 7.872974502266941e-05}
{"train_loss": 0.18089759349822998, "global_step": 83183, "epoch": 934, "lr": 7.872927054229312e-05}
{"train_loss": 0.21679890155792236, "global_step": 83184, "epoch": 934, "lr": 7.872879605805453e-05}
{"train_loss": 0.22813943028450012, "global_step": 83185, "epoch": 934, "lr": 7.872832156995371e-05}
{"train_loss": 0.15693674981594086, "global_step": 83186, "epoch": 934, "lr": 7.872784707799071e-05}
{"train_loss": 0.272283136844635, "global_step": 83187, "epoch": 934, "lr": 7.872737258216561e-05}
{"train_loss": 0.22772599756717682, "global_step": 83188, "epoch": 934, "lr": 7.872689808247847e-05}
{"train_loss": 0.23846080899238586, "global_step": 83189, "epoch": 934, "lr": 7.872642357892931e-05}
{"train_loss": 0.18859198689460754, "global_step": 83190, "epoch": 934, "lr": 7.872594907151826e-05}
{"train_loss": 0.18457411229610443, "global_step": 83191, "epoch": 934, "lr": 7.872547456024536e-05}
{"train_loss": 0.20879434049129486, "global_step": 83192, "epoch": 934, "lr": 7.872500004511067e-05}
{"train_loss": 0.19510434567928314, "global_step": 83193, "epoch": 934, "lr": 7.872452552611427e-05}
{"train_loss": 0.18076366186141968, "global_step": 83194, "epoch": 934, "lr": 7.872405100325618e-05}
{"train_loss": 0.16879402101039886, "global_step": 83195, "epoch": 934, "lr": 7.87235764765365e-05}
{"train_loss": 0.1740606278181076, "global_step": 83196, "epoch": 934, "lr": 7.87231019459553e-05}
{"train_loss": 0.19768840074539185, "global_step": 83197, "epoch": 934, "lr": 7.87226274115126e-05}
{"train_loss": 0.18588043749332428, "global_step": 83198, "epoch": 934, "lr": 7.872215287320853e-05}
{"train_loss": 0.14328056573867798, "global_step": 83199, "epoch": 934, "lr": 7.872167833104308e-05}
{"train_loss": 0.23930923640727997, "global_step": 83200, "epoch": 934, "lr": 7.872120378501636e-05}
{"train_loss": 0.1688217967748642, "global_step": 83201, "epoch": 934, "lr": 7.872072923512844e-05}
{"train_loss": 0.18575121462345123, "global_step": 83202, "epoch": 934, "lr": 7.872025468137936e-05}
{"train_loss": 0.15850163996219635, "global_step": 83203, "epoch": 934, "lr": 7.871978012376918e-05}
{"train_loss": 0.1955910474061966, "global_step": 83204, "epoch": 934, "lr": 7.871930556229799e-05}
{"train_loss": 0.26025763154029846, "global_step": 83205, "epoch": 934, "lr": 7.871883099696582e-05}
{"train_loss": 0.16703547537326813, "global_step": 83206, "epoch": 934, "lr": 7.871835642777277e-05}
{"train_loss": 0.22757934033870697, "global_step": 83207, "epoch": 934, "lr": 7.871788185471889e-05}
{"train_loss": 0.25282788276672363, "global_step": 83208, "epoch": 934, "lr": 7.871740727780424e-05}
{"train_loss": 0.12244139611721039, "global_step": 83209, "epoch": 934, "lr": 7.871693269702887e-05}
{"train_loss": 0.262462854385376, "global_step": 83210, "epoch": 934, "lr": 7.871645811239287e-05}
{"train_loss": 0.09852008521556854, "global_step": 83211, "epoch": 934, "lr": 7.871598352389628e-05}
{"train_loss": 0.1857912838459015, "global_step": 83212, "epoch": 934, "lr": 7.871550893153919e-05}
{"train_loss": 0.16258485615253448, "global_step": 83213, "epoch": 934, "lr": 7.871503433532165e-05}
{"train_loss": 0.19739508051215932, "global_step": 83214, "epoch": 934, "lr": 7.871455973524372e-05, "val_loss": 4.052907466888428}
{"train_loss": 0.10743667185306549, "global_step": 83215, "epoch": 935, "lr": 7.871408513130547e-05}
{"train_loss": 0.2365986704826355, "global_step": 83216, "epoch": 935, "lr": 7.871361052350696e-05}
{"train_loss": 0.21141259372234344, "global_step": 83217, "epoch": 935, "lr": 7.871313591184825e-05}
{"train_loss": 0.13855543732643127, "global_step": 83218, "epoch": 935, "lr": 7.871266129632941e-05}
{"train_loss": 0.22686061263084412, "global_step": 83219, "epoch": 935, "lr": 7.871218667695051e-05}
{"train_loss": 0.20740173757076263, "global_step": 83220, "epoch": 935, "lr": 7.871171205371159e-05}
{"train_loss": 0.22843199968338013, "global_step": 83221, "epoch": 935, "lr": 7.871123742661275e-05}
{"train_loss": 0.19679421186447144, "global_step": 83222, "epoch": 935, "lr": 7.871076279565403e-05}
{"train_loss": 0.194570854306221, "global_step": 83223, "epoch": 935, "lr": 7.871028816083548e-05}
{"train_loss": 0.15834876894950867, "global_step": 83224, "epoch": 935, "lr": 7.87098135221572e-05}
{"train_loss": 0.23943786323070526, "global_step": 83225, "epoch": 935, "lr": 7.870933887961923e-05}
{"train_loss": 0.17139993607997894, "global_step": 83226, "epoch": 935, "lr": 7.870886423322164e-05}
{"train_loss": 0.1508815437555313, "global_step": 83227, "epoch": 935, "lr": 7.870838958296449e-05}
{"train_loss": 0.2325710505247116, "global_step": 83228, "epoch": 935, "lr": 7.870791492884786e-05}
{"train_loss": 0.19113683700561523, "global_step": 83229, "epoch": 935, "lr": 7.870744027087178e-05}
{"train_loss": 0.16908027231693268, "global_step": 83230, "epoch": 935, "lr": 7.870696560903635e-05}
{"train_loss": 0.221061110496521, "global_step": 83231, "epoch": 935, "lr": 7.870649094334162e-05}
{"train_loss": 0.20531922578811646, "global_step": 83232, "epoch": 935, "lr": 7.870601627378764e-05}
{"train_loss": 0.1894749402999878, "global_step": 83233, "epoch": 935, "lr": 7.870554160037449e-05}
{"train_loss": 0.159762442111969, "global_step": 83234, "epoch": 935, "lr": 7.870506692310222e-05}
{"train_loss": 0.23080086708068848, "global_step": 83235, "epoch": 935, "lr": 7.870459224197093e-05}
{"train_loss": 0.12890246510505676, "global_step": 83236, "epoch": 935, "lr": 7.870411755698064e-05}
{"train_loss": 0.18206295371055603, "global_step": 83237, "epoch": 935, "lr": 7.870364286813143e-05}
{"train_loss": 0.22665388882160187, "global_step": 83238, "epoch": 935, "lr": 7.870316817542338e-05}
{"train_loss": 0.14614911377429962, "global_step": 83239, "epoch": 935, "lr": 7.870269347885653e-05}
{"train_loss": 0.11872182786464691, "global_step": 83240, "epoch": 935, "lr": 7.870221877843094e-05}
{"train_loss": 0.13635939359664917, "global_step": 83241, "epoch": 935, "lr": 7.870174407414669e-05}
{"train_loss": 0.1828361302614212, "global_step": 83242, "epoch": 935, "lr": 7.870126936600384e-05}
{"train_loss": 0.18847528100013733, "global_step": 83243, "epoch": 935, "lr": 7.870079465400247e-05}
{"train_loss": 0.1998913288116455, "global_step": 83244, "epoch": 935, "lr": 7.870031993814262e-05}
{"train_loss": 0.16328667104244232, "global_step": 83245, "epoch": 935, "lr": 7.869984521842436e-05}
{"train_loss": 0.12127210944890976, "global_step": 83246, "epoch": 935, "lr": 7.869937049484777e-05}
{"train_loss": 0.22719725966453552, "global_step": 83247, "epoch": 935, "lr": 7.869889576741287e-05}
{"train_loss": 0.15246547758579254, "global_step": 83248, "epoch": 935, "lr": 7.869842103611977e-05}
{"train_loss": 0.16642111539840698, "global_step": 83249, "epoch": 935, "lr": 7.869794630096852e-05}
{"train_loss": 0.2716955244541168, "global_step": 83250, "epoch": 935, "lr": 7.869747156195918e-05}
{"train_loss": 0.13583777844905853, "global_step": 83251, "epoch": 935, "lr": 7.869699681909182e-05}
{"train_loss": 0.1793670803308487, "global_step": 83252, "epoch": 935, "lr": 7.869652207236649e-05}
{"train_loss": 0.18195877969264984, "global_step": 83253, "epoch": 935, "lr": 7.869604732178326e-05}
{"train_loss": 0.18186292052268982, "global_step": 83254, "epoch": 935, "lr": 7.869557256734219e-05}
{"train_loss": 0.1579006463289261, "global_step": 83255, "epoch": 935, "lr": 7.869509780904337e-05}
{"train_loss": 0.19943316280841827, "global_step": 83256, "epoch": 935, "lr": 7.869462304688684e-05}
{"train_loss": 0.12709403038024902, "global_step": 83257, "epoch": 935, "lr": 7.869414828087265e-05}
{"train_loss": 0.16974994540214539, "global_step": 83258, "epoch": 935, "lr": 7.86936735110009e-05}
{"train_loss": 0.20148785412311554, "global_step": 83259, "epoch": 935, "lr": 7.869319873727162e-05}
{"train_loss": 0.12674041092395782, "global_step": 83260, "epoch": 935, "lr": 7.869272395968488e-05}
{"train_loss": 0.21817328035831451, "global_step": 83261, "epoch": 935, "lr": 7.869224917824078e-05}
{"train_loss": 0.19577088952064514, "global_step": 83262, "epoch": 935, "lr": 7.869177439293935e-05}
{"train_loss": 0.17300038039684296, "global_step": 83263, "epoch": 935, "lr": 7.869129960378065e-05}
{"train_loss": 0.083330437541008, "global_step": 83264, "epoch": 935, "lr": 7.869082481076478e-05}
{"train_loss": 0.14720532298088074, "global_step": 83265, "epoch": 935, "lr": 7.869035001389175e-05}
{"train_loss": 0.15967555344104767, "global_step": 83266, "epoch": 935, "lr": 7.868987521316167e-05}
{"train_loss": 0.11921525001525879, "global_step": 83267, "epoch": 935, "lr": 7.868940040857457e-05}
{"train_loss": 0.18351943790912628, "global_step": 83268, "epoch": 935, "lr": 7.868892560013055e-05}
{"train_loss": 0.172324538230896, "global_step": 83269, "epoch": 935, "lr": 7.868845078782962e-05}
{"train_loss": 0.16083315014839172, "global_step": 83270, "epoch": 935, "lr": 7.868797597167192e-05}
{"train_loss": 0.24902479350566864, "global_step": 83271, "epoch": 935, "lr": 7.868750115165745e-05}
{"train_loss": 0.18839578330516815, "global_step": 83272, "epoch": 935, "lr": 7.868702632778628e-05}
{"train_loss": 0.19283220171928406, "global_step": 83273, "epoch": 935, "lr": 7.86865515000585e-05}
{"train_loss": 0.24021272361278534, "global_step": 83274, "epoch": 935, "lr": 7.868607666847418e-05}
{"train_loss": 0.16937130689620972, "global_step": 83275, "epoch": 935, "lr": 7.868560183303336e-05}
{"train_loss": 0.1673995554447174, "global_step": 83276, "epoch": 935, "lr": 7.86851269937361e-05}
{"train_loss": 0.14421063661575317, "global_step": 83277, "epoch": 935, "lr": 7.868465215058247e-05}
{"train_loss": 0.16089195013046265, "global_step": 83278, "epoch": 935, "lr": 7.868417730357253e-05}
{"train_loss": 0.15777277946472168, "global_step": 83279, "epoch": 935, "lr": 7.868370245270637e-05}
{"train_loss": 0.18863272666931152, "global_step": 83280, "epoch": 935, "lr": 7.868322759798406e-05}
{"train_loss": 0.15811316668987274, "global_step": 83281, "epoch": 935, "lr": 7.86827527394056e-05}
{"train_loss": 0.17025409638881683, "global_step": 83282, "epoch": 935, "lr": 7.868227787697111e-05}
{"train_loss": 0.09050606191158295, "global_step": 83283, "epoch": 935, "lr": 7.868180301068064e-05}
{"train_loss": 0.14139120280742645, "global_step": 83284, "epoch": 935, "lr": 7.868132814053424e-05}
{"train_loss": 0.21131908893585205, "global_step": 83285, "epoch": 935, "lr": 7.8680853266532e-05}
{"train_loss": 0.3291965126991272, "global_step": 83286, "epoch": 935, "lr": 7.868037838867395e-05}
{"train_loss": 0.18256805837154388, "global_step": 83287, "epoch": 935, "lr": 7.867990350696019e-05}
{"train_loss": 0.179860919713974, "global_step": 83288, "epoch": 935, "lr": 7.867942862139075e-05}
{"train_loss": 0.16550500690937042, "global_step": 83289, "epoch": 935, "lr": 7.867895373196573e-05}
{"train_loss": 0.13758651912212372, "global_step": 83290, "epoch": 935, "lr": 7.867847883868516e-05}
{"train_loss": 0.19545811414718628, "global_step": 83291, "epoch": 935, "lr": 7.867800394154912e-05}
{"train_loss": 0.13887128233909607, "global_step": 83292, "epoch": 935, "lr": 7.867752904055768e-05}
{"train_loss": 0.20426496863365173, "global_step": 83293, "epoch": 935, "lr": 7.86770541357109e-05}
{"train_loss": 0.2343464493751526, "global_step": 83294, "epoch": 935, "lr": 7.867657922700882e-05}
{"train_loss": 0.2279963493347168, "global_step": 83295, "epoch": 935, "lr": 7.867610431445154e-05}
{"train_loss": 0.20690752565860748, "global_step": 83296, "epoch": 935, "lr": 7.867562939803907e-05}
{"train_loss": 0.21818526089191437, "global_step": 83297, "epoch": 935, "lr": 7.867515447777155e-05}
{"train_loss": 0.20243661105632782, "global_step": 83298, "epoch": 935, "lr": 7.8674679553649e-05}
{"train_loss": 0.20401640236377716, "global_step": 83299, "epoch": 935, "lr": 7.867420462567148e-05}
{"train_loss": 0.15394113957881927, "global_step": 83300, "epoch": 935, "lr": 7.867372969383907e-05}
{"train_loss": 0.19897834956645966, "global_step": 83301, "epoch": 935, "lr": 7.867325475815184e-05}
{"train_loss": 0.21713055670261383, "global_step": 83302, "epoch": 935, "lr": 7.867277981860981e-05}
{"train_loss": 0.18099497434463394, "global_step": 83303, "epoch": 935, "lr": 7.867230487521308e-05, "val_loss": 4.261709690093994, "train_action_mse_error": 40.74237823486328}
{"train_loss": 0.12804444134235382, "global_step": 83304, "epoch": 936, "lr": 7.867182992796173e-05}
{"train_loss": 0.180578351020813, "global_step": 83305, "epoch": 936, "lr": 7.867135497685578e-05}
{"train_loss": 0.1384124606847763, "global_step": 83306, "epoch": 936, "lr": 7.867088002189534e-05}
{"train_loss": 0.17350414395332336, "global_step": 83307, "epoch": 936, "lr": 7.867040506308043e-05}
{"train_loss": 0.23085789382457733, "global_step": 83308, "epoch": 936, "lr": 7.866993010041112e-05}
{"train_loss": 0.2527189254760742, "global_step": 83309, "epoch": 936, "lr": 7.866945513388752e-05}
{"train_loss": 0.1375715285539627, "global_step": 83310, "epoch": 936, "lr": 7.866898016350964e-05}
{"train_loss": 0.1666586995124817, "global_step": 83311, "epoch": 936, "lr": 7.866850518927758e-05}
{"train_loss": 0.18932987749576569, "global_step": 83312, "epoch": 936, "lr": 7.866803021119138e-05}
{"train_loss": 0.33956611156463623, "global_step": 83313, "epoch": 936, "lr": 7.866755522925111e-05}
{"train_loss": 0.09850818663835526, "global_step": 83314, "epoch": 936, "lr": 7.866708024345684e-05}
{"train_loss": 0.13431808352470398, "global_step": 83315, "epoch": 936, "lr": 7.866660525380863e-05}
{"train_loss": 0.13663356006145477, "global_step": 83316, "epoch": 936, "lr": 7.866613026030654e-05}
{"train_loss": 0.21208536624908447, "global_step": 83317, "epoch": 936, "lr": 7.866565526295065e-05}
{"train_loss": 0.18189549446105957, "global_step": 83318, "epoch": 936, "lr": 7.866518026174099e-05}
{"train_loss": 0.2631353437900543, "global_step": 83319, "epoch": 936, "lr": 7.866470525667766e-05}
{"train_loss": 0.24492987990379333, "global_step": 83320, "epoch": 936, "lr": 7.866423024776072e-05}
{"train_loss": 0.16669344902038574, "global_step": 83321, "epoch": 936, "lr": 7.866375523499022e-05}
{"train_loss": 0.12406321614980698, "global_step": 83322, "epoch": 936, "lr": 7.866328021836621e-05}
{"train_loss": 0.16018304228782654, "global_step": 83323, "epoch": 936, "lr": 7.866280519788878e-05}
{"train_loss": 0.20903654396533966, "global_step": 83324, "epoch": 936, "lr": 7.8662330173558e-05}
{"train_loss": 0.19055116176605225, "global_step": 83325, "epoch": 936, "lr": 7.86618551453739e-05}
{"train_loss": 0.12159550189971924, "global_step": 83326, "epoch": 936, "lr": 7.866138011333657e-05}
{"train_loss": 0.20057906210422516, "global_step": 83327, "epoch": 936, "lr": 7.866090507744605e-05}
{"train_loss": 0.20897777378559113, "global_step": 83328, "epoch": 936, "lr": 7.866043003770245e-05}
{"train_loss": 0.1863900125026703, "global_step": 83329, "epoch": 936, "lr": 7.865995499410579e-05}
{"train_loss": 0.1764204204082489, "global_step": 83330, "epoch": 936, "lr": 7.865947994665614e-05}
{"train_loss": 0.21370646357536316, "global_step": 83331, "epoch": 936, "lr": 7.865900489535359e-05}
{"train_loss": 0.22353868186473846, "global_step": 83332, "epoch": 936, "lr": 7.865852984019817e-05}
{"train_loss": 0.337319016456604, "global_step": 83333, "epoch": 936, "lr": 7.865805478118997e-05}
{"train_loss": 0.20614805817604065, "global_step": 83334, "epoch": 936, "lr": 7.865757971832905e-05}
{"train_loss": 0.1805073469877243, "global_step": 83335, "epoch": 936, "lr": 7.865710465161546e-05}
{"train_loss": 0.1947401762008667, "global_step": 83336, "epoch": 936, "lr": 7.865662958104927e-05}
{"train_loss": 0.2073998898267746, "global_step": 83337, "epoch": 936, "lr": 7.865615450663054e-05}
{"train_loss": 0.18817085027694702, "global_step": 83338, "epoch": 936, "lr": 7.865567942835934e-05}
{"train_loss": 0.2559818625450134, "global_step": 83339, "epoch": 936, "lr": 7.865520434623575e-05}
{"train_loss": 0.1808367669582367, "global_step": 83340, "epoch": 936, "lr": 7.86547292602598e-05}
{"train_loss": 0.18974623084068298, "global_step": 83341, "epoch": 936, "lr": 7.865425417043158e-05}
{"train_loss": 0.21025043725967407, "global_step": 83342, "epoch": 936, "lr": 7.865377907675116e-05}
{"train_loss": 0.14292371273040771, "global_step": 83343, "epoch": 936, "lr": 7.865330397921855e-05}
{"train_loss": 0.1635531336069107, "global_step": 83344, "epoch": 936, "lr": 7.865282887783387e-05}
{"train_loss": 0.13547970354557037, "global_step": 83345, "epoch": 936, "lr": 7.865235377259719e-05}
{"train_loss": 0.17512847483158112, "global_step": 83346, "epoch": 936, "lr": 7.865187866350852e-05}
{"train_loss": 0.16726243495941162, "global_step": 83347, "epoch": 936, "lr": 7.865140355056797e-05}
{"train_loss": 0.1968214064836502, "global_step": 83348, "epoch": 936, "lr": 7.865092843377557e-05}
{"train_loss": 0.2860623002052307, "global_step": 83349, "epoch": 936, "lr": 7.865045331313142e-05}
{"train_loss": 0.2206859141588211, "global_step": 83350, "epoch": 936, "lr": 7.864997818863556e-05}
{"train_loss": 0.13136504590511322, "global_step": 83351, "epoch": 936, "lr": 7.864950306028805e-05}
{"train_loss": 0.20876461267471313, "global_step": 83352, "epoch": 936, "lr": 7.864902792808897e-05}
{"train_loss": 0.17075154185295105, "global_step": 83353, "epoch": 936, "lr": 7.864855279203838e-05}
{"train_loss": 0.1814355105161667, "global_step": 83354, "epoch": 936, "lr": 7.864807765213635e-05}
{"train_loss": 0.17171110212802887, "global_step": 83355, "epoch": 936, "lr": 7.864760250838293e-05}
{"train_loss": 0.18510814011096954, "global_step": 83356, "epoch": 936, "lr": 7.864712736077817e-05}
{"train_loss": 0.196967214345932, "global_step": 83357, "epoch": 936, "lr": 7.864665220932217e-05}
{"train_loss": 0.3047664761543274, "global_step": 83358, "epoch": 936, "lr": 7.864617705401498e-05}
{"train_loss": 0.2510089874267578, "global_step": 83359, "epoch": 936, "lr": 7.864570189485667e-05}
{"train_loss": 0.13761863112449646, "global_step": 83360, "epoch": 936, "lr": 7.864522673184726e-05}
{"train_loss": 0.23064476251602173, "global_step": 83361, "epoch": 936, "lr": 7.864475156498687e-05}
{"train_loss": 0.1091194748878479, "global_step": 83362, "epoch": 936, "lr": 7.864427639427554e-05}
{"train_loss": 0.15337656438350677, "global_step": 83363, "epoch": 936, "lr": 7.864380121971334e-05}
{"train_loss": 0.25911587476730347, "global_step": 83364, "epoch": 936, "lr": 7.864332604130033e-05}
{"train_loss": 0.30079615116119385, "global_step": 83365, "epoch": 936, "lr": 7.864285085903657e-05}
{"train_loss": 0.27671298384666443, "global_step": 83366, "epoch": 936, "lr": 7.864237567292213e-05}
{"train_loss": 0.15049825608730316, "global_step": 83367, "epoch": 936, "lr": 7.864190048295708e-05}
{"train_loss": 0.16841939091682434, "global_step": 83368, "epoch": 936, "lr": 7.864142528914145e-05}
{"train_loss": 0.31122612953186035, "global_step": 83369, "epoch": 936, "lr": 7.864095009147533e-05}
{"train_loss": 0.24340735375881195, "global_step": 83370, "epoch": 936, "lr": 7.864047488995881e-05}
{"train_loss": 0.2624781131744385, "global_step": 83371, "epoch": 936, "lr": 7.863999968459192e-05}
{"train_loss": 0.20765924453735352, "global_step": 83372, "epoch": 936, "lr": 7.863952447537473e-05}
{"train_loss": 0.2413330227136612, "global_step": 83373, "epoch": 936, "lr": 7.863904926230729e-05}
{"train_loss": 0.26595398783683777, "global_step": 83374, "epoch": 936, "lr": 7.86385740453897e-05}
{"train_loss": 0.1927478015422821, "global_step": 83375, "epoch": 936, "lr": 7.8638098824622e-05}
{"train_loss": 0.11256545037031174, "global_step": 83376, "epoch": 936, "lr": 7.863762360000424e-05}
{"train_loss": 0.26731759309768677, "global_step": 83377, "epoch": 936, "lr": 7.863714837153652e-05}
{"train_loss": 0.1394239217042923, "global_step": 83378, "epoch": 936, "lr": 7.863667313921887e-05}
{"train_loss": 0.17244794964790344, "global_step": 83379, "epoch": 936, "lr": 7.863619790305138e-05}
{"train_loss": 0.25933852791786194, "global_step": 83380, "epoch": 936, "lr": 7.863572266303409e-05}
{"train_loss": 0.21096710860729218, "global_step": 83381, "epoch": 936, "lr": 7.863524741916709e-05}
{"train_loss": 0.29156094789505005, "global_step": 83382, "epoch": 936, "lr": 7.863477217145042e-05}
{"train_loss": 0.2136039286851883, "global_step": 83383, "epoch": 936, "lr": 7.863429691988415e-05}
{"train_loss": 0.23850175738334656, "global_step": 83384, "epoch": 936, "lr": 7.863382166446835e-05}
{"train_loss": 0.16094090044498444, "global_step": 83385, "epoch": 936, "lr": 7.86333464052031e-05}
{"train_loss": 0.2066301703453064, "global_step": 83386, "epoch": 936, "lr": 7.863287114208843e-05}
{"train_loss": 0.1777205467224121, "global_step": 83387, "epoch": 936, "lr": 7.86323958751244e-05}
{"train_loss": 0.2393028736114502, "global_step": 83388, "epoch": 936, "lr": 7.863192060431113e-05}
{"train_loss": 0.20392654836177826, "global_step": 83389, "epoch": 936, "lr": 7.863144532964863e-05}
{"train_loss": 0.1715114712715149, "global_step": 83390, "epoch": 936, "lr": 7.863097005113698e-05}
{"train_loss": 0.16068732738494873, "global_step": 83391, "epoch": 936, "lr": 7.863049476877624e-05}
{"train_loss": 0.20002392126937932, "global_step": 83392, "epoch": 936, "lr": 7.863001948256649e-05, "val_loss": 4.002331733703613}
{"train_loss": 0.1978302150964737, "global_step": 83393, "epoch": 937, "lr": 7.862954419250778e-05}
{"train_loss": 0.18748177587985992, "global_step": 83394, "epoch": 937, "lr": 7.862906889860017e-05}
{"train_loss": 0.1773400604724884, "global_step": 83395, "epoch": 937, "lr": 7.862859360084374e-05}
{"train_loss": 0.18864881992340088, "global_step": 83396, "epoch": 937, "lr": 7.862811829923855e-05}
{"train_loss": 0.21163111925125122, "global_step": 83397, "epoch": 937, "lr": 7.862764299378465e-05}
{"train_loss": 0.19165067374706268, "global_step": 83398, "epoch": 937, "lr": 7.862716768448211e-05}
{"train_loss": 0.2060510665178299, "global_step": 83399, "epoch": 937, "lr": 7.8626692371331e-05}
{"train_loss": 0.20780983567237854, "global_step": 83400, "epoch": 937, "lr": 7.862621705433138e-05}
{"train_loss": 0.20935702323913574, "global_step": 83401, "epoch": 937, "lr": 7.862574173348331e-05}
{"train_loss": 0.1996317356824875, "global_step": 83402, "epoch": 937, "lr": 7.862526640878686e-05}
{"train_loss": 0.18670019507408142, "global_step": 83403, "epoch": 937, "lr": 7.862479108024208e-05}
{"train_loss": 0.18010829389095306, "global_step": 83404, "epoch": 937, "lr": 7.862431574784906e-05}
{"train_loss": 0.19890086352825165, "global_step": 83405, "epoch": 937, "lr": 7.862384041160784e-05}
{"train_loss": 0.20011049509048462, "global_step": 83406, "epoch": 937, "lr": 7.86233650715185e-05}
{"train_loss": 0.19568878412246704, "global_step": 83407, "epoch": 937, "lr": 7.86228897275811e-05}
{"train_loss": 0.22293634712696075, "global_step": 83408, "epoch": 937, "lr": 7.86224143797957e-05}
{"train_loss": 0.2558688223361969, "global_step": 83409, "epoch": 937, "lr": 7.862193902816238e-05}
{"train_loss": 0.18643806874752045, "global_step": 83410, "epoch": 937, "lr": 7.862146367268116e-05}
{"train_loss": 0.16777245700359344, "global_step": 83411, "epoch": 937, "lr": 7.862098831335215e-05}
{"train_loss": 0.11220607906579971, "global_step": 83412, "epoch": 937, "lr": 7.86205129501754e-05}
{"train_loss": 0.2090020775794983, "global_step": 83413, "epoch": 937, "lr": 7.862003758315095e-05}
{"train_loss": 0.14619342982769012, "global_step": 83414, "epoch": 937, "lr": 7.86195622122789e-05}
{"train_loss": 0.19792936742305756, "global_step": 83415, "epoch": 937, "lr": 7.86190868375593e-05}
{"train_loss": 0.20228905975818634, "global_step": 83416, "epoch": 937, "lr": 7.861861145899222e-05}
{"train_loss": 0.22230462729930878, "global_step": 83417, "epoch": 937, "lr": 7.861813607657771e-05}
{"train_loss": 0.1555650681257248, "global_step": 83418, "epoch": 937, "lr": 7.861766069031585e-05}
{"train_loss": 0.1944543570280075, "global_step": 83419, "epoch": 937, "lr": 7.861718530020668e-05}
{"train_loss": 0.27080342173576355, "global_step": 83420, "epoch": 937, "lr": 7.861670990625027e-05}
{"train_loss": 0.14634133875370026, "global_step": 83421, "epoch": 937, "lr": 7.861623450844671e-05}
{"train_loss": 0.22448033094406128, "global_step": 83422, "epoch": 937, "lr": 7.861575910679602e-05}
{"train_loss": 0.17646424472332, "global_step": 83423, "epoch": 937, "lr": 7.861528370129831e-05}
{"train_loss": 0.20469839870929718, "global_step": 83424, "epoch": 937, "lr": 7.861480829195362e-05}
{"train_loss": 0.14724291861057281, "global_step": 83425, "epoch": 937, "lr": 7.861433287876203e-05}
{"train_loss": 0.17829708755016327, "global_step": 83426, "epoch": 937, "lr": 7.861385746172359e-05}
{"train_loss": 0.18714402616024017, "global_step": 83427, "epoch": 937, "lr": 7.861338204083835e-05}
{"train_loss": 0.19465143978595734, "global_step": 83428, "epoch": 937, "lr": 7.861290661610638e-05}
{"train_loss": 0.10998287051916122, "global_step": 83429, "epoch": 937, "lr": 7.861243118752778e-05}
{"train_loss": 0.1172945573925972, "global_step": 83430, "epoch": 937, "lr": 7.861195575510258e-05}
{"train_loss": 0.15828339755535126, "global_step": 83431, "epoch": 937, "lr": 7.861148031883085e-05}
{"train_loss": 0.1683819442987442, "global_step": 83432, "epoch": 937, "lr": 7.861100487871265e-05}
{"train_loss": 0.20366404950618744, "global_step": 83433, "epoch": 937, "lr": 7.861052943474804e-05}
{"train_loss": 0.2485223114490509, "global_step": 83434, "epoch": 937, "lr": 7.861005398693712e-05}
{"train_loss": 0.13243520259857178, "global_step": 83435, "epoch": 937, "lr": 7.860957853527991e-05}
{"train_loss": 0.19199615716934204, "global_step": 83436, "epoch": 937, "lr": 7.860910307977649e-05}
{"train_loss": 0.17695564031600952, "global_step": 83437, "epoch": 937, "lr": 7.860862762042693e-05}
{"train_loss": 0.13520634174346924, "global_step": 83438, "epoch": 937, "lr": 7.860815215723129e-05}
{"train_loss": 0.2200549989938736, "global_step": 83439, "epoch": 937, "lr": 7.860767669018964e-05}
{"train_loss": 0.13675446808338165, "global_step": 83440, "epoch": 937, "lr": 7.860720121930203e-05}
{"train_loss": 0.1873188316822052, "global_step": 83441, "epoch": 937, "lr": 7.860672574456851e-05}
{"train_loss": 0.2351662665605545, "global_step": 83442, "epoch": 937, "lr": 7.860625026598918e-05}
{"train_loss": 0.1596311330795288, "global_step": 83443, "epoch": 937, "lr": 7.86057747835641e-05}
{"train_loss": 0.15261369943618774, "global_step": 83444, "epoch": 937, "lr": 7.860529929729331e-05}
{"train_loss": 0.0975288599729538, "global_step": 83445, "epoch": 937, "lr": 7.860482380717689e-05}
{"train_loss": 0.18279550969600677, "global_step": 83446, "epoch": 937, "lr": 7.860434831321491e-05}
{"train_loss": 0.1855090856552124, "global_step": 83447, "epoch": 937, "lr": 7.86038728154074e-05}
{"train_loss": 0.18194280564785004, "global_step": 83448, "epoch": 937, "lr": 7.860339731375445e-05}
{"train_loss": 0.2978612184524536, "global_step": 83449, "epoch": 937, "lr": 7.860292180825614e-05}
{"train_loss": 0.22112712264060974, "global_step": 83450, "epoch": 937, "lr": 7.860244629891252e-05}
{"train_loss": 0.19912010431289673, "global_step": 83451, "epoch": 937, "lr": 7.860197078572362e-05}
{"train_loss": 0.1532481610774994, "global_step": 83452, "epoch": 937, "lr": 7.860149526868956e-05}
{"train_loss": 0.23452895879745483, "global_step": 83453, "epoch": 937, "lr": 7.860101974781036e-05}
{"train_loss": 0.2198668122291565, "global_step": 83454, "epoch": 937, "lr": 7.860054422308611e-05}
{"train_loss": 0.1464916318655014, "global_step": 83455, "epoch": 937, "lr": 7.860006869451688e-05}
{"train_loss": 0.23680086433887482, "global_step": 83456, "epoch": 937, "lr": 7.85995931621027e-05}
{"train_loss": 0.1154395043849945, "global_step": 83457, "epoch": 937, "lr": 7.859911762584366e-05}
{"train_loss": 0.15844786167144775, "global_step": 83458, "epoch": 937, "lr": 7.859864208573981e-05}
{"train_loss": 0.16644032299518585, "global_step": 83459, "epoch": 937, "lr": 7.859816654179123e-05}
{"train_loss": 0.2511713206768036, "global_step": 83460, "epoch": 937, "lr": 7.859769099399796e-05}
{"train_loss": 0.1839388906955719, "global_step": 83461, "epoch": 937, "lr": 7.85972154423601e-05}
{"train_loss": 0.13312162458896637, "global_step": 83462, "epoch": 937, "lr": 7.859673988687768e-05}
{"train_loss": 0.20729127526283264, "global_step": 83463, "epoch": 937, "lr": 7.859626432755078e-05}
{"train_loss": 0.1847945898771286, "global_step": 83464, "epoch": 937, "lr": 7.859578876437947e-05}
{"train_loss": 0.17756803333759308, "global_step": 83465, "epoch": 937, "lr": 7.85953131973638e-05}
{"train_loss": 0.1842806339263916, "global_step": 83466, "epoch": 937, "lr": 7.859483762650384e-05}
{"train_loss": 0.1618223488330841, "global_step": 83467, "epoch": 937, "lr": 7.859436205179965e-05}
{"train_loss": 0.13871225714683533, "global_step": 83468, "epoch": 937, "lr": 7.859388647325128e-05}
{"train_loss": 0.17739227414131165, "global_step": 83469, "epoch": 937, "lr": 7.859341089085884e-05}
{"train_loss": 0.35671138763427734, "global_step": 83470, "epoch": 937, "lr": 7.859293530462233e-05}
{"train_loss": 0.3050520718097687, "global_step": 83471, "epoch": 937, "lr": 7.859245971454187e-05}
{"train_loss": 0.23110590875148773, "global_step": 83472, "epoch": 937, "lr": 7.85919841206175e-05}
{"train_loss": 0.15016888082027435, "global_step": 83473, "epoch": 937, "lr": 7.859150852284928e-05}
{"train_loss": 0.18995177745819092, "global_step": 83474, "epoch": 937, "lr": 7.859103292123729e-05}
{"train_loss": 0.24848660826683044, "global_step": 83475, "epoch": 937, "lr": 7.859055731578158e-05}
{"train_loss": 0.20265570282936096, "global_step": 83476, "epoch": 937, "lr": 7.859008170648223e-05}
{"train_loss": 0.2279943972826004, "global_step": 83477, "epoch": 937, "lr": 7.858960609333927e-05}
{"train_loss": 0.21000520884990692, "global_step": 83478, "epoch": 937, "lr": 7.858913047635279e-05}
{"train_loss": 0.14571651816368103, "global_step": 83479, "epoch": 937, "lr": 7.858865485552286e-05}
{"train_loss": 0.2184121459722519, "global_step": 83480, "epoch": 937, "lr": 7.858817923084954e-05}
{"train_loss": 0.18982594175620027, "global_step": 83481, "epoch": 937, "lr": 7.858770360233288e-05, "val_loss": 4.20601224899292}
{"train_loss": 0.17425550520420074, "global_step": 83482, "epoch": 938, "lr": 7.858722796997293e-05}
{"train_loss": 0.1954185515642166, "global_step": 83483, "epoch": 938, "lr": 7.85867523337698e-05}
{"train_loss": 0.2108411192893982, "global_step": 83484, "epoch": 938, "lr": 7.858627669372353e-05}
{"train_loss": 0.18941417336463928, "global_step": 83485, "epoch": 938, "lr": 7.858580104983418e-05}
{"train_loss": 0.20296911895275116, "global_step": 83486, "epoch": 938, "lr": 7.858532540210182e-05}
{"train_loss": 0.2356962263584137, "global_step": 83487, "epoch": 938, "lr": 7.858484975052649e-05}
{"train_loss": 0.22558966279029846, "global_step": 83488, "epoch": 938, "lr": 7.85843740951083e-05}
{"train_loss": 0.15708082914352417, "global_step": 83489, "epoch": 938, "lr": 7.858389843584729e-05}
{"train_loss": 0.251770943403244, "global_step": 83490, "epoch": 938, "lr": 7.85834227727435e-05}
{"train_loss": 0.12020524591207504, "global_step": 83491, "epoch": 938, "lr": 7.858294710579703e-05}
{"train_loss": 0.18400296568870544, "global_step": 83492, "epoch": 938, "lr": 7.858247143500795e-05}
{"train_loss": 0.16711731255054474, "global_step": 83493, "epoch": 938, "lr": 7.858199576037628e-05}
{"train_loss": 0.1328345537185669, "global_step": 83494, "epoch": 938, "lr": 7.858152008190212e-05}
{"train_loss": 0.16618946194648743, "global_step": 83495, "epoch": 938, "lr": 7.858104439958552e-05}
{"train_loss": 0.21169719099998474, "global_step": 83496, "epoch": 938, "lr": 7.858056871342654e-05}
{"train_loss": 0.22753259539604187, "global_step": 83497, "epoch": 938, "lr": 7.858009302342526e-05}
{"train_loss": 0.15903626382350922, "global_step": 83498, "epoch": 938, "lr": 7.857961732958172e-05}
{"train_loss": 0.14754435420036316, "global_step": 83499, "epoch": 938, "lr": 7.857914163189601e-05}
{"train_loss": 0.23774345219135284, "global_step": 83500, "epoch": 938, "lr": 7.85786659303682e-05}
{"train_loss": 0.2126389443874359, "global_step": 83501, "epoch": 938, "lr": 7.857819022499832e-05}
{"train_loss": 0.17167073488235474, "global_step": 83502, "epoch": 938, "lr": 7.857771451578645e-05}
{"train_loss": 0.17753636837005615, "global_step": 83503, "epoch": 938, "lr": 7.857723880273265e-05}
{"train_loss": 0.11330997943878174, "global_step": 83504, "epoch": 938, "lr": 7.8576763085837e-05}
{"train_loss": 0.20047512650489807, "global_step": 83505, "epoch": 938, "lr": 7.857628736509954e-05}
{"train_loss": 0.2548958957195282, "global_step": 83506, "epoch": 938, "lr": 7.857581164052036e-05}
{"train_loss": 0.23067229986190796, "global_step": 83507, "epoch": 938, "lr": 7.857533591209951e-05}
{"train_loss": 0.147291362285614, "global_step": 83508, "epoch": 938, "lr": 7.857486017983704e-05}
{"train_loss": 0.17833511531352997, "global_step": 83509, "epoch": 938, "lr": 7.857438444373304e-05}
{"train_loss": 0.1882089227437973, "global_step": 83510, "epoch": 938, "lr": 7.857390870378756e-05}
{"train_loss": 0.18202358484268188, "global_step": 83511, "epoch": 938, "lr": 7.857343296000066e-05}
{"train_loss": 0.25147855281829834, "global_step": 83512, "epoch": 938, "lr": 7.857295721237243e-05}
{"train_loss": 0.1866012066602707, "global_step": 83513, "epoch": 938, "lr": 7.857248146090289e-05}
{"train_loss": 0.2049907147884369, "global_step": 83514, "epoch": 938, "lr": 7.857200570559213e-05}
{"train_loss": 0.23972399532794952, "global_step": 83515, "epoch": 938, "lr": 7.857152994644021e-05}
{"train_loss": 0.17185628414154053, "global_step": 83516, "epoch": 938, "lr": 7.857105418344722e-05}
{"train_loss": 0.16013960540294647, "global_step": 83517, "epoch": 938, "lr": 7.857057841661318e-05}
{"train_loss": 0.22000224888324738, "global_step": 83518, "epoch": 938, "lr": 7.857010264593818e-05}
{"train_loss": 0.1753225475549698, "global_step": 83519, "epoch": 938, "lr": 7.856962687142228e-05}
{"train_loss": 0.25600144267082214, "global_step": 83520, "epoch": 938, "lr": 7.856915109306554e-05}
{"train_loss": 0.20779643952846527, "global_step": 83521, "epoch": 938, "lr": 7.856867531086803e-05}
{"train_loss": 0.3173590302467346, "global_step": 83522, "epoch": 938, "lr": 7.856819952482982e-05}
{"train_loss": 0.1715981364250183, "global_step": 83523, "epoch": 938, "lr": 7.856772373495094e-05}
{"train_loss": 0.15341970324516296, "global_step": 83524, "epoch": 938, "lr": 7.856724794123148e-05}
{"train_loss": 0.2818146347999573, "global_step": 83525, "epoch": 938, "lr": 7.85667721436715e-05}
{"train_loss": 0.19654791057109833, "global_step": 83526, "epoch": 938, "lr": 7.856629634227108e-05}
{"train_loss": 0.16964781284332275, "global_step": 83527, "epoch": 938, "lr": 7.856582053703027e-05}
{"train_loss": 0.18629330396652222, "global_step": 83528, "epoch": 938, "lr": 7.856534472794913e-05}
{"train_loss": 0.19371233880519867, "global_step": 83529, "epoch": 938, "lr": 7.856486891502773e-05}
{"train_loss": 0.1169646829366684, "global_step": 83530, "epoch": 938, "lr": 7.856439309826611e-05}
{"train_loss": 0.1556898057460785, "global_step": 83531, "epoch": 938, "lr": 7.856391727766437e-05}
{"train_loss": 0.18410725891590118, "global_step": 83532, "epoch": 938, "lr": 7.856344145322257e-05}
{"train_loss": 0.15906259417533875, "global_step": 83533, "epoch": 938, "lr": 7.856296562494074e-05}
{"train_loss": 0.27233394980430603, "global_step": 83534, "epoch": 938, "lr": 7.8562489792819e-05}
{"train_loss": 0.17914031445980072, "global_step": 83535, "epoch": 938, "lr": 7.856201395685736e-05}
{"train_loss": 0.17321078479290009, "global_step": 83536, "epoch": 938, "lr": 7.856153811705592e-05}
{"train_loss": 0.18058481812477112, "global_step": 83537, "epoch": 938, "lr": 7.85610622734147e-05}
{"train_loss": 0.1779887080192566, "global_step": 83538, "epoch": 938, "lr": 7.856058642593381e-05}
{"train_loss": 0.15733441710472107, "global_step": 83539, "epoch": 938, "lr": 7.85601105746133e-05}
{"train_loss": 0.13650193810462952, "global_step": 83540, "epoch": 938, "lr": 7.855963471945323e-05}
{"train_loss": 0.15670020878314972, "global_step": 83541, "epoch": 938, "lr": 7.855915886045367e-05}
{"train_loss": 0.18580366671085358, "global_step": 83542, "epoch": 938, "lr": 7.855868299761467e-05}
{"train_loss": 0.151038259267807, "global_step": 83543, "epoch": 938, "lr": 7.85582071309363e-05}
{"train_loss": 0.2028331160545349, "global_step": 83544, "epoch": 938, "lr": 7.855773126041866e-05}
{"train_loss": 0.2548162341117859, "global_step": 83545, "epoch": 938, "lr": 7.855725538606175e-05}
{"train_loss": 0.29174765944480896, "global_step": 83546, "epoch": 938, "lr": 7.855677950786566e-05}
{"train_loss": 0.18431738018989563, "global_step": 83547, "epoch": 938, "lr": 7.855630362583048e-05}
{"train_loss": 0.1887170970439911, "global_step": 83548, "epoch": 938, "lr": 7.855582773995624e-05}
{"train_loss": 0.21389555931091309, "global_step": 83549, "epoch": 938, "lr": 7.855535185024302e-05}
{"train_loss": 0.18311186134815216, "global_step": 83550, "epoch": 938, "lr": 7.855487595669087e-05}
{"train_loss": 0.23691947758197784, "global_step": 83551, "epoch": 938, "lr": 7.855440005929988e-05}
{"train_loss": 0.16682080924510956, "global_step": 83552, "epoch": 938, "lr": 7.85539241580701e-05}
{"train_loss": 0.14059801399707794, "global_step": 83553, "epoch": 938, "lr": 7.855344825300158e-05}
{"train_loss": 0.17616751790046692, "global_step": 83554, "epoch": 938, "lr": 7.85529723440944e-05}
{"train_loss": 0.24787317216396332, "global_step": 83555, "epoch": 938, "lr": 7.855249643134863e-05}
{"train_loss": 0.19108115136623383, "global_step": 83556, "epoch": 938, "lr": 7.855202051476433e-05}
{"train_loss": 0.20987658202648163, "global_step": 83557, "epoch": 938, "lr": 7.855154459434155e-05}
{"train_loss": 0.1791113168001175, "global_step": 83558, "epoch": 938, "lr": 7.855106867008035e-05}
{"train_loss": 0.13371619582176208, "global_step": 83559, "epoch": 938, "lr": 7.855059274198082e-05}
{"train_loss": 0.20955875515937805, "global_step": 83560, "epoch": 938, "lr": 7.855011681004301e-05}
{"train_loss": 0.23353311419487, "global_step": 83561, "epoch": 938, "lr": 7.8549640874267e-05}
{"train_loss": 0.22337129712104797, "global_step": 83562, "epoch": 938, "lr": 7.85491649346528e-05}
{"train_loss": 0.13909143209457397, "global_step": 83563, "epoch": 938, "lr": 7.854868899120055e-05}
{"train_loss": 0.18257811665534973, "global_step": 83564, "epoch": 938, "lr": 7.854821304391027e-05}
{"train_loss": 0.25880613923072815, "global_step": 83565, "epoch": 938, "lr": 7.854773709278203e-05}
{"train_loss": 0.24534672498703003, "global_step": 83566, "epoch": 938, "lr": 7.854726113781587e-05}
{"train_loss": 0.27195408940315247, "global_step": 83567, "epoch": 938, "lr": 7.85467851790119e-05}
{"train_loss": 0.1751229614019394, "global_step": 83568, "epoch": 938, "lr": 7.854630921637016e-05}
{"train_loss": 0.15890398621559143, "global_step": 83569, "epoch": 938, "lr": 7.85458332498907e-05}
{"train_loss": 0.19417923111259267, "global_step": 83570, "epoch": 938, "lr": 7.854535727957363e-05, "val_loss": 4.225793361663818}
{"train_loss": 0.1653299480676651, "global_step": 83571, "epoch": 939, "lr": 7.854488130541897e-05}
{"train_loss": 0.18022596836090088, "global_step": 83572, "epoch": 939, "lr": 7.85444053274268e-05}
{"train_loss": 0.14019320905208588, "global_step": 83573, "epoch": 939, "lr": 7.854392934559718e-05}
{"train_loss": 0.17042158544063568, "global_step": 83574, "epoch": 939, "lr": 7.854345335993019e-05}
{"train_loss": 0.23519758880138397, "global_step": 83575, "epoch": 939, "lr": 7.854297737042587e-05}
{"train_loss": 0.19682490825653076, "global_step": 83576, "epoch": 939, "lr": 7.85425013770843e-05}
{"train_loss": 0.15080486238002777, "global_step": 83577, "epoch": 939, "lr": 7.854202537990551e-05}
{"train_loss": 0.1488945186138153, "global_step": 83578, "epoch": 939, "lr": 7.854154937888962e-05}
{"train_loss": 0.22333689033985138, "global_step": 83579, "epoch": 939, "lr": 7.854107337403667e-05}
{"train_loss": 0.20052146911621094, "global_step": 83580, "epoch": 939, "lr": 7.854059736534671e-05}
{"train_loss": 0.37721917033195496, "global_step": 83581, "epoch": 939, "lr": 7.854012135281983e-05}
{"train_loss": 0.255016952753067, "global_step": 83582, "epoch": 939, "lr": 7.853964533645605e-05}
{"train_loss": 0.15722927451133728, "global_step": 83583, "epoch": 939, "lr": 7.853916931625549e-05}
{"train_loss": 0.16228510439395905, "global_step": 83584, "epoch": 939, "lr": 7.853869329221816e-05}
{"train_loss": 0.18052369356155396, "global_step": 83585, "epoch": 939, "lr": 7.853821726434418e-05}
{"train_loss": 0.2107308954000473, "global_step": 83586, "epoch": 939, "lr": 7.853774123263355e-05}
{"train_loss": 0.20967768132686615, "global_step": 83587, "epoch": 939, "lr": 7.85372651970864e-05}
{"train_loss": 0.22593174874782562, "global_step": 83588, "epoch": 939, "lr": 7.853678915770275e-05}
{"train_loss": 0.16372713446617126, "global_step": 83589, "epoch": 939, "lr": 7.853631311448269e-05}
{"train_loss": 0.24216924607753754, "global_step": 83590, "epoch": 939, "lr": 7.853583706742624e-05}
{"train_loss": 0.1950284242630005, "global_step": 83591, "epoch": 939, "lr": 7.853536101653351e-05}
{"train_loss": 0.196786031126976, "global_step": 83592, "epoch": 939, "lr": 7.853488496180457e-05}
{"train_loss": 0.1646953970193863, "global_step": 83593, "epoch": 939, "lr": 7.853440890323944e-05}
{"train_loss": 0.21084679663181305, "global_step": 83594, "epoch": 939, "lr": 7.853393284083821e-05}
{"train_loss": 0.1690017431974411, "global_step": 83595, "epoch": 939, "lr": 7.853345677460093e-05}
{"train_loss": 0.16951051354408264, "global_step": 83596, "epoch": 939, "lr": 7.853298070452768e-05}
{"train_loss": 0.18582148849964142, "global_step": 83597, "epoch": 939, "lr": 7.853250463061853e-05}
{"train_loss": 0.20091386139392853, "global_step": 83598, "epoch": 939, "lr": 7.853202855287354e-05}
{"train_loss": 0.1878145933151245, "global_step": 83599, "epoch": 939, "lr": 7.853155247129275e-05}
{"train_loss": 0.18004190921783447, "global_step": 83600, "epoch": 939, "lr": 7.853107638587623e-05}
{"train_loss": 0.2143622189760208, "global_step": 83601, "epoch": 939, "lr": 7.853060029662406e-05}
{"train_loss": 0.2249143421649933, "global_step": 83602, "epoch": 939, "lr": 7.853012420353631e-05}
{"train_loss": 0.1734035313129425, "global_step": 83603, "epoch": 939, "lr": 7.852964810661303e-05}
{"train_loss": 0.1493050754070282, "global_step": 83604, "epoch": 939, "lr": 7.852917200585428e-05}
{"train_loss": 0.142099067568779, "global_step": 83605, "epoch": 939, "lr": 7.852869590126015e-05}
{"train_loss": 0.17477022111415863, "global_step": 83606, "epoch": 939, "lr": 7.852821979283067e-05}
{"train_loss": 0.19101375341415405, "global_step": 83607, "epoch": 939, "lr": 7.852774368056592e-05}
{"train_loss": 0.19057735800743103, "global_step": 83608, "epoch": 939, "lr": 7.852726756446596e-05}
{"train_loss": 0.2448032647371292, "global_step": 83609, "epoch": 939, "lr": 7.852679144453086e-05}
{"train_loss": 0.1869811713695526, "global_step": 83610, "epoch": 939, "lr": 7.852631532076067e-05}
{"train_loss": 0.21214021742343903, "global_step": 83611, "epoch": 939, "lr": 7.852583919315548e-05}
{"train_loss": 0.22827287018299103, "global_step": 83612, "epoch": 939, "lr": 7.852536306171533e-05}
{"train_loss": 0.1446298509836197, "global_step": 83613, "epoch": 939, "lr": 7.852488692644029e-05}
{"train_loss": 0.17956238985061646, "global_step": 83614, "epoch": 939, "lr": 7.852441078733044e-05}
{"train_loss": 0.08950179070234299, "global_step": 83615, "epoch": 939, "lr": 7.85239346443858e-05}
{"train_loss": 0.1652078926563263, "global_step": 83616, "epoch": 939, "lr": 7.852345849760649e-05}
{"train_loss": 0.23306314647197723, "global_step": 83617, "epoch": 939, "lr": 7.852298234699256e-05}
{"train_loss": 0.15147548913955688, "global_step": 83618, "epoch": 939, "lr": 7.852250619254405e-05}
{"train_loss": 0.16822636127471924, "global_step": 83619, "epoch": 939, "lr": 7.852203003426103e-05}
{"train_loss": 0.17654502391815186, "global_step": 83620, "epoch": 939, "lr": 7.852155387214357e-05}
{"train_loss": 0.15728184580802917, "global_step": 83621, "epoch": 939, "lr": 7.852107770619174e-05}
{"train_loss": 0.18795077502727509, "global_step": 83622, "epoch": 939, "lr": 7.85206015364056e-05}
{"train_loss": 0.18498273193836212, "global_step": 83623, "epoch": 939, "lr": 7.852012536278522e-05}
{"train_loss": 0.17512591183185577, "global_step": 83624, "epoch": 939, "lr": 7.851964918533064e-05}
{"train_loss": 0.1438736766576767, "global_step": 83625, "epoch": 939, "lr": 7.851917300404194e-05}
{"train_loss": 0.2079007625579834, "global_step": 83626, "epoch": 939, "lr": 7.85186968189192e-05}
{"train_loss": 0.09561750292778015, "global_step": 83627, "epoch": 939, "lr": 7.851822062996246e-05}
{"train_loss": 0.14066879451274872, "global_step": 83628, "epoch": 939, "lr": 7.851774443717179e-05}
{"train_loss": 0.1623581200838089, "global_step": 83629, "epoch": 939, "lr": 7.851726824054727e-05}
{"train_loss": 0.23873183131217957, "global_step": 83630, "epoch": 939, "lr": 7.851679204008894e-05}
{"train_loss": 0.24038764834403992, "global_step": 83631, "epoch": 939, "lr": 7.851631583579688e-05}
{"train_loss": 0.17918165028095245, "global_step": 83632, "epoch": 939, "lr": 7.851583962767114e-05}
{"train_loss": 0.24959217011928558, "global_step": 83633, "epoch": 939, "lr": 7.85153634157118e-05}
{"train_loss": 0.16293780505657196, "global_step": 83634, "epoch": 939, "lr": 7.851488719991893e-05}
{"train_loss": 0.1393754482269287, "global_step": 83635, "epoch": 939, "lr": 7.851441098029255e-05}
{"train_loss": 0.15786999464035034, "global_step": 83636, "epoch": 939, "lr": 7.851393475683277e-05}
{"train_loss": 0.18694789707660675, "global_step": 83637, "epoch": 939, "lr": 7.851345852953967e-05}
{"train_loss": 0.12807798385620117, "global_step": 83638, "epoch": 939, "lr": 7.851298229841323e-05}
{"train_loss": 0.15532177686691284, "global_step": 83639, "epoch": 939, "lr": 7.85125060634536e-05}
{"train_loss": 0.21056470274925232, "global_step": 83640, "epoch": 939, "lr": 7.85120298246608e-05}
{"train_loss": 0.1893729567527771, "global_step": 83641, "epoch": 939, "lr": 7.85115535820349e-05}
{"train_loss": 0.19775211811065674, "global_step": 83642, "epoch": 939, "lr": 7.8511077335576e-05}
{"train_loss": 0.12081213295459747, "global_step": 83643, "epoch": 939, "lr": 7.851060108528409e-05}
{"train_loss": 0.20045270025730133, "global_step": 83644, "epoch": 939, "lr": 7.85101248311593e-05}
{"train_loss": 0.13611909747123718, "global_step": 83645, "epoch": 939, "lr": 7.850964857320167e-05}
{"train_loss": 0.15832440555095673, "global_step": 83646, "epoch": 939, "lr": 7.850917231141126e-05}
{"train_loss": 0.172693133354187, "global_step": 83647, "epoch": 939, "lr": 7.850869604578815e-05}
{"train_loss": 0.2775428593158722, "global_step": 83648, "epoch": 939, "lr": 7.850821977633237e-05}
{"train_loss": 0.20542074739933014, "global_step": 83649, "epoch": 939, "lr": 7.850774350304404e-05}
{"train_loss": 0.23486272990703583, "global_step": 83650, "epoch": 939, "lr": 7.850726722592316e-05}
{"train_loss": 0.23622719943523407, "global_step": 83651, "epoch": 939, "lr": 7.850679094496985e-05}
{"train_loss": 0.14434491097927094, "global_step": 83652, "epoch": 939, "lr": 7.850631466018415e-05}
{"train_loss": 0.23858287930488586, "global_step": 83653, "epoch": 939, "lr": 7.850583837156611e-05}
{"train_loss": 0.12444788217544556, "global_step": 83654, "epoch": 939, "lr": 7.85053620791158e-05}
{"train_loss": 0.14963294565677643, "global_step": 83655, "epoch": 939, "lr": 7.85048857828333e-05}
{"train_loss": 0.1346597820520401, "global_step": 83656, "epoch": 939, "lr": 7.850440948271868e-05}
{"train_loss": 0.15972086787223816, "global_step": 83657, "epoch": 939, "lr": 7.850393317877198e-05}
{"train_loss": 0.21360640227794647, "global_step": 83658, "epoch": 939, "lr": 7.850345687099326e-05}
{"train_loss": 0.18420515661494116, "global_step": 83659, "epoch": 939, "lr": 7.850298055938261e-05, "val_loss": 4.201672077178955}
{"train_loss": 0.17921677231788635, "global_step": 83660, "epoch": 940, "lr": 7.850250424394007e-05}
{"train_loss": 0.14270395040512085, "global_step": 83661, "epoch": 940, "lr": 7.850202792466573e-05}
{"train_loss": 0.13075751066207886, "global_step": 83662, "epoch": 940, "lr": 7.850155160155963e-05}
{"train_loss": 0.14543516933918, "global_step": 83663, "epoch": 940, "lr": 7.850107527462186e-05}
{"train_loss": 0.17393361032009125, "global_step": 83664, "epoch": 940, "lr": 7.850059894385245e-05}
{"train_loss": 0.19045986235141754, "global_step": 83665, "epoch": 940, "lr": 7.85001226092515e-05}
{"train_loss": 0.15466390550136566, "global_step": 83666, "epoch": 940, "lr": 7.849964627081904e-05}
{"train_loss": 0.2253999263048172, "global_step": 83667, "epoch": 940, "lr": 7.849916992855515e-05}
{"train_loss": 0.17441770434379578, "global_step": 83668, "epoch": 940, "lr": 7.84986935824599e-05}
{"train_loss": 0.24583861231803894, "global_step": 83669, "epoch": 940, "lr": 7.849821723253335e-05}
{"train_loss": 0.24418359994888306, "global_step": 83670, "epoch": 940, "lr": 7.849774087877556e-05}
{"train_loss": 0.22185169160366058, "global_step": 83671, "epoch": 940, "lr": 7.84972645211866e-05}
{"train_loss": 0.15673872828483582, "global_step": 83672, "epoch": 940, "lr": 7.849678815976651e-05}
{"train_loss": 0.16100749373435974, "global_step": 83673, "epoch": 940, "lr": 7.849631179451539e-05}
{"train_loss": 0.17574414610862732, "global_step": 83674, "epoch": 940, "lr": 7.84958354254333e-05}
{"train_loss": 0.1719333529472351, "global_step": 83675, "epoch": 940, "lr": 7.849535905252027e-05}
{"train_loss": 0.1424829661846161, "global_step": 83676, "epoch": 940, "lr": 7.849488267577641e-05}
{"train_loss": 0.10042170435190201, "global_step": 83677, "epoch": 940, "lr": 7.849440629520173e-05}
{"train_loss": 0.2034100443124771, "global_step": 83678, "epoch": 940, "lr": 7.849392991079634e-05}
{"train_loss": 0.19080223143100739, "global_step": 83679, "epoch": 940, "lr": 7.84934535225603e-05}
{"train_loss": 0.1756395846605301, "global_step": 83680, "epoch": 940, "lr": 7.849297713049364e-05}
{"train_loss": 0.15898053348064423, "global_step": 83681, "epoch": 940, "lr": 7.849250073459647e-05}
{"train_loss": 0.12139032781124115, "global_step": 83682, "epoch": 940, "lr": 7.84920243348688e-05}
{"train_loss": 0.15956129133701324, "global_step": 83683, "epoch": 940, "lr": 7.849154793131075e-05}
{"train_loss": 0.2496722936630249, "global_step": 83684, "epoch": 940, "lr": 7.849107152392236e-05}
{"train_loss": 0.146192267537117, "global_step": 83685, "epoch": 940, "lr": 7.849059511270368e-05}
{"train_loss": 0.1781369298696518, "global_step": 83686, "epoch": 940, "lr": 7.849011869765479e-05}
{"train_loss": 0.24339182674884796, "global_step": 83687, "epoch": 940, "lr": 7.848964227877575e-05}
{"train_loss": 0.19395823776721954, "global_step": 83688, "epoch": 940, "lr": 7.848916585606663e-05}
{"train_loss": 0.18005965650081635, "global_step": 83689, "epoch": 940, "lr": 7.848868942952749e-05}
{"train_loss": 0.11960382759571075, "global_step": 83690, "epoch": 940, "lr": 7.848821299915839e-05}
{"train_loss": 0.14651526510715485, "global_step": 83691, "epoch": 940, "lr": 7.848773656495938e-05}
{"train_loss": 0.17222712934017181, "global_step": 83692, "epoch": 940, "lr": 7.848726012693056e-05}
{"train_loss": 0.22958627343177795, "global_step": 83693, "epoch": 940, "lr": 7.848678368507198e-05}
{"train_loss": 0.18327383697032928, "global_step": 83694, "epoch": 940, "lr": 7.848630723938368e-05}
{"train_loss": 0.268318235874176, "global_step": 83695, "epoch": 940, "lr": 7.848583078986575e-05}
{"train_loss": 0.13983571529388428, "global_step": 83696, "epoch": 940, "lr": 7.848535433651826e-05}
{"train_loss": 0.19579243659973145, "global_step": 83697, "epoch": 940, "lr": 7.848487787934125e-05}
{"train_loss": 0.18310819566249847, "global_step": 83698, "epoch": 940, "lr": 7.84844014183348e-05}
{"train_loss": 0.14804930984973907, "global_step": 83699, "epoch": 940, "lr": 7.848392495349897e-05}
{"train_loss": 0.18515262007713318, "global_step": 83700, "epoch": 940, "lr": 7.848344848483383e-05}
{"train_loss": 0.21733050048351288, "global_step": 83701, "epoch": 940, "lr": 7.84829720123394e-05}
{"train_loss": 0.22072792053222656, "global_step": 83702, "epoch": 940, "lr": 7.848249553601584e-05}
{"train_loss": 0.14979013800621033, "global_step": 83703, "epoch": 940, "lr": 7.848201905586311e-05}
{"train_loss": 0.20654334127902985, "global_step": 83704, "epoch": 940, "lr": 7.848154257188134e-05}
{"train_loss": 0.1935308575630188, "global_step": 83705, "epoch": 940, "lr": 7.848106608407057e-05}
{"train_loss": 0.11954297125339508, "global_step": 83706, "epoch": 940, "lr": 7.848058959243086e-05}
{"train_loss": 0.2993910014629364, "global_step": 83707, "epoch": 940, "lr": 7.848011309696229e-05}
{"train_loss": 0.17626909911632538, "global_step": 83708, "epoch": 940, "lr": 7.84796365976649e-05}
{"train_loss": 0.13197173178195953, "global_step": 83709, "epoch": 940, "lr": 7.84791600945388e-05}
{"train_loss": 0.21286816895008087, "global_step": 83710, "epoch": 940, "lr": 7.847868358758402e-05}
{"train_loss": 0.15186414122581482, "global_step": 83711, "epoch": 940, "lr": 7.84782070768006e-05}
{"train_loss": 0.21375055611133575, "global_step": 83712, "epoch": 940, "lr": 7.847773056218865e-05}
{"train_loss": 0.16182434558868408, "global_step": 83713, "epoch": 940, "lr": 7.847725404374821e-05}
{"train_loss": 0.1889466792345047, "global_step": 83714, "epoch": 940, "lr": 7.847677752147936e-05}
{"train_loss": 0.24654842913150787, "global_step": 83715, "epoch": 940, "lr": 7.847630099538214e-05}
{"train_loss": 0.19077199697494507, "global_step": 83716, "epoch": 940, "lr": 7.847582446545665e-05}
{"train_loss": 0.19142600893974304, "global_step": 83717, "epoch": 940, "lr": 7.847534793170291e-05}
{"train_loss": 0.18602681159973145, "global_step": 83718, "epoch": 940, "lr": 7.847487139412103e-05}
{"train_loss": 0.1517491489648819, "global_step": 83719, "epoch": 940, "lr": 7.847439485271103e-05}
{"train_loss": 0.16428637504577637, "global_step": 83720, "epoch": 940, "lr": 7.847391830747301e-05}
{"train_loss": 0.14079444110393524, "global_step": 83721, "epoch": 940, "lr": 7.8473441758407e-05}
{"train_loss": 0.24442046880722046, "global_step": 83722, "epoch": 940, "lr": 7.847296520551311e-05}
{"train_loss": 0.2580752968788147, "global_step": 83723, "epoch": 940, "lr": 7.847248864879136e-05}
{"train_loss": 0.11837013065814972, "global_step": 83724, "epoch": 940, "lr": 7.847201208824182e-05}
{"train_loss": 0.15041178464889526, "global_step": 83725, "epoch": 940, "lr": 7.847153552386458e-05}
{"train_loss": 0.2599432170391083, "global_step": 83726, "epoch": 940, "lr": 7.847105895565969e-05}
{"train_loss": 0.15978698432445526, "global_step": 83727, "epoch": 940, "lr": 7.847058238362722e-05}
{"train_loss": 0.17975756525993347, "global_step": 83728, "epoch": 940, "lr": 7.84701058077672e-05}
{"train_loss": 0.1587149053812027, "global_step": 83729, "epoch": 940, "lr": 7.846962922807974e-05}
{"train_loss": 0.304096519947052, "global_step": 83730, "epoch": 940, "lr": 7.846915264456489e-05}
{"train_loss": 0.1871292144060135, "global_step": 83731, "epoch": 940, "lr": 7.846867605722268e-05}
{"train_loss": 0.2326921671628952, "global_step": 83732, "epoch": 940, "lr": 7.846819946605323e-05}
{"train_loss": 0.15638893842697144, "global_step": 83733, "epoch": 940, "lr": 7.846772287105657e-05}
{"train_loss": 0.13564810156822205, "global_step": 83734, "epoch": 940, "lr": 7.846724627223278e-05}
{"train_loss": 0.24791233241558075, "global_step": 83735, "epoch": 940, "lr": 7.846676966958191e-05}
{"train_loss": 0.21381208300590515, "global_step": 83736, "epoch": 940, "lr": 7.846629306310402e-05}
{"train_loss": 0.14173975586891174, "global_step": 83737, "epoch": 940, "lr": 7.84658164527992e-05}
{"train_loss": 0.23885869979858398, "global_step": 83738, "epoch": 940, "lr": 7.846533983866749e-05}
{"train_loss": 0.2700810432434082, "global_step": 83739, "epoch": 940, "lr": 7.846486322070896e-05}
{"train_loss": 0.15614579617977142, "global_step": 83740, "epoch": 940, "lr": 7.846438659892369e-05}
{"train_loss": 0.11327309906482697, "global_step": 83741, "epoch": 940, "lr": 7.846390997331172e-05}
{"train_loss": 0.2729732096195221, "global_step": 83742, "epoch": 940, "lr": 7.846343334387313e-05}
{"train_loss": 0.15750375390052795, "global_step": 83743, "epoch": 940, "lr": 7.846295671060797e-05}
{"train_loss": 0.20974406599998474, "global_step": 83744, "epoch": 940, "lr": 7.846248007351631e-05}
{"train_loss": 0.21883392333984375, "global_step": 83745, "epoch": 940, "lr": 7.846200343259824e-05}
{"train_loss": 0.1884593367576599, "global_step": 83746, "epoch": 940, "lr": 7.846152678785377e-05}
{"train_loss": 0.24390867352485657, "global_step": 83747, "epoch": 940, "lr": 7.8461050139283e-05}
{"train_loss": 0.18663925677537918, "global_step": 83748, "epoch": 940, "lr": 7.846057348688601e-05, "val_loss": 4.122034072875977, "train_action_mse_error": 15.832564353942871}
{"train_loss": 0.2569691240787506, "global_step": 83749, "epoch": 941, "lr": 7.846009683066282e-05}
{"train_loss": 0.20504435896873474, "global_step": 83750, "epoch": 941, "lr": 7.845962017061353e-05}
{"train_loss": 0.17473924160003662, "global_step": 83751, "epoch": 941, "lr": 7.845914350673818e-05}
{"train_loss": 0.19521860778331757, "global_step": 83752, "epoch": 941, "lr": 7.845866683903684e-05}
{"train_loss": 0.2335529327392578, "global_step": 83753, "epoch": 941, "lr": 7.845819016750959e-05}
{"train_loss": 0.28783926367759705, "global_step": 83754, "epoch": 941, "lr": 7.84577134921565e-05}
{"train_loss": 0.1972031593322754, "global_step": 83755, "epoch": 941, "lr": 7.845723681297759e-05}
{"train_loss": 0.24847115576267242, "global_step": 83756, "epoch": 941, "lr": 7.845676012997296e-05}
{"train_loss": 0.23422102630138397, "global_step": 83757, "epoch": 941, "lr": 7.845628344314266e-05}
{"train_loss": 0.18791432678699493, "global_step": 83758, "epoch": 941, "lr": 7.845580675248678e-05}
{"train_loss": 0.12678192555904388, "global_step": 83759, "epoch": 941, "lr": 7.845533005800535e-05}
{"train_loss": 0.16406622529029846, "global_step": 83760, "epoch": 941, "lr": 7.845485335969845e-05}
{"train_loss": 0.1862083226442337, "global_step": 83761, "epoch": 941, "lr": 7.845437665756614e-05}
{"train_loss": 0.18912549316883087, "global_step": 83762, "epoch": 941, "lr": 7.845389995160848e-05}
{"train_loss": 0.27382540702819824, "global_step": 83763, "epoch": 941, "lr": 7.845342324182555e-05}
{"train_loss": 0.16351206600666046, "global_step": 83764, "epoch": 941, "lr": 7.845294652821739e-05}
{"train_loss": 0.18744425475597382, "global_step": 83765, "epoch": 941, "lr": 7.845246981078411e-05}
{"train_loss": 0.2765098512172699, "global_step": 83766, "epoch": 941, "lr": 7.84519930895257e-05}
{"train_loss": 0.17249079048633575, "global_step": 83767, "epoch": 941, "lr": 7.84515163644423e-05}
{"train_loss": 0.23212271928787231, "global_step": 83768, "epoch": 941, "lr": 7.845103963553392e-05}
{"train_loss": 0.1679382175207138, "global_step": 83769, "epoch": 941, "lr": 7.845056290280065e-05}
{"train_loss": 0.210723415017128, "global_step": 83770, "epoch": 941, "lr": 7.845008616624254e-05}
{"train_loss": 0.16712266206741333, "global_step": 83771, "epoch": 941, "lr": 7.844960942585968e-05}
{"train_loss": 0.21096903085708618, "global_step": 83772, "epoch": 941, "lr": 7.84491326816521e-05}
{"train_loss": 0.17652061581611633, "global_step": 83773, "epoch": 941, "lr": 7.84486559336199e-05}
{"train_loss": 0.19882263243198395, "global_step": 83774, "epoch": 941, "lr": 7.844817918176311e-05}
{"train_loss": 0.24437418580055237, "global_step": 83775, "epoch": 941, "lr": 7.844770242608182e-05}
{"train_loss": 0.2095605880022049, "global_step": 83776, "epoch": 941, "lr": 7.844722566657607e-05}
{"train_loss": 0.2346048802137375, "global_step": 83777, "epoch": 941, "lr": 7.844674890324595e-05}
{"train_loss": 0.24864913523197174, "global_step": 83778, "epoch": 941, "lr": 7.844627213609151e-05}
{"train_loss": 0.17684604227542877, "global_step": 83779, "epoch": 941, "lr": 7.84457953651128e-05}
{"train_loss": 0.2781944274902344, "global_step": 83780, "epoch": 941, "lr": 7.844531859030992e-05}
{"train_loss": 0.15961985290050507, "global_step": 83781, "epoch": 941, "lr": 7.844484181168291e-05}
{"train_loss": 0.23927074670791626, "global_step": 83782, "epoch": 941, "lr": 7.844436502923183e-05}
{"train_loss": 0.1896507292985916, "global_step": 83783, "epoch": 941, "lr": 7.844388824295675e-05}
{"train_loss": 0.14087094366550446, "global_step": 83784, "epoch": 941, "lr": 7.844341145285774e-05}
{"train_loss": 0.19688628613948822, "global_step": 83785, "epoch": 941, "lr": 7.844293465893486e-05}
{"train_loss": 0.17863363027572632, "global_step": 83786, "epoch": 941, "lr": 7.844245786118816e-05}
{"train_loss": 0.16658245027065277, "global_step": 83787, "epoch": 941, "lr": 7.844198105961776e-05}
{"train_loss": 0.11465778946876526, "global_step": 83788, "epoch": 941, "lr": 7.844150425422364e-05}
{"train_loss": 0.18892206251621246, "global_step": 83789, "epoch": 941, "lr": 7.844102744500592e-05}
{"train_loss": 0.20857539772987366, "global_step": 83790, "epoch": 941, "lr": 7.844055063196465e-05}
{"train_loss": 0.1949654370546341, "global_step": 83791, "epoch": 941, "lr": 7.844007381509991e-05}
{"train_loss": 0.23590905964374542, "global_step": 83792, "epoch": 941, "lr": 7.843959699441172e-05}
{"train_loss": 0.15416768193244934, "global_step": 83793, "epoch": 941, "lr": 7.84391201699002e-05}
{"train_loss": 0.29893243312835693, "global_step": 83794, "epoch": 941, "lr": 7.843864334156536e-05}
{"train_loss": 0.1837136298418045, "global_step": 83795, "epoch": 941, "lr": 7.843816650940732e-05}
{"train_loss": 0.12379556149244308, "global_step": 83796, "epoch": 941, "lr": 7.84376896734261e-05}
{"train_loss": 0.16452182829380035, "global_step": 83797, "epoch": 941, "lr": 7.843721283362177e-05}
{"train_loss": 0.21124230325222015, "global_step": 83798, "epoch": 941, "lr": 7.843673598999442e-05}
{"train_loss": 0.1322138011455536, "global_step": 83799, "epoch": 941, "lr": 7.843625914254409e-05}
{"train_loss": 0.23190411925315857, "global_step": 83800, "epoch": 941, "lr": 7.843578229127086e-05}
{"train_loss": 0.292568176984787, "global_step": 83801, "epoch": 941, "lr": 7.843530543617477e-05}
{"train_loss": 0.17484022676944733, "global_step": 83802, "epoch": 941, "lr": 7.843482857725591e-05}
{"train_loss": 0.24884523451328278, "global_step": 83803, "epoch": 941, "lr": 7.843435171451433e-05}
{"train_loss": 0.16842088103294373, "global_step": 83804, "epoch": 941, "lr": 7.843387484795008e-05}
{"train_loss": 0.2191518098115921, "global_step": 83805, "epoch": 941, "lr": 7.843339797756327e-05}
{"train_loss": 0.26469460129737854, "global_step": 83806, "epoch": 941, "lr": 7.843292110335393e-05}
{"train_loss": 0.16880343854427338, "global_step": 83807, "epoch": 941, "lr": 7.843244422532213e-05}
{"train_loss": 0.15153692662715912, "global_step": 83808, "epoch": 941, "lr": 7.843196734346794e-05}
{"train_loss": 0.15506553649902344, "global_step": 83809, "epoch": 941, "lr": 7.843149045779141e-05}
{"train_loss": 0.12474169582128525, "global_step": 83810, "epoch": 941, "lr": 7.843101356829262e-05}
{"train_loss": 0.18643003702163696, "global_step": 83811, "epoch": 941, "lr": 7.843053667497161e-05}
{"train_loss": 0.20606374740600586, "global_step": 83812, "epoch": 941, "lr": 7.843005977782847e-05}
{"train_loss": 0.13180255889892578, "global_step": 83813, "epoch": 941, "lr": 7.842958287686326e-05}
{"train_loss": 0.19595752656459808, "global_step": 83814, "epoch": 941, "lr": 7.842910597207603e-05}
{"train_loss": 0.17853814363479614, "global_step": 83815, "epoch": 941, "lr": 7.842862906346686e-05}
{"train_loss": 0.13547441363334656, "global_step": 83816, "epoch": 941, "lr": 7.84281521510358e-05}
{"train_loss": 0.16732214391231537, "global_step": 83817, "epoch": 941, "lr": 7.842767523478294e-05}
{"train_loss": 0.17482754588127136, "global_step": 83818, "epoch": 941, "lr": 7.842719831470831e-05}
{"train_loss": 0.15569035708904266, "global_step": 83819, "epoch": 941, "lr": 7.842672139081199e-05}
{"train_loss": 0.17222712934017181, "global_step": 83820, "epoch": 941, "lr": 7.842624446309405e-05}
{"train_loss": 0.15651775896549225, "global_step": 83821, "epoch": 941, "lr": 7.842576753155454e-05}
{"train_loss": 0.19344939291477203, "global_step": 83822, "epoch": 941, "lr": 7.842529059619352e-05}
{"train_loss": 0.14496134221553802, "global_step": 83823, "epoch": 941, "lr": 7.842481365701109e-05}
{"train_loss": 0.15610764920711517, "global_step": 83824, "epoch": 941, "lr": 7.842433671400728e-05}
{"train_loss": 0.2141444981098175, "global_step": 83825, "epoch": 941, "lr": 7.842385976718219e-05}
{"train_loss": 0.14821483194828033, "global_step": 83826, "epoch": 941, "lr": 7.842338281653583e-05}
{"train_loss": 0.17679762840270996, "global_step": 83827, "epoch": 941, "lr": 7.842290586206829e-05}
{"train_loss": 0.1373189389705658, "global_step": 83828, "epoch": 941, "lr": 7.842242890377964e-05}
{"train_loss": 0.19010958075523376, "global_step": 83829, "epoch": 941, "lr": 7.842195194166994e-05}
{"train_loss": 0.19406430423259735, "global_step": 83830, "epoch": 941, "lr": 7.842147497573927e-05}
{"train_loss": 0.24475249648094177, "global_step": 83831, "epoch": 941, "lr": 7.842099800598768e-05}
{"train_loss": 0.15245841443538666, "global_step": 83832, "epoch": 941, "lr": 7.842052103241522e-05}
{"train_loss": 0.18557821214199066, "global_step": 83833, "epoch": 941, "lr": 7.842004405502197e-05}
{"train_loss": 0.17768731713294983, "global_step": 83834, "epoch": 941, "lr": 7.841956707380799e-05}
{"train_loss": 0.11887305229902267, "global_step": 83835, "epoch": 941, "lr": 7.841909008877335e-05}
{"train_loss": 0.17985586822032928, "global_step": 83836, "epoch": 941, "lr": 7.84186130999181e-05}
{"train_loss": 0.19191263942571168, "global_step": 83837, "epoch": 941, "lr": 7.841813610724232e-05, "val_loss": 4.197658061981201}
{"train_loss": 0.17703337967395782, "global_step": 83838, "epoch": 942, "lr": 7.841765911074608e-05}
{"train_loss": 0.20782305300235748, "global_step": 83839, "epoch": 942, "lr": 7.841718211042942e-05}
{"train_loss": 0.23683691024780273, "global_step": 83840, "epoch": 942, "lr": 7.84167051062924e-05}
{"train_loss": 0.14743053913116455, "global_step": 83841, "epoch": 942, "lr": 7.841622809833512e-05}
{"train_loss": 0.14122223854064941, "global_step": 83842, "epoch": 942, "lr": 7.841575108655763e-05}
{"train_loss": 0.14149129390716553, "global_step": 83843, "epoch": 942, "lr": 7.841527407095998e-05}
{"train_loss": 0.1687808334827423, "global_step": 83844, "epoch": 942, "lr": 7.841479705154224e-05}
{"train_loss": 0.23860467970371246, "global_step": 83845, "epoch": 942, "lr": 7.841432002830449e-05}
{"train_loss": 0.1540789157152176, "global_step": 83846, "epoch": 942, "lr": 7.841384300124677e-05}
{"train_loss": 0.15709282457828522, "global_step": 83847, "epoch": 942, "lr": 7.841336597036915e-05}
{"train_loss": 0.10710106790065765, "global_step": 83848, "epoch": 942, "lr": 7.841288893567172e-05}
{"train_loss": 0.12227898836135864, "global_step": 83849, "epoch": 942, "lr": 7.84124118971545e-05}
{"train_loss": 0.1554856300354004, "global_step": 83850, "epoch": 942, "lr": 7.841193485481759e-05}
{"train_loss": 0.14894181489944458, "global_step": 83851, "epoch": 942, "lr": 7.841145780866104e-05}
{"train_loss": 0.17607903480529785, "global_step": 83852, "epoch": 942, "lr": 7.841098075868491e-05}
{"train_loss": 0.1782447248697281, "global_step": 83853, "epoch": 942, "lr": 7.841050370488927e-05}
{"train_loss": 0.16827735304832458, "global_step": 83854, "epoch": 942, "lr": 7.84100266472742e-05}
{"train_loss": 0.22785639762878418, "global_step": 83855, "epoch": 942, "lr": 7.840954958583973e-05}
{"train_loss": 0.20166435837745667, "global_step": 83856, "epoch": 942, "lr": 7.840907252058593e-05}
{"train_loss": 0.25258803367614746, "global_step": 83857, "epoch": 942, "lr": 7.840859545151291e-05}
{"train_loss": 0.12399090081453323, "global_step": 83858, "epoch": 942, "lr": 7.840811837862068e-05}
{"train_loss": 0.17600007355213165, "global_step": 83859, "epoch": 942, "lr": 7.840764130190932e-05}
{"train_loss": 0.15069760382175446, "global_step": 83860, "epoch": 942, "lr": 7.840716422137891e-05}
{"train_loss": 0.15287883579730988, "global_step": 83861, "epoch": 942, "lr": 7.84066871370295e-05}
{"train_loss": 0.1354127824306488, "global_step": 83862, "epoch": 942, "lr": 7.840621004886118e-05}
{"train_loss": 0.23173771798610687, "global_step": 83863, "epoch": 942, "lr": 7.840573295687395e-05}
{"train_loss": 0.1928175985813141, "global_step": 83864, "epoch": 942, "lr": 7.840525586106795e-05}
{"train_loss": 0.16311226785182953, "global_step": 83865, "epoch": 942, "lr": 7.84047787614432e-05}
{"train_loss": 0.18966519832611084, "global_step": 83866, "epoch": 942, "lr": 7.840430165799977e-05}
{"train_loss": 0.12798917293548584, "global_step": 83867, "epoch": 942, "lr": 7.840382455073772e-05}
{"train_loss": 0.24028366804122925, "global_step": 83868, "epoch": 942, "lr": 7.840334743965712e-05}
{"train_loss": 0.22101524472236633, "global_step": 83869, "epoch": 942, "lr": 7.840287032475806e-05}
{"train_loss": 0.1816926747560501, "global_step": 83870, "epoch": 942, "lr": 7.840239320604056e-05}
{"train_loss": 0.26110899448394775, "global_step": 83871, "epoch": 942, "lr": 7.84019160835047e-05}
{"train_loss": 0.18395927548408508, "global_step": 83872, "epoch": 942, "lr": 7.840143895715057e-05}
{"train_loss": 0.24001596868038177, "global_step": 83873, "epoch": 942, "lr": 7.84009618269782e-05}
{"train_loss": 0.16557635366916656, "global_step": 83874, "epoch": 942, "lr": 7.840048469298766e-05}
{"train_loss": 0.22929376363754272, "global_step": 83875, "epoch": 942, "lr": 7.840000755517904e-05}
{"train_loss": 0.14676588773727417, "global_step": 83876, "epoch": 942, "lr": 7.839953041355237e-05}
{"train_loss": 0.16952455043792725, "global_step": 83877, "epoch": 942, "lr": 7.839905326810773e-05}
{"train_loss": 0.14698565006256104, "global_step": 83878, "epoch": 942, "lr": 7.839857611884518e-05}
{"train_loss": 0.14566835761070251, "global_step": 83879, "epoch": 942, "lr": 7.83980989657648e-05}
{"train_loss": 0.21475879848003387, "global_step": 83880, "epoch": 942, "lr": 7.839762180886663e-05}
{"train_loss": 0.2651050090789795, "global_step": 83881, "epoch": 942, "lr": 7.839714464815076e-05}
{"train_loss": 0.1875593662261963, "global_step": 83882, "epoch": 942, "lr": 7.839666748361722e-05}
{"train_loss": 0.2311299741268158, "global_step": 83883, "epoch": 942, "lr": 7.839619031526612e-05}
{"train_loss": 0.17392651736736298, "global_step": 83884, "epoch": 942, "lr": 7.839571314309748e-05}
{"train_loss": 0.2411944568157196, "global_step": 83885, "epoch": 942, "lr": 7.83952359671114e-05}
{"train_loss": 0.2043614387512207, "global_step": 83886, "epoch": 942, "lr": 7.839475878730791e-05}
{"train_loss": 0.16899387538433075, "global_step": 83887, "epoch": 942, "lr": 7.839428160368709e-05}
{"train_loss": 0.08748035132884979, "global_step": 83888, "epoch": 942, "lr": 7.8393804416249e-05}
{"train_loss": 0.18219606578350067, "global_step": 83889, "epoch": 942, "lr": 7.839332722499373e-05}
{"train_loss": 0.22205086052417755, "global_step": 83890, "epoch": 942, "lr": 7.83928500299213e-05}
{"train_loss": 0.15053346753120422, "global_step": 83891, "epoch": 942, "lr": 7.839237283103181e-05}
{"train_loss": 0.1476922482252121, "global_step": 83892, "epoch": 942, "lr": 7.839189562832532e-05}
{"train_loss": 0.20800116658210754, "global_step": 83893, "epoch": 942, "lr": 7.839141842180186e-05}
{"train_loss": 0.20304957032203674, "global_step": 83894, "epoch": 942, "lr": 7.839094121146155e-05}
{"train_loss": 0.20455878973007202, "global_step": 83895, "epoch": 942, "lr": 7.83904639973044e-05}
{"train_loss": 0.26401588320732117, "global_step": 83896, "epoch": 942, "lr": 7.838998677933051e-05}
{"train_loss": 0.26628896594047546, "global_step": 83897, "epoch": 942, "lr": 7.838950955753993e-05}
{"train_loss": 0.33779212832450867, "global_step": 83898, "epoch": 942, "lr": 7.838903233193273e-05}
{"train_loss": 0.22679664194583893, "global_step": 83899, "epoch": 942, "lr": 7.838855510250897e-05}
{"train_loss": 0.2308100163936615, "global_step": 83900, "epoch": 942, "lr": 7.83880778692687e-05}
{"train_loss": 0.16252931952476501, "global_step": 83901, "epoch": 942, "lr": 7.838760063221202e-05}
{"train_loss": 0.18888600170612335, "global_step": 83902, "epoch": 942, "lr": 7.838712339133896e-05}
{"train_loss": 0.2816295325756073, "global_step": 83903, "epoch": 942, "lr": 7.838664614664962e-05}
{"train_loss": 0.18297941982746124, "global_step": 83904, "epoch": 942, "lr": 7.838616889814401e-05}
{"train_loss": 0.16364946961402893, "global_step": 83905, "epoch": 942, "lr": 7.838569164582223e-05}
{"train_loss": 0.17010222375392914, "global_step": 83906, "epoch": 942, "lr": 7.838521438968435e-05}
{"train_loss": 0.21573962271213531, "global_step": 83907, "epoch": 942, "lr": 7.838473712973042e-05}
{"train_loss": 0.15190401673316956, "global_step": 83908, "epoch": 942, "lr": 7.838425986596052e-05}
{"train_loss": 0.17729419469833374, "global_step": 83909, "epoch": 942, "lr": 7.838378259837469e-05}
{"train_loss": 0.24453721940517426, "global_step": 83910, "epoch": 942, "lr": 7.838330532697301e-05}
{"train_loss": 0.23539389669895172, "global_step": 83911, "epoch": 942, "lr": 7.838282805175554e-05}
{"train_loss": 0.20343345403671265, "global_step": 83912, "epoch": 942, "lr": 7.838235077272235e-05}
{"train_loss": 0.23528724908828735, "global_step": 83913, "epoch": 942, "lr": 7.838187348987348e-05}
{"train_loss": 0.19920925796031952, "global_step": 83914, "epoch": 942, "lr": 7.838139620320903e-05}
{"train_loss": 0.15454508364200592, "global_step": 83915, "epoch": 942, "lr": 7.838091891272904e-05}
{"train_loss": 0.19910871982574463, "global_step": 83916, "epoch": 942, "lr": 7.838044161843359e-05}
{"train_loss": 0.1309807300567627, "global_step": 83917, "epoch": 942, "lr": 7.837996432032273e-05}
{"train_loss": 0.15235504508018494, "global_step": 83918, "epoch": 942, "lr": 7.837948701839652e-05}
{"train_loss": 0.16111578047275543, "global_step": 83919, "epoch": 942, "lr": 7.837900971265505e-05}
{"train_loss": 0.208492711186409, "global_step": 83920, "epoch": 942, "lr": 7.837853240309836e-05}
{"train_loss": 0.12785328924655914, "global_step": 83921, "epoch": 942, "lr": 7.837805508972652e-05}
{"train_loss": 0.26602497696876526, "global_step": 83922, "epoch": 942, "lr": 7.83775777725396e-05}
{"train_loss": 0.17986956238746643, "global_step": 83923, "epoch": 942, "lr": 7.837710045153767e-05}
{"train_loss": 0.2370329052209854, "global_step": 83924, "epoch": 942, "lr": 7.837662312672078e-05}
{"train_loss": 0.2548592686653137, "global_step": 83925, "epoch": 942, "lr": 7.837614579808898e-05}
{"train_loss": 0.19009868524382623, "global_step": 83926, "epoch": 942, "lr": 7.837566846564237e-05, "val_loss": 4.16101598739624}
{"train_loss": 0.11518994718790054, "global_step": 83927, "epoch": 943, "lr": 7.837519112938099e-05}
{"train_loss": 0.29472923278808594, "global_step": 83928, "epoch": 943, "lr": 7.837471378930492e-05}
{"train_loss": 0.1734939068555832, "global_step": 83929, "epoch": 943, "lr": 7.837423644541421e-05}
{"train_loss": 0.25176864862442017, "global_step": 83930, "epoch": 943, "lr": 7.837375909770894e-05}
{"train_loss": 0.1326156109571457, "global_step": 83931, "epoch": 943, "lr": 7.837328174618914e-05}
{"train_loss": 0.23248156905174255, "global_step": 83932, "epoch": 943, "lr": 7.837280439085492e-05}
{"train_loss": 0.1358531415462494, "global_step": 83933, "epoch": 943, "lr": 7.837232703170632e-05}
{"train_loss": 0.1901882290840149, "global_step": 83934, "epoch": 943, "lr": 7.83718496687434e-05}
{"train_loss": 0.25842949748039246, "global_step": 83935, "epoch": 943, "lr": 7.837137230196623e-05}
{"train_loss": 0.20504392683506012, "global_step": 83936, "epoch": 943, "lr": 7.837089493137488e-05}
{"train_loss": 0.20521405339241028, "global_step": 83937, "epoch": 943, "lr": 7.837041755696941e-05}
{"train_loss": 0.1523086130619049, "global_step": 83938, "epoch": 943, "lr": 7.836994017874988e-05}
{"train_loss": 0.10936965048313141, "global_step": 83939, "epoch": 943, "lr": 7.836946279671635e-05}
{"train_loss": 0.19304119050502777, "global_step": 83940, "epoch": 943, "lr": 7.83689854108689e-05}
{"train_loss": 0.14897021651268005, "global_step": 83941, "epoch": 943, "lr": 7.836850802120759e-05}
{"train_loss": 0.21150238811969757, "global_step": 83942, "epoch": 943, "lr": 7.836803062773248e-05}
{"train_loss": 0.17324988543987274, "global_step": 83943, "epoch": 943, "lr": 7.836755323044362e-05}
{"train_loss": 0.17583461105823517, "global_step": 83944, "epoch": 943, "lr": 7.83670758293411e-05}
{"train_loss": 0.16885028779506683, "global_step": 83945, "epoch": 943, "lr": 7.836659842442496e-05}
{"train_loss": 0.20110931992530823, "global_step": 83946, "epoch": 943, "lr": 7.83661210156953e-05}
{"train_loss": 0.203129380941391, "global_step": 83947, "epoch": 943, "lr": 7.836564360315212e-05}
{"train_loss": 0.16333027184009552, "global_step": 83948, "epoch": 943, "lr": 7.836516618679557e-05}
{"train_loss": 0.1958772838115692, "global_step": 83949, "epoch": 943, "lr": 7.836468876662564e-05}
{"train_loss": 0.24865514039993286, "global_step": 83950, "epoch": 943, "lr": 7.836421134264244e-05}
{"train_loss": 0.17877605557441711, "global_step": 83951, "epoch": 943, "lr": 7.8363733914846e-05}
{"train_loss": 0.21417419612407684, "global_step": 83952, "epoch": 943, "lr": 7.836325648323642e-05}
{"train_loss": 0.18276795744895935, "global_step": 83953, "epoch": 943, "lr": 7.836277904781373e-05}
{"train_loss": 0.18175353109836578, "global_step": 83954, "epoch": 943, "lr": 7.836230160857803e-05}
{"train_loss": 0.26508742570877075, "global_step": 83955, "epoch": 943, "lr": 7.836182416552935e-05}
{"train_loss": 0.1218005120754242, "global_step": 83956, "epoch": 943, "lr": 7.836134671866778e-05}
{"train_loss": 0.16875168681144714, "global_step": 83957, "epoch": 943, "lr": 7.836086926799335e-05}
{"train_loss": 0.15774202346801758, "global_step": 83958, "epoch": 943, "lr": 7.836039181350616e-05}
{"train_loss": 0.12807485461235046, "global_step": 83959, "epoch": 943, "lr": 7.835991435520627e-05}
{"train_loss": 0.27076590061187744, "global_step": 83960, "epoch": 943, "lr": 7.835943689309372e-05}
{"train_loss": 0.17202478647232056, "global_step": 83961, "epoch": 943, "lr": 7.83589594271686e-05}
{"train_loss": 0.18280285596847534, "global_step": 83962, "epoch": 943, "lr": 7.835848195743097e-05}
{"train_loss": 0.12202180176973343, "global_step": 83963, "epoch": 943, "lr": 7.835800448388086e-05}
{"train_loss": 0.13586755096912384, "global_step": 83964, "epoch": 943, "lr": 7.835752700651839e-05}
{"train_loss": 0.17580492794513702, "global_step": 83965, "epoch": 943, "lr": 7.835704952534358e-05}
{"train_loss": 0.14561353623867035, "global_step": 83966, "epoch": 943, "lr": 7.835657204035651e-05}
{"train_loss": 0.23252177238464355, "global_step": 83967, "epoch": 943, "lr": 7.835609455155725e-05}
{"train_loss": 0.17236894369125366, "global_step": 83968, "epoch": 943, "lr": 7.835561705894587e-05}
{"train_loss": 0.20551545917987823, "global_step": 83969, "epoch": 943, "lr": 7.83551395625224e-05}
{"train_loss": 0.1695767641067505, "global_step": 83970, "epoch": 943, "lr": 7.835466206228694e-05}
{"train_loss": 0.16095542907714844, "global_step": 83971, "epoch": 943, "lr": 7.835418455823954e-05}
{"train_loss": 0.1636422574520111, "global_step": 83972, "epoch": 943, "lr": 7.835370705038028e-05}
{"train_loss": 0.19034363329410553, "global_step": 83973, "epoch": 943, "lr": 7.835322953870918e-05}
{"train_loss": 0.17192292213439941, "global_step": 83974, "epoch": 943, "lr": 7.835275202322635e-05}
{"train_loss": 0.18958613276481628, "global_step": 83975, "epoch": 943, "lr": 7.835227450393183e-05}
{"train_loss": 0.20050948858261108, "global_step": 83976, "epoch": 943, "lr": 7.83517969808257e-05}
{"train_loss": 0.15466031432151794, "global_step": 83977, "epoch": 943, "lr": 7.835131945390802e-05}
{"train_loss": 0.19462347030639648, "global_step": 83978, "epoch": 943, "lr": 7.835084192317886e-05}
{"train_loss": 0.16396979987621307, "global_step": 83979, "epoch": 943, "lr": 7.835036438863825e-05}
{"train_loss": 0.15969058871269226, "global_step": 83980, "epoch": 943, "lr": 7.83498868502863e-05}
{"train_loss": 0.22222301363945007, "global_step": 83981, "epoch": 943, "lr": 7.834940930812304e-05}
{"train_loss": 0.26400551199913025, "global_step": 83982, "epoch": 943, "lr": 7.834893176214855e-05}
{"train_loss": 0.1463586390018463, "global_step": 83983, "epoch": 943, "lr": 7.83484542123629e-05}
{"train_loss": 0.20605546236038208, "global_step": 83984, "epoch": 943, "lr": 7.834797665876614e-05}
{"train_loss": 0.18335658311843872, "global_step": 83985, "epoch": 943, "lr": 7.834749910135832e-05}
{"train_loss": 0.13903413712978363, "global_step": 83986, "epoch": 943, "lr": 7.834702154013956e-05}
{"train_loss": 0.18355034291744232, "global_step": 83987, "epoch": 943, "lr": 7.834654397510986e-05}
{"train_loss": 0.16153249144554138, "global_step": 83988, "epoch": 943, "lr": 7.834606640626933e-05}
{"train_loss": 0.14173699915409088, "global_step": 83989, "epoch": 943, "lr": 7.8345588833618e-05}
{"train_loss": 0.19843952357769012, "global_step": 83990, "epoch": 943, "lr": 7.834511125715596e-05}
{"train_loss": 0.19127950072288513, "global_step": 83991, "epoch": 943, "lr": 7.834463367688327e-05}
{"train_loss": 0.15095986425876617, "global_step": 83992, "epoch": 943, "lr": 7.834415609279999e-05}
{"train_loss": 0.14294098317623138, "global_step": 83993, "epoch": 943, "lr": 7.834367850490618e-05}
{"train_loss": 0.20650887489318848, "global_step": 83994, "epoch": 943, "lr": 7.83432009132019e-05}
{"train_loss": 0.13831518590450287, "global_step": 83995, "epoch": 943, "lr": 7.834272331768724e-05}
{"train_loss": 0.27153393626213074, "global_step": 83996, "epoch": 943, "lr": 7.834224571836225e-05}
{"train_loss": 0.14503756165504456, "global_step": 83997, "epoch": 943, "lr": 7.834176811522697e-05}
{"train_loss": 0.19598248600959778, "global_step": 83998, "epoch": 943, "lr": 7.83412905082815e-05}
{"train_loss": 0.1930142492055893, "global_step": 83999, "epoch": 943, "lr": 7.834081289752588e-05}
{"train_loss": 0.22151628136634827, "global_step": 84000, "epoch": 943, "lr": 7.834033528296018e-05}
{"train_loss": 0.25821083784103394, "global_step": 84001, "epoch": 943, "lr": 7.833985766458448e-05}
{"train_loss": 0.19865155220031738, "global_step": 84002, "epoch": 943, "lr": 7.83393800423988e-05}
{"train_loss": 0.2110360860824585, "global_step": 84003, "epoch": 943, "lr": 7.833890241640326e-05}
{"train_loss": 0.1538086235523224, "global_step": 84004, "epoch": 943, "lr": 7.83384247865979e-05}
{"train_loss": 0.17561359703540802, "global_step": 84005, "epoch": 943, "lr": 7.833794715298279e-05}
{"train_loss": 0.2782950699329376, "global_step": 84006, "epoch": 943, "lr": 7.833746951555798e-05}
{"train_loss": 0.2560548186302185, "global_step": 84007, "epoch": 943, "lr": 7.833699187432354e-05}
{"train_loss": 0.17999742925167084, "global_step": 84008, "epoch": 943, "lr": 7.833651422927955e-05}
{"train_loss": 0.19619151949882507, "global_step": 84009, "epoch": 943, "lr": 7.833603658042604e-05}
{"train_loss": 0.16249977052211761, "global_step": 84010, "epoch": 943, "lr": 7.833555892776311e-05}
{"train_loss": 0.18987727165222168, "global_step": 84011, "epoch": 943, "lr": 7.83350812712908e-05}
{"train_loss": 0.14031994342803955, "global_step": 84012, "epoch": 943, "lr": 7.83346036110092e-05}
{"train_loss": 0.19346356391906738, "global_step": 84013, "epoch": 943, "lr": 7.833412594691834e-05}
{"train_loss": 0.14470750093460083, "global_step": 84014, "epoch": 943, "lr": 7.833364827901831e-05}
{"train_loss": 0.1850487809167819, "global_step": 84015, "epoch": 943, "lr": 7.833317060730917e-05, "val_loss": 4.2358317375183105}
{"train_loss": 0.1115756407380104, "global_step": 84016, "epoch": 944, "lr": 7.833269293179097e-05}
{"train_loss": 0.14873211085796356, "global_step": 84017, "epoch": 944, "lr": 7.83322152524638e-05}
{"train_loss": 0.14579808712005615, "global_step": 84018, "epoch": 944, "lr": 7.83317375693277e-05}
{"train_loss": 0.19860170781612396, "global_step": 84019, "epoch": 944, "lr": 7.833125988238275e-05}
{"train_loss": 0.22266021370887756, "global_step": 84020, "epoch": 944, "lr": 7.833078219162899e-05}
{"train_loss": 0.21214821934700012, "global_step": 84021, "epoch": 944, "lr": 7.833030449706652e-05}
{"train_loss": 0.23437801003456116, "global_step": 84022, "epoch": 944, "lr": 7.832982679869537e-05}
{"train_loss": 0.1807202249765396, "global_step": 84023, "epoch": 944, "lr": 7.832934909651564e-05}
{"train_loss": 0.0664258822798729, "global_step": 84024, "epoch": 944, "lr": 7.832887139052736e-05}
{"train_loss": 0.151262566447258, "global_step": 84025, "epoch": 944, "lr": 7.83283936807306e-05}
{"train_loss": 0.22136928141117096, "global_step": 84026, "epoch": 944, "lr": 7.832791596712546e-05}
{"train_loss": 0.16726942360401154, "global_step": 84027, "epoch": 944, "lr": 7.832743824971196e-05}
{"train_loss": 0.2232300043106079, "global_step": 84028, "epoch": 944, "lr": 7.832696052849018e-05}
{"train_loss": 0.17565076053142548, "global_step": 84029, "epoch": 944, "lr": 7.832648280346018e-05}
{"train_loss": 0.16606028378009796, "global_step": 84030, "epoch": 944, "lr": 7.832600507462205e-05}
{"train_loss": 0.31452834606170654, "global_step": 84031, "epoch": 944, "lr": 7.832552734197581e-05}
{"train_loss": 0.22115489840507507, "global_step": 84032, "epoch": 944, "lr": 7.832504960552156e-05}
{"train_loss": 0.2890184223651886, "global_step": 84033, "epoch": 944, "lr": 7.832457186525936e-05}
{"train_loss": 0.21936260163784027, "global_step": 84034, "epoch": 944, "lr": 7.832409412118927e-05}
{"train_loss": 0.14464539289474487, "global_step": 84035, "epoch": 944, "lr": 7.832361637331132e-05}
{"train_loss": 0.15577886998653412, "global_step": 84036, "epoch": 944, "lr": 7.832313862162563e-05}
{"train_loss": 0.16271434724330902, "global_step": 84037, "epoch": 944, "lr": 7.832266086613223e-05}
{"train_loss": 0.16236785054206848, "global_step": 84038, "epoch": 944, "lr": 7.83221831068312e-05}
{"train_loss": 0.27122315764427185, "global_step": 84039, "epoch": 944, "lr": 7.83217053437226e-05}
{"train_loss": 0.13529887795448303, "global_step": 84040, "epoch": 944, "lr": 7.832122757680646e-05}
{"train_loss": 0.1571587324142456, "global_step": 84041, "epoch": 944, "lr": 7.832074980608291e-05}
{"train_loss": 0.20294001698493958, "global_step": 84042, "epoch": 944, "lr": 7.832027203155198e-05}
{"train_loss": 0.11187148094177246, "global_step": 84043, "epoch": 944, "lr": 7.831979425321372e-05}
{"train_loss": 0.17225977778434753, "global_step": 84044, "epoch": 944, "lr": 7.831931647106823e-05}
{"train_loss": 0.1793038696050644, "global_step": 84045, "epoch": 944, "lr": 7.831883868511554e-05}
{"train_loss": 0.14677001535892487, "global_step": 84046, "epoch": 944, "lr": 7.831836089535572e-05}
{"train_loss": 0.19406017661094666, "global_step": 84047, "epoch": 944, "lr": 7.831788310178884e-05}
{"train_loss": 0.22091366350650787, "global_step": 84048, "epoch": 944, "lr": 7.831740530441498e-05}
{"train_loss": 0.25565478205680847, "global_step": 84049, "epoch": 944, "lr": 7.831692750323418e-05}
{"train_loss": 0.17783191800117493, "global_step": 84050, "epoch": 944, "lr": 7.831644969824654e-05}
{"train_loss": 0.2173861265182495, "global_step": 84051, "epoch": 944, "lr": 7.831597188945207e-05}
{"train_loss": 0.2749044895172119, "global_step": 84052, "epoch": 944, "lr": 7.831549407685085e-05}
{"train_loss": 0.1473589390516281, "global_step": 84053, "epoch": 944, "lr": 7.8315016260443e-05}
{"train_loss": 0.11124248802661896, "global_step": 84054, "epoch": 944, "lr": 7.831453844022852e-05}
{"train_loss": 0.2273849993944168, "global_step": 84055, "epoch": 944, "lr": 7.831406061620747e-05}
{"train_loss": 0.1961699277162552, "global_step": 84056, "epoch": 944, "lr": 7.831358278837997e-05}
{"train_loss": 0.2537720799446106, "global_step": 84057, "epoch": 944, "lr": 7.831310495674605e-05}
{"train_loss": 0.1829124093055725, "global_step": 84058, "epoch": 944, "lr": 7.831262712130578e-05}
{"train_loss": 0.1575879156589508, "global_step": 84059, "epoch": 944, "lr": 7.831214928205921e-05}
{"train_loss": 0.28989896178245544, "global_step": 84060, "epoch": 944, "lr": 7.831167143900642e-05}
{"train_loss": 0.09722345322370529, "global_step": 84061, "epoch": 944, "lr": 7.831119359214747e-05}
{"train_loss": 0.17121998965740204, "global_step": 84062, "epoch": 944, "lr": 7.831071574148243e-05}
{"train_loss": 0.23536701500415802, "global_step": 84063, "epoch": 944, "lr": 7.831023788701136e-05}
{"train_loss": 0.22089192271232605, "global_step": 84064, "epoch": 944, "lr": 7.830976002873434e-05}
{"train_loss": 0.21125836670398712, "global_step": 84065, "epoch": 944, "lr": 7.830928216665138e-05}
{"train_loss": 0.17921233177185059, "global_step": 84066, "epoch": 944, "lr": 7.830880430076261e-05}
{"train_loss": 0.14897485077381134, "global_step": 84067, "epoch": 944, "lr": 7.830832643106807e-05}
{"train_loss": 0.15882183611392975, "global_step": 84068, "epoch": 944, "lr": 7.830784855756781e-05}
{"train_loss": 0.1919908970594406, "global_step": 84069, "epoch": 944, "lr": 7.83073706802619e-05}
{"train_loss": 0.1876148283481598, "global_step": 84070, "epoch": 944, "lr": 7.83068927991504e-05}
{"train_loss": 0.16206875443458557, "global_step": 84071, "epoch": 944, "lr": 7.83064149142334e-05}
{"train_loss": 0.24890050292015076, "global_step": 84072, "epoch": 944, "lr": 7.830593702551094e-05}
{"train_loss": 0.16629526019096375, "global_step": 84073, "epoch": 944, "lr": 7.83054591329831e-05}
{"train_loss": 0.23815451562404633, "global_step": 84074, "epoch": 944, "lr": 7.830498123664993e-05}
{"train_loss": 0.1632087230682373, "global_step": 84075, "epoch": 944, "lr": 7.830450333651151e-05}
{"train_loss": 0.24431805312633514, "global_step": 84076, "epoch": 944, "lr": 7.830402543256788e-05}
{"train_loss": 0.2631450891494751, "global_step": 84077, "epoch": 944, "lr": 7.830354752481912e-05}
{"train_loss": 0.2885696589946747, "global_step": 84078, "epoch": 944, "lr": 7.83030696132653e-05}
{"train_loss": 0.24402445554733276, "global_step": 84079, "epoch": 944, "lr": 7.830259169790647e-05}
{"train_loss": 0.2012445032596588, "global_step": 84080, "epoch": 944, "lr": 7.830211377874272e-05}
{"train_loss": 0.24410037696361542, "global_step": 84081, "epoch": 944, "lr": 7.830163585577407e-05}
{"train_loss": 0.21356385946273804, "global_step": 84082, "epoch": 944, "lr": 7.830115792900064e-05}
{"train_loss": 0.1169176772236824, "global_step": 84083, "epoch": 944, "lr": 7.830067999842243e-05}
{"train_loss": 0.23198780417442322, "global_step": 84084, "epoch": 944, "lr": 7.830020206403957e-05}
{"train_loss": 0.2012879103422165, "global_step": 84085, "epoch": 944, "lr": 7.829972412585207e-05}
{"train_loss": 0.18177808821201324, "global_step": 84086, "epoch": 944, "lr": 7.829924618386002e-05}
{"train_loss": 0.2010745108127594, "global_step": 84087, "epoch": 944, "lr": 7.82987682380635e-05}
{"train_loss": 0.17553061246871948, "global_step": 84088, "epoch": 944, "lr": 7.829829028846254e-05}
{"train_loss": 0.20438992977142334, "global_step": 84089, "epoch": 944, "lr": 7.829781233505723e-05}
{"train_loss": 0.18997934460639954, "global_step": 84090, "epoch": 944, "lr": 7.829733437784761e-05}
{"train_loss": 0.23477427661418915, "global_step": 84091, "epoch": 944, "lr": 7.829685641683376e-05}
{"train_loss": 0.16034777462482452, "global_step": 84092, "epoch": 944, "lr": 7.829637845201574e-05}
{"train_loss": 0.2792075276374817, "global_step": 84093, "epoch": 944, "lr": 7.829590048339363e-05}
{"train_loss": 0.26345983147621155, "global_step": 84094, "epoch": 944, "lr": 7.829542251096746e-05}
{"train_loss": 0.20861394703388214, "global_step": 84095, "epoch": 944, "lr": 7.829494453473734e-05}
{"train_loss": 0.22502756118774414, "global_step": 84096, "epoch": 944, "lr": 7.82944665547033e-05}
{"train_loss": 0.13121812045574188, "global_step": 84097, "epoch": 944, "lr": 7.82939885708654e-05}
{"train_loss": 0.182338684797287, "global_step": 84098, "epoch": 944, "lr": 7.829351058322374e-05}
{"train_loss": 0.2086331993341446, "global_step": 84099, "epoch": 944, "lr": 7.829303259177836e-05}
{"train_loss": 0.1799757182598114, "global_step": 84100, "epoch": 944, "lr": 7.82925545965293e-05}
{"train_loss": 0.24757380783557892, "global_step": 84101, "epoch": 944, "lr": 7.829207659747668e-05}
{"train_loss": 0.15746229887008667, "global_step": 84102, "epoch": 944, "lr": 7.829159859462053e-05}
{"train_loss": 0.17551498115062714, "global_step": 84103, "epoch": 944, "lr": 7.829112058796092e-05}
{"train_loss": 0.19502991873226808, "global_step": 84104, "epoch": 944, "lr": 7.829064257749792e-05, "val_loss": 4.200808525085449}
{"train_loss": 0.2460833638906479, "global_step": 84105, "epoch": 945, "lr": 7.829016456323157e-05}
{"train_loss": 0.20567691326141357, "global_step": 84106, "epoch": 945, "lr": 7.828968654516196e-05}
{"train_loss": 0.2409180849790573, "global_step": 84107, "epoch": 945, "lr": 7.828920852328914e-05}
{"train_loss": 0.1674368530511856, "global_step": 84108, "epoch": 945, "lr": 7.828873049761318e-05}
{"train_loss": 0.19905701279640198, "global_step": 84109, "epoch": 945, "lr": 7.828825246813416e-05}
{"train_loss": 0.1970387101173401, "global_step": 84110, "epoch": 945, "lr": 7.828777443485212e-05}
{"train_loss": 0.18239346146583557, "global_step": 84111, "epoch": 945, "lr": 7.828729639776713e-05}
{"train_loss": 0.15315505862236023, "global_step": 84112, "epoch": 945, "lr": 7.828681835687927e-05}
{"train_loss": 0.21910616755485535, "global_step": 84113, "epoch": 945, "lr": 7.828634031218857e-05}
{"train_loss": 0.21129807829856873, "global_step": 84114, "epoch": 945, "lr": 7.828586226369514e-05}
{"train_loss": 0.21347998082637787, "global_step": 84115, "epoch": 945, "lr": 7.828538421139901e-05}
{"train_loss": 0.2066686600446701, "global_step": 84116, "epoch": 945, "lr": 7.828490615530025e-05}
{"train_loss": 0.20611266791820526, "global_step": 84117, "epoch": 945, "lr": 7.828442809539894e-05}
{"train_loss": 0.2003173679113388, "global_step": 84118, "epoch": 945, "lr": 7.828395003169513e-05}
{"train_loss": 0.22438977658748627, "global_step": 84119, "epoch": 945, "lr": 7.828347196418888e-05}
{"train_loss": 0.17576994001865387, "global_step": 84120, "epoch": 945, "lr": 7.828299389288028e-05}
{"train_loss": 0.14707347750663757, "global_step": 84121, "epoch": 945, "lr": 7.828251581776935e-05}
{"train_loss": 0.17928075790405273, "global_step": 84122, "epoch": 945, "lr": 7.828203773885621e-05}
{"train_loss": 0.1193152591586113, "global_step": 84123, "epoch": 945, "lr": 7.828155965614088e-05}
{"train_loss": 0.20852810144424438, "global_step": 84124, "epoch": 945, "lr": 7.828108156962344e-05}
{"train_loss": 0.21979548037052155, "global_step": 84125, "epoch": 945, "lr": 7.828060347930395e-05}
{"train_loss": 0.10696924477815628, "global_step": 84126, "epoch": 945, "lr": 7.828012538518248e-05}
{"train_loss": 0.2020748406648636, "global_step": 84127, "epoch": 945, "lr": 7.827964728725908e-05}
{"train_loss": 0.2597656548023224, "global_step": 84128, "epoch": 945, "lr": 7.827916918553384e-05}
{"train_loss": 0.20442081987857819, "global_step": 84129, "epoch": 945, "lr": 7.82786910800068e-05}
{"train_loss": 0.24326685070991516, "global_step": 84130, "epoch": 945, "lr": 7.827821297067804e-05}
{"train_loss": 0.16682180762290955, "global_step": 84131, "epoch": 945, "lr": 7.827773485754764e-05}
{"train_loss": 0.18395814299583435, "global_step": 84132, "epoch": 945, "lr": 7.827725674061561e-05}
{"train_loss": 0.2297443300485611, "global_step": 84133, "epoch": 945, "lr": 7.827677861988206e-05}
{"train_loss": 0.1365077793598175, "global_step": 84134, "epoch": 945, "lr": 7.827630049534704e-05}
{"train_loss": 0.16487805545330048, "global_step": 84135, "epoch": 945, "lr": 7.827582236701064e-05}
{"train_loss": 0.2040150910615921, "global_step": 84136, "epoch": 945, "lr": 7.827534423487286e-05}
{"train_loss": 0.18103142082691193, "global_step": 84137, "epoch": 945, "lr": 7.827486609893383e-05}
{"train_loss": 0.16626089811325073, "global_step": 84138, "epoch": 945, "lr": 7.827438795919358e-05}
{"train_loss": 0.1405528336763382, "global_step": 84139, "epoch": 945, "lr": 7.82739098156522e-05}
{"train_loss": 0.17387263476848602, "global_step": 84140, "epoch": 945, "lr": 7.827343166830971e-05}
{"train_loss": 0.1374083161354065, "global_step": 84141, "epoch": 945, "lr": 7.827295351716621e-05}
{"train_loss": 0.22318971157073975, "global_step": 84142, "epoch": 945, "lr": 7.827247536222178e-05}
{"train_loss": 0.16131645441055298, "global_step": 84143, "epoch": 945, "lr": 7.827199720347645e-05}
{"train_loss": 0.19708780944347382, "global_step": 84144, "epoch": 945, "lr": 7.827151904093027e-05}
{"train_loss": 0.22126786410808563, "global_step": 84145, "epoch": 945, "lr": 7.827104087458335e-05}
{"train_loss": 0.16151711344718933, "global_step": 84146, "epoch": 945, "lr": 7.827056270443574e-05}
{"train_loss": 0.2635502219200134, "global_step": 84147, "epoch": 945, "lr": 7.827008453048748e-05}
{"train_loss": 0.2513425350189209, "global_step": 84148, "epoch": 945, "lr": 7.826960635273865e-05}
{"train_loss": 0.18325857818126678, "global_step": 84149, "epoch": 945, "lr": 7.826912817118933e-05}
{"train_loss": 0.2284444123506546, "global_step": 84150, "epoch": 945, "lr": 7.826864998583956e-05}
{"train_loss": 0.145313560962677, "global_step": 84151, "epoch": 945, "lr": 7.826817179668941e-05}
{"train_loss": 0.1806466281414032, "global_step": 84152, "epoch": 945, "lr": 7.826769360373895e-05}
{"train_loss": 0.16199186444282532, "global_step": 84153, "epoch": 945, "lr": 7.826721540698825e-05}
{"train_loss": 0.10358582437038422, "global_step": 84154, "epoch": 945, "lr": 7.826673720643737e-05}
{"train_loss": 0.20417746901512146, "global_step": 84155, "epoch": 945, "lr": 7.826625900208638e-05}
{"train_loss": 0.1311226338148117, "global_step": 84156, "epoch": 945, "lr": 7.826578079393532e-05}
{"train_loss": 0.24526748061180115, "global_step": 84157, "epoch": 945, "lr": 7.826530258198427e-05}
{"train_loss": 0.23340480029582977, "global_step": 84158, "epoch": 945, "lr": 7.82648243662333e-05}
{"train_loss": 0.1416206955909729, "global_step": 84159, "epoch": 945, "lr": 7.826434614668248e-05}
{"train_loss": 0.25162795186042786, "global_step": 84160, "epoch": 945, "lr": 7.826386792333184e-05}
{"train_loss": 0.2630743682384491, "global_step": 84161, "epoch": 945, "lr": 7.82633896961815e-05}
{"train_loss": 0.18893523514270782, "global_step": 84162, "epoch": 945, "lr": 7.826291146523146e-05}
{"train_loss": 0.2009807974100113, "global_step": 84163, "epoch": 945, "lr": 7.826243323048183e-05}
{"train_loss": 0.17328546941280365, "global_step": 84164, "epoch": 945, "lr": 7.826195499193266e-05}
{"train_loss": 0.15839852392673492, "global_step": 84165, "epoch": 945, "lr": 7.826147674958401e-05}
{"train_loss": 0.17707224190235138, "global_step": 84166, "epoch": 945, "lr": 7.826099850343595e-05}
{"train_loss": 0.2130133956670761, "global_step": 84167, "epoch": 945, "lr": 7.826052025348854e-05}
{"train_loss": 0.21348229050636292, "global_step": 84168, "epoch": 945, "lr": 7.826004199974186e-05}
{"train_loss": 0.24059239029884338, "global_step": 84169, "epoch": 945, "lr": 7.825956374219595e-05}
{"train_loss": 0.2726237177848816, "global_step": 84170, "epoch": 945, "lr": 7.82590854808509e-05}
{"train_loss": 0.19975970685482025, "global_step": 84171, "epoch": 945, "lr": 7.825860721570675e-05}
{"train_loss": 0.21803708374500275, "global_step": 84172, "epoch": 945, "lr": 7.825812894676357e-05}
{"train_loss": 0.19216854870319366, "global_step": 84173, "epoch": 945, "lr": 7.825765067402142e-05}
{"train_loss": 0.1736660748720169, "global_step": 84174, "epoch": 945, "lr": 7.82571723974804e-05}
{"train_loss": 0.23782505095005035, "global_step": 84175, "epoch": 945, "lr": 7.825669411714053e-05}
{"train_loss": 0.13964501023292542, "global_step": 84176, "epoch": 945, "lr": 7.82562158330019e-05}
{"train_loss": 0.15982234477996826, "global_step": 84177, "epoch": 945, "lr": 7.825573754506456e-05}
{"train_loss": 0.18900568783283234, "global_step": 84178, "epoch": 945, "lr": 7.825525925332858e-05}
{"train_loss": 0.1867535561323166, "global_step": 84179, "epoch": 945, "lr": 7.825478095779404e-05}
{"train_loss": 0.24816542863845825, "global_step": 84180, "epoch": 945, "lr": 7.825430265846097e-05}
{"train_loss": 0.18350166082382202, "global_step": 84181, "epoch": 945, "lr": 7.825382435532947e-05}
{"train_loss": 0.18892869353294373, "global_step": 84182, "epoch": 945, "lr": 7.825334604839958e-05}
{"train_loss": 0.1700952798128128, "global_step": 84183, "epoch": 945, "lr": 7.825286773767136e-05}
{"train_loss": 0.14800511300563812, "global_step": 84184, "epoch": 945, "lr": 7.82523894231449e-05}
{"train_loss": 0.2409248948097229, "global_step": 84185, "epoch": 945, "lr": 7.825191110482024e-05}
{"train_loss": 0.19785790145397186, "global_step": 84186, "epoch": 945, "lr": 7.825143278269746e-05}
{"train_loss": 0.12733547389507294, "global_step": 84187, "epoch": 945, "lr": 7.825095445677661e-05}
{"train_loss": 0.18981210887432098, "global_step": 84188, "epoch": 945, "lr": 7.825047612705779e-05}
{"train_loss": 0.240199476480484, "global_step": 84189, "epoch": 945, "lr": 7.824999779354103e-05}
{"train_loss": 0.1338777244091034, "global_step": 84190, "epoch": 945, "lr": 7.824951945622638e-05}
{"train_loss": 0.184806227684021, "global_step": 84191, "epoch": 945, "lr": 7.824904111511395e-05}
{"train_loss": 0.2027328610420227, "global_step": 84192, "epoch": 945, "lr": 7.824856277020377e-05}
{"train_loss": 0.19248262382625195, "global_step": 84193, "epoch": 945, "lr": 7.824808442149592e-05, "val_loss": 4.203746318817139, "train_action_mse_error": 40.211761474609375}
{"train_loss": 0.17293471097946167, "global_step": 84194, "epoch": 946, "lr": 7.824760606899046e-05}
{"train_loss": 0.16251932084560394, "global_step": 84195, "epoch": 946, "lr": 7.824712771268744e-05}
{"train_loss": 0.16839095950126648, "global_step": 84196, "epoch": 946, "lr": 7.824664935258694e-05}
{"train_loss": 0.2424190193414688, "global_step": 84197, "epoch": 946, "lr": 7.824617098868904e-05}
{"train_loss": 0.12507067620754242, "global_step": 84198, "epoch": 946, "lr": 7.824569262099377e-05}
{"train_loss": 0.16230948269367218, "global_step": 84199, "epoch": 946, "lr": 7.824521424950122e-05}
{"train_loss": 0.2644590735435486, "global_step": 84200, "epoch": 946, "lr": 7.824473587421145e-05}
{"train_loss": 0.1624506413936615, "global_step": 84201, "epoch": 946, "lr": 7.824425749512451e-05}
{"train_loss": 0.12510709464550018, "global_step": 84202, "epoch": 946, "lr": 7.824377911224048e-05}
{"train_loss": 0.21013350784778595, "global_step": 84203, "epoch": 946, "lr": 7.824330072555941e-05}
{"train_loss": 0.1876005083322525, "global_step": 84204, "epoch": 946, "lr": 7.824282233508138e-05}
{"train_loss": 0.2272670418024063, "global_step": 84205, "epoch": 946, "lr": 7.824234394080643e-05}
{"train_loss": 0.20398922264575958, "global_step": 84206, "epoch": 946, "lr": 7.824186554273466e-05}
{"train_loss": 0.1872253715991974, "global_step": 84207, "epoch": 946, "lr": 7.82413871408661e-05}
{"train_loss": 0.2011995166540146, "global_step": 84208, "epoch": 946, "lr": 7.824090873520085e-05}
{"train_loss": 0.13613580167293549, "global_step": 84209, "epoch": 946, "lr": 7.824043032573896e-05}
{"train_loss": 0.20625054836273193, "global_step": 84210, "epoch": 946, "lr": 7.823995191248046e-05}
{"train_loss": 0.11794768273830414, "global_step": 84211, "epoch": 946, "lr": 7.823947349542545e-05}
{"train_loss": 0.18780378997325897, "global_step": 84212, "epoch": 946, "lr": 7.8238995074574e-05}
{"train_loss": 0.18370307981967926, "global_step": 84213, "epoch": 946, "lr": 7.823851664992615e-05}
{"train_loss": 0.26789939403533936, "global_step": 84214, "epoch": 946, "lr": 7.823803822148197e-05}
{"train_loss": 0.21269018948078156, "global_step": 84215, "epoch": 946, "lr": 7.823755978924155e-05}
{"train_loss": 0.12630771100521088, "global_step": 84216, "epoch": 946, "lr": 7.823708135320492e-05}
{"train_loss": 0.14281626045703888, "global_step": 84217, "epoch": 946, "lr": 7.823660291337216e-05}
{"train_loss": 0.279523640871048, "global_step": 84218, "epoch": 946, "lr": 7.823612446974334e-05}
{"train_loss": 0.15079298615455627, "global_step": 84219, "epoch": 946, "lr": 7.823564602231851e-05}
{"train_loss": 0.1569303423166275, "global_step": 84220, "epoch": 946, "lr": 7.823516757109774e-05}
{"train_loss": 0.15199941396713257, "global_step": 84221, "epoch": 946, "lr": 7.823468911608109e-05}
{"train_loss": 0.16664913296699524, "global_step": 84222, "epoch": 946, "lr": 7.823421065726863e-05}
{"train_loss": 0.2994416356086731, "global_step": 84223, "epoch": 946, "lr": 7.823373219466045e-05}
{"train_loss": 0.2792961895465851, "global_step": 84224, "epoch": 946, "lr": 7.823325372825657e-05}
{"train_loss": 0.21543973684310913, "global_step": 84225, "epoch": 946, "lr": 7.823277525805708e-05}
{"train_loss": 0.18959356844425201, "global_step": 84226, "epoch": 946, "lr": 7.823229678406203e-05}
{"train_loss": 0.18392419815063477, "global_step": 84227, "epoch": 946, "lr": 7.82318183062715e-05}
{"train_loss": 0.19185972213745117, "global_step": 84228, "epoch": 946, "lr": 7.823133982468554e-05}
{"train_loss": 0.3326314389705658, "global_step": 84229, "epoch": 946, "lr": 7.823086133930422e-05}
{"train_loss": 0.19364219903945923, "global_step": 84230, "epoch": 946, "lr": 7.82303828501276e-05}
{"train_loss": 0.1374252736568451, "global_step": 84231, "epoch": 946, "lr": 7.822990435715576e-05}
{"train_loss": 0.16551919281482697, "global_step": 84232, "epoch": 946, "lr": 7.822942586038875e-05}
{"train_loss": 0.19731780886650085, "global_step": 84233, "epoch": 946, "lr": 7.822894735982664e-05}
{"train_loss": 0.25926339626312256, "global_step": 84234, "epoch": 946, "lr": 7.822846885546949e-05}
{"train_loss": 0.13932493329048157, "global_step": 84235, "epoch": 946, "lr": 7.822799034731736e-05}
{"train_loss": 0.1862986534833908, "global_step": 84236, "epoch": 946, "lr": 7.822751183537032e-05}
{"train_loss": 0.1512255221605301, "global_step": 84237, "epoch": 946, "lr": 7.822703331962844e-05}
{"train_loss": 0.12924611568450928, "global_step": 84238, "epoch": 946, "lr": 7.822655480009179e-05}
{"train_loss": 0.13190872967243195, "global_step": 84239, "epoch": 946, "lr": 7.822607627676041e-05}
{"train_loss": 0.11433622241020203, "global_step": 84240, "epoch": 946, "lr": 7.822559774963438e-05}
{"train_loss": 0.19871900975704193, "global_step": 84241, "epoch": 946, "lr": 7.822511921871376e-05}
{"train_loss": 0.1720951944589615, "global_step": 84242, "epoch": 946, "lr": 7.822464068399861e-05}
{"train_loss": 0.21680064499378204, "global_step": 84243, "epoch": 946, "lr": 7.822416214548903e-05}
{"train_loss": 0.2105654776096344, "global_step": 84244, "epoch": 946, "lr": 7.822368360318504e-05}
{"train_loss": 0.1987287849187851, "global_step": 84245, "epoch": 946, "lr": 7.822320505708671e-05}
{"train_loss": 0.19323194026947021, "global_step": 84246, "epoch": 946, "lr": 7.822272650719414e-05}
{"train_loss": 0.17803342640399933, "global_step": 84247, "epoch": 946, "lr": 7.822224795350735e-05}
{"train_loss": 0.1518380492925644, "global_step": 84248, "epoch": 946, "lr": 7.822176939602642e-05}
{"train_loss": 0.18068912625312805, "global_step": 84249, "epoch": 946, "lr": 7.82212908347514e-05}
{"train_loss": 0.20921571552753448, "global_step": 84250, "epoch": 946, "lr": 7.82208122696824e-05}
{"train_loss": 0.2622283101081848, "global_step": 84251, "epoch": 946, "lr": 7.822033370081945e-05}
{"train_loss": 0.15409153699874878, "global_step": 84252, "epoch": 946, "lr": 7.821985512816261e-05}
{"train_loss": 0.16626225411891937, "global_step": 84253, "epoch": 946, "lr": 7.821937655171196e-05}
{"train_loss": 0.23297624289989471, "global_step": 84254, "epoch": 946, "lr": 7.821889797146756e-05}
{"train_loss": 0.15945221483707428, "global_step": 84255, "epoch": 946, "lr": 7.821841938742949e-05}
{"train_loss": 0.25763386487960815, "global_step": 84256, "epoch": 946, "lr": 7.821794079959777e-05}
{"train_loss": 0.16917595267295837, "global_step": 84257, "epoch": 946, "lr": 7.82174622079725e-05}
{"train_loss": 0.2240006923675537, "global_step": 84258, "epoch": 946, "lr": 7.821698361255374e-05}
{"train_loss": 0.19360235333442688, "global_step": 84259, "epoch": 946, "lr": 7.821650501334154e-05}
{"train_loss": 0.2139737755060196, "global_step": 84260, "epoch": 946, "lr": 7.821602641033598e-05}
{"train_loss": 0.2909353971481323, "global_step": 84261, "epoch": 946, "lr": 7.821554780353712e-05}
{"train_loss": 0.2712484300136566, "global_step": 84262, "epoch": 946, "lr": 7.821506919294503e-05}
{"train_loss": 0.17519135773181915, "global_step": 84263, "epoch": 946, "lr": 7.821459057855978e-05}
{"train_loss": 0.15406298637390137, "global_step": 84264, "epoch": 946, "lr": 7.82141119603814e-05}
{"train_loss": 0.21446149051189423, "global_step": 84265, "epoch": 946, "lr": 7.821363333840999e-05}
{"train_loss": 0.19754715263843536, "global_step": 84266, "epoch": 946, "lr": 7.821315471264559e-05}
{"train_loss": 0.20566314458847046, "global_step": 84267, "epoch": 946, "lr": 7.821267608308829e-05}
{"train_loss": 0.19215764105319977, "global_step": 84268, "epoch": 946, "lr": 7.821219744973811e-05}
{"train_loss": 0.2511126399040222, "global_step": 84269, "epoch": 946, "lr": 7.821171881259517e-05}
{"train_loss": 0.137980654835701, "global_step": 84270, "epoch": 946, "lr": 7.821124017165951e-05}
{"train_loss": 0.14865684509277344, "global_step": 84271, "epoch": 946, "lr": 7.821076152693118e-05}
{"train_loss": 0.2768701910972595, "global_step": 84272, "epoch": 946, "lr": 7.821028287841026e-05}
{"train_loss": 0.22883129119873047, "global_step": 84273, "epoch": 946, "lr": 7.82098042260968e-05}
{"train_loss": 0.22553683817386627, "global_step": 84274, "epoch": 946, "lr": 7.82093255699909e-05}
{"train_loss": 0.11809200793504715, "global_step": 84275, "epoch": 946, "lr": 7.820884691009259e-05}
{"train_loss": 0.1404510885477066, "global_step": 84276, "epoch": 946, "lr": 7.820836824640193e-05}
{"train_loss": 0.17511031031608582, "global_step": 84277, "epoch": 946, "lr": 7.820788957891901e-05}
{"train_loss": 0.17037419974803925, "global_step": 84278, "epoch": 946, "lr": 7.820741090764388e-05}
{"train_loss": 0.16978207230567932, "global_step": 84279, "epoch": 946, "lr": 7.820693223257662e-05}
{"train_loss": 0.1502050757408142, "global_step": 84280, "epoch": 946, "lr": 7.820645355371728e-05}
{"train_loss": 0.1794232875108719, "global_step": 84281, "epoch": 946, "lr": 7.820597487106591e-05}
{"train_loss": 0.19201473967078025, "global_step": 84282, "epoch": 946, "lr": 7.820549618462259e-05, "val_loss": 4.190672397613525}
{"train_loss": 0.15605391561985016, "global_step": 84283, "epoch": 947, "lr": 7.82050174943874e-05}
{"train_loss": 0.15632738173007965, "global_step": 84284, "epoch": 947, "lr": 7.820453880036038e-05}
{"train_loss": 0.17494139075279236, "global_step": 84285, "epoch": 947, "lr": 7.820406010254162e-05}
{"train_loss": 0.1358117312192917, "global_step": 84286, "epoch": 947, "lr": 7.820358140093115e-05}
{"train_loss": 0.19075345993041992, "global_step": 84287, "epoch": 947, "lr": 7.820310269552907e-05}
{"train_loss": 0.1684068888425827, "global_step": 84288, "epoch": 947, "lr": 7.82026239863354e-05}
{"train_loss": 0.14498451352119446, "global_step": 84289, "epoch": 947, "lr": 7.820214527335024e-05}
{"train_loss": 0.18716876208782196, "global_step": 84290, "epoch": 947, "lr": 7.820166655657365e-05}
{"train_loss": 0.1579112857580185, "global_step": 84291, "epoch": 947, "lr": 7.82011878360057e-05}
{"train_loss": 0.2146681696176529, "global_step": 84292, "epoch": 947, "lr": 7.820070911164643e-05}
{"train_loss": 0.19966909289360046, "global_step": 84293, "epoch": 947, "lr": 7.820023038349592e-05}
{"train_loss": 0.16016049683094025, "global_step": 84294, "epoch": 947, "lr": 7.819975165155423e-05}
{"train_loss": 0.18155743181705475, "global_step": 84295, "epoch": 947, "lr": 7.819927291582143e-05}
{"train_loss": 0.1779000163078308, "global_step": 84296, "epoch": 947, "lr": 7.819879417629759e-05}
{"train_loss": 0.26036569476127625, "global_step": 84297, "epoch": 947, "lr": 7.819831543298275e-05}
{"train_loss": 0.23078207671642303, "global_step": 84298, "epoch": 947, "lr": 7.8197836685877e-05}
{"train_loss": 0.18027718365192413, "global_step": 84299, "epoch": 947, "lr": 7.81973579349804e-05}
{"train_loss": 0.22030292451381683, "global_step": 84300, "epoch": 947, "lr": 7.819687918029299e-05}
{"train_loss": 0.16150115430355072, "global_step": 84301, "epoch": 947, "lr": 7.819640042181488e-05}
{"train_loss": 0.13936804234981537, "global_step": 84302, "epoch": 947, "lr": 7.819592165954609e-05}
{"train_loss": 0.1586914211511612, "global_step": 84303, "epoch": 947, "lr": 7.81954428934867e-05}
{"train_loss": 0.21240514516830444, "global_step": 84304, "epoch": 947, "lr": 7.819496412363678e-05}
{"train_loss": 0.175911545753479, "global_step": 84305, "epoch": 947, "lr": 7.81944853499964e-05}
{"train_loss": 0.22862420976161957, "global_step": 84306, "epoch": 947, "lr": 7.819400657256562e-05}
{"train_loss": 0.16229724884033203, "global_step": 84307, "epoch": 947, "lr": 7.819352779134448e-05}
{"train_loss": 0.2252146601676941, "global_step": 84308, "epoch": 947, "lr": 7.819304900633306e-05}
{"train_loss": 0.18312475085258484, "global_step": 84309, "epoch": 947, "lr": 7.819257021753145e-05}
{"train_loss": 0.16861644387245178, "global_step": 84310, "epoch": 947, "lr": 7.819209142493968e-05}
{"train_loss": 0.12898537516593933, "global_step": 84311, "epoch": 947, "lr": 7.819161262855783e-05}
{"train_loss": 0.2433643341064453, "global_step": 84312, "epoch": 947, "lr": 7.819113382838595e-05}
{"train_loss": 0.2635689377784729, "global_step": 84313, "epoch": 947, "lr": 7.819065502442413e-05}
{"train_loss": 0.12514449656009674, "global_step": 84314, "epoch": 947, "lr": 7.819017621667242e-05}
{"train_loss": 0.15270860493183136, "global_step": 84315, "epoch": 947, "lr": 7.818969740513087e-05}
{"train_loss": 0.13657909631729126, "global_step": 84316, "epoch": 947, "lr": 7.818921858979957e-05}
{"train_loss": 0.24872273206710815, "global_step": 84317, "epoch": 947, "lr": 7.818873977067857e-05}
{"train_loss": 0.18108835816383362, "global_step": 84318, "epoch": 947, "lr": 7.818826094776794e-05}
{"train_loss": 0.11337767541408539, "global_step": 84319, "epoch": 947, "lr": 7.818778212106774e-05}
{"train_loss": 0.24101793766021729, "global_step": 84320, "epoch": 947, "lr": 7.818730329057804e-05}
{"train_loss": 0.19474972784519196, "global_step": 84321, "epoch": 947, "lr": 7.81868244562989e-05}
{"train_loss": 0.1875554919242859, "global_step": 84322, "epoch": 947, "lr": 7.818634561823038e-05}
{"train_loss": 0.15681979060173035, "global_step": 84323, "epoch": 947, "lr": 7.818586677637256e-05}
{"train_loss": 0.17126820981502533, "global_step": 84324, "epoch": 947, "lr": 7.818538793072549e-05}
{"train_loss": 0.19419535994529724, "global_step": 84325, "epoch": 947, "lr": 7.818490908128923e-05}
{"train_loss": 0.11852766573429108, "global_step": 84326, "epoch": 947, "lr": 7.818443022806385e-05}
{"train_loss": 0.2425253689289093, "global_step": 84327, "epoch": 947, "lr": 7.818395137104944e-05}
{"train_loss": 0.25634416937828064, "global_step": 84328, "epoch": 947, "lr": 7.818347251024601e-05}
{"train_loss": 0.14071623980998993, "global_step": 84329, "epoch": 947, "lr": 7.818299364565366e-05}
{"train_loss": 0.19533008337020874, "global_step": 84330, "epoch": 947, "lr": 7.818251477727248e-05}
{"train_loss": 0.10114292800426483, "global_step": 84331, "epoch": 947, "lr": 7.818203590510246e-05}
{"train_loss": 0.27075836062431335, "global_step": 84332, "epoch": 947, "lr": 7.818155702914375e-05}
{"train_loss": 0.11889581382274628, "global_step": 84333, "epoch": 947, "lr": 7.818107814939635e-05}
{"train_loss": 0.19812729954719543, "global_step": 84334, "epoch": 947, "lr": 7.818059926586036e-05}
{"train_loss": 0.16462138295173645, "global_step": 84335, "epoch": 947, "lr": 7.818012037853582e-05}
{"train_loss": 0.13236072659492493, "global_step": 84336, "epoch": 947, "lr": 7.817964148742283e-05}
{"train_loss": 0.1730584353208542, "global_step": 84337, "epoch": 947, "lr": 7.817916259252141e-05}
{"train_loss": 0.1848195642232895, "global_step": 84338, "epoch": 947, "lr": 7.817868369383165e-05}
{"train_loss": 0.23203526437282562, "global_step": 84339, "epoch": 947, "lr": 7.81782047913536e-05}
{"train_loss": 0.16861972212791443, "global_step": 84340, "epoch": 947, "lr": 7.817772588508734e-05}
{"train_loss": 0.17468467354774475, "global_step": 84341, "epoch": 947, "lr": 7.817724697503292e-05}
{"train_loss": 0.12406859546899796, "global_step": 84342, "epoch": 947, "lr": 7.817676806119043e-05}
{"train_loss": 0.1690986007452011, "global_step": 84343, "epoch": 947, "lr": 7.817628914355992e-05}
{"train_loss": 0.15357381105422974, "global_step": 84344, "epoch": 947, "lr": 7.817581022214144e-05}
{"train_loss": 0.2470797300338745, "global_step": 84345, "epoch": 947, "lr": 7.817533129693507e-05}
{"train_loss": 0.13559816777706146, "global_step": 84346, "epoch": 947, "lr": 7.817485236794086e-05}
{"train_loss": 0.1956329196691513, "global_step": 84347, "epoch": 947, "lr": 7.81743734351589e-05}
{"train_loss": 0.12187446653842926, "global_step": 84348, "epoch": 947, "lr": 7.817389449858921e-05}
{"train_loss": 0.23031072318553925, "global_step": 84349, "epoch": 947, "lr": 7.817341555823191e-05}
{"train_loss": 0.1771996170282364, "global_step": 84350, "epoch": 947, "lr": 7.817293661408704e-05}
{"train_loss": 0.09825049340724945, "global_step": 84351, "epoch": 947, "lr": 7.817245766615464e-05}
{"train_loss": 0.23642054200172424, "global_step": 84352, "epoch": 947, "lr": 7.817197871443481e-05}
{"train_loss": 0.1451067477464676, "global_step": 84353, "epoch": 947, "lr": 7.817149975892761e-05}
{"train_loss": 0.17398099601268768, "global_step": 84354, "epoch": 947, "lr": 7.817102079963307e-05}
{"train_loss": 0.16303548216819763, "global_step": 84355, "epoch": 947, "lr": 7.81705418365513e-05}
{"train_loss": 0.1637406200170517, "global_step": 84356, "epoch": 947, "lr": 7.817006286968234e-05}
{"train_loss": 0.22528047859668732, "global_step": 84357, "epoch": 947, "lr": 7.816958389902626e-05}
{"train_loss": 0.14708000421524048, "global_step": 84358, "epoch": 947, "lr": 7.816910492458312e-05}
{"train_loss": 0.25355905294418335, "global_step": 84359, "epoch": 947, "lr": 7.816862594635299e-05}
{"train_loss": 0.2383124828338623, "global_step": 84360, "epoch": 947, "lr": 7.816814696433592e-05}
{"train_loss": 0.2191476821899414, "global_step": 84361, "epoch": 947, "lr": 7.8167667978532e-05}
{"train_loss": 0.12344209104776382, "global_step": 84362, "epoch": 947, "lr": 7.816718898894129e-05}
{"train_loss": 0.14044764637947083, "global_step": 84363, "epoch": 947, "lr": 7.816670999556382e-05}
{"train_loss": 0.1770845204591751, "global_step": 84364, "epoch": 947, "lr": 7.816623099839969e-05}
{"train_loss": 0.1901540607213974, "global_step": 84365, "epoch": 947, "lr": 7.816575199744895e-05}
{"train_loss": 0.16843226552009583, "global_step": 84366, "epoch": 947, "lr": 7.816527299271166e-05}
{"train_loss": 0.22954784333705902, "global_step": 84367, "epoch": 947, "lr": 7.81647939841879e-05}
{"train_loss": 0.24691987037658691, "global_step": 84368, "epoch": 947, "lr": 7.816431497187773e-05}
{"train_loss": 0.24481697380542755, "global_step": 84369, "epoch": 947, "lr": 7.816383595578123e-05}
{"train_loss": 0.2711091935634613, "global_step": 84370, "epoch": 947, "lr": 7.816335693589843e-05}
{"train_loss": 0.18361180518450362, "global_step": 84371, "epoch": 947, "lr": 7.816287791222939e-05, "val_loss": 4.21854829788208}
{"train_loss": 0.18995888531208038, "global_step": 84372, "epoch": 948, "lr": 7.816239888477421e-05}
{"train_loss": 0.16842348873615265, "global_step": 84373, "epoch": 948, "lr": 7.816191985353294e-05}
{"train_loss": 0.11155898869037628, "global_step": 84374, "epoch": 948, "lr": 7.816144081850565e-05}
{"train_loss": 0.1562299132347107, "global_step": 84375, "epoch": 948, "lr": 7.816096177969238e-05}
{"train_loss": 0.1736307144165039, "global_step": 84376, "epoch": 948, "lr": 7.816048273709322e-05}
{"train_loss": 0.17636947333812714, "global_step": 84377, "epoch": 948, "lr": 7.816000369070824e-05}
{"train_loss": 0.175149604678154, "global_step": 84378, "epoch": 948, "lr": 7.815952464053747e-05}
{"train_loss": 0.1921534389257431, "global_step": 84379, "epoch": 948, "lr": 7.8159045586581e-05}
{"train_loss": 0.21999554336071014, "global_step": 84380, "epoch": 948, "lr": 7.81585665288389e-05}
{"train_loss": 0.23377196490764618, "global_step": 84381, "epoch": 948, "lr": 7.815808746731122e-05}
{"train_loss": 0.16851258277893066, "global_step": 84382, "epoch": 948, "lr": 7.815760840199803e-05}
{"train_loss": 0.18890847265720367, "global_step": 84383, "epoch": 948, "lr": 7.815712933289939e-05}
{"train_loss": 0.19309408962726593, "global_step": 84384, "epoch": 948, "lr": 7.815665026001535e-05}
{"train_loss": 0.2299354523420334, "global_step": 84385, "epoch": 948, "lr": 7.815617118334601e-05}
{"train_loss": 0.1869155764579773, "global_step": 84386, "epoch": 948, "lr": 7.815569210289142e-05}
{"train_loss": 0.22412478923797607, "global_step": 84387, "epoch": 948, "lr": 7.815521301865163e-05}
{"train_loss": 0.10797775536775589, "global_step": 84388, "epoch": 948, "lr": 7.815473393062672e-05}
{"train_loss": 0.16060450673103333, "global_step": 84389, "epoch": 948, "lr": 7.815425483881675e-05}
{"train_loss": 0.1837535798549652, "global_step": 84390, "epoch": 948, "lr": 7.815377574322178e-05}
{"train_loss": 0.24139511585235596, "global_step": 84391, "epoch": 948, "lr": 7.815329664384188e-05}
{"train_loss": 0.2149903029203415, "global_step": 84392, "epoch": 948, "lr": 7.815281754067712e-05}
{"train_loss": 0.21835489571094513, "global_step": 84393, "epoch": 948, "lr": 7.815233843372753e-05}
{"train_loss": 0.2297758311033249, "global_step": 84394, "epoch": 948, "lr": 7.815185932299322e-05}
{"train_loss": 0.1590002179145813, "global_step": 84395, "epoch": 948, "lr": 7.815138020847425e-05}
{"train_loss": 0.20431658625602722, "global_step": 84396, "epoch": 948, "lr": 7.815090109017065e-05}
{"train_loss": 0.1970142126083374, "global_step": 84397, "epoch": 948, "lr": 7.815042196808251e-05}
{"train_loss": 0.1574271023273468, "global_step": 84398, "epoch": 948, "lr": 7.814994284220989e-05}
{"train_loss": 0.22680306434631348, "global_step": 84399, "epoch": 948, "lr": 7.814946371255285e-05}
{"train_loss": 0.2037801593542099, "global_step": 84400, "epoch": 948, "lr": 7.814898457911146e-05}
{"train_loss": 0.21802572906017303, "global_step": 84401, "epoch": 948, "lr": 7.814850544188579e-05}
{"train_loss": 0.1813804656267166, "global_step": 84402, "epoch": 948, "lr": 7.814802630087588e-05}
{"train_loss": 0.2022019624710083, "global_step": 84403, "epoch": 948, "lr": 7.814754715608182e-05}
{"train_loss": 0.34011974930763245, "global_step": 84404, "epoch": 948, "lr": 7.814706800750366e-05}
{"train_loss": 0.20657649636268616, "global_step": 84405, "epoch": 948, "lr": 7.814658885514148e-05}
{"train_loss": 0.20869289338588715, "global_step": 84406, "epoch": 948, "lr": 7.814610969899532e-05}
{"train_loss": 0.1643025130033493, "global_step": 84407, "epoch": 948, "lr": 7.814563053906527e-05}
{"train_loss": 0.15188688039779663, "global_step": 84408, "epoch": 948, "lr": 7.814515137535139e-05}
{"train_loss": 0.1518002450466156, "global_step": 84409, "epoch": 948, "lr": 7.814467220785371e-05}
{"train_loss": 0.15591861307621002, "global_step": 84410, "epoch": 948, "lr": 7.814419303657234e-05}
{"train_loss": 0.21818262338638306, "global_step": 84411, "epoch": 948, "lr": 7.814371386150734e-05}
{"train_loss": 0.14605507254600525, "global_step": 84412, "epoch": 948, "lr": 7.814323468265874e-05}
{"train_loss": 0.2047150880098343, "global_step": 84413, "epoch": 948, "lr": 7.814275550002663e-05}
{"train_loss": 0.16848213970661163, "global_step": 84414, "epoch": 948, "lr": 7.814227631361107e-05}
{"train_loss": 0.18558886647224426, "global_step": 84415, "epoch": 948, "lr": 7.814179712341213e-05}
{"train_loss": 0.1443488746881485, "global_step": 84416, "epoch": 948, "lr": 7.814131792942986e-05}
{"train_loss": 0.188497856259346, "global_step": 84417, "epoch": 948, "lr": 7.814083873166433e-05}
{"train_loss": 0.21086442470550537, "global_step": 84418, "epoch": 948, "lr": 7.814035953011562e-05}
{"train_loss": 0.23903262615203857, "global_step": 84419, "epoch": 948, "lr": 7.813988032478378e-05}
{"train_loss": 0.25773268938064575, "global_step": 84420, "epoch": 948, "lr": 7.813940111566887e-05}
{"train_loss": 0.20220646262168884, "global_step": 84421, "epoch": 948, "lr": 7.813892190277095e-05}
{"train_loss": 0.26341167092323303, "global_step": 84422, "epoch": 948, "lr": 7.813844268609011e-05}
{"train_loss": 0.2028026133775711, "global_step": 84423, "epoch": 948, "lr": 7.81379634656264e-05}
{"train_loss": 0.2082337737083435, "global_step": 84424, "epoch": 948, "lr": 7.813748424137988e-05}
{"train_loss": 0.12346246093511581, "global_step": 84425, "epoch": 948, "lr": 7.813700501335062e-05}
{"train_loss": 0.22047793865203857, "global_step": 84426, "epoch": 948, "lr": 7.813652578153867e-05}
{"train_loss": 0.1606587916612625, "global_step": 84427, "epoch": 948, "lr": 7.813604654594413e-05}
{"train_loss": 0.1699482500553131, "global_step": 84428, "epoch": 948, "lr": 7.813556730656702e-05}
{"train_loss": 0.21426738798618317, "global_step": 84429, "epoch": 948, "lr": 7.813508806340743e-05}
{"train_loss": 0.22105716168880463, "global_step": 84430, "epoch": 948, "lr": 7.813460881646544e-05}
{"train_loss": 0.20206989347934723, "global_step": 84431, "epoch": 948, "lr": 7.813412956574108e-05}
{"train_loss": 0.2477157562971115, "global_step": 84432, "epoch": 948, "lr": 7.813365031123443e-05}
{"train_loss": 0.143914133310318, "global_step": 84433, "epoch": 948, "lr": 7.813317105294555e-05}
{"train_loss": 0.1977401077747345, "global_step": 84434, "epoch": 948, "lr": 7.813269179087451e-05}
{"train_loss": 0.21703793108463287, "global_step": 84435, "epoch": 948, "lr": 7.813221252502136e-05}
{"train_loss": 0.20153354108333588, "global_step": 84436, "epoch": 948, "lr": 7.81317332553862e-05}
{"train_loss": 0.23178833723068237, "global_step": 84437, "epoch": 948, "lr": 7.813125398196906e-05}
{"train_loss": 0.19882817566394806, "global_step": 84438, "epoch": 948, "lr": 7.813077470477002e-05}
{"train_loss": 0.2585597038269043, "global_step": 84439, "epoch": 948, "lr": 7.813029542378914e-05}
{"train_loss": 0.1570732593536377, "global_step": 84440, "epoch": 948, "lr": 7.812981613902647e-05}
{"train_loss": 0.17360429465770721, "global_step": 84441, "epoch": 948, "lr": 7.81293368504821e-05}
{"train_loss": 0.3180091679096222, "global_step": 84442, "epoch": 948, "lr": 7.812885755815609e-05}
{"train_loss": 0.21234452724456787, "global_step": 84443, "epoch": 948, "lr": 7.81283782620485e-05}
{"train_loss": 0.208139568567276, "global_step": 84444, "epoch": 948, "lr": 7.812789896215937e-05}
{"train_loss": 0.19825337827205658, "global_step": 84445, "epoch": 948, "lr": 7.81274196584888e-05}
{"train_loss": 0.17964230477809906, "global_step": 84446, "epoch": 948, "lr": 7.812694035103686e-05}
{"train_loss": 0.27705851197242737, "global_step": 84447, "epoch": 948, "lr": 7.812646103980357e-05}
{"train_loss": 0.28378039598464966, "global_step": 84448, "epoch": 948, "lr": 7.812598172478904e-05}
{"train_loss": 0.1615772396326065, "global_step": 84449, "epoch": 948, "lr": 7.81255024059933e-05}
{"train_loss": 0.19116374850273132, "global_step": 84450, "epoch": 948, "lr": 7.812502308341643e-05}
{"train_loss": 0.18685315549373627, "global_step": 84451, "epoch": 948, "lr": 7.81245437570585e-05}
{"train_loss": 0.1136186346411705, "global_step": 84452, "epoch": 948, "lr": 7.812406442691957e-05}
{"train_loss": 0.1837715357542038, "global_step": 84453, "epoch": 948, "lr": 7.812358509299968e-05}
{"train_loss": 0.22325651347637177, "global_step": 84454, "epoch": 948, "lr": 7.812310575529894e-05}
{"train_loss": 0.2732902467250824, "global_step": 84455, "epoch": 948, "lr": 7.812262641381738e-05}
{"train_loss": 0.23124384880065918, "global_step": 84456, "epoch": 948, "lr": 7.812214706855508e-05}
{"train_loss": 0.10739755630493164, "global_step": 84457, "epoch": 948, "lr": 7.812166771951211e-05}
{"train_loss": 0.1733628809452057, "global_step": 84458, "epoch": 948, "lr": 7.81211883666885e-05}
{"train_loss": 0.1287163347005844, "global_step": 84459, "epoch": 948, "lr": 7.812070901008436e-05}
{"train_loss": 0.19739359324232916, "global_step": 84460, "epoch": 948, "lr": 7.812022964969972e-05, "val_loss": 3.94360613822937}
{"train_loss": 0.22573208808898926, "global_step": 84461, "epoch": 949, "lr": 7.811975028553467e-05}
{"train_loss": 0.14492474496364594, "global_step": 84462, "epoch": 949, "lr": 7.811927091758925e-05}
{"train_loss": 0.17821148037910461, "global_step": 84463, "epoch": 949, "lr": 7.811879154586355e-05}
{"train_loss": 0.21626149117946625, "global_step": 84464, "epoch": 949, "lr": 7.811831217035761e-05}
{"train_loss": 0.23692955076694489, "global_step": 84465, "epoch": 949, "lr": 7.811783279107152e-05}
{"train_loss": 0.1620086133480072, "global_step": 84466, "epoch": 949, "lr": 7.811735340800532e-05}
{"train_loss": 0.15959309041500092, "global_step": 84467, "epoch": 949, "lr": 7.811687402115909e-05}
{"train_loss": 0.153342604637146, "global_step": 84468, "epoch": 949, "lr": 7.811639463053289e-05}
{"train_loss": 0.15830020606517792, "global_step": 84469, "epoch": 949, "lr": 7.811591523612676e-05}
{"train_loss": 0.20957008004188538, "global_step": 84470, "epoch": 949, "lr": 7.81154358379408e-05}
{"train_loss": 0.18211613595485687, "global_step": 84471, "epoch": 949, "lr": 7.811495643597509e-05}
{"train_loss": 0.17083945870399475, "global_step": 84472, "epoch": 949, "lr": 7.811447703022963e-05}
{"train_loss": 0.19537651538848877, "global_step": 84473, "epoch": 949, "lr": 7.811399762070454e-05}
{"train_loss": 0.13462959229946136, "global_step": 84474, "epoch": 949, "lr": 7.811351820739986e-05}
{"train_loss": 0.1777319759130478, "global_step": 84475, "epoch": 949, "lr": 7.811303879031566e-05}
{"train_loss": 0.15130625665187836, "global_step": 84476, "epoch": 949, "lr": 7.8112559369452e-05}
{"train_loss": 0.2049494832754135, "global_step": 84477, "epoch": 949, "lr": 7.811207994480895e-05}
{"train_loss": 0.23148056864738464, "global_step": 84478, "epoch": 949, "lr": 7.811160051638659e-05}
{"train_loss": 0.19088411331176758, "global_step": 84479, "epoch": 949, "lr": 7.811112108418494e-05}
{"train_loss": 0.1519538015127182, "global_step": 84480, "epoch": 949, "lr": 7.811064164820412e-05}
{"train_loss": 0.14064837992191315, "global_step": 84481, "epoch": 949, "lr": 7.811016220844416e-05}
{"train_loss": 0.1324644237756729, "global_step": 84482, "epoch": 949, "lr": 7.810968276490513e-05}
{"train_loss": 0.15731623768806458, "global_step": 84483, "epoch": 949, "lr": 7.810920331758708e-05}
{"train_loss": 0.1992151439189911, "global_step": 84484, "epoch": 949, "lr": 7.810872386649011e-05}
{"train_loss": 0.19085092842578888, "global_step": 84485, "epoch": 949, "lr": 7.810824441161426e-05}
{"train_loss": 0.23502002656459808, "global_step": 84486, "epoch": 949, "lr": 7.81077649529596e-05}
{"train_loss": 0.11926598101854324, "global_step": 84487, "epoch": 949, "lr": 7.810728549052619e-05}
{"train_loss": 0.1884121596813202, "global_step": 84488, "epoch": 949, "lr": 7.81068060243141e-05}
{"train_loss": 0.2021842896938324, "global_step": 84489, "epoch": 949, "lr": 7.810632655432339e-05}
{"train_loss": 0.29443690180778503, "global_step": 84490, "epoch": 949, "lr": 7.810584708055412e-05}
{"train_loss": 0.23856951296329498, "global_step": 84491, "epoch": 949, "lr": 7.810536760300637e-05}
{"train_loss": 0.18407480418682098, "global_step": 84492, "epoch": 949, "lr": 7.81048881216802e-05}
{"train_loss": 0.21845389902591705, "global_step": 84493, "epoch": 949, "lr": 7.810440863657566e-05}
{"train_loss": 0.24017861485481262, "global_step": 84494, "epoch": 949, "lr": 7.810392914769283e-05}
{"train_loss": 0.16756631433963776, "global_step": 84495, "epoch": 949, "lr": 7.810344965503178e-05}
{"train_loss": 0.2522542178630829, "global_step": 84496, "epoch": 949, "lr": 7.810297015859255e-05}
{"train_loss": 0.25473469495773315, "global_step": 84497, "epoch": 949, "lr": 7.810249065837522e-05}
{"train_loss": 0.1920452117919922, "global_step": 84498, "epoch": 949, "lr": 7.810201115437985e-05}
{"train_loss": 0.15534070134162903, "global_step": 84499, "epoch": 949, "lr": 7.810153164660652e-05}
{"train_loss": 0.24021033942699432, "global_step": 84500, "epoch": 949, "lr": 7.810105213505527e-05}
{"train_loss": 0.217374786734581, "global_step": 84501, "epoch": 949, "lr": 7.810057261972619e-05}
{"train_loss": 0.2714385390281677, "global_step": 84502, "epoch": 949, "lr": 7.810009310061931e-05}
{"train_loss": 0.2631152272224426, "global_step": 84503, "epoch": 949, "lr": 7.809961357773472e-05}
{"train_loss": 0.18133936822414398, "global_step": 84504, "epoch": 949, "lr": 7.80991340510725e-05}
{"train_loss": 0.1418153941631317, "global_step": 84505, "epoch": 949, "lr": 7.809865452063268e-05}
{"train_loss": 0.1740296185016632, "global_step": 84506, "epoch": 949, "lr": 7.809817498641531e-05}
{"train_loss": 0.2627283036708832, "global_step": 84507, "epoch": 949, "lr": 7.809769544842053e-05}
{"train_loss": 0.12234964966773987, "global_step": 84508, "epoch": 949, "lr": 7.809721590664833e-05}
{"train_loss": 0.21096579730510712, "global_step": 84509, "epoch": 949, "lr": 7.809673636109881e-05}
{"train_loss": 0.2519562244415283, "global_step": 84510, "epoch": 949, "lr": 7.809625681177202e-05}
{"train_loss": 0.2615852355957031, "global_step": 84511, "epoch": 949, "lr": 7.809577725866804e-05}
{"train_loss": 0.19013245403766632, "global_step": 84512, "epoch": 949, "lr": 7.809529770178692e-05}
{"train_loss": 0.2598305940628052, "global_step": 84513, "epoch": 949, "lr": 7.809481814112873e-05}
{"train_loss": 0.13599085807800293, "global_step": 84514, "epoch": 949, "lr": 7.809433857669352e-05}
{"train_loss": 0.19981993734836578, "global_step": 84515, "epoch": 949, "lr": 7.809385900848138e-05}
{"train_loss": 0.2737290561199188, "global_step": 84516, "epoch": 949, "lr": 7.809337943649237e-05}
{"train_loss": 0.26414987444877625, "global_step": 84517, "epoch": 949, "lr": 7.809289986072654e-05}
{"train_loss": 0.34991922974586487, "global_step": 84518, "epoch": 949, "lr": 7.809242028118397e-05}
{"train_loss": 0.22897469997406006, "global_step": 84519, "epoch": 949, "lr": 7.809194069786472e-05}
{"train_loss": 0.2592001259326935, "global_step": 84520, "epoch": 949, "lr": 7.809146111076884e-05}
{"train_loss": 0.250674843788147, "global_step": 84521, "epoch": 949, "lr": 7.809098151989641e-05}
{"train_loss": 0.2967512905597687, "global_step": 84522, "epoch": 949, "lr": 7.809050192524749e-05}
{"train_loss": 0.21424943208694458, "global_step": 84523, "epoch": 949, "lr": 7.809002232682215e-05}
{"train_loss": 0.18916627764701843, "global_step": 84524, "epoch": 949, "lr": 7.808954272462043e-05}
{"train_loss": 0.15572577714920044, "global_step": 84525, "epoch": 949, "lr": 7.808906311864244e-05}
{"train_loss": 0.19354258477687836, "global_step": 84526, "epoch": 949, "lr": 7.80885835088882e-05}
{"train_loss": 0.30741000175476074, "global_step": 84527, "epoch": 949, "lr": 7.808810389535779e-05}
{"train_loss": 0.23849636316299438, "global_step": 84528, "epoch": 949, "lr": 7.808762427805128e-05}
{"train_loss": 0.23365004360675812, "global_step": 84529, "epoch": 949, "lr": 7.808714465696874e-05}
{"train_loss": 0.2869528830051422, "global_step": 84530, "epoch": 949, "lr": 7.808666503211019e-05}
{"train_loss": 0.2279517948627472, "global_step": 84531, "epoch": 949, "lr": 7.808618540347577e-05}
{"train_loss": 0.1696217656135559, "global_step": 84532, "epoch": 949, "lr": 7.80857057710655e-05}
{"train_loss": 0.2561200261116028, "global_step": 84533, "epoch": 949, "lr": 7.808522613487944e-05}
{"train_loss": 0.24278877675533295, "global_step": 84534, "epoch": 949, "lr": 7.808474649491767e-05}
{"train_loss": 0.18926328420639038, "global_step": 84535, "epoch": 949, "lr": 7.808426685118024e-05}
{"train_loss": 0.16682645678520203, "global_step": 84536, "epoch": 949, "lr": 7.808378720366722e-05}
{"train_loss": 0.18368440866470337, "global_step": 84537, "epoch": 949, "lr": 7.80833075523787e-05}
{"train_loss": 0.19321012496948242, "global_step": 84538, "epoch": 949, "lr": 7.80828278973147e-05}
{"train_loss": 0.2428053468465805, "global_step": 84539, "epoch": 949, "lr": 7.808234823847532e-05}
{"train_loss": 0.24661700427532196, "global_step": 84540, "epoch": 949, "lr": 7.80818685758606e-05}
{"train_loss": 0.2671917676925659, "global_step": 84541, "epoch": 949, "lr": 7.808138890947061e-05}
{"train_loss": 0.20253312587738037, "global_step": 84542, "epoch": 949, "lr": 7.808090923930543e-05}
{"train_loss": 0.20671184360980988, "global_step": 84543, "epoch": 949, "lr": 7.808042956536512e-05}
{"train_loss": 0.21440812945365906, "global_step": 84544, "epoch": 949, "lr": 7.807994988764973e-05}
{"train_loss": 0.20295782387256622, "global_step": 84545, "epoch": 949, "lr": 7.807947020615934e-05}
{"train_loss": 0.28469815850257874, "global_step": 84546, "epoch": 949, "lr": 7.807899052089398e-05}
{"train_loss": 0.36410513520240784, "global_step": 84547, "epoch": 949, "lr": 7.807851083185377e-05}
{"train_loss": 0.2705617845058441, "global_step": 84548, "epoch": 949, "lr": 7.807803113903873e-05}
{"train_loss": 0.21204895877771163, "global_step": 84549, "epoch": 949, "lr": 7.807755144244895e-05, "val_loss": 3.97591233253479}
{"train_loss": 0.1834959089756012, "global_step": 84550, "epoch": 950, "lr": 7.80770717420845e-05}
{"train_loss": 0.13454537093639374, "global_step": 84551, "epoch": 950, "lr": 7.807659203794539e-05}
{"train_loss": 0.16286057233810425, "global_step": 84552, "epoch": 950, "lr": 7.807611233003175e-05}
{"train_loss": 0.2600861191749573, "global_step": 84553, "epoch": 950, "lr": 7.807563261834363e-05}
{"train_loss": 0.23884892463684082, "global_step": 84554, "epoch": 950, "lr": 7.807515290288107e-05}
{"train_loss": 0.1755039095878601, "global_step": 84555, "epoch": 950, "lr": 7.807467318364415e-05}
{"train_loss": 0.26196831464767456, "global_step": 84556, "epoch": 950, "lr": 7.807419346063294e-05}
{"train_loss": 0.20863085985183716, "global_step": 84557, "epoch": 950, "lr": 7.807371373384749e-05}
{"train_loss": 0.2375679314136505, "global_step": 84558, "epoch": 950, "lr": 7.807323400328785e-05}
{"train_loss": 0.19813670217990875, "global_step": 84559, "epoch": 950, "lr": 7.807275426895413e-05}
{"train_loss": 0.23858897387981415, "global_step": 84560, "epoch": 950, "lr": 7.807227453084637e-05}
{"train_loss": 0.23098671436309814, "global_step": 84561, "epoch": 950, "lr": 7.807179478896463e-05}
{"train_loss": 0.19725842773914337, "global_step": 84562, "epoch": 950, "lr": 7.807131504330899e-05}
{"train_loss": 0.27423587441444397, "global_step": 84563, "epoch": 950, "lr": 7.807083529387949e-05}
{"train_loss": 0.26953446865081787, "global_step": 84564, "epoch": 950, "lr": 7.807035554067621e-05}
{"train_loss": 0.23650208115577698, "global_step": 84565, "epoch": 950, "lr": 7.806987578369922e-05}
{"train_loss": 0.19704671204090118, "global_step": 84566, "epoch": 950, "lr": 7.806939602294857e-05}
{"train_loss": 0.18035684525966644, "global_step": 84567, "epoch": 950, "lr": 7.806891625842433e-05}
{"train_loss": 0.21311043202877045, "global_step": 84568, "epoch": 950, "lr": 7.806843649012657e-05}
{"train_loss": 0.20592689514160156, "global_step": 84569, "epoch": 950, "lr": 7.806795671805536e-05}
{"train_loss": 0.21648874878883362, "global_step": 84570, "epoch": 950, "lr": 7.806747694221075e-05}
{"train_loss": 0.23064102232456207, "global_step": 84571, "epoch": 950, "lr": 7.806699716259281e-05}
{"train_loss": 0.18371669948101044, "global_step": 84572, "epoch": 950, "lr": 7.806651737920158e-05}
{"train_loss": 0.212555393576622, "global_step": 84573, "epoch": 950, "lr": 7.806603759203718e-05}
{"train_loss": 0.20308728516101837, "global_step": 84574, "epoch": 950, "lr": 7.806555780109963e-05}
{"train_loss": 0.14116208255290985, "global_step": 84575, "epoch": 950, "lr": 7.806507800638902e-05}
{"train_loss": 0.17864462733268738, "global_step": 84576, "epoch": 950, "lr": 7.806459820790539e-05}
{"train_loss": 0.1464509814977646, "global_step": 84577, "epoch": 950, "lr": 7.806411840564883e-05}
{"train_loss": 0.21801072359085083, "global_step": 84578, "epoch": 950, "lr": 7.806363859961937e-05}
{"train_loss": 0.21297895908355713, "global_step": 84579, "epoch": 950, "lr": 7.80631587898171e-05}
{"train_loss": 0.20607775449752808, "global_step": 84580, "epoch": 950, "lr": 7.806267897624209e-05}
{"train_loss": 0.20971117913722992, "global_step": 84581, "epoch": 950, "lr": 7.806219915889439e-05}
{"train_loss": 0.20279759168624878, "global_step": 84582, "epoch": 950, "lr": 7.806171933777409e-05}
{"train_loss": 0.13092759251594543, "global_step": 84583, "epoch": 950, "lr": 7.80612395128812e-05}
{"train_loss": 0.16299711167812347, "global_step": 84584, "epoch": 950, "lr": 7.806075968421584e-05}
{"train_loss": 0.17368249595165253, "global_step": 84585, "epoch": 950, "lr": 7.806027985177804e-05}
{"train_loss": 0.15477889776229858, "global_step": 84586, "epoch": 950, "lr": 7.805980001556789e-05}
{"train_loss": 0.2001803070306778, "global_step": 84587, "epoch": 950, "lr": 7.805932017558543e-05}
{"train_loss": 0.23100750148296356, "global_step": 84588, "epoch": 950, "lr": 7.805884033183074e-05}
{"train_loss": 0.1795877367258072, "global_step": 84589, "epoch": 950, "lr": 7.805836048430388e-05}
{"train_loss": 0.16711540520191193, "global_step": 84590, "epoch": 950, "lr": 7.805788063300493e-05}
{"train_loss": 0.166752889752388, "global_step": 84591, "epoch": 950, "lr": 7.805740077793393e-05}
{"train_loss": 0.1999649852514267, "global_step": 84592, "epoch": 950, "lr": 7.805692091909096e-05}
{"train_loss": 0.2051648199558258, "global_step": 84593, "epoch": 950, "lr": 7.805644105647608e-05}
{"train_loss": 0.2673049569129944, "global_step": 84594, "epoch": 950, "lr": 7.805596119008934e-05}
{"train_loss": 0.13253304362297058, "global_step": 84595, "epoch": 950, "lr": 7.805548131993081e-05}
{"train_loss": 0.22057676315307617, "global_step": 84596, "epoch": 950, "lr": 7.805500144600057e-05}
{"train_loss": 0.13017162680625916, "global_step": 84597, "epoch": 950, "lr": 7.805452156829869e-05}
{"train_loss": 0.1912311613559723, "global_step": 84598, "epoch": 950, "lr": 7.805404168682522e-05}
{"train_loss": 0.1380750983953476, "global_step": 84599, "epoch": 950, "lr": 7.805356180158021e-05}
{"train_loss": 0.15177184343338013, "global_step": 84600, "epoch": 950, "lr": 7.805308191256375e-05}
{"train_loss": 0.13483549654483795, "global_step": 84601, "epoch": 950, "lr": 7.805260201977591e-05}
{"train_loss": 0.22044874727725983, "global_step": 84602, "epoch": 950, "lr": 7.805212212321671e-05}
{"train_loss": 0.25615400075912476, "global_step": 84603, "epoch": 950, "lr": 7.805164222288627e-05}
{"train_loss": 0.21673640608787537, "global_step": 84604, "epoch": 950, "lr": 7.805116231878462e-05}
{"train_loss": 0.24140286445617676, "global_step": 84605, "epoch": 950, "lr": 7.805068241091182e-05}
{"train_loss": 0.16551582515239716, "global_step": 84606, "epoch": 950, "lr": 7.805020249926796e-05}
{"train_loss": 0.13074848055839539, "global_step": 84607, "epoch": 950, "lr": 7.80497225838531e-05}
{"train_loss": 0.23639583587646484, "global_step": 84608, "epoch": 950, "lr": 7.804924266466729e-05}
{"train_loss": 0.2037578523159027, "global_step": 84609, "epoch": 950, "lr": 7.804876274171061e-05}
{"train_loss": 0.15884116291999817, "global_step": 84610, "epoch": 950, "lr": 7.804828281498308e-05}
{"train_loss": 0.2007625550031662, "global_step": 84611, "epoch": 950, "lr": 7.804780288448483e-05}
{"train_loss": 0.18441741168498993, "global_step": 84612, "epoch": 950, "lr": 7.80473229502159e-05}
{"train_loss": 0.19867928326129913, "global_step": 84613, "epoch": 950, "lr": 7.804684301217635e-05}
{"train_loss": 0.1536768078804016, "global_step": 84614, "epoch": 950, "lr": 7.804636307036623e-05}
{"train_loss": 0.1714506894350052, "global_step": 84615, "epoch": 950, "lr": 7.804588312478561e-05}
{"train_loss": 0.2738145887851715, "global_step": 84616, "epoch": 950, "lr": 7.804540317543458e-05}
{"train_loss": 0.13636139035224915, "global_step": 84617, "epoch": 950, "lr": 7.804492322231319e-05}
{"train_loss": 0.202456533908844, "global_step": 84618, "epoch": 950, "lr": 7.804444326542148e-05}
{"train_loss": 0.19210997223854065, "global_step": 84619, "epoch": 950, "lr": 7.804396330475955e-05}
{"train_loss": 0.15525364875793457, "global_step": 84620, "epoch": 950, "lr": 7.804348334032745e-05}
{"train_loss": 0.1518617421388626, "global_step": 84621, "epoch": 950, "lr": 7.804300337212525e-05}
{"train_loss": 0.21475549042224884, "global_step": 84622, "epoch": 950, "lr": 7.8042523400153e-05}
{"train_loss": 0.20126201212406158, "global_step": 84623, "epoch": 950, "lr": 7.804204342441078e-05}
{"train_loss": 0.22888362407684326, "global_step": 84624, "epoch": 950, "lr": 7.804156344489866e-05}
{"train_loss": 0.11813944578170776, "global_step": 84625, "epoch": 950, "lr": 7.804108346161667e-05}
{"train_loss": 0.2158350795507431, "global_step": 84626, "epoch": 950, "lr": 7.804060347456492e-05}
{"train_loss": 0.2996523678302765, "global_step": 84627, "epoch": 950, "lr": 7.804012348374346e-05}
{"train_loss": 0.23294803500175476, "global_step": 84628, "epoch": 950, "lr": 7.803964348915232e-05}
{"train_loss": 0.28082776069641113, "global_step": 84629, "epoch": 950, "lr": 7.80391634907916e-05}
{"train_loss": 0.12200229614973068, "global_step": 84630, "epoch": 950, "lr": 7.803868348866137e-05}
{"train_loss": 0.18827027082443237, "global_step": 84631, "epoch": 950, "lr": 7.803820348276166e-05}
{"train_loss": 0.16956448554992676, "global_step": 84632, "epoch": 950, "lr": 7.803772347309258e-05}
{"train_loss": 0.2377290278673172, "global_step": 84633, "epoch": 950, "lr": 7.803724345965414e-05}
{"train_loss": 0.12819573283195496, "global_step": 84634, "epoch": 950, "lr": 7.803676344244646e-05}
{"train_loss": 0.1762920618057251, "global_step": 84635, "epoch": 950, "lr": 7.803628342146956e-05}
{"train_loss": 0.20724666118621826, "global_step": 84636, "epoch": 950, "lr": 7.803580339672354e-05}
{"train_loss": 0.16254614293575287, "global_step": 84637, "epoch": 950, "lr": 7.803532336820842e-05}
{"train_loss": 0.19566504163353632, "global_step": 84638, "epoch": 950, "lr": 7.803484333592432e-05, "train/sim_max_reward_0": 0.5870295146599736, "train/sim_max_reward_1": 0.5391368534040151, "train/sim_max_reward_2": 0.67244238735979, "train/sim_max_reward_3": 0.5520022059319455, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9021201377167325, "test/sim_max_reward_4500000": 0.9964893197896647, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.3171761277751452, "test/sim_max_reward_4500003": 0.7212657273353718, "test/sim_max_reward_4500004": 0.960066951386366, "test/sim_max_reward_4500005": 0.977838669793644, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.2621886622682455, "test/sim_max_reward_4500009": 0.050818248395612345, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9622995727560143, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.601474832086475, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9918294210952894, "test/sim_max_reward_4500016": 1.0, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 0.1511689531214938, "test/sim_max_reward_4500019": 0.9852266738257077, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 1.0, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9889919156115532, "test/sim_max_reward_4500026": 0.6201049685740686, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.4521534973217031, "test/sim_max_reward_4500029": 1.0, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.5428925727690191, "test/sim_max_reward_4500032": 1.0, "test/sim_max_reward_4500033": 0.7258974188997843, "test/sim_max_reward_4500034": 0.5499511708620335, "test/sim_max_reward_4500035": 0.7215655230385208, "test/sim_max_reward_4500036": 0.15366037124113888, "test/sim_max_reward_4500037": 0.9564029096572174, "test/sim_max_reward_4500038": 0.625031602172228, "test/sim_max_reward_4500039": 0.9945306583560214, "test/sim_max_reward_4500040": 0.9847826160066764, "test/sim_max_reward_4500041": 0.7083099653916574, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.6976378204624212, "test/sim_max_reward_4500045": 0.1509024179127279, "test/sim_max_reward_4500046": 0.4908842789660534, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9190652318112043, "test/sim_max_reward_4500049": 0.12512634911434875, "train/mean_score": 0.7087885165120761, "test/mean_score": 0.6277146889559482, "val_loss": 4.07770299911499, "train_action_mse_error": 14.034894943237305}
{"train_loss": 0.15356098115444183, "global_step": 84639, "epoch": 951, "lr": 7.803436329987127e-05}
{"train_loss": 0.19139906764030457, "global_step": 84640, "epoch": 951, "lr": 7.803388326004935e-05}
{"train_loss": 0.16914954781532288, "global_step": 84641, "epoch": 951, "lr": 7.80334032164586e-05}
{"train_loss": 0.24288450181484222, "global_step": 84642, "epoch": 951, "lr": 7.803292316909911e-05}
{"train_loss": 0.1537487655878067, "global_step": 84643, "epoch": 951, "lr": 7.803244311797093e-05}
{"train_loss": 0.22290927171707153, "global_step": 84644, "epoch": 951, "lr": 7.803196306307413e-05}
{"train_loss": 0.19034907221794128, "global_step": 84645, "epoch": 951, "lr": 7.803148300440877e-05}
{"train_loss": 0.23569421470165253, "global_step": 84646, "epoch": 951, "lr": 7.803100294197492e-05}
{"train_loss": 0.17199604213237762, "global_step": 84647, "epoch": 951, "lr": 7.803052287577266e-05}
{"train_loss": 0.1473817229270935, "global_step": 84648, "epoch": 951, "lr": 7.803004280580203e-05}
{"train_loss": 0.1910918951034546, "global_step": 84649, "epoch": 951, "lr": 7.80295627320631e-05}
{"train_loss": 0.19860173761844635, "global_step": 84650, "epoch": 951, "lr": 7.802908265455592e-05}
{"train_loss": 0.1846877932548523, "global_step": 84651, "epoch": 951, "lr": 7.80286025732806e-05}
{"train_loss": 0.15067942440509796, "global_step": 84652, "epoch": 951, "lr": 7.802812248823717e-05}
{"train_loss": 0.137423574924469, "global_step": 84653, "epoch": 951, "lr": 7.802764239942568e-05}
{"train_loss": 0.19595888257026672, "global_step": 84654, "epoch": 951, "lr": 7.802716230684623e-05}
{"train_loss": 0.23671351373195648, "global_step": 84655, "epoch": 951, "lr": 7.802668221049887e-05}
{"train_loss": 0.20976920425891876, "global_step": 84656, "epoch": 951, "lr": 7.802620211038367e-05}
{"train_loss": 0.1704246699810028, "global_step": 84657, "epoch": 951, "lr": 7.802572200650067e-05}
{"train_loss": 0.25478166341781616, "global_step": 84658, "epoch": 951, "lr": 7.802524189884996e-05}
{"train_loss": 0.2711074948310852, "global_step": 84659, "epoch": 951, "lr": 7.802476178743161e-05}
{"train_loss": 0.2041802555322647, "global_step": 84660, "epoch": 951, "lr": 7.802428167224566e-05}
{"train_loss": 0.24734336137771606, "global_step": 84661, "epoch": 951, "lr": 7.80238015532922e-05}
{"train_loss": 0.14944098889827728, "global_step": 84662, "epoch": 951, "lr": 7.802332143057127e-05}
{"train_loss": 0.18671606481075287, "global_step": 84663, "epoch": 951, "lr": 7.802284130408295e-05}
{"train_loss": 0.20716404914855957, "global_step": 84664, "epoch": 951, "lr": 7.80223611738273e-05}
{"train_loss": 0.15960995852947235, "global_step": 84665, "epoch": 951, "lr": 7.802188103980439e-05}
{"train_loss": 0.1982731819152832, "global_step": 84666, "epoch": 951, "lr": 7.802140090201428e-05}
{"train_loss": 0.18439257144927979, "global_step": 84667, "epoch": 951, "lr": 7.802092076045702e-05}
{"train_loss": 0.23426619172096252, "global_step": 84668, "epoch": 951, "lr": 7.802044061513269e-05}
{"train_loss": 0.18440498411655426, "global_step": 84669, "epoch": 951, "lr": 7.801996046604136e-05}
{"train_loss": 0.3644886910915375, "global_step": 84670, "epoch": 951, "lr": 7.80194803131831e-05}
{"train_loss": 0.18013788759708405, "global_step": 84671, "epoch": 951, "lr": 7.801900015655795e-05}
{"train_loss": 0.1950458437204361, "global_step": 84672, "epoch": 951, "lr": 7.801851999616598e-05}
{"train_loss": 0.18699073791503906, "global_step": 84673, "epoch": 951, "lr": 7.801803983200728e-05}
{"train_loss": 0.19571085274219513, "global_step": 84674, "epoch": 951, "lr": 7.801755966408188e-05}
{"train_loss": 0.1592869609594345, "global_step": 84675, "epoch": 951, "lr": 7.801707949238987e-05}
{"train_loss": 0.19000443816184998, "global_step": 84676, "epoch": 951, "lr": 7.80165993169313e-05}
{"train_loss": 0.19109483063220978, "global_step": 84677, "epoch": 951, "lr": 7.801611913770623e-05}
{"train_loss": 0.1481623500585556, "global_step": 84678, "epoch": 951, "lr": 7.801563895471476e-05}
{"train_loss": 0.2201617807149887, "global_step": 84679, "epoch": 951, "lr": 7.80151587679569e-05}
{"train_loss": 0.2302306741476059, "global_step": 84680, "epoch": 951, "lr": 7.801467857743276e-05}
{"train_loss": 0.34774747490882874, "global_step": 84681, "epoch": 951, "lr": 7.80141983831424e-05}
{"train_loss": 0.2174767553806305, "global_step": 84682, "epoch": 951, "lr": 7.801371818508585e-05}
{"train_loss": 0.21493659913539886, "global_step": 84683, "epoch": 951, "lr": 7.80132379832632e-05}
{"train_loss": 0.18597370386123657, "global_step": 84684, "epoch": 951, "lr": 7.801275777767451e-05}
{"train_loss": 0.13473349809646606, "global_step": 84685, "epoch": 951, "lr": 7.801227756831986e-05}
{"train_loss": 0.1101197749376297, "global_step": 84686, "epoch": 951, "lr": 7.80117973551993e-05}
{"train_loss": 0.16129226982593536, "global_step": 84687, "epoch": 951, "lr": 7.80113171383129e-05}
{"train_loss": 0.24296538531780243, "global_step": 84688, "epoch": 951, "lr": 7.80108369176607e-05}
{"train_loss": 0.2164693921804428, "global_step": 84689, "epoch": 951, "lr": 7.801035669324278e-05}
{"train_loss": 0.17696118354797363, "global_step": 84690, "epoch": 951, "lr": 7.800987646505921e-05}
{"train_loss": 0.18417388200759888, "global_step": 84691, "epoch": 951, "lr": 7.800939623311008e-05}
{"train_loss": 0.19601315259933472, "global_step": 84692, "epoch": 951, "lr": 7.800891599739541e-05}
{"train_loss": 0.1853622943162918, "global_step": 84693, "epoch": 951, "lr": 7.800843575791528e-05}
{"train_loss": 0.20091557502746582, "global_step": 84694, "epoch": 951, "lr": 7.800795551466977e-05}
{"train_loss": 0.15536187589168549, "global_step": 84695, "epoch": 951, "lr": 7.800747526765893e-05}
{"train_loss": 0.15787659585475922, "global_step": 84696, "epoch": 951, "lr": 7.800699501688281e-05}
{"train_loss": 0.2063833624124527, "global_step": 84697, "epoch": 951, "lr": 7.800651476234152e-05}
{"train_loss": 0.18273994326591492, "global_step": 84698, "epoch": 951, "lr": 7.800603450403506e-05}
{"train_loss": 0.20109793543815613, "global_step": 84699, "epoch": 951, "lr": 7.800555424196355e-05}
{"train_loss": 0.16679789125919342, "global_step": 84700, "epoch": 951, "lr": 7.800507397612703e-05}
{"train_loss": 0.155427485704422, "global_step": 84701, "epoch": 951, "lr": 7.800459370652557e-05}
{"train_loss": 0.18669471144676208, "global_step": 84702, "epoch": 951, "lr": 7.800411343315924e-05}
{"train_loss": 0.17443892359733582, "global_step": 84703, "epoch": 951, "lr": 7.800363315602809e-05}
{"train_loss": 0.21086621284484863, "global_step": 84704, "epoch": 951, "lr": 7.80031528751322e-05}
{"train_loss": 0.16759787499904633, "global_step": 84705, "epoch": 951, "lr": 7.800267259047163e-05}
{"train_loss": 0.20853929221630096, "global_step": 84706, "epoch": 951, "lr": 7.800219230204643e-05}
{"train_loss": 0.15610209107398987, "global_step": 84707, "epoch": 951, "lr": 7.800171200985667e-05}
{"train_loss": 0.1693018078804016, "global_step": 84708, "epoch": 951, "lr": 7.800123171390243e-05}
{"train_loss": 0.20162270963191986, "global_step": 84709, "epoch": 951, "lr": 7.800075141418378e-05}
{"train_loss": 0.17343661189079285, "global_step": 84710, "epoch": 951, "lr": 7.800027111070077e-05}
{"train_loss": 0.19223684072494507, "global_step": 84711, "epoch": 951, "lr": 7.799979080345344e-05}
{"train_loss": 0.13567999005317688, "global_step": 84712, "epoch": 951, "lr": 7.79993104924419e-05}
{"train_loss": 0.19888366758823395, "global_step": 84713, "epoch": 951, "lr": 7.799883017766619e-05}
{"train_loss": 0.25273439288139343, "global_step": 84714, "epoch": 951, "lr": 7.799834985912636e-05}
{"train_loss": 0.21802103519439697, "global_step": 84715, "epoch": 951, "lr": 7.799786953682252e-05}
{"train_loss": 0.23276464641094208, "global_step": 84716, "epoch": 951, "lr": 7.799738921075468e-05}
{"train_loss": 0.17522482573986053, "global_step": 84717, "epoch": 951, "lr": 7.799690888092295e-05}
{"train_loss": 0.19443774223327637, "global_step": 84718, "epoch": 951, "lr": 7.799642854732739e-05}
{"train_loss": 0.14917297661304474, "global_step": 84719, "epoch": 951, "lr": 7.799594820996802e-05}
{"train_loss": 0.1359972208738327, "global_step": 84720, "epoch": 951, "lr": 7.799546786884495e-05}
{"train_loss": 0.22403235733509064, "global_step": 84721, "epoch": 951, "lr": 7.799498752395824e-05}
{"train_loss": 0.35666295886039734, "global_step": 84722, "epoch": 951, "lr": 7.799450717530793e-05}
{"train_loss": 0.1367207020521164, "global_step": 84723, "epoch": 951, "lr": 7.799402682289411e-05}
{"train_loss": 0.15318924188613892, "global_step": 84724, "epoch": 951, "lr": 7.799354646671683e-05}
{"train_loss": 0.15363013744354248, "global_step": 84725, "epoch": 951, "lr": 7.799306610677616e-05}
{"train_loss": 0.13294069468975067, "global_step": 84726, "epoch": 951, "lr": 7.799258574307218e-05}
{"train_loss": 0.19289866709307338, "global_step": 84727, "epoch": 951, "lr": 7.799210537560492e-05, "val_loss": 4.2603068351745605}
{"train_loss": 0.19985240697860718, "global_step": 84728, "epoch": 952, "lr": 7.799162500437447e-05}
{"train_loss": 0.1720067411661148, "global_step": 84729, "epoch": 952, "lr": 7.799114462938088e-05}
{"train_loss": 0.16764606535434723, "global_step": 84730, "epoch": 952, "lr": 7.799066425062422e-05}
{"train_loss": 0.1431395709514618, "global_step": 84731, "epoch": 952, "lr": 7.799018386810456e-05}
{"train_loss": 0.12288811802864075, "global_step": 84732, "epoch": 952, "lr": 7.798970348182196e-05}
{"train_loss": 0.2614210844039917, "global_step": 84733, "epoch": 952, "lr": 7.79892230917765e-05}
{"train_loss": 0.25685814023017883, "global_step": 84734, "epoch": 952, "lr": 7.798874269796822e-05}
{"train_loss": 0.1897708624601364, "global_step": 84735, "epoch": 952, "lr": 7.798826230039717e-05}
{"train_loss": 0.19655898213386536, "global_step": 84736, "epoch": 952, "lr": 7.798778189906348e-05}
{"train_loss": 0.14715251326560974, "global_step": 84737, "epoch": 952, "lr": 7.798730149396715e-05}
{"train_loss": 0.15876492857933044, "global_step": 84738, "epoch": 952, "lr": 7.798682108510827e-05}
{"train_loss": 0.19626682996749878, "global_step": 84739, "epoch": 952, "lr": 7.798634067248692e-05}
{"train_loss": 0.16633956134319305, "global_step": 84740, "epoch": 952, "lr": 7.798586025610312e-05}
{"train_loss": 0.16137446463108063, "global_step": 84741, "epoch": 952, "lr": 7.798537983595698e-05}
{"train_loss": 0.15166230499744415, "global_step": 84742, "epoch": 952, "lr": 7.798489941204855e-05}
{"train_loss": 0.22628052532672882, "global_step": 84743, "epoch": 952, "lr": 7.798441898437788e-05}
{"train_loss": 0.25484544038772583, "global_step": 84744, "epoch": 952, "lr": 7.798393855294504e-05}
{"train_loss": 0.19281837344169617, "global_step": 84745, "epoch": 952, "lr": 7.798345811775012e-05}
{"train_loss": 0.19236865639686584, "global_step": 84746, "epoch": 952, "lr": 7.798297767879316e-05}
{"train_loss": 0.16268855333328247, "global_step": 84747, "epoch": 952, "lr": 7.79824972360742e-05}
{"train_loss": 0.16060101985931396, "global_step": 84748, "epoch": 952, "lr": 7.798201678959337e-05}
{"train_loss": 0.27617233991622925, "global_step": 84749, "epoch": 952, "lr": 7.79815363393507e-05}
{"train_loss": 0.1489395946264267, "global_step": 84750, "epoch": 952, "lr": 7.798105588534623e-05}
{"train_loss": 0.17347706854343414, "global_step": 84751, "epoch": 952, "lr": 7.798057542758007e-05}
{"train_loss": 0.11177980154752731, "global_step": 84752, "epoch": 952, "lr": 7.798009496605226e-05}
{"train_loss": 0.1793411672115326, "global_step": 84753, "epoch": 952, "lr": 7.797961450076285e-05}
{"train_loss": 0.16420824825763702, "global_step": 84754, "epoch": 952, "lr": 7.797913403171194e-05}
{"train_loss": 0.21131737530231476, "global_step": 84755, "epoch": 952, "lr": 7.797865355889956e-05}
{"train_loss": 0.12582872807979584, "global_step": 84756, "epoch": 952, "lr": 7.79781730823258e-05}
{"train_loss": 0.16838321089744568, "global_step": 84757, "epoch": 952, "lr": 7.797769260199071e-05}
{"train_loss": 0.22470150887966156, "global_step": 84758, "epoch": 952, "lr": 7.797721211789438e-05}
{"train_loss": 0.14488890767097473, "global_step": 84759, "epoch": 952, "lr": 7.797673163003683e-05}
{"train_loss": 0.18694345653057098, "global_step": 84760, "epoch": 952, "lr": 7.797625113841815e-05}
{"train_loss": 0.2077244073152542, "global_step": 84761, "epoch": 952, "lr": 7.797577064303842e-05}
{"train_loss": 0.15009371936321259, "global_step": 84762, "epoch": 952, "lr": 7.797529014389769e-05}
{"train_loss": 0.20857954025268555, "global_step": 84763, "epoch": 952, "lr": 7.797480964099602e-05}
{"train_loss": 0.2136392891407013, "global_step": 84764, "epoch": 952, "lr": 7.797432913433347e-05}
{"train_loss": 0.16700969636440277, "global_step": 84765, "epoch": 952, "lr": 7.797384862391011e-05}
{"train_loss": 0.162260964512825, "global_step": 84766, "epoch": 952, "lr": 7.797336810972602e-05}
{"train_loss": 0.18204258382320404, "global_step": 84767, "epoch": 952, "lr": 7.797288759178123e-05}
{"train_loss": 0.1741969883441925, "global_step": 84768, "epoch": 952, "lr": 7.797240707007585e-05}
{"train_loss": 0.17937710881233215, "global_step": 84769, "epoch": 952, "lr": 7.79719265446099e-05}
{"train_loss": 0.15570832788944244, "global_step": 84770, "epoch": 952, "lr": 7.797144601538348e-05}
{"train_loss": 0.1641937494277954, "global_step": 84771, "epoch": 952, "lr": 7.797096548239664e-05}
{"train_loss": 0.19201594591140747, "global_step": 84772, "epoch": 952, "lr": 7.797048494564945e-05}
{"train_loss": 0.16481748223304749, "global_step": 84773, "epoch": 952, "lr": 7.797000440514196e-05}
{"train_loss": 0.18775857985019684, "global_step": 84774, "epoch": 952, "lr": 7.796952386087423e-05}
{"train_loss": 0.19378647208213806, "global_step": 84775, "epoch": 952, "lr": 7.796904331284636e-05}
{"train_loss": 0.21935389935970306, "global_step": 84776, "epoch": 952, "lr": 7.79685627610584e-05}
{"train_loss": 0.33638638257980347, "global_step": 84777, "epoch": 952, "lr": 7.796808220551038e-05}
{"train_loss": 0.20923681557178497, "global_step": 84778, "epoch": 952, "lr": 7.79676016462024e-05}
{"train_loss": 0.22633662819862366, "global_step": 84779, "epoch": 952, "lr": 7.796712108313453e-05}
{"train_loss": 0.2039192169904709, "global_step": 84780, "epoch": 952, "lr": 7.796664051630681e-05}
{"train_loss": 0.20307670533657074, "global_step": 84781, "epoch": 952, "lr": 7.796615994571933e-05}
{"train_loss": 0.13638465106487274, "global_step": 84782, "epoch": 952, "lr": 7.796567937137214e-05}
{"train_loss": 0.1635943502187729, "global_step": 84783, "epoch": 952, "lr": 7.796519879326529e-05}
{"train_loss": 0.13820880651474, "global_step": 84784, "epoch": 952, "lr": 7.796471821139886e-05}
{"train_loss": 0.16844947636127472, "global_step": 84785, "epoch": 952, "lr": 7.796423762577291e-05}
{"train_loss": 0.2010326087474823, "global_step": 84786, "epoch": 952, "lr": 7.796375703638752e-05}
{"train_loss": 0.17319262027740479, "global_step": 84787, "epoch": 952, "lr": 7.796327644324275e-05}
{"train_loss": 0.15823818743228912, "global_step": 84788, "epoch": 952, "lr": 7.796279584633866e-05}
{"train_loss": 0.1411881446838379, "global_step": 84789, "epoch": 952, "lr": 7.79623152456753e-05}
{"train_loss": 0.15342958271503448, "global_step": 84790, "epoch": 952, "lr": 7.796183464125275e-05}
{"train_loss": 0.22022321820259094, "global_step": 84791, "epoch": 952, "lr": 7.796135403307108e-05}
{"train_loss": 0.1568007469177246, "global_step": 84792, "epoch": 952, "lr": 7.796087342113035e-05}
{"train_loss": 0.23107752203941345, "global_step": 84793, "epoch": 952, "lr": 7.796039280543059e-05}
{"train_loss": 0.14472180604934692, "global_step": 84794, "epoch": 952, "lr": 7.795991218597193e-05}
{"train_loss": 0.19888019561767578, "global_step": 84795, "epoch": 952, "lr": 7.795943156275438e-05}
{"train_loss": 0.0933527946472168, "global_step": 84796, "epoch": 952, "lr": 7.795895093577804e-05}
{"train_loss": 0.15539538860321045, "global_step": 84797, "epoch": 952, "lr": 7.795847030504296e-05}
{"train_loss": 0.240658700466156, "global_step": 84798, "epoch": 952, "lr": 7.795798967054919e-05}
{"train_loss": 0.23459863662719727, "global_step": 84799, "epoch": 952, "lr": 7.795750903229682e-05}
{"train_loss": 0.22182485461235046, "global_step": 84800, "epoch": 952, "lr": 7.79570283902859e-05}
{"train_loss": 0.14598466455936432, "global_step": 84801, "epoch": 952, "lr": 7.79565477445165e-05}
{"train_loss": 0.3114717900753021, "global_step": 84802, "epoch": 952, "lr": 7.795606709498869e-05}
{"train_loss": 0.12354186922311783, "global_step": 84803, "epoch": 952, "lr": 7.795558644170251e-05}
{"train_loss": 0.15419578552246094, "global_step": 84804, "epoch": 952, "lr": 7.795510578465806e-05}
{"train_loss": 0.16453878581523895, "global_step": 84805, "epoch": 952, "lr": 7.795462512385537e-05}
{"train_loss": 0.16654720902442932, "global_step": 84806, "epoch": 952, "lr": 7.795414445929452e-05}
{"train_loss": 0.16864120960235596, "global_step": 84807, "epoch": 952, "lr": 7.79536637909756e-05}
{"train_loss": 0.14515365660190582, "global_step": 84808, "epoch": 952, "lr": 7.795318311889863e-05}
{"train_loss": 0.1425534337759018, "global_step": 84809, "epoch": 952, "lr": 7.79527024430637e-05}
{"train_loss": 0.1803833544254303, "global_step": 84810, "epoch": 952, "lr": 7.795222176347088e-05}
{"train_loss": 0.12905184924602509, "global_step": 84811, "epoch": 952, "lr": 7.795174108012022e-05}
{"train_loss": 0.1651577204465866, "global_step": 84812, "epoch": 952, "lr": 7.795126039301177e-05}
{"train_loss": 0.2764930725097656, "global_step": 84813, "epoch": 952, "lr": 7.795077970214563e-05}
{"train_loss": 0.16554906964302063, "global_step": 84814, "epoch": 952, "lr": 7.795029900752185e-05}
{"train_loss": 0.16643187403678894, "global_step": 84815, "epoch": 952, "lr": 7.79498183091405e-05}
{"train_loss": 0.18193161219693302, "global_step": 84816, "epoch": 952, "lr": 7.794933760700162e-05, "val_loss": 4.212191104888916}
{"train_loss": 0.14005401730537415, "global_step": 84817, "epoch": 953, "lr": 7.794885690110531e-05}
{"train_loss": 0.15098848938941956, "global_step": 84818, "epoch": 953, "lr": 7.79483761914516e-05}
{"train_loss": 0.14113089442253113, "global_step": 84819, "epoch": 953, "lr": 7.794789547804057e-05}
{"train_loss": 0.21235539019107819, "global_step": 84820, "epoch": 953, "lr": 7.794741476087231e-05}
{"train_loss": 0.14265039563179016, "global_step": 84821, "epoch": 953, "lr": 7.794693403994684e-05}
{"train_loss": 0.12369772791862488, "global_step": 84822, "epoch": 953, "lr": 7.794645331526425e-05}
{"train_loss": 0.1583445966243744, "global_step": 84823, "epoch": 953, "lr": 7.79459725868246e-05}
{"train_loss": 0.1398778110742569, "global_step": 84824, "epoch": 953, "lr": 7.794549185462795e-05}
{"train_loss": 0.18362775444984436, "global_step": 84825, "epoch": 953, "lr": 7.794501111867437e-05}
{"train_loss": 0.1380893886089325, "global_step": 84826, "epoch": 953, "lr": 7.794453037896393e-05}
{"train_loss": 0.25417572259902954, "global_step": 84827, "epoch": 953, "lr": 7.794404963549668e-05}
{"train_loss": 0.14461173117160797, "global_step": 84828, "epoch": 953, "lr": 7.794356888827271e-05}
{"train_loss": 0.19536390900611877, "global_step": 84829, "epoch": 953, "lr": 7.794308813729206e-05}
{"train_loss": 0.2672503888607025, "global_step": 84830, "epoch": 953, "lr": 7.79426073825548e-05}
{"train_loss": 0.1929454654455185, "global_step": 84831, "epoch": 953, "lr": 7.7942126624061e-05}
{"train_loss": 0.1327139288187027, "global_step": 84832, "epoch": 953, "lr": 7.794164586181072e-05}
{"train_loss": 0.15803313255310059, "global_step": 84833, "epoch": 953, "lr": 7.794116509580403e-05}
{"train_loss": 0.19920165836811066, "global_step": 84834, "epoch": 953, "lr": 7.794068432604099e-05}
{"train_loss": 0.12785230576992035, "global_step": 84835, "epoch": 953, "lr": 7.794020355252166e-05}
{"train_loss": 0.20526765286922455, "global_step": 84836, "epoch": 953, "lr": 7.793972277524611e-05}
{"train_loss": 0.1811673939228058, "global_step": 84837, "epoch": 953, "lr": 7.79392419942144e-05}
{"train_loss": 0.16247133910655975, "global_step": 84838, "epoch": 953, "lr": 7.79387612094266e-05}
{"train_loss": 0.08813206106424332, "global_step": 84839, "epoch": 953, "lr": 7.793828042088277e-05}
{"train_loss": 0.13605670630931854, "global_step": 84840, "epoch": 953, "lr": 7.793779962858299e-05}
{"train_loss": 0.14892475306987762, "global_step": 84841, "epoch": 953, "lr": 7.793731883252731e-05}
{"train_loss": 0.11740731447935104, "global_step": 84842, "epoch": 953, "lr": 7.793683803271581e-05}
{"train_loss": 0.14703553915023804, "global_step": 84843, "epoch": 953, "lr": 7.793635722914853e-05}
{"train_loss": 0.18258307874202728, "global_step": 84844, "epoch": 953, "lr": 7.793587642182555e-05}
{"train_loss": 0.10297214984893799, "global_step": 84845, "epoch": 953, "lr": 7.793539561074693e-05}
{"train_loss": 0.1461021602153778, "global_step": 84846, "epoch": 953, "lr": 7.793491479591274e-05}
{"train_loss": 0.16593140363693237, "global_step": 84847, "epoch": 953, "lr": 7.793443397732303e-05}
{"train_loss": 0.12028691172599792, "global_step": 84848, "epoch": 953, "lr": 7.793395315497789e-05}
{"train_loss": 0.13619470596313477, "global_step": 84849, "epoch": 953, "lr": 7.793347232887737e-05}
{"train_loss": 0.15428215265274048, "global_step": 84850, "epoch": 953, "lr": 7.793299149902153e-05}
{"train_loss": 0.21074065566062927, "global_step": 84851, "epoch": 953, "lr": 7.793251066541044e-05}
{"train_loss": 0.1884237825870514, "global_step": 84852, "epoch": 953, "lr": 7.793202982804417e-05}
{"train_loss": 0.17629139125347137, "global_step": 84853, "epoch": 953, "lr": 7.793154898692276e-05}
{"train_loss": 0.1563645303249359, "global_step": 84854, "epoch": 953, "lr": 7.793106814204631e-05}
{"train_loss": 0.2136698216199875, "global_step": 84855, "epoch": 953, "lr": 7.793058729341488e-05}
{"train_loss": 0.18367885053157806, "global_step": 84856, "epoch": 953, "lr": 7.79301064410285e-05}
{"train_loss": 0.191403329372406, "global_step": 84857, "epoch": 953, "lr": 7.792962558488727e-05}
{"train_loss": 0.15080080926418304, "global_step": 84858, "epoch": 953, "lr": 7.792914472499123e-05}
{"train_loss": 0.10906396806240082, "global_step": 84859, "epoch": 953, "lr": 7.792866386134047e-05}
{"train_loss": 0.18765613436698914, "global_step": 84860, "epoch": 953, "lr": 7.792818299393504e-05}
{"train_loss": 0.16994507610797882, "global_step": 84861, "epoch": 953, "lr": 7.7927702122775e-05}
{"train_loss": 0.23434166610240936, "global_step": 84862, "epoch": 953, "lr": 7.792722124786043e-05}
{"train_loss": 0.2188870757818222, "global_step": 84863, "epoch": 953, "lr": 7.792674036919138e-05}
{"train_loss": 0.1830165833234787, "global_step": 84864, "epoch": 953, "lr": 7.792625948676791e-05}
{"train_loss": 0.15858244895935059, "global_step": 84865, "epoch": 953, "lr": 7.792577860059011e-05}
{"train_loss": 0.2076219916343689, "global_step": 84866, "epoch": 953, "lr": 7.792529771065803e-05}
{"train_loss": 0.16271193325519562, "global_step": 84867, "epoch": 953, "lr": 7.792481681697172e-05}
{"train_loss": 0.09858985245227814, "global_step": 84868, "epoch": 953, "lr": 7.792433591953128e-05}
{"train_loss": 0.26347815990448, "global_step": 84869, "epoch": 953, "lr": 7.792385501833674e-05}
{"train_loss": 0.20286139845848083, "global_step": 84870, "epoch": 953, "lr": 7.792337411338817e-05}
{"train_loss": 0.1269267350435257, "global_step": 84871, "epoch": 953, "lr": 7.792289320468566e-05}
{"train_loss": 0.1207088828086853, "global_step": 84872, "epoch": 953, "lr": 7.792241229222924e-05}
{"train_loss": 0.17710378766059875, "global_step": 84873, "epoch": 953, "lr": 7.7921931376019e-05}
{"train_loss": 0.22358602285385132, "global_step": 84874, "epoch": 953, "lr": 7.792145045605499e-05}
{"train_loss": 0.15571148693561554, "global_step": 84875, "epoch": 953, "lr": 7.79209695323373e-05}
{"train_loss": 0.31391090154647827, "global_step": 84876, "epoch": 953, "lr": 7.792048860486596e-05}
{"train_loss": 0.20056308805942535, "global_step": 84877, "epoch": 953, "lr": 7.792000767364105e-05}
{"train_loss": 0.30026620626449585, "global_step": 84878, "epoch": 953, "lr": 7.791952673866264e-05}
{"train_loss": 0.18741777539253235, "global_step": 84879, "epoch": 953, "lr": 7.79190457999308e-05}
{"train_loss": 0.150114044547081, "global_step": 84880, "epoch": 953, "lr": 7.791856485744556e-05}
{"train_loss": 0.16101695597171783, "global_step": 84881, "epoch": 953, "lr": 7.791808391120703e-05}
{"train_loss": 0.21785327792167664, "global_step": 84882, "epoch": 953, "lr": 7.791760296121524e-05}
{"train_loss": 0.15481573343276978, "global_step": 84883, "epoch": 953, "lr": 7.791712200747028e-05}
{"train_loss": 0.1778564304113388, "global_step": 84884, "epoch": 953, "lr": 7.791664104997221e-05}
{"train_loss": 0.1666967272758484, "global_step": 84885, "epoch": 953, "lr": 7.791616008872107e-05}
{"train_loss": 0.2650410234928131, "global_step": 84886, "epoch": 953, "lr": 7.791567912371695e-05}
{"train_loss": 0.3151099681854248, "global_step": 84887, "epoch": 953, "lr": 7.791519815495992e-05}
{"train_loss": 0.23472827672958374, "global_step": 84888, "epoch": 953, "lr": 7.791471718245e-05}
{"train_loss": 0.1666153073310852, "global_step": 84889, "epoch": 953, "lr": 7.791423620618731e-05}
{"train_loss": 0.23277157545089722, "global_step": 84890, "epoch": 953, "lr": 7.79137552261719e-05}
{"train_loss": 0.2937406003475189, "global_step": 84891, "epoch": 953, "lr": 7.79132742424038e-05}
{"train_loss": 0.26055672764778137, "global_step": 84892, "epoch": 953, "lr": 7.791279325488311e-05}
{"train_loss": 0.17134328186511993, "global_step": 84893, "epoch": 953, "lr": 7.791231226360989e-05}
{"train_loss": 0.21828576922416687, "global_step": 84894, "epoch": 953, "lr": 7.791183126858419e-05}
{"train_loss": 0.19207826256752014, "global_step": 84895, "epoch": 953, "lr": 7.79113502698061e-05}
{"train_loss": 0.16607920825481415, "global_step": 84896, "epoch": 953, "lr": 7.791086926727566e-05}
{"train_loss": 0.21221008896827698, "global_step": 84897, "epoch": 953, "lr": 7.791038826099295e-05}
{"train_loss": 0.14152055978775024, "global_step": 84898, "epoch": 953, "lr": 7.790990725095801e-05}
{"train_loss": 0.1257973462343216, "global_step": 84899, "epoch": 953, "lr": 7.790942623717094e-05}
{"train_loss": 0.2240939885377884, "global_step": 84900, "epoch": 953, "lr": 7.790894521963179e-05}
{"train_loss": 0.21314571797847748, "global_step": 84901, "epoch": 953, "lr": 7.790846419834061e-05}
{"train_loss": 0.2052948921918869, "global_step": 84902, "epoch": 953, "lr": 7.790798317329749e-05}
{"train_loss": 0.23478658497333527, "global_step": 84903, "epoch": 953, "lr": 7.790750214450249e-05}
{"train_loss": 0.17471030354499817, "global_step": 84904, "epoch": 953, "lr": 7.790702111195565e-05}
{"train_loss": 0.18234482340598374, "global_step": 84905, "epoch": 953, "lr": 7.790654007565706e-05, "val_loss": 4.202307224273682}
{"train_loss": 0.14745092391967773, "global_step": 84906, "epoch": 954, "lr": 7.790605903560676e-05}
{"train_loss": 0.3112486004829407, "global_step": 84907, "epoch": 954, "lr": 7.790557799180484e-05}
{"train_loss": 0.23190340399742126, "global_step": 84908, "epoch": 954, "lr": 7.790509694425135e-05}
{"train_loss": 0.2124629020690918, "global_step": 84909, "epoch": 954, "lr": 7.790461589294638e-05}
{"train_loss": 0.2838125228881836, "global_step": 84910, "epoch": 954, "lr": 7.790413483788995e-05}
{"train_loss": 0.17786651849746704, "global_step": 84911, "epoch": 954, "lr": 7.790365377908216e-05}
{"train_loss": 0.19509588181972504, "global_step": 84912, "epoch": 954, "lr": 7.790317271652305e-05}
{"train_loss": 0.2657819390296936, "global_step": 84913, "epoch": 954, "lr": 7.790269165021272e-05}
{"train_loss": 0.16740016639232635, "global_step": 84914, "epoch": 954, "lr": 7.79022105801512e-05}
{"train_loss": 0.17935681343078613, "global_step": 84915, "epoch": 954, "lr": 7.790172950633858e-05}
{"train_loss": 0.23219609260559082, "global_step": 84916, "epoch": 954, "lr": 7.790124842877489e-05}
{"train_loss": 0.2172686755657196, "global_step": 84917, "epoch": 954, "lr": 7.790076734746024e-05}
{"train_loss": 0.17559349536895752, "global_step": 84918, "epoch": 954, "lr": 7.790028626239467e-05}
{"train_loss": 0.25310054421424866, "global_step": 84919, "epoch": 954, "lr": 7.789980517357825e-05}
{"train_loss": 0.20643644034862518, "global_step": 84920, "epoch": 954, "lr": 7.789932408101102e-05}
{"train_loss": 0.20148199796676636, "global_step": 84921, "epoch": 954, "lr": 7.789884298469308e-05}
{"train_loss": 0.24590392410755157, "global_step": 84922, "epoch": 954, "lr": 7.789836188462447e-05}
{"train_loss": 0.21089401841163635, "global_step": 84923, "epoch": 954, "lr": 7.789788078080528e-05}
{"train_loss": 0.17435042560100555, "global_step": 84924, "epoch": 954, "lr": 7.789739967323555e-05}
{"train_loss": 0.19174641370773315, "global_step": 84925, "epoch": 954, "lr": 7.789691856191536e-05}
{"train_loss": 0.15779666602611542, "global_step": 84926, "epoch": 954, "lr": 7.789643744684477e-05}
{"train_loss": 0.21085040271282196, "global_step": 84927, "epoch": 954, "lr": 7.789595632802385e-05}
{"train_loss": 0.1907835751771927, "global_step": 84928, "epoch": 954, "lr": 7.789547520545264e-05}
{"train_loss": 0.19879929721355438, "global_step": 84929, "epoch": 954, "lr": 7.789499407913124e-05}
{"train_loss": 0.17904186248779297, "global_step": 84930, "epoch": 954, "lr": 7.789451294905968e-05}
{"train_loss": 0.23423263430595398, "global_step": 84931, "epoch": 954, "lr": 7.789403181523805e-05}
{"train_loss": 0.20673760771751404, "global_step": 84932, "epoch": 954, "lr": 7.789355067766641e-05}
{"train_loss": 0.16719798743724823, "global_step": 84933, "epoch": 954, "lr": 7.789306953634483e-05}
{"train_loss": 0.228923961520195, "global_step": 84934, "epoch": 954, "lr": 7.789258839127336e-05}
{"train_loss": 0.1619253009557724, "global_step": 84935, "epoch": 954, "lr": 7.789210724245207e-05}
{"train_loss": 0.17111149430274963, "global_step": 84936, "epoch": 954, "lr": 7.789162608988103e-05}
{"train_loss": 0.13304229080677032, "global_step": 84937, "epoch": 954, "lr": 7.78911449335603e-05}
{"train_loss": 0.21272683143615723, "global_step": 84938, "epoch": 954, "lr": 7.789066377348995e-05}
{"train_loss": 0.2297676056623459, "global_step": 84939, "epoch": 954, "lr": 7.789018260967002e-05}
{"train_loss": 0.11234939843416214, "global_step": 84940, "epoch": 954, "lr": 7.788970144210061e-05}
{"train_loss": 0.1592978686094284, "global_step": 84941, "epoch": 954, "lr": 7.788922027078178e-05}
{"train_loss": 0.20668578147888184, "global_step": 84942, "epoch": 954, "lr": 7.788873909571357e-05}
{"train_loss": 0.15903373062610626, "global_step": 84943, "epoch": 954, "lr": 7.788825791689606e-05}
{"train_loss": 0.24672184884548187, "global_step": 84944, "epoch": 954, "lr": 7.788777673432932e-05}
{"train_loss": 0.18536025285720825, "global_step": 84945, "epoch": 954, "lr": 7.788729554801341e-05}
{"train_loss": 0.1497810035943985, "global_step": 84946, "epoch": 954, "lr": 7.788681435794838e-05}
{"train_loss": 0.16091611981391907, "global_step": 84947, "epoch": 954, "lr": 7.788633316413432e-05}
{"train_loss": 0.12019101530313492, "global_step": 84948, "epoch": 954, "lr": 7.788585196657127e-05}
{"train_loss": 0.14221470057964325, "global_step": 84949, "epoch": 954, "lr": 7.788537076525932e-05}
{"train_loss": 0.1308988779783249, "global_step": 84950, "epoch": 954, "lr": 7.788488956019851e-05}
{"train_loss": 0.16877631843090057, "global_step": 84951, "epoch": 954, "lr": 7.788440835138893e-05}
{"train_loss": 0.19007980823516846, "global_step": 84952, "epoch": 954, "lr": 7.788392713883062e-05}
{"train_loss": 0.22185838222503662, "global_step": 84953, "epoch": 954, "lr": 7.788344592252365e-05}
{"train_loss": 0.21891292929649353, "global_step": 84954, "epoch": 954, "lr": 7.788296470246811e-05}
{"train_loss": 0.2147117555141449, "global_step": 84955, "epoch": 954, "lr": 7.788248347866404e-05}
{"train_loss": 0.23066793382167816, "global_step": 84956, "epoch": 954, "lr": 7.788200225111151e-05}
{"train_loss": 0.23218634724617004, "global_step": 84957, "epoch": 954, "lr": 7.788152101981057e-05}
{"train_loss": 0.1594710797071457, "global_step": 84958, "epoch": 954, "lr": 7.788103978476131e-05}
{"train_loss": 0.22640235722064972, "global_step": 84959, "epoch": 954, "lr": 7.788055854596379e-05}
{"train_loss": 0.1721237599849701, "global_step": 84960, "epoch": 954, "lr": 7.788007730341805e-05}
{"train_loss": 0.263962060213089, "global_step": 84961, "epoch": 954, "lr": 7.78795960571242e-05}
{"train_loss": 0.17403945326805115, "global_step": 84962, "epoch": 954, "lr": 7.787911480708225e-05}
{"train_loss": 0.171141117811203, "global_step": 84963, "epoch": 954, "lr": 7.787863355329231e-05}
{"train_loss": 0.14433221518993378, "global_step": 84964, "epoch": 954, "lr": 7.787815229575443e-05}
{"train_loss": 0.16522593796253204, "global_step": 84965, "epoch": 954, "lr": 7.787767103446866e-05}
{"train_loss": 0.2214931845664978, "global_step": 84966, "epoch": 954, "lr": 7.787718976943508e-05}
{"train_loss": 0.17903898656368256, "global_step": 84967, "epoch": 954, "lr": 7.787670850065376e-05}
{"train_loss": 0.1985037475824356, "global_step": 84968, "epoch": 954, "lr": 7.787622722812474e-05}
{"train_loss": 0.13572177290916443, "global_step": 84969, "epoch": 954, "lr": 7.787574595184812e-05}
{"train_loss": 0.1469445526599884, "global_step": 84970, "epoch": 954, "lr": 7.787526467182394e-05}
{"train_loss": 0.1782878041267395, "global_step": 84971, "epoch": 954, "lr": 7.787478338805227e-05}
{"train_loss": 0.30761778354644775, "global_step": 84972, "epoch": 954, "lr": 7.787430210053317e-05}
{"train_loss": 0.2122783064842224, "global_step": 84973, "epoch": 954, "lr": 7.787382080926671e-05}
{"train_loss": 0.20913204550743103, "global_step": 84974, "epoch": 954, "lr": 7.787333951425297e-05}
{"train_loss": 0.18387950956821442, "global_step": 84975, "epoch": 954, "lr": 7.787285821549197e-05}
{"train_loss": 0.16053640842437744, "global_step": 84976, "epoch": 954, "lr": 7.787237691298383e-05}
{"train_loss": 0.09639205038547516, "global_step": 84977, "epoch": 954, "lr": 7.787189560672859e-05}
{"train_loss": 0.16244018077850342, "global_step": 84978, "epoch": 954, "lr": 7.787141429672628e-05}
{"train_loss": 0.14098817110061646, "global_step": 84979, "epoch": 954, "lr": 7.787093298297703e-05}
{"train_loss": 0.0862475112080574, "global_step": 84980, "epoch": 954, "lr": 7.787045166548085e-05}
{"train_loss": 0.15778963267803192, "global_step": 84981, "epoch": 954, "lr": 7.786997034423786e-05}
{"train_loss": 0.17873631417751312, "global_step": 84982, "epoch": 954, "lr": 7.786948901924807e-05}
{"train_loss": 0.13387365639209747, "global_step": 84983, "epoch": 954, "lr": 7.786900769051157e-05}
{"train_loss": 0.2233899086713791, "global_step": 84984, "epoch": 954, "lr": 7.786852635802842e-05}
{"train_loss": 0.2539576590061188, "global_step": 84985, "epoch": 954, "lr": 7.786804502179869e-05}
{"train_loss": 0.19964657723903656, "global_step": 84986, "epoch": 954, "lr": 7.786756368182243e-05}
{"train_loss": 0.11980739235877991, "global_step": 84987, "epoch": 954, "lr": 7.786708233809972e-05}
{"train_loss": 0.26128047704696655, "global_step": 84988, "epoch": 954, "lr": 7.786660099063064e-05}
{"train_loss": 0.15715433657169342, "global_step": 84989, "epoch": 954, "lr": 7.786611963941521e-05}
{"train_loss": 0.1917407512664795, "global_step": 84990, "epoch": 954, "lr": 7.786563828445353e-05}
{"train_loss": 0.20698577165603638, "global_step": 84991, "epoch": 954, "lr": 7.786515692574566e-05}
{"train_loss": 0.24998661875724792, "global_step": 84992, "epoch": 954, "lr": 7.786467556329165e-05}
{"train_loss": 0.2058778554201126, "global_step": 84993, "epoch": 954, "lr": 7.786419419709159e-05}
{"train_loss": 0.19131138929155436, "global_step": 84994, "epoch": 954, "lr": 7.78637128271455e-05, "val_loss": 4.163650035858154}
{"train_loss": 0.17002055048942566, "global_step": 84995, "epoch": 955, "lr": 7.78632314534535e-05}
{"train_loss": 0.12638260424137115, "global_step": 84996, "epoch": 955, "lr": 7.786275007601562e-05}
{"train_loss": 0.23806989192962646, "global_step": 84997, "epoch": 955, "lr": 7.786226869483194e-05}
{"train_loss": 0.1920400857925415, "global_step": 84998, "epoch": 955, "lr": 7.78617873099025e-05}
{"train_loss": 0.18865567445755005, "global_step": 84999, "epoch": 955, "lr": 7.786130592122739e-05}
{"train_loss": 0.20522698760032654, "global_step": 85000, "epoch": 955, "lr": 7.786082452880667e-05}
{"train_loss": 0.20965735614299774, "global_step": 85001, "epoch": 955, "lr": 7.786034313264039e-05}
{"train_loss": 0.2029627561569214, "global_step": 85002, "epoch": 955, "lr": 7.785986173272863e-05}
{"train_loss": 0.19716492295265198, "global_step": 85003, "epoch": 955, "lr": 7.785938032907146e-05}
{"train_loss": 0.18667031824588776, "global_step": 85004, "epoch": 955, "lr": 7.785889892166894e-05}
{"train_loss": 0.16030040383338928, "global_step": 85005, "epoch": 955, "lr": 7.785841751052112e-05}
{"train_loss": 0.11536739021539688, "global_step": 85006, "epoch": 955, "lr": 7.785793609562808e-05}
{"train_loss": 0.162235289812088, "global_step": 85007, "epoch": 955, "lr": 7.785745467698989e-05}
{"train_loss": 0.16297300159931183, "global_step": 85008, "epoch": 955, "lr": 7.78569732546066e-05}
{"train_loss": 0.22252881526947021, "global_step": 85009, "epoch": 955, "lr": 7.785649182847826e-05}
{"train_loss": 0.24251072108745575, "global_step": 85010, "epoch": 955, "lr": 7.785601039860497e-05}
{"train_loss": 0.1656729131937027, "global_step": 85011, "epoch": 955, "lr": 7.785552896498678e-05}
{"train_loss": 0.17797620594501495, "global_step": 85012, "epoch": 955, "lr": 7.785504752762374e-05}
{"train_loss": 0.12995725870132446, "global_step": 85013, "epoch": 955, "lr": 7.785456608651595e-05}
{"train_loss": 0.13876494765281677, "global_step": 85014, "epoch": 955, "lr": 7.785408464166343e-05}
{"train_loss": 0.19181746244430542, "global_step": 85015, "epoch": 955, "lr": 7.785360319306628e-05}
{"train_loss": 0.14238755404949188, "global_step": 85016, "epoch": 955, "lr": 7.785312174072456e-05}
{"train_loss": 0.13275134563446045, "global_step": 85017, "epoch": 955, "lr": 7.785264028463832e-05}
{"train_loss": 0.11219611763954163, "global_step": 85018, "epoch": 955, "lr": 7.785215882480763e-05}
{"train_loss": 0.20530365407466888, "global_step": 85019, "epoch": 955, "lr": 7.785167736123256e-05}
{"train_loss": 0.1520126461982727, "global_step": 85020, "epoch": 955, "lr": 7.785119589391316e-05}
{"train_loss": 0.12941820919513702, "global_step": 85021, "epoch": 955, "lr": 7.785071442284951e-05}
{"train_loss": 0.19716376066207886, "global_step": 85022, "epoch": 955, "lr": 7.785023294804168e-05}
{"train_loss": 0.15753404796123505, "global_step": 85023, "epoch": 955, "lr": 7.784975146948972e-05}
{"train_loss": 0.1294327676296234, "global_step": 85024, "epoch": 955, "lr": 7.784926998719371e-05}
{"train_loss": 0.16759175062179565, "global_step": 85025, "epoch": 955, "lr": 7.78487885011537e-05}
{"train_loss": 0.134520024061203, "global_step": 85026, "epoch": 955, "lr": 7.784830701136975e-05}
{"train_loss": 0.10071185976266861, "global_step": 85027, "epoch": 955, "lr": 7.784782551784194e-05}
{"train_loss": 0.19257380068302155, "global_step": 85028, "epoch": 955, "lr": 7.784734402057035e-05}
{"train_loss": 0.18310682475566864, "global_step": 85029, "epoch": 955, "lr": 7.784686251955499e-05}
{"train_loss": 0.1738443821668625, "global_step": 85030, "epoch": 955, "lr": 7.784638101479598e-05}
{"train_loss": 0.23067772388458252, "global_step": 85031, "epoch": 955, "lr": 7.784589950629335e-05}
{"train_loss": 0.21293260157108307, "global_step": 85032, "epoch": 955, "lr": 7.784541799404718e-05}
{"train_loss": 0.11758670210838318, "global_step": 85033, "epoch": 955, "lr": 7.784493647805754e-05}
{"train_loss": 0.1911618411540985, "global_step": 85034, "epoch": 955, "lr": 7.784445495832448e-05}
{"train_loss": 0.18709029257297516, "global_step": 85035, "epoch": 955, "lr": 7.784397343484809e-05}
{"train_loss": 0.2407122254371643, "global_step": 85036, "epoch": 955, "lr": 7.78434919076284e-05}
{"train_loss": 0.09341662377119064, "global_step": 85037, "epoch": 955, "lr": 7.78430103766655e-05}
{"train_loss": 0.20996145904064178, "global_step": 85038, "epoch": 955, "lr": 7.784252884195945e-05}
{"train_loss": 0.22659055888652802, "global_step": 85039, "epoch": 955, "lr": 7.78420473035103e-05}
{"train_loss": 0.23352955281734467, "global_step": 85040, "epoch": 955, "lr": 7.784156576131814e-05}
{"train_loss": 0.15531066060066223, "global_step": 85041, "epoch": 955, "lr": 7.784108421538301e-05}
{"train_loss": 0.12561491131782532, "global_step": 85042, "epoch": 955, "lr": 7.784060266570499e-05}
{"train_loss": 0.21510003507137299, "global_step": 85043, "epoch": 955, "lr": 7.784012111228414e-05}
{"train_loss": 0.20729410648345947, "global_step": 85044, "epoch": 955, "lr": 7.783963955512052e-05}
{"train_loss": 0.2489020824432373, "global_step": 85045, "epoch": 955, "lr": 7.78391579942142e-05}
{"train_loss": 0.24300037324428558, "global_step": 85046, "epoch": 955, "lr": 7.783867642956524e-05}
{"train_loss": 0.21943069994449615, "global_step": 85047, "epoch": 955, "lr": 7.783819486117371e-05}
{"train_loss": 0.27001333236694336, "global_step": 85048, "epoch": 955, "lr": 7.78377132890397e-05}
{"train_loss": 0.1463426649570465, "global_step": 85049, "epoch": 955, "lr": 7.783723171316322e-05}
{"train_loss": 0.28305870294570923, "global_step": 85050, "epoch": 955, "lr": 7.783675013354438e-05}
{"train_loss": 0.14999403059482574, "global_step": 85051, "epoch": 955, "lr": 7.783626855018322e-05}
{"train_loss": 0.13497856259346008, "global_step": 85052, "epoch": 955, "lr": 7.783578696307981e-05}
{"train_loss": 0.2165367752313614, "global_step": 85053, "epoch": 955, "lr": 7.783530537223423e-05}
{"train_loss": 0.19193553924560547, "global_step": 85054, "epoch": 955, "lr": 7.783482377764652e-05}
{"train_loss": 0.16173788905143738, "global_step": 85055, "epoch": 955, "lr": 7.783434217931676e-05}
{"train_loss": 0.16178619861602783, "global_step": 85056, "epoch": 955, "lr": 7.783386057724502e-05}
{"train_loss": 0.1471896767616272, "global_step": 85057, "epoch": 955, "lr": 7.783337897143134e-05}
{"train_loss": 0.23702378571033478, "global_step": 85058, "epoch": 955, "lr": 7.783289736187582e-05}
{"train_loss": 0.31652647256851196, "global_step": 85059, "epoch": 955, "lr": 7.783241574857849e-05}
{"train_loss": 0.24911274015903473, "global_step": 85060, "epoch": 955, "lr": 7.783193413153944e-05}
{"train_loss": 0.22036783397197723, "global_step": 85061, "epoch": 955, "lr": 7.783145251075872e-05}
{"train_loss": 0.23787952959537506, "global_step": 85062, "epoch": 955, "lr": 7.783097088623642e-05}
{"train_loss": 0.20045457780361176, "global_step": 85063, "epoch": 955, "lr": 7.783048925797256e-05}
{"train_loss": 0.1571357101202011, "global_step": 85064, "epoch": 955, "lr": 7.783000762596723e-05}
{"train_loss": 0.27844366431236267, "global_step": 85065, "epoch": 955, "lr": 7.782952599022051e-05}
{"train_loss": 0.29510897397994995, "global_step": 85066, "epoch": 955, "lr": 7.782904435073245e-05}
{"train_loss": 0.19850103557109833, "global_step": 85067, "epoch": 955, "lr": 7.782856270750311e-05}
{"train_loss": 0.20817476511001587, "global_step": 85068, "epoch": 955, "lr": 7.782808106053257e-05}
{"train_loss": 0.24452705681324005, "global_step": 85069, "epoch": 955, "lr": 7.782759940982085e-05}
{"train_loss": 0.22999821603298187, "global_step": 85070, "epoch": 955, "lr": 7.782711775536809e-05}
{"train_loss": 0.23380553722381592, "global_step": 85071, "epoch": 955, "lr": 7.782663609717429e-05}
{"train_loss": 0.18493197858333588, "global_step": 85072, "epoch": 955, "lr": 7.782615443523952e-05}
{"train_loss": 0.3351600468158722, "global_step": 85073, "epoch": 955, "lr": 7.782567276956389e-05}
{"train_loss": 0.24330006539821625, "global_step": 85074, "epoch": 955, "lr": 7.782519110014743e-05}
{"train_loss": 0.16214746236801147, "global_step": 85075, "epoch": 955, "lr": 7.782470942699021e-05}
{"train_loss": 0.23787693679332733, "global_step": 85076, "epoch": 955, "lr": 7.782422775009231e-05}
{"train_loss": 0.24617771804332733, "global_step": 85077, "epoch": 955, "lr": 7.782374606945376e-05}
{"train_loss": 0.21705873310565948, "global_step": 85078, "epoch": 955, "lr": 7.782326438507465e-05}
{"train_loss": 0.17708532512187958, "global_step": 85079, "epoch": 955, "lr": 7.782278269695506e-05}
{"train_loss": 0.13649578392505646, "global_step": 85080, "epoch": 955, "lr": 7.782230100509502e-05}
{"train_loss": 0.16936929523944855, "global_step": 85081, "epoch": 955, "lr": 7.782181930949461e-05}
{"train_loss": 0.21711446344852448, "global_step": 85082, "epoch": 955, "lr": 7.782133761015392e-05}
{"train_loss": 0.19274426703707556, "global_step": 85083, "epoch": 955, "lr": 7.782085590707297e-05, "val_loss": 4.067564964294434, "train_action_mse_error": 16.731212615966797}
{"train_loss": 0.2042037546634674, "global_step": 85084, "epoch": 956, "lr": 7.782037420025184e-05}
{"train_loss": 0.19135010242462158, "global_step": 85085, "epoch": 956, "lr": 7.78198924896906e-05}
{"train_loss": 0.23055188357830048, "global_step": 85086, "epoch": 956, "lr": 7.781941077538934e-05}
{"train_loss": 0.18227557837963104, "global_step": 85087, "epoch": 956, "lr": 7.781892905734807e-05}
{"train_loss": 0.18533743917942047, "global_step": 85088, "epoch": 956, "lr": 7.78184473355669e-05}
{"train_loss": 0.2212657928466797, "global_step": 85089, "epoch": 956, "lr": 7.781796561004586e-05}
{"train_loss": 0.15285161137580872, "global_step": 85090, "epoch": 956, "lr": 7.781748388078505e-05}
{"train_loss": 0.14349524676799774, "global_step": 85091, "epoch": 956, "lr": 7.781700214778452e-05}
{"train_loss": 0.17165875434875488, "global_step": 85092, "epoch": 956, "lr": 7.781652041104432e-05}
{"train_loss": 0.17271612584590912, "global_step": 85093, "epoch": 956, "lr": 7.781603867056454e-05}
{"train_loss": 0.13161446154117584, "global_step": 85094, "epoch": 956, "lr": 7.781555692634524e-05}
{"train_loss": 0.17827872931957245, "global_step": 85095, "epoch": 956, "lr": 7.781507517838645e-05}
{"train_loss": 0.1769818365573883, "global_step": 85096, "epoch": 956, "lr": 7.781459342668829e-05}
{"train_loss": 0.1852102428674698, "global_step": 85097, "epoch": 956, "lr": 7.781411167125078e-05}
{"train_loss": 0.19843821227550507, "global_step": 85098, "epoch": 956, "lr": 7.781362991207401e-05}
{"train_loss": 0.1812872439622879, "global_step": 85099, "epoch": 956, "lr": 7.781314814915803e-05}
{"train_loss": 0.19391506910324097, "global_step": 85100, "epoch": 956, "lr": 7.781266638250291e-05}
{"train_loss": 0.22120648622512817, "global_step": 85101, "epoch": 956, "lr": 7.781218461210873e-05}
{"train_loss": 0.14763948321342468, "global_step": 85102, "epoch": 956, "lr": 7.781170283797551e-05}
{"train_loss": 0.2295466661453247, "global_step": 85103, "epoch": 956, "lr": 7.781122106010339e-05}
{"train_loss": 0.22454118728637695, "global_step": 85104, "epoch": 956, "lr": 7.781073927849236e-05}
{"train_loss": 0.22157634794712067, "global_step": 85105, "epoch": 956, "lr": 7.781025749314252e-05}
{"train_loss": 0.4235121011734009, "global_step": 85106, "epoch": 956, "lr": 7.780977570405394e-05}
{"train_loss": 0.17574140429496765, "global_step": 85107, "epoch": 956, "lr": 7.780929391122667e-05}
{"train_loss": 0.2520714998245239, "global_step": 85108, "epoch": 956, "lr": 7.780881211466077e-05}
{"train_loss": 0.1825987547636032, "global_step": 85109, "epoch": 956, "lr": 7.780833031435632e-05}
{"train_loss": 0.16696010529994965, "global_step": 85110, "epoch": 956, "lr": 7.780784851031337e-05}
{"train_loss": 0.16899745166301727, "global_step": 85111, "epoch": 956, "lr": 7.7807366702532e-05}
{"train_loss": 0.14286109805107117, "global_step": 85112, "epoch": 956, "lr": 7.780688489101226e-05}
{"train_loss": 0.17631679773330688, "global_step": 85113, "epoch": 956, "lr": 7.780640307575426e-05}
{"train_loss": 0.21078051626682281, "global_step": 85114, "epoch": 956, "lr": 7.780592125675799e-05}
{"train_loss": 0.26810672879219055, "global_step": 85115, "epoch": 956, "lr": 7.780543943402356e-05}
{"train_loss": 0.13349898159503937, "global_step": 85116, "epoch": 956, "lr": 7.780495760755104e-05}
{"train_loss": 0.20643894374370575, "global_step": 85117, "epoch": 956, "lr": 7.780447577734047e-05}
{"train_loss": 0.2149783819913864, "global_step": 85118, "epoch": 956, "lr": 7.780399394339193e-05}
{"train_loss": 0.1612463891506195, "global_step": 85119, "epoch": 956, "lr": 7.780351210570548e-05}
{"train_loss": 0.19360025227069855, "global_step": 85120, "epoch": 956, "lr": 7.780303026428121e-05}
{"train_loss": 0.17755834758281708, "global_step": 85121, "epoch": 956, "lr": 7.780254841911914e-05}
{"train_loss": 0.22641639411449432, "global_step": 85122, "epoch": 956, "lr": 7.780206657021936e-05}
{"train_loss": 0.1728973090648651, "global_step": 85123, "epoch": 956, "lr": 7.780158471758193e-05}
{"train_loss": 0.15623092651367188, "global_step": 85124, "epoch": 956, "lr": 7.780110286120691e-05}
{"train_loss": 0.22725428640842438, "global_step": 85125, "epoch": 956, "lr": 7.780062100109437e-05}
{"train_loss": 0.21944645047187805, "global_step": 85126, "epoch": 956, "lr": 7.78001391372444e-05}
{"train_loss": 0.12513385713100433, "global_step": 85127, "epoch": 956, "lr": 7.779965726965701e-05}
{"train_loss": 0.1429402083158493, "global_step": 85128, "epoch": 956, "lr": 7.77991753983323e-05}
{"train_loss": 0.23244544863700867, "global_step": 85129, "epoch": 956, "lr": 7.779869352327034e-05}
{"train_loss": 0.16517020761966705, "global_step": 85130, "epoch": 956, "lr": 7.779821164447117e-05}
{"train_loss": 0.20047684013843536, "global_step": 85131, "epoch": 956, "lr": 7.77977297619349e-05}
{"train_loss": 0.23725390434265137, "global_step": 85132, "epoch": 956, "lr": 7.779724787566154e-05}
{"train_loss": 0.16750317811965942, "global_step": 85133, "epoch": 956, "lr": 7.779676598565118e-05}
{"train_loss": 0.11784347891807556, "global_step": 85134, "epoch": 956, "lr": 7.779628409190389e-05}
{"train_loss": 0.10594773292541504, "global_step": 85135, "epoch": 956, "lr": 7.779580219441972e-05}
{"train_loss": 0.16870397329330444, "global_step": 85136, "epoch": 956, "lr": 7.779532029319875e-05}
{"train_loss": 0.22398187220096588, "global_step": 85137, "epoch": 956, "lr": 7.779483838824103e-05}
{"train_loss": 0.20584720373153687, "global_step": 85138, "epoch": 956, "lr": 7.779435647954665e-05}
{"train_loss": 0.2343742698431015, "global_step": 85139, "epoch": 956, "lr": 7.779387456711565e-05}
{"train_loss": 0.1134105995297432, "global_step": 85140, "epoch": 956, "lr": 7.779339265094808e-05}
{"train_loss": 0.13940829038619995, "global_step": 85141, "epoch": 956, "lr": 7.779291073104405e-05}
{"train_loss": 0.12840662896633148, "global_step": 85142, "epoch": 956, "lr": 7.779242880740361e-05}
{"train_loss": 0.2144133746623993, "global_step": 85143, "epoch": 956, "lr": 7.77919468800268e-05}
{"train_loss": 0.22537194192409515, "global_step": 85144, "epoch": 956, "lr": 7.77914649489137e-05}
{"train_loss": 0.17411139607429504, "global_step": 85145, "epoch": 956, "lr": 7.779098301406438e-05}
{"train_loss": 0.21258145570755005, "global_step": 85146, "epoch": 956, "lr": 7.779050107547889e-05}
{"train_loss": 0.2627118229866028, "global_step": 85147, "epoch": 956, "lr": 7.779001913315734e-05}
{"train_loss": 0.18408440053462982, "global_step": 85148, "epoch": 956, "lr": 7.778953718709973e-05}
{"train_loss": 0.15720319747924805, "global_step": 85149, "epoch": 956, "lr": 7.778905523730615e-05}
{"train_loss": 0.24265621602535248, "global_step": 85150, "epoch": 956, "lr": 7.778857328377669e-05}
{"train_loss": 0.2270158976316452, "global_step": 85151, "epoch": 956, "lr": 7.77880913265114e-05}
{"train_loss": 0.22200243175029755, "global_step": 85152, "epoch": 956, "lr": 7.778760936551033e-05}
{"train_loss": 0.1723831743001938, "global_step": 85153, "epoch": 956, "lr": 7.778712740077356e-05}
{"train_loss": 0.16410741209983826, "global_step": 85154, "epoch": 956, "lr": 7.778664543230114e-05}
{"train_loss": 0.18090836703777313, "global_step": 85155, "epoch": 956, "lr": 7.778616346009315e-05}
{"train_loss": 0.15761759877204895, "global_step": 85156, "epoch": 956, "lr": 7.778568148414964e-05}
{"train_loss": 0.17378678917884827, "global_step": 85157, "epoch": 956, "lr": 7.77851995044707e-05}
{"train_loss": 0.23279404640197754, "global_step": 85158, "epoch": 956, "lr": 7.778471752105636e-05}
{"train_loss": 0.15709421038627625, "global_step": 85159, "epoch": 956, "lr": 7.778423553390671e-05}
{"train_loss": 0.20689889788627625, "global_step": 85160, "epoch": 956, "lr": 7.778375354302181e-05}
{"train_loss": 0.18062111735343933, "global_step": 85161, "epoch": 956, "lr": 7.778327154840172e-05}
{"train_loss": 0.11773727834224701, "global_step": 85162, "epoch": 956, "lr": 7.778278955004653e-05}
{"train_loss": 0.1634201556444168, "global_step": 85163, "epoch": 956, "lr": 7.778230754795625e-05}
{"train_loss": 0.1483384221792221, "global_step": 85164, "epoch": 956, "lr": 7.7781825542131e-05}
{"train_loss": 0.22674047946929932, "global_step": 85165, "epoch": 956, "lr": 7.77813435325708e-05}
{"train_loss": 0.18599089980125427, "global_step": 85166, "epoch": 956, "lr": 7.778086151927575e-05}
{"train_loss": 0.21089962124824524, "global_step": 85167, "epoch": 956, "lr": 7.77803795022459e-05}
{"train_loss": 0.1468781679868698, "global_step": 85168, "epoch": 956, "lr": 7.777989748148133e-05}
{"train_loss": 0.21989966928958893, "global_step": 85169, "epoch": 956, "lr": 7.777941545698209e-05}
{"train_loss": 0.16277459263801575, "global_step": 85170, "epoch": 956, "lr": 7.777893342874822e-05}
{"train_loss": 0.16777978837490082, "global_step": 85171, "epoch": 956, "lr": 7.777845139677984e-05}
{"train_loss": 0.1886663002411971, "global_step": 85172, "epoch": 956, "lr": 7.777796936107697e-05, "val_loss": 4.315785884857178}
{"train_loss": 0.19032570719718933, "global_step": 85173, "epoch": 957, "lr": 7.777748732163969e-05}
{"train_loss": 0.10675396770238876, "global_step": 85174, "epoch": 957, "lr": 7.777700527846808e-05}
{"train_loss": 0.14815741777420044, "global_step": 85175, "epoch": 957, "lr": 7.777652323156217e-05}
{"train_loss": 0.2931908965110779, "global_step": 85176, "epoch": 957, "lr": 7.777604118092206e-05}
{"train_loss": 0.28063148260116577, "global_step": 85177, "epoch": 957, "lr": 7.777555912654779e-05}
{"train_loss": 0.21523559093475342, "global_step": 85178, "epoch": 957, "lr": 7.777507706843945e-05}
{"train_loss": 0.17440354824066162, "global_step": 85179, "epoch": 957, "lr": 7.777459500659707e-05}
{"train_loss": 0.1814521700143814, "global_step": 85180, "epoch": 957, "lr": 7.777411294102074e-05}
{"train_loss": 0.18840590119361877, "global_step": 85181, "epoch": 957, "lr": 7.777363087171052e-05}
{"train_loss": 0.2657022476196289, "global_step": 85182, "epoch": 957, "lr": 7.777314879866648e-05}
{"train_loss": 0.1499721258878708, "global_step": 85183, "epoch": 957, "lr": 7.777266672188868e-05}
{"train_loss": 0.15547102689743042, "global_step": 85184, "epoch": 957, "lr": 7.777218464137717e-05}
{"train_loss": 0.18455708026885986, "global_step": 85185, "epoch": 957, "lr": 7.777170255713203e-05}
{"train_loss": 0.1869935244321823, "global_step": 85186, "epoch": 957, "lr": 7.777122046915332e-05}
{"train_loss": 0.24862119555473328, "global_step": 85187, "epoch": 957, "lr": 7.777073837744114e-05}
{"train_loss": 0.24718092381954193, "global_step": 85188, "epoch": 957, "lr": 7.777025628199549e-05}
{"train_loss": 0.23853608965873718, "global_step": 85189, "epoch": 957, "lr": 7.776977418281648e-05}
{"train_loss": 0.2919234037399292, "global_step": 85190, "epoch": 957, "lr": 7.776929207990416e-05}
{"train_loss": 0.1879536509513855, "global_step": 85191, "epoch": 957, "lr": 7.77688099732586e-05}
{"train_loss": 0.2859039008617401, "global_step": 85192, "epoch": 957, "lr": 7.776832786287985e-05}
{"train_loss": 0.12407912313938141, "global_step": 85193, "epoch": 957, "lr": 7.776784574876799e-05}
{"train_loss": 0.20836381614208221, "global_step": 85194, "epoch": 957, "lr": 7.776736363092309e-05}
{"train_loss": 0.14315035939216614, "global_step": 85195, "epoch": 957, "lr": 7.776688150934521e-05}
{"train_loss": 0.18405064940452576, "global_step": 85196, "epoch": 957, "lr": 7.77663993840344e-05}
{"train_loss": 0.21951176226139069, "global_step": 85197, "epoch": 957, "lr": 7.776591725499074e-05}
{"train_loss": 0.17833617329597473, "global_step": 85198, "epoch": 957, "lr": 7.77654351222143e-05}
{"train_loss": 0.20166538655757904, "global_step": 85199, "epoch": 957, "lr": 7.776495298570512e-05}
{"train_loss": 0.23594486713409424, "global_step": 85200, "epoch": 957, "lr": 7.776447084546328e-05}
{"train_loss": 0.18878856301307678, "global_step": 85201, "epoch": 957, "lr": 7.776398870148886e-05}
{"train_loss": 0.1787584275007248, "global_step": 85202, "epoch": 957, "lr": 7.77635065537819e-05}
{"train_loss": 0.1537012904882431, "global_step": 85203, "epoch": 957, "lr": 7.776302440234247e-05}
{"train_loss": 0.26805126667022705, "global_step": 85204, "epoch": 957, "lr": 7.776254224717066e-05}
{"train_loss": 0.18781939148902893, "global_step": 85205, "epoch": 957, "lr": 7.77620600882665e-05}
{"train_loss": 0.2946227192878723, "global_step": 85206, "epoch": 957, "lr": 7.776157792563007e-05}
{"train_loss": 0.1624860316514969, "global_step": 85207, "epoch": 957, "lr": 7.776109575926143e-05}
{"train_loss": 0.20828013122081757, "global_step": 85208, "epoch": 957, "lr": 7.776061358916067e-05}
{"train_loss": 0.22128315269947052, "global_step": 85209, "epoch": 957, "lr": 7.776013141532781e-05}
{"train_loss": 0.17632625997066498, "global_step": 85210, "epoch": 957, "lr": 7.775964923776296e-05}
{"train_loss": 0.1547241359949112, "global_step": 85211, "epoch": 957, "lr": 7.775916705646614e-05}
{"train_loss": 0.2344433069229126, "global_step": 85212, "epoch": 957, "lr": 7.775868487143744e-05}
{"train_loss": 0.23887383937835693, "global_step": 85213, "epoch": 957, "lr": 7.775820268267695e-05}
{"train_loss": 0.22738249599933624, "global_step": 85214, "epoch": 957, "lr": 7.775772049018468e-05}
{"train_loss": 0.2426978349685669, "global_step": 85215, "epoch": 957, "lr": 7.775723829396073e-05}
{"train_loss": 0.14181043207645416, "global_step": 85216, "epoch": 957, "lr": 7.775675609400517e-05}
{"train_loss": 0.2137712836265564, "global_step": 85217, "epoch": 957, "lr": 7.775627389031804e-05}
{"train_loss": 0.24368168413639069, "global_step": 85218, "epoch": 957, "lr": 7.775579168289942e-05}
{"train_loss": 0.19140881299972534, "global_step": 85219, "epoch": 957, "lr": 7.775530947174938e-05}
{"train_loss": 0.20370757579803467, "global_step": 85220, "epoch": 957, "lr": 7.775482725686796e-05}
{"train_loss": 0.19036242365837097, "global_step": 85221, "epoch": 957, "lr": 7.775434503825525e-05}
{"train_loss": 0.23569966852664948, "global_step": 85222, "epoch": 957, "lr": 7.77538628159113e-05}
{"train_loss": 0.2324237823486328, "global_step": 85223, "epoch": 957, "lr": 7.775338058983619e-05}
{"train_loss": 0.19706755876541138, "global_step": 85224, "epoch": 957, "lr": 7.775289836002999e-05}
{"train_loss": 0.13608071208000183, "global_step": 85225, "epoch": 957, "lr": 7.775241612649273e-05}
{"train_loss": 0.20628704130649567, "global_step": 85226, "epoch": 957, "lr": 7.77519338892245e-05}
{"train_loss": 0.21375812590122223, "global_step": 85227, "epoch": 957, "lr": 7.775145164822536e-05}
{"train_loss": 0.14688727259635925, "global_step": 85228, "epoch": 957, "lr": 7.775096940349538e-05}
{"train_loss": 0.21044543385505676, "global_step": 85229, "epoch": 957, "lr": 7.775048715503463e-05}
{"train_loss": 0.22688652575016022, "global_step": 85230, "epoch": 957, "lr": 7.775000490284314e-05}
{"train_loss": 0.13011056184768677, "global_step": 85231, "epoch": 957, "lr": 7.774952264692103e-05}
{"train_loss": 0.25671130418777466, "global_step": 85232, "epoch": 957, "lr": 7.774904038726831e-05}
{"train_loss": 0.20758502185344696, "global_step": 85233, "epoch": 957, "lr": 7.774855812388508e-05}
{"train_loss": 0.20986196398735046, "global_step": 85234, "epoch": 957, "lr": 7.77480758567714e-05}
{"train_loss": 0.2755230963230133, "global_step": 85235, "epoch": 957, "lr": 7.774759358592732e-05}
{"train_loss": 0.18396379053592682, "global_step": 85236, "epoch": 957, "lr": 7.774711131135292e-05}
{"train_loss": 0.1657101809978485, "global_step": 85237, "epoch": 957, "lr": 7.774662903304826e-05}
{"train_loss": 0.17906704545021057, "global_step": 85238, "epoch": 957, "lr": 7.77461467510134e-05}
{"train_loss": 0.2724129259586334, "global_step": 85239, "epoch": 957, "lr": 7.77456644652484e-05}
{"train_loss": 0.20052506029605865, "global_step": 85240, "epoch": 957, "lr": 7.774518217575335e-05}
{"train_loss": 0.10535991191864014, "global_step": 85241, "epoch": 957, "lr": 7.774469988252828e-05}
{"train_loss": 0.17127040028572083, "global_step": 85242, "epoch": 957, "lr": 7.774421758557329e-05}
{"train_loss": 0.15684008598327637, "global_step": 85243, "epoch": 957, "lr": 7.774373528488841e-05}
{"train_loss": 0.23751944303512573, "global_step": 85244, "epoch": 957, "lr": 7.774325298047375e-05}
{"train_loss": 0.18674157559871674, "global_step": 85245, "epoch": 957, "lr": 7.774277067232934e-05}
{"train_loss": 0.21353012323379517, "global_step": 85246, "epoch": 957, "lr": 7.774228836045523e-05}
{"train_loss": 0.22693026065826416, "global_step": 85247, "epoch": 957, "lr": 7.774180604485152e-05}
{"train_loss": 0.17919211089611053, "global_step": 85248, "epoch": 957, "lr": 7.774132372551827e-05}
{"train_loss": 0.146226704120636, "global_step": 85249, "epoch": 957, "lr": 7.774084140245555e-05}
{"train_loss": 0.15844565629959106, "global_step": 85250, "epoch": 957, "lr": 7.774035907566338e-05}
{"train_loss": 0.1992989480495453, "global_step": 85251, "epoch": 957, "lr": 7.773987674514187e-05}
{"train_loss": 0.15920086205005646, "global_step": 85252, "epoch": 957, "lr": 7.773939441089108e-05}
{"train_loss": 0.15089793503284454, "global_step": 85253, "epoch": 957, "lr": 7.773891207291106e-05}
{"train_loss": 0.16590538620948792, "global_step": 85254, "epoch": 957, "lr": 7.773842973120188e-05}
{"train_loss": 0.17919312417507172, "global_step": 85255, "epoch": 957, "lr": 7.77379473857636e-05}
{"train_loss": 0.19114860892295837, "global_step": 85256, "epoch": 957, "lr": 7.77374650365963e-05}
{"train_loss": 0.19187068939208984, "global_step": 85257, "epoch": 957, "lr": 7.773698268370003e-05}
{"train_loss": 0.16858167946338654, "global_step": 85258, "epoch": 957, "lr": 7.773650032707487e-05}
{"train_loss": 0.11613243818283081, "global_step": 85259, "epoch": 957, "lr": 7.773601796672087e-05}
{"train_loss": 0.22382231056690216, "global_step": 85260, "epoch": 957, "lr": 7.77355356026381e-05}
{"train_loss": 0.19782126828860702, "global_step": 85261, "epoch": 957, "lr": 7.773505323482662e-05, "val_loss": 4.0498175621032715}
{"train_loss": 0.1818748563528061, "global_step": 85262, "epoch": 958, "lr": 7.77345708632865e-05}
{"train_loss": 0.1709822118282318, "global_step": 85263, "epoch": 958, "lr": 7.773408848801781e-05}
{"train_loss": 0.19173666834831238, "global_step": 85264, "epoch": 958, "lr": 7.773360610902064e-05}
{"train_loss": 0.14037422835826874, "global_step": 85265, "epoch": 958, "lr": 7.773312372629498e-05}
{"train_loss": 0.2419215440750122, "global_step": 85266, "epoch": 958, "lr": 7.773264133984096e-05}
{"train_loss": 0.2026856690645218, "global_step": 85267, "epoch": 958, "lr": 7.773215894965862e-05}
{"train_loss": 0.1115180030465126, "global_step": 85268, "epoch": 958, "lr": 7.773167655574802e-05}
{"train_loss": 0.16159263253211975, "global_step": 85269, "epoch": 958, "lr": 7.773119415810926e-05}
{"train_loss": 0.10575047135353088, "global_step": 85270, "epoch": 958, "lr": 7.773071175674236e-05}
{"train_loss": 0.1772194802761078, "global_step": 85271, "epoch": 958, "lr": 7.77302293516474e-05}
{"train_loss": 0.1638573259115219, "global_step": 85272, "epoch": 958, "lr": 7.772974694282446e-05}
{"train_loss": 0.1563234180212021, "global_step": 85273, "epoch": 958, "lr": 7.772926453027357e-05}
{"train_loss": 0.15750859677791595, "global_step": 85274, "epoch": 958, "lr": 7.772878211399485e-05}
{"train_loss": 0.16300441324710846, "global_step": 85275, "epoch": 958, "lr": 7.772829969398832e-05}
{"train_loss": 0.21049003303050995, "global_step": 85276, "epoch": 958, "lr": 7.772781727025405e-05}
{"train_loss": 0.250945121049881, "global_step": 85277, "epoch": 958, "lr": 7.772733484279212e-05}
{"train_loss": 0.24563804268836975, "global_step": 85278, "epoch": 958, "lr": 7.772685241160259e-05}
{"train_loss": 0.17167973518371582, "global_step": 85279, "epoch": 958, "lr": 7.772636997668553e-05}
{"train_loss": 0.1538902223110199, "global_step": 85280, "epoch": 958, "lr": 7.772588753804098e-05}
{"train_loss": 0.18925587832927704, "global_step": 85281, "epoch": 958, "lr": 7.772540509566904e-05}
{"train_loss": 0.17960424721240997, "global_step": 85282, "epoch": 958, "lr": 7.772492264956975e-05}
{"train_loss": 0.1837025284767151, "global_step": 85283, "epoch": 958, "lr": 7.772444019974317e-05}
{"train_loss": 0.17867016792297363, "global_step": 85284, "epoch": 958, "lr": 7.772395774618939e-05}
{"train_loss": 0.2064155489206314, "global_step": 85285, "epoch": 958, "lr": 7.772347528890845e-05}
{"train_loss": 0.17381511628627777, "global_step": 85286, "epoch": 958, "lr": 7.772299282790044e-05}
{"train_loss": 0.2356109917163849, "global_step": 85287, "epoch": 958, "lr": 7.772251036316541e-05}
{"train_loss": 0.09647568315267563, "global_step": 85288, "epoch": 958, "lr": 7.772202789470342e-05}
{"train_loss": 0.1681220978498459, "global_step": 85289, "epoch": 958, "lr": 7.772154542251455e-05}
{"train_loss": 0.2129748910665512, "global_step": 85290, "epoch": 958, "lr": 7.772106294659884e-05}
{"train_loss": 0.1879350244998932, "global_step": 85291, "epoch": 958, "lr": 7.77205804669564e-05}
{"train_loss": 0.17425931990146637, "global_step": 85292, "epoch": 958, "lr": 7.772009798358724e-05}
{"train_loss": 0.13909552991390228, "global_step": 85293, "epoch": 958, "lr": 7.771961549649145e-05}
{"train_loss": 0.16403110325336456, "global_step": 85294, "epoch": 958, "lr": 7.77191330056691e-05}
{"train_loss": 0.19902387261390686, "global_step": 85295, "epoch": 958, "lr": 7.771865051112026e-05}
{"train_loss": 0.2137976884841919, "global_step": 85296, "epoch": 958, "lr": 7.771816801284497e-05}
{"train_loss": 0.13261407613754272, "global_step": 85297, "epoch": 958, "lr": 7.771768551084334e-05}
{"train_loss": 0.24304665625095367, "global_step": 85298, "epoch": 958, "lr": 7.771720300511538e-05}
{"train_loss": 0.21872252225875854, "global_step": 85299, "epoch": 958, "lr": 7.771672049566118e-05}
{"train_loss": 0.2320285439491272, "global_step": 85300, "epoch": 958, "lr": 7.77162379824808e-05}
{"train_loss": 0.16635826230049133, "global_step": 85301, "epoch": 958, "lr": 7.771575546557432e-05}
{"train_loss": 0.171123206615448, "global_step": 85302, "epoch": 958, "lr": 7.771527294494179e-05}
{"train_loss": 0.17549073696136475, "global_step": 85303, "epoch": 958, "lr": 7.77147904205833e-05}
{"train_loss": 0.19042757153511047, "global_step": 85304, "epoch": 958, "lr": 7.771430789249887e-05}
{"train_loss": 0.19520194828510284, "global_step": 85305, "epoch": 958, "lr": 7.77138253606886e-05}
{"train_loss": 0.17476683855056763, "global_step": 85306, "epoch": 958, "lr": 7.771334282515255e-05}
{"train_loss": 0.18354180455207825, "global_step": 85307, "epoch": 958, "lr": 7.771286028589076e-05}
{"train_loss": 0.12808725237846375, "global_step": 85308, "epoch": 958, "lr": 7.771237774290331e-05}
{"train_loss": 0.19993725419044495, "global_step": 85309, "epoch": 958, "lr": 7.771189519619029e-05}
{"train_loss": 0.21800880134105682, "global_step": 85310, "epoch": 958, "lr": 7.771141264575173e-05}
{"train_loss": 0.15038585662841797, "global_step": 85311, "epoch": 958, "lr": 7.77109300915877e-05}
{"train_loss": 0.2063830941915512, "global_step": 85312, "epoch": 958, "lr": 7.771044753369829e-05}
{"train_loss": 0.12391189485788345, "global_step": 85313, "epoch": 958, "lr": 7.770996497208356e-05}
{"train_loss": 0.18842551112174988, "global_step": 85314, "epoch": 958, "lr": 7.770948240674354e-05}
{"train_loss": 0.13141146302223206, "global_step": 85315, "epoch": 958, "lr": 7.770899983767832e-05}
{"train_loss": 0.271479070186615, "global_step": 85316, "epoch": 958, "lr": 7.770851726488797e-05}
{"train_loss": 0.2387971729040146, "global_step": 85317, "epoch": 958, "lr": 7.770803468837256e-05}
{"train_loss": 0.1965343952178955, "global_step": 85318, "epoch": 958, "lr": 7.770755210813211e-05}
{"train_loss": 0.17008192837238312, "global_step": 85319, "epoch": 958, "lr": 7.770706952416674e-05}
{"train_loss": 0.2611638903617859, "global_step": 85320, "epoch": 958, "lr": 7.770658693647649e-05}
{"train_loss": 0.22111251950263977, "global_step": 85321, "epoch": 958, "lr": 7.770610434506143e-05}
{"train_loss": 0.21640753746032715, "global_step": 85322, "epoch": 958, "lr": 7.770562174992161e-05}
{"train_loss": 0.07307546585798264, "global_step": 85323, "epoch": 958, "lr": 7.770513915105711e-05}
{"train_loss": 0.14908337593078613, "global_step": 85324, "epoch": 958, "lr": 7.770465654846799e-05}
{"train_loss": 0.22007404267787933, "global_step": 85325, "epoch": 958, "lr": 7.770417394215434e-05}
{"train_loss": 0.2029944658279419, "global_step": 85326, "epoch": 958, "lr": 7.770369133211617e-05}
{"train_loss": 0.18886713683605194, "global_step": 85327, "epoch": 958, "lr": 7.770320871835359e-05}
{"train_loss": 0.24768079817295074, "global_step": 85328, "epoch": 958, "lr": 7.770272610086664e-05}
{"train_loss": 0.23349207639694214, "global_step": 85329, "epoch": 958, "lr": 7.77022434796554e-05}
{"train_loss": 0.19955934584140778, "global_step": 85330, "epoch": 958, "lr": 7.770176085471995e-05}
{"train_loss": 0.22536179423332214, "global_step": 85331, "epoch": 958, "lr": 7.770127822606032e-05}
{"train_loss": 0.1508626639842987, "global_step": 85332, "epoch": 958, "lr": 7.77007955936766e-05}
{"train_loss": 0.201912060379982, "global_step": 85333, "epoch": 958, "lr": 7.770031295756884e-05}
{"train_loss": 0.2893911302089691, "global_step": 85334, "epoch": 958, "lr": 7.769983031773711e-05}
{"train_loss": 0.24490314722061157, "global_step": 85335, "epoch": 958, "lr": 7.769934767418148e-05}
{"train_loss": 0.1418657749891281, "global_step": 85336, "epoch": 958, "lr": 7.7698865026902e-05}
{"train_loss": 0.19686609506607056, "global_step": 85337, "epoch": 958, "lr": 7.769838237589875e-05}
{"train_loss": 0.17261479794979095, "global_step": 85338, "epoch": 958, "lr": 7.769789972117179e-05}
{"train_loss": 0.14080817997455597, "global_step": 85339, "epoch": 958, "lr": 7.76974170627212e-05}
{"train_loss": 0.1321466863155365, "global_step": 85340, "epoch": 958, "lr": 7.769693440054702e-05}
{"train_loss": 0.14427368342876434, "global_step": 85341, "epoch": 958, "lr": 7.769645173464932e-05}
{"train_loss": 0.27481505274772644, "global_step": 85342, "epoch": 958, "lr": 7.769596906502817e-05}
{"train_loss": 0.17146092653274536, "global_step": 85343, "epoch": 958, "lr": 7.769548639168362e-05}
{"train_loss": 0.19663622975349426, "global_step": 85344, "epoch": 958, "lr": 7.769500371461578e-05}
{"train_loss": 0.13035549223423004, "global_step": 85345, "epoch": 958, "lr": 7.769452103382467e-05}
{"train_loss": 0.22661471366882324, "global_step": 85346, "epoch": 958, "lr": 7.769403834931035e-05}
{"train_loss": 0.13736414909362793, "global_step": 85347, "epoch": 958, "lr": 7.769355566107293e-05}
{"train_loss": 0.22351117432117462, "global_step": 85348, "epoch": 958, "lr": 7.769307296911243e-05}
{"train_loss": 0.13797713816165924, "global_step": 85349, "epoch": 958, "lr": 7.769259027342895e-05}
{"train_loss": 0.18506729586070844, "global_step": 85350, "epoch": 958, "lr": 7.769210757402253e-05, "val_loss": 4.19504451751709}
{"train_loss": 0.19234231114387512, "global_step": 85351, "epoch": 959, "lr": 7.769162487089325e-05}
{"train_loss": 0.24388355016708374, "global_step": 85352, "epoch": 959, "lr": 7.769114216404117e-05}
{"train_loss": 0.1192011684179306, "global_step": 85353, "epoch": 959, "lr": 7.769065945346636e-05}
{"train_loss": 0.15809422731399536, "global_step": 85354, "epoch": 959, "lr": 7.769017673916886e-05}
{"train_loss": 0.2153434008359909, "global_step": 85355, "epoch": 959, "lr": 7.768969402114876e-05}
{"train_loss": 0.2456093728542328, "global_step": 85356, "epoch": 959, "lr": 7.768921129940612e-05}
{"train_loss": 0.15461546182632446, "global_step": 85357, "epoch": 959, "lr": 7.768872857394101e-05}
{"train_loss": 0.17259906232357025, "global_step": 85358, "epoch": 959, "lr": 7.768824584475347e-05}
{"train_loss": 0.21930445730686188, "global_step": 85359, "epoch": 959, "lr": 7.76877631118436e-05}
{"train_loss": 0.11998217552900314, "global_step": 85360, "epoch": 959, "lr": 7.768728037521144e-05}
{"train_loss": 0.22753332555294037, "global_step": 85361, "epoch": 959, "lr": 7.768679763485706e-05}
{"train_loss": 0.18390272557735443, "global_step": 85362, "epoch": 959, "lr": 7.768631489078053e-05}
{"train_loss": 0.1582174003124237, "global_step": 85363, "epoch": 959, "lr": 7.768583214298191e-05}
{"train_loss": 0.22222360968589783, "global_step": 85364, "epoch": 959, "lr": 7.768534939146126e-05}
{"train_loss": 0.15600883960723877, "global_step": 85365, "epoch": 959, "lr": 7.768486663621868e-05}
{"train_loss": 0.17965562641620636, "global_step": 85366, "epoch": 959, "lr": 7.768438387725419e-05}
{"train_loss": 0.0992501899600029, "global_step": 85367, "epoch": 959, "lr": 7.768390111456787e-05}
{"train_loss": 0.132229283452034, "global_step": 85368, "epoch": 959, "lr": 7.768341834815978e-05}
{"train_loss": 0.24624961614608765, "global_step": 85369, "epoch": 959, "lr": 7.768293557803e-05}
{"train_loss": 0.1794988065958023, "global_step": 85370, "epoch": 959, "lr": 7.768245280417858e-05}
{"train_loss": 0.18410509824752808, "global_step": 85371, "epoch": 959, "lr": 7.768197002660561e-05}
{"train_loss": 0.20429620146751404, "global_step": 85372, "epoch": 959, "lr": 7.768148724531112e-05}
{"train_loss": 0.22540345788002014, "global_step": 85373, "epoch": 959, "lr": 7.76810044602952e-05}
{"train_loss": 0.13455726206302643, "global_step": 85374, "epoch": 959, "lr": 7.768052167155791e-05}
{"train_loss": 0.16571927070617676, "global_step": 85375, "epoch": 959, "lr": 7.768003887909931e-05}
{"train_loss": 0.23490718007087708, "global_step": 85376, "epoch": 959, "lr": 7.767955608291946e-05}
{"train_loss": 0.2073078602552414, "global_step": 85377, "epoch": 959, "lr": 7.767907328301842e-05}
{"train_loss": 0.1813531517982483, "global_step": 85378, "epoch": 959, "lr": 7.767859047939628e-05}
{"train_loss": 0.2166825383901596, "global_step": 85379, "epoch": 959, "lr": 7.767810767205309e-05}
{"train_loss": 0.18963967263698578, "global_step": 85380, "epoch": 959, "lr": 7.76776248609889e-05}
{"train_loss": 0.2435227930545807, "global_step": 85381, "epoch": 959, "lr": 7.767714204620381e-05}
{"train_loss": 0.19861236214637756, "global_step": 85382, "epoch": 959, "lr": 7.767665922769787e-05}
{"train_loss": 0.1611040085554123, "global_step": 85383, "epoch": 959, "lr": 7.767617640547113e-05}
{"train_loss": 0.14403116703033447, "global_step": 85384, "epoch": 959, "lr": 7.767569357952367e-05}
{"train_loss": 0.08234727382659912, "global_step": 85385, "epoch": 959, "lr": 7.767521074985553e-05}
{"train_loss": 0.1528732180595398, "global_step": 85386, "epoch": 959, "lr": 7.767472791646681e-05}
{"train_loss": 0.1876266449689865, "global_step": 85387, "epoch": 959, "lr": 7.767424507935757e-05}
{"train_loss": 0.1902170032262802, "global_step": 85388, "epoch": 959, "lr": 7.767376223852786e-05}
{"train_loss": 0.15453393757343292, "global_step": 85389, "epoch": 959, "lr": 7.767327939397775e-05}
{"train_loss": 0.21381542086601257, "global_step": 85390, "epoch": 959, "lr": 7.76727965457073e-05}
{"train_loss": 0.18263563513755798, "global_step": 85391, "epoch": 959, "lr": 7.767231369371659e-05}
{"train_loss": 0.21102112531661987, "global_step": 85392, "epoch": 959, "lr": 7.767183083800566e-05}
{"train_loss": 0.2220405489206314, "global_step": 85393, "epoch": 959, "lr": 7.767134797857461e-05}
{"train_loss": 0.18833833932876587, "global_step": 85394, "epoch": 959, "lr": 7.767086511542345e-05}
{"train_loss": 0.1305123120546341, "global_step": 85395, "epoch": 959, "lr": 7.76703822485523e-05}
{"train_loss": 0.225521981716156, "global_step": 85396, "epoch": 959, "lr": 7.76698993779612e-05}
{"train_loss": 0.184273362159729, "global_step": 85397, "epoch": 959, "lr": 7.766941650365025e-05}
{"train_loss": 0.14996421337127686, "global_step": 85398, "epoch": 959, "lr": 7.766893362561944e-05}
{"train_loss": 0.13884802162647247, "global_step": 85399, "epoch": 959, "lr": 7.766845074386891e-05}
{"train_loss": 0.17120738327503204, "global_step": 85400, "epoch": 959, "lr": 7.766796785839868e-05}
{"train_loss": 0.21790140867233276, "global_step": 85401, "epoch": 959, "lr": 7.766748496920882e-05}
{"train_loss": 0.18466225266456604, "global_step": 85402, "epoch": 959, "lr": 7.76670020762994e-05}
{"train_loss": 0.15889374911785126, "global_step": 85403, "epoch": 959, "lr": 7.766651917967051e-05}
{"train_loss": 0.1423322856426239, "global_step": 85404, "epoch": 959, "lr": 7.766603627932218e-05}
{"train_loss": 0.17609266936779022, "global_step": 85405, "epoch": 959, "lr": 7.766555337525451e-05}
{"train_loss": 0.1651361584663391, "global_step": 85406, "epoch": 959, "lr": 7.766507046746751e-05}
{"train_loss": 0.16733701527118683, "global_step": 85407, "epoch": 959, "lr": 7.766458755596132e-05}
{"train_loss": 0.14579181373119354, "global_step": 85408, "epoch": 959, "lr": 7.766410464073593e-05}
{"train_loss": 0.16202905774116516, "global_step": 85409, "epoch": 959, "lr": 7.766362172179145e-05}
{"train_loss": 0.12751030921936035, "global_step": 85410, "epoch": 959, "lr": 7.766313879912794e-05}
{"train_loss": 0.20767752826213837, "global_step": 85411, "epoch": 959, "lr": 7.766265587274544e-05}
{"train_loss": 0.18297697603702545, "global_step": 85412, "epoch": 959, "lr": 7.766217294264405e-05}
{"train_loss": 0.25208595395088196, "global_step": 85413, "epoch": 959, "lr": 7.766169000882381e-05}
{"train_loss": 0.17904479801654816, "global_step": 85414, "epoch": 959, "lr": 7.766120707128479e-05}
{"train_loss": 0.27303051948547363, "global_step": 85415, "epoch": 959, "lr": 7.766072413002707e-05}
{"train_loss": 0.2916910648345947, "global_step": 85416, "epoch": 959, "lr": 7.766024118505069e-05}
{"train_loss": 0.1744251400232315, "global_step": 85417, "epoch": 959, "lr": 7.765975823635574e-05}
{"train_loss": 0.15803498029708862, "global_step": 85418, "epoch": 959, "lr": 7.765927528394226e-05}
{"train_loss": 0.20135125517845154, "global_step": 85419, "epoch": 959, "lr": 7.765879232781033e-05}
{"train_loss": 0.1951371729373932, "global_step": 85420, "epoch": 959, "lr": 7.765830936796001e-05}
{"train_loss": 0.15864115953445435, "global_step": 85421, "epoch": 959, "lr": 7.765782640439136e-05}
{"train_loss": 0.20292894542217255, "global_step": 85422, "epoch": 959, "lr": 7.765734343710448e-05}
{"train_loss": 0.21181105077266693, "global_step": 85423, "epoch": 959, "lr": 7.765686046609939e-05}
{"train_loss": 0.20604680478572845, "global_step": 85424, "epoch": 959, "lr": 7.765637749137617e-05}
{"train_loss": 0.2162148356437683, "global_step": 85425, "epoch": 959, "lr": 7.765589451293489e-05}
{"train_loss": 0.15522190928459167, "global_step": 85426, "epoch": 959, "lr": 7.765541153077562e-05}
{"train_loss": 0.1828102469444275, "global_step": 85427, "epoch": 959, "lr": 7.765492854489841e-05}
{"train_loss": 0.23597806692123413, "global_step": 85428, "epoch": 959, "lr": 7.765444555530332e-05}
{"train_loss": 0.17431871592998505, "global_step": 85429, "epoch": 959, "lr": 7.765396256199045e-05}
{"train_loss": 0.1586904227733612, "global_step": 85430, "epoch": 959, "lr": 7.765347956495983e-05}
{"train_loss": 0.24933955073356628, "global_step": 85431, "epoch": 959, "lr": 7.765299656421155e-05}
{"train_loss": 0.15634597837924957, "global_step": 85432, "epoch": 959, "lr": 7.765251355974564e-05}
{"train_loss": 0.15082186460494995, "global_step": 85433, "epoch": 959, "lr": 7.765203055156221e-05}
{"train_loss": 0.2414509803056717, "global_step": 85434, "epoch": 959, "lr": 7.765154753966128e-05}
{"train_loss": 0.21174411475658417, "global_step": 85435, "epoch": 959, "lr": 7.765106452404293e-05}
{"train_loss": 0.15061844885349274, "global_step": 85436, "epoch": 959, "lr": 7.765058150470725e-05}
{"train_loss": 0.12447986006736755, "global_step": 85437, "epoch": 959, "lr": 7.765009848165427e-05}
{"train_loss": 0.12864167988300323, "global_step": 85438, "epoch": 959, "lr": 7.764961545488409e-05}
{"train_loss": 0.18326462838756905, "global_step": 85439, "epoch": 959, "lr": 7.764913242439675e-05, "val_loss": 4.1280083656311035}
{"train_loss": 0.19933462142944336, "global_step": 85440, "epoch": 960, "lr": 7.764864939019232e-05}
{"train_loss": 0.15657468140125275, "global_step": 85441, "epoch": 960, "lr": 7.764816635227088e-05}
{"train_loss": 0.1487473100423813, "global_step": 85442, "epoch": 960, "lr": 7.764768331063245e-05}
{"train_loss": 0.18091332912445068, "global_step": 85443, "epoch": 960, "lr": 7.764720026527715e-05}
{"train_loss": 0.1656993180513382, "global_step": 85444, "epoch": 960, "lr": 7.764671721620503e-05}
{"train_loss": 0.20298470556735992, "global_step": 85445, "epoch": 960, "lr": 7.764623416341612e-05}
{"train_loss": 0.21145778894424438, "global_step": 85446, "epoch": 960, "lr": 7.764575110691052e-05}
{"train_loss": 0.14972977340221405, "global_step": 85447, "epoch": 960, "lr": 7.764526804668829e-05}
{"train_loss": 0.19232654571533203, "global_step": 85448, "epoch": 960, "lr": 7.764478498274947e-05}
{"train_loss": 0.21114742755889893, "global_step": 85449, "epoch": 960, "lr": 7.764430191509418e-05}
{"train_loss": 0.25153520703315735, "global_step": 85450, "epoch": 960, "lr": 7.764381884372243e-05}
{"train_loss": 0.14208266139030457, "global_step": 85451, "epoch": 960, "lr": 7.764333576863432e-05}
{"train_loss": 0.264685720205307, "global_step": 85452, "epoch": 960, "lr": 7.764285268982988e-05}
{"train_loss": 0.10724030435085297, "global_step": 85453, "epoch": 960, "lr": 7.764236960730921e-05}
{"train_loss": 0.2166309505701065, "global_step": 85454, "epoch": 960, "lr": 7.764188652107234e-05}
{"train_loss": 0.1601037085056305, "global_step": 85455, "epoch": 960, "lr": 7.764140343111936e-05}
{"train_loss": 0.14134481549263, "global_step": 85456, "epoch": 960, "lr": 7.764092033745036e-05}
{"train_loss": 0.18796080350875854, "global_step": 85457, "epoch": 960, "lr": 7.764043724006536e-05}
{"train_loss": 0.1882512867450714, "global_step": 85458, "epoch": 960, "lr": 7.763995413896443e-05}
{"train_loss": 0.1984911561012268, "global_step": 85459, "epoch": 960, "lr": 7.763947103414765e-05}
{"train_loss": 0.21759742498397827, "global_step": 85460, "epoch": 960, "lr": 7.763898792561508e-05}
{"train_loss": 0.1730158030986786, "global_step": 85461, "epoch": 960, "lr": 7.763850481336679e-05}
{"train_loss": 0.15167170763015747, "global_step": 85462, "epoch": 960, "lr": 7.763802169740283e-05}
{"train_loss": 0.13898490369319916, "global_step": 85463, "epoch": 960, "lr": 7.763753857772327e-05}
{"train_loss": 0.16397014260292053, "global_step": 85464, "epoch": 960, "lr": 7.76370554543282e-05}
{"train_loss": 0.13841992616653442, "global_step": 85465, "epoch": 960, "lr": 7.763657232721766e-05}
{"train_loss": 0.20796091854572296, "global_step": 85466, "epoch": 960, "lr": 7.763608919639171e-05}
{"train_loss": 0.2739342153072357, "global_step": 85467, "epoch": 960, "lr": 7.763560606185043e-05}
{"train_loss": 0.14638246595859528, "global_step": 85468, "epoch": 960, "lr": 7.763512292359388e-05}
{"train_loss": 0.17564299702644348, "global_step": 85469, "epoch": 960, "lr": 7.763463978162213e-05}
{"train_loss": 0.20250660181045532, "global_step": 85470, "epoch": 960, "lr": 7.763415663593523e-05}
{"train_loss": 0.17249226570129395, "global_step": 85471, "epoch": 960, "lr": 7.763367348653326e-05}
{"train_loss": 0.18561667203903198, "global_step": 85472, "epoch": 960, "lr": 7.763319033341627e-05}
{"train_loss": 0.10276943445205688, "global_step": 85473, "epoch": 960, "lr": 7.763270717658433e-05}
{"train_loss": 0.18703681230545044, "global_step": 85474, "epoch": 960, "lr": 7.763222401603752e-05}
{"train_loss": 0.19062617421150208, "global_step": 85475, "epoch": 960, "lr": 7.763174085177591e-05}
{"train_loss": 0.12869292497634888, "global_step": 85476, "epoch": 960, "lr": 7.763125768379953e-05}
{"train_loss": 0.17426368594169617, "global_step": 85477, "epoch": 960, "lr": 7.763077451210847e-05}
{"train_loss": 0.20026056468486786, "global_step": 85478, "epoch": 960, "lr": 7.763029133670278e-05}
{"train_loss": 0.24928748607635498, "global_step": 85479, "epoch": 960, "lr": 7.762980815758253e-05}
{"train_loss": 0.14906063675880432, "global_step": 85480, "epoch": 960, "lr": 7.762932497474782e-05}
{"train_loss": 0.2617396414279938, "global_step": 85481, "epoch": 960, "lr": 7.762884178819867e-05}
{"train_loss": 0.2142709195613861, "global_step": 85482, "epoch": 960, "lr": 7.762835859793514e-05}
{"train_loss": 0.2211182713508606, "global_step": 85483, "epoch": 960, "lr": 7.762787540395732e-05}
{"train_loss": 0.12198837101459503, "global_step": 85484, "epoch": 960, "lr": 7.762739220626529e-05}
{"train_loss": 0.14207136631011963, "global_step": 85485, "epoch": 960, "lr": 7.762690900485907e-05}
{"train_loss": 0.24878168106079102, "global_step": 85486, "epoch": 960, "lr": 7.762642579973876e-05}
{"train_loss": 0.16943861544132233, "global_step": 85487, "epoch": 960, "lr": 7.762594259090443e-05}
{"train_loss": 0.17533764243125916, "global_step": 85488, "epoch": 960, "lr": 7.76254593783561e-05}
{"train_loss": 0.17459774017333984, "global_step": 85489, "epoch": 960, "lr": 7.762497616209388e-05}
{"train_loss": 0.13285283744335175, "global_step": 85490, "epoch": 960, "lr": 7.762449294211782e-05}
{"train_loss": 0.17666231095790863, "global_step": 85491, "epoch": 960, "lr": 7.762400971842795e-05}
{"train_loss": 0.13201653957366943, "global_step": 85492, "epoch": 960, "lr": 7.76235264910244e-05}
{"train_loss": 0.22951477766036987, "global_step": 85493, "epoch": 960, "lr": 7.762304325990721e-05}
{"train_loss": 0.21814793348312378, "global_step": 85494, "epoch": 960, "lr": 7.762256002507643e-05}
{"train_loss": 0.19909170269966125, "global_step": 85495, "epoch": 960, "lr": 7.762207678653213e-05}
{"train_loss": 0.173728808760643, "global_step": 85496, "epoch": 960, "lr": 7.762159354427438e-05}
{"train_loss": 0.14486901462078094, "global_step": 85497, "epoch": 960, "lr": 7.762111029830324e-05}
{"train_loss": 0.16739317774772644, "global_step": 85498, "epoch": 960, "lr": 7.762062704861879e-05}
{"train_loss": 0.18425698578357697, "global_step": 85499, "epoch": 960, "lr": 7.762014379522107e-05}
{"train_loss": 0.175096333026886, "global_step": 85500, "epoch": 960, "lr": 7.761966053811016e-05}
{"train_loss": 0.29508981108665466, "global_step": 85501, "epoch": 960, "lr": 7.761917727728614e-05}
{"train_loss": 0.25519177317619324, "global_step": 85502, "epoch": 960, "lr": 7.761869401274903e-05}
{"train_loss": 0.19218964874744415, "global_step": 85503, "epoch": 960, "lr": 7.761821074449895e-05}
{"train_loss": 0.25975722074508667, "global_step": 85504, "epoch": 960, "lr": 7.761772747253593e-05}
{"train_loss": 0.1390407234430313, "global_step": 85505, "epoch": 960, "lr": 7.761724419686002e-05}
{"train_loss": 0.21173927187919617, "global_step": 85506, "epoch": 960, "lr": 7.761676091747135e-05}
{"train_loss": 0.16649596393108368, "global_step": 85507, "epoch": 960, "lr": 7.761627763436992e-05}
{"train_loss": 0.16702835261821747, "global_step": 85508, "epoch": 960, "lr": 7.761579434755581e-05}
{"train_loss": 0.13565050065517426, "global_step": 85509, "epoch": 960, "lr": 7.76153110570291e-05}
{"train_loss": 0.11539781093597412, "global_step": 85510, "epoch": 960, "lr": 7.761482776278984e-05}
{"train_loss": 0.25294357538223267, "global_step": 85511, "epoch": 960, "lr": 7.761434446483812e-05}
{"train_loss": 0.19498196244239807, "global_step": 85512, "epoch": 960, "lr": 7.761386116317398e-05}
{"train_loss": 0.17825524508953094, "global_step": 85513, "epoch": 960, "lr": 7.76133778577975e-05}
{"train_loss": 0.13629430532455444, "global_step": 85514, "epoch": 960, "lr": 7.761289454870872e-05}
{"train_loss": 0.1327463984489441, "global_step": 85515, "epoch": 960, "lr": 7.761241123590774e-05}
{"train_loss": 0.17876534163951874, "global_step": 85516, "epoch": 960, "lr": 7.76119279193946e-05}
{"train_loss": 0.15493716299533844, "global_step": 85517, "epoch": 960, "lr": 7.761144459916938e-05}
{"train_loss": 0.24200502038002014, "global_step": 85518, "epoch": 960, "lr": 7.761096127523213e-05}
{"train_loss": 0.11432375758886337, "global_step": 85519, "epoch": 960, "lr": 7.761047794758293e-05}
{"train_loss": 0.1633736491203308, "global_step": 85520, "epoch": 960, "lr": 7.760999461622183e-05}
{"train_loss": 0.2301081418991089, "global_step": 85521, "epoch": 960, "lr": 7.760951128114891e-05}
{"train_loss": 0.192379429936409, "global_step": 85522, "epoch": 960, "lr": 7.760902794236422e-05}
{"train_loss": 0.12588563561439514, "global_step": 85523, "epoch": 960, "lr": 7.760854459986784e-05}
{"train_loss": 0.1938103586435318, "global_step": 85524, "epoch": 960, "lr": 7.760806125365982e-05}
{"train_loss": 0.20419776439666748, "global_step": 85525, "epoch": 960, "lr": 7.760757790374025e-05}
{"train_loss": 0.2474360167980194, "global_step": 85526, "epoch": 960, "lr": 7.760709455010916e-05}
{"train_loss": 0.2242952436208725, "global_step": 85527, "epoch": 960, "lr": 7.760661119276662e-05}
{"train_loss": 0.1833311922931939, "global_step": 85528, "epoch": 960, "lr": 7.760612783171273e-05, "val_loss": 4.3885884284973145, "train_action_mse_error": 30.68077278137207}
{"train_loss": 0.17821043729782104, "global_step": 85529, "epoch": 961, "lr": 7.760564446694754e-05}
{"train_loss": 0.201622873544693, "global_step": 85530, "epoch": 961, "lr": 7.76051610984711e-05}
{"train_loss": 0.12135356664657593, "global_step": 85531, "epoch": 961, "lr": 7.760467772628346e-05}
{"train_loss": 0.21541212499141693, "global_step": 85532, "epoch": 961, "lr": 7.760419435038473e-05}
{"train_loss": 0.18376441299915314, "global_step": 85533, "epoch": 961, "lr": 7.760371097077495e-05}
{"train_loss": 0.10993977636098862, "global_step": 85534, "epoch": 961, "lr": 7.760322758745418e-05}
{"train_loss": 0.2037869542837143, "global_step": 85535, "epoch": 961, "lr": 7.760274420042251e-05}
{"train_loss": 0.1644345372915268, "global_step": 85536, "epoch": 961, "lr": 7.760226080967996e-05}
{"train_loss": 0.21949659287929535, "global_step": 85537, "epoch": 961, "lr": 7.760177741522665e-05}
{"train_loss": 0.18064218759536743, "global_step": 85538, "epoch": 961, "lr": 7.76012940170626e-05}
{"train_loss": 0.12647205591201782, "global_step": 85539, "epoch": 961, "lr": 7.760081061518789e-05}
{"train_loss": 0.18232876062393188, "global_step": 85540, "epoch": 961, "lr": 7.76003272096026e-05}
{"train_loss": 0.07353465259075165, "global_step": 85541, "epoch": 961, "lr": 7.759984380030678e-05}
{"train_loss": 0.19759289920330048, "global_step": 85542, "epoch": 961, "lr": 7.759936038730049e-05}
{"train_loss": 0.16895030438899994, "global_step": 85543, "epoch": 961, "lr": 7.759887697058381e-05}
{"train_loss": 0.2858348786830902, "global_step": 85544, "epoch": 961, "lr": 7.759839355015679e-05}
{"train_loss": 0.14051049947738647, "global_step": 85545, "epoch": 961, "lr": 7.759791012601951e-05}
{"train_loss": 0.2000878006219864, "global_step": 85546, "epoch": 961, "lr": 7.759742669817203e-05}
{"train_loss": 0.15138235688209534, "global_step": 85547, "epoch": 961, "lr": 7.75969432666144e-05}
{"train_loss": 0.17146515846252441, "global_step": 85548, "epoch": 961, "lr": 7.75964598313467e-05}
{"train_loss": 0.20255623757839203, "global_step": 85549, "epoch": 961, "lr": 7.7595976392369e-05}
{"train_loss": 0.1574128419160843, "global_step": 85550, "epoch": 961, "lr": 7.759549294968136e-05}
{"train_loss": 0.11362428218126297, "global_step": 85551, "epoch": 961, "lr": 7.759500950328383e-05}
{"train_loss": 0.13142025470733643, "global_step": 85552, "epoch": 961, "lr": 7.75945260531765e-05}
{"train_loss": 0.24827535450458527, "global_step": 85553, "epoch": 961, "lr": 7.759404259935941e-05}
{"train_loss": 0.16050004959106445, "global_step": 85554, "epoch": 961, "lr": 7.759355914183263e-05}
{"train_loss": 0.21680767834186554, "global_step": 85555, "epoch": 961, "lr": 7.759307568059626e-05}
{"train_loss": 0.171891450881958, "global_step": 85556, "epoch": 961, "lr": 7.759259221565031e-05}
{"train_loss": 0.26469242572784424, "global_step": 85557, "epoch": 961, "lr": 7.75921087469949e-05}
{"train_loss": 0.2808269262313843, "global_step": 85558, "epoch": 961, "lr": 7.759162527463004e-05}
{"train_loss": 0.17303010821342468, "global_step": 85559, "epoch": 961, "lr": 7.759114179855584e-05}
{"train_loss": 0.2677752375602722, "global_step": 85560, "epoch": 961, "lr": 7.759065831877235e-05}
{"train_loss": 0.2500280737876892, "global_step": 85561, "epoch": 961, "lr": 7.759017483527962e-05}
{"train_loss": 0.18393270671367645, "global_step": 85562, "epoch": 961, "lr": 7.758969134807774e-05}
{"train_loss": 0.21792253851890564, "global_step": 85563, "epoch": 961, "lr": 7.758920785716675e-05}
{"train_loss": 0.13243368268013, "global_step": 85564, "epoch": 961, "lr": 7.758872436254671e-05}
{"train_loss": 0.1745053082704544, "global_step": 85565, "epoch": 961, "lr": 7.758824086421773e-05}
{"train_loss": 0.15628421306610107, "global_step": 85566, "epoch": 961, "lr": 7.758775736217985e-05}
{"train_loss": 0.20394039154052734, "global_step": 85567, "epoch": 961, "lr": 7.758727385643312e-05}
{"train_loss": 0.23503296077251434, "global_step": 85568, "epoch": 961, "lr": 7.758679034697761e-05}
{"train_loss": 0.25077998638153076, "global_step": 85569, "epoch": 961, "lr": 7.75863068338134e-05}
{"train_loss": 0.1917486935853958, "global_step": 85570, "epoch": 961, "lr": 7.758582331694055e-05}
{"train_loss": 0.17396165430545807, "global_step": 85571, "epoch": 961, "lr": 7.758533979635912e-05}
{"train_loss": 0.17854353785514832, "global_step": 85572, "epoch": 961, "lr": 7.75848562720692e-05}
{"train_loss": 0.13467028737068176, "global_step": 85573, "epoch": 961, "lr": 7.758437274407079e-05}
{"train_loss": 0.19247576594352722, "global_step": 85574, "epoch": 961, "lr": 7.758388921236403e-05}
{"train_loss": 0.202678382396698, "global_step": 85575, "epoch": 961, "lr": 7.758340567694893e-05}
{"train_loss": 0.22030985355377197, "global_step": 85576, "epoch": 961, "lr": 7.75829221378256e-05}
{"train_loss": 0.14997926354408264, "global_step": 85577, "epoch": 961, "lr": 7.758243859499407e-05}
{"train_loss": 0.2393670678138733, "global_step": 85578, "epoch": 961, "lr": 7.758195504845441e-05}
{"train_loss": 0.15883350372314453, "global_step": 85579, "epoch": 961, "lr": 7.75814714982067e-05}
{"train_loss": 0.23411113023757935, "global_step": 85580, "epoch": 961, "lr": 7.7580987944251e-05}
{"train_loss": 0.23498672246932983, "global_step": 85581, "epoch": 961, "lr": 7.758050438658737e-05}
{"train_loss": 0.2447047233581543, "global_step": 85582, "epoch": 961, "lr": 7.758002082521587e-05}
{"train_loss": 0.16871334612369537, "global_step": 85583, "epoch": 961, "lr": 7.757953726013658e-05}
{"train_loss": 0.16198202967643738, "global_step": 85584, "epoch": 961, "lr": 7.757905369134955e-05}
{"train_loss": 0.15735772252082825, "global_step": 85585, "epoch": 961, "lr": 7.757857011885487e-05}
{"train_loss": 0.10110392421483994, "global_step": 85586, "epoch": 961, "lr": 7.757808654265257e-05}
{"train_loss": 0.23137681186199188, "global_step": 85587, "epoch": 961, "lr": 7.757760296274274e-05}
{"train_loss": 0.15623317658901215, "global_step": 85588, "epoch": 961, "lr": 7.757711937912545e-05}
{"train_loss": 0.21403130888938904, "global_step": 85589, "epoch": 961, "lr": 7.757663579180072e-05}
{"train_loss": 0.23853525519371033, "global_step": 85590, "epoch": 961, "lr": 7.757615220076868e-05}
{"train_loss": 0.19734899699687958, "global_step": 85591, "epoch": 961, "lr": 7.757566860602934e-05}
{"train_loss": 0.2708005905151367, "global_step": 85592, "epoch": 961, "lr": 7.757518500758281e-05}
{"train_loss": 0.27333787083625793, "global_step": 85593, "epoch": 961, "lr": 7.757470140542911e-05}
{"train_loss": 0.20799779891967773, "global_step": 85594, "epoch": 961, "lr": 7.757421779956832e-05}
{"train_loss": 0.22516722977161407, "global_step": 85595, "epoch": 961, "lr": 7.757373419000054e-05}
{"train_loss": 0.19012200832366943, "global_step": 85596, "epoch": 961, "lr": 7.75732505767258e-05}
{"train_loss": 0.18139603734016418, "global_step": 85597, "epoch": 961, "lr": 7.757276695974416e-05}
{"train_loss": 0.15894463658332825, "global_step": 85598, "epoch": 961, "lr": 7.75722833390557e-05}
{"train_loss": 0.15199705958366394, "global_step": 85599, "epoch": 961, "lr": 7.75717997146605e-05}
{"train_loss": 0.16940535604953766, "global_step": 85600, "epoch": 961, "lr": 7.757131608655859e-05}
{"train_loss": 0.21565724909305573, "global_step": 85601, "epoch": 961, "lr": 7.757083245475005e-05}
{"train_loss": 0.1707107573747635, "global_step": 85602, "epoch": 961, "lr": 7.757034881923498e-05}
{"train_loss": 0.13437388837337494, "global_step": 85603, "epoch": 961, "lr": 7.756986518001337e-05}
{"train_loss": 0.20647220313549042, "global_step": 85604, "epoch": 961, "lr": 7.756938153708535e-05}
{"train_loss": 0.18141739070415497, "global_step": 85605, "epoch": 961, "lr": 7.756889789045097e-05}
{"train_loss": 0.18718138337135315, "global_step": 85606, "epoch": 961, "lr": 7.756841424011027e-05}
{"train_loss": 0.1852220892906189, "global_step": 85607, "epoch": 961, "lr": 7.756793058606332e-05}
{"train_loss": 0.1909724324941635, "global_step": 85608, "epoch": 961, "lr": 7.756744692831024e-05}
{"train_loss": 0.16989687085151672, "global_step": 85609, "epoch": 961, "lr": 7.756696326685102e-05}
{"train_loss": 0.15071001648902893, "global_step": 85610, "epoch": 961, "lr": 7.756647960168577e-05}
{"train_loss": 0.21382373571395874, "global_step": 85611, "epoch": 961, "lr": 7.756599593281455e-05}
{"train_loss": 0.22575919330120087, "global_step": 85612, "epoch": 961, "lr": 7.756551226023741e-05}
{"train_loss": 0.2134263664484024, "global_step": 85613, "epoch": 961, "lr": 7.756502858395442e-05}
{"train_loss": 0.18657024204730988, "global_step": 85614, "epoch": 961, "lr": 7.756454490396565e-05}
{"train_loss": 0.21055901050567627, "global_step": 85615, "epoch": 961, "lr": 7.756406122027115e-05}
{"train_loss": 0.18000540137290955, "global_step": 85616, "epoch": 961, "lr": 7.756357753287103e-05}
{"train_loss": 0.18970035075136785, "global_step": 85617, "epoch": 961, "lr": 7.75630938417653e-05, "val_loss": 4.0102009773254395}
{"train_loss": 0.13150924444198608, "global_step": 85618, "epoch": 962, "lr": 7.756261014695404e-05}
{"train_loss": 0.2027817666530609, "global_step": 85619, "epoch": 962, "lr": 7.756212644843733e-05}
{"train_loss": 0.24017266929149628, "global_step": 85620, "epoch": 962, "lr": 7.756164274621523e-05}
{"train_loss": 0.1967729777097702, "global_step": 85621, "epoch": 962, "lr": 7.756115904028781e-05}
{"train_loss": 0.2135327011346817, "global_step": 85622, "epoch": 962, "lr": 7.756067533065512e-05}
{"train_loss": 0.1819602996110916, "global_step": 85623, "epoch": 962, "lr": 7.756019161731725e-05}
{"train_loss": 0.20661576092243195, "global_step": 85624, "epoch": 962, "lr": 7.755970790027423e-05}
{"train_loss": 0.17926371097564697, "global_step": 85625, "epoch": 962, "lr": 7.755922417952615e-05}
{"train_loss": 0.1936085969209671, "global_step": 85626, "epoch": 962, "lr": 7.755874045507305e-05}
{"train_loss": 0.18183700740337372, "global_step": 85627, "epoch": 962, "lr": 7.755825672691504e-05}
{"train_loss": 0.20608671009540558, "global_step": 85628, "epoch": 962, "lr": 7.755777299505214e-05}
{"train_loss": 0.29153120517730713, "global_step": 85629, "epoch": 962, "lr": 7.755728925948445e-05}
{"train_loss": 0.15236029028892517, "global_step": 85630, "epoch": 962, "lr": 7.7556805520212e-05}
{"train_loss": 0.2720285654067993, "global_step": 85631, "epoch": 962, "lr": 7.755632177723489e-05}
{"train_loss": 0.12755602598190308, "global_step": 85632, "epoch": 962, "lr": 7.755583803055317e-05}
{"train_loss": 0.12140708416700363, "global_step": 85633, "epoch": 962, "lr": 7.755535428016689e-05}
{"train_loss": 0.2489718347787857, "global_step": 85634, "epoch": 962, "lr": 7.755487052607613e-05}
{"train_loss": 0.1781173050403595, "global_step": 85635, "epoch": 962, "lr": 7.755438676828095e-05}
{"train_loss": 0.19693832099437714, "global_step": 85636, "epoch": 962, "lr": 7.755390300678143e-05}
{"train_loss": 0.18682484328746796, "global_step": 85637, "epoch": 962, "lr": 7.755341924157762e-05}
{"train_loss": 0.13902375102043152, "global_step": 85638, "epoch": 962, "lr": 7.755293547266958e-05}
{"train_loss": 0.17095834016799927, "global_step": 85639, "epoch": 962, "lr": 7.755245170005738e-05}
{"train_loss": 0.11481934785842896, "global_step": 85640, "epoch": 962, "lr": 7.75519679237411e-05}
{"train_loss": 0.1001785397529602, "global_step": 85641, "epoch": 962, "lr": 7.75514841437208e-05}
{"train_loss": 0.20048020780086517, "global_step": 85642, "epoch": 962, "lr": 7.755100035999654e-05}
{"train_loss": 0.14903981983661652, "global_step": 85643, "epoch": 962, "lr": 7.755051657256837e-05}
{"train_loss": 0.1684502363204956, "global_step": 85644, "epoch": 962, "lr": 7.755003278143638e-05}
{"train_loss": 0.17005561292171478, "global_step": 85645, "epoch": 962, "lr": 7.75495489866006e-05}
{"train_loss": 0.18932019174098969, "global_step": 85646, "epoch": 962, "lr": 7.754906518806115e-05}
{"train_loss": 0.14573800563812256, "global_step": 85647, "epoch": 962, "lr": 7.754858138581804e-05}
{"train_loss": 0.2316761016845703, "global_step": 85648, "epoch": 962, "lr": 7.754809757987139e-05}
{"train_loss": 0.16897594928741455, "global_step": 85649, "epoch": 962, "lr": 7.754761377022122e-05}
{"train_loss": 0.1569705456495285, "global_step": 85650, "epoch": 962, "lr": 7.75471299568676e-05}
{"train_loss": 0.15356245636940002, "global_step": 85651, "epoch": 962, "lr": 7.75466461398106e-05}
{"train_loss": 0.22939041256904602, "global_step": 85652, "epoch": 962, "lr": 7.75461623190503e-05}
{"train_loss": 0.15111994743347168, "global_step": 85653, "epoch": 962, "lr": 7.754567849458674e-05}
{"train_loss": 0.24933379888534546, "global_step": 85654, "epoch": 962, "lr": 7.754519466642e-05}
{"train_loss": 0.17553985118865967, "global_step": 85655, "epoch": 962, "lr": 7.754471083455016e-05}
{"train_loss": 0.1805875450372696, "global_step": 85656, "epoch": 962, "lr": 7.754422699897727e-05}
{"train_loss": 0.12591435015201569, "global_step": 85657, "epoch": 962, "lr": 7.754374315970139e-05}
{"train_loss": 0.13125485181808472, "global_step": 85658, "epoch": 962, "lr": 7.75432593167226e-05}
{"train_loss": 0.14243026077747345, "global_step": 85659, "epoch": 962, "lr": 7.754277547004093e-05}
{"train_loss": 0.15010984241962433, "global_step": 85660, "epoch": 962, "lr": 7.754229161965648e-05}
{"train_loss": 0.18943646550178528, "global_step": 85661, "epoch": 962, "lr": 7.754180776556931e-05}
{"train_loss": 0.18254850804805756, "global_step": 85662, "epoch": 962, "lr": 7.754132390777947e-05}
{"train_loss": 0.14422111213207245, "global_step": 85663, "epoch": 962, "lr": 7.754084004628704e-05}
{"train_loss": 0.1464427411556244, "global_step": 85664, "epoch": 962, "lr": 7.754035618109207e-05}
{"train_loss": 0.1836351901292801, "global_step": 85665, "epoch": 962, "lr": 7.753987231219466e-05}
{"train_loss": 0.15422895550727844, "global_step": 85666, "epoch": 962, "lr": 7.753938843959482e-05}
{"train_loss": 0.18173100054264069, "global_step": 85667, "epoch": 962, "lr": 7.753890456329267e-05}
{"train_loss": 0.17732900381088257, "global_step": 85668, "epoch": 962, "lr": 7.753842068328824e-05}
{"train_loss": 0.19312292337417603, "global_step": 85669, "epoch": 962, "lr": 7.75379367995816e-05}
{"train_loss": 0.22226572036743164, "global_step": 85670, "epoch": 962, "lr": 7.753745291217284e-05}
{"train_loss": 0.16473674774169922, "global_step": 85671, "epoch": 962, "lr": 7.753696902106198e-05}
{"train_loss": 0.1774180680513382, "global_step": 85672, "epoch": 962, "lr": 7.753648512624912e-05}
{"train_loss": 0.18149660527706146, "global_step": 85673, "epoch": 962, "lr": 7.753600122773432e-05}
{"train_loss": 0.15339511632919312, "global_step": 85674, "epoch": 962, "lr": 7.753551732551762e-05}
{"train_loss": 0.15758398175239563, "global_step": 85675, "epoch": 962, "lr": 7.753503341959913e-05}
{"train_loss": 0.22178810834884644, "global_step": 85676, "epoch": 962, "lr": 7.753454950997888e-05}
{"train_loss": 0.19242234528064728, "global_step": 85677, "epoch": 962, "lr": 7.753406559665695e-05}
{"train_loss": 0.23496542870998383, "global_step": 85678, "epoch": 962, "lr": 7.753358167963339e-05}
{"train_loss": 0.14614279568195343, "global_step": 85679, "epoch": 962, "lr": 7.753309775890829e-05}
{"train_loss": 0.20399659872055054, "global_step": 85680, "epoch": 962, "lr": 7.753261383448169e-05}
{"train_loss": 0.14013831317424774, "global_step": 85681, "epoch": 962, "lr": 7.753212990635366e-05}
{"train_loss": 0.2886642515659332, "global_step": 85682, "epoch": 962, "lr": 7.753164597452429e-05}
{"train_loss": 0.2940802276134491, "global_step": 85683, "epoch": 962, "lr": 7.753116203899362e-05}
{"train_loss": 0.13013234734535217, "global_step": 85684, "epoch": 962, "lr": 7.753067809976171e-05}
{"train_loss": 0.14822210371494293, "global_step": 85685, "epoch": 962, "lr": 7.753019415682865e-05}
{"train_loss": 0.1907726228237152, "global_step": 85686, "epoch": 962, "lr": 7.752971021019447e-05}
{"train_loss": 0.14330840110778809, "global_step": 85687, "epoch": 962, "lr": 7.752922625985927e-05}
{"train_loss": 0.1395258754491806, "global_step": 85688, "epoch": 962, "lr": 7.752874230582311e-05}
{"train_loss": 0.219760000705719, "global_step": 85689, "epoch": 962, "lr": 7.752825834808604e-05}
{"train_loss": 0.18197861313819885, "global_step": 85690, "epoch": 962, "lr": 7.752777438664813e-05}
{"train_loss": 0.22048325836658478, "global_step": 85691, "epoch": 962, "lr": 7.752729042150942e-05}
{"train_loss": 0.24353885650634766, "global_step": 85692, "epoch": 962, "lr": 7.752680645267003e-05}
{"train_loss": 0.20993995666503906, "global_step": 85693, "epoch": 962, "lr": 7.752632248012999e-05}
{"train_loss": 0.13110166788101196, "global_step": 85694, "epoch": 962, "lr": 7.752583850388937e-05}
{"train_loss": 0.19869346916675568, "global_step": 85695, "epoch": 962, "lr": 7.752535452394824e-05}
{"train_loss": 0.13635599613189697, "global_step": 85696, "epoch": 962, "lr": 7.752487054030666e-05}
{"train_loss": 0.1934014856815338, "global_step": 85697, "epoch": 962, "lr": 7.75243865529647e-05}
{"train_loss": 0.2799004018306732, "global_step": 85698, "epoch": 962, "lr": 7.752390256192243e-05}
{"train_loss": 0.157753586769104, "global_step": 85699, "epoch": 962, "lr": 7.752341856717988e-05}
{"train_loss": 0.11694569140672684, "global_step": 85700, "epoch": 962, "lr": 7.752293456873716e-05}
{"train_loss": 0.2075154036283493, "global_step": 85701, "epoch": 962, "lr": 7.752245056659432e-05}
{"train_loss": 0.194752037525177, "global_step": 85702, "epoch": 962, "lr": 7.752196656075142e-05}
{"train_loss": 0.266697496175766, "global_step": 85703, "epoch": 962, "lr": 7.752148255120852e-05}
{"train_loss": 0.1558232605457306, "global_step": 85704, "epoch": 962, "lr": 7.752099853796571e-05}
{"train_loss": 0.1488242745399475, "global_step": 85705, "epoch": 962, "lr": 7.752051452102301e-05}
{"train_loss": 0.18318677501062328, "global_step": 85706, "epoch": 962, "lr": 7.752003050038053e-05, "val_loss": 4.381970405578613}
{"train_loss": 0.1658421754837036, "global_step": 85707, "epoch": 963, "lr": 7.751954647603832e-05}
{"train_loss": 0.21271273493766785, "global_step": 85708, "epoch": 963, "lr": 7.751906244799643e-05}
{"train_loss": 0.09459453076124191, "global_step": 85709, "epoch": 963, "lr": 7.751857841625493e-05}
{"train_loss": 0.1770678013563156, "global_step": 85710, "epoch": 963, "lr": 7.75180943808139e-05}
{"train_loss": 0.14132870733737946, "global_step": 85711, "epoch": 963, "lr": 7.751761034167341e-05}
{"train_loss": 0.2252129316329956, "global_step": 85712, "epoch": 963, "lr": 7.75171262988335e-05}
{"train_loss": 0.19410647451877594, "global_step": 85713, "epoch": 963, "lr": 7.751664225229424e-05}
{"train_loss": 0.1651073843240738, "global_step": 85714, "epoch": 963, "lr": 7.751615820205574e-05}
{"train_loss": 0.1807098537683487, "global_step": 85715, "epoch": 963, "lr": 7.7515674148118e-05}
{"train_loss": 0.15738053619861603, "global_step": 85716, "epoch": 963, "lr": 7.751519009048112e-05}
{"train_loss": 0.22036266326904297, "global_step": 85717, "epoch": 963, "lr": 7.751470602914514e-05}
{"train_loss": 0.1971970945596695, "global_step": 85718, "epoch": 963, "lr": 7.751422196411017e-05}
{"train_loss": 0.136520117521286, "global_step": 85719, "epoch": 963, "lr": 7.751373789537623e-05}
{"train_loss": 0.20861978828907013, "global_step": 85720, "epoch": 963, "lr": 7.75132538229434e-05}
{"train_loss": 0.22603805363178253, "global_step": 85721, "epoch": 963, "lr": 7.751276974681177e-05}
{"train_loss": 0.1381969004869461, "global_step": 85722, "epoch": 963, "lr": 7.751228566698136e-05}
{"train_loss": 0.18900170922279358, "global_step": 85723, "epoch": 963, "lr": 7.751180158345228e-05}
{"train_loss": 0.1290694624185562, "global_step": 85724, "epoch": 963, "lr": 7.751131749622455e-05}
{"train_loss": 0.22894152998924255, "global_step": 85725, "epoch": 963, "lr": 7.751083340529827e-05}
{"train_loss": 0.28047141432762146, "global_step": 85726, "epoch": 963, "lr": 7.751034931067351e-05}
{"train_loss": 0.1893520951271057, "global_step": 85727, "epoch": 963, "lr": 7.750986521235031e-05}
{"train_loss": 0.21369175612926483, "global_step": 85728, "epoch": 963, "lr": 7.750938111032873e-05}
{"train_loss": 0.17024271190166473, "global_step": 85729, "epoch": 963, "lr": 7.750889700460886e-05}
{"train_loss": 0.17121203243732452, "global_step": 85730, "epoch": 963, "lr": 7.750841289519075e-05}
{"train_loss": 0.1727764755487442, "global_step": 85731, "epoch": 963, "lr": 7.750792878207447e-05}
{"train_loss": 0.17828097939491272, "global_step": 85732, "epoch": 963, "lr": 7.750744466526007e-05}
{"train_loss": 0.11156085878610611, "global_step": 85733, "epoch": 963, "lr": 7.750696054474766e-05}
{"train_loss": 0.22691819071769714, "global_step": 85734, "epoch": 963, "lr": 7.750647642053726e-05}
{"train_loss": 0.2220689356327057, "global_step": 85735, "epoch": 963, "lr": 7.750599229262895e-05}
{"train_loss": 0.2503737211227417, "global_step": 85736, "epoch": 963, "lr": 7.75055081610228e-05}
{"train_loss": 0.2530268132686615, "global_step": 85737, "epoch": 963, "lr": 7.750502402571886e-05}
{"train_loss": 0.15276412665843964, "global_step": 85738, "epoch": 963, "lr": 7.75045398867172e-05}
{"train_loss": 0.17159512639045715, "global_step": 85739, "epoch": 963, "lr": 7.750405574401789e-05}
{"train_loss": 0.2231636345386505, "global_step": 85740, "epoch": 963, "lr": 7.750357159762101e-05}
{"train_loss": 0.22416377067565918, "global_step": 85741, "epoch": 963, "lr": 7.75030874475266e-05}
{"train_loss": 0.13603176176548004, "global_step": 85742, "epoch": 963, "lr": 7.750260329373473e-05}
{"train_loss": 0.11011485755443573, "global_step": 85743, "epoch": 963, "lr": 7.750211913624548e-05}
{"train_loss": 0.16724692285060883, "global_step": 85744, "epoch": 963, "lr": 7.750163497505892e-05}
{"train_loss": 0.17013201117515564, "global_step": 85745, "epoch": 963, "lr": 7.750115081017506e-05}
{"train_loss": 0.25674736499786377, "global_step": 85746, "epoch": 963, "lr": 7.750066664159403e-05}
{"train_loss": 0.15643469989299774, "global_step": 85747, "epoch": 963, "lr": 7.750018246931587e-05}
{"train_loss": 0.2425248771905899, "global_step": 85748, "epoch": 963, "lr": 7.749969829334066e-05}
{"train_loss": 0.2227567434310913, "global_step": 85749, "epoch": 963, "lr": 7.749921411366843e-05}
{"train_loss": 0.21326963603496552, "global_step": 85750, "epoch": 963, "lr": 7.749872993029926e-05}
{"train_loss": 0.2122390866279602, "global_step": 85751, "epoch": 963, "lr": 7.749824574323324e-05}
{"train_loss": 0.1797102391719818, "global_step": 85752, "epoch": 963, "lr": 7.749776155247041e-05}
{"train_loss": 0.22248074412345886, "global_step": 85753, "epoch": 963, "lr": 7.749727735801085e-05}
{"train_loss": 0.2509876489639282, "global_step": 85754, "epoch": 963, "lr": 7.749679315985459e-05}
{"train_loss": 0.21348701417446136, "global_step": 85755, "epoch": 963, "lr": 7.749630895800174e-05}
{"train_loss": 0.14627006649971008, "global_step": 85756, "epoch": 963, "lr": 7.749582475245234e-05}
{"train_loss": 0.19615107774734497, "global_step": 85757, "epoch": 963, "lr": 7.749534054320646e-05}
{"train_loss": 0.21976955235004425, "global_step": 85758, "epoch": 963, "lr": 7.749485633026417e-05}
{"train_loss": 0.2554399073123932, "global_step": 85759, "epoch": 963, "lr": 7.749437211362553e-05}
{"train_loss": 0.2138649970293045, "global_step": 85760, "epoch": 963, "lr": 7.749388789329061e-05}
{"train_loss": 0.11467885971069336, "global_step": 85761, "epoch": 963, "lr": 7.749340366925948e-05}
{"train_loss": 0.18159806728363037, "global_step": 85762, "epoch": 963, "lr": 7.749291944153218e-05}
{"train_loss": 0.1729586124420166, "global_step": 85763, "epoch": 963, "lr": 7.74924352101088e-05}
{"train_loss": 0.25002235174179077, "global_step": 85764, "epoch": 963, "lr": 7.74919509749894e-05}
{"train_loss": 0.21219073235988617, "global_step": 85765, "epoch": 963, "lr": 7.749146673617405e-05}
{"train_loss": 0.14106762409210205, "global_step": 85766, "epoch": 963, "lr": 7.74909824936628e-05}
{"train_loss": 0.2528795599937439, "global_step": 85767, "epoch": 963, "lr": 7.749049824745571e-05}
{"train_loss": 0.13216586410999298, "global_step": 85768, "epoch": 963, "lr": 7.749001399755286e-05}
{"train_loss": 0.24414928257465363, "global_step": 85769, "epoch": 963, "lr": 7.748952974395434e-05}
{"train_loss": 0.20687218010425568, "global_step": 85770, "epoch": 963, "lr": 7.748904548666016e-05}
{"train_loss": 0.3113853335380554, "global_step": 85771, "epoch": 963, "lr": 7.748856122567042e-05}
{"train_loss": 0.13490243256092072, "global_step": 85772, "epoch": 963, "lr": 7.748807696098518e-05}
{"train_loss": 0.14355218410491943, "global_step": 85773, "epoch": 963, "lr": 7.748759269260451e-05}
{"train_loss": 0.17336827516555786, "global_step": 85774, "epoch": 963, "lr": 7.748710842052845e-05}
{"train_loss": 0.10789420455694199, "global_step": 85775, "epoch": 963, "lr": 7.748662414475709e-05}
{"train_loss": 0.17348536849021912, "global_step": 85776, "epoch": 963, "lr": 7.74861398652905e-05}
{"train_loss": 0.30095499753952026, "global_step": 85777, "epoch": 963, "lr": 7.748565558212873e-05}
{"train_loss": 0.2686595320701599, "global_step": 85778, "epoch": 963, "lr": 7.748517129527184e-05}
{"train_loss": 0.1858258843421936, "global_step": 85779, "epoch": 963, "lr": 7.74846870047199e-05}
{"train_loss": 0.25850170850753784, "global_step": 85780, "epoch": 963, "lr": 7.748420271047299e-05}
{"train_loss": 0.19882793724536896, "global_step": 85781, "epoch": 963, "lr": 7.748371841253114e-05}
{"train_loss": 0.19105872511863708, "global_step": 85782, "epoch": 963, "lr": 7.748323411089448e-05}
{"train_loss": 0.10202335566282272, "global_step": 85783, "epoch": 963, "lr": 7.7482749805563e-05}
{"train_loss": 0.15959012508392334, "global_step": 85784, "epoch": 963, "lr": 7.748226549653683e-05}
{"train_loss": 0.11466607451438904, "global_step": 85785, "epoch": 963, "lr": 7.748178118381597e-05}
{"train_loss": 0.256332665681839, "global_step": 85786, "epoch": 963, "lr": 7.748129686740055e-05}
{"train_loss": 0.2438802868127823, "global_step": 85787, "epoch": 963, "lr": 7.748081254729058e-05}
{"train_loss": 0.13937175273895264, "global_step": 85788, "epoch": 963, "lr": 7.748032822348616e-05}
{"train_loss": 0.22868844866752625, "global_step": 85789, "epoch": 963, "lr": 7.747984389598735e-05}
{"train_loss": 0.15069326758384705, "global_step": 85790, "epoch": 963, "lr": 7.747935956479421e-05}
{"train_loss": 0.19169478118419647, "global_step": 85791, "epoch": 963, "lr": 7.74788752299068e-05}
{"train_loss": 0.23612454533576965, "global_step": 85792, "epoch": 963, "lr": 7.747839089132519e-05}
{"train_loss": 0.24376066029071808, "global_step": 85793, "epoch": 963, "lr": 7.747790654904946e-05}
{"train_loss": 0.13279227912425995, "global_step": 85794, "epoch": 963, "lr": 7.747742220307963e-05}
{"train_loss": 0.19298278180400977, "global_step": 85795, "epoch": 963, "lr": 7.74769378534158e-05, "val_loss": 4.104490756988525}
{"train_loss": 0.13802409172058105, "global_step": 85796, "epoch": 964, "lr": 7.747645350005805e-05}
{"train_loss": 0.1830788552761078, "global_step": 85797, "epoch": 964, "lr": 7.747596914300643e-05}
{"train_loss": 0.16845722496509552, "global_step": 85798, "epoch": 964, "lr": 7.747548478226099e-05}
{"train_loss": 0.22133341431617737, "global_step": 85799, "epoch": 964, "lr": 7.747500041782181e-05}
{"train_loss": 0.24071025848388672, "global_step": 85800, "epoch": 964, "lr": 7.747451604968894e-05}
{"train_loss": 0.19569657742977142, "global_step": 85801, "epoch": 964, "lr": 7.747403167786245e-05}
{"train_loss": 0.13755568861961365, "global_step": 85802, "epoch": 964, "lr": 7.747354730234243e-05}
{"train_loss": 0.1741587221622467, "global_step": 85803, "epoch": 964, "lr": 7.747306292312893e-05}
{"train_loss": 0.28381267189979553, "global_step": 85804, "epoch": 964, "lr": 7.747257854022199e-05}
{"train_loss": 0.19050335884094238, "global_step": 85805, "epoch": 964, "lr": 7.747209415362171e-05}
{"train_loss": 0.24842482805252075, "global_step": 85806, "epoch": 964, "lr": 7.747160976332815e-05}
{"train_loss": 0.18690767884254456, "global_step": 85807, "epoch": 964, "lr": 7.747112536934135e-05}
{"train_loss": 0.15202903747558594, "global_step": 85808, "epoch": 964, "lr": 7.74706409716614e-05}
{"train_loss": 0.10335064679384232, "global_step": 85809, "epoch": 964, "lr": 7.747015657028835e-05}
{"train_loss": 0.2529371380805969, "global_step": 85810, "epoch": 964, "lr": 7.746967216522227e-05}
{"train_loss": 0.24425135552883148, "global_step": 85811, "epoch": 964, "lr": 7.746918775646325e-05}
{"train_loss": 0.204916313290596, "global_step": 85812, "epoch": 964, "lr": 7.746870334401131e-05}
{"train_loss": 0.19674165546894073, "global_step": 85813, "epoch": 964, "lr": 7.746821892786656e-05}
{"train_loss": 0.14025358855724335, "global_step": 85814, "epoch": 964, "lr": 7.746773450802902e-05}
{"train_loss": 0.1771731823682785, "global_step": 85815, "epoch": 964, "lr": 7.746725008449879e-05}
{"train_loss": 0.16206568479537964, "global_step": 85816, "epoch": 964, "lr": 7.746676565727591e-05}
{"train_loss": 0.2345588207244873, "global_step": 85817, "epoch": 964, "lr": 7.746628122636047e-05}
{"train_loss": 0.16822001338005066, "global_step": 85818, "epoch": 964, "lr": 7.746579679175252e-05}
{"train_loss": 0.1694999635219574, "global_step": 85819, "epoch": 964, "lr": 7.746531235345214e-05}
{"train_loss": 0.21368591487407684, "global_step": 85820, "epoch": 964, "lr": 7.746482791145936e-05}
{"train_loss": 0.18369174003601074, "global_step": 85821, "epoch": 964, "lr": 7.746434346577428e-05}
{"train_loss": 0.19833387434482574, "global_step": 85822, "epoch": 964, "lr": 7.746385901639696e-05}
{"train_loss": 0.18209928274154663, "global_step": 85823, "epoch": 964, "lr": 7.746337456332746e-05}
{"train_loss": 0.24274520576000214, "global_step": 85824, "epoch": 964, "lr": 7.746289010656582e-05}
{"train_loss": 0.21193379163742065, "global_step": 85825, "epoch": 964, "lr": 7.746240564611216e-05}
{"train_loss": 0.20204289257526398, "global_step": 85826, "epoch": 964, "lr": 7.746192118196649e-05}
{"train_loss": 0.2601572573184967, "global_step": 85827, "epoch": 964, "lr": 7.746143671412891e-05}
{"train_loss": 0.2558131515979767, "global_step": 85828, "epoch": 964, "lr": 7.746095224259948e-05}
{"train_loss": 0.1625002920627594, "global_step": 85829, "epoch": 964, "lr": 7.746046776737825e-05}
{"train_loss": 0.19978806376457214, "global_step": 85830, "epoch": 964, "lr": 7.74599832884653e-05}
{"train_loss": 0.1658535897731781, "global_step": 85831, "epoch": 964, "lr": 7.74594988058607e-05}
{"train_loss": 0.231415256857872, "global_step": 85832, "epoch": 964, "lr": 7.74590143195645e-05}
{"train_loss": 0.2224164456129074, "global_step": 85833, "epoch": 964, "lr": 7.745852982957677e-05}
{"train_loss": 0.2765550911426544, "global_step": 85834, "epoch": 964, "lr": 7.745804533589756e-05}
{"train_loss": 0.20601791143417358, "global_step": 85835, "epoch": 964, "lr": 7.745756083852696e-05}
{"train_loss": 0.2617351710796356, "global_step": 85836, "epoch": 964, "lr": 7.745707633746502e-05}
{"train_loss": 0.19033494591712952, "global_step": 85837, "epoch": 964, "lr": 7.745659183271182e-05}
{"train_loss": 0.20440416038036346, "global_step": 85838, "epoch": 964, "lr": 7.745610732426742e-05}
{"train_loss": 0.22967568039894104, "global_step": 85839, "epoch": 964, "lr": 7.745562281213187e-05}
{"train_loss": 0.16342049837112427, "global_step": 85840, "epoch": 964, "lr": 7.745513829630526e-05}
{"train_loss": 0.1567908525466919, "global_step": 85841, "epoch": 964, "lr": 7.745465377678762e-05}
{"train_loss": 0.17009706795215607, "global_step": 85842, "epoch": 964, "lr": 7.745416925357904e-05}
{"train_loss": 0.21387815475463867, "global_step": 85843, "epoch": 964, "lr": 7.74536847266796e-05}
{"train_loss": 0.1737455278635025, "global_step": 85844, "epoch": 964, "lr": 7.745320019608934e-05}
{"train_loss": 0.29884010553359985, "global_step": 85845, "epoch": 964, "lr": 7.745271566180831e-05}
{"train_loss": 0.1932985931634903, "global_step": 85846, "epoch": 964, "lr": 7.745223112383663e-05}
{"train_loss": 0.21707284450531006, "global_step": 85847, "epoch": 964, "lr": 7.745174658217431e-05}
{"train_loss": 0.13750125467777252, "global_step": 85848, "epoch": 964, "lr": 7.745126203682144e-05}
{"train_loss": 0.18297068774700165, "global_step": 85849, "epoch": 964, "lr": 7.74507774877781e-05}
{"train_loss": 0.11157404631376266, "global_step": 85850, "epoch": 964, "lr": 7.745029293504433e-05}
{"train_loss": 0.13207091391086578, "global_step": 85851, "epoch": 964, "lr": 7.74498083786202e-05}
{"train_loss": 0.17904356122016907, "global_step": 85852, "epoch": 964, "lr": 7.744932381850577e-05}
{"train_loss": 0.13106009364128113, "global_step": 85853, "epoch": 964, "lr": 7.744883925470112e-05}
{"train_loss": 0.22139482200145721, "global_step": 85854, "epoch": 964, "lr": 7.744835468720629e-05}
{"train_loss": 0.27373576164245605, "global_step": 85855, "epoch": 964, "lr": 7.744787011602138e-05}
{"train_loss": 0.20149342715740204, "global_step": 85856, "epoch": 964, "lr": 7.744738554114643e-05}
{"train_loss": 0.19117604196071625, "global_step": 85857, "epoch": 964, "lr": 7.744690096258154e-05}
{"train_loss": 0.1818636655807495, "global_step": 85858, "epoch": 964, "lr": 7.744641638032672e-05}
{"train_loss": 0.13792242109775543, "global_step": 85859, "epoch": 964, "lr": 7.744593179438208e-05}
{"train_loss": 0.20048971474170685, "global_step": 85860, "epoch": 964, "lr": 7.744544720474767e-05}
{"train_loss": 0.2113160640001297, "global_step": 85861, "epoch": 964, "lr": 7.744496261142353e-05}
{"train_loss": 0.18070146441459656, "global_step": 85862, "epoch": 964, "lr": 7.744447801440977e-05}
{"train_loss": 0.18861109018325806, "global_step": 85863, "epoch": 964, "lr": 7.744399341370642e-05}
{"train_loss": 0.13244862854480743, "global_step": 85864, "epoch": 964, "lr": 7.74435088093136e-05}
{"train_loss": 0.22694779932498932, "global_step": 85865, "epoch": 964, "lr": 7.744302420123128e-05}
{"train_loss": 0.18708178400993347, "global_step": 85866, "epoch": 964, "lr": 7.744253958945962e-05}
{"train_loss": 0.18267594277858734, "global_step": 85867, "epoch": 964, "lr": 7.744205497399863e-05}
{"train_loss": 0.16668683290481567, "global_step": 85868, "epoch": 964, "lr": 7.744157035484839e-05}
{"train_loss": 0.20863300561904907, "global_step": 85869, "epoch": 964, "lr": 7.744108573200897e-05}
{"train_loss": 0.17042891681194305, "global_step": 85870, "epoch": 964, "lr": 7.744060110548042e-05}
{"train_loss": 0.1961807757616043, "global_step": 85871, "epoch": 964, "lr": 7.744011647526282e-05}
{"train_loss": 0.13323144614696503, "global_step": 85872, "epoch": 964, "lr": 7.743963184135623e-05}
{"train_loss": 0.2819081246852875, "global_step": 85873, "epoch": 964, "lr": 7.743914720376071e-05}
{"train_loss": 0.18819598853588104, "global_step": 85874, "epoch": 964, "lr": 7.743866256247633e-05}
{"train_loss": 0.13911253213882446, "global_step": 85875, "epoch": 964, "lr": 7.743817791750317e-05}
{"train_loss": 0.1492328941822052, "global_step": 85876, "epoch": 964, "lr": 7.74376932688413e-05}
{"train_loss": 0.1518193781375885, "global_step": 85877, "epoch": 964, "lr": 7.743720861649072e-05}
{"train_loss": 0.20478205382823944, "global_step": 85878, "epoch": 964, "lr": 7.743672396045159e-05}
{"train_loss": 0.17480890452861786, "global_step": 85879, "epoch": 964, "lr": 7.743623930072389e-05}
{"train_loss": 0.09040293842554092, "global_step": 85880, "epoch": 964, "lr": 7.743575463730773e-05}
{"train_loss": 0.19432249665260315, "global_step": 85881, "epoch": 964, "lr": 7.743526997020317e-05}
{"train_loss": 0.24505726993083954, "global_step": 85882, "epoch": 964, "lr": 7.743478529941029e-05}
{"train_loss": 0.13957585394382477, "global_step": 85883, "epoch": 964, "lr": 7.743430062492912e-05}
{"train_loss": 0.19151475548409344, "global_step": 85884, "epoch": 964, "lr": 7.743381594675973e-05, "val_loss": 4.342576026916504}
{"train_loss": 0.19547122716903687, "global_step": 85885, "epoch": 965, "lr": 7.743333126490222e-05}
{"train_loss": 0.19720765948295593, "global_step": 85886, "epoch": 965, "lr": 7.743284657935661e-05}
{"train_loss": 0.2237900197505951, "global_step": 85887, "epoch": 965, "lr": 7.743236189012301e-05}
{"train_loss": 0.12608978152275085, "global_step": 85888, "epoch": 965, "lr": 7.743187719720146e-05}
{"train_loss": 0.21100124716758728, "global_step": 85889, "epoch": 965, "lr": 7.743139250059203e-05}
{"train_loss": 0.1622302383184433, "global_step": 85890, "epoch": 965, "lr": 7.743090780029476e-05}
{"train_loss": 0.196763813495636, "global_step": 85891, "epoch": 965, "lr": 7.743042309630977e-05}
{"train_loss": 0.17036277055740356, "global_step": 85892, "epoch": 965, "lr": 7.742993838863708e-05}
{"train_loss": 0.18373402953147888, "global_step": 85893, "epoch": 965, "lr": 7.742945367727677e-05}
{"train_loss": 0.16440601646900177, "global_step": 85894, "epoch": 965, "lr": 7.74289689622289e-05}
{"train_loss": 0.220173180103302, "global_step": 85895, "epoch": 965, "lr": 7.742848424349353e-05}
{"train_loss": 0.1435297578573227, "global_step": 85896, "epoch": 965, "lr": 7.742799952107076e-05}
{"train_loss": 0.14970359206199646, "global_step": 85897, "epoch": 965, "lr": 7.742751479496062e-05}
{"train_loss": 0.18472030758857727, "global_step": 85898, "epoch": 965, "lr": 7.742703006516318e-05}
{"train_loss": 0.20723488926887512, "global_step": 85899, "epoch": 965, "lr": 7.742654533167852e-05}
{"train_loss": 0.25372299551963806, "global_step": 85900, "epoch": 965, "lr": 7.742606059450668e-05}
{"train_loss": 0.19233867526054382, "global_step": 85901, "epoch": 965, "lr": 7.742557585364776e-05}
{"train_loss": 0.18034611642360687, "global_step": 85902, "epoch": 965, "lr": 7.742509110910179e-05}
{"train_loss": 0.17725926637649536, "global_step": 85903, "epoch": 965, "lr": 7.742460636086886e-05}
{"train_loss": 0.1777849793434143, "global_step": 85904, "epoch": 965, "lr": 7.742412160894901e-05}
{"train_loss": 0.16664406657218933, "global_step": 85905, "epoch": 965, "lr": 7.742363685334233e-05}
{"train_loss": 0.17270514369010925, "global_step": 85906, "epoch": 965, "lr": 7.742315209404888e-05}
{"train_loss": 0.18711644411087036, "global_step": 85907, "epoch": 965, "lr": 7.742266733106873e-05}
{"train_loss": 0.18593259155750275, "global_step": 85908, "epoch": 965, "lr": 7.742218256440193e-05}
{"train_loss": 0.17771273851394653, "global_step": 85909, "epoch": 965, "lr": 7.742169779404853e-05}
{"train_loss": 0.253844678401947, "global_step": 85910, "epoch": 965, "lr": 7.742121302000865e-05}
{"train_loss": 0.2859925329685211, "global_step": 85911, "epoch": 965, "lr": 7.74207282422823e-05}
{"train_loss": 0.19400303065776825, "global_step": 85912, "epoch": 965, "lr": 7.742024346086958e-05}
{"train_loss": 0.08344008028507233, "global_step": 85913, "epoch": 965, "lr": 7.741975867577052e-05}
{"train_loss": 0.17606127262115479, "global_step": 85914, "epoch": 965, "lr": 7.741927388698523e-05}
{"train_loss": 0.1493319869041443, "global_step": 85915, "epoch": 965, "lr": 7.741878909451375e-05}
{"train_loss": 0.15376421809196472, "global_step": 85916, "epoch": 965, "lr": 7.741830429835615e-05}
{"train_loss": 0.23317283391952515, "global_step": 85917, "epoch": 965, "lr": 7.741781949851249e-05}
{"train_loss": 0.13430340588092804, "global_step": 85918, "epoch": 965, "lr": 7.741733469498284e-05}
{"train_loss": 0.18879079818725586, "global_step": 85919, "epoch": 965, "lr": 7.741684988776725e-05}
{"train_loss": 0.21637333929538727, "global_step": 85920, "epoch": 965, "lr": 7.741636507686581e-05}
{"train_loss": 0.2155943065881729, "global_step": 85921, "epoch": 965, "lr": 7.741588026227857e-05}
{"train_loss": 0.15138404071331024, "global_step": 85922, "epoch": 965, "lr": 7.741539544400561e-05}
{"train_loss": 0.1908808797597885, "global_step": 85923, "epoch": 965, "lr": 7.741491062204698e-05}
{"train_loss": 0.17784662544727325, "global_step": 85924, "epoch": 965, "lr": 7.741442579640275e-05}
{"train_loss": 0.14793609082698822, "global_step": 85925, "epoch": 965, "lr": 7.741394096707298e-05}
{"train_loss": 0.1933647096157074, "global_step": 85926, "epoch": 965, "lr": 7.741345613405773e-05}
{"train_loss": 0.24012823402881622, "global_step": 85927, "epoch": 965, "lr": 7.741297129735709e-05}
{"train_loss": 0.1746096909046173, "global_step": 85928, "epoch": 965, "lr": 7.74124864569711e-05}
{"train_loss": 0.21777284145355225, "global_step": 85929, "epoch": 965, "lr": 7.741200161289985e-05}
{"train_loss": 0.18947894871234894, "global_step": 85930, "epoch": 965, "lr": 7.741151676514338e-05}
{"train_loss": 0.15014046430587769, "global_step": 85931, "epoch": 965, "lr": 7.741103191370176e-05}
{"train_loss": 0.22994549572467804, "global_step": 85932, "epoch": 965, "lr": 7.741054705857507e-05}
{"train_loss": 0.12926310300827026, "global_step": 85933, "epoch": 965, "lr": 7.741006219976337e-05}
{"train_loss": 0.26855674386024475, "global_step": 85934, "epoch": 965, "lr": 7.74095773372667e-05}
{"train_loss": 0.15836939215660095, "global_step": 85935, "epoch": 965, "lr": 7.740909247108517e-05}
{"train_loss": 0.139033243060112, "global_step": 85936, "epoch": 965, "lr": 7.740860760121882e-05}
{"train_loss": 0.15338923037052155, "global_step": 85937, "epoch": 965, "lr": 7.740812272766771e-05}
{"train_loss": 0.17042995989322662, "global_step": 85938, "epoch": 965, "lr": 7.74076378504319e-05}
{"train_loss": 0.19215072691440582, "global_step": 85939, "epoch": 965, "lr": 7.740715296951149e-05}
{"train_loss": 0.17530487477779388, "global_step": 85940, "epoch": 965, "lr": 7.740666808490651e-05}
{"train_loss": 0.2230415940284729, "global_step": 85941, "epoch": 965, "lr": 7.740618319661703e-05}
{"train_loss": 0.2697705328464508, "global_step": 85942, "epoch": 965, "lr": 7.740569830464312e-05}
{"train_loss": 0.20568086206912994, "global_step": 85943, "epoch": 965, "lr": 7.740521340898488e-05}
{"train_loss": 0.1843293011188507, "global_step": 85944, "epoch": 965, "lr": 7.740472850964231e-05}
{"train_loss": 0.2360088974237442, "global_step": 85945, "epoch": 965, "lr": 7.740424360661552e-05}
{"train_loss": 0.18905717134475708, "global_step": 85946, "epoch": 965, "lr": 7.740375869990457e-05}
{"train_loss": 0.1772032380104065, "global_step": 85947, "epoch": 965, "lr": 7.740327378950951e-05}
{"train_loss": 0.1758776754140854, "global_step": 85948, "epoch": 965, "lr": 7.740278887543042e-05}
{"train_loss": 0.25276196002960205, "global_step": 85949, "epoch": 965, "lr": 7.740230395766735e-05}
{"train_loss": 0.2179381549358368, "global_step": 85950, "epoch": 965, "lr": 7.740181903622038e-05}
{"train_loss": 0.25764188170433044, "global_step": 85951, "epoch": 965, "lr": 7.740133411108956e-05}
{"train_loss": 0.22317487001419067, "global_step": 85952, "epoch": 965, "lr": 7.740084918227498e-05}
{"train_loss": 0.22714416682720184, "global_step": 85953, "epoch": 965, "lr": 7.740036424977669e-05}
{"train_loss": 0.19878363609313965, "global_step": 85954, "epoch": 965, "lr": 7.739987931359475e-05}
{"train_loss": 0.18276219069957733, "global_step": 85955, "epoch": 965, "lr": 7.739939437372923e-05}
{"train_loss": 0.17111735045909882, "global_step": 85956, "epoch": 965, "lr": 7.739890943018019e-05}
{"train_loss": 0.2537853419780731, "global_step": 85957, "epoch": 965, "lr": 7.739842448294771e-05}
{"train_loss": 0.1253463327884674, "global_step": 85958, "epoch": 965, "lr": 7.739793953203182e-05}
{"train_loss": 0.3457220494747162, "global_step": 85959, "epoch": 965, "lr": 7.739745457743264e-05}
{"train_loss": 0.16746723651885986, "global_step": 85960, "epoch": 965, "lr": 7.73969696191502e-05}
{"train_loss": 0.20175911486148834, "global_step": 85961, "epoch": 965, "lr": 7.739648465718456e-05}
{"train_loss": 0.2720477879047394, "global_step": 85962, "epoch": 965, "lr": 7.739599969153581e-05}
{"train_loss": 0.1852930784225464, "global_step": 85963, "epoch": 965, "lr": 7.7395514722204e-05}
{"train_loss": 0.25076737999916077, "global_step": 85964, "epoch": 965, "lr": 7.73950297491892e-05}
{"train_loss": 0.15419307351112366, "global_step": 85965, "epoch": 965, "lr": 7.739454477249147e-05}
{"train_loss": 0.293226033449173, "global_step": 85966, "epoch": 965, "lr": 7.739405979211086e-05}
{"train_loss": 0.13104209303855896, "global_step": 85967, "epoch": 965, "lr": 7.739357480804747e-05}
{"train_loss": 0.2511524558067322, "global_step": 85968, "epoch": 965, "lr": 7.739308982030134e-05}
{"train_loss": 0.18122707307338715, "global_step": 85969, "epoch": 965, "lr": 7.739260482887255e-05}
{"train_loss": 0.21386469900608063, "global_step": 85970, "epoch": 965, "lr": 7.739211983376115e-05}
{"train_loss": 0.20025432109832764, "global_step": 85971, "epoch": 965, "lr": 7.739163483496723e-05}
{"train_loss": 0.17490828037261963, "global_step": 85972, "epoch": 965, "lr": 7.739114983249083e-05}
{"train_loss": 0.19574165963724757, "global_step": 85973, "epoch": 965, "lr": 7.739066482633201e-05, "val_loss": 4.120913505554199, "train_action_mse_error": 15.963653564453125}
{"train_loss": 0.19245333969593048, "global_step": 85974, "epoch": 966, "lr": 7.739017981649086e-05}
{"train_loss": 0.25373321771621704, "global_step": 85975, "epoch": 966, "lr": 7.738969480296744e-05}
{"train_loss": 0.1619539111852646, "global_step": 85976, "epoch": 966, "lr": 7.73892097857618e-05}
{"train_loss": 0.13991108536720276, "global_step": 85977, "epoch": 966, "lr": 7.738872476487402e-05}
{"train_loss": 0.139445960521698, "global_step": 85978, "epoch": 966, "lr": 7.738823974030416e-05}
{"train_loss": 0.17862680554389954, "global_step": 85979, "epoch": 966, "lr": 7.738775471205229e-05}
{"train_loss": 0.23541076481342316, "global_step": 85980, "epoch": 966, "lr": 7.738726968011844e-05}
{"train_loss": 0.12781131267547607, "global_step": 85981, "epoch": 966, "lr": 7.738678464450273e-05}
{"train_loss": 0.21774648129940033, "global_step": 85982, "epoch": 966, "lr": 7.738629960520521e-05}
{"train_loss": 0.18585416674613953, "global_step": 85983, "epoch": 966, "lr": 7.738581456222592e-05}
{"train_loss": 0.24170605838298798, "global_step": 85984, "epoch": 966, "lr": 7.738532951556495e-05}
{"train_loss": 0.09055027365684509, "global_step": 85985, "epoch": 966, "lr": 7.738484446522233e-05}
{"train_loss": 0.15956996381282806, "global_step": 85986, "epoch": 966, "lr": 7.73843594111982e-05}
{"train_loss": 0.2020820826292038, "global_step": 85987, "epoch": 966, "lr": 7.738387435349255e-05}
{"train_loss": 0.2087557315826416, "global_step": 85988, "epoch": 966, "lr": 7.738338929210547e-05}
{"train_loss": 0.23458179831504822, "global_step": 85989, "epoch": 966, "lr": 7.738290422703702e-05}
{"train_loss": 0.13522782921791077, "global_step": 85990, "epoch": 966, "lr": 7.738241915828727e-05}
{"train_loss": 0.1449323147535324, "global_step": 85991, "epoch": 966, "lr": 7.738193408585631e-05}
{"train_loss": 0.1553342342376709, "global_step": 85992, "epoch": 966, "lr": 7.738144900974418e-05}
{"train_loss": 0.17751505970954895, "global_step": 85993, "epoch": 966, "lr": 7.738096392995094e-05}
{"train_loss": 0.27353742718696594, "global_step": 85994, "epoch": 966, "lr": 7.738047884647665e-05}
{"train_loss": 0.15981245040893555, "global_step": 85995, "epoch": 966, "lr": 7.737999375932141e-05}
{"train_loss": 0.1835528016090393, "global_step": 85996, "epoch": 966, "lr": 7.737950866848524e-05}
{"train_loss": 0.21290676295757294, "global_step": 85997, "epoch": 966, "lr": 7.737902357396826e-05}
{"train_loss": 0.17046180367469788, "global_step": 85998, "epoch": 966, "lr": 7.737853847577047e-05}
{"train_loss": 0.18489781022071838, "global_step": 85999, "epoch": 966, "lr": 7.737805337389199e-05}
{"train_loss": 0.1596221923828125, "global_step": 86000, "epoch": 966, "lr": 7.737756826833286e-05}
{"train_loss": 0.2533622980117798, "global_step": 86001, "epoch": 966, "lr": 7.737708315909315e-05}
{"train_loss": 0.11193367093801498, "global_step": 86002, "epoch": 966, "lr": 7.737659804617293e-05}
{"train_loss": 0.20564357936382294, "global_step": 86003, "epoch": 966, "lr": 7.737611292957225e-05}
{"train_loss": 0.18347996473312378, "global_step": 86004, "epoch": 966, "lr": 7.737562780929119e-05}
{"train_loss": 0.20257940888404846, "global_step": 86005, "epoch": 966, "lr": 7.737514268532982e-05}
{"train_loss": 0.17933639883995056, "global_step": 86006, "epoch": 966, "lr": 7.737465755768819e-05}
{"train_loss": 0.1885586977005005, "global_step": 86007, "epoch": 966, "lr": 7.737417242636635e-05}
{"train_loss": 0.1206323429942131, "global_step": 86008, "epoch": 966, "lr": 7.737368729136441e-05}
{"train_loss": 0.21195557713508606, "global_step": 86009, "epoch": 966, "lr": 7.73732021526824e-05}
{"train_loss": 0.15557880699634552, "global_step": 86010, "epoch": 966, "lr": 7.737271701032041e-05}
{"train_loss": 0.21880118548870087, "global_step": 86011, "epoch": 966, "lr": 7.737223186427849e-05}
{"train_loss": 0.15310214459896088, "global_step": 86012, "epoch": 966, "lr": 7.73717467145567e-05}
{"train_loss": 0.3444553017616272, "global_step": 86013, "epoch": 966, "lr": 7.73712615611551e-05}
{"train_loss": 0.22429336607456207, "global_step": 86014, "epoch": 966, "lr": 7.73707764040738e-05}
{"train_loss": 0.19484220445156097, "global_step": 86015, "epoch": 966, "lr": 7.73702912433128e-05}
{"train_loss": 0.16162465512752533, "global_step": 86016, "epoch": 966, "lr": 7.736980607887221e-05}
{"train_loss": 0.1811274141073227, "global_step": 86017, "epoch": 966, "lr": 7.736932091075208e-05}
{"train_loss": 0.1099168211221695, "global_step": 86018, "epoch": 966, "lr": 7.736883573895248e-05}
{"train_loss": 0.15786512196063995, "global_step": 86019, "epoch": 966, "lr": 7.736835056347348e-05}
{"train_loss": 0.2149282544851303, "global_step": 86020, "epoch": 966, "lr": 7.736786538431513e-05}
{"train_loss": 0.23130548000335693, "global_step": 86021, "epoch": 966, "lr": 7.736738020147751e-05}
{"train_loss": 0.19013860821723938, "global_step": 86022, "epoch": 966, "lr": 7.736689501496068e-05}
{"train_loss": 0.18288852274417877, "global_step": 86023, "epoch": 966, "lr": 7.73664098247647e-05}
{"train_loss": 0.23936578631401062, "global_step": 86024, "epoch": 966, "lr": 7.736592463088965e-05}
{"train_loss": 0.1392671763896942, "global_step": 86025, "epoch": 966, "lr": 7.736543943333559e-05}
{"train_loss": 0.22413362562656403, "global_step": 86026, "epoch": 966, "lr": 7.736495423210255e-05}
{"train_loss": 0.10283124446868896, "global_step": 86027, "epoch": 966, "lr": 7.736446902719066e-05}
{"train_loss": 0.15328966081142426, "global_step": 86028, "epoch": 966, "lr": 7.736398381859992e-05}
{"train_loss": 0.2706281542778015, "global_step": 86029, "epoch": 966, "lr": 7.736349860633044e-05}
{"train_loss": 0.20617452263832092, "global_step": 86030, "epoch": 966, "lr": 7.736301339038228e-05}
{"train_loss": 0.13094370067119598, "global_step": 86031, "epoch": 966, "lr": 7.736252817075549e-05}
{"train_loss": 0.19158069789409637, "global_step": 86032, "epoch": 966, "lr": 7.736204294745013e-05}
{"train_loss": 0.14839893579483032, "global_step": 86033, "epoch": 966, "lr": 7.736155772046629e-05}
{"train_loss": 0.15130436420440674, "global_step": 86034, "epoch": 966, "lr": 7.736107248980402e-05}
{"train_loss": 0.18006163835525513, "global_step": 86035, "epoch": 966, "lr": 7.736058725546338e-05}
{"train_loss": 0.198513463139534, "global_step": 86036, "epoch": 966, "lr": 7.736010201744444e-05}
{"train_loss": 0.176815465092659, "global_step": 86037, "epoch": 966, "lr": 7.735961677574729e-05}
{"train_loss": 0.2007908970117569, "global_step": 86038, "epoch": 966, "lr": 7.735913153037195e-05}
{"train_loss": 0.21954114735126495, "global_step": 86039, "epoch": 966, "lr": 7.735864628131851e-05}
{"train_loss": 0.19816471636295319, "global_step": 86040, "epoch": 966, "lr": 7.735816102858704e-05}
{"train_loss": 0.18658529222011566, "global_step": 86041, "epoch": 966, "lr": 7.735767577217761e-05}
{"train_loss": 0.19620496034622192, "global_step": 86042, "epoch": 966, "lr": 7.735719051209025e-05}
{"train_loss": 0.13412787020206451, "global_step": 86043, "epoch": 966, "lr": 7.735670524832507e-05}
{"train_loss": 0.19074836373329163, "global_step": 86044, "epoch": 966, "lr": 7.735621998088211e-05}
{"train_loss": 0.26101154088974, "global_step": 86045, "epoch": 966, "lr": 7.735573470976143e-05}
{"train_loss": 0.28838345408439636, "global_step": 86046, "epoch": 966, "lr": 7.73552494349631e-05}
{"train_loss": 0.16956868767738342, "global_step": 86047, "epoch": 966, "lr": 7.735476415648721e-05}
{"train_loss": 0.1770447939634323, "global_step": 86048, "epoch": 966, "lr": 7.73542788743338e-05}
{"train_loss": 0.19497092068195343, "global_step": 86049, "epoch": 966, "lr": 7.735379358850293e-05}
{"train_loss": 0.24657419323921204, "global_step": 86050, "epoch": 966, "lr": 7.73533082989947e-05}
{"train_loss": 0.18666577339172363, "global_step": 86051, "epoch": 966, "lr": 7.735282300580913e-05}
{"train_loss": 0.12547633051872253, "global_step": 86052, "epoch": 966, "lr": 7.73523377089463e-05}
{"train_loss": 0.2030075192451477, "global_step": 86053, "epoch": 966, "lr": 7.73518524084063e-05}
{"train_loss": 0.19786055386066437, "global_step": 86054, "epoch": 966, "lr": 7.735136710418915e-05}
{"train_loss": 0.22307100892066956, "global_step": 86055, "epoch": 966, "lr": 7.735088179629497e-05}
{"train_loss": 0.16768237948417664, "global_step": 86056, "epoch": 966, "lr": 7.735039648472379e-05}
{"train_loss": 0.19196411967277527, "global_step": 86057, "epoch": 966, "lr": 7.734991116947568e-05}
{"train_loss": 0.1676052212715149, "global_step": 86058, "epoch": 966, "lr": 7.73494258505507e-05}
{"train_loss": 0.19222556054592133, "global_step": 86059, "epoch": 966, "lr": 7.734894052794894e-05}
{"train_loss": 0.2467166781425476, "global_step": 86060, "epoch": 966, "lr": 7.734845520167044e-05}
{"train_loss": 0.16037575900554657, "global_step": 86061, "epoch": 966, "lr": 7.734796987171526e-05}
{"train_loss": 0.1892827344074678, "global_step": 86062, "epoch": 966, "lr": 7.73474845380835e-05, "val_loss": 4.245323181152344}
{"train_loss": 0.20505429804325104, "global_step": 86063, "epoch": 967, "lr": 7.73469992007752e-05}
{"train_loss": 0.13724423944950104, "global_step": 86064, "epoch": 967, "lr": 7.734651385979041e-05}
{"train_loss": 0.19922509789466858, "global_step": 86065, "epoch": 967, "lr": 7.734602851512924e-05}
{"train_loss": 0.19224996864795685, "global_step": 86066, "epoch": 967, "lr": 7.734554316679172e-05}
{"train_loss": 0.14111486077308655, "global_step": 86067, "epoch": 967, "lr": 7.734505781477792e-05}
{"train_loss": 0.24325406551361084, "global_step": 86068, "epoch": 967, "lr": 7.734457245908793e-05}
{"train_loss": 0.21526165306568146, "global_step": 86069, "epoch": 967, "lr": 7.734408709972177e-05}
{"train_loss": 0.2150246500968933, "global_step": 86070, "epoch": 967, "lr": 7.734360173667954e-05}
{"train_loss": 0.1704583764076233, "global_step": 86071, "epoch": 967, "lr": 7.73431163699613e-05}
{"train_loss": 0.16886326670646667, "global_step": 86072, "epoch": 967, "lr": 7.73426309995671e-05}
{"train_loss": 0.16284491121768951, "global_step": 86073, "epoch": 967, "lr": 7.734214562549702e-05}
{"train_loss": 0.14922171831130981, "global_step": 86074, "epoch": 967, "lr": 7.734166024775113e-05}
{"train_loss": 0.1803913712501526, "global_step": 86075, "epoch": 967, "lr": 7.734117486632948e-05}
{"train_loss": 0.11146936565637589, "global_step": 86076, "epoch": 967, "lr": 7.734068948123216e-05}
{"train_loss": 0.15529367327690125, "global_step": 86077, "epoch": 967, "lr": 7.73402040924592e-05}
{"train_loss": 0.27133840322494507, "global_step": 86078, "epoch": 967, "lr": 7.733971870001069e-05}
{"train_loss": 0.2751717269420624, "global_step": 86079, "epoch": 967, "lr": 7.733923330388668e-05}
{"train_loss": 0.17816351354122162, "global_step": 86080, "epoch": 967, "lr": 7.733874790408725e-05}
{"train_loss": 0.22347907721996307, "global_step": 86081, "epoch": 967, "lr": 7.733826250061246e-05}
{"train_loss": 0.1719292849302292, "global_step": 86082, "epoch": 967, "lr": 7.733777709346237e-05}
{"train_loss": 0.19020800292491913, "global_step": 86083, "epoch": 967, "lr": 7.733729168263704e-05}
{"train_loss": 0.1827041655778885, "global_step": 86084, "epoch": 967, "lr": 7.733680626813656e-05}
{"train_loss": 0.1521090716123581, "global_step": 86085, "epoch": 967, "lr": 7.733632084996096e-05}
{"train_loss": 0.19681088626384735, "global_step": 86086, "epoch": 967, "lr": 7.733583542811035e-05}
{"train_loss": 0.2440238744020462, "global_step": 86087, "epoch": 967, "lr": 7.733535000258477e-05}
{"train_loss": 0.19233107566833496, "global_step": 86088, "epoch": 967, "lr": 7.733486457338425e-05}
{"train_loss": 0.167787566781044, "global_step": 86089, "epoch": 967, "lr": 7.733437914050893e-05}
{"train_loss": 0.2539273798465729, "global_step": 86090, "epoch": 967, "lr": 7.73338937039588e-05}
{"train_loss": 0.16976985335350037, "global_step": 86091, "epoch": 967, "lr": 7.733340826373399e-05}
{"train_loss": 0.1563590168952942, "global_step": 86092, "epoch": 967, "lr": 7.733292281983453e-05}
{"train_loss": 0.2160244584083557, "global_step": 86093, "epoch": 967, "lr": 7.733243737226048e-05}
{"train_loss": 0.1020738035440445, "global_step": 86094, "epoch": 967, "lr": 7.733195192101193e-05}
{"train_loss": 0.19295431673526764, "global_step": 86095, "epoch": 967, "lr": 7.733146646608894e-05}
{"train_loss": 0.21419188380241394, "global_step": 86096, "epoch": 967, "lr": 7.733098100749154e-05}
{"train_loss": 0.20311881601810455, "global_step": 86097, "epoch": 967, "lr": 7.733049554521986e-05}
{"train_loss": 0.2743799388408661, "global_step": 86098, "epoch": 967, "lr": 7.73300100792739e-05}
{"train_loss": 0.20155243575572968, "global_step": 86099, "epoch": 967, "lr": 7.732952460965375e-05}
{"train_loss": 0.1927023082971573, "global_step": 86100, "epoch": 967, "lr": 7.73290391363595e-05}
{"train_loss": 0.23459826409816742, "global_step": 86101, "epoch": 967, "lr": 7.732855365939118e-05}
{"train_loss": 0.09666179120540619, "global_step": 86102, "epoch": 967, "lr": 7.732806817874886e-05}
{"train_loss": 0.12386313825845718, "global_step": 86103, "epoch": 967, "lr": 7.732758269443264e-05}
{"train_loss": 0.25769808888435364, "global_step": 86104, "epoch": 967, "lr": 7.732709720644253e-05}
{"train_loss": 0.13597644865512848, "global_step": 86105, "epoch": 967, "lr": 7.732661171477864e-05}
{"train_loss": 0.23186908662319183, "global_step": 86106, "epoch": 967, "lr": 7.732612621944102e-05}
{"train_loss": 0.14426949620246887, "global_step": 86107, "epoch": 967, "lr": 7.732564072042974e-05}
{"train_loss": 0.14268887042999268, "global_step": 86108, "epoch": 967, "lr": 7.732515521774485e-05}
{"train_loss": 0.21508757770061493, "global_step": 86109, "epoch": 967, "lr": 7.732466971138644e-05}
{"train_loss": 0.26341521739959717, "global_step": 86110, "epoch": 967, "lr": 7.732418420135455e-05}
{"train_loss": 0.15110896527767181, "global_step": 86111, "epoch": 967, "lr": 7.732369868764925e-05}
{"train_loss": 0.13090862333774567, "global_step": 86112, "epoch": 967, "lr": 7.732321317027062e-05}
{"train_loss": 0.2008671760559082, "global_step": 86113, "epoch": 967, "lr": 7.732272764921871e-05}
{"train_loss": 0.199421226978302, "global_step": 86114, "epoch": 967, "lr": 7.732224212449362e-05}
{"train_loss": 0.24205568432807922, "global_step": 86115, "epoch": 967, "lr": 7.732175659609536e-05}
{"train_loss": 0.21019339561462402, "global_step": 86116, "epoch": 967, "lr": 7.732127106402402e-05}
{"train_loss": 0.26142793893814087, "global_step": 86117, "epoch": 967, "lr": 7.73207855282797e-05}
{"train_loss": 0.16697195172309875, "global_step": 86118, "epoch": 967, "lr": 7.732029998886242e-05}
{"train_loss": 0.1763932704925537, "global_step": 86119, "epoch": 967, "lr": 7.731981444577223e-05}
{"train_loss": 0.19210785627365112, "global_step": 86120, "epoch": 967, "lr": 7.731932889900926e-05}
{"train_loss": 0.23541203141212463, "global_step": 86121, "epoch": 967, "lr": 7.73188433485735e-05}
{"train_loss": 0.15276220440864563, "global_step": 86122, "epoch": 967, "lr": 7.73183577944651e-05}
{"train_loss": 0.1379265934228897, "global_step": 86123, "epoch": 967, "lr": 7.731787223668405e-05}
{"train_loss": 0.12229865044355392, "global_step": 86124, "epoch": 967, "lr": 7.731738667523046e-05}
{"train_loss": 0.10188726335763931, "global_step": 86125, "epoch": 967, "lr": 7.731690111010437e-05}
{"train_loss": 0.16729921102523804, "global_step": 86126, "epoch": 967, "lr": 7.731641554130585e-05}
{"train_loss": 0.21913129091262817, "global_step": 86127, "epoch": 967, "lr": 7.731592996883499e-05}
{"train_loss": 0.30033472180366516, "global_step": 86128, "epoch": 967, "lr": 7.731544439269182e-05}
{"train_loss": 0.250401109457016, "global_step": 86129, "epoch": 967, "lr": 7.731495881287642e-05}
{"train_loss": 0.1791619509458542, "global_step": 86130, "epoch": 967, "lr": 7.731447322938887e-05}
{"train_loss": 0.17936700582504272, "global_step": 86131, "epoch": 967, "lr": 7.731398764222921e-05}
{"train_loss": 0.11588093638420105, "global_step": 86132, "epoch": 967, "lr": 7.731350205139754e-05}
{"train_loss": 0.16057968139648438, "global_step": 86133, "epoch": 967, "lr": 7.731301645689388e-05}
{"train_loss": 0.1864054799079895, "global_step": 86134, "epoch": 967, "lr": 7.731253085871832e-05}
{"train_loss": 0.16202321648597717, "global_step": 86135, "epoch": 967, "lr": 7.731204525687092e-05}
{"train_loss": 0.18010327219963074, "global_step": 86136, "epoch": 967, "lr": 7.731155965135177e-05}
{"train_loss": 0.28792694211006165, "global_step": 86137, "epoch": 967, "lr": 7.731107404216089e-05}
{"train_loss": 0.14435212314128876, "global_step": 86138, "epoch": 967, "lr": 7.731058842929839e-05}
{"train_loss": 0.20627355575561523, "global_step": 86139, "epoch": 967, "lr": 7.73101028127643e-05}
{"train_loss": 0.195302352309227, "global_step": 86140, "epoch": 967, "lr": 7.730961719255869e-05}
{"train_loss": 0.1663534939289093, "global_step": 86141, "epoch": 967, "lr": 7.730913156868166e-05}
{"train_loss": 0.16810719668865204, "global_step": 86142, "epoch": 967, "lr": 7.730864594113322e-05}
{"train_loss": 0.18075944483280182, "global_step": 86143, "epoch": 967, "lr": 7.73081603099135e-05}
{"train_loss": 0.18869060277938843, "global_step": 86144, "epoch": 967, "lr": 7.73076746750225e-05}
{"train_loss": 0.2012224942445755, "global_step": 86145, "epoch": 967, "lr": 7.730718903646032e-05}
{"train_loss": 0.2048848569393158, "global_step": 86146, "epoch": 967, "lr": 7.730670339422704e-05}
{"train_loss": 0.17826464772224426, "global_step": 86147, "epoch": 967, "lr": 7.730621774832269e-05}
{"train_loss": 0.15497970581054688, "global_step": 86148, "epoch": 967, "lr": 7.730573209874737e-05}
{"train_loss": 0.1362900733947754, "global_step": 86149, "epoch": 967, "lr": 7.730524644550111e-05}
{"train_loss": 0.18283753097057343, "global_step": 86150, "epoch": 967, "lr": 7.730476078858399e-05}
{"train_loss": 0.18677469133661034, "global_step": 86151, "epoch": 967, "lr": 7.73042751279961e-05, "val_loss": 4.1078290939331055}
{"train_loss": 0.18988783657550812, "global_step": 86152, "epoch": 968, "lr": 7.730378946373747e-05}
{"train_loss": 0.135584756731987, "global_step": 86153, "epoch": 968, "lr": 7.730330379580817e-05}
{"train_loss": 0.17648737132549286, "global_step": 86154, "epoch": 968, "lr": 7.730281812420829e-05}
{"train_loss": 0.18646715581417084, "global_step": 86155, "epoch": 968, "lr": 7.730233244893787e-05}
{"train_loss": 0.16227108240127563, "global_step": 86156, "epoch": 968, "lr": 7.730184676999699e-05}
{"train_loss": 0.2718612849712372, "global_step": 86157, "epoch": 968, "lr": 7.73013610873857e-05}
{"train_loss": 0.198963925242424, "global_step": 86158, "epoch": 968, "lr": 7.730087540110407e-05}
{"train_loss": 0.2985197901725769, "global_step": 86159, "epoch": 968, "lr": 7.730038971115219e-05}
{"train_loss": 0.17063336074352264, "global_step": 86160, "epoch": 968, "lr": 7.729990401753008e-05}
{"train_loss": 0.2059556245803833, "global_step": 86161, "epoch": 968, "lr": 7.729941832023784e-05}
{"train_loss": 0.2656412124633789, "global_step": 86162, "epoch": 968, "lr": 7.729893261927554e-05}
{"train_loss": 0.22499576210975647, "global_step": 86163, "epoch": 968, "lr": 7.729844691464321e-05}
{"train_loss": 0.1229218989610672, "global_step": 86164, "epoch": 968, "lr": 7.729796120634095e-05}
{"train_loss": 0.2292407751083374, "global_step": 86165, "epoch": 968, "lr": 7.729747549436882e-05}
{"train_loss": 0.1941167563199997, "global_step": 86166, "epoch": 968, "lr": 7.729698977872686e-05}
{"train_loss": 0.3055480718612671, "global_step": 86167, "epoch": 968, "lr": 7.729650405941515e-05}
{"train_loss": 0.1817864030599594, "global_step": 86168, "epoch": 968, "lr": 7.729601833643378e-05}
{"train_loss": 0.18610766530036926, "global_step": 86169, "epoch": 968, "lr": 7.729553260978277e-05}
{"train_loss": 0.19178198277950287, "global_step": 86170, "epoch": 968, "lr": 7.729504687946222e-05}
{"train_loss": 0.18393772840499878, "global_step": 86171, "epoch": 968, "lr": 7.729456114547219e-05}
{"train_loss": 0.16534367203712463, "global_step": 86172, "epoch": 968, "lr": 7.729407540781272e-05}
{"train_loss": 0.21546262502670288, "global_step": 86173, "epoch": 968, "lr": 7.729358966648393e-05}
{"train_loss": 0.14690415561199188, "global_step": 86174, "epoch": 968, "lr": 7.72931039214858e-05}
{"train_loss": 0.17336180806159973, "global_step": 86175, "epoch": 968, "lr": 7.729261817281848e-05}
{"train_loss": 0.3014675974845886, "global_step": 86176, "epoch": 968, "lr": 7.729213242048199e-05}
{"train_loss": 0.16987015306949615, "global_step": 86177, "epoch": 968, "lr": 7.72916466644764e-05}
{"train_loss": 0.2485882192850113, "global_step": 86178, "epoch": 968, "lr": 7.72911609048018e-05}
{"train_loss": 0.24932920932769775, "global_step": 86179, "epoch": 968, "lr": 7.729067514145822e-05}
{"train_loss": 0.2660363018512726, "global_step": 86180, "epoch": 968, "lr": 7.729018937444574e-05}
{"train_loss": 0.20067563652992249, "global_step": 86181, "epoch": 968, "lr": 7.728970360376442e-05}
{"train_loss": 0.23461836576461792, "global_step": 86182, "epoch": 968, "lr": 7.728921782941434e-05}
{"train_loss": 0.20100301504135132, "global_step": 86183, "epoch": 968, "lr": 7.728873205139556e-05}
{"train_loss": 0.1593734174966812, "global_step": 86184, "epoch": 968, "lr": 7.728824626970814e-05}
{"train_loss": 0.23398320376873016, "global_step": 86185, "epoch": 968, "lr": 7.728776048435216e-05}
{"train_loss": 0.30464479327201843, "global_step": 86186, "epoch": 968, "lr": 7.728727469532766e-05}
{"train_loss": 0.18430954217910767, "global_step": 86187, "epoch": 968, "lr": 7.728678890263472e-05}
{"train_loss": 0.19680781662464142, "global_step": 86188, "epoch": 968, "lr": 7.72863031062734e-05}
{"train_loss": 0.26109778881073, "global_step": 86189, "epoch": 968, "lr": 7.728581730624378e-05}
{"train_loss": 0.23393514752388, "global_step": 86190, "epoch": 968, "lr": 7.728533150254592e-05}
{"train_loss": 0.16290758550167084, "global_step": 86191, "epoch": 968, "lr": 7.728484569517985e-05}
{"train_loss": 0.17818233370780945, "global_step": 86192, "epoch": 968, "lr": 7.728435988414569e-05}
{"train_loss": 0.19906377792358398, "global_step": 86193, "epoch": 968, "lr": 7.728387406944348e-05}
{"train_loss": 0.21529266238212585, "global_step": 86194, "epoch": 968, "lr": 7.728338825107327e-05}
{"train_loss": 0.2142363041639328, "global_step": 86195, "epoch": 968, "lr": 7.728290242903515e-05}
{"train_loss": 0.238990917801857, "global_step": 86196, "epoch": 968, "lr": 7.728241660332918e-05}
{"train_loss": 0.2753552496433258, "global_step": 86197, "epoch": 968, "lr": 7.728193077395542e-05}
{"train_loss": 0.17897161841392517, "global_step": 86198, "epoch": 968, "lr": 7.728144494091394e-05}
{"train_loss": 0.2158452272415161, "global_step": 86199, "epoch": 968, "lr": 7.728095910420478e-05}
{"train_loss": 0.2358960509300232, "global_step": 86200, "epoch": 968, "lr": 7.728047326382806e-05}
{"train_loss": 0.14622455835342407, "global_step": 86201, "epoch": 968, "lr": 7.72799874197838e-05}
{"train_loss": 0.19800648093223572, "global_step": 86202, "epoch": 968, "lr": 7.727950157207206e-05}
{"train_loss": 0.13707436621189117, "global_step": 86203, "epoch": 968, "lr": 7.727901572069294e-05}
{"train_loss": 0.11906330287456512, "global_step": 86204, "epoch": 968, "lr": 7.72785298656465e-05}
{"train_loss": 0.2537720799446106, "global_step": 86205, "epoch": 968, "lr": 7.727804400693277e-05}
{"train_loss": 0.13749919831752777, "global_step": 86206, "epoch": 968, "lr": 7.727755814455187e-05}
{"train_loss": 0.24514615535736084, "global_step": 86207, "epoch": 968, "lr": 7.72770722785038e-05}
{"train_loss": 0.1340910792350769, "global_step": 86208, "epoch": 968, "lr": 7.727658640878867e-05}
{"train_loss": 0.20958885550498962, "global_step": 86209, "epoch": 968, "lr": 7.727610053540655e-05}
{"train_loss": 0.16225551068782806, "global_step": 86210, "epoch": 968, "lr": 7.727561465835749e-05}
{"train_loss": 0.13417567312717438, "global_step": 86211, "epoch": 968, "lr": 7.727512877764155e-05}
{"train_loss": 0.2656799256801605, "global_step": 86212, "epoch": 968, "lr": 7.72746428932588e-05}
{"train_loss": 0.23843643069267273, "global_step": 86213, "epoch": 968, "lr": 7.727415700520931e-05}
{"train_loss": 0.14748093485832214, "global_step": 86214, "epoch": 968, "lr": 7.727367111349312e-05}
{"train_loss": 0.15081317722797394, "global_step": 86215, "epoch": 968, "lr": 7.727318521811036e-05}
{"train_loss": 0.20398016273975372, "global_step": 86216, "epoch": 968, "lr": 7.727269931906102e-05}
{"train_loss": 0.21166974306106567, "global_step": 86217, "epoch": 968, "lr": 7.727221341634521e-05}
{"train_loss": 0.1905185580253601, "global_step": 86218, "epoch": 968, "lr": 7.727172750996297e-05}
{"train_loss": 0.16603396832942963, "global_step": 86219, "epoch": 968, "lr": 7.727124159991439e-05}
{"train_loss": 0.2057936191558838, "global_step": 86220, "epoch": 968, "lr": 7.727075568619953e-05}
{"train_loss": 0.245110884308815, "global_step": 86221, "epoch": 968, "lr": 7.727026976881843e-05}
{"train_loss": 0.27494654059410095, "global_step": 86222, "epoch": 968, "lr": 7.726978384777119e-05}
{"train_loss": 0.13079628348350525, "global_step": 86223, "epoch": 968, "lr": 7.726929792305785e-05}
{"train_loss": 0.19014355540275574, "global_step": 86224, "epoch": 968, "lr": 7.72688119946785e-05}
{"train_loss": 0.13616174459457397, "global_step": 86225, "epoch": 968, "lr": 7.726832606263319e-05}
{"train_loss": 0.164001926779747, "global_step": 86226, "epoch": 968, "lr": 7.726784012692197e-05}
{"train_loss": 0.15434031188488007, "global_step": 86227, "epoch": 968, "lr": 7.726735418754494e-05}
{"train_loss": 0.1757119745016098, "global_step": 86228, "epoch": 968, "lr": 7.726686824450213e-05}
{"train_loss": 0.22473540902137756, "global_step": 86229, "epoch": 968, "lr": 7.726638229779364e-05}
{"train_loss": 0.13963226974010468, "global_step": 86230, "epoch": 968, "lr": 7.726589634741949e-05}
{"train_loss": 0.17865939438343048, "global_step": 86231, "epoch": 968, "lr": 7.72654103933798e-05}
{"train_loss": 0.16414135694503784, "global_step": 86232, "epoch": 968, "lr": 7.726492443567459e-05}
{"train_loss": 0.19700771570205688, "global_step": 86233, "epoch": 968, "lr": 7.726443847430395e-05}
{"train_loss": 0.18405351042747498, "global_step": 86234, "epoch": 968, "lr": 7.726395250926795e-05}
{"train_loss": 0.19172722101211548, "global_step": 86235, "epoch": 968, "lr": 7.726346654056662e-05}
{"train_loss": 0.15871651470661163, "global_step": 86236, "epoch": 968, "lr": 7.726298056820005e-05}
{"train_loss": 0.1667437106370926, "global_step": 86237, "epoch": 968, "lr": 7.726249459216832e-05}
{"train_loss": 0.23760144412517548, "global_step": 86238, "epoch": 968, "lr": 7.726200861247149e-05}
{"train_loss": 0.10369683802127838, "global_step": 86239, "epoch": 968, "lr": 7.726152262910959e-05}
{"train_loss": 0.19675584179296923, "global_step": 86240, "epoch": 968, "lr": 7.726103664208272e-05, "val_loss": 4.201237678527832}
{"train_loss": 0.28873223066329956, "global_step": 86241, "epoch": 969, "lr": 7.726055065139094e-05}
{"train_loss": 0.15395145118236542, "global_step": 86242, "epoch": 969, "lr": 7.72600646570343e-05}
{"train_loss": 0.13944391906261444, "global_step": 86243, "epoch": 969, "lr": 7.72595786590129e-05}
{"train_loss": 0.24079914391040802, "global_step": 86244, "epoch": 969, "lr": 7.725909265732676e-05}
{"train_loss": 0.16777218878269196, "global_step": 86245, "epoch": 969, "lr": 7.725860665197598e-05}
{"train_loss": 0.16340214014053345, "global_step": 86246, "epoch": 969, "lr": 7.725812064296061e-05}
{"train_loss": 0.11527659744024277, "global_step": 86247, "epoch": 969, "lr": 7.725763463028071e-05}
{"train_loss": 0.18984420597553253, "global_step": 86248, "epoch": 969, "lr": 7.725714861393637e-05}
{"train_loss": 0.17309392988681793, "global_step": 86249, "epoch": 969, "lr": 7.725666259392761e-05}
{"train_loss": 0.14366838335990906, "global_step": 86250, "epoch": 969, "lr": 7.725617657025455e-05}
{"train_loss": 0.16454634070396423, "global_step": 86251, "epoch": 969, "lr": 7.725569054291722e-05}
{"train_loss": 0.22282765805721283, "global_step": 86252, "epoch": 969, "lr": 7.725520451191568e-05}
{"train_loss": 0.15257973968982697, "global_step": 86253, "epoch": 969, "lr": 7.725471847725001e-05}
{"train_loss": 0.2015305459499359, "global_step": 86254, "epoch": 969, "lr": 7.725423243892029e-05}
{"train_loss": 0.1687086969614029, "global_step": 86255, "epoch": 969, "lr": 7.725374639692658e-05}
{"train_loss": 0.15408460795879364, "global_step": 86256, "epoch": 969, "lr": 7.725326035126892e-05}
{"train_loss": 0.1939646303653717, "global_step": 86257, "epoch": 969, "lr": 7.72527743019474e-05}
{"train_loss": 0.19930927455425262, "global_step": 86258, "epoch": 969, "lr": 7.725228824896208e-05}
{"train_loss": 0.24378937482833862, "global_step": 86259, "epoch": 969, "lr": 7.725180219231302e-05}
{"train_loss": 0.11702457815408707, "global_step": 86260, "epoch": 969, "lr": 7.725131613200027e-05}
{"train_loss": 0.09510155022144318, "global_step": 86261, "epoch": 969, "lr": 7.725083006802394e-05}
{"train_loss": 0.18323896825313568, "global_step": 86262, "epoch": 969, "lr": 7.725034400038404e-05}
{"train_loss": 0.15111909806728363, "global_step": 86263, "epoch": 969, "lr": 7.724985792908069e-05}
{"train_loss": 0.2329902946949005, "global_step": 86264, "epoch": 969, "lr": 7.724937185411391e-05}
{"train_loss": 0.2446308434009552, "global_step": 86265, "epoch": 969, "lr": 7.72488857754838e-05}
{"train_loss": 0.143705353140831, "global_step": 86266, "epoch": 969, "lr": 7.72483996931904e-05}
{"train_loss": 0.13467146456241608, "global_step": 86267, "epoch": 969, "lr": 7.724791360723378e-05}
{"train_loss": 0.16518160700798035, "global_step": 86268, "epoch": 969, "lr": 7.724742751761402e-05}
{"train_loss": 0.12354887276887894, "global_step": 86269, "epoch": 969, "lr": 7.724694142433115e-05}
{"train_loss": 0.15938593447208405, "global_step": 86270, "epoch": 969, "lr": 7.72464553273853e-05}
{"train_loss": 0.19791972637176514, "global_step": 86271, "epoch": 969, "lr": 7.724596922677647e-05}
{"train_loss": 0.14371997117996216, "global_step": 86272, "epoch": 969, "lr": 7.724548312250475e-05}
{"train_loss": 0.1562454253435135, "global_step": 86273, "epoch": 969, "lr": 7.724499701457022e-05}
{"train_loss": 0.19447900354862213, "global_step": 86274, "epoch": 969, "lr": 7.724451090297293e-05}
{"train_loss": 0.2148222029209137, "global_step": 86275, "epoch": 969, "lr": 7.724402478771295e-05}
{"train_loss": 0.18866488337516785, "global_step": 86276, "epoch": 969, "lr": 7.724353866879032e-05}
{"train_loss": 0.2138718217611313, "global_step": 86277, "epoch": 969, "lr": 7.724305254620516e-05}
{"train_loss": 0.19564929604530334, "global_step": 86278, "epoch": 969, "lr": 7.724256641995749e-05}
{"train_loss": 0.2129746824502945, "global_step": 86279, "epoch": 969, "lr": 7.724208029004739e-05}
{"train_loss": 0.18843595683574677, "global_step": 86280, "epoch": 969, "lr": 7.724159415647492e-05}
{"train_loss": 0.16059890389442444, "global_step": 86281, "epoch": 969, "lr": 7.724110801924016e-05}
{"train_loss": 0.2131311595439911, "global_step": 86282, "epoch": 969, "lr": 7.724062187834316e-05}
{"train_loss": 0.1458854228258133, "global_step": 86283, "epoch": 969, "lr": 7.724013573378398e-05}
{"train_loss": 0.19049443304538727, "global_step": 86284, "epoch": 969, "lr": 7.723964958556272e-05}
{"train_loss": 0.2398528754711151, "global_step": 86285, "epoch": 969, "lr": 7.723916343367939e-05}
{"train_loss": 0.16796649992465973, "global_step": 86286, "epoch": 969, "lr": 7.72386772781341e-05}
{"train_loss": 0.167093887925148, "global_step": 86287, "epoch": 969, "lr": 7.72381911189269e-05}
{"train_loss": 0.11866621673107147, "global_step": 86288, "epoch": 969, "lr": 7.723770495605786e-05}
{"train_loss": 0.1751750260591507, "global_step": 86289, "epoch": 969, "lr": 7.723721878952704e-05}
{"train_loss": 0.1613842099905014, "global_step": 86290, "epoch": 969, "lr": 7.723673261933452e-05}
{"train_loss": 0.19198313355445862, "global_step": 86291, "epoch": 969, "lr": 7.723624644548035e-05}
{"train_loss": 0.2355462610721588, "global_step": 86292, "epoch": 969, "lr": 7.723576026796458e-05}
{"train_loss": 0.11540113389492035, "global_step": 86293, "epoch": 969, "lr": 7.723527408678731e-05}
{"train_loss": 0.24728429317474365, "global_step": 86294, "epoch": 969, "lr": 7.723478790194859e-05}
{"train_loss": 0.12510381639003754, "global_step": 86295, "epoch": 969, "lr": 7.723430171344848e-05}
{"train_loss": 0.13617900013923645, "global_step": 86296, "epoch": 969, "lr": 7.723381552128705e-05}
{"train_loss": 0.18334122002124786, "global_step": 86297, "epoch": 969, "lr": 7.723332932546437e-05}
{"train_loss": 0.1976848542690277, "global_step": 86298, "epoch": 969, "lr": 7.723284312598049e-05}
{"train_loss": 0.31543031334877014, "global_step": 86299, "epoch": 969, "lr": 7.723235692283551e-05}
{"train_loss": 0.1969396471977234, "global_step": 86300, "epoch": 969, "lr": 7.723187071602944e-05}
{"train_loss": 0.23431846499443054, "global_step": 86301, "epoch": 969, "lr": 7.723138450556239e-05}
{"train_loss": 0.18593379855155945, "global_step": 86302, "epoch": 969, "lr": 7.723089829143442e-05}
{"train_loss": 0.1729004830121994, "global_step": 86303, "epoch": 969, "lr": 7.723041207364558e-05}
{"train_loss": 0.14115238189697266, "global_step": 86304, "epoch": 969, "lr": 7.722992585219595e-05}
{"train_loss": 0.1369018405675888, "global_step": 86305, "epoch": 969, "lr": 7.722943962708559e-05}
{"train_loss": 0.11451618373394012, "global_step": 86306, "epoch": 969, "lr": 7.722895339831455e-05}
{"train_loss": 0.16667011380195618, "global_step": 86307, "epoch": 969, "lr": 7.722846716588292e-05}
{"train_loss": 0.17290954291820526, "global_step": 86308, "epoch": 969, "lr": 7.722798092979075e-05}
{"train_loss": 0.1662253886461258, "global_step": 86309, "epoch": 969, "lr": 7.72274946900381e-05}
{"train_loss": 0.20816634595394135, "global_step": 86310, "epoch": 969, "lr": 7.722700844662506e-05}
{"train_loss": 0.205508291721344, "global_step": 86311, "epoch": 969, "lr": 7.722652219955168e-05}
{"train_loss": 0.18357041478157043, "global_step": 86312, "epoch": 969, "lr": 7.722603594881801e-05}
{"train_loss": 0.2331593781709671, "global_step": 86313, "epoch": 969, "lr": 7.722554969442415e-05}
{"train_loss": 0.15928739309310913, "global_step": 86314, "epoch": 969, "lr": 7.722506343637015e-05}
{"train_loss": 0.1315312534570694, "global_step": 86315, "epoch": 969, "lr": 7.722457717465606e-05}
{"train_loss": 0.23144683241844177, "global_step": 86316, "epoch": 969, "lr": 7.722409090928195e-05}
{"train_loss": 0.15071538090705872, "global_step": 86317, "epoch": 969, "lr": 7.722360464024792e-05}
{"train_loss": 0.22178474068641663, "global_step": 86318, "epoch": 969, "lr": 7.722311836755399e-05}
{"train_loss": 0.2911122441291809, "global_step": 86319, "epoch": 969, "lr": 7.722263209120024e-05}
{"train_loss": 0.19617396593093872, "global_step": 86320, "epoch": 969, "lr": 7.722214581118675e-05}
{"train_loss": 0.13538163900375366, "global_step": 86321, "epoch": 969, "lr": 7.722165952751355e-05}
{"train_loss": 0.19474509358406067, "global_step": 86322, "epoch": 969, "lr": 7.722117324018074e-05}
{"train_loss": 0.19772326946258545, "global_step": 86323, "epoch": 969, "lr": 7.722068694918839e-05}
{"train_loss": 0.11810246855020523, "global_step": 86324, "epoch": 969, "lr": 7.722020065453655e-05}
{"train_loss": 0.2080111801624298, "global_step": 86325, "epoch": 969, "lr": 7.721971435622528e-05}
{"train_loss": 0.15805453062057495, "global_step": 86326, "epoch": 969, "lr": 7.721922805425463e-05}
{"train_loss": 0.1718733012676239, "global_step": 86327, "epoch": 969, "lr": 7.721874174862471e-05}
{"train_loss": 0.25259414315223694, "global_step": 86328, "epoch": 969, "lr": 7.721825543933557e-05}
{"train_loss": 0.18164899727601683, "global_step": 86329, "epoch": 969, "lr": 7.721776912638726e-05, "val_loss": 4.372735023498535}
{"train_loss": 0.19709517061710358, "global_step": 86330, "epoch": 970, "lr": 7.721728280977983e-05}
{"train_loss": 0.23146796226501465, "global_step": 86331, "epoch": 970, "lr": 7.721679648951338e-05}
{"train_loss": 0.14024442434310913, "global_step": 86332, "epoch": 970, "lr": 7.721631016558798e-05}
{"train_loss": 0.16242066025733948, "global_step": 86333, "epoch": 970, "lr": 7.721582383800366e-05}
{"train_loss": 0.17343561351299286, "global_step": 86334, "epoch": 970, "lr": 7.721533750676052e-05}
{"train_loss": 0.20848526060581207, "global_step": 86335, "epoch": 970, "lr": 7.72148511718586e-05}
{"train_loss": 0.24555480480194092, "global_step": 86336, "epoch": 970, "lr": 7.721436483329798e-05}
{"train_loss": 0.12627267837524414, "global_step": 86337, "epoch": 970, "lr": 7.721387849107872e-05}
{"train_loss": 0.23798705637454987, "global_step": 86338, "epoch": 970, "lr": 7.721339214520088e-05}
{"train_loss": 0.2227545976638794, "global_step": 86339, "epoch": 970, "lr": 7.721290579566454e-05}
{"train_loss": 0.152458056807518, "global_step": 86340, "epoch": 970, "lr": 7.721241944246975e-05}
{"train_loss": 0.1372116208076477, "global_step": 86341, "epoch": 970, "lr": 7.721193308561659e-05}
{"train_loss": 0.20150533318519592, "global_step": 86342, "epoch": 970, "lr": 7.72114467251051e-05}
{"train_loss": 0.24973855912685394, "global_step": 86343, "epoch": 970, "lr": 7.721096036093538e-05}
{"train_loss": 0.2113741636276245, "global_step": 86344, "epoch": 970, "lr": 7.721047399310746e-05}
{"train_loss": 0.2241724729537964, "global_step": 86345, "epoch": 970, "lr": 7.720998762162143e-05}
{"train_loss": 0.21346059441566467, "global_step": 86346, "epoch": 970, "lr": 7.720950124647736e-05}
{"train_loss": 0.16470977663993835, "global_step": 86347, "epoch": 970, "lr": 7.720901486767529e-05}
{"train_loss": 0.11039657145738602, "global_step": 86348, "epoch": 970, "lr": 7.720852848521532e-05}
{"train_loss": 0.2011934220790863, "global_step": 86349, "epoch": 970, "lr": 7.720804209909747e-05}
{"train_loss": 0.12387353181838989, "global_step": 86350, "epoch": 970, "lr": 7.720755570932183e-05}
{"train_loss": 0.13694044947624207, "global_step": 86351, "epoch": 970, "lr": 7.720706931588849e-05}
{"train_loss": 0.14539054036140442, "global_step": 86352, "epoch": 970, "lr": 7.720658291879748e-05}
{"train_loss": 0.17406690120697021, "global_step": 86353, "epoch": 970, "lr": 7.720609651804886e-05}
{"train_loss": 0.2344125509262085, "global_step": 86354, "epoch": 970, "lr": 7.720561011364273e-05}
{"train_loss": 0.1434582620859146, "global_step": 86355, "epoch": 970, "lr": 7.720512370557915e-05}
{"train_loss": 0.15431855618953705, "global_step": 86356, "epoch": 970, "lr": 7.720463729385815e-05}
{"train_loss": 0.19297772645950317, "global_step": 86357, "epoch": 970, "lr": 7.720415087847984e-05}
{"train_loss": 0.17625990509986877, "global_step": 86358, "epoch": 970, "lr": 7.720366445944424e-05}
{"train_loss": 0.19251090288162231, "global_step": 86359, "epoch": 970, "lr": 7.720317803675145e-05}
{"train_loss": 0.16610126197338104, "global_step": 86360, "epoch": 970, "lr": 7.720269161040153e-05}
{"train_loss": 0.15483424067497253, "global_step": 86361, "epoch": 970, "lr": 7.720220518039452e-05}
{"train_loss": 0.1567392498254776, "global_step": 86362, "epoch": 970, "lr": 7.720171874673052e-05}
{"train_loss": 0.16911087930202484, "global_step": 86363, "epoch": 970, "lr": 7.720123230940956e-05}
{"train_loss": 0.10650575160980225, "global_step": 86364, "epoch": 970, "lr": 7.720074586843176e-05}
{"train_loss": 0.13917537033557892, "global_step": 86365, "epoch": 970, "lr": 7.720025942379712e-05}
{"train_loss": 0.1783074289560318, "global_step": 86366, "epoch": 970, "lr": 7.719977297550576e-05}
{"train_loss": 0.20917369425296783, "global_step": 86367, "epoch": 970, "lr": 7.71992865235577e-05}
{"train_loss": 0.2164556235074997, "global_step": 86368, "epoch": 970, "lr": 7.719880006795305e-05}
{"train_loss": 0.19163575768470764, "global_step": 86369, "epoch": 970, "lr": 7.719831360869185e-05}
{"train_loss": 0.23456160724163055, "global_step": 86370, "epoch": 970, "lr": 7.719782714577416e-05}
{"train_loss": 0.1665044128894806, "global_step": 86371, "epoch": 970, "lr": 7.719734067920005e-05}
{"train_loss": 0.2090950608253479, "global_step": 86372, "epoch": 970, "lr": 7.719685420896959e-05}
{"train_loss": 0.21683494746685028, "global_step": 86373, "epoch": 970, "lr": 7.719636773508285e-05}
{"train_loss": 0.16746647655963898, "global_step": 86374, "epoch": 970, "lr": 7.719588125753989e-05}
{"train_loss": 0.2534452974796295, "global_step": 86375, "epoch": 970, "lr": 7.719539477634076e-05}
{"train_loss": 0.16120563447475433, "global_step": 86376, "epoch": 970, "lr": 7.719490829148557e-05}
{"train_loss": 0.1469166874885559, "global_step": 86377, "epoch": 970, "lr": 7.719442180297434e-05}
{"train_loss": 0.15761631727218628, "global_step": 86378, "epoch": 970, "lr": 7.719393531080715e-05}
{"train_loss": 0.1907491683959961, "global_step": 86379, "epoch": 970, "lr": 7.719344881498407e-05}
{"train_loss": 0.16799359023571014, "global_step": 86380, "epoch": 970, "lr": 7.719296231550514e-05}
{"train_loss": 0.16167588531970978, "global_step": 86381, "epoch": 970, "lr": 7.719247581237048e-05}
{"train_loss": 0.20024779438972473, "global_step": 86382, "epoch": 970, "lr": 7.719198930558011e-05}
{"train_loss": 0.14692634344100952, "global_step": 86383, "epoch": 970, "lr": 7.719150279513411e-05}
{"train_loss": 0.2081245630979538, "global_step": 86384, "epoch": 970, "lr": 7.719101628103255e-05}
{"train_loss": 0.2178049087524414, "global_step": 86385, "epoch": 970, "lr": 7.719052976327548e-05}
{"train_loss": 0.20004147291183472, "global_step": 86386, "epoch": 970, "lr": 7.719004324186297e-05}
{"train_loss": 0.15676262974739075, "global_step": 86387, "epoch": 970, "lr": 7.718955671679511e-05}
{"train_loss": 0.20315593481063843, "global_step": 86388, "epoch": 970, "lr": 7.718907018807193e-05}
{"train_loss": 0.12077511847019196, "global_step": 86389, "epoch": 970, "lr": 7.718858365569351e-05}
{"train_loss": 0.1645805835723877, "global_step": 86390, "epoch": 970, "lr": 7.718809711965994e-05}
{"train_loss": 0.27653390169143677, "global_step": 86391, "epoch": 970, "lr": 7.718761057997123e-05}
{"train_loss": 0.15596644580364227, "global_step": 86392, "epoch": 970, "lr": 7.718712403662749e-05}
{"train_loss": 0.20878663659095764, "global_step": 86393, "epoch": 970, "lr": 7.718663748962878e-05}
{"train_loss": 0.18863403797149658, "global_step": 86394, "epoch": 970, "lr": 7.718615093897515e-05}
{"train_loss": 0.14756453037261963, "global_step": 86395, "epoch": 970, "lr": 7.718566438466665e-05}
{"train_loss": 0.21071498095989227, "global_step": 86396, "epoch": 970, "lr": 7.718517782670341e-05}
{"train_loss": 0.19538374245166779, "global_step": 86397, "epoch": 970, "lr": 7.718469126508543e-05}
{"train_loss": 0.13803493976593018, "global_step": 86398, "epoch": 970, "lr": 7.71842046998128e-05}
{"train_loss": 0.1589909940958023, "global_step": 86399, "epoch": 970, "lr": 7.718371813088559e-05}
{"train_loss": 0.14143876731395721, "global_step": 86400, "epoch": 970, "lr": 7.718323155830387e-05}
{"train_loss": 0.144626647233963, "global_step": 86401, "epoch": 970, "lr": 7.718274498206767e-05}
{"train_loss": 0.19130168855190277, "global_step": 86402, "epoch": 970, "lr": 7.71822584021771e-05}
{"train_loss": 0.21453380584716797, "global_step": 86403, "epoch": 970, "lr": 7.718177181863222e-05}
{"train_loss": 0.299709290266037, "global_step": 86404, "epoch": 970, "lr": 7.718128523143307e-05}
{"train_loss": 0.16792859137058258, "global_step": 86405, "epoch": 970, "lr": 7.718079864057972e-05}
{"train_loss": 0.23355962336063385, "global_step": 86406, "epoch": 970, "lr": 7.718031204607224e-05}
{"train_loss": 0.22856204211711884, "global_step": 86407, "epoch": 970, "lr": 7.717982544791071e-05}
{"train_loss": 0.2541104853153229, "global_step": 86408, "epoch": 970, "lr": 7.717933884609519e-05}
{"train_loss": 0.19159914553165436, "global_step": 86409, "epoch": 970, "lr": 7.717885224062571e-05}
{"train_loss": 0.1388300359249115, "global_step": 86410, "epoch": 970, "lr": 7.717836563150238e-05}
{"train_loss": 0.15428230166435242, "global_step": 86411, "epoch": 970, "lr": 7.717787901872526e-05}
{"train_loss": 0.1822897046804428, "global_step": 86412, "epoch": 970, "lr": 7.71773924022944e-05}
{"train_loss": 0.2152252495288849, "global_step": 86413, "epoch": 970, "lr": 7.717690578220988e-05}
{"train_loss": 0.29056376218795776, "global_step": 86414, "epoch": 970, "lr": 7.717641915847173e-05}
{"train_loss": 0.2035018801689148, "global_step": 86415, "epoch": 970, "lr": 7.717593253108006e-05}
{"train_loss": 0.16490291059017181, "global_step": 86416, "epoch": 970, "lr": 7.717544590003492e-05}
{"train_loss": 0.1653323918581009, "global_step": 86417, "epoch": 970, "lr": 7.717495926533636e-05}
{"train_loss": 0.18464472144842148, "global_step": 86418, "epoch": 970, "lr": 7.717447262698447e-05, "val_loss": 4.241110801696777, "train_action_mse_error": 27.403640747070312}
{"train_loss": 0.22296199202537537, "global_step": 86419, "epoch": 971, "lr": 7.717398598497929e-05}
{"train_loss": 0.1903090924024582, "global_step": 86420, "epoch": 971, "lr": 7.717349933932092e-05}
{"train_loss": 0.21318618953227997, "global_step": 86421, "epoch": 971, "lr": 7.717301269000938e-05}
{"train_loss": 0.20463232696056366, "global_step": 86422, "epoch": 971, "lr": 7.717252603704477e-05}
{"train_loss": 0.23936067521572113, "global_step": 86423, "epoch": 971, "lr": 7.717203938042715e-05}
{"train_loss": 0.21019810438156128, "global_step": 86424, "epoch": 971, "lr": 7.717155272015657e-05}
{"train_loss": 0.19550254940986633, "global_step": 86425, "epoch": 971, "lr": 7.717106605623313e-05}
{"train_loss": 0.11624033749103546, "global_step": 86426, "epoch": 971, "lr": 7.717057938865684e-05}
{"train_loss": 0.13021151721477509, "global_step": 86427, "epoch": 971, "lr": 7.71700927174278e-05}
{"train_loss": 0.13195353746414185, "global_step": 86428, "epoch": 971, "lr": 7.71696060425461e-05}
{"train_loss": 0.20603786408901215, "global_step": 86429, "epoch": 971, "lr": 7.716911936401175e-05}
{"train_loss": 0.1538255512714386, "global_step": 86430, "epoch": 971, "lr": 7.716863268182486e-05}
{"train_loss": 0.1584128886461258, "global_step": 86431, "epoch": 971, "lr": 7.716814599598547e-05}
{"train_loss": 0.2469273805618286, "global_step": 86432, "epoch": 971, "lr": 7.716765930649365e-05}
{"train_loss": 0.20838043093681335, "global_step": 86433, "epoch": 971, "lr": 7.716717261334949e-05}
{"train_loss": 0.18124352395534515, "global_step": 86434, "epoch": 971, "lr": 7.716668591655302e-05}
{"train_loss": 0.17391762137413025, "global_step": 86435, "epoch": 971, "lr": 7.71661992161043e-05}
{"train_loss": 0.14208176732063293, "global_step": 86436, "epoch": 971, "lr": 7.716571251200344e-05}
{"train_loss": 0.22432048618793488, "global_step": 86437, "epoch": 971, "lr": 7.716522580425047e-05}
{"train_loss": 0.1613360047340393, "global_step": 86438, "epoch": 971, "lr": 7.716473909284549e-05}
{"train_loss": 0.20778566598892212, "global_step": 86439, "epoch": 971, "lr": 7.716425237778853e-05}
{"train_loss": 0.15871265530586243, "global_step": 86440, "epoch": 971, "lr": 7.716376565907966e-05}
{"train_loss": 0.13765622675418854, "global_step": 86441, "epoch": 971, "lr": 7.716327893671896e-05}
{"train_loss": 0.1867065578699112, "global_step": 86442, "epoch": 971, "lr": 7.716279221070648e-05}
{"train_loss": 0.18684005737304688, "global_step": 86443, "epoch": 971, "lr": 7.716230548104228e-05}
{"train_loss": 0.1769576221704483, "global_step": 86444, "epoch": 971, "lr": 7.716181874772647e-05}
{"train_loss": 0.21846874058246613, "global_step": 86445, "epoch": 971, "lr": 7.716133201075907e-05}
{"train_loss": 0.16579864919185638, "global_step": 86446, "epoch": 971, "lr": 7.716084527014016e-05}
{"train_loss": 0.1732795238494873, "global_step": 86447, "epoch": 971, "lr": 7.71603585258698e-05}
{"train_loss": 0.20270361006259918, "global_step": 86448, "epoch": 971, "lr": 7.715987177794807e-05}
{"train_loss": 0.1434425711631775, "global_step": 86449, "epoch": 971, "lr": 7.715938502637502e-05}
{"train_loss": 0.1867125928401947, "global_step": 86450, "epoch": 971, "lr": 7.715889827115072e-05}
{"train_loss": 0.1541367918252945, "global_step": 86451, "epoch": 971, "lr": 7.715841151227525e-05}
{"train_loss": 0.17184022068977356, "global_step": 86452, "epoch": 971, "lr": 7.715792474974865e-05}
{"train_loss": 0.11775172501802444, "global_step": 86453, "epoch": 971, "lr": 7.715743798357101e-05}
{"train_loss": 0.21093007922172546, "global_step": 86454, "epoch": 971, "lr": 7.715695121374237e-05}
{"train_loss": 0.16947247087955475, "global_step": 86455, "epoch": 971, "lr": 7.715646444026281e-05}
{"train_loss": 0.15872161090373993, "global_step": 86456, "epoch": 971, "lr": 7.71559776631324e-05}
{"train_loss": 0.20417985320091248, "global_step": 86457, "epoch": 971, "lr": 7.71554908823512e-05}
{"train_loss": 0.16524089872837067, "global_step": 86458, "epoch": 971, "lr": 7.715500409791929e-05}
{"train_loss": 0.13358885049819946, "global_step": 86459, "epoch": 971, "lr": 7.715451730983669e-05}
{"train_loss": 0.17449302971363068, "global_step": 86460, "epoch": 971, "lr": 7.715403051810352e-05}
{"train_loss": 0.199861079454422, "global_step": 86461, "epoch": 971, "lr": 7.715354372271981e-05}
{"train_loss": 0.1766786128282547, "global_step": 86462, "epoch": 971, "lr": 7.715305692368564e-05}
{"train_loss": 0.18118630349636078, "global_step": 86463, "epoch": 971, "lr": 7.715257012100108e-05}
{"train_loss": 0.11949827522039413, "global_step": 86464, "epoch": 971, "lr": 7.715208331466617e-05}
{"train_loss": 0.1614755392074585, "global_step": 86465, "epoch": 971, "lr": 7.7151596504681e-05}
{"train_loss": 0.16518184542655945, "global_step": 86466, "epoch": 971, "lr": 7.715110969104564e-05}
{"train_loss": 0.12318387627601624, "global_step": 86467, "epoch": 971, "lr": 7.715062287376014e-05}
{"train_loss": 0.12644432485103607, "global_step": 86468, "epoch": 971, "lr": 7.715013605282456e-05}
{"train_loss": 0.1380363553762436, "global_step": 86469, "epoch": 971, "lr": 7.714964922823898e-05}
{"train_loss": 0.16304942965507507, "global_step": 86470, "epoch": 971, "lr": 7.714916240000347e-05}
{"train_loss": 0.1753188669681549, "global_step": 86471, "epoch": 971, "lr": 7.714867556811806e-05}
{"train_loss": 0.2098938524723053, "global_step": 86472, "epoch": 971, "lr": 7.714818873258287e-05}
{"train_loss": 0.17634978890419006, "global_step": 86473, "epoch": 971, "lr": 7.714770189339793e-05}
{"train_loss": 0.20058855414390564, "global_step": 86474, "epoch": 971, "lr": 7.71472150505633e-05}
{"train_loss": 0.21519674360752106, "global_step": 86475, "epoch": 971, "lr": 7.714672820407906e-05}
{"train_loss": 0.18469864130020142, "global_step": 86476, "epoch": 971, "lr": 7.714624135394528e-05}
{"train_loss": 0.18073418736457825, "global_step": 86477, "epoch": 971, "lr": 7.714575450016202e-05}
{"train_loss": 0.12252891808748245, "global_step": 86478, "epoch": 971, "lr": 7.714526764272936e-05}
{"train_loss": 0.20253777503967285, "global_step": 86479, "epoch": 971, "lr": 7.714478078164732e-05}
{"train_loss": 0.23475304245948792, "global_step": 86480, "epoch": 971, "lr": 7.714429391691601e-05}
{"train_loss": 0.12058431655168533, "global_step": 86481, "epoch": 971, "lr": 7.714380704853548e-05}
{"train_loss": 0.17772139608860016, "global_step": 86482, "epoch": 971, "lr": 7.714332017650578e-05}
{"train_loss": 0.16266785562038422, "global_step": 86483, "epoch": 971, "lr": 7.714283330082701e-05}
{"train_loss": 0.22241716086864471, "global_step": 86484, "epoch": 971, "lr": 7.714234642149922e-05}
{"train_loss": 0.17829939723014832, "global_step": 86485, "epoch": 971, "lr": 7.714185953852247e-05}
{"train_loss": 0.12683220207691193, "global_step": 86486, "epoch": 971, "lr": 7.714137265189682e-05}
{"train_loss": 0.22517487406730652, "global_step": 86487, "epoch": 971, "lr": 7.714088576162234e-05}
{"train_loss": 0.11475516855716705, "global_step": 86488, "epoch": 971, "lr": 7.71403988676991e-05}
{"train_loss": 0.18614380061626434, "global_step": 86489, "epoch": 971, "lr": 7.713991197012716e-05}
{"train_loss": 0.22428937256336212, "global_step": 86490, "epoch": 971, "lr": 7.713942506890661e-05}
{"train_loss": 0.18748606741428375, "global_step": 86491, "epoch": 971, "lr": 7.713893816403748e-05}
{"train_loss": 0.1510923057794571, "global_step": 86492, "epoch": 971, "lr": 7.713845125551985e-05}
{"train_loss": 0.16808901727199554, "global_step": 86493, "epoch": 971, "lr": 7.713796434335379e-05}
{"train_loss": 0.14961998164653778, "global_step": 86494, "epoch": 971, "lr": 7.713747742753936e-05}
{"train_loss": 0.16121749579906464, "global_step": 86495, "epoch": 971, "lr": 7.713699050807663e-05}
{"train_loss": 0.1899958997964859, "global_step": 86496, "epoch": 971, "lr": 7.713650358496565e-05}
{"train_loss": 0.12658029794692993, "global_step": 86497, "epoch": 971, "lr": 7.713601665820653e-05}
{"train_loss": 0.17576774954795837, "global_step": 86498, "epoch": 971, "lr": 7.713552972779928e-05}
{"train_loss": 0.1478813737630844, "global_step": 86499, "epoch": 971, "lr": 7.713504279374398e-05}
{"train_loss": 0.16777896881103516, "global_step": 86500, "epoch": 971, "lr": 7.713455585604072e-05}
{"train_loss": 0.20061801373958588, "global_step": 86501, "epoch": 971, "lr": 7.713406891468954e-05}
{"train_loss": 0.15029819309711456, "global_step": 86502, "epoch": 971, "lr": 7.713358196969052e-05}
{"train_loss": 0.16339199244976044, "global_step": 86503, "epoch": 971, "lr": 7.713309502104372e-05}
{"train_loss": 0.20909500122070312, "global_step": 86504, "epoch": 971, "lr": 7.71326080687492e-05}
{"train_loss": 0.26440519094467163, "global_step": 86505, "epoch": 971, "lr": 7.713212111280705e-05}
{"train_loss": 0.11992140859365463, "global_step": 86506, "epoch": 971, "lr": 7.713163415321729e-05}
{"train_loss": 0.17609751885861494, "global_step": 86507, "epoch": 971, "lr": 7.713114718998003e-05, "val_loss": 4.310980319976807}
{"train_loss": 0.1831962615251541, "global_step": 86508, "epoch": 972, "lr": 7.713066022309531e-05}
{"train_loss": 0.16607744991779327, "global_step": 86509, "epoch": 972, "lr": 7.713017325256322e-05}
{"train_loss": 0.15142101049423218, "global_step": 86510, "epoch": 972, "lr": 7.71296862783838e-05}
{"train_loss": 0.12250331789255142, "global_step": 86511, "epoch": 972, "lr": 7.712919930055711e-05}
{"train_loss": 0.1626729518175125, "global_step": 86512, "epoch": 972, "lr": 7.712871231908325e-05}
{"train_loss": 0.11686982959508896, "global_step": 86513, "epoch": 972, "lr": 7.712822533396225e-05}
{"train_loss": 0.23928019404411316, "global_step": 86514, "epoch": 972, "lr": 7.71277383451942e-05}
{"train_loss": 0.21024000644683838, "global_step": 86515, "epoch": 972, "lr": 7.712725135277915e-05}
{"train_loss": 0.1380230039358139, "global_step": 86516, "epoch": 972, "lr": 7.712676435671717e-05}
{"train_loss": 0.12913882732391357, "global_step": 86517, "epoch": 972, "lr": 7.712627735700835e-05}
{"train_loss": 0.19519838690757751, "global_step": 86518, "epoch": 972, "lr": 7.71257903536527e-05}
{"train_loss": 0.2117079794406891, "global_step": 86519, "epoch": 972, "lr": 7.712530334665033e-05}
{"train_loss": 0.14471931755542755, "global_step": 86520, "epoch": 972, "lr": 7.712481633600131e-05}
{"train_loss": 0.16132459044456482, "global_step": 86521, "epoch": 972, "lr": 7.712432932170568e-05}
{"train_loss": 0.23991088569164276, "global_step": 86522, "epoch": 972, "lr": 7.712384230376351e-05}
{"train_loss": 0.15811048448085785, "global_step": 86523, "epoch": 972, "lr": 7.712335528217487e-05}
{"train_loss": 0.09732108563184738, "global_step": 86524, "epoch": 972, "lr": 7.712286825693981e-05}
{"train_loss": 0.1922595500946045, "global_step": 86525, "epoch": 972, "lr": 7.712238122805844e-05}
{"train_loss": 0.2580172121524811, "global_step": 86526, "epoch": 972, "lr": 7.712189419553078e-05}
{"train_loss": 0.10401281714439392, "global_step": 86527, "epoch": 972, "lr": 7.712140715935691e-05}
{"train_loss": 0.1221245601773262, "global_step": 86528, "epoch": 972, "lr": 7.712092011953691e-05}
{"train_loss": 0.18256400525569916, "global_step": 86529, "epoch": 972, "lr": 7.712043307607082e-05}
{"train_loss": 0.19098469614982605, "global_step": 86530, "epoch": 972, "lr": 7.711994602895872e-05}
{"train_loss": 0.18050143122673035, "global_step": 86531, "epoch": 972, "lr": 7.711945897820068e-05}
{"train_loss": 0.15786239504814148, "global_step": 86532, "epoch": 972, "lr": 7.711897192379675e-05}
{"train_loss": 0.18312855064868927, "global_step": 86533, "epoch": 972, "lr": 7.711848486574702e-05}
{"train_loss": 0.16134829819202423, "global_step": 86534, "epoch": 972, "lr": 7.711799780405153e-05}
{"train_loss": 0.16558122634887695, "global_step": 86535, "epoch": 972, "lr": 7.711751073871036e-05}
{"train_loss": 0.15918445587158203, "global_step": 86536, "epoch": 972, "lr": 7.711702366972357e-05}
{"train_loss": 0.16265127062797546, "global_step": 86537, "epoch": 972, "lr": 7.711653659709122e-05}
{"train_loss": 0.14609162509441376, "global_step": 86538, "epoch": 972, "lr": 7.711604952081339e-05}
{"train_loss": 0.13875004649162292, "global_step": 86539, "epoch": 972, "lr": 7.711556244089014e-05}
{"train_loss": 0.1346348077058792, "global_step": 86540, "epoch": 972, "lr": 7.711507535732152e-05}
{"train_loss": 0.1089455783367157, "global_step": 86541, "epoch": 972, "lr": 7.711458827010763e-05}
{"train_loss": 0.12933394312858582, "global_step": 86542, "epoch": 972, "lr": 7.711410117924849e-05}
{"train_loss": 0.2010045051574707, "global_step": 86543, "epoch": 972, "lr": 7.71136140847442e-05}
{"train_loss": 0.10617100447416306, "global_step": 86544, "epoch": 972, "lr": 7.71131269865948e-05}
{"train_loss": 0.08806668221950531, "global_step": 86545, "epoch": 972, "lr": 7.71126398848004e-05}
{"train_loss": 0.23944519460201263, "global_step": 86546, "epoch": 972, "lr": 7.711215277936102e-05}
{"train_loss": 0.17419786751270294, "global_step": 86547, "epoch": 972, "lr": 7.711166567027674e-05}
{"train_loss": 0.21204306185245514, "global_step": 86548, "epoch": 972, "lr": 7.711117855754764e-05}
{"train_loss": 0.1910175085067749, "global_step": 86549, "epoch": 972, "lr": 7.711069144117375e-05}
{"train_loss": 0.18017584085464478, "global_step": 86550, "epoch": 972, "lr": 7.71102043211552e-05}
{"train_loss": 0.22372567653656006, "global_step": 86551, "epoch": 972, "lr": 7.710971719749196e-05}
{"train_loss": 0.34859246015548706, "global_step": 86552, "epoch": 972, "lr": 7.710923007018417e-05}
{"train_loss": 0.11324794590473175, "global_step": 86553, "epoch": 972, "lr": 7.71087429392319e-05}
{"train_loss": 0.12419012933969498, "global_step": 86554, "epoch": 972, "lr": 7.710825580463516e-05}
{"train_loss": 0.13575731217861176, "global_step": 86555, "epoch": 972, "lr": 7.710776866639405e-05}
{"train_loss": 0.2485462725162506, "global_step": 86556, "epoch": 972, "lr": 7.710728152450864e-05}
{"train_loss": 0.22778810560703278, "global_step": 86557, "epoch": 972, "lr": 7.710679437897897e-05}
{"train_loss": 0.19593873620033264, "global_step": 86558, "epoch": 972, "lr": 7.710630722980514e-05}
{"train_loss": 0.2547641694545746, "global_step": 86559, "epoch": 972, "lr": 7.71058200769872e-05}
{"train_loss": 0.22677721083164215, "global_step": 86560, "epoch": 972, "lr": 7.71053329205252e-05}
{"train_loss": 0.13594204187393188, "global_step": 86561, "epoch": 972, "lr": 7.71048457604192e-05}
{"train_loss": 0.18939875066280365, "global_step": 86562, "epoch": 972, "lr": 7.710435859666931e-05}
{"train_loss": 0.2386104166507721, "global_step": 86563, "epoch": 972, "lr": 7.710387142927557e-05}
{"train_loss": 0.16220204532146454, "global_step": 86564, "epoch": 972, "lr": 7.710338425823805e-05}
{"train_loss": 0.24459706246852875, "global_step": 86565, "epoch": 972, "lr": 7.710289708355678e-05}
{"train_loss": 0.16246475279331207, "global_step": 86566, "epoch": 972, "lr": 7.710240990523188e-05}
{"train_loss": 0.2692272663116455, "global_step": 86567, "epoch": 972, "lr": 7.710192272326339e-05}
{"train_loss": 0.162178173661232, "global_step": 86568, "epoch": 972, "lr": 7.710143553765137e-05}
{"train_loss": 0.20900987088680267, "global_step": 86569, "epoch": 972, "lr": 7.71009483483959e-05}
{"train_loss": 0.1433860808610916, "global_step": 86570, "epoch": 972, "lr": 7.710046115549703e-05}
{"train_loss": 0.18151624500751495, "global_step": 86571, "epoch": 972, "lr": 7.709997395895483e-05}
{"train_loss": 0.19207313656806946, "global_step": 86572, "epoch": 972, "lr": 7.709948675876937e-05}
{"train_loss": 0.21186529099941254, "global_step": 86573, "epoch": 972, "lr": 7.709899955494073e-05}
{"train_loss": 0.19355815649032593, "global_step": 86574, "epoch": 972, "lr": 7.709851234746895e-05}
{"train_loss": 0.1364033967256546, "global_step": 86575, "epoch": 972, "lr": 7.709802513635411e-05}
{"train_loss": 0.17184917628765106, "global_step": 86576, "epoch": 972, "lr": 7.709753792159626e-05}
{"train_loss": 0.21443147957324982, "global_step": 86577, "epoch": 972, "lr": 7.709705070319548e-05}
{"train_loss": 0.17411339282989502, "global_step": 86578, "epoch": 972, "lr": 7.709656348115184e-05}
{"train_loss": 0.17570294439792633, "global_step": 86579, "epoch": 972, "lr": 7.70960762554654e-05}
{"train_loss": 0.15036721527576447, "global_step": 86580, "epoch": 972, "lr": 7.709558902613621e-05}
{"train_loss": 0.1741310954093933, "global_step": 86581, "epoch": 972, "lr": 7.709510179316434e-05}
{"train_loss": 0.17309218645095825, "global_step": 86582, "epoch": 972, "lr": 7.709461455654989e-05}
{"train_loss": 0.21262042224407196, "global_step": 86583, "epoch": 972, "lr": 7.709412731629287e-05}
{"train_loss": 0.16031001508235931, "global_step": 86584, "epoch": 972, "lr": 7.709364007239339e-05}
{"train_loss": 0.27162081003189087, "global_step": 86585, "epoch": 972, "lr": 7.70931528248515e-05}
{"train_loss": 0.1849118173122406, "global_step": 86586, "epoch": 972, "lr": 7.709266557366728e-05}
{"train_loss": 0.24516917765140533, "global_step": 86587, "epoch": 972, "lr": 7.709217831884076e-05}
{"train_loss": 0.2543368339538574, "global_step": 86588, "epoch": 972, "lr": 7.709169106037204e-05}
{"train_loss": 0.3130260705947876, "global_step": 86589, "epoch": 972, "lr": 7.709120379826116e-05}
{"train_loss": 0.24113883078098297, "global_step": 86590, "epoch": 972, "lr": 7.709071653250821e-05}
{"train_loss": 0.15405139327049255, "global_step": 86591, "epoch": 972, "lr": 7.709022926311324e-05}
{"train_loss": 0.16341355443000793, "global_step": 86592, "epoch": 972, "lr": 7.708974199007631e-05}
{"train_loss": 0.20127174258232117, "global_step": 86593, "epoch": 972, "lr": 7.70892547133975e-05}
{"train_loss": 0.28743866086006165, "global_step": 86594, "epoch": 972, "lr": 7.708876743307686e-05}
{"train_loss": 0.21533708274364471, "global_step": 86595, "epoch": 972, "lr": 7.708828014911448e-05}
{"train_loss": 0.1844866702730736, "global_step": 86596, "epoch": 972, "lr": 7.708779286151042e-05, "val_loss": 4.19073486328125}
{"train_loss": 0.2271626740694046, "global_step": 86597, "epoch": 973, "lr": 7.708730557026472e-05}
{"train_loss": 0.1936538815498352, "global_step": 86598, "epoch": 973, "lr": 7.708681827537746e-05}
{"train_loss": 0.15879705548286438, "global_step": 86599, "epoch": 973, "lr": 7.708633097684873e-05}
{"train_loss": 0.17876307666301727, "global_step": 86600, "epoch": 973, "lr": 7.708584367467855e-05}
{"train_loss": 0.2479466199874878, "global_step": 86601, "epoch": 973, "lr": 7.708535636886702e-05}
{"train_loss": 0.20011970400810242, "global_step": 86602, "epoch": 973, "lr": 7.708486905941418e-05}
{"train_loss": 0.18844184279441833, "global_step": 86603, "epoch": 973, "lr": 7.708438174632011e-05}
{"train_loss": 0.20069164037704468, "global_step": 86604, "epoch": 973, "lr": 7.708389442958489e-05}
{"train_loss": 0.1472686231136322, "global_step": 86605, "epoch": 973, "lr": 7.708340710920856e-05}
{"train_loss": 0.19652311503887177, "global_step": 86606, "epoch": 973, "lr": 7.70829197851912e-05}
{"train_loss": 0.2129838466644287, "global_step": 86607, "epoch": 973, "lr": 7.708243245753287e-05}
{"train_loss": 0.1860685795545578, "global_step": 86608, "epoch": 973, "lr": 7.708194512623364e-05}
{"train_loss": 0.21698953211307526, "global_step": 86609, "epoch": 973, "lr": 7.708145779129356e-05}
{"train_loss": 0.17599157989025116, "global_step": 86610, "epoch": 973, "lr": 7.708097045271272e-05}
{"train_loss": 0.15575654804706573, "global_step": 86611, "epoch": 973, "lr": 7.708048311049119e-05}
{"train_loss": 0.2095963954925537, "global_step": 86612, "epoch": 973, "lr": 7.707999576462898e-05}
{"train_loss": 0.14315956830978394, "global_step": 86613, "epoch": 973, "lr": 7.707950841512623e-05}
{"train_loss": 0.14572718739509583, "global_step": 86614, "epoch": 973, "lr": 7.707902106198295e-05}
{"train_loss": 0.19566413760185242, "global_step": 86615, "epoch": 973, "lr": 7.707853370519924e-05}
{"train_loss": 0.18125632405281067, "global_step": 86616, "epoch": 973, "lr": 7.707804634477514e-05}
{"train_loss": 0.16848473250865936, "global_step": 86617, "epoch": 973, "lr": 7.707755898071071e-05}
{"train_loss": 0.18071258068084717, "global_step": 86618, "epoch": 973, "lr": 7.707707161300607e-05}
{"train_loss": 0.20612666010856628, "global_step": 86619, "epoch": 973, "lr": 7.707658424166123e-05}
{"train_loss": 0.13330812752246857, "global_step": 86620, "epoch": 973, "lr": 7.707609686667628e-05}
{"train_loss": 0.1901230663061142, "global_step": 86621, "epoch": 973, "lr": 7.707560948805126e-05}
{"train_loss": 0.2528350353240967, "global_step": 86622, "epoch": 973, "lr": 7.707512210578627e-05}
{"train_loss": 0.20661556720733643, "global_step": 86623, "epoch": 973, "lr": 7.707463471988135e-05}
{"train_loss": 0.14323851466178894, "global_step": 86624, "epoch": 973, "lr": 7.707414733033658e-05}
{"train_loss": 0.15655913949012756, "global_step": 86625, "epoch": 973, "lr": 7.707365993715203e-05}
{"train_loss": 0.201262965798378, "global_step": 86626, "epoch": 973, "lr": 7.707317254032775e-05}
{"train_loss": 0.19438788294792175, "global_step": 86627, "epoch": 973, "lr": 7.707268513986381e-05}
{"train_loss": 0.18323403596878052, "global_step": 86628, "epoch": 973, "lr": 7.707219773576027e-05}
{"train_loss": 0.10761448740959167, "global_step": 86629, "epoch": 973, "lr": 7.707171032801721e-05}
{"train_loss": 0.2387419492006302, "global_step": 86630, "epoch": 973, "lr": 7.70712229166347e-05}
{"train_loss": 0.19318172335624695, "global_step": 86631, "epoch": 973, "lr": 7.707073550161277e-05}
{"train_loss": 0.2221115529537201, "global_step": 86632, "epoch": 973, "lr": 7.707024808295154e-05}
{"train_loss": 0.18747662007808685, "global_step": 86633, "epoch": 973, "lr": 7.706976066065102e-05}
{"train_loss": 0.16318151354789734, "global_step": 86634, "epoch": 973, "lr": 7.706927323471131e-05}
{"train_loss": 0.33460983633995056, "global_step": 86635, "epoch": 973, "lr": 7.706878580513246e-05}
{"train_loss": 0.19201692938804626, "global_step": 86636, "epoch": 973, "lr": 7.706829837191455e-05}
{"train_loss": 0.1449403315782547, "global_step": 86637, "epoch": 973, "lr": 7.706781093505764e-05}
{"train_loss": 0.17474065721035004, "global_step": 86638, "epoch": 973, "lr": 7.706732349456179e-05}
{"train_loss": 0.1975424438714981, "global_step": 86639, "epoch": 973, "lr": 7.706683605042706e-05}
{"train_loss": 0.16751466691493988, "global_step": 86640, "epoch": 973, "lr": 7.706634860265352e-05}
{"train_loss": 0.12952563166618347, "global_step": 86641, "epoch": 973, "lr": 7.706586115124125e-05}
{"train_loss": 0.2121555060148239, "global_step": 86642, "epoch": 973, "lr": 7.70653736961903e-05}
{"train_loss": 0.15726564824581146, "global_step": 86643, "epoch": 973, "lr": 7.706488623750075e-05}
{"train_loss": 0.19352789223194122, "global_step": 86644, "epoch": 973, "lr": 7.706439877517264e-05}
{"train_loss": 0.15656740963459015, "global_step": 86645, "epoch": 973, "lr": 7.706391130920607e-05}
{"train_loss": 0.22197215259075165, "global_step": 86646, "epoch": 973, "lr": 7.706342383960107e-05}
{"train_loss": 0.2684016525745392, "global_step": 86647, "epoch": 973, "lr": 7.706293636635773e-05}
{"train_loss": 0.21699634194374084, "global_step": 86648, "epoch": 973, "lr": 7.706244888947612e-05}
{"train_loss": 0.20283758640289307, "global_step": 86649, "epoch": 973, "lr": 7.706196140895627e-05}
{"train_loss": 0.19299852848052979, "global_step": 86650, "epoch": 973, "lr": 7.706147392479828e-05}
{"train_loss": 0.15872368216514587, "global_step": 86651, "epoch": 973, "lr": 7.706098643700222e-05}
{"train_loss": 0.18536390364170074, "global_step": 86652, "epoch": 973, "lr": 7.70604989455681e-05}
{"train_loss": 0.25481197237968445, "global_step": 86653, "epoch": 973, "lr": 7.706001145049607e-05}
{"train_loss": 0.13418254256248474, "global_step": 86654, "epoch": 973, "lr": 7.705952395178613e-05}
{"train_loss": 0.20279565453529358, "global_step": 86655, "epoch": 973, "lr": 7.705903644943837e-05}
{"train_loss": 0.18449588119983673, "global_step": 86656, "epoch": 973, "lr": 7.705854894345285e-05}
{"train_loss": 0.1508629322052002, "global_step": 86657, "epoch": 973, "lr": 7.705806143382963e-05}
{"train_loss": 0.16328036785125732, "global_step": 86658, "epoch": 973, "lr": 7.70575739205688e-05}
{"train_loss": 0.17338798940181732, "global_step": 86659, "epoch": 973, "lr": 7.70570864036704e-05}
{"train_loss": 0.19827093183994293, "global_step": 86660, "epoch": 973, "lr": 7.70565988831345e-05}
{"train_loss": 0.19918982684612274, "global_step": 86661, "epoch": 973, "lr": 7.705611135896118e-05}
{"train_loss": 0.18095435202121735, "global_step": 86662, "epoch": 973, "lr": 7.70556238311505e-05}
{"train_loss": 0.17821398377418518, "global_step": 86663, "epoch": 973, "lr": 7.705513629970251e-05}
{"train_loss": 0.13490501046180725, "global_step": 86664, "epoch": 973, "lr": 7.70546487646173e-05}
{"train_loss": 0.2034730166196823, "global_step": 86665, "epoch": 973, "lr": 7.705416122589491e-05}
{"train_loss": 0.21511919796466827, "global_step": 86666, "epoch": 973, "lr": 7.705367368353543e-05}
{"train_loss": 0.2088291049003601, "global_step": 86667, "epoch": 973, "lr": 7.705318613753889e-05}
{"train_loss": 0.2348480075597763, "global_step": 86668, "epoch": 973, "lr": 7.70526985879054e-05}
{"train_loss": 0.17082640528678894, "global_step": 86669, "epoch": 973, "lr": 7.705221103463498e-05}
{"train_loss": 0.11663476377725601, "global_step": 86670, "epoch": 973, "lr": 7.705172347772774e-05}
{"train_loss": 0.2824985980987549, "global_step": 86671, "epoch": 973, "lr": 7.705123591718373e-05}
{"train_loss": 0.1999960094690323, "global_step": 86672, "epoch": 973, "lr": 7.705074835300298e-05}
{"train_loss": 0.19589222967624664, "global_step": 86673, "epoch": 973, "lr": 7.705026078518562e-05}
{"train_loss": 0.10322893410921097, "global_step": 86674, "epoch": 973, "lr": 7.704977321373168e-05}
{"train_loss": 0.23217745125293732, "global_step": 86675, "epoch": 973, "lr": 7.704928563864123e-05}
{"train_loss": 0.16848298907279968, "global_step": 86676, "epoch": 973, "lr": 7.704879805991431e-05}
{"train_loss": 0.18279340863227844, "global_step": 86677, "epoch": 973, "lr": 7.704831047755102e-05}
{"train_loss": 0.21086373925209045, "global_step": 86678, "epoch": 973, "lr": 7.704782289155142e-05}
{"train_loss": 0.3029388189315796, "global_step": 86679, "epoch": 973, "lr": 7.704733530191557e-05}
{"train_loss": 0.16295886039733887, "global_step": 86680, "epoch": 973, "lr": 7.704684770864352e-05}
{"train_loss": 0.23186150193214417, "global_step": 86681, "epoch": 973, "lr": 7.704636011173536e-05}
{"train_loss": 0.17964760959148407, "global_step": 86682, "epoch": 973, "lr": 7.704587251119115e-05}
{"train_loss": 0.18187037110328674, "global_step": 86683, "epoch": 973, "lr": 7.704538490701096e-05}
{"train_loss": 0.18543407320976257, "global_step": 86684, "epoch": 973, "lr": 7.704489729919484e-05}
{"train_loss": 0.18991785990388205, "global_step": 86685, "epoch": 973, "lr": 7.704440968774285e-05, "val_loss": 4.193371772766113}
{"train_loss": 0.18458639085292816, "global_step": 86686, "epoch": 974, "lr": 7.704392207265509e-05}
{"train_loss": 0.16117848455905914, "global_step": 86687, "epoch": 974, "lr": 7.704343445393158e-05}
{"train_loss": 0.15369972586631775, "global_step": 86688, "epoch": 974, "lr": 7.704294683157243e-05}
{"train_loss": 0.14633700251579285, "global_step": 86689, "epoch": 974, "lr": 7.704245920557768e-05}
{"train_loss": 0.16379159688949585, "global_step": 86690, "epoch": 974, "lr": 7.70419715759474e-05}
{"train_loss": 0.2232586294412613, "global_step": 86691, "epoch": 974, "lr": 7.704148394268166e-05}
{"train_loss": 0.10061788558959961, "global_step": 86692, "epoch": 974, "lr": 7.704099630578052e-05}
{"train_loss": 0.21406227350234985, "global_step": 86693, "epoch": 974, "lr": 7.704050866524403e-05}
{"train_loss": 0.19380047917366028, "global_step": 86694, "epoch": 974, "lr": 7.70400210210723e-05}
{"train_loss": 0.2057160884141922, "global_step": 86695, "epoch": 974, "lr": 7.703953337326537e-05}
{"train_loss": 0.17580075562000275, "global_step": 86696, "epoch": 974, "lr": 7.703904572182329e-05}
{"train_loss": 0.1096537858247757, "global_step": 86697, "epoch": 974, "lr": 7.703855806674614e-05}
{"train_loss": 0.18944238126277924, "global_step": 86698, "epoch": 974, "lr": 7.7038070408034e-05}
{"train_loss": 0.18711940944194794, "global_step": 86699, "epoch": 974, "lr": 7.703758274568691e-05}
{"train_loss": 0.2148234099149704, "global_step": 86700, "epoch": 974, "lr": 7.703709507970493e-05}
{"train_loss": 0.1407538801431656, "global_step": 86701, "epoch": 974, "lr": 7.703660741008817e-05}
{"train_loss": 0.14858806133270264, "global_step": 86702, "epoch": 974, "lr": 7.703611973683666e-05}
{"train_loss": 0.20565567910671234, "global_step": 86703, "epoch": 974, "lr": 7.703563205995046e-05}
{"train_loss": 0.20016774535179138, "global_step": 86704, "epoch": 974, "lr": 7.703514437942966e-05}
{"train_loss": 0.16105885803699493, "global_step": 86705, "epoch": 974, "lr": 7.703465669527431e-05}
{"train_loss": 0.1367734968662262, "global_step": 86706, "epoch": 974, "lr": 7.703416900748447e-05}
{"train_loss": 0.1979142427444458, "global_step": 86707, "epoch": 974, "lr": 7.703368131606023e-05}
{"train_loss": 0.10614155977964401, "global_step": 86708, "epoch": 974, "lr": 7.703319362100164e-05}
{"train_loss": 0.13313673436641693, "global_step": 86709, "epoch": 974, "lr": 7.703270592230877e-05}
{"train_loss": 0.13651412725448608, "global_step": 86710, "epoch": 974, "lr": 7.703221821998168e-05}
{"train_loss": 0.1764582395553589, "global_step": 86711, "epoch": 974, "lr": 7.703173051402044e-05}
{"train_loss": 0.24630720913410187, "global_step": 86712, "epoch": 974, "lr": 7.703124280442511e-05}
{"train_loss": 0.15772511065006256, "global_step": 86713, "epoch": 974, "lr": 7.703075509119575e-05}
{"train_loss": 0.14702628552913666, "global_step": 86714, "epoch": 974, "lr": 7.703026737433243e-05}
{"train_loss": 0.1360856294631958, "global_step": 86715, "epoch": 974, "lr": 7.702977965383524e-05}
{"train_loss": 0.3183889091014862, "global_step": 86716, "epoch": 974, "lr": 7.702929192970422e-05}
{"train_loss": 0.2320353090763092, "global_step": 86717, "epoch": 974, "lr": 7.702880420193944e-05}
{"train_loss": 0.15161658823490143, "global_step": 86718, "epoch": 974, "lr": 7.702831647054096e-05}
{"train_loss": 0.15609967708587646, "global_step": 86719, "epoch": 974, "lr": 7.702782873550887e-05}
{"train_loss": 0.19656355679035187, "global_step": 86720, "epoch": 974, "lr": 7.70273409968432e-05}
{"train_loss": 0.12655675411224365, "global_step": 86721, "epoch": 974, "lr": 7.702685325454404e-05}
{"train_loss": 0.11442507803440094, "global_step": 86722, "epoch": 974, "lr": 7.702636550861145e-05}
{"train_loss": 0.1610485315322876, "global_step": 86723, "epoch": 974, "lr": 7.702587775904549e-05}
{"train_loss": 0.23209913074970245, "global_step": 86724, "epoch": 974, "lr": 7.702539000584624e-05}
{"train_loss": 0.10891641676425934, "global_step": 86725, "epoch": 974, "lr": 7.702490224901373e-05}
{"train_loss": 0.20141606032848358, "global_step": 86726, "epoch": 974, "lr": 7.702441448854807e-05}
{"train_loss": 0.19838300347328186, "global_step": 86727, "epoch": 974, "lr": 7.702392672444932e-05}
{"train_loss": 0.24782468378543854, "global_step": 86728, "epoch": 974, "lr": 7.70234389567175e-05}
{"train_loss": 0.14454083144664764, "global_step": 86729, "epoch": 974, "lr": 7.702295118535273e-05}
{"train_loss": 0.17924882471561432, "global_step": 86730, "epoch": 974, "lr": 7.702246341035506e-05}
{"train_loss": 0.12488625198602676, "global_step": 86731, "epoch": 974, "lr": 7.702197563172453e-05}
{"train_loss": 0.17441532015800476, "global_step": 86732, "epoch": 974, "lr": 7.702148784946123e-05}
{"train_loss": 0.21191971004009247, "global_step": 86733, "epoch": 974, "lr": 7.702100006356524e-05}
{"train_loss": 0.2478310614824295, "global_step": 86734, "epoch": 974, "lr": 7.702051227403658e-05}
{"train_loss": 0.2512296736240387, "global_step": 86735, "epoch": 974, "lr": 7.702002448087535e-05}
{"train_loss": 0.12587995827198029, "global_step": 86736, "epoch": 974, "lr": 7.701953668408162e-05}
{"train_loss": 0.14459654688835144, "global_step": 86737, "epoch": 974, "lr": 7.701904888365543e-05}
{"train_loss": 0.26732587814331055, "global_step": 86738, "epoch": 974, "lr": 7.701856107959686e-05}
{"train_loss": 0.1665290892124176, "global_step": 86739, "epoch": 974, "lr": 7.701807327190597e-05}
{"train_loss": 0.15917371213436127, "global_step": 86740, "epoch": 974, "lr": 7.701758546058283e-05}
{"train_loss": 0.18205980956554413, "global_step": 86741, "epoch": 974, "lr": 7.701709764562752e-05}
{"train_loss": 0.19015854597091675, "global_step": 86742, "epoch": 974, "lr": 7.701660982704007e-05}
{"train_loss": 0.22225847840309143, "global_step": 86743, "epoch": 974, "lr": 7.701612200482056e-05}
{"train_loss": 0.18975470960140228, "global_step": 86744, "epoch": 974, "lr": 7.701563417896908e-05}
{"train_loss": 0.12524686753749847, "global_step": 86745, "epoch": 974, "lr": 7.701514634948568e-05}
{"train_loss": 0.17991919815540314, "global_step": 86746, "epoch": 974, "lr": 7.701465851637041e-05}
{"train_loss": 0.19266316294670105, "global_step": 86747, "epoch": 974, "lr": 7.701417067962335e-05}
{"train_loss": 0.21051530539989471, "global_step": 86748, "epoch": 974, "lr": 7.701368283924457e-05}
{"train_loss": 0.20924349129199982, "global_step": 86749, "epoch": 974, "lr": 7.701319499523414e-05}
{"train_loss": 0.32087284326553345, "global_step": 86750, "epoch": 974, "lr": 7.70127071475921e-05}
{"train_loss": 0.10072637349367142, "global_step": 86751, "epoch": 974, "lr": 7.701221929631855e-05}
{"train_loss": 0.21686875820159912, "global_step": 86752, "epoch": 974, "lr": 7.70117314414135e-05}
{"train_loss": 0.18585601449012756, "global_step": 86753, "epoch": 974, "lr": 7.701124358287708e-05}
{"train_loss": 0.18616150319576263, "global_step": 86754, "epoch": 974, "lr": 7.701075572070933e-05}
{"train_loss": 0.1499471217393875, "global_step": 86755, "epoch": 974, "lr": 7.701026785491029e-05}
{"train_loss": 0.1676495522260666, "global_step": 86756, "epoch": 974, "lr": 7.700977998548007e-05}
{"train_loss": 0.21285736560821533, "global_step": 86757, "epoch": 974, "lr": 7.70092921124187e-05}
{"train_loss": 0.20555603504180908, "global_step": 86758, "epoch": 974, "lr": 7.700880423572627e-05}
{"train_loss": 0.18800412118434906, "global_step": 86759, "epoch": 974, "lr": 7.700831635540282e-05}
{"train_loss": 0.2034166306257248, "global_step": 86760, "epoch": 974, "lr": 7.700782847144846e-05}
{"train_loss": 0.24117283523082733, "global_step": 86761, "epoch": 974, "lr": 7.700734058386318e-05}
{"train_loss": 0.2268839031457901, "global_step": 86762, "epoch": 974, "lr": 7.700685269264714e-05}
{"train_loss": 0.17491523921489716, "global_step": 86763, "epoch": 974, "lr": 7.700636479780033e-05}
{"train_loss": 0.13991019129753113, "global_step": 86764, "epoch": 974, "lr": 7.700587689932285e-05}
{"train_loss": 0.185127854347229, "global_step": 86765, "epoch": 974, "lr": 7.700538899721477e-05}
{"train_loss": 0.200681671500206, "global_step": 86766, "epoch": 974, "lr": 7.700490109147613e-05}
{"train_loss": 0.2504805326461792, "global_step": 86767, "epoch": 974, "lr": 7.700441318210701e-05}
{"train_loss": 0.1790532022714615, "global_step": 86768, "epoch": 974, "lr": 7.700392526910749e-05}
{"train_loss": 0.15509702265262604, "global_step": 86769, "epoch": 974, "lr": 7.700343735247762e-05}
{"train_loss": 0.19424986839294434, "global_step": 86770, "epoch": 974, "lr": 7.700294943221746e-05}
{"train_loss": 0.20169001817703247, "global_step": 86771, "epoch": 974, "lr": 7.700246150832708e-05}
{"train_loss": 0.3025911748409271, "global_step": 86772, "epoch": 974, "lr": 7.700197358080655e-05}
{"train_loss": 0.243733748793602, "global_step": 86773, "epoch": 974, "lr": 7.700148564965594e-05}
{"train_loss": 0.18508144802926632, "global_step": 86774, "epoch": 974, "lr": 7.70009977148753e-05, "val_loss": 4.316974639892578}
{"train_loss": 0.15105336904525757, "global_step": 86775, "epoch": 975, "lr": 7.700050977646471e-05}
{"train_loss": 0.1259794682264328, "global_step": 86776, "epoch": 975, "lr": 7.700002183442423e-05}
{"train_loss": 0.2240578681230545, "global_step": 86777, "epoch": 975, "lr": 7.699953388875392e-05}
{"train_loss": 0.14978614449501038, "global_step": 86778, "epoch": 975, "lr": 7.699904593945386e-05}
{"train_loss": 0.2363487184047699, "global_step": 86779, "epoch": 975, "lr": 7.69985579865241e-05}
{"train_loss": 0.22162757813930511, "global_step": 86780, "epoch": 975, "lr": 7.699807002996471e-05}
{"train_loss": 0.17039509117603302, "global_step": 86781, "epoch": 975, "lr": 7.699758206977577e-05}
{"train_loss": 0.16855253279209137, "global_step": 86782, "epoch": 975, "lr": 7.699709410595733e-05}
{"train_loss": 0.1966305375099182, "global_step": 86783, "epoch": 975, "lr": 7.699660613850947e-05}
{"train_loss": 0.25911980867385864, "global_step": 86784, "epoch": 975, "lr": 7.699611816743223e-05}
{"train_loss": 0.12621332705020905, "global_step": 86785, "epoch": 975, "lr": 7.699563019272568e-05}
{"train_loss": 0.25152140855789185, "global_step": 86786, "epoch": 975, "lr": 7.699514221438993e-05}
{"train_loss": 0.18771497905254364, "global_step": 86787, "epoch": 975, "lr": 7.6994654232425e-05}
{"train_loss": 0.19054563343524933, "global_step": 86788, "epoch": 975, "lr": 7.699416624683097e-05}
{"train_loss": 0.16274122893810272, "global_step": 86789, "epoch": 975, "lr": 7.699367825760789e-05}
{"train_loss": 0.1578272581100464, "global_step": 86790, "epoch": 975, "lr": 7.699319026475584e-05}
{"train_loss": 0.1405973881483078, "global_step": 86791, "epoch": 975, "lr": 7.69927022682749e-05}
{"train_loss": 0.1460009664297104, "global_step": 86792, "epoch": 975, "lr": 7.699221426816511e-05}
{"train_loss": 0.18745598196983337, "global_step": 86793, "epoch": 975, "lr": 7.699172626442655e-05}
{"train_loss": 0.11628760397434235, "global_step": 86794, "epoch": 975, "lr": 7.699123825705927e-05}
{"train_loss": 0.1917761266231537, "global_step": 86795, "epoch": 975, "lr": 7.699075024606335e-05}
{"train_loss": 0.2210533320903778, "global_step": 86796, "epoch": 975, "lr": 7.699026223143885e-05}
{"train_loss": 0.13294970989227295, "global_step": 86797, "epoch": 975, "lr": 7.698977421318585e-05}
{"train_loss": 0.20399023592472076, "global_step": 86798, "epoch": 975, "lr": 7.69892861913044e-05}
{"train_loss": 0.1684357076883316, "global_step": 86799, "epoch": 975, "lr": 7.698879816579457e-05}
{"train_loss": 0.19817396998405457, "global_step": 86800, "epoch": 975, "lr": 7.698831013665642e-05}
{"train_loss": 0.1580405980348587, "global_step": 86801, "epoch": 975, "lr": 7.698782210389002e-05}
{"train_loss": 0.1385272592306137, "global_step": 86802, "epoch": 975, "lr": 7.698733406749544e-05}
{"train_loss": 0.15944097936153412, "global_step": 86803, "epoch": 975, "lr": 7.698684602747275e-05}
{"train_loss": 0.21086344122886658, "global_step": 86804, "epoch": 975, "lr": 7.698635798382199e-05}
{"train_loss": 0.1540290266275406, "global_step": 86805, "epoch": 975, "lr": 7.698586993654326e-05}
{"train_loss": 0.17623931169509888, "global_step": 86806, "epoch": 975, "lr": 7.69853818856366e-05}
{"train_loss": 0.17606188356876373, "global_step": 86807, "epoch": 975, "lr": 7.69848938311021e-05}
{"train_loss": 0.26307132840156555, "global_step": 86808, "epoch": 975, "lr": 7.698440577293979e-05}
{"train_loss": 0.15608914196491241, "global_step": 86809, "epoch": 975, "lr": 7.698391771114977e-05}
{"train_loss": 0.13429686427116394, "global_step": 86810, "epoch": 975, "lr": 7.698342964573207e-05}
{"train_loss": 0.2042781561613083, "global_step": 86811, "epoch": 975, "lr": 7.69829415766868e-05}
{"train_loss": 0.14229752123355865, "global_step": 86812, "epoch": 975, "lr": 7.698245350401398e-05}
{"train_loss": 0.19710715115070343, "global_step": 86813, "epoch": 975, "lr": 7.698196542771373e-05}
{"train_loss": 0.2136969417333603, "global_step": 86814, "epoch": 975, "lr": 7.698147734778604e-05}
{"train_loss": 0.17457664012908936, "global_step": 86815, "epoch": 975, "lr": 7.698098926423105e-05}
{"train_loss": 0.2138027548789978, "global_step": 86816, "epoch": 975, "lr": 7.698050117704878e-05}
{"train_loss": 0.1173378974199295, "global_step": 86817, "epoch": 975, "lr": 7.698001308623931e-05}
{"train_loss": 0.17862215638160706, "global_step": 86818, "epoch": 975, "lr": 7.697952499180273e-05}
{"train_loss": 0.19615931808948517, "global_step": 86819, "epoch": 975, "lr": 7.697903689373905e-05}
{"train_loss": 0.24952790141105652, "global_step": 86820, "epoch": 975, "lr": 7.697854879204838e-05}
{"train_loss": 0.24859663844108582, "global_step": 86821, "epoch": 975, "lr": 7.697806068673078e-05}
{"train_loss": 0.22596299648284912, "global_step": 86822, "epoch": 975, "lr": 7.697757257778631e-05}
{"train_loss": 0.12360583990812302, "global_step": 86823, "epoch": 975, "lr": 7.697708446521502e-05}
{"train_loss": 0.2021736204624176, "global_step": 86824, "epoch": 975, "lr": 7.697659634901699e-05}
{"train_loss": 0.16296091675758362, "global_step": 86825, "epoch": 975, "lr": 7.69761082291923e-05}
{"train_loss": 0.1546381264925003, "global_step": 86826, "epoch": 975, "lr": 7.6975620105741e-05}
{"train_loss": 0.1873827576637268, "global_step": 86827, "epoch": 975, "lr": 7.697513197866316e-05}
{"train_loss": 0.2061518430709839, "global_step": 86828, "epoch": 975, "lr": 7.697464384795883e-05}
{"train_loss": 0.3455589711666107, "global_step": 86829, "epoch": 975, "lr": 7.697415571362809e-05}
{"train_loss": 0.214299738407135, "global_step": 86830, "epoch": 975, "lr": 7.697366757567102e-05}
{"train_loss": 0.10867021977901459, "global_step": 86831, "epoch": 975, "lr": 7.697317943408763e-05}
{"train_loss": 0.20395930111408234, "global_step": 86832, "epoch": 975, "lr": 7.697269128887806e-05}
{"train_loss": 0.12802159786224365, "global_step": 86833, "epoch": 975, "lr": 7.697220314004233e-05}
{"train_loss": 0.16026300191879272, "global_step": 86834, "epoch": 975, "lr": 7.697171498758051e-05}
{"train_loss": 0.19981518387794495, "global_step": 86835, "epoch": 975, "lr": 7.697122683149267e-05}
{"train_loss": 0.19590631127357483, "global_step": 86836, "epoch": 975, "lr": 7.697073867177888e-05}
{"train_loss": 0.27582281827926636, "global_step": 86837, "epoch": 975, "lr": 7.697025050843922e-05}
{"train_loss": 0.18584203720092773, "global_step": 86838, "epoch": 975, "lr": 7.696976234147372e-05}
{"train_loss": 0.12706835567951202, "global_step": 86839, "epoch": 975, "lr": 7.696927417088246e-05}
{"train_loss": 0.1780921220779419, "global_step": 86840, "epoch": 975, "lr": 7.696878599666553e-05}
{"train_loss": 0.15530084073543549, "global_step": 86841, "epoch": 975, "lr": 7.696829781882296e-05}
{"train_loss": 0.21683143079280853, "global_step": 86842, "epoch": 975, "lr": 7.696780963735484e-05}
{"train_loss": 0.16732148826122284, "global_step": 86843, "epoch": 975, "lr": 7.696732145226122e-05}
{"train_loss": 0.1603984385728836, "global_step": 86844, "epoch": 975, "lr": 7.696683326354219e-05}
{"train_loss": 0.2673914432525635, "global_step": 86845, "epoch": 975, "lr": 7.696634507119778e-05}
{"train_loss": 0.1730562299489975, "global_step": 86846, "epoch": 975, "lr": 7.696585687522809e-05}
{"train_loss": 0.18030202388763428, "global_step": 86847, "epoch": 975, "lr": 7.696536867563316e-05}
{"train_loss": 0.19118188321590424, "global_step": 86848, "epoch": 975, "lr": 7.696488047241305e-05}
{"train_loss": 0.1855776011943817, "global_step": 86849, "epoch": 975, "lr": 7.696439226556784e-05}
{"train_loss": 0.22754168510437012, "global_step": 86850, "epoch": 975, "lr": 7.696390405509762e-05}
{"train_loss": 0.17522673308849335, "global_step": 86851, "epoch": 975, "lr": 7.696341584100242e-05}
{"train_loss": 0.1738666146993637, "global_step": 86852, "epoch": 975, "lr": 7.69629276232823e-05}
{"train_loss": 0.1261909008026123, "global_step": 86853, "epoch": 975, "lr": 7.696243940193737e-05}
{"train_loss": 0.1933472454547882, "global_step": 86854, "epoch": 975, "lr": 7.696195117696764e-05}
{"train_loss": 0.17090003192424774, "global_step": 86855, "epoch": 975, "lr": 7.696146294837324e-05}
{"train_loss": 0.18616162240505219, "global_step": 86856, "epoch": 975, "lr": 7.696097471615418e-05}
{"train_loss": 0.2330552488565445, "global_step": 86857, "epoch": 975, "lr": 7.696048648031054e-05}
{"train_loss": 0.17571979761123657, "global_step": 86858, "epoch": 975, "lr": 7.69599982408424e-05}
{"train_loss": 0.1957726627588272, "global_step": 86859, "epoch": 975, "lr": 7.695950999774981e-05}
{"train_loss": 0.1375352293252945, "global_step": 86860, "epoch": 975, "lr": 7.695902175103285e-05}
{"train_loss": 0.20618370175361633, "global_step": 86861, "epoch": 975, "lr": 7.695853350069157e-05}
{"train_loss": 0.12166464328765869, "global_step": 86862, "epoch": 975, "lr": 7.695804524672606e-05}
{"train_loss": 0.18341962951287794, "global_step": 86863, "epoch": 975, "lr": 7.695755698913636e-05, "val_loss": 4.359239101409912, "train_action_mse_error": 22.749479293823242}
{"train_loss": 0.27879634499549866, "global_step": 86864, "epoch": 976, "lr": 7.695706872792254e-05}
{"train_loss": 0.19924989342689514, "global_step": 86865, "epoch": 976, "lr": 7.695658046308469e-05}
{"train_loss": 0.2384684532880783, "global_step": 86866, "epoch": 976, "lr": 7.695609219462284e-05}
{"train_loss": 0.16620218753814697, "global_step": 86867, "epoch": 976, "lr": 7.695560392253709e-05}
{"train_loss": 0.18566225469112396, "global_step": 86868, "epoch": 976, "lr": 7.695511564682745e-05}
{"train_loss": 0.1981901377439499, "global_step": 86869, "epoch": 976, "lr": 7.695462736749406e-05}
{"train_loss": 0.11937972158193588, "global_step": 86870, "epoch": 976, "lr": 7.695413908453692e-05}
{"train_loss": 0.20373767614364624, "global_step": 86871, "epoch": 976, "lr": 7.695365079795613e-05}
{"train_loss": 0.15752245485782623, "global_step": 86872, "epoch": 976, "lr": 7.695316250775177e-05}
{"train_loss": 0.2293463498353958, "global_step": 86873, "epoch": 976, "lr": 7.695267421392387e-05}
{"train_loss": 0.34035545587539673, "global_step": 86874, "epoch": 976, "lr": 7.69521859164725e-05}
{"train_loss": 0.17830206453800201, "global_step": 86875, "epoch": 976, "lr": 7.695169761539777e-05}
{"train_loss": 0.16418907046318054, "global_step": 86876, "epoch": 976, "lr": 7.695120931069969e-05}
{"train_loss": 0.15746742486953735, "global_step": 86877, "epoch": 976, "lr": 7.695072100237835e-05}
{"train_loss": 0.21695904433727264, "global_step": 86878, "epoch": 976, "lr": 7.695023269043382e-05}
{"train_loss": 0.16394183039665222, "global_step": 86879, "epoch": 976, "lr": 7.694974437486615e-05}
{"train_loss": 0.15791667997837067, "global_step": 86880, "epoch": 976, "lr": 7.694925605567544e-05}
{"train_loss": 0.29542726278305054, "global_step": 86881, "epoch": 976, "lr": 7.694876773286171e-05}
{"train_loss": 0.17382365465164185, "global_step": 86882, "epoch": 976, "lr": 7.694827940642504e-05}
{"train_loss": 0.1673986315727234, "global_step": 86883, "epoch": 976, "lr": 7.694779107636551e-05}
{"train_loss": 0.19205951690673828, "global_step": 86884, "epoch": 976, "lr": 7.694730274268318e-05}
{"train_loss": 0.12784956395626068, "global_step": 86885, "epoch": 976, "lr": 7.694681440537813e-05}
{"train_loss": 0.2809613347053528, "global_step": 86886, "epoch": 976, "lr": 7.69463260644504e-05}
{"train_loss": 0.19492864608764648, "global_step": 86887, "epoch": 976, "lr": 7.694583771990004e-05}
{"train_loss": 0.15958090126514435, "global_step": 86888, "epoch": 976, "lr": 7.694534937172716e-05}
{"train_loss": 0.16018922626972198, "global_step": 86889, "epoch": 976, "lr": 7.694486101993181e-05}
{"train_loss": 0.14472155272960663, "global_step": 86890, "epoch": 976, "lr": 7.694437266451404e-05}
{"train_loss": 0.17661052942276, "global_step": 86891, "epoch": 976, "lr": 7.694388430547395e-05}
{"train_loss": 0.11815675348043442, "global_step": 86892, "epoch": 976, "lr": 7.694339594281156e-05}
{"train_loss": 0.17338664829730988, "global_step": 86893, "epoch": 976, "lr": 7.694290757652696e-05}
{"train_loss": 0.1399276852607727, "global_step": 86894, "epoch": 976, "lr": 7.694241920662022e-05}
{"train_loss": 0.19240601360797882, "global_step": 86895, "epoch": 976, "lr": 7.69419308330914e-05}
{"train_loss": 0.22860747575759888, "global_step": 86896, "epoch": 976, "lr": 7.694144245594057e-05}
{"train_loss": 0.12363720685243607, "global_step": 86897, "epoch": 976, "lr": 7.694095407516779e-05}
{"train_loss": 0.17023810744285583, "global_step": 86898, "epoch": 976, "lr": 7.694046569077313e-05}
{"train_loss": 0.1615062803030014, "global_step": 86899, "epoch": 976, "lr": 7.693997730275664e-05}
{"train_loss": 0.15832766890525818, "global_step": 86900, "epoch": 976, "lr": 7.693948891111841e-05}
{"train_loss": 0.16826245188713074, "global_step": 86901, "epoch": 976, "lr": 7.693900051585849e-05}
{"train_loss": 0.14990191161632538, "global_step": 86902, "epoch": 976, "lr": 7.693851211697695e-05}
{"train_loss": 0.2272915244102478, "global_step": 86903, "epoch": 976, "lr": 7.693802371447385e-05}
{"train_loss": 0.274929404258728, "global_step": 86904, "epoch": 976, "lr": 7.693753530834928e-05}
{"train_loss": 0.18147312104701996, "global_step": 86905, "epoch": 976, "lr": 7.693704689860327e-05}
{"train_loss": 0.1986391544342041, "global_step": 86906, "epoch": 976, "lr": 7.69365584852359e-05}
{"train_loss": 0.2505624294281006, "global_step": 86907, "epoch": 976, "lr": 7.693607006824725e-05}
{"train_loss": 0.18984057009220123, "global_step": 86908, "epoch": 976, "lr": 7.693558164763737e-05}
{"train_loss": 0.16875874996185303, "global_step": 86909, "epoch": 976, "lr": 7.693509322340632e-05}
{"train_loss": 0.16334611177444458, "global_step": 86910, "epoch": 976, "lr": 7.693460479555418e-05}
{"train_loss": 0.12209915369749069, "global_step": 86911, "epoch": 976, "lr": 7.693411636408101e-05}
{"train_loss": 0.20344455540180206, "global_step": 86912, "epoch": 976, "lr": 7.693362792898689e-05}
{"train_loss": 0.17781896889209747, "global_step": 86913, "epoch": 976, "lr": 7.693313949027185e-05}
{"train_loss": 0.14033417403697968, "global_step": 86914, "epoch": 976, "lr": 7.693265104793598e-05}
{"train_loss": 0.16109530627727509, "global_step": 86915, "epoch": 976, "lr": 7.693216260197935e-05}
{"train_loss": 0.31725791096687317, "global_step": 86916, "epoch": 976, "lr": 7.693167415240203e-05}
{"train_loss": 0.14811915159225464, "global_step": 86917, "epoch": 976, "lr": 7.693118569920407e-05}
{"train_loss": 0.11566153168678284, "global_step": 86918, "epoch": 976, "lr": 7.693069724238553e-05}
{"train_loss": 0.1250818371772766, "global_step": 86919, "epoch": 976, "lr": 7.69302087819465e-05}
{"train_loss": 0.21390178799629211, "global_step": 86920, "epoch": 976, "lr": 7.692972031788701e-05}
{"train_loss": 0.22197586297988892, "global_step": 86921, "epoch": 976, "lr": 7.692923185020717e-05}
{"train_loss": 0.1821526437997818, "global_step": 86922, "epoch": 976, "lr": 7.692874337890701e-05}
{"train_loss": 0.2114199548959732, "global_step": 86923, "epoch": 976, "lr": 7.692825490398662e-05}
{"train_loss": 0.1620362251996994, "global_step": 86924, "epoch": 976, "lr": 7.692776642544603e-05}
{"train_loss": 0.21918319165706635, "global_step": 86925, "epoch": 976, "lr": 7.692727794328536e-05}
{"train_loss": 0.19460737705230713, "global_step": 86926, "epoch": 976, "lr": 7.692678945750462e-05}
{"train_loss": 0.17708471417427063, "global_step": 86927, "epoch": 976, "lr": 7.692630096810392e-05}
{"train_loss": 0.1908443570137024, "global_step": 86928, "epoch": 976, "lr": 7.69258124750833e-05}
{"train_loss": 0.1731548309326172, "global_step": 86929, "epoch": 976, "lr": 7.692532397844283e-05}
{"train_loss": 0.26467835903167725, "global_step": 86930, "epoch": 976, "lr": 7.692483547818259e-05}
{"train_loss": 0.1791795939207077, "global_step": 86931, "epoch": 976, "lr": 7.692434697430261e-05}
{"train_loss": 0.1608472466468811, "global_step": 86932, "epoch": 976, "lr": 7.692385846680301e-05}
{"train_loss": 0.21277974545955658, "global_step": 86933, "epoch": 976, "lr": 7.692336995568381e-05}
{"train_loss": 0.2128264605998993, "global_step": 86934, "epoch": 976, "lr": 7.692288144094509e-05}
{"train_loss": 0.23135899007320404, "global_step": 86935, "epoch": 976, "lr": 7.692239292258692e-05}
{"train_loss": 0.20211119949817657, "global_step": 86936, "epoch": 976, "lr": 7.692190440060935e-05}
{"train_loss": 0.1571890115737915, "global_step": 86937, "epoch": 976, "lr": 7.692141587501248e-05}
{"train_loss": 0.25356101989746094, "global_step": 86938, "epoch": 976, "lr": 7.692092734579635e-05}
{"train_loss": 0.18779541552066803, "global_step": 86939, "epoch": 976, "lr": 7.692043881296101e-05}
{"train_loss": 0.15791304409503937, "global_step": 86940, "epoch": 976, "lr": 7.691995027650656e-05}
{"train_loss": 0.1464949995279312, "global_step": 86941, "epoch": 976, "lr": 7.691946173643305e-05}
{"train_loss": 0.25364410877227783, "global_step": 86942, "epoch": 976, "lr": 7.691897319274056e-05}
{"train_loss": 0.2717215120792389, "global_step": 86943, "epoch": 976, "lr": 7.69184846454291e-05}
{"train_loss": 0.13789400458335876, "global_step": 86944, "epoch": 976, "lr": 7.691799609449882e-05}
{"train_loss": 0.14713431894779205, "global_step": 86945, "epoch": 976, "lr": 7.691750753994973e-05}
{"train_loss": 0.10664823651313782, "global_step": 86946, "epoch": 976, "lr": 7.69170189817819e-05}
{"train_loss": 0.13997311890125275, "global_step": 86947, "epoch": 976, "lr": 7.691653041999541e-05}
{"train_loss": 0.1918523758649826, "global_step": 86948, "epoch": 976, "lr": 7.691604185459033e-05}
{"train_loss": 0.17227572202682495, "global_step": 86949, "epoch": 976, "lr": 7.691555328556671e-05}
{"train_loss": 0.17879395186901093, "global_step": 86950, "epoch": 976, "lr": 7.691506471292463e-05}
{"train_loss": 0.15650597214698792, "global_step": 86951, "epoch": 976, "lr": 7.691457613666414e-05}
{"train_loss": 0.18651491437065468, "global_step": 86952, "epoch": 976, "lr": 7.69140875567853e-05, "val_loss": 4.242950439453125}
{"train_loss": 0.24509193003177643, "global_step": 86953, "epoch": 977, "lr": 7.69135989732882e-05}
{"train_loss": 0.12634067237377167, "global_step": 86954, "epoch": 977, "lr": 7.69131103861729e-05}
{"train_loss": 0.1494571417570114, "global_step": 86955, "epoch": 977, "lr": 7.691262179543945e-05}
{"train_loss": 0.12813560664653778, "global_step": 86956, "epoch": 977, "lr": 7.691213320108793e-05}
{"train_loss": 0.15696103870868683, "global_step": 86957, "epoch": 977, "lr": 7.691164460311841e-05}
{"train_loss": 0.14962856471538544, "global_step": 86958, "epoch": 977, "lr": 7.691115600153094e-05}
{"train_loss": 0.18911106884479523, "global_step": 86959, "epoch": 977, "lr": 7.69106673963256e-05}
{"train_loss": 0.14168451726436615, "global_step": 86960, "epoch": 977, "lr": 7.691017878750244e-05}
{"train_loss": 0.1720913052558899, "global_step": 86961, "epoch": 977, "lr": 7.690969017506153e-05}
{"train_loss": 0.1483522206544876, "global_step": 86962, "epoch": 977, "lr": 7.690920155900295e-05}
{"train_loss": 0.19422203302383423, "global_step": 86963, "epoch": 977, "lr": 7.690871293932676e-05}
{"train_loss": 0.20760801434516907, "global_step": 86964, "epoch": 977, "lr": 7.690822431603301e-05}
{"train_loss": 0.16163329780101776, "global_step": 86965, "epoch": 977, "lr": 7.690773568912178e-05}
{"train_loss": 0.20306739211082458, "global_step": 86966, "epoch": 977, "lr": 7.690724705859313e-05}
{"train_loss": 0.19810311496257782, "global_step": 86967, "epoch": 977, "lr": 7.690675842444714e-05}
{"train_loss": 0.1037198007106781, "global_step": 86968, "epoch": 977, "lr": 7.690626978668385e-05}
{"train_loss": 0.11798139661550522, "global_step": 86969, "epoch": 977, "lr": 7.690578114530336e-05}
{"train_loss": 0.19830463826656342, "global_step": 86970, "epoch": 977, "lr": 7.69052925003057e-05}
{"train_loss": 0.13858698308467865, "global_step": 86971, "epoch": 977, "lr": 7.690480385169095e-05}
{"train_loss": 0.14678992331027985, "global_step": 86972, "epoch": 977, "lr": 7.690431519945918e-05}
{"train_loss": 0.15535883605480194, "global_step": 86973, "epoch": 977, "lr": 7.690382654361046e-05}
{"train_loss": 0.12302294373512268, "global_step": 86974, "epoch": 977, "lr": 7.690333788414484e-05}
{"train_loss": 0.17377588152885437, "global_step": 86975, "epoch": 977, "lr": 7.690284922106238e-05}
{"train_loss": 0.13804891705513, "global_step": 86976, "epoch": 977, "lr": 7.69023605543632e-05}
{"train_loss": 0.20719340443611145, "global_step": 86977, "epoch": 977, "lr": 7.690187188404729e-05}
{"train_loss": 0.17229063808918, "global_step": 86978, "epoch": 977, "lr": 7.690138321011478e-05}
{"train_loss": 0.12694108486175537, "global_step": 86979, "epoch": 977, "lr": 7.690089453256568e-05}
{"train_loss": 0.2637368440628052, "global_step": 86980, "epoch": 977, "lr": 7.69004058514001e-05}
{"train_loss": 0.1574176698923111, "global_step": 86981, "epoch": 977, "lr": 7.689991716661807e-05}
{"train_loss": 0.17952018976211548, "global_step": 86982, "epoch": 977, "lr": 7.68994284782197e-05}
{"train_loss": 0.13163413107395172, "global_step": 86983, "epoch": 977, "lr": 7.6898939786205e-05}
{"train_loss": 0.16177378594875336, "global_step": 86984, "epoch": 977, "lr": 7.689845109057408e-05}
{"train_loss": 0.14375029504299164, "global_step": 86985, "epoch": 977, "lr": 7.6897962391327e-05}
{"train_loss": 0.14330194890499115, "global_step": 86986, "epoch": 977, "lr": 7.689747368846381e-05}
{"train_loss": 0.1906261146068573, "global_step": 86987, "epoch": 977, "lr": 7.689698498198458e-05}
{"train_loss": 0.2464030385017395, "global_step": 86988, "epoch": 977, "lr": 7.689649627188938e-05}
{"train_loss": 0.18016940355300903, "global_step": 86989, "epoch": 977, "lr": 7.689600755817828e-05}
{"train_loss": 0.2230999916791916, "global_step": 86990, "epoch": 977, "lr": 7.689551884085132e-05}
{"train_loss": 0.17490829527378082, "global_step": 86991, "epoch": 977, "lr": 7.689503011990859e-05}
{"train_loss": 0.17820003628730774, "global_step": 86992, "epoch": 977, "lr": 7.689454139535017e-05}
{"train_loss": 0.21491999924182892, "global_step": 86993, "epoch": 977, "lr": 7.689405266717608e-05}
{"train_loss": 0.1870366781949997, "global_step": 86994, "epoch": 977, "lr": 7.689356393538645e-05}
{"train_loss": 0.1591295748949051, "global_step": 86995, "epoch": 977, "lr": 7.689307519998128e-05}
{"train_loss": 0.17836140096187592, "global_step": 86996, "epoch": 977, "lr": 7.689258646096067e-05}
{"train_loss": 0.16697709262371063, "global_step": 86997, "epoch": 977, "lr": 7.689209771832468e-05}
{"train_loss": 0.22363507747650146, "global_step": 86998, "epoch": 977, "lr": 7.689160897207338e-05}
{"train_loss": 0.2326820343732834, "global_step": 86999, "epoch": 977, "lr": 7.689112022220682e-05}
{"train_loss": 0.09823750704526901, "global_step": 87000, "epoch": 977, "lr": 7.689063146872508e-05}
{"train_loss": 0.16018056869506836, "global_step": 87001, "epoch": 977, "lr": 7.689014271162822e-05}
{"train_loss": 0.21476440131664276, "global_step": 87002, "epoch": 977, "lr": 7.688965395091632e-05}
{"train_loss": 0.20827651023864746, "global_step": 87003, "epoch": 977, "lr": 7.688916518658941e-05}
{"train_loss": 0.17379868030548096, "global_step": 87004, "epoch": 977, "lr": 7.688867641864762e-05}
{"train_loss": 0.1341119259595871, "global_step": 87005, "epoch": 977, "lr": 7.688818764709094e-05}
{"train_loss": 0.13088594377040863, "global_step": 87006, "epoch": 977, "lr": 7.688769887191948e-05}
{"train_loss": 0.1867152899503708, "global_step": 87007, "epoch": 977, "lr": 7.68872100931333e-05}
{"train_loss": 0.1966131031513214, "global_step": 87008, "epoch": 977, "lr": 7.688672131073247e-05}
{"train_loss": 0.11886519938707352, "global_step": 87009, "epoch": 977, "lr": 7.688623252471703e-05}
{"train_loss": 0.19820863008499146, "global_step": 87010, "epoch": 977, "lr": 7.688574373508707e-05}
{"train_loss": 0.22257980704307556, "global_step": 87011, "epoch": 977, "lr": 7.688525494184265e-05}
{"train_loss": 0.1864302009344101, "global_step": 87012, "epoch": 977, "lr": 7.688476614498383e-05}
{"train_loss": 0.18954190611839294, "global_step": 87013, "epoch": 977, "lr": 7.68842773445107e-05}
{"train_loss": 0.18693886697292328, "global_step": 87014, "epoch": 977, "lr": 7.68837885404233e-05}
{"train_loss": 0.1599017083644867, "global_step": 87015, "epoch": 977, "lr": 7.688329973272169e-05}
{"train_loss": 0.18512260913848877, "global_step": 87016, "epoch": 977, "lr": 7.688281092140595e-05}
{"train_loss": 0.20556318759918213, "global_step": 87017, "epoch": 977, "lr": 7.688232210647616e-05}
{"train_loss": 0.2820218801498413, "global_step": 87018, "epoch": 977, "lr": 7.688183328793235e-05}
{"train_loss": 0.15305697917938232, "global_step": 87019, "epoch": 977, "lr": 7.688134446577461e-05}
{"train_loss": 0.18587477505207062, "global_step": 87020, "epoch": 977, "lr": 7.6880855640003e-05}
{"train_loss": 0.14854736626148224, "global_step": 87021, "epoch": 977, "lr": 7.68803668106176e-05}
{"train_loss": 0.21118146181106567, "global_step": 87022, "epoch": 977, "lr": 7.687987797761846e-05}
{"train_loss": 0.135011687874794, "global_step": 87023, "epoch": 977, "lr": 7.687938914100563e-05}
{"train_loss": 0.21628466248512268, "global_step": 87024, "epoch": 977, "lr": 7.68789003007792e-05}
{"train_loss": 0.1936715990304947, "global_step": 87025, "epoch": 977, "lr": 7.687841145693923e-05}
{"train_loss": 0.20690034329891205, "global_step": 87026, "epoch": 977, "lr": 7.687792260948581e-05}
{"train_loss": 0.14755764603614807, "global_step": 87027, "epoch": 977, "lr": 7.687743375841896e-05}
{"train_loss": 0.19348008930683136, "global_step": 87028, "epoch": 977, "lr": 7.687694490373876e-05}
{"train_loss": 0.20091480016708374, "global_step": 87029, "epoch": 977, "lr": 7.687645604544529e-05}
{"train_loss": 0.18997281789779663, "global_step": 87030, "epoch": 977, "lr": 7.68759671835386e-05}
{"train_loss": 0.16942428052425385, "global_step": 87031, "epoch": 977, "lr": 7.687547831801878e-05}
{"train_loss": 0.2177732139825821, "global_step": 87032, "epoch": 977, "lr": 7.687498944888587e-05}
{"train_loss": 0.18744197487831116, "global_step": 87033, "epoch": 977, "lr": 7.687450057613994e-05}
{"train_loss": 0.20169200003147125, "global_step": 87034, "epoch": 977, "lr": 7.687401169978107e-05}
{"train_loss": 0.20237937569618225, "global_step": 87035, "epoch": 977, "lr": 7.687352281980932e-05}
{"train_loss": 0.20429456233978271, "global_step": 87036, "epoch": 977, "lr": 7.687303393622474e-05}
{"train_loss": 0.21783138811588287, "global_step": 87037, "epoch": 977, "lr": 7.687254504902741e-05}
{"train_loss": 0.24432604014873505, "global_step": 87038, "epoch": 977, "lr": 7.68720561582174e-05}
{"train_loss": 0.19735395908355713, "global_step": 87039, "epoch": 977, "lr": 7.687156726379478e-05}
{"train_loss": 0.1856755018234253, "global_step": 87040, "epoch": 977, "lr": 7.687107836575959e-05}
{"train_loss": 0.17914624453595515, "global_step": 87041, "epoch": 977, "lr": 7.687058946411191e-05, "val_loss": 4.176016330718994}
{"train_loss": 0.16416187584400177, "global_step": 87042, "epoch": 978, "lr": 7.687010055885181e-05}
{"train_loss": 0.1368786096572876, "global_step": 87043, "epoch": 978, "lr": 7.686961164997935e-05}
{"train_loss": 0.21124579012393951, "global_step": 87044, "epoch": 978, "lr": 7.68691227374946e-05}
{"train_loss": 0.16770945489406586, "global_step": 87045, "epoch": 978, "lr": 7.686863382139763e-05}
{"train_loss": 0.14802710711956024, "global_step": 87046, "epoch": 978, "lr": 7.68681449016885e-05}
{"train_loss": 0.2163604348897934, "global_step": 87047, "epoch": 978, "lr": 7.686765597836727e-05}
{"train_loss": 0.23901724815368652, "global_step": 87048, "epoch": 978, "lr": 7.686716705143401e-05}
{"train_loss": 0.13268211483955383, "global_step": 87049, "epoch": 978, "lr": 7.68666781208888e-05}
{"train_loss": 0.17816811800003052, "global_step": 87050, "epoch": 978, "lr": 7.686618918673166e-05}
{"train_loss": 0.10392899811267853, "global_step": 87051, "epoch": 978, "lr": 7.68657002489627e-05}
{"train_loss": 0.19579194486141205, "global_step": 87052, "epoch": 978, "lr": 7.686521130758199e-05}
{"train_loss": 0.17568014562129974, "global_step": 87053, "epoch": 978, "lr": 7.686472236258957e-05}
{"train_loss": 0.20294079184532166, "global_step": 87054, "epoch": 978, "lr": 7.686423341398553e-05}
{"train_loss": 0.18844090402126312, "global_step": 87055, "epoch": 978, "lr": 7.686374446176991e-05}
{"train_loss": 0.24154673516750336, "global_step": 87056, "epoch": 978, "lr": 7.686325550594278e-05}
{"train_loss": 0.1912933886051178, "global_step": 87057, "epoch": 978, "lr": 7.686276654650422e-05}
{"train_loss": 0.20209448039531708, "global_step": 87058, "epoch": 978, "lr": 7.686227758345431e-05}
{"train_loss": 0.18232174217700958, "global_step": 87059, "epoch": 978, "lr": 7.686178861679305e-05}
{"train_loss": 0.21223439276218414, "global_step": 87060, "epoch": 978, "lr": 7.686129964652058e-05}
{"train_loss": 0.2230975180864334, "global_step": 87061, "epoch": 978, "lr": 7.686081067263694e-05}
{"train_loss": 0.18809674680233002, "global_step": 87062, "epoch": 978, "lr": 7.686032169514217e-05}
{"train_loss": 0.17054258286952972, "global_step": 87063, "epoch": 978, "lr": 7.685983271403636e-05}
{"train_loss": 0.23616351187229156, "global_step": 87064, "epoch": 978, "lr": 7.685934372931957e-05}
{"train_loss": 0.14488045871257782, "global_step": 87065, "epoch": 978, "lr": 7.685885474099188e-05}
{"train_loss": 0.2188938558101654, "global_step": 87066, "epoch": 978, "lr": 7.685836574905335e-05}
{"train_loss": 0.22528472542762756, "global_step": 87067, "epoch": 978, "lr": 7.685787675350403e-05}
{"train_loss": 0.15606525540351868, "global_step": 87068, "epoch": 978, "lr": 7.685738775434399e-05}
{"train_loss": 0.26339590549468994, "global_step": 87069, "epoch": 978, "lr": 7.685689875157331e-05}
{"train_loss": 0.21050547063350677, "global_step": 87070, "epoch": 978, "lr": 7.685640974519204e-05}
{"train_loss": 0.2595556974411011, "global_step": 87071, "epoch": 978, "lr": 7.685592073520028e-05}
{"train_loss": 0.19646918773651123, "global_step": 87072, "epoch": 978, "lr": 7.685543172159804e-05}
{"train_loss": 0.20007875561714172, "global_step": 87073, "epoch": 978, "lr": 7.685494270438543e-05}
{"train_loss": 0.17186762392520905, "global_step": 87074, "epoch": 978, "lr": 7.685445368356249e-05}
{"train_loss": 0.16634723544120789, "global_step": 87075, "epoch": 978, "lr": 7.68539646591293e-05}
{"train_loss": 0.2098139375448227, "global_step": 87076, "epoch": 978, "lr": 7.685347563108591e-05}
{"train_loss": 0.15841442346572876, "global_step": 87077, "epoch": 978, "lr": 7.685298659943242e-05}
{"train_loss": 0.17341791093349457, "global_step": 87078, "epoch": 978, "lr": 7.685249756416885e-05}
{"train_loss": 0.1550741195678711, "global_step": 87079, "epoch": 978, "lr": 7.68520085252953e-05}
{"train_loss": 0.20499934256076813, "global_step": 87080, "epoch": 978, "lr": 7.685151948281184e-05}
{"train_loss": 0.1341312676668167, "global_step": 87081, "epoch": 978, "lr": 7.685103043671851e-05}
{"train_loss": 0.17872732877731323, "global_step": 87082, "epoch": 978, "lr": 7.685054138701538e-05}
{"train_loss": 0.2009521871805191, "global_step": 87083, "epoch": 978, "lr": 7.685005233370253e-05}
{"train_loss": 0.16924458742141724, "global_step": 87084, "epoch": 978, "lr": 7.684956327678e-05}
{"train_loss": 0.17352403700351715, "global_step": 87085, "epoch": 978, "lr": 7.68490742162479e-05}
{"train_loss": 0.16679005324840546, "global_step": 87086, "epoch": 978, "lr": 7.684858515210625e-05}
{"train_loss": 0.17617879807949066, "global_step": 87087, "epoch": 978, "lr": 7.684809608435512e-05}
{"train_loss": 0.16832272708415985, "global_step": 87088, "epoch": 978, "lr": 7.684760701299462e-05}
{"train_loss": 0.18086162209510803, "global_step": 87089, "epoch": 978, "lr": 7.684711793802478e-05}
{"train_loss": 0.25838273763656616, "global_step": 87090, "epoch": 978, "lr": 7.684662885944565e-05}
{"train_loss": 0.18737244606018066, "global_step": 87091, "epoch": 978, "lr": 7.684613977725734e-05}
{"train_loss": 0.19893485307693481, "global_step": 87092, "epoch": 978, "lr": 7.68456506914599e-05}
{"train_loss": 0.17957791686058044, "global_step": 87093, "epoch": 978, "lr": 7.68451616020534e-05}
{"train_loss": 0.26130515336990356, "global_step": 87094, "epoch": 978, "lr": 7.684467250903786e-05}
{"train_loss": 0.10481899231672287, "global_step": 87095, "epoch": 978, "lr": 7.68441834124134e-05}
{"train_loss": 0.18395258486270905, "global_step": 87096, "epoch": 978, "lr": 7.684369431218007e-05}
{"train_loss": 0.16867361962795258, "global_step": 87097, "epoch": 978, "lr": 7.684320520833792e-05}
{"train_loss": 0.16111783683300018, "global_step": 87098, "epoch": 978, "lr": 7.684271610088702e-05}
{"train_loss": 0.22542470693588257, "global_step": 87099, "epoch": 978, "lr": 7.684222698982747e-05}
{"train_loss": 0.20002062618732452, "global_step": 87100, "epoch": 978, "lr": 7.68417378751593e-05}
{"train_loss": 0.2049359530210495, "global_step": 87101, "epoch": 978, "lr": 7.684124875688258e-05}
{"train_loss": 0.2479434758424759, "global_step": 87102, "epoch": 978, "lr": 7.684075963499738e-05}
{"train_loss": 0.14787404239177704, "global_step": 87103, "epoch": 978, "lr": 7.684027050950377e-05}
{"train_loss": 0.16808536648750305, "global_step": 87104, "epoch": 978, "lr": 7.683978138040182e-05}
{"train_loss": 0.17256967723369598, "global_step": 87105, "epoch": 978, "lr": 7.683929224769157e-05}
{"train_loss": 0.17993126809597015, "global_step": 87106, "epoch": 978, "lr": 7.683880311137313e-05}
{"train_loss": 0.13228844106197357, "global_step": 87107, "epoch": 978, "lr": 7.683831397144652e-05}
{"train_loss": 0.1993284374475479, "global_step": 87108, "epoch": 978, "lr": 7.683782482791183e-05}
{"train_loss": 0.21363094449043274, "global_step": 87109, "epoch": 978, "lr": 7.683733568076912e-05}
{"train_loss": 0.17057210206985474, "global_step": 87110, "epoch": 978, "lr": 7.683684653001845e-05}
{"train_loss": 0.19366580247879028, "global_step": 87111, "epoch": 978, "lr": 7.683635737565992e-05}
{"train_loss": 0.21725516021251678, "global_step": 87112, "epoch": 978, "lr": 7.683586821769356e-05}
{"train_loss": 0.18051548302173615, "global_step": 87113, "epoch": 978, "lr": 7.683537905611944e-05}
{"train_loss": 0.15923282504081726, "global_step": 87114, "epoch": 978, "lr": 7.683488989093763e-05}
{"train_loss": 0.1913590133190155, "global_step": 87115, "epoch": 978, "lr": 7.683440072214819e-05}
{"train_loss": 0.23932668566703796, "global_step": 87116, "epoch": 978, "lr": 7.68339115497512e-05}
{"train_loss": 0.13947182893753052, "global_step": 87117, "epoch": 978, "lr": 7.683342237374671e-05}
{"train_loss": 0.1856609433889389, "global_step": 87118, "epoch": 978, "lr": 7.68329331941348e-05}
{"train_loss": 0.20816627144813538, "global_step": 87119, "epoch": 978, "lr": 7.683244401091553e-05}
{"train_loss": 0.10636927932500839, "global_step": 87120, "epoch": 978, "lr": 7.683195482408896e-05}
{"train_loss": 0.12822307646274567, "global_step": 87121, "epoch": 978, "lr": 7.683146563365517e-05}
{"train_loss": 0.20869258046150208, "global_step": 87122, "epoch": 978, "lr": 7.683097643961422e-05}
{"train_loss": 0.26799142360687256, "global_step": 87123, "epoch": 978, "lr": 7.683048724196614e-05}
{"train_loss": 0.2694264054298401, "global_step": 87124, "epoch": 978, "lr": 7.682999804071105e-05}
{"train_loss": 0.2834846079349518, "global_step": 87125, "epoch": 978, "lr": 7.682950883584899e-05}
{"train_loss": 0.18189001083374023, "global_step": 87126, "epoch": 978, "lr": 7.682901962738002e-05}
{"train_loss": 0.18841013312339783, "global_step": 87127, "epoch": 978, "lr": 7.682853041530424e-05}
{"train_loss": 0.1557772010564804, "global_step": 87128, "epoch": 978, "lr": 7.682804119962168e-05}
{"train_loss": 0.14454951882362366, "global_step": 87129, "epoch": 978, "lr": 7.682755198033241e-05}
{"train_loss": 0.18946147366855923, "global_step": 87130, "epoch": 978, "lr": 7.682706275743651e-05, "val_loss": 4.3705034255981445}
{"train_loss": 0.24663521349430084, "global_step": 87131, "epoch": 979, "lr": 7.682657353093405e-05}
{"train_loss": 0.21500270068645477, "global_step": 87132, "epoch": 979, "lr": 7.682608430082507e-05}
{"train_loss": 0.18118613958358765, "global_step": 87133, "epoch": 979, "lr": 7.682559506710964e-05}
{"train_loss": 0.19095012545585632, "global_step": 87134, "epoch": 979, "lr": 7.682510582978786e-05}
{"train_loss": 0.19252486526966095, "global_step": 87135, "epoch": 979, "lr": 7.682461658885976e-05}
{"train_loss": 0.16514421999454498, "global_step": 87136, "epoch": 979, "lr": 7.68241273443254e-05}
{"train_loss": 0.23717087507247925, "global_step": 87137, "epoch": 979, "lr": 7.682363809618489e-05}
{"train_loss": 0.2260379046201706, "global_step": 87138, "epoch": 979, "lr": 7.682314884443826e-05}
{"train_loss": 0.18952149152755737, "global_step": 87139, "epoch": 979, "lr": 7.682265958908557e-05}
{"train_loss": 0.1158008947968483, "global_step": 87140, "epoch": 979, "lr": 7.68221703301269e-05}
{"train_loss": 0.17484638094902039, "global_step": 87141, "epoch": 979, "lr": 7.682168106756235e-05}
{"train_loss": 0.13810916244983673, "global_step": 87142, "epoch": 979, "lr": 7.682119180139191e-05}
{"train_loss": 0.12141495943069458, "global_step": 87143, "epoch": 979, "lr": 7.682070253161571e-05}
{"train_loss": 0.09787700325250626, "global_step": 87144, "epoch": 979, "lr": 7.682021325823379e-05}
{"train_loss": 0.19510087370872498, "global_step": 87145, "epoch": 979, "lr": 7.681972398124622e-05}
{"train_loss": 0.2356683313846588, "global_step": 87146, "epoch": 979, "lr": 7.681923470065306e-05}
{"train_loss": 0.1959686577320099, "global_step": 87147, "epoch": 979, "lr": 7.68187454164544e-05}
{"train_loss": 0.1891564428806305, "global_step": 87148, "epoch": 979, "lr": 7.681825612865027e-05}
{"train_loss": 0.19606125354766846, "global_step": 87149, "epoch": 979, "lr": 7.681776683724075e-05}
{"train_loss": 0.20145395398139954, "global_step": 87150, "epoch": 979, "lr": 7.681727754222592e-05}
{"train_loss": 0.17401628196239471, "global_step": 87151, "epoch": 979, "lr": 7.681678824360584e-05}
{"train_loss": 0.15671920776367188, "global_step": 87152, "epoch": 979, "lr": 7.681629894138055e-05}
{"train_loss": 0.23863562941551208, "global_step": 87153, "epoch": 979, "lr": 7.681580963555013e-05}
{"train_loss": 0.20044030249118805, "global_step": 87154, "epoch": 979, "lr": 7.681532032611468e-05}
{"train_loss": 0.1766120046377182, "global_step": 87155, "epoch": 979, "lr": 7.68148310130742e-05}
{"train_loss": 0.20712345838546753, "global_step": 87156, "epoch": 979, "lr": 7.681434169642883e-05}
{"train_loss": 0.1451295167207718, "global_step": 87157, "epoch": 979, "lr": 7.681385237617858e-05}
{"train_loss": 0.14283105731010437, "global_step": 87158, "epoch": 979, "lr": 7.681336305232354e-05}
{"train_loss": 0.17726081609725952, "global_step": 87159, "epoch": 979, "lr": 7.681287372486376e-05}
{"train_loss": 0.14242956042289734, "global_step": 87160, "epoch": 979, "lr": 7.681238439379933e-05}
{"train_loss": 0.17234830558300018, "global_step": 87161, "epoch": 979, "lr": 7.681189505913028e-05}
{"train_loss": 0.12944278120994568, "global_step": 87162, "epoch": 979, "lr": 7.681140572085671e-05}
{"train_loss": 0.20886778831481934, "global_step": 87163, "epoch": 979, "lr": 7.681091637897869e-05}
{"train_loss": 0.1336384117603302, "global_step": 87164, "epoch": 979, "lr": 7.681042703349623e-05}
{"train_loss": 0.1493002474308014, "global_step": 87165, "epoch": 979, "lr": 7.680993768440948e-05}
{"train_loss": 0.19631728529930115, "global_step": 87166, "epoch": 979, "lr": 7.680944833171842e-05}
{"train_loss": 0.16828322410583496, "global_step": 87167, "epoch": 979, "lr": 7.680895897542317e-05}
{"train_loss": 0.24546043574810028, "global_step": 87168, "epoch": 979, "lr": 7.68084696155238e-05}
{"train_loss": 0.19097089767456055, "global_step": 87169, "epoch": 979, "lr": 7.680798025202034e-05}
{"train_loss": 0.12749417126178741, "global_step": 87170, "epoch": 979, "lr": 7.680749088491289e-05}
{"train_loss": 0.2626951038837433, "global_step": 87171, "epoch": 979, "lr": 7.68070015142015e-05}
{"train_loss": 0.13985894620418549, "global_step": 87172, "epoch": 979, "lr": 7.680651213988621e-05}
{"train_loss": 0.1679188311100006, "global_step": 87173, "epoch": 979, "lr": 7.680602276196714e-05}
{"train_loss": 0.2478693127632141, "global_step": 87174, "epoch": 979, "lr": 7.680553338044431e-05}
{"train_loss": 0.14533549547195435, "global_step": 87175, "epoch": 979, "lr": 7.68050439953178e-05}
{"train_loss": 0.2019096463918686, "global_step": 87176, "epoch": 979, "lr": 7.680455460658768e-05}
{"train_loss": 0.15341778099536896, "global_step": 87177, "epoch": 979, "lr": 7.680406521425403e-05}
{"train_loss": 0.1729794293642044, "global_step": 87178, "epoch": 979, "lr": 7.680357581831687e-05}
{"train_loss": 0.21223433315753937, "global_step": 87179, "epoch": 979, "lr": 7.680308641877631e-05}
{"train_loss": 0.22726988792419434, "global_step": 87180, "epoch": 979, "lr": 7.680259701563242e-05}
{"train_loss": 0.1926463395357132, "global_step": 87181, "epoch": 979, "lr": 7.680210760888522e-05}
{"train_loss": 0.28382501006126404, "global_step": 87182, "epoch": 979, "lr": 7.680161819853482e-05}
{"train_loss": 0.18135662376880646, "global_step": 87183, "epoch": 979, "lr": 7.680112878458128e-05}
{"train_loss": 0.12524977326393127, "global_step": 87184, "epoch": 979, "lr": 7.680063936702463e-05}
{"train_loss": 0.19422592222690582, "global_step": 87185, "epoch": 979, "lr": 7.680014994586498e-05}
{"train_loss": 0.17417828738689423, "global_step": 87186, "epoch": 979, "lr": 7.679966052110237e-05}
{"train_loss": 0.1852077841758728, "global_step": 87187, "epoch": 979, "lr": 7.679917109273687e-05}
{"train_loss": 0.19975897669792175, "global_step": 87188, "epoch": 979, "lr": 7.679868166076854e-05}
{"train_loss": 0.13081781566143036, "global_step": 87189, "epoch": 979, "lr": 7.679819222519748e-05}
{"train_loss": 0.20846517384052277, "global_step": 87190, "epoch": 979, "lr": 7.67977027860237e-05}
{"train_loss": 0.14674942195415497, "global_step": 87191, "epoch": 979, "lr": 7.679721334324731e-05}
{"train_loss": 0.18609265983104706, "global_step": 87192, "epoch": 979, "lr": 7.679672389686837e-05}
{"train_loss": 0.1731296330690384, "global_step": 87193, "epoch": 979, "lr": 7.679623444688691e-05}
{"train_loss": 0.17753474414348602, "global_step": 87194, "epoch": 979, "lr": 7.679574499330304e-05}
{"train_loss": 0.18232199549674988, "global_step": 87195, "epoch": 979, "lr": 7.679525553611682e-05}
{"train_loss": 0.13946017622947693, "global_step": 87196, "epoch": 979, "lr": 7.679476607532828e-05}
{"train_loss": 0.11246953159570694, "global_step": 87197, "epoch": 979, "lr": 7.679427661093752e-05}
{"train_loss": 0.14563986659049988, "global_step": 87198, "epoch": 979, "lr": 7.67937871429446e-05}
{"train_loss": 0.194289892911911, "global_step": 87199, "epoch": 979, "lr": 7.679329767134957e-05}
{"train_loss": 0.16396832466125488, "global_step": 87200, "epoch": 979, "lr": 7.679280819615253e-05}
{"train_loss": 0.15938851237297058, "global_step": 87201, "epoch": 979, "lr": 7.67923187173535e-05}
{"train_loss": 0.15286171436309814, "global_step": 87202, "epoch": 979, "lr": 7.679182923495259e-05}
{"train_loss": 0.16900530457496643, "global_step": 87203, "epoch": 979, "lr": 7.679133974894983e-05}
{"train_loss": 0.1701376736164093, "global_step": 87204, "epoch": 979, "lr": 7.679085025934531e-05}
{"train_loss": 0.2151985764503479, "global_step": 87205, "epoch": 979, "lr": 7.679036076613908e-05}
{"train_loss": 0.2269383668899536, "global_step": 87206, "epoch": 979, "lr": 7.678987126933122e-05}
{"train_loss": 0.21725577116012573, "global_step": 87207, "epoch": 979, "lr": 7.678938176892179e-05}
{"train_loss": 0.16879862546920776, "global_step": 87208, "epoch": 979, "lr": 7.678889226491085e-05}
{"train_loss": 0.15176215767860413, "global_step": 87209, "epoch": 979, "lr": 7.678840275729846e-05}
{"train_loss": 0.20960025489330292, "global_step": 87210, "epoch": 979, "lr": 7.678791324608472e-05}
{"train_loss": 0.13293497264385223, "global_step": 87211, "epoch": 979, "lr": 7.678742373126964e-05}
{"train_loss": 0.15346413850784302, "global_step": 87212, "epoch": 979, "lr": 7.678693421285335e-05}
{"train_loss": 0.1940467655658722, "global_step": 87213, "epoch": 979, "lr": 7.678644469083585e-05}
{"train_loss": 0.14727607369422913, "global_step": 87214, "epoch": 979, "lr": 7.678595516521727e-05}
{"train_loss": 0.21274051070213318, "global_step": 87215, "epoch": 979, "lr": 7.678546563599762e-05}
{"train_loss": 0.1531064659357071, "global_step": 87216, "epoch": 979, "lr": 7.6784976103177e-05}
{"train_loss": 0.20462751388549805, "global_step": 87217, "epoch": 979, "lr": 7.678448656675546e-05}
{"train_loss": 0.09528639167547226, "global_step": 87218, "epoch": 979, "lr": 7.678399702673308e-05}
{"train_loss": 0.17908253696527374, "global_step": 87219, "epoch": 979, "lr": 7.678350748310992e-05, "val_loss": 4.375064849853516}
{"train_loss": 0.1598546952009201, "global_step": 87220, "epoch": 980, "lr": 7.678301793588605e-05}
{"train_loss": 0.24755734205245972, "global_step": 87221, "epoch": 980, "lr": 7.67825283850615e-05}
{"train_loss": 0.1698642373085022, "global_step": 87222, "epoch": 980, "lr": 7.678203883063639e-05}
{"train_loss": 0.2005843222141266, "global_step": 87223, "epoch": 980, "lr": 7.678154927261076e-05}
{"train_loss": 0.21054013073444366, "global_step": 87224, "epoch": 980, "lr": 7.678105971098469e-05}
{"train_loss": 0.15340688824653625, "global_step": 87225, "epoch": 980, "lr": 7.678057014575821e-05}
{"train_loss": 0.15706072747707367, "global_step": 87226, "epoch": 980, "lr": 7.678008057693142e-05}
{"train_loss": 0.18137921392917633, "global_step": 87227, "epoch": 980, "lr": 7.677959100450436e-05}
{"train_loss": 0.14102160930633545, "global_step": 87228, "epoch": 980, "lr": 7.677910142847713e-05}
{"train_loss": 0.22417086362838745, "global_step": 87229, "epoch": 980, "lr": 7.677861184884979e-05}
{"train_loss": 0.14927595853805542, "global_step": 87230, "epoch": 980, "lr": 7.677812226562235e-05}
{"train_loss": 0.1536499410867691, "global_step": 87231, "epoch": 980, "lr": 7.677763267879495e-05}
{"train_loss": 0.1717498004436493, "global_step": 87232, "epoch": 980, "lr": 7.677714308836761e-05}
{"train_loss": 0.16538000106811523, "global_step": 87233, "epoch": 980, "lr": 7.67766534943404e-05}
{"train_loss": 0.3139205276966095, "global_step": 87234, "epoch": 980, "lr": 7.67761638967134e-05}
{"train_loss": 0.11356427520513535, "global_step": 87235, "epoch": 980, "lr": 7.677567429548667e-05}
{"train_loss": 0.22815006971359253, "global_step": 87236, "epoch": 980, "lr": 7.677518469066029e-05}
{"train_loss": 0.18781383335590363, "global_step": 87237, "epoch": 980, "lr": 7.677469508223432e-05}
{"train_loss": 0.22311601042747498, "global_step": 87238, "epoch": 980, "lr": 7.67742054702088e-05}
{"train_loss": 0.1618019938468933, "global_step": 87239, "epoch": 980, "lr": 7.677371585458381e-05}
{"train_loss": 0.18681710958480835, "global_step": 87240, "epoch": 980, "lr": 7.677322623535943e-05}
{"train_loss": 0.15494689345359802, "global_step": 87241, "epoch": 980, "lr": 7.677273661253573e-05}
{"train_loss": 0.26352667808532715, "global_step": 87242, "epoch": 980, "lr": 7.677224698611274e-05}
{"train_loss": 0.15682300925254822, "global_step": 87243, "epoch": 980, "lr": 7.677175735609055e-05}
{"train_loss": 0.16203317046165466, "global_step": 87244, "epoch": 980, "lr": 7.677126772246924e-05}
{"train_loss": 0.2995651960372925, "global_step": 87245, "epoch": 980, "lr": 7.677077808524885e-05}
{"train_loss": 0.16332916915416718, "global_step": 87246, "epoch": 980, "lr": 7.677028844442945e-05}
{"train_loss": 0.15615518391132355, "global_step": 87247, "epoch": 980, "lr": 7.676979880001112e-05}
{"train_loss": 0.19426125288009644, "global_step": 87248, "epoch": 980, "lr": 7.676930915199392e-05}
{"train_loss": 0.2805236577987671, "global_step": 87249, "epoch": 980, "lr": 7.676881950037788e-05}
{"train_loss": 0.22551831603050232, "global_step": 87250, "epoch": 980, "lr": 7.676832984516314e-05}
{"train_loss": 0.18533585965633392, "global_step": 87251, "epoch": 980, "lr": 7.67678401863497e-05}
{"train_loss": 0.16925393044948578, "global_step": 87252, "epoch": 980, "lr": 7.676735052393765e-05}
{"train_loss": 0.24812975525856018, "global_step": 87253, "epoch": 980, "lr": 7.676686085792706e-05}
{"train_loss": 0.19100596010684967, "global_step": 87254, "epoch": 980, "lr": 7.6766371188318e-05}
{"train_loss": 0.160488560795784, "global_step": 87255, "epoch": 980, "lr": 7.676588151511053e-05}
{"train_loss": 0.17820927500724792, "global_step": 87256, "epoch": 980, "lr": 7.676539183830468e-05}
{"train_loss": 0.2087450921535492, "global_step": 87257, "epoch": 980, "lr": 7.67649021579006e-05}
{"train_loss": 0.2501148283481598, "global_step": 87258, "epoch": 980, "lr": 7.676441247389827e-05}
{"train_loss": 0.1945338100194931, "global_step": 87259, "epoch": 980, "lr": 7.67639227862978e-05}
{"train_loss": 0.12979523837566376, "global_step": 87260, "epoch": 980, "lr": 7.676343309509926e-05}
{"train_loss": 0.16848315298557281, "global_step": 87261, "epoch": 980, "lr": 7.676294340030268e-05}
{"train_loss": 0.21991388499736786, "global_step": 87262, "epoch": 980, "lr": 7.676245370190816e-05}
{"train_loss": 0.1564444899559021, "global_step": 87263, "epoch": 980, "lr": 7.676196399991576e-05}
{"train_loss": 0.09507768601179123, "global_step": 87264, "epoch": 980, "lr": 7.676147429432553e-05}
{"train_loss": 0.2566128969192505, "global_step": 87265, "epoch": 980, "lr": 7.676098458513756e-05}
{"train_loss": 0.23010176420211792, "global_step": 87266, "epoch": 980, "lr": 7.676049487235188e-05}
{"train_loss": 0.20551253855228424, "global_step": 87267, "epoch": 980, "lr": 7.676000515596859e-05}
{"train_loss": 0.20249038934707642, "global_step": 87268, "epoch": 980, "lr": 7.675951543598776e-05}
{"train_loss": 0.19578422605991364, "global_step": 87269, "epoch": 980, "lr": 7.67590257124094e-05}
{"train_loss": 0.19987457990646362, "global_step": 87270, "epoch": 980, "lr": 7.675853598523364e-05}
{"train_loss": 0.2775910198688507, "global_step": 87271, "epoch": 980, "lr": 7.675804625446052e-05}
{"train_loss": 0.07641886919736862, "global_step": 87272, "epoch": 980, "lr": 7.675755652009011e-05}
{"train_loss": 0.13646896183490753, "global_step": 87273, "epoch": 980, "lr": 7.675706678212247e-05}
{"train_loss": 0.15652012825012207, "global_step": 87274, "epoch": 980, "lr": 7.675657704055765e-05}
{"train_loss": 0.14957037568092346, "global_step": 87275, "epoch": 980, "lr": 7.675608729539576e-05}
{"train_loss": 0.18930037319660187, "global_step": 87276, "epoch": 980, "lr": 7.675559754663682e-05}
{"train_loss": 0.15113627910614014, "global_step": 87277, "epoch": 980, "lr": 7.675510779428094e-05}
{"train_loss": 0.1055225059390068, "global_step": 87278, "epoch": 980, "lr": 7.675461803832815e-05}
{"train_loss": 0.14979548752307892, "global_step": 87279, "epoch": 980, "lr": 7.675412827877853e-05}
{"train_loss": 0.13094942271709442, "global_step": 87280, "epoch": 980, "lr": 7.675363851563213e-05}
{"train_loss": 0.1480218470096588, "global_step": 87281, "epoch": 980, "lr": 7.675314874888904e-05}
{"train_loss": 0.13167113065719604, "global_step": 87282, "epoch": 980, "lr": 7.675265897854931e-05}
{"train_loss": 0.1772424727678299, "global_step": 87283, "epoch": 980, "lr": 7.675216920461301e-05}
{"train_loss": 0.16431310772895813, "global_step": 87284, "epoch": 980, "lr": 7.675167942708021e-05}
{"train_loss": 0.17232874035835266, "global_step": 87285, "epoch": 980, "lr": 7.675118964595099e-05}
{"train_loss": 0.1556912362575531, "global_step": 87286, "epoch": 980, "lr": 7.675069986122538e-05}
{"train_loss": 0.24004440009593964, "global_step": 87287, "epoch": 980, "lr": 7.675021007290346e-05}
{"train_loss": 0.2198694944381714, "global_step": 87288, "epoch": 980, "lr": 7.674972028098532e-05}
{"train_loss": 0.10496615618467331, "global_step": 87289, "epoch": 980, "lr": 7.674923048547098e-05}
{"train_loss": 0.1951712816953659, "global_step": 87290, "epoch": 980, "lr": 7.674874068636056e-05}
{"train_loss": 0.2641568183898926, "global_step": 87291, "epoch": 980, "lr": 7.674825088365407e-05}
{"train_loss": 0.17156736552715302, "global_step": 87292, "epoch": 980, "lr": 7.674776107735162e-05}
{"train_loss": 0.24688085913658142, "global_step": 87293, "epoch": 980, "lr": 7.674727126745327e-05}
{"train_loss": 0.17843590676784515, "global_step": 87294, "epoch": 980, "lr": 7.674678145395906e-05}
{"train_loss": 0.20560261607170105, "global_step": 87295, "epoch": 980, "lr": 7.674629163686907e-05}
{"train_loss": 0.15100865066051483, "global_step": 87296, "epoch": 980, "lr": 7.674580181618337e-05}
{"train_loss": 0.17525991797447205, "global_step": 87297, "epoch": 980, "lr": 7.674531199190202e-05}
{"train_loss": 0.1896485686302185, "global_step": 87298, "epoch": 980, "lr": 7.67448221640251e-05}
{"train_loss": 0.1938738226890564, "global_step": 87299, "epoch": 980, "lr": 7.674433233255265e-05}
{"train_loss": 0.15784777700901031, "global_step": 87300, "epoch": 980, "lr": 7.674384249748476e-05}
{"train_loss": 0.18497124314308167, "global_step": 87301, "epoch": 980, "lr": 7.674335265882149e-05}
{"train_loss": 0.24144354462623596, "global_step": 87302, "epoch": 980, "lr": 7.674286281656288e-05}
{"train_loss": 0.13608035445213318, "global_step": 87303, "epoch": 980, "lr": 7.674237297070905e-05}
{"train_loss": 0.1659710556268692, "global_step": 87304, "epoch": 980, "lr": 7.674188312126002e-05}
{"train_loss": 0.22489964962005615, "global_step": 87305, "epoch": 980, "lr": 7.674139326821586e-05}
{"train_loss": 0.1533021479845047, "global_step": 87306, "epoch": 980, "lr": 7.674090341157665e-05}
{"train_loss": 0.1852812021970749, "global_step": 87307, "epoch": 980, "lr": 7.674041355134248e-05}
{"train_loss": 0.18430022133535215, "global_step": 87308, "epoch": 980, "lr": 7.673992368751334e-05, "val_loss": 4.382376194000244, "train_action_mse_error": 14.10598373413086}
{"train_loss": 0.22901137173175812, "global_step": 87309, "epoch": 981, "lr": 7.673943382008937e-05}
{"train_loss": 0.16792374849319458, "global_step": 87310, "epoch": 981, "lr": 7.673894394907061e-05}
{"train_loss": 0.16632229089736938, "global_step": 87311, "epoch": 981, "lr": 7.673845407445712e-05}
{"train_loss": 0.18102657794952393, "global_step": 87312, "epoch": 981, "lr": 7.673796419624897e-05}
{"train_loss": 0.21108609437942505, "global_step": 87313, "epoch": 981, "lr": 7.673747431444622e-05}
{"train_loss": 0.20030778646469116, "global_step": 87314, "epoch": 981, "lr": 7.673698442904896e-05}
{"train_loss": 0.09467893838882446, "global_step": 87315, "epoch": 981, "lr": 7.673649454005723e-05}
{"train_loss": 0.1879124641418457, "global_step": 87316, "epoch": 981, "lr": 7.67360046474711e-05}
{"train_loss": 0.2076234668493271, "global_step": 87317, "epoch": 981, "lr": 7.673551475129064e-05}
{"train_loss": 0.17481288313865662, "global_step": 87318, "epoch": 981, "lr": 7.673502485151591e-05}
{"train_loss": 0.12437291443347931, "global_step": 87319, "epoch": 981, "lr": 7.673453494814699e-05}
{"train_loss": 0.2236669957637787, "global_step": 87320, "epoch": 981, "lr": 7.673404504118393e-05}
{"train_loss": 0.22153744101524353, "global_step": 87321, "epoch": 981, "lr": 7.673355513062682e-05}
{"train_loss": 0.1863984614610672, "global_step": 87322, "epoch": 981, "lr": 7.673306521647569e-05}
{"train_loss": 0.18685901165008545, "global_step": 87323, "epoch": 981, "lr": 7.673257529873065e-05}
{"train_loss": 0.12974345684051514, "global_step": 87324, "epoch": 981, "lr": 7.673208537739172e-05}
{"train_loss": 0.2295103818178177, "global_step": 87325, "epoch": 981, "lr": 7.6731595452459e-05}
{"train_loss": 0.210499569773674, "global_step": 87326, "epoch": 981, "lr": 7.673110552393254e-05}
{"train_loss": 0.15371128916740417, "global_step": 87327, "epoch": 981, "lr": 7.67306155918124e-05}
{"train_loss": 0.15584327280521393, "global_step": 87328, "epoch": 981, "lr": 7.673012565609865e-05}
{"train_loss": 0.18666934967041016, "global_step": 87329, "epoch": 981, "lr": 7.672963571679137e-05}
{"train_loss": 0.2012322098016739, "global_step": 87330, "epoch": 981, "lr": 7.672914577389062e-05}
{"train_loss": 0.12343145906925201, "global_step": 87331, "epoch": 981, "lr": 7.672865582739646e-05}
{"train_loss": 0.15270155668258667, "global_step": 87332, "epoch": 981, "lr": 7.672816587730896e-05}
{"train_loss": 0.1802651286125183, "global_step": 87333, "epoch": 981, "lr": 7.672767592362817e-05}
{"train_loss": 0.10060776770114899, "global_step": 87334, "epoch": 981, "lr": 7.672718596635419e-05}
{"train_loss": 0.10673549026250839, "global_step": 87335, "epoch": 981, "lr": 7.672669600548705e-05}
{"train_loss": 0.1664782017469406, "global_step": 87336, "epoch": 981, "lr": 7.672620604102683e-05}
{"train_loss": 0.19115692377090454, "global_step": 87337, "epoch": 981, "lr": 7.67257160729736e-05}
{"train_loss": 0.21775339543819427, "global_step": 87338, "epoch": 981, "lr": 7.672522610132745e-05}
{"train_loss": 0.15010981261730194, "global_step": 87339, "epoch": 981, "lr": 7.67247361260884e-05}
{"train_loss": 0.22952330112457275, "global_step": 87340, "epoch": 981, "lr": 7.672424614725651e-05}
{"train_loss": 0.2726742923259735, "global_step": 87341, "epoch": 981, "lr": 7.67237561648319e-05}
{"train_loss": 0.16996628046035767, "global_step": 87342, "epoch": 981, "lr": 7.67232661788146e-05}
{"train_loss": 0.21558384597301483, "global_step": 87343, "epoch": 981, "lr": 7.672277618920468e-05}
{"train_loss": 0.19016483426094055, "global_step": 87344, "epoch": 981, "lr": 7.672228619600222e-05}
{"train_loss": 0.14115947484970093, "global_step": 87345, "epoch": 981, "lr": 7.672179619920726e-05}
{"train_loss": 0.26344209909439087, "global_step": 87346, "epoch": 981, "lr": 7.672130619881988e-05}
{"train_loss": 0.11489900201559067, "global_step": 87347, "epoch": 981, "lr": 7.672081619484015e-05}
{"train_loss": 0.1606072634458542, "global_step": 87348, "epoch": 981, "lr": 7.672032618726814e-05}
{"train_loss": 0.2143879532814026, "global_step": 87349, "epoch": 981, "lr": 7.67198361761039e-05}
{"train_loss": 0.2025192826986313, "global_step": 87350, "epoch": 981, "lr": 7.671934616134749e-05}
{"train_loss": 0.2023879736661911, "global_step": 87351, "epoch": 981, "lr": 7.6718856142999e-05}
{"train_loss": 0.16941550374031067, "global_step": 87352, "epoch": 981, "lr": 7.67183661210585e-05}
{"train_loss": 0.2170688360929489, "global_step": 87353, "epoch": 981, "lr": 7.671787609552603e-05}
{"train_loss": 0.10094703733921051, "global_step": 87354, "epoch": 981, "lr": 7.671738606640167e-05}
{"train_loss": 0.16234657168388367, "global_step": 87355, "epoch": 981, "lr": 7.671689603368548e-05}
{"train_loss": 0.25552770495414734, "global_step": 87356, "epoch": 981, "lr": 7.671640599737752e-05}
{"train_loss": 0.19021156430244446, "global_step": 87357, "epoch": 981, "lr": 7.671591595747787e-05}
{"train_loss": 0.12729144096374512, "global_step": 87358, "epoch": 981, "lr": 7.67154259139866e-05}
{"train_loss": 0.1882815957069397, "global_step": 87359, "epoch": 981, "lr": 7.671493586690377e-05}
{"train_loss": 0.267762154340744, "global_step": 87360, "epoch": 981, "lr": 7.671444581622943e-05}
{"train_loss": 0.1730247586965561, "global_step": 87361, "epoch": 981, "lr": 7.671395576196367e-05}
{"train_loss": 0.2163912057876587, "global_step": 87362, "epoch": 981, "lr": 7.671346570410654e-05}
{"train_loss": 0.2086292803287506, "global_step": 87363, "epoch": 981, "lr": 7.671297564265809e-05}
{"train_loss": 0.15227290987968445, "global_step": 87364, "epoch": 981, "lr": 7.671248557761843e-05}
{"train_loss": 0.30397772789001465, "global_step": 87365, "epoch": 981, "lr": 7.67119955089876e-05}
{"train_loss": 0.2645375430583954, "global_step": 87366, "epoch": 981, "lr": 7.671150543676566e-05}
{"train_loss": 0.14895497262477875, "global_step": 87367, "epoch": 981, "lr": 7.671101536095268e-05}
{"train_loss": 0.19739685952663422, "global_step": 87368, "epoch": 981, "lr": 7.671052528154874e-05}
{"train_loss": 0.24258296191692352, "global_step": 87369, "epoch": 981, "lr": 7.671003519855389e-05}
{"train_loss": 0.12650661170482635, "global_step": 87370, "epoch": 981, "lr": 7.67095451119682e-05}
{"train_loss": 0.17561458051204681, "global_step": 87371, "epoch": 981, "lr": 7.670905502179174e-05}
{"train_loss": 0.22402280569076538, "global_step": 87372, "epoch": 981, "lr": 7.670856492802458e-05}
{"train_loss": 0.20151561498641968, "global_step": 87373, "epoch": 981, "lr": 7.670807483066678e-05}
{"train_loss": 0.12000555545091629, "global_step": 87374, "epoch": 981, "lr": 7.670758472971839e-05}
{"train_loss": 0.18669982254505157, "global_step": 87375, "epoch": 981, "lr": 7.670709462517949e-05}
{"train_loss": 0.21218135952949524, "global_step": 87376, "epoch": 981, "lr": 7.670660451705017e-05}
{"train_loss": 0.14754916727542877, "global_step": 87377, "epoch": 981, "lr": 7.670611440533044e-05}
{"train_loss": 0.17803369462490082, "global_step": 87378, "epoch": 981, "lr": 7.670562429002041e-05}
{"train_loss": 0.1360667496919632, "global_step": 87379, "epoch": 981, "lr": 7.670513417112016e-05}
{"train_loss": 0.2606862783432007, "global_step": 87380, "epoch": 981, "lr": 7.67046440486297e-05}
{"train_loss": 0.19808585941791534, "global_step": 87381, "epoch": 981, "lr": 7.670415392254912e-05}
{"train_loss": 0.21727533638477325, "global_step": 87382, "epoch": 981, "lr": 7.670366379287852e-05}
{"train_loss": 0.165358766913414, "global_step": 87383, "epoch": 981, "lr": 7.670317365961792e-05}
{"train_loss": 0.25856220722198486, "global_step": 87384, "epoch": 981, "lr": 7.670268352276741e-05}
{"train_loss": 0.1614806056022644, "global_step": 87385, "epoch": 981, "lr": 7.670219338232706e-05}
{"train_loss": 0.15080894529819489, "global_step": 87386, "epoch": 981, "lr": 7.670170323829691e-05}
{"train_loss": 0.1748405247926712, "global_step": 87387, "epoch": 981, "lr": 7.670121309067704e-05}
{"train_loss": 0.16017358005046844, "global_step": 87388, "epoch": 981, "lr": 7.670072293946753e-05}
{"train_loss": 0.19386236369609833, "global_step": 87389, "epoch": 981, "lr": 7.670023278466842e-05}
{"train_loss": 0.17900030314922333, "global_step": 87390, "epoch": 981, "lr": 7.66997426262798e-05}
{"train_loss": 0.16190402209758759, "global_step": 87391, "epoch": 981, "lr": 7.669925246430172e-05}
{"train_loss": 0.17675825953483582, "global_step": 87392, "epoch": 981, "lr": 7.669876229873424e-05}
{"train_loss": 0.18191489577293396, "global_step": 87393, "epoch": 981, "lr": 7.669827212957746e-05}
{"train_loss": 0.1414535641670227, "global_step": 87394, "epoch": 981, "lr": 7.669778195683141e-05}
{"train_loss": 0.22847606241703033, "global_step": 87395, "epoch": 981, "lr": 7.669729178049617e-05}
{"train_loss": 0.18434740602970123, "global_step": 87396, "epoch": 981, "lr": 7.669680160057181e-05}
{"train_loss": 0.184500843872515, "global_step": 87397, "epoch": 981, "lr": 7.669631141705837e-05, "val_loss": 4.287691593170166}
{"train_loss": 0.12673647701740265, "global_step": 87398, "epoch": 982, "lr": 7.669582122995597e-05}
{"train_loss": 0.23531128466129303, "global_step": 87399, "epoch": 982, "lr": 7.669533103926461e-05}
{"train_loss": 0.22519098222255707, "global_step": 87400, "epoch": 982, "lr": 7.669484084498441e-05}
{"train_loss": 0.18079641461372375, "global_step": 87401, "epoch": 982, "lr": 7.669435064711541e-05}
{"train_loss": 0.1380147784948349, "global_step": 87402, "epoch": 982, "lr": 7.669386044565767e-05}
{"train_loss": 0.23848626017570496, "global_step": 87403, "epoch": 982, "lr": 7.669337024061127e-05}
{"train_loss": 0.20803789794445038, "global_step": 87404, "epoch": 982, "lr": 7.669288003197628e-05}
{"train_loss": 0.24965503811836243, "global_step": 87405, "epoch": 982, "lr": 7.669238981975275e-05}
{"train_loss": 0.22411881387233734, "global_step": 87406, "epoch": 982, "lr": 7.669189960394074e-05}
{"train_loss": 0.16758377850055695, "global_step": 87407, "epoch": 982, "lr": 7.669140938454036e-05}
{"train_loss": 0.17197124660015106, "global_step": 87408, "epoch": 982, "lr": 7.669091916155163e-05}
{"train_loss": 0.22195039689540863, "global_step": 87409, "epoch": 982, "lr": 7.669042893497463e-05}
{"train_loss": 0.17351140081882477, "global_step": 87410, "epoch": 982, "lr": 7.668993870480941e-05}
{"train_loss": 0.2513434588909149, "global_step": 87411, "epoch": 982, "lr": 7.668944847105608e-05}
{"train_loss": 0.13127762079238892, "global_step": 87412, "epoch": 982, "lr": 7.668895823371467e-05}
{"train_loss": 0.13668960332870483, "global_step": 87413, "epoch": 982, "lr": 7.668846799278525e-05}
{"train_loss": 0.1521710306406021, "global_step": 87414, "epoch": 982, "lr": 7.668797774826788e-05}
{"train_loss": 0.20853473246097565, "global_step": 87415, "epoch": 982, "lr": 7.668748750016265e-05}
{"train_loss": 0.2325173318386078, "global_step": 87416, "epoch": 982, "lr": 7.668699724846964e-05}
{"train_loss": 0.12197309732437134, "global_step": 87417, "epoch": 982, "lr": 7.668650699318885e-05}
{"train_loss": 0.22597259283065796, "global_step": 87418, "epoch": 982, "lr": 7.66860167343204e-05}
{"train_loss": 0.15202638506889343, "global_step": 87419, "epoch": 982, "lr": 7.668552647186433e-05}
{"train_loss": 0.14273639023303986, "global_step": 87420, "epoch": 982, "lr": 7.668503620582073e-05}
{"train_loss": 0.1777002513408661, "global_step": 87421, "epoch": 982, "lr": 7.668454593618965e-05}
{"train_loss": 0.18035538494586945, "global_step": 87422, "epoch": 982, "lr": 7.668405566297114e-05}
{"train_loss": 0.20903967320919037, "global_step": 87423, "epoch": 982, "lr": 7.66835653861653e-05}
{"train_loss": 0.18894824385643005, "global_step": 87424, "epoch": 982, "lr": 7.668307510577218e-05}
{"train_loss": 0.1951177418231964, "global_step": 87425, "epoch": 982, "lr": 7.668258482179185e-05}
{"train_loss": 0.14123283326625824, "global_step": 87426, "epoch": 982, "lr": 7.668209453422437e-05}
{"train_loss": 0.2587588429450989, "global_step": 87427, "epoch": 982, "lr": 7.66816042430698e-05}
{"train_loss": 0.12852202355861664, "global_step": 87428, "epoch": 982, "lr": 7.668111394832821e-05}
{"train_loss": 0.24723222851753235, "global_step": 87429, "epoch": 982, "lr": 7.668062364999969e-05}
{"train_loss": 0.1552312970161438, "global_step": 87430, "epoch": 982, "lr": 7.668013334808426e-05}
{"train_loss": 0.2534498870372772, "global_step": 87431, "epoch": 982, "lr": 7.667964304258202e-05}
{"train_loss": 0.19926784932613373, "global_step": 87432, "epoch": 982, "lr": 7.667915273349302e-05}
{"train_loss": 0.2258244901895523, "global_step": 87433, "epoch": 982, "lr": 7.667866242081734e-05}
{"train_loss": 0.20950058102607727, "global_step": 87434, "epoch": 982, "lr": 7.667817210455504e-05}
{"train_loss": 0.184220090508461, "global_step": 87435, "epoch": 982, "lr": 7.66776817847062e-05}
{"train_loss": 0.21730417013168335, "global_step": 87436, "epoch": 982, "lr": 7.667719146127085e-05}
{"train_loss": 0.19291752576828003, "global_step": 87437, "epoch": 982, "lr": 7.66767011342491e-05}
{"train_loss": 0.22872696816921234, "global_step": 87438, "epoch": 982, "lr": 7.667621080364096e-05}
{"train_loss": 0.17292940616607666, "global_step": 87439, "epoch": 982, "lr": 7.667572046944656e-05}
{"train_loss": 0.2158493548631668, "global_step": 87440, "epoch": 982, "lr": 7.667523013166591e-05}
{"train_loss": 0.1317259818315506, "global_step": 87441, "epoch": 982, "lr": 7.667473979029911e-05}
{"train_loss": 0.15800777077674866, "global_step": 87442, "epoch": 982, "lr": 7.667424944534622e-05}
{"train_loss": 0.229136660695076, "global_step": 87443, "epoch": 982, "lr": 7.667375909680729e-05}
{"train_loss": 0.2736786901950836, "global_step": 87444, "epoch": 982, "lr": 7.667326874468241e-05}
{"train_loss": 0.1597670465707779, "global_step": 87445, "epoch": 982, "lr": 7.667277838897163e-05}
{"train_loss": 0.19375357031822205, "global_step": 87446, "epoch": 982, "lr": 7.667228802967502e-05}
{"train_loss": 0.10749544948339462, "global_step": 87447, "epoch": 982, "lr": 7.667179766679264e-05}
{"train_loss": 0.21479487419128418, "global_step": 87448, "epoch": 982, "lr": 7.667130730032458e-05}
{"train_loss": 0.1537453681230545, "global_step": 87449, "epoch": 982, "lr": 7.667081693027088e-05}
{"train_loss": 0.15506385266780853, "global_step": 87450, "epoch": 982, "lr": 7.66703265566316e-05}
{"train_loss": 0.21615692973136902, "global_step": 87451, "epoch": 982, "lr": 7.666983617940682e-05}
{"train_loss": 0.16992738842964172, "global_step": 87452, "epoch": 982, "lr": 7.666934579859662e-05}
{"train_loss": 0.2238086462020874, "global_step": 87453, "epoch": 982, "lr": 7.666885541420105e-05}
{"train_loss": 0.22208920121192932, "global_step": 87454, "epoch": 982, "lr": 7.666836502622017e-05}
{"train_loss": 0.22060392796993256, "global_step": 87455, "epoch": 982, "lr": 7.666787463465407e-05}
{"train_loss": 0.18474361300468445, "global_step": 87456, "epoch": 982, "lr": 7.66673842395028e-05}
{"train_loss": 0.14626255631446838, "global_step": 87457, "epoch": 982, "lr": 7.666689384076639e-05}
{"train_loss": 0.1464518904685974, "global_step": 87458, "epoch": 982, "lr": 7.666640343844498e-05}
{"train_loss": 0.18592287600040436, "global_step": 87459, "epoch": 982, "lr": 7.666591303253858e-05}
{"train_loss": 0.17965535819530487, "global_step": 87460, "epoch": 982, "lr": 7.666542262304726e-05}
{"train_loss": 0.20743435621261597, "global_step": 87461, "epoch": 982, "lr": 7.666493220997113e-05}
{"train_loss": 0.1522432416677475, "global_step": 87462, "epoch": 982, "lr": 7.666444179331021e-05}
{"train_loss": 0.16857968270778656, "global_step": 87463, "epoch": 982, "lr": 7.666395137306456e-05}
{"train_loss": 0.19062180817127228, "global_step": 87464, "epoch": 982, "lr": 7.666346094923429e-05}
{"train_loss": 0.16584278643131256, "global_step": 87465, "epoch": 982, "lr": 7.666297052181944e-05}
{"train_loss": 0.1699438840150833, "global_step": 87466, "epoch": 982, "lr": 7.666248009082008e-05}
{"train_loss": 0.17813047766685486, "global_step": 87467, "epoch": 982, "lr": 7.666198965623626e-05}
{"train_loss": 0.14942790567874908, "global_step": 87468, "epoch": 982, "lr": 7.666149921806808e-05}
{"train_loss": 0.2380588799715042, "global_step": 87469, "epoch": 982, "lr": 7.666100877631557e-05}
{"train_loss": 0.12280558049678802, "global_step": 87470, "epoch": 982, "lr": 7.666051833097882e-05}
{"train_loss": 0.22095277905464172, "global_step": 87471, "epoch": 982, "lr": 7.666002788205788e-05}
{"train_loss": 0.1643909215927124, "global_step": 87472, "epoch": 982, "lr": 7.665953742955283e-05}
{"train_loss": 0.2085905373096466, "global_step": 87473, "epoch": 982, "lr": 7.665904697346373e-05}
{"train_loss": 0.25220435857772827, "global_step": 87474, "epoch": 982, "lr": 7.665855651379067e-05}
{"train_loss": 0.27486860752105713, "global_step": 87475, "epoch": 982, "lr": 7.665806605053368e-05}
{"train_loss": 0.22381974756717682, "global_step": 87476, "epoch": 982, "lr": 7.665757558369282e-05}
{"train_loss": 0.13289599120616913, "global_step": 87477, "epoch": 982, "lr": 7.665708511326818e-05}
{"train_loss": 0.19304488599300385, "global_step": 87478, "epoch": 982, "lr": 7.665659463925982e-05}
{"train_loss": 0.20776070654392242, "global_step": 87479, "epoch": 982, "lr": 7.665610416166781e-05}
{"train_loss": 0.17700642347335815, "global_step": 87480, "epoch": 982, "lr": 7.665561368049221e-05}
{"train_loss": 0.18117576837539673, "global_step": 87481, "epoch": 982, "lr": 7.665512319573309e-05}
{"train_loss": 0.3003422021865845, "global_step": 87482, "epoch": 982, "lr": 7.665463270739051e-05}
{"train_loss": 0.17798441648483276, "global_step": 87483, "epoch": 982, "lr": 7.665414221546454e-05}
{"train_loss": 0.23417522013187408, "global_step": 87484, "epoch": 982, "lr": 7.665365171995524e-05}
{"train_loss": 0.19404488801956177, "global_step": 87485, "epoch": 982, "lr": 7.66531612208627e-05}
{"train_loss": 0.19128700094611456, "global_step": 87486, "epoch": 982, "lr": 7.665267071818693e-05, "val_loss": 4.358687877655029}
{"train_loss": 0.19417926669120789, "global_step": 87487, "epoch": 983, "lr": 7.665218021192806e-05}
{"train_loss": 0.19625568389892578, "global_step": 87488, "epoch": 983, "lr": 7.665168970208613e-05}
{"train_loss": 0.25680485367774963, "global_step": 87489, "epoch": 983, "lr": 7.665119918866119e-05}
{"train_loss": 0.15113204717636108, "global_step": 87490, "epoch": 983, "lr": 7.665070867165334e-05}
{"train_loss": 0.20435731112957, "global_step": 87491, "epoch": 983, "lr": 7.665021815106262e-05}
{"train_loss": 0.17093013226985931, "global_step": 87492, "epoch": 983, "lr": 7.664972762688911e-05}
{"train_loss": 0.14719036221504211, "global_step": 87493, "epoch": 983, "lr": 7.664923709913288e-05}
{"train_loss": 0.18299414217472076, "global_step": 87494, "epoch": 983, "lr": 7.664874656779396e-05}
{"train_loss": 0.15883632004261017, "global_step": 87495, "epoch": 983, "lr": 7.664825603287245e-05}
{"train_loss": 0.20860636234283447, "global_step": 87496, "epoch": 983, "lr": 7.66477654943684e-05}
{"train_loss": 0.20063500106334686, "global_step": 87497, "epoch": 983, "lr": 7.664727495228189e-05}
{"train_loss": 0.2318210005760193, "global_step": 87498, "epoch": 983, "lr": 7.664678440661297e-05}
{"train_loss": 0.26650893688201904, "global_step": 87499, "epoch": 983, "lr": 7.664629385736173e-05}
{"train_loss": 0.146407350897789, "global_step": 87500, "epoch": 983, "lr": 7.66458033045282e-05}
{"train_loss": 0.17647531628608704, "global_step": 87501, "epoch": 983, "lr": 7.664531274811249e-05}
{"train_loss": 0.2021610587835312, "global_step": 87502, "epoch": 983, "lr": 7.664482218811462e-05}
{"train_loss": 0.21197383105754852, "global_step": 87503, "epoch": 983, "lr": 7.66443316245347e-05}
{"train_loss": 0.20636415481567383, "global_step": 87504, "epoch": 983, "lr": 7.664384105737276e-05}
{"train_loss": 0.23417821526527405, "global_step": 87505, "epoch": 983, "lr": 7.664335048662887e-05}
{"train_loss": 0.21149282157421112, "global_step": 87506, "epoch": 983, "lr": 7.664285991230311e-05}
{"train_loss": 0.14597757160663605, "global_step": 87507, "epoch": 983, "lr": 7.664236933439556e-05}
{"train_loss": 0.1458314061164856, "global_step": 87508, "epoch": 983, "lr": 7.664187875290626e-05}
{"train_loss": 0.2130294144153595, "global_step": 87509, "epoch": 983, "lr": 7.664138816783528e-05}
{"train_loss": 0.17395688593387604, "global_step": 87510, "epoch": 983, "lr": 7.664089757918269e-05}
{"train_loss": 0.24349163472652435, "global_step": 87511, "epoch": 983, "lr": 7.664040698694856e-05}
{"train_loss": 0.22634997963905334, "global_step": 87512, "epoch": 983, "lr": 7.663991639113296e-05}
{"train_loss": 0.2481938600540161, "global_step": 87513, "epoch": 983, "lr": 7.663942579173596e-05}
{"train_loss": 0.14937463402748108, "global_step": 87514, "epoch": 983, "lr": 7.663893518875758e-05}
{"train_loss": 0.25336313247680664, "global_step": 87515, "epoch": 983, "lr": 7.663844458219793e-05}
{"train_loss": 0.20715409517288208, "global_step": 87516, "epoch": 983, "lr": 7.663795397205708e-05}
{"train_loss": 0.17434848845005035, "global_step": 87517, "epoch": 983, "lr": 7.663746335833508e-05}
{"train_loss": 0.12020484358072281, "global_step": 87518, "epoch": 983, "lr": 7.663697274103197e-05}
{"train_loss": 0.12154383957386017, "global_step": 87519, "epoch": 983, "lr": 7.663648212014787e-05}
{"train_loss": 0.13512514531612396, "global_step": 87520, "epoch": 983, "lr": 7.663599149568281e-05}
{"train_loss": 0.17244522273540497, "global_step": 87521, "epoch": 983, "lr": 7.663550086763686e-05}
{"train_loss": 0.18170632421970367, "global_step": 87522, "epoch": 983, "lr": 7.663501023601012e-05}
{"train_loss": 0.277948796749115, "global_step": 87523, "epoch": 983, "lr": 7.66345196008026e-05}
{"train_loss": 0.19796234369277954, "global_step": 87524, "epoch": 983, "lr": 7.66340289620144e-05}
{"train_loss": 0.1711016595363617, "global_step": 87525, "epoch": 983, "lr": 7.663353831964557e-05}
{"train_loss": 0.124128058552742, "global_step": 87526, "epoch": 983, "lr": 7.663304767369619e-05}
{"train_loss": 0.19846481084823608, "global_step": 87527, "epoch": 983, "lr": 7.663255702416633e-05}
{"train_loss": 0.3120819926261902, "global_step": 87528, "epoch": 983, "lr": 7.663206637105604e-05}
{"train_loss": 0.1625235229730606, "global_step": 87529, "epoch": 983, "lr": 7.66315757143654e-05}
{"train_loss": 0.19585907459259033, "global_step": 87530, "epoch": 983, "lr": 7.663108505409447e-05}
{"train_loss": 0.1376398652791977, "global_step": 87531, "epoch": 983, "lr": 7.663059439024331e-05}
{"train_loss": 0.1717185229063034, "global_step": 87532, "epoch": 983, "lr": 7.663010372281201e-05}
{"train_loss": 0.17817668616771698, "global_step": 87533, "epoch": 983, "lr": 7.662961305180058e-05}
{"train_loss": 0.19400563836097717, "global_step": 87534, "epoch": 983, "lr": 7.662912237720916e-05}
{"train_loss": 0.12178696691989899, "global_step": 87535, "epoch": 983, "lr": 7.662863169903775e-05}
{"train_loss": 0.16493305563926697, "global_step": 87536, "epoch": 983, "lr": 7.662814101728646e-05}
{"train_loss": 0.174992173910141, "global_step": 87537, "epoch": 983, "lr": 7.662765033195533e-05}
{"train_loss": 0.2271326184272766, "global_step": 87538, "epoch": 983, "lr": 7.662715964304445e-05}
{"train_loss": 0.21616007387638092, "global_step": 87539, "epoch": 983, "lr": 7.662666895055388e-05}
{"train_loss": 0.13821646571159363, "global_step": 87540, "epoch": 983, "lr": 7.662617825448366e-05}
{"train_loss": 0.22005441784858704, "global_step": 87541, "epoch": 983, "lr": 7.662568755483389e-05}
{"train_loss": 0.14583583176136017, "global_step": 87542, "epoch": 983, "lr": 7.66251968516046e-05}
{"train_loss": 0.262549489736557, "global_step": 87543, "epoch": 983, "lr": 7.662470614479589e-05}
{"train_loss": 0.20655372738838196, "global_step": 87544, "epoch": 983, "lr": 7.662421543440781e-05}
{"train_loss": 0.2751806974411011, "global_step": 87545, "epoch": 983, "lr": 7.662372472044044e-05}
{"train_loss": 0.17751045525074005, "global_step": 87546, "epoch": 983, "lr": 7.662323400289382e-05}
{"train_loss": 0.14182248711585999, "global_step": 87547, "epoch": 983, "lr": 7.662274328176804e-05}
{"train_loss": 0.1840491145849228, "global_step": 87548, "epoch": 983, "lr": 7.662225255706314e-05}
{"train_loss": 0.16944867372512817, "global_step": 87549, "epoch": 983, "lr": 7.662176182877923e-05}
{"train_loss": 0.20615898072719574, "global_step": 87550, "epoch": 983, "lr": 7.662127109691633e-05}
{"train_loss": 0.22765985131263733, "global_step": 87551, "epoch": 983, "lr": 7.662078036147453e-05}
{"train_loss": 0.13495957851409912, "global_step": 87552, "epoch": 983, "lr": 7.662028962245389e-05}
{"train_loss": 0.23057733476161957, "global_step": 87553, "epoch": 983, "lr": 7.661979887985449e-05}
{"train_loss": 0.12997393310070038, "global_step": 87554, "epoch": 983, "lr": 7.661930813367636e-05}
{"train_loss": 0.15975321829319, "global_step": 87555, "epoch": 983, "lr": 7.66188173839196e-05}
{"train_loss": 0.20158909261226654, "global_step": 87556, "epoch": 983, "lr": 7.661832663058426e-05}
{"train_loss": 0.1688985973596573, "global_step": 87557, "epoch": 983, "lr": 7.661783587367041e-05}
{"train_loss": 0.1616852730512619, "global_step": 87558, "epoch": 983, "lr": 7.661734511317812e-05}
{"train_loss": 0.16110624372959137, "global_step": 87559, "epoch": 983, "lr": 7.661685434910744e-05}
{"train_loss": 0.22290512919425964, "global_step": 87560, "epoch": 983, "lr": 7.661636358145845e-05}
{"train_loss": 0.21587790548801422, "global_step": 87561, "epoch": 983, "lr": 7.661587281023123e-05}
{"train_loss": 0.1200767457485199, "global_step": 87562, "epoch": 983, "lr": 7.661538203542582e-05}
{"train_loss": 0.14206211268901825, "global_step": 87563, "epoch": 983, "lr": 7.661489125704231e-05}
{"train_loss": 0.14247755706310272, "global_step": 87564, "epoch": 983, "lr": 7.661440047508073e-05}
{"train_loss": 0.11100149154663086, "global_step": 87565, "epoch": 983, "lr": 7.661390968954119e-05}
{"train_loss": 0.17856180667877197, "global_step": 87566, "epoch": 983, "lr": 7.661341890042371e-05}
{"train_loss": 0.19689907133579254, "global_step": 87567, "epoch": 983, "lr": 7.66129281077284e-05}
{"train_loss": 0.28625887632369995, "global_step": 87568, "epoch": 983, "lr": 7.661243731145531e-05}
{"train_loss": 0.15665212273597717, "global_step": 87569, "epoch": 983, "lr": 7.661194651160449e-05}
{"train_loss": 0.1416601687669754, "global_step": 87570, "epoch": 983, "lr": 7.661145570817602e-05}
{"train_loss": 0.19186699390411377, "global_step": 87571, "epoch": 983, "lr": 7.661096490116996e-05}
{"train_loss": 0.19932973384857178, "global_step": 87572, "epoch": 983, "lr": 7.66104740905864e-05}
{"train_loss": 0.17579682171344757, "global_step": 87573, "epoch": 983, "lr": 7.660998327642537e-05}
{"train_loss": 0.22027160227298737, "global_step": 87574, "epoch": 983, "lr": 7.660949245868696e-05}
{"train_loss": 0.18744356653998406, "global_step": 87575, "epoch": 983, "lr": 7.660900163737121e-05, "val_loss": 4.348776817321777}
{"train_loss": 0.19774924218654633, "global_step": 87576, "epoch": 984, "lr": 7.660851081247822e-05}
{"train_loss": 0.16421106457710266, "global_step": 87577, "epoch": 984, "lr": 7.660801998400804e-05}
{"train_loss": 0.2160186767578125, "global_step": 87578, "epoch": 984, "lr": 7.660752915196073e-05}
{"train_loss": 0.1757727563381195, "global_step": 87579, "epoch": 984, "lr": 7.660703831633636e-05}
{"train_loss": 0.2683629095554352, "global_step": 87580, "epoch": 984, "lr": 7.660654747713499e-05}
{"train_loss": 0.2069980949163437, "global_step": 87581, "epoch": 984, "lr": 7.660605663435671e-05}
{"train_loss": 0.179177924990654, "global_step": 87582, "epoch": 984, "lr": 7.660556578800156e-05}
{"train_loss": 0.23387399315834045, "global_step": 87583, "epoch": 984, "lr": 7.660507493806963e-05}
{"train_loss": 0.2193034589290619, "global_step": 87584, "epoch": 984, "lr": 7.660458408456096e-05}
{"train_loss": 0.1868383288383484, "global_step": 87585, "epoch": 984, "lr": 7.660409322747564e-05}
{"train_loss": 0.15345337986946106, "global_step": 87586, "epoch": 984, "lr": 7.66036023668137e-05}
{"train_loss": 0.11921133100986481, "global_step": 87587, "epoch": 984, "lr": 7.660311150257525e-05}
{"train_loss": 0.15932613611221313, "global_step": 87588, "epoch": 984, "lr": 7.660262063476034e-05}
{"train_loss": 0.24904240667819977, "global_step": 87589, "epoch": 984, "lr": 7.660212976336902e-05}
{"train_loss": 0.19996805489063263, "global_step": 87590, "epoch": 984, "lr": 7.660163888840138e-05}
{"train_loss": 0.14371752738952637, "global_step": 87591, "epoch": 984, "lr": 7.660114800985746e-05}
{"train_loss": 0.1248970627784729, "global_step": 87592, "epoch": 984, "lr": 7.660065712773733e-05}
{"train_loss": 0.2195630669593811, "global_step": 87593, "epoch": 984, "lr": 7.66001662420411e-05}
{"train_loss": 0.19220755994319916, "global_step": 87594, "epoch": 984, "lr": 7.659967535276877e-05}
{"train_loss": 0.11314190924167633, "global_step": 87595, "epoch": 984, "lr": 7.659918445992046e-05}
{"train_loss": 0.13482540845870972, "global_step": 87596, "epoch": 984, "lr": 7.65986935634962e-05}
{"train_loss": 0.1765199452638626, "global_step": 87597, "epoch": 984, "lr": 7.659820266349607e-05}
{"train_loss": 0.10470099747180939, "global_step": 87598, "epoch": 984, "lr": 7.659771175992014e-05}
{"train_loss": 0.2475212663412094, "global_step": 87599, "epoch": 984, "lr": 7.659722085276847e-05}
{"train_loss": 0.19616460800170898, "global_step": 87600, "epoch": 984, "lr": 7.659672994204112e-05}
{"train_loss": 0.29371383786201477, "global_step": 87601, "epoch": 984, "lr": 7.659623902773818e-05}
{"train_loss": 0.1579218953847885, "global_step": 87602, "epoch": 984, "lr": 7.65957481098597e-05}
{"train_loss": 0.22644411027431488, "global_step": 87603, "epoch": 984, "lr": 7.659525718840574e-05}
{"train_loss": 0.205875426530838, "global_step": 87604, "epoch": 984, "lr": 7.659476626337636e-05}
{"train_loss": 0.11513518542051315, "global_step": 87605, "epoch": 984, "lr": 7.659427533477166e-05}
{"train_loss": 0.19249039888381958, "global_step": 87606, "epoch": 984, "lr": 7.659378440259166e-05}
{"train_loss": 0.3042012155056, "global_step": 87607, "epoch": 984, "lr": 7.659329346683647e-05}
{"train_loss": 0.17120355367660522, "global_step": 87608, "epoch": 984, "lr": 7.659280252750612e-05}
{"train_loss": 0.14953625202178955, "global_step": 87609, "epoch": 984, "lr": 7.659231158460072e-05}
{"train_loss": 0.22067870199680328, "global_step": 87610, "epoch": 984, "lr": 7.659182063812028e-05}
{"train_loss": 0.1162518784403801, "global_step": 87611, "epoch": 984, "lr": 7.65913296880649e-05}
{"train_loss": 0.18622750043869019, "global_step": 87612, "epoch": 984, "lr": 7.659083873443464e-05}
{"train_loss": 0.1758909374475479, "global_step": 87613, "epoch": 984, "lr": 7.659034777722958e-05}
{"train_loss": 0.18982911109924316, "global_step": 87614, "epoch": 984, "lr": 7.658985681644975e-05}
{"train_loss": 0.17713488638401031, "global_step": 87615, "epoch": 984, "lr": 7.658936585209524e-05}
{"train_loss": 0.1916530281305313, "global_step": 87616, "epoch": 984, "lr": 7.658887488416612e-05}
{"train_loss": 0.13962292671203613, "global_step": 87617, "epoch": 984, "lr": 7.658838391266245e-05}
{"train_loss": 0.1418115794658661, "global_step": 87618, "epoch": 984, "lr": 7.658789293758431e-05}
{"train_loss": 0.17358949780464172, "global_step": 87619, "epoch": 984, "lr": 7.658740195893172e-05}
{"train_loss": 0.18788214027881622, "global_step": 87620, "epoch": 984, "lr": 7.658691097670481e-05}
{"train_loss": 0.14141739904880524, "global_step": 87621, "epoch": 984, "lr": 7.658641999090359e-05}
{"train_loss": 0.20543716847896576, "global_step": 87622, "epoch": 984, "lr": 7.658592900152817e-05}
{"train_loss": 0.22846993803977966, "global_step": 87623, "epoch": 984, "lr": 7.658543800857858e-05}
{"train_loss": 0.1385611742734909, "global_step": 87624, "epoch": 984, "lr": 7.658494701205491e-05}
{"train_loss": 0.28539878129959106, "global_step": 87625, "epoch": 984, "lr": 7.658445601195721e-05}
{"train_loss": 0.16612565517425537, "global_step": 87626, "epoch": 984, "lr": 7.658396500828557e-05}
{"train_loss": 0.21437306702136993, "global_step": 87627, "epoch": 984, "lr": 7.658347400104002e-05}
{"train_loss": 0.17711497843265533, "global_step": 87628, "epoch": 984, "lr": 7.658298299022067e-05}
{"train_loss": 0.16565506160259247, "global_step": 87629, "epoch": 984, "lr": 7.658249197582754e-05}
{"train_loss": 0.1451573669910431, "global_step": 87630, "epoch": 984, "lr": 7.658200095786072e-05}
{"train_loss": 0.15828709304332733, "global_step": 87631, "epoch": 984, "lr": 7.658150993632029e-05}
{"train_loss": 0.29257088899612427, "global_step": 87632, "epoch": 984, "lr": 7.658101891120628e-05}
{"train_loss": 0.1337171047925949, "global_step": 87633, "epoch": 984, "lr": 7.658052788251879e-05}
{"train_loss": 0.1527322679758072, "global_step": 87634, "epoch": 984, "lr": 7.658003685025785e-05}
{"train_loss": 0.10471833497285843, "global_step": 87635, "epoch": 984, "lr": 7.657954581442357e-05}
{"train_loss": 0.152327299118042, "global_step": 87636, "epoch": 984, "lr": 7.657905477501598e-05}
{"train_loss": 0.19017960131168365, "global_step": 87637, "epoch": 984, "lr": 7.657856373203518e-05}
{"train_loss": 0.23122237622737885, "global_step": 87638, "epoch": 984, "lr": 7.657807268548121e-05}
{"train_loss": 0.1489003747701645, "global_step": 87639, "epoch": 984, "lr": 7.657758163535413e-05}
{"train_loss": 0.20594234764575958, "global_step": 87640, "epoch": 984, "lr": 7.657709058165403e-05}
{"train_loss": 0.17667482793331146, "global_step": 87641, "epoch": 984, "lr": 7.657659952438097e-05}
{"train_loss": 0.15821613371372223, "global_step": 87642, "epoch": 984, "lr": 7.6576108463535e-05}
{"train_loss": 0.20732037723064423, "global_step": 87643, "epoch": 984, "lr": 7.657561739911618e-05}
{"train_loss": 0.22583769261837006, "global_step": 87644, "epoch": 984, "lr": 7.657512633112462e-05}
{"train_loss": 0.12969838082790375, "global_step": 87645, "epoch": 984, "lr": 7.657463525956034e-05}
{"train_loss": 0.1726030707359314, "global_step": 87646, "epoch": 984, "lr": 7.657414418442344e-05}
{"train_loss": 0.21124932169914246, "global_step": 87647, "epoch": 984, "lr": 7.657365310571395e-05}
{"train_loss": 0.28943347930908203, "global_step": 87648, "epoch": 984, "lr": 7.657316202343197e-05}
{"train_loss": 0.19844748079776764, "global_step": 87649, "epoch": 984, "lr": 7.657267093757756e-05}
{"train_loss": 0.16799958050251007, "global_step": 87650, "epoch": 984, "lr": 7.657217984815075e-05}
{"train_loss": 0.20956431329250336, "global_step": 87651, "epoch": 984, "lr": 7.657168875515166e-05}
{"train_loss": 0.13004304468631744, "global_step": 87652, "epoch": 984, "lr": 7.657119765858032e-05}
{"train_loss": 0.2407839596271515, "global_step": 87653, "epoch": 984, "lr": 7.65707065584368e-05}
{"train_loss": 0.3023512363433838, "global_step": 87654, "epoch": 984, "lr": 7.657021545472118e-05}
{"train_loss": 0.13998155295848846, "global_step": 87655, "epoch": 984, "lr": 7.656972434743352e-05}
{"train_loss": 0.14572562277317047, "global_step": 87656, "epoch": 984, "lr": 7.656923323657389e-05}
{"train_loss": 0.17535904049873352, "global_step": 87657, "epoch": 984, "lr": 7.656874212214233e-05}
{"train_loss": 0.18920299410820007, "global_step": 87658, "epoch": 984, "lr": 7.656825100413896e-05}
{"train_loss": 0.16425950825214386, "global_step": 87659, "epoch": 984, "lr": 7.656775988256379e-05}
{"train_loss": 0.1576196849346161, "global_step": 87660, "epoch": 984, "lr": 7.656726875741692e-05}
{"train_loss": 0.21436108648777008, "global_step": 87661, "epoch": 984, "lr": 7.656677762869838e-05}
{"train_loss": 0.22043319046497345, "global_step": 87662, "epoch": 984, "lr": 7.656628649640827e-05}
{"train_loss": 0.16753624379634857, "global_step": 87663, "epoch": 984, "lr": 7.656579536054667e-05}
{"train_loss": 0.18657348154301054, "global_step": 87664, "epoch": 984, "lr": 7.656530422111359e-05, "val_loss": 4.328670978546143}
{"train_loss": 0.17188847064971924, "global_step": 87665, "epoch": 985, "lr": 7.656481307810915e-05}
{"train_loss": 0.1399235725402832, "global_step": 87666, "epoch": 985, "lr": 7.656432193153339e-05}
{"train_loss": 0.13449354469776154, "global_step": 87667, "epoch": 985, "lr": 7.656383078138639e-05}
{"train_loss": 0.18628934025764465, "global_step": 87668, "epoch": 985, "lr": 7.656333962766818e-05}
{"train_loss": 0.1355690211057663, "global_step": 87669, "epoch": 985, "lr": 7.656284847037887e-05}
{"train_loss": 0.20674118399620056, "global_step": 87670, "epoch": 985, "lr": 7.656235730951851e-05}
{"train_loss": 0.18830829858779907, "global_step": 87671, "epoch": 985, "lr": 7.656186614508716e-05}
{"train_loss": 0.2380225658416748, "global_step": 87672, "epoch": 985, "lr": 7.656137497708489e-05}
{"train_loss": 0.16855785250663757, "global_step": 87673, "epoch": 985, "lr": 7.656088380551178e-05}
{"train_loss": 0.16248628497123718, "global_step": 87674, "epoch": 985, "lr": 7.656039263036787e-05}
{"train_loss": 0.11862578988075256, "global_step": 87675, "epoch": 985, "lr": 7.655990145165324e-05}
{"train_loss": 0.22421453893184662, "global_step": 87676, "epoch": 985, "lr": 7.655941026936795e-05}
{"train_loss": 0.17925645411014557, "global_step": 87677, "epoch": 985, "lr": 7.655891908351209e-05}
{"train_loss": 0.14514318108558655, "global_step": 87678, "epoch": 985, "lr": 7.655842789408569e-05}
{"train_loss": 0.2726830840110779, "global_step": 87679, "epoch": 985, "lr": 7.655793670108884e-05}
{"train_loss": 0.23954355716705322, "global_step": 87680, "epoch": 985, "lr": 7.65574455045216e-05}
{"train_loss": 0.14506052434444427, "global_step": 87681, "epoch": 985, "lr": 7.655695430438403e-05}
{"train_loss": 0.22405250370502472, "global_step": 87682, "epoch": 985, "lr": 7.65564631006762e-05}
{"train_loss": 0.18976286053657532, "global_step": 87683, "epoch": 985, "lr": 7.655597189339819e-05}
{"train_loss": 0.20754054188728333, "global_step": 87684, "epoch": 985, "lr": 7.655548068255003e-05}
{"train_loss": 0.24641764163970947, "global_step": 87685, "epoch": 985, "lr": 7.655498946813182e-05}
{"train_loss": 0.2072887122631073, "global_step": 87686, "epoch": 985, "lr": 7.655449825014363e-05}
{"train_loss": 0.15521091222763062, "global_step": 87687, "epoch": 985, "lr": 7.655400702858549e-05}
{"train_loss": 0.24181105196475983, "global_step": 87688, "epoch": 985, "lr": 7.65535158034575e-05}
{"train_loss": 0.08591878414154053, "global_step": 87689, "epoch": 985, "lr": 7.655302457475972e-05}
{"train_loss": 0.206583172082901, "global_step": 87690, "epoch": 985, "lr": 7.65525333424922e-05}
{"train_loss": 0.173532173037529, "global_step": 87691, "epoch": 985, "lr": 7.6552042106655e-05}
{"train_loss": 0.22172822058200836, "global_step": 87692, "epoch": 985, "lr": 7.655155086724824e-05}
{"train_loss": 0.23691248893737793, "global_step": 87693, "epoch": 985, "lr": 7.655105962427192e-05}
{"train_loss": 0.13439781963825226, "global_step": 87694, "epoch": 985, "lr": 7.655056837772614e-05}
{"train_loss": 0.24090959131717682, "global_step": 87695, "epoch": 985, "lr": 7.655007712761096e-05}
{"train_loss": 0.18097254633903503, "global_step": 87696, "epoch": 985, "lr": 7.654958587392643e-05}
{"train_loss": 0.19487029314041138, "global_step": 87697, "epoch": 985, "lr": 7.654909461667265e-05}
{"train_loss": 0.14277398586273193, "global_step": 87698, "epoch": 985, "lr": 7.654860335584966e-05}
{"train_loss": 0.1929088681936264, "global_step": 87699, "epoch": 985, "lr": 7.654811209145753e-05}
{"train_loss": 0.24533098936080933, "global_step": 87700, "epoch": 985, "lr": 7.654762082349634e-05}
{"train_loss": 0.19398599863052368, "global_step": 87701, "epoch": 985, "lr": 7.654712955196614e-05}
{"train_loss": 0.1060117781162262, "global_step": 87702, "epoch": 985, "lr": 7.654663827686701e-05}
{"train_loss": 0.2247975766658783, "global_step": 87703, "epoch": 985, "lr": 7.654614699819901e-05}
{"train_loss": 0.16225767135620117, "global_step": 87704, "epoch": 985, "lr": 7.654565571596219e-05}
{"train_loss": 0.19234994053840637, "global_step": 87705, "epoch": 985, "lr": 7.654516443015664e-05}
{"train_loss": 0.19858276844024658, "global_step": 87706, "epoch": 985, "lr": 7.654467314078242e-05}
{"train_loss": 0.26027899980545044, "global_step": 87707, "epoch": 985, "lr": 7.654418184783958e-05}
{"train_loss": 0.24432161450386047, "global_step": 87708, "epoch": 985, "lr": 7.654369055132821e-05}
{"train_loss": 0.14881351590156555, "global_step": 87709, "epoch": 985, "lr": 7.654319925124836e-05}
{"train_loss": 0.27000123262405396, "global_step": 87710, "epoch": 985, "lr": 7.65427079476001e-05}
{"train_loss": 0.2534789741039276, "global_step": 87711, "epoch": 985, "lr": 7.654221664038349e-05}
{"train_loss": 0.20899608731269836, "global_step": 87712, "epoch": 985, "lr": 7.654172532959861e-05}
{"train_loss": 0.1157519519329071, "global_step": 87713, "epoch": 985, "lr": 7.65412340152455e-05}
{"train_loss": 0.22332972288131714, "global_step": 87714, "epoch": 985, "lr": 7.654074269732428e-05}
{"train_loss": 0.23628003895282745, "global_step": 87715, "epoch": 985, "lr": 7.654025137583497e-05}
{"train_loss": 0.1486445814371109, "global_step": 87716, "epoch": 985, "lr": 7.653976005077763e-05}
{"train_loss": 0.18493002653121948, "global_step": 87717, "epoch": 985, "lr": 7.653926872215235e-05}
{"train_loss": 0.2788207232952118, "global_step": 87718, "epoch": 985, "lr": 7.653877738995919e-05}
{"train_loss": 0.1562235951423645, "global_step": 87719, "epoch": 985, "lr": 7.653828605419821e-05}
{"train_loss": 0.17006073892116547, "global_step": 87720, "epoch": 985, "lr": 7.653779471486948e-05}
{"train_loss": 0.22137998044490814, "global_step": 87721, "epoch": 985, "lr": 7.653730337197308e-05}
{"train_loss": 0.21313151717185974, "global_step": 87722, "epoch": 985, "lr": 7.653681202550906e-05}
{"train_loss": 0.21549926698207855, "global_step": 87723, "epoch": 985, "lr": 7.653632067547748e-05}
{"train_loss": 0.15453793108463287, "global_step": 87724, "epoch": 985, "lr": 7.653582932187842e-05}
{"train_loss": 0.17476961016654968, "global_step": 87725, "epoch": 985, "lr": 7.653533796471195e-05}
{"train_loss": 0.24133071303367615, "global_step": 87726, "epoch": 985, "lr": 7.653484660397812e-05}
{"train_loss": 0.2636853754520416, "global_step": 87727, "epoch": 985, "lr": 7.653435523967699e-05}
{"train_loss": 0.1890583038330078, "global_step": 87728, "epoch": 985, "lr": 7.653386387180865e-05}
{"train_loss": 0.21130001544952393, "global_step": 87729, "epoch": 985, "lr": 7.653337250037315e-05}
{"train_loss": 0.1891869157552719, "global_step": 87730, "epoch": 985, "lr": 7.653288112537057e-05}
{"train_loss": 0.24453899264335632, "global_step": 87731, "epoch": 985, "lr": 7.653238974680097e-05}
{"train_loss": 0.1525907963514328, "global_step": 87732, "epoch": 985, "lr": 7.653189836466441e-05}
{"train_loss": 0.1675991415977478, "global_step": 87733, "epoch": 985, "lr": 7.653140697896095e-05}
{"train_loss": 0.18029223382472992, "global_step": 87734, "epoch": 985, "lr": 7.653091558969068e-05}
{"train_loss": 0.16203410923480988, "global_step": 87735, "epoch": 985, "lr": 7.653042419685366e-05}
{"train_loss": 0.2121885120868683, "global_step": 87736, "epoch": 985, "lr": 7.652993280044992e-05}
{"train_loss": 0.19452111423015594, "global_step": 87737, "epoch": 985, "lr": 7.652944140047957e-05}
{"train_loss": 0.22007179260253906, "global_step": 87738, "epoch": 985, "lr": 7.652894999694265e-05}
{"train_loss": 0.1874687224626541, "global_step": 87739, "epoch": 985, "lr": 7.652845858983926e-05}
{"train_loss": 0.14522847533226013, "global_step": 87740, "epoch": 985, "lr": 7.652796717916942e-05}
{"train_loss": 0.1694505363702774, "global_step": 87741, "epoch": 985, "lr": 7.652747576493322e-05}
{"train_loss": 0.15523695945739746, "global_step": 87742, "epoch": 985, "lr": 7.652698434713072e-05}
{"train_loss": 0.17029011249542236, "global_step": 87743, "epoch": 985, "lr": 7.652649292576201e-05}
{"train_loss": 0.17941772937774658, "global_step": 87744, "epoch": 985, "lr": 7.652600150082711e-05}
{"train_loss": 0.20177198946475983, "global_step": 87745, "epoch": 985, "lr": 7.652551007232614e-05}
{"train_loss": 0.22435548901557922, "global_step": 87746, "epoch": 985, "lr": 7.652501864025911e-05}
{"train_loss": 0.20872387290000916, "global_step": 87747, "epoch": 985, "lr": 7.652452720462615e-05}
{"train_loss": 0.19473448395729065, "global_step": 87748, "epoch": 985, "lr": 7.652403576542726e-05}
{"train_loss": 0.1460338979959488, "global_step": 87749, "epoch": 985, "lr": 7.652354432266256e-05}
{"train_loss": 0.1719914972782135, "global_step": 87750, "epoch": 985, "lr": 7.652305287633207e-05}
{"train_loss": 0.14932264387607574, "global_step": 87751, "epoch": 985, "lr": 7.65225614264359e-05}
{"train_loss": 0.14648477733135223, "global_step": 87752, "epoch": 985, "lr": 7.652206997297408e-05}
{"train_loss": 0.19116330280732574, "global_step": 87753, "epoch": 985, "lr": 7.652157851594669e-05, "val_loss": 4.191702365875244, "train_action_mse_error": 35.46861267089844}
{"train_loss": 0.17443682253360748, "global_step": 87754, "epoch": 986, "lr": 7.65210870553538e-05}
{"train_loss": 0.20212477445602417, "global_step": 87755, "epoch": 986, "lr": 7.652059559119548e-05}
{"train_loss": 0.2184114158153534, "global_step": 87756, "epoch": 986, "lr": 7.652010412347179e-05}
{"train_loss": 0.23782742023468018, "global_step": 87757, "epoch": 986, "lr": 7.651961265218277e-05}
{"train_loss": 0.10577734559774399, "global_step": 87758, "epoch": 986, "lr": 7.651912117732854e-05}
{"train_loss": 0.25041139125823975, "global_step": 87759, "epoch": 986, "lr": 7.651862969890913e-05}
{"train_loss": 0.20972830057144165, "global_step": 87760, "epoch": 986, "lr": 7.65181382169246e-05}
{"train_loss": 0.14733096957206726, "global_step": 87761, "epoch": 986, "lr": 7.651764673137505e-05}
{"train_loss": 0.17354285717010498, "global_step": 87762, "epoch": 986, "lr": 7.65171552422605e-05}
{"train_loss": 0.2124481201171875, "global_step": 87763, "epoch": 986, "lr": 7.651666374958108e-05}
{"train_loss": 0.15965193510055542, "global_step": 87764, "epoch": 986, "lr": 7.651617225333678e-05}
{"train_loss": 0.21431569755077362, "global_step": 87765, "epoch": 986, "lr": 7.651568075352774e-05}
{"train_loss": 0.26113906502723694, "global_step": 87766, "epoch": 986, "lr": 7.651518925015395e-05}
{"train_loss": 0.24723941087722778, "global_step": 87767, "epoch": 986, "lr": 7.651469774321553e-05}
{"train_loss": 0.15092846751213074, "global_step": 87768, "epoch": 986, "lr": 7.651420623271255e-05}
{"train_loss": 0.1404259353876114, "global_step": 87769, "epoch": 986, "lr": 7.651371471864504e-05}
{"train_loss": 0.1804182082414627, "global_step": 87770, "epoch": 986, "lr": 7.65132232010131e-05}
{"train_loss": 0.2185358703136444, "global_step": 87771, "epoch": 986, "lr": 7.651273167981677e-05}
{"train_loss": 0.19337469339370728, "global_step": 87772, "epoch": 986, "lr": 7.651224015505612e-05}
{"train_loss": 0.16516482830047607, "global_step": 87773, "epoch": 986, "lr": 7.651174862673123e-05}
{"train_loss": 0.18679922819137573, "global_step": 87774, "epoch": 986, "lr": 7.651125709484216e-05}
{"train_loss": 0.1545410007238388, "global_step": 87775, "epoch": 986, "lr": 7.651076555938898e-05}
{"train_loss": 0.16715145111083984, "global_step": 87776, "epoch": 986, "lr": 7.651027402037174e-05}
{"train_loss": 0.13590283691883087, "global_step": 87777, "epoch": 986, "lr": 7.65097824777905e-05}
{"train_loss": 0.13124285638332367, "global_step": 87778, "epoch": 986, "lr": 7.650929093164539e-05}
{"train_loss": 0.21801629662513733, "global_step": 87779, "epoch": 986, "lr": 7.65087993819364e-05}
{"train_loss": 0.1765328347682953, "global_step": 87780, "epoch": 986, "lr": 7.650830782866362e-05}
{"train_loss": 0.17580509185791016, "global_step": 87781, "epoch": 986, "lr": 7.650781627182713e-05}
{"train_loss": 0.15243996679782867, "global_step": 87782, "epoch": 986, "lr": 7.650732471142698e-05}
{"train_loss": 0.191275954246521, "global_step": 87783, "epoch": 986, "lr": 7.650683314746325e-05}
{"train_loss": 0.1255197823047638, "global_step": 87784, "epoch": 986, "lr": 7.6506341579936e-05}
{"train_loss": 0.11862237751483917, "global_step": 87785, "epoch": 986, "lr": 7.650585000884529e-05}
{"train_loss": 0.10564855486154556, "global_step": 87786, "epoch": 986, "lr": 7.650535843419118e-05}
{"train_loss": 0.21227744221687317, "global_step": 87787, "epoch": 986, "lr": 7.650486685597377e-05}
{"train_loss": 0.26419785618782043, "global_step": 87788, "epoch": 986, "lr": 7.65043752741931e-05}
{"train_loss": 0.19596020877361298, "global_step": 87789, "epoch": 986, "lr": 7.650388368884924e-05}
{"train_loss": 0.2050909399986267, "global_step": 87790, "epoch": 986, "lr": 7.650339209994225e-05}
{"train_loss": 0.21146264672279358, "global_step": 87791, "epoch": 986, "lr": 7.65029005074722e-05}
{"train_loss": 0.1908046454191208, "global_step": 87792, "epoch": 986, "lr": 7.650240891143917e-05}
{"train_loss": 0.19870997965335846, "global_step": 87793, "epoch": 986, "lr": 7.65019173118432e-05}
{"train_loss": 0.20431753993034363, "global_step": 87794, "epoch": 986, "lr": 7.650142570868437e-05}
{"train_loss": 0.1636984497308731, "global_step": 87795, "epoch": 986, "lr": 7.650093410196275e-05}
{"train_loss": 0.22441869974136353, "global_step": 87796, "epoch": 986, "lr": 7.65004424916784e-05}
{"train_loss": 0.23730973899364471, "global_step": 87797, "epoch": 986, "lr": 7.649995087783139e-05}
{"train_loss": 0.18458320200443268, "global_step": 87798, "epoch": 986, "lr": 7.64994592604218e-05}
{"train_loss": 0.2456827312707901, "global_step": 87799, "epoch": 986, "lr": 7.649896763944966e-05}
{"train_loss": 0.1558220535516739, "global_step": 87800, "epoch": 986, "lr": 7.649847601491506e-05}
{"train_loss": 0.23444537818431854, "global_step": 87801, "epoch": 986, "lr": 7.649798438681808e-05}
{"train_loss": 0.26242539286613464, "global_step": 87802, "epoch": 986, "lr": 7.649749275515876e-05}
{"train_loss": 0.1443394273519516, "global_step": 87803, "epoch": 986, "lr": 7.649700111993716e-05}
{"train_loss": 0.16822722554206848, "global_step": 87804, "epoch": 986, "lr": 7.649650948115339e-05}
{"train_loss": 0.24714872241020203, "global_step": 87805, "epoch": 986, "lr": 7.649601783880747e-05}
{"train_loss": 0.1971702128648758, "global_step": 87806, "epoch": 986, "lr": 7.649552619289948e-05}
{"train_loss": 0.22225765883922577, "global_step": 87807, "epoch": 986, "lr": 7.649503454342948e-05}
{"train_loss": 0.211078941822052, "global_step": 87808, "epoch": 986, "lr": 7.649454289039757e-05}
{"train_loss": 0.1404673308134079, "global_step": 87809, "epoch": 986, "lr": 7.649405123380377e-05}
{"train_loss": 0.17864036560058594, "global_step": 87810, "epoch": 986, "lr": 7.649355957364819e-05}
{"train_loss": 0.15489333868026733, "global_step": 87811, "epoch": 986, "lr": 7.649306790993085e-05}
{"train_loss": 0.17257533967494965, "global_step": 87812, "epoch": 986, "lr": 7.649257624265186e-05}
{"train_loss": 0.1814076006412506, "global_step": 87813, "epoch": 986, "lr": 7.649208457181125e-05}
{"train_loss": 0.18284744024276733, "global_step": 87814, "epoch": 986, "lr": 7.64915928974091e-05}
{"train_loss": 0.17403271794319153, "global_step": 87815, "epoch": 986, "lr": 7.649110121944549e-05}
{"train_loss": 0.1733694076538086, "global_step": 87816, "epoch": 986, "lr": 7.649060953792047e-05}
{"train_loss": 0.20605580508708954, "global_step": 87817, "epoch": 986, "lr": 7.649011785283413e-05}
{"train_loss": 0.207939013838768, "global_step": 87818, "epoch": 986, "lr": 7.64896261641865e-05}
{"train_loss": 0.16476480662822723, "global_step": 87819, "epoch": 986, "lr": 7.648913447197767e-05}
{"train_loss": 0.21382883191108704, "global_step": 87820, "epoch": 986, "lr": 7.648864277620768e-05}
{"train_loss": 0.20746079087257385, "global_step": 87821, "epoch": 986, "lr": 7.648815107687664e-05}
{"train_loss": 0.17579609155654907, "global_step": 87822, "epoch": 986, "lr": 7.648765937398456e-05}
{"train_loss": 0.22457128763198853, "global_step": 87823, "epoch": 986, "lr": 7.648716766753155e-05}
{"train_loss": 0.18143413960933685, "global_step": 87824, "epoch": 986, "lr": 7.648667595751767e-05}
{"train_loss": 0.09704899042844772, "global_step": 87825, "epoch": 986, "lr": 7.648618424394298e-05}
{"train_loss": 0.18348297476768494, "global_step": 87826, "epoch": 986, "lr": 7.648569252680754e-05}
{"train_loss": 0.2083505541086197, "global_step": 87827, "epoch": 986, "lr": 7.648520080611143e-05}
{"train_loss": 0.20139776170253754, "global_step": 87828, "epoch": 986, "lr": 7.64847090818547e-05}
{"train_loss": 0.20062342286109924, "global_step": 87829, "epoch": 986, "lr": 7.648421735403742e-05}
{"train_loss": 0.22938470542430878, "global_step": 87830, "epoch": 986, "lr": 7.648372562265966e-05}
{"train_loss": 0.16557446122169495, "global_step": 87831, "epoch": 986, "lr": 7.648323388772149e-05}
{"train_loss": 0.15451976656913757, "global_step": 87832, "epoch": 986, "lr": 7.648274214922297e-05}
{"train_loss": 0.24027658998966217, "global_step": 87833, "epoch": 986, "lr": 7.648225040716416e-05}
{"train_loss": 0.12774303555488586, "global_step": 87834, "epoch": 986, "lr": 7.648175866154514e-05}
{"train_loss": 0.2800222337245941, "global_step": 87835, "epoch": 986, "lr": 7.648126691236599e-05}
{"train_loss": 0.210391104221344, "global_step": 87836, "epoch": 986, "lr": 7.648077515962675e-05}
{"train_loss": 0.17073559761047363, "global_step": 87837, "epoch": 986, "lr": 7.648028340332747e-05}
{"train_loss": 0.1563008576631546, "global_step": 87838, "epoch": 986, "lr": 7.647979164346825e-05}
{"train_loss": 0.16189195215702057, "global_step": 87839, "epoch": 986, "lr": 7.647929988004916e-05}
{"train_loss": 0.18886122107505798, "global_step": 87840, "epoch": 986, "lr": 7.647880811307023e-05}
{"train_loss": 0.18763576447963715, "global_step": 87841, "epoch": 986, "lr": 7.647831634253155e-05}
{"train_loss": 0.18884893439793854, "global_step": 87842, "epoch": 986, "lr": 7.647782456843319e-05, "val_loss": 4.2968597412109375}
{"train_loss": 0.11671645194292068, "global_step": 87843, "epoch": 987, "lr": 7.647733279077521e-05}
{"train_loss": 0.11615724861621857, "global_step": 87844, "epoch": 987, "lr": 7.647684100955768e-05}
{"train_loss": 0.1303814798593521, "global_step": 87845, "epoch": 987, "lr": 7.647634922478064e-05}
{"train_loss": 0.15590952336788177, "global_step": 87846, "epoch": 987, "lr": 7.64758574364442e-05}
{"train_loss": 0.22421856224536896, "global_step": 87847, "epoch": 987, "lr": 7.647536564454839e-05}
{"train_loss": 0.19976140558719635, "global_step": 87848, "epoch": 987, "lr": 7.647487384909329e-05}
{"train_loss": 0.2612050175666809, "global_step": 87849, "epoch": 987, "lr": 7.647438205007898e-05}
{"train_loss": 0.16014622151851654, "global_step": 87850, "epoch": 987, "lr": 7.647389024750549e-05}
{"train_loss": 0.1967281550168991, "global_step": 87851, "epoch": 987, "lr": 7.647339844137291e-05}
{"train_loss": 0.23284085094928741, "global_step": 87852, "epoch": 987, "lr": 7.647290663168132e-05}
{"train_loss": 0.14541184902191162, "global_step": 87853, "epoch": 987, "lr": 7.647241481843077e-05}
{"train_loss": 0.25167688727378845, "global_step": 87854, "epoch": 987, "lr": 7.647192300162132e-05}
{"train_loss": 0.2135080248117447, "global_step": 87855, "epoch": 987, "lr": 7.647143118125304e-05}
{"train_loss": 0.16427528858184814, "global_step": 87856, "epoch": 987, "lr": 7.6470939357326e-05}
{"train_loss": 0.2191777378320694, "global_step": 87857, "epoch": 987, "lr": 7.647044752984029e-05}
{"train_loss": 0.1138622835278511, "global_step": 87858, "epoch": 987, "lr": 7.646995569879593e-05}
{"train_loss": 0.20710726082324982, "global_step": 87859, "epoch": 987, "lr": 7.6469463864193e-05}
{"train_loss": 0.15439525246620178, "global_step": 87860, "epoch": 987, "lr": 7.646897202603158e-05}
{"train_loss": 0.17496167123317719, "global_step": 87861, "epoch": 987, "lr": 7.646848018431172e-05}
{"train_loss": 0.1779634803533554, "global_step": 87862, "epoch": 987, "lr": 7.64679883390335e-05}
{"train_loss": 0.16946715116500854, "global_step": 87863, "epoch": 987, "lr": 7.6467496490197e-05}
{"train_loss": 0.19906271994113922, "global_step": 87864, "epoch": 987, "lr": 7.646700463780225e-05}
{"train_loss": 0.15189340710639954, "global_step": 87865, "epoch": 987, "lr": 7.646651278184933e-05}
{"train_loss": 0.17043904960155487, "global_step": 87866, "epoch": 987, "lr": 7.646602092233832e-05}
{"train_loss": 0.18966491520404816, "global_step": 87867, "epoch": 987, "lr": 7.646552905926927e-05}
{"train_loss": 0.12214729189872742, "global_step": 87868, "epoch": 987, "lr": 7.646503719264226e-05}
{"train_loss": 0.15491139888763428, "global_step": 87869, "epoch": 987, "lr": 7.646454532245733e-05}
{"train_loss": 0.14510269463062286, "global_step": 87870, "epoch": 987, "lr": 7.646405344871459e-05}
{"train_loss": 0.13402707874774933, "global_step": 87871, "epoch": 987, "lr": 7.646356157141406e-05}
{"train_loss": 0.18710090219974518, "global_step": 87872, "epoch": 987, "lr": 7.646306969055583e-05}
{"train_loss": 0.3045385777950287, "global_step": 87873, "epoch": 987, "lr": 7.646257780613997e-05}
{"train_loss": 0.17794452607631683, "global_step": 87874, "epoch": 987, "lr": 7.646208591816654e-05}
{"train_loss": 0.1650206297636032, "global_step": 87875, "epoch": 987, "lr": 7.646159402663562e-05}
{"train_loss": 0.1576760858297348, "global_step": 87876, "epoch": 987, "lr": 7.646110213154724e-05}
{"train_loss": 0.187088280916214, "global_step": 87877, "epoch": 987, "lr": 7.646061023290149e-05}
{"train_loss": 0.2517913281917572, "global_step": 87878, "epoch": 987, "lr": 7.646011833069843e-05}
{"train_loss": 0.16034959256649017, "global_step": 87879, "epoch": 987, "lr": 7.645962642493813e-05}
{"train_loss": 0.144279345870018, "global_step": 87880, "epoch": 987, "lr": 7.645913451562066e-05}
{"train_loss": 0.2519418001174927, "global_step": 87881, "epoch": 987, "lr": 7.645864260274609e-05}
{"train_loss": 0.30285191535949707, "global_step": 87882, "epoch": 987, "lr": 7.645815068631445e-05}
{"train_loss": 0.17719338834285736, "global_step": 87883, "epoch": 987, "lr": 7.645765876632586e-05}
{"train_loss": 0.1283998191356659, "global_step": 87884, "epoch": 987, "lr": 7.645716684278036e-05}
{"train_loss": 0.2018895298242569, "global_step": 87885, "epoch": 987, "lr": 7.6456674915678e-05}
{"train_loss": 0.10610495507717133, "global_step": 87886, "epoch": 987, "lr": 7.645618298501886e-05}
{"train_loss": 0.16817958652973175, "global_step": 87887, "epoch": 987, "lr": 7.645569105080303e-05}
{"train_loss": 0.15746232867240906, "global_step": 87888, "epoch": 987, "lr": 7.645519911303055e-05}
{"train_loss": 0.1994415670633316, "global_step": 87889, "epoch": 987, "lr": 7.645470717170148e-05}
{"train_loss": 0.21070058643817902, "global_step": 87890, "epoch": 987, "lr": 7.645421522681591e-05}
{"train_loss": 0.1704423874616623, "global_step": 87891, "epoch": 987, "lr": 7.645372327837387e-05}
{"train_loss": 0.16126756370067596, "global_step": 87892, "epoch": 987, "lr": 7.645323132637548e-05}
{"train_loss": 0.18895530700683594, "global_step": 87893, "epoch": 987, "lr": 7.645273937082076e-05}
{"train_loss": 0.1545456349849701, "global_step": 87894, "epoch": 987, "lr": 7.64522474117098e-05}
{"train_loss": 0.1372835636138916, "global_step": 87895, "epoch": 987, "lr": 7.645175544904267e-05}
{"train_loss": 0.16049648821353912, "global_step": 87896, "epoch": 987, "lr": 7.645126348281939e-05}
{"train_loss": 0.08886950463056564, "global_step": 87897, "epoch": 987, "lr": 7.645077151304007e-05}
{"train_loss": 0.23238733410835266, "global_step": 87898, "epoch": 987, "lr": 7.645027953970478e-05}
{"train_loss": 0.17665867507457733, "global_step": 87899, "epoch": 987, "lr": 7.644978756281356e-05}
{"train_loss": 0.2316117137670517, "global_step": 87900, "epoch": 987, "lr": 7.644929558236648e-05}
{"train_loss": 0.16071343421936035, "global_step": 87901, "epoch": 987, "lr": 7.644880359836364e-05}
{"train_loss": 0.2872985005378723, "global_step": 87902, "epoch": 987, "lr": 7.644831161080507e-05}
{"train_loss": 0.17282861471176147, "global_step": 87903, "epoch": 987, "lr": 7.644781961969084e-05}
{"train_loss": 0.12636947631835938, "global_step": 87904, "epoch": 987, "lr": 7.644732762502102e-05}
{"train_loss": 0.1877092868089676, "global_step": 87905, "epoch": 987, "lr": 7.644683562679568e-05}
{"train_loss": 0.2456434965133667, "global_step": 87906, "epoch": 987, "lr": 7.644634362501491e-05}
{"train_loss": 0.30176591873168945, "global_step": 87907, "epoch": 987, "lr": 7.644585161967873e-05}
{"train_loss": 0.22373297810554504, "global_step": 87908, "epoch": 987, "lr": 7.644535961078722e-05}
{"train_loss": 0.1690686047077179, "global_step": 87909, "epoch": 987, "lr": 7.644486759834046e-05}
{"train_loss": 0.25873127579689026, "global_step": 87910, "epoch": 987, "lr": 7.644437558233852e-05}
{"train_loss": 0.20492471754550934, "global_step": 87911, "epoch": 987, "lr": 7.644388356278145e-05}
{"train_loss": 0.16151754558086395, "global_step": 87912, "epoch": 987, "lr": 7.644339153966933e-05}
{"train_loss": 0.18043482303619385, "global_step": 87913, "epoch": 987, "lr": 7.644289951300221e-05}
{"train_loss": 0.28772783279418945, "global_step": 87914, "epoch": 987, "lr": 7.644240748278017e-05}
{"train_loss": 0.42927345633506775, "global_step": 87915, "epoch": 987, "lr": 7.644191544900326e-05}
{"train_loss": 0.2521970272064209, "global_step": 87916, "epoch": 987, "lr": 7.644142341167156e-05}
{"train_loss": 0.2080412060022354, "global_step": 87917, "epoch": 987, "lr": 7.644093137078514e-05}
{"train_loss": 0.22099070250988007, "global_step": 87918, "epoch": 987, "lr": 7.644043932634406e-05}
{"train_loss": 0.13516588509082794, "global_step": 87919, "epoch": 987, "lr": 7.643994727834838e-05}
{"train_loss": 0.10346635431051254, "global_step": 87920, "epoch": 987, "lr": 7.643945522679816e-05}
{"train_loss": 0.23890593647956848, "global_step": 87921, "epoch": 987, "lr": 7.64389631716935e-05}
{"train_loss": 0.21277153491973877, "global_step": 87922, "epoch": 987, "lr": 7.643847111303441e-05}
{"train_loss": 0.23255513608455658, "global_step": 87923, "epoch": 987, "lr": 7.643797905082102e-05}
{"train_loss": 0.16404518485069275, "global_step": 87924, "epoch": 987, "lr": 7.643748698505336e-05}
{"train_loss": 0.18317224085330963, "global_step": 87925, "epoch": 987, "lr": 7.643699491573149e-05}
{"train_loss": 0.17564302682876587, "global_step": 87926, "epoch": 987, "lr": 7.64365028428555e-05}
{"train_loss": 0.24165204167366028, "global_step": 87927, "epoch": 987, "lr": 7.643601076642544e-05}
{"train_loss": 0.13882037997245789, "global_step": 87928, "epoch": 987, "lr": 7.643551868644138e-05}
{"train_loss": 0.22976890206336975, "global_step": 87929, "epoch": 987, "lr": 7.64350266029034e-05}
{"train_loss": 0.21139028668403625, "global_step": 87930, "epoch": 987, "lr": 7.643453451581154e-05}
{"train_loss": 0.1900359025497115, "global_step": 87931, "epoch": 987, "lr": 7.643404242516587e-05, "val_loss": 4.100401401519775}
{"train_loss": 0.1411643922328949, "global_step": 87932, "epoch": 988, "lr": 7.643355033096649e-05}
{"train_loss": 0.13906696438789368, "global_step": 87933, "epoch": 988, "lr": 7.643305823321344e-05}
{"train_loss": 0.15618827939033508, "global_step": 87934, "epoch": 988, "lr": 7.643256613190676e-05}
{"train_loss": 0.17150898277759552, "global_step": 87935, "epoch": 988, "lr": 7.643207402704658e-05}
{"train_loss": 0.21102985739707947, "global_step": 87936, "epoch": 988, "lr": 7.64315819186329e-05}
{"train_loss": 0.14018750190734863, "global_step": 87937, "epoch": 988, "lr": 7.643108980666582e-05}
{"train_loss": 0.23564639687538147, "global_step": 87938, "epoch": 988, "lr": 7.643059769114542e-05}
{"train_loss": 0.18092338740825653, "global_step": 87939, "epoch": 988, "lr": 7.643010557207173e-05}
{"train_loss": 0.14135020971298218, "global_step": 87940, "epoch": 988, "lr": 7.642961344944484e-05}
{"train_loss": 0.10423906147480011, "global_step": 87941, "epoch": 988, "lr": 7.642912132326481e-05}
{"train_loss": 0.18457268178462982, "global_step": 87942, "epoch": 988, "lr": 7.64286291935317e-05}
{"train_loss": 0.16238044202327728, "global_step": 87943, "epoch": 988, "lr": 7.642813706024558e-05}
{"train_loss": 0.18697364628314972, "global_step": 87944, "epoch": 988, "lr": 7.642764492340654e-05}
{"train_loss": 0.11094561964273453, "global_step": 87945, "epoch": 988, "lr": 7.642715278301461e-05}
{"train_loss": 0.18158188462257385, "global_step": 87946, "epoch": 988, "lr": 7.642666063906987e-05}
{"train_loss": 0.1108875572681427, "global_step": 87947, "epoch": 988, "lr": 7.642616849157241e-05}
{"train_loss": 0.2011956125497818, "global_step": 87948, "epoch": 988, "lr": 7.642567634052224e-05}
{"train_loss": 0.126039519906044, "global_step": 87949, "epoch": 988, "lr": 7.642518418591949e-05}
{"train_loss": 0.1320970207452774, "global_step": 87950, "epoch": 988, "lr": 7.64246920277642e-05}
{"train_loss": 0.21080467104911804, "global_step": 87951, "epoch": 988, "lr": 7.64241998660564e-05}
{"train_loss": 0.1821601688861847, "global_step": 87952, "epoch": 988, "lr": 7.642370770079621e-05}
{"train_loss": 0.16225437819957733, "global_step": 87953, "epoch": 988, "lr": 7.642321553198366e-05}
{"train_loss": 0.12507902085781097, "global_step": 87954, "epoch": 988, "lr": 7.642272335961885e-05}
{"train_loss": 0.16745427250862122, "global_step": 87955, "epoch": 988, "lr": 7.642223118370182e-05}
{"train_loss": 0.1664412021636963, "global_step": 87956, "epoch": 988, "lr": 7.642173900423264e-05}
{"train_loss": 0.18214747309684753, "global_step": 87957, "epoch": 988, "lr": 7.642124682121137e-05}
{"train_loss": 0.19833074510097504, "global_step": 87958, "epoch": 988, "lr": 7.64207546346381e-05}
{"train_loss": 0.18406979739665985, "global_step": 87959, "epoch": 988, "lr": 7.642026244451286e-05}
{"train_loss": 0.23339247703552246, "global_step": 87960, "epoch": 988, "lr": 7.641977025083576e-05}
{"train_loss": 0.2339354306459427, "global_step": 87961, "epoch": 988, "lr": 7.641927805360682e-05}
{"train_loss": 0.20795968174934387, "global_step": 87962, "epoch": 988, "lr": 7.641878585282616e-05}
{"train_loss": 0.16988405585289001, "global_step": 87963, "epoch": 988, "lr": 7.64182936484938e-05}
{"train_loss": 0.2131166309118271, "global_step": 87964, "epoch": 988, "lr": 7.641780144060982e-05}
{"train_loss": 0.1434781551361084, "global_step": 87965, "epoch": 988, "lr": 7.64173092291743e-05}
{"train_loss": 0.17830374836921692, "global_step": 87966, "epoch": 988, "lr": 7.64168170141873e-05}
{"train_loss": 0.23739399015903473, "global_step": 87967, "epoch": 988, "lr": 7.641632479564887e-05}
{"train_loss": 0.17220473289489746, "global_step": 87968, "epoch": 988, "lr": 7.641583257355908e-05}
{"train_loss": 0.1510525941848755, "global_step": 87969, "epoch": 988, "lr": 7.641534034791801e-05}
{"train_loss": 0.18607059121131897, "global_step": 87970, "epoch": 988, "lr": 7.641484811872573e-05}
{"train_loss": 0.15571552515029907, "global_step": 87971, "epoch": 988, "lr": 7.641435588598227e-05}
{"train_loss": 0.17676202952861786, "global_step": 87972, "epoch": 988, "lr": 7.641386364968775e-05}
{"train_loss": 0.2036406546831131, "global_step": 87973, "epoch": 988, "lr": 7.641337140984219e-05}
{"train_loss": 0.1239936500787735, "global_step": 87974, "epoch": 988, "lr": 7.641287916644568e-05}
{"train_loss": 0.21797257661819458, "global_step": 87975, "epoch": 988, "lr": 7.641238691949829e-05}
{"train_loss": 0.23141607642173767, "global_step": 87976, "epoch": 988, "lr": 7.641189466900006e-05}
{"train_loss": 0.170293390750885, "global_step": 87977, "epoch": 988, "lr": 7.641140241495108e-05}
{"train_loss": 0.23921014368534088, "global_step": 87978, "epoch": 988, "lr": 7.64109101573514e-05}
{"train_loss": 0.26001811027526855, "global_step": 87979, "epoch": 988, "lr": 7.64104178962011e-05}
{"train_loss": 0.13163812458515167, "global_step": 87980, "epoch": 988, "lr": 7.640992563150025e-05}
{"train_loss": 0.22083070874214172, "global_step": 87981, "epoch": 988, "lr": 7.64094333632489e-05}
{"train_loss": 0.10170530527830124, "global_step": 87982, "epoch": 988, "lr": 7.640894109144711e-05}
{"train_loss": 0.26565393805503845, "global_step": 87983, "epoch": 988, "lr": 7.6408448816095e-05}
{"train_loss": 0.19518014788627625, "global_step": 87984, "epoch": 988, "lr": 7.640795653719255e-05}
{"train_loss": 0.12186990678310394, "global_step": 87985, "epoch": 988, "lr": 7.64074642547399e-05}
{"train_loss": 0.14600789546966553, "global_step": 87986, "epoch": 988, "lr": 7.640697196873708e-05}
{"train_loss": 0.23129884898662567, "global_step": 87987, "epoch": 988, "lr": 7.640647967918418e-05}
{"train_loss": 0.14849291741847992, "global_step": 87988, "epoch": 988, "lr": 7.640598738608122e-05}
{"train_loss": 0.13030798733234406, "global_step": 87989, "epoch": 988, "lr": 7.640549508942831e-05}
{"train_loss": 0.18804921209812164, "global_step": 87990, "epoch": 988, "lr": 7.640500278922551e-05}
{"train_loss": 0.15471363067626953, "global_step": 87991, "epoch": 988, "lr": 7.640451048547289e-05}
{"train_loss": 0.16410064697265625, "global_step": 87992, "epoch": 988, "lr": 7.640401817817049e-05}
{"train_loss": 0.1536971926689148, "global_step": 87993, "epoch": 988, "lr": 7.640352586731838e-05}
{"train_loss": 0.29044100642204285, "global_step": 87994, "epoch": 988, "lr": 7.640303355291666e-05}
{"train_loss": 0.19695642590522766, "global_step": 87995, "epoch": 988, "lr": 7.640254123496535e-05}
{"train_loss": 0.12549421191215515, "global_step": 87996, "epoch": 988, "lr": 7.640204891346456e-05}
{"train_loss": 0.282866507768631, "global_step": 87997, "epoch": 988, "lr": 7.640155658841433e-05}
{"train_loss": 0.18107903003692627, "global_step": 87998, "epoch": 988, "lr": 7.640106425981473e-05}
{"train_loss": 0.1356351524591446, "global_step": 87999, "epoch": 988, "lr": 7.640057192766583e-05}
{"train_loss": 0.1448328197002411, "global_step": 88000, "epoch": 988, "lr": 7.64000795919677e-05}
{"train_loss": 0.19825008511543274, "global_step": 88001, "epoch": 988, "lr": 7.63995872527204e-05}
{"train_loss": 0.21635936200618744, "global_step": 88002, "epoch": 988, "lr": 7.639909490992398e-05}
{"train_loss": 0.18288670480251312, "global_step": 88003, "epoch": 988, "lr": 7.639860256357855e-05}
{"train_loss": 0.13528403639793396, "global_step": 88004, "epoch": 988, "lr": 7.639811021368414e-05}
{"train_loss": 0.21492999792099, "global_step": 88005, "epoch": 988, "lr": 7.639761786024083e-05}
{"train_loss": 0.21949538588523865, "global_step": 88006, "epoch": 988, "lr": 7.639712550324867e-05}
{"train_loss": 0.17819540202617645, "global_step": 88007, "epoch": 988, "lr": 7.639663314270774e-05}
{"train_loss": 0.15786941349506378, "global_step": 88008, "epoch": 988, "lr": 7.639614077861811e-05}
{"train_loss": 0.24441243708133698, "global_step": 88009, "epoch": 988, "lr": 7.639564841097983e-05}
{"train_loss": 0.19580596685409546, "global_step": 88010, "epoch": 988, "lr": 7.6395156039793e-05}
{"train_loss": 0.13839395344257355, "global_step": 88011, "epoch": 988, "lr": 7.639466366505766e-05}
{"train_loss": 0.16725851595401764, "global_step": 88012, "epoch": 988, "lr": 7.639417128677387e-05}
{"train_loss": 0.1597229689359665, "global_step": 88013, "epoch": 988, "lr": 7.639367890494169e-05}
{"train_loss": 0.23217429220676422, "global_step": 88014, "epoch": 988, "lr": 7.63931865195612e-05}
{"train_loss": 0.1654825359582901, "global_step": 88015, "epoch": 988, "lr": 7.639269413063249e-05}
{"train_loss": 0.14750784635543823, "global_step": 88016, "epoch": 988, "lr": 7.63922017381556e-05}
{"train_loss": 0.1693774163722992, "global_step": 88017, "epoch": 988, "lr": 7.639170934213059e-05}
{"train_loss": 0.15848688781261444, "global_step": 88018, "epoch": 988, "lr": 7.639121694255755e-05}
{"train_loss": 0.21932560205459595, "global_step": 88019, "epoch": 988, "lr": 7.639072453943652e-05}
{"train_loss": 0.17804473822706202, "global_step": 88020, "epoch": 988, "lr": 7.639023213276759e-05, "val_loss": 4.344656944274902}
{"train_loss": 0.19570814073085785, "global_step": 88021, "epoch": 989, "lr": 7.638973972255081e-05}
{"train_loss": 0.20131956040859222, "global_step": 88022, "epoch": 989, "lr": 7.638924730878625e-05}
{"train_loss": 0.16398100554943085, "global_step": 88023, "epoch": 989, "lr": 7.638875489147397e-05}
{"train_loss": 0.20788690447807312, "global_step": 88024, "epoch": 989, "lr": 7.638826247061406e-05}
{"train_loss": 0.2002631276845932, "global_step": 88025, "epoch": 989, "lr": 7.638777004620655e-05}
{"train_loss": 0.17720195651054382, "global_step": 88026, "epoch": 989, "lr": 7.638727761825155e-05}
{"train_loss": 0.16847454011440277, "global_step": 88027, "epoch": 989, "lr": 7.638678518674909e-05}
{"train_loss": 0.2439633011817932, "global_step": 88028, "epoch": 989, "lr": 7.638629275169924e-05}
{"train_loss": 0.17186230421066284, "global_step": 88029, "epoch": 989, "lr": 7.638580031310208e-05}
{"train_loss": 0.13627547025680542, "global_step": 88030, "epoch": 989, "lr": 7.638530787095769e-05}
{"train_loss": 0.1664648950099945, "global_step": 88031, "epoch": 989, "lr": 7.638481542526608e-05}
{"train_loss": 0.1380605846643448, "global_step": 88032, "epoch": 989, "lr": 7.638432297602738e-05}
{"train_loss": 0.1460515707731247, "global_step": 88033, "epoch": 989, "lr": 7.638383052324161e-05}
{"train_loss": 0.14215029776096344, "global_step": 88034, "epoch": 989, "lr": 7.638333806690888e-05}
{"train_loss": 0.1447322964668274, "global_step": 88035, "epoch": 989, "lr": 7.638284560702922e-05}
{"train_loss": 0.22424018383026123, "global_step": 88036, "epoch": 989, "lr": 7.63823531436027e-05}
{"train_loss": 0.1624089777469635, "global_step": 88037, "epoch": 989, "lr": 7.63818606766294e-05}
{"train_loss": 0.19273021817207336, "global_step": 88038, "epoch": 989, "lr": 7.63813682061094e-05}
{"train_loss": 0.17753179371356964, "global_step": 88039, "epoch": 989, "lr": 7.63808757320427e-05}
{"train_loss": 0.21013560891151428, "global_step": 88040, "epoch": 989, "lr": 7.638038325442945e-05}
{"train_loss": 0.12711118161678314, "global_step": 88041, "epoch": 989, "lr": 7.637989077326968e-05}
{"train_loss": 0.12167128920555115, "global_step": 88042, "epoch": 989, "lr": 7.637939828856345e-05}
{"train_loss": 0.23253099620342255, "global_step": 88043, "epoch": 989, "lr": 7.637890580031082e-05}
{"train_loss": 0.14566828310489655, "global_step": 88044, "epoch": 989, "lr": 7.637841330851188e-05}
{"train_loss": 0.12441691011190414, "global_step": 88045, "epoch": 989, "lr": 7.637792081316667e-05}
{"train_loss": 0.16023442149162292, "global_step": 88046, "epoch": 989, "lr": 7.637742831427527e-05}
{"train_loss": 0.20186938345432281, "global_step": 88047, "epoch": 989, "lr": 7.637693581183776e-05}
{"train_loss": 0.17725880444049835, "global_step": 88048, "epoch": 989, "lr": 7.63764433058542e-05}
{"train_loss": 0.18390439450740814, "global_step": 88049, "epoch": 989, "lr": 7.637595079632462e-05}
{"train_loss": 0.2052752822637558, "global_step": 88050, "epoch": 989, "lr": 7.637545828324914e-05}
{"train_loss": 0.21823924779891968, "global_step": 88051, "epoch": 989, "lr": 7.637496576662779e-05}
{"train_loss": 0.1826881468296051, "global_step": 88052, "epoch": 989, "lr": 7.637447324646065e-05}
{"train_loss": 0.18456724286079407, "global_step": 88053, "epoch": 989, "lr": 7.637398072274779e-05}
{"train_loss": 0.22833506762981415, "global_step": 88054, "epoch": 989, "lr": 7.637348819548927e-05}
{"train_loss": 0.14305219054222107, "global_step": 88055, "epoch": 989, "lr": 7.637299566468515e-05}
{"train_loss": 0.21190007030963898, "global_step": 88056, "epoch": 989, "lr": 7.637250313033548e-05}
{"train_loss": 0.17606253921985626, "global_step": 88057, "epoch": 989, "lr": 7.637201059244037e-05}
{"train_loss": 0.16092953085899353, "global_step": 88058, "epoch": 989, "lr": 7.637151805099986e-05}
{"train_loss": 0.18879389762878418, "global_step": 88059, "epoch": 989, "lr": 7.637102550601403e-05}
{"train_loss": 0.20777076482772827, "global_step": 88060, "epoch": 989, "lr": 7.637053295748293e-05}
{"train_loss": 0.11602606624364853, "global_step": 88061, "epoch": 989, "lr": 7.637004040540664e-05}
{"train_loss": 0.15539461374282837, "global_step": 88062, "epoch": 989, "lr": 7.63695478497852e-05}
{"train_loss": 0.18110139667987823, "global_step": 88063, "epoch": 989, "lr": 7.636905529061874e-05}
{"train_loss": 0.11667031794786453, "global_step": 88064, "epoch": 989, "lr": 7.636856272790724e-05}
{"train_loss": 0.20777268707752228, "global_step": 88065, "epoch": 989, "lr": 7.636807016165082e-05}
{"train_loss": 0.2272639274597168, "global_step": 88066, "epoch": 989, "lr": 7.636757759184953e-05}
{"train_loss": 0.1195506900548935, "global_step": 88067, "epoch": 989, "lr": 7.636708501850345e-05}
{"train_loss": 0.1985878348350525, "global_step": 88068, "epoch": 989, "lr": 7.636659244161263e-05}
{"train_loss": 0.21971675753593445, "global_step": 88069, "epoch": 989, "lr": 7.636609986117714e-05}
{"train_loss": 0.2276976853609085, "global_step": 88070, "epoch": 989, "lr": 7.636560727719705e-05}
{"train_loss": 0.1432715505361557, "global_step": 88071, "epoch": 989, "lr": 7.636511468967244e-05}
{"train_loss": 0.1902925819158554, "global_step": 88072, "epoch": 989, "lr": 7.636462209860335e-05}
{"train_loss": 0.11320603638887405, "global_step": 88073, "epoch": 989, "lr": 7.636412950398985e-05}
{"train_loss": 0.1836090385913849, "global_step": 88074, "epoch": 989, "lr": 7.636363690583203e-05}
{"train_loss": 0.1829773187637329, "global_step": 88075, "epoch": 989, "lr": 7.636314430412993e-05}
{"train_loss": 0.14796610176563263, "global_step": 88076, "epoch": 989, "lr": 7.636265169888363e-05}
{"train_loss": 0.1862364262342453, "global_step": 88077, "epoch": 989, "lr": 7.636215909009319e-05}
{"train_loss": 0.21399714052677155, "global_step": 88078, "epoch": 989, "lr": 7.636166647775869e-05}
{"train_loss": 0.19604697823524475, "global_step": 88079, "epoch": 989, "lr": 7.636117386188017e-05}
{"train_loss": 0.13052348792552948, "global_step": 88080, "epoch": 989, "lr": 7.636068124245772e-05}
{"train_loss": 0.1915796995162964, "global_step": 88081, "epoch": 989, "lr": 7.636018861949138e-05}
{"train_loss": 0.25630804896354675, "global_step": 88082, "epoch": 989, "lr": 7.635969599298127e-05}
{"train_loss": 0.20341192185878754, "global_step": 88083, "epoch": 989, "lr": 7.63592033629274e-05}
{"train_loss": 0.13840846717357635, "global_step": 88084, "epoch": 989, "lr": 7.635871072932986e-05}
{"train_loss": 0.2064792811870575, "global_step": 88085, "epoch": 989, "lr": 7.635821809218871e-05}
{"train_loss": 0.14481960237026215, "global_step": 88086, "epoch": 989, "lr": 7.635772545150402e-05}
{"train_loss": 0.1854362189769745, "global_step": 88087, "epoch": 989, "lr": 7.635723280727585e-05}
{"train_loss": 0.15427462756633759, "global_step": 88088, "epoch": 989, "lr": 7.635674015950427e-05}
{"train_loss": 0.26872387528419495, "global_step": 88089, "epoch": 989, "lr": 7.635624750818934e-05}
{"train_loss": 0.17802214622497559, "global_step": 88090, "epoch": 989, "lr": 7.635575485333115e-05}
{"train_loss": 0.11764228343963623, "global_step": 88091, "epoch": 989, "lr": 7.635526219492976e-05}
{"train_loss": 0.11433538049459457, "global_step": 88092, "epoch": 989, "lr": 7.635476953298521e-05}
{"train_loss": 0.2734401226043701, "global_step": 88093, "epoch": 989, "lr": 7.635427686749759e-05}
{"train_loss": 0.18905878067016602, "global_step": 88094, "epoch": 989, "lr": 7.635378419846695e-05}
{"train_loss": 0.1844319999217987, "global_step": 88095, "epoch": 989, "lr": 7.635329152589337e-05}
{"train_loss": 0.2040860652923584, "global_step": 88096, "epoch": 989, "lr": 7.635279884977691e-05}
{"train_loss": 0.1324804425239563, "global_step": 88097, "epoch": 989, "lr": 7.635230617011762e-05}
{"train_loss": 0.12751424312591553, "global_step": 88098, "epoch": 989, "lr": 7.635181348691561e-05}
{"train_loss": 0.2592768371105194, "global_step": 88099, "epoch": 989, "lr": 7.635132080017091e-05}
{"train_loss": 0.2563701868057251, "global_step": 88100, "epoch": 989, "lr": 7.635082810988359e-05}
{"train_loss": 0.1376565843820572, "global_step": 88101, "epoch": 989, "lr": 7.635033541605374e-05}
{"train_loss": 0.17275084555149078, "global_step": 88102, "epoch": 989, "lr": 7.63498427186814e-05}
{"train_loss": 0.19547994434833527, "global_step": 88103, "epoch": 989, "lr": 7.634935001776663e-05}
{"train_loss": 0.21142303943634033, "global_step": 88104, "epoch": 989, "lr": 7.634885731330952e-05}
{"train_loss": 0.13534966111183167, "global_step": 88105, "epoch": 989, "lr": 7.634836460531014e-05}
{"train_loss": 0.14275214076042175, "global_step": 88106, "epoch": 989, "lr": 7.634787189376853e-05}
{"train_loss": 0.15915976464748383, "global_step": 88107, "epoch": 989, "lr": 7.634737917868478e-05}
{"train_loss": 0.2612011730670929, "global_step": 88108, "epoch": 989, "lr": 7.634688646005893e-05}
{"train_loss": 0.17930335317100032, "global_step": 88109, "epoch": 989, "lr": 7.634639373789108e-05, "val_loss": 4.222689151763916}
{"train_loss": 0.15207460522651672, "global_step": 88110, "epoch": 990, "lr": 7.634590101218126e-05}
{"train_loss": 0.1669454425573349, "global_step": 88111, "epoch": 990, "lr": 7.634540828292957e-05}
{"train_loss": 0.15154489874839783, "global_step": 88112, "epoch": 990, "lr": 7.634491555013606e-05}
{"train_loss": 0.21898533403873444, "global_step": 88113, "epoch": 990, "lr": 7.634442281380079e-05}
{"train_loss": 0.2079484462738037, "global_step": 88114, "epoch": 990, "lr": 7.634393007392384e-05}
{"train_loss": 0.1854294091463089, "global_step": 88115, "epoch": 990, "lr": 7.634343733050526e-05}
{"train_loss": 0.16283109784126282, "global_step": 88116, "epoch": 990, "lr": 7.634294458354514e-05}
{"train_loss": 0.20370535552501678, "global_step": 88117, "epoch": 990, "lr": 7.634245183304351e-05}
{"train_loss": 0.14168718457221985, "global_step": 88118, "epoch": 990, "lr": 7.634195907900048e-05}
{"train_loss": 0.1680569350719452, "global_step": 88119, "epoch": 990, "lr": 7.634146632141609e-05}
{"train_loss": 0.30280742049217224, "global_step": 88120, "epoch": 990, "lr": 7.634097356029041e-05}
{"train_loss": 0.20726585388183594, "global_step": 88121, "epoch": 990, "lr": 7.63404807956235e-05}
{"train_loss": 0.19925178587436676, "global_step": 88122, "epoch": 990, "lr": 7.633998802741545e-05}
{"train_loss": 0.20207984745502472, "global_step": 88123, "epoch": 990, "lr": 7.633949525566629e-05}
{"train_loss": 0.1566123068332672, "global_step": 88124, "epoch": 990, "lr": 7.633900248037611e-05}
{"train_loss": 0.15505391359329224, "global_step": 88125, "epoch": 990, "lr": 7.633850970154498e-05}
{"train_loss": 0.24683518707752228, "global_step": 88126, "epoch": 990, "lr": 7.633801691917294e-05}
{"train_loss": 0.22329960763454437, "global_step": 88127, "epoch": 990, "lr": 7.63375241332601e-05}
{"train_loss": 0.1574939489364624, "global_step": 88128, "epoch": 990, "lr": 7.63370313438065e-05}
{"train_loss": 0.2198878526687622, "global_step": 88129, "epoch": 990, "lr": 7.63365385508122e-05}
{"train_loss": 0.18159924447536469, "global_step": 88130, "epoch": 990, "lr": 7.633604575427726e-05}
{"train_loss": 0.14181800186634064, "global_step": 88131, "epoch": 990, "lr": 7.633555295420178e-05}
{"train_loss": 0.29979950189590454, "global_step": 88132, "epoch": 990, "lr": 7.633506015058579e-05}
{"train_loss": 0.15236735343933105, "global_step": 88133, "epoch": 990, "lr": 7.633456734342939e-05}
{"train_loss": 0.14865811169147491, "global_step": 88134, "epoch": 990, "lr": 7.63340745327326e-05}
{"train_loss": 0.16203632950782776, "global_step": 88135, "epoch": 990, "lr": 7.633358171849553e-05}
{"train_loss": 0.22125956416130066, "global_step": 88136, "epoch": 990, "lr": 7.633308890071824e-05}
{"train_loss": 0.16836586594581604, "global_step": 88137, "epoch": 990, "lr": 7.633259607940078e-05}
{"train_loss": 0.24884635210037231, "global_step": 88138, "epoch": 990, "lr": 7.633210325454324e-05}
{"train_loss": 0.15624865889549255, "global_step": 88139, "epoch": 990, "lr": 7.633161042614564e-05}
{"train_loss": 0.131802037358284, "global_step": 88140, "epoch": 990, "lr": 7.633111759420808e-05}
{"train_loss": 0.17977923154830933, "global_step": 88141, "epoch": 990, "lr": 7.633062475873064e-05}
{"train_loss": 0.16277232766151428, "global_step": 88142, "epoch": 990, "lr": 7.633013191971335e-05}
{"train_loss": 0.1911294013261795, "global_step": 88143, "epoch": 990, "lr": 7.632963907715631e-05}
{"train_loss": 0.1327282190322876, "global_step": 88144, "epoch": 990, "lr": 7.632914623105956e-05}
{"train_loss": 0.23326484858989716, "global_step": 88145, "epoch": 990, "lr": 7.632865338142319e-05}
{"train_loss": 0.1652488112449646, "global_step": 88146, "epoch": 990, "lr": 7.632816052824725e-05}
{"train_loss": 0.1443980485200882, "global_step": 88147, "epoch": 990, "lr": 7.63276676715318e-05}
{"train_loss": 0.23768658936023712, "global_step": 88148, "epoch": 990, "lr": 7.632717481127691e-05}
{"train_loss": 0.12466983497142792, "global_step": 88149, "epoch": 990, "lr": 7.632668194748267e-05}
{"train_loss": 0.16935187578201294, "global_step": 88150, "epoch": 990, "lr": 7.632618908014912e-05}
{"train_loss": 0.2710426449775696, "global_step": 88151, "epoch": 990, "lr": 7.632569620927634e-05}
{"train_loss": 0.1519402414560318, "global_step": 88152, "epoch": 990, "lr": 7.632520333486439e-05}
{"train_loss": 0.10193664580583572, "global_step": 88153, "epoch": 990, "lr": 7.632471045691333e-05}
{"train_loss": 0.13505734503269196, "global_step": 88154, "epoch": 990, "lr": 7.632421757542325e-05}
{"train_loss": 0.14127306640148163, "global_step": 88155, "epoch": 990, "lr": 7.632372469039418e-05}
{"train_loss": 0.1835128366947174, "global_step": 88156, "epoch": 990, "lr": 7.632323180182621e-05}
{"train_loss": 0.15538319945335388, "global_step": 88157, "epoch": 990, "lr": 7.632273890971941e-05}
{"train_loss": 0.1946660727262497, "global_step": 88158, "epoch": 990, "lr": 7.632224601407382e-05}
{"train_loss": 0.1876784712076187, "global_step": 88159, "epoch": 990, "lr": 7.632175311488955e-05}
{"train_loss": 0.15975332260131836, "global_step": 88160, "epoch": 990, "lr": 7.632126021216664e-05}
{"train_loss": 0.1706257462501526, "global_step": 88161, "epoch": 990, "lr": 7.632076730590515e-05}
{"train_loss": 0.17022372782230377, "global_step": 88162, "epoch": 990, "lr": 7.632027439610515e-05}
{"train_loss": 0.1264619678258896, "global_step": 88163, "epoch": 990, "lr": 7.631978148276672e-05}
{"train_loss": 0.20499487221240997, "global_step": 88164, "epoch": 990, "lr": 7.63192885658899e-05}
{"train_loss": 0.1349729597568512, "global_step": 88165, "epoch": 990, "lr": 7.631879564547478e-05}
{"train_loss": 0.09121108055114746, "global_step": 88166, "epoch": 990, "lr": 7.631830272152143e-05}
{"train_loss": 0.08382059633731842, "global_step": 88167, "epoch": 990, "lr": 7.63178097940299e-05}
{"train_loss": 0.2501649558544159, "global_step": 88168, "epoch": 990, "lr": 7.631731686300024e-05}
{"train_loss": 0.1180858314037323, "global_step": 88169, "epoch": 990, "lr": 7.631682392843255e-05}
{"train_loss": 0.09754297137260437, "global_step": 88170, "epoch": 990, "lr": 7.63163309903269e-05}
{"train_loss": 0.18958531320095062, "global_step": 88171, "epoch": 990, "lr": 7.631583804868332e-05}
{"train_loss": 0.15564262866973877, "global_step": 88172, "epoch": 990, "lr": 7.631534510350192e-05}
{"train_loss": 0.1934426873922348, "global_step": 88173, "epoch": 990, "lr": 7.63148521547827e-05}
{"train_loss": 0.245578795671463, "global_step": 88174, "epoch": 990, "lr": 7.631435920252581e-05}
{"train_loss": 0.1140332892537117, "global_step": 88175, "epoch": 990, "lr": 7.631386624673126e-05}
{"train_loss": 0.18318621814250946, "global_step": 88176, "epoch": 990, "lr": 7.631337328739912e-05}
{"train_loss": 0.21997228264808655, "global_step": 88177, "epoch": 990, "lr": 7.631288032452948e-05}
{"train_loss": 0.16769033670425415, "global_step": 88178, "epoch": 990, "lr": 7.631238735812238e-05}
{"train_loss": 0.18797530233860016, "global_step": 88179, "epoch": 990, "lr": 7.631189438817793e-05}
{"train_loss": 0.18413902819156647, "global_step": 88180, "epoch": 990, "lr": 7.631140141469614e-05}
{"train_loss": 0.24678510427474976, "global_step": 88181, "epoch": 990, "lr": 7.631090843767713e-05}
{"train_loss": 0.25324976444244385, "global_step": 88182, "epoch": 990, "lr": 7.631041545712091e-05}
{"train_loss": 0.15529833734035492, "global_step": 88183, "epoch": 990, "lr": 7.630992247302759e-05}
{"train_loss": 0.1959012895822525, "global_step": 88184, "epoch": 990, "lr": 7.630942948539723e-05}
{"train_loss": 0.1849019080400467, "global_step": 88185, "epoch": 990, "lr": 7.630893649422988e-05}
{"train_loss": 0.18375539779663086, "global_step": 88186, "epoch": 990, "lr": 7.63084434995256e-05}
{"train_loss": 0.18626222014427185, "global_step": 88187, "epoch": 990, "lr": 7.630795050128448e-05}
{"train_loss": 0.1451123058795929, "global_step": 88188, "epoch": 990, "lr": 7.630745749950658e-05}
{"train_loss": 0.0942174643278122, "global_step": 88189, "epoch": 990, "lr": 7.630696449419196e-05}
{"train_loss": 0.17048010230064392, "global_step": 88190, "epoch": 990, "lr": 7.630647148534069e-05}
{"train_loss": 0.14115147292613983, "global_step": 88191, "epoch": 990, "lr": 7.630597847295284e-05}
{"train_loss": 0.16890035569667816, "global_step": 88192, "epoch": 990, "lr": 7.630548545702847e-05}
{"train_loss": 0.18311886489391327, "global_step": 88193, "epoch": 990, "lr": 7.630499243756764e-05}
{"train_loss": 0.17843639850616455, "global_step": 88194, "epoch": 990, "lr": 7.630449941457041e-05}
{"train_loss": 0.1795380860567093, "global_step": 88195, "epoch": 990, "lr": 7.63040063880369e-05}
{"train_loss": 0.17770271003246307, "global_step": 88196, "epoch": 990, "lr": 7.63035133579671e-05}
{"train_loss": 0.16138403117656708, "global_step": 88197, "epoch": 990, "lr": 7.630302032436114e-05}
{"train_loss": 0.17711056567979663, "global_step": 88198, "epoch": 990, "lr": 7.630252728721904e-05, "val_loss": 4.280978679656982, "train_action_mse_error": 32.38896179199219}
{"train_loss": 0.09677591919898987, "global_step": 88199, "epoch": 991, "lr": 7.630203424654092e-05}
{"train_loss": 0.15884673595428467, "global_step": 88200, "epoch": 991, "lr": 7.630154120232678e-05}
{"train_loss": 0.15538136661052704, "global_step": 88201, "epoch": 991, "lr": 7.630104815457672e-05}
{"train_loss": 0.21266613900661469, "global_step": 88202, "epoch": 991, "lr": 7.630055510329082e-05}
{"train_loss": 0.21167276799678802, "global_step": 88203, "epoch": 991, "lr": 7.630006204846912e-05}
{"train_loss": 0.2171201854944229, "global_step": 88204, "epoch": 991, "lr": 7.629956899011171e-05}
{"train_loss": 0.1570974886417389, "global_step": 88205, "epoch": 991, "lr": 7.629907592821862e-05}
{"train_loss": 0.20327962934970856, "global_step": 88206, "epoch": 991, "lr": 7.629858286278994e-05}
{"train_loss": 0.11980362236499786, "global_step": 88207, "epoch": 991, "lr": 7.629808979382576e-05}
{"train_loss": 0.1674068123102188, "global_step": 88208, "epoch": 991, "lr": 7.629759672132611e-05}
{"train_loss": 0.1995837390422821, "global_step": 88209, "epoch": 991, "lr": 7.629710364529109e-05}
{"train_loss": 0.17010577023029327, "global_step": 88210, "epoch": 991, "lr": 7.62966105657207e-05}
{"train_loss": 0.13498003780841827, "global_step": 88211, "epoch": 991, "lr": 7.629611748261508e-05}
{"train_loss": 0.16143979132175446, "global_step": 88212, "epoch": 991, "lr": 7.629562439597427e-05}
{"train_loss": 0.21252186596393585, "global_step": 88213, "epoch": 991, "lr": 7.629513130579831e-05}
{"train_loss": 0.13506187498569489, "global_step": 88214, "epoch": 991, "lr": 7.629463821208732e-05}
{"train_loss": 0.27727368474006653, "global_step": 88215, "epoch": 991, "lr": 7.629414511484132e-05}
{"train_loss": 0.1485881358385086, "global_step": 88216, "epoch": 991, "lr": 7.629365201406039e-05}
{"train_loss": 0.18596075475215912, "global_step": 88217, "epoch": 991, "lr": 7.62931589097446e-05}
{"train_loss": 0.14331938326358795, "global_step": 88218, "epoch": 991, "lr": 7.629266580189404e-05}
{"train_loss": 0.1494741439819336, "global_step": 88219, "epoch": 991, "lr": 7.629217269050872e-05}
{"train_loss": 0.15613551437854767, "global_step": 88220, "epoch": 991, "lr": 7.629167957558876e-05}
{"train_loss": 0.18953131139278412, "global_step": 88221, "epoch": 991, "lr": 7.629118645713419e-05}
{"train_loss": 0.18128705024719238, "global_step": 88222, "epoch": 991, "lr": 7.629069333514509e-05}
{"train_loss": 0.17844974994659424, "global_step": 88223, "epoch": 991, "lr": 7.629020020962155e-05}
{"train_loss": 0.15192130208015442, "global_step": 88224, "epoch": 991, "lr": 7.628970708056358e-05}
{"train_loss": 0.15029366314411163, "global_step": 88225, "epoch": 991, "lr": 7.628921394797129e-05}
{"train_loss": 0.18224801123142242, "global_step": 88226, "epoch": 991, "lr": 7.628872081184475e-05}
{"train_loss": 0.16654403507709503, "global_step": 88227, "epoch": 991, "lr": 7.6288227672184e-05}
{"train_loss": 0.19134758412837982, "global_step": 88228, "epoch": 991, "lr": 7.628773452898912e-05}
{"train_loss": 0.11643372476100922, "global_step": 88229, "epoch": 991, "lr": 7.628724138226016e-05}
{"train_loss": 0.19762197136878967, "global_step": 88230, "epoch": 991, "lr": 7.628674823199722e-05}
{"train_loss": 0.15675988793373108, "global_step": 88231, "epoch": 991, "lr": 7.628625507820034e-05}
{"train_loss": 0.21514391899108887, "global_step": 88232, "epoch": 991, "lr": 7.628576192086958e-05}
{"train_loss": 0.2222687304019928, "global_step": 88233, "epoch": 991, "lr": 7.628526876000503e-05}
{"train_loss": 0.2034791111946106, "global_step": 88234, "epoch": 991, "lr": 7.628477559560676e-05}
{"train_loss": 0.16967269778251648, "global_step": 88235, "epoch": 991, "lr": 7.628428242767481e-05}
{"train_loss": 0.12328307330608368, "global_step": 88236, "epoch": 991, "lr": 7.628378925620927e-05}
{"train_loss": 0.30735069513320923, "global_step": 88237, "epoch": 991, "lr": 7.628329608121018e-05}
{"train_loss": 0.1180666908621788, "global_step": 88238, "epoch": 991, "lr": 7.628280290267764e-05}
{"train_loss": 0.17616429924964905, "global_step": 88239, "epoch": 991, "lr": 7.628230972061168e-05}
{"train_loss": 0.18879149854183197, "global_step": 88240, "epoch": 991, "lr": 7.628181653501239e-05}
{"train_loss": 0.14369060099124908, "global_step": 88241, "epoch": 991, "lr": 7.628132334587983e-05}
{"train_loss": 0.20350825786590576, "global_step": 88242, "epoch": 991, "lr": 7.628083015321407e-05}
{"train_loss": 0.20179720222949982, "global_step": 88243, "epoch": 991, "lr": 7.628033695701517e-05}
{"train_loss": 0.15602295100688934, "global_step": 88244, "epoch": 991, "lr": 7.627984375728319e-05}
{"train_loss": 0.15759927034378052, "global_step": 88245, "epoch": 991, "lr": 7.627935055401822e-05}
{"train_loss": 0.2067607045173645, "global_step": 88246, "epoch": 991, "lr": 7.62788573472203e-05}
{"train_loss": 0.1941138058900833, "global_step": 88247, "epoch": 991, "lr": 7.627836413688951e-05}
{"train_loss": 0.20804835855960846, "global_step": 88248, "epoch": 991, "lr": 7.627787092302592e-05}
{"train_loss": 0.2407674342393875, "global_step": 88249, "epoch": 991, "lr": 7.627737770562956e-05}
{"train_loss": 0.1399664580821991, "global_step": 88250, "epoch": 991, "lr": 7.627688448470057e-05}
{"train_loss": 0.14107348024845123, "global_step": 88251, "epoch": 991, "lr": 7.627639126023896e-05}
{"train_loss": 0.20634430646896362, "global_step": 88252, "epoch": 991, "lr": 7.62758980322448e-05}
{"train_loss": 0.13659369945526123, "global_step": 88253, "epoch": 991, "lr": 7.627540480071817e-05}
{"train_loss": 0.2244051694869995, "global_step": 88254, "epoch": 991, "lr": 7.627491156565913e-05}
{"train_loss": 0.1843363642692566, "global_step": 88255, "epoch": 991, "lr": 7.627441832706775e-05}
{"train_loss": 0.12300604581832886, "global_step": 88256, "epoch": 991, "lr": 7.62739250849441e-05}
{"train_loss": 0.20986932516098022, "global_step": 88257, "epoch": 991, "lr": 7.627343183928825e-05}
{"train_loss": 0.13192076981067657, "global_step": 88258, "epoch": 991, "lr": 7.627293859010024e-05}
{"train_loss": 0.1994277387857437, "global_step": 88259, "epoch": 991, "lr": 7.627244533738016e-05}
{"train_loss": 0.19064906239509583, "global_step": 88260, "epoch": 991, "lr": 7.627195208112808e-05}
{"train_loss": 0.23854964971542358, "global_step": 88261, "epoch": 991, "lr": 7.627145882134404e-05}
{"train_loss": 0.13618580996990204, "global_step": 88262, "epoch": 991, "lr": 7.627096555802813e-05}
{"train_loss": 0.184615358710289, "global_step": 88263, "epoch": 991, "lr": 7.627047229118041e-05}
{"train_loss": 0.29869669675827026, "global_step": 88264, "epoch": 991, "lr": 7.626997902080094e-05}
{"train_loss": 0.16451001167297363, "global_step": 88265, "epoch": 991, "lr": 7.626948574688978e-05}
{"train_loss": 0.23420390486717224, "global_step": 88266, "epoch": 991, "lr": 7.626899246944703e-05}
{"train_loss": 0.15504154562950134, "global_step": 88267, "epoch": 991, "lr": 7.626849918847273e-05}
{"train_loss": 0.1447814702987671, "global_step": 88268, "epoch": 991, "lr": 7.626800590396693e-05}
{"train_loss": 0.20880183577537537, "global_step": 88269, "epoch": 991, "lr": 7.626751261592973e-05}
{"train_loss": 0.20925191044807434, "global_step": 88270, "epoch": 991, "lr": 7.626701932436118e-05}
{"train_loss": 0.1962125152349472, "global_step": 88271, "epoch": 991, "lr": 7.626652602926136e-05}
{"train_loss": 0.2036045640707016, "global_step": 88272, "epoch": 991, "lr": 7.626603273063032e-05}
{"train_loss": 0.22844453155994415, "global_step": 88273, "epoch": 991, "lr": 7.626553942846814e-05}
{"train_loss": 0.13162057101726532, "global_step": 88274, "epoch": 991, "lr": 7.626504612277486e-05}
{"train_loss": 0.11232037097215652, "global_step": 88275, "epoch": 991, "lr": 7.626455281355058e-05}
{"train_loss": 0.120905801653862, "global_step": 88276, "epoch": 991, "lr": 7.626405950079536e-05}
{"train_loss": 0.176008939743042, "global_step": 88277, "epoch": 991, "lr": 7.626356618450924e-05}
{"train_loss": 0.1956964135169983, "global_step": 88278, "epoch": 991, "lr": 7.62630728646923e-05}
{"train_loss": 0.20013552904129028, "global_step": 88279, "epoch": 991, "lr": 7.626257954134462e-05}
{"train_loss": 0.18783646821975708, "global_step": 88280, "epoch": 991, "lr": 7.626208621446627e-05}
{"train_loss": 0.16943863034248352, "global_step": 88281, "epoch": 991, "lr": 7.626159288405727e-05}
{"train_loss": 0.25682345032691956, "global_step": 88282, "epoch": 991, "lr": 7.626109955011774e-05}
{"train_loss": 0.14663997292518616, "global_step": 88283, "epoch": 991, "lr": 7.626060621264773e-05}
{"train_loss": 0.13294553756713867, "global_step": 88284, "epoch": 991, "lr": 7.626011287164728e-05}
{"train_loss": 0.3045627176761627, "global_step": 88285, "epoch": 991, "lr": 7.625961952711649e-05}
{"train_loss": 0.19916518032550812, "global_step": 88286, "epoch": 991, "lr": 7.625912617905541e-05}
{"train_loss": 0.18246277178941148, "global_step": 88287, "epoch": 991, "lr": 7.625863282746411e-05, "val_loss": 4.251069068908691}
{"train_loss": 0.17786194384098053, "global_step": 88288, "epoch": 992, "lr": 7.625813947234267e-05}
{"train_loss": 0.17786608636379242, "global_step": 88289, "epoch": 992, "lr": 7.625764611369113e-05}
{"train_loss": 0.18117809295654297, "global_step": 88290, "epoch": 992, "lr": 7.625715275150957e-05}
{"train_loss": 0.153606578707695, "global_step": 88291, "epoch": 992, "lr": 7.625665938579806e-05}
{"train_loss": 0.18118618428707123, "global_step": 88292, "epoch": 992, "lr": 7.625616601655667e-05}
{"train_loss": 0.15829738974571228, "global_step": 88293, "epoch": 992, "lr": 7.625567264378545e-05}
{"train_loss": 0.18455973267555237, "global_step": 88294, "epoch": 992, "lr": 7.625517926748446e-05}
{"train_loss": 0.34391602873802185, "global_step": 88295, "epoch": 992, "lr": 7.625468588765382e-05}
{"train_loss": 0.15806105732917786, "global_step": 88296, "epoch": 992, "lr": 7.625419250429353e-05}
{"train_loss": 0.23031270503997803, "global_step": 88297, "epoch": 992, "lr": 7.62536991174037e-05}
{"train_loss": 0.20687733590602875, "global_step": 88298, "epoch": 992, "lr": 7.625320572698436e-05}
{"train_loss": 0.17799857258796692, "global_step": 88299, "epoch": 992, "lr": 7.62527123330356e-05}
{"train_loss": 0.25561079382896423, "global_step": 88300, "epoch": 992, "lr": 7.62522189355575e-05}
{"train_loss": 0.06598842889070511, "global_step": 88301, "epoch": 992, "lr": 7.62517255345501e-05}
{"train_loss": 0.22984246909618378, "global_step": 88302, "epoch": 992, "lr": 7.625123213001347e-05}
{"train_loss": 0.13439595699310303, "global_step": 88303, "epoch": 992, "lr": 7.625073872194766e-05}
{"train_loss": 0.1082846000790596, "global_step": 88304, "epoch": 992, "lr": 7.625024531035278e-05}
{"train_loss": 0.23620903491973877, "global_step": 88305, "epoch": 992, "lr": 7.624975189522888e-05}
{"train_loss": 0.2201806753873825, "global_step": 88306, "epoch": 992, "lr": 7.624925847657602e-05}
{"train_loss": 0.12716451287269592, "global_step": 88307, "epoch": 992, "lr": 7.624876505439427e-05}
{"train_loss": 0.19258050620555878, "global_step": 88308, "epoch": 992, "lr": 7.624827162868369e-05}
{"train_loss": 0.25472137331962585, "global_step": 88309, "epoch": 992, "lr": 7.624777819944434e-05}
{"train_loss": 0.13937214016914368, "global_step": 88310, "epoch": 992, "lr": 7.624728476667631e-05}
{"train_loss": 0.22872278094291687, "global_step": 88311, "epoch": 992, "lr": 7.624679133037964e-05}
{"train_loss": 0.15465766191482544, "global_step": 88312, "epoch": 992, "lr": 7.624629789055442e-05}
{"train_loss": 0.22417022287845612, "global_step": 88313, "epoch": 992, "lr": 7.624580444720069e-05}
{"train_loss": 0.14662116765975952, "global_step": 88314, "epoch": 992, "lr": 7.624531100031856e-05}
{"train_loss": 0.26696956157684326, "global_step": 88315, "epoch": 992, "lr": 7.624481754990805e-05}
{"train_loss": 0.2825157046318054, "global_step": 88316, "epoch": 992, "lr": 7.624432409596924e-05}
{"train_loss": 0.23957930505275726, "global_step": 88317, "epoch": 992, "lr": 7.62438306385022e-05}
{"train_loss": 0.19382934272289276, "global_step": 88318, "epoch": 992, "lr": 7.6243337177507e-05}
{"train_loss": 0.2115691751241684, "global_step": 88319, "epoch": 992, "lr": 7.624284371298371e-05}
{"train_loss": 0.22002935409545898, "global_step": 88320, "epoch": 992, "lr": 7.624235024493239e-05}
{"train_loss": 0.14583493769168854, "global_step": 88321, "epoch": 992, "lr": 7.624185677335309e-05}
{"train_loss": 0.08482979238033295, "global_step": 88322, "epoch": 992, "lr": 7.624136329824589e-05}
{"train_loss": 0.15124867856502533, "global_step": 88323, "epoch": 992, "lr": 7.624086981961087e-05}
{"train_loss": 0.11384060233831406, "global_step": 88324, "epoch": 992, "lr": 7.624037633744809e-05}
{"train_loss": 0.23790809512138367, "global_step": 88325, "epoch": 992, "lr": 7.62398828517576e-05}
{"train_loss": 0.18405240774154663, "global_step": 88326, "epoch": 992, "lr": 7.62393893625395e-05}
{"train_loss": 0.11919756233692169, "global_step": 88327, "epoch": 992, "lr": 7.62388958697938e-05}
{"train_loss": 0.15595927834510803, "global_step": 88328, "epoch": 992, "lr": 7.623840237352062e-05}
{"train_loss": 0.13096654415130615, "global_step": 88329, "epoch": 992, "lr": 7.623790887372003e-05}
{"train_loss": 0.19920483231544495, "global_step": 88330, "epoch": 992, "lr": 7.623741537039203e-05}
{"train_loss": 0.2341298758983612, "global_step": 88331, "epoch": 992, "lr": 7.623692186353675e-05}
{"train_loss": 0.2099292278289795, "global_step": 88332, "epoch": 992, "lr": 7.623642835315424e-05}
{"train_loss": 0.11626330018043518, "global_step": 88333, "epoch": 992, "lr": 7.623593483924454e-05}
{"train_loss": 0.16650938987731934, "global_step": 88334, "epoch": 992, "lr": 7.623544132180777e-05}
{"train_loss": 0.24224556982517242, "global_step": 88335, "epoch": 992, "lr": 7.623494780084394e-05}
{"train_loss": 0.29465731978416443, "global_step": 88336, "epoch": 992, "lr": 7.623445427635316e-05}
{"train_loss": 0.15750347077846527, "global_step": 88337, "epoch": 992, "lr": 7.623396074833546e-05}
{"train_loss": 0.2840489149093628, "global_step": 88338, "epoch": 992, "lr": 7.623346721679093e-05}
{"train_loss": 0.210277259349823, "global_step": 88339, "epoch": 992, "lr": 7.623297368171963e-05}
{"train_loss": 0.1464322954416275, "global_step": 88340, "epoch": 992, "lr": 7.623248014312162e-05}
{"train_loss": 0.16041482985019684, "global_step": 88341, "epoch": 992, "lr": 7.623198660099698e-05}
{"train_loss": 0.21880462765693665, "global_step": 88342, "epoch": 992, "lr": 7.623149305534576e-05}
{"train_loss": 0.12285265326499939, "global_step": 88343, "epoch": 992, "lr": 7.623099950616805e-05}
{"train_loss": 0.15964044630527496, "global_step": 88344, "epoch": 992, "lr": 7.623050595346389e-05}
{"train_loss": 0.15155892074108124, "global_step": 88345, "epoch": 992, "lr": 7.623001239723338e-05}
{"train_loss": 0.28113025426864624, "global_step": 88346, "epoch": 992, "lr": 7.622951883747653e-05}
{"train_loss": 0.22981826961040497, "global_step": 88347, "epoch": 992, "lr": 7.622902527419347e-05}
{"train_loss": 0.1720641851425171, "global_step": 88348, "epoch": 992, "lr": 7.622853170738422e-05}
{"train_loss": 0.21861578524112701, "global_step": 88349, "epoch": 992, "lr": 7.622803813704887e-05}
{"train_loss": 0.16606707870960236, "global_step": 88350, "epoch": 992, "lr": 7.622754456318749e-05}
{"train_loss": 0.15694470703601837, "global_step": 88351, "epoch": 992, "lr": 7.622705098580012e-05}
{"train_loss": 0.18204225599765778, "global_step": 88352, "epoch": 992, "lr": 7.622655740488684e-05}
{"train_loss": 0.13571806252002716, "global_step": 88353, "epoch": 992, "lr": 7.622606382044774e-05}
{"train_loss": 0.19769248366355896, "global_step": 88354, "epoch": 992, "lr": 7.622557023248285e-05}
{"train_loss": 0.17156842350959778, "global_step": 88355, "epoch": 992, "lr": 7.622507664099225e-05}
{"train_loss": 0.22338241338729858, "global_step": 88356, "epoch": 992, "lr": 7.622458304597602e-05}
{"train_loss": 0.19480571150779724, "global_step": 88357, "epoch": 992, "lr": 7.622408944743422e-05}
{"train_loss": 0.13274416327476501, "global_step": 88358, "epoch": 992, "lr": 7.622359584536688e-05}
{"train_loss": 0.23548570275306702, "global_step": 88359, "epoch": 992, "lr": 7.622310223977412e-05}
{"train_loss": 0.2082660049200058, "global_step": 88360, "epoch": 992, "lr": 7.622260863065598e-05}
{"train_loss": 0.19643068313598633, "global_step": 88361, "epoch": 992, "lr": 7.622211501801253e-05}
{"train_loss": 0.12940095365047455, "global_step": 88362, "epoch": 992, "lr": 7.622162140184384e-05}
{"train_loss": 0.2031211256980896, "global_step": 88363, "epoch": 992, "lr": 7.622112778214996e-05}
{"train_loss": 0.18584603071212769, "global_step": 88364, "epoch": 992, "lr": 7.622063415893097e-05}
{"train_loss": 0.1664917916059494, "global_step": 88365, "epoch": 992, "lr": 7.622014053218695e-05}
{"train_loss": 0.17593808472156525, "global_step": 88366, "epoch": 992, "lr": 7.621964690191793e-05}
{"train_loss": 0.11258518695831299, "global_step": 88367, "epoch": 992, "lr": 7.621915326812402e-05}
{"train_loss": 0.19644178450107574, "global_step": 88368, "epoch": 992, "lr": 7.621865963080525e-05}
{"train_loss": 0.09816224128007889, "global_step": 88369, "epoch": 992, "lr": 7.621816598996172e-05}
{"train_loss": 0.19075535237789154, "global_step": 88370, "epoch": 992, "lr": 7.621767234559346e-05}
{"train_loss": 0.17924486100673676, "global_step": 88371, "epoch": 992, "lr": 7.621717869770055e-05}
{"train_loss": 0.23877963423728943, "global_step": 88372, "epoch": 992, "lr": 7.621668504628308e-05}
{"train_loss": 0.16382896900177002, "global_step": 88373, "epoch": 992, "lr": 7.621619139134108e-05}
{"train_loss": 0.20017071068286896, "global_step": 88374, "epoch": 992, "lr": 7.621569773287465e-05}
{"train_loss": 0.17189718782901764, "global_step": 88375, "epoch": 992, "lr": 7.621520407088382e-05}
{"train_loss": 0.18645325169134674, "global_step": 88376, "epoch": 992, "lr": 7.621471040536867e-05, "val_loss": 4.215169429779053}
{"train_loss": 0.2198689579963684, "global_step": 88377, "epoch": 993, "lr": 7.621421673632928e-05}
{"train_loss": 0.1425800770521164, "global_step": 88378, "epoch": 993, "lr": 7.62137230637657e-05}
{"train_loss": 0.20338399708271027, "global_step": 88379, "epoch": 993, "lr": 7.621322938767803e-05}
{"train_loss": 0.16428305208683014, "global_step": 88380, "epoch": 993, "lr": 7.621273570806628e-05}
{"train_loss": 0.12554408609867096, "global_step": 88381, "epoch": 993, "lr": 7.621224202493058e-05}
{"train_loss": 0.08091361820697784, "global_step": 88382, "epoch": 993, "lr": 7.621174833827094e-05}
{"train_loss": 0.2736417055130005, "global_step": 88383, "epoch": 993, "lr": 7.621125464808745e-05}
{"train_loss": 0.21324461698532104, "global_step": 88384, "epoch": 993, "lr": 7.621076095438019e-05}
{"train_loss": 0.3011331260204315, "global_step": 88385, "epoch": 993, "lr": 7.621026725714921e-05}
{"train_loss": 0.1299639642238617, "global_step": 88386, "epoch": 993, "lr": 7.620977355639457e-05}
{"train_loss": 0.13625353574752808, "global_step": 88387, "epoch": 993, "lr": 7.620927985211635e-05}
{"train_loss": 0.18628613650798798, "global_step": 88388, "epoch": 993, "lr": 7.620878614431462e-05}
{"train_loss": 0.20798498392105103, "global_step": 88389, "epoch": 993, "lr": 7.620829243298942e-05}
{"train_loss": 0.12997491657733917, "global_step": 88390, "epoch": 993, "lr": 7.620779871814085e-05}
{"train_loss": 0.18697910010814667, "global_step": 88391, "epoch": 993, "lr": 7.620730499976894e-05}
{"train_loss": 0.239299938082695, "global_step": 88392, "epoch": 993, "lr": 7.620681127787381e-05}
{"train_loss": 0.1560307741165161, "global_step": 88393, "epoch": 993, "lr": 7.620631755245548e-05}
{"train_loss": 0.2006712555885315, "global_step": 88394, "epoch": 993, "lr": 7.620582382351401e-05}
{"train_loss": 0.18389186263084412, "global_step": 88395, "epoch": 993, "lr": 7.620533009104951e-05}
{"train_loss": 0.24301232397556305, "global_step": 88396, "epoch": 993, "lr": 7.620483635506202e-05}
{"train_loss": 0.24576936662197113, "global_step": 88397, "epoch": 993, "lr": 7.620434261555159e-05}
{"train_loss": 0.10889896005392075, "global_step": 88398, "epoch": 993, "lr": 7.620384887251833e-05}
{"train_loss": 0.08515570312738419, "global_step": 88399, "epoch": 993, "lr": 7.620335512596226e-05}
{"train_loss": 0.19784504175186157, "global_step": 88400, "epoch": 993, "lr": 7.620286137588349e-05}
{"train_loss": 0.10866648703813553, "global_step": 88401, "epoch": 993, "lr": 7.620236762228206e-05}
{"train_loss": 0.19967235624790192, "global_step": 88402, "epoch": 993, "lr": 7.620187386515802e-05}
{"train_loss": 0.18478016555309296, "global_step": 88403, "epoch": 993, "lr": 7.620138010451148e-05}
{"train_loss": 0.26885855197906494, "global_step": 88404, "epoch": 993, "lr": 7.620088634034248e-05}
{"train_loss": 0.1839965134859085, "global_step": 88405, "epoch": 993, "lr": 7.620039257265109e-05}
{"train_loss": 0.1826099008321762, "global_step": 88406, "epoch": 993, "lr": 7.619989880143736e-05}
{"train_loss": 0.18665637075901031, "global_step": 88407, "epoch": 993, "lr": 7.61994050267014e-05}
{"train_loss": 0.18592262268066406, "global_step": 88408, "epoch": 993, "lr": 7.619891124844323e-05}
{"train_loss": 0.17513170838356018, "global_step": 88409, "epoch": 993, "lr": 7.619841746666294e-05}
{"train_loss": 0.23847991228103638, "global_step": 88410, "epoch": 993, "lr": 7.619792368136058e-05}
{"train_loss": 0.21335570514202118, "global_step": 88411, "epoch": 993, "lr": 7.619742989253627e-05}
{"train_loss": 0.11917545646429062, "global_step": 88412, "epoch": 993, "lr": 7.619693610018999e-05}
{"train_loss": 0.2623509168624878, "global_step": 88413, "epoch": 993, "lr": 7.619644230432187e-05}
{"train_loss": 0.20181412994861603, "global_step": 88414, "epoch": 993, "lr": 7.619594850493194e-05}
{"train_loss": 0.16878394782543182, "global_step": 88415, "epoch": 993, "lr": 7.619545470202032e-05}
{"train_loss": 0.13279011845588684, "global_step": 88416, "epoch": 993, "lr": 7.619496089558703e-05}
{"train_loss": 0.22363443672657013, "global_step": 88417, "epoch": 993, "lr": 7.619446708563212e-05}
{"train_loss": 0.17692720890045166, "global_step": 88418, "epoch": 993, "lr": 7.619397327215569e-05}
{"train_loss": 0.16636520624160767, "global_step": 88419, "epoch": 993, "lr": 7.619347945515782e-05}
{"train_loss": 0.18407942354679108, "global_step": 88420, "epoch": 993, "lr": 7.619298563463854e-05}
{"train_loss": 0.16137245297431946, "global_step": 88421, "epoch": 993, "lr": 7.619249181059795e-05}
{"train_loss": 0.2672429084777832, "global_step": 88422, "epoch": 993, "lr": 7.619199798303608e-05}
{"train_loss": 0.20384611189365387, "global_step": 88423, "epoch": 993, "lr": 7.619150415195302e-05}
{"train_loss": 0.22124595940113068, "global_step": 88424, "epoch": 993, "lr": 7.619101031734883e-05}
{"train_loss": 0.17398130893707275, "global_step": 88425, "epoch": 993, "lr": 7.619051647922358e-05}
{"train_loss": 0.1923440396785736, "global_step": 88426, "epoch": 993, "lr": 7.619002263757735e-05}
{"train_loss": 0.12395471334457397, "global_step": 88427, "epoch": 993, "lr": 7.618952879241017e-05}
{"train_loss": 0.13657301664352417, "global_step": 88428, "epoch": 993, "lr": 7.618903494372214e-05}
{"train_loss": 0.18885573744773865, "global_step": 88429, "epoch": 993, "lr": 7.61885410915133e-05}
{"train_loss": 0.12055543065071106, "global_step": 88430, "epoch": 993, "lr": 7.618804723578374e-05}
{"train_loss": 0.2402815967798233, "global_step": 88431, "epoch": 993, "lr": 7.61875533765335e-05}
{"train_loss": 0.18185290694236755, "global_step": 88432, "epoch": 993, "lr": 7.618705951376269e-05}
{"train_loss": 0.19150644540786743, "global_step": 88433, "epoch": 993, "lr": 7.618656564747133e-05}
{"train_loss": 0.15346640348434448, "global_step": 88434, "epoch": 993, "lr": 7.61860717776595e-05}
{"train_loss": 0.14394672214984894, "global_step": 88435, "epoch": 993, "lr": 7.618557790432729e-05}
{"train_loss": 0.1390225738286972, "global_step": 88436, "epoch": 993, "lr": 7.618508402747473e-05}
{"train_loss": 0.15203431248664856, "global_step": 88437, "epoch": 993, "lr": 7.618459014710192e-05}
{"train_loss": 0.11001662909984589, "global_step": 88438, "epoch": 993, "lr": 7.61840962632089e-05}
{"train_loss": 0.222361221909523, "global_step": 88439, "epoch": 993, "lr": 7.618360237579575e-05}
{"train_loss": 0.22606800496578217, "global_step": 88440, "epoch": 993, "lr": 7.618310848486254e-05}
{"train_loss": 0.12086717784404755, "global_step": 88441, "epoch": 993, "lr": 7.618261459040932e-05}
{"train_loss": 0.19475369155406952, "global_step": 88442, "epoch": 993, "lr": 7.618212069243617e-05}
{"train_loss": 0.12555642426013947, "global_step": 88443, "epoch": 993, "lr": 7.618162679094317e-05}
{"train_loss": 0.2746261954307556, "global_step": 88444, "epoch": 993, "lr": 7.618113288593036e-05}
{"train_loss": 0.17901282012462616, "global_step": 88445, "epoch": 993, "lr": 7.61806389773978e-05}
{"train_loss": 0.17964273691177368, "global_step": 88446, "epoch": 993, "lr": 7.618014506534559e-05}
{"train_loss": 0.22588381171226501, "global_step": 88447, "epoch": 993, "lr": 7.617965114977377e-05}
{"train_loss": 0.1870950311422348, "global_step": 88448, "epoch": 993, "lr": 7.617915723068242e-05}
{"train_loss": 0.29421335458755493, "global_step": 88449, "epoch": 993, "lr": 7.61786633080716e-05}
{"train_loss": 0.14843638241291046, "global_step": 88450, "epoch": 993, "lr": 7.617816938194138e-05}
{"train_loss": 0.27642619609832764, "global_step": 88451, "epoch": 993, "lr": 7.617767545229182e-05}
{"train_loss": 0.207710400223732, "global_step": 88452, "epoch": 993, "lr": 7.617718151912299e-05}
{"train_loss": 0.1759381890296936, "global_step": 88453, "epoch": 993, "lr": 7.617668758243494e-05}
{"train_loss": 0.20532894134521484, "global_step": 88454, "epoch": 993, "lr": 7.617619364222779e-05}
{"train_loss": 0.27091771364212036, "global_step": 88455, "epoch": 993, "lr": 7.617569969850154e-05}
{"train_loss": 0.1458812803030014, "global_step": 88456, "epoch": 993, "lr": 7.61752057512563e-05}
{"train_loss": 0.13122500479221344, "global_step": 88457, "epoch": 993, "lr": 7.617471180049213e-05}
{"train_loss": 0.20874838531017303, "global_step": 88458, "epoch": 993, "lr": 7.617421784620907e-05}
{"train_loss": 0.1711788773536682, "global_step": 88459, "epoch": 993, "lr": 7.617372388840723e-05}
{"train_loss": 0.13316521048545837, "global_step": 88460, "epoch": 993, "lr": 7.617322992708662e-05}
{"train_loss": 0.11727011203765869, "global_step": 88461, "epoch": 993, "lr": 7.617273596224736e-05}
{"train_loss": 0.21200278401374817, "global_step": 88462, "epoch": 993, "lr": 7.617224199388949e-05}
{"train_loss": 0.1780310869216919, "global_step": 88463, "epoch": 993, "lr": 7.617174802201308e-05}
{"train_loss": 0.11919788271188736, "global_step": 88464, "epoch": 993, "lr": 7.61712540466182e-05}
{"train_loss": 0.18359465671054434, "global_step": 88465, "epoch": 993, "lr": 7.617076006770491e-05, "val_loss": 4.199999809265137}
{"train_loss": 0.1873476654291153, "global_step": 88466, "epoch": 994, "lr": 7.617026608527328e-05}
{"train_loss": 0.15268996357917786, "global_step": 88467, "epoch": 994, "lr": 7.616977209932338e-05}
{"train_loss": 0.14830201864242554, "global_step": 88468, "epoch": 994, "lr": 7.616927810985528e-05}
{"train_loss": 0.22297880053520203, "global_step": 88469, "epoch": 994, "lr": 7.616878411686903e-05}
{"train_loss": 0.15425431728363037, "global_step": 88470, "epoch": 994, "lr": 7.616829012036471e-05}
{"train_loss": 0.1594001054763794, "global_step": 88471, "epoch": 994, "lr": 7.616779612034237e-05}
{"train_loss": 0.14547842741012573, "global_step": 88472, "epoch": 994, "lr": 7.61673021168021e-05}
{"train_loss": 0.2141488939523697, "global_step": 88473, "epoch": 994, "lr": 7.616680810974395e-05}
{"train_loss": 0.13936010003089905, "global_step": 88474, "epoch": 994, "lr": 7.6166314099168e-05}
{"train_loss": 0.13854408264160156, "global_step": 88475, "epoch": 994, "lr": 7.61658200850743e-05}
{"train_loss": 0.08867144584655762, "global_step": 88476, "epoch": 994, "lr": 7.616532606746293e-05}
{"train_loss": 0.26075828075408936, "global_step": 88477, "epoch": 994, "lr": 7.616483204633394e-05}
{"train_loss": 0.20770852267742157, "global_step": 88478, "epoch": 994, "lr": 7.616433802168742e-05}
{"train_loss": 0.18228588998317719, "global_step": 88479, "epoch": 994, "lr": 7.616384399352342e-05}
{"train_loss": 0.2368117719888687, "global_step": 88480, "epoch": 994, "lr": 7.616334996184201e-05}
{"train_loss": 0.201563760638237, "global_step": 88481, "epoch": 994, "lr": 7.616285592664325e-05}
{"train_loss": 0.26990026235580444, "global_step": 88482, "epoch": 994, "lr": 7.616236188792722e-05}
{"train_loss": 0.14381729066371918, "global_step": 88483, "epoch": 994, "lr": 7.6161867845694e-05}
{"train_loss": 0.15798227488994598, "global_step": 88484, "epoch": 994, "lr": 7.616137379994359e-05}
{"train_loss": 0.13820095360279083, "global_step": 88485, "epoch": 994, "lr": 7.616087975067613e-05}
{"train_loss": 0.17438022792339325, "global_step": 88486, "epoch": 994, "lr": 7.616038569789165e-05}
{"train_loss": 0.20478138327598572, "global_step": 88487, "epoch": 994, "lr": 7.615989164159023e-05}
{"train_loss": 0.1292957067489624, "global_step": 88488, "epoch": 994, "lr": 7.615939758177194e-05}
{"train_loss": 0.19664157927036285, "global_step": 88489, "epoch": 994, "lr": 7.615890351843682e-05}
{"train_loss": 0.17543689906597137, "global_step": 88490, "epoch": 994, "lr": 7.615840945158497e-05}
{"train_loss": 0.14905421435832977, "global_step": 88491, "epoch": 994, "lr": 7.615791538121643e-05}
{"train_loss": 0.15743669867515564, "global_step": 88492, "epoch": 994, "lr": 7.615742130733128e-05}
{"train_loss": 0.14444702863693237, "global_step": 88493, "epoch": 994, "lr": 7.615692722992959e-05}
{"train_loss": 0.14770829677581787, "global_step": 88494, "epoch": 994, "lr": 7.615643314901142e-05}
{"train_loss": 0.15117371082305908, "global_step": 88495, "epoch": 994, "lr": 7.615593906457682e-05}
{"train_loss": 0.19878211617469788, "global_step": 88496, "epoch": 994, "lr": 7.615544497662588e-05}
{"train_loss": 0.1855606585741043, "global_step": 88497, "epoch": 994, "lr": 7.615495088515867e-05}
{"train_loss": 0.12648522853851318, "global_step": 88498, "epoch": 994, "lr": 7.615445679017523e-05}
{"train_loss": 0.143358513712883, "global_step": 88499, "epoch": 994, "lr": 7.615396269167565e-05}
{"train_loss": 0.1812945157289505, "global_step": 88500, "epoch": 994, "lr": 7.615346858966e-05}
{"train_loss": 0.16005723178386688, "global_step": 88501, "epoch": 994, "lr": 7.615297448412833e-05}
{"train_loss": 0.13150128722190857, "global_step": 88502, "epoch": 994, "lr": 7.61524803750807e-05}
{"train_loss": 0.14455145597457886, "global_step": 88503, "epoch": 994, "lr": 7.615198626251718e-05}
{"train_loss": 0.1603943407535553, "global_step": 88504, "epoch": 994, "lr": 7.615149214643787e-05}
{"train_loss": 0.1511949747800827, "global_step": 88505, "epoch": 994, "lr": 7.615099802684279e-05}
{"train_loss": 0.21462863683700562, "global_step": 88506, "epoch": 994, "lr": 7.615050390373204e-05}
{"train_loss": 0.2025524079799652, "global_step": 88507, "epoch": 994, "lr": 7.615000977710567e-05}
{"train_loss": 0.22889851033687592, "global_step": 88508, "epoch": 994, "lr": 7.614951564696375e-05}
{"train_loss": 0.20891940593719482, "global_step": 88509, "epoch": 994, "lr": 7.614902151330635e-05}
{"train_loss": 0.1594012975692749, "global_step": 88510, "epoch": 994, "lr": 7.614852737613352e-05}
{"train_loss": 0.19417734444141388, "global_step": 88511, "epoch": 994, "lr": 7.614803323544534e-05}
{"train_loss": 0.20007096230983734, "global_step": 88512, "epoch": 994, "lr": 7.614753909124188e-05}
{"train_loss": 0.1265537440776825, "global_step": 88513, "epoch": 994, "lr": 7.61470449435232e-05}
{"train_loss": 0.15881970524787903, "global_step": 88514, "epoch": 994, "lr": 7.614655079228938e-05}
{"train_loss": 0.17629067599773407, "global_step": 88515, "epoch": 994, "lr": 7.614605663754047e-05}
{"train_loss": 0.1752091348171234, "global_step": 88516, "epoch": 994, "lr": 7.614556247927653e-05}
{"train_loss": 0.1376335173845291, "global_step": 88517, "epoch": 994, "lr": 7.614506831749765e-05}
{"train_loss": 0.2669958472251892, "global_step": 88518, "epoch": 994, "lr": 7.614457415220388e-05}
{"train_loss": 0.08136457204818726, "global_step": 88519, "epoch": 994, "lr": 7.61440799833953e-05}
{"train_loss": 0.1743301898241043, "global_step": 88520, "epoch": 994, "lr": 7.614358581107195e-05}
{"train_loss": 0.20470526814460754, "global_step": 88521, "epoch": 994, "lr": 7.614309163523393e-05}
{"train_loss": 0.22570349276065826, "global_step": 88522, "epoch": 994, "lr": 7.614259745588128e-05}
{"train_loss": 0.15489499270915985, "global_step": 88523, "epoch": 994, "lr": 7.614210327301407e-05}
{"train_loss": 0.15410877764225006, "global_step": 88524, "epoch": 994, "lr": 7.61416090866324e-05}
{"train_loss": 0.20644977688789368, "global_step": 88525, "epoch": 994, "lr": 7.614111489673628e-05}
{"train_loss": 0.18274490535259247, "global_step": 88526, "epoch": 994, "lr": 7.614062070332581e-05}
{"train_loss": 0.15464402735233307, "global_step": 88527, "epoch": 994, "lr": 7.614012650640106e-05}
{"train_loss": 0.21415694057941437, "global_step": 88528, "epoch": 994, "lr": 7.613963230596209e-05}
{"train_loss": 0.17107252776622772, "global_step": 88529, "epoch": 994, "lr": 7.613913810200897e-05}
{"train_loss": 0.13182775676250458, "global_step": 88530, "epoch": 994, "lr": 7.613864389454175e-05}
{"train_loss": 0.12421462684869766, "global_step": 88531, "epoch": 994, "lr": 7.61381496835605e-05}
{"train_loss": 0.17457982897758484, "global_step": 88532, "epoch": 994, "lr": 7.61376554690653e-05}
{"train_loss": 0.21700678765773773, "global_step": 88533, "epoch": 994, "lr": 7.613716125105623e-05}
{"train_loss": 0.14841149747371674, "global_step": 88534, "epoch": 994, "lr": 7.613666702953332e-05}
{"train_loss": 0.13304805755615234, "global_step": 88535, "epoch": 994, "lr": 7.613617280449665e-05}
{"train_loss": 0.19837047159671783, "global_step": 88536, "epoch": 994, "lr": 7.61356785759463e-05}
{"train_loss": 0.1534997820854187, "global_step": 88537, "epoch": 994, "lr": 7.613518434388234e-05}
{"train_loss": 0.15320836007595062, "global_step": 88538, "epoch": 994, "lr": 7.613469010830481e-05}
{"train_loss": 0.1559906154870987, "global_step": 88539, "epoch": 994, "lr": 7.613419586921377e-05}
{"train_loss": 0.1755335032939911, "global_step": 88540, "epoch": 994, "lr": 7.613370162660932e-05}
{"train_loss": 0.13592229783535004, "global_step": 88541, "epoch": 994, "lr": 7.613320738049152e-05}
{"train_loss": 0.3014404773712158, "global_step": 88542, "epoch": 994, "lr": 7.613271313086043e-05}
{"train_loss": 0.15978039801120758, "global_step": 88543, "epoch": 994, "lr": 7.61322188777161e-05}
{"train_loss": 0.14806418120861053, "global_step": 88544, "epoch": 994, "lr": 7.613172462105863e-05}
{"train_loss": 0.15394799411296844, "global_step": 88545, "epoch": 994, "lr": 7.613123036088805e-05}
{"train_loss": 0.15745560824871063, "global_step": 88546, "epoch": 994, "lr": 7.613073609720446e-05}
{"train_loss": 0.2552538812160492, "global_step": 88547, "epoch": 994, "lr": 7.61302418300079e-05}
{"train_loss": 0.12505193054676056, "global_step": 88548, "epoch": 994, "lr": 7.612974755929844e-05}
{"train_loss": 0.22923168540000916, "global_step": 88549, "epoch": 994, "lr": 7.612925328507618e-05}
{"train_loss": 0.21221724152565002, "global_step": 88550, "epoch": 994, "lr": 7.612875900734115e-05}
{"train_loss": 0.09531062096357346, "global_step": 88551, "epoch": 994, "lr": 7.612826472609342e-05}
{"train_loss": 0.14690545201301575, "global_step": 88552, "epoch": 994, "lr": 7.612777044133307e-05}
{"train_loss": 0.24727404117584229, "global_step": 88553, "epoch": 994, "lr": 7.612727615306015e-05}
{"train_loss": 0.1744022307436118, "global_step": 88554, "epoch": 994, "lr": 7.612678186127475e-05, "val_loss": 4.3728108406066895}
{"train_loss": 0.2618396282196045, "global_step": 88555, "epoch": 995, "lr": 7.61262875659769e-05}
{"train_loss": 0.1651923656463623, "global_step": 88556, "epoch": 995, "lr": 7.612579326716671e-05}
{"train_loss": 0.16145141422748566, "global_step": 88557, "epoch": 995, "lr": 7.612529896484422e-05}
{"train_loss": 0.12380658090114594, "global_step": 88558, "epoch": 995, "lr": 7.61248046590095e-05}
{"train_loss": 0.16354990005493164, "global_step": 88559, "epoch": 995, "lr": 7.612431034966261e-05}
{"train_loss": 0.17144303023815155, "global_step": 88560, "epoch": 995, "lr": 7.612381603680363e-05}
{"train_loss": 0.15695902705192566, "global_step": 88561, "epoch": 995, "lr": 7.612332172043263e-05}
{"train_loss": 0.2083941400051117, "global_step": 88562, "epoch": 995, "lr": 7.612282740054967e-05}
{"train_loss": 0.1875665783882141, "global_step": 88563, "epoch": 995, "lr": 7.612233307715482e-05}
{"train_loss": 0.25258901715278625, "global_step": 88564, "epoch": 995, "lr": 7.612183875024811e-05}
{"train_loss": 0.22447583079338074, "global_step": 88565, "epoch": 995, "lr": 7.612134441982968e-05}
{"train_loss": 0.11746257543563843, "global_step": 88566, "epoch": 995, "lr": 7.612085008589952e-05}
{"train_loss": 0.07726132869720459, "global_step": 88567, "epoch": 995, "lr": 7.612035574845773e-05}
{"train_loss": 0.19328773021697998, "global_step": 88568, "epoch": 995, "lr": 7.61198614075044e-05}
{"train_loss": 0.17163346707820892, "global_step": 88569, "epoch": 995, "lr": 7.611936706303954e-05}
{"train_loss": 0.12236932665109634, "global_step": 88570, "epoch": 995, "lr": 7.611887271506328e-05}
{"train_loss": 0.21653316915035248, "global_step": 88571, "epoch": 995, "lr": 7.611837836357565e-05}
{"train_loss": 0.27437129616737366, "global_step": 88572, "epoch": 995, "lr": 7.611788400857671e-05}
{"train_loss": 0.17616711556911469, "global_step": 88573, "epoch": 995, "lr": 7.611738965006656e-05}
{"train_loss": 0.15685978531837463, "global_step": 88574, "epoch": 995, "lr": 7.611689528804523e-05}
{"train_loss": 0.17389988899230957, "global_step": 88575, "epoch": 995, "lr": 7.611640092251277e-05}
{"train_loss": 0.16767750680446625, "global_step": 88576, "epoch": 995, "lr": 7.611590655346932e-05}
{"train_loss": 0.15323379635810852, "global_step": 88577, "epoch": 995, "lr": 7.61154121809149e-05}
{"train_loss": 0.19246932864189148, "global_step": 88578, "epoch": 995, "lr": 7.611491780484959e-05}
{"train_loss": 0.1706586480140686, "global_step": 88579, "epoch": 995, "lr": 7.611442342527342e-05}
{"train_loss": 0.21520578861236572, "global_step": 88580, "epoch": 995, "lr": 7.61139290421865e-05}
{"train_loss": 0.16002970933914185, "global_step": 88581, "epoch": 995, "lr": 7.611343465558889e-05}
{"train_loss": 0.16833677887916565, "global_step": 88582, "epoch": 995, "lr": 7.611294026548063e-05}
{"train_loss": 0.17081037163734436, "global_step": 88583, "epoch": 995, "lr": 7.611244587186182e-05}
{"train_loss": 0.14473825693130493, "global_step": 88584, "epoch": 995, "lr": 7.611195147473249e-05}
{"train_loss": 0.15760233998298645, "global_step": 88585, "epoch": 995, "lr": 7.611145707409274e-05}
{"train_loss": 0.16907498240470886, "global_step": 88586, "epoch": 995, "lr": 7.611096266994262e-05}
{"train_loss": 0.17543210089206696, "global_step": 88587, "epoch": 995, "lr": 7.611046826228218e-05}
{"train_loss": 0.10585608333349228, "global_step": 88588, "epoch": 995, "lr": 7.610997385111154e-05}
{"train_loss": 0.17047706246376038, "global_step": 88589, "epoch": 995, "lr": 7.610947943643071e-05}
{"train_loss": 0.17423073947429657, "global_step": 88590, "epoch": 995, "lr": 7.610898501823978e-05}
{"train_loss": 0.16173872351646423, "global_step": 88591, "epoch": 995, "lr": 7.610849059653883e-05}
{"train_loss": 0.21622556447982788, "global_step": 88592, "epoch": 995, "lr": 7.61079961713279e-05}
{"train_loss": 0.195669025182724, "global_step": 88593, "epoch": 995, "lr": 7.610750174260707e-05}
{"train_loss": 0.21157242357730865, "global_step": 88594, "epoch": 995, "lr": 7.610700731037639e-05}
{"train_loss": 0.12193694710731506, "global_step": 88595, "epoch": 995, "lr": 7.610651287463596e-05}
{"train_loss": 0.1284656524658203, "global_step": 88596, "epoch": 995, "lr": 7.610601843538583e-05}
{"train_loss": 0.10446781665086746, "global_step": 88597, "epoch": 995, "lr": 7.610552399262605e-05}
{"train_loss": 0.22664019465446472, "global_step": 88598, "epoch": 995, "lr": 7.61050295463567e-05}
{"train_loss": 0.13933251798152924, "global_step": 88599, "epoch": 995, "lr": 7.610453509657786e-05}
{"train_loss": 0.12885229289531708, "global_step": 88600, "epoch": 995, "lr": 7.610404064328958e-05}
{"train_loss": 0.12707969546318054, "global_step": 88601, "epoch": 995, "lr": 7.610354618649194e-05}
{"train_loss": 0.18459951877593994, "global_step": 88602, "epoch": 995, "lr": 7.610305172618497e-05}
{"train_loss": 0.10900359600782394, "global_step": 88603, "epoch": 995, "lr": 7.610255726236879e-05}
{"train_loss": 0.16670656204223633, "global_step": 88604, "epoch": 995, "lr": 7.610206279504341e-05}
{"train_loss": 0.12418345361948013, "global_step": 88605, "epoch": 995, "lr": 7.610156832420895e-05}
{"train_loss": 0.15192961692810059, "global_step": 88606, "epoch": 995, "lr": 7.610107384986544e-05}
{"train_loss": 0.12419102340936661, "global_step": 88607, "epoch": 995, "lr": 7.610057937201296e-05}
{"train_loss": 0.13161222636699677, "global_step": 88608, "epoch": 995, "lr": 7.610008489065159e-05}
{"train_loss": 0.16444392502307892, "global_step": 88609, "epoch": 995, "lr": 7.609959040578136e-05}
{"train_loss": 0.13256683945655823, "global_step": 88610, "epoch": 995, "lr": 7.609909591740238e-05}
{"train_loss": 0.20532575249671936, "global_step": 88611, "epoch": 995, "lr": 7.609860142551468e-05}
{"train_loss": 0.23911699652671814, "global_step": 88612, "epoch": 995, "lr": 7.609810693011833e-05}
{"train_loss": 0.18290258944034576, "global_step": 88613, "epoch": 995, "lr": 7.609761243121342e-05}
{"train_loss": 0.17763151228427887, "global_step": 88614, "epoch": 995, "lr": 7.609711792880001e-05}
{"train_loss": 0.14254696667194366, "global_step": 88615, "epoch": 995, "lr": 7.609662342287815e-05}
{"train_loss": 0.15672621130943298, "global_step": 88616, "epoch": 995, "lr": 7.609612891344793e-05}
{"train_loss": 0.1538623571395874, "global_step": 88617, "epoch": 995, "lr": 7.60956344005094e-05}
{"train_loss": 0.15926775336265564, "global_step": 88618, "epoch": 995, "lr": 7.609513988406262e-05}
{"train_loss": 0.12978337705135345, "global_step": 88619, "epoch": 995, "lr": 7.609464536410769e-05}
{"train_loss": 0.14077040553092957, "global_step": 88620, "epoch": 995, "lr": 7.609415084064463e-05}
{"train_loss": 0.2574557662010193, "global_step": 88621, "epoch": 995, "lr": 7.609365631367354e-05}
{"train_loss": 0.15352606773376465, "global_step": 88622, "epoch": 995, "lr": 7.609316178319447e-05}
{"train_loss": 0.1666373610496521, "global_step": 88623, "epoch": 995, "lr": 7.60926672492075e-05}
{"train_loss": 0.15990810096263885, "global_step": 88624, "epoch": 995, "lr": 7.609217271171268e-05}
{"train_loss": 0.28915244340896606, "global_step": 88625, "epoch": 995, "lr": 7.609167817071007e-05}
{"train_loss": 0.2903870642185211, "global_step": 88626, "epoch": 995, "lr": 7.609118362619978e-05}
{"train_loss": 0.20643499493598938, "global_step": 88627, "epoch": 995, "lr": 7.609068907818184e-05}
{"train_loss": 0.2436157912015915, "global_step": 88628, "epoch": 995, "lr": 7.609019452665632e-05}
{"train_loss": 0.13076695799827576, "global_step": 88629, "epoch": 995, "lr": 7.60896999716233e-05}
{"train_loss": 0.25467389822006226, "global_step": 88630, "epoch": 995, "lr": 7.608920541308282e-05}
{"train_loss": 0.09815190732479095, "global_step": 88631, "epoch": 995, "lr": 7.608871085103497e-05}
{"train_loss": 0.16518250107765198, "global_step": 88632, "epoch": 995, "lr": 7.608821628547982e-05}
{"train_loss": 0.16182994842529297, "global_step": 88633, "epoch": 995, "lr": 7.608772171641743e-05}
{"train_loss": 0.22216850519180298, "global_step": 88634, "epoch": 995, "lr": 7.608722714384786e-05}
{"train_loss": 0.17560657858848572, "global_step": 88635, "epoch": 995, "lr": 7.608673256777118e-05}
{"train_loss": 0.2347617745399475, "global_step": 88636, "epoch": 995, "lr": 7.608623798818746e-05}
{"train_loss": 0.21928799152374268, "global_step": 88637, "epoch": 995, "lr": 7.608574340509676e-05}
{"train_loss": 0.14822109043598175, "global_step": 88638, "epoch": 995, "lr": 7.608524881849914e-05}
{"train_loss": 0.1672680377960205, "global_step": 88639, "epoch": 995, "lr": 7.608475422839469e-05}
{"train_loss": 0.1712992936372757, "global_step": 88640, "epoch": 995, "lr": 7.608425963478347e-05}
{"train_loss": 0.18809643387794495, "global_step": 88641, "epoch": 995, "lr": 7.608376503766552e-05}
{"train_loss": 0.26243793964385986, "global_step": 88642, "epoch": 995, "lr": 7.608327043704094e-05}
{"train_loss": 0.17493281572052602, "global_step": 88643, "epoch": 995, "lr": 7.608277583290976e-05, "val_loss": 4.445579528808594, "train_action_mse_error": 28.561519622802734}
{"train_loss": 0.22578126192092896, "global_step": 88644, "epoch": 996, "lr": 7.60822812252721e-05}
{"train_loss": 0.18054507672786713, "global_step": 88645, "epoch": 996, "lr": 7.608178661412797e-05}
{"train_loss": 0.20483317971229553, "global_step": 88646, "epoch": 996, "lr": 7.608129199947747e-05}
{"train_loss": 0.17558541893959045, "global_step": 88647, "epoch": 996, "lr": 7.608079738132066e-05}
{"train_loss": 0.24307219684123993, "global_step": 88648, "epoch": 996, "lr": 7.60803027596576e-05}
{"train_loss": 0.1548130214214325, "global_step": 88649, "epoch": 996, "lr": 7.607980813448837e-05}
{"train_loss": 0.2028627097606659, "global_step": 88650, "epoch": 996, "lr": 7.607931350581301e-05}
{"train_loss": 0.22209706902503967, "global_step": 88651, "epoch": 996, "lr": 7.607881887363162e-05}
{"train_loss": 0.19450737535953522, "global_step": 88652, "epoch": 996, "lr": 7.607832423794425e-05}
{"train_loss": 0.2537674307823181, "global_step": 88653, "epoch": 996, "lr": 7.607782959875097e-05}
{"train_loss": 0.14408178627490997, "global_step": 88654, "epoch": 996, "lr": 7.607733495605182e-05}
{"train_loss": 0.18462777137756348, "global_step": 88655, "epoch": 996, "lr": 7.607684030984691e-05}
{"train_loss": 0.2077074646949768, "global_step": 88656, "epoch": 996, "lr": 7.607634566013628e-05}
{"train_loss": 0.1329188346862793, "global_step": 88657, "epoch": 996, "lr": 7.607585100692001e-05}
{"train_loss": 0.17306514084339142, "global_step": 88658, "epoch": 996, "lr": 7.607535635019818e-05}
{"train_loss": 0.1686352789402008, "global_step": 88659, "epoch": 996, "lr": 7.60748616899708e-05}
{"train_loss": 0.22683613002300262, "global_step": 88660, "epoch": 996, "lr": 7.607436702623799e-05}
{"train_loss": 0.23399774730205536, "global_step": 88661, "epoch": 996, "lr": 7.60738723589998e-05}
{"train_loss": 0.2564627230167389, "global_step": 88662, "epoch": 996, "lr": 7.607337768825628e-05}
{"train_loss": 0.2233031988143921, "global_step": 88663, "epoch": 996, "lr": 7.607288301400752e-05}
{"train_loss": 0.13816066086292267, "global_step": 88664, "epoch": 996, "lr": 7.607238833625359e-05}
{"train_loss": 0.1897878646850586, "global_step": 88665, "epoch": 996, "lr": 7.607189365499454e-05}
{"train_loss": 0.17788371443748474, "global_step": 88666, "epoch": 996, "lr": 7.607139897023044e-05}
{"train_loss": 0.20308272540569305, "global_step": 88667, "epoch": 996, "lr": 7.607090428196136e-05}
{"train_loss": 0.15745258331298828, "global_step": 88668, "epoch": 996, "lr": 7.607040959018734e-05}
{"train_loss": 0.1407298445701599, "global_step": 88669, "epoch": 996, "lr": 7.60699148949085e-05}
{"train_loss": 0.21342992782592773, "global_step": 88670, "epoch": 996, "lr": 7.606942019612486e-05}
{"train_loss": 0.13403618335723877, "global_step": 88671, "epoch": 996, "lr": 7.606892549383653e-05}
{"train_loss": 0.17691902816295624, "global_step": 88672, "epoch": 996, "lr": 7.606843078804353e-05}
{"train_loss": 0.13581696152687073, "global_step": 88673, "epoch": 996, "lr": 7.606793607874595e-05}
{"train_loss": 0.19944429397583008, "global_step": 88674, "epoch": 996, "lr": 7.606744136594386e-05}
{"train_loss": 0.1424447000026703, "global_step": 88675, "epoch": 996, "lr": 7.606694664963731e-05}
{"train_loss": 0.19188183546066284, "global_step": 88676, "epoch": 996, "lr": 7.606645192982639e-05}
{"train_loss": 0.19811519980430603, "global_step": 88677, "epoch": 996, "lr": 7.606595720651115e-05}
{"train_loss": 0.13294649124145508, "global_step": 88678, "epoch": 996, "lr": 7.606546247969167e-05}
{"train_loss": 0.2083846926689148, "global_step": 88679, "epoch": 996, "lr": 7.606496774936798e-05}
{"train_loss": 0.22747263312339783, "global_step": 88680, "epoch": 996, "lr": 7.606447301554018e-05}
{"train_loss": 0.19488316774368286, "global_step": 88681, "epoch": 996, "lr": 7.606397827820835e-05}
{"train_loss": 0.2032778114080429, "global_step": 88682, "epoch": 996, "lr": 7.606348353737252e-05}
{"train_loss": 0.14652283489704132, "global_step": 88683, "epoch": 996, "lr": 7.606298879303277e-05}
{"train_loss": 0.17860309779644012, "global_step": 88684, "epoch": 996, "lr": 7.606249404518915e-05}
{"train_loss": 0.185001939535141, "global_step": 88685, "epoch": 996, "lr": 7.606199929384177e-05}
{"train_loss": 0.21433238685131073, "global_step": 88686, "epoch": 996, "lr": 7.606150453899067e-05}
{"train_loss": 0.1621229648590088, "global_step": 88687, "epoch": 996, "lr": 7.606100978063591e-05}
{"train_loss": 0.23209068179130554, "global_step": 88688, "epoch": 996, "lr": 7.606051501877758e-05}
{"train_loss": 0.16405123472213745, "global_step": 88689, "epoch": 996, "lr": 7.606002025341573e-05}
{"train_loss": 0.20098808407783508, "global_step": 88690, "epoch": 996, "lr": 7.605952548455042e-05}
{"train_loss": 0.15773266553878784, "global_step": 88691, "epoch": 996, "lr": 7.605903071218172e-05}
{"train_loss": 0.27350056171417236, "global_step": 88692, "epoch": 996, "lr": 7.60585359363097e-05}
{"train_loss": 0.1929677277803421, "global_step": 88693, "epoch": 996, "lr": 7.605804115693445e-05}
{"train_loss": 0.09069081395864487, "global_step": 88694, "epoch": 996, "lr": 7.605754637405599e-05}
{"train_loss": 0.20557792484760284, "global_step": 88695, "epoch": 996, "lr": 7.605705158767442e-05}
{"train_loss": 0.20177367329597473, "global_step": 88696, "epoch": 996, "lr": 7.60565567977898e-05}
{"train_loss": 0.2551037669181824, "global_step": 88697, "epoch": 996, "lr": 7.605606200440219e-05}
{"train_loss": 0.15718437731266022, "global_step": 88698, "epoch": 996, "lr": 7.605556720751166e-05}
{"train_loss": 0.3117130994796753, "global_step": 88699, "epoch": 996, "lr": 7.605507240711827e-05}
{"train_loss": 0.14554595947265625, "global_step": 88700, "epoch": 996, "lr": 7.60545776032221e-05}
{"train_loss": 0.17699278891086578, "global_step": 88701, "epoch": 996, "lr": 7.605408279582322e-05}
{"train_loss": 0.15919943153858185, "global_step": 88702, "epoch": 996, "lr": 7.605358798492166e-05}
{"train_loss": 0.1641673743724823, "global_step": 88703, "epoch": 996, "lr": 7.605309317051752e-05}
{"train_loss": 0.19129882752895355, "global_step": 88704, "epoch": 996, "lr": 7.605259835261088e-05}
{"train_loss": 0.12445315718650818, "global_step": 88705, "epoch": 996, "lr": 7.605210353120176e-05}
{"train_loss": 0.10915321111679077, "global_step": 88706, "epoch": 996, "lr": 7.605160870629027e-05}
{"train_loss": 0.18693791329860687, "global_step": 88707, "epoch": 996, "lr": 7.605111387787644e-05}
{"train_loss": 0.16295620799064636, "global_step": 88708, "epoch": 996, "lr": 7.605061904596039e-05}
{"train_loss": 0.19574250280857086, "global_step": 88709, "epoch": 996, "lr": 7.605012421054211e-05}
{"train_loss": 0.19570744037628174, "global_step": 88710, "epoch": 996, "lr": 7.604962937162173e-05}
{"train_loss": 0.14553780853748322, "global_step": 88711, "epoch": 996, "lr": 7.60491345291993e-05}
{"train_loss": 0.11736442893743515, "global_step": 88712, "epoch": 996, "lr": 7.604863968327487e-05}
{"train_loss": 0.15119695663452148, "global_step": 88713, "epoch": 996, "lr": 7.604814483384853e-05}
{"train_loss": 0.20019343495368958, "global_step": 88714, "epoch": 996, "lr": 7.604764998092031e-05}
{"train_loss": 0.23717330396175385, "global_step": 88715, "epoch": 996, "lr": 7.604715512449032e-05}
{"train_loss": 0.1749831736087799, "global_step": 88716, "epoch": 996, "lr": 7.604666026455863e-05}
{"train_loss": 0.127282053232193, "global_step": 88717, "epoch": 996, "lr": 7.604616540112525e-05}
{"train_loss": 0.1368650496006012, "global_step": 88718, "epoch": 996, "lr": 7.60456705341903e-05}
{"train_loss": 0.08530641347169876, "global_step": 88719, "epoch": 996, "lr": 7.60451756637538e-05}
{"train_loss": 0.1998412162065506, "global_step": 88720, "epoch": 996, "lr": 7.604468078981586e-05}
{"train_loss": 0.1813727468252182, "global_step": 88721, "epoch": 996, "lr": 7.604418591237652e-05}
{"train_loss": 0.2926364541053772, "global_step": 88722, "epoch": 996, "lr": 7.604369103143587e-05}
{"train_loss": 0.16823799908161163, "global_step": 88723, "epoch": 996, "lr": 7.604319614699396e-05}
{"train_loss": 0.2023642361164093, "global_step": 88724, "epoch": 996, "lr": 7.604270125905087e-05}
{"train_loss": 0.19681692123413086, "global_step": 88725, "epoch": 996, "lr": 7.604220636760664e-05}
{"train_loss": 0.22343972325325012, "global_step": 88726, "epoch": 996, "lr": 7.604171147266134e-05}
{"train_loss": 0.15156862139701843, "global_step": 88727, "epoch": 996, "lr": 7.604121657421508e-05}
{"train_loss": 0.18490323424339294, "global_step": 88728, "epoch": 996, "lr": 7.604072167226788e-05}
{"train_loss": 0.18874776363372803, "global_step": 88729, "epoch": 996, "lr": 7.604022676681984e-05}
{"train_loss": 0.27401551604270935, "global_step": 88730, "epoch": 996, "lr": 7.603973185787099e-05}
{"train_loss": 0.14187441766262054, "global_step": 88731, "epoch": 996, "lr": 7.603923694542141e-05}
{"train_loss": 0.1852560414189703, "global_step": 88732, "epoch": 996, "lr": 7.603874202947119e-05, "val_loss": 4.32149600982666}
{"train_loss": 0.10538925230503082, "global_step": 88733, "epoch": 997, "lr": 7.603824711002037e-05}
{"train_loss": 0.14901103079319, "global_step": 88734, "epoch": 997, "lr": 7.603775218706904e-05}
{"train_loss": 0.1923496127128601, "global_step": 88735, "epoch": 997, "lr": 7.603725726061723e-05}
{"train_loss": 0.14991550147533417, "global_step": 88736, "epoch": 997, "lr": 7.603676233066504e-05}
{"train_loss": 0.21254009008407593, "global_step": 88737, "epoch": 997, "lr": 7.603626739721253e-05}
{"train_loss": 0.24872447550296783, "global_step": 88738, "epoch": 997, "lr": 7.603577246025975e-05}
{"train_loss": 0.14190445840358734, "global_step": 88739, "epoch": 997, "lr": 7.603527751980676e-05}
{"train_loss": 0.16254658997058868, "global_step": 88740, "epoch": 997, "lr": 7.603478257585366e-05}
{"train_loss": 0.16917988657951355, "global_step": 88741, "epoch": 997, "lr": 7.60342876284005e-05}
{"train_loss": 0.1995542198419571, "global_step": 88742, "epoch": 997, "lr": 7.603379267744737e-05}
{"train_loss": 0.19364109635353088, "global_step": 88743, "epoch": 997, "lr": 7.603329772299428e-05}
{"train_loss": 0.14669732749462128, "global_step": 88744, "epoch": 997, "lr": 7.603280276504135e-05}
{"train_loss": 0.1795019805431366, "global_step": 88745, "epoch": 997, "lr": 7.603230780358863e-05}
{"train_loss": 0.185511976480484, "global_step": 88746, "epoch": 997, "lr": 7.603181283863618e-05}
{"train_loss": 0.1559540182352066, "global_step": 88747, "epoch": 997, "lr": 7.603131787018406e-05}
{"train_loss": 0.20394986867904663, "global_step": 88748, "epoch": 997, "lr": 7.603082289823236e-05}
{"train_loss": 0.19157223403453827, "global_step": 88749, "epoch": 997, "lr": 7.603032792278112e-05}
{"train_loss": 0.3046363890171051, "global_step": 88750, "epoch": 997, "lr": 7.602983294383043e-05}
{"train_loss": 0.1526043564081192, "global_step": 88751, "epoch": 997, "lr": 7.602933796138034e-05}
{"train_loss": 0.19257378578186035, "global_step": 88752, "epoch": 997, "lr": 7.602884297543091e-05}
{"train_loss": 0.19373764097690582, "global_step": 88753, "epoch": 997, "lr": 7.602834798598224e-05}
{"train_loss": 0.16474340856075287, "global_step": 88754, "epoch": 997, "lr": 7.602785299303438e-05}
{"train_loss": 0.16256432235240936, "global_step": 88755, "epoch": 997, "lr": 7.602735799658738e-05}
{"train_loss": 0.1648852378129959, "global_step": 88756, "epoch": 997, "lr": 7.602686299664132e-05}
{"train_loss": 0.200410395860672, "global_step": 88757, "epoch": 997, "lr": 7.602636799319627e-05}
{"train_loss": 0.15141525864601135, "global_step": 88758, "epoch": 997, "lr": 7.602587298625228e-05}
{"train_loss": 0.2339605987071991, "global_step": 88759, "epoch": 997, "lr": 7.602537797580944e-05}
{"train_loss": 0.24104632437229156, "global_step": 88760, "epoch": 997, "lr": 7.602488296186781e-05}
{"train_loss": 0.13661141693592072, "global_step": 88761, "epoch": 997, "lr": 7.602438794442744e-05}
{"train_loss": 0.2198762744665146, "global_step": 88762, "epoch": 997, "lr": 7.602389292348842e-05}
{"train_loss": 0.1754607856273651, "global_step": 88763, "epoch": 997, "lr": 7.60233978990508e-05}
{"train_loss": 0.18304206430912018, "global_step": 88764, "epoch": 997, "lr": 7.602290287111465e-05}
{"train_loss": 0.2218274474143982, "global_step": 88765, "epoch": 997, "lr": 7.602240783968002e-05}
{"train_loss": 0.21502482891082764, "global_step": 88766, "epoch": 997, "lr": 7.602191280474701e-05}
{"train_loss": 0.23418954014778137, "global_step": 88767, "epoch": 997, "lr": 7.602141776631568e-05}
{"train_loss": 0.09351996332406998, "global_step": 88768, "epoch": 997, "lr": 7.602092272438608e-05}
{"train_loss": 0.16242995858192444, "global_step": 88769, "epoch": 997, "lr": 7.602042767895827e-05}
{"train_loss": 0.13034187257289886, "global_step": 88770, "epoch": 997, "lr": 7.601993263003236e-05}
{"train_loss": 0.14621250331401825, "global_step": 88771, "epoch": 997, "lr": 7.601943757760836e-05}
{"train_loss": 0.13344506919384003, "global_step": 88772, "epoch": 997, "lr": 7.601894252168638e-05}
{"train_loss": 0.10970351099967957, "global_step": 88773, "epoch": 997, "lr": 7.601844746226647e-05}
{"train_loss": 0.23875652253627777, "global_step": 88774, "epoch": 997, "lr": 7.60179523993487e-05}
{"train_loss": 0.20363734662532806, "global_step": 88775, "epoch": 997, "lr": 7.601745733293311e-05}
{"train_loss": 0.21841977536678314, "global_step": 88776, "epoch": 997, "lr": 7.601696226301981e-05}
{"train_loss": 0.18753284215927124, "global_step": 88777, "epoch": 997, "lr": 7.601646718960884e-05}
{"train_loss": 0.15919414162635803, "global_step": 88778, "epoch": 997, "lr": 7.601597211270028e-05}
{"train_loss": 0.1793379783630371, "global_step": 88779, "epoch": 997, "lr": 7.601547703229418e-05}
{"train_loss": 0.2291228473186493, "global_step": 88780, "epoch": 997, "lr": 7.601498194839064e-05}
{"train_loss": 0.23850445449352264, "global_step": 88781, "epoch": 997, "lr": 7.601448686098967e-05}
{"train_loss": 0.2163345366716385, "global_step": 88782, "epoch": 997, "lr": 7.601399177009139e-05}
{"train_loss": 0.20235829055309296, "global_step": 88783, "epoch": 997, "lr": 7.601349667569583e-05}
{"train_loss": 0.16153758764266968, "global_step": 88784, "epoch": 997, "lr": 7.601300157780309e-05}
{"train_loss": 0.2847760021686554, "global_step": 88785, "epoch": 997, "lr": 7.601250647641321e-05}
{"train_loss": 0.18717198073863983, "global_step": 88786, "epoch": 997, "lr": 7.601201137152626e-05}
{"train_loss": 0.273563951253891, "global_step": 88787, "epoch": 997, "lr": 7.601151626314232e-05}
{"train_loss": 0.15347793698310852, "global_step": 88788, "epoch": 997, "lr": 7.601102115126143e-05}
{"train_loss": 0.16999822854995728, "global_step": 88789, "epoch": 997, "lr": 7.60105260358837e-05}
{"train_loss": 0.20579363405704498, "global_step": 88790, "epoch": 997, "lr": 7.601003091700917e-05}
{"train_loss": 0.14469730854034424, "global_step": 88791, "epoch": 997, "lr": 7.60095357946379e-05}
{"train_loss": 0.1602068394422531, "global_step": 88792, "epoch": 997, "lr": 7.600904066876997e-05}
{"train_loss": 0.15408386290073395, "global_step": 88793, "epoch": 997, "lr": 7.600854553940545e-05}
{"train_loss": 0.1830611675977707, "global_step": 88794, "epoch": 997, "lr": 7.600805040654437e-05}
{"train_loss": 0.13948889076709747, "global_step": 88795, "epoch": 997, "lr": 7.600755527018684e-05}
{"train_loss": 0.15314453840255737, "global_step": 88796, "epoch": 997, "lr": 7.600706013033292e-05}
{"train_loss": 0.22494058310985565, "global_step": 88797, "epoch": 997, "lr": 7.600656498698266e-05}
{"train_loss": 0.10932702571153641, "global_step": 88798, "epoch": 997, "lr": 7.600606984013615e-05}
{"train_loss": 0.12454412132501602, "global_step": 88799, "epoch": 997, "lr": 7.600557468979341e-05}
{"train_loss": 0.22444477677345276, "global_step": 88800, "epoch": 997, "lr": 7.600507953595455e-05}
{"train_loss": 0.16451044380664825, "global_step": 88801, "epoch": 997, "lr": 7.600458437861963e-05}
{"train_loss": 0.17167356610298157, "global_step": 88802, "epoch": 997, "lr": 7.60040892177887e-05}
{"train_loss": 0.23262393474578857, "global_step": 88803, "epoch": 997, "lr": 7.600359405346184e-05}
{"train_loss": 0.11823795735836029, "global_step": 88804, "epoch": 997, "lr": 7.600309888563912e-05}
{"train_loss": 0.20217710733413696, "global_step": 88805, "epoch": 997, "lr": 7.600260371432061e-05}
{"train_loss": 0.12954166531562805, "global_step": 88806, "epoch": 997, "lr": 7.600210853950636e-05}
{"train_loss": 0.19988884031772614, "global_step": 88807, "epoch": 997, "lr": 7.600161336119643e-05}
{"train_loss": 0.17555736005306244, "global_step": 88808, "epoch": 997, "lr": 7.600111817939091e-05}
{"train_loss": 0.18032577633857727, "global_step": 88809, "epoch": 997, "lr": 7.600062299408986e-05}
{"train_loss": 0.12472622096538544, "global_step": 88810, "epoch": 997, "lr": 7.600012780529333e-05}
{"train_loss": 0.1556202471256256, "global_step": 88811, "epoch": 997, "lr": 7.599963261300141e-05}
{"train_loss": 0.19625958800315857, "global_step": 88812, "epoch": 997, "lr": 7.599913741721417e-05}
{"train_loss": 0.11941089481115341, "global_step": 88813, "epoch": 997, "lr": 7.599864221793164e-05}
{"train_loss": 0.15045058727264404, "global_step": 88814, "epoch": 997, "lr": 7.59981470151539e-05}
{"train_loss": 0.22488933801651, "global_step": 88815, "epoch": 997, "lr": 7.599765180888105e-05}
{"train_loss": 0.2571062743663788, "global_step": 88816, "epoch": 997, "lr": 7.599715659911312e-05}
{"train_loss": 0.18727461993694305, "global_step": 88817, "epoch": 997, "lr": 7.59966613858502e-05}
{"train_loss": 0.18140169978141785, "global_step": 88818, "epoch": 997, "lr": 7.599616616909236e-05}
{"train_loss": 0.17969894409179688, "global_step": 88819, "epoch": 997, "lr": 7.599567094883961e-05}
{"train_loss": 0.21157878637313843, "global_step": 88820, "epoch": 997, "lr": 7.599517572509209e-05}
{"train_loss": 0.1826564870523603, "global_step": 88821, "epoch": 997, "lr": 7.599468049784983e-05, "val_loss": 4.223546504974365}
{"train_loss": 0.16609013080596924, "global_step": 88822, "epoch": 998, "lr": 7.59941852671129e-05}
{"train_loss": 0.18653686344623566, "global_step": 88823, "epoch": 998, "lr": 7.599369003288137e-05}
{"train_loss": 0.19202636182308197, "global_step": 88824, "epoch": 998, "lr": 7.59931947951553e-05}
{"train_loss": 0.16012775897979736, "global_step": 88825, "epoch": 998, "lr": 7.599269955393476e-05}
{"train_loss": 0.1711895763874054, "global_step": 88826, "epoch": 998, "lr": 7.599220430921982e-05}
{"train_loss": 0.1375473141670227, "global_step": 88827, "epoch": 998, "lr": 7.599170906101055e-05}
{"train_loss": 0.13477426767349243, "global_step": 88828, "epoch": 998, "lr": 7.599121380930701e-05}
{"train_loss": 0.16595812141895294, "global_step": 88829, "epoch": 998, "lr": 7.599071855410926e-05}
{"train_loss": 0.18783104419708252, "global_step": 88830, "epoch": 998, "lr": 7.599022329541739e-05}
{"train_loss": 0.19797612726688385, "global_step": 88831, "epoch": 998, "lr": 7.598972803323145e-05}
{"train_loss": 0.13485752046108246, "global_step": 88832, "epoch": 998, "lr": 7.59892327675515e-05}
{"train_loss": 0.1729114055633545, "global_step": 88833, "epoch": 998, "lr": 7.59887374983776e-05}
{"train_loss": 0.24763484299182892, "global_step": 88834, "epoch": 998, "lr": 7.598824222570986e-05}
{"train_loss": 0.21010291576385498, "global_step": 88835, "epoch": 998, "lr": 7.59877469495483e-05}
{"train_loss": 0.19336937367916107, "global_step": 88836, "epoch": 998, "lr": 7.598725166989302e-05}
{"train_loss": 0.24892129004001617, "global_step": 88837, "epoch": 998, "lr": 7.598675638674405e-05}
{"train_loss": 0.2720259726047516, "global_step": 88838, "epoch": 998, "lr": 7.598626110010149e-05}
{"train_loss": 0.21180224418640137, "global_step": 88839, "epoch": 998, "lr": 7.59857658099654e-05}
{"train_loss": 0.23167963325977325, "global_step": 88840, "epoch": 998, "lr": 7.598527051633581e-05}
{"train_loss": 0.26376307010650635, "global_step": 88841, "epoch": 998, "lr": 7.598477521921285e-05}
{"train_loss": 0.2599080801010132, "global_step": 88842, "epoch": 998, "lr": 7.598427991859653e-05}
{"train_loss": 0.1919422745704651, "global_step": 88843, "epoch": 998, "lr": 7.598378461448697e-05}
{"train_loss": 0.17445047199726105, "global_step": 88844, "epoch": 998, "lr": 7.598328930688419e-05}
{"train_loss": 0.18227612972259521, "global_step": 88845, "epoch": 998, "lr": 7.598279399578827e-05}
{"train_loss": 0.2062091827392578, "global_step": 88846, "epoch": 998, "lr": 7.598229868119929e-05}
{"train_loss": 0.18945150077342987, "global_step": 88847, "epoch": 998, "lr": 7.598180336311728e-05}
{"train_loss": 0.17803293466567993, "global_step": 88848, "epoch": 998, "lr": 7.598130804154235e-05}
{"train_loss": 0.23092809319496155, "global_step": 88849, "epoch": 998, "lr": 7.598081271647456e-05}
{"train_loss": 0.1525319516658783, "global_step": 88850, "epoch": 998, "lr": 7.598031738791395e-05}
{"train_loss": 0.22273139655590057, "global_step": 88851, "epoch": 998, "lr": 7.597982205586061e-05}
{"train_loss": 0.14838431775569916, "global_step": 88852, "epoch": 998, "lr": 7.59793267203146e-05}
{"train_loss": 0.12444350868463516, "global_step": 88853, "epoch": 998, "lr": 7.5978831381276e-05}
{"train_loss": 0.17401131987571716, "global_step": 88854, "epoch": 998, "lr": 7.597833603874484e-05}
{"train_loss": 0.15300330519676208, "global_step": 88855, "epoch": 998, "lr": 7.597784069272123e-05}
{"train_loss": 0.1559859812259674, "global_step": 88856, "epoch": 998, "lr": 7.59773453432052e-05}
{"train_loss": 0.18739962577819824, "global_step": 88857, "epoch": 998, "lr": 7.597684999019684e-05}
{"train_loss": 0.17067766189575195, "global_step": 88858, "epoch": 998, "lr": 7.597635463369621e-05}
{"train_loss": 0.2546096444129944, "global_step": 88859, "epoch": 998, "lr": 7.597585927370337e-05}
{"train_loss": 0.1957152783870697, "global_step": 88860, "epoch": 998, "lr": 7.59753639102184e-05}
{"train_loss": 0.2210695743560791, "global_step": 88861, "epoch": 998, "lr": 7.597486854324134e-05}
{"train_loss": 0.15746498107910156, "global_step": 88862, "epoch": 998, "lr": 7.59743731727723e-05}
{"train_loss": 0.13281524181365967, "global_step": 88863, "epoch": 998, "lr": 7.597387779881131e-05}
{"train_loss": 0.20362550020217896, "global_step": 88864, "epoch": 998, "lr": 7.597338242135845e-05}
{"train_loss": 0.20144079625606537, "global_step": 88865, "epoch": 998, "lr": 7.597288704041379e-05}
{"train_loss": 0.25691360235214233, "global_step": 88866, "epoch": 998, "lr": 7.597239165597738e-05}
{"train_loss": 0.24520403146743774, "global_step": 88867, "epoch": 998, "lr": 7.59718962680493e-05}
{"train_loss": 0.18765100836753845, "global_step": 88868, "epoch": 998, "lr": 7.597140087662963e-05}
{"train_loss": 0.20731431245803833, "global_step": 88869, "epoch": 998, "lr": 7.59709054817184e-05}
{"train_loss": 0.1399017572402954, "global_step": 88870, "epoch": 998, "lr": 7.597041008331573e-05}
{"train_loss": 0.21486057341098785, "global_step": 88871, "epoch": 998, "lr": 7.596991468142163e-05}
{"train_loss": 0.1347213238477707, "global_step": 88872, "epoch": 998, "lr": 7.59694192760362e-05}
{"train_loss": 0.18941307067871094, "global_step": 88873, "epoch": 998, "lr": 7.59689238671595e-05}
{"train_loss": 0.18763458728790283, "global_step": 88874, "epoch": 998, "lr": 7.596842845479157e-05}
{"train_loss": 0.2507542371749878, "global_step": 88875, "epoch": 998, "lr": 7.596793303893252e-05}
{"train_loss": 0.19729866087436676, "global_step": 88876, "epoch": 998, "lr": 7.59674376195824e-05}
{"train_loss": 0.22791172564029694, "global_step": 88877, "epoch": 998, "lr": 7.596694219674129e-05}
{"train_loss": 0.23104634881019592, "global_step": 88878, "epoch": 998, "lr": 7.596644677040923e-05}
{"train_loss": 0.16090214252471924, "global_step": 88879, "epoch": 998, "lr": 7.596595134058629e-05}
{"train_loss": 0.15706543624401093, "global_step": 88880, "epoch": 998, "lr": 7.596545590727254e-05}
{"train_loss": 0.2542305886745453, "global_step": 88881, "epoch": 998, "lr": 7.596496047046806e-05}
{"train_loss": 0.2483167201280594, "global_step": 88882, "epoch": 998, "lr": 7.59644650301729e-05}
{"train_loss": 0.23964105546474457, "global_step": 88883, "epoch": 998, "lr": 7.596396958638713e-05}
{"train_loss": 0.1603148877620697, "global_step": 88884, "epoch": 998, "lr": 7.596347413911084e-05}
{"train_loss": 0.17277956008911133, "global_step": 88885, "epoch": 998, "lr": 7.596297868834407e-05}
{"train_loss": 0.245022714138031, "global_step": 88886, "epoch": 998, "lr": 7.59624832340869e-05}
{"train_loss": 0.1960495263338089, "global_step": 88887, "epoch": 998, "lr": 7.596198777633939e-05}
{"train_loss": 0.1279251128435135, "global_step": 88888, "epoch": 998, "lr": 7.59614923151016e-05}
{"train_loss": 0.18724264204502106, "global_step": 88889, "epoch": 998, "lr": 7.59609968503736e-05}
{"train_loss": 0.15401200950145721, "global_step": 88890, "epoch": 998, "lr": 7.596050138215545e-05}
{"train_loss": 0.19351807236671448, "global_step": 88891, "epoch": 998, "lr": 7.596000591044725e-05}
{"train_loss": 0.148895263671875, "global_step": 88892, "epoch": 998, "lr": 7.595951043524902e-05}
{"train_loss": 0.18137285113334656, "global_step": 88893, "epoch": 998, "lr": 7.595901495656087e-05}
{"train_loss": 0.1280716061592102, "global_step": 88894, "epoch": 998, "lr": 7.595851947438284e-05}
{"train_loss": 0.24811871349811554, "global_step": 88895, "epoch": 998, "lr": 7.5958023988715e-05}
{"train_loss": 0.23200391232967377, "global_step": 88896, "epoch": 998, "lr": 7.595752849955744e-05}
{"train_loss": 0.11058235168457031, "global_step": 88897, "epoch": 998, "lr": 7.595703300691018e-05}
{"train_loss": 0.1932349056005478, "global_step": 88898, "epoch": 998, "lr": 7.595653751077333e-05}
{"train_loss": 0.1801602840423584, "global_step": 88899, "epoch": 998, "lr": 7.595604201114694e-05}
{"train_loss": 0.15772996842861176, "global_step": 88900, "epoch": 998, "lr": 7.595554650803108e-05}
{"train_loss": 0.26716578006744385, "global_step": 88901, "epoch": 998, "lr": 7.59550510014258e-05}
{"train_loss": 0.21947823464870453, "global_step": 88902, "epoch": 998, "lr": 7.595455549133117e-05}
{"train_loss": 0.13799628615379333, "global_step": 88903, "epoch": 998, "lr": 7.595405997774727e-05}
{"train_loss": 0.15385791659355164, "global_step": 88904, "epoch": 998, "lr": 7.595356446067419e-05}
{"train_loss": 0.16614963114261627, "global_step": 88905, "epoch": 998, "lr": 7.595306894011194e-05}
{"train_loss": 0.20167770981788635, "global_step": 88906, "epoch": 998, "lr": 7.595257341606064e-05}
{"train_loss": 0.21940068900585175, "global_step": 88907, "epoch": 998, "lr": 7.595207788852032e-05}
{"train_loss": 0.23189616203308105, "global_step": 88908, "epoch": 998, "lr": 7.595158235749107e-05}
{"train_loss": 0.18511255085468292, "global_step": 88909, "epoch": 998, "lr": 7.595108682297294e-05}
{"train_loss": 0.191735143993008, "global_step": 88910, "epoch": 998, "lr": 7.595059128496599e-05, "val_loss": 4.266739845275879}
{"train_loss": 0.18661373853683472, "global_step": 88911, "epoch": 999, "lr": 7.595009574347031e-05}
{"train_loss": 0.12526950240135193, "global_step": 88912, "epoch": 999, "lr": 7.594960019848596e-05}
{"train_loss": 0.1438305377960205, "global_step": 88913, "epoch": 999, "lr": 7.5949104650013e-05}
{"train_loss": 0.15444210171699524, "global_step": 88914, "epoch": 999, "lr": 7.59486090980515e-05}
{"train_loss": 0.20612595975399017, "global_step": 88915, "epoch": 999, "lr": 7.594811354260152e-05}
{"train_loss": 0.12622477114200592, "global_step": 88916, "epoch": 999, "lr": 7.594761798366313e-05}
{"train_loss": 0.13090527057647705, "global_step": 88917, "epoch": 999, "lr": 7.594712242123641e-05}
{"train_loss": 0.1623462438583374, "global_step": 88918, "epoch": 999, "lr": 7.59466268553214e-05}
{"train_loss": 0.17605410516262054, "global_step": 88919, "epoch": 999, "lr": 7.59461312859182e-05}
{"train_loss": 0.17519132792949677, "global_step": 88920, "epoch": 999, "lr": 7.594563571302684e-05}
{"train_loss": 0.16965855658054352, "global_step": 88921, "epoch": 999, "lr": 7.594514013664742e-05}
{"train_loss": 0.16936159133911133, "global_step": 88922, "epoch": 999, "lr": 7.594464455677998e-05}
{"train_loss": 0.19212408363819122, "global_step": 88923, "epoch": 999, "lr": 7.594414897342461e-05}
{"train_loss": 0.1639811396598816, "global_step": 88924, "epoch": 999, "lr": 7.594365338658137e-05}
{"train_loss": 0.12666462361812592, "global_step": 88925, "epoch": 999, "lr": 7.594315779625032e-05}
{"train_loss": 0.27180466055870056, "global_step": 88926, "epoch": 999, "lr": 7.594266220243152e-05}
{"train_loss": 0.15170791745185852, "global_step": 88927, "epoch": 999, "lr": 7.594216660512505e-05}
{"train_loss": 0.197667196393013, "global_step": 88928, "epoch": 999, "lr": 7.594167100433097e-05}
{"train_loss": 0.12482695281505585, "global_step": 88929, "epoch": 999, "lr": 7.594117540004935e-05}
{"train_loss": 0.18143068253993988, "global_step": 88930, "epoch": 999, "lr": 7.594067979228026e-05}
{"train_loss": 0.24266549944877625, "global_step": 88931, "epoch": 999, "lr": 7.594018418102375e-05}
{"train_loss": 0.16389676928520203, "global_step": 88932, "epoch": 999, "lr": 7.593968856627991e-05}
{"train_loss": 0.13839960098266602, "global_step": 88933, "epoch": 999, "lr": 7.593919294804879e-05}
{"train_loss": 0.17763324081897736, "global_step": 88934, "epoch": 999, "lr": 7.593869732633046e-05}
{"train_loss": 0.23925717175006866, "global_step": 88935, "epoch": 999, "lr": 7.593820170112498e-05}
{"train_loss": 0.17347195744514465, "global_step": 88936, "epoch": 999, "lr": 7.593770607243244e-05}
{"train_loss": 0.2210199236869812, "global_step": 88937, "epoch": 999, "lr": 7.593721044025289e-05}
{"train_loss": 0.1443900316953659, "global_step": 88938, "epoch": 999, "lr": 7.593671480458637e-05}
{"train_loss": 0.21079888939857483, "global_step": 88939, "epoch": 999, "lr": 7.593621916543301e-05}
{"train_loss": 0.17343279719352722, "global_step": 88940, "epoch": 999, "lr": 7.593572352279284e-05}
{"train_loss": 0.18694208562374115, "global_step": 88941, "epoch": 999, "lr": 7.593522787666591e-05}
{"train_loss": 0.21201574802398682, "global_step": 88942, "epoch": 999, "lr": 7.593473222705232e-05}
{"train_loss": 0.252204954624176, "global_step": 88943, "epoch": 999, "lr": 7.593423657395211e-05}
{"train_loss": 0.23420487344264984, "global_step": 88944, "epoch": 999, "lr": 7.593374091736536e-05}
{"train_loss": 0.15924511849880219, "global_step": 88945, "epoch": 999, "lr": 7.593324525729216e-05}
{"train_loss": 0.13928313553333282, "global_step": 88946, "epoch": 999, "lr": 7.593274959373251e-05}
{"train_loss": 0.15640473365783691, "global_step": 88947, "epoch": 999, "lr": 7.593225392668655e-05}
{"train_loss": 0.18074984848499298, "global_step": 88948, "epoch": 999, "lr": 7.593175825615428e-05}
{"train_loss": 0.21454308927059174, "global_step": 88949, "epoch": 999, "lr": 7.593126258213583e-05}
{"train_loss": 0.2523581385612488, "global_step": 88950, "epoch": 999, "lr": 7.593076690463123e-05}
{"train_loss": 0.17985260486602783, "global_step": 88951, "epoch": 999, "lr": 7.593027122364057e-05}
{"train_loss": 0.17435528337955475, "global_step": 88952, "epoch": 999, "lr": 7.592977553916388e-05}
{"train_loss": 0.25477373600006104, "global_step": 88953, "epoch": 999, "lr": 7.592927985120124e-05}
{"train_loss": 0.20302847027778625, "global_step": 88954, "epoch": 999, "lr": 7.592878415975274e-05}
{"train_loss": 0.14203938841819763, "global_step": 88955, "epoch": 999, "lr": 7.592828846481843e-05}
{"train_loss": 0.16115523874759674, "global_step": 88956, "epoch": 999, "lr": 7.592779276639836e-05}
{"train_loss": 0.2090035378932953, "global_step": 88957, "epoch": 999, "lr": 7.592729706449264e-05}
{"train_loss": 0.16993604600429535, "global_step": 88958, "epoch": 999, "lr": 7.59268013591013e-05}
{"train_loss": 0.13262613117694855, "global_step": 88959, "epoch": 999, "lr": 7.592630565022441e-05}
{"train_loss": 0.12718333303928375, "global_step": 88960, "epoch": 999, "lr": 7.592580993786207e-05}
{"train_loss": 0.16899473965168, "global_step": 88961, "epoch": 999, "lr": 7.59253142220143e-05}
{"train_loss": 0.16437935829162598, "global_step": 88962, "epoch": 999, "lr": 7.59248185026812e-05}
{"train_loss": 0.2968185245990753, "global_step": 88963, "epoch": 999, "lr": 7.592432277986282e-05}
{"train_loss": 0.20911304652690887, "global_step": 88964, "epoch": 999, "lr": 7.592382705355922e-05}
{"train_loss": 0.22396236658096313, "global_step": 88965, "epoch": 999, "lr": 7.592333132377047e-05}
{"train_loss": 0.16837525367736816, "global_step": 88966, "epoch": 999, "lr": 7.592283559049667e-05}
{"train_loss": 0.18773195147514343, "global_step": 88967, "epoch": 999, "lr": 7.592233985373784e-05}
{"train_loss": 0.16271033883094788, "global_step": 88968, "epoch": 999, "lr": 7.592184411349408e-05}
{"train_loss": 0.20171797275543213, "global_step": 88969, "epoch": 999, "lr": 7.592134836976544e-05}
{"train_loss": 0.1359129250049591, "global_step": 88970, "epoch": 999, "lr": 7.592085262255199e-05}
{"train_loss": 0.14082470536231995, "global_step": 88971, "epoch": 999, "lr": 7.592035687185381e-05}
{"train_loss": 0.11878100782632828, "global_step": 88972, "epoch": 999, "lr": 7.591986111767093e-05}
{"train_loss": 0.20914459228515625, "global_step": 88973, "epoch": 999, "lr": 7.591936536000346e-05}
{"train_loss": 0.14622966945171356, "global_step": 88974, "epoch": 999, "lr": 7.591886959885144e-05}
{"train_loss": 0.1864752471446991, "global_step": 88975, "epoch": 999, "lr": 7.591837383421493e-05}
{"train_loss": 0.31687450408935547, "global_step": 88976, "epoch": 999, "lr": 7.591787806609402e-05}
{"train_loss": 0.10034842789173126, "global_step": 88977, "epoch": 999, "lr": 7.591738229448877e-05}
{"train_loss": 0.16686400771141052, "global_step": 88978, "epoch": 999, "lr": 7.591688651939925e-05}
{"train_loss": 0.19716224074363708, "global_step": 88979, "epoch": 999, "lr": 7.591639074082552e-05}
{"train_loss": 0.26863983273506165, "global_step": 88980, "epoch": 999, "lr": 7.591589495876762e-05}
{"train_loss": 0.1745264083147049, "global_step": 88981, "epoch": 999, "lr": 7.59153991732257e-05}
{"train_loss": 0.2055768072605133, "global_step": 88982, "epoch": 999, "lr": 7.591490338419973e-05}
{"train_loss": 0.13014784455299377, "global_step": 88983, "epoch": 999, "lr": 7.591440759168982e-05}
{"train_loss": 0.17140527069568634, "global_step": 88984, "epoch": 999, "lr": 7.591391179569605e-05}
{"train_loss": 0.16588377952575684, "global_step": 88985, "epoch": 999, "lr": 7.591341599621845e-05}
{"train_loss": 0.16454145312309265, "global_step": 88986, "epoch": 999, "lr": 7.591292019325711e-05}
{"train_loss": 0.19838374853134155, "global_step": 88987, "epoch": 999, "lr": 7.59124243868121e-05}
{"train_loss": 0.1297045648097992, "global_step": 88988, "epoch": 999, "lr": 7.591192857688349e-05}
{"train_loss": 0.16385330259799957, "global_step": 88989, "epoch": 999, "lr": 7.591143276347132e-05}
{"train_loss": 0.14333046972751617, "global_step": 88990, "epoch": 999, "lr": 7.591093694657567e-05}
{"train_loss": 0.19473563134670258, "global_step": 88991, "epoch": 999, "lr": 7.591044112619663e-05}
{"train_loss": 0.16110500693321228, "global_step": 88992, "epoch": 999, "lr": 7.590994530233423e-05}
{"train_loss": 0.15907025337219238, "global_step": 88993, "epoch": 999, "lr": 7.590944947498855e-05}
{"train_loss": 0.1291813999414444, "global_step": 88994, "epoch": 999, "lr": 7.590895364415969e-05}
{"train_loss": 0.18677453696727753, "global_step": 88995, "epoch": 999, "lr": 7.590845780984765e-05}
{"train_loss": 0.2928829491138458, "global_step": 88996, "epoch": 999, "lr": 7.590796197205257e-05}
{"train_loss": 0.1511254757642746, "global_step": 88997, "epoch": 999, "lr": 7.590746613077444e-05}
{"train_loss": 0.2310703545808792, "global_step": 88998, "epoch": 999, "lr": 7.59069702860134e-05}
{"train_loss": 0.18003228549542052, "global_step": 88999, "epoch": 999, "lr": 7.590647443776948e-05, "val_loss": 4.366181373596191}
{"train_loss": 0.20211030542850494, "global_step": 89000, "epoch": 1000, "lr": 7.590597858604275e-05}
{"train_loss": 0.171965554356575, "global_step": 89001, "epoch": 1000, "lr": 7.590548273083327e-05}
{"train_loss": 0.2334349900484085, "global_step": 89002, "epoch": 1000, "lr": 7.590498687214112e-05}
{"train_loss": 0.2084154635667801, "global_step": 89003, "epoch": 1000, "lr": 7.590449100996635e-05}
{"train_loss": 0.1968153417110443, "global_step": 89004, "epoch": 1000, "lr": 7.590399514430903e-05}
{"train_loss": 0.21204806864261627, "global_step": 89005, "epoch": 1000, "lr": 7.590349927516924e-05}
{"train_loss": 0.21934732794761658, "global_step": 89006, "epoch": 1000, "lr": 7.590300340254705e-05}
{"train_loss": 0.16804903745651245, "global_step": 89007, "epoch": 1000, "lr": 7.590250752644251e-05}
{"train_loss": 0.16344556212425232, "global_step": 89008, "epoch": 1000, "lr": 7.59020116468557e-05}
{"train_loss": 0.1466626077890396, "global_step": 89009, "epoch": 1000, "lr": 7.590151576378668e-05}
{"train_loss": 0.13466417789459229, "global_step": 89010, "epoch": 1000, "lr": 7.59010198772355e-05}
{"train_loss": 0.23518837988376617, "global_step": 89011, "epoch": 1000, "lr": 7.590052398720225e-05}
{"train_loss": 0.1889897584915161, "global_step": 89012, "epoch": 1000, "lr": 7.590002809368699e-05}
{"train_loss": 0.17505905032157898, "global_step": 89013, "epoch": 1000, "lr": 7.589953219668981e-05}
{"train_loss": 0.1328306496143341, "global_step": 89014, "epoch": 1000, "lr": 7.589903629621073e-05}
{"train_loss": 0.14304272830486298, "global_step": 89015, "epoch": 1000, "lr": 7.589854039224984e-05}
{"train_loss": 0.2005375176668167, "global_step": 89016, "epoch": 1000, "lr": 7.589804448480722e-05}
{"train_loss": 0.15576735138893127, "global_step": 89017, "epoch": 1000, "lr": 7.589754857388292e-05}
{"train_loss": 0.202353835105896, "global_step": 89018, "epoch": 1000, "lr": 7.5897052659477e-05}
{"train_loss": 0.29635679721832275, "global_step": 89019, "epoch": 1000, "lr": 7.589655674158955e-05}
{"train_loss": 0.23662026226520538, "global_step": 89020, "epoch": 1000, "lr": 7.589606082022062e-05}
{"train_loss": 0.21867632865905762, "global_step": 89021, "epoch": 1000, "lr": 7.589556489537029e-05}
{"train_loss": 0.14807257056236267, "global_step": 89022, "epoch": 1000, "lr": 7.58950689670386e-05}
{"train_loss": 0.19986040890216827, "global_step": 89023, "epoch": 1000, "lr": 7.589457303522564e-05}
{"train_loss": 0.14399726688861847, "global_step": 89024, "epoch": 1000, "lr": 7.589407709993146e-05}
{"train_loss": 0.23556461930274963, "global_step": 89025, "epoch": 1000, "lr": 7.589358116115615e-05}
{"train_loss": 0.22442162036895752, "global_step": 89026, "epoch": 1000, "lr": 7.589308521889977e-05}
{"train_loss": 0.1399562656879425, "global_step": 89027, "epoch": 1000, "lr": 7.589258927316237e-05}
{"train_loss": 0.18503469228744507, "global_step": 89028, "epoch": 1000, "lr": 7.589209332394402e-05}
{"train_loss": 0.19352596998214722, "global_step": 89029, "epoch": 1000, "lr": 7.589159737124478e-05}
{"train_loss": 0.1491040289402008, "global_step": 89030, "epoch": 1000, "lr": 7.589110141506476e-05}
{"train_loss": 0.14291375875473022, "global_step": 89031, "epoch": 1000, "lr": 7.589060545540399e-05}
{"train_loss": 0.09999305009841919, "global_step": 89032, "epoch": 1000, "lr": 7.589010949226254e-05}
{"train_loss": 0.14206458628177643, "global_step": 89033, "epoch": 1000, "lr": 7.588961352564048e-05}
{"train_loss": 0.17002788186073303, "global_step": 89034, "epoch": 1000, "lr": 7.588911755553787e-05}
{"train_loss": 0.17659823596477509, "global_step": 89035, "epoch": 1000, "lr": 7.58886215819548e-05}
{"train_loss": 0.12227872014045715, "global_step": 89036, "epoch": 1000, "lr": 7.588812560489132e-05}
{"train_loss": 0.19964167475700378, "global_step": 89037, "epoch": 1000, "lr": 7.588762962434748e-05}
{"train_loss": 0.19308620691299438, "global_step": 89038, "epoch": 1000, "lr": 7.588713364032338e-05}
{"train_loss": 0.23533663153648376, "global_step": 89039, "epoch": 1000, "lr": 7.588663765281907e-05}
{"train_loss": 0.2849067151546478, "global_step": 89040, "epoch": 1000, "lr": 7.58861416618346e-05}
{"train_loss": 0.14246897399425507, "global_step": 89041, "epoch": 1000, "lr": 7.588564566737007e-05}
{"train_loss": 0.21150512993335724, "global_step": 89042, "epoch": 1000, "lr": 7.588514966942553e-05}
{"train_loss": 0.13164567947387695, "global_step": 89043, "epoch": 1000, "lr": 7.588465366800104e-05}
{"train_loss": 0.13498760759830475, "global_step": 89044, "epoch": 1000, "lr": 7.588415766309668e-05}
{"train_loss": 0.13597926497459412, "global_step": 89045, "epoch": 1000, "lr": 7.58836616547125e-05}
{"train_loss": 0.2066112756729126, "global_step": 89046, "epoch": 1000, "lr": 7.588316564284859e-05}
{"train_loss": 0.24448318779468536, "global_step": 89047, "epoch": 1000, "lr": 7.588266962750499e-05}
{"train_loss": 0.1723993420600891, "global_step": 89048, "epoch": 1000, "lr": 7.588217360868178e-05}
{"train_loss": 0.23559781908988953, "global_step": 89049, "epoch": 1000, "lr": 7.588167758637904e-05}
{"train_loss": 0.2052466869354248, "global_step": 89050, "epoch": 1000, "lr": 7.588118156059681e-05}
{"train_loss": 0.18688634037971497, "global_step": 89051, "epoch": 1000, "lr": 7.588068553133519e-05}
{"train_loss": 0.267591267824173, "global_step": 89052, "epoch": 1000, "lr": 7.588018949859421e-05}
{"train_loss": 0.1632349193096161, "global_step": 89053, "epoch": 1000, "lr": 7.587969346237396e-05}
{"train_loss": 0.12273087352514267, "global_step": 89054, "epoch": 1000, "lr": 7.58791974226745e-05}
{"train_loss": 0.2107265293598175, "global_step": 89055, "epoch": 1000, "lr": 7.587870137949591e-05}
{"train_loss": 0.16046689450740814, "global_step": 89056, "epoch": 1000, "lr": 7.587820533283822e-05}
{"train_loss": 0.09988265484571457, "global_step": 89057, "epoch": 1000, "lr": 7.587770928270155e-05}
{"train_loss": 0.18676182627677917, "global_step": 89058, "epoch": 1000, "lr": 7.587721322908592e-05}
{"train_loss": 0.1790560930967331, "global_step": 89059, "epoch": 1000, "lr": 7.587671717199142e-05}
{"train_loss": 0.3106444776058197, "global_step": 89060, "epoch": 1000, "lr": 7.58762211114181e-05}
{"train_loss": 0.20049899816513062, "global_step": 89061, "epoch": 1000, "lr": 7.587572504736605e-05}
{"train_loss": 0.21068741381168365, "global_step": 89062, "epoch": 1000, "lr": 7.587522897983532e-05}
{"train_loss": 0.13592050969600677, "global_step": 89063, "epoch": 1000, "lr": 7.587473290882597e-05}
{"train_loss": 0.13658317923545837, "global_step": 89064, "epoch": 1000, "lr": 7.58742368343381e-05}
{"train_loss": 0.14901073276996613, "global_step": 89065, "epoch": 1000, "lr": 7.587374075637173e-05}
{"train_loss": 0.23426471650600433, "global_step": 89066, "epoch": 1000, "lr": 7.587324467492696e-05}
{"train_loss": 0.1875232756137848, "global_step": 89067, "epoch": 1000, "lr": 7.587274859000384e-05}
{"train_loss": 0.1430932581424713, "global_step": 89068, "epoch": 1000, "lr": 7.587225250160245e-05}
{"train_loss": 0.1953401118516922, "global_step": 89069, "epoch": 1000, "lr": 7.587175640972285e-05}
{"train_loss": 0.11563194543123245, "global_step": 89070, "epoch": 1000, "lr": 7.587126031436511e-05}
{"train_loss": 0.1889909952878952, "global_step": 89071, "epoch": 1000, "lr": 7.58707642155293e-05}
{"train_loss": 0.14571066200733185, "global_step": 89072, "epoch": 1000, "lr": 7.587026811321548e-05}
{"train_loss": 0.17281509935855865, "global_step": 89073, "epoch": 1000, "lr": 7.586977200742372e-05}
{"train_loss": 0.15744104981422424, "global_step": 89074, "epoch": 1000, "lr": 7.586927589815407e-05}
{"train_loss": 0.1770833283662796, "global_step": 89075, "epoch": 1000, "lr": 7.586877978540663e-05}
{"train_loss": 0.13826529681682587, "global_step": 89076, "epoch": 1000, "lr": 7.586828366918143e-05}
{"train_loss": 0.17713434994220734, "global_step": 89077, "epoch": 1000, "lr": 7.586778754947856e-05}
{"train_loss": 0.12038972973823547, "global_step": 89078, "epoch": 1000, "lr": 7.586729142629809e-05}
{"train_loss": 0.20855796337127686, "global_step": 89079, "epoch": 1000, "lr": 7.586679529964008e-05}
{"train_loss": 0.2988940477371216, "global_step": 89080, "epoch": 1000, "lr": 7.586629916950458e-05}
{"train_loss": 0.18284088373184204, "global_step": 89081, "epoch": 1000, "lr": 7.586580303589168e-05}
{"train_loss": 0.1321074515581131, "global_step": 89082, "epoch": 1000, "lr": 7.586530689880145e-05}
{"train_loss": 0.19714994728565216, "global_step": 89083, "epoch": 1000, "lr": 7.586481075823391e-05}
{"train_loss": 0.1751391440629959, "global_step": 89084, "epoch": 1000, "lr": 7.58643146141892e-05}
{"train_loss": 0.1720520406961441, "global_step": 89085, "epoch": 1000, "lr": 7.586381846666733e-05}
{"train_loss": 0.3395608365535736, "global_step": 89086, "epoch": 1000, "lr": 7.586332231566838e-05}
{"train_loss": 0.15119224786758423, "global_step": 89087, "epoch": 1000, "lr": 7.586282616119242e-05}
{"train_loss": 0.18291565717271205, "global_step": 89088, "epoch": 1000, "lr": 7.586233000323952e-05, "train/sim_max_reward_0": 0.5633622080711012, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.6072871837665634, "train/sim_max_reward_3": 0.4617428745040765, "train/sim_max_reward_4": 0.9849577486396643, "train/sim_max_reward_5": 0.8037402811188781, "test/sim_max_reward_4500000": 0.9505256546871903, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.98135208967467, "test/sim_max_reward_4500003": 0.7977953381021128, "test/sim_max_reward_4500004": 0.48283445935380787, "test/sim_max_reward_4500005": 0.45055549411703916, "test/sim_max_reward_4500006": 0.9205673571372556, "test/sim_max_reward_4500007": 0.9938325408738273, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.9817341082848186, "test/sim_max_reward_4500010": 0.015189030052918653, "test/sim_max_reward_4500011": 0.9291956725637143, "test/sim_max_reward_4500012": 1.0, "test/sim_max_reward_4500013": 0.969837744332928, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.41880980406906304, "test/sim_max_reward_4500016": 0.4335780312481267, "test/sim_max_reward_4500017": 0.0850102435834523, "test/sim_max_reward_4500018": 0.9832832428103959, "test/sim_max_reward_4500019": 0.9376754393481859, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9547283683057163, "test/sim_max_reward_4500022": 0.9940059659353538, "test/sim_max_reward_4500023": 0.7051391738964862, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9977750898032078, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 0.9993406177162912, "test/sim_max_reward_4500028": 0.988910844789887, "test/sim_max_reward_4500029": 0.6609371172595662, "test/sim_max_reward_4500030": 0.9664066232753631, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.7291117761825984, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.8284763781123647, "test/sim_max_reward_4500035": 0.6187120241534556, "test/sim_max_reward_4500036": 0.9442655905811063, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.7707091896037249, "test/sim_max_reward_4500039": 0.9801300792945561, "test/sim_max_reward_4500040": 0.9817600240254695, "test/sim_max_reward_4500041": 0.18520679021768974, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.5730992102716782, "test/sim_max_reward_4500044": 0.9911493164077142, "test/sim_max_reward_4500045": 0.6043262839937126, "test/sim_max_reward_4500046": 0.3502627178128462, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9695929922989631, "test/sim_max_reward_4500049": 0.9250849033776877, "train/mean_score": 0.7368483826833806, "test/mean_score": 0.6810181465510989, "val_loss": 4.285877704620361, "train_action_mse_error": 16.62441635131836}
{"train_loss": 0.10672392696142197, "global_step": 89089, "epoch": 1001, "lr": 7.586183384180977e-05}
{"train_loss": 0.16086287796497345, "global_step": 89090, "epoch": 1001, "lr": 7.586133767690317e-05}
{"train_loss": 0.24046459794044495, "global_step": 89091, "epoch": 1001, "lr": 7.586084150851986e-05}
{"train_loss": 0.18087249994277954, "global_step": 89092, "epoch": 1001, "lr": 7.586034533665987e-05}
{"train_loss": 0.20265427231788635, "global_step": 89093, "epoch": 1001, "lr": 7.585984916132325e-05}
{"train_loss": 0.15453621745109558, "global_step": 89094, "epoch": 1001, "lr": 7.585935298251012e-05}
{"train_loss": 0.21719105541706085, "global_step": 89095, "epoch": 1001, "lr": 7.585885680022048e-05}
{"train_loss": 0.1518106907606125, "global_step": 89096, "epoch": 1001, "lr": 7.585836061445445e-05}
{"train_loss": 0.22118644416332245, "global_step": 89097, "epoch": 1001, "lr": 7.58578644252121e-05}
{"train_loss": 0.16468876600265503, "global_step": 89098, "epoch": 1001, "lr": 7.585736823249346e-05}
{"train_loss": 0.15719814598560333, "global_step": 89099, "epoch": 1001, "lr": 7.585687203629861e-05}
{"train_loss": 0.16595996916294098, "global_step": 89100, "epoch": 1001, "lr": 7.585637583662761e-05}
{"train_loss": 0.15141277015209198, "global_step": 89101, "epoch": 1001, "lr": 7.585587963348055e-05}
{"train_loss": 0.13141612708568573, "global_step": 89102, "epoch": 1001, "lr": 7.585538342685747e-05}
{"train_loss": 0.25717470049858093, "global_step": 89103, "epoch": 1001, "lr": 7.585488721675844e-05}
{"train_loss": 0.16401417553424835, "global_step": 89104, "epoch": 1001, "lr": 7.585439100318356e-05}
{"train_loss": 0.14880287647247314, "global_step": 89105, "epoch": 1001, "lr": 7.585389478613285e-05}
{"train_loss": 0.17070135474205017, "global_step": 89106, "epoch": 1001, "lr": 7.585339856560642e-05}
{"train_loss": 0.21955890953540802, "global_step": 89107, "epoch": 1001, "lr": 7.58529023416043e-05}
{"train_loss": 0.15466925501823425, "global_step": 89108, "epoch": 1001, "lr": 7.585240611412659e-05}
{"train_loss": 0.19499331712722778, "global_step": 89109, "epoch": 1001, "lr": 7.585190988317332e-05}
{"train_loss": 0.11328741163015366, "global_step": 89110, "epoch": 1001, "lr": 7.585141364874457e-05}
{"train_loss": 0.1561049073934555, "global_step": 89111, "epoch": 1001, "lr": 7.585091741084044e-05}
{"train_loss": 0.1412188857793808, "global_step": 89112, "epoch": 1001, "lr": 7.585042116946095e-05}
{"train_loss": 0.19825100898742676, "global_step": 89113, "epoch": 1001, "lr": 7.584992492460618e-05}
{"train_loss": 0.16069357097148895, "global_step": 89114, "epoch": 1001, "lr": 7.584942867627622e-05}
{"train_loss": 0.12441271543502808, "global_step": 89115, "epoch": 1001, "lr": 7.584893242447113e-05}
{"train_loss": 0.17460058629512787, "global_step": 89116, "epoch": 1001, "lr": 7.584843616919094e-05}
{"train_loss": 0.0909956693649292, "global_step": 89117, "epoch": 1001, "lr": 7.584793991043574e-05}
{"train_loss": 0.14580541849136353, "global_step": 89118, "epoch": 1001, "lr": 7.584744364820562e-05}
{"train_loss": 0.06180708855390549, "global_step": 89119, "epoch": 1001, "lr": 7.584694738250062e-05}
{"train_loss": 0.1817285716533661, "global_step": 89120, "epoch": 1001, "lr": 7.58464511133208e-05}
{"train_loss": 0.17670407891273499, "global_step": 89121, "epoch": 1001, "lr": 7.584595484066625e-05}
{"train_loss": 0.15770220756530762, "global_step": 89122, "epoch": 1001, "lr": 7.584545856453704e-05}
{"train_loss": 0.23262040317058563, "global_step": 89123, "epoch": 1001, "lr": 7.58449622849332e-05}
{"train_loss": 0.29499614238739014, "global_step": 89124, "epoch": 1001, "lr": 7.584446600185484e-05}
{"train_loss": 0.13706448674201965, "global_step": 89125, "epoch": 1001, "lr": 7.584396971530199e-05}
{"train_loss": 0.22336769104003906, "global_step": 89126, "epoch": 1001, "lr": 7.584347342527473e-05}
{"train_loss": 0.14550673961639404, "global_step": 89127, "epoch": 1001, "lr": 7.584297713177316e-05}
{"train_loss": 0.1958075314760208, "global_step": 89128, "epoch": 1001, "lr": 7.584248083479728e-05}
{"train_loss": 0.1333610862493515, "global_step": 89129, "epoch": 1001, "lr": 7.58419845343472e-05}
{"train_loss": 0.20552277565002441, "global_step": 89130, "epoch": 1001, "lr": 7.5841488230423e-05}
{"train_loss": 0.17970767617225647, "global_step": 89131, "epoch": 1001, "lr": 7.584099192302472e-05}
{"train_loss": 0.14263752102851868, "global_step": 89132, "epoch": 1001, "lr": 7.584049561215243e-05}
{"train_loss": 0.10183613747358322, "global_step": 89133, "epoch": 1001, "lr": 7.58399992978062e-05}
{"train_loss": 0.23499998450279236, "global_step": 89134, "epoch": 1001, "lr": 7.58395029799861e-05}
{"train_loss": 0.13785526156425476, "global_step": 89135, "epoch": 1001, "lr": 7.58390066586922e-05}
{"train_loss": 0.17671534419059753, "global_step": 89136, "epoch": 1001, "lr": 7.583851033392455e-05}
{"train_loss": 0.2079191654920578, "global_step": 89137, "epoch": 1001, "lr": 7.583801400568322e-05}
{"train_loss": 0.18516966700553894, "global_step": 89138, "epoch": 1001, "lr": 7.58375176739683e-05}
{"train_loss": 0.17693503201007843, "global_step": 89139, "epoch": 1001, "lr": 7.583702133877982e-05}
{"train_loss": 0.15086455643177032, "global_step": 89140, "epoch": 1001, "lr": 7.583652500011788e-05}
{"train_loss": 0.12357146292924881, "global_step": 89141, "epoch": 1001, "lr": 7.583602865798253e-05}
{"train_loss": 0.2168348729610443, "global_step": 89142, "epoch": 1001, "lr": 7.583553231237386e-05}
{"train_loss": 0.14528873562812805, "global_step": 89143, "epoch": 1001, "lr": 7.58350359632919e-05}
{"train_loss": 0.19402509927749634, "global_step": 89144, "epoch": 1001, "lr": 7.583453961073673e-05}
{"train_loss": 0.32793107628822327, "global_step": 89145, "epoch": 1001, "lr": 7.583404325470843e-05}
{"train_loss": 0.13713201880455017, "global_step": 89146, "epoch": 1001, "lr": 7.583354689520705e-05}
{"train_loss": 0.1791858822107315, "global_step": 89147, "epoch": 1001, "lr": 7.583305053223269e-05}
{"train_loss": 0.194020614027977, "global_step": 89148, "epoch": 1001, "lr": 7.583255416578536e-05}
{"train_loss": 0.13192161917686462, "global_step": 89149, "epoch": 1001, "lr": 7.583205779586517e-05}
{"train_loss": 0.16482992470264435, "global_step": 89150, "epoch": 1001, "lr": 7.583156142247218e-05}
{"train_loss": 0.10351598262786865, "global_step": 89151, "epoch": 1001, "lr": 7.583106504560645e-05}
{"train_loss": 0.16614091396331787, "global_step": 89152, "epoch": 1001, "lr": 7.583056866526803e-05}
{"train_loss": 0.30998149514198303, "global_step": 89153, "epoch": 1001, "lr": 7.583007228145702e-05}
{"train_loss": 0.11479048430919647, "global_step": 89154, "epoch": 1001, "lr": 7.582957589417348e-05}
{"train_loss": 0.2206851989030838, "global_step": 89155, "epoch": 1001, "lr": 7.582907950341746e-05}
{"train_loss": 0.18005456030368805, "global_step": 89156, "epoch": 1001, "lr": 7.582858310918903e-05}
{"train_loss": 0.21952643990516663, "global_step": 89157, "epoch": 1001, "lr": 7.582808671148827e-05}
{"train_loss": 0.1192818209528923, "global_step": 89158, "epoch": 1001, "lr": 7.582759031031523e-05}
{"train_loss": 0.16657879948616028, "global_step": 89159, "epoch": 1001, "lr": 7.582709390566999e-05}
{"train_loss": 0.12956757843494415, "global_step": 89160, "epoch": 1001, "lr": 7.58265974975526e-05}
{"train_loss": 0.13986395299434662, "global_step": 89161, "epoch": 1001, "lr": 7.582610108596316e-05}
{"train_loss": 0.12574975192546844, "global_step": 89162, "epoch": 1001, "lr": 7.58256046709017e-05}
{"train_loss": 0.10679186135530472, "global_step": 89163, "epoch": 1001, "lr": 7.582510825236832e-05}
{"train_loss": 0.18194547295570374, "global_step": 89164, "epoch": 1001, "lr": 7.582461183036305e-05}
{"train_loss": 0.18716813623905182, "global_step": 89165, "epoch": 1001, "lr": 7.582411540488599e-05}
{"train_loss": 0.26134705543518066, "global_step": 89166, "epoch": 1001, "lr": 7.582361897593717e-05}
{"train_loss": 0.14089815318584442, "global_step": 89167, "epoch": 1001, "lr": 7.582312254351671e-05}
{"train_loss": 0.21261855959892273, "global_step": 89168, "epoch": 1001, "lr": 7.582262610762463e-05}
{"train_loss": 0.14256331324577332, "global_step": 89169, "epoch": 1001, "lr": 7.582212966826101e-05}
{"train_loss": 0.12348896265029907, "global_step": 89170, "epoch": 1001, "lr": 7.582163322542593e-05}
{"train_loss": 0.14501304924488068, "global_step": 89171, "epoch": 1001, "lr": 7.582113677911943e-05}
{"train_loss": 0.14843206107616425, "global_step": 89172, "epoch": 1001, "lr": 7.58206403293416e-05}
{"train_loss": 0.15980824828147888, "global_step": 89173, "epoch": 1001, "lr": 7.58201438760925e-05}
{"train_loss": 0.14278897643089294, "global_step": 89174, "epoch": 1001, "lr": 7.58196474193722e-05}
{"train_loss": 0.08918239921331406, "global_step": 89175, "epoch": 1001, "lr": 7.581915095918076e-05}
{"train_loss": 0.1687643975019455, "global_step": 89176, "epoch": 1001, "lr": 7.581865449551825e-05}
{"train_loss": 0.1683230714014407, "global_step": 89177, "epoch": 1001, "lr": 7.581815802838473e-05, "val_loss": 4.517510414123535}
{"train_loss": 0.1688624620437622, "global_step": 89178, "epoch": 1002, "lr": 7.581766155778029e-05}
{"train_loss": 0.17715580761432648, "global_step": 89179, "epoch": 1002, "lr": 7.581716508370496e-05}
{"train_loss": 0.2262708842754364, "global_step": 89180, "epoch": 1002, "lr": 7.581666860615883e-05}
{"train_loss": 0.09801299124956131, "global_step": 89181, "epoch": 1002, "lr": 7.581617212514199e-05}
{"train_loss": 0.17584262788295746, "global_step": 89182, "epoch": 1002, "lr": 7.581567564065446e-05}
{"train_loss": 0.15865488350391388, "global_step": 89183, "epoch": 1002, "lr": 7.581517915269633e-05}
{"train_loss": 0.24630072712898254, "global_step": 89184, "epoch": 1002, "lr": 7.581468266126766e-05}
{"train_loss": 0.1858014166355133, "global_step": 89185, "epoch": 1002, "lr": 7.581418616636851e-05}
{"train_loss": 0.1765531748533249, "global_step": 89186, "epoch": 1002, "lr": 7.581368966799898e-05}
{"train_loss": 0.14168712496757507, "global_step": 89187, "epoch": 1002, "lr": 7.581319316615911e-05}
{"train_loss": 0.20619823038578033, "global_step": 89188, "epoch": 1002, "lr": 7.581269666084896e-05}
{"train_loss": 0.11714533716440201, "global_step": 89189, "epoch": 1002, "lr": 7.58122001520686e-05}
{"train_loss": 0.12349513918161392, "global_step": 89190, "epoch": 1002, "lr": 7.581170363981813e-05}
{"train_loss": 0.16030827164649963, "global_step": 89191, "epoch": 1002, "lr": 7.581120712409757e-05}
{"train_loss": 0.21901562809944153, "global_step": 89192, "epoch": 1002, "lr": 7.581071060490701e-05}
{"train_loss": 0.18973450362682343, "global_step": 89193, "epoch": 1002, "lr": 7.581021408224654e-05}
{"train_loss": 0.24711380898952484, "global_step": 89194, "epoch": 1002, "lr": 7.580971755611616e-05}
{"train_loss": 0.16081666946411133, "global_step": 89195, "epoch": 1002, "lr": 7.580922102651601e-05}
{"train_loss": 0.20571501553058624, "global_step": 89196, "epoch": 1002, "lr": 7.580872449344611e-05}
{"train_loss": 0.12491080164909363, "global_step": 89197, "epoch": 1002, "lr": 7.580822795690654e-05}
{"train_loss": 0.18886499106884003, "global_step": 89198, "epoch": 1002, "lr": 7.580773141689739e-05}
{"train_loss": 0.17033934593200684, "global_step": 89199, "epoch": 1002, "lr": 7.580723487341868e-05}
{"train_loss": 0.1572970747947693, "global_step": 89200, "epoch": 1002, "lr": 7.580673832647052e-05}
{"train_loss": 0.1998560130596161, "global_step": 89201, "epoch": 1002, "lr": 7.580624177605294e-05}
{"train_loss": 0.23479299247264862, "global_step": 89202, "epoch": 1002, "lr": 7.580574522216604e-05}
{"train_loss": 0.20735801756381989, "global_step": 89203, "epoch": 1002, "lr": 7.580524866480986e-05}
{"train_loss": 0.21857284009456635, "global_step": 89204, "epoch": 1002, "lr": 7.580475210398449e-05}
{"train_loss": 0.1848163604736328, "global_step": 89205, "epoch": 1002, "lr": 7.580425553968998e-05}
{"train_loss": 0.12771444022655487, "global_step": 89206, "epoch": 1002, "lr": 7.58037589719264e-05}
{"train_loss": 0.10341939330101013, "global_step": 89207, "epoch": 1002, "lr": 7.580326240069382e-05}
{"train_loss": 0.16107267141342163, "global_step": 89208, "epoch": 1002, "lr": 7.58027658259923e-05}
{"train_loss": 0.11722195893526077, "global_step": 89209, "epoch": 1002, "lr": 7.580226924782193e-05}
{"train_loss": 0.2961912453174591, "global_step": 89210, "epoch": 1002, "lr": 7.580177266618276e-05}
{"train_loss": 0.21738334000110626, "global_step": 89211, "epoch": 1002, "lr": 7.580127608107483e-05}
{"train_loss": 0.28384071588516235, "global_step": 89212, "epoch": 1002, "lr": 7.580077949249826e-05}
{"train_loss": 0.17825236916542053, "global_step": 89213, "epoch": 1002, "lr": 7.580028290045308e-05}
{"train_loss": 0.18855924904346466, "global_step": 89214, "epoch": 1002, "lr": 7.579978630493936e-05}
{"train_loss": 0.14218609035015106, "global_step": 89215, "epoch": 1002, "lr": 7.579928970595718e-05}
{"train_loss": 0.15676115453243256, "global_step": 89216, "epoch": 1002, "lr": 7.579879310350661e-05}
{"train_loss": 0.20911377668380737, "global_step": 89217, "epoch": 1002, "lr": 7.57982964975877e-05}
{"train_loss": 0.18870140612125397, "global_step": 89218, "epoch": 1002, "lr": 7.579779988820051e-05}
{"train_loss": 0.23113742470741272, "global_step": 89219, "epoch": 1002, "lr": 7.579730327534514e-05}
{"train_loss": 0.10730951279401779, "global_step": 89220, "epoch": 1002, "lr": 7.57968066590216e-05}
{"train_loss": 0.14786314964294434, "global_step": 89221, "epoch": 1002, "lr": 7.579631003923004e-05}
{"train_loss": 0.2103779911994934, "global_step": 89222, "epoch": 1002, "lr": 7.579581341597045e-05}
{"train_loss": 0.15727998316287994, "global_step": 89223, "epoch": 1002, "lr": 7.579531678924294e-05}
{"train_loss": 0.26978641748428345, "global_step": 89224, "epoch": 1002, "lr": 7.579482015904757e-05}
{"train_loss": 0.19339528679847717, "global_step": 89225, "epoch": 1002, "lr": 7.579432352538438e-05}
{"train_loss": 0.18912279605865479, "global_step": 89226, "epoch": 1002, "lr": 7.579382688825348e-05}
{"train_loss": 0.18349771201610565, "global_step": 89227, "epoch": 1002, "lr": 7.579333024765489e-05}
{"train_loss": 0.1660798192024231, "global_step": 89228, "epoch": 1002, "lr": 7.579283360358873e-05}
{"train_loss": 0.19410845637321472, "global_step": 89229, "epoch": 1002, "lr": 7.579233695605503e-05}
{"train_loss": 0.10205865651369095, "global_step": 89230, "epoch": 1002, "lr": 7.579184030505385e-05}
{"train_loss": 0.17499054968357086, "global_step": 89231, "epoch": 1002, "lr": 7.579134365058528e-05}
{"train_loss": 0.20953822135925293, "global_step": 89232, "epoch": 1002, "lr": 7.579084699264937e-05}
{"train_loss": 0.15031160414218903, "global_step": 89233, "epoch": 1002, "lr": 7.579035033124621e-05}
{"train_loss": 0.14660778641700745, "global_step": 89234, "epoch": 1002, "lr": 7.578985366637584e-05}
{"train_loss": 0.2158668339252472, "global_step": 89235, "epoch": 1002, "lr": 7.578935699803835e-05}
{"train_loss": 0.20082268118858337, "global_step": 89236, "epoch": 1002, "lr": 7.578886032623378e-05}
{"train_loss": 0.25557154417037964, "global_step": 89237, "epoch": 1002, "lr": 7.578836365096222e-05}
{"train_loss": 0.13780364394187927, "global_step": 89238, "epoch": 1002, "lr": 7.578786697222373e-05}
{"train_loss": 0.19357630610466003, "global_step": 89239, "epoch": 1002, "lr": 7.578737029001837e-05}
{"train_loss": 0.24302180111408234, "global_step": 89240, "epoch": 1002, "lr": 7.578687360434622e-05}
{"train_loss": 0.17937718331813812, "global_step": 89241, "epoch": 1002, "lr": 7.578637691520733e-05}
{"train_loss": 0.15527203679084778, "global_step": 89242, "epoch": 1002, "lr": 7.57858802226018e-05}
{"train_loss": 0.18739472329616547, "global_step": 89243, "epoch": 1002, "lr": 7.578538352652965e-05}
{"train_loss": 0.20113009214401245, "global_step": 89244, "epoch": 1002, "lr": 7.578488682699097e-05}
{"train_loss": 0.30792436003685, "global_step": 89245, "epoch": 1002, "lr": 7.578439012398584e-05}
{"train_loss": 0.19696974754333496, "global_step": 89246, "epoch": 1002, "lr": 7.57838934175143e-05}
{"train_loss": 0.29551035165786743, "global_step": 89247, "epoch": 1002, "lr": 7.578339670757643e-05}
{"train_loss": 0.15517212450504303, "global_step": 89248, "epoch": 1002, "lr": 7.57828999941723e-05}
{"train_loss": 0.2755698263645172, "global_step": 89249, "epoch": 1002, "lr": 7.578240327730197e-05}
{"train_loss": 0.1825394630432129, "global_step": 89250, "epoch": 1002, "lr": 7.578190655696552e-05}
{"train_loss": 0.1538088470697403, "global_step": 89251, "epoch": 1002, "lr": 7.5781409833163e-05}
{"train_loss": 0.15336564183235168, "global_step": 89252, "epoch": 1002, "lr": 7.578091310589448e-05}
{"train_loss": 0.23323290050029755, "global_step": 89253, "epoch": 1002, "lr": 7.578041637516004e-05}
{"train_loss": 0.228702113032341, "global_step": 89254, "epoch": 1002, "lr": 7.577991964095973e-05}
{"train_loss": 0.15594570338726044, "global_step": 89255, "epoch": 1002, "lr": 7.577942290329362e-05}
{"train_loss": 0.16515317559242249, "global_step": 89256, "epoch": 1002, "lr": 7.577892616216179e-05}
{"train_loss": 0.1462341994047165, "global_step": 89257, "epoch": 1002, "lr": 7.57784294175643e-05}
{"train_loss": 0.21874849498271942, "global_step": 89258, "epoch": 1002, "lr": 7.57779326695012e-05}
{"train_loss": 0.17275798320770264, "global_step": 89259, "epoch": 1002, "lr": 7.577743591797258e-05}
{"train_loss": 0.1980687379837036, "global_step": 89260, "epoch": 1002, "lr": 7.577693916297849e-05}
{"train_loss": 0.1739683747291565, "global_step": 89261, "epoch": 1002, "lr": 7.577644240451903e-05}
{"train_loss": 0.1718122959136963, "global_step": 89262, "epoch": 1002, "lr": 7.577594564259422e-05}
{"train_loss": 0.24073636531829834, "global_step": 89263, "epoch": 1002, "lr": 7.577544887720415e-05}
{"train_loss": 0.1795331835746765, "global_step": 89264, "epoch": 1002, "lr": 7.577495210834889e-05}
{"train_loss": 0.15846498310565948, "global_step": 89265, "epoch": 1002, "lr": 7.57744553360285e-05}
{"train_loss": 0.18637849641649912, "global_step": 89266, "epoch": 1002, "lr": 7.577395856024304e-05, "val_loss": 4.224726676940918}
{"train_loss": 0.19076122343540192, "global_step": 89267, "epoch": 1003, "lr": 7.577346178099259e-05}
{"train_loss": 0.16476286947727203, "global_step": 89268, "epoch": 1003, "lr": 7.577296499827721e-05}
{"train_loss": 0.15409906208515167, "global_step": 89269, "epoch": 1003, "lr": 7.577246821209697e-05}
{"train_loss": 0.20217153429985046, "global_step": 89270, "epoch": 1003, "lr": 7.577197142245196e-05}
{"train_loss": 0.2413739264011383, "global_step": 89271, "epoch": 1003, "lr": 7.57714746293422e-05}
{"train_loss": 0.14136484265327454, "global_step": 89272, "epoch": 1003, "lr": 7.577097783276779e-05}
{"train_loss": 0.19978708028793335, "global_step": 89273, "epoch": 1003, "lr": 7.577048103272877e-05}
{"train_loss": 0.22650465369224548, "global_step": 89274, "epoch": 1003, "lr": 7.576998422922523e-05}
{"train_loss": 0.208764910697937, "global_step": 89275, "epoch": 1003, "lr": 7.576948742225722e-05}
{"train_loss": 0.17482802271842957, "global_step": 89276, "epoch": 1003, "lr": 7.576899061182483e-05}
{"train_loss": 0.14150746166706085, "global_step": 89277, "epoch": 1003, "lr": 7.576849379792812e-05}
{"train_loss": 0.23016458749771118, "global_step": 89278, "epoch": 1003, "lr": 7.576799698056714e-05}
{"train_loss": 0.08679262548685074, "global_step": 89279, "epoch": 1003, "lr": 7.576750015974197e-05}
{"train_loss": 0.1537678837776184, "global_step": 89280, "epoch": 1003, "lr": 7.576700333545266e-05}
{"train_loss": 0.22165018320083618, "global_step": 89281, "epoch": 1003, "lr": 7.576650650769931e-05}
{"train_loss": 0.18886440992355347, "global_step": 89282, "epoch": 1003, "lr": 7.576600967648195e-05}
{"train_loss": 0.2272840291261673, "global_step": 89283, "epoch": 1003, "lr": 7.576551284180067e-05}
{"train_loss": 0.22879420220851898, "global_step": 89284, "epoch": 1003, "lr": 7.576501600365553e-05}
{"train_loss": 0.2232726514339447, "global_step": 89285, "epoch": 1003, "lr": 7.57645191620466e-05}
{"train_loss": 0.18608708679676056, "global_step": 89286, "epoch": 1003, "lr": 7.576402231697395e-05}
{"train_loss": 0.18532036244869232, "global_step": 89287, "epoch": 1003, "lr": 7.576352546843764e-05}
{"train_loss": 0.17585060000419617, "global_step": 89288, "epoch": 1003, "lr": 7.576302861643774e-05}
{"train_loss": 0.168594628572464, "global_step": 89289, "epoch": 1003, "lr": 7.576253176097431e-05}
{"train_loss": 0.1390145868062973, "global_step": 89290, "epoch": 1003, "lr": 7.576203490204744e-05}
{"train_loss": 0.17817915976047516, "global_step": 89291, "epoch": 1003, "lr": 7.576153803965715e-05}
{"train_loss": 0.19354847073554993, "global_step": 89292, "epoch": 1003, "lr": 7.576104117380354e-05}
{"train_loss": 0.2627333700656891, "global_step": 89293, "epoch": 1003, "lr": 7.576054430448668e-05}
{"train_loss": 0.20614229142665863, "global_step": 89294, "epoch": 1003, "lr": 7.576004743170661e-05}
{"train_loss": 0.28450629115104675, "global_step": 89295, "epoch": 1003, "lr": 7.575955055546345e-05}
{"train_loss": 0.21806611120700836, "global_step": 89296, "epoch": 1003, "lr": 7.57590536757572e-05}
{"train_loss": 0.2357630431652069, "global_step": 89297, "epoch": 1003, "lr": 7.575855679258798e-05}
{"train_loss": 0.12199670821428299, "global_step": 89298, "epoch": 1003, "lr": 7.575805990595582e-05}
{"train_loss": 0.20201964676380157, "global_step": 89299, "epoch": 1003, "lr": 7.575756301586082e-05}
{"train_loss": 0.18198662996292114, "global_step": 89300, "epoch": 1003, "lr": 7.575706612230303e-05}
{"train_loss": 0.14347223937511444, "global_step": 89301, "epoch": 1003, "lr": 7.57565692252825e-05}
{"train_loss": 0.2326633632183075, "global_step": 89302, "epoch": 1003, "lr": 7.57560723247993e-05}
{"train_loss": 0.16315601766109467, "global_step": 89303, "epoch": 1003, "lr": 7.575557542085354e-05}
{"train_loss": 0.10321839153766632, "global_step": 89304, "epoch": 1003, "lr": 7.575507851344525e-05}
{"train_loss": 0.12671539187431335, "global_step": 89305, "epoch": 1003, "lr": 7.575458160257451e-05}
{"train_loss": 0.1733117401599884, "global_step": 89306, "epoch": 1003, "lr": 7.575408468824137e-05}
{"train_loss": 0.1816639006137848, "global_step": 89307, "epoch": 1003, "lr": 7.575358777044591e-05}
{"train_loss": 0.22262811660766602, "global_step": 89308, "epoch": 1003, "lr": 7.575309084918819e-05}
{"train_loss": 0.12044579535722733, "global_step": 89309, "epoch": 1003, "lr": 7.575259392446828e-05}
{"train_loss": 0.1786208152770996, "global_step": 89310, "epoch": 1003, "lr": 7.575209699628626e-05}
{"train_loss": 0.20111802220344543, "global_step": 89311, "epoch": 1003, "lr": 7.575160006464217e-05}
{"train_loss": 0.1939888894557953, "global_step": 89312, "epoch": 1003, "lr": 7.575110312953611e-05}
{"train_loss": 0.15878361463546753, "global_step": 89313, "epoch": 1003, "lr": 7.57506061909681e-05}
{"train_loss": 0.12484049052000046, "global_step": 89314, "epoch": 1003, "lr": 7.575010924893826e-05}
{"train_loss": 0.17089679837226868, "global_step": 89315, "epoch": 1003, "lr": 7.574961230344662e-05}
{"train_loss": 0.1263435035943985, "global_step": 89316, "epoch": 1003, "lr": 7.574911535449325e-05}
{"train_loss": 0.1604650616645813, "global_step": 89317, "epoch": 1003, "lr": 7.574861840207824e-05}
{"train_loss": 0.2186487466096878, "global_step": 89318, "epoch": 1003, "lr": 7.574812144620165e-05}
{"train_loss": 0.17678216099739075, "global_step": 89319, "epoch": 1003, "lr": 7.57476244868635e-05}
{"train_loss": 0.18857000768184662, "global_step": 89320, "epoch": 1003, "lr": 7.574712752406392e-05}
{"train_loss": 0.16866837441921234, "global_step": 89321, "epoch": 1003, "lr": 7.574663055780296e-05}
{"train_loss": 0.23416492342948914, "global_step": 89322, "epoch": 1003, "lr": 7.574613358808066e-05}
{"train_loss": 0.134948268532753, "global_step": 89323, "epoch": 1003, "lr": 7.574563661489712e-05}
{"train_loss": 0.17324958741664886, "global_step": 89324, "epoch": 1003, "lr": 7.574513963825243e-05}
{"train_loss": 0.230371356010437, "global_step": 89325, "epoch": 1003, "lr": 7.574464265814659e-05}
{"train_loss": 0.12770415842533112, "global_step": 89326, "epoch": 1003, "lr": 7.574414567457967e-05}
{"train_loss": 0.21833188831806183, "global_step": 89327, "epoch": 1003, "lr": 7.574364868755178e-05}
{"train_loss": 0.20338787138462067, "global_step": 89328, "epoch": 1003, "lr": 7.574315169706297e-05}
{"train_loss": 0.13630302250385284, "global_step": 89329, "epoch": 1003, "lr": 7.574265470311332e-05}
{"train_loss": 0.22020070254802704, "global_step": 89330, "epoch": 1003, "lr": 7.574215770570287e-05}
{"train_loss": 0.18942396342754364, "global_step": 89331, "epoch": 1003, "lr": 7.574166070483171e-05}
{"train_loss": 0.2672020494937897, "global_step": 89332, "epoch": 1003, "lr": 7.574116370049988e-05}
{"train_loss": 0.21086156368255615, "global_step": 89333, "epoch": 1003, "lr": 7.574066669270749e-05}
{"train_loss": 0.19371487200260162, "global_step": 89334, "epoch": 1003, "lr": 7.574016968145456e-05}
{"train_loss": 0.17966090142726898, "global_step": 89335, "epoch": 1003, "lr": 7.573967266674118e-05}
{"train_loss": 0.21852731704711914, "global_step": 89336, "epoch": 1003, "lr": 7.573917564856743e-05}
{"train_loss": 0.1488197147846222, "global_step": 89337, "epoch": 1003, "lr": 7.573867862693333e-05}
{"train_loss": 0.15452232956886292, "global_step": 89338, "epoch": 1003, "lr": 7.573818160183901e-05}
{"train_loss": 0.19271792471408844, "global_step": 89339, "epoch": 1003, "lr": 7.573768457328449e-05}
{"train_loss": 0.19640246033668518, "global_step": 89340, "epoch": 1003, "lr": 7.573718754126986e-05}
{"train_loss": 0.18849597871303558, "global_step": 89341, "epoch": 1003, "lr": 7.573669050579517e-05}
{"train_loss": 0.15238933265209198, "global_step": 89342, "epoch": 1003, "lr": 7.57361934668605e-05}
{"train_loss": 0.18155525624752045, "global_step": 89343, "epoch": 1003, "lr": 7.573569642446593e-05}
{"train_loss": 0.10354200750589371, "global_step": 89344, "epoch": 1003, "lr": 7.573519937861149e-05}
{"train_loss": 0.12683263421058655, "global_step": 89345, "epoch": 1003, "lr": 7.573470232929726e-05}
{"train_loss": 0.15570081770420074, "global_step": 89346, "epoch": 1003, "lr": 7.573420527652332e-05}
{"train_loss": 0.20655055344104767, "global_step": 89347, "epoch": 1003, "lr": 7.573370822028976e-05}
{"train_loss": 0.1707824021577835, "global_step": 89348, "epoch": 1003, "lr": 7.573321116059658e-05}
{"train_loss": 0.1901961714029312, "global_step": 89349, "epoch": 1003, "lr": 7.57327140974439e-05}
{"train_loss": 0.18526721000671387, "global_step": 89350, "epoch": 1003, "lr": 7.573221703083175e-05}
{"train_loss": 0.15285460650920868, "global_step": 89351, "epoch": 1003, "lr": 7.573171996076022e-05}
{"train_loss": 0.15710021555423737, "global_step": 89352, "epoch": 1003, "lr": 7.57312228872294e-05}
{"train_loss": 0.1118553876876831, "global_step": 89353, "epoch": 1003, "lr": 7.57307258102393e-05}
{"train_loss": 0.15728016197681427, "global_step": 89354, "epoch": 1003, "lr": 7.573022872979003e-05}
{"train_loss": 0.18153206309240855, "global_step": 89355, "epoch": 1003, "lr": 7.572973164588166e-05, "val_loss": 4.273754119873047}
{"train_loss": 0.12175347656011581, "global_step": 89356, "epoch": 1004, "lr": 7.572923455851422e-05}
{"train_loss": 0.19897378981113434, "global_step": 89357, "epoch": 1004, "lr": 7.57287374676878e-05}
{"train_loss": 0.18982642889022827, "global_step": 89358, "epoch": 1004, "lr": 7.572824037340247e-05}
{"train_loss": 0.1569593846797943, "global_step": 89359, "epoch": 1004, "lr": 7.57277432756583e-05}
{"train_loss": 0.13782134652137756, "global_step": 89360, "epoch": 1004, "lr": 7.572724617445532e-05}
{"train_loss": 0.18724925816059113, "global_step": 89361, "epoch": 1004, "lr": 7.572674906979366e-05}
{"train_loss": 0.27212634682655334, "global_step": 89362, "epoch": 1004, "lr": 7.572625196167333e-05}
{"train_loss": 0.12572261691093445, "global_step": 89363, "epoch": 1004, "lr": 7.572575485009443e-05}
{"train_loss": 0.09334834665060043, "global_step": 89364, "epoch": 1004, "lr": 7.572525773505702e-05}
{"train_loss": 0.16834871470928192, "global_step": 89365, "epoch": 1004, "lr": 7.572476061656116e-05}
{"train_loss": 0.18217647075653076, "global_step": 89366, "epoch": 1004, "lr": 7.572426349460691e-05}
{"train_loss": 0.19386309385299683, "global_step": 89367, "epoch": 1004, "lr": 7.572376636919435e-05}
{"train_loss": 0.17302320897579193, "global_step": 89368, "epoch": 1004, "lr": 7.572326924032355e-05}
{"train_loss": 0.2256181538105011, "global_step": 89369, "epoch": 1004, "lr": 7.572277210799456e-05}
{"train_loss": 0.1832815706729889, "global_step": 89370, "epoch": 1004, "lr": 7.572227497220746e-05}
{"train_loss": 0.14565622806549072, "global_step": 89371, "epoch": 1004, "lr": 7.572177783296232e-05}
{"train_loss": 0.2904391586780548, "global_step": 89372, "epoch": 1004, "lr": 7.572128069025919e-05}
{"train_loss": 0.136773020029068, "global_step": 89373, "epoch": 1004, "lr": 7.572078354409815e-05}
{"train_loss": 0.15479989349842072, "global_step": 89374, "epoch": 1004, "lr": 7.572028639447926e-05}
{"train_loss": 0.17743085324764252, "global_step": 89375, "epoch": 1004, "lr": 7.571978924140259e-05}
{"train_loss": 0.160820871591568, "global_step": 89376, "epoch": 1004, "lr": 7.571929208486823e-05}
{"train_loss": 0.1478940099477768, "global_step": 89377, "epoch": 1004, "lr": 7.571879492487619e-05}
{"train_loss": 0.1604674607515335, "global_step": 89378, "epoch": 1004, "lr": 7.571829776142661e-05}
{"train_loss": 0.2077288180589676, "global_step": 89379, "epoch": 1004, "lr": 7.571780059451949e-05}
{"train_loss": 0.17407238483428955, "global_step": 89380, "epoch": 1004, "lr": 7.571730342415492e-05}
{"train_loss": 0.1284760981798172, "global_step": 89381, "epoch": 1004, "lr": 7.5716806250333e-05}
{"train_loss": 0.21310947835445404, "global_step": 89382, "epoch": 1004, "lr": 7.571630907305376e-05}
{"train_loss": 0.19673776626586914, "global_step": 89383, "epoch": 1004, "lr": 7.571581189231727e-05}
{"train_loss": 0.16103976964950562, "global_step": 89384, "epoch": 1004, "lr": 7.57153147081236e-05}
{"train_loss": 0.1180277019739151, "global_step": 89385, "epoch": 1004, "lr": 7.571481752047283e-05}
{"train_loss": 0.1727457195520401, "global_step": 89386, "epoch": 1004, "lr": 7.5714320329365e-05}
{"train_loss": 0.16961321234703064, "global_step": 89387, "epoch": 1004, "lr": 7.571382313480023e-05}
{"train_loss": 0.22139695286750793, "global_step": 89388, "epoch": 1004, "lr": 7.571332593677851e-05}
{"train_loss": 0.16731727123260498, "global_step": 89389, "epoch": 1004, "lr": 7.571282873529996e-05}
{"train_loss": 0.16043424606323242, "global_step": 89390, "epoch": 1004, "lr": 7.571233153036463e-05}
{"train_loss": 0.2358989268541336, "global_step": 89391, "epoch": 1004, "lr": 7.57118343219726e-05}
{"train_loss": 0.12458858639001846, "global_step": 89392, "epoch": 1004, "lr": 7.571133711012391e-05}
{"train_loss": 0.17865660786628723, "global_step": 89393, "epoch": 1004, "lr": 7.571083989481864e-05}
{"train_loss": 0.16864480078220367, "global_step": 89394, "epoch": 1004, "lr": 7.571034267605687e-05}
{"train_loss": 0.26044028997421265, "global_step": 89395, "epoch": 1004, "lr": 7.570984545383868e-05}
{"train_loss": 0.11975085735321045, "global_step": 89396, "epoch": 1004, "lr": 7.57093482281641e-05}
{"train_loss": 0.17899926006793976, "global_step": 89397, "epoch": 1004, "lr": 7.570885099903319e-05}
{"train_loss": 0.15593861043453217, "global_step": 89398, "epoch": 1004, "lr": 7.570835376644607e-05}
{"train_loss": 0.18684789538383484, "global_step": 89399, "epoch": 1004, "lr": 7.570785653040276e-05}
{"train_loss": 0.15288342535495758, "global_step": 89400, "epoch": 1004, "lr": 7.570735929090334e-05}
{"train_loss": 0.13777871429920197, "global_step": 89401, "epoch": 1004, "lr": 7.570686204794789e-05}
{"train_loss": 0.1678926795721054, "global_step": 89402, "epoch": 1004, "lr": 7.570636480153647e-05}
{"train_loss": 0.1777278482913971, "global_step": 89403, "epoch": 1004, "lr": 7.570586755166913e-05}
{"train_loss": 0.14784502983093262, "global_step": 89404, "epoch": 1004, "lr": 7.570537029834595e-05}
{"train_loss": 0.21046948432922363, "global_step": 89405, "epoch": 1004, "lr": 7.5704873041567e-05}
{"train_loss": 0.1800900548696518, "global_step": 89406, "epoch": 1004, "lr": 7.570437578133234e-05}
{"train_loss": 0.15473636984825134, "global_step": 89407, "epoch": 1004, "lr": 7.570387851764205e-05}
{"train_loss": 0.14036527276039124, "global_step": 89408, "epoch": 1004, "lr": 7.570338125049617e-05}
{"train_loss": 0.21096202731132507, "global_step": 89409, "epoch": 1004, "lr": 7.570288397989479e-05}
{"train_loss": 0.20363451540470123, "global_step": 89410, "epoch": 1004, "lr": 7.570238670583795e-05}
{"train_loss": 0.10752854496240616, "global_step": 89411, "epoch": 1004, "lr": 7.570188942832576e-05}
{"train_loss": 0.1834677755832672, "global_step": 89412, "epoch": 1004, "lr": 7.570139214735827e-05}
{"train_loss": 0.11152402311563492, "global_step": 89413, "epoch": 1004, "lr": 7.570089486293554e-05}
{"train_loss": 0.1530035436153412, "global_step": 89414, "epoch": 1004, "lr": 7.570039757505763e-05}
{"train_loss": 0.302634060382843, "global_step": 89415, "epoch": 1004, "lr": 7.569990028372463e-05}
{"train_loss": 0.17220081388950348, "global_step": 89416, "epoch": 1004, "lr": 7.569940298893657e-05}
{"train_loss": 0.22186098992824554, "global_step": 89417, "epoch": 1004, "lr": 7.569890569069354e-05}
{"train_loss": 0.3494272828102112, "global_step": 89418, "epoch": 1004, "lr": 7.569840838899561e-05}
{"train_loss": 0.2478545606136322, "global_step": 89419, "epoch": 1004, "lr": 7.569791108384285e-05}
{"train_loss": 0.19619892537593842, "global_step": 89420, "epoch": 1004, "lr": 7.569741377523531e-05}
{"train_loss": 0.17170222103595734, "global_step": 89421, "epoch": 1004, "lr": 7.569691646317308e-05}
{"train_loss": 0.19886267185211182, "global_step": 89422, "epoch": 1004, "lr": 7.56964191476562e-05}
{"train_loss": 0.17849259078502655, "global_step": 89423, "epoch": 1004, "lr": 7.569592182868476e-05}
{"train_loss": 0.14180287718772888, "global_step": 89424, "epoch": 1004, "lr": 7.569542450625881e-05}
{"train_loss": 0.1083216667175293, "global_step": 89425, "epoch": 1004, "lr": 7.569492718037841e-05}
{"train_loss": 0.15606947243213654, "global_step": 89426, "epoch": 1004, "lr": 7.569442985104366e-05}
{"train_loss": 0.1483488529920578, "global_step": 89427, "epoch": 1004, "lr": 7.56939325182546e-05}
{"train_loss": 0.14455850422382355, "global_step": 89428, "epoch": 1004, "lr": 7.569343518201129e-05}
{"train_loss": 0.2264421284198761, "global_step": 89429, "epoch": 1004, "lr": 7.569293784231382e-05}
{"train_loss": 0.2211800515651703, "global_step": 89430, "epoch": 1004, "lr": 7.569244049916226e-05}
{"train_loss": 0.22459052503108978, "global_step": 89431, "epoch": 1004, "lr": 7.569194315255665e-05}
{"train_loss": 0.1970970332622528, "global_step": 89432, "epoch": 1004, "lr": 7.569144580249707e-05}
{"train_loss": 0.15554533898830414, "global_step": 89433, "epoch": 1004, "lr": 7.56909484489836e-05}
{"train_loss": 0.15607474744319916, "global_step": 89434, "epoch": 1004, "lr": 7.56904510920163e-05}
{"train_loss": 0.12469013035297394, "global_step": 89435, "epoch": 1004, "lr": 7.568995373159522e-05}
{"train_loss": 0.1401609182357788, "global_step": 89436, "epoch": 1004, "lr": 7.568945636772044e-05}
{"train_loss": 0.1874556988477707, "global_step": 89437, "epoch": 1004, "lr": 7.568895900039204e-05}
{"train_loss": 0.1578221470117569, "global_step": 89438, "epoch": 1004, "lr": 7.568846162961005e-05}
{"train_loss": 0.10566875338554382, "global_step": 89439, "epoch": 1004, "lr": 7.568796425537457e-05}
{"train_loss": 0.15799064934253693, "global_step": 89440, "epoch": 1004, "lr": 7.568746687768566e-05}
{"train_loss": 0.18625909090042114, "global_step": 89441, "epoch": 1004, "lr": 7.56869694965434e-05}
{"train_loss": 0.18994829058647156, "global_step": 89442, "epoch": 1004, "lr": 7.568647211194781e-05}
{"train_loss": 0.14699658751487732, "global_step": 89443, "epoch": 1004, "lr": 7.568597472389898e-05}
{"train_loss": 0.17541552284795248, "global_step": 89444, "epoch": 1004, "lr": 7.568547733239702e-05, "val_loss": 4.4246439933776855}
{"train_loss": 0.20326529443264008, "global_step": 89445, "epoch": 1005, "lr": 7.568497993744193e-05}
{"train_loss": 0.14079178869724274, "global_step": 89446, "epoch": 1005, "lr": 7.568448253903381e-05}
{"train_loss": 0.19496430456638336, "global_step": 89447, "epoch": 1005, "lr": 7.568398513717272e-05}
{"train_loss": 0.2247740924358368, "global_step": 89448, "epoch": 1005, "lr": 7.568348773185877e-05}
{"train_loss": 0.1686665266752243, "global_step": 89449, "epoch": 1005, "lr": 7.568299032309194e-05}
{"train_loss": 0.1881183236837387, "global_step": 89450, "epoch": 1005, "lr": 7.568249291087237e-05}
{"train_loss": 0.11269919574260712, "global_step": 89451, "epoch": 1005, "lr": 7.56819954952001e-05}
{"train_loss": 0.19864459335803986, "global_step": 89452, "epoch": 1005, "lr": 7.56814980760752e-05}
{"train_loss": 0.16179488599300385, "global_step": 89453, "epoch": 1005, "lr": 7.568100065349772e-05}
{"train_loss": 0.2472616583108902, "global_step": 89454, "epoch": 1005, "lr": 7.568050322746776e-05}
{"train_loss": 0.21230579912662506, "global_step": 89455, "epoch": 1005, "lr": 7.568000579798538e-05}
{"train_loss": 0.1134132668375969, "global_step": 89456, "epoch": 1005, "lr": 7.567950836505061e-05}
{"train_loss": 0.27260708808898926, "global_step": 89457, "epoch": 1005, "lr": 7.567901092866355e-05}
{"train_loss": 0.12375231087207794, "global_step": 89458, "epoch": 1005, "lr": 7.567851348882427e-05}
{"train_loss": 0.09561798721551895, "global_step": 89459, "epoch": 1005, "lr": 7.567801604553281e-05}
{"train_loss": 0.1718512773513794, "global_step": 89460, "epoch": 1005, "lr": 7.567751859878929e-05}
{"train_loss": 0.1885448694229126, "global_step": 89461, "epoch": 1005, "lr": 7.567702114859371e-05}
{"train_loss": 0.20556648075580597, "global_step": 89462, "epoch": 1005, "lr": 7.567652369494617e-05}
{"train_loss": 0.20463956892490387, "global_step": 89463, "epoch": 1005, "lr": 7.567602623784673e-05}
{"train_loss": 0.22803495824337006, "global_step": 89464, "epoch": 1005, "lr": 7.567552877729547e-05}
{"train_loss": 0.188319131731987, "global_step": 89465, "epoch": 1005, "lr": 7.567503131329244e-05}
{"train_loss": 0.10850660502910614, "global_step": 89466, "epoch": 1005, "lr": 7.567453384583771e-05}
{"train_loss": 0.19041304290294647, "global_step": 89467, "epoch": 1005, "lr": 7.567403637493138e-05}
{"train_loss": 0.13943596184253693, "global_step": 89468, "epoch": 1005, "lr": 7.567353890057348e-05}
{"train_loss": 0.17136923968791962, "global_step": 89469, "epoch": 1005, "lr": 7.567304142276406e-05}
{"train_loss": 0.19160348176956177, "global_step": 89470, "epoch": 1005, "lr": 7.567254394150322e-05}
{"train_loss": 0.22157827019691467, "global_step": 89471, "epoch": 1005, "lr": 7.567204645679104e-05}
{"train_loss": 0.17805033922195435, "global_step": 89472, "epoch": 1005, "lr": 7.567154896862756e-05}
{"train_loss": 0.11348981410264969, "global_step": 89473, "epoch": 1005, "lr": 7.567105147701284e-05}
{"train_loss": 0.1965571939945221, "global_step": 89474, "epoch": 1005, "lr": 7.567055398194698e-05}
{"train_loss": 0.1670493185520172, "global_step": 89475, "epoch": 1005, "lr": 7.567005648343e-05}
{"train_loss": 0.18826602399349213, "global_step": 89476, "epoch": 1005, "lr": 7.566955898146201e-05}
{"train_loss": 0.15108533203601837, "global_step": 89477, "epoch": 1005, "lr": 7.566906147604307e-05}
{"train_loss": 0.21474617719650269, "global_step": 89478, "epoch": 1005, "lr": 7.566856396717322e-05}
{"train_loss": 0.16990092396736145, "global_step": 89479, "epoch": 1005, "lr": 7.566806645485256e-05}
{"train_loss": 0.19211402535438538, "global_step": 89480, "epoch": 1005, "lr": 7.566756893908113e-05}
{"train_loss": 0.14598941802978516, "global_step": 89481, "epoch": 1005, "lr": 7.566707141985902e-05}
{"train_loss": 0.1419237107038498, "global_step": 89482, "epoch": 1005, "lr": 7.566657389718627e-05}
{"train_loss": 0.19804081320762634, "global_step": 89483, "epoch": 1005, "lr": 7.566607637106296e-05}
{"train_loss": 0.2025107443332672, "global_step": 89484, "epoch": 1005, "lr": 7.566557884148917e-05}
{"train_loss": 0.13889366388320923, "global_step": 89485, "epoch": 1005, "lr": 7.566508130846494e-05}
{"train_loss": 0.2121201455593109, "global_step": 89486, "epoch": 1005, "lr": 7.566458377199038e-05}
{"train_loss": 0.18958431482315063, "global_step": 89487, "epoch": 1005, "lr": 7.566408623206551e-05}
{"train_loss": 0.1794283092021942, "global_step": 89488, "epoch": 1005, "lr": 7.566358868869043e-05}
{"train_loss": 0.21574385464191437, "global_step": 89489, "epoch": 1005, "lr": 7.566309114186518e-05}
{"train_loss": 0.15273116528987885, "global_step": 89490, "epoch": 1005, "lr": 7.566259359158984e-05}
{"train_loss": 0.16316355764865875, "global_step": 89491, "epoch": 1005, "lr": 7.566209603786449e-05}
{"train_loss": 0.1305881291627884, "global_step": 89492, "epoch": 1005, "lr": 7.566159848068918e-05}
{"train_loss": 0.11594630777835846, "global_step": 89493, "epoch": 1005, "lr": 7.566110092006398e-05}
{"train_loss": 0.22392232716083527, "global_step": 89494, "epoch": 1005, "lr": 7.566060335598894e-05}
{"train_loss": 0.12657319009304047, "global_step": 89495, "epoch": 1005, "lr": 7.566010578846416e-05}
{"train_loss": 0.13639339804649353, "global_step": 89496, "epoch": 1005, "lr": 7.56596082174897e-05}
{"train_loss": 0.1753595769405365, "global_step": 89497, "epoch": 1005, "lr": 7.565911064306561e-05}
{"train_loss": 0.23255176842212677, "global_step": 89498, "epoch": 1005, "lr": 7.565861306519197e-05}
{"train_loss": 0.14975787699222565, "global_step": 89499, "epoch": 1005, "lr": 7.565811548386883e-05}
{"train_loss": 0.1417645961046219, "global_step": 89500, "epoch": 1005, "lr": 7.56576178990963e-05}
{"train_loss": 0.15140758454799652, "global_step": 89501, "epoch": 1005, "lr": 7.565712031087437e-05}
{"train_loss": 0.15231677889823914, "global_step": 89502, "epoch": 1005, "lr": 7.565662271920317e-05}
{"train_loss": 0.13053885102272034, "global_step": 89503, "epoch": 1005, "lr": 7.565612512408276e-05}
{"train_loss": 0.09673713147640228, "global_step": 89504, "epoch": 1005, "lr": 7.565562752551318e-05}
{"train_loss": 0.15202276408672333, "global_step": 89505, "epoch": 1005, "lr": 7.565512992349452e-05}
{"train_loss": 0.13612815737724304, "global_step": 89506, "epoch": 1005, "lr": 7.565463231802685e-05}
{"train_loss": 0.14248740673065186, "global_step": 89507, "epoch": 1005, "lr": 7.565413470911023e-05}
{"train_loss": 0.15172812342643738, "global_step": 89508, "epoch": 1005, "lr": 7.56536370967447e-05}
{"train_loss": 0.2428552657365799, "global_step": 89509, "epoch": 1005, "lr": 7.565313948093038e-05}
{"train_loss": 0.16782422363758087, "global_step": 89510, "epoch": 1005, "lr": 7.565264186166729e-05}
{"train_loss": 0.14916373789310455, "global_step": 89511, "epoch": 1005, "lr": 7.565214423895552e-05}
{"train_loss": 0.3154829740524292, "global_step": 89512, "epoch": 1005, "lr": 7.565164661279514e-05}
{"train_loss": 0.21213626861572266, "global_step": 89513, "epoch": 1005, "lr": 7.56511489831862e-05}
{"train_loss": 0.14083331823349, "global_step": 89514, "epoch": 1005, "lr": 7.565065135012879e-05}
{"train_loss": 0.13070103526115417, "global_step": 89515, "epoch": 1005, "lr": 7.565015371362293e-05}
{"train_loss": 0.20025132596492767, "global_step": 89516, "epoch": 1005, "lr": 7.564965607366875e-05}
{"train_loss": 0.12066462635993958, "global_step": 89517, "epoch": 1005, "lr": 7.564915843026628e-05}
{"train_loss": 0.17155329883098602, "global_step": 89518, "epoch": 1005, "lr": 7.564866078341559e-05}
{"train_loss": 0.1565859615802765, "global_step": 89519, "epoch": 1005, "lr": 7.564816313311675e-05}
{"train_loss": 0.15379540622234344, "global_step": 89520, "epoch": 1005, "lr": 7.564766547936983e-05}
{"train_loss": 0.16483716666698456, "global_step": 89521, "epoch": 1005, "lr": 7.564716782217487e-05}
{"train_loss": 0.17122556269168854, "global_step": 89522, "epoch": 1005, "lr": 7.5646670161532e-05}
{"train_loss": 0.20119723677635193, "global_step": 89523, "epoch": 1005, "lr": 7.564617249744122e-05}
{"train_loss": 0.2209426909685135, "global_step": 89524, "epoch": 1005, "lr": 7.564567482990264e-05}
{"train_loss": 0.14662176370620728, "global_step": 89525, "epoch": 1005, "lr": 7.56451771589163e-05}
{"train_loss": 0.13783499598503113, "global_step": 89526, "epoch": 1005, "lr": 7.564467948448228e-05}
{"train_loss": 0.11795306950807571, "global_step": 89527, "epoch": 1005, "lr": 7.564418180660064e-05}
{"train_loss": 0.13970616459846497, "global_step": 89528, "epoch": 1005, "lr": 7.564368412527147e-05}
{"train_loss": 0.23796963691711426, "global_step": 89529, "epoch": 1005, "lr": 7.564318644049482e-05}
{"train_loss": 0.23414163291454315, "global_step": 89530, "epoch": 1005, "lr": 7.564268875227074e-05}
{"train_loss": 0.11227168142795563, "global_step": 89531, "epoch": 1005, "lr": 7.564219106059933e-05}
{"train_loss": 0.14426928758621216, "global_step": 89532, "epoch": 1005, "lr": 7.564169336548062e-05}
{"train_loss": 0.1711590728732977, "global_step": 89533, "epoch": 1005, "lr": 7.564119566691471e-05, "val_loss": 4.469101905822754, "train_action_mse_error": 20.885543823242188}
{"train_loss": 0.2232034057378769, "global_step": 89534, "epoch": 1006, "lr": 7.564069796490166e-05}
{"train_loss": 0.15589122474193573, "global_step": 89535, "epoch": 1006, "lr": 7.56402002594415e-05}
{"train_loss": 0.16161325573921204, "global_step": 89536, "epoch": 1006, "lr": 7.563970255053436e-05}
{"train_loss": 0.2769494354724884, "global_step": 89537, "epoch": 1006, "lr": 7.563920483818026e-05}
{"train_loss": 0.1629287749528885, "global_step": 89538, "epoch": 1006, "lr": 7.563870712237928e-05}
{"train_loss": 0.2513798773288727, "global_step": 89539, "epoch": 1006, "lr": 7.56382094031315e-05}
{"train_loss": 0.1250167042016983, "global_step": 89540, "epoch": 1006, "lr": 7.563771168043696e-05}
{"train_loss": 0.2514454126358032, "global_step": 89541, "epoch": 1006, "lr": 7.563721395429575e-05}
{"train_loss": 0.15751610696315765, "global_step": 89542, "epoch": 1006, "lr": 7.563671622470794e-05}
{"train_loss": 0.16765867173671722, "global_step": 89543, "epoch": 1006, "lr": 7.563621849167357e-05}
{"train_loss": 0.1339591145515442, "global_step": 89544, "epoch": 1006, "lr": 7.563572075519273e-05}
{"train_loss": 0.11477544903755188, "global_step": 89545, "epoch": 1006, "lr": 7.563522301526548e-05}
{"train_loss": 0.21009808778762817, "global_step": 89546, "epoch": 1006, "lr": 7.563472527189189e-05}
{"train_loss": 0.15947893261909485, "global_step": 89547, "epoch": 1006, "lr": 7.563422752507201e-05}
{"train_loss": 0.24414949119091034, "global_step": 89548, "epoch": 1006, "lr": 7.563372977480593e-05}
{"train_loss": 0.20215117931365967, "global_step": 89549, "epoch": 1006, "lr": 7.56332320210937e-05}
{"train_loss": 0.1386687308549881, "global_step": 89550, "epoch": 1006, "lr": 7.56327342639354e-05}
{"train_loss": 0.16100960969924927, "global_step": 89551, "epoch": 1006, "lr": 7.56322365033311e-05}
{"train_loss": 0.17232340574264526, "global_step": 89552, "epoch": 1006, "lr": 7.563173873928084e-05}
{"train_loss": 0.18883895874023438, "global_step": 89553, "epoch": 1006, "lr": 7.563124097178472e-05}
{"train_loss": 0.22731807827949524, "global_step": 89554, "epoch": 1006, "lr": 7.563074320084279e-05}
{"train_loss": 0.24205352365970612, "global_step": 89555, "epoch": 1006, "lr": 7.563024542645511e-05}
{"train_loss": 0.16972507536411285, "global_step": 89556, "epoch": 1006, "lr": 7.562974764862177e-05}
{"train_loss": 0.21518026292324066, "global_step": 89557, "epoch": 1006, "lr": 7.562924986734282e-05}
{"train_loss": 0.23250795900821686, "global_step": 89558, "epoch": 1006, "lr": 7.562875208261833e-05}
{"train_loss": 0.1646135449409485, "global_step": 89559, "epoch": 1006, "lr": 7.562825429444835e-05}
{"train_loss": 0.20488332211971283, "global_step": 89560, "epoch": 1006, "lr": 7.562775650283298e-05}
{"train_loss": 0.1958678960800171, "global_step": 89561, "epoch": 1006, "lr": 7.562725870777226e-05}
{"train_loss": 0.1420707106590271, "global_step": 89562, "epoch": 1006, "lr": 7.562676090926628e-05}
{"train_loss": 0.12002064287662506, "global_step": 89563, "epoch": 1006, "lr": 7.56262631073151e-05}
{"train_loss": 0.21815615892410278, "global_step": 89564, "epoch": 1006, "lr": 7.562576530191876e-05}
{"train_loss": 0.2103116810321808, "global_step": 89565, "epoch": 1006, "lr": 7.562526749307734e-05}
{"train_loss": 0.20226053893566132, "global_step": 89566, "epoch": 1006, "lr": 7.562476968079094e-05}
{"train_loss": 0.24758602678775787, "global_step": 89567, "epoch": 1006, "lr": 7.562427186505959e-05}
{"train_loss": 0.1431293785572052, "global_step": 89568, "epoch": 1006, "lr": 7.562377404588338e-05}
{"train_loss": 0.11784006655216217, "global_step": 89569, "epoch": 1006, "lr": 7.562327622326235e-05}
{"train_loss": 0.22330772876739502, "global_step": 89570, "epoch": 1006, "lr": 7.562277839719661e-05}
{"train_loss": 0.2099178433418274, "global_step": 89571, "epoch": 1006, "lr": 7.562228056768616e-05}
{"train_loss": 0.19442369043827057, "global_step": 89572, "epoch": 1006, "lr": 7.562178273473113e-05}
{"train_loss": 0.23416973650455475, "global_step": 89573, "epoch": 1006, "lr": 7.562128489833157e-05}
{"train_loss": 0.240648552775383, "global_step": 89574, "epoch": 1006, "lr": 7.562078705848753e-05}
{"train_loss": 0.2612341344356537, "global_step": 89575, "epoch": 1006, "lr": 7.562028921519908e-05}
{"train_loss": 0.21811437606811523, "global_step": 89576, "epoch": 1006, "lr": 7.56197913684663e-05}
{"train_loss": 0.15857629477977753, "global_step": 89577, "epoch": 1006, "lr": 7.561929351828927e-05}
{"train_loss": 0.20019395649433136, "global_step": 89578, "epoch": 1006, "lr": 7.561879566466802e-05}
{"train_loss": 0.15647146105766296, "global_step": 89579, "epoch": 1006, "lr": 7.561829780760263e-05}
{"train_loss": 0.1690349131822586, "global_step": 89580, "epoch": 1006, "lr": 7.561779994709319e-05}
{"train_loss": 0.21031524240970612, "global_step": 89581, "epoch": 1006, "lr": 7.561730208313974e-05}
{"train_loss": 0.1681404858827591, "global_step": 89582, "epoch": 1006, "lr": 7.561680421574236e-05}
{"train_loss": 0.2067849487066269, "global_step": 89583, "epoch": 1006, "lr": 7.561630634490111e-05}
{"train_loss": 0.2742535173892975, "global_step": 89584, "epoch": 1006, "lr": 7.561580847061604e-05}
{"train_loss": 0.1431604027748108, "global_step": 89585, "epoch": 1006, "lr": 7.561531059288727e-05}
{"train_loss": 0.1394117921590805, "global_step": 89586, "epoch": 1006, "lr": 7.561481271171482e-05}
{"train_loss": 0.16989168524742126, "global_step": 89587, "epoch": 1006, "lr": 7.561431482709877e-05}
{"train_loss": 0.19008661806583405, "global_step": 89588, "epoch": 1006, "lr": 7.561381693903919e-05}
{"train_loss": 0.22601424157619476, "global_step": 89589, "epoch": 1006, "lr": 7.561331904753613e-05}
{"train_loss": 0.18061839044094086, "global_step": 89590, "epoch": 1006, "lr": 7.561282115258969e-05}
{"train_loss": 0.12805046141147614, "global_step": 89591, "epoch": 1006, "lr": 7.561232325419992e-05}
{"train_loss": 0.2129509598016739, "global_step": 89592, "epoch": 1006, "lr": 7.561182535236686e-05}
{"train_loss": 0.16617713868618011, "global_step": 89593, "epoch": 1006, "lr": 7.561132744709063e-05}
{"train_loss": 0.17587995529174805, "global_step": 89594, "epoch": 1006, "lr": 7.561082953837125e-05}
{"train_loss": 0.16489477455615997, "global_step": 89595, "epoch": 1006, "lr": 7.561033162620882e-05}
{"train_loss": 0.1887759417295456, "global_step": 89596, "epoch": 1006, "lr": 7.560983371060339e-05}
{"train_loss": 0.1408311128616333, "global_step": 89597, "epoch": 1006, "lr": 7.560933579155503e-05}
{"train_loss": 0.23158708214759827, "global_step": 89598, "epoch": 1006, "lr": 7.56088378690638e-05}
{"train_loss": 0.21409811079502106, "global_step": 89599, "epoch": 1006, "lr": 7.560833994312977e-05}
{"train_loss": 0.21547125279903412, "global_step": 89600, "epoch": 1006, "lr": 7.560784201375302e-05}
{"train_loss": 0.16873565316200256, "global_step": 89601, "epoch": 1006, "lr": 7.560734408093362e-05}
{"train_loss": 0.13044258952140808, "global_step": 89602, "epoch": 1006, "lr": 7.560684614467159e-05}
{"train_loss": 0.22767296433448792, "global_step": 89603, "epoch": 1006, "lr": 7.560634820496706e-05}
{"train_loss": 0.10817976295948029, "global_step": 89604, "epoch": 1006, "lr": 7.560585026182006e-05}
{"train_loss": 0.1317261904478073, "global_step": 89605, "epoch": 1006, "lr": 7.560535231523067e-05}
{"train_loss": 0.16302257776260376, "global_step": 89606, "epoch": 1006, "lr": 7.560485436519895e-05}
{"train_loss": 0.17266324162483215, "global_step": 89607, "epoch": 1006, "lr": 7.560435641172498e-05}
{"train_loss": 0.1848105490207672, "global_step": 89608, "epoch": 1006, "lr": 7.56038584548088e-05}
{"train_loss": 0.27143406867980957, "global_step": 89609, "epoch": 1006, "lr": 7.56033604944505e-05}
{"train_loss": 0.18956296145915985, "global_step": 89610, "epoch": 1006, "lr": 7.560286253065015e-05}
{"train_loss": 0.10580332577228546, "global_step": 89611, "epoch": 1006, "lr": 7.56023645634078e-05}
{"train_loss": 0.18774673342704773, "global_step": 89612, "epoch": 1006, "lr": 7.560186659272351e-05}
{"train_loss": 0.1872914880514145, "global_step": 89613, "epoch": 1006, "lr": 7.560136861859737e-05}
{"train_loss": 0.19567468762397766, "global_step": 89614, "epoch": 1006, "lr": 7.560087064102943e-05}
{"train_loss": 0.16231374442577362, "global_step": 89615, "epoch": 1006, "lr": 7.560037266001978e-05}
{"train_loss": 0.19023734331130981, "global_step": 89616, "epoch": 1006, "lr": 7.559987467556846e-05}
{"train_loss": 0.23701730370521545, "global_step": 89617, "epoch": 1006, "lr": 7.559937668767556e-05}
{"train_loss": 0.18936291337013245, "global_step": 89618, "epoch": 1006, "lr": 7.559887869634113e-05}
{"train_loss": 0.19052189588546753, "global_step": 89619, "epoch": 1006, "lr": 7.559838070156525e-05}
{"train_loss": 0.18194246292114258, "global_step": 89620, "epoch": 1006, "lr": 7.559788270334797e-05}
{"train_loss": 0.19590170681476593, "global_step": 89621, "epoch": 1006, "lr": 7.559738470168937e-05}
{"train_loss": 0.18790660698092385, "global_step": 89622, "epoch": 1006, "lr": 7.559688669658951e-05, "val_loss": 4.249353885650635}
{"train_loss": 0.19689148664474487, "global_step": 89623, "epoch": 1007, "lr": 7.559638868804846e-05}
{"train_loss": 0.1261197328567505, "global_step": 89624, "epoch": 1007, "lr": 7.559589067606629e-05}
{"train_loss": 0.20363810658454895, "global_step": 89625, "epoch": 1007, "lr": 7.559539266064307e-05}
{"train_loss": 0.12842218577861786, "global_step": 89626, "epoch": 1007, "lr": 7.559489464177887e-05}
{"train_loss": 0.1422082483768463, "global_step": 89627, "epoch": 1007, "lr": 7.559439661947372e-05}
{"train_loss": 0.16650401055812836, "global_step": 89628, "epoch": 1007, "lr": 7.559389859372774e-05}
{"train_loss": 0.21857403218746185, "global_step": 89629, "epoch": 1007, "lr": 7.559340056454096e-05}
{"train_loss": 0.15292681753635406, "global_step": 89630, "epoch": 1007, "lr": 7.559290253191347e-05}
{"train_loss": 0.19677919149398804, "global_step": 89631, "epoch": 1007, "lr": 7.559240449584531e-05}
{"train_loss": 0.19201096892356873, "global_step": 89632, "epoch": 1007, "lr": 7.559190645633657e-05}
{"train_loss": 0.1992260217666626, "global_step": 89633, "epoch": 1007, "lr": 7.559140841338733e-05}
{"train_loss": 0.18310271203517914, "global_step": 89634, "epoch": 1007, "lr": 7.559091036699762e-05}
{"train_loss": 0.22438327968120575, "global_step": 89635, "epoch": 1007, "lr": 7.559041231716751e-05}
{"train_loss": 0.21778671443462372, "global_step": 89636, "epoch": 1007, "lr": 7.55899142638971e-05}
{"train_loss": 0.1475905478000641, "global_step": 89637, "epoch": 1007, "lr": 7.558941620718643e-05}
{"train_loss": 0.20043762028217316, "global_step": 89638, "epoch": 1007, "lr": 7.558891814703557e-05}
{"train_loss": 0.2050645798444748, "global_step": 89639, "epoch": 1007, "lr": 7.55884200834446e-05}
{"train_loss": 0.173309326171875, "global_step": 89640, "epoch": 1007, "lr": 7.558792201641358e-05}
{"train_loss": 0.23823007941246033, "global_step": 89641, "epoch": 1007, "lr": 7.558742394594258e-05}
{"train_loss": 0.1256982684135437, "global_step": 89642, "epoch": 1007, "lr": 7.558692587203165e-05}
{"train_loss": 0.19507679343223572, "global_step": 89643, "epoch": 1007, "lr": 7.558642779468088e-05}
{"train_loss": 0.21854032576084137, "global_step": 89644, "epoch": 1007, "lr": 7.558592971389031e-05}
{"train_loss": 0.1808573603630066, "global_step": 89645, "epoch": 1007, "lr": 7.558543162966005e-05}
{"train_loss": 0.11407394707202911, "global_step": 89646, "epoch": 1007, "lr": 7.558493354199012e-05}
{"train_loss": 0.16642799973487854, "global_step": 89647, "epoch": 1007, "lr": 7.558443545088061e-05}
{"train_loss": 0.21551916003227234, "global_step": 89648, "epoch": 1007, "lr": 7.55839373563316e-05}
{"train_loss": 0.16044504940509796, "global_step": 89649, "epoch": 1007, "lr": 7.558343925834313e-05}
{"train_loss": 0.1652929186820984, "global_step": 89650, "epoch": 1007, "lr": 7.558294115691528e-05}
{"train_loss": 0.23517632484436035, "global_step": 89651, "epoch": 1007, "lr": 7.558244305204812e-05}
{"train_loss": 0.26356977224349976, "global_step": 89652, "epoch": 1007, "lr": 7.558194494374168e-05}
{"train_loss": 0.09289409220218658, "global_step": 89653, "epoch": 1007, "lr": 7.558144683199611e-05}
{"train_loss": 0.22511397302150726, "global_step": 89654, "epoch": 1007, "lr": 7.55809487168114e-05}
{"train_loss": 0.15246504545211792, "global_step": 89655, "epoch": 1007, "lr": 7.558045059818763e-05}
{"train_loss": 0.17356248199939728, "global_step": 89656, "epoch": 1007, "lr": 7.557995247612491e-05}
{"train_loss": 0.10540125519037247, "global_step": 89657, "epoch": 1007, "lr": 7.557945435062326e-05}
{"train_loss": 0.18631385266780853, "global_step": 89658, "epoch": 1007, "lr": 7.557895622168277e-05}
{"train_loss": 0.15324777364730835, "global_step": 89659, "epoch": 1007, "lr": 7.55784580893035e-05}
{"train_loss": 0.1453603208065033, "global_step": 89660, "epoch": 1007, "lr": 7.557795995348552e-05}
{"train_loss": 0.24123121798038483, "global_step": 89661, "epoch": 1007, "lr": 7.557746181422888e-05}
{"train_loss": 0.21182318031787872, "global_step": 89662, "epoch": 1007, "lr": 7.557696367153367e-05}
{"train_loss": 0.11721981316804886, "global_step": 89663, "epoch": 1007, "lr": 7.557646552539995e-05}
{"train_loss": 0.1309334635734558, "global_step": 89664, "epoch": 1007, "lr": 7.557596737582778e-05}
{"train_loss": 0.14667461812496185, "global_step": 89665, "epoch": 1007, "lr": 7.557546922281724e-05}
{"train_loss": 0.17788784205913544, "global_step": 89666, "epoch": 1007, "lr": 7.55749710663684e-05}
{"train_loss": 0.2055637240409851, "global_step": 89667, "epoch": 1007, "lr": 7.557447290648131e-05}
{"train_loss": 0.13612520694732666, "global_step": 89668, "epoch": 1007, "lr": 7.557397474315604e-05}
{"train_loss": 0.1622850000858307, "global_step": 89669, "epoch": 1007, "lr": 7.557347657639266e-05}
{"train_loss": 0.24232347309589386, "global_step": 89670, "epoch": 1007, "lr": 7.557297840619124e-05}
{"train_loss": 0.1609954982995987, "global_step": 89671, "epoch": 1007, "lr": 7.557248023255185e-05}
{"train_loss": 0.11038340628147125, "global_step": 89672, "epoch": 1007, "lr": 7.557198205547454e-05}
{"train_loss": 0.15172027051448822, "global_step": 89673, "epoch": 1007, "lr": 7.55714838749594e-05}
{"train_loss": 0.14520150423049927, "global_step": 89674, "epoch": 1007, "lr": 7.557098569100647e-05}
{"train_loss": 0.23601163923740387, "global_step": 89675, "epoch": 1007, "lr": 7.557048750361585e-05}
{"train_loss": 0.15341848134994507, "global_step": 89676, "epoch": 1007, "lr": 7.556998931278758e-05}
{"train_loss": 0.14918352663516998, "global_step": 89677, "epoch": 1007, "lr": 7.556949111852174e-05}
{"train_loss": 0.16008572280406952, "global_step": 89678, "epoch": 1007, "lr": 7.55689929208184e-05}
{"train_loss": 0.1643487513065338, "global_step": 89679, "epoch": 1007, "lr": 7.556849471967762e-05}
{"train_loss": 0.15442249178886414, "global_step": 89680, "epoch": 1007, "lr": 7.556799651509945e-05}
{"train_loss": 0.16328606009483337, "global_step": 89681, "epoch": 1007, "lr": 7.556749830708398e-05}
{"train_loss": 0.16285249590873718, "global_step": 89682, "epoch": 1007, "lr": 7.556700009563127e-05}
{"train_loss": 0.1317255049943924, "global_step": 89683, "epoch": 1007, "lr": 7.556650188074141e-05}
{"train_loss": 0.11036592721939087, "global_step": 89684, "epoch": 1007, "lr": 7.556600366241442e-05}
{"train_loss": 0.17736472189426422, "global_step": 89685, "epoch": 1007, "lr": 7.55655054406504e-05}
{"train_loss": 0.11299573630094528, "global_step": 89686, "epoch": 1007, "lr": 7.556500721544943e-05}
{"train_loss": 0.1266525536775589, "global_step": 89687, "epoch": 1007, "lr": 7.556450898681153e-05}
{"train_loss": 0.23193536698818207, "global_step": 89688, "epoch": 1007, "lr": 7.55640107547368e-05}
{"train_loss": 0.14134427905082703, "global_step": 89689, "epoch": 1007, "lr": 7.556351251922531e-05}
{"train_loss": 0.16160714626312256, "global_step": 89690, "epoch": 1007, "lr": 7.556301428027711e-05}
{"train_loss": 0.19202174246311188, "global_step": 89691, "epoch": 1007, "lr": 7.556251603789229e-05}
{"train_loss": 0.19888687133789062, "global_step": 89692, "epoch": 1007, "lr": 7.556201779207087e-05}
{"train_loss": 0.1673944741487503, "global_step": 89693, "epoch": 1007, "lr": 7.556151954281296e-05}
{"train_loss": 0.15448254346847534, "global_step": 89694, "epoch": 1007, "lr": 7.556102129011862e-05}
{"train_loss": 0.19119124114513397, "global_step": 89695, "epoch": 1007, "lr": 7.556052303398792e-05}
{"train_loss": 0.10316751897335052, "global_step": 89696, "epoch": 1007, "lr": 7.55600247744209e-05}
{"train_loss": 0.17564797401428223, "global_step": 89697, "epoch": 1007, "lr": 7.555952651141766e-05}
{"train_loss": 0.13166269659996033, "global_step": 89698, "epoch": 1007, "lr": 7.555902824497826e-05}
{"train_loss": 0.19179069995880127, "global_step": 89699, "epoch": 1007, "lr": 7.555852997510275e-05}
{"train_loss": 0.2203531563282013, "global_step": 89700, "epoch": 1007, "lr": 7.555803170179119e-05}
{"train_loss": 0.2082727998495102, "global_step": 89701, "epoch": 1007, "lr": 7.555753342504367e-05}
{"train_loss": 0.14479771256446838, "global_step": 89702, "epoch": 1007, "lr": 7.555703514486028e-05}
{"train_loss": 0.283156156539917, "global_step": 89703, "epoch": 1007, "lr": 7.555653686124104e-05}
{"train_loss": 0.2724606990814209, "global_step": 89704, "epoch": 1007, "lr": 7.555603857418603e-05}
{"train_loss": 0.1801849603652954, "global_step": 89705, "epoch": 1007, "lr": 7.555554028369534e-05}
{"train_loss": 0.15280084311962128, "global_step": 89706, "epoch": 1007, "lr": 7.555504198976902e-05}
{"train_loss": 0.14393523335456848, "global_step": 89707, "epoch": 1007, "lr": 7.555454369240711e-05}
{"train_loss": 0.12521903216838837, "global_step": 89708, "epoch": 1007, "lr": 7.555404539160972e-05}
{"train_loss": 0.17503373324871063, "global_step": 89709, "epoch": 1007, "lr": 7.555354708737689e-05}
{"train_loss": 0.19978612661361694, "global_step": 89710, "epoch": 1007, "lr": 7.55530487797087e-05}
{"train_loss": 0.17317994667238065, "global_step": 89711, "epoch": 1007, "lr": 7.555255046860523e-05, "val_loss": 4.28162145614624}
{"train_loss": 0.12757448852062225, "global_step": 89712, "epoch": 1008, "lr": 7.55520521540665e-05}
{"train_loss": 0.2899564802646637, "global_step": 89713, "epoch": 1008, "lr": 7.555155383609265e-05}
{"train_loss": 0.18513844907283783, "global_step": 89714, "epoch": 1008, "lr": 7.555105551468367e-05}
{"train_loss": 0.2096787691116333, "global_step": 89715, "epoch": 1008, "lr": 7.555055718983967e-05}
{"train_loss": 0.16514264047145844, "global_step": 89716, "epoch": 1008, "lr": 7.555005886156072e-05}
{"train_loss": 0.1928204596042633, "global_step": 89717, "epoch": 1008, "lr": 7.554956052984688e-05}
{"train_loss": 0.1462346762418747, "global_step": 89718, "epoch": 1008, "lr": 7.55490621946982e-05}
{"train_loss": 0.11684083938598633, "global_step": 89719, "epoch": 1008, "lr": 7.554856385611474e-05}
{"train_loss": 0.20419564843177795, "global_step": 89720, "epoch": 1008, "lr": 7.55480655140966e-05}
{"train_loss": 0.17762857675552368, "global_step": 89721, "epoch": 1008, "lr": 7.554756716864385e-05}
{"train_loss": 0.15673889219760895, "global_step": 89722, "epoch": 1008, "lr": 7.554706881975653e-05}
{"train_loss": 0.10778700560331345, "global_step": 89723, "epoch": 1008, "lr": 7.554657046743474e-05}
{"train_loss": 0.16965489089488983, "global_step": 89724, "epoch": 1008, "lr": 7.55460721116785e-05}
{"train_loss": 0.2074369639158249, "global_step": 89725, "epoch": 1008, "lr": 7.554557375248793e-05}
{"train_loss": 0.16941313445568085, "global_step": 89726, "epoch": 1008, "lr": 7.554507538986304e-05}
{"train_loss": 0.12662193179130554, "global_step": 89727, "epoch": 1008, "lr": 7.554457702380392e-05}
{"train_loss": 0.11804014444351196, "global_step": 89728, "epoch": 1008, "lr": 7.554407865431065e-05}
{"train_loss": 0.2286440134048462, "global_step": 89729, "epoch": 1008, "lr": 7.55435802813833e-05}
{"train_loss": 0.22818322479724884, "global_step": 89730, "epoch": 1008, "lr": 7.554308190502193e-05}
{"train_loss": 0.21658955514431, "global_step": 89731, "epoch": 1008, "lr": 7.55425835252266e-05}
{"train_loss": 0.1578095257282257, "global_step": 89732, "epoch": 1008, "lr": 7.554208514199736e-05}
{"train_loss": 0.19470517337322235, "global_step": 89733, "epoch": 1008, "lr": 7.554158675533434e-05}
{"train_loss": 0.17816559970378876, "global_step": 89734, "epoch": 1008, "lr": 7.554108836523752e-05}
{"train_loss": 0.18873295187950134, "global_step": 89735, "epoch": 1008, "lr": 7.554058997170704e-05}
{"train_loss": 0.09737985581159592, "global_step": 89736, "epoch": 1008, "lr": 7.554009157474293e-05}
{"train_loss": 0.16811765730381012, "global_step": 89737, "epoch": 1008, "lr": 7.553959317434526e-05}
{"train_loss": 0.25566643476486206, "global_step": 89738, "epoch": 1008, "lr": 7.55390947705141e-05}
{"train_loss": 0.16027116775512695, "global_step": 89739, "epoch": 1008, "lr": 7.553859636324955e-05}
{"train_loss": 0.09888966381549835, "global_step": 89740, "epoch": 1008, "lr": 7.553809795255162e-05}
{"train_loss": 0.10216102749109268, "global_step": 89741, "epoch": 1008, "lr": 7.55375995384204e-05}
{"train_loss": 0.099073126912117, "global_step": 89742, "epoch": 1008, "lr": 7.5537101120856e-05}
{"train_loss": 0.1519545465707779, "global_step": 89743, "epoch": 1008, "lr": 7.553660269985841e-05}
{"train_loss": 0.17470316588878632, "global_step": 89744, "epoch": 1008, "lr": 7.553610427542776e-05}
{"train_loss": 0.15137633681297302, "global_step": 89745, "epoch": 1008, "lr": 7.553560584756407e-05}
{"train_loss": 0.26096466183662415, "global_step": 89746, "epoch": 1008, "lr": 7.553510741626744e-05}
{"train_loss": 0.15095825493335724, "global_step": 89747, "epoch": 1008, "lr": 7.553460898153793e-05}
{"train_loss": 0.23481479287147522, "global_step": 89748, "epoch": 1008, "lr": 7.553411054337559e-05}
{"train_loss": 0.2419901192188263, "global_step": 89749, "epoch": 1008, "lr": 7.55336121017805e-05}
{"train_loss": 0.1779092252254486, "global_step": 89750, "epoch": 1008, "lr": 7.553311365675273e-05}
{"train_loss": 0.12675292789936066, "global_step": 89751, "epoch": 1008, "lr": 7.553261520829236e-05}
{"train_loss": 0.16150227189064026, "global_step": 89752, "epoch": 1008, "lr": 7.553211675639942e-05}
{"train_loss": 0.11524885147809982, "global_step": 89753, "epoch": 1008, "lr": 7.5531618301074e-05}
{"train_loss": 0.18272247910499573, "global_step": 89754, "epoch": 1008, "lr": 7.553111984231618e-05}
{"train_loss": 0.3094545006752014, "global_step": 89755, "epoch": 1008, "lr": 7.553062138012599e-05}
{"train_loss": 0.2699367105960846, "global_step": 89756, "epoch": 1008, "lr": 7.553012291450354e-05}
{"train_loss": 0.2637752592563629, "global_step": 89757, "epoch": 1008, "lr": 7.552962444544887e-05}
{"train_loss": 0.18485264480113983, "global_step": 89758, "epoch": 1008, "lr": 7.552912597296204e-05}
{"train_loss": 0.2348616123199463, "global_step": 89759, "epoch": 1008, "lr": 7.552862749704315e-05}
{"train_loss": 0.11288673430681229, "global_step": 89760, "epoch": 1008, "lr": 7.552812901769224e-05}
{"train_loss": 0.1561209112405777, "global_step": 89761, "epoch": 1008, "lr": 7.552763053490938e-05}
{"train_loss": 0.18662607669830322, "global_step": 89762, "epoch": 1008, "lr": 7.552713204869465e-05}
{"train_loss": 0.269370973110199, "global_step": 89763, "epoch": 1008, "lr": 7.55266335590481e-05}
{"train_loss": 0.25390854477882385, "global_step": 89764, "epoch": 1008, "lr": 7.552613506596981e-05}
{"train_loss": 0.20548659563064575, "global_step": 89765, "epoch": 1008, "lr": 7.552563656945986e-05}
{"train_loss": 0.19858649373054504, "global_step": 89766, "epoch": 1008, "lr": 7.552513806951827e-05}
{"train_loss": 0.16996383666992188, "global_step": 89767, "epoch": 1008, "lr": 7.552463956614515e-05}
{"train_loss": 0.17001183331012726, "global_step": 89768, "epoch": 1008, "lr": 7.552414105934055e-05}
{"train_loss": 0.12296219915151596, "global_step": 89769, "epoch": 1008, "lr": 7.552364254910453e-05}
{"train_loss": 0.16129539906978607, "global_step": 89770, "epoch": 1008, "lr": 7.552314403543718e-05}
{"train_loss": 0.15766656398773193, "global_step": 89771, "epoch": 1008, "lr": 7.552264551833856e-05}
{"train_loss": 0.1272868812084198, "global_step": 89772, "epoch": 1008, "lr": 7.552214699780871e-05}
{"train_loss": 0.1490849405527115, "global_step": 89773, "epoch": 1008, "lr": 7.552164847384772e-05}
{"train_loss": 0.20286831259727478, "global_step": 89774, "epoch": 1008, "lr": 7.552114994645567e-05}
{"train_loss": 0.24897883832454681, "global_step": 89775, "epoch": 1008, "lr": 7.552065141563261e-05}
{"train_loss": 0.2590307593345642, "global_step": 89776, "epoch": 1008, "lr": 7.552015288137861e-05}
{"train_loss": 0.10341502726078033, "global_step": 89777, "epoch": 1008, "lr": 7.551965434369375e-05}
{"train_loss": 0.15790338814258575, "global_step": 89778, "epoch": 1008, "lr": 7.551915580257805e-05}
{"train_loss": 0.25569286942481995, "global_step": 89779, "epoch": 1008, "lr": 7.551865725803164e-05}
{"train_loss": 0.256052166223526, "global_step": 89780, "epoch": 1008, "lr": 7.551815871005454e-05}
{"train_loss": 0.12404817342758179, "global_step": 89781, "epoch": 1008, "lr": 7.551766015864685e-05}
{"train_loss": 0.14920948445796967, "global_step": 89782, "epoch": 1008, "lr": 7.551716160380861e-05}
{"train_loss": 0.22749564051628113, "global_step": 89783, "epoch": 1008, "lr": 7.55166630455399e-05}
{"train_loss": 0.16541725397109985, "global_step": 89784, "epoch": 1008, "lr": 7.55161644838408e-05}
{"train_loss": 0.1695118546485901, "global_step": 89785, "epoch": 1008, "lr": 7.551566591871135e-05}
{"train_loss": 0.22966675460338593, "global_step": 89786, "epoch": 1008, "lr": 7.551516735015162e-05}
{"train_loss": 0.1361846923828125, "global_step": 89787, "epoch": 1008, "lr": 7.55146687781617e-05}
{"train_loss": 0.1725001335144043, "global_step": 89788, "epoch": 1008, "lr": 7.551417020274164e-05}
{"train_loss": 0.191014364361763, "global_step": 89789, "epoch": 1008, "lr": 7.551367162389152e-05}
{"train_loss": 0.09550550580024719, "global_step": 89790, "epoch": 1008, "lr": 7.551317304161139e-05}
{"train_loss": 0.1897125393152237, "global_step": 89791, "epoch": 1008, "lr": 7.551267445590132e-05}
{"train_loss": 0.21292412281036377, "global_step": 89792, "epoch": 1008, "lr": 7.551217586676137e-05}
{"train_loss": 0.12075650691986084, "global_step": 89793, "epoch": 1008, "lr": 7.551167727419163e-05}
{"train_loss": 0.20743829011917114, "global_step": 89794, "epoch": 1008, "lr": 7.551117867819217e-05}
{"train_loss": 0.16823485493659973, "global_step": 89795, "epoch": 1008, "lr": 7.551068007876303e-05}
{"train_loss": 0.23229557275772095, "global_step": 89796, "epoch": 1008, "lr": 7.551018147590429e-05}
{"train_loss": 0.16579285264015198, "global_step": 89797, "epoch": 1008, "lr": 7.550968286961602e-05}
{"train_loss": 0.10758933424949646, "global_step": 89798, "epoch": 1008, "lr": 7.550918425989829e-05}
{"train_loss": 0.1652991771697998, "global_step": 89799, "epoch": 1008, "lr": 7.550868564675115e-05}
{"train_loss": 0.1801680220981662, "global_step": 89800, "epoch": 1008, "lr": 7.550818703017468e-05, "val_loss": 4.364258289337158}
{"train_loss": 0.26782405376434326, "global_step": 89801, "epoch": 1009, "lr": 7.550768841016895e-05}
{"train_loss": 0.14198173582553864, "global_step": 89802, "epoch": 1009, "lr": 7.550718978673404e-05}
{"train_loss": 0.09697487205266953, "global_step": 89803, "epoch": 1009, "lr": 7.550669115986997e-05}
{"train_loss": 0.1505638062953949, "global_step": 89804, "epoch": 1009, "lr": 7.550619252957684e-05}
{"train_loss": 0.16809584200382233, "global_step": 89805, "epoch": 1009, "lr": 7.550569389585473e-05}
{"train_loss": 0.11767660826444626, "global_step": 89806, "epoch": 1009, "lr": 7.550519525870367e-05}
{"train_loss": 0.199004665017128, "global_step": 89807, "epoch": 1009, "lr": 7.550469661812377e-05}
{"train_loss": 0.2006671130657196, "global_step": 89808, "epoch": 1009, "lr": 7.550419797411505e-05}
{"train_loss": 0.1373216211795807, "global_step": 89809, "epoch": 1009, "lr": 7.55036993266776e-05}
{"train_loss": 0.21100416779518127, "global_step": 89810, "epoch": 1009, "lr": 7.550320067581151e-05}
{"train_loss": 0.19553029537200928, "global_step": 89811, "epoch": 1009, "lr": 7.550270202151683e-05}
{"train_loss": 0.13439694046974182, "global_step": 89812, "epoch": 1009, "lr": 7.550220336379362e-05}
{"train_loss": 0.27647003531455994, "global_step": 89813, "epoch": 1009, "lr": 7.550170470264192e-05}
{"train_loss": 0.1332501620054245, "global_step": 89814, "epoch": 1009, "lr": 7.550120603806185e-05}
{"train_loss": 0.23253388702869415, "global_step": 89815, "epoch": 1009, "lr": 7.550070737005345e-05}
{"train_loss": 0.24992790818214417, "global_step": 89816, "epoch": 1009, "lr": 7.550020869861678e-05}
{"train_loss": 0.13363099098205566, "global_step": 89817, "epoch": 1009, "lr": 7.549971002375194e-05}
{"train_loss": 0.10986393690109253, "global_step": 89818, "epoch": 1009, "lr": 7.549921134545896e-05}
{"train_loss": 0.16990144550800323, "global_step": 89819, "epoch": 1009, "lr": 7.549871266373792e-05}
{"train_loss": 0.23586638271808624, "global_step": 89820, "epoch": 1009, "lr": 7.549821397858889e-05}
{"train_loss": 0.10525457561016083, "global_step": 89821, "epoch": 1009, "lr": 7.549771529001195e-05}
{"train_loss": 0.18996980786323547, "global_step": 89822, "epoch": 1009, "lr": 7.549721659800715e-05}
{"train_loss": 0.1920003592967987, "global_step": 89823, "epoch": 1009, "lr": 7.549671790257456e-05}
{"train_loss": 0.18942779302597046, "global_step": 89824, "epoch": 1009, "lr": 7.549621920371423e-05}
{"train_loss": 0.25499609112739563, "global_step": 89825, "epoch": 1009, "lr": 7.549572050142626e-05}
{"train_loss": 0.12766814231872559, "global_step": 89826, "epoch": 1009, "lr": 7.549522179571068e-05}
{"train_loss": 0.1670304834842682, "global_step": 89827, "epoch": 1009, "lr": 7.54947230865676e-05}
{"train_loss": 0.22524358332157135, "global_step": 89828, "epoch": 1009, "lr": 7.549422437399705e-05}
{"train_loss": 0.1463848054409027, "global_step": 89829, "epoch": 1009, "lr": 7.549372565799912e-05}
{"train_loss": 0.1871427595615387, "global_step": 89830, "epoch": 1009, "lr": 7.549322693857387e-05}
{"train_loss": 0.18313293159008026, "global_step": 89831, "epoch": 1009, "lr": 7.549272821572137e-05}
{"train_loss": 0.20865283906459808, "global_step": 89832, "epoch": 1009, "lr": 7.549222948944168e-05}
{"train_loss": 0.20871713757514954, "global_step": 89833, "epoch": 1009, "lr": 7.549173075973487e-05}
{"train_loss": 0.1270604431629181, "global_step": 89834, "epoch": 1009, "lr": 7.5491232026601e-05}
{"train_loss": 0.23130318522453308, "global_step": 89835, "epoch": 1009, "lr": 7.549073329004015e-05}
{"train_loss": 0.22139573097229004, "global_step": 89836, "epoch": 1009, "lr": 7.549023455005239e-05}
{"train_loss": 0.11458103358745575, "global_step": 89837, "epoch": 1009, "lr": 7.548973580663777e-05}
{"train_loss": 0.17531684041023254, "global_step": 89838, "epoch": 1009, "lr": 7.548923705979638e-05}
{"train_loss": 0.2292124629020691, "global_step": 89839, "epoch": 1009, "lr": 7.548873830952825e-05}
{"train_loss": 0.11824508756399155, "global_step": 89840, "epoch": 1009, "lr": 7.548823955583348e-05}
{"train_loss": 0.17930860817432404, "global_step": 89841, "epoch": 1009, "lr": 7.548774079871213e-05}
{"train_loss": 0.1298971176147461, "global_step": 89842, "epoch": 1009, "lr": 7.548724203816427e-05}
{"train_loss": 0.22304564714431763, "global_step": 89843, "epoch": 1009, "lr": 7.548674327418995e-05}
{"train_loss": 0.16638438403606415, "global_step": 89844, "epoch": 1009, "lr": 7.548624450678925e-05}
{"train_loss": 0.17299596965312958, "global_step": 89845, "epoch": 1009, "lr": 7.548574573596223e-05}
{"train_loss": 0.23419496417045593, "global_step": 89846, "epoch": 1009, "lr": 7.548524696170895e-05}
{"train_loss": 0.160847470164299, "global_step": 89847, "epoch": 1009, "lr": 7.548474818402951e-05}
{"train_loss": 0.2371782660484314, "global_step": 89848, "epoch": 1009, "lr": 7.548424940292396e-05}
{"train_loss": 0.20947782695293427, "global_step": 89849, "epoch": 1009, "lr": 7.548375061839236e-05}
{"train_loss": 0.12125059217214584, "global_step": 89850, "epoch": 1009, "lr": 7.548325183043477e-05}
{"train_loss": 0.1281786561012268, "global_step": 89851, "epoch": 1009, "lr": 7.548275303905127e-05}
{"train_loss": 0.22010773420333862, "global_step": 89852, "epoch": 1009, "lr": 7.548225424424193e-05}
{"train_loss": 0.11497978121042252, "global_step": 89853, "epoch": 1009, "lr": 7.548175544600682e-05}
{"train_loss": 0.17547085881233215, "global_step": 89854, "epoch": 1009, "lr": 7.548125664434597e-05}
{"train_loss": 0.17687463760375977, "global_step": 89855, "epoch": 1009, "lr": 7.54807578392595e-05}
{"train_loss": 0.16646072268486023, "global_step": 89856, "epoch": 1009, "lr": 7.548025903074744e-05}
{"train_loss": 0.18634523451328278, "global_step": 89857, "epoch": 1009, "lr": 7.547976021880989e-05}
{"train_loss": 0.14374449849128723, "global_step": 89858, "epoch": 1009, "lr": 7.547926140344688e-05}
{"train_loss": 0.17451153695583344, "global_step": 89859, "epoch": 1009, "lr": 7.54787625846585e-05}
{"train_loss": 0.19136269390583038, "global_step": 89860, "epoch": 1009, "lr": 7.54782637624448e-05}
{"train_loss": 0.12729550898075104, "global_step": 89861, "epoch": 1009, "lr": 7.547776493680588e-05}
{"train_loss": 0.1592881828546524, "global_step": 89862, "epoch": 1009, "lr": 7.547726610774177e-05}
{"train_loss": 0.236272931098938, "global_step": 89863, "epoch": 1009, "lr": 7.547676727525256e-05}
{"train_loss": 0.1537419706583023, "global_step": 89864, "epoch": 1009, "lr": 7.547626843933828e-05}
{"train_loss": 0.19196821749210358, "global_step": 89865, "epoch": 1009, "lr": 7.547576959999906e-05}
{"train_loss": 0.16065016388893127, "global_step": 89866, "epoch": 1009, "lr": 7.547527075723492e-05}
{"train_loss": 0.12014085054397583, "global_step": 89867, "epoch": 1009, "lr": 7.547477191104594e-05}
{"train_loss": 0.20192255079746246, "global_step": 89868, "epoch": 1009, "lr": 7.547427306143219e-05}
{"train_loss": 0.15886925160884857, "global_step": 89869, "epoch": 1009, "lr": 7.547377420839374e-05}
{"train_loss": 0.1396634876728058, "global_step": 89870, "epoch": 1009, "lr": 7.547327535193064e-05}
{"train_loss": 0.13167177140712738, "global_step": 89871, "epoch": 1009, "lr": 7.547277649204297e-05}
{"train_loss": 0.2151164561510086, "global_step": 89872, "epoch": 1009, "lr": 7.547227762873081e-05}
{"train_loss": 0.19753998517990112, "global_step": 89873, "epoch": 1009, "lr": 7.547177876199419e-05}
{"train_loss": 0.1734985113143921, "global_step": 89874, "epoch": 1009, "lr": 7.547127989183323e-05}
{"train_loss": 0.1505589336156845, "global_step": 89875, "epoch": 1009, "lr": 7.547078101824794e-05}
{"train_loss": 0.1815165877342224, "global_step": 89876, "epoch": 1009, "lr": 7.547028214123841e-05}
{"train_loss": 0.22934354841709137, "global_step": 89877, "epoch": 1009, "lr": 7.546978326080475e-05}
{"train_loss": 0.25960999727249146, "global_step": 89878, "epoch": 1009, "lr": 7.546928437694694e-05}
{"train_loss": 0.20403292775154114, "global_step": 89879, "epoch": 1009, "lr": 7.546878548966512e-05}
{"train_loss": 0.17166215181350708, "global_step": 89880, "epoch": 1009, "lr": 7.546828659895934e-05}
{"train_loss": 0.16645246744155884, "global_step": 89881, "epoch": 1009, "lr": 7.546778770482965e-05}
{"train_loss": 0.08775729686021805, "global_step": 89882, "epoch": 1009, "lr": 7.546728880727611e-05}
{"train_loss": 0.17270871996879578, "global_step": 89883, "epoch": 1009, "lr": 7.546678990629883e-05}
{"train_loss": 0.14264346659183502, "global_step": 89884, "epoch": 1009, "lr": 7.546629100189784e-05}
{"train_loss": 0.19661231338977814, "global_step": 89885, "epoch": 1009, "lr": 7.54657920940732e-05}
{"train_loss": 0.11638416349887848, "global_step": 89886, "epoch": 1009, "lr": 7.546529318282502e-05}
{"train_loss": 0.1380908191204071, "global_step": 89887, "epoch": 1009, "lr": 7.546479426815334e-05}
{"train_loss": 0.2053574025630951, "global_step": 89888, "epoch": 1009, "lr": 7.546429535005821e-05}
{"train_loss": 0.17518573480375696, "global_step": 89889, "epoch": 1009, "lr": 7.546379642853974e-05, "val_loss": 4.443680763244629}
{"train_loss": 0.15264980494976044, "global_step": 89890, "epoch": 1010, "lr": 7.546329750359795e-05}
{"train_loss": 0.17173944413661957, "global_step": 89891, "epoch": 1010, "lr": 7.546279857523294e-05}
{"train_loss": 0.2253400832414627, "global_step": 89892, "epoch": 1010, "lr": 7.546229964344477e-05}
{"train_loss": 0.1261962205171585, "global_step": 89893, "epoch": 1010, "lr": 7.54618007082335e-05}
{"train_loss": 0.22032858431339264, "global_step": 89894, "epoch": 1010, "lr": 7.54613017695992e-05}
{"train_loss": 0.08877236396074295, "global_step": 89895, "epoch": 1010, "lr": 7.546080282754194e-05}
{"train_loss": 0.12724646925926208, "global_step": 89896, "epoch": 1010, "lr": 7.546030388206179e-05}
{"train_loss": 0.18354959785938263, "global_step": 89897, "epoch": 1010, "lr": 7.54598049331588e-05}
{"train_loss": 0.16971205174922943, "global_step": 89898, "epoch": 1010, "lr": 7.545930598083306e-05}
{"train_loss": 0.16796451807022095, "global_step": 89899, "epoch": 1010, "lr": 7.545880702508464e-05}
{"train_loss": 0.12845554947853088, "global_step": 89900, "epoch": 1010, "lr": 7.545830806591358e-05}
{"train_loss": 0.20622365176677704, "global_step": 89901, "epoch": 1010, "lr": 7.545780910331995e-05}
{"train_loss": 0.19236156344413757, "global_step": 89902, "epoch": 1010, "lr": 7.545731013730384e-05}
{"train_loss": 0.23547649383544922, "global_step": 89903, "epoch": 1010, "lr": 7.54568111678653e-05}
{"train_loss": 0.2555813789367676, "global_step": 89904, "epoch": 1010, "lr": 7.54563121950044e-05}
{"train_loss": 0.184585839509964, "global_step": 89905, "epoch": 1010, "lr": 7.545581321872122e-05}
{"train_loss": 0.16354888677597046, "global_step": 89906, "epoch": 1010, "lr": 7.545531423901582e-05}
{"train_loss": 0.19637152552604675, "global_step": 89907, "epoch": 1010, "lr": 7.545481525588824e-05}
{"train_loss": 0.1495201736688614, "global_step": 89908, "epoch": 1010, "lr": 7.545431626933859e-05}
{"train_loss": 0.19997677206993103, "global_step": 89909, "epoch": 1010, "lr": 7.545381727936691e-05}
{"train_loss": 0.2022051066160202, "global_step": 89910, "epoch": 1010, "lr": 7.545331828597327e-05}
{"train_loss": 0.18555884063243866, "global_step": 89911, "epoch": 1010, "lr": 7.545281928915775e-05}
{"train_loss": 0.08713763952255249, "global_step": 89912, "epoch": 1010, "lr": 7.54523202889204e-05}
{"train_loss": 0.2770484983921051, "global_step": 89913, "epoch": 1010, "lr": 7.545182128526131e-05}
{"train_loss": 0.17917990684509277, "global_step": 89914, "epoch": 1010, "lr": 7.545132227818052e-05}
{"train_loss": 0.16783200204372406, "global_step": 89915, "epoch": 1010, "lr": 7.545082326767811e-05}
{"train_loss": 0.24375736713409424, "global_step": 89916, "epoch": 1010, "lr": 7.545032425375415e-05}
{"train_loss": 0.19648101925849915, "global_step": 89917, "epoch": 1010, "lr": 7.544982523640871e-05}
{"train_loss": 0.16178162395954132, "global_step": 89918, "epoch": 1010, "lr": 7.544932621564185e-05}
{"train_loss": 0.16497313976287842, "global_step": 89919, "epoch": 1010, "lr": 7.544882719145363e-05}
{"train_loss": 0.17550840973854065, "global_step": 89920, "epoch": 1010, "lr": 7.544832816384413e-05}
{"train_loss": 0.19385768473148346, "global_step": 89921, "epoch": 1010, "lr": 7.54478291328134e-05}
{"train_loss": 0.13407789170742035, "global_step": 89922, "epoch": 1010, "lr": 7.544733009836155e-05}
{"train_loss": 0.17026032507419586, "global_step": 89923, "epoch": 1010, "lr": 7.54468310604886e-05}
{"train_loss": 0.21127350628376007, "global_step": 89924, "epoch": 1010, "lr": 7.544633201919462e-05}
{"train_loss": 0.16220669448375702, "global_step": 89925, "epoch": 1010, "lr": 7.544583297447971e-05}
{"train_loss": 0.11416706442832947, "global_step": 89926, "epoch": 1010, "lr": 7.544533392634391e-05}
{"train_loss": 0.14045482873916626, "global_step": 89927, "epoch": 1010, "lr": 7.54448348747873e-05}
{"train_loss": 0.19819049537181854, "global_step": 89928, "epoch": 1010, "lr": 7.544433581980993e-05}
{"train_loss": 0.17753979563713074, "global_step": 89929, "epoch": 1010, "lr": 7.544383676141188e-05}
{"train_loss": 0.138742133975029, "global_step": 89930, "epoch": 1010, "lr": 7.544333769959324e-05}
{"train_loss": 0.11806245893239975, "global_step": 89931, "epoch": 1010, "lr": 7.544283863435403e-05}
{"train_loss": 0.21031703054904938, "global_step": 89932, "epoch": 1010, "lr": 7.544233956569436e-05}
{"train_loss": 0.24059250950813293, "global_step": 89933, "epoch": 1010, "lr": 7.544184049361424e-05}
{"train_loss": 0.1515956073999405, "global_step": 89934, "epoch": 1010, "lr": 7.54413414181138e-05}
{"train_loss": 0.24297046661376953, "global_step": 89935, "epoch": 1010, "lr": 7.544084233919309e-05}
{"train_loss": 0.1574682593345642, "global_step": 89936, "epoch": 1010, "lr": 7.544034325685216e-05}
{"train_loss": 0.20194661617279053, "global_step": 89937, "epoch": 1010, "lr": 7.543984417109108e-05}
{"train_loss": 0.20114488899707794, "global_step": 89938, "epoch": 1010, "lr": 7.543934508190994e-05}
{"train_loss": 0.16497094929218292, "global_step": 89939, "epoch": 1010, "lr": 7.543884598930878e-05}
{"train_loss": 0.1667330265045166, "global_step": 89940, "epoch": 1010, "lr": 7.543834689328769e-05}
{"train_loss": 0.14935621619224548, "global_step": 89941, "epoch": 1010, "lr": 7.54378477938467e-05}
{"train_loss": 0.23109780251979828, "global_step": 89942, "epoch": 1010, "lr": 7.54373486909859e-05}
{"train_loss": 0.16298286616802216, "global_step": 89943, "epoch": 1010, "lr": 7.543684958470538e-05}
{"train_loss": 0.18461191654205322, "global_step": 89944, "epoch": 1010, "lr": 7.543635047500517e-05}
{"train_loss": 0.23048612475395203, "global_step": 89945, "epoch": 1010, "lr": 7.543585136188535e-05}
{"train_loss": 0.11197046935558319, "global_step": 89946, "epoch": 1010, "lr": 7.543535224534601e-05}
{"train_loss": 0.1820482462644577, "global_step": 89947, "epoch": 1010, "lr": 7.543485312538718e-05}
{"train_loss": 0.18493086099624634, "global_step": 89948, "epoch": 1010, "lr": 7.543435400200895e-05}
{"train_loss": 0.24046935141086578, "global_step": 89949, "epoch": 1010, "lr": 7.543385487521139e-05}
{"train_loss": 0.2929457426071167, "global_step": 89950, "epoch": 1010, "lr": 7.543335574499455e-05}
{"train_loss": 0.2065272182226181, "global_step": 89951, "epoch": 1010, "lr": 7.54328566113585e-05}
{"train_loss": 0.2217356115579605, "global_step": 89952, "epoch": 1010, "lr": 7.543235747430331e-05}
{"train_loss": 0.14303059875965118, "global_step": 89953, "epoch": 1010, "lr": 7.543185833382906e-05}
{"train_loss": 0.16195029020309448, "global_step": 89954, "epoch": 1010, "lr": 7.54313591899358e-05}
{"train_loss": 0.1386270672082901, "global_step": 89955, "epoch": 1010, "lr": 7.543086004262361e-05}
{"train_loss": 0.184453547000885, "global_step": 89956, "epoch": 1010, "lr": 7.543036089189254e-05}
{"train_loss": 0.199809730052948, "global_step": 89957, "epoch": 1010, "lr": 7.542986173774269e-05}
{"train_loss": 0.28940239548683167, "global_step": 89958, "epoch": 1010, "lr": 7.542936258017408e-05}
{"train_loss": 0.26179245114326477, "global_step": 89959, "epoch": 1010, "lr": 7.542886341918682e-05}
{"train_loss": 0.24373102188110352, "global_step": 89960, "epoch": 1010, "lr": 7.542836425478094e-05}
{"train_loss": 0.10858955979347229, "global_step": 89961, "epoch": 1010, "lr": 7.542786508695655e-05}
{"train_loss": 0.13210195302963257, "global_step": 89962, "epoch": 1010, "lr": 7.542736591571367e-05}
{"train_loss": 0.18721534311771393, "global_step": 89963, "epoch": 1010, "lr": 7.542686674105242e-05}
{"train_loss": 0.1468387395143509, "global_step": 89964, "epoch": 1010, "lr": 7.542636756297281e-05}
{"train_loss": 0.16902610659599304, "global_step": 89965, "epoch": 1010, "lr": 7.542586838147494e-05}
{"train_loss": 0.23039427399635315, "global_step": 89966, "epoch": 1010, "lr": 7.542536919655888e-05}
{"train_loss": 0.17102377116680145, "global_step": 89967, "epoch": 1010, "lr": 7.542487000822468e-05}
{"train_loss": 0.21701157093048096, "global_step": 89968, "epoch": 1010, "lr": 7.542437081647242e-05}
{"train_loss": 0.10588715970516205, "global_step": 89969, "epoch": 1010, "lr": 7.542387162130218e-05}
{"train_loss": 0.16147273778915405, "global_step": 89970, "epoch": 1010, "lr": 7.542337242271398e-05}
{"train_loss": 0.223281592130661, "global_step": 89971, "epoch": 1010, "lr": 7.542287322070794e-05}
{"train_loss": 0.22064541280269623, "global_step": 89972, "epoch": 1010, "lr": 7.542237401528409e-05}
{"train_loss": 0.21685105562210083, "global_step": 89973, "epoch": 1010, "lr": 7.542187480644252e-05}
{"train_loss": 0.22048944234848022, "global_step": 89974, "epoch": 1010, "lr": 7.54213755941833e-05}
{"train_loss": 0.21932195127010345, "global_step": 89975, "epoch": 1010, "lr": 7.542087637850646e-05}
{"train_loss": 0.21646396815776825, "global_step": 89976, "epoch": 1010, "lr": 7.54203771594121e-05}
{"train_loss": 0.14292481541633606, "global_step": 89977, "epoch": 1010, "lr": 7.54198779369003e-05}
{"train_loss": 0.18360297616278187, "global_step": 89978, "epoch": 1010, "lr": 7.541937871097109e-05, "val_loss": 4.361655235290527, "train_action_mse_error": 31.208662033081055}
{"train_loss": 0.17590317130088806, "global_step": 89979, "epoch": 1011, "lr": 7.541887948162456e-05}
{"train_loss": 0.1346825659275055, "global_step": 89980, "epoch": 1011, "lr": 7.541838024886076e-05}
{"train_loss": 0.12659852206707, "global_step": 89981, "epoch": 1011, "lr": 7.541788101267979e-05}
{"train_loss": 0.16068598628044128, "global_step": 89982, "epoch": 1011, "lr": 7.541738177308168e-05}
{"train_loss": 0.12756562232971191, "global_step": 89983, "epoch": 1011, "lr": 7.541688253006651e-05}
{"train_loss": 0.13456067442893982, "global_step": 89984, "epoch": 1011, "lr": 7.541638328363436e-05}
{"train_loss": 0.153049498796463, "global_step": 89985, "epoch": 1011, "lr": 7.541588403378528e-05}
{"train_loss": 0.1592029631137848, "global_step": 89986, "epoch": 1011, "lr": 7.541538478051935e-05}
{"train_loss": 0.14364124834537506, "global_step": 89987, "epoch": 1011, "lr": 7.541488552383664e-05}
{"train_loss": 0.12580537796020508, "global_step": 89988, "epoch": 1011, "lr": 7.541438626373719e-05}
{"train_loss": 0.17768248915672302, "global_step": 89989, "epoch": 1011, "lr": 7.54138870002211e-05}
{"train_loss": 0.11639833450317383, "global_step": 89990, "epoch": 1011, "lr": 7.54133877332884e-05}
{"train_loss": 0.1563320904970169, "global_step": 89991, "epoch": 1011, "lr": 7.541288846293921e-05}
{"train_loss": 0.21243487298488617, "global_step": 89992, "epoch": 1011, "lr": 7.541238918917355e-05}
{"train_loss": 0.24100391566753387, "global_step": 89993, "epoch": 1011, "lr": 7.541188991199151e-05}
{"train_loss": 0.2262772172689438, "global_step": 89994, "epoch": 1011, "lr": 7.541139063139315e-05}
{"train_loss": 0.20463523268699646, "global_step": 89995, "epoch": 1011, "lr": 7.541089134737853e-05}
{"train_loss": 0.18197327852249146, "global_step": 89996, "epoch": 1011, "lr": 7.541039205994774e-05}
{"train_loss": 0.17238907516002655, "global_step": 89997, "epoch": 1011, "lr": 7.540989276910082e-05}
{"train_loss": 0.2489178329706192, "global_step": 89998, "epoch": 1011, "lr": 7.540939347483785e-05}
{"train_loss": 0.2073759138584137, "global_step": 89999, "epoch": 1011, "lr": 7.540889417715889e-05}
{"train_loss": 0.09281525015830994, "global_step": 90000, "epoch": 1011, "lr": 7.540839487606404e-05}
{"train_loss": 0.20088960230350494, "global_step": 90001, "epoch": 1011, "lr": 7.540789557155332e-05}
{"train_loss": 0.13483421504497528, "global_step": 90002, "epoch": 1011, "lr": 7.540739626362682e-05}
{"train_loss": 0.1810697764158249, "global_step": 90003, "epoch": 1011, "lr": 7.54068969522846e-05}
{"train_loss": 0.22250553965568542, "global_step": 90004, "epoch": 1011, "lr": 7.540639763752675e-05}
{"train_loss": 0.16323140263557434, "global_step": 90005, "epoch": 1011, "lr": 7.540589831935331e-05}
{"train_loss": 0.14990568161010742, "global_step": 90006, "epoch": 1011, "lr": 7.540539899776435e-05}
{"train_loss": 0.16332828998565674, "global_step": 90007, "epoch": 1011, "lr": 7.540489967275995e-05}
{"train_loss": 0.1917533576488495, "global_step": 90008, "epoch": 1011, "lr": 7.540440034434017e-05}
{"train_loss": 0.15616437792778015, "global_step": 90009, "epoch": 1011, "lr": 7.540390101250508e-05}
{"train_loss": 0.10024652630090714, "global_step": 90010, "epoch": 1011, "lr": 7.540340167725475e-05}
{"train_loss": 0.18114811182022095, "global_step": 90011, "epoch": 1011, "lr": 7.540290233858925e-05}
{"train_loss": 0.13461269438266754, "global_step": 90012, "epoch": 1011, "lr": 7.540240299650861e-05}
{"train_loss": 0.2248162478208542, "global_step": 90013, "epoch": 1011, "lr": 7.540190365101294e-05}
{"train_loss": 0.18369106948375702, "global_step": 90014, "epoch": 1011, "lr": 7.54014043021023e-05}
{"train_loss": 0.20501570403575897, "global_step": 90015, "epoch": 1011, "lr": 7.540090494977674e-05}
{"train_loss": 0.20970110595226288, "global_step": 90016, "epoch": 1011, "lr": 7.540040559403635e-05}
{"train_loss": 0.157136932015419, "global_step": 90017, "epoch": 1011, "lr": 7.539990623488118e-05}
{"train_loss": 0.1236947700381279, "global_step": 90018, "epoch": 1011, "lr": 7.53994068723113e-05}
{"train_loss": 0.13354924321174622, "global_step": 90019, "epoch": 1011, "lr": 7.539890750632679e-05}
{"train_loss": 0.19852431118488312, "global_step": 90020, "epoch": 1011, "lr": 7.53984081369277e-05}
{"train_loss": 0.1399610936641693, "global_step": 90021, "epoch": 1011, "lr": 7.53979087641141e-05}
{"train_loss": 0.196461483836174, "global_step": 90022, "epoch": 1011, "lr": 7.539740938788607e-05}
{"train_loss": 0.21018768846988678, "global_step": 90023, "epoch": 1011, "lr": 7.539691000824365e-05}
{"train_loss": 0.1864694058895111, "global_step": 90024, "epoch": 1011, "lr": 7.539641062518694e-05}
{"train_loss": 0.12558673322200775, "global_step": 90025, "epoch": 1011, "lr": 7.539591123871599e-05}
{"train_loss": 0.14496371150016785, "global_step": 90026, "epoch": 1011, "lr": 7.539541184883086e-05}
{"train_loss": 0.16605639457702637, "global_step": 90027, "epoch": 1011, "lr": 7.539491245553161e-05}
{"train_loss": 0.1774894744157791, "global_step": 90028, "epoch": 1011, "lr": 7.539441305881835e-05}
{"train_loss": 0.1950773298740387, "global_step": 90029, "epoch": 1011, "lr": 7.539391365869113e-05}
{"train_loss": 0.1534084975719452, "global_step": 90030, "epoch": 1011, "lr": 7.539341425515e-05}
{"train_loss": 0.11232562363147736, "global_step": 90031, "epoch": 1011, "lr": 7.539291484819503e-05}
{"train_loss": 0.20720456540584564, "global_step": 90032, "epoch": 1011, "lr": 7.53924154378263e-05}
{"train_loss": 0.13801999390125275, "global_step": 90033, "epoch": 1011, "lr": 7.539191602404385e-05}
{"train_loss": 0.17312532663345337, "global_step": 90034, "epoch": 1011, "lr": 7.539141660684778e-05}
{"train_loss": 0.250169575214386, "global_step": 90035, "epoch": 1011, "lr": 7.539091718623814e-05}
{"train_loss": 0.21561960875988007, "global_step": 90036, "epoch": 1011, "lr": 7.539041776221501e-05}
{"train_loss": 0.15675602853298187, "global_step": 90037, "epoch": 1011, "lr": 7.538991833477843e-05}
{"train_loss": 0.15569408237934113, "global_step": 90038, "epoch": 1011, "lr": 7.53894189039285e-05}
{"train_loss": 0.16051967442035675, "global_step": 90039, "epoch": 1011, "lr": 7.538891946966526e-05}
{"train_loss": 0.1735895723104477, "global_step": 90040, "epoch": 1011, "lr": 7.538842003198881e-05}
{"train_loss": 0.10900252312421799, "global_step": 90041, "epoch": 1011, "lr": 7.538792059089918e-05}
{"train_loss": 0.18672609329223633, "global_step": 90042, "epoch": 1011, "lr": 7.538742114639646e-05}
{"train_loss": 0.17659202218055725, "global_step": 90043, "epoch": 1011, "lr": 7.53869216984807e-05}
{"train_loss": 0.08618146181106567, "global_step": 90044, "epoch": 1011, "lr": 7.538642224715198e-05}
{"train_loss": 0.1828935444355011, "global_step": 90045, "epoch": 1011, "lr": 7.538592279241035e-05}
{"train_loss": 0.23136238753795624, "global_step": 90046, "epoch": 1011, "lr": 7.538542333425593e-05}
{"train_loss": 0.11383914947509766, "global_step": 90047, "epoch": 1011, "lr": 7.538492387268871e-05}
{"train_loss": 0.13721446692943573, "global_step": 90048, "epoch": 1011, "lr": 7.538442440770883e-05}
{"train_loss": 0.17287206649780273, "global_step": 90049, "epoch": 1011, "lr": 7.538392493931632e-05}
{"train_loss": 0.19979414343833923, "global_step": 90050, "epoch": 1011, "lr": 7.538342546751123e-05}
{"train_loss": 0.16975758969783783, "global_step": 90051, "epoch": 1011, "lr": 7.538292599229366e-05}
{"train_loss": 0.20851460099220276, "global_step": 90052, "epoch": 1011, "lr": 7.538242651366368e-05}
{"train_loss": 0.14961878955364227, "global_step": 90053, "epoch": 1011, "lr": 7.538192703162132e-05}
{"train_loss": 0.20601333677768707, "global_step": 90054, "epoch": 1011, "lr": 7.538142754616667e-05}
{"train_loss": 0.24426133930683136, "global_step": 90055, "epoch": 1011, "lr": 7.53809280572998e-05}
{"train_loss": 0.19315628707408905, "global_step": 90056, "epoch": 1011, "lr": 7.538042856502078e-05}
{"train_loss": 0.1402512788772583, "global_step": 90057, "epoch": 1011, "lr": 7.537992906932967e-05}
{"train_loss": 0.19783444702625275, "global_step": 90058, "epoch": 1011, "lr": 7.537942957022653e-05}
{"train_loss": 0.14979282021522522, "global_step": 90059, "epoch": 1011, "lr": 7.537893006771144e-05}
{"train_loss": 0.10188739746809006, "global_step": 90060, "epoch": 1011, "lr": 7.537843056178446e-05}
{"train_loss": 0.1881093978881836, "global_step": 90061, "epoch": 1011, "lr": 7.537793105244566e-05}
{"train_loss": 0.1568269282579422, "global_step": 90062, "epoch": 1011, "lr": 7.537743153969511e-05}
{"train_loss": 0.22999438643455505, "global_step": 90063, "epoch": 1011, "lr": 7.537693202353286e-05}
{"train_loss": 0.19781193137168884, "global_step": 90064, "epoch": 1011, "lr": 7.5376432503959e-05}
{"train_loss": 0.2175837904214859, "global_step": 90065, "epoch": 1011, "lr": 7.537593298097358e-05}
{"train_loss": 0.22006693482398987, "global_step": 90066, "epoch": 1011, "lr": 7.537543345457669e-05}
{"train_loss": 0.17194737326563075, "global_step": 90067, "epoch": 1011, "lr": 7.537493392476837e-05, "val_loss": 4.445292949676514}
{"train_loss": 0.178719162940979, "global_step": 90068, "epoch": 1012, "lr": 7.537443439154872e-05}
{"train_loss": 0.19963806867599487, "global_step": 90069, "epoch": 1012, "lr": 7.537393485491777e-05}
{"train_loss": 0.1840035766363144, "global_step": 90070, "epoch": 1012, "lr": 7.537343531487562e-05}
{"train_loss": 0.1493183672428131, "global_step": 90071, "epoch": 1012, "lr": 7.53729357714223e-05}
{"train_loss": 0.15306219458580017, "global_step": 90072, "epoch": 1012, "lr": 7.537243622455791e-05}
{"train_loss": 0.14289818704128265, "global_step": 90073, "epoch": 1012, "lr": 7.53719366742825e-05}
{"train_loss": 0.11479362845420837, "global_step": 90074, "epoch": 1012, "lr": 7.537143712059613e-05}
{"train_loss": 0.14320094883441925, "global_step": 90075, "epoch": 1012, "lr": 7.537093756349892e-05}
{"train_loss": 0.268246591091156, "global_step": 90076, "epoch": 1012, "lr": 7.537043800299086e-05}
{"train_loss": 0.19532373547554016, "global_step": 90077, "epoch": 1012, "lr": 7.536993843907205e-05}
{"train_loss": 0.16711197793483734, "global_step": 90078, "epoch": 1012, "lr": 7.536943887174257e-05}
{"train_loss": 0.17482595145702362, "global_step": 90079, "epoch": 1012, "lr": 7.536893930100248e-05}
{"train_loss": 0.16392163932323456, "global_step": 90080, "epoch": 1012, "lr": 7.536843972685186e-05}
{"train_loss": 0.21113240718841553, "global_step": 90081, "epoch": 1012, "lr": 7.536794014929073e-05}
{"train_loss": 0.1713055819272995, "global_step": 90082, "epoch": 1012, "lr": 7.536744056831921e-05}
{"train_loss": 0.13233567774295807, "global_step": 90083, "epoch": 1012, "lr": 7.536694098393735e-05}
{"train_loss": 0.1567969024181366, "global_step": 90084, "epoch": 1012, "lr": 7.536644139614522e-05}
{"train_loss": 0.12646785378456116, "global_step": 90085, "epoch": 1012, "lr": 7.536594180494287e-05}
{"train_loss": 0.1533001959323883, "global_step": 90086, "epoch": 1012, "lr": 7.536544221033037e-05}
{"train_loss": 0.12360473722219467, "global_step": 90087, "epoch": 1012, "lr": 7.536494261230782e-05}
{"train_loss": 0.18261529505252838, "global_step": 90088, "epoch": 1012, "lr": 7.536444301087524e-05}
{"train_loss": 0.114716537296772, "global_step": 90089, "epoch": 1012, "lr": 7.536394340603273e-05}
{"train_loss": 0.17192457616329193, "global_step": 90090, "epoch": 1012, "lr": 7.536344379778035e-05}
{"train_loss": 0.17855648696422577, "global_step": 90091, "epoch": 1012, "lr": 7.536294418611815e-05}
{"train_loss": 0.248691588640213, "global_step": 90092, "epoch": 1012, "lr": 7.536244457104622e-05}
{"train_loss": 0.22334074974060059, "global_step": 90093, "epoch": 1012, "lr": 7.536194495256463e-05}
{"train_loss": 0.16705462336540222, "global_step": 90094, "epoch": 1012, "lr": 7.536144533067342e-05}
{"train_loss": 0.09888327866792679, "global_step": 90095, "epoch": 1012, "lr": 7.536094570537268e-05}
{"train_loss": 0.08480200171470642, "global_step": 90096, "epoch": 1012, "lr": 7.536044607666246e-05}
{"train_loss": 0.12559622526168823, "global_step": 90097, "epoch": 1012, "lr": 7.535994644454285e-05}
{"train_loss": 0.12233595550060272, "global_step": 90098, "epoch": 1012, "lr": 7.53594468090139e-05}
{"train_loss": 0.17262516915798187, "global_step": 90099, "epoch": 1012, "lr": 7.535894717007568e-05}
{"train_loss": 0.22833409905433655, "global_step": 90100, "epoch": 1012, "lr": 7.535844752772825e-05}
{"train_loss": 0.14745977520942688, "global_step": 90101, "epoch": 1012, "lr": 7.535794788197167e-05}
{"train_loss": 0.13513551652431488, "global_step": 90102, "epoch": 1012, "lr": 7.535744823280606e-05}
{"train_loss": 0.19214344024658203, "global_step": 90103, "epoch": 1012, "lr": 7.535694858023144e-05}
{"train_loss": 0.1885271668434143, "global_step": 90104, "epoch": 1012, "lr": 7.535644892424789e-05}
{"train_loss": 0.15833976864814758, "global_step": 90105, "epoch": 1012, "lr": 7.535594926485547e-05}
{"train_loss": 0.20982882380485535, "global_step": 90106, "epoch": 1012, "lr": 7.535544960205425e-05}
{"train_loss": 0.19502341747283936, "global_step": 90107, "epoch": 1012, "lr": 7.535494993584431e-05}
{"train_loss": 0.20149169862270355, "global_step": 90108, "epoch": 1012, "lr": 7.53544502662257e-05}
{"train_loss": 0.12106085568666458, "global_step": 90109, "epoch": 1012, "lr": 7.535395059319847e-05}
{"train_loss": 0.16530945897102356, "global_step": 90110, "epoch": 1012, "lr": 7.535345091676273e-05}
{"train_loss": 0.16386380791664124, "global_step": 90111, "epoch": 1012, "lr": 7.535295123691853e-05}
{"train_loss": 0.1441035121679306, "global_step": 90112, "epoch": 1012, "lr": 7.535245155366594e-05}
{"train_loss": 0.15964031219482422, "global_step": 90113, "epoch": 1012, "lr": 7.535195186700501e-05}
{"train_loss": 0.16187110543251038, "global_step": 90114, "epoch": 1012, "lr": 7.535145217693582e-05}
{"train_loss": 0.18376781046390533, "global_step": 90115, "epoch": 1012, "lr": 7.535095248345844e-05}
{"train_loss": 0.16056683659553528, "global_step": 90116, "epoch": 1012, "lr": 7.535045278657294e-05}
{"train_loss": 0.15010413527488708, "global_step": 90117, "epoch": 1012, "lr": 7.534995308627937e-05}
{"train_loss": 0.24714304506778717, "global_step": 90118, "epoch": 1012, "lr": 7.534945338257779e-05}
{"train_loss": 0.2034546583890915, "global_step": 90119, "epoch": 1012, "lr": 7.53489536754683e-05}
{"train_loss": 0.19410991668701172, "global_step": 90120, "epoch": 1012, "lr": 7.534845396495096e-05}
{"train_loss": 0.1380101591348648, "global_step": 90121, "epoch": 1012, "lr": 7.534795425102584e-05}
{"train_loss": 0.16424812376499176, "global_step": 90122, "epoch": 1012, "lr": 7.534745453369297e-05}
{"train_loss": 0.09751898795366287, "global_step": 90123, "epoch": 1012, "lr": 7.534695481295246e-05}
{"train_loss": 0.16837099194526672, "global_step": 90124, "epoch": 1012, "lr": 7.534645508880436e-05}
{"train_loss": 0.1928224116563797, "global_step": 90125, "epoch": 1012, "lr": 7.534595536124873e-05}
{"train_loss": 0.14648303389549255, "global_step": 90126, "epoch": 1012, "lr": 7.534545563028566e-05}
{"train_loss": 0.12346642464399338, "global_step": 90127, "epoch": 1012, "lr": 7.534495589591519e-05}
{"train_loss": 0.24405838549137115, "global_step": 90128, "epoch": 1012, "lr": 7.534445615813742e-05}
{"train_loss": 0.172173410654068, "global_step": 90129, "epoch": 1012, "lr": 7.534395641695238e-05}
{"train_loss": 0.20461341738700867, "global_step": 90130, "epoch": 1012, "lr": 7.534345667236015e-05}
{"train_loss": 0.08233930915594101, "global_step": 90131, "epoch": 1012, "lr": 7.53429569243608e-05}
{"train_loss": 0.16262875497341156, "global_step": 90132, "epoch": 1012, "lr": 7.534245717295441e-05}
{"train_loss": 0.18979057669639587, "global_step": 90133, "epoch": 1012, "lr": 7.534195741814103e-05}
{"train_loss": 0.17331115901470184, "global_step": 90134, "epoch": 1012, "lr": 7.534145765992074e-05}
{"train_loss": 0.1575377732515335, "global_step": 90135, "epoch": 1012, "lr": 7.534095789829357e-05}
{"train_loss": 0.17562521994113922, "global_step": 90136, "epoch": 1012, "lr": 7.534045813325963e-05}
{"train_loss": 0.19019100069999695, "global_step": 90137, "epoch": 1012, "lr": 7.533995836481899e-05}
{"train_loss": 0.2057860940694809, "global_step": 90138, "epoch": 1012, "lr": 7.53394585929717e-05}
{"train_loss": 0.24631446599960327, "global_step": 90139, "epoch": 1012, "lr": 7.533895881771782e-05}
{"train_loss": 0.25566765666007996, "global_step": 90140, "epoch": 1012, "lr": 7.533845903905743e-05}
{"train_loss": 0.16319887340068817, "global_step": 90141, "epoch": 1012, "lr": 7.533795925699059e-05}
{"train_loss": 0.16637645661830902, "global_step": 90142, "epoch": 1012, "lr": 7.533745947151737e-05}
{"train_loss": 0.1365731954574585, "global_step": 90143, "epoch": 1012, "lr": 7.533695968263783e-05}
{"train_loss": 0.18222932517528534, "global_step": 90144, "epoch": 1012, "lr": 7.533645989035205e-05}
{"train_loss": 0.21731218695640564, "global_step": 90145, "epoch": 1012, "lr": 7.53359600946601e-05}
{"train_loss": 0.20525142550468445, "global_step": 90146, "epoch": 1012, "lr": 7.533546029556204e-05}
{"train_loss": 0.2559667229652405, "global_step": 90147, "epoch": 1012, "lr": 7.533496049305794e-05}
{"train_loss": 0.14579160511493683, "global_step": 90148, "epoch": 1012, "lr": 7.533446068714785e-05}
{"train_loss": 0.18430215120315552, "global_step": 90149, "epoch": 1012, "lr": 7.533396087783187e-05}
{"train_loss": 0.10936877876520157, "global_step": 90150, "epoch": 1012, "lr": 7.533346106511002e-05}
{"train_loss": 0.12396324425935745, "global_step": 90151, "epoch": 1012, "lr": 7.533296124898242e-05}
{"train_loss": 0.1863919049501419, "global_step": 90152, "epoch": 1012, "lr": 7.533246142944908e-05}
{"train_loss": 0.21047593653202057, "global_step": 90153, "epoch": 1012, "lr": 7.533196160651013e-05}
{"train_loss": 0.2645888924598694, "global_step": 90154, "epoch": 1012, "lr": 7.533146178016558e-05}
{"train_loss": 0.18521316349506378, "global_step": 90155, "epoch": 1012, "lr": 7.533096195041554e-05}
{"train_loss": 0.17229570573970174, "global_step": 90156, "epoch": 1012, "lr": 7.533046211726005e-05, "val_loss": 4.394625186920166}
{"train_loss": 0.1807498037815094, "global_step": 90157, "epoch": 1013, "lr": 7.532996228069921e-05}
{"train_loss": 0.1660289615392685, "global_step": 90158, "epoch": 1013, "lr": 7.532946244073304e-05}
{"train_loss": 0.20728160440921783, "global_step": 90159, "epoch": 1013, "lr": 7.532896259736166e-05}
{"train_loss": 0.1351465880870819, "global_step": 90160, "epoch": 1013, "lr": 7.532846275058509e-05}
{"train_loss": 0.18106162548065186, "global_step": 90161, "epoch": 1013, "lr": 7.532796290040342e-05}
{"train_loss": 0.1371714323759079, "global_step": 90162, "epoch": 1013, "lr": 7.532746304681673e-05}
{"train_loss": 0.1467895805835724, "global_step": 90163, "epoch": 1013, "lr": 7.532696318982505e-05}
{"train_loss": 0.17561852931976318, "global_step": 90164, "epoch": 1013, "lr": 7.532646332942847e-05}
{"train_loss": 0.20542879402637482, "global_step": 90165, "epoch": 1013, "lr": 7.532596346562708e-05}
{"train_loss": 0.24969208240509033, "global_step": 90166, "epoch": 1013, "lr": 7.53254635984209e-05}
{"train_loss": 0.19559764862060547, "global_step": 90167, "epoch": 1013, "lr": 7.532496372781002e-05}
{"train_loss": 0.1855309009552002, "global_step": 90168, "epoch": 1013, "lr": 7.532446385379453e-05}
{"train_loss": 0.1471167802810669, "global_step": 90169, "epoch": 1013, "lr": 7.532396397637444e-05}
{"train_loss": 0.1706644743680954, "global_step": 90170, "epoch": 1013, "lr": 7.532346409554987e-05}
{"train_loss": 0.16403773427009583, "global_step": 90171, "epoch": 1013, "lr": 7.532296421132086e-05}
{"train_loss": 0.1939111202955246, "global_step": 90172, "epoch": 1013, "lr": 7.532246432368749e-05}
{"train_loss": 0.19157880544662476, "global_step": 90173, "epoch": 1013, "lr": 7.532196443264982e-05}
{"train_loss": 0.202558234333992, "global_step": 90174, "epoch": 1013, "lr": 7.532146453820794e-05}
{"train_loss": 0.12006668001413345, "global_step": 90175, "epoch": 1013, "lr": 7.532096464036188e-05}
{"train_loss": 0.14282284677028656, "global_step": 90176, "epoch": 1013, "lr": 7.532046473911174e-05}
{"train_loss": 0.19405260682106018, "global_step": 90177, "epoch": 1013, "lr": 7.531996483445756e-05}
{"train_loss": 0.14671054482460022, "global_step": 90178, "epoch": 1013, "lr": 7.531946492639941e-05}
{"train_loss": 0.19777923822402954, "global_step": 90179, "epoch": 1013, "lr": 7.531896501493738e-05}
{"train_loss": 0.19782200455665588, "global_step": 90180, "epoch": 1013, "lr": 7.531846510007152e-05}
{"train_loss": 0.2712382674217224, "global_step": 90181, "epoch": 1013, "lr": 7.531796518180191e-05}
{"train_loss": 0.19329574704170227, "global_step": 90182, "epoch": 1013, "lr": 7.53174652601286e-05}
{"train_loss": 0.21292005479335785, "global_step": 90183, "epoch": 1013, "lr": 7.531696533505166e-05}
{"train_loss": 0.2151309847831726, "global_step": 90184, "epoch": 1013, "lr": 7.531646540657117e-05}
{"train_loss": 0.1645851880311966, "global_step": 90185, "epoch": 1013, "lr": 7.53159654746872e-05}
{"train_loss": 0.10880599915981293, "global_step": 90186, "epoch": 1013, "lr": 7.531546553939979e-05}
{"train_loss": 0.13300272822380066, "global_step": 90187, "epoch": 1013, "lr": 7.531496560070903e-05}
{"train_loss": 0.16913729906082153, "global_step": 90188, "epoch": 1013, "lr": 7.531446565861498e-05}
{"train_loss": 0.2848948538303375, "global_step": 90189, "epoch": 1013, "lr": 7.531396571311772e-05}
{"train_loss": 0.1416821926832199, "global_step": 90190, "epoch": 1013, "lr": 7.531346576421729e-05}
{"train_loss": 0.1506931185722351, "global_step": 90191, "epoch": 1013, "lr": 7.531296581191378e-05}
{"train_loss": 0.19795773923397064, "global_step": 90192, "epoch": 1013, "lr": 7.531246585620723e-05}
{"train_loss": 0.2327536940574646, "global_step": 90193, "epoch": 1013, "lr": 7.531196589709776e-05}
{"train_loss": 0.1678105741739273, "global_step": 90194, "epoch": 1013, "lr": 7.531146593458537e-05}
{"train_loss": 0.1510915607213974, "global_step": 90195, "epoch": 1013, "lr": 7.53109659686702e-05}
{"train_loss": 0.23479700088500977, "global_step": 90196, "epoch": 1013, "lr": 7.531046599935226e-05}
{"train_loss": 0.11607059836387634, "global_step": 90197, "epoch": 1013, "lr": 7.530996602663163e-05}
{"train_loss": 0.16850633919239044, "global_step": 90198, "epoch": 1013, "lr": 7.53094660505084e-05}
{"train_loss": 0.23363058269023895, "global_step": 90199, "epoch": 1013, "lr": 7.530896607098262e-05}
{"train_loss": 0.15504759550094604, "global_step": 90200, "epoch": 1013, "lr": 7.530846608805435e-05}
{"train_loss": 0.20368200540542603, "global_step": 90201, "epoch": 1013, "lr": 7.530796610172366e-05}
{"train_loss": 0.19792363047599792, "global_step": 90202, "epoch": 1013, "lr": 7.530746611199063e-05}
{"train_loss": 0.19710089266300201, "global_step": 90203, "epoch": 1013, "lr": 7.530696611885531e-05}
{"train_loss": 0.1213274821639061, "global_step": 90204, "epoch": 1013, "lr": 7.53064661223178e-05}
{"train_loss": 0.20015400648117065, "global_step": 90205, "epoch": 1013, "lr": 7.530596612237813e-05}
{"train_loss": 0.15950582921504974, "global_step": 90206, "epoch": 1013, "lr": 7.530546611903638e-05}
{"train_loss": 0.2054789960384369, "global_step": 90207, "epoch": 1013, "lr": 7.530496611229262e-05}
{"train_loss": 0.1945403665304184, "global_step": 90208, "epoch": 1013, "lr": 7.530446610214689e-05}
{"train_loss": 0.09971288591623306, "global_step": 90209, "epoch": 1013, "lr": 7.530396608859932e-05}
{"train_loss": 0.2091865986585617, "global_step": 90210, "epoch": 1013, "lr": 7.530346607164994e-05}
{"train_loss": 0.1807575672864914, "global_step": 90211, "epoch": 1013, "lr": 7.530296605129879e-05}
{"train_loss": 0.14380130171775818, "global_step": 90212, "epoch": 1013, "lr": 7.530246602754599e-05}
{"train_loss": 0.1557326465845108, "global_step": 90213, "epoch": 1013, "lr": 7.530196600039157e-05}
{"train_loss": 0.20707444846630096, "global_step": 90214, "epoch": 1013, "lr": 7.530146596983561e-05}
{"train_loss": 0.25151848793029785, "global_step": 90215, "epoch": 1013, "lr": 7.530096593587816e-05}
{"train_loss": 0.242926225066185, "global_step": 90216, "epoch": 1013, "lr": 7.530046589851934e-05}
{"train_loss": 0.15341782569885254, "global_step": 90217, "epoch": 1013, "lr": 7.529996585775915e-05}
{"train_loss": 0.11749742180109024, "global_step": 90218, "epoch": 1013, "lr": 7.529946581359771e-05}
{"train_loss": 0.08642072230577469, "global_step": 90219, "epoch": 1013, "lr": 7.529896576603505e-05}
{"train_loss": 0.1569591760635376, "global_step": 90220, "epoch": 1013, "lr": 7.529846571507125e-05}
{"train_loss": 0.149824321269989, "global_step": 90221, "epoch": 1013, "lr": 7.52979656607064e-05}
{"train_loss": 0.1885780245065689, "global_step": 90222, "epoch": 1013, "lr": 7.529746560294051e-05}
{"train_loss": 0.19487695395946503, "global_step": 90223, "epoch": 1013, "lr": 7.529696554177372e-05}
{"train_loss": 0.21919041872024536, "global_step": 90224, "epoch": 1013, "lr": 7.529646547720606e-05}
{"train_loss": 0.27673807740211487, "global_step": 90225, "epoch": 1013, "lr": 7.529596540923758e-05}
{"train_loss": 0.23578095436096191, "global_step": 90226, "epoch": 1013, "lr": 7.529546533786838e-05}
{"train_loss": 0.15202246606349945, "global_step": 90227, "epoch": 1013, "lr": 7.52949652630985e-05}
{"train_loss": 0.15307536721229553, "global_step": 90228, "epoch": 1013, "lr": 7.529446518492801e-05}
{"train_loss": 0.1994204968214035, "global_step": 90229, "epoch": 1013, "lr": 7.529396510335701e-05}
{"train_loss": 0.1954350620508194, "global_step": 90230, "epoch": 1013, "lr": 7.529346501838555e-05}
{"train_loss": 0.1359720379114151, "global_step": 90231, "epoch": 1013, "lr": 7.529296493001368e-05}
{"train_loss": 0.1527729481458664, "global_step": 90232, "epoch": 1013, "lr": 7.529246483824146e-05}
{"train_loss": 0.23082245886325836, "global_step": 90233, "epoch": 1013, "lr": 7.529196474306901e-05}
{"train_loss": 0.20324192941188812, "global_step": 90234, "epoch": 1013, "lr": 7.529146464449635e-05}
{"train_loss": 0.1480785459280014, "global_step": 90235, "epoch": 1013, "lr": 7.529096454252356e-05}
{"train_loss": 0.217482790350914, "global_step": 90236, "epoch": 1013, "lr": 7.52904644371507e-05}
{"train_loss": 0.24884450435638428, "global_step": 90237, "epoch": 1013, "lr": 7.528996432837784e-05}
{"train_loss": 0.16793984174728394, "global_step": 90238, "epoch": 1013, "lr": 7.528946421620508e-05}
{"train_loss": 0.17881587147712708, "global_step": 90239, "epoch": 1013, "lr": 7.528896410063245e-05}
{"train_loss": 0.2217468023300171, "global_step": 90240, "epoch": 1013, "lr": 7.528846398166002e-05}
{"train_loss": 0.15315596759319305, "global_step": 90241, "epoch": 1013, "lr": 7.528796385928787e-05}
{"train_loss": 0.21882766485214233, "global_step": 90242, "epoch": 1013, "lr": 7.528746373351606e-05}
{"train_loss": 0.2055957019329071, "global_step": 90243, "epoch": 1013, "lr": 7.528696360434466e-05}
{"train_loss": 0.12494327127933502, "global_step": 90244, "epoch": 1013, "lr": 7.528646347177373e-05}
{"train_loss": 0.1824292903033535, "global_step": 90245, "epoch": 1013, "lr": 7.528596333580333e-05, "val_loss": 4.451895713806152}
{"train_loss": 0.14439623057842255, "global_step": 90246, "epoch": 1014, "lr": 7.528546319643357e-05}
{"train_loss": 0.1356181502342224, "global_step": 90247, "epoch": 1014, "lr": 7.528496305366446e-05}
{"train_loss": 0.10964540392160416, "global_step": 90248, "epoch": 1014, "lr": 7.528446290749613e-05}
{"train_loss": 0.14470335841178894, "global_step": 90249, "epoch": 1014, "lr": 7.528396275792859e-05}
{"train_loss": 0.24033123254776, "global_step": 90250, "epoch": 1014, "lr": 7.528346260496192e-05}
{"train_loss": 0.21615010499954224, "global_step": 90251, "epoch": 1014, "lr": 7.52829624485962e-05}
{"train_loss": 0.20604775846004486, "global_step": 90252, "epoch": 1014, "lr": 7.528246228883151e-05}
{"train_loss": 0.15383675694465637, "global_step": 90253, "epoch": 1014, "lr": 7.52819621256679e-05}
{"train_loss": 0.25040897727012634, "global_step": 90254, "epoch": 1014, "lr": 7.528146195910542e-05}
{"train_loss": 0.15912437438964844, "global_step": 90255, "epoch": 1014, "lr": 7.528096178914416e-05}
{"train_loss": 0.1468309760093689, "global_step": 90256, "epoch": 1014, "lr": 7.52804616157842e-05}
{"train_loss": 0.13439704477787018, "global_step": 90257, "epoch": 1014, "lr": 7.527996143902558e-05}
{"train_loss": 0.2513585090637207, "global_step": 90258, "epoch": 1014, "lr": 7.527946125886838e-05}
{"train_loss": 0.14468182623386383, "global_step": 90259, "epoch": 1014, "lr": 7.527896107531264e-05}
{"train_loss": 0.26348555088043213, "global_step": 90260, "epoch": 1014, "lr": 7.527846088835848e-05}
{"train_loss": 0.1520206332206726, "global_step": 90261, "epoch": 1014, "lr": 7.527796069800595e-05}
{"train_loss": 0.23949186503887177, "global_step": 90262, "epoch": 1014, "lr": 7.527746050425508e-05}
{"train_loss": 0.21031932532787323, "global_step": 90263, "epoch": 1014, "lr": 7.527696030710597e-05}
{"train_loss": 0.18013060092926025, "global_step": 90264, "epoch": 1014, "lr": 7.527646010655869e-05}
{"train_loss": 0.20905087888240814, "global_step": 90265, "epoch": 1014, "lr": 7.527595990261327e-05}
{"train_loss": 0.20704993605613708, "global_step": 90266, "epoch": 1014, "lr": 7.527545969526983e-05}
{"train_loss": 0.18522945046424866, "global_step": 90267, "epoch": 1014, "lr": 7.527495948452841e-05}
{"train_loss": 0.2423143982887268, "global_step": 90268, "epoch": 1014, "lr": 7.527445927038908e-05}
{"train_loss": 0.21331660449504852, "global_step": 90269, "epoch": 1014, "lr": 7.52739590528519e-05}
{"train_loss": 0.2454547882080078, "global_step": 90270, "epoch": 1014, "lr": 7.527345883191696e-05}
{"train_loss": 0.15247319638729095, "global_step": 90271, "epoch": 1014, "lr": 7.52729586075843e-05}
{"train_loss": 0.20194192230701447, "global_step": 90272, "epoch": 1014, "lr": 7.527245837985399e-05}
{"train_loss": 0.20272183418273926, "global_step": 90273, "epoch": 1014, "lr": 7.527195814872612e-05}
{"train_loss": 0.22705082595348358, "global_step": 90274, "epoch": 1014, "lr": 7.527145791420074e-05}
{"train_loss": 0.19747017323970795, "global_step": 90275, "epoch": 1014, "lr": 7.527095767627792e-05}
{"train_loss": 0.18048402667045593, "global_step": 90276, "epoch": 1014, "lr": 7.527045743495773e-05}
{"train_loss": 0.1837107539176941, "global_step": 90277, "epoch": 1014, "lr": 7.526995719024022e-05}
{"train_loss": 0.15111437439918518, "global_step": 90278, "epoch": 1014, "lr": 7.52694569421255e-05}
{"train_loss": 0.16663546860218048, "global_step": 90279, "epoch": 1014, "lr": 7.526895669061358e-05}
{"train_loss": 0.18512040376663208, "global_step": 90280, "epoch": 1014, "lr": 7.526845643570457e-05}
{"train_loss": 0.17012149095535278, "global_step": 90281, "epoch": 1014, "lr": 7.526795617739853e-05}
{"train_loss": 0.17742012441158295, "global_step": 90282, "epoch": 1014, "lr": 7.526745591569552e-05}
{"train_loss": 0.12892286479473114, "global_step": 90283, "epoch": 1014, "lr": 7.526695565059561e-05}
{"train_loss": 0.17069296538829803, "global_step": 90284, "epoch": 1014, "lr": 7.526645538209884e-05}
{"train_loss": 0.18643292784690857, "global_step": 90285, "epoch": 1014, "lr": 7.526595511020533e-05}
{"train_loss": 0.20092542469501495, "global_step": 90286, "epoch": 1014, "lr": 7.526545483491511e-05}
{"train_loss": 0.2233862280845642, "global_step": 90287, "epoch": 1014, "lr": 7.526495455622827e-05}
{"train_loss": 0.15590785443782806, "global_step": 90288, "epoch": 1014, "lr": 7.526445427414486e-05}
{"train_loss": 0.22637678682804108, "global_step": 90289, "epoch": 1014, "lr": 7.526395398866493e-05}
{"train_loss": 0.18015709519386292, "global_step": 90290, "epoch": 1014, "lr": 7.52634536997886e-05}
{"train_loss": 0.22472499310970306, "global_step": 90291, "epoch": 1014, "lr": 7.526295340751588e-05}
{"train_loss": 0.1870073527097702, "global_step": 90292, "epoch": 1014, "lr": 7.526245311184688e-05}
{"train_loss": 0.2563548982143402, "global_step": 90293, "epoch": 1014, "lr": 7.526195281278166e-05}
{"train_loss": 0.1472124606370926, "global_step": 90294, "epoch": 1014, "lr": 7.526145251032027e-05}
{"train_loss": 0.18793143332004547, "global_step": 90295, "epoch": 1014, "lr": 7.526095220446277e-05}
{"train_loss": 0.1375586986541748, "global_step": 90296, "epoch": 1014, "lr": 7.526045189520926e-05}
{"train_loss": 0.1277419626712799, "global_step": 90297, "epoch": 1014, "lr": 7.525995158255978e-05}
{"train_loss": 0.20073607563972473, "global_step": 90298, "epoch": 1014, "lr": 7.525945126651441e-05}
{"train_loss": 0.1483238935470581, "global_step": 90299, "epoch": 1014, "lr": 7.525895094707322e-05}
{"train_loss": 0.17207102477550507, "global_step": 90300, "epoch": 1014, "lr": 7.525845062423628e-05}
{"train_loss": 0.220842644572258, "global_step": 90301, "epoch": 1014, "lr": 7.525795029800363e-05}
{"train_loss": 0.1761718988418579, "global_step": 90302, "epoch": 1014, "lr": 7.525744996837536e-05}
{"train_loss": 0.144158735871315, "global_step": 90303, "epoch": 1014, "lr": 7.525694963535154e-05}
{"train_loss": 0.22229717671871185, "global_step": 90304, "epoch": 1014, "lr": 7.525644929893223e-05}
{"train_loss": 0.15996153652668, "global_step": 90305, "epoch": 1014, "lr": 7.52559489591175e-05}
{"train_loss": 0.18733800947666168, "global_step": 90306, "epoch": 1014, "lr": 7.525544861590741e-05}
{"train_loss": 0.1465112566947937, "global_step": 90307, "epoch": 1014, "lr": 7.525494826930204e-05}
{"train_loss": 0.19747310876846313, "global_step": 90308, "epoch": 1014, "lr": 7.525444791930146e-05}
{"train_loss": 0.13142450153827667, "global_step": 90309, "epoch": 1014, "lr": 7.52539475659057e-05}
{"train_loss": 0.20794230699539185, "global_step": 90310, "epoch": 1014, "lr": 7.525344720911487e-05}
{"train_loss": 0.2010020911693573, "global_step": 90311, "epoch": 1014, "lr": 7.525294684892902e-05}
{"train_loss": 0.19015386700630188, "global_step": 90312, "epoch": 1014, "lr": 7.525244648534822e-05}
{"train_loss": 0.24071478843688965, "global_step": 90313, "epoch": 1014, "lr": 7.525194611837255e-05}
{"train_loss": 0.1174248531460762, "global_step": 90314, "epoch": 1014, "lr": 7.525144574800203e-05}
{"train_loss": 0.21618397533893585, "global_step": 90315, "epoch": 1014, "lr": 7.525094537423678e-05}
{"train_loss": 0.14196164906024933, "global_step": 90316, "epoch": 1014, "lr": 7.525044499707685e-05}
{"train_loss": 0.23412223160266876, "global_step": 90317, "epoch": 1014, "lr": 7.52499446165223e-05}
{"train_loss": 0.1794036328792572, "global_step": 90318, "epoch": 1014, "lr": 7.524944423257321e-05}
{"train_loss": 0.13266587257385254, "global_step": 90319, "epoch": 1014, "lr": 7.524894384522963e-05}
{"train_loss": 0.19268064200878143, "global_step": 90320, "epoch": 1014, "lr": 7.524844345449164e-05}
{"train_loss": 0.2640543282032013, "global_step": 90321, "epoch": 1014, "lr": 7.524794306035932e-05}
{"train_loss": 0.13042588531970978, "global_step": 90322, "epoch": 1014, "lr": 7.52474426628327e-05}
{"train_loss": 0.19633230566978455, "global_step": 90323, "epoch": 1014, "lr": 7.524694226191189e-05}
{"train_loss": 0.3604668974876404, "global_step": 90324, "epoch": 1014, "lr": 7.524644185759692e-05}
{"train_loss": 0.13997246325016022, "global_step": 90325, "epoch": 1014, "lr": 7.524594144988789e-05}
{"train_loss": 0.19243241846561432, "global_step": 90326, "epoch": 1014, "lr": 7.524544103878483e-05}
{"train_loss": 0.23356200754642487, "global_step": 90327, "epoch": 1014, "lr": 7.524494062428784e-05}
{"train_loss": 0.12439858168363571, "global_step": 90328, "epoch": 1014, "lr": 7.524444020639696e-05}
{"train_loss": 0.23736852407455444, "global_step": 90329, "epoch": 1014, "lr": 7.524393978511229e-05}
{"train_loss": 0.14195415377616882, "global_step": 90330, "epoch": 1014, "lr": 7.524343936043387e-05}
{"train_loss": 0.201771080493927, "global_step": 90331, "epoch": 1014, "lr": 7.524293893236179e-05}
{"train_loss": 0.23299503326416016, "global_step": 90332, "epoch": 1014, "lr": 7.52424385008961e-05}
{"train_loss": 0.24638432264328003, "global_step": 90333, "epoch": 1014, "lr": 7.524193806603687e-05}
{"train_loss": 0.18939352948009297, "global_step": 90334, "epoch": 1014, "lr": 7.524143762778416e-05, "val_loss": 4.378035068511963}
{"train_loss": 0.20211152732372284, "global_step": 90335, "epoch": 1015, "lr": 7.524093718613807e-05}
{"train_loss": 0.1834758222103119, "global_step": 90336, "epoch": 1015, "lr": 7.524043674109861e-05}
{"train_loss": 0.157037153840065, "global_step": 90337, "epoch": 1015, "lr": 7.52399362926659e-05}
{"train_loss": 0.15220002830028534, "global_step": 90338, "epoch": 1015, "lr": 7.523943584083998e-05}
{"train_loss": 0.15707921981811523, "global_step": 90339, "epoch": 1015, "lr": 7.523893538562094e-05}
{"train_loss": 0.23614032566547394, "global_step": 90340, "epoch": 1015, "lr": 7.523843492700884e-05}
{"train_loss": 0.34486517310142517, "global_step": 90341, "epoch": 1015, "lr": 7.52379344650037e-05}
{"train_loss": 0.1861632615327835, "global_step": 90342, "epoch": 1015, "lr": 7.523743399960567e-05}
{"train_loss": 0.1678641438484192, "global_step": 90343, "epoch": 1015, "lr": 7.523693353081475e-05}
{"train_loss": 0.1548287272453308, "global_step": 90344, "epoch": 1015, "lr": 7.523643305863104e-05}
{"train_loss": 0.10939998924732208, "global_step": 90345, "epoch": 1015, "lr": 7.523593258305461e-05}
{"train_loss": 0.15635046362876892, "global_step": 90346, "epoch": 1015, "lr": 7.52354321040855e-05}
{"train_loss": 0.14625534415245056, "global_step": 90347, "epoch": 1015, "lr": 7.52349316217238e-05}
{"train_loss": 0.21362873911857605, "global_step": 90348, "epoch": 1015, "lr": 7.523443113596957e-05}
{"train_loss": 0.19739918410778046, "global_step": 90349, "epoch": 1015, "lr": 7.523393064682287e-05}
{"train_loss": 0.218780055642128, "global_step": 90350, "epoch": 1015, "lr": 7.523343015428379e-05}
{"train_loss": 0.11376417428255081, "global_step": 90351, "epoch": 1015, "lr": 7.523292965835238e-05}
{"train_loss": 0.11681757867336273, "global_step": 90352, "epoch": 1015, "lr": 7.523242915902871e-05}
{"train_loss": 0.15302206575870514, "global_step": 90353, "epoch": 1015, "lr": 7.523192865631283e-05}
{"train_loss": 0.15774761140346527, "global_step": 90354, "epoch": 1015, "lr": 7.523142815020485e-05}
{"train_loss": 0.17866568267345428, "global_step": 90355, "epoch": 1015, "lr": 7.52309276407048e-05}
{"train_loss": 0.22285084426403046, "global_step": 90356, "epoch": 1015, "lr": 7.523042712781277e-05}
{"train_loss": 0.1957778036594391, "global_step": 90357, "epoch": 1015, "lr": 7.522992661152879e-05}
{"train_loss": 0.16308899223804474, "global_step": 90358, "epoch": 1015, "lr": 7.522942609185297e-05}
{"train_loss": 0.1918138563632965, "global_step": 90359, "epoch": 1015, "lr": 7.522892556878537e-05}
{"train_loss": 0.26344189047813416, "global_step": 90360, "epoch": 1015, "lr": 7.522842504232603e-05}
{"train_loss": 0.13200804591178894, "global_step": 90361, "epoch": 1015, "lr": 7.522792451247504e-05}
{"train_loss": 0.23207217454910278, "global_step": 90362, "epoch": 1015, "lr": 7.522742397923248e-05}
{"train_loss": 0.25232940912246704, "global_step": 90363, "epoch": 1015, "lr": 7.522692344259839e-05}
{"train_loss": 0.22240859270095825, "global_step": 90364, "epoch": 1015, "lr": 7.522642290257285e-05}
{"train_loss": 0.20306263864040375, "global_step": 90365, "epoch": 1015, "lr": 7.522592235915592e-05}
{"train_loss": 0.23511505126953125, "global_step": 90366, "epoch": 1015, "lr": 7.522542181234768e-05}
{"train_loss": 0.2071210741996765, "global_step": 90367, "epoch": 1015, "lr": 7.522492126214818e-05}
{"train_loss": 0.23496130108833313, "global_step": 90368, "epoch": 1015, "lr": 7.522442070855749e-05}
{"train_loss": 0.22677072882652283, "global_step": 90369, "epoch": 1015, "lr": 7.522392015157571e-05}
{"train_loss": 0.14591337740421295, "global_step": 90370, "epoch": 1015, "lr": 7.522341959120286e-05}
{"train_loss": 0.21894729137420654, "global_step": 90371, "epoch": 1015, "lr": 7.522291902743904e-05}
{"train_loss": 0.17213885486125946, "global_step": 90372, "epoch": 1015, "lr": 7.522241846028429e-05}
{"train_loss": 0.23046599328517914, "global_step": 90373, "epoch": 1015, "lr": 7.522191788973871e-05}
{"train_loss": 0.19257748126983643, "global_step": 90374, "epoch": 1015, "lr": 7.522141731580235e-05}
{"train_loss": 0.22708779573440552, "global_step": 90375, "epoch": 1015, "lr": 7.522091673847527e-05}
{"train_loss": 0.19355592131614685, "global_step": 90376, "epoch": 1015, "lr": 7.522041615775755e-05}
{"train_loss": 0.18723995983600616, "global_step": 90377, "epoch": 1015, "lr": 7.521991557364924e-05}
{"train_loss": 0.1761147826910019, "global_step": 90378, "epoch": 1015, "lr": 7.521941498615042e-05}
{"train_loss": 0.1396011859178543, "global_step": 90379, "epoch": 1015, "lr": 7.521891439526118e-05}
{"train_loss": 0.15758295357227325, "global_step": 90380, "epoch": 1015, "lr": 7.521841380098154e-05}
{"train_loss": 0.2261306196451187, "global_step": 90381, "epoch": 1015, "lr": 7.52179132033116e-05}
{"train_loss": 0.16056577861309052, "global_step": 90382, "epoch": 1015, "lr": 7.521741260225143e-05}
{"train_loss": 0.1714627891778946, "global_step": 90383, "epoch": 1015, "lr": 7.521691199780107e-05}
{"train_loss": 0.18803314864635468, "global_step": 90384, "epoch": 1015, "lr": 7.521641138996062e-05}
{"train_loss": 0.10471660643815994, "global_step": 90385, "epoch": 1015, "lr": 7.521591077873011e-05}
{"train_loss": 0.12093189358711243, "global_step": 90386, "epoch": 1015, "lr": 7.521541016410964e-05}
{"train_loss": 0.12996134161949158, "global_step": 90387, "epoch": 1015, "lr": 7.521490954609927e-05}
{"train_loss": 0.18116843700408936, "global_step": 90388, "epoch": 1015, "lr": 7.521440892469906e-05}
{"train_loss": 0.2177523672580719, "global_step": 90389, "epoch": 1015, "lr": 7.521390829990908e-05}
{"train_loss": 0.2594197392463684, "global_step": 90390, "epoch": 1015, "lr": 7.52134076717294e-05}
{"train_loss": 0.17532099783420563, "global_step": 90391, "epoch": 1015, "lr": 7.521290704016006e-05}
{"train_loss": 0.17987871170043945, "global_step": 90392, "epoch": 1015, "lr": 7.52124064052012e-05}
{"train_loss": 0.1462574601173401, "global_step": 90393, "epoch": 1015, "lr": 7.521190576685279e-05}
{"train_loss": 0.1338219791650772, "global_step": 90394, "epoch": 1015, "lr": 7.521140512511497e-05}
{"train_loss": 0.16340097784996033, "global_step": 90395, "epoch": 1015, "lr": 7.52109044799878e-05}
{"train_loss": 0.23317177593708038, "global_step": 90396, "epoch": 1015, "lr": 7.521040383147131e-05}
{"train_loss": 0.14588741958141327, "global_step": 90397, "epoch": 1015, "lr": 7.520990317956559e-05}
{"train_loss": 0.1372155398130417, "global_step": 90398, "epoch": 1015, "lr": 7.520940252427071e-05}
{"train_loss": 0.170301616191864, "global_step": 90399, "epoch": 1015, "lr": 7.520890186558672e-05}
{"train_loss": 0.14429596066474915, "global_step": 90400, "epoch": 1015, "lr": 7.520840120351372e-05}
{"train_loss": 0.24031531810760498, "global_step": 90401, "epoch": 1015, "lr": 7.520790053805175e-05}
{"train_loss": 0.30613473057746887, "global_step": 90402, "epoch": 1015, "lr": 7.520739986920088e-05}
{"train_loss": 0.13193348050117493, "global_step": 90403, "epoch": 1015, "lr": 7.520689919696118e-05}
{"train_loss": 0.13517557084560394, "global_step": 90404, "epoch": 1015, "lr": 7.520639852133273e-05}
{"train_loss": 0.14666059613227844, "global_step": 90405, "epoch": 1015, "lr": 7.520589784231557e-05}
{"train_loss": 0.23844923079013824, "global_step": 90406, "epoch": 1015, "lr": 7.520539715990981e-05}
{"train_loss": 0.159819096326828, "global_step": 90407, "epoch": 1015, "lr": 7.520489647411549e-05}
{"train_loss": 0.17810845375061035, "global_step": 90408, "epoch": 1015, "lr": 7.520439578493265e-05}
{"train_loss": 0.23217615485191345, "global_step": 90409, "epoch": 1015, "lr": 7.52038950923614e-05}
{"train_loss": 0.2162913680076599, "global_step": 90410, "epoch": 1015, "lr": 7.52033943964018e-05}
{"train_loss": 0.15694110095500946, "global_step": 90411, "epoch": 1015, "lr": 7.520289369705392e-05}
{"train_loss": 0.1567368507385254, "global_step": 90412, "epoch": 1015, "lr": 7.52023929943178e-05}
{"train_loss": 0.10887254774570465, "global_step": 90413, "epoch": 1015, "lr": 7.520189228819351e-05}
{"train_loss": 0.18590839207172394, "global_step": 90414, "epoch": 1015, "lr": 7.520139157868118e-05}
{"train_loss": 0.21874822676181793, "global_step": 90415, "epoch": 1015, "lr": 7.520089086578079e-05}
{"train_loss": 0.2020946443080902, "global_step": 90416, "epoch": 1015, "lr": 7.520039014949248e-05}
{"train_loss": 0.16631700098514557, "global_step": 90417, "epoch": 1015, "lr": 7.519988942981627e-05}
{"train_loss": 0.14827100932598114, "global_step": 90418, "epoch": 1015, "lr": 7.519938870675223e-05}
{"train_loss": 0.21763759851455688, "global_step": 90419, "epoch": 1015, "lr": 7.519888798030044e-05}
{"train_loss": 0.17774461209774017, "global_step": 90420, "epoch": 1015, "lr": 7.519838725046098e-05}
{"train_loss": 0.10502823442220688, "global_step": 90421, "epoch": 1015, "lr": 7.51978865172339e-05}
{"train_loss": 0.23519529402256012, "global_step": 90422, "epoch": 1015, "lr": 7.519738578061926e-05}
{"train_loss": 0.1831693114357048, "global_step": 90423, "epoch": 1015, "lr": 7.519688504061714e-05, "val_loss": 4.614508628845215, "train_action_mse_error": 21.465919494628906}
{"train_loss": 0.1346014440059662, "global_step": 90424, "epoch": 1016, "lr": 7.519638429722762e-05}
{"train_loss": 0.11850551515817642, "global_step": 90425, "epoch": 1016, "lr": 7.519588355045075e-05}
{"train_loss": 0.15827418863773346, "global_step": 90426, "epoch": 1016, "lr": 7.519538280028659e-05}
{"train_loss": 0.19009588658809662, "global_step": 90427, "epoch": 1016, "lr": 7.519488204673521e-05}
{"train_loss": 0.18694916367530823, "global_step": 90428, "epoch": 1016, "lr": 7.519438128979672e-05}
{"train_loss": 0.1515401303768158, "global_step": 90429, "epoch": 1016, "lr": 7.519388052947112e-05}
{"train_loss": 0.11374812573194504, "global_step": 90430, "epoch": 1016, "lr": 7.519337976575854e-05}
{"train_loss": 0.15247571468353271, "global_step": 90431, "epoch": 1016, "lr": 7.5192878998659e-05}
{"train_loss": 0.16265664994716644, "global_step": 90432, "epoch": 1016, "lr": 7.519237822817257e-05}
{"train_loss": 0.22044837474822998, "global_step": 90433, "epoch": 1016, "lr": 7.519187745429937e-05}
{"train_loss": 0.19774307310581207, "global_step": 90434, "epoch": 1016, "lr": 7.519137667703939e-05}
{"train_loss": 0.1705072969198227, "global_step": 90435, "epoch": 1016, "lr": 7.519087589639276e-05}
{"train_loss": 0.09389036148786545, "global_step": 90436, "epoch": 1016, "lr": 7.519037511235951e-05}
{"train_loss": 0.25452321767807007, "global_step": 90437, "epoch": 1016, "lr": 7.518987432493973e-05}
{"train_loss": 0.16210174560546875, "global_step": 90438, "epoch": 1016, "lr": 7.518937353413348e-05}
{"train_loss": 0.23207561671733856, "global_step": 90439, "epoch": 1016, "lr": 7.518887273994082e-05}
{"train_loss": 0.22108963131904602, "global_step": 90440, "epoch": 1016, "lr": 7.518837194236181e-05}
{"train_loss": 0.1370076835155487, "global_step": 90441, "epoch": 1016, "lr": 7.518787114139656e-05}
{"train_loss": 0.16941088438034058, "global_step": 90442, "epoch": 1016, "lr": 7.518737033704508e-05}
{"train_loss": 0.21816711127758026, "global_step": 90443, "epoch": 1016, "lr": 7.518686952930749e-05}
{"train_loss": 0.13413529098033905, "global_step": 90444, "epoch": 1016, "lr": 7.518636871818381e-05}
{"train_loss": 0.14181208610534668, "global_step": 90445, "epoch": 1016, "lr": 7.518586790367413e-05}
{"train_loss": 0.18898126482963562, "global_step": 90446, "epoch": 1016, "lr": 7.518536708577853e-05}
{"train_loss": 0.2988989055156708, "global_step": 90447, "epoch": 1016, "lr": 7.518486626449706e-05}
{"train_loss": 0.2240048497915268, "global_step": 90448, "epoch": 1016, "lr": 7.51843654398298e-05}
{"train_loss": 0.199710875749588, "global_step": 90449, "epoch": 1016, "lr": 7.51838646117768e-05}
{"train_loss": 0.10996846109628677, "global_step": 90450, "epoch": 1016, "lr": 7.518336378033814e-05}
{"train_loss": 0.1823875904083252, "global_step": 90451, "epoch": 1016, "lr": 7.518286294551389e-05}
{"train_loss": 0.1659725308418274, "global_step": 90452, "epoch": 1016, "lr": 7.518236210730409e-05}
{"train_loss": 0.15451917052268982, "global_step": 90453, "epoch": 1016, "lr": 7.518186126570885e-05}
{"train_loss": 0.2146950662136078, "global_step": 90454, "epoch": 1016, "lr": 7.518136042072821e-05}
{"train_loss": 0.11595764756202698, "global_step": 90455, "epoch": 1016, "lr": 7.518085957236223e-05}
{"train_loss": 0.22247804701328278, "global_step": 90456, "epoch": 1016, "lr": 7.5180358720611e-05}
{"train_loss": 0.17088817059993744, "global_step": 90457, "epoch": 1016, "lr": 7.517985786547457e-05}
{"train_loss": 0.22187943756580353, "global_step": 90458, "epoch": 1016, "lr": 7.517935700695303e-05}
{"train_loss": 0.15768086910247803, "global_step": 90459, "epoch": 1016, "lr": 7.517885614504642e-05}
{"train_loss": 0.15213057398796082, "global_step": 90460, "epoch": 1016, "lr": 7.517835527975483e-05}
{"train_loss": 0.17524036765098572, "global_step": 90461, "epoch": 1016, "lr": 7.517785441107832e-05}
{"train_loss": 0.2120533287525177, "global_step": 90462, "epoch": 1016, "lr": 7.517735353901694e-05}
{"train_loss": 0.2707119286060333, "global_step": 90463, "epoch": 1016, "lr": 7.517685266357077e-05}
{"train_loss": 0.1185518428683281, "global_step": 90464, "epoch": 1016, "lr": 7.517635178473989e-05}
{"train_loss": 0.16481314599514008, "global_step": 90465, "epoch": 1016, "lr": 7.517585090252434e-05}
{"train_loss": 0.24560566246509552, "global_step": 90466, "epoch": 1016, "lr": 7.517535001692421e-05}
{"train_loss": 0.14668679237365723, "global_step": 90467, "epoch": 1016, "lr": 7.517484912793959e-05}
{"train_loss": 0.22467732429504395, "global_step": 90468, "epoch": 1016, "lr": 7.517434823557048e-05}
{"train_loss": 0.2361728996038437, "global_step": 90469, "epoch": 1016, "lr": 7.517384733981703e-05}
{"train_loss": 0.17172247171401978, "global_step": 90470, "epoch": 1016, "lr": 7.517334644067922e-05}
{"train_loss": 0.12382746487855911, "global_step": 90471, "epoch": 1016, "lr": 7.517284553815718e-05}
{"train_loss": 0.1795489341020584, "global_step": 90472, "epoch": 1016, "lr": 7.517234463225095e-05}
{"train_loss": 0.16027739644050598, "global_step": 90473, "epoch": 1016, "lr": 7.517184372296061e-05}
{"train_loss": 0.13407473266124725, "global_step": 90474, "epoch": 1016, "lr": 7.517134281028621e-05}
{"train_loss": 0.17525604367256165, "global_step": 90475, "epoch": 1016, "lr": 7.517084189422785e-05}
{"train_loss": 0.18679681420326233, "global_step": 90476, "epoch": 1016, "lr": 7.517034097478557e-05}
{"train_loss": 0.16586489975452423, "global_step": 90477, "epoch": 1016, "lr": 7.516984005195944e-05}
{"train_loss": 0.22551897168159485, "global_step": 90478, "epoch": 1016, "lr": 7.516933912574952e-05}
{"train_loss": 0.1623765230178833, "global_step": 90479, "epoch": 1016, "lr": 7.516883819615593e-05}
{"train_loss": 0.14889147877693176, "global_step": 90480, "epoch": 1016, "lr": 7.516833726317866e-05}
{"train_loss": 0.22626212239265442, "global_step": 90481, "epoch": 1016, "lr": 7.516783632681781e-05}
{"train_loss": 0.1461252123117447, "global_step": 90482, "epoch": 1016, "lr": 7.516733538707345e-05}
{"train_loss": 0.17772717773914337, "global_step": 90483, "epoch": 1016, "lr": 7.516683444394568e-05}
{"train_loss": 0.16621091961860657, "global_step": 90484, "epoch": 1016, "lr": 7.516633349743452e-05}
{"train_loss": 0.19063647091388702, "global_step": 90485, "epoch": 1016, "lr": 7.516583254754004e-05}
{"train_loss": 0.20238056778907776, "global_step": 90486, "epoch": 1016, "lr": 7.516533159426233e-05}
{"train_loss": 0.15912887454032898, "global_step": 90487, "epoch": 1016, "lr": 7.516483063760145e-05}
{"train_loss": 0.17864039540290833, "global_step": 90488, "epoch": 1016, "lr": 7.516432967755748e-05}
{"train_loss": 0.15221138298511505, "global_step": 90489, "epoch": 1016, "lr": 7.516382871413045e-05}
{"train_loss": 0.20038969814777374, "global_step": 90490, "epoch": 1016, "lr": 7.516332774732047e-05}
{"train_loss": 0.16088314354419708, "global_step": 90491, "epoch": 1016, "lr": 7.516282677712756e-05}
{"train_loss": 0.20941293239593506, "global_step": 90492, "epoch": 1016, "lr": 7.516232580355184e-05}
{"train_loss": 0.19124533236026764, "global_step": 90493, "epoch": 1016, "lr": 7.516182482659335e-05}
{"train_loss": 0.14632776379585266, "global_step": 90494, "epoch": 1016, "lr": 7.516132384625214e-05}
{"train_loss": 0.14722417294979095, "global_step": 90495, "epoch": 1016, "lr": 7.516082286252831e-05}
{"train_loss": 0.11954431980848312, "global_step": 90496, "epoch": 1016, "lr": 7.516032187542191e-05}
{"train_loss": 0.1366352140903473, "global_step": 90497, "epoch": 1016, "lr": 7.515982088493302e-05}
{"train_loss": 0.16668693721294403, "global_step": 90498, "epoch": 1016, "lr": 7.515931989106168e-05}
{"train_loss": 0.2203313708305359, "global_step": 90499, "epoch": 1016, "lr": 7.515881889380798e-05}
{"train_loss": 0.16598448157310486, "global_step": 90500, "epoch": 1016, "lr": 7.5158317893172e-05}
{"train_loss": 0.16628751158714294, "global_step": 90501, "epoch": 1016, "lr": 7.515781688915377e-05}
{"train_loss": 0.19311083853244781, "global_step": 90502, "epoch": 1016, "lr": 7.515731588175341e-05}
{"train_loss": 0.26304519176483154, "global_step": 90503, "epoch": 1016, "lr": 7.515681487097092e-05}
{"train_loss": 0.13623149693012238, "global_step": 90504, "epoch": 1016, "lr": 7.515631385680643e-05}
{"train_loss": 0.20147831737995148, "global_step": 90505, "epoch": 1016, "lr": 7.515581283925997e-05}
{"train_loss": 0.18184639513492584, "global_step": 90506, "epoch": 1016, "lr": 7.515531181833162e-05}
{"train_loss": 0.16010110080242157, "global_step": 90507, "epoch": 1016, "lr": 7.515481079402143e-05}
{"train_loss": 0.17499330639839172, "global_step": 90508, "epoch": 1016, "lr": 7.515430976632951e-05}
{"train_loss": 0.26440727710723877, "global_step": 90509, "epoch": 1016, "lr": 7.515380873525588e-05}
{"train_loss": 0.273714154958725, "global_step": 90510, "epoch": 1016, "lr": 7.515330770080062e-05}
{"train_loss": 0.18034476041793823, "global_step": 90511, "epoch": 1016, "lr": 7.515280666296383e-05}
{"train_loss": 0.17963630354471422, "global_step": 90512, "epoch": 1016, "lr": 7.515230562174554e-05, "val_loss": 4.411946773529053}
{"train_loss": 0.2040175497531891, "global_step": 90513, "epoch": 1017, "lr": 7.51518045771458e-05}
{"train_loss": 0.16506214439868927, "global_step": 90514, "epoch": 1017, "lr": 7.515130352916475e-05}
{"train_loss": 0.1580106019973755, "global_step": 90515, "epoch": 1017, "lr": 7.51508024778024e-05}
{"train_loss": 0.11573529243469238, "global_step": 90516, "epoch": 1017, "lr": 7.515030142305882e-05}
{"train_loss": 0.1976424902677536, "global_step": 90517, "epoch": 1017, "lr": 7.514980036493409e-05}
{"train_loss": 0.16247273981571198, "global_step": 90518, "epoch": 1017, "lr": 7.514929930342828e-05}
{"train_loss": 0.15458188951015472, "global_step": 90519, "epoch": 1017, "lr": 7.514879823854146e-05}
{"train_loss": 0.14970895648002625, "global_step": 90520, "epoch": 1017, "lr": 7.514829717027368e-05}
{"train_loss": 0.19725298881530762, "global_step": 90521, "epoch": 1017, "lr": 7.514779609862503e-05}
{"train_loss": 0.16859763860702515, "global_step": 90522, "epoch": 1017, "lr": 7.514729502359554e-05}
{"train_loss": 0.21272271871566772, "global_step": 90523, "epoch": 1017, "lr": 7.514679394518534e-05}
{"train_loss": 0.18921321630477905, "global_step": 90524, "epoch": 1017, "lr": 7.514629286339443e-05}
{"train_loss": 0.09414859116077423, "global_step": 90525, "epoch": 1017, "lr": 7.514579177822291e-05}
{"train_loss": 0.16818073391914368, "global_step": 90526, "epoch": 1017, "lr": 7.514529068967086e-05}
{"train_loss": 0.20886653661727905, "global_step": 90527, "epoch": 1017, "lr": 7.514478959773835e-05}
{"train_loss": 0.15311121940612793, "global_step": 90528, "epoch": 1017, "lr": 7.51442885024254e-05}
{"train_loss": 0.16457092761993408, "global_step": 90529, "epoch": 1017, "lr": 7.514378740373211e-05}
{"train_loss": 0.20123767852783203, "global_step": 90530, "epoch": 1017, "lr": 7.514328630165855e-05}
{"train_loss": 0.18384552001953125, "global_step": 90531, "epoch": 1017, "lr": 7.514278519620477e-05}
{"train_loss": 0.15758644044399261, "global_step": 90532, "epoch": 1017, "lr": 7.514228408737085e-05}
{"train_loss": 0.1618371605873108, "global_step": 90533, "epoch": 1017, "lr": 7.514178297515687e-05}
{"train_loss": 0.18780310451984406, "global_step": 90534, "epoch": 1017, "lr": 7.514128185956289e-05}
{"train_loss": 0.15960495173931122, "global_step": 90535, "epoch": 1017, "lr": 7.514078074058894e-05}
{"train_loss": 0.1598302721977234, "global_step": 90536, "epoch": 1017, "lr": 7.514027961823513e-05}
{"train_loss": 0.15923945605754852, "global_step": 90537, "epoch": 1017, "lr": 7.513977849250153e-05}
{"train_loss": 0.19036982953548431, "global_step": 90538, "epoch": 1017, "lr": 7.513927736338817e-05}
{"train_loss": 0.26998403668403625, "global_step": 90539, "epoch": 1017, "lr": 7.513877623089516e-05}
{"train_loss": 0.27332064509391785, "global_step": 90540, "epoch": 1017, "lr": 7.513827509502255e-05}
{"train_loss": 0.14805085957050323, "global_step": 90541, "epoch": 1017, "lr": 7.51377739557704e-05}
{"train_loss": 0.27606335282325745, "global_step": 90542, "epoch": 1017, "lr": 7.513727281313879e-05}
{"train_loss": 0.1439816951751709, "global_step": 90543, "epoch": 1017, "lr": 7.513677166712776e-05}
{"train_loss": 0.14833930134773254, "global_step": 90544, "epoch": 1017, "lr": 7.513627051773742e-05}
{"train_loss": 0.1361130326986313, "global_step": 90545, "epoch": 1017, "lr": 7.51357693649678e-05}
{"train_loss": 0.12272264063358307, "global_step": 90546, "epoch": 1017, "lr": 7.5135268208819e-05}
{"train_loss": 0.14851336181163788, "global_step": 90547, "epoch": 1017, "lr": 7.513476704929106e-05}
{"train_loss": 0.18862251937389374, "global_step": 90548, "epoch": 1017, "lr": 7.513426588638404e-05}
{"train_loss": 0.14430159330368042, "global_step": 90549, "epoch": 1017, "lr": 7.513376472009804e-05}
{"train_loss": 0.14507228136062622, "global_step": 90550, "epoch": 1017, "lr": 7.51332635504331e-05}
{"train_loss": 0.14253543317317963, "global_step": 90551, "epoch": 1017, "lr": 7.513276237738931e-05}
{"train_loss": 0.16815151274204254, "global_step": 90552, "epoch": 1017, "lr": 7.513226120096672e-05}
{"train_loss": 0.15159578621387482, "global_step": 90553, "epoch": 1017, "lr": 7.51317600211654e-05}
{"train_loss": 0.1765456348657608, "global_step": 90554, "epoch": 1017, "lr": 7.513125883798544e-05}
{"train_loss": 0.1350003331899643, "global_step": 90555, "epoch": 1017, "lr": 7.513075765142687e-05}
{"train_loss": 0.179024338722229, "global_step": 90556, "epoch": 1017, "lr": 7.513025646148979e-05}
{"train_loss": 0.21975503861904144, "global_step": 90557, "epoch": 1017, "lr": 7.512975526817425e-05}
{"train_loss": 0.16611915826797485, "global_step": 90558, "epoch": 1017, "lr": 7.51292540714803e-05}
{"train_loss": 0.1073603481054306, "global_step": 90559, "epoch": 1017, "lr": 7.512875287140806e-05}
{"train_loss": 0.18313461542129517, "global_step": 90560, "epoch": 1017, "lr": 7.512825166795755e-05}
{"train_loss": 0.24208299815654755, "global_step": 90561, "epoch": 1017, "lr": 7.512775046112886e-05}
{"train_loss": 0.1666998267173767, "global_step": 90562, "epoch": 1017, "lr": 7.512724925092203e-05}
{"train_loss": 0.22193880379199982, "global_step": 90563, "epoch": 1017, "lr": 7.512674803733716e-05}
{"train_loss": 0.20056132972240448, "global_step": 90564, "epoch": 1017, "lr": 7.512624682037432e-05}
{"train_loss": 0.25557783246040344, "global_step": 90565, "epoch": 1017, "lr": 7.512574560003353e-05}
{"train_loss": 0.15366831421852112, "global_step": 90566, "epoch": 1017, "lr": 7.512524437631493e-05}
{"train_loss": 0.16174882650375366, "global_step": 90567, "epoch": 1017, "lr": 7.512474314921854e-05}
{"train_loss": 0.21521353721618652, "global_step": 90568, "epoch": 1017, "lr": 7.51242419187444e-05}
{"train_loss": 0.16941328346729279, "global_step": 90569, "epoch": 1017, "lr": 7.512374068489264e-05}
{"train_loss": 0.20454367995262146, "global_step": 90570, "epoch": 1017, "lr": 7.512323944766328e-05}
{"train_loss": 0.26387661695480347, "global_step": 90571, "epoch": 1017, "lr": 7.512273820705641e-05}
{"train_loss": 0.2283596396446228, "global_step": 90572, "epoch": 1017, "lr": 7.51222369630721e-05}
{"train_loss": 0.1736813336610794, "global_step": 90573, "epoch": 1017, "lr": 7.512173571571043e-05}
{"train_loss": 0.14411485195159912, "global_step": 90574, "epoch": 1017, "lr": 7.512123446497143e-05}
{"train_loss": 0.2037859857082367, "global_step": 90575, "epoch": 1017, "lr": 7.512073321085519e-05}
{"train_loss": 0.1448696404695511, "global_step": 90576, "epoch": 1017, "lr": 7.512023195336178e-05}
{"train_loss": 0.1743033528327942, "global_step": 90577, "epoch": 1017, "lr": 7.511973069249125e-05}
{"train_loss": 0.23967774212360382, "global_step": 90578, "epoch": 1017, "lr": 7.511922942824368e-05}
{"train_loss": 0.15843962132930756, "global_step": 90579, "epoch": 1017, "lr": 7.511872816061913e-05}
{"train_loss": 0.27332931756973267, "global_step": 90580, "epoch": 1017, "lr": 7.511822688961768e-05}
{"train_loss": 0.12715847790241241, "global_step": 90581, "epoch": 1017, "lr": 7.51177256152394e-05}
{"train_loss": 0.15577203035354614, "global_step": 90582, "epoch": 1017, "lr": 7.511722433748433e-05}
{"train_loss": 0.1929008662700653, "global_step": 90583, "epoch": 1017, "lr": 7.511672305635256e-05}
{"train_loss": 0.1062861829996109, "global_step": 90584, "epoch": 1017, "lr": 7.511622177184417e-05}
{"train_loss": 0.11560880392789841, "global_step": 90585, "epoch": 1017, "lr": 7.511572048395921e-05}
{"train_loss": 0.14115789532661438, "global_step": 90586, "epoch": 1017, "lr": 7.511521919269773e-05}
{"train_loss": 0.12590497732162476, "global_step": 90587, "epoch": 1017, "lr": 7.511471789805982e-05}
{"train_loss": 0.13599902391433716, "global_step": 90588, "epoch": 1017, "lr": 7.511421660004554e-05}
{"train_loss": 0.09637971222400665, "global_step": 90589, "epoch": 1017, "lr": 7.511371529865495e-05}
{"train_loss": 0.14274518191814423, "global_step": 90590, "epoch": 1017, "lr": 7.511321399388815e-05}
{"train_loss": 0.26575446128845215, "global_step": 90591, "epoch": 1017, "lr": 7.511271268574516e-05}
{"train_loss": 0.1261175125837326, "global_step": 90592, "epoch": 1017, "lr": 7.511221137422609e-05}
{"train_loss": 0.18785251677036285, "global_step": 90593, "epoch": 1017, "lr": 7.511171005933099e-05}
{"train_loss": 0.23918908834457397, "global_step": 90594, "epoch": 1017, "lr": 7.511120874105991e-05}
{"train_loss": 0.22391565144062042, "global_step": 90595, "epoch": 1017, "lr": 7.511070741941296e-05}
{"train_loss": 0.16401837766170502, "global_step": 90596, "epoch": 1017, "lr": 7.511020609439016e-05}
{"train_loss": 0.1119171530008316, "global_step": 90597, "epoch": 1017, "lr": 7.51097047659916e-05}
{"train_loss": 0.13732504844665527, "global_step": 90598, "epoch": 1017, "lr": 7.510920343421735e-05}
{"train_loss": 0.1441257745027542, "global_step": 90599, "epoch": 1017, "lr": 7.510870209906748e-05}
{"train_loss": 0.1338047832250595, "global_step": 90600, "epoch": 1017, "lr": 7.510820076054205e-05}
{"train_loss": 0.1732910592736823, "global_step": 90601, "epoch": 1017, "lr": 7.510769941864112e-05, "val_loss": 4.379567623138428}
{"train_loss": 0.12403520941734314, "global_step": 90602, "epoch": 1018, "lr": 7.510719807336477e-05}
{"train_loss": 0.14185979962348938, "global_step": 90603, "epoch": 1018, "lr": 7.510669672471306e-05}
{"train_loss": 0.17275217175483704, "global_step": 90604, "epoch": 1018, "lr": 7.510619537268606e-05}
{"train_loss": 0.10437913984060287, "global_step": 90605, "epoch": 1018, "lr": 7.510569401728386e-05}
{"train_loss": 0.1850343495607376, "global_step": 90606, "epoch": 1018, "lr": 7.510519265850648e-05}
{"train_loss": 0.2385631948709488, "global_step": 90607, "epoch": 1018, "lr": 7.510469129635403e-05}
{"train_loss": 0.1309121698141098, "global_step": 90608, "epoch": 1018, "lr": 7.510418993082654e-05}
{"train_loss": 0.1862488090991974, "global_step": 90609, "epoch": 1018, "lr": 7.510368856192411e-05}
{"train_loss": 0.26249611377716064, "global_step": 90610, "epoch": 1018, "lr": 7.51031871896468e-05}
{"train_loss": 0.17556284368038177, "global_step": 90611, "epoch": 1018, "lr": 7.510268581399467e-05}
{"train_loss": 0.1485559493303299, "global_step": 90612, "epoch": 1018, "lr": 7.510218443496777e-05}
{"train_loss": 0.15261080861091614, "global_step": 90613, "epoch": 1018, "lr": 7.510168305256621e-05}
{"train_loss": 0.18934869766235352, "global_step": 90614, "epoch": 1018, "lr": 7.510118166679003e-05}
{"train_loss": 0.1327631026506424, "global_step": 90615, "epoch": 1018, "lr": 7.51006802776393e-05}
{"train_loss": 0.20540088415145874, "global_step": 90616, "epoch": 1018, "lr": 7.510017888511409e-05}
{"train_loss": 0.19603461027145386, "global_step": 90617, "epoch": 1018, "lr": 7.509967748921446e-05}
{"train_loss": 0.15425078570842743, "global_step": 90618, "epoch": 1018, "lr": 7.509917608994049e-05}
{"train_loss": 0.1535908430814743, "global_step": 90619, "epoch": 1018, "lr": 7.509867468729223e-05}
{"train_loss": 0.21846164762973785, "global_step": 90620, "epoch": 1018, "lr": 7.509817328126978e-05}
{"train_loss": 0.15150406956672668, "global_step": 90621, "epoch": 1018, "lr": 7.509767187187318e-05}
{"train_loss": 0.166978120803833, "global_step": 90622, "epoch": 1018, "lr": 7.509717045910251e-05}
{"train_loss": 0.17146490514278412, "global_step": 90623, "epoch": 1018, "lr": 7.509666904295782e-05}
{"train_loss": 0.21418102085590363, "global_step": 90624, "epoch": 1018, "lr": 7.50961676234392e-05}
{"train_loss": 0.20926687121391296, "global_step": 90625, "epoch": 1018, "lr": 7.50956662005467e-05}
{"train_loss": 0.16364508867263794, "global_step": 90626, "epoch": 1018, "lr": 7.509516477428039e-05}
{"train_loss": 0.20522062480449677, "global_step": 90627, "epoch": 1018, "lr": 7.509466334464032e-05}
{"train_loss": 0.18894462287425995, "global_step": 90628, "epoch": 1018, "lr": 7.509416191162661e-05}
{"train_loss": 0.18445755541324615, "global_step": 90629, "epoch": 1018, "lr": 7.509366047523929e-05}
{"train_loss": 0.24835923314094543, "global_step": 90630, "epoch": 1018, "lr": 7.509315903547842e-05}
{"train_loss": 0.2236672043800354, "global_step": 90631, "epoch": 1018, "lr": 7.50926575923441e-05}
{"train_loss": 0.2335783988237381, "global_step": 90632, "epoch": 1018, "lr": 7.509215614583636e-05}
{"train_loss": 0.14148077368736267, "global_step": 90633, "epoch": 1018, "lr": 7.50916546959553e-05}
{"train_loss": 0.170906201004982, "global_step": 90634, "epoch": 1018, "lr": 7.509115324270098e-05}
{"train_loss": 0.1998514086008072, "global_step": 90635, "epoch": 1018, "lr": 7.509065178607344e-05}
{"train_loss": 0.1525280922651291, "global_step": 90636, "epoch": 1018, "lr": 7.509015032607277e-05}
{"train_loss": 0.14870360493659973, "global_step": 90637, "epoch": 1018, "lr": 7.508964886269904e-05}
{"train_loss": 0.27659595012664795, "global_step": 90638, "epoch": 1018, "lr": 7.508914739595232e-05}
{"train_loss": 0.24791811406612396, "global_step": 90639, "epoch": 1018, "lr": 7.508864592583266e-05}
{"train_loss": 0.18928460776805878, "global_step": 90640, "epoch": 1018, "lr": 7.508814445234014e-05}
{"train_loss": 0.1294981986284256, "global_step": 90641, "epoch": 1018, "lr": 7.508764297547482e-05}
{"train_loss": 0.15993022918701172, "global_step": 90642, "epoch": 1018, "lr": 7.508714149523677e-05}
{"train_loss": 0.14782333374023438, "global_step": 90643, "epoch": 1018, "lr": 7.508664001162609e-05}
{"train_loss": 0.36084499955177307, "global_step": 90644, "epoch": 1018, "lr": 7.508613852464277e-05}
{"train_loss": 0.3222105801105499, "global_step": 90645, "epoch": 1018, "lr": 7.508563703428695e-05}
{"train_loss": 0.23095236718654633, "global_step": 90646, "epoch": 1018, "lr": 7.508513554055867e-05}
{"train_loss": 0.1662021279335022, "global_step": 90647, "epoch": 1018, "lr": 7.5084634043458e-05}
{"train_loss": 0.17206136882305145, "global_step": 90648, "epoch": 1018, "lr": 7.508413254298501e-05}
{"train_loss": 0.16478294134140015, "global_step": 90649, "epoch": 1018, "lr": 7.508363103913975e-05}
{"train_loss": 0.21291129291057587, "global_step": 90650, "epoch": 1018, "lr": 7.508312953192231e-05}
{"train_loss": 0.18365156650543213, "global_step": 90651, "epoch": 1018, "lr": 7.508262802133276e-05}
{"train_loss": 0.16475407779216766, "global_step": 90652, "epoch": 1018, "lr": 7.508212650737114e-05}
{"train_loss": 0.26244229078292847, "global_step": 90653, "epoch": 1018, "lr": 7.508162499003754e-05}
{"train_loss": 0.22265589237213135, "global_step": 90654, "epoch": 1018, "lr": 7.508112346933203e-05}
{"train_loss": 0.17762908339500427, "global_step": 90655, "epoch": 1018, "lr": 7.508062194525465e-05}
{"train_loss": 0.28738075494766235, "global_step": 90656, "epoch": 1018, "lr": 7.50801204178055e-05}
{"train_loss": 0.24657975137233734, "global_step": 90657, "epoch": 1018, "lr": 7.507961888698463e-05}
{"train_loss": 0.20126250386238098, "global_step": 90658, "epoch": 1018, "lr": 7.507911735279212e-05}
{"train_loss": 0.1401805430650711, "global_step": 90659, "epoch": 1018, "lr": 7.507861581522801e-05}
{"train_loss": 0.18310853838920593, "global_step": 90660, "epoch": 1018, "lr": 7.507811427429238e-05}
{"train_loss": 0.21068927645683289, "global_step": 90661, "epoch": 1018, "lr": 7.507761272998532e-05}
{"train_loss": 0.18853981792926788, "global_step": 90662, "epoch": 1018, "lr": 7.507711118230689e-05}
{"train_loss": 0.22003109753131866, "global_step": 90663, "epoch": 1018, "lr": 7.507660963125713e-05}
{"train_loss": 0.14029255509376526, "global_step": 90664, "epoch": 1018, "lr": 7.507610807683613e-05}
{"train_loss": 0.197587788105011, "global_step": 90665, "epoch": 1018, "lr": 7.507560651904397e-05}
{"train_loss": 0.2245384305715561, "global_step": 90666, "epoch": 1018, "lr": 7.507510495788068e-05}
{"train_loss": 0.23125059902668, "global_step": 90667, "epoch": 1018, "lr": 7.507460339334635e-05}
{"train_loss": 0.23276934027671814, "global_step": 90668, "epoch": 1018, "lr": 7.507410182544106e-05}
{"train_loss": 0.17549006640911102, "global_step": 90669, "epoch": 1018, "lr": 7.507360025416483e-05}
{"train_loss": 0.31604090332984924, "global_step": 90670, "epoch": 1018, "lr": 7.50730986795178e-05}
{"train_loss": 0.26962077617645264, "global_step": 90671, "epoch": 1018, "lr": 7.507259710149996e-05}
{"train_loss": 0.09530956298112869, "global_step": 90672, "epoch": 1018, "lr": 7.507209552011144e-05}
{"train_loss": 0.20072415471076965, "global_step": 90673, "epoch": 1018, "lr": 7.507159393535228e-05}
{"train_loss": 0.16614790260791779, "global_step": 90674, "epoch": 1018, "lr": 7.507109234722253e-05}
{"train_loss": 0.20222266018390656, "global_step": 90675, "epoch": 1018, "lr": 7.50705907557223e-05}
{"train_loss": 0.32631537318229675, "global_step": 90676, "epoch": 1018, "lr": 7.507008916085163e-05}
{"train_loss": 0.13150042295455933, "global_step": 90677, "epoch": 1018, "lr": 7.50695875626106e-05}
{"train_loss": 0.1992790549993515, "global_step": 90678, "epoch": 1018, "lr": 7.506908596099925e-05}
{"train_loss": 0.1249251738190651, "global_step": 90679, "epoch": 1018, "lr": 7.506858435601767e-05}
{"train_loss": 0.20286451280117035, "global_step": 90680, "epoch": 1018, "lr": 7.506808274766593e-05}
{"train_loss": 0.18511724472045898, "global_step": 90681, "epoch": 1018, "lr": 7.50675811359441e-05}
{"train_loss": 0.1670137494802475, "global_step": 90682, "epoch": 1018, "lr": 7.506707952085223e-05}
{"train_loss": 0.17523278295993805, "global_step": 90683, "epoch": 1018, "lr": 7.50665779023904e-05}
{"train_loss": 0.14546073973178864, "global_step": 90684, "epoch": 1018, "lr": 7.506607628055867e-05}
{"train_loss": 0.23054155707359314, "global_step": 90685, "epoch": 1018, "lr": 7.506557465535711e-05}
{"train_loss": 0.1966288983821869, "global_step": 90686, "epoch": 1018, "lr": 7.506507302678579e-05}
{"train_loss": 0.1250963807106018, "global_step": 90687, "epoch": 1018, "lr": 7.506457139484478e-05}
{"train_loss": 0.16608178615570068, "global_step": 90688, "epoch": 1018, "lr": 7.506406975953415e-05}
{"train_loss": 0.20854274928569794, "global_step": 90689, "epoch": 1018, "lr": 7.506356812085396e-05}
{"train_loss": 0.1927931805172663, "global_step": 90690, "epoch": 1018, "lr": 7.506306647880426e-05, "val_loss": 4.291771411895752}
{"train_loss": 0.14535826444625854, "global_step": 90691, "epoch": 1019, "lr": 7.506256483338515e-05}
{"train_loss": 0.19221988320350647, "global_step": 90692, "epoch": 1019, "lr": 7.506206318459668e-05}
{"train_loss": 0.1769724190235138, "global_step": 90693, "epoch": 1019, "lr": 7.506156153243893e-05}
{"train_loss": 0.1648140549659729, "global_step": 90694, "epoch": 1019, "lr": 7.506105987691195e-05}
{"train_loss": 0.1870337426662445, "global_step": 90695, "epoch": 1019, "lr": 7.506055821801581e-05}
{"train_loss": 0.16837146878242493, "global_step": 90696, "epoch": 1019, "lr": 7.50600565557506e-05}
{"train_loss": 0.20235484838485718, "global_step": 90697, "epoch": 1019, "lr": 7.505955489011636e-05}
{"train_loss": 0.09638413041830063, "global_step": 90698, "epoch": 1019, "lr": 7.505905322111317e-05}
{"train_loss": 0.18969455361366272, "global_step": 90699, "epoch": 1019, "lr": 7.505855154874111e-05}
{"train_loss": 0.31902194023132324, "global_step": 90700, "epoch": 1019, "lr": 7.50580498730002e-05}
{"train_loss": 0.20039749145507812, "global_step": 90701, "epoch": 1019, "lr": 7.505754819389057e-05}
{"train_loss": 0.14954379200935364, "global_step": 90702, "epoch": 1019, "lr": 7.505704651141226e-05}
{"train_loss": 0.1497470885515213, "global_step": 90703, "epoch": 1019, "lr": 7.505654482556531e-05}
{"train_loss": 0.17709662020206451, "global_step": 90704, "epoch": 1019, "lr": 7.505604313634983e-05}
{"train_loss": 0.14367303252220154, "global_step": 90705, "epoch": 1019, "lr": 7.505554144376586e-05}
{"train_loss": 0.18434713780879974, "global_step": 90706, "epoch": 1019, "lr": 7.50550397478135e-05}
{"train_loss": 0.16739021241664886, "global_step": 90707, "epoch": 1019, "lr": 7.505453804849277e-05}
{"train_loss": 0.23785816133022308, "global_step": 90708, "epoch": 1019, "lr": 7.505403634580377e-05}
{"train_loss": 0.19636885821819305, "global_step": 90709, "epoch": 1019, "lr": 7.505353463974657e-05}
{"train_loss": 0.1203543022274971, "global_step": 90710, "epoch": 1019, "lr": 7.505303293032122e-05}
{"train_loss": 0.14636653661727905, "global_step": 90711, "epoch": 1019, "lr": 7.50525312175278e-05}
{"train_loss": 0.16380909085273743, "global_step": 90712, "epoch": 1019, "lr": 7.505202950136636e-05}
{"train_loss": 0.167815700173378, "global_step": 90713, "epoch": 1019, "lr": 7.505152778183699e-05}
{"train_loss": 0.1390443593263626, "global_step": 90714, "epoch": 1019, "lr": 7.505102605893974e-05}
{"train_loss": 0.16972053050994873, "global_step": 90715, "epoch": 1019, "lr": 7.50505243326747e-05}
{"train_loss": 0.1462220698595047, "global_step": 90716, "epoch": 1019, "lr": 7.50500226030419e-05}
{"train_loss": 0.16294071078300476, "global_step": 90717, "epoch": 1019, "lr": 7.504952087004144e-05}
{"train_loss": 0.19297583401203156, "global_step": 90718, "epoch": 1019, "lr": 7.50490191336734e-05}
{"train_loss": 0.18755657970905304, "global_step": 90719, "epoch": 1019, "lr": 7.504851739393778e-05}
{"train_loss": 0.17322760820388794, "global_step": 90720, "epoch": 1019, "lr": 7.504801565083473e-05}
{"train_loss": 0.15862539410591125, "global_step": 90721, "epoch": 1019, "lr": 7.504751390436427e-05}
{"train_loss": 0.20059789717197418, "global_step": 90722, "epoch": 1019, "lr": 7.504701215452648e-05}
{"train_loss": 0.21869836747646332, "global_step": 90723, "epoch": 1019, "lr": 7.504651040132141e-05}
{"train_loss": 0.1505189836025238, "global_step": 90724, "epoch": 1019, "lr": 7.504600864474915e-05}
{"train_loss": 0.11077526211738586, "global_step": 90725, "epoch": 1019, "lr": 7.504550688480976e-05}
{"train_loss": 0.10935080051422119, "global_step": 90726, "epoch": 1019, "lr": 7.50450051215033e-05}
{"train_loss": 0.14180387556552887, "global_step": 90727, "epoch": 1019, "lr": 7.504450335482986e-05}
{"train_loss": 0.1635938286781311, "global_step": 90728, "epoch": 1019, "lr": 7.504400158478948e-05}
{"train_loss": 0.0942603051662445, "global_step": 90729, "epoch": 1019, "lr": 7.504349981138223e-05}
{"train_loss": 0.15254946053028107, "global_step": 90730, "epoch": 1019, "lr": 7.504299803460821e-05}
{"train_loss": 0.14205658435821533, "global_step": 90731, "epoch": 1019, "lr": 7.504249625446744e-05}
{"train_loss": 0.15152320265769958, "global_step": 90732, "epoch": 1019, "lr": 7.504199447096004e-05}
{"train_loss": 0.16743220388889313, "global_step": 90733, "epoch": 1019, "lr": 7.504149268408602e-05}
{"train_loss": 0.17902465164661407, "global_step": 90734, "epoch": 1019, "lr": 7.504099089384549e-05}
{"train_loss": 0.17069070041179657, "global_step": 90735, "epoch": 1019, "lr": 7.504048910023851e-05}
{"train_loss": 0.11486618220806122, "global_step": 90736, "epoch": 1019, "lr": 7.503998730326514e-05}
{"train_loss": 0.1367524117231369, "global_step": 90737, "epoch": 1019, "lr": 7.503948550292543e-05}
{"train_loss": 0.21400433778762817, "global_step": 90738, "epoch": 1019, "lr": 7.503898369921949e-05}
{"train_loss": 0.17393141984939575, "global_step": 90739, "epoch": 1019, "lr": 7.503848189214736e-05}
{"train_loss": 0.2824972867965698, "global_step": 90740, "epoch": 1019, "lr": 7.50379800817091e-05}
{"train_loss": 0.1014135330915451, "global_step": 90741, "epoch": 1019, "lr": 7.503747826790479e-05}
{"train_loss": 0.1400146782398224, "global_step": 90742, "epoch": 1019, "lr": 7.503697645073451e-05}
{"train_loss": 0.1841210424900055, "global_step": 90743, "epoch": 1019, "lr": 7.50364746301983e-05}
{"train_loss": 0.1420954167842865, "global_step": 90744, "epoch": 1019, "lr": 7.503597280629624e-05}
{"train_loss": 0.1659257560968399, "global_step": 90745, "epoch": 1019, "lr": 7.503547097902842e-05}
{"train_loss": 0.16224202513694763, "global_step": 90746, "epoch": 1019, "lr": 7.503496914839485e-05}
{"train_loss": 0.251353919506073, "global_step": 90747, "epoch": 1019, "lr": 7.503446731439565e-05}
{"train_loss": 0.15461483597755432, "global_step": 90748, "epoch": 1019, "lr": 7.503396547703088e-05}
{"train_loss": 0.09883556514978409, "global_step": 90749, "epoch": 1019, "lr": 7.503346363630059e-05}
{"train_loss": 0.2800889313220978, "global_step": 90750, "epoch": 1019, "lr": 7.503296179220488e-05}
{"train_loss": 0.1529732048511505, "global_step": 90751, "epoch": 1019, "lr": 7.503245994474375e-05}
{"train_loss": 0.10390102863311768, "global_step": 90752, "epoch": 1019, "lr": 7.503195809391735e-05}
{"train_loss": 0.1499270647764206, "global_step": 90753, "epoch": 1019, "lr": 7.503145623972567e-05}
{"train_loss": 0.12344083935022354, "global_step": 90754, "epoch": 1019, "lr": 7.503095438216885e-05}
{"train_loss": 0.14433062076568604, "global_step": 90755, "epoch": 1019, "lr": 7.50304525212469e-05}
{"train_loss": 0.21388350427150726, "global_step": 90756, "epoch": 1019, "lr": 7.502995065695992e-05}
{"train_loss": 0.2431371957063675, "global_step": 90757, "epoch": 1019, "lr": 7.502944878930798e-05}
{"train_loss": 0.22541570663452148, "global_step": 90758, "epoch": 1019, "lr": 7.502894691829112e-05}
{"train_loss": 0.14464999735355377, "global_step": 90759, "epoch": 1019, "lr": 7.502844504390943e-05}
{"train_loss": 0.21871499717235565, "global_step": 90760, "epoch": 1019, "lr": 7.502794316616297e-05}
{"train_loss": 0.2885266840457916, "global_step": 90761, "epoch": 1019, "lr": 7.50274412850518e-05}
{"train_loss": 0.19352208077907562, "global_step": 90762, "epoch": 1019, "lr": 7.502693940057601e-05}
{"train_loss": 0.23763906955718994, "global_step": 90763, "epoch": 1019, "lr": 7.502643751273563e-05}
{"train_loss": 0.10935113579034805, "global_step": 90764, "epoch": 1019, "lr": 7.502593562153078e-05}
{"train_loss": 0.14319530129432678, "global_step": 90765, "epoch": 1019, "lr": 7.502543372696149e-05}
{"train_loss": 0.23534905910491943, "global_step": 90766, "epoch": 1019, "lr": 7.502493182902782e-05}
{"train_loss": 0.22441250085830688, "global_step": 90767, "epoch": 1019, "lr": 7.502442992772987e-05}
{"train_loss": 0.1488272100687027, "global_step": 90768, "epoch": 1019, "lr": 7.502392802306768e-05}
{"train_loss": 0.12656761705875397, "global_step": 90769, "epoch": 1019, "lr": 7.502342611504133e-05}
{"train_loss": 0.16099078953266144, "global_step": 90770, "epoch": 1019, "lr": 7.50229242036509e-05}
{"train_loss": 0.22530010342597961, "global_step": 90771, "epoch": 1019, "lr": 7.50224222888964e-05}
{"train_loss": 0.2507254481315613, "global_step": 90772, "epoch": 1019, "lr": 7.502192037077798e-05}
{"train_loss": 0.17351454496383667, "global_step": 90773, "epoch": 1019, "lr": 7.502141844929565e-05}
{"train_loss": 0.1696050763130188, "global_step": 90774, "epoch": 1019, "lr": 7.502091652444951e-05}
{"train_loss": 0.15767934918403625, "global_step": 90775, "epoch": 1019, "lr": 7.50204145962396e-05}
{"train_loss": 0.2216586172580719, "global_step": 90776, "epoch": 1019, "lr": 7.501991266466602e-05}
{"train_loss": 0.13770833611488342, "global_step": 90777, "epoch": 1019, "lr": 7.501941072972879e-05}
{"train_loss": 0.137200266122818, "global_step": 90778, "epoch": 1019, "lr": 7.501890879142801e-05}
{"train_loss": 0.17307494019859293, "global_step": 90779, "epoch": 1019, "lr": 7.501840684976375e-05, "val_loss": 4.48989200592041}
{"train_loss": 0.13353128731250763, "global_step": 90780, "epoch": 1020, "lr": 7.501790490473608e-05}
{"train_loss": 0.16818885505199432, "global_step": 90781, "epoch": 1020, "lr": 7.501740295634505e-05}
{"train_loss": 0.19052810966968536, "global_step": 90782, "epoch": 1020, "lr": 7.501690100459074e-05}
{"train_loss": 0.158133864402771, "global_step": 90783, "epoch": 1020, "lr": 7.50163990494732e-05}
{"train_loss": 0.1722368746995926, "global_step": 90784, "epoch": 1020, "lr": 7.501589709099251e-05}
{"train_loss": 0.11663440614938736, "global_step": 90785, "epoch": 1020, "lr": 7.501539512914875e-05}
{"train_loss": 0.22680333256721497, "global_step": 90786, "epoch": 1020, "lr": 7.501489316394197e-05}
{"train_loss": 0.24364009499549866, "global_step": 90787, "epoch": 1020, "lr": 7.501439119537224e-05}
{"train_loss": 0.15664340555667877, "global_step": 90788, "epoch": 1020, "lr": 7.501388922343963e-05}
{"train_loss": 0.18403857946395874, "global_step": 90789, "epoch": 1020, "lr": 7.501338724814419e-05}
{"train_loss": 0.19899322092533112, "global_step": 90790, "epoch": 1020, "lr": 7.501288526948602e-05}
{"train_loss": 0.1365751475095749, "global_step": 90791, "epoch": 1020, "lr": 7.501238328746518e-05}
{"train_loss": 0.12141714990139008, "global_step": 90792, "epoch": 1020, "lr": 7.501188130208173e-05}
{"train_loss": 0.13475842773914337, "global_step": 90793, "epoch": 1020, "lr": 7.501137931333575e-05}
{"train_loss": 0.19440947473049164, "global_step": 90794, "epoch": 1020, "lr": 7.501087732122726e-05}
{"train_loss": 0.10555515438318253, "global_step": 90795, "epoch": 1020, "lr": 7.501037532575638e-05}
{"train_loss": 0.22869905829429626, "global_step": 90796, "epoch": 1020, "lr": 7.500987332692318e-05}
{"train_loss": 0.14460188150405884, "global_step": 90797, "epoch": 1020, "lr": 7.500937132472768e-05}
{"train_loss": 0.29565250873565674, "global_step": 90798, "epoch": 1020, "lr": 7.500886931916998e-05}
{"train_loss": 0.14309997856616974, "global_step": 90799, "epoch": 1020, "lr": 7.500836731025015e-05}
{"train_loss": 0.18353505432605743, "global_step": 90800, "epoch": 1020, "lr": 7.500786529796824e-05}
{"train_loss": 0.15653835237026215, "global_step": 90801, "epoch": 1020, "lr": 7.500736328232434e-05}
{"train_loss": 0.1890661120414734, "global_step": 90802, "epoch": 1020, "lr": 7.500686126331848e-05}
{"train_loss": 0.1920951008796692, "global_step": 90803, "epoch": 1020, "lr": 7.500635924095078e-05}
{"train_loss": 0.11458157747983932, "global_step": 90804, "epoch": 1020, "lr": 7.500585721522127e-05}
{"train_loss": 0.18321073055267334, "global_step": 90805, "epoch": 1020, "lr": 7.500535518613004e-05}
{"train_loss": 0.15607301890850067, "global_step": 90806, "epoch": 1020, "lr": 7.500485315367712e-05}
{"train_loss": 0.09774598479270935, "global_step": 90807, "epoch": 1020, "lr": 7.50043511178626e-05}
{"train_loss": 0.13317625224590302, "global_step": 90808, "epoch": 1020, "lr": 7.500384907868656e-05}
{"train_loss": 0.14278548955917358, "global_step": 90809, "epoch": 1020, "lr": 7.500334703614906e-05}
{"train_loss": 0.17676366865634918, "global_step": 90810, "epoch": 1020, "lr": 7.500284499025016e-05}
{"train_loss": 0.13361571729183197, "global_step": 90811, "epoch": 1020, "lr": 7.500234294098995e-05}
{"train_loss": 0.10694940388202667, "global_step": 90812, "epoch": 1020, "lr": 7.500184088836848e-05}
{"train_loss": 0.14443878829479218, "global_step": 90813, "epoch": 1020, "lr": 7.500133883238581e-05}
{"train_loss": 0.11547813564538956, "global_step": 90814, "epoch": 1020, "lr": 7.500083677304199e-05}
{"train_loss": 0.22231633961200714, "global_step": 90815, "epoch": 1020, "lr": 7.500033471033714e-05}
{"train_loss": 0.1342802792787552, "global_step": 90816, "epoch": 1020, "lr": 7.499983264427129e-05}
{"train_loss": 0.2163516879081726, "global_step": 90817, "epoch": 1020, "lr": 7.499933057484452e-05}
{"train_loss": 0.19213618338108063, "global_step": 90818, "epoch": 1020, "lr": 7.499882850205688e-05}
{"train_loss": 0.2163306027650833, "global_step": 90819, "epoch": 1020, "lr": 7.499832642590847e-05}
{"train_loss": 0.2121366411447525, "global_step": 90820, "epoch": 1020, "lr": 7.499782434639934e-05}
{"train_loss": 0.14824749529361725, "global_step": 90821, "epoch": 1020, "lr": 7.499732226352954e-05}
{"train_loss": 0.13996116816997528, "global_step": 90822, "epoch": 1020, "lr": 7.499682017729918e-05}
{"train_loss": 0.1517452895641327, "global_step": 90823, "epoch": 1020, "lr": 7.499631808770828e-05}
{"train_loss": 0.16100561618804932, "global_step": 90824, "epoch": 1020, "lr": 7.499581599475694e-05}
{"train_loss": 0.1828705072402954, "global_step": 90825, "epoch": 1020, "lr": 7.499531389844521e-05}
{"train_loss": 0.12444138526916504, "global_step": 90826, "epoch": 1020, "lr": 7.499481179877316e-05}
{"train_loss": 0.17744019627571106, "global_step": 90827, "epoch": 1020, "lr": 7.499430969574089e-05}
{"train_loss": 0.14796985685825348, "global_step": 90828, "epoch": 1020, "lr": 7.499380758934841e-05}
{"train_loss": 0.1619413048028946, "global_step": 90829, "epoch": 1020, "lr": 7.499330547959583e-05}
{"train_loss": 0.14909885823726654, "global_step": 90830, "epoch": 1020, "lr": 7.49928033664832e-05}
{"train_loss": 0.15996888279914856, "global_step": 90831, "epoch": 1020, "lr": 7.499230125001062e-05}
{"train_loss": 0.24599376320838928, "global_step": 90832, "epoch": 1020, "lr": 7.49917991301781e-05}
{"train_loss": 0.15543709695339203, "global_step": 90833, "epoch": 1020, "lr": 7.499129700698574e-05}
{"train_loss": 0.18380334973335266, "global_step": 90834, "epoch": 1020, "lr": 7.499079488043361e-05}
{"train_loss": 0.1920040100812912, "global_step": 90835, "epoch": 1020, "lr": 7.499029275052177e-05}
{"train_loss": 0.2155677080154419, "global_step": 90836, "epoch": 1020, "lr": 7.49897906172503e-05}
{"train_loss": 0.10852920264005661, "global_step": 90837, "epoch": 1020, "lr": 7.498928848061923e-05}
{"train_loss": 0.16379939019680023, "global_step": 90838, "epoch": 1020, "lr": 7.498878634062869e-05}
{"train_loss": 0.21457970142364502, "global_step": 90839, "epoch": 1020, "lr": 7.498828419727867e-05}
{"train_loss": 0.1486464887857437, "global_step": 90840, "epoch": 1020, "lr": 7.498778205056932e-05}
{"train_loss": 0.17283941805362701, "global_step": 90841, "epoch": 1020, "lr": 7.498727990050065e-05}
{"train_loss": 0.15211893618106842, "global_step": 90842, "epoch": 1020, "lr": 7.498677774707275e-05}
{"train_loss": 0.18378718197345734, "global_step": 90843, "epoch": 1020, "lr": 7.498627559028566e-05}
{"train_loss": 0.2124568223953247, "global_step": 90844, "epoch": 1020, "lr": 7.498577343013949e-05}
{"train_loss": 0.18341517448425293, "global_step": 90845, "epoch": 1020, "lr": 7.498527126663428e-05}
{"train_loss": 0.21351327002048492, "global_step": 90846, "epoch": 1020, "lr": 7.498476909977012e-05}
{"train_loss": 0.16182728111743927, "global_step": 90847, "epoch": 1020, "lr": 7.498426692954704e-05}
{"train_loss": 0.22252874076366425, "global_step": 90848, "epoch": 1020, "lr": 7.498376475596515e-05}
{"train_loss": 0.19478197395801544, "global_step": 90849, "epoch": 1020, "lr": 7.498326257902448e-05}
{"train_loss": 0.10686475038528442, "global_step": 90850, "epoch": 1020, "lr": 7.498276039872514e-05}
{"train_loss": 0.1497240662574768, "global_step": 90851, "epoch": 1020, "lr": 7.498225821506716e-05}
{"train_loss": 0.24716362357139587, "global_step": 90852, "epoch": 1020, "lr": 7.498175602805062e-05}
{"train_loss": 0.17778380215168, "global_step": 90853, "epoch": 1020, "lr": 7.498125383767558e-05}
{"train_loss": 0.20012781023979187, "global_step": 90854, "epoch": 1020, "lr": 7.498075164394212e-05}
{"train_loss": 0.13407161831855774, "global_step": 90855, "epoch": 1020, "lr": 7.498024944685029e-05}
{"train_loss": 0.13886277377605438, "global_step": 90856, "epoch": 1020, "lr": 7.497974724640019e-05}
{"train_loss": 0.19984938204288483, "global_step": 90857, "epoch": 1020, "lr": 7.497924504259186e-05}
{"train_loss": 0.14173483848571777, "global_step": 90858, "epoch": 1020, "lr": 7.497874283542537e-05}
{"train_loss": 0.20076176524162292, "global_step": 90859, "epoch": 1020, "lr": 7.49782406249008e-05}
{"train_loss": 0.24759089946746826, "global_step": 90860, "epoch": 1020, "lr": 7.49777384110182e-05}
{"train_loss": 0.277872771024704, "global_step": 90861, "epoch": 1020, "lr": 7.497723619377765e-05}
{"train_loss": 0.17947688698768616, "global_step": 90862, "epoch": 1020, "lr": 7.497673397317921e-05}
{"train_loss": 0.1947382241487503, "global_step": 90863, "epoch": 1020, "lr": 7.497623174922297e-05}
{"train_loss": 0.19618409872055054, "global_step": 90864, "epoch": 1020, "lr": 7.497572952190897e-05}
{"train_loss": 0.18183763325214386, "global_step": 90865, "epoch": 1020, "lr": 7.49752272912373e-05}
{"train_loss": 0.1252463310956955, "global_step": 90866, "epoch": 1020, "lr": 7.497472505720799e-05}
{"train_loss": 0.23341846466064453, "global_step": 90867, "epoch": 1020, "lr": 7.497422281982115e-05}
{"train_loss": 0.17441110204109986, "global_step": 90868, "epoch": 1020, "lr": 7.497372057907682e-05, "val_loss": 4.576228618621826, "train_action_mse_error": 25.273487091064453}
{"train_loss": 0.1389922797679901, "global_step": 90869, "epoch": 1021, "lr": 7.497321833497509e-05}
{"train_loss": 0.18422378599643707, "global_step": 90870, "epoch": 1021, "lr": 7.497271608751601e-05}
{"train_loss": 0.14178073406219482, "global_step": 90871, "epoch": 1021, "lr": 7.497221383669966e-05}
{"train_loss": 0.18318884074687958, "global_step": 90872, "epoch": 1021, "lr": 7.49717115825261e-05}
{"train_loss": 0.16435867547988892, "global_step": 90873, "epoch": 1021, "lr": 7.497120932499538e-05}
{"train_loss": 0.242540642619133, "global_step": 90874, "epoch": 1021, "lr": 7.49707070641076e-05}
{"train_loss": 0.17898571491241455, "global_step": 90875, "epoch": 1021, "lr": 7.497020479986281e-05}
{"train_loss": 0.18352113664150238, "global_step": 90876, "epoch": 1021, "lr": 7.496970253226108e-05}
{"train_loss": 0.3430967628955841, "global_step": 90877, "epoch": 1021, "lr": 7.496920026130248e-05}
{"train_loss": 0.14408574998378754, "global_step": 90878, "epoch": 1021, "lr": 7.496869798698707e-05}
{"train_loss": 0.2324485331773758, "global_step": 90879, "epoch": 1021, "lr": 7.496819570931492e-05}
{"train_loss": 0.126765638589859, "global_step": 90880, "epoch": 1021, "lr": 7.49676934282861e-05}
{"train_loss": 0.17846837639808655, "global_step": 90881, "epoch": 1021, "lr": 7.496719114390068e-05}
{"train_loss": 0.24094225466251373, "global_step": 90882, "epoch": 1021, "lr": 7.496668885615872e-05}
{"train_loss": 0.18816867470741272, "global_step": 90883, "epoch": 1021, "lr": 7.49661865650603e-05}
{"train_loss": 0.2351718544960022, "global_step": 90884, "epoch": 1021, "lr": 7.496568427060548e-05}
{"train_loss": 0.16109426319599152, "global_step": 90885, "epoch": 1021, "lr": 7.496518197279433e-05}
{"train_loss": 0.17627038061618805, "global_step": 90886, "epoch": 1021, "lr": 7.496467967162693e-05}
{"train_loss": 0.15808451175689697, "global_step": 90887, "epoch": 1021, "lr": 7.49641773671033e-05}
{"train_loss": 0.20441751182079315, "global_step": 90888, "epoch": 1021, "lr": 7.496367505922357e-05}
{"train_loss": 0.20503628253936768, "global_step": 90889, "epoch": 1021, "lr": 7.496317274798777e-05}
{"train_loss": 0.19060544669628143, "global_step": 90890, "epoch": 1021, "lr": 7.496267043339598e-05}
{"train_loss": 0.22924797236919403, "global_step": 90891, "epoch": 1021, "lr": 7.496216811544824e-05}
{"train_loss": 0.2448701709508896, "global_step": 90892, "epoch": 1021, "lr": 7.496166579414467e-05}
{"train_loss": 0.27420300245285034, "global_step": 90893, "epoch": 1021, "lr": 7.496116346948529e-05}
{"train_loss": 0.23437723517417908, "global_step": 90894, "epoch": 1021, "lr": 7.49606611414702e-05}
{"train_loss": 0.23465116322040558, "global_step": 90895, "epoch": 1021, "lr": 7.496015881009944e-05}
{"train_loss": 0.2196255773305893, "global_step": 90896, "epoch": 1021, "lr": 7.49596564753731e-05}
{"train_loss": 0.14423473179340363, "global_step": 90897, "epoch": 1021, "lr": 7.495915413729123e-05}
{"train_loss": 0.2208605408668518, "global_step": 90898, "epoch": 1021, "lr": 7.495865179585391e-05}
{"train_loss": 0.19018563628196716, "global_step": 90899, "epoch": 1021, "lr": 7.49581494510612e-05}
{"train_loss": 0.19271911680698395, "global_step": 90900, "epoch": 1021, "lr": 7.495764710291317e-05}
{"train_loss": 0.15865233540534973, "global_step": 90901, "epoch": 1021, "lr": 7.49571447514099e-05}
{"train_loss": 0.18495279550552368, "global_step": 90902, "epoch": 1021, "lr": 7.495664239655143e-05}
{"train_loss": 0.2239023745059967, "global_step": 90903, "epoch": 1021, "lr": 7.495614003833785e-05}
{"train_loss": 0.23695021867752075, "global_step": 90904, "epoch": 1021, "lr": 7.495563767676923e-05}
{"train_loss": 0.16456346213817596, "global_step": 90905, "epoch": 1021, "lr": 7.495513531184562e-05}
{"train_loss": 0.24977828562259674, "global_step": 90906, "epoch": 1021, "lr": 7.495463294356709e-05}
{"train_loss": 0.19860225915908813, "global_step": 90907, "epoch": 1021, "lr": 7.495413057193372e-05}
{"train_loss": 0.1742243617773056, "global_step": 90908, "epoch": 1021, "lr": 7.495362819694558e-05}
{"train_loss": 0.1653030663728714, "global_step": 90909, "epoch": 1021, "lr": 7.495312581860273e-05}
{"train_loss": 0.21007445454597473, "global_step": 90910, "epoch": 1021, "lr": 7.495262343690523e-05}
{"train_loss": 0.23146949708461761, "global_step": 90911, "epoch": 1021, "lr": 7.495212105185315e-05}
{"train_loss": 0.16614675521850586, "global_step": 90912, "epoch": 1021, "lr": 7.495161866344657e-05}
{"train_loss": 0.19449883699417114, "global_step": 90913, "epoch": 1021, "lr": 7.495111627168554e-05}
{"train_loss": 0.15859287977218628, "global_step": 90914, "epoch": 1021, "lr": 7.495061387657015e-05}
{"train_loss": 0.15985873341560364, "global_step": 90915, "epoch": 1021, "lr": 7.495011147810044e-05}
{"train_loss": 0.16813038289546967, "global_step": 90916, "epoch": 1021, "lr": 7.494960907627648e-05}
{"train_loss": 0.21146564185619354, "global_step": 90917, "epoch": 1021, "lr": 7.494910667109836e-05}
{"train_loss": 0.12706254422664642, "global_step": 90918, "epoch": 1021, "lr": 7.494860426256615e-05}
{"train_loss": 0.1802978813648224, "global_step": 90919, "epoch": 1021, "lr": 7.494810185067988e-05}
{"train_loss": 0.2025124579668045, "global_step": 90920, "epoch": 1021, "lr": 7.494759943543968e-05}
{"train_loss": 0.21267297863960266, "global_step": 90921, "epoch": 1021, "lr": 7.494709701684555e-05}
{"train_loss": 0.21174755692481995, "global_step": 90922, "epoch": 1021, "lr": 7.49465945948976e-05}
{"train_loss": 0.28035128116607666, "global_step": 90923, "epoch": 1021, "lr": 7.494609216959587e-05}
{"train_loss": 0.28937074542045593, "global_step": 90924, "epoch": 1021, "lr": 7.494558974094044e-05}
{"train_loss": 0.2611277401447296, "global_step": 90925, "epoch": 1021, "lr": 7.49450873089314e-05}
{"train_loss": 0.2524520456790924, "global_step": 90926, "epoch": 1021, "lr": 7.494458487356879e-05}
{"train_loss": 0.17940233647823334, "global_step": 90927, "epoch": 1021, "lr": 7.494408243485266e-05}
{"train_loss": 0.2512489855289459, "global_step": 90928, "epoch": 1021, "lr": 7.494357999278315e-05}
{"train_loss": 0.1524081975221634, "global_step": 90929, "epoch": 1021, "lr": 7.494307754736025e-05}
{"train_loss": 0.1700432002544403, "global_step": 90930, "epoch": 1021, "lr": 7.494257509858406e-05}
{"train_loss": 0.19234706461429596, "global_step": 90931, "epoch": 1021, "lr": 7.494207264645464e-05}
{"train_loss": 0.24732521176338196, "global_step": 90932, "epoch": 1021, "lr": 7.494157019097206e-05}
{"train_loss": 0.19075731933116913, "global_step": 90933, "epoch": 1021, "lr": 7.49410677321364e-05}
{"train_loss": 0.24480341374874115, "global_step": 90934, "epoch": 1021, "lr": 7.494056526994771e-05}
{"train_loss": 0.2064952552318573, "global_step": 90935, "epoch": 1021, "lr": 7.494006280440607e-05}
{"train_loss": 0.1748208850622177, "global_step": 90936, "epoch": 1021, "lr": 7.493956033551153e-05}
{"train_loss": 0.15404056012630463, "global_step": 90937, "epoch": 1021, "lr": 7.49390578632642e-05}
{"train_loss": 0.20322321355342865, "global_step": 90938, "epoch": 1021, "lr": 7.493855538766409e-05}
{"train_loss": 0.20947106182575226, "global_step": 90939, "epoch": 1021, "lr": 7.493805290871131e-05}
{"train_loss": 0.18503572046756744, "global_step": 90940, "epoch": 1021, "lr": 7.49375504264059e-05}
{"train_loss": 0.325219064950943, "global_step": 90941, "epoch": 1021, "lr": 7.493704794074795e-05}
{"train_loss": 0.16153456270694733, "global_step": 90942, "epoch": 1021, "lr": 7.493654545173752e-05}
{"train_loss": 0.16512173414230347, "global_step": 90943, "epoch": 1021, "lr": 7.493604295937467e-05}
{"train_loss": 0.15608614683151245, "global_step": 90944, "epoch": 1021, "lr": 7.493554046365948e-05}
{"train_loss": 0.1321394145488739, "global_step": 90945, "epoch": 1021, "lr": 7.493503796459199e-05}
{"train_loss": 0.2402605414390564, "global_step": 90946, "epoch": 1021, "lr": 7.493453546217231e-05}
{"train_loss": 0.21685223281383514, "global_step": 90947, "epoch": 1021, "lr": 7.493403295640049e-05}
{"train_loss": 0.18734531104564667, "global_step": 90948, "epoch": 1021, "lr": 7.493353044727658e-05}
{"train_loss": 0.14558614790439606, "global_step": 90949, "epoch": 1021, "lr": 7.493302793480065e-05}
{"train_loss": 0.13980500400066376, "global_step": 90950, "epoch": 1021, "lr": 7.49325254189728e-05}
{"train_loss": 0.12804698944091797, "global_step": 90951, "epoch": 1021, "lr": 7.493202289979306e-05}
{"train_loss": 0.2516840696334839, "global_step": 90952, "epoch": 1021, "lr": 7.493152037726151e-05}
{"train_loss": 0.261756032705307, "global_step": 90953, "epoch": 1021, "lr": 7.493101785137824e-05}
{"train_loss": 0.26169484853744507, "global_step": 90954, "epoch": 1021, "lr": 7.493051532214328e-05}
{"train_loss": 0.16598449647426605, "global_step": 90955, "epoch": 1021, "lr": 7.493001278955673e-05}
{"train_loss": 0.1577087938785553, "global_step": 90956, "epoch": 1021, "lr": 7.492951025361864e-05}
{"train_loss": 0.20017870456984874, "global_step": 90957, "epoch": 1021, "lr": 7.492900771432908e-05, "val_loss": 4.373863220214844}
{"train_loss": 0.21589022874832153, "global_step": 90958, "epoch": 1022, "lr": 7.492850517168812e-05}
{"train_loss": 0.17036330699920654, "global_step": 90959, "epoch": 1022, "lr": 7.492800262569582e-05}
{"train_loss": 0.1933344304561615, "global_step": 90960, "epoch": 1022, "lr": 7.492750007635226e-05}
{"train_loss": 0.14691299200057983, "global_step": 90961, "epoch": 1022, "lr": 7.49269975236575e-05}
{"train_loss": 0.15253973007202148, "global_step": 90962, "epoch": 1022, "lr": 7.492649496761161e-05}
{"train_loss": 0.17020614445209503, "global_step": 90963, "epoch": 1022, "lr": 7.492599240821465e-05}
{"train_loss": 0.15964756906032562, "global_step": 90964, "epoch": 1022, "lr": 7.49254898454667e-05}
{"train_loss": 0.23116770386695862, "global_step": 90965, "epoch": 1022, "lr": 7.492498727936783e-05}
{"train_loss": 0.24170352518558502, "global_step": 90966, "epoch": 1022, "lr": 7.492448470991808e-05}
{"train_loss": 0.30991265177726746, "global_step": 90967, "epoch": 1022, "lr": 7.492398213711757e-05}
{"train_loss": 0.22977033257484436, "global_step": 90968, "epoch": 1022, "lr": 7.49234795609663e-05}
{"train_loss": 0.13396209478378296, "global_step": 90969, "epoch": 1022, "lr": 7.492297698146438e-05}
{"train_loss": 0.10543182492256165, "global_step": 90970, "epoch": 1022, "lr": 7.492247439861188e-05}
{"train_loss": 0.20599579811096191, "global_step": 90971, "epoch": 1022, "lr": 7.492197181240885e-05}
{"train_loss": 0.24915711581707, "global_step": 90972, "epoch": 1022, "lr": 7.492146922285535e-05}
{"train_loss": 0.23144379258155823, "global_step": 90973, "epoch": 1022, "lr": 7.492096662995148e-05}
{"train_loss": 0.19201312959194183, "global_step": 90974, "epoch": 1022, "lr": 7.492046403369728e-05}
{"train_loss": 0.19270718097686768, "global_step": 90975, "epoch": 1022, "lr": 7.491996143409285e-05}
{"train_loss": 0.2244940847158432, "global_step": 90976, "epoch": 1022, "lr": 7.491945883113822e-05}
{"train_loss": 0.18527370691299438, "global_step": 90977, "epoch": 1022, "lr": 7.491895622483348e-05}
{"train_loss": 0.15354250371456146, "global_step": 90978, "epoch": 1022, "lr": 7.491845361517868e-05}
{"train_loss": 0.1401902288198471, "global_step": 90979, "epoch": 1022, "lr": 7.491795100217389e-05}
{"train_loss": 0.18974246084690094, "global_step": 90980, "epoch": 1022, "lr": 7.49174483858192e-05}
{"train_loss": 0.19234971702098846, "global_step": 90981, "epoch": 1022, "lr": 7.491694576611466e-05}
{"train_loss": 0.15971441566944122, "global_step": 90982, "epoch": 1022, "lr": 7.491644314306033e-05}
{"train_loss": 0.20498022437095642, "global_step": 90983, "epoch": 1022, "lr": 7.49159405166563e-05}
{"train_loss": 0.27037379145622253, "global_step": 90984, "epoch": 1022, "lr": 7.491543788690261e-05}
{"train_loss": 0.21997970342636108, "global_step": 90985, "epoch": 1022, "lr": 7.491493525379937e-05}
{"train_loss": 0.25762420892715454, "global_step": 90986, "epoch": 1022, "lr": 7.49144326173466e-05}
{"train_loss": 0.2717035412788391, "global_step": 90987, "epoch": 1022, "lr": 7.49139299775444e-05}
{"train_loss": 0.19753813743591309, "global_step": 90988, "epoch": 1022, "lr": 7.491342733439281e-05}
{"train_loss": 0.3082507848739624, "global_step": 90989, "epoch": 1022, "lr": 7.491292468789193e-05}
{"train_loss": 0.1471339762210846, "global_step": 90990, "epoch": 1022, "lr": 7.491242203804179e-05}
{"train_loss": 0.14313605427742004, "global_step": 90991, "epoch": 1022, "lr": 7.49119193848425e-05}
{"train_loss": 0.21433255076408386, "global_step": 90992, "epoch": 1022, "lr": 7.491141672829409e-05}
{"train_loss": 0.2170189917087555, "global_step": 90993, "epoch": 1022, "lr": 7.491091406839666e-05}
{"train_loss": 0.25660255551338196, "global_step": 90994, "epoch": 1022, "lr": 7.491041140515025e-05}
{"train_loss": 0.11518353223800659, "global_step": 90995, "epoch": 1022, "lr": 7.490990873855494e-05}
{"train_loss": 0.1896507292985916, "global_step": 90996, "epoch": 1022, "lr": 7.49094060686108e-05}
{"train_loss": 0.1906057745218277, "global_step": 90997, "epoch": 1022, "lr": 7.490890339531788e-05}
{"train_loss": 0.13213689625263214, "global_step": 90998, "epoch": 1022, "lr": 7.490840071867628e-05}
{"train_loss": 0.155544251203537, "global_step": 90999, "epoch": 1022, "lr": 7.490789803868605e-05}
{"train_loss": 0.2423890233039856, "global_step": 91000, "epoch": 1022, "lr": 7.490739535534725e-05}
{"train_loss": 0.1562838852405548, "global_step": 91001, "epoch": 1022, "lr": 7.490689266865996e-05}
{"train_loss": 0.16505269706249237, "global_step": 91002, "epoch": 1022, "lr": 7.490638997862423e-05}
{"train_loss": 0.14828146994113922, "global_step": 91003, "epoch": 1022, "lr": 7.490588728524014e-05}
{"train_loss": 0.14527961611747742, "global_step": 91004, "epoch": 1022, "lr": 7.490538458850777e-05}
{"train_loss": 0.12851910293102264, "global_step": 91005, "epoch": 1022, "lr": 7.490488188842715e-05}
{"train_loss": 0.13565614819526672, "global_step": 91006, "epoch": 1022, "lr": 7.490437918499841e-05}
{"train_loss": 0.2134356051683426, "global_step": 91007, "epoch": 1022, "lr": 7.490387647822154e-05}
{"train_loss": 0.21143393218517303, "global_step": 91008, "epoch": 1022, "lr": 7.490337376809667e-05}
{"train_loss": 0.2108619213104248, "global_step": 91009, "epoch": 1022, "lr": 7.490287105462384e-05}
{"train_loss": 0.14056192338466644, "global_step": 91010, "epoch": 1022, "lr": 7.490236833780312e-05}
{"train_loss": 0.16218027472496033, "global_step": 91011, "epoch": 1022, "lr": 7.49018656176346e-05}
{"train_loss": 0.24890251457691193, "global_step": 91012, "epoch": 1022, "lr": 7.49013628941183e-05}
{"train_loss": 0.256662517786026, "global_step": 91013, "epoch": 1022, "lr": 7.490086016725432e-05}
{"train_loss": 0.18105313181877136, "global_step": 91014, "epoch": 1022, "lr": 7.490035743704274e-05}
{"train_loss": 0.10275281965732574, "global_step": 91015, "epoch": 1022, "lr": 7.48998547034836e-05}
{"train_loss": 0.19694818556308746, "global_step": 91016, "epoch": 1022, "lr": 7.489935196657697e-05}
{"train_loss": 0.10071663558483124, "global_step": 91017, "epoch": 1022, "lr": 7.489884922632293e-05}
{"train_loss": 0.13475064933300018, "global_step": 91018, "epoch": 1022, "lr": 7.489834648272154e-05}
{"train_loss": 0.21776533126831055, "global_step": 91019, "epoch": 1022, "lr": 7.489784373577288e-05}
{"train_loss": 0.2552623748779297, "global_step": 91020, "epoch": 1022, "lr": 7.489734098547702e-05}
{"train_loss": 0.19492319226264954, "global_step": 91021, "epoch": 1022, "lr": 7.4896838231834e-05}
{"train_loss": 0.2046169638633728, "global_step": 91022, "epoch": 1022, "lr": 7.489633547484389e-05}
{"train_loss": 0.22286491096019745, "global_step": 91023, "epoch": 1022, "lr": 7.48958327145068e-05}
{"train_loss": 0.124957375228405, "global_step": 91024, "epoch": 1022, "lr": 7.489532995082276e-05}
{"train_loss": 0.2019624561071396, "global_step": 91025, "epoch": 1022, "lr": 7.489482718379183e-05}
{"train_loss": 0.15362650156021118, "global_step": 91026, "epoch": 1022, "lr": 7.48943244134141e-05}
{"train_loss": 0.22331219911575317, "global_step": 91027, "epoch": 1022, "lr": 7.489382163968964e-05}
{"train_loss": 0.2910657823085785, "global_step": 91028, "epoch": 1022, "lr": 7.489331886261851e-05}
{"train_loss": 0.13893790543079376, "global_step": 91029, "epoch": 1022, "lr": 7.489281608220077e-05}
{"train_loss": 0.15738821029663086, "global_step": 91030, "epoch": 1022, "lr": 7.48923132984365e-05}
{"train_loss": 0.13249005377292633, "global_step": 91031, "epoch": 1022, "lr": 7.489181051132575e-05}
{"train_loss": 0.12375518679618835, "global_step": 91032, "epoch": 1022, "lr": 7.489130772086861e-05}
{"train_loss": 0.12316292524337769, "global_step": 91033, "epoch": 1022, "lr": 7.489080492706514e-05}
{"train_loss": 0.20511649549007416, "global_step": 91034, "epoch": 1022, "lr": 7.48903021299154e-05}
{"train_loss": 0.17437586188316345, "global_step": 91035, "epoch": 1022, "lr": 7.488979932941946e-05}
{"train_loss": 0.15147928893566132, "global_step": 91036, "epoch": 1022, "lr": 7.488929652557739e-05}
{"train_loss": 0.2100011259317398, "global_step": 91037, "epoch": 1022, "lr": 7.488879371838926e-05}
{"train_loss": 0.18086642026901245, "global_step": 91038, "epoch": 1022, "lr": 7.488829090785514e-05}
{"train_loss": 0.12216487526893616, "global_step": 91039, "epoch": 1022, "lr": 7.488778809397508e-05}
{"train_loss": 0.18050874769687653, "global_step": 91040, "epoch": 1022, "lr": 7.488728527674918e-05}
{"train_loss": 0.19666194915771484, "global_step": 91041, "epoch": 1022, "lr": 7.488678245617746e-05}
{"train_loss": 0.22192908823490143, "global_step": 91042, "epoch": 1022, "lr": 7.488627963226004e-05}
{"train_loss": 0.20409643650054932, "global_step": 91043, "epoch": 1022, "lr": 7.488577680499694e-05}
{"train_loss": 0.13521645963191986, "global_step": 91044, "epoch": 1022, "lr": 7.488527397438827e-05}
{"train_loss": 0.13168443739414215, "global_step": 91045, "epoch": 1022, "lr": 7.488477114043407e-05}
{"train_loss": 0.18599202056949057, "global_step": 91046, "epoch": 1022, "lr": 7.488426830313441e-05, "val_loss": 4.234493255615234}
{"train_loss": 0.16468071937561035, "global_step": 91047, "epoch": 1023, "lr": 7.488376546248938e-05}
{"train_loss": 0.21434837579727173, "global_step": 91048, "epoch": 1023, "lr": 7.488326261849903e-05}
{"train_loss": 0.1715485006570816, "global_step": 91049, "epoch": 1023, "lr": 7.488275977116342e-05}
{"train_loss": 0.16531158983707428, "global_step": 91050, "epoch": 1023, "lr": 7.488225692048263e-05}
{"train_loss": 0.24459515511989594, "global_step": 91051, "epoch": 1023, "lr": 7.488175406645671e-05}
{"train_loss": 0.2279626578092575, "global_step": 91052, "epoch": 1023, "lr": 7.488125120908576e-05}
{"train_loss": 0.1999414712190628, "global_step": 91053, "epoch": 1023, "lr": 7.488074834836982e-05}
{"train_loss": 0.19681070744991302, "global_step": 91054, "epoch": 1023, "lr": 7.488024548430896e-05}
{"train_loss": 0.19524218142032623, "global_step": 91055, "epoch": 1023, "lr": 7.487974261690327e-05}
{"train_loss": 0.17670270800590515, "global_step": 91056, "epoch": 1023, "lr": 7.48792397461528e-05}
{"train_loss": 0.1790350079536438, "global_step": 91057, "epoch": 1023, "lr": 7.48787368720576e-05}
{"train_loss": 0.22396552562713623, "global_step": 91058, "epoch": 1023, "lr": 7.487823399461777e-05}
{"train_loss": 0.15317440032958984, "global_step": 91059, "epoch": 1023, "lr": 7.487773111383337e-05}
{"train_loss": 0.21522614359855652, "global_step": 91060, "epoch": 1023, "lr": 7.487722822970447e-05}
{"train_loss": 0.09345918148756027, "global_step": 91061, "epoch": 1023, "lr": 7.487672534223111e-05}
{"train_loss": 0.19017848372459412, "global_step": 91062, "epoch": 1023, "lr": 7.487622245141337e-05}
{"train_loss": 0.16189266741275787, "global_step": 91063, "epoch": 1023, "lr": 7.487571955725135e-05}
{"train_loss": 0.17831404507160187, "global_step": 91064, "epoch": 1023, "lr": 7.487521665974508e-05}
{"train_loss": 0.2398429811000824, "global_step": 91065, "epoch": 1023, "lr": 7.487471375889465e-05}
{"train_loss": 0.1579979658126831, "global_step": 91066, "epoch": 1023, "lr": 7.487421085470012e-05}
{"train_loss": 0.15409381687641144, "global_step": 91067, "epoch": 1023, "lr": 7.487370794716155e-05}
{"train_loss": 0.13176463544368744, "global_step": 91068, "epoch": 1023, "lr": 7.4873205036279e-05}
{"train_loss": 0.19640469551086426, "global_step": 91069, "epoch": 1023, "lr": 7.487270212205257e-05}
{"train_loss": 0.14301125705242157, "global_step": 91070, "epoch": 1023, "lr": 7.487219920448228e-05}
{"train_loss": 0.16642962396144867, "global_step": 91071, "epoch": 1023, "lr": 7.487169628356825e-05}
{"train_loss": 0.18098929524421692, "global_step": 91072, "epoch": 1023, "lr": 7.487119335931052e-05}
{"train_loss": 0.1752646565437317, "global_step": 91073, "epoch": 1023, "lr": 7.487069043170915e-05}
{"train_loss": 0.16504362225532532, "global_step": 91074, "epoch": 1023, "lr": 7.487018750076424e-05}
{"train_loss": 0.15831685066223145, "global_step": 91075, "epoch": 1023, "lr": 7.486968456647583e-05}
{"train_loss": 0.18230797350406647, "global_step": 91076, "epoch": 1023, "lr": 7.486918162884398e-05}
{"train_loss": 0.11962847411632538, "global_step": 91077, "epoch": 1023, "lr": 7.486867868786878e-05}
{"train_loss": 0.17397868633270264, "global_step": 91078, "epoch": 1023, "lr": 7.486817574355029e-05}
{"train_loss": 0.16087377071380615, "global_step": 91079, "epoch": 1023, "lr": 7.486767279588858e-05}
{"train_loss": 0.24253402650356293, "global_step": 91080, "epoch": 1023, "lr": 7.486716984488371e-05}
{"train_loss": 0.22116932272911072, "global_step": 91081, "epoch": 1023, "lr": 7.486666689053575e-05}
{"train_loss": 0.18260641396045685, "global_step": 91082, "epoch": 1023, "lr": 7.486616393284478e-05}
{"train_loss": 0.15872327983379364, "global_step": 91083, "epoch": 1023, "lr": 7.486566097181085e-05}
{"train_loss": 0.15629711747169495, "global_step": 91084, "epoch": 1023, "lr": 7.486515800743404e-05}
{"train_loss": 0.1588929444551468, "global_step": 91085, "epoch": 1023, "lr": 7.48646550397144e-05}
{"train_loss": 0.23197664320468903, "global_step": 91086, "epoch": 1023, "lr": 7.486415206865202e-05}
{"train_loss": 0.15281540155410767, "global_step": 91087, "epoch": 1023, "lr": 7.486364909424696e-05}
{"train_loss": 0.1803869903087616, "global_step": 91088, "epoch": 1023, "lr": 7.486314611649928e-05}
{"train_loss": 0.17873670160770416, "global_step": 91089, "epoch": 1023, "lr": 7.486264313540905e-05}
{"train_loss": 0.1843709796667099, "global_step": 91090, "epoch": 1023, "lr": 7.486214015097634e-05}
{"train_loss": 0.2119516283273697, "global_step": 91091, "epoch": 1023, "lr": 7.486163716320123e-05}
{"train_loss": 0.19151932001113892, "global_step": 91092, "epoch": 1023, "lr": 7.486113417208376e-05}
{"train_loss": 0.2572648525238037, "global_step": 91093, "epoch": 1023, "lr": 7.486063117762403e-05}
{"train_loss": 0.22921879589557648, "global_step": 91094, "epoch": 1023, "lr": 7.486012817982209e-05}
{"train_loss": 0.1820598989725113, "global_step": 91095, "epoch": 1023, "lr": 7.4859625178678e-05}
{"train_loss": 0.30754512548446655, "global_step": 91096, "epoch": 1023, "lr": 7.485912217419183e-05}
{"train_loss": 0.12704996764659882, "global_step": 91097, "epoch": 1023, "lr": 7.485861916636367e-05}
{"train_loss": 0.18521174788475037, "global_step": 91098, "epoch": 1023, "lr": 7.485811615519356e-05}
{"train_loss": 0.21059578657150269, "global_step": 91099, "epoch": 1023, "lr": 7.485761314068158e-05}
{"train_loss": 0.16400687396526337, "global_step": 91100, "epoch": 1023, "lr": 7.485711012282781e-05}
{"train_loss": 0.18830151855945587, "global_step": 91101, "epoch": 1023, "lr": 7.485660710163229e-05}
{"train_loss": 0.16759218275547028, "global_step": 91102, "epoch": 1023, "lr": 7.485610407709512e-05}
{"train_loss": 0.14348113536834717, "global_step": 91103, "epoch": 1023, "lr": 7.485560104921633e-05}
{"train_loss": 0.2149546593427658, "global_step": 91104, "epoch": 1023, "lr": 7.485509801799602e-05}
{"train_loss": 0.14482326805591583, "global_step": 91105, "epoch": 1023, "lr": 7.485459498343424e-05}
{"train_loss": 0.19326992332935333, "global_step": 91106, "epoch": 1023, "lr": 7.485409194553106e-05}
{"train_loss": 0.26001131534576416, "global_step": 91107, "epoch": 1023, "lr": 7.485358890428654e-05}
{"train_loss": 0.15030336380004883, "global_step": 91108, "epoch": 1023, "lr": 7.485308585970077e-05}
{"train_loss": 0.17989815771579742, "global_step": 91109, "epoch": 1023, "lr": 7.485258281177381e-05}
{"train_loss": 0.17815692722797394, "global_step": 91110, "epoch": 1023, "lr": 7.485207976050572e-05}
{"train_loss": 0.16069844365119934, "global_step": 91111, "epoch": 1023, "lr": 7.485157670589657e-05}
{"train_loss": 0.1918017417192459, "global_step": 91112, "epoch": 1023, "lr": 7.485107364794643e-05}
{"train_loss": 0.1099768579006195, "global_step": 91113, "epoch": 1023, "lr": 7.485057058665536e-05}
{"train_loss": 0.17365548014640808, "global_step": 91114, "epoch": 1023, "lr": 7.485006752202343e-05}
{"train_loss": 0.10381079465150833, "global_step": 91115, "epoch": 1023, "lr": 7.484956445405073e-05}
{"train_loss": 0.2217177301645279, "global_step": 91116, "epoch": 1023, "lr": 7.48490613827373e-05}
{"train_loss": 0.1580166071653366, "global_step": 91117, "epoch": 1023, "lr": 7.484855830808322e-05}
{"train_loss": 0.16219866275787354, "global_step": 91118, "epoch": 1023, "lr": 7.484805523008854e-05}
{"train_loss": 0.1192486509680748, "global_step": 91119, "epoch": 1023, "lr": 7.484755214875335e-05}
{"train_loss": 0.1378583312034607, "global_step": 91120, "epoch": 1023, "lr": 7.484704906407771e-05}
{"train_loss": 0.09852082282304764, "global_step": 91121, "epoch": 1023, "lr": 7.484654597606169e-05}
{"train_loss": 0.1409638524055481, "global_step": 91122, "epoch": 1023, "lr": 7.484604288470536e-05}
{"train_loss": 0.18304947018623352, "global_step": 91123, "epoch": 1023, "lr": 7.484553979000877e-05}
{"train_loss": 0.13126526772975922, "global_step": 91124, "epoch": 1023, "lr": 7.4845036691972e-05}
{"train_loss": 0.2062777429819107, "global_step": 91125, "epoch": 1023, "lr": 7.484453359059512e-05}
{"train_loss": 0.16323299705982208, "global_step": 91126, "epoch": 1023, "lr": 7.48440304858782e-05}
{"train_loss": 0.1636720597743988, "global_step": 91127, "epoch": 1023, "lr": 7.484352737782129e-05}
{"train_loss": 0.1870100051164627, "global_step": 91128, "epoch": 1023, "lr": 7.484302426642448e-05}
{"train_loss": 0.2166636735200882, "global_step": 91129, "epoch": 1023, "lr": 7.484252115168784e-05}
{"train_loss": 0.1356857866048813, "global_step": 91130, "epoch": 1023, "lr": 7.484201803361141e-05}
{"train_loss": 0.16637985408306122, "global_step": 91131, "epoch": 1023, "lr": 7.484151491219529e-05}
{"train_loss": 0.16980299353599548, "global_step": 91132, "epoch": 1023, "lr": 7.484101178743952e-05}
{"train_loss": 0.23072639107704163, "global_step": 91133, "epoch": 1023, "lr": 7.484050865934417e-05}
{"train_loss": 0.10814744979143143, "global_step": 91134, "epoch": 1023, "lr": 7.484000552790932e-05}
{"train_loss": 0.17793246927890885, "global_step": 91135, "epoch": 1023, "lr": 7.483950239313503e-05, "val_loss": 4.524713039398193}
{"train_loss": 0.1573195904493332, "global_step": 91136, "epoch": 1024, "lr": 7.48389992550214e-05}
{"train_loss": 0.16352710127830505, "global_step": 91137, "epoch": 1024, "lr": 7.483849611356844e-05}
{"train_loss": 0.19303469359874725, "global_step": 91138, "epoch": 1024, "lr": 7.483799296877626e-05}
{"train_loss": 0.2212478071451187, "global_step": 91139, "epoch": 1024, "lr": 7.483748982064492e-05}
{"train_loss": 0.14018240571022034, "global_step": 91140, "epoch": 1024, "lr": 7.483698666917446e-05}
{"train_loss": 0.15707911550998688, "global_step": 91141, "epoch": 1024, "lr": 7.483648351436498e-05}
{"train_loss": 0.1986599564552307, "global_step": 91142, "epoch": 1024, "lr": 7.483598035621655e-05}
{"train_loss": 0.18982267379760742, "global_step": 91143, "epoch": 1024, "lr": 7.48354771947292e-05}
{"train_loss": 0.18182648718357086, "global_step": 91144, "epoch": 1024, "lr": 7.483497402990305e-05}
{"train_loss": 0.16077618300914764, "global_step": 91145, "epoch": 1024, "lr": 7.483447086173813e-05}
{"train_loss": 0.1229453831911087, "global_step": 91146, "epoch": 1024, "lr": 7.483396769023452e-05}
{"train_loss": 0.18232190608978271, "global_step": 91147, "epoch": 1024, "lr": 7.483346451539228e-05}
{"train_loss": 0.1312294453382492, "global_step": 91148, "epoch": 1024, "lr": 7.483296133721149e-05}
{"train_loss": 0.23325151205062866, "global_step": 91149, "epoch": 1024, "lr": 7.48324581556922e-05}
{"train_loss": 0.21915382146835327, "global_step": 91150, "epoch": 1024, "lr": 7.48319549708345e-05}
{"train_loss": 0.1441177874803543, "global_step": 91151, "epoch": 1024, "lr": 7.483145178263844e-05}
{"train_loss": 0.13915453851222992, "global_step": 91152, "epoch": 1024, "lr": 7.483094859110409e-05}
{"train_loss": 0.0995803028345108, "global_step": 91153, "epoch": 1024, "lr": 7.483044539623152e-05}
{"train_loss": 0.23247763514518738, "global_step": 91154, "epoch": 1024, "lr": 7.482994219802081e-05}
{"train_loss": 0.32060572504997253, "global_step": 91155, "epoch": 1024, "lr": 7.482943899647203e-05}
{"train_loss": 0.14102481305599213, "global_step": 91156, "epoch": 1024, "lr": 7.482893579158521e-05}
{"train_loss": 0.19753628969192505, "global_step": 91157, "epoch": 1024, "lr": 7.482843258336046e-05}
{"train_loss": 0.20910495519638062, "global_step": 91158, "epoch": 1024, "lr": 7.482792937179782e-05}
{"train_loss": 0.1968851387500763, "global_step": 91159, "epoch": 1024, "lr": 7.482742615689736e-05}
{"train_loss": 0.17070463299751282, "global_step": 91160, "epoch": 1024, "lr": 7.482692293865916e-05}
{"train_loss": 0.13552764058113098, "global_step": 91161, "epoch": 1024, "lr": 7.482641971708329e-05}
{"train_loss": 0.18254783749580383, "global_step": 91162, "epoch": 1024, "lr": 7.482591649216981e-05}
{"train_loss": 0.17189638316631317, "global_step": 91163, "epoch": 1024, "lr": 7.482541326391879e-05}
{"train_loss": 0.19433888792991638, "global_step": 91164, "epoch": 1024, "lr": 7.482491003233028e-05}
{"train_loss": 0.17302392423152924, "global_step": 91165, "epoch": 1024, "lr": 7.482440679740438e-05}
{"train_loss": 0.12776221334934235, "global_step": 91166, "epoch": 1024, "lr": 7.482390355914113e-05}
{"train_loss": 0.2321341633796692, "global_step": 91167, "epoch": 1024, "lr": 7.482340031754062e-05}
{"train_loss": 0.17591196298599243, "global_step": 91168, "epoch": 1024, "lr": 7.48228970726029e-05}
{"train_loss": 0.22489969432353973, "global_step": 91169, "epoch": 1024, "lr": 7.482239382432804e-05}
{"train_loss": 0.17064383625984192, "global_step": 91170, "epoch": 1024, "lr": 7.48218905727161e-05}
{"train_loss": 0.22876180708408356, "global_step": 91171, "epoch": 1024, "lr": 7.482138731776718e-05}
{"train_loss": 0.2552546262741089, "global_step": 91172, "epoch": 1024, "lr": 7.482088405948132e-05}
{"train_loss": 0.2460595667362213, "global_step": 91173, "epoch": 1024, "lr": 7.48203807978586e-05}
{"train_loss": 0.12358949333429337, "global_step": 91174, "epoch": 1024, "lr": 7.481987753289909e-05}
{"train_loss": 0.2126888632774353, "global_step": 91175, "epoch": 1024, "lr": 7.481937426460283e-05}
{"train_loss": 0.15993686020374298, "global_step": 91176, "epoch": 1024, "lr": 7.481887099296991e-05}
{"train_loss": 0.193192720413208, "global_step": 91177, "epoch": 1024, "lr": 7.481836771800042e-05}
{"train_loss": 0.194181427359581, "global_step": 91178, "epoch": 1024, "lr": 7.481786443969437e-05}
{"train_loss": 0.2845788598060608, "global_step": 91179, "epoch": 1024, "lr": 7.481736115805188e-05}
{"train_loss": 0.14289243519306183, "global_step": 91180, "epoch": 1024, "lr": 7.4816857873073e-05}
{"train_loss": 0.21076886355876923, "global_step": 91181, "epoch": 1024, "lr": 7.481635458475779e-05}
{"train_loss": 0.2081790566444397, "global_step": 91182, "epoch": 1024, "lr": 7.481585129310631e-05}
{"train_loss": 0.21449026465415955, "global_step": 91183, "epoch": 1024, "lr": 7.481534799811867e-05}
{"train_loss": 0.2699448764324188, "global_step": 91184, "epoch": 1024, "lr": 7.481484469979488e-05}
{"train_loss": 0.1884000152349472, "global_step": 91185, "epoch": 1024, "lr": 7.481434139813506e-05}
{"train_loss": 0.18454326689243317, "global_step": 91186, "epoch": 1024, "lr": 7.481383809313924e-05}
{"train_loss": 0.17370159924030304, "global_step": 91187, "epoch": 1024, "lr": 7.48133347848075e-05}
{"train_loss": 0.2320854663848877, "global_step": 91188, "epoch": 1024, "lr": 7.481283147313991e-05}
{"train_loss": 0.22013558447360992, "global_step": 91189, "epoch": 1024, "lr": 7.481232815813654e-05}
{"train_loss": 0.23637785017490387, "global_step": 91190, "epoch": 1024, "lr": 7.481182483979746e-05}
{"train_loss": 0.18858247995376587, "global_step": 91191, "epoch": 1024, "lr": 7.481132151812274e-05}
{"train_loss": 0.14015692472457886, "global_step": 91192, "epoch": 1024, "lr": 7.481081819311242e-05}
{"train_loss": 0.26483258605003357, "global_step": 91193, "epoch": 1024, "lr": 7.481031486476661e-05}
{"train_loss": 0.15890946984291077, "global_step": 91194, "epoch": 1024, "lr": 7.480981153308536e-05}
{"train_loss": 0.15395860373973846, "global_step": 91195, "epoch": 1024, "lr": 7.48093081980687e-05}
{"train_loss": 0.20317791402339935, "global_step": 91196, "epoch": 1024, "lr": 7.480880485971675e-05}
{"train_loss": 0.12431924790143967, "global_step": 91197, "epoch": 1024, "lr": 7.480830151802956e-05}
{"train_loss": 0.17868544161319733, "global_step": 91198, "epoch": 1024, "lr": 7.48077981730072e-05}
{"train_loss": 0.23721282184123993, "global_step": 91199, "epoch": 1024, "lr": 7.480729482464972e-05}
{"train_loss": 0.14546868205070496, "global_step": 91200, "epoch": 1024, "lr": 7.48067914729572e-05}
{"train_loss": 0.25950783491134644, "global_step": 91201, "epoch": 1024, "lr": 7.480628811792973e-05}
{"train_loss": 0.12077653408050537, "global_step": 91202, "epoch": 1024, "lr": 7.480578475956735e-05}
{"train_loss": 0.24479445815086365, "global_step": 91203, "epoch": 1024, "lr": 7.480528139787012e-05}
{"train_loss": 0.1351718306541443, "global_step": 91204, "epoch": 1024, "lr": 7.480477803283813e-05}
{"train_loss": 0.17940661311149597, "global_step": 91205, "epoch": 1024, "lr": 7.480427466447143e-05}
{"train_loss": 0.16709482669830322, "global_step": 91206, "epoch": 1024, "lr": 7.480377129277011e-05}
{"train_loss": 0.1652270406484604, "global_step": 91207, "epoch": 1024, "lr": 7.480326791773422e-05}
{"train_loss": 0.18291690945625305, "global_step": 91208, "epoch": 1024, "lr": 7.480276453936384e-05}
{"train_loss": 0.23537315428256989, "global_step": 91209, "epoch": 1024, "lr": 7.480226115765902e-05}
{"train_loss": 0.22729215025901794, "global_step": 91210, "epoch": 1024, "lr": 7.480175777261986e-05}
{"train_loss": 0.19188056886196136, "global_step": 91211, "epoch": 1024, "lr": 7.480125438424639e-05}
{"train_loss": 0.18659371137619019, "global_step": 91212, "epoch": 1024, "lr": 7.480075099253869e-05}
{"train_loss": 0.15806777775287628, "global_step": 91213, "epoch": 1024, "lr": 7.480024759749686e-05}
{"train_loss": 0.18163201212882996, "global_step": 91214, "epoch": 1024, "lr": 7.479974419912091e-05}
{"train_loss": 0.20292510092258453, "global_step": 91215, "epoch": 1024, "lr": 7.479924079741094e-05}
{"train_loss": 0.1558898687362671, "global_step": 91216, "epoch": 1024, "lr": 7.479873739236701e-05}
{"train_loss": 0.16324591636657715, "global_step": 91217, "epoch": 1024, "lr": 7.479823398398919e-05}
{"train_loss": 0.1585363745689392, "global_step": 91218, "epoch": 1024, "lr": 7.479773057227757e-05}
{"train_loss": 0.11504137516021729, "global_step": 91219, "epoch": 1024, "lr": 7.479722715723219e-05}
{"train_loss": 0.24333244562149048, "global_step": 91220, "epoch": 1024, "lr": 7.479672373885312e-05}
{"train_loss": 0.19625519216060638, "global_step": 91221, "epoch": 1024, "lr": 7.479622031714043e-05}
{"train_loss": 0.17288555204868317, "global_step": 91222, "epoch": 1024, "lr": 7.47957168920942e-05}
{"train_loss": 0.11237752437591553, "global_step": 91223, "epoch": 1024, "lr": 7.479521346371447e-05}
{"train_loss": 0.18635512419630973, "global_step": 91224, "epoch": 1024, "lr": 7.479471003200134e-05, "val_loss": 4.483263969421387}
{"train_loss": 0.1261064112186432, "global_step": 91225, "epoch": 1025, "lr": 7.479420659695486e-05}
{"train_loss": 0.14833872020244598, "global_step": 91226, "epoch": 1025, "lr": 7.47937031585751e-05}
{"train_loss": 0.19012023508548737, "global_step": 91227, "epoch": 1025, "lr": 7.479319971686213e-05}
{"train_loss": 0.1533196121454239, "global_step": 91228, "epoch": 1025, "lr": 7.479269627181603e-05}
{"train_loss": 0.2381141185760498, "global_step": 91229, "epoch": 1025, "lr": 7.479219282343684e-05}
{"train_loss": 0.12269924581050873, "global_step": 91230, "epoch": 1025, "lr": 7.479168937172464e-05}
{"train_loss": 0.1461164802312851, "global_step": 91231, "epoch": 1025, "lr": 7.47911859166795e-05}
{"train_loss": 0.2657431364059448, "global_step": 91232, "epoch": 1025, "lr": 7.47906824583015e-05}
{"train_loss": 0.1284054070711136, "global_step": 91233, "epoch": 1025, "lr": 7.479017899659068e-05}
{"train_loss": 0.1728636771440506, "global_step": 91234, "epoch": 1025, "lr": 7.478967553154713e-05}
{"train_loss": 0.17437613010406494, "global_step": 91235, "epoch": 1025, "lr": 7.478917206317091e-05}
{"train_loss": 0.176502525806427, "global_step": 91236, "epoch": 1025, "lr": 7.478866859146209e-05}
{"train_loss": 0.152433842420578, "global_step": 91237, "epoch": 1025, "lr": 7.478816511642073e-05}
{"train_loss": 0.12407290190458298, "global_step": 91238, "epoch": 1025, "lr": 7.47876616380469e-05}
{"train_loss": 0.15306979417800903, "global_step": 91239, "epoch": 1025, "lr": 7.478715815634067e-05}
{"train_loss": 0.1696043759584427, "global_step": 91240, "epoch": 1025, "lr": 7.478665467130213e-05}
{"train_loss": 0.19360463321208954, "global_step": 91241, "epoch": 1025, "lr": 7.478615118293131e-05}
{"train_loss": 0.19056686758995056, "global_step": 91242, "epoch": 1025, "lr": 7.478564769122828e-05}
{"train_loss": 0.15885254740715027, "global_step": 91243, "epoch": 1025, "lr": 7.478514419619314e-05}
{"train_loss": 0.15063424408435822, "global_step": 91244, "epoch": 1025, "lr": 7.478464069782594e-05}
{"train_loss": 0.1435510516166687, "global_step": 91245, "epoch": 1025, "lr": 7.478413719612674e-05}
{"train_loss": 0.15346454083919525, "global_step": 91246, "epoch": 1025, "lr": 7.478363369109563e-05}
{"train_loss": 0.15652208030223846, "global_step": 91247, "epoch": 1025, "lr": 7.478313018273264e-05}
{"train_loss": 0.1580943465232849, "global_step": 91248, "epoch": 1025, "lr": 7.478262667103787e-05}
{"train_loss": 0.10654862225055695, "global_step": 91249, "epoch": 1025, "lr": 7.478212315601138e-05}
{"train_loss": 0.17892113327980042, "global_step": 91250, "epoch": 1025, "lr": 7.478161963765323e-05}
{"train_loss": 0.177557110786438, "global_step": 91251, "epoch": 1025, "lr": 7.47811161159635e-05}
{"train_loss": 0.17535017430782318, "global_step": 91252, "epoch": 1025, "lr": 7.478061259094227e-05}
{"train_loss": 0.18249697983264923, "global_step": 91253, "epoch": 1025, "lr": 7.478010906258957e-05}
{"train_loss": 0.18124639987945557, "global_step": 91254, "epoch": 1025, "lr": 7.477960553090547e-05}
{"train_loss": 0.1762877106666565, "global_step": 91255, "epoch": 1025, "lr": 7.477910199589007e-05}
{"train_loss": 0.13184727728366852, "global_step": 91256, "epoch": 1025, "lr": 7.477859845754343e-05}
{"train_loss": 0.1577187180519104, "global_step": 91257, "epoch": 1025, "lr": 7.47780949158656e-05}
{"train_loss": 0.17708326876163483, "global_step": 91258, "epoch": 1025, "lr": 7.477759137085664e-05}
{"train_loss": 0.17148181796073914, "global_step": 91259, "epoch": 1025, "lr": 7.477708782251667e-05}
{"train_loss": 0.20424266159534454, "global_step": 91260, "epoch": 1025, "lr": 7.47765842708457e-05}
{"train_loss": 0.17584075033664703, "global_step": 91261, "epoch": 1025, "lr": 7.477608071584382e-05}
{"train_loss": 0.12535180151462555, "global_step": 91262, "epoch": 1025, "lr": 7.477557715751109e-05}
{"train_loss": 0.18437938392162323, "global_step": 91263, "epoch": 1025, "lr": 7.47750735958476e-05}
{"train_loss": 0.16302159428596497, "global_step": 91264, "epoch": 1025, "lr": 7.47745700308534e-05}
{"train_loss": 0.18241244554519653, "global_step": 91265, "epoch": 1025, "lr": 7.477406646252855e-05}
{"train_loss": 0.1532447636127472, "global_step": 91266, "epoch": 1025, "lr": 7.477356289087315e-05}
{"train_loss": 0.18779681622982025, "global_step": 91267, "epoch": 1025, "lr": 7.477305931588724e-05}
{"train_loss": 0.16321678459644318, "global_step": 91268, "epoch": 1025, "lr": 7.477255573757089e-05}
{"train_loss": 0.21488657593727112, "global_step": 91269, "epoch": 1025, "lr": 7.477205215592417e-05}
{"train_loss": 0.13458257913589478, "global_step": 91270, "epoch": 1025, "lr": 7.477154857094716e-05}
{"train_loss": 0.15409809350967407, "global_step": 91271, "epoch": 1025, "lr": 7.477104498263992e-05}
{"train_loss": 0.1571924090385437, "global_step": 91272, "epoch": 1025, "lr": 7.477054139100251e-05}
{"train_loss": 0.1981240063905716, "global_step": 91273, "epoch": 1025, "lr": 7.477003779603498e-05}
{"train_loss": 0.19120089709758759, "global_step": 91274, "epoch": 1025, "lr": 7.476953419773744e-05}
{"train_loss": 0.15967492759227753, "global_step": 91275, "epoch": 1025, "lr": 7.476903059610993e-05}
{"train_loss": 0.13163936138153076, "global_step": 91276, "epoch": 1025, "lr": 7.476852699115254e-05}
{"train_loss": 0.15923011302947998, "global_step": 91277, "epoch": 1025, "lr": 7.476802338286532e-05}
{"train_loss": 0.14789226651191711, "global_step": 91278, "epoch": 1025, "lr": 7.476751977124834e-05}
{"train_loss": 0.2094096690416336, "global_step": 91279, "epoch": 1025, "lr": 7.476701615630165e-05}
{"train_loss": 0.1812024861574173, "global_step": 91280, "epoch": 1025, "lr": 7.476651253802535e-05}
{"train_loss": 0.16767628490924835, "global_step": 91281, "epoch": 1025, "lr": 7.476600891641949e-05}
{"train_loss": 0.12178540974855423, "global_step": 91282, "epoch": 1025, "lr": 7.476550529148414e-05}
{"train_loss": 0.07778801023960114, "global_step": 91283, "epoch": 1025, "lr": 7.476500166321939e-05}
{"train_loss": 0.1880190670490265, "global_step": 91284, "epoch": 1025, "lr": 7.476449803162528e-05}
{"train_loss": 0.2274055778980255, "global_step": 91285, "epoch": 1025, "lr": 7.476399439670188e-05}
{"train_loss": 0.15921373665332794, "global_step": 91286, "epoch": 1025, "lr": 7.476349075844926e-05}
{"train_loss": 0.15674032270908356, "global_step": 91287, "epoch": 1025, "lr": 7.47629871168675e-05}
{"train_loss": 0.15383611619472504, "global_step": 91288, "epoch": 1025, "lr": 7.476248347195664e-05}
{"train_loss": 0.1781865954399109, "global_step": 91289, "epoch": 1025, "lr": 7.476197982371678e-05}
{"train_loss": 0.231596902012825, "global_step": 91290, "epoch": 1025, "lr": 7.476147617214799e-05}
{"train_loss": 0.17775492370128632, "global_step": 91291, "epoch": 1025, "lr": 7.476097251725029e-05}
{"train_loss": 0.12139279395341873, "global_step": 91292, "epoch": 1025, "lr": 7.476046885902379e-05}
{"train_loss": 0.19028037786483765, "global_step": 91293, "epoch": 1025, "lr": 7.475996519746855e-05}
{"train_loss": 0.13714276254177094, "global_step": 91294, "epoch": 1025, "lr": 7.475946153258463e-05}
{"train_loss": 0.20322826504707336, "global_step": 91295, "epoch": 1025, "lr": 7.475895786437211e-05}
{"train_loss": 0.12215986847877502, "global_step": 91296, "epoch": 1025, "lr": 7.475845419283104e-05}
{"train_loss": 0.16165205836296082, "global_step": 91297, "epoch": 1025, "lr": 7.475795051796149e-05}
{"train_loss": 0.20410014688968658, "global_step": 91298, "epoch": 1025, "lr": 7.475744683976356e-05}
{"train_loss": 0.2177402377128601, "global_step": 91299, "epoch": 1025, "lr": 7.475694315823727e-05}
{"train_loss": 0.13024961948394775, "global_step": 91300, "epoch": 1025, "lr": 7.475643947338272e-05}
{"train_loss": 0.12578746676445007, "global_step": 91301, "epoch": 1025, "lr": 7.475593578519996e-05}
{"train_loss": 0.15591122210025787, "global_step": 91302, "epoch": 1025, "lr": 7.475543209368909e-05}
{"train_loss": 0.1361193209886551, "global_step": 91303, "epoch": 1025, "lr": 7.475492839885013e-05}
{"train_loss": 0.21834933757781982, "global_step": 91304, "epoch": 1025, "lr": 7.475442470068319e-05}
{"train_loss": 0.22697599232196808, "global_step": 91305, "epoch": 1025, "lr": 7.475392099918831e-05}
{"train_loss": 0.17015229165554047, "global_step": 91306, "epoch": 1025, "lr": 7.475341729436557e-05}
{"train_loss": 0.08813425153493881, "global_step": 91307, "epoch": 1025, "lr": 7.475291358621504e-05}
{"train_loss": 0.13419441878795624, "global_step": 91308, "epoch": 1025, "lr": 7.475240987473676e-05}
{"train_loss": 0.2257806360721588, "global_step": 91309, "epoch": 1025, "lr": 7.475190615993084e-05}
{"train_loss": 0.18323826789855957, "global_step": 91310, "epoch": 1025, "lr": 7.475140244179732e-05}
{"train_loss": 0.12134047597646713, "global_step": 91311, "epoch": 1025, "lr": 7.47508987203363e-05}
{"train_loss": 0.19082944095134735, "global_step": 91312, "epoch": 1025, "lr": 7.475039499554779e-05}
{"train_loss": 0.16838636031646406, "global_step": 91313, "epoch": 1025, "lr": 7.47498912674319e-05, "val_loss": 4.453167915344238, "train_action_mse_error": 31.594806671142578}
{"train_loss": 0.14161115884780884, "global_step": 91314, "epoch": 1026, "lr": 7.474938753598869e-05}
{"train_loss": 0.17757613956928253, "global_step": 91315, "epoch": 1026, "lr": 7.474888380121823e-05}
{"train_loss": 0.143511563539505, "global_step": 91316, "epoch": 1026, "lr": 7.474838006312057e-05}
{"train_loss": 0.17736104130744934, "global_step": 91317, "epoch": 1026, "lr": 7.474787632169581e-05}
{"train_loss": 0.18656086921691895, "global_step": 91318, "epoch": 1026, "lr": 7.4747372576944e-05}
{"train_loss": 0.15304116904735565, "global_step": 91319, "epoch": 1026, "lr": 7.47468688288652e-05}
{"train_loss": 0.15045034885406494, "global_step": 91320, "epoch": 1026, "lr": 7.474636507745948e-05}
{"train_loss": 0.16453047096729279, "global_step": 91321, "epoch": 1026, "lr": 7.474586132272692e-05}
{"train_loss": 0.30922144651412964, "global_step": 91322, "epoch": 1026, "lr": 7.474535756466759e-05}
{"train_loss": 0.20918354392051697, "global_step": 91323, "epoch": 1026, "lr": 7.474485380328154e-05}
{"train_loss": 0.15334247052669525, "global_step": 91324, "epoch": 1026, "lr": 7.474435003856884e-05}
{"train_loss": 0.14247940480709076, "global_step": 91325, "epoch": 1026, "lr": 7.474384627052956e-05}
{"train_loss": 0.13237327337265015, "global_step": 91326, "epoch": 1026, "lr": 7.474334249916378e-05}
{"train_loss": 0.289506733417511, "global_step": 91327, "epoch": 1026, "lr": 7.474283872447156e-05}
{"train_loss": 0.1643056422472, "global_step": 91328, "epoch": 1026, "lr": 7.474233494645296e-05}
{"train_loss": 0.1509961485862732, "global_step": 91329, "epoch": 1026, "lr": 7.474183116510806e-05}
{"train_loss": 0.1486380398273468, "global_step": 91330, "epoch": 1026, "lr": 7.474132738043694e-05}
{"train_loss": 0.13753893971443176, "global_step": 91331, "epoch": 1026, "lr": 7.474082359243963e-05}
{"train_loss": 0.21220743656158447, "global_step": 91332, "epoch": 1026, "lr": 7.474031980111622e-05}
{"train_loss": 0.24771526455879211, "global_step": 91333, "epoch": 1026, "lr": 7.473981600646676e-05}
{"train_loss": 0.22664423286914825, "global_step": 91334, "epoch": 1026, "lr": 7.473931220849135e-05}
{"train_loss": 0.21829912066459656, "global_step": 91335, "epoch": 1026, "lr": 7.473880840719004e-05}
{"train_loss": 0.20778924226760864, "global_step": 91336, "epoch": 1026, "lr": 7.47383046025629e-05}
{"train_loss": 0.17399457097053528, "global_step": 91337, "epoch": 1026, "lr": 7.473780079460998e-05}
{"train_loss": 0.18439263105392456, "global_step": 91338, "epoch": 1026, "lr": 7.47372969833314e-05}
{"train_loss": 0.17064251005649567, "global_step": 91339, "epoch": 1026, "lr": 7.473679316872717e-05}
{"train_loss": 0.1874072253704071, "global_step": 91340, "epoch": 1026, "lr": 7.473628935079738e-05}
{"train_loss": 0.1245374083518982, "global_step": 91341, "epoch": 1026, "lr": 7.47357855295421e-05}
{"train_loss": 0.1676260083913803, "global_step": 91342, "epoch": 1026, "lr": 7.47352817049614e-05}
{"train_loss": 0.18479834496974945, "global_step": 91343, "epoch": 1026, "lr": 7.473477787705532e-05}
{"train_loss": 0.2024010717868805, "global_step": 91344, "epoch": 1026, "lr": 7.473427404582398e-05}
{"train_loss": 0.23652581870555878, "global_step": 91345, "epoch": 1026, "lr": 7.47337702112674e-05}
{"train_loss": 0.22330661118030548, "global_step": 91346, "epoch": 1026, "lr": 7.473326637338567e-05}
{"train_loss": 0.1760960817337036, "global_step": 91347, "epoch": 1026, "lr": 7.473276253217885e-05}
{"train_loss": 0.19619828462600708, "global_step": 91348, "epoch": 1026, "lr": 7.473225868764703e-05}
{"train_loss": 0.1533198207616806, "global_step": 91349, "epoch": 1026, "lr": 7.473175483979024e-05}
{"train_loss": 0.18940286338329315, "global_step": 91350, "epoch": 1026, "lr": 7.473125098860859e-05}
{"train_loss": 0.2069128006696701, "global_step": 91351, "epoch": 1026, "lr": 7.473074713410209e-05}
{"train_loss": 0.14647886157035828, "global_step": 91352, "epoch": 1026, "lr": 7.473024327627086e-05}
{"train_loss": 0.1928063929080963, "global_step": 91353, "epoch": 1026, "lr": 7.472973941511495e-05}
{"train_loss": 0.19596171379089355, "global_step": 91354, "epoch": 1026, "lr": 7.472923555063443e-05}
{"train_loss": 0.1819363832473755, "global_step": 91355, "epoch": 1026, "lr": 7.472873168282936e-05}
{"train_loss": 0.15990008413791656, "global_step": 91356, "epoch": 1026, "lr": 7.472822781169981e-05}
{"train_loss": 0.19596661627292633, "global_step": 91357, "epoch": 1026, "lr": 7.472772393724586e-05}
{"train_loss": 0.15769758820533752, "global_step": 91358, "epoch": 1026, "lr": 7.472722005946757e-05}
{"train_loss": 0.1334456354379654, "global_step": 91359, "epoch": 1026, "lr": 7.472671617836501e-05}
{"train_loss": 0.18011866509914398, "global_step": 91360, "epoch": 1026, "lr": 7.472621229393824e-05}
{"train_loss": 0.21191644668579102, "global_step": 91361, "epoch": 1026, "lr": 7.472570840618732e-05}
{"train_loss": 0.15468396246433258, "global_step": 91362, "epoch": 1026, "lr": 7.472520451511235e-05}
{"train_loss": 0.1193002387881279, "global_step": 91363, "epoch": 1026, "lr": 7.472470062071337e-05}
{"train_loss": 0.18596962094306946, "global_step": 91364, "epoch": 1026, "lr": 7.472419672299045e-05}
{"train_loss": 0.21546028554439545, "global_step": 91365, "epoch": 1026, "lr": 7.472369282194366e-05}
{"train_loss": 0.16553311049938202, "global_step": 91366, "epoch": 1026, "lr": 7.472318891757308e-05}
{"train_loss": 0.20392736792564392, "global_step": 91367, "epoch": 1026, "lr": 7.472268500987877e-05}
{"train_loss": 0.19428393244743347, "global_step": 91368, "epoch": 1026, "lr": 7.47221810988608e-05}
{"train_loss": 0.1627039611339569, "global_step": 91369, "epoch": 1026, "lr": 7.472167718451923e-05}
{"train_loss": 0.25811970233917236, "global_step": 91370, "epoch": 1026, "lr": 7.472117326685412e-05}
{"train_loss": 0.1851469725370407, "global_step": 91371, "epoch": 1026, "lr": 7.472066934586556e-05}
{"train_loss": 0.22837162017822266, "global_step": 91372, "epoch": 1026, "lr": 7.47201654215536e-05}
{"train_loss": 0.20832699537277222, "global_step": 91373, "epoch": 1026, "lr": 7.471966149391832e-05}
{"train_loss": 0.19006897509098053, "global_step": 91374, "epoch": 1026, "lr": 7.471915756295979e-05}
{"train_loss": 0.20993572473526, "global_step": 91375, "epoch": 1026, "lr": 7.471865362867805e-05}
{"train_loss": 0.16270215809345245, "global_step": 91376, "epoch": 1026, "lr": 7.47181496910732e-05}
{"train_loss": 0.1199352890253067, "global_step": 91377, "epoch": 1026, "lr": 7.47176457501453e-05}
{"train_loss": 0.21984146535396576, "global_step": 91378, "epoch": 1026, "lr": 7.471714180589442e-05}
{"train_loss": 0.15817837417125702, "global_step": 91379, "epoch": 1026, "lr": 7.471663785832061e-05}
{"train_loss": 0.16927705705165863, "global_step": 91380, "epoch": 1026, "lr": 7.471613390742396e-05}
{"train_loss": 0.19236017763614655, "global_step": 91381, "epoch": 1026, "lr": 7.471562995320451e-05}
{"train_loss": 0.27361705899238586, "global_step": 91382, "epoch": 1026, "lr": 7.471512599566235e-05}
{"train_loss": 0.1398719847202301, "global_step": 91383, "epoch": 1026, "lr": 7.471462203479755e-05}
{"train_loss": 0.14839722216129303, "global_step": 91384, "epoch": 1026, "lr": 7.471411807061018e-05}
{"train_loss": 0.1632714420557022, "global_step": 91385, "epoch": 1026, "lr": 7.471361410310027e-05}
{"train_loss": 0.24243128299713135, "global_step": 91386, "epoch": 1026, "lr": 7.471311013226794e-05}
{"train_loss": 0.18313747644424438, "global_step": 91387, "epoch": 1026, "lr": 7.471260615811323e-05}
{"train_loss": 0.2451070100069046, "global_step": 91388, "epoch": 1026, "lr": 7.47121021806362e-05}
{"train_loss": 0.227638840675354, "global_step": 91389, "epoch": 1026, "lr": 7.471159819983692e-05}
{"train_loss": 0.24160444736480713, "global_step": 91390, "epoch": 1026, "lr": 7.47110942157155e-05}
{"train_loss": 0.21222130954265594, "global_step": 91391, "epoch": 1026, "lr": 7.471059022827195e-05}
{"train_loss": 0.23534822463989258, "global_step": 91392, "epoch": 1026, "lr": 7.471008623750637e-05}
{"train_loss": 0.20621341466903687, "global_step": 91393, "epoch": 1026, "lr": 7.470958224341882e-05}
{"train_loss": 0.17530114948749542, "global_step": 91394, "epoch": 1026, "lr": 7.470907824600937e-05}
{"train_loss": 0.23697267472743988, "global_step": 91395, "epoch": 1026, "lr": 7.470857424527807e-05}
{"train_loss": 0.16146549582481384, "global_step": 91396, "epoch": 1026, "lr": 7.470807024122502e-05}
{"train_loss": 0.12543533742427826, "global_step": 91397, "epoch": 1026, "lr": 7.470756623385027e-05}
{"train_loss": 0.21996036171913147, "global_step": 91398, "epoch": 1026, "lr": 7.470706222315388e-05}
{"train_loss": 0.1594201773405075, "global_step": 91399, "epoch": 1026, "lr": 7.470655820913594e-05}
{"train_loss": 0.21289458870887756, "global_step": 91400, "epoch": 1026, "lr": 7.470605419179649e-05}
{"train_loss": 0.17713017761707306, "global_step": 91401, "epoch": 1026, "lr": 7.470555017113562e-05}
{"train_loss": 0.18785246478372744, "global_step": 91402, "epoch": 1026, "lr": 7.470504614715338e-05, "val_loss": 4.502687454223633}
{"train_loss": 0.2158072590827942, "global_step": 91403, "epoch": 1027, "lr": 7.470454211984986e-05}
{"train_loss": 0.16083940863609314, "global_step": 91404, "epoch": 1027, "lr": 7.470403808922511e-05}
{"train_loss": 0.15208959579467773, "global_step": 91405, "epoch": 1027, "lr": 7.470353405527921e-05}
{"train_loss": 0.24632737040519714, "global_step": 91406, "epoch": 1027, "lr": 7.470303001801222e-05}
{"train_loss": 0.21788738667964935, "global_step": 91407, "epoch": 1027, "lr": 7.47025259774242e-05}
{"train_loss": 0.2144668698310852, "global_step": 91408, "epoch": 1027, "lr": 7.470202193351523e-05}
{"train_loss": 0.1434192806482315, "global_step": 91409, "epoch": 1027, "lr": 7.470151788628537e-05}
{"train_loss": 0.1571529060602188, "global_step": 91410, "epoch": 1027, "lr": 7.47010138357347e-05}
{"train_loss": 0.18863332271575928, "global_step": 91411, "epoch": 1027, "lr": 7.470050978186329e-05}
{"train_loss": 0.07115493714809418, "global_step": 91412, "epoch": 1027, "lr": 7.470000572467117e-05}
{"train_loss": 0.16965562105178833, "global_step": 91413, "epoch": 1027, "lr": 7.469950166415845e-05}
{"train_loss": 0.18077720701694489, "global_step": 91414, "epoch": 1027, "lr": 7.469899760032518e-05}
{"train_loss": 0.12579290568828583, "global_step": 91415, "epoch": 1027, "lr": 7.469849353317144e-05}
{"train_loss": 0.12616421282291412, "global_step": 91416, "epoch": 1027, "lr": 7.469798946269728e-05}
{"train_loss": 0.15513162314891815, "global_step": 91417, "epoch": 1027, "lr": 7.469748538890277e-05}
{"train_loss": 0.13303223252296448, "global_step": 91418, "epoch": 1027, "lr": 7.4696981311788e-05}
{"train_loss": 0.1561720073223114, "global_step": 91419, "epoch": 1027, "lr": 7.469647723135302e-05}
{"train_loss": 0.2086360901594162, "global_step": 91420, "epoch": 1027, "lr": 7.469597314759788e-05}
{"train_loss": 0.13859814405441284, "global_step": 91421, "epoch": 1027, "lr": 7.46954690605227e-05}
{"train_loss": 0.1906580626964569, "global_step": 91422, "epoch": 1027, "lr": 7.469496497012749e-05}
{"train_loss": 0.1579878181219101, "global_step": 91423, "epoch": 1027, "lr": 7.469446087641234e-05}
{"train_loss": 0.21732327342033386, "global_step": 91424, "epoch": 1027, "lr": 7.469395677937734e-05}
{"train_loss": 0.23587054014205933, "global_step": 91425, "epoch": 1027, "lr": 7.469345267902252e-05}
{"train_loss": 0.24773623049259186, "global_step": 91426, "epoch": 1027, "lr": 7.469294857534799e-05}
{"train_loss": 0.1772029548883438, "global_step": 91427, "epoch": 1027, "lr": 7.469244446835377e-05}
{"train_loss": 0.1369614601135254, "global_step": 91428, "epoch": 1027, "lr": 7.469194035803996e-05}
{"train_loss": 0.21105019748210907, "global_step": 91429, "epoch": 1027, "lr": 7.469143624440663e-05}
{"train_loss": 0.17460671067237854, "global_step": 91430, "epoch": 1027, "lr": 7.469093212745381e-05}
{"train_loss": 0.180227592587471, "global_step": 91431, "epoch": 1027, "lr": 7.469042800718162e-05}
{"train_loss": 0.12306768447160721, "global_step": 91432, "epoch": 1027, "lr": 7.468992388359009e-05}
{"train_loss": 0.1629011183977127, "global_step": 91433, "epoch": 1027, "lr": 7.468941975667931e-05}
{"train_loss": 0.15120521187782288, "global_step": 91434, "epoch": 1027, "lr": 7.468891562644934e-05}
{"train_loss": 0.16213567554950714, "global_step": 91435, "epoch": 1027, "lr": 7.468841149290025e-05}
{"train_loss": 0.24648264050483704, "global_step": 91436, "epoch": 1027, "lr": 7.468790735603208e-05}
{"train_loss": 0.17049159109592438, "global_step": 91437, "epoch": 1027, "lr": 7.468740321584495e-05}
{"train_loss": 0.19796417653560638, "global_step": 91438, "epoch": 1027, "lr": 7.468689907233887e-05}
{"train_loss": 0.11263787001371384, "global_step": 91439, "epoch": 1027, "lr": 7.468639492551397e-05}
{"train_loss": 0.1613517701625824, "global_step": 91440, "epoch": 1027, "lr": 7.468589077537026e-05}
{"train_loss": 0.1626758575439453, "global_step": 91441, "epoch": 1027, "lr": 7.468538662190785e-05}
{"train_loss": 0.11619612574577332, "global_step": 91442, "epoch": 1027, "lr": 7.468488246512678e-05}
{"train_loss": 0.18659228086471558, "global_step": 91443, "epoch": 1027, "lr": 7.468437830502712e-05}
{"train_loss": 0.21060366928577423, "global_step": 91444, "epoch": 1027, "lr": 7.468387414160897e-05}
{"train_loss": 0.1813192516565323, "global_step": 91445, "epoch": 1027, "lr": 7.468336997487236e-05}
{"train_loss": 0.15142765641212463, "global_step": 91446, "epoch": 1027, "lr": 7.468286580481737e-05}
{"train_loss": 0.15205809473991394, "global_step": 91447, "epoch": 1027, "lr": 7.468236163144407e-05}
{"train_loss": 0.2017345130443573, "global_step": 91448, "epoch": 1027, "lr": 7.468185745475253e-05}
{"train_loss": 0.18513016402721405, "global_step": 91449, "epoch": 1027, "lr": 7.468135327474282e-05}
{"train_loss": 0.2021273672580719, "global_step": 91450, "epoch": 1027, "lr": 7.4680849091415e-05}
{"train_loss": 0.19644910097122192, "global_step": 91451, "epoch": 1027, "lr": 7.468034490476913e-05}
{"train_loss": 0.27302441000938416, "global_step": 91452, "epoch": 1027, "lr": 7.467984071480529e-05}
{"train_loss": 0.11037936061620712, "global_step": 91453, "epoch": 1027, "lr": 7.467933652152356e-05}
{"train_loss": 0.14032021164894104, "global_step": 91454, "epoch": 1027, "lr": 7.467883232492398e-05}
{"train_loss": 0.15579625964164734, "global_step": 91455, "epoch": 1027, "lr": 7.467832812500665e-05}
{"train_loss": 0.1571749746799469, "global_step": 91456, "epoch": 1027, "lr": 7.46778239217716e-05}
{"train_loss": 0.1482284665107727, "global_step": 91457, "epoch": 1027, "lr": 7.467731971521893e-05}
{"train_loss": 0.1841316819190979, "global_step": 91458, "epoch": 1027, "lr": 7.467681550534869e-05}
{"train_loss": 0.15382692217826843, "global_step": 91459, "epoch": 1027, "lr": 7.467631129216095e-05}
{"train_loss": 0.22147831320762634, "global_step": 91460, "epoch": 1027, "lr": 7.467580707565578e-05}
{"train_loss": 0.14942993223667145, "global_step": 91461, "epoch": 1027, "lr": 7.467530285583327e-05}
{"train_loss": 0.1574941724538803, "global_step": 91462, "epoch": 1027, "lr": 7.467479863269343e-05}
{"train_loss": 0.19549787044525146, "global_step": 91463, "epoch": 1027, "lr": 7.467429440623639e-05}
{"train_loss": 0.12284733355045319, "global_step": 91464, "epoch": 1027, "lr": 7.46737901764622e-05}
{"train_loss": 0.18215444684028625, "global_step": 91465, "epoch": 1027, "lr": 7.467328594337089e-05}
{"train_loss": 0.11761071532964706, "global_step": 91466, "epoch": 1027, "lr": 7.467278170696257e-05}
{"train_loss": 0.1950172334909439, "global_step": 91467, "epoch": 1027, "lr": 7.467227746723729e-05}
{"train_loss": 0.2300715297460556, "global_step": 91468, "epoch": 1027, "lr": 7.467177322419514e-05}
{"train_loss": 0.19055499136447906, "global_step": 91469, "epoch": 1027, "lr": 7.467126897783616e-05}
{"train_loss": 0.17474448680877686, "global_step": 91470, "epoch": 1027, "lr": 7.467076472816042e-05}
{"train_loss": 0.10834559798240662, "global_step": 91471, "epoch": 1027, "lr": 7.467026047516801e-05}
{"train_loss": 0.16795580089092255, "global_step": 91472, "epoch": 1027, "lr": 7.466975621885899e-05}
{"train_loss": 0.1345849484205246, "global_step": 91473, "epoch": 1027, "lr": 7.466925195923341e-05}
{"train_loss": 0.17863087356090546, "global_step": 91474, "epoch": 1027, "lr": 7.466874769629135e-05}
{"train_loss": 0.2043691873550415, "global_step": 91475, "epoch": 1027, "lr": 7.466824343003288e-05}
{"train_loss": 0.16653324663639069, "global_step": 91476, "epoch": 1027, "lr": 7.466773916045806e-05}
{"train_loss": 0.24867503345012665, "global_step": 91477, "epoch": 1027, "lr": 7.466723488756698e-05}
{"train_loss": 0.13620352745056152, "global_step": 91478, "epoch": 1027, "lr": 7.466673061135965e-05}
{"train_loss": 0.19456540048122406, "global_step": 91479, "epoch": 1027, "lr": 7.466622633183623e-05}
{"train_loss": 0.12119420617818832, "global_step": 91480, "epoch": 1027, "lr": 7.466572204899671e-05}
{"train_loss": 0.19954580068588257, "global_step": 91481, "epoch": 1027, "lr": 7.466521776284118e-05}
{"train_loss": 0.1633811891078949, "global_step": 91482, "epoch": 1027, "lr": 7.466471347336973e-05}
{"train_loss": 0.21333418786525726, "global_step": 91483, "epoch": 1027, "lr": 7.46642091805824e-05}
{"train_loss": 0.20010234415531158, "global_step": 91484, "epoch": 1027, "lr": 7.466370488447926e-05}
{"train_loss": 0.18721675872802734, "global_step": 91485, "epoch": 1027, "lr": 7.46632005850604e-05}
{"train_loss": 0.17134912312030792, "global_step": 91486, "epoch": 1027, "lr": 7.466269628232587e-05}
{"train_loss": 0.2121383547782898, "global_step": 91487, "epoch": 1027, "lr": 7.466219197627573e-05}
{"train_loss": 0.16066278517246246, "global_step": 91488, "epoch": 1027, "lr": 7.466168766691008e-05}
{"train_loss": 0.2033674120903015, "global_step": 91489, "epoch": 1027, "lr": 7.466118335422894e-05}
{"train_loss": 0.1377655416727066, "global_step": 91490, "epoch": 1027, "lr": 7.466067903823242e-05}
{"train_loss": 0.17406416985760914, "global_step": 91491, "epoch": 1027, "lr": 7.466017471892057e-05, "val_loss": 4.510740756988525}
{"train_loss": 0.22378943860530853, "global_step": 91492, "epoch": 1028, "lr": 7.465967039629347e-05}
{"train_loss": 0.1563350409269333, "global_step": 91493, "epoch": 1028, "lr": 7.465916607035116e-05}
{"train_loss": 0.21968068182468414, "global_step": 91494, "epoch": 1028, "lr": 7.465866174109374e-05}
{"train_loss": 0.15903329849243164, "global_step": 91495, "epoch": 1028, "lr": 7.465815740852125e-05}
{"train_loss": 0.232985258102417, "global_step": 91496, "epoch": 1028, "lr": 7.465765307263379e-05}
{"train_loss": 0.1962362825870514, "global_step": 91497, "epoch": 1028, "lr": 7.465714873343139e-05}
{"train_loss": 0.16606883704662323, "global_step": 91498, "epoch": 1028, "lr": 7.465664439091415e-05}
{"train_loss": 0.15920068323612213, "global_step": 91499, "epoch": 1028, "lr": 7.465614004508211e-05}
{"train_loss": 0.15685772895812988, "global_step": 91500, "epoch": 1028, "lr": 7.465563569593537e-05}
{"train_loss": 0.19788891077041626, "global_step": 91501, "epoch": 1028, "lr": 7.465513134347398e-05}
{"train_loss": 0.14202557504177094, "global_step": 91502, "epoch": 1028, "lr": 7.465462698769799e-05}
{"train_loss": 0.13766270875930786, "global_step": 91503, "epoch": 1028, "lr": 7.46541226286075e-05}
{"train_loss": 0.2089708149433136, "global_step": 91504, "epoch": 1028, "lr": 7.465361826620255e-05}
{"train_loss": 0.17899461090564728, "global_step": 91505, "epoch": 1028, "lr": 7.465311390048323e-05}
{"train_loss": 0.15279139578342438, "global_step": 91506, "epoch": 1028, "lr": 7.465260953144961e-05}
{"train_loss": 0.24602971971035004, "global_step": 91507, "epoch": 1028, "lr": 7.465210515910173e-05}
{"train_loss": 0.21261736750602722, "global_step": 91508, "epoch": 1028, "lr": 7.46516007834397e-05}
{"train_loss": 0.28554049134254456, "global_step": 91509, "epoch": 1028, "lr": 7.465109640446353e-05}
{"train_loss": 0.17369917035102844, "global_step": 91510, "epoch": 1028, "lr": 7.465059202217333e-05}
{"train_loss": 0.21426984667778015, "global_step": 91511, "epoch": 1028, "lr": 7.465008763656917e-05}
{"train_loss": 0.16364318132400513, "global_step": 91512, "epoch": 1028, "lr": 7.46495832476511e-05}
{"train_loss": 0.11723251640796661, "global_step": 91513, "epoch": 1028, "lr": 7.464907885541919e-05}
{"train_loss": 0.16099664568901062, "global_step": 91514, "epoch": 1028, "lr": 7.464857445987352e-05}
{"train_loss": 0.17275768518447876, "global_step": 91515, "epoch": 1028, "lr": 7.464807006101413e-05}
{"train_loss": 0.1649111658334732, "global_step": 91516, "epoch": 1028, "lr": 7.464756565884113e-05}
{"train_loss": 0.13487151265144348, "global_step": 91517, "epoch": 1028, "lr": 7.464706125335457e-05}
{"train_loss": 0.15562567114830017, "global_step": 91518, "epoch": 1028, "lr": 7.46465568445545e-05}
{"train_loss": 0.16055874526500702, "global_step": 91519, "epoch": 1028, "lr": 7.4646052432441e-05}
{"train_loss": 0.1814497858285904, "global_step": 91520, "epoch": 1028, "lr": 7.464554801701415e-05}
{"train_loss": 0.18178479373455048, "global_step": 91521, "epoch": 1028, "lr": 7.464504359827399e-05}
{"train_loss": 0.1765047311782837, "global_step": 91522, "epoch": 1028, "lr": 7.464453917622061e-05}
{"train_loss": 0.15626701712608337, "global_step": 91523, "epoch": 1028, "lr": 7.464403475085408e-05}
{"train_loss": 0.15703478455543518, "global_step": 91524, "epoch": 1028, "lr": 7.464353032217445e-05}
{"train_loss": 0.1558752804994583, "global_step": 91525, "epoch": 1028, "lr": 7.46430258901818e-05}
{"train_loss": 0.16085174679756165, "global_step": 91526, "epoch": 1028, "lr": 7.46425214548762e-05}
{"train_loss": 0.122044138610363, "global_step": 91527, "epoch": 1028, "lr": 7.46420170162577e-05}
{"train_loss": 0.21290679275989532, "global_step": 91528, "epoch": 1028, "lr": 7.464151257432639e-05}
{"train_loss": 0.1489161252975464, "global_step": 91529, "epoch": 1028, "lr": 7.464100812908233e-05}
{"train_loss": 0.20004302263259888, "global_step": 91530, "epoch": 1028, "lr": 7.46405036805256e-05}
{"train_loss": 0.1306547373533249, "global_step": 91531, "epoch": 1028, "lr": 7.463999922865623e-05}
{"train_loss": 0.1658082753419876, "global_step": 91532, "epoch": 1028, "lr": 7.463949477347432e-05}
{"train_loss": 0.16986559331417084, "global_step": 91533, "epoch": 1028, "lr": 7.463899031497994e-05}
{"train_loss": 0.12598152458667755, "global_step": 91534, "epoch": 1028, "lr": 7.463848585317314e-05}
{"train_loss": 0.23649494349956512, "global_step": 91535, "epoch": 1028, "lr": 7.463798138805399e-05}
{"train_loss": 0.14704513549804688, "global_step": 91536, "epoch": 1028, "lr": 7.463747691962258e-05}
{"train_loss": 0.10420811176300049, "global_step": 91537, "epoch": 1028, "lr": 7.463697244787896e-05}
{"train_loss": 0.16166669130325317, "global_step": 91538, "epoch": 1028, "lr": 7.46364679728232e-05}
{"train_loss": 0.18676243722438812, "global_step": 91539, "epoch": 1028, "lr": 7.463596349445535e-05}
{"train_loss": 0.11959917843341827, "global_step": 91540, "epoch": 1028, "lr": 7.463545901277551e-05}
{"train_loss": 0.18812070786952972, "global_step": 91541, "epoch": 1028, "lr": 7.463495452778372e-05}
{"train_loss": 0.21453356742858887, "global_step": 91542, "epoch": 1028, "lr": 7.463445003948008e-05}
{"train_loss": 0.18965798616409302, "global_step": 91543, "epoch": 1028, "lr": 7.463394554786462e-05}
{"train_loss": 0.21364159882068634, "global_step": 91544, "epoch": 1028, "lr": 7.463344105293743e-05}
{"train_loss": 0.18148519098758698, "global_step": 91545, "epoch": 1028, "lr": 7.463293655469858e-05}
{"train_loss": 0.1781940907239914, "global_step": 91546, "epoch": 1028, "lr": 7.463243205314813e-05}
{"train_loss": 0.24426355957984924, "global_step": 91547, "epoch": 1028, "lr": 7.463192754828614e-05}
{"train_loss": 0.15950186550617218, "global_step": 91548, "epoch": 1028, "lr": 7.46314230401127e-05}
{"train_loss": 0.19757284224033356, "global_step": 91549, "epoch": 1028, "lr": 7.463091852862787e-05}
{"train_loss": 0.11090034991502762, "global_step": 91550, "epoch": 1028, "lr": 7.463041401383169e-05}
{"train_loss": 0.1342100352048874, "global_step": 91551, "epoch": 1028, "lr": 7.462990949572426e-05}
{"train_loss": 0.27454888820648193, "global_step": 91552, "epoch": 1028, "lr": 7.462940497430566e-05}
{"train_loss": 0.17873349785804749, "global_step": 91553, "epoch": 1028, "lr": 7.462890044957591e-05}
{"train_loss": 0.24007639288902283, "global_step": 91554, "epoch": 1028, "lr": 7.462839592153512e-05}
{"train_loss": 0.1534513384103775, "global_step": 91555, "epoch": 1028, "lr": 7.462789139018333e-05}
{"train_loss": 0.23196005821228027, "global_step": 91556, "epoch": 1028, "lr": 7.462738685552064e-05}
{"train_loss": 0.1963951140642166, "global_step": 91557, "epoch": 1028, "lr": 7.462688231754708e-05}
{"train_loss": 0.20041322708129883, "global_step": 91558, "epoch": 1028, "lr": 7.462637777626276e-05}
{"train_loss": 0.19026246666908264, "global_step": 91559, "epoch": 1028, "lr": 7.462587323166769e-05}
{"train_loss": 0.19216127693653107, "global_step": 91560, "epoch": 1028, "lr": 7.4625368683762e-05}
{"train_loss": 0.19628658890724182, "global_step": 91561, "epoch": 1028, "lr": 7.462486413254571e-05}
{"train_loss": 0.17343340814113617, "global_step": 91562, "epoch": 1028, "lr": 7.462435957801893e-05}
{"train_loss": 0.2259373962879181, "global_step": 91563, "epoch": 1028, "lr": 7.462385502018167e-05}
{"train_loss": 0.1512642353773117, "global_step": 91564, "epoch": 1028, "lr": 7.462335045903406e-05}
{"train_loss": 0.23959699273109436, "global_step": 91565, "epoch": 1028, "lr": 7.462284589457614e-05}
{"train_loss": 0.1894165426492691, "global_step": 91566, "epoch": 1028, "lr": 7.462234132680797e-05}
{"train_loss": 0.16947565972805023, "global_step": 91567, "epoch": 1028, "lr": 7.462183675572962e-05}
{"train_loss": 0.17280307412147522, "global_step": 91568, "epoch": 1028, "lr": 7.462133218134118e-05}
{"train_loss": 0.19870984554290771, "global_step": 91569, "epoch": 1028, "lr": 7.462082760364267e-05}
{"train_loss": 0.19012948870658875, "global_step": 91570, "epoch": 1028, "lr": 7.462032302263422e-05}
{"train_loss": 0.24772511422634125, "global_step": 91571, "epoch": 1028, "lr": 7.461981843831586e-05}
{"train_loss": 0.26875755190849304, "global_step": 91572, "epoch": 1028, "lr": 7.461931385068767e-05}
{"train_loss": 0.17722266912460327, "global_step": 91573, "epoch": 1028, "lr": 7.46188092597497e-05}
{"train_loss": 0.2879045903682709, "global_step": 91574, "epoch": 1028, "lr": 7.461830466550204e-05}
{"train_loss": 0.13984407484531403, "global_step": 91575, "epoch": 1028, "lr": 7.461780006794476e-05}
{"train_loss": 0.12978552281856537, "global_step": 91576, "epoch": 1028, "lr": 7.461729546707791e-05}
{"train_loss": 0.19375626742839813, "global_step": 91577, "epoch": 1028, "lr": 7.461679086290157e-05}
{"train_loss": 0.18738187849521637, "global_step": 91578, "epoch": 1028, "lr": 7.461628625541579e-05}
{"train_loss": 0.23446863889694214, "global_step": 91579, "epoch": 1028, "lr": 7.461578164462065e-05}
{"train_loss": 0.18354211915075108, "global_step": 91580, "epoch": 1028, "lr": 7.461527703051622e-05, "val_loss": 4.601254940032959}
{"train_loss": 0.22983674705028534, "global_step": 91581, "epoch": 1029, "lr": 7.461477241310257e-05}
{"train_loss": 0.1704220175743103, "global_step": 91582, "epoch": 1029, "lr": 7.461426779237977e-05}
{"train_loss": 0.19125622510910034, "global_step": 91583, "epoch": 1029, "lr": 7.461376316834788e-05}
{"train_loss": 0.15195751190185547, "global_step": 91584, "epoch": 1029, "lr": 7.461325854100695e-05}
{"train_loss": 0.18050885200500488, "global_step": 91585, "epoch": 1029, "lr": 7.461275391035709e-05}
{"train_loss": 0.230963796377182, "global_step": 91586, "epoch": 1029, "lr": 7.461224927639832e-05}
{"train_loss": 0.2054053694009781, "global_step": 91587, "epoch": 1029, "lr": 7.461174463913075e-05}
{"train_loss": 0.22906845808029175, "global_step": 91588, "epoch": 1029, "lr": 7.461123999855443e-05}
{"train_loss": 0.21143430471420288, "global_step": 91589, "epoch": 1029, "lr": 7.461073535466943e-05}
{"train_loss": 0.23049546778202057, "global_step": 91590, "epoch": 1029, "lr": 7.461023070747582e-05}
{"train_loss": 0.2259715497493744, "global_step": 91591, "epoch": 1029, "lr": 7.460972605697365e-05}
{"train_loss": 0.2097325474023819, "global_step": 91592, "epoch": 1029, "lr": 7.4609221403163e-05}
{"train_loss": 0.2497631311416626, "global_step": 91593, "epoch": 1029, "lr": 7.460871674604396e-05}
{"train_loss": 0.15214429795742035, "global_step": 91594, "epoch": 1029, "lr": 7.460821208561657e-05}
{"train_loss": 0.13695110380649567, "global_step": 91595, "epoch": 1029, "lr": 7.460770742188092e-05}
{"train_loss": 0.21214956045150757, "global_step": 91596, "epoch": 1029, "lr": 7.460720275483705e-05}
{"train_loss": 0.2265007644891739, "global_step": 91597, "epoch": 1029, "lr": 7.460669808448504e-05}
{"train_loss": 0.11570782959461212, "global_step": 91598, "epoch": 1029, "lr": 7.460619341082495e-05}
{"train_loss": 0.16744659841060638, "global_step": 91599, "epoch": 1029, "lr": 7.460568873385687e-05}
{"train_loss": 0.21178163588047028, "global_step": 91600, "epoch": 1029, "lr": 7.460518405358086e-05}
{"train_loss": 0.23722001910209656, "global_step": 91601, "epoch": 1029, "lr": 7.460467936999698e-05}
{"train_loss": 0.18234699964523315, "global_step": 91602, "epoch": 1029, "lr": 7.46041746831053e-05}
{"train_loss": 0.2234218865633011, "global_step": 91603, "epoch": 1029, "lr": 7.460366999290587e-05}
{"train_loss": 0.13171428442001343, "global_step": 91604, "epoch": 1029, "lr": 7.46031652993988e-05}
{"train_loss": 0.16262903809547424, "global_step": 91605, "epoch": 1029, "lr": 7.460266060258411e-05}
{"train_loss": 0.12193954735994339, "global_step": 91606, "epoch": 1029, "lr": 7.460215590246191e-05}
{"train_loss": 0.16613797843456268, "global_step": 91607, "epoch": 1029, "lr": 7.460165119903225e-05}
{"train_loss": 0.19160765409469604, "global_step": 91608, "epoch": 1029, "lr": 7.46011464922952e-05}
{"train_loss": 0.20101451873779297, "global_step": 91609, "epoch": 1029, "lr": 7.460064178225081e-05}
{"train_loss": 0.16395384073257446, "global_step": 91610, "epoch": 1029, "lr": 7.460013706889918e-05}
{"train_loss": 0.18301650881767273, "global_step": 91611, "epoch": 1029, "lr": 7.459963235224036e-05}
{"train_loss": 0.17599128186702728, "global_step": 91612, "epoch": 1029, "lr": 7.459912763227441e-05}
{"train_loss": 0.20353104174137115, "global_step": 91613, "epoch": 1029, "lr": 7.459862290900142e-05}
{"train_loss": 0.23304149508476257, "global_step": 91614, "epoch": 1029, "lr": 7.459811818242144e-05}
{"train_loss": 0.1619933843612671, "global_step": 91615, "epoch": 1029, "lr": 7.459761345253454e-05}
{"train_loss": 0.22169671952724457, "global_step": 91616, "epoch": 1029, "lr": 7.45971087193408e-05}
{"train_loss": 0.27973294258117676, "global_step": 91617, "epoch": 1029, "lr": 7.459660398284027e-05}
{"train_loss": 0.20385967195034027, "global_step": 91618, "epoch": 1029, "lr": 7.459609924303302e-05}
{"train_loss": 0.24756227433681488, "global_step": 91619, "epoch": 1029, "lr": 7.459559449991913e-05}
{"train_loss": 0.22037136554718018, "global_step": 91620, "epoch": 1029, "lr": 7.459508975349867e-05}
{"train_loss": 0.19829468429088593, "global_step": 91621, "epoch": 1029, "lr": 7.459458500377169e-05}
{"train_loss": 0.1967056393623352, "global_step": 91622, "epoch": 1029, "lr": 7.459408025073826e-05}
{"train_loss": 0.2827281653881073, "global_step": 91623, "epoch": 1029, "lr": 7.459357549439847e-05}
{"train_loss": 0.1295211911201477, "global_step": 91624, "epoch": 1029, "lr": 7.459307073475237e-05}
{"train_loss": 0.25958481431007385, "global_step": 91625, "epoch": 1029, "lr": 7.459256597180002e-05}
{"train_loss": 0.2489788681268692, "global_step": 91626, "epoch": 1029, "lr": 7.459206120554152e-05}
{"train_loss": 0.21378546953201294, "global_step": 91627, "epoch": 1029, "lr": 7.45915564359769e-05}
{"train_loss": 0.2926252782344818, "global_step": 91628, "epoch": 1029, "lr": 7.459105166310625e-05}
{"train_loss": 0.16677288711071014, "global_step": 91629, "epoch": 1029, "lr": 7.459054688692964e-05}
{"train_loss": 0.19511154294013977, "global_step": 91630, "epoch": 1029, "lr": 7.459004210744712e-05}
{"train_loss": 0.1923218071460724, "global_step": 91631, "epoch": 1029, "lr": 7.458953732465878e-05}
{"train_loss": 0.13238507509231567, "global_step": 91632, "epoch": 1029, "lr": 7.458903253856466e-05}
{"train_loss": 0.20343072712421417, "global_step": 91633, "epoch": 1029, "lr": 7.458852774916487e-05}
{"train_loss": 0.21959498524665833, "global_step": 91634, "epoch": 1029, "lr": 7.458802295645943e-05}
{"train_loss": 0.16994905471801758, "global_step": 91635, "epoch": 1029, "lr": 7.458751816044845e-05}
{"train_loss": 0.14320963621139526, "global_step": 91636, "epoch": 1029, "lr": 7.458701336113197e-05}
{"train_loss": 0.1695999801158905, "global_step": 91637, "epoch": 1029, "lr": 7.458650855851006e-05}
{"train_loss": 0.23097503185272217, "global_step": 91638, "epoch": 1029, "lr": 7.458600375258279e-05}
{"train_loss": 0.139691561460495, "global_step": 91639, "epoch": 1029, "lr": 7.458549894335022e-05}
{"train_loss": 0.18347522616386414, "global_step": 91640, "epoch": 1029, "lr": 7.458499413081246e-05}
{"train_loss": 0.11413735151290894, "global_step": 91641, "epoch": 1029, "lr": 7.458448931496953e-05}
{"train_loss": 0.1763041466474533, "global_step": 91642, "epoch": 1029, "lr": 7.458398449582151e-05}
{"train_loss": 0.2840045690536499, "global_step": 91643, "epoch": 1029, "lr": 7.458347967336848e-05}
{"train_loss": 0.13284476101398468, "global_step": 91644, "epoch": 1029, "lr": 7.45829748476105e-05}
{"train_loss": 0.20516717433929443, "global_step": 91645, "epoch": 1029, "lr": 7.458247001854764e-05}
{"train_loss": 0.22320176661014557, "global_step": 91646, "epoch": 1029, "lr": 7.458196518617997e-05}
{"train_loss": 0.14299243688583374, "global_step": 91647, "epoch": 1029, "lr": 7.458146035050755e-05}
{"train_loss": 0.2645373046398163, "global_step": 91648, "epoch": 1029, "lr": 7.458095551153045e-05}
{"train_loss": 0.22599411010742188, "global_step": 91649, "epoch": 1029, "lr": 7.458045066924875e-05}
{"train_loss": 0.16182297468185425, "global_step": 91650, "epoch": 1029, "lr": 7.457994582366249e-05}
{"train_loss": 0.16414152085781097, "global_step": 91651, "epoch": 1029, "lr": 7.457944097477176e-05}
{"train_loss": 0.09156735241413116, "global_step": 91652, "epoch": 1029, "lr": 7.457893612257664e-05}
{"train_loss": 0.18412524461746216, "global_step": 91653, "epoch": 1029, "lr": 7.457843126707717e-05}
{"train_loss": 0.17432552576065063, "global_step": 91654, "epoch": 1029, "lr": 7.457792640827345e-05}
{"train_loss": 0.22460900247097015, "global_step": 91655, "epoch": 1029, "lr": 7.45774215461655e-05}
{"train_loss": 0.14545831084251404, "global_step": 91656, "epoch": 1029, "lr": 7.457691668075343e-05}
{"train_loss": 0.19397972524166107, "global_step": 91657, "epoch": 1029, "lr": 7.457641181203728e-05}
{"train_loss": 0.15266698598861694, "global_step": 91658, "epoch": 1029, "lr": 7.457590694001714e-05}
{"train_loss": 0.13092532753944397, "global_step": 91659, "epoch": 1029, "lr": 7.457540206469307e-05}
{"train_loss": 0.1713491678237915, "global_step": 91660, "epoch": 1029, "lr": 7.457489718606511e-05}
{"train_loss": 0.1586543768644333, "global_step": 91661, "epoch": 1029, "lr": 7.457439230413339e-05}
{"train_loss": 0.2345394343137741, "global_step": 91662, "epoch": 1029, "lr": 7.457388741889793e-05}
{"train_loss": 0.14050237834453583, "global_step": 91663, "epoch": 1029, "lr": 7.457338253035882e-05}
{"train_loss": 0.14828714728355408, "global_step": 91664, "epoch": 1029, "lr": 7.45728776385161e-05}
{"train_loss": 0.18180027604103088, "global_step": 91665, "epoch": 1029, "lr": 7.457237274336986e-05}
{"train_loss": 0.12773540616035461, "global_step": 91666, "epoch": 1029, "lr": 7.457186784492018e-05}
{"train_loss": 0.17220304906368256, "global_step": 91667, "epoch": 1029, "lr": 7.45713629431671e-05}
{"train_loss": 0.1820029318332672, "global_step": 91668, "epoch": 1029, "lr": 7.457085803811071e-05}
{"train_loss": 0.19022114533052015, "global_step": 91669, "epoch": 1029, "lr": 7.457035312975105e-05, "val_loss": 4.503338813781738}
{"train_loss": 0.13047119975090027, "global_step": 91670, "epoch": 1030, "lr": 7.456984821808821e-05}
{"train_loss": 0.15802906453609467, "global_step": 91671, "epoch": 1030, "lr": 7.456934330312227e-05}
{"train_loss": 0.15938405692577362, "global_step": 91672, "epoch": 1030, "lr": 7.456883838485326e-05}
{"train_loss": 0.14125774800777435, "global_step": 91673, "epoch": 1030, "lr": 7.456833346328129e-05}
{"train_loss": 0.24450640380382538, "global_step": 91674, "epoch": 1030, "lr": 7.456782853840639e-05}
{"train_loss": 0.2501978874206543, "global_step": 91675, "epoch": 1030, "lr": 7.456732361022865e-05}
{"train_loss": 0.20629489421844482, "global_step": 91676, "epoch": 1030, "lr": 7.456681867874813e-05}
{"train_loss": 0.13908198475837708, "global_step": 91677, "epoch": 1030, "lr": 7.456631374396491e-05}
{"train_loss": 0.1392013281583786, "global_step": 91678, "epoch": 1030, "lr": 7.456580880587903e-05}
{"train_loss": 0.1602989137172699, "global_step": 91679, "epoch": 1030, "lr": 7.456530386449059e-05}
{"train_loss": 0.116126649081707, "global_step": 91680, "epoch": 1030, "lr": 7.456479891979963e-05}
{"train_loss": 0.11809343844652176, "global_step": 91681, "epoch": 1030, "lr": 7.456429397180625e-05}
{"train_loss": 0.21445217728614807, "global_step": 91682, "epoch": 1030, "lr": 7.45637890205105e-05}
{"train_loss": 0.10258851945400238, "global_step": 91683, "epoch": 1030, "lr": 7.456328406591245e-05}
{"train_loss": 0.2842831015586853, "global_step": 91684, "epoch": 1030, "lr": 7.456277910801215e-05}
{"train_loss": 0.12983004748821259, "global_step": 91685, "epoch": 1030, "lr": 7.45622741468097e-05}
{"train_loss": 0.12957851588726044, "global_step": 91686, "epoch": 1030, "lr": 7.456176918230514e-05}
{"train_loss": 0.14757028222084045, "global_step": 91687, "epoch": 1030, "lr": 7.456126421449857e-05}
{"train_loss": 0.14594611525535583, "global_step": 91688, "epoch": 1030, "lr": 7.456075924339e-05}
{"train_loss": 0.2379952073097229, "global_step": 91689, "epoch": 1030, "lr": 7.456025426897957e-05}
{"train_loss": 0.1823035478591919, "global_step": 91690, "epoch": 1030, "lr": 7.455974929126729e-05}
{"train_loss": 0.1485375314950943, "global_step": 91691, "epoch": 1030, "lr": 7.455924431025328e-05}
{"train_loss": 0.13124507665634155, "global_step": 91692, "epoch": 1030, "lr": 7.455873932593755e-05}
{"train_loss": 0.2437451183795929, "global_step": 91693, "epoch": 1030, "lr": 7.45582343383202e-05}
{"train_loss": 0.16312305629253387, "global_step": 91694, "epoch": 1030, "lr": 7.45577293474013e-05}
{"train_loss": 0.10045535117387772, "global_step": 91695, "epoch": 1030, "lr": 7.455722435318091e-05}
{"train_loss": 0.17788273096084595, "global_step": 91696, "epoch": 1030, "lr": 7.455671935565912e-05}
{"train_loss": 0.1105070486664772, "global_step": 91697, "epoch": 1030, "lr": 7.455621435483595e-05}
{"train_loss": 0.22802719473838806, "global_step": 91698, "epoch": 1030, "lr": 7.45557093507115e-05}
{"train_loss": 0.23740993440151215, "global_step": 91699, "epoch": 1030, "lr": 7.455520434328584e-05}
{"train_loss": 0.20094060897827148, "global_step": 91700, "epoch": 1030, "lr": 7.455469933255904e-05}
{"train_loss": 0.22904235124588013, "global_step": 91701, "epoch": 1030, "lr": 7.455419431853116e-05}
{"train_loss": 0.16899050772190094, "global_step": 91702, "epoch": 1030, "lr": 7.455368930120226e-05}
{"train_loss": 0.10453625023365021, "global_step": 91703, "epoch": 1030, "lr": 7.455318428057241e-05}
{"train_loss": 0.1671278178691864, "global_step": 91704, "epoch": 1030, "lr": 7.45526792566417e-05}
{"train_loss": 0.14320969581604004, "global_step": 91705, "epoch": 1030, "lr": 7.455217422941018e-05}
{"train_loss": 0.15014581382274628, "global_step": 91706, "epoch": 1030, "lr": 7.455166919887791e-05}
{"train_loss": 0.1795492172241211, "global_step": 91707, "epoch": 1030, "lr": 7.455116416504497e-05}
{"train_loss": 0.23845243453979492, "global_step": 91708, "epoch": 1030, "lr": 7.455065912791143e-05}
{"train_loss": 0.2739925980567932, "global_step": 91709, "epoch": 1030, "lr": 7.455015408747734e-05}
{"train_loss": 0.2273704558610916, "global_step": 91710, "epoch": 1030, "lr": 7.45496490437428e-05}
{"train_loss": 0.2189844399690628, "global_step": 91711, "epoch": 1030, "lr": 7.454914399670785e-05}
{"train_loss": 0.12691406905651093, "global_step": 91712, "epoch": 1030, "lr": 7.454863894637257e-05}
{"train_loss": 0.14524729549884796, "global_step": 91713, "epoch": 1030, "lr": 7.454813389273702e-05}
{"train_loss": 0.17558197677135468, "global_step": 91714, "epoch": 1030, "lr": 7.454762883580126e-05}
{"train_loss": 0.22145721316337585, "global_step": 91715, "epoch": 1030, "lr": 7.454712377556539e-05}
{"train_loss": 0.15680918097496033, "global_step": 91716, "epoch": 1030, "lr": 7.454661871202945e-05}
{"train_loss": 0.12909604609012604, "global_step": 91717, "epoch": 1030, "lr": 7.454611364519352e-05}
{"train_loss": 0.19451361894607544, "global_step": 91718, "epoch": 1030, "lr": 7.454560857505766e-05}
{"train_loss": 0.0968603566288948, "global_step": 91719, "epoch": 1030, "lr": 7.454510350162194e-05}
{"train_loss": 0.267372727394104, "global_step": 91720, "epoch": 1030, "lr": 7.454459842488644e-05}
{"train_loss": 0.23615069687366486, "global_step": 91721, "epoch": 1030, "lr": 7.454409334485121e-05}
{"train_loss": 0.20762187242507935, "global_step": 91722, "epoch": 1030, "lr": 7.454358826151633e-05}
{"train_loss": 0.19515272974967957, "global_step": 91723, "epoch": 1030, "lr": 7.454308317488185e-05}
{"train_loss": 0.20292045176029205, "global_step": 91724, "epoch": 1030, "lr": 7.454257808494786e-05}
{"train_loss": 0.12287719547748566, "global_step": 91725, "epoch": 1030, "lr": 7.454207299171443e-05}
{"train_loss": 0.24770475924015045, "global_step": 91726, "epoch": 1030, "lr": 7.454156789518161e-05}
{"train_loss": 0.3203403949737549, "global_step": 91727, "epoch": 1030, "lr": 7.454106279534948e-05}
{"train_loss": 0.21923507750034332, "global_step": 91728, "epoch": 1030, "lr": 7.454055769221809e-05}
{"train_loss": 0.20990213751792908, "global_step": 91729, "epoch": 1030, "lr": 7.454005258578753e-05}
{"train_loss": 0.1475318968296051, "global_step": 91730, "epoch": 1030, "lr": 7.453954747605787e-05}
{"train_loss": 0.14331094920635223, "global_step": 91731, "epoch": 1030, "lr": 7.453904236302915e-05}
{"train_loss": 0.22132012248039246, "global_step": 91732, "epoch": 1030, "lr": 7.453853724670146e-05}
{"train_loss": 0.17198166251182556, "global_step": 91733, "epoch": 1030, "lr": 7.453803212707487e-05}
{"train_loss": 0.1745934933423996, "global_step": 91734, "epoch": 1030, "lr": 7.453752700414942e-05}
{"train_loss": 0.18024343252182007, "global_step": 91735, "epoch": 1030, "lr": 7.453702187792522e-05}
{"train_loss": 0.2936399579048157, "global_step": 91736, "epoch": 1030, "lr": 7.453651674840231e-05}
{"train_loss": 0.2906131148338318, "global_step": 91737, "epoch": 1030, "lr": 7.453601161558076e-05}
{"train_loss": 0.23201489448547363, "global_step": 91738, "epoch": 1030, "lr": 7.453550647946064e-05}
{"train_loss": 0.28815174102783203, "global_step": 91739, "epoch": 1030, "lr": 7.453500134004203e-05}
{"train_loss": 0.19396255910396576, "global_step": 91740, "epoch": 1030, "lr": 7.453449619732498e-05}
{"train_loss": 0.2828230559825897, "global_step": 91741, "epoch": 1030, "lr": 7.453399105130958e-05}
{"train_loss": 0.1343117654323578, "global_step": 91742, "epoch": 1030, "lr": 7.453348590199587e-05}
{"train_loss": 0.09670546650886536, "global_step": 91743, "epoch": 1030, "lr": 7.453298074938392e-05}
{"train_loss": 0.10828937590122223, "global_step": 91744, "epoch": 1030, "lr": 7.453247559347383e-05}
{"train_loss": 0.16714777052402496, "global_step": 91745, "epoch": 1030, "lr": 7.453197043426566e-05}
{"train_loss": 0.15003752708435059, "global_step": 91746, "epoch": 1030, "lr": 7.453146527175944e-05}
{"train_loss": 0.16522926092147827, "global_step": 91747, "epoch": 1030, "lr": 7.453096010595528e-05}
{"train_loss": 0.19954578578472137, "global_step": 91748, "epoch": 1030, "lr": 7.453045493685322e-05}
{"train_loss": 0.21564289927482605, "global_step": 91749, "epoch": 1030, "lr": 7.452994976445334e-05}
{"train_loss": 0.19666129350662231, "global_step": 91750, "epoch": 1030, "lr": 7.452944458875572e-05}
{"train_loss": 0.2001839131116867, "global_step": 91751, "epoch": 1030, "lr": 7.45289394097604e-05}
{"train_loss": 0.21249310672283173, "global_step": 91752, "epoch": 1030, "lr": 7.452843422746748e-05}
{"train_loss": 0.20935183763504028, "global_step": 91753, "epoch": 1030, "lr": 7.452792904187698e-05}
{"train_loss": 0.21014975011348724, "global_step": 91754, "epoch": 1030, "lr": 7.452742385298903e-05}
{"train_loss": 0.14580702781677246, "global_step": 91755, "epoch": 1030, "lr": 7.452691866080365e-05}
{"train_loss": 0.1783725619316101, "global_step": 91756, "epoch": 1030, "lr": 7.452641346532094e-05}
{"train_loss": 0.19683244824409485, "global_step": 91757, "epoch": 1030, "lr": 7.452590826654094e-05}
{"train_loss": 0.18465504041883382, "global_step": 91758, "epoch": 1030, "lr": 7.452540306446373e-05, "val_loss": 4.4270243644714355, "train_action_mse_error": 19.368606567382812}
{"train_loss": 0.1694992482662201, "global_step": 91759, "epoch": 1031, "lr": 7.452489785908938e-05}
{"train_loss": 0.16145385801792145, "global_step": 91760, "epoch": 1031, "lr": 7.452439265041795e-05}
{"train_loss": 0.15015359222888947, "global_step": 91761, "epoch": 1031, "lr": 7.452388743844953e-05}
{"train_loss": 0.20100271701812744, "global_step": 91762, "epoch": 1031, "lr": 7.452338222318415e-05}
{"train_loss": 0.15632109344005585, "global_step": 91763, "epoch": 1031, "lr": 7.452287700462192e-05}
{"train_loss": 0.21761366724967957, "global_step": 91764, "epoch": 1031, "lr": 7.452237178276289e-05}
{"train_loss": 0.16596142947673798, "global_step": 91765, "epoch": 1031, "lr": 7.452186655760711e-05}
{"train_loss": 0.18258056044578552, "global_step": 91766, "epoch": 1031, "lr": 7.452136132915468e-05}
{"train_loss": 0.24440208077430725, "global_step": 91767, "epoch": 1031, "lr": 7.452085609740565e-05}
{"train_loss": 0.18851608037948608, "global_step": 91768, "epoch": 1031, "lr": 7.452035086236008e-05}
{"train_loss": 0.17469502985477448, "global_step": 91769, "epoch": 1031, "lr": 7.451984562401804e-05}
{"train_loss": 0.2564450204372406, "global_step": 91770, "epoch": 1031, "lr": 7.451934038237963e-05}
{"train_loss": 0.26364442706108093, "global_step": 91771, "epoch": 1031, "lr": 7.451883513744487e-05}
{"train_loss": 0.1768081933259964, "global_step": 91772, "epoch": 1031, "lr": 7.451832988921388e-05}
{"train_loss": 0.18135537207126617, "global_step": 91773, "epoch": 1031, "lr": 7.451782463768667e-05}
{"train_loss": 0.132591113448143, "global_step": 91774, "epoch": 1031, "lr": 7.451731938286336e-05}
{"train_loss": 0.14739519357681274, "global_step": 91775, "epoch": 1031, "lr": 7.451681412474398e-05}
{"train_loss": 0.22175303101539612, "global_step": 91776, "epoch": 1031, "lr": 7.45163088633286e-05}
{"train_loss": 0.1552993208169937, "global_step": 91777, "epoch": 1031, "lr": 7.451580359861734e-05}
{"train_loss": 0.14616931974887848, "global_step": 91778, "epoch": 1031, "lr": 7.451529833061019e-05}
{"train_loss": 0.17482930421829224, "global_step": 91779, "epoch": 1031, "lr": 7.451479305930728e-05}
{"train_loss": 0.1370050609111786, "global_step": 91780, "epoch": 1031, "lr": 7.451428778470866e-05}
{"train_loss": 0.19210602343082428, "global_step": 91781, "epoch": 1031, "lr": 7.451378250681437e-05}
{"train_loss": 0.21105733513832092, "global_step": 91782, "epoch": 1031, "lr": 7.451327722562452e-05}
{"train_loss": 0.2071782350540161, "global_step": 91783, "epoch": 1031, "lr": 7.451277194113915e-05}
{"train_loss": 0.2277863323688507, "global_step": 91784, "epoch": 1031, "lr": 7.451226665335834e-05}
{"train_loss": 0.14233949780464172, "global_step": 91785, "epoch": 1031, "lr": 7.451176136228214e-05}
{"train_loss": 0.15522687137126923, "global_step": 91786, "epoch": 1031, "lr": 7.451125606791066e-05}
{"train_loss": 0.19204680621623993, "global_step": 91787, "epoch": 1031, "lr": 7.451075077024392e-05}
{"train_loss": 0.1676000952720642, "global_step": 91788, "epoch": 1031, "lr": 7.451024546928203e-05}
{"train_loss": 0.14791712164878845, "global_step": 91789, "epoch": 1031, "lr": 7.450974016502502e-05}
{"train_loss": 0.2162940949201584, "global_step": 91790, "epoch": 1031, "lr": 7.450923485747296e-05}
{"train_loss": 0.17429478466510773, "global_step": 91791, "epoch": 1031, "lr": 7.450872954662595e-05}
{"train_loss": 0.27590495347976685, "global_step": 91792, "epoch": 1031, "lr": 7.450822423248403e-05}
{"train_loss": 0.1594167798757553, "global_step": 91793, "epoch": 1031, "lr": 7.450771891504728e-05}
{"train_loss": 0.17218594253063202, "global_step": 91794, "epoch": 1031, "lr": 7.450721359431579e-05}
{"train_loss": 0.17019309103488922, "global_step": 91795, "epoch": 1031, "lr": 7.450670827028957e-05}
{"train_loss": 0.23142418265342712, "global_step": 91796, "epoch": 1031, "lr": 7.450620294296874e-05}
{"train_loss": 0.13992778956890106, "global_step": 91797, "epoch": 1031, "lr": 7.450569761235335e-05}
{"train_loss": 0.15937259793281555, "global_step": 91798, "epoch": 1031, "lr": 7.450519227844346e-05}
{"train_loss": 0.14865009486675262, "global_step": 91799, "epoch": 1031, "lr": 7.450468694123915e-05}
{"train_loss": 0.2388811707496643, "global_step": 91800, "epoch": 1031, "lr": 7.450418160074048e-05}
{"train_loss": 0.1694900095462799, "global_step": 91801, "epoch": 1031, "lr": 7.450367625694752e-05}
{"train_loss": 0.1417512148618698, "global_step": 91802, "epoch": 1031, "lr": 7.450317090986035e-05}
{"train_loss": 0.11019207537174225, "global_step": 91803, "epoch": 1031, "lr": 7.450266555947903e-05}
{"train_loss": 0.22471822798252106, "global_step": 91804, "epoch": 1031, "lr": 7.45021602058036e-05}
{"train_loss": 0.14328746497631073, "global_step": 91805, "epoch": 1031, "lr": 7.450165484883419e-05}
{"train_loss": 0.1688394397497177, "global_step": 91806, "epoch": 1031, "lr": 7.450114948857078e-05}
{"train_loss": 0.1947837769985199, "global_step": 91807, "epoch": 1031, "lr": 7.450064412501353e-05}
{"train_loss": 0.17609155178070068, "global_step": 91808, "epoch": 1031, "lr": 7.450013875816247e-05}
{"train_loss": 0.13471704721450806, "global_step": 91809, "epoch": 1031, "lr": 7.449963338801765e-05}
{"train_loss": 0.18833529949188232, "global_step": 91810, "epoch": 1031, "lr": 7.449912801457914e-05}
{"train_loss": 0.1595718413591385, "global_step": 91811, "epoch": 1031, "lr": 7.449862263784703e-05}
{"train_loss": 0.09802432358264923, "global_step": 91812, "epoch": 1031, "lr": 7.449811725782139e-05}
{"train_loss": 0.15560239553451538, "global_step": 91813, "epoch": 1031, "lr": 7.449761187450227e-05}
{"train_loss": 0.07048080861568451, "global_step": 91814, "epoch": 1031, "lr": 7.449710648788974e-05}
{"train_loss": 0.22169914841651917, "global_step": 91815, "epoch": 1031, "lr": 7.449660109798389e-05}
{"train_loss": 0.20241251587867737, "global_step": 91816, "epoch": 1031, "lr": 7.449609570478475e-05}
{"train_loss": 0.17565315961837769, "global_step": 91817, "epoch": 1031, "lr": 7.449559030829241e-05}
{"train_loss": 0.24414169788360596, "global_step": 91818, "epoch": 1031, "lr": 7.449508490850695e-05}
{"train_loss": 0.20311374962329865, "global_step": 91819, "epoch": 1031, "lr": 7.449457950542842e-05}
{"train_loss": 0.14015649259090424, "global_step": 91820, "epoch": 1031, "lr": 7.449407409905688e-05}
{"train_loss": 0.16408585011959076, "global_step": 91821, "epoch": 1031, "lr": 7.449356868939243e-05}
{"train_loss": 0.1921127736568451, "global_step": 91822, "epoch": 1031, "lr": 7.449306327643511e-05}
{"train_loss": 0.15927207469940186, "global_step": 91823, "epoch": 1031, "lr": 7.4492557860185e-05}
{"train_loss": 0.15367774665355682, "global_step": 91824, "epoch": 1031, "lr": 7.449205244064215e-05}
{"train_loss": 0.10539086163043976, "global_step": 91825, "epoch": 1031, "lr": 7.449154701780665e-05}
{"train_loss": 0.16872382164001465, "global_step": 91826, "epoch": 1031, "lr": 7.449104159167856e-05}
{"train_loss": 0.16352584958076477, "global_step": 91827, "epoch": 1031, "lr": 7.449053616225796e-05}
{"train_loss": 0.19559703767299652, "global_step": 91828, "epoch": 1031, "lr": 7.449003072954489e-05}
{"train_loss": 0.17056375741958618, "global_step": 91829, "epoch": 1031, "lr": 7.448952529353943e-05}
{"train_loss": 0.1485128551721573, "global_step": 91830, "epoch": 1031, "lr": 7.448901985424166e-05}
{"train_loss": 0.2003546953201294, "global_step": 91831, "epoch": 1031, "lr": 7.448851441165164e-05}
{"train_loss": 0.12335779517889023, "global_step": 91832, "epoch": 1031, "lr": 7.448800896576943e-05}
{"train_loss": 0.1543094515800476, "global_step": 91833, "epoch": 1031, "lr": 7.448750351659512e-05}
{"train_loss": 0.22232069075107574, "global_step": 91834, "epoch": 1031, "lr": 7.448699806412877e-05}
{"train_loss": 0.21521519124507904, "global_step": 91835, "epoch": 1031, "lr": 7.448649260837041e-05}
{"train_loss": 0.14930804073810577, "global_step": 91836, "epoch": 1031, "lr": 7.448598714932017e-05}
{"train_loss": 0.17254191637039185, "global_step": 91837, "epoch": 1031, "lr": 7.448548168697808e-05}
{"train_loss": 0.19452594220638275, "global_step": 91838, "epoch": 1031, "lr": 7.44849762213442e-05}
{"train_loss": 0.2484586089849472, "global_step": 91839, "epoch": 1031, "lr": 7.448447075241863e-05}
{"train_loss": 0.16810031235218048, "global_step": 91840, "epoch": 1031, "lr": 7.448396528020142e-05}
{"train_loss": 0.12000562995672226, "global_step": 91841, "epoch": 1031, "lr": 7.448345980469264e-05}
{"train_loss": 0.15787813067436218, "global_step": 91842, "epoch": 1031, "lr": 7.448295432589234e-05}
{"train_loss": 0.16627384722232819, "global_step": 91843, "epoch": 1031, "lr": 7.448244884380062e-05}
{"train_loss": 0.1184801310300827, "global_step": 91844, "epoch": 1031, "lr": 7.448194335841754e-05}
{"train_loss": 0.19541724026203156, "global_step": 91845, "epoch": 1031, "lr": 7.448143786974314e-05}
{"train_loss": 0.1600920855998993, "global_step": 91846, "epoch": 1031, "lr": 7.448093237777753e-05}
{"train_loss": 0.17453480520275202, "global_step": 91847, "epoch": 1031, "lr": 7.448042688252075e-05, "val_loss": 4.510280609130859}
{"train_loss": 0.1761597841978073, "global_step": 91848, "epoch": 1032, "lr": 7.447992138397286e-05}
{"train_loss": 0.20558735728263855, "global_step": 91849, "epoch": 1032, "lr": 7.447941588213396e-05}
{"train_loss": 0.17922036349773407, "global_step": 91850, "epoch": 1032, "lr": 7.44789103770041e-05}
{"train_loss": 0.13480786979198456, "global_step": 91851, "epoch": 1032, "lr": 7.447840486858333e-05}
{"train_loss": 0.20441250503063202, "global_step": 91852, "epoch": 1032, "lr": 7.447789935687177e-05}
{"train_loss": 0.13247212767601013, "global_step": 91853, "epoch": 1032, "lr": 7.447739384186943e-05}
{"train_loss": 0.22292205691337585, "global_step": 91854, "epoch": 1032, "lr": 7.447688832357641e-05}
{"train_loss": 0.16812579333782196, "global_step": 91855, "epoch": 1032, "lr": 7.447638280199278e-05}
{"train_loss": 0.21339485049247742, "global_step": 91856, "epoch": 1032, "lr": 7.447587727711858e-05}
{"train_loss": 0.10722415149211884, "global_step": 91857, "epoch": 1032, "lr": 7.447537174895392e-05}
{"train_loss": 0.17379160225391388, "global_step": 91858, "epoch": 1032, "lr": 7.447486621749882e-05}
{"train_loss": 0.16862183809280396, "global_step": 91859, "epoch": 1032, "lr": 7.447436068275338e-05}
{"train_loss": 0.19258858263492584, "global_step": 91860, "epoch": 1032, "lr": 7.447385514471767e-05}
{"train_loss": 0.1738670915365219, "global_step": 91861, "epoch": 1032, "lr": 7.447334960339174e-05}
{"train_loss": 0.1740129142999649, "global_step": 91862, "epoch": 1032, "lr": 7.447284405877566e-05}
{"train_loss": 0.11080709844827652, "global_step": 91863, "epoch": 1032, "lr": 7.447233851086952e-05}
{"train_loss": 0.15759840607643127, "global_step": 91864, "epoch": 1032, "lr": 7.447183295967336e-05}
{"train_loss": 0.1851149946451187, "global_step": 91865, "epoch": 1032, "lr": 7.447132740518727e-05}
{"train_loss": 0.16849464178085327, "global_step": 91866, "epoch": 1032, "lr": 7.44708218474113e-05}
{"train_loss": 0.16291669011116028, "global_step": 91867, "epoch": 1032, "lr": 7.447031628634553e-05}
{"train_loss": 0.14609266817569733, "global_step": 91868, "epoch": 1032, "lr": 7.446981072199002e-05}
{"train_loss": 0.15668773651123047, "global_step": 91869, "epoch": 1032, "lr": 7.446930515434485e-05}
{"train_loss": 0.16233062744140625, "global_step": 91870, "epoch": 1032, "lr": 7.446879958341008e-05}
{"train_loss": 0.15737205743789673, "global_step": 91871, "epoch": 1032, "lr": 7.446829400918578e-05}
{"train_loss": 0.1801796853542328, "global_step": 91872, "epoch": 1032, "lr": 7.4467788431672e-05}
{"train_loss": 0.148997962474823, "global_step": 91873, "epoch": 1032, "lr": 7.446728285086883e-05}
{"train_loss": 0.18213583528995514, "global_step": 91874, "epoch": 1032, "lr": 7.446677726677635e-05}
{"train_loss": 0.13961273431777954, "global_step": 91875, "epoch": 1032, "lr": 7.44662716793946e-05}
{"train_loss": 0.21080024540424347, "global_step": 91876, "epoch": 1032, "lr": 7.446576608872364e-05}
{"train_loss": 0.14352378249168396, "global_step": 91877, "epoch": 1032, "lr": 7.446526049476358e-05}
{"train_loss": 0.18439218401908875, "global_step": 91878, "epoch": 1032, "lr": 7.446475489751445e-05}
{"train_loss": 0.1412736028432846, "global_step": 91879, "epoch": 1032, "lr": 7.446424929697633e-05}
{"train_loss": 0.12967413663864136, "global_step": 91880, "epoch": 1032, "lr": 7.446374369314931e-05}
{"train_loss": 0.15998192131519318, "global_step": 91881, "epoch": 1032, "lr": 7.446323808603341e-05}
{"train_loss": 0.13381682336330414, "global_step": 91882, "epoch": 1032, "lr": 7.446273247562875e-05}
{"train_loss": 0.1313704550266266, "global_step": 91883, "epoch": 1032, "lr": 7.446222686193536e-05}
{"train_loss": 0.17000749707221985, "global_step": 91884, "epoch": 1032, "lr": 7.446172124495333e-05}
{"train_loss": 0.12348967790603638, "global_step": 91885, "epoch": 1032, "lr": 7.446121562468272e-05}
{"train_loss": 0.10364480316638947, "global_step": 91886, "epoch": 1032, "lr": 7.44607100011236e-05}
{"train_loss": 0.2778650224208832, "global_step": 91887, "epoch": 1032, "lr": 7.446020437427602e-05}
{"train_loss": 0.17923936247825623, "global_step": 91888, "epoch": 1032, "lr": 7.445969874414008e-05}
{"train_loss": 0.19866475462913513, "global_step": 91889, "epoch": 1032, "lr": 7.445919311071581e-05}
{"train_loss": 0.15237224102020264, "global_step": 91890, "epoch": 1032, "lr": 7.445868747400331e-05}
{"train_loss": 0.19641347229480743, "global_step": 91891, "epoch": 1032, "lr": 7.445818183400265e-05}
{"train_loss": 0.1761547476053238, "global_step": 91892, "epoch": 1032, "lr": 7.445767619071388e-05}
{"train_loss": 0.11063405871391296, "global_step": 91893, "epoch": 1032, "lr": 7.445717054413706e-05}
{"train_loss": 0.19351540505886078, "global_step": 91894, "epoch": 1032, "lr": 7.445666489427228e-05}
{"train_loss": 0.18293681740760803, "global_step": 91895, "epoch": 1032, "lr": 7.44561592411196e-05}
{"train_loss": 0.11079949140548706, "global_step": 91896, "epoch": 1032, "lr": 7.445565358467909e-05}
{"train_loss": 0.14100150763988495, "global_step": 91897, "epoch": 1032, "lr": 7.44551479249508e-05}
{"train_loss": 0.15438559651374817, "global_step": 91898, "epoch": 1032, "lr": 7.445464226193484e-05}
{"train_loss": 0.17157283425331116, "global_step": 91899, "epoch": 1032, "lr": 7.445413659563125e-05}
{"train_loss": 0.13201652467250824, "global_step": 91900, "epoch": 1032, "lr": 7.445363092604009e-05}
{"train_loss": 0.18851375579833984, "global_step": 91901, "epoch": 1032, "lr": 7.445312525316144e-05}
{"train_loss": 0.16836227476596832, "global_step": 91902, "epoch": 1032, "lr": 7.445261957699535e-05}
{"train_loss": 0.16004246473312378, "global_step": 91903, "epoch": 1032, "lr": 7.445211389754192e-05}
{"train_loss": 0.20310935378074646, "global_step": 91904, "epoch": 1032, "lr": 7.445160821480119e-05}
{"train_loss": 0.1439085602760315, "global_step": 91905, "epoch": 1032, "lr": 7.445110252877325e-05}
{"train_loss": 0.19127601385116577, "global_step": 91906, "epoch": 1032, "lr": 7.445059683945815e-05}
{"train_loss": 0.13583067059516907, "global_step": 91907, "epoch": 1032, "lr": 7.445009114685597e-05}
{"train_loss": 0.183488130569458, "global_step": 91908, "epoch": 1032, "lr": 7.444958545096676e-05}
{"train_loss": 0.1594429761171341, "global_step": 91909, "epoch": 1032, "lr": 7.444907975179061e-05}
{"train_loss": 0.19223587214946747, "global_step": 91910, "epoch": 1032, "lr": 7.44485740493276e-05}
{"train_loss": 0.20362071692943573, "global_step": 91911, "epoch": 1032, "lr": 7.444806834357775e-05}
{"train_loss": 0.25341054797172546, "global_step": 91912, "epoch": 1032, "lr": 7.444756263454117e-05}
{"train_loss": 0.191802978515625, "global_step": 91913, "epoch": 1032, "lr": 7.44470569222179e-05}
{"train_loss": 0.2188626378774643, "global_step": 91914, "epoch": 1032, "lr": 7.444655120660801e-05}
{"train_loss": 0.2264019250869751, "global_step": 91915, "epoch": 1032, "lr": 7.444604548771162e-05}
{"train_loss": 0.22327496111392975, "global_step": 91916, "epoch": 1032, "lr": 7.444553976552872e-05}
{"train_loss": 0.16249780356884003, "global_step": 91917, "epoch": 1032, "lr": 7.444503404005944e-05}
{"train_loss": 0.17354816198349, "global_step": 91918, "epoch": 1032, "lr": 7.444452831130382e-05}
{"train_loss": 0.18293523788452148, "global_step": 91919, "epoch": 1032, "lr": 7.444402257926191e-05}
{"train_loss": 0.2960270047187805, "global_step": 91920, "epoch": 1032, "lr": 7.444351684393383e-05}
{"train_loss": 0.20549307763576508, "global_step": 91921, "epoch": 1032, "lr": 7.444301110531959e-05}
{"train_loss": 0.22044864296913147, "global_step": 91922, "epoch": 1032, "lr": 7.44425053634193e-05}
{"train_loss": 0.18894316256046295, "global_step": 91923, "epoch": 1032, "lr": 7.444199961823301e-05}
{"train_loss": 0.14724010229110718, "global_step": 91924, "epoch": 1032, "lr": 7.444149386976078e-05}
{"train_loss": 0.12311279773712158, "global_step": 91925, "epoch": 1032, "lr": 7.44409881180027e-05}
{"train_loss": 0.21982111036777496, "global_step": 91926, "epoch": 1032, "lr": 7.444048236295883e-05}
{"train_loss": 0.18654023110866547, "global_step": 91927, "epoch": 1032, "lr": 7.443997660462922e-05}
{"train_loss": 0.20468436181545258, "global_step": 91928, "epoch": 1032, "lr": 7.443947084301397e-05}
{"train_loss": 0.20739924907684326, "global_step": 91929, "epoch": 1032, "lr": 7.443896507811312e-05}
{"train_loss": 0.1805655062198639, "global_step": 91930, "epoch": 1032, "lr": 7.443845930992676e-05}
{"train_loss": 0.13474911451339722, "global_step": 91931, "epoch": 1032, "lr": 7.443795353845495e-05}
{"train_loss": 0.19586485624313354, "global_step": 91932, "epoch": 1032, "lr": 7.443744776369772e-05}
{"train_loss": 0.16251184046268463, "global_step": 91933, "epoch": 1032, "lr": 7.443694198565521e-05}
{"train_loss": 0.21553637087345123, "global_step": 91934, "epoch": 1032, "lr": 7.443643620432744e-05}
{"train_loss": 0.1248251274228096, "global_step": 91935, "epoch": 1032, "lr": 7.443593041971449e-05}
{"train_loss": 0.1736182811554898, "global_step": 91936, "epoch": 1032, "lr": 7.44354246318164e-05, "val_loss": 4.541580677032471}
{"train_loss": 0.15355455875396729, "global_step": 91937, "epoch": 1033, "lr": 7.44349188406333e-05}
{"train_loss": 0.1856323927640915, "global_step": 91938, "epoch": 1033, "lr": 7.443441304616524e-05}
{"train_loss": 0.1215762048959732, "global_step": 91939, "epoch": 1033, "lr": 7.443390724841224e-05}
{"train_loss": 0.24932488799095154, "global_step": 91940, "epoch": 1033, "lr": 7.443340144737439e-05}
{"train_loss": 0.18594451248645782, "global_step": 91941, "epoch": 1033, "lr": 7.443289564305178e-05}
{"train_loss": 0.2164946049451828, "global_step": 91942, "epoch": 1033, "lr": 7.443238983544446e-05}
{"train_loss": 0.1856658160686493, "global_step": 91943, "epoch": 1033, "lr": 7.443188402455253e-05}
{"train_loss": 0.1832810342311859, "global_step": 91944, "epoch": 1033, "lr": 7.4431378210376e-05}
{"train_loss": 0.1546720713376999, "global_step": 91945, "epoch": 1033, "lr": 7.443087239291498e-05}
{"train_loss": 0.18075886368751526, "global_step": 91946, "epoch": 1033, "lr": 7.443036657216951e-05}
{"train_loss": 0.1737925410270691, "global_step": 91947, "epoch": 1033, "lr": 7.44298607481397e-05}
{"train_loss": 0.16291609406471252, "global_step": 91948, "epoch": 1033, "lr": 7.442935492082559e-05}
{"train_loss": 0.1284937560558319, "global_step": 91949, "epoch": 1033, "lr": 7.442884909022723e-05}
{"train_loss": 0.1669161021709442, "global_step": 91950, "epoch": 1033, "lr": 7.442834325634472e-05}
{"train_loss": 0.11991150677204132, "global_step": 91951, "epoch": 1033, "lr": 7.442783741917812e-05}
{"train_loss": 0.19596335291862488, "global_step": 91952, "epoch": 1033, "lr": 7.44273315787275e-05}
{"train_loss": 0.15333592891693115, "global_step": 91953, "epoch": 1033, "lr": 7.442682573499291e-05}
{"train_loss": 0.17608357965946198, "global_step": 91954, "epoch": 1033, "lr": 7.442631988797444e-05}
{"train_loss": 0.16460831463336945, "global_step": 91955, "epoch": 1033, "lr": 7.442581403767214e-05}
{"train_loss": 0.1735697239637375, "global_step": 91956, "epoch": 1033, "lr": 7.442530818408611e-05}
{"train_loss": 0.18582673370838165, "global_step": 91957, "epoch": 1033, "lr": 7.442480232721637e-05}
{"train_loss": 0.15070484578609467, "global_step": 91958, "epoch": 1033, "lr": 7.442429646706303e-05}
{"train_loss": 0.18981841206550598, "global_step": 91959, "epoch": 1033, "lr": 7.442379060362615e-05}
{"train_loss": 0.18728145956993103, "global_step": 91960, "epoch": 1033, "lr": 7.442328473690577e-05}
{"train_loss": 0.20147521793842316, "global_step": 91961, "epoch": 1033, "lr": 7.442277886690198e-05}
{"train_loss": 0.22481928765773773, "global_step": 91962, "epoch": 1033, "lr": 7.442227299361484e-05}
{"train_loss": 0.22940798103809357, "global_step": 91963, "epoch": 1033, "lr": 7.442176711704443e-05}
{"train_loss": 0.2195819914340973, "global_step": 91964, "epoch": 1033, "lr": 7.442126123719081e-05}
{"train_loss": 0.17830818891525269, "global_step": 91965, "epoch": 1033, "lr": 7.442075535405404e-05}
{"train_loss": 0.18982751667499542, "global_step": 91966, "epoch": 1033, "lr": 7.442024946763422e-05}
{"train_loss": 0.16944502294063568, "global_step": 91967, "epoch": 1033, "lr": 7.441974357793137e-05}
{"train_loss": 0.12137695401906967, "global_step": 91968, "epoch": 1033, "lr": 7.441923768494558e-05}
{"train_loss": 0.1585439294576645, "global_step": 91969, "epoch": 1033, "lr": 7.441873178867694e-05}
{"train_loss": 0.14410291612148285, "global_step": 91970, "epoch": 1033, "lr": 7.44182258891255e-05}
{"train_loss": 0.1959136426448822, "global_step": 91971, "epoch": 1033, "lr": 7.44177199862913e-05}
{"train_loss": 0.17392301559448242, "global_step": 91972, "epoch": 1033, "lr": 7.441721408017448e-05}
{"train_loss": 0.24994033575057983, "global_step": 91973, "epoch": 1033, "lr": 7.441670817077503e-05}
{"train_loss": 0.2800120711326599, "global_step": 91974, "epoch": 1033, "lr": 7.441620225809306e-05}
{"train_loss": 0.15069960057735443, "global_step": 91975, "epoch": 1033, "lr": 7.441569634212864e-05}
{"train_loss": 0.13689352571964264, "global_step": 91976, "epoch": 1033, "lr": 7.441519042288182e-05}
{"train_loss": 0.19816039502620697, "global_step": 91977, "epoch": 1033, "lr": 7.441468450035268e-05}
{"train_loss": 0.19446295499801636, "global_step": 91978, "epoch": 1033, "lr": 7.441417857454127e-05}
{"train_loss": 0.16563281416893005, "global_step": 91979, "epoch": 1033, "lr": 7.441367264544768e-05}
{"train_loss": 0.06493974477052689, "global_step": 91980, "epoch": 1033, "lr": 7.441316671307196e-05}
{"train_loss": 0.1408538669347763, "global_step": 91981, "epoch": 1033, "lr": 7.44126607774142e-05}
{"train_loss": 0.14644533395767212, "global_step": 91982, "epoch": 1033, "lr": 7.441215483847443e-05}
{"train_loss": 0.18413543701171875, "global_step": 91983, "epoch": 1033, "lr": 7.441164889625277e-05}
{"train_loss": 0.18571554124355316, "global_step": 91984, "epoch": 1033, "lr": 7.441114295074926e-05}
{"train_loss": 0.17139604687690735, "global_step": 91985, "epoch": 1033, "lr": 7.441063700196395e-05}
{"train_loss": 0.17819344997406006, "global_step": 91986, "epoch": 1033, "lr": 7.441013104989694e-05}
{"train_loss": 0.19941401481628418, "global_step": 91987, "epoch": 1033, "lr": 7.440962509454828e-05}
{"train_loss": 0.1899946630001068, "global_step": 91988, "epoch": 1033, "lr": 7.440911913591806e-05}
{"train_loss": 0.21713197231292725, "global_step": 91989, "epoch": 1033, "lr": 7.440861317400632e-05}
{"train_loss": 0.1868584007024765, "global_step": 91990, "epoch": 1033, "lr": 7.440810720881313e-05}
{"train_loss": 0.12126291543245316, "global_step": 91991, "epoch": 1033, "lr": 7.440760124033858e-05}
{"train_loss": 0.1869698315858841, "global_step": 91992, "epoch": 1033, "lr": 7.440709526858273e-05}
{"train_loss": 0.1777532994747162, "global_step": 91993, "epoch": 1033, "lr": 7.440658929354562e-05}
{"train_loss": 0.18282866477966309, "global_step": 91994, "epoch": 1033, "lr": 7.440608331522738e-05}
{"train_loss": 0.12635286152362823, "global_step": 91995, "epoch": 1033, "lr": 7.440557733362801e-05}
{"train_loss": 0.2077668160200119, "global_step": 91996, "epoch": 1033, "lr": 7.440507134874762e-05}
{"train_loss": 0.11216482520103455, "global_step": 91997, "epoch": 1033, "lr": 7.440456536058627e-05}
{"train_loss": 0.1618809849023819, "global_step": 91998, "epoch": 1033, "lr": 7.4404059369144e-05}
{"train_loss": 0.24730144441127777, "global_step": 91999, "epoch": 1033, "lr": 7.440355337442092e-05}
{"train_loss": 0.18837453424930573, "global_step": 92000, "epoch": 1033, "lr": 7.440304737641708e-05}
{"train_loss": 0.16454410552978516, "global_step": 92001, "epoch": 1033, "lr": 7.440254137513255e-05}
{"train_loss": 0.20766231417655945, "global_step": 92002, "epoch": 1033, "lr": 7.440203537056738e-05}
{"train_loss": 0.16898787021636963, "global_step": 92003, "epoch": 1033, "lr": 7.440152936272166e-05}
{"train_loss": 0.1790294051170349, "global_step": 92004, "epoch": 1033, "lr": 7.440102335159546e-05}
{"train_loss": 0.149881511926651, "global_step": 92005, "epoch": 1033, "lr": 7.440051733718883e-05}
{"train_loss": 0.16378077864646912, "global_step": 92006, "epoch": 1033, "lr": 7.440001131950183e-05}
{"train_loss": 0.15382391214370728, "global_step": 92007, "epoch": 1033, "lr": 7.439950529853457e-05}
{"train_loss": 0.1401311308145523, "global_step": 92008, "epoch": 1033, "lr": 7.439899927428711e-05}
{"train_loss": 0.18458881974220276, "global_step": 92009, "epoch": 1033, "lr": 7.439849324675946e-05}
{"train_loss": 0.1573648899793625, "global_step": 92010, "epoch": 1033, "lr": 7.439798721595177e-05}
{"train_loss": 0.2193874567747116, "global_step": 92011, "epoch": 1033, "lr": 7.439748118186405e-05}
{"train_loss": 0.16356730461120605, "global_step": 92012, "epoch": 1033, "lr": 7.43969751444964e-05}
{"train_loss": 0.11555860191583633, "global_step": 92013, "epoch": 1033, "lr": 7.439646910384885e-05}
{"train_loss": 0.13831040263175964, "global_step": 92014, "epoch": 1033, "lr": 7.43959630599215e-05}
{"train_loss": 0.1255083978176117, "global_step": 92015, "epoch": 1033, "lr": 7.439545701271444e-05}
{"train_loss": 0.18345408141613007, "global_step": 92016, "epoch": 1033, "lr": 7.439495096222767e-05}
{"train_loss": 0.06646482646465302, "global_step": 92017, "epoch": 1033, "lr": 7.439444490846132e-05}
{"train_loss": 0.23130446672439575, "global_step": 92018, "epoch": 1033, "lr": 7.439393885141541e-05}
{"train_loss": 0.19107270240783691, "global_step": 92019, "epoch": 1033, "lr": 7.439343279109004e-05}
{"train_loss": 0.1424068808555603, "global_step": 92020, "epoch": 1033, "lr": 7.439292672748528e-05}
{"train_loss": 0.17283415794372559, "global_step": 92021, "epoch": 1033, "lr": 7.439242066060119e-05}
{"train_loss": 0.21042490005493164, "global_step": 92022, "epoch": 1033, "lr": 7.439191459043783e-05}
{"train_loss": 0.17651420831680298, "global_step": 92023, "epoch": 1033, "lr": 7.439140851699527e-05}
{"train_loss": 0.2545909583568573, "global_step": 92024, "epoch": 1033, "lr": 7.439090244027358e-05}
{"train_loss": 0.17442534930920334, "global_step": 92025, "epoch": 1033, "lr": 7.439039636027283e-05, "val_loss": 4.568644046783447}
{"train_loss": 0.20122750103473663, "global_step": 92026, "epoch": 1034, "lr": 7.43898902769931e-05}
{"train_loss": 0.17267104983329773, "global_step": 92027, "epoch": 1034, "lr": 7.438938419043444e-05}
{"train_loss": 0.13275094330310822, "global_step": 92028, "epoch": 1034, "lr": 7.438887810059692e-05}
{"train_loss": 0.17267343401908875, "global_step": 92029, "epoch": 1034, "lr": 7.438837200748064e-05}
{"train_loss": 0.19410425424575806, "global_step": 92030, "epoch": 1034, "lr": 7.438786591108561e-05}
{"train_loss": 0.1368337869644165, "global_step": 92031, "epoch": 1034, "lr": 7.438735981141194e-05}
{"train_loss": 0.11274377256631851, "global_step": 92032, "epoch": 1034, "lr": 7.438685370845968e-05}
{"train_loss": 0.16585199534893036, "global_step": 92033, "epoch": 1034, "lr": 7.438634760222892e-05}
{"train_loss": 0.18224984407424927, "global_step": 92034, "epoch": 1034, "lr": 7.43858414927197e-05}
{"train_loss": 0.2172052264213562, "global_step": 92035, "epoch": 1034, "lr": 7.43853353799321e-05}
{"train_loss": 0.1415410339832306, "global_step": 92036, "epoch": 1034, "lr": 7.438482926386619e-05}
{"train_loss": 0.19769978523254395, "global_step": 92037, "epoch": 1034, "lr": 7.438432314452204e-05}
{"train_loss": 0.11646328121423721, "global_step": 92038, "epoch": 1034, "lr": 7.438381702189973e-05}
{"train_loss": 0.16599105298519135, "global_step": 92039, "epoch": 1034, "lr": 7.438331089599929e-05}
{"train_loss": 0.1366748809814453, "global_step": 92040, "epoch": 1034, "lr": 7.438280476682082e-05}
{"train_loss": 0.13830292224884033, "global_step": 92041, "epoch": 1034, "lr": 7.438229863436437e-05}
{"train_loss": 0.12321637570858002, "global_step": 92042, "epoch": 1034, "lr": 7.438179249863003e-05}
{"train_loss": 0.2642802298069, "global_step": 92043, "epoch": 1034, "lr": 7.438128635961784e-05}
{"train_loss": 0.21808937191963196, "global_step": 92044, "epoch": 1034, "lr": 7.43807802173279e-05}
{"train_loss": 0.1704799383878708, "global_step": 92045, "epoch": 1034, "lr": 7.438027407176024e-05}
{"train_loss": 0.13096608221530914, "global_step": 92046, "epoch": 1034, "lr": 7.437976792291497e-05}
{"train_loss": 0.16299480199813843, "global_step": 92047, "epoch": 1034, "lr": 7.437926177079214e-05}
{"train_loss": 0.19339123368263245, "global_step": 92048, "epoch": 1034, "lr": 7.43787556153918e-05}
{"train_loss": 0.09243415296077728, "global_step": 92049, "epoch": 1034, "lr": 7.437824945671403e-05}
{"train_loss": 0.15895366668701172, "global_step": 92050, "epoch": 1034, "lr": 7.437774329475892e-05}
{"train_loss": 0.15552660822868347, "global_step": 92051, "epoch": 1034, "lr": 7.437723712952651e-05}
{"train_loss": 0.09701734781265259, "global_step": 92052, "epoch": 1034, "lr": 7.437673096101689e-05}
{"train_loss": 0.19954723119735718, "global_step": 92053, "epoch": 1034, "lr": 7.43762247892301e-05}
{"train_loss": 0.16093307733535767, "global_step": 92054, "epoch": 1034, "lr": 7.437571861416624e-05}
{"train_loss": 0.15231850743293762, "global_step": 92055, "epoch": 1034, "lr": 7.437521243582534e-05}
{"train_loss": 0.1242789775133133, "global_step": 92056, "epoch": 1034, "lr": 7.43747062542075e-05}
{"train_loss": 0.14901408553123474, "global_step": 92057, "epoch": 1034, "lr": 7.437420006931278e-05}
{"train_loss": 0.15336325764656067, "global_step": 92058, "epoch": 1034, "lr": 7.437369388114124e-05}
{"train_loss": 0.238266721367836, "global_step": 92059, "epoch": 1034, "lr": 7.437318768969294e-05}
{"train_loss": 0.15849879384040833, "global_step": 92060, "epoch": 1034, "lr": 7.437268149496799e-05}
{"train_loss": 0.16102543473243713, "global_step": 92061, "epoch": 1034, "lr": 7.437217529696642e-05}
{"train_loss": 0.1620614230632782, "global_step": 92062, "epoch": 1034, "lr": 7.43716690956883e-05}
{"train_loss": 0.18958379328250885, "global_step": 92063, "epoch": 1034, "lr": 7.437116289113371e-05}
{"train_loss": 0.16351130604743958, "global_step": 92064, "epoch": 1034, "lr": 7.437065668330271e-05}
{"train_loss": 0.08422479033470154, "global_step": 92065, "epoch": 1034, "lr": 7.43701504721954e-05}
{"train_loss": 0.12146244198083878, "global_step": 92066, "epoch": 1034, "lr": 7.436964425781179e-05}
{"train_loss": 0.2141840010881424, "global_step": 92067, "epoch": 1034, "lr": 7.436913804015199e-05}
{"train_loss": 0.221712127327919, "global_step": 92068, "epoch": 1034, "lr": 7.436863181921605e-05}
{"train_loss": 0.12116014212369919, "global_step": 92069, "epoch": 1034, "lr": 7.436812559500406e-05}
{"train_loss": 0.19654300808906555, "global_step": 92070, "epoch": 1034, "lr": 7.436761936751606e-05}
{"train_loss": 0.32340940833091736, "global_step": 92071, "epoch": 1034, "lr": 7.436711313675212e-05}
{"train_loss": 0.16148404777050018, "global_step": 92072, "epoch": 1034, "lr": 7.436660690271233e-05}
{"train_loss": 0.14424075186252594, "global_step": 92073, "epoch": 1034, "lr": 7.436610066539677e-05}
{"train_loss": 0.15313701331615448, "global_step": 92074, "epoch": 1034, "lr": 7.436559442480546e-05}
{"train_loss": 0.14317657053470612, "global_step": 92075, "epoch": 1034, "lr": 7.436508818093849e-05}
{"train_loss": 0.1730324923992157, "global_step": 92076, "epoch": 1034, "lr": 7.436458193379595e-05}
{"train_loss": 0.20121325552463531, "global_step": 92077, "epoch": 1034, "lr": 7.436407568337785e-05}
{"train_loss": 0.20459234714508057, "global_step": 92078, "epoch": 1034, "lr": 7.436356942968432e-05}
{"train_loss": 0.19940067827701569, "global_step": 92079, "epoch": 1034, "lr": 7.436306317271541e-05}
{"train_loss": 0.16635526716709137, "global_step": 92080, "epoch": 1034, "lr": 7.436255691247118e-05}
{"train_loss": 0.21418696641921997, "global_step": 92081, "epoch": 1034, "lr": 7.43620506489517e-05}
{"train_loss": 0.1855008602142334, "global_step": 92082, "epoch": 1034, "lr": 7.436154438215704e-05}
{"train_loss": 0.1419512927532196, "global_step": 92083, "epoch": 1034, "lr": 7.436103811208727e-05}
{"train_loss": 0.1552666574716568, "global_step": 92084, "epoch": 1034, "lr": 7.436053183874245e-05}
{"train_loss": 0.14883412420749664, "global_step": 92085, "epoch": 1034, "lr": 7.436002556212266e-05}
{"train_loss": 0.1633487492799759, "global_step": 92086, "epoch": 1034, "lr": 7.435951928222795e-05}
{"train_loss": 0.23973727226257324, "global_step": 92087, "epoch": 1034, "lr": 7.435901299905841e-05}
{"train_loss": 0.1983710378408432, "global_step": 92088, "epoch": 1034, "lr": 7.435850671261411e-05}
{"train_loss": 0.20752528309822083, "global_step": 92089, "epoch": 1034, "lr": 7.435800042289509e-05}
{"train_loss": 0.1295049637556076, "global_step": 92090, "epoch": 1034, "lr": 7.435749412990143e-05}
{"train_loss": 0.20367005467414856, "global_step": 92091, "epoch": 1034, "lr": 7.43569878336332e-05}
{"train_loss": 0.12151570618152618, "global_step": 92092, "epoch": 1034, "lr": 7.435648153409048e-05}
{"train_loss": 0.19616805016994476, "global_step": 92093, "epoch": 1034, "lr": 7.435597523127333e-05}
{"train_loss": 0.133839949965477, "global_step": 92094, "epoch": 1034, "lr": 7.435546892518181e-05}
{"train_loss": 0.17643259465694427, "global_step": 92095, "epoch": 1034, "lr": 7.435496261581598e-05}
{"train_loss": 0.1773291528224945, "global_step": 92096, "epoch": 1034, "lr": 7.435445630317595e-05}
{"train_loss": 0.16426007449626923, "global_step": 92097, "epoch": 1034, "lr": 7.435394998726175e-05}
{"train_loss": 0.1459839642047882, "global_step": 92098, "epoch": 1034, "lr": 7.435344366807344e-05}
{"train_loss": 0.34368181228637695, "global_step": 92099, "epoch": 1034, "lr": 7.435293734561113e-05}
{"train_loss": 0.17137546837329865, "global_step": 92100, "epoch": 1034, "lr": 7.435243101987484e-05}
{"train_loss": 0.1768694370985031, "global_step": 92101, "epoch": 1034, "lr": 7.435192469086468e-05}
{"train_loss": 0.11374164372682571, "global_step": 92102, "epoch": 1034, "lr": 7.435141835858071e-05}
{"train_loss": 0.22139683365821838, "global_step": 92103, "epoch": 1034, "lr": 7.435091202302297e-05}
{"train_loss": 0.1970444768667221, "global_step": 92104, "epoch": 1034, "lr": 7.435040568419155e-05}
{"train_loss": 0.1602790802717209, "global_step": 92105, "epoch": 1034, "lr": 7.434989934208651e-05}
{"train_loss": 0.13876424729824066, "global_step": 92106, "epoch": 1034, "lr": 7.434939299670795e-05}
{"train_loss": 0.16824880242347717, "global_step": 92107, "epoch": 1034, "lr": 7.434888664805588e-05}
{"train_loss": 0.18830780684947968, "global_step": 92108, "epoch": 1034, "lr": 7.434838029613042e-05}
{"train_loss": 0.14046888053417206, "global_step": 92109, "epoch": 1034, "lr": 7.43478739409316e-05}
{"train_loss": 0.15460756421089172, "global_step": 92110, "epoch": 1034, "lr": 7.434736758245952e-05}
{"train_loss": 0.08126292377710342, "global_step": 92111, "epoch": 1034, "lr": 7.434686122071422e-05}
{"train_loss": 0.1665782332420349, "global_step": 92112, "epoch": 1034, "lr": 7.43463548556958e-05}
{"train_loss": 0.12534165382385254, "global_step": 92113, "epoch": 1034, "lr": 7.434584848740429e-05}
{"train_loss": 0.16989900573585812, "global_step": 92114, "epoch": 1034, "lr": 7.434534211583977e-05, "val_loss": 4.4268012046813965}
{"train_loss": 0.17665179073810577, "global_step": 92115, "epoch": 1035, "lr": 7.434483574100234e-05}
{"train_loss": 0.1784549057483673, "global_step": 92116, "epoch": 1035, "lr": 7.434432936289204e-05}
{"train_loss": 0.14926958084106445, "global_step": 92117, "epoch": 1035, "lr": 7.434382298150893e-05}
{"train_loss": 0.22991971671581268, "global_step": 92118, "epoch": 1035, "lr": 7.43433165968531e-05}
{"train_loss": 0.185878723859787, "global_step": 92119, "epoch": 1035, "lr": 7.43428102089246e-05}
{"train_loss": 0.18167346715927124, "global_step": 92120, "epoch": 1035, "lr": 7.43423038177235e-05}
{"train_loss": 0.1810884177684784, "global_step": 92121, "epoch": 1035, "lr": 7.43417974232499e-05}
{"train_loss": 0.23140104115009308, "global_step": 92122, "epoch": 1035, "lr": 7.434129102550381e-05}
{"train_loss": 0.13730080425739288, "global_step": 92123, "epoch": 1035, "lr": 7.434078462448535e-05}
{"train_loss": 0.18707415461540222, "global_step": 92124, "epoch": 1035, "lr": 7.434027822019456e-05}
{"train_loss": 0.21562565863132477, "global_step": 92125, "epoch": 1035, "lr": 7.433977181263151e-05}
{"train_loss": 0.21034783124923706, "global_step": 92126, "epoch": 1035, "lr": 7.433926540179629e-05}
{"train_loss": 0.10738302767276764, "global_step": 92127, "epoch": 1035, "lr": 7.433875898768893e-05}
{"train_loss": 0.24141089618206024, "global_step": 92128, "epoch": 1035, "lr": 7.433825257030955e-05}
{"train_loss": 0.2516084909439087, "global_step": 92129, "epoch": 1035, "lr": 7.433774614965818e-05}
{"train_loss": 0.11112795770168304, "global_step": 92130, "epoch": 1035, "lr": 7.433723972573489e-05}
{"train_loss": 0.25975558161735535, "global_step": 92131, "epoch": 1035, "lr": 7.433673329853976e-05}
{"train_loss": 0.15943199396133423, "global_step": 92132, "epoch": 1035, "lr": 7.433622686807285e-05}
{"train_loss": 0.16665303707122803, "global_step": 92133, "epoch": 1035, "lr": 7.433572043433422e-05}
{"train_loss": 0.1707119345664978, "global_step": 92134, "epoch": 1035, "lr": 7.433521399732396e-05}
{"train_loss": 0.2052914798259735, "global_step": 92135, "epoch": 1035, "lr": 7.433470755704213e-05}
{"train_loss": 0.12928783893585205, "global_step": 92136, "epoch": 1035, "lr": 7.433420111348879e-05}
{"train_loss": 0.19176791608333588, "global_step": 92137, "epoch": 1035, "lr": 7.433369466666402e-05}
{"train_loss": 0.19838792085647583, "global_step": 92138, "epoch": 1035, "lr": 7.433318821656787e-05}
{"train_loss": 0.19562511146068573, "global_step": 92139, "epoch": 1035, "lr": 7.433268176320042e-05}
{"train_loss": 0.23114542663097382, "global_step": 92140, "epoch": 1035, "lr": 7.433217530656174e-05}
{"train_loss": 0.17398099601268768, "global_step": 92141, "epoch": 1035, "lr": 7.433166884665191e-05}
{"train_loss": 0.24890904128551483, "global_step": 92142, "epoch": 1035, "lr": 7.433116238347096e-05}
{"train_loss": 0.20550230145454407, "global_step": 92143, "epoch": 1035, "lr": 7.4330655917019e-05}
{"train_loss": 0.17513836920261383, "global_step": 92144, "epoch": 1035, "lr": 7.433014944729607e-05}
{"train_loss": 0.15271927416324615, "global_step": 92145, "epoch": 1035, "lr": 7.432964297430225e-05}
{"train_loss": 0.1949428766965866, "global_step": 92146, "epoch": 1035, "lr": 7.43291364980376e-05}
{"train_loss": 0.1631881445646286, "global_step": 92147, "epoch": 1035, "lr": 7.432863001850221e-05}
{"train_loss": 0.13942907750606537, "global_step": 92148, "epoch": 1035, "lr": 7.432812353569613e-05}
{"train_loss": 0.26456156373023987, "global_step": 92149, "epoch": 1035, "lr": 7.432761704961943e-05}
{"train_loss": 0.10949854552745819, "global_step": 92150, "epoch": 1035, "lr": 7.432711056027217e-05}
{"train_loss": 0.21920111775398254, "global_step": 92151, "epoch": 1035, "lr": 7.432660406765442e-05}
{"train_loss": 0.1767071932554245, "global_step": 92152, "epoch": 1035, "lr": 7.432609757176627e-05}
{"train_loss": 0.1818051040172577, "global_step": 92153, "epoch": 1035, "lr": 7.432559107260776e-05}
{"train_loss": 0.16519193351268768, "global_step": 92154, "epoch": 1035, "lr": 7.432508457017899e-05}
{"train_loss": 0.15935583412647247, "global_step": 92155, "epoch": 1035, "lr": 7.432457806447999e-05}
{"train_loss": 0.22733542323112488, "global_step": 92156, "epoch": 1035, "lr": 7.432407155551084e-05}
{"train_loss": 0.1945299208164215, "global_step": 92157, "epoch": 1035, "lr": 7.432356504327164e-05}
{"train_loss": 0.12142868340015411, "global_step": 92158, "epoch": 1035, "lr": 7.432305852776243e-05}
{"train_loss": 0.1828453093767166, "global_step": 92159, "epoch": 1035, "lr": 7.432255200898325e-05}
{"train_loss": 0.1343691349029541, "global_step": 92160, "epoch": 1035, "lr": 7.432204548693423e-05}
{"train_loss": 0.17862209677696228, "global_step": 92161, "epoch": 1035, "lr": 7.43215389616154e-05}
{"train_loss": 0.12025955319404602, "global_step": 92162, "epoch": 1035, "lr": 7.432103243302684e-05}
{"train_loss": 0.16702309250831604, "global_step": 92163, "epoch": 1035, "lr": 7.432052590116861e-05}
{"train_loss": 0.1383390575647354, "global_step": 92164, "epoch": 1035, "lr": 7.432001936604077e-05}
{"train_loss": 0.1705496609210968, "global_step": 92165, "epoch": 1035, "lr": 7.43195128276434e-05}
{"train_loss": 0.19906175136566162, "global_step": 92166, "epoch": 1035, "lr": 7.431900628597659e-05}
{"train_loss": 0.21954895555973053, "global_step": 92167, "epoch": 1035, "lr": 7.431849974104037e-05}
{"train_loss": 0.1576870232820511, "global_step": 92168, "epoch": 1035, "lr": 7.431799319283483e-05}
{"train_loss": 0.19133713841438293, "global_step": 92169, "epoch": 1035, "lr": 7.431748664136004e-05}
{"train_loss": 0.17058420181274414, "global_step": 92170, "epoch": 1035, "lr": 7.431698008661604e-05}
{"train_loss": 0.15622161328792572, "global_step": 92171, "epoch": 1035, "lr": 7.431647352860294e-05}
{"train_loss": 0.19635552167892456, "global_step": 92172, "epoch": 1035, "lr": 7.431596696732077e-05}
{"train_loss": 0.1456061601638794, "global_step": 92173, "epoch": 1035, "lr": 7.431546040276962e-05}
{"train_loss": 0.21044251322746277, "global_step": 92174, "epoch": 1035, "lr": 7.431495383494955e-05}
{"train_loss": 0.1859855055809021, "global_step": 92175, "epoch": 1035, "lr": 7.431444726386063e-05}
{"train_loss": 0.16341713070869446, "global_step": 92176, "epoch": 1035, "lr": 7.431394068950294e-05}
{"train_loss": 0.20512725412845612, "global_step": 92177, "epoch": 1035, "lr": 7.431343411187652e-05}
{"train_loss": 0.15520596504211426, "global_step": 92178, "epoch": 1035, "lr": 7.431292753098147e-05}
{"train_loss": 0.24539126455783844, "global_step": 92179, "epoch": 1035, "lr": 7.431242094681782e-05}
{"train_loss": 0.17132890224456787, "global_step": 92180, "epoch": 1035, "lr": 7.431191435938569e-05}
{"train_loss": 0.21251945197582245, "global_step": 92181, "epoch": 1035, "lr": 7.431140776868511e-05}
{"train_loss": 0.12912526726722717, "global_step": 92182, "epoch": 1035, "lr": 7.431090117471615e-05}
{"train_loss": 0.15132562816143036, "global_step": 92183, "epoch": 1035, "lr": 7.43103945774789e-05}
{"train_loss": 0.19168078899383545, "global_step": 92184, "epoch": 1035, "lr": 7.43098879769734e-05}
{"train_loss": 0.18183201551437378, "global_step": 92185, "epoch": 1035, "lr": 7.430938137319974e-05}
{"train_loss": 0.20115157961845398, "global_step": 92186, "epoch": 1035, "lr": 7.430887476615798e-05}
{"train_loss": 0.17751429975032806, "global_step": 92187, "epoch": 1035, "lr": 7.430836815584818e-05}
{"train_loss": 0.14447052776813507, "global_step": 92188, "epoch": 1035, "lr": 7.430786154227043e-05}
{"train_loss": 0.11899672448635101, "global_step": 92189, "epoch": 1035, "lr": 7.430735492542478e-05}
{"train_loss": 0.15077783167362213, "global_step": 92190, "epoch": 1035, "lr": 7.43068483053113e-05}
{"train_loss": 0.18564355373382568, "global_step": 92191, "epoch": 1035, "lr": 7.430634168193005e-05}
{"train_loss": 0.09254312515258789, "global_step": 92192, "epoch": 1035, "lr": 7.430583505528111e-05}
{"train_loss": 0.19802521169185638, "global_step": 92193, "epoch": 1035, "lr": 7.430532842536456e-05}
{"train_loss": 0.17287229001522064, "global_step": 92194, "epoch": 1035, "lr": 7.430482179218045e-05}
{"train_loss": 0.20196479558944702, "global_step": 92195, "epoch": 1035, "lr": 7.430431515572884e-05}
{"train_loss": 0.17047546803951263, "global_step": 92196, "epoch": 1035, "lr": 7.430380851600984e-05}
{"train_loss": 0.19163449108600616, "global_step": 92197, "epoch": 1035, "lr": 7.430330187302347e-05}
{"train_loss": 0.13566678762435913, "global_step": 92198, "epoch": 1035, "lr": 7.430279522676982e-05}
{"train_loss": 0.21188029646873474, "global_step": 92199, "epoch": 1035, "lr": 7.430228857724895e-05}
{"train_loss": 0.15582792460918427, "global_step": 92200, "epoch": 1035, "lr": 7.430178192446094e-05}
{"train_loss": 0.20289777219295502, "global_step": 92201, "epoch": 1035, "lr": 7.430127526840584e-05}
{"train_loss": 0.1325925588607788, "global_step": 92202, "epoch": 1035, "lr": 7.430076860908374e-05}
{"train_loss": 0.17903120179524581, "global_step": 92203, "epoch": 1035, "lr": 7.43002619464947e-05, "val_loss": 4.402312755584717, "train_action_mse_error": 36.10521697998047}
{"train_loss": 0.20353354513645172, "global_step": 92204, "epoch": 1036, "lr": 7.429975528063879e-05}
{"train_loss": 0.12487823516130447, "global_step": 92205, "epoch": 1036, "lr": 7.429924861151606e-05}
{"train_loss": 0.11480513960123062, "global_step": 92206, "epoch": 1036, "lr": 7.42987419391266e-05}
{"train_loss": 0.11065534502267838, "global_step": 92207, "epoch": 1036, "lr": 7.429823526347047e-05}
{"train_loss": 0.11997844278812408, "global_step": 92208, "epoch": 1036, "lr": 7.429772858454774e-05}
{"train_loss": 0.1283944696187973, "global_step": 92209, "epoch": 1036, "lr": 7.429722190235848e-05}
{"train_loss": 0.23112203180789948, "global_step": 92210, "epoch": 1036, "lr": 7.429671521690275e-05}
{"train_loss": 0.18408478796482086, "global_step": 92211, "epoch": 1036, "lr": 7.429620852818063e-05}
{"train_loss": 0.18588510155677795, "global_step": 92212, "epoch": 1036, "lr": 7.429570183619217e-05}
{"train_loss": 0.17160366475582123, "global_step": 92213, "epoch": 1036, "lr": 7.429519514093745e-05}
{"train_loss": 0.1631953865289688, "global_step": 92214, "epoch": 1036, "lr": 7.429468844241654e-05}
{"train_loss": 0.16381807625293732, "global_step": 92215, "epoch": 1036, "lr": 7.42941817406295e-05}
{"train_loss": 0.13453195989131927, "global_step": 92216, "epoch": 1036, "lr": 7.42936750355764e-05}
{"train_loss": 0.19150756299495697, "global_step": 92217, "epoch": 1036, "lr": 7.429316832725733e-05}
{"train_loss": 0.12664251029491425, "global_step": 92218, "epoch": 1036, "lr": 7.429266161567232e-05}
{"train_loss": 0.12783163785934448, "global_step": 92219, "epoch": 1036, "lr": 7.429215490082146e-05}
{"train_loss": 0.1296653151512146, "global_step": 92220, "epoch": 1036, "lr": 7.429164818270483e-05}
{"train_loss": 0.19155143201351166, "global_step": 92221, "epoch": 1036, "lr": 7.429114146132247e-05}
{"train_loss": 0.1331523209810257, "global_step": 92222, "epoch": 1036, "lr": 7.429063473667447e-05}
{"train_loss": 0.1659303456544876, "global_step": 92223, "epoch": 1036, "lr": 7.42901280087609e-05}
{"train_loss": 0.14470016956329346, "global_step": 92224, "epoch": 1036, "lr": 7.42896212775818e-05}
{"train_loss": 0.17664223909378052, "global_step": 92225, "epoch": 1036, "lr": 7.428911454313727e-05}
{"train_loss": 0.22393672168254852, "global_step": 92226, "epoch": 1036, "lr": 7.428860780542736e-05}
{"train_loss": 0.11226779967546463, "global_step": 92227, "epoch": 1036, "lr": 7.428810106445213e-05}
{"train_loss": 0.1913335919380188, "global_step": 92228, "epoch": 1036, "lr": 7.428759432021167e-05}
{"train_loss": 0.21326357126235962, "global_step": 92229, "epoch": 1036, "lr": 7.428708757270605e-05}
{"train_loss": 0.1490483433008194, "global_step": 92230, "epoch": 1036, "lr": 7.428658082193532e-05}
{"train_loss": 0.13284724950790405, "global_step": 92231, "epoch": 1036, "lr": 7.428607406789955e-05}
{"train_loss": 0.10297570377588272, "global_step": 92232, "epoch": 1036, "lr": 7.428556731059882e-05}
{"train_loss": 0.21335068345069885, "global_step": 92233, "epoch": 1036, "lr": 7.428506055003318e-05}
{"train_loss": 0.12264825403690338, "global_step": 92234, "epoch": 1036, "lr": 7.428455378620271e-05}
{"train_loss": 0.15290872752666473, "global_step": 92235, "epoch": 1036, "lr": 7.428404701910749e-05}
{"train_loss": 0.13935574889183044, "global_step": 92236, "epoch": 1036, "lr": 7.428354024874757e-05}
{"train_loss": 0.12681880593299866, "global_step": 92237, "epoch": 1036, "lr": 7.428303347512303e-05}
{"train_loss": 0.159264475107193, "global_step": 92238, "epoch": 1036, "lr": 7.428252669823393e-05}
{"train_loss": 0.1808566153049469, "global_step": 92239, "epoch": 1036, "lr": 7.428201991808033e-05}
{"train_loss": 0.13333803415298462, "global_step": 92240, "epoch": 1036, "lr": 7.428151313466231e-05}
{"train_loss": 0.1507674604654312, "global_step": 92241, "epoch": 1036, "lr": 7.428100634797994e-05}
{"train_loss": 0.21340996026992798, "global_step": 92242, "epoch": 1036, "lr": 7.42804995580333e-05}
{"train_loss": 0.15415827929973602, "global_step": 92243, "epoch": 1036, "lr": 7.427999276482243e-05}
{"train_loss": 0.17681600153446198, "global_step": 92244, "epoch": 1036, "lr": 7.427948596834742e-05}
{"train_loss": 0.16966204345226288, "global_step": 92245, "epoch": 1036, "lr": 7.42789791686083e-05}
{"train_loss": 0.1493123322725296, "global_step": 92246, "epoch": 1036, "lr": 7.42784723656052e-05}
{"train_loss": 0.1440841555595398, "global_step": 92247, "epoch": 1036, "lr": 7.427796555933815e-05}
{"train_loss": 0.0963084027171135, "global_step": 92248, "epoch": 1036, "lr": 7.42774587498072e-05}
{"train_loss": 0.18474626541137695, "global_step": 92249, "epoch": 1036, "lr": 7.427695193701246e-05}
{"train_loss": 0.1298503279685974, "global_step": 92250, "epoch": 1036, "lr": 7.427644512095398e-05}
{"train_loss": 0.2148740142583847, "global_step": 92251, "epoch": 1036, "lr": 7.427593830163182e-05}
{"train_loss": 0.18049609661102295, "global_step": 92252, "epoch": 1036, "lr": 7.427543147904606e-05}
{"train_loss": 0.10693571716547012, "global_step": 92253, "epoch": 1036, "lr": 7.427492465319677e-05}
{"train_loss": 0.14001184701919556, "global_step": 92254, "epoch": 1036, "lr": 7.427441782408401e-05}
{"train_loss": 0.17631752789020538, "global_step": 92255, "epoch": 1036, "lr": 7.427391099170784e-05}
{"train_loss": 0.1462850719690323, "global_step": 92256, "epoch": 1036, "lr": 7.427340415606835e-05}
{"train_loss": 0.19975146651268005, "global_step": 92257, "epoch": 1036, "lr": 7.42728973171656e-05}
{"train_loss": 0.16766257584095, "global_step": 92258, "epoch": 1036, "lr": 7.427239047499965e-05}
{"train_loss": 0.11979006975889206, "global_step": 92259, "epoch": 1036, "lr": 7.427188362957055e-05}
{"train_loss": 0.2293553203344345, "global_step": 92260, "epoch": 1036, "lr": 7.427137678087841e-05}
{"train_loss": 0.14228084683418274, "global_step": 92261, "epoch": 1036, "lr": 7.42708699289233e-05}
{"train_loss": 0.21679989993572235, "global_step": 92262, "epoch": 1036, "lr": 7.427036307370523e-05}
{"train_loss": 0.21360182762145996, "global_step": 92263, "epoch": 1036, "lr": 7.426985621522433e-05}
{"train_loss": 0.1440713256597519, "global_step": 92264, "epoch": 1036, "lr": 7.426934935348065e-05}
{"train_loss": 0.22544041275978088, "global_step": 92265, "epoch": 1036, "lr": 7.426884248847424e-05}
{"train_loss": 0.18037891387939453, "global_step": 92266, "epoch": 1036, "lr": 7.426833562020516e-05}
{"train_loss": 0.16757330298423767, "global_step": 92267, "epoch": 1036, "lr": 7.426782874867353e-05}
{"train_loss": 0.17780888080596924, "global_step": 92268, "epoch": 1036, "lr": 7.426732187387937e-05}
{"train_loss": 0.15072371065616608, "global_step": 92269, "epoch": 1036, "lr": 7.426681499582277e-05}
{"train_loss": 0.18140390515327454, "global_step": 92270, "epoch": 1036, "lr": 7.42663081145038e-05}
{"train_loss": 0.231644868850708, "global_step": 92271, "epoch": 1036, "lr": 7.42658012299225e-05}
{"train_loss": 0.13197478652000427, "global_step": 92272, "epoch": 1036, "lr": 7.426529434207896e-05}
{"train_loss": 0.1555715650320053, "global_step": 92273, "epoch": 1036, "lr": 7.426478745097326e-05}
{"train_loss": 0.14853636920452118, "global_step": 92274, "epoch": 1036, "lr": 7.426428055660546e-05}
{"train_loss": 0.15484870970249176, "global_step": 92275, "epoch": 1036, "lr": 7.42637736589756e-05}
{"train_loss": 0.23630118370056152, "global_step": 92276, "epoch": 1036, "lr": 7.426326675808379e-05}
{"train_loss": 0.1683567315340042, "global_step": 92277, "epoch": 1036, "lr": 7.426275985393008e-05}
{"train_loss": 0.2228616178035736, "global_step": 92278, "epoch": 1036, "lr": 7.426225294651452e-05}
{"train_loss": 0.1545337438583374, "global_step": 92279, "epoch": 1036, "lr": 7.426174603583722e-05}
{"train_loss": 0.22156985104084015, "global_step": 92280, "epoch": 1036, "lr": 7.426123912189821e-05}
{"train_loss": 0.21158862113952637, "global_step": 92281, "epoch": 1036, "lr": 7.426073220469759e-05}
{"train_loss": 0.13491316139698029, "global_step": 92282, "epoch": 1036, "lr": 7.426022528423538e-05}
{"train_loss": 0.18767645955085754, "global_step": 92283, "epoch": 1036, "lr": 7.42597183605117e-05}
{"train_loss": 0.15711286664009094, "global_step": 92284, "epoch": 1036, "lr": 7.425921143352659e-05}
{"train_loss": 0.13337795436382294, "global_step": 92285, "epoch": 1036, "lr": 7.425870450328013e-05}
{"train_loss": 0.15404564142227173, "global_step": 92286, "epoch": 1036, "lr": 7.425819756977237e-05}
{"train_loss": 0.1891757845878601, "global_step": 92287, "epoch": 1036, "lr": 7.425769063300341e-05}
{"train_loss": 0.17117884755134583, "global_step": 92288, "epoch": 1036, "lr": 7.425718369297327e-05}
{"train_loss": 0.21438899636268616, "global_step": 92289, "epoch": 1036, "lr": 7.425667674968208e-05}
{"train_loss": 0.13376259803771973, "global_step": 92290, "epoch": 1036, "lr": 7.425616980312987e-05}
{"train_loss": 0.2149253487586975, "global_step": 92291, "epoch": 1036, "lr": 7.425566285331668e-05}
{"train_loss": 0.16521234241094482, "global_step": 92292, "epoch": 1036, "lr": 7.425515590024264e-05, "val_loss": 4.298839092254639}
{"train_loss": 0.19630053639411926, "global_step": 92293, "epoch": 1037, "lr": 7.425464894390779e-05}
{"train_loss": 0.1775617003440857, "global_step": 92294, "epoch": 1037, "lr": 7.42541419843122e-05}
{"train_loss": 0.1965133249759674, "global_step": 92295, "epoch": 1037, "lr": 7.425363502145592e-05}
{"train_loss": 0.11807066947221756, "global_step": 92296, "epoch": 1037, "lr": 7.425312805533903e-05}
{"train_loss": 0.17168454825878143, "global_step": 92297, "epoch": 1037, "lr": 7.425262108596163e-05}
{"train_loss": 0.12176366150379181, "global_step": 92298, "epoch": 1037, "lr": 7.425211411332375e-05}
{"train_loss": 0.24839000403881073, "global_step": 92299, "epoch": 1037, "lr": 7.425160713742548e-05}
{"train_loss": 0.1479654759168625, "global_step": 92300, "epoch": 1037, "lr": 7.425110015826686e-05}
{"train_loss": 0.11149509996175766, "global_step": 92301, "epoch": 1037, "lr": 7.425059317584797e-05}
{"train_loss": 0.14151373505592346, "global_step": 92302, "epoch": 1037, "lr": 7.42500861901689e-05}
{"train_loss": 0.1508374810218811, "global_step": 92303, "epoch": 1037, "lr": 7.424957920122969e-05}
{"train_loss": 0.13670803606510162, "global_step": 92304, "epoch": 1037, "lr": 7.424907220903042e-05}
{"train_loss": 0.21166688203811646, "global_step": 92305, "epoch": 1037, "lr": 7.424856521357116e-05}
{"train_loss": 0.12321358919143677, "global_step": 92306, "epoch": 1037, "lr": 7.424805821485199e-05}
{"train_loss": 0.17415514588356018, "global_step": 92307, "epoch": 1037, "lr": 7.424755121287295e-05}
{"train_loss": 0.14224936068058014, "global_step": 92308, "epoch": 1037, "lr": 7.424704420763412e-05}
{"train_loss": 0.1460474133491516, "global_step": 92309, "epoch": 1037, "lr": 7.424653719913556e-05}
{"train_loss": 0.15179042518138885, "global_step": 92310, "epoch": 1037, "lr": 7.424603018737739e-05}
{"train_loss": 0.23936232924461365, "global_step": 92311, "epoch": 1037, "lr": 7.424552317235959e-05}
{"train_loss": 0.18173642456531525, "global_step": 92312, "epoch": 1037, "lr": 7.424501615408231e-05}
{"train_loss": 0.2221129834651947, "global_step": 92313, "epoch": 1037, "lr": 7.424450913254555e-05}
{"train_loss": 0.1981872320175171, "global_step": 92314, "epoch": 1037, "lr": 7.424400210774942e-05}
{"train_loss": 0.26469072699546814, "global_step": 92315, "epoch": 1037, "lr": 7.4243495079694e-05}
{"train_loss": 0.09807974100112915, "global_step": 92316, "epoch": 1037, "lr": 7.424298804837933e-05}
{"train_loss": 0.18318259716033936, "global_step": 92317, "epoch": 1037, "lr": 7.424248101380546e-05}
{"train_loss": 0.2087867110967636, "global_step": 92318, "epoch": 1037, "lr": 7.424197397597253e-05}
{"train_loss": 0.23159874975681305, "global_step": 92319, "epoch": 1037, "lr": 7.424146693488053e-05}
{"train_loss": 0.1931983083486557, "global_step": 92320, "epoch": 1037, "lr": 7.424095989052958e-05}
{"train_loss": 0.19381007552146912, "global_step": 92321, "epoch": 1037, "lr": 7.42404528429197e-05}
{"train_loss": 0.14897973835468292, "global_step": 92322, "epoch": 1037, "lr": 7.423994579205102e-05}
{"train_loss": 0.22645631432533264, "global_step": 92323, "epoch": 1037, "lr": 7.423943873792355e-05}
{"train_loss": 0.19105710089206696, "global_step": 92324, "epoch": 1037, "lr": 7.423893168053738e-05}
{"train_loss": 0.1693013608455658, "global_step": 92325, "epoch": 1037, "lr": 7.42384246198926e-05}
{"train_loss": 0.1942826509475708, "global_step": 92326, "epoch": 1037, "lr": 7.423791755598925e-05}
{"train_loss": 0.15474143624305725, "global_step": 92327, "epoch": 1037, "lr": 7.42374104888274e-05}
{"train_loss": 0.15036612749099731, "global_step": 92328, "epoch": 1037, "lr": 7.423690341840714e-05}
{"train_loss": 0.17488786578178406, "global_step": 92329, "epoch": 1037, "lr": 7.423639634472852e-05}
{"train_loss": 0.2148950845003128, "global_step": 92330, "epoch": 1037, "lr": 7.42358892677916e-05}
{"train_loss": 0.14937424659729004, "global_step": 92331, "epoch": 1037, "lr": 7.423538218759647e-05}
{"train_loss": 0.13726797699928284, "global_step": 92332, "epoch": 1037, "lr": 7.423487510414317e-05}
{"train_loss": 0.18916800618171692, "global_step": 92333, "epoch": 1037, "lr": 7.423436801743182e-05}
{"train_loss": 0.3055608570575714, "global_step": 92334, "epoch": 1037, "lr": 7.423386092746242e-05}
{"train_loss": 0.15847322344779968, "global_step": 92335, "epoch": 1037, "lr": 7.423335383423511e-05}
{"train_loss": 0.12775543332099915, "global_step": 92336, "epoch": 1037, "lr": 7.423284673774989e-05}
{"train_loss": 0.19138787686824799, "global_step": 92337, "epoch": 1037, "lr": 7.423233963800688e-05}
{"train_loss": 0.14960701763629913, "global_step": 92338, "epoch": 1037, "lr": 7.423183253500612e-05}
{"train_loss": 0.22409456968307495, "global_step": 92339, "epoch": 1037, "lr": 7.423132542874769e-05}
{"train_loss": 0.13878439366817474, "global_step": 92340, "epoch": 1037, "lr": 7.423081831923164e-05}
{"train_loss": 0.14328600466251373, "global_step": 92341, "epoch": 1037, "lr": 7.423031120645806e-05}
{"train_loss": 0.20940712094306946, "global_step": 92342, "epoch": 1037, "lr": 7.422980409042701e-05}
{"train_loss": 0.2135818898677826, "global_step": 92343, "epoch": 1037, "lr": 7.422929697113857e-05}
{"train_loss": 0.2914191484451294, "global_step": 92344, "epoch": 1037, "lr": 7.422878984859279e-05}
{"train_loss": 0.12105967104434967, "global_step": 92345, "epoch": 1037, "lr": 7.422828272278973e-05}
{"train_loss": 0.2113339602947235, "global_step": 92346, "epoch": 1037, "lr": 7.422777559372948e-05}
{"train_loss": 0.2676490545272827, "global_step": 92347, "epoch": 1037, "lr": 7.42272684614121e-05}
{"train_loss": 0.17765077948570251, "global_step": 92348, "epoch": 1037, "lr": 7.422676132583766e-05}
{"train_loss": 0.15703937411308289, "global_step": 92349, "epoch": 1037, "lr": 7.422625418700622e-05}
{"train_loss": 0.21432997286319733, "global_step": 92350, "epoch": 1037, "lr": 7.422574704491786e-05}
{"train_loss": 0.1933305710554123, "global_step": 92351, "epoch": 1037, "lr": 7.422523989957265e-05}
{"train_loss": 0.172955721616745, "global_step": 92352, "epoch": 1037, "lr": 7.422473275097064e-05}
{"train_loss": 0.15541145205497742, "global_step": 92353, "epoch": 1037, "lr": 7.422422559911193e-05}
{"train_loss": 0.19836293160915375, "global_step": 92354, "epoch": 1037, "lr": 7.422371844399655e-05}
{"train_loss": 0.20396281778812408, "global_step": 92355, "epoch": 1037, "lr": 7.42232112856246e-05}
{"train_loss": 0.22266632318496704, "global_step": 92356, "epoch": 1037, "lr": 7.422270412399612e-05}
{"train_loss": 0.19729940593242645, "global_step": 92357, "epoch": 1037, "lr": 7.42221969591112e-05}
{"train_loss": 0.17805105447769165, "global_step": 92358, "epoch": 1037, "lr": 7.422168979096991e-05}
{"train_loss": 0.183069109916687, "global_step": 92359, "epoch": 1037, "lr": 7.42211826195723e-05}
{"train_loss": 0.2296266406774521, "global_step": 92360, "epoch": 1037, "lr": 7.422067544491845e-05}
{"train_loss": 0.1887151300907135, "global_step": 92361, "epoch": 1037, "lr": 7.422016826700843e-05}
{"train_loss": 0.12627504765987396, "global_step": 92362, "epoch": 1037, "lr": 7.42196610858423e-05}
{"train_loss": 0.23771578073501587, "global_step": 92363, "epoch": 1037, "lr": 7.421915390142012e-05}
{"train_loss": 0.179277241230011, "global_step": 92364, "epoch": 1037, "lr": 7.421864671374198e-05}
{"train_loss": 0.23841921985149384, "global_step": 92365, "epoch": 1037, "lr": 7.421813952280794e-05}
{"train_loss": 0.18817414343357086, "global_step": 92366, "epoch": 1037, "lr": 7.421763232861805e-05}
{"train_loss": 0.23908597230911255, "global_step": 92367, "epoch": 1037, "lr": 7.42171251311724e-05}
{"train_loss": 0.16757351160049438, "global_step": 92368, "epoch": 1037, "lr": 7.421661793047107e-05}
{"train_loss": 0.2628437876701355, "global_step": 92369, "epoch": 1037, "lr": 7.42161107265141e-05}
{"train_loss": 0.16559956967830658, "global_step": 92370, "epoch": 1037, "lr": 7.421560351930157e-05}
{"train_loss": 0.207500159740448, "global_step": 92371, "epoch": 1037, "lr": 7.421509630883354e-05}
{"train_loss": 0.14903753995895386, "global_step": 92372, "epoch": 1037, "lr": 7.42145890951101e-05}
{"train_loss": 0.1615259200334549, "global_step": 92373, "epoch": 1037, "lr": 7.421408187813129e-05}
{"train_loss": 0.21121488511562347, "global_step": 92374, "epoch": 1037, "lr": 7.42135746578972e-05}
{"train_loss": 0.19167615473270416, "global_step": 92375, "epoch": 1037, "lr": 7.421306743440789e-05}
{"train_loss": 0.1464710682630539, "global_step": 92376, "epoch": 1037, "lr": 7.421256020766343e-05}
{"train_loss": 0.20353780686855316, "global_step": 92377, "epoch": 1037, "lr": 7.42120529776639e-05}
{"train_loss": 0.22783592343330383, "global_step": 92378, "epoch": 1037, "lr": 7.421154574440932e-05}
{"train_loss": 0.2424992173910141, "global_step": 92379, "epoch": 1037, "lr": 7.421103850789982e-05}
{"train_loss": 0.20942337810993195, "global_step": 92380, "epoch": 1037, "lr": 7.421053126813544e-05}
{"train_loss": 0.1860488040393658, "global_step": 92381, "epoch": 1037, "lr": 7.421002402511623e-05, "val_loss": 4.215951442718506}
{"train_loss": 0.18708869814872742, "global_step": 92382, "epoch": 1038, "lr": 7.420951677884229e-05}
{"train_loss": 0.17821525037288666, "global_step": 92383, "epoch": 1038, "lr": 7.420900952931367e-05}
{"train_loss": 0.20167414844036102, "global_step": 92384, "epoch": 1038, "lr": 7.420850227653044e-05}
{"train_loss": 0.1999935656785965, "global_step": 92385, "epoch": 1038, "lr": 7.420799502049268e-05}
{"train_loss": 0.2464529573917389, "global_step": 92386, "epoch": 1038, "lr": 7.420748776120045e-05}
{"train_loss": 0.17587552964687347, "global_step": 92387, "epoch": 1038, "lr": 7.420698049865381e-05}
{"train_loss": 0.12704825401306152, "global_step": 92388, "epoch": 1038, "lr": 7.420647323285283e-05}
{"train_loss": 0.18370403349399567, "global_step": 92389, "epoch": 1038, "lr": 7.420596596379761e-05}
{"train_loss": 0.10461359471082687, "global_step": 92390, "epoch": 1038, "lr": 7.420545869148818e-05}
{"train_loss": 0.15929828584194183, "global_step": 92391, "epoch": 1038, "lr": 7.420495141592462e-05}
{"train_loss": 0.23534272611141205, "global_step": 92392, "epoch": 1038, "lr": 7.420444413710701e-05}
{"train_loss": 0.18571045994758606, "global_step": 92393, "epoch": 1038, "lr": 7.42039368550354e-05}
{"train_loss": 0.14146660268306732, "global_step": 92394, "epoch": 1038, "lr": 7.420342956970986e-05}
{"train_loss": 0.1433078944683075, "global_step": 92395, "epoch": 1038, "lr": 7.420292228113047e-05}
{"train_loss": 0.2047998160123825, "global_step": 92396, "epoch": 1038, "lr": 7.420241498929729e-05}
{"train_loss": 0.1574297398328781, "global_step": 92397, "epoch": 1038, "lr": 7.42019076942104e-05}
{"train_loss": 0.15782952308654785, "global_step": 92398, "epoch": 1038, "lr": 7.420140039586985e-05}
{"train_loss": 0.23503883183002472, "global_step": 92399, "epoch": 1038, "lr": 7.420089309427571e-05}
{"train_loss": 0.10328290611505508, "global_step": 92400, "epoch": 1038, "lr": 7.420038578942806e-05}
{"train_loss": 0.13907311856746674, "global_step": 92401, "epoch": 1038, "lr": 7.419987848132696e-05}
{"train_loss": 0.13273434340953827, "global_step": 92402, "epoch": 1038, "lr": 7.41993711699725e-05}
{"train_loss": 0.18490324914455414, "global_step": 92403, "epoch": 1038, "lr": 7.41988638553647e-05}
{"train_loss": 0.2652621865272522, "global_step": 92404, "epoch": 1038, "lr": 7.419835653750367e-05}
{"train_loss": 0.12132900953292847, "global_step": 92405, "epoch": 1038, "lr": 7.419784921638946e-05}
{"train_loss": 0.15325377881526947, "global_step": 92406, "epoch": 1038, "lr": 7.419734189202216e-05}
{"train_loss": 0.13384583592414856, "global_step": 92407, "epoch": 1038, "lr": 7.419683456440182e-05}
{"train_loss": 0.1674555540084839, "global_step": 92408, "epoch": 1038, "lr": 7.41963272335285e-05}
{"train_loss": 0.11061950027942657, "global_step": 92409, "epoch": 1038, "lr": 7.419581989940228e-05}
{"train_loss": 0.1415347009897232, "global_step": 92410, "epoch": 1038, "lr": 7.419531256202325e-05}
{"train_loss": 0.17474080622196198, "global_step": 92411, "epoch": 1038, "lr": 7.419480522139142e-05}
{"train_loss": 0.1686466634273529, "global_step": 92412, "epoch": 1038, "lr": 7.419429787750691e-05}
{"train_loss": 0.17269931733608246, "global_step": 92413, "epoch": 1038, "lr": 7.419379053036978e-05}
{"train_loss": 0.2103874683380127, "global_step": 92414, "epoch": 1038, "lr": 7.419328317998007e-05}
{"train_loss": 0.17743772268295288, "global_step": 92415, "epoch": 1038, "lr": 7.419277582633789e-05}
{"train_loss": 0.25465694069862366, "global_step": 92416, "epoch": 1038, "lr": 7.419226846944329e-05}
{"train_loss": 0.20187492668628693, "global_step": 92417, "epoch": 1038, "lr": 7.419176110929631e-05}
{"train_loss": 0.1280214488506317, "global_step": 92418, "epoch": 1038, "lr": 7.419125374589706e-05}
{"train_loss": 0.1992107480764389, "global_step": 92419, "epoch": 1038, "lr": 7.419074637924558e-05}
{"train_loss": 0.1620381623506546, "global_step": 92420, "epoch": 1038, "lr": 7.419023900934195e-05}
{"train_loss": 0.14444339275360107, "global_step": 92421, "epoch": 1038, "lr": 7.418973163618624e-05}
{"train_loss": 0.19744279980659485, "global_step": 92422, "epoch": 1038, "lr": 7.41892242597785e-05}
{"train_loss": 0.19539906084537506, "global_step": 92423, "epoch": 1038, "lr": 7.418871688011884e-05}
{"train_loss": 0.2312469184398651, "global_step": 92424, "epoch": 1038, "lr": 7.418820949720729e-05}
{"train_loss": 0.20576772093772888, "global_step": 92425, "epoch": 1038, "lr": 7.418770211104392e-05}
{"train_loss": 0.14399008452892303, "global_step": 92426, "epoch": 1038, "lr": 7.418719472162882e-05}
{"train_loss": 0.22879627346992493, "global_step": 92427, "epoch": 1038, "lr": 7.418668732896204e-05}
{"train_loss": 0.20326630771160126, "global_step": 92428, "epoch": 1038, "lr": 7.418617993304367e-05}
{"train_loss": 0.13501903414726257, "global_step": 92429, "epoch": 1038, "lr": 7.418567253387373e-05}
{"train_loss": 0.1027296632528305, "global_step": 92430, "epoch": 1038, "lr": 7.418516513145234e-05}
{"train_loss": 0.1290132850408554, "global_step": 92431, "epoch": 1038, "lr": 7.418465772577957e-05}
{"train_loss": 0.16290996968746185, "global_step": 92432, "epoch": 1038, "lr": 7.418415031685545e-05}
{"train_loss": 0.2083224356174469, "global_step": 92433, "epoch": 1038, "lr": 7.418364290468006e-05}
{"train_loss": 0.1912887841463089, "global_step": 92434, "epoch": 1038, "lr": 7.418313548925349e-05}
{"train_loss": 0.10971839725971222, "global_step": 92435, "epoch": 1038, "lr": 7.418262807057579e-05}
{"train_loss": 0.15194886922836304, "global_step": 92436, "epoch": 1038, "lr": 7.418212064864703e-05}
{"train_loss": 0.21478241682052612, "global_step": 92437, "epoch": 1038, "lr": 7.418161322346726e-05}
{"train_loss": 0.20362454652786255, "global_step": 92438, "epoch": 1038, "lr": 7.418110579503658e-05}
{"train_loss": 0.19953131675720215, "global_step": 92439, "epoch": 1038, "lr": 7.418059836335505e-05}
{"train_loss": 0.1286408007144928, "global_step": 92440, "epoch": 1038, "lr": 7.418009092842271e-05}
{"train_loss": 0.16467192769050598, "global_step": 92441, "epoch": 1038, "lr": 7.417958349023967e-05}
{"train_loss": 0.16628210246562958, "global_step": 92442, "epoch": 1038, "lr": 7.417907604880597e-05}
{"train_loss": 0.14788062870502472, "global_step": 92443, "epoch": 1038, "lr": 7.41785686041217e-05}
{"train_loss": 0.24153858423233032, "global_step": 92444, "epoch": 1038, "lr": 7.41780611561869e-05}
{"train_loss": 0.18319734930992126, "global_step": 92445, "epoch": 1038, "lr": 7.417755370500168e-05}
{"train_loss": 0.1809692233800888, "global_step": 92446, "epoch": 1038, "lr": 7.417704625056606e-05}
{"train_loss": 0.1351321041584015, "global_step": 92447, "epoch": 1038, "lr": 7.417653879288015e-05}
{"train_loss": 0.15090841054916382, "global_step": 92448, "epoch": 1038, "lr": 7.417603133194397e-05}
{"train_loss": 0.2101842314004898, "global_step": 92449, "epoch": 1038, "lr": 7.417552386775765e-05}
{"train_loss": 0.18569296598434448, "global_step": 92450, "epoch": 1038, "lr": 7.41750164003212e-05}
{"train_loss": 0.20116066932678223, "global_step": 92451, "epoch": 1038, "lr": 7.417450892963473e-05}
{"train_loss": 0.23675371706485748, "global_step": 92452, "epoch": 1038, "lr": 7.417400145569829e-05}
{"train_loss": 0.24302765727043152, "global_step": 92453, "epoch": 1038, "lr": 7.417349397851194e-05}
{"train_loss": 0.18705245852470398, "global_step": 92454, "epoch": 1038, "lr": 7.417298649807577e-05}
{"train_loss": 0.16436336934566498, "global_step": 92455, "epoch": 1038, "lr": 7.417247901438983e-05}
{"train_loss": 0.2063918262720108, "global_step": 92456, "epoch": 1038, "lr": 7.417197152745421e-05}
{"train_loss": 0.12305369973182678, "global_step": 92457, "epoch": 1038, "lr": 7.417146403726894e-05}
{"train_loss": 0.14032964408397675, "global_step": 92458, "epoch": 1038, "lr": 7.417095654383412e-05}
{"train_loss": 0.15270571410655975, "global_step": 92459, "epoch": 1038, "lr": 7.417044904714981e-05}
{"train_loss": 0.1921457201242447, "global_step": 92460, "epoch": 1038, "lr": 7.416994154721608e-05}
{"train_loss": 0.15169654786586761, "global_step": 92461, "epoch": 1038, "lr": 7.416943404403299e-05}
{"train_loss": 0.20852148532867432, "global_step": 92462, "epoch": 1038, "lr": 7.416892653760062e-05}
{"train_loss": 0.1542254388332367, "global_step": 92463, "epoch": 1038, "lr": 7.416841902791904e-05}
{"train_loss": 0.16516980528831482, "global_step": 92464, "epoch": 1038, "lr": 7.41679115149883e-05}
{"train_loss": 0.11537177860736847, "global_step": 92465, "epoch": 1038, "lr": 7.416740399880847e-05}
{"train_loss": 0.14179423451423645, "global_step": 92466, "epoch": 1038, "lr": 7.416689647937964e-05}
{"train_loss": 0.20836378633975983, "global_step": 92467, "epoch": 1038, "lr": 7.416638895670186e-05}
{"train_loss": 0.12653344869613647, "global_step": 92468, "epoch": 1038, "lr": 7.41658814307752e-05}
{"train_loss": 0.20569995045661926, "global_step": 92469, "epoch": 1038, "lr": 7.416537390159974e-05}
{"train_loss": 0.17402280431784942, "global_step": 92470, "epoch": 1038, "lr": 7.416486636917554e-05, "val_loss": 4.472131252288818}
{"train_loss": 0.2372761368751526, "global_step": 92471, "epoch": 1039, "lr": 7.416435883350267e-05}
{"train_loss": 0.13290716707706451, "global_step": 92472, "epoch": 1039, "lr": 7.416385129458119e-05}
{"train_loss": 0.18475812673568726, "global_step": 92473, "epoch": 1039, "lr": 7.416334375241118e-05}
{"train_loss": 0.1966957449913025, "global_step": 92474, "epoch": 1039, "lr": 7.416283620699271e-05}
{"train_loss": 0.1859586536884308, "global_step": 92475, "epoch": 1039, "lr": 7.416232865832583e-05}
{"train_loss": 0.22053587436676025, "global_step": 92476, "epoch": 1039, "lr": 7.416182110641062e-05}
{"train_loss": 0.20290662348270416, "global_step": 92477, "epoch": 1039, "lr": 7.416131355124714e-05}
{"train_loss": 0.1524854153394699, "global_step": 92478, "epoch": 1039, "lr": 7.416080599283548e-05}
{"train_loss": 0.11947229504585266, "global_step": 92479, "epoch": 1039, "lr": 7.416029843117568e-05}
{"train_loss": 0.24284908175468445, "global_step": 92480, "epoch": 1039, "lr": 7.415979086626783e-05}
{"train_loss": 0.23932567238807678, "global_step": 92481, "epoch": 1039, "lr": 7.415928329811199e-05}
{"train_loss": 0.2732754051685333, "global_step": 92482, "epoch": 1039, "lr": 7.415877572670824e-05}
{"train_loss": 0.22328568994998932, "global_step": 92483, "epoch": 1039, "lr": 7.415826815205662e-05}
{"train_loss": 0.1532888412475586, "global_step": 92484, "epoch": 1039, "lr": 7.415776057415722e-05}
{"train_loss": 0.17714594304561615, "global_step": 92485, "epoch": 1039, "lr": 7.41572529930101e-05}
{"train_loss": 0.16745682060718536, "global_step": 92486, "epoch": 1039, "lr": 7.415674540861535e-05}
{"train_loss": 0.21324743330478668, "global_step": 92487, "epoch": 1039, "lr": 7.415623782097301e-05}
{"train_loss": 0.2617396116256714, "global_step": 92488, "epoch": 1039, "lr": 7.415573023008316e-05}
{"train_loss": 0.2531454563140869, "global_step": 92489, "epoch": 1039, "lr": 7.415522263594585e-05}
{"train_loss": 0.1422513723373413, "global_step": 92490, "epoch": 1039, "lr": 7.415471503856118e-05}
{"train_loss": 0.2736108601093292, "global_step": 92491, "epoch": 1039, "lr": 7.41542074379292e-05}
{"train_loss": 0.24821233749389648, "global_step": 92492, "epoch": 1039, "lr": 7.415369983404998e-05}
{"train_loss": 0.164058119058609, "global_step": 92493, "epoch": 1039, "lr": 7.415319222692361e-05}
{"train_loss": 0.1565447449684143, "global_step": 92494, "epoch": 1039, "lr": 7.41526846165501e-05}
{"train_loss": 0.11805706471204758, "global_step": 92495, "epoch": 1039, "lr": 7.415217700292958e-05}
{"train_loss": 0.1780557930469513, "global_step": 92496, "epoch": 1039, "lr": 7.415166938606208e-05}
{"train_loss": 0.20356354117393494, "global_step": 92497, "epoch": 1039, "lr": 7.41511617659477e-05}
{"train_loss": 0.155966117978096, "global_step": 92498, "epoch": 1039, "lr": 7.415065414258648e-05}
{"train_loss": 0.2357347458600998, "global_step": 92499, "epoch": 1039, "lr": 7.415014651597851e-05}
{"train_loss": 0.17913353443145752, "global_step": 92500, "epoch": 1039, "lr": 7.414963888612383e-05}
{"train_loss": 0.15443584322929382, "global_step": 92501, "epoch": 1039, "lr": 7.414913125302253e-05}
{"train_loss": 0.14498630166053772, "global_step": 92502, "epoch": 1039, "lr": 7.414862361667467e-05}
{"train_loss": 0.1374732106924057, "global_step": 92503, "epoch": 1039, "lr": 7.414811597708034e-05}
{"train_loss": 0.19520100951194763, "global_step": 92504, "epoch": 1039, "lr": 7.414760833423957e-05}
{"train_loss": 0.23325400054454803, "global_step": 92505, "epoch": 1039, "lr": 7.414710068815246e-05}
{"train_loss": 0.15659889578819275, "global_step": 92506, "epoch": 1039, "lr": 7.414659303881907e-05}
{"train_loss": 0.18585488200187683, "global_step": 92507, "epoch": 1039, "lr": 7.414608538623947e-05}
{"train_loss": 0.08958940207958221, "global_step": 92508, "epoch": 1039, "lr": 7.41455777304137e-05}
{"train_loss": 0.1578318178653717, "global_step": 92509, "epoch": 1039, "lr": 7.414507007134187e-05}
{"train_loss": 0.17713963985443115, "global_step": 92510, "epoch": 1039, "lr": 7.414456240902402e-05}
{"train_loss": 0.1100417748093605, "global_step": 92511, "epoch": 1039, "lr": 7.414405474346023e-05}
{"train_loss": 0.15614928305149078, "global_step": 92512, "epoch": 1039, "lr": 7.414354707465057e-05}
{"train_loss": 0.22279547154903412, "global_step": 92513, "epoch": 1039, "lr": 7.414303940259509e-05}
{"train_loss": 0.1712711751461029, "global_step": 92514, "epoch": 1039, "lr": 7.414253172729388e-05}
{"train_loss": 0.17232397198677063, "global_step": 92515, "epoch": 1039, "lr": 7.414202404874701e-05}
{"train_loss": 0.14706367254257202, "global_step": 92516, "epoch": 1039, "lr": 7.414151636695453e-05}
{"train_loss": 0.18453602492809296, "global_step": 92517, "epoch": 1039, "lr": 7.414100868191652e-05}
{"train_loss": 0.15229107439517975, "global_step": 92518, "epoch": 1039, "lr": 7.414050099363306e-05}
{"train_loss": 0.14916454255580902, "global_step": 92519, "epoch": 1039, "lr": 7.413999330210419e-05}
{"train_loss": 0.10229634493589401, "global_step": 92520, "epoch": 1039, "lr": 7.413948560732999e-05}
{"train_loss": 0.104802705347538, "global_step": 92521, "epoch": 1039, "lr": 7.413897790931052e-05}
{"train_loss": 0.20042064785957336, "global_step": 92522, "epoch": 1039, "lr": 7.413847020804588e-05}
{"train_loss": 0.18611876666545868, "global_step": 92523, "epoch": 1039, "lr": 7.413796250353611e-05}
{"train_loss": 0.19958175718784332, "global_step": 92524, "epoch": 1039, "lr": 7.413745479578129e-05}
{"train_loss": 0.1308615505695343, "global_step": 92525, "epoch": 1039, "lr": 7.413694708478147e-05}
{"train_loss": 0.11769679188728333, "global_step": 92526, "epoch": 1039, "lr": 7.413643937053674e-05}
{"train_loss": 0.22875283658504486, "global_step": 92527, "epoch": 1039, "lr": 7.413593165304717e-05}
{"train_loss": 0.16344724595546722, "global_step": 92528, "epoch": 1039, "lr": 7.413542393231279e-05}
{"train_loss": 0.19195018708705902, "global_step": 92529, "epoch": 1039, "lr": 7.413491620833372e-05}
{"train_loss": 0.11340982466936111, "global_step": 92530, "epoch": 1039, "lr": 7.413440848111e-05}
{"train_loss": 0.1592537760734558, "global_step": 92531, "epoch": 1039, "lr": 7.413390075064171e-05}
{"train_loss": 0.19730186462402344, "global_step": 92532, "epoch": 1039, "lr": 7.41333930169289e-05}
{"train_loss": 0.197982519865036, "global_step": 92533, "epoch": 1039, "lr": 7.413288527997165e-05}
{"train_loss": 0.14289695024490356, "global_step": 92534, "epoch": 1039, "lr": 7.413237753977002e-05}
{"train_loss": 0.1282171607017517, "global_step": 92535, "epoch": 1039, "lr": 7.413186979632409e-05}
{"train_loss": 0.158443421125412, "global_step": 92536, "epoch": 1039, "lr": 7.413136204963394e-05}
{"train_loss": 0.16019828617572784, "global_step": 92537, "epoch": 1039, "lr": 7.413085429969961e-05}
{"train_loss": 0.21782492101192474, "global_step": 92538, "epoch": 1039, "lr": 7.413034654652119e-05}
{"train_loss": 0.1669558733701706, "global_step": 92539, "epoch": 1039, "lr": 7.412983879009873e-05}
{"train_loss": 0.12869970500469208, "global_step": 92540, "epoch": 1039, "lr": 7.412933103043231e-05}
{"train_loss": 0.20420394837856293, "global_step": 92541, "epoch": 1039, "lr": 7.412882326752199e-05}
{"train_loss": 0.12083704769611359, "global_step": 92542, "epoch": 1039, "lr": 7.412831550136785e-05}
{"train_loss": 0.13827462494373322, "global_step": 92543, "epoch": 1039, "lr": 7.412780773196995e-05}
{"train_loss": 0.1959087997674942, "global_step": 92544, "epoch": 1039, "lr": 7.412729995932836e-05}
{"train_loss": 0.10095752775669098, "global_step": 92545, "epoch": 1039, "lr": 7.412679218344315e-05}
{"train_loss": 0.16249696910381317, "global_step": 92546, "epoch": 1039, "lr": 7.412628440431438e-05}
{"train_loss": 0.19817012548446655, "global_step": 92547, "epoch": 1039, "lr": 7.412577662194213e-05}
{"train_loss": 0.12617142498493195, "global_step": 92548, "epoch": 1039, "lr": 7.412526883632646e-05}
{"train_loss": 0.12148630619049072, "global_step": 92549, "epoch": 1039, "lr": 7.412476104746746e-05}
{"train_loss": 0.2094586342573166, "global_step": 92550, "epoch": 1039, "lr": 7.412425325536516e-05}
{"train_loss": 0.1868792176246643, "global_step": 92551, "epoch": 1039, "lr": 7.412374546001965e-05}
{"train_loss": 0.16000838577747345, "global_step": 92552, "epoch": 1039, "lr": 7.412323766143102e-05}
{"train_loss": 0.11512113362550735, "global_step": 92553, "epoch": 1039, "lr": 7.412272985959928e-05}
{"train_loss": 0.1791066825389862, "global_step": 92554, "epoch": 1039, "lr": 7.412222205452455e-05}
{"train_loss": 0.1951235979795456, "global_step": 92555, "epoch": 1039, "lr": 7.412171424620689e-05}
{"train_loss": 0.1107686460018158, "global_step": 92556, "epoch": 1039, "lr": 7.412120643464635e-05}
{"train_loss": 0.20227402448654175, "global_step": 92557, "epoch": 1039, "lr": 7.4120698619843e-05}
{"train_loss": 0.11719866096973419, "global_step": 92558, "epoch": 1039, "lr": 7.412019080179693e-05}
{"train_loss": 0.17312822683473653, "global_step": 92559, "epoch": 1039, "lr": 7.411968298050818e-05, "val_loss": 4.440242767333984}
{"train_loss": 0.156626895070076, "global_step": 92560, "epoch": 1040, "lr": 7.411917515597684e-05}
{"train_loss": 0.09814683347940445, "global_step": 92561, "epoch": 1040, "lr": 7.411866732820298e-05}
{"train_loss": 0.15429194271564484, "global_step": 92562, "epoch": 1040, "lr": 7.411815949718664e-05}
{"train_loss": 0.1585671752691269, "global_step": 92563, "epoch": 1040, "lr": 7.411765166292792e-05}
{"train_loss": 0.17750266194343567, "global_step": 92564, "epoch": 1040, "lr": 7.411714382542686e-05}
{"train_loss": 0.10671280324459076, "global_step": 92565, "epoch": 1040, "lr": 7.411663598468357e-05}
{"train_loss": 0.15680822730064392, "global_step": 92566, "epoch": 1040, "lr": 7.411612814069808e-05}
{"train_loss": 0.15778295695781708, "global_step": 92567, "epoch": 1040, "lr": 7.411562029347047e-05}
{"train_loss": 0.1444096863269806, "global_step": 92568, "epoch": 1040, "lr": 7.411511244300081e-05}
{"train_loss": 0.1776179075241089, "global_step": 92569, "epoch": 1040, "lr": 7.411460458928916e-05}
{"train_loss": 0.1837848573923111, "global_step": 92570, "epoch": 1040, "lr": 7.41140967323356e-05}
{"train_loss": 0.13543543219566345, "global_step": 92571, "epoch": 1040, "lr": 7.41135888721402e-05}
{"train_loss": 0.10199739038944244, "global_step": 92572, "epoch": 1040, "lr": 7.411308100870303e-05}
{"train_loss": 0.1660454422235489, "global_step": 92573, "epoch": 1040, "lr": 7.411257314202413e-05}
{"train_loss": 0.19166100025177002, "global_step": 92574, "epoch": 1040, "lr": 7.41120652721036e-05}
{"train_loss": 0.10365462303161621, "global_step": 92575, "epoch": 1040, "lr": 7.411155739894149e-05}
{"train_loss": 0.12363749742507935, "global_step": 92576, "epoch": 1040, "lr": 7.411104952253787e-05}
{"train_loss": 0.2175649255514145, "global_step": 92577, "epoch": 1040, "lr": 7.411054164289282e-05}
{"train_loss": 0.13247010111808777, "global_step": 92578, "epoch": 1040, "lr": 7.411003376000641e-05}
{"train_loss": 0.149778351187706, "global_step": 92579, "epoch": 1040, "lr": 7.410952587387868e-05}
{"train_loss": 0.19714218378067017, "global_step": 92580, "epoch": 1040, "lr": 7.410901798450973e-05}
{"train_loss": 0.14386340975761414, "global_step": 92581, "epoch": 1040, "lr": 7.410851009189962e-05}
{"train_loss": 0.1181943342089653, "global_step": 92582, "epoch": 1040, "lr": 7.410800219604841e-05}
{"train_loss": 0.17816881835460663, "global_step": 92583, "epoch": 1040, "lr": 7.410749429695618e-05}
{"train_loss": 0.16974583268165588, "global_step": 92584, "epoch": 1040, "lr": 7.410698639462298e-05}
{"train_loss": 0.18021325767040253, "global_step": 92585, "epoch": 1040, "lr": 7.410647848904889e-05}
{"train_loss": 0.15119953453540802, "global_step": 92586, "epoch": 1040, "lr": 7.410597058023399e-05}
{"train_loss": 0.20647315680980682, "global_step": 92587, "epoch": 1040, "lr": 7.410546266817833e-05}
{"train_loss": 0.1770590990781784, "global_step": 92588, "epoch": 1040, "lr": 7.410495475288197e-05}
{"train_loss": 0.1345977634191513, "global_step": 92589, "epoch": 1040, "lr": 7.4104446834345e-05}
{"train_loss": 0.1444227546453476, "global_step": 92590, "epoch": 1040, "lr": 7.410393891256749e-05}
{"train_loss": 0.14977528154850006, "global_step": 92591, "epoch": 1040, "lr": 7.41034309875495e-05}
{"train_loss": 0.184592604637146, "global_step": 92592, "epoch": 1040, "lr": 7.41029230592911e-05}
{"train_loss": 0.18967454135417938, "global_step": 92593, "epoch": 1040, "lr": 7.410241512779235e-05}
{"train_loss": 0.16645285487174988, "global_step": 92594, "epoch": 1040, "lr": 7.410190719305332e-05}
{"train_loss": 0.20622701942920685, "global_step": 92595, "epoch": 1040, "lr": 7.410139925507408e-05}
{"train_loss": 0.2548464834690094, "global_step": 92596, "epoch": 1040, "lr": 7.410089131385471e-05}
{"train_loss": 0.1914558708667755, "global_step": 92597, "epoch": 1040, "lr": 7.410038336939527e-05}
{"train_loss": 0.08452305197715759, "global_step": 92598, "epoch": 1040, "lr": 7.409987542169582e-05}
{"train_loss": 0.14673474431037903, "global_step": 92599, "epoch": 1040, "lr": 7.409936747075643e-05}
{"train_loss": 0.11730876564979553, "global_step": 92600, "epoch": 1040, "lr": 7.409885951657718e-05}
{"train_loss": 0.16433703899383545, "global_step": 92601, "epoch": 1040, "lr": 7.409835155915814e-05}
{"train_loss": 0.2396559715270996, "global_step": 92602, "epoch": 1040, "lr": 7.409784359849936e-05}
{"train_loss": 0.17620345950126648, "global_step": 92603, "epoch": 1040, "lr": 7.40973356346009e-05}
{"train_loss": 0.22236646711826324, "global_step": 92604, "epoch": 1040, "lr": 7.409682766746289e-05}
{"train_loss": 0.15332549810409546, "global_step": 92605, "epoch": 1040, "lr": 7.409631969708534e-05}
{"train_loss": 0.1388138085603714, "global_step": 92606, "epoch": 1040, "lr": 7.409581172346832e-05}
{"train_loss": 0.19355084002017975, "global_step": 92607, "epoch": 1040, "lr": 7.409530374661192e-05}
{"train_loss": 0.13116271793842316, "global_step": 92608, "epoch": 1040, "lr": 7.40947957665162e-05}
{"train_loss": 0.16719211637973785, "global_step": 92609, "epoch": 1040, "lr": 7.409428778318122e-05}
{"train_loss": 0.23762552440166473, "global_step": 92610, "epoch": 1040, "lr": 7.409377979660707e-05}
{"train_loss": 0.2078769952058792, "global_step": 92611, "epoch": 1040, "lr": 7.409327180679381e-05}
{"train_loss": 0.11296505481004715, "global_step": 92612, "epoch": 1040, "lr": 7.409276381374148e-05}
{"train_loss": 0.2119930237531662, "global_step": 92613, "epoch": 1040, "lr": 7.40922558174502e-05}
{"train_loss": 0.163142591714859, "global_step": 92614, "epoch": 1040, "lr": 7.409174781792e-05}
{"train_loss": 0.2537900507450104, "global_step": 92615, "epoch": 1040, "lr": 7.409123981515094e-05}
{"train_loss": 0.12301980704069138, "global_step": 92616, "epoch": 1040, "lr": 7.409073180914313e-05}
{"train_loss": 0.15055882930755615, "global_step": 92617, "epoch": 1040, "lr": 7.409022379989659e-05}
{"train_loss": 0.18272773921489716, "global_step": 92618, "epoch": 1040, "lr": 7.408971578741143e-05}
{"train_loss": 0.1682828962802887, "global_step": 92619, "epoch": 1040, "lr": 7.40892077716877e-05}
{"train_loss": 0.17221753299236298, "global_step": 92620, "epoch": 1040, "lr": 7.408869975272546e-05}
{"train_loss": 0.15750539302825928, "global_step": 92621, "epoch": 1040, "lr": 7.40881917305248e-05}
{"train_loss": 0.265439510345459, "global_step": 92622, "epoch": 1040, "lr": 7.408768370508576e-05}
{"train_loss": 0.18813559412956238, "global_step": 92623, "epoch": 1040, "lr": 7.408717567640844e-05}
{"train_loss": 0.12315915524959564, "global_step": 92624, "epoch": 1040, "lr": 7.408666764449291e-05}
{"train_loss": 0.15510976314544678, "global_step": 92625, "epoch": 1040, "lr": 7.40861596093392e-05}
{"train_loss": 0.2665359377861023, "global_step": 92626, "epoch": 1040, "lr": 7.408565157094739e-05}
{"train_loss": 0.23381966352462769, "global_step": 92627, "epoch": 1040, "lr": 7.408514352931757e-05}
{"train_loss": 0.2389097511768341, "global_step": 92628, "epoch": 1040, "lr": 7.408463548444979e-05}
{"train_loss": 0.16514527797698975, "global_step": 92629, "epoch": 1040, "lr": 7.408412743634411e-05}
{"train_loss": 0.1670318990945816, "global_step": 92630, "epoch": 1040, "lr": 7.408361938500065e-05}
{"train_loss": 0.14667287468910217, "global_step": 92631, "epoch": 1040, "lr": 7.408311133041941e-05}
{"train_loss": 0.21047186851501465, "global_step": 92632, "epoch": 1040, "lr": 7.408260327260051e-05}
{"train_loss": 0.17208294570446014, "global_step": 92633, "epoch": 1040, "lr": 7.408209521154398e-05}
{"train_loss": 0.246648371219635, "global_step": 92634, "epoch": 1040, "lr": 7.408158714724993e-05}
{"train_loss": 0.22617965936660767, "global_step": 92635, "epoch": 1040, "lr": 7.408107907971837e-05}
{"train_loss": 0.14975349605083466, "global_step": 92636, "epoch": 1040, "lr": 7.408057100894943e-05}
{"train_loss": 0.18689553439617157, "global_step": 92637, "epoch": 1040, "lr": 7.408006293494314e-05}
{"train_loss": 0.12799079716205597, "global_step": 92638, "epoch": 1040, "lr": 7.407955485769957e-05}
{"train_loss": 0.24139690399169922, "global_step": 92639, "epoch": 1040, "lr": 7.407904677721882e-05}
{"train_loss": 0.23303912580013275, "global_step": 92640, "epoch": 1040, "lr": 7.40785386935009e-05}
{"train_loss": 0.14384746551513672, "global_step": 92641, "epoch": 1040, "lr": 7.407803060654595e-05}
{"train_loss": 0.1851167380809784, "global_step": 92642, "epoch": 1040, "lr": 7.407752251635399e-05}
{"train_loss": 0.24892285466194153, "global_step": 92643, "epoch": 1040, "lr": 7.407701442292511e-05}
{"train_loss": 0.15259350836277008, "global_step": 92644, "epoch": 1040, "lr": 7.407650632625935e-05}
{"train_loss": 0.1718973070383072, "global_step": 92645, "epoch": 1040, "lr": 7.40759982263568e-05}
{"train_loss": 0.24959316849708557, "global_step": 92646, "epoch": 1040, "lr": 7.407549012321754e-05}
{"train_loss": 0.23731543123722076, "global_step": 92647, "epoch": 1040, "lr": 7.407498201684162e-05}
{"train_loss": 0.17371222293109037, "global_step": 92648, "epoch": 1040, "lr": 7.40744739072291e-05, "val_loss": 4.314225196838379, "train_action_mse_error": 27.88250160217285}
{"train_loss": 0.1914791613817215, "global_step": 92649, "epoch": 1041, "lr": 7.407396579438006e-05}
{"train_loss": 0.1873260736465454, "global_step": 92650, "epoch": 1041, "lr": 7.407345767829459e-05}
{"train_loss": 0.16384443640708923, "global_step": 92651, "epoch": 1041, "lr": 7.407294955897272e-05}
{"train_loss": 0.22828258574008942, "global_step": 92652, "epoch": 1041, "lr": 7.407244143641454e-05}
{"train_loss": 0.23020774126052856, "global_step": 92653, "epoch": 1041, "lr": 7.407193331062012e-05}
{"train_loss": 0.20418822765350342, "global_step": 92654, "epoch": 1041, "lr": 7.407142518158951e-05}
{"train_loss": 0.16428294777870178, "global_step": 92655, "epoch": 1041, "lr": 7.407091704932279e-05}
{"train_loss": 0.18970271944999695, "global_step": 92656, "epoch": 1041, "lr": 7.407040891382003e-05}
{"train_loss": 0.16319985687732697, "global_step": 92657, "epoch": 1041, "lr": 7.40699007750813e-05}
{"train_loss": 0.1682542860507965, "global_step": 92658, "epoch": 1041, "lr": 7.406939263310666e-05}
{"train_loss": 0.13940489292144775, "global_step": 92659, "epoch": 1041, "lr": 7.406888448789618e-05}
{"train_loss": 0.15168921649456024, "global_step": 92660, "epoch": 1041, "lr": 7.406837633944994e-05}
{"train_loss": 0.09915012121200562, "global_step": 92661, "epoch": 1041, "lr": 7.406786818776801e-05}
{"train_loss": 0.1285042017698288, "global_step": 92662, "epoch": 1041, "lr": 7.406736003285044e-05}
{"train_loss": 0.19310548901557922, "global_step": 92663, "epoch": 1041, "lr": 7.406685187469731e-05}
{"train_loss": 0.1603328287601471, "global_step": 92664, "epoch": 1041, "lr": 7.406634371330869e-05}
{"train_loss": 0.17379657924175262, "global_step": 92665, "epoch": 1041, "lr": 7.406583554868462e-05}
{"train_loss": 0.12385335564613342, "global_step": 92666, "epoch": 1041, "lr": 7.406532738082521e-05}
{"train_loss": 0.21049511432647705, "global_step": 92667, "epoch": 1041, "lr": 7.40648192097305e-05}
{"train_loss": 0.1463782787322998, "global_step": 92668, "epoch": 1041, "lr": 7.406431103540059e-05}
{"train_loss": 0.15015670657157898, "global_step": 92669, "epoch": 1041, "lr": 7.406380285783551e-05}
{"train_loss": 0.156411811709404, "global_step": 92670, "epoch": 1041, "lr": 7.406329467703535e-05}
{"train_loss": 0.16494323313236237, "global_step": 92671, "epoch": 1041, "lr": 7.406278649300016e-05}
{"train_loss": 0.1643637716770172, "global_step": 92672, "epoch": 1041, "lr": 7.406227830573003e-05}
{"train_loss": 0.18575215339660645, "global_step": 92673, "epoch": 1041, "lr": 7.406177011522503e-05}
{"train_loss": 0.1090179905295372, "global_step": 92674, "epoch": 1041, "lr": 7.40612619214852e-05}
{"train_loss": 0.29172250628471375, "global_step": 92675, "epoch": 1041, "lr": 7.406075372451064e-05}
{"train_loss": 0.23211029171943665, "global_step": 92676, "epoch": 1041, "lr": 7.40602455243014e-05}
{"train_loss": 0.2248798906803131, "global_step": 92677, "epoch": 1041, "lr": 7.405973732085754e-05}
{"train_loss": 0.14016921818256378, "global_step": 92678, "epoch": 1041, "lr": 7.405922911417916e-05}
{"train_loss": 0.16021355986595154, "global_step": 92679, "epoch": 1041, "lr": 7.405872090426631e-05}
{"train_loss": 0.16355077922344208, "global_step": 92680, "epoch": 1041, "lr": 7.405821269111906e-05}
{"train_loss": 0.13382071256637573, "global_step": 92681, "epoch": 1041, "lr": 7.405770447473747e-05}
{"train_loss": 0.16480939090251923, "global_step": 92682, "epoch": 1041, "lr": 7.405719625512162e-05}
{"train_loss": 0.19142001867294312, "global_step": 92683, "epoch": 1041, "lr": 7.405668803227157e-05}
{"train_loss": 0.18995214998722076, "global_step": 92684, "epoch": 1041, "lr": 7.405617980618739e-05}
{"train_loss": 0.15517444908618927, "global_step": 92685, "epoch": 1041, "lr": 7.405567157686914e-05}
{"train_loss": 0.17910104990005493, "global_step": 92686, "epoch": 1041, "lr": 7.405516334431691e-05}
{"train_loss": 0.1550992876291275, "global_step": 92687, "epoch": 1041, "lr": 7.405465510853076e-05}
{"train_loss": 0.1691146194934845, "global_step": 92688, "epoch": 1041, "lr": 7.405414686951075e-05}
{"train_loss": 0.21793213486671448, "global_step": 92689, "epoch": 1041, "lr": 7.405363862725694e-05}
{"train_loss": 0.1529313176870346, "global_step": 92690, "epoch": 1041, "lr": 7.405313038176944e-05}
{"train_loss": 0.16542653739452362, "global_step": 92691, "epoch": 1041, "lr": 7.405262213304827e-05}
{"train_loss": 0.19023187458515167, "global_step": 92692, "epoch": 1041, "lr": 7.405211388109352e-05}
{"train_loss": 0.15922723710536957, "global_step": 92693, "epoch": 1041, "lr": 7.405160562590526e-05}
{"train_loss": 0.21572841703891754, "global_step": 92694, "epoch": 1041, "lr": 7.405109736748354e-05}
{"train_loss": 0.23464339971542358, "global_step": 92695, "epoch": 1041, "lr": 7.405058910582846e-05}
{"train_loss": 0.12314924597740173, "global_step": 92696, "epoch": 1041, "lr": 7.405008084094007e-05}
{"train_loss": 0.18077681958675385, "global_step": 92697, "epoch": 1041, "lr": 7.404957257281843e-05}
{"train_loss": 0.12717850506305695, "global_step": 92698, "epoch": 1041, "lr": 7.404906430146363e-05}
{"train_loss": 0.12970218062400818, "global_step": 92699, "epoch": 1041, "lr": 7.404855602687573e-05}
{"train_loss": 0.1264777034521103, "global_step": 92700, "epoch": 1041, "lr": 7.404804774905478e-05}
{"train_loss": 0.13525205850601196, "global_step": 92701, "epoch": 1041, "lr": 7.404753946800087e-05}
{"train_loss": 0.1666157841682434, "global_step": 92702, "epoch": 1041, "lr": 7.404703118371406e-05}
{"train_loss": 0.10237782448530197, "global_step": 92703, "epoch": 1041, "lr": 7.404652289619443e-05}
{"train_loss": 0.22467157244682312, "global_step": 92704, "epoch": 1041, "lr": 7.404601460544202e-05}
{"train_loss": 0.1541486233472824, "global_step": 92705, "epoch": 1041, "lr": 7.404550631145694e-05}
{"train_loss": 0.14213405549526215, "global_step": 92706, "epoch": 1041, "lr": 7.404499801423921e-05}
{"train_loss": 0.20222602784633636, "global_step": 92707, "epoch": 1041, "lr": 7.404448971378893e-05}
{"train_loss": 0.20216797292232513, "global_step": 92708, "epoch": 1041, "lr": 7.404398141010617e-05}
{"train_loss": 0.25016307830810547, "global_step": 92709, "epoch": 1041, "lr": 7.404347310319099e-05}
{"train_loss": 0.168711319565773, "global_step": 92710, "epoch": 1041, "lr": 7.404296479304345e-05}
{"train_loss": 0.1687384843826294, "global_step": 92711, "epoch": 1041, "lr": 7.404245647966362e-05}
{"train_loss": 0.1501799374818802, "global_step": 92712, "epoch": 1041, "lr": 7.404194816305158e-05}
{"train_loss": 0.24540509283542633, "global_step": 92713, "epoch": 1041, "lr": 7.404143984320739e-05}
{"train_loss": 0.2265045940876007, "global_step": 92714, "epoch": 1041, "lr": 7.404093152013113e-05}
{"train_loss": 0.21191108226776123, "global_step": 92715, "epoch": 1041, "lr": 7.404042319382286e-05}
{"train_loss": 0.14936520159244537, "global_step": 92716, "epoch": 1041, "lr": 7.403991486428265e-05}
{"train_loss": 0.20272493362426758, "global_step": 92717, "epoch": 1041, "lr": 7.403940653151056e-05}
{"train_loss": 0.1406092345714569, "global_step": 92718, "epoch": 1041, "lr": 7.403889819550667e-05}
{"train_loss": 0.19183661043643951, "global_step": 92719, "epoch": 1041, "lr": 7.403838985627104e-05}
{"train_loss": 0.12589700520038605, "global_step": 92720, "epoch": 1041, "lr": 7.403788151380375e-05}
{"train_loss": 0.18631981313228607, "global_step": 92721, "epoch": 1041, "lr": 7.403737316810485e-05}
{"train_loss": 0.15806934237480164, "global_step": 92722, "epoch": 1041, "lr": 7.403686481917442e-05}
{"train_loss": 0.25452736020088196, "global_step": 92723, "epoch": 1041, "lr": 7.403635646701252e-05}
{"train_loss": 0.2131762057542801, "global_step": 92724, "epoch": 1041, "lr": 7.403584811161924e-05}
{"train_loss": 0.18761055171489716, "global_step": 92725, "epoch": 1041, "lr": 7.403533975299462e-05}
{"train_loss": 0.2972559332847595, "global_step": 92726, "epoch": 1041, "lr": 7.403483139113874e-05}
{"train_loss": 0.10175283998250961, "global_step": 92727, "epoch": 1041, "lr": 7.403432302605168e-05}
{"train_loss": 0.07909764349460602, "global_step": 92728, "epoch": 1041, "lr": 7.40338146577335e-05}
{"train_loss": 0.18125823140144348, "global_step": 92729, "epoch": 1041, "lr": 7.403330628618426e-05}
{"train_loss": 0.17362460494041443, "global_step": 92730, "epoch": 1041, "lr": 7.403279791140403e-05}
{"train_loss": 0.21558576822280884, "global_step": 92731, "epoch": 1041, "lr": 7.403228953339288e-05}
{"train_loss": 0.20038771629333496, "global_step": 92732, "epoch": 1041, "lr": 7.403178115215091e-05}
{"train_loss": 0.2663148045539856, "global_step": 92733, "epoch": 1041, "lr": 7.403127276767812e-05}
{"train_loss": 0.19723369181156158, "global_step": 92734, "epoch": 1041, "lr": 7.403076437997465e-05}
{"train_loss": 0.16741277277469635, "global_step": 92735, "epoch": 1041, "lr": 7.403025598904052e-05}
{"train_loss": 0.15234163403511047, "global_step": 92736, "epoch": 1041, "lr": 7.402974759487582e-05}
{"train_loss": 0.17612252802996153, "global_step": 92737, "epoch": 1041, "lr": 7.402923919748062e-05, "val_loss": 4.374413967132568}
{"train_loss": 0.18791581690311432, "global_step": 92738, "epoch": 1042, "lr": 7.402873079685497e-05}
{"train_loss": 0.10810694843530655, "global_step": 92739, "epoch": 1042, "lr": 7.402822239299896e-05}
{"train_loss": 0.1507476270198822, "global_step": 92740, "epoch": 1042, "lr": 7.402771398591264e-05}
{"train_loss": 0.1662190705537796, "global_step": 92741, "epoch": 1042, "lr": 7.40272055755961e-05}
{"train_loss": 0.1838214099407196, "global_step": 92742, "epoch": 1042, "lr": 7.402669716204939e-05}
{"train_loss": 0.1691683530807495, "global_step": 92743, "epoch": 1042, "lr": 7.402618874527257e-05}
{"train_loss": 0.13114066421985626, "global_step": 92744, "epoch": 1042, "lr": 7.402568032526574e-05}
{"train_loss": 0.18055815994739532, "global_step": 92745, "epoch": 1042, "lr": 7.402517190202893e-05}
{"train_loss": 0.1928030550479889, "global_step": 92746, "epoch": 1042, "lr": 7.402466347556225e-05}
{"train_loss": 0.24928061664104462, "global_step": 92747, "epoch": 1042, "lr": 7.402415504586574e-05}
{"train_loss": 0.2167324721813202, "global_step": 92748, "epoch": 1042, "lr": 7.402364661293944e-05}
{"train_loss": 0.12968569993972778, "global_step": 92749, "epoch": 1042, "lr": 7.402313817678348e-05}
{"train_loss": 0.18762679398059845, "global_step": 92750, "epoch": 1042, "lr": 7.402262973739792e-05}
{"train_loss": 0.2623898684978485, "global_step": 92751, "epoch": 1042, "lr": 7.40221212947828e-05}
{"train_loss": 0.17002490162849426, "global_step": 92752, "epoch": 1042, "lr": 7.40216128489382e-05}
{"train_loss": 0.155659481883049, "global_step": 92753, "epoch": 1042, "lr": 7.402110439986418e-05}
{"train_loss": 0.21499942243099213, "global_step": 92754, "epoch": 1042, "lr": 7.402059594756082e-05}
{"train_loss": 0.20251955091953278, "global_step": 92755, "epoch": 1042, "lr": 7.402008749202816e-05}
{"train_loss": 0.17450356483459473, "global_step": 92756, "epoch": 1042, "lr": 7.401957903326633e-05}
{"train_loss": 0.14720025658607483, "global_step": 92757, "epoch": 1042, "lr": 7.401907057127534e-05}
{"train_loss": 0.24314361810684204, "global_step": 92758, "epoch": 1042, "lr": 7.401856210605527e-05}
{"train_loss": 0.15769578516483307, "global_step": 92759, "epoch": 1042, "lr": 7.401805363760623e-05}
{"train_loss": 0.13235965371131897, "global_step": 92760, "epoch": 1042, "lr": 7.401754516592822e-05}
{"train_loss": 0.18904249370098114, "global_step": 92761, "epoch": 1042, "lr": 7.401703669102136e-05}
{"train_loss": 0.3013012111186981, "global_step": 92762, "epoch": 1042, "lr": 7.401652821288572e-05}
{"train_loss": 0.23274588584899902, "global_step": 92763, "epoch": 1042, "lr": 7.401601973152133e-05}
{"train_loss": 0.17705558240413666, "global_step": 92764, "epoch": 1042, "lr": 7.401551124692827e-05}
{"train_loss": 0.1612812578678131, "global_step": 92765, "epoch": 1042, "lr": 7.401500275910663e-05}
{"train_loss": 0.1668248325586319, "global_step": 92766, "epoch": 1042, "lr": 7.401449426805647e-05}
{"train_loss": 0.19966310262680054, "global_step": 92767, "epoch": 1042, "lr": 7.401398577377783e-05}
{"train_loss": 0.23633109033107758, "global_step": 92768, "epoch": 1042, "lr": 7.401347727627082e-05}
{"train_loss": 0.23380516469478607, "global_step": 92769, "epoch": 1042, "lr": 7.401296877553548e-05}
{"train_loss": 0.1806151121854782, "global_step": 92770, "epoch": 1042, "lr": 7.401246027157191e-05}
{"train_loss": 0.12132550776004791, "global_step": 92771, "epoch": 1042, "lr": 7.401195176438014e-05}
{"train_loss": 0.1133531853556633, "global_step": 92772, "epoch": 1042, "lr": 7.401144325396026e-05}
{"train_loss": 0.18822775781154633, "global_step": 92773, "epoch": 1042, "lr": 7.401093474031234e-05}
{"train_loss": 0.10796792805194855, "global_step": 92774, "epoch": 1042, "lr": 7.401042622343643e-05}
{"train_loss": 0.14168661832809448, "global_step": 92775, "epoch": 1042, "lr": 7.400991770333263e-05}
{"train_loss": 0.10098046809434891, "global_step": 92776, "epoch": 1042, "lr": 7.400940918000097e-05}
{"train_loss": 0.16415831446647644, "global_step": 92777, "epoch": 1042, "lr": 7.400890065344154e-05}
{"train_loss": 0.2775590419769287, "global_step": 92778, "epoch": 1042, "lr": 7.400839212365443e-05}
{"train_loss": 0.17567946016788483, "global_step": 92779, "epoch": 1042, "lr": 7.400788359063967e-05}
{"train_loss": 0.21389976143836975, "global_step": 92780, "epoch": 1042, "lr": 7.400737505439734e-05}
{"train_loss": 0.18924163281917572, "global_step": 92781, "epoch": 1042, "lr": 7.400686651492752e-05}
{"train_loss": 0.2283477634191513, "global_step": 92782, "epoch": 1042, "lr": 7.400635797223025e-05}
{"train_loss": 0.18255189061164856, "global_step": 92783, "epoch": 1042, "lr": 7.400584942630561e-05}
{"train_loss": 0.1447199434041977, "global_step": 92784, "epoch": 1042, "lr": 7.40053408771537e-05}
{"train_loss": 0.2824722230434418, "global_step": 92785, "epoch": 1042, "lr": 7.400483232477455e-05}
{"train_loss": 0.19306467473506927, "global_step": 92786, "epoch": 1042, "lr": 7.400432376916826e-05}
{"train_loss": 0.17637322843074799, "global_step": 92787, "epoch": 1042, "lr": 7.400381521033487e-05}
{"train_loss": 0.157114177942276, "global_step": 92788, "epoch": 1042, "lr": 7.400330664827446e-05}
{"train_loss": 0.25313475728034973, "global_step": 92789, "epoch": 1042, "lr": 7.40027980829871e-05}
{"train_loss": 0.19461612403392792, "global_step": 92790, "epoch": 1042, "lr": 7.400228951447286e-05}
{"train_loss": 0.1525842547416687, "global_step": 92791, "epoch": 1042, "lr": 7.400178094273179e-05}
{"train_loss": 0.1677199751138687, "global_step": 92792, "epoch": 1042, "lr": 7.400127236776399e-05}
{"train_loss": 0.20788557827472687, "global_step": 92793, "epoch": 1042, "lr": 7.400076378956951e-05}
{"train_loss": 0.18458235263824463, "global_step": 92794, "epoch": 1042, "lr": 7.40002552081484e-05}
{"train_loss": 0.2174181342124939, "global_step": 92795, "epoch": 1042, "lr": 7.399974662350077e-05}
{"train_loss": 0.21961964666843414, "global_step": 92796, "epoch": 1042, "lr": 7.399923803562666e-05}
{"train_loss": 0.24529209733009338, "global_step": 92797, "epoch": 1042, "lr": 7.399872944452615e-05}
{"train_loss": 0.20684023201465607, "global_step": 92798, "epoch": 1042, "lr": 7.39982208501993e-05}
{"train_loss": 0.22598345577716827, "global_step": 92799, "epoch": 1042, "lr": 7.399771225264617e-05}
{"train_loss": 0.13622666895389557, "global_step": 92800, "epoch": 1042, "lr": 7.399720365186686e-05}
{"train_loss": 0.1816304624080658, "global_step": 92801, "epoch": 1042, "lr": 7.39966950478614e-05}
{"train_loss": 0.2588912546634674, "global_step": 92802, "epoch": 1042, "lr": 7.399618644062989e-05}
{"train_loss": 0.17312398552894592, "global_step": 92803, "epoch": 1042, "lr": 7.399567783017237e-05}
{"train_loss": 0.16142892837524414, "global_step": 92804, "epoch": 1042, "lr": 7.399516921648894e-05}
{"train_loss": 0.1734108328819275, "global_step": 92805, "epoch": 1042, "lr": 7.399466059957965e-05}
{"train_loss": 0.12521201372146606, "global_step": 92806, "epoch": 1042, "lr": 7.399415197944456e-05}
{"train_loss": 0.19398021697998047, "global_step": 92807, "epoch": 1042, "lr": 7.399364335608375e-05}
{"train_loss": 0.23929691314697266, "global_step": 92808, "epoch": 1042, "lr": 7.399313472949729e-05}
{"train_loss": 0.21327108144760132, "global_step": 92809, "epoch": 1042, "lr": 7.399262609968525e-05}
{"train_loss": 0.3029603362083435, "global_step": 92810, "epoch": 1042, "lr": 7.399211746664769e-05}
{"train_loss": 0.2794002890586853, "global_step": 92811, "epoch": 1042, "lr": 7.399160883038468e-05}
{"train_loss": 0.21079526841640472, "global_step": 92812, "epoch": 1042, "lr": 7.39911001908963e-05}
{"train_loss": 0.12999477982521057, "global_step": 92813, "epoch": 1042, "lr": 7.39905915481826e-05}
{"train_loss": 0.2059241682291031, "global_step": 92814, "epoch": 1042, "lr": 7.399008290224367e-05}
{"train_loss": 0.15303683280944824, "global_step": 92815, "epoch": 1042, "lr": 7.398957425307955e-05}
{"train_loss": 0.17882901430130005, "global_step": 92816, "epoch": 1042, "lr": 7.398906560069034e-05}
{"train_loss": 0.20541243255138397, "global_step": 92817, "epoch": 1042, "lr": 7.398855694507609e-05}
{"train_loss": 0.127716526389122, "global_step": 92818, "epoch": 1042, "lr": 7.398804828623685e-05}
{"train_loss": 0.1851329207420349, "global_step": 92819, "epoch": 1042, "lr": 7.398753962417273e-05}
{"train_loss": 0.10973099619150162, "global_step": 92820, "epoch": 1042, "lr": 7.398703095888378e-05}
{"train_loss": 0.1746266782283783, "global_step": 92821, "epoch": 1042, "lr": 7.398652229037004e-05}
{"train_loss": 0.23905009031295776, "global_step": 92822, "epoch": 1042, "lr": 7.398601361863162e-05}
{"train_loss": 0.10613840818405151, "global_step": 92823, "epoch": 1042, "lr": 7.398550494366857e-05}
{"train_loss": 0.1590280681848526, "global_step": 92824, "epoch": 1042, "lr": 7.398499626548096e-05}
{"train_loss": 0.1702340692281723, "global_step": 92825, "epoch": 1042, "lr": 7.398448758406888e-05}
{"train_loss": 0.1869380423861943, "global_step": 92826, "epoch": 1042, "lr": 7.398397889943237e-05, "val_loss": 4.358723163604736}
{"train_loss": 0.13453713059425354, "global_step": 92827, "epoch": 1043, "lr": 7.398347021157149e-05}
{"train_loss": 0.24412256479263306, "global_step": 92828, "epoch": 1043, "lr": 7.398296152048633e-05}
{"train_loss": 0.22407609224319458, "global_step": 92829, "epoch": 1043, "lr": 7.398245282617696e-05}
{"train_loss": 0.1708991527557373, "global_step": 92830, "epoch": 1043, "lr": 7.398194412864344e-05}
{"train_loss": 0.1321319043636322, "global_step": 92831, "epoch": 1043, "lr": 7.398143542788584e-05}
{"train_loss": 0.13938093185424805, "global_step": 92832, "epoch": 1043, "lr": 7.398092672390422e-05}
{"train_loss": 0.16313005983829498, "global_step": 92833, "epoch": 1043, "lr": 7.398041801669865e-05}
{"train_loss": 0.16742894053459167, "global_step": 92834, "epoch": 1043, "lr": 7.397990930626923e-05}
{"train_loss": 0.16795913875102997, "global_step": 92835, "epoch": 1043, "lr": 7.397940059261598e-05}
{"train_loss": 0.20666508376598358, "global_step": 92836, "epoch": 1043, "lr": 7.397889187573904e-05}
{"train_loss": 0.18588010966777802, "global_step": 92837, "epoch": 1043, "lr": 7.397838315563839e-05}
{"train_loss": 0.1297972947359085, "global_step": 92838, "epoch": 1043, "lr": 7.397787443231413e-05}
{"train_loss": 0.2132902443408966, "global_step": 92839, "epoch": 1043, "lr": 7.397736570576634e-05}
{"train_loss": 0.19188661873340607, "global_step": 92840, "epoch": 1043, "lr": 7.39768569759951e-05}
{"train_loss": 0.1943250149488449, "global_step": 92841, "epoch": 1043, "lr": 7.397634824300046e-05}
{"train_loss": 0.23197577893733978, "global_step": 92842, "epoch": 1043, "lr": 7.397583950678249e-05}
{"train_loss": 0.17497646808624268, "global_step": 92843, "epoch": 1043, "lr": 7.397533076734126e-05}
{"train_loss": 0.2143128216266632, "global_step": 92844, "epoch": 1043, "lr": 7.397482202467684e-05}
{"train_loss": 0.17083795368671417, "global_step": 92845, "epoch": 1043, "lr": 7.39743132787893e-05}
{"train_loss": 0.18786662817001343, "global_step": 92846, "epoch": 1043, "lr": 7.397380452967871e-05}
{"train_loss": 0.1690157949924469, "global_step": 92847, "epoch": 1043, "lr": 7.397329577734513e-05}
{"train_loss": 0.16350366175174713, "global_step": 92848, "epoch": 1043, "lr": 7.397278702178863e-05}
{"train_loss": 0.2825378477573395, "global_step": 92849, "epoch": 1043, "lr": 7.397227826300928e-05}
{"train_loss": 0.12758055329322815, "global_step": 92850, "epoch": 1043, "lr": 7.397176950100714e-05}
{"train_loss": 0.13571757078170776, "global_step": 92851, "epoch": 1043, "lr": 7.39712607357823e-05}
{"train_loss": 0.1499881148338318, "global_step": 92852, "epoch": 1043, "lr": 7.397075196733482e-05}
{"train_loss": 0.17290222644805908, "global_step": 92853, "epoch": 1043, "lr": 7.397024319566476e-05}
{"train_loss": 0.21671302616596222, "global_step": 92854, "epoch": 1043, "lr": 7.39697344207722e-05}
{"train_loss": 0.17193324863910675, "global_step": 92855, "epoch": 1043, "lr": 7.396922564265718e-05}
{"train_loss": 0.24691548943519592, "global_step": 92856, "epoch": 1043, "lr": 7.396871686131982e-05}
{"train_loss": 0.1724368929862976, "global_step": 92857, "epoch": 1043, "lr": 7.396820807676012e-05}
{"train_loss": 0.1483832597732544, "global_step": 92858, "epoch": 1043, "lr": 7.396769928897822e-05}
{"train_loss": 0.1326567828655243, "global_step": 92859, "epoch": 1043, "lr": 7.396719049797413e-05}
{"train_loss": 0.16711817681789398, "global_step": 92860, "epoch": 1043, "lr": 7.396668170374797e-05}
{"train_loss": 0.2735188603401184, "global_step": 92861, "epoch": 1043, "lr": 7.396617290629977e-05}
{"train_loss": 0.20492342114448547, "global_step": 92862, "epoch": 1043, "lr": 7.39656641056296e-05}
{"train_loss": 0.21515139937400818, "global_step": 92863, "epoch": 1043, "lr": 7.396515530173755e-05}
{"train_loss": 0.14302490651607513, "global_step": 92864, "epoch": 1043, "lr": 7.396464649462368e-05}
{"train_loss": 0.11357900500297546, "global_step": 92865, "epoch": 1043, "lr": 7.396413768428806e-05}
{"train_loss": 0.163628488779068, "global_step": 92866, "epoch": 1043, "lr": 7.396362887073073e-05}
{"train_loss": 0.2376994788646698, "global_step": 92867, "epoch": 1043, "lr": 7.396312005395179e-05}
{"train_loss": 0.17493543028831482, "global_step": 92868, "epoch": 1043, "lr": 7.39626112339513e-05}
{"train_loss": 0.15373702347278595, "global_step": 92869, "epoch": 1043, "lr": 7.396210241072935e-05}
{"train_loss": 0.05925186350941658, "global_step": 92870, "epoch": 1043, "lr": 7.396159358428596e-05}
{"train_loss": 0.18831267952919006, "global_step": 92871, "epoch": 1043, "lr": 7.396108475462124e-05}
{"train_loss": 0.21254156529903412, "global_step": 92872, "epoch": 1043, "lr": 7.396057592173525e-05}
{"train_loss": 0.1646929532289505, "global_step": 92873, "epoch": 1043, "lr": 7.396006708562803e-05}
{"train_loss": 0.14570848643779755, "global_step": 92874, "epoch": 1043, "lr": 7.395955824629969e-05}
{"train_loss": 0.16752883791923523, "global_step": 92875, "epoch": 1043, "lr": 7.395904940375028e-05}
{"train_loss": 0.178806334733963, "global_step": 92876, "epoch": 1043, "lr": 7.395854055797986e-05}
{"train_loss": 0.19023315608501434, "global_step": 92877, "epoch": 1043, "lr": 7.395803170898852e-05}
{"train_loss": 0.22181101143360138, "global_step": 92878, "epoch": 1043, "lr": 7.395752285677629e-05}
{"train_loss": 0.1361187994480133, "global_step": 92879, "epoch": 1043, "lr": 7.395701400134328e-05}
{"train_loss": 0.16671520471572876, "global_step": 92880, "epoch": 1043, "lr": 7.395650514268955e-05}
{"train_loss": 0.15122881531715393, "global_step": 92881, "epoch": 1043, "lr": 7.395599628081515e-05}
{"train_loss": 0.1307254582643509, "global_step": 92882, "epoch": 1043, "lr": 7.395548741572015e-05}
{"train_loss": 0.12975554168224335, "global_step": 92883, "epoch": 1043, "lr": 7.395497854740464e-05}
{"train_loss": 0.15005731582641602, "global_step": 92884, "epoch": 1043, "lr": 7.395446967586866e-05}
{"train_loss": 0.13309001922607422, "global_step": 92885, "epoch": 1043, "lr": 7.39539608011123e-05}
{"train_loss": 0.20057956874370575, "global_step": 92886, "epoch": 1043, "lr": 7.395345192313563e-05}
{"train_loss": 0.12379097938537598, "global_step": 92887, "epoch": 1043, "lr": 7.39529430419387e-05}
{"train_loss": 0.14234216511249542, "global_step": 92888, "epoch": 1043, "lr": 7.39524341575216e-05}
{"train_loss": 0.24871189892292023, "global_step": 92889, "epoch": 1043, "lr": 7.395192526988438e-05}
{"train_loss": 0.17838320136070251, "global_step": 92890, "epoch": 1043, "lr": 7.395141637902712e-05}
{"train_loss": 0.20690986514091492, "global_step": 92891, "epoch": 1043, "lr": 7.395090748494987e-05}
{"train_loss": 0.1831384152173996, "global_step": 92892, "epoch": 1043, "lr": 7.395039858765272e-05}
{"train_loss": 0.1583297848701477, "global_step": 92893, "epoch": 1043, "lr": 7.394988968713574e-05}
{"train_loss": 0.23077890276908875, "global_step": 92894, "epoch": 1043, "lr": 7.394938078339899e-05}
{"train_loss": 0.22232507169246674, "global_step": 92895, "epoch": 1043, "lr": 7.394887187644252e-05}
{"train_loss": 0.15726737678050995, "global_step": 92896, "epoch": 1043, "lr": 7.394836296626642e-05}
{"train_loss": 0.14798803627490997, "global_step": 92897, "epoch": 1043, "lr": 7.394785405287077e-05}
{"train_loss": 0.1672360599040985, "global_step": 92898, "epoch": 1043, "lr": 7.394734513625561e-05}
{"train_loss": 0.22349053621292114, "global_step": 92899, "epoch": 1043, "lr": 7.394683621642102e-05}
{"train_loss": 0.15812303125858307, "global_step": 92900, "epoch": 1043, "lr": 7.394632729336708e-05}
{"train_loss": 0.18741890788078308, "global_step": 92901, "epoch": 1043, "lr": 7.394581836709384e-05}
{"train_loss": 0.24948923289775848, "global_step": 92902, "epoch": 1043, "lr": 7.394530943760138e-05}
{"train_loss": 0.21042554080486298, "global_step": 92903, "epoch": 1043, "lr": 7.394480050488976e-05}
{"train_loss": 0.2543069124221802, "global_step": 92904, "epoch": 1043, "lr": 7.394429156895906e-05}
{"train_loss": 0.21710310876369476, "global_step": 92905, "epoch": 1043, "lr": 7.394378262980934e-05}
{"train_loss": 0.2925800085067749, "global_step": 92906, "epoch": 1043, "lr": 7.394327368744068e-05}
{"train_loss": 0.23013447225093842, "global_step": 92907, "epoch": 1043, "lr": 7.394276474185313e-05}
{"train_loss": 0.21504034101963043, "global_step": 92908, "epoch": 1043, "lr": 7.394225579304676e-05}
{"train_loss": 0.13857638835906982, "global_step": 92909, "epoch": 1043, "lr": 7.394174684102165e-05}
{"train_loss": 0.1388482302427292, "global_step": 92910, "epoch": 1043, "lr": 7.394123788577786e-05}
{"train_loss": 0.21401545405387878, "global_step": 92911, "epoch": 1043, "lr": 7.394072892731547e-05}
{"train_loss": 0.154526486992836, "global_step": 92912, "epoch": 1043, "lr": 7.394021996563453e-05}
{"train_loss": 0.17876440286636353, "global_step": 92913, "epoch": 1043, "lr": 7.393971100073513e-05}
{"train_loss": 0.23670056462287903, "global_step": 92914, "epoch": 1043, "lr": 7.393920203261733e-05}
{"train_loss": 0.181089891351006, "global_step": 92915, "epoch": 1043, "lr": 7.393869306128119e-05, "val_loss": 4.529345512390137}
{"train_loss": 0.15584968030452728, "global_step": 92916, "epoch": 1044, "lr": 7.393818408672677e-05}
{"train_loss": 0.21762442588806152, "global_step": 92917, "epoch": 1044, "lr": 7.393767510895418e-05}
{"train_loss": 0.159896582365036, "global_step": 92918, "epoch": 1044, "lr": 7.393716612796346e-05}
{"train_loss": 0.16683857142925262, "global_step": 92919, "epoch": 1044, "lr": 7.393665714375466e-05}
{"train_loss": 0.14554069936275482, "global_step": 92920, "epoch": 1044, "lr": 7.393614815632788e-05}
{"train_loss": 0.19781552255153656, "global_step": 92921, "epoch": 1044, "lr": 7.393563916568318e-05}
{"train_loss": 0.1974559724330902, "global_step": 92922, "epoch": 1044, "lr": 7.39351301718206e-05}
{"train_loss": 0.18595433235168457, "global_step": 92923, "epoch": 1044, "lr": 7.393462117474027e-05}
{"train_loss": 0.10012085735797882, "global_step": 92924, "epoch": 1044, "lr": 7.393411217444221e-05}
{"train_loss": 0.1306152194738388, "global_step": 92925, "epoch": 1044, "lr": 7.393360317092648e-05}
{"train_loss": 0.1670542061328888, "global_step": 92926, "epoch": 1044, "lr": 7.393309416419318e-05}
{"train_loss": 0.1831938773393631, "global_step": 92927, "epoch": 1044, "lr": 7.393258515424239e-05}
{"train_loss": 0.19782017171382904, "global_step": 92928, "epoch": 1044, "lr": 7.393207614107413e-05}
{"train_loss": 0.19962449371814728, "global_step": 92929, "epoch": 1044, "lr": 7.39315671246885e-05}
{"train_loss": 0.15917660295963287, "global_step": 92930, "epoch": 1044, "lr": 7.393105810508556e-05}
{"train_loss": 0.10288936644792557, "global_step": 92931, "epoch": 1044, "lr": 7.393054908226539e-05}
{"train_loss": 0.24967345595359802, "global_step": 92932, "epoch": 1044, "lr": 7.393004005622803e-05}
{"train_loss": 0.20671169459819794, "global_step": 92933, "epoch": 1044, "lr": 7.39295310269736e-05}
{"train_loss": 0.15799705684185028, "global_step": 92934, "epoch": 1044, "lr": 7.392902199450212e-05}
{"train_loss": 0.21505795419216156, "global_step": 92935, "epoch": 1044, "lr": 7.392851295881366e-05}
{"train_loss": 0.2024192214012146, "global_step": 92936, "epoch": 1044, "lr": 7.392800391990832e-05}
{"train_loss": 0.2055833339691162, "global_step": 92937, "epoch": 1044, "lr": 7.392749487778614e-05}
{"train_loss": 0.12692686915397644, "global_step": 92938, "epoch": 1044, "lr": 7.392698583244722e-05}
{"train_loss": 0.1726977676153183, "global_step": 92939, "epoch": 1044, "lr": 7.392647678389158e-05}
{"train_loss": 0.22781667113304138, "global_step": 92940, "epoch": 1044, "lr": 7.392596773211934e-05}
{"train_loss": 0.19826455414295197, "global_step": 92941, "epoch": 1044, "lr": 7.392545867713053e-05}
{"train_loss": 0.21592040359973907, "global_step": 92942, "epoch": 1044, "lr": 7.392494961892525e-05}
{"train_loss": 0.1218821108341217, "global_step": 92943, "epoch": 1044, "lr": 7.392444055750354e-05}
{"train_loss": 0.1679937094449997, "global_step": 92944, "epoch": 1044, "lr": 7.392393149286548e-05}
{"train_loss": 0.20054130256175995, "global_step": 92945, "epoch": 1044, "lr": 7.392342242501115e-05}
{"train_loss": 0.27220630645751953, "global_step": 92946, "epoch": 1044, "lr": 7.392291335394059e-05}
{"train_loss": 0.11977566033601761, "global_step": 92947, "epoch": 1044, "lr": 7.39224042796539e-05}
{"train_loss": 0.15226180851459503, "global_step": 92948, "epoch": 1044, "lr": 7.392189520215112e-05}
{"train_loss": 0.12887009978294373, "global_step": 92949, "epoch": 1044, "lr": 7.392138612143235e-05}
{"train_loss": 0.19007402658462524, "global_step": 92950, "epoch": 1044, "lr": 7.392087703749764e-05}
{"train_loss": 0.25951725244522095, "global_step": 92951, "epoch": 1044, "lr": 7.392036795034704e-05}
{"train_loss": 0.18365466594696045, "global_step": 92952, "epoch": 1044, "lr": 7.391985885998066e-05}
{"train_loss": 0.1417747288942337, "global_step": 92953, "epoch": 1044, "lr": 7.391934976639854e-05}
{"train_loss": 0.21514104306697845, "global_step": 92954, "epoch": 1044, "lr": 7.391884066960075e-05}
{"train_loss": 0.1454688161611557, "global_step": 92955, "epoch": 1044, "lr": 7.391833156958737e-05}
{"train_loss": 0.13876749575138092, "global_step": 92956, "epoch": 1044, "lr": 7.391782246635845e-05}
{"train_loss": 0.19542333483695984, "global_step": 92957, "epoch": 1044, "lr": 7.391731335991408e-05}
{"train_loss": 0.16821914911270142, "global_step": 92958, "epoch": 1044, "lr": 7.391680425025433e-05}
{"train_loss": 0.18116258084774017, "global_step": 92959, "epoch": 1044, "lr": 7.391629513737925e-05}
{"train_loss": 0.19245505332946777, "global_step": 92960, "epoch": 1044, "lr": 7.39157860212889e-05}
{"train_loss": 0.1553322970867157, "global_step": 92961, "epoch": 1044, "lr": 7.391527690198337e-05}
{"train_loss": 0.24073757231235504, "global_step": 92962, "epoch": 1044, "lr": 7.391476777946274e-05}
{"train_loss": 0.16709043085575104, "global_step": 92963, "epoch": 1044, "lr": 7.391425865372704e-05}
{"train_loss": 0.17048171162605286, "global_step": 92964, "epoch": 1044, "lr": 7.391374952477636e-05}
{"train_loss": 0.12199084460735321, "global_step": 92965, "epoch": 1044, "lr": 7.391324039261079e-05}
{"train_loss": 0.09337829053401947, "global_step": 92966, "epoch": 1044, "lr": 7.391273125723035e-05}
{"train_loss": 0.13227048516273499, "global_step": 92967, "epoch": 1044, "lr": 7.391222211863515e-05}
{"train_loss": 0.13258980214595795, "global_step": 92968, "epoch": 1044, "lr": 7.391171297682525e-05}
{"train_loss": 0.1600176990032196, "global_step": 92969, "epoch": 1044, "lr": 7.391120383180069e-05}
{"train_loss": 0.1295151263475418, "global_step": 92970, "epoch": 1044, "lr": 7.391069468356157e-05}
{"train_loss": 0.1416090875864029, "global_step": 92971, "epoch": 1044, "lr": 7.391018553210796e-05}
{"train_loss": 0.12070627510547638, "global_step": 92972, "epoch": 1044, "lr": 7.39096763774399e-05}
{"train_loss": 0.13205210864543915, "global_step": 92973, "epoch": 1044, "lr": 7.390916721955747e-05}
{"train_loss": 0.24392592906951904, "global_step": 92974, "epoch": 1044, "lr": 7.390865805846076e-05}
{"train_loss": 0.22851738333702087, "global_step": 92975, "epoch": 1044, "lr": 7.390814889414981e-05}
{"train_loss": 0.1288350373506546, "global_step": 92976, "epoch": 1044, "lr": 7.390763972662472e-05}
{"train_loss": 0.18363364040851593, "global_step": 92977, "epoch": 1044, "lr": 7.390713055588553e-05}
{"train_loss": 0.11753004044294357, "global_step": 92978, "epoch": 1044, "lr": 7.390662138193231e-05}
{"train_loss": 0.10388258844614029, "global_step": 92979, "epoch": 1044, "lr": 7.390611220476515e-05}
{"train_loss": 0.10699810087680817, "global_step": 92980, "epoch": 1044, "lr": 7.39056030243841e-05}
{"train_loss": 0.16179952025413513, "global_step": 92981, "epoch": 1044, "lr": 7.390509384078923e-05}
{"train_loss": 0.14062096178531647, "global_step": 92982, "epoch": 1044, "lr": 7.39045846539806e-05}
{"train_loss": 0.18805477023124695, "global_step": 92983, "epoch": 1044, "lr": 7.390407546395831e-05}
{"train_loss": 0.18047775328159332, "global_step": 92984, "epoch": 1044, "lr": 7.390356627072238e-05}
{"train_loss": 0.14226488769054413, "global_step": 92985, "epoch": 1044, "lr": 7.390305707427295e-05}
{"train_loss": 0.1946903020143509, "global_step": 92986, "epoch": 1044, "lr": 7.390254787461002e-05}
{"train_loss": 0.13981878757476807, "global_step": 92987, "epoch": 1044, "lr": 7.390203867173369e-05}
{"train_loss": 0.16010065376758575, "global_step": 92988, "epoch": 1044, "lr": 7.390152946564401e-05}
{"train_loss": 0.12380813807249069, "global_step": 92989, "epoch": 1044, "lr": 7.390102025634107e-05}
{"train_loss": 0.23366743326187134, "global_step": 92990, "epoch": 1044, "lr": 7.390051104382493e-05}
{"train_loss": 0.17038768529891968, "global_step": 92991, "epoch": 1044, "lr": 7.390000182809564e-05}
{"train_loss": 0.132258340716362, "global_step": 92992, "epoch": 1044, "lr": 7.38994926091533e-05}
{"train_loss": 0.20195885002613068, "global_step": 92993, "epoch": 1044, "lr": 7.389898338699797e-05}
{"train_loss": 0.14358985424041748, "global_step": 92994, "epoch": 1044, "lr": 7.389847416162972e-05}
{"train_loss": 0.18725088238716125, "global_step": 92995, "epoch": 1044, "lr": 7.389796493304859e-05}
{"train_loss": 0.25017049908638, "global_step": 92996, "epoch": 1044, "lr": 7.389745570125469e-05}
{"train_loss": 0.19570137560367584, "global_step": 92997, "epoch": 1044, "lr": 7.389694646624805e-05}
{"train_loss": 0.18562763929367065, "global_step": 92998, "epoch": 1044, "lr": 7.389643722802876e-05}
{"train_loss": 0.09006907790899277, "global_step": 92999, "epoch": 1044, "lr": 7.38959279865969e-05}
{"train_loss": 0.18671119213104248, "global_step": 93000, "epoch": 1044, "lr": 7.389541874195251e-05}
{"train_loss": 0.18169648945331573, "global_step": 93001, "epoch": 1044, "lr": 7.389490949409567e-05}
{"train_loss": 0.12380748987197876, "global_step": 93002, "epoch": 1044, "lr": 7.389440024302645e-05}
{"train_loss": 0.15622054040431976, "global_step": 93003, "epoch": 1044, "lr": 7.389389098874493e-05}
{"train_loss": 0.1694626145148545, "global_step": 93004, "epoch": 1044, "lr": 7.389338173125116e-05, "val_loss": 4.71461820602417}
{"train_loss": 0.1864713877439499, "global_step": 93005, "epoch": 1045, "lr": 7.389287247054524e-05}
{"train_loss": 0.1682894229888916, "global_step": 93006, "epoch": 1045, "lr": 7.389236320662719e-05}
{"train_loss": 0.14754018187522888, "global_step": 93007, "epoch": 1045, "lr": 7.389185393949712e-05}
{"train_loss": 0.13674134016036987, "global_step": 93008, "epoch": 1045, "lr": 7.389134466915506e-05}
{"train_loss": 0.20397289097309113, "global_step": 93009, "epoch": 1045, "lr": 7.389083539560111e-05}
{"train_loss": 0.1669953465461731, "global_step": 93010, "epoch": 1045, "lr": 7.389032611883532e-05}
{"train_loss": 0.18321889638900757, "global_step": 93011, "epoch": 1045, "lr": 7.388981683885777e-05}
{"train_loss": 0.12272541970014572, "global_step": 93012, "epoch": 1045, "lr": 7.388930755566854e-05}
{"train_loss": 0.10154086351394653, "global_step": 93013, "epoch": 1045, "lr": 7.388879826926767e-05}
{"train_loss": 0.1508258730173111, "global_step": 93014, "epoch": 1045, "lr": 7.388828897965524e-05}
{"train_loss": 0.1431446522474289, "global_step": 93015, "epoch": 1045, "lr": 7.388777968683133e-05}
{"train_loss": 0.2276759147644043, "global_step": 93016, "epoch": 1045, "lr": 7.3887270390796e-05}
{"train_loss": 0.16128548979759216, "global_step": 93017, "epoch": 1045, "lr": 7.388676109154931e-05}
{"train_loss": 0.18612849712371826, "global_step": 93018, "epoch": 1045, "lr": 7.388625178909133e-05}
{"train_loss": 0.1458396166563034, "global_step": 93019, "epoch": 1045, "lr": 7.388574248342214e-05}
{"train_loss": 0.17457331717014313, "global_step": 93020, "epoch": 1045, "lr": 7.388523317454181e-05}
{"train_loss": 0.1513579934835434, "global_step": 93021, "epoch": 1045, "lr": 7.38847238624504e-05}
{"train_loss": 0.15753546357154846, "global_step": 93022, "epoch": 1045, "lr": 7.388421454714796e-05}
{"train_loss": 0.18777024745941162, "global_step": 93023, "epoch": 1045, "lr": 7.38837052286346e-05}
{"train_loss": 0.12801077961921692, "global_step": 93024, "epoch": 1045, "lr": 7.388319590691037e-05}
{"train_loss": 0.19671252369880676, "global_step": 93025, "epoch": 1045, "lr": 7.388268658197534e-05}
{"train_loss": 0.14901232719421387, "global_step": 93026, "epoch": 1045, "lr": 7.388217725382956e-05}
{"train_loss": 0.17715787887573242, "global_step": 93027, "epoch": 1045, "lr": 7.388166792247311e-05}
{"train_loss": 0.13543905317783356, "global_step": 93028, "epoch": 1045, "lr": 7.388115858790608e-05}
{"train_loss": 0.157570019364357, "global_step": 93029, "epoch": 1045, "lr": 7.38806492501285e-05}
{"train_loss": 0.195926234126091, "global_step": 93030, "epoch": 1045, "lr": 7.388013990914046e-05}
{"train_loss": 0.189535453915596, "global_step": 93031, "epoch": 1045, "lr": 7.387963056494202e-05}
{"train_loss": 0.1939515322446823, "global_step": 93032, "epoch": 1045, "lr": 7.387912121753327e-05}
{"train_loss": 0.19981913268566132, "global_step": 93033, "epoch": 1045, "lr": 7.387861186691426e-05}
{"train_loss": 0.1799321323633194, "global_step": 93034, "epoch": 1045, "lr": 7.387810251308506e-05}
{"train_loss": 0.1488073319196701, "global_step": 93035, "epoch": 1045, "lr": 7.387759315604574e-05}
{"train_loss": 0.15671761333942413, "global_step": 93036, "epoch": 1045, "lr": 7.387708379579636e-05}
{"train_loss": 0.182607963681221, "global_step": 93037, "epoch": 1045, "lr": 7.3876574432337e-05}
{"train_loss": 0.13044363260269165, "global_step": 93038, "epoch": 1045, "lr": 7.387606506566773e-05}
{"train_loss": 0.2688809633255005, "global_step": 93039, "epoch": 1045, "lr": 7.387555569578861e-05}
{"train_loss": 0.19330309331417084, "global_step": 93040, "epoch": 1045, "lr": 7.387504632269972e-05}
{"train_loss": 0.19038039445877075, "global_step": 93041, "epoch": 1045, "lr": 7.387453694640112e-05}
{"train_loss": 0.1471104621887207, "global_step": 93042, "epoch": 1045, "lr": 7.387402756689288e-05}
{"train_loss": 0.12499716877937317, "global_step": 93043, "epoch": 1045, "lr": 7.387351818417506e-05}
{"train_loss": 0.15190868079662323, "global_step": 93044, "epoch": 1045, "lr": 7.387300879824775e-05}
{"train_loss": 0.21384266018867493, "global_step": 93045, "epoch": 1045, "lr": 7.387249940911099e-05}
{"train_loss": 0.18685650825500488, "global_step": 93046, "epoch": 1045, "lr": 7.387199001676488e-05}
{"train_loss": 0.1925852745771408, "global_step": 93047, "epoch": 1045, "lr": 7.387148062120945e-05}
{"train_loss": 0.1499720960855484, "global_step": 93048, "epoch": 1045, "lr": 7.387097122244482e-05}
{"train_loss": 0.22141005098819733, "global_step": 93049, "epoch": 1045, "lr": 7.387046182047101e-05}
{"train_loss": 0.17252033948898315, "global_step": 93050, "epoch": 1045, "lr": 7.38699524152881e-05}
{"train_loss": 0.11378104984760284, "global_step": 93051, "epoch": 1045, "lr": 7.386944300689618e-05}
{"train_loss": 0.19028086960315704, "global_step": 93052, "epoch": 1045, "lr": 7.38689335952953e-05}
{"train_loss": 0.13810129463672638, "global_step": 93053, "epoch": 1045, "lr": 7.386842418048553e-05}
{"train_loss": 0.16756939888000488, "global_step": 93054, "epoch": 1045, "lr": 7.386791476246694e-05}
{"train_loss": 0.18851570785045624, "global_step": 93055, "epoch": 1045, "lr": 7.386740534123961e-05}
{"train_loss": 0.10753452777862549, "global_step": 93056, "epoch": 1045, "lr": 7.386689591680358e-05}
{"train_loss": 0.17952287197113037, "global_step": 93057, "epoch": 1045, "lr": 7.386638648915893e-05}
{"train_loss": 0.14925438165664673, "global_step": 93058, "epoch": 1045, "lr": 7.386587705830577e-05}
{"train_loss": 0.19991189241409302, "global_step": 93059, "epoch": 1045, "lr": 7.386536762424411e-05}
{"train_loss": 0.16041964292526245, "global_step": 93060, "epoch": 1045, "lr": 7.386485818697405e-05}
{"train_loss": 0.14516068994998932, "global_step": 93061, "epoch": 1045, "lr": 7.386434874649565e-05}
{"train_loss": 0.17538340389728546, "global_step": 93062, "epoch": 1045, "lr": 7.386383930280898e-05}
{"train_loss": 0.20102275907993317, "global_step": 93063, "epoch": 1045, "lr": 7.386332985591412e-05}
{"train_loss": 0.16873309016227722, "global_step": 93064, "epoch": 1045, "lr": 7.386282040581111e-05}
{"train_loss": 0.22078533470630646, "global_step": 93065, "epoch": 1045, "lr": 7.386231095250003e-05}
{"train_loss": 0.16710422933101654, "global_step": 93066, "epoch": 1045, "lr": 7.386180149598097e-05}
{"train_loss": 0.09816311299800873, "global_step": 93067, "epoch": 1045, "lr": 7.386129203625396e-05}
{"train_loss": 0.17127375304698944, "global_step": 93068, "epoch": 1045, "lr": 7.38607825733191e-05}
{"train_loss": 0.17278890311717987, "global_step": 93069, "epoch": 1045, "lr": 7.386027310717644e-05}
{"train_loss": 0.16856488585472107, "global_step": 93070, "epoch": 1045, "lr": 7.385976363782608e-05}
{"train_loss": 0.17823483049869537, "global_step": 93071, "epoch": 1045, "lr": 7.385925416526805e-05}
{"train_loss": 0.16661478579044342, "global_step": 93072, "epoch": 1045, "lr": 7.385874468950244e-05}
{"train_loss": 0.19927164912223816, "global_step": 93073, "epoch": 1045, "lr": 7.385823521052931e-05}
{"train_loss": 0.17293842136859894, "global_step": 93074, "epoch": 1045, "lr": 7.385772572834872e-05}
{"train_loss": 0.11297591775655746, "global_step": 93075, "epoch": 1045, "lr": 7.385721624296076e-05}
{"train_loss": 0.1323750913143158, "global_step": 93076, "epoch": 1045, "lr": 7.385670675436549e-05}
{"train_loss": 0.19608472287654877, "global_step": 93077, "epoch": 1045, "lr": 7.385619726256298e-05}
{"train_loss": 0.11781961470842361, "global_step": 93078, "epoch": 1045, "lr": 7.385568776755328e-05}
{"train_loss": 0.1473264843225479, "global_step": 93079, "epoch": 1045, "lr": 7.385517826933648e-05}
{"train_loss": 0.15263108909130096, "global_step": 93080, "epoch": 1045, "lr": 7.385466876791265e-05}
{"train_loss": 0.22860220074653625, "global_step": 93081, "epoch": 1045, "lr": 7.385415926328186e-05}
{"train_loss": 0.16312769055366516, "global_step": 93082, "epoch": 1045, "lr": 7.385364975544414e-05}
{"train_loss": 0.15570935606956482, "global_step": 93083, "epoch": 1045, "lr": 7.385314024439962e-05}
{"train_loss": 0.2664574384689331, "global_step": 93084, "epoch": 1045, "lr": 7.385263073014832e-05}
{"train_loss": 0.2089591771364212, "global_step": 93085, "epoch": 1045, "lr": 7.385212121269031e-05}
{"train_loss": 0.13131144642829895, "global_step": 93086, "epoch": 1045, "lr": 7.385161169202571e-05}
{"train_loss": 0.15021167695522308, "global_step": 93087, "epoch": 1045, "lr": 7.385110216815452e-05}
{"train_loss": 0.16181860864162445, "global_step": 93088, "epoch": 1045, "lr": 7.385059264107684e-05}
{"train_loss": 0.18639230728149414, "global_step": 93089, "epoch": 1045, "lr": 7.385008311079275e-05}
{"train_loss": 0.254901647567749, "global_step": 93090, "epoch": 1045, "lr": 7.38495735773023e-05}
{"train_loss": 0.25929826498031616, "global_step": 93091, "epoch": 1045, "lr": 7.384906404060557e-05}
{"train_loss": 0.17218460142612457, "global_step": 93092, "epoch": 1045, "lr": 7.384855450070263e-05}
{"train_loss": 0.17158344681008478, "global_step": 93093, "epoch": 1045, "lr": 7.384804495759352e-05, "val_loss": 4.461490631103516, "train_action_mse_error": 15.848638534545898}
{"train_loss": 0.1183667853474617, "global_step": 93094, "epoch": 1046, "lr": 7.384753541127834e-05}
{"train_loss": 0.1479465663433075, "global_step": 93095, "epoch": 1046, "lr": 7.384702586175715e-05}
{"train_loss": 0.13878664374351501, "global_step": 93096, "epoch": 1046, "lr": 7.384651630903003e-05}
{"train_loss": 0.1592780500650406, "global_step": 93097, "epoch": 1046, "lr": 7.384600675309703e-05}
{"train_loss": 0.19927877187728882, "global_step": 93098, "epoch": 1046, "lr": 7.384549719395824e-05}
{"train_loss": 0.1959058940410614, "global_step": 93099, "epoch": 1046, "lr": 7.384498763161369e-05}
{"train_loss": 0.2195451408624649, "global_step": 93100, "epoch": 1046, "lr": 7.384447806606348e-05}
{"train_loss": 0.16411402821540833, "global_step": 93101, "epoch": 1046, "lr": 7.384396849730768e-05}
{"train_loss": 0.1311422735452652, "global_step": 93102, "epoch": 1046, "lr": 7.384345892534634e-05}
{"train_loss": 0.22121180593967438, "global_step": 93103, "epoch": 1046, "lr": 7.384294935017955e-05}
{"train_loss": 0.21883755922317505, "global_step": 93104, "epoch": 1046, "lr": 7.384243977180735e-05}
{"train_loss": 0.1603493094444275, "global_step": 93105, "epoch": 1046, "lr": 7.384193019022983e-05}
{"train_loss": 0.18719787895679474, "global_step": 93106, "epoch": 1046, "lr": 7.384142060544705e-05}
{"train_loss": 0.13117419183254242, "global_step": 93107, "epoch": 1046, "lr": 7.384091101745907e-05}
{"train_loss": 0.21166321635246277, "global_step": 93108, "epoch": 1046, "lr": 7.384040142626599e-05}
{"train_loss": 0.24606367945671082, "global_step": 93109, "epoch": 1046, "lr": 7.383989183186786e-05}
{"train_loss": 0.19390009343624115, "global_step": 93110, "epoch": 1046, "lr": 7.383938223426474e-05}
{"train_loss": 0.20077015459537506, "global_step": 93111, "epoch": 1046, "lr": 7.383887263345669e-05}
{"train_loss": 0.20832066237926483, "global_step": 93112, "epoch": 1046, "lr": 7.383836302944381e-05}
{"train_loss": 0.1858367919921875, "global_step": 93113, "epoch": 1046, "lr": 7.383785342222615e-05}
{"train_loss": 0.14442941546440125, "global_step": 93114, "epoch": 1046, "lr": 7.383734381180378e-05}
{"train_loss": 0.12918798625469208, "global_step": 93115, "epoch": 1046, "lr": 7.383683419817678e-05}
{"train_loss": 0.19648736715316772, "global_step": 93116, "epoch": 1046, "lr": 7.383632458134519e-05}
{"train_loss": 0.13026346266269684, "global_step": 93117, "epoch": 1046, "lr": 7.383581496130912e-05}
{"train_loss": 0.1463107019662857, "global_step": 93118, "epoch": 1046, "lr": 7.383530533806859e-05}
{"train_loss": 0.20296387374401093, "global_step": 93119, "epoch": 1046, "lr": 7.383479571162372e-05}
{"train_loss": 0.13793767988681793, "global_step": 93120, "epoch": 1046, "lr": 7.383428608197454e-05}
{"train_loss": 0.1478877067565918, "global_step": 93121, "epoch": 1046, "lr": 7.383377644912113e-05}
{"train_loss": 0.1366364061832428, "global_step": 93122, "epoch": 1046, "lr": 7.383326681306356e-05}
{"train_loss": 0.1761987954378128, "global_step": 93123, "epoch": 1046, "lr": 7.383275717380191e-05}
{"train_loss": 0.17473861575126648, "global_step": 93124, "epoch": 1046, "lr": 7.383224753133622e-05}
{"train_loss": 0.17849448323249817, "global_step": 93125, "epoch": 1046, "lr": 7.383173788566657e-05}
{"train_loss": 0.2895210087299347, "global_step": 93126, "epoch": 1046, "lr": 7.383122823679304e-05}
{"train_loss": 0.1629215031862259, "global_step": 93127, "epoch": 1046, "lr": 7.383071858471569e-05}
{"train_loss": 0.22878234088420868, "global_step": 93128, "epoch": 1046, "lr": 7.38302089294346e-05}
{"train_loss": 0.20702725648880005, "global_step": 93129, "epoch": 1046, "lr": 7.38296992709498e-05}
{"train_loss": 0.14554744958877563, "global_step": 93130, "epoch": 1046, "lr": 7.382918960926141e-05}
{"train_loss": 0.19014102220535278, "global_step": 93131, "epoch": 1046, "lr": 7.382867994436948e-05}
{"train_loss": 0.1563730537891388, "global_step": 93132, "epoch": 1046, "lr": 7.382817027627406e-05}
{"train_loss": 0.17157113552093506, "global_step": 93133, "epoch": 1046, "lr": 7.382766060497526e-05}
{"train_loss": 0.23998774588108063, "global_step": 93134, "epoch": 1046, "lr": 7.38271509304731e-05}
{"train_loss": 0.11131197214126587, "global_step": 93135, "epoch": 1046, "lr": 7.382664125276766e-05}
{"train_loss": 0.1731339544057846, "global_step": 93136, "epoch": 1046, "lr": 7.382613157185905e-05}
{"train_loss": 0.1760079711675644, "global_step": 93137, "epoch": 1046, "lr": 7.382562188774729e-05}
{"train_loss": 0.26980602741241455, "global_step": 93138, "epoch": 1046, "lr": 7.382511220043245e-05}
{"train_loss": 0.1563771814107895, "global_step": 93139, "epoch": 1046, "lr": 7.382460250991463e-05}
{"train_loss": 0.1649848222732544, "global_step": 93140, "epoch": 1046, "lr": 7.382409281619389e-05}
{"train_loss": 0.21875084936618805, "global_step": 93141, "epoch": 1046, "lr": 7.382358311927029e-05}
{"train_loss": 0.17831553518772125, "global_step": 93142, "epoch": 1046, "lr": 7.38230734191439e-05}
{"train_loss": 0.10247083008289337, "global_step": 93143, "epoch": 1046, "lr": 7.382256371581478e-05}
{"train_loss": 0.15932369232177734, "global_step": 93144, "epoch": 1046, "lr": 7.3822054009283e-05}
{"train_loss": 0.25272539258003235, "global_step": 93145, "epoch": 1046, "lr": 7.382154429954864e-05}
{"train_loss": 0.18996118009090424, "global_step": 93146, "epoch": 1046, "lr": 7.382103458661177e-05}
{"train_loss": 0.2182297706604004, "global_step": 93147, "epoch": 1046, "lr": 7.382052487047244e-05}
{"train_loss": 0.15995286405086517, "global_step": 93148, "epoch": 1046, "lr": 7.382001515113072e-05}
{"train_loss": 0.18478648364543915, "global_step": 93149, "epoch": 1046, "lr": 7.381950542858673e-05}
{"train_loss": 0.19835643470287323, "global_step": 93150, "epoch": 1046, "lr": 7.381899570284047e-05}
{"train_loss": 0.20146556198596954, "global_step": 93151, "epoch": 1046, "lr": 7.381848597389203e-05}
{"train_loss": 0.10596548765897751, "global_step": 93152, "epoch": 1046, "lr": 7.38179762417415e-05}
{"train_loss": 0.176936537027359, "global_step": 93153, "epoch": 1046, "lr": 7.381746650638894e-05}
{"train_loss": 0.1558426022529602, "global_step": 93154, "epoch": 1046, "lr": 7.38169567678344e-05}
{"train_loss": 0.21160143613815308, "global_step": 93155, "epoch": 1046, "lr": 7.381644702607796e-05}
{"train_loss": 0.13653817772865295, "global_step": 93156, "epoch": 1046, "lr": 7.38159372811197e-05}
{"train_loss": 0.17307288944721222, "global_step": 93157, "epoch": 1046, "lr": 7.381542753295967e-05}
{"train_loss": 0.17241458594799042, "global_step": 93158, "epoch": 1046, "lr": 7.381491778159793e-05}
{"train_loss": 0.13427357375621796, "global_step": 93159, "epoch": 1046, "lr": 7.381440802703459e-05}
{"train_loss": 0.11599979549646378, "global_step": 93160, "epoch": 1046, "lr": 7.38138982692697e-05}
{"train_loss": 0.13685539364814758, "global_step": 93161, "epoch": 1046, "lr": 7.38133885083033e-05}
{"train_loss": 0.24725501239299774, "global_step": 93162, "epoch": 1046, "lr": 7.381287874413549e-05}
{"train_loss": 0.15685658156871796, "global_step": 93163, "epoch": 1046, "lr": 7.381236897676631e-05}
{"train_loss": 0.2125817984342575, "global_step": 93164, "epoch": 1046, "lr": 7.381185920619587e-05}
{"train_loss": 0.17693257331848145, "global_step": 93165, "epoch": 1046, "lr": 7.38113494324242e-05}
{"train_loss": 0.14218516647815704, "global_step": 93166, "epoch": 1046, "lr": 7.38108396554514e-05}
{"train_loss": 0.2521989345550537, "global_step": 93167, "epoch": 1046, "lr": 7.38103298752775e-05}
{"train_loss": 0.2065969854593277, "global_step": 93168, "epoch": 1046, "lr": 7.38098200919026e-05}
{"train_loss": 0.1503485143184662, "global_step": 93169, "epoch": 1046, "lr": 7.380931030532676e-05}
{"train_loss": 0.13507647812366486, "global_step": 93170, "epoch": 1046, "lr": 7.380880051555006e-05}
{"train_loss": 0.2279512882232666, "global_step": 93171, "epoch": 1046, "lr": 7.380829072257254e-05}
{"train_loss": 0.1343737542629242, "global_step": 93172, "epoch": 1046, "lr": 7.38077809263943e-05}
{"train_loss": 0.2230074256658554, "global_step": 93173, "epoch": 1046, "lr": 7.38072711270154e-05}
{"train_loss": 0.18335527181625366, "global_step": 93174, "epoch": 1046, "lr": 7.380676132443589e-05}
{"train_loss": 0.20418748259544373, "global_step": 93175, "epoch": 1046, "lr": 7.380625151865586e-05}
{"train_loss": 0.21662698686122894, "global_step": 93176, "epoch": 1046, "lr": 7.380574170967536e-05}
{"train_loss": 0.1606685221195221, "global_step": 93177, "epoch": 1046, "lr": 7.380523189749447e-05}
{"train_loss": 0.13271082937717438, "global_step": 93178, "epoch": 1046, "lr": 7.380472208211326e-05}
{"train_loss": 0.17546093463897705, "global_step": 93179, "epoch": 1046, "lr": 7.38042122635318e-05}
{"train_loss": 0.1339343637228012, "global_step": 93180, "epoch": 1046, "lr": 7.380370244175015e-05}
{"train_loss": 0.18332752585411072, "global_step": 93181, "epoch": 1046, "lr": 7.380319261676839e-05}
{"train_loss": 0.1768448900305823, "global_step": 93182, "epoch": 1046, "lr": 7.380268278858656e-05, "val_loss": 4.449029922485352}
{"train_loss": 0.14468175172805786, "global_step": 93183, "epoch": 1047, "lr": 7.380217295720475e-05}
{"train_loss": 0.15837860107421875, "global_step": 93184, "epoch": 1047, "lr": 7.380166312262304e-05}
{"train_loss": 0.13976436853408813, "global_step": 93185, "epoch": 1047, "lr": 7.380115328484148e-05}
{"train_loss": 0.21755336225032806, "global_step": 93186, "epoch": 1047, "lr": 7.380064344386016e-05}
{"train_loss": 0.1957358419895172, "global_step": 93187, "epoch": 1047, "lr": 7.380013359967911e-05}
{"train_loss": 0.133788600564003, "global_step": 93188, "epoch": 1047, "lr": 7.379962375229843e-05}
{"train_loss": 0.21179163455963135, "global_step": 93189, "epoch": 1047, "lr": 7.379911390171819e-05}
{"train_loss": 0.15839658677577972, "global_step": 93190, "epoch": 1047, "lr": 7.379860404793845e-05}
{"train_loss": 0.12921911478042603, "global_step": 93191, "epoch": 1047, "lr": 7.379809419095927e-05}
{"train_loss": 0.16013619303703308, "global_step": 93192, "epoch": 1047, "lr": 7.379758433078073e-05}
{"train_loss": 0.20938549935817719, "global_step": 93193, "epoch": 1047, "lr": 7.379707446740289e-05}
{"train_loss": 0.24604053795337677, "global_step": 93194, "epoch": 1047, "lr": 7.379656460082582e-05}
{"train_loss": 0.15113995969295502, "global_step": 93195, "epoch": 1047, "lr": 7.37960547310496e-05}
{"train_loss": 0.17928726971149445, "global_step": 93196, "epoch": 1047, "lr": 7.379554485807429e-05}
{"train_loss": 0.21187233924865723, "global_step": 93197, "epoch": 1047, "lr": 7.379503498189997e-05}
{"train_loss": 0.12426667660474777, "global_step": 93198, "epoch": 1047, "lr": 7.379452510252668e-05}
{"train_loss": 0.126893550157547, "global_step": 93199, "epoch": 1047, "lr": 7.379401521995452e-05}
{"train_loss": 0.11555854976177216, "global_step": 93200, "epoch": 1047, "lr": 7.379350533418355e-05}
{"train_loss": 0.1998124122619629, "global_step": 93201, "epoch": 1047, "lr": 7.379299544521381e-05}
{"train_loss": 0.1732940822839737, "global_step": 93202, "epoch": 1047, "lr": 7.379248555304539e-05}
{"train_loss": 0.15701690316200256, "global_step": 93203, "epoch": 1047, "lr": 7.379197565767837e-05}
{"train_loss": 0.20701301097869873, "global_step": 93204, "epoch": 1047, "lr": 7.379146575911281e-05}
{"train_loss": 0.24284371733665466, "global_step": 93205, "epoch": 1047, "lr": 7.379095585734878e-05}
{"train_loss": 0.1631728559732437, "global_step": 93206, "epoch": 1047, "lr": 7.379044595238634e-05}
{"train_loss": 0.1776784509420395, "global_step": 93207, "epoch": 1047, "lr": 7.378993604422558e-05}
{"train_loss": 0.08928747475147247, "global_step": 93208, "epoch": 1047, "lr": 7.378942613286654e-05}
{"train_loss": 0.17182351648807526, "global_step": 93209, "epoch": 1047, "lr": 7.37889162183093e-05}
{"train_loss": 0.1074691116809845, "global_step": 93210, "epoch": 1047, "lr": 7.378840630055393e-05}
{"train_loss": 0.1424637883901596, "global_step": 93211, "epoch": 1047, "lr": 7.378789637960051e-05}
{"train_loss": 0.28145724534988403, "global_step": 93212, "epoch": 1047, "lr": 7.37873864554491e-05}
{"train_loss": 0.23826837539672852, "global_step": 93213, "epoch": 1047, "lr": 7.378687652809975e-05}
{"train_loss": 0.1806214451789856, "global_step": 93214, "epoch": 1047, "lr": 7.378636659755254e-05}
{"train_loss": 0.1342986524105072, "global_step": 93215, "epoch": 1047, "lr": 7.378585666380754e-05}
{"train_loss": 0.1099119558930397, "global_step": 93216, "epoch": 1047, "lr": 7.378534672686484e-05}
{"train_loss": 0.13327811658382416, "global_step": 93217, "epoch": 1047, "lr": 7.37848367867245e-05}
{"train_loss": 0.16176603734493256, "global_step": 93218, "epoch": 1047, "lr": 7.378432684338657e-05}
{"train_loss": 0.10065940767526627, "global_step": 93219, "epoch": 1047, "lr": 7.378381689685112e-05}
{"train_loss": 0.16705387830734253, "global_step": 93220, "epoch": 1047, "lr": 7.378330694711822e-05}
{"train_loss": 0.26053309440612793, "global_step": 93221, "epoch": 1047, "lr": 7.378279699418794e-05}
{"train_loss": 0.16901814937591553, "global_step": 93222, "epoch": 1047, "lr": 7.378228703806036e-05}
{"train_loss": 0.2122889757156372, "global_step": 93223, "epoch": 1047, "lr": 7.378177707873555e-05}
{"train_loss": 0.1313849538564682, "global_step": 93224, "epoch": 1047, "lr": 7.378126711621357e-05}
{"train_loss": 0.19645363092422485, "global_step": 93225, "epoch": 1047, "lr": 7.378075715049447e-05}
{"train_loss": 0.1239953488111496, "global_step": 93226, "epoch": 1047, "lr": 7.378024718157835e-05}
{"train_loss": 0.20801053941249847, "global_step": 93227, "epoch": 1047, "lr": 7.377973720946527e-05}
{"train_loss": 0.22605280578136444, "global_step": 93228, "epoch": 1047, "lr": 7.377922723415529e-05}
{"train_loss": 0.1832263171672821, "global_step": 93229, "epoch": 1047, "lr": 7.377871725564848e-05}
{"train_loss": 0.21267440915107727, "global_step": 93230, "epoch": 1047, "lr": 7.377820727394491e-05}
{"train_loss": 0.1602509617805481, "global_step": 93231, "epoch": 1047, "lr": 7.377769728904466e-05}
{"train_loss": 0.1372733861207962, "global_step": 93232, "epoch": 1047, "lr": 7.377718730094778e-05}
{"train_loss": 0.1922309845685959, "global_step": 93233, "epoch": 1047, "lr": 7.377667730965435e-05}
{"train_loss": 0.10092973709106445, "global_step": 93234, "epoch": 1047, "lr": 7.377616731516443e-05}
{"train_loss": 0.14438089728355408, "global_step": 93235, "epoch": 1047, "lr": 7.37756573174781e-05}
{"train_loss": 0.17387370765209198, "global_step": 93236, "epoch": 1047, "lr": 7.377514731659542e-05}
{"train_loss": 0.13330870866775513, "global_step": 93237, "epoch": 1047, "lr": 7.377463731251647e-05}
{"train_loss": 0.1405225694179535, "global_step": 93238, "epoch": 1047, "lr": 7.377412730524129e-05}
{"train_loss": 0.14932848513126373, "global_step": 93239, "epoch": 1047, "lr": 7.377361729476999e-05}
{"train_loss": 0.14842477440834045, "global_step": 93240, "epoch": 1047, "lr": 7.377310728110261e-05}
{"train_loss": 0.18713778257369995, "global_step": 93241, "epoch": 1047, "lr": 7.377259726423922e-05}
{"train_loss": 0.14055858552455902, "global_step": 93242, "epoch": 1047, "lr": 7.377208724417988e-05}
{"train_loss": 0.18659350275993347, "global_step": 93243, "epoch": 1047, "lr": 7.37715772209247e-05}
{"train_loss": 0.1434696912765503, "global_step": 93244, "epoch": 1047, "lr": 7.377106719447371e-05}
{"train_loss": 0.2550714612007141, "global_step": 93245, "epoch": 1047, "lr": 7.377055716482699e-05}
{"train_loss": 0.17179766297340393, "global_step": 93246, "epoch": 1047, "lr": 7.377004713198461e-05}
{"train_loss": 0.16548679769039154, "global_step": 93247, "epoch": 1047, "lr": 7.376953709594663e-05}
{"train_loss": 0.16487769782543182, "global_step": 93248, "epoch": 1047, "lr": 7.376902705671313e-05}
{"train_loss": 0.13187119364738464, "global_step": 93249, "epoch": 1047, "lr": 7.376851701428418e-05}
{"train_loss": 0.27727892994880676, "global_step": 93250, "epoch": 1047, "lr": 7.376800696865984e-05}
{"train_loss": 0.16446417570114136, "global_step": 93251, "epoch": 1047, "lr": 7.376749691984017e-05}
{"train_loss": 0.18653947114944458, "global_step": 93252, "epoch": 1047, "lr": 7.376698686782526e-05}
{"train_loss": 0.181909441947937, "global_step": 93253, "epoch": 1047, "lr": 7.376647681261517e-05}
{"train_loss": 0.17873933911323547, "global_step": 93254, "epoch": 1047, "lr": 7.376596675420997e-05}
{"train_loss": 0.17140784859657288, "global_step": 93255, "epoch": 1047, "lr": 7.376545669260971e-05}
{"train_loss": 0.2010977417230606, "global_step": 93256, "epoch": 1047, "lr": 7.376494662781449e-05}
{"train_loss": 0.1508362740278244, "global_step": 93257, "epoch": 1047, "lr": 7.376443655982435e-05}
{"train_loss": 0.15772214531898499, "global_step": 93258, "epoch": 1047, "lr": 7.376392648863938e-05}
{"train_loss": 0.17957055568695068, "global_step": 93259, "epoch": 1047, "lr": 7.376341641425963e-05}
{"train_loss": 0.17292684316635132, "global_step": 93260, "epoch": 1047, "lr": 7.376290633668519e-05}
{"train_loss": 0.16455870866775513, "global_step": 93261, "epoch": 1047, "lr": 7.376239625591612e-05}
{"train_loss": 0.24237340688705444, "global_step": 93262, "epoch": 1047, "lr": 7.376188617195247e-05}
{"train_loss": 0.1922014206647873, "global_step": 93263, "epoch": 1047, "lr": 7.376137608479433e-05}
{"train_loss": 0.11969047039747238, "global_step": 93264, "epoch": 1047, "lr": 7.376086599444176e-05}
{"train_loss": 0.16836054623126984, "global_step": 93265, "epoch": 1047, "lr": 7.376035590089484e-05}
{"train_loss": 0.18632975220680237, "global_step": 93266, "epoch": 1047, "lr": 7.375984580415363e-05}
{"train_loss": 0.1645689308643341, "global_step": 93267, "epoch": 1047, "lr": 7.375933570421818e-05}
{"train_loss": 0.12682369351387024, "global_step": 93268, "epoch": 1047, "lr": 7.37588256010886e-05}
{"train_loss": 0.18560263514518738, "global_step": 93269, "epoch": 1047, "lr": 7.375831549476492e-05}
{"train_loss": 0.25516757369041443, "global_step": 93270, "epoch": 1047, "lr": 7.375780538524724e-05}
{"train_loss": 0.17212699403923548, "global_step": 93271, "epoch": 1047, "lr": 7.375729527253558e-05, "val_loss": 4.516993522644043}
{"train_loss": 0.10553636401891708, "global_step": 93272, "epoch": 1048, "lr": 7.375678515663008e-05}
{"train_loss": 0.1770729124546051, "global_step": 93273, "epoch": 1048, "lr": 7.375627503753076e-05}
{"train_loss": 0.13514719903469086, "global_step": 93274, "epoch": 1048, "lr": 7.375576491523768e-05}
{"train_loss": 0.24805010855197906, "global_step": 93275, "epoch": 1048, "lr": 7.375525478975094e-05}
{"train_loss": 0.20685617625713348, "global_step": 93276, "epoch": 1048, "lr": 7.37547446610706e-05}
{"train_loss": 0.12428098171949387, "global_step": 93277, "epoch": 1048, "lr": 7.375423452919671e-05}
{"train_loss": 0.14944986999034882, "global_step": 93278, "epoch": 1048, "lr": 7.375372439412935e-05}
{"train_loss": 0.18920208513736725, "global_step": 93279, "epoch": 1048, "lr": 7.37532142558686e-05}
{"train_loss": 0.14768144488334656, "global_step": 93280, "epoch": 1048, "lr": 7.375270411441453e-05}
{"train_loss": 0.09357314556837082, "global_step": 93281, "epoch": 1048, "lr": 7.375219396976719e-05}
{"train_loss": 0.268398255109787, "global_step": 93282, "epoch": 1048, "lr": 7.375168382192666e-05}
{"train_loss": 0.244206503033638, "global_step": 93283, "epoch": 1048, "lr": 7.375117367089301e-05}
{"train_loss": 0.13693834841251373, "global_step": 93284, "epoch": 1048, "lr": 7.375066351666628e-05}
{"train_loss": 0.18559743463993073, "global_step": 93285, "epoch": 1048, "lr": 7.375015335924658e-05}
{"train_loss": 0.15151111781597137, "global_step": 93286, "epoch": 1048, "lr": 7.374964319863398e-05}
{"train_loss": 0.13262170553207397, "global_step": 93287, "epoch": 1048, "lr": 7.37491330348285e-05}
{"train_loss": 0.23863252997398376, "global_step": 93288, "epoch": 1048, "lr": 7.374862286783026e-05}
{"train_loss": 0.1345341056585312, "global_step": 93289, "epoch": 1048, "lr": 7.374811269763928e-05}
{"train_loss": 0.12165102362632751, "global_step": 93290, "epoch": 1048, "lr": 7.374760252425568e-05}
{"train_loss": 0.1463042050600052, "global_step": 93291, "epoch": 1048, "lr": 7.37470923476795e-05}
{"train_loss": 0.14852078258991241, "global_step": 93292, "epoch": 1048, "lr": 7.374658216791081e-05}
{"train_loss": 0.1778421849012375, "global_step": 93293, "epoch": 1048, "lr": 7.374607198494967e-05}
{"train_loss": 0.18433153629302979, "global_step": 93294, "epoch": 1048, "lr": 7.374556179879618e-05}
{"train_loss": 0.10380133241415024, "global_step": 93295, "epoch": 1048, "lr": 7.374505160945038e-05}
{"train_loss": 0.21901541948318481, "global_step": 93296, "epoch": 1048, "lr": 7.374454141691235e-05}
{"train_loss": 0.13351789116859436, "global_step": 93297, "epoch": 1048, "lr": 7.374403122118214e-05}
{"train_loss": 0.12039651721715927, "global_step": 93298, "epoch": 1048, "lr": 7.374352102225986e-05}
{"train_loss": 0.21314014494419098, "global_step": 93299, "epoch": 1048, "lr": 7.374301082014554e-05}
{"train_loss": 0.21751639246940613, "global_step": 93300, "epoch": 1048, "lr": 7.374250061483926e-05}
{"train_loss": 0.21208055317401886, "global_step": 93301, "epoch": 1048, "lr": 7.374199040634108e-05}
{"train_loss": 0.15621966123580933, "global_step": 93302, "epoch": 1048, "lr": 7.37414801946511e-05}
{"train_loss": 0.15279433131217957, "global_step": 93303, "epoch": 1048, "lr": 7.374096997976935e-05}
{"train_loss": 0.15308520197868347, "global_step": 93304, "epoch": 1048, "lr": 7.374045976169594e-05}
{"train_loss": 0.14734017848968506, "global_step": 93305, "epoch": 1048, "lr": 7.373994954043088e-05}
{"train_loss": 0.27758511900901794, "global_step": 93306, "epoch": 1048, "lr": 7.37394393159743e-05}
{"train_loss": 0.14572951197624207, "global_step": 93307, "epoch": 1048, "lr": 7.373892908832623e-05}
{"train_loss": 0.15362043678760529, "global_step": 93308, "epoch": 1048, "lr": 7.373841885748676e-05}
{"train_loss": 0.1953493356704712, "global_step": 93309, "epoch": 1048, "lr": 7.373790862345594e-05}
{"train_loss": 0.24144677817821503, "global_step": 93310, "epoch": 1048, "lr": 7.373739838623386e-05}
{"train_loss": 0.07974167168140411, "global_step": 93311, "epoch": 1048, "lr": 7.373688814582056e-05}
{"train_loss": 0.25397124886512756, "global_step": 93312, "epoch": 1048, "lr": 7.373637790221613e-05}
{"train_loss": 0.2362324297428131, "global_step": 93313, "epoch": 1048, "lr": 7.373586765542063e-05}
{"train_loss": 0.16026923060417175, "global_step": 93314, "epoch": 1048, "lr": 7.373535740543413e-05}
{"train_loss": 0.13568158447742462, "global_step": 93315, "epoch": 1048, "lr": 7.373484715225672e-05}
{"train_loss": 0.11555957049131393, "global_step": 93316, "epoch": 1048, "lr": 7.373433689588843e-05}
{"train_loss": 0.20060943067073822, "global_step": 93317, "epoch": 1048, "lr": 7.373382663632935e-05}
{"train_loss": 0.17410700023174286, "global_step": 93318, "epoch": 1048, "lr": 7.373331637357955e-05}
{"train_loss": 0.1653199940919876, "global_step": 93319, "epoch": 1048, "lr": 7.37328061076391e-05}
{"train_loss": 0.1583956778049469, "global_step": 93320, "epoch": 1048, "lr": 7.373229583850806e-05}
{"train_loss": 0.21564842760562897, "global_step": 93321, "epoch": 1048, "lr": 7.37317855661865e-05}
{"train_loss": 0.21135684847831726, "global_step": 93322, "epoch": 1048, "lr": 7.373127529067448e-05}
{"train_loss": 0.14066626131534576, "global_step": 93323, "epoch": 1048, "lr": 7.373076501197209e-05}
{"train_loss": 0.18274210393428802, "global_step": 93324, "epoch": 1048, "lr": 7.37302547300794e-05}
{"train_loss": 0.21596626937389374, "global_step": 93325, "epoch": 1048, "lr": 7.372974444499645e-05}
{"train_loss": 0.1457708775997162, "global_step": 93326, "epoch": 1048, "lr": 7.372923415672332e-05}
{"train_loss": 0.1699257493019104, "global_step": 93327, "epoch": 1048, "lr": 7.37287238652601e-05}
{"train_loss": 0.14289627969264984, "global_step": 93328, "epoch": 1048, "lr": 7.372821357060685e-05}
{"train_loss": 0.18114043772220612, "global_step": 93329, "epoch": 1048, "lr": 7.37277032727636e-05}
{"train_loss": 0.17915791273117065, "global_step": 93330, "epoch": 1048, "lr": 7.372719297173048e-05}
{"train_loss": 0.21533232927322388, "global_step": 93331, "epoch": 1048, "lr": 7.37266826675075e-05}
{"train_loss": 0.18817275762557983, "global_step": 93332, "epoch": 1048, "lr": 7.372617236009478e-05}
{"train_loss": 0.2651427090167999, "global_step": 93333, "epoch": 1048, "lr": 7.372566204949235e-05}
{"train_loss": 0.278238981962204, "global_step": 93334, "epoch": 1048, "lr": 7.372515173570029e-05}
{"train_loss": 0.18512724339962006, "global_step": 93335, "epoch": 1048, "lr": 7.372464141871869e-05}
{"train_loss": 0.15080349147319794, "global_step": 93336, "epoch": 1048, "lr": 7.372413109854759e-05}
{"train_loss": 0.1873321682214737, "global_step": 93337, "epoch": 1048, "lr": 7.372362077518707e-05}
{"train_loss": 0.1503315567970276, "global_step": 93338, "epoch": 1048, "lr": 7.372311044863721e-05}
{"train_loss": 0.19615307450294495, "global_step": 93339, "epoch": 1048, "lr": 7.372260011889805e-05}
{"train_loss": 0.18735961616039276, "global_step": 93340, "epoch": 1048, "lr": 7.37220897859697e-05}
{"train_loss": 0.1524774581193924, "global_step": 93341, "epoch": 1048, "lr": 7.372157944985219e-05}
{"train_loss": 0.24341513216495514, "global_step": 93342, "epoch": 1048, "lr": 7.37210691105456e-05}
{"train_loss": 0.25808054208755493, "global_step": 93343, "epoch": 1048, "lr": 7.372055876805e-05}
{"train_loss": 0.16673311591148376, "global_step": 93344, "epoch": 1048, "lr": 7.372004842236546e-05}
{"train_loss": 0.16748735308647156, "global_step": 93345, "epoch": 1048, "lr": 7.371953807349204e-05}
{"train_loss": 0.25049784779548645, "global_step": 93346, "epoch": 1048, "lr": 7.371902772142984e-05}
{"train_loss": 0.18757358193397522, "global_step": 93347, "epoch": 1048, "lr": 7.371851736617887e-05}
{"train_loss": 0.11697567254304886, "global_step": 93348, "epoch": 1048, "lr": 7.371800700773928e-05}
{"train_loss": 0.13024508953094482, "global_step": 93349, "epoch": 1048, "lr": 7.371749664611108e-05}
{"train_loss": 0.2607898712158203, "global_step": 93350, "epoch": 1048, "lr": 7.371698628129433e-05}
{"train_loss": 0.14444860816001892, "global_step": 93351, "epoch": 1048, "lr": 7.371647591328913e-05}
{"train_loss": 0.1939280778169632, "global_step": 93352, "epoch": 1048, "lr": 7.371596554209554e-05}
{"train_loss": 0.08071059733629227, "global_step": 93353, "epoch": 1048, "lr": 7.371545516771363e-05}
{"train_loss": 0.23408013582229614, "global_step": 93354, "epoch": 1048, "lr": 7.371494479014346e-05}
{"train_loss": 0.1341356635093689, "global_step": 93355, "epoch": 1048, "lr": 7.371443440938512e-05}
{"train_loss": 0.14973674714565277, "global_step": 93356, "epoch": 1048, "lr": 7.371392402543866e-05}
{"train_loss": 0.1453956961631775, "global_step": 93357, "epoch": 1048, "lr": 7.371341363830414e-05}
{"train_loss": 0.23066070675849915, "global_step": 93358, "epoch": 1048, "lr": 7.371290324798165e-05}
{"train_loss": 0.16695313155651093, "global_step": 93359, "epoch": 1048, "lr": 7.371239285447125e-05}
{"train_loss": 0.17693433282750376, "global_step": 93360, "epoch": 1048, "lr": 7.371188245777299e-05, "val_loss": 4.408931255340576}
{"train_loss": 0.2475515604019165, "global_step": 93361, "epoch": 1049, "lr": 7.371137205788698e-05}
{"train_loss": 0.18269094824790955, "global_step": 93362, "epoch": 1049, "lr": 7.371086165481325e-05}
{"train_loss": 0.21438424289226532, "global_step": 93363, "epoch": 1049, "lr": 7.371035124855189e-05}
{"train_loss": 0.19488859176635742, "global_step": 93364, "epoch": 1049, "lr": 7.370984083910296e-05}
{"train_loss": 0.1742088347673416, "global_step": 93365, "epoch": 1049, "lr": 7.370933042646652e-05}
{"train_loss": 0.17540618777275085, "global_step": 93366, "epoch": 1049, "lr": 7.370882001064266e-05}
{"train_loss": 0.21819105744361877, "global_step": 93367, "epoch": 1049, "lr": 7.370830959163144e-05}
{"train_loss": 0.16571073234081268, "global_step": 93368, "epoch": 1049, "lr": 7.370779916943293e-05}
{"train_loss": 0.16038447618484497, "global_step": 93369, "epoch": 1049, "lr": 7.37072887440472e-05}
{"train_loss": 0.3003065884113312, "global_step": 93370, "epoch": 1049, "lr": 7.37067783154743e-05}
{"train_loss": 0.13472850620746613, "global_step": 93371, "epoch": 1049, "lr": 7.370626788371433e-05}
{"train_loss": 0.14528688788414001, "global_step": 93372, "epoch": 1049, "lr": 7.370575744876732e-05}
{"train_loss": 0.22555264830589294, "global_step": 93373, "epoch": 1049, "lr": 7.370524701063337e-05}
{"train_loss": 0.1930702179670334, "global_step": 93374, "epoch": 1049, "lr": 7.370473656931254e-05}
{"train_loss": 0.14552994072437286, "global_step": 93375, "epoch": 1049, "lr": 7.37042261248049e-05}
{"train_loss": 0.1514884978532791, "global_step": 93376, "epoch": 1049, "lr": 7.37037156771105e-05}
{"train_loss": 0.14332567155361176, "global_step": 93377, "epoch": 1049, "lr": 7.370320522622944e-05}
{"train_loss": 0.18956103920936584, "global_step": 93378, "epoch": 1049, "lr": 7.370269477216177e-05}
{"train_loss": 0.18625977635383606, "global_step": 93379, "epoch": 1049, "lr": 7.370218431490755e-05}
{"train_loss": 0.159352108836174, "global_step": 93380, "epoch": 1049, "lr": 7.370167385446687e-05}
{"train_loss": 0.17428898811340332, "global_step": 93381, "epoch": 1049, "lr": 7.370116339083979e-05}
{"train_loss": 0.17132890224456787, "global_step": 93382, "epoch": 1049, "lr": 7.370065292402638e-05}
{"train_loss": 0.17251747846603394, "global_step": 93383, "epoch": 1049, "lr": 7.370014245402669e-05}
{"train_loss": 0.25333428382873535, "global_step": 93384, "epoch": 1049, "lr": 7.369963198084083e-05}
{"train_loss": 0.3273732364177704, "global_step": 93385, "epoch": 1049, "lr": 7.369912150446881e-05}
{"train_loss": 0.13426080346107483, "global_step": 93386, "epoch": 1049, "lr": 7.369861102491076e-05}
{"train_loss": 0.2400834560394287, "global_step": 93387, "epoch": 1049, "lr": 7.369810054216673e-05}
{"train_loss": 0.15456920862197876, "global_step": 93388, "epoch": 1049, "lr": 7.369759005623675e-05}
{"train_loss": 0.20902541279792786, "global_step": 93389, "epoch": 1049, "lr": 7.369707956712093e-05}
{"train_loss": 0.18296092748641968, "global_step": 93390, "epoch": 1049, "lr": 7.369656907481933e-05}
{"train_loss": 0.14993827044963837, "global_step": 93391, "epoch": 1049, "lr": 7.369605857933201e-05}
{"train_loss": 0.13758869469165802, "global_step": 93392, "epoch": 1049, "lr": 7.369554808065904e-05}
{"train_loss": 0.20544098317623138, "global_step": 93393, "epoch": 1049, "lr": 7.36950375788005e-05}
{"train_loss": 0.28904393315315247, "global_step": 93394, "epoch": 1049, "lr": 7.369452707375644e-05}
{"train_loss": 0.13126957416534424, "global_step": 93395, "epoch": 1049, "lr": 7.369401656552696e-05}
{"train_loss": 0.1379210203886032, "global_step": 93396, "epoch": 1049, "lr": 7.36935060541121e-05}
{"train_loss": 0.17296525835990906, "global_step": 93397, "epoch": 1049, "lr": 7.369299553951192e-05}
{"train_loss": 0.17802052199840546, "global_step": 93398, "epoch": 1049, "lr": 7.369248502172652e-05}
{"train_loss": 0.10116926580667496, "global_step": 93399, "epoch": 1049, "lr": 7.369197450075595e-05}
{"train_loss": 0.18279248476028442, "global_step": 93400, "epoch": 1049, "lr": 7.36914639766003e-05}
{"train_loss": 0.16555377840995789, "global_step": 93401, "epoch": 1049, "lr": 7.36909534492596e-05}
{"train_loss": 0.1850721687078476, "global_step": 93402, "epoch": 1049, "lr": 7.369044291873394e-05}
{"train_loss": 0.16497497260570526, "global_step": 93403, "epoch": 1049, "lr": 7.368993238502339e-05}
{"train_loss": 0.1581040471792221, "global_step": 93404, "epoch": 1049, "lr": 7.368942184812802e-05}
{"train_loss": 0.2133278101682663, "global_step": 93405, "epoch": 1049, "lr": 7.368891130804791e-05}
{"train_loss": 0.18464800715446472, "global_step": 93406, "epoch": 1049, "lr": 7.368840076478311e-05}
{"train_loss": 0.23380309343338013, "global_step": 93407, "epoch": 1049, "lr": 7.368789021833367e-05}
{"train_loss": 0.18720729649066925, "global_step": 93408, "epoch": 1049, "lr": 7.36873796686997e-05}
{"train_loss": 0.1343645453453064, "global_step": 93409, "epoch": 1049, "lr": 7.368686911588125e-05}
{"train_loss": 0.1417161077260971, "global_step": 93410, "epoch": 1049, "lr": 7.36863585598784e-05}
{"train_loss": 0.20463499426841736, "global_step": 93411, "epoch": 1049, "lr": 7.368584800069119e-05}
{"train_loss": 0.16580575704574585, "global_step": 93412, "epoch": 1049, "lr": 7.368533743831971e-05}
{"train_loss": 0.17384007573127747, "global_step": 93413, "epoch": 1049, "lr": 7.368482687276403e-05}
{"train_loss": 0.1849883496761322, "global_step": 93414, "epoch": 1049, "lr": 7.368431630402421e-05}
{"train_loss": 0.0970214456319809, "global_step": 93415, "epoch": 1049, "lr": 7.368380573210034e-05}
{"train_loss": 0.1758277714252472, "global_step": 93416, "epoch": 1049, "lr": 7.368329515699244e-05}
{"train_loss": 0.1915288120508194, "global_step": 93417, "epoch": 1049, "lr": 7.368278457870062e-05}
{"train_loss": 0.11082248389720917, "global_step": 93418, "epoch": 1049, "lr": 7.368227399722495e-05}
{"train_loss": 0.24751931428909302, "global_step": 93419, "epoch": 1049, "lr": 7.368176341256548e-05}
{"train_loss": 0.21312811970710754, "global_step": 93420, "epoch": 1049, "lr": 7.368125282472227e-05}
{"train_loss": 0.20619161427021027, "global_step": 93421, "epoch": 1049, "lr": 7.368074223369542e-05}
{"train_loss": 0.228305384516716, "global_step": 93422, "epoch": 1049, "lr": 7.368023163948499e-05}
{"train_loss": 0.2039252519607544, "global_step": 93423, "epoch": 1049, "lr": 7.367972104209102e-05}
{"train_loss": 0.1902141571044922, "global_step": 93424, "epoch": 1049, "lr": 7.367921044151362e-05}
{"train_loss": 0.2048458755016327, "global_step": 93425, "epoch": 1049, "lr": 7.367869983775283e-05}
{"train_loss": 0.1715010404586792, "global_step": 93426, "epoch": 1049, "lr": 7.367818923080874e-05}
{"train_loss": 0.17032171785831451, "global_step": 93427, "epoch": 1049, "lr": 7.36776786206814e-05}
{"train_loss": 0.1416001319885254, "global_step": 93428, "epoch": 1049, "lr": 7.367716800737088e-05}
{"train_loss": 0.18152561783790588, "global_step": 93429, "epoch": 1049, "lr": 7.367665739087725e-05}
{"train_loss": 0.11995839327573776, "global_step": 93430, "epoch": 1049, "lr": 7.367614677120058e-05}
{"train_loss": 0.1694142073392868, "global_step": 93431, "epoch": 1049, "lr": 7.367563614834096e-05}
{"train_loss": 0.1486588418483734, "global_step": 93432, "epoch": 1049, "lr": 7.367512552229842e-05}
{"train_loss": 0.15585559606552124, "global_step": 93433, "epoch": 1049, "lr": 7.367461489307306e-05}
{"train_loss": 0.2952992618083954, "global_step": 93434, "epoch": 1049, "lr": 7.367410426066492e-05}
{"train_loss": 0.2575571537017822, "global_step": 93435, "epoch": 1049, "lr": 7.36735936250741e-05}
{"train_loss": 0.28898191452026367, "global_step": 93436, "epoch": 1049, "lr": 7.367308298630065e-05}
{"train_loss": 0.20706267654895782, "global_step": 93437, "epoch": 1049, "lr": 7.367257234434464e-05}
{"train_loss": 0.21544893085956573, "global_step": 93438, "epoch": 1049, "lr": 7.367206169920612e-05}
{"train_loss": 0.1330646574497223, "global_step": 93439, "epoch": 1049, "lr": 7.367155105088521e-05}
{"train_loss": 0.18854248523712158, "global_step": 93440, "epoch": 1049, "lr": 7.367104039938195e-05}
{"train_loss": 0.2713266909122467, "global_step": 93441, "epoch": 1049, "lr": 7.36705297446964e-05}
{"train_loss": 0.18728412687778473, "global_step": 93442, "epoch": 1049, "lr": 7.367001908682863e-05}
{"train_loss": 0.201076477766037, "global_step": 93443, "epoch": 1049, "lr": 7.366950842577872e-05}
{"train_loss": 0.18007592856884003, "global_step": 93444, "epoch": 1049, "lr": 7.366899776154674e-05}
{"train_loss": 0.2640029788017273, "global_step": 93445, "epoch": 1049, "lr": 7.366848709413274e-05}
{"train_loss": 0.15024596452713013, "global_step": 93446, "epoch": 1049, "lr": 7.36679764235368e-05}
{"train_loss": 0.22878985106945038, "global_step": 93447, "epoch": 1049, "lr": 7.3667465749759e-05}
{"train_loss": 0.11830844730138779, "global_step": 93448, "epoch": 1049, "lr": 7.36669550727994e-05}
{"train_loss": 0.18687396211905427, "global_step": 93449, "epoch": 1049, "lr": 7.366644439265804e-05, "val_loss": 4.33919620513916}
{"train_loss": 0.2143043875694275, "global_step": 93450, "epoch": 1050, "lr": 7.366593370933504e-05}
{"train_loss": 0.1660548746585846, "global_step": 93451, "epoch": 1050, "lr": 7.366542302283044e-05}
{"train_loss": 0.18090347945690155, "global_step": 93452, "epoch": 1050, "lr": 7.36649123331443e-05}
{"train_loss": 0.1128193736076355, "global_step": 93453, "epoch": 1050, "lr": 7.366440164027672e-05}
{"train_loss": 0.3163481652736664, "global_step": 93454, "epoch": 1050, "lr": 7.366389094422773e-05}
{"train_loss": 0.2734217643737793, "global_step": 93455, "epoch": 1050, "lr": 7.366338024499742e-05}
{"train_loss": 0.18551424145698547, "global_step": 93456, "epoch": 1050, "lr": 7.366286954258586e-05}
{"train_loss": 0.15445862710475922, "global_step": 93457, "epoch": 1050, "lr": 7.366235883699312e-05}
{"train_loss": 0.1423248052597046, "global_step": 93458, "epoch": 1050, "lr": 7.366184812821925e-05}
{"train_loss": 0.18917390704154968, "global_step": 93459, "epoch": 1050, "lr": 7.366133741626435e-05}
{"train_loss": 0.14975081384181976, "global_step": 93460, "epoch": 1050, "lr": 7.366082670112849e-05}
{"train_loss": 0.1418895721435547, "global_step": 93461, "epoch": 1050, "lr": 7.366031598281169e-05}
{"train_loss": 0.17084281146526337, "global_step": 93462, "epoch": 1050, "lr": 7.365980526131406e-05}
{"train_loss": 0.1513463705778122, "global_step": 93463, "epoch": 1050, "lr": 7.365929453663565e-05}
{"train_loss": 0.13378401100635529, "global_step": 93464, "epoch": 1050, "lr": 7.365878380877656e-05}
{"train_loss": 0.26426562666893005, "global_step": 93465, "epoch": 1050, "lr": 7.36582730777368e-05}
{"train_loss": 0.20167414844036102, "global_step": 93466, "epoch": 1050, "lr": 7.36577623435165e-05}
{"train_loss": 0.18613576889038086, "global_step": 93467, "epoch": 1050, "lr": 7.365725160611568e-05}
{"train_loss": 0.20284922420978546, "global_step": 93468, "epoch": 1050, "lr": 7.365674086553445e-05}
{"train_loss": 0.2002304494380951, "global_step": 93469, "epoch": 1050, "lr": 7.365623012177285e-05}
{"train_loss": 0.20018775761127472, "global_step": 93470, "epoch": 1050, "lr": 7.365571937483096e-05}
{"train_loss": 0.22532179951667786, "global_step": 93471, "epoch": 1050, "lr": 7.365520862470884e-05}
{"train_loss": 0.13920976221561432, "global_step": 93472, "epoch": 1050, "lr": 7.365469787140657e-05}
{"train_loss": 0.19710877537727356, "global_step": 93473, "epoch": 1050, "lr": 7.365418711492422e-05}
{"train_loss": 0.21883675456047058, "global_step": 93474, "epoch": 1050, "lr": 7.365367635526184e-05}
{"train_loss": 0.17352965474128723, "global_step": 93475, "epoch": 1050, "lr": 7.365316559241953e-05}
{"train_loss": 0.18488061428070068, "global_step": 93476, "epoch": 1050, "lr": 7.365265482639732e-05}
{"train_loss": 0.27907994389533997, "global_step": 93477, "epoch": 1050, "lr": 7.365214405719531e-05}
{"train_loss": 0.1779150664806366, "global_step": 93478, "epoch": 1050, "lr": 7.365163328481358e-05}
{"train_loss": 0.20212103426456451, "global_step": 93479, "epoch": 1050, "lr": 7.365112250925215e-05}
{"train_loss": 0.17344217002391815, "global_step": 93480, "epoch": 1050, "lr": 7.365061173051111e-05}
{"train_loss": 0.1267484426498413, "global_step": 93481, "epoch": 1050, "lr": 7.365010094859054e-05}
{"train_loss": 0.23348000645637512, "global_step": 93482, "epoch": 1050, "lr": 7.364959016349052e-05}
{"train_loss": 0.14164744317531586, "global_step": 93483, "epoch": 1050, "lr": 7.364907937521107e-05}
{"train_loss": 0.16194851696491241, "global_step": 93484, "epoch": 1050, "lr": 7.36485685837523e-05}
{"train_loss": 0.18660616874694824, "global_step": 93485, "epoch": 1050, "lr": 7.364805778911428e-05}
{"train_loss": 0.19215607643127441, "global_step": 93486, "epoch": 1050, "lr": 7.364754699129707e-05}
{"train_loss": 0.17060735821723938, "global_step": 93487, "epoch": 1050, "lr": 7.364703619030072e-05}
{"train_loss": 0.17921759188175201, "global_step": 93488, "epoch": 1050, "lr": 7.364652538612532e-05}
{"train_loss": 0.20012985169887543, "global_step": 93489, "epoch": 1050, "lr": 7.364601457877094e-05}
{"train_loss": 0.1980048567056656, "global_step": 93490, "epoch": 1050, "lr": 7.364550376823762e-05}
{"train_loss": 0.2089385837316513, "global_step": 93491, "epoch": 1050, "lr": 7.364499295452548e-05}
{"train_loss": 0.19643738865852356, "global_step": 93492, "epoch": 1050, "lr": 7.364448213763453e-05}
{"train_loss": 0.18008065223693848, "global_step": 93493, "epoch": 1050, "lr": 7.364397131756487e-05}
{"train_loss": 0.13260695338249207, "global_step": 93494, "epoch": 1050, "lr": 7.364346049431659e-05}
{"train_loss": 0.13895894587039948, "global_step": 93495, "epoch": 1050, "lr": 7.364294966788972e-05}
{"train_loss": 0.13841092586517334, "global_step": 93496, "epoch": 1050, "lr": 7.364243883828434e-05}
{"train_loss": 0.19024339318275452, "global_step": 93497, "epoch": 1050, "lr": 7.364192800550052e-05}
{"train_loss": 0.12211738526821136, "global_step": 93498, "epoch": 1050, "lr": 7.364141716953834e-05}
{"train_loss": 0.13553409278392792, "global_step": 93499, "epoch": 1050, "lr": 7.364090633039786e-05}
{"train_loss": 0.20018549263477325, "global_step": 93500, "epoch": 1050, "lr": 7.364039548807915e-05}
{"train_loss": 0.17668618261814117, "global_step": 93501, "epoch": 1050, "lr": 7.363988464258227e-05}
{"train_loss": 0.19008031487464905, "global_step": 93502, "epoch": 1050, "lr": 7.36393737939073e-05}
{"train_loss": 0.21958939731121063, "global_step": 93503, "epoch": 1050, "lr": 7.363886294205431e-05}
{"train_loss": 0.15596148371696472, "global_step": 93504, "epoch": 1050, "lr": 7.363835208702337e-05}
{"train_loss": 0.16768084466457367, "global_step": 93505, "epoch": 1050, "lr": 7.363784122881451e-05}
{"train_loss": 0.2574322819709778, "global_step": 93506, "epoch": 1050, "lr": 7.363733036742786e-05}
{"train_loss": 0.21223177015781403, "global_step": 93507, "epoch": 1050, "lr": 7.363681950286345e-05}
{"train_loss": 0.0921192467212677, "global_step": 93508, "epoch": 1050, "lr": 7.363630863512135e-05}
{"train_loss": 0.20389701426029205, "global_step": 93509, "epoch": 1050, "lr": 7.363579776420165e-05}
{"train_loss": 0.17320190370082855, "global_step": 93510, "epoch": 1050, "lr": 7.363528689010439e-05}
{"train_loss": 0.25342801213264465, "global_step": 93511, "epoch": 1050, "lr": 7.363477601282964e-05}
{"train_loss": 0.16277402639389038, "global_step": 93512, "epoch": 1050, "lr": 7.363426513237751e-05}
{"train_loss": 0.14801235496997833, "global_step": 93513, "epoch": 1050, "lr": 7.363375424874803e-05}
{"train_loss": 0.23837155103683472, "global_step": 93514, "epoch": 1050, "lr": 7.363324336194128e-05}
{"train_loss": 0.312835156917572, "global_step": 93515, "epoch": 1050, "lr": 7.363273247195734e-05}
{"train_loss": 0.15322288870811462, "global_step": 93516, "epoch": 1050, "lr": 7.363222157879626e-05}
{"train_loss": 0.2116721272468567, "global_step": 93517, "epoch": 1050, "lr": 7.363171068245811e-05}
{"train_loss": 0.13744887709617615, "global_step": 93518, "epoch": 1050, "lr": 7.363119978294297e-05}
{"train_loss": 0.11749909818172455, "global_step": 93519, "epoch": 1050, "lr": 7.363068888025091e-05}
{"train_loss": 0.1200844943523407, "global_step": 93520, "epoch": 1050, "lr": 7.363017797438199e-05}
{"train_loss": 0.15456295013427734, "global_step": 93521, "epoch": 1050, "lr": 7.362966706533627e-05}
{"train_loss": 0.1505085676908493, "global_step": 93522, "epoch": 1050, "lr": 7.362915615311385e-05}
{"train_loss": 0.13534928858280182, "global_step": 93523, "epoch": 1050, "lr": 7.362864523771476e-05}
{"train_loss": 0.19544494152069092, "global_step": 93524, "epoch": 1050, "lr": 7.362813431913908e-05}
{"train_loss": 0.1695227324962616, "global_step": 93525, "epoch": 1050, "lr": 7.36276233973869e-05}
{"train_loss": 0.16521255671977997, "global_step": 93526, "epoch": 1050, "lr": 7.362711247245827e-05}
{"train_loss": 0.1463824212551117, "global_step": 93527, "epoch": 1050, "lr": 7.362660154435326e-05}
{"train_loss": 0.24606691300868988, "global_step": 93528, "epoch": 1050, "lr": 7.362609061307194e-05}
{"train_loss": 0.21191464364528656, "global_step": 93529, "epoch": 1050, "lr": 7.362557967861438e-05}
{"train_loss": 0.18284787237644196, "global_step": 93530, "epoch": 1050, "lr": 7.362506874098065e-05}
{"train_loss": 0.23059676587581635, "global_step": 93531, "epoch": 1050, "lr": 7.362455780017081e-05}
{"train_loss": 0.1702880710363388, "global_step": 93532, "epoch": 1050, "lr": 7.362404685618495e-05}
{"train_loss": 0.19794736802577972, "global_step": 93533, "epoch": 1050, "lr": 7.362353590902313e-05}
{"train_loss": 0.19004756212234497, "global_step": 93534, "epoch": 1050, "lr": 7.36230249586854e-05}
{"train_loss": 0.24394798278808594, "global_step": 93535, "epoch": 1050, "lr": 7.362251400517183e-05}
{"train_loss": 0.15851879119873047, "global_step": 93536, "epoch": 1050, "lr": 7.362200304848252e-05}
{"train_loss": 0.13675621151924133, "global_step": 93537, "epoch": 1050, "lr": 7.362149208861753e-05}
{"train_loss": 0.18324139245440452, "global_step": 93538, "epoch": 1050, "lr": 7.362098112557689e-05, "train/sim_max_reward_0": 0.5308347243667011, "train/sim_max_reward_1": 0.4886914499759492, "train/sim_max_reward_2": 0.419131024436123, "train/sim_max_reward_3": 0.15553950266767397, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.705237204602562, "test/sim_max_reward_4500000": 0.42431579196910973, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9413146961245555, "test/sim_max_reward_4500003": 0.6294427138477154, "test/sim_max_reward_4500004": 0.992011378494715, "test/sim_max_reward_4500005": 0.40365679925195824, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.5424263742744286, "test/sim_max_reward_4500009": 0.40972292395157534, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9042143990912864, "test/sim_max_reward_4500012": 0.3875099899520281, "test/sim_max_reward_4500013": 1.0, "test/sim_max_reward_4500014": 0.9438931092569343, "test/sim_max_reward_4500015": 0.0, "test/sim_max_reward_4500016": 0.6320718655870232, "test/sim_max_reward_4500017": 0.5533807798457241, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.5800868616584649, "test/sim_max_reward_4500020": 0.5904873484389952, "test/sim_max_reward_4500021": 0.972363417578888, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 1.0, "test/sim_max_reward_4500024": 0.7521020547178572, "test/sim_max_reward_4500025": 0.9989640195236946, "test/sim_max_reward_4500026": 0.9078966243759671, "test/sim_max_reward_4500027": 0.9871456141139626, "test/sim_max_reward_4500028": 0.4984870128935873, "test/sim_max_reward_4500029": 0.28262177622731544, "test/sim_max_reward_4500030": 0.9327495593717288, "test/sim_max_reward_4500031": 0.5163729304752814, "test/sim_max_reward_4500032": 0.6954962274017154, "test/sim_max_reward_4500033": 0.6163103253501617, "test/sim_max_reward_4500034": 0.8199137480767814, "test/sim_max_reward_4500035": 1.0, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.6716401670537656, "test/sim_max_reward_4500038": 0.37446150277769363, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9878023980957295, "test/sim_max_reward_4500041": 0.44657421694186106, "test/sim_max_reward_4500042": 0.43510133689754205, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.7022868276902491, "test/sim_max_reward_4500046": 0.3562485659782866, "test/sim_max_reward_4500047": 0.6499510906196273, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.6301504659846834, "train/mean_score": 0.5499056510081682, "test/mean_score": 0.6833834982778179, "val_loss": 4.35257625579834, "train_action_mse_error": 29.52131462097168}
{"train_loss": 0.22452661395072937, "global_step": 93539, "epoch": 1051, "lr": 7.36204701593607e-05}
{"train_loss": 0.12686315178871155, "global_step": 93540, "epoch": 1051, "lr": 7.361995918996906e-05}
{"train_loss": 0.17483936250209808, "global_step": 93541, "epoch": 1051, "lr": 7.361944821740197e-05}
{"train_loss": 0.16296353936195374, "global_step": 93542, "epoch": 1051, "lr": 7.361893724165955e-05}
{"train_loss": 0.10829273611307144, "global_step": 93543, "epoch": 1051, "lr": 7.361842626274186e-05}
{"train_loss": 0.161383718252182, "global_step": 93544, "epoch": 1051, "lr": 7.361791528064894e-05}
{"train_loss": 0.134726420044899, "global_step": 93545, "epoch": 1051, "lr": 7.361740429538089e-05}
{"train_loss": 0.15391261875629425, "global_step": 93546, "epoch": 1051, "lr": 7.361689330693776e-05}
{"train_loss": 0.12397587299346924, "global_step": 93547, "epoch": 1051, "lr": 7.361638231531962e-05}
{"train_loss": 0.16005411744117737, "global_step": 93548, "epoch": 1051, "lr": 7.361587132052656e-05}
{"train_loss": 0.16920620203018188, "global_step": 93549, "epoch": 1051, "lr": 7.361536032255862e-05}
{"train_loss": 0.1194930151104927, "global_step": 93550, "epoch": 1051, "lr": 7.361484932141591e-05}
{"train_loss": 0.18705742061138153, "global_step": 93551, "epoch": 1051, "lr": 7.361433831709845e-05}
{"train_loss": 0.19939035177230835, "global_step": 93552, "epoch": 1051, "lr": 7.361382730960633e-05}
{"train_loss": 0.18867965042591095, "global_step": 93553, "epoch": 1051, "lr": 7.361331629893961e-05}
{"train_loss": 0.24696438014507294, "global_step": 93554, "epoch": 1051, "lr": 7.36128052850984e-05}
{"train_loss": 0.20494802296161652, "global_step": 93555, "epoch": 1051, "lr": 7.36122942680827e-05}
{"train_loss": 0.19512543082237244, "global_step": 93556, "epoch": 1051, "lr": 7.361178324789263e-05}
{"train_loss": 0.1769103854894638, "global_step": 93557, "epoch": 1051, "lr": 7.361127222452825e-05}
{"train_loss": 0.2507120668888092, "global_step": 93558, "epoch": 1051, "lr": 7.361076119798962e-05}
{"train_loss": 0.17558903992176056, "global_step": 93559, "epoch": 1051, "lr": 7.361025016827681e-05}
{"train_loss": 0.21537885069847107, "global_step": 93560, "epoch": 1051, "lr": 7.360973913538988e-05}
{"train_loss": 0.2089536041021347, "global_step": 93561, "epoch": 1051, "lr": 7.360922809932893e-05}
{"train_loss": 0.23462456464767456, "global_step": 93562, "epoch": 1051, "lr": 7.360871706009399e-05}
{"train_loss": 0.23410235345363617, "global_step": 93563, "epoch": 1051, "lr": 7.360820601768515e-05}
{"train_loss": 0.2607642710208893, "global_step": 93564, "epoch": 1051, "lr": 7.360769497210246e-05}
{"train_loss": 0.18742094933986664, "global_step": 93565, "epoch": 1051, "lr": 7.360718392334601e-05}
{"train_loss": 0.11994486302137375, "global_step": 93566, "epoch": 1051, "lr": 7.360667287141587e-05}
{"train_loss": 0.1543077975511551, "global_step": 93567, "epoch": 1051, "lr": 7.36061618163121e-05}
{"train_loss": 0.19115561246871948, "global_step": 93568, "epoch": 1051, "lr": 7.360565075803476e-05}
{"train_loss": 0.13698938488960266, "global_step": 93569, "epoch": 1051, "lr": 7.360513969658394e-05}
{"train_loss": 0.25308141112327576, "global_step": 93570, "epoch": 1051, "lr": 7.360462863195969e-05}
{"train_loss": 0.15283316373825073, "global_step": 93571, "epoch": 1051, "lr": 7.36041175641621e-05}
{"train_loss": 0.23259033262729645, "global_step": 93572, "epoch": 1051, "lr": 7.36036064931912e-05}
{"train_loss": 0.1502707600593567, "global_step": 93573, "epoch": 1051, "lr": 7.36030954190471e-05}
{"train_loss": 0.12092800438404083, "global_step": 93574, "epoch": 1051, "lr": 7.360258434172985e-05}
{"train_loss": 0.18014059960842133, "global_step": 93575, "epoch": 1051, "lr": 7.360207326123951e-05}
{"train_loss": 0.18475356698036194, "global_step": 93576, "epoch": 1051, "lr": 7.360156217757617e-05}
{"train_loss": 0.1298171579837799, "global_step": 93577, "epoch": 1051, "lr": 7.360105109073988e-05}
{"train_loss": 0.15295720100402832, "global_step": 93578, "epoch": 1051, "lr": 7.360054000073071e-05}
{"train_loss": 0.2358963042497635, "global_step": 93579, "epoch": 1051, "lr": 7.360002890754875e-05}
{"train_loss": 0.16231924295425415, "global_step": 93580, "epoch": 1051, "lr": 7.359951781119405e-05}
{"train_loss": 0.1631905734539032, "global_step": 93581, "epoch": 1051, "lr": 7.359900671166669e-05}
{"train_loss": 0.21425235271453857, "global_step": 93582, "epoch": 1051, "lr": 7.359849560896672e-05}
{"train_loss": 0.17440661787986755, "global_step": 93583, "epoch": 1051, "lr": 7.359798450309421e-05}
{"train_loss": 0.13758237659931183, "global_step": 93584, "epoch": 1051, "lr": 7.359747339404926e-05}
{"train_loss": 0.19191144406795502, "global_step": 93585, "epoch": 1051, "lr": 7.35969622818319e-05}
{"train_loss": 0.15334759652614594, "global_step": 93586, "epoch": 1051, "lr": 7.359645116644222e-05}
{"train_loss": 0.18765608966350555, "global_step": 93587, "epoch": 1051, "lr": 7.359594004788028e-05}
{"train_loss": 0.18203715980052948, "global_step": 93588, "epoch": 1051, "lr": 7.359542892614618e-05}
{"train_loss": 0.15569640696048737, "global_step": 93589, "epoch": 1051, "lr": 7.359491780123993e-05}
{"train_loss": 0.1658579409122467, "global_step": 93590, "epoch": 1051, "lr": 7.359440667316164e-05}
{"train_loss": 0.17784754931926727, "global_step": 93591, "epoch": 1051, "lr": 7.359389554191138e-05}
{"train_loss": 0.1591159850358963, "global_step": 93592, "epoch": 1051, "lr": 7.359338440748921e-05}
{"train_loss": 0.22846455872058868, "global_step": 93593, "epoch": 1051, "lr": 7.359287326989518e-05}
{"train_loss": 0.21828459203243256, "global_step": 93594, "epoch": 1051, "lr": 7.359236212912939e-05}
{"train_loss": 0.18420815467834473, "global_step": 93595, "epoch": 1051, "lr": 7.35918509851919e-05}
{"train_loss": 0.13209544122219086, "global_step": 93596, "epoch": 1051, "lr": 7.359133983808276e-05}
{"train_loss": 0.13928550481796265, "global_step": 93597, "epoch": 1051, "lr": 7.359082868780205e-05}
{"train_loss": 0.1901276707649231, "global_step": 93598, "epoch": 1051, "lr": 7.359031753434984e-05}
{"train_loss": 0.12891800701618195, "global_step": 93599, "epoch": 1051, "lr": 7.35898063777262e-05}
{"train_loss": 0.18614651262760162, "global_step": 93600, "epoch": 1051, "lr": 7.358929521793123e-05}
{"train_loss": 0.09745871275663376, "global_step": 93601, "epoch": 1051, "lr": 7.358878405496492e-05}
{"train_loss": 0.2763766646385193, "global_step": 93602, "epoch": 1051, "lr": 7.358827288882741e-05}
{"train_loss": 0.18809756636619568, "global_step": 93603, "epoch": 1051, "lr": 7.358776171951873e-05}
{"train_loss": 0.17091618478298187, "global_step": 93604, "epoch": 1051, "lr": 7.358725054703897e-05}
{"train_loss": 0.1438027173280716, "global_step": 93605, "epoch": 1051, "lr": 7.358673937138821e-05}
{"train_loss": 0.1665690541267395, "global_step": 93606, "epoch": 1051, "lr": 7.358622819256647e-05}
{"train_loss": 0.13969328999519348, "global_step": 93607, "epoch": 1051, "lr": 7.358571701057386e-05}
{"train_loss": 0.15792329609394073, "global_step": 93608, "epoch": 1051, "lr": 7.358520582541045e-05}
{"train_loss": 0.21574494242668152, "global_step": 93609, "epoch": 1051, "lr": 7.358469463707629e-05}
{"train_loss": 0.15250125527381897, "global_step": 93610, "epoch": 1051, "lr": 7.358418344557144e-05}
{"train_loss": 0.1878257393836975, "global_step": 93611, "epoch": 1051, "lr": 7.3583672250896e-05}
{"train_loss": 0.14220324158668518, "global_step": 93612, "epoch": 1051, "lr": 7.358316105305002e-05}
{"train_loss": 0.1481817364692688, "global_step": 93613, "epoch": 1051, "lr": 7.358264985203357e-05}
{"train_loss": 0.15907461941242218, "global_step": 93614, "epoch": 1051, "lr": 7.358213864784672e-05}
{"train_loss": 0.16678951680660248, "global_step": 93615, "epoch": 1051, "lr": 7.358162744048954e-05}
{"train_loss": 0.23548203706741333, "global_step": 93616, "epoch": 1051, "lr": 7.35811162299621e-05}
{"train_loss": 0.2153259664773941, "global_step": 93617, "epoch": 1051, "lr": 7.358060501626447e-05}
{"train_loss": 0.18398989737033844, "global_step": 93618, "epoch": 1051, "lr": 7.35800937993967e-05}
{"train_loss": 0.11409799009561539, "global_step": 93619, "epoch": 1051, "lr": 7.357958257935888e-05}
{"train_loss": 0.1673753559589386, "global_step": 93620, "epoch": 1051, "lr": 7.357907135615107e-05}
{"train_loss": 0.19202637672424316, "global_step": 93621, "epoch": 1051, "lr": 7.357856012977335e-05}
{"train_loss": 0.09543675184249878, "global_step": 93622, "epoch": 1051, "lr": 7.357804890022578e-05}
{"train_loss": 0.2598922550678253, "global_step": 93623, "epoch": 1051, "lr": 7.357753766750843e-05}
{"train_loss": 0.22882017493247986, "global_step": 93624, "epoch": 1051, "lr": 7.357702643162137e-05}
{"train_loss": 0.15568386018276215, "global_step": 93625, "epoch": 1051, "lr": 7.357651519256465e-05}
{"train_loss": 0.20120616257190704, "global_step": 93626, "epoch": 1051, "lr": 7.357600395033836e-05}
{"train_loss": 0.17749978760989865, "global_step": 93627, "epoch": 1051, "lr": 7.357549270494257e-05, "val_loss": 4.475411415100098}
{"train_loss": 0.142462357878685, "global_step": 93628, "epoch": 1052, "lr": 7.357498145637733e-05}
{"train_loss": 0.13865870237350464, "global_step": 93629, "epoch": 1052, "lr": 7.357447020464274e-05}
{"train_loss": 0.24583855271339417, "global_step": 93630, "epoch": 1052, "lr": 7.357395894973884e-05}
{"train_loss": 0.14425872266292572, "global_step": 93631, "epoch": 1052, "lr": 7.35734476916657e-05}
{"train_loss": 0.1887669712305069, "global_step": 93632, "epoch": 1052, "lr": 7.357293643042342e-05}
{"train_loss": 0.14430604875087738, "global_step": 93633, "epoch": 1052, "lr": 7.357242516601202e-05}
{"train_loss": 0.13455364108085632, "global_step": 93634, "epoch": 1052, "lr": 7.357191389843161e-05}
{"train_loss": 0.20095950365066528, "global_step": 93635, "epoch": 1052, "lr": 7.357140262768226e-05}
{"train_loss": 0.23690414428710938, "global_step": 93636, "epoch": 1052, "lr": 7.3570891353764e-05}
{"train_loss": 0.13389652967453003, "global_step": 93637, "epoch": 1052, "lr": 7.357038007667692e-05}
{"train_loss": 0.1426820456981659, "global_step": 93638, "epoch": 1052, "lr": 7.356986879642109e-05}
{"train_loss": 0.18576066195964813, "global_step": 93639, "epoch": 1052, "lr": 7.356935751299658e-05}
{"train_loss": 0.16506540775299072, "global_step": 93640, "epoch": 1052, "lr": 7.356884622640346e-05}
{"train_loss": 0.19482262432575226, "global_step": 93641, "epoch": 1052, "lr": 7.356833493664179e-05}
{"train_loss": 0.20073673129081726, "global_step": 93642, "epoch": 1052, "lr": 7.356782364371165e-05}
{"train_loss": 0.1516779512166977, "global_step": 93643, "epoch": 1052, "lr": 7.35673123476131e-05}
{"train_loss": 0.20159879326820374, "global_step": 93644, "epoch": 1052, "lr": 7.356680104834622e-05}
{"train_loss": 0.19382432103157043, "global_step": 93645, "epoch": 1052, "lr": 7.356628974591107e-05}
{"train_loss": 0.2120867520570755, "global_step": 93646, "epoch": 1052, "lr": 7.356577844030772e-05}
{"train_loss": 0.17951355874538422, "global_step": 93647, "epoch": 1052, "lr": 7.356526713153622e-05}
{"train_loss": 0.14808498322963715, "global_step": 93648, "epoch": 1052, "lr": 7.356475581959668e-05}
{"train_loss": 0.15498165786266327, "global_step": 93649, "epoch": 1052, "lr": 7.356424450448913e-05}
{"train_loss": 0.1326156109571457, "global_step": 93650, "epoch": 1052, "lr": 7.356373318621366e-05}
{"train_loss": 0.15955618023872375, "global_step": 93651, "epoch": 1052, "lr": 7.356322186477034e-05}
{"train_loss": 0.19852274656295776, "global_step": 93652, "epoch": 1052, "lr": 7.356271054015922e-05}
{"train_loss": 0.11695364862680435, "global_step": 93653, "epoch": 1052, "lr": 7.35621992123804e-05}
{"train_loss": 0.15344569087028503, "global_step": 93654, "epoch": 1052, "lr": 7.356168788143391e-05}
{"train_loss": 0.17265968024730682, "global_step": 93655, "epoch": 1052, "lr": 7.356117654731985e-05}
{"train_loss": 0.16687072813510895, "global_step": 93656, "epoch": 1052, "lr": 7.356066521003826e-05}
{"train_loss": 0.21591143310070038, "global_step": 93657, "epoch": 1052, "lr": 7.356015386958924e-05}
{"train_loss": 0.17371593415737152, "global_step": 93658, "epoch": 1052, "lr": 7.355964252597285e-05}
{"train_loss": 0.17457589507102966, "global_step": 93659, "epoch": 1052, "lr": 7.355913117918914e-05}
{"train_loss": 0.16559511423110962, "global_step": 93660, "epoch": 1052, "lr": 7.355861982923819e-05}
{"train_loss": 0.15918198227882385, "global_step": 93661, "epoch": 1052, "lr": 7.355810847612008e-05}
{"train_loss": 0.17689447104930878, "global_step": 93662, "epoch": 1052, "lr": 7.355759711983487e-05}
{"train_loss": 0.14389106631278992, "global_step": 93663, "epoch": 1052, "lr": 7.355708576038263e-05}
{"train_loss": 0.22998788952827454, "global_step": 93664, "epoch": 1052, "lr": 7.355657439776343e-05}
{"train_loss": 0.1320691704750061, "global_step": 93665, "epoch": 1052, "lr": 7.355606303197731e-05}
{"train_loss": 0.13962174952030182, "global_step": 93666, "epoch": 1052, "lr": 7.35555516630244e-05}
{"train_loss": 0.1468139886856079, "global_step": 93667, "epoch": 1052, "lr": 7.355504029090471e-05}
{"train_loss": 0.15653657913208008, "global_step": 93668, "epoch": 1052, "lr": 7.355452891561834e-05}
{"train_loss": 0.13692310452461243, "global_step": 93669, "epoch": 1052, "lr": 7.355401753716534e-05}
{"train_loss": 0.1669611632823944, "global_step": 93670, "epoch": 1052, "lr": 7.35535061555458e-05}
{"train_loss": 0.1610720008611679, "global_step": 93671, "epoch": 1052, "lr": 7.355299477075977e-05}
{"train_loss": 0.1768943816423416, "global_step": 93672, "epoch": 1052, "lr": 7.355248338280734e-05}
{"train_loss": 0.2341611683368683, "global_step": 93673, "epoch": 1052, "lr": 7.355197199168857e-05}
{"train_loss": 0.22771349549293518, "global_step": 93674, "epoch": 1052, "lr": 7.35514605974035e-05}
{"train_loss": 0.2611958682537079, "global_step": 93675, "epoch": 1052, "lr": 7.355094919995224e-05}
{"train_loss": 0.15912118554115295, "global_step": 93676, "epoch": 1052, "lr": 7.355043779933484e-05}
{"train_loss": 0.2160950005054474, "global_step": 93677, "epoch": 1052, "lr": 7.354992639555136e-05}
{"train_loss": 0.15115147829055786, "global_step": 93678, "epoch": 1052, "lr": 7.35494149886019e-05}
{"train_loss": 0.1745748072862625, "global_step": 93679, "epoch": 1052, "lr": 7.35489035784865e-05}
{"train_loss": 0.1208949089050293, "global_step": 93680, "epoch": 1052, "lr": 7.354839216520521e-05}
{"train_loss": 0.16169454157352448, "global_step": 93681, "epoch": 1052, "lr": 7.354788074875815e-05}
{"train_loss": 0.27021580934524536, "global_step": 93682, "epoch": 1052, "lr": 7.354736932914537e-05}
{"train_loss": 0.1735558658838272, "global_step": 93683, "epoch": 1052, "lr": 7.354685790636692e-05}
{"train_loss": 0.231526181101799, "global_step": 93684, "epoch": 1052, "lr": 7.35463464804229e-05}
{"train_loss": 0.1678324192762375, "global_step": 93685, "epoch": 1052, "lr": 7.354583505131335e-05}
{"train_loss": 0.18765652179718018, "global_step": 93686, "epoch": 1052, "lr": 7.354532361903835e-05}
{"train_loss": 0.1432047337293625, "global_step": 93687, "epoch": 1052, "lr": 7.354481218359796e-05}
{"train_loss": 0.250893235206604, "global_step": 93688, "epoch": 1052, "lr": 7.354430074499226e-05}
{"train_loss": 0.2363036423921585, "global_step": 93689, "epoch": 1052, "lr": 7.354378930322132e-05}
{"train_loss": 0.10327833145856857, "global_step": 93690, "epoch": 1052, "lr": 7.35432778582852e-05}
{"train_loss": 0.2322167009115219, "global_step": 93691, "epoch": 1052, "lr": 7.3542766410184e-05}
{"train_loss": 0.14555451273918152, "global_step": 93692, "epoch": 1052, "lr": 7.354225495891773e-05}
{"train_loss": 0.16302064061164856, "global_step": 93693, "epoch": 1052, "lr": 7.35417435044865e-05}
{"train_loss": 0.14788831770420074, "global_step": 93694, "epoch": 1052, "lr": 7.354123204689036e-05}
{"train_loss": 0.13475939631462097, "global_step": 93695, "epoch": 1052, "lr": 7.354072058612941e-05}
{"train_loss": 0.10892973095178604, "global_step": 93696, "epoch": 1052, "lr": 7.354020912220369e-05}
{"train_loss": 0.2174420952796936, "global_step": 93697, "epoch": 1052, "lr": 7.353969765511326e-05}
{"train_loss": 0.16130156815052032, "global_step": 93698, "epoch": 1052, "lr": 7.353918618485822e-05}
{"train_loss": 0.16858695447444916, "global_step": 93699, "epoch": 1052, "lr": 7.353867471143862e-05}
{"train_loss": 0.19777445495128632, "global_step": 93700, "epoch": 1052, "lr": 7.353816323485453e-05}
{"train_loss": 0.15642693638801575, "global_step": 93701, "epoch": 1052, "lr": 7.353765175510602e-05}
{"train_loss": 0.17047032713890076, "global_step": 93702, "epoch": 1052, "lr": 7.353714027219316e-05}
{"train_loss": 0.12242957204580307, "global_step": 93703, "epoch": 1052, "lr": 7.353662878611602e-05}
{"train_loss": 0.1595207005739212, "global_step": 93704, "epoch": 1052, "lr": 7.353611729687467e-05}
{"train_loss": 0.17324407398700714, "global_step": 93705, "epoch": 1052, "lr": 7.353560580446916e-05}
{"train_loss": 0.20009459555149078, "global_step": 93706, "epoch": 1052, "lr": 7.35350943088996e-05}
{"train_loss": 0.295145183801651, "global_step": 93707, "epoch": 1052, "lr": 7.353458281016602e-05}
{"train_loss": 0.1407671421766281, "global_step": 93708, "epoch": 1052, "lr": 7.35340713082685e-05}
{"train_loss": 0.23040924966335297, "global_step": 93709, "epoch": 1052, "lr": 7.353355980320711e-05}
{"train_loss": 0.114698626101017, "global_step": 93710, "epoch": 1052, "lr": 7.353304829498192e-05}
{"train_loss": 0.1535026729106903, "global_step": 93711, "epoch": 1052, "lr": 7.353253678359301e-05}
{"train_loss": 0.1608273684978485, "global_step": 93712, "epoch": 1052, "lr": 7.353202526904043e-05}
{"train_loss": 0.2161606401205063, "global_step": 93713, "epoch": 1052, "lr": 7.353151375132425e-05}
{"train_loss": 0.2626826763153076, "global_step": 93714, "epoch": 1052, "lr": 7.353100223044456e-05}
{"train_loss": 0.15542912483215332, "global_step": 93715, "epoch": 1052, "lr": 7.35304907064014e-05}
{"train_loss": 0.17518194262566192, "global_step": 93716, "epoch": 1052, "lr": 7.352997917919485e-05, "val_loss": 4.4149370193481445}
{"train_loss": 0.134812593460083, "global_step": 93717, "epoch": 1053, "lr": 7.352946764882499e-05}
{"train_loss": 0.13300223648548126, "global_step": 93718, "epoch": 1053, "lr": 7.352895611529188e-05}
{"train_loss": 0.14265316724777222, "global_step": 93719, "epoch": 1053, "lr": 7.352844457859558e-05}
{"train_loss": 0.09719724953174591, "global_step": 93720, "epoch": 1053, "lr": 7.352793303873617e-05}
{"train_loss": 0.24138611555099487, "global_step": 93721, "epoch": 1053, "lr": 7.352742149571373e-05}
{"train_loss": 0.1858506053686142, "global_step": 93722, "epoch": 1053, "lr": 7.35269099495283e-05}
{"train_loss": 0.19350476562976837, "global_step": 93723, "epoch": 1053, "lr": 7.352639840017997e-05}
{"train_loss": 0.17633914947509766, "global_step": 93724, "epoch": 1053, "lr": 7.35258868476688e-05}
{"train_loss": 0.27292191982269287, "global_step": 93725, "epoch": 1053, "lr": 7.352537529199485e-05}
{"train_loss": 0.18940554559230804, "global_step": 93726, "epoch": 1053, "lr": 7.352486373315822e-05}
{"train_loss": 0.16799604892730713, "global_step": 93727, "epoch": 1053, "lr": 7.352435217115894e-05}
{"train_loss": 0.20085015892982483, "global_step": 93728, "epoch": 1053, "lr": 7.35238406059971e-05}
{"train_loss": 0.13560865819454193, "global_step": 93729, "epoch": 1053, "lr": 7.352332903767278e-05}
{"train_loss": 0.21056444942951202, "global_step": 93730, "epoch": 1053, "lr": 7.352281746618604e-05}
{"train_loss": 0.17841440439224243, "global_step": 93731, "epoch": 1053, "lr": 7.352230589153693e-05}
{"train_loss": 0.1199876219034195, "global_step": 93732, "epoch": 1053, "lr": 7.352179431372554e-05}
{"train_loss": 0.13791921734809875, "global_step": 93733, "epoch": 1053, "lr": 7.352128273275193e-05}
{"train_loss": 0.1713305115699768, "global_step": 93734, "epoch": 1053, "lr": 7.352077114861617e-05}
{"train_loss": 0.14015446603298187, "global_step": 93735, "epoch": 1053, "lr": 7.352025956131832e-05}
{"train_loss": 0.17190857231616974, "global_step": 93736, "epoch": 1053, "lr": 7.351974797085847e-05}
{"train_loss": 0.22737035155296326, "global_step": 93737, "epoch": 1053, "lr": 7.351923637723667e-05}
{"train_loss": 0.23315228521823883, "global_step": 93738, "epoch": 1053, "lr": 7.351872478045299e-05}
{"train_loss": 0.15494076907634735, "global_step": 93739, "epoch": 1053, "lr": 7.351821318050752e-05}
{"train_loss": 0.21326041221618652, "global_step": 93740, "epoch": 1053, "lr": 7.35177015774003e-05}
{"train_loss": 0.22263124585151672, "global_step": 93741, "epoch": 1053, "lr": 7.351718997113142e-05}
{"train_loss": 0.11918709427118301, "global_step": 93742, "epoch": 1053, "lr": 7.351667836170092e-05}
{"train_loss": 0.23153141140937805, "global_step": 93743, "epoch": 1053, "lr": 7.351616674910891e-05}
{"train_loss": 0.16388720273971558, "global_step": 93744, "epoch": 1053, "lr": 7.351565513335543e-05}
{"train_loss": 0.15742360055446625, "global_step": 93745, "epoch": 1053, "lr": 7.351514351444057e-05}
{"train_loss": 0.19634687900543213, "global_step": 93746, "epoch": 1053, "lr": 7.351463189236436e-05}
{"train_loss": 0.18241196870803833, "global_step": 93747, "epoch": 1053, "lr": 7.35141202671269e-05}
{"train_loss": 0.16994795203208923, "global_step": 93748, "epoch": 1053, "lr": 7.351360863872827e-05}
{"train_loss": 0.20639072358608246, "global_step": 93749, "epoch": 1053, "lr": 7.351309700716853e-05}
{"train_loss": 0.13542823493480682, "global_step": 93750, "epoch": 1053, "lr": 7.351258537244772e-05}
{"train_loss": 0.2562204897403717, "global_step": 93751, "epoch": 1053, "lr": 7.351207373456594e-05}
{"train_loss": 0.1931522637605667, "global_step": 93752, "epoch": 1053, "lr": 7.351156209352324e-05}
{"train_loss": 0.15254627168178558, "global_step": 93753, "epoch": 1053, "lr": 7.35110504493197e-05}
{"train_loss": 0.11314032226800919, "global_step": 93754, "epoch": 1053, "lr": 7.351053880195538e-05}
{"train_loss": 0.12164768576622009, "global_step": 93755, "epoch": 1053, "lr": 7.351002715143036e-05}
{"train_loss": 0.19281049072742462, "global_step": 93756, "epoch": 1053, "lr": 7.35095154977447e-05}
{"train_loss": 0.12248773872852325, "global_step": 93757, "epoch": 1053, "lr": 7.350900384089848e-05}
{"train_loss": 0.1323554515838623, "global_step": 93758, "epoch": 1053, "lr": 7.350849218089176e-05}
{"train_loss": 0.1333225518465042, "global_step": 93759, "epoch": 1053, "lr": 7.35079805177246e-05}
{"train_loss": 0.19083310663700104, "global_step": 93760, "epoch": 1053, "lr": 7.35074688513971e-05}
{"train_loss": 0.17505046725273132, "global_step": 93761, "epoch": 1053, "lr": 7.350695718190929e-05}
{"train_loss": 0.1864742487668991, "global_step": 93762, "epoch": 1053, "lr": 7.350644550926126e-05}
{"train_loss": 0.1571628749370575, "global_step": 93763, "epoch": 1053, "lr": 7.350593383345306e-05}
{"train_loss": 0.18953143060207367, "global_step": 93764, "epoch": 1053, "lr": 7.350542215448479e-05}
{"train_loss": 0.1561412513256073, "global_step": 93765, "epoch": 1053, "lr": 7.350491047235649e-05}
{"train_loss": 0.1372799277305603, "global_step": 93766, "epoch": 1053, "lr": 7.350439878706825e-05}
{"train_loss": 0.15668553113937378, "global_step": 93767, "epoch": 1053, "lr": 7.350388709862014e-05}
{"train_loss": 0.2148929238319397, "global_step": 93768, "epoch": 1053, "lr": 7.35033754070122e-05}
{"train_loss": 0.261055052280426, "global_step": 93769, "epoch": 1053, "lr": 7.350286371224454e-05}
{"train_loss": 0.2118154764175415, "global_step": 93770, "epoch": 1053, "lr": 7.350235201431719e-05}
{"train_loss": 0.1578206717967987, "global_step": 93771, "epoch": 1053, "lr": 7.350184031323024e-05}
{"train_loss": 0.1927582323551178, "global_step": 93772, "epoch": 1053, "lr": 7.350132860898376e-05}
{"train_loss": 0.16412244737148285, "global_step": 93773, "epoch": 1053, "lr": 7.35008169015778e-05}
{"train_loss": 0.11684400588274002, "global_step": 93774, "epoch": 1053, "lr": 7.350030519101245e-05}
{"train_loss": 0.1945417821407318, "global_step": 93775, "epoch": 1053, "lr": 7.349979347728778e-05}
{"train_loss": 0.16008271276950836, "global_step": 93776, "epoch": 1053, "lr": 7.349928176040382e-05}
{"train_loss": 0.1636648178100586, "global_step": 93777, "epoch": 1053, "lr": 7.349877004036069e-05}
{"train_loss": 0.19776183366775513, "global_step": 93778, "epoch": 1053, "lr": 7.349825831715843e-05}
{"train_loss": 0.1994045376777649, "global_step": 93779, "epoch": 1053, "lr": 7.349774659079712e-05}
{"train_loss": 0.13254013657569885, "global_step": 93780, "epoch": 1053, "lr": 7.349723486127682e-05}
{"train_loss": 0.14824828505516052, "global_step": 93781, "epoch": 1053, "lr": 7.349672312859761e-05}
{"train_loss": 0.15781119465827942, "global_step": 93782, "epoch": 1053, "lr": 7.349621139275955e-05}
{"train_loss": 0.1392267942428589, "global_step": 93783, "epoch": 1053, "lr": 7.34956996537627e-05}
{"train_loss": 0.15356579422950745, "global_step": 93784, "epoch": 1053, "lr": 7.349518791160715e-05}
{"train_loss": 0.13578550517559052, "global_step": 93785, "epoch": 1053, "lr": 7.349467616629296e-05}
{"train_loss": 0.16154645383358002, "global_step": 93786, "epoch": 1053, "lr": 7.349416441782018e-05}
{"train_loss": 0.1394483745098114, "global_step": 93787, "epoch": 1053, "lr": 7.349365266618891e-05}
{"train_loss": 0.16567528247833252, "global_step": 93788, "epoch": 1053, "lr": 7.34931409113992e-05}
{"train_loss": 0.13419024646282196, "global_step": 93789, "epoch": 1053, "lr": 7.349262915345115e-05}
{"train_loss": 0.1813722550868988, "global_step": 93790, "epoch": 1053, "lr": 7.349211739234477e-05}
{"train_loss": 0.15540578961372375, "global_step": 93791, "epoch": 1053, "lr": 7.349160562808018e-05}
{"train_loss": 0.10437460988759995, "global_step": 93792, "epoch": 1053, "lr": 7.349109386065741e-05}
{"train_loss": 0.24512067437171936, "global_step": 93793, "epoch": 1053, "lr": 7.349058209007655e-05}
{"train_loss": 0.21751384437084198, "global_step": 93794, "epoch": 1053, "lr": 7.349007031633768e-05}
{"train_loss": 0.15178416669368744, "global_step": 93795, "epoch": 1053, "lr": 7.348955853944087e-05}
{"train_loss": 0.103218212723732, "global_step": 93796, "epoch": 1053, "lr": 7.348904675938615e-05}
{"train_loss": 0.19479665160179138, "global_step": 93797, "epoch": 1053, "lr": 7.348853497617362e-05}
{"train_loss": 0.14341148734092712, "global_step": 93798, "epoch": 1053, "lr": 7.348802318980335e-05}
{"train_loss": 0.24573595821857452, "global_step": 93799, "epoch": 1053, "lr": 7.34875114002754e-05}
{"train_loss": 0.1831892728805542, "global_step": 93800, "epoch": 1053, "lr": 7.348699960758983e-05}
{"train_loss": 0.11432362347841263, "global_step": 93801, "epoch": 1053, "lr": 7.348648781174672e-05}
{"train_loss": 0.14509135484695435, "global_step": 93802, "epoch": 1053, "lr": 7.348597601274613e-05}
{"train_loss": 0.24889224767684937, "global_step": 93803, "epoch": 1053, "lr": 7.348546421058815e-05}
{"train_loss": 0.24256490170955658, "global_step": 93804, "epoch": 1053, "lr": 7.348495240527284e-05}
{"train_loss": 0.17395303532313766, "global_step": 93805, "epoch": 1053, "lr": 7.348444059680026e-05, "val_loss": 4.660536289215088}
{"train_loss": 0.15527717769145966, "global_step": 93806, "epoch": 1054, "lr": 7.348392878517049e-05}
{"train_loss": 0.1504417359828949, "global_step": 93807, "epoch": 1054, "lr": 7.348341697038358e-05}
{"train_loss": 0.17603279650211334, "global_step": 93808, "epoch": 1054, "lr": 7.348290515243963e-05}
{"train_loss": 0.14432913064956665, "global_step": 93809, "epoch": 1054, "lr": 7.348239333133868e-05}
{"train_loss": 0.23056159913539886, "global_step": 93810, "epoch": 1054, "lr": 7.34818815070808e-05}
{"train_loss": 0.13304127752780914, "global_step": 93811, "epoch": 1054, "lr": 7.348136967966608e-05}
{"train_loss": 0.1617869734764099, "global_step": 93812, "epoch": 1054, "lr": 7.348085784909457e-05}
{"train_loss": 0.2147039920091629, "global_step": 93813, "epoch": 1054, "lr": 7.348034601536632e-05}
{"train_loss": 0.1874493658542633, "global_step": 93814, "epoch": 1054, "lr": 7.347983417848145e-05}
{"train_loss": 0.10227406024932861, "global_step": 93815, "epoch": 1054, "lr": 7.347932233844002e-05}
{"train_loss": 0.1935848444700241, "global_step": 93816, "epoch": 1054, "lr": 7.347881049524205e-05}
{"train_loss": 0.19768919050693512, "global_step": 93817, "epoch": 1054, "lr": 7.347829864888765e-05}
{"train_loss": 0.15947414934635162, "global_step": 93818, "epoch": 1054, "lr": 7.347778679937689e-05}
{"train_loss": 0.15340213477611542, "global_step": 93819, "epoch": 1054, "lr": 7.34772749467098e-05}
{"train_loss": 0.2616249918937683, "global_step": 93820, "epoch": 1054, "lr": 7.347676309088649e-05}
{"train_loss": 0.26293179392814636, "global_step": 93821, "epoch": 1054, "lr": 7.347625123190702e-05}
{"train_loss": 0.21484149992465973, "global_step": 93822, "epoch": 1054, "lr": 7.347573936977146e-05}
{"train_loss": 0.2195550501346588, "global_step": 93823, "epoch": 1054, "lr": 7.347522750447987e-05}
{"train_loss": 0.13351128995418549, "global_step": 93824, "epoch": 1054, "lr": 7.347471563603231e-05}
{"train_loss": 0.19074609875679016, "global_step": 93825, "epoch": 1054, "lr": 7.347420376442887e-05}
{"train_loss": 0.281339168548584, "global_step": 93826, "epoch": 1054, "lr": 7.34736918896696e-05}
{"train_loss": 0.2082156091928482, "global_step": 93827, "epoch": 1054, "lr": 7.34731800117546e-05}
{"train_loss": 0.19070957601070404, "global_step": 93828, "epoch": 1054, "lr": 7.34726681306839e-05}
{"train_loss": 0.30232539772987366, "global_step": 93829, "epoch": 1054, "lr": 7.347215624645759e-05}
{"train_loss": 0.2807459235191345, "global_step": 93830, "epoch": 1054, "lr": 7.347164435907573e-05}
{"train_loss": 0.22339758276939392, "global_step": 93831, "epoch": 1054, "lr": 7.34711324685384e-05}
{"train_loss": 0.14805865287780762, "global_step": 93832, "epoch": 1054, "lr": 7.347062057484566e-05}
{"train_loss": 0.14626356959342957, "global_step": 93833, "epoch": 1054, "lr": 7.347010867799759e-05}
{"train_loss": 0.13211293518543243, "global_step": 93834, "epoch": 1054, "lr": 7.346959677799423e-05}
{"train_loss": 0.24758729338645935, "global_step": 93835, "epoch": 1054, "lr": 7.346908487483569e-05}
{"train_loss": 0.22811736166477203, "global_step": 93836, "epoch": 1054, "lr": 7.3468572968522e-05}
{"train_loss": 0.11376632004976273, "global_step": 93837, "epoch": 1054, "lr": 7.346806105905324e-05}
{"train_loss": 0.2027074247598648, "global_step": 93838, "epoch": 1054, "lr": 7.346754914642951e-05}
{"train_loss": 0.22787253558635712, "global_step": 93839, "epoch": 1054, "lr": 7.346703723065083e-05}
{"train_loss": 0.19932587444782257, "global_step": 93840, "epoch": 1054, "lr": 7.346652531171731e-05}
{"train_loss": 0.17310529947280884, "global_step": 93841, "epoch": 1054, "lr": 7.3466013389629e-05}
{"train_loss": 0.167249858379364, "global_step": 93842, "epoch": 1054, "lr": 7.346550146438597e-05}
{"train_loss": 0.18440616130828857, "global_step": 93843, "epoch": 1054, "lr": 7.346498953598828e-05}
{"train_loss": 0.14104439318180084, "global_step": 93844, "epoch": 1054, "lr": 7.346447760443602e-05}
{"train_loss": 0.22932814061641693, "global_step": 93845, "epoch": 1054, "lr": 7.346396566972924e-05}
{"train_loss": 0.13930289447307587, "global_step": 93846, "epoch": 1054, "lr": 7.346345373186802e-05}
{"train_loss": 0.22695571184158325, "global_step": 93847, "epoch": 1054, "lr": 7.346294179085243e-05}
{"train_loss": 0.1751927137374878, "global_step": 93848, "epoch": 1054, "lr": 7.346242984668252e-05}
{"train_loss": 0.21027831733226776, "global_step": 93849, "epoch": 1054, "lr": 7.346191789935838e-05}
{"train_loss": 0.2145509272813797, "global_step": 93850, "epoch": 1054, "lr": 7.346140594888007e-05}
{"train_loss": 0.19268077611923218, "global_step": 93851, "epoch": 1054, "lr": 7.346089399524765e-05}
{"train_loss": 0.16419577598571777, "global_step": 93852, "epoch": 1054, "lr": 7.346038203846121e-05}
{"train_loss": 0.17169474065303802, "global_step": 93853, "epoch": 1054, "lr": 7.345987007852082e-05}
{"train_loss": 0.23752130568027496, "global_step": 93854, "epoch": 1054, "lr": 7.345935811542652e-05}
{"train_loss": 0.17859293520450592, "global_step": 93855, "epoch": 1054, "lr": 7.345884614917838e-05}
{"train_loss": 0.23053821921348572, "global_step": 93856, "epoch": 1054, "lr": 7.34583341797765e-05}
{"train_loss": 0.2376329004764557, "global_step": 93857, "epoch": 1054, "lr": 7.345782220722094e-05}
{"train_loss": 0.2003728300333023, "global_step": 93858, "epoch": 1054, "lr": 7.345731023151176e-05}
{"train_loss": 0.10747034847736359, "global_step": 93859, "epoch": 1054, "lr": 7.345679825264902e-05}
{"train_loss": 0.1941947638988495, "global_step": 93860, "epoch": 1054, "lr": 7.345628627063282e-05}
{"train_loss": 0.1721464991569519, "global_step": 93861, "epoch": 1054, "lr": 7.345577428546319e-05}
{"train_loss": 0.1557227075099945, "global_step": 93862, "epoch": 1054, "lr": 7.345526229714022e-05}
{"train_loss": 0.1443161964416504, "global_step": 93863, "epoch": 1054, "lr": 7.345475030566397e-05}
{"train_loss": 0.18287739157676697, "global_step": 93864, "epoch": 1054, "lr": 7.345423831103453e-05}
{"train_loss": 0.1855657696723938, "global_step": 93865, "epoch": 1054, "lr": 7.345372631325195e-05}
{"train_loss": 0.16769903898239136, "global_step": 93866, "epoch": 1054, "lr": 7.34532143123163e-05}
{"train_loss": 0.230708047747612, "global_step": 93867, "epoch": 1054, "lr": 7.345270230822765e-05}
{"train_loss": 0.21017470955848694, "global_step": 93868, "epoch": 1054, "lr": 7.345219030098607e-05}
{"train_loss": 0.1631404012441635, "global_step": 93869, "epoch": 1054, "lr": 7.345167829059162e-05}
{"train_loss": 0.11543600261211395, "global_step": 93870, "epoch": 1054, "lr": 7.345116627704438e-05}
{"train_loss": 0.20414963364601135, "global_step": 93871, "epoch": 1054, "lr": 7.345065426034442e-05}
{"train_loss": 0.23807163536548615, "global_step": 93872, "epoch": 1054, "lr": 7.345014224049182e-05}
{"train_loss": 0.1549016833305359, "global_step": 93873, "epoch": 1054, "lr": 7.34496302174866e-05}
{"train_loss": 0.08331939578056335, "global_step": 93874, "epoch": 1054, "lr": 7.34491181913289e-05}
{"train_loss": 0.13401462137699127, "global_step": 93875, "epoch": 1054, "lr": 7.344860616201872e-05}
{"train_loss": 0.15046434104442596, "global_step": 93876, "epoch": 1054, "lr": 7.344809412955618e-05}
{"train_loss": 0.2254556566476822, "global_step": 93877, "epoch": 1054, "lr": 7.344758209394133e-05}
{"train_loss": 0.14129218459129333, "global_step": 93878, "epoch": 1054, "lr": 7.344707005517424e-05}
{"train_loss": 0.21198770403862, "global_step": 93879, "epoch": 1054, "lr": 7.344655801325496e-05}
{"train_loss": 0.1496536135673523, "global_step": 93880, "epoch": 1054, "lr": 7.344604596818359e-05}
{"train_loss": 0.1464177817106247, "global_step": 93881, "epoch": 1054, "lr": 7.344553391996019e-05}
{"train_loss": 0.12916399538516998, "global_step": 93882, "epoch": 1054, "lr": 7.344502186858482e-05}
{"train_loss": 0.21087506413459778, "global_step": 93883, "epoch": 1054, "lr": 7.344450981405754e-05}
{"train_loss": 0.24494722485542297, "global_step": 93884, "epoch": 1054, "lr": 7.344399775637845e-05}
{"train_loss": 0.23261354863643646, "global_step": 93885, "epoch": 1054, "lr": 7.34434856955476e-05}
{"train_loss": 0.09865092486143112, "global_step": 93886, "epoch": 1054, "lr": 7.344297363156505e-05}
{"train_loss": 0.11712070554494858, "global_step": 93887, "epoch": 1054, "lr": 7.344246156443087e-05}
{"train_loss": 0.15298371016979218, "global_step": 93888, "epoch": 1054, "lr": 7.344194949414515e-05}
{"train_loss": 0.1255885511636734, "global_step": 93889, "epoch": 1054, "lr": 7.344143742070794e-05}
{"train_loss": 0.17664949595928192, "global_step": 93890, "epoch": 1054, "lr": 7.344092534411931e-05}
{"train_loss": 0.19539852440357208, "global_step": 93891, "epoch": 1054, "lr": 7.344041326437934e-05}
{"train_loss": 0.15613067150115967, "global_step": 93892, "epoch": 1054, "lr": 7.343990118148808e-05}
{"train_loss": 0.10529351979494095, "global_step": 93893, "epoch": 1054, "lr": 7.343938909544563e-05}
{"train_loss": 0.18194100947192546, "global_step": 93894, "epoch": 1054, "lr": 7.343887700625203e-05, "val_loss": 4.537318229675293}
{"train_loss": 0.14774034917354584, "global_step": 93895, "epoch": 1055, "lr": 7.343836491390736e-05}
{"train_loss": 0.14575311541557312, "global_step": 93896, "epoch": 1055, "lr": 7.343785281841168e-05}
{"train_loss": 0.19015531241893768, "global_step": 93897, "epoch": 1055, "lr": 7.343734071976507e-05}
{"train_loss": 0.21524836122989655, "global_step": 93898, "epoch": 1055, "lr": 7.34368286179676e-05}
{"train_loss": 0.26362308859825134, "global_step": 93899, "epoch": 1055, "lr": 7.343631651301933e-05}
{"train_loss": 0.14526832103729248, "global_step": 93900, "epoch": 1055, "lr": 7.343580440492034e-05}
{"train_loss": 0.1454610526561737, "global_step": 93901, "epoch": 1055, "lr": 7.343529229367069e-05}
{"train_loss": 0.14729644358158112, "global_step": 93902, "epoch": 1055, "lr": 7.343478017927045e-05}
{"train_loss": 0.2316042184829712, "global_step": 93903, "epoch": 1055, "lr": 7.343426806171969e-05}
{"train_loss": 0.15430733561515808, "global_step": 93904, "epoch": 1055, "lr": 7.343375594101848e-05}
{"train_loss": 0.1777457296848297, "global_step": 93905, "epoch": 1055, "lr": 7.343324381716688e-05}
{"train_loss": 0.15518762171268463, "global_step": 93906, "epoch": 1055, "lr": 7.343273169016497e-05}
{"train_loss": 0.20626665651798248, "global_step": 93907, "epoch": 1055, "lr": 7.343221956001281e-05}
{"train_loss": 0.14366450905799866, "global_step": 93908, "epoch": 1055, "lr": 7.343170742671048e-05}
{"train_loss": 0.16513487696647644, "global_step": 93909, "epoch": 1055, "lr": 7.343119529025804e-05}
{"train_loss": 0.19691024720668793, "global_step": 93910, "epoch": 1055, "lr": 7.343068315065556e-05}
{"train_loss": 0.12439515441656113, "global_step": 93911, "epoch": 1055, "lr": 7.34301710079031e-05}
{"train_loss": 0.20227542519569397, "global_step": 93912, "epoch": 1055, "lr": 7.342965886200076e-05}
{"train_loss": 0.10315079241991043, "global_step": 93913, "epoch": 1055, "lr": 7.342914671294857e-05}
{"train_loss": 0.14158891141414642, "global_step": 93914, "epoch": 1055, "lr": 7.342863456074663e-05}
{"train_loss": 0.1231408640742302, "global_step": 93915, "epoch": 1055, "lr": 7.3428122405395e-05}
{"train_loss": 0.2344035506248474, "global_step": 93916, "epoch": 1055, "lr": 7.342761024689373e-05}
{"train_loss": 0.1094081774353981, "global_step": 93917, "epoch": 1055, "lr": 7.34270980852429e-05}
{"train_loss": 0.15423141419887543, "global_step": 93918, "epoch": 1055, "lr": 7.342658592044259e-05}
{"train_loss": 0.14644773304462433, "global_step": 93919, "epoch": 1055, "lr": 7.342607375249287e-05}
{"train_loss": 0.15414290130138397, "global_step": 93920, "epoch": 1055, "lr": 7.342556158139378e-05}
{"train_loss": 0.2079354077577591, "global_step": 93921, "epoch": 1055, "lr": 7.342504940714543e-05}
{"train_loss": 0.21517124772071838, "global_step": 93922, "epoch": 1055, "lr": 7.342453722974786e-05}
{"train_loss": 0.16168224811553955, "global_step": 93923, "epoch": 1055, "lr": 7.342402504920115e-05}
{"train_loss": 0.1550443470478058, "global_step": 93924, "epoch": 1055, "lr": 7.342351286550536e-05}
{"train_loss": 0.0727134719491005, "global_step": 93925, "epoch": 1055, "lr": 7.342300067866057e-05}
{"train_loss": 0.1647513508796692, "global_step": 93926, "epoch": 1055, "lr": 7.342248848866684e-05}
{"train_loss": 0.112140953540802, "global_step": 93927, "epoch": 1055, "lr": 7.342197629552423e-05}
{"train_loss": 0.14445139467716217, "global_step": 93928, "epoch": 1055, "lr": 7.342146409923283e-05}
{"train_loss": 0.15321528911590576, "global_step": 93929, "epoch": 1055, "lr": 7.34209518997927e-05}
{"train_loss": 0.11259637773036957, "global_step": 93930, "epoch": 1055, "lr": 7.34204396972039e-05}
{"train_loss": 0.16643427312374115, "global_step": 93931, "epoch": 1055, "lr": 7.34199274914665e-05}
{"train_loss": 0.19172681868076324, "global_step": 93932, "epoch": 1055, "lr": 7.34194152825806e-05}
{"train_loss": 0.13163645565509796, "global_step": 93933, "epoch": 1055, "lr": 7.341890307054623e-05}
{"train_loss": 0.22637633979320526, "global_step": 93934, "epoch": 1055, "lr": 7.341839085536349e-05}
{"train_loss": 0.173141747713089, "global_step": 93935, "epoch": 1055, "lr": 7.341787863703244e-05}
{"train_loss": 0.2097620666027069, "global_step": 93936, "epoch": 1055, "lr": 7.341736641555311e-05}
{"train_loss": 0.1565849781036377, "global_step": 93937, "epoch": 1055, "lr": 7.341685419092562e-05}
{"train_loss": 0.16006402671337128, "global_step": 93938, "epoch": 1055, "lr": 7.341634196315002e-05}
{"train_loss": 0.1492440402507782, "global_step": 93939, "epoch": 1055, "lr": 7.341582973222637e-05}
{"train_loss": 0.13845224678516388, "global_step": 93940, "epoch": 1055, "lr": 7.341531749815475e-05}
{"train_loss": 0.1472960263490677, "global_step": 93941, "epoch": 1055, "lr": 7.341480526093523e-05}
{"train_loss": 0.2103973627090454, "global_step": 93942, "epoch": 1055, "lr": 7.341429302056786e-05}
{"train_loss": 0.12945568561553955, "global_step": 93943, "epoch": 1055, "lr": 7.341378077705275e-05}
{"train_loss": 0.19838252663612366, "global_step": 93944, "epoch": 1055, "lr": 7.341326853038992e-05}
{"train_loss": 0.17166532576084137, "global_step": 93945, "epoch": 1055, "lr": 7.341275628057947e-05}
{"train_loss": 0.17324134707450867, "global_step": 93946, "epoch": 1055, "lr": 7.341224402762146e-05}
{"train_loss": 0.14487430453300476, "global_step": 93947, "epoch": 1055, "lr": 7.341173177151596e-05}
{"train_loss": 0.19927752017974854, "global_step": 93948, "epoch": 1055, "lr": 7.341121951226303e-05}
{"train_loss": 0.1829821616411209, "global_step": 93949, "epoch": 1055, "lr": 7.341070724986275e-05}
{"train_loss": 0.19650989770889282, "global_step": 93950, "epoch": 1055, "lr": 7.34101949843152e-05}
{"train_loss": 0.18011383712291718, "global_step": 93951, "epoch": 1055, "lr": 7.340968271562041e-05}
{"train_loss": 0.13462993502616882, "global_step": 93952, "epoch": 1055, "lr": 7.340917044377849e-05}
{"train_loss": 0.15250976383686066, "global_step": 93953, "epoch": 1055, "lr": 7.340865816878949e-05}
{"train_loss": 0.14248499274253845, "global_step": 93954, "epoch": 1055, "lr": 7.340814589065348e-05}
{"train_loss": 0.1487816870212555, "global_step": 93955, "epoch": 1055, "lr": 7.340763360937053e-05}
{"train_loss": 0.17380669713020325, "global_step": 93956, "epoch": 1055, "lr": 7.34071213249407e-05}
{"train_loss": 0.10695052146911621, "global_step": 93957, "epoch": 1055, "lr": 7.340660903736409e-05}
{"train_loss": 0.20591656863689423, "global_step": 93958, "epoch": 1055, "lr": 7.340609674664072e-05}
{"train_loss": 0.16251322627067566, "global_step": 93959, "epoch": 1055, "lr": 7.340558445277072e-05}
{"train_loss": 0.17074963450431824, "global_step": 93960, "epoch": 1055, "lr": 7.34050721557541e-05}
{"train_loss": 0.1904866099357605, "global_step": 93961, "epoch": 1055, "lr": 7.340455985559095e-05}
{"train_loss": 0.12510573863983154, "global_step": 93962, "epoch": 1055, "lr": 7.340404755228136e-05}
{"train_loss": 0.15823350846767426, "global_step": 93963, "epoch": 1055, "lr": 7.340353524582538e-05}
{"train_loss": 0.15972615778446198, "global_step": 93964, "epoch": 1055, "lr": 7.340302293622306e-05}
{"train_loss": 0.21038207411766052, "global_step": 93965, "epoch": 1055, "lr": 7.340251062347449e-05}
{"train_loss": 0.15359842777252197, "global_step": 93966, "epoch": 1055, "lr": 7.340199830757975e-05}
{"train_loss": 0.2095886915922165, "global_step": 93967, "epoch": 1055, "lr": 7.340148598853889e-05}
{"train_loss": 0.29119041562080383, "global_step": 93968, "epoch": 1055, "lr": 7.3400973666352e-05}
{"train_loss": 0.2725411653518677, "global_step": 93969, "epoch": 1055, "lr": 7.340046134101912e-05}
{"train_loss": 0.1709311157464981, "global_step": 93970, "epoch": 1055, "lr": 7.339994901254033e-05}
{"train_loss": 0.16036652028560638, "global_step": 93971, "epoch": 1055, "lr": 7.339943668091572e-05}
{"train_loss": 0.1409747451543808, "global_step": 93972, "epoch": 1055, "lr": 7.339892434614533e-05}
{"train_loss": 0.10327109694480896, "global_step": 93973, "epoch": 1055, "lr": 7.339841200822923e-05}
{"train_loss": 0.22702565789222717, "global_step": 93974, "epoch": 1055, "lr": 7.339789966716751e-05}
{"train_loss": 0.1958409696817398, "global_step": 93975, "epoch": 1055, "lr": 7.339738732296022e-05}
{"train_loss": 0.15113957226276398, "global_step": 93976, "epoch": 1055, "lr": 7.339687497560744e-05}
{"train_loss": 0.16265784204006195, "global_step": 93977, "epoch": 1055, "lr": 7.339636262510923e-05}
{"train_loss": 0.19412963092327118, "global_step": 93978, "epoch": 1055, "lr": 7.339585027146569e-05}
{"train_loss": 0.14548639953136444, "global_step": 93979, "epoch": 1055, "lr": 7.339533791467683e-05}
{"train_loss": 0.17873792350292206, "global_step": 93980, "epoch": 1055, "lr": 7.339482555474278e-05}
{"train_loss": 0.1498967707157135, "global_step": 93981, "epoch": 1055, "lr": 7.339431319166357e-05}
{"train_loss": 0.2047549933195114, "global_step": 93982, "epoch": 1055, "lr": 7.339380082543927e-05}
{"train_loss": 0.16807958642753323, "global_step": 93983, "epoch": 1055, "lr": 7.339328845606996e-05, "val_loss": 4.537777900695801, "train_action_mse_error": 14.486299514770508}
{"train_loss": 0.2653454840183258, "global_step": 93984, "epoch": 1056, "lr": 7.339277608355573e-05}
{"train_loss": 0.20699839293956757, "global_step": 93985, "epoch": 1056, "lr": 7.33922637078966e-05}
{"train_loss": 0.17204008996486664, "global_step": 93986, "epoch": 1056, "lr": 7.339175132909267e-05}
{"train_loss": 0.1514482945203781, "global_step": 93987, "epoch": 1056, "lr": 7.339123894714402e-05}
{"train_loss": 0.14419244229793549, "global_step": 93988, "epoch": 1056, "lr": 7.339072656205068e-05}
{"train_loss": 0.15217715501785278, "global_step": 93989, "epoch": 1056, "lr": 7.339021417381276e-05}
{"train_loss": 0.1601010411977768, "global_step": 93990, "epoch": 1056, "lr": 7.33897017824303e-05}
{"train_loss": 0.2895658612251282, "global_step": 93991, "epoch": 1056, "lr": 7.338918938790339e-05}
{"train_loss": 0.29654932022094727, "global_step": 93992, "epoch": 1056, "lr": 7.338867699023209e-05}
{"train_loss": 0.19359523057937622, "global_step": 93993, "epoch": 1056, "lr": 7.338816458941645e-05}
{"train_loss": 0.17367461323738098, "global_step": 93994, "epoch": 1056, "lr": 7.338765218545657e-05}
{"train_loss": 0.17640765011310577, "global_step": 93995, "epoch": 1056, "lr": 7.33871397783525e-05}
{"train_loss": 0.2206237018108368, "global_step": 93996, "epoch": 1056, "lr": 7.338662736810432e-05}
{"train_loss": 0.21211601793766022, "global_step": 93997, "epoch": 1056, "lr": 7.338611495471208e-05}
{"train_loss": 0.1760425567626953, "global_step": 93998, "epoch": 1056, "lr": 7.338560253817588e-05}
{"train_loss": 0.21235403418540955, "global_step": 93999, "epoch": 1056, "lr": 7.338509011849576e-05}
{"train_loss": 0.1517244577407837, "global_step": 94000, "epoch": 1056, "lr": 7.338457769567179e-05}
{"train_loss": 0.2340157926082611, "global_step": 94001, "epoch": 1056, "lr": 7.338406526970407e-05}
{"train_loss": 0.296967476606369, "global_step": 94002, "epoch": 1056, "lr": 7.338355284059264e-05}
{"train_loss": 0.2810826599597931, "global_step": 94003, "epoch": 1056, "lr": 7.338304040833756e-05}
{"train_loss": 0.1764524281024933, "global_step": 94004, "epoch": 1056, "lr": 7.338252797293894e-05}
{"train_loss": 0.24276383221149445, "global_step": 94005, "epoch": 1056, "lr": 7.33820155343968e-05}
{"train_loss": 0.1789321005344391, "global_step": 94006, "epoch": 1056, "lr": 7.338150309271126e-05}
{"train_loss": 0.20060966908931732, "global_step": 94007, "epoch": 1056, "lr": 7.338099064788234e-05}
{"train_loss": 0.2509418725967407, "global_step": 94008, "epoch": 1056, "lr": 7.338047819991014e-05}
{"train_loss": 0.250479131937027, "global_step": 94009, "epoch": 1056, "lr": 7.337996574879473e-05}
{"train_loss": 0.245042085647583, "global_step": 94010, "epoch": 1056, "lr": 7.337945329453615e-05}
{"train_loss": 0.17446908354759216, "global_step": 94011, "epoch": 1056, "lr": 7.337894083713449e-05}
{"train_loss": 0.15332160890102386, "global_step": 94012, "epoch": 1056, "lr": 7.337842837658983e-05}
{"train_loss": 0.17125239968299866, "global_step": 94013, "epoch": 1056, "lr": 7.33779159129022e-05}
{"train_loss": 0.23250223696231842, "global_step": 94014, "epoch": 1056, "lr": 7.337740344607173e-05}
{"train_loss": 0.19615784287452698, "global_step": 94015, "epoch": 1056, "lr": 7.337689097609843e-05}
{"train_loss": 0.20576657354831696, "global_step": 94016, "epoch": 1056, "lr": 7.337637850298238e-05}
{"train_loss": 0.21336869895458221, "global_step": 94017, "epoch": 1056, "lr": 7.337586602672368e-05}
{"train_loss": 0.18786604702472687, "global_step": 94018, "epoch": 1056, "lr": 7.337535354732238e-05}
{"train_loss": 0.2492312788963318, "global_step": 94019, "epoch": 1056, "lr": 7.337484106477854e-05}
{"train_loss": 0.1916089951992035, "global_step": 94020, "epoch": 1056, "lr": 7.337432857909224e-05}
{"train_loss": 0.13987113535404205, "global_step": 94021, "epoch": 1056, "lr": 7.337381609026355e-05}
{"train_loss": 0.15629985928535461, "global_step": 94022, "epoch": 1056, "lr": 7.337330359829255e-05}
{"train_loss": 0.19215212762355804, "global_step": 94023, "epoch": 1056, "lr": 7.337279110317928e-05}
{"train_loss": 0.21039274334907532, "global_step": 94024, "epoch": 1056, "lr": 7.33722786049238e-05}
{"train_loss": 0.13805419206619263, "global_step": 94025, "epoch": 1056, "lr": 7.337176610352622e-05}
{"train_loss": 0.2062930017709732, "global_step": 94026, "epoch": 1056, "lr": 7.33712535989866e-05}
{"train_loss": 0.23981459438800812, "global_step": 94027, "epoch": 1056, "lr": 7.337074109130499e-05}
{"train_loss": 0.15671731531620026, "global_step": 94028, "epoch": 1056, "lr": 7.337022858048148e-05}
{"train_loss": 0.1452348828315735, "global_step": 94029, "epoch": 1056, "lr": 7.336971606651612e-05}
{"train_loss": 0.17253214120864868, "global_step": 94030, "epoch": 1056, "lr": 7.336920354940899e-05}
{"train_loss": 0.194380521774292, "global_step": 94031, "epoch": 1056, "lr": 7.336869102916016e-05}
{"train_loss": 0.18352600932121277, "global_step": 94032, "epoch": 1056, "lr": 7.336817850576967e-05}
{"train_loss": 0.14041826128959656, "global_step": 94033, "epoch": 1056, "lr": 7.336766597923764e-05}
{"train_loss": 0.19057638943195343, "global_step": 94034, "epoch": 1056, "lr": 7.33671534495641e-05}
{"train_loss": 0.23734183609485626, "global_step": 94035, "epoch": 1056, "lr": 7.336664091674913e-05}
{"train_loss": 0.10633757710456848, "global_step": 94036, "epoch": 1056, "lr": 7.33661283807928e-05}
{"train_loss": 0.16795317828655243, "global_step": 94037, "epoch": 1056, "lr": 7.336561584169518e-05}
{"train_loss": 0.19949576258659363, "global_step": 94038, "epoch": 1056, "lr": 7.336510329945634e-05}
{"train_loss": 0.1473754644393921, "global_step": 94039, "epoch": 1056, "lr": 7.336459075407635e-05}
{"train_loss": 0.1535414755344391, "global_step": 94040, "epoch": 1056, "lr": 7.336407820555527e-05}
{"train_loss": 0.21795572340488434, "global_step": 94041, "epoch": 1056, "lr": 7.336356565389317e-05}
{"train_loss": 0.19962744414806366, "global_step": 94042, "epoch": 1056, "lr": 7.336305309909013e-05}
{"train_loss": 0.17978253960609436, "global_step": 94043, "epoch": 1056, "lr": 7.336254054114622e-05}
{"train_loss": 0.2215038239955902, "global_step": 94044, "epoch": 1056, "lr": 7.336202798006148e-05}
{"train_loss": 0.18543799221515656, "global_step": 94045, "epoch": 1056, "lr": 7.336151541583602e-05}
{"train_loss": 0.20094884932041168, "global_step": 94046, "epoch": 1056, "lr": 7.336100284846988e-05}
{"train_loss": 0.20431175827980042, "global_step": 94047, "epoch": 1056, "lr": 7.336049027796314e-05}
{"train_loss": 0.15878087282180786, "global_step": 94048, "epoch": 1056, "lr": 7.335997770431586e-05}
{"train_loss": 0.13683095574378967, "global_step": 94049, "epoch": 1056, "lr": 7.335946512752813e-05}
{"train_loss": 0.1835470199584961, "global_step": 94050, "epoch": 1056, "lr": 7.335895254759999e-05}
{"train_loss": 0.12380306422710419, "global_step": 94051, "epoch": 1056, "lr": 7.335843996453153e-05}
{"train_loss": 0.168191060423851, "global_step": 94052, "epoch": 1056, "lr": 7.335792737832281e-05}
{"train_loss": 0.16107326745986938, "global_step": 94053, "epoch": 1056, "lr": 7.335741478897392e-05}
{"train_loss": 0.18607954680919647, "global_step": 94054, "epoch": 1056, "lr": 7.335690219648487e-05}
{"train_loss": 0.18388761579990387, "global_step": 94055, "epoch": 1056, "lr": 7.33563896008558e-05}
{"train_loss": 0.11116749793291092, "global_step": 94056, "epoch": 1056, "lr": 7.335587700208675e-05}
{"train_loss": 0.2313164919614792, "global_step": 94057, "epoch": 1056, "lr": 7.335536440017778e-05}
{"train_loss": 0.14823076128959656, "global_step": 94058, "epoch": 1056, "lr": 7.335485179512897e-05}
{"train_loss": 0.14560841023921967, "global_step": 94059, "epoch": 1056, "lr": 7.335433918694036e-05}
{"train_loss": 0.19335202872753143, "global_step": 94060, "epoch": 1056, "lr": 7.335382657561207e-05}
{"train_loss": 0.19378803670406342, "global_step": 94061, "epoch": 1056, "lr": 7.335331396114414e-05}
{"train_loss": 0.2281922698020935, "global_step": 94062, "epoch": 1056, "lr": 7.335280134353664e-05}
{"train_loss": 0.21459490060806274, "global_step": 94063, "epoch": 1056, "lr": 7.335228872278963e-05}
{"train_loss": 0.22173632681369781, "global_step": 94064, "epoch": 1056, "lr": 7.33517760989032e-05}
{"train_loss": 0.2509899139404297, "global_step": 94065, "epoch": 1056, "lr": 7.335126347187741e-05}
{"train_loss": 0.19674953818321228, "global_step": 94066, "epoch": 1056, "lr": 7.335075084171233e-05}
{"train_loss": 0.17496982216835022, "global_step": 94067, "epoch": 1056, "lr": 7.3350238208408e-05}
{"train_loss": 0.11556762456893921, "global_step": 94068, "epoch": 1056, "lr": 7.334972557196455e-05}
{"train_loss": 0.20673803985118866, "global_step": 94069, "epoch": 1056, "lr": 7.334921293238199e-05}
{"train_loss": 0.203502357006073, "global_step": 94070, "epoch": 1056, "lr": 7.334870028966042e-05}
{"train_loss": 0.10761281102895737, "global_step": 94071, "epoch": 1056, "lr": 7.33481876437999e-05}
{"train_loss": 0.19067199829589115, "global_step": 94072, "epoch": 1056, "lr": 7.33476749948005e-05, "val_loss": 4.349422931671143}
{"train_loss": 0.1212274432182312, "global_step": 94073, "epoch": 1057, "lr": 7.33471623426623e-05}
{"train_loss": 0.19839248061180115, "global_step": 94074, "epoch": 1057, "lr": 7.334664968738535e-05}
{"train_loss": 0.16444283723831177, "global_step": 94075, "epoch": 1057, "lr": 7.334613702896973e-05}
{"train_loss": 0.09063124656677246, "global_step": 94076, "epoch": 1057, "lr": 7.33456243674155e-05}
{"train_loss": 0.16444960236549377, "global_step": 94077, "epoch": 1057, "lr": 7.334511170272275e-05}
{"train_loss": 0.17550109326839447, "global_step": 94078, "epoch": 1057, "lr": 7.334459903489152e-05}
{"train_loss": 0.19727827608585358, "global_step": 94079, "epoch": 1057, "lr": 7.33440863639219e-05}
{"train_loss": 0.195970818400383, "global_step": 94080, "epoch": 1057, "lr": 7.334357368981395e-05}
{"train_loss": 0.1581009179353714, "global_step": 94081, "epoch": 1057, "lr": 7.334306101256774e-05}
{"train_loss": 0.2079198658466339, "global_step": 94082, "epoch": 1057, "lr": 7.334254833218335e-05}
{"train_loss": 0.1663103699684143, "global_step": 94083, "epoch": 1057, "lr": 7.334203564866083e-05}
{"train_loss": 0.20048344135284424, "global_step": 94084, "epoch": 1057, "lr": 7.334152296200026e-05}
{"train_loss": 0.16902807354927063, "global_step": 94085, "epoch": 1057, "lr": 7.33410102722017e-05}
{"train_loss": 0.17342393100261688, "global_step": 94086, "epoch": 1057, "lr": 7.334049757926522e-05}
{"train_loss": 0.18122529983520508, "global_step": 94087, "epoch": 1057, "lr": 7.333998488319091e-05}
{"train_loss": 0.11911904066801071, "global_step": 94088, "epoch": 1057, "lr": 7.333947218397882e-05}
{"train_loss": 0.21059876680374146, "global_step": 94089, "epoch": 1057, "lr": 7.333895948162902e-05}
{"train_loss": 0.1399446576833725, "global_step": 94090, "epoch": 1057, "lr": 7.333844677614159e-05}
{"train_loss": 0.1603740006685257, "global_step": 94091, "epoch": 1057, "lr": 7.333793406751656e-05}
{"train_loss": 0.20392222702503204, "global_step": 94092, "epoch": 1057, "lr": 7.333742135575406e-05}
{"train_loss": 0.2543182671070099, "global_step": 94093, "epoch": 1057, "lr": 7.333690864085413e-05}
{"train_loss": 0.17039193212985992, "global_step": 94094, "epoch": 1057, "lr": 7.333639592281682e-05}
{"train_loss": 0.20175111293792725, "global_step": 94095, "epoch": 1057, "lr": 7.333588320164223e-05}
{"train_loss": 0.19234150648117065, "global_step": 94096, "epoch": 1057, "lr": 7.333537047733042e-05}
{"train_loss": 0.1797969788312912, "global_step": 94097, "epoch": 1057, "lr": 7.333485774988143e-05}
{"train_loss": 0.15856730937957764, "global_step": 94098, "epoch": 1057, "lr": 7.333434501929537e-05}
{"train_loss": 0.13535268604755402, "global_step": 94099, "epoch": 1057, "lr": 7.333383228557227e-05}
{"train_loss": 0.176923468708992, "global_step": 94100, "epoch": 1057, "lr": 7.333331954871224e-05}
{"train_loss": 0.1866118162870407, "global_step": 94101, "epoch": 1057, "lr": 7.333280680871533e-05}
{"train_loss": 0.22202934324741364, "global_step": 94102, "epoch": 1057, "lr": 7.333229406558161e-05}
{"train_loss": 0.2116677612066269, "global_step": 94103, "epoch": 1057, "lr": 7.333178131931115e-05}
{"train_loss": 0.30749452114105225, "global_step": 94104, "epoch": 1057, "lr": 7.3331268569904e-05}
{"train_loss": 0.18566037714481354, "global_step": 94105, "epoch": 1057, "lr": 7.333075581736025e-05}
{"train_loss": 0.20049142837524414, "global_step": 94106, "epoch": 1057, "lr": 7.333024306167997e-05}
{"train_loss": 0.18486471474170685, "global_step": 94107, "epoch": 1057, "lr": 7.332973030286323e-05}
{"train_loss": 0.21384039521217346, "global_step": 94108, "epoch": 1057, "lr": 7.332921754091008e-05}
{"train_loss": 0.17539075016975403, "global_step": 94109, "epoch": 1057, "lr": 7.332870477582059e-05}
{"train_loss": 0.22599568963050842, "global_step": 94110, "epoch": 1057, "lr": 7.332819200759487e-05}
{"train_loss": 0.23947639763355255, "global_step": 94111, "epoch": 1057, "lr": 7.332767923623294e-05}
{"train_loss": 0.19695138931274414, "global_step": 94112, "epoch": 1057, "lr": 7.33271664617349e-05}
{"train_loss": 0.1824697107076645, "global_step": 94113, "epoch": 1057, "lr": 7.332665368410078e-05}
{"train_loss": 0.14966697990894318, "global_step": 94114, "epoch": 1057, "lr": 7.33261409033307e-05}
{"train_loss": 0.19154813885688782, "global_step": 94115, "epoch": 1057, "lr": 7.332562811942471e-05}
{"train_loss": 0.1600329428911209, "global_step": 94116, "epoch": 1057, "lr": 7.332511533238285e-05}
{"train_loss": 0.13182269036769867, "global_step": 94117, "epoch": 1057, "lr": 7.332460254220524e-05}
{"train_loss": 0.13518162071704865, "global_step": 94118, "epoch": 1057, "lr": 7.33240897488919e-05}
{"train_loss": 0.22078558802604675, "global_step": 94119, "epoch": 1057, "lr": 7.332357695244292e-05}
{"train_loss": 0.21715208888053894, "global_step": 94120, "epoch": 1057, "lr": 7.332306415285838e-05}
{"train_loss": 0.17071132361888885, "global_step": 94121, "epoch": 1057, "lr": 7.332255135013833e-05}
{"train_loss": 0.19380535185337067, "global_step": 94122, "epoch": 1057, "lr": 7.332203854428285e-05}
{"train_loss": 0.24617470800876617, "global_step": 94123, "epoch": 1057, "lr": 7.332152573529202e-05}
{"train_loss": 0.17704623937606812, "global_step": 94124, "epoch": 1057, "lr": 7.332101292316587e-05}
{"train_loss": 0.17125113308429718, "global_step": 94125, "epoch": 1057, "lr": 7.33205001079045e-05}
{"train_loss": 0.13101062178611755, "global_step": 94126, "epoch": 1057, "lr": 7.3319987289508e-05}
{"train_loss": 0.24125215411186218, "global_step": 94127, "epoch": 1057, "lr": 7.331947446797638e-05}
{"train_loss": 0.16221491992473602, "global_step": 94128, "epoch": 1057, "lr": 7.331896164330975e-05}
{"train_loss": 0.19365248084068298, "global_step": 94129, "epoch": 1057, "lr": 7.331844881550817e-05}
{"train_loss": 0.16995003819465637, "global_step": 94130, "epoch": 1057, "lr": 7.331793598457171e-05}
{"train_loss": 0.20448023080825806, "global_step": 94131, "epoch": 1057, "lr": 7.331742315050045e-05}
{"train_loss": 0.16458918154239655, "global_step": 94132, "epoch": 1057, "lr": 7.331691031329444e-05}
{"train_loss": 0.22561874985694885, "global_step": 94133, "epoch": 1057, "lr": 7.331639747295375e-05}
{"train_loss": 0.12150312960147858, "global_step": 94134, "epoch": 1057, "lr": 7.331588462947846e-05}
{"train_loss": 0.23738448321819305, "global_step": 94135, "epoch": 1057, "lr": 7.331537178286862e-05}
{"train_loss": 0.1189965307712555, "global_step": 94136, "epoch": 1057, "lr": 7.331485893312433e-05}
{"train_loss": 0.1341807097196579, "global_step": 94137, "epoch": 1057, "lr": 7.331434608024563e-05}
{"train_loss": 0.15929701924324036, "global_step": 94138, "epoch": 1057, "lr": 7.331383322423259e-05}
{"train_loss": 0.19054235517978668, "global_step": 94139, "epoch": 1057, "lr": 7.331332036508531e-05}
{"train_loss": 0.16719256341457367, "global_step": 94140, "epoch": 1057, "lr": 7.331280750280383e-05}
{"train_loss": 0.2565837502479553, "global_step": 94141, "epoch": 1057, "lr": 7.331229463738824e-05}
{"train_loss": 0.1356554627418518, "global_step": 94142, "epoch": 1057, "lr": 7.331178176883858e-05}
{"train_loss": 0.14042799174785614, "global_step": 94143, "epoch": 1057, "lr": 7.331126889715494e-05}
{"train_loss": 0.11886803805828094, "global_step": 94144, "epoch": 1057, "lr": 7.331075602233739e-05}
{"train_loss": 0.14836402237415314, "global_step": 94145, "epoch": 1057, "lr": 7.331024314438597e-05}
{"train_loss": 0.16987299919128418, "global_step": 94146, "epoch": 1057, "lr": 7.33097302633008e-05}
{"train_loss": 0.2702598571777344, "global_step": 94147, "epoch": 1057, "lr": 7.33092173790819e-05}
{"train_loss": 0.1418331116437912, "global_step": 94148, "epoch": 1057, "lr": 7.330870449172937e-05}
{"train_loss": 0.1384916603565216, "global_step": 94149, "epoch": 1057, "lr": 7.330819160124326e-05}
{"train_loss": 0.233443021774292, "global_step": 94150, "epoch": 1057, "lr": 7.330767870762367e-05}
{"train_loss": 0.18277999758720398, "global_step": 94151, "epoch": 1057, "lr": 7.330716581087063e-05}
{"train_loss": 0.14608196914196014, "global_step": 94152, "epoch": 1057, "lr": 7.330665291098424e-05}
{"train_loss": 0.21176213026046753, "global_step": 94153, "epoch": 1057, "lr": 7.330614000796452e-05}
{"train_loss": 0.11739812046289444, "global_step": 94154, "epoch": 1057, "lr": 7.330562710181162e-05}
{"train_loss": 0.15829628705978394, "global_step": 94155, "epoch": 1057, "lr": 7.330511419252553e-05}
{"train_loss": 0.19465041160583496, "global_step": 94156, "epoch": 1057, "lr": 7.330460128010635e-05}
{"train_loss": 0.21603921055793762, "global_step": 94157, "epoch": 1057, "lr": 7.330408836455417e-05}
{"train_loss": 0.21432721614837646, "global_step": 94158, "epoch": 1057, "lr": 7.330357544586903e-05}
{"train_loss": 0.20745190978050232, "global_step": 94159, "epoch": 1057, "lr": 7.3303062524051e-05}
{"train_loss": 0.1728297621011734, "global_step": 94160, "epoch": 1057, "lr": 7.330254959910017e-05}
{"train_loss": 0.1811359923207358, "global_step": 94161, "epoch": 1057, "lr": 7.330203667101659e-05, "val_loss": 4.351891040802002}
{"train_loss": 0.2704751491546631, "global_step": 94162, "epoch": 1058, "lr": 7.330152373980034e-05}
{"train_loss": 0.1231110468506813, "global_step": 94163, "epoch": 1058, "lr": 7.330101080545147e-05}
{"train_loss": 0.2014595866203308, "global_step": 94164, "epoch": 1058, "lr": 7.330049786797007e-05}
{"train_loss": 0.13822922110557556, "global_step": 94165, "epoch": 1058, "lr": 7.329998492735622e-05}
{"train_loss": 0.1633000522851944, "global_step": 94166, "epoch": 1058, "lr": 7.329947198360994e-05}
{"train_loss": 0.15177640318870544, "global_step": 94167, "epoch": 1058, "lr": 7.329895903673135e-05}
{"train_loss": 0.2169630229473114, "global_step": 94168, "epoch": 1058, "lr": 7.32984460867205e-05}
{"train_loss": 0.1914564073085785, "global_step": 94169, "epoch": 1058, "lr": 7.329793313357746e-05}
{"train_loss": 0.18261446058750153, "global_step": 94170, "epoch": 1058, "lr": 7.32974201773023e-05}
{"train_loss": 0.18836183845996857, "global_step": 94171, "epoch": 1058, "lr": 7.329690721789507e-05}
{"train_loss": 0.1524592489004135, "global_step": 94172, "epoch": 1058, "lr": 7.329639425535586e-05}
{"train_loss": 0.2115427851676941, "global_step": 94173, "epoch": 1058, "lr": 7.329588128968474e-05}
{"train_loss": 0.16687731444835663, "global_step": 94174, "epoch": 1058, "lr": 7.329536832088176e-05}
{"train_loss": 0.17415735125541687, "global_step": 94175, "epoch": 1058, "lr": 7.3294855348947e-05}
{"train_loss": 0.14308035373687744, "global_step": 94176, "epoch": 1058, "lr": 7.329434237388056e-05}
{"train_loss": 0.2190762609243393, "global_step": 94177, "epoch": 1058, "lr": 7.329382939568245e-05}
{"train_loss": 0.12563030421733856, "global_step": 94178, "epoch": 1058, "lr": 7.329331641435277e-05}
{"train_loss": 0.14381355047225952, "global_step": 94179, "epoch": 1058, "lr": 7.329280342989162e-05}
{"train_loss": 0.16319149732589722, "global_step": 94180, "epoch": 1058, "lr": 7.3292290442299e-05}
{"train_loss": 0.20084427297115326, "global_step": 94181, "epoch": 1058, "lr": 7.329177745157503e-05}
{"train_loss": 0.17000547051429749, "global_step": 94182, "epoch": 1058, "lr": 7.329126445771975e-05}
{"train_loss": 0.0943322405219078, "global_step": 94183, "epoch": 1058, "lr": 7.329075146073325e-05}
{"train_loss": 0.12415661662817001, "global_step": 94184, "epoch": 1058, "lr": 7.32902384606156e-05}
{"train_loss": 0.158540740609169, "global_step": 94185, "epoch": 1058, "lr": 7.328972545736685e-05}
{"train_loss": 0.19633927941322327, "global_step": 94186, "epoch": 1058, "lr": 7.32892124509871e-05}
{"train_loss": 0.09432948380708694, "global_step": 94187, "epoch": 1058, "lr": 7.328869944147638e-05}
{"train_loss": 0.13895322382450104, "global_step": 94188, "epoch": 1058, "lr": 7.328818642883479e-05}
{"train_loss": 0.18739427626132965, "global_step": 94189, "epoch": 1058, "lr": 7.328767341306239e-05}
{"train_loss": 0.21142537891864777, "global_step": 94190, "epoch": 1058, "lr": 7.328716039415925e-05}
{"train_loss": 0.17209215462207794, "global_step": 94191, "epoch": 1058, "lr": 7.328664737212543e-05}
{"train_loss": 0.22606031596660614, "global_step": 94192, "epoch": 1058, "lr": 7.3286134346961e-05}
{"train_loss": 0.2228716015815735, "global_step": 94193, "epoch": 1058, "lr": 7.328562131866602e-05}
{"train_loss": 0.21259361505508423, "global_step": 94194, "epoch": 1058, "lr": 7.328510828724058e-05}
{"train_loss": 0.2691926062107086, "global_step": 94195, "epoch": 1058, "lr": 7.328459525268475e-05}
{"train_loss": 0.2080254852771759, "global_step": 94196, "epoch": 1058, "lr": 7.328408221499859e-05}
{"train_loss": 0.12823383510112762, "global_step": 94197, "epoch": 1058, "lr": 7.328356917418216e-05}
{"train_loss": 0.18008361756801605, "global_step": 94198, "epoch": 1058, "lr": 7.328305613023553e-05}
{"train_loss": 0.2177453637123108, "global_step": 94199, "epoch": 1058, "lr": 7.32825430831588e-05}
{"train_loss": 0.09755431115627289, "global_step": 94200, "epoch": 1058, "lr": 7.3282030032952e-05}
{"train_loss": 0.1578001081943512, "global_step": 94201, "epoch": 1058, "lr": 7.328151697961522e-05}
{"train_loss": 0.19405780732631683, "global_step": 94202, "epoch": 1058, "lr": 7.328100392314851e-05}
{"train_loss": 0.13978035748004913, "global_step": 94203, "epoch": 1058, "lr": 7.328049086355197e-05}
{"train_loss": 0.14578494429588318, "global_step": 94204, "epoch": 1058, "lr": 7.327997780082565e-05}
{"train_loss": 0.1359775811433792, "global_step": 94205, "epoch": 1058, "lr": 7.327946473496962e-05}
{"train_loss": 0.19786116480827332, "global_step": 94206, "epoch": 1058, "lr": 7.327895166598396e-05}
{"train_loss": 0.1693393439054489, "global_step": 94207, "epoch": 1058, "lr": 7.327843859386871e-05}
{"train_loss": 0.2572783827781677, "global_step": 94208, "epoch": 1058, "lr": 7.327792551862398e-05}
{"train_loss": 0.20095817744731903, "global_step": 94209, "epoch": 1058, "lr": 7.32774124402498e-05}
{"train_loss": 0.11355952173471451, "global_step": 94210, "epoch": 1058, "lr": 7.327689935874626e-05}
{"train_loss": 0.12992681562900543, "global_step": 94211, "epoch": 1058, "lr": 7.327638627411341e-05}
{"train_loss": 0.20684275031089783, "global_step": 94212, "epoch": 1058, "lr": 7.327587318635136e-05}
{"train_loss": 0.1337958127260208, "global_step": 94213, "epoch": 1058, "lr": 7.327536009546012e-05}
{"train_loss": 0.15360334515571594, "global_step": 94214, "epoch": 1058, "lr": 7.327484700143981e-05}
{"train_loss": 0.15512487292289734, "global_step": 94215, "epoch": 1058, "lr": 7.327433390429048e-05}
{"train_loss": 0.2520924508571625, "global_step": 94216, "epoch": 1058, "lr": 7.32738208040122e-05}
{"train_loss": 0.1817624568939209, "global_step": 94217, "epoch": 1058, "lr": 7.327330770060504e-05}
{"train_loss": 0.19593262672424316, "global_step": 94218, "epoch": 1058, "lr": 7.327279459406906e-05}
{"train_loss": 0.21127896010875702, "global_step": 94219, "epoch": 1058, "lr": 7.327228148440433e-05}
{"train_loss": 0.17054425179958344, "global_step": 94220, "epoch": 1058, "lr": 7.327176837161095e-05}
{"train_loss": 0.1336587816476822, "global_step": 94221, "epoch": 1058, "lr": 7.327125525568894e-05}
{"train_loss": 0.2116318941116333, "global_step": 94222, "epoch": 1058, "lr": 7.327074213663842e-05}
{"train_loss": 0.16232042014598846, "global_step": 94223, "epoch": 1058, "lr": 7.327022901445941e-05}
{"train_loss": 0.15514664351940155, "global_step": 94224, "epoch": 1058, "lr": 7.326971588915201e-05}
{"train_loss": 0.14990980923175812, "global_step": 94225, "epoch": 1058, "lr": 7.326920276071628e-05}
{"train_loss": 0.21410350501537323, "global_step": 94226, "epoch": 1058, "lr": 7.326868962915229e-05}
{"train_loss": 0.13634228706359863, "global_step": 94227, "epoch": 1058, "lr": 7.326817649446011e-05}
{"train_loss": 0.1468854546546936, "global_step": 94228, "epoch": 1058, "lr": 7.32676633566398e-05}
{"train_loss": 0.17581094801425934, "global_step": 94229, "epoch": 1058, "lr": 7.326715021569144e-05}
{"train_loss": 0.15144062042236328, "global_step": 94230, "epoch": 1058, "lr": 7.326663707161509e-05}
{"train_loss": 0.21780447661876678, "global_step": 94231, "epoch": 1058, "lr": 7.326612392441084e-05}
{"train_loss": 0.1312136948108673, "global_step": 94232, "epoch": 1058, "lr": 7.326561077407873e-05}
{"train_loss": 0.16193833947181702, "global_step": 94233, "epoch": 1058, "lr": 7.326509762061885e-05}
{"train_loss": 0.24453529715538025, "global_step": 94234, "epoch": 1058, "lr": 7.326458446403124e-05}
{"train_loss": 0.1837736964225769, "global_step": 94235, "epoch": 1058, "lr": 7.326407130431602e-05}
{"train_loss": 0.1526387482881546, "global_step": 94236, "epoch": 1058, "lr": 7.32635581414732e-05}
{"train_loss": 0.09775134176015854, "global_step": 94237, "epoch": 1058, "lr": 7.326304497550289e-05}
{"train_loss": 0.13532239198684692, "global_step": 94238, "epoch": 1058, "lr": 7.326253180640515e-05}
{"train_loss": 0.17894379794597626, "global_step": 94239, "epoch": 1058, "lr": 7.326201863418005e-05}
{"train_loss": 0.18737491965293884, "global_step": 94240, "epoch": 1058, "lr": 7.326150545882767e-05}
{"train_loss": 0.20991624891757965, "global_step": 94241, "epoch": 1058, "lr": 7.326099228034803e-05}
{"train_loss": 0.13983334600925446, "global_step": 94242, "epoch": 1058, "lr": 7.326047909874126e-05}
{"train_loss": 0.20706403255462646, "global_step": 94243, "epoch": 1058, "lr": 7.325996591400737e-05}
{"train_loss": 0.14269037544727325, "global_step": 94244, "epoch": 1058, "lr": 7.32594527261465e-05}
{"train_loss": 0.1600358635187149, "global_step": 94245, "epoch": 1058, "lr": 7.325893953515865e-05}
{"train_loss": 0.15860001742839813, "global_step": 94246, "epoch": 1058, "lr": 7.325842634104394e-05}
{"train_loss": 0.10149139165878296, "global_step": 94247, "epoch": 1058, "lr": 7.325791314380241e-05}
{"train_loss": 0.17120103538036346, "global_step": 94248, "epoch": 1058, "lr": 7.325739994343414e-05}
{"train_loss": 0.15515966713428497, "global_step": 94249, "epoch": 1058, "lr": 7.32568867399392e-05}
{"train_loss": 0.17061304016394563, "global_step": 94250, "epoch": 1058, "lr": 7.325637353331764e-05, "val_loss": 4.491343975067139}
{"train_loss": 0.15913783013820648, "global_step": 94251, "epoch": 1059, "lr": 7.325586032356955e-05}
{"train_loss": 0.14455285668373108, "global_step": 94252, "epoch": 1059, "lr": 7.3255347110695e-05}
{"train_loss": 0.13063254952430725, "global_step": 94253, "epoch": 1059, "lr": 7.325483389469405e-05}
{"train_loss": 0.185822993516922, "global_step": 94254, "epoch": 1059, "lr": 7.325432067556676e-05}
{"train_loss": 0.14084424078464508, "global_step": 94255, "epoch": 1059, "lr": 7.325380745331321e-05}
{"train_loss": 0.16402709484100342, "global_step": 94256, "epoch": 1059, "lr": 7.325329422793348e-05}
{"train_loss": 0.1438826322555542, "global_step": 94257, "epoch": 1059, "lr": 7.325278099942763e-05}
{"train_loss": 0.20230120420455933, "global_step": 94258, "epoch": 1059, "lr": 7.325226776779571e-05}
{"train_loss": 0.2116628885269165, "global_step": 94259, "epoch": 1059, "lr": 7.325175453303782e-05}
{"train_loss": 0.10020927339792252, "global_step": 94260, "epoch": 1059, "lr": 7.325124129515402e-05}
{"train_loss": 0.1499810814857483, "global_step": 94261, "epoch": 1059, "lr": 7.325072805414436e-05}
{"train_loss": 0.1675424426794052, "global_step": 94262, "epoch": 1059, "lr": 7.325021481000894e-05}
{"train_loss": 0.14931011199951172, "global_step": 94263, "epoch": 1059, "lr": 7.32497015627478e-05}
{"train_loss": 0.11628267168998718, "global_step": 94264, "epoch": 1059, "lr": 7.324918831236101e-05}
{"train_loss": 0.24884521961212158, "global_step": 94265, "epoch": 1059, "lr": 7.324867505884866e-05}
{"train_loss": 0.2062542736530304, "global_step": 94266, "epoch": 1059, "lr": 7.324816180221082e-05}
{"train_loss": 0.23226025700569153, "global_step": 94267, "epoch": 1059, "lr": 7.324764854244754e-05}
{"train_loss": 0.22043180465698242, "global_step": 94268, "epoch": 1059, "lr": 7.324713527955892e-05}
{"train_loss": 0.20996227860450745, "global_step": 94269, "epoch": 1059, "lr": 7.324662201354497e-05}
{"train_loss": 0.21583469212055206, "global_step": 94270, "epoch": 1059, "lr": 7.324610874440581e-05}
{"train_loss": 0.18002170324325562, "global_step": 94271, "epoch": 1059, "lr": 7.324559547214148e-05}
{"train_loss": 0.12448384612798691, "global_step": 94272, "epoch": 1059, "lr": 7.324508219675207e-05}
{"train_loss": 0.2117116004228592, "global_step": 94273, "epoch": 1059, "lr": 7.324456891823764e-05}
{"train_loss": 0.23463478684425354, "global_step": 94274, "epoch": 1059, "lr": 7.324405563659828e-05}
{"train_loss": 0.1162739172577858, "global_step": 94275, "epoch": 1059, "lr": 7.324354235183402e-05}
{"train_loss": 0.1511228084564209, "global_step": 94276, "epoch": 1059, "lr": 7.324302906394495e-05}
{"train_loss": 0.11917097866535187, "global_step": 94277, "epoch": 1059, "lr": 7.324251577293115e-05}
{"train_loss": 0.18418945372104645, "global_step": 94278, "epoch": 1059, "lr": 7.324200247879267e-05}
{"train_loss": 0.15410418808460236, "global_step": 94279, "epoch": 1059, "lr": 7.324148918152959e-05}
{"train_loss": 0.15018433332443237, "global_step": 94280, "epoch": 1059, "lr": 7.324097588114198e-05}
{"train_loss": 0.21431465446949005, "global_step": 94281, "epoch": 1059, "lr": 7.324046257762988e-05}
{"train_loss": 0.15749284625053406, "global_step": 94282, "epoch": 1059, "lr": 7.323994927099342e-05}
{"train_loss": 0.1846451312303543, "global_step": 94283, "epoch": 1059, "lr": 7.32394359612326e-05}
{"train_loss": 0.15463167428970337, "global_step": 94284, "epoch": 1059, "lr": 7.323892264834754e-05}
{"train_loss": 0.15490075945854187, "global_step": 94285, "epoch": 1059, "lr": 7.323840933233828e-05}
{"train_loss": 0.20940952003002167, "global_step": 94286, "epoch": 1059, "lr": 7.323789601320492e-05}
{"train_loss": 0.18456260859966278, "global_step": 94287, "epoch": 1059, "lr": 7.323738269094748e-05}
{"train_loss": 0.18812644481658936, "global_step": 94288, "epoch": 1059, "lr": 7.323686936556607e-05}
{"train_loss": 0.22394324839115143, "global_step": 94289, "epoch": 1059, "lr": 7.323635603706074e-05}
{"train_loss": 0.1596364676952362, "global_step": 94290, "epoch": 1059, "lr": 7.323584270543157e-05}
{"train_loss": 0.18008799850940704, "global_step": 94291, "epoch": 1059, "lr": 7.323532937067861e-05}
{"train_loss": 0.14710679650306702, "global_step": 94292, "epoch": 1059, "lr": 7.323481603280197e-05}
{"train_loss": 0.15053202211856842, "global_step": 94293, "epoch": 1059, "lr": 7.323430269180167e-05}
{"train_loss": 0.1185971349477768, "global_step": 94294, "epoch": 1059, "lr": 7.323378934767782e-05}
{"train_loss": 0.24639108777046204, "global_step": 94295, "epoch": 1059, "lr": 7.323327600043045e-05}
{"train_loss": 0.18277719616889954, "global_step": 94296, "epoch": 1059, "lr": 7.323276265005966e-05}
{"train_loss": 0.19591663777828217, "global_step": 94297, "epoch": 1059, "lr": 7.323224929656551e-05}
{"train_loss": 0.19387246668338776, "global_step": 94298, "epoch": 1059, "lr": 7.323173593994807e-05}
{"train_loss": 0.26568642258644104, "global_step": 94299, "epoch": 1059, "lr": 7.323122258020739e-05}
{"train_loss": 0.1366736888885498, "global_step": 94300, "epoch": 1059, "lr": 7.323070921734357e-05}
{"train_loss": 0.12760108709335327, "global_step": 94301, "epoch": 1059, "lr": 7.323019585135666e-05}
{"train_loss": 0.1817517727613449, "global_step": 94302, "epoch": 1059, "lr": 7.322968248224673e-05}
{"train_loss": 0.2508498728275299, "global_step": 94303, "epoch": 1059, "lr": 7.322916911001386e-05}
{"train_loss": 0.14481684565544128, "global_step": 94304, "epoch": 1059, "lr": 7.322865573465811e-05}
{"train_loss": 0.26888081431388855, "global_step": 94305, "epoch": 1059, "lr": 7.322814235617955e-05}
{"train_loss": 0.21782554686069489, "global_step": 94306, "epoch": 1059, "lr": 7.322762897457825e-05}
{"train_loss": 0.15067444741725922, "global_step": 94307, "epoch": 1059, "lr": 7.322711558985428e-05}
{"train_loss": 0.18880675733089447, "global_step": 94308, "epoch": 1059, "lr": 7.32266022020077e-05}
{"train_loss": 0.21691885590553284, "global_step": 94309, "epoch": 1059, "lr": 7.322608881103858e-05}
{"train_loss": 0.14064590632915497, "global_step": 94310, "epoch": 1059, "lr": 7.3225575416947e-05}
{"train_loss": 0.1900261491537094, "global_step": 94311, "epoch": 1059, "lr": 7.322506201973302e-05}
{"train_loss": 0.2352050542831421, "global_step": 94312, "epoch": 1059, "lr": 7.322454861939674e-05}
{"train_loss": 0.19795118272304535, "global_step": 94313, "epoch": 1059, "lr": 7.322403521593819e-05}
{"train_loss": 0.149732306599617, "global_step": 94314, "epoch": 1059, "lr": 7.322352180935743e-05}
{"train_loss": 0.17833516001701355, "global_step": 94315, "epoch": 1059, "lr": 7.322300839965457e-05}
{"train_loss": 0.1455821543931961, "global_step": 94316, "epoch": 1059, "lr": 7.322249498682965e-05}
{"train_loss": 0.21940647065639496, "global_step": 94317, "epoch": 1059, "lr": 7.322198157088277e-05}
{"train_loss": 0.1614183783531189, "global_step": 94318, "epoch": 1059, "lr": 7.322146815181397e-05}
{"train_loss": 0.26087382435798645, "global_step": 94319, "epoch": 1059, "lr": 7.322095472962331e-05}
{"train_loss": 0.11177820712327957, "global_step": 94320, "epoch": 1059, "lr": 7.322044130431088e-05}
{"train_loss": 0.15756231546401978, "global_step": 94321, "epoch": 1059, "lr": 7.321992787587675e-05}
{"train_loss": 0.1808367818593979, "global_step": 94322, "epoch": 1059, "lr": 7.321941444432099e-05}
{"train_loss": 0.12085958570241928, "global_step": 94323, "epoch": 1059, "lr": 7.321890100964366e-05}
{"train_loss": 0.2356387972831726, "global_step": 94324, "epoch": 1059, "lr": 7.321838757184483e-05}
{"train_loss": 0.14431366324424744, "global_step": 94325, "epoch": 1059, "lr": 7.321787413092457e-05}
{"train_loss": 0.11175043135881424, "global_step": 94326, "epoch": 1059, "lr": 7.321736068688295e-05}
{"train_loss": 0.19881704449653625, "global_step": 94327, "epoch": 1059, "lr": 7.321684723972002e-05}
{"train_loss": 0.17361709475517273, "global_step": 94328, "epoch": 1059, "lr": 7.321633378943588e-05}
{"train_loss": 0.1648234724998474, "global_step": 94329, "epoch": 1059, "lr": 7.32158203360306e-05}
{"train_loss": 0.21671009063720703, "global_step": 94330, "epoch": 1059, "lr": 7.321530687950423e-05}
{"train_loss": 0.1355997920036316, "global_step": 94331, "epoch": 1059, "lr": 7.321479341985684e-05}
{"train_loss": 0.18369877338409424, "global_step": 94332, "epoch": 1059, "lr": 7.321427995708851e-05}
{"train_loss": 0.2475394904613495, "global_step": 94333, "epoch": 1059, "lr": 7.32137664911993e-05}
{"train_loss": 0.09912213683128357, "global_step": 94334, "epoch": 1059, "lr": 7.321325302218929e-05}
{"train_loss": 0.22384974360466003, "global_step": 94335, "epoch": 1059, "lr": 7.321273955005852e-05}
{"train_loss": 0.08962399512529373, "global_step": 94336, "epoch": 1059, "lr": 7.32122260748071e-05}
{"train_loss": 0.1627238541841507, "global_step": 94337, "epoch": 1059, "lr": 7.321171259643508e-05}
{"train_loss": 0.1444849967956543, "global_step": 94338, "epoch": 1059, "lr": 7.321119911494253e-05}
{"train_loss": 0.17595609143543778, "global_step": 94339, "epoch": 1059, "lr": 7.321068563032949e-05, "val_loss": 4.356892108917236}
{"train_loss": 0.09600463509559631, "global_step": 94340, "epoch": 1060, "lr": 7.321017214259607e-05}
{"train_loss": 0.15693654119968414, "global_step": 94341, "epoch": 1060, "lr": 7.320965865174234e-05}
{"train_loss": 0.22749508917331696, "global_step": 94342, "epoch": 1060, "lr": 7.320914515776834e-05}
{"train_loss": 0.17154814302921295, "global_step": 94343, "epoch": 1060, "lr": 7.320863166067416e-05}
{"train_loss": 0.2169092893600464, "global_step": 94344, "epoch": 1060, "lr": 7.320811816045987e-05}
{"train_loss": 0.17167912423610687, "global_step": 94345, "epoch": 1060, "lr": 7.320760465712551e-05}
{"train_loss": 0.17609010636806488, "global_step": 94346, "epoch": 1060, "lr": 7.320709115067118e-05}
{"train_loss": 0.14352907240390778, "global_step": 94347, "epoch": 1060, "lr": 7.320657764109695e-05}
{"train_loss": 0.2010224610567093, "global_step": 94348, "epoch": 1060, "lr": 7.320606412840285e-05}
{"train_loss": 0.23300723731517792, "global_step": 94349, "epoch": 1060, "lr": 7.3205550612589e-05}
{"train_loss": 0.2124544233083725, "global_step": 94350, "epoch": 1060, "lr": 7.320503709365545e-05}
{"train_loss": 0.20218166708946228, "global_step": 94351, "epoch": 1060, "lr": 7.320452357160225e-05}
{"train_loss": 0.10740027576684952, "global_step": 94352, "epoch": 1060, "lr": 7.32040100464295e-05}
{"train_loss": 0.2104906439781189, "global_step": 94353, "epoch": 1060, "lr": 7.320349651813724e-05}
{"train_loss": 0.16009582579135895, "global_step": 94354, "epoch": 1060, "lr": 7.320298298672557e-05}
{"train_loss": 0.18007208406925201, "global_step": 94355, "epoch": 1060, "lr": 7.320246945219453e-05}
{"train_loss": 0.17240102589130402, "global_step": 94356, "epoch": 1060, "lr": 7.32019559145442e-05}
{"train_loss": 0.1662033349275589, "global_step": 94357, "epoch": 1060, "lr": 7.320144237377465e-05}
{"train_loss": 0.172218918800354, "global_step": 94358, "epoch": 1060, "lr": 7.320092882988595e-05}
{"train_loss": 0.14099614322185516, "global_step": 94359, "epoch": 1060, "lr": 7.320041528287817e-05}
{"train_loss": 0.1510009616613388, "global_step": 94360, "epoch": 1060, "lr": 7.319990173275138e-05}
{"train_loss": 0.14929549396038055, "global_step": 94361, "epoch": 1060, "lr": 7.319938817950563e-05}
{"train_loss": 0.15164795517921448, "global_step": 94362, "epoch": 1060, "lr": 7.319887462314103e-05}
{"train_loss": 0.15414267778396606, "global_step": 94363, "epoch": 1060, "lr": 7.319836106365761e-05}
{"train_loss": 0.17881041765213013, "global_step": 94364, "epoch": 1060, "lr": 7.319784750105546e-05}
{"train_loss": 0.11362618207931519, "global_step": 94365, "epoch": 1060, "lr": 7.319733393533462e-05}
{"train_loss": 0.19610169529914856, "global_step": 94366, "epoch": 1060, "lr": 7.31968203664952e-05}
{"train_loss": 0.12646552920341492, "global_step": 94367, "epoch": 1060, "lr": 7.319630679453725e-05}
{"train_loss": 0.10540755093097687, "global_step": 94368, "epoch": 1060, "lr": 7.319579321946084e-05}
{"train_loss": 0.16614152491092682, "global_step": 94369, "epoch": 1060, "lr": 7.319527964126603e-05}
{"train_loss": 0.21148288249969482, "global_step": 94370, "epoch": 1060, "lr": 7.31947660599529e-05}
{"train_loss": 0.1579141765832901, "global_step": 94371, "epoch": 1060, "lr": 7.319425247552152e-05}
{"train_loss": 0.15198341012001038, "global_step": 94372, "epoch": 1060, "lr": 7.319373888797195e-05}
{"train_loss": 0.16703565418720245, "global_step": 94373, "epoch": 1060, "lr": 7.319322529730428e-05}
{"train_loss": 0.16354107856750488, "global_step": 94374, "epoch": 1060, "lr": 7.319271170351855e-05}
{"train_loss": 0.21000519394874573, "global_step": 94375, "epoch": 1060, "lr": 7.319219810661484e-05}
{"train_loss": 0.12319376319646835, "global_step": 94376, "epoch": 1060, "lr": 7.319168450659322e-05}
{"train_loss": 0.1584705412387848, "global_step": 94377, "epoch": 1060, "lr": 7.319117090345378e-05}
{"train_loss": 0.16381748020648956, "global_step": 94378, "epoch": 1060, "lr": 7.319065729719655e-05}
{"train_loss": 0.18630020320415497, "global_step": 94379, "epoch": 1060, "lr": 7.319014368782162e-05}
{"train_loss": 0.1792038083076477, "global_step": 94380, "epoch": 1060, "lr": 7.318963007532908e-05}
{"train_loss": 0.15146556496620178, "global_step": 94381, "epoch": 1060, "lr": 7.318911645971895e-05}
{"train_loss": 0.15055394172668457, "global_step": 94382, "epoch": 1060, "lr": 7.318860284099134e-05}
{"train_loss": 0.1760953813791275, "global_step": 94383, "epoch": 1060, "lr": 7.31880892191463e-05}
{"train_loss": 0.2287018895149231, "global_step": 94384, "epoch": 1060, "lr": 7.31875755941839e-05}
{"train_loss": 0.15660567581653595, "global_step": 94385, "epoch": 1060, "lr": 7.318706196610422e-05}
{"train_loss": 0.22827720642089844, "global_step": 94386, "epoch": 1060, "lr": 7.318654833490732e-05}
{"train_loss": 0.17009079456329346, "global_step": 94387, "epoch": 1060, "lr": 7.318603470059327e-05}
{"train_loss": 0.19705666601657867, "global_step": 94388, "epoch": 1060, "lr": 7.318552106316214e-05}
{"train_loss": 0.23006704449653625, "global_step": 94389, "epoch": 1060, "lr": 7.318500742261401e-05}
{"train_loss": 0.15614616870880127, "global_step": 94390, "epoch": 1060, "lr": 7.318449377894893e-05}
{"train_loss": 0.2197977751493454, "global_step": 94391, "epoch": 1060, "lr": 7.318398013216698e-05}
{"train_loss": 0.16110332310199738, "global_step": 94392, "epoch": 1060, "lr": 7.318346648226821e-05}
{"train_loss": 0.1354473978281021, "global_step": 94393, "epoch": 1060, "lr": 7.318295282925273e-05}
{"train_loss": 0.14926260709762573, "global_step": 94394, "epoch": 1060, "lr": 7.318243917312058e-05}
{"train_loss": 0.15874281525611877, "global_step": 94395, "epoch": 1060, "lr": 7.318192551387182e-05}
{"train_loss": 0.2204940915107727, "global_step": 94396, "epoch": 1060, "lr": 7.318141185150653e-05}
{"train_loss": 0.15286056697368622, "global_step": 94397, "epoch": 1060, "lr": 7.318089818602481e-05}
{"train_loss": 0.19646455347537994, "global_step": 94398, "epoch": 1060, "lr": 7.318038451742668e-05}
{"train_loss": 0.1437448263168335, "global_step": 94399, "epoch": 1060, "lr": 7.317987084571223e-05}
{"train_loss": 0.1513815075159073, "global_step": 94400, "epoch": 1060, "lr": 7.317935717088152e-05}
{"train_loss": 0.15024183690547943, "global_step": 94401, "epoch": 1060, "lr": 7.317884349293465e-05}
{"train_loss": 0.16861368715763092, "global_step": 94402, "epoch": 1060, "lr": 7.317832981187164e-05}
{"train_loss": 0.13707850873470306, "global_step": 94403, "epoch": 1060, "lr": 7.317781612769262e-05}
{"train_loss": 0.14747002720832825, "global_step": 94404, "epoch": 1060, "lr": 7.31773024403976e-05}
{"train_loss": 0.17716485261917114, "global_step": 94405, "epoch": 1060, "lr": 7.317678874998668e-05}
{"train_loss": 0.15130819380283356, "global_step": 94406, "epoch": 1060, "lr": 7.317627505645991e-05}
{"train_loss": 0.17764787375926971, "global_step": 94407, "epoch": 1060, "lr": 7.31757613598174e-05}
{"train_loss": 0.12759223580360413, "global_step": 94408, "epoch": 1060, "lr": 7.317524766005917e-05}
{"train_loss": 0.16802600026130676, "global_step": 94409, "epoch": 1060, "lr": 7.317473395718533e-05}
{"train_loss": 0.12737469375133514, "global_step": 94410, "epoch": 1060, "lr": 7.317422025119592e-05}
{"train_loss": 0.16578885912895203, "global_step": 94411, "epoch": 1060, "lr": 7.317370654209101e-05}
{"train_loss": 0.1098000705242157, "global_step": 94412, "epoch": 1060, "lr": 7.317319282987068e-05}
{"train_loss": 0.12853635847568512, "global_step": 94413, "epoch": 1060, "lr": 7.3172679114535e-05}
{"train_loss": 0.16160674393177032, "global_step": 94414, "epoch": 1060, "lr": 7.317216539608405e-05}
{"train_loss": 0.1275411993265152, "global_step": 94415, "epoch": 1060, "lr": 7.317165167451786e-05}
{"train_loss": 0.17587675154209137, "global_step": 94416, "epoch": 1060, "lr": 7.317113794983654e-05}
{"train_loss": 0.2604290246963501, "global_step": 94417, "epoch": 1060, "lr": 7.317062422204014e-05}
{"train_loss": 0.1867896020412445, "global_step": 94418, "epoch": 1060, "lr": 7.317011049112873e-05}
{"train_loss": 0.14102225005626678, "global_step": 94419, "epoch": 1060, "lr": 7.316959675710238e-05}
{"train_loss": 0.15329094231128693, "global_step": 94420, "epoch": 1060, "lr": 7.316908301996116e-05}
{"train_loss": 0.14776454865932465, "global_step": 94421, "epoch": 1060, "lr": 7.316856927970515e-05}
{"train_loss": 0.18810983002185822, "global_step": 94422, "epoch": 1060, "lr": 7.316805553633439e-05}
{"train_loss": 0.16631996631622314, "global_step": 94423, "epoch": 1060, "lr": 7.316754178984897e-05}
{"train_loss": 0.19723427295684814, "global_step": 94424, "epoch": 1060, "lr": 7.316702804024897e-05}
{"train_loss": 0.14475484192371368, "global_step": 94425, "epoch": 1060, "lr": 7.316651428753444e-05}
{"train_loss": 0.2017163187265396, "global_step": 94426, "epoch": 1060, "lr": 7.316600053170544e-05}
{"train_loss": 0.2014278620481491, "global_step": 94427, "epoch": 1060, "lr": 7.316548677276207e-05}
{"train_loss": 0.16865290081902837, "global_step": 94428, "epoch": 1060, "lr": 7.316497301070439e-05, "val_loss": 4.738714694976807, "train_action_mse_error": 40.21590042114258}
{"train_loss": 0.26690641045570374, "global_step": 94429, "epoch": 1061, "lr": 7.316445924553244e-05}
{"train_loss": 0.25578951835632324, "global_step": 94430, "epoch": 1061, "lr": 7.316394547724632e-05}
{"train_loss": 0.23887309432029724, "global_step": 94431, "epoch": 1061, "lr": 7.31634317058461e-05}
{"train_loss": 0.16650231182575226, "global_step": 94432, "epoch": 1061, "lr": 7.316291793133182e-05}
{"train_loss": 0.14116056263446808, "global_step": 94433, "epoch": 1061, "lr": 7.316240415370357e-05}
{"train_loss": 0.13179734349250793, "global_step": 94434, "epoch": 1061, "lr": 7.316189037296144e-05}
{"train_loss": 0.1290917992591858, "global_step": 94435, "epoch": 1061, "lr": 7.316137658910545e-05}
{"train_loss": 0.1543615609407425, "global_step": 94436, "epoch": 1061, "lr": 7.316086280213571e-05}
{"train_loss": 0.20736399292945862, "global_step": 94437, "epoch": 1061, "lr": 7.316034901205227e-05}
{"train_loss": 0.17833833396434784, "global_step": 94438, "epoch": 1061, "lr": 7.315983521885521e-05}
{"train_loss": 0.19678343832492828, "global_step": 94439, "epoch": 1061, "lr": 7.315932142254458e-05}
{"train_loss": 0.16986000537872314, "global_step": 94440, "epoch": 1061, "lr": 7.315880762312047e-05}
{"train_loss": 0.25402912497520447, "global_step": 94441, "epoch": 1061, "lr": 7.315829382058294e-05}
{"train_loss": 0.14212366938591003, "global_step": 94442, "epoch": 1061, "lr": 7.315778001493206e-05}
{"train_loss": 0.18501144647598267, "global_step": 94443, "epoch": 1061, "lr": 7.31572662061679e-05}
{"train_loss": 0.21022911369800568, "global_step": 94444, "epoch": 1061, "lr": 7.315675239429052e-05}
{"train_loss": 0.15578381717205048, "global_step": 94445, "epoch": 1061, "lr": 7.315623857930001e-05}
{"train_loss": 0.19360403716564178, "global_step": 94446, "epoch": 1061, "lr": 7.315572476119641e-05}
{"train_loss": 0.17827312648296356, "global_step": 94447, "epoch": 1061, "lr": 7.315521093997981e-05}
{"train_loss": 0.1782718449831009, "global_step": 94448, "epoch": 1061, "lr": 7.315469711565028e-05}
{"train_loss": 0.24654440581798553, "global_step": 94449, "epoch": 1061, "lr": 7.315418328820788e-05}
{"train_loss": 0.1271568238735199, "global_step": 94450, "epoch": 1061, "lr": 7.315366945765268e-05}
{"train_loss": 0.1517154723405838, "global_step": 94451, "epoch": 1061, "lr": 7.315315562398475e-05}
{"train_loss": 0.1298980414867401, "global_step": 94452, "epoch": 1061, "lr": 7.315264178720417e-05}
{"train_loss": 0.16840286552906036, "global_step": 94453, "epoch": 1061, "lr": 7.315212794731098e-05}
{"train_loss": 0.20051686465740204, "global_step": 94454, "epoch": 1061, "lr": 7.31516141043053e-05}
{"train_loss": 0.12477275729179382, "global_step": 94455, "epoch": 1061, "lr": 7.315110025818715e-05}
{"train_loss": 0.16191616654396057, "global_step": 94456, "epoch": 1061, "lr": 7.31505864089566e-05}
{"train_loss": 0.23599597811698914, "global_step": 94457, "epoch": 1061, "lr": 7.315007255661377e-05}
{"train_loss": 0.1373300850391388, "global_step": 94458, "epoch": 1061, "lr": 7.314955870115867e-05}
{"train_loss": 0.19724036753177643, "global_step": 94459, "epoch": 1061, "lr": 7.314904484259139e-05}
{"train_loss": 0.14166444540023804, "global_step": 94460, "epoch": 1061, "lr": 7.314853098091203e-05}
{"train_loss": 0.19644397497177124, "global_step": 94461, "epoch": 1061, "lr": 7.31480171161206e-05}
{"train_loss": 0.13151668012142181, "global_step": 94462, "epoch": 1061, "lr": 7.314750324821724e-05}
{"train_loss": 0.1547842025756836, "global_step": 94463, "epoch": 1061, "lr": 7.314698937720194e-05}
{"train_loss": 0.17233610153198242, "global_step": 94464, "epoch": 1061, "lr": 7.314647550307483e-05}
{"train_loss": 0.20318029820919037, "global_step": 94465, "epoch": 1061, "lr": 7.314596162583596e-05}
{"train_loss": 0.17523613572120667, "global_step": 94466, "epoch": 1061, "lr": 7.31454477454854e-05}
{"train_loss": 0.1358114629983902, "global_step": 94467, "epoch": 1061, "lr": 7.31449338620232e-05}
{"train_loss": 0.1936745047569275, "global_step": 94468, "epoch": 1061, "lr": 7.314441997544945e-05}
{"train_loss": 0.21675080060958862, "global_step": 94469, "epoch": 1061, "lr": 7.314390608576423e-05}
{"train_loss": 0.16175992786884308, "global_step": 94470, "epoch": 1061, "lr": 7.314339219296758e-05}
{"train_loss": 0.13373102247714996, "global_step": 94471, "epoch": 1061, "lr": 7.314287829705959e-05}
{"train_loss": 0.21238696575164795, "global_step": 94472, "epoch": 1061, "lr": 7.31423643980403e-05}
{"train_loss": 0.13764964044094086, "global_step": 94473, "epoch": 1061, "lr": 7.314185049590982e-05}
{"train_loss": 0.10421393811702728, "global_step": 94474, "epoch": 1061, "lr": 7.314133659066822e-05}
{"train_loss": 0.15029509365558624, "global_step": 94475, "epoch": 1061, "lr": 7.314082268231554e-05}
{"train_loss": 0.21533040702342987, "global_step": 94476, "epoch": 1061, "lr": 7.314030877085184e-05}
{"train_loss": 0.16270627081394196, "global_step": 94477, "epoch": 1061, "lr": 7.313979485627723e-05}
{"train_loss": 0.24680736660957336, "global_step": 94478, "epoch": 1061, "lr": 7.313928093859173e-05}
{"train_loss": 0.14787015318870544, "global_step": 94479, "epoch": 1061, "lr": 7.313876701779545e-05}
{"train_loss": 0.15843364596366882, "global_step": 94480, "epoch": 1061, "lr": 7.313825309388845e-05}
{"train_loss": 0.1497449427843094, "global_step": 94481, "epoch": 1061, "lr": 7.313773916687078e-05}
{"train_loss": 0.18711715936660767, "global_step": 94482, "epoch": 1061, "lr": 7.313722523674253e-05}
{"train_loss": 0.2122918963432312, "global_step": 94483, "epoch": 1061, "lr": 7.313671130350375e-05}
{"train_loss": 0.21939492225646973, "global_step": 94484, "epoch": 1061, "lr": 7.313619736715455e-05}
{"train_loss": 0.1582449972629547, "global_step": 94485, "epoch": 1061, "lr": 7.313568342769494e-05}
{"train_loss": 0.18565070629119873, "global_step": 94486, "epoch": 1061, "lr": 7.313516948512504e-05}
{"train_loss": 0.20468783378601074, "global_step": 94487, "epoch": 1061, "lr": 7.313465553944489e-05}
{"train_loss": 0.17496468126773834, "global_step": 94488, "epoch": 1061, "lr": 7.313414159065457e-05}
{"train_loss": 0.15810495615005493, "global_step": 94489, "epoch": 1061, "lr": 7.313362763875415e-05}
{"train_loss": 0.16538693010807037, "global_step": 94490, "epoch": 1061, "lr": 7.313311368374369e-05}
{"train_loss": 0.2446667104959488, "global_step": 94491, "epoch": 1061, "lr": 7.313259972562325e-05}
{"train_loss": 0.13460151851177216, "global_step": 94492, "epoch": 1061, "lr": 7.313208576439293e-05}
{"train_loss": 0.11012927442789078, "global_step": 94493, "epoch": 1061, "lr": 7.31315718000528e-05}
{"train_loss": 0.17000652849674225, "global_step": 94494, "epoch": 1061, "lr": 7.31310578326029e-05}
{"train_loss": 0.19469675421714783, "global_step": 94495, "epoch": 1061, "lr": 7.31305438620433e-05}
{"train_loss": 0.14828535914421082, "global_step": 94496, "epoch": 1061, "lr": 7.313002988837408e-05}
{"train_loss": 0.11949662119150162, "global_step": 94497, "epoch": 1061, "lr": 7.312951591159533e-05}
{"train_loss": 0.24507999420166016, "global_step": 94498, "epoch": 1061, "lr": 7.312900193170707e-05}
{"train_loss": 0.1701584905385971, "global_step": 94499, "epoch": 1061, "lr": 7.312848794870943e-05}
{"train_loss": 0.2711848020553589, "global_step": 94500, "epoch": 1061, "lr": 7.312797396260243e-05}
{"train_loss": 0.16117103397846222, "global_step": 94501, "epoch": 1061, "lr": 7.312745997338614e-05}
{"train_loss": 0.2410198152065277, "global_step": 94502, "epoch": 1061, "lr": 7.312694598106067e-05}
{"train_loss": 0.13332898914813995, "global_step": 94503, "epoch": 1061, "lr": 7.312643198562605e-05}
{"train_loss": 0.15002292394638062, "global_step": 94504, "epoch": 1061, "lr": 7.312591798708237e-05}
{"train_loss": 0.12141741812229156, "global_step": 94505, "epoch": 1061, "lr": 7.312540398542969e-05}
{"train_loss": 0.2564733922481537, "global_step": 94506, "epoch": 1061, "lr": 7.312488998066807e-05}
{"train_loss": 0.141132190823555, "global_step": 94507, "epoch": 1061, "lr": 7.31243759727976e-05}
{"train_loss": 0.15609461069107056, "global_step": 94508, "epoch": 1061, "lr": 7.312386196181833e-05}
{"train_loss": 0.2104169875383377, "global_step": 94509, "epoch": 1061, "lr": 7.312334794773035e-05}
{"train_loss": 0.12706416845321655, "global_step": 94510, "epoch": 1061, "lr": 7.312283393053371e-05}
{"train_loss": 0.10426948219537735, "global_step": 94511, "epoch": 1061, "lr": 7.31223199102285e-05}
{"train_loss": 0.15743136405944824, "global_step": 94512, "epoch": 1061, "lr": 7.312180588681477e-05}
{"train_loss": 0.13009530305862427, "global_step": 94513, "epoch": 1061, "lr": 7.31212918602926e-05}
{"train_loss": 0.17235395312309265, "global_step": 94514, "epoch": 1061, "lr": 7.312077783066205e-05}
{"train_loss": 0.21307411789894104, "global_step": 94515, "epoch": 1061, "lr": 7.31202637979232e-05}
{"train_loss": 0.18483887612819672, "global_step": 94516, "epoch": 1061, "lr": 7.311974976207609e-05}
{"train_loss": 0.175261500594991, "global_step": 94517, "epoch": 1061, "lr": 7.311923572312083e-05, "val_loss": 4.4941606521606445}
{"train_loss": 0.15695683658123016, "global_step": 94518, "epoch": 1062, "lr": 7.311872168105747e-05}
{"train_loss": 0.14115922152996063, "global_step": 94519, "epoch": 1062, "lr": 7.311820763588606e-05}
{"train_loss": 0.11025050282478333, "global_step": 94520, "epoch": 1062, "lr": 7.311769358760671e-05}
{"train_loss": 0.2023886889219284, "global_step": 94521, "epoch": 1062, "lr": 7.311717953621947e-05}
{"train_loss": 0.14585737884044647, "global_step": 94522, "epoch": 1062, "lr": 7.31166654817244e-05}
{"train_loss": 0.17761485278606415, "global_step": 94523, "epoch": 1062, "lr": 7.311615142412158e-05}
{"train_loss": 0.21773917973041534, "global_step": 94524, "epoch": 1062, "lr": 7.311563736341107e-05}
{"train_loss": 0.1995803862810135, "global_step": 94525, "epoch": 1062, "lr": 7.311512329959294e-05}
{"train_loss": 0.15641100704669952, "global_step": 94526, "epoch": 1062, "lr": 7.311460923266726e-05}
{"train_loss": 0.0937117263674736, "global_step": 94527, "epoch": 1062, "lr": 7.311409516263411e-05}
{"train_loss": 0.1583782434463501, "global_step": 94528, "epoch": 1062, "lr": 7.311358108949356e-05}
{"train_loss": 0.17119954526424408, "global_step": 94529, "epoch": 1062, "lr": 7.311306701324565e-05}
{"train_loss": 0.1455531120300293, "global_step": 94530, "epoch": 1062, "lr": 7.311255293389049e-05}
{"train_loss": 0.23738588392734528, "global_step": 94531, "epoch": 1062, "lr": 7.311203885142812e-05}
{"train_loss": 0.1894659698009491, "global_step": 94532, "epoch": 1062, "lr": 7.311152476585863e-05}
{"train_loss": 0.25027498602867126, "global_step": 94533, "epoch": 1062, "lr": 7.311101067718206e-05}
{"train_loss": 0.20572571456432343, "global_step": 94534, "epoch": 1062, "lr": 7.31104965853985e-05}
{"train_loss": 0.16018730401992798, "global_step": 94535, "epoch": 1062, "lr": 7.310998249050803e-05}
{"train_loss": 0.14052297174930573, "global_step": 94536, "epoch": 1062, "lr": 7.31094683925107e-05}
{"train_loss": 0.17539364099502563, "global_step": 94537, "epoch": 1062, "lr": 7.310895429140656e-05}
{"train_loss": 0.16782328486442566, "global_step": 94538, "epoch": 1062, "lr": 7.310844018719573e-05}
{"train_loss": 0.1862327605485916, "global_step": 94539, "epoch": 1062, "lr": 7.310792607987825e-05}
{"train_loss": 0.10505081713199615, "global_step": 94540, "epoch": 1062, "lr": 7.310741196945416e-05}
{"train_loss": 0.1718091368675232, "global_step": 94541, "epoch": 1062, "lr": 7.310689785592359e-05}
{"train_loss": 0.14193052053451538, "global_step": 94542, "epoch": 1062, "lr": 7.310638373928658e-05}
{"train_loss": 0.18736228346824646, "global_step": 94543, "epoch": 1062, "lr": 7.310586961954319e-05}
{"train_loss": 0.22118814289569855, "global_step": 94544, "epoch": 1062, "lr": 7.31053554966935e-05}
{"train_loss": 0.19811098277568817, "global_step": 94545, "epoch": 1062, "lr": 7.310484137073755e-05}
{"train_loss": 0.15006738901138306, "global_step": 94546, "epoch": 1062, "lr": 7.310432724167546e-05}
{"train_loss": 0.37807318568229675, "global_step": 94547, "epoch": 1062, "lr": 7.310381310950728e-05}
{"train_loss": 0.2049538493156433, "global_step": 94548, "epoch": 1062, "lr": 7.310329897423307e-05}
{"train_loss": 0.18115699291229248, "global_step": 94549, "epoch": 1062, "lr": 7.31027848358529e-05}
{"train_loss": 0.10751467198133469, "global_step": 94550, "epoch": 1062, "lr": 7.310227069436684e-05}
{"train_loss": 0.15057438611984253, "global_step": 94551, "epoch": 1062, "lr": 7.310175654977498e-05}
{"train_loss": 0.2696530818939209, "global_step": 94552, "epoch": 1062, "lr": 7.310124240207735e-05}
{"train_loss": 0.11648356914520264, "global_step": 94553, "epoch": 1062, "lr": 7.310072825127405e-05}
{"train_loss": 0.17018763720989227, "global_step": 94554, "epoch": 1062, "lr": 7.310021409736514e-05}
{"train_loss": 0.22093315422534943, "global_step": 94555, "epoch": 1062, "lr": 7.309969994035067e-05}
{"train_loss": 0.13878285884857178, "global_step": 94556, "epoch": 1062, "lr": 7.309918578023074e-05}
{"train_loss": 0.18502643704414368, "global_step": 94557, "epoch": 1062, "lr": 7.309867161700542e-05}
{"train_loss": 0.15373875200748444, "global_step": 94558, "epoch": 1062, "lr": 7.309815745067474e-05}
{"train_loss": 0.13642661273479462, "global_step": 94559, "epoch": 1062, "lr": 7.30976432812388e-05}
{"train_loss": 0.11943324655294418, "global_step": 94560, "epoch": 1062, "lr": 7.309712910869766e-05}
{"train_loss": 0.17819547653198242, "global_step": 94561, "epoch": 1062, "lr": 7.30966149330514e-05}
{"train_loss": 0.14108414947986603, "global_step": 94562, "epoch": 1062, "lr": 7.309610075430009e-05}
{"train_loss": 0.102345772087574, "global_step": 94563, "epoch": 1062, "lr": 7.309558657244378e-05}
{"train_loss": 0.19757027924060822, "global_step": 94564, "epoch": 1062, "lr": 7.309507238748254e-05}
{"train_loss": 0.14569908380508423, "global_step": 94565, "epoch": 1062, "lr": 7.309455819941646e-05}
{"train_loss": 0.15550854802131653, "global_step": 94566, "epoch": 1062, "lr": 7.30940440082456e-05}
{"train_loss": 0.1443856656551361, "global_step": 94567, "epoch": 1062, "lr": 7.309352981397003e-05}
{"train_loss": 0.11355315893888474, "global_step": 94568, "epoch": 1062, "lr": 7.309301561658981e-05}
{"train_loss": 0.1348412185907364, "global_step": 94569, "epoch": 1062, "lr": 7.309250141610502e-05}
{"train_loss": 0.13848544657230377, "global_step": 94570, "epoch": 1062, "lr": 7.309198721251572e-05}
{"train_loss": 0.2118493616580963, "global_step": 94571, "epoch": 1062, "lr": 7.309147300582199e-05}
{"train_loss": 0.22604455053806305, "global_step": 94572, "epoch": 1062, "lr": 7.30909587960239e-05}
{"train_loss": 0.16504491865634918, "global_step": 94573, "epoch": 1062, "lr": 7.30904445831215e-05}
{"train_loss": 0.14850135147571564, "global_step": 94574, "epoch": 1062, "lr": 7.308993036711487e-05}
{"train_loss": 0.13282440602779388, "global_step": 94575, "epoch": 1062, "lr": 7.308941614800408e-05}
{"train_loss": 0.15718595683574677, "global_step": 94576, "epoch": 1062, "lr": 7.308890192578921e-05}
{"train_loss": 0.2485714852809906, "global_step": 94577, "epoch": 1062, "lr": 7.308838770047031e-05}
{"train_loss": 0.14402607083320618, "global_step": 94578, "epoch": 1062, "lr": 7.308787347204746e-05}
{"train_loss": 0.10704539716243744, "global_step": 94579, "epoch": 1062, "lr": 7.308735924052073e-05}
{"train_loss": 0.17688974738121033, "global_step": 94580, "epoch": 1062, "lr": 7.308684500589017e-05}
{"train_loss": 0.15527592599391937, "global_step": 94581, "epoch": 1062, "lr": 7.308633076815586e-05}
{"train_loss": 0.15194296836853027, "global_step": 94582, "epoch": 1062, "lr": 7.30858165273179e-05}
{"train_loss": 0.1469547599554062, "global_step": 94583, "epoch": 1062, "lr": 7.308530228337632e-05}
{"train_loss": 0.22964543104171753, "global_step": 94584, "epoch": 1062, "lr": 7.308478803633121e-05}
{"train_loss": 0.14283332228660583, "global_step": 94585, "epoch": 1062, "lr": 7.308427378618262e-05}
{"train_loss": 0.19563955068588257, "global_step": 94586, "epoch": 1062, "lr": 7.308375953293064e-05}
{"train_loss": 0.14788782596588135, "global_step": 94587, "epoch": 1062, "lr": 7.308324527657533e-05}
{"train_loss": 0.21507026255130768, "global_step": 94588, "epoch": 1062, "lr": 7.308273101711676e-05}
{"train_loss": 0.2935420274734497, "global_step": 94589, "epoch": 1062, "lr": 7.3082216754555e-05}
{"train_loss": 0.10594002157449722, "global_step": 94590, "epoch": 1062, "lr": 7.308170248889012e-05}
{"train_loss": 0.12825101613998413, "global_step": 94591, "epoch": 1062, "lr": 7.308118822012217e-05}
{"train_loss": 0.16492711007595062, "global_step": 94592, "epoch": 1062, "lr": 7.308067394825125e-05}
{"train_loss": 0.11014582961797714, "global_step": 94593, "epoch": 1062, "lr": 7.308015967327742e-05}
{"train_loss": 0.16996267437934875, "global_step": 94594, "epoch": 1062, "lr": 7.307964539520073e-05}
{"train_loss": 0.14949139952659607, "global_step": 94595, "epoch": 1062, "lr": 7.307913111402126e-05}
{"train_loss": 0.13539758324623108, "global_step": 94596, "epoch": 1062, "lr": 7.30786168297391e-05}
{"train_loss": 0.163652703166008, "global_step": 94597, "epoch": 1062, "lr": 7.307810254235428e-05}
{"train_loss": 0.30129683017730713, "global_step": 94598, "epoch": 1062, "lr": 7.30775882518669e-05}
{"train_loss": 0.17908310890197754, "global_step": 94599, "epoch": 1062, "lr": 7.307707395827701e-05}
{"train_loss": 0.11982899159193039, "global_step": 94600, "epoch": 1062, "lr": 7.30765596615847e-05}
{"train_loss": 0.19095191359519958, "global_step": 94601, "epoch": 1062, "lr": 7.307604536179002e-05}
{"train_loss": 0.11700049787759781, "global_step": 94602, "epoch": 1062, "lr": 7.307553105889307e-05}
{"train_loss": 0.09596406668424606, "global_step": 94603, "epoch": 1062, "lr": 7.307501675289387e-05}
{"train_loss": 0.21642929315567017, "global_step": 94604, "epoch": 1062, "lr": 7.307450244379253e-05}
{"train_loss": 0.1313389092683792, "global_step": 94605, "epoch": 1062, "lr": 7.30739881315891e-05}
{"train_loss": 0.16933117304625137, "global_step": 94606, "epoch": 1062, "lr": 7.307347381628365e-05, "val_loss": 4.710961818695068}
{"train_loss": 0.1902172714471817, "global_step": 94607, "epoch": 1063, "lr": 7.307295949787625e-05}
{"train_loss": 0.19296413660049438, "global_step": 94608, "epoch": 1063, "lr": 7.307244517636698e-05}
{"train_loss": 0.14891840517520905, "global_step": 94609, "epoch": 1063, "lr": 7.307193085175591e-05}
{"train_loss": 0.10820760577917099, "global_step": 94610, "epoch": 1063, "lr": 7.307141652404309e-05}
{"train_loss": 0.2408260703086853, "global_step": 94611, "epoch": 1063, "lr": 7.30709021932286e-05}
{"train_loss": 0.13341470062732697, "global_step": 94612, "epoch": 1063, "lr": 7.30703878593125e-05}
{"train_loss": 0.20472411811351776, "global_step": 94613, "epoch": 1063, "lr": 7.306987352229488e-05}
{"train_loss": 0.12523114681243896, "global_step": 94614, "epoch": 1063, "lr": 7.306935918217578e-05}
{"train_loss": 0.18804031610488892, "global_step": 94615, "epoch": 1063, "lr": 7.30688448389553e-05}
{"train_loss": 0.219755157828331, "global_step": 94616, "epoch": 1063, "lr": 7.306833049263349e-05}
{"train_loss": 0.14604991674423218, "global_step": 94617, "epoch": 1063, "lr": 7.306781614321041e-05}
{"train_loss": 0.14474137127399445, "global_step": 94618, "epoch": 1063, "lr": 7.306730179068616e-05}
{"train_loss": 0.14997892081737518, "global_step": 94619, "epoch": 1063, "lr": 7.30667874350608e-05}
{"train_loss": 0.154412180185318, "global_step": 94620, "epoch": 1063, "lr": 7.306627307633438e-05}
{"train_loss": 0.13107429444789886, "global_step": 94621, "epoch": 1063, "lr": 7.306575871450698e-05}
{"train_loss": 0.0775516927242279, "global_step": 94622, "epoch": 1063, "lr": 7.306524434957866e-05}
{"train_loss": 0.19775137305259705, "global_step": 94623, "epoch": 1063, "lr": 7.306472998154952e-05}
{"train_loss": 0.1833621710538864, "global_step": 94624, "epoch": 1063, "lr": 7.30642156104196e-05}
{"train_loss": 0.2071394920349121, "global_step": 94625, "epoch": 1063, "lr": 7.306370123618898e-05}
{"train_loss": 0.19163084030151367, "global_step": 94626, "epoch": 1063, "lr": 7.306318685885772e-05}
{"train_loss": 0.15519750118255615, "global_step": 94627, "epoch": 1063, "lr": 7.30626724784259e-05}
{"train_loss": 0.13105303049087524, "global_step": 94628, "epoch": 1063, "lr": 7.306215809489358e-05}
{"train_loss": 0.18033938109874725, "global_step": 94629, "epoch": 1063, "lr": 7.306164370826085e-05}
{"train_loss": 0.2367205172777176, "global_step": 94630, "epoch": 1063, "lr": 7.306112931852775e-05}
{"train_loss": 0.1767757534980774, "global_step": 94631, "epoch": 1063, "lr": 7.306061492569437e-05}
{"train_loss": 0.12723734974861145, "global_step": 94632, "epoch": 1063, "lr": 7.306010052976077e-05}
{"train_loss": 0.12708507478237152, "global_step": 94633, "epoch": 1063, "lr": 7.305958613072701e-05}
{"train_loss": 0.1444714218378067, "global_step": 94634, "epoch": 1063, "lr": 7.305907172859318e-05}
{"train_loss": 0.21032176911830902, "global_step": 94635, "epoch": 1063, "lr": 7.305855732335933e-05}
{"train_loss": 0.14898350834846497, "global_step": 94636, "epoch": 1063, "lr": 7.305804291502554e-05}
{"train_loss": 0.18214042484760284, "global_step": 94637, "epoch": 1063, "lr": 7.305752850359187e-05}
{"train_loss": 0.15769648551940918, "global_step": 94638, "epoch": 1063, "lr": 7.305701408905841e-05}
{"train_loss": 0.21001984179019928, "global_step": 94639, "epoch": 1063, "lr": 7.305649967142521e-05}
{"train_loss": 0.1644044816493988, "global_step": 94640, "epoch": 1063, "lr": 7.305598525069235e-05}
{"train_loss": 0.17562653124332428, "global_step": 94641, "epoch": 1063, "lr": 7.305547082685988e-05}
{"train_loss": 0.2029339224100113, "global_step": 94642, "epoch": 1063, "lr": 7.30549563999279e-05}
{"train_loss": 0.24986697733402252, "global_step": 94643, "epoch": 1063, "lr": 7.305444196989645e-05}
{"train_loss": 0.07616666704416275, "global_step": 94644, "epoch": 1063, "lr": 7.305392753676561e-05}
{"train_loss": 0.18668895959854126, "global_step": 94645, "epoch": 1063, "lr": 7.305341310053546e-05}
{"train_loss": 0.1799764335155487, "global_step": 94646, "epoch": 1063, "lr": 7.305289866120606e-05}
{"train_loss": 0.1560140699148178, "global_step": 94647, "epoch": 1063, "lr": 7.305238421877747e-05}
{"train_loss": 0.1396487057209015, "global_step": 94648, "epoch": 1063, "lr": 7.305186977324978e-05}
{"train_loss": 0.16438056528568268, "global_step": 94649, "epoch": 1063, "lr": 7.305135532462304e-05}
{"train_loss": 0.13008807599544525, "global_step": 94650, "epoch": 1063, "lr": 7.305084087289733e-05}
{"train_loss": 0.19620001316070557, "global_step": 94651, "epoch": 1063, "lr": 7.30503264180727e-05}
{"train_loss": 0.22274348139762878, "global_step": 94652, "epoch": 1063, "lr": 7.304981196014926e-05}
{"train_loss": 0.11652177572250366, "global_step": 94653, "epoch": 1063, "lr": 7.304929749912703e-05}
{"train_loss": 0.16718418896198273, "global_step": 94654, "epoch": 1063, "lr": 7.304878303500611e-05}
{"train_loss": 0.14939641952514648, "global_step": 94655, "epoch": 1063, "lr": 7.304826856778656e-05}
{"train_loss": 0.13030925393104553, "global_step": 94656, "epoch": 1063, "lr": 7.304775409746846e-05}
{"train_loss": 0.18211525678634644, "global_step": 94657, "epoch": 1063, "lr": 7.304723962405186e-05}
{"train_loss": 0.2840671241283417, "global_step": 94658, "epoch": 1063, "lr": 7.304672514753686e-05}
{"train_loss": 0.13599921762943268, "global_step": 94659, "epoch": 1063, "lr": 7.304621066792349e-05}
{"train_loss": 0.15236718952655792, "global_step": 94660, "epoch": 1063, "lr": 7.304569618521183e-05}
{"train_loss": 0.18393318355083466, "global_step": 94661, "epoch": 1063, "lr": 7.304518169940197e-05}
{"train_loss": 0.16076785326004028, "global_step": 94662, "epoch": 1063, "lr": 7.304466721049397e-05}
{"train_loss": 0.2408827245235443, "global_step": 94663, "epoch": 1063, "lr": 7.30441527184879e-05}
{"train_loss": 0.2378651201725006, "global_step": 94664, "epoch": 1063, "lr": 7.304363822338381e-05}
{"train_loss": 0.24846693873405457, "global_step": 94665, "epoch": 1063, "lr": 7.304312372518178e-05}
{"train_loss": 0.13334114849567413, "global_step": 94666, "epoch": 1063, "lr": 7.30426092238819e-05}
{"train_loss": 0.17404484748840332, "global_step": 94667, "epoch": 1063, "lr": 7.304209471948422e-05}
{"train_loss": 0.24710384011268616, "global_step": 94668, "epoch": 1063, "lr": 7.30415802119888e-05}
{"train_loss": 0.2189936488866806, "global_step": 94669, "epoch": 1063, "lr": 7.304106570139573e-05}
{"train_loss": 0.09895943850278854, "global_step": 94670, "epoch": 1063, "lr": 7.304055118770506e-05}
{"train_loss": 0.1630600392818451, "global_step": 94671, "epoch": 1063, "lr": 7.304003667091688e-05}
{"train_loss": 0.16411620378494263, "global_step": 94672, "epoch": 1063, "lr": 7.303952215103122e-05}
{"train_loss": 0.17780685424804688, "global_step": 94673, "epoch": 1063, "lr": 7.303900762804821e-05}
{"train_loss": 0.14990784227848053, "global_step": 94674, "epoch": 1063, "lr": 7.303849310196787e-05}
{"train_loss": 0.16704097390174866, "global_step": 94675, "epoch": 1063, "lr": 7.303797857279029e-05}
{"train_loss": 0.20342083275318146, "global_step": 94676, "epoch": 1063, "lr": 7.303746404051554e-05}
{"train_loss": 0.1574351191520691, "global_step": 94677, "epoch": 1063, "lr": 7.303694950514368e-05}
{"train_loss": 0.2584962844848633, "global_step": 94678, "epoch": 1063, "lr": 7.303643496667478e-05}
{"train_loss": 0.1647099256515503, "global_step": 94679, "epoch": 1063, "lr": 7.303592042510891e-05}
{"train_loss": 0.25096720457077026, "global_step": 94680, "epoch": 1063, "lr": 7.303540588044615e-05}
{"train_loss": 0.1598050892353058, "global_step": 94681, "epoch": 1063, "lr": 7.303489133268655e-05}
{"train_loss": 0.11076489090919495, "global_step": 94682, "epoch": 1063, "lr": 7.303437678183019e-05}
{"train_loss": 0.2333056479692459, "global_step": 94683, "epoch": 1063, "lr": 7.303386222787715e-05}
{"train_loss": 0.18169526755809784, "global_step": 94684, "epoch": 1063, "lr": 7.303334767082747e-05}
{"train_loss": 0.17364992201328278, "global_step": 94685, "epoch": 1063, "lr": 7.303283311068126e-05}
{"train_loss": 0.30418065190315247, "global_step": 94686, "epoch": 1063, "lr": 7.303231854743856e-05}
{"train_loss": 0.15165072679519653, "global_step": 94687, "epoch": 1063, "lr": 7.303180398109944e-05}
{"train_loss": 0.14856357872486115, "global_step": 94688, "epoch": 1063, "lr": 7.303128941166399e-05}
{"train_loss": 0.17483161389827728, "global_step": 94689, "epoch": 1063, "lr": 7.303077483913224e-05}
{"train_loss": 0.10859508067369461, "global_step": 94690, "epoch": 1063, "lr": 7.303026026350428e-05}
{"train_loss": 0.1653212010860443, "global_step": 94691, "epoch": 1063, "lr": 7.302974568478021e-05}
{"train_loss": 0.16066581010818481, "global_step": 94692, "epoch": 1063, "lr": 7.302923110296005e-05}
{"train_loss": 0.1477973312139511, "global_step": 94693, "epoch": 1063, "lr": 7.302871651804389e-05}
{"train_loss": 0.15257664024829865, "global_step": 94694, "epoch": 1063, "lr": 7.30282019300318e-05}
{"train_loss": 0.17355115058716764, "global_step": 94695, "epoch": 1063, "lr": 7.302768733892387e-05, "val_loss": 4.431357383728027}
{"train_loss": 0.1761203110218048, "global_step": 94696, "epoch": 1064, "lr": 7.302717274472013e-05}
{"train_loss": 0.18515509366989136, "global_step": 94697, "epoch": 1064, "lr": 7.302665814742067e-05}
{"train_loss": 0.20696067810058594, "global_step": 94698, "epoch": 1064, "lr": 7.302614354702555e-05}
{"train_loss": 0.1746322363615036, "global_step": 94699, "epoch": 1064, "lr": 7.302562894353485e-05}
{"train_loss": 0.14072656631469727, "global_step": 94700, "epoch": 1064, "lr": 7.302511433694864e-05}
{"train_loss": 0.1807691901922226, "global_step": 94701, "epoch": 1064, "lr": 7.302459972726699e-05}
{"train_loss": 0.12816782295703888, "global_step": 94702, "epoch": 1064, "lr": 7.302408511448994e-05}
{"train_loss": 0.12382936477661133, "global_step": 94703, "epoch": 1064, "lr": 7.302357049861761e-05}
{"train_loss": 0.14771048724651337, "global_step": 94704, "epoch": 1064, "lr": 7.302305587965002e-05}
{"train_loss": 0.24102900922298431, "global_step": 94705, "epoch": 1064, "lr": 7.302254125758727e-05}
{"train_loss": 0.19886785745620728, "global_step": 94706, "epoch": 1064, "lr": 7.302202663242944e-05}
{"train_loss": 0.13415874540805817, "global_step": 94707, "epoch": 1064, "lr": 7.302151200417656e-05}
{"train_loss": 0.16945652663707733, "global_step": 94708, "epoch": 1064, "lr": 7.30209973728287e-05}
{"train_loss": 0.20200221240520477, "global_step": 94709, "epoch": 1064, "lr": 7.302048273838597e-05}
{"train_loss": 0.22593934834003448, "global_step": 94710, "epoch": 1064, "lr": 7.30199681008484e-05}
{"train_loss": 0.13298094272613525, "global_step": 94711, "epoch": 1064, "lr": 7.301945346021609e-05}
{"train_loss": 0.20406587421894073, "global_step": 94712, "epoch": 1064, "lr": 7.301893881648909e-05}
{"train_loss": 0.23023773729801178, "global_step": 94713, "epoch": 1064, "lr": 7.301842416966747e-05}
{"train_loss": 0.13808588683605194, "global_step": 94714, "epoch": 1064, "lr": 7.301790951975131e-05}
{"train_loss": 0.11556476354598999, "global_step": 94715, "epoch": 1064, "lr": 7.301739486674067e-05}
{"train_loss": 0.17020165920257568, "global_step": 94716, "epoch": 1064, "lr": 7.301688021063563e-05}
{"train_loss": 0.1247786432504654, "global_step": 94717, "epoch": 1064, "lr": 7.301636555143626e-05}
{"train_loss": 0.18775832653045654, "global_step": 94718, "epoch": 1064, "lr": 7.301585088914258e-05}
{"train_loss": 0.17222359776496887, "global_step": 94719, "epoch": 1064, "lr": 7.301533622375473e-05}
{"train_loss": 0.16752341389656067, "global_step": 94720, "epoch": 1064, "lr": 7.301482155527275e-05}
{"train_loss": 0.13609638810157776, "global_step": 94721, "epoch": 1064, "lr": 7.301430688369671e-05}
{"train_loss": 0.14267180860042572, "global_step": 94722, "epoch": 1064, "lr": 7.301379220902665e-05}
{"train_loss": 0.15147417783737183, "global_step": 94723, "epoch": 1064, "lr": 7.301327753126269e-05}
{"train_loss": 0.1258765310049057, "global_step": 94724, "epoch": 1064, "lr": 7.301276285040487e-05}
{"train_loss": 0.11628768593072891, "global_step": 94725, "epoch": 1064, "lr": 7.301224816645327e-05}
{"train_loss": 0.23547980189323425, "global_step": 94726, "epoch": 1064, "lr": 7.301173347940794e-05}
{"train_loss": 0.19732502102851868, "global_step": 94727, "epoch": 1064, "lr": 7.301121878926898e-05}
{"train_loss": 0.16168847680091858, "global_step": 94728, "epoch": 1064, "lr": 7.301070409603642e-05}
{"train_loss": 0.15401741862297058, "global_step": 94729, "epoch": 1064, "lr": 7.301018939971036e-05}
{"train_loss": 0.13858386874198914, "global_step": 94730, "epoch": 1064, "lr": 7.300967470029088e-05}
{"train_loss": 0.19496913254261017, "global_step": 94731, "epoch": 1064, "lr": 7.3009159997778e-05}
{"train_loss": 0.15996402502059937, "global_step": 94732, "epoch": 1064, "lr": 7.300864529217185e-05}
{"train_loss": 0.10645692050457001, "global_step": 94733, "epoch": 1064, "lr": 7.300813058347244e-05}
{"train_loss": 0.17081692814826965, "global_step": 94734, "epoch": 1064, "lr": 7.300761587167988e-05}
{"train_loss": 0.15452693402767181, "global_step": 94735, "epoch": 1064, "lr": 7.300710115679422e-05}
{"train_loss": 0.17059125006198883, "global_step": 94736, "epoch": 1064, "lr": 7.300658643881554e-05}
{"train_loss": 0.14054767787456512, "global_step": 94737, "epoch": 1064, "lr": 7.300607171774391e-05}
{"train_loss": 0.13674810528755188, "global_step": 94738, "epoch": 1064, "lr": 7.300555699357939e-05}
{"train_loss": 0.22108399868011475, "global_step": 94739, "epoch": 1064, "lr": 7.300504226632204e-05}
{"train_loss": 0.23545272648334503, "global_step": 94740, "epoch": 1064, "lr": 7.300452753597196e-05}
{"train_loss": 0.2074955254793167, "global_step": 94741, "epoch": 1064, "lr": 7.30040128025292e-05}
{"train_loss": 0.16337378323078156, "global_step": 94742, "epoch": 1064, "lr": 7.300349806599382e-05}
{"train_loss": 0.12054409831762314, "global_step": 94743, "epoch": 1064, "lr": 7.300298332636592e-05}
{"train_loss": 0.20572063326835632, "global_step": 94744, "epoch": 1064, "lr": 7.300246858364552e-05}
{"train_loss": 0.20073442161083221, "global_step": 94745, "epoch": 1064, "lr": 7.300195383783275e-05}
{"train_loss": 0.12428959459066391, "global_step": 94746, "epoch": 1064, "lr": 7.300143908892763e-05}
{"train_loss": 0.13421756029129028, "global_step": 94747, "epoch": 1064, "lr": 7.300092433693024e-05}
{"train_loss": 0.20264866948127747, "global_step": 94748, "epoch": 1064, "lr": 7.300040958184066e-05}
{"train_loss": 0.1455262303352356, "global_step": 94749, "epoch": 1064, "lr": 7.299989482365896e-05}
{"train_loss": 0.16544851660728455, "global_step": 94750, "epoch": 1064, "lr": 7.299938006238522e-05}
{"train_loss": 0.1138436421751976, "global_step": 94751, "epoch": 1064, "lr": 7.299886529801947e-05}
{"train_loss": 0.12660688161849976, "global_step": 94752, "epoch": 1064, "lr": 7.299835053056182e-05}
{"train_loss": 0.19558386504650116, "global_step": 94753, "epoch": 1064, "lr": 7.299783576001231e-05}
{"train_loss": 0.18745970726013184, "global_step": 94754, "epoch": 1064, "lr": 7.299732098637102e-05}
{"train_loss": 0.1625785529613495, "global_step": 94755, "epoch": 1064, "lr": 7.299680620963803e-05}
{"train_loss": 0.1063152626156807, "global_step": 94756, "epoch": 1064, "lr": 7.29962914298134e-05}
{"train_loss": 0.20359106361865997, "global_step": 94757, "epoch": 1064, "lr": 7.29957766468972e-05}
{"train_loss": 0.18345296382904053, "global_step": 94758, "epoch": 1064, "lr": 7.299526186088949e-05}
{"train_loss": 0.2290450483560562, "global_step": 94759, "epoch": 1064, "lr": 7.299474707179035e-05}
{"train_loss": 0.210903599858284, "global_step": 94760, "epoch": 1064, "lr": 7.299423227959984e-05}
{"train_loss": 0.1882355660200119, "global_step": 94761, "epoch": 1064, "lr": 7.299371748431805e-05}
{"train_loss": 0.11349663883447647, "global_step": 94762, "epoch": 1064, "lr": 7.299320268594501e-05}
{"train_loss": 0.1804114431142807, "global_step": 94763, "epoch": 1064, "lr": 7.299268788448083e-05}
{"train_loss": 0.18588592112064362, "global_step": 94764, "epoch": 1064, "lr": 7.299217307992558e-05}
{"train_loss": 0.14549121260643005, "global_step": 94765, "epoch": 1064, "lr": 7.299165827227929e-05}
{"train_loss": 0.13823939859867096, "global_step": 94766, "epoch": 1064, "lr": 7.299114346154205e-05}
{"train_loss": 0.1900896281003952, "global_step": 94767, "epoch": 1064, "lr": 7.299062864771395e-05}
{"train_loss": 0.2015211284160614, "global_step": 94768, "epoch": 1064, "lr": 7.299011383079504e-05}
{"train_loss": 0.21310600638389587, "global_step": 94769, "epoch": 1064, "lr": 7.298959901078537e-05}
{"train_loss": 0.14320547878742218, "global_step": 94770, "epoch": 1064, "lr": 7.298908418768504e-05}
{"train_loss": 0.19434061646461487, "global_step": 94771, "epoch": 1064, "lr": 7.298856936149411e-05}
{"train_loss": 0.14781907200813293, "global_step": 94772, "epoch": 1064, "lr": 7.298805453221263e-05}
{"train_loss": 0.13807103037834167, "global_step": 94773, "epoch": 1064, "lr": 7.298753969984071e-05}
{"train_loss": 0.23921118676662445, "global_step": 94774, "epoch": 1064, "lr": 7.298702486437839e-05}
{"train_loss": 0.1998654156923294, "global_step": 94775, "epoch": 1064, "lr": 7.298651002582573e-05}
{"train_loss": 0.1522693932056427, "global_step": 94776, "epoch": 1064, "lr": 7.298599518418282e-05}
{"train_loss": 0.19404254853725433, "global_step": 94777, "epoch": 1064, "lr": 7.298548033944973e-05}
{"train_loss": 0.18944337964057922, "global_step": 94778, "epoch": 1064, "lr": 7.298496549162653e-05}
{"train_loss": 0.2629460394382477, "global_step": 94779, "epoch": 1064, "lr": 7.298445064071327e-05}
{"train_loss": 0.13067132234573364, "global_step": 94780, "epoch": 1064, "lr": 7.298393578671002e-05}
{"train_loss": 0.20715929567813873, "global_step": 94781, "epoch": 1064, "lr": 7.298342092961688e-05}
{"train_loss": 0.22315393388271332, "global_step": 94782, "epoch": 1064, "lr": 7.298290606943388e-05}
{"train_loss": 0.37370428442955017, "global_step": 94783, "epoch": 1064, "lr": 7.298239120616112e-05}
{"train_loss": 0.17346492292505972, "global_step": 94784, "epoch": 1064, "lr": 7.298187633979865e-05, "val_loss": 4.511867046356201}
{"train_loss": 0.20462466776371002, "global_step": 94785, "epoch": 1065, "lr": 7.298136147034656e-05}
{"train_loss": 0.16824452579021454, "global_step": 94786, "epoch": 1065, "lr": 7.29808465978049e-05}
{"train_loss": 0.20952530205249786, "global_step": 94787, "epoch": 1065, "lr": 7.298033172217375e-05}
{"train_loss": 0.15816333889961243, "global_step": 94788, "epoch": 1065, "lr": 7.297981684345317e-05}
{"train_loss": 0.1652732938528061, "global_step": 94789, "epoch": 1065, "lr": 7.297930196164322e-05}
{"train_loss": 0.13281258940696716, "global_step": 94790, "epoch": 1065, "lr": 7.2978787076744e-05}
{"train_loss": 0.16841700673103333, "global_step": 94791, "epoch": 1065, "lr": 7.297827218875556e-05}
{"train_loss": 0.09158214181661606, "global_step": 94792, "epoch": 1065, "lr": 7.297775729767797e-05}
{"train_loss": 0.19541949033737183, "global_step": 94793, "epoch": 1065, "lr": 7.297724240351129e-05}
{"train_loss": 0.13987158238887787, "global_step": 94794, "epoch": 1065, "lr": 7.297672750625561e-05}
{"train_loss": 0.19251739978790283, "global_step": 94795, "epoch": 1065, "lr": 7.297621260591099e-05}
{"train_loss": 0.19456058740615845, "global_step": 94796, "epoch": 1065, "lr": 7.29756977024775e-05}
{"train_loss": 0.21291622519493103, "global_step": 94797, "epoch": 1065, "lr": 7.29751827959552e-05}
{"train_loss": 0.1215655729174614, "global_step": 94798, "epoch": 1065, "lr": 7.297466788634417e-05}
{"train_loss": 0.16955508291721344, "global_step": 94799, "epoch": 1065, "lr": 7.297415297364448e-05}
{"train_loss": 0.12581051886081696, "global_step": 94800, "epoch": 1065, "lr": 7.297363805785619e-05}
{"train_loss": 0.13650551438331604, "global_step": 94801, "epoch": 1065, "lr": 7.297312313897938e-05}
{"train_loss": 0.18901652097702026, "global_step": 94802, "epoch": 1065, "lr": 7.29726082170141e-05}
{"train_loss": 0.1840333640575409, "global_step": 94803, "epoch": 1065, "lr": 7.297209329196045e-05}
{"train_loss": 0.14470668137073517, "global_step": 94804, "epoch": 1065, "lr": 7.297157836381847e-05}
{"train_loss": 0.17719395458698273, "global_step": 94805, "epoch": 1065, "lr": 7.297106343258825e-05}
{"train_loss": 0.24151168763637543, "global_step": 94806, "epoch": 1065, "lr": 7.297054849826985e-05}
{"train_loss": 0.12399481981992722, "global_step": 94807, "epoch": 1065, "lr": 7.297003356086333e-05}
{"train_loss": 0.1326679289340973, "global_step": 94808, "epoch": 1065, "lr": 7.296951862036878e-05}
{"train_loss": 0.21652768552303314, "global_step": 94809, "epoch": 1065, "lr": 7.296900367678625e-05}
{"train_loss": 0.08952037245035172, "global_step": 94810, "epoch": 1065, "lr": 7.296848873011582e-05}
{"train_loss": 0.17688694596290588, "global_step": 94811, "epoch": 1065, "lr": 7.296797378035756e-05}
{"train_loss": 0.23792870342731476, "global_step": 94812, "epoch": 1065, "lr": 7.296745882751154e-05}
{"train_loss": 0.22040516138076782, "global_step": 94813, "epoch": 1065, "lr": 7.296694387157781e-05}
{"train_loss": 0.19014030694961548, "global_step": 94814, "epoch": 1065, "lr": 7.296642891255646e-05}
{"train_loss": 0.22420422732830048, "global_step": 94815, "epoch": 1065, "lr": 7.296591395044757e-05}
{"train_loss": 0.18937182426452637, "global_step": 94816, "epoch": 1065, "lr": 7.296539898525116e-05}
{"train_loss": 0.2113674432039261, "global_step": 94817, "epoch": 1065, "lr": 7.296488401696736e-05}
{"train_loss": 0.2207385152578354, "global_step": 94818, "epoch": 1065, "lr": 7.296436904559621e-05}
{"train_loss": 0.2719368040561676, "global_step": 94819, "epoch": 1065, "lr": 7.296385407113777e-05}
{"train_loss": 0.1885654628276825, "global_step": 94820, "epoch": 1065, "lr": 7.296333909359213e-05}
{"train_loss": 0.1552242934703827, "global_step": 94821, "epoch": 1065, "lr": 7.296282411295932e-05}
{"train_loss": 0.11287146806716919, "global_step": 94822, "epoch": 1065, "lr": 7.296230912923947e-05}
{"train_loss": 0.17795845866203308, "global_step": 94823, "epoch": 1065, "lr": 7.296179414243261e-05}
{"train_loss": 0.19597266614437103, "global_step": 94824, "epoch": 1065, "lr": 7.296127915253881e-05}
{"train_loss": 0.14701174199581146, "global_step": 94825, "epoch": 1065, "lr": 7.296076415955815e-05}
{"train_loss": 0.1689816266298294, "global_step": 94826, "epoch": 1065, "lr": 7.29602491634907e-05}
{"train_loss": 0.16215185821056366, "global_step": 94827, "epoch": 1065, "lr": 7.29597341643365e-05}
{"train_loss": 0.20048251748085022, "global_step": 94828, "epoch": 1065, "lr": 7.295921916209567e-05}
{"train_loss": 0.10075975954532623, "global_step": 94829, "epoch": 1065, "lr": 7.295870415676826e-05}
{"train_loss": 0.16141492128372192, "global_step": 94830, "epoch": 1065, "lr": 7.29581891483543e-05}
{"train_loss": 0.19026196002960205, "global_step": 94831, "epoch": 1065, "lr": 7.295767413685391e-05}
{"train_loss": 0.11973482370376587, "global_step": 94832, "epoch": 1065, "lr": 7.295715912226715e-05}
{"train_loss": 0.15318864583969116, "global_step": 94833, "epoch": 1065, "lr": 7.295664410459407e-05}
{"train_loss": 0.15914055705070496, "global_step": 94834, "epoch": 1065, "lr": 7.295612908383475e-05}
{"train_loss": 0.1373683661222458, "global_step": 94835, "epoch": 1065, "lr": 7.295561405998924e-05}
{"train_loss": 0.18631760776042938, "global_step": 94836, "epoch": 1065, "lr": 7.295509903305766e-05}
{"train_loss": 0.15416987240314484, "global_step": 94837, "epoch": 1065, "lr": 7.295458400304003e-05}
{"train_loss": 0.18214361369609833, "global_step": 94838, "epoch": 1065, "lr": 7.295406896993643e-05}
{"train_loss": 0.15563714504241943, "global_step": 94839, "epoch": 1065, "lr": 7.295355393374694e-05}
{"train_loss": 0.21423444151878357, "global_step": 94840, "epoch": 1065, "lr": 7.295303889447163e-05}
{"train_loss": 0.1671801060438156, "global_step": 94841, "epoch": 1065, "lr": 7.295252385211055e-05}
{"train_loss": 0.22699517011642456, "global_step": 94842, "epoch": 1065, "lr": 7.29520088066638e-05}
{"train_loss": 0.18042230606079102, "global_step": 94843, "epoch": 1065, "lr": 7.295149375813143e-05}
{"train_loss": 0.20750460028648376, "global_step": 94844, "epoch": 1065, "lr": 7.295097870651351e-05}
{"train_loss": 0.14176999032497406, "global_step": 94845, "epoch": 1065, "lr": 7.29504636518101e-05}
{"train_loss": 0.15687640011310577, "global_step": 94846, "epoch": 1065, "lr": 7.29499485940213e-05}
{"train_loss": 0.17619211971759796, "global_step": 94847, "epoch": 1065, "lr": 7.294943353314712e-05}
{"train_loss": 0.253873735666275, "global_step": 94848, "epoch": 1065, "lr": 7.29489184691877e-05}
{"train_loss": 0.18845227360725403, "global_step": 94849, "epoch": 1065, "lr": 7.294840340214308e-05}
{"train_loss": 0.21183085441589355, "global_step": 94850, "epoch": 1065, "lr": 7.294788833201331e-05}
{"train_loss": 0.11047776788473129, "global_step": 94851, "epoch": 1065, "lr": 7.29473732587985e-05}
{"train_loss": 0.16043353080749512, "global_step": 94852, "epoch": 1065, "lr": 7.294685818249867e-05}
{"train_loss": 0.15307877957820892, "global_step": 94853, "epoch": 1065, "lr": 7.294634310311392e-05}
{"train_loss": 0.1515195220708847, "global_step": 94854, "epoch": 1065, "lr": 7.294582802064432e-05}
{"train_loss": 0.2873329818248749, "global_step": 94855, "epoch": 1065, "lr": 7.294531293508992e-05}
{"train_loss": 0.21428583562374115, "global_step": 94856, "epoch": 1065, "lr": 7.294479784645082e-05}
{"train_loss": 0.09547184407711029, "global_step": 94857, "epoch": 1065, "lr": 7.294428275472707e-05}
{"train_loss": 0.20324550569057465, "global_step": 94858, "epoch": 1065, "lr": 7.294376765991874e-05}
{"train_loss": 0.13976328074932098, "global_step": 94859, "epoch": 1065, "lr": 7.29432525620259e-05}
{"train_loss": 0.19934767484664917, "global_step": 94860, "epoch": 1065, "lr": 7.294273746104862e-05}
{"train_loss": 0.13007009029388428, "global_step": 94861, "epoch": 1065, "lr": 7.294222235698696e-05}
{"train_loss": 0.15385866165161133, "global_step": 94862, "epoch": 1065, "lr": 7.2941707249841e-05}
{"train_loss": 0.1893412470817566, "global_step": 94863, "epoch": 1065, "lr": 7.294119213961081e-05}
{"train_loss": 0.17415332794189453, "global_step": 94864, "epoch": 1065, "lr": 7.294067702629645e-05}
{"train_loss": 0.1370808482170105, "global_step": 94865, "epoch": 1065, "lr": 7.294016190989801e-05}
{"train_loss": 0.13366536796092987, "global_step": 94866, "epoch": 1065, "lr": 7.293964679041554e-05}
{"train_loss": 0.11398223787546158, "global_step": 94867, "epoch": 1065, "lr": 7.293913166784911e-05}
{"train_loss": 0.10823925584554672, "global_step": 94868, "epoch": 1065, "lr": 7.293861654219879e-05}
{"train_loss": 0.24907156825065613, "global_step": 94869, "epoch": 1065, "lr": 7.293810141346466e-05}
{"train_loss": 0.12887875735759735, "global_step": 94870, "epoch": 1065, "lr": 7.293758628164678e-05}
{"train_loss": 0.1668878197669983, "global_step": 94871, "epoch": 1065, "lr": 7.293707114674519e-05}
{"train_loss": 0.1829894483089447, "global_step": 94872, "epoch": 1065, "lr": 7.293655600876002e-05}
{"train_loss": 0.17238024705916308, "global_step": 94873, "epoch": 1065, "lr": 7.293604086769132e-05, "val_loss": 4.378527641296387, "train_action_mse_error": 21.38100242614746}
{"train_loss": 0.12005085498094559, "global_step": 94874, "epoch": 1066, "lr": 7.293552572353913e-05}
{"train_loss": 0.22454942762851715, "global_step": 94875, "epoch": 1066, "lr": 7.293501057630356e-05}
{"train_loss": 0.09845353662967682, "global_step": 94876, "epoch": 1066, "lr": 7.293449542598465e-05}
{"train_loss": 0.15955357253551483, "global_step": 94877, "epoch": 1066, "lr": 7.293398027258247e-05}
{"train_loss": 0.2095097005367279, "global_step": 94878, "epoch": 1066, "lr": 7.293346511609709e-05}
{"train_loss": 0.19657455384731293, "global_step": 94879, "epoch": 1066, "lr": 7.29329499565286e-05}
{"train_loss": 0.14517788589000702, "global_step": 94880, "epoch": 1066, "lr": 7.293243479387704e-05}
{"train_loss": 0.11460448801517487, "global_step": 94881, "epoch": 1066, "lr": 7.29319196281425e-05}
{"train_loss": 0.13473886251449585, "global_step": 94882, "epoch": 1066, "lr": 7.293140445932504e-05}
{"train_loss": 0.1773429661989212, "global_step": 94883, "epoch": 1066, "lr": 7.293088928742473e-05}
{"train_loss": 0.20307661592960358, "global_step": 94884, "epoch": 1066, "lr": 7.293037411244165e-05}
{"train_loss": 0.12613780796527863, "global_step": 94885, "epoch": 1066, "lr": 7.292985893437585e-05}
{"train_loss": 0.17960751056671143, "global_step": 94886, "epoch": 1066, "lr": 7.292934375322741e-05}
{"train_loss": 0.21350999176502228, "global_step": 94887, "epoch": 1066, "lr": 7.29288285689964e-05}
{"train_loss": 0.20781047642230988, "global_step": 94888, "epoch": 1066, "lr": 7.292831338168291e-05}
{"train_loss": 0.2042606920003891, "global_step": 94889, "epoch": 1066, "lr": 7.292779819128697e-05}
{"train_loss": 0.15371814370155334, "global_step": 94890, "epoch": 1066, "lr": 7.292728299780866e-05}
{"train_loss": 0.16451047360897064, "global_step": 94891, "epoch": 1066, "lr": 7.292676780124806e-05}
{"train_loss": 0.1684982180595398, "global_step": 94892, "epoch": 1066, "lr": 7.292625260160525e-05}
{"train_loss": 0.16517971456050873, "global_step": 94893, "epoch": 1066, "lr": 7.292573739888026e-05}
{"train_loss": 0.17239665985107422, "global_step": 94894, "epoch": 1066, "lr": 7.292522219307322e-05}
{"train_loss": 0.1116533875465393, "global_step": 94895, "epoch": 1066, "lr": 7.292470698418415e-05}
{"train_loss": 0.17567642033100128, "global_step": 94896, "epoch": 1066, "lr": 7.292419177221312e-05}
{"train_loss": 0.23540492355823517, "global_step": 94897, "epoch": 1066, "lr": 7.292367655716021e-05}
{"train_loss": 0.14217787981033325, "global_step": 94898, "epoch": 1066, "lr": 7.29231613390255e-05}
{"train_loss": 0.09461872279644012, "global_step": 94899, "epoch": 1066, "lr": 7.292264611780905e-05}
{"train_loss": 0.20997533202171326, "global_step": 94900, "epoch": 1066, "lr": 7.292213089351093e-05}
{"train_loss": 0.15030337870121002, "global_step": 94901, "epoch": 1066, "lr": 7.292161566613121e-05}
{"train_loss": 0.18450713157653809, "global_step": 94902, "epoch": 1066, "lr": 7.292110043566996e-05}
{"train_loss": 0.10763462632894516, "global_step": 94903, "epoch": 1066, "lr": 7.292058520212724e-05}
{"train_loss": 0.15069493651390076, "global_step": 94904, "epoch": 1066, "lr": 7.292006996550312e-05}
{"train_loss": 0.21725736558437347, "global_step": 94905, "epoch": 1066, "lr": 7.291955472579768e-05}
{"train_loss": 0.14273779094219208, "global_step": 94906, "epoch": 1066, "lr": 7.2919039483011e-05}
{"train_loss": 0.1565200537443161, "global_step": 94907, "epoch": 1066, "lr": 7.291852423714311e-05}
{"train_loss": 0.12804219126701355, "global_step": 94908, "epoch": 1066, "lr": 7.291800898819412e-05}
{"train_loss": 0.1079740822315216, "global_step": 94909, "epoch": 1066, "lr": 7.291749373616407e-05}
{"train_loss": 0.15630096197128296, "global_step": 94910, "epoch": 1066, "lr": 7.291697848105306e-05}
{"train_loss": 0.13222424685955048, "global_step": 94911, "epoch": 1066, "lr": 7.291646322286114e-05}
{"train_loss": 0.17789599299430847, "global_step": 94912, "epoch": 1066, "lr": 7.291594796158838e-05}
{"train_loss": 0.1463456004858017, "global_step": 94913, "epoch": 1066, "lr": 7.291543269723484e-05}
{"train_loss": 0.2084628939628601, "global_step": 94914, "epoch": 1066, "lr": 7.291491742980061e-05}
{"train_loss": 0.08624632656574249, "global_step": 94915, "epoch": 1066, "lr": 7.291440215928576e-05}
{"train_loss": 0.14071783423423767, "global_step": 94916, "epoch": 1066, "lr": 7.291388688569032e-05}
{"train_loss": 0.1468728482723236, "global_step": 94917, "epoch": 1066, "lr": 7.29133716090144e-05}
{"train_loss": 0.1576704978942871, "global_step": 94918, "epoch": 1066, "lr": 7.291285632925806e-05}
{"train_loss": 0.13364796340465546, "global_step": 94919, "epoch": 1066, "lr": 7.291234104642136e-05}
{"train_loss": 0.22744129598140717, "global_step": 94920, "epoch": 1066, "lr": 7.291182576050437e-05}
{"train_loss": 0.19404900074005127, "global_step": 94921, "epoch": 1066, "lr": 7.291131047150718e-05}
{"train_loss": 0.12922975420951843, "global_step": 94922, "epoch": 1066, "lr": 7.291079517942982e-05}
{"train_loss": 0.18071569502353668, "global_step": 94923, "epoch": 1066, "lr": 7.291027988427242e-05}
{"train_loss": 0.1552717387676239, "global_step": 94924, "epoch": 1066, "lr": 7.290976458603498e-05}
{"train_loss": 0.09539762139320374, "global_step": 94925, "epoch": 1066, "lr": 7.290924928471761e-05}
{"train_loss": 0.17491982877254486, "global_step": 94926, "epoch": 1066, "lr": 7.290873398032037e-05}
{"train_loss": 0.12694957852363586, "global_step": 94927, "epoch": 1066, "lr": 7.290821867284332e-05}
{"train_loss": 0.12534166872501373, "global_step": 94928, "epoch": 1066, "lr": 7.290770336228654e-05}
{"train_loss": 0.2293134331703186, "global_step": 94929, "epoch": 1066, "lr": 7.290718804865012e-05}
{"train_loss": 0.20886695384979248, "global_step": 94930, "epoch": 1066, "lr": 7.290667273193411e-05}
{"train_loss": 0.22905196249485016, "global_step": 94931, "epoch": 1066, "lr": 7.290615741213855e-05}
{"train_loss": 0.09193823486566544, "global_step": 94932, "epoch": 1066, "lr": 7.290564208926355e-05}
{"train_loss": 0.15794780850410461, "global_step": 94933, "epoch": 1066, "lr": 7.290512676330917e-05}
{"train_loss": 0.18825942277908325, "global_step": 94934, "epoch": 1066, "lr": 7.290461143427548e-05}
{"train_loss": 0.18095886707305908, "global_step": 94935, "epoch": 1066, "lr": 7.290409610216253e-05}
{"train_loss": 0.0871719941496849, "global_step": 94936, "epoch": 1066, "lr": 7.290358076697042e-05}
{"train_loss": 0.1584237664937973, "global_step": 94937, "epoch": 1066, "lr": 7.290306542869918e-05}
{"train_loss": 0.14480909705162048, "global_step": 94938, "epoch": 1066, "lr": 7.290255008734892e-05}
{"train_loss": 0.15011657774448395, "global_step": 94939, "epoch": 1066, "lr": 7.290203474291968e-05}
{"train_loss": 0.18227921426296234, "global_step": 94940, "epoch": 1066, "lr": 7.290151939541154e-05}
{"train_loss": 0.13005219399929047, "global_step": 94941, "epoch": 1066, "lr": 7.290100404482459e-05}
{"train_loss": 0.13039849698543549, "global_step": 94942, "epoch": 1066, "lr": 7.290048869115885e-05}
{"train_loss": 0.1299314647912979, "global_step": 94943, "epoch": 1066, "lr": 7.289997333441443e-05}
{"train_loss": 0.15440797805786133, "global_step": 94944, "epoch": 1066, "lr": 7.289945797459139e-05}
{"train_loss": 0.181191548705101, "global_step": 94945, "epoch": 1066, "lr": 7.289894261168979e-05}
{"train_loss": 0.17071714997291565, "global_step": 94946, "epoch": 1066, "lr": 7.289842724570971e-05}
{"train_loss": 0.19041885435581207, "global_step": 94947, "epoch": 1066, "lr": 7.289791187665123e-05}
{"train_loss": 0.16435539722442627, "global_step": 94948, "epoch": 1066, "lr": 7.289739650451438e-05}
{"train_loss": 0.16138926148414612, "global_step": 94949, "epoch": 1066, "lr": 7.289688112929927e-05}
{"train_loss": 0.10747096687555313, "global_step": 94950, "epoch": 1066, "lr": 7.289636575100595e-05}
{"train_loss": 0.16932609677314758, "global_step": 94951, "epoch": 1066, "lr": 7.28958503696345e-05}
{"train_loss": 0.14577484130859375, "global_step": 94952, "epoch": 1066, "lr": 7.289533498518498e-05}
{"train_loss": 0.20851366221904755, "global_step": 94953, "epoch": 1066, "lr": 7.289481959765747e-05}
{"train_loss": 0.11474684625864029, "global_step": 94954, "epoch": 1066, "lr": 7.289430420705202e-05}
{"train_loss": 0.25332391262054443, "global_step": 94955, "epoch": 1066, "lr": 7.289378881336869e-05}
{"train_loss": 0.2444603592157364, "global_step": 94956, "epoch": 1066, "lr": 7.289327341660759e-05}
{"train_loss": 0.1385556012392044, "global_step": 94957, "epoch": 1066, "lr": 7.289275801676877e-05}
{"train_loss": 0.15003836154937744, "global_step": 94958, "epoch": 1066, "lr": 7.289224261385229e-05}
{"train_loss": 0.1174505203962326, "global_step": 94959, "epoch": 1066, "lr": 7.289172720785823e-05}
{"train_loss": 0.12978559732437134, "global_step": 94960, "epoch": 1066, "lr": 7.289121179878666e-05}
{"train_loss": 0.15756231546401978, "global_step": 94961, "epoch": 1066, "lr": 7.289069638663763e-05}
{"train_loss": 0.16003132586398822, "global_step": 94962, "epoch": 1066, "lr": 7.289018097141123e-05, "val_loss": 4.448904514312744}
{"train_loss": 0.14864173531532288, "global_step": 94963, "epoch": 1067, "lr": 7.288966555310754e-05}
{"train_loss": 0.13313721120357513, "global_step": 94964, "epoch": 1067, "lr": 7.288915013172659e-05}
{"train_loss": 0.21633227169513702, "global_step": 94965, "epoch": 1067, "lr": 7.28886347072685e-05}
{"train_loss": 0.12716488540172577, "global_step": 94966, "epoch": 1067, "lr": 7.288811927973329e-05}
{"train_loss": 0.18001006543636322, "global_step": 94967, "epoch": 1067, "lr": 7.288760384912105e-05}
{"train_loss": 0.20283734798431396, "global_step": 94968, "epoch": 1067, "lr": 7.288708841543187e-05}
{"train_loss": 0.1858300268650055, "global_step": 94969, "epoch": 1067, "lr": 7.28865729786658e-05}
{"train_loss": 0.09883140027523041, "global_step": 94970, "epoch": 1067, "lr": 7.288605753882289e-05}
{"train_loss": 0.1280611753463745, "global_step": 94971, "epoch": 1067, "lr": 7.288554209590323e-05}
{"train_loss": 0.1602446436882019, "global_step": 94972, "epoch": 1067, "lr": 7.288502664990692e-05}
{"train_loss": 0.16728733479976654, "global_step": 94973, "epoch": 1067, "lr": 7.288451120083397e-05}
{"train_loss": 0.2548536956310272, "global_step": 94974, "epoch": 1067, "lr": 7.288399574868448e-05}
{"train_loss": 0.226983904838562, "global_step": 94975, "epoch": 1067, "lr": 7.288348029345851e-05}
{"train_loss": 0.3441430926322937, "global_step": 94976, "epoch": 1067, "lr": 7.288296483515614e-05}
{"train_loss": 0.12552063167095184, "global_step": 94977, "epoch": 1067, "lr": 7.288244937377744e-05}
{"train_loss": 0.1862330287694931, "global_step": 94978, "epoch": 1067, "lr": 7.288193390932247e-05}
{"train_loss": 0.1625983864068985, "global_step": 94979, "epoch": 1067, "lr": 7.28814184417913e-05}
{"train_loss": 0.1400901824235916, "global_step": 94980, "epoch": 1067, "lr": 7.2880902971184e-05}
{"train_loss": 0.20552819967269897, "global_step": 94981, "epoch": 1067, "lr": 7.288038749750063e-05}
{"train_loss": 0.2474065124988556, "global_step": 94982, "epoch": 1067, "lr": 7.287987202074129e-05}
{"train_loss": 0.20598934590816498, "global_step": 94983, "epoch": 1067, "lr": 7.287935654090603e-05}
{"train_loss": 0.1606014370918274, "global_step": 94984, "epoch": 1067, "lr": 7.287884105799491e-05}
{"train_loss": 0.18788419663906097, "global_step": 94985, "epoch": 1067, "lr": 7.287832557200802e-05}
{"train_loss": 0.15895357728004456, "global_step": 94986, "epoch": 1067, "lr": 7.287781008294541e-05}
{"train_loss": 0.1355716586112976, "global_step": 94987, "epoch": 1067, "lr": 7.287729459080715e-05}
{"train_loss": 0.14947554469108582, "global_step": 94988, "epoch": 1067, "lr": 7.287677909559332e-05}
{"train_loss": 0.15696801245212555, "global_step": 94989, "epoch": 1067, "lr": 7.2876263597304e-05}
{"train_loss": 0.17870941758155823, "global_step": 94990, "epoch": 1067, "lr": 7.287574809593925e-05}
{"train_loss": 0.13831739127635956, "global_step": 94991, "epoch": 1067, "lr": 7.287523259149912e-05}
{"train_loss": 0.17223794758319855, "global_step": 94992, "epoch": 1067, "lr": 7.28747170839837e-05}
{"train_loss": 0.10903113335371017, "global_step": 94993, "epoch": 1067, "lr": 7.287420157339303e-05}
{"train_loss": 0.1199326366186142, "global_step": 94994, "epoch": 1067, "lr": 7.287368605972724e-05}
{"train_loss": 0.1655684858560562, "global_step": 94995, "epoch": 1067, "lr": 7.287317054298634e-05}
{"train_loss": 0.13147468864917755, "global_step": 94996, "epoch": 1067, "lr": 7.287265502317042e-05}
{"train_loss": 0.1394103616476059, "global_step": 94997, "epoch": 1067, "lr": 7.287213950027956e-05}
{"train_loss": 0.17611107230186462, "global_step": 94998, "epoch": 1067, "lr": 7.28716239743138e-05}
{"train_loss": 0.19491876661777496, "global_step": 94999, "epoch": 1067, "lr": 7.287110844527324e-05}
{"train_loss": 0.11948452889919281, "global_step": 95000, "epoch": 1067, "lr": 7.287059291315794e-05}
{"train_loss": 0.1606322079896927, "global_step": 95001, "epoch": 1067, "lr": 7.287007737796796e-05}
{"train_loss": 0.11503056436777115, "global_step": 95002, "epoch": 1067, "lr": 7.286956183970339e-05}
{"train_loss": 0.2111181616783142, "global_step": 95003, "epoch": 1067, "lr": 7.286904629836428e-05}
{"train_loss": 0.19975155591964722, "global_step": 95004, "epoch": 1067, "lr": 7.28685307539507e-05}
{"train_loss": 0.1048395112156868, "global_step": 95005, "epoch": 1067, "lr": 7.286801520646274e-05}
{"train_loss": 0.0966038927435875, "global_step": 95006, "epoch": 1067, "lr": 7.286749965590044e-05}
{"train_loss": 0.10341569036245346, "global_step": 95007, "epoch": 1067, "lr": 7.286698410226389e-05}
{"train_loss": 0.1670343577861786, "global_step": 95008, "epoch": 1067, "lr": 7.286646854555315e-05}
{"train_loss": 0.154315784573555, "global_step": 95009, "epoch": 1067, "lr": 7.286595298576829e-05}
{"train_loss": 0.14194922149181366, "global_step": 95010, "epoch": 1067, "lr": 7.28654374229094e-05}
{"train_loss": 0.15320312976837158, "global_step": 95011, "epoch": 1067, "lr": 7.286492185697651e-05}
{"train_loss": 0.15609745681285858, "global_step": 95012, "epoch": 1067, "lr": 7.286440628796971e-05}
{"train_loss": 0.24179063737392426, "global_step": 95013, "epoch": 1067, "lr": 7.286389071588908e-05}
{"train_loss": 0.11236050724983215, "global_step": 95014, "epoch": 1067, "lr": 7.286337514073466e-05}
{"train_loss": 0.13966545462608337, "global_step": 95015, "epoch": 1067, "lr": 7.286285956250656e-05}
{"train_loss": 0.14751560986042023, "global_step": 95016, "epoch": 1067, "lr": 7.286234398120482e-05}
{"train_loss": 0.14950719475746155, "global_step": 95017, "epoch": 1067, "lr": 7.28618283968295e-05}
{"train_loss": 0.16315443813800812, "global_step": 95018, "epoch": 1067, "lr": 7.286131280938071e-05}
{"train_loss": 0.18122808635234833, "global_step": 95019, "epoch": 1067, "lr": 7.286079721885848e-05}
{"train_loss": 0.25527748465538025, "global_step": 95020, "epoch": 1067, "lr": 7.286028162526291e-05}
{"train_loss": 0.20183445513248444, "global_step": 95021, "epoch": 1067, "lr": 7.285976602859403e-05}
{"train_loss": 0.2042621672153473, "global_step": 95022, "epoch": 1067, "lr": 7.285925042885196e-05}
{"train_loss": 0.1803838610649109, "global_step": 95023, "epoch": 1067, "lr": 7.285873482603672e-05}
{"train_loss": 0.19813527166843414, "global_step": 95024, "epoch": 1067, "lr": 7.285821922014841e-05}
{"train_loss": 0.2029130607843399, "global_step": 95025, "epoch": 1067, "lr": 7.285770361118709e-05}
{"train_loss": 0.15683868527412415, "global_step": 95026, "epoch": 1067, "lr": 7.285718799915284e-05}
{"train_loss": 0.16580511629581451, "global_step": 95027, "epoch": 1067, "lr": 7.28566723840457e-05}
{"train_loss": 0.1853025257587433, "global_step": 95028, "epoch": 1067, "lr": 7.285615676586577e-05}
{"train_loss": 0.16598635911941528, "global_step": 95029, "epoch": 1067, "lr": 7.285564114461311e-05}
{"train_loss": 0.13193678855895996, "global_step": 95030, "epoch": 1067, "lr": 7.28551255202878e-05}
{"train_loss": 0.2363445907831192, "global_step": 95031, "epoch": 1067, "lr": 7.285460989288988e-05}
{"train_loss": 0.1411677747964859, "global_step": 95032, "epoch": 1067, "lr": 7.285409426241945e-05}
{"train_loss": 0.22337070107460022, "global_step": 95033, "epoch": 1067, "lr": 7.285357862887655e-05}
{"train_loss": 0.13834033906459808, "global_step": 95034, "epoch": 1067, "lr": 7.285306299226126e-05}
{"train_loss": 0.1852189004421234, "global_step": 95035, "epoch": 1067, "lr": 7.285254735257367e-05}
{"train_loss": 0.10572770982980728, "global_step": 95036, "epoch": 1067, "lr": 7.285203170981381e-05}
{"train_loss": 0.19938796758651733, "global_step": 95037, "epoch": 1067, "lr": 7.28515160639818e-05}
{"train_loss": 0.16068288683891296, "global_step": 95038, "epoch": 1067, "lr": 7.285100041507768e-05}
{"train_loss": 0.1688438206911087, "global_step": 95039, "epoch": 1067, "lr": 7.285048476310151e-05}
{"train_loss": 0.13691376149654388, "global_step": 95040, "epoch": 1067, "lr": 7.284996910805337e-05}
{"train_loss": 0.15879905223846436, "global_step": 95041, "epoch": 1067, "lr": 7.284945344993333e-05}
{"train_loss": 0.1487847864627838, "global_step": 95042, "epoch": 1067, "lr": 7.284893778874148e-05}
{"train_loss": 0.0912691280245781, "global_step": 95043, "epoch": 1067, "lr": 7.284842212447785e-05}
{"train_loss": 0.12589047849178314, "global_step": 95044, "epoch": 1067, "lr": 7.284790645714253e-05}
{"train_loss": 0.20899750292301178, "global_step": 95045, "epoch": 1067, "lr": 7.284739078673559e-05}
{"train_loss": 0.17835915088653564, "global_step": 95046, "epoch": 1067, "lr": 7.28468751132571e-05}
{"train_loss": 0.19161072373390198, "global_step": 95047, "epoch": 1067, "lr": 7.284635943670712e-05}
{"train_loss": 0.2662939727306366, "global_step": 95048, "epoch": 1067, "lr": 7.284584375708572e-05}
{"train_loss": 0.13779400289058685, "global_step": 95049, "epoch": 1067, "lr": 7.2845328074393e-05}
{"train_loss": 0.21792015433311462, "global_step": 95050, "epoch": 1067, "lr": 7.284481238862898e-05}
{"train_loss": 0.16812034729826317, "global_step": 95051, "epoch": 1067, "lr": 7.284429669979376e-05, "val_loss": 4.368371486663818}
{"train_loss": 0.14218010008335114, "global_step": 95052, "epoch": 1068, "lr": 7.28437810078874e-05}
{"train_loss": 0.12012044340372086, "global_step": 95053, "epoch": 1068, "lr": 7.284326531290997e-05}
{"train_loss": 0.14988498389720917, "global_step": 95054, "epoch": 1068, "lr": 7.284274961486155e-05}
{"train_loss": 0.1842852234840393, "global_step": 95055, "epoch": 1068, "lr": 7.28422339137422e-05}
{"train_loss": 0.1477593183517456, "global_step": 95056, "epoch": 1068, "lr": 7.284171820955197e-05}
{"train_loss": 0.16142772138118744, "global_step": 95057, "epoch": 1068, "lr": 7.284120250229098e-05}
{"train_loss": 0.1664166897535324, "global_step": 95058, "epoch": 1068, "lr": 7.284068679195926e-05}
{"train_loss": 0.12794944643974304, "global_step": 95059, "epoch": 1068, "lr": 7.284017107855688e-05}
{"train_loss": 0.16624951362609863, "global_step": 95060, "epoch": 1068, "lr": 7.283965536208391e-05}
{"train_loss": 0.12207456678152084, "global_step": 95061, "epoch": 1068, "lr": 7.283913964254045e-05}
{"train_loss": 0.12556369602680206, "global_step": 95062, "epoch": 1068, "lr": 7.283862391992653e-05}
{"train_loss": 0.15434516966342926, "global_step": 95063, "epoch": 1068, "lr": 7.283810819424224e-05}
{"train_loss": 0.1487135887145996, "global_step": 95064, "epoch": 1068, "lr": 7.283759246548763e-05}
{"train_loss": 0.154952272772789, "global_step": 95065, "epoch": 1068, "lr": 7.283707673366281e-05}
{"train_loss": 0.1991666555404663, "global_step": 95066, "epoch": 1068, "lr": 7.283656099876782e-05}
{"train_loss": 0.14084385335445404, "global_step": 95067, "epoch": 1068, "lr": 7.28360452608027e-05}
{"train_loss": 0.1870115101337433, "global_step": 95068, "epoch": 1068, "lr": 7.28355295197676e-05}
{"train_loss": 0.17640502750873566, "global_step": 95069, "epoch": 1068, "lr": 7.283501377566251e-05}
{"train_loss": 0.1302221417427063, "global_step": 95070, "epoch": 1068, "lr": 7.283449802848754e-05}
{"train_loss": 0.18059654533863068, "global_step": 95071, "epoch": 1068, "lr": 7.283398227824274e-05}
{"train_loss": 0.12671472132205963, "global_step": 95072, "epoch": 1068, "lr": 7.28334665249282e-05}
{"train_loss": 0.18950971961021423, "global_step": 95073, "epoch": 1068, "lr": 7.283295076854397e-05}
{"train_loss": 0.21844659745693207, "global_step": 95074, "epoch": 1068, "lr": 7.283243500909013e-05}
{"train_loss": 0.1594451367855072, "global_step": 95075, "epoch": 1068, "lr": 7.283191924656676e-05}
{"train_loss": 0.1587909609079361, "global_step": 95076, "epoch": 1068, "lr": 7.28314034809739e-05}
{"train_loss": 0.16126319766044617, "global_step": 95077, "epoch": 1068, "lr": 7.283088771231165e-05}
{"train_loss": 0.14053376019001007, "global_step": 95078, "epoch": 1068, "lr": 7.283037194058005e-05}
{"train_loss": 0.1452607810497284, "global_step": 95079, "epoch": 1068, "lr": 7.28298561657792e-05}
{"train_loss": 0.18760119378566742, "global_step": 95080, "epoch": 1068, "lr": 7.282934038790915e-05}
{"train_loss": 0.1820310801267624, "global_step": 95081, "epoch": 1068, "lr": 7.282882460696997e-05}
{"train_loss": 0.1725635528564453, "global_step": 95082, "epoch": 1068, "lr": 7.282830882296174e-05}
{"train_loss": 0.1406768411397934, "global_step": 95083, "epoch": 1068, "lr": 7.282779303588452e-05}
{"train_loss": 0.2086196094751358, "global_step": 95084, "epoch": 1068, "lr": 7.282727724573837e-05}
{"train_loss": 0.18490886688232422, "global_step": 95085, "epoch": 1068, "lr": 7.282676145252339e-05}
{"train_loss": 0.11887140572071075, "global_step": 95086, "epoch": 1068, "lr": 7.282624565623961e-05}
{"train_loss": 0.19425839185714722, "global_step": 95087, "epoch": 1068, "lr": 7.282572985688712e-05}
{"train_loss": 0.172856867313385, "global_step": 95088, "epoch": 1068, "lr": 7.2825214054466e-05}
{"train_loss": 0.18106302618980408, "global_step": 95089, "epoch": 1068, "lr": 7.28246982489763e-05}
{"train_loss": 0.1533660739660263, "global_step": 95090, "epoch": 1068, "lr": 7.282418244041811e-05}
{"train_loss": 0.14038428664207458, "global_step": 95091, "epoch": 1068, "lr": 7.282366662879148e-05}
{"train_loss": 0.19257409870624542, "global_step": 95092, "epoch": 1068, "lr": 7.282315081409649e-05}
{"train_loss": 0.16173478960990906, "global_step": 95093, "epoch": 1068, "lr": 7.28226349963332e-05}
{"train_loss": 0.17109696567058563, "global_step": 95094, "epoch": 1068, "lr": 7.282211917550168e-05}
{"train_loss": 0.158918559551239, "global_step": 95095, "epoch": 1068, "lr": 7.282160335160202e-05}
{"train_loss": 0.07547269016504288, "global_step": 95096, "epoch": 1068, "lr": 7.282108752463425e-05}
{"train_loss": 0.21570253372192383, "global_step": 95097, "epoch": 1068, "lr": 7.282057169459848e-05}
{"train_loss": 0.154469296336174, "global_step": 95098, "epoch": 1068, "lr": 7.282005586149476e-05}
{"train_loss": 0.15067966282367706, "global_step": 95099, "epoch": 1068, "lr": 7.281954002532316e-05}
{"train_loss": 0.189475417137146, "global_step": 95100, "epoch": 1068, "lr": 7.281902418608377e-05}
{"train_loss": 0.111654132604599, "global_step": 95101, "epoch": 1068, "lr": 7.281850834377661e-05}
{"train_loss": 0.1604258269071579, "global_step": 95102, "epoch": 1068, "lr": 7.28179924984018e-05}
{"train_loss": 0.12017287313938141, "global_step": 95103, "epoch": 1068, "lr": 7.281747664995937e-05}
{"train_loss": 0.15494711697101593, "global_step": 95104, "epoch": 1068, "lr": 7.281696079844943e-05}
{"train_loss": 0.1912822425365448, "global_step": 95105, "epoch": 1068, "lr": 7.2816444943872e-05}
{"train_loss": 0.1524442434310913, "global_step": 95106, "epoch": 1068, "lr": 7.281592908622721e-05}
{"train_loss": 0.09666819870471954, "global_step": 95107, "epoch": 1068, "lr": 7.281541322551508e-05}
{"train_loss": 0.2410295158624649, "global_step": 95108, "epoch": 1068, "lr": 7.28148973617357e-05}
{"train_loss": 0.18331550061702728, "global_step": 95109, "epoch": 1068, "lr": 7.281438149488913e-05}
{"train_loss": 0.18920183181762695, "global_step": 95110, "epoch": 1068, "lr": 7.281386562497544e-05}
{"train_loss": 0.16920195519924164, "global_step": 95111, "epoch": 1068, "lr": 7.281334975199472e-05}
{"train_loss": 0.216678649187088, "global_step": 95112, "epoch": 1068, "lr": 7.2812833875947e-05}
{"train_loss": 0.09513445198535919, "global_step": 95113, "epoch": 1068, "lr": 7.281231799683239e-05}
{"train_loss": 0.14231719076633453, "global_step": 95114, "epoch": 1068, "lr": 7.281180211465094e-05}
{"train_loss": 0.11752686649560928, "global_step": 95115, "epoch": 1068, "lr": 7.281128622940272e-05}
{"train_loss": 0.23754730820655823, "global_step": 95116, "epoch": 1068, "lr": 7.28107703410878e-05}
{"train_loss": 0.11044714599847794, "global_step": 95117, "epoch": 1068, "lr": 7.281025444970625e-05}
{"train_loss": 0.16430708765983582, "global_step": 95118, "epoch": 1068, "lr": 7.280973855525813e-05}
{"train_loss": 0.13980141282081604, "global_step": 95119, "epoch": 1068, "lr": 7.280922265774353e-05}
{"train_loss": 0.14208146929740906, "global_step": 95120, "epoch": 1068, "lr": 7.280870675716252e-05}
{"train_loss": 0.20033273100852966, "global_step": 95121, "epoch": 1068, "lr": 7.280819085351513e-05}
{"train_loss": 0.12937483191490173, "global_step": 95122, "epoch": 1068, "lr": 7.280767494680147e-05}
{"train_loss": 0.14824245870113373, "global_step": 95123, "epoch": 1068, "lr": 7.28071590370216e-05}
{"train_loss": 0.18356597423553467, "global_step": 95124, "epoch": 1068, "lr": 7.280664312417557e-05}
{"train_loss": 0.17323291301727295, "global_step": 95125, "epoch": 1068, "lr": 7.280612720826349e-05}
{"train_loss": 0.18986718356609344, "global_step": 95126, "epoch": 1068, "lr": 7.280561128928538e-05}
{"train_loss": 0.1693607121706009, "global_step": 95127, "epoch": 1068, "lr": 7.280509536724134e-05}
{"train_loss": 0.16405752301216125, "global_step": 95128, "epoch": 1068, "lr": 7.280457944213143e-05}
{"train_loss": 0.16097180545330048, "global_step": 95129, "epoch": 1068, "lr": 7.280406351395575e-05}
{"train_loss": 0.1630789041519165, "global_step": 95130, "epoch": 1068, "lr": 7.280354758271432e-05}
{"train_loss": 0.22203388810157776, "global_step": 95131, "epoch": 1068, "lr": 7.280303164840723e-05}
{"train_loss": 0.1842261254787445, "global_step": 95132, "epoch": 1068, "lr": 7.280251571103454e-05}
{"train_loss": 0.12239930778741837, "global_step": 95133, "epoch": 1068, "lr": 7.280199977059634e-05}
{"train_loss": 0.13767220079898834, "global_step": 95134, "epoch": 1068, "lr": 7.28014838270927e-05}
{"train_loss": 0.2597283124923706, "global_step": 95135, "epoch": 1068, "lr": 7.280096788052366e-05}
{"train_loss": 0.12996311485767365, "global_step": 95136, "epoch": 1068, "lr": 7.280045193088933e-05}
{"train_loss": 0.2042304426431656, "global_step": 95137, "epoch": 1068, "lr": 7.279993597818973e-05}
{"train_loss": 0.21765844523906708, "global_step": 95138, "epoch": 1068, "lr": 7.279942002242497e-05}
{"train_loss": 0.1281735748052597, "global_step": 95139, "epoch": 1068, "lr": 7.279890406359511e-05}
{"train_loss": 0.1619567812493678, "global_step": 95140, "epoch": 1068, "lr": 7.279838810170021e-05, "val_loss": 4.429117679595947}
{"train_loss": 0.1424298882484436, "global_step": 95141, "epoch": 1069, "lr": 7.279787213674034e-05}
{"train_loss": 0.17503590881824493, "global_step": 95142, "epoch": 1069, "lr": 7.279735616871559e-05}
{"train_loss": 0.1950879544019699, "global_step": 95143, "epoch": 1069, "lr": 7.2796840197626e-05}
{"train_loss": 0.2010948359966278, "global_step": 95144, "epoch": 1069, "lr": 7.279632422347165e-05}
{"train_loss": 0.13464690744876862, "global_step": 95145, "epoch": 1069, "lr": 7.279580824625262e-05}
{"train_loss": 0.18069815635681152, "global_step": 95146, "epoch": 1069, "lr": 7.279529226596898e-05}
{"train_loss": 0.15847979485988617, "global_step": 95147, "epoch": 1069, "lr": 7.279477628262077e-05}
{"train_loss": 0.15487916767597198, "global_step": 95148, "epoch": 1069, "lr": 7.27942602962081e-05}
{"train_loss": 0.17197835445404053, "global_step": 95149, "epoch": 1069, "lr": 7.2793744306731e-05}
{"train_loss": 0.11839071661233902, "global_step": 95150, "epoch": 1069, "lr": 7.279322831418958e-05}
{"train_loss": 0.17917300760746002, "global_step": 95151, "epoch": 1069, "lr": 7.279271231858387e-05}
{"train_loss": 0.14461638033390045, "global_step": 95152, "epoch": 1069, "lr": 7.279219631991397e-05}
{"train_loss": 0.1300348937511444, "global_step": 95153, "epoch": 1069, "lr": 7.279168031817993e-05}
{"train_loss": 0.15928877890110016, "global_step": 95154, "epoch": 1069, "lr": 7.279116431338182e-05}
{"train_loss": 0.14756231009960175, "global_step": 95155, "epoch": 1069, "lr": 7.279064830551973e-05}
{"train_loss": 0.14239464700222015, "global_step": 95156, "epoch": 1069, "lr": 7.279013229459371e-05}
{"train_loss": 0.07895456999540329, "global_step": 95157, "epoch": 1069, "lr": 7.278961628060382e-05}
{"train_loss": 0.2220749706029892, "global_step": 95158, "epoch": 1069, "lr": 7.278910026355017e-05}
{"train_loss": 0.19452068209648132, "global_step": 95159, "epoch": 1069, "lr": 7.278858424343278e-05}
{"train_loss": 0.21924720704555511, "global_step": 95160, "epoch": 1069, "lr": 7.278806822025176e-05}
{"train_loss": 0.14278224110603333, "global_step": 95161, "epoch": 1069, "lr": 7.278755219400715e-05}
{"train_loss": 0.27266719937324524, "global_step": 95162, "epoch": 1069, "lr": 7.278703616469904e-05}
{"train_loss": 0.14584137499332428, "global_step": 95163, "epoch": 1069, "lr": 7.278652013232748e-05}
{"train_loss": 0.1795412302017212, "global_step": 95164, "epoch": 1069, "lr": 7.278600409689256e-05}
{"train_loss": 0.14751280844211578, "global_step": 95165, "epoch": 1069, "lr": 7.278548805839433e-05}
{"train_loss": 0.13583701848983765, "global_step": 95166, "epoch": 1069, "lr": 7.278497201683288e-05}
{"train_loss": 0.15377266705036163, "global_step": 95167, "epoch": 1069, "lr": 7.278445597220826e-05}
{"train_loss": 0.1713639199733734, "global_step": 95168, "epoch": 1069, "lr": 7.278393992452056e-05}
{"train_loss": 0.23504206538200378, "global_step": 95169, "epoch": 1069, "lr": 7.278342387376984e-05}
{"train_loss": 0.20936299860477448, "global_step": 95170, "epoch": 1069, "lr": 7.278290781995614e-05}
{"train_loss": 0.17302031815052032, "global_step": 95171, "epoch": 1069, "lr": 7.278239176307957e-05}
{"train_loss": 0.1490955352783203, "global_step": 95172, "epoch": 1069, "lr": 7.278187570314019e-05}
{"train_loss": 0.22076044976711273, "global_step": 95173, "epoch": 1069, "lr": 7.278135964013806e-05}
{"train_loss": 0.1948242038488388, "global_step": 95174, "epoch": 1069, "lr": 7.278084357407326e-05}
{"train_loss": 0.17306405305862427, "global_step": 95175, "epoch": 1069, "lr": 7.278032750494583e-05}
{"train_loss": 0.1315356343984604, "global_step": 95176, "epoch": 1069, "lr": 7.277981143275588e-05}
{"train_loss": 0.17028826475143433, "global_step": 95177, "epoch": 1069, "lr": 7.277929535750347e-05}
{"train_loss": 0.1394093930721283, "global_step": 95178, "epoch": 1069, "lr": 7.277877927918865e-05}
{"train_loss": 0.16097313165664673, "global_step": 95179, "epoch": 1069, "lr": 7.27782631978115e-05}
{"train_loss": 0.17797859013080597, "global_step": 95180, "epoch": 1069, "lr": 7.27777471133721e-05}
{"train_loss": 0.15374808013439178, "global_step": 95181, "epoch": 1069, "lr": 7.277723102587051e-05}
{"train_loss": 0.15584686398506165, "global_step": 95182, "epoch": 1069, "lr": 7.277671493530678e-05}
{"train_loss": 0.1896408498287201, "global_step": 95183, "epoch": 1069, "lr": 7.2776198841681e-05}
{"train_loss": 0.16200339794158936, "global_step": 95184, "epoch": 1069, "lr": 7.277568274499327e-05}
{"train_loss": 0.20223964750766754, "global_step": 95185, "epoch": 1069, "lr": 7.277516664524359e-05}
{"train_loss": 0.17283916473388672, "global_step": 95186, "epoch": 1069, "lr": 7.277465054243209e-05}
{"train_loss": 0.15471069514751434, "global_step": 95187, "epoch": 1069, "lr": 7.27741344365588e-05}
{"train_loss": 0.173021137714386, "global_step": 95188, "epoch": 1069, "lr": 7.277361832762382e-05}
{"train_loss": 0.13960282504558563, "global_step": 95189, "epoch": 1069, "lr": 7.27731022156272e-05}
{"train_loss": 0.1319899559020996, "global_step": 95190, "epoch": 1069, "lr": 7.277258610056901e-05}
{"train_loss": 0.14111453294754028, "global_step": 95191, "epoch": 1069, "lr": 7.277206998244932e-05}
{"train_loss": 0.16058708727359772, "global_step": 95192, "epoch": 1069, "lr": 7.277155386126822e-05}
{"train_loss": 0.21212461590766907, "global_step": 95193, "epoch": 1069, "lr": 7.277103773702574e-05}
{"train_loss": 0.1678011417388916, "global_step": 95194, "epoch": 1069, "lr": 7.277052160972198e-05}
{"train_loss": 0.1526845544576645, "global_step": 95195, "epoch": 1069, "lr": 7.277000547935701e-05}
{"train_loss": 0.17160427570343018, "global_step": 95196, "epoch": 1069, "lr": 7.276948934593088e-05}
{"train_loss": 0.25697943568229675, "global_step": 95197, "epoch": 1069, "lr": 7.276897320944368e-05}
{"train_loss": 0.10851588845252991, "global_step": 95198, "epoch": 1069, "lr": 7.276845706989543e-05}
{"train_loss": 0.09037646651268005, "global_step": 95199, "epoch": 1069, "lr": 7.27679409272863e-05}
{"train_loss": 0.2272670567035675, "global_step": 95200, "epoch": 1069, "lr": 7.276742478161626e-05}
{"train_loss": 0.25196725130081177, "global_step": 95201, "epoch": 1069, "lr": 7.276690863288542e-05}
{"train_loss": 0.13629953563213348, "global_step": 95202, "epoch": 1069, "lr": 7.276639248109385e-05}
{"train_loss": 0.18186689913272858, "global_step": 95203, "epoch": 1069, "lr": 7.27658763262416e-05}
{"train_loss": 0.23618923127651215, "global_step": 95204, "epoch": 1069, "lr": 7.27653601683288e-05}
{"train_loss": 0.2601493299007416, "global_step": 95205, "epoch": 1069, "lr": 7.276484400735544e-05}
{"train_loss": 0.23823904991149902, "global_step": 95206, "epoch": 1069, "lr": 7.276432784332163e-05}
{"train_loss": 0.1917528361082077, "global_step": 95207, "epoch": 1069, "lr": 7.276381167622743e-05}
{"train_loss": 0.1702626645565033, "global_step": 95208, "epoch": 1069, "lr": 7.276329550607292e-05}
{"train_loss": 0.11858774721622467, "global_step": 95209, "epoch": 1069, "lr": 7.276277933285815e-05}
{"train_loss": 0.17914055287837982, "global_step": 95210, "epoch": 1069, "lr": 7.276226315658323e-05}
{"train_loss": 0.19614970684051514, "global_step": 95211, "epoch": 1069, "lr": 7.276174697724818e-05}
{"train_loss": 0.10631906986236572, "global_step": 95212, "epoch": 1069, "lr": 7.276123079485308e-05}
{"train_loss": 0.1606220304965973, "global_step": 95213, "epoch": 1069, "lr": 7.276071460939804e-05}
{"train_loss": 0.17431093752384186, "global_step": 95214, "epoch": 1069, "lr": 7.276019842088308e-05}
{"train_loss": 0.21893295645713806, "global_step": 95215, "epoch": 1069, "lr": 7.27596822293083e-05}
{"train_loss": 0.16803471744060516, "global_step": 95216, "epoch": 1069, "lr": 7.275916603467373e-05}
{"train_loss": 0.14786690473556519, "global_step": 95217, "epoch": 1069, "lr": 7.275864983697949e-05}
{"train_loss": 0.16877314448356628, "global_step": 95218, "epoch": 1069, "lr": 7.275813363622563e-05}
{"train_loss": 0.1505347192287445, "global_step": 95219, "epoch": 1069, "lr": 7.275761743241221e-05}
{"train_loss": 0.18890775740146637, "global_step": 95220, "epoch": 1069, "lr": 7.275710122553931e-05}
{"train_loss": 0.2274927943944931, "global_step": 95221, "epoch": 1069, "lr": 7.2756585015607e-05}
{"train_loss": 0.13561999797821045, "global_step": 95222, "epoch": 1069, "lr": 7.275606880261533e-05}
{"train_loss": 0.17193715274333954, "global_step": 95223, "epoch": 1069, "lr": 7.275555258656439e-05}
{"train_loss": 0.14812660217285156, "global_step": 95224, "epoch": 1069, "lr": 7.275503636745425e-05}
{"train_loss": 0.16324001550674438, "global_step": 95225, "epoch": 1069, "lr": 7.275452014528496e-05}
{"train_loss": 0.2489672601222992, "global_step": 95226, "epoch": 1069, "lr": 7.27540039200566e-05}
{"train_loss": 0.15421819686889648, "global_step": 95227, "epoch": 1069, "lr": 7.275348769176925e-05}
{"train_loss": 0.10222813487052917, "global_step": 95228, "epoch": 1069, "lr": 7.275297146042297e-05}
{"train_loss": 0.17122349762514735, "global_step": 95229, "epoch": 1069, "lr": 7.275245522601784e-05, "val_loss": 4.493058681488037}
{"train_loss": 0.16480667889118195, "global_step": 95230, "epoch": 1070, "lr": 7.27519389885539e-05}
{"train_loss": 0.2301308810710907, "global_step": 95231, "epoch": 1070, "lr": 7.275142274803127e-05}
{"train_loss": 0.20793385803699493, "global_step": 95232, "epoch": 1070, "lr": 7.275090650444996e-05}
{"train_loss": 0.1665591299533844, "global_step": 95233, "epoch": 1070, "lr": 7.275039025781007e-05}
{"train_loss": 0.15183120965957642, "global_step": 95234, "epoch": 1070, "lr": 7.274987400811166e-05}
{"train_loss": 0.17028462886810303, "global_step": 95235, "epoch": 1070, "lr": 7.274935775535483e-05}
{"train_loss": 0.14054161310195923, "global_step": 95236, "epoch": 1070, "lr": 7.274884149953961e-05}
{"train_loss": 0.23542171716690063, "global_step": 95237, "epoch": 1070, "lr": 7.274832524066611e-05}
{"train_loss": 0.156777486205101, "global_step": 95238, "epoch": 1070, "lr": 7.274780897873434e-05}
{"train_loss": 0.15086416900157928, "global_step": 95239, "epoch": 1070, "lr": 7.274729271374442e-05}
{"train_loss": 0.23547545075416565, "global_step": 95240, "epoch": 1070, "lr": 7.27467764456964e-05}
{"train_loss": 0.11465160548686981, "global_step": 95241, "epoch": 1070, "lr": 7.274626017459038e-05}
{"train_loss": 0.20631887018680573, "global_step": 95242, "epoch": 1070, "lr": 7.274574390042636e-05}
{"train_loss": 0.18117555975914001, "global_step": 95243, "epoch": 1070, "lr": 7.274522762320449e-05}
{"train_loss": 0.21381604671478271, "global_step": 95244, "epoch": 1070, "lr": 7.274471134292477e-05}
{"train_loss": 0.2101486623287201, "global_step": 95245, "epoch": 1070, "lr": 7.274419505958732e-05}
{"train_loss": 0.14459823071956635, "global_step": 95246, "epoch": 1070, "lr": 7.274367877319219e-05}
{"train_loss": 0.1567479968070984, "global_step": 95247, "epoch": 1070, "lr": 7.274316248373944e-05}
{"train_loss": 0.1627703160047531, "global_step": 95248, "epoch": 1070, "lr": 7.274264619122916e-05}
{"train_loss": 0.20062954723834991, "global_step": 95249, "epoch": 1070, "lr": 7.274212989566141e-05}
{"train_loss": 0.1686285436153412, "global_step": 95250, "epoch": 1070, "lr": 7.274161359703625e-05}
{"train_loss": 0.11395212262868881, "global_step": 95251, "epoch": 1070, "lr": 7.274109729535374e-05}
{"train_loss": 0.1572784185409546, "global_step": 95252, "epoch": 1070, "lr": 7.274058099061399e-05}
{"train_loss": 0.22864499688148499, "global_step": 95253, "epoch": 1070, "lr": 7.274006468281702e-05}
{"train_loss": 0.16534756124019623, "global_step": 95254, "epoch": 1070, "lr": 7.273954837196295e-05}
{"train_loss": 0.222699373960495, "global_step": 95255, "epoch": 1070, "lr": 7.273903205805182e-05}
{"train_loss": 0.19719266891479492, "global_step": 95256, "epoch": 1070, "lr": 7.27385157410837e-05}
{"train_loss": 0.2652646005153656, "global_step": 95257, "epoch": 1070, "lr": 7.273799942105868e-05}
{"train_loss": 0.14009536802768707, "global_step": 95258, "epoch": 1070, "lr": 7.273748309797682e-05}
{"train_loss": 0.16208213567733765, "global_step": 95259, "epoch": 1070, "lr": 7.273696677183814e-05}
{"train_loss": 0.17485760152339935, "global_step": 95260, "epoch": 1070, "lr": 7.273645044264278e-05}
{"train_loss": 0.2463017851114273, "global_step": 95261, "epoch": 1070, "lr": 7.273593411039078e-05}
{"train_loss": 0.22984978556632996, "global_step": 95262, "epoch": 1070, "lr": 7.273541777508222e-05}
{"train_loss": 0.15754446387290955, "global_step": 95263, "epoch": 1070, "lr": 7.273490143671714e-05}
{"train_loss": 0.21794113516807556, "global_step": 95264, "epoch": 1070, "lr": 7.273438509529563e-05}
{"train_loss": 0.17911410331726074, "global_step": 95265, "epoch": 1070, "lr": 7.273386875081777e-05}
{"train_loss": 0.13508552312850952, "global_step": 95266, "epoch": 1070, "lr": 7.273335240328362e-05}
{"train_loss": 0.12487637251615524, "global_step": 95267, "epoch": 1070, "lr": 7.273283605269323e-05}
{"train_loss": 0.16948992013931274, "global_step": 95268, "epoch": 1070, "lr": 7.27323196990467e-05}
{"train_loss": 0.2116585075855255, "global_step": 95269, "epoch": 1070, "lr": 7.27318033423441e-05}
{"train_loss": 0.23858104646205902, "global_step": 95270, "epoch": 1070, "lr": 7.273128698258546e-05}
{"train_loss": 0.17991846799850464, "global_step": 95271, "epoch": 1070, "lr": 7.273077061977088e-05}
{"train_loss": 0.18935056030750275, "global_step": 95272, "epoch": 1070, "lr": 7.273025425390045e-05}
{"train_loss": 0.22507360577583313, "global_step": 95273, "epoch": 1070, "lr": 7.272973788497419e-05}
{"train_loss": 0.3485461175441742, "global_step": 95274, "epoch": 1070, "lr": 7.27292215129922e-05}
{"train_loss": 0.1534181535243988, "global_step": 95275, "epoch": 1070, "lr": 7.272870513795455e-05}
{"train_loss": 0.11425464600324631, "global_step": 95276, "epoch": 1070, "lr": 7.27281887598613e-05}
{"train_loss": 0.21239756047725677, "global_step": 95277, "epoch": 1070, "lr": 7.272767237871252e-05}
{"train_loss": 0.2408265918493271, "global_step": 95278, "epoch": 1070, "lr": 7.272715599450828e-05}
{"train_loss": 0.22274962067604065, "global_step": 95279, "epoch": 1070, "lr": 7.272663960724864e-05}
{"train_loss": 0.1664309948682785, "global_step": 95280, "epoch": 1070, "lr": 7.272612321693371e-05}
{"train_loss": 0.23832732439041138, "global_step": 95281, "epoch": 1070, "lr": 7.27256068235635e-05}
{"train_loss": 0.2153456062078476, "global_step": 95282, "epoch": 1070, "lr": 7.272509042713813e-05}
{"train_loss": 0.15167436003684998, "global_step": 95283, "epoch": 1070, "lr": 7.272457402765763e-05}
{"train_loss": 0.1329064816236496, "global_step": 95284, "epoch": 1070, "lr": 7.27240576251221e-05}
{"train_loss": 0.14452648162841797, "global_step": 95285, "epoch": 1070, "lr": 7.27235412195316e-05}
{"train_loss": 0.16524456441402435, "global_step": 95286, "epoch": 1070, "lr": 7.272302481088619e-05}
{"train_loss": 0.239082470536232, "global_step": 95287, "epoch": 1070, "lr": 7.272250839918595e-05}
{"train_loss": 0.15987306833267212, "global_step": 95288, "epoch": 1070, "lr": 7.272199198443093e-05}
{"train_loss": 0.15199865400791168, "global_step": 95289, "epoch": 1070, "lr": 7.272147556662124e-05}
{"train_loss": 0.20207609236240387, "global_step": 95290, "epoch": 1070, "lr": 7.27209591457569e-05}
{"train_loss": 0.18239884078502655, "global_step": 95291, "epoch": 1070, "lr": 7.272044272183803e-05}
{"train_loss": 0.1782623678445816, "global_step": 95292, "epoch": 1070, "lr": 7.271992629486464e-05}
{"train_loss": 0.19168682396411896, "global_step": 95293, "epoch": 1070, "lr": 7.271940986483686e-05}
{"train_loss": 0.11997353285551071, "global_step": 95294, "epoch": 1070, "lr": 7.271889343175473e-05}
{"train_loss": 0.22851227223873138, "global_step": 95295, "epoch": 1070, "lr": 7.271837699561832e-05}
{"train_loss": 0.13642066717147827, "global_step": 95296, "epoch": 1070, "lr": 7.271786055642771e-05}
{"train_loss": 0.20070607960224152, "global_step": 95297, "epoch": 1070, "lr": 7.271734411418294e-05}
{"train_loss": 0.19106712937355042, "global_step": 95298, "epoch": 1070, "lr": 7.271682766888411e-05}
{"train_loss": 0.12954974174499512, "global_step": 95299, "epoch": 1070, "lr": 7.271631122053128e-05}
{"train_loss": 0.18827678263187408, "global_step": 95300, "epoch": 1070, "lr": 7.27157947691245e-05}
{"train_loss": 0.13586078584194183, "global_step": 95301, "epoch": 1070, "lr": 7.271527831466389e-05}
{"train_loss": 0.20272597670555115, "global_step": 95302, "epoch": 1070, "lr": 7.271476185714947e-05}
{"train_loss": 0.15808191895484924, "global_step": 95303, "epoch": 1070, "lr": 7.271424539658133e-05}
{"train_loss": 0.21148307621479034, "global_step": 95304, "epoch": 1070, "lr": 7.271372893295955e-05}
{"train_loss": 0.12894980609416962, "global_step": 95305, "epoch": 1070, "lr": 7.271321246628417e-05}
{"train_loss": 0.14356546103954315, "global_step": 95306, "epoch": 1070, "lr": 7.271269599655527e-05}
{"train_loss": 0.14953012764453888, "global_step": 95307, "epoch": 1070, "lr": 7.271217952377293e-05}
{"train_loss": 0.19466617703437805, "global_step": 95308, "epoch": 1070, "lr": 7.271166304793722e-05}
{"train_loss": 0.13424615561962128, "global_step": 95309, "epoch": 1070, "lr": 7.271114656904819e-05}
{"train_loss": 0.1979149729013443, "global_step": 95310, "epoch": 1070, "lr": 7.271063008710594e-05}
{"train_loss": 0.20813387632369995, "global_step": 95311, "epoch": 1070, "lr": 7.271011360211052e-05}
{"train_loss": 0.18744006752967834, "global_step": 95312, "epoch": 1070, "lr": 7.2709597114062e-05}
{"train_loss": 0.17985162138938904, "global_step": 95313, "epoch": 1070, "lr": 7.270908062296044e-05}
{"train_loss": 0.2625962197780609, "global_step": 95314, "epoch": 1070, "lr": 7.270856412880593e-05}
{"train_loss": 0.2055278718471527, "global_step": 95315, "epoch": 1070, "lr": 7.270804763159856e-05}
{"train_loss": 0.17886340618133545, "global_step": 95316, "epoch": 1070, "lr": 7.270753113133833e-05}
{"train_loss": 0.14536772668361664, "global_step": 95317, "epoch": 1070, "lr": 7.270701462802536e-05}
{"train_loss": 0.1838574806291066, "global_step": 95318, "epoch": 1070, "lr": 7.27064981216597e-05, "val_loss": 4.439914703369141, "train_action_mse_error": 6.6821441650390625}
{"train_loss": 0.17031285166740417, "global_step": 95319, "epoch": 1071, "lr": 7.270598161224144e-05}
{"train_loss": 0.22051343321800232, "global_step": 95320, "epoch": 1071, "lr": 7.270546509977064e-05}
{"train_loss": 0.15620602667331696, "global_step": 95321, "epoch": 1071, "lr": 7.270494858424737e-05}
{"train_loss": 0.15755850076675415, "global_step": 95322, "epoch": 1071, "lr": 7.270443206567168e-05}
{"train_loss": 0.254631906747818, "global_step": 95323, "epoch": 1071, "lr": 7.270391554404367e-05}
{"train_loss": 0.2523684501647949, "global_step": 95324, "epoch": 1071, "lr": 7.270339901936338e-05}
{"train_loss": 0.2295098751783371, "global_step": 95325, "epoch": 1071, "lr": 7.27028824916309e-05}
{"train_loss": 0.22327548265457153, "global_step": 95326, "epoch": 1071, "lr": 7.27023659608463e-05}
{"train_loss": 0.11056756973266602, "global_step": 95327, "epoch": 1071, "lr": 7.270184942700964e-05}
{"train_loss": 0.20310984551906586, "global_step": 95328, "epoch": 1071, "lr": 7.2701332890121e-05}
{"train_loss": 0.12842866778373718, "global_step": 95329, "epoch": 1071, "lr": 7.270081635018044e-05}
{"train_loss": 0.1264445036649704, "global_step": 95330, "epoch": 1071, "lr": 7.270029980718802e-05}
{"train_loss": 0.1813613325357437, "global_step": 95331, "epoch": 1071, "lr": 7.269978326114383e-05}
{"train_loss": 0.24855343997478485, "global_step": 95332, "epoch": 1071, "lr": 7.269926671204793e-05}
{"train_loss": 0.21634845435619354, "global_step": 95333, "epoch": 1071, "lr": 7.26987501599004e-05}
{"train_loss": 0.2279629111289978, "global_step": 95334, "epoch": 1071, "lr": 7.26982336047013e-05}
{"train_loss": 0.1283334493637085, "global_step": 95335, "epoch": 1071, "lr": 7.269771704645068e-05}
{"train_loss": 0.14327146112918854, "global_step": 95336, "epoch": 1071, "lr": 7.269720048514864e-05}
{"train_loss": 0.17439693212509155, "global_step": 95337, "epoch": 1071, "lr": 7.269668392079524e-05}
{"train_loss": 0.21568000316619873, "global_step": 95338, "epoch": 1071, "lr": 7.269616735339054e-05}
{"train_loss": 0.23862110078334808, "global_step": 95339, "epoch": 1071, "lr": 7.269565078293463e-05}
{"train_loss": 0.11036527901887894, "global_step": 95340, "epoch": 1071, "lr": 7.269513420942756e-05}
{"train_loss": 0.1922653764486313, "global_step": 95341, "epoch": 1071, "lr": 7.26946176328694e-05}
{"train_loss": 0.1493230164051056, "global_step": 95342, "epoch": 1071, "lr": 7.269410105326022e-05}
{"train_loss": 0.13130970299243927, "global_step": 95343, "epoch": 1071, "lr": 7.269358447060011e-05}
{"train_loss": 0.1420898586511612, "global_step": 95344, "epoch": 1071, "lr": 7.26930678848891e-05}
{"train_loss": 0.1922241747379303, "global_step": 95345, "epoch": 1071, "lr": 7.269255129612731e-05}
{"train_loss": 0.1331174373626709, "global_step": 95346, "epoch": 1071, "lr": 7.269203470431477e-05}
{"train_loss": 0.16712959110736847, "global_step": 95347, "epoch": 1071, "lr": 7.269151810945157e-05}
{"train_loss": 0.12809458374977112, "global_step": 95348, "epoch": 1071, "lr": 7.269100151153777e-05}
{"train_loss": 0.11431483179330826, "global_step": 95349, "epoch": 1071, "lr": 7.269048491057344e-05}
{"train_loss": 0.11562270671129227, "global_step": 95350, "epoch": 1071, "lr": 7.268996830655865e-05}
{"train_loss": 0.1693991720676422, "global_step": 95351, "epoch": 1071, "lr": 7.268945169949348e-05}
{"train_loss": 0.17866922914981842, "global_step": 95352, "epoch": 1071, "lr": 7.268893508937798e-05}
{"train_loss": 0.19109997153282166, "global_step": 95353, "epoch": 1071, "lr": 7.268841847621224e-05}
{"train_loss": 0.1788298785686493, "global_step": 95354, "epoch": 1071, "lr": 7.26879018599963e-05}
{"train_loss": 0.1620066612958908, "global_step": 95355, "epoch": 1071, "lr": 7.268738524073027e-05}
{"train_loss": 0.16918955743312836, "global_step": 95356, "epoch": 1071, "lr": 7.268686861841418e-05}
{"train_loss": 0.15154147148132324, "global_step": 95357, "epoch": 1071, "lr": 7.268635199304811e-05}
{"train_loss": 0.1647839993238449, "global_step": 95358, "epoch": 1071, "lr": 7.268583536463216e-05}
{"train_loss": 0.11297304928302765, "global_step": 95359, "epoch": 1071, "lr": 7.268531873316637e-05}
{"train_loss": 0.13515859842300415, "global_step": 95360, "epoch": 1071, "lr": 7.26848020986508e-05}
{"train_loss": 0.13279084861278534, "global_step": 95361, "epoch": 1071, "lr": 7.268428546108554e-05}
{"train_loss": 0.17957550287246704, "global_step": 95362, "epoch": 1071, "lr": 7.268376882047066e-05}
{"train_loss": 0.14988982677459717, "global_step": 95363, "epoch": 1071, "lr": 7.268325217680622e-05}
{"train_loss": 0.23524679243564606, "global_step": 95364, "epoch": 1071, "lr": 7.26827355300923e-05}
{"train_loss": 0.20939724147319794, "global_step": 95365, "epoch": 1071, "lr": 7.268221888032896e-05}
{"train_loss": 0.16235196590423584, "global_step": 95366, "epoch": 1071, "lr": 7.268170222751627e-05}
{"train_loss": 0.17217080295085907, "global_step": 95367, "epoch": 1071, "lr": 7.268118557165429e-05}
{"train_loss": 0.12362032383680344, "global_step": 95368, "epoch": 1071, "lr": 7.268066891274312e-05}
{"train_loss": 0.1737014204263687, "global_step": 95369, "epoch": 1071, "lr": 7.26801522507828e-05}
{"train_loss": 0.1690276712179184, "global_step": 95370, "epoch": 1071, "lr": 7.267963558577342e-05}
{"train_loss": 0.13264012336730957, "global_step": 95371, "epoch": 1071, "lr": 7.267911891771503e-05}
{"train_loss": 0.21316517889499664, "global_step": 95372, "epoch": 1071, "lr": 7.267860224660771e-05}
{"train_loss": 0.16876845061779022, "global_step": 95373, "epoch": 1071, "lr": 7.267808557245153e-05}
{"train_loss": 0.142486572265625, "global_step": 95374, "epoch": 1071, "lr": 7.267756889524657e-05}
{"train_loss": 0.17568513751029968, "global_step": 95375, "epoch": 1071, "lr": 7.267705221499288e-05}
{"train_loss": 0.2022022306919098, "global_step": 95376, "epoch": 1071, "lr": 7.267653553169052e-05}
{"train_loss": 0.17355546355247498, "global_step": 95377, "epoch": 1071, "lr": 7.267601884533959e-05}
{"train_loss": 0.1676214188337326, "global_step": 95378, "epoch": 1071, "lr": 7.267550215594014e-05}
{"train_loss": 0.1781514436006546, "global_step": 95379, "epoch": 1071, "lr": 7.267498546349225e-05}
{"train_loss": 0.17037725448608398, "global_step": 95380, "epoch": 1071, "lr": 7.267446876799598e-05}
{"train_loss": 0.16997332870960236, "global_step": 95381, "epoch": 1071, "lr": 7.267395206945141e-05}
{"train_loss": 0.22694113850593567, "global_step": 95382, "epoch": 1071, "lr": 7.26734353678586e-05}
{"train_loss": 0.18013069033622742, "global_step": 95383, "epoch": 1071, "lr": 7.267291866321763e-05}
{"train_loss": 0.1472303420305252, "global_step": 95384, "epoch": 1071, "lr": 7.267240195552855e-05}
{"train_loss": 0.13025325536727905, "global_step": 95385, "epoch": 1071, "lr": 7.267188524479145e-05}
{"train_loss": 0.1502823829650879, "global_step": 95386, "epoch": 1071, "lr": 7.26713685310064e-05}
{"train_loss": 0.16225531697273254, "global_step": 95387, "epoch": 1071, "lr": 7.267085181417343e-05}
{"train_loss": 0.17949794232845306, "global_step": 95388, "epoch": 1071, "lr": 7.267033509429268e-05}
{"train_loss": 0.20403242111206055, "global_step": 95389, "epoch": 1071, "lr": 7.266981837136416e-05}
{"train_loss": 0.1029597744345665, "global_step": 95390, "epoch": 1071, "lr": 7.266930164538795e-05}
{"train_loss": 0.10189585387706757, "global_step": 95391, "epoch": 1071, "lr": 7.266878491636414e-05}
{"train_loss": 0.14898626506328583, "global_step": 95392, "epoch": 1071, "lr": 7.26682681842928e-05}
{"train_loss": 0.1763523519039154, "global_step": 95393, "epoch": 1071, "lr": 7.266775144917398e-05}
{"train_loss": 0.14476338028907776, "global_step": 95394, "epoch": 1071, "lr": 7.266723471100774e-05}
{"train_loss": 0.2501296401023865, "global_step": 95395, "epoch": 1071, "lr": 7.266671796979418e-05}
{"train_loss": 0.1560172140598297, "global_step": 95396, "epoch": 1071, "lr": 7.266620122553336e-05}
{"train_loss": 0.15138143301010132, "global_step": 95397, "epoch": 1071, "lr": 7.266568447822533e-05}
{"train_loss": 0.15547795593738556, "global_step": 95398, "epoch": 1071, "lr": 7.266516772787018e-05}
{"train_loss": 0.17436325550079346, "global_step": 95399, "epoch": 1071, "lr": 7.266465097446797e-05}
{"train_loss": 0.2530020475387573, "global_step": 95400, "epoch": 1071, "lr": 7.266413421801879e-05}
{"train_loss": 0.2331482172012329, "global_step": 95401, "epoch": 1071, "lr": 7.26636174585227e-05}
{"train_loss": 0.13925734162330627, "global_step": 95402, "epoch": 1071, "lr": 7.266310069597973e-05}
{"train_loss": 0.21150635182857513, "global_step": 95403, "epoch": 1071, "lr": 7.266258393039e-05}
{"train_loss": 0.24845506250858307, "global_step": 95404, "epoch": 1071, "lr": 7.266206716175357e-05}
{"train_loss": 0.1713753342628479, "global_step": 95405, "epoch": 1071, "lr": 7.26615503900705e-05}
{"train_loss": 0.2389492392539978, "global_step": 95406, "epoch": 1071, "lr": 7.266103361534086e-05}
{"train_loss": 0.172398810976007, "global_step": 95407, "epoch": 1071, "lr": 7.26605168375647e-05, "val_loss": 4.561318397521973}
{"train_loss": 0.12095813453197479, "global_step": 95408, "epoch": 1072, "lr": 7.266000005674214e-05}
{"train_loss": 0.18050013482570648, "global_step": 95409, "epoch": 1072, "lr": 7.26594832728732e-05}
{"train_loss": 0.17038516700267792, "global_step": 95410, "epoch": 1072, "lr": 7.265896648595796e-05}
{"train_loss": 0.11075729131698608, "global_step": 95411, "epoch": 1072, "lr": 7.265844969599653e-05}
{"train_loss": 0.18476438522338867, "global_step": 95412, "epoch": 1072, "lr": 7.265793290298893e-05}
{"train_loss": 0.14897114038467407, "global_step": 95413, "epoch": 1072, "lr": 7.265741610693525e-05}
{"train_loss": 0.13635462522506714, "global_step": 95414, "epoch": 1072, "lr": 7.265689930783555e-05}
{"train_loss": 0.1442635953426361, "global_step": 95415, "epoch": 1072, "lr": 7.26563825056899e-05}
{"train_loss": 0.09314596652984619, "global_step": 95416, "epoch": 1072, "lr": 7.265586570049838e-05}
{"train_loss": 0.2035418301820755, "global_step": 95417, "epoch": 1072, "lr": 7.265534889226107e-05}
{"train_loss": 0.16703425347805023, "global_step": 95418, "epoch": 1072, "lr": 7.265483208097801e-05}
{"train_loss": 0.16698388755321503, "global_step": 95419, "epoch": 1072, "lr": 7.265431526664929e-05}
{"train_loss": 0.19482898712158203, "global_step": 95420, "epoch": 1072, "lr": 7.265379844927497e-05}
{"train_loss": 0.22235007584095, "global_step": 95421, "epoch": 1072, "lr": 7.265328162885512e-05}
{"train_loss": 0.1996060311794281, "global_step": 95422, "epoch": 1072, "lr": 7.265276480538982e-05}
{"train_loss": 0.19690671563148499, "global_step": 95423, "epoch": 1072, "lr": 7.265224797887913e-05}
{"train_loss": 0.22134828567504883, "global_step": 95424, "epoch": 1072, "lr": 7.265173114932311e-05}
{"train_loss": 0.11951738595962524, "global_step": 95425, "epoch": 1072, "lr": 7.265121431672186e-05}
{"train_loss": 0.18684230744838715, "global_step": 95426, "epoch": 1072, "lr": 7.265069748107541e-05}
{"train_loss": 0.12244413048028946, "global_step": 95427, "epoch": 1072, "lr": 7.265018064238387e-05}
{"train_loss": 0.17382366955280304, "global_step": 95428, "epoch": 1072, "lr": 7.264966380064728e-05}
{"train_loss": 0.1462768167257309, "global_step": 95429, "epoch": 1072, "lr": 7.26491469558657e-05}
{"train_loss": 0.17251673340797424, "global_step": 95430, "epoch": 1072, "lr": 7.264863010803925e-05}
{"train_loss": 0.18087045848369598, "global_step": 95431, "epoch": 1072, "lr": 7.264811325716797e-05}
{"train_loss": 0.21658360958099365, "global_step": 95432, "epoch": 1072, "lr": 7.264759640325191e-05}
{"train_loss": 0.19034343957901, "global_step": 95433, "epoch": 1072, "lr": 7.264707954629116e-05}
{"train_loss": 0.17053155601024628, "global_step": 95434, "epoch": 1072, "lr": 7.264656268628577e-05}
{"train_loss": 0.19415238499641418, "global_step": 95435, "epoch": 1072, "lr": 7.264604582323584e-05}
{"train_loss": 0.2345157116651535, "global_step": 95436, "epoch": 1072, "lr": 7.264552895714143e-05}
{"train_loss": 0.15817150473594666, "global_step": 95437, "epoch": 1072, "lr": 7.264501208800259e-05}
{"train_loss": 0.189301535487175, "global_step": 95438, "epoch": 1072, "lr": 7.264449521581942e-05}
{"train_loss": 0.2522532343864441, "global_step": 95439, "epoch": 1072, "lr": 7.264397834059197e-05}
{"train_loss": 0.16478902101516724, "global_step": 95440, "epoch": 1072, "lr": 7.26434614623203e-05}
{"train_loss": 0.11788169294595718, "global_step": 95441, "epoch": 1072, "lr": 7.264294458100452e-05}
{"train_loss": 0.15587833523750305, "global_step": 95442, "epoch": 1072, "lr": 7.264242769664465e-05}
{"train_loss": 0.2003777176141739, "global_step": 95443, "epoch": 1072, "lr": 7.264191080924078e-05}
{"train_loss": 0.09554774314165115, "global_step": 95444, "epoch": 1072, "lr": 7.264139391879298e-05}
{"train_loss": 0.1891043782234192, "global_step": 95445, "epoch": 1072, "lr": 7.264087702530134e-05}
{"train_loss": 0.21673284471035004, "global_step": 95446, "epoch": 1072, "lr": 7.264036012876589e-05}
{"train_loss": 0.16469894349575043, "global_step": 95447, "epoch": 1072, "lr": 7.263984322918673e-05}
{"train_loss": 0.10059249401092529, "global_step": 95448, "epoch": 1072, "lr": 7.263932632656392e-05}
{"train_loss": 0.18921774625778198, "global_step": 95449, "epoch": 1072, "lr": 7.263880942089754e-05}
{"train_loss": 0.14157308638095856, "global_step": 95450, "epoch": 1072, "lr": 7.263829251218764e-05}
{"train_loss": 0.10379059612751007, "global_step": 95451, "epoch": 1072, "lr": 7.263777560043429e-05}
{"train_loss": 0.16642417013645172, "global_step": 95452, "epoch": 1072, "lr": 7.263725868563756e-05}
{"train_loss": 0.15558664500713348, "global_step": 95453, "epoch": 1072, "lr": 7.263674176779754e-05}
{"train_loss": 0.1098514199256897, "global_step": 95454, "epoch": 1072, "lr": 7.263622484691427e-05}
{"train_loss": 0.12173309177160263, "global_step": 95455, "epoch": 1072, "lr": 7.263570792298785e-05}
{"train_loss": 0.16816626489162445, "global_step": 95456, "epoch": 1072, "lr": 7.263519099601833e-05}
{"train_loss": 0.1645982712507248, "global_step": 95457, "epoch": 1072, "lr": 7.263467406600578e-05}
{"train_loss": 0.20550990104675293, "global_step": 95458, "epoch": 1072, "lr": 7.263415713295028e-05}
{"train_loss": 0.18453359603881836, "global_step": 95459, "epoch": 1072, "lr": 7.263364019685189e-05}
{"train_loss": 0.18281154334545135, "global_step": 95460, "epoch": 1072, "lr": 7.263312325771069e-05}
{"train_loss": 0.13926030695438385, "global_step": 95461, "epoch": 1072, "lr": 7.263260631552674e-05}
{"train_loss": 0.11402804404497147, "global_step": 95462, "epoch": 1072, "lr": 7.26320893703001e-05}
{"train_loss": 0.16131919622421265, "global_step": 95463, "epoch": 1072, "lr": 7.263157242203087e-05}
{"train_loss": 0.166250541806221, "global_step": 95464, "epoch": 1072, "lr": 7.263105547071909e-05}
{"train_loss": 0.12612417340278625, "global_step": 95465, "epoch": 1072, "lr": 7.263053851636486e-05}
{"train_loss": 0.21304696798324585, "global_step": 95466, "epoch": 1072, "lr": 7.26300215589682e-05}
{"train_loss": 0.18158601224422455, "global_step": 95467, "epoch": 1072, "lr": 7.262950459852922e-05}
{"train_loss": 0.2022729218006134, "global_step": 95468, "epoch": 1072, "lr": 7.262898763504799e-05}
{"train_loss": 0.17515090107917786, "global_step": 95469, "epoch": 1072, "lr": 7.262847066852456e-05}
{"train_loss": 0.2238919734954834, "global_step": 95470, "epoch": 1072, "lr": 7.2627953698959e-05}
{"train_loss": 0.15491622686386108, "global_step": 95471, "epoch": 1072, "lr": 7.262743672635139e-05}
{"train_loss": 0.25431278347969055, "global_step": 95472, "epoch": 1072, "lr": 7.262691975070182e-05}
{"train_loss": 0.1788913905620575, "global_step": 95473, "epoch": 1072, "lr": 7.262640277201032e-05}
{"train_loss": 0.11220240592956543, "global_step": 95474, "epoch": 1072, "lr": 7.262588579027696e-05}
{"train_loss": 0.14935706555843353, "global_step": 95475, "epoch": 1072, "lr": 7.262536880550183e-05}
{"train_loss": 0.1273982673883438, "global_step": 95476, "epoch": 1072, "lr": 7.262485181768502e-05}
{"train_loss": 0.12943273782730103, "global_step": 95477, "epoch": 1072, "lr": 7.262433482682656e-05}
{"train_loss": 0.12683501839637756, "global_step": 95478, "epoch": 1072, "lr": 7.262381783292653e-05}
{"train_loss": 0.1288180947303772, "global_step": 95479, "epoch": 1072, "lr": 7.262330083598501e-05}
{"train_loss": 0.13016647100448608, "global_step": 95480, "epoch": 1072, "lr": 7.262278383600206e-05}
{"train_loss": 0.30453699827194214, "global_step": 95481, "epoch": 1072, "lr": 7.262226683297775e-05}
{"train_loss": 0.15958529710769653, "global_step": 95482, "epoch": 1072, "lr": 7.262174982691216e-05}
{"train_loss": 0.10966283082962036, "global_step": 95483, "epoch": 1072, "lr": 7.262123281780534e-05}
{"train_loss": 0.20321373641490936, "global_step": 95484, "epoch": 1072, "lr": 7.262071580565738e-05}
{"train_loss": 0.23039767146110535, "global_step": 95485, "epoch": 1072, "lr": 7.262019879046833e-05}
{"train_loss": 0.19501036405563354, "global_step": 95486, "epoch": 1072, "lr": 7.261968177223828e-05}
{"train_loss": 0.12089941650629044, "global_step": 95487, "epoch": 1072, "lr": 7.261916475096728e-05}
{"train_loss": 0.250102162361145, "global_step": 95488, "epoch": 1072, "lr": 7.261864772665542e-05}
{"train_loss": 0.10728664696216583, "global_step": 95489, "epoch": 1072, "lr": 7.261813069930275e-05}
{"train_loss": 0.1413750946521759, "global_step": 95490, "epoch": 1072, "lr": 7.261761366890936e-05}
{"train_loss": 0.22404631972312927, "global_step": 95491, "epoch": 1072, "lr": 7.26170966354753e-05}
{"train_loss": 0.1262364387512207, "global_step": 95492, "epoch": 1072, "lr": 7.261657959900064e-05}
{"train_loss": 0.18085061013698578, "global_step": 95493, "epoch": 1072, "lr": 7.261606255948546e-05}
{"train_loss": 0.18098925054073334, "global_step": 95494, "epoch": 1072, "lr": 7.261554551692984e-05}
{"train_loss": 0.16431525349617004, "global_step": 95495, "epoch": 1072, "lr": 7.26150284713338e-05}
{"train_loss": 0.167893588710367, "global_step": 95496, "epoch": 1072, "lr": 7.261451142269748e-05, "val_loss": 4.667893886566162}
{"train_loss": 0.15996596217155457, "global_step": 95497, "epoch": 1073, "lr": 7.261399437102091e-05}
{"train_loss": 0.24608981609344482, "global_step": 95498, "epoch": 1073, "lr": 7.261347731630416e-05}
{"train_loss": 0.17955100536346436, "global_step": 95499, "epoch": 1073, "lr": 7.261296025854731e-05}
{"train_loss": 0.15485262870788574, "global_step": 95500, "epoch": 1073, "lr": 7.26124431977504e-05}
{"train_loss": 0.14765852689743042, "global_step": 95501, "epoch": 1073, "lr": 7.261192613391355e-05}
{"train_loss": 0.08708221465349197, "global_step": 95502, "epoch": 1073, "lr": 7.261140906703678e-05}
{"train_loss": 0.18342764675617218, "global_step": 95503, "epoch": 1073, "lr": 7.26108919971202e-05}
{"train_loss": 0.18944156169891357, "global_step": 95504, "epoch": 1073, "lr": 7.261037492416384e-05}
{"train_loss": 0.1403670310974121, "global_step": 95505, "epoch": 1073, "lr": 7.260985784816781e-05}
{"train_loss": 0.16978919506072998, "global_step": 95506, "epoch": 1073, "lr": 7.260934076913215e-05}
{"train_loss": 0.18440568447113037, "global_step": 95507, "epoch": 1073, "lr": 7.260882368705695e-05}
{"train_loss": 0.21978837251663208, "global_step": 95508, "epoch": 1073, "lr": 7.260830660194226e-05}
{"train_loss": 0.09562879800796509, "global_step": 95509, "epoch": 1073, "lr": 7.260778951378816e-05}
{"train_loss": 0.1379982978105545, "global_step": 95510, "epoch": 1073, "lr": 7.260727242259471e-05}
{"train_loss": 0.17353984713554382, "global_step": 95511, "epoch": 1073, "lr": 7.2606755328362e-05}
{"train_loss": 0.17588841915130615, "global_step": 95512, "epoch": 1073, "lr": 7.260623823109009e-05}
{"train_loss": 0.17974743247032166, "global_step": 95513, "epoch": 1073, "lr": 7.260572113077903e-05}
{"train_loss": 0.15058179199695587, "global_step": 95514, "epoch": 1073, "lr": 7.260520402742893e-05}
{"train_loss": 0.13124091923236847, "global_step": 95515, "epoch": 1073, "lr": 7.260468692103982e-05}
{"train_loss": 0.1802523136138916, "global_step": 95516, "epoch": 1073, "lr": 7.260416981161179e-05}
{"train_loss": 0.19649235904216766, "global_step": 95517, "epoch": 1073, "lr": 7.26036526991449e-05}
{"train_loss": 0.12083442509174347, "global_step": 95518, "epoch": 1073, "lr": 7.260313558363923e-05}
{"train_loss": 0.17176608741283417, "global_step": 95519, "epoch": 1073, "lr": 7.260261846509485e-05}
{"train_loss": 0.14495885372161865, "global_step": 95520, "epoch": 1073, "lr": 7.260210134351181e-05}
{"train_loss": 0.22733405232429504, "global_step": 95521, "epoch": 1073, "lr": 7.260158421889019e-05}
{"train_loss": 0.1591235250234604, "global_step": 95522, "epoch": 1073, "lr": 7.260106709123008e-05}
{"train_loss": 0.12348701804876328, "global_step": 95523, "epoch": 1073, "lr": 7.260054996053152e-05}
{"train_loss": 0.13762584328651428, "global_step": 95524, "epoch": 1073, "lr": 7.260003282679459e-05}
{"train_loss": 0.16684028506278992, "global_step": 95525, "epoch": 1073, "lr": 7.259951569001937e-05}
{"train_loss": 0.06766857951879501, "global_step": 95526, "epoch": 1073, "lr": 7.259899855020591e-05}
{"train_loss": 0.2619219422340393, "global_step": 95527, "epoch": 1073, "lr": 7.25984814073543e-05}
{"train_loss": 0.2136702835559845, "global_step": 95528, "epoch": 1073, "lr": 7.259796426146459e-05}
{"train_loss": 0.14877554774284363, "global_step": 95529, "epoch": 1073, "lr": 7.259744711253686e-05}
{"train_loss": 0.159125417470932, "global_step": 95530, "epoch": 1073, "lr": 7.259692996057119e-05}
{"train_loss": 0.13987880945205688, "global_step": 95531, "epoch": 1073, "lr": 7.259641280556764e-05}
{"train_loss": 0.15393738448619843, "global_step": 95532, "epoch": 1073, "lr": 7.259589564752627e-05}
{"train_loss": 0.16081565618515015, "global_step": 95533, "epoch": 1073, "lr": 7.259537848644716e-05}
{"train_loss": 0.15282465517520905, "global_step": 95534, "epoch": 1073, "lr": 7.259486132233037e-05}
{"train_loss": 0.1177397072315216, "global_step": 95535, "epoch": 1073, "lr": 7.259434415517597e-05}
{"train_loss": 0.2213256061077118, "global_step": 95536, "epoch": 1073, "lr": 7.259382698498405e-05}
{"train_loss": 0.22302918136119843, "global_step": 95537, "epoch": 1073, "lr": 7.259330981175465e-05}
{"train_loss": 0.2564103305339813, "global_step": 95538, "epoch": 1073, "lr": 7.259279263548787e-05}
{"train_loss": 0.3411216735839844, "global_step": 95539, "epoch": 1073, "lr": 7.259227545618377e-05}
{"train_loss": 0.26847249269485474, "global_step": 95540, "epoch": 1073, "lr": 7.25917582738424e-05}
{"train_loss": 0.10479436069726944, "global_step": 95541, "epoch": 1073, "lr": 7.259124108846385e-05}
{"train_loss": 0.18326279520988464, "global_step": 95542, "epoch": 1073, "lr": 7.259072390004818e-05}
{"train_loss": 0.19900327920913696, "global_step": 95543, "epoch": 1073, "lr": 7.259020670859545e-05}
{"train_loss": 0.10918024182319641, "global_step": 95544, "epoch": 1073, "lr": 7.258968951410576e-05}
{"train_loss": 0.1616320163011551, "global_step": 95545, "epoch": 1073, "lr": 7.258917231657916e-05}
{"train_loss": 0.17815636098384857, "global_step": 95546, "epoch": 1073, "lr": 7.258865511601572e-05}
{"train_loss": 0.2772824764251709, "global_step": 95547, "epoch": 1073, "lr": 7.25881379124155e-05}
{"train_loss": 0.24373120069503784, "global_step": 95548, "epoch": 1073, "lr": 7.258762070577859e-05}
{"train_loss": 0.22027285397052765, "global_step": 95549, "epoch": 1073, "lr": 7.258710349610505e-05}
{"train_loss": 0.19325000047683716, "global_step": 95550, "epoch": 1073, "lr": 7.258658628339495e-05}
{"train_loss": 0.17363855242729187, "global_step": 95551, "epoch": 1073, "lr": 7.258606906764836e-05}
{"train_loss": 0.2555777132511139, "global_step": 95552, "epoch": 1073, "lr": 7.258555184886533e-05}
{"train_loss": 0.17775045335292816, "global_step": 95553, "epoch": 1073, "lr": 7.258503462704597e-05}
{"train_loss": 0.09413181245326996, "global_step": 95554, "epoch": 1073, "lr": 7.258451740219033e-05}
{"train_loss": 0.21085073053836823, "global_step": 95555, "epoch": 1073, "lr": 7.258400017429845e-05}
{"train_loss": 0.12576402723789215, "global_step": 95556, "epoch": 1073, "lr": 7.258348294337045e-05}
{"train_loss": 0.21676959097385406, "global_step": 95557, "epoch": 1073, "lr": 7.258296570940637e-05}
{"train_loss": 0.1486629992723465, "global_step": 95558, "epoch": 1073, "lr": 7.258244847240628e-05}
{"train_loss": 0.1800631731748581, "global_step": 95559, "epoch": 1073, "lr": 7.258193123237026e-05}
{"train_loss": 0.19491247832775116, "global_step": 95560, "epoch": 1073, "lr": 7.258141398929837e-05}
{"train_loss": 0.16444581747055054, "global_step": 95561, "epoch": 1073, "lr": 7.258089674319069e-05}
{"train_loss": 0.1439639925956726, "global_step": 95562, "epoch": 1073, "lr": 7.258037949404728e-05}
{"train_loss": 0.14957265555858612, "global_step": 95563, "epoch": 1073, "lr": 7.257986224186821e-05}
{"train_loss": 0.19355133175849915, "global_step": 95564, "epoch": 1073, "lr": 7.257934498665357e-05}
{"train_loss": 0.26783227920532227, "global_step": 95565, "epoch": 1073, "lr": 7.257882772840339e-05}
{"train_loss": 0.12569111585617065, "global_step": 95566, "epoch": 1073, "lr": 7.257831046711779e-05}
{"train_loss": 0.17778442800045013, "global_step": 95567, "epoch": 1073, "lr": 7.257779320279679e-05}
{"train_loss": 0.11824321746826172, "global_step": 95568, "epoch": 1073, "lr": 7.257727593544046e-05}
{"train_loss": 0.22749219834804535, "global_step": 95569, "epoch": 1073, "lr": 7.257675866504893e-05}
{"train_loss": 0.2670621871948242, "global_step": 95570, "epoch": 1073, "lr": 7.257624139162222e-05}
{"train_loss": 0.18489886820316315, "global_step": 95571, "epoch": 1073, "lr": 7.257572411516039e-05}
{"train_loss": 0.14222560822963715, "global_step": 95572, "epoch": 1073, "lr": 7.257520683566355e-05}
{"train_loss": 0.2669035792350769, "global_step": 95573, "epoch": 1073, "lr": 7.257468955313173e-05}
{"train_loss": 0.20665085315704346, "global_step": 95574, "epoch": 1073, "lr": 7.257417226756504e-05}
{"train_loss": 0.14677895605564117, "global_step": 95575, "epoch": 1073, "lr": 7.25736549789635e-05}
{"train_loss": 0.2248293161392212, "global_step": 95576, "epoch": 1073, "lr": 7.257313768732724e-05}
{"train_loss": 0.129185751080513, "global_step": 95577, "epoch": 1073, "lr": 7.257262039265628e-05}
{"train_loss": 0.20376726984977722, "global_step": 95578, "epoch": 1073, "lr": 7.25721030949507e-05}
{"train_loss": 0.26346614956855774, "global_step": 95579, "epoch": 1073, "lr": 7.257158579421058e-05}
{"train_loss": 0.15772780776023865, "global_step": 95580, "epoch": 1073, "lr": 7.2571068490436e-05}
{"train_loss": 0.24726371467113495, "global_step": 95581, "epoch": 1073, "lr": 7.257055118362701e-05}
{"train_loss": 0.11509151011705399, "global_step": 95582, "epoch": 1073, "lr": 7.257003387378367e-05}
{"train_loss": 0.14238956570625305, "global_step": 95583, "epoch": 1073, "lr": 7.256951656090607e-05}
{"train_loss": 0.14676062762737274, "global_step": 95584, "epoch": 1073, "lr": 7.256899924499428e-05}
{"train_loss": 0.17879200124003913, "global_step": 95585, "epoch": 1073, "lr": 7.256848192604836e-05, "val_loss": 4.5448126792907715}
{"train_loss": 0.19687527418136597, "global_step": 95586, "epoch": 1074, "lr": 7.256796460406839e-05}
{"train_loss": 0.1343604475259781, "global_step": 95587, "epoch": 1074, "lr": 7.256744727905442e-05}
{"train_loss": 0.11808163672685623, "global_step": 95588, "epoch": 1074, "lr": 7.256692995100654e-05}
{"train_loss": 0.15006418526172638, "global_step": 95589, "epoch": 1074, "lr": 7.256641261992481e-05}
{"train_loss": 0.16612035036087036, "global_step": 95590, "epoch": 1074, "lr": 7.256589528580928e-05}
{"train_loss": 0.1600922793149948, "global_step": 95591, "epoch": 1074, "lr": 7.256537794866006e-05}
{"train_loss": 0.19863468408584595, "global_step": 95592, "epoch": 1074, "lr": 7.25648606084772e-05}
{"train_loss": 0.2714102268218994, "global_step": 95593, "epoch": 1074, "lr": 7.256434326526077e-05}
{"train_loss": 0.17120559513568878, "global_step": 95594, "epoch": 1074, "lr": 7.256382591901084e-05}
{"train_loss": 0.14001436531543732, "global_step": 95595, "epoch": 1074, "lr": 7.256330856972747e-05}
{"train_loss": 0.18442629277706146, "global_step": 95596, "epoch": 1074, "lr": 7.256279121741074e-05}
{"train_loss": 0.1621415615081787, "global_step": 95597, "epoch": 1074, "lr": 7.256227386206072e-05}
{"train_loss": 0.25631576776504517, "global_step": 95598, "epoch": 1074, "lr": 7.256175650367748e-05}
{"train_loss": 0.1468629390001297, "global_step": 95599, "epoch": 1074, "lr": 7.256123914226109e-05}
{"train_loss": 0.19155287742614746, "global_step": 95600, "epoch": 1074, "lr": 7.256072177781161e-05}
{"train_loss": 0.15255720913410187, "global_step": 95601, "epoch": 1074, "lr": 7.256020441032913e-05}
{"train_loss": 0.23165881633758545, "global_step": 95602, "epoch": 1074, "lr": 7.255968703981368e-05}
{"train_loss": 0.18315576016902924, "global_step": 95603, "epoch": 1074, "lr": 7.255916966626537e-05}
{"train_loss": 0.18821528553962708, "global_step": 95604, "epoch": 1074, "lr": 7.255865228968425e-05}
{"train_loss": 0.11945585161447525, "global_step": 95605, "epoch": 1074, "lr": 7.255813491007041e-05}
{"train_loss": 0.1768593043088913, "global_step": 95606, "epoch": 1074, "lr": 7.255761752742388e-05}
{"train_loss": 0.1633564680814743, "global_step": 95607, "epoch": 1074, "lr": 7.255710014174477e-05}
{"train_loss": 0.11580116301774979, "global_step": 95608, "epoch": 1074, "lr": 7.255658275303311e-05}
{"train_loss": 0.20162388682365417, "global_step": 95609, "epoch": 1074, "lr": 7.255606536128903e-05}
{"train_loss": 0.1959100365638733, "global_step": 95610, "epoch": 1074, "lr": 7.255554796651254e-05}
{"train_loss": 0.17924126982688904, "global_step": 95611, "epoch": 1074, "lr": 7.255503056870374e-05}
{"train_loss": 0.17237326502799988, "global_step": 95612, "epoch": 1074, "lr": 7.255451316786267e-05}
{"train_loss": 0.25445830821990967, "global_step": 95613, "epoch": 1074, "lr": 7.255399576398943e-05}
{"train_loss": 0.1788838803768158, "global_step": 95614, "epoch": 1074, "lr": 7.25534783570841e-05}
{"train_loss": 0.1842094212770462, "global_step": 95615, "epoch": 1074, "lr": 7.25529609471467e-05}
{"train_loss": 0.24629199504852295, "global_step": 95616, "epoch": 1074, "lr": 7.255244353417734e-05}
{"train_loss": 0.1624661535024643, "global_step": 95617, "epoch": 1074, "lr": 7.255192611817609e-05}
{"train_loss": 0.1371796727180481, "global_step": 95618, "epoch": 1074, "lr": 7.255140869914299e-05}
{"train_loss": 0.171123668551445, "global_step": 95619, "epoch": 1074, "lr": 7.255089127707814e-05}
{"train_loss": 0.11709639430046082, "global_step": 95620, "epoch": 1074, "lr": 7.25503738519816e-05}
{"train_loss": 0.15005484223365784, "global_step": 95621, "epoch": 1074, "lr": 7.254985642385344e-05}
{"train_loss": 0.1646287441253662, "global_step": 95622, "epoch": 1074, "lr": 7.254933899269371e-05}
{"train_loss": 0.1346559226512909, "global_step": 95623, "epoch": 1074, "lr": 7.254882155850251e-05}
{"train_loss": 0.18435746431350708, "global_step": 95624, "epoch": 1074, "lr": 7.25483041212799e-05}
{"train_loss": 0.27372199296951294, "global_step": 95625, "epoch": 1074, "lr": 7.254778668102593e-05}
{"train_loss": 0.16446241736412048, "global_step": 95626, "epoch": 1074, "lr": 7.25472692377407e-05}
{"train_loss": 0.12747977674007416, "global_step": 95627, "epoch": 1074, "lr": 7.254675179142425e-05}
{"train_loss": 0.1761729121208191, "global_step": 95628, "epoch": 1074, "lr": 7.254623434207667e-05}
{"train_loss": 0.16010698676109314, "global_step": 95629, "epoch": 1074, "lr": 7.254571688969803e-05}
{"train_loss": 0.1614907830953598, "global_step": 95630, "epoch": 1074, "lr": 7.25451994342884e-05}
{"train_loss": 0.20618219673633575, "global_step": 95631, "epoch": 1074, "lr": 7.254468197584781e-05}
{"train_loss": 0.1613703817129135, "global_step": 95632, "epoch": 1074, "lr": 7.25441645143764e-05}
{"train_loss": 0.09159564971923828, "global_step": 95633, "epoch": 1074, "lr": 7.254364704987417e-05}
{"train_loss": 0.18751947581768036, "global_step": 95634, "epoch": 1074, "lr": 7.254312958234124e-05}
{"train_loss": 0.14123062789440155, "global_step": 95635, "epoch": 1074, "lr": 7.254261211177764e-05}
{"train_loss": 0.20472334325313568, "global_step": 95636, "epoch": 1074, "lr": 7.254209463818349e-05}
{"train_loss": 0.15893292427062988, "global_step": 95637, "epoch": 1074, "lr": 7.25415771615588e-05}
{"train_loss": 0.18864907324314117, "global_step": 95638, "epoch": 1074, "lr": 7.25410596819037e-05}
{"train_loss": 0.1979120522737503, "global_step": 95639, "epoch": 1074, "lr": 7.254054219921821e-05}
{"train_loss": 0.17625457048416138, "global_step": 95640, "epoch": 1074, "lr": 7.254002471350244e-05}
{"train_loss": 0.23214435577392578, "global_step": 95641, "epoch": 1074, "lr": 7.253950722475642e-05}
{"train_loss": 0.16133743524551392, "global_step": 95642, "epoch": 1074, "lr": 7.253898973298024e-05}
{"train_loss": 0.15138226747512817, "global_step": 95643, "epoch": 1074, "lr": 7.253847223817397e-05}
{"train_loss": 0.2056064009666443, "global_step": 95644, "epoch": 1074, "lr": 7.253795474033768e-05}
{"train_loss": 0.18208816647529602, "global_step": 95645, "epoch": 1074, "lr": 7.253743723947142e-05}
{"train_loss": 0.13924002647399902, "global_step": 95646, "epoch": 1074, "lr": 7.253691973557531e-05}
{"train_loss": 0.18995243310928345, "global_step": 95647, "epoch": 1074, "lr": 7.253640222864935e-05}
{"train_loss": 0.24721452593803406, "global_step": 95648, "epoch": 1074, "lr": 7.253588471869366e-05}
{"train_loss": 0.15138600766658783, "global_step": 95649, "epoch": 1074, "lr": 7.25353672057083e-05}
{"train_loss": 0.1249844953417778, "global_step": 95650, "epoch": 1074, "lr": 7.253484968969333e-05}
{"train_loss": 0.21396611630916595, "global_step": 95651, "epoch": 1074, "lr": 7.253433217064881e-05}
{"train_loss": 0.16306337714195251, "global_step": 95652, "epoch": 1074, "lr": 7.253381464857484e-05}
{"train_loss": 0.19545277953147888, "global_step": 95653, "epoch": 1074, "lr": 7.253329712347148e-05}
{"train_loss": 0.14622074365615845, "global_step": 95654, "epoch": 1074, "lr": 7.253277959533877e-05}
{"train_loss": 0.1386445164680481, "global_step": 95655, "epoch": 1074, "lr": 7.253226206417683e-05}
{"train_loss": 0.17289502918720245, "global_step": 95656, "epoch": 1074, "lr": 7.253174452998568e-05}
{"train_loss": 0.1495533585548401, "global_step": 95657, "epoch": 1074, "lr": 7.253122699276542e-05}
{"train_loss": 0.24789048731327057, "global_step": 95658, "epoch": 1074, "lr": 7.253070945251611e-05}
{"train_loss": 0.11689203232526779, "global_step": 95659, "epoch": 1074, "lr": 7.253019190923783e-05}
{"train_loss": 0.21290121972560883, "global_step": 95660, "epoch": 1074, "lr": 7.252967436293063e-05}
{"train_loss": 0.18510867655277252, "global_step": 95661, "epoch": 1074, "lr": 7.25291568135946e-05}
{"train_loss": 0.17945823073387146, "global_step": 95662, "epoch": 1074, "lr": 7.252863926122979e-05}
{"train_loss": 0.17484505474567413, "global_step": 95663, "epoch": 1074, "lr": 7.252812170583627e-05}
{"train_loss": 0.09501461684703827, "global_step": 95664, "epoch": 1074, "lr": 7.252760414741414e-05}
{"train_loss": 0.13289998471736908, "global_step": 95665, "epoch": 1074, "lr": 7.252708658596343e-05}
{"train_loss": 0.15593229234218597, "global_step": 95666, "epoch": 1074, "lr": 7.252656902148423e-05}
{"train_loss": 0.17085061967372894, "global_step": 95667, "epoch": 1074, "lr": 7.252605145397662e-05}
{"train_loss": 0.2013009786605835, "global_step": 95668, "epoch": 1074, "lr": 7.252553388344066e-05}
{"train_loss": 0.14355556666851044, "global_step": 95669, "epoch": 1074, "lr": 7.252501630987638e-05}
{"train_loss": 0.20485258102416992, "global_step": 95670, "epoch": 1074, "lr": 7.252449873328392e-05}
{"train_loss": 0.16652849316596985, "global_step": 95671, "epoch": 1074, "lr": 7.25239811536633e-05}
{"train_loss": 0.15652193129062653, "global_step": 95672, "epoch": 1074, "lr": 7.252346357101462e-05}
{"train_loss": 0.16150006651878357, "global_step": 95673, "epoch": 1074, "lr": 7.252294598533794e-05}
{"train_loss": 0.17360250007235603, "global_step": 95674, "epoch": 1074, "lr": 7.25224283966333e-05, "val_loss": 4.559043884277344}
{"train_loss": 0.15316009521484375, "global_step": 95675, "epoch": 1075, "lr": 7.252191080490081e-05}
{"train_loss": 0.18300746381282806, "global_step": 95676, "epoch": 1075, "lr": 7.25213932101405e-05}
{"train_loss": 0.16722320020198822, "global_step": 95677, "epoch": 1075, "lr": 7.25208756123525e-05}
{"train_loss": 0.17404071986675262, "global_step": 95678, "epoch": 1075, "lr": 7.252035801153683e-05}
{"train_loss": 0.12470470368862152, "global_step": 95679, "epoch": 1075, "lr": 7.251984040769358e-05}
{"train_loss": 0.19436611235141754, "global_step": 95680, "epoch": 1075, "lr": 7.251932280082279e-05}
{"train_loss": 0.14835457503795624, "global_step": 95681, "epoch": 1075, "lr": 7.251880519092455e-05}
{"train_loss": 0.2248174548149109, "global_step": 95682, "epoch": 1075, "lr": 7.251828757799896e-05}
{"train_loss": 0.14305363595485687, "global_step": 95683, "epoch": 1075, "lr": 7.251776996204604e-05}
{"train_loss": 0.23243682086467743, "global_step": 95684, "epoch": 1075, "lr": 7.251725234306587e-05}
{"train_loss": 0.12480639666318893, "global_step": 95685, "epoch": 1075, "lr": 7.251673472105855e-05}
{"train_loss": 0.1630220115184784, "global_step": 95686, "epoch": 1075, "lr": 7.251621709602414e-05}
{"train_loss": 0.13175331056118011, "global_step": 95687, "epoch": 1075, "lr": 7.251569946796267e-05}
{"train_loss": 0.14676551520824432, "global_step": 95688, "epoch": 1075, "lr": 7.251518183687423e-05}
{"train_loss": 0.186295747756958, "global_step": 95689, "epoch": 1075, "lr": 7.251466420275893e-05}
{"train_loss": 0.2253902405500412, "global_step": 95690, "epoch": 1075, "lr": 7.25141465656168e-05}
{"train_loss": 0.2180289626121521, "global_step": 95691, "epoch": 1075, "lr": 7.251362892544791e-05}
{"train_loss": 0.15172450244426727, "global_step": 95692, "epoch": 1075, "lr": 7.251311128225234e-05}
{"train_loss": 0.18255986273288727, "global_step": 95693, "epoch": 1075, "lr": 7.251259363603016e-05}
{"train_loss": 0.1328829526901245, "global_step": 95694, "epoch": 1075, "lr": 7.251207598678143e-05}
{"train_loss": 0.1961624175310135, "global_step": 95695, "epoch": 1075, "lr": 7.251155833450623e-05}
{"train_loss": 0.19281361997127533, "global_step": 95696, "epoch": 1075, "lr": 7.251104067920463e-05}
{"train_loss": 0.1540997475385666, "global_step": 95697, "epoch": 1075, "lr": 7.25105230208767e-05}
{"train_loss": 0.12892821431159973, "global_step": 95698, "epoch": 1075, "lr": 7.251000535952249e-05}
{"train_loss": 0.09747020900249481, "global_step": 95699, "epoch": 1075, "lr": 7.250948769514209e-05}
{"train_loss": 0.150457501411438, "global_step": 95700, "epoch": 1075, "lr": 7.250897002773556e-05}
{"train_loss": 0.17217645049095154, "global_step": 95701, "epoch": 1075, "lr": 7.250845235730297e-05}
{"train_loss": 0.21157033741474152, "global_step": 95702, "epoch": 1075, "lr": 7.25079346838444e-05}
{"train_loss": 0.2300247997045517, "global_step": 95703, "epoch": 1075, "lr": 7.25074170073599e-05}
{"train_loss": 0.18028555810451508, "global_step": 95704, "epoch": 1075, "lr": 7.250689932784958e-05}
{"train_loss": 0.13472720980644226, "global_step": 95705, "epoch": 1075, "lr": 7.250638164531346e-05}
{"train_loss": 0.1844533532857895, "global_step": 95706, "epoch": 1075, "lr": 7.250586395975162e-05}
{"train_loss": 0.15964269638061523, "global_step": 95707, "epoch": 1075, "lr": 7.250534627116417e-05}
{"train_loss": 0.13342037796974182, "global_step": 95708, "epoch": 1075, "lr": 7.250482857955113e-05}
{"train_loss": 0.20347003638744354, "global_step": 95709, "epoch": 1075, "lr": 7.250431088491259e-05}
{"train_loss": 0.14233164489269257, "global_step": 95710, "epoch": 1075, "lr": 7.250379318724864e-05}
{"train_loss": 0.21105670928955078, "global_step": 95711, "epoch": 1075, "lr": 7.250327548655931e-05}
{"train_loss": 0.23033246397972107, "global_step": 95712, "epoch": 1075, "lr": 7.25027577828447e-05}
{"train_loss": 0.2080022692680359, "global_step": 95713, "epoch": 1075, "lr": 7.250224007610486e-05}
{"train_loss": 0.1591072678565979, "global_step": 95714, "epoch": 1075, "lr": 7.250172236633987e-05}
{"train_loss": 0.16331206262111664, "global_step": 95715, "epoch": 1075, "lr": 7.25012046535498e-05}
{"train_loss": 0.21546122431755066, "global_step": 95716, "epoch": 1075, "lr": 7.250068693773474e-05}
{"train_loss": 0.2188279628753662, "global_step": 95717, "epoch": 1075, "lr": 7.250016921889472e-05}
{"train_loss": 0.16525550186634064, "global_step": 95718, "epoch": 1075, "lr": 7.249965149702981e-05}
{"train_loss": 0.20126540958881378, "global_step": 95719, "epoch": 1075, "lr": 7.24991337721401e-05}
{"train_loss": 0.10918103903532028, "global_step": 95720, "epoch": 1075, "lr": 7.249861604422566e-05}
{"train_loss": 0.17794430255889893, "global_step": 95721, "epoch": 1075, "lr": 7.249809831328656e-05}
{"train_loss": 0.1109793484210968, "global_step": 95722, "epoch": 1075, "lr": 7.249758057932287e-05}
{"train_loss": 0.14924293756484985, "global_step": 95723, "epoch": 1075, "lr": 7.249706284233464e-05}
{"train_loss": 0.1343398541212082, "global_step": 95724, "epoch": 1075, "lr": 7.249654510232196e-05}
{"train_loss": 0.13682293891906738, "global_step": 95725, "epoch": 1075, "lr": 7.249602735928489e-05}
{"train_loss": 0.1560353934764862, "global_step": 95726, "epoch": 1075, "lr": 7.24955096132235e-05}
{"train_loss": 0.15590715408325195, "global_step": 95727, "epoch": 1075, "lr": 7.249499186413787e-05}
{"train_loss": 0.19268648326396942, "global_step": 95728, "epoch": 1075, "lr": 7.249447411202807e-05}
{"train_loss": 0.10250595957040787, "global_step": 95729, "epoch": 1075, "lr": 7.249395635689416e-05}
{"train_loss": 0.16863089799880981, "global_step": 95730, "epoch": 1075, "lr": 7.24934385987362e-05}
{"train_loss": 0.14904139935970306, "global_step": 95731, "epoch": 1075, "lr": 7.249292083755428e-05}
{"train_loss": 0.20062106847763062, "global_step": 95732, "epoch": 1075, "lr": 7.249240307334846e-05}
{"train_loss": 0.11794929206371307, "global_step": 95733, "epoch": 1075, "lr": 7.249188530611881e-05}
{"train_loss": 0.16694357991218567, "global_step": 95734, "epoch": 1075, "lr": 7.24913675358654e-05}
{"train_loss": 0.11482164263725281, "global_step": 95735, "epoch": 1075, "lr": 7.24908497625883e-05}
{"train_loss": 0.2187611162662506, "global_step": 95736, "epoch": 1075, "lr": 7.249033198628759e-05}
{"train_loss": 0.1709975302219391, "global_step": 95737, "epoch": 1075, "lr": 7.248981420696332e-05}
{"train_loss": 0.13117757439613342, "global_step": 95738, "epoch": 1075, "lr": 7.248929642461556e-05}
{"train_loss": 0.1517219841480255, "global_step": 95739, "epoch": 1075, "lr": 7.248877863924439e-05}
{"train_loss": 0.13699866831302643, "global_step": 95740, "epoch": 1075, "lr": 7.248826085084989e-05}
{"train_loss": 0.1514066755771637, "global_step": 95741, "epoch": 1075, "lr": 7.248774305943211e-05}
{"train_loss": 0.1586260050535202, "global_step": 95742, "epoch": 1075, "lr": 7.248722526499112e-05}
{"train_loss": 0.1319975107908249, "global_step": 95743, "epoch": 1075, "lr": 7.248670746752701e-05}
{"train_loss": 0.23196806013584137, "global_step": 95744, "epoch": 1075, "lr": 7.248618966703982e-05}
{"train_loss": 0.1857820600271225, "global_step": 95745, "epoch": 1075, "lr": 7.248567186352965e-05}
{"train_loss": 0.13474801182746887, "global_step": 95746, "epoch": 1075, "lr": 7.248515405699655e-05}
{"train_loss": 0.2164570540189743, "global_step": 95747, "epoch": 1075, "lr": 7.248463624744059e-05}
{"train_loss": 0.16737833619117737, "global_step": 95748, "epoch": 1075, "lr": 7.248411843486186e-05}
{"train_loss": 0.11590057611465454, "global_step": 95749, "epoch": 1075, "lr": 7.24836006192604e-05}
{"train_loss": 0.1867026686668396, "global_step": 95750, "epoch": 1075, "lr": 7.24830828006363e-05}
{"train_loss": 0.12727119028568268, "global_step": 95751, "epoch": 1075, "lr": 7.248256497898962e-05}
{"train_loss": 0.1191113293170929, "global_step": 95752, "epoch": 1075, "lr": 7.248204715432045e-05}
{"train_loss": 0.1619531363248825, "global_step": 95753, "epoch": 1075, "lr": 7.248152932662882e-05}
{"train_loss": 0.09617311507463455, "global_step": 95754, "epoch": 1075, "lr": 7.248101149591484e-05}
{"train_loss": 0.18118435144424438, "global_step": 95755, "epoch": 1075, "lr": 7.248049366217856e-05}
{"train_loss": 0.1738543063402176, "global_step": 95756, "epoch": 1075, "lr": 7.247997582542004e-05}
{"train_loss": 0.2001274824142456, "global_step": 95757, "epoch": 1075, "lr": 7.247945798563937e-05}
{"train_loss": 0.14335443079471588, "global_step": 95758, "epoch": 1075, "lr": 7.247894014283661e-05}
{"train_loss": 0.17027197778224945, "global_step": 95759, "epoch": 1075, "lr": 7.247842229701183e-05}
{"train_loss": 0.15420706570148468, "global_step": 95760, "epoch": 1075, "lr": 7.247790444816508e-05}
{"train_loss": 0.2623286545276642, "global_step": 95761, "epoch": 1075, "lr": 7.247738659629647e-05}
{"train_loss": 0.201590433716774, "global_step": 95762, "epoch": 1075, "lr": 7.247686874140604e-05}
{"train_loss": 0.16694217920303345, "global_step": 95763, "epoch": 1075, "lr": 7.247635088349388e-05, "val_loss": 4.531651020050049, "train_action_mse_error": 12.160604476928711}
{"train_loss": 0.11230513453483582, "global_step": 95764, "epoch": 1076, "lr": 7.247583302256004e-05}
{"train_loss": 0.12844207882881165, "global_step": 95765, "epoch": 1076, "lr": 7.24753151586046e-05}
{"train_loss": 0.08485777676105499, "global_step": 95766, "epoch": 1076, "lr": 7.247479729162764e-05}
{"train_loss": 0.16414308547973633, "global_step": 95767, "epoch": 1076, "lr": 7.247427942162921e-05}
{"train_loss": 0.144306018948555, "global_step": 95768, "epoch": 1076, "lr": 7.247376154860939e-05}
{"train_loss": 0.1818915456533432, "global_step": 95769, "epoch": 1076, "lr": 7.247324367256823e-05}
{"train_loss": 0.2447817176580429, "global_step": 95770, "epoch": 1076, "lr": 7.247272579350584e-05}
{"train_loss": 0.16587497293949127, "global_step": 95771, "epoch": 1076, "lr": 7.247220791142224e-05}
{"train_loss": 0.20437665283679962, "global_step": 95772, "epoch": 1076, "lr": 7.247169002631756e-05}
{"train_loss": 0.153123140335083, "global_step": 95773, "epoch": 1076, "lr": 7.24711721381918e-05}
{"train_loss": 0.12577420473098755, "global_step": 95774, "epoch": 1076, "lr": 7.247065424704509e-05}
{"train_loss": 0.11285217106342316, "global_step": 95775, "epoch": 1076, "lr": 7.247013635287746e-05}
{"train_loss": 0.17956481873989105, "global_step": 95776, "epoch": 1076, "lr": 7.2469618455689e-05}
{"train_loss": 0.24612145125865936, "global_step": 95777, "epoch": 1076, "lr": 7.246910055547977e-05}
{"train_loss": 0.1797734498977661, "global_step": 95778, "epoch": 1076, "lr": 7.246858265224984e-05}
{"train_loss": 0.2969973385334015, "global_step": 95779, "epoch": 1076, "lr": 7.246806474599929e-05}
{"train_loss": 0.27387797832489014, "global_step": 95780, "epoch": 1076, "lr": 7.246754683672817e-05}
{"train_loss": 0.20534518361091614, "global_step": 95781, "epoch": 1076, "lr": 7.246702892443657e-05}
{"train_loss": 0.1879124939441681, "global_step": 95782, "epoch": 1076, "lr": 7.246651100912456e-05}
{"train_loss": 0.24278314411640167, "global_step": 95783, "epoch": 1076, "lr": 7.24659930907922e-05}
{"train_loss": 0.13145394623279572, "global_step": 95784, "epoch": 1076, "lr": 7.246547516943955e-05}
{"train_loss": 0.10314811766147614, "global_step": 95785, "epoch": 1076, "lr": 7.24649572450667e-05}
{"train_loss": 0.11731626838445663, "global_step": 95786, "epoch": 1076, "lr": 7.24644393176737e-05}
{"train_loss": 0.2354159951210022, "global_step": 95787, "epoch": 1076, "lr": 7.246392138726065e-05}
{"train_loss": 0.18798868358135223, "global_step": 95788, "epoch": 1076, "lr": 7.246340345382758e-05}
{"train_loss": 0.20734404027462006, "global_step": 95789, "epoch": 1076, "lr": 7.24628855173746e-05}
{"train_loss": 0.16399936378002167, "global_step": 95790, "epoch": 1076, "lr": 7.246236757790174e-05}
{"train_loss": 0.2546705901622772, "global_step": 95791, "epoch": 1076, "lr": 7.246184963540909e-05}
{"train_loss": 0.19052202999591827, "global_step": 95792, "epoch": 1076, "lr": 7.246133168989672e-05}
{"train_loss": 0.17981866002082825, "global_step": 95793, "epoch": 1076, "lr": 7.24608137413647e-05}
{"train_loss": 0.1117875874042511, "global_step": 95794, "epoch": 1076, "lr": 7.24602957898131e-05}
{"train_loss": 0.22128310799598694, "global_step": 95795, "epoch": 1076, "lr": 7.245977783524198e-05}
{"train_loss": 0.1517103761434555, "global_step": 95796, "epoch": 1076, "lr": 7.245925987765141e-05}
{"train_loss": 0.16290269792079926, "global_step": 95797, "epoch": 1076, "lr": 7.245874191704148e-05}
{"train_loss": 0.1363849639892578, "global_step": 95798, "epoch": 1076, "lr": 7.245822395341226e-05}
{"train_loss": 0.2547188401222229, "global_step": 95799, "epoch": 1076, "lr": 7.245770598676377e-05}
{"train_loss": 0.1960247904062271, "global_step": 95800, "epoch": 1076, "lr": 7.245718801709613e-05}
{"train_loss": 0.19873934984207153, "global_step": 95801, "epoch": 1076, "lr": 7.245667004440941e-05}
{"train_loss": 0.22148513793945312, "global_step": 95802, "epoch": 1076, "lr": 7.245615206870366e-05}
{"train_loss": 0.20875580608844757, "global_step": 95803, "epoch": 1076, "lr": 7.245563408997893e-05}
{"train_loss": 0.2086917757987976, "global_step": 95804, "epoch": 1076, "lr": 7.245511610823533e-05}
{"train_loss": 0.17429809272289276, "global_step": 95805, "epoch": 1076, "lr": 7.245459812347292e-05}
{"train_loss": 0.20788021385669708, "global_step": 95806, "epoch": 1076, "lr": 7.245408013569176e-05}
{"train_loss": 0.17580339312553406, "global_step": 95807, "epoch": 1076, "lr": 7.245356214489192e-05}
{"train_loss": 0.19780360162258148, "global_step": 95808, "epoch": 1076, "lr": 7.245304415107347e-05}
{"train_loss": 0.22560295462608337, "global_step": 95809, "epoch": 1076, "lr": 7.245252615423648e-05}
{"train_loss": 0.15703538060188293, "global_step": 95810, "epoch": 1076, "lr": 7.245200815438102e-05}
{"train_loss": 0.1860540509223938, "global_step": 95811, "epoch": 1076, "lr": 7.245149015150718e-05}
{"train_loss": 0.13067273795604706, "global_step": 95812, "epoch": 1076, "lr": 7.2450972145615e-05}
{"train_loss": 0.2298511266708374, "global_step": 95813, "epoch": 1076, "lr": 7.245045413670455e-05}
{"train_loss": 0.22537773847579956, "global_step": 95814, "epoch": 1076, "lr": 7.244993612477592e-05}
{"train_loss": 0.09135688096284866, "global_step": 95815, "epoch": 1076, "lr": 7.244941810982916e-05}
{"train_loss": 0.15057307481765747, "global_step": 95816, "epoch": 1076, "lr": 7.244890009186436e-05}
{"train_loss": 0.1478900909423828, "global_step": 95817, "epoch": 1076, "lr": 7.244838207088157e-05}
{"train_loss": 0.1269063651561737, "global_step": 95818, "epoch": 1076, "lr": 7.244786404688087e-05}
{"train_loss": 0.14481422305107117, "global_step": 95819, "epoch": 1076, "lr": 7.244734601986233e-05}
{"train_loss": 0.17978554964065552, "global_step": 95820, "epoch": 1076, "lr": 7.244682798982603e-05}
{"train_loss": 0.14652830362319946, "global_step": 95821, "epoch": 1076, "lr": 7.244630995677202e-05}
{"train_loss": 0.18493671715259552, "global_step": 95822, "epoch": 1076, "lr": 7.244579192070036e-05}
{"train_loss": 0.18490555882453918, "global_step": 95823, "epoch": 1076, "lr": 7.244527388161116e-05}
{"train_loss": 0.1939094513654709, "global_step": 95824, "epoch": 1076, "lr": 7.244475583950446e-05}
{"train_loss": 0.19330176711082458, "global_step": 95825, "epoch": 1076, "lr": 7.244423779438034e-05}
{"train_loss": 0.16866661608219147, "global_step": 95826, "epoch": 1076, "lr": 7.244371974623884e-05}
{"train_loss": 0.17198669910430908, "global_step": 95827, "epoch": 1076, "lr": 7.244320169508008e-05}
{"train_loss": 0.16269612312316895, "global_step": 95828, "epoch": 1076, "lr": 7.244268364090411e-05}
{"train_loss": 0.13768890500068665, "global_step": 95829, "epoch": 1076, "lr": 7.244216558371099e-05}
{"train_loss": 0.14340642094612122, "global_step": 95830, "epoch": 1076, "lr": 7.244164752350078e-05}
{"train_loss": 0.18849441409111023, "global_step": 95831, "epoch": 1076, "lr": 7.244112946027358e-05}
{"train_loss": 0.19901885092258453, "global_step": 95832, "epoch": 1076, "lr": 7.244061139402944e-05}
{"train_loss": 0.24991753697395325, "global_step": 95833, "epoch": 1076, "lr": 7.244009332476842e-05}
{"train_loss": 0.2174636870622635, "global_step": 95834, "epoch": 1076, "lr": 7.243957525249061e-05}
{"train_loss": 0.11753819137811661, "global_step": 95835, "epoch": 1076, "lr": 7.243905717719609e-05}
{"train_loss": 0.2300531417131424, "global_step": 95836, "epoch": 1076, "lr": 7.243853909888489e-05}
{"train_loss": 0.171066015958786, "global_step": 95837, "epoch": 1076, "lr": 7.24380210175571e-05}
{"train_loss": 0.20805399119853973, "global_step": 95838, "epoch": 1076, "lr": 7.243750293321281e-05}
{"train_loss": 0.14726164937019348, "global_step": 95839, "epoch": 1076, "lr": 7.243698484585207e-05}
{"train_loss": 0.23535941541194916, "global_step": 95840, "epoch": 1076, "lr": 7.243646675547494e-05}
{"train_loss": 0.16397573053836823, "global_step": 95841, "epoch": 1076, "lr": 7.24359486620815e-05}
{"train_loss": 0.24835552275180817, "global_step": 95842, "epoch": 1076, "lr": 7.243543056567184e-05}
{"train_loss": 0.1729806512594223, "global_step": 95843, "epoch": 1076, "lr": 7.243491246624599e-05}
{"train_loss": 0.1800665557384491, "global_step": 95844, "epoch": 1076, "lr": 7.243439436380404e-05}
{"train_loss": 0.17383667826652527, "global_step": 95845, "epoch": 1076, "lr": 7.243387625834605e-05}
{"train_loss": 0.19216032326221466, "global_step": 95846, "epoch": 1076, "lr": 7.243335814987212e-05}
{"train_loss": 0.14094996452331543, "global_step": 95847, "epoch": 1076, "lr": 7.24328400383823e-05}
{"train_loss": 0.13007166981697083, "global_step": 95848, "epoch": 1076, "lr": 7.243232192387665e-05}
{"train_loss": 0.1722613126039505, "global_step": 95849, "epoch": 1076, "lr": 7.243180380635524e-05}
{"train_loss": 0.13329894840717316, "global_step": 95850, "epoch": 1076, "lr": 7.243128568581815e-05}
{"train_loss": 0.25234195590019226, "global_step": 95851, "epoch": 1076, "lr": 7.243076756226545e-05}
{"train_loss": 0.1799716581956724, "global_step": 95852, "epoch": 1076, "lr": 7.24302494356972e-05, "val_loss": 4.5701727867126465}
{"train_loss": 0.1612624078989029, "global_step": 95853, "epoch": 1077, "lr": 7.242973130611349e-05}
{"train_loss": 0.11045204102993011, "global_step": 95854, "epoch": 1077, "lr": 7.242921317351437e-05}
{"train_loss": 0.13837498426437378, "global_step": 95855, "epoch": 1077, "lr": 7.242869503789992e-05}
{"train_loss": 0.1490863561630249, "global_step": 95856, "epoch": 1077, "lr": 7.24281768992702e-05}
{"train_loss": 0.1604476273059845, "global_step": 95857, "epoch": 1077, "lr": 7.242765875762529e-05}
{"train_loss": 0.1876116842031479, "global_step": 95858, "epoch": 1077, "lr": 7.242714061296524e-05}
{"train_loss": 0.1509673148393631, "global_step": 95859, "epoch": 1077, "lr": 7.242662246529016e-05}
{"train_loss": 0.12851019203662872, "global_step": 95860, "epoch": 1077, "lr": 7.242610431460006e-05}
{"train_loss": 0.2300543338060379, "global_step": 95861, "epoch": 1077, "lr": 7.242558616089506e-05}
{"train_loss": 0.21094918251037598, "global_step": 95862, "epoch": 1077, "lr": 7.242506800417522e-05}
{"train_loss": 0.19611766934394836, "global_step": 95863, "epoch": 1077, "lr": 7.24245498444406e-05}
{"train_loss": 0.14007648825645447, "global_step": 95864, "epoch": 1077, "lr": 7.242403168169127e-05}
{"train_loss": 0.1740468293428421, "global_step": 95865, "epoch": 1077, "lr": 7.242351351592731e-05}
{"train_loss": 0.13570980727672577, "global_step": 95866, "epoch": 1077, "lr": 7.242299534714877e-05}
{"train_loss": 0.20052967965602875, "global_step": 95867, "epoch": 1077, "lr": 7.242247717535573e-05}
{"train_loss": 0.23979081213474274, "global_step": 95868, "epoch": 1077, "lr": 7.242195900054826e-05}
{"train_loss": 0.131633460521698, "global_step": 95869, "epoch": 1077, "lr": 7.242144082272646e-05}
{"train_loss": 0.12284397333860397, "global_step": 95870, "epoch": 1077, "lr": 7.242092264189034e-05}
{"train_loss": 0.14819714426994324, "global_step": 95871, "epoch": 1077, "lr": 7.242040445804002e-05}
{"train_loss": 0.26887473464012146, "global_step": 95872, "epoch": 1077, "lr": 7.241988627117553e-05}
{"train_loss": 0.20717035233974457, "global_step": 95873, "epoch": 1077, "lr": 7.241936808129697e-05}
{"train_loss": 0.1396993100643158, "global_step": 95874, "epoch": 1077, "lr": 7.24188498884044e-05}
{"train_loss": 0.12509530782699585, "global_step": 95875, "epoch": 1077, "lr": 7.241833169249789e-05}
{"train_loss": 0.11897476762533188, "global_step": 95876, "epoch": 1077, "lr": 7.241781349357751e-05}
{"train_loss": 0.1511506587266922, "global_step": 95877, "epoch": 1077, "lr": 7.241729529164332e-05}
{"train_loss": 0.14517074823379517, "global_step": 95878, "epoch": 1077, "lr": 7.24167770866954e-05}
{"train_loss": 0.1735905408859253, "global_step": 95879, "epoch": 1077, "lr": 7.241625887873382e-05}
{"train_loss": 0.2224576771259308, "global_step": 95880, "epoch": 1077, "lr": 7.241574066775864e-05}
{"train_loss": 0.18119080364704132, "global_step": 95881, "epoch": 1077, "lr": 7.241522245376995e-05}
{"train_loss": 0.2699621319770813, "global_step": 95882, "epoch": 1077, "lr": 7.24147042367678e-05}
{"train_loss": 0.1836068034172058, "global_step": 95883, "epoch": 1077, "lr": 7.241418601675226e-05}
{"train_loss": 0.17519500851631165, "global_step": 95884, "epoch": 1077, "lr": 7.241366779372341e-05}
{"train_loss": 0.1564454436302185, "global_step": 95885, "epoch": 1077, "lr": 7.241314956768132e-05}
{"train_loss": 0.21622434258460999, "global_step": 95886, "epoch": 1077, "lr": 7.241263133862604e-05}
{"train_loss": 0.1391972005367279, "global_step": 95887, "epoch": 1077, "lr": 7.241211310655767e-05}
{"train_loss": 0.11686370521783829, "global_step": 95888, "epoch": 1077, "lr": 7.241159487147628e-05}
{"train_loss": 0.1843705177307129, "global_step": 95889, "epoch": 1077, "lr": 7.241107663338189e-05}
{"train_loss": 0.17711788415908813, "global_step": 95890, "epoch": 1077, "lr": 7.241055839227463e-05}
{"train_loss": 0.10309664160013199, "global_step": 95891, "epoch": 1077, "lr": 7.241004014815453e-05}
{"train_loss": 0.1653420478105545, "global_step": 95892, "epoch": 1077, "lr": 7.240952190102168e-05}
{"train_loss": 0.1504759043455124, "global_step": 95893, "epoch": 1077, "lr": 7.240900365087614e-05}
{"train_loss": 0.18995116651058197, "global_step": 95894, "epoch": 1077, "lr": 7.240848539771798e-05}
{"train_loss": 0.11797367036342621, "global_step": 95895, "epoch": 1077, "lr": 7.240796714154727e-05}
{"train_loss": 0.18116328120231628, "global_step": 95896, "epoch": 1077, "lr": 7.24074488823641e-05}
{"train_loss": 0.18094152212142944, "global_step": 95897, "epoch": 1077, "lr": 7.24069306201685e-05}
{"train_loss": 0.13791128993034363, "global_step": 95898, "epoch": 1077, "lr": 7.240641235496057e-05}
{"train_loss": 0.11462089419364929, "global_step": 95899, "epoch": 1077, "lr": 7.240589408674036e-05}
{"train_loss": 0.14204832911491394, "global_step": 95900, "epoch": 1077, "lr": 7.240537581550797e-05}
{"train_loss": 0.13204313814640045, "global_step": 95901, "epoch": 1077, "lr": 7.240485754126344e-05}
{"train_loss": 0.10902060568332672, "global_step": 95902, "epoch": 1077, "lr": 7.240433926400685e-05}
{"train_loss": 0.1951802521944046, "global_step": 95903, "epoch": 1077, "lr": 7.240382098373829e-05}
{"train_loss": 0.12544573843479156, "global_step": 95904, "epoch": 1077, "lr": 7.240330270045778e-05}
{"train_loss": 0.1938202679157257, "global_step": 95905, "epoch": 1077, "lr": 7.240278441416542e-05}
{"train_loss": 0.1645091474056244, "global_step": 95906, "epoch": 1077, "lr": 7.24022661248613e-05}
{"train_loss": 0.2881876528263092, "global_step": 95907, "epoch": 1077, "lr": 7.240174783254546e-05}
{"train_loss": 0.20138652622699738, "global_step": 95908, "epoch": 1077, "lr": 7.240122953721797e-05}
{"train_loss": 0.12784020602703094, "global_step": 95909, "epoch": 1077, "lr": 7.24007112388789e-05}
{"train_loss": 0.12243323028087616, "global_step": 95910, "epoch": 1077, "lr": 7.240019293752834e-05}
{"train_loss": 0.13405121862888336, "global_step": 95911, "epoch": 1077, "lr": 7.239967463316634e-05}
{"train_loss": 0.2151995450258255, "global_step": 95912, "epoch": 1077, "lr": 7.239915632579299e-05}
{"train_loss": 0.1946977823972702, "global_step": 95913, "epoch": 1077, "lr": 7.239863801540835e-05}
{"train_loss": 0.24216631054878235, "global_step": 95914, "epoch": 1077, "lr": 7.239811970201247e-05}
{"train_loss": 0.17575135827064514, "global_step": 95915, "epoch": 1077, "lr": 7.239760138560544e-05}
{"train_loss": 0.18919698894023895, "global_step": 95916, "epoch": 1077, "lr": 7.239708306618733e-05}
{"train_loss": 0.13588422536849976, "global_step": 95917, "epoch": 1077, "lr": 7.23965647437582e-05}
{"train_loss": 0.08512412011623383, "global_step": 95918, "epoch": 1077, "lr": 7.239604641831813e-05}
{"train_loss": 0.19778411090373993, "global_step": 95919, "epoch": 1077, "lr": 7.239552808986718e-05}
{"train_loss": 0.16216681897640228, "global_step": 95920, "epoch": 1077, "lr": 7.239500975840543e-05}
{"train_loss": 0.20468072593212128, "global_step": 95921, "epoch": 1077, "lr": 7.239449142393294e-05}
{"train_loss": 0.1360771507024765, "global_step": 95922, "epoch": 1077, "lr": 7.239397308644978e-05}
{"train_loss": 0.18334311246871948, "global_step": 95923, "epoch": 1077, "lr": 7.239345474595603e-05}
{"train_loss": 0.14307282865047455, "global_step": 95924, "epoch": 1077, "lr": 7.239293640245175e-05}
{"train_loss": 0.15889281034469604, "global_step": 95925, "epoch": 1077, "lr": 7.239241805593702e-05}
{"train_loss": 0.12181216478347778, "global_step": 95926, "epoch": 1077, "lr": 7.239189970641189e-05}
{"train_loss": 0.19257354736328125, "global_step": 95927, "epoch": 1077, "lr": 7.239138135387645e-05}
{"train_loss": 0.2143046259880066, "global_step": 95928, "epoch": 1077, "lr": 7.239086299833076e-05}
{"train_loss": 0.191252663731575, "global_step": 95929, "epoch": 1077, "lr": 7.239034463977488e-05}
{"train_loss": 0.24472381174564362, "global_step": 95930, "epoch": 1077, "lr": 7.23898262782089e-05}
{"train_loss": 0.11259260773658752, "global_step": 95931, "epoch": 1077, "lr": 7.238930791363289e-05}
{"train_loss": 0.18048326671123505, "global_step": 95932, "epoch": 1077, "lr": 7.23887895460469e-05}
{"train_loss": 0.18307232856750488, "global_step": 95933, "epoch": 1077, "lr": 7.2388271175451e-05}
{"train_loss": 0.24249140918254852, "global_step": 95934, "epoch": 1077, "lr": 7.238775280184528e-05}
{"train_loss": 0.1746842861175537, "global_step": 95935, "epoch": 1077, "lr": 7.238723442522981e-05}
{"train_loss": 0.12386025488376617, "global_step": 95936, "epoch": 1077, "lr": 7.238671604560464e-05}
{"train_loss": 0.2170923352241516, "global_step": 95937, "epoch": 1077, "lr": 7.238619766296986e-05}
{"train_loss": 0.20079673826694489, "global_step": 95938, "epoch": 1077, "lr": 7.238567927732552e-05}
{"train_loss": 0.13700160384178162, "global_step": 95939, "epoch": 1077, "lr": 7.23851608886717e-05}
{"train_loss": 0.16740193963050842, "global_step": 95940, "epoch": 1077, "lr": 7.238464249700845e-05}
{"train_loss": 0.16936532634027887, "global_step": 95941, "epoch": 1077, "lr": 7.238412410233587e-05, "val_loss": 4.4136271476745605}
{"train_loss": 0.15151454508304596, "global_step": 95942, "epoch": 1078, "lr": 7.238360570465402e-05}
{"train_loss": 0.16615350544452667, "global_step": 95943, "epoch": 1078, "lr": 7.238308730396297e-05}
{"train_loss": 0.1537773609161377, "global_step": 95944, "epoch": 1078, "lr": 7.238256890026278e-05}
{"train_loss": 0.11988920718431473, "global_step": 95945, "epoch": 1078, "lr": 7.238205049355354e-05}
{"train_loss": 0.14680074155330658, "global_step": 95946, "epoch": 1078, "lr": 7.23815320838353e-05}
{"train_loss": 0.18753765523433685, "global_step": 95947, "epoch": 1078, "lr": 7.238101367110812e-05}
{"train_loss": 0.17796437442302704, "global_step": 95948, "epoch": 1078, "lr": 7.23804952553721e-05}
{"train_loss": 0.13117365539073944, "global_step": 95949, "epoch": 1078, "lr": 7.237997683662729e-05}
{"train_loss": 0.19864961504936218, "global_step": 95950, "epoch": 1078, "lr": 7.237945841487378e-05}
{"train_loss": 0.19459028542041779, "global_step": 95951, "epoch": 1078, "lr": 7.237893999011159e-05}
{"train_loss": 0.1604830026626587, "global_step": 95952, "epoch": 1078, "lr": 7.237842156234085e-05}
{"train_loss": 0.16500236093997955, "global_step": 95953, "epoch": 1078, "lr": 7.23779031315616e-05}
{"train_loss": 0.1844598948955536, "global_step": 95954, "epoch": 1078, "lr": 7.237738469777391e-05}
{"train_loss": 0.11390592902898788, "global_step": 95955, "epoch": 1078, "lr": 7.237686626097787e-05}
{"train_loss": 0.2463913857936859, "global_step": 95956, "epoch": 1078, "lr": 7.23763478211735e-05}
{"train_loss": 0.22009845077991486, "global_step": 95957, "epoch": 1078, "lr": 7.237582937836094e-05}
{"train_loss": 0.14536283910274506, "global_step": 95958, "epoch": 1078, "lr": 7.23753109325402e-05}
{"train_loss": 0.12485463917255402, "global_step": 95959, "epoch": 1078, "lr": 7.237479248371138e-05}
{"train_loss": 0.15386012196540833, "global_step": 95960, "epoch": 1078, "lr": 7.237427403187453e-05}
{"train_loss": 0.26927992701530457, "global_step": 95961, "epoch": 1078, "lr": 7.237375557702973e-05}
{"train_loss": 0.2074035406112671, "global_step": 95962, "epoch": 1078, "lr": 7.237323711917707e-05}
{"train_loss": 0.17583584785461426, "global_step": 95963, "epoch": 1078, "lr": 7.23727186583166e-05}
{"train_loss": 0.14122213423252106, "global_step": 95964, "epoch": 1078, "lr": 7.23722001944484e-05}
{"train_loss": 0.19779928028583527, "global_step": 95965, "epoch": 1078, "lr": 7.23716817275725e-05}
{"train_loss": 0.2465740293264389, "global_step": 95966, "epoch": 1078, "lr": 7.237116325768901e-05}
{"train_loss": 0.10470210760831833, "global_step": 95967, "epoch": 1078, "lr": 7.237064478479801e-05}
{"train_loss": 0.1641659438610077, "global_step": 95968, "epoch": 1078, "lr": 7.237012630889952e-05}
{"train_loss": 0.1322205662727356, "global_step": 95969, "epoch": 1078, "lr": 7.236960782999365e-05}
{"train_loss": 0.19780440628528595, "global_step": 95970, "epoch": 1078, "lr": 7.236908934808046e-05}
{"train_loss": 0.16911689937114716, "global_step": 95971, "epoch": 1078, "lr": 7.236857086316003e-05}
{"train_loss": 0.11542476713657379, "global_step": 95972, "epoch": 1078, "lr": 7.23680523752324e-05}
{"train_loss": 0.17511612176895142, "global_step": 95973, "epoch": 1078, "lr": 7.236753388429767e-05}
{"train_loss": 0.12936067581176758, "global_step": 95974, "epoch": 1078, "lr": 7.236701539035588e-05}
{"train_loss": 0.19123248755931854, "global_step": 95975, "epoch": 1078, "lr": 7.236649689340715e-05}
{"train_loss": 0.17579467594623566, "global_step": 95976, "epoch": 1078, "lr": 7.236597839345149e-05}
{"train_loss": 0.31796789169311523, "global_step": 95977, "epoch": 1078, "lr": 7.2365459890489e-05}
{"train_loss": 0.18831177055835724, "global_step": 95978, "epoch": 1078, "lr": 7.236494138451975e-05}
{"train_loss": 0.25231754779815674, "global_step": 95979, "epoch": 1078, "lr": 7.236442287554381e-05}
{"train_loss": 0.17417234182357788, "global_step": 95980, "epoch": 1078, "lr": 7.236390436356124e-05}
{"train_loss": 0.1594906896352768, "global_step": 95981, "epoch": 1078, "lr": 7.236338584857213e-05}
{"train_loss": 0.16569718718528748, "global_step": 95982, "epoch": 1078, "lr": 7.236286733057652e-05}
{"train_loss": 0.11597144603729248, "global_step": 95983, "epoch": 1078, "lr": 7.236234880957451e-05}
{"train_loss": 0.2474791556596756, "global_step": 95984, "epoch": 1078, "lr": 7.236183028556614e-05}
{"train_loss": 0.1776202768087387, "global_step": 95985, "epoch": 1078, "lr": 7.23613117585515e-05}
{"train_loss": 0.18342164158821106, "global_step": 95986, "epoch": 1078, "lr": 7.236079322853065e-05}
{"train_loss": 0.17716488242149353, "global_step": 95987, "epoch": 1078, "lr": 7.236027469550367e-05}
{"train_loss": 0.20893944799900055, "global_step": 95988, "epoch": 1078, "lr": 7.235975615947061e-05}
{"train_loss": 0.14191308617591858, "global_step": 95989, "epoch": 1078, "lr": 7.235923762043158e-05}
{"train_loss": 0.13565433025360107, "global_step": 95990, "epoch": 1078, "lr": 7.235871907838661e-05}
{"train_loss": 0.1463385671377182, "global_step": 95991, "epoch": 1078, "lr": 7.235820053333577e-05}
{"train_loss": 0.17406409978866577, "global_step": 95992, "epoch": 1078, "lr": 7.235768198527915e-05}
{"train_loss": 0.17376987636089325, "global_step": 95993, "epoch": 1078, "lr": 7.235716343421682e-05}
{"train_loss": 0.11920272558927536, "global_step": 95994, "epoch": 1078, "lr": 7.235664488014884e-05}
{"train_loss": 0.1581113487482071, "global_step": 95995, "epoch": 1078, "lr": 7.235612632307527e-05}
{"train_loss": 0.290640264749527, "global_step": 95996, "epoch": 1078, "lr": 7.235560776299619e-05}
{"train_loss": 0.1267903596162796, "global_step": 95997, "epoch": 1078, "lr": 7.235508919991168e-05}
{"train_loss": 0.20330575108528137, "global_step": 95998, "epoch": 1078, "lr": 7.23545706338218e-05}
{"train_loss": 0.16744691133499146, "global_step": 95999, "epoch": 1078, "lr": 7.235405206472663e-05}
{"train_loss": 0.1972481608390808, "global_step": 96000, "epoch": 1078, "lr": 7.235353349262622e-05}
{"train_loss": 0.19983501732349396, "global_step": 96001, "epoch": 1078, "lr": 7.235301491752066e-05}
{"train_loss": 0.21115976572036743, "global_step": 96002, "epoch": 1078, "lr": 7.235249633940999e-05}
{"train_loss": 0.16189154982566833, "global_step": 96003, "epoch": 1078, "lr": 7.235197775829432e-05}
{"train_loss": 0.08671247214078903, "global_step": 96004, "epoch": 1078, "lr": 7.235145917417369e-05}
{"train_loss": 0.13610678911209106, "global_step": 96005, "epoch": 1078, "lr": 7.235094058704817e-05}
{"train_loss": 0.19200728833675385, "global_step": 96006, "epoch": 1078, "lr": 7.235042199691785e-05}
{"train_loss": 0.14935225248336792, "global_step": 96007, "epoch": 1078, "lr": 7.234990340378278e-05}
{"train_loss": 0.2253304272890091, "global_step": 96008, "epoch": 1078, "lr": 7.234938480764305e-05}
{"train_loss": 0.16617721319198608, "global_step": 96009, "epoch": 1078, "lr": 7.23488662084987e-05}
{"train_loss": 0.15700717270374298, "global_step": 96010, "epoch": 1078, "lr": 7.234834760634982e-05}
{"train_loss": 0.263643354177475, "global_step": 96011, "epoch": 1078, "lr": 7.23478290011965e-05}
{"train_loss": 0.19829347729682922, "global_step": 96012, "epoch": 1078, "lr": 7.234731039303878e-05}
{"train_loss": 0.2208988070487976, "global_step": 96013, "epoch": 1078, "lr": 7.234679178187672e-05}
{"train_loss": 0.11654651165008545, "global_step": 96014, "epoch": 1078, "lr": 7.23462731677104e-05}
{"train_loss": 0.18716028332710266, "global_step": 96015, "epoch": 1078, "lr": 7.234575455053991e-05}
{"train_loss": 0.16266676783561707, "global_step": 96016, "epoch": 1078, "lr": 7.23452359303653e-05}
{"train_loss": 0.12615352869033813, "global_step": 96017, "epoch": 1078, "lr": 7.234471730718665e-05}
{"train_loss": 0.2393709123134613, "global_step": 96018, "epoch": 1078, "lr": 7.234419868100403e-05}
{"train_loss": 0.13920462131500244, "global_step": 96019, "epoch": 1078, "lr": 7.23436800518175e-05}
{"train_loss": 0.1491822600364685, "global_step": 96020, "epoch": 1078, "lr": 7.234316141962716e-05}
{"train_loss": 0.24456705152988434, "global_step": 96021, "epoch": 1078, "lr": 7.234264278443303e-05}
{"train_loss": 0.20717263221740723, "global_step": 96022, "epoch": 1078, "lr": 7.23421241462352e-05}
{"train_loss": 0.16945341229438782, "global_step": 96023, "epoch": 1078, "lr": 7.234160550503374e-05}
{"train_loss": 0.15431752800941467, "global_step": 96024, "epoch": 1078, "lr": 7.234108686082874e-05}
{"train_loss": 0.11897216737270355, "global_step": 96025, "epoch": 1078, "lr": 7.234056821362024e-05}
{"train_loss": 0.15321822464466095, "global_step": 96026, "epoch": 1078, "lr": 7.234004956340832e-05}
{"train_loss": 0.14684638381004333, "global_step": 96027, "epoch": 1078, "lr": 7.233953091019306e-05}
{"train_loss": 0.09823384881019592, "global_step": 96028, "epoch": 1078, "lr": 7.233901225397453e-05}
{"train_loss": 0.1766733080148697, "global_step": 96029, "epoch": 1078, "lr": 7.233849359475278e-05}
{"train_loss": 0.17379166847199537, "global_step": 96030, "epoch": 1078, "lr": 7.233797493252789e-05, "val_loss": 4.419061183929443}
{"train_loss": 0.14241273701190948, "global_step": 96031, "epoch": 1079, "lr": 7.233745626729995e-05}
{"train_loss": 0.19097773730754852, "global_step": 96032, "epoch": 1079, "lr": 7.233693759906899e-05}
{"train_loss": 0.1741604059934616, "global_step": 96033, "epoch": 1079, "lr": 7.233641892783511e-05}
{"train_loss": 0.15097977221012115, "global_step": 96034, "epoch": 1079, "lr": 7.233590025359837e-05}
{"train_loss": 0.1576293557882309, "global_step": 96035, "epoch": 1079, "lr": 7.233538157635884e-05}
{"train_loss": 0.16079643368721008, "global_step": 96036, "epoch": 1079, "lr": 7.233486289611658e-05}
{"train_loss": 0.15198762714862823, "global_step": 96037, "epoch": 1079, "lr": 7.233434421287168e-05}
{"train_loss": 0.1421484798192978, "global_step": 96038, "epoch": 1079, "lr": 7.23338255266242e-05}
{"train_loss": 0.1202094778418541, "global_step": 96039, "epoch": 1079, "lr": 7.233330683737422e-05}
{"train_loss": 0.31423380970954895, "global_step": 96040, "epoch": 1079, "lr": 7.23327881451218e-05}
{"train_loss": 0.14189346134662628, "global_step": 96041, "epoch": 1079, "lr": 7.2332269449867e-05}
{"train_loss": 0.15545853972434998, "global_step": 96042, "epoch": 1079, "lr": 7.23317507516099e-05}
{"train_loss": 0.28257355093955994, "global_step": 96043, "epoch": 1079, "lr": 7.233123205035056e-05}
{"train_loss": 0.11563650518655777, "global_step": 96044, "epoch": 1079, "lr": 7.233071334608906e-05}
{"train_loss": 0.2019348442554474, "global_step": 96045, "epoch": 1079, "lr": 7.233019463882546e-05}
{"train_loss": 0.17357152700424194, "global_step": 96046, "epoch": 1079, "lr": 7.232967592855986e-05}
{"train_loss": 0.16468338668346405, "global_step": 96047, "epoch": 1079, "lr": 7.23291572152923e-05}
{"train_loss": 0.21705935895442963, "global_step": 96048, "epoch": 1079, "lr": 7.232863849902285e-05}
{"train_loss": 0.17037560045719147, "global_step": 96049, "epoch": 1079, "lr": 7.232811977975158e-05}
{"train_loss": 0.18121035397052765, "global_step": 96050, "epoch": 1079, "lr": 7.232760105747858e-05}
{"train_loss": 0.13390560448169708, "global_step": 96051, "epoch": 1079, "lr": 7.23270823322039e-05}
{"train_loss": 0.16131974756717682, "global_step": 96052, "epoch": 1079, "lr": 7.23265636039276e-05}
{"train_loss": 0.1441487967967987, "global_step": 96053, "epoch": 1079, "lr": 7.232604487264978e-05}
{"train_loss": 0.2557550072669983, "global_step": 96054, "epoch": 1079, "lr": 7.232552613837051e-05}
{"train_loss": 0.20377406477928162, "global_step": 96055, "epoch": 1079, "lr": 7.232500740108984e-05}
{"train_loss": 0.16354601085186005, "global_step": 96056, "epoch": 1079, "lr": 7.232448866080783e-05}
{"train_loss": 0.15480650961399078, "global_step": 96057, "epoch": 1079, "lr": 7.232396991752457e-05}
{"train_loss": 0.20070765912532806, "global_step": 96058, "epoch": 1079, "lr": 7.232345117124013e-05}
{"train_loss": 0.13130567967891693, "global_step": 96059, "epoch": 1079, "lr": 7.232293242195458e-05}
{"train_loss": 0.12080276012420654, "global_step": 96060, "epoch": 1079, "lr": 7.232241366966798e-05}
{"train_loss": 0.1683395653963089, "global_step": 96061, "epoch": 1079, "lr": 7.23218949143804e-05}
{"train_loss": 0.1348901093006134, "global_step": 96062, "epoch": 1079, "lr": 7.232137615609191e-05}
{"train_loss": 0.20825321972370148, "global_step": 96063, "epoch": 1079, "lr": 7.232085739480259e-05}
{"train_loss": 0.13818120956420898, "global_step": 96064, "epoch": 1079, "lr": 7.232033863051249e-05}
{"train_loss": 0.1640039086341858, "global_step": 96065, "epoch": 1079, "lr": 7.23198198632217e-05}
{"train_loss": 0.17294791340827942, "global_step": 96066, "epoch": 1079, "lr": 7.231930109293029e-05}
{"train_loss": 0.1869724541902542, "global_step": 96067, "epoch": 1079, "lr": 7.231878231963832e-05}
{"train_loss": 0.1536080539226532, "global_step": 96068, "epoch": 1079, "lr": 7.231826354334586e-05}
{"train_loss": 0.20581834018230438, "global_step": 96069, "epoch": 1079, "lr": 7.231774476405295e-05}
{"train_loss": 0.13868862390518188, "global_step": 96070, "epoch": 1079, "lr": 7.231722598175973e-05}
{"train_loss": 0.2196262925863266, "global_step": 96071, "epoch": 1079, "lr": 7.231670719646621e-05}
{"train_loss": 0.1746169775724411, "global_step": 96072, "epoch": 1079, "lr": 7.231618840817249e-05}
{"train_loss": 0.11696210503578186, "global_step": 96073, "epoch": 1079, "lr": 7.231566961687863e-05}
{"train_loss": 0.15857145190238953, "global_step": 96074, "epoch": 1079, "lr": 7.231515082258472e-05}
{"train_loss": 0.2144935429096222, "global_step": 96075, "epoch": 1079, "lr": 7.231463202529079e-05}
{"train_loss": 0.1989762783050537, "global_step": 96076, "epoch": 1079, "lr": 7.231411322499692e-05}
{"train_loss": 0.1618874967098236, "global_step": 96077, "epoch": 1079, "lr": 7.23135944217032e-05}
{"train_loss": 0.12990783154964447, "global_step": 96078, "epoch": 1079, "lr": 7.23130756154097e-05}
{"train_loss": 0.18879157304763794, "global_step": 96079, "epoch": 1079, "lr": 7.231255680611647e-05}
{"train_loss": 0.135250985622406, "global_step": 96080, "epoch": 1079, "lr": 7.231203799382359e-05}
{"train_loss": 0.15523773431777954, "global_step": 96081, "epoch": 1079, "lr": 7.231151917853113e-05}
{"train_loss": 0.18752117455005646, "global_step": 96082, "epoch": 1079, "lr": 7.231100036023915e-05}
{"train_loss": 0.24729016423225403, "global_step": 96083, "epoch": 1079, "lr": 7.231048153894773e-05}
{"train_loss": 0.16839943826198578, "global_step": 96084, "epoch": 1079, "lr": 7.230996271465694e-05}
{"train_loss": 0.20763731002807617, "global_step": 96085, "epoch": 1079, "lr": 7.230944388736685e-05}
{"train_loss": 0.18544909358024597, "global_step": 96086, "epoch": 1079, "lr": 7.230892505707753e-05}
{"train_loss": 0.21951493620872498, "global_step": 96087, "epoch": 1079, "lr": 7.230840622378902e-05}
{"train_loss": 0.21217899024486542, "global_step": 96088, "epoch": 1079, "lr": 7.230788738750145e-05}
{"train_loss": 0.2681776285171509, "global_step": 96089, "epoch": 1079, "lr": 7.230736854821484e-05}
{"train_loss": 0.18283723294734955, "global_step": 96090, "epoch": 1079, "lr": 7.230684970592928e-05}
{"train_loss": 0.16725321114063263, "global_step": 96091, "epoch": 1079, "lr": 7.230633086064484e-05}
{"train_loss": 0.20018838346004486, "global_step": 96092, "epoch": 1079, "lr": 7.230581201236157e-05}
{"train_loss": 0.19333931803703308, "global_step": 96093, "epoch": 1079, "lr": 7.230529316107957e-05}
{"train_loss": 0.13611584901809692, "global_step": 96094, "epoch": 1079, "lr": 7.23047743067989e-05}
{"train_loss": 0.1737058013677597, "global_step": 96095, "epoch": 1079, "lr": 7.230425544951961e-05}
{"train_loss": 0.13785839080810547, "global_step": 96096, "epoch": 1079, "lr": 7.23037365892418e-05}
{"train_loss": 0.13921357691287994, "global_step": 96097, "epoch": 1079, "lr": 7.230321772596553e-05}
{"train_loss": 0.15567046403884888, "global_step": 96098, "epoch": 1079, "lr": 7.230269885969085e-05}
{"train_loss": 0.15818791091442108, "global_step": 96099, "epoch": 1079, "lr": 7.230217999041783e-05}
{"train_loss": 0.15860798954963684, "global_step": 96100, "epoch": 1079, "lr": 7.230166111814658e-05}
{"train_loss": 0.08209803700447083, "global_step": 96101, "epoch": 1079, "lr": 7.230114224287713e-05}
{"train_loss": 0.1863107979297638, "global_step": 96102, "epoch": 1079, "lr": 7.230062336460956e-05}
{"train_loss": 0.14531372487545013, "global_step": 96103, "epoch": 1079, "lr": 7.230010448334396e-05}
{"train_loss": 0.199741393327713, "global_step": 96104, "epoch": 1079, "lr": 7.229958559908037e-05}
{"train_loss": 0.2010907679796219, "global_step": 96105, "epoch": 1079, "lr": 7.229906671181887e-05}
{"train_loss": 0.21834246814250946, "global_step": 96106, "epoch": 1079, "lr": 7.229854782155952e-05}
{"train_loss": 0.23868729174137115, "global_step": 96107, "epoch": 1079, "lr": 7.229802892830243e-05}
{"train_loss": 0.1733444482088089, "global_step": 96108, "epoch": 1079, "lr": 7.229751003204762e-05}
{"train_loss": 0.1310344636440277, "global_step": 96109, "epoch": 1079, "lr": 7.22969911327952e-05}
{"train_loss": 0.20447997748851776, "global_step": 96110, "epoch": 1079, "lr": 7.229647223054521e-05}
{"train_loss": 0.125772625207901, "global_step": 96111, "epoch": 1079, "lr": 7.229595332529773e-05}
{"train_loss": 0.15383896231651306, "global_step": 96112, "epoch": 1079, "lr": 7.229543441705284e-05}
{"train_loss": 0.15969395637512207, "global_step": 96113, "epoch": 1079, "lr": 7.229491550581059e-05}
{"train_loss": 0.17413978278636932, "global_step": 96114, "epoch": 1079, "lr": 7.229439659157106e-05}
{"train_loss": 0.16651204228401184, "global_step": 96115, "epoch": 1079, "lr": 7.229387767433433e-05}
{"train_loss": 0.1198224425315857, "global_step": 96116, "epoch": 1079, "lr": 7.229335875410046e-05}
{"train_loss": 0.21642494201660156, "global_step": 96117, "epoch": 1079, "lr": 7.229283983086953e-05}
{"train_loss": 0.12981854379177094, "global_step": 96118, "epoch": 1079, "lr": 7.229232090464157e-05}
{"train_loss": 0.17307359549436677, "global_step": 96119, "epoch": 1079, "lr": 7.22918019754167e-05, "val_loss": 4.656105041503906}
{"train_loss": 0.1748633086681366, "global_step": 96120, "epoch": 1080, "lr": 7.229128304319496e-05}
{"train_loss": 0.21135902404785156, "global_step": 96121, "epoch": 1080, "lr": 7.229076410797644e-05}
{"train_loss": 0.15757271647453308, "global_step": 96122, "epoch": 1080, "lr": 7.229024516976119e-05}
{"train_loss": 0.21053805947303772, "global_step": 96123, "epoch": 1080, "lr": 7.228972622854929e-05}
{"train_loss": 0.23649846017360687, "global_step": 96124, "epoch": 1080, "lr": 7.228920728434079e-05}
{"train_loss": 0.19518761336803436, "global_step": 96125, "epoch": 1080, "lr": 7.228868833713581e-05}
{"train_loss": 0.13004179298877716, "global_step": 96126, "epoch": 1080, "lr": 7.228816938693435e-05}
{"train_loss": 0.1258430778980255, "global_step": 96127, "epoch": 1080, "lr": 7.228765043373654e-05}
{"train_loss": 0.17792965471744537, "global_step": 96128, "epoch": 1080, "lr": 7.228713147754243e-05}
{"train_loss": 0.14966250956058502, "global_step": 96129, "epoch": 1080, "lr": 7.228661251835208e-05}
{"train_loss": 0.13676691055297852, "global_step": 96130, "epoch": 1080, "lr": 7.228609355616557e-05}
{"train_loss": 0.17203737795352936, "global_step": 96131, "epoch": 1080, "lr": 7.228557459098298e-05}
{"train_loss": 0.15374423563480377, "global_step": 96132, "epoch": 1080, "lr": 7.228505562280434e-05}
{"train_loss": 0.22895583510398865, "global_step": 96133, "epoch": 1080, "lr": 7.228453665162975e-05}
{"train_loss": 0.16516676545143127, "global_step": 96134, "epoch": 1080, "lr": 7.22840176774593e-05}
{"train_loss": 0.1760525107383728, "global_step": 96135, "epoch": 1080, "lr": 7.2283498700293e-05}
{"train_loss": 0.1680300235748291, "global_step": 96136, "epoch": 1080, "lr": 7.228297972013099e-05}
{"train_loss": 0.14173907041549683, "global_step": 96137, "epoch": 1080, "lr": 7.22824607369733e-05}
{"train_loss": 0.12300460785627365, "global_step": 96138, "epoch": 1080, "lr": 7.228194175081998e-05}
{"train_loss": 0.12856410443782806, "global_step": 96139, "epoch": 1080, "lr": 7.228142276167115e-05}
{"train_loss": 0.09192375838756561, "global_step": 96140, "epoch": 1080, "lr": 7.228090376952683e-05}
{"train_loss": 0.26712262630462646, "global_step": 96141, "epoch": 1080, "lr": 7.228038477438712e-05}
{"train_loss": 0.07319309562444687, "global_step": 96142, "epoch": 1080, "lr": 7.22798657762521e-05}
{"train_loss": 0.22812707722187042, "global_step": 96143, "epoch": 1080, "lr": 7.22793467751218e-05}
{"train_loss": 0.1705445796251297, "global_step": 96144, "epoch": 1080, "lr": 7.227882777099632e-05}
{"train_loss": 0.15161089599132538, "global_step": 96145, "epoch": 1080, "lr": 7.227830876387574e-05}
{"train_loss": 0.13704997301101685, "global_step": 96146, "epoch": 1080, "lr": 7.227778975376011e-05}
{"train_loss": 0.28303709626197815, "global_step": 96147, "epoch": 1080, "lr": 7.227727074064949e-05}
{"train_loss": 0.24867050349712372, "global_step": 96148, "epoch": 1080, "lr": 7.227675172454396e-05}
{"train_loss": 0.14684002101421356, "global_step": 96149, "epoch": 1080, "lr": 7.227623270544362e-05}
{"train_loss": 0.194880872964859, "global_step": 96150, "epoch": 1080, "lr": 7.227571368334848e-05}
{"train_loss": 0.29481586813926697, "global_step": 96151, "epoch": 1080, "lr": 7.227519465825866e-05}
{"train_loss": 0.13194678723812103, "global_step": 96152, "epoch": 1080, "lr": 7.227467563017421e-05}
{"train_loss": 0.19680224359035492, "global_step": 96153, "epoch": 1080, "lr": 7.227415659909519e-05}
{"train_loss": 0.23515771329402924, "global_step": 96154, "epoch": 1080, "lr": 7.22736375650217e-05}
{"train_loss": 0.1437663733959198, "global_step": 96155, "epoch": 1080, "lr": 7.227311852795377e-05}
{"train_loss": 0.15411920845508575, "global_step": 96156, "epoch": 1080, "lr": 7.227259948789151e-05}
{"train_loss": 0.17159028351306915, "global_step": 96157, "epoch": 1080, "lr": 7.227208044483496e-05}
{"train_loss": 0.1587853729724884, "global_step": 96158, "epoch": 1080, "lr": 7.227156139878419e-05}
{"train_loss": 0.1889021098613739, "global_step": 96159, "epoch": 1080, "lr": 7.227104234973929e-05}
{"train_loss": 0.1507684737443924, "global_step": 96160, "epoch": 1080, "lr": 7.227052329770032e-05}
{"train_loss": 0.1659305840730667, "global_step": 96161, "epoch": 1080, "lr": 7.227000424266735e-05}
{"train_loss": 0.15834279358386993, "global_step": 96162, "epoch": 1080, "lr": 7.226948518464045e-05}
{"train_loss": 0.15621092915534973, "global_step": 96163, "epoch": 1080, "lr": 7.226896612361968e-05}
{"train_loss": 0.1470588594675064, "global_step": 96164, "epoch": 1080, "lr": 7.226844705960514e-05}
{"train_loss": 0.0843120589852333, "global_step": 96165, "epoch": 1080, "lr": 7.226792799259686e-05}
{"train_loss": 0.15980510413646698, "global_step": 96166, "epoch": 1080, "lr": 7.226740892259494e-05}
{"train_loss": 0.1932137906551361, "global_step": 96167, "epoch": 1080, "lr": 7.226688984959943e-05}
{"train_loss": 0.17112068831920624, "global_step": 96168, "epoch": 1080, "lr": 7.226637077361041e-05}
{"train_loss": 0.15097352862358093, "global_step": 96169, "epoch": 1080, "lr": 7.226585169462795e-05}
{"train_loss": 0.14136667549610138, "global_step": 96170, "epoch": 1080, "lr": 7.226533261265212e-05}
{"train_loss": 0.17374561727046967, "global_step": 96171, "epoch": 1080, "lr": 7.226481352768299e-05}
{"train_loss": 0.22558091580867767, "global_step": 96172, "epoch": 1080, "lr": 7.226429443972061e-05}
{"train_loss": 0.11804989725351334, "global_step": 96173, "epoch": 1080, "lr": 7.226377534876509e-05}
{"train_loss": 0.1444455087184906, "global_step": 96174, "epoch": 1080, "lr": 7.226325625481646e-05}
{"train_loss": 0.2107994705438614, "global_step": 96175, "epoch": 1080, "lr": 7.226273715787481e-05}
{"train_loss": 0.2153824120759964, "global_step": 96176, "epoch": 1080, "lr": 7.226221805794021e-05}
{"train_loss": 0.14456115663051605, "global_step": 96177, "epoch": 1080, "lr": 7.226169895501272e-05}
{"train_loss": 0.13151122629642487, "global_step": 96178, "epoch": 1080, "lr": 7.226117984909242e-05}
{"train_loss": 0.1423843950033188, "global_step": 96179, "epoch": 1080, "lr": 7.226066074017937e-05}
{"train_loss": 0.2067485898733139, "global_step": 96180, "epoch": 1080, "lr": 7.226014162827366e-05}
{"train_loss": 0.11145184189081192, "global_step": 96181, "epoch": 1080, "lr": 7.225962251337532e-05}
{"train_loss": 0.202662393450737, "global_step": 96182, "epoch": 1080, "lr": 7.225910339548448e-05}
{"train_loss": 0.17803239822387695, "global_step": 96183, "epoch": 1080, "lr": 7.225858427460115e-05}
{"train_loss": 0.18252356350421906, "global_step": 96184, "epoch": 1080, "lr": 7.225806515072543e-05}
{"train_loss": 0.1406441628932953, "global_step": 96185, "epoch": 1080, "lr": 7.225754602385738e-05}
{"train_loss": 0.08267883211374283, "global_step": 96186, "epoch": 1080, "lr": 7.225702689399708e-05}
{"train_loss": 0.1445092260837555, "global_step": 96187, "epoch": 1080, "lr": 7.22565077611446e-05}
{"train_loss": 0.16925758123397827, "global_step": 96188, "epoch": 1080, "lr": 7.22559886253e-05}
{"train_loss": 0.1491222232580185, "global_step": 96189, "epoch": 1080, "lr": 7.225546948646334e-05}
{"train_loss": 0.19457973539829254, "global_step": 96190, "epoch": 1080, "lr": 7.225495034463472e-05}
{"train_loss": 0.16477903723716736, "global_step": 96191, "epoch": 1080, "lr": 7.22544311998142e-05}
{"train_loss": 0.2073468267917633, "global_step": 96192, "epoch": 1080, "lr": 7.225391205200182e-05}
{"train_loss": 0.15992362797260284, "global_step": 96193, "epoch": 1080, "lr": 7.225339290119768e-05}
{"train_loss": 0.23652946949005127, "global_step": 96194, "epoch": 1080, "lr": 7.225287374740186e-05}
{"train_loss": 0.10185565799474716, "global_step": 96195, "epoch": 1080, "lr": 7.22523545906144e-05}
{"train_loss": 0.18697959184646606, "global_step": 96196, "epoch": 1080, "lr": 7.225183543083536e-05}
{"train_loss": 0.11115967482328415, "global_step": 96197, "epoch": 1080, "lr": 7.225131626806485e-05}
{"train_loss": 0.08027473837137222, "global_step": 96198, "epoch": 1080, "lr": 7.225079710230293e-05}
{"train_loss": 0.17085453867912292, "global_step": 96199, "epoch": 1080, "lr": 7.225027793354966e-05}
{"train_loss": 0.11367817968130112, "global_step": 96200, "epoch": 1080, "lr": 7.22497587618051e-05}
{"train_loss": 0.16149865090847015, "global_step": 96201, "epoch": 1080, "lr": 7.224923958706933e-05}
{"train_loss": 0.18309862911701202, "global_step": 96202, "epoch": 1080, "lr": 7.224872040934244e-05}
{"train_loss": 0.26028740406036377, "global_step": 96203, "epoch": 1080, "lr": 7.224820122862447e-05}
{"train_loss": 0.17315252125263214, "global_step": 96204, "epoch": 1080, "lr": 7.22476820449155e-05}
{"train_loss": 0.17779172956943512, "global_step": 96205, "epoch": 1080, "lr": 7.22471628582156e-05}
{"train_loss": 0.15635330975055695, "global_step": 96206, "epoch": 1080, "lr": 7.224664366852484e-05}
{"train_loss": 0.18210554122924805, "global_step": 96207, "epoch": 1080, "lr": 7.224612447584329e-05}
{"train_loss": 0.1683509157279904, "global_step": 96208, "epoch": 1080, "lr": 7.224560528017102e-05, "val_loss": 4.555721282958984, "train_action_mse_error": 27.65964698791504}
{"train_loss": 0.2122056633234024, "global_step": 96209, "epoch": 1081, "lr": 7.224508608150809e-05}
{"train_loss": 0.13687510788440704, "global_step": 96210, "epoch": 1081, "lr": 7.224456687985459e-05}
{"train_loss": 0.11791203171014786, "global_step": 96211, "epoch": 1081, "lr": 7.224404767521058e-05}
{"train_loss": 0.17578889429569244, "global_step": 96212, "epoch": 1081, "lr": 7.224352846757613e-05}
{"train_loss": 0.12702035903930664, "global_step": 96213, "epoch": 1081, "lr": 7.22430092569513e-05}
{"train_loss": 0.1719365268945694, "global_step": 96214, "epoch": 1081, "lr": 7.224249004333618e-05}
{"train_loss": 0.1821117252111435, "global_step": 96215, "epoch": 1081, "lr": 7.224197082673082e-05}
{"train_loss": 0.12878619134426117, "global_step": 96216, "epoch": 1081, "lr": 7.22414516071353e-05}
{"train_loss": 0.2087176889181137, "global_step": 96217, "epoch": 1081, "lr": 7.22409323845497e-05}
{"train_loss": 0.2159365713596344, "global_step": 96218, "epoch": 1081, "lr": 7.224041315897405e-05}
{"train_loss": 0.14669577777385712, "global_step": 96219, "epoch": 1081, "lr": 7.223989393040847e-05}
{"train_loss": 0.12682084739208221, "global_step": 96220, "epoch": 1081, "lr": 7.2239374698853e-05}
{"train_loss": 0.15734857320785522, "global_step": 96221, "epoch": 1081, "lr": 7.223885546430772e-05}
{"train_loss": 0.13941901922225952, "global_step": 96222, "epoch": 1081, "lr": 7.22383362267727e-05}
{"train_loss": 0.17024457454681396, "global_step": 96223, "epoch": 1081, "lr": 7.2237816986248e-05}
{"train_loss": 0.13712520897388458, "global_step": 96224, "epoch": 1081, "lr": 7.223729774273369e-05}
{"train_loss": 0.18932051956653595, "global_step": 96225, "epoch": 1081, "lr": 7.223677849622986e-05}
{"train_loss": 0.23274363577365875, "global_step": 96226, "epoch": 1081, "lr": 7.223625924673656e-05}
{"train_loss": 0.14098626375198364, "global_step": 96227, "epoch": 1081, "lr": 7.223573999425386e-05}
{"train_loss": 0.14524999260902405, "global_step": 96228, "epoch": 1081, "lr": 7.223522073878186e-05}
{"train_loss": 0.09163773059844971, "global_step": 96229, "epoch": 1081, "lr": 7.223470148032058e-05}
{"train_loss": 0.1905602216720581, "global_step": 96230, "epoch": 1081, "lr": 7.223418221887013e-05}
{"train_loss": 0.17069584131240845, "global_step": 96231, "epoch": 1081, "lr": 7.223366295443055e-05}
{"train_loss": 0.1364986151456833, "global_step": 96232, "epoch": 1081, "lr": 7.223314368700194e-05}
{"train_loss": 0.17758086323738098, "global_step": 96233, "epoch": 1081, "lr": 7.223262441658434e-05}
{"train_loss": 0.18265938758850098, "global_step": 96234, "epoch": 1081, "lr": 7.223210514317784e-05}
{"train_loss": 0.18984144926071167, "global_step": 96235, "epoch": 1081, "lr": 7.223158586678252e-05}
{"train_loss": 0.22336924076080322, "global_step": 96236, "epoch": 1081, "lr": 7.223106658739842e-05}
{"train_loss": 0.22635072469711304, "global_step": 96237, "epoch": 1081, "lr": 7.223054730502562e-05}
{"train_loss": 0.10662790387868881, "global_step": 96238, "epoch": 1081, "lr": 7.22300280196642e-05}
{"train_loss": 0.16721279919147491, "global_step": 96239, "epoch": 1081, "lr": 7.222950873131423e-05}
{"train_loss": 0.1948297768831253, "global_step": 96240, "epoch": 1081, "lr": 7.222898943997575e-05}
{"train_loss": 0.16801181435585022, "global_step": 96241, "epoch": 1081, "lr": 7.222847014564888e-05}
{"train_loss": 0.12657293677330017, "global_step": 96242, "epoch": 1081, "lr": 7.222795084833365e-05}
{"train_loss": 0.1982637345790863, "global_step": 96243, "epoch": 1081, "lr": 7.222743154803015e-05}
{"train_loss": 0.17684310674667358, "global_step": 96244, "epoch": 1081, "lr": 7.222691224473844e-05}
{"train_loss": 0.175995871424675, "global_step": 96245, "epoch": 1081, "lr": 7.222639293845858e-05}
{"train_loss": 0.1222827285528183, "global_step": 96246, "epoch": 1081, "lr": 7.222587362919067e-05}
{"train_loss": 0.22986453771591187, "global_step": 96247, "epoch": 1081, "lr": 7.222535431693475e-05}
{"train_loss": 0.19741840660572052, "global_step": 96248, "epoch": 1081, "lr": 7.222483500169091e-05}
{"train_loss": 0.16641193628311157, "global_step": 96249, "epoch": 1081, "lr": 7.222431568345922e-05}
{"train_loss": 0.21603193879127502, "global_step": 96250, "epoch": 1081, "lr": 7.222379636223973e-05}
{"train_loss": 0.1086161881685257, "global_step": 96251, "epoch": 1081, "lr": 7.222327703803252e-05}
{"train_loss": 0.18484067916870117, "global_step": 96252, "epoch": 1081, "lr": 7.222275771083766e-05}
{"train_loss": 0.15335166454315186, "global_step": 96253, "epoch": 1081, "lr": 7.222223838065522e-05}
{"train_loss": 0.18253834545612335, "global_step": 96254, "epoch": 1081, "lr": 7.222171904748529e-05}
{"train_loss": 0.13998664915561676, "global_step": 96255, "epoch": 1081, "lr": 7.22211997113279e-05}
{"train_loss": 0.1477966159582138, "global_step": 96256, "epoch": 1081, "lr": 7.222068037218316e-05}
{"train_loss": 0.2414097636938095, "global_step": 96257, "epoch": 1081, "lr": 7.22201610300511e-05}
{"train_loss": 0.24077844619750977, "global_step": 96258, "epoch": 1081, "lr": 7.221964168493183e-05}
{"train_loss": 0.16478699445724487, "global_step": 96259, "epoch": 1081, "lr": 7.221912233682537e-05}
{"train_loss": 0.11251690238714218, "global_step": 96260, "epoch": 1081, "lr": 7.221860298573184e-05}
{"train_loss": 0.2688613831996918, "global_step": 96261, "epoch": 1081, "lr": 7.221808363165129e-05}
{"train_loss": 0.275441974401474, "global_step": 96262, "epoch": 1081, "lr": 7.22175642745838e-05}
{"train_loss": 0.14191894233226776, "global_step": 96263, "epoch": 1081, "lr": 7.22170449145294e-05}
{"train_loss": 0.09208616614341736, "global_step": 96264, "epoch": 1081, "lr": 7.221652555148821e-05}
{"train_loss": 0.17028184235095978, "global_step": 96265, "epoch": 1081, "lr": 7.221600618546027e-05}
{"train_loss": 0.19615314900875092, "global_step": 96266, "epoch": 1081, "lr": 7.221548681644567e-05}
{"train_loss": 0.21175599098205566, "global_step": 96267, "epoch": 1081, "lr": 7.221496744444446e-05}
{"train_loss": 0.11020048707723618, "global_step": 96268, "epoch": 1081, "lr": 7.221444806945672e-05}
{"train_loss": 0.15224888920783997, "global_step": 96269, "epoch": 1081, "lr": 7.221392869148252e-05}
{"train_loss": 0.19937600195407867, "global_step": 96270, "epoch": 1081, "lr": 7.221340931052193e-05}
{"train_loss": 0.16890107095241547, "global_step": 96271, "epoch": 1081, "lr": 7.221288992657501e-05}
{"train_loss": 0.2026376575231552, "global_step": 96272, "epoch": 1081, "lr": 7.221237053964184e-05}
{"train_loss": 0.2177092581987381, "global_step": 96273, "epoch": 1081, "lr": 7.221185114972248e-05}
{"train_loss": 0.21203520894050598, "global_step": 96274, "epoch": 1081, "lr": 7.221133175681702e-05}
{"train_loss": 0.20210407674312592, "global_step": 96275, "epoch": 1081, "lr": 7.221081236092552e-05}
{"train_loss": 0.17969751358032227, "global_step": 96276, "epoch": 1081, "lr": 7.221029296204804e-05}
{"train_loss": 0.1699175387620926, "global_step": 96277, "epoch": 1081, "lr": 7.220977356018466e-05}
{"train_loss": 0.18598903715610504, "global_step": 96278, "epoch": 1081, "lr": 7.220925415533544e-05}
{"train_loss": 0.15185078978538513, "global_step": 96279, "epoch": 1081, "lr": 7.220873474750046e-05}
{"train_loss": 0.21116489171981812, "global_step": 96280, "epoch": 1081, "lr": 7.220821533667979e-05}
{"train_loss": 0.14467500150203705, "global_step": 96281, "epoch": 1081, "lr": 7.22076959228735e-05}
{"train_loss": 0.1209339052438736, "global_step": 96282, "epoch": 1081, "lr": 7.220717650608164e-05}
{"train_loss": 0.18625900149345398, "global_step": 96283, "epoch": 1081, "lr": 7.22066570863043e-05}
{"train_loss": 0.11671599000692368, "global_step": 96284, "epoch": 1081, "lr": 7.220613766354156e-05}
{"train_loss": 0.18887177109718323, "global_step": 96285, "epoch": 1081, "lr": 7.220561823779346e-05}
{"train_loss": 0.22119468450546265, "global_step": 96286, "epoch": 1081, "lr": 7.22050988090601e-05}
{"train_loss": 0.19306650757789612, "global_step": 96287, "epoch": 1081, "lr": 7.220457937734152e-05}
{"train_loss": 0.13515211641788483, "global_step": 96288, "epoch": 1081, "lr": 7.220405994263781e-05}
{"train_loss": 0.12078780680894852, "global_step": 96289, "epoch": 1081, "lr": 7.220354050494904e-05}
{"train_loss": 0.17395475506782532, "global_step": 96290, "epoch": 1081, "lr": 7.220302106427527e-05}
{"train_loss": 0.26397931575775146, "global_step": 96291, "epoch": 1081, "lr": 7.220250162061657e-05}
{"train_loss": 0.15503086149692535, "global_step": 96292, "epoch": 1081, "lr": 7.220198217397302e-05}
{"train_loss": 0.23573197424411774, "global_step": 96293, "epoch": 1081, "lr": 7.220146272434469e-05}
{"train_loss": 0.2405228465795517, "global_step": 96294, "epoch": 1081, "lr": 7.220094327173163e-05}
{"train_loss": 0.18890810012817383, "global_step": 96295, "epoch": 1081, "lr": 7.220042381613393e-05}
{"train_loss": 0.206239253282547, "global_step": 96296, "epoch": 1081, "lr": 7.219990435755166e-05}
{"train_loss": 0.17490988383802136, "global_step": 96297, "epoch": 1081, "lr": 7.219938489598488e-05, "val_loss": 4.568860054016113}
{"train_loss": 0.15870508551597595, "global_step": 96298, "epoch": 1082, "lr": 7.219886543143366e-05}
{"train_loss": 0.14629633724689484, "global_step": 96299, "epoch": 1082, "lr": 7.219834596389808e-05}
{"train_loss": 0.18834926187992096, "global_step": 96300, "epoch": 1082, "lr": 7.21978264933782e-05}
{"train_loss": 0.2000284641981125, "global_step": 96301, "epoch": 1082, "lr": 7.219730701987408e-05}
{"train_loss": 0.1728862226009369, "global_step": 96302, "epoch": 1082, "lr": 7.219678754338581e-05}
{"train_loss": 0.11815505474805832, "global_step": 96303, "epoch": 1082, "lr": 7.219626806391346e-05}
{"train_loss": 0.17974631488323212, "global_step": 96304, "epoch": 1082, "lr": 7.219574858145708e-05}
{"train_loss": 0.14502333104610443, "global_step": 96305, "epoch": 1082, "lr": 7.219522909601677e-05}
{"train_loss": 0.13853637874126434, "global_step": 96306, "epoch": 1082, "lr": 7.219470960759257e-05}
{"train_loss": 0.1655094176530838, "global_step": 96307, "epoch": 1082, "lr": 7.219419011618456e-05}
{"train_loss": 0.166772723197937, "global_step": 96308, "epoch": 1082, "lr": 7.219367062179282e-05}
{"train_loss": 0.1682838648557663, "global_step": 96309, "epoch": 1082, "lr": 7.21931511244174e-05}
{"train_loss": 0.19820180535316467, "global_step": 96310, "epoch": 1082, "lr": 7.21926316240584e-05}
{"train_loss": 0.11704277992248535, "global_step": 96311, "epoch": 1082, "lr": 7.219211212071586e-05}
{"train_loss": 0.29993826150894165, "global_step": 96312, "epoch": 1082, "lr": 7.219159261438985e-05}
{"train_loss": 0.16186274588108063, "global_step": 96313, "epoch": 1082, "lr": 7.219107310508046e-05}
{"train_loss": 0.12702497839927673, "global_step": 96314, "epoch": 1082, "lr": 7.219055359278775e-05}
{"train_loss": 0.2668071389198303, "global_step": 96315, "epoch": 1082, "lr": 7.21900340775118e-05}
{"train_loss": 0.14243119955062866, "global_step": 96316, "epoch": 1082, "lr": 7.218951455925265e-05}
{"train_loss": 0.11602946370840073, "global_step": 96317, "epoch": 1082, "lr": 7.218899503801042e-05}
{"train_loss": 0.22699253261089325, "global_step": 96318, "epoch": 1082, "lr": 7.218847551378512e-05}
{"train_loss": 0.14030563831329346, "global_step": 96319, "epoch": 1082, "lr": 7.218795598657687e-05}
{"train_loss": 0.21279871463775635, "global_step": 96320, "epoch": 1082, "lr": 7.218743645638571e-05}
{"train_loss": 0.19242098927497864, "global_step": 96321, "epoch": 1082, "lr": 7.218691692321172e-05}
{"train_loss": 0.1317831575870514, "global_step": 96322, "epoch": 1082, "lr": 7.218639738705497e-05}
{"train_loss": 0.08989501744508743, "global_step": 96323, "epoch": 1082, "lr": 7.218587784791554e-05}
{"train_loss": 0.14326758682727814, "global_step": 96324, "epoch": 1082, "lr": 7.218535830579347e-05}
{"train_loss": 0.19087006151676178, "global_step": 96325, "epoch": 1082, "lr": 7.218483876068887e-05}
{"train_loss": 0.16322042047977448, "global_step": 96326, "epoch": 1082, "lr": 7.218431921260178e-05}
{"train_loss": 0.3192395269870758, "global_step": 96327, "epoch": 1082, "lr": 7.218379966153227e-05}
{"train_loss": 0.19047875702381134, "global_step": 96328, "epoch": 1082, "lr": 7.218328010748043e-05}
{"train_loss": 0.13762150704860687, "global_step": 96329, "epoch": 1082, "lr": 7.21827605504463e-05}
{"train_loss": 0.11215724796056747, "global_step": 96330, "epoch": 1082, "lr": 7.218224099043e-05}
{"train_loss": 0.1871916502714157, "global_step": 96331, "epoch": 1082, "lr": 7.218172142743154e-05}
{"train_loss": 0.09890240430831909, "global_step": 96332, "epoch": 1082, "lr": 7.218120186145104e-05}
{"train_loss": 0.17681846022605896, "global_step": 96333, "epoch": 1082, "lr": 7.218068229248852e-05}
{"train_loss": 0.19089576601982117, "global_step": 96334, "epoch": 1082, "lr": 7.21801627205441e-05}
{"train_loss": 0.23990404605865479, "global_step": 96335, "epoch": 1082, "lr": 7.217964314561782e-05}
{"train_loss": 0.18528667092323303, "global_step": 96336, "epoch": 1082, "lr": 7.217912356770975e-05}
{"train_loss": 0.15984861552715302, "global_step": 96337, "epoch": 1082, "lr": 7.217860398681998e-05}
{"train_loss": 0.14273063838481903, "global_step": 96338, "epoch": 1082, "lr": 7.217808440294856e-05}
{"train_loss": 0.20752698183059692, "global_step": 96339, "epoch": 1082, "lr": 7.217756481609557e-05}
{"train_loss": 0.15793150663375854, "global_step": 96340, "epoch": 1082, "lr": 7.217704522626107e-05}
{"train_loss": 0.22128325700759888, "global_step": 96341, "epoch": 1082, "lr": 7.217652563344513e-05}
{"train_loss": 0.15986807644367218, "global_step": 96342, "epoch": 1082, "lr": 7.217600603764785e-05}
{"train_loss": 0.15903617441654205, "global_step": 96343, "epoch": 1082, "lr": 7.217548643886927e-05}
{"train_loss": 0.19648057222366333, "global_step": 96344, "epoch": 1082, "lr": 7.217496683710945e-05}
{"train_loss": 0.21409879624843597, "global_step": 96345, "epoch": 1082, "lr": 7.21744472323685e-05}
{"train_loss": 0.07811277359724045, "global_step": 96346, "epoch": 1082, "lr": 7.217392762464645e-05}
{"train_loss": 0.1504482924938202, "global_step": 96347, "epoch": 1082, "lr": 7.21734080139434e-05}
{"train_loss": 0.15575505793094635, "global_step": 96348, "epoch": 1082, "lr": 7.217288840025939e-05}
{"train_loss": 0.16759614646434784, "global_step": 96349, "epoch": 1082, "lr": 7.21723687835945e-05}
{"train_loss": 0.1487431675195694, "global_step": 96350, "epoch": 1082, "lr": 7.217184916394882e-05}
{"train_loss": 0.12365040183067322, "global_step": 96351, "epoch": 1082, "lr": 7.21713295413224e-05}
{"train_loss": 0.21448208391666412, "global_step": 96352, "epoch": 1082, "lr": 7.217080991571531e-05}
{"train_loss": 0.13652120530605316, "global_step": 96353, "epoch": 1082, "lr": 7.217029028712762e-05}
{"train_loss": 0.18729275465011597, "global_step": 96354, "epoch": 1082, "lr": 7.216977065555943e-05}
{"train_loss": 0.23637118935585022, "global_step": 96355, "epoch": 1082, "lr": 7.216925102101076e-05}
{"train_loss": 0.14779967069625854, "global_step": 96356, "epoch": 1082, "lr": 7.216873138348172e-05}
{"train_loss": 0.14992377161979675, "global_step": 96357, "epoch": 1082, "lr": 7.216821174297236e-05}
{"train_loss": 0.150784432888031, "global_step": 96358, "epoch": 1082, "lr": 7.216769209948275e-05}
{"train_loss": 0.13026538491249084, "global_step": 96359, "epoch": 1082, "lr": 7.216717245301296e-05}
{"train_loss": 0.2859584093093872, "global_step": 96360, "epoch": 1082, "lr": 7.216665280356306e-05}
{"train_loss": 0.15960794687271118, "global_step": 96361, "epoch": 1082, "lr": 7.216613315113314e-05}
{"train_loss": 0.16796384751796722, "global_step": 96362, "epoch": 1082, "lr": 7.216561349572324e-05}
{"train_loss": 0.16498753428459167, "global_step": 96363, "epoch": 1082, "lr": 7.216509383733346e-05}
{"train_loss": 0.12320400774478912, "global_step": 96364, "epoch": 1082, "lr": 7.216457417596385e-05}
{"train_loss": 0.13731412589550018, "global_step": 96365, "epoch": 1082, "lr": 7.216405451161447e-05}
{"train_loss": 0.22122521698474884, "global_step": 96366, "epoch": 1082, "lr": 7.216353484428542e-05}
{"train_loss": 0.11767943203449249, "global_step": 96367, "epoch": 1082, "lr": 7.216301517397673e-05}
{"train_loss": 0.17543651163578033, "global_step": 96368, "epoch": 1082, "lr": 7.216249550068851e-05}
{"train_loss": 0.16162647306919098, "global_step": 96369, "epoch": 1082, "lr": 7.216197582442081e-05}
{"train_loss": 0.16424046456813812, "global_step": 96370, "epoch": 1082, "lr": 7.21614561451737e-05}
{"train_loss": 0.1910632848739624, "global_step": 96371, "epoch": 1082, "lr": 7.216093646294725e-05}
{"train_loss": 0.24357235431671143, "global_step": 96372, "epoch": 1082, "lr": 7.216041677774154e-05}
{"train_loss": 0.2161407768726349, "global_step": 96373, "epoch": 1082, "lr": 7.215989708955662e-05}
{"train_loss": 0.2041180282831192, "global_step": 96374, "epoch": 1082, "lr": 7.215937739839258e-05}
{"train_loss": 0.15372829139232635, "global_step": 96375, "epoch": 1082, "lr": 7.215885770424948e-05}
{"train_loss": 0.12811854481697083, "global_step": 96376, "epoch": 1082, "lr": 7.21583380071274e-05}
{"train_loss": 0.19938279688358307, "global_step": 96377, "epoch": 1082, "lr": 7.215781830702637e-05}
{"train_loss": 0.16284307837486267, "global_step": 96378, "epoch": 1082, "lr": 7.215729860394652e-05}
{"train_loss": 0.1574251353740692, "global_step": 96379, "epoch": 1082, "lr": 7.215677889788788e-05}
{"train_loss": 0.18693874776363373, "global_step": 96380, "epoch": 1082, "lr": 7.215625918885055e-05}
{"train_loss": 0.17122331261634827, "global_step": 96381, "epoch": 1082, "lr": 7.215573947683456e-05}
{"train_loss": 0.21833328902721405, "global_step": 96382, "epoch": 1082, "lr": 7.215521976184002e-05}
{"train_loss": 0.16304966807365417, "global_step": 96383, "epoch": 1082, "lr": 7.215470004386698e-05}
{"train_loss": 0.19520919024944305, "global_step": 96384, "epoch": 1082, "lr": 7.21541803229155e-05}
{"train_loss": 0.1764736920595169, "global_step": 96385, "epoch": 1082, "lr": 7.215366059898565e-05}
{"train_loss": 0.1729599310273535, "global_step": 96386, "epoch": 1082, "lr": 7.215314087207752e-05, "val_loss": 4.441161155700684}
{"train_loss": 0.11663555353879929, "global_step": 96387, "epoch": 1083, "lr": 7.215262114219119e-05}
{"train_loss": 0.2103644609451294, "global_step": 96388, "epoch": 1083, "lr": 7.215210140932668e-05}
{"train_loss": 0.10706036537885666, "global_step": 96389, "epoch": 1083, "lr": 7.21515816734841e-05}
{"train_loss": 0.17908811569213867, "global_step": 96390, "epoch": 1083, "lr": 7.215106193466351e-05}
{"train_loss": 0.19733455777168274, "global_step": 96391, "epoch": 1083, "lr": 7.215054219286499e-05}
{"train_loss": 0.18812841176986694, "global_step": 96392, "epoch": 1083, "lr": 7.215002244808859e-05}
{"train_loss": 0.1434440016746521, "global_step": 96393, "epoch": 1083, "lr": 7.214950270033438e-05}
{"train_loss": 0.22993458807468414, "global_step": 96394, "epoch": 1083, "lr": 7.214898294960246e-05}
{"train_loss": 0.15378165245056152, "global_step": 96395, "epoch": 1083, "lr": 7.214846319589286e-05}
{"train_loss": 0.11058811843395233, "global_step": 96396, "epoch": 1083, "lr": 7.214794343920568e-05}
{"train_loss": 0.2016756683588028, "global_step": 96397, "epoch": 1083, "lr": 7.214742367954097e-05}
{"train_loss": 0.13180400431156158, "global_step": 96398, "epoch": 1083, "lr": 7.214690391689882e-05}
{"train_loss": 0.11296062916517258, "global_step": 96399, "epoch": 1083, "lr": 7.214638415127928e-05}
{"train_loss": 0.16046419739723206, "global_step": 96400, "epoch": 1083, "lr": 7.214586438268244e-05}
{"train_loss": 0.11783555895090103, "global_step": 96401, "epoch": 1083, "lr": 7.214534461110835e-05}
{"train_loss": 0.17000341415405273, "global_step": 96402, "epoch": 1083, "lr": 7.21448248365571e-05}
{"train_loss": 0.11587470769882202, "global_step": 96403, "epoch": 1083, "lr": 7.214430505902874e-05}
{"train_loss": 0.1755090057849884, "global_step": 96404, "epoch": 1083, "lr": 7.214378527852335e-05}
{"train_loss": 0.15101371705532074, "global_step": 96405, "epoch": 1083, "lr": 7.214326549504098e-05}
{"train_loss": 0.2156788557767868, "global_step": 96406, "epoch": 1083, "lr": 7.214274570858173e-05}
{"train_loss": 0.16395936906337738, "global_step": 96407, "epoch": 1083, "lr": 7.214222591914567e-05}
{"train_loss": 0.13862141966819763, "global_step": 96408, "epoch": 1083, "lr": 7.214170612673285e-05}
{"train_loss": 0.14707383513450623, "global_step": 96409, "epoch": 1083, "lr": 7.214118633134334e-05}
{"train_loss": 0.22551001608371735, "global_step": 96410, "epoch": 1083, "lr": 7.214066653297722e-05}
{"train_loss": 0.1523805856704712, "global_step": 96411, "epoch": 1083, "lr": 7.214014673163456e-05}
{"train_loss": 0.08538006246089935, "global_step": 96412, "epoch": 1083, "lr": 7.213962692731544e-05}
{"train_loss": 0.15526998043060303, "global_step": 96413, "epoch": 1083, "lr": 7.213910712001988e-05}
{"train_loss": 0.1700746864080429, "global_step": 96414, "epoch": 1083, "lr": 7.2138587309748e-05}
{"train_loss": 0.12757135927677155, "global_step": 96415, "epoch": 1083, "lr": 7.213806749649988e-05}
{"train_loss": 0.12913630902767181, "global_step": 96416, "epoch": 1083, "lr": 7.213754768027555e-05}
{"train_loss": 0.08870311081409454, "global_step": 96417, "epoch": 1083, "lr": 7.21370278610751e-05}
{"train_loss": 0.12682190537452698, "global_step": 96418, "epoch": 1083, "lr": 7.21365080388986e-05}
{"train_loss": 0.11664288491010666, "global_step": 96419, "epoch": 1083, "lr": 7.21359882137461e-05}
{"train_loss": 0.12176672369241714, "global_step": 96420, "epoch": 1083, "lr": 7.213546838561771e-05}
{"train_loss": 0.1553221195936203, "global_step": 96421, "epoch": 1083, "lr": 7.213494855451348e-05}
{"train_loss": 0.19369426369667053, "global_step": 96422, "epoch": 1083, "lr": 7.213442872043346e-05}
{"train_loss": 0.12213689833879471, "global_step": 96423, "epoch": 1083, "lr": 7.213390888337773e-05}
{"train_loss": 0.1550629734992981, "global_step": 96424, "epoch": 1083, "lr": 7.213338904334638e-05}
{"train_loss": 0.15624623000621796, "global_step": 96425, "epoch": 1083, "lr": 7.213286920033946e-05}
{"train_loss": 0.13892200589179993, "global_step": 96426, "epoch": 1083, "lr": 7.213234935435704e-05}
{"train_loss": 0.1613440066576004, "global_step": 96427, "epoch": 1083, "lr": 7.21318295053992e-05}
{"train_loss": 0.20657901465892792, "global_step": 96428, "epoch": 1083, "lr": 7.2131309653466e-05}
{"train_loss": 0.13604269921779633, "global_step": 96429, "epoch": 1083, "lr": 7.213078979855752e-05}
{"train_loss": 0.10876966267824173, "global_step": 96430, "epoch": 1083, "lr": 7.213026994067382e-05}
{"train_loss": 0.22294117510318756, "global_step": 96431, "epoch": 1083, "lr": 7.212975007981498e-05}
{"train_loss": 0.2581771910190582, "global_step": 96432, "epoch": 1083, "lr": 7.212923021598106e-05}
{"train_loss": 0.1026117131114006, "global_step": 96433, "epoch": 1083, "lr": 7.212871034917214e-05}
{"train_loss": 0.1850770264863968, "global_step": 96434, "epoch": 1083, "lr": 7.212819047938828e-05}
{"train_loss": 0.18029451370239258, "global_step": 96435, "epoch": 1083, "lr": 7.212767060662956e-05}
{"train_loss": 0.1562431901693344, "global_step": 96436, "epoch": 1083, "lr": 7.212715073089605e-05}
{"train_loss": 0.1722937375307083, "global_step": 96437, "epoch": 1083, "lr": 7.212663085218781e-05}
{"train_loss": 0.19435366988182068, "global_step": 96438, "epoch": 1083, "lr": 7.21261109705049e-05}
{"train_loss": 0.23716053366661072, "global_step": 96439, "epoch": 1083, "lr": 7.212559108584743e-05}
{"train_loss": 0.12217776477336884, "global_step": 96440, "epoch": 1083, "lr": 7.212507119821543e-05}
{"train_loss": 0.09008410573005676, "global_step": 96441, "epoch": 1083, "lr": 7.212455130760898e-05}
{"train_loss": 0.15508390963077545, "global_step": 96442, "epoch": 1083, "lr": 7.212403141402816e-05}
{"train_loss": 0.23989814519882202, "global_step": 96443, "epoch": 1083, "lr": 7.212351151747303e-05}
{"train_loss": 0.1507154405117035, "global_step": 96444, "epoch": 1083, "lr": 7.212299161794365e-05}
{"train_loss": 0.13901905715465546, "global_step": 96445, "epoch": 1083, "lr": 7.212247171544013e-05}
{"train_loss": 0.12507736682891846, "global_step": 96446, "epoch": 1083, "lr": 7.212195180996251e-05}
{"train_loss": 0.09417994320392609, "global_step": 96447, "epoch": 1083, "lr": 7.212143190151084e-05}
{"train_loss": 0.19519421458244324, "global_step": 96448, "epoch": 1083, "lr": 7.212091199008522e-05}
{"train_loss": 0.16286051273345947, "global_step": 96449, "epoch": 1083, "lr": 7.212039207568573e-05}
{"train_loss": 0.18287596106529236, "global_step": 96450, "epoch": 1083, "lr": 7.211987215831241e-05}
{"train_loss": 0.21662701666355133, "global_step": 96451, "epoch": 1083, "lr": 7.211935223796535e-05}
{"train_loss": 0.17272454500198364, "global_step": 96452, "epoch": 1083, "lr": 7.21188323146446e-05}
{"train_loss": 0.15961360931396484, "global_step": 96453, "epoch": 1083, "lr": 7.211831238835026e-05}
{"train_loss": 0.1561063975095749, "global_step": 96454, "epoch": 1083, "lr": 7.211779245908236e-05}
{"train_loss": 0.18034964799880981, "global_step": 96455, "epoch": 1083, "lr": 7.2117272526841e-05}
{"train_loss": 0.16146835684776306, "global_step": 96456, "epoch": 1083, "lr": 7.211675259162625e-05}
{"train_loss": 0.22220994532108307, "global_step": 96457, "epoch": 1083, "lr": 7.211623265343818e-05}
{"train_loss": 0.17559780180454254, "global_step": 96458, "epoch": 1083, "lr": 7.211571271227683e-05}
{"train_loss": 0.19499678909778595, "global_step": 96459, "epoch": 1083, "lr": 7.211519276814232e-05}
{"train_loss": 0.23500610888004303, "global_step": 96460, "epoch": 1083, "lr": 7.211467282103468e-05}
{"train_loss": 0.15011286735534668, "global_step": 96461, "epoch": 1083, "lr": 7.2114152870954e-05}
{"train_loss": 0.1301289200782776, "global_step": 96462, "epoch": 1083, "lr": 7.211363291790032e-05}
{"train_loss": 0.1505645513534546, "global_step": 96463, "epoch": 1083, "lr": 7.211311296187375e-05}
{"train_loss": 0.27792513370513916, "global_step": 96464, "epoch": 1083, "lr": 7.211259300287432e-05}
{"train_loss": 0.18081842362880707, "global_step": 96465, "epoch": 1083, "lr": 7.211207304090213e-05}
{"train_loss": 0.1497492790222168, "global_step": 96466, "epoch": 1083, "lr": 7.211155307595725e-05}
{"train_loss": 0.25612232089042664, "global_step": 96467, "epoch": 1083, "lr": 7.211103310803973e-05}
{"train_loss": 0.19023166596889496, "global_step": 96468, "epoch": 1083, "lr": 7.211051313714965e-05}
{"train_loss": 0.1929766684770584, "global_step": 96469, "epoch": 1083, "lr": 7.21099931632871e-05}
{"train_loss": 0.20261706411838531, "global_step": 96470, "epoch": 1083, "lr": 7.210947318645211e-05}
{"train_loss": 0.17244572937488556, "global_step": 96471, "epoch": 1083, "lr": 7.210895320664478e-05}
{"train_loss": 0.1498149335384369, "global_step": 96472, "epoch": 1083, "lr": 7.210843322386518e-05}
{"train_loss": 0.19065997004508972, "global_step": 96473, "epoch": 1083, "lr": 7.210791323811335e-05}
{"train_loss": 0.2144680768251419, "global_step": 96474, "epoch": 1083, "lr": 7.21073932493894e-05}
{"train_loss": 0.1666976222663783, "global_step": 96475, "epoch": 1083, "lr": 7.210687325769337e-05, "val_loss": 4.786978721618652}
{"train_loss": 0.2746042013168335, "global_step": 96476, "epoch": 1084, "lr": 7.210635326302535e-05}
{"train_loss": 0.20689398050308228, "global_step": 96477, "epoch": 1084, "lr": 7.210583326538538e-05}
{"train_loss": 0.18653371930122375, "global_step": 96478, "epoch": 1084, "lr": 7.210531326477357e-05}
{"train_loss": 0.14968900382518768, "global_step": 96479, "epoch": 1084, "lr": 7.210479326118997e-05}
{"train_loss": 0.16858446598052979, "global_step": 96480, "epoch": 1084, "lr": 7.210427325463464e-05}
{"train_loss": 0.21411371231079102, "global_step": 96481, "epoch": 1084, "lr": 7.210375324510767e-05}
{"train_loss": 0.13847331702709198, "global_step": 96482, "epoch": 1084, "lr": 7.210323323260911e-05}
{"train_loss": 0.18334852159023285, "global_step": 96483, "epoch": 1084, "lr": 7.210271321713906e-05}
{"train_loss": 0.1940886676311493, "global_step": 96484, "epoch": 1084, "lr": 7.210219319869754e-05}
{"train_loss": 0.188789963722229, "global_step": 96485, "epoch": 1084, "lr": 7.210167317728467e-05}
{"train_loss": 0.23090560734272003, "global_step": 96486, "epoch": 1084, "lr": 7.210115315290047e-05}
{"train_loss": 0.14360105991363525, "global_step": 96487, "epoch": 1084, "lr": 7.210063312554506e-05}
{"train_loss": 0.17451876401901245, "global_step": 96488, "epoch": 1084, "lr": 7.21001130952185e-05}
{"train_loss": 0.16812290251255035, "global_step": 96489, "epoch": 1084, "lr": 7.209959306192084e-05}
{"train_loss": 0.1826622188091278, "global_step": 96490, "epoch": 1084, "lr": 7.209907302565215e-05}
{"train_loss": 0.15148454904556274, "global_step": 96491, "epoch": 1084, "lr": 7.209855298641252e-05}
{"train_loss": 0.12457039952278137, "global_step": 96492, "epoch": 1084, "lr": 7.209803294420201e-05}
{"train_loss": 0.1562834233045578, "global_step": 96493, "epoch": 1084, "lr": 7.209751289902068e-05}
{"train_loss": 0.13860361278057098, "global_step": 96494, "epoch": 1084, "lr": 7.209699285086861e-05}
{"train_loss": 0.1864996999502182, "global_step": 96495, "epoch": 1084, "lr": 7.209647279974587e-05}
{"train_loss": 0.1585279405117035, "global_step": 96496, "epoch": 1084, "lr": 7.209595274565253e-05}
{"train_loss": 0.13216671347618103, "global_step": 96497, "epoch": 1084, "lr": 7.209543268858866e-05}
{"train_loss": 0.16120529174804688, "global_step": 96498, "epoch": 1084, "lr": 7.209491262855434e-05}
{"train_loss": 0.22170837223529816, "global_step": 96499, "epoch": 1084, "lr": 7.209439256554961e-05}
{"train_loss": 0.10722395777702332, "global_step": 96500, "epoch": 1084, "lr": 7.209387249957457e-05}
{"train_loss": 0.1927926242351532, "global_step": 96501, "epoch": 1084, "lr": 7.209335243062927e-05}
{"train_loss": 0.09927213937044144, "global_step": 96502, "epoch": 1084, "lr": 7.209283235871379e-05}
{"train_loss": 0.20141659677028656, "global_step": 96503, "epoch": 1084, "lr": 7.20923122838282e-05}
{"train_loss": 0.17054511606693268, "global_step": 96504, "epoch": 1084, "lr": 7.209179220597256e-05}
{"train_loss": 0.2064271867275238, "global_step": 96505, "epoch": 1084, "lr": 7.209127212514694e-05}
{"train_loss": 0.1705905944108963, "global_step": 96506, "epoch": 1084, "lr": 7.209075204135144e-05}
{"train_loss": 0.20037560164928436, "global_step": 96507, "epoch": 1084, "lr": 7.209023195458611e-05}
{"train_loss": 0.29766780138015747, "global_step": 96508, "epoch": 1084, "lr": 7.208971186485101e-05}
{"train_loss": 0.17210650444030762, "global_step": 96509, "epoch": 1084, "lr": 7.20891917721462e-05}
{"train_loss": 0.21846385300159454, "global_step": 96510, "epoch": 1084, "lr": 7.208867167647178e-05}
{"train_loss": 0.2063789665699005, "global_step": 96511, "epoch": 1084, "lr": 7.208815157782781e-05}
{"train_loss": 0.22636133432388306, "global_step": 96512, "epoch": 1084, "lr": 7.208763147621435e-05}
{"train_loss": 0.23473939299583435, "global_step": 96513, "epoch": 1084, "lr": 7.208711137163148e-05}
{"train_loss": 0.15163134038448334, "global_step": 96514, "epoch": 1084, "lr": 7.208659126407928e-05}
{"train_loss": 0.2101362943649292, "global_step": 96515, "epoch": 1084, "lr": 7.20860711535578e-05}
{"train_loss": 0.1406836360692978, "global_step": 96516, "epoch": 1084, "lr": 7.208555104006712e-05}
{"train_loss": 0.10440780222415924, "global_step": 96517, "epoch": 1084, "lr": 7.20850309236073e-05}
{"train_loss": 0.16463367640972137, "global_step": 96518, "epoch": 1084, "lr": 7.208451080417842e-05}
{"train_loss": 0.15467359125614166, "global_step": 96519, "epoch": 1084, "lr": 7.208399068178056e-05}
{"train_loss": 0.17245803773403168, "global_step": 96520, "epoch": 1084, "lr": 7.208347055641376e-05}
{"train_loss": 0.17533567547798157, "global_step": 96521, "epoch": 1084, "lr": 7.20829504280781e-05}
{"train_loss": 0.10900649428367615, "global_step": 96522, "epoch": 1084, "lr": 7.208243029677365e-05}
{"train_loss": 0.17860928177833557, "global_step": 96523, "epoch": 1084, "lr": 7.20819101625005e-05}
{"train_loss": 0.09775091707706451, "global_step": 96524, "epoch": 1084, "lr": 7.208139002525872e-05}
{"train_loss": 0.08980295062065125, "global_step": 96525, "epoch": 1084, "lr": 7.208086988504834e-05}
{"train_loss": 0.15476076304912567, "global_step": 96526, "epoch": 1084, "lr": 7.208034974186949e-05}
{"train_loss": 0.13003741204738617, "global_step": 96527, "epoch": 1084, "lr": 7.207982959572217e-05}
{"train_loss": 0.15487468242645264, "global_step": 96528, "epoch": 1084, "lr": 7.207930944660651e-05}
{"train_loss": 0.18046867847442627, "global_step": 96529, "epoch": 1084, "lr": 7.207878929452255e-05}
{"train_loss": 0.10718248039484024, "global_step": 96530, "epoch": 1084, "lr": 7.207826913947035e-05}
{"train_loss": 0.187026247382164, "global_step": 96531, "epoch": 1084, "lr": 7.207774898145002e-05}
{"train_loss": 0.15072444081306458, "global_step": 96532, "epoch": 1084, "lr": 7.207722882046158e-05}
{"train_loss": 0.21372771263122559, "global_step": 96533, "epoch": 1084, "lr": 7.207670865650515e-05}
{"train_loss": 0.16535134613513947, "global_step": 96534, "epoch": 1084, "lr": 7.207618848958076e-05}
{"train_loss": 0.17353172600269318, "global_step": 96535, "epoch": 1084, "lr": 7.207566831968851e-05}
{"train_loss": 0.12027949094772339, "global_step": 96536, "epoch": 1084, "lr": 7.207514814682844e-05}
{"train_loss": 0.16152238845825195, "global_step": 96537, "epoch": 1084, "lr": 7.207462797100065e-05}
{"train_loss": 0.14110678434371948, "global_step": 96538, "epoch": 1084, "lr": 7.207410779220519e-05}
{"train_loss": 0.20569024980068207, "global_step": 96539, "epoch": 1084, "lr": 7.207358761044214e-05}
{"train_loss": 0.18460266292095184, "global_step": 96540, "epoch": 1084, "lr": 7.207306742571154e-05}
{"train_loss": 0.16397027671337128, "global_step": 96541, "epoch": 1084, "lr": 7.207254723801351e-05}
{"train_loss": 0.17918290197849274, "global_step": 96542, "epoch": 1084, "lr": 7.20720270473481e-05}
{"train_loss": 0.1802626997232437, "global_step": 96543, "epoch": 1084, "lr": 7.207150685371535e-05}
{"train_loss": 0.17648576200008392, "global_step": 96544, "epoch": 1084, "lr": 7.207098665711536e-05}
{"train_loss": 0.10969969630241394, "global_step": 96545, "epoch": 1084, "lr": 7.207046645754821e-05}
{"train_loss": 0.1197582259774208, "global_step": 96546, "epoch": 1084, "lr": 7.206994625501395e-05}
{"train_loss": 0.16470666229724884, "global_step": 96547, "epoch": 1084, "lr": 7.206942604951265e-05}
{"train_loss": 0.14598190784454346, "global_step": 96548, "epoch": 1084, "lr": 7.206890584104439e-05}
{"train_loss": 0.10850352793931961, "global_step": 96549, "epoch": 1084, "lr": 7.206838562960922e-05}
{"train_loss": 0.17400217056274414, "global_step": 96550, "epoch": 1084, "lr": 7.206786541520724e-05}
{"train_loss": 0.16740871965885162, "global_step": 96551, "epoch": 1084, "lr": 7.206734519783851e-05}
{"train_loss": 0.2549229860305786, "global_step": 96552, "epoch": 1084, "lr": 7.206682497750308e-05}
{"train_loss": 0.176059290766716, "global_step": 96553, "epoch": 1084, "lr": 7.206630475420104e-05}
{"train_loss": 0.2147715836763382, "global_step": 96554, "epoch": 1084, "lr": 7.206578452793245e-05}
{"train_loss": 0.16894987225532532, "global_step": 96555, "epoch": 1084, "lr": 7.20652642986974e-05}
{"train_loss": 0.2009638100862503, "global_step": 96556, "epoch": 1084, "lr": 7.206474406649594e-05}
{"train_loss": 0.09052864462137222, "global_step": 96557, "epoch": 1084, "lr": 7.206422383132814e-05}
{"train_loss": 0.14312413334846497, "global_step": 96558, "epoch": 1084, "lr": 7.206370359319406e-05}
{"train_loss": 0.12954887747764587, "global_step": 96559, "epoch": 1084, "lr": 7.206318335209379e-05}
{"train_loss": 0.13561584055423737, "global_step": 96560, "epoch": 1084, "lr": 7.20626631080274e-05}
{"train_loss": 0.16017988324165344, "global_step": 96561, "epoch": 1084, "lr": 7.206214286099495e-05}
{"train_loss": 0.09537644684314728, "global_step": 96562, "epoch": 1084, "lr": 7.206162261099652e-05}
{"train_loss": 0.09682752937078476, "global_step": 96563, "epoch": 1084, "lr": 7.206110235803217e-05}
{"train_loss": 0.166943815149618, "global_step": 96564, "epoch": 1084, "lr": 7.206058210210196e-05, "val_loss": 4.641944408416748}
{"train_loss": 0.1084827408194542, "global_step": 96565, "epoch": 1085, "lr": 7.2060061843206e-05}
{"train_loss": 0.19650611281394958, "global_step": 96566, "epoch": 1085, "lr": 7.205954158134431e-05}
{"train_loss": 0.21553170680999756, "global_step": 96567, "epoch": 1085, "lr": 7.2059021316517e-05}
{"train_loss": 0.13399529457092285, "global_step": 96568, "epoch": 1085, "lr": 7.205850104872411e-05}
{"train_loss": 0.13477422297000885, "global_step": 96569, "epoch": 1085, "lr": 7.205798077796574e-05}
{"train_loss": 0.17471309006214142, "global_step": 96570, "epoch": 1085, "lr": 7.205746050424194e-05}
{"train_loss": 0.16287767887115479, "global_step": 96571, "epoch": 1085, "lr": 7.205694022755277e-05}
{"train_loss": 0.21556727588176727, "global_step": 96572, "epoch": 1085, "lr": 7.205641994789832e-05}
{"train_loss": 0.12788331508636475, "global_step": 96573, "epoch": 1085, "lr": 7.205589966527866e-05}
{"train_loss": 0.14488105475902557, "global_step": 96574, "epoch": 1085, "lr": 7.205537937969384e-05}
{"train_loss": 0.11768101900815964, "global_step": 96575, "epoch": 1085, "lr": 7.205485909114395e-05}
{"train_loss": 0.23116113245487213, "global_step": 96576, "epoch": 1085, "lr": 7.205433879962907e-05}
{"train_loss": 0.20942775905132294, "global_step": 96577, "epoch": 1085, "lr": 7.205381850514923e-05}
{"train_loss": 0.15247493982315063, "global_step": 96578, "epoch": 1085, "lr": 7.205329820770452e-05}
{"train_loss": 0.1942172795534134, "global_step": 96579, "epoch": 1085, "lr": 7.205277790729504e-05}
{"train_loss": 0.13671104609966278, "global_step": 96580, "epoch": 1085, "lr": 7.20522576039208e-05}
{"train_loss": 0.19997075200080872, "global_step": 96581, "epoch": 1085, "lr": 7.205173729758192e-05}
{"train_loss": 0.13143576681613922, "global_step": 96582, "epoch": 1085, "lr": 7.205121698827845e-05}
{"train_loss": 0.1648799628019333, "global_step": 96583, "epoch": 1085, "lr": 7.205069667601046e-05}
{"train_loss": 0.1588229387998581, "global_step": 96584, "epoch": 1085, "lr": 7.205017636077802e-05}
{"train_loss": 0.24865685403347015, "global_step": 96585, "epoch": 1085, "lr": 7.204965604258121e-05}
{"train_loss": 0.1719537377357483, "global_step": 96586, "epoch": 1085, "lr": 7.204913572142009e-05}
{"train_loss": 0.12991555035114288, "global_step": 96587, "epoch": 1085, "lr": 7.204861539729474e-05}
{"train_loss": 0.1535634994506836, "global_step": 96588, "epoch": 1085, "lr": 7.20480950702052e-05}
{"train_loss": 0.11978774517774582, "global_step": 96589, "epoch": 1085, "lr": 7.204757474015159e-05}
{"train_loss": 0.19870714843273163, "global_step": 96590, "epoch": 1085, "lr": 7.204705440713395e-05}
{"train_loss": 0.1841275542974472, "global_step": 96591, "epoch": 1085, "lr": 7.204653407115234e-05}
{"train_loss": 0.11232104897499084, "global_step": 96592, "epoch": 1085, "lr": 7.204601373220684e-05}
{"train_loss": 0.16864261031150818, "global_step": 96593, "epoch": 1085, "lr": 7.204549339029752e-05}
{"train_loss": 0.20419374108314514, "global_step": 96594, "epoch": 1085, "lr": 7.204497304542446e-05}
{"train_loss": 0.18094931542873383, "global_step": 96595, "epoch": 1085, "lr": 7.204445269758773e-05}
{"train_loss": 0.16320472955703735, "global_step": 96596, "epoch": 1085, "lr": 7.204393234678737e-05}
{"train_loss": 0.15765047073364258, "global_step": 96597, "epoch": 1085, "lr": 7.204341199302349e-05}
{"train_loss": 0.19523528218269348, "global_step": 96598, "epoch": 1085, "lr": 7.204289163629614e-05}
{"train_loss": 0.16570362448692322, "global_step": 96599, "epoch": 1085, "lr": 7.204237127660538e-05}
{"train_loss": 0.11397607624530792, "global_step": 96600, "epoch": 1085, "lr": 7.20418509139513e-05}
{"train_loss": 0.17103977501392365, "global_step": 96601, "epoch": 1085, "lr": 7.204133054833397e-05}
{"train_loss": 0.18630529940128326, "global_step": 96602, "epoch": 1085, "lr": 7.204081017975344e-05}
{"train_loss": 0.12745290994644165, "global_step": 96603, "epoch": 1085, "lr": 7.204028980820979e-05}
{"train_loss": 0.08709566295146942, "global_step": 96604, "epoch": 1085, "lr": 7.203976943370309e-05}
{"train_loss": 0.19539573788642883, "global_step": 96605, "epoch": 1085, "lr": 7.203924905623342e-05}
{"train_loss": 0.1533016562461853, "global_step": 96606, "epoch": 1085, "lr": 7.203872867580085e-05}
{"train_loss": 0.15035037696361542, "global_step": 96607, "epoch": 1085, "lr": 7.203820829240544e-05}
{"train_loss": 0.17826144397258759, "global_step": 96608, "epoch": 1085, "lr": 7.203768790604724e-05}
{"train_loss": 0.12568646669387817, "global_step": 96609, "epoch": 1085, "lr": 7.203716751672636e-05}
{"train_loss": 0.1347568929195404, "global_step": 96610, "epoch": 1085, "lr": 7.203664712444284e-05}
{"train_loss": 0.16346465051174164, "global_step": 96611, "epoch": 1085, "lr": 7.203612672919677e-05}
{"train_loss": 0.14801764488220215, "global_step": 96612, "epoch": 1085, "lr": 7.203560633098821e-05}
{"train_loss": 0.11527294665575027, "global_step": 96613, "epoch": 1085, "lr": 7.203508592981724e-05}
{"train_loss": 0.19305731356143951, "global_step": 96614, "epoch": 1085, "lr": 7.203456552568391e-05}
{"train_loss": 0.1465640813112259, "global_step": 96615, "epoch": 1085, "lr": 7.20340451185883e-05}
{"train_loss": 0.16487360000610352, "global_step": 96616, "epoch": 1085, "lr": 7.203352470853048e-05}
{"train_loss": 0.12624570727348328, "global_step": 96617, "epoch": 1085, "lr": 7.203300429551053e-05}
{"train_loss": 0.1497206687927246, "global_step": 96618, "epoch": 1085, "lr": 7.203248387952852e-05}
{"train_loss": 0.1283302903175354, "global_step": 96619, "epoch": 1085, "lr": 7.203196346058448e-05}
{"train_loss": 0.18702246248722076, "global_step": 96620, "epoch": 1085, "lr": 7.203144303867854e-05}
{"train_loss": 0.23370152711868286, "global_step": 96621, "epoch": 1085, "lr": 7.203092261381072e-05}
{"train_loss": 0.2051977515220642, "global_step": 96622, "epoch": 1085, "lr": 7.203040218598113e-05}
{"train_loss": 0.22359445691108704, "global_step": 96623, "epoch": 1085, "lr": 7.202988175518982e-05}
{"train_loss": 0.21976856887340546, "global_step": 96624, "epoch": 1085, "lr": 7.202936132143684e-05}
{"train_loss": 0.13057930767536163, "global_step": 96625, "epoch": 1085, "lr": 7.20288408847223e-05}
{"train_loss": 0.24456514418125153, "global_step": 96626, "epoch": 1085, "lr": 7.202832044504624e-05}
{"train_loss": 0.22960737347602844, "global_step": 96627, "epoch": 1085, "lr": 7.202780000240874e-05}
{"train_loss": 0.14565542340278625, "global_step": 96628, "epoch": 1085, "lr": 7.202727955680988e-05}
{"train_loss": 0.22874876856803894, "global_step": 96629, "epoch": 1085, "lr": 7.202675910824971e-05}
{"train_loss": 0.19648902118206024, "global_step": 96630, "epoch": 1085, "lr": 7.202623865672832e-05}
{"train_loss": 0.13155797123908997, "global_step": 96631, "epoch": 1085, "lr": 7.202571820224577e-05}
{"train_loss": 0.0972064882516861, "global_step": 96632, "epoch": 1085, "lr": 7.202519774480212e-05}
{"train_loss": 0.15696819126605988, "global_step": 96633, "epoch": 1085, "lr": 7.202467728439746e-05}
{"train_loss": 0.15782950818538666, "global_step": 96634, "epoch": 1085, "lr": 7.202415682103186e-05}
{"train_loss": 0.1986871063709259, "global_step": 96635, "epoch": 1085, "lr": 7.202363635470536e-05}
{"train_loss": 0.16535347700119019, "global_step": 96636, "epoch": 1085, "lr": 7.202311588541805e-05}
{"train_loss": 0.10012371838092804, "global_step": 96637, "epoch": 1085, "lr": 7.202259541317003e-05}
{"train_loss": 0.21679387986660004, "global_step": 96638, "epoch": 1085, "lr": 7.202207493796131e-05}
{"train_loss": 0.12226442992687225, "global_step": 96639, "epoch": 1085, "lr": 7.202155445979201e-05}
{"train_loss": 0.13618746399879456, "global_step": 96640, "epoch": 1085, "lr": 7.202103397866218e-05}
{"train_loss": 0.2286522388458252, "global_step": 96641, "epoch": 1085, "lr": 7.202051349457187e-05}
{"train_loss": 0.11731132864952087, "global_step": 96642, "epoch": 1085, "lr": 7.201999300752118e-05}
{"train_loss": 0.18580695986747742, "global_step": 96643, "epoch": 1085, "lr": 7.201947251751018e-05}
{"train_loss": 0.18674734234809875, "global_step": 96644, "epoch": 1085, "lr": 7.201895202453892e-05}
{"train_loss": 0.24027356505393982, "global_step": 96645, "epoch": 1085, "lr": 7.201843152860749e-05}
{"train_loss": 0.19467635452747345, "global_step": 96646, "epoch": 1085, "lr": 7.201791102971594e-05}
{"train_loss": 0.2279985547065735, "global_step": 96647, "epoch": 1085, "lr": 7.201739052786436e-05}
{"train_loss": 0.11952899396419525, "global_step": 96648, "epoch": 1085, "lr": 7.20168700230528e-05}
{"train_loss": 0.2930326759815216, "global_step": 96649, "epoch": 1085, "lr": 7.201634951528135e-05}
{"train_loss": 0.14532339572906494, "global_step": 96650, "epoch": 1085, "lr": 7.201582900455006e-05}
{"train_loss": 0.23547662794589996, "global_step": 96651, "epoch": 1085, "lr": 7.201530849085902e-05}
{"train_loss": 0.17158888280391693, "global_step": 96652, "epoch": 1085, "lr": 7.201478797420828e-05}
{"train_loss": 0.16948437255420043, "global_step": 96653, "epoch": 1085, "lr": 7.201426745459793e-05, "val_loss": 4.498856067657471, "train_action_mse_error": 34.530235290527344}
{"train_loss": 0.16421379148960114, "global_step": 96654, "epoch": 1086, "lr": 7.201374693202803e-05}
{"train_loss": 0.08769001811742783, "global_step": 96655, "epoch": 1086, "lr": 7.201322640649864e-05}
{"train_loss": 0.12851214408874512, "global_step": 96656, "epoch": 1086, "lr": 7.201270587800984e-05}
{"train_loss": 0.14359010756015778, "global_step": 96657, "epoch": 1086, "lr": 7.201218534656171e-05}
{"train_loss": 0.13991287350654602, "global_step": 96658, "epoch": 1086, "lr": 7.201166481215431e-05}
{"train_loss": 0.13446193933486938, "global_step": 96659, "epoch": 1086, "lr": 7.20111442747877e-05}
{"train_loss": 0.1456301510334015, "global_step": 96660, "epoch": 1086, "lr": 7.201062373446197e-05}
{"train_loss": 0.22161105275154114, "global_step": 96661, "epoch": 1086, "lr": 7.201010319117717e-05}
{"train_loss": 0.25092706084251404, "global_step": 96662, "epoch": 1086, "lr": 7.200958264493338e-05}
{"train_loss": 0.15641754865646362, "global_step": 96663, "epoch": 1086, "lr": 7.200906209573068e-05}
{"train_loss": 0.21856598556041718, "global_step": 96664, "epoch": 1086, "lr": 7.200854154356914e-05}
{"train_loss": 0.23253780603408813, "global_step": 96665, "epoch": 1086, "lr": 7.200802098844879e-05}
{"train_loss": 0.12019889801740646, "global_step": 96666, "epoch": 1086, "lr": 7.200750043036974e-05}
{"train_loss": 0.15874327719211578, "global_step": 96667, "epoch": 1086, "lr": 7.200697986933206e-05}
{"train_loss": 0.1065034344792366, "global_step": 96668, "epoch": 1086, "lr": 7.20064593053358e-05}
{"train_loss": 0.10076632350683212, "global_step": 96669, "epoch": 1086, "lr": 7.200593873838105e-05}
{"train_loss": 0.17856375873088837, "global_step": 96670, "epoch": 1086, "lr": 7.200541816846787e-05}
{"train_loss": 0.10989916324615479, "global_step": 96671, "epoch": 1086, "lr": 7.200489759559632e-05}
{"train_loss": 0.15179334580898285, "global_step": 96672, "epoch": 1086, "lr": 7.200437701976648e-05}
{"train_loss": 0.12174371629953384, "global_step": 96673, "epoch": 1086, "lr": 7.200385644097842e-05}
{"train_loss": 0.17223010957241058, "global_step": 96674, "epoch": 1086, "lr": 7.200333585923221e-05}
{"train_loss": 0.11688386648893356, "global_step": 96675, "epoch": 1086, "lr": 7.200281527452792e-05}
{"train_loss": 0.14364506304264069, "global_step": 96676, "epoch": 1086, "lr": 7.200229468686563e-05}
{"train_loss": 0.15955312550067902, "global_step": 96677, "epoch": 1086, "lr": 7.200177409624539e-05}
{"train_loss": 0.15456321835517883, "global_step": 96678, "epoch": 1086, "lr": 7.200125350266728e-05}
{"train_loss": 0.20364752411842346, "global_step": 96679, "epoch": 1086, "lr": 7.200073290613138e-05}
{"train_loss": 0.13828937709331512, "global_step": 96680, "epoch": 1086, "lr": 7.200021230663773e-05}
{"train_loss": 0.19642522931098938, "global_step": 96681, "epoch": 1086, "lr": 7.199969170418643e-05}
{"train_loss": 0.15600457787513733, "global_step": 96682, "epoch": 1086, "lr": 7.199917109877753e-05}
{"train_loss": 0.18557672202587128, "global_step": 96683, "epoch": 1086, "lr": 7.199865049041113e-05}
{"train_loss": 0.16659343242645264, "global_step": 96684, "epoch": 1086, "lr": 7.199812987908727e-05}
{"train_loss": 0.24126914143562317, "global_step": 96685, "epoch": 1086, "lr": 7.199760926480601e-05}
{"train_loss": 0.14865514636039734, "global_step": 96686, "epoch": 1086, "lr": 7.199708864756747e-05}
{"train_loss": 0.18884515762329102, "global_step": 96687, "epoch": 1086, "lr": 7.199656802737169e-05}
{"train_loss": 0.17380699515342712, "global_step": 96688, "epoch": 1086, "lr": 7.199604740421871e-05}
{"train_loss": 0.1807561218738556, "global_step": 96689, "epoch": 1086, "lr": 7.199552677810866e-05}
{"train_loss": 0.1378035992383957, "global_step": 96690, "epoch": 1086, "lr": 7.199500614904155e-05}
{"train_loss": 0.1470164656639099, "global_step": 96691, "epoch": 1086, "lr": 7.19944855170175e-05}
{"train_loss": 0.15558230876922607, "global_step": 96692, "epoch": 1086, "lr": 7.199396488203656e-05}
{"train_loss": 0.25954708456993103, "global_step": 96693, "epoch": 1086, "lr": 7.19934442440988e-05}
{"train_loss": 0.15254245698451996, "global_step": 96694, "epoch": 1086, "lr": 7.199292360320428e-05}
{"train_loss": 0.13415119051933289, "global_step": 96695, "epoch": 1086, "lr": 7.199240295935308e-05}
{"train_loss": 0.22940950095653534, "global_step": 96696, "epoch": 1086, "lr": 7.199188231254527e-05}
{"train_loss": 0.1442284733057022, "global_step": 96697, "epoch": 1086, "lr": 7.199136166278094e-05}
{"train_loss": 0.1490822732448578, "global_step": 96698, "epoch": 1086, "lr": 7.19908410100601e-05}
{"train_loss": 0.17312759160995483, "global_step": 96699, "epoch": 1086, "lr": 7.19903203543829e-05}
{"train_loss": 0.13868024945259094, "global_step": 96700, "epoch": 1086, "lr": 7.198979969574934e-05}
{"train_loss": 0.1290145069360733, "global_step": 96701, "epoch": 1086, "lr": 7.198927903415953e-05}
{"train_loss": 0.1560596078634262, "global_step": 96702, "epoch": 1086, "lr": 7.198875836961353e-05}
{"train_loss": 0.1464182734489441, "global_step": 96703, "epoch": 1086, "lr": 7.19882377021114e-05}
{"train_loss": 0.24596795439720154, "global_step": 96704, "epoch": 1086, "lr": 7.198771703165322e-05}
{"train_loss": 0.17951855063438416, "global_step": 96705, "epoch": 1086, "lr": 7.198719635823908e-05}
{"train_loss": 0.19229276478290558, "global_step": 96706, "epoch": 1086, "lr": 7.198667568186902e-05}
{"train_loss": 0.247291699051857, "global_step": 96707, "epoch": 1086, "lr": 7.198615500254312e-05}
{"train_loss": 0.1397896111011505, "global_step": 96708, "epoch": 1086, "lr": 7.198563432026145e-05}
{"train_loss": 0.12646397948265076, "global_step": 96709, "epoch": 1086, "lr": 7.198511363502408e-05}
{"train_loss": 0.19752782583236694, "global_step": 96710, "epoch": 1086, "lr": 7.198459294683106e-05}
{"train_loss": 0.13441038131713867, "global_step": 96711, "epoch": 1086, "lr": 7.19840722556825e-05}
{"train_loss": 0.21981288492679596, "global_step": 96712, "epoch": 1086, "lr": 7.198355156157844e-05}
{"train_loss": 0.11213263124227524, "global_step": 96713, "epoch": 1086, "lr": 7.198303086451897e-05}
{"train_loss": 0.3025732636451721, "global_step": 96714, "epoch": 1086, "lr": 7.198251016450415e-05}
{"train_loss": 0.16479969024658203, "global_step": 96715, "epoch": 1086, "lr": 7.198198946153403e-05}
{"train_loss": 0.1802397072315216, "global_step": 96716, "epoch": 1086, "lr": 7.198146875560871e-05}
{"train_loss": 0.19393137097358704, "global_step": 96717, "epoch": 1086, "lr": 7.198094804672825e-05}
{"train_loss": 0.0913391187787056, "global_step": 96718, "epoch": 1086, "lr": 7.198042733489272e-05}
{"train_loss": 0.20033518970012665, "global_step": 96719, "epoch": 1086, "lr": 7.197990662010219e-05}
{"train_loss": 0.14546845853328705, "global_step": 96720, "epoch": 1086, "lr": 7.197938590235674e-05}
{"train_loss": 0.14107652008533478, "global_step": 96721, "epoch": 1086, "lr": 7.197886518165641e-05}
{"train_loss": 0.14878731966018677, "global_step": 96722, "epoch": 1086, "lr": 7.197834445800129e-05}
{"train_loss": 0.1949937492609024, "global_step": 96723, "epoch": 1086, "lr": 7.197782373139145e-05}
{"train_loss": 0.24302631616592407, "global_step": 96724, "epoch": 1086, "lr": 7.197730300182698e-05}
{"train_loss": 0.17604492604732513, "global_step": 96725, "epoch": 1086, "lr": 7.197678226930793e-05}
{"train_loss": 0.1284727305173874, "global_step": 96726, "epoch": 1086, "lr": 7.197626153383435e-05}
{"train_loss": 0.16990821063518524, "global_step": 96727, "epoch": 1086, "lr": 7.197574079540633e-05}
{"train_loss": 0.16874580085277557, "global_step": 96728, "epoch": 1086, "lr": 7.197522005402395e-05}
{"train_loss": 0.1927344799041748, "global_step": 96729, "epoch": 1086, "lr": 7.197469930968726e-05}
{"train_loss": 0.17601065337657928, "global_step": 96730, "epoch": 1086, "lr": 7.197417856239634e-05}
{"train_loss": 0.18960003554821014, "global_step": 96731, "epoch": 1086, "lr": 7.197365781215128e-05}
{"train_loss": 0.06575244665145874, "global_step": 96732, "epoch": 1086, "lr": 7.197313705895212e-05}
{"train_loss": 0.21811795234680176, "global_step": 96733, "epoch": 1086, "lr": 7.197261630279892e-05}
{"train_loss": 0.21058647334575653, "global_step": 96734, "epoch": 1086, "lr": 7.197209554369178e-05}
{"train_loss": 0.16165606677532196, "global_step": 96735, "epoch": 1086, "lr": 7.197157478163077e-05}
{"train_loss": 0.20753082633018494, "global_step": 96736, "epoch": 1086, "lr": 7.197105401661593e-05}
{"train_loss": 0.17316590249538422, "global_step": 96737, "epoch": 1086, "lr": 7.197053324864736e-05}
{"train_loss": 0.20161239802837372, "global_step": 96738, "epoch": 1086, "lr": 7.197001247772513e-05}
{"train_loss": 0.2178936004638672, "global_step": 96739, "epoch": 1086, "lr": 7.19694917038493e-05}
{"train_loss": 0.18109360337257385, "global_step": 96740, "epoch": 1086, "lr": 7.196897092701991e-05}
{"train_loss": 0.16796660423278809, "global_step": 96741, "epoch": 1086, "lr": 7.196845014723707e-05}
{"train_loss": 0.1682393845547451, "global_step": 96742, "epoch": 1086, "lr": 7.196792936450085e-05, "val_loss": 4.4239277839660645}
{"train_loss": 0.11808450520038605, "global_step": 96743, "epoch": 1087, "lr": 7.196740857881131e-05}
{"train_loss": 0.2122291624546051, "global_step": 96744, "epoch": 1087, "lr": 7.196688779016854e-05}
{"train_loss": 0.15930470824241638, "global_step": 96745, "epoch": 1087, "lr": 7.196636699857256e-05}
{"train_loss": 0.1453583985567093, "global_step": 96746, "epoch": 1087, "lr": 7.196584620402348e-05}
{"train_loss": 0.14378105103969574, "global_step": 96747, "epoch": 1087, "lr": 7.196532540652136e-05}
{"train_loss": 0.13326439261436462, "global_step": 96748, "epoch": 1087, "lr": 7.196480460606627e-05}
{"train_loss": 0.1817210465669632, "global_step": 96749, "epoch": 1087, "lr": 7.196428380265827e-05}
{"train_loss": 0.20080436766147614, "global_step": 96750, "epoch": 1087, "lr": 7.196376299629745e-05}
{"train_loss": 0.2005908489227295, "global_step": 96751, "epoch": 1087, "lr": 7.196324218698388e-05}
{"train_loss": 0.14430087804794312, "global_step": 96752, "epoch": 1087, "lr": 7.19627213747176e-05}
{"train_loss": 0.1365301012992859, "global_step": 96753, "epoch": 1087, "lr": 7.19622005594987e-05}
{"train_loss": 0.11908417195081711, "global_step": 96754, "epoch": 1087, "lr": 7.196167974132726e-05}
{"train_loss": 0.22633029520511627, "global_step": 96755, "epoch": 1087, "lr": 7.196115892020333e-05}
{"train_loss": 0.19844591617584229, "global_step": 96756, "epoch": 1087, "lr": 7.1960638096127e-05}
{"train_loss": 0.13423453271389008, "global_step": 96757, "epoch": 1087, "lr": 7.196011726909833e-05}
{"train_loss": 0.18957045674324036, "global_step": 96758, "epoch": 1087, "lr": 7.195959643911735e-05}
{"train_loss": 0.10440519452095032, "global_step": 96759, "epoch": 1087, "lr": 7.195907560618423e-05}
{"train_loss": 0.15801405906677246, "global_step": 96760, "epoch": 1087, "lr": 7.195855477029895e-05}
{"train_loss": 0.11852574348449707, "global_step": 96761, "epoch": 1087, "lr": 7.195803393146161e-05}
{"train_loss": 0.1851772964000702, "global_step": 96762, "epoch": 1087, "lr": 7.195751308967229e-05}
{"train_loss": 0.1361619085073471, "global_step": 96763, "epoch": 1087, "lr": 7.195699224493105e-05}
{"train_loss": 0.12154320627450943, "global_step": 96764, "epoch": 1087, "lr": 7.195647139723796e-05}
{"train_loss": 0.1673237383365631, "global_step": 96765, "epoch": 1087, "lr": 7.195595054659308e-05}
{"train_loss": 0.12218847870826721, "global_step": 96766, "epoch": 1087, "lr": 7.195542969299649e-05}
{"train_loss": 0.17256642878055573, "global_step": 96767, "epoch": 1087, "lr": 7.195490883644826e-05}
{"train_loss": 0.15158990025520325, "global_step": 96768, "epoch": 1087, "lr": 7.195438797694848e-05}
{"train_loss": 0.1301596611738205, "global_step": 96769, "epoch": 1087, "lr": 7.195386711449718e-05}
{"train_loss": 0.07889487594366074, "global_step": 96770, "epoch": 1087, "lr": 7.195334624909445e-05}
{"train_loss": 0.23983041942119598, "global_step": 96771, "epoch": 1087, "lr": 7.195282538074038e-05}
{"train_loss": 0.21872395277023315, "global_step": 96772, "epoch": 1087, "lr": 7.195230450943499e-05}
{"train_loss": 0.11908133327960968, "global_step": 96773, "epoch": 1087, "lr": 7.195178363517841e-05}
{"train_loss": 0.1596740186214447, "global_step": 96774, "epoch": 1087, "lr": 7.195126275797066e-05}
{"train_loss": 0.14233066141605377, "global_step": 96775, "epoch": 1087, "lr": 7.195074187781183e-05}
{"train_loss": 0.1706313043832779, "global_step": 96776, "epoch": 1087, "lr": 7.195022099470199e-05}
{"train_loss": 0.22012992203235626, "global_step": 96777, "epoch": 1087, "lr": 7.194970010864123e-05}
{"train_loss": 0.14344452321529388, "global_step": 96778, "epoch": 1087, "lr": 7.194917921962959e-05}
{"train_loss": 0.1643935889005661, "global_step": 96779, "epoch": 1087, "lr": 7.194865832766714e-05}
{"train_loss": 0.1883200705051422, "global_step": 96780, "epoch": 1087, "lr": 7.194813743275398e-05}
{"train_loss": 0.16765199601650238, "global_step": 96781, "epoch": 1087, "lr": 7.194761653489014e-05}
{"train_loss": 0.14386436343193054, "global_step": 96782, "epoch": 1087, "lr": 7.194709563407571e-05}
{"train_loss": 0.15564654767513275, "global_step": 96783, "epoch": 1087, "lr": 7.194657473031079e-05}
{"train_loss": 0.27624526619911194, "global_step": 96784, "epoch": 1087, "lr": 7.19460538235954e-05}
{"train_loss": 0.14142054319381714, "global_step": 96785, "epoch": 1087, "lr": 7.194553291392962e-05}
{"train_loss": 0.19193360209465027, "global_step": 96786, "epoch": 1087, "lr": 7.194501200131355e-05}
{"train_loss": 0.1772129088640213, "global_step": 96787, "epoch": 1087, "lr": 7.194449108574724e-05}
{"train_loss": 0.10336082428693771, "global_step": 96788, "epoch": 1087, "lr": 7.194397016723074e-05}
{"train_loss": 0.12075404077768326, "global_step": 96789, "epoch": 1087, "lr": 7.194344924576415e-05}
{"train_loss": 0.3034818172454834, "global_step": 96790, "epoch": 1087, "lr": 7.194292832134755e-05}
{"train_loss": 0.15388695895671844, "global_step": 96791, "epoch": 1087, "lr": 7.194240739398097e-05}
{"train_loss": 0.22590163350105286, "global_step": 96792, "epoch": 1087, "lr": 7.19418864636645e-05}
{"train_loss": 0.13670942187309265, "global_step": 96793, "epoch": 1087, "lr": 7.194136553039822e-05}
{"train_loss": 0.16403326392173767, "global_step": 96794, "epoch": 1087, "lr": 7.194084459418218e-05}
{"train_loss": 0.22100229561328888, "global_step": 96795, "epoch": 1087, "lr": 7.194032365501646e-05}
{"train_loss": 0.18485292792320251, "global_step": 96796, "epoch": 1087, "lr": 7.193980271290114e-05}
{"train_loss": 0.18367555737495422, "global_step": 96797, "epoch": 1087, "lr": 7.193928176783628e-05}
{"train_loss": 0.17168430984020233, "global_step": 96798, "epoch": 1087, "lr": 7.193876081982195e-05}
{"train_loss": 0.2068881094455719, "global_step": 96799, "epoch": 1087, "lr": 7.193823986885823e-05}
{"train_loss": 0.1080346405506134, "global_step": 96800, "epoch": 1087, "lr": 7.193771891494517e-05}
{"train_loss": 0.18102534115314484, "global_step": 96801, "epoch": 1087, "lr": 7.193719795808284e-05}
{"train_loss": 0.14043699204921722, "global_step": 96802, "epoch": 1087, "lr": 7.193667699827135e-05}
{"train_loss": 0.15944071114063263, "global_step": 96803, "epoch": 1087, "lr": 7.193615603551074e-05}
{"train_loss": 0.1572234034538269, "global_step": 96804, "epoch": 1087, "lr": 7.193563506980105e-05}
{"train_loss": 0.10607496649026871, "global_step": 96805, "epoch": 1087, "lr": 7.193511410114242e-05}
{"train_loss": 0.18685467541217804, "global_step": 96806, "epoch": 1087, "lr": 7.193459312953486e-05}
{"train_loss": 0.1789129227399826, "global_step": 96807, "epoch": 1087, "lr": 7.193407215497845e-05}
{"train_loss": 0.23938055336475372, "global_step": 96808, "epoch": 1087, "lr": 7.193355117747328e-05}
{"train_loss": 0.26591596007347107, "global_step": 96809, "epoch": 1087, "lr": 7.193303019701941e-05}
{"train_loss": 0.1360791027545929, "global_step": 96810, "epoch": 1087, "lr": 7.193250921361691e-05}
{"train_loss": 0.13123610615730286, "global_step": 96811, "epoch": 1087, "lr": 7.193198822726586e-05}
{"train_loss": 0.15989890694618225, "global_step": 96812, "epoch": 1087, "lr": 7.193146723796632e-05}
{"train_loss": 0.15255562961101532, "global_step": 96813, "epoch": 1087, "lr": 7.193094624571836e-05}
{"train_loss": 0.15079785883426666, "global_step": 96814, "epoch": 1087, "lr": 7.193042525052205e-05}
{"train_loss": 0.1014476865530014, "global_step": 96815, "epoch": 1087, "lr": 7.192990425237747e-05}
{"train_loss": 0.10848278552293777, "global_step": 96816, "epoch": 1087, "lr": 7.192938325128467e-05}
{"train_loss": 0.25616344809532166, "global_step": 96817, "epoch": 1087, "lr": 7.192886224724373e-05}
{"train_loss": 0.20847737789154053, "global_step": 96818, "epoch": 1087, "lr": 7.192834124025474e-05}
{"train_loss": 0.20902042090892792, "global_step": 96819, "epoch": 1087, "lr": 7.192782023031774e-05}
{"train_loss": 0.13734044134616852, "global_step": 96820, "epoch": 1087, "lr": 7.19272992174328e-05}
{"train_loss": 0.16449446976184845, "global_step": 96821, "epoch": 1087, "lr": 7.192677820160002e-05}
{"train_loss": 0.11638965457677841, "global_step": 96822, "epoch": 1087, "lr": 7.192625718281945e-05}
{"train_loss": 0.15202248096466064, "global_step": 96823, "epoch": 1087, "lr": 7.192573616109117e-05}
{"train_loss": 0.20897503197193146, "global_step": 96824, "epoch": 1087, "lr": 7.192521513641522e-05}
{"train_loss": 0.10030162334442139, "global_step": 96825, "epoch": 1087, "lr": 7.19246941087917e-05}
{"train_loss": 0.15419821441173553, "global_step": 96826, "epoch": 1087, "lr": 7.192417307822069e-05}
{"train_loss": 0.18871068954467773, "global_step": 96827, "epoch": 1087, "lr": 7.192365204470222e-05}
{"train_loss": 0.14329062402248383, "global_step": 96828, "epoch": 1087, "lr": 7.192313100823637e-05}
{"train_loss": 0.12318601459264755, "global_step": 96829, "epoch": 1087, "lr": 7.192260996882325e-05}
{"train_loss": 0.3004222810268402, "global_step": 96830, "epoch": 1087, "lr": 7.19220889264629e-05}
{"train_loss": 0.16557640038179547, "global_step": 96831, "epoch": 1087, "lr": 7.192156788115537e-05, "val_loss": 4.574066162109375}
{"train_loss": 0.14806438982486725, "global_step": 96832, "epoch": 1088, "lr": 7.192104683290076e-05}
{"train_loss": 0.10489336401224136, "global_step": 96833, "epoch": 1088, "lr": 7.192052578169915e-05}
{"train_loss": 0.13030211627483368, "global_step": 96834, "epoch": 1088, "lr": 7.192000472755057e-05}
{"train_loss": 0.2030227780342102, "global_step": 96835, "epoch": 1088, "lr": 7.191948367045513e-05}
{"train_loss": 0.1383773535490036, "global_step": 96836, "epoch": 1088, "lr": 7.191896261041287e-05}
{"train_loss": 0.1429145634174347, "global_step": 96837, "epoch": 1088, "lr": 7.19184415474239e-05}
{"train_loss": 0.22352656722068787, "global_step": 96838, "epoch": 1088, "lr": 7.191792048148823e-05}
{"train_loss": 0.1443772315979004, "global_step": 96839, "epoch": 1088, "lr": 7.191739941260599e-05}
{"train_loss": 0.15126094222068787, "global_step": 96840, "epoch": 1088, "lr": 7.191687834077719e-05}
{"train_loss": 0.17484647035598755, "global_step": 96841, "epoch": 1088, "lr": 7.191635726600196e-05}
{"train_loss": 0.16530302166938782, "global_step": 96842, "epoch": 1088, "lr": 7.191583618828034e-05}
{"train_loss": 0.2298687845468521, "global_step": 96843, "epoch": 1088, "lr": 7.19153151076124e-05}
{"train_loss": 0.13734613358974457, "global_step": 96844, "epoch": 1088, "lr": 7.191479402399821e-05}
{"train_loss": 0.1259748637676239, "global_step": 96845, "epoch": 1088, "lr": 7.191427293743784e-05}
{"train_loss": 0.20741279423236847, "global_step": 96846, "epoch": 1088, "lr": 7.191375184793137e-05}
{"train_loss": 0.14723609387874603, "global_step": 96847, "epoch": 1088, "lr": 7.191323075547885e-05}
{"train_loss": 0.13111969828605652, "global_step": 96848, "epoch": 1088, "lr": 7.191270966008037e-05}
{"train_loss": 0.14321625232696533, "global_step": 96849, "epoch": 1088, "lr": 7.1912188561736e-05}
{"train_loss": 0.09852371364831924, "global_step": 96850, "epoch": 1088, "lr": 7.19116674604458e-05}
{"train_loss": 0.1062493771314621, "global_step": 96851, "epoch": 1088, "lr": 7.191114635620984e-05}
{"train_loss": 0.1538545936346054, "global_step": 96852, "epoch": 1088, "lr": 7.191062524902818e-05}
{"train_loss": 0.22086745500564575, "global_step": 96853, "epoch": 1088, "lr": 7.191010413890092e-05}
{"train_loss": 0.183614581823349, "global_step": 96854, "epoch": 1088, "lr": 7.190958302582812e-05}
{"train_loss": 0.20525142550468445, "global_step": 96855, "epoch": 1088, "lr": 7.190906190980984e-05}
{"train_loss": 0.20090346038341522, "global_step": 96856, "epoch": 1088, "lr": 7.190854079084615e-05}
{"train_loss": 0.13229461014270782, "global_step": 96857, "epoch": 1088, "lr": 7.190801966893712e-05}
{"train_loss": 0.1232033297419548, "global_step": 96858, "epoch": 1088, "lr": 7.190749854408284e-05}
{"train_loss": 0.12032212316989899, "global_step": 96859, "epoch": 1088, "lr": 7.190697741628334e-05}
{"train_loss": 0.15887700021266937, "global_step": 96860, "epoch": 1088, "lr": 7.190645628553873e-05}
{"train_loss": 0.16239021718502045, "global_step": 96861, "epoch": 1088, "lr": 7.190593515184907e-05}
{"train_loss": 0.2146352231502533, "global_step": 96862, "epoch": 1088, "lr": 7.190541401521442e-05}
{"train_loss": 0.1461860090494156, "global_step": 96863, "epoch": 1088, "lr": 7.190489287563483e-05}
{"train_loss": 0.20720645785331726, "global_step": 96864, "epoch": 1088, "lr": 7.190437173311042e-05}
{"train_loss": 0.19203762710094452, "global_step": 96865, "epoch": 1088, "lr": 7.190385058764122e-05}
{"train_loss": 0.17191632091999054, "global_step": 96866, "epoch": 1088, "lr": 7.190332943922733e-05}
{"train_loss": 0.17323417961597443, "global_step": 96867, "epoch": 1088, "lr": 7.190280828786877e-05}
{"train_loss": 0.24559178948402405, "global_step": 96868, "epoch": 1088, "lr": 7.190228713356568e-05}
{"train_loss": 0.13510003685951233, "global_step": 96869, "epoch": 1088, "lr": 7.190176597631809e-05}
{"train_loss": 0.1897374838590622, "global_step": 96870, "epoch": 1088, "lr": 7.190124481612606e-05}
{"train_loss": 0.19760197401046753, "global_step": 96871, "epoch": 1088, "lr": 7.190072365298968e-05}
{"train_loss": 0.1792883574962616, "global_step": 96872, "epoch": 1088, "lr": 7.190020248690902e-05}
{"train_loss": 0.1762748807668686, "global_step": 96873, "epoch": 1088, "lr": 7.189968131788413e-05}
{"train_loss": 0.15724724531173706, "global_step": 96874, "epoch": 1088, "lr": 7.189916014591511e-05}
{"train_loss": 0.12769629061222076, "global_step": 96875, "epoch": 1088, "lr": 7.189863897100201e-05}
{"train_loss": 0.1824386566877365, "global_step": 96876, "epoch": 1088, "lr": 7.18981177931449e-05}
{"train_loss": 0.1834671050310135, "global_step": 96877, "epoch": 1088, "lr": 7.189759661234387e-05}
{"train_loss": 0.2056368589401245, "global_step": 96878, "epoch": 1088, "lr": 7.189707542859897e-05}
{"train_loss": 0.17754624783992767, "global_step": 96879, "epoch": 1088, "lr": 7.189655424191026e-05}
{"train_loss": 0.17870208621025085, "global_step": 96880, "epoch": 1088, "lr": 7.189603305227785e-05}
{"train_loss": 0.22969579696655273, "global_step": 96881, "epoch": 1088, "lr": 7.189551185970177e-05}
{"train_loss": 0.18128618597984314, "global_step": 96882, "epoch": 1088, "lr": 7.189499066418212e-05}
{"train_loss": 0.16630959510803223, "global_step": 96883, "epoch": 1088, "lr": 7.189446946571894e-05}
{"train_loss": 0.2167692631483078, "global_step": 96884, "epoch": 1088, "lr": 7.18939482643123e-05}
{"train_loss": 0.1968359500169754, "global_step": 96885, "epoch": 1088, "lr": 7.18934270599623e-05}
{"train_loss": 0.1335751861333847, "global_step": 96886, "epoch": 1088, "lr": 7.189290585266901e-05}
{"train_loss": 0.24555037915706635, "global_step": 96887, "epoch": 1088, "lr": 7.189238464243247e-05}
{"train_loss": 0.21818438172340393, "global_step": 96888, "epoch": 1088, "lr": 7.189186342925277e-05}
{"train_loss": 0.22290493547916412, "global_step": 96889, "epoch": 1088, "lr": 7.189134221312998e-05}
{"train_loss": 0.207638218998909, "global_step": 96890, "epoch": 1088, "lr": 7.189082099406416e-05}
{"train_loss": 0.16909459233283997, "global_step": 96891, "epoch": 1088, "lr": 7.189029977205538e-05}
{"train_loss": 0.18642792105674744, "global_step": 96892, "epoch": 1088, "lr": 7.188977854710373e-05}
{"train_loss": 0.15805746614933014, "global_step": 96893, "epoch": 1088, "lr": 7.188925731920926e-05}
{"train_loss": 0.13543494045734406, "global_step": 96894, "epoch": 1088, "lr": 7.188873608837205e-05}
{"train_loss": 0.16385214030742645, "global_step": 96895, "epoch": 1088, "lr": 7.188821485459215e-05}
{"train_loss": 0.19390445947647095, "global_step": 96896, "epoch": 1088, "lr": 7.188769361786967e-05}
{"train_loss": 0.1841520518064499, "global_step": 96897, "epoch": 1088, "lr": 7.188717237820463e-05}
{"train_loss": 0.14216674864292145, "global_step": 96898, "epoch": 1088, "lr": 7.188665113559715e-05}
{"train_loss": 0.1776670664548874, "global_step": 96899, "epoch": 1088, "lr": 7.188612989004728e-05}
{"train_loss": 0.14820073544979095, "global_step": 96900, "epoch": 1088, "lr": 7.188560864155506e-05}
{"train_loss": 0.19283612072467804, "global_step": 96901, "epoch": 1088, "lr": 7.18850873901206e-05}
{"train_loss": 0.19404058158397675, "global_step": 96902, "epoch": 1088, "lr": 7.188456613574398e-05}
{"train_loss": 0.18081028759479523, "global_step": 96903, "epoch": 1088, "lr": 7.18840448784252e-05}
{"train_loss": 0.14492526650428772, "global_step": 96904, "epoch": 1088, "lr": 7.18835236181644e-05}
{"train_loss": 0.26289016008377075, "global_step": 96905, "epoch": 1088, "lr": 7.188300235496163e-05}
{"train_loss": 0.14425136148929596, "global_step": 96906, "epoch": 1088, "lr": 7.188248108881695e-05}
{"train_loss": 0.20779092609882355, "global_step": 96907, "epoch": 1088, "lr": 7.188195981973045e-05}
{"train_loss": 0.14858190715312958, "global_step": 96908, "epoch": 1088, "lr": 7.188143854770218e-05}
{"train_loss": 0.11387647688388824, "global_step": 96909, "epoch": 1088, "lr": 7.188091727273221e-05}
{"train_loss": 0.1677582561969757, "global_step": 96910, "epoch": 1088, "lr": 7.188039599482064e-05}
{"train_loss": 0.15254902839660645, "global_step": 96911, "epoch": 1088, "lr": 7.187987471396748e-05}
{"train_loss": 0.2695814371109009, "global_step": 96912, "epoch": 1088, "lr": 7.187935343017287e-05}
{"train_loss": 0.2182152420282364, "global_step": 96913, "epoch": 1088, "lr": 7.187883214343683e-05}
{"train_loss": 0.2583829462528229, "global_step": 96914, "epoch": 1088, "lr": 7.187831085375947e-05}
{"train_loss": 0.09920912981033325, "global_step": 96915, "epoch": 1088, "lr": 7.187778956114082e-05}
{"train_loss": 0.18117138743400574, "global_step": 96916, "epoch": 1088, "lr": 7.187726826558096e-05}
{"train_loss": 0.10706658661365509, "global_step": 96917, "epoch": 1088, "lr": 7.187674696707999e-05}
{"train_loss": 0.18062880635261536, "global_step": 96918, "epoch": 1088, "lr": 7.187622566563794e-05}
{"train_loss": 0.14849154651165009, "global_step": 96919, "epoch": 1088, "lr": 7.187570436125491e-05}
{"train_loss": 0.17224407371845138, "global_step": 96920, "epoch": 1088, "lr": 7.187518305393097e-05, "val_loss": 4.796233177185059}
{"train_loss": 0.2365151047706604, "global_step": 96921, "epoch": 1089, "lr": 7.187466174366614e-05}
{"train_loss": 0.2591540515422821, "global_step": 96922, "epoch": 1089, "lr": 7.187414043046055e-05}
{"train_loss": 0.17523373663425446, "global_step": 96923, "epoch": 1089, "lr": 7.187361911431425e-05}
{"train_loss": 0.19022394716739655, "global_step": 96924, "epoch": 1089, "lr": 7.18730977952273e-05}
{"train_loss": 0.19393505156040192, "global_step": 96925, "epoch": 1089, "lr": 7.187257647319978e-05}
{"train_loss": 0.17855216562747955, "global_step": 96926, "epoch": 1089, "lr": 7.187205514823177e-05}
{"train_loss": 0.1952260434627533, "global_step": 96927, "epoch": 1089, "lr": 7.187153382032332e-05}
{"train_loss": 0.19576340913772583, "global_step": 96928, "epoch": 1089, "lr": 7.18710124894745e-05}
{"train_loss": 0.12924064695835114, "global_step": 96929, "epoch": 1089, "lr": 7.187049115568542e-05}
{"train_loss": 0.17259752750396729, "global_step": 96930, "epoch": 1089, "lr": 7.186996981895609e-05}
{"train_loss": 0.2797802686691284, "global_step": 96931, "epoch": 1089, "lr": 7.186944847928661e-05}
{"train_loss": 0.34015458822250366, "global_step": 96932, "epoch": 1089, "lr": 7.186892713667707e-05}
{"train_loss": 0.16335061192512512, "global_step": 96933, "epoch": 1089, "lr": 7.18684057911275e-05}
{"train_loss": 0.2125566154718399, "global_step": 96934, "epoch": 1089, "lr": 7.1867884442638e-05}
{"train_loss": 0.14029213786125183, "global_step": 96935, "epoch": 1089, "lr": 7.186736309120862e-05}
{"train_loss": 0.2090332806110382, "global_step": 96936, "epoch": 1089, "lr": 7.186684173683946e-05}
{"train_loss": 0.20831429958343506, "global_step": 96937, "epoch": 1089, "lr": 7.186632037953054e-05}
{"train_loss": 0.19462502002716064, "global_step": 96938, "epoch": 1089, "lr": 7.186579901928199e-05}
{"train_loss": 0.19388367235660553, "global_step": 96939, "epoch": 1089, "lr": 7.186527765609383e-05}
{"train_loss": 0.16182973980903625, "global_step": 96940, "epoch": 1089, "lr": 7.186475628996613e-05}
{"train_loss": 0.12290330976247787, "global_step": 96941, "epoch": 1089, "lr": 7.186423492089902e-05}
{"train_loss": 0.17884434759616852, "global_step": 96942, "epoch": 1089, "lr": 7.186371354889253e-05}
{"train_loss": 0.2555953562259674, "global_step": 96943, "epoch": 1089, "lr": 7.18631921739467e-05}
{"train_loss": 0.1838611215353012, "global_step": 96944, "epoch": 1089, "lr": 7.186267079606164e-05}
{"train_loss": 0.18552614748477936, "global_step": 96945, "epoch": 1089, "lr": 7.186214941523741e-05}
{"train_loss": 0.17433737218379974, "global_step": 96946, "epoch": 1089, "lr": 7.18616280314741e-05}
{"train_loss": 0.11996780335903168, "global_step": 96947, "epoch": 1089, "lr": 7.186110664477173e-05}
{"train_loss": 0.19269388914108276, "global_step": 96948, "epoch": 1089, "lr": 7.186058525513042e-05}
{"train_loss": 0.20967668294906616, "global_step": 96949, "epoch": 1089, "lr": 7.186006386255022e-05}
{"train_loss": 0.13758669793605804, "global_step": 96950, "epoch": 1089, "lr": 7.18595424670312e-05}
{"train_loss": 0.11373220384120941, "global_step": 96951, "epoch": 1089, "lr": 7.185902106857343e-05}
{"train_loss": 0.16057687997817993, "global_step": 96952, "epoch": 1089, "lr": 7.185849966717698e-05}
{"train_loss": 0.17097662389278412, "global_step": 96953, "epoch": 1089, "lr": 7.185797826284192e-05}
{"train_loss": 0.2336590737104416, "global_step": 96954, "epoch": 1089, "lr": 7.185745685556832e-05}
{"train_loss": 0.21828356385231018, "global_step": 96955, "epoch": 1089, "lr": 7.185693544535625e-05}
{"train_loss": 0.16125787794589996, "global_step": 96956, "epoch": 1089, "lr": 7.18564140322058e-05}
{"train_loss": 0.15919235348701477, "global_step": 96957, "epoch": 1089, "lr": 7.185589261611702e-05}
{"train_loss": 0.10028761625289917, "global_step": 96958, "epoch": 1089, "lr": 7.185537119708995e-05}
{"train_loss": 0.1282082498073578, "global_step": 96959, "epoch": 1089, "lr": 7.18548497751247e-05}
{"train_loss": 0.1599472463130951, "global_step": 96960, "epoch": 1089, "lr": 7.185432835022135e-05}
{"train_loss": 0.12568143010139465, "global_step": 96961, "epoch": 1089, "lr": 7.185380692237995e-05}
{"train_loss": 0.24536563456058502, "global_step": 96962, "epoch": 1089, "lr": 7.185328549160057e-05}
{"train_loss": 0.1968328207731247, "global_step": 96963, "epoch": 1089, "lr": 7.185276405788328e-05}
{"train_loss": 0.15170882642269135, "global_step": 96964, "epoch": 1089, "lr": 7.185224262122814e-05}
{"train_loss": 0.1341850757598877, "global_step": 96965, "epoch": 1089, "lr": 7.185172118163524e-05}
{"train_loss": 0.15224596858024597, "global_step": 96966, "epoch": 1089, "lr": 7.185119973910465e-05}
{"train_loss": 0.25007978081703186, "global_step": 96967, "epoch": 1089, "lr": 7.185067829363643e-05}
{"train_loss": 0.20305858552455902, "global_step": 96968, "epoch": 1089, "lr": 7.185015684523065e-05}
{"train_loss": 0.11361192166805267, "global_step": 96969, "epoch": 1089, "lr": 7.184963539388738e-05}
{"train_loss": 0.13275162875652313, "global_step": 96970, "epoch": 1089, "lr": 7.184911393960669e-05}
{"train_loss": 0.16941498219966888, "global_step": 96971, "epoch": 1089, "lr": 7.184859248238867e-05}
{"train_loss": 0.2208772748708725, "global_step": 96972, "epoch": 1089, "lr": 7.184807102223335e-05}
{"train_loss": 0.158835306763649, "global_step": 96973, "epoch": 1089, "lr": 7.184754955914083e-05}
{"train_loss": 0.29011741280555725, "global_step": 96974, "epoch": 1089, "lr": 7.184702809311119e-05}
{"train_loss": 0.16101202368736267, "global_step": 96975, "epoch": 1089, "lr": 7.184650662414447e-05}
{"train_loss": 0.19228102266788483, "global_step": 96976, "epoch": 1089, "lr": 7.184598515224076e-05}
{"train_loss": 0.20878338813781738, "global_step": 96977, "epoch": 1089, "lr": 7.184546367740012e-05}
{"train_loss": 0.17380036413669586, "global_step": 96978, "epoch": 1089, "lr": 7.184494219962261e-05}
{"train_loss": 0.1470969021320343, "global_step": 96979, "epoch": 1089, "lr": 7.184442071890833e-05}
{"train_loss": 0.1780676394701004, "global_step": 96980, "epoch": 1089, "lr": 7.184389923525734e-05}
{"train_loss": 0.1349925398826599, "global_step": 96981, "epoch": 1089, "lr": 7.184337774866969e-05}
{"train_loss": 0.18371307849884033, "global_step": 96982, "epoch": 1089, "lr": 7.184285625914546e-05}
{"train_loss": 0.21325896680355072, "global_step": 96983, "epoch": 1089, "lr": 7.184233476668473e-05}
{"train_loss": 0.1895231455564499, "global_step": 96984, "epoch": 1089, "lr": 7.184181327128757e-05}
{"train_loss": 0.1429753303527832, "global_step": 96985, "epoch": 1089, "lr": 7.184129177295404e-05}
{"train_loss": 0.21145422756671906, "global_step": 96986, "epoch": 1089, "lr": 7.184077027168422e-05}
{"train_loss": 0.18501733243465424, "global_step": 96987, "epoch": 1089, "lr": 7.184024876747816e-05}
{"train_loss": 0.15454579889774323, "global_step": 96988, "epoch": 1089, "lr": 7.183972726033597e-05}
{"train_loss": 0.13447540998458862, "global_step": 96989, "epoch": 1089, "lr": 7.183920575025768e-05}
{"train_loss": 0.20445406436920166, "global_step": 96990, "epoch": 1089, "lr": 7.183868423724337e-05}
{"train_loss": 0.1488824188709259, "global_step": 96991, "epoch": 1089, "lr": 7.183816272129312e-05}
{"train_loss": 0.20445656776428223, "global_step": 96992, "epoch": 1089, "lr": 7.183764120240699e-05}
{"train_loss": 0.14050795137882233, "global_step": 96993, "epoch": 1089, "lr": 7.183711968058506e-05}
{"train_loss": 0.23269356787204742, "global_step": 96994, "epoch": 1089, "lr": 7.18365981558274e-05}
{"train_loss": 0.2490587830543518, "global_step": 96995, "epoch": 1089, "lr": 7.183607662813408e-05}
{"train_loss": 0.13932418823242188, "global_step": 96996, "epoch": 1089, "lr": 7.183555509750514e-05}
{"train_loss": 0.1962503045797348, "global_step": 96997, "epoch": 1089, "lr": 7.18350335639407e-05}
{"train_loss": 0.17162063717842102, "global_step": 96998, "epoch": 1089, "lr": 7.183451202744081e-05}
{"train_loss": 0.2864214777946472, "global_step": 96999, "epoch": 1089, "lr": 7.183399048800551e-05}
{"train_loss": 0.19331969320774078, "global_step": 97000, "epoch": 1089, "lr": 7.183346894563492e-05}
{"train_loss": 0.18821552395820618, "global_step": 97001, "epoch": 1089, "lr": 7.183294740032906e-05}
{"train_loss": 0.12752853333950043, "global_step": 97002, "epoch": 1089, "lr": 7.183242585208805e-05}
{"train_loss": 0.13571979105472565, "global_step": 97003, "epoch": 1089, "lr": 7.183190430091193e-05}
{"train_loss": 0.22323894500732422, "global_step": 97004, "epoch": 1089, "lr": 7.183138274680078e-05}
{"train_loss": 0.14660058915615082, "global_step": 97005, "epoch": 1089, "lr": 7.183086118975466e-05}
{"train_loss": 0.18682894110679626, "global_step": 97006, "epoch": 1089, "lr": 7.183033962977364e-05}
{"train_loss": 0.1331862062215805, "global_step": 97007, "epoch": 1089, "lr": 7.182981806685782e-05}
{"train_loss": 0.2844846248626709, "global_step": 97008, "epoch": 1089, "lr": 7.182929650100722e-05}
{"train_loss": 0.1839914573879724, "global_step": 97009, "epoch": 1089, "lr": 7.182877493222197e-05, "val_loss": 4.522927761077881}
{"train_loss": 0.15830087661743164, "global_step": 97010, "epoch": 1090, "lr": 7.182825336050207e-05}
{"train_loss": 0.17076709866523743, "global_step": 97011, "epoch": 1090, "lr": 7.182773178584765e-05}
{"train_loss": 0.17725035548210144, "global_step": 97012, "epoch": 1090, "lr": 7.182721020825876e-05}
{"train_loss": 0.19866544008255005, "global_step": 97013, "epoch": 1090, "lr": 7.182668862773545e-05}
{"train_loss": 0.14534713327884674, "global_step": 97014, "epoch": 1090, "lr": 7.182616704427782e-05}
{"train_loss": 0.13973653316497803, "global_step": 97015, "epoch": 1090, "lr": 7.182564545788592e-05}
{"train_loss": 0.17156532406806946, "global_step": 97016, "epoch": 1090, "lr": 7.182512386855984e-05}
{"train_loss": 0.21590648591518402, "global_step": 97017, "epoch": 1090, "lr": 7.182460227629963e-05}
{"train_loss": 0.10636259615421295, "global_step": 97018, "epoch": 1090, "lr": 7.182408068110537e-05}
{"train_loss": 0.2163543552160263, "global_step": 97019, "epoch": 1090, "lr": 7.182355908297713e-05}
{"train_loss": 0.13948190212249756, "global_step": 97020, "epoch": 1090, "lr": 7.182303748191499e-05}
{"train_loss": 0.21223123371601105, "global_step": 97021, "epoch": 1090, "lr": 7.1822515877919e-05}
{"train_loss": 0.17577975988388062, "global_step": 97022, "epoch": 1090, "lr": 7.182199427098922e-05}
{"train_loss": 0.1714051067829132, "global_step": 97023, "epoch": 1090, "lr": 7.182147266112575e-05}
{"train_loss": 0.14182515442371368, "global_step": 97024, "epoch": 1090, "lr": 7.182095104832867e-05}
{"train_loss": 0.14939960837364197, "global_step": 97025, "epoch": 1090, "lr": 7.1820429432598e-05}
{"train_loss": 0.17681699991226196, "global_step": 97026, "epoch": 1090, "lr": 7.181990781393385e-05}
{"train_loss": 0.1358608603477478, "global_step": 97027, "epoch": 1090, "lr": 7.181938619233627e-05}
{"train_loss": 0.17955225706100464, "global_step": 97028, "epoch": 1090, "lr": 7.181886456780535e-05}
{"train_loss": 0.15979844331741333, "global_step": 97029, "epoch": 1090, "lr": 7.181834294034116e-05}
{"train_loss": 0.15421530604362488, "global_step": 97030, "epoch": 1090, "lr": 7.181782130994374e-05}
{"train_loss": 0.17698200047016144, "global_step": 97031, "epoch": 1090, "lr": 7.181729967661319e-05}
{"train_loss": 0.21527764201164246, "global_step": 97032, "epoch": 1090, "lr": 7.181677804034957e-05}
{"train_loss": 0.21179087460041046, "global_step": 97033, "epoch": 1090, "lr": 7.181625640115295e-05}
{"train_loss": 0.11845830082893372, "global_step": 97034, "epoch": 1090, "lr": 7.18157347590234e-05}
{"train_loss": 0.1622050702571869, "global_step": 97035, "epoch": 1090, "lr": 7.181521311396098e-05}
{"train_loss": 0.10425125807523727, "global_step": 97036, "epoch": 1090, "lr": 7.181469146596578e-05}
{"train_loss": 0.15604065358638763, "global_step": 97037, "epoch": 1090, "lr": 7.181416981503787e-05}
{"train_loss": 0.1862531304359436, "global_step": 97038, "epoch": 1090, "lr": 7.18136481611773e-05}
{"train_loss": 0.1707005500793457, "global_step": 97039, "epoch": 1090, "lr": 7.181312650438416e-05}
{"train_loss": 0.24044087529182434, "global_step": 97040, "epoch": 1090, "lr": 7.18126048446585e-05}
{"train_loss": 0.1561359316110611, "global_step": 97041, "epoch": 1090, "lr": 7.18120831820004e-05}
{"train_loss": 0.17107096314430237, "global_step": 97042, "epoch": 1090, "lr": 7.181156151640993e-05}
{"train_loss": 0.1625445932149887, "global_step": 97043, "epoch": 1090, "lr": 7.181103984788718e-05}
{"train_loss": 0.16286006569862366, "global_step": 97044, "epoch": 1090, "lr": 7.181051817643219e-05}
{"train_loss": 0.12987901270389557, "global_step": 97045, "epoch": 1090, "lr": 7.180999650204503e-05}
{"train_loss": 0.13658708333969116, "global_step": 97046, "epoch": 1090, "lr": 7.180947482472579e-05}
{"train_loss": 0.19938135147094727, "global_step": 97047, "epoch": 1090, "lr": 7.180895314447454e-05}
{"train_loss": 0.20349416136741638, "global_step": 97048, "epoch": 1090, "lr": 7.180843146129132e-05}
{"train_loss": 0.07791061699390411, "global_step": 97049, "epoch": 1090, "lr": 7.180790977517625e-05}
{"train_loss": 0.3000848591327667, "global_step": 97050, "epoch": 1090, "lr": 7.180738808612937e-05}
{"train_loss": 0.234051913022995, "global_step": 97051, "epoch": 1090, "lr": 7.180686639415073e-05}
{"train_loss": 0.21842369437217712, "global_step": 97052, "epoch": 1090, "lr": 7.180634469924043e-05}
{"train_loss": 0.19130954146385193, "global_step": 97053, "epoch": 1090, "lr": 7.180582300139855e-05}
{"train_loss": 0.17285272479057312, "global_step": 97054, "epoch": 1090, "lr": 7.180530130062512e-05}
{"train_loss": 0.14690802991390228, "global_step": 97055, "epoch": 1090, "lr": 7.180477959692024e-05}
{"train_loss": 0.12653276324272156, "global_step": 97056, "epoch": 1090, "lr": 7.180425789028399e-05}
{"train_loss": 0.13322047889232635, "global_step": 97057, "epoch": 1090, "lr": 7.180373618071641e-05}
{"train_loss": 0.1420334279537201, "global_step": 97058, "epoch": 1090, "lr": 7.180321446821759e-05}
{"train_loss": 0.1967901736497879, "global_step": 97059, "epoch": 1090, "lr": 7.180269275278758e-05}
{"train_loss": 0.111823670566082, "global_step": 97060, "epoch": 1090, "lr": 7.180217103442646e-05}
{"train_loss": 0.1755581945180893, "global_step": 97061, "epoch": 1090, "lr": 7.180164931313433e-05}
{"train_loss": 0.20948879420757294, "global_step": 97062, "epoch": 1090, "lr": 7.180112758891122e-05}
{"train_loss": 0.21143533289432526, "global_step": 97063, "epoch": 1090, "lr": 7.18006058617572e-05}
{"train_loss": 0.20567969977855682, "global_step": 97064, "epoch": 1090, "lr": 7.180008413167237e-05}
{"train_loss": 0.20116093754768372, "global_step": 97065, "epoch": 1090, "lr": 7.17995623986568e-05}
{"train_loss": 0.16195069253444672, "global_step": 97066, "epoch": 1090, "lr": 7.17990406627105e-05}
{"train_loss": 0.2014790177345276, "global_step": 97067, "epoch": 1090, "lr": 7.179851892383361e-05}
{"train_loss": 0.12739427387714386, "global_step": 97068, "epoch": 1090, "lr": 7.179799718202619e-05}
{"train_loss": 0.10639679431915283, "global_step": 97069, "epoch": 1090, "lr": 7.179747543728827e-05}
{"train_loss": 0.16173671185970306, "global_step": 97070, "epoch": 1090, "lr": 7.179695368961997e-05}
{"train_loss": 0.12065654247999191, "global_step": 97071, "epoch": 1090, "lr": 7.179643193902132e-05}
{"train_loss": 0.14503447711467743, "global_step": 97072, "epoch": 1090, "lr": 7.17959101854924e-05}
{"train_loss": 0.152084618806839, "global_step": 97073, "epoch": 1090, "lr": 7.17953884290333e-05}
{"train_loss": 0.23421192169189453, "global_step": 97074, "epoch": 1090, "lr": 7.179486666964406e-05}
{"train_loss": 0.1196494996547699, "global_step": 97075, "epoch": 1090, "lr": 7.179434490732477e-05}
{"train_loss": 0.19856815040111542, "global_step": 97076, "epoch": 1090, "lr": 7.17938231420755e-05}
{"train_loss": 0.17928220331668854, "global_step": 97077, "epoch": 1090, "lr": 7.179330137389632e-05}
{"train_loss": 0.16612450778484344, "global_step": 97078, "epoch": 1090, "lr": 7.17927796027873e-05}
{"train_loss": 0.1725984662771225, "global_step": 97079, "epoch": 1090, "lr": 7.179225782874848e-05}
{"train_loss": 0.2173202484846115, "global_step": 97080, "epoch": 1090, "lr": 7.179173605178e-05}
{"train_loss": 0.1360192745923996, "global_step": 97081, "epoch": 1090, "lr": 7.179121427188184e-05}
{"train_loss": 0.1796300858259201, "global_step": 97082, "epoch": 1090, "lr": 7.179069248905415e-05}
{"train_loss": 0.14925029873847961, "global_step": 97083, "epoch": 1090, "lr": 7.179017070329696e-05}
{"train_loss": 0.11944397538900375, "global_step": 97084, "epoch": 1090, "lr": 7.178964891461033e-05}
{"train_loss": 0.18169541656970978, "global_step": 97085, "epoch": 1090, "lr": 7.178912712299435e-05}
{"train_loss": 0.15243978798389435, "global_step": 97086, "epoch": 1090, "lr": 7.17886053284491e-05}
{"train_loss": 0.16446267068386078, "global_step": 97087, "epoch": 1090, "lr": 7.178808353097464e-05}
{"train_loss": 0.16681738197803497, "global_step": 97088, "epoch": 1090, "lr": 7.178756173057104e-05}
{"train_loss": 0.19952704012393951, "global_step": 97089, "epoch": 1090, "lr": 7.178703992723836e-05}
{"train_loss": 0.18176597356796265, "global_step": 97090, "epoch": 1090, "lr": 7.178651812097667e-05}
{"train_loss": 0.13710656762123108, "global_step": 97091, "epoch": 1090, "lr": 7.178599631178606e-05}
{"train_loss": 0.19586293399333954, "global_step": 97092, "epoch": 1090, "lr": 7.178547449966659e-05}
{"train_loss": 0.15321333706378937, "global_step": 97093, "epoch": 1090, "lr": 7.178495268461832e-05}
{"train_loss": 0.17469331622123718, "global_step": 97094, "epoch": 1090, "lr": 7.178443086664133e-05}
{"train_loss": 0.12288663536310196, "global_step": 97095, "epoch": 1090, "lr": 7.178390904573569e-05}
{"train_loss": 0.23004424571990967, "global_step": 97096, "epoch": 1090, "lr": 7.178338722190148e-05}
{"train_loss": 0.12114240229129791, "global_step": 97097, "epoch": 1090, "lr": 7.178286539513874e-05}
{"train_loss": 0.16850009016441495, "global_step": 97098, "epoch": 1090, "lr": 7.178234356544757e-05, "val_loss": 4.528677940368652, "train_action_mse_error": 20.10306739807129}
{"train_loss": 0.14323849976062775, "global_step": 97099, "epoch": 1091, "lr": 7.178182173282802e-05}
{"train_loss": 0.15125198662281036, "global_step": 97100, "epoch": 1091, "lr": 7.178129989728018e-05}
{"train_loss": 0.1043122187256813, "global_step": 97101, "epoch": 1091, "lr": 7.17807780588041e-05}
{"train_loss": 0.1731434017419815, "global_step": 97102, "epoch": 1091, "lr": 7.178025621739988e-05}
{"train_loss": 0.1278005838394165, "global_step": 97103, "epoch": 1091, "lr": 7.177973437306754e-05}
{"train_loss": 0.14527195692062378, "global_step": 97104, "epoch": 1091, "lr": 7.177921252580719e-05}
{"train_loss": 0.15013954043388367, "global_step": 97105, "epoch": 1091, "lr": 7.177869067561891e-05}
{"train_loss": 0.10450759530067444, "global_step": 97106, "epoch": 1091, "lr": 7.177816882250273e-05}
{"train_loss": 0.15467488765716553, "global_step": 97107, "epoch": 1091, "lr": 7.177764696645877e-05}
{"train_loss": 0.2553740441799164, "global_step": 97108, "epoch": 1091, "lr": 7.177712510748705e-05}
{"train_loss": 0.16876348853111267, "global_step": 97109, "epoch": 1091, "lr": 7.177660324558765e-05}
{"train_loss": 0.1537541300058365, "global_step": 97110, "epoch": 1091, "lr": 7.177608138076066e-05}
{"train_loss": 0.1829296499490738, "global_step": 97111, "epoch": 1091, "lr": 7.177555951300614e-05}
{"train_loss": 0.15805275738239288, "global_step": 97112, "epoch": 1091, "lr": 7.177503764232416e-05}
{"train_loss": 0.20977507531642914, "global_step": 97113, "epoch": 1091, "lr": 7.17745157687148e-05}
{"train_loss": 0.160464808344841, "global_step": 97114, "epoch": 1091, "lr": 7.177399389217812e-05}
{"train_loss": 0.11192020773887634, "global_step": 97115, "epoch": 1091, "lr": 7.177347201271418e-05}
{"train_loss": 0.20757482945919037, "global_step": 97116, "epoch": 1091, "lr": 7.177295013032308e-05}
{"train_loss": 0.2073056697845459, "global_step": 97117, "epoch": 1091, "lr": 7.177242824500486e-05}
{"train_loss": 0.13849744200706482, "global_step": 97118, "epoch": 1091, "lr": 7.177190635675961e-05}
{"train_loss": 0.09522232413291931, "global_step": 97119, "epoch": 1091, "lr": 7.177138446558739e-05}
{"train_loss": 0.11068608611822128, "global_step": 97120, "epoch": 1091, "lr": 7.177086257148826e-05}
{"train_loss": 0.13226386904716492, "global_step": 97121, "epoch": 1091, "lr": 7.177034067446231e-05}
{"train_loss": 0.16636809706687927, "global_step": 97122, "epoch": 1091, "lr": 7.176981877450961e-05}
{"train_loss": 0.14984148740768433, "global_step": 97123, "epoch": 1091, "lr": 7.176929687163022e-05}
{"train_loss": 0.18728859722614288, "global_step": 97124, "epoch": 1091, "lr": 7.176877496582421e-05}
{"train_loss": 0.12085834890604019, "global_step": 97125, "epoch": 1091, "lr": 7.176825305709166e-05}
{"train_loss": 0.2087586671113968, "global_step": 97126, "epoch": 1091, "lr": 7.176773114543264e-05}
{"train_loss": 0.19165684282779694, "global_step": 97127, "epoch": 1091, "lr": 7.17672092308472e-05}
{"train_loss": 0.20691876113414764, "global_step": 97128, "epoch": 1091, "lr": 7.176668731333544e-05}
{"train_loss": 0.1343689262866974, "global_step": 97129, "epoch": 1091, "lr": 7.17661653928974e-05}
{"train_loss": 0.17402999103069305, "global_step": 97130, "epoch": 1091, "lr": 7.176564346953317e-05}
{"train_loss": 0.13198474049568176, "global_step": 97131, "epoch": 1091, "lr": 7.17651215432428e-05}
{"train_loss": 0.2110271453857422, "global_step": 97132, "epoch": 1091, "lr": 7.176459961402639e-05}
{"train_loss": 0.14269216358661652, "global_step": 97133, "epoch": 1091, "lr": 7.1764077681884e-05}
{"train_loss": 0.09763786941766739, "global_step": 97134, "epoch": 1091, "lr": 7.176355574681568e-05}
{"train_loss": 0.21612098813056946, "global_step": 97135, "epoch": 1091, "lr": 7.176303380882152e-05}
{"train_loss": 0.21768824756145477, "global_step": 97136, "epoch": 1091, "lr": 7.176251186790159e-05}
{"train_loss": 0.18677204847335815, "global_step": 97137, "epoch": 1091, "lr": 7.176198992405594e-05}
{"train_loss": 0.18593968451023102, "global_step": 97138, "epoch": 1091, "lr": 7.176146797728467e-05}
{"train_loss": 0.20406988263130188, "global_step": 97139, "epoch": 1091, "lr": 7.176094602758783e-05}
{"train_loss": 0.144755557179451, "global_step": 97140, "epoch": 1091, "lr": 7.17604240749655e-05}
{"train_loss": 0.18326060473918915, "global_step": 97141, "epoch": 1091, "lr": 7.175990211941774e-05}
{"train_loss": 0.11688560247421265, "global_step": 97142, "epoch": 1091, "lr": 7.175938016094464e-05}
{"train_loss": 0.17058371007442474, "global_step": 97143, "epoch": 1091, "lr": 7.175885819954624e-05}
{"train_loss": 0.19726672768592834, "global_step": 97144, "epoch": 1091, "lr": 7.175833623522263e-05}
{"train_loss": 0.16980263590812683, "global_step": 97145, "epoch": 1091, "lr": 7.17578142679739e-05}
{"train_loss": 0.08093363791704178, "global_step": 97146, "epoch": 1091, "lr": 7.175729229780008e-05}
{"train_loss": 0.13715417683124542, "global_step": 97147, "epoch": 1091, "lr": 7.175677032470127e-05}
{"train_loss": 0.1434991955757141, "global_step": 97148, "epoch": 1091, "lr": 7.17562483486775e-05}
{"train_loss": 0.24157002568244934, "global_step": 97149, "epoch": 1091, "lr": 7.175572636972888e-05}
{"train_loss": 0.10726667195558548, "global_step": 97150, "epoch": 1091, "lr": 7.175520438785548e-05}
{"train_loss": 0.23879694938659668, "global_step": 97151, "epoch": 1091, "lr": 7.175468240305734e-05}
{"train_loss": 0.15535672008991241, "global_step": 97152, "epoch": 1091, "lr": 7.175416041533456e-05}
{"train_loss": 0.16463226079940796, "global_step": 97153, "epoch": 1091, "lr": 7.175363842468718e-05}
{"train_loss": 0.186757892370224, "global_step": 97154, "epoch": 1091, "lr": 7.175311643111531e-05}
{"train_loss": 0.23081301152706146, "global_step": 97155, "epoch": 1091, "lr": 7.175259443461899e-05}
{"train_loss": 0.2020644247531891, "global_step": 97156, "epoch": 1091, "lr": 7.17520724351983e-05}
{"train_loss": 0.12587985396385193, "global_step": 97157, "epoch": 1091, "lr": 7.17515504328533e-05}
{"train_loss": 0.1637287437915802, "global_step": 97158, "epoch": 1091, "lr": 7.175102842758408e-05}
{"train_loss": 0.13771943747997284, "global_step": 97159, "epoch": 1091, "lr": 7.17505064193907e-05}
{"train_loss": 0.16340354084968567, "global_step": 97160, "epoch": 1091, "lr": 7.174998440827323e-05}
{"train_loss": 0.147846519947052, "global_step": 97161, "epoch": 1091, "lr": 7.174946239423173e-05}
{"train_loss": 0.08276743441820145, "global_step": 97162, "epoch": 1091, "lr": 7.174894037726628e-05}
{"train_loss": 0.1570805311203003, "global_step": 97163, "epoch": 1091, "lr": 7.174841835737696e-05}
{"train_loss": 0.12625883519649506, "global_step": 97164, "epoch": 1091, "lr": 7.174789633456383e-05}
{"train_loss": 0.1832805871963501, "global_step": 97165, "epoch": 1091, "lr": 7.174737430882696e-05}
{"train_loss": 0.1754300445318222, "global_step": 97166, "epoch": 1091, "lr": 7.174685228016642e-05}
{"train_loss": 0.2270350605249405, "global_step": 97167, "epoch": 1091, "lr": 7.174633024858228e-05}
{"train_loss": 0.1409313678741455, "global_step": 97168, "epoch": 1091, "lr": 7.174580821407462e-05}
{"train_loss": 0.1752818375825882, "global_step": 97169, "epoch": 1091, "lr": 7.174528617664348e-05}
{"train_loss": 0.18839684128761292, "global_step": 97170, "epoch": 1091, "lr": 7.174476413628896e-05}
{"train_loss": 0.16067631542682648, "global_step": 97171, "epoch": 1091, "lr": 7.174424209301112e-05}
{"train_loss": 0.12685003876686096, "global_step": 97172, "epoch": 1091, "lr": 7.174372004681005e-05}
{"train_loss": 0.14280694723129272, "global_step": 97173, "epoch": 1091, "lr": 7.174319799768577e-05}
{"train_loss": 0.1552240401506424, "global_step": 97174, "epoch": 1091, "lr": 7.17426759456384e-05}
{"train_loss": 0.15557222068309784, "global_step": 97175, "epoch": 1091, "lr": 7.174215389066798e-05}
{"train_loss": 0.12624993920326233, "global_step": 97176, "epoch": 1091, "lr": 7.174163183277462e-05}
{"train_loss": 0.21965450048446655, "global_step": 97177, "epoch": 1091, "lr": 7.174110977195833e-05}
{"train_loss": 0.1344229280948639, "global_step": 97178, "epoch": 1091, "lr": 7.174058770821924e-05}
{"train_loss": 0.15387046337127686, "global_step": 97179, "epoch": 1091, "lr": 7.174006564155738e-05}
{"train_loss": 0.24817858636379242, "global_step": 97180, "epoch": 1091, "lr": 7.173954357197283e-05}
{"train_loss": 0.2222396582365036, "global_step": 97181, "epoch": 1091, "lr": 7.173902149946567e-05}
{"train_loss": 0.17453143000602722, "global_step": 97182, "epoch": 1091, "lr": 7.173849942403596e-05}
{"train_loss": 0.21406254172325134, "global_step": 97183, "epoch": 1091, "lr": 7.173797734568378e-05}
{"train_loss": 0.1902199685573578, "global_step": 97184, "epoch": 1091, "lr": 7.17374552644092e-05}
{"train_loss": 0.11691145598888397, "global_step": 97185, "epoch": 1091, "lr": 7.173693318021227e-05}
{"train_loss": 0.27073654532432556, "global_step": 97186, "epoch": 1091, "lr": 7.173641109309308e-05}
{"train_loss": 0.16465946094373637, "global_step": 97187, "epoch": 1091, "lr": 7.173588900305169e-05, "val_loss": 4.630542278289795}
{"train_loss": 0.20074842870235443, "global_step": 97188, "epoch": 1092, "lr": 7.173536691008817e-05}
{"train_loss": 0.19028066098690033, "global_step": 97189, "epoch": 1092, "lr": 7.17348448142026e-05}
{"train_loss": 0.19199272990226746, "global_step": 97190, "epoch": 1092, "lr": 7.173432271539506e-05}
{"train_loss": 0.17197959125041962, "global_step": 97191, "epoch": 1092, "lr": 7.173380061366559e-05}
{"train_loss": 0.11608874052762985, "global_step": 97192, "epoch": 1092, "lr": 7.173327850901428e-05}
{"train_loss": 0.18196311593055725, "global_step": 97193, "epoch": 1092, "lr": 7.173275640144118e-05}
{"train_loss": 0.10412998497486115, "global_step": 97194, "epoch": 1092, "lr": 7.173223429094638e-05}
{"train_loss": 0.19056588411331177, "global_step": 97195, "epoch": 1092, "lr": 7.173171217752997e-05}
{"train_loss": 0.15107963979244232, "global_step": 97196, "epoch": 1092, "lr": 7.173119006119197e-05}
{"train_loss": 0.16531918942928314, "global_step": 97197, "epoch": 1092, "lr": 7.173066794193248e-05}
{"train_loss": 0.16220302879810333, "global_step": 97198, "epoch": 1092, "lr": 7.173014581975158e-05}
{"train_loss": 0.19383032619953156, "global_step": 97199, "epoch": 1092, "lr": 7.172962369464933e-05}
{"train_loss": 0.14479923248291016, "global_step": 97200, "epoch": 1092, "lr": 7.172910156662577e-05}
{"train_loss": 0.21118395030498505, "global_step": 97201, "epoch": 1092, "lr": 7.172857943568104e-05}
{"train_loss": 0.18283683061599731, "global_step": 97202, "epoch": 1092, "lr": 7.172805730181514e-05}
{"train_loss": 0.09132294356822968, "global_step": 97203, "epoch": 1092, "lr": 7.172753516502816e-05}
{"train_loss": 0.1288408488035202, "global_step": 97204, "epoch": 1092, "lr": 7.172701302532019e-05}
{"train_loss": 0.13445578515529633, "global_step": 97205, "epoch": 1092, "lr": 7.172649088269129e-05}
{"train_loss": 0.18712888658046722, "global_step": 97206, "epoch": 1092, "lr": 7.172596873714153e-05}
{"train_loss": 0.08164346218109131, "global_step": 97207, "epoch": 1092, "lr": 7.172544658867096e-05}
{"train_loss": 0.17320087552070618, "global_step": 97208, "epoch": 1092, "lr": 7.172492443727969e-05}
{"train_loss": 0.13186468183994293, "global_step": 97209, "epoch": 1092, "lr": 7.172440228296775e-05}
{"train_loss": 0.1781798154115677, "global_step": 97210, "epoch": 1092, "lr": 7.172388012573524e-05}
{"train_loss": 0.23207643628120422, "global_step": 97211, "epoch": 1092, "lr": 7.172335796558222e-05}
{"train_loss": 0.1806173026561737, "global_step": 97212, "epoch": 1092, "lr": 7.172283580250873e-05}
{"train_loss": 0.14541448652744293, "global_step": 97213, "epoch": 1092, "lr": 7.17223136365149e-05}
{"train_loss": 0.2115461677312851, "global_step": 97214, "epoch": 1092, "lr": 7.172179146760077e-05}
{"train_loss": 0.214266836643219, "global_step": 97215, "epoch": 1092, "lr": 7.172126929576642e-05}
{"train_loss": 0.17725712060928345, "global_step": 97216, "epoch": 1092, "lr": 7.172074712101189e-05}
{"train_loss": 0.19166038930416107, "global_step": 97217, "epoch": 1092, "lr": 7.172022494333727e-05}
{"train_loss": 0.13121731579303741, "global_step": 97218, "epoch": 1092, "lr": 7.171970276274263e-05}
{"train_loss": 0.1801399290561676, "global_step": 97219, "epoch": 1092, "lr": 7.171918057922805e-05}
{"train_loss": 0.10691940784454346, "global_step": 97220, "epoch": 1092, "lr": 7.171865839279359e-05}
{"train_loss": 0.21407929062843323, "global_step": 97221, "epoch": 1092, "lr": 7.171813620343931e-05}
{"train_loss": 0.15123385190963745, "global_step": 97222, "epoch": 1092, "lr": 7.171761401116531e-05}
{"train_loss": 0.2232426255941391, "global_step": 97223, "epoch": 1092, "lr": 7.171709181597163e-05}
{"train_loss": 0.1747330278158188, "global_step": 97224, "epoch": 1092, "lr": 7.171656961785835e-05}
{"train_loss": 0.15512266755104065, "global_step": 97225, "epoch": 1092, "lr": 7.171604741682555e-05}
{"train_loss": 0.1225959062576294, "global_step": 97226, "epoch": 1092, "lr": 7.171552521287329e-05}
{"train_loss": 0.20383603870868683, "global_step": 97227, "epoch": 1092, "lr": 7.171500300600164e-05}
{"train_loss": 0.131844162940979, "global_step": 97228, "epoch": 1092, "lr": 7.171448079621068e-05}
{"train_loss": 0.17892543971538544, "global_step": 97229, "epoch": 1092, "lr": 7.171395858350046e-05}
{"train_loss": 0.16193653643131256, "global_step": 97230, "epoch": 1092, "lr": 7.171343636787106e-05}
{"train_loss": 0.12846575677394867, "global_step": 97231, "epoch": 1092, "lr": 7.171291414932256e-05}
{"train_loss": 0.1408497542142868, "global_step": 97232, "epoch": 1092, "lr": 7.171239192785502e-05}
{"train_loss": 0.16598264873027802, "global_step": 97233, "epoch": 1092, "lr": 7.171186970346851e-05}
{"train_loss": 0.20453442633152008, "global_step": 97234, "epoch": 1092, "lr": 7.171134747616311e-05}
{"train_loss": 0.18542024493217468, "global_step": 97235, "epoch": 1092, "lr": 7.171082524593889e-05}
{"train_loss": 0.13562649488449097, "global_step": 97236, "epoch": 1092, "lr": 7.171030301279592e-05}
{"train_loss": 0.1279798448085785, "global_step": 97237, "epoch": 1092, "lr": 7.170978077673424e-05}
{"train_loss": 0.2060060203075409, "global_step": 97238, "epoch": 1092, "lr": 7.170925853775396e-05}
{"train_loss": 0.2051396369934082, "global_step": 97239, "epoch": 1092, "lr": 7.170873629585514e-05}
{"train_loss": 0.18567129969596863, "global_step": 97240, "epoch": 1092, "lr": 7.170821405103784e-05}
{"train_loss": 0.15296654403209686, "global_step": 97241, "epoch": 1092, "lr": 7.170769180330213e-05}
{"train_loss": 0.15828542411327362, "global_step": 97242, "epoch": 1092, "lr": 7.17071695526481e-05}
{"train_loss": 0.1867864578962326, "global_step": 97243, "epoch": 1092, "lr": 7.170664729907578e-05}
{"train_loss": 0.19338805973529816, "global_step": 97244, "epoch": 1092, "lr": 7.170612504258529e-05}
{"train_loss": 0.13316692411899567, "global_step": 97245, "epoch": 1092, "lr": 7.170560278317668e-05}
{"train_loss": 0.19794991612434387, "global_step": 97246, "epoch": 1092, "lr": 7.170508052084999e-05}
{"train_loss": 0.14012041687965393, "global_step": 97247, "epoch": 1092, "lr": 7.170455825560533e-05}
{"train_loss": 0.21644067764282227, "global_step": 97248, "epoch": 1092, "lr": 7.170403598744275e-05}
{"train_loss": 0.23522910475730896, "global_step": 97249, "epoch": 1092, "lr": 7.170351371636232e-05}
{"train_loss": 0.1545971781015396, "global_step": 97250, "epoch": 1092, "lr": 7.170299144236415e-05}
{"train_loss": 0.16223680973052979, "global_step": 97251, "epoch": 1092, "lr": 7.170246916544824e-05}
{"train_loss": 0.18837745487689972, "global_step": 97252, "epoch": 1092, "lr": 7.170194688561473e-05}
{"train_loss": 0.11975138634443283, "global_step": 97253, "epoch": 1092, "lr": 7.170142460286364e-05}
{"train_loss": 0.23088714480400085, "global_step": 97254, "epoch": 1092, "lr": 7.170090231719505e-05}
{"train_loss": 0.20243330299854279, "global_step": 97255, "epoch": 1092, "lr": 7.170038002860905e-05}
{"train_loss": 0.20229171216487885, "global_step": 97256, "epoch": 1092, "lr": 7.16998577371057e-05}
{"train_loss": 0.1851986050605774, "global_step": 97257, "epoch": 1092, "lr": 7.169933544268507e-05}
{"train_loss": 0.15580283105373383, "global_step": 97258, "epoch": 1092, "lr": 7.169881314534722e-05}
{"train_loss": 0.15478333830833435, "global_step": 97259, "epoch": 1092, "lr": 7.169829084509224e-05}
{"train_loss": 0.14337800443172455, "global_step": 97260, "epoch": 1092, "lr": 7.169776854192017e-05}
{"train_loss": 0.13129112124443054, "global_step": 97261, "epoch": 1092, "lr": 7.169724623583111e-05}
{"train_loss": 0.1490878462791443, "global_step": 97262, "epoch": 1092, "lr": 7.169672392682514e-05}
{"train_loss": 0.188606858253479, "global_step": 97263, "epoch": 1092, "lr": 7.169620161490229e-05}
{"train_loss": 0.2018469274044037, "global_step": 97264, "epoch": 1092, "lr": 7.169567930006265e-05}
{"train_loss": 0.18627750873565674, "global_step": 97265, "epoch": 1092, "lr": 7.169515698230629e-05}
{"train_loss": 0.16997087001800537, "global_step": 97266, "epoch": 1092, "lr": 7.169463466163327e-05}
{"train_loss": 0.15170371532440186, "global_step": 97267, "epoch": 1092, "lr": 7.169411233804369e-05}
{"train_loss": 0.11153970658779144, "global_step": 97268, "epoch": 1092, "lr": 7.16935900115376e-05}
{"train_loss": 0.1425214409828186, "global_step": 97269, "epoch": 1092, "lr": 7.169306768211506e-05}
{"train_loss": 0.20245523750782013, "global_step": 97270, "epoch": 1092, "lr": 7.169254534977617e-05}
{"train_loss": 0.1467803418636322, "global_step": 97271, "epoch": 1092, "lr": 7.169202301452096e-05}
{"train_loss": 0.19256658852100372, "global_step": 97272, "epoch": 1092, "lr": 7.169150067634953e-05}
{"train_loss": 0.23752912878990173, "global_step": 97273, "epoch": 1092, "lr": 7.169097833526194e-05}
{"train_loss": 0.1313285529613495, "global_step": 97274, "epoch": 1092, "lr": 7.169045599125826e-05}
{"train_loss": 0.17763952910900116, "global_step": 97275, "epoch": 1092, "lr": 7.168993364433857e-05}
{"train_loss": 0.16827365873234995, "global_step": 97276, "epoch": 1092, "lr": 7.168941129450293e-05, "val_loss": 4.485989093780518}
{"train_loss": 0.18958696722984314, "global_step": 97277, "epoch": 1093, "lr": 7.168888894175142e-05}
{"train_loss": 0.12186422944068909, "global_step": 97278, "epoch": 1093, "lr": 7.168836658608409e-05}
{"train_loss": 0.22027552127838135, "global_step": 97279, "epoch": 1093, "lr": 7.168784422750104e-05}
{"train_loss": 0.16462412476539612, "global_step": 97280, "epoch": 1093, "lr": 7.168732186600231e-05}
{"train_loss": 0.15355440974235535, "global_step": 97281, "epoch": 1093, "lr": 7.1686799501588e-05}
{"train_loss": 0.18698585033416748, "global_step": 97282, "epoch": 1093, "lr": 7.168627713425815e-05}
{"train_loss": 0.22954213619232178, "global_step": 97283, "epoch": 1093, "lr": 7.168575476401285e-05}
{"train_loss": 0.23830628395080566, "global_step": 97284, "epoch": 1093, "lr": 7.168523239085214e-05}
{"train_loss": 0.18939848244190216, "global_step": 97285, "epoch": 1093, "lr": 7.168471001477614e-05}
{"train_loss": 0.1333208829164505, "global_step": 97286, "epoch": 1093, "lr": 7.16841876357849e-05}
{"train_loss": 0.17295360565185547, "global_step": 97287, "epoch": 1093, "lr": 7.168366525387846e-05}
{"train_loss": 0.21958443522453308, "global_step": 97288, "epoch": 1093, "lr": 7.168314286905694e-05}
{"train_loss": 0.11443813145160675, "global_step": 97289, "epoch": 1093, "lr": 7.168262048132037e-05}
{"train_loss": 0.16352558135986328, "global_step": 97290, "epoch": 1093, "lr": 7.168209809066884e-05}
{"train_loss": 0.23364269733428955, "global_step": 97291, "epoch": 1093, "lr": 7.168157569710244e-05}
{"train_loss": 0.1565731316804886, "global_step": 97292, "epoch": 1093, "lr": 7.168105330062118e-05}
{"train_loss": 0.15461134910583496, "global_step": 97293, "epoch": 1093, "lr": 7.168053090122518e-05}
{"train_loss": 0.2230326384305954, "global_step": 97294, "epoch": 1093, "lr": 7.168000849891451e-05}
{"train_loss": 0.15581375360488892, "global_step": 97295, "epoch": 1093, "lr": 7.167948609368922e-05}
{"train_loss": 0.1785394698381424, "global_step": 97296, "epoch": 1093, "lr": 7.167896368554938e-05}
{"train_loss": 0.17793777585029602, "global_step": 97297, "epoch": 1093, "lr": 7.167844127449507e-05}
{"train_loss": 0.16610321402549744, "global_step": 97298, "epoch": 1093, "lr": 7.167791886052636e-05}
{"train_loss": 0.16349515318870544, "global_step": 97299, "epoch": 1093, "lr": 7.167739644364332e-05}
{"train_loss": 0.21789845824241638, "global_step": 97300, "epoch": 1093, "lr": 7.167687402384602e-05}
{"train_loss": 0.18640954792499542, "global_step": 97301, "epoch": 1093, "lr": 7.167635160113453e-05}
{"train_loss": 0.1944122165441513, "global_step": 97302, "epoch": 1093, "lr": 7.16758291755089e-05}
{"train_loss": 0.21830381453037262, "global_step": 97303, "epoch": 1093, "lr": 7.167530674696922e-05}
{"train_loss": 0.13965904712677002, "global_step": 97304, "epoch": 1093, "lr": 7.167478431551557e-05}
{"train_loss": 0.1537032276391983, "global_step": 97305, "epoch": 1093, "lr": 7.167426188114803e-05}
{"train_loss": 0.1204979345202446, "global_step": 97306, "epoch": 1093, "lr": 7.167373944386661e-05}
{"train_loss": 0.21826975047588348, "global_step": 97307, "epoch": 1093, "lr": 7.167321700367145e-05}
{"train_loss": 0.11339785158634186, "global_step": 97308, "epoch": 1093, "lr": 7.167269456056258e-05}
{"train_loss": 0.19496333599090576, "global_step": 97309, "epoch": 1093, "lr": 7.167217211454008e-05}
{"train_loss": 0.13323557376861572, "global_step": 97310, "epoch": 1093, "lr": 7.167164966560403e-05}
{"train_loss": 0.1174573078751564, "global_step": 97311, "epoch": 1093, "lr": 7.167112721375448e-05}
{"train_loss": 0.18574264645576477, "global_step": 97312, "epoch": 1093, "lr": 7.16706047589915e-05}
{"train_loss": 0.16026490926742554, "global_step": 97313, "epoch": 1093, "lr": 7.167008230131519e-05}
{"train_loss": 0.11231827735900879, "global_step": 97314, "epoch": 1093, "lr": 7.16695598407256e-05}
{"train_loss": 0.17409870028495789, "global_step": 97315, "epoch": 1093, "lr": 7.16690373772228e-05}
{"train_loss": 0.1728263646364212, "global_step": 97316, "epoch": 1093, "lr": 7.166851491080686e-05}
{"train_loss": 0.20833280682563782, "global_step": 97317, "epoch": 1093, "lr": 7.166799244147784e-05}
{"train_loss": 0.18709984421730042, "global_step": 97318, "epoch": 1093, "lr": 7.166746996923584e-05}
{"train_loss": 0.17562809586524963, "global_step": 97319, "epoch": 1093, "lr": 7.16669474940809e-05}
{"train_loss": 0.14802607893943787, "global_step": 97320, "epoch": 1093, "lr": 7.166642501601312e-05}
{"train_loss": 0.13608619570732117, "global_step": 97321, "epoch": 1093, "lr": 7.166590253503254e-05}
{"train_loss": 0.15855933725833893, "global_step": 97322, "epoch": 1093, "lr": 7.166538005113924e-05}
{"train_loss": 0.10222911834716797, "global_step": 97323, "epoch": 1093, "lr": 7.16648575643333e-05}
{"train_loss": 0.22087706625461578, "global_step": 97324, "epoch": 1093, "lr": 7.166433507461478e-05}
{"train_loss": 0.17028586566448212, "global_step": 97325, "epoch": 1093, "lr": 7.166381258198376e-05}
{"train_loss": 0.12184031307697296, "global_step": 97326, "epoch": 1093, "lr": 7.166329008644031e-05}
{"train_loss": 0.19233979284763336, "global_step": 97327, "epoch": 1093, "lr": 7.166276758798449e-05}
{"train_loss": 0.14433170855045319, "global_step": 97328, "epoch": 1093, "lr": 7.166224508661636e-05}
{"train_loss": 0.17518414556980133, "global_step": 97329, "epoch": 1093, "lr": 7.166172258233604e-05}
{"train_loss": 0.13837283849716187, "global_step": 97330, "epoch": 1093, "lr": 7.166120007514354e-05}
{"train_loss": 0.20204564929008484, "global_step": 97331, "epoch": 1093, "lr": 7.166067756503895e-05}
{"train_loss": 0.16109713912010193, "global_step": 97332, "epoch": 1093, "lr": 7.166015505202234e-05}
{"train_loss": 0.10747978836297989, "global_step": 97333, "epoch": 1093, "lr": 7.165963253609382e-05}
{"train_loss": 0.12257204204797745, "global_step": 97334, "epoch": 1093, "lr": 7.16591100172534e-05}
{"train_loss": 0.194695845246315, "global_step": 97335, "epoch": 1093, "lr": 7.165858749550118e-05}
{"train_loss": 0.15265437960624695, "global_step": 97336, "epoch": 1093, "lr": 7.165806497083723e-05}
{"train_loss": 0.13536913692951202, "global_step": 97337, "epoch": 1093, "lr": 7.165754244326162e-05}
{"train_loss": 0.1484779715538025, "global_step": 97338, "epoch": 1093, "lr": 7.165701991277443e-05}
{"train_loss": 0.1729123592376709, "global_step": 97339, "epoch": 1093, "lr": 7.16564973793757e-05}
{"train_loss": 0.10872078686952591, "global_step": 97340, "epoch": 1093, "lr": 7.165597484306551e-05}
{"train_loss": 0.125828817486763, "global_step": 97341, "epoch": 1093, "lr": 7.165545230384395e-05}
{"train_loss": 0.14494626224040985, "global_step": 97342, "epoch": 1093, "lr": 7.165492976171109e-05}
{"train_loss": 0.14895494282245636, "global_step": 97343, "epoch": 1093, "lr": 7.165440721666696e-05}
{"train_loss": 0.10955691337585449, "global_step": 97344, "epoch": 1093, "lr": 7.165388466871167e-05}
{"train_loss": 0.10806494951248169, "global_step": 97345, "epoch": 1093, "lr": 7.16533621178453e-05}
{"train_loss": 0.09491557627916336, "global_step": 97346, "epoch": 1093, "lr": 7.165283956406787e-05}
{"train_loss": 0.20385167002677917, "global_step": 97347, "epoch": 1093, "lr": 7.16523170073795e-05}
{"train_loss": 0.12037397176027298, "global_step": 97348, "epoch": 1093, "lr": 7.165179444778023e-05}
{"train_loss": 0.09965507686138153, "global_step": 97349, "epoch": 1093, "lr": 7.165127188527014e-05}
{"train_loss": 0.18238213658332825, "global_step": 97350, "epoch": 1093, "lr": 7.165074931984931e-05}
{"train_loss": 0.18606029450893402, "global_step": 97351, "epoch": 1093, "lr": 7.16502267515178e-05}
{"train_loss": 0.16953438520431519, "global_step": 97352, "epoch": 1093, "lr": 7.164970418027567e-05}
{"train_loss": 0.09654127061367035, "global_step": 97353, "epoch": 1093, "lr": 7.164918160612301e-05}
{"train_loss": 0.12446079403162003, "global_step": 97354, "epoch": 1093, "lr": 7.164865902905987e-05}
{"train_loss": 0.0867360383272171, "global_step": 97355, "epoch": 1093, "lr": 7.164813644908634e-05}
{"train_loss": 0.19731327891349792, "global_step": 97356, "epoch": 1093, "lr": 7.164761386620249e-05}
{"train_loss": 0.1760091930627823, "global_step": 97357, "epoch": 1093, "lr": 7.164709128040838e-05}
{"train_loss": 0.14575520157814026, "global_step": 97358, "epoch": 1093, "lr": 7.164656869170406e-05}
{"train_loss": 0.1359393447637558, "global_step": 97359, "epoch": 1093, "lr": 7.164604610008965e-05}
{"train_loss": 0.07448908686637878, "global_step": 97360, "epoch": 1093, "lr": 7.164552350556517e-05}
{"train_loss": 0.17272703349590302, "global_step": 97361, "epoch": 1093, "lr": 7.164500090813073e-05}
{"train_loss": 0.12781499326229095, "global_step": 97362, "epoch": 1093, "lr": 7.164447830778638e-05}
{"train_loss": 0.19458405673503876, "global_step": 97363, "epoch": 1093, "lr": 7.16439557045322e-05}
{"train_loss": 0.1563037782907486, "global_step": 97364, "epoch": 1093, "lr": 7.164343309836822e-05}
{"train_loss": 0.1602576132235902, "global_step": 97365, "epoch": 1093, "lr": 7.164291048929458e-05, "val_loss": 4.57395601272583}
{"train_loss": 0.13911022245883942, "global_step": 97366, "epoch": 1094, "lr": 7.164238787731131e-05}
{"train_loss": 0.21312694251537323, "global_step": 97367, "epoch": 1094, "lr": 7.164186526241848e-05}
{"train_loss": 0.17769789695739746, "global_step": 97368, "epoch": 1094, "lr": 7.164134264461616e-05}
{"train_loss": 0.12259526550769806, "global_step": 97369, "epoch": 1094, "lr": 7.164082002390444e-05}
{"train_loss": 0.19675421714782715, "global_step": 97370, "epoch": 1094, "lr": 7.164029740028335e-05}
{"train_loss": 0.10523910820484161, "global_step": 97371, "epoch": 1094, "lr": 7.1639774773753e-05}
{"train_loss": 0.10247521102428436, "global_step": 97372, "epoch": 1094, "lr": 7.163925214431345e-05}
{"train_loss": 0.10489204525947571, "global_step": 97373, "epoch": 1094, "lr": 7.163872951196476e-05}
{"train_loss": 0.16388411819934845, "global_step": 97374, "epoch": 1094, "lr": 7.163820687670702e-05}
{"train_loss": 0.11500861495733261, "global_step": 97375, "epoch": 1094, "lr": 7.163768423854028e-05}
{"train_loss": 0.16624616086483002, "global_step": 97376, "epoch": 1094, "lr": 7.163716159746461e-05}
{"train_loss": 0.12532052397727966, "global_step": 97377, "epoch": 1094, "lr": 7.163663895348008e-05}
{"train_loss": 0.09447664767503738, "global_step": 97378, "epoch": 1094, "lr": 7.163611630658679e-05}
{"train_loss": 0.1433660238981247, "global_step": 97379, "epoch": 1094, "lr": 7.163559365678477e-05}
{"train_loss": 0.13736097514629364, "global_step": 97380, "epoch": 1094, "lr": 7.163507100407412e-05}
{"train_loss": 0.21949660778045654, "global_step": 97381, "epoch": 1094, "lr": 7.163454834845489e-05}
{"train_loss": 0.1701369434595108, "global_step": 97382, "epoch": 1094, "lr": 7.163402568992717e-05}
{"train_loss": 0.20308776199817657, "global_step": 97383, "epoch": 1094, "lr": 7.1633503028491e-05}
{"train_loss": 0.2763140797615051, "global_step": 97384, "epoch": 1094, "lr": 7.163298036414648e-05}
{"train_loss": 0.2216911017894745, "global_step": 97385, "epoch": 1094, "lr": 7.163245769689366e-05}
{"train_loss": 0.1521405577659607, "global_step": 97386, "epoch": 1094, "lr": 7.163193502673264e-05}
{"train_loss": 0.17089752852916718, "global_step": 97387, "epoch": 1094, "lr": 7.163141235366345e-05}
{"train_loss": 0.13366274535655975, "global_step": 97388, "epoch": 1094, "lr": 7.16308896776862e-05}
{"train_loss": 0.15550783276557922, "global_step": 97389, "epoch": 1094, "lr": 7.163036699880093e-05}
{"train_loss": 0.15312686562538147, "global_step": 97390, "epoch": 1094, "lr": 7.162984431700771e-05}
{"train_loss": 0.1411055475473404, "global_step": 97391, "epoch": 1094, "lr": 7.162932163230663e-05}
{"train_loss": 0.1318054348230362, "global_step": 97392, "epoch": 1094, "lr": 7.162879894469777e-05}
{"train_loss": 0.17639054358005524, "global_step": 97393, "epoch": 1094, "lr": 7.162827625418116e-05}
{"train_loss": 0.15720386803150177, "global_step": 97394, "epoch": 1094, "lr": 7.162775356075688e-05}
{"train_loss": 0.11572223156690598, "global_step": 97395, "epoch": 1094, "lr": 7.162723086442504e-05}
{"train_loss": 0.21777740120887756, "global_step": 97396, "epoch": 1094, "lr": 7.162670816518567e-05}
{"train_loss": 0.11131133139133453, "global_step": 97397, "epoch": 1094, "lr": 7.162618546303886e-05}
{"train_loss": 0.08702154457569122, "global_step": 97398, "epoch": 1094, "lr": 7.162566275798466e-05}
{"train_loss": 0.2544432282447815, "global_step": 97399, "epoch": 1094, "lr": 7.162514005002315e-05}
{"train_loss": 0.1315394788980484, "global_step": 97400, "epoch": 1094, "lr": 7.162461733915442e-05}
{"train_loss": 0.13885661959648132, "global_step": 97401, "epoch": 1094, "lr": 7.162409462537851e-05}
{"train_loss": 0.1738917976617813, "global_step": 97402, "epoch": 1094, "lr": 7.162357190869552e-05}
{"train_loss": 0.18109863996505737, "global_step": 97403, "epoch": 1094, "lr": 7.16230491891055e-05}
{"train_loss": 0.17320407927036285, "global_step": 97404, "epoch": 1094, "lr": 7.162252646660852e-05}
{"train_loss": 0.20292037725448608, "global_step": 97405, "epoch": 1094, "lr": 7.162200374120465e-05}
{"train_loss": 0.11217397451400757, "global_step": 97406, "epoch": 1094, "lr": 7.162148101289397e-05}
{"train_loss": 0.22733044624328613, "global_step": 97407, "epoch": 1094, "lr": 7.162095828167654e-05}
{"train_loss": 0.13277561962604523, "global_step": 97408, "epoch": 1094, "lr": 7.162043554755245e-05}
{"train_loss": 0.19951234757900238, "global_step": 97409, "epoch": 1094, "lr": 7.161991281052174e-05}
{"train_loss": 0.16201527416706085, "global_step": 97410, "epoch": 1094, "lr": 7.16193900705845e-05}
{"train_loss": 0.097398541867733, "global_step": 97411, "epoch": 1094, "lr": 7.16188673277408e-05}
{"train_loss": 0.17278926074504852, "global_step": 97412, "epoch": 1094, "lr": 7.16183445819907e-05}
{"train_loss": 0.22945286333560944, "global_step": 97413, "epoch": 1094, "lr": 7.16178218333343e-05}
{"train_loss": 0.13443955779075623, "global_step": 97414, "epoch": 1094, "lr": 7.161729908177161e-05}
{"train_loss": 0.15576960146427155, "global_step": 97415, "epoch": 1094, "lr": 7.161677632730278e-05}
{"train_loss": 0.09954942762851715, "global_step": 97416, "epoch": 1094, "lr": 7.16162535699278e-05}
{"train_loss": 0.18090319633483887, "global_step": 97417, "epoch": 1094, "lr": 7.161573080964679e-05}
{"train_loss": 0.21464985609054565, "global_step": 97418, "epoch": 1094, "lr": 7.161520804645983e-05}
{"train_loss": 0.20780830085277557, "global_step": 97419, "epoch": 1094, "lr": 7.161468528036694e-05}
{"train_loss": 0.14465084671974182, "global_step": 97420, "epoch": 1094, "lr": 7.161416251136824e-05}
{"train_loss": 0.1420370191335678, "global_step": 97421, "epoch": 1094, "lr": 7.161363973946376e-05}
{"train_loss": 0.17076507210731506, "global_step": 97422, "epoch": 1094, "lr": 7.16131169646536e-05}
{"train_loss": 0.1647530198097229, "global_step": 97423, "epoch": 1094, "lr": 7.161259418693782e-05}
{"train_loss": 0.13341853022575378, "global_step": 97424, "epoch": 1094, "lr": 7.161207140631649e-05}
{"train_loss": 0.15926644206047058, "global_step": 97425, "epoch": 1094, "lr": 7.161154862278968e-05}
{"train_loss": 0.14580200612545013, "global_step": 97426, "epoch": 1094, "lr": 7.161102583635746e-05}
{"train_loss": 0.1415710151195526, "global_step": 97427, "epoch": 1094, "lr": 7.16105030470199e-05}
{"train_loss": 0.19313707947731018, "global_step": 97428, "epoch": 1094, "lr": 7.160998025477707e-05}
{"train_loss": 0.14445596933364868, "global_step": 97429, "epoch": 1094, "lr": 7.160945745962904e-05}
{"train_loss": 0.1036069393157959, "global_step": 97430, "epoch": 1094, "lr": 7.16089346615759e-05}
{"train_loss": 0.16903027892112732, "global_step": 97431, "epoch": 1094, "lr": 7.160841186061768e-05}
{"train_loss": 0.17474164068698883, "global_step": 97432, "epoch": 1094, "lr": 7.160788905675448e-05}
{"train_loss": 0.23144178092479706, "global_step": 97433, "epoch": 1094, "lr": 7.160736624998637e-05}
{"train_loss": 0.13981041312217712, "global_step": 97434, "epoch": 1094, "lr": 7.16068434403134e-05}
{"train_loss": 0.15169818699359894, "global_step": 97435, "epoch": 1094, "lr": 7.160632062773567e-05}
{"train_loss": 0.1389627605676651, "global_step": 97436, "epoch": 1094, "lr": 7.160579781225322e-05}
{"train_loss": 0.19716915488243103, "global_step": 97437, "epoch": 1094, "lr": 7.160527499386613e-05}
{"train_loss": 0.15500836074352264, "global_step": 97438, "epoch": 1094, "lr": 7.160475217257448e-05}
{"train_loss": 0.2193194180727005, "global_step": 97439, "epoch": 1094, "lr": 7.160422934837834e-05}
{"train_loss": 0.1698285937309265, "global_step": 97440, "epoch": 1094, "lr": 7.160370652127777e-05}
{"train_loss": 0.12202009558677673, "global_step": 97441, "epoch": 1094, "lr": 7.160318369127284e-05}
{"train_loss": 0.15507690608501434, "global_step": 97442, "epoch": 1094, "lr": 7.160266085836363e-05}
{"train_loss": 0.18326841294765472, "global_step": 97443, "epoch": 1094, "lr": 7.160213802255021e-05}
{"train_loss": 0.10558103024959564, "global_step": 97444, "epoch": 1094, "lr": 7.160161518383265e-05}
{"train_loss": 0.18890035152435303, "global_step": 97445, "epoch": 1094, "lr": 7.160109234221101e-05}
{"train_loss": 0.1306183934211731, "global_step": 97446, "epoch": 1094, "lr": 7.160056949768536e-05}
{"train_loss": 0.1037701964378357, "global_step": 97447, "epoch": 1094, "lr": 7.160004665025577e-05}
{"train_loss": 0.1639498770236969, "global_step": 97448, "epoch": 1094, "lr": 7.159952379992233e-05}
{"train_loss": 0.22186583280563354, "global_step": 97449, "epoch": 1094, "lr": 7.159900094668509e-05}
{"train_loss": 0.2146511673927307, "global_step": 97450, "epoch": 1094, "lr": 7.159847809054414e-05}
{"train_loss": 0.161296546459198, "global_step": 97451, "epoch": 1094, "lr": 7.159795523149953e-05}
{"train_loss": 0.18100830912590027, "global_step": 97452, "epoch": 1094, "lr": 7.159743236955135e-05}
{"train_loss": 0.15988101065158844, "global_step": 97453, "epoch": 1094, "lr": 7.159690950469965e-05}
{"train_loss": 0.1612609921211607, "global_step": 97454, "epoch": 1094, "lr": 7.15963866369445e-05, "val_loss": 4.679940223693848}
{"train_loss": 0.15549243986606598, "global_step": 97455, "epoch": 1095, "lr": 7.159586376628599e-05}
{"train_loss": 0.20832154154777527, "global_step": 97456, "epoch": 1095, "lr": 7.159534089272417e-05}
{"train_loss": 0.17050568759441376, "global_step": 97457, "epoch": 1095, "lr": 7.159481801625912e-05}
{"train_loss": 0.14773723483085632, "global_step": 97458, "epoch": 1095, "lr": 7.159429513689091e-05}
{"train_loss": 0.17118746042251587, "global_step": 97459, "epoch": 1095, "lr": 7.159377225461962e-05}
{"train_loss": 0.15690752863883972, "global_step": 97460, "epoch": 1095, "lr": 7.15932493694453e-05}
{"train_loss": 0.15643496811389923, "global_step": 97461, "epoch": 1095, "lr": 7.159272648136804e-05}
{"train_loss": 0.16142798960208893, "global_step": 97462, "epoch": 1095, "lr": 7.159220359038789e-05}
{"train_loss": 0.2605808675289154, "global_step": 97463, "epoch": 1095, "lr": 7.159168069650495e-05}
{"train_loss": 0.14133845269680023, "global_step": 97464, "epoch": 1095, "lr": 7.159115779971925e-05}
{"train_loss": 0.18322980403900146, "global_step": 97465, "epoch": 1095, "lr": 7.159063490003089e-05}
{"train_loss": 0.1680377572774887, "global_step": 97466, "epoch": 1095, "lr": 7.159011199743992e-05}
{"train_loss": 0.23024749755859375, "global_step": 97467, "epoch": 1095, "lr": 7.158958909194644e-05}
{"train_loss": 0.19180741906166077, "global_step": 97468, "epoch": 1095, "lr": 7.15890661835505e-05}
{"train_loss": 0.09878972172737122, "global_step": 97469, "epoch": 1095, "lr": 7.158854327225218e-05}
{"train_loss": 0.18306760489940643, "global_step": 97470, "epoch": 1095, "lr": 7.158802035805154e-05}
{"train_loss": 0.12427164614200592, "global_step": 97471, "epoch": 1095, "lr": 7.158749744094865e-05}
{"train_loss": 0.08489023149013519, "global_step": 97472, "epoch": 1095, "lr": 7.158697452094359e-05}
{"train_loss": 0.09352954477071762, "global_step": 97473, "epoch": 1095, "lr": 7.158645159803642e-05}
{"train_loss": 0.21709603071212769, "global_step": 97474, "epoch": 1095, "lr": 7.158592867222721e-05}
{"train_loss": 0.14167582988739014, "global_step": 97475, "epoch": 1095, "lr": 7.158540574351604e-05}
{"train_loss": 0.24623842537403107, "global_step": 97476, "epoch": 1095, "lr": 7.158488281190296e-05}
{"train_loss": 0.21125862002372742, "global_step": 97477, "epoch": 1095, "lr": 7.158435987738807e-05}
{"train_loss": 0.17490322887897491, "global_step": 97478, "epoch": 1095, "lr": 7.158383693997144e-05}
{"train_loss": 0.1473405957221985, "global_step": 97479, "epoch": 1095, "lr": 7.15833139996531e-05}
{"train_loss": 0.11962887644767761, "global_step": 97480, "epoch": 1095, "lr": 7.158279105643316e-05}
{"train_loss": 0.12416554987430573, "global_step": 97481, "epoch": 1095, "lr": 7.158226811031168e-05}
{"train_loss": 0.11804734915494919, "global_step": 97482, "epoch": 1095, "lr": 7.158174516128872e-05}
{"train_loss": 0.19831055402755737, "global_step": 97483, "epoch": 1095, "lr": 7.158122220936436e-05}
{"train_loss": 0.21979649364948273, "global_step": 97484, "epoch": 1095, "lr": 7.158069925453865e-05}
{"train_loss": 0.16121254861354828, "global_step": 97485, "epoch": 1095, "lr": 7.158017629681169e-05}
{"train_loss": 0.1634203940629959, "global_step": 97486, "epoch": 1095, "lr": 7.157965333618355e-05}
{"train_loss": 0.14945214986801147, "global_step": 97487, "epoch": 1095, "lr": 7.157913037265428e-05}
{"train_loss": 0.14060476422309875, "global_step": 97488, "epoch": 1095, "lr": 7.157860740622398e-05}
{"train_loss": 0.19262322783470154, "global_step": 97489, "epoch": 1095, "lr": 7.157808443689267e-05}
{"train_loss": 0.12310805171728134, "global_step": 97490, "epoch": 1095, "lr": 7.157756146466046e-05}
{"train_loss": 0.20651693642139435, "global_step": 97491, "epoch": 1095, "lr": 7.157703848952741e-05}
{"train_loss": 0.15175125002861023, "global_step": 97492, "epoch": 1095, "lr": 7.15765155114936e-05}
{"train_loss": 0.1519729495048523, "global_step": 97493, "epoch": 1095, "lr": 7.157599253055906e-05}
{"train_loss": 0.12117024511098862, "global_step": 97494, "epoch": 1095, "lr": 7.157546954672392e-05}
{"train_loss": 0.2125723659992218, "global_step": 97495, "epoch": 1095, "lr": 7.15749465599882e-05}
{"train_loss": 0.20118527114391327, "global_step": 97496, "epoch": 1095, "lr": 7.157442357035201e-05}
{"train_loss": 0.1575431376695633, "global_step": 97497, "epoch": 1095, "lr": 7.15739005778154e-05}
{"train_loss": 0.12128730118274689, "global_step": 97498, "epoch": 1095, "lr": 7.157337758237844e-05}
{"train_loss": 0.2787327170372009, "global_step": 97499, "epoch": 1095, "lr": 7.15728545840412e-05}
{"train_loss": 0.19030702114105225, "global_step": 97500, "epoch": 1095, "lr": 7.157233158280375e-05}
{"train_loss": 0.20365917682647705, "global_step": 97501, "epoch": 1095, "lr": 7.157180857866616e-05}
{"train_loss": 0.14511160552501678, "global_step": 97502, "epoch": 1095, "lr": 7.15712855716285e-05}
{"train_loss": 0.19129785895347595, "global_step": 97503, "epoch": 1095, "lr": 7.157076256169086e-05}
{"train_loss": 0.1852087676525116, "global_step": 97504, "epoch": 1095, "lr": 7.157023954885327e-05}
{"train_loss": 0.17011089622974396, "global_step": 97505, "epoch": 1095, "lr": 7.156971653311583e-05}
{"train_loss": 0.21238476037979126, "global_step": 97506, "epoch": 1095, "lr": 7.156919351447863e-05}
{"train_loss": 0.14821897447109222, "global_step": 97507, "epoch": 1095, "lr": 7.156867049294169e-05}
{"train_loss": 0.19693534076213837, "global_step": 97508, "epoch": 1095, "lr": 7.156814746850513e-05}
{"train_loss": 0.2310033142566681, "global_step": 97509, "epoch": 1095, "lr": 7.156762444116898e-05}
{"train_loss": 0.1190761923789978, "global_step": 97510, "epoch": 1095, "lr": 7.156710141093332e-05}
{"train_loss": 0.17735938727855682, "global_step": 97511, "epoch": 1095, "lr": 7.156657837779824e-05}
{"train_loss": 0.13624191284179688, "global_step": 97512, "epoch": 1095, "lr": 7.156605534176378e-05}
{"train_loss": 0.20617231726646423, "global_step": 97513, "epoch": 1095, "lr": 7.156553230283004e-05}
{"train_loss": 0.21379944682121277, "global_step": 97514, "epoch": 1095, "lr": 7.156500926099708e-05}
{"train_loss": 0.18042443692684174, "global_step": 97515, "epoch": 1095, "lr": 7.156448621626494e-05}
{"train_loss": 0.17603415250778198, "global_step": 97516, "epoch": 1095, "lr": 7.156396316863374e-05}
{"train_loss": 0.1280360221862793, "global_step": 97517, "epoch": 1095, "lr": 7.156344011810353e-05}
{"train_loss": 0.1950841099023819, "global_step": 97518, "epoch": 1095, "lr": 7.156291706467438e-05}
{"train_loss": 0.23206935822963715, "global_step": 97519, "epoch": 1095, "lr": 7.156239400834634e-05}
{"train_loss": 0.14643020927906036, "global_step": 97520, "epoch": 1095, "lr": 7.156187094911951e-05}
{"train_loss": 0.11611129343509674, "global_step": 97521, "epoch": 1095, "lr": 7.156134788699394e-05}
{"train_loss": 0.12409175932407379, "global_step": 97522, "epoch": 1095, "lr": 7.156082482196973e-05}
{"train_loss": 0.1356215626001358, "global_step": 97523, "epoch": 1095, "lr": 7.15603017540469e-05}
{"train_loss": 0.18389061093330383, "global_step": 97524, "epoch": 1095, "lr": 7.155977868322559e-05}
{"train_loss": 0.1443709284067154, "global_step": 97525, "epoch": 1095, "lr": 7.155925560950581e-05}
{"train_loss": 0.18342547118663788, "global_step": 97526, "epoch": 1095, "lr": 7.155873253288764e-05}
{"train_loss": 0.1544889211654663, "global_step": 97527, "epoch": 1095, "lr": 7.155820945337119e-05}
{"train_loss": 0.1764201521873474, "global_step": 97528, "epoch": 1095, "lr": 7.155768637095648e-05}
{"train_loss": 0.2112923413515091, "global_step": 97529, "epoch": 1095, "lr": 7.15571632856436e-05}
{"train_loss": 0.22488266229629517, "global_step": 97530, "epoch": 1095, "lr": 7.155664019743263e-05}
{"train_loss": 0.1711450070142746, "global_step": 97531, "epoch": 1095, "lr": 7.155611710632363e-05}
{"train_loss": 0.1987386792898178, "global_step": 97532, "epoch": 1095, "lr": 7.155559401231667e-05}
{"train_loss": 0.16981980204582214, "global_step": 97533, "epoch": 1095, "lr": 7.155507091541182e-05}
{"train_loss": 0.14576414227485657, "global_step": 97534, "epoch": 1095, "lr": 7.155454781560916e-05}
{"train_loss": 0.15336595475673676, "global_step": 97535, "epoch": 1095, "lr": 7.155402471290875e-05}
{"train_loss": 0.22332507371902466, "global_step": 97536, "epoch": 1095, "lr": 7.155350160731067e-05}
{"train_loss": 0.10518274456262589, "global_step": 97537, "epoch": 1095, "lr": 7.155297849881498e-05}
{"train_loss": 0.17650942504405975, "global_step": 97538, "epoch": 1095, "lr": 7.155245538742175e-05}
{"train_loss": 0.15528711676597595, "global_step": 97539, "epoch": 1095, "lr": 7.155193227313106e-05}
{"train_loss": 0.1809690296649933, "global_step": 97540, "epoch": 1095, "lr": 7.155140915594298e-05}
{"train_loss": 0.1173761785030365, "global_step": 97541, "epoch": 1095, "lr": 7.155088603585756e-05}
{"train_loss": 0.14317169785499573, "global_step": 97542, "epoch": 1095, "lr": 7.155036291287491e-05}
{"train_loss": 0.16877019346764918, "global_step": 97543, "epoch": 1095, "lr": 7.154983978699506e-05, "val_loss": 4.599028587341309, "train_action_mse_error": 31.63509178161621}
{"train_loss": 0.11749731749296188, "global_step": 97544, "epoch": 1096, "lr": 7.15493166582181e-05}
{"train_loss": 0.1991645097732544, "global_step": 97545, "epoch": 1096, "lr": 7.15487935265441e-05}
{"train_loss": 0.12752321362495422, "global_step": 97546, "epoch": 1096, "lr": 7.154827039197312e-05}
{"train_loss": 0.13715551793575287, "global_step": 97547, "epoch": 1096, "lr": 7.154774725450526e-05}
{"train_loss": 0.23309391736984253, "global_step": 97548, "epoch": 1096, "lr": 7.154722411414056e-05}
{"train_loss": 0.21127980947494507, "global_step": 97549, "epoch": 1096, "lr": 7.154670097087908e-05}
{"train_loss": 0.19361688196659088, "global_step": 97550, "epoch": 1096, "lr": 7.154617782472093e-05}
{"train_loss": 0.2162139117717743, "global_step": 97551, "epoch": 1096, "lr": 7.154565467566614e-05}
{"train_loss": 0.17347434163093567, "global_step": 97552, "epoch": 1096, "lr": 7.15451315237148e-05}
{"train_loss": 0.19066010415554047, "global_step": 97553, "epoch": 1096, "lr": 7.1544608368867e-05}
{"train_loss": 0.19702257215976715, "global_step": 97554, "epoch": 1096, "lr": 7.154408521112276e-05}
{"train_loss": 0.12946856021881104, "global_step": 97555, "epoch": 1096, "lr": 7.154356205048221e-05}
{"train_loss": 0.1604522317647934, "global_step": 97556, "epoch": 1096, "lr": 7.154303888694536e-05}
{"train_loss": 0.10534589737653732, "global_step": 97557, "epoch": 1096, "lr": 7.154251572051233e-05}
{"train_loss": 0.11623504757881165, "global_step": 97558, "epoch": 1096, "lr": 7.154199255118317e-05}
{"train_loss": 0.0850793793797493, "global_step": 97559, "epoch": 1096, "lr": 7.154146937895795e-05}
{"train_loss": 0.19631706178188324, "global_step": 97560, "epoch": 1096, "lr": 7.154094620383674e-05}
{"train_loss": 0.1666889637708664, "global_step": 97561, "epoch": 1096, "lr": 7.154042302581962e-05}
{"train_loss": 0.14172446727752686, "global_step": 97562, "epoch": 1096, "lr": 7.153989984490666e-05}
{"train_loss": 0.1832328587770462, "global_step": 97563, "epoch": 1096, "lr": 7.15393766610979e-05}
{"train_loss": 0.17248201370239258, "global_step": 97564, "epoch": 1096, "lr": 7.153885347439345e-05}
{"train_loss": 0.16176949441432953, "global_step": 97565, "epoch": 1096, "lr": 7.153833028479335e-05}
{"train_loss": 0.15488141775131226, "global_step": 97566, "epoch": 1096, "lr": 7.15378070922977e-05}
{"train_loss": 0.20651772618293762, "global_step": 97567, "epoch": 1096, "lr": 7.153728389690654e-05}
{"train_loss": 0.10571303963661194, "global_step": 97568, "epoch": 1096, "lr": 7.153676069861998e-05}
{"train_loss": 0.21440823376178741, "global_step": 97569, "epoch": 1096, "lr": 7.153623749743804e-05}
{"train_loss": 0.2186802178621292, "global_step": 97570, "epoch": 1096, "lr": 7.153571429336083e-05}
{"train_loss": 0.2296227365732193, "global_step": 97571, "epoch": 1096, "lr": 7.153519108638838e-05}
{"train_loss": 0.17535413801670074, "global_step": 97572, "epoch": 1096, "lr": 7.153466787652081e-05}
{"train_loss": 0.13240475952625275, "global_step": 97573, "epoch": 1096, "lr": 7.153414466375817e-05}
{"train_loss": 0.2855016589164734, "global_step": 97574, "epoch": 1096, "lr": 7.15336214481005e-05}
{"train_loss": 0.06325262784957886, "global_step": 97575, "epoch": 1096, "lr": 7.153309822954792e-05}
{"train_loss": 0.14541320502758026, "global_step": 97576, "epoch": 1096, "lr": 7.153257500810047e-05}
{"train_loss": 0.19357092678546906, "global_step": 97577, "epoch": 1096, "lr": 7.153205178375822e-05}
{"train_loss": 0.2501205503940582, "global_step": 97578, "epoch": 1096, "lr": 7.153152855652125e-05}
{"train_loss": 0.13459935784339905, "global_step": 97579, "epoch": 1096, "lr": 7.153100532638964e-05}
{"train_loss": 0.20737315714359283, "global_step": 97580, "epoch": 1096, "lr": 7.153048209336345e-05}
{"train_loss": 0.13787581026554108, "global_step": 97581, "epoch": 1096, "lr": 7.152995885744273e-05}
{"train_loss": 0.14490313827991486, "global_step": 97582, "epoch": 1096, "lr": 7.15294356186276e-05}
{"train_loss": 0.13917501270771027, "global_step": 97583, "epoch": 1096, "lr": 7.152891237691807e-05}
{"train_loss": 0.16605244576931, "global_step": 97584, "epoch": 1096, "lr": 7.152838913231426e-05}
{"train_loss": 0.11757858097553253, "global_step": 97585, "epoch": 1096, "lr": 7.152786588481621e-05}
{"train_loss": 0.11745204776525497, "global_step": 97586, "epoch": 1096, "lr": 7.152734263442402e-05}
{"train_loss": 0.16963879764080048, "global_step": 97587, "epoch": 1096, "lr": 7.152681938113772e-05}
{"train_loss": 0.23966747522354126, "global_step": 97588, "epoch": 1096, "lr": 7.15262961249574e-05}
{"train_loss": 0.14093568921089172, "global_step": 97589, "epoch": 1096, "lr": 7.152577286588314e-05}
{"train_loss": 0.16009646654129028, "global_step": 97590, "epoch": 1096, "lr": 7.1525249603915e-05}
{"train_loss": 0.24147111177444458, "global_step": 97591, "epoch": 1096, "lr": 7.152472633905306e-05}
{"train_loss": 0.22120030224323273, "global_step": 97592, "epoch": 1096, "lr": 7.152420307129737e-05}
{"train_loss": 0.10228945314884186, "global_step": 97593, "epoch": 1096, "lr": 7.152367980064801e-05}
{"train_loss": 0.11227140575647354, "global_step": 97594, "epoch": 1096, "lr": 7.152315652710507e-05}
{"train_loss": 0.1842552274465561, "global_step": 97595, "epoch": 1096, "lr": 7.152263325066859e-05}
{"train_loss": 0.19330409169197083, "global_step": 97596, "epoch": 1096, "lr": 7.152210997133866e-05}
{"train_loss": 0.13755270838737488, "global_step": 97597, "epoch": 1096, "lr": 7.152158668911535e-05}
{"train_loss": 0.18820001184940338, "global_step": 97598, "epoch": 1096, "lr": 7.152106340399872e-05}
{"train_loss": 0.13711793720722198, "global_step": 97599, "epoch": 1096, "lr": 7.152054011598885e-05}
{"train_loss": 0.184550479054451, "global_step": 97600, "epoch": 1096, "lr": 7.15200168250858e-05}
{"train_loss": 0.14003382623195648, "global_step": 97601, "epoch": 1096, "lr": 7.151949353128964e-05}
{"train_loss": 0.19039888679981232, "global_step": 97602, "epoch": 1096, "lr": 7.151897023460045e-05}
{"train_loss": 0.1751377135515213, "global_step": 97603, "epoch": 1096, "lr": 7.151844693501831e-05}
{"train_loss": 0.1872156709432602, "global_step": 97604, "epoch": 1096, "lr": 7.151792363254326e-05}
{"train_loss": 0.14262332022190094, "global_step": 97605, "epoch": 1096, "lr": 7.151740032717541e-05}
{"train_loss": 0.24773851037025452, "global_step": 97606, "epoch": 1096, "lr": 7.151687701891479e-05}
{"train_loss": 0.1583538055419922, "global_step": 97607, "epoch": 1096, "lr": 7.15163537077615e-05}
{"train_loss": 0.15910668671131134, "global_step": 97608, "epoch": 1096, "lr": 7.151583039371559e-05}
{"train_loss": 0.1004471704363823, "global_step": 97609, "epoch": 1096, "lr": 7.151530707677714e-05}
{"train_loss": 0.13606171309947968, "global_step": 97610, "epoch": 1096, "lr": 7.151478375694622e-05}
{"train_loss": 0.10136596858501434, "global_step": 97611, "epoch": 1096, "lr": 7.151426043422287e-05}
{"train_loss": 0.23504406213760376, "global_step": 97612, "epoch": 1096, "lr": 7.151373710860723e-05}
{"train_loss": 0.16030046343803406, "global_step": 97613, "epoch": 1096, "lr": 7.151321378009931e-05}
{"train_loss": 0.1247684583067894, "global_step": 97614, "epoch": 1096, "lr": 7.151269044869922e-05}
{"train_loss": 0.18554288148880005, "global_step": 97615, "epoch": 1096, "lr": 7.151216711440699e-05}
{"train_loss": 0.20852482318878174, "global_step": 97616, "epoch": 1096, "lr": 7.151164377722272e-05}
{"train_loss": 0.19156526029109955, "global_step": 97617, "epoch": 1096, "lr": 7.151112043714648e-05}
{"train_loss": 0.2458731085062027, "global_step": 97618, "epoch": 1096, "lr": 7.151059709417831e-05}
{"train_loss": 0.12381716072559357, "global_step": 97619, "epoch": 1096, "lr": 7.151007374831832e-05}
{"train_loss": 0.19337095320224762, "global_step": 97620, "epoch": 1096, "lr": 7.150955039956656e-05}
{"train_loss": 0.15024927258491516, "global_step": 97621, "epoch": 1096, "lr": 7.15090270479231e-05}
{"train_loss": 0.22326481342315674, "global_step": 97622, "epoch": 1096, "lr": 7.150850369338802e-05}
{"train_loss": 0.17425838112831116, "global_step": 97623, "epoch": 1096, "lr": 7.150798033596138e-05}
{"train_loss": 0.17112600803375244, "global_step": 97624, "epoch": 1096, "lr": 7.150745697564326e-05}
{"train_loss": 0.2268574982881546, "global_step": 97625, "epoch": 1096, "lr": 7.150693361243373e-05}
{"train_loss": 0.1893179714679718, "global_step": 97626, "epoch": 1096, "lr": 7.150641024633284e-05}
{"train_loss": 0.21189278364181519, "global_step": 97627, "epoch": 1096, "lr": 7.150588687734068e-05}
{"train_loss": 0.11439703404903412, "global_step": 97628, "epoch": 1096, "lr": 7.150536350545732e-05}
{"train_loss": 0.13242082297801971, "global_step": 97629, "epoch": 1096, "lr": 7.15048401306828e-05}
{"train_loss": 0.19558313488960266, "global_step": 97630, "epoch": 1096, "lr": 7.150431675301723e-05}
{"train_loss": 0.2574401795864105, "global_step": 97631, "epoch": 1096, "lr": 7.150379337246069e-05}
{"train_loss": 0.1694762629320782, "global_step": 97632, "epoch": 1096, "lr": 7.15032699890132e-05, "val_loss": 4.577155113220215}
{"train_loss": 0.17341792583465576, "global_step": 97633, "epoch": 1097, "lr": 7.150274660267486e-05}
{"train_loss": 0.15798647701740265, "global_step": 97634, "epoch": 1097, "lr": 7.150222321344575e-05}
{"train_loss": 0.15871545672416687, "global_step": 97635, "epoch": 1097, "lr": 7.150169982132594e-05}
{"train_loss": 0.14303243160247803, "global_step": 97636, "epoch": 1097, "lr": 7.150117642631547e-05}
{"train_loss": 0.19878387451171875, "global_step": 97637, "epoch": 1097, "lr": 7.150065302841443e-05}
{"train_loss": 0.2054632008075714, "global_step": 97638, "epoch": 1097, "lr": 7.150012962762288e-05}
{"train_loss": 0.21381086111068726, "global_step": 97639, "epoch": 1097, "lr": 7.149960622394091e-05}
{"train_loss": 0.16266648471355438, "global_step": 97640, "epoch": 1097, "lr": 7.149908281736858e-05}
{"train_loss": 0.21704566478729248, "global_step": 97641, "epoch": 1097, "lr": 7.149855940790597e-05}
{"train_loss": 0.18908396363258362, "global_step": 97642, "epoch": 1097, "lr": 7.149803599555311e-05}
{"train_loss": 0.14352713525295258, "global_step": 97643, "epoch": 1097, "lr": 7.149751258031014e-05}
{"train_loss": 0.18084298074245453, "global_step": 97644, "epoch": 1097, "lr": 7.149698916217708e-05}
{"train_loss": 0.1537141054868698, "global_step": 97645, "epoch": 1097, "lr": 7.149646574115401e-05}
{"train_loss": 0.1878773719072342, "global_step": 97646, "epoch": 1097, "lr": 7.1495942317241e-05}
{"train_loss": 0.14380711317062378, "global_step": 97647, "epoch": 1097, "lr": 7.14954188904381e-05}
{"train_loss": 0.18269865214824677, "global_step": 97648, "epoch": 1097, "lr": 7.149489546074542e-05}
{"train_loss": 0.1180892139673233, "global_step": 97649, "epoch": 1097, "lr": 7.149437202816303e-05}
{"train_loss": 0.1546046882867813, "global_step": 97650, "epoch": 1097, "lr": 7.149384859269096e-05}
{"train_loss": 0.17725567519664764, "global_step": 97651, "epoch": 1097, "lr": 7.149332515432931e-05}
{"train_loss": 0.16102847456932068, "global_step": 97652, "epoch": 1097, "lr": 7.149280171307815e-05}
{"train_loss": 0.20879623293876648, "global_step": 97653, "epoch": 1097, "lr": 7.149227826893754e-05}
{"train_loss": 0.1449848711490631, "global_step": 97654, "epoch": 1097, "lr": 7.149175482190756e-05}
{"train_loss": 0.12965083122253418, "global_step": 97655, "epoch": 1097, "lr": 7.149123137198827e-05}
{"train_loss": 0.14342951774597168, "global_step": 97656, "epoch": 1097, "lr": 7.149070791917976e-05}
{"train_loss": 0.21005722880363464, "global_step": 97657, "epoch": 1097, "lr": 7.149018446348208e-05}
{"train_loss": 0.19248196482658386, "global_step": 97658, "epoch": 1097, "lr": 7.148966100489529e-05}
{"train_loss": 0.18461382389068604, "global_step": 97659, "epoch": 1097, "lr": 7.148913754341949e-05}
{"train_loss": 0.2324114441871643, "global_step": 97660, "epoch": 1097, "lr": 7.148861407905474e-05}
{"train_loss": 0.19909296929836273, "global_step": 97661, "epoch": 1097, "lr": 7.14880906118011e-05}
{"train_loss": 0.20010100305080414, "global_step": 97662, "epoch": 1097, "lr": 7.148756714165866e-05}
{"train_loss": 0.14149628579616547, "global_step": 97663, "epoch": 1097, "lr": 7.148704366862749e-05}
{"train_loss": 0.1614651083946228, "global_step": 97664, "epoch": 1097, "lr": 7.148652019270762e-05}
{"train_loss": 0.10791855305433273, "global_step": 97665, "epoch": 1097, "lr": 7.148599671389916e-05}
{"train_loss": 0.12126071006059647, "global_step": 97666, "epoch": 1097, "lr": 7.148547323220216e-05}
{"train_loss": 0.17658033967018127, "global_step": 97667, "epoch": 1097, "lr": 7.14849497476167e-05}
{"train_loss": 0.17555946111679077, "global_step": 97668, "epoch": 1097, "lr": 7.148442626014287e-05}
{"train_loss": 0.17878690361976624, "global_step": 97669, "epoch": 1097, "lr": 7.14839027697807e-05}
{"train_loss": 0.15582816302776337, "global_step": 97670, "epoch": 1097, "lr": 7.14833792765303e-05}
{"train_loss": 0.21457353234291077, "global_step": 97671, "epoch": 1097, "lr": 7.14828557803917e-05}
{"train_loss": 0.17522791028022766, "global_step": 97672, "epoch": 1097, "lr": 7.148233228136501e-05}
{"train_loss": 0.20031321048736572, "global_step": 97673, "epoch": 1097, "lr": 7.148180877945029e-05}
{"train_loss": 0.20765657722949982, "global_step": 97674, "epoch": 1097, "lr": 7.148128527464759e-05}
{"train_loss": 0.19974122941493988, "global_step": 97675, "epoch": 1097, "lr": 7.1480761766957e-05}
{"train_loss": 0.12233313918113708, "global_step": 97676, "epoch": 1097, "lr": 7.148023825637856e-05}
{"train_loss": 0.08544301241636276, "global_step": 97677, "epoch": 1097, "lr": 7.147971474291238e-05}
{"train_loss": 0.1730653941631317, "global_step": 97678, "epoch": 1097, "lr": 7.147919122655851e-05}
{"train_loss": 0.12894268333911896, "global_step": 97679, "epoch": 1097, "lr": 7.147866770731703e-05}
{"train_loss": 0.20372235774993896, "global_step": 97680, "epoch": 1097, "lr": 7.147814418518801e-05}
{"train_loss": 0.17927561700344086, "global_step": 97681, "epoch": 1097, "lr": 7.147762066017151e-05}
{"train_loss": 0.19296705722808838, "global_step": 97682, "epoch": 1097, "lr": 7.147709713226761e-05}
{"train_loss": 0.19635535776615143, "global_step": 97683, "epoch": 1097, "lr": 7.147657360147635e-05}
{"train_loss": 0.2215336561203003, "global_step": 97684, "epoch": 1097, "lr": 7.147605006779787e-05}
{"train_loss": 0.18206150829792023, "global_step": 97685, "epoch": 1097, "lr": 7.147552653123216e-05}
{"train_loss": 0.19485493004322052, "global_step": 97686, "epoch": 1097, "lr": 7.147500299177934e-05}
{"train_loss": 0.13902994990348816, "global_step": 97687, "epoch": 1097, "lr": 7.147447944943947e-05}
{"train_loss": 0.19827425479888916, "global_step": 97688, "epoch": 1097, "lr": 7.147395590421262e-05}
{"train_loss": 0.16306349635124207, "global_step": 97689, "epoch": 1097, "lr": 7.147343235609885e-05}
{"train_loss": 0.17312310636043549, "global_step": 97690, "epoch": 1097, "lr": 7.147290880509824e-05}
{"train_loss": 0.10993926227092743, "global_step": 97691, "epoch": 1097, "lr": 7.147238525121087e-05}
{"train_loss": 0.1475732922554016, "global_step": 97692, "epoch": 1097, "lr": 7.147186169443679e-05}
{"train_loss": 0.14995864033699036, "global_step": 97693, "epoch": 1097, "lr": 7.147133813477609e-05}
{"train_loss": 0.18534816801548004, "global_step": 97694, "epoch": 1097, "lr": 7.147081457222882e-05}
{"train_loss": 0.26487329602241516, "global_step": 97695, "epoch": 1097, "lr": 7.147029100679506e-05}
{"train_loss": 0.1726723164319992, "global_step": 97696, "epoch": 1097, "lr": 7.146976743847489e-05}
{"train_loss": 0.1825401932001114, "global_step": 97697, "epoch": 1097, "lr": 7.146924386726837e-05}
{"train_loss": 0.177817240357399, "global_step": 97698, "epoch": 1097, "lr": 7.146872029317556e-05}
{"train_loss": 0.16726554930210114, "global_step": 97699, "epoch": 1097, "lr": 7.146819671619654e-05}
{"train_loss": 0.19642353057861328, "global_step": 97700, "epoch": 1097, "lr": 7.146767313633141e-05}
{"train_loss": 0.17428605258464813, "global_step": 97701, "epoch": 1097, "lr": 7.146714955358018e-05}
{"train_loss": 0.1992645561695099, "global_step": 97702, "epoch": 1097, "lr": 7.146662596794298e-05}
{"train_loss": 0.17195752263069153, "global_step": 97703, "epoch": 1097, "lr": 7.146610237941984e-05}
{"train_loss": 0.18858008086681366, "global_step": 97704, "epoch": 1097, "lr": 7.146557878801084e-05}
{"train_loss": 0.15823717415332794, "global_step": 97705, "epoch": 1097, "lr": 7.146505519371607e-05}
{"train_loss": 0.2199387401342392, "global_step": 97706, "epoch": 1097, "lr": 7.146453159653557e-05}
{"train_loss": 0.11265204101800919, "global_step": 97707, "epoch": 1097, "lr": 7.146400799646943e-05}
{"train_loss": 0.14097101986408234, "global_step": 97708, "epoch": 1097, "lr": 7.146348439351772e-05}
{"train_loss": 0.16038881242275238, "global_step": 97709, "epoch": 1097, "lr": 7.14629607876805e-05}
{"train_loss": 0.2732365131378174, "global_step": 97710, "epoch": 1097, "lr": 7.146243717895786e-05}
{"train_loss": 0.22064070403575897, "global_step": 97711, "epoch": 1097, "lr": 7.146191356734983e-05}
{"train_loss": 0.12941254675388336, "global_step": 97712, "epoch": 1097, "lr": 7.146138995285654e-05}
{"train_loss": 0.19583582878112793, "global_step": 97713, "epoch": 1097, "lr": 7.1460866335478e-05}
{"train_loss": 0.18238939344882965, "global_step": 97714, "epoch": 1097, "lr": 7.146034271521432e-05}
{"train_loss": 0.16579373180866241, "global_step": 97715, "epoch": 1097, "lr": 7.145981909206556e-05}
{"train_loss": 0.23394104838371277, "global_step": 97716, "epoch": 1097, "lr": 7.145929546603179e-05}
{"train_loss": 0.12881897389888763, "global_step": 97717, "epoch": 1097, "lr": 7.145877183711307e-05}
{"train_loss": 0.16545122861862183, "global_step": 97718, "epoch": 1097, "lr": 7.145824820530949e-05}
{"train_loss": 0.23410505056381226, "global_step": 97719, "epoch": 1097, "lr": 7.14577245706211e-05}
{"train_loss": 0.17442257702350616, "global_step": 97720, "epoch": 1097, "lr": 7.1457200933048e-05}
{"train_loss": 0.17481020595250504, "global_step": 97721, "epoch": 1097, "lr": 7.145667729259022e-05, "val_loss": 4.480071544647217}
{"train_loss": 0.30515822768211365, "global_step": 97722, "epoch": 1098, "lr": 7.145615364924785e-05}
{"train_loss": 0.11291834712028503, "global_step": 97723, "epoch": 1098, "lr": 7.145563000302098e-05}
{"train_loss": 0.14722171425819397, "global_step": 97724, "epoch": 1098, "lr": 7.145510635390965e-05}
{"train_loss": 0.16113290190696716, "global_step": 97725, "epoch": 1098, "lr": 7.145458270191395e-05}
{"train_loss": 0.1832689344882965, "global_step": 97726, "epoch": 1098, "lr": 7.145405904703394e-05}
{"train_loss": 0.15964943170547485, "global_step": 97727, "epoch": 1098, "lr": 7.145353538926968e-05}
{"train_loss": 0.16338787972927094, "global_step": 97728, "epoch": 1098, "lr": 7.145301172862128e-05}
{"train_loss": 0.1479753702878952, "global_step": 97729, "epoch": 1098, "lr": 7.145248806508876e-05}
{"train_loss": 0.14224031567573547, "global_step": 97730, "epoch": 1098, "lr": 7.145196439867223e-05}
{"train_loss": 0.12953364849090576, "global_step": 97731, "epoch": 1098, "lr": 7.145144072937173e-05}
{"train_loss": 0.2032671719789505, "global_step": 97732, "epoch": 1098, "lr": 7.145091705718737e-05}
{"train_loss": 0.16108031570911407, "global_step": 97733, "epoch": 1098, "lr": 7.145039338211916e-05}
{"train_loss": 0.1267378181219101, "global_step": 97734, "epoch": 1098, "lr": 7.144986970416723e-05}
{"train_loss": 0.17150111496448517, "global_step": 97735, "epoch": 1098, "lr": 7.144934602333162e-05}
{"train_loss": 0.22346970438957214, "global_step": 97736, "epoch": 1098, "lr": 7.14488223396124e-05}
{"train_loss": 0.13136711716651917, "global_step": 97737, "epoch": 1098, "lr": 7.144829865300966e-05}
{"train_loss": 0.20993873476982117, "global_step": 97738, "epoch": 1098, "lr": 7.144777496352346e-05}
{"train_loss": 0.13177432119846344, "global_step": 97739, "epoch": 1098, "lr": 7.144725127115386e-05}
{"train_loss": 0.17100322246551514, "global_step": 97740, "epoch": 1098, "lr": 7.144672757590094e-05}
{"train_loss": 0.21717189252376556, "global_step": 97741, "epoch": 1098, "lr": 7.144620387776478e-05}
{"train_loss": 0.11457917094230652, "global_step": 97742, "epoch": 1098, "lr": 7.144568017674542e-05}
{"train_loss": 0.15758442878723145, "global_step": 97743, "epoch": 1098, "lr": 7.144515647284296e-05}
{"train_loss": 0.17522616684436798, "global_step": 97744, "epoch": 1098, "lr": 7.144463276605745e-05}
{"train_loss": 0.20114274322986603, "global_step": 97745, "epoch": 1098, "lr": 7.144410905638899e-05}
{"train_loss": 0.12428867816925049, "global_step": 97746, "epoch": 1098, "lr": 7.144358534383761e-05}
{"train_loss": 0.17885373532772064, "global_step": 97747, "epoch": 1098, "lr": 7.14430616284034e-05}
{"train_loss": 0.13967207074165344, "global_step": 97748, "epoch": 1098, "lr": 7.144253791008645e-05}
{"train_loss": 0.1349569708108902, "global_step": 97749, "epoch": 1098, "lr": 7.144201418888681e-05}
{"train_loss": 0.16598182916641235, "global_step": 97750, "epoch": 1098, "lr": 7.144149046480455e-05}
{"train_loss": 0.1791180670261383, "global_step": 97751, "epoch": 1098, "lr": 7.144096673783973e-05}
{"train_loss": 0.21610207855701447, "global_step": 97752, "epoch": 1098, "lr": 7.144044300799244e-05}
{"train_loss": 0.16915263235569, "global_step": 97753, "epoch": 1098, "lr": 7.143991927526276e-05}
{"train_loss": 0.08966910094022751, "global_step": 97754, "epoch": 1098, "lr": 7.143939553965072e-05}
{"train_loss": 0.1680060178041458, "global_step": 97755, "epoch": 1098, "lr": 7.143887180115643e-05}
{"train_loss": 0.1356930136680603, "global_step": 97756, "epoch": 1098, "lr": 7.143834805977993e-05}
{"train_loss": 0.20800526440143585, "global_step": 97757, "epoch": 1098, "lr": 7.143782431552133e-05}
{"train_loss": 0.1980128437280655, "global_step": 97758, "epoch": 1098, "lr": 7.143730056838066e-05}
{"train_loss": 0.11564502865076065, "global_step": 97759, "epoch": 1098, "lr": 7.143677681835801e-05}
{"train_loss": 0.15056708455085754, "global_step": 97760, "epoch": 1098, "lr": 7.143625306545344e-05}
{"train_loss": 0.17906109988689423, "global_step": 97761, "epoch": 1098, "lr": 7.143572930966703e-05}
{"train_loss": 0.1689753383398056, "global_step": 97762, "epoch": 1098, "lr": 7.143520555099885e-05}
{"train_loss": 0.16450664401054382, "global_step": 97763, "epoch": 1098, "lr": 7.143468178944896e-05}
{"train_loss": 0.1739625632762909, "global_step": 97764, "epoch": 1098, "lr": 7.143415802501745e-05}
{"train_loss": 0.15938261151313782, "global_step": 97765, "epoch": 1098, "lr": 7.143363425770437e-05}
{"train_loss": 0.14769244194030762, "global_step": 97766, "epoch": 1098, "lr": 7.143311048750981e-05}
{"train_loss": 0.12802191078662872, "global_step": 97767, "epoch": 1098, "lr": 7.143258671443382e-05}
{"train_loss": 0.13352881371974945, "global_step": 97768, "epoch": 1098, "lr": 7.143206293847648e-05}
{"train_loss": 0.1552431732416153, "global_step": 97769, "epoch": 1098, "lr": 7.143153915963788e-05}
{"train_loss": 0.1275005042552948, "global_step": 97770, "epoch": 1098, "lr": 7.143101537791804e-05}
{"train_loss": 0.1449599415063858, "global_step": 97771, "epoch": 1098, "lr": 7.143049159331708e-05}
{"train_loss": 0.1349237859249115, "global_step": 97772, "epoch": 1098, "lr": 7.142996780583504e-05}
{"train_loss": 0.27128538489341736, "global_step": 97773, "epoch": 1098, "lr": 7.142944401547201e-05}
{"train_loss": 0.19783872365951538, "global_step": 97774, "epoch": 1098, "lr": 7.142892022222806e-05}
{"train_loss": 0.1164635494351387, "global_step": 97775, "epoch": 1098, "lr": 7.142839642610324e-05}
{"train_loss": 0.1701740175485611, "global_step": 97776, "epoch": 1098, "lr": 7.142787262709763e-05}
{"train_loss": 0.12799270451068878, "global_step": 97777, "epoch": 1098, "lr": 7.142734882521132e-05}
{"train_loss": 0.15668544173240662, "global_step": 97778, "epoch": 1098, "lr": 7.142682502044435e-05}
{"train_loss": 0.13481715321540833, "global_step": 97779, "epoch": 1098, "lr": 7.142630121279681e-05}
{"train_loss": 0.22208550572395325, "global_step": 97780, "epoch": 1098, "lr": 7.142577740226877e-05}
{"train_loss": 0.1921655237674713, "global_step": 97781, "epoch": 1098, "lr": 7.142525358886028e-05}
{"train_loss": 0.19718578457832336, "global_step": 97782, "epoch": 1098, "lr": 7.142472977257145e-05}
{"train_loss": 0.10662954300642014, "global_step": 97783, "epoch": 1098, "lr": 7.14242059534023e-05}
{"train_loss": 0.2082265317440033, "global_step": 97784, "epoch": 1098, "lr": 7.142368213135295e-05}
{"train_loss": 0.1209779679775238, "global_step": 97785, "epoch": 1098, "lr": 7.142315830642343e-05}
{"train_loss": 0.20976297557353973, "global_step": 97786, "epoch": 1098, "lr": 7.142263447861383e-05}
{"train_loss": 0.206058070063591, "global_step": 97787, "epoch": 1098, "lr": 7.142211064792422e-05}
{"train_loss": 0.1089087426662445, "global_step": 97788, "epoch": 1098, "lr": 7.142158681435467e-05}
{"train_loss": 0.21129091084003448, "global_step": 97789, "epoch": 1098, "lr": 7.142106297790525e-05}
{"train_loss": 0.19392158091068268, "global_step": 97790, "epoch": 1098, "lr": 7.142053913857603e-05}
{"train_loss": 0.14387722313404083, "global_step": 97791, "epoch": 1098, "lr": 7.142001529636707e-05}
{"train_loss": 0.14014002680778503, "global_step": 97792, "epoch": 1098, "lr": 7.141949145127845e-05}
{"train_loss": 0.2523125112056732, "global_step": 97793, "epoch": 1098, "lr": 7.141896760331026e-05}
{"train_loss": 0.1768849641084671, "global_step": 97794, "epoch": 1098, "lr": 7.141844375246254e-05}
{"train_loss": 0.23973235487937927, "global_step": 97795, "epoch": 1098, "lr": 7.141791989873536e-05}
{"train_loss": 0.12467679381370544, "global_step": 97796, "epoch": 1098, "lr": 7.141739604212882e-05}
{"train_loss": 0.1170920729637146, "global_step": 97797, "epoch": 1098, "lr": 7.141687218264296e-05}
{"train_loss": 0.12954674661159515, "global_step": 97798, "epoch": 1098, "lr": 7.141634832027786e-05}
{"train_loss": 0.19390363991260529, "global_step": 97799, "epoch": 1098, "lr": 7.141582445503361e-05}
{"train_loss": 0.187889963388443, "global_step": 97800, "epoch": 1098, "lr": 7.141530058691025e-05}
{"train_loss": 0.2107664942741394, "global_step": 97801, "epoch": 1098, "lr": 7.141477671590786e-05}
{"train_loss": 0.16929508745670319, "global_step": 97802, "epoch": 1098, "lr": 7.141425284202652e-05}
{"train_loss": 0.14471137523651123, "global_step": 97803, "epoch": 1098, "lr": 7.14137289652663e-05}
{"train_loss": 0.17198555171489716, "global_step": 97804, "epoch": 1098, "lr": 7.141320508562726e-05}
{"train_loss": 0.1250157207250595, "global_step": 97805, "epoch": 1098, "lr": 7.141268120310948e-05}
{"train_loss": 0.13216012716293335, "global_step": 97806, "epoch": 1098, "lr": 7.141215731771302e-05}
{"train_loss": 0.21222034096717834, "global_step": 97807, "epoch": 1098, "lr": 7.141163342943795e-05}
{"train_loss": 0.12074990570545197, "global_step": 97808, "epoch": 1098, "lr": 7.141110953828437e-05}
{"train_loss": 0.25185397267341614, "global_step": 97809, "epoch": 1098, "lr": 7.14105856442523e-05}
{"train_loss": 0.16657855453785886, "global_step": 97810, "epoch": 1098, "lr": 7.141006174734184e-05, "val_loss": 4.578671932220459}
{"train_loss": 0.22344288229942322, "global_step": 97811, "epoch": 1099, "lr": 7.140953784755307e-05}
{"train_loss": 0.2416703999042511, "global_step": 97812, "epoch": 1099, "lr": 7.140901394488607e-05}
{"train_loss": 0.10822094231843948, "global_step": 97813, "epoch": 1099, "lr": 7.140849003934084e-05}
{"train_loss": 0.17995038628578186, "global_step": 97814, "epoch": 1099, "lr": 7.140796613091754e-05}
{"train_loss": 0.2004869431257248, "global_step": 97815, "epoch": 1099, "lr": 7.140744221961618e-05}
{"train_loss": 0.12246184051036835, "global_step": 97816, "epoch": 1099, "lr": 7.140691830543686e-05}
{"train_loss": 0.13847403228282928, "global_step": 97817, "epoch": 1099, "lr": 7.140639438837963e-05}
{"train_loss": 0.16943885385990143, "global_step": 97818, "epoch": 1099, "lr": 7.140587046844458e-05}
{"train_loss": 0.13972169160842896, "global_step": 97819, "epoch": 1099, "lr": 7.140534654563176e-05}
{"train_loss": 0.18004946410655975, "global_step": 97820, "epoch": 1099, "lr": 7.140482261994126e-05}
{"train_loss": 0.1680351346731186, "global_step": 97821, "epoch": 1099, "lr": 7.140429869137314e-05}
{"train_loss": 0.18903587758541107, "global_step": 97822, "epoch": 1099, "lr": 7.140377475992749e-05}
{"train_loss": 0.17575334012508392, "global_step": 97823, "epoch": 1099, "lr": 7.140325082560435e-05}
{"train_loss": 0.17239157855510712, "global_step": 97824, "epoch": 1099, "lr": 7.14027268884038e-05}
{"train_loss": 0.17043115198612213, "global_step": 97825, "epoch": 1099, "lr": 7.140220294832593e-05}
{"train_loss": 0.1552165448665619, "global_step": 97826, "epoch": 1099, "lr": 7.140167900537079e-05}
{"train_loss": 0.1484357863664627, "global_step": 97827, "epoch": 1099, "lr": 7.140115505953844e-05}
{"train_loss": 0.1978420466184616, "global_step": 97828, "epoch": 1099, "lr": 7.140063111082895e-05}
{"train_loss": 0.05990482121706009, "global_step": 97829, "epoch": 1099, "lr": 7.140010715924243e-05}
{"train_loss": 0.19926254451274872, "global_step": 97830, "epoch": 1099, "lr": 7.139958320477893e-05}
{"train_loss": 0.2222222089767456, "global_step": 97831, "epoch": 1099, "lr": 7.139905924743852e-05}
{"train_loss": 0.21259242296218872, "global_step": 97832, "epoch": 1099, "lr": 7.139853528722127e-05}
{"train_loss": 0.12681977450847626, "global_step": 97833, "epoch": 1099, "lr": 7.139801132412722e-05}
{"train_loss": 0.17850038409233093, "global_step": 97834, "epoch": 1099, "lr": 7.13974873581565e-05}
{"train_loss": 0.22232381999492645, "global_step": 97835, "epoch": 1099, "lr": 7.139696338930913e-05}
{"train_loss": 0.11017728596925735, "global_step": 97836, "epoch": 1099, "lr": 7.139643941758521e-05}
{"train_loss": 0.13569198548793793, "global_step": 97837, "epoch": 1099, "lr": 7.139591544298478e-05}
{"train_loss": 0.23621004819869995, "global_step": 97838, "epoch": 1099, "lr": 7.139539146550795e-05}
{"train_loss": 0.15871724486351013, "global_step": 97839, "epoch": 1099, "lr": 7.139486748515476e-05}
{"train_loss": 0.12835627794265747, "global_step": 97840, "epoch": 1099, "lr": 7.13943435019253e-05}
{"train_loss": 0.19874051213264465, "global_step": 97841, "epoch": 1099, "lr": 7.139381951581963e-05}
{"train_loss": 0.1359391212463379, "global_step": 97842, "epoch": 1099, "lr": 7.139329552683781e-05}
{"train_loss": 0.1718616485595703, "global_step": 97843, "epoch": 1099, "lr": 7.139277153497993e-05}
{"train_loss": 0.23393258452415466, "global_step": 97844, "epoch": 1099, "lr": 7.139224754024604e-05}
{"train_loss": 0.164958193898201, "global_step": 97845, "epoch": 1099, "lr": 7.139172354263624e-05}
{"train_loss": 0.17757706344127655, "global_step": 97846, "epoch": 1099, "lr": 7.139119954215059e-05}
{"train_loss": 0.16446606814861298, "global_step": 97847, "epoch": 1099, "lr": 7.139067553878913e-05}
{"train_loss": 0.17678740620613098, "global_step": 97848, "epoch": 1099, "lr": 7.139015153255196e-05}
{"train_loss": 0.13370104134082794, "global_step": 97849, "epoch": 1099, "lr": 7.138962752343915e-05}
{"train_loss": 0.16430023312568665, "global_step": 97850, "epoch": 1099, "lr": 7.138910351145077e-05}
{"train_loss": 0.21089893579483032, "global_step": 97851, "epoch": 1099, "lr": 7.13885794965869e-05}
{"train_loss": 0.17710161209106445, "global_step": 97852, "epoch": 1099, "lr": 7.138805547884756e-05}
{"train_loss": 0.2553168833255768, "global_step": 97853, "epoch": 1099, "lr": 7.138753145823289e-05}
{"train_loss": 0.12871618568897247, "global_step": 97854, "epoch": 1099, "lr": 7.138700743474291e-05}
{"train_loss": 0.19227102398872375, "global_step": 97855, "epoch": 1099, "lr": 7.13864834083777e-05}
{"train_loss": 0.16815553605556488, "global_step": 97856, "epoch": 1099, "lr": 7.138595937913735e-05}
{"train_loss": 0.15279535949230194, "global_step": 97857, "epoch": 1099, "lr": 7.138543534702192e-05}
{"train_loss": 0.14021369814872742, "global_step": 97858, "epoch": 1099, "lr": 7.138491131203147e-05}
{"train_loss": 0.15432600677013397, "global_step": 97859, "epoch": 1099, "lr": 7.13843872741661e-05}
{"train_loss": 0.16870243847370148, "global_step": 97860, "epoch": 1099, "lr": 7.138386323342584e-05}
{"train_loss": 0.16899824142456055, "global_step": 97861, "epoch": 1099, "lr": 7.138333918981078e-05}
{"train_loss": 0.20793992280960083, "global_step": 97862, "epoch": 1099, "lr": 7.1382815143321e-05}
{"train_loss": 0.22961917519569397, "global_step": 97863, "epoch": 1099, "lr": 7.138229109395656e-05}
{"train_loss": 0.10270765423774719, "global_step": 97864, "epoch": 1099, "lr": 7.138176704171752e-05}
{"train_loss": 0.20917057991027832, "global_step": 97865, "epoch": 1099, "lr": 7.138124298660395e-05}
{"train_loss": 0.12071047723293304, "global_step": 97866, "epoch": 1099, "lr": 7.138071892861596e-05}
{"train_loss": 0.20189984142780304, "global_step": 97867, "epoch": 1099, "lr": 7.138019486775359e-05}
{"train_loss": 0.15871356427669525, "global_step": 97868, "epoch": 1099, "lr": 7.13796708040169e-05}
{"train_loss": 0.1787499636411667, "global_step": 97869, "epoch": 1099, "lr": 7.1379146737406e-05}
{"train_loss": 0.10469278693199158, "global_step": 97870, "epoch": 1099, "lr": 7.137862266792092e-05}
{"train_loss": 0.24269038438796997, "global_step": 97871, "epoch": 1099, "lr": 7.137809859556173e-05}
{"train_loss": 0.2268211990594864, "global_step": 97872, "epoch": 1099, "lr": 7.137757452032853e-05}
{"train_loss": 0.16287711262702942, "global_step": 97873, "epoch": 1099, "lr": 7.137705044222138e-05}
{"train_loss": 0.15470507740974426, "global_step": 97874, "epoch": 1099, "lr": 7.137652636124034e-05}
{"train_loss": 0.1892848014831543, "global_step": 97875, "epoch": 1099, "lr": 7.137600227738548e-05}
{"train_loss": 0.15172430872917175, "global_step": 97876, "epoch": 1099, "lr": 7.137547819065689e-05}
{"train_loss": 0.2048308104276657, "global_step": 97877, "epoch": 1099, "lr": 7.137495410105462e-05}
{"train_loss": 0.15315081179141998, "global_step": 97878, "epoch": 1099, "lr": 7.137443000857874e-05}
{"train_loss": 0.1488817036151886, "global_step": 97879, "epoch": 1099, "lr": 7.137390591322933e-05}
{"train_loss": 0.12543921172618866, "global_step": 97880, "epoch": 1099, "lr": 7.137338181500647e-05}
{"train_loss": 0.18273523449897766, "global_step": 97881, "epoch": 1099, "lr": 7.137285771391021e-05}
{"train_loss": 0.1245245635509491, "global_step": 97882, "epoch": 1099, "lr": 7.137233360994065e-05}
{"train_loss": 0.13377758860588074, "global_step": 97883, "epoch": 1099, "lr": 7.13718095030978e-05}
{"train_loss": 0.2088453620672226, "global_step": 97884, "epoch": 1099, "lr": 7.13712853933818e-05}
{"train_loss": 0.20389889180660248, "global_step": 97885, "epoch": 1099, "lr": 7.137076128079268e-05}
{"train_loss": 0.1462850719690323, "global_step": 97886, "epoch": 1099, "lr": 7.137023716533054e-05}
{"train_loss": 0.22293436527252197, "global_step": 97887, "epoch": 1099, "lr": 7.136971304699542e-05}
{"train_loss": 0.09022815525531769, "global_step": 97888, "epoch": 1099, "lr": 7.13691889257874e-05}
{"train_loss": 0.20503844320774078, "global_step": 97889, "epoch": 1099, "lr": 7.136866480170656e-05}
{"train_loss": 0.13947395980358124, "global_step": 97890, "epoch": 1099, "lr": 7.136814067475296e-05}
{"train_loss": 0.13231819868087769, "global_step": 97891, "epoch": 1099, "lr": 7.136761654492667e-05}
{"train_loss": 0.22476324439048767, "global_step": 97892, "epoch": 1099, "lr": 7.136709241222777e-05}
{"train_loss": 0.19015246629714966, "global_step": 97893, "epoch": 1099, "lr": 7.136656827665632e-05}
{"train_loss": 0.17529653012752533, "global_step": 97894, "epoch": 1099, "lr": 7.13660441382124e-05}
{"train_loss": 0.15354907512664795, "global_step": 97895, "epoch": 1099, "lr": 7.136551999689607e-05}
{"train_loss": 0.15862619876861572, "global_step": 97896, "epoch": 1099, "lr": 7.13649958527074e-05}
{"train_loss": 0.14762656390666962, "global_step": 97897, "epoch": 1099, "lr": 7.136447170564648e-05}
{"train_loss": 0.12366233021020889, "global_step": 97898, "epoch": 1099, "lr": 7.136394755571336e-05}
{"train_loss": 0.16923263953642897, "global_step": 97899, "epoch": 1099, "lr": 7.136342340290814e-05, "val_loss": 4.675788402557373}
{"train_loss": 0.18475879728794098, "global_step": 97900, "epoch": 1100, "lr": 7.136289924723084e-05}
{"train_loss": 0.12061094492673874, "global_step": 97901, "epoch": 1100, "lr": 7.136237508868155e-05}
{"train_loss": 0.20034341514110565, "global_step": 97902, "epoch": 1100, "lr": 7.136185092726037e-05}
{"train_loss": 0.10408927500247955, "global_step": 97903, "epoch": 1100, "lr": 7.136132676296735e-05}
{"train_loss": 0.19740474224090576, "global_step": 97904, "epoch": 1100, "lr": 7.136080259580256e-05}
{"train_loss": 0.20518726110458374, "global_step": 97905, "epoch": 1100, "lr": 7.136027842576605e-05}
{"train_loss": 0.23069830238819122, "global_step": 97906, "epoch": 1100, "lr": 7.135975425285793e-05}
{"train_loss": 0.16553911566734314, "global_step": 97907, "epoch": 1100, "lr": 7.135923007707826e-05}
{"train_loss": 0.21509914100170135, "global_step": 97908, "epoch": 1100, "lr": 7.135870589842709e-05}
{"train_loss": 0.1850493848323822, "global_step": 97909, "epoch": 1100, "lr": 7.135818171690452e-05}
{"train_loss": 0.15161395072937012, "global_step": 97910, "epoch": 1100, "lr": 7.135765753251058e-05}
{"train_loss": 0.19275610148906708, "global_step": 97911, "epoch": 1100, "lr": 7.135713334524537e-05}
{"train_loss": 0.148768350481987, "global_step": 97912, "epoch": 1100, "lr": 7.135660915510895e-05}
{"train_loss": 0.13332660496234894, "global_step": 97913, "epoch": 1100, "lr": 7.13560849621014e-05}
{"train_loss": 0.20430181920528412, "global_step": 97914, "epoch": 1100, "lr": 7.135556076622277e-05}
{"train_loss": 0.14255008101463318, "global_step": 97915, "epoch": 1100, "lr": 7.135503656747317e-05}
{"train_loss": 0.11710765212774277, "global_step": 97916, "epoch": 1100, "lr": 7.135451236585262e-05}
{"train_loss": 0.14486363530158997, "global_step": 97917, "epoch": 1100, "lr": 7.135398816136123e-05}
{"train_loss": 0.21319565176963806, "global_step": 97918, "epoch": 1100, "lr": 7.135346395399906e-05}
{"train_loss": 0.19318482279777527, "global_step": 97919, "epoch": 1100, "lr": 7.135293974376615e-05}
{"train_loss": 0.19551120698451996, "global_step": 97920, "epoch": 1100, "lr": 7.135241553066263e-05}
{"train_loss": 0.20429611206054688, "global_step": 97921, "epoch": 1100, "lr": 7.135189131468853e-05}
{"train_loss": 0.12771892547607422, "global_step": 97922, "epoch": 1100, "lr": 7.135136709584392e-05}
{"train_loss": 0.14668509364128113, "global_step": 97923, "epoch": 1100, "lr": 7.135084287412889e-05}
{"train_loss": 0.184050053358078, "global_step": 97924, "epoch": 1100, "lr": 7.135031864954349e-05}
{"train_loss": 0.19311445951461792, "global_step": 97925, "epoch": 1100, "lr": 7.13497944220878e-05}
{"train_loss": 0.14929941296577454, "global_step": 97926, "epoch": 1100, "lr": 7.134927019176189e-05}
{"train_loss": 0.1759389191865921, "global_step": 97927, "epoch": 1100, "lr": 7.134874595856584e-05}
{"train_loss": 0.17134220898151398, "global_step": 97928, "epoch": 1100, "lr": 7.134822172249973e-05}
{"train_loss": 0.17130053043365479, "global_step": 97929, "epoch": 1100, "lr": 7.134769748356357e-05}
{"train_loss": 0.2635042369365692, "global_step": 97930, "epoch": 1100, "lr": 7.134717324175751e-05}
{"train_loss": 0.1453712433576584, "global_step": 97931, "epoch": 1100, "lr": 7.134664899708156e-05}
{"train_loss": 0.16717161238193512, "global_step": 97932, "epoch": 1100, "lr": 7.134612474953582e-05}
{"train_loss": 0.19944895803928375, "global_step": 97933, "epoch": 1100, "lr": 7.134560049912035e-05}
{"train_loss": 0.2023421674966812, "global_step": 97934, "epoch": 1100, "lr": 7.134507624583522e-05}
{"train_loss": 0.14831720292568207, "global_step": 97935, "epoch": 1100, "lr": 7.134455198968051e-05}
{"train_loss": 0.20110398530960083, "global_step": 97936, "epoch": 1100, "lr": 7.134402773065629e-05}
{"train_loss": 0.1806419938802719, "global_step": 97937, "epoch": 1100, "lr": 7.13435034687626e-05}
{"train_loss": 0.15857496857643127, "global_step": 97938, "epoch": 1100, "lr": 7.134297920399956e-05}
{"train_loss": 0.15978199243545532, "global_step": 97939, "epoch": 1100, "lr": 7.13424549363672e-05}
{"train_loss": 0.18232862651348114, "global_step": 97940, "epoch": 1100, "lr": 7.134193066586563e-05}
{"train_loss": 0.1590350866317749, "global_step": 97941, "epoch": 1100, "lr": 7.134140639249489e-05}
{"train_loss": 0.1736469864845276, "global_step": 97942, "epoch": 1100, "lr": 7.134088211625505e-05}
{"train_loss": 0.13270914554595947, "global_step": 97943, "epoch": 1100, "lr": 7.13403578371462e-05}
{"train_loss": 0.18535497784614563, "global_step": 97944, "epoch": 1100, "lr": 7.133983355516839e-05}
{"train_loss": 0.18826666474342346, "global_step": 97945, "epoch": 1100, "lr": 7.13393092703217e-05}
{"train_loss": 0.1943957805633545, "global_step": 97946, "epoch": 1100, "lr": 7.133878498260619e-05}
{"train_loss": 0.2372591346502304, "global_step": 97947, "epoch": 1100, "lr": 7.133826069202197e-05}
{"train_loss": 0.1286928504705429, "global_step": 97948, "epoch": 1100, "lr": 7.133773639856906e-05}
{"train_loss": 0.1737322360277176, "global_step": 97949, "epoch": 1100, "lr": 7.133721210224756e-05}
{"train_loss": 0.25562137365341187, "global_step": 97950, "epoch": 1100, "lr": 7.133668780305753e-05}
{"train_loss": 0.13319671154022217, "global_step": 97951, "epoch": 1100, "lr": 7.133616350099904e-05}
{"train_loss": 0.1945566087961197, "global_step": 97952, "epoch": 1100, "lr": 7.133563919607215e-05}
{"train_loss": 0.15803690254688263, "global_step": 97953, "epoch": 1100, "lr": 7.133511488827697e-05}
{"train_loss": 0.15349051356315613, "global_step": 97954, "epoch": 1100, "lr": 7.133459057761352e-05}
{"train_loss": 0.13562753796577454, "global_step": 97955, "epoch": 1100, "lr": 7.133406626408191e-05}
{"train_loss": 0.10458969324827194, "global_step": 97956, "epoch": 1100, "lr": 7.133354194768218e-05}
{"train_loss": 0.17913788557052612, "global_step": 97957, "epoch": 1100, "lr": 7.133301762841442e-05}
{"train_loss": 0.12861773371696472, "global_step": 97958, "epoch": 1100, "lr": 7.133249330627871e-05}
{"train_loss": 0.20676438510417938, "global_step": 97959, "epoch": 1100, "lr": 7.133196898127509e-05}
{"train_loss": 0.25228819251060486, "global_step": 97960, "epoch": 1100, "lr": 7.133144465340365e-05}
{"train_loss": 0.16016785800457, "global_step": 97961, "epoch": 1100, "lr": 7.133092032266447e-05}
{"train_loss": 0.12348298728466034, "global_step": 97962, "epoch": 1100, "lr": 7.13303959890576e-05}
{"train_loss": 0.09167499095201492, "global_step": 97963, "epoch": 1100, "lr": 7.13298716525831e-05}
{"train_loss": 0.13850383460521698, "global_step": 97964, "epoch": 1100, "lr": 7.132934731324108e-05}
{"train_loss": 0.13509440422058105, "global_step": 97965, "epoch": 1100, "lr": 7.132882297103158e-05}
{"train_loss": 0.11255625635385513, "global_step": 97966, "epoch": 1100, "lr": 7.13282986259547e-05}
{"train_loss": 0.11407903581857681, "global_step": 97967, "epoch": 1100, "lr": 7.132777427801047e-05}
{"train_loss": 0.16293835639953613, "global_step": 97968, "epoch": 1100, "lr": 7.132724992719899e-05}
{"train_loss": 0.13108915090560913, "global_step": 97969, "epoch": 1100, "lr": 7.132672557352032e-05}
{"train_loss": 0.2639552354812622, "global_step": 97970, "epoch": 1100, "lr": 7.132620121697452e-05}
{"train_loss": 0.1603185087442398, "global_step": 97971, "epoch": 1100, "lr": 7.132567685756168e-05}
{"train_loss": 0.13630464673042297, "global_step": 97972, "epoch": 1100, "lr": 7.132515249528186e-05}
{"train_loss": 0.14797081053256989, "global_step": 97973, "epoch": 1100, "lr": 7.132462813013514e-05}
{"train_loss": 0.11629485338926315, "global_step": 97974, "epoch": 1100, "lr": 7.132410376212156e-05}
{"train_loss": 0.16844956576824188, "global_step": 97975, "epoch": 1100, "lr": 7.132357939124124e-05}
{"train_loss": 0.12115630507469177, "global_step": 97976, "epoch": 1100, "lr": 7.132305501749422e-05}
{"train_loss": 0.16188578307628632, "global_step": 97977, "epoch": 1100, "lr": 7.132253064088057e-05}
{"train_loss": 0.12608583271503448, "global_step": 97978, "epoch": 1100, "lr": 7.132200626140037e-05}
{"train_loss": 0.14732733368873596, "global_step": 97979, "epoch": 1100, "lr": 7.132148187905369e-05}
{"train_loss": 0.12857745587825775, "global_step": 97980, "epoch": 1100, "lr": 7.132095749384058e-05}
{"train_loss": 0.1408122479915619, "global_step": 97981, "epoch": 1100, "lr": 7.132043310576114e-05}
{"train_loss": 0.18678244948387146, "global_step": 97982, "epoch": 1100, "lr": 7.131990871481543e-05}
{"train_loss": 0.24193613231182098, "global_step": 97983, "epoch": 1100, "lr": 7.131938432100351e-05}
{"train_loss": 0.21178758144378662, "global_step": 97984, "epoch": 1100, "lr": 7.131885992432547e-05}
{"train_loss": 0.20293109118938446, "global_step": 97985, "epoch": 1100, "lr": 7.131833552478135e-05}
{"train_loss": 0.15056678652763367, "global_step": 97986, "epoch": 1100, "lr": 7.131781112237126e-05}
{"train_loss": 0.1604977697134018, "global_step": 97987, "epoch": 1100, "lr": 7.131728671709525e-05}
{"train_loss": 0.16775642276814814, "global_step": 97988, "epoch": 1100, "lr": 7.13167623089534e-05, "train/sim_max_reward_0": 0.6942257418096723, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.6500628820583921, "train/sim_max_reward_3": 0.42084646632872735, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.7042322558442891, "test/sim_max_reward_4500000": 0.44584162295418495, "test/sim_max_reward_4500001": 0.04937632018482877, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.4927746489450131, "test/sim_max_reward_4500004": 0.5748406674153251, "test/sim_max_reward_4500005": 0.9378022907724646, "test/sim_max_reward_4500006": 0.04317389840462805, "test/sim_max_reward_4500007": 0.8521828275854627, "test/sim_max_reward_4500008": 0.5495707765868054, "test/sim_max_reward_4500009": 0.04964428152118653, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.8975547029550055, "test/sim_max_reward_4500013": 0.7097353085229076, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.9178956273619762, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 0.9997286241274266, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.5047857413034531, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9933431675965881, "test/sim_max_reward_4500026": 0.9913408224913046, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.40631352721417346, "test/sim_max_reward_4500029": 0.9653107857344786, "test/sim_max_reward_4500030": 0.982643450439468, "test/sim_max_reward_4500031": 0.408268559758505, "test/sim_max_reward_4500032": 0.9896350925440427, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.7604187565907156, "test/sim_max_reward_4500035": 0.5793176290018985, "test/sim_max_reward_4500036": 0.30617690206643855, "test/sim_max_reward_4500037": 0.5864073928839066, "test/sim_max_reward_4500038": 0.9809951073202648, "test/sim_max_reward_4500039": 0.7585510185269454, "test/sim_max_reward_4500040": 0.24247050786222002, "test/sim_max_reward_4500041": 0.20558790770902613, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9821076909411464, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.6979855308274364, "test/sim_max_reward_4500046": 1.0, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9850282021870232, "test/sim_max_reward_4500049": 0.6694087915332477, "train/mean_score": 0.7448945576735134, "test/mean_score": 0.61032436363739, "val_loss": 4.766086578369141, "train_action_mse_error": 27.26731300354004}
{"train_loss": 0.12782777845859528, "global_step": 97989, "epoch": 1101, "lr": 7.131623789794573e-05}
{"train_loss": 0.11429709941148758, "global_step": 97990, "epoch": 1101, "lr": 7.131571348407238e-05}
{"train_loss": 0.1646677851676941, "global_step": 97991, "epoch": 1101, "lr": 7.13151890673334e-05}
{"train_loss": 0.1627846360206604, "global_step": 97992, "epoch": 1101, "lr": 7.131466464772882e-05}
{"train_loss": 0.13005408644676208, "global_step": 97993, "epoch": 1101, "lr": 7.131414022525876e-05}
{"train_loss": 0.13648200035095215, "global_step": 97994, "epoch": 1101, "lr": 7.131361579992329e-05}
{"train_loss": 0.178644061088562, "global_step": 97995, "epoch": 1101, "lr": 7.131309137172244e-05}
{"train_loss": 0.1048155203461647, "global_step": 97996, "epoch": 1101, "lr": 7.131256694065631e-05}
{"train_loss": 0.13187825679779053, "global_step": 97997, "epoch": 1101, "lr": 7.131204250672497e-05}
{"train_loss": 0.21810363233089447, "global_step": 97998, "epoch": 1101, "lr": 7.131151806992849e-05}
{"train_loss": 0.16081319749355316, "global_step": 97999, "epoch": 1101, "lr": 7.131099363026693e-05}
{"train_loss": 0.19700302183628082, "global_step": 98000, "epoch": 1101, "lr": 7.131046918774037e-05}
{"train_loss": 0.2145623117685318, "global_step": 98001, "epoch": 1101, "lr": 7.130994474234888e-05}
{"train_loss": 0.20144718885421753, "global_step": 98002, "epoch": 1101, "lr": 7.130942029409252e-05}
{"train_loss": 0.17493997514247894, "global_step": 98003, "epoch": 1101, "lr": 7.130889584297138e-05}
{"train_loss": 0.20571669936180115, "global_step": 98004, "epoch": 1101, "lr": 7.130837138898551e-05}
{"train_loss": 0.15872494876384735, "global_step": 98005, "epoch": 1101, "lr": 7.130784693213499e-05}
{"train_loss": 0.10980100184679031, "global_step": 98006, "epoch": 1101, "lr": 7.130732247241989e-05}
{"train_loss": 0.1565539687871933, "global_step": 98007, "epoch": 1101, "lr": 7.130679800984028e-05}
{"train_loss": 0.17771123349666595, "global_step": 98008, "epoch": 1101, "lr": 7.130627354439623e-05}
{"train_loss": 0.17876243591308594, "global_step": 98009, "epoch": 1101, "lr": 7.130574907608782e-05}
{"train_loss": 0.18949130177497864, "global_step": 98010, "epoch": 1101, "lr": 7.13052246049151e-05}
{"train_loss": 0.09354506433010101, "global_step": 98011, "epoch": 1101, "lr": 7.130470013087814e-05}
{"train_loss": 0.13794930279254913, "global_step": 98012, "epoch": 1101, "lr": 7.130417565397705e-05}
{"train_loss": 0.1959320604801178, "global_step": 98013, "epoch": 1101, "lr": 7.130365117421185e-05}
{"train_loss": 0.255058228969574, "global_step": 98014, "epoch": 1101, "lr": 7.130312669158266e-05}
{"train_loss": 0.12680603563785553, "global_step": 98015, "epoch": 1101, "lr": 7.13026022060895e-05}
{"train_loss": 0.23333273828029633, "global_step": 98016, "epoch": 1101, "lr": 7.130207771773248e-05}
{"train_loss": 0.14473438262939453, "global_step": 98017, "epoch": 1101, "lr": 7.130155322651164e-05}
{"train_loss": 0.1891053169965744, "global_step": 98018, "epoch": 1101, "lr": 7.130102873242708e-05}
{"train_loss": 0.11705601215362549, "global_step": 98019, "epoch": 1101, "lr": 7.130050423547885e-05}
{"train_loss": 0.138499915599823, "global_step": 98020, "epoch": 1101, "lr": 7.129997973566703e-05}
{"train_loss": 0.16176347434520721, "global_step": 98021, "epoch": 1101, "lr": 7.129945523299168e-05}
{"train_loss": 0.08911480009555817, "global_step": 98022, "epoch": 1101, "lr": 7.129893072745288e-05}
{"train_loss": 0.1821780651807785, "global_step": 98023, "epoch": 1101, "lr": 7.12984062190507e-05}
{"train_loss": 0.3080804646015167, "global_step": 98024, "epoch": 1101, "lr": 7.129788170778523e-05}
{"train_loss": 0.1894010305404663, "global_step": 98025, "epoch": 1101, "lr": 7.12973571936565e-05}
{"train_loss": 0.15891428291797638, "global_step": 98026, "epoch": 1101, "lr": 7.12968326766646e-05}
{"train_loss": 0.16557270288467407, "global_step": 98027, "epoch": 1101, "lr": 7.12963081568096e-05}
{"train_loss": 0.22181011736392975, "global_step": 98028, "epoch": 1101, "lr": 7.129578363409156e-05}
{"train_loss": 0.1802644580602646, "global_step": 98029, "epoch": 1101, "lr": 7.129525910851058e-05}
{"train_loss": 0.19463685154914856, "global_step": 98030, "epoch": 1101, "lr": 7.129473458006671e-05}
{"train_loss": 0.21716266870498657, "global_step": 98031, "epoch": 1101, "lr": 7.129421004876001e-05}
{"train_loss": 0.15720638632774353, "global_step": 98032, "epoch": 1101, "lr": 7.129368551459056e-05}
{"train_loss": 0.14442475140094757, "global_step": 98033, "epoch": 1101, "lr": 7.129316097755846e-05}
{"train_loss": 0.1661050170660019, "global_step": 98034, "epoch": 1101, "lr": 7.129263643766373e-05}
{"train_loss": 0.2108379602432251, "global_step": 98035, "epoch": 1101, "lr": 7.129211189490648e-05}
{"train_loss": 0.09408707171678543, "global_step": 98036, "epoch": 1101, "lr": 7.129158734928675e-05}
{"train_loss": 0.18179339170455933, "global_step": 98037, "epoch": 1101, "lr": 7.129106280080464e-05}
{"train_loss": 0.1704910397529602, "global_step": 98038, "epoch": 1101, "lr": 7.12905382494602e-05}
{"train_loss": 0.17323556542396545, "global_step": 98039, "epoch": 1101, "lr": 7.12900136952535e-05}
{"train_loss": 0.18633730709552765, "global_step": 98040, "epoch": 1101, "lr": 7.128948913818463e-05}
{"train_loss": 0.13547664880752563, "global_step": 98041, "epoch": 1101, "lr": 7.128896457825364e-05}
{"train_loss": 0.1464250385761261, "global_step": 98042, "epoch": 1101, "lr": 7.12884400154606e-05}
{"train_loss": 0.19962067902088165, "global_step": 98043, "epoch": 1101, "lr": 7.12879154498056e-05}
{"train_loss": 0.1824236810207367, "global_step": 98044, "epoch": 1101, "lr": 7.12873908812887e-05}
{"train_loss": 0.2561294138431549, "global_step": 98045, "epoch": 1101, "lr": 7.128686630990997e-05}
{"train_loss": 0.24017179012298584, "global_step": 98046, "epoch": 1101, "lr": 7.128634173566947e-05}
{"train_loss": 0.18123292922973633, "global_step": 98047, "epoch": 1101, "lr": 7.128581715856729e-05}
{"train_loss": 0.23444928228855133, "global_step": 98048, "epoch": 1101, "lr": 7.12852925786035e-05}
{"train_loss": 0.15925230085849762, "global_step": 98049, "epoch": 1101, "lr": 7.128476799577814e-05}
{"train_loss": 0.13298748433589935, "global_step": 98050, "epoch": 1101, "lr": 7.128424341009131e-05}
{"train_loss": 0.22924456000328064, "global_step": 98051, "epoch": 1101, "lr": 7.128371882154308e-05}
{"train_loss": 0.1667260378599167, "global_step": 98052, "epoch": 1101, "lr": 7.128319423013351e-05}
{"train_loss": 0.18957854807376862, "global_step": 98053, "epoch": 1101, "lr": 7.128266963586267e-05}
{"train_loss": 0.1284104585647583, "global_step": 98054, "epoch": 1101, "lr": 7.128214503873065e-05}
{"train_loss": 0.16048653423786163, "global_step": 98055, "epoch": 1101, "lr": 7.128162043873748e-05}
{"train_loss": 0.12830503284931183, "global_step": 98056, "epoch": 1101, "lr": 7.128109583588327e-05}
{"train_loss": 0.1253785789012909, "global_step": 98057, "epoch": 1101, "lr": 7.128057123016807e-05}
{"train_loss": 0.19076307117938995, "global_step": 98058, "epoch": 1101, "lr": 7.128004662159197e-05}
{"train_loss": 0.1583126187324524, "global_step": 98059, "epoch": 1101, "lr": 7.127952201015502e-05}
{"train_loss": 0.11997342854738235, "global_step": 98060, "epoch": 1101, "lr": 7.127899739585729e-05}
{"train_loss": 0.09149838984012604, "global_step": 98061, "epoch": 1101, "lr": 7.127847277869887e-05}
{"train_loss": 0.11260731518268585, "global_step": 98062, "epoch": 1101, "lr": 7.127794815867981e-05}
{"train_loss": 0.18224117159843445, "global_step": 98063, "epoch": 1101, "lr": 7.12774235358002e-05}
{"train_loss": 0.12806876003742218, "global_step": 98064, "epoch": 1101, "lr": 7.127689891006009e-05}
{"train_loss": 0.1568504273891449, "global_step": 98065, "epoch": 1101, "lr": 7.127637428145956e-05}
{"train_loss": 0.21299217641353607, "global_step": 98066, "epoch": 1101, "lr": 7.127584964999867e-05}
{"train_loss": 0.18468429148197174, "global_step": 98067, "epoch": 1101, "lr": 7.127532501567752e-05}
{"train_loss": 0.11383367329835892, "global_step": 98068, "epoch": 1101, "lr": 7.127480037849615e-05}
{"train_loss": 0.11097317188978195, "global_step": 98069, "epoch": 1101, "lr": 7.127427573845465e-05}
{"train_loss": 0.21387135982513428, "global_step": 98070, "epoch": 1101, "lr": 7.127375109555308e-05}
{"train_loss": 0.1819797307252884, "global_step": 98071, "epoch": 1101, "lr": 7.127322644979152e-05}
{"train_loss": 0.151405930519104, "global_step": 98072, "epoch": 1101, "lr": 7.127270180117003e-05}
{"train_loss": 0.17747803032398224, "global_step": 98073, "epoch": 1101, "lr": 7.127217714968868e-05}
{"train_loss": 0.1377706229686737, "global_step": 98074, "epoch": 1101, "lr": 7.127165249534755e-05}
{"train_loss": 0.16473881900310516, "global_step": 98075, "epoch": 1101, "lr": 7.127112783814671e-05}
{"train_loss": 0.13025657832622528, "global_step": 98076, "epoch": 1101, "lr": 7.127060317808621e-05}
{"train_loss": 0.1668612914306394, "global_step": 98077, "epoch": 1101, "lr": 7.127007851516614e-05, "val_loss": 4.673018455505371}
{"train_loss": 0.20728547871112823, "global_step": 98078, "epoch": 1102, "lr": 7.126955384938659e-05}
{"train_loss": 0.22479388117790222, "global_step": 98079, "epoch": 1102, "lr": 7.126902918074758e-05}
{"train_loss": 0.1678885817527771, "global_step": 98080, "epoch": 1102, "lr": 7.126850450924922e-05}
{"train_loss": 0.12503328919410706, "global_step": 98081, "epoch": 1102, "lr": 7.126797983489158e-05}
{"train_loss": 0.1897774338722229, "global_step": 98082, "epoch": 1102, "lr": 7.12674551576747e-05}
{"train_loss": 0.11260468512773514, "global_step": 98083, "epoch": 1102, "lr": 7.126693047759868e-05}
{"train_loss": 0.1275612711906433, "global_step": 98084, "epoch": 1102, "lr": 7.126640579466358e-05}
{"train_loss": 0.15023410320281982, "global_step": 98085, "epoch": 1102, "lr": 7.126588110886948e-05}
{"train_loss": 0.2428838312625885, "global_step": 98086, "epoch": 1102, "lr": 7.126535642021642e-05}
{"train_loss": 0.18333956599235535, "global_step": 98087, "epoch": 1102, "lr": 7.12648317287045e-05}
{"train_loss": 0.2241518497467041, "global_step": 98088, "epoch": 1102, "lr": 7.126430703433379e-05}
{"train_loss": 0.14680081605911255, "global_step": 98089, "epoch": 1102, "lr": 7.126378233710435e-05}
{"train_loss": 0.19496867060661316, "global_step": 98090, "epoch": 1102, "lr": 7.126325763701624e-05}
{"train_loss": 0.10486756265163422, "global_step": 98091, "epoch": 1102, "lr": 7.126273293406957e-05}
{"train_loss": 0.18404042720794678, "global_step": 98092, "epoch": 1102, "lr": 7.126220822826437e-05}
{"train_loss": 0.18715481460094452, "global_step": 98093, "epoch": 1102, "lr": 7.126168351960072e-05}
{"train_loss": 0.13921795785427094, "global_step": 98094, "epoch": 1102, "lr": 7.126115880807871e-05}
{"train_loss": 0.19616949558258057, "global_step": 98095, "epoch": 1102, "lr": 7.126063409369838e-05}
{"train_loss": 0.10927919298410416, "global_step": 98096, "epoch": 1102, "lr": 7.126010937645983e-05}
{"train_loss": 0.19781197607517242, "global_step": 98097, "epoch": 1102, "lr": 7.125958465636311e-05}
{"train_loss": 0.1345806121826172, "global_step": 98098, "epoch": 1102, "lr": 7.12590599334083e-05}
{"train_loss": 0.15886567533016205, "global_step": 98099, "epoch": 1102, "lr": 7.125853520759546e-05}
{"train_loss": 0.2124360352754593, "global_step": 98100, "epoch": 1102, "lr": 7.125801047892468e-05}
{"train_loss": 0.1803753674030304, "global_step": 98101, "epoch": 1102, "lr": 7.125748574739603e-05}
{"train_loss": 0.15437734127044678, "global_step": 98102, "epoch": 1102, "lr": 7.125696101300955e-05}
{"train_loss": 0.1457093060016632, "global_step": 98103, "epoch": 1102, "lr": 7.125643627576533e-05}
{"train_loss": 0.13752481341362, "global_step": 98104, "epoch": 1102, "lr": 7.125591153566346e-05}
{"train_loss": 0.15863743424415588, "global_step": 98105, "epoch": 1102, "lr": 7.125538679270398e-05}
{"train_loss": 0.14948725700378418, "global_step": 98106, "epoch": 1102, "lr": 7.125486204688698e-05}
{"train_loss": 0.1031515970826149, "global_step": 98107, "epoch": 1102, "lr": 7.125433729821252e-05}
{"train_loss": 0.2296992689371109, "global_step": 98108, "epoch": 1102, "lr": 7.125381254668066e-05}
{"train_loss": 0.17601464688777924, "global_step": 98109, "epoch": 1102, "lr": 7.125328779229151e-05}
{"train_loss": 0.18084174394607544, "global_step": 98110, "epoch": 1102, "lr": 7.125276303504509e-05}
{"train_loss": 0.18579041957855225, "global_step": 98111, "epoch": 1102, "lr": 7.12522382749415e-05}
{"train_loss": 0.1777268946170807, "global_step": 98112, "epoch": 1102, "lr": 7.125171351198081e-05}
{"train_loss": 0.17915669083595276, "global_step": 98113, "epoch": 1102, "lr": 7.125118874616308e-05}
{"train_loss": 0.18994101881980896, "global_step": 98114, "epoch": 1102, "lr": 7.12506639774884e-05}
{"train_loss": 0.18739716708660126, "global_step": 98115, "epoch": 1102, "lr": 7.125013920595682e-05}
{"train_loss": 0.15440833568572998, "global_step": 98116, "epoch": 1102, "lr": 7.124961443156842e-05}
{"train_loss": 0.18030130863189697, "global_step": 98117, "epoch": 1102, "lr": 7.124908965432325e-05}
{"train_loss": 0.09387632459402084, "global_step": 98118, "epoch": 1102, "lr": 7.124856487422142e-05}
{"train_loss": 0.18685391545295715, "global_step": 98119, "epoch": 1102, "lr": 7.124804009126297e-05}
{"train_loss": 0.21178846061229706, "global_step": 98120, "epoch": 1102, "lr": 7.124751530544799e-05}
{"train_loss": 0.2066303789615631, "global_step": 98121, "epoch": 1102, "lr": 7.124699051677653e-05}
{"train_loss": 0.14807623624801636, "global_step": 98122, "epoch": 1102, "lr": 7.124646572524866e-05}
{"train_loss": 0.18163152039051056, "global_step": 98123, "epoch": 1102, "lr": 7.12459409308645e-05}
{"train_loss": 0.2056000828742981, "global_step": 98124, "epoch": 1102, "lr": 7.124541613362406e-05}
{"train_loss": 0.20524348318576813, "global_step": 98125, "epoch": 1102, "lr": 7.124489133352742e-05}
{"train_loss": 0.17578700184822083, "global_step": 98126, "epoch": 1102, "lr": 7.124436653057467e-05}
{"train_loss": 0.17576196789741516, "global_step": 98127, "epoch": 1102, "lr": 7.124384172476588e-05}
{"train_loss": 0.2160785049200058, "global_step": 98128, "epoch": 1102, "lr": 7.124331691610112e-05}
{"train_loss": 0.24895232915878296, "global_step": 98129, "epoch": 1102, "lr": 7.124279210458045e-05}
{"train_loss": 0.1880546510219574, "global_step": 98130, "epoch": 1102, "lr": 7.124226729020395e-05}
{"train_loss": 0.20779792964458466, "global_step": 98131, "epoch": 1102, "lr": 7.124174247297167e-05}
{"train_loss": 0.21153463423252106, "global_step": 98132, "epoch": 1102, "lr": 7.12412176528837e-05}
{"train_loss": 0.1397012621164322, "global_step": 98133, "epoch": 1102, "lr": 7.124069282994012e-05}
{"train_loss": 0.1284293830394745, "global_step": 98134, "epoch": 1102, "lr": 7.124016800414098e-05}
{"train_loss": 0.20750363171100616, "global_step": 98135, "epoch": 1102, "lr": 7.123964317548635e-05}
{"train_loss": 0.13088804483413696, "global_step": 98136, "epoch": 1102, "lr": 7.123911834397633e-05}
{"train_loss": 0.18119046092033386, "global_step": 98137, "epoch": 1102, "lr": 7.123859350961096e-05}
{"train_loss": 0.20190536975860596, "global_step": 98138, "epoch": 1102, "lr": 7.123806867239031e-05}
{"train_loss": 0.15381455421447754, "global_step": 98139, "epoch": 1102, "lr": 7.123754383231448e-05}
{"train_loss": 0.16337601840496063, "global_step": 98140, "epoch": 1102, "lr": 7.123701898938351e-05}
{"train_loss": 0.14521752297878265, "global_step": 98141, "epoch": 1102, "lr": 7.123649414359748e-05}
{"train_loss": 0.16346627473831177, "global_step": 98142, "epoch": 1102, "lr": 7.123596929495646e-05}
{"train_loss": 0.11057469993829727, "global_step": 98143, "epoch": 1102, "lr": 7.123544444346053e-05}
{"train_loss": 0.12258432060480118, "global_step": 98144, "epoch": 1102, "lr": 7.123491958910975e-05}
{"train_loss": 0.14559170603752136, "global_step": 98145, "epoch": 1102, "lr": 7.123439473190419e-05}
{"train_loss": 0.17619909346103668, "global_step": 98146, "epoch": 1102, "lr": 7.123386987184392e-05}
{"train_loss": 0.18414746224880219, "global_step": 98147, "epoch": 1102, "lr": 7.123334500892903e-05}
{"train_loss": 0.17814229428768158, "global_step": 98148, "epoch": 1102, "lr": 7.123282014315957e-05}
{"train_loss": 0.2461840659379959, "global_step": 98149, "epoch": 1102, "lr": 7.123229527453561e-05}
{"train_loss": 0.24153029918670654, "global_step": 98150, "epoch": 1102, "lr": 7.123177040305724e-05}
{"train_loss": 0.15399689972400665, "global_step": 98151, "epoch": 1102, "lr": 7.12312455287245e-05}
{"train_loss": 0.2647222876548767, "global_step": 98152, "epoch": 1102, "lr": 7.123072065153749e-05}
{"train_loss": 0.1285383403301239, "global_step": 98153, "epoch": 1102, "lr": 7.123019577149626e-05}
{"train_loss": 0.1483287215232849, "global_step": 98154, "epoch": 1102, "lr": 7.122967088860089e-05}
{"train_loss": 0.1956392228603363, "global_step": 98155, "epoch": 1102, "lr": 7.122914600285146e-05}
{"train_loss": 0.17791159451007843, "global_step": 98156, "epoch": 1102, "lr": 7.122862111424802e-05}
{"train_loss": 0.1883385330438614, "global_step": 98157, "epoch": 1102, "lr": 7.122809622279066e-05}
{"train_loss": 0.10204645991325378, "global_step": 98158, "epoch": 1102, "lr": 7.122757132847944e-05}
{"train_loss": 0.14481130242347717, "global_step": 98159, "epoch": 1102, "lr": 7.122704643131441e-05}
{"train_loss": 0.19449439644813538, "global_step": 98160, "epoch": 1102, "lr": 7.122652153129569e-05}
{"train_loss": 0.22037473320960999, "global_step": 98161, "epoch": 1102, "lr": 7.122599662842331e-05}
{"train_loss": 0.1482393443584442, "global_step": 98162, "epoch": 1102, "lr": 7.122547172269736e-05}
{"train_loss": 0.0921383798122406, "global_step": 98163, "epoch": 1102, "lr": 7.122494681411789e-05}
{"train_loss": 0.10394206643104553, "global_step": 98164, "epoch": 1102, "lr": 7.1224421902685e-05}
{"train_loss": 0.19529077410697937, "global_step": 98165, "epoch": 1102, "lr": 7.122389698839872e-05}
{"train_loss": 0.1714204524675112, "global_step": 98166, "epoch": 1102, "lr": 7.122337207125917e-05, "val_loss": 4.56910514831543}
{"train_loss": 0.12427789717912674, "global_step": 98167, "epoch": 1103, "lr": 7.12228471512664e-05}
{"train_loss": 0.11923027038574219, "global_step": 98168, "epoch": 1103, "lr": 7.122232222842046e-05}
{"train_loss": 0.16205738484859467, "global_step": 98169, "epoch": 1103, "lr": 7.122179730272145e-05}
{"train_loss": 0.15934713184833527, "global_step": 98170, "epoch": 1103, "lr": 7.122127237416942e-05}
{"train_loss": 0.19833406805992126, "global_step": 98171, "epoch": 1103, "lr": 7.122074744276446e-05}
{"train_loss": 0.221973717212677, "global_step": 98172, "epoch": 1103, "lr": 7.122022250850661e-05}
{"train_loss": 0.16372022032737732, "global_step": 98173, "epoch": 1103, "lr": 7.121969757139598e-05}
{"train_loss": 0.15879467129707336, "global_step": 98174, "epoch": 1103, "lr": 7.121917263143262e-05}
{"train_loss": 0.16984310746192932, "global_step": 98175, "epoch": 1103, "lr": 7.121864768861659e-05}
{"train_loss": 0.11749833077192307, "global_step": 98176, "epoch": 1103, "lr": 7.121812274294796e-05}
{"train_loss": 0.1364232748746872, "global_step": 98177, "epoch": 1103, "lr": 7.121759779442684e-05}
{"train_loss": 0.10353156924247742, "global_step": 98178, "epoch": 1103, "lr": 7.121707284305325e-05}
{"train_loss": 0.10013598948717117, "global_step": 98179, "epoch": 1103, "lr": 7.12165478888273e-05}
{"train_loss": 0.1596170961856842, "global_step": 98180, "epoch": 1103, "lr": 7.121602293174903e-05}
{"train_loss": 0.182527557015419, "global_step": 98181, "epoch": 1103, "lr": 7.121549797181853e-05}
{"train_loss": 0.1622421145439148, "global_step": 98182, "epoch": 1103, "lr": 7.121497300903585e-05}
{"train_loss": 0.1609387844800949, "global_step": 98183, "epoch": 1103, "lr": 7.12144480434011e-05}
{"train_loss": 0.17169509828090668, "global_step": 98184, "epoch": 1103, "lr": 7.121392307491433e-05}
{"train_loss": 0.130734384059906, "global_step": 98185, "epoch": 1103, "lr": 7.121339810357557e-05}
{"train_loss": 0.11616308987140656, "global_step": 98186, "epoch": 1103, "lr": 7.121287312938496e-05}
{"train_loss": 0.18182715773582458, "global_step": 98187, "epoch": 1103, "lr": 7.121234815234252e-05}
{"train_loss": 0.15942132472991943, "global_step": 98188, "epoch": 1103, "lr": 7.121182317244836e-05}
{"train_loss": 0.13182350993156433, "global_step": 98189, "epoch": 1103, "lr": 7.121129818970251e-05}
{"train_loss": 0.15249435603618622, "global_step": 98190, "epoch": 1103, "lr": 7.121077320410506e-05}
{"train_loss": 0.09664586931467056, "global_step": 98191, "epoch": 1103, "lr": 7.121024821565607e-05}
{"train_loss": 0.21083681285381317, "global_step": 98192, "epoch": 1103, "lr": 7.120972322435564e-05}
{"train_loss": 0.23431020975112915, "global_step": 98193, "epoch": 1103, "lr": 7.120919823020382e-05}
{"train_loss": 0.1427270770072937, "global_step": 98194, "epoch": 1103, "lr": 7.120867323320069e-05}
{"train_loss": 0.10151855647563934, "global_step": 98195, "epoch": 1103, "lr": 7.120814823334629e-05}
{"train_loss": 0.17015878856182098, "global_step": 98196, "epoch": 1103, "lr": 7.120762323064073e-05}
{"train_loss": 0.1362432837486267, "global_step": 98197, "epoch": 1103, "lr": 7.120709822508406e-05}
{"train_loss": 0.18313069641590118, "global_step": 98198, "epoch": 1103, "lr": 7.120657321667634e-05}
{"train_loss": 0.16621731221675873, "global_step": 98199, "epoch": 1103, "lr": 7.120604820541768e-05}
{"train_loss": 0.14990399777889252, "global_step": 98200, "epoch": 1103, "lr": 7.12055231913081e-05}
{"train_loss": 0.1252908557653427, "global_step": 98201, "epoch": 1103, "lr": 7.12049981743477e-05}
{"train_loss": 0.15393701195716858, "global_step": 98202, "epoch": 1103, "lr": 7.120447315453655e-05}
{"train_loss": 0.16445215046405792, "global_step": 98203, "epoch": 1103, "lr": 7.120394813187473e-05}
{"train_loss": 0.1639399230480194, "global_step": 98204, "epoch": 1103, "lr": 7.120342310636229e-05}
{"train_loss": 0.1376904547214508, "global_step": 98205, "epoch": 1103, "lr": 7.12028980779993e-05}
{"train_loss": 0.1300225704908371, "global_step": 98206, "epoch": 1103, "lr": 7.120237304678586e-05}
{"train_loss": 0.23750637471675873, "global_step": 98207, "epoch": 1103, "lr": 7.1201848012722e-05}
{"train_loss": 0.09266620874404907, "global_step": 98208, "epoch": 1103, "lr": 7.12013229758078e-05}
{"train_loss": 0.1573919653892517, "global_step": 98209, "epoch": 1103, "lr": 7.120079793604337e-05}
{"train_loss": 0.11722360551357269, "global_step": 98210, "epoch": 1103, "lr": 7.120027289342872e-05}
{"train_loss": 0.14138084650039673, "global_step": 98211, "epoch": 1103, "lr": 7.119974784796396e-05}
{"train_loss": 0.23394037783145905, "global_step": 98212, "epoch": 1103, "lr": 7.119922279964917e-05}
{"train_loss": 0.1769762635231018, "global_step": 98213, "epoch": 1103, "lr": 7.11986977484844e-05}
{"train_loss": 0.18972527980804443, "global_step": 98214, "epoch": 1103, "lr": 7.11981726944697e-05}
{"train_loss": 0.21936051547527313, "global_step": 98215, "epoch": 1103, "lr": 7.11976476376052e-05}
{"train_loss": 0.13132651150226593, "global_step": 98216, "epoch": 1103, "lr": 7.11971225778909e-05}
{"train_loss": 0.13358290493488312, "global_step": 98217, "epoch": 1103, "lr": 7.119659751532692e-05}
{"train_loss": 0.15556979179382324, "global_step": 98218, "epoch": 1103, "lr": 7.119607244991332e-05}
{"train_loss": 0.15987005829811096, "global_step": 98219, "epoch": 1103, "lr": 7.119554738165016e-05}
{"train_loss": 0.13537214696407318, "global_step": 98220, "epoch": 1103, "lr": 7.119502231053752e-05}
{"train_loss": 0.20340852439403534, "global_step": 98221, "epoch": 1103, "lr": 7.119449723657548e-05}
{"train_loss": 0.19665935635566711, "global_step": 98222, "epoch": 1103, "lr": 7.119397215976407e-05}
{"train_loss": 0.14529170095920563, "global_step": 98223, "epoch": 1103, "lr": 7.11934470801034e-05}
{"train_loss": 0.12866199016571045, "global_step": 98224, "epoch": 1103, "lr": 7.119292199759354e-05}
{"train_loss": 0.08789804577827454, "global_step": 98225, "epoch": 1103, "lr": 7.119239691223454e-05}
{"train_loss": 0.2366928905248642, "global_step": 98226, "epoch": 1103, "lr": 7.11918718240265e-05}
{"train_loss": 0.174352765083313, "global_step": 98227, "epoch": 1103, "lr": 7.119134673296945e-05}
{"train_loss": 0.16524744033813477, "global_step": 98228, "epoch": 1103, "lr": 7.119082163906347e-05}
{"train_loss": 0.24829405546188354, "global_step": 98229, "epoch": 1103, "lr": 7.119029654230865e-05}
{"train_loss": 0.09778014570474625, "global_step": 98230, "epoch": 1103, "lr": 7.118977144270507e-05}
{"train_loss": 0.17691665887832642, "global_step": 98231, "epoch": 1103, "lr": 7.118924634025277e-05}
{"train_loss": 0.24907442927360535, "global_step": 98232, "epoch": 1103, "lr": 7.118872123495184e-05}
{"train_loss": 0.17087936401367188, "global_step": 98233, "epoch": 1103, "lr": 7.118819612680235e-05}
{"train_loss": 0.16593815386295319, "global_step": 98234, "epoch": 1103, "lr": 7.118767101580437e-05}
{"train_loss": 0.12799987196922302, "global_step": 98235, "epoch": 1103, "lr": 7.118714590195795e-05}
{"train_loss": 0.18553170561790466, "global_step": 98236, "epoch": 1103, "lr": 7.118662078526318e-05}
{"train_loss": 0.20530445873737335, "global_step": 98237, "epoch": 1103, "lr": 7.118609566572013e-05}
{"train_loss": 0.09757418930530548, "global_step": 98238, "epoch": 1103, "lr": 7.118557054332887e-05}
{"train_loss": 0.19175750017166138, "global_step": 98239, "epoch": 1103, "lr": 7.118504541808946e-05}
{"train_loss": 0.16183535754680634, "global_step": 98240, "epoch": 1103, "lr": 7.118452029000197e-05}
{"train_loss": 0.31546565890312195, "global_step": 98241, "epoch": 1103, "lr": 7.118399515906651e-05}
{"train_loss": 0.23406386375427246, "global_step": 98242, "epoch": 1103, "lr": 7.11834700252831e-05}
{"train_loss": 0.14455974102020264, "global_step": 98243, "epoch": 1103, "lr": 7.118294488865183e-05}
{"train_loss": 0.15617170929908752, "global_step": 98244, "epoch": 1103, "lr": 7.118241974917278e-05}
{"train_loss": 0.202082559466362, "global_step": 98245, "epoch": 1103, "lr": 7.1181894606846e-05}
{"train_loss": 0.16215857863426208, "global_step": 98246, "epoch": 1103, "lr": 7.118136946167158e-05}
{"train_loss": 0.1849534958600998, "global_step": 98247, "epoch": 1103, "lr": 7.118084431364957e-05}
{"train_loss": 0.29734593629837036, "global_step": 98248, "epoch": 1103, "lr": 7.118031916278007e-05}
{"train_loss": 0.23231588304042816, "global_step": 98249, "epoch": 1103, "lr": 7.117979400906313e-05}
{"train_loss": 0.1551908403635025, "global_step": 98250, "epoch": 1103, "lr": 7.117926885249882e-05}
{"train_loss": 0.17654623091220856, "global_step": 98251, "epoch": 1103, "lr": 7.117874369308723e-05}
{"train_loss": 0.17121167480945587, "global_step": 98252, "epoch": 1103, "lr": 7.117821853082842e-05}
{"train_loss": 0.12241227179765701, "global_step": 98253, "epoch": 1103, "lr": 7.117769336572244e-05}
{"train_loss": 0.1914432942867279, "global_step": 98254, "epoch": 1103, "lr": 7.117716819776938e-05}
{"train_loss": 0.16467637769626767, "global_step": 98255, "epoch": 1103, "lr": 7.11766430269693e-05, "val_loss": 4.654786109924316}
{"train_loss": 0.1356421411037445, "global_step": 98256, "epoch": 1104, "lr": 7.117611785332229e-05}
{"train_loss": 0.18147213757038116, "global_step": 98257, "epoch": 1104, "lr": 7.11755926768284e-05}
{"train_loss": 0.15672194957733154, "global_step": 98258, "epoch": 1104, "lr": 7.117506749748771e-05}
{"train_loss": 0.14589449763298035, "global_step": 98259, "epoch": 1104, "lr": 7.11745423153003e-05}
{"train_loss": 0.24858801066875458, "global_step": 98260, "epoch": 1104, "lr": 7.117401713026622e-05}
{"train_loss": 0.2243930846452713, "global_step": 98261, "epoch": 1104, "lr": 7.117349194238557e-05}
{"train_loss": 0.1263001561164856, "global_step": 98262, "epoch": 1104, "lr": 7.117296675165838e-05}
{"train_loss": 0.10090174525976181, "global_step": 98263, "epoch": 1104, "lr": 7.117244155808476e-05}
{"train_loss": 0.16790400445461273, "global_step": 98264, "epoch": 1104, "lr": 7.117191636166473e-05}
{"train_loss": 0.1282295435667038, "global_step": 98265, "epoch": 1104, "lr": 7.117139116239844e-05}
{"train_loss": 0.15710479021072388, "global_step": 98266, "epoch": 1104, "lr": 7.117086596028588e-05}
{"train_loss": 0.19782567024230957, "global_step": 98267, "epoch": 1104, "lr": 7.117034075532718e-05}
{"train_loss": 0.18890729546546936, "global_step": 98268, "epoch": 1104, "lr": 7.116981554752237e-05}
{"train_loss": 0.20874372124671936, "global_step": 98269, "epoch": 1104, "lr": 7.116929033687154e-05}
{"train_loss": 0.16689318418502808, "global_step": 98270, "epoch": 1104, "lr": 7.116876512337475e-05}
{"train_loss": 0.19725598394870758, "global_step": 98271, "epoch": 1104, "lr": 7.116823990703211e-05}
{"train_loss": 0.11309071630239487, "global_step": 98272, "epoch": 1104, "lr": 7.116771468784362e-05}
{"train_loss": 0.23880037665367126, "global_step": 98273, "epoch": 1104, "lr": 7.116718946580942e-05}
{"train_loss": 0.21805346012115479, "global_step": 98274, "epoch": 1104, "lr": 7.116666424092953e-05}
{"train_loss": 0.17903468012809753, "global_step": 98275, "epoch": 1104, "lr": 7.116613901320404e-05}
{"train_loss": 0.17570126056671143, "global_step": 98276, "epoch": 1104, "lr": 7.116561378263302e-05}
{"train_loss": 0.1847982108592987, "global_step": 98277, "epoch": 1104, "lr": 7.116508854921654e-05}
{"train_loss": 0.24122782051563263, "global_step": 98278, "epoch": 1104, "lr": 7.116456331295468e-05}
{"train_loss": 0.23156240582466125, "global_step": 98279, "epoch": 1104, "lr": 7.11640380738475e-05}
{"train_loss": 0.22735680639743805, "global_step": 98280, "epoch": 1104, "lr": 7.116351283189507e-05}
{"train_loss": 0.16465085744857788, "global_step": 98281, "epoch": 1104, "lr": 7.116298758709746e-05}
{"train_loss": 0.2056366354227066, "global_step": 98282, "epoch": 1104, "lr": 7.116246233945473e-05}
{"train_loss": 0.20989681780338287, "global_step": 98283, "epoch": 1104, "lr": 7.116193708896699e-05}
{"train_loss": 0.19514133036136627, "global_step": 98284, "epoch": 1104, "lr": 7.116141183563429e-05}
{"train_loss": 0.1442026048898697, "global_step": 98285, "epoch": 1104, "lr": 7.116088657945668e-05}
{"train_loss": 0.18445611000061035, "global_step": 98286, "epoch": 1104, "lr": 7.116036132043424e-05}
{"train_loss": 0.14167562127113342, "global_step": 98287, "epoch": 1104, "lr": 7.115983605856706e-05}
{"train_loss": 0.23257257044315338, "global_step": 98288, "epoch": 1104, "lr": 7.115931079385521e-05}
{"train_loss": 0.14041882753372192, "global_step": 98289, "epoch": 1104, "lr": 7.115878552629874e-05}
{"train_loss": 0.1290847659111023, "global_step": 98290, "epoch": 1104, "lr": 7.115826025589772e-05}
{"train_loss": 0.21721677482128143, "global_step": 98291, "epoch": 1104, "lr": 7.115773498265225e-05}
{"train_loss": 0.21364708244800568, "global_step": 98292, "epoch": 1104, "lr": 7.115720970656236e-05}
{"train_loss": 0.2746237516403198, "global_step": 98293, "epoch": 1104, "lr": 7.115668442762815e-05}
{"train_loss": 0.22696301341056824, "global_step": 98294, "epoch": 1104, "lr": 7.115615914584968e-05}
{"train_loss": 0.2584938406944275, "global_step": 98295, "epoch": 1104, "lr": 7.115563386122701e-05}
{"train_loss": 0.2639203667640686, "global_step": 98296, "epoch": 1104, "lr": 7.115510857376023e-05}
{"train_loss": 0.2754143476486206, "global_step": 98297, "epoch": 1104, "lr": 7.115458328344942e-05}
{"train_loss": 0.19184133410453796, "global_step": 98298, "epoch": 1104, "lr": 7.11540579902946e-05}
{"train_loss": 0.1264389306306839, "global_step": 98299, "epoch": 1104, "lr": 7.115353269429592e-05}
{"train_loss": 0.24632485210895538, "global_step": 98300, "epoch": 1104, "lr": 7.115300739545336e-05}
{"train_loss": 0.13712970912456512, "global_step": 98301, "epoch": 1104, "lr": 7.115248209376706e-05}
{"train_loss": 0.13454069197177887, "global_step": 98302, "epoch": 1104, "lr": 7.115195678923706e-05}
{"train_loss": 0.2432951182126999, "global_step": 98303, "epoch": 1104, "lr": 7.115143148186345e-05}
{"train_loss": 0.1308809518814087, "global_step": 98304, "epoch": 1104, "lr": 7.115090617164627e-05}
{"train_loss": 0.15489228069782257, "global_step": 98305, "epoch": 1104, "lr": 7.11503808585856e-05}
{"train_loss": 0.17013680934906006, "global_step": 98306, "epoch": 1104, "lr": 7.114985554268156e-05}
{"train_loss": 0.14903877675533295, "global_step": 98307, "epoch": 1104, "lr": 7.114933022393415e-05}
{"train_loss": 0.1256849765777588, "global_step": 98308, "epoch": 1104, "lr": 7.114880490234347e-05}
{"train_loss": 0.1284000277519226, "global_step": 98309, "epoch": 1104, "lr": 7.114827957790959e-05}
{"train_loss": 0.2010933756828308, "global_step": 98310, "epoch": 1104, "lr": 7.11477542506326e-05}
{"train_loss": 0.14627136290073395, "global_step": 98311, "epoch": 1104, "lr": 7.114722892051255e-05}
{"train_loss": 0.20728915929794312, "global_step": 98312, "epoch": 1104, "lr": 7.11467035875495e-05}
{"train_loss": 0.15123099088668823, "global_step": 98313, "epoch": 1104, "lr": 7.114617825174354e-05}
{"train_loss": 0.1905885934829712, "global_step": 98314, "epoch": 1104, "lr": 7.114565291309474e-05}
{"train_loss": 0.18758398294448853, "global_step": 98315, "epoch": 1104, "lr": 7.114512757160314e-05}
{"train_loss": 0.15816549956798553, "global_step": 98316, "epoch": 1104, "lr": 7.114460222726885e-05}
{"train_loss": 0.17389072477817535, "global_step": 98317, "epoch": 1104, "lr": 7.114407688009194e-05}
{"train_loss": 0.12112509459257126, "global_step": 98318, "epoch": 1104, "lr": 7.114355153007243e-05}
{"train_loss": 0.12380190193653107, "global_step": 98319, "epoch": 1104, "lr": 7.114302617721046e-05}
{"train_loss": 0.1543017327785492, "global_step": 98320, "epoch": 1104, "lr": 7.114250082150606e-05}
{"train_loss": 0.20381443202495575, "global_step": 98321, "epoch": 1104, "lr": 7.114197546295931e-05}
{"train_loss": 0.13166621327400208, "global_step": 98322, "epoch": 1104, "lr": 7.114145010157028e-05}
{"train_loss": 0.15896975994110107, "global_step": 98323, "epoch": 1104, "lr": 7.114092473733902e-05}
{"train_loss": 0.12950864434242249, "global_step": 98324, "epoch": 1104, "lr": 7.114039937026564e-05}
{"train_loss": 0.19262854754924774, "global_step": 98325, "epoch": 1104, "lr": 7.113987400035019e-05}
{"train_loss": 0.18024961650371552, "global_step": 98326, "epoch": 1104, "lr": 7.113934862759273e-05}
{"train_loss": 0.14566542208194733, "global_step": 98327, "epoch": 1104, "lr": 7.113882325199336e-05}
{"train_loss": 0.1962345987558365, "global_step": 98328, "epoch": 1104, "lr": 7.113829787355213e-05}
{"train_loss": 0.13116323947906494, "global_step": 98329, "epoch": 1104, "lr": 7.11377724922691e-05}
{"train_loss": 0.18460680544376373, "global_step": 98330, "epoch": 1104, "lr": 7.113724710814436e-05}
{"train_loss": 0.18697337806224823, "global_step": 98331, "epoch": 1104, "lr": 7.113672172117798e-05}
{"train_loss": 0.14679406583309174, "global_step": 98332, "epoch": 1104, "lr": 7.113619633137003e-05}
{"train_loss": 0.1833963841199875, "global_step": 98333, "epoch": 1104, "lr": 7.113567093872057e-05}
{"train_loss": 0.19075851142406464, "global_step": 98334, "epoch": 1104, "lr": 7.113514554322967e-05}
{"train_loss": 0.1394275426864624, "global_step": 98335, "epoch": 1104, "lr": 7.113462014489742e-05}
{"train_loss": 0.20295986533164978, "global_step": 98336, "epoch": 1104, "lr": 7.113409474372385e-05}
{"train_loss": 0.21283824741840363, "global_step": 98337, "epoch": 1104, "lr": 7.113356933970908e-05}
{"train_loss": 0.2571626305580139, "global_step": 98338, "epoch": 1104, "lr": 7.113304393285316e-05}
{"train_loss": 0.12002326548099518, "global_step": 98339, "epoch": 1104, "lr": 7.113251852315616e-05}
{"train_loss": 0.09336838126182556, "global_step": 98340, "epoch": 1104, "lr": 7.113199311061815e-05}
{"train_loss": 0.1909245103597641, "global_step": 98341, "epoch": 1104, "lr": 7.113146769523919e-05}
{"train_loss": 0.1478731334209442, "global_step": 98342, "epoch": 1104, "lr": 7.113094227701937e-05}
{"train_loss": 0.18004029989242554, "global_step": 98343, "epoch": 1104, "lr": 7.113041685595876e-05}
{"train_loss": 0.178270535438918, "global_step": 98344, "epoch": 1104, "lr": 7.11298914320574e-05, "val_loss": 4.714313983917236}
{"train_loss": 0.1735047847032547, "global_step": 98345, "epoch": 1105, "lr": 7.112936600531541e-05}
{"train_loss": 0.17313610017299652, "global_step": 98346, "epoch": 1105, "lr": 7.112884057573283e-05}
{"train_loss": 0.15950830280780792, "global_step": 98347, "epoch": 1105, "lr": 7.112831514330972e-05}
{"train_loss": 0.10512544214725494, "global_step": 98348, "epoch": 1105, "lr": 7.112778970804617e-05}
{"train_loss": 0.14150509238243103, "global_step": 98349, "epoch": 1105, "lr": 7.112726426994226e-05}
{"train_loss": 0.19313013553619385, "global_step": 98350, "epoch": 1105, "lr": 7.112673882899805e-05}
{"train_loss": 0.21535846590995789, "global_step": 98351, "epoch": 1105, "lr": 7.112621338521359e-05}
{"train_loss": 0.17304298281669617, "global_step": 98352, "epoch": 1105, "lr": 7.112568793858897e-05}
{"train_loss": 0.13067279756069183, "global_step": 98353, "epoch": 1105, "lr": 7.112516248912427e-05}
{"train_loss": 0.18393714725971222, "global_step": 98354, "epoch": 1105, "lr": 7.112463703681953e-05}
{"train_loss": 0.24373038113117218, "global_step": 98355, "epoch": 1105, "lr": 7.112411158167484e-05}
{"train_loss": 0.14692053198814392, "global_step": 98356, "epoch": 1105, "lr": 7.112358612369028e-05}
{"train_loss": 0.2653466761112213, "global_step": 98357, "epoch": 1105, "lr": 7.112306066286591e-05}
{"train_loss": 0.15416422486305237, "global_step": 98358, "epoch": 1105, "lr": 7.112253519920181e-05}
{"train_loss": 0.20967766642570496, "global_step": 98359, "epoch": 1105, "lr": 7.112200973269803e-05}
{"train_loss": 0.10846563428640366, "global_step": 98360, "epoch": 1105, "lr": 7.112148426335466e-05}
{"train_loss": 0.14218944311141968, "global_step": 98361, "epoch": 1105, "lr": 7.112095879117178e-05}
{"train_loss": 0.1668291985988617, "global_step": 98362, "epoch": 1105, "lr": 7.112043331614941e-05}
{"train_loss": 0.20862051844596863, "global_step": 98363, "epoch": 1105, "lr": 7.111990783828768e-05}
{"train_loss": 0.15985269844532013, "global_step": 98364, "epoch": 1105, "lr": 7.111938235758664e-05}
{"train_loss": 0.18553239107131958, "global_step": 98365, "epoch": 1105, "lr": 7.111885687404633e-05}
{"train_loss": 0.20186707377433777, "global_step": 98366, "epoch": 1105, "lr": 7.111833138766687e-05}
{"train_loss": 0.19394052028656006, "global_step": 98367, "epoch": 1105, "lr": 7.111780589844831e-05}
{"train_loss": 0.10657776892185211, "global_step": 98368, "epoch": 1105, "lr": 7.111728040639072e-05}
{"train_loss": 0.1885656714439392, "global_step": 98369, "epoch": 1105, "lr": 7.111675491149416e-05}
{"train_loss": 0.17819571495056152, "global_step": 98370, "epoch": 1105, "lr": 7.11162294137587e-05}
{"train_loss": 0.10480579733848572, "global_step": 98371, "epoch": 1105, "lr": 7.111570391318443e-05}
{"train_loss": 0.19051475822925568, "global_step": 98372, "epoch": 1105, "lr": 7.111517840977141e-05}
{"train_loss": 0.15407440066337585, "global_step": 98373, "epoch": 1105, "lr": 7.11146529035197e-05}
{"train_loss": 0.14478617906570435, "global_step": 98374, "epoch": 1105, "lr": 7.11141273944294e-05}
{"train_loss": 0.17319254577159882, "global_step": 98375, "epoch": 1105, "lr": 7.111360188250054e-05}
{"train_loss": 0.1803579032421112, "global_step": 98376, "epoch": 1105, "lr": 7.111307636773323e-05}
{"train_loss": 0.20226095616817474, "global_step": 98377, "epoch": 1105, "lr": 7.111255085012754e-05}
{"train_loss": 0.2172202318906784, "global_step": 98378, "epoch": 1105, "lr": 7.11120253296835e-05}
{"train_loss": 0.2047305554151535, "global_step": 98379, "epoch": 1105, "lr": 7.111149980640121e-05}
{"train_loss": 0.1519642174243927, "global_step": 98380, "epoch": 1105, "lr": 7.111097428028075e-05}
{"train_loss": 0.16314281523227692, "global_step": 98381, "epoch": 1105, "lr": 7.111044875132217e-05}
{"train_loss": 0.25673407316207886, "global_step": 98382, "epoch": 1105, "lr": 7.110992321952554e-05}
{"train_loss": 0.2098187357187271, "global_step": 98383, "epoch": 1105, "lr": 7.110939768489095e-05}
{"train_loss": 0.13567106425762177, "global_step": 98384, "epoch": 1105, "lr": 7.110887214741845e-05}
{"train_loss": 0.12943829596042633, "global_step": 98385, "epoch": 1105, "lr": 7.110834660710812e-05}
{"train_loss": 0.11032573133707047, "global_step": 98386, "epoch": 1105, "lr": 7.110782106396004e-05}
{"train_loss": 0.16880880296230316, "global_step": 98387, "epoch": 1105, "lr": 7.110729551797426e-05}
{"train_loss": 0.13883017003536224, "global_step": 98388, "epoch": 1105, "lr": 7.110676996915086e-05}
{"train_loss": 0.1443365067243576, "global_step": 98389, "epoch": 1105, "lr": 7.110624441748992e-05}
{"train_loss": 0.2028142362833023, "global_step": 98390, "epoch": 1105, "lr": 7.11057188629915e-05}
{"train_loss": 0.11747342348098755, "global_step": 98391, "epoch": 1105, "lr": 7.110519330565565e-05}
{"train_loss": 0.1936185508966446, "global_step": 98392, "epoch": 1105, "lr": 7.11046677454825e-05}
{"train_loss": 0.11941969394683838, "global_step": 98393, "epoch": 1105, "lr": 7.110414218247207e-05}
{"train_loss": 0.1471315622329712, "global_step": 98394, "epoch": 1105, "lr": 7.110361661662443e-05}
{"train_loss": 0.1535695195198059, "global_step": 98395, "epoch": 1105, "lr": 7.110309104793968e-05}
{"train_loss": 0.18147066235542297, "global_step": 98396, "epoch": 1105, "lr": 7.11025654764179e-05}
{"train_loss": 0.13008391857147217, "global_step": 98397, "epoch": 1105, "lr": 7.11020399020591e-05}
{"train_loss": 0.16656291484832764, "global_step": 98398, "epoch": 1105, "lr": 7.11015143248634e-05}
{"train_loss": 0.12942887842655182, "global_step": 98399, "epoch": 1105, "lr": 7.110098874483086e-05}
{"train_loss": 0.12401819229125977, "global_step": 98400, "epoch": 1105, "lr": 7.110046316196156e-05}
{"train_loss": 0.15702605247497559, "global_step": 98401, "epoch": 1105, "lr": 7.109993757625554e-05}
{"train_loss": 0.14466704428195953, "global_step": 98402, "epoch": 1105, "lr": 7.109941198771291e-05}
{"train_loss": 0.0853172019124031, "global_step": 98403, "epoch": 1105, "lr": 7.109888639633372e-05}
{"train_loss": 0.2069985270500183, "global_step": 98404, "epoch": 1105, "lr": 7.109836080211803e-05}
{"train_loss": 0.16351400315761566, "global_step": 98405, "epoch": 1105, "lr": 7.109783520506592e-05}
{"train_loss": 0.21798737347126007, "global_step": 98406, "epoch": 1105, "lr": 7.109730960517747e-05}
{"train_loss": 0.1381397396326065, "global_step": 98407, "epoch": 1105, "lr": 7.109678400245276e-05}
{"train_loss": 0.16366954147815704, "global_step": 98408, "epoch": 1105, "lr": 7.109625839689183e-05}
{"train_loss": 0.22516722977161407, "global_step": 98409, "epoch": 1105, "lr": 7.109573278849476e-05}
{"train_loss": 0.15322816371917725, "global_step": 98410, "epoch": 1105, "lr": 7.109520717726163e-05}
{"train_loss": 0.15732817351818085, "global_step": 98411, "epoch": 1105, "lr": 7.10946815631925e-05}
{"train_loss": 0.1145089715719223, "global_step": 98412, "epoch": 1105, "lr": 7.109415594628745e-05}
{"train_loss": 0.19903157651424408, "global_step": 98413, "epoch": 1105, "lr": 7.109363032654655e-05}
{"train_loss": 0.12483468651771545, "global_step": 98414, "epoch": 1105, "lr": 7.109310470396988e-05}
{"train_loss": 0.27482447028160095, "global_step": 98415, "epoch": 1105, "lr": 7.109257907855749e-05}
{"train_loss": 0.2345607876777649, "global_step": 98416, "epoch": 1105, "lr": 7.109205345030945e-05}
{"train_loss": 0.2741852402687073, "global_step": 98417, "epoch": 1105, "lr": 7.109152781922585e-05}
{"train_loss": 0.1330653578042984, "global_step": 98418, "epoch": 1105, "lr": 7.109100218530676e-05}
{"train_loss": 0.17660944163799286, "global_step": 98419, "epoch": 1105, "lr": 7.109047654855223e-05}
{"train_loss": 0.10886800289154053, "global_step": 98420, "epoch": 1105, "lr": 7.108995090896233e-05}
{"train_loss": 0.14653430879116058, "global_step": 98421, "epoch": 1105, "lr": 7.108942526653716e-05}
{"train_loss": 0.2898018956184387, "global_step": 98422, "epoch": 1105, "lr": 7.108889962127679e-05}
{"train_loss": 0.1886317878961563, "global_step": 98423, "epoch": 1105, "lr": 7.108837397318125e-05}
{"train_loss": 0.250324010848999, "global_step": 98424, "epoch": 1105, "lr": 7.108784832225064e-05}
{"train_loss": 0.13018645346164703, "global_step": 98425, "epoch": 1105, "lr": 7.108732266848502e-05}
{"train_loss": 0.18240706622600555, "global_step": 98426, "epoch": 1105, "lr": 7.108679701188449e-05}
{"train_loss": 0.2703539729118347, "global_step": 98427, "epoch": 1105, "lr": 7.108627135244907e-05}
{"train_loss": 0.17384110391139984, "global_step": 98428, "epoch": 1105, "lr": 7.108574569017888e-05}
{"train_loss": 0.19818511605262756, "global_step": 98429, "epoch": 1105, "lr": 7.108522002507394e-05}
{"train_loss": 0.2451414316892624, "global_step": 98430, "epoch": 1105, "lr": 7.108469435713437e-05}
{"train_loss": 0.12140201777219772, "global_step": 98431, "epoch": 1105, "lr": 7.108416868636022e-05}
{"train_loss": 0.16140995919704437, "global_step": 98432, "epoch": 1105, "lr": 7.108364301275156e-05}
{"train_loss": 0.1721093433626582, "global_step": 98433, "epoch": 1105, "lr": 7.108311733630845e-05, "val_loss": 4.499545574188232, "train_action_mse_error": 28.25901985168457}
{"train_loss": 0.14827577769756317, "global_step": 98434, "epoch": 1106, "lr": 7.108259165703098e-05}
{"train_loss": 0.176545187830925, "global_step": 98435, "epoch": 1106, "lr": 7.108206597491921e-05}
{"train_loss": 0.15526767075061798, "global_step": 98436, "epoch": 1106, "lr": 7.108154028997321e-05}
{"train_loss": 0.1288141906261444, "global_step": 98437, "epoch": 1106, "lr": 7.108101460219306e-05}
{"train_loss": 0.08768238872289658, "global_step": 98438, "epoch": 1106, "lr": 7.108048891157884e-05}
{"train_loss": 0.14306218922138214, "global_step": 98439, "epoch": 1106, "lr": 7.107996321813057e-05}
{"train_loss": 0.17032621800899506, "global_step": 98440, "epoch": 1106, "lr": 7.107943752184838e-05}
{"train_loss": 0.16161775588989258, "global_step": 98441, "epoch": 1106, "lr": 7.107891182273232e-05}
{"train_loss": 0.1742560714483261, "global_step": 98442, "epoch": 1106, "lr": 7.107838612078244e-05}
{"train_loss": 0.18760225176811218, "global_step": 98443, "epoch": 1106, "lr": 7.107786041599884e-05}
{"train_loss": 0.1877482682466507, "global_step": 98444, "epoch": 1106, "lr": 7.107733470838159e-05}
{"train_loss": 0.1556805670261383, "global_step": 98445, "epoch": 1106, "lr": 7.107680899793072e-05}
{"train_loss": 0.15926218032836914, "global_step": 98446, "epoch": 1106, "lr": 7.107628328464635e-05}
{"train_loss": 0.2286641150712967, "global_step": 98447, "epoch": 1106, "lr": 7.107575756852852e-05}
{"train_loss": 0.10445736348628998, "global_step": 98448, "epoch": 1106, "lr": 7.107523184957732e-05}
{"train_loss": 0.15144158899784088, "global_step": 98449, "epoch": 1106, "lr": 7.10747061277928e-05}
{"train_loss": 0.10344509035348892, "global_step": 98450, "epoch": 1106, "lr": 7.107418040317506e-05}
{"train_loss": 0.15660960972309113, "global_step": 98451, "epoch": 1106, "lr": 7.107365467572413e-05}
{"train_loss": 0.17621609568595886, "global_step": 98452, "epoch": 1106, "lr": 7.107312894544013e-05}
{"train_loss": 0.14385242760181427, "global_step": 98453, "epoch": 1106, "lr": 7.107260321232308e-05}
{"train_loss": 0.13690032064914703, "global_step": 98454, "epoch": 1106, "lr": 7.10720774763731e-05}
{"train_loss": 0.09292831271886826, "global_step": 98455, "epoch": 1106, "lr": 7.107155173759021e-05}
{"train_loss": 0.12298830598592758, "global_step": 98456, "epoch": 1106, "lr": 7.107102599597452e-05}
{"train_loss": 0.16418100893497467, "global_step": 98457, "epoch": 1106, "lr": 7.10705002515261e-05}
{"train_loss": 0.24978533387184143, "global_step": 98458, "epoch": 1106, "lr": 7.106997450424499e-05}
{"train_loss": 0.13940826058387756, "global_step": 98459, "epoch": 1106, "lr": 7.106944875413129e-05}
{"train_loss": 0.1477588713169098, "global_step": 98460, "epoch": 1106, "lr": 7.106892300118505e-05}
{"train_loss": 0.1513535976409912, "global_step": 98461, "epoch": 1106, "lr": 7.106839724540636e-05}
{"train_loss": 0.1463880091905594, "global_step": 98462, "epoch": 1106, "lr": 7.106787148679527e-05}
{"train_loss": 0.12582100927829742, "global_step": 98463, "epoch": 1106, "lr": 7.106734572535187e-05}
{"train_loss": 0.17094841599464417, "global_step": 98464, "epoch": 1106, "lr": 7.106681996107621e-05}
{"train_loss": 0.14038847386837006, "global_step": 98465, "epoch": 1106, "lr": 7.106629419396838e-05}
{"train_loss": 0.1940114051103592, "global_step": 98466, "epoch": 1106, "lr": 7.106576842402845e-05}
{"train_loss": 0.12924623489379883, "global_step": 98467, "epoch": 1106, "lr": 7.106524265125648e-05}
{"train_loss": 0.12126108258962631, "global_step": 98468, "epoch": 1106, "lr": 7.106471687565255e-05}
{"train_loss": 0.12530986964702606, "global_step": 98469, "epoch": 1106, "lr": 7.106419109721671e-05}
{"train_loss": 0.20089510083198547, "global_step": 98470, "epoch": 1106, "lr": 7.106366531594906e-05}
{"train_loss": 0.09886748343706131, "global_step": 98471, "epoch": 1106, "lr": 7.106313953184965e-05}
{"train_loss": 0.1639837920665741, "global_step": 98472, "epoch": 1106, "lr": 7.106261374491857e-05}
{"train_loss": 0.19895663857460022, "global_step": 98473, "epoch": 1106, "lr": 7.106208795515586e-05}
{"train_loss": 0.1042562797665596, "global_step": 98474, "epoch": 1106, "lr": 7.106156216256162e-05}
{"train_loss": 0.13806501030921936, "global_step": 98475, "epoch": 1106, "lr": 7.106103636713591e-05}
{"train_loss": 0.20858590304851532, "global_step": 98476, "epoch": 1106, "lr": 7.10605105688788e-05}
{"train_loss": 0.13799609243869781, "global_step": 98477, "epoch": 1106, "lr": 7.105998476779036e-05}
{"train_loss": 0.24132557213306427, "global_step": 98478, "epoch": 1106, "lr": 7.105945896387065e-05}
{"train_loss": 0.1501404196023941, "global_step": 98479, "epoch": 1106, "lr": 7.105893315711977e-05}
{"train_loss": 0.12131808698177338, "global_step": 98480, "epoch": 1106, "lr": 7.105840734753776e-05}
{"train_loss": 0.13754627108573914, "global_step": 98481, "epoch": 1106, "lr": 7.105788153512472e-05}
{"train_loss": 0.23923027515411377, "global_step": 98482, "epoch": 1106, "lr": 7.10573557198807e-05}
{"train_loss": 0.24793994426727295, "global_step": 98483, "epoch": 1106, "lr": 7.105682990180576e-05}
{"train_loss": 0.14275747537612915, "global_step": 98484, "epoch": 1106, "lr": 7.105630408089999e-05}
{"train_loss": 0.19401319324970245, "global_step": 98485, "epoch": 1106, "lr": 7.105577825716347e-05}
{"train_loss": 0.14969606697559357, "global_step": 98486, "epoch": 1106, "lr": 7.105525243059626e-05}
{"train_loss": 0.16439612209796906, "global_step": 98487, "epoch": 1106, "lr": 7.105472660119841e-05}
{"train_loss": 0.17850211262702942, "global_step": 98488, "epoch": 1106, "lr": 7.105420076897e-05}
{"train_loss": 0.16094818711280823, "global_step": 98489, "epoch": 1106, "lr": 7.105367493391113e-05}
{"train_loss": 0.12297307699918747, "global_step": 98490, "epoch": 1106, "lr": 7.105314909602185e-05}
{"train_loss": 0.20253880321979523, "global_step": 98491, "epoch": 1106, "lr": 7.105262325530222e-05}
{"train_loss": 0.13368792831897736, "global_step": 98492, "epoch": 1106, "lr": 7.105209741175234e-05}
{"train_loss": 0.23834753036499023, "global_step": 98493, "epoch": 1106, "lr": 7.105157156537224e-05}
{"train_loss": 0.13200592994689941, "global_step": 98494, "epoch": 1106, "lr": 7.105104571616202e-05}
{"train_loss": 0.13319887220859528, "global_step": 98495, "epoch": 1106, "lr": 7.105051986412174e-05}
{"train_loss": 0.14470747113227844, "global_step": 98496, "epoch": 1106, "lr": 7.104999400925149e-05}
{"train_loss": 0.1653529703617096, "global_step": 98497, "epoch": 1106, "lr": 7.104946815155131e-05}
{"train_loss": 0.1103309765458107, "global_step": 98498, "epoch": 1106, "lr": 7.10489422910213e-05}
{"train_loss": 0.1763400137424469, "global_step": 98499, "epoch": 1106, "lr": 7.10484164276615e-05}
{"train_loss": 0.09834396839141846, "global_step": 98500, "epoch": 1106, "lr": 7.1047890561472e-05}
{"train_loss": 0.1408321112394333, "global_step": 98501, "epoch": 1106, "lr": 7.104736469245287e-05}
{"train_loss": 0.19764618575572968, "global_step": 98502, "epoch": 1106, "lr": 7.104683882060419e-05}
{"train_loss": 0.11377755552530289, "global_step": 98503, "epoch": 1106, "lr": 7.1046312945926e-05}
{"train_loss": 0.17083802819252014, "global_step": 98504, "epoch": 1106, "lr": 7.10457870684184e-05}
{"train_loss": 0.15743333101272583, "global_step": 98505, "epoch": 1106, "lr": 7.104526118808145e-05}
{"train_loss": 0.15024837851524353, "global_step": 98506, "epoch": 1106, "lr": 7.104473530491523e-05}
{"train_loss": 0.11570406705141068, "global_step": 98507, "epoch": 1106, "lr": 7.104420941891978e-05}
{"train_loss": 0.19226709008216858, "global_step": 98508, "epoch": 1106, "lr": 7.104368353009521e-05}
{"train_loss": 0.179100900888443, "global_step": 98509, "epoch": 1106, "lr": 7.104315763844156e-05}
{"train_loss": 0.13585016131401062, "global_step": 98510, "epoch": 1106, "lr": 7.104263174395892e-05}
{"train_loss": 0.13234783709049225, "global_step": 98511, "epoch": 1106, "lr": 7.104210584664735e-05}
{"train_loss": 0.1175190657377243, "global_step": 98512, "epoch": 1106, "lr": 7.104157994650694e-05}
{"train_loss": 0.1405036896467209, "global_step": 98513, "epoch": 1106, "lr": 7.104105404353774e-05}
{"train_loss": 0.20529918372631073, "global_step": 98514, "epoch": 1106, "lr": 7.104052813773982e-05}
{"train_loss": 0.2241157591342926, "global_step": 98515, "epoch": 1106, "lr": 7.104000222911327e-05}
{"train_loss": 0.253813236951828, "global_step": 98516, "epoch": 1106, "lr": 7.103947631765812e-05}
{"train_loss": 0.17296470701694489, "global_step": 98517, "epoch": 1106, "lr": 7.103895040337451e-05}
{"train_loss": 0.15162286162376404, "global_step": 98518, "epoch": 1106, "lr": 7.103842448626243e-05}
{"train_loss": 0.19363248348236084, "global_step": 98519, "epoch": 1106, "lr": 7.103789856632201e-05}
{"train_loss": 0.13404329121112823, "global_step": 98520, "epoch": 1106, "lr": 7.103737264355331e-05}
{"train_loss": 0.2489154040813446, "global_step": 98521, "epoch": 1106, "lr": 7.103684671795638e-05}
{"train_loss": 0.15941114246510388, "global_step": 98522, "epoch": 1106, "lr": 7.103632078953131e-05, "val_loss": 4.693126201629639}
{"train_loss": 0.13245771825313568, "global_step": 98523, "epoch": 1107, "lr": 7.103579485827814e-05}
{"train_loss": 0.21150511503219604, "global_step": 98524, "epoch": 1107, "lr": 7.103526892419698e-05}
{"train_loss": 0.15871717035770416, "global_step": 98525, "epoch": 1107, "lr": 7.103474298728789e-05}
{"train_loss": 0.21523252129554749, "global_step": 98526, "epoch": 1107, "lr": 7.103421704755094e-05}
{"train_loss": 0.20937448740005493, "global_step": 98527, "epoch": 1107, "lr": 7.103369110498618e-05}
{"train_loss": 0.14700572192668915, "global_step": 98528, "epoch": 1107, "lr": 7.103316515959369e-05}
{"train_loss": 0.18356826901435852, "global_step": 98529, "epoch": 1107, "lr": 7.103263921137358e-05}
{"train_loss": 0.1597120463848114, "global_step": 98530, "epoch": 1107, "lr": 7.103211326032586e-05}
{"train_loss": 0.21554860472679138, "global_step": 98531, "epoch": 1107, "lr": 7.103158730645063e-05}
{"train_loss": 0.14425019919872284, "global_step": 98532, "epoch": 1107, "lr": 7.103106134974797e-05}
{"train_loss": 0.19467365741729736, "global_step": 98533, "epoch": 1107, "lr": 7.103053539021793e-05}
{"train_loss": 0.17703528702259064, "global_step": 98534, "epoch": 1107, "lr": 7.10300094278606e-05}
{"train_loss": 0.11511287838220596, "global_step": 98535, "epoch": 1107, "lr": 7.102948346267603e-05}
{"train_loss": 0.29147833585739136, "global_step": 98536, "epoch": 1107, "lr": 7.102895749466431e-05}
{"train_loss": 0.0984344482421875, "global_step": 98537, "epoch": 1107, "lr": 7.102843152382551e-05}
{"train_loss": 0.22692713141441345, "global_step": 98538, "epoch": 1107, "lr": 7.102790555015969e-05}
{"train_loss": 0.16304843127727509, "global_step": 98539, "epoch": 1107, "lr": 7.102737957366691e-05}
{"train_loss": 0.22621558606624603, "global_step": 98540, "epoch": 1107, "lr": 7.102685359434727e-05}
{"train_loss": 0.18041755259037018, "global_step": 98541, "epoch": 1107, "lr": 7.102632761220083e-05}
{"train_loss": 0.18290139734745026, "global_step": 98542, "epoch": 1107, "lr": 7.102580162722766e-05}
{"train_loss": 0.12005458772182465, "global_step": 98543, "epoch": 1107, "lr": 7.10252756394278e-05}
{"train_loss": 0.24733483791351318, "global_step": 98544, "epoch": 1107, "lr": 7.102474964880137e-05}
{"train_loss": 0.16296829283237457, "global_step": 98545, "epoch": 1107, "lr": 7.102422365534841e-05}
{"train_loss": 0.27613168954849243, "global_step": 98546, "epoch": 1107, "lr": 7.1023697659069e-05}
{"train_loss": 0.1718156635761261, "global_step": 98547, "epoch": 1107, "lr": 7.102317165996322e-05}
{"train_loss": 0.21056634187698364, "global_step": 98548, "epoch": 1107, "lr": 7.102264565803112e-05}
{"train_loss": 0.1935996115207672, "global_step": 98549, "epoch": 1107, "lr": 7.102211965327279e-05}
{"train_loss": 0.19722965359687805, "global_step": 98550, "epoch": 1107, "lr": 7.102159364568828e-05}
{"train_loss": 0.23332256078720093, "global_step": 98551, "epoch": 1107, "lr": 7.102106763527768e-05}
{"train_loss": 0.1379089504480362, "global_step": 98552, "epoch": 1107, "lr": 7.102054162204106e-05}
{"train_loss": 0.1545373499393463, "global_step": 98553, "epoch": 1107, "lr": 7.102001560597847e-05}
{"train_loss": 0.14122149348258972, "global_step": 98554, "epoch": 1107, "lr": 7.101948958708999e-05}
{"train_loss": 0.13299740850925446, "global_step": 98555, "epoch": 1107, "lr": 7.101896356537571e-05}
{"train_loss": 0.16802650690078735, "global_step": 98556, "epoch": 1107, "lr": 7.10184375408357e-05}
{"train_loss": 0.27507537603378296, "global_step": 98557, "epoch": 1107, "lr": 7.101791151347e-05}
{"train_loss": 0.14681033790111542, "global_step": 98558, "epoch": 1107, "lr": 7.10173854832787e-05}
{"train_loss": 0.18126705288887024, "global_step": 98559, "epoch": 1107, "lr": 7.101685945026188e-05}
{"train_loss": 0.19163575768470764, "global_step": 98560, "epoch": 1107, "lr": 7.101633341441958e-05}
{"train_loss": 0.19580014050006866, "global_step": 98561, "epoch": 1107, "lr": 7.101580737575189e-05}
{"train_loss": 0.1755121946334839, "global_step": 98562, "epoch": 1107, "lr": 7.10152813342589e-05}
{"train_loss": 0.14945940673351288, "global_step": 98563, "epoch": 1107, "lr": 7.101475528994065e-05}
{"train_loss": 0.10267447680234909, "global_step": 98564, "epoch": 1107, "lr": 7.101422924279723e-05}
{"train_loss": 0.13293959200382233, "global_step": 98565, "epoch": 1107, "lr": 7.101370319282869e-05}
{"train_loss": 0.21731503307819366, "global_step": 98566, "epoch": 1107, "lr": 7.101317714003512e-05}
{"train_loss": 0.19267837703227997, "global_step": 98567, "epoch": 1107, "lr": 7.101265108441658e-05}
{"train_loss": 0.10713399946689606, "global_step": 98568, "epoch": 1107, "lr": 7.101212502597315e-05}
{"train_loss": 0.11582377552986145, "global_step": 98569, "epoch": 1107, "lr": 7.10115989647049e-05}
{"train_loss": 0.21077916026115417, "global_step": 98570, "epoch": 1107, "lr": 7.101107290061189e-05}
{"train_loss": 0.14730523526668549, "global_step": 98571, "epoch": 1107, "lr": 7.10105468336942e-05}
{"train_loss": 0.1303214728832245, "global_step": 98572, "epoch": 1107, "lr": 7.101002076395189e-05}
{"train_loss": 0.11936325579881668, "global_step": 98573, "epoch": 1107, "lr": 7.100949469138505e-05}
{"train_loss": 0.22982436418533325, "global_step": 98574, "epoch": 1107, "lr": 7.100896861599374e-05}
{"train_loss": 0.14787031710147858, "global_step": 98575, "epoch": 1107, "lr": 7.100844253777802e-05}
{"train_loss": 0.17888370156288147, "global_step": 98576, "epoch": 1107, "lr": 7.100791645673799e-05}
{"train_loss": 0.19071638584136963, "global_step": 98577, "epoch": 1107, "lr": 7.100739037287368e-05}
{"train_loss": 0.15684182941913605, "global_step": 98578, "epoch": 1107, "lr": 7.100686428618519e-05}
{"train_loss": 0.19829072058200836, "global_step": 98579, "epoch": 1107, "lr": 7.100633819667259e-05}
{"train_loss": 0.22400015592575073, "global_step": 98580, "epoch": 1107, "lr": 7.100581210433594e-05}
{"train_loss": 0.20897027850151062, "global_step": 98581, "epoch": 1107, "lr": 7.10052860091753e-05}
{"train_loss": 0.18419073522090912, "global_step": 98582, "epoch": 1107, "lr": 7.100475991119078e-05}
{"train_loss": 0.17592284083366394, "global_step": 98583, "epoch": 1107, "lr": 7.100423381038242e-05}
{"train_loss": 0.15135373175144196, "global_step": 98584, "epoch": 1107, "lr": 7.100370770675028e-05}
{"train_loss": 0.17094840109348297, "global_step": 98585, "epoch": 1107, "lr": 7.100318160029447e-05}
{"train_loss": 0.11814247816801071, "global_step": 98586, "epoch": 1107, "lr": 7.100265549101502e-05}
{"train_loss": 0.14261901378631592, "global_step": 98587, "epoch": 1107, "lr": 7.100212937891203e-05}
{"train_loss": 0.19073061645030975, "global_step": 98588, "epoch": 1107, "lr": 7.100160326398555e-05}
{"train_loss": 0.12261532247066498, "global_step": 98589, "epoch": 1107, "lr": 7.100107714623567e-05}
{"train_loss": 0.16191180050373077, "global_step": 98590, "epoch": 1107, "lr": 7.100055102566244e-05}
{"train_loss": 0.13629794120788574, "global_step": 98591, "epoch": 1107, "lr": 7.100002490226594e-05}
{"train_loss": 0.18748065829277039, "global_step": 98592, "epoch": 1107, "lr": 7.099949877604626e-05}
{"train_loss": 0.22914661467075348, "global_step": 98593, "epoch": 1107, "lr": 7.099897264700344e-05}
{"train_loss": 0.258987158536911, "global_step": 98594, "epoch": 1107, "lr": 7.099844651513757e-05}
{"train_loss": 0.1760643720626831, "global_step": 98595, "epoch": 1107, "lr": 7.099792038044872e-05}
{"train_loss": 0.102462537586689, "global_step": 98596, "epoch": 1107, "lr": 7.099739424293695e-05}
{"train_loss": 0.17026008665561676, "global_step": 98597, "epoch": 1107, "lr": 7.099686810260233e-05}
{"train_loss": 0.18777132034301758, "global_step": 98598, "epoch": 1107, "lr": 7.099634195944494e-05}
{"train_loss": 0.12494391202926636, "global_step": 98599, "epoch": 1107, "lr": 7.099581581346486e-05}
{"train_loss": 0.16883321106433868, "global_step": 98600, "epoch": 1107, "lr": 7.099528966466214e-05}
{"train_loss": 0.12622323632240295, "global_step": 98601, "epoch": 1107, "lr": 7.099476351303686e-05}
{"train_loss": 0.12166427075862885, "global_step": 98602, "epoch": 1107, "lr": 7.099423735858907e-05}
{"train_loss": 0.20647555589675903, "global_step": 98603, "epoch": 1107, "lr": 7.09937112013189e-05}
{"train_loss": 0.1426863670349121, "global_step": 98604, "epoch": 1107, "lr": 7.099318504122635e-05}
{"train_loss": 0.08041511476039886, "global_step": 98605, "epoch": 1107, "lr": 7.099265887831153e-05}
{"train_loss": 0.14890511333942413, "global_step": 98606, "epoch": 1107, "lr": 7.09921327125745e-05}
{"train_loss": 0.18186482787132263, "global_step": 98607, "epoch": 1107, "lr": 7.099160654401534e-05}
{"train_loss": 0.14162783324718475, "global_step": 98608, "epoch": 1107, "lr": 7.099108037263411e-05}
{"train_loss": 0.16130082309246063, "global_step": 98609, "epoch": 1107, "lr": 7.099055419843088e-05}
{"train_loss": 0.15239383280277252, "global_step": 98610, "epoch": 1107, "lr": 7.099002802140574e-05}
{"train_loss": 0.1714736068349206, "global_step": 98611, "epoch": 1107, "lr": 7.098950184155874e-05, "val_loss": 4.6352314949035645}
{"train_loss": 0.16221895813941956, "global_step": 98612, "epoch": 1108, "lr": 7.098897565888996e-05}
{"train_loss": 0.23075690865516663, "global_step": 98613, "epoch": 1108, "lr": 7.098844947339946e-05}
{"train_loss": 0.22712481021881104, "global_step": 98614, "epoch": 1108, "lr": 7.098792328508734e-05}
{"train_loss": 0.13976828753948212, "global_step": 98615, "epoch": 1108, "lr": 7.098739709395363e-05}
{"train_loss": 0.16355448961257935, "global_step": 98616, "epoch": 1108, "lr": 7.098687089999843e-05}
{"train_loss": 0.17351624369621277, "global_step": 98617, "epoch": 1108, "lr": 7.098634470322179e-05}
{"train_loss": 0.1610715687274933, "global_step": 98618, "epoch": 1108, "lr": 7.09858185036238e-05}
{"train_loss": 0.13873136043548584, "global_step": 98619, "epoch": 1108, "lr": 7.098529230120452e-05}
{"train_loss": 0.13055159151554108, "global_step": 98620, "epoch": 1108, "lr": 7.098476609596402e-05}
{"train_loss": 0.17092351615428925, "global_step": 98621, "epoch": 1108, "lr": 7.098423988790237e-05}
{"train_loss": 0.15117117762565613, "global_step": 98622, "epoch": 1108, "lr": 7.098371367701966e-05}
{"train_loss": 0.12856285274028778, "global_step": 98623, "epoch": 1108, "lr": 7.098318746331594e-05}
{"train_loss": 0.21535994112491608, "global_step": 98624, "epoch": 1108, "lr": 7.098266124679127e-05}
{"train_loss": 0.13575948774814606, "global_step": 98625, "epoch": 1108, "lr": 7.098213502744575e-05}
{"train_loss": 0.2623612582683563, "global_step": 98626, "epoch": 1108, "lr": 7.098160880527944e-05}
{"train_loss": 0.16482283174991608, "global_step": 98627, "epoch": 1108, "lr": 7.098108258029241e-05}
{"train_loss": 0.18336991965770721, "global_step": 98628, "epoch": 1108, "lr": 7.098055635248471e-05}
{"train_loss": 0.10190679132938385, "global_step": 98629, "epoch": 1108, "lr": 7.098003012185645e-05}
{"train_loss": 0.136435404419899, "global_step": 98630, "epoch": 1108, "lr": 7.097950388840768e-05}
{"train_loss": 0.08080528676509857, "global_step": 98631, "epoch": 1108, "lr": 7.097897765213847e-05}
{"train_loss": 0.20802052319049835, "global_step": 98632, "epoch": 1108, "lr": 7.097845141304889e-05}
{"train_loss": 0.12366078048944473, "global_step": 98633, "epoch": 1108, "lr": 7.097792517113902e-05}
{"train_loss": 0.1300060749053955, "global_step": 98634, "epoch": 1108, "lr": 7.097739892640891e-05}
{"train_loss": 0.21306227147579193, "global_step": 98635, "epoch": 1108, "lr": 7.097687267885865e-05}
{"train_loss": 0.20546308159828186, "global_step": 98636, "epoch": 1108, "lr": 7.09763464284883e-05}
{"train_loss": 0.13605597615242004, "global_step": 98637, "epoch": 1108, "lr": 7.097582017529795e-05}
{"train_loss": 0.12395210564136505, "global_step": 98638, "epoch": 1108, "lr": 7.097529391928765e-05}
{"train_loss": 0.19570013880729675, "global_step": 98639, "epoch": 1108, "lr": 7.097476766045747e-05}
{"train_loss": 0.21018821001052856, "global_step": 98640, "epoch": 1108, "lr": 7.09742413988075e-05}
{"train_loss": 0.16640540957450867, "global_step": 98641, "epoch": 1108, "lr": 7.097371513433779e-05}
{"train_loss": 0.16273687779903412, "global_step": 98642, "epoch": 1108, "lr": 7.097318886704842e-05}
{"train_loss": 0.18244872987270355, "global_step": 98643, "epoch": 1108, "lr": 7.097266259693947e-05}
{"train_loss": 0.13621532917022705, "global_step": 98644, "epoch": 1108, "lr": 7.097213632401098e-05}
{"train_loss": 0.19406673312187195, "global_step": 98645, "epoch": 1108, "lr": 7.097161004826305e-05}
{"train_loss": 0.2633160352706909, "global_step": 98646, "epoch": 1108, "lr": 7.097108376969575e-05}
{"train_loss": 0.20830555260181427, "global_step": 98647, "epoch": 1108, "lr": 7.097055748830914e-05}
{"train_loss": 0.19910240173339844, "global_step": 98648, "epoch": 1108, "lr": 7.09700312041033e-05}
{"train_loss": 0.18190236389636993, "global_step": 98649, "epoch": 1108, "lr": 7.096950491707829e-05}
{"train_loss": 0.10881365090608597, "global_step": 98650, "epoch": 1108, "lr": 7.096897862723418e-05}
{"train_loss": 0.16076549887657166, "global_step": 98651, "epoch": 1108, "lr": 7.096845233457106e-05}
{"train_loss": 0.14623583853244781, "global_step": 98652, "epoch": 1108, "lr": 7.096792603908897e-05}
{"train_loss": 0.14383217692375183, "global_step": 98653, "epoch": 1108, "lr": 7.096739974078802e-05}
{"train_loss": 0.15749871730804443, "global_step": 98654, "epoch": 1108, "lr": 7.096687343966825e-05}
{"train_loss": 0.15344730019569397, "global_step": 98655, "epoch": 1108, "lr": 7.096634713572973e-05}
{"train_loss": 0.1511216163635254, "global_step": 98656, "epoch": 1108, "lr": 7.096582082897255e-05}
{"train_loss": 0.15758903324604034, "global_step": 98657, "epoch": 1108, "lr": 7.096529451939677e-05}
{"train_loss": 0.1355116069316864, "global_step": 98658, "epoch": 1108, "lr": 7.096476820700245e-05}
{"train_loss": 0.18960259854793549, "global_step": 98659, "epoch": 1108, "lr": 7.096424189178969e-05}
{"train_loss": 0.1536448895931244, "global_step": 98660, "epoch": 1108, "lr": 7.096371557375852e-05}
{"train_loss": 0.1683344542980194, "global_step": 98661, "epoch": 1108, "lr": 7.096318925290906e-05}
{"train_loss": 0.12118671089410782, "global_step": 98662, "epoch": 1108, "lr": 7.096266292924134e-05}
{"train_loss": 0.19547337293624878, "global_step": 98663, "epoch": 1108, "lr": 7.096213660275544e-05}
{"train_loss": 0.1479923278093338, "global_step": 98664, "epoch": 1108, "lr": 7.096161027345144e-05}
{"train_loss": 0.17290939390659332, "global_step": 98665, "epoch": 1108, "lr": 7.09610839413294e-05}
{"train_loss": 0.1373288333415985, "global_step": 98666, "epoch": 1108, "lr": 7.096055760638942e-05}
{"train_loss": 0.21569542586803436, "global_step": 98667, "epoch": 1108, "lr": 7.096003126863155e-05}
{"train_loss": 0.15180423855781555, "global_step": 98668, "epoch": 1108, "lr": 7.095950492805582e-05}
{"train_loss": 0.15201735496520996, "global_step": 98669, "epoch": 1108, "lr": 7.095897858466238e-05}
{"train_loss": 0.13933679461479187, "global_step": 98670, "epoch": 1108, "lr": 7.095845223845126e-05}
{"train_loss": 0.2498789131641388, "global_step": 98671, "epoch": 1108, "lr": 7.095792588942251e-05}
{"train_loss": 0.14354917407035828, "global_step": 98672, "epoch": 1108, "lr": 7.095739953757625e-05}
{"train_loss": 0.13248945772647858, "global_step": 98673, "epoch": 1108, "lr": 7.09568731829125e-05}
{"train_loss": 0.16668173670768738, "global_step": 98674, "epoch": 1108, "lr": 7.095634682543135e-05}
{"train_loss": 0.1335587203502655, "global_step": 98675, "epoch": 1108, "lr": 7.09558204651329e-05}
{"train_loss": 0.16361382603645325, "global_step": 98676, "epoch": 1108, "lr": 7.095529410201717e-05}
{"train_loss": 0.15127842128276825, "global_step": 98677, "epoch": 1108, "lr": 7.095476773608426e-05}
{"train_loss": 0.14896419644355774, "global_step": 98678, "epoch": 1108, "lr": 7.095424136733424e-05}
{"train_loss": 0.1064324602484703, "global_step": 98679, "epoch": 1108, "lr": 7.095371499576718e-05}
{"train_loss": 0.17155864834785461, "global_step": 98680, "epoch": 1108, "lr": 7.095318862138314e-05}
{"train_loss": 0.17982080578804016, "global_step": 98681, "epoch": 1108, "lr": 7.09526622441822e-05}
{"train_loss": 0.11527934670448303, "global_step": 98682, "epoch": 1108, "lr": 7.095213586416445e-05}
{"train_loss": 0.18619763851165771, "global_step": 98683, "epoch": 1108, "lr": 7.095160948132991e-05}
{"train_loss": 0.14400005340576172, "global_step": 98684, "epoch": 1108, "lr": 7.09510830956787e-05}
{"train_loss": 0.16454899311065674, "global_step": 98685, "epoch": 1108, "lr": 7.095055670721088e-05}
{"train_loss": 0.15131348371505737, "global_step": 98686, "epoch": 1108, "lr": 7.09500303159265e-05}
{"train_loss": 0.2029125839471817, "global_step": 98687, "epoch": 1108, "lr": 7.094950392182564e-05}
{"train_loss": 0.11489345878362656, "global_step": 98688, "epoch": 1108, "lr": 7.094897752490839e-05}
{"train_loss": 0.1821955442428589, "global_step": 98689, "epoch": 1108, "lr": 7.09484511251748e-05}
{"train_loss": 0.10664808005094528, "global_step": 98690, "epoch": 1108, "lr": 7.094792472262495e-05}
{"train_loss": 0.17534750699996948, "global_step": 98691, "epoch": 1108, "lr": 7.09473983172589e-05}
{"train_loss": 0.10126881301403046, "global_step": 98692, "epoch": 1108, "lr": 7.094687190907673e-05}
{"train_loss": 0.1425236463546753, "global_step": 98693, "epoch": 1108, "lr": 7.09463454980785e-05}
{"train_loss": 0.19720883667469025, "global_step": 98694, "epoch": 1108, "lr": 7.094581908426431e-05}
{"train_loss": 0.17340482771396637, "global_step": 98695, "epoch": 1108, "lr": 7.094529266763419e-05}
{"train_loss": 0.1599714607000351, "global_step": 98696, "epoch": 1108, "lr": 7.094476624818824e-05}
{"train_loss": 0.21605996787548065, "global_step": 98697, "epoch": 1108, "lr": 7.094423982592652e-05}
{"train_loss": 0.2274259328842163, "global_step": 98698, "epoch": 1108, "lr": 7.094371340084908e-05}
{"train_loss": 0.07924755662679672, "global_step": 98699, "epoch": 1108, "lr": 7.094318697295603e-05}
{"train_loss": 0.16328483024674856, "global_step": 98700, "epoch": 1108, "lr": 7.094266054224743e-05, "val_loss": 4.703190326690674}
{"train_loss": 0.1388903260231018, "global_step": 98701, "epoch": 1109, "lr": 7.094213410872335e-05}
{"train_loss": 0.1457182615995407, "global_step": 98702, "epoch": 1109, "lr": 7.094160767238384e-05}
{"train_loss": 0.173453688621521, "global_step": 98703, "epoch": 1109, "lr": 7.094108123322899e-05}
{"train_loss": 0.22150921821594238, "global_step": 98704, "epoch": 1109, "lr": 7.094055479125887e-05}
{"train_loss": 0.18458032608032227, "global_step": 98705, "epoch": 1109, "lr": 7.094002834647355e-05}
{"train_loss": 0.25016921758651733, "global_step": 98706, "epoch": 1109, "lr": 7.09395018988731e-05}
{"train_loss": 0.17575037479400635, "global_step": 98707, "epoch": 1109, "lr": 7.093897544845758e-05}
{"train_loss": 0.13318227231502533, "global_step": 98708, "epoch": 1109, "lr": 7.093844899522708e-05}
{"train_loss": 0.1734902262687683, "global_step": 98709, "epoch": 1109, "lr": 7.093792253918166e-05}
{"train_loss": 0.16549932956695557, "global_step": 98710, "epoch": 1109, "lr": 7.093739608032137e-05}
{"train_loss": 0.2148059606552124, "global_step": 98711, "epoch": 1109, "lr": 7.093686961864634e-05}
{"train_loss": 0.17667622864246368, "global_step": 98712, "epoch": 1109, "lr": 7.093634315415658e-05}
{"train_loss": 0.2054247111082077, "global_step": 98713, "epoch": 1109, "lr": 7.093581668685219e-05}
{"train_loss": 0.15587401390075684, "global_step": 98714, "epoch": 1109, "lr": 7.093529021673322e-05}
{"train_loss": 0.20620043575763702, "global_step": 98715, "epoch": 1109, "lr": 7.093476374379977e-05}
{"train_loss": 0.12489741295576096, "global_step": 98716, "epoch": 1109, "lr": 7.09342372680519e-05}
{"train_loss": 0.1425902098417282, "global_step": 98717, "epoch": 1109, "lr": 7.093371078948966e-05}
{"train_loss": 0.21029484272003174, "global_step": 98718, "epoch": 1109, "lr": 7.093318430811314e-05}
{"train_loss": 0.14249934256076813, "global_step": 98719, "epoch": 1109, "lr": 7.093265782392243e-05}
{"train_loss": 0.16170471906661987, "global_step": 98720, "epoch": 1109, "lr": 7.093213133691756e-05}
{"train_loss": 0.20094318687915802, "global_step": 98721, "epoch": 1109, "lr": 7.093160484709862e-05}
{"train_loss": 0.187827005982399, "global_step": 98722, "epoch": 1109, "lr": 7.093107835446569e-05}
{"train_loss": 0.11608389765024185, "global_step": 98723, "epoch": 1109, "lr": 7.093055185901882e-05}
{"train_loss": 0.1471146047115326, "global_step": 98724, "epoch": 1109, "lr": 7.09300253607581e-05}
{"train_loss": 0.20753517746925354, "global_step": 98725, "epoch": 1109, "lr": 7.092949885968358e-05}
{"train_loss": 0.2482364922761917, "global_step": 98726, "epoch": 1109, "lr": 7.092897235579536e-05}
{"train_loss": 0.15326926112174988, "global_step": 98727, "epoch": 1109, "lr": 7.092844584909351e-05}
{"train_loss": 0.2795463502407074, "global_step": 98728, "epoch": 1109, "lr": 7.092791933957805e-05}
{"train_loss": 0.27652236819267273, "global_step": 98729, "epoch": 1109, "lr": 7.092739282724912e-05}
{"train_loss": 0.18088893592357635, "global_step": 98730, "epoch": 1109, "lr": 7.092686631210675e-05}
{"train_loss": 0.184878408908844, "global_step": 98731, "epoch": 1109, "lr": 7.092633979415099e-05}
{"train_loss": 0.1947045773267746, "global_step": 98732, "epoch": 1109, "lr": 7.092581327338196e-05}
{"train_loss": 0.19309955835342407, "global_step": 98733, "epoch": 1109, "lr": 7.092528674979972e-05}
{"train_loss": 0.15104660391807556, "global_step": 98734, "epoch": 1109, "lr": 7.09247602234043e-05}
{"train_loss": 0.22075486183166504, "global_step": 98735, "epoch": 1109, "lr": 7.092423369419583e-05}
{"train_loss": 0.18870878219604492, "global_step": 98736, "epoch": 1109, "lr": 7.092370716217433e-05}
{"train_loss": 0.14784586429595947, "global_step": 98737, "epoch": 1109, "lr": 7.09231806273399e-05}
{"train_loss": 0.1550980806350708, "global_step": 98738, "epoch": 1109, "lr": 7.09226540896926e-05}
{"train_loss": 0.15166248381137848, "global_step": 98739, "epoch": 1109, "lr": 7.092212754923252e-05}
{"train_loss": 0.1615031361579895, "global_step": 98740, "epoch": 1109, "lr": 7.09216010059597e-05}
{"train_loss": 0.21352531015872955, "global_step": 98741, "epoch": 1109, "lr": 7.092107445987424e-05}
{"train_loss": 0.2072899490594864, "global_step": 98742, "epoch": 1109, "lr": 7.092054791097617e-05}
{"train_loss": 0.15386275947093964, "global_step": 98743, "epoch": 1109, "lr": 7.092002135926561e-05}
{"train_loss": 0.22100630402565002, "global_step": 98744, "epoch": 1109, "lr": 7.09194948047426e-05}
{"train_loss": 0.17588208615779877, "global_step": 98745, "epoch": 1109, "lr": 7.091896824740722e-05}
{"train_loss": 0.15645910799503326, "global_step": 98746, "epoch": 1109, "lr": 7.091844168725955e-05}
{"train_loss": 0.20687805116176605, "global_step": 98747, "epoch": 1109, "lr": 7.091791512429963e-05}
{"train_loss": 0.1421125829219818, "global_step": 98748, "epoch": 1109, "lr": 7.091738855852757e-05}
{"train_loss": 0.22594189643859863, "global_step": 98749, "epoch": 1109, "lr": 7.091686198994342e-05}
{"train_loss": 0.25457072257995605, "global_step": 98750, "epoch": 1109, "lr": 7.091633541854725e-05}
{"train_loss": 0.18408873677253723, "global_step": 98751, "epoch": 1109, "lr": 7.091580884433914e-05}
{"train_loss": 0.14592987298965454, "global_step": 98752, "epoch": 1109, "lr": 7.091528226731913e-05}
{"train_loss": 0.14272616803646088, "global_step": 98753, "epoch": 1109, "lr": 7.091475568748735e-05}
{"train_loss": 0.10768335312604904, "global_step": 98754, "epoch": 1109, "lr": 7.09142291048438e-05}
{"train_loss": 0.22446458041667938, "global_step": 98755, "epoch": 1109, "lr": 7.09137025193886e-05}
{"train_loss": 0.12853197753429413, "global_step": 98756, "epoch": 1109, "lr": 7.09131759311218e-05}
{"train_loss": 0.12264375388622284, "global_step": 98757, "epoch": 1109, "lr": 7.091264934004351e-05}
{"train_loss": 0.15582939982414246, "global_step": 98758, "epoch": 1109, "lr": 7.091212274615374e-05}
{"train_loss": 0.09767094254493713, "global_step": 98759, "epoch": 1109, "lr": 7.091159614945261e-05}
{"train_loss": 0.11798115074634552, "global_step": 98760, "epoch": 1109, "lr": 7.091106954994016e-05}
{"train_loss": 0.2304946929216385, "global_step": 98761, "epoch": 1109, "lr": 7.091054294761647e-05}
{"train_loss": 0.12696386873722076, "global_step": 98762, "epoch": 1109, "lr": 7.09100163424816e-05}
{"train_loss": 0.16823197901248932, "global_step": 98763, "epoch": 1109, "lr": 7.090948973453566e-05}
{"train_loss": 0.1199955865740776, "global_step": 98764, "epoch": 1109, "lr": 7.090896312377868e-05}
{"train_loss": 0.11413919180631638, "global_step": 98765, "epoch": 1109, "lr": 7.090843651021074e-05}
{"train_loss": 0.1451348513364792, "global_step": 98766, "epoch": 1109, "lr": 7.090790989383193e-05}
{"train_loss": 0.1267506629228592, "global_step": 98767, "epoch": 1109, "lr": 7.090738327464229e-05}
{"train_loss": 0.14718174934387207, "global_step": 98768, "epoch": 1109, "lr": 7.090685665264192e-05}
{"train_loss": 0.19071316719055176, "global_step": 98769, "epoch": 1109, "lr": 7.090633002783088e-05}
{"train_loss": 0.15629202127456665, "global_step": 98770, "epoch": 1109, "lr": 7.090580340020922e-05}
{"train_loss": 0.1902099996805191, "global_step": 98771, "epoch": 1109, "lr": 7.090527676977705e-05}
{"train_loss": 0.194920614361763, "global_step": 98772, "epoch": 1109, "lr": 7.09047501365344e-05}
{"train_loss": 0.17146527767181396, "global_step": 98773, "epoch": 1109, "lr": 7.090422350048137e-05}
{"train_loss": 0.14692218601703644, "global_step": 98774, "epoch": 1109, "lr": 7.090369686161803e-05}
{"train_loss": 0.2115810662508011, "global_step": 98775, "epoch": 1109, "lr": 7.090317021994443e-05}
{"train_loss": 0.18594683706760406, "global_step": 98776, "epoch": 1109, "lr": 7.090264357546067e-05}
{"train_loss": 0.1546999216079712, "global_step": 98777, "epoch": 1109, "lr": 7.09021169281668e-05}
{"train_loss": 0.16207894682884216, "global_step": 98778, "epoch": 1109, "lr": 7.090159027806288e-05}
{"train_loss": 0.1740161031484604, "global_step": 98779, "epoch": 1109, "lr": 7.090106362514902e-05}
{"train_loss": 0.1906810998916626, "global_step": 98780, "epoch": 1109, "lr": 7.090053696942525e-05}
{"train_loss": 0.1683799773454666, "global_step": 98781, "epoch": 1109, "lr": 7.090001031089166e-05}
{"train_loss": 0.09104149788618088, "global_step": 98782, "epoch": 1109, "lr": 7.089948364954831e-05}
{"train_loss": 0.1556418240070343, "global_step": 98783, "epoch": 1109, "lr": 7.089895698539528e-05}
{"train_loss": 0.09602080285549164, "global_step": 98784, "epoch": 1109, "lr": 7.089843031843265e-05}
{"train_loss": 0.1143939420580864, "global_step": 98785, "epoch": 1109, "lr": 7.089790364866048e-05}
{"train_loss": 0.1988227218389511, "global_step": 98786, "epoch": 1109, "lr": 7.089737697607885e-05}
{"train_loss": 0.19595962762832642, "global_step": 98787, "epoch": 1109, "lr": 7.08968503006878e-05}
{"train_loss": 0.19011946022510529, "global_step": 98788, "epoch": 1109, "lr": 7.089632362248744e-05}
{"train_loss": 0.1719907821061906, "global_step": 98789, "epoch": 1109, "lr": 7.089579694147783e-05, "val_loss": 4.705111980438232}
{"train_loss": 0.10332850366830826, "global_step": 98790, "epoch": 1110, "lr": 7.0895270257659e-05}
{"train_loss": 0.24158638715744019, "global_step": 98791, "epoch": 1110, "lr": 7.08947435710311e-05}
{"train_loss": 0.18001465499401093, "global_step": 98792, "epoch": 1110, "lr": 7.089421688159413e-05}
{"train_loss": 0.13823959231376648, "global_step": 98793, "epoch": 1110, "lr": 7.089369018934819e-05}
{"train_loss": 0.10355187207460403, "global_step": 98794, "epoch": 1110, "lr": 7.089316349429335e-05}
{"train_loss": 0.22271466255187988, "global_step": 98795, "epoch": 1110, "lr": 7.089263679642969e-05}
{"train_loss": 0.17023633420467377, "global_step": 98796, "epoch": 1110, "lr": 7.089211009575726e-05}
{"train_loss": 0.16284224390983582, "global_step": 98797, "epoch": 1110, "lr": 7.089158339227613e-05}
{"train_loss": 0.11008797585964203, "global_step": 98798, "epoch": 1110, "lr": 7.08910566859864e-05}
{"train_loss": 0.17642365396022797, "global_step": 98799, "epoch": 1110, "lr": 7.08905299768881e-05}
{"train_loss": 0.1444888859987259, "global_step": 98800, "epoch": 1110, "lr": 7.089000326498135e-05}
{"train_loss": 0.1510729342699051, "global_step": 98801, "epoch": 1110, "lr": 7.088947655026617e-05}
{"train_loss": 0.1478099524974823, "global_step": 98802, "epoch": 1110, "lr": 7.088894983274268e-05}
{"train_loss": 0.23180486261844635, "global_step": 98803, "epoch": 1110, "lr": 7.08884231124109e-05}
{"train_loss": 0.13986842334270477, "global_step": 98804, "epoch": 1110, "lr": 7.088789638927093e-05}
{"train_loss": 0.13121016323566437, "global_step": 98805, "epoch": 1110, "lr": 7.088736966332286e-05}
{"train_loss": 0.15900211036205292, "global_step": 98806, "epoch": 1110, "lr": 7.088684293456671e-05}
{"train_loss": 0.20320716500282288, "global_step": 98807, "epoch": 1110, "lr": 7.08863162030026e-05}
{"train_loss": 0.09852004796266556, "global_step": 98808, "epoch": 1110, "lr": 7.088578946863056e-05}
{"train_loss": 0.16942189633846283, "global_step": 98809, "epoch": 1110, "lr": 7.088526273145069e-05}
{"train_loss": 0.11103496700525284, "global_step": 98810, "epoch": 1110, "lr": 7.088473599146304e-05}
{"train_loss": 0.1297272890806198, "global_step": 98811, "epoch": 1110, "lr": 7.08842092486677e-05}
{"train_loss": 0.15707844495773315, "global_step": 98812, "epoch": 1110, "lr": 7.088368250306474e-05}
{"train_loss": 0.15740302205085754, "global_step": 98813, "epoch": 1110, "lr": 7.088315575465423e-05}
{"train_loss": 0.13763722777366638, "global_step": 98814, "epoch": 1110, "lr": 7.088262900343621e-05}
{"train_loss": 0.17789043486118317, "global_step": 98815, "epoch": 1110, "lr": 7.088210224941079e-05}
{"train_loss": 0.11807382106781006, "global_step": 98816, "epoch": 1110, "lr": 7.088157549257802e-05}
{"train_loss": 0.22015076875686646, "global_step": 98817, "epoch": 1110, "lr": 7.088104873293799e-05}
{"train_loss": 0.0785539448261261, "global_step": 98818, "epoch": 1110, "lr": 7.088052197049074e-05}
{"train_loss": 0.21846337616443634, "global_step": 98819, "epoch": 1110, "lr": 7.087999520523637e-05}
{"train_loss": 0.15136322379112244, "global_step": 98820, "epoch": 1110, "lr": 7.087946843717494e-05}
{"train_loss": 0.264078825712204, "global_step": 98821, "epoch": 1110, "lr": 7.08789416663065e-05}
{"train_loss": 0.1884424239397049, "global_step": 98822, "epoch": 1110, "lr": 7.087841489263116e-05}
{"train_loss": 0.15692907571792603, "global_step": 98823, "epoch": 1110, "lr": 7.087788811614897e-05}
{"train_loss": 0.15700355172157288, "global_step": 98824, "epoch": 1110, "lr": 7.087736133686e-05}
{"train_loss": 0.1276097297668457, "global_step": 98825, "epoch": 1110, "lr": 7.087683455476433e-05}
{"train_loss": 0.17729321122169495, "global_step": 98826, "epoch": 1110, "lr": 7.087630776986201e-05}
{"train_loss": 0.15192574262619019, "global_step": 98827, "epoch": 1110, "lr": 7.087578098215312e-05}
{"train_loss": 0.2401386946439743, "global_step": 98828, "epoch": 1110, "lr": 7.087525419163775e-05}
{"train_loss": 0.15712152421474457, "global_step": 98829, "epoch": 1110, "lr": 7.087472739831595e-05}
{"train_loss": 0.21044081449508667, "global_step": 98830, "epoch": 1110, "lr": 7.087420060218781e-05}
{"train_loss": 0.14664055407047272, "global_step": 98831, "epoch": 1110, "lr": 7.087367380325338e-05}
{"train_loss": 0.2143641859292984, "global_step": 98832, "epoch": 1110, "lr": 7.087314700151274e-05}
{"train_loss": 0.1296917200088501, "global_step": 98833, "epoch": 1110, "lr": 7.087262019696595e-05}
{"train_loss": 0.2222689539194107, "global_step": 98834, "epoch": 1110, "lr": 7.087209338961311e-05}
{"train_loss": 0.18144236505031586, "global_step": 98835, "epoch": 1110, "lr": 7.087156657945426e-05}
{"train_loss": 0.19180263578891754, "global_step": 98836, "epoch": 1110, "lr": 7.087103976648948e-05}
{"train_loss": 0.13849550485610962, "global_step": 98837, "epoch": 1110, "lr": 7.087051295071884e-05}
{"train_loss": 0.17086325585842133, "global_step": 98838, "epoch": 1110, "lr": 7.086998613214242e-05}
{"train_loss": 0.11782795190811157, "global_step": 98839, "epoch": 1110, "lr": 7.086945931076027e-05}
{"train_loss": 0.16854824125766754, "global_step": 98840, "epoch": 1110, "lr": 7.086893248657251e-05}
{"train_loss": 0.13678833842277527, "global_step": 98841, "epoch": 1110, "lr": 7.086840565957916e-05}
{"train_loss": 0.1424986869096756, "global_step": 98842, "epoch": 1110, "lr": 7.086787882978028e-05}
{"train_loss": 0.1719614863395691, "global_step": 98843, "epoch": 1110, "lr": 7.0867351997176e-05}
{"train_loss": 0.11035075783729553, "global_step": 98844, "epoch": 1110, "lr": 7.086682516176634e-05}
{"train_loss": 0.12189766764640808, "global_step": 98845, "epoch": 1110, "lr": 7.08662983235514e-05}
{"train_loss": 0.1603821963071823, "global_step": 98846, "epoch": 1110, "lr": 7.086577148253125e-05}
{"train_loss": 0.2305069863796234, "global_step": 98847, "epoch": 1110, "lr": 7.086524463870595e-05}
{"train_loss": 0.17529858648777008, "global_step": 98848, "epoch": 1110, "lr": 7.086471779207555e-05}
{"train_loss": 0.1702125370502472, "global_step": 98849, "epoch": 1110, "lr": 7.086419094264016e-05}
{"train_loss": 0.11083705723285675, "global_step": 98850, "epoch": 1110, "lr": 7.086366409039983e-05}
{"train_loss": 0.17203055322170258, "global_step": 98851, "epoch": 1110, "lr": 7.086313723535463e-05}
{"train_loss": 0.13278719782829285, "global_step": 98852, "epoch": 1110, "lr": 7.086261037750466e-05}
{"train_loss": 0.18114854395389557, "global_step": 98853, "epoch": 1110, "lr": 7.086208351684993e-05}
{"train_loss": 0.1418127566576004, "global_step": 98854, "epoch": 1110, "lr": 7.086155665339058e-05}
{"train_loss": 0.15890727937221527, "global_step": 98855, "epoch": 1110, "lr": 7.086102978712662e-05}
{"train_loss": 0.15950500965118408, "global_step": 98856, "epoch": 1110, "lr": 7.086050291805817e-05}
{"train_loss": 0.12684185802936554, "global_step": 98857, "epoch": 1110, "lr": 7.085997604618526e-05}
{"train_loss": 0.16508421301841736, "global_step": 98858, "epoch": 1110, "lr": 7.085944917150799e-05}
{"train_loss": 0.16900113224983215, "global_step": 98859, "epoch": 1110, "lr": 7.085892229402643e-05}
{"train_loss": 0.1781681776046753, "global_step": 98860, "epoch": 1110, "lr": 7.085839541374063e-05}
{"train_loss": 0.1367477923631668, "global_step": 98861, "epoch": 1110, "lr": 7.085786853065068e-05}
{"train_loss": 0.13884608447551727, "global_step": 98862, "epoch": 1110, "lr": 7.085734164475664e-05}
{"train_loss": 0.09758464246988297, "global_step": 98863, "epoch": 1110, "lr": 7.085681475605859e-05}
{"train_loss": 0.11603909730911255, "global_step": 98864, "epoch": 1110, "lr": 7.085628786455659e-05}
{"train_loss": 0.16076786816120148, "global_step": 98865, "epoch": 1110, "lr": 7.085576097025072e-05}
{"train_loss": 0.21076712012290955, "global_step": 98866, "epoch": 1110, "lr": 7.085523407314104e-05}
{"train_loss": 0.14666065573692322, "global_step": 98867, "epoch": 1110, "lr": 7.085470717322763e-05}
{"train_loss": 0.16247810423374176, "global_step": 98868, "epoch": 1110, "lr": 7.085418027051057e-05}
{"train_loss": 0.12165693193674088, "global_step": 98869, "epoch": 1110, "lr": 7.08536533649899e-05}
{"train_loss": 0.13900932669639587, "global_step": 98870, "epoch": 1110, "lr": 7.085312645666573e-05}
{"train_loss": 0.2305029034614563, "global_step": 98871, "epoch": 1110, "lr": 7.08525995455381e-05}
{"train_loss": 0.14832371473312378, "global_step": 98872, "epoch": 1110, "lr": 7.085207263160709e-05}
{"train_loss": 0.21572834253311157, "global_step": 98873, "epoch": 1110, "lr": 7.085154571487278e-05}
{"train_loss": 0.22476038336753845, "global_step": 98874, "epoch": 1110, "lr": 7.085101879533523e-05}
{"train_loss": 0.14911745488643646, "global_step": 98875, "epoch": 1110, "lr": 7.085049187299451e-05}
{"train_loss": 0.09500385075807571, "global_step": 98876, "epoch": 1110, "lr": 7.084996494785071e-05}
{"train_loss": 0.16748304665088654, "global_step": 98877, "epoch": 1110, "lr": 7.084943801990388e-05}
{"train_loss": 0.16084544703866657, "global_step": 98878, "epoch": 1110, "lr": 7.08489110891541e-05, "val_loss": 4.703232765197754, "train_action_mse_error": 43.34333801269531}
{"train_loss": 0.15117056667804718, "global_step": 98879, "epoch": 1111, "lr": 7.084838415560144e-05}
{"train_loss": 0.1436983048915863, "global_step": 98880, "epoch": 1111, "lr": 7.084785721924595e-05}
{"train_loss": 0.1881374567747116, "global_step": 98881, "epoch": 1111, "lr": 7.084733028008774e-05}
{"train_loss": 0.14538323879241943, "global_step": 98882, "epoch": 1111, "lr": 7.084680333812685e-05}
{"train_loss": 0.21143044531345367, "global_step": 98883, "epoch": 1111, "lr": 7.084627639336336e-05}
{"train_loss": 0.14565768837928772, "global_step": 98884, "epoch": 1111, "lr": 7.084574944579735e-05}
{"train_loss": 0.19376282393932343, "global_step": 98885, "epoch": 1111, "lr": 7.084522249542888e-05}
{"train_loss": 0.17599022388458252, "global_step": 98886, "epoch": 1111, "lr": 7.084469554225803e-05}
{"train_loss": 0.1483270823955536, "global_step": 98887, "epoch": 1111, "lr": 7.084416858628484e-05}
{"train_loss": 0.19752632081508636, "global_step": 98888, "epoch": 1111, "lr": 7.084364162750942e-05}
{"train_loss": 0.13314585387706757, "global_step": 98889, "epoch": 1111, "lr": 7.084311466593185e-05}
{"train_loss": 0.20670031011104584, "global_step": 98890, "epoch": 1111, "lr": 7.084258770155215e-05}
{"train_loss": 0.1432088017463684, "global_step": 98891, "epoch": 1111, "lr": 7.084206073437042e-05}
{"train_loss": 0.15177874267101288, "global_step": 98892, "epoch": 1111, "lr": 7.084153376438674e-05}
{"train_loss": 0.16007646918296814, "global_step": 98893, "epoch": 1111, "lr": 7.084100679160116e-05}
{"train_loss": 0.2262699455022812, "global_step": 98894, "epoch": 1111, "lr": 7.084047981601376e-05}
{"train_loss": 0.25433963537216187, "global_step": 98895, "epoch": 1111, "lr": 7.083995283762463e-05}
{"train_loss": 0.19952543079853058, "global_step": 98896, "epoch": 1111, "lr": 7.083942585643379e-05}
{"train_loss": 0.21474149823188782, "global_step": 98897, "epoch": 1111, "lr": 7.083889887244138e-05}
{"train_loss": 0.16252124309539795, "global_step": 98898, "epoch": 1111, "lr": 7.083837188564741e-05}
{"train_loss": 0.19634883105754852, "global_step": 98899, "epoch": 1111, "lr": 7.083784489605198e-05}
{"train_loss": 0.2517181932926178, "global_step": 98900, "epoch": 1111, "lr": 7.083731790365516e-05}
{"train_loss": 0.10839853435754776, "global_step": 98901, "epoch": 1111, "lr": 7.083679090845701e-05}
{"train_loss": 0.1947261095046997, "global_step": 98902, "epoch": 1111, "lr": 7.083626391045761e-05}
{"train_loss": 0.14964111149311066, "global_step": 98903, "epoch": 1111, "lr": 7.083573690965703e-05}
{"train_loss": 0.10478733479976654, "global_step": 98904, "epoch": 1111, "lr": 7.083520990605533e-05}
{"train_loss": 0.2383754551410675, "global_step": 98905, "epoch": 1111, "lr": 7.08346828996526e-05}
{"train_loss": 0.20851363241672516, "global_step": 98906, "epoch": 1111, "lr": 7.083415589044891e-05}
{"train_loss": 0.1746206283569336, "global_step": 98907, "epoch": 1111, "lr": 7.08336288784443e-05}
{"train_loss": 0.22157402336597443, "global_step": 98908, "epoch": 1111, "lr": 7.083310186363888e-05}
{"train_loss": 0.29814255237579346, "global_step": 98909, "epoch": 1111, "lr": 7.08325748460327e-05}
{"train_loss": 0.19698812067508698, "global_step": 98910, "epoch": 1111, "lr": 7.083204782562583e-05}
{"train_loss": 0.16343773901462555, "global_step": 98911, "epoch": 1111, "lr": 7.083152080241834e-05}
{"train_loss": 0.21865734457969666, "global_step": 98912, "epoch": 1111, "lr": 7.083099377641031e-05}
{"train_loss": 0.1474934071302414, "global_step": 98913, "epoch": 1111, "lr": 7.083046674760182e-05}
{"train_loss": 0.19980181753635406, "global_step": 98914, "epoch": 1111, "lr": 7.08299397159929e-05}
{"train_loss": 0.16509538888931274, "global_step": 98915, "epoch": 1111, "lr": 7.082941268158366e-05}
{"train_loss": 0.15113714337348938, "global_step": 98916, "epoch": 1111, "lr": 7.082888564437418e-05}
{"train_loss": 0.16559556126594543, "global_step": 98917, "epoch": 1111, "lr": 7.082835860436447e-05}
{"train_loss": 0.17164696753025055, "global_step": 98918, "epoch": 1111, "lr": 7.082783156155469e-05}
{"train_loss": 0.1610935926437378, "global_step": 98919, "epoch": 1111, "lr": 7.082730451594483e-05}
{"train_loss": 0.19575266540050507, "global_step": 98920, "epoch": 1111, "lr": 7.082677746753499e-05}
{"train_loss": 0.13888275623321533, "global_step": 98921, "epoch": 1111, "lr": 7.082625041632525e-05}
{"train_loss": 0.19212442636489868, "global_step": 98922, "epoch": 1111, "lr": 7.082572336231568e-05}
{"train_loss": 0.11749052256345749, "global_step": 98923, "epoch": 1111, "lr": 7.082519630550634e-05}
{"train_loss": 0.19251051545143127, "global_step": 98924, "epoch": 1111, "lr": 7.082466924589732e-05}
{"train_loss": 0.1526753157377243, "global_step": 98925, "epoch": 1111, "lr": 7.082414218348867e-05}
{"train_loss": 0.20930571854114532, "global_step": 98926, "epoch": 1111, "lr": 7.082361511828046e-05}
{"train_loss": 0.209327831864357, "global_step": 98927, "epoch": 1111, "lr": 7.082308805027277e-05}
{"train_loss": 0.1572752743959427, "global_step": 98928, "epoch": 1111, "lr": 7.082256097946567e-05}
{"train_loss": 0.18517790734767914, "global_step": 98929, "epoch": 1111, "lr": 7.082203390585925e-05}
{"train_loss": 0.07172872871160507, "global_step": 98930, "epoch": 1111, "lr": 7.082150682945353e-05}
{"train_loss": 0.12199625372886658, "global_step": 98931, "epoch": 1111, "lr": 7.082097975024863e-05}
{"train_loss": 0.2334025353193283, "global_step": 98932, "epoch": 1111, "lr": 7.082045266824459e-05}
{"train_loss": 0.1921197921037674, "global_step": 98933, "epoch": 1111, "lr": 7.08199255834415e-05}
{"train_loss": 0.17223703861236572, "global_step": 98934, "epoch": 1111, "lr": 7.081939849583944e-05}
{"train_loss": 0.19106291234493256, "global_step": 98935, "epoch": 1111, "lr": 7.081887140543843e-05}
{"train_loss": 0.1316337138414383, "global_step": 98936, "epoch": 1111, "lr": 7.081834431223862e-05}
{"train_loss": 0.19184716045856476, "global_step": 98937, "epoch": 1111, "lr": 7.081781721624001e-05}
{"train_loss": 0.1020134687423706, "global_step": 98938, "epoch": 1111, "lr": 7.081729011744272e-05}
{"train_loss": 0.1841694861650467, "global_step": 98939, "epoch": 1111, "lr": 7.081676301584677e-05}
{"train_loss": 0.10855884104967117, "global_step": 98940, "epoch": 1111, "lr": 7.081623591145229e-05}
{"train_loss": 0.10143276304006577, "global_step": 98941, "epoch": 1111, "lr": 7.08157088042593e-05}
{"train_loss": 0.164606511592865, "global_step": 98942, "epoch": 1111, "lr": 7.08151816942679e-05}
{"train_loss": 0.15381209552288055, "global_step": 98943, "epoch": 1111, "lr": 7.081465458147815e-05}
{"train_loss": 0.14128464460372925, "global_step": 98944, "epoch": 1111, "lr": 7.081412746589013e-05}
{"train_loss": 0.1278495192527771, "global_step": 98945, "epoch": 1111, "lr": 7.081360034750389e-05}
{"train_loss": 0.12694428861141205, "global_step": 98946, "epoch": 1111, "lr": 7.081307322631951e-05}
{"train_loss": 0.16589806973934174, "global_step": 98947, "epoch": 1111, "lr": 7.08125461023371e-05}
{"train_loss": 0.1579713225364685, "global_step": 98948, "epoch": 1111, "lr": 7.081201897555667e-05}
{"train_loss": 0.24418814480304718, "global_step": 98949, "epoch": 1111, "lr": 7.081149184597832e-05}
{"train_loss": 0.1093156561255455, "global_step": 98950, "epoch": 1111, "lr": 7.081096471360212e-05}
{"train_loss": 0.17018727958202362, "global_step": 98951, "epoch": 1111, "lr": 7.081043757842813e-05}
{"train_loss": 0.18960188329219818, "global_step": 98952, "epoch": 1111, "lr": 7.080991044045646e-05}
{"train_loss": 0.1799323707818985, "global_step": 98953, "epoch": 1111, "lr": 7.080938329968712e-05}
{"train_loss": 0.16044440865516663, "global_step": 98954, "epoch": 1111, "lr": 7.080885615612022e-05}
{"train_loss": 0.12798646092414856, "global_step": 98955, "epoch": 1111, "lr": 7.080832900975583e-05}
{"train_loss": 0.2381906360387802, "global_step": 98956, "epoch": 1111, "lr": 7.080780186059401e-05}
{"train_loss": 0.2555307149887085, "global_step": 98957, "epoch": 1111, "lr": 7.080727470863484e-05}
{"train_loss": 0.09701432287693024, "global_step": 98958, "epoch": 1111, "lr": 7.080674755387838e-05}
{"train_loss": 0.11253856122493744, "global_step": 98959, "epoch": 1111, "lr": 7.080622039632473e-05}
{"train_loss": 0.09751417487859726, "global_step": 98960, "epoch": 1111, "lr": 7.08056932359739e-05}
{"train_loss": 0.17403867840766907, "global_step": 98961, "epoch": 1111, "lr": 7.080516607282601e-05}
{"train_loss": 0.14361605048179626, "global_step": 98962, "epoch": 1111, "lr": 7.080463890688113e-05}
{"train_loss": 0.20428721606731415, "global_step": 98963, "epoch": 1111, "lr": 7.08041117381393e-05}
{"train_loss": 0.14744709432125092, "global_step": 98964, "epoch": 1111, "lr": 7.080358456660063e-05}
{"train_loss": 0.1597874015569687, "global_step": 98965, "epoch": 1111, "lr": 7.080305739226515e-05}
{"train_loss": 0.16314253211021423, "global_step": 98966, "epoch": 1111, "lr": 7.080253021513298e-05}
{"train_loss": 0.1710201119271557, "global_step": 98967, "epoch": 1111, "lr": 7.080200303520415e-05, "val_loss": 4.525521278381348}
{"train_loss": 0.21221111714839935, "global_step": 98968, "epoch": 1112, "lr": 7.080147585247875e-05}
{"train_loss": 0.12283115833997726, "global_step": 98969, "epoch": 1112, "lr": 7.080094866695684e-05}
{"train_loss": 0.16805581748485565, "global_step": 98970, "epoch": 1112, "lr": 7.080042147863849e-05}
{"train_loss": 0.15352332592010498, "global_step": 98971, "epoch": 1112, "lr": 7.079989428752378e-05}
{"train_loss": 0.12481841444969177, "global_step": 98972, "epoch": 1112, "lr": 7.079936709361276e-05}
{"train_loss": 0.19596602022647858, "global_step": 98973, "epoch": 1112, "lr": 7.079883989690556e-05}
{"train_loss": 0.1048840805888176, "global_step": 98974, "epoch": 1112, "lr": 7.079831269740217e-05}
{"train_loss": 0.21783749759197235, "global_step": 98975, "epoch": 1112, "lr": 7.079778549510273e-05}
{"train_loss": 0.17072850465774536, "global_step": 98976, "epoch": 1112, "lr": 7.079725829000727e-05}
{"train_loss": 0.21036297082901, "global_step": 98977, "epoch": 1112, "lr": 7.079673108211588e-05}
{"train_loss": 0.20953534543514252, "global_step": 98978, "epoch": 1112, "lr": 7.079620387142863e-05}
{"train_loss": 0.15685150027275085, "global_step": 98979, "epoch": 1112, "lr": 7.079567665794555e-05}
{"train_loss": 0.15814611315727234, "global_step": 98980, "epoch": 1112, "lr": 7.079514944166677e-05}
{"train_loss": 0.16940632462501526, "global_step": 98981, "epoch": 1112, "lr": 7.079462222259233e-05}
{"train_loss": 0.18044854700565338, "global_step": 98982, "epoch": 1112, "lr": 7.079409500072231e-05}
{"train_loss": 0.1269635558128357, "global_step": 98983, "epoch": 1112, "lr": 7.079356777605677e-05}
{"train_loss": 0.17276696860790253, "global_step": 98984, "epoch": 1112, "lr": 7.07930405485958e-05}
{"train_loss": 0.13007503747940063, "global_step": 98985, "epoch": 1112, "lr": 7.079251331833944e-05}
{"train_loss": 0.131869837641716, "global_step": 98986, "epoch": 1112, "lr": 7.079198608528779e-05}
{"train_loss": 0.2192254364490509, "global_step": 98987, "epoch": 1112, "lr": 7.079145884944091e-05}
{"train_loss": 0.18931566178798676, "global_step": 98988, "epoch": 1112, "lr": 7.079093161079887e-05}
{"train_loss": 0.15578210353851318, "global_step": 98989, "epoch": 1112, "lr": 7.079040436936176e-05}
{"train_loss": 0.18202738463878632, "global_step": 98990, "epoch": 1112, "lr": 7.07898771251296e-05}
{"train_loss": 0.13831409811973572, "global_step": 98991, "epoch": 1112, "lr": 7.078934987810251e-05}
{"train_loss": 0.17574407160282135, "global_step": 98992, "epoch": 1112, "lr": 7.078882262828055e-05}
{"train_loss": 0.21500195562839508, "global_step": 98993, "epoch": 1112, "lr": 7.078829537566379e-05}
{"train_loss": 0.18103328347206116, "global_step": 98994, "epoch": 1112, "lr": 7.07877681202523e-05}
{"train_loss": 0.20381948351860046, "global_step": 98995, "epoch": 1112, "lr": 7.078724086204614e-05}
{"train_loss": 0.17117421329021454, "global_step": 98996, "epoch": 1112, "lr": 7.078671360104538e-05}
{"train_loss": 0.19897177815437317, "global_step": 98997, "epoch": 1112, "lr": 7.078618633725011e-05}
{"train_loss": 0.17451030015945435, "global_step": 98998, "epoch": 1112, "lr": 7.078565907066039e-05}
{"train_loss": 0.14325255155563354, "global_step": 98999, "epoch": 1112, "lr": 7.07851318012763e-05}
{"train_loss": 0.17525216937065125, "global_step": 99000, "epoch": 1112, "lr": 7.07846045290979e-05}
{"train_loss": 0.14450988173484802, "global_step": 99001, "epoch": 1112, "lr": 7.078407725412525e-05}
{"train_loss": 0.15260790288448334, "global_step": 99002, "epoch": 1112, "lr": 7.078354997635843e-05}
{"train_loss": 0.1685710847377777, "global_step": 99003, "epoch": 1112, "lr": 7.078302269579753e-05}
{"train_loss": 0.1390911340713501, "global_step": 99004, "epoch": 1112, "lr": 7.078249541244261e-05}
{"train_loss": 0.15128327906131744, "global_step": 99005, "epoch": 1112, "lr": 7.078196812629372e-05}
{"train_loss": 0.22677846252918243, "global_step": 99006, "epoch": 1112, "lr": 7.078144083735097e-05}
{"train_loss": 0.13574407994747162, "global_step": 99007, "epoch": 1112, "lr": 7.078091354561438e-05}
{"train_loss": 0.10561922192573547, "global_step": 99008, "epoch": 1112, "lr": 7.078038625108405e-05}
{"train_loss": 0.14890369772911072, "global_step": 99009, "epoch": 1112, "lr": 7.077985895376007e-05}
{"train_loss": 0.1785268634557724, "global_step": 99010, "epoch": 1112, "lr": 7.077933165364249e-05}
{"train_loss": 0.16060973703861237, "global_step": 99011, "epoch": 1112, "lr": 7.077880435073137e-05}
{"train_loss": 0.10362524539232254, "global_step": 99012, "epoch": 1112, "lr": 7.07782770450268e-05}
{"train_loss": 0.09115855395793915, "global_step": 99013, "epoch": 1112, "lr": 7.077774973652886e-05}
{"train_loss": 0.15553386509418488, "global_step": 99014, "epoch": 1112, "lr": 7.077722242523757e-05}
{"train_loss": 0.16421808302402496, "global_step": 99015, "epoch": 1112, "lr": 7.077669511115307e-05}
{"train_loss": 0.15356171131134033, "global_step": 99016, "epoch": 1112, "lr": 7.077616779427538e-05}
{"train_loss": 0.1466090828180313, "global_step": 99017, "epoch": 1112, "lr": 7.07756404746046e-05}
{"train_loss": 0.1858496367931366, "global_step": 99018, "epoch": 1112, "lr": 7.077511315214077e-05}
{"train_loss": 0.20402981340885162, "global_step": 99019, "epoch": 1112, "lr": 7.077458582688399e-05}
{"train_loss": 0.10349436104297638, "global_step": 99020, "epoch": 1112, "lr": 7.077405849883431e-05}
{"train_loss": 0.1823107898235321, "global_step": 99021, "epoch": 1112, "lr": 7.077353116799183e-05}
{"train_loss": 0.14740058779716492, "global_step": 99022, "epoch": 1112, "lr": 7.077300383435659e-05}
{"train_loss": 0.1662023812532425, "global_step": 99023, "epoch": 1112, "lr": 7.077247649792866e-05}
{"train_loss": 0.15974611043930054, "global_step": 99024, "epoch": 1112, "lr": 7.077194915870815e-05}
{"train_loss": 0.11198496073484421, "global_step": 99025, "epoch": 1112, "lr": 7.077142181669509e-05}
{"train_loss": 0.1952049881219864, "global_step": 99026, "epoch": 1112, "lr": 7.077089447188954e-05}
{"train_loss": 0.1871451735496521, "global_step": 99027, "epoch": 1112, "lr": 7.077036712429163e-05}
{"train_loss": 0.14900986850261688, "global_step": 99028, "epoch": 1112, "lr": 7.076983977390139e-05}
{"train_loss": 0.12428367137908936, "global_step": 99029, "epoch": 1112, "lr": 7.07693124207189e-05}
{"train_loss": 0.11740705370903015, "global_step": 99030, "epoch": 1112, "lr": 7.076878506474422e-05}
{"train_loss": 0.23518994450569153, "global_step": 99031, "epoch": 1112, "lr": 7.076825770597745e-05}
{"train_loss": 0.20131836831569672, "global_step": 99032, "epoch": 1112, "lr": 7.076773034441863e-05}
{"train_loss": 0.12755495309829712, "global_step": 99033, "epoch": 1112, "lr": 7.076720298006783e-05}
{"train_loss": 0.16325493156909943, "global_step": 99034, "epoch": 1112, "lr": 7.076667561292516e-05}
{"train_loss": 0.20638157427310944, "global_step": 99035, "epoch": 1112, "lr": 7.076614824299064e-05}
{"train_loss": 0.11427069455385208, "global_step": 99036, "epoch": 1112, "lr": 7.076562087026437e-05}
{"train_loss": 0.19656412303447723, "global_step": 99037, "epoch": 1112, "lr": 7.076509349474643e-05}
{"train_loss": 0.08577565103769302, "global_step": 99038, "epoch": 1112, "lr": 7.076456611643685e-05}
{"train_loss": 0.11474469304084778, "global_step": 99039, "epoch": 1112, "lr": 7.076403873533574e-05}
{"train_loss": 0.1365489810705185, "global_step": 99040, "epoch": 1112, "lr": 7.076351135144316e-05}
{"train_loss": 0.19512499868869781, "global_step": 99041, "epoch": 1112, "lr": 7.076298396475918e-05}
{"train_loss": 0.10759630054235458, "global_step": 99042, "epoch": 1112, "lr": 7.076245657528386e-05}
{"train_loss": 0.21249674260616302, "global_step": 99043, "epoch": 1112, "lr": 7.07619291830173e-05}
{"train_loss": 0.1709679663181305, "global_step": 99044, "epoch": 1112, "lr": 7.076140178795952e-05}
{"train_loss": 0.12019170820713043, "global_step": 99045, "epoch": 1112, "lr": 7.076087439011064e-05}
{"train_loss": 0.11395888030529022, "global_step": 99046, "epoch": 1112, "lr": 7.076034698947072e-05}
{"train_loss": 0.12837132811546326, "global_step": 99047, "epoch": 1112, "lr": 7.075981958603983e-05}
{"train_loss": 0.23428955674171448, "global_step": 99048, "epoch": 1112, "lr": 7.075929217981801e-05}
{"train_loss": 0.22440485656261444, "global_step": 99049, "epoch": 1112, "lr": 7.075876477080537e-05}
{"train_loss": 0.24356035888195038, "global_step": 99050, "epoch": 1112, "lr": 7.075823735900198e-05}
{"train_loss": 0.2144109308719635, "global_step": 99051, "epoch": 1112, "lr": 7.075770994440788e-05}
{"train_loss": 0.12832079827785492, "global_step": 99052, "epoch": 1112, "lr": 7.075718252702317e-05}
{"train_loss": 0.17038613557815552, "global_step": 99053, "epoch": 1112, "lr": 7.07566551068479e-05}
{"train_loss": 0.1577329784631729, "global_step": 99054, "epoch": 1112, "lr": 7.075612768388217e-05}
{"train_loss": 0.16065679490566254, "global_step": 99055, "epoch": 1112, "lr": 7.075560025812601e-05}
{"train_loss": 0.16317015995135467, "global_step": 99056, "epoch": 1112, "lr": 7.075507282957952e-05, "val_loss": 4.657005786895752}
{"train_loss": 0.1697833389043808, "global_step": 99057, "epoch": 1113, "lr": 7.075454539824277e-05}
{"train_loss": 0.1719474047422409, "global_step": 99058, "epoch": 1113, "lr": 7.07540179641158e-05}
{"train_loss": 0.1477271169424057, "global_step": 99059, "epoch": 1113, "lr": 7.075349052719873e-05}
{"train_loss": 0.15296266973018646, "global_step": 99060, "epoch": 1113, "lr": 7.07529630874916e-05}
{"train_loss": 0.155313178896904, "global_step": 99061, "epoch": 1113, "lr": 7.075243564499449e-05}
{"train_loss": 0.16875053942203522, "global_step": 99062, "epoch": 1113, "lr": 7.075190819970746e-05}
{"train_loss": 0.1923740953207016, "global_step": 99063, "epoch": 1113, "lr": 7.075138075163058e-05}
{"train_loss": 0.1275586485862732, "global_step": 99064, "epoch": 1113, "lr": 7.075085330076394e-05}
{"train_loss": 0.13549739122390747, "global_step": 99065, "epoch": 1113, "lr": 7.075032584710761e-05}
{"train_loss": 0.21413767337799072, "global_step": 99066, "epoch": 1113, "lr": 7.074979839066164e-05}
{"train_loss": 0.07781514525413513, "global_step": 99067, "epoch": 1113, "lr": 7.074927093142612e-05}
{"train_loss": 0.1067245602607727, "global_step": 99068, "epoch": 1113, "lr": 7.074874346940111e-05}
{"train_loss": 0.16846244037151337, "global_step": 99069, "epoch": 1113, "lr": 7.074821600458669e-05}
{"train_loss": 0.10987848788499832, "global_step": 99070, "epoch": 1113, "lr": 7.074768853698291e-05}
{"train_loss": 0.13340428471565247, "global_step": 99071, "epoch": 1113, "lr": 7.074716106658987e-05}
{"train_loss": 0.14621296525001526, "global_step": 99072, "epoch": 1113, "lr": 7.074663359340763e-05}
{"train_loss": 0.17049863934516907, "global_step": 99073, "epoch": 1113, "lr": 7.074610611743627e-05}
{"train_loss": 0.14603744447231293, "global_step": 99074, "epoch": 1113, "lr": 7.074557863867581e-05}
{"train_loss": 0.20475836098194122, "global_step": 99075, "epoch": 1113, "lr": 7.074505115712639e-05}
{"train_loss": 0.16597595810890198, "global_step": 99076, "epoch": 1113, "lr": 7.074452367278805e-05}
{"train_loss": 0.09032903611660004, "global_step": 99077, "epoch": 1113, "lr": 7.074399618566084e-05}
{"train_loss": 0.16554434597492218, "global_step": 99078, "epoch": 1113, "lr": 7.074346869574486e-05}
{"train_loss": 0.149399995803833, "global_step": 99079, "epoch": 1113, "lr": 7.07429412030402e-05}
{"train_loss": 0.15270785987377167, "global_step": 99080, "epoch": 1113, "lr": 7.074241370754687e-05}
{"train_loss": 0.10141213238239288, "global_step": 99081, "epoch": 1113, "lr": 7.074188620926497e-05}
{"train_loss": 0.18706423044204712, "global_step": 99082, "epoch": 1113, "lr": 7.074135870819461e-05}
{"train_loss": 0.16477584838867188, "global_step": 99083, "epoch": 1113, "lr": 7.07408312043358e-05}
{"train_loss": 0.2398974448442459, "global_step": 99084, "epoch": 1113, "lr": 7.074030369768865e-05}
{"train_loss": 0.17795060575008392, "global_step": 99085, "epoch": 1113, "lr": 7.073977618825323e-05}
{"train_loss": 0.19140660762786865, "global_step": 99086, "epoch": 1113, "lr": 7.073924867602959e-05}
{"train_loss": 0.10168342292308807, "global_step": 99087, "epoch": 1113, "lr": 7.07387211610178e-05}
{"train_loss": 0.23683606088161469, "global_step": 99088, "epoch": 1113, "lr": 7.073819364321794e-05}
{"train_loss": 0.22478510439395905, "global_step": 99089, "epoch": 1113, "lr": 7.07376661226301e-05}
{"train_loss": 0.19054530560970306, "global_step": 99090, "epoch": 1113, "lr": 7.073713859925433e-05}
{"train_loss": 0.14337356388568878, "global_step": 99091, "epoch": 1113, "lr": 7.07366110730907e-05}
{"train_loss": 0.16467130184173584, "global_step": 99092, "epoch": 1113, "lr": 7.07360835441393e-05}
{"train_loss": 0.1206701472401619, "global_step": 99093, "epoch": 1113, "lr": 7.073555601240018e-05}
{"train_loss": 0.1793733686208725, "global_step": 99094, "epoch": 1113, "lr": 7.07350284778734e-05}
{"train_loss": 0.14954030513763428, "global_step": 99095, "epoch": 1113, "lr": 7.073450094055906e-05}
{"train_loss": 0.11546212434768677, "global_step": 99096, "epoch": 1113, "lr": 7.073397340045722e-05}
{"train_loss": 0.15630628168582916, "global_step": 99097, "epoch": 1113, "lr": 7.073344585756795e-05}
{"train_loss": 0.18822766840457916, "global_step": 99098, "epoch": 1113, "lr": 7.073291831189132e-05}
{"train_loss": 0.16473358869552612, "global_step": 99099, "epoch": 1113, "lr": 7.073239076342738e-05}
{"train_loss": 0.1497069001197815, "global_step": 99100, "epoch": 1113, "lr": 7.073186321217624e-05}
{"train_loss": 0.15420320630073547, "global_step": 99101, "epoch": 1113, "lr": 7.073133565813796e-05}
{"train_loss": 0.1279713362455368, "global_step": 99102, "epoch": 1113, "lr": 7.07308081013126e-05}
{"train_loss": 0.17175425589084625, "global_step": 99103, "epoch": 1113, "lr": 7.073028054170024e-05}
{"train_loss": 0.1990448534488678, "global_step": 99104, "epoch": 1113, "lr": 7.072975297930095e-05}
{"train_loss": 0.20924752950668335, "global_step": 99105, "epoch": 1113, "lr": 7.072922541411479e-05}
{"train_loss": 0.12166700512170792, "global_step": 99106, "epoch": 1113, "lr": 7.072869784614184e-05}
{"train_loss": 0.12962985038757324, "global_step": 99107, "epoch": 1113, "lr": 7.072817027538216e-05}
{"train_loss": 0.17045724391937256, "global_step": 99108, "epoch": 1113, "lr": 7.072764270183585e-05}
{"train_loss": 0.1582046002149582, "global_step": 99109, "epoch": 1113, "lr": 7.072711512550294e-05}
{"train_loss": 0.18166543543338776, "global_step": 99110, "epoch": 1113, "lr": 7.072658754638352e-05}
{"train_loss": 0.22210456430912018, "global_step": 99111, "epoch": 1113, "lr": 7.072605996447769e-05}
{"train_loss": 0.21847672760486603, "global_step": 99112, "epoch": 1113, "lr": 7.072553237978549e-05}
{"train_loss": 0.16135616600513458, "global_step": 99113, "epoch": 1113, "lr": 7.072500479230697e-05}
{"train_loss": 0.12499597668647766, "global_step": 99114, "epoch": 1113, "lr": 7.072447720204225e-05}
{"train_loss": 0.25467586517333984, "global_step": 99115, "epoch": 1113, "lr": 7.072394960899136e-05}
{"train_loss": 0.17779216170310974, "global_step": 99116, "epoch": 1113, "lr": 7.07234220131544e-05}
{"train_loss": 0.15006031095981598, "global_step": 99117, "epoch": 1113, "lr": 7.07228944145314e-05}
{"train_loss": 0.13912180066108704, "global_step": 99118, "epoch": 1113, "lr": 7.072236681312248e-05}
{"train_loss": 0.1762438714504242, "global_step": 99119, "epoch": 1113, "lr": 7.07218392089277e-05}
{"train_loss": 0.13013672828674316, "global_step": 99120, "epoch": 1113, "lr": 7.07213116019471e-05}
{"train_loss": 0.16800127923488617, "global_step": 99121, "epoch": 1113, "lr": 7.07207839921808e-05}
{"train_loss": 0.20101825892925262, "global_step": 99122, "epoch": 1113, "lr": 7.072025637962882e-05}
{"train_loss": 0.16028936207294464, "global_step": 99123, "epoch": 1113, "lr": 7.071972876429128e-05}
{"train_loss": 0.1810993254184723, "global_step": 99124, "epoch": 1113, "lr": 7.071920114616819e-05}
{"train_loss": 0.13373151421546936, "global_step": 99125, "epoch": 1113, "lr": 7.071867352525968e-05}
{"train_loss": 0.19528880715370178, "global_step": 99126, "epoch": 1113, "lr": 7.071814590156578e-05}
{"train_loss": 0.22276321053504944, "global_step": 99127, "epoch": 1113, "lr": 7.07176182750866e-05}
{"train_loss": 0.1454552561044693, "global_step": 99128, "epoch": 1113, "lr": 7.071709064582217e-05}
{"train_loss": 0.14453153312206268, "global_step": 99129, "epoch": 1113, "lr": 7.071656301377258e-05}
{"train_loss": 0.15934062004089355, "global_step": 99130, "epoch": 1113, "lr": 7.07160353789379e-05}
{"train_loss": 0.10682430118322372, "global_step": 99131, "epoch": 1113, "lr": 7.071550774131823e-05}
{"train_loss": 0.23518066108226776, "global_step": 99132, "epoch": 1113, "lr": 7.071498010091358e-05}
{"train_loss": 0.20997266471385956, "global_step": 99133, "epoch": 1113, "lr": 7.071445245772408e-05}
{"train_loss": 0.15838392078876495, "global_step": 99134, "epoch": 1113, "lr": 7.071392481174975e-05}
{"train_loss": 0.1511373370885849, "global_step": 99135, "epoch": 1113, "lr": 7.071339716299069e-05}
{"train_loss": 0.16991335153579712, "global_step": 99136, "epoch": 1113, "lr": 7.071286951144697e-05}
{"train_loss": 0.14984296262264252, "global_step": 99137, "epoch": 1113, "lr": 7.071234185711865e-05}
{"train_loss": 0.18217386305332184, "global_step": 99138, "epoch": 1113, "lr": 7.07118142000058e-05}
{"train_loss": 0.10944005101919174, "global_step": 99139, "epoch": 1113, "lr": 7.071128654010853e-05}
{"train_loss": 0.08733505010604858, "global_step": 99140, "epoch": 1113, "lr": 7.071075887742686e-05}
{"train_loss": 0.12803444266319275, "global_step": 99141, "epoch": 1113, "lr": 7.071023121196089e-05}
{"train_loss": 0.17315953969955444, "global_step": 99142, "epoch": 1113, "lr": 7.070970354371067e-05}
{"train_loss": 0.11945231258869171, "global_step": 99143, "epoch": 1113, "lr": 7.070917587267629e-05}
{"train_loss": 0.18390777707099915, "global_step": 99144, "epoch": 1113, "lr": 7.070864819885782e-05}
{"train_loss": 0.1623440533039275, "global_step": 99145, "epoch": 1113, "lr": 7.070812052225532e-05, "val_loss": 4.580025672912598}
{"train_loss": 0.18888577818870544, "global_step": 99146, "epoch": 1114, "lr": 7.070759284286886e-05}
{"train_loss": 0.14409111440181732, "global_step": 99147, "epoch": 1114, "lr": 7.070706516069852e-05}
{"train_loss": 0.13214857876300812, "global_step": 99148, "epoch": 1114, "lr": 7.070653747574436e-05}
{"train_loss": 0.20828655362129211, "global_step": 99149, "epoch": 1114, "lr": 7.070600978800648e-05}
{"train_loss": 0.19672362506389618, "global_step": 99150, "epoch": 1114, "lr": 7.07054820974849e-05}
{"train_loss": 0.14079856872558594, "global_step": 99151, "epoch": 1114, "lr": 7.070495440417974e-05}
{"train_loss": 0.12786296010017395, "global_step": 99152, "epoch": 1114, "lr": 7.070442670809105e-05}
{"train_loss": 0.08463671058416367, "global_step": 99153, "epoch": 1114, "lr": 7.070389900921889e-05}
{"train_loss": 0.13518160581588745, "global_step": 99154, "epoch": 1114, "lr": 7.070337130756335e-05}
{"train_loss": 0.17966002225875854, "global_step": 99155, "epoch": 1114, "lr": 7.070284360312452e-05}
{"train_loss": 0.2262837141752243, "global_step": 99156, "epoch": 1114, "lr": 7.07023158959024e-05}
{"train_loss": 0.09317043423652649, "global_step": 99157, "epoch": 1114, "lr": 7.070178818589714e-05}
{"train_loss": 0.1155511885881424, "global_step": 99158, "epoch": 1114, "lr": 7.070126047310877e-05}
{"train_loss": 0.17754323780536652, "global_step": 99159, "epoch": 1114, "lr": 7.070073275753734e-05}
{"train_loss": 0.20782823860645294, "global_step": 99160, "epoch": 1114, "lr": 7.070020503918298e-05}
{"train_loss": 0.188948854804039, "global_step": 99161, "epoch": 1114, "lr": 7.069967731804572e-05}
{"train_loss": 0.18913480639457703, "global_step": 99162, "epoch": 1114, "lr": 7.069914959412564e-05}
{"train_loss": 0.2856835424900055, "global_step": 99163, "epoch": 1114, "lr": 7.069862186742282e-05}
{"train_loss": 0.20280180871486664, "global_step": 99164, "epoch": 1114, "lr": 7.06980941379373e-05}
{"train_loss": 0.19395391643047333, "global_step": 99165, "epoch": 1114, "lr": 7.06975664056692e-05}
{"train_loss": 0.2121470719575882, "global_step": 99166, "epoch": 1114, "lr": 7.069703867061855e-05}
{"train_loss": 0.1545235812664032, "global_step": 99167, "epoch": 1114, "lr": 7.069651093278545e-05}
{"train_loss": 0.13172511756420135, "global_step": 99168, "epoch": 1114, "lr": 7.069598319216993e-05}
{"train_loss": 0.1603947877883911, "global_step": 99169, "epoch": 1114, "lr": 7.069545544877211e-05}
{"train_loss": 0.11311730742454529, "global_step": 99170, "epoch": 1114, "lr": 7.069492770259204e-05}
{"train_loss": 0.1716381460428238, "global_step": 99171, "epoch": 1114, "lr": 7.069439995362977e-05}
{"train_loss": 0.15177905559539795, "global_step": 99172, "epoch": 1114, "lr": 7.069387220188539e-05}
{"train_loss": 0.21312576532363892, "global_step": 99173, "epoch": 1114, "lr": 7.069334444735899e-05}
{"train_loss": 0.14171554148197174, "global_step": 99174, "epoch": 1114, "lr": 7.069281669005061e-05}
{"train_loss": 0.16661417484283447, "global_step": 99175, "epoch": 1114, "lr": 7.069228892996034e-05}
{"train_loss": 0.18401439487934113, "global_step": 99176, "epoch": 1114, "lr": 7.069176116708824e-05}
{"train_loss": 0.16637882590293884, "global_step": 99177, "epoch": 1114, "lr": 7.069123340143439e-05}
{"train_loss": 0.14345470070838928, "global_step": 99178, "epoch": 1114, "lr": 7.069070563299884e-05}
{"train_loss": 0.12798738479614258, "global_step": 99179, "epoch": 1114, "lr": 7.069017786178169e-05}
{"train_loss": 0.14941570162773132, "global_step": 99180, "epoch": 1114, "lr": 7.0689650087783e-05}
{"train_loss": 0.23990172147750854, "global_step": 99181, "epoch": 1114, "lr": 7.068912231100282e-05}
{"train_loss": 0.1615074723958969, "global_step": 99182, "epoch": 1114, "lr": 7.068859453144126e-05}
{"train_loss": 0.18358676135540009, "global_step": 99183, "epoch": 1114, "lr": 7.068806674909837e-05}
{"train_loss": 0.12888659536838531, "global_step": 99184, "epoch": 1114, "lr": 7.068753896397422e-05}
{"train_loss": 0.15874932706356049, "global_step": 99185, "epoch": 1114, "lr": 7.068701117606887e-05}
{"train_loss": 0.16234368085861206, "global_step": 99186, "epoch": 1114, "lr": 7.068648338538242e-05}
{"train_loss": 0.15418057143688202, "global_step": 99187, "epoch": 1114, "lr": 7.068595559191492e-05}
{"train_loss": 0.09505555033683777, "global_step": 99188, "epoch": 1114, "lr": 7.068542779566644e-05}
{"train_loss": 0.12598542869091034, "global_step": 99189, "epoch": 1114, "lr": 7.068489999663706e-05}
{"train_loss": 0.21721065044403076, "global_step": 99190, "epoch": 1114, "lr": 7.068437219482684e-05}
{"train_loss": 0.1657237410545349, "global_step": 99191, "epoch": 1114, "lr": 7.068384439023586e-05}
{"train_loss": 0.159736767411232, "global_step": 99192, "epoch": 1114, "lr": 7.06833165828642e-05}
{"train_loss": 0.10141615569591522, "global_step": 99193, "epoch": 1114, "lr": 7.06827887727119e-05}
{"train_loss": 0.2211524099111557, "global_step": 99194, "epoch": 1114, "lr": 7.068226095977908e-05}
{"train_loss": 0.18245773017406464, "global_step": 99195, "epoch": 1114, "lr": 7.068173314406577e-05}
{"train_loss": 0.14511360228061676, "global_step": 99196, "epoch": 1114, "lr": 7.068120532557202e-05}
{"train_loss": 0.13097591698169708, "global_step": 99197, "epoch": 1114, "lr": 7.068067750429799e-05}
{"train_loss": 0.13131359219551086, "global_step": 99198, "epoch": 1114, "lr": 7.068014968024366e-05}
{"train_loss": 0.1768156737089157, "global_step": 99199, "epoch": 1114, "lr": 7.067962185340914e-05}
{"train_loss": 0.1456989198923111, "global_step": 99200, "epoch": 1114, "lr": 7.06790940237945e-05}
{"train_loss": 0.1548667997121811, "global_step": 99201, "epoch": 1114, "lr": 7.067856619139981e-05}
{"train_loss": 0.17067325115203857, "global_step": 99202, "epoch": 1114, "lr": 7.067803835622514e-05}
{"train_loss": 0.1695057451725006, "global_step": 99203, "epoch": 1114, "lr": 7.067751051827056e-05}
{"train_loss": 0.15184034407138824, "global_step": 99204, "epoch": 1114, "lr": 7.067698267753614e-05}
{"train_loss": 0.192704975605011, "global_step": 99205, "epoch": 1114, "lr": 7.067645483402194e-05}
{"train_loss": 0.12277405709028244, "global_step": 99206, "epoch": 1114, "lr": 7.067592698772806e-05}
{"train_loss": 0.16562582552433014, "global_step": 99207, "epoch": 1114, "lr": 7.067539913865455e-05}
{"train_loss": 0.13504460453987122, "global_step": 99208, "epoch": 1114, "lr": 7.067487128680148e-05}
{"train_loss": 0.19162335991859436, "global_step": 99209, "epoch": 1114, "lr": 7.067434343216891e-05}
{"train_loss": 0.168185755610466, "global_step": 99210, "epoch": 1114, "lr": 7.067381557475696e-05}
{"train_loss": 0.17326374351978302, "global_step": 99211, "epoch": 1114, "lr": 7.067328771456565e-05}
{"train_loss": 0.17522650957107544, "global_step": 99212, "epoch": 1114, "lr": 7.067275985159507e-05}
{"train_loss": 0.18050429224967957, "global_step": 99213, "epoch": 1114, "lr": 7.067223198584528e-05}
{"train_loss": 0.17900153994560242, "global_step": 99214, "epoch": 1114, "lr": 7.067170411731637e-05}
{"train_loss": 0.15266555547714233, "global_step": 99215, "epoch": 1114, "lr": 7.06711762460084e-05}
{"train_loss": 0.17325040698051453, "global_step": 99216, "epoch": 1114, "lr": 7.067064837192145e-05}
{"train_loss": 0.13680361211299896, "global_step": 99217, "epoch": 1114, "lr": 7.067012049505557e-05}
{"train_loss": 0.147455096244812, "global_step": 99218, "epoch": 1114, "lr": 7.066959261541086e-05}
{"train_loss": 0.13635945320129395, "global_step": 99219, "epoch": 1114, "lr": 7.066906473298736e-05}
{"train_loss": 0.1195925772190094, "global_step": 99220, "epoch": 1114, "lr": 7.066853684778516e-05}
{"train_loss": 0.1837136149406433, "global_step": 99221, "epoch": 1114, "lr": 7.066800895980432e-05}
{"train_loss": 0.11714164167642593, "global_step": 99222, "epoch": 1114, "lr": 7.066748106904493e-05}
{"train_loss": 0.09481212496757507, "global_step": 99223, "epoch": 1114, "lr": 7.066695317550705e-05}
{"train_loss": 0.14685310423374176, "global_step": 99224, "epoch": 1114, "lr": 7.066642527919075e-05}
{"train_loss": 0.1515994668006897, "global_step": 99225, "epoch": 1114, "lr": 7.06658973800961e-05}
{"train_loss": 0.1625712364912033, "global_step": 99226, "epoch": 1114, "lr": 7.066536947822315e-05}
{"train_loss": 0.1543808877468109, "global_step": 99227, "epoch": 1114, "lr": 7.066484157357203e-05}
{"train_loss": 0.1616804152727127, "global_step": 99228, "epoch": 1114, "lr": 7.066431366614274e-05}
{"train_loss": 0.17881952226161957, "global_step": 99229, "epoch": 1114, "lr": 7.066378575593541e-05}
{"train_loss": 0.15834370255470276, "global_step": 99230, "epoch": 1114, "lr": 7.066325784295007e-05}
{"train_loss": 0.15995293855667114, "global_step": 99231, "epoch": 1114, "lr": 7.06627299271868e-05}
{"train_loss": 0.11597955971956253, "global_step": 99232, "epoch": 1114, "lr": 7.06622020086457e-05}
{"train_loss": 0.08317508548498154, "global_step": 99233, "epoch": 1114, "lr": 7.066167408732682e-05}
{"train_loss": 0.159147443647465, "global_step": 99234, "epoch": 1114, "lr": 7.066114616323022e-05, "val_loss": 4.612303733825684}
{"train_loss": 0.11800540238618851, "global_step": 99235, "epoch": 1115, "lr": 7.066061823635597e-05}
{"train_loss": 0.17252200841903687, "global_step": 99236, "epoch": 1115, "lr": 7.066009030670416e-05}
{"train_loss": 0.1390189826488495, "global_step": 99237, "epoch": 1115, "lr": 7.065956237427486e-05}
{"train_loss": 0.12987521290779114, "global_step": 99238, "epoch": 1115, "lr": 7.065903443906812e-05}
{"train_loss": 0.1756952404975891, "global_step": 99239, "epoch": 1115, "lr": 7.065850650108405e-05}
{"train_loss": 0.14889250695705414, "global_step": 99240, "epoch": 1115, "lr": 7.065797856032267e-05}
{"train_loss": 0.17893573641777039, "global_step": 99241, "epoch": 1115, "lr": 7.065745061678408e-05}
{"train_loss": 0.16234885156154633, "global_step": 99242, "epoch": 1115, "lr": 7.065692267046834e-05}
{"train_loss": 0.15554754436016083, "global_step": 99243, "epoch": 1115, "lr": 7.065639472137555e-05}
{"train_loss": 0.15023300051689148, "global_step": 99244, "epoch": 1115, "lr": 7.065586676950575e-05}
{"train_loss": 0.11663201451301575, "global_step": 99245, "epoch": 1115, "lr": 7.065533881485903e-05}
{"train_loss": 0.1328858584165573, "global_step": 99246, "epoch": 1115, "lr": 7.065481085743544e-05}
{"train_loss": 0.16361139714717865, "global_step": 99247, "epoch": 1115, "lr": 7.065428289723508e-05}
{"train_loss": 0.12624084949493408, "global_step": 99248, "epoch": 1115, "lr": 7.065375493425797e-05}
{"train_loss": 0.1593530774116516, "global_step": 99249, "epoch": 1115, "lr": 7.065322696850424e-05}
{"train_loss": 0.1085917204618454, "global_step": 99250, "epoch": 1115, "lr": 7.065269899997394e-05}
{"train_loss": 0.15039309859275818, "global_step": 99251, "epoch": 1115, "lr": 7.065217102866711e-05}
{"train_loss": 0.21668373048305511, "global_step": 99252, "epoch": 1115, "lr": 7.065164305458386e-05}
{"train_loss": 0.1666502058506012, "global_step": 99253, "epoch": 1115, "lr": 7.065111507772424e-05}
{"train_loss": 0.18828924000263214, "global_step": 99254, "epoch": 1115, "lr": 7.065058709808835e-05}
{"train_loss": 0.1270720213651657, "global_step": 99255, "epoch": 1115, "lr": 7.065005911567623e-05}
{"train_loss": 0.16512231528759003, "global_step": 99256, "epoch": 1115, "lr": 7.064953113048796e-05}
{"train_loss": 0.16325537860393524, "global_step": 99257, "epoch": 1115, "lr": 7.06490031425236e-05}
{"train_loss": 0.17243199050426483, "global_step": 99258, "epoch": 1115, "lr": 7.064847515178325e-05}
{"train_loss": 0.19938606023788452, "global_step": 99259, "epoch": 1115, "lr": 7.064794715826697e-05}
{"train_loss": 0.08981865644454956, "global_step": 99260, "epoch": 1115, "lr": 7.064741916197481e-05}
{"train_loss": 0.1711178421974182, "global_step": 99261, "epoch": 1115, "lr": 7.064689116290687e-05}
{"train_loss": 0.0918215811252594, "global_step": 99262, "epoch": 1115, "lr": 7.06463631610632e-05}
{"train_loss": 0.14951711893081665, "global_step": 99263, "epoch": 1115, "lr": 7.064583515644388e-05}
{"train_loss": 0.10250242054462433, "global_step": 99264, "epoch": 1115, "lr": 7.064530714904896e-05}
{"train_loss": 0.18649804592132568, "global_step": 99265, "epoch": 1115, "lr": 7.064477913887855e-05}
{"train_loss": 0.12682120501995087, "global_step": 99266, "epoch": 1115, "lr": 7.064425112593271e-05}
{"train_loss": 0.18606655299663544, "global_step": 99267, "epoch": 1115, "lr": 7.064372311021147e-05}
{"train_loss": 0.09927273541688919, "global_step": 99268, "epoch": 1115, "lr": 7.064319509171497e-05}
{"train_loss": 0.20872445404529572, "global_step": 99269, "epoch": 1115, "lr": 7.064266707044323e-05}
{"train_loss": 0.18172036111354828, "global_step": 99270, "epoch": 1115, "lr": 7.064213904639631e-05}
{"train_loss": 0.1431204229593277, "global_step": 99271, "epoch": 1115, "lr": 7.064161101957434e-05}
{"train_loss": 0.1551465541124344, "global_step": 99272, "epoch": 1115, "lr": 7.064108298997733e-05}
{"train_loss": 0.2345934957265854, "global_step": 99273, "epoch": 1115, "lr": 7.06405549576054e-05}
{"train_loss": 0.17259837687015533, "global_step": 99274, "epoch": 1115, "lr": 7.06400269224586e-05}
{"train_loss": 0.13703332841396332, "global_step": 99275, "epoch": 1115, "lr": 7.063949888453698e-05}
{"train_loss": 0.20320147275924683, "global_step": 99276, "epoch": 1115, "lr": 7.063897084384065e-05}
{"train_loss": 0.15545959770679474, "global_step": 99277, "epoch": 1115, "lr": 7.063844280036965e-05}
{"train_loss": 0.1741667240858078, "global_step": 99278, "epoch": 1115, "lr": 7.063791475412407e-05}
{"train_loss": 0.14376899600028992, "global_step": 99279, "epoch": 1115, "lr": 7.063738670510397e-05}
{"train_loss": 0.1520986557006836, "global_step": 99280, "epoch": 1115, "lr": 7.063685865330942e-05}
{"train_loss": 0.26993387937545776, "global_step": 99281, "epoch": 1115, "lr": 7.06363305987405e-05}
{"train_loss": 0.16581623256206512, "global_step": 99282, "epoch": 1115, "lr": 7.063580254139727e-05}
{"train_loss": 0.2084222137928009, "global_step": 99283, "epoch": 1115, "lr": 7.063527448127983e-05}
{"train_loss": 0.25551891326904297, "global_step": 99284, "epoch": 1115, "lr": 7.063474641838821e-05}
{"train_loss": 0.0691756010055542, "global_step": 99285, "epoch": 1115, "lr": 7.063421835272251e-05}
{"train_loss": 0.1304253488779068, "global_step": 99286, "epoch": 1115, "lr": 7.063369028428278e-05}
{"train_loss": 0.12699179351329803, "global_step": 99287, "epoch": 1115, "lr": 7.063316221306912e-05}
{"train_loss": 0.1947958767414093, "global_step": 99288, "epoch": 1115, "lr": 7.063263413908156e-05}
{"train_loss": 0.17573198676109314, "global_step": 99289, "epoch": 1115, "lr": 7.06321060623202e-05}
{"train_loss": 0.1298012137413025, "global_step": 99290, "epoch": 1115, "lr": 7.063157798278512e-05}
{"train_loss": 0.16028697788715363, "global_step": 99291, "epoch": 1115, "lr": 7.063104990047636e-05}
{"train_loss": 0.17799164354801178, "global_step": 99292, "epoch": 1115, "lr": 7.063052181539402e-05}
{"train_loss": 0.1424277424812317, "global_step": 99293, "epoch": 1115, "lr": 7.062999372753814e-05}
{"train_loss": 0.1663377285003662, "global_step": 99294, "epoch": 1115, "lr": 7.062946563690882e-05}
{"train_loss": 0.1518893539905548, "global_step": 99295, "epoch": 1115, "lr": 7.062893754350613e-05}
{"train_loss": 0.24617090821266174, "global_step": 99296, "epoch": 1115, "lr": 7.062840944733011e-05}
{"train_loss": 0.20878265798091888, "global_step": 99297, "epoch": 1115, "lr": 7.062788134838086e-05}
{"train_loss": 0.13710622489452362, "global_step": 99298, "epoch": 1115, "lr": 7.062735324665843e-05}
{"train_loss": 0.13663703203201294, "global_step": 99299, "epoch": 1115, "lr": 7.062682514216293e-05}
{"train_loss": 0.15014001727104187, "global_step": 99300, "epoch": 1115, "lr": 7.06262970348944e-05}
{"train_loss": 0.16593118011951447, "global_step": 99301, "epoch": 1115, "lr": 7.06257689248529e-05}
{"train_loss": 0.23067443072795868, "global_step": 99302, "epoch": 1115, "lr": 7.062524081203852e-05}
{"train_loss": 0.21950504183769226, "global_step": 99303, "epoch": 1115, "lr": 7.062471269645133e-05}
{"train_loss": 0.2320331484079361, "global_step": 99304, "epoch": 1115, "lr": 7.062418457809141e-05}
{"train_loss": 0.182850182056427, "global_step": 99305, "epoch": 1115, "lr": 7.062365645695881e-05}
{"train_loss": 0.08090588450431824, "global_step": 99306, "epoch": 1115, "lr": 7.062312833305361e-05}
{"train_loss": 0.11835306882858276, "global_step": 99307, "epoch": 1115, "lr": 7.062260020637588e-05}
{"train_loss": 0.13621459901332855, "global_step": 99308, "epoch": 1115, "lr": 7.06220720769257e-05}
{"train_loss": 0.18524309992790222, "global_step": 99309, "epoch": 1115, "lr": 7.062154394470312e-05}
{"train_loss": 0.11188381910324097, "global_step": 99310, "epoch": 1115, "lr": 7.062101580970824e-05}
{"train_loss": 0.19102051854133606, "global_step": 99311, "epoch": 1115, "lr": 7.062048767194112e-05}
{"train_loss": 0.15508203208446503, "global_step": 99312, "epoch": 1115, "lr": 7.061995953140181e-05}
{"train_loss": 0.1971694380044937, "global_step": 99313, "epoch": 1115, "lr": 7.06194313880904e-05}
{"train_loss": 0.24101166427135468, "global_step": 99314, "epoch": 1115, "lr": 7.061890324200697e-05}
{"train_loss": 0.156291201710701, "global_step": 99315, "epoch": 1115, "lr": 7.061837509315157e-05}
{"train_loss": 0.10929036140441895, "global_step": 99316, "epoch": 1115, "lr": 7.061784694152427e-05}
{"train_loss": 0.21595589816570282, "global_step": 99317, "epoch": 1115, "lr": 7.061731878712516e-05}
{"train_loss": 0.12248360365629196, "global_step": 99318, "epoch": 1115, "lr": 7.061679062995432e-05}
{"train_loss": 0.1050257533788681, "global_step": 99319, "epoch": 1115, "lr": 7.06162624700118e-05}
{"train_loss": 0.1405491977930069, "global_step": 99320, "epoch": 1115, "lr": 7.061573430729767e-05}
{"train_loss": 0.241842582821846, "global_step": 99321, "epoch": 1115, "lr": 7.061520614181201e-05}
{"train_loss": 0.12353826314210892, "global_step": 99322, "epoch": 1115, "lr": 7.061467797355487e-05}
{"train_loss": 0.16090718899550063, "global_step": 99323, "epoch": 1115, "lr": 7.061414980252635e-05, "val_loss": 4.601212978363037, "train_action_mse_error": 24.33282470703125}
{"train_loss": 0.16756029427051544, "global_step": 99324, "epoch": 1116, "lr": 7.061362162872652e-05}
{"train_loss": 0.10740182548761368, "global_step": 99325, "epoch": 1116, "lr": 7.061309345215541e-05}
{"train_loss": 0.1432725340127945, "global_step": 99326, "epoch": 1116, "lr": 7.061256527281313e-05}
{"train_loss": 0.1778109073638916, "global_step": 99327, "epoch": 1116, "lr": 7.061203709069975e-05}
{"train_loss": 0.15912866592407227, "global_step": 99328, "epoch": 1116, "lr": 7.061150890581533e-05}
{"train_loss": 0.08805900812149048, "global_step": 99329, "epoch": 1116, "lr": 7.061098071815996e-05}
{"train_loss": 0.10234284400939941, "global_step": 99330, "epoch": 1116, "lr": 7.061045252773367e-05}
{"train_loss": 0.22214342653751373, "global_step": 99331, "epoch": 1116, "lr": 7.060992433453657e-05}
{"train_loss": 0.22932733595371246, "global_step": 99332, "epoch": 1116, "lr": 7.060939613856871e-05}
{"train_loss": 0.23792427778244019, "global_step": 99333, "epoch": 1116, "lr": 7.060886793983017e-05}
{"train_loss": 0.13774466514587402, "global_step": 99334, "epoch": 1116, "lr": 7.060833973832101e-05}
{"train_loss": 0.1849893033504486, "global_step": 99335, "epoch": 1116, "lr": 7.060781153404133e-05}
{"train_loss": 0.15513235330581665, "global_step": 99336, "epoch": 1116, "lr": 7.060728332699116e-05}
{"train_loss": 0.15599782764911652, "global_step": 99337, "epoch": 1116, "lr": 7.060675511717061e-05}
{"train_loss": 0.12477871030569077, "global_step": 99338, "epoch": 1116, "lr": 7.060622690457971e-05}
{"train_loss": 0.1876765489578247, "global_step": 99339, "epoch": 1116, "lr": 7.060569868921859e-05}
{"train_loss": 0.1231624186038971, "global_step": 99340, "epoch": 1116, "lr": 7.060517047108727e-05}
{"train_loss": 0.14073002338409424, "global_step": 99341, "epoch": 1116, "lr": 7.060464225018583e-05}
{"train_loss": 0.07900767773389816, "global_step": 99342, "epoch": 1116, "lr": 7.060411402651435e-05}
{"train_loss": 0.14881925284862518, "global_step": 99343, "epoch": 1116, "lr": 7.06035858000729e-05}
{"train_loss": 0.13740696012973785, "global_step": 99344, "epoch": 1116, "lr": 7.060305757086155e-05}
{"train_loss": 0.1194169744849205, "global_step": 99345, "epoch": 1116, "lr": 7.060252933888036e-05}
{"train_loss": 0.14218617975711823, "global_step": 99346, "epoch": 1116, "lr": 7.06020011041294e-05}
{"train_loss": 0.23578785359859467, "global_step": 99347, "epoch": 1116, "lr": 7.060147286660878e-05}
{"train_loss": 0.15524031221866608, "global_step": 99348, "epoch": 1116, "lr": 7.060094462631853e-05}
{"train_loss": 0.1729135811328888, "global_step": 99349, "epoch": 1116, "lr": 7.060041638325873e-05}
{"train_loss": 0.16945336759090424, "global_step": 99350, "epoch": 1116, "lr": 7.059988813742946e-05}
{"train_loss": 0.16487880051136017, "global_step": 99351, "epoch": 1116, "lr": 7.059935988883078e-05}
{"train_loss": 0.1689332127571106, "global_step": 99352, "epoch": 1116, "lr": 7.059883163746278e-05}
{"train_loss": 0.12308672070503235, "global_step": 99353, "epoch": 1116, "lr": 7.05983033833255e-05}
{"train_loss": 0.17764830589294434, "global_step": 99354, "epoch": 1116, "lr": 7.059777512641904e-05}
{"train_loss": 0.16674508154392242, "global_step": 99355, "epoch": 1116, "lr": 7.059724686674347e-05}
{"train_loss": 0.14940114319324493, "global_step": 99356, "epoch": 1116, "lr": 7.059671860429884e-05}
{"train_loss": 0.1387016475200653, "global_step": 99357, "epoch": 1116, "lr": 7.059619033908522e-05}
{"train_loss": 0.14357852935791016, "global_step": 99358, "epoch": 1116, "lr": 7.059566207110273e-05}
{"train_loss": 0.1861657053232193, "global_step": 99359, "epoch": 1116, "lr": 7.059513380035137e-05}
{"train_loss": 0.18967248499393463, "global_step": 99360, "epoch": 1116, "lr": 7.059460552683125e-05}
{"train_loss": 0.1989513486623764, "global_step": 99361, "epoch": 1116, "lr": 7.059407725054245e-05}
{"train_loss": 0.1280142217874527, "global_step": 99362, "epoch": 1116, "lr": 7.059354897148503e-05}
{"train_loss": 0.14040258526802063, "global_step": 99363, "epoch": 1116, "lr": 7.059302068965904e-05}
{"train_loss": 0.1586553007364273, "global_step": 99364, "epoch": 1116, "lr": 7.059249240506458e-05}
{"train_loss": 0.20578117668628693, "global_step": 99365, "epoch": 1116, "lr": 7.059196411770172e-05}
{"train_loss": 0.15092812478542328, "global_step": 99366, "epoch": 1116, "lr": 7.05914358275705e-05}
{"train_loss": 0.22578899562358856, "global_step": 99367, "epoch": 1116, "lr": 7.059090753467103e-05}
{"train_loss": 0.15924426913261414, "global_step": 99368, "epoch": 1116, "lr": 7.059037923900335e-05}
{"train_loss": 0.19246502220630646, "global_step": 99369, "epoch": 1116, "lr": 7.058985094056755e-05}
{"train_loss": 0.1908484399318695, "global_step": 99370, "epoch": 1116, "lr": 7.05893226393637e-05}
{"train_loss": 0.18711724877357483, "global_step": 99371, "epoch": 1116, "lr": 7.058879433539185e-05}
{"train_loss": 0.07272088527679443, "global_step": 99372, "epoch": 1116, "lr": 7.058826602865209e-05}
{"train_loss": 0.1546494960784912, "global_step": 99373, "epoch": 1116, "lr": 7.058773771914451e-05}
{"train_loss": 0.14439105987548828, "global_step": 99374, "epoch": 1116, "lr": 7.058720940686913e-05}
{"train_loss": 0.19478514790534973, "global_step": 99375, "epoch": 1116, "lr": 7.058668109182608e-05}
{"train_loss": 0.11997310817241669, "global_step": 99376, "epoch": 1116, "lr": 7.05861527740154e-05}
{"train_loss": 0.1658325493335724, "global_step": 99377, "epoch": 1116, "lr": 7.058562445343715e-05}
{"train_loss": 0.1851574033498764, "global_step": 99378, "epoch": 1116, "lr": 7.058509613009142e-05}
{"train_loss": 0.18003296852111816, "global_step": 99379, "epoch": 1116, "lr": 7.058456780397827e-05}
{"train_loss": 0.1342422366142273, "global_step": 99380, "epoch": 1116, "lr": 7.058403947509778e-05}
{"train_loss": 0.2148772031068802, "global_step": 99381, "epoch": 1116, "lr": 7.058351114345e-05}
{"train_loss": 0.19765923917293549, "global_step": 99382, "epoch": 1116, "lr": 7.058298280903505e-05}
{"train_loss": 0.14070487022399902, "global_step": 99383, "epoch": 1116, "lr": 7.058245447185294e-05}
{"train_loss": 0.173237144947052, "global_step": 99384, "epoch": 1116, "lr": 7.058192613190379e-05}
{"train_loss": 0.17083770036697388, "global_step": 99385, "epoch": 1116, "lr": 7.058139778918764e-05}
{"train_loss": 0.1120331883430481, "global_step": 99386, "epoch": 1116, "lr": 7.058086944370456e-05}
{"train_loss": 0.12371648848056793, "global_step": 99387, "epoch": 1116, "lr": 7.058034109545466e-05}
{"train_loss": 0.21237397193908691, "global_step": 99388, "epoch": 1116, "lr": 7.057981274443797e-05}
{"train_loss": 0.1562124639749527, "global_step": 99389, "epoch": 1116, "lr": 7.057928439065457e-05}
{"train_loss": 0.12690813839435577, "global_step": 99390, "epoch": 1116, "lr": 7.057875603410453e-05}
{"train_loss": 0.22643545269966125, "global_step": 99391, "epoch": 1116, "lr": 7.057822767478795e-05}
{"train_loss": 0.1544632762670517, "global_step": 99392, "epoch": 1116, "lr": 7.057769931270486e-05}
{"train_loss": 0.14431610703468323, "global_step": 99393, "epoch": 1116, "lr": 7.057717094785536e-05}
{"train_loss": 0.12802979350090027, "global_step": 99394, "epoch": 1116, "lr": 7.05766425802395e-05}
{"train_loss": 0.13462930917739868, "global_step": 99395, "epoch": 1116, "lr": 7.057611420985737e-05}
{"train_loss": 0.18448933959007263, "global_step": 99396, "epoch": 1116, "lr": 7.057558583670903e-05}
{"train_loss": 0.09773516654968262, "global_step": 99397, "epoch": 1116, "lr": 7.057505746079455e-05}
{"train_loss": 0.16640107333660126, "global_step": 99398, "epoch": 1116, "lr": 7.057452908211401e-05}
{"train_loss": 0.17503759264945984, "global_step": 99399, "epoch": 1116, "lr": 7.057400070066748e-05}
{"train_loss": 0.1396336555480957, "global_step": 99400, "epoch": 1116, "lr": 7.057347231645502e-05}
{"train_loss": 0.06315551698207855, "global_step": 99401, "epoch": 1116, "lr": 7.05729439294767e-05}
{"train_loss": 0.16524772346019745, "global_step": 99402, "epoch": 1116, "lr": 7.05724155397326e-05}
{"train_loss": 0.12571513652801514, "global_step": 99403, "epoch": 1116, "lr": 7.05718871472228e-05}
{"train_loss": 0.1827923208475113, "global_step": 99404, "epoch": 1116, "lr": 7.057135875194736e-05}
{"train_loss": 0.1487007588148117, "global_step": 99405, "epoch": 1116, "lr": 7.057083035390633e-05}
{"train_loss": 0.2793973386287689, "global_step": 99406, "epoch": 1116, "lr": 7.057030195309982e-05}
{"train_loss": 0.22873011231422424, "global_step": 99407, "epoch": 1116, "lr": 7.056977354952787e-05}
{"train_loss": 0.19431759417057037, "global_step": 99408, "epoch": 1116, "lr": 7.056924514319059e-05}
{"train_loss": 0.16301274299621582, "global_step": 99409, "epoch": 1116, "lr": 7.0568716734088e-05}
{"train_loss": 0.12466759979724884, "global_step": 99410, "epoch": 1116, "lr": 7.056818832222022e-05}
{"train_loss": 0.17239250242710114, "global_step": 99411, "epoch": 1116, "lr": 7.056765990758728e-05}
{"train_loss": 0.1603360859195838, "global_step": 99412, "epoch": 1116, "lr": 7.056713149018928e-05, "val_loss": 4.605827808380127}
{"train_loss": 0.12254074215888977, "global_step": 99413, "epoch": 1117, "lr": 7.056660307002626e-05}
{"train_loss": 0.1876436471939087, "global_step": 99414, "epoch": 1117, "lr": 7.056607464709835e-05}
{"train_loss": 0.20570722222328186, "global_step": 99415, "epoch": 1117, "lr": 7.056554622140555e-05}
{"train_loss": 0.21847185492515564, "global_step": 99416, "epoch": 1117, "lr": 7.056501779294797e-05}
{"train_loss": 0.14384157955646515, "global_step": 99417, "epoch": 1117, "lr": 7.056448936172569e-05}
{"train_loss": 0.1676514446735382, "global_step": 99418, "epoch": 1117, "lr": 7.056396092773874e-05}
{"train_loss": 0.22938451170921326, "global_step": 99419, "epoch": 1117, "lr": 7.056343249098722e-05}
{"train_loss": 0.11586920917034149, "global_step": 99420, "epoch": 1117, "lr": 7.056290405147121e-05}
{"train_loss": 0.16495855152606964, "global_step": 99421, "epoch": 1117, "lr": 7.056237560919077e-05}
{"train_loss": 0.12557299435138702, "global_step": 99422, "epoch": 1117, "lr": 7.056184716414595e-05}
{"train_loss": 0.1008477658033371, "global_step": 99423, "epoch": 1117, "lr": 7.056131871633686e-05}
{"train_loss": 0.15307915210723877, "global_step": 99424, "epoch": 1117, "lr": 7.056079026576354e-05}
{"train_loss": 0.15083478391170502, "global_step": 99425, "epoch": 1117, "lr": 7.056026181242607e-05}
{"train_loss": 0.16369962692260742, "global_step": 99426, "epoch": 1117, "lr": 7.055973335632452e-05}
{"train_loss": 0.12146264314651489, "global_step": 99427, "epoch": 1117, "lr": 7.055920489745897e-05}
{"train_loss": 0.2387048304080963, "global_step": 99428, "epoch": 1117, "lr": 7.05586764358295e-05}
{"train_loss": 0.13589374721050262, "global_step": 99429, "epoch": 1117, "lr": 7.055814797143615e-05}
{"train_loss": 0.1933661550283432, "global_step": 99430, "epoch": 1117, "lr": 7.055761950427902e-05}
{"train_loss": 0.1266813576221466, "global_step": 99431, "epoch": 1117, "lr": 7.055709103435815e-05}
{"train_loss": 0.17866356670856476, "global_step": 99432, "epoch": 1117, "lr": 7.055656256167364e-05}
{"train_loss": 0.19219675660133362, "global_step": 99433, "epoch": 1117, "lr": 7.055603408622555e-05}
{"train_loss": 0.2207607477903366, "global_step": 99434, "epoch": 1117, "lr": 7.055550560801396e-05}
{"train_loss": 0.24002204835414886, "global_step": 99435, "epoch": 1117, "lr": 7.055497712703894e-05}
{"train_loss": 0.20273713767528534, "global_step": 99436, "epoch": 1117, "lr": 7.055444864330053e-05}
{"train_loss": 0.2386465072631836, "global_step": 99437, "epoch": 1117, "lr": 7.055392015679883e-05}
{"train_loss": 0.139007568359375, "global_step": 99438, "epoch": 1117, "lr": 7.055339166753392e-05}
{"train_loss": 0.106133371591568, "global_step": 99439, "epoch": 1117, "lr": 7.055286317550584e-05}
{"train_loss": 0.15343543887138367, "global_step": 99440, "epoch": 1117, "lr": 7.05523346807147e-05}
{"train_loss": 0.18372103571891785, "global_step": 99441, "epoch": 1117, "lr": 7.055180618316054e-05}
{"train_loss": 0.22628293931484222, "global_step": 99442, "epoch": 1117, "lr": 7.055127768284342e-05}
{"train_loss": 0.1821734607219696, "global_step": 99443, "epoch": 1117, "lr": 7.055074917976343e-05}
{"train_loss": 0.17502261698246002, "global_step": 99444, "epoch": 1117, "lr": 7.055022067392066e-05}
{"train_loss": 0.12288276851177216, "global_step": 99445, "epoch": 1117, "lr": 7.054969216531517e-05}
{"train_loss": 0.21376241743564606, "global_step": 99446, "epoch": 1117, "lr": 7.054916365394702e-05}
{"train_loss": 0.10414622724056244, "global_step": 99447, "epoch": 1117, "lr": 7.054863513981627e-05}
{"train_loss": 0.16113847494125366, "global_step": 99448, "epoch": 1117, "lr": 7.054810662292302e-05}
{"train_loss": 0.13092312216758728, "global_step": 99449, "epoch": 1117, "lr": 7.054757810326732e-05}
{"train_loss": 0.17814569175243378, "global_step": 99450, "epoch": 1117, "lr": 7.054704958084926e-05}
{"train_loss": 0.26541683077812195, "global_step": 99451, "epoch": 1117, "lr": 7.054652105566889e-05}
{"train_loss": 0.06942318379878998, "global_step": 99452, "epoch": 1117, "lr": 7.05459925277263e-05}
{"train_loss": 0.16769036650657654, "global_step": 99453, "epoch": 1117, "lr": 7.054546399702155e-05}
{"train_loss": 0.17256483435630798, "global_step": 99454, "epoch": 1117, "lr": 7.054493546355469e-05}
{"train_loss": 0.16085895895957947, "global_step": 99455, "epoch": 1117, "lr": 7.054440692732585e-05}
{"train_loss": 0.12456659227609634, "global_step": 99456, "epoch": 1117, "lr": 7.054387838833506e-05}
{"train_loss": 0.22779500484466553, "global_step": 99457, "epoch": 1117, "lr": 7.054334984658237e-05}
{"train_loss": 0.22222645580768585, "global_step": 99458, "epoch": 1117, "lr": 7.05428213020679e-05}
{"train_loss": 0.20110824704170227, "global_step": 99459, "epoch": 1117, "lr": 7.054229275479168e-05}
{"train_loss": 0.15603385865688324, "global_step": 99460, "epoch": 1117, "lr": 7.054176420475381e-05}
{"train_loss": 0.14100879430770874, "global_step": 99461, "epoch": 1117, "lr": 7.054123565195434e-05}
{"train_loss": 0.19604533910751343, "global_step": 99462, "epoch": 1117, "lr": 7.054070709639336e-05}
{"train_loss": 0.20695938169956207, "global_step": 99463, "epoch": 1117, "lr": 7.054017853807094e-05}
{"train_loss": 0.1766633689403534, "global_step": 99464, "epoch": 1117, "lr": 7.053964997698712e-05}
{"train_loss": 0.12806099653244019, "global_step": 99465, "epoch": 1117, "lr": 7.0539121413142e-05}
{"train_loss": 0.12452816218137741, "global_step": 99466, "epoch": 1117, "lr": 7.053859284653567e-05}
{"train_loss": 0.08798963576555252, "global_step": 99467, "epoch": 1117, "lr": 7.053806427716816e-05}
{"train_loss": 0.11202372610569, "global_step": 99468, "epoch": 1117, "lr": 7.053753570503957e-05}
{"train_loss": 0.21715569496154785, "global_step": 99469, "epoch": 1117, "lr": 7.053700713014994e-05}
{"train_loss": 0.21134813129901886, "global_step": 99470, "epoch": 1117, "lr": 7.053647855249936e-05}
{"train_loss": 0.1603815257549286, "global_step": 99471, "epoch": 1117, "lr": 7.05359499720879e-05}
{"train_loss": 0.16486628353595734, "global_step": 99472, "epoch": 1117, "lr": 7.053542138891565e-05}
{"train_loss": 0.18099988996982574, "global_step": 99473, "epoch": 1117, "lr": 7.053489280298264e-05}
{"train_loss": 0.20787324011325836, "global_step": 99474, "epoch": 1117, "lr": 7.053436421428899e-05}
{"train_loss": 0.1463909149169922, "global_step": 99475, "epoch": 1117, "lr": 7.053383562283473e-05}
{"train_loss": 0.1838182508945465, "global_step": 99476, "epoch": 1117, "lr": 7.053330702861996e-05}
{"train_loss": 0.1074729859828949, "global_step": 99477, "epoch": 1117, "lr": 7.053277843164471e-05}
{"train_loss": 0.1995902806520462, "global_step": 99478, "epoch": 1117, "lr": 7.05322498319091e-05}
{"train_loss": 0.14459194242954254, "global_step": 99479, "epoch": 1117, "lr": 7.053172122941317e-05}
{"train_loss": 0.18840965628623962, "global_step": 99480, "epoch": 1117, "lr": 7.053119262415699e-05}
{"train_loss": 0.1381409764289856, "global_step": 99481, "epoch": 1117, "lr": 7.053066401614066e-05}
{"train_loss": 0.1823086142539978, "global_step": 99482, "epoch": 1117, "lr": 7.053013540536423e-05}
{"train_loss": 0.2587020695209503, "global_step": 99483, "epoch": 1117, "lr": 7.052960679182776e-05}
{"train_loss": 0.13306397199630737, "global_step": 99484, "epoch": 1117, "lr": 7.052907817553134e-05}
{"train_loss": 0.15777139365673065, "global_step": 99485, "epoch": 1117, "lr": 7.052854955647506e-05}
{"train_loss": 0.14132580161094666, "global_step": 99486, "epoch": 1117, "lr": 7.052802093465894e-05}
{"train_loss": 0.14266270399093628, "global_step": 99487, "epoch": 1117, "lr": 7.052749231008308e-05}
{"train_loss": 0.12821683287620544, "global_step": 99488, "epoch": 1117, "lr": 7.052696368274756e-05}
{"train_loss": 0.20328117907047272, "global_step": 99489, "epoch": 1117, "lr": 7.052643505265243e-05}
{"train_loss": 0.12646827101707458, "global_step": 99490, "epoch": 1117, "lr": 7.052590641979777e-05}
{"train_loss": 0.23606464266777039, "global_step": 99491, "epoch": 1117, "lr": 7.052537778418365e-05}
{"train_loss": 0.21396832168102264, "global_step": 99492, "epoch": 1117, "lr": 7.052484914581016e-05}
{"train_loss": 0.13937263190746307, "global_step": 99493, "epoch": 1117, "lr": 7.052432050467734e-05}
{"train_loss": 0.11454804986715317, "global_step": 99494, "epoch": 1117, "lr": 7.052379186078528e-05}
{"train_loss": 0.12253647297620773, "global_step": 99495, "epoch": 1117, "lr": 7.052326321413404e-05}
{"train_loss": 0.12509186565876007, "global_step": 99496, "epoch": 1117, "lr": 7.052273456472372e-05}
{"train_loss": 0.1322309374809265, "global_step": 99497, "epoch": 1117, "lr": 7.052220591255436e-05}
{"train_loss": 0.20868371427059174, "global_step": 99498, "epoch": 1117, "lr": 7.052167725762601e-05}
{"train_loss": 0.1939755231142044, "global_step": 99499, "epoch": 1117, "lr": 7.052114859993879e-05}
{"train_loss": 0.14814062416553497, "global_step": 99500, "epoch": 1117, "lr": 7.052061993949275e-05}
{"train_loss": 0.16760606754027055, "global_step": 99501, "epoch": 1117, "lr": 7.052009127628797e-05, "val_loss": 4.6730780601501465}
{"train_loss": 0.1195639818906784, "global_step": 99502, "epoch": 1118, "lr": 7.05195626103245e-05}
{"train_loss": 0.22195196151733398, "global_step": 99503, "epoch": 1118, "lr": 7.051903394160244e-05}
{"train_loss": 0.2095441073179245, "global_step": 99504, "epoch": 1118, "lr": 7.051850527012184e-05}
{"train_loss": 0.15335732698440552, "global_step": 99505, "epoch": 1118, "lr": 7.051797659588278e-05}
{"train_loss": 0.12615244090557098, "global_step": 99506, "epoch": 1118, "lr": 7.051744791888532e-05}
{"train_loss": 0.22738458216190338, "global_step": 99507, "epoch": 1118, "lr": 7.051691923912956e-05}
{"train_loss": 0.1373949497938156, "global_step": 99508, "epoch": 1118, "lr": 7.051639055661552e-05}
{"train_loss": 0.11850754171609879, "global_step": 99509, "epoch": 1118, "lr": 7.051586187134331e-05}
{"train_loss": 0.1428578644990921, "global_step": 99510, "epoch": 1118, "lr": 7.051533318331301e-05}
{"train_loss": 0.15160207450389862, "global_step": 99511, "epoch": 1118, "lr": 7.051480449252465e-05}
{"train_loss": 0.1374932825565338, "global_step": 99512, "epoch": 1118, "lr": 7.051427579897835e-05}
{"train_loss": 0.13625678420066833, "global_step": 99513, "epoch": 1118, "lr": 7.051374710267415e-05}
{"train_loss": 0.13233362138271332, "global_step": 99514, "epoch": 1118, "lr": 7.051321840361211e-05}
{"train_loss": 0.18848037719726562, "global_step": 99515, "epoch": 1118, "lr": 7.051268970179233e-05}
{"train_loss": 0.16406919062137604, "global_step": 99516, "epoch": 1118, "lr": 7.051216099721486e-05}
{"train_loss": 0.1519084870815277, "global_step": 99517, "epoch": 1118, "lr": 7.051163228987979e-05}
{"train_loss": 0.19585837423801422, "global_step": 99518, "epoch": 1118, "lr": 7.051110357978718e-05}
{"train_loss": 0.15050913393497467, "global_step": 99519, "epoch": 1118, "lr": 7.05105748669371e-05}
{"train_loss": 0.20408043265342712, "global_step": 99520, "epoch": 1118, "lr": 7.051004615132961e-05}
{"train_loss": 0.12756755948066711, "global_step": 99521, "epoch": 1118, "lr": 7.050951743296481e-05}
{"train_loss": 0.15816284716129303, "global_step": 99522, "epoch": 1118, "lr": 7.050898871184275e-05}
{"train_loss": 0.10995328426361084, "global_step": 99523, "epoch": 1118, "lr": 7.050845998796351e-05}
{"train_loss": 0.20382876694202423, "global_step": 99524, "epoch": 1118, "lr": 7.050793126132715e-05}
{"train_loss": 0.16396914422512054, "global_step": 99525, "epoch": 1118, "lr": 7.050740253193376e-05}
{"train_loss": 0.24409016966819763, "global_step": 99526, "epoch": 1118, "lr": 7.05068737997834e-05}
{"train_loss": 0.11326298117637634, "global_step": 99527, "epoch": 1118, "lr": 7.050634506487613e-05}
{"train_loss": 0.17095522582530975, "global_step": 99528, "epoch": 1118, "lr": 7.050581632721203e-05}
{"train_loss": 0.13630414009094238, "global_step": 99529, "epoch": 1118, "lr": 7.050528758679119e-05}
{"train_loss": 0.17043381929397583, "global_step": 99530, "epoch": 1118, "lr": 7.050475884361365e-05}
{"train_loss": 0.19940578937530518, "global_step": 99531, "epoch": 1118, "lr": 7.05042300976795e-05}
{"train_loss": 0.20771914720535278, "global_step": 99532, "epoch": 1118, "lr": 7.050370134898881e-05}
{"train_loss": 0.16282691061496735, "global_step": 99533, "epoch": 1118, "lr": 7.050317259754165e-05}
{"train_loss": 0.1346430480480194, "global_step": 99534, "epoch": 1118, "lr": 7.050264384333807e-05}
{"train_loss": 0.19569753110408783, "global_step": 99535, "epoch": 1118, "lr": 7.050211508637817e-05}
{"train_loss": 0.14363686740398407, "global_step": 99536, "epoch": 1118, "lr": 7.050158632666201e-05}
{"train_loss": 0.13973361253738403, "global_step": 99537, "epoch": 1118, "lr": 7.050105756418966e-05}
{"train_loss": 0.22241851687431335, "global_step": 99538, "epoch": 1118, "lr": 7.05005287989612e-05}
{"train_loss": 0.20126979053020477, "global_step": 99539, "epoch": 1118, "lr": 7.050000003097669e-05}
{"train_loss": 0.19113552570343018, "global_step": 99540, "epoch": 1118, "lr": 7.049947126023619e-05}
{"train_loss": 0.16833671927452087, "global_step": 99541, "epoch": 1118, "lr": 7.04989424867398e-05}
{"train_loss": 0.1881885975599289, "global_step": 99542, "epoch": 1118, "lr": 7.049841371048758e-05}
{"train_loss": 0.12159907072782516, "global_step": 99543, "epoch": 1118, "lr": 7.04978849314796e-05}
{"train_loss": 0.11727648228406906, "global_step": 99544, "epoch": 1118, "lr": 7.049735614971592e-05}
{"train_loss": 0.22038304805755615, "global_step": 99545, "epoch": 1118, "lr": 7.049682736519663e-05}
{"train_loss": 0.1851864457130432, "global_step": 99546, "epoch": 1118, "lr": 7.049629857792178e-05}
{"train_loss": 0.16684816777706146, "global_step": 99547, "epoch": 1118, "lr": 7.049576978789145e-05}
{"train_loss": 0.13741691410541534, "global_step": 99548, "epoch": 1118, "lr": 7.049524099510573e-05}
{"train_loss": 0.08799844235181808, "global_step": 99549, "epoch": 1118, "lr": 7.049471219956467e-05}
{"train_loss": 0.17633220553398132, "global_step": 99550, "epoch": 1118, "lr": 7.049418340126834e-05}
{"train_loss": 0.15902772545814514, "global_step": 99551, "epoch": 1118, "lr": 7.049365460021682e-05}
{"train_loss": 0.18740376830101013, "global_step": 99552, "epoch": 1118, "lr": 7.049312579641017e-05}
{"train_loss": 0.18929509818553925, "global_step": 99553, "epoch": 1118, "lr": 7.049259698984848e-05}
{"train_loss": 0.15365168452262878, "global_step": 99554, "epoch": 1118, "lr": 7.04920681805318e-05}
{"train_loss": 0.14541321992874146, "global_step": 99555, "epoch": 1118, "lr": 7.049153936846023e-05}
{"train_loss": 0.1696605235338211, "global_step": 99556, "epoch": 1118, "lr": 7.04910105536338e-05}
{"train_loss": 0.11975277215242386, "global_step": 99557, "epoch": 1118, "lr": 7.049048173605262e-05}
{"train_loss": 0.11901753395795822, "global_step": 99558, "epoch": 1118, "lr": 7.048995291571673e-05}
{"train_loss": 0.19635671377182007, "global_step": 99559, "epoch": 1118, "lr": 7.048942409262622e-05}
{"train_loss": 0.22694645822048187, "global_step": 99560, "epoch": 1118, "lr": 7.048889526678117e-05}
{"train_loss": 0.16982945799827576, "global_step": 99561, "epoch": 1118, "lr": 7.048836643818162e-05}
{"train_loss": 0.17732787132263184, "global_step": 99562, "epoch": 1118, "lr": 7.048783760682767e-05}
{"train_loss": 0.16280989348888397, "global_step": 99563, "epoch": 1118, "lr": 7.048730877271937e-05}
{"train_loss": 0.18524302542209625, "global_step": 99564, "epoch": 1118, "lr": 7.04867799358568e-05}
{"train_loss": 0.11995852738618851, "global_step": 99565, "epoch": 1118, "lr": 7.048625109624003e-05}
{"train_loss": 0.11022815108299255, "global_step": 99566, "epoch": 1118, "lr": 7.048572225386915e-05}
{"train_loss": 0.13320019841194153, "global_step": 99567, "epoch": 1118, "lr": 7.048519340874421e-05}
{"train_loss": 0.14908312261104584, "global_step": 99568, "epoch": 1118, "lr": 7.048466456086529e-05}
{"train_loss": 0.15310724079608917, "global_step": 99569, "epoch": 1118, "lr": 7.048413571023245e-05}
{"train_loss": 0.09944655001163483, "global_step": 99570, "epoch": 1118, "lr": 7.048360685684576e-05}
{"train_loss": 0.132597416639328, "global_step": 99571, "epoch": 1118, "lr": 7.048307800070531e-05}
{"train_loss": 0.20331983268260956, "global_step": 99572, "epoch": 1118, "lr": 7.048254914181115e-05}
{"train_loss": 0.18919242918491364, "global_step": 99573, "epoch": 1118, "lr": 7.048202028016337e-05}
{"train_loss": 0.21749824285507202, "global_step": 99574, "epoch": 1118, "lr": 7.048149141576202e-05}
{"train_loss": 0.14715230464935303, "global_step": 99575, "epoch": 1118, "lr": 7.048096254860719e-05}
{"train_loss": 0.07113408297300339, "global_step": 99576, "epoch": 1118, "lr": 7.048043367869894e-05}
{"train_loss": 0.15953759849071503, "global_step": 99577, "epoch": 1118, "lr": 7.047990480603736e-05}
{"train_loss": 0.19312532246112823, "global_step": 99578, "epoch": 1118, "lr": 7.047937593062249e-05}
{"train_loss": 0.1650097817182541, "global_step": 99579, "epoch": 1118, "lr": 7.047884705245442e-05}
{"train_loss": 0.15648579597473145, "global_step": 99580, "epoch": 1118, "lr": 7.047831817153323e-05}
{"train_loss": 0.1684054285287857, "global_step": 99581, "epoch": 1118, "lr": 7.047778928785897e-05}
{"train_loss": 0.2352822870016098, "global_step": 99582, "epoch": 1118, "lr": 7.047726040143172e-05}
{"train_loss": 0.10823281854391098, "global_step": 99583, "epoch": 1118, "lr": 7.047673151225156e-05}
{"train_loss": 0.21508288383483887, "global_step": 99584, "epoch": 1118, "lr": 7.047620262031855e-05}
{"train_loss": 0.1390639692544937, "global_step": 99585, "epoch": 1118, "lr": 7.047567372563276e-05}
{"train_loss": 0.09241122752428055, "global_step": 99586, "epoch": 1118, "lr": 7.047514482819427e-05}
{"train_loss": 0.24887600541114807, "global_step": 99587, "epoch": 1118, "lr": 7.047461592800314e-05}
{"train_loss": 0.2021637260913849, "global_step": 99588, "epoch": 1118, "lr": 7.047408702505946e-05}
{"train_loss": 0.12176045030355453, "global_step": 99589, "epoch": 1118, "lr": 7.047355811936327e-05}
{"train_loss": 0.16250154099772485, "global_step": 99590, "epoch": 1118, "lr": 7.047302921091468e-05, "val_loss": 4.753063201904297}
{"train_loss": 0.14359529316425323, "global_step": 99591, "epoch": 1119, "lr": 7.047250029971372e-05}
{"train_loss": 0.17231710255146027, "global_step": 99592, "epoch": 1119, "lr": 7.04719713857605e-05}
{"train_loss": 0.1975240707397461, "global_step": 99593, "epoch": 1119, "lr": 7.047144246905506e-05}
{"train_loss": 0.20043693482875824, "global_step": 99594, "epoch": 1119, "lr": 7.04709135495975e-05}
{"train_loss": 0.09710961580276489, "global_step": 99595, "epoch": 1119, "lr": 7.047038462738785e-05}
{"train_loss": 0.18988479673862457, "global_step": 99596, "epoch": 1119, "lr": 7.046985570242623e-05}
{"train_loss": 0.16413983702659607, "global_step": 99597, "epoch": 1119, "lr": 7.046932677471267e-05}
{"train_loss": 0.15209488570690155, "global_step": 99598, "epoch": 1119, "lr": 7.046879784424727e-05}
{"train_loss": 0.1098928451538086, "global_step": 99599, "epoch": 1119, "lr": 7.046826891103008e-05}
{"train_loss": 0.19419723749160767, "global_step": 99600, "epoch": 1119, "lr": 7.04677399750612e-05}
{"train_loss": 0.20047760009765625, "global_step": 99601, "epoch": 1119, "lr": 7.046721103634067e-05}
{"train_loss": 0.11141077429056168, "global_step": 99602, "epoch": 1119, "lr": 7.046668209486857e-05}
{"train_loss": 0.10097021609544754, "global_step": 99603, "epoch": 1119, "lr": 7.046615315064497e-05}
{"train_loss": 0.21981269121170044, "global_step": 99604, "epoch": 1119, "lr": 7.046562420366995e-05}
{"train_loss": 0.21350128948688507, "global_step": 99605, "epoch": 1119, "lr": 7.046509525394358e-05}
{"train_loss": 0.1375861018896103, "global_step": 99606, "epoch": 1119, "lr": 7.046456630146592e-05}
{"train_loss": 0.24125343561172485, "global_step": 99607, "epoch": 1119, "lr": 7.046403734623706e-05}
{"train_loss": 0.18046024441719055, "global_step": 99608, "epoch": 1119, "lr": 7.046350838825705e-05}
{"train_loss": 0.166758194565773, "global_step": 99609, "epoch": 1119, "lr": 7.046297942752597e-05}
{"train_loss": 0.17999117076396942, "global_step": 99610, "epoch": 1119, "lr": 7.046245046404389e-05}
{"train_loss": 0.10271808505058289, "global_step": 99611, "epoch": 1119, "lr": 7.04619214978109e-05}
{"train_loss": 0.2215212732553482, "global_step": 99612, "epoch": 1119, "lr": 7.046139252882704e-05}
{"train_loss": 0.1781260073184967, "global_step": 99613, "epoch": 1119, "lr": 7.04608635570924e-05}
{"train_loss": 0.16653376817703247, "global_step": 99614, "epoch": 1119, "lr": 7.046033458260704e-05}
{"train_loss": 0.1852179914712906, "global_step": 99615, "epoch": 1119, "lr": 7.045980560537104e-05}
{"train_loss": 0.18738052248954773, "global_step": 99616, "epoch": 1119, "lr": 7.045927662538446e-05}
{"train_loss": 0.16007918119430542, "global_step": 99617, "epoch": 1119, "lr": 7.04587476426474e-05}
{"train_loss": 0.1416502743959427, "global_step": 99618, "epoch": 1119, "lr": 7.045821865715989e-05}
{"train_loss": 0.1289599984884262, "global_step": 99619, "epoch": 1119, "lr": 7.045768966892204e-05}
{"train_loss": 0.18229222297668457, "global_step": 99620, "epoch": 1119, "lr": 7.045716067793388e-05}
{"train_loss": 0.09541623294353485, "global_step": 99621, "epoch": 1119, "lr": 7.045663168419553e-05}
{"train_loss": 0.21775920689105988, "global_step": 99622, "epoch": 1119, "lr": 7.045610268770703e-05}
{"train_loss": 0.1410568356513977, "global_step": 99623, "epoch": 1119, "lr": 7.045557368846845e-05}
{"train_loss": 0.12061990052461624, "global_step": 99624, "epoch": 1119, "lr": 7.045504468647986e-05}
{"train_loss": 0.16753897070884705, "global_step": 99625, "epoch": 1119, "lr": 7.045451568174135e-05}
{"train_loss": 0.12612998485565186, "global_step": 99626, "epoch": 1119, "lr": 7.045398667425297e-05}
{"train_loss": 0.21027234196662903, "global_step": 99627, "epoch": 1119, "lr": 7.045345766401481e-05}
{"train_loss": 0.22107283771038055, "global_step": 99628, "epoch": 1119, "lr": 7.045292865102692e-05}
{"train_loss": 0.13288213312625885, "global_step": 99629, "epoch": 1119, "lr": 7.04523996352894e-05}
{"train_loss": 0.11319949477910995, "global_step": 99630, "epoch": 1119, "lr": 7.045187061680229e-05}
{"train_loss": 0.21464383602142334, "global_step": 99631, "epoch": 1119, "lr": 7.045134159556568e-05}
{"train_loss": 0.07483623176813126, "global_step": 99632, "epoch": 1119, "lr": 7.045081257157963e-05}
{"train_loss": 0.1618047058582306, "global_step": 99633, "epoch": 1119, "lr": 7.045028354484424e-05}
{"train_loss": 0.18069154024124146, "global_step": 99634, "epoch": 1119, "lr": 7.044975451535955e-05}
{"train_loss": 0.09998942911624908, "global_step": 99635, "epoch": 1119, "lr": 7.044922548312562e-05}
{"train_loss": 0.15104809403419495, "global_step": 99636, "epoch": 1119, "lr": 7.044869644814257e-05}
{"train_loss": 0.13765816390514374, "global_step": 99637, "epoch": 1119, "lr": 7.044816741041042e-05}
{"train_loss": 0.17395031452178955, "global_step": 99638, "epoch": 1119, "lr": 7.044763836992926e-05}
{"train_loss": 0.1547120362520218, "global_step": 99639, "epoch": 1119, "lr": 7.044710932669918e-05}
{"train_loss": 0.15121987462043762, "global_step": 99640, "epoch": 1119, "lr": 7.044658028072023e-05}
{"train_loss": 0.12669003009796143, "global_step": 99641, "epoch": 1119, "lr": 7.044605123199248e-05}
{"train_loss": 0.1353623867034912, "global_step": 99642, "epoch": 1119, "lr": 7.044552218051601e-05}
{"train_loss": 0.1767086684703827, "global_step": 99643, "epoch": 1119, "lr": 7.044499312629089e-05}
{"train_loss": 0.14953015744686127, "global_step": 99644, "epoch": 1119, "lr": 7.04444640693172e-05}
{"train_loss": 0.12726351618766785, "global_step": 99645, "epoch": 1119, "lr": 7.044393500959498e-05}
{"train_loss": 0.13955193758010864, "global_step": 99646, "epoch": 1119, "lr": 7.044340594712435e-05}
{"train_loss": 0.15024130046367645, "global_step": 99647, "epoch": 1119, "lr": 7.044287688190534e-05}
{"train_loss": 0.10208868980407715, "global_step": 99648, "epoch": 1119, "lr": 7.044234781393802e-05}
{"train_loss": 0.18619109690189362, "global_step": 99649, "epoch": 1119, "lr": 7.04418187432225e-05}
{"train_loss": 0.13452249765396118, "global_step": 99650, "epoch": 1119, "lr": 7.04412896697588e-05}
{"train_loss": 0.13807713985443115, "global_step": 99651, "epoch": 1119, "lr": 7.044076059354704e-05}
{"train_loss": 0.12927144765853882, "global_step": 99652, "epoch": 1119, "lr": 7.044023151458727e-05}
{"train_loss": 0.14507384598255157, "global_step": 99653, "epoch": 1119, "lr": 7.043970243287954e-05}
{"train_loss": 0.22109001874923706, "global_step": 99654, "epoch": 1119, "lr": 7.043917334842396e-05}
{"train_loss": 0.23410329222679138, "global_step": 99655, "epoch": 1119, "lr": 7.043864426122058e-05}
{"train_loss": 0.16422834992408752, "global_step": 99656, "epoch": 1119, "lr": 7.043811517126946e-05}
{"train_loss": 0.11925871670246124, "global_step": 99657, "epoch": 1119, "lr": 7.043758607857069e-05}
{"train_loss": 0.19031605124473572, "global_step": 99658, "epoch": 1119, "lr": 7.043705698312435e-05}
{"train_loss": 0.15558426082134247, "global_step": 99659, "epoch": 1119, "lr": 7.043652788493047e-05}
{"train_loss": 0.12726959586143494, "global_step": 99660, "epoch": 1119, "lr": 7.043599878398916e-05}
{"train_loss": 0.1604727804660797, "global_step": 99661, "epoch": 1119, "lr": 7.043546968030048e-05}
{"train_loss": 0.20693762600421906, "global_step": 99662, "epoch": 1119, "lr": 7.04349405738645e-05}
{"train_loss": 0.1712435483932495, "global_step": 99663, "epoch": 1119, "lr": 7.043441146468131e-05}
{"train_loss": 0.14862094819545746, "global_step": 99664, "epoch": 1119, "lr": 7.043388235275093e-05}
{"train_loss": 0.14572803676128387, "global_step": 99665, "epoch": 1119, "lr": 7.043335323807348e-05}
{"train_loss": 0.21086296439170837, "global_step": 99666, "epoch": 1119, "lr": 7.043282412064902e-05}
{"train_loss": 0.17574773728847504, "global_step": 99667, "epoch": 1119, "lr": 7.043229500047761e-05}
{"train_loss": 0.22885864973068237, "global_step": 99668, "epoch": 1119, "lr": 7.043176587755933e-05}
{"train_loss": 0.1262095868587494, "global_step": 99669, "epoch": 1119, "lr": 7.043123675189424e-05}
{"train_loss": 0.1219092383980751, "global_step": 99670, "epoch": 1119, "lr": 7.043070762348242e-05}
{"train_loss": 0.18001297116279602, "global_step": 99671, "epoch": 1119, "lr": 7.043017849232395e-05}
{"train_loss": 0.19594120979309082, "global_step": 99672, "epoch": 1119, "lr": 7.04296493584189e-05}
{"train_loss": 0.10772427916526794, "global_step": 99673, "epoch": 1119, "lr": 7.042912022176731e-05}
{"train_loss": 0.2186446636915207, "global_step": 99674, "epoch": 1119, "lr": 7.042859108236927e-05}
{"train_loss": 0.13513700664043427, "global_step": 99675, "epoch": 1119, "lr": 7.042806194022488e-05}
{"train_loss": 0.25129759311676025, "global_step": 99676, "epoch": 1119, "lr": 7.042753279533417e-05}
{"train_loss": 0.1153266504406929, "global_step": 99677, "epoch": 1119, "lr": 7.042700364769724e-05}
{"train_loss": 0.1559896171092987, "global_step": 99678, "epoch": 1119, "lr": 7.042647449731413e-05}
{"train_loss": 0.1620271930031562, "global_step": 99679, "epoch": 1119, "lr": 7.042594534418494e-05, "val_loss": 4.7619404792785645}
{"train_loss": 0.1520540416240692, "global_step": 99680, "epoch": 1120, "lr": 7.042541618830972e-05}
{"train_loss": 0.1895940899848938, "global_step": 99681, "epoch": 1120, "lr": 7.042488702968855e-05}
{"train_loss": 0.13133586943149567, "global_step": 99682, "epoch": 1120, "lr": 7.042435786832154e-05}
{"train_loss": 0.159517303109169, "global_step": 99683, "epoch": 1120, "lr": 7.042382870420869e-05}
{"train_loss": 0.1518699675798416, "global_step": 99684, "epoch": 1120, "lr": 7.042329953735011e-05}
{"train_loss": 0.15745212137699127, "global_step": 99685, "epoch": 1120, "lr": 7.042277036774587e-05}
{"train_loss": 0.12775297462940216, "global_step": 99686, "epoch": 1120, "lr": 7.042224119539604e-05}
{"train_loss": 0.19065387547016144, "global_step": 99687, "epoch": 1120, "lr": 7.042171202030068e-05}
{"train_loss": 0.21663136780261993, "global_step": 99688, "epoch": 1120, "lr": 7.042118284245988e-05}
{"train_loss": 0.1467658430337906, "global_step": 99689, "epoch": 1120, "lr": 7.042065366187369e-05}
{"train_loss": 0.16529949009418488, "global_step": 99690, "epoch": 1120, "lr": 7.04201244785422e-05}
{"train_loss": 0.1250670999288559, "global_step": 99691, "epoch": 1120, "lr": 7.041959529246547e-05}
{"train_loss": 0.17201215028762817, "global_step": 99692, "epoch": 1120, "lr": 7.041906610364357e-05}
{"train_loss": 0.15514548122882843, "global_step": 99693, "epoch": 1120, "lr": 7.041853691207658e-05}
{"train_loss": 0.16667650640010834, "global_step": 99694, "epoch": 1120, "lr": 7.041800771776457e-05}
{"train_loss": 0.09462248533964157, "global_step": 99695, "epoch": 1120, "lr": 7.041747852070761e-05}
{"train_loss": 0.16320054233074188, "global_step": 99696, "epoch": 1120, "lr": 7.041694932090576e-05}
{"train_loss": 0.18998722732067108, "global_step": 99697, "epoch": 1120, "lr": 7.04164201183591e-05}
{"train_loss": 0.17571017146110535, "global_step": 99698, "epoch": 1120, "lr": 7.041589091306771e-05}
{"train_loss": 0.117812879383564, "global_step": 99699, "epoch": 1120, "lr": 7.041536170503163e-05}
{"train_loss": 0.13431261479854584, "global_step": 99700, "epoch": 1120, "lr": 7.041483249425098e-05}
{"train_loss": 0.16237208247184753, "global_step": 99701, "epoch": 1120, "lr": 7.04143032807258e-05}
{"train_loss": 0.20518551766872406, "global_step": 99702, "epoch": 1120, "lr": 7.041377406445616e-05}
{"train_loss": 0.19695943593978882, "global_step": 99703, "epoch": 1120, "lr": 7.041324484544214e-05}
{"train_loss": 0.1384253054857254, "global_step": 99704, "epoch": 1120, "lr": 7.041271562368381e-05}
{"train_loss": 0.22537517547607422, "global_step": 99705, "epoch": 1120, "lr": 7.041218639918124e-05}
{"train_loss": 0.149521604180336, "global_step": 99706, "epoch": 1120, "lr": 7.041165717193449e-05}
{"train_loss": 0.17272219061851501, "global_step": 99707, "epoch": 1120, "lr": 7.041112794194365e-05}
{"train_loss": 0.1868053823709488, "global_step": 99708, "epoch": 1120, "lr": 7.041059870920877e-05}
{"train_loss": 0.20949512720108032, "global_step": 99709, "epoch": 1120, "lr": 7.041006947372995e-05}
{"train_loss": 0.20635071396827698, "global_step": 99710, "epoch": 1120, "lr": 7.040954023550723e-05}
{"train_loss": 0.12397628277540207, "global_step": 99711, "epoch": 1120, "lr": 7.040901099454071e-05}
{"train_loss": 0.0964280292391777, "global_step": 99712, "epoch": 1120, "lr": 7.040848175083045e-05}
{"train_loss": 0.1729753613471985, "global_step": 99713, "epoch": 1120, "lr": 7.040795250437651e-05}
{"train_loss": 0.2105523943901062, "global_step": 99714, "epoch": 1120, "lr": 7.040742325517896e-05}
{"train_loss": 0.1308824121952057, "global_step": 99715, "epoch": 1120, "lr": 7.040689400323788e-05}
{"train_loss": 0.14791466295719147, "global_step": 99716, "epoch": 1120, "lr": 7.040636474855336e-05}
{"train_loss": 0.19219468533992767, "global_step": 99717, "epoch": 1120, "lr": 7.040583549112544e-05}
{"train_loss": 0.20444561541080475, "global_step": 99718, "epoch": 1120, "lr": 7.040530623095421e-05}
{"train_loss": 0.18369153141975403, "global_step": 99719, "epoch": 1120, "lr": 7.040477696803974e-05}
{"train_loss": 0.19210375845432281, "global_step": 99720, "epoch": 1120, "lr": 7.04042477023821e-05}
{"train_loss": 0.1124788299202919, "global_step": 99721, "epoch": 1120, "lr": 7.040371843398134e-05}
{"train_loss": 0.10411569476127625, "global_step": 99722, "epoch": 1120, "lr": 7.040318916283756e-05}
{"train_loss": 0.20530684292316437, "global_step": 99723, "epoch": 1120, "lr": 7.040265988895082e-05}
{"train_loss": 0.1414126455783844, "global_step": 99724, "epoch": 1120, "lr": 7.040213061232118e-05}
{"train_loss": 0.10803784430027008, "global_step": 99725, "epoch": 1120, "lr": 7.040160133294873e-05}
{"train_loss": 0.15109436213970184, "global_step": 99726, "epoch": 1120, "lr": 7.040107205083354e-05}
{"train_loss": 0.12973399460315704, "global_step": 99727, "epoch": 1120, "lr": 7.040054276597567e-05}
{"train_loss": 0.296330064535141, "global_step": 99728, "epoch": 1120, "lr": 7.04000134783752e-05}
{"train_loss": 0.15536609292030334, "global_step": 99729, "epoch": 1120, "lr": 7.039948418803219e-05}
{"train_loss": 0.11190374940633774, "global_step": 99730, "epoch": 1120, "lr": 7.039895489494671e-05}
{"train_loss": 0.14711417257785797, "global_step": 99731, "epoch": 1120, "lr": 7.039842559911884e-05}
{"train_loss": 0.13083747029304504, "global_step": 99732, "epoch": 1120, "lr": 7.039789630054867e-05}
{"train_loss": 0.09873891621828079, "global_step": 99733, "epoch": 1120, "lr": 7.039736699923624e-05}
{"train_loss": 0.14505061507225037, "global_step": 99734, "epoch": 1120, "lr": 7.039683769518162e-05}
{"train_loss": 0.13718301057815552, "global_step": 99735, "epoch": 1120, "lr": 7.03963083883849e-05}
{"train_loss": 0.11346390098333359, "global_step": 99736, "epoch": 1120, "lr": 7.039577907884616e-05}
{"train_loss": 0.12911686301231384, "global_step": 99737, "epoch": 1120, "lr": 7.039524976656544e-05}
{"train_loss": 0.17479930818080902, "global_step": 99738, "epoch": 1120, "lr": 7.039472045154283e-05}
{"train_loss": 0.1862223893404007, "global_step": 99739, "epoch": 1120, "lr": 7.03941911337784e-05}
{"train_loss": 0.25327199697494507, "global_step": 99740, "epoch": 1120, "lr": 7.039366181327224e-05}
{"train_loss": 0.16476485133171082, "global_step": 99741, "epoch": 1120, "lr": 7.039313249002438e-05}
{"train_loss": 0.1424509733915329, "global_step": 99742, "epoch": 1120, "lr": 7.039260316403493e-05}
{"train_loss": 0.09427197277545929, "global_step": 99743, "epoch": 1120, "lr": 7.039207383530392e-05}
{"train_loss": 0.17932909727096558, "global_step": 99744, "epoch": 1120, "lr": 7.039154450383145e-05}
{"train_loss": 0.17969895899295807, "global_step": 99745, "epoch": 1120, "lr": 7.03910151696176e-05}
{"train_loss": 0.15862669050693512, "global_step": 99746, "epoch": 1120, "lr": 7.03904858326624e-05}
{"train_loss": 0.10559207201004028, "global_step": 99747, "epoch": 1120, "lr": 7.038995649296596e-05}
{"train_loss": 0.14168834686279297, "global_step": 99748, "epoch": 1120, "lr": 7.038942715052835e-05}
{"train_loss": 0.16764701902866364, "global_step": 99749, "epoch": 1120, "lr": 7.038889780534961e-05}
{"train_loss": 0.17277570068836212, "global_step": 99750, "epoch": 1120, "lr": 7.038836845742984e-05}
{"train_loss": 0.13346657156944275, "global_step": 99751, "epoch": 1120, "lr": 7.038783910676911e-05}
{"train_loss": 0.24924801290035248, "global_step": 99752, "epoch": 1120, "lr": 7.038730975336748e-05}
{"train_loss": 0.1244819387793541, "global_step": 99753, "epoch": 1120, "lr": 7.0386780397225e-05}
{"train_loss": 0.08531653881072998, "global_step": 99754, "epoch": 1120, "lr": 7.03862510383418e-05}
{"train_loss": 0.15572582185268402, "global_step": 99755, "epoch": 1120, "lr": 7.03857216767179e-05}
{"train_loss": 0.14203296601772308, "global_step": 99756, "epoch": 1120, "lr": 7.03851923123534e-05}
{"train_loss": 0.11501792818307877, "global_step": 99757, "epoch": 1120, "lr": 7.038466294524835e-05}
{"train_loss": 0.19252033531665802, "global_step": 99758, "epoch": 1120, "lr": 7.038413357540284e-05}
{"train_loss": 0.15722985565662384, "global_step": 99759, "epoch": 1120, "lr": 7.038360420281692e-05}
{"train_loss": 0.13148808479309082, "global_step": 99760, "epoch": 1120, "lr": 7.038307482749069e-05}
{"train_loss": 0.1531289517879486, "global_step": 99761, "epoch": 1120, "lr": 7.038254544942418e-05}
{"train_loss": 0.22228345274925232, "global_step": 99762, "epoch": 1120, "lr": 7.038201606861751e-05}
{"train_loss": 0.10734494775533676, "global_step": 99763, "epoch": 1120, "lr": 7.03814866850707e-05}
{"train_loss": 0.14609940350055695, "global_step": 99764, "epoch": 1120, "lr": 7.038095729878386e-05}
{"train_loss": 0.22583551704883575, "global_step": 99765, "epoch": 1120, "lr": 7.038042790975705e-05}
{"train_loss": 0.09816864877939224, "global_step": 99766, "epoch": 1120, "lr": 7.037989851799036e-05}
{"train_loss": 0.09216257184743881, "global_step": 99767, "epoch": 1120, "lr": 7.037936912348381e-05}
{"train_loss": 0.15911443305484366, "global_step": 99768, "epoch": 1120, "lr": 7.037883972623752e-05, "val_loss": 4.874972343444824, "train_action_mse_error": 6.336394309997559}
{"train_loss": 0.14526192843914032, "global_step": 99769, "epoch": 1121, "lr": 7.037831032625154e-05}
{"train_loss": 0.1741972267627716, "global_step": 99770, "epoch": 1121, "lr": 7.037778092352591e-05}
{"train_loss": 0.1084461584687233, "global_step": 99771, "epoch": 1121, "lr": 7.037725151806077e-05}
{"train_loss": 0.14358699321746826, "global_step": 99772, "epoch": 1121, "lr": 7.037672210985616e-05}
{"train_loss": 0.24659809470176697, "global_step": 99773, "epoch": 1121, "lr": 7.037619269891213e-05}
{"train_loss": 0.18025776743888855, "global_step": 99774, "epoch": 1121, "lr": 7.037566328522879e-05}
{"train_loss": 0.16737015545368195, "global_step": 99775, "epoch": 1121, "lr": 7.037513386880618e-05}
{"train_loss": 0.17890368402004242, "global_step": 99776, "epoch": 1121, "lr": 7.037460444964439e-05}
{"train_loss": 0.18486765027046204, "global_step": 99777, "epoch": 1121, "lr": 7.037407502774347e-05}
{"train_loss": 0.1255260407924652, "global_step": 99778, "epoch": 1121, "lr": 7.037354560310352e-05}
{"train_loss": 0.13049942255020142, "global_step": 99779, "epoch": 1121, "lr": 7.037301617572458e-05}
{"train_loss": 0.15155476331710815, "global_step": 99780, "epoch": 1121, "lr": 7.037248674560675e-05}
{"train_loss": 0.1520329713821411, "global_step": 99781, "epoch": 1121, "lr": 7.037195731275008e-05}
{"train_loss": 0.16004900634288788, "global_step": 99782, "epoch": 1121, "lr": 7.037142787715464e-05}
{"train_loss": 0.13810773193836212, "global_step": 99783, "epoch": 1121, "lr": 7.037089843882052e-05}
{"train_loss": 0.0852617621421814, "global_step": 99784, "epoch": 1121, "lr": 7.037036899774777e-05}
{"train_loss": 0.2111584097146988, "global_step": 99785, "epoch": 1121, "lr": 7.036983955393648e-05}
{"train_loss": 0.19210129976272583, "global_step": 99786, "epoch": 1121, "lr": 7.036931010738673e-05}
{"train_loss": 0.1324792206287384, "global_step": 99787, "epoch": 1121, "lr": 7.036878065809854e-05}
{"train_loss": 0.16432054340839386, "global_step": 99788, "epoch": 1121, "lr": 7.036825120607202e-05}
{"train_loss": 0.21674415469169617, "global_step": 99789, "epoch": 1121, "lr": 7.036772175130726e-05}
{"train_loss": 0.13454119861125946, "global_step": 99790, "epoch": 1121, "lr": 7.036719229380429e-05}
{"train_loss": 0.1626335233449936, "global_step": 99791, "epoch": 1121, "lr": 7.036666283356321e-05}
{"train_loss": 0.12301608175039291, "global_step": 99792, "epoch": 1121, "lr": 7.036613337058409e-05}
{"train_loss": 0.18688130378723145, "global_step": 99793, "epoch": 1121, "lr": 7.036560390486697e-05}
{"train_loss": 0.17162442207336426, "global_step": 99794, "epoch": 1121, "lr": 7.036507443641194e-05}
{"train_loss": 0.15458914637565613, "global_step": 99795, "epoch": 1121, "lr": 7.036454496521909e-05}
{"train_loss": 0.09878090023994446, "global_step": 99796, "epoch": 1121, "lr": 7.036401549128849e-05}
{"train_loss": 0.1570887714624405, "global_step": 99797, "epoch": 1121, "lr": 7.036348601462017e-05}
{"train_loss": 0.1541813462972641, "global_step": 99798, "epoch": 1121, "lr": 7.036295653521424e-05}
{"train_loss": 0.2785537838935852, "global_step": 99799, "epoch": 1121, "lr": 7.036242705307075e-05}
{"train_loss": 0.12516997754573822, "global_step": 99800, "epoch": 1121, "lr": 7.036189756818979e-05}
{"train_loss": 0.11482201516628265, "global_step": 99801, "epoch": 1121, "lr": 7.03613680805714e-05}
{"train_loss": 0.13632717728614807, "global_step": 99802, "epoch": 1121, "lr": 7.036083859021571e-05}
{"train_loss": 0.1473173350095749, "global_step": 99803, "epoch": 1121, "lr": 7.036030909712272e-05}
{"train_loss": 0.13066402077674866, "global_step": 99804, "epoch": 1121, "lr": 7.035977960129255e-05}
{"train_loss": 0.1985015571117401, "global_step": 99805, "epoch": 1121, "lr": 7.035925010272525e-05}
{"train_loss": 0.13062861561775208, "global_step": 99806, "epoch": 1121, "lr": 7.035872060142089e-05}
{"train_loss": 0.1801796853542328, "global_step": 99807, "epoch": 1121, "lr": 7.035819109737956e-05}
{"train_loss": 0.20250654220581055, "global_step": 99808, "epoch": 1121, "lr": 7.03576615906013e-05}
{"train_loss": 0.18911969661712646, "global_step": 99809, "epoch": 1121, "lr": 7.035713208108623e-05}
{"train_loss": 0.20007944107055664, "global_step": 99810, "epoch": 1121, "lr": 7.035660256883438e-05}
{"train_loss": 0.13006675243377686, "global_step": 99811, "epoch": 1121, "lr": 7.035607305384581e-05}
{"train_loss": 0.15883281826972961, "global_step": 99812, "epoch": 1121, "lr": 7.035554353612065e-05}
{"train_loss": 0.2044452726840973, "global_step": 99813, "epoch": 1121, "lr": 7.035501401565891e-05}
{"train_loss": 0.15686650574207306, "global_step": 99814, "epoch": 1121, "lr": 7.03544844924607e-05}
{"train_loss": 0.06797640025615692, "global_step": 99815, "epoch": 1121, "lr": 7.035395496652609e-05}
{"train_loss": 0.15847909450531006, "global_step": 99816, "epoch": 1121, "lr": 7.035342543785512e-05}
{"train_loss": 0.13100311160087585, "global_step": 99817, "epoch": 1121, "lr": 7.035289590644789e-05}
{"train_loss": 0.1409490406513214, "global_step": 99818, "epoch": 1121, "lr": 7.035236637230445e-05}
{"train_loss": 0.18151281774044037, "global_step": 99819, "epoch": 1121, "lr": 7.035183683542488e-05}
{"train_loss": 0.19825339317321777, "global_step": 99820, "epoch": 1121, "lr": 7.035130729580926e-05}
{"train_loss": 0.16893279552459717, "global_step": 99821, "epoch": 1121, "lr": 7.035077775345767e-05}
{"train_loss": 0.19877666234970093, "global_step": 99822, "epoch": 1121, "lr": 7.035024820837015e-05}
{"train_loss": 0.17271079123020172, "global_step": 99823, "epoch": 1121, "lr": 7.034971866054678e-05}
{"train_loss": 0.14935708045959473, "global_step": 99824, "epoch": 1121, "lr": 7.034918910998764e-05}
{"train_loss": 0.25557440519332886, "global_step": 99825, "epoch": 1121, "lr": 7.03486595566928e-05}
{"train_loss": 0.147237166762352, "global_step": 99826, "epoch": 1121, "lr": 7.034813000066233e-05}
{"train_loss": 0.1663850098848343, "global_step": 99827, "epoch": 1121, "lr": 7.034760044189631e-05}
{"train_loss": 0.2408723533153534, "global_step": 99828, "epoch": 1121, "lr": 7.034707088039481e-05}
{"train_loss": 0.21931743621826172, "global_step": 99829, "epoch": 1121, "lr": 7.034654131615788e-05}
{"train_loss": 0.21668648719787598, "global_step": 99830, "epoch": 1121, "lr": 7.034601174918562e-05}
{"train_loss": 0.16025665402412415, "global_step": 99831, "epoch": 1121, "lr": 7.034548217947807e-05}
{"train_loss": 0.14352205395698547, "global_step": 99832, "epoch": 1121, "lr": 7.034495260703533e-05}
{"train_loss": 0.14504358172416687, "global_step": 99833, "epoch": 1121, "lr": 7.034442303185746e-05}
{"train_loss": 0.12182646989822388, "global_step": 99834, "epoch": 1121, "lr": 7.034389345394452e-05}
{"train_loss": 0.1053030788898468, "global_step": 99835, "epoch": 1121, "lr": 7.034336387329659e-05}
{"train_loss": 0.1850183755159378, "global_step": 99836, "epoch": 1121, "lr": 7.034283428991376e-05}
{"train_loss": 0.237823948264122, "global_step": 99837, "epoch": 1121, "lr": 7.034230470379607e-05}
{"train_loss": 0.19231107831001282, "global_step": 99838, "epoch": 1121, "lr": 7.034177511494362e-05}
{"train_loss": 0.3104332387447357, "global_step": 99839, "epoch": 1121, "lr": 7.034124552335646e-05}
{"train_loss": 0.3134654462337494, "global_step": 99840, "epoch": 1121, "lr": 7.034071592903465e-05}
{"train_loss": 0.23996102809906006, "global_step": 99841, "epoch": 1121, "lr": 7.03401863319783e-05}
{"train_loss": 0.21597695350646973, "global_step": 99842, "epoch": 1121, "lr": 7.033965673218744e-05}
{"train_loss": 0.10804641246795654, "global_step": 99843, "epoch": 1121, "lr": 7.033912712966216e-05}
{"train_loss": 0.18634767830371857, "global_step": 99844, "epoch": 1121, "lr": 7.033859752440255e-05}
{"train_loss": 0.19253668189048767, "global_step": 99845, "epoch": 1121, "lr": 7.033806791640865e-05}
{"train_loss": 0.23151017725467682, "global_step": 99846, "epoch": 1121, "lr": 7.033753830568056e-05}
{"train_loss": 0.2128099948167801, "global_step": 99847, "epoch": 1121, "lr": 7.033700869221833e-05}
{"train_loss": 0.25043773651123047, "global_step": 99848, "epoch": 1121, "lr": 7.033647907602203e-05}
{"train_loss": 0.13048362731933594, "global_step": 99849, "epoch": 1121, "lr": 7.033594945709175e-05}
{"train_loss": 0.16516247391700745, "global_step": 99850, "epoch": 1121, "lr": 7.033541983542754e-05}
{"train_loss": 0.1679975390434265, "global_step": 99851, "epoch": 1121, "lr": 7.033489021102948e-05}
{"train_loss": 0.2772652804851532, "global_step": 99852, "epoch": 1121, "lr": 7.033436058389766e-05}
{"train_loss": 0.19083340466022491, "global_step": 99853, "epoch": 1121, "lr": 7.033383095403211e-05}
{"train_loss": 0.175207257270813, "global_step": 99854, "epoch": 1121, "lr": 7.033330132143293e-05}
{"train_loss": 0.24057632684707642, "global_step": 99855, "epoch": 1121, "lr": 7.033277168610018e-05}
{"train_loss": 0.25093644857406616, "global_step": 99856, "epoch": 1121, "lr": 7.033224204803396e-05}
{"train_loss": 0.17616568522506884, "global_step": 99857, "epoch": 1121, "lr": 7.033171240723431e-05, "val_loss": 4.638954162597656}
{"train_loss": 0.19471749663352966, "global_step": 99858, "epoch": 1122, "lr": 7.03311827637013e-05}
{"train_loss": 0.20219364762306213, "global_step": 99859, "epoch": 1122, "lr": 7.033065311743501e-05}
{"train_loss": 0.15489140152931213, "global_step": 99860, "epoch": 1122, "lr": 7.033012346843552e-05}
{"train_loss": 0.08890481293201447, "global_step": 99861, "epoch": 1122, "lr": 7.032959381670288e-05}
{"train_loss": 0.18949076533317566, "global_step": 99862, "epoch": 1122, "lr": 7.032906416223717e-05}
{"train_loss": 0.21632707118988037, "global_step": 99863, "epoch": 1122, "lr": 7.032853450503848e-05}
{"train_loss": 0.14611534774303436, "global_step": 99864, "epoch": 1122, "lr": 7.032800484510685e-05}
{"train_loss": 0.14072155952453613, "global_step": 99865, "epoch": 1122, "lr": 7.032747518244238e-05}
{"train_loss": 0.2140951156616211, "global_step": 99866, "epoch": 1122, "lr": 7.032694551704513e-05}
{"train_loss": 0.23611678183078766, "global_step": 99867, "epoch": 1122, "lr": 7.032641584891518e-05}
{"train_loss": 0.14619767665863037, "global_step": 99868, "epoch": 1122, "lr": 7.032588617805257e-05}
{"train_loss": 0.15429924428462982, "global_step": 99869, "epoch": 1122, "lr": 7.03253565044574e-05}
{"train_loss": 0.20990318059921265, "global_step": 99870, "epoch": 1122, "lr": 7.032482682812973e-05}
{"train_loss": 0.14198434352874756, "global_step": 99871, "epoch": 1122, "lr": 7.032429714906965e-05}
{"train_loss": 0.20809757709503174, "global_step": 99872, "epoch": 1122, "lr": 7.03237674672772e-05}
{"train_loss": 0.2038053721189499, "global_step": 99873, "epoch": 1122, "lr": 7.032323778275247e-05}
{"train_loss": 0.2595991790294647, "global_step": 99874, "epoch": 1122, "lr": 7.032270809549554e-05}
{"train_loss": 0.16678708791732788, "global_step": 99875, "epoch": 1122, "lr": 7.032217840550647e-05}
{"train_loss": 0.1455032229423523, "global_step": 99876, "epoch": 1122, "lr": 7.032164871278532e-05}
{"train_loss": 0.177289679646492, "global_step": 99877, "epoch": 1122, "lr": 7.032111901733217e-05}
{"train_loss": 0.19229687750339508, "global_step": 99878, "epoch": 1122, "lr": 7.03205893191471e-05}
{"train_loss": 0.19681136310100555, "global_step": 99879, "epoch": 1122, "lr": 7.032005961823016e-05}
{"train_loss": 0.1712038218975067, "global_step": 99880, "epoch": 1122, "lr": 7.031952991458144e-05}
{"train_loss": 0.15570877492427826, "global_step": 99881, "epoch": 1122, "lr": 7.031900020820101e-05}
{"train_loss": 0.13017870485782623, "global_step": 99882, "epoch": 1122, "lr": 7.031847049908895e-05}
{"train_loss": 0.1989426612854004, "global_step": 99883, "epoch": 1122, "lr": 7.031794078724532e-05}
{"train_loss": 0.13232246041297913, "global_step": 99884, "epoch": 1122, "lr": 7.031741107267016e-05}
{"train_loss": 0.13974377512931824, "global_step": 99885, "epoch": 1122, "lr": 7.03168813553636e-05}
{"train_loss": 0.17861053347587585, "global_step": 99886, "epoch": 1122, "lr": 7.031635163532568e-05}
{"train_loss": 0.16042762994766235, "global_step": 99887, "epoch": 1122, "lr": 7.031582191255648e-05}
{"train_loss": 0.1076413244009018, "global_step": 99888, "epoch": 1122, "lr": 7.031529218705605e-05}
{"train_loss": 0.1954745352268219, "global_step": 99889, "epoch": 1122, "lr": 7.031476245882448e-05}
{"train_loss": 0.14183016121387482, "global_step": 99890, "epoch": 1122, "lr": 7.031423272786183e-05}
{"train_loss": 0.2103479951620102, "global_step": 99891, "epoch": 1122, "lr": 7.03137029941682e-05}
{"train_loss": 0.21910732984542847, "global_step": 99892, "epoch": 1122, "lr": 7.031317325774363e-05}
{"train_loss": 0.2551302909851074, "global_step": 99893, "epoch": 1122, "lr": 7.031264351858819e-05}
{"train_loss": 0.18358542025089264, "global_step": 99894, "epoch": 1122, "lr": 7.031211377670198e-05}
{"train_loss": 0.2118009775876999, "global_step": 99895, "epoch": 1122, "lr": 7.031158403208506e-05}
{"train_loss": 0.16920259594917297, "global_step": 99896, "epoch": 1122, "lr": 7.031105428473747e-05}
{"train_loss": 0.21748150885105133, "global_step": 99897, "epoch": 1122, "lr": 7.031052453465932e-05}
{"train_loss": 0.15656690299510956, "global_step": 99898, "epoch": 1122, "lr": 7.030999478185066e-05}
{"train_loss": 0.11871927976608276, "global_step": 99899, "epoch": 1122, "lr": 7.030946502631159e-05}
{"train_loss": 0.19262264668941498, "global_step": 99900, "epoch": 1122, "lr": 7.030893526804214e-05}
{"train_loss": 0.14294831454753876, "global_step": 99901, "epoch": 1122, "lr": 7.03084055070424e-05}
{"train_loss": 0.15699145197868347, "global_step": 99902, "epoch": 1122, "lr": 7.030787574331245e-05}
{"train_loss": 0.1927991807460785, "global_step": 99903, "epoch": 1122, "lr": 7.030734597685235e-05}
{"train_loss": 0.19558119773864746, "global_step": 99904, "epoch": 1122, "lr": 7.030681620766217e-05}
{"train_loss": 0.10022072494029999, "global_step": 99905, "epoch": 1122, "lr": 7.0306286435742e-05}
{"train_loss": 0.2166721224784851, "global_step": 99906, "epoch": 1122, "lr": 7.03057566610919e-05}
{"train_loss": 0.2041883021593094, "global_step": 99907, "epoch": 1122, "lr": 7.030522688371193e-05}
{"train_loss": 0.12504884600639343, "global_step": 99908, "epoch": 1122, "lr": 7.030469710360216e-05}
{"train_loss": 0.17155112326145172, "global_step": 99909, "epoch": 1122, "lr": 7.030416732076268e-05}
{"train_loss": 0.22152434289455414, "global_step": 99910, "epoch": 1122, "lr": 7.030363753519355e-05}
{"train_loss": 0.16710121929645538, "global_step": 99911, "epoch": 1122, "lr": 7.030310774689484e-05}
{"train_loss": 0.14260601997375488, "global_step": 99912, "epoch": 1122, "lr": 7.030257795586663e-05}
{"train_loss": 0.1847224235534668, "global_step": 99913, "epoch": 1122, "lr": 7.030204816210899e-05}
{"train_loss": 0.11092300713062286, "global_step": 99914, "epoch": 1122, "lr": 7.0301518365622e-05}
{"train_loss": 0.20273099839687347, "global_step": 99915, "epoch": 1122, "lr": 7.030098856640568e-05}
{"train_loss": 0.13766756653785706, "global_step": 99916, "epoch": 1122, "lr": 7.030045876446016e-05}
{"train_loss": 0.16881510615348816, "global_step": 99917, "epoch": 1122, "lr": 7.02999289597855e-05}
{"train_loss": 0.16288012266159058, "global_step": 99918, "epoch": 1122, "lr": 7.029939915238175e-05}
{"train_loss": 0.1954103261232376, "global_step": 99919, "epoch": 1122, "lr": 7.0298869342249e-05}
{"train_loss": 0.09912143647670746, "global_step": 99920, "epoch": 1122, "lr": 7.029833952938731e-05}
{"train_loss": 0.1395920217037201, "global_step": 99921, "epoch": 1122, "lr": 7.029780971379675e-05}
{"train_loss": 0.1669309288263321, "global_step": 99922, "epoch": 1122, "lr": 7.029727989547741e-05}
{"train_loss": 0.19116365909576416, "global_step": 99923, "epoch": 1122, "lr": 7.029675007442933e-05}
{"train_loss": 0.09201423078775406, "global_step": 99924, "epoch": 1122, "lr": 7.029622025065262e-05}
{"train_loss": 0.17961549758911133, "global_step": 99925, "epoch": 1122, "lr": 7.029569042414731e-05}
{"train_loss": 0.1710669994354248, "global_step": 99926, "epoch": 1122, "lr": 7.029516059491351e-05}
{"train_loss": 0.11156564205884933, "global_step": 99927, "epoch": 1122, "lr": 7.029463076295126e-05}
{"train_loss": 0.20553237199783325, "global_step": 99928, "epoch": 1122, "lr": 7.029410092826065e-05}
{"train_loss": 0.11408861726522446, "global_step": 99929, "epoch": 1122, "lr": 7.029357109084176e-05}
{"train_loss": 0.16158437728881836, "global_step": 99930, "epoch": 1122, "lr": 7.029304125069462e-05}
{"train_loss": 0.21359777450561523, "global_step": 99931, "epoch": 1122, "lr": 7.029251140781934e-05}
{"train_loss": 0.1931591033935547, "global_step": 99932, "epoch": 1122, "lr": 7.0291981562216e-05}
{"train_loss": 0.1101517304778099, "global_step": 99933, "epoch": 1122, "lr": 7.029145171388462e-05}
{"train_loss": 0.15633204579353333, "global_step": 99934, "epoch": 1122, "lr": 7.029092186282531e-05}
{"train_loss": 0.18330328166484833, "global_step": 99935, "epoch": 1122, "lr": 7.029039200903812e-05}
{"train_loss": 0.15159223973751068, "global_step": 99936, "epoch": 1122, "lr": 7.028986215252315e-05}
{"train_loss": 0.15836191177368164, "global_step": 99937, "epoch": 1122, "lr": 7.028933229328046e-05}
{"train_loss": 0.16253265738487244, "global_step": 99938, "epoch": 1122, "lr": 7.028880243131012e-05}
{"train_loss": 0.16227537393569946, "global_step": 99939, "epoch": 1122, "lr": 7.028827256661219e-05}
{"train_loss": 0.15896914899349213, "global_step": 99940, "epoch": 1122, "lr": 7.028774269918674e-05}
{"train_loss": 0.13608969748020172, "global_step": 99941, "epoch": 1122, "lr": 7.028721282903386e-05}
{"train_loss": 0.20486241579055786, "global_step": 99942, "epoch": 1122, "lr": 7.028668295615361e-05}
{"train_loss": 0.19043248891830444, "global_step": 99943, "epoch": 1122, "lr": 7.028615308054605e-05}
{"train_loss": 0.1459653675556183, "global_step": 99944, "epoch": 1122, "lr": 7.028562320221128e-05}
{"train_loss": 0.24138608574867249, "global_step": 99945, "epoch": 1122, "lr": 7.028509332114935e-05}
{"train_loss": 0.1710620371813185, "global_step": 99946, "epoch": 1122, "lr": 7.028456343736035e-05, "val_loss": 4.695284366607666}
{"train_loss": 0.16739420592784882, "global_step": 99947, "epoch": 1123, "lr": 7.028403355084432e-05}
{"train_loss": 0.20097820460796356, "global_step": 99948, "epoch": 1123, "lr": 7.028350366160135e-05}
{"train_loss": 0.14672888815402985, "global_step": 99949, "epoch": 1123, "lr": 7.028297376963151e-05}
{"train_loss": 0.2000269740819931, "global_step": 99950, "epoch": 1123, "lr": 7.028244387493488e-05}
{"train_loss": 0.15404005348682404, "global_step": 99951, "epoch": 1123, "lr": 7.028191397751151e-05}
{"train_loss": 0.20832385122776031, "global_step": 99952, "epoch": 1123, "lr": 7.02813840773615e-05}
{"train_loss": 0.08395016938447952, "global_step": 99953, "epoch": 1123, "lr": 7.02808541744849e-05}
{"train_loss": 0.1309703290462494, "global_step": 99954, "epoch": 1123, "lr": 7.028032426888178e-05}
{"train_loss": 0.20083867013454437, "global_step": 99955, "epoch": 1123, "lr": 7.027979436055223e-05}
{"train_loss": 0.1216246709227562, "global_step": 99956, "epoch": 1123, "lr": 7.02792644494963e-05}
{"train_loss": 0.21773844957351685, "global_step": 99957, "epoch": 1123, "lr": 7.027873453571407e-05}
{"train_loss": 0.20341575145721436, "global_step": 99958, "epoch": 1123, "lr": 7.02782046192056e-05}
{"train_loss": 0.10992953181266785, "global_step": 99959, "epoch": 1123, "lr": 7.0277674699971e-05}
{"train_loss": 0.14774812757968903, "global_step": 99960, "epoch": 1123, "lr": 7.027714477801028e-05}
{"train_loss": 0.1694014072418213, "global_step": 99961, "epoch": 1123, "lr": 7.027661485332357e-05}
{"train_loss": 0.16416411101818085, "global_step": 99962, "epoch": 1123, "lr": 7.027608492591091e-05}
{"train_loss": 0.16626191139221191, "global_step": 99963, "epoch": 1123, "lr": 7.027555499577238e-05}
{"train_loss": 0.2024957239627838, "global_step": 99964, "epoch": 1123, "lr": 7.027502506290803e-05}
{"train_loss": 0.1528298705816269, "global_step": 99965, "epoch": 1123, "lr": 7.027449512731797e-05}
{"train_loss": 0.19022446870803833, "global_step": 99966, "epoch": 1123, "lr": 7.027396518900226e-05}
{"train_loss": 0.23283874988555908, "global_step": 99967, "epoch": 1123, "lr": 7.027343524796093e-05}
{"train_loss": 0.19138383865356445, "global_step": 99968, "epoch": 1123, "lr": 7.027290530419411e-05}
{"train_loss": 0.21641917526721954, "global_step": 99969, "epoch": 1123, "lr": 7.027237535770184e-05}
{"train_loss": 0.19483381509780884, "global_step": 99970, "epoch": 1123, "lr": 7.027184540848419e-05}
{"train_loss": 0.15238186717033386, "global_step": 99971, "epoch": 1123, "lr": 7.027131545654124e-05}
{"train_loss": 0.27115488052368164, "global_step": 99972, "epoch": 1123, "lr": 7.027078550187308e-05}
{"train_loss": 0.13125619292259216, "global_step": 99973, "epoch": 1123, "lr": 7.027025554447972e-05}
{"train_loss": 0.2560083568096161, "global_step": 99974, "epoch": 1123, "lr": 7.02697255843613e-05}
{"train_loss": 0.14888089895248413, "global_step": 99975, "epoch": 1123, "lr": 7.026919562151786e-05}
{"train_loss": 0.1501261591911316, "global_step": 99976, "epoch": 1123, "lr": 7.026866565594947e-05}
{"train_loss": 0.17162859439849854, "global_step": 99977, "epoch": 1123, "lr": 7.026813568765619e-05}
{"train_loss": 0.18660125136375427, "global_step": 99978, "epoch": 1123, "lr": 7.026760571663811e-05}
{"train_loss": 0.19894862174987793, "global_step": 99979, "epoch": 1123, "lr": 7.026707574289531e-05}
{"train_loss": 0.20943056046962738, "global_step": 99980, "epoch": 1123, "lr": 7.026654576642785e-05}
{"train_loss": 0.14566968381404877, "global_step": 99981, "epoch": 1123, "lr": 7.02660157872358e-05}
{"train_loss": 0.1639937162399292, "global_step": 99982, "epoch": 1123, "lr": 7.026548580531922e-05}
{"train_loss": 0.15112975239753723, "global_step": 99983, "epoch": 1123, "lr": 7.02649558206782e-05}
{"train_loss": 0.1060137078166008, "global_step": 99984, "epoch": 1123, "lr": 7.02644258333128e-05}
{"train_loss": 0.17261619865894318, "global_step": 99985, "epoch": 1123, "lr": 7.02638958432231e-05}
{"train_loss": 0.21386770904064178, "global_step": 99986, "epoch": 1123, "lr": 7.026336585040916e-05}
{"train_loss": 0.19815757870674133, "global_step": 99987, "epoch": 1123, "lr": 7.026283585487106e-05}
{"train_loss": 0.16903279721736908, "global_step": 99988, "epoch": 1123, "lr": 7.026230585660886e-05}
{"train_loss": 0.22748838365077972, "global_step": 99989, "epoch": 1123, "lr": 7.026177585562266e-05}
{"train_loss": 0.22501103579998016, "global_step": 99990, "epoch": 1123, "lr": 7.026124585191249e-05}
{"train_loss": 0.18046829104423523, "global_step": 99991, "epoch": 1123, "lr": 7.026071584547847e-05}
{"train_loss": 0.1837293654680252, "global_step": 99992, "epoch": 1123, "lr": 7.026018583632061e-05}
{"train_loss": 0.15766462683677673, "global_step": 99993, "epoch": 1123, "lr": 7.025965582443905e-05}
{"train_loss": 0.13867385685443878, "global_step": 99994, "epoch": 1123, "lr": 7.02591258098338e-05}
{"train_loss": 0.16464129090309143, "global_step": 99995, "epoch": 1123, "lr": 7.025859579250496e-05}
{"train_loss": 0.19412964582443237, "global_step": 99996, "epoch": 1123, "lr": 7.02580657724526e-05}
{"train_loss": 0.1371656209230423, "global_step": 99997, "epoch": 1123, "lr": 7.02575357496768e-05}
{"train_loss": 0.3244529664516449, "global_step": 99998, "epoch": 1123, "lr": 7.025700572417762e-05}
{"train_loss": 0.13989700376987457, "global_step": 99999, "epoch": 1123, "lr": 7.025647569595512e-05}
{"train_loss": 0.2453163117170334, "global_step": 100000, "epoch": 1123, "lr": 7.025594566500939e-05}
{"train_loss": 0.21904030442237854, "global_step": 100001, "epoch": 1123, "lr": 7.02554156313405e-05}
{"train_loss": 0.11327842622995377, "global_step": 100002, "epoch": 1123, "lr": 7.025488559494851e-05}
{"train_loss": 0.18774890899658203, "global_step": 100003, "epoch": 1123, "lr": 7.025435555583351e-05}
{"train_loss": 0.20387567579746246, "global_step": 100004, "epoch": 1123, "lr": 7.025382551399555e-05}
{"train_loss": 0.1583544760942459, "global_step": 100005, "epoch": 1123, "lr": 7.02532954694347e-05}
{"train_loss": 0.14916501939296722, "global_step": 100006, "epoch": 1123, "lr": 7.025276542215104e-05}
{"train_loss": 0.16266167163848877, "global_step": 100007, "epoch": 1123, "lr": 7.025223537214466e-05}
{"train_loss": 0.20394612848758698, "global_step": 100008, "epoch": 1123, "lr": 7.02517053194156e-05}
{"train_loss": 0.1666756421327591, "global_step": 100009, "epoch": 1123, "lr": 7.025117526396396e-05}
{"train_loss": 0.19162259995937347, "global_step": 100010, "epoch": 1123, "lr": 7.025064520578978e-05}
{"train_loss": 0.271461546421051, "global_step": 100011, "epoch": 1123, "lr": 7.025011514489316e-05}
{"train_loss": 0.19122298061847687, "global_step": 100012, "epoch": 1123, "lr": 7.024958508127414e-05}
{"train_loss": 0.14203587174415588, "global_step": 100013, "epoch": 1123, "lr": 7.024905501493285e-05}
{"train_loss": 0.17242024838924408, "global_step": 100014, "epoch": 1123, "lr": 7.02485249458693e-05}
{"train_loss": 0.17057867348194122, "global_step": 100015, "epoch": 1123, "lr": 7.024799487408357e-05}
{"train_loss": 0.12825122475624084, "global_step": 100016, "epoch": 1123, "lr": 7.024746479957575e-05}
{"train_loss": 0.17009122669696808, "global_step": 100017, "epoch": 1123, "lr": 7.02469347223459e-05}
{"train_loss": 0.13411006331443787, "global_step": 100018, "epoch": 1123, "lr": 7.02464046423941e-05}
{"train_loss": 0.15405379235744476, "global_step": 100019, "epoch": 1123, "lr": 7.024587455972042e-05}
{"train_loss": 0.16395613551139832, "global_step": 100020, "epoch": 1123, "lr": 7.024534447432494e-05}
{"train_loss": 0.17359185218811035, "global_step": 100021, "epoch": 1123, "lr": 7.024481438620771e-05}
{"train_loss": 0.11930833011865616, "global_step": 100022, "epoch": 1123, "lr": 7.024428429536881e-05}
{"train_loss": 0.15672051906585693, "global_step": 100023, "epoch": 1123, "lr": 7.02437542018083e-05}
{"train_loss": 0.14247086644172668, "global_step": 100024, "epoch": 1123, "lr": 7.024322410552629e-05}
{"train_loss": 0.17526568472385406, "global_step": 100025, "epoch": 1123, "lr": 7.024269400652282e-05}
{"train_loss": 0.1282423734664917, "global_step": 100026, "epoch": 1123, "lr": 7.024216390479795e-05}
{"train_loss": 0.15088780224323273, "global_step": 100027, "epoch": 1123, "lr": 7.024163380035178e-05}
{"train_loss": 0.17497263848781586, "global_step": 100028, "epoch": 1123, "lr": 7.024110369318438e-05}
{"train_loss": 0.17894481122493744, "global_step": 100029, "epoch": 1123, "lr": 7.024057358329579e-05}
{"train_loss": 0.11435213685035706, "global_step": 100030, "epoch": 1123, "lr": 7.024004347068611e-05}
{"train_loss": 0.11825414001941681, "global_step": 100031, "epoch": 1123, "lr": 7.02395133553554e-05}
{"train_loss": 0.08202432096004486, "global_step": 100032, "epoch": 1123, "lr": 7.023898323730374e-05}
{"train_loss": 0.1267746388912201, "global_step": 100033, "epoch": 1123, "lr": 7.02384531165312e-05}
{"train_loss": 0.12580302357673645, "global_step": 100034, "epoch": 1123, "lr": 7.023792299303782e-05}
{"train_loss": 0.17175840251566318, "global_step": 100035, "epoch": 1123, "lr": 7.023739286682372e-05, "val_loss": 4.6157684326171875}
{"train_loss": 0.12863880395889282, "global_step": 100036, "epoch": 1124, "lr": 7.023686273788894e-05}
{"train_loss": 0.12307727336883545, "global_step": 100037, "epoch": 1124, "lr": 7.023633260623357e-05}
{"train_loss": 0.09123862534761429, "global_step": 100038, "epoch": 1124, "lr": 7.023580247185768e-05}
{"train_loss": 0.09918864816427231, "global_step": 100039, "epoch": 1124, "lr": 7.02352723347613e-05}
{"train_loss": 0.19601310789585114, "global_step": 100040, "epoch": 1124, "lr": 7.023474219494457e-05}
{"train_loss": 0.11104080080986023, "global_step": 100041, "epoch": 1124, "lr": 7.02342120524075e-05}
{"train_loss": 0.12992684543132782, "global_step": 100042, "epoch": 1124, "lr": 7.023368190715019e-05}
{"train_loss": 0.1725616306066513, "global_step": 100043, "epoch": 1124, "lr": 7.023315175917271e-05}
{"train_loss": 0.12227599322795868, "global_step": 100044, "epoch": 1124, "lr": 7.023262160847514e-05}
{"train_loss": 0.15926283597946167, "global_step": 100045, "epoch": 1124, "lr": 7.023209145505754e-05}
{"train_loss": 0.18136143684387207, "global_step": 100046, "epoch": 1124, "lr": 7.023156129891998e-05}
{"train_loss": 0.10603447258472443, "global_step": 100047, "epoch": 1124, "lr": 7.023103114006253e-05}
{"train_loss": 0.1569119244813919, "global_step": 100048, "epoch": 1124, "lr": 7.023050097848527e-05}
{"train_loss": 0.13481082022190094, "global_step": 100049, "epoch": 1124, "lr": 7.022997081418826e-05}
{"train_loss": 0.14331012964248657, "global_step": 100050, "epoch": 1124, "lr": 7.022944064717158e-05}
{"train_loss": 0.1339893490076065, "global_step": 100051, "epoch": 1124, "lr": 7.02289104774353e-05}
{"train_loss": 0.10527472198009491, "global_step": 100052, "epoch": 1124, "lr": 7.022838030497946e-05}
{"train_loss": 0.19666358828544617, "global_step": 100053, "epoch": 1124, "lr": 7.022785012980419e-05}
{"train_loss": 0.12477291375398636, "global_step": 100054, "epoch": 1124, "lr": 7.022731995190955e-05}
{"train_loss": 0.10139630734920502, "global_step": 100055, "epoch": 1124, "lr": 7.022678977129555e-05}
{"train_loss": 0.1975494623184204, "global_step": 100056, "epoch": 1124, "lr": 7.022625958796233e-05}
{"train_loss": 0.21460790932178497, "global_step": 100057, "epoch": 1124, "lr": 7.022572940190994e-05}
{"train_loss": 0.1050100028514862, "global_step": 100058, "epoch": 1124, "lr": 7.022519921313844e-05}
{"train_loss": 0.16185477375984192, "global_step": 100059, "epoch": 1124, "lr": 7.02246690216479e-05}
{"train_loss": 0.20813925564289093, "global_step": 100060, "epoch": 1124, "lr": 7.02241388274384e-05}
{"train_loss": 0.1896049529314041, "global_step": 100061, "epoch": 1124, "lr": 7.022360863051001e-05}
{"train_loss": 0.24153263866901398, "global_step": 100062, "epoch": 1124, "lr": 7.022307843086282e-05}
{"train_loss": 0.17349061369895935, "global_step": 100063, "epoch": 1124, "lr": 7.022254822849688e-05}
{"train_loss": 0.16111375391483307, "global_step": 100064, "epoch": 1124, "lr": 7.022201802341226e-05}
{"train_loss": 0.14392177760601044, "global_step": 100065, "epoch": 1124, "lr": 7.022148781560904e-05}
{"train_loss": 0.1678200364112854, "global_step": 100066, "epoch": 1124, "lr": 7.022095760508728e-05}
{"train_loss": 0.11119700223207474, "global_step": 100067, "epoch": 1124, "lr": 7.022042739184707e-05}
{"train_loss": 0.198466494679451, "global_step": 100068, "epoch": 1124, "lr": 7.021989717588844e-05}
{"train_loss": 0.10892893373966217, "global_step": 100069, "epoch": 1124, "lr": 7.021936695721152e-05}
{"train_loss": 0.1370847076177597, "global_step": 100070, "epoch": 1124, "lr": 7.021883673581635e-05}
{"train_loss": 0.20028120279312134, "global_step": 100071, "epoch": 1124, "lr": 7.0218306511703e-05}
{"train_loss": 0.16461153328418732, "global_step": 100072, "epoch": 1124, "lr": 7.021777628487154e-05}
{"train_loss": 0.24493981897830963, "global_step": 100073, "epoch": 1124, "lr": 7.021724605532205e-05}
{"train_loss": 0.2411756068468094, "global_step": 100074, "epoch": 1124, "lr": 7.02167158230546e-05}
{"train_loss": 0.21013009548187256, "global_step": 100075, "epoch": 1124, "lr": 7.021618558806927e-05}
{"train_loss": 0.12932465970516205, "global_step": 100076, "epoch": 1124, "lr": 7.021565535036609e-05}
{"train_loss": 0.17771349847316742, "global_step": 100077, "epoch": 1124, "lr": 7.02151251099452e-05}
{"train_loss": 0.18086306750774384, "global_step": 100078, "epoch": 1124, "lr": 7.021459486680659e-05}
{"train_loss": 0.11949507892131805, "global_step": 100079, "epoch": 1124, "lr": 7.02140646209504e-05}
{"train_loss": 0.12244356423616409, "global_step": 100080, "epoch": 1124, "lr": 7.021353437237667e-05}
{"train_loss": 0.11987327039241791, "global_step": 100081, "epoch": 1124, "lr": 7.021300412108547e-05}
{"train_loss": 0.2234543263912201, "global_step": 100082, "epoch": 1124, "lr": 7.021247386707688e-05}
{"train_loss": 0.12281771004199982, "global_step": 100083, "epoch": 1124, "lr": 7.0211943610351e-05}
{"train_loss": 0.2575254440307617, "global_step": 100084, "epoch": 1124, "lr": 7.021141335090785e-05}
{"train_loss": 0.17114762961864471, "global_step": 100085, "epoch": 1124, "lr": 7.021088308874751e-05}
{"train_loss": 0.1512032151222229, "global_step": 100086, "epoch": 1124, "lr": 7.021035282387006e-05}
{"train_loss": 0.20112691819667816, "global_step": 100087, "epoch": 1124, "lr": 7.020982255627559e-05}
{"train_loss": 0.1629234254360199, "global_step": 100088, "epoch": 1124, "lr": 7.020929228596415e-05}
{"train_loss": 0.19010047614574432, "global_step": 100089, "epoch": 1124, "lr": 7.020876201293583e-05}
{"train_loss": 0.12970681488513947, "global_step": 100090, "epoch": 1124, "lr": 7.020823173719066e-05}
{"train_loss": 0.16066494584083557, "global_step": 100091, "epoch": 1124, "lr": 7.020770145872876e-05}
{"train_loss": 0.14379668235778809, "global_step": 100092, "epoch": 1124, "lr": 7.020717117755018e-05}
{"train_loss": 0.12733307480812073, "global_step": 100093, "epoch": 1124, "lr": 7.020664089365498e-05}
{"train_loss": 0.24360725283622742, "global_step": 100094, "epoch": 1124, "lr": 7.020611060704326e-05}
{"train_loss": 0.13503728806972504, "global_step": 100095, "epoch": 1124, "lr": 7.020558031771506e-05}
{"train_loss": 0.2756226658821106, "global_step": 100096, "epoch": 1124, "lr": 7.020505002567046e-05}
{"train_loss": 0.26015791296958923, "global_step": 100097, "epoch": 1124, "lr": 7.020451973090956e-05}
{"train_loss": 0.16404859721660614, "global_step": 100098, "epoch": 1124, "lr": 7.02039894334324e-05}
{"train_loss": 0.15764674544334412, "global_step": 100099, "epoch": 1124, "lr": 7.020345913323904e-05}
{"train_loss": 0.1546023041009903, "global_step": 100100, "epoch": 1124, "lr": 7.02029288303296e-05}
{"train_loss": 0.19000881910324097, "global_step": 100101, "epoch": 1124, "lr": 7.020239852470411e-05}
{"train_loss": 0.15520651638507843, "global_step": 100102, "epoch": 1124, "lr": 7.020186821636267e-05}
{"train_loss": 0.18312686681747437, "global_step": 100103, "epoch": 1124, "lr": 7.020133790530531e-05}
{"train_loss": 0.10959785431623459, "global_step": 100104, "epoch": 1124, "lr": 7.020080759153215e-05}
{"train_loss": 0.19710440933704376, "global_step": 100105, "epoch": 1124, "lr": 7.020027727504322e-05}
{"train_loss": 0.17190170288085938, "global_step": 100106, "epoch": 1124, "lr": 7.019974695583862e-05}
{"train_loss": 0.13231509923934937, "global_step": 100107, "epoch": 1124, "lr": 7.01992166339184e-05}
{"train_loss": 0.2017916440963745, "global_step": 100108, "epoch": 1124, "lr": 7.019868630928263e-05}
{"train_loss": 0.14841242134571075, "global_step": 100109, "epoch": 1124, "lr": 7.019815598193143e-05}
{"train_loss": 0.1324046552181244, "global_step": 100110, "epoch": 1124, "lr": 7.019762565186481e-05}
{"train_loss": 0.12228520959615707, "global_step": 100111, "epoch": 1124, "lr": 7.019709531908286e-05}
{"train_loss": 0.21314002573490143, "global_step": 100112, "epoch": 1124, "lr": 7.019656498358566e-05}
{"train_loss": 0.1506175845861435, "global_step": 100113, "epoch": 1124, "lr": 7.019603464537328e-05}
{"train_loss": 0.1358591765165329, "global_step": 100114, "epoch": 1124, "lr": 7.019550430444579e-05}
{"train_loss": 0.12579354643821716, "global_step": 100115, "epoch": 1124, "lr": 7.019497396080326e-05}
{"train_loss": 0.137893408536911, "global_step": 100116, "epoch": 1124, "lr": 7.019444361444575e-05}
{"train_loss": 0.13029241561889648, "global_step": 100117, "epoch": 1124, "lr": 7.019391326537337e-05}
{"train_loss": 0.14595749974250793, "global_step": 100118, "epoch": 1124, "lr": 7.019338291358614e-05}
{"train_loss": 0.13668005168437958, "global_step": 100119, "epoch": 1124, "lr": 7.019285255908417e-05}
{"train_loss": 0.11889977008104324, "global_step": 100120, "epoch": 1124, "lr": 7.019232220186752e-05}
{"train_loss": 0.14648832380771637, "global_step": 100121, "epoch": 1124, "lr": 7.019179184193628e-05}
{"train_loss": 0.1367519050836563, "global_step": 100122, "epoch": 1124, "lr": 7.019126147929046e-05}
{"train_loss": 0.14425644278526306, "global_step": 100123, "epoch": 1124, "lr": 7.019073111393018e-05}
{"train_loss": 0.15978659722912178, "global_step": 100124, "epoch": 1124, "lr": 7.019020074585552e-05, "val_loss": 4.776017665863037}
{"train_loss": 0.09086418896913528, "global_step": 100125, "epoch": 1125, "lr": 7.018967037506651e-05}
{"train_loss": 0.22671182453632355, "global_step": 100126, "epoch": 1125, "lr": 7.018914000156325e-05}
{"train_loss": 0.1498839259147644, "global_step": 100127, "epoch": 1125, "lr": 7.018860962534581e-05}
{"train_loss": 0.1476084440946579, "global_step": 100128, "epoch": 1125, "lr": 7.018807924641427e-05}
{"train_loss": 0.15770193934440613, "global_step": 100129, "epoch": 1125, "lr": 7.018754886476867e-05}
{"train_loss": 0.11558227986097336, "global_step": 100130, "epoch": 1125, "lr": 7.018701848040911e-05}
{"train_loss": 0.15315809845924377, "global_step": 100131, "epoch": 1125, "lr": 7.018648809333564e-05}
{"train_loss": 0.17979922890663147, "global_step": 100132, "epoch": 1125, "lr": 7.018595770354835e-05}
{"train_loss": 0.11135376989841461, "global_step": 100133, "epoch": 1125, "lr": 7.01854273110473e-05}
{"train_loss": 0.08297783136367798, "global_step": 100134, "epoch": 1125, "lr": 7.018489691583257e-05}
{"train_loss": 0.12013469636440277, "global_step": 100135, "epoch": 1125, "lr": 7.018436651790423e-05}
{"train_loss": 0.16453894972801208, "global_step": 100136, "epoch": 1125, "lr": 7.018383611726234e-05}
{"train_loss": 0.18966996669769287, "global_step": 100137, "epoch": 1125, "lr": 7.018330571390699e-05}
{"train_loss": 0.16447767615318298, "global_step": 100138, "epoch": 1125, "lr": 7.018277530783823e-05}
{"train_loss": 0.13331806659698486, "global_step": 100139, "epoch": 1125, "lr": 7.018224489905615e-05}
{"train_loss": 0.16667203605175018, "global_step": 100140, "epoch": 1125, "lr": 7.018171448756082e-05}
{"train_loss": 0.17884334921836853, "global_step": 100141, "epoch": 1125, "lr": 7.01811840733523e-05}
{"train_loss": 0.192538321018219, "global_step": 100142, "epoch": 1125, "lr": 7.018065365643065e-05}
{"train_loss": 0.17143958806991577, "global_step": 100143, "epoch": 1125, "lr": 7.018012323679597e-05}
{"train_loss": 0.15883591771125793, "global_step": 100144, "epoch": 1125, "lr": 7.017959281444832e-05}
{"train_loss": 0.13993096351623535, "global_step": 100145, "epoch": 1125, "lr": 7.017906238938776e-05}
{"train_loss": 0.12464752793312073, "global_step": 100146, "epoch": 1125, "lr": 7.017853196161438e-05}
{"train_loss": 0.0996779128909111, "global_step": 100147, "epoch": 1125, "lr": 7.017800153112823e-05}
{"train_loss": 0.11924445629119873, "global_step": 100148, "epoch": 1125, "lr": 7.017747109792941e-05}
{"train_loss": 0.17423434555530548, "global_step": 100149, "epoch": 1125, "lr": 7.017694066201796e-05}
{"train_loss": 0.09559416025876999, "global_step": 100150, "epoch": 1125, "lr": 7.017641022339399e-05}
{"train_loss": 0.2338462769985199, "global_step": 100151, "epoch": 1125, "lr": 7.01758797820575e-05}
{"train_loss": 0.15019477903842926, "global_step": 100152, "epoch": 1125, "lr": 7.017534933800866e-05}
{"train_loss": 0.1681346595287323, "global_step": 100153, "epoch": 1125, "lr": 7.017481889124746e-05}
{"train_loss": 0.2720484137535095, "global_step": 100154, "epoch": 1125, "lr": 7.017428844177402e-05}
{"train_loss": 0.14180971682071686, "global_step": 100155, "epoch": 1125, "lr": 7.017375798958838e-05}
{"train_loss": 0.10360884666442871, "global_step": 100156, "epoch": 1125, "lr": 7.017322753469063e-05}
{"train_loss": 0.12836025655269623, "global_step": 100157, "epoch": 1125, "lr": 7.017269707708082e-05}
{"train_loss": 0.13635748624801636, "global_step": 100158, "epoch": 1125, "lr": 7.017216661675906e-05}
{"train_loss": 0.2732061445713043, "global_step": 100159, "epoch": 1125, "lr": 7.017163615372539e-05}
{"train_loss": 0.14437784254550934, "global_step": 100160, "epoch": 1125, "lr": 7.01711056879799e-05}
{"train_loss": 0.2098797857761383, "global_step": 100161, "epoch": 1125, "lr": 7.017057521952264e-05}
{"train_loss": 0.2621224820613861, "global_step": 100162, "epoch": 1125, "lr": 7.017004474835369e-05}
{"train_loss": 0.19326888024806976, "global_step": 100163, "epoch": 1125, "lr": 7.016951427447313e-05}
{"train_loss": 0.23325012624263763, "global_step": 100164, "epoch": 1125, "lr": 7.016898379788102e-05}
{"train_loss": 0.16716268658638, "global_step": 100165, "epoch": 1125, "lr": 7.016845331857741e-05}
{"train_loss": 0.10212802141904831, "global_step": 100166, "epoch": 1125, "lr": 7.016792283656244e-05}
{"train_loss": 0.2236575484275818, "global_step": 100167, "epoch": 1125, "lr": 7.016739235183611e-05}
{"train_loss": 0.11090527474880219, "global_step": 100168, "epoch": 1125, "lr": 7.016686186439854e-05}
{"train_loss": 0.10706218332052231, "global_step": 100169, "epoch": 1125, "lr": 7.016633137424976e-05}
{"train_loss": 0.17539626359939575, "global_step": 100170, "epoch": 1125, "lr": 7.016580088138988e-05}
{"train_loss": 0.13348932564258575, "global_step": 100171, "epoch": 1125, "lr": 7.016527038581895e-05}
{"train_loss": 0.17216205596923828, "global_step": 100172, "epoch": 1125, "lr": 7.016473988753704e-05}
{"train_loss": 0.1518741101026535, "global_step": 100173, "epoch": 1125, "lr": 7.016420938654422e-05}
{"train_loss": 0.13858309388160706, "global_step": 100174, "epoch": 1125, "lr": 7.01636788828406e-05}
{"train_loss": 0.15608808398246765, "global_step": 100175, "epoch": 1125, "lr": 7.016314837642618e-05}
{"train_loss": 0.2407013177871704, "global_step": 100176, "epoch": 1125, "lr": 7.016261786730109e-05}
{"train_loss": 0.10263802856206894, "global_step": 100177, "epoch": 1125, "lr": 7.016208735546537e-05}
{"train_loss": 0.19024163484573364, "global_step": 100178, "epoch": 1125, "lr": 7.016155684091911e-05}
{"train_loss": 0.18742893636226654, "global_step": 100179, "epoch": 1125, "lr": 7.016102632366239e-05}
{"train_loss": 0.1131344810128212, "global_step": 100180, "epoch": 1125, "lr": 7.016049580369527e-05}
{"train_loss": 0.15761823952198029, "global_step": 100181, "epoch": 1125, "lr": 7.015996528101779e-05}
{"train_loss": 0.2204524427652359, "global_step": 100182, "epoch": 1125, "lr": 7.015943475563006e-05}
{"train_loss": 0.21009397506713867, "global_step": 100183, "epoch": 1125, "lr": 7.015890422753213e-05}
{"train_loss": 0.09212307631969452, "global_step": 100184, "epoch": 1125, "lr": 7.015837369672409e-05}
{"train_loss": 0.13159313797950745, "global_step": 100185, "epoch": 1125, "lr": 7.0157843163206e-05}
{"train_loss": 0.15007668733596802, "global_step": 100186, "epoch": 1125, "lr": 7.015731262697793e-05}
{"train_loss": 0.17817270755767822, "global_step": 100187, "epoch": 1125, "lr": 7.015678208803996e-05}
{"train_loss": 0.15889371931552887, "global_step": 100188, "epoch": 1125, "lr": 7.015625154639215e-05}
{"train_loss": 0.2394133061170578, "global_step": 100189, "epoch": 1125, "lr": 7.015572100203457e-05}
{"train_loss": 0.11735890805721283, "global_step": 100190, "epoch": 1125, "lr": 7.015519045496732e-05}
{"train_loss": 0.17331907153129578, "global_step": 100191, "epoch": 1125, "lr": 7.015465990519046e-05}
{"train_loss": 0.10175219178199768, "global_step": 100192, "epoch": 1125, "lr": 7.015412935270402e-05}
{"train_loss": 0.14707571268081665, "global_step": 100193, "epoch": 1125, "lr": 7.015359879750812e-05}
{"train_loss": 0.131373792886734, "global_step": 100194, "epoch": 1125, "lr": 7.015306823960282e-05}
{"train_loss": 0.18146727979183197, "global_step": 100195, "epoch": 1125, "lr": 7.015253767898817e-05}
{"train_loss": 0.11812349408864975, "global_step": 100196, "epoch": 1125, "lr": 7.015200711566426e-05}
{"train_loss": 0.21152982115745544, "global_step": 100197, "epoch": 1125, "lr": 7.015147654963116e-05}
{"train_loss": 0.18069449067115784, "global_step": 100198, "epoch": 1125, "lr": 7.015094598088894e-05}
{"train_loss": 0.21917706727981567, "global_step": 100199, "epoch": 1125, "lr": 7.01504154094377e-05}
{"train_loss": 0.17899969220161438, "global_step": 100200, "epoch": 1125, "lr": 7.014988483527744e-05}
{"train_loss": 0.11716011166572571, "global_step": 100201, "epoch": 1125, "lr": 7.014935425840829e-05}
{"train_loss": 0.17802056670188904, "global_step": 100202, "epoch": 1125, "lr": 7.014882367883031e-05}
{"train_loss": 0.15616776049137115, "global_step": 100203, "epoch": 1125, "lr": 7.014829309654356e-05}
{"train_loss": 0.1518002152442932, "global_step": 100204, "epoch": 1125, "lr": 7.014776251154812e-05}
{"train_loss": 0.13836663961410522, "global_step": 100205, "epoch": 1125, "lr": 7.014723192384405e-05}
{"train_loss": 0.2457147240638733, "global_step": 100206, "epoch": 1125, "lr": 7.014670133343144e-05}
{"train_loss": 0.1832999885082245, "global_step": 100207, "epoch": 1125, "lr": 7.014617074031033e-05}
{"train_loss": 0.14663201570510864, "global_step": 100208, "epoch": 1125, "lr": 7.014564014448084e-05}
{"train_loss": 0.17939122021198273, "global_step": 100209, "epoch": 1125, "lr": 7.0145109545943e-05}
{"train_loss": 0.19392472505569458, "global_step": 100210, "epoch": 1125, "lr": 7.014457894469691e-05}
{"train_loss": 0.1366955190896988, "global_step": 100211, "epoch": 1125, "lr": 7.014404834074262e-05}
{"train_loss": 0.18688616156578064, "global_step": 100212, "epoch": 1125, "lr": 7.01435177340802e-05}
{"train_loss": 0.16113882838340288, "global_step": 100213, "epoch": 1125, "lr": 7.014298712470974e-05, "val_loss": 4.695184230804443, "train_action_mse_error": 13.981497764587402}
{"train_loss": 0.15136465430259705, "global_step": 100214, "epoch": 1126, "lr": 7.014245651263129e-05}
{"train_loss": 0.18386442959308624, "global_step": 100215, "epoch": 1126, "lr": 7.014192589784494e-05}
{"train_loss": 0.12530893087387085, "global_step": 100216, "epoch": 1126, "lr": 7.014139528035075e-05}
{"train_loss": 0.11152679473161697, "global_step": 100217, "epoch": 1126, "lr": 7.01408646601488e-05}
{"train_loss": 0.20205380022525787, "global_step": 100218, "epoch": 1126, "lr": 7.014033403723917e-05}
{"train_loss": 0.17823423445224762, "global_step": 100219, "epoch": 1126, "lr": 7.01398034116219e-05}
{"train_loss": 0.15367408096790314, "global_step": 100220, "epoch": 1126, "lr": 7.013927278329708e-05}
{"train_loss": 0.12675990164279938, "global_step": 100221, "epoch": 1126, "lr": 7.013874215226477e-05}
{"train_loss": 0.1581472009420395, "global_step": 100222, "epoch": 1126, "lr": 7.013821151852506e-05}
{"train_loss": 0.1432778537273407, "global_step": 100223, "epoch": 1126, "lr": 7.013768088207801e-05}
{"train_loss": 0.16725341975688934, "global_step": 100224, "epoch": 1126, "lr": 7.01371502429237e-05}
{"train_loss": 0.17182940244674683, "global_step": 100225, "epoch": 1126, "lr": 7.01366196010622e-05}
{"train_loss": 0.2646648585796356, "global_step": 100226, "epoch": 1126, "lr": 7.013608895649356e-05}
{"train_loss": 0.17710980772972107, "global_step": 100227, "epoch": 1126, "lr": 7.013555830921789e-05}
{"train_loss": 0.19010619819164276, "global_step": 100228, "epoch": 1126, "lr": 7.013502765923522e-05}
{"train_loss": 0.11668210476636887, "global_step": 100229, "epoch": 1126, "lr": 7.013449700654565e-05}
{"train_loss": 0.15269185602664948, "global_step": 100230, "epoch": 1126, "lr": 7.013396635114923e-05}
{"train_loss": 0.19849495589733124, "global_step": 100231, "epoch": 1126, "lr": 7.013343569304607e-05}
{"train_loss": 0.13478942215442657, "global_step": 100232, "epoch": 1126, "lr": 7.013290503223619e-05}
{"train_loss": 0.20679768919944763, "global_step": 100233, "epoch": 1126, "lr": 7.013237436871971e-05}
{"train_loss": 0.14381663501262665, "global_step": 100234, "epoch": 1126, "lr": 7.013184370249666e-05}
{"train_loss": 0.2485680878162384, "global_step": 100235, "epoch": 1126, "lr": 7.013131303356714e-05}
{"train_loss": 0.1565016210079193, "global_step": 100236, "epoch": 1126, "lr": 7.01307823619312e-05}
{"train_loss": 0.17719481885433197, "global_step": 100237, "epoch": 1126, "lr": 7.013025168758894e-05}
{"train_loss": 0.15707387030124664, "global_step": 100238, "epoch": 1126, "lr": 7.01297210105404e-05}
{"train_loss": 0.11855366826057434, "global_step": 100239, "epoch": 1126, "lr": 7.012919033078566e-05}
{"train_loss": 0.0912095382809639, "global_step": 100240, "epoch": 1126, "lr": 7.01286596483248e-05}
{"train_loss": 0.16994239389896393, "global_step": 100241, "epoch": 1126, "lr": 7.012812896315788e-05}
{"train_loss": 0.15900897979736328, "global_step": 100242, "epoch": 1126, "lr": 7.012759827528497e-05}
{"train_loss": 0.28018826246261597, "global_step": 100243, "epoch": 1126, "lr": 7.012706758470617e-05}
{"train_loss": 0.259859561920166, "global_step": 100244, "epoch": 1126, "lr": 7.012653689142152e-05}
{"train_loss": 0.190414696931839, "global_step": 100245, "epoch": 1126, "lr": 7.012600619543112e-05}
{"train_loss": 0.13640347123146057, "global_step": 100246, "epoch": 1126, "lr": 7.0125475496735e-05}
{"train_loss": 0.14269644021987915, "global_step": 100247, "epoch": 1126, "lr": 7.012494479533327e-05}
{"train_loss": 0.13642890751361847, "global_step": 100248, "epoch": 1126, "lr": 7.012441409122598e-05}
{"train_loss": 0.13288505375385284, "global_step": 100249, "epoch": 1126, "lr": 7.012388338441322e-05}
{"train_loss": 0.20878301560878754, "global_step": 100250, "epoch": 1126, "lr": 7.012335267489503e-05}
{"train_loss": 0.1781543642282486, "global_step": 100251, "epoch": 1126, "lr": 7.012282196267151e-05}
{"train_loss": 0.12490343302488327, "global_step": 100252, "epoch": 1126, "lr": 7.012229124774272e-05}
{"train_loss": 0.16530989110469818, "global_step": 100253, "epoch": 1126, "lr": 7.012176053010875e-05}
{"train_loss": 0.18929053843021393, "global_step": 100254, "epoch": 1126, "lr": 7.012122980976963e-05}
{"train_loss": 0.23264648020267487, "global_step": 100255, "epoch": 1126, "lr": 7.012069908672547e-05}
{"train_loss": 0.18139877915382385, "global_step": 100256, "epoch": 1126, "lr": 7.012016836097632e-05}
{"train_loss": 0.18323884904384613, "global_step": 100257, "epoch": 1126, "lr": 7.011963763252227e-05}
{"train_loss": 0.18487045168876648, "global_step": 100258, "epoch": 1126, "lr": 7.011910690136337e-05}
{"train_loss": 0.10728374123573303, "global_step": 100259, "epoch": 1126, "lr": 7.01185761674997e-05}
{"train_loss": 0.18486103415489197, "global_step": 100260, "epoch": 1126, "lr": 7.011804543093134e-05}
{"train_loss": 0.13770611584186554, "global_step": 100261, "epoch": 1126, "lr": 7.011751469165835e-05}
{"train_loss": 0.25408223271369934, "global_step": 100262, "epoch": 1126, "lr": 7.01169839496808e-05}
{"train_loss": 0.15817636251449585, "global_step": 100263, "epoch": 1126, "lr": 7.011645320499876e-05}
{"train_loss": 0.21989378333091736, "global_step": 100264, "epoch": 1126, "lr": 7.011592245761233e-05}
{"train_loss": 0.24701723456382751, "global_step": 100265, "epoch": 1126, "lr": 7.011539170752155e-05}
{"train_loss": 0.14901696145534515, "global_step": 100266, "epoch": 1126, "lr": 7.01148609547265e-05}
{"train_loss": 0.19719268381595612, "global_step": 100267, "epoch": 1126, "lr": 7.011433019922725e-05}
{"train_loss": 0.17696382105350494, "global_step": 100268, "epoch": 1126, "lr": 7.011379944102388e-05}
{"train_loss": 0.1878662407398224, "global_step": 100269, "epoch": 1126, "lr": 7.011326868011646e-05}
{"train_loss": 0.23684698343276978, "global_step": 100270, "epoch": 1126, "lr": 7.011273791650504e-05}
{"train_loss": 0.1898345798254013, "global_step": 100271, "epoch": 1126, "lr": 7.011220715018971e-05}
{"train_loss": 0.15005014836788177, "global_step": 100272, "epoch": 1126, "lr": 7.011167638117055e-05}
{"train_loss": 0.1314263641834259, "global_step": 100273, "epoch": 1126, "lr": 7.01111456094476e-05}
{"train_loss": 0.13413113355636597, "global_step": 100274, "epoch": 1126, "lr": 7.011061483502097e-05}
{"train_loss": 0.1584511250257492, "global_step": 100275, "epoch": 1126, "lr": 7.01100840578907e-05}
{"train_loss": 0.11731250584125519, "global_step": 100276, "epoch": 1126, "lr": 7.010955327805689e-05}
{"train_loss": 0.13932308554649353, "global_step": 100277, "epoch": 1126, "lr": 7.01090224955196e-05}
{"train_loss": 0.2540508508682251, "global_step": 100278, "epoch": 1126, "lr": 7.010849171027887e-05}
{"train_loss": 0.15877828001976013, "global_step": 100279, "epoch": 1126, "lr": 7.010796092233481e-05}
{"train_loss": 0.16370292007923126, "global_step": 100280, "epoch": 1126, "lr": 7.010743013168748e-05}
{"train_loss": 0.17255949974060059, "global_step": 100281, "epoch": 1126, "lr": 7.010689933833697e-05}
{"train_loss": 0.15268780291080475, "global_step": 100282, "epoch": 1126, "lr": 7.01063685422833e-05}
{"train_loss": 0.11300045251846313, "global_step": 100283, "epoch": 1126, "lr": 7.010583774352659e-05}
{"train_loss": 0.16714611649513245, "global_step": 100284, "epoch": 1126, "lr": 7.010530694206689e-05}
{"train_loss": 0.2309858202934265, "global_step": 100285, "epoch": 1126, "lr": 7.010477613790429e-05}
{"train_loss": 0.1578134000301361, "global_step": 100286, "epoch": 1126, "lr": 7.010424533103884e-05}
{"train_loss": 0.257906436920166, "global_step": 100287, "epoch": 1126, "lr": 7.010371452147063e-05}
{"train_loss": 0.1177954375743866, "global_step": 100288, "epoch": 1126, "lr": 7.010318370919971e-05}
{"train_loss": 0.22313235700130463, "global_step": 100289, "epoch": 1126, "lr": 7.010265289422615e-05}
{"train_loss": 0.11487556248903275, "global_step": 100290, "epoch": 1126, "lr": 7.010212207655004e-05}
{"train_loss": 0.12363212555646896, "global_step": 100291, "epoch": 1126, "lr": 7.010159125617146e-05}
{"train_loss": 0.21640250086784363, "global_step": 100292, "epoch": 1126, "lr": 7.010106043309044e-05}
{"train_loss": 0.2502351999282837, "global_step": 100293, "epoch": 1126, "lr": 7.01005296073071e-05}
{"train_loss": 0.11859563738107681, "global_step": 100294, "epoch": 1126, "lr": 7.009999877882148e-05}
{"train_loss": 0.22286048531532288, "global_step": 100295, "epoch": 1126, "lr": 7.009946794763367e-05}
{"train_loss": 0.2503208816051483, "global_step": 100296, "epoch": 1126, "lr": 7.009893711374372e-05}
{"train_loss": 0.14122982323169708, "global_step": 100297, "epoch": 1126, "lr": 7.009840627715172e-05}
{"train_loss": 0.16833147406578064, "global_step": 100298, "epoch": 1126, "lr": 7.009787543785773e-05}
{"train_loss": 0.24425329267978668, "global_step": 100299, "epoch": 1126, "lr": 7.009734459586183e-05}
{"train_loss": 0.21131131052970886, "global_step": 100300, "epoch": 1126, "lr": 7.009681375116408e-05}
{"train_loss": 0.17647895216941833, "global_step": 100301, "epoch": 1126, "lr": 7.009628290376456e-05}
{"train_loss": 0.17398232966661453, "global_step": 100302, "epoch": 1126, "lr": 7.009575205366335e-05, "val_loss": 4.652027606964111}
{"train_loss": 0.17252837121486664, "global_step": 100303, "epoch": 1127, "lr": 7.00952212008605e-05}
{"train_loss": 0.08589956909418106, "global_step": 100304, "epoch": 1127, "lr": 7.009469034535609e-05}
{"train_loss": 0.21851257979869843, "global_step": 100305, "epoch": 1127, "lr": 7.00941594871502e-05}
{"train_loss": 0.165972501039505, "global_step": 100306, "epoch": 1127, "lr": 7.009362862624289e-05}
{"train_loss": 0.0970129668712616, "global_step": 100307, "epoch": 1127, "lr": 7.009309776263425e-05}
{"train_loss": 0.13386572897434235, "global_step": 100308, "epoch": 1127, "lr": 7.009256689632432e-05}
{"train_loss": 0.13342167437076569, "global_step": 100309, "epoch": 1127, "lr": 7.009203602731319e-05}
{"train_loss": 0.20126745104789734, "global_step": 100310, "epoch": 1127, "lr": 7.009150515560095e-05}
{"train_loss": 0.17139944434165955, "global_step": 100311, "epoch": 1127, "lr": 7.009097428118763e-05}
{"train_loss": 0.20979724824428558, "global_step": 100312, "epoch": 1127, "lr": 7.009044340407333e-05}
{"train_loss": 0.13965050876140594, "global_step": 100313, "epoch": 1127, "lr": 7.008991252425813e-05}
{"train_loss": 0.18590563535690308, "global_step": 100314, "epoch": 1127, "lr": 7.008938164174207e-05}
{"train_loss": 0.156339630484581, "global_step": 100315, "epoch": 1127, "lr": 7.008885075652524e-05}
{"train_loss": 0.11638709902763367, "global_step": 100316, "epoch": 1127, "lr": 7.008831986860771e-05}
{"train_loss": 0.10498970001935959, "global_step": 100317, "epoch": 1127, "lr": 7.008778897798956e-05}
{"train_loss": 0.17909401655197144, "global_step": 100318, "epoch": 1127, "lr": 7.008725808467083e-05}
{"train_loss": 0.18114282190799713, "global_step": 100319, "epoch": 1127, "lr": 7.008672718865164e-05}
{"train_loss": 0.14930559694766998, "global_step": 100320, "epoch": 1127, "lr": 7.0086196289932e-05}
{"train_loss": 0.11767861992120743, "global_step": 100321, "epoch": 1127, "lr": 7.008566538851205e-05}
{"train_loss": 0.13598380982875824, "global_step": 100322, "epoch": 1127, "lr": 7.008513448439181e-05}
{"train_loss": 0.1004614308476448, "global_step": 100323, "epoch": 1127, "lr": 7.008460357757137e-05}
{"train_loss": 0.14267098903656006, "global_step": 100324, "epoch": 1127, "lr": 7.008407266805079e-05}
{"train_loss": 0.22613336145877838, "global_step": 100325, "epoch": 1127, "lr": 7.008354175583018e-05}
{"train_loss": 0.21034719049930573, "global_step": 100326, "epoch": 1127, "lr": 7.008301084090956e-05}
{"train_loss": 0.11421139538288116, "global_step": 100327, "epoch": 1127, "lr": 7.008247992328903e-05}
{"train_loss": 0.23480840027332306, "global_step": 100328, "epoch": 1127, "lr": 7.008194900296865e-05}
{"train_loss": 0.22757776081562042, "global_step": 100329, "epoch": 1127, "lr": 7.00814180799485e-05}
{"train_loss": 0.1379106640815735, "global_step": 100330, "epoch": 1127, "lr": 7.008088715422865e-05}
{"train_loss": 0.23589888215065002, "global_step": 100331, "epoch": 1127, "lr": 7.008035622580918e-05}
{"train_loss": 0.10674164444208145, "global_step": 100332, "epoch": 1127, "lr": 7.007982529469012e-05}
{"train_loss": 0.26720255613327026, "global_step": 100333, "epoch": 1127, "lr": 7.007929436087159e-05}
{"train_loss": 0.16965484619140625, "global_step": 100334, "epoch": 1127, "lr": 7.007876342435366e-05}
{"train_loss": 0.1684264987707138, "global_step": 100335, "epoch": 1127, "lr": 7.007823248513637e-05}
{"train_loss": 0.1953861564397812, "global_step": 100336, "epoch": 1127, "lr": 7.00777015432198e-05}
{"train_loss": 0.1460544615983963, "global_step": 100337, "epoch": 1127, "lr": 7.007717059860404e-05}
{"train_loss": 0.17564503848552704, "global_step": 100338, "epoch": 1127, "lr": 7.007663965128914e-05}
{"train_loss": 0.120008684694767, "global_step": 100339, "epoch": 1127, "lr": 7.007610870127518e-05}
{"train_loss": 0.19913697242736816, "global_step": 100340, "epoch": 1127, "lr": 7.007557774856224e-05}
{"train_loss": 0.22169983386993408, "global_step": 100341, "epoch": 1127, "lr": 7.007504679315038e-05}
{"train_loss": 0.1551278829574585, "global_step": 100342, "epoch": 1127, "lr": 7.007451583503966e-05}
{"train_loss": 0.20140066742897034, "global_step": 100343, "epoch": 1127, "lr": 7.007398487423019e-05}
{"train_loss": 0.16639827191829681, "global_step": 100344, "epoch": 1127, "lr": 7.0073453910722e-05}
{"train_loss": 0.2603908181190491, "global_step": 100345, "epoch": 1127, "lr": 7.00729229445152e-05}
{"train_loss": 0.12153893709182739, "global_step": 100346, "epoch": 1127, "lr": 7.007239197560982e-05}
{"train_loss": 0.25240468978881836, "global_step": 100347, "epoch": 1127, "lr": 7.007186100400595e-05}
{"train_loss": 0.16506467759609222, "global_step": 100348, "epoch": 1127, "lr": 7.007133002970368e-05}
{"train_loss": 0.18047650158405304, "global_step": 100349, "epoch": 1127, "lr": 7.007079905270304e-05}
{"train_loss": 0.1846088320016861, "global_step": 100350, "epoch": 1127, "lr": 7.007026807300416e-05}
{"train_loss": 0.11304927617311478, "global_step": 100351, "epoch": 1127, "lr": 7.006973709060705e-05}
{"train_loss": 0.19208131730556488, "global_step": 100352, "epoch": 1127, "lr": 7.006920610551182e-05}
{"train_loss": 0.11321216076612473, "global_step": 100353, "epoch": 1127, "lr": 7.006867511771852e-05}
{"train_loss": 0.2285785675048828, "global_step": 100354, "epoch": 1127, "lr": 7.006814412722724e-05}
{"train_loss": 0.1596800535917282, "global_step": 100355, "epoch": 1127, "lr": 7.006761313403804e-05}
{"train_loss": 0.10135287046432495, "global_step": 100356, "epoch": 1127, "lr": 7.006708213815098e-05}
{"train_loss": 0.1764981597661972, "global_step": 100357, "epoch": 1127, "lr": 7.006655113956617e-05}
{"train_loss": 0.14375634491443634, "global_step": 100358, "epoch": 1127, "lr": 7.006602013828365e-05}
{"train_loss": 0.15777577459812164, "global_step": 100359, "epoch": 1127, "lr": 7.006548913430348e-05}
{"train_loss": 0.1486043483018875, "global_step": 100360, "epoch": 1127, "lr": 7.006495812762577e-05}
{"train_loss": 0.11094518005847931, "global_step": 100361, "epoch": 1127, "lr": 7.006442711825057e-05}
{"train_loss": 0.15011116862297058, "global_step": 100362, "epoch": 1127, "lr": 7.006389610617795e-05}
{"train_loss": 0.16873914003372192, "global_step": 100363, "epoch": 1127, "lr": 7.006336509140798e-05}
{"train_loss": 0.11188088357448578, "global_step": 100364, "epoch": 1127, "lr": 7.006283407394073e-05}
{"train_loss": 0.1088772565126419, "global_step": 100365, "epoch": 1127, "lr": 7.006230305377628e-05}
{"train_loss": 0.1739799827337265, "global_step": 100366, "epoch": 1127, "lr": 7.006177203091471e-05}
{"train_loss": 0.2181251496076584, "global_step": 100367, "epoch": 1127, "lr": 7.006124100535606e-05}
{"train_loss": 0.1545177847146988, "global_step": 100368, "epoch": 1127, "lr": 7.006070997710043e-05}
{"train_loss": 0.15149088203907013, "global_step": 100369, "epoch": 1127, "lr": 7.006017894614787e-05}
{"train_loss": 0.15660463273525238, "global_step": 100370, "epoch": 1127, "lr": 7.005964791249849e-05}
{"train_loss": 0.1156291663646698, "global_step": 100371, "epoch": 1127, "lr": 7.005911687615233e-05}
{"train_loss": 0.14329278469085693, "global_step": 100372, "epoch": 1127, "lr": 7.005858583710944e-05}
{"train_loss": 0.20260737836360931, "global_step": 100373, "epoch": 1127, "lr": 7.005805479536994e-05}
{"train_loss": 0.14544980227947235, "global_step": 100374, "epoch": 1127, "lr": 7.005752375093387e-05}
{"train_loss": 0.17996980249881744, "global_step": 100375, "epoch": 1127, "lr": 7.005699270380132e-05}
{"train_loss": 0.19411538541316986, "global_step": 100376, "epoch": 1127, "lr": 7.005646165397234e-05}
{"train_loss": 0.1566619575023651, "global_step": 100377, "epoch": 1127, "lr": 7.005593060144702e-05}
{"train_loss": 0.21919168531894684, "global_step": 100378, "epoch": 1127, "lr": 7.005539954622542e-05}
{"train_loss": 0.10548749566078186, "global_step": 100379, "epoch": 1127, "lr": 7.005486848830763e-05}
{"train_loss": 0.09955409914255142, "global_step": 100380, "epoch": 1127, "lr": 7.00543374276937e-05}
{"train_loss": 0.17730599641799927, "global_step": 100381, "epoch": 1127, "lr": 7.005380636438369e-05}
{"train_loss": 0.18251408636569977, "global_step": 100382, "epoch": 1127, "lr": 7.005327529837771e-05}
{"train_loss": 0.16612814366817474, "global_step": 100383, "epoch": 1127, "lr": 7.005274422967581e-05}
{"train_loss": 0.21825803816318512, "global_step": 100384, "epoch": 1127, "lr": 7.005221315827806e-05}
{"train_loss": 0.12189372628927231, "global_step": 100385, "epoch": 1127, "lr": 7.005168208418453e-05}
{"train_loss": 0.15048055350780487, "global_step": 100386, "epoch": 1127, "lr": 7.005115100739529e-05}
{"train_loss": 0.17282634973526, "global_step": 100387, "epoch": 1127, "lr": 7.005061992791043e-05}
{"train_loss": 0.17345930635929108, "global_step": 100388, "epoch": 1127, "lr": 7.005008884573002e-05}
{"train_loss": 0.1802307367324829, "global_step": 100389, "epoch": 1127, "lr": 7.00495577608541e-05}
{"train_loss": 0.11459137499332428, "global_step": 100390, "epoch": 1127, "lr": 7.004902667328277e-05}
{"train_loss": 0.1631241264972794, "global_step": 100391, "epoch": 1127, "lr": 7.004849558301609e-05, "val_loss": 4.709888458251953}
{"train_loss": 0.10977784544229507, "global_step": 100392, "epoch": 1128, "lr": 7.004796449005414e-05}
{"train_loss": 0.10140947252511978, "global_step": 100393, "epoch": 1128, "lr": 7.004743339439696e-05}
{"train_loss": 0.16620351374149323, "global_step": 100394, "epoch": 1128, "lr": 7.004690229604467e-05}
{"train_loss": 0.19477365911006927, "global_step": 100395, "epoch": 1128, "lr": 7.004637119499732e-05}
{"train_loss": 0.19238148629665375, "global_step": 100396, "epoch": 1128, "lr": 7.004584009125497e-05}
{"train_loss": 0.19968275725841522, "global_step": 100397, "epoch": 1128, "lr": 7.004530898481771e-05}
{"train_loss": 0.09471495449542999, "global_step": 100398, "epoch": 1128, "lr": 7.00447778756856e-05}
{"train_loss": 0.23176369071006775, "global_step": 100399, "epoch": 1128, "lr": 7.00442467638587e-05}
{"train_loss": 0.11529841274023056, "global_step": 100400, "epoch": 1128, "lr": 7.004371564933712e-05}
{"train_loss": 0.1687707006931305, "global_step": 100401, "epoch": 1128, "lr": 7.004318453212089e-05}
{"train_loss": 0.14351478219032288, "global_step": 100402, "epoch": 1128, "lr": 7.00426534122101e-05}
{"train_loss": 0.27742287516593933, "global_step": 100403, "epoch": 1128, "lr": 7.004212228960481e-05}
{"train_loss": 0.14017480611801147, "global_step": 100404, "epoch": 1128, "lr": 7.004159116430512e-05}
{"train_loss": 0.16911481320858002, "global_step": 100405, "epoch": 1128, "lr": 7.004106003631106e-05}
{"train_loss": 0.11697416752576828, "global_step": 100406, "epoch": 1128, "lr": 7.004052890562276e-05}
{"train_loss": 0.12060936540365219, "global_step": 100407, "epoch": 1128, "lr": 7.003999777224023e-05}
{"train_loss": 0.1452556550502777, "global_step": 100408, "epoch": 1128, "lr": 7.003946663616356e-05}
{"train_loss": 0.1680566966533661, "global_step": 100409, "epoch": 1128, "lr": 7.003893549739284e-05}
{"train_loss": 0.3235740661621094, "global_step": 100410, "epoch": 1128, "lr": 7.003840435592813e-05}
{"train_loss": 0.14679816365242004, "global_step": 100411, "epoch": 1128, "lr": 7.00378732117695e-05}
{"train_loss": 0.22317719459533691, "global_step": 100412, "epoch": 1128, "lr": 7.003734206491702e-05}
{"train_loss": 0.11379624158143997, "global_step": 100413, "epoch": 1128, "lr": 7.003681091537076e-05}
{"train_loss": 0.12242627888917923, "global_step": 100414, "epoch": 1128, "lr": 7.003627976313081e-05}
{"train_loss": 0.1732998937368393, "global_step": 100415, "epoch": 1128, "lr": 7.003574860819722e-05}
{"train_loss": 0.15806341171264648, "global_step": 100416, "epoch": 1128, "lr": 7.003521745057006e-05}
{"train_loss": 0.14837674796581268, "global_step": 100417, "epoch": 1128, "lr": 7.00346862902494e-05}
{"train_loss": 0.20945484936237335, "global_step": 100418, "epoch": 1128, "lr": 7.003415512723534e-05}
{"train_loss": 0.20025378465652466, "global_step": 100419, "epoch": 1128, "lr": 7.003362396152792e-05}
{"train_loss": 0.22587966918945312, "global_step": 100420, "epoch": 1128, "lr": 7.003309279312723e-05}
{"train_loss": 0.12976519763469696, "global_step": 100421, "epoch": 1128, "lr": 7.003256162203333e-05}
{"train_loss": 0.11997843533754349, "global_step": 100422, "epoch": 1128, "lr": 7.003203044824631e-05}
{"train_loss": 0.13915085792541504, "global_step": 100423, "epoch": 1128, "lr": 7.00314992717662e-05}
{"train_loss": 0.15945769846439362, "global_step": 100424, "epoch": 1128, "lr": 7.003096809259312e-05}
{"train_loss": 0.14751587808132172, "global_step": 100425, "epoch": 1128, "lr": 7.003043691072711e-05}
{"train_loss": 0.25792738795280457, "global_step": 100426, "epoch": 1128, "lr": 7.002990572616828e-05}
{"train_loss": 0.16763098537921906, "global_step": 100427, "epoch": 1128, "lr": 7.002937453891665e-05}
{"train_loss": 0.1443185657262802, "global_step": 100428, "epoch": 1128, "lr": 7.002884334897232e-05}
{"train_loss": 0.16401812434196472, "global_step": 100429, "epoch": 1128, "lr": 7.002831215633534e-05}
{"train_loss": 0.13247479498386383, "global_step": 100430, "epoch": 1128, "lr": 7.002778096100582e-05}
{"train_loss": 0.1299382448196411, "global_step": 100431, "epoch": 1128, "lr": 7.00272497629838e-05}
{"train_loss": 0.16993431746959686, "global_step": 100432, "epoch": 1128, "lr": 7.002671856226936e-05}
{"train_loss": 0.2453015297651291, "global_step": 100433, "epoch": 1128, "lr": 7.002618735886257e-05}
{"train_loss": 0.19619238376617432, "global_step": 100434, "epoch": 1128, "lr": 7.00256561527635e-05}
{"train_loss": 0.12351024150848389, "global_step": 100435, "epoch": 1128, "lr": 7.002512494397224e-05}
{"train_loss": 0.1408681422472, "global_step": 100436, "epoch": 1128, "lr": 7.002459373248883e-05}
{"train_loss": 0.18122674524784088, "global_step": 100437, "epoch": 1128, "lr": 7.002406251831335e-05}
{"train_loss": 0.2765445113182068, "global_step": 100438, "epoch": 1128, "lr": 7.00235313014459e-05}
{"train_loss": 0.20939987897872925, "global_step": 100439, "epoch": 1128, "lr": 7.002300008188653e-05}
{"train_loss": 0.2301110178232193, "global_step": 100440, "epoch": 1128, "lr": 7.002246885963531e-05}
{"train_loss": 0.2120470404624939, "global_step": 100441, "epoch": 1128, "lr": 7.002193763469229e-05}
{"train_loss": 0.14290571212768555, "global_step": 100442, "epoch": 1128, "lr": 7.002140640705758e-05}
{"train_loss": 0.1910754144191742, "global_step": 100443, "epoch": 1128, "lr": 7.002087517673124e-05}
{"train_loss": 0.16270869970321655, "global_step": 100444, "epoch": 1128, "lr": 7.002034394371332e-05}
{"train_loss": 0.1497398465871811, "global_step": 100445, "epoch": 1128, "lr": 7.001981270800394e-05}
{"train_loss": 0.14073161780834198, "global_step": 100446, "epoch": 1128, "lr": 7.001928146960313e-05}
{"train_loss": 0.23742645978927612, "global_step": 100447, "epoch": 1128, "lr": 7.001875022851095e-05}
{"train_loss": 0.19085922837257385, "global_step": 100448, "epoch": 1128, "lr": 7.00182189847275e-05}
{"train_loss": 0.17130404710769653, "global_step": 100449, "epoch": 1128, "lr": 7.001768773825287e-05}
{"train_loss": 0.11484745889902115, "global_step": 100450, "epoch": 1128, "lr": 7.001715648908709e-05}
{"train_loss": 0.2084198296070099, "global_step": 100451, "epoch": 1128, "lr": 7.001662523723025e-05}
{"train_loss": 0.18228474259376526, "global_step": 100452, "epoch": 1128, "lr": 7.001609398268241e-05}
{"train_loss": 0.20984981954097748, "global_step": 100453, "epoch": 1128, "lr": 7.001556272544365e-05}
{"train_loss": 0.1777449995279312, "global_step": 100454, "epoch": 1128, "lr": 7.001503146551405e-05}
{"train_loss": 0.11481431126594543, "global_step": 100455, "epoch": 1128, "lr": 7.001450020289366e-05}
{"train_loss": 0.21918733417987823, "global_step": 100456, "epoch": 1128, "lr": 7.001396893758259e-05}
{"train_loss": 0.18847857415676117, "global_step": 100457, "epoch": 1128, "lr": 7.001343766958086e-05}
{"train_loss": 0.23494374752044678, "global_step": 100458, "epoch": 1128, "lr": 7.001290639888858e-05}
{"train_loss": 0.1365237534046173, "global_step": 100459, "epoch": 1128, "lr": 7.001237512550579e-05}
{"train_loss": 0.17650294303894043, "global_step": 100460, "epoch": 1128, "lr": 7.00118438494326e-05}
{"train_loss": 0.1454407274723053, "global_step": 100461, "epoch": 1128, "lr": 7.001131257066905e-05}
{"train_loss": 0.12783902883529663, "global_step": 100462, "epoch": 1128, "lr": 7.001078128921524e-05}
{"train_loss": 0.13794384896755219, "global_step": 100463, "epoch": 1128, "lr": 7.001025000507121e-05}
{"train_loss": 0.20970995724201202, "global_step": 100464, "epoch": 1128, "lr": 7.000971871823705e-05}
{"train_loss": 0.2102741003036499, "global_step": 100465, "epoch": 1128, "lr": 7.000918742871284e-05}
{"train_loss": 0.20603930950164795, "global_step": 100466, "epoch": 1128, "lr": 7.000865613649862e-05}
{"train_loss": 0.13040584325790405, "global_step": 100467, "epoch": 1128, "lr": 7.000812484159448e-05}
{"train_loss": 0.1596141904592514, "global_step": 100468, "epoch": 1128, "lr": 7.000759354400051e-05}
{"train_loss": 0.15396277606487274, "global_step": 100469, "epoch": 1128, "lr": 7.000706224371674e-05}
{"train_loss": 0.09140084683895111, "global_step": 100470, "epoch": 1128, "lr": 7.000653094074328e-05}
{"train_loss": 0.14702188968658447, "global_step": 100471, "epoch": 1128, "lr": 7.000599963508017e-05}
{"train_loss": 0.11195719987154007, "global_step": 100472, "epoch": 1128, "lr": 7.000546832672751e-05}
{"train_loss": 0.12119731307029724, "global_step": 100473, "epoch": 1128, "lr": 7.000493701568536e-05}
{"train_loss": 0.17611558735370636, "global_step": 100474, "epoch": 1128, "lr": 7.000440570195379e-05}
{"train_loss": 0.12107905000448227, "global_step": 100475, "epoch": 1128, "lr": 7.000387438553286e-05}
{"train_loss": 0.12810839712619781, "global_step": 100476, "epoch": 1128, "lr": 7.000334306642267e-05}
{"train_loss": 0.1314580887556076, "global_step": 100477, "epoch": 1128, "lr": 7.000281174462325e-05}
{"train_loss": 0.13649019598960876, "global_step": 100478, "epoch": 1128, "lr": 7.00022804201347e-05}
{"train_loss": 0.18077415227890015, "global_step": 100479, "epoch": 1128, "lr": 7.000174909295711e-05}
{"train_loss": 0.16714437624041953, "global_step": 100480, "epoch": 1128, "lr": 7.000121776309051e-05, "val_loss": 4.641117572784424}
{"train_loss": 0.15456019341945648, "global_step": 100481, "epoch": 1129, "lr": 7.0000686430535e-05}
{"train_loss": 0.17456847429275513, "global_step": 100482, "epoch": 1129, "lr": 7.000015509529064e-05}
{"train_loss": 0.16499875485897064, "global_step": 100483, "epoch": 1129, "lr": 6.99996237573575e-05}
{"train_loss": 0.09972317516803741, "global_step": 100484, "epoch": 1129, "lr": 6.999909241673566e-05}
{"train_loss": 0.16595177352428436, "global_step": 100485, "epoch": 1129, "lr": 6.999856107342518e-05}
{"train_loss": 0.20873689651489258, "global_step": 100486, "epoch": 1129, "lr": 6.999802972742614e-05}
{"train_loss": 0.1659102886915207, "global_step": 100487, "epoch": 1129, "lr": 6.999749837873862e-05}
{"train_loss": 0.16911283135414124, "global_step": 100488, "epoch": 1129, "lr": 6.999696702736264e-05}
{"train_loss": 0.14832079410552979, "global_step": 100489, "epoch": 1129, "lr": 6.999643567329835e-05}
{"train_loss": 0.10259434580802917, "global_step": 100490, "epoch": 1129, "lr": 6.999590431654577e-05}
{"train_loss": 0.15241460502147675, "global_step": 100491, "epoch": 1129, "lr": 6.999537295710499e-05}
{"train_loss": 0.09596927464008331, "global_step": 100492, "epoch": 1129, "lr": 6.999484159497606e-05}
{"train_loss": 0.11841708421707153, "global_step": 100493, "epoch": 1129, "lr": 6.999431023015909e-05}
{"train_loss": 0.1480656862258911, "global_step": 100494, "epoch": 1129, "lr": 6.999377886265412e-05}
{"train_loss": 0.17124958336353302, "global_step": 100495, "epoch": 1129, "lr": 6.999324749246122e-05}
{"train_loss": 0.18942518532276154, "global_step": 100496, "epoch": 1129, "lr": 6.999271611958047e-05}
{"train_loss": 0.15307103097438812, "global_step": 100497, "epoch": 1129, "lr": 6.999218474401196e-05}
{"train_loss": 0.14939109981060028, "global_step": 100498, "epoch": 1129, "lr": 6.999165336575573e-05}
{"train_loss": 0.16634950041770935, "global_step": 100499, "epoch": 1129, "lr": 6.999112198481187e-05}
{"train_loss": 0.12083498388528824, "global_step": 100500, "epoch": 1129, "lr": 6.999059060118045e-05}
{"train_loss": 0.15039432048797607, "global_step": 100501, "epoch": 1129, "lr": 6.999005921486154e-05}
{"train_loss": 0.11112494021654129, "global_step": 100502, "epoch": 1129, "lr": 6.99895278258552e-05}
{"train_loss": 0.1652868539094925, "global_step": 100503, "epoch": 1129, "lr": 6.998899643416155e-05}
{"train_loss": 0.11715275794267654, "global_step": 100504, "epoch": 1129, "lr": 6.998846503978058e-05}
{"train_loss": 0.1299474686384201, "global_step": 100505, "epoch": 1129, "lr": 6.998793364271242e-05}
{"train_loss": 0.11749009042978287, "global_step": 100506, "epoch": 1129, "lr": 6.998740224295713e-05}
{"train_loss": 0.07835707068443298, "global_step": 100507, "epoch": 1129, "lr": 6.998687084051476e-05}
{"train_loss": 0.14192907512187958, "global_step": 100508, "epoch": 1129, "lr": 6.998633943538542e-05}
{"train_loss": 0.22426295280456543, "global_step": 100509, "epoch": 1129, "lr": 6.998580802756915e-05}
{"train_loss": 0.13620416820049286, "global_step": 100510, "epoch": 1129, "lr": 6.998527661706603e-05}
{"train_loss": 0.18162192404270172, "global_step": 100511, "epoch": 1129, "lr": 6.998474520387613e-05}
{"train_loss": 0.12025731801986694, "global_step": 100512, "epoch": 1129, "lr": 6.998421378799954e-05}
{"train_loss": 0.12143898010253906, "global_step": 100513, "epoch": 1129, "lr": 6.99836823694363e-05}
{"train_loss": 0.1436842530965805, "global_step": 100514, "epoch": 1129, "lr": 6.99831509481865e-05}
{"train_loss": 0.1339404433965683, "global_step": 100515, "epoch": 1129, "lr": 6.99826195242502e-05}
{"train_loss": 0.12286859005689621, "global_step": 100516, "epoch": 1129, "lr": 6.99820880976275e-05}
{"train_loss": 0.1378334015607834, "global_step": 100517, "epoch": 1129, "lr": 6.998155666831844e-05}
{"train_loss": 0.21251049637794495, "global_step": 100518, "epoch": 1129, "lr": 6.99810252363231e-05}
{"train_loss": 0.15215735137462616, "global_step": 100519, "epoch": 1129, "lr": 6.998049380164156e-05}
{"train_loss": 0.12541192770004272, "global_step": 100520, "epoch": 1129, "lr": 6.99799623642739e-05}
{"train_loss": 0.1420496553182602, "global_step": 100521, "epoch": 1129, "lr": 6.997943092422016e-05}
{"train_loss": 0.12498719245195389, "global_step": 100522, "epoch": 1129, "lr": 6.997889948148044e-05}
{"train_loss": 0.09253600239753723, "global_step": 100523, "epoch": 1129, "lr": 6.99783680360548e-05}
{"train_loss": 0.13701583445072174, "global_step": 100524, "epoch": 1129, "lr": 6.997783658794331e-05}
{"train_loss": 0.10598503798246384, "global_step": 100525, "epoch": 1129, "lr": 6.997730513714604e-05}
{"train_loss": 0.1039109081029892, "global_step": 100526, "epoch": 1129, "lr": 6.997677368366307e-05}
{"train_loss": 0.14383156597614288, "global_step": 100527, "epoch": 1129, "lr": 6.997624222749446e-05}
{"train_loss": 0.12880784273147583, "global_step": 100528, "epoch": 1129, "lr": 6.99757107686403e-05}
{"train_loss": 0.1657392382621765, "global_step": 100529, "epoch": 1129, "lr": 6.997517930710063e-05}
{"train_loss": 0.14836262166500092, "global_step": 100530, "epoch": 1129, "lr": 6.997464784287555e-05}
{"train_loss": 0.18761730194091797, "global_step": 100531, "epoch": 1129, "lr": 6.997411637596512e-05}
{"train_loss": 0.136421337723732, "global_step": 100532, "epoch": 1129, "lr": 6.997358490636941e-05}
{"train_loss": 0.10697513073682785, "global_step": 100533, "epoch": 1129, "lr": 6.997305343408848e-05}
{"train_loss": 0.21061603724956512, "global_step": 100534, "epoch": 1129, "lr": 6.997252195912245e-05}
{"train_loss": 0.1025596335530281, "global_step": 100535, "epoch": 1129, "lr": 6.997199048147133e-05}
{"train_loss": 0.20033536851406097, "global_step": 100536, "epoch": 1129, "lr": 6.997145900113525e-05}
{"train_loss": 0.1930909901857376, "global_step": 100537, "epoch": 1129, "lr": 6.997092751811423e-05}
{"train_loss": 0.15509049594402313, "global_step": 100538, "epoch": 1129, "lr": 6.997039603240836e-05}
{"train_loss": 0.15173862874507904, "global_step": 100539, "epoch": 1129, "lr": 6.996986454401772e-05}
{"train_loss": 0.14206767082214355, "global_step": 100540, "epoch": 1129, "lr": 6.996933305294238e-05}
{"train_loss": 0.14037024974822998, "global_step": 100541, "epoch": 1129, "lr": 6.99688015591824e-05}
{"train_loss": 0.20784862339496613, "global_step": 100542, "epoch": 1129, "lr": 6.996827006273787e-05}
{"train_loss": 0.23508352041244507, "global_step": 100543, "epoch": 1129, "lr": 6.996773856360885e-05}
{"train_loss": 0.167579784989357, "global_step": 100544, "epoch": 1129, "lr": 6.996720706179539e-05}
{"train_loss": 0.13457533717155457, "global_step": 100545, "epoch": 1129, "lr": 6.996667555729759e-05}
{"train_loss": 0.13462653756141663, "global_step": 100546, "epoch": 1129, "lr": 6.996614405011552e-05}
{"train_loss": 0.17399367690086365, "global_step": 100547, "epoch": 1129, "lr": 6.996561254024923e-05}
{"train_loss": 0.12425351142883301, "global_step": 100548, "epoch": 1129, "lr": 6.996508102769883e-05}
{"train_loss": 0.14612628519535065, "global_step": 100549, "epoch": 1129, "lr": 6.996454951246435e-05}
{"train_loss": 0.24069879949092865, "global_step": 100550, "epoch": 1129, "lr": 6.996401799454589e-05}
{"train_loss": 0.08082509785890579, "global_step": 100551, "epoch": 1129, "lr": 6.996348647394349e-05}
{"train_loss": 0.0949440449476242, "global_step": 100552, "epoch": 1129, "lr": 6.996295495065727e-05}
{"train_loss": 0.16206607222557068, "global_step": 100553, "epoch": 1129, "lr": 6.996242342468726e-05}
{"train_loss": 0.1701633781194687, "global_step": 100554, "epoch": 1129, "lr": 6.996189189603355e-05}
{"train_loss": 0.1688017100095749, "global_step": 100555, "epoch": 1129, "lr": 6.99613603646962e-05}
{"train_loss": 0.11343929916620255, "global_step": 100556, "epoch": 1129, "lr": 6.996082883067529e-05}
{"train_loss": 0.10019522905349731, "global_step": 100557, "epoch": 1129, "lr": 6.99602972939709e-05}
{"train_loss": 0.17105865478515625, "global_step": 100558, "epoch": 1129, "lr": 6.99597657545831e-05}
{"train_loss": 0.13367366790771484, "global_step": 100559, "epoch": 1129, "lr": 6.995923421251193e-05}
{"train_loss": 0.15397761762142181, "global_step": 100560, "epoch": 1129, "lr": 6.995870266775749e-05}
{"train_loss": 0.22120064496994019, "global_step": 100561, "epoch": 1129, "lr": 6.995817112031986e-05}
{"train_loss": 0.1072038933634758, "global_step": 100562, "epoch": 1129, "lr": 6.995763957019908e-05}
{"train_loss": 0.18139053881168365, "global_step": 100563, "epoch": 1129, "lr": 6.995710801739525e-05}
{"train_loss": 0.1531309187412262, "global_step": 100564, "epoch": 1129, "lr": 6.995657646190842e-05}
{"train_loss": 0.15624624490737915, "global_step": 100565, "epoch": 1129, "lr": 6.995604490373868e-05}
{"train_loss": 0.18933121860027313, "global_step": 100566, "epoch": 1129, "lr": 6.995551334288608e-05}
{"train_loss": 0.14665229618549347, "global_step": 100567, "epoch": 1129, "lr": 6.995498177935072e-05}
{"train_loss": 0.180242121219635, "global_step": 100568, "epoch": 1129, "lr": 6.995445021313265e-05}
{"train_loss": 0.1499611179480392, "global_step": 100569, "epoch": 1129, "lr": 6.995391864423193e-05, "val_loss": 4.848191261291504}
{"train_loss": 0.10383230447769165, "global_step": 100570, "epoch": 1130, "lr": 6.995338707264867e-05}
{"train_loss": 0.18094484508037567, "global_step": 100571, "epoch": 1130, "lr": 6.995285549838292e-05}
{"train_loss": 0.12303969264030457, "global_step": 100572, "epoch": 1130, "lr": 6.995232392143474e-05}
{"train_loss": 0.13835972547531128, "global_step": 100573, "epoch": 1130, "lr": 6.995179234180422e-05}
{"train_loss": 0.2507636249065399, "global_step": 100574, "epoch": 1130, "lr": 6.995126075949143e-05}
{"train_loss": 0.16138634085655212, "global_step": 100575, "epoch": 1130, "lr": 6.995072917449644e-05}
{"train_loss": 0.18846522271633148, "global_step": 100576, "epoch": 1130, "lr": 6.99501975868193e-05}
{"train_loss": 0.12044638395309448, "global_step": 100577, "epoch": 1130, "lr": 6.99496659964601e-05}
{"train_loss": 0.15135592222213745, "global_step": 100578, "epoch": 1130, "lr": 6.994913440341893e-05}
{"train_loss": 0.16575156152248383, "global_step": 100579, "epoch": 1130, "lr": 6.994860280769582e-05}
{"train_loss": 0.13337652385234833, "global_step": 100580, "epoch": 1130, "lr": 6.99480712092909e-05}
{"train_loss": 0.18426138162612915, "global_step": 100581, "epoch": 1130, "lr": 6.994753960820418e-05}
{"train_loss": 0.1535145491361618, "global_step": 100582, "epoch": 1130, "lr": 6.994700800443574e-05}
{"train_loss": 0.10371650755405426, "global_step": 100583, "epoch": 1130, "lr": 6.99464763979857e-05}
{"train_loss": 0.16191115975379944, "global_step": 100584, "epoch": 1130, "lr": 6.994594478885408e-05}
{"train_loss": 0.25181612372398376, "global_step": 100585, "epoch": 1130, "lr": 6.994541317704097e-05}
{"train_loss": 0.1574832648038864, "global_step": 100586, "epoch": 1130, "lr": 6.994488156254644e-05}
{"train_loss": 0.13354799151420593, "global_step": 100587, "epoch": 1130, "lr": 6.994434994537055e-05}
{"train_loss": 0.14681300520896912, "global_step": 100588, "epoch": 1130, "lr": 6.994381832551342e-05}
{"train_loss": 0.13212977349758148, "global_step": 100589, "epoch": 1130, "lr": 6.994328670297507e-05}
{"train_loss": 0.12082822620868683, "global_step": 100590, "epoch": 1130, "lr": 6.994275507775559e-05}
{"train_loss": 0.14130976796150208, "global_step": 100591, "epoch": 1130, "lr": 6.994222344985504e-05}
{"train_loss": 0.19482079148292542, "global_step": 100592, "epoch": 1130, "lr": 6.994169181927353e-05}
{"train_loss": 0.16980183124542236, "global_step": 100593, "epoch": 1130, "lr": 6.994116018601106e-05}
{"train_loss": 0.17748762667179108, "global_step": 100594, "epoch": 1130, "lr": 6.994062855006777e-05}
{"train_loss": 0.11745560169219971, "global_step": 100595, "epoch": 1130, "lr": 6.994009691144373e-05}
{"train_loss": 0.11880820244550705, "global_step": 100596, "epoch": 1130, "lr": 6.993956527013893e-05}
{"train_loss": 0.14380431175231934, "global_step": 100597, "epoch": 1130, "lr": 6.993903362615353e-05}
{"train_loss": 0.12695452570915222, "global_step": 100598, "epoch": 1130, "lr": 6.993850197948758e-05}
{"train_loss": 0.11943072080612183, "global_step": 100599, "epoch": 1130, "lr": 6.993797033014114e-05}
{"train_loss": 0.18119004368782043, "global_step": 100600, "epoch": 1130, "lr": 6.993743867811428e-05}
{"train_loss": 0.13384976983070374, "global_step": 100601, "epoch": 1130, "lr": 6.993690702340705e-05}
{"train_loss": 0.09666392207145691, "global_step": 100602, "epoch": 1130, "lr": 6.993637536601957e-05}
{"train_loss": 0.2702258229255676, "global_step": 100603, "epoch": 1130, "lr": 6.993584370595188e-05}
{"train_loss": 0.2035515457391739, "global_step": 100604, "epoch": 1130, "lr": 6.993531204320405e-05}
{"train_loss": 0.1366073489189148, "global_step": 100605, "epoch": 1130, "lr": 6.993478037777616e-05}
{"train_loss": 0.2173614352941513, "global_step": 100606, "epoch": 1130, "lr": 6.993424870966831e-05}
{"train_loss": 0.10468143969774246, "global_step": 100607, "epoch": 1130, "lr": 6.993371703888052e-05}
{"train_loss": 0.1795901209115982, "global_step": 100608, "epoch": 1130, "lr": 6.993318536541289e-05}
{"train_loss": 0.2054753303527832, "global_step": 100609, "epoch": 1130, "lr": 6.993265368926549e-05}
{"train_loss": 0.15815040469169617, "global_step": 100610, "epoch": 1130, "lr": 6.993212201043837e-05}
{"train_loss": 0.15294355154037476, "global_step": 100611, "epoch": 1130, "lr": 6.993159032893164e-05}
{"train_loss": 0.16317705810070038, "global_step": 100612, "epoch": 1130, "lr": 6.993105864474534e-05}
{"train_loss": 0.14229080080986023, "global_step": 100613, "epoch": 1130, "lr": 6.993052695787954e-05}
{"train_loss": 0.17081183195114136, "global_step": 100614, "epoch": 1130, "lr": 6.992999526833434e-05}
{"train_loss": 0.14744672179222107, "global_step": 100615, "epoch": 1130, "lr": 6.99294635761098e-05}
{"train_loss": 0.24111226201057434, "global_step": 100616, "epoch": 1130, "lr": 6.992893188120599e-05}
{"train_loss": 0.1280827671289444, "global_step": 100617, "epoch": 1130, "lr": 6.992840018362296e-05}
{"train_loss": 0.13585104048252106, "global_step": 100618, "epoch": 1130, "lr": 6.99278684833608e-05}
{"train_loss": 0.12298908084630966, "global_step": 100619, "epoch": 1130, "lr": 6.99273367804196e-05}
{"train_loss": 0.1727701872587204, "global_step": 100620, "epoch": 1130, "lr": 6.992680507479941e-05}
{"train_loss": 0.13647732138633728, "global_step": 100621, "epoch": 1130, "lr": 6.992627336650029e-05}
{"train_loss": 0.16960549354553223, "global_step": 100622, "epoch": 1130, "lr": 6.992574165552232e-05}
{"train_loss": 0.19132940471172333, "global_step": 100623, "epoch": 1130, "lr": 6.992520994186559e-05}
{"train_loss": 0.1807904988527298, "global_step": 100624, "epoch": 1130, "lr": 6.992467822553014e-05}
{"train_loss": 0.20531004667282104, "global_step": 100625, "epoch": 1130, "lr": 6.992414650651608e-05}
{"train_loss": 0.13756690919399261, "global_step": 100626, "epoch": 1130, "lr": 6.992361478482346e-05}
{"train_loss": 0.07630573213100433, "global_step": 100627, "epoch": 1130, "lr": 6.992308306045236e-05}
{"train_loss": 0.09436294436454773, "global_step": 100628, "epoch": 1130, "lr": 6.992255133340283e-05}
{"train_loss": 0.1039905920624733, "global_step": 100629, "epoch": 1130, "lr": 6.992201960367496e-05}
{"train_loss": 0.2781487703323364, "global_step": 100630, "epoch": 1130, "lr": 6.992148787126883e-05}
{"train_loss": 0.23468725383281708, "global_step": 100631, "epoch": 1130, "lr": 6.992095613618446e-05}
{"train_loss": 0.11107927560806274, "global_step": 100632, "epoch": 1130, "lr": 6.992042439842199e-05}
{"train_loss": 0.212538480758667, "global_step": 100633, "epoch": 1130, "lr": 6.991989265798146e-05}
{"train_loss": 0.17481575906276703, "global_step": 100634, "epoch": 1130, "lr": 6.991936091486294e-05}
{"train_loss": 0.16107657551765442, "global_step": 100635, "epoch": 1130, "lr": 6.991882916906651e-05}
{"train_loss": 0.20831403136253357, "global_step": 100636, "epoch": 1130, "lr": 6.991829742059223e-05}
{"train_loss": 0.09069370478391647, "global_step": 100637, "epoch": 1130, "lr": 6.991776566944018e-05}
{"train_loss": 0.19797439873218536, "global_step": 100638, "epoch": 1130, "lr": 6.991723391561043e-05}
{"train_loss": 0.19002005457878113, "global_step": 100639, "epoch": 1130, "lr": 6.991670215910305e-05}
{"train_loss": 0.2304006665945053, "global_step": 100640, "epoch": 1130, "lr": 6.99161703999181e-05}
{"train_loss": 0.2389337420463562, "global_step": 100641, "epoch": 1130, "lr": 6.991563863805568e-05}
{"train_loss": 0.24628426134586334, "global_step": 100642, "epoch": 1130, "lr": 6.991510687351584e-05}
{"train_loss": 0.1750452071428299, "global_step": 100643, "epoch": 1130, "lr": 6.991457510629867e-05}
{"train_loss": 0.20471438765525818, "global_step": 100644, "epoch": 1130, "lr": 6.99140433364042e-05}
{"train_loss": 0.21054184436798096, "global_step": 100645, "epoch": 1130, "lr": 6.991351156383255e-05}
{"train_loss": 0.26785343885421753, "global_step": 100646, "epoch": 1130, "lr": 6.991297978858377e-05}
{"train_loss": 0.18414825201034546, "global_step": 100647, "epoch": 1130, "lr": 6.991244801065793e-05}
{"train_loss": 0.19622258841991425, "global_step": 100648, "epoch": 1130, "lr": 6.99119162300551e-05}
{"train_loss": 0.25430452823638916, "global_step": 100649, "epoch": 1130, "lr": 6.991138444677535e-05}
{"train_loss": 0.10010019689798355, "global_step": 100650, "epoch": 1130, "lr": 6.991085266081875e-05}
{"train_loss": 0.20845679938793182, "global_step": 100651, "epoch": 1130, "lr": 6.99103208721854e-05}
{"train_loss": 0.14513495564460754, "global_step": 100652, "epoch": 1130, "lr": 6.990978908087536e-05}
{"train_loss": 0.17093637585639954, "global_step": 100653, "epoch": 1130, "lr": 6.990925728688866e-05}
{"train_loss": 0.2203052043914795, "global_step": 100654, "epoch": 1130, "lr": 6.990872549022543e-05}
{"train_loss": 0.14628998935222626, "global_step": 100655, "epoch": 1130, "lr": 6.990819369088569e-05}
{"train_loss": 0.1126670241355896, "global_step": 100656, "epoch": 1130, "lr": 6.990766188886955e-05}
{"train_loss": 0.19722388684749603, "global_step": 100657, "epoch": 1130, "lr": 6.990713008417707e-05}
{"train_loss": 0.16719012883272064, "global_step": 100658, "epoch": 1130, "lr": 6.990659827680832e-05, "val_loss": 4.676231384277344, "train_action_mse_error": 27.65228271484375}
{"train_loss": 0.1423887461423874, "global_step": 100659, "epoch": 1131, "lr": 6.990606646676336e-05}
{"train_loss": 0.15164202451705933, "global_step": 100660, "epoch": 1131, "lr": 6.990553465404229e-05}
{"train_loss": 0.25363945960998535, "global_step": 100661, "epoch": 1131, "lr": 6.990500283864513e-05}
{"train_loss": 0.187392920255661, "global_step": 100662, "epoch": 1131, "lr": 6.990447102057202e-05}
{"train_loss": 0.18842196464538574, "global_step": 100663, "epoch": 1131, "lr": 6.990393919982297e-05}
{"train_loss": 0.24816127121448517, "global_step": 100664, "epoch": 1131, "lr": 6.99034073763981e-05}
{"train_loss": 0.2656351923942566, "global_step": 100665, "epoch": 1131, "lr": 6.990287555029745e-05}
{"train_loss": 0.13657739758491516, "global_step": 100666, "epoch": 1131, "lr": 6.990234372152109e-05}
{"train_loss": 0.1798306107521057, "global_step": 100667, "epoch": 1131, "lr": 6.990181189006912e-05}
{"train_loss": 0.13203364610671997, "global_step": 100668, "epoch": 1131, "lr": 6.990128005594159e-05}
{"train_loss": 0.1383725106716156, "global_step": 100669, "epoch": 1131, "lr": 6.990074821913856e-05}
{"train_loss": 0.1545266956090927, "global_step": 100670, "epoch": 1131, "lr": 6.990021637966014e-05}
{"train_loss": 0.1916990876197815, "global_step": 100671, "epoch": 1131, "lr": 6.989968453750637e-05}
{"train_loss": 0.12149563431739807, "global_step": 100672, "epoch": 1131, "lr": 6.989915269267733e-05}
{"train_loss": 0.11804210394620895, "global_step": 100673, "epoch": 1131, "lr": 6.989862084517308e-05}
{"train_loss": 0.1900319904088974, "global_step": 100674, "epoch": 1131, "lr": 6.989808899499372e-05}
{"train_loss": 0.18546366691589355, "global_step": 100675, "epoch": 1131, "lr": 6.989755714213929e-05}
{"train_loss": 0.16021417081356049, "global_step": 100676, "epoch": 1131, "lr": 6.989702528660988e-05}
{"train_loss": 0.22264966368675232, "global_step": 100677, "epoch": 1131, "lr": 6.989649342840557e-05}
{"train_loss": 0.18014858663082123, "global_step": 100678, "epoch": 1131, "lr": 6.98959615675264e-05}
{"train_loss": 0.16224941611289978, "global_step": 100679, "epoch": 1131, "lr": 6.989542970397247e-05}
{"train_loss": 0.21564316749572754, "global_step": 100680, "epoch": 1131, "lr": 6.989489783774384e-05}
{"train_loss": 0.13195474445819855, "global_step": 100681, "epoch": 1131, "lr": 6.989436596884059e-05}
{"train_loss": 0.14828325808048248, "global_step": 100682, "epoch": 1131, "lr": 6.989383409726278e-05}
{"train_loss": 0.18741203844547272, "global_step": 100683, "epoch": 1131, "lr": 6.98933022230105e-05}
{"train_loss": 0.14864042401313782, "global_step": 100684, "epoch": 1131, "lr": 6.989277034608379e-05}
{"train_loss": 0.1999056339263916, "global_step": 100685, "epoch": 1131, "lr": 6.989223846648274e-05}
{"train_loss": 0.22482016682624817, "global_step": 100686, "epoch": 1131, "lr": 6.989170658420742e-05}
{"train_loss": 0.20267538726329803, "global_step": 100687, "epoch": 1131, "lr": 6.98911746992579e-05}
{"train_loss": 0.17495208978652954, "global_step": 100688, "epoch": 1131, "lr": 6.989064281163427e-05}
{"train_loss": 0.3244101405143738, "global_step": 100689, "epoch": 1131, "lr": 6.989011092133657e-05}
{"train_loss": 0.09258662164211273, "global_step": 100690, "epoch": 1131, "lr": 6.988957902836491e-05}
{"train_loss": 0.12475030869245529, "global_step": 100691, "epoch": 1131, "lr": 6.988904713271931e-05}
{"train_loss": 0.1447320133447647, "global_step": 100692, "epoch": 1131, "lr": 6.988851523439988e-05}
{"train_loss": 0.13764846324920654, "global_step": 100693, "epoch": 1131, "lr": 6.988798333340669e-05}
{"train_loss": 0.19039444625377655, "global_step": 100694, "epoch": 1131, "lr": 6.98874514297398e-05}
{"train_loss": 0.29886654019355774, "global_step": 100695, "epoch": 1131, "lr": 6.988691952339928e-05}
{"train_loss": 0.13497072458267212, "global_step": 100696, "epoch": 1131, "lr": 6.988638761438521e-05}
{"train_loss": 0.14124393463134766, "global_step": 100697, "epoch": 1131, "lr": 6.988585570269765e-05}
{"train_loss": 0.17709162831306458, "global_step": 100698, "epoch": 1131, "lr": 6.988532378833668e-05}
{"train_loss": 0.11020241677761078, "global_step": 100699, "epoch": 1131, "lr": 6.988479187130237e-05}
{"train_loss": 0.20296402275562286, "global_step": 100700, "epoch": 1131, "lr": 6.988425995159479e-05}
{"train_loss": 0.14878550171852112, "global_step": 100701, "epoch": 1131, "lr": 6.988372802921403e-05}
{"train_loss": 0.10070047527551651, "global_step": 100702, "epoch": 1131, "lr": 6.988319610416014e-05}
{"train_loss": 0.16563370823860168, "global_step": 100703, "epoch": 1131, "lr": 6.988266417643319e-05}
{"train_loss": 0.0908646285533905, "global_step": 100704, "epoch": 1131, "lr": 6.988213224603326e-05}
{"train_loss": 0.21781562268733978, "global_step": 100705, "epoch": 1131, "lr": 6.988160031296042e-05}
{"train_loss": 0.1337735652923584, "global_step": 100706, "epoch": 1131, "lr": 6.988106837721474e-05}
{"train_loss": 0.196952685713768, "global_step": 100707, "epoch": 1131, "lr": 6.98805364387963e-05}
{"train_loss": 0.14220011234283447, "global_step": 100708, "epoch": 1131, "lr": 6.988000449770515e-05}
{"train_loss": 0.21206365525722504, "global_step": 100709, "epoch": 1131, "lr": 6.98794725539414e-05}
{"train_loss": 0.16311001777648926, "global_step": 100710, "epoch": 1131, "lr": 6.987894060750508e-05}
{"train_loss": 0.14973624050617218, "global_step": 100711, "epoch": 1131, "lr": 6.987840865839628e-05}
{"train_loss": 0.12855182588100433, "global_step": 100712, "epoch": 1131, "lr": 6.987787670661507e-05}
{"train_loss": 0.1256176382303238, "global_step": 100713, "epoch": 1131, "lr": 6.98773447521615e-05}
{"train_loss": 0.2471446543931961, "global_step": 100714, "epoch": 1131, "lr": 6.98768127950357e-05}
{"train_loss": 0.1946173906326294, "global_step": 100715, "epoch": 1131, "lr": 6.987628083523768e-05}
{"train_loss": 0.1583646684885025, "global_step": 100716, "epoch": 1131, "lr": 6.987574887276755e-05}
{"train_loss": 0.16535961627960205, "global_step": 100717, "epoch": 1131, "lr": 6.987521690762537e-05}
{"train_loss": 0.10841775685548782, "global_step": 100718, "epoch": 1131, "lr": 6.98746849398112e-05}
{"train_loss": 0.14231091737747192, "global_step": 100719, "epoch": 1131, "lr": 6.987415296932513e-05}
{"train_loss": 0.1382128745317459, "global_step": 100720, "epoch": 1131, "lr": 6.987362099616721e-05}
{"train_loss": 0.14047107100486755, "global_step": 100721, "epoch": 1131, "lr": 6.987308902033753e-05}
{"train_loss": 0.16006603837013245, "global_step": 100722, "epoch": 1131, "lr": 6.987255704183614e-05}
{"train_loss": 0.13245059549808502, "global_step": 100723, "epoch": 1131, "lr": 6.987202506066316e-05}
{"train_loss": 0.1490907222032547, "global_step": 100724, "epoch": 1131, "lr": 6.987149307681861e-05}
{"train_loss": 0.21971750259399414, "global_step": 100725, "epoch": 1131, "lr": 6.987096109030258e-05}
{"train_loss": 0.19964076578617096, "global_step": 100726, "epoch": 1131, "lr": 6.987042910111515e-05}
{"train_loss": 0.1947885900735855, "global_step": 100727, "epoch": 1131, "lr": 6.986989710925637e-05}
{"train_loss": 0.11495005339384079, "global_step": 100728, "epoch": 1131, "lr": 6.986936511472633e-05}
{"train_loss": 0.1443650871515274, "global_step": 100729, "epoch": 1131, "lr": 6.98688331175251e-05}
{"train_loss": 0.10907933115959167, "global_step": 100730, "epoch": 1131, "lr": 6.986830111765275e-05}
{"train_loss": 0.1003015860915184, "global_step": 100731, "epoch": 1131, "lr": 6.986776911510933e-05}
{"train_loss": 0.13214904069900513, "global_step": 100732, "epoch": 1131, "lr": 6.986723710989495e-05}
{"train_loss": 0.17883339524269104, "global_step": 100733, "epoch": 1131, "lr": 6.986670510200966e-05}
{"train_loss": 0.17506591975688934, "global_step": 100734, "epoch": 1131, "lr": 6.986617309145351e-05}
{"train_loss": 0.13375991582870483, "global_step": 100735, "epoch": 1131, "lr": 6.986564107822662e-05}
{"train_loss": 0.1910870373249054, "global_step": 100736, "epoch": 1131, "lr": 6.986510906232904e-05}
{"train_loss": 0.21537292003631592, "global_step": 100737, "epoch": 1131, "lr": 6.986457704376083e-05}
{"train_loss": 0.1724659949541092, "global_step": 100738, "epoch": 1131, "lr": 6.986404502252207e-05}
{"train_loss": 0.1643308401107788, "global_step": 100739, "epoch": 1131, "lr": 6.986351299861282e-05}
{"train_loss": 0.13668595254421234, "global_step": 100740, "epoch": 1131, "lr": 6.986298097203318e-05}
{"train_loss": 0.1473747342824936, "global_step": 100741, "epoch": 1131, "lr": 6.98624489427832e-05}
{"train_loss": 0.2170417457818985, "global_step": 100742, "epoch": 1131, "lr": 6.986191691086295e-05}
{"train_loss": 0.1895340532064438, "global_step": 100743, "epoch": 1131, "lr": 6.986138487627251e-05}
{"train_loss": 0.15580658614635468, "global_step": 100744, "epoch": 1131, "lr": 6.986085283901195e-05}
{"train_loss": 0.1385723054409027, "global_step": 100745, "epoch": 1131, "lr": 6.986032079908134e-05}
{"train_loss": 0.12299464643001556, "global_step": 100746, "epoch": 1131, "lr": 6.985978875648075e-05}
{"train_loss": 0.16659942872068856, "global_step": 100747, "epoch": 1131, "lr": 6.985925671121026e-05, "val_loss": 4.511877536773682}
{"train_loss": 0.14850418269634247, "global_step": 100748, "epoch": 1132, "lr": 6.985872466326992e-05}
{"train_loss": 0.10373640805482864, "global_step": 100749, "epoch": 1132, "lr": 6.985819261265982e-05}
{"train_loss": 0.2015525996685028, "global_step": 100750, "epoch": 1132, "lr": 6.985766055938004e-05}
{"train_loss": 0.24090790748596191, "global_step": 100751, "epoch": 1132, "lr": 6.985712850343064e-05}
{"train_loss": 0.12651517987251282, "global_step": 100752, "epoch": 1132, "lr": 6.985659644481169e-05}
{"train_loss": 0.17771510779857635, "global_step": 100753, "epoch": 1132, "lr": 6.985606438352324e-05}
{"train_loss": 0.15940280258655548, "global_step": 100754, "epoch": 1132, "lr": 6.985553231956541e-05}
{"train_loss": 0.27067863941192627, "global_step": 100755, "epoch": 1132, "lr": 6.985500025293824e-05}
{"train_loss": 0.15042714774608612, "global_step": 100756, "epoch": 1132, "lr": 6.985446818364181e-05}
{"train_loss": 0.17917108535766602, "global_step": 100757, "epoch": 1132, "lr": 6.985393611167618e-05}
{"train_loss": 0.1811501383781433, "global_step": 100758, "epoch": 1132, "lr": 6.985340403704144e-05}
{"train_loss": 0.1416863650083542, "global_step": 100759, "epoch": 1132, "lr": 6.985287195973764e-05}
{"train_loss": 0.21068966388702393, "global_step": 100760, "epoch": 1132, "lr": 6.985233987976487e-05}
{"train_loss": 0.135654479265213, "global_step": 100761, "epoch": 1132, "lr": 6.98518077971232e-05}
{"train_loss": 0.17159877717494965, "global_step": 100762, "epoch": 1132, "lr": 6.985127571181268e-05}
{"train_loss": 0.16041667759418488, "global_step": 100763, "epoch": 1132, "lr": 6.985074362383341e-05}
{"train_loss": 0.15215592086315155, "global_step": 100764, "epoch": 1132, "lr": 6.985021153318545e-05}
{"train_loss": 0.21626603603363037, "global_step": 100765, "epoch": 1132, "lr": 6.984967943986887e-05}
{"train_loss": 0.15465006232261658, "global_step": 100766, "epoch": 1132, "lr": 6.984914734388373e-05}
{"train_loss": 0.18797385692596436, "global_step": 100767, "epoch": 1132, "lr": 6.984861524523013e-05}
{"train_loss": 0.18351863324642181, "global_step": 100768, "epoch": 1132, "lr": 6.984808314390811e-05}
{"train_loss": 0.08398149907588959, "global_step": 100769, "epoch": 1132, "lr": 6.984755103991778e-05}
{"train_loss": 0.2645496129989624, "global_step": 100770, "epoch": 1132, "lr": 6.984701893325917e-05}
{"train_loss": 0.09182890504598618, "global_step": 100771, "epoch": 1132, "lr": 6.984648682393237e-05}
{"train_loss": 0.11440499871969223, "global_step": 100772, "epoch": 1132, "lr": 6.984595471193747e-05}
{"train_loss": 0.1517234593629837, "global_step": 100773, "epoch": 1132, "lr": 6.984542259727451e-05}
{"train_loss": 0.12309832125902176, "global_step": 100774, "epoch": 1132, "lr": 6.984489047994358e-05}
{"train_loss": 0.16243499517440796, "global_step": 100775, "epoch": 1132, "lr": 6.984435835994474e-05}
{"train_loss": 0.13578544557094574, "global_step": 100776, "epoch": 1132, "lr": 6.984382623727807e-05}
{"train_loss": 0.13172376155853271, "global_step": 100777, "epoch": 1132, "lr": 6.984329411194365e-05}
{"train_loss": 0.1178610771894455, "global_step": 100778, "epoch": 1132, "lr": 6.984276198394153e-05}
{"train_loss": 0.19446802139282227, "global_step": 100779, "epoch": 1132, "lr": 6.984222985327179e-05}
{"train_loss": 0.1961469054222107, "global_step": 100780, "epoch": 1132, "lr": 6.98416977199345e-05}
{"train_loss": 0.11366064846515656, "global_step": 100781, "epoch": 1132, "lr": 6.984116558392974e-05}
{"train_loss": 0.14342617988586426, "global_step": 100782, "epoch": 1132, "lr": 6.98406334452576e-05}
{"train_loss": 0.1618051379919052, "global_step": 100783, "epoch": 1132, "lr": 6.98401013039181e-05}
{"train_loss": 0.1767316460609436, "global_step": 100784, "epoch": 1132, "lr": 6.983956915991136e-05}
{"train_loss": 0.1271742284297943, "global_step": 100785, "epoch": 1132, "lr": 6.983903701323741e-05}
{"train_loss": 0.2354627549648285, "global_step": 100786, "epoch": 1132, "lr": 6.983850486389635e-05}
{"train_loss": 0.1789361834526062, "global_step": 100787, "epoch": 1132, "lr": 6.983797271188824e-05}
{"train_loss": 0.1487446129322052, "global_step": 100788, "epoch": 1132, "lr": 6.983744055721319e-05}
{"train_loss": 0.15210998058319092, "global_step": 100789, "epoch": 1132, "lr": 6.983690839987121e-05}
{"train_loss": 0.0780349001288414, "global_step": 100790, "epoch": 1132, "lr": 6.98363762398624e-05}
{"train_loss": 0.17165187001228333, "global_step": 100791, "epoch": 1132, "lr": 6.983584407718683e-05}
{"train_loss": 0.14189918339252472, "global_step": 100792, "epoch": 1132, "lr": 6.983531191184458e-05}
{"train_loss": 0.1092187762260437, "global_step": 100793, "epoch": 1132, "lr": 6.983477974383571e-05}
{"train_loss": 0.1651362180709839, "global_step": 100794, "epoch": 1132, "lr": 6.983424757316029e-05}
{"train_loss": 0.25578054785728455, "global_step": 100795, "epoch": 1132, "lr": 6.983371539981841e-05}
{"train_loss": 0.1501452624797821, "global_step": 100796, "epoch": 1132, "lr": 6.983318322381011e-05}
{"train_loss": 0.18341201543807983, "global_step": 100797, "epoch": 1132, "lr": 6.98326510451355e-05}
{"train_loss": 0.16549283266067505, "global_step": 100798, "epoch": 1132, "lr": 6.983211886379463e-05}
{"train_loss": 0.16996318101882935, "global_step": 100799, "epoch": 1132, "lr": 6.983158667978756e-05}
{"train_loss": 0.14774446189403534, "global_step": 100800, "epoch": 1132, "lr": 6.983105449311438e-05}
{"train_loss": 0.11719131469726562, "global_step": 100801, "epoch": 1132, "lr": 6.983052230377517e-05}
{"train_loss": 0.15929940342903137, "global_step": 100802, "epoch": 1132, "lr": 6.982999011176997e-05}
{"train_loss": 0.20349523425102234, "global_step": 100803, "epoch": 1132, "lr": 6.982945791709888e-05}
{"train_loss": 0.1303231418132782, "global_step": 100804, "epoch": 1132, "lr": 6.982892571976195e-05}
{"train_loss": 0.1908208280801773, "global_step": 100805, "epoch": 1132, "lr": 6.982839351975927e-05}
{"train_loss": 0.1609395295381546, "global_step": 100806, "epoch": 1132, "lr": 6.98278613170909e-05}
{"train_loss": 0.15020273625850677, "global_step": 100807, "epoch": 1132, "lr": 6.982732911175694e-05}
{"train_loss": 0.11784335225820541, "global_step": 100808, "epoch": 1132, "lr": 6.982679690375741e-05}
{"train_loss": 0.14691466093063354, "global_step": 100809, "epoch": 1132, "lr": 6.982626469309241e-05}
{"train_loss": 0.12694302201271057, "global_step": 100810, "epoch": 1132, "lr": 6.982573247976201e-05}
{"train_loss": 0.16379337012767792, "global_step": 100811, "epoch": 1132, "lr": 6.98252002637663e-05}
{"train_loss": 0.1908515989780426, "global_step": 100812, "epoch": 1132, "lr": 6.982466804510532e-05}
{"train_loss": 0.1481092870235443, "global_step": 100813, "epoch": 1132, "lr": 6.982413582377916e-05}
{"train_loss": 0.11798006296157837, "global_step": 100814, "epoch": 1132, "lr": 6.982360359978788e-05}
{"train_loss": 0.1910063922405243, "global_step": 100815, "epoch": 1132, "lr": 6.982307137313157e-05}
{"train_loss": 0.10979243367910385, "global_step": 100816, "epoch": 1132, "lr": 6.982253914381027e-05}
{"train_loss": 0.22159285843372345, "global_step": 100817, "epoch": 1132, "lr": 6.982200691182408e-05}
{"train_loss": 0.2282973676919937, "global_step": 100818, "epoch": 1132, "lr": 6.982147467717308e-05}
{"train_loss": 0.23730668425559998, "global_step": 100819, "epoch": 1132, "lr": 6.98209424398573e-05}
{"train_loss": 0.22809281945228577, "global_step": 100820, "epoch": 1132, "lr": 6.982041019987684e-05}
{"train_loss": 0.10824006795883179, "global_step": 100821, "epoch": 1132, "lr": 6.981987795723177e-05}
{"train_loss": 0.18273751437664032, "global_step": 100822, "epoch": 1132, "lr": 6.981934571192216e-05}
{"train_loss": 0.1775968223810196, "global_step": 100823, "epoch": 1132, "lr": 6.981881346394807e-05}
{"train_loss": 0.10088948160409927, "global_step": 100824, "epoch": 1132, "lr": 6.98182812133096e-05}
{"train_loss": 0.13926145434379578, "global_step": 100825, "epoch": 1132, "lr": 6.98177489600068e-05}
{"train_loss": 0.16426749527454376, "global_step": 100826, "epoch": 1132, "lr": 6.981721670403975e-05}
{"train_loss": 0.15902438759803772, "global_step": 100827, "epoch": 1132, "lr": 6.981668444540852e-05}
{"train_loss": 0.1317506581544876, "global_step": 100828, "epoch": 1132, "lr": 6.981615218411315e-05}
{"train_loss": 0.16131754219532013, "global_step": 100829, "epoch": 1132, "lr": 6.981561992015378e-05}
{"train_loss": 0.15905921161174774, "global_step": 100830, "epoch": 1132, "lr": 6.981508765353041e-05}
{"train_loss": 0.12574906647205353, "global_step": 100831, "epoch": 1132, "lr": 6.981455538424316e-05}
{"train_loss": 0.16820944845676422, "global_step": 100832, "epoch": 1132, "lr": 6.981402311229207e-05}
{"train_loss": 0.1609140932559967, "global_step": 100833, "epoch": 1132, "lr": 6.981349083767723e-05}
{"train_loss": 0.22621648013591766, "global_step": 100834, "epoch": 1132, "lr": 6.981295856039871e-05}
{"train_loss": 0.2421041876077652, "global_step": 100835, "epoch": 1132, "lr": 6.981242628045659e-05}
{"train_loss": 0.16296022192815715, "global_step": 100836, "epoch": 1132, "lr": 6.981189399785091e-05, "val_loss": 4.794915199279785}
{"train_loss": 0.1881399303674698, "global_step": 100837, "epoch": 1133, "lr": 6.981136171258177e-05}
{"train_loss": 0.16078555583953857, "global_step": 100838, "epoch": 1133, "lr": 6.981082942464925e-05}
{"train_loss": 0.18824060261249542, "global_step": 100839, "epoch": 1133, "lr": 6.981029713405338e-05}
{"train_loss": 0.14125332236289978, "global_step": 100840, "epoch": 1133, "lr": 6.980976484079426e-05}
{"train_loss": 0.13020123541355133, "global_step": 100841, "epoch": 1133, "lr": 6.980923254487197e-05}
{"train_loss": 0.1344742476940155, "global_step": 100842, "epoch": 1133, "lr": 6.980870024628657e-05}
{"train_loss": 0.2197866439819336, "global_step": 100843, "epoch": 1133, "lr": 6.980816794503813e-05}
{"train_loss": 0.24034824967384338, "global_step": 100844, "epoch": 1133, "lr": 6.980763564112672e-05}
{"train_loss": 0.22733579576015472, "global_step": 100845, "epoch": 1133, "lr": 6.980710333455243e-05}
{"train_loss": 0.18253786861896515, "global_step": 100846, "epoch": 1133, "lr": 6.98065710253153e-05}
{"train_loss": 0.15234550833702087, "global_step": 100847, "epoch": 1133, "lr": 6.980603871341542e-05}
{"train_loss": 0.09474051743745804, "global_step": 100848, "epoch": 1133, "lr": 6.980550639885288e-05}
{"train_loss": 0.12530243396759033, "global_step": 100849, "epoch": 1133, "lr": 6.98049740816277e-05}
{"train_loss": 0.21490280330181122, "global_step": 100850, "epoch": 1133, "lr": 6.980444176173999e-05}
{"train_loss": 0.10229302197694778, "global_step": 100851, "epoch": 1133, "lr": 6.980390943918983e-05}
{"train_loss": 0.15894325077533722, "global_step": 100852, "epoch": 1133, "lr": 6.980337711397726e-05}
{"train_loss": 0.18281544744968414, "global_step": 100853, "epoch": 1133, "lr": 6.980284478610237e-05}
{"train_loss": 0.19283969700336456, "global_step": 100854, "epoch": 1133, "lr": 6.980231245556524e-05}
{"train_loss": 0.18731608986854553, "global_step": 100855, "epoch": 1133, "lr": 6.980178012236591e-05}
{"train_loss": 0.18034611642360687, "global_step": 100856, "epoch": 1133, "lr": 6.980124778650448e-05}
{"train_loss": 0.12616923451423645, "global_step": 100857, "epoch": 1133, "lr": 6.980071544798103e-05}
{"train_loss": 0.12165552377700806, "global_step": 100858, "epoch": 1133, "lr": 6.98001831067956e-05}
{"train_loss": 0.19715116918087006, "global_step": 100859, "epoch": 1133, "lr": 6.979965076294826e-05}
{"train_loss": 0.13324220478534698, "global_step": 100860, "epoch": 1133, "lr": 6.979911841643913e-05}
{"train_loss": 0.1464153379201889, "global_step": 100861, "epoch": 1133, "lr": 6.979858606726824e-05}
{"train_loss": 0.11393359303474426, "global_step": 100862, "epoch": 1133, "lr": 6.979805371543566e-05}
{"train_loss": 0.1549423485994339, "global_step": 100863, "epoch": 1133, "lr": 6.979752136094149e-05}
{"train_loss": 0.13165509700775146, "global_step": 100864, "epoch": 1133, "lr": 6.979698900378578e-05}
{"train_loss": 0.23700718581676483, "global_step": 100865, "epoch": 1133, "lr": 6.97964566439686e-05}
{"train_loss": 0.20230963826179504, "global_step": 100866, "epoch": 1133, "lr": 6.979592428149003e-05}
{"train_loss": 0.1644488275051117, "global_step": 100867, "epoch": 1133, "lr": 6.979539191635014e-05}
{"train_loss": 0.15532146394252777, "global_step": 100868, "epoch": 1133, "lr": 6.979485954854901e-05}
{"train_loss": 0.13871799409389496, "global_step": 100869, "epoch": 1133, "lr": 6.979432717808668e-05}
{"train_loss": 0.16937030851840973, "global_step": 100870, "epoch": 1133, "lr": 6.979379480496326e-05}
{"train_loss": 0.23520225286483765, "global_step": 100871, "epoch": 1133, "lr": 6.979326242917881e-05}
{"train_loss": 0.1802177131175995, "global_step": 100872, "epoch": 1133, "lr": 6.979273005073338e-05}
{"train_loss": 0.21430101990699768, "global_step": 100873, "epoch": 1133, "lr": 6.979219766962707e-05}
{"train_loss": 0.1470193862915039, "global_step": 100874, "epoch": 1133, "lr": 6.979166528585994e-05}
{"train_loss": 0.15858455002307892, "global_step": 100875, "epoch": 1133, "lr": 6.979113289943207e-05}
{"train_loss": 0.15407684445381165, "global_step": 100876, "epoch": 1133, "lr": 6.979060051034351e-05}
{"train_loss": 0.18921467661857605, "global_step": 100877, "epoch": 1133, "lr": 6.979006811859434e-05}
{"train_loss": 0.17793075740337372, "global_step": 100878, "epoch": 1133, "lr": 6.978953572418462e-05}
{"train_loss": 0.13588152825832367, "global_step": 100879, "epoch": 1133, "lr": 6.978900332711448e-05}
{"train_loss": 0.13254442811012268, "global_step": 100880, "epoch": 1133, "lr": 6.978847092738392e-05}
{"train_loss": 0.13482666015625, "global_step": 100881, "epoch": 1133, "lr": 6.978793852499307e-05}
{"train_loss": 0.19288086891174316, "global_step": 100882, "epoch": 1133, "lr": 6.978740611994196e-05}
{"train_loss": 0.12091328203678131, "global_step": 100883, "epoch": 1133, "lr": 6.978687371223066e-05}
{"train_loss": 0.11598639190196991, "global_step": 100884, "epoch": 1133, "lr": 6.978634130185926e-05}
{"train_loss": 0.1433134377002716, "global_step": 100885, "epoch": 1133, "lr": 6.978580888882785e-05}
{"train_loss": 0.15726596117019653, "global_step": 100886, "epoch": 1133, "lr": 6.978527647313647e-05}
{"train_loss": 0.10954540222883224, "global_step": 100887, "epoch": 1133, "lr": 6.978474405478519e-05}
{"train_loss": 0.1430223137140274, "global_step": 100888, "epoch": 1133, "lr": 6.97842116337741e-05}
{"train_loss": 0.158803328871727, "global_step": 100889, "epoch": 1133, "lr": 6.978367921010326e-05}
{"train_loss": 0.23007138073444366, "global_step": 100890, "epoch": 1133, "lr": 6.978314678377275e-05}
{"train_loss": 0.2409319132566452, "global_step": 100891, "epoch": 1133, "lr": 6.978261435478264e-05}
{"train_loss": 0.22114680707454681, "global_step": 100892, "epoch": 1133, "lr": 6.978208192313298e-05}
{"train_loss": 0.20636098086833954, "global_step": 100893, "epoch": 1133, "lr": 6.978154948882387e-05}
{"train_loss": 0.15464606881141663, "global_step": 100894, "epoch": 1133, "lr": 6.978101705185539e-05}
{"train_loss": 0.1268727034330368, "global_step": 100895, "epoch": 1133, "lr": 6.978048461222756e-05}
{"train_loss": 0.1691201776266098, "global_step": 100896, "epoch": 1133, "lr": 6.977995216994049e-05}
{"train_loss": 0.20335470139980316, "global_step": 100897, "epoch": 1133, "lr": 6.977941972499426e-05}
{"train_loss": 0.16662302613258362, "global_step": 100898, "epoch": 1133, "lr": 6.977888727738894e-05}
{"train_loss": 0.22770117223262787, "global_step": 100899, "epoch": 1133, "lr": 6.977835482712457e-05}
{"train_loss": 0.17459547519683838, "global_step": 100900, "epoch": 1133, "lr": 6.977782237420125e-05}
{"train_loss": 0.18343600630760193, "global_step": 100901, "epoch": 1133, "lr": 6.977728991861904e-05}
{"train_loss": 0.11046041548252106, "global_step": 100902, "epoch": 1133, "lr": 6.977675746037802e-05}
{"train_loss": 0.16653895378112793, "global_step": 100903, "epoch": 1133, "lr": 6.977622499947824e-05}
{"train_loss": 0.17911271750926971, "global_step": 100904, "epoch": 1133, "lr": 6.977569253591981e-05}
{"train_loss": 0.1555033177137375, "global_step": 100905, "epoch": 1133, "lr": 6.977516006970277e-05}
{"train_loss": 0.11838095635175705, "global_step": 100906, "epoch": 1133, "lr": 6.97746276008272e-05}
{"train_loss": 0.2529262900352478, "global_step": 100907, "epoch": 1133, "lr": 6.977409512929317e-05}
{"train_loss": 0.167464479804039, "global_step": 100908, "epoch": 1133, "lr": 6.977356265510076e-05}
{"train_loss": 0.14802807569503784, "global_step": 100909, "epoch": 1133, "lr": 6.977303017825003e-05}
{"train_loss": 0.12223239988088608, "global_step": 100910, "epoch": 1133, "lr": 6.977249769874106e-05}
{"train_loss": 0.17587465047836304, "global_step": 100911, "epoch": 1133, "lr": 6.977196521657391e-05}
{"train_loss": 0.15013238787651062, "global_step": 100912, "epoch": 1133, "lr": 6.977143273174866e-05}
{"train_loss": 0.16653349995613098, "global_step": 100913, "epoch": 1133, "lr": 6.977090024426538e-05}
{"train_loss": 0.14906352758407593, "global_step": 100914, "epoch": 1133, "lr": 6.977036775412415e-05}
{"train_loss": 0.09375274926424026, "global_step": 100915, "epoch": 1133, "lr": 6.976983526132505e-05}
{"train_loss": 0.23220138251781464, "global_step": 100916, "epoch": 1133, "lr": 6.976930276586811e-05}
{"train_loss": 0.13971088826656342, "global_step": 100917, "epoch": 1133, "lr": 6.976877026775344e-05}
{"train_loss": 0.15540693700313568, "global_step": 100918, "epoch": 1133, "lr": 6.976823776698111e-05}
{"train_loss": 0.20352724194526672, "global_step": 100919, "epoch": 1133, "lr": 6.976770526355117e-05}
{"train_loss": 0.15405365824699402, "global_step": 100920, "epoch": 1133, "lr": 6.97671727574637e-05}
{"train_loss": 0.2466387003660202, "global_step": 100921, "epoch": 1133, "lr": 6.97666402487188e-05}
{"train_loss": 0.19243118166923523, "global_step": 100922, "epoch": 1133, "lr": 6.976610773731649e-05}
{"train_loss": 0.18388262391090393, "global_step": 100923, "epoch": 1133, "lr": 6.976557522325688e-05}
{"train_loss": 0.133588507771492, "global_step": 100924, "epoch": 1133, "lr": 6.976504270654003e-05}
{"train_loss": 0.16630306086513433, "global_step": 100925, "epoch": 1133, "lr": 6.9764510187166e-05, "val_loss": 4.767066478729248}
{"train_loss": 0.1397387981414795, "global_step": 100926, "epoch": 1134, "lr": 6.976397766513487e-05}
{"train_loss": 0.227162167429924, "global_step": 100927, "epoch": 1134, "lr": 6.976344514044672e-05}
{"train_loss": 0.1710294485092163, "global_step": 100928, "epoch": 1134, "lr": 6.976291261310161e-05}
{"train_loss": 0.2276185005903244, "global_step": 100929, "epoch": 1134, "lr": 6.976238008309964e-05}
{"train_loss": 0.16646674275398254, "global_step": 100930, "epoch": 1134, "lr": 6.976184755044084e-05}
{"train_loss": 0.21893411874771118, "global_step": 100931, "epoch": 1134, "lr": 6.976131501512528e-05}
{"train_loss": 0.1565372347831726, "global_step": 100932, "epoch": 1134, "lr": 6.976078247715308e-05}
{"train_loss": 0.1336672157049179, "global_step": 100933, "epoch": 1134, "lr": 6.976024993652429e-05}
{"train_loss": 0.22918367385864258, "global_step": 100934, "epoch": 1134, "lr": 6.975971739323895e-05}
{"train_loss": 0.17922645807266235, "global_step": 100935, "epoch": 1134, "lr": 6.975918484729717e-05}
{"train_loss": 0.18649473786354065, "global_step": 100936, "epoch": 1134, "lr": 6.975865229869903e-05}
{"train_loss": 0.24375639855861664, "global_step": 100937, "epoch": 1134, "lr": 6.975811974744455e-05}
{"train_loss": 0.15385879576206207, "global_step": 100938, "epoch": 1134, "lr": 6.975758719353383e-05}
{"train_loss": 0.1679655909538269, "global_step": 100939, "epoch": 1134, "lr": 6.975705463696695e-05}
{"train_loss": 0.19238871335983276, "global_step": 100940, "epoch": 1134, "lr": 6.975652207774399e-05}
{"train_loss": 0.20040635764598846, "global_step": 100941, "epoch": 1134, "lr": 6.9755989515865e-05}
{"train_loss": 0.18950968980789185, "global_step": 100942, "epoch": 1134, "lr": 6.975545695133006e-05}
{"train_loss": 0.15893620252609253, "global_step": 100943, "epoch": 1134, "lr": 6.975492438413923e-05}
{"train_loss": 0.1830417364835739, "global_step": 100944, "epoch": 1134, "lr": 6.97543918142926e-05}
{"train_loss": 0.1497652381658554, "global_step": 100945, "epoch": 1134, "lr": 6.975385924179023e-05}
{"train_loss": 0.11085968464612961, "global_step": 100946, "epoch": 1134, "lr": 6.975332666663219e-05}
{"train_loss": 0.24160754680633545, "global_step": 100947, "epoch": 1134, "lr": 6.975279408881855e-05}
{"train_loss": 0.09969978779554367, "global_step": 100948, "epoch": 1134, "lr": 6.975226150834939e-05}
{"train_loss": 0.2128806710243225, "global_step": 100949, "epoch": 1134, "lr": 6.975172892522478e-05}
{"train_loss": 0.15422102808952332, "global_step": 100950, "epoch": 1134, "lr": 6.97511963394448e-05}
{"train_loss": 0.1197008341550827, "global_step": 100951, "epoch": 1134, "lr": 6.97506637510095e-05}
{"train_loss": 0.13004273176193237, "global_step": 100952, "epoch": 1134, "lr": 6.975013115991897e-05}
{"train_loss": 0.1652577519416809, "global_step": 100953, "epoch": 1134, "lr": 6.974959856617328e-05}
{"train_loss": 0.15815958380699158, "global_step": 100954, "epoch": 1134, "lr": 6.97490659697725e-05}
{"train_loss": 0.11989723145961761, "global_step": 100955, "epoch": 1134, "lr": 6.974853337071669e-05}
{"train_loss": 0.1885889172554016, "global_step": 100956, "epoch": 1134, "lr": 6.974800076900594e-05}
{"train_loss": 0.1630813479423523, "global_step": 100957, "epoch": 1134, "lr": 6.974746816464031e-05}
{"train_loss": 0.15511029958724976, "global_step": 100958, "epoch": 1134, "lr": 6.974693555761987e-05}
{"train_loss": 0.12597621977329254, "global_step": 100959, "epoch": 1134, "lr": 6.974640294794468e-05}
{"train_loss": 0.2562105059623718, "global_step": 100960, "epoch": 1134, "lr": 6.974587033561483e-05}
{"train_loss": 0.19730129837989807, "global_step": 100961, "epoch": 1134, "lr": 6.974533772063041e-05}
{"train_loss": 0.17913250625133514, "global_step": 100962, "epoch": 1134, "lr": 6.974480510299148e-05}
{"train_loss": 0.15867698192596436, "global_step": 100963, "epoch": 1134, "lr": 6.974427248269807e-05}
{"train_loss": 0.16377609968185425, "global_step": 100964, "epoch": 1134, "lr": 6.974373985975029e-05}
{"train_loss": 0.16947303712368011, "global_step": 100965, "epoch": 1134, "lr": 6.974320723414822e-05}
{"train_loss": 0.13825291395187378, "global_step": 100966, "epoch": 1134, "lr": 6.97426746058919e-05}
{"train_loss": 0.10739795863628387, "global_step": 100967, "epoch": 1134, "lr": 6.974214197498141e-05}
{"train_loss": 0.2026987224817276, "global_step": 100968, "epoch": 1134, "lr": 6.974160934141685e-05}
{"train_loss": 0.1614820659160614, "global_step": 100969, "epoch": 1134, "lr": 6.974107670519826e-05}
{"train_loss": 0.1453913450241089, "global_step": 100970, "epoch": 1134, "lr": 6.974054406632573e-05}
{"train_loss": 0.1511499583721161, "global_step": 100971, "epoch": 1134, "lr": 6.974001142479932e-05}
{"train_loss": 0.1295740008354187, "global_step": 100972, "epoch": 1134, "lr": 6.97394787806191e-05}
{"train_loss": 0.13592001795768738, "global_step": 100973, "epoch": 1134, "lr": 6.973894613378517e-05}
{"train_loss": 0.08863503485918045, "global_step": 100974, "epoch": 1134, "lr": 6.973841348429756e-05}
{"train_loss": 0.20812463760375977, "global_step": 100975, "epoch": 1134, "lr": 6.973788083215637e-05}
{"train_loss": 0.1314309686422348, "global_step": 100976, "epoch": 1134, "lr": 6.973734817736164e-05}
{"train_loss": 0.21744704246520996, "global_step": 100977, "epoch": 1134, "lr": 6.973681551991349e-05}
{"train_loss": 0.13666120171546936, "global_step": 100978, "epoch": 1134, "lr": 6.973628285981196e-05}
{"train_loss": 0.2072412371635437, "global_step": 100979, "epoch": 1134, "lr": 6.973575019705713e-05}
{"train_loss": 0.1625044047832489, "global_step": 100980, "epoch": 1134, "lr": 6.973521753164906e-05}
{"train_loss": 0.19559426605701447, "global_step": 100981, "epoch": 1134, "lr": 6.973468486358784e-05}
{"train_loss": 0.20246821641921997, "global_step": 100982, "epoch": 1134, "lr": 6.973415219287353e-05}
{"train_loss": 0.12040494382381439, "global_step": 100983, "epoch": 1134, "lr": 6.97336195195062e-05}
{"train_loss": 0.13484150171279907, "global_step": 100984, "epoch": 1134, "lr": 6.973308684348593e-05}
{"train_loss": 0.09816500544548035, "global_step": 100985, "epoch": 1134, "lr": 6.973255416481277e-05}
{"train_loss": 0.12594081461429596, "global_step": 100986, "epoch": 1134, "lr": 6.973202148348683e-05}
{"train_loss": 0.14938385784626007, "global_step": 100987, "epoch": 1134, "lr": 6.973148879950815e-05}
{"train_loss": 0.12123528122901917, "global_step": 100988, "epoch": 1134, "lr": 6.973095611287682e-05}
{"train_loss": 0.1335054636001587, "global_step": 100989, "epoch": 1134, "lr": 6.973042342359288e-05}
{"train_loss": 0.15154826641082764, "global_step": 100990, "epoch": 1134, "lr": 6.972989073165644e-05}
{"train_loss": 0.15633459389209747, "global_step": 100991, "epoch": 1134, "lr": 6.972935803706756e-05}
{"train_loss": 0.17747513949871063, "global_step": 100992, "epoch": 1134, "lr": 6.972882533982632e-05}
{"train_loss": 0.20086795091629028, "global_step": 100993, "epoch": 1134, "lr": 6.972829263993276e-05}
{"train_loss": 0.1568622589111328, "global_step": 100994, "epoch": 1134, "lr": 6.972775993738698e-05}
{"train_loss": 0.17186197638511658, "global_step": 100995, "epoch": 1134, "lr": 6.972722723218903e-05}
{"train_loss": 0.2534874975681305, "global_step": 100996, "epoch": 1134, "lr": 6.9726694524339e-05}
{"train_loss": 0.12557490170001984, "global_step": 100997, "epoch": 1134, "lr": 6.972616181383698e-05}
{"train_loss": 0.23438821732997894, "global_step": 100998, "epoch": 1134, "lr": 6.972562910068299e-05}
{"train_loss": 0.11918412148952484, "global_step": 100999, "epoch": 1134, "lr": 6.972509638487714e-05}
{"train_loss": 0.13495881855487823, "global_step": 101000, "epoch": 1134, "lr": 6.97245636664195e-05}
{"train_loss": 0.12752573192119598, "global_step": 101001, "epoch": 1134, "lr": 6.972403094531013e-05}
{"train_loss": 0.10522307455539703, "global_step": 101002, "epoch": 1134, "lr": 6.97234982215491e-05}
{"train_loss": 0.18236875534057617, "global_step": 101003, "epoch": 1134, "lr": 6.972296549513648e-05}
{"train_loss": 0.27152785658836365, "global_step": 101004, "epoch": 1134, "lr": 6.972243276607235e-05}
{"train_loss": 0.15198911726474762, "global_step": 101005, "epoch": 1134, "lr": 6.972190003435678e-05}
{"train_loss": 0.17348210513591766, "global_step": 101006, "epoch": 1134, "lr": 6.972136729998984e-05}
{"train_loss": 0.17668572068214417, "global_step": 101007, "epoch": 1134, "lr": 6.97208345629716e-05}
{"train_loss": 0.13213692605495453, "global_step": 101008, "epoch": 1134, "lr": 6.972030182330215e-05}
{"train_loss": 0.2036055326461792, "global_step": 101009, "epoch": 1134, "lr": 6.971976908098152e-05}
{"train_loss": 0.17044702172279358, "global_step": 101010, "epoch": 1134, "lr": 6.971923633600983e-05}
{"train_loss": 0.2698749005794525, "global_step": 101011, "epoch": 1134, "lr": 6.971870358838712e-05}
{"train_loss": 0.1834162324666977, "global_step": 101012, "epoch": 1134, "lr": 6.971817083811347e-05}
{"train_loss": 0.1590389907360077, "global_step": 101013, "epoch": 1134, "lr": 6.971763808518896e-05}
{"train_loss": 0.1675262436940429, "global_step": 101014, "epoch": 1134, "lr": 6.971710532961364e-05, "val_loss": 4.697056293487549}
{"train_loss": 0.14519524574279785, "global_step": 101015, "epoch": 1135, "lr": 6.971657257138762e-05}
{"train_loss": 0.21955333650112152, "global_step": 101016, "epoch": 1135, "lr": 6.971603981051092e-05}
{"train_loss": 0.13059690594673157, "global_step": 101017, "epoch": 1135, "lr": 6.971550704698366e-05}
{"train_loss": 0.14407919347286224, "global_step": 101018, "epoch": 1135, "lr": 6.971497428080587e-05}
{"train_loss": 0.15671095252037048, "global_step": 101019, "epoch": 1135, "lr": 6.971444151197766e-05}
{"train_loss": 0.15776975452899933, "global_step": 101020, "epoch": 1135, "lr": 6.971390874049908e-05}
{"train_loss": 0.164072185754776, "global_step": 101021, "epoch": 1135, "lr": 6.97133759663702e-05}
{"train_loss": 0.18341508507728577, "global_step": 101022, "epoch": 1135, "lr": 6.971284318959108e-05}
{"train_loss": 0.1907237470149994, "global_step": 101023, "epoch": 1135, "lr": 6.971231041016183e-05}
{"train_loss": 0.16753576695919037, "global_step": 101024, "epoch": 1135, "lr": 6.97117776280825e-05}
{"train_loss": 0.14012271165847778, "global_step": 101025, "epoch": 1135, "lr": 6.971124484335316e-05}
{"train_loss": 0.14867399632930756, "global_step": 101026, "epoch": 1135, "lr": 6.971071205597388e-05}
{"train_loss": 0.12338037043809891, "global_step": 101027, "epoch": 1135, "lr": 6.971017926594473e-05}
{"train_loss": 0.21773073077201843, "global_step": 101028, "epoch": 1135, "lr": 6.97096464732658e-05}
{"train_loss": 0.11172235012054443, "global_step": 101029, "epoch": 1135, "lr": 6.970911367793713e-05}
{"train_loss": 0.3108135759830475, "global_step": 101030, "epoch": 1135, "lr": 6.970858087995883e-05}
{"train_loss": 0.20033708214759827, "global_step": 101031, "epoch": 1135, "lr": 6.970804807933094e-05}
{"train_loss": 0.12766717374324799, "global_step": 101032, "epoch": 1135, "lr": 6.970751527605354e-05}
{"train_loss": 0.15116886794567108, "global_step": 101033, "epoch": 1135, "lr": 6.970698247012672e-05}
{"train_loss": 0.1664804220199585, "global_step": 101034, "epoch": 1135, "lr": 6.970644966155053e-05}
{"train_loss": 0.18076379597187042, "global_step": 101035, "epoch": 1135, "lr": 6.970591685032504e-05}
{"train_loss": 0.1262027472257614, "global_step": 101036, "epoch": 1135, "lr": 6.970538403645034e-05}
{"train_loss": 0.18782266974449158, "global_step": 101037, "epoch": 1135, "lr": 6.970485121992648e-05}
{"train_loss": 0.19939765334129333, "global_step": 101038, "epoch": 1135, "lr": 6.970431840075355e-05}
{"train_loss": 0.1943320482969284, "global_step": 101039, "epoch": 1135, "lr": 6.970378557893162e-05}
{"train_loss": 0.14711274206638336, "global_step": 101040, "epoch": 1135, "lr": 6.970325275446075e-05}
{"train_loss": 0.1323154866695404, "global_step": 101041, "epoch": 1135, "lr": 6.970271992734101e-05}
{"train_loss": 0.18407553434371948, "global_step": 101042, "epoch": 1135, "lr": 6.970218709757249e-05}
{"train_loss": 0.13704916834831238, "global_step": 101043, "epoch": 1135, "lr": 6.970165426515525e-05}
{"train_loss": 0.17159713804721832, "global_step": 101044, "epoch": 1135, "lr": 6.970112143008936e-05}
{"train_loss": 0.13421876728534698, "global_step": 101045, "epoch": 1135, "lr": 6.97005885923749e-05}
{"train_loss": 0.2192779779434204, "global_step": 101046, "epoch": 1135, "lr": 6.970005575201193e-05}
{"train_loss": 0.17517130076885223, "global_step": 101047, "epoch": 1135, "lr": 6.969952290900052e-05}
{"train_loss": 0.16977928578853607, "global_step": 101048, "epoch": 1135, "lr": 6.969899006334077e-05}
{"train_loss": 0.13263684511184692, "global_step": 101049, "epoch": 1135, "lr": 6.96984572150327e-05}
{"train_loss": 0.10984816402196884, "global_step": 101050, "epoch": 1135, "lr": 6.969792436407643e-05}
{"train_loss": 0.2008170634508133, "global_step": 101051, "epoch": 1135, "lr": 6.969739151047203e-05}
{"train_loss": 0.15657269954681396, "global_step": 101052, "epoch": 1135, "lr": 6.969685865421954e-05}
{"train_loss": 0.15380267798900604, "global_step": 101053, "epoch": 1135, "lr": 6.969632579531905e-05}
{"train_loss": 0.12617987394332886, "global_step": 101054, "epoch": 1135, "lr": 6.969579293377062e-05}
{"train_loss": 0.14497366547584534, "global_step": 101055, "epoch": 1135, "lr": 6.969526006957434e-05}
{"train_loss": 0.1001356765627861, "global_step": 101056, "epoch": 1135, "lr": 6.969472720273027e-05}
{"train_loss": 0.16461613774299622, "global_step": 101057, "epoch": 1135, "lr": 6.969419433323848e-05}
{"train_loss": 0.14229124784469604, "global_step": 101058, "epoch": 1135, "lr": 6.969366146109905e-05}
{"train_loss": 0.17947155237197876, "global_step": 101059, "epoch": 1135, "lr": 6.969312858631205e-05}
{"train_loss": 0.10356582701206207, "global_step": 101060, "epoch": 1135, "lr": 6.969259570887755e-05}
{"train_loss": 0.10256969183683395, "global_step": 101061, "epoch": 1135, "lr": 6.969206282879562e-05}
{"train_loss": 0.14591369032859802, "global_step": 101062, "epoch": 1135, "lr": 6.969152994606633e-05}
{"train_loss": 0.19514809548854828, "global_step": 101063, "epoch": 1135, "lr": 6.969099706068975e-05}
{"train_loss": 0.19433817267417908, "global_step": 101064, "epoch": 1135, "lr": 6.969046417266595e-05}
{"train_loss": 0.12312474101781845, "global_step": 101065, "epoch": 1135, "lr": 6.968993128199502e-05}
{"train_loss": 0.10875869542360306, "global_step": 101066, "epoch": 1135, "lr": 6.968939838867702e-05}
{"train_loss": 0.15930253267288208, "global_step": 101067, "epoch": 1135, "lr": 6.9688865492712e-05}
{"train_loss": 0.24041156470775604, "global_step": 101068, "epoch": 1135, "lr": 6.968833259410008e-05}
{"train_loss": 0.17081457376480103, "global_step": 101069, "epoch": 1135, "lr": 6.96877996928413e-05}
{"train_loss": 0.16970773041248322, "global_step": 101070, "epoch": 1135, "lr": 6.968726678893571e-05}
{"train_loss": 0.12283308058977127, "global_step": 101071, "epoch": 1135, "lr": 6.968673388238342e-05}
{"train_loss": 0.1100568026304245, "global_step": 101072, "epoch": 1135, "lr": 6.96862009731845e-05}
{"train_loss": 0.1736280471086502, "global_step": 101073, "epoch": 1135, "lr": 6.9685668061339e-05}
{"train_loss": 0.08583109825849533, "global_step": 101074, "epoch": 1135, "lr": 6.968513514684701e-05}
{"train_loss": 0.14432069659233093, "global_step": 101075, "epoch": 1135, "lr": 6.968460222970859e-05}
{"train_loss": 0.19792447984218597, "global_step": 101076, "epoch": 1135, "lr": 6.96840693099238e-05}
{"train_loss": 0.16369275748729706, "global_step": 101077, "epoch": 1135, "lr": 6.968353638749275e-05}
{"train_loss": 0.1843957006931305, "global_step": 101078, "epoch": 1135, "lr": 6.968300346241546e-05}
{"train_loss": 0.14125239849090576, "global_step": 101079, "epoch": 1135, "lr": 6.968247053469206e-05}
{"train_loss": 0.10194103419780731, "global_step": 101080, "epoch": 1135, "lr": 6.968193760432258e-05}
{"train_loss": 0.15187764167785645, "global_step": 101081, "epoch": 1135, "lr": 6.96814046713071e-05}
{"train_loss": 0.12334268540143967, "global_step": 101082, "epoch": 1135, "lr": 6.968087173564569e-05}
{"train_loss": 0.12383513152599335, "global_step": 101083, "epoch": 1135, "lr": 6.968033879733843e-05}
{"train_loss": 0.18692287802696228, "global_step": 101084, "epoch": 1135, "lr": 6.96798058563854e-05}
{"train_loss": 0.25079408288002014, "global_step": 101085, "epoch": 1135, "lr": 6.967927291278666e-05}
{"train_loss": 0.13730096817016602, "global_step": 101086, "epoch": 1135, "lr": 6.967873996654226e-05}
{"train_loss": 0.10376850515604019, "global_step": 101087, "epoch": 1135, "lr": 6.967820701765231e-05}
{"train_loss": 0.1922510266304016, "global_step": 101088, "epoch": 1135, "lr": 6.967767406611687e-05}
{"train_loss": 0.10695872455835342, "global_step": 101089, "epoch": 1135, "lr": 6.9677141111936e-05}
{"train_loss": 0.12038534134626389, "global_step": 101090, "epoch": 1135, "lr": 6.967660815510977e-05}
{"train_loss": 0.18970279395580292, "global_step": 101091, "epoch": 1135, "lr": 6.967607519563826e-05}
{"train_loss": 0.17138513922691345, "global_step": 101092, "epoch": 1135, "lr": 6.967554223352155e-05}
{"train_loss": 0.1577652096748352, "global_step": 101093, "epoch": 1135, "lr": 6.967500926875971e-05}
{"train_loss": 0.2178429216146469, "global_step": 101094, "epoch": 1135, "lr": 6.96744763013528e-05}
{"train_loss": 0.1850765347480774, "global_step": 101095, "epoch": 1135, "lr": 6.967394333130089e-05}
{"train_loss": 0.17047037184238434, "global_step": 101096, "epoch": 1135, "lr": 6.967341035860405e-05}
{"train_loss": 0.1657385677099228, "global_step": 101097, "epoch": 1135, "lr": 6.967287738326238e-05}
{"train_loss": 0.15343035757541656, "global_step": 101098, "epoch": 1135, "lr": 6.967234440527591e-05}
{"train_loss": 0.2500661015510559, "global_step": 101099, "epoch": 1135, "lr": 6.967181142464475e-05}
{"train_loss": 0.1133146733045578, "global_step": 101100, "epoch": 1135, "lr": 6.967127844136895e-05}
{"train_loss": 0.15853269398212433, "global_step": 101101, "epoch": 1135, "lr": 6.967074545544858e-05}
{"train_loss": 0.12039351463317871, "global_step": 101102, "epoch": 1135, "lr": 6.967021246688373e-05}
{"train_loss": 0.15960666654484995, "global_step": 101103, "epoch": 1135, "lr": 6.966967947567446e-05, "val_loss": 4.7640767097473145, "train_action_mse_error": 19.847904205322266}
{"train_loss": 0.12321078777313232, "global_step": 101104, "epoch": 1136, "lr": 6.966914648182084e-05}
{"train_loss": 0.13143528997898102, "global_step": 101105, "epoch": 1136, "lr": 6.966861348532294e-05}
{"train_loss": 0.20063593983650208, "global_step": 101106, "epoch": 1136, "lr": 6.966808048618081e-05}
{"train_loss": 0.13990306854248047, "global_step": 101107, "epoch": 1136, "lr": 6.966754748439458e-05}
{"train_loss": 0.13715112209320068, "global_step": 101108, "epoch": 1136, "lr": 6.966701447996427e-05}
{"train_loss": 0.10744056105613708, "global_step": 101109, "epoch": 1136, "lr": 6.966648147288999e-05}
{"train_loss": 0.15483500063419342, "global_step": 101110, "epoch": 1136, "lr": 6.966594846317178e-05}
{"train_loss": 0.16938211023807526, "global_step": 101111, "epoch": 1136, "lr": 6.966541545080972e-05}
{"train_loss": 0.20822453498840332, "global_step": 101112, "epoch": 1136, "lr": 6.966488243580387e-05}
{"train_loss": 0.16885891556739807, "global_step": 101113, "epoch": 1136, "lr": 6.966434941815434e-05}
{"train_loss": 0.20394372940063477, "global_step": 101114, "epoch": 1136, "lr": 6.966381639786117e-05}
{"train_loss": 0.14529626071453094, "global_step": 101115, "epoch": 1136, "lr": 6.966328337492444e-05}
{"train_loss": 0.13990333676338196, "global_step": 101116, "epoch": 1136, "lr": 6.966275034934422e-05}
{"train_loss": 0.18436318635940552, "global_step": 101117, "epoch": 1136, "lr": 6.966221732112058e-05}
{"train_loss": 0.14375902712345123, "global_step": 101118, "epoch": 1136, "lr": 6.96616842902536e-05}
{"train_loss": 0.17810183763504028, "global_step": 101119, "epoch": 1136, "lr": 6.966115125674335e-05}
{"train_loss": 0.12690389156341553, "global_step": 101120, "epoch": 1136, "lr": 6.96606182205899e-05}
{"train_loss": 0.18827645480632782, "global_step": 101121, "epoch": 1136, "lr": 6.96600851817933e-05}
{"train_loss": 0.13381098210811615, "global_step": 101122, "epoch": 1136, "lr": 6.965955214035364e-05}
{"train_loss": 0.20032961666584015, "global_step": 101123, "epoch": 1136, "lr": 6.965901909627101e-05}
{"train_loss": 0.19872400164604187, "global_step": 101124, "epoch": 1136, "lr": 6.965848604954546e-05}
{"train_loss": 0.15746524930000305, "global_step": 101125, "epoch": 1136, "lr": 6.965795300017706e-05}
{"train_loss": 0.13423466682434082, "global_step": 101126, "epoch": 1136, "lr": 6.96574199481659e-05}
{"train_loss": 0.16431961953639984, "global_step": 101127, "epoch": 1136, "lr": 6.965688689351204e-05}
{"train_loss": 0.1759590357542038, "global_step": 101128, "epoch": 1136, "lr": 6.965635383621553e-05}
{"train_loss": 0.14407604932785034, "global_step": 101129, "epoch": 1136, "lr": 6.965582077627648e-05}
{"train_loss": 0.27761149406433105, "global_step": 101130, "epoch": 1136, "lr": 6.965528771369495e-05}
{"train_loss": 0.10904816538095474, "global_step": 101131, "epoch": 1136, "lr": 6.9654754648471e-05}
{"train_loss": 0.0902465209364891, "global_step": 101132, "epoch": 1136, "lr": 6.96542215806047e-05}
{"train_loss": 0.14776577055454254, "global_step": 101133, "epoch": 1136, "lr": 6.965368851009614e-05}
{"train_loss": 0.16081777215003967, "global_step": 101134, "epoch": 1136, "lr": 6.965315543694537e-05}
{"train_loss": 0.2061873972415924, "global_step": 101135, "epoch": 1136, "lr": 6.965262236115249e-05}
{"train_loss": 0.09780304878950119, "global_step": 101136, "epoch": 1136, "lr": 6.965208928271753e-05}
{"train_loss": 0.22130869328975677, "global_step": 101137, "epoch": 1136, "lr": 6.96515562016406e-05}
{"train_loss": 0.13067945837974548, "global_step": 101138, "epoch": 1136, "lr": 6.965102311792175e-05}
{"train_loss": 0.17393144965171814, "global_step": 101139, "epoch": 1136, "lr": 6.965049003156108e-05}
{"train_loss": 0.178279310464859, "global_step": 101140, "epoch": 1136, "lr": 6.964995694255862e-05}
{"train_loss": 0.1840381622314453, "global_step": 101141, "epoch": 1136, "lr": 6.964942385091447e-05}
{"train_loss": 0.15311935544013977, "global_step": 101142, "epoch": 1136, "lr": 6.96488907566287e-05}
{"train_loss": 0.16451050341129303, "global_step": 101143, "epoch": 1136, "lr": 6.964835765970137e-05}
{"train_loss": 0.15176790952682495, "global_step": 101144, "epoch": 1136, "lr": 6.964782456013255e-05}
{"train_loss": 0.16884435713291168, "global_step": 101145, "epoch": 1136, "lr": 6.964729145792234e-05}
{"train_loss": 0.14934243261814117, "global_step": 101146, "epoch": 1136, "lr": 6.964675835307077e-05}
{"train_loss": 0.15860725939273834, "global_step": 101147, "epoch": 1136, "lr": 6.964622524557796e-05}
{"train_loss": 0.15743033587932587, "global_step": 101148, "epoch": 1136, "lr": 6.964569213544393e-05}
{"train_loss": 0.16704770922660828, "global_step": 101149, "epoch": 1136, "lr": 6.964515902266877e-05}
{"train_loss": 0.20664051175117493, "global_step": 101150, "epoch": 1136, "lr": 6.964462590725258e-05}
{"train_loss": 0.19440484046936035, "global_step": 101151, "epoch": 1136, "lr": 6.964409278919541e-05}
{"train_loss": 0.1639149785041809, "global_step": 101152, "epoch": 1136, "lr": 6.964355966849732e-05}
{"train_loss": 0.13758200407028198, "global_step": 101153, "epoch": 1136, "lr": 6.964302654515839e-05}
{"train_loss": 0.14507706463336945, "global_step": 101154, "epoch": 1136, "lr": 6.96424934191787e-05}
{"train_loss": 0.1510777622461319, "global_step": 101155, "epoch": 1136, "lr": 6.964196029055831e-05}
{"train_loss": 0.11441079527139664, "global_step": 101156, "epoch": 1136, "lr": 6.96414271592973e-05}
{"train_loss": 0.13413023948669434, "global_step": 101157, "epoch": 1136, "lr": 6.964089402539574e-05}
{"train_loss": 0.1658981740474701, "global_step": 101158, "epoch": 1136, "lr": 6.96403608888537e-05}
{"train_loss": 0.10360178351402283, "global_step": 101159, "epoch": 1136, "lr": 6.963982774967126e-05}
{"train_loss": 0.25389233231544495, "global_step": 101160, "epoch": 1136, "lr": 6.963929460784848e-05}
{"train_loss": 0.18323877453804016, "global_step": 101161, "epoch": 1136, "lr": 6.963876146338544e-05}
{"train_loss": 0.14568942785263062, "global_step": 101162, "epoch": 1136, "lr": 6.963822831628221e-05}
{"train_loss": 0.09756751358509064, "global_step": 101163, "epoch": 1136, "lr": 6.963769516653885e-05}
{"train_loss": 0.23195385932922363, "global_step": 101164, "epoch": 1136, "lr": 6.963716201415544e-05}
{"train_loss": 0.16135308146476746, "global_step": 101165, "epoch": 1136, "lr": 6.963662885913207e-05}
{"train_loss": 0.2522776126861572, "global_step": 101166, "epoch": 1136, "lr": 6.963609570146879e-05}
{"train_loss": 0.1927136778831482, "global_step": 101167, "epoch": 1136, "lr": 6.963556254116566e-05}
{"train_loss": 0.18876945972442627, "global_step": 101168, "epoch": 1136, "lr": 6.963502937822278e-05}
{"train_loss": 0.20099474489688873, "global_step": 101169, "epoch": 1136, "lr": 6.96344962126402e-05}
{"train_loss": 0.2179233431816101, "global_step": 101170, "epoch": 1136, "lr": 6.9633963044418e-05}
{"train_loss": 0.17682607471942902, "global_step": 101171, "epoch": 1136, "lr": 6.963342987355628e-05}
{"train_loss": 0.25512611865997314, "global_step": 101172, "epoch": 1136, "lr": 6.963289670005506e-05}
{"train_loss": 0.14460709691047668, "global_step": 101173, "epoch": 1136, "lr": 6.963236352391445e-05}
{"train_loss": 0.1666879802942276, "global_step": 101174, "epoch": 1136, "lr": 6.96318303451345e-05}
{"train_loss": 0.1516340672969818, "global_step": 101175, "epoch": 1136, "lr": 6.963129716371528e-05}
{"train_loss": 0.20626123249530792, "global_step": 101176, "epoch": 1136, "lr": 6.963076397965689e-05}
{"train_loss": 0.14522229135036469, "global_step": 101177, "epoch": 1136, "lr": 6.963023079295937e-05}
{"train_loss": 0.15955664217472076, "global_step": 101178, "epoch": 1136, "lr": 6.962969760362281e-05}
{"train_loss": 0.1430267095565796, "global_step": 101179, "epoch": 1136, "lr": 6.962916441164727e-05}
{"train_loss": 0.1637437492609024, "global_step": 101180, "epoch": 1136, "lr": 6.962863121703283e-05}
{"train_loss": 0.15470945835113525, "global_step": 101181, "epoch": 1136, "lr": 6.962809801977957e-05}
{"train_loss": 0.2097480595111847, "global_step": 101182, "epoch": 1136, "lr": 6.962756481988755e-05}
{"train_loss": 0.22198323905467987, "global_step": 101183, "epoch": 1136, "lr": 6.962703161735683e-05}
{"train_loss": 0.1538676619529724, "global_step": 101184, "epoch": 1136, "lr": 6.96264984121875e-05}
{"train_loss": 0.1402583122253418, "global_step": 101185, "epoch": 1136, "lr": 6.962596520437962e-05}
{"train_loss": 0.13009805977344513, "global_step": 101186, "epoch": 1136, "lr": 6.962543199393327e-05}
{"train_loss": 0.14325836300849915, "global_step": 101187, "epoch": 1136, "lr": 6.962489878084853e-05}
{"train_loss": 0.1417941451072693, "global_step": 101188, "epoch": 1136, "lr": 6.962436556512546e-05}
{"train_loss": 0.1111711785197258, "global_step": 101189, "epoch": 1136, "lr": 6.962383234676415e-05}
{"train_loss": 0.14571566879749298, "global_step": 101190, "epoch": 1136, "lr": 6.962329912576464e-05}
{"train_loss": 0.14137893915176392, "global_step": 101191, "epoch": 1136, "lr": 6.9622765902127e-05}
{"train_loss": 0.16399556507220428, "global_step": 101192, "epoch": 1136, "lr": 6.962223267585134e-05, "val_loss": 4.610487461090088}
{"train_loss": 0.14727437496185303, "global_step": 101193, "epoch": 1137, "lr": 6.962169944693771e-05}
{"train_loss": 0.12344759702682495, "global_step": 101194, "epoch": 1137, "lr": 6.962116621538617e-05}
{"train_loss": 0.11955252289772034, "global_step": 101195, "epoch": 1137, "lr": 6.962063298119681e-05}
{"train_loss": 0.184817373752594, "global_step": 101196, "epoch": 1137, "lr": 6.962009974436968e-05}
{"train_loss": 0.19494453072547913, "global_step": 101197, "epoch": 1137, "lr": 6.961956650490489e-05}
{"train_loss": 0.18304483592510223, "global_step": 101198, "epoch": 1137, "lr": 6.961903326280246e-05}
{"train_loss": 0.17345775663852692, "global_step": 101199, "epoch": 1137, "lr": 6.961850001806251e-05}
{"train_loss": 0.15761889517307281, "global_step": 101200, "epoch": 1137, "lr": 6.96179667706851e-05}
{"train_loss": 0.1601366251707077, "global_step": 101201, "epoch": 1137, "lr": 6.961743352067028e-05}
{"train_loss": 0.163802832365036, "global_step": 101202, "epoch": 1137, "lr": 6.961690026801814e-05}
{"train_loss": 0.17828680574893951, "global_step": 101203, "epoch": 1137, "lr": 6.961636701272874e-05}
{"train_loss": 0.1740783005952835, "global_step": 101204, "epoch": 1137, "lr": 6.961583375480217e-05}
{"train_loss": 0.17057380080223083, "global_step": 101205, "epoch": 1137, "lr": 6.961530049423846e-05}
{"train_loss": 0.18342085182666779, "global_step": 101206, "epoch": 1137, "lr": 6.961476723103775e-05}
{"train_loss": 0.14670300483703613, "global_step": 101207, "epoch": 1137, "lr": 6.961423396520007e-05}
{"train_loss": 0.1932869255542755, "global_step": 101208, "epoch": 1137, "lr": 6.961370069672548e-05}
{"train_loss": 0.15787319839000702, "global_step": 101209, "epoch": 1137, "lr": 6.961316742561406e-05}
{"train_loss": 0.1582421064376831, "global_step": 101210, "epoch": 1137, "lr": 6.96126341518659e-05}
{"train_loss": 0.18287549912929535, "global_step": 101211, "epoch": 1137, "lr": 6.961210087548107e-05}
{"train_loss": 0.0936194509267807, "global_step": 101212, "epoch": 1137, "lr": 6.961156759645961e-05}
{"train_loss": 0.20736122131347656, "global_step": 101213, "epoch": 1137, "lr": 6.961103431480163e-05}
{"train_loss": 0.14639875292778015, "global_step": 101214, "epoch": 1137, "lr": 6.96105010305072e-05}
{"train_loss": 0.17060057818889618, "global_step": 101215, "epoch": 1137, "lr": 6.960996774357635e-05}
{"train_loss": 0.12614741921424866, "global_step": 101216, "epoch": 1137, "lr": 6.960943445400918e-05}
{"train_loss": 0.1047888994216919, "global_step": 101217, "epoch": 1137, "lr": 6.960890116180577e-05}
{"train_loss": 0.1597970426082611, "global_step": 101218, "epoch": 1137, "lr": 6.960836786696618e-05}
{"train_loss": 0.09762564301490784, "global_step": 101219, "epoch": 1137, "lr": 6.960783456949049e-05}
{"train_loss": 0.16132968664169312, "global_step": 101220, "epoch": 1137, "lr": 6.960730126937876e-05}
{"train_loss": 0.19219034910202026, "global_step": 101221, "epoch": 1137, "lr": 6.960676796663105e-05}
{"train_loss": 0.17879748344421387, "global_step": 101222, "epoch": 1137, "lr": 6.960623466124748e-05}
{"train_loss": 0.15935349464416504, "global_step": 101223, "epoch": 1137, "lr": 6.960570135322807e-05}
{"train_loss": 0.2058347463607788, "global_step": 101224, "epoch": 1137, "lr": 6.960516804257291e-05}
{"train_loss": 0.20816351473331451, "global_step": 101225, "epoch": 1137, "lr": 6.96046347292821e-05}
{"train_loss": 0.14262545108795166, "global_step": 101226, "epoch": 1137, "lr": 6.960410141335567e-05}
{"train_loss": 0.25744935870170593, "global_step": 101227, "epoch": 1137, "lr": 6.960356809479369e-05}
{"train_loss": 0.09269335120916367, "global_step": 101228, "epoch": 1137, "lr": 6.960303477359629e-05}
{"train_loss": 0.17323242127895355, "global_step": 101229, "epoch": 1137, "lr": 6.960250144976347e-05}
{"train_loss": 0.20273582637310028, "global_step": 101230, "epoch": 1137, "lr": 6.960196812329533e-05}
{"train_loss": 0.18574190139770508, "global_step": 101231, "epoch": 1137, "lr": 6.960143479419197e-05}
{"train_loss": 0.15738065540790558, "global_step": 101232, "epoch": 1137, "lr": 6.960090146245342e-05}
{"train_loss": 0.17517822980880737, "global_step": 101233, "epoch": 1137, "lr": 6.960036812807976e-05}
{"train_loss": 0.15109558403491974, "global_step": 101234, "epoch": 1137, "lr": 6.959983479107108e-05}
{"train_loss": 0.21588195860385895, "global_step": 101235, "epoch": 1137, "lr": 6.959930145142743e-05}
{"train_loss": 0.16078925132751465, "global_step": 101236, "epoch": 1137, "lr": 6.959876810914891e-05}
{"train_loss": 0.24028077721595764, "global_step": 101237, "epoch": 1137, "lr": 6.959823476423556e-05}
{"train_loss": 0.17161524295806885, "global_step": 101238, "epoch": 1137, "lr": 6.959770141668747e-05}
{"train_loss": 0.12617960572242737, "global_step": 101239, "epoch": 1137, "lr": 6.95971680665047e-05}
{"train_loss": 0.16969117522239685, "global_step": 101240, "epoch": 1137, "lr": 6.959663471368733e-05}
{"train_loss": 0.213470458984375, "global_step": 101241, "epoch": 1137, "lr": 6.959610135823545e-05}
{"train_loss": 0.1941414624452591, "global_step": 101242, "epoch": 1137, "lr": 6.95955680001491e-05}
{"train_loss": 0.11882440745830536, "global_step": 101243, "epoch": 1137, "lr": 6.959503463942837e-05}
{"train_loss": 0.1304018497467041, "global_step": 101244, "epoch": 1137, "lr": 6.959450127607331e-05}
{"train_loss": 0.21751311421394348, "global_step": 101245, "epoch": 1137, "lr": 6.959396791008404e-05}
{"train_loss": 0.20631557703018188, "global_step": 101246, "epoch": 1137, "lr": 6.959343454146058e-05}
{"train_loss": 0.14929109811782837, "global_step": 101247, "epoch": 1137, "lr": 6.959290117020303e-05}
{"train_loss": 0.24675628542900085, "global_step": 101248, "epoch": 1137, "lr": 6.959236779631145e-05}
{"train_loss": 0.25516775250434875, "global_step": 101249, "epoch": 1137, "lr": 6.959183441978592e-05}
{"train_loss": 0.24926018714904785, "global_step": 101250, "epoch": 1137, "lr": 6.95913010406265e-05}
{"train_loss": 0.1542462557554245, "global_step": 101251, "epoch": 1137, "lr": 6.959076765883326e-05}
{"train_loss": 0.22915774583816528, "global_step": 101252, "epoch": 1137, "lr": 6.959023427440629e-05}
{"train_loss": 0.20063768327236176, "global_step": 101253, "epoch": 1137, "lr": 6.958970088734564e-05}
{"train_loss": 0.1760587841272354, "global_step": 101254, "epoch": 1137, "lr": 6.958916749765141e-05}
{"train_loss": 0.20613431930541992, "global_step": 101255, "epoch": 1137, "lr": 6.958863410532364e-05}
{"train_loss": 0.2913036048412323, "global_step": 101256, "epoch": 1137, "lr": 6.958810071036244e-05}
{"train_loss": 0.19511447846889496, "global_step": 101257, "epoch": 1137, "lr": 6.958756731276783e-05}
{"train_loss": 0.12424733489751816, "global_step": 101258, "epoch": 1137, "lr": 6.958703391253991e-05}
{"train_loss": 0.20278970897197723, "global_step": 101259, "epoch": 1137, "lr": 6.958650050967878e-05}
{"train_loss": 0.27421262860298157, "global_step": 101260, "epoch": 1137, "lr": 6.958596710418446e-05}
{"train_loss": 0.10351832956075668, "global_step": 101261, "epoch": 1137, "lr": 6.958543369605706e-05}
{"train_loss": 0.24084095656871796, "global_step": 101262, "epoch": 1137, "lr": 6.958490028529663e-05}
{"train_loss": 0.1148933619260788, "global_step": 101263, "epoch": 1137, "lr": 6.958436687190325e-05}
{"train_loss": 0.13606970012187958, "global_step": 101264, "epoch": 1137, "lr": 6.958383345587699e-05}
{"train_loss": 0.17686666548252106, "global_step": 101265, "epoch": 1137, "lr": 6.958330003721792e-05}
{"train_loss": 0.17612367868423462, "global_step": 101266, "epoch": 1137, "lr": 6.958276661592613e-05}
{"train_loss": 0.25542765855789185, "global_step": 101267, "epoch": 1137, "lr": 6.958223319200166e-05}
{"train_loss": 0.17662906646728516, "global_step": 101268, "epoch": 1137, "lr": 6.95816997654446e-05}
{"train_loss": 0.1542113572359085, "global_step": 101269, "epoch": 1137, "lr": 6.958116633625501e-05}
{"train_loss": 0.1584186851978302, "global_step": 101270, "epoch": 1137, "lr": 6.9580632904433e-05}
{"train_loss": 0.22868485748767853, "global_step": 101271, "epoch": 1137, "lr": 6.95800994699786e-05}
{"train_loss": 0.2706640064716339, "global_step": 101272, "epoch": 1137, "lr": 6.957956603289187e-05}
{"train_loss": 0.18222980201244354, "global_step": 101273, "epoch": 1137, "lr": 6.957903259317293e-05}
{"train_loss": 0.19193851947784424, "global_step": 101274, "epoch": 1137, "lr": 6.957849915082182e-05}
{"train_loss": 0.21587307751178741, "global_step": 101275, "epoch": 1137, "lr": 6.957796570583862e-05}
{"train_loss": 0.17657256126403809, "global_step": 101276, "epoch": 1137, "lr": 6.95774322582234e-05}
{"train_loss": 0.2106115072965622, "global_step": 101277, "epoch": 1137, "lr": 6.957689880797623e-05}
{"train_loss": 0.21780355274677277, "global_step": 101278, "epoch": 1137, "lr": 6.957636535509718e-05}
{"train_loss": 0.16206610202789307, "global_step": 101279, "epoch": 1137, "lr": 6.957583189958634e-05}
{"train_loss": 0.18721668422222137, "global_step": 101280, "epoch": 1137, "lr": 6.957529844144377e-05}
{"train_loss": 0.1786674466193392, "global_step": 101281, "epoch": 1137, "lr": 6.957476498066953e-05, "val_loss": 4.5329508781433105}
{"train_loss": 0.16657961905002594, "global_step": 101282, "epoch": 1138, "lr": 6.95742315172637e-05}
{"train_loss": 0.18685543537139893, "global_step": 101283, "epoch": 1138, "lr": 6.957369805122635e-05}
{"train_loss": 0.17106452584266663, "global_step": 101284, "epoch": 1138, "lr": 6.957316458255758e-05}
{"train_loss": 0.20592591166496277, "global_step": 101285, "epoch": 1138, "lr": 6.957263111125742e-05}
{"train_loss": 0.13434404134750366, "global_step": 101286, "epoch": 1138, "lr": 6.957209763732596e-05}
{"train_loss": 0.12478159368038177, "global_step": 101287, "epoch": 1138, "lr": 6.957156416076327e-05}
{"train_loss": 0.13953888416290283, "global_step": 101288, "epoch": 1138, "lr": 6.957103068156942e-05}
{"train_loss": 0.17721883952617645, "global_step": 101289, "epoch": 1138, "lr": 6.957049719974448e-05}
{"train_loss": 0.14330458641052246, "global_step": 101290, "epoch": 1138, "lr": 6.956996371528852e-05}
{"train_loss": 0.12971070408821106, "global_step": 101291, "epoch": 1138, "lr": 6.956943022820164e-05}
{"train_loss": 0.15279115736484528, "global_step": 101292, "epoch": 1138, "lr": 6.956889673848387e-05}
{"train_loss": 0.15449704229831696, "global_step": 101293, "epoch": 1138, "lr": 6.95683632461353e-05}
{"train_loss": 0.16271066665649414, "global_step": 101294, "epoch": 1138, "lr": 6.9567829751156e-05}
{"train_loss": 0.17107830941677094, "global_step": 101295, "epoch": 1138, "lr": 6.956729625354605e-05}
{"train_loss": 0.13781331479549408, "global_step": 101296, "epoch": 1138, "lr": 6.956676275330551e-05}
{"train_loss": 0.162991464138031, "global_step": 101297, "epoch": 1138, "lr": 6.956622925043446e-05}
{"train_loss": 0.2283749133348465, "global_step": 101298, "epoch": 1138, "lr": 6.956569574493297e-05}
{"train_loss": 0.10147733241319656, "global_step": 101299, "epoch": 1138, "lr": 6.956516223680112e-05}
{"train_loss": 0.20629660785198212, "global_step": 101300, "epoch": 1138, "lr": 6.956462872603896e-05}
{"train_loss": 0.11855648458003998, "global_step": 101301, "epoch": 1138, "lr": 6.956409521264658e-05}
{"train_loss": 0.21885842084884644, "global_step": 101302, "epoch": 1138, "lr": 6.956356169662405e-05}
{"train_loss": 0.1980687528848648, "global_step": 101303, "epoch": 1138, "lr": 6.956302817797143e-05}
{"train_loss": 0.1735360324382782, "global_step": 101304, "epoch": 1138, "lr": 6.95624946566888e-05}
{"train_loss": 0.12469594180583954, "global_step": 101305, "epoch": 1138, "lr": 6.956196113277624e-05}
{"train_loss": 0.11785836517810822, "global_step": 101306, "epoch": 1138, "lr": 6.956142760623381e-05}
{"train_loss": 0.12185697257518768, "global_step": 101307, "epoch": 1138, "lr": 6.956089407706156e-05}
{"train_loss": 0.17313669621944427, "global_step": 101308, "epoch": 1138, "lr": 6.956036054525961e-05}
{"train_loss": 0.1849321573972702, "global_step": 101309, "epoch": 1138, "lr": 6.955982701082801e-05}
{"train_loss": 0.148299440741539, "global_step": 101310, "epoch": 1138, "lr": 6.955929347376683e-05}
{"train_loss": 0.13162881135940552, "global_step": 101311, "epoch": 1138, "lr": 6.955875993407613e-05}
{"train_loss": 0.09231195598840714, "global_step": 101312, "epoch": 1138, "lr": 6.955822639175598e-05}
{"train_loss": 0.1693425178527832, "global_step": 101313, "epoch": 1138, "lr": 6.955769284680648e-05}
{"train_loss": 0.12636657059192657, "global_step": 101314, "epoch": 1138, "lr": 6.95571592992277e-05}
{"train_loss": 0.217803493142128, "global_step": 101315, "epoch": 1138, "lr": 6.95566257490197e-05}
{"train_loss": 0.17722608149051666, "global_step": 101316, "epoch": 1138, "lr": 6.955609219618253e-05}
{"train_loss": 0.16727283596992493, "global_step": 101317, "epoch": 1138, "lr": 6.955555864071628e-05}
{"train_loss": 0.1389111429452896, "global_step": 101318, "epoch": 1138, "lr": 6.955502508262103e-05}
{"train_loss": 0.12843593955039978, "global_step": 101319, "epoch": 1138, "lr": 6.955449152189687e-05}
{"train_loss": 0.2150183767080307, "global_step": 101320, "epoch": 1138, "lr": 6.955395795854382e-05}
{"train_loss": 0.19506530463695526, "global_step": 101321, "epoch": 1138, "lr": 6.955342439256198e-05}
{"train_loss": 0.11796614527702332, "global_step": 101322, "epoch": 1138, "lr": 6.955289082395142e-05}
{"train_loss": 0.19124050438404083, "global_step": 101323, "epoch": 1138, "lr": 6.955235725271221e-05}
{"train_loss": 0.17672204971313477, "global_step": 101324, "epoch": 1138, "lr": 6.955182367884444e-05}
{"train_loss": 0.17578333616256714, "global_step": 101325, "epoch": 1138, "lr": 6.955129010234815e-05}
{"train_loss": 0.1596873700618744, "global_step": 101326, "epoch": 1138, "lr": 6.955075652322344e-05}
{"train_loss": 0.15176565945148468, "global_step": 101327, "epoch": 1138, "lr": 6.955022294147036e-05}
{"train_loss": 0.15768557786941528, "global_step": 101328, "epoch": 1138, "lr": 6.9549689357089e-05}
{"train_loss": 0.15969015657901764, "global_step": 101329, "epoch": 1138, "lr": 6.95491557700794e-05}
{"train_loss": 0.11481291800737381, "global_step": 101330, "epoch": 1138, "lr": 6.954862218044167e-05}
{"train_loss": 0.13169409334659576, "global_step": 101331, "epoch": 1138, "lr": 6.954808858817588e-05}
{"train_loss": 0.18124854564666748, "global_step": 101332, "epoch": 1138, "lr": 6.954755499328206e-05}
{"train_loss": 0.14188626408576965, "global_step": 101333, "epoch": 1138, "lr": 6.954702139576033e-05}
{"train_loss": 0.13213396072387695, "global_step": 101334, "epoch": 1138, "lr": 6.954648779561072e-05}
{"train_loss": 0.19086016714572906, "global_step": 101335, "epoch": 1138, "lr": 6.954595419283335e-05}
{"train_loss": 0.18501615524291992, "global_step": 101336, "epoch": 1138, "lr": 6.954542058742826e-05}
{"train_loss": 0.13556832075119019, "global_step": 101337, "epoch": 1138, "lr": 6.954488697939551e-05}
{"train_loss": 0.16895554959774017, "global_step": 101338, "epoch": 1138, "lr": 6.954435336873519e-05}
{"train_loss": 0.1887369453907013, "global_step": 101339, "epoch": 1138, "lr": 6.954381975544738e-05}
{"train_loss": 0.15081673860549927, "global_step": 101340, "epoch": 1138, "lr": 6.954328613953213e-05}
{"train_loss": 0.1498151570558548, "global_step": 101341, "epoch": 1138, "lr": 6.954275252098954e-05}
{"train_loss": 0.17749544978141785, "global_step": 101342, "epoch": 1138, "lr": 6.954221889981965e-05}
{"train_loss": 0.2099129855632782, "global_step": 101343, "epoch": 1138, "lr": 6.954168527602256e-05}
{"train_loss": 0.12761755287647247, "global_step": 101344, "epoch": 1138, "lr": 6.954115164959833e-05}
{"train_loss": 0.11788935214281082, "global_step": 101345, "epoch": 1138, "lr": 6.954061802054703e-05}
{"train_loss": 0.23730605840682983, "global_step": 101346, "epoch": 1138, "lr": 6.954008438886872e-05}
{"train_loss": 0.14942823350429535, "global_step": 101347, "epoch": 1138, "lr": 6.953955075456348e-05}
{"train_loss": 0.14110155403614044, "global_step": 101348, "epoch": 1138, "lr": 6.95390171176314e-05}
{"train_loss": 0.13105422258377075, "global_step": 101349, "epoch": 1138, "lr": 6.953848347807253e-05}
{"train_loss": 0.143008753657341, "global_step": 101350, "epoch": 1138, "lr": 6.953794983588694e-05}
{"train_loss": 0.12830068171024323, "global_step": 101351, "epoch": 1138, "lr": 6.953741619107472e-05}
{"train_loss": 0.13184385001659393, "global_step": 101352, "epoch": 1138, "lr": 6.953688254363593e-05}
{"train_loss": 0.1626170575618744, "global_step": 101353, "epoch": 1138, "lr": 6.953634889357066e-05}
{"train_loss": 0.13665658235549927, "global_step": 101354, "epoch": 1138, "lr": 6.953581524087895e-05}
{"train_loss": 0.16815751791000366, "global_step": 101355, "epoch": 1138, "lr": 6.953528158556089e-05}
{"train_loss": 0.14197342097759247, "global_step": 101356, "epoch": 1138, "lr": 6.953474792761656e-05}
{"train_loss": 0.18426358699798584, "global_step": 101357, "epoch": 1138, "lr": 6.9534214267046e-05}
{"train_loss": 0.15719915926456451, "global_step": 101358, "epoch": 1138, "lr": 6.953368060384932e-05}
{"train_loss": 0.15575328469276428, "global_step": 101359, "epoch": 1138, "lr": 6.953314693802657e-05}
{"train_loss": 0.17285306751728058, "global_step": 101360, "epoch": 1138, "lr": 6.953261326957782e-05}
{"train_loss": 0.17898985743522644, "global_step": 101361, "epoch": 1138, "lr": 6.953207959850315e-05}
{"train_loss": 0.237276092171669, "global_step": 101362, "epoch": 1138, "lr": 6.953154592480265e-05}
{"train_loss": 0.1523597687482834, "global_step": 101363, "epoch": 1138, "lr": 6.953101224847636e-05}
{"train_loss": 0.17990583181381226, "global_step": 101364, "epoch": 1138, "lr": 6.953047856952436e-05}
{"train_loss": 0.12643514573574066, "global_step": 101365, "epoch": 1138, "lr": 6.952994488794671e-05}
{"train_loss": 0.15817418694496155, "global_step": 101366, "epoch": 1138, "lr": 6.95294112037435e-05}
{"train_loss": 0.12796972692012787, "global_step": 101367, "epoch": 1138, "lr": 6.952887751691482e-05}
{"train_loss": 0.11200670897960663, "global_step": 101368, "epoch": 1138, "lr": 6.952834382746071e-05}
{"train_loss": 0.163467675447464, "global_step": 101369, "epoch": 1138, "lr": 6.952781013538125e-05}
{"train_loss": 0.15834906965159298, "global_step": 101370, "epoch": 1138, "lr": 6.952727644067653e-05, "val_loss": 4.601349830627441}
{"train_loss": 0.20988090336322784, "global_step": 101371, "epoch": 1139, "lr": 6.952674274334659e-05}
{"train_loss": 0.17093926668167114, "global_step": 101372, "epoch": 1139, "lr": 6.952620904339151e-05}
{"train_loss": 0.11292943358421326, "global_step": 101373, "epoch": 1139, "lr": 6.952567534081139e-05}
{"train_loss": 0.12842121720314026, "global_step": 101374, "epoch": 1139, "lr": 6.952514163560626e-05}
{"train_loss": 0.10732225328683853, "global_step": 101375, "epoch": 1139, "lr": 6.952460792777623e-05}
{"train_loss": 0.13542625308036804, "global_step": 101376, "epoch": 1139, "lr": 6.952407421732134e-05}
{"train_loss": 0.19643498957157135, "global_step": 101377, "epoch": 1139, "lr": 6.95235405042417e-05}
{"train_loss": 0.122499980032444, "global_step": 101378, "epoch": 1139, "lr": 6.952300678853733e-05}
{"train_loss": 0.18882520496845245, "global_step": 101379, "epoch": 1139, "lr": 6.952247307020834e-05}
{"train_loss": 0.08716684579849243, "global_step": 101380, "epoch": 1139, "lr": 6.95219393492548e-05}
{"train_loss": 0.12886619567871094, "global_step": 101381, "epoch": 1139, "lr": 6.952140562567677e-05}
{"train_loss": 0.1897861659526825, "global_step": 101382, "epoch": 1139, "lr": 6.952087189947432e-05}
{"train_loss": 0.16811588406562805, "global_step": 101383, "epoch": 1139, "lr": 6.952033817064754e-05}
{"train_loss": 0.10701107233762741, "global_step": 101384, "epoch": 1139, "lr": 6.951980443919647e-05}
{"train_loss": 0.17994600534439087, "global_step": 101385, "epoch": 1139, "lr": 6.951927070512118e-05}
{"train_loss": 0.14626258611679077, "global_step": 101386, "epoch": 1139, "lr": 6.951873696842179e-05}
{"train_loss": 0.12018303573131561, "global_step": 101387, "epoch": 1139, "lr": 6.951820322909835e-05}
{"train_loss": 0.16658206284046173, "global_step": 101388, "epoch": 1139, "lr": 6.951766948715092e-05}
{"train_loss": 0.20125895738601685, "global_step": 101389, "epoch": 1139, "lr": 6.951713574257958e-05}
{"train_loss": 0.19531799852848053, "global_step": 101390, "epoch": 1139, "lr": 6.951660199538439e-05}
{"train_loss": 0.15018004179000854, "global_step": 101391, "epoch": 1139, "lr": 6.951606824556542e-05}
{"train_loss": 0.0879087969660759, "global_step": 101392, "epoch": 1139, "lr": 6.951553449312278e-05}
{"train_loss": 0.2846481502056122, "global_step": 101393, "epoch": 1139, "lr": 6.951500073805651e-05}
{"train_loss": 0.1168026477098465, "global_step": 101394, "epoch": 1139, "lr": 6.951446698036666e-05}
{"train_loss": 0.13118626177310944, "global_step": 101395, "epoch": 1139, "lr": 6.951393322005335e-05}
{"train_loss": 0.18166112899780273, "global_step": 101396, "epoch": 1139, "lr": 6.951339945711662e-05}
{"train_loss": 0.21037942171096802, "global_step": 101397, "epoch": 1139, "lr": 6.951286569155656e-05}
{"train_loss": 0.25083497166633606, "global_step": 101398, "epoch": 1139, "lr": 6.951233192337323e-05}
{"train_loss": 0.16955435276031494, "global_step": 101399, "epoch": 1139, "lr": 6.95117981525667e-05}
{"train_loss": 0.1559067964553833, "global_step": 101400, "epoch": 1139, "lr": 6.951126437913705e-05}
{"train_loss": 0.2141721099615097, "global_step": 101401, "epoch": 1139, "lr": 6.951073060308436e-05}
{"train_loss": 0.08548585325479507, "global_step": 101402, "epoch": 1139, "lr": 6.951019682440868e-05}
{"train_loss": 0.12686020135879517, "global_step": 101403, "epoch": 1139, "lr": 6.950966304311008e-05}
{"train_loss": 0.12344563007354736, "global_step": 101404, "epoch": 1139, "lr": 6.950912925918865e-05}
{"train_loss": 0.09889895468950272, "global_step": 101405, "epoch": 1139, "lr": 6.950859547264446e-05}
{"train_loss": 0.16462714970111847, "global_step": 101406, "epoch": 1139, "lr": 6.950806168347758e-05}
{"train_loss": 0.2230783849954605, "global_step": 101407, "epoch": 1139, "lr": 6.950752789168807e-05}
{"train_loss": 0.11669047176837921, "global_step": 101408, "epoch": 1139, "lr": 6.950699409727601e-05}
{"train_loss": 0.15006953477859497, "global_step": 101409, "epoch": 1139, "lr": 6.950646030024148e-05}
{"train_loss": 0.14333970844745636, "global_step": 101410, "epoch": 1139, "lr": 6.950592650058455e-05}
{"train_loss": 0.19609007239341736, "global_step": 101411, "epoch": 1139, "lr": 6.950539269830528e-05}
{"train_loss": 0.13564461469650269, "global_step": 101412, "epoch": 1139, "lr": 6.950485889340375e-05}
{"train_loss": 0.11570734530687332, "global_step": 101413, "epoch": 1139, "lr": 6.950432508588001e-05}
{"train_loss": 0.1945493370294571, "global_step": 101414, "epoch": 1139, "lr": 6.950379127573417e-05}
{"train_loss": 0.2368880808353424, "global_step": 101415, "epoch": 1139, "lr": 6.950325746296629e-05}
{"train_loss": 0.16300641000270844, "global_step": 101416, "epoch": 1139, "lr": 6.950272364757642e-05}
{"train_loss": 0.12724706530570984, "global_step": 101417, "epoch": 1139, "lr": 6.950218982956465e-05}
{"train_loss": 0.2375916838645935, "global_step": 101418, "epoch": 1139, "lr": 6.950165600893105e-05}
{"train_loss": 0.20779812335968018, "global_step": 101419, "epoch": 1139, "lr": 6.950112218567568e-05}
{"train_loss": 0.1417064219713211, "global_step": 101420, "epoch": 1139, "lr": 6.950058835979863e-05}
{"train_loss": 0.17054159939289093, "global_step": 101421, "epoch": 1139, "lr": 6.950005453129998e-05}
{"train_loss": 0.24220778048038483, "global_step": 101422, "epoch": 1139, "lr": 6.949952070017977e-05}
{"train_loss": 0.1748296171426773, "global_step": 101423, "epoch": 1139, "lr": 6.949898686643807e-05}
{"train_loss": 0.14134401082992554, "global_step": 101424, "epoch": 1139, "lr": 6.949845303007499e-05}
{"train_loss": 0.12780210375785828, "global_step": 101425, "epoch": 1139, "lr": 6.949791919109058e-05}
{"train_loss": 0.13941732048988342, "global_step": 101426, "epoch": 1139, "lr": 6.949738534948491e-05}
{"train_loss": 0.14753147959709167, "global_step": 101427, "epoch": 1139, "lr": 6.949685150525805e-05}
{"train_loss": 0.13464896380901337, "global_step": 101428, "epoch": 1139, "lr": 6.949631765841009e-05}
{"train_loss": 0.1985434889793396, "global_step": 101429, "epoch": 1139, "lr": 6.949578380894108e-05}
{"train_loss": 0.14959339797496796, "global_step": 101430, "epoch": 1139, "lr": 6.949524995685109e-05}
{"train_loss": 0.2711856961250305, "global_step": 101431, "epoch": 1139, "lr": 6.94947161021402e-05}
{"train_loss": 0.11864175647497177, "global_step": 101432, "epoch": 1139, "lr": 6.949418224480849e-05}
{"train_loss": 0.17413410544395447, "global_step": 101433, "epoch": 1139, "lr": 6.949364838485603e-05}
{"train_loss": 0.23388709127902985, "global_step": 101434, "epoch": 1139, "lr": 6.949311452228289e-05}
{"train_loss": 0.1445891261100769, "global_step": 101435, "epoch": 1139, "lr": 6.949258065708912e-05}
{"train_loss": 0.1457551270723343, "global_step": 101436, "epoch": 1139, "lr": 6.949204678927482e-05}
{"train_loss": 0.2355199009180069, "global_step": 101437, "epoch": 1139, "lr": 6.949151291884005e-05}
{"train_loss": 0.11811459809541702, "global_step": 101438, "epoch": 1139, "lr": 6.949097904578489e-05}
{"train_loss": 0.18875689804553986, "global_step": 101439, "epoch": 1139, "lr": 6.94904451701094e-05}
{"train_loss": 0.12661662697792053, "global_step": 101440, "epoch": 1139, "lr": 6.948991129181366e-05}
{"train_loss": 0.20777976512908936, "global_step": 101441, "epoch": 1139, "lr": 6.948937741089774e-05}
{"train_loss": 0.2185228168964386, "global_step": 101442, "epoch": 1139, "lr": 6.94888435273617e-05}
{"train_loss": 0.15254047513008118, "global_step": 101443, "epoch": 1139, "lr": 6.948830964120563e-05}
{"train_loss": 0.1848609447479248, "global_step": 101444, "epoch": 1139, "lr": 6.94877757524296e-05}
{"train_loss": 0.20556285977363586, "global_step": 101445, "epoch": 1139, "lr": 6.948724186103366e-05}
{"train_loss": 0.21279476583003998, "global_step": 101446, "epoch": 1139, "lr": 6.948670796701791e-05}
{"train_loss": 0.15920181572437286, "global_step": 101447, "epoch": 1139, "lr": 6.948617407038241e-05}
{"train_loss": 0.1408156305551529, "global_step": 101448, "epoch": 1139, "lr": 6.948564017112724e-05}
{"train_loss": 0.2793979048728943, "global_step": 101449, "epoch": 1139, "lr": 6.948510626925245e-05}
{"train_loss": 0.16672347486019135, "global_step": 101450, "epoch": 1139, "lr": 6.948457236475813e-05}
{"train_loss": 0.21260270476341248, "global_step": 101451, "epoch": 1139, "lr": 6.948403845764435e-05}
{"train_loss": 0.23371872305870056, "global_step": 101452, "epoch": 1139, "lr": 6.948350454791117e-05}
{"train_loss": 0.1929982602596283, "global_step": 101453, "epoch": 1139, "lr": 6.948297063555867e-05}
{"train_loss": 0.1594773530960083, "global_step": 101454, "epoch": 1139, "lr": 6.948243672058694e-05}
{"train_loss": 0.18223722279071808, "global_step": 101455, "epoch": 1139, "lr": 6.9481902802996e-05}
{"train_loss": 0.15082675218582153, "global_step": 101456, "epoch": 1139, "lr": 6.948136888278598e-05}
{"train_loss": 0.20863372087478638, "global_step": 101457, "epoch": 1139, "lr": 6.948083495995692e-05}
{"train_loss": 0.22234173119068146, "global_step": 101458, "epoch": 1139, "lr": 6.94803010345089e-05}
{"train_loss": 0.16854874119999702, "global_step": 101459, "epoch": 1139, "lr": 6.9479767106442e-05, "val_loss": 4.807707786560059}
{"train_loss": 0.23704804480075836, "global_step": 101460, "epoch": 1140, "lr": 6.947923317575628e-05}
{"train_loss": 0.17279991507530212, "global_step": 101461, "epoch": 1140, "lr": 6.947869924245181e-05}
{"train_loss": 0.17992447316646576, "global_step": 101462, "epoch": 1140, "lr": 6.947816530652867e-05}
{"train_loss": 0.15589047968387604, "global_step": 101463, "epoch": 1140, "lr": 6.947763136798692e-05}
{"train_loss": 0.12263848632574081, "global_step": 101464, "epoch": 1140, "lr": 6.947709742682664e-05}
{"train_loss": 0.11899464577436447, "global_step": 101465, "epoch": 1140, "lr": 6.947656348304791e-05}
{"train_loss": 0.11161147803068161, "global_step": 101466, "epoch": 1140, "lr": 6.947602953665078e-05}
{"train_loss": 0.13106325268745422, "global_step": 101467, "epoch": 1140, "lr": 6.947549558763535e-05}
{"train_loss": 0.18321464955806732, "global_step": 101468, "epoch": 1140, "lr": 6.947496163600167e-05}
{"train_loss": 0.1276036947965622, "global_step": 101469, "epoch": 1140, "lr": 6.947442768174982e-05}
{"train_loss": 0.13572487235069275, "global_step": 101470, "epoch": 1140, "lr": 6.947389372487987e-05}
{"train_loss": 0.314716100692749, "global_step": 101471, "epoch": 1140, "lr": 6.947335976539189e-05}
{"train_loss": 0.1073271781206131, "global_step": 101472, "epoch": 1140, "lr": 6.947282580328596e-05}
{"train_loss": 0.17585356533527374, "global_step": 101473, "epoch": 1140, "lr": 6.947229183856215e-05}
{"train_loss": 0.19529208540916443, "global_step": 101474, "epoch": 1140, "lr": 6.94717578712205e-05}
{"train_loss": 0.13480640947818756, "global_step": 101475, "epoch": 1140, "lr": 6.947122390126115e-05}
{"train_loss": 0.19656722247600555, "global_step": 101476, "epoch": 1140, "lr": 6.947068992868411e-05}
{"train_loss": 0.18568289279937744, "global_step": 101477, "epoch": 1140, "lr": 6.947015595348947e-05}
{"train_loss": 0.11294866353273392, "global_step": 101478, "epoch": 1140, "lr": 6.946962197567731e-05}
{"train_loss": 0.18600711226463318, "global_step": 101479, "epoch": 1140, "lr": 6.94690879952477e-05}
{"train_loss": 0.20287998020648956, "global_step": 101480, "epoch": 1140, "lr": 6.94685540122007e-05}
{"train_loss": 0.20230470597743988, "global_step": 101481, "epoch": 1140, "lr": 6.94680200265364e-05}
{"train_loss": 0.1444217562675476, "global_step": 101482, "epoch": 1140, "lr": 6.946748603825485e-05}
{"train_loss": 0.16436131298542023, "global_step": 101483, "epoch": 1140, "lr": 6.946695204735614e-05}
{"train_loss": 0.14794772863388062, "global_step": 101484, "epoch": 1140, "lr": 6.946641805384034e-05}
{"train_loss": 0.14461632072925568, "global_step": 101485, "epoch": 1140, "lr": 6.94658840577075e-05}
{"train_loss": 0.12100912630558014, "global_step": 101486, "epoch": 1140, "lr": 6.946535005895773e-05}
{"train_loss": 0.21102668344974518, "global_step": 101487, "epoch": 1140, "lr": 6.946481605759107e-05}
{"train_loss": 0.17022448778152466, "global_step": 101488, "epoch": 1140, "lr": 6.946428205360763e-05}
{"train_loss": 0.20668010413646698, "global_step": 101489, "epoch": 1140, "lr": 6.946374804700742e-05}
{"train_loss": 0.14131686091423035, "global_step": 101490, "epoch": 1140, "lr": 6.946321403779055e-05}
{"train_loss": 0.20618849992752075, "global_step": 101491, "epoch": 1140, "lr": 6.946268002595709e-05}
{"train_loss": 0.24370524287223816, "global_step": 101492, "epoch": 1140, "lr": 6.946214601150712e-05}
{"train_loss": 0.1744072288274765, "global_step": 101493, "epoch": 1140, "lr": 6.94616119944407e-05}
{"train_loss": 0.13157528638839722, "global_step": 101494, "epoch": 1140, "lr": 6.94610779747579e-05}
{"train_loss": 0.17709267139434814, "global_step": 101495, "epoch": 1140, "lr": 6.94605439524588e-05}
{"train_loss": 0.13853561878204346, "global_step": 101496, "epoch": 1140, "lr": 6.946000992754346e-05}
{"train_loss": 0.1776120662689209, "global_step": 101497, "epoch": 1140, "lr": 6.945947590001196e-05}
{"train_loss": 0.1364729404449463, "global_step": 101498, "epoch": 1140, "lr": 6.945894186986437e-05}
{"train_loss": 0.20879678428173065, "global_step": 101499, "epoch": 1140, "lr": 6.945840783710077e-05}
{"train_loss": 0.1819770187139511, "global_step": 101500, "epoch": 1140, "lr": 6.945787380172122e-05}
{"train_loss": 0.21451415121555328, "global_step": 101501, "epoch": 1140, "lr": 6.945733976372579e-05}
{"train_loss": 0.13934646546840668, "global_step": 101502, "epoch": 1140, "lr": 6.945680572311456e-05}
{"train_loss": 0.20280134677886963, "global_step": 101503, "epoch": 1140, "lr": 6.945627167988759e-05}
{"train_loss": 0.1995740532875061, "global_step": 101504, "epoch": 1140, "lr": 6.945573763404498e-05}
{"train_loss": 0.22410249710083008, "global_step": 101505, "epoch": 1140, "lr": 6.945520358558678e-05}
{"train_loss": 0.11126799881458282, "global_step": 101506, "epoch": 1140, "lr": 6.945466953451307e-05}
{"train_loss": 0.21822524070739746, "global_step": 101507, "epoch": 1140, "lr": 6.94541354808239e-05}
{"train_loss": 0.07195623219013214, "global_step": 101508, "epoch": 1140, "lr": 6.945360142451937e-05}
{"train_loss": 0.15703144669532776, "global_step": 101509, "epoch": 1140, "lr": 6.945306736559954e-05}
{"train_loss": 0.18606820702552795, "global_step": 101510, "epoch": 1140, "lr": 6.945253330406448e-05}
{"train_loss": 0.11604864150285721, "global_step": 101511, "epoch": 1140, "lr": 6.945199923991425e-05}
{"train_loss": 0.15638026595115662, "global_step": 101512, "epoch": 1140, "lr": 6.945146517314895e-05}
{"train_loss": 0.11618733406066895, "global_step": 101513, "epoch": 1140, "lr": 6.945093110376865e-05}
{"train_loss": 0.20833227038383484, "global_step": 101514, "epoch": 1140, "lr": 6.945039703177339e-05}
{"train_loss": 0.21935443580150604, "global_step": 101515, "epoch": 1140, "lr": 6.944986295716327e-05}
{"train_loss": 0.16231386363506317, "global_step": 101516, "epoch": 1140, "lr": 6.944932887993836e-05}
{"train_loss": 0.1444794088602066, "global_step": 101517, "epoch": 1140, "lr": 6.944879480009872e-05}
{"train_loss": 0.17795433104038239, "global_step": 101518, "epoch": 1140, "lr": 6.94482607176444e-05}
{"train_loss": 0.1074116975069046, "global_step": 101519, "epoch": 1140, "lr": 6.944772663257552e-05}
{"train_loss": 0.11787184327840805, "global_step": 101520, "epoch": 1140, "lr": 6.944719254489214e-05}
{"train_loss": 0.17286935448646545, "global_step": 101521, "epoch": 1140, "lr": 6.944665845459431e-05}
{"train_loss": 0.2303498089313507, "global_step": 101522, "epoch": 1140, "lr": 6.944612436168212e-05}
{"train_loss": 0.1689915955066681, "global_step": 101523, "epoch": 1140, "lr": 6.944559026615564e-05}
{"train_loss": 0.1596159040927887, "global_step": 101524, "epoch": 1140, "lr": 6.944505616801492e-05}
{"train_loss": 0.1518966257572174, "global_step": 101525, "epoch": 1140, "lr": 6.944452206726006e-05}
{"train_loss": 0.12824328243732452, "global_step": 101526, "epoch": 1140, "lr": 6.944398796389112e-05}
{"train_loss": 0.14041662216186523, "global_step": 101527, "epoch": 1140, "lr": 6.944345385790817e-05}
{"train_loss": 0.12569065392017365, "global_step": 101528, "epoch": 1140, "lr": 6.944291974931129e-05}
{"train_loss": 0.20596881210803986, "global_step": 101529, "epoch": 1140, "lr": 6.944238563810054e-05}
{"train_loss": 0.17962276935577393, "global_step": 101530, "epoch": 1140, "lr": 6.944185152427599e-05}
{"train_loss": 0.14403310418128967, "global_step": 101531, "epoch": 1140, "lr": 6.944131740783774e-05}
{"train_loss": 0.21687936782836914, "global_step": 101532, "epoch": 1140, "lr": 6.944078328878584e-05}
{"train_loss": 0.1557648479938507, "global_step": 101533, "epoch": 1140, "lr": 6.944024916712035e-05}
{"train_loss": 0.13675914704799652, "global_step": 101534, "epoch": 1140, "lr": 6.943971504284137e-05}
{"train_loss": 0.13743455708026886, "global_step": 101535, "epoch": 1140, "lr": 6.943918091594894e-05}
{"train_loss": 0.1549246907234192, "global_step": 101536, "epoch": 1140, "lr": 6.943864678644316e-05}
{"train_loss": 0.170014426112175, "global_step": 101537, "epoch": 1140, "lr": 6.94381126543241e-05}
{"train_loss": 0.15476426482200623, "global_step": 101538, "epoch": 1140, "lr": 6.94375785195918e-05}
{"train_loss": 0.2063056379556656, "global_step": 101539, "epoch": 1140, "lr": 6.943704438224638e-05}
{"train_loss": 0.12143570929765701, "global_step": 101540, "epoch": 1140, "lr": 6.943651024228786e-05}
{"train_loss": 0.15061110258102417, "global_step": 101541, "epoch": 1140, "lr": 6.943597609971636e-05}
{"train_loss": 0.15451689064502716, "global_step": 101542, "epoch": 1140, "lr": 6.943544195453192e-05}
{"train_loss": 0.19132542610168457, "global_step": 101543, "epoch": 1140, "lr": 6.943490780673462e-05}
{"train_loss": 0.19949816167354584, "global_step": 101544, "epoch": 1140, "lr": 6.943437365632455e-05}
{"train_loss": 0.20094437897205353, "global_step": 101545, "epoch": 1140, "lr": 6.943383950330175e-05}
{"train_loss": 0.15489932894706726, "global_step": 101546, "epoch": 1140, "lr": 6.94333053476663e-05}
{"train_loss": 0.15739770233631134, "global_step": 101547, "epoch": 1140, "lr": 6.943277118941828e-05}
{"train_loss": 0.16698404552226656, "global_step": 101548, "epoch": 1140, "lr": 6.943223702855777e-05, "val_loss": 4.662888050079346, "train_action_mse_error": 39.617088317871094}
{"train_loss": 0.16442985832691193, "global_step": 101549, "epoch": 1141, "lr": 6.943170286508483e-05}
{"train_loss": 0.22223466634750366, "global_step": 101550, "epoch": 1141, "lr": 6.943116869899953e-05}
{"train_loss": 0.22222043573856354, "global_step": 101551, "epoch": 1141, "lr": 6.943063453030196e-05}
{"train_loss": 0.11958570778369904, "global_step": 101552, "epoch": 1141, "lr": 6.943010035899218e-05}
{"train_loss": 0.14649346470832825, "global_step": 101553, "epoch": 1141, "lr": 6.942956618507025e-05}
{"train_loss": 0.11908397078514099, "global_step": 101554, "epoch": 1141, "lr": 6.942903200853626e-05}
{"train_loss": 0.12069223076105118, "global_step": 101555, "epoch": 1141, "lr": 6.942849782939025e-05}
{"train_loss": 0.23358504474163055, "global_step": 101556, "epoch": 1141, "lr": 6.942796364763234e-05}
{"train_loss": 0.15988662838935852, "global_step": 101557, "epoch": 1141, "lr": 6.942742946326256e-05}
{"train_loss": 0.13293805718421936, "global_step": 101558, "epoch": 1141, "lr": 6.942689527628102e-05}
{"train_loss": 0.18873856961727142, "global_step": 101559, "epoch": 1141, "lr": 6.942636108668774e-05}
{"train_loss": 0.09386802464723587, "global_step": 101560, "epoch": 1141, "lr": 6.942582689448285e-05}
{"train_loss": 0.2019253969192505, "global_step": 101561, "epoch": 1141, "lr": 6.942529269966638e-05}
{"train_loss": 0.11462503671646118, "global_step": 101562, "epoch": 1141, "lr": 6.942475850223841e-05}
{"train_loss": 0.1405257284641266, "global_step": 101563, "epoch": 1141, "lr": 6.942422430219904e-05}
{"train_loss": 0.1611722707748413, "global_step": 101564, "epoch": 1141, "lr": 6.94236900995483e-05}
{"train_loss": 0.10387755930423737, "global_step": 101565, "epoch": 1141, "lr": 6.942315589428628e-05}
{"train_loss": 0.09730169922113419, "global_step": 101566, "epoch": 1141, "lr": 6.942262168641306e-05}
{"train_loss": 0.1615733802318573, "global_step": 101567, "epoch": 1141, "lr": 6.942208747592871e-05}
{"train_loss": 0.13064859807491302, "global_step": 101568, "epoch": 1141, "lr": 6.942155326283329e-05}
{"train_loss": 0.16052651405334473, "global_step": 101569, "epoch": 1141, "lr": 6.942101904712688e-05}
{"train_loss": 0.14450442790985107, "global_step": 101570, "epoch": 1141, "lr": 6.942048482880957e-05}
{"train_loss": 0.1633085459470749, "global_step": 101571, "epoch": 1141, "lr": 6.94199506078814e-05}
{"train_loss": 0.24542351067066193, "global_step": 101572, "epoch": 1141, "lr": 6.941941638434245e-05}
{"train_loss": 0.19125504791736603, "global_step": 101573, "epoch": 1141, "lr": 6.94188821581928e-05}
{"train_loss": 0.19809120893478394, "global_step": 101574, "epoch": 1141, "lr": 6.941834792943252e-05}
{"train_loss": 0.17046965658664703, "global_step": 101575, "epoch": 1141, "lr": 6.941781369806166e-05}
{"train_loss": 0.1438441127538681, "global_step": 101576, "epoch": 1141, "lr": 6.941727946408033e-05}
{"train_loss": 0.10781963169574738, "global_step": 101577, "epoch": 1141, "lr": 6.941674522748858e-05}
{"train_loss": 0.15194563567638397, "global_step": 101578, "epoch": 1141, "lr": 6.94162109882865e-05}
{"train_loss": 0.16147355735301971, "global_step": 101579, "epoch": 1141, "lr": 6.941567674647413e-05}
{"train_loss": 0.15831363201141357, "global_step": 101580, "epoch": 1141, "lr": 6.941514250205155e-05}
{"train_loss": 0.06924745440483093, "global_step": 101581, "epoch": 1141, "lr": 6.941460825501886e-05}
{"train_loss": 0.13566291332244873, "global_step": 101582, "epoch": 1141, "lr": 6.94140740053761e-05}
{"train_loss": 0.1482657641172409, "global_step": 101583, "epoch": 1141, "lr": 6.941353975312336e-05}
{"train_loss": 0.16987177729606628, "global_step": 101584, "epoch": 1141, "lr": 6.94130054982607e-05}
{"train_loss": 0.13694535195827484, "global_step": 101585, "epoch": 1141, "lr": 6.94124712407882e-05}
{"train_loss": 0.09358736127614975, "global_step": 101586, "epoch": 1141, "lr": 6.941193698070594e-05}
{"train_loss": 0.15600423514842987, "global_step": 101587, "epoch": 1141, "lr": 6.941140271801397e-05}
{"train_loss": 0.13779275119304657, "global_step": 101588, "epoch": 1141, "lr": 6.941086845271237e-05}
{"train_loss": 0.11082620918750763, "global_step": 101589, "epoch": 1141, "lr": 6.941033418480123e-05}
{"train_loss": 0.13101394474506378, "global_step": 101590, "epoch": 1141, "lr": 6.94097999142806e-05}
{"train_loss": 0.15035587549209595, "global_step": 101591, "epoch": 1141, "lr": 6.940926564115057e-05}
{"train_loss": 0.16267381608486176, "global_step": 101592, "epoch": 1141, "lr": 6.94087313654112e-05}
{"train_loss": 0.239241361618042, "global_step": 101593, "epoch": 1141, "lr": 6.940819708706254e-05}
{"train_loss": 0.18354377150535583, "global_step": 101594, "epoch": 1141, "lr": 6.94076628061047e-05}
{"train_loss": 0.1234297901391983, "global_step": 101595, "epoch": 1141, "lr": 6.940712852253774e-05}
{"train_loss": 0.13051475584506989, "global_step": 101596, "epoch": 1141, "lr": 6.940659423636172e-05}
{"train_loss": 0.13596540689468384, "global_step": 101597, "epoch": 1141, "lr": 6.940605994757672e-05}
{"train_loss": 0.14610257744789124, "global_step": 101598, "epoch": 1141, "lr": 6.940552565618281e-05}
{"train_loss": 0.28047990798950195, "global_step": 101599, "epoch": 1141, "lr": 6.940499136218007e-05}
{"train_loss": 0.14139176905155182, "global_step": 101600, "epoch": 1141, "lr": 6.940445706556857e-05}
{"train_loss": 0.1658976674079895, "global_step": 101601, "epoch": 1141, "lr": 6.940392276634836e-05}
{"train_loss": 0.23214523494243622, "global_step": 101602, "epoch": 1141, "lr": 6.940338846451953e-05}
{"train_loss": 0.13768157362937927, "global_step": 101603, "epoch": 1141, "lr": 6.940285416008217e-05}
{"train_loss": 0.14775484800338745, "global_step": 101604, "epoch": 1141, "lr": 6.94023198530363e-05}
{"train_loss": 0.2280205488204956, "global_step": 101605, "epoch": 1141, "lr": 6.940178554338206e-05}
{"train_loss": 0.1551840603351593, "global_step": 101606, "epoch": 1141, "lr": 6.940125123111948e-05}
{"train_loss": 0.09518815577030182, "global_step": 101607, "epoch": 1141, "lr": 6.940071691624864e-05}
{"train_loss": 0.0935695543885231, "global_step": 101608, "epoch": 1141, "lr": 6.94001825987696e-05}
{"train_loss": 0.18775884807109833, "global_step": 101609, "epoch": 1141, "lr": 6.939964827868244e-05}
{"train_loss": 0.12419017404317856, "global_step": 101610, "epoch": 1141, "lr": 6.939911395598724e-05}
{"train_loss": 0.08933199942111969, "global_step": 101611, "epoch": 1141, "lr": 6.939857963068407e-05}
{"train_loss": 0.1670997589826584, "global_step": 101612, "epoch": 1141, "lr": 6.939804530277298e-05}
{"train_loss": 0.1455494463443756, "global_step": 101613, "epoch": 1141, "lr": 6.939751097225408e-05}
{"train_loss": 0.24886757135391235, "global_step": 101614, "epoch": 1141, "lr": 6.939697663912741e-05}
{"train_loss": 0.10656903684139252, "global_step": 101615, "epoch": 1141, "lr": 6.939644230339305e-05}
{"train_loss": 0.24424374103546143, "global_step": 101616, "epoch": 1141, "lr": 6.939590796505108e-05}
{"train_loss": 0.19333136081695557, "global_step": 101617, "epoch": 1141, "lr": 6.939537362410157e-05}
{"train_loss": 0.16728103160858154, "global_step": 101618, "epoch": 1141, "lr": 6.939483928054458e-05}
{"train_loss": 0.09967054426670074, "global_step": 101619, "epoch": 1141, "lr": 6.939430493438019e-05}
{"train_loss": 0.2432679831981659, "global_step": 101620, "epoch": 1141, "lr": 6.939377058560847e-05}
{"train_loss": 0.2337510585784912, "global_step": 101621, "epoch": 1141, "lr": 6.93932362342295e-05}
{"train_loss": 0.11193110793828964, "global_step": 101622, "epoch": 1141, "lr": 6.939270188024334e-05}
{"train_loss": 0.09253516793251038, "global_step": 101623, "epoch": 1141, "lr": 6.939216752365008e-05}
{"train_loss": 0.16391074657440186, "global_step": 101624, "epoch": 1141, "lr": 6.939163316444977e-05}
{"train_loss": 0.13846328854560852, "global_step": 101625, "epoch": 1141, "lr": 6.93910988026425e-05}
{"train_loss": 0.22085589170455933, "global_step": 101626, "epoch": 1141, "lr": 6.939056443822832e-05}
{"train_loss": 0.12960872054100037, "global_step": 101627, "epoch": 1141, "lr": 6.939003007120732e-05}
{"train_loss": 0.18709412217140198, "global_step": 101628, "epoch": 1141, "lr": 6.938949570157958e-05}
{"train_loss": 0.20649588108062744, "global_step": 101629, "epoch": 1141, "lr": 6.938896132934514e-05}
{"train_loss": 0.1011657565832138, "global_step": 101630, "epoch": 1141, "lr": 6.938842695450411e-05}
{"train_loss": 0.23580332100391388, "global_step": 101631, "epoch": 1141, "lr": 6.938789257705653e-05}
{"train_loss": 0.15818731486797333, "global_step": 101632, "epoch": 1141, "lr": 6.938735819700249e-05}
{"train_loss": 0.14578211307525635, "global_step": 101633, "epoch": 1141, "lr": 6.938682381434204e-05}
{"train_loss": 0.16911043226718903, "global_step": 101634, "epoch": 1141, "lr": 6.938628942907527e-05}
{"train_loss": 0.27838027477264404, "global_step": 101635, "epoch": 1141, "lr": 6.938575504120228e-05}
{"train_loss": 0.1286434680223465, "global_step": 101636, "epoch": 1141, "lr": 6.938522065072308e-05}
{"train_loss": 0.15918845314992947, "global_step": 101637, "epoch": 1141, "lr": 6.938468625763778e-05, "val_loss": 4.849889278411865}
{"train_loss": 0.1469600796699524, "global_step": 101638, "epoch": 1142, "lr": 6.938415186194645e-05}
{"train_loss": 0.21941445767879486, "global_step": 101639, "epoch": 1142, "lr": 6.938361746364914e-05}
{"train_loss": 0.09889379888772964, "global_step": 101640, "epoch": 1142, "lr": 6.938308306274595e-05}
{"train_loss": 0.21073883771896362, "global_step": 101641, "epoch": 1142, "lr": 6.938254865923695e-05}
{"train_loss": 0.10376682132482529, "global_step": 101642, "epoch": 1142, "lr": 6.93820142531222e-05}
{"train_loss": 0.2395554482936859, "global_step": 101643, "epoch": 1142, "lr": 6.938147984440178e-05}
{"train_loss": 0.1871253252029419, "global_step": 101644, "epoch": 1142, "lr": 6.938094543307574e-05}
{"train_loss": 0.17775945365428925, "global_step": 101645, "epoch": 1142, "lr": 6.938041101914418e-05}
{"train_loss": 0.149717777967453, "global_step": 101646, "epoch": 1142, "lr": 6.937987660260715e-05}
{"train_loss": 0.14905624091625214, "global_step": 101647, "epoch": 1142, "lr": 6.937934218346474e-05}
{"train_loss": 0.16001646220684052, "global_step": 101648, "epoch": 1142, "lr": 6.937880776171701e-05}
{"train_loss": 0.14059950411319733, "global_step": 101649, "epoch": 1142, "lr": 6.937827333736404e-05}
{"train_loss": 0.142495259642601, "global_step": 101650, "epoch": 1142, "lr": 6.93777389104059e-05}
{"train_loss": 0.1604452133178711, "global_step": 101651, "epoch": 1142, "lr": 6.937720448084266e-05}
{"train_loss": 0.13743938505649567, "global_step": 101652, "epoch": 1142, "lr": 6.937667004867438e-05}
{"train_loss": 0.06868540495634079, "global_step": 101653, "epoch": 1142, "lr": 6.937613561390115e-05}
{"train_loss": 0.10023590177297592, "global_step": 101654, "epoch": 1142, "lr": 6.937560117652302e-05}
{"train_loss": 0.19397622346878052, "global_step": 101655, "epoch": 1142, "lr": 6.93750667365401e-05}
{"train_loss": 0.14208132028579712, "global_step": 101656, "epoch": 1142, "lr": 6.937453229395242e-05}
{"train_loss": 0.09850487112998962, "global_step": 101657, "epoch": 1142, "lr": 6.937399784876007e-05}
{"train_loss": 0.1990641951560974, "global_step": 101658, "epoch": 1142, "lr": 6.937346340096314e-05}
{"train_loss": 0.16451837122440338, "global_step": 101659, "epoch": 1142, "lr": 6.937292895056167e-05}
{"train_loss": 0.13210324943065643, "global_step": 101660, "epoch": 1142, "lr": 6.937239449755574e-05}
{"train_loss": 0.14220713078975677, "global_step": 101661, "epoch": 1142, "lr": 6.937186004194544e-05}
{"train_loss": 0.14422334730625153, "global_step": 101662, "epoch": 1142, "lr": 6.937132558373083e-05}
{"train_loss": 0.27481716871261597, "global_step": 101663, "epoch": 1142, "lr": 6.937079112291197e-05}
{"train_loss": 0.16679416596889496, "global_step": 101664, "epoch": 1142, "lr": 6.937025665948896e-05}
{"train_loss": 0.16411402821540833, "global_step": 101665, "epoch": 1142, "lr": 6.936972219346185e-05}
{"train_loss": 0.1539556235074997, "global_step": 101666, "epoch": 1142, "lr": 6.93691877248307e-05}
{"train_loss": 0.22152885794639587, "global_step": 101667, "epoch": 1142, "lr": 6.936865325359563e-05}
{"train_loss": 0.14811885356903076, "global_step": 101668, "epoch": 1142, "lr": 6.936811877975666e-05}
{"train_loss": 0.1897638738155365, "global_step": 101669, "epoch": 1142, "lr": 6.936758430331388e-05}
{"train_loss": 0.12863509356975555, "global_step": 101670, "epoch": 1142, "lr": 6.936704982426737e-05}
{"train_loss": 0.16275052726268768, "global_step": 101671, "epoch": 1142, "lr": 6.93665153426172e-05}
{"train_loss": 0.17010720074176788, "global_step": 101672, "epoch": 1142, "lr": 6.936598085836343e-05}
{"train_loss": 0.2075314074754715, "global_step": 101673, "epoch": 1142, "lr": 6.936544637150615e-05}
{"train_loss": 0.14408372342586517, "global_step": 101674, "epoch": 1142, "lr": 6.936491188204541e-05}
{"train_loss": 0.15831969678401947, "global_step": 101675, "epoch": 1142, "lr": 6.936437738998128e-05}
{"train_loss": 0.13930730521678925, "global_step": 101676, "epoch": 1142, "lr": 6.936384289531387e-05}
{"train_loss": 0.12300461530685425, "global_step": 101677, "epoch": 1142, "lr": 6.936330839804322e-05}
{"train_loss": 0.1931917518377304, "global_step": 101678, "epoch": 1142, "lr": 6.93627738981694e-05}
{"train_loss": 0.15063083171844482, "global_step": 101679, "epoch": 1142, "lr": 6.936223939569252e-05}
{"train_loss": 0.17279140651226044, "global_step": 101680, "epoch": 1142, "lr": 6.93617048906126e-05}
{"train_loss": 0.23104798793792725, "global_step": 101681, "epoch": 1142, "lr": 6.936117038292975e-05}
{"train_loss": 0.11396949738264084, "global_step": 101682, "epoch": 1142, "lr": 6.936063587264401e-05}
{"train_loss": 0.0871003121137619, "global_step": 101683, "epoch": 1142, "lr": 6.936010135975547e-05}
{"train_loss": 0.11649318784475327, "global_step": 101684, "epoch": 1142, "lr": 6.93595668442642e-05}
{"train_loss": 0.13563230633735657, "global_step": 101685, "epoch": 1142, "lr": 6.935903232617028e-05}
{"train_loss": 0.09969779849052429, "global_step": 101686, "epoch": 1142, "lr": 6.935849780547377e-05}
{"train_loss": 0.12100397795438766, "global_step": 101687, "epoch": 1142, "lr": 6.935796328217476e-05}
{"train_loss": 0.1664811670780182, "global_step": 101688, "epoch": 1142, "lr": 6.935742875627329e-05}
{"train_loss": 0.09979848563671112, "global_step": 101689, "epoch": 1142, "lr": 6.935689422776946e-05}
{"train_loss": 0.16483701765537262, "global_step": 101690, "epoch": 1142, "lr": 6.935635969666333e-05}
{"train_loss": 0.12328127771615982, "global_step": 101691, "epoch": 1142, "lr": 6.935582516295496e-05}
{"train_loss": 0.1612076312303543, "global_step": 101692, "epoch": 1142, "lr": 6.935529062664446e-05}
{"train_loss": 0.1680469661951065, "global_step": 101693, "epoch": 1142, "lr": 6.935475608773184e-05}
{"train_loss": 0.1799265444278717, "global_step": 101694, "epoch": 1142, "lr": 6.935422154621723e-05}
{"train_loss": 0.18882693350315094, "global_step": 101695, "epoch": 1142, "lr": 6.935368700210069e-05}
{"train_loss": 0.19013774394989014, "global_step": 101696, "epoch": 1142, "lr": 6.935315245538227e-05}
{"train_loss": 0.14657515287399292, "global_step": 101697, "epoch": 1142, "lr": 6.935261790606206e-05}
{"train_loss": 0.20958487689495087, "global_step": 101698, "epoch": 1142, "lr": 6.935208335414011e-05}
{"train_loss": 0.10330798476934433, "global_step": 101699, "epoch": 1142, "lr": 6.935154879961654e-05}
{"train_loss": 0.16479314863681793, "global_step": 101700, "epoch": 1142, "lr": 6.935101424249138e-05}
{"train_loss": 0.13266566395759583, "global_step": 101701, "epoch": 1142, "lr": 6.935047968276469e-05}
{"train_loss": 0.2815527617931366, "global_step": 101702, "epoch": 1142, "lr": 6.934994512043659e-05}
{"train_loss": 0.17215357720851898, "global_step": 101703, "epoch": 1142, "lr": 6.934941055550712e-05}
{"train_loss": 0.14880190789699554, "global_step": 101704, "epoch": 1142, "lr": 6.934887598797634e-05}
{"train_loss": 0.2417878657579422, "global_step": 101705, "epoch": 1142, "lr": 6.934834141784436e-05}
{"train_loss": 0.20298677682876587, "global_step": 101706, "epoch": 1142, "lr": 6.934780684511123e-05}
{"train_loss": 0.12375565618276596, "global_step": 101707, "epoch": 1142, "lr": 6.934727226977702e-05}
{"train_loss": 0.1848578304052353, "global_step": 101708, "epoch": 1142, "lr": 6.93467376918418e-05}
{"train_loss": 0.15963202714920044, "global_step": 101709, "epoch": 1142, "lr": 6.934620311130564e-05}
{"train_loss": 0.19907209277153015, "global_step": 101710, "epoch": 1142, "lr": 6.934566852816864e-05}
{"train_loss": 0.12670139968395233, "global_step": 101711, "epoch": 1142, "lr": 6.934513394243082e-05}
{"train_loss": 0.1443527787923813, "global_step": 101712, "epoch": 1142, "lr": 6.93445993540923e-05}
{"train_loss": 0.17854242026805878, "global_step": 101713, "epoch": 1142, "lr": 6.934406476315314e-05}
{"train_loss": 0.23186933994293213, "global_step": 101714, "epoch": 1142, "lr": 6.93435301696134e-05}
{"train_loss": 0.15287171304225922, "global_step": 101715, "epoch": 1142, "lr": 6.934299557347316e-05}
{"train_loss": 0.17221949994564056, "global_step": 101716, "epoch": 1142, "lr": 6.93424609747325e-05}
{"train_loss": 0.17996494472026825, "global_step": 101717, "epoch": 1142, "lr": 6.934192637339145e-05}
{"train_loss": 0.16160130500793457, "global_step": 101718, "epoch": 1142, "lr": 6.934139176945013e-05}
{"train_loss": 0.22047190368175507, "global_step": 101719, "epoch": 1142, "lr": 6.934085716290861e-05}
{"train_loss": 0.16433578729629517, "global_step": 101720, "epoch": 1142, "lr": 6.934032255376695e-05}
{"train_loss": 0.17339825630187988, "global_step": 101721, "epoch": 1142, "lr": 6.93397879420252e-05}
{"train_loss": 0.16578388214111328, "global_step": 101722, "epoch": 1142, "lr": 6.933925332768345e-05}
{"train_loss": 0.15415334701538086, "global_step": 101723, "epoch": 1142, "lr": 6.933871871074179e-05}
{"train_loss": 0.15587562322616577, "global_step": 101724, "epoch": 1142, "lr": 6.933818409120027e-05}
{"train_loss": 0.2437896579504013, "global_step": 101725, "epoch": 1142, "lr": 6.933764946905897e-05}
{"train_loss": 0.16219897618454493, "global_step": 101726, "epoch": 1142, "lr": 6.933711484431795e-05, "val_loss": 4.742298126220703}
{"train_loss": 0.1656518280506134, "global_step": 101727, "epoch": 1143, "lr": 6.93365802169773e-05}
{"train_loss": 0.14146563410758972, "global_step": 101728, "epoch": 1143, "lr": 6.933604558703706e-05}
{"train_loss": 0.1651214212179184, "global_step": 101729, "epoch": 1143, "lr": 6.933551095449734e-05}
{"train_loss": 0.10787700861692429, "global_step": 101730, "epoch": 1143, "lr": 6.933497631935819e-05}
{"train_loss": 0.11780000478029251, "global_step": 101731, "epoch": 1143, "lr": 6.93344416816197e-05}
{"train_loss": 0.15429317951202393, "global_step": 101732, "epoch": 1143, "lr": 6.933390704128192e-05}
{"train_loss": 0.11371763795614243, "global_step": 101733, "epoch": 1143, "lr": 6.933337239834494e-05}
{"train_loss": 0.08718441426753998, "global_step": 101734, "epoch": 1143, "lr": 6.933283775280882e-05}
{"train_loss": 0.17015936970710754, "global_step": 101735, "epoch": 1143, "lr": 6.933230310467364e-05}
{"train_loss": 0.1140688955783844, "global_step": 101736, "epoch": 1143, "lr": 6.933176845393946e-05}
{"train_loss": 0.16037391126155853, "global_step": 101737, "epoch": 1143, "lr": 6.933123380060635e-05}
{"train_loss": 0.11180121451616287, "global_step": 101738, "epoch": 1143, "lr": 6.933069914467442e-05}
{"train_loss": 0.20211201906204224, "global_step": 101739, "epoch": 1143, "lr": 6.933016448614369e-05}
{"train_loss": 0.08819995820522308, "global_step": 101740, "epoch": 1143, "lr": 6.932962982501426e-05}
{"train_loss": 0.17918306589126587, "global_step": 101741, "epoch": 1143, "lr": 6.93290951612862e-05}
{"train_loss": 0.16549436748027802, "global_step": 101742, "epoch": 1143, "lr": 6.932856049495957e-05}
{"train_loss": 0.13337388634681702, "global_step": 101743, "epoch": 1143, "lr": 6.932802582603446e-05}
{"train_loss": 0.18694467842578888, "global_step": 101744, "epoch": 1143, "lr": 6.932749115451094e-05}
{"train_loss": 0.13513688743114471, "global_step": 101745, "epoch": 1143, "lr": 6.932695648038907e-05}
{"train_loss": 0.15927419066429138, "global_step": 101746, "epoch": 1143, "lr": 6.932642180366892e-05}
{"train_loss": 0.26902028918266296, "global_step": 101747, "epoch": 1143, "lr": 6.932588712435055e-05}
{"train_loss": 0.2330101728439331, "global_step": 101748, "epoch": 1143, "lr": 6.932535244243407e-05}
{"train_loss": 0.18264269828796387, "global_step": 101749, "epoch": 1143, "lr": 6.932481775791954e-05}
{"train_loss": 0.12201740592718124, "global_step": 101750, "epoch": 1143, "lr": 6.9324283070807e-05}
{"train_loss": 0.22796353697776794, "global_step": 101751, "epoch": 1143, "lr": 6.932374838109658e-05}
{"train_loss": 0.12222204357385635, "global_step": 101752, "epoch": 1143, "lr": 6.932321368878829e-05}
{"train_loss": 0.10690313577651978, "global_step": 101753, "epoch": 1143, "lr": 6.932267899388224e-05}
{"train_loss": 0.11988214403390884, "global_step": 101754, "epoch": 1143, "lr": 6.932214429637849e-05}
{"train_loss": 0.10807324945926666, "global_step": 101755, "epoch": 1143, "lr": 6.93216095962771e-05}
{"train_loss": 0.15209269523620605, "global_step": 101756, "epoch": 1143, "lr": 6.932107489357818e-05}
{"train_loss": 0.14512673020362854, "global_step": 101757, "epoch": 1143, "lr": 6.932054018828175e-05}
{"train_loss": 0.14429029822349548, "global_step": 101758, "epoch": 1143, "lr": 6.932000548038793e-05}
{"train_loss": 0.15177688002586365, "global_step": 101759, "epoch": 1143, "lr": 6.931947076989677e-05}
{"train_loss": 0.15069660544395447, "global_step": 101760, "epoch": 1143, "lr": 6.931893605680832e-05}
{"train_loss": 0.14126229286193848, "global_step": 101761, "epoch": 1143, "lr": 6.93184013411227e-05}
{"train_loss": 0.13372868299484253, "global_step": 101762, "epoch": 1143, "lr": 6.931786662283994e-05}
{"train_loss": 0.13189905881881714, "global_step": 101763, "epoch": 1143, "lr": 6.931733190196015e-05}
{"train_loss": 0.13682125508785248, "global_step": 101764, "epoch": 1143, "lr": 6.931679717848337e-05}
{"train_loss": 0.10305475443601608, "global_step": 101765, "epoch": 1143, "lr": 6.931626245240967e-05}
{"train_loss": 0.1891258805990219, "global_step": 101766, "epoch": 1143, "lr": 6.931572772373913e-05}
{"train_loss": 0.1597374975681305, "global_step": 101767, "epoch": 1143, "lr": 6.931519299247184e-05}
{"train_loss": 0.11633287370204926, "global_step": 101768, "epoch": 1143, "lr": 6.931465825860785e-05}
{"train_loss": 0.14127883315086365, "global_step": 101769, "epoch": 1143, "lr": 6.931412352214726e-05}
{"train_loss": 0.1816934198141098, "global_step": 101770, "epoch": 1143, "lr": 6.93135887830901e-05}
{"train_loss": 0.16541741788387299, "global_step": 101771, "epoch": 1143, "lr": 6.931305404143647e-05}
{"train_loss": 0.24613551795482635, "global_step": 101772, "epoch": 1143, "lr": 6.931251929718643e-05}
{"train_loss": 0.1494540423154831, "global_step": 101773, "epoch": 1143, "lr": 6.931198455034007e-05}
{"train_loss": 0.14532215893268585, "global_step": 101774, "epoch": 1143, "lr": 6.931144980089743e-05}
{"train_loss": 0.10633407533168793, "global_step": 101775, "epoch": 1143, "lr": 6.931091504885861e-05}
{"train_loss": 0.16537971794605255, "global_step": 101776, "epoch": 1143, "lr": 6.931038029422368e-05}
{"train_loss": 0.2479870468378067, "global_step": 101777, "epoch": 1143, "lr": 6.93098455369927e-05}
{"train_loss": 0.18597282469272614, "global_step": 101778, "epoch": 1143, "lr": 6.930931077716574e-05}
{"train_loss": 0.14822229743003845, "global_step": 101779, "epoch": 1143, "lr": 6.93087760147429e-05}
{"train_loss": 0.24427054822444916, "global_step": 101780, "epoch": 1143, "lr": 6.930824124972421e-05}
{"train_loss": 0.15967382490634918, "global_step": 101781, "epoch": 1143, "lr": 6.930770648210977e-05}
{"train_loss": 0.13224007189273834, "global_step": 101782, "epoch": 1143, "lr": 6.930717171189964e-05}
{"train_loss": 0.12383746355772018, "global_step": 101783, "epoch": 1143, "lr": 6.93066369390939e-05}
{"train_loss": 0.16354811191558838, "global_step": 101784, "epoch": 1143, "lr": 6.930610216369262e-05}
{"train_loss": 0.23759669065475464, "global_step": 101785, "epoch": 1143, "lr": 6.930556738569586e-05}
{"train_loss": 0.15037718415260315, "global_step": 101786, "epoch": 1143, "lr": 6.93050326051037e-05}
{"train_loss": 0.16838285326957703, "global_step": 101787, "epoch": 1143, "lr": 6.930449782191623e-05}
{"train_loss": 0.16572023928165436, "global_step": 101788, "epoch": 1143, "lr": 6.930396303613349e-05}
{"train_loss": 0.21635109186172485, "global_step": 101789, "epoch": 1143, "lr": 6.930342824775557e-05}
{"train_loss": 0.17347657680511475, "global_step": 101790, "epoch": 1143, "lr": 6.930289345678255e-05}
{"train_loss": 0.22101138532161713, "global_step": 101791, "epoch": 1143, "lr": 6.930235866321448e-05}
{"train_loss": 0.1384841352701187, "global_step": 101792, "epoch": 1143, "lr": 6.930182386705145e-05}
{"train_loss": 0.19947189092636108, "global_step": 101793, "epoch": 1143, "lr": 6.930128906829352e-05}
{"train_loss": 0.1414947360754013, "global_step": 101794, "epoch": 1143, "lr": 6.930075426694077e-05}
{"train_loss": 0.18161389231681824, "global_step": 101795, "epoch": 1143, "lr": 6.930021946299327e-05}
{"train_loss": 0.18893393874168396, "global_step": 101796, "epoch": 1143, "lr": 6.92996846564511e-05}
{"train_loss": 0.17996034026145935, "global_step": 101797, "epoch": 1143, "lr": 6.92991498473143e-05}
{"train_loss": 0.1456933170557022, "global_step": 101798, "epoch": 1143, "lr": 6.929861503558298e-05}
{"train_loss": 0.14989863336086273, "global_step": 101799, "epoch": 1143, "lr": 6.92980802212572e-05}
{"train_loss": 0.2786329686641693, "global_step": 101800, "epoch": 1143, "lr": 6.929754540433702e-05}
{"train_loss": 0.16243410110473633, "global_step": 101801, "epoch": 1143, "lr": 6.929701058482253e-05}
{"train_loss": 0.16074120998382568, "global_step": 101802, "epoch": 1143, "lr": 6.929647576271377e-05}
{"train_loss": 0.15501585602760315, "global_step": 101803, "epoch": 1143, "lr": 6.929594093801084e-05}
{"train_loss": 0.157944917678833, "global_step": 101804, "epoch": 1143, "lr": 6.929540611071382e-05}
{"train_loss": 0.1143341064453125, "global_step": 101805, "epoch": 1143, "lr": 6.929487128082276e-05}
{"train_loss": 0.17411862313747406, "global_step": 101806, "epoch": 1143, "lr": 6.929433644833774e-05}
{"train_loss": 0.21330955624580383, "global_step": 101807, "epoch": 1143, "lr": 6.929380161325883e-05}
{"train_loss": 0.1369982808828354, "global_step": 101808, "epoch": 1143, "lr": 6.929326677558611e-05}
{"train_loss": 0.15959620475769043, "global_step": 101809, "epoch": 1143, "lr": 6.929273193531964e-05}
{"train_loss": 0.16697658598423004, "global_step": 101810, "epoch": 1143, "lr": 6.92921970924595e-05}
{"train_loss": 0.1744871288537979, "global_step": 101811, "epoch": 1143, "lr": 6.929166224700577e-05}
{"train_loss": 0.2062312662601471, "global_step": 101812, "epoch": 1143, "lr": 6.929112739895849e-05}
{"train_loss": 0.16548994183540344, "global_step": 101813, "epoch": 1143, "lr": 6.929059254831775e-05}
{"train_loss": 0.10424361377954483, "global_step": 101814, "epoch": 1143, "lr": 6.929005769508364e-05}
{"train_loss": 0.16041101984093697, "global_step": 101815, "epoch": 1143, "lr": 6.928952283925623e-05, "val_loss": 4.643917560577393}
{"train_loss": 0.13879387080669403, "global_step": 101816, "epoch": 1144, "lr": 6.928898798083555e-05}
{"train_loss": 0.16354142129421234, "global_step": 101817, "epoch": 1144, "lr": 6.928845311982171e-05}
{"train_loss": 0.11302603036165237, "global_step": 101818, "epoch": 1144, "lr": 6.928791825621477e-05}
{"train_loss": 0.0871872827410698, "global_step": 101819, "epoch": 1144, "lr": 6.928738339001482e-05}
{"train_loss": 0.12790176272392273, "global_step": 101820, "epoch": 1144, "lr": 6.928684852122191e-05}
{"train_loss": 0.12774278223514557, "global_step": 101821, "epoch": 1144, "lr": 6.928631364983611e-05}
{"train_loss": 0.14330095052719116, "global_step": 101822, "epoch": 1144, "lr": 6.928577877585751e-05}
{"train_loss": 0.23121067881584167, "global_step": 101823, "epoch": 1144, "lr": 6.928524389928617e-05}
{"train_loss": 0.11787865310907364, "global_step": 101824, "epoch": 1144, "lr": 6.928470902012216e-05}
{"train_loss": 0.12713780999183655, "global_step": 101825, "epoch": 1144, "lr": 6.928417413836555e-05}
{"train_loss": 0.2198338657617569, "global_step": 101826, "epoch": 1144, "lr": 6.928363925401643e-05}
{"train_loss": 0.14210525155067444, "global_step": 101827, "epoch": 1144, "lr": 6.928310436707486e-05}
{"train_loss": 0.14732317626476288, "global_step": 101828, "epoch": 1144, "lr": 6.928256947754091e-05}
{"train_loss": 0.21386748552322388, "global_step": 101829, "epoch": 1144, "lr": 6.928203458541464e-05}
{"train_loss": 0.17633135616779327, "global_step": 101830, "epoch": 1144, "lr": 6.928149969069615e-05}
{"train_loss": 0.14666855335235596, "global_step": 101831, "epoch": 1144, "lr": 6.92809647933855e-05}
{"train_loss": 0.18180547654628754, "global_step": 101832, "epoch": 1144, "lr": 6.928042989348275e-05}
{"train_loss": 0.18191304802894592, "global_step": 101833, "epoch": 1144, "lr": 6.927989499098798e-05}
{"train_loss": 0.19629471004009247, "global_step": 101834, "epoch": 1144, "lr": 6.927936008590127e-05}
{"train_loss": 0.1178588718175888, "global_step": 101835, "epoch": 1144, "lr": 6.927882517822268e-05}
{"train_loss": 0.21080544590950012, "global_step": 101836, "epoch": 1144, "lr": 6.927829026795229e-05}
{"train_loss": 0.16656240820884705, "global_step": 101837, "epoch": 1144, "lr": 6.927775535509016e-05}
{"train_loss": 0.13585253059864044, "global_step": 101838, "epoch": 1144, "lr": 6.927722043963639e-05}
{"train_loss": 0.14008137583732605, "global_step": 101839, "epoch": 1144, "lr": 6.927668552159101e-05}
{"train_loss": 0.19941002130508423, "global_step": 101840, "epoch": 1144, "lr": 6.927615060095413e-05}
{"train_loss": 0.15933918952941895, "global_step": 101841, "epoch": 1144, "lr": 6.92756156777258e-05}
{"train_loss": 0.1720113903284073, "global_step": 101842, "epoch": 1144, "lr": 6.92750807519061e-05}
{"train_loss": 0.1553047001361847, "global_step": 101843, "epoch": 1144, "lr": 6.927454582349511e-05}
{"train_loss": 0.20287549495697021, "global_step": 101844, "epoch": 1144, "lr": 6.927401089249288e-05}
{"train_loss": 0.11169389635324478, "global_step": 101845, "epoch": 1144, "lr": 6.927347595889949e-05}
{"train_loss": 0.1187017634510994, "global_step": 101846, "epoch": 1144, "lr": 6.927294102271502e-05}
{"train_loss": 0.14017489552497864, "global_step": 101847, "epoch": 1144, "lr": 6.927240608393954e-05}
{"train_loss": 0.16002942621707916, "global_step": 101848, "epoch": 1144, "lr": 6.927187114257313e-05}
{"train_loss": 0.20834624767303467, "global_step": 101849, "epoch": 1144, "lr": 6.927133619861583e-05}
{"train_loss": 0.1275385618209839, "global_step": 101850, "epoch": 1144, "lr": 6.927080125206775e-05}
{"train_loss": 0.20633941888809204, "global_step": 101851, "epoch": 1144, "lr": 6.927026630292895e-05}
{"train_loss": 0.2117038369178772, "global_step": 101852, "epoch": 1144, "lr": 6.926973135119948e-05}
{"train_loss": 0.15574997663497925, "global_step": 101853, "epoch": 1144, "lr": 6.926919639687945e-05}
{"train_loss": 0.20655833184719086, "global_step": 101854, "epoch": 1144, "lr": 6.926866143996889e-05}
{"train_loss": 0.1558036357164383, "global_step": 101855, "epoch": 1144, "lr": 6.92681264804679e-05}
{"train_loss": 0.14268465340137482, "global_step": 101856, "epoch": 1144, "lr": 6.926759151837656e-05}
{"train_loss": 0.19687286019325256, "global_step": 101857, "epoch": 1144, "lr": 6.926705655369493e-05}
{"train_loss": 0.2061595618724823, "global_step": 101858, "epoch": 1144, "lr": 6.926652158642306e-05}
{"train_loss": 0.19268058240413666, "global_step": 101859, "epoch": 1144, "lr": 6.926598661656104e-05}
{"train_loss": 0.14431799948215485, "global_step": 101860, "epoch": 1144, "lr": 6.926545164410895e-05}
{"train_loss": 0.09817003458738327, "global_step": 101861, "epoch": 1144, "lr": 6.926491666906684e-05}
{"train_loss": 0.093898244202137, "global_step": 101862, "epoch": 1144, "lr": 6.926438169143483e-05}
{"train_loss": 0.19783428311347961, "global_step": 101863, "epoch": 1144, "lr": 6.926384671121293e-05}
{"train_loss": 0.13833196461200714, "global_step": 101864, "epoch": 1144, "lr": 6.926331172840126e-05}
{"train_loss": 0.1347600668668747, "global_step": 101865, "epoch": 1144, "lr": 6.926277674299986e-05}
{"train_loss": 0.14887939393520355, "global_step": 101866, "epoch": 1144, "lr": 6.926224175500882e-05}
{"train_loss": 0.11988166719675064, "global_step": 101867, "epoch": 1144, "lr": 6.92617067644282e-05}
{"train_loss": 0.15137718617916107, "global_step": 101868, "epoch": 1144, "lr": 6.926117177125808e-05}
{"train_loss": 0.15895798802375793, "global_step": 101869, "epoch": 1144, "lr": 6.926063677549854e-05}
{"train_loss": 0.1851256787776947, "global_step": 101870, "epoch": 1144, "lr": 6.926010177714964e-05}
{"train_loss": 0.21799133718013763, "global_step": 101871, "epoch": 1144, "lr": 6.925956677621144e-05}
{"train_loss": 0.18291564285755157, "global_step": 101872, "epoch": 1144, "lr": 6.925903177268402e-05}
{"train_loss": 0.1997959166765213, "global_step": 101873, "epoch": 1144, "lr": 6.925849676656748e-05}
{"train_loss": 0.13878008723258972, "global_step": 101874, "epoch": 1144, "lr": 6.925796175786186e-05}
{"train_loss": 0.11490317434072495, "global_step": 101875, "epoch": 1144, "lr": 6.925742674656725e-05}
{"train_loss": 0.1724882423877716, "global_step": 101876, "epoch": 1144, "lr": 6.925689173268371e-05}
{"train_loss": 0.1963932067155838, "global_step": 101877, "epoch": 1144, "lr": 6.925635671621131e-05}
{"train_loss": 0.20482924580574036, "global_step": 101878, "epoch": 1144, "lr": 6.925582169715013e-05}
{"train_loss": 0.19021663069725037, "global_step": 101879, "epoch": 1144, "lr": 6.925528667550025e-05}
{"train_loss": 0.1450863927602768, "global_step": 101880, "epoch": 1144, "lr": 6.925475165126171e-05}
{"train_loss": 0.20647285878658295, "global_step": 101881, "epoch": 1144, "lr": 6.925421662443462e-05}
{"train_loss": 0.10930752754211426, "global_step": 101882, "epoch": 1144, "lr": 6.925368159501903e-05}
{"train_loss": 0.09146205335855484, "global_step": 101883, "epoch": 1144, "lr": 6.925314656301502e-05}
{"train_loss": 0.1324750930070877, "global_step": 101884, "epoch": 1144, "lr": 6.925261152842265e-05}
{"train_loss": 0.11920641362667084, "global_step": 101885, "epoch": 1144, "lr": 6.925207649124202e-05}
{"train_loss": 0.12428057938814163, "global_step": 101886, "epoch": 1144, "lr": 6.925154145147315e-05}
{"train_loss": 0.1482536941766739, "global_step": 101887, "epoch": 1144, "lr": 6.925100640911618e-05}
{"train_loss": 0.16229629516601562, "global_step": 101888, "epoch": 1144, "lr": 6.925047136417113e-05}
{"train_loss": 0.1820550411939621, "global_step": 101889, "epoch": 1144, "lr": 6.924993631663809e-05}
{"train_loss": 0.20058047771453857, "global_step": 101890, "epoch": 1144, "lr": 6.924940126651714e-05}
{"train_loss": 0.23724062740802765, "global_step": 101891, "epoch": 1144, "lr": 6.924886621380832e-05}
{"train_loss": 0.30764180421829224, "global_step": 101892, "epoch": 1144, "lr": 6.924833115851173e-05}
{"train_loss": 0.11922624707221985, "global_step": 101893, "epoch": 1144, "lr": 6.924779610062745e-05}
{"train_loss": 0.12112919986248016, "global_step": 101894, "epoch": 1144, "lr": 6.924726104015554e-05}
{"train_loss": 0.15211205184459686, "global_step": 101895, "epoch": 1144, "lr": 6.924672597709606e-05}
{"train_loss": 0.13745665550231934, "global_step": 101896, "epoch": 1144, "lr": 6.92461909114491e-05}
{"train_loss": 0.17364704608917236, "global_step": 101897, "epoch": 1144, "lr": 6.924565584321472e-05}
{"train_loss": 0.15467973053455353, "global_step": 101898, "epoch": 1144, "lr": 6.9245120772393e-05}
{"train_loss": 0.1765897423028946, "global_step": 101899, "epoch": 1144, "lr": 6.9244585698984e-05}
{"train_loss": 0.18673360347747803, "global_step": 101900, "epoch": 1144, "lr": 6.924405062298779e-05}
{"train_loss": 0.1578863561153412, "global_step": 101901, "epoch": 1144, "lr": 6.924351554440448e-05}
{"train_loss": 0.18108223378658295, "global_step": 101902, "epoch": 1144, "lr": 6.924298046323409e-05}
{"train_loss": 0.1486591100692749, "global_step": 101903, "epoch": 1144, "lr": 6.924244537947673e-05}
{"train_loss": 0.16116011896160212, "global_step": 101904, "epoch": 1144, "lr": 6.924191029313247e-05, "val_loss": 4.761302947998047}
{"train_loss": 0.20384114980697632, "global_step": 101905, "epoch": 1145, "lr": 6.924137520420135e-05}
{"train_loss": 0.197719544172287, "global_step": 101906, "epoch": 1145, "lr": 6.924084011268346e-05}
{"train_loss": 0.15361900627613068, "global_step": 101907, "epoch": 1145, "lr": 6.924030501857888e-05}
{"train_loss": 0.1724822074174881, "global_step": 101908, "epoch": 1145, "lr": 6.923976992188767e-05}
{"train_loss": 0.3022676110267639, "global_step": 101909, "epoch": 1145, "lr": 6.923923482260992e-05}
{"train_loss": 0.1359473168849945, "global_step": 101910, "epoch": 1145, "lr": 6.923869972074568e-05}
{"train_loss": 0.2295648604631424, "global_step": 101911, "epoch": 1145, "lr": 6.923816461629502e-05}
{"train_loss": 0.1452876329421997, "global_step": 101912, "epoch": 1145, "lr": 6.923762950925804e-05}
{"train_loss": 0.14078496396541595, "global_step": 101913, "epoch": 1145, "lr": 6.92370943996348e-05}
{"train_loss": 0.11770530045032501, "global_step": 101914, "epoch": 1145, "lr": 6.923655928742536e-05}
{"train_loss": 0.12045237421989441, "global_step": 101915, "epoch": 1145, "lr": 6.92360241726298e-05}
{"train_loss": 0.13437815010547638, "global_step": 101916, "epoch": 1145, "lr": 6.923548905524819e-05}
{"train_loss": 0.1661045402288437, "global_step": 101917, "epoch": 1145, "lr": 6.923495393528059e-05}
{"train_loss": 0.17239437997341156, "global_step": 101918, "epoch": 1145, "lr": 6.92344188127271e-05}
{"train_loss": 0.18282148241996765, "global_step": 101919, "epoch": 1145, "lr": 6.923388368758778e-05}
{"train_loss": 0.10953419655561447, "global_step": 101920, "epoch": 1145, "lr": 6.923334855986269e-05}
{"train_loss": 0.15541225671768188, "global_step": 101921, "epoch": 1145, "lr": 6.92328134295519e-05}
{"train_loss": 0.1861625760793686, "global_step": 101922, "epoch": 1145, "lr": 6.92322782966555e-05}
{"train_loss": 0.17683908343315125, "global_step": 101923, "epoch": 1145, "lr": 6.923174316117357e-05}
{"train_loss": 0.18071380257606506, "global_step": 101924, "epoch": 1145, "lr": 6.923120802310615e-05}
{"train_loss": 0.22639212012290955, "global_step": 101925, "epoch": 1145, "lr": 6.923067288245334e-05}
{"train_loss": 0.14624756574630737, "global_step": 101926, "epoch": 1145, "lr": 6.92301377392152e-05}
{"train_loss": 0.17733512818813324, "global_step": 101927, "epoch": 1145, "lr": 6.92296025933918e-05}
{"train_loss": 0.23483362793922424, "global_step": 101928, "epoch": 1145, "lr": 6.92290674449832e-05}
{"train_loss": 0.11604612320661545, "global_step": 101929, "epoch": 1145, "lr": 6.92285322939895e-05}
{"train_loss": 0.12239958345890045, "global_step": 101930, "epoch": 1145, "lr": 6.922799714041075e-05}
{"train_loss": 0.2032514214515686, "global_step": 101931, "epoch": 1145, "lr": 6.922746198424703e-05}
{"train_loss": 0.09229188412427902, "global_step": 101932, "epoch": 1145, "lr": 6.922692682549843e-05}
{"train_loss": 0.0982280895113945, "global_step": 101933, "epoch": 1145, "lr": 6.9226391664165e-05}
{"train_loss": 0.12850864231586456, "global_step": 101934, "epoch": 1145, "lr": 6.92258565002468e-05}
{"train_loss": 0.22767053544521332, "global_step": 101935, "epoch": 1145, "lr": 6.922532133374393e-05}
{"train_loss": 0.07112989574670792, "global_step": 101936, "epoch": 1145, "lr": 6.922478616465645e-05}
{"train_loss": 0.14267204701900482, "global_step": 101937, "epoch": 1145, "lr": 6.922425099298441e-05}
{"train_loss": 0.11088511347770691, "global_step": 101938, "epoch": 1145, "lr": 6.922371581872793e-05}
{"train_loss": 0.14035509526729584, "global_step": 101939, "epoch": 1145, "lr": 6.922318064188705e-05}
{"train_loss": 0.10339025408029556, "global_step": 101940, "epoch": 1145, "lr": 6.922264546246185e-05}
{"train_loss": 0.12564054131507874, "global_step": 101941, "epoch": 1145, "lr": 6.92221102804524e-05}
{"train_loss": 0.19049988687038422, "global_step": 101942, "epoch": 1145, "lr": 6.922157509585876e-05}
{"train_loss": 0.12543316185474396, "global_step": 101943, "epoch": 1145, "lr": 6.922103990868102e-05}
{"train_loss": 0.14161744713783264, "global_step": 101944, "epoch": 1145, "lr": 6.922050471891924e-05}
{"train_loss": 0.20761160552501678, "global_step": 101945, "epoch": 1145, "lr": 6.92199695265735e-05}
{"train_loss": 0.1176563948392868, "global_step": 101946, "epoch": 1145, "lr": 6.921943433164388e-05}
{"train_loss": 0.17851240932941437, "global_step": 101947, "epoch": 1145, "lr": 6.921889913413042e-05}
{"train_loss": 0.16461403667926788, "global_step": 101948, "epoch": 1145, "lr": 6.921836393403322e-05}
{"train_loss": 0.12937085330486298, "global_step": 101949, "epoch": 1145, "lr": 6.921782873135237e-05}
{"train_loss": 0.1113240048289299, "global_step": 101950, "epoch": 1145, "lr": 6.921729352608789e-05}
{"train_loss": 0.14788125455379486, "global_step": 101951, "epoch": 1145, "lr": 6.921675831823988e-05}
{"train_loss": 0.18663112819194794, "global_step": 101952, "epoch": 1145, "lr": 6.921622310780844e-05}
{"train_loss": 0.13027852773666382, "global_step": 101953, "epoch": 1145, "lr": 6.921568789479358e-05}
{"train_loss": 0.22083374857902527, "global_step": 101954, "epoch": 1145, "lr": 6.921515267919543e-05}
{"train_loss": 0.21043241024017334, "global_step": 101955, "epoch": 1145, "lr": 6.921461746101402e-05}
{"train_loss": 0.10764613747596741, "global_step": 101956, "epoch": 1145, "lr": 6.921408224024945e-05}
{"train_loss": 0.17771883308887482, "global_step": 101957, "epoch": 1145, "lr": 6.921354701690178e-05}
{"train_loss": 0.26462066173553467, "global_step": 101958, "epoch": 1145, "lr": 6.921301179097106e-05}
{"train_loss": 0.1313556581735611, "global_step": 101959, "epoch": 1145, "lr": 6.921247656245741e-05}
{"train_loss": 0.12497661262750626, "global_step": 101960, "epoch": 1145, "lr": 6.921194133136087e-05}
{"train_loss": 0.13937488198280334, "global_step": 101961, "epoch": 1145, "lr": 6.921140609768151e-05}
{"train_loss": 0.1364058554172516, "global_step": 101962, "epoch": 1145, "lr": 6.921087086141942e-05}
{"train_loss": 0.16096597909927368, "global_step": 101963, "epoch": 1145, "lr": 6.921033562257466e-05}
{"train_loss": 0.12560950219631195, "global_step": 101964, "epoch": 1145, "lr": 6.92098003811473e-05}
{"train_loss": 0.13617727160453796, "global_step": 101965, "epoch": 1145, "lr": 6.920926513713742e-05}
{"train_loss": 0.11767750978469849, "global_step": 101966, "epoch": 1145, "lr": 6.920872989054508e-05}
{"train_loss": 0.1320178359746933, "global_step": 101967, "epoch": 1145, "lr": 6.920819464137037e-05}
{"train_loss": 0.10697420686483383, "global_step": 101968, "epoch": 1145, "lr": 6.920765938961335e-05}
{"train_loss": 0.1870945543050766, "global_step": 101969, "epoch": 1145, "lr": 6.92071241352741e-05}
{"train_loss": 0.13565604388713837, "global_step": 101970, "epoch": 1145, "lr": 6.920658887835268e-05}
{"train_loss": 0.09681141376495361, "global_step": 101971, "epoch": 1145, "lr": 6.920605361884918e-05}
{"train_loss": 0.17319834232330322, "global_step": 101972, "epoch": 1145, "lr": 6.920551835676365e-05}
{"train_loss": 0.2747059464454651, "global_step": 101973, "epoch": 1145, "lr": 6.920498309209617e-05}
{"train_loss": 0.1807778775691986, "global_step": 101974, "epoch": 1145, "lr": 6.920444782484684e-05}
{"train_loss": 0.1808992624282837, "global_step": 101975, "epoch": 1145, "lr": 6.920391255501566e-05}
{"train_loss": 0.18531715869903564, "global_step": 101976, "epoch": 1145, "lr": 6.920337728260278e-05}
{"train_loss": 0.14852803945541382, "global_step": 101977, "epoch": 1145, "lr": 6.920284200760823e-05}
{"train_loss": 0.15166954696178436, "global_step": 101978, "epoch": 1145, "lr": 6.92023067300321e-05}
{"train_loss": 0.16224534809589386, "global_step": 101979, "epoch": 1145, "lr": 6.920177144987444e-05}
{"train_loss": 0.14520132541656494, "global_step": 101980, "epoch": 1145, "lr": 6.920123616713534e-05}
{"train_loss": 0.1901426464319229, "global_step": 101981, "epoch": 1145, "lr": 6.920070088181487e-05}
{"train_loss": 0.24368984997272491, "global_step": 101982, "epoch": 1145, "lr": 6.920016559391311e-05}
{"train_loss": 0.12431789934635162, "global_step": 101983, "epoch": 1145, "lr": 6.91996303034301e-05}
{"train_loss": 0.12390326708555222, "global_step": 101984, "epoch": 1145, "lr": 6.919909501036595e-05}
{"train_loss": 0.12275876104831696, "global_step": 101985, "epoch": 1145, "lr": 6.91985597147207e-05}
{"train_loss": 0.19254831969738007, "global_step": 101986, "epoch": 1145, "lr": 6.919802441649446e-05}
{"train_loss": 0.2189767211675644, "global_step": 101987, "epoch": 1145, "lr": 6.919748911568726e-05}
{"train_loss": 0.13177251815795898, "global_step": 101988, "epoch": 1145, "lr": 6.919695381229921e-05}
{"train_loss": 0.14729344844818115, "global_step": 101989, "epoch": 1145, "lr": 6.919641850633035e-05}
{"train_loss": 0.08865763992071152, "global_step": 101990, "epoch": 1145, "lr": 6.919588319778077e-05}
{"train_loss": 0.20416612923145294, "global_step": 101991, "epoch": 1145, "lr": 6.919534788665053e-05}
{"train_loss": 0.19938279688358307, "global_step": 101992, "epoch": 1145, "lr": 6.919481257293972e-05}
{"train_loss": 0.1588945049750671, "global_step": 101993, "epoch": 1145, "lr": 6.919427725664841e-05, "val_loss": 4.644132137298584, "train_action_mse_error": 43.569427490234375}
{"train_loss": 0.20322178304195404, "global_step": 101994, "epoch": 1146, "lr": 6.919374193777665e-05}
{"train_loss": 0.1315600723028183, "global_step": 101995, "epoch": 1146, "lr": 6.919320661632453e-05}
{"train_loss": 0.13699643313884735, "global_step": 101996, "epoch": 1146, "lr": 6.919267129229211e-05}
{"train_loss": 0.1740584373474121, "global_step": 101997, "epoch": 1146, "lr": 6.919213596567947e-05}
{"train_loss": 0.13184639811515808, "global_step": 101998, "epoch": 1146, "lr": 6.919160063648668e-05}
{"train_loss": 0.13619141280651093, "global_step": 101999, "epoch": 1146, "lr": 6.919106530471382e-05}
{"train_loss": 0.20541499555110931, "global_step": 102000, "epoch": 1146, "lr": 6.919052997036095e-05}
{"train_loss": 0.1941889077425003, "global_step": 102001, "epoch": 1146, "lr": 6.918999463342814e-05}
{"train_loss": 0.13890622556209564, "global_step": 102002, "epoch": 1146, "lr": 6.918945929391549e-05}
{"train_loss": 0.11579819768667221, "global_step": 102003, "epoch": 1146, "lr": 6.918892395182304e-05}
{"train_loss": 0.15163160860538483, "global_step": 102004, "epoch": 1146, "lr": 6.918838860715087e-05}
{"train_loss": 0.14585790038108826, "global_step": 102005, "epoch": 1146, "lr": 6.918785325989904e-05}
{"train_loss": 0.18275323510169983, "global_step": 102006, "epoch": 1146, "lr": 6.918731791006767e-05}
{"train_loss": 0.2064070999622345, "global_step": 102007, "epoch": 1146, "lr": 6.918678255765678e-05}
{"train_loss": 0.1882575899362564, "global_step": 102008, "epoch": 1146, "lr": 6.918624720266647e-05}
{"train_loss": 0.19670303165912628, "global_step": 102009, "epoch": 1146, "lr": 6.918571184509681e-05}
{"train_loss": 0.16395846009254456, "global_step": 102010, "epoch": 1146, "lr": 6.918517648494784e-05}
{"train_loss": 0.1735055148601532, "global_step": 102011, "epoch": 1146, "lr": 6.918464112221968e-05}
{"train_loss": 0.14891096949577332, "global_step": 102012, "epoch": 1146, "lr": 6.918410575691236e-05}
{"train_loss": 0.11897878348827362, "global_step": 102013, "epoch": 1146, "lr": 6.918357038902599e-05}
{"train_loss": 0.140432670712471, "global_step": 102014, "epoch": 1146, "lr": 6.91830350185606e-05}
{"train_loss": 0.10669273883104324, "global_step": 102015, "epoch": 1146, "lr": 6.91824996455163e-05}
{"train_loss": 0.11906033009290695, "global_step": 102016, "epoch": 1146, "lr": 6.918196426989315e-05}
{"train_loss": 0.16040486097335815, "global_step": 102017, "epoch": 1146, "lr": 6.918142889169122e-05}
{"train_loss": 0.13596035540103912, "global_step": 102018, "epoch": 1146, "lr": 6.918089351091058e-05}
{"train_loss": 0.17903316020965576, "global_step": 102019, "epoch": 1146, "lr": 6.918035812755128e-05}
{"train_loss": 0.13074930012226105, "global_step": 102020, "epoch": 1146, "lr": 6.917982274161343e-05}
{"train_loss": 0.14599378407001495, "global_step": 102021, "epoch": 1146, "lr": 6.917928735309708e-05}
{"train_loss": 0.1647709161043167, "global_step": 102022, "epoch": 1146, "lr": 6.917875196200232e-05}
{"train_loss": 0.1854616105556488, "global_step": 102023, "epoch": 1146, "lr": 6.91782165683292e-05}
{"train_loss": 0.14332249760627747, "global_step": 102024, "epoch": 1146, "lr": 6.917768117207781e-05}
{"train_loss": 0.14735053479671478, "global_step": 102025, "epoch": 1146, "lr": 6.917714577324821e-05}
{"train_loss": 0.1701420098543167, "global_step": 102026, "epoch": 1146, "lr": 6.917661037184048e-05}
{"train_loss": 0.16307419538497925, "global_step": 102027, "epoch": 1146, "lr": 6.917607496785468e-05}
{"train_loss": 0.17559608817100525, "global_step": 102028, "epoch": 1146, "lr": 6.91755395612909e-05}
{"train_loss": 0.17375218868255615, "global_step": 102029, "epoch": 1146, "lr": 6.917500415214922e-05}
{"train_loss": 0.13015137612819672, "global_step": 102030, "epoch": 1146, "lr": 6.917446874042968e-05}
{"train_loss": 0.18804849684238434, "global_step": 102031, "epoch": 1146, "lr": 6.917393332613236e-05}
{"train_loss": 0.1872563362121582, "global_step": 102032, "epoch": 1146, "lr": 6.917339790925733e-05}
{"train_loss": 0.2072248011827469, "global_step": 102033, "epoch": 1146, "lr": 6.917286248980468e-05}
{"train_loss": 0.20369303226470947, "global_step": 102034, "epoch": 1146, "lr": 6.917232706777448e-05}
{"train_loss": 0.13104534149169922, "global_step": 102035, "epoch": 1146, "lr": 6.917179164316678e-05}
{"train_loss": 0.13180695474147797, "global_step": 102036, "epoch": 1146, "lr": 6.917125621598168e-05}
{"train_loss": 0.13755950331687927, "global_step": 102037, "epoch": 1146, "lr": 6.917072078621922e-05}
{"train_loss": 0.2144990861415863, "global_step": 102038, "epoch": 1146, "lr": 6.91701853538795e-05}
{"train_loss": 0.1019749641418457, "global_step": 102039, "epoch": 1146, "lr": 6.916964991896259e-05}
{"train_loss": 0.12066822499036789, "global_step": 102040, "epoch": 1146, "lr": 6.916911448146855e-05}
{"train_loss": 0.18536509573459625, "global_step": 102041, "epoch": 1146, "lr": 6.916857904139744e-05}
{"train_loss": 0.13757774233818054, "global_step": 102042, "epoch": 1146, "lr": 6.916804359874938e-05}
{"train_loss": 0.1507153958082199, "global_step": 102043, "epoch": 1146, "lr": 6.916750815352438e-05}
{"train_loss": 0.26365306973457336, "global_step": 102044, "epoch": 1146, "lr": 6.916697270572255e-05}
{"train_loss": 0.20887918770313263, "global_step": 102045, "epoch": 1146, "lr": 6.916643725534396e-05}
{"train_loss": 0.07448787242174149, "global_step": 102046, "epoch": 1146, "lr": 6.916590180238868e-05}
{"train_loss": 0.17639905214309692, "global_step": 102047, "epoch": 1146, "lr": 6.916536634685677e-05}
{"train_loss": 0.1846248358488083, "global_step": 102048, "epoch": 1146, "lr": 6.916483088874832e-05}
{"train_loss": 0.11851060390472412, "global_step": 102049, "epoch": 1146, "lr": 6.916429542806339e-05}
{"train_loss": 0.15667711198329926, "global_step": 102050, "epoch": 1146, "lr": 6.916375996480205e-05}
{"train_loss": 0.1836835741996765, "global_step": 102051, "epoch": 1146, "lr": 6.916322449896437e-05}
{"train_loss": 0.1454065889120102, "global_step": 102052, "epoch": 1146, "lr": 6.916268903055044e-05}
{"train_loss": 0.2486938238143921, "global_step": 102053, "epoch": 1146, "lr": 6.916215355956032e-05}
{"train_loss": 0.21141257882118225, "global_step": 102054, "epoch": 1146, "lr": 6.916161808599407e-05}
{"train_loss": 0.16290722787380219, "global_step": 102055, "epoch": 1146, "lr": 6.916108260985178e-05}
{"train_loss": 0.1776212602853775, "global_step": 102056, "epoch": 1146, "lr": 6.916054713113351e-05}
{"train_loss": 0.1286756843328476, "global_step": 102057, "epoch": 1146, "lr": 6.916001164983935e-05}
{"train_loss": 0.09657030552625656, "global_step": 102058, "epoch": 1146, "lr": 6.915947616596935e-05}
{"train_loss": 0.22837169468402863, "global_step": 102059, "epoch": 1146, "lr": 6.915894067952359e-05}
{"train_loss": 0.1367340236902237, "global_step": 102060, "epoch": 1146, "lr": 6.915840519050215e-05}
{"train_loss": 0.13638587296009064, "global_step": 102061, "epoch": 1146, "lr": 6.91578696989051e-05}
{"train_loss": 0.1693173050880432, "global_step": 102062, "epoch": 1146, "lr": 6.91573342047325e-05}
{"train_loss": 0.1577715277671814, "global_step": 102063, "epoch": 1146, "lr": 6.915679870798442e-05}
{"train_loss": 0.16679774224758148, "global_step": 102064, "epoch": 1146, "lr": 6.915626320866096e-05}
{"train_loss": 0.20696021616458893, "global_step": 102065, "epoch": 1146, "lr": 6.915572770676217e-05}
{"train_loss": 0.1295543909072876, "global_step": 102066, "epoch": 1146, "lr": 6.915519220228813e-05}
{"train_loss": 0.1533571481704712, "global_step": 102067, "epoch": 1146, "lr": 6.91546566952389e-05}
{"train_loss": 0.15534421801567078, "global_step": 102068, "epoch": 1146, "lr": 6.915412118561457e-05}
{"train_loss": 0.11537174880504608, "global_step": 102069, "epoch": 1146, "lr": 6.91535856734152e-05}
{"train_loss": 0.18185196816921234, "global_step": 102070, "epoch": 1146, "lr": 6.915305015864085e-05}
{"train_loss": 0.08131124079227448, "global_step": 102071, "epoch": 1146, "lr": 6.915251464129162e-05}
{"train_loss": 0.11709515750408173, "global_step": 102072, "epoch": 1146, "lr": 6.915197912136756e-05}
{"train_loss": 0.17898136377334595, "global_step": 102073, "epoch": 1146, "lr": 6.915144359886875e-05}
{"train_loss": 0.12637723982334137, "global_step": 102074, "epoch": 1146, "lr": 6.915090807379526e-05}
{"train_loss": 0.1530400961637497, "global_step": 102075, "epoch": 1146, "lr": 6.915037254614717e-05}
{"train_loss": 0.1496642678976059, "global_step": 102076, "epoch": 1146, "lr": 6.914983701592455e-05}
{"train_loss": 0.2055954486131668, "global_step": 102077, "epoch": 1146, "lr": 6.914930148312744e-05}
{"train_loss": 0.2130604237318039, "global_step": 102078, "epoch": 1146, "lr": 6.914876594775597e-05}
{"train_loss": 0.1814679354429245, "global_step": 102079, "epoch": 1146, "lr": 6.914823040981018e-05}
{"train_loss": 0.16141349077224731, "global_step": 102080, "epoch": 1146, "lr": 6.914769486929014e-05}
{"train_loss": 0.08521471172571182, "global_step": 102081, "epoch": 1146, "lr": 6.914715932619591e-05}
{"train_loss": 0.1588956963145331, "global_step": 102082, "epoch": 1146, "lr": 6.914662378052759e-05, "val_loss": 4.772064208984375}
{"train_loss": 0.11113501340150833, "global_step": 102083, "epoch": 1147, "lr": 6.914608823228524e-05}
{"train_loss": 0.1968851089477539, "global_step": 102084, "epoch": 1147, "lr": 6.914555268146894e-05}
{"train_loss": 0.1338305026292801, "global_step": 102085, "epoch": 1147, "lr": 6.914501712807874e-05}
{"train_loss": 0.2100042998790741, "global_step": 102086, "epoch": 1147, "lr": 6.914448157211473e-05}
{"train_loss": 0.12217042595148087, "global_step": 102087, "epoch": 1147, "lr": 6.914394601357698e-05}
{"train_loss": 0.13487502932548523, "global_step": 102088, "epoch": 1147, "lr": 6.914341045246558e-05}
{"train_loss": 0.14614999294281006, "global_step": 102089, "epoch": 1147, "lr": 6.914287488878055e-05}
{"train_loss": 0.14292767643928528, "global_step": 102090, "epoch": 1147, "lr": 6.914233932252201e-05}
{"train_loss": 0.16757413744926453, "global_step": 102091, "epoch": 1147, "lr": 6.914180375369001e-05}
{"train_loss": 0.11905301362276077, "global_step": 102092, "epoch": 1147, "lr": 6.914126818228461e-05}
{"train_loss": 0.15927881002426147, "global_step": 102093, "epoch": 1147, "lr": 6.914073260830591e-05}
{"train_loss": 0.2158779501914978, "global_step": 102094, "epoch": 1147, "lr": 6.914019703175398e-05}
{"train_loss": 0.10670331865549088, "global_step": 102095, "epoch": 1147, "lr": 6.913966145262889e-05}
{"train_loss": 0.15384720265865326, "global_step": 102096, "epoch": 1147, "lr": 6.913912587093069e-05}
{"train_loss": 0.21524670720100403, "global_step": 102097, "epoch": 1147, "lr": 6.913859028665949e-05}
{"train_loss": 0.22578337788581848, "global_step": 102098, "epoch": 1147, "lr": 6.913805469981533e-05}
{"train_loss": 0.08512279391288757, "global_step": 102099, "epoch": 1147, "lr": 6.913751911039829e-05}
{"train_loss": 0.14053115248680115, "global_step": 102100, "epoch": 1147, "lr": 6.913698351840844e-05}
{"train_loss": 0.2368321567773819, "global_step": 102101, "epoch": 1147, "lr": 6.913644792384587e-05}
{"train_loss": 0.1789827048778534, "global_step": 102102, "epoch": 1147, "lr": 6.913591232671062e-05}
{"train_loss": 0.07284141331911087, "global_step": 102103, "epoch": 1147, "lr": 6.913537672700278e-05}
{"train_loss": 0.16960017383098602, "global_step": 102104, "epoch": 1147, "lr": 6.913484112472242e-05}
{"train_loss": 0.15954473614692688, "global_step": 102105, "epoch": 1147, "lr": 6.913430551986962e-05}
{"train_loss": 0.13633863627910614, "global_step": 102106, "epoch": 1147, "lr": 6.913376991244446e-05}
{"train_loss": 0.22792991995811462, "global_step": 102107, "epoch": 1147, "lr": 6.9133234302447e-05}
{"train_loss": 0.18084049224853516, "global_step": 102108, "epoch": 1147, "lr": 6.913269868987729e-05}
{"train_loss": 0.0845809131860733, "global_step": 102109, "epoch": 1147, "lr": 6.913216307473543e-05}
{"train_loss": 0.15194784104824066, "global_step": 102110, "epoch": 1147, "lr": 6.913162745702147e-05}
{"train_loss": 0.15821990370750427, "global_step": 102111, "epoch": 1147, "lr": 6.91310918367355e-05}
{"train_loss": 0.17468056082725525, "global_step": 102112, "epoch": 1147, "lr": 6.913055621387759e-05}
{"train_loss": 0.1655270904302597, "global_step": 102113, "epoch": 1147, "lr": 6.913002058844781e-05}
{"train_loss": 0.2386915236711502, "global_step": 102114, "epoch": 1147, "lr": 6.912948496044624e-05}
{"train_loss": 0.12491666525602341, "global_step": 102115, "epoch": 1147, "lr": 6.912894932987295e-05}
{"train_loss": 0.2838478684425354, "global_step": 102116, "epoch": 1147, "lr": 6.912841369672799e-05}
{"train_loss": 0.1563444435596466, "global_step": 102117, "epoch": 1147, "lr": 6.912787806101145e-05}
{"train_loss": 0.13745756447315216, "global_step": 102118, "epoch": 1147, "lr": 6.912734242272341e-05}
{"train_loss": 0.1885639876127243, "global_step": 102119, "epoch": 1147, "lr": 6.912680678186393e-05}
{"train_loss": 0.1799461394548416, "global_step": 102120, "epoch": 1147, "lr": 6.912627113843308e-05}
{"train_loss": 0.15710066258907318, "global_step": 102121, "epoch": 1147, "lr": 6.912573549243093e-05}
{"train_loss": 0.19279654324054718, "global_step": 102122, "epoch": 1147, "lr": 6.912519984385757e-05}
{"train_loss": 0.20266059041023254, "global_step": 102123, "epoch": 1147, "lr": 6.912466419271305e-05}
{"train_loss": 0.18495896458625793, "global_step": 102124, "epoch": 1147, "lr": 6.912412853899745e-05}
{"train_loss": 0.2065362185239792, "global_step": 102125, "epoch": 1147, "lr": 6.912359288271086e-05}
{"train_loss": 0.261286199092865, "global_step": 102126, "epoch": 1147, "lr": 6.912305722385332e-05}
{"train_loss": 0.23085545003414154, "global_step": 102127, "epoch": 1147, "lr": 6.912252156242494e-05}
{"train_loss": 0.12754547595977783, "global_step": 102128, "epoch": 1147, "lr": 6.912198589842573e-05}
{"train_loss": 0.21851947903633118, "global_step": 102129, "epoch": 1147, "lr": 6.912145023185583e-05}
{"train_loss": 0.1197589784860611, "global_step": 102130, "epoch": 1147, "lr": 6.912091456271528e-05}
{"train_loss": 0.194383904337883, "global_step": 102131, "epoch": 1147, "lr": 6.912037889100416e-05}
{"train_loss": 0.24327868223190308, "global_step": 102132, "epoch": 1147, "lr": 6.911984321672253e-05}
{"train_loss": 0.12568414211273193, "global_step": 102133, "epoch": 1147, "lr": 6.911930753987048e-05}
{"train_loss": 0.2020687460899353, "global_step": 102134, "epoch": 1147, "lr": 6.911877186044806e-05}
{"train_loss": 0.1616116762161255, "global_step": 102135, "epoch": 1147, "lr": 6.911823617845536e-05}
{"train_loss": 0.2098730355501175, "global_step": 102136, "epoch": 1147, "lr": 6.911770049389246e-05}
{"train_loss": 0.2064896523952484, "global_step": 102137, "epoch": 1147, "lr": 6.91171648067594e-05}
{"train_loss": 0.1319892257452011, "global_step": 102138, "epoch": 1147, "lr": 6.911662911705628e-05}
{"train_loss": 0.12029079347848892, "global_step": 102139, "epoch": 1147, "lr": 6.911609342478316e-05}
{"train_loss": 0.15391552448272705, "global_step": 102140, "epoch": 1147, "lr": 6.91155577299401e-05}
{"train_loss": 0.26076966524124146, "global_step": 102141, "epoch": 1147, "lr": 6.91150220325272e-05}
{"train_loss": 0.13038095831871033, "global_step": 102142, "epoch": 1147, "lr": 6.911448633254451e-05}
{"train_loss": 0.12494014203548431, "global_step": 102143, "epoch": 1147, "lr": 6.911395062999212e-05}
{"train_loss": 0.18885932862758636, "global_step": 102144, "epoch": 1147, "lr": 6.911341492487008e-05}
{"train_loss": 0.1545369178056717, "global_step": 102145, "epoch": 1147, "lr": 6.91128792171785e-05}
{"train_loss": 0.2526378929615021, "global_step": 102146, "epoch": 1147, "lr": 6.911234350691741e-05}
{"train_loss": 0.21758130192756653, "global_step": 102147, "epoch": 1147, "lr": 6.91118077940869e-05}
{"train_loss": 0.17000019550323486, "global_step": 102148, "epoch": 1147, "lr": 6.911127207868702e-05}
{"train_loss": 0.19129319489002228, "global_step": 102149, "epoch": 1147, "lr": 6.911073636071789e-05}
{"train_loss": 0.1089620441198349, "global_step": 102150, "epoch": 1147, "lr": 6.911020064017954e-05}
{"train_loss": 0.1408851146697998, "global_step": 102151, "epoch": 1147, "lr": 6.910966491707206e-05}
{"train_loss": 0.09516537189483643, "global_step": 102152, "epoch": 1147, "lr": 6.910912919139553e-05}
{"train_loss": 0.19231683015823364, "global_step": 102153, "epoch": 1147, "lr": 6.910859346314999e-05}
{"train_loss": 0.18226027488708496, "global_step": 102154, "epoch": 1147, "lr": 6.910805773233556e-05}
{"train_loss": 0.21720722317695618, "global_step": 102155, "epoch": 1147, "lr": 6.910752199895226e-05}
{"train_loss": 0.17434585094451904, "global_step": 102156, "epoch": 1147, "lr": 6.910698626300018e-05}
{"train_loss": 0.18500596284866333, "global_step": 102157, "epoch": 1147, "lr": 6.910645052447943e-05}
{"train_loss": 0.10554187744855881, "global_step": 102158, "epoch": 1147, "lr": 6.910591478339003e-05}
{"train_loss": 0.16301143169403076, "global_step": 102159, "epoch": 1147, "lr": 6.910537903973208e-05}
{"train_loss": 0.14634360373020172, "global_step": 102160, "epoch": 1147, "lr": 6.910484329350564e-05}
{"train_loss": 0.22636838257312775, "global_step": 102161, "epoch": 1147, "lr": 6.910430754471078e-05}
{"train_loss": 0.16364271938800812, "global_step": 102162, "epoch": 1147, "lr": 6.910377179334758e-05}
{"train_loss": 0.1892085075378418, "global_step": 102163, "epoch": 1147, "lr": 6.910323603941611e-05}
{"train_loss": 0.16427937150001526, "global_step": 102164, "epoch": 1147, "lr": 6.910270028291646e-05}
{"train_loss": 0.16578702628612518, "global_step": 102165, "epoch": 1147, "lr": 6.910216452384868e-05}
{"train_loss": 0.14039306342601776, "global_step": 102166, "epoch": 1147, "lr": 6.910162876221283e-05}
{"train_loss": 0.17044037580490112, "global_step": 102167, "epoch": 1147, "lr": 6.910109299800902e-05}
{"train_loss": 0.1300646811723709, "global_step": 102168, "epoch": 1147, "lr": 6.910055723123728e-05}
{"train_loss": 0.2377789467573166, "global_step": 102169, "epoch": 1147, "lr": 6.910002146189773e-05}
{"train_loss": 0.17282792925834656, "global_step": 102170, "epoch": 1147, "lr": 6.909948568999038e-05}
{"train_loss": 0.16929143457935097, "global_step": 102171, "epoch": 1147, "lr": 6.909894991551537e-05, "val_loss": 4.798773288726807}
{"train_loss": 0.15100520849227905, "global_step": 102172, "epoch": 1148, "lr": 6.909841413847271e-05}
{"train_loss": 0.2402675747871399, "global_step": 102173, "epoch": 1148, "lr": 6.909787835886252e-05}
{"train_loss": 0.18871460855007172, "global_step": 102174, "epoch": 1148, "lr": 6.909734257668485e-05}
{"train_loss": 0.16000446677207947, "global_step": 102175, "epoch": 1148, "lr": 6.909680679193978e-05}
{"train_loss": 0.2283412367105484, "global_step": 102176, "epoch": 1148, "lr": 6.909627100462737e-05}
{"train_loss": 0.2335805892944336, "global_step": 102177, "epoch": 1148, "lr": 6.909573521474769e-05}
{"train_loss": 0.21795524656772614, "global_step": 102178, "epoch": 1148, "lr": 6.909519942230083e-05}
{"train_loss": 0.15243475139141083, "global_step": 102179, "epoch": 1148, "lr": 6.909466362728686e-05}
{"train_loss": 0.2139064222574234, "global_step": 102180, "epoch": 1148, "lr": 6.909412782970583e-05}
{"train_loss": 0.143888920545578, "global_step": 102181, "epoch": 1148, "lr": 6.909359202955783e-05}
{"train_loss": 0.19075876474380493, "global_step": 102182, "epoch": 1148, "lr": 6.909305622684295e-05}
{"train_loss": 0.1382600963115692, "global_step": 102183, "epoch": 1148, "lr": 6.909252042156123e-05}
{"train_loss": 0.14439915120601654, "global_step": 102184, "epoch": 1148, "lr": 6.909198461371274e-05}
{"train_loss": 0.15742573142051697, "global_step": 102185, "epoch": 1148, "lr": 6.909144880329759e-05}
{"train_loss": 0.12769007682800293, "global_step": 102186, "epoch": 1148, "lr": 6.90909129903158e-05}
{"train_loss": 0.11940403282642365, "global_step": 102187, "epoch": 1148, "lr": 6.909037717476748e-05}
{"train_loss": 0.16476868093013763, "global_step": 102188, "epoch": 1148, "lr": 6.908984135665271e-05}
{"train_loss": 0.1379712074995041, "global_step": 102189, "epoch": 1148, "lr": 6.908930553597151e-05}
{"train_loss": 0.1448706090450287, "global_step": 102190, "epoch": 1148, "lr": 6.9088769712724e-05}
{"train_loss": 0.12338533997535706, "global_step": 102191, "epoch": 1148, "lr": 6.908823388691026e-05}
{"train_loss": 0.16228517889976501, "global_step": 102192, "epoch": 1148, "lr": 6.908769805853032e-05}
{"train_loss": 0.198256716132164, "global_step": 102193, "epoch": 1148, "lr": 6.908716222758427e-05}
{"train_loss": 0.13307389616966248, "global_step": 102194, "epoch": 1148, "lr": 6.908662639407217e-05}
{"train_loss": 0.15877167880535126, "global_step": 102195, "epoch": 1148, "lr": 6.908609055799414e-05}
{"train_loss": 0.2244470715522766, "global_step": 102196, "epoch": 1148, "lr": 6.908555471935019e-05}
{"train_loss": 0.11682115495204926, "global_step": 102197, "epoch": 1148, "lr": 6.908501887814043e-05}
{"train_loss": 0.08609874546527863, "global_step": 102198, "epoch": 1148, "lr": 6.908448303436493e-05}
{"train_loss": 0.16968649625778198, "global_step": 102199, "epoch": 1148, "lr": 6.908394718802374e-05}
{"train_loss": 0.15428729355335236, "global_step": 102200, "epoch": 1148, "lr": 6.908341133911695e-05}
{"train_loss": 0.1233101412653923, "global_step": 102201, "epoch": 1148, "lr": 6.908287548764463e-05}
{"train_loss": 0.29918283224105835, "global_step": 102202, "epoch": 1148, "lr": 6.908233963360684e-05}
{"train_loss": 0.10754410177469254, "global_step": 102203, "epoch": 1148, "lr": 6.908180377700367e-05}
{"train_loss": 0.15329952538013458, "global_step": 102204, "epoch": 1148, "lr": 6.908126791783519e-05}
{"train_loss": 0.1573614925146103, "global_step": 102205, "epoch": 1148, "lr": 6.908073205610146e-05}
{"train_loss": 0.1973574012517929, "global_step": 102206, "epoch": 1148, "lr": 6.908019619180255e-05}
{"train_loss": 0.1512296497821808, "global_step": 102207, "epoch": 1148, "lr": 6.907966032493856e-05}
{"train_loss": 0.20021380484104156, "global_step": 102208, "epoch": 1148, "lr": 6.907912445550953e-05}
{"train_loss": 0.26123616099357605, "global_step": 102209, "epoch": 1148, "lr": 6.907858858351554e-05}
{"train_loss": 0.14521919190883636, "global_step": 102210, "epoch": 1148, "lr": 6.907805270895667e-05}
{"train_loss": 0.10793566703796387, "global_step": 102211, "epoch": 1148, "lr": 6.907751683183299e-05}
{"train_loss": 0.1519141048192978, "global_step": 102212, "epoch": 1148, "lr": 6.907698095214456e-05}
{"train_loss": 0.1022363007068634, "global_step": 102213, "epoch": 1148, "lr": 6.907644506989148e-05}
{"train_loss": 0.20182198286056519, "global_step": 102214, "epoch": 1148, "lr": 6.90759091850738e-05}
{"train_loss": 0.11575309932231903, "global_step": 102215, "epoch": 1148, "lr": 6.90753732976916e-05}
{"train_loss": 0.2073267698287964, "global_step": 102216, "epoch": 1148, "lr": 6.907483740774494e-05}
{"train_loss": 0.17747953534126282, "global_step": 102217, "epoch": 1148, "lr": 6.907430151523389e-05}
{"train_loss": 0.15257859230041504, "global_step": 102218, "epoch": 1148, "lr": 6.907376562015855e-05}
{"train_loss": 0.08839806914329529, "global_step": 102219, "epoch": 1148, "lr": 6.907322972251897e-05}
{"train_loss": 0.16379724442958832, "global_step": 102220, "epoch": 1148, "lr": 6.907269382231523e-05}
{"train_loss": 0.1732129156589508, "global_step": 102221, "epoch": 1148, "lr": 6.90721579195474e-05}
{"train_loss": 0.14672793447971344, "global_step": 102222, "epoch": 1148, "lr": 6.907162201421554e-05}
{"train_loss": 0.11726228147745132, "global_step": 102223, "epoch": 1148, "lr": 6.907108610631974e-05}
{"train_loss": 0.15571990609169006, "global_step": 102224, "epoch": 1148, "lr": 6.907055019586006e-05}
{"train_loss": 0.10986924171447754, "global_step": 102225, "epoch": 1148, "lr": 6.907001428283659e-05}
{"train_loss": 0.18845447897911072, "global_step": 102226, "epoch": 1148, "lr": 6.906947836724938e-05}
{"train_loss": 0.16232912242412567, "global_step": 102227, "epoch": 1148, "lr": 6.906894244909851e-05}
{"train_loss": 0.2166169434785843, "global_step": 102228, "epoch": 1148, "lr": 6.906840652838406e-05}
{"train_loss": 0.17009080946445465, "global_step": 102229, "epoch": 1148, "lr": 6.906787060510608e-05}
{"train_loss": 0.11579027026891708, "global_step": 102230, "epoch": 1148, "lr": 6.906733467926467e-05}
{"train_loss": 0.112673819065094, "global_step": 102231, "epoch": 1148, "lr": 6.906679875085988e-05}
{"train_loss": 0.14843422174453735, "global_step": 102232, "epoch": 1148, "lr": 6.90662628198918e-05}
{"train_loss": 0.17387372255325317, "global_step": 102233, "epoch": 1148, "lr": 6.906572688636049e-05}
{"train_loss": 0.16402368247509003, "global_step": 102234, "epoch": 1148, "lr": 6.906519095026602e-05}
{"train_loss": 0.1318548619747162, "global_step": 102235, "epoch": 1148, "lr": 6.906465501160848e-05}
{"train_loss": 0.19836880266666412, "global_step": 102236, "epoch": 1148, "lr": 6.906411907038792e-05}
{"train_loss": 0.10635767877101898, "global_step": 102237, "epoch": 1148, "lr": 6.906358312660443e-05}
{"train_loss": 0.19984284043312073, "global_step": 102238, "epoch": 1148, "lr": 6.906304718025807e-05}
{"train_loss": 0.0996706634759903, "global_step": 102239, "epoch": 1148, "lr": 6.90625112313489e-05}
{"train_loss": 0.10291478037834167, "global_step": 102240, "epoch": 1148, "lr": 6.906197527987704e-05}
{"train_loss": 0.19082437455654144, "global_step": 102241, "epoch": 1148, "lr": 6.90614393258425e-05}
{"train_loss": 0.15012669563293457, "global_step": 102242, "epoch": 1148, "lr": 6.90609033692454e-05}
{"train_loss": 0.2148396521806717, "global_step": 102243, "epoch": 1148, "lr": 6.906036741008577e-05}
{"train_loss": 0.2408754825592041, "global_step": 102244, "epoch": 1148, "lr": 6.905983144836374e-05}
{"train_loss": 0.19583724439144135, "global_step": 102245, "epoch": 1148, "lr": 6.905929548407932e-05}
{"train_loss": 0.14915770292282104, "global_step": 102246, "epoch": 1148, "lr": 6.905875951723262e-05}
{"train_loss": 0.10600036382675171, "global_step": 102247, "epoch": 1148, "lr": 6.905822354782371e-05}
{"train_loss": 0.16697387397289276, "global_step": 102248, "epoch": 1148, "lr": 6.905768757585264e-05}
{"train_loss": 0.08965229988098145, "global_step": 102249, "epoch": 1148, "lr": 6.905715160131951e-05}
{"train_loss": 0.18657250702381134, "global_step": 102250, "epoch": 1148, "lr": 6.905661562422438e-05}
{"train_loss": 0.1579040288925171, "global_step": 102251, "epoch": 1148, "lr": 6.90560796445673e-05}
{"train_loss": 0.10048577934503555, "global_step": 102252, "epoch": 1148, "lr": 6.905554366234837e-05}
{"train_loss": 0.15991917252540588, "global_step": 102253, "epoch": 1148, "lr": 6.905500767756766e-05}
{"train_loss": 0.11343167722225189, "global_step": 102254, "epoch": 1148, "lr": 6.905447169022524e-05}
{"train_loss": 0.07707693427801132, "global_step": 102255, "epoch": 1148, "lr": 6.905393570032116e-05}
{"train_loss": 0.14866769313812256, "global_step": 102256, "epoch": 1148, "lr": 6.905339970785552e-05}
{"train_loss": 0.15705497562885284, "global_step": 102257, "epoch": 1148, "lr": 6.90528637128284e-05}
{"train_loss": 0.2066524177789688, "global_step": 102258, "epoch": 1148, "lr": 6.905232771523984e-05}
{"train_loss": 0.12392178177833557, "global_step": 102259, "epoch": 1148, "lr": 6.905179171508993e-05}
{"train_loss": 0.15953264410576123, "global_step": 102260, "epoch": 1148, "lr": 6.905125571237873e-05, "val_loss": 4.636826992034912}
{"train_loss": 0.18001443147659302, "global_step": 102261, "epoch": 1149, "lr": 6.905071970710634e-05}
{"train_loss": 0.0904984176158905, "global_step": 102262, "epoch": 1149, "lr": 6.905018369927279e-05}
{"train_loss": 0.15422095358371735, "global_step": 102263, "epoch": 1149, "lr": 6.90496476888782e-05}
{"train_loss": 0.1460610181093216, "global_step": 102264, "epoch": 1149, "lr": 6.90491116759226e-05}
{"train_loss": 0.20643308758735657, "global_step": 102265, "epoch": 1149, "lr": 6.904857566040608e-05}
{"train_loss": 0.12885646522045135, "global_step": 102266, "epoch": 1149, "lr": 6.90480396423287e-05}
{"train_loss": 0.1434132307767868, "global_step": 102267, "epoch": 1149, "lr": 6.904750362169057e-05}
{"train_loss": 0.12194610387086868, "global_step": 102268, "epoch": 1149, "lr": 6.904696759849172e-05}
{"train_loss": 0.13779553771018982, "global_step": 102269, "epoch": 1149, "lr": 6.904643157273224e-05}
{"train_loss": 0.142744243144989, "global_step": 102270, "epoch": 1149, "lr": 6.90458955444122e-05}
{"train_loss": 0.1509713977575302, "global_step": 102271, "epoch": 1149, "lr": 6.904535951353167e-05}
{"train_loss": 0.15713368356227875, "global_step": 102272, "epoch": 1149, "lr": 6.904482348009072e-05}
{"train_loss": 0.11330083012580872, "global_step": 102273, "epoch": 1149, "lr": 6.904428744408943e-05}
{"train_loss": 0.1899890899658203, "global_step": 102274, "epoch": 1149, "lr": 6.904375140552786e-05}
{"train_loss": 0.17256753146648407, "global_step": 102275, "epoch": 1149, "lr": 6.904321536440609e-05}
{"train_loss": 0.1339026838541031, "global_step": 102276, "epoch": 1149, "lr": 6.904267932072422e-05}
{"train_loss": 0.1586640328168869, "global_step": 102277, "epoch": 1149, "lr": 6.904214327448227e-05}
{"train_loss": 0.22149431705474854, "global_step": 102278, "epoch": 1149, "lr": 6.904160722568033e-05}
{"train_loss": 0.12146688997745514, "global_step": 102279, "epoch": 1149, "lr": 6.904107117431848e-05}
{"train_loss": 0.14239586889743805, "global_step": 102280, "epoch": 1149, "lr": 6.904053512039681e-05}
{"train_loss": 0.27166011929512024, "global_step": 102281, "epoch": 1149, "lr": 6.903999906391535e-05}
{"train_loss": 0.09672094136476517, "global_step": 102282, "epoch": 1149, "lr": 6.90394630048742e-05}
{"train_loss": 0.13595318794250488, "global_step": 102283, "epoch": 1149, "lr": 6.903892694327343e-05}
{"train_loss": 0.1410827338695526, "global_step": 102284, "epoch": 1149, "lr": 6.903839087911312e-05}
{"train_loss": 0.1682068109512329, "global_step": 102285, "epoch": 1149, "lr": 6.903785481239331e-05}
{"train_loss": 0.18415646255016327, "global_step": 102286, "epoch": 1149, "lr": 6.90373187431141e-05}
{"train_loss": 0.192769855260849, "global_step": 102287, "epoch": 1149, "lr": 6.903678267127556e-05}
{"train_loss": 0.2425725907087326, "global_step": 102288, "epoch": 1149, "lr": 6.903624659687775e-05}
{"train_loss": 0.16880390048027039, "global_step": 102289, "epoch": 1149, "lr": 6.903571051992074e-05}
{"train_loss": 0.21640470623970032, "global_step": 102290, "epoch": 1149, "lr": 6.903517444040462e-05}
{"train_loss": 0.17557549476623535, "global_step": 102291, "epoch": 1149, "lr": 6.903463835832946e-05}
{"train_loss": 0.1847267746925354, "global_step": 102292, "epoch": 1149, "lr": 6.90341022736953e-05}
{"train_loss": 0.10511580854654312, "global_step": 102293, "epoch": 1149, "lr": 6.903356618650225e-05}
{"train_loss": 0.18770591914653778, "global_step": 102294, "epoch": 1149, "lr": 6.903303009675038e-05}
{"train_loss": 0.1541796326637268, "global_step": 102295, "epoch": 1149, "lr": 6.903249400443976e-05}
{"train_loss": 0.22912955284118652, "global_step": 102296, "epoch": 1149, "lr": 6.903195790957043e-05}
{"train_loss": 0.18785454332828522, "global_step": 102297, "epoch": 1149, "lr": 6.903142181214249e-05}
{"train_loss": 0.2628057599067688, "global_step": 102298, "epoch": 1149, "lr": 6.903088571215601e-05}
{"train_loss": 0.1351148933172226, "global_step": 102299, "epoch": 1149, "lr": 6.903034960961107e-05}
{"train_loss": 0.22853919863700867, "global_step": 102300, "epoch": 1149, "lr": 6.902981350450773e-05}
{"train_loss": 0.18872463703155518, "global_step": 102301, "epoch": 1149, "lr": 6.902927739684605e-05}
{"train_loss": 0.1531040072441101, "global_step": 102302, "epoch": 1149, "lr": 6.902874128662612e-05}
{"train_loss": 0.1724320650100708, "global_step": 102303, "epoch": 1149, "lr": 6.902820517384801e-05}
{"train_loss": 0.17473450303077698, "global_step": 102304, "epoch": 1149, "lr": 6.902766905851177e-05}
{"train_loss": 0.15922296047210693, "global_step": 102305, "epoch": 1149, "lr": 6.902713294061753e-05}
{"train_loss": 0.188339963555336, "global_step": 102306, "epoch": 1149, "lr": 6.90265968201653e-05}
{"train_loss": 0.16094903647899628, "global_step": 102307, "epoch": 1149, "lr": 6.902606069715519e-05}
{"train_loss": 0.22386595606803894, "global_step": 102308, "epoch": 1149, "lr": 6.902552457158723e-05}
{"train_loss": 0.14412571489810944, "global_step": 102309, "epoch": 1149, "lr": 6.902498844346154e-05}
{"train_loss": 0.1370551735162735, "global_step": 102310, "epoch": 1149, "lr": 6.902445231277817e-05}
{"train_loss": 0.0972718745470047, "global_step": 102311, "epoch": 1149, "lr": 6.902391617953719e-05}
{"train_loss": 0.22505217790603638, "global_step": 102312, "epoch": 1149, "lr": 6.902338004373867e-05}
{"train_loss": 0.18254265189170837, "global_step": 102313, "epoch": 1149, "lr": 6.90228439053827e-05}
{"train_loss": 0.19338060915470123, "global_step": 102314, "epoch": 1149, "lr": 6.902230776446934e-05}
{"train_loss": 0.20144686102867126, "global_step": 102315, "epoch": 1149, "lr": 6.902177162099868e-05}
{"train_loss": 0.2529871165752411, "global_step": 102316, "epoch": 1149, "lr": 6.902123547497074e-05}
{"train_loss": 0.1681329905986786, "global_step": 102317, "epoch": 1149, "lr": 6.902069932638566e-05}
{"train_loss": 0.09361889958381653, "global_step": 102318, "epoch": 1149, "lr": 6.902016317524346e-05}
{"train_loss": 0.12158370018005371, "global_step": 102319, "epoch": 1149, "lr": 6.901962702154424e-05}
{"train_loss": 0.14538581669330597, "global_step": 102320, "epoch": 1149, "lr": 6.901909086528806e-05}
{"train_loss": 0.15701863169670105, "global_step": 102321, "epoch": 1149, "lr": 6.901855470647498e-05}
{"train_loss": 0.11325787007808685, "global_step": 102322, "epoch": 1149, "lr": 6.90180185451051e-05}
{"train_loss": 0.1620512157678604, "global_step": 102323, "epoch": 1149, "lr": 6.901748238117848e-05}
{"train_loss": 0.15619702637195587, "global_step": 102324, "epoch": 1149, "lr": 6.901694621469518e-05}
{"train_loss": 0.18381674587726593, "global_step": 102325, "epoch": 1149, "lr": 6.901641004565529e-05}
{"train_loss": 0.17521046102046967, "global_step": 102326, "epoch": 1149, "lr": 6.901587387405888e-05}
{"train_loss": 0.1392841339111328, "global_step": 102327, "epoch": 1149, "lr": 6.901533769990602e-05}
{"train_loss": 0.15671660006046295, "global_step": 102328, "epoch": 1149, "lr": 6.901480152319676e-05}
{"train_loss": 0.13373953104019165, "global_step": 102329, "epoch": 1149, "lr": 6.901426534393121e-05}
{"train_loss": 0.19830475747585297, "global_step": 102330, "epoch": 1149, "lr": 6.901372916210942e-05}
{"train_loss": 0.14719578623771667, "global_step": 102331, "epoch": 1149, "lr": 6.901319297773146e-05}
{"train_loss": 0.15324683487415314, "global_step": 102332, "epoch": 1149, "lr": 6.901265679079743e-05}
{"train_loss": 0.18754646182060242, "global_step": 102333, "epoch": 1149, "lr": 6.901212060130735e-05}
{"train_loss": 0.11881177872419357, "global_step": 102334, "epoch": 1149, "lr": 6.901158440926135e-05}
{"train_loss": 0.1778346300125122, "global_step": 102335, "epoch": 1149, "lr": 6.901104821465946e-05}
{"train_loss": 0.16908249258995056, "global_step": 102336, "epoch": 1149, "lr": 6.901051201750177e-05}
{"train_loss": 0.1385544240474701, "global_step": 102337, "epoch": 1149, "lr": 6.900997581778834e-05}
{"train_loss": 0.14298571646213531, "global_step": 102338, "epoch": 1149, "lr": 6.900943961551926e-05}
{"train_loss": 0.1607261300086975, "global_step": 102339, "epoch": 1149, "lr": 6.900890341069457e-05}
{"train_loss": 0.1273912936449051, "global_step": 102340, "epoch": 1149, "lr": 6.90083672033144e-05}
{"train_loss": 0.12274790555238724, "global_step": 102341, "epoch": 1149, "lr": 6.900783099337876e-05}
{"train_loss": 0.19686609506607056, "global_step": 102342, "epoch": 1149, "lr": 6.900729478088777e-05}
{"train_loss": 0.12934555113315582, "global_step": 102343, "epoch": 1149, "lr": 6.900675856584147e-05}
{"train_loss": 0.16305193305015564, "global_step": 102344, "epoch": 1149, "lr": 6.900622234823994e-05}
{"train_loss": 0.09059514105319977, "global_step": 102345, "epoch": 1149, "lr": 6.900568612808324e-05}
{"train_loss": 0.15701626241207123, "global_step": 102346, "epoch": 1149, "lr": 6.900514990537148e-05}
{"train_loss": 0.12101442366838455, "global_step": 102347, "epoch": 1149, "lr": 6.90046136801047e-05}
{"train_loss": 0.19767005741596222, "global_step": 102348, "epoch": 1149, "lr": 6.900407745228298e-05}
{"train_loss": 0.16298302388593053, "global_step": 102349, "epoch": 1149, "lr": 6.900354122190639e-05, "val_loss": 4.748807430267334}
{"train_loss": 0.149323508143425, "global_step": 102350, "epoch": 1150, "lr": 6.900300498897503e-05}
{"train_loss": 0.11250803619623184, "global_step": 102351, "epoch": 1150, "lr": 6.900246875348892e-05}
{"train_loss": 0.1257134974002838, "global_step": 102352, "epoch": 1150, "lr": 6.900193251544817e-05}
{"train_loss": 0.21805156767368317, "global_step": 102353, "epoch": 1150, "lr": 6.900139627485286e-05}
{"train_loss": 0.13980528712272644, "global_step": 102354, "epoch": 1150, "lr": 6.900086003170302e-05}
{"train_loss": 0.22891753911972046, "global_step": 102355, "epoch": 1150, "lr": 6.900032378599877e-05}
{"train_loss": 0.13998475670814514, "global_step": 102356, "epoch": 1150, "lr": 6.899978753774012e-05}
{"train_loss": 0.16061627864837646, "global_step": 102357, "epoch": 1150, "lr": 6.899925128692721e-05}
{"train_loss": 0.13782663643360138, "global_step": 102358, "epoch": 1150, "lr": 6.899871503356008e-05}
{"train_loss": 0.21578526496887207, "global_step": 102359, "epoch": 1150, "lr": 6.899817877763879e-05}
{"train_loss": 0.1365463137626648, "global_step": 102360, "epoch": 1150, "lr": 6.899764251916343e-05}
{"train_loss": 0.12467554211616516, "global_step": 102361, "epoch": 1150, "lr": 6.899710625813408e-05}
{"train_loss": 0.12812258303165436, "global_step": 102362, "epoch": 1150, "lr": 6.899656999455079e-05}
{"train_loss": 0.156904399394989, "global_step": 102363, "epoch": 1150, "lr": 6.899603372841365e-05}
{"train_loss": 0.18031635880470276, "global_step": 102364, "epoch": 1150, "lr": 6.89954974597227e-05}
{"train_loss": 0.16662178933620453, "global_step": 102365, "epoch": 1150, "lr": 6.899496118847806e-05}
{"train_loss": 0.16354118287563324, "global_step": 102366, "epoch": 1150, "lr": 6.899442491467977e-05}
{"train_loss": 0.16717401146888733, "global_step": 102367, "epoch": 1150, "lr": 6.899388863832792e-05}
{"train_loss": 0.11760130524635315, "global_step": 102368, "epoch": 1150, "lr": 6.899335235942257e-05}
{"train_loss": 0.10794348269701004, "global_step": 102369, "epoch": 1150, "lr": 6.89928160779638e-05}
{"train_loss": 0.2418043613433838, "global_step": 102370, "epoch": 1150, "lr": 6.899227979395167e-05}
{"train_loss": 0.14465394616127014, "global_step": 102371, "epoch": 1150, "lr": 6.899174350738626e-05}
{"train_loss": 0.13076463341712952, "global_step": 102372, "epoch": 1150, "lr": 6.899120721826765e-05}
{"train_loss": 0.10790364444255829, "global_step": 102373, "epoch": 1150, "lr": 6.899067092659591e-05}
{"train_loss": 0.18867579102516174, "global_step": 102374, "epoch": 1150, "lr": 6.899013463237109e-05}
{"train_loss": 0.2268216460943222, "global_step": 102375, "epoch": 1150, "lr": 6.898959833559328e-05}
{"train_loss": 0.10707802325487137, "global_step": 102376, "epoch": 1150, "lr": 6.898906203626255e-05}
{"train_loss": 0.21496617794036865, "global_step": 102377, "epoch": 1150, "lr": 6.898852573437898e-05}
{"train_loss": 0.13514789938926697, "global_step": 102378, "epoch": 1150, "lr": 6.898798942994262e-05}
{"train_loss": 0.17382654547691345, "global_step": 102379, "epoch": 1150, "lr": 6.898745312295357e-05}
{"train_loss": 0.15578971803188324, "global_step": 102380, "epoch": 1150, "lr": 6.898691681341188e-05}
{"train_loss": 0.25734826922416687, "global_step": 102381, "epoch": 1150, "lr": 6.898638050131764e-05}
{"train_loss": 0.15985217690467834, "global_step": 102382, "epoch": 1150, "lr": 6.898584418667089e-05}
{"train_loss": 0.2920691967010498, "global_step": 102383, "epoch": 1150, "lr": 6.898530786947174e-05}
{"train_loss": 0.16724835336208344, "global_step": 102384, "epoch": 1150, "lr": 6.898477154972025e-05}
{"train_loss": 0.22004802525043488, "global_step": 102385, "epoch": 1150, "lr": 6.898423522741649e-05}
{"train_loss": 0.25904372334480286, "global_step": 102386, "epoch": 1150, "lr": 6.898369890256051e-05}
{"train_loss": 0.1455443799495697, "global_step": 102387, "epoch": 1150, "lr": 6.898316257515242e-05}
{"train_loss": 0.14388082921504974, "global_step": 102388, "epoch": 1150, "lr": 6.898262624519228e-05}
{"train_loss": 0.19117160141468048, "global_step": 102389, "epoch": 1150, "lr": 6.898208991268016e-05}
{"train_loss": 0.21294580399990082, "global_step": 102390, "epoch": 1150, "lr": 6.898155357761612e-05}
{"train_loss": 0.17620056867599487, "global_step": 102391, "epoch": 1150, "lr": 6.898101724000024e-05}
{"train_loss": 0.12053979188203812, "global_step": 102392, "epoch": 1150, "lr": 6.89804808998326e-05}
{"train_loss": 0.1688750833272934, "global_step": 102393, "epoch": 1150, "lr": 6.897994455711326e-05}
{"train_loss": 0.13490919768810272, "global_step": 102394, "epoch": 1150, "lr": 6.89794082118423e-05}
{"train_loss": 0.12352686375379562, "global_step": 102395, "epoch": 1150, "lr": 6.897887186401978e-05}
{"train_loss": 0.18143145740032196, "global_step": 102396, "epoch": 1150, "lr": 6.897833551364581e-05}
{"train_loss": 0.11926368623971939, "global_step": 102397, "epoch": 1150, "lr": 6.897779916072042e-05}
{"train_loss": 0.1868128925561905, "global_step": 102398, "epoch": 1150, "lr": 6.897726280524369e-05}
{"train_loss": 0.1880475878715515, "global_step": 102399, "epoch": 1150, "lr": 6.89767264472157e-05}
{"train_loss": 0.16020624339580536, "global_step": 102400, "epoch": 1150, "lr": 6.897619008663652e-05}
{"train_loss": 0.15004467964172363, "global_step": 102401, "epoch": 1150, "lr": 6.89756537235062e-05}
{"train_loss": 0.12240344285964966, "global_step": 102402, "epoch": 1150, "lr": 6.897511735782487e-05}
{"train_loss": 0.1415853202342987, "global_step": 102403, "epoch": 1150, "lr": 6.897458098959256e-05}
{"train_loss": 0.15226109325885773, "global_step": 102404, "epoch": 1150, "lr": 6.897404461880933e-05}
{"train_loss": 0.23554806411266327, "global_step": 102405, "epoch": 1150, "lr": 6.897350824547528e-05}
{"train_loss": 0.09365507960319519, "global_step": 102406, "epoch": 1150, "lr": 6.897297186959049e-05}
{"train_loss": 0.13904701173305511, "global_step": 102407, "epoch": 1150, "lr": 6.8972435491155e-05}
{"train_loss": 0.16143731772899628, "global_step": 102408, "epoch": 1150, "lr": 6.89718991101689e-05}
{"train_loss": 0.11740980297327042, "global_step": 102409, "epoch": 1150, "lr": 6.897136272663226e-05}
{"train_loss": 0.15123960375785828, "global_step": 102410, "epoch": 1150, "lr": 6.897082634054514e-05}
{"train_loss": 0.14178530871868134, "global_step": 102411, "epoch": 1150, "lr": 6.897028995190766e-05}
{"train_loss": 0.11171168088912964, "global_step": 102412, "epoch": 1150, "lr": 6.896975356071982e-05}
{"train_loss": 0.15119455754756927, "global_step": 102413, "epoch": 1150, "lr": 6.896921716698175e-05}
{"train_loss": 0.25107091665267944, "global_step": 102414, "epoch": 1150, "lr": 6.896868077069347e-05}
{"train_loss": 0.2138676941394806, "global_step": 102415, "epoch": 1150, "lr": 6.896814437185511e-05}
{"train_loss": 0.2034754604101181, "global_step": 102416, "epoch": 1150, "lr": 6.896760797046671e-05}
{"train_loss": 0.12970302999019623, "global_step": 102417, "epoch": 1150, "lr": 6.896707156652833e-05}
{"train_loss": 0.20588834583759308, "global_step": 102418, "epoch": 1150, "lr": 6.896653516004007e-05}
{"train_loss": 0.2126922905445099, "global_step": 102419, "epoch": 1150, "lr": 6.896599875100199e-05}
{"train_loss": 0.1119939386844635, "global_step": 102420, "epoch": 1150, "lr": 6.896546233941415e-05}
{"train_loss": 0.18403995037078857, "global_step": 102421, "epoch": 1150, "lr": 6.896492592527666e-05}
{"train_loss": 0.1489468365907669, "global_step": 102422, "epoch": 1150, "lr": 6.896438950858954e-05}
{"train_loss": 0.18154725432395935, "global_step": 102423, "epoch": 1150, "lr": 6.89638530893529e-05}
{"train_loss": 0.14259284734725952, "global_step": 102424, "epoch": 1150, "lr": 6.896331666756681e-05}
{"train_loss": 0.15276461839675903, "global_step": 102425, "epoch": 1150, "lr": 6.896278024323133e-05}
{"train_loss": 0.2146243304014206, "global_step": 102426, "epoch": 1150, "lr": 6.896224381634653e-05}
{"train_loss": 0.16175228357315063, "global_step": 102427, "epoch": 1150, "lr": 6.89617073869125e-05}
{"train_loss": 0.08871733397245407, "global_step": 102428, "epoch": 1150, "lr": 6.896117095492928e-05}
{"train_loss": 0.11654055863618851, "global_step": 102429, "epoch": 1150, "lr": 6.896063452039698e-05}
{"train_loss": 0.15317799150943756, "global_step": 102430, "epoch": 1150, "lr": 6.896009808331564e-05}
{"train_loss": 0.19145898520946503, "global_step": 102431, "epoch": 1150, "lr": 6.895956164368535e-05}
{"train_loss": 0.1888345181941986, "global_step": 102432, "epoch": 1150, "lr": 6.89590252015062e-05}
{"train_loss": 0.1368199735879898, "global_step": 102433, "epoch": 1150, "lr": 6.89584887567782e-05}
{"train_loss": 0.11884785443544388, "global_step": 102434, "epoch": 1150, "lr": 6.895795230950149e-05}
{"train_loss": 0.19337977468967438, "global_step": 102435, "epoch": 1150, "lr": 6.89574158596761e-05}
{"train_loss": 0.12601995468139648, "global_step": 102436, "epoch": 1150, "lr": 6.895687940730213e-05}
{"train_loss": 0.1853361427783966, "global_step": 102437, "epoch": 1150, "lr": 6.895634295237961e-05}
{"train_loss": 0.16264061805572402, "global_step": 102438, "epoch": 1150, "lr": 6.895580649490866e-05, "train/sim_max_reward_0": 0.6179477496834584, "train/sim_max_reward_1": 0.5127942803373965, "train/sim_max_reward_2": 0.7171523106003781, "train/sim_max_reward_3": 0.9861828016469713, "train/sim_max_reward_4": 0.9008044846303612, "train/sim_max_reward_5": 0.5499673604115118, "test/sim_max_reward_4500000": 0.9808792694295122, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9481962728018012, "test/sim_max_reward_4500003": 0.5275333986182781, "test/sim_max_reward_4500004": 0.6031370925825973, "test/sim_max_reward_4500005": 0.1623356040877551, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.973209758852365, "test/sim_max_reward_4500008": 0.16004177357797605, "test/sim_max_reward_4500009": 0.04909981911080175, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.24126871362780633, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.753488239910425, "test/sim_max_reward_4500016": 0.5235468451761736, "test/sim_max_reward_4500017": 0.5992212683529667, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9520461584652602, "test/sim_max_reward_4500020": 0.43346973146627155, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9828338331290154, "test/sim_max_reward_4500023": 0.6031785635913219, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.5267980173210477, "test/sim_max_reward_4500026": 0.393868561336843, "test/sim_max_reward_4500027": 0.9785301807086043, "test/sim_max_reward_4500028": 0.6714366425737263, "test/sim_max_reward_4500029": 0.7111399680361, "test/sim_max_reward_4500030": 0.99255878226423, "test/sim_max_reward_4500031": 0.4433335402971338, "test/sim_max_reward_4500032": 0.5629740201095437, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 1.0, "test/sim_max_reward_4500036": 0.4753352260157061, "test/sim_max_reward_4500037": 0.4893938814401381, "test/sim_max_reward_4500038": 0.6362238844833233, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9993838005675282, "test/sim_max_reward_4500041": 0.355586343157984, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.6821578043832872, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.6503625909142193, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9951371688482931, "test/sim_max_reward_4500049": 0.5046755508126963, "train/mean_score": 0.7141414978850129, "test/mean_score": 0.6112476461210146, "val_loss": 4.6669392585754395, "train_action_mse_error": 63.0425910949707}
{"train_loss": 0.13810083270072937, "global_step": 102439, "epoch": 1151, "lr": 6.895527003488934e-05}
{"train_loss": 0.17556904256343842, "global_step": 102440, "epoch": 1151, "lr": 6.89547335723217e-05}
{"train_loss": 0.1132974922657013, "global_step": 102441, "epoch": 1151, "lr": 6.895419710720585e-05}
{"train_loss": 0.11105313897132874, "global_step": 102442, "epoch": 1151, "lr": 6.895366063954181e-05}
{"train_loss": 0.19306401908397675, "global_step": 102443, "epoch": 1151, "lr": 6.895312416932971e-05}
{"train_loss": 0.14067795872688293, "global_step": 102444, "epoch": 1151, "lr": 6.895258769656956e-05}
{"train_loss": 0.1287863701581955, "global_step": 102445, "epoch": 1151, "lr": 6.89520512212615e-05}
{"train_loss": 0.15109343826770782, "global_step": 102446, "epoch": 1151, "lr": 6.895151474340556e-05}
{"train_loss": 0.19430018961429596, "global_step": 102447, "epoch": 1151, "lr": 6.895097826300181e-05}
{"train_loss": 0.15211676061153412, "global_step": 102448, "epoch": 1151, "lr": 6.895044178005034e-05}
{"train_loss": 0.1704663634300232, "global_step": 102449, "epoch": 1151, "lr": 6.89499052945512e-05}
{"train_loss": 0.14835615456104279, "global_step": 102450, "epoch": 1151, "lr": 6.89493688065045e-05}
{"train_loss": 0.1354541927576065, "global_step": 102451, "epoch": 1151, "lr": 6.894883231591028e-05}
{"train_loss": 0.13078807294368744, "global_step": 102452, "epoch": 1151, "lr": 6.89482958227686e-05}
{"train_loss": 0.14560015499591827, "global_step": 102453, "epoch": 1151, "lr": 6.894775932707957e-05}
{"train_loss": 0.18171261250972748, "global_step": 102454, "epoch": 1151, "lr": 6.894722282884324e-05}
{"train_loss": 0.1478329300880432, "global_step": 102455, "epoch": 1151, "lr": 6.894668632805968e-05}
{"train_loss": 0.13763675093650818, "global_step": 102456, "epoch": 1151, "lr": 6.894614982472897e-05}
{"train_loss": 0.2280111014842987, "global_step": 102457, "epoch": 1151, "lr": 6.89456133188512e-05}
{"train_loss": 0.11261840909719467, "global_step": 102458, "epoch": 1151, "lr": 6.894507681042641e-05}
{"train_loss": 0.16524729132652283, "global_step": 102459, "epoch": 1151, "lr": 6.894454029945468e-05}
{"train_loss": 0.13063500821590424, "global_step": 102460, "epoch": 1151, "lr": 6.89440037859361e-05}
{"train_loss": 0.13368918001651764, "global_step": 102461, "epoch": 1151, "lr": 6.894346726987073e-05}
{"train_loss": 0.13304316997528076, "global_step": 102462, "epoch": 1151, "lr": 6.894293075125862e-05}
{"train_loss": 0.16221506893634796, "global_step": 102463, "epoch": 1151, "lr": 6.894239423009989e-05}
{"train_loss": 0.13874530792236328, "global_step": 102464, "epoch": 1151, "lr": 6.894185770639457e-05}
{"train_loss": 0.15308213233947754, "global_step": 102465, "epoch": 1151, "lr": 6.894132118014276e-05}
{"train_loss": 0.12112750858068466, "global_step": 102466, "epoch": 1151, "lr": 6.894078465134451e-05}
{"train_loss": 0.12496725469827652, "global_step": 102467, "epoch": 1151, "lr": 6.89402481199999e-05}
{"train_loss": 0.23578940331935883, "global_step": 102468, "epoch": 1151, "lr": 6.893971158610902e-05}
{"train_loss": 0.16482113301753998, "global_step": 102469, "epoch": 1151, "lr": 6.893917504967193e-05}
{"train_loss": 0.1611408144235611, "global_step": 102470, "epoch": 1151, "lr": 6.89386385106887e-05}
{"train_loss": 0.22648575901985168, "global_step": 102471, "epoch": 1151, "lr": 6.893810196915938e-05}
{"train_loss": 0.13388019800186157, "global_step": 102472, "epoch": 1151, "lr": 6.893756542508407e-05}
{"train_loss": 0.23252426087856293, "global_step": 102473, "epoch": 1151, "lr": 6.893702887846282e-05}
{"train_loss": 0.1272018700838089, "global_step": 102474, "epoch": 1151, "lr": 6.893649232929573e-05}
{"train_loss": 0.22881878912448883, "global_step": 102475, "epoch": 1151, "lr": 6.893595577758285e-05}
{"train_loss": 0.11223389208316803, "global_step": 102476, "epoch": 1151, "lr": 6.893541922332428e-05}
{"train_loss": 0.14210379123687744, "global_step": 102477, "epoch": 1151, "lr": 6.893488266652008e-05}
{"train_loss": 0.20331916213035583, "global_step": 102478, "epoch": 1151, "lr": 6.89343461071703e-05}
{"train_loss": 0.21786494553089142, "global_step": 102479, "epoch": 1151, "lr": 6.893380954527501e-05}
{"train_loss": 0.1307729333639145, "global_step": 102480, "epoch": 1151, "lr": 6.893327298083433e-05}
{"train_loss": 0.2551421523094177, "global_step": 102481, "epoch": 1151, "lr": 6.89327364138483e-05}
{"train_loss": 0.1359321027994156, "global_step": 102482, "epoch": 1151, "lr": 6.893219984431698e-05}
{"train_loss": 0.18009738624095917, "global_step": 102483, "epoch": 1151, "lr": 6.893166327224047e-05}
{"train_loss": 0.11924406886100769, "global_step": 102484, "epoch": 1151, "lr": 6.893112669761882e-05}
{"train_loss": 0.1400207281112671, "global_step": 102485, "epoch": 1151, "lr": 6.89305901204521e-05}
{"train_loss": 0.17165507376194, "global_step": 102486, "epoch": 1151, "lr": 6.893005354074042e-05}
{"train_loss": 0.17181448638439178, "global_step": 102487, "epoch": 1151, "lr": 6.892951695848382e-05}
{"train_loss": 0.1831628829240799, "global_step": 102488, "epoch": 1151, "lr": 6.892898037368236e-05}
{"train_loss": 0.14202767610549927, "global_step": 102489, "epoch": 1151, "lr": 6.892844378633615e-05}
{"train_loss": 0.13851942121982574, "global_step": 102490, "epoch": 1151, "lr": 6.892790719644523e-05}
{"train_loss": 0.16944876313209534, "global_step": 102491, "epoch": 1151, "lr": 6.892737060400967e-05}
{"train_loss": 0.21773116290569305, "global_step": 102492, "epoch": 1151, "lr": 6.892683400902957e-05}
{"train_loss": 0.15379540622234344, "global_step": 102493, "epoch": 1151, "lr": 6.892629741150498e-05}
{"train_loss": 0.1839674562215805, "global_step": 102494, "epoch": 1151, "lr": 6.8925760811436e-05}
{"train_loss": 0.194834366440773, "global_step": 102495, "epoch": 1151, "lr": 6.892522420882266e-05}
{"train_loss": 0.2335178107023239, "global_step": 102496, "epoch": 1151, "lr": 6.892468760366507e-05}
{"train_loss": 0.1375313550233841, "global_step": 102497, "epoch": 1151, "lr": 6.892415099596327e-05}
{"train_loss": 0.1734085977077484, "global_step": 102498, "epoch": 1151, "lr": 6.892361438571737e-05}
{"train_loss": 0.1888037621974945, "global_step": 102499, "epoch": 1151, "lr": 6.89230777729274e-05}
{"train_loss": 0.15172331035137177, "global_step": 102500, "epoch": 1151, "lr": 6.892254115759346e-05}
{"train_loss": 0.1456160545349121, "global_step": 102501, "epoch": 1151, "lr": 6.892200453971561e-05}
{"train_loss": 0.13004550337791443, "global_step": 102502, "epoch": 1151, "lr": 6.892146791929394e-05}
{"train_loss": 0.1376018524169922, "global_step": 102503, "epoch": 1151, "lr": 6.89209312963285e-05}
{"train_loss": 0.2080567181110382, "global_step": 102504, "epoch": 1151, "lr": 6.892039467081938e-05}
{"train_loss": 0.13522744178771973, "global_step": 102505, "epoch": 1151, "lr": 6.891985804276663e-05}
{"train_loss": 0.17252451181411743, "global_step": 102506, "epoch": 1151, "lr": 6.891932141217035e-05}
{"train_loss": 0.12095490097999573, "global_step": 102507, "epoch": 1151, "lr": 6.891878477903059e-05}
{"train_loss": 0.16956350207328796, "global_step": 102508, "epoch": 1151, "lr": 6.891824814334744e-05}
{"train_loss": 0.14559203386306763, "global_step": 102509, "epoch": 1151, "lr": 6.891771150512094e-05}
{"train_loss": 0.1779552549123764, "global_step": 102510, "epoch": 1151, "lr": 6.89171748643512e-05}
{"train_loss": 0.1488879919052124, "global_step": 102511, "epoch": 1151, "lr": 6.891663822103826e-05}
{"train_loss": 0.18519289791584015, "global_step": 102512, "epoch": 1151, "lr": 6.891610157518224e-05}
{"train_loss": 0.16204701364040375, "global_step": 102513, "epoch": 1151, "lr": 6.891556492678316e-05}
{"train_loss": 0.2188076376914978, "global_step": 102514, "epoch": 1151, "lr": 6.891502827584111e-05}
{"train_loss": 0.20346668362617493, "global_step": 102515, "epoch": 1151, "lr": 6.891449162235617e-05}
{"train_loss": 0.16870944201946259, "global_step": 102516, "epoch": 1151, "lr": 6.891395496632841e-05}
{"train_loss": 0.2120353877544403, "global_step": 102517, "epoch": 1151, "lr": 6.891341830775791e-05}
{"train_loss": 0.1273556798696518, "global_step": 102518, "epoch": 1151, "lr": 6.891288164664471e-05}
{"train_loss": 0.13729538023471832, "global_step": 102519, "epoch": 1151, "lr": 6.89123449829889e-05}
{"train_loss": 0.13060234487056732, "global_step": 102520, "epoch": 1151, "lr": 6.891180831679058e-05}
{"train_loss": 0.22203825414180756, "global_step": 102521, "epoch": 1151, "lr": 6.891127164804978e-05}
{"train_loss": 0.1584010124206543, "global_step": 102522, "epoch": 1151, "lr": 6.89107349767666e-05}
{"train_loss": 0.12066479027271271, "global_step": 102523, "epoch": 1151, "lr": 6.89101983029411e-05}
{"train_loss": 0.15056076645851135, "global_step": 102524, "epoch": 1151, "lr": 6.890966162657335e-05}
{"train_loss": 0.13511978089809418, "global_step": 102525, "epoch": 1151, "lr": 6.890912494766342e-05}
{"train_loss": 0.05820631608366966, "global_step": 102526, "epoch": 1151, "lr": 6.89085882662114e-05}
{"train_loss": 0.16090801178237027, "global_step": 102527, "epoch": 1151, "lr": 6.890805158221736e-05, "val_loss": 4.634737491607666}
{"train_loss": 0.15937680006027222, "global_step": 102528, "epoch": 1152, "lr": 6.890751489568134e-05}
{"train_loss": 0.1348927617073059, "global_step": 102529, "epoch": 1152, "lr": 6.890697820660345e-05}
{"train_loss": 0.15495790541172028, "global_step": 102530, "epoch": 1152, "lr": 6.890644151498375e-05}
{"train_loss": 0.10942167788743973, "global_step": 102531, "epoch": 1152, "lr": 6.89059048208223e-05}
{"train_loss": 0.1976774036884308, "global_step": 102532, "epoch": 1152, "lr": 6.890536812411918e-05}
{"train_loss": 0.1423686146736145, "global_step": 102533, "epoch": 1152, "lr": 6.890483142487447e-05}
{"train_loss": 0.14878655970096588, "global_step": 102534, "epoch": 1152, "lr": 6.890429472308824e-05}
{"train_loss": 0.17530100047588348, "global_step": 102535, "epoch": 1152, "lr": 6.890375801876054e-05}
{"train_loss": 0.142889142036438, "global_step": 102536, "epoch": 1152, "lr": 6.890322131189148e-05}
{"train_loss": 0.12055394053459167, "global_step": 102537, "epoch": 1152, "lr": 6.890268460248111e-05}
{"train_loss": 0.16029685735702515, "global_step": 102538, "epoch": 1152, "lr": 6.890214789052949e-05}
{"train_loss": 0.13201555609703064, "global_step": 102539, "epoch": 1152, "lr": 6.890161117603672e-05}
{"train_loss": 0.09938530623912811, "global_step": 102540, "epoch": 1152, "lr": 6.890107445900284e-05}
{"train_loss": 0.17802993953227997, "global_step": 102541, "epoch": 1152, "lr": 6.890053773942796e-05}
{"train_loss": 0.10659248381853104, "global_step": 102542, "epoch": 1152, "lr": 6.890000101731214e-05}
{"train_loss": 0.12567219138145447, "global_step": 102543, "epoch": 1152, "lr": 6.889946429265542e-05}
{"train_loss": 0.13670134544372559, "global_step": 102544, "epoch": 1152, "lr": 6.889892756545792e-05}
{"train_loss": 0.1475222408771515, "global_step": 102545, "epoch": 1152, "lr": 6.889839083571967e-05}
{"train_loss": 0.1746014654636383, "global_step": 102546, "epoch": 1152, "lr": 6.889785410344078e-05}
{"train_loss": 0.12966959178447723, "global_step": 102547, "epoch": 1152, "lr": 6.88973173686213e-05}
{"train_loss": 0.14993920922279358, "global_step": 102548, "epoch": 1152, "lr": 6.889678063126129e-05}
{"train_loss": 0.16633917391300201, "global_step": 102549, "epoch": 1152, "lr": 6.889624389136086e-05}
{"train_loss": 0.14191459119319916, "global_step": 102550, "epoch": 1152, "lr": 6.889570714892004e-05}
{"train_loss": 0.12430278211832047, "global_step": 102551, "epoch": 1152, "lr": 6.889517040393894e-05}
{"train_loss": 0.09245579689741135, "global_step": 102552, "epoch": 1152, "lr": 6.88946336564176e-05}
{"train_loss": 0.22153222560882568, "global_step": 102553, "epoch": 1152, "lr": 6.889409690635612e-05}
{"train_loss": 0.1888980120420456, "global_step": 102554, "epoch": 1152, "lr": 6.889356015375455e-05}
{"train_loss": 0.20291219651699066, "global_step": 102555, "epoch": 1152, "lr": 6.889302339861298e-05}
{"train_loss": 0.16112163662910461, "global_step": 102556, "epoch": 1152, "lr": 6.889248664093147e-05}
{"train_loss": 0.1451873481273651, "global_step": 102557, "epoch": 1152, "lr": 6.889194988071009e-05}
{"train_loss": 0.20169787108898163, "global_step": 102558, "epoch": 1152, "lr": 6.889141311794892e-05}
{"train_loss": 0.09776699542999268, "global_step": 102559, "epoch": 1152, "lr": 6.889087635264803e-05}
{"train_loss": 0.19273579120635986, "global_step": 102560, "epoch": 1152, "lr": 6.88903395848075e-05}
{"train_loss": 0.11565130203962326, "global_step": 102561, "epoch": 1152, "lr": 6.888980281442739e-05}
{"train_loss": 0.15742547810077667, "global_step": 102562, "epoch": 1152, "lr": 6.888926604150776e-05}
{"train_loss": 0.1188497468829155, "global_step": 102563, "epoch": 1152, "lr": 6.888872926604871e-05}
{"train_loss": 0.1304224729537964, "global_step": 102564, "epoch": 1152, "lr": 6.88881924880503e-05}
{"train_loss": 0.14999276399612427, "global_step": 102565, "epoch": 1152, "lr": 6.888765570751261e-05}
{"train_loss": 0.13335253298282623, "global_step": 102566, "epoch": 1152, "lr": 6.88871189244357e-05}
{"train_loss": 0.13811232149600983, "global_step": 102567, "epoch": 1152, "lr": 6.888658213881964e-05}
{"train_loss": 0.08485600352287292, "global_step": 102568, "epoch": 1152, "lr": 6.888604535066451e-05}
{"train_loss": 0.1687363088130951, "global_step": 102569, "epoch": 1152, "lr": 6.88855085599704e-05}
{"train_loss": 0.09566699713468552, "global_step": 102570, "epoch": 1152, "lr": 6.888497176673734e-05}
{"train_loss": 0.17442536354064941, "global_step": 102571, "epoch": 1152, "lr": 6.888443497096544e-05}
{"train_loss": 0.13104590773582458, "global_step": 102572, "epoch": 1152, "lr": 6.888389817265475e-05}
{"train_loss": 0.10929136723279953, "global_step": 102573, "epoch": 1152, "lr": 6.888336137180536e-05}
{"train_loss": 0.1271314024925232, "global_step": 102574, "epoch": 1152, "lr": 6.888282456841733e-05}
{"train_loss": 0.1517145335674286, "global_step": 102575, "epoch": 1152, "lr": 6.888228776249072e-05}
{"train_loss": 0.16623654961585999, "global_step": 102576, "epoch": 1152, "lr": 6.888175095402564e-05}
{"train_loss": 0.1318453997373581, "global_step": 102577, "epoch": 1152, "lr": 6.888121414302211e-05}
{"train_loss": 0.22637805342674255, "global_step": 102578, "epoch": 1152, "lr": 6.888067732948025e-05}
{"train_loss": 0.14274917542934418, "global_step": 102579, "epoch": 1152, "lr": 6.888014051340011e-05}
{"train_loss": 0.19770613312721252, "global_step": 102580, "epoch": 1152, "lr": 6.887960369478178e-05}
{"train_loss": 0.1257299780845642, "global_step": 102581, "epoch": 1152, "lr": 6.887906687362529e-05}
{"train_loss": 0.17633379995822906, "global_step": 102582, "epoch": 1152, "lr": 6.887853004993076e-05}
{"train_loss": 0.15384703874588013, "global_step": 102583, "epoch": 1152, "lr": 6.887799322369823e-05}
{"train_loss": 0.18045243620872498, "global_step": 102584, "epoch": 1152, "lr": 6.887745639492779e-05}
{"train_loss": 0.16012762486934662, "global_step": 102585, "epoch": 1152, "lr": 6.88769195636195e-05}
{"train_loss": 0.21922263503074646, "global_step": 102586, "epoch": 1152, "lr": 6.887638272977343e-05}
{"train_loss": 0.14695531129837036, "global_step": 102587, "epoch": 1152, "lr": 6.887584589338968e-05}
{"train_loss": 0.14998605847358704, "global_step": 102588, "epoch": 1152, "lr": 6.88753090544683e-05}
{"train_loss": 0.15264077484607697, "global_step": 102589, "epoch": 1152, "lr": 6.887477221300935e-05}
{"train_loss": 0.1901455521583557, "global_step": 102590, "epoch": 1152, "lr": 6.887423536901294e-05}
{"train_loss": 0.1094997227191925, "global_step": 102591, "epoch": 1152, "lr": 6.88736985224791e-05}
{"train_loss": 0.24313825368881226, "global_step": 102592, "epoch": 1152, "lr": 6.887316167340793e-05}
{"train_loss": 0.27406275272369385, "global_step": 102593, "epoch": 1152, "lr": 6.88726248217995e-05}
{"train_loss": 0.22285085916519165, "global_step": 102594, "epoch": 1152, "lr": 6.887208796765386e-05}
{"train_loss": 0.1454525589942932, "global_step": 102595, "epoch": 1152, "lr": 6.88715511109711e-05}
{"train_loss": 0.22663159668445587, "global_step": 102596, "epoch": 1152, "lr": 6.887101425175128e-05}
{"train_loss": 0.15837682783603668, "global_step": 102597, "epoch": 1152, "lr": 6.88704773899945e-05}
{"train_loss": 0.2503507733345032, "global_step": 102598, "epoch": 1152, "lr": 6.886994052570081e-05}
{"train_loss": 0.13877061009407043, "global_step": 102599, "epoch": 1152, "lr": 6.88694036588703e-05}
{"train_loss": 0.12709639966487885, "global_step": 102600, "epoch": 1152, "lr": 6.886886678950301e-05}
{"train_loss": 0.2530409097671509, "global_step": 102601, "epoch": 1152, "lr": 6.886832991759904e-05}
{"train_loss": 0.21358947455883026, "global_step": 102602, "epoch": 1152, "lr": 6.886779304315845e-05}
{"train_loss": 0.12052121013402939, "global_step": 102603, "epoch": 1152, "lr": 6.886725616618132e-05}
{"train_loss": 0.15212275087833405, "global_step": 102604, "epoch": 1152, "lr": 6.88667192866677e-05}
{"train_loss": 0.11913425475358963, "global_step": 102605, "epoch": 1152, "lr": 6.88661824046177e-05}
{"train_loss": 0.14133551716804504, "global_step": 102606, "epoch": 1152, "lr": 6.886564552003136e-05}
{"train_loss": 0.1658274084329605, "global_step": 102607, "epoch": 1152, "lr": 6.886510863290877e-05}
{"train_loss": 0.19954465329647064, "global_step": 102608, "epoch": 1152, "lr": 6.886457174324999e-05}
{"train_loss": 0.20757636427879333, "global_step": 102609, "epoch": 1152, "lr": 6.88640348510551e-05}
{"train_loss": 0.18552649021148682, "global_step": 102610, "epoch": 1152, "lr": 6.886349795632417e-05}
{"train_loss": 0.23627348244190216, "global_step": 102611, "epoch": 1152, "lr": 6.886296105905728e-05}
{"train_loss": 0.14782968163490295, "global_step": 102612, "epoch": 1152, "lr": 6.886242415925448e-05}
{"train_loss": 0.1500057727098465, "global_step": 102613, "epoch": 1152, "lr": 6.886188725691586e-05}
{"train_loss": 0.15740269422531128, "global_step": 102614, "epoch": 1152, "lr": 6.88613503520415e-05}
{"train_loss": 0.1860184371471405, "global_step": 102615, "epoch": 1152, "lr": 6.886081344463147e-05}
{"train_loss": 0.15877562325991942, "global_step": 102616, "epoch": 1152, "lr": 6.88602765346858e-05, "val_loss": 4.529812335968018}
{"train_loss": 0.1306096315383911, "global_step": 102617, "epoch": 1153, "lr": 6.885973962220463e-05}
{"train_loss": 0.2830100953578949, "global_step": 102618, "epoch": 1153, "lr": 6.885920270718798e-05}
{"train_loss": 0.15489411354064941, "global_step": 102619, "epoch": 1153, "lr": 6.885866578963594e-05}
{"train_loss": 0.22639162838459015, "global_step": 102620, "epoch": 1153, "lr": 6.885812886954859e-05}
{"train_loss": 0.13830120861530304, "global_step": 102621, "epoch": 1153, "lr": 6.885759194692599e-05}
{"train_loss": 0.18779747188091278, "global_step": 102622, "epoch": 1153, "lr": 6.885705502176822e-05}
{"train_loss": 0.22111453115940094, "global_step": 102623, "epoch": 1153, "lr": 6.885651809407535e-05}
{"train_loss": 0.17840926349163055, "global_step": 102624, "epoch": 1153, "lr": 6.885598116384744e-05}
{"train_loss": 0.1981021761894226, "global_step": 102625, "epoch": 1153, "lr": 6.885544423108459e-05}
{"train_loss": 0.21060511469841003, "global_step": 102626, "epoch": 1153, "lr": 6.885490729578684e-05}
{"train_loss": 0.14357993006706238, "global_step": 102627, "epoch": 1153, "lr": 6.885437035795428e-05}
{"train_loss": 0.17124995589256287, "global_step": 102628, "epoch": 1153, "lr": 6.885383341758699e-05}
{"train_loss": 0.16205336153507233, "global_step": 102629, "epoch": 1153, "lr": 6.885329647468501e-05}
{"train_loss": 0.17787033319473267, "global_step": 102630, "epoch": 1153, "lr": 6.885275952924845e-05}
{"train_loss": 0.17764350771903992, "global_step": 102631, "epoch": 1153, "lr": 6.885222258127737e-05}
{"train_loss": 0.1430855244398117, "global_step": 102632, "epoch": 1153, "lr": 6.885168563077184e-05}
{"train_loss": 0.15188980102539062, "global_step": 102633, "epoch": 1153, "lr": 6.885114867773192e-05}
{"train_loss": 0.15342611074447632, "global_step": 102634, "epoch": 1153, "lr": 6.88506117221577e-05}
{"train_loss": 0.12000113725662231, "global_step": 102635, "epoch": 1153, "lr": 6.885007476404923e-05}
{"train_loss": 0.1178266778588295, "global_step": 102636, "epoch": 1153, "lr": 6.88495378034066e-05}
{"train_loss": 0.16386841237545013, "global_step": 102637, "epoch": 1153, "lr": 6.88490008402299e-05}
{"train_loss": 0.2138187140226364, "global_step": 102638, "epoch": 1153, "lr": 6.884846387451918e-05}
{"train_loss": 0.11091577261686325, "global_step": 102639, "epoch": 1153, "lr": 6.88479269062745e-05}
{"train_loss": 0.19589415192604065, "global_step": 102640, "epoch": 1153, "lr": 6.884738993549595e-05}
{"train_loss": 0.15594765543937683, "global_step": 102641, "epoch": 1153, "lr": 6.884685296218359e-05}
{"train_loss": 0.17380894720554352, "global_step": 102642, "epoch": 1153, "lr": 6.884631598633752e-05}
{"train_loss": 0.1474698930978775, "global_step": 102643, "epoch": 1153, "lr": 6.884577900795777e-05}
{"train_loss": 0.1353961080312729, "global_step": 102644, "epoch": 1153, "lr": 6.884524202704445e-05}
{"train_loss": 0.13088856637477875, "global_step": 102645, "epoch": 1153, "lr": 6.884470504359761e-05}
{"train_loss": 0.20501965284347534, "global_step": 102646, "epoch": 1153, "lr": 6.884416805761734e-05}
{"train_loss": 0.15345558524131775, "global_step": 102647, "epoch": 1153, "lr": 6.884363106910369e-05}
{"train_loss": 0.1355535387992859, "global_step": 102648, "epoch": 1153, "lr": 6.884309407805674e-05}
{"train_loss": 0.13060563802719116, "global_step": 102649, "epoch": 1153, "lr": 6.884255708447657e-05}
{"train_loss": 0.14443299174308777, "global_step": 102650, "epoch": 1153, "lr": 6.884202008836325e-05}
{"train_loss": 0.11217623203992844, "global_step": 102651, "epoch": 1153, "lr": 6.884148308971686e-05}
{"train_loss": 0.14711180329322815, "global_step": 102652, "epoch": 1153, "lr": 6.884094608853744e-05}
{"train_loss": 0.09718362987041473, "global_step": 102653, "epoch": 1153, "lr": 6.88404090848251e-05}
{"train_loss": 0.19395262002944946, "global_step": 102654, "epoch": 1153, "lr": 6.883987207857989e-05}
{"train_loss": 0.11785359680652618, "global_step": 102655, "epoch": 1153, "lr": 6.88393350698019e-05}
{"train_loss": 0.227895587682724, "global_step": 102656, "epoch": 1153, "lr": 6.883879805849117e-05}
{"train_loss": 0.17742805182933807, "global_step": 102657, "epoch": 1153, "lr": 6.88382610446478e-05}
{"train_loss": 0.17092885076999664, "global_step": 102658, "epoch": 1153, "lr": 6.883772402827187e-05}
{"train_loss": 0.1585186868906021, "global_step": 102659, "epoch": 1153, "lr": 6.883718700936343e-05}
{"train_loss": 0.1505114734172821, "global_step": 102660, "epoch": 1153, "lr": 6.883664998792256e-05}
{"train_loss": 0.12907695770263672, "global_step": 102661, "epoch": 1153, "lr": 6.883611296394934e-05}
{"train_loss": 0.1309356689453125, "global_step": 102662, "epoch": 1153, "lr": 6.883557593744381e-05}
{"train_loss": 0.18965841829776764, "global_step": 102663, "epoch": 1153, "lr": 6.883503890840609e-05}
{"train_loss": 0.24189059436321259, "global_step": 102664, "epoch": 1153, "lr": 6.883450187683622e-05}
{"train_loss": 0.15454883873462677, "global_step": 102665, "epoch": 1153, "lr": 6.883396484273428e-05}
{"train_loss": 0.1893661916255951, "global_step": 102666, "epoch": 1153, "lr": 6.883342780610034e-05}
{"train_loss": 0.20473481714725494, "global_step": 102667, "epoch": 1153, "lr": 6.883289076693449e-05}
{"train_loss": 0.17338702082633972, "global_step": 102668, "epoch": 1153, "lr": 6.883235372523677e-05}
{"train_loss": 0.17090857028961182, "global_step": 102669, "epoch": 1153, "lr": 6.883181668100728e-05}
{"train_loss": 0.19207003712654114, "global_step": 102670, "epoch": 1153, "lr": 6.883127963424608e-05}
{"train_loss": 0.10160939395427704, "global_step": 102671, "epoch": 1153, "lr": 6.883074258495323e-05}
{"train_loss": 0.2625678479671478, "global_step": 102672, "epoch": 1153, "lr": 6.883020553312884e-05}
{"train_loss": 0.17651785910129547, "global_step": 102673, "epoch": 1153, "lr": 6.882966847877292e-05}
{"train_loss": 0.12464900314807892, "global_step": 102674, "epoch": 1153, "lr": 6.882913142188562e-05}
{"train_loss": 0.1332254409790039, "global_step": 102675, "epoch": 1153, "lr": 6.882859436246696e-05}
{"train_loss": 0.23743674159049988, "global_step": 102676, "epoch": 1153, "lr": 6.882805730051702e-05}
{"train_loss": 0.2058771848678589, "global_step": 102677, "epoch": 1153, "lr": 6.882752023603589e-05}
{"train_loss": 0.12351525574922562, "global_step": 102678, "epoch": 1153, "lr": 6.882698316902363e-05}
{"train_loss": 0.183782160282135, "global_step": 102679, "epoch": 1153, "lr": 6.882644609948031e-05}
{"train_loss": 0.23997238278388977, "global_step": 102680, "epoch": 1153, "lr": 6.882590902740598e-05}
{"train_loss": 0.1692304164171219, "global_step": 102681, "epoch": 1153, "lr": 6.882537195280076e-05}
{"train_loss": 0.13138043880462646, "global_step": 102682, "epoch": 1153, "lr": 6.882483487566469e-05}
{"train_loss": 0.1841050237417221, "global_step": 102683, "epoch": 1153, "lr": 6.882429779599785e-05}
{"train_loss": 0.16583308577537537, "global_step": 102684, "epoch": 1153, "lr": 6.882376071380032e-05}
{"train_loss": 0.1342027187347412, "global_step": 102685, "epoch": 1153, "lr": 6.882322362907214e-05}
{"train_loss": 0.2173384428024292, "global_step": 102686, "epoch": 1153, "lr": 6.882268654181343e-05}
{"train_loss": 0.1362254023551941, "global_step": 102687, "epoch": 1153, "lr": 6.882214945202424e-05}
{"train_loss": 0.16581715643405914, "global_step": 102688, "epoch": 1153, "lr": 6.882161235970463e-05}
{"train_loss": 0.2270233929157257, "global_step": 102689, "epoch": 1153, "lr": 6.882107526485469e-05}
{"train_loss": 0.1631712168455124, "global_step": 102690, "epoch": 1153, "lr": 6.882053816747449e-05}
{"train_loss": 0.13193053007125854, "global_step": 102691, "epoch": 1153, "lr": 6.882000106756408e-05}
{"train_loss": 0.14103133976459503, "global_step": 102692, "epoch": 1153, "lr": 6.881946396512357e-05}
{"train_loss": 0.23525561392307281, "global_step": 102693, "epoch": 1153, "lr": 6.8818926860153e-05}
{"train_loss": 0.13739757239818573, "global_step": 102694, "epoch": 1153, "lr": 6.881838975265247e-05}
{"train_loss": 0.19482260942459106, "global_step": 102695, "epoch": 1153, "lr": 6.881785264262202e-05}
{"train_loss": 0.18307779729366302, "global_step": 102696, "epoch": 1153, "lr": 6.881731553006177e-05}
{"train_loss": 0.1735159158706665, "global_step": 102697, "epoch": 1153, "lr": 6.881677841497172e-05}
{"train_loss": 0.13004598021507263, "global_step": 102698, "epoch": 1153, "lr": 6.8816241297352e-05}
{"train_loss": 0.15325157344341278, "global_step": 102699, "epoch": 1153, "lr": 6.881570417720268e-05}
{"train_loss": 0.16948918998241425, "global_step": 102700, "epoch": 1153, "lr": 6.88151670545238e-05}
{"train_loss": 0.1601647436618805, "global_step": 102701, "epoch": 1153, "lr": 6.881462992931545e-05}
{"train_loss": 0.17136985063552856, "global_step": 102702, "epoch": 1153, "lr": 6.881409280157771e-05}
{"train_loss": 0.14847686886787415, "global_step": 102703, "epoch": 1153, "lr": 6.881355567131065e-05}
{"train_loss": 0.22070178389549255, "global_step": 102704, "epoch": 1153, "lr": 6.881301853851432e-05}
{"train_loss": 0.1679565780953075, "global_step": 102705, "epoch": 1153, "lr": 6.881248140318883e-05, "val_loss": 4.7598748207092285}
{"train_loss": 0.14720964431762695, "global_step": 102706, "epoch": 1154, "lr": 6.881194426533422e-05}
{"train_loss": 0.11295357346534729, "global_step": 102707, "epoch": 1154, "lr": 6.881140712495056e-05}
{"train_loss": 0.09282679110765457, "global_step": 102708, "epoch": 1154, "lr": 6.881086998203795e-05}
{"train_loss": 0.18261659145355225, "global_step": 102709, "epoch": 1154, "lr": 6.881033283659643e-05}
{"train_loss": 0.1607942432165146, "global_step": 102710, "epoch": 1154, "lr": 6.880979568862611e-05}
{"train_loss": 0.1705905646085739, "global_step": 102711, "epoch": 1154, "lr": 6.880925853812704e-05}
{"train_loss": 0.189122274518013, "global_step": 102712, "epoch": 1154, "lr": 6.880872138509929e-05}
{"train_loss": 0.2379661500453949, "global_step": 102713, "epoch": 1154, "lr": 6.880818422954294e-05}
{"train_loss": 0.24137340486049652, "global_step": 102714, "epoch": 1154, "lr": 6.880764707145805e-05}
{"train_loss": 0.18480883538722992, "global_step": 102715, "epoch": 1154, "lr": 6.880710991084471e-05}
{"train_loss": 0.1518290489912033, "global_step": 102716, "epoch": 1154, "lr": 6.8806572747703e-05}
{"train_loss": 0.20512144267559052, "global_step": 102717, "epoch": 1154, "lr": 6.880603558203296e-05}
{"train_loss": 0.1350431740283966, "global_step": 102718, "epoch": 1154, "lr": 6.880549841383468e-05}
{"train_loss": 0.12622982263565063, "global_step": 102719, "epoch": 1154, "lr": 6.880496124310822e-05}
{"train_loss": 0.1444382220506668, "global_step": 102720, "epoch": 1154, "lr": 6.880442406985367e-05}
{"train_loss": 0.198835089802742, "global_step": 102721, "epoch": 1154, "lr": 6.88038868940711e-05}
{"train_loss": 0.09874453395605087, "global_step": 102722, "epoch": 1154, "lr": 6.880334971576055e-05}
{"train_loss": 0.13495193421840668, "global_step": 102723, "epoch": 1154, "lr": 6.880281253492214e-05}
{"train_loss": 0.12352286279201508, "global_step": 102724, "epoch": 1154, "lr": 6.880227535155592e-05}
{"train_loss": 0.23668834567070007, "global_step": 102725, "epoch": 1154, "lr": 6.880173816566197e-05}
{"train_loss": 0.1680377572774887, "global_step": 102726, "epoch": 1154, "lr": 6.880120097724034e-05}
{"train_loss": 0.08188921958208084, "global_step": 102727, "epoch": 1154, "lr": 6.880066378629112e-05}
{"train_loss": 0.19152599573135376, "global_step": 102728, "epoch": 1154, "lr": 6.880012659281439e-05}
{"train_loss": 0.14838942885398865, "global_step": 102729, "epoch": 1154, "lr": 6.879958939681021e-05}
{"train_loss": 0.16341856122016907, "global_step": 102730, "epoch": 1154, "lr": 6.879905219827865e-05}
{"train_loss": 0.1990433633327484, "global_step": 102731, "epoch": 1154, "lr": 6.879851499721977e-05}
{"train_loss": 0.19381579756736755, "global_step": 102732, "epoch": 1154, "lr": 6.879797779363368e-05}
{"train_loss": 0.1676381379365921, "global_step": 102733, "epoch": 1154, "lr": 6.879744058752043e-05}
{"train_loss": 0.1745278388261795, "global_step": 102734, "epoch": 1154, "lr": 6.87969033788801e-05}
{"train_loss": 0.08827964961528778, "global_step": 102735, "epoch": 1154, "lr": 6.879636616771275e-05}
{"train_loss": 0.1307312250137329, "global_step": 102736, "epoch": 1154, "lr": 6.879582895401846e-05}
{"train_loss": 0.11078996956348419, "global_step": 102737, "epoch": 1154, "lr": 6.879529173779729e-05}
{"train_loss": 0.1371467113494873, "global_step": 102738, "epoch": 1154, "lr": 6.879475451904934e-05}
{"train_loss": 0.12696413695812225, "global_step": 102739, "epoch": 1154, "lr": 6.879421729777463e-05}
{"train_loss": 0.23209719359874725, "global_step": 102740, "epoch": 1154, "lr": 6.879368007397328e-05}
{"train_loss": 0.10386821627616882, "global_step": 102741, "epoch": 1154, "lr": 6.879314284764536e-05}
{"train_loss": 0.13822375237941742, "global_step": 102742, "epoch": 1154, "lr": 6.879260561879093e-05}
{"train_loss": 0.09037529677152634, "global_step": 102743, "epoch": 1154, "lr": 6.879206838741006e-05}
{"train_loss": 0.1678880751132965, "global_step": 102744, "epoch": 1154, "lr": 6.879153115350283e-05}
{"train_loss": 0.12800154089927673, "global_step": 102745, "epoch": 1154, "lr": 6.879099391706929e-05}
{"train_loss": 0.1799769252538681, "global_step": 102746, "epoch": 1154, "lr": 6.879045667810955e-05}
{"train_loss": 0.134269118309021, "global_step": 102747, "epoch": 1154, "lr": 6.878991943662365e-05}
{"train_loss": 0.15860489010810852, "global_step": 102748, "epoch": 1154, "lr": 6.878938219261169e-05}
{"train_loss": 0.11755452305078506, "global_step": 102749, "epoch": 1154, "lr": 6.87888449460737e-05}
{"train_loss": 0.16311535239219666, "global_step": 102750, "epoch": 1154, "lr": 6.87883076970098e-05}
{"train_loss": 0.2018425017595291, "global_step": 102751, "epoch": 1154, "lr": 6.878777044542004e-05}
{"train_loss": 0.15231136977672577, "global_step": 102752, "epoch": 1154, "lr": 6.878723319130449e-05}
{"train_loss": 0.15915247797966003, "global_step": 102753, "epoch": 1154, "lr": 6.878669593466322e-05}
{"train_loss": 0.17665156722068787, "global_step": 102754, "epoch": 1154, "lr": 6.878615867549631e-05}
{"train_loss": 0.11180324852466583, "global_step": 102755, "epoch": 1154, "lr": 6.878562141380385e-05}
{"train_loss": 0.1833042949438095, "global_step": 102756, "epoch": 1154, "lr": 6.878508414958587e-05}
{"train_loss": 0.1374908983707428, "global_step": 102757, "epoch": 1154, "lr": 6.878454688284246e-05}
{"train_loss": 0.18075427412986755, "global_step": 102758, "epoch": 1154, "lr": 6.878400961357371e-05}
{"train_loss": 0.0990423783659935, "global_step": 102759, "epoch": 1154, "lr": 6.878347234177967e-05}
{"train_loss": 0.12414610385894775, "global_step": 102760, "epoch": 1154, "lr": 6.878293506746043e-05}
{"train_loss": 0.11135730147361755, "global_step": 102761, "epoch": 1154, "lr": 6.878239779061605e-05}
{"train_loss": 0.13263405859470367, "global_step": 102762, "epoch": 1154, "lr": 6.87818605112466e-05}
{"train_loss": 0.17666666209697723, "global_step": 102763, "epoch": 1154, "lr": 6.878132322935216e-05}
{"train_loss": 0.19660122692584991, "global_step": 102764, "epoch": 1154, "lr": 6.878078594493281e-05}
{"train_loss": 0.11699292808771133, "global_step": 102765, "epoch": 1154, "lr": 6.878024865798858e-05}
{"train_loss": 0.21806231141090393, "global_step": 102766, "epoch": 1154, "lr": 6.877971136851962e-05}
{"train_loss": 0.11069437861442566, "global_step": 102767, "epoch": 1154, "lr": 6.877917407652593e-05}
{"train_loss": 0.10454802215099335, "global_step": 102768, "epoch": 1154, "lr": 6.877863678200761e-05}
{"train_loss": 0.2138400375843048, "global_step": 102769, "epoch": 1154, "lr": 6.877809948496472e-05}
{"train_loss": 0.11038624495267868, "global_step": 102770, "epoch": 1154, "lr": 6.877756218539736e-05}
{"train_loss": 0.25271061062812805, "global_step": 102771, "epoch": 1154, "lr": 6.87770248833056e-05}
{"train_loss": 0.06359962373971939, "global_step": 102772, "epoch": 1154, "lr": 6.877648757868948e-05}
{"train_loss": 0.11065122485160828, "global_step": 102773, "epoch": 1154, "lr": 6.877595027154909e-05}
{"train_loss": 0.15087316930294037, "global_step": 102774, "epoch": 1154, "lr": 6.877541296188453e-05}
{"train_loss": 0.16704343259334564, "global_step": 102775, "epoch": 1154, "lr": 6.877487564969581e-05}
{"train_loss": 0.18178243935108185, "global_step": 102776, "epoch": 1154, "lr": 6.877433833498305e-05}
{"train_loss": 0.16493600606918335, "global_step": 102777, "epoch": 1154, "lr": 6.87738010177463e-05}
{"train_loss": 0.14377683401107788, "global_step": 102778, "epoch": 1154, "lr": 6.877326369798565e-05}
{"train_loss": 0.22194801270961761, "global_step": 102779, "epoch": 1154, "lr": 6.877272637570116e-05}
{"train_loss": 0.1944514960050583, "global_step": 102780, "epoch": 1154, "lr": 6.87721890508929e-05}
{"train_loss": 0.10177113115787506, "global_step": 102781, "epoch": 1154, "lr": 6.877165172356095e-05}
{"train_loss": 0.16336683928966522, "global_step": 102782, "epoch": 1154, "lr": 6.877111439370539e-05}
{"train_loss": 0.13158223032951355, "global_step": 102783, "epoch": 1154, "lr": 6.877057706132626e-05}
{"train_loss": 0.18660059571266174, "global_step": 102784, "epoch": 1154, "lr": 6.877003972642367e-05}
{"train_loss": 0.16651052236557007, "global_step": 102785, "epoch": 1154, "lr": 6.876950238899769e-05}
{"train_loss": 0.1535572111606598, "global_step": 102786, "epoch": 1154, "lr": 6.876896504904837e-05}
{"train_loss": 0.1450045108795166, "global_step": 102787, "epoch": 1154, "lr": 6.876842770657578e-05}
{"train_loss": 0.14393164217472076, "global_step": 102788, "epoch": 1154, "lr": 6.876789036158001e-05}
{"train_loss": 0.18008622527122498, "global_step": 102789, "epoch": 1154, "lr": 6.876735301406113e-05}
{"train_loss": 0.17124269902706146, "global_step": 102790, "epoch": 1154, "lr": 6.876681566401921e-05}
{"train_loss": 0.10626663267612457, "global_step": 102791, "epoch": 1154, "lr": 6.876627831145431e-05}
{"train_loss": 0.17612358927726746, "global_step": 102792, "epoch": 1154, "lr": 6.876574095636652e-05}
{"train_loss": 0.20405679941177368, "global_step": 102793, "epoch": 1154, "lr": 6.87652035987559e-05}
{"train_loss": 0.1556014096301593, "global_step": 102794, "epoch": 1154, "lr": 6.876466623862255e-05, "val_loss": 4.840212345123291}
{"train_loss": 0.13008272647857666, "global_step": 102795, "epoch": 1155, "lr": 6.876412887596649e-05}
{"train_loss": 0.16295327246189117, "global_step": 102796, "epoch": 1155, "lr": 6.876359151078784e-05}
{"train_loss": 0.19004662334918976, "global_step": 102797, "epoch": 1155, "lr": 6.876305414308665e-05}
{"train_loss": 0.17549435794353485, "global_step": 102798, "epoch": 1155, "lr": 6.876251677286299e-05}
{"train_loss": 0.22715184092521667, "global_step": 102799, "epoch": 1155, "lr": 6.876197940011694e-05}
{"train_loss": 0.1624782234430313, "global_step": 102800, "epoch": 1155, "lr": 6.876144202484856e-05}
{"train_loss": 0.2522481083869934, "global_step": 102801, "epoch": 1155, "lr": 6.876090464705793e-05}
{"train_loss": 0.11145171523094177, "global_step": 102802, "epoch": 1155, "lr": 6.876036726674515e-05}
{"train_loss": 0.14482754468917847, "global_step": 102803, "epoch": 1155, "lr": 6.875982988391026e-05}
{"train_loss": 0.16538359224796295, "global_step": 102804, "epoch": 1155, "lr": 6.875929249855333e-05}
{"train_loss": 0.19370412826538086, "global_step": 102805, "epoch": 1155, "lr": 6.875875511067445e-05}
{"train_loss": 0.17035700380802155, "global_step": 102806, "epoch": 1155, "lr": 6.875821772027368e-05}
{"train_loss": 0.22054144740104675, "global_step": 102807, "epoch": 1155, "lr": 6.875768032735109e-05}
{"train_loss": 0.2673322260379791, "global_step": 102808, "epoch": 1155, "lr": 6.875714293190677e-05}
{"train_loss": 0.13057151436805725, "global_step": 102809, "epoch": 1155, "lr": 6.875660553394078e-05}
{"train_loss": 0.22146418690681458, "global_step": 102810, "epoch": 1155, "lr": 6.875606813345319e-05}
{"train_loss": 0.16710609197616577, "global_step": 102811, "epoch": 1155, "lr": 6.875553073044408e-05}
{"train_loss": 0.1843753606081009, "global_step": 102812, "epoch": 1155, "lr": 6.87549933249135e-05}
{"train_loss": 0.13340328633785248, "global_step": 102813, "epoch": 1155, "lr": 6.875445591686155e-05}
{"train_loss": 0.13437047600746155, "global_step": 102814, "epoch": 1155, "lr": 6.87539185062883e-05}
{"train_loss": 0.16262897849082947, "global_step": 102815, "epoch": 1155, "lr": 6.875338109319382e-05}
{"train_loss": 0.11686201393604279, "global_step": 102816, "epoch": 1155, "lr": 6.875284367757817e-05}
{"train_loss": 0.24507170915603638, "global_step": 102817, "epoch": 1155, "lr": 6.875230625944143e-05}
{"train_loss": 0.11734920740127563, "global_step": 102818, "epoch": 1155, "lr": 6.875176883878365e-05}
{"train_loss": 0.13174331188201904, "global_step": 102819, "epoch": 1155, "lr": 6.875123141560493e-05}
{"train_loss": 0.20802342891693115, "global_step": 102820, "epoch": 1155, "lr": 6.875069398990534e-05}
{"train_loss": 0.18497323989868164, "global_step": 102821, "epoch": 1155, "lr": 6.875015656168495e-05}
{"train_loss": 0.09097824990749359, "global_step": 102822, "epoch": 1155, "lr": 6.874961913094383e-05}
{"train_loss": 0.16402295231819153, "global_step": 102823, "epoch": 1155, "lr": 6.874908169768206e-05}
{"train_loss": 0.1241038516163826, "global_step": 102824, "epoch": 1155, "lr": 6.87485442618997e-05}
{"train_loss": 0.16557003557682037, "global_step": 102825, "epoch": 1155, "lr": 6.874800682359682e-05}
{"train_loss": 0.1304805427789688, "global_step": 102826, "epoch": 1155, "lr": 6.87474693827735e-05}
{"train_loss": 0.17086689174175262, "global_step": 102827, "epoch": 1155, "lr": 6.874693193942981e-05}
{"train_loss": 0.2422298640012741, "global_step": 102828, "epoch": 1155, "lr": 6.874639449356583e-05}
{"train_loss": 0.2329847514629364, "global_step": 102829, "epoch": 1155, "lr": 6.874585704518163e-05}
{"train_loss": 0.26841044425964355, "global_step": 102830, "epoch": 1155, "lr": 6.874531959427726e-05}
{"train_loss": 0.17454412579536438, "global_step": 102831, "epoch": 1155, "lr": 6.874478214085283e-05}
{"train_loss": 0.1188950389623642, "global_step": 102832, "epoch": 1155, "lr": 6.87442446849084e-05}
{"train_loss": 0.24168792366981506, "global_step": 102833, "epoch": 1155, "lr": 6.874370722644402e-05}
{"train_loss": 0.187123641371727, "global_step": 102834, "epoch": 1155, "lr": 6.874316976545978e-05}
{"train_loss": 0.18352475762367249, "global_step": 102835, "epoch": 1155, "lr": 6.874263230195575e-05}
{"train_loss": 0.2414533495903015, "global_step": 102836, "epoch": 1155, "lr": 6.874209483593198e-05}
{"train_loss": 0.14713455736637115, "global_step": 102837, "epoch": 1155, "lr": 6.874155736738858e-05}
{"train_loss": 0.20633642375469208, "global_step": 102838, "epoch": 1155, "lr": 6.874101989632559e-05}
{"train_loss": 0.16367468237876892, "global_step": 102839, "epoch": 1155, "lr": 6.874048242274313e-05}
{"train_loss": 0.1584281474351883, "global_step": 102840, "epoch": 1155, "lr": 6.873994494664122e-05}
{"train_loss": 0.22101782262325287, "global_step": 102841, "epoch": 1155, "lr": 6.873940746801997e-05}
{"train_loss": 0.19850458204746246, "global_step": 102842, "epoch": 1155, "lr": 6.873886998687942e-05}
{"train_loss": 0.2825973331928253, "global_step": 102843, "epoch": 1155, "lr": 6.873833250321966e-05}
{"train_loss": 0.15435370802879333, "global_step": 102844, "epoch": 1155, "lr": 6.873779501704077e-05}
{"train_loss": 0.1587507724761963, "global_step": 102845, "epoch": 1155, "lr": 6.87372575283428e-05}
{"train_loss": 0.20073673129081726, "global_step": 102846, "epoch": 1155, "lr": 6.873672003712583e-05}
{"train_loss": 0.13008296489715576, "global_step": 102847, "epoch": 1155, "lr": 6.873618254338994e-05}
{"train_loss": 0.13478097319602966, "global_step": 102848, "epoch": 1155, "lr": 6.87356450471352e-05}
{"train_loss": 0.16797205805778503, "global_step": 102849, "epoch": 1155, "lr": 6.87351075483617e-05}
{"train_loss": 0.13928961753845215, "global_step": 102850, "epoch": 1155, "lr": 6.873457004706945e-05}
{"train_loss": 0.197395458817482, "global_step": 102851, "epoch": 1155, "lr": 6.873403254325861e-05}
{"train_loss": 0.14600208401679993, "global_step": 102852, "epoch": 1155, "lr": 6.873349503692918e-05}
{"train_loss": 0.18195292353630066, "global_step": 102853, "epoch": 1155, "lr": 6.873295752808127e-05}
{"train_loss": 0.1842711716890335, "global_step": 102854, "epoch": 1155, "lr": 6.873242001671494e-05}
{"train_loss": 0.11859779059886932, "global_step": 102855, "epoch": 1155, "lr": 6.873188250283027e-05}
{"train_loss": 0.18538309633731842, "global_step": 102856, "epoch": 1155, "lr": 6.873134498642731e-05}
{"train_loss": 0.17795610427856445, "global_step": 102857, "epoch": 1155, "lr": 6.873080746750618e-05}
{"train_loss": 0.10160902887582779, "global_step": 102858, "epoch": 1155, "lr": 6.87302699460669e-05}
{"train_loss": 0.24895250797271729, "global_step": 102859, "epoch": 1155, "lr": 6.872973242210956e-05}
{"train_loss": 0.17786085605621338, "global_step": 102860, "epoch": 1155, "lr": 6.872919489563424e-05}
{"train_loss": 0.11832529306411743, "global_step": 102861, "epoch": 1155, "lr": 6.8728657366641e-05}
{"train_loss": 0.11042086780071259, "global_step": 102862, "epoch": 1155, "lr": 6.872811983512993e-05}
{"train_loss": 0.19383499026298523, "global_step": 102863, "epoch": 1155, "lr": 6.87275823011011e-05}
{"train_loss": 0.18638306856155396, "global_step": 102864, "epoch": 1155, "lr": 6.872704476455457e-05}
{"train_loss": 0.09726624190807343, "global_step": 102865, "epoch": 1155, "lr": 6.872650722549042e-05}
{"train_loss": 0.22802481055259705, "global_step": 102866, "epoch": 1155, "lr": 6.87259696839087e-05}
{"train_loss": 0.22070476412773132, "global_step": 102867, "epoch": 1155, "lr": 6.872543213980952e-05}
{"train_loss": 0.1454296112060547, "global_step": 102868, "epoch": 1155, "lr": 6.872489459319293e-05}
{"train_loss": 0.20473474264144897, "global_step": 102869, "epoch": 1155, "lr": 6.8724357044059e-05}
{"train_loss": 0.1821029782295227, "global_step": 102870, "epoch": 1155, "lr": 6.872381949240782e-05}
{"train_loss": 0.10625410079956055, "global_step": 102871, "epoch": 1155, "lr": 6.872328193823945e-05}
{"train_loss": 0.08645937591791153, "global_step": 102872, "epoch": 1155, "lr": 6.872274438155395e-05}
{"train_loss": 0.08783257007598877, "global_step": 102873, "epoch": 1155, "lr": 6.872220682235141e-05}
{"train_loss": 0.10252445191144943, "global_step": 102874, "epoch": 1155, "lr": 6.872166926063191e-05}
{"train_loss": 0.11455410718917847, "global_step": 102875, "epoch": 1155, "lr": 6.87211316963955e-05}
{"train_loss": 0.10424970090389252, "global_step": 102876, "epoch": 1155, "lr": 6.872059412964225e-05}
{"train_loss": 0.14579002559185028, "global_step": 102877, "epoch": 1155, "lr": 6.872005656037226e-05}
{"train_loss": 0.16819022595882416, "global_step": 102878, "epoch": 1155, "lr": 6.871951898858558e-05}
{"train_loss": 0.1932234764099121, "global_step": 102879, "epoch": 1155, "lr": 6.87189814142823e-05}
{"train_loss": 0.17487645149230957, "global_step": 102880, "epoch": 1155, "lr": 6.871844383746247e-05}
{"train_loss": 0.18585728108882904, "global_step": 102881, "epoch": 1155, "lr": 6.871790625812617e-05}
{"train_loss": 0.15364119410514832, "global_step": 102882, "epoch": 1155, "lr": 6.871736867627347e-05}
{"train_loss": 0.16897840571872305, "global_step": 102883, "epoch": 1155, "lr": 6.871683109190448e-05, "val_loss": 4.655364513397217, "train_action_mse_error": 45.063079833984375}
{"train_loss": 0.14560654759407043, "global_step": 102884, "epoch": 1156, "lr": 6.871629350501921e-05}
{"train_loss": 0.18275083601474762, "global_step": 102885, "epoch": 1156, "lr": 6.871575591561777e-05}
{"train_loss": 0.16820700466632843, "global_step": 102886, "epoch": 1156, "lr": 6.871521832370023e-05}
{"train_loss": 0.18143919110298157, "global_step": 102887, "epoch": 1156, "lr": 6.871468072926665e-05}
{"train_loss": 0.16888093948364258, "global_step": 102888, "epoch": 1156, "lr": 6.871414313231712e-05}
{"train_loss": 0.16567131876945496, "global_step": 102889, "epoch": 1156, "lr": 6.871360553285169e-05}
{"train_loss": 0.16709664463996887, "global_step": 102890, "epoch": 1156, "lr": 6.871306793087045e-05}
{"train_loss": 0.11851881444454193, "global_step": 102891, "epoch": 1156, "lr": 6.871253032637346e-05}
{"train_loss": 0.14142988622188568, "global_step": 102892, "epoch": 1156, "lr": 6.87119927193608e-05}
{"train_loss": 0.1471652090549469, "global_step": 102893, "epoch": 1156, "lr": 6.871145510983255e-05}
{"train_loss": 0.14735627174377441, "global_step": 102894, "epoch": 1156, "lr": 6.871091749778876e-05}
{"train_loss": 0.14037850499153137, "global_step": 102895, "epoch": 1156, "lr": 6.871037988322953e-05}
{"train_loss": 0.14711527526378632, "global_step": 102896, "epoch": 1156, "lr": 6.870984226615492e-05}
{"train_loss": 0.18548241257667542, "global_step": 102897, "epoch": 1156, "lr": 6.870930464656498e-05}
{"train_loss": 0.10365448147058487, "global_step": 102898, "epoch": 1156, "lr": 6.870876702445981e-05}
{"train_loss": 0.10282668471336365, "global_step": 102899, "epoch": 1156, "lr": 6.870822939983948e-05}
{"train_loss": 0.10391800850629807, "global_step": 102900, "epoch": 1156, "lr": 6.870769177270404e-05}
{"train_loss": 0.19429366290569305, "global_step": 102901, "epoch": 1156, "lr": 6.87071541430536e-05}
{"train_loss": 0.12389092892408371, "global_step": 102902, "epoch": 1156, "lr": 6.870661651088819e-05}
{"train_loss": 0.19235362112522125, "global_step": 102903, "epoch": 1156, "lr": 6.870607887620792e-05}
{"train_loss": 0.08492954075336456, "global_step": 102904, "epoch": 1156, "lr": 6.870554123901284e-05}
{"train_loss": 0.2066429704427719, "global_step": 102905, "epoch": 1156, "lr": 6.870500359930303e-05}
{"train_loss": 0.13245290517807007, "global_step": 102906, "epoch": 1156, "lr": 6.870446595707856e-05}
{"train_loss": 0.19925183057785034, "global_step": 102907, "epoch": 1156, "lr": 6.870392831233951e-05}
{"train_loss": 0.14275449514389038, "global_step": 102908, "epoch": 1156, "lr": 6.870339066508593e-05}
{"train_loss": 0.12446600943803787, "global_step": 102909, "epoch": 1156, "lr": 6.870285301531792e-05}
{"train_loss": 0.0852125883102417, "global_step": 102910, "epoch": 1156, "lr": 6.870231536303552e-05}
{"train_loss": 0.15728315711021423, "global_step": 102911, "epoch": 1156, "lr": 6.870177770823884e-05}
{"train_loss": 0.11224783211946487, "global_step": 102912, "epoch": 1156, "lr": 6.870124005092793e-05}
{"train_loss": 0.1027546301484108, "global_step": 102913, "epoch": 1156, "lr": 6.870070239110287e-05}
{"train_loss": 0.18035341799259186, "global_step": 102914, "epoch": 1156, "lr": 6.870016472876371e-05}
{"train_loss": 0.1783854067325592, "global_step": 102915, "epoch": 1156, "lr": 6.869962706391055e-05}
{"train_loss": 0.12680822610855103, "global_step": 102916, "epoch": 1156, "lr": 6.869908939654346e-05}
{"train_loss": 0.19207599759101868, "global_step": 102917, "epoch": 1156, "lr": 6.869855172666251e-05}
{"train_loss": 0.14003358781337738, "global_step": 102918, "epoch": 1156, "lr": 6.869801405426776e-05}
{"train_loss": 0.13813908398151398, "global_step": 102919, "epoch": 1156, "lr": 6.869747637935928e-05}
{"train_loss": 0.12844887375831604, "global_step": 102920, "epoch": 1156, "lr": 6.869693870193717e-05}
{"train_loss": 0.13863857090473175, "global_step": 102921, "epoch": 1156, "lr": 6.869640102200148e-05}
{"train_loss": 0.19590751826763153, "global_step": 102922, "epoch": 1156, "lr": 6.869586333955228e-05}
{"train_loss": 0.2187042236328125, "global_step": 102923, "epoch": 1156, "lr": 6.869532565458964e-05}
{"train_loss": 0.09772241860628128, "global_step": 102924, "epoch": 1156, "lr": 6.869478796711367e-05}
{"train_loss": 0.12353228032588959, "global_step": 102925, "epoch": 1156, "lr": 6.86942502771244e-05}
{"train_loss": 0.1880597472190857, "global_step": 102926, "epoch": 1156, "lr": 6.869371258462191e-05}
{"train_loss": 0.23291684687137604, "global_step": 102927, "epoch": 1156, "lr": 6.869317488960627e-05}
{"train_loss": 0.13185454905033112, "global_step": 102928, "epoch": 1156, "lr": 6.869263719207758e-05}
{"train_loss": 0.116179458796978, "global_step": 102929, "epoch": 1156, "lr": 6.869209949203588e-05}
{"train_loss": 0.1358911693096161, "global_step": 102930, "epoch": 1156, "lr": 6.869156178948126e-05}
{"train_loss": 0.06813501566648483, "global_step": 102931, "epoch": 1156, "lr": 6.86910240844138e-05}
{"train_loss": 0.18420471251010895, "global_step": 102932, "epoch": 1156, "lr": 6.869048637683352e-05}
{"train_loss": 0.10191957652568817, "global_step": 102933, "epoch": 1156, "lr": 6.868994866674057e-05}
{"train_loss": 0.16549399495124817, "global_step": 102934, "epoch": 1156, "lr": 6.868941095413499e-05}
{"train_loss": 0.2504129111766815, "global_step": 102935, "epoch": 1156, "lr": 6.868887323901681e-05}
{"train_loss": 0.12392089515924454, "global_step": 102936, "epoch": 1156, "lr": 6.868833552138616e-05}
{"train_loss": 0.15362247824668884, "global_step": 102937, "epoch": 1156, "lr": 6.86877978012431e-05}
{"train_loss": 0.13512161374092102, "global_step": 102938, "epoch": 1156, "lr": 6.868726007858768e-05}
{"train_loss": 0.16561031341552734, "global_step": 102939, "epoch": 1156, "lr": 6.868672235341999e-05}
{"train_loss": 0.16534292697906494, "global_step": 102940, "epoch": 1156, "lr": 6.86861846257401e-05}
{"train_loss": 0.15044546127319336, "global_step": 102941, "epoch": 1156, "lr": 6.868564689554806e-05}
{"train_loss": 0.14585378766059875, "global_step": 102942, "epoch": 1156, "lr": 6.868510916284399e-05}
{"train_loss": 0.18765507638454437, "global_step": 102943, "epoch": 1156, "lr": 6.868457142762792e-05}
{"train_loss": 0.19409716129302979, "global_step": 102944, "epoch": 1156, "lr": 6.868403368989995e-05}
{"train_loss": 0.24731487035751343, "global_step": 102945, "epoch": 1156, "lr": 6.868349594966013e-05}
{"train_loss": 0.1965462863445282, "global_step": 102946, "epoch": 1156, "lr": 6.868295820690853e-05}
{"train_loss": 0.11041083186864853, "global_step": 102947, "epoch": 1156, "lr": 6.868242046164524e-05}
{"train_loss": 0.10310737788677216, "global_step": 102948, "epoch": 1156, "lr": 6.868188271387034e-05}
{"train_loss": 0.10820893198251724, "global_step": 102949, "epoch": 1156, "lr": 6.868134496358388e-05}
{"train_loss": 0.186137855052948, "global_step": 102950, "epoch": 1156, "lr": 6.868080721078594e-05}
{"train_loss": 0.15150046348571777, "global_step": 102951, "epoch": 1156, "lr": 6.86802694554766e-05}
{"train_loss": 0.12475413084030151, "global_step": 102952, "epoch": 1156, "lr": 6.867973169765591e-05}
{"train_loss": 0.09772617369890213, "global_step": 102953, "epoch": 1156, "lr": 6.867919393732397e-05}
{"train_loss": 0.141577810049057, "global_step": 102954, "epoch": 1156, "lr": 6.867865617448084e-05}
{"train_loss": 0.14259205758571625, "global_step": 102955, "epoch": 1156, "lr": 6.867811840912659e-05}
{"train_loss": 0.14090414345264435, "global_step": 102956, "epoch": 1156, "lr": 6.86775806412613e-05}
{"train_loss": 0.16980445384979248, "global_step": 102957, "epoch": 1156, "lr": 6.867704287088503e-05}
{"train_loss": 0.1309380978345871, "global_step": 102958, "epoch": 1156, "lr": 6.867650509799786e-05}
{"train_loss": 0.2143026739358902, "global_step": 102959, "epoch": 1156, "lr": 6.867596732259987e-05}
{"train_loss": 0.23613397777080536, "global_step": 102960, "epoch": 1156, "lr": 6.86754295446911e-05}
{"train_loss": 0.1496594250202179, "global_step": 102961, "epoch": 1156, "lr": 6.867489176427167e-05}
{"train_loss": 0.20594027638435364, "global_step": 102962, "epoch": 1156, "lr": 6.867435398134162e-05}
{"train_loss": 0.1752421110868454, "global_step": 102963, "epoch": 1156, "lr": 6.867381619590102e-05}
{"train_loss": 0.1366906315088272, "global_step": 102964, "epoch": 1156, "lr": 6.867327840794997e-05}
{"train_loss": 0.10543564707040787, "global_step": 102965, "epoch": 1156, "lr": 6.867274061748853e-05}
{"train_loss": 0.2639355957508087, "global_step": 102966, "epoch": 1156, "lr": 6.867220282451675e-05}
{"train_loss": 0.1556321382522583, "global_step": 102967, "epoch": 1156, "lr": 6.867166502903472e-05}
{"train_loss": 0.13747632503509521, "global_step": 102968, "epoch": 1156, "lr": 6.867112723104254e-05}
{"train_loss": 0.20890098810195923, "global_step": 102969, "epoch": 1156, "lr": 6.867058943054022e-05}
{"train_loss": 0.21210578083992004, "global_step": 102970, "epoch": 1156, "lr": 6.867005162752789e-05}
{"train_loss": 0.16630059480667114, "global_step": 102971, "epoch": 1156, "lr": 6.866951382200558e-05}
{"train_loss": 0.15439313139473454, "global_step": 102972, "epoch": 1156, "lr": 6.866897601397339e-05, "val_loss": 4.836241245269775}
{"train_loss": 0.1980164498090744, "global_step": 102973, "epoch": 1157, "lr": 6.86684382034314e-05}
{"train_loss": 0.10704206675291061, "global_step": 102974, "epoch": 1157, "lr": 6.866790039037964e-05}
{"train_loss": 0.12015223503112793, "global_step": 102975, "epoch": 1157, "lr": 6.866736257481823e-05}
{"train_loss": 0.22270146012306213, "global_step": 102976, "epoch": 1157, "lr": 6.86668247567472e-05}
{"train_loss": 0.2272752970457077, "global_step": 102977, "epoch": 1157, "lr": 6.866628693616666e-05}
{"train_loss": 0.14786875247955322, "global_step": 102978, "epoch": 1157, "lr": 6.866574911307666e-05}
{"train_loss": 0.1625155210494995, "global_step": 102979, "epoch": 1157, "lr": 6.866521128747727e-05}
{"train_loss": 0.16243255138397217, "global_step": 102980, "epoch": 1157, "lr": 6.866467345936857e-05}
{"train_loss": 0.15211358666419983, "global_step": 102981, "epoch": 1157, "lr": 6.866413562875065e-05}
{"train_loss": 0.16212418675422668, "global_step": 102982, "epoch": 1157, "lr": 6.866359779562355e-05}
{"train_loss": 0.23451673984527588, "global_step": 102983, "epoch": 1157, "lr": 6.866305995998735e-05}
{"train_loss": 0.13020411133766174, "global_step": 102984, "epoch": 1157, "lr": 6.866252212184214e-05}
{"train_loss": 0.2431287169456482, "global_step": 102985, "epoch": 1157, "lr": 6.866198428118798e-05}
{"train_loss": 0.17823433876037598, "global_step": 102986, "epoch": 1157, "lr": 6.866144643802495e-05}
{"train_loss": 0.16174429655075073, "global_step": 102987, "epoch": 1157, "lr": 6.86609085923531e-05}
{"train_loss": 0.1575351059436798, "global_step": 102988, "epoch": 1157, "lr": 6.866037074417253e-05}
{"train_loss": 0.12099531292915344, "global_step": 102989, "epoch": 1157, "lr": 6.86598328934833e-05}
{"train_loss": 0.17032648622989655, "global_step": 102990, "epoch": 1157, "lr": 6.865929504028548e-05}
{"train_loss": 0.23825916647911072, "global_step": 102991, "epoch": 1157, "lr": 6.865875718457915e-05}
{"train_loss": 0.1516352891921997, "global_step": 102992, "epoch": 1157, "lr": 6.865821932636437e-05}
{"train_loss": 0.1480703502893448, "global_step": 102993, "epoch": 1157, "lr": 6.865768146564122e-05}
{"train_loss": 0.183816596865654, "global_step": 102994, "epoch": 1157, "lr": 6.865714360240978e-05}
{"train_loss": 0.1301020234823227, "global_step": 102995, "epoch": 1157, "lr": 6.865660573667011e-05}
{"train_loss": 0.09107040613889694, "global_step": 102996, "epoch": 1157, "lr": 6.86560678684223e-05}
{"train_loss": 0.18350914120674133, "global_step": 102997, "epoch": 1157, "lr": 6.865552999766637e-05}
{"train_loss": 0.15608079731464386, "global_step": 102998, "epoch": 1157, "lr": 6.865499212440247e-05}
{"train_loss": 0.1921301931142807, "global_step": 102999, "epoch": 1157, "lr": 6.865445424863062e-05}
{"train_loss": 0.1469714343547821, "global_step": 103000, "epoch": 1157, "lr": 6.86539163703509e-05}
{"train_loss": 0.2547174394130707, "global_step": 103001, "epoch": 1157, "lr": 6.86533784895634e-05}
{"train_loss": 0.21202895045280457, "global_step": 103002, "epoch": 1157, "lr": 6.865284060626818e-05}
{"train_loss": 0.19438636302947998, "global_step": 103003, "epoch": 1157, "lr": 6.86523027204653e-05}
{"train_loss": 0.1937694251537323, "global_step": 103004, "epoch": 1157, "lr": 6.865176483215486e-05}
{"train_loss": 0.2299419492483139, "global_step": 103005, "epoch": 1157, "lr": 6.865122694133692e-05}
{"train_loss": 0.17814688384532928, "global_step": 103006, "epoch": 1157, "lr": 6.865068904801153e-05}
{"train_loss": 0.11520084738731384, "global_step": 103007, "epoch": 1157, "lr": 6.86501511521788e-05}
{"train_loss": 0.1781143993139267, "global_step": 103008, "epoch": 1157, "lr": 6.864961325383877e-05}
{"train_loss": 0.16582567989826202, "global_step": 103009, "epoch": 1157, "lr": 6.864907535299155e-05}
{"train_loss": 0.12375194579362869, "global_step": 103010, "epoch": 1157, "lr": 6.864853744963716e-05}
{"train_loss": 0.09954558312892914, "global_step": 103011, "epoch": 1157, "lr": 6.864799954377572e-05}
{"train_loss": 0.1528819054365158, "global_step": 103012, "epoch": 1157, "lr": 6.864746163540728e-05}
{"train_loss": 0.12530645728111267, "global_step": 103013, "epoch": 1157, "lr": 6.864692372453192e-05}
{"train_loss": 0.12043648213148117, "global_step": 103014, "epoch": 1157, "lr": 6.86463858111497e-05}
{"train_loss": 0.14274822175502777, "global_step": 103015, "epoch": 1157, "lr": 6.86458478952607e-05}
{"train_loss": 0.1282981038093567, "global_step": 103016, "epoch": 1157, "lr": 6.864530997686501e-05}
{"train_loss": 0.16894540190696716, "global_step": 103017, "epoch": 1157, "lr": 6.864477205596265e-05}
{"train_loss": 0.13055863976478577, "global_step": 103018, "epoch": 1157, "lr": 6.864423413255375e-05}
{"train_loss": 0.2534584403038025, "global_step": 103019, "epoch": 1157, "lr": 6.864369620663836e-05}
{"train_loss": 0.22909541428089142, "global_step": 103020, "epoch": 1157, "lr": 6.864315827821657e-05}
{"train_loss": 0.20138993859291077, "global_step": 103021, "epoch": 1157, "lr": 6.864262034728841e-05}
{"train_loss": 0.15837372839450836, "global_step": 103022, "epoch": 1157, "lr": 6.864208241385399e-05}
{"train_loss": 0.1977461278438568, "global_step": 103023, "epoch": 1157, "lr": 6.864154447791335e-05}
{"train_loss": 0.13675649464130402, "global_step": 103024, "epoch": 1157, "lr": 6.86410065394666e-05}
{"train_loss": 0.17294944822788239, "global_step": 103025, "epoch": 1157, "lr": 6.86404685985138e-05}
{"train_loss": 0.10769103467464447, "global_step": 103026, "epoch": 1157, "lr": 6.863993065505499e-05}
{"train_loss": 0.09122894704341888, "global_step": 103027, "epoch": 1157, "lr": 6.86393927090903e-05}
{"train_loss": 0.15200382471084595, "global_step": 103028, "epoch": 1157, "lr": 6.863885476061974e-05}
{"train_loss": 0.14523395895957947, "global_step": 103029, "epoch": 1157, "lr": 6.863831680964344e-05}
{"train_loss": 0.10654859244823456, "global_step": 103030, "epoch": 1157, "lr": 6.863777885616142e-05}
{"train_loss": 0.12954489886760712, "global_step": 103031, "epoch": 1157, "lr": 6.863724090017379e-05}
{"train_loss": 0.16933287680149078, "global_step": 103032, "epoch": 1157, "lr": 6.863670294168062e-05}
{"train_loss": 0.1549604833126068, "global_step": 103033, "epoch": 1157, "lr": 6.863616498068196e-05}
{"train_loss": 0.24746431410312653, "global_step": 103034, "epoch": 1157, "lr": 6.86356270171779e-05}
{"train_loss": 0.16711102426052094, "global_step": 103035, "epoch": 1157, "lr": 6.86350890511685e-05}
{"train_loss": 0.09331949055194855, "global_step": 103036, "epoch": 1157, "lr": 6.863455108265384e-05}
{"train_loss": 0.14582912623882294, "global_step": 103037, "epoch": 1157, "lr": 6.8634013111634e-05}
{"train_loss": 0.19223634898662567, "global_step": 103038, "epoch": 1157, "lr": 6.863347513810902e-05}
{"train_loss": 0.11929978430271149, "global_step": 103039, "epoch": 1157, "lr": 6.863293716207904e-05}
{"train_loss": 0.1461634784936905, "global_step": 103040, "epoch": 1157, "lr": 6.863239918354407e-05}
{"train_loss": 0.1588667333126068, "global_step": 103041, "epoch": 1157, "lr": 6.86318612025042e-05}
{"train_loss": 0.2682245373725891, "global_step": 103042, "epoch": 1157, "lr": 6.863132321895949e-05}
{"train_loss": 0.2212766706943512, "global_step": 103043, "epoch": 1157, "lr": 6.863078523291004e-05}
{"train_loss": 0.13657696545124054, "global_step": 103044, "epoch": 1157, "lr": 6.86302472443559e-05}
{"train_loss": 0.17166219651699066, "global_step": 103045, "epoch": 1157, "lr": 6.862970925329715e-05}
{"train_loss": 0.14743399620056152, "global_step": 103046, "epoch": 1157, "lr": 6.862917125973388e-05}
{"train_loss": 0.17681361734867096, "global_step": 103047, "epoch": 1157, "lr": 6.862863326366612e-05}
{"train_loss": 0.1059010922908783, "global_step": 103048, "epoch": 1157, "lr": 6.862809526509398e-05}
{"train_loss": 0.12395090609788895, "global_step": 103049, "epoch": 1157, "lr": 6.862755726401752e-05}
{"train_loss": 0.1441345363855362, "global_step": 103050, "epoch": 1157, "lr": 6.862701926043681e-05}
{"train_loss": 0.1413516253232956, "global_step": 103051, "epoch": 1157, "lr": 6.862648125435192e-05}
{"train_loss": 0.2057725489139557, "global_step": 103052, "epoch": 1157, "lr": 6.862594324576294e-05}
{"train_loss": 0.16568629443645477, "global_step": 103053, "epoch": 1157, "lr": 6.862540523466992e-05}
{"train_loss": 0.17537812888622284, "global_step": 103054, "epoch": 1157, "lr": 6.862486722107293e-05}
{"train_loss": 0.19254249334335327, "global_step": 103055, "epoch": 1157, "lr": 6.862432920497207e-05}
{"train_loss": 0.17294862866401672, "global_step": 103056, "epoch": 1157, "lr": 6.862379118636738e-05}
{"train_loss": 0.1333283632993698, "global_step": 103057, "epoch": 1157, "lr": 6.862325316525896e-05}
{"train_loss": 0.18209823966026306, "global_step": 103058, "epoch": 1157, "lr": 6.862271514164686e-05}
{"train_loss": 0.19489596784114838, "global_step": 103059, "epoch": 1157, "lr": 6.862217711553119e-05}
{"train_loss": 0.15491081774234772, "global_step": 103060, "epoch": 1157, "lr": 6.862163908691198e-05}
{"train_loss": 0.16397278462902884, "global_step": 103061, "epoch": 1157, "lr": 6.862110105578931e-05, "val_loss": 4.698594570159912}
{"train_loss": 0.15395954251289368, "global_step": 103062, "epoch": 1158, "lr": 6.862056302216328e-05}
{"train_loss": 0.18688209354877472, "global_step": 103063, "epoch": 1158, "lr": 6.862002498603393e-05}
{"train_loss": 0.190989688038826, "global_step": 103064, "epoch": 1158, "lr": 6.861948694740133e-05}
{"train_loss": 0.2104371339082718, "global_step": 103065, "epoch": 1158, "lr": 6.861894890626557e-05}
{"train_loss": 0.1699681133031845, "global_step": 103066, "epoch": 1158, "lr": 6.861841086262674e-05}
{"train_loss": 0.17551405727863312, "global_step": 103067, "epoch": 1158, "lr": 6.861787281648488e-05}
{"train_loss": 0.09946806728839874, "global_step": 103068, "epoch": 1158, "lr": 6.861733476784007e-05}
{"train_loss": 0.12134841829538345, "global_step": 103069, "epoch": 1158, "lr": 6.861679671669238e-05}
{"train_loss": 0.09599046409130096, "global_step": 103070, "epoch": 1158, "lr": 6.861625866304191e-05}
{"train_loss": 0.2274985909461975, "global_step": 103071, "epoch": 1158, "lr": 6.86157206068887e-05}
{"train_loss": 0.12617778778076172, "global_step": 103072, "epoch": 1158, "lr": 6.861518254823283e-05}
{"train_loss": 0.16214869916439056, "global_step": 103073, "epoch": 1158, "lr": 6.861464448707438e-05}
{"train_loss": 0.16289375722408295, "global_step": 103074, "epoch": 1158, "lr": 6.861410642341342e-05}
{"train_loss": 0.15935993194580078, "global_step": 103075, "epoch": 1158, "lr": 6.861356835725002e-05}
{"train_loss": 0.15096794068813324, "global_step": 103076, "epoch": 1158, "lr": 6.861303028858424e-05}
{"train_loss": 0.16988296806812286, "global_step": 103077, "epoch": 1158, "lr": 6.861249221741617e-05}
{"train_loss": 0.18306976556777954, "global_step": 103078, "epoch": 1158, "lr": 6.861195414374589e-05}
{"train_loss": 0.12439088523387909, "global_step": 103079, "epoch": 1158, "lr": 6.861141606757346e-05}
{"train_loss": 0.16477327048778534, "global_step": 103080, "epoch": 1158, "lr": 6.861087798889893e-05}
{"train_loss": 0.17161206901073456, "global_step": 103081, "epoch": 1158, "lr": 6.86103399077224e-05}
{"train_loss": 0.137091726064682, "global_step": 103082, "epoch": 1158, "lr": 6.860980182404396e-05}
{"train_loss": 0.24805079400539398, "global_step": 103083, "epoch": 1158, "lr": 6.860926373786365e-05}
{"train_loss": 0.06663067638874054, "global_step": 103084, "epoch": 1158, "lr": 6.860872564918154e-05}
{"train_loss": 0.1905342936515808, "global_step": 103085, "epoch": 1158, "lr": 6.860818755799772e-05}
{"train_loss": 0.19691318273544312, "global_step": 103086, "epoch": 1158, "lr": 6.860764946431225e-05}
{"train_loss": 0.11417286098003387, "global_step": 103087, "epoch": 1158, "lr": 6.860711136812521e-05}
{"train_loss": 0.08573578298091888, "global_step": 103088, "epoch": 1158, "lr": 6.860657326943667e-05}
{"train_loss": 0.13011369109153748, "global_step": 103089, "epoch": 1158, "lr": 6.86060351682467e-05}
{"train_loss": 0.11541666090488434, "global_step": 103090, "epoch": 1158, "lr": 6.860549706455538e-05}
{"train_loss": 0.13715247809886932, "global_step": 103091, "epoch": 1158, "lr": 6.860495895836279e-05}
{"train_loss": 0.18576979637145996, "global_step": 103092, "epoch": 1158, "lr": 6.860442084966897e-05}
{"train_loss": 0.11650305986404419, "global_step": 103093, "epoch": 1158, "lr": 6.860388273847403e-05}
{"train_loss": 0.10101237893104553, "global_step": 103094, "epoch": 1158, "lr": 6.860334462477801e-05}
{"train_loss": 0.11360066384077072, "global_step": 103095, "epoch": 1158, "lr": 6.8602806508581e-05}
{"train_loss": 0.10817784816026688, "global_step": 103096, "epoch": 1158, "lr": 6.860226838988308e-05}
{"train_loss": 0.20004068315029144, "global_step": 103097, "epoch": 1158, "lr": 6.860173026868431e-05}
{"train_loss": 0.14890064299106598, "global_step": 103098, "epoch": 1158, "lr": 6.860119214498477e-05}
{"train_loss": 0.14321401715278625, "global_step": 103099, "epoch": 1158, "lr": 6.860065401878451e-05}
{"train_loss": 0.14475801587104797, "global_step": 103100, "epoch": 1158, "lr": 6.860011589008362e-05}
{"train_loss": 0.17779365181922913, "global_step": 103101, "epoch": 1158, "lr": 6.859957775888217e-05}
{"train_loss": 0.1808801144361496, "global_step": 103102, "epoch": 1158, "lr": 6.859903962518024e-05}
{"train_loss": 0.21557606756687164, "global_step": 103103, "epoch": 1158, "lr": 6.859850148897789e-05}
{"train_loss": 0.11721853166818619, "global_step": 103104, "epoch": 1158, "lr": 6.859796335027521e-05}
{"train_loss": 0.14797060191631317, "global_step": 103105, "epoch": 1158, "lr": 6.859742520907224e-05}
{"train_loss": 0.10975485295057297, "global_step": 103106, "epoch": 1158, "lr": 6.85968870653691e-05}
{"train_loss": 0.14083147048950195, "global_step": 103107, "epoch": 1158, "lr": 6.859634891916582e-05}
{"train_loss": 0.09493151307106018, "global_step": 103108, "epoch": 1158, "lr": 6.859581077046248e-05}
{"train_loss": 0.13216015696525574, "global_step": 103109, "epoch": 1158, "lr": 6.859527261925916e-05}
{"train_loss": 0.14298538863658905, "global_step": 103110, "epoch": 1158, "lr": 6.859473446555595e-05}
{"train_loss": 0.08718186616897583, "global_step": 103111, "epoch": 1158, "lr": 6.859419630935289e-05}
{"train_loss": 0.17018434405326843, "global_step": 103112, "epoch": 1158, "lr": 6.859365815065007e-05}
{"train_loss": 0.14915354549884796, "global_step": 103113, "epoch": 1158, "lr": 6.859311998944756e-05}
{"train_loss": 0.19235830008983612, "global_step": 103114, "epoch": 1158, "lr": 6.859258182574544e-05}
{"train_loss": 0.08100629597902298, "global_step": 103115, "epoch": 1158, "lr": 6.859204365954376e-05}
{"train_loss": 0.14557798206806183, "global_step": 103116, "epoch": 1158, "lr": 6.85915054908426e-05}
{"train_loss": 0.17321857810020447, "global_step": 103117, "epoch": 1158, "lr": 6.859096731964208e-05}
{"train_loss": 0.07829426974058151, "global_step": 103118, "epoch": 1158, "lr": 6.85904291459422e-05}
{"train_loss": 0.1375626176595688, "global_step": 103119, "epoch": 1158, "lr": 6.858989096974307e-05}
{"train_loss": 0.09393706917762756, "global_step": 103120, "epoch": 1158, "lr": 6.858935279104475e-05}
{"train_loss": 0.19176089763641357, "global_step": 103121, "epoch": 1158, "lr": 6.858881460984732e-05}
{"train_loss": 0.15956270694732666, "global_step": 103122, "epoch": 1158, "lr": 6.858827642615084e-05}
{"train_loss": 0.15140612423419952, "global_step": 103123, "epoch": 1158, "lr": 6.858773823995541e-05}
{"train_loss": 0.15840981900691986, "global_step": 103124, "epoch": 1158, "lr": 6.858720005126108e-05}
{"train_loss": 0.1447928249835968, "global_step": 103125, "epoch": 1158, "lr": 6.858666186006793e-05}
{"train_loss": 0.21557891368865967, "global_step": 103126, "epoch": 1158, "lr": 6.8586123666376e-05}
{"train_loss": 0.15537163615226746, "global_step": 103127, "epoch": 1158, "lr": 6.858558547018543e-05}
{"train_loss": 0.2213478684425354, "global_step": 103128, "epoch": 1158, "lr": 6.858504727149624e-05}
{"train_loss": 0.12029479444026947, "global_step": 103129, "epoch": 1158, "lr": 6.858450907030852e-05}
{"train_loss": 0.14758846163749695, "global_step": 103130, "epoch": 1158, "lr": 6.858397086662234e-05}
{"train_loss": 0.0891905277967453, "global_step": 103131, "epoch": 1158, "lr": 6.858343266043777e-05}
{"train_loss": 0.1493273824453354, "global_step": 103132, "epoch": 1158, "lr": 6.858289445175489e-05}
{"train_loss": 0.16147677600383759, "global_step": 103133, "epoch": 1158, "lr": 6.858235624057376e-05}
{"train_loss": 0.15536804497241974, "global_step": 103134, "epoch": 1158, "lr": 6.858181802689445e-05}
{"train_loss": 0.13871653378009796, "global_step": 103135, "epoch": 1158, "lr": 6.858127981071706e-05}
{"train_loss": 0.12726853787899017, "global_step": 103136, "epoch": 1158, "lr": 6.858074159204164e-05}
{"train_loss": 0.17591185867786407, "global_step": 103137, "epoch": 1158, "lr": 6.858020337086825e-05}
{"train_loss": 0.10857602208852768, "global_step": 103138, "epoch": 1158, "lr": 6.8579665147197e-05}
{"train_loss": 0.14237608015537262, "global_step": 103139, "epoch": 1158, "lr": 6.857912692102792e-05}
{"train_loss": 0.21657957136631012, "global_step": 103140, "epoch": 1158, "lr": 6.857858869236111e-05}
{"train_loss": 0.13282303512096405, "global_step": 103141, "epoch": 1158, "lr": 6.857805046119664e-05}
{"train_loss": 0.20799823105335236, "global_step": 103142, "epoch": 1158, "lr": 6.857751222753458e-05}
{"train_loss": 0.09244430065155029, "global_step": 103143, "epoch": 1158, "lr": 6.8576973991375e-05}
{"train_loss": 0.2501569986343384, "global_step": 103144, "epoch": 1158, "lr": 6.857643575271795e-05}
{"train_loss": 0.21304163336753845, "global_step": 103145, "epoch": 1158, "lr": 6.857589751156354e-05}
{"train_loss": 0.18407195806503296, "global_step": 103146, "epoch": 1158, "lr": 6.857535926791183e-05}
{"train_loss": 0.11262598633766174, "global_step": 103147, "epoch": 1158, "lr": 6.857482102176288e-05}
{"train_loss": 0.12951916456222534, "global_step": 103148, "epoch": 1158, "lr": 6.857428277311678e-05}
{"train_loss": 0.18826240301132202, "global_step": 103149, "epoch": 1158, "lr": 6.85737445219736e-05}
{"train_loss": 0.15127154598745068, "global_step": 103150, "epoch": 1158, "lr": 6.85732062683334e-05, "val_loss": 4.915450572967529}
{"train_loss": 0.07970207929611206, "global_step": 103151, "epoch": 1159, "lr": 6.857266801219625e-05}
{"train_loss": 0.1915871500968933, "global_step": 103152, "epoch": 1159, "lr": 6.857212975356224e-05}
{"train_loss": 0.17299328744411469, "global_step": 103153, "epoch": 1159, "lr": 6.857159149243143e-05}
{"train_loss": 0.1529061198234558, "global_step": 103154, "epoch": 1159, "lr": 6.857105322880391e-05}
{"train_loss": 0.20795638859272003, "global_step": 103155, "epoch": 1159, "lr": 6.857051496267972e-05}
{"train_loss": 0.14518466591835022, "global_step": 103156, "epoch": 1159, "lr": 6.856997669405896e-05}
{"train_loss": 0.15813636779785156, "global_step": 103157, "epoch": 1159, "lr": 6.85694384229417e-05}
{"train_loss": 0.21985316276550293, "global_step": 103158, "epoch": 1159, "lr": 6.8568900149328e-05}
{"train_loss": 0.17013110220432281, "global_step": 103159, "epoch": 1159, "lr": 6.856836187321793e-05}
{"train_loss": 0.1558198630809784, "global_step": 103160, "epoch": 1159, "lr": 6.856782359461157e-05}
{"train_loss": 0.1512485295534134, "global_step": 103161, "epoch": 1159, "lr": 6.8567285313509e-05}
{"train_loss": 0.14303819835186005, "global_step": 103162, "epoch": 1159, "lr": 6.856674702991027e-05}
{"train_loss": 0.15740972757339478, "global_step": 103163, "epoch": 1159, "lr": 6.856620874381548e-05}
{"train_loss": 0.21539929509162903, "global_step": 103164, "epoch": 1159, "lr": 6.856567045522468e-05}
{"train_loss": 0.13733692467212677, "global_step": 103165, "epoch": 1159, "lr": 6.856513216413796e-05}
{"train_loss": 0.1536695808172226, "global_step": 103166, "epoch": 1159, "lr": 6.856459387055538e-05}
{"train_loss": 0.19538617134094238, "global_step": 103167, "epoch": 1159, "lr": 6.856405557447703e-05}
{"train_loss": 0.226144939661026, "global_step": 103168, "epoch": 1159, "lr": 6.856351727590295e-05}
{"train_loss": 0.14631347358226776, "global_step": 103169, "epoch": 1159, "lr": 6.856297897483323e-05}
{"train_loss": 0.14557555317878723, "global_step": 103170, "epoch": 1159, "lr": 6.856244067126798e-05}
{"train_loss": 0.1869972050189972, "global_step": 103171, "epoch": 1159, "lr": 6.85619023652072e-05}
{"train_loss": 0.17664198577404022, "global_step": 103172, "epoch": 1159, "lr": 6.8561364056651e-05}
{"train_loss": 0.23244236409664154, "global_step": 103173, "epoch": 1159, "lr": 6.856082574559947e-05}
{"train_loss": 0.20161263644695282, "global_step": 103174, "epoch": 1159, "lr": 6.856028743205265e-05}
{"train_loss": 0.12908637523651123, "global_step": 103175, "epoch": 1159, "lr": 6.855974911601064e-05}
{"train_loss": 0.12902110815048218, "global_step": 103176, "epoch": 1159, "lr": 6.855921079747349e-05}
{"train_loss": 0.11952503770589828, "global_step": 103177, "epoch": 1159, "lr": 6.855867247644128e-05}
{"train_loss": 0.20169921219348907, "global_step": 103178, "epoch": 1159, "lr": 6.855813415291407e-05}
{"train_loss": 0.1400197446346283, "global_step": 103179, "epoch": 1159, "lr": 6.855759582689195e-05}
{"train_loss": 0.1751984804868698, "global_step": 103180, "epoch": 1159, "lr": 6.8557057498375e-05}
{"train_loss": 0.1611657589673996, "global_step": 103181, "epoch": 1159, "lr": 6.855651916736326e-05}
{"train_loss": 0.1210070475935936, "global_step": 103182, "epoch": 1159, "lr": 6.855598083385684e-05}
{"train_loss": 0.17269369959831238, "global_step": 103183, "epoch": 1159, "lr": 6.855544249785578e-05}
{"train_loss": 0.22521913051605225, "global_step": 103184, "epoch": 1159, "lr": 6.855490415936018e-05}
{"train_loss": 0.12585678696632385, "global_step": 103185, "epoch": 1159, "lr": 6.85543658183701e-05}
{"train_loss": 0.16460593044757843, "global_step": 103186, "epoch": 1159, "lr": 6.855382747488559e-05}
{"train_loss": 0.1498299390077591, "global_step": 103187, "epoch": 1159, "lr": 6.855328912890678e-05}
{"train_loss": 0.16884948313236237, "global_step": 103188, "epoch": 1159, "lr": 6.855275078043368e-05}
{"train_loss": 0.235580712556839, "global_step": 103189, "epoch": 1159, "lr": 6.85522124294664e-05}
{"train_loss": 0.1457148939371109, "global_step": 103190, "epoch": 1159, "lr": 6.855167407600498e-05}
{"train_loss": 0.1972971111536026, "global_step": 103191, "epoch": 1159, "lr": 6.855113572004954e-05}
{"train_loss": 0.14961585402488708, "global_step": 103192, "epoch": 1159, "lr": 6.855059736160011e-05}
{"train_loss": 0.2059212327003479, "global_step": 103193, "epoch": 1159, "lr": 6.855005900065679e-05}
{"train_loss": 0.14152434468269348, "global_step": 103194, "epoch": 1159, "lr": 6.854952063721964e-05}
{"train_loss": 0.2085571587085724, "global_step": 103195, "epoch": 1159, "lr": 6.854898227128873e-05}
{"train_loss": 0.18861986696720123, "global_step": 103196, "epoch": 1159, "lr": 6.854844390286413e-05}
{"train_loss": 0.12348172068595886, "global_step": 103197, "epoch": 1159, "lr": 6.854790553194592e-05}
{"train_loss": 0.11137588322162628, "global_step": 103198, "epoch": 1159, "lr": 6.854736715853419e-05}
{"train_loss": 0.20504604279994965, "global_step": 103199, "epoch": 1159, "lr": 6.854682878262894e-05}
{"train_loss": 0.2040826678276062, "global_step": 103200, "epoch": 1159, "lr": 6.854629040423034e-05}
{"train_loss": 0.18381808698177338, "global_step": 103201, "epoch": 1159, "lr": 6.85457520233384e-05}
{"train_loss": 0.21630141139030457, "global_step": 103202, "epoch": 1159, "lr": 6.854521363995322e-05}
{"train_loss": 0.21769315004348755, "global_step": 103203, "epoch": 1159, "lr": 6.854467525407486e-05}
{"train_loss": 0.18302731215953827, "global_step": 103204, "epoch": 1159, "lr": 6.85441368657034e-05}
{"train_loss": 0.190108522772789, "global_step": 103205, "epoch": 1159, "lr": 6.854359847483888e-05}
{"train_loss": 0.1657833606004715, "global_step": 103206, "epoch": 1159, "lr": 6.854306008148143e-05}
{"train_loss": 0.19156253337860107, "global_step": 103207, "epoch": 1159, "lr": 6.854252168563108e-05}
{"train_loss": 0.17251840233802795, "global_step": 103208, "epoch": 1159, "lr": 6.854198328728792e-05}
{"train_loss": 0.09807676076889038, "global_step": 103209, "epoch": 1159, "lr": 6.8541444886452e-05}
{"train_loss": 0.15183423459529877, "global_step": 103210, "epoch": 1159, "lr": 6.854090648312342e-05}
{"train_loss": 0.19113387167453766, "global_step": 103211, "epoch": 1159, "lr": 6.854036807730224e-05}
{"train_loss": 0.18629421293735504, "global_step": 103212, "epoch": 1159, "lr": 6.853982966898853e-05}
{"train_loss": 0.17603911459445953, "global_step": 103213, "epoch": 1159, "lr": 6.853929125818238e-05}
{"train_loss": 0.15750449895858765, "global_step": 103214, "epoch": 1159, "lr": 6.853875284488385e-05}
{"train_loss": 0.34101730585098267, "global_step": 103215, "epoch": 1159, "lr": 6.853821442909298e-05}
{"train_loss": 0.20279543101787567, "global_step": 103216, "epoch": 1159, "lr": 6.853767601080991e-05}
{"train_loss": 0.12853293120861053, "global_step": 103217, "epoch": 1159, "lr": 6.853713759003463e-05}
{"train_loss": 0.14319656789302826, "global_step": 103218, "epoch": 1159, "lr": 6.853659916676729e-05}
{"train_loss": 0.14819011092185974, "global_step": 103219, "epoch": 1159, "lr": 6.853606074100792e-05}
{"train_loss": 0.15174326300621033, "global_step": 103220, "epoch": 1159, "lr": 6.853552231275661e-05}
{"train_loss": 0.2269519865512848, "global_step": 103221, "epoch": 1159, "lr": 6.853498388201342e-05}
{"train_loss": 0.14590658247470856, "global_step": 103222, "epoch": 1159, "lr": 6.853444544877844e-05}
{"train_loss": 0.1808600127696991, "global_step": 103223, "epoch": 1159, "lr": 6.853390701305172e-05}
{"train_loss": 0.11774192005395889, "global_step": 103224, "epoch": 1159, "lr": 6.853336857483334e-05}
{"train_loss": 0.1643408238887787, "global_step": 103225, "epoch": 1159, "lr": 6.853283013412338e-05}
{"train_loss": 0.19529874622821808, "global_step": 103226, "epoch": 1159, "lr": 6.85322916909219e-05}
{"train_loss": 0.2087317258119583, "global_step": 103227, "epoch": 1159, "lr": 6.853175324522898e-05}
{"train_loss": 0.20175649225711823, "global_step": 103228, "epoch": 1159, "lr": 6.853121479704469e-05}
{"train_loss": 0.1546545922756195, "global_step": 103229, "epoch": 1159, "lr": 6.853067634636911e-05}
{"train_loss": 0.18471167981624603, "global_step": 103230, "epoch": 1159, "lr": 6.85301378932023e-05}
{"train_loss": 0.09960748255252838, "global_step": 103231, "epoch": 1159, "lr": 6.852959943754434e-05}
{"train_loss": 0.11393117159605026, "global_step": 103232, "epoch": 1159, "lr": 6.852906097939531e-05}
{"train_loss": 0.20920300483703613, "global_step": 103233, "epoch": 1159, "lr": 6.852852251875528e-05}
{"train_loss": 0.1688416302204132, "global_step": 103234, "epoch": 1159, "lr": 6.85279840556243e-05}
{"train_loss": 0.16595858335494995, "global_step": 103235, "epoch": 1159, "lr": 6.852744559000245e-05}
{"train_loss": 0.14184480905532837, "global_step": 103236, "epoch": 1159, "lr": 6.852690712188983e-05}
{"train_loss": 0.1972382664680481, "global_step": 103237, "epoch": 1159, "lr": 6.852636865128647e-05}
{"train_loss": 0.14974653720855713, "global_step": 103238, "epoch": 1159, "lr": 6.852583017819249e-05}
{"train_loss": 0.17097332069043364, "global_step": 103239, "epoch": 1159, "lr": 6.852529170260794e-05, "val_loss": 4.624613285064697}
{"train_loss": 0.31505072116851807, "global_step": 103240, "epoch": 1160, "lr": 6.852475322453287e-05}
{"train_loss": 0.15276309847831726, "global_step": 103241, "epoch": 1160, "lr": 6.852421474396738e-05}
{"train_loss": 0.12506133317947388, "global_step": 103242, "epoch": 1160, "lr": 6.852367626091154e-05}
{"train_loss": 0.1886136531829834, "global_step": 103243, "epoch": 1160, "lr": 6.852313777536542e-05}
{"train_loss": 0.16687098145484924, "global_step": 103244, "epoch": 1160, "lr": 6.852259928732908e-05}
{"train_loss": 0.1843062937259674, "global_step": 103245, "epoch": 1160, "lr": 6.852206079680262e-05}
{"train_loss": 0.12508520483970642, "global_step": 103246, "epoch": 1160, "lr": 6.852152230378607e-05}
{"train_loss": 0.11755293607711792, "global_step": 103247, "epoch": 1160, "lr": 6.852098380827954e-05}
{"train_loss": 0.1461063027381897, "global_step": 103248, "epoch": 1160, "lr": 6.852044531028309e-05}
{"train_loss": 0.17733116447925568, "global_step": 103249, "epoch": 1160, "lr": 6.851990680979678e-05}
{"train_loss": 0.17187273502349854, "global_step": 103250, "epoch": 1160, "lr": 6.851936830682071e-05}
{"train_loss": 0.16190175712108612, "global_step": 103251, "epoch": 1160, "lr": 6.851882980135493e-05}
{"train_loss": 0.1588626205921173, "global_step": 103252, "epoch": 1160, "lr": 6.851829129339953e-05}
{"train_loss": 0.22646492719650269, "global_step": 103253, "epoch": 1160, "lr": 6.851775278295456e-05}
{"train_loss": 0.1603817492723465, "global_step": 103254, "epoch": 1160, "lr": 6.85172142700201e-05}
{"train_loss": 0.16077718138694763, "global_step": 103255, "epoch": 1160, "lr": 6.851667575459623e-05}
{"train_loss": 0.12941530346870422, "global_step": 103256, "epoch": 1160, "lr": 6.851613723668302e-05}
{"train_loss": 0.15647053718566895, "global_step": 103257, "epoch": 1160, "lr": 6.851559871628055e-05}
{"train_loss": 0.1490715891122818, "global_step": 103258, "epoch": 1160, "lr": 6.851506019338886e-05}
{"train_loss": 0.11153848469257355, "global_step": 103259, "epoch": 1160, "lr": 6.851452166800807e-05}
{"train_loss": 0.19046291708946228, "global_step": 103260, "epoch": 1160, "lr": 6.851398314013821e-05}
{"train_loss": 0.1292644888162613, "global_step": 103261, "epoch": 1160, "lr": 6.851344460977939e-05}
{"train_loss": 0.13330480456352234, "global_step": 103262, "epoch": 1160, "lr": 6.851290607693165e-05}
{"train_loss": 0.1976337432861328, "global_step": 103263, "epoch": 1160, "lr": 6.851236754159508e-05}
{"train_loss": 0.21165955066680908, "global_step": 103264, "epoch": 1160, "lr": 6.851182900376975e-05}
{"train_loss": 0.12020412087440491, "global_step": 103265, "epoch": 1160, "lr": 6.851129046345572e-05}
{"train_loss": 0.15532304346561432, "global_step": 103266, "epoch": 1160, "lr": 6.851075192065308e-05}
{"train_loss": 0.17035020887851715, "global_step": 103267, "epoch": 1160, "lr": 6.85102133753619e-05}
{"train_loss": 0.17692308127880096, "global_step": 103268, "epoch": 1160, "lr": 6.850967482758223e-05}
{"train_loss": 0.13039764761924744, "global_step": 103269, "epoch": 1160, "lr": 6.850913627731419e-05}
{"train_loss": 0.152422234416008, "global_step": 103270, "epoch": 1160, "lr": 6.850859772455779e-05}
{"train_loss": 0.13863883912563324, "global_step": 103271, "epoch": 1160, "lr": 6.850805916931314e-05}
{"train_loss": 0.15963728725910187, "global_step": 103272, "epoch": 1160, "lr": 6.850752061158032e-05}
{"train_loss": 0.2364668846130371, "global_step": 103273, "epoch": 1160, "lr": 6.850698205135939e-05}
{"train_loss": 0.1337401568889618, "global_step": 103274, "epoch": 1160, "lr": 6.850644348865041e-05}
{"train_loss": 0.15608027577400208, "global_step": 103275, "epoch": 1160, "lr": 6.850590492345347e-05}
{"train_loss": 0.14542415738105774, "global_step": 103276, "epoch": 1160, "lr": 6.850536635576865e-05}
{"train_loss": 0.17378194630146027, "global_step": 103277, "epoch": 1160, "lr": 6.850482778559598e-05}
{"train_loss": 0.1533556878566742, "global_step": 103278, "epoch": 1160, "lr": 6.850428921293558e-05}
{"train_loss": 0.13887174427509308, "global_step": 103279, "epoch": 1160, "lr": 6.85037506377875e-05}
{"train_loss": 0.1330883651971817, "global_step": 103280, "epoch": 1160, "lr": 6.850321206015182e-05}
{"train_loss": 0.17407351732254028, "global_step": 103281, "epoch": 1160, "lr": 6.850267348002861e-05}
{"train_loss": 0.13829755783081055, "global_step": 103282, "epoch": 1160, "lr": 6.850213489741793e-05}
{"train_loss": 0.1403246521949768, "global_step": 103283, "epoch": 1160, "lr": 6.850159631231988e-05}
{"train_loss": 0.09974715858697891, "global_step": 103284, "epoch": 1160, "lr": 6.85010577247345e-05}
{"train_loss": 0.20368580520153046, "global_step": 103285, "epoch": 1160, "lr": 6.850051913466189e-05}
{"train_loss": 0.1694602519273758, "global_step": 103286, "epoch": 1160, "lr": 6.84999805421021e-05}
{"train_loss": 0.22097763419151306, "global_step": 103287, "epoch": 1160, "lr": 6.849944194705521e-05}
{"train_loss": 0.11746243387460709, "global_step": 103288, "epoch": 1160, "lr": 6.849890334952132e-05}
{"train_loss": 0.18147225677967072, "global_step": 103289, "epoch": 1160, "lr": 6.849836474950045e-05}
{"train_loss": 0.1492854356765747, "global_step": 103290, "epoch": 1160, "lr": 6.849782614699271e-05}
{"train_loss": 0.12600113451480865, "global_step": 103291, "epoch": 1160, "lr": 6.849728754199818e-05}
{"train_loss": 0.1975494921207428, "global_step": 103292, "epoch": 1160, "lr": 6.849674893451689e-05}
{"train_loss": 0.18276606500148773, "global_step": 103293, "epoch": 1160, "lr": 6.849621032454894e-05}
{"train_loss": 0.16859349608421326, "global_step": 103294, "epoch": 1160, "lr": 6.849567171209442e-05}
{"train_loss": 0.1384887844324112, "global_step": 103295, "epoch": 1160, "lr": 6.849513309715336e-05}
{"train_loss": 0.09863495081663132, "global_step": 103296, "epoch": 1160, "lr": 6.849459447972587e-05}
{"train_loss": 0.10373478382825851, "global_step": 103297, "epoch": 1160, "lr": 6.849405585981201e-05}
{"train_loss": 0.14272253215312958, "global_step": 103298, "epoch": 1160, "lr": 6.849351723741185e-05}
{"train_loss": 0.13554467260837555, "global_step": 103299, "epoch": 1160, "lr": 6.849297861252544e-05}
{"train_loss": 0.16039368510246277, "global_step": 103300, "epoch": 1160, "lr": 6.849243998515289e-05}
{"train_loss": 0.12252071499824524, "global_step": 103301, "epoch": 1160, "lr": 6.849190135529425e-05}
{"train_loss": 0.11940520256757736, "global_step": 103302, "epoch": 1160, "lr": 6.849136272294961e-05}
{"train_loss": 0.1187252625823021, "global_step": 103303, "epoch": 1160, "lr": 6.849082408811902e-05}
{"train_loss": 0.14680521190166473, "global_step": 103304, "epoch": 1160, "lr": 6.849028545080257e-05}
{"train_loss": 0.1291407346725464, "global_step": 103305, "epoch": 1160, "lr": 6.848974681100033e-05}
{"train_loss": 0.17443585395812988, "global_step": 103306, "epoch": 1160, "lr": 6.848920816871236e-05}
{"train_loss": 0.12015020102262497, "global_step": 103307, "epoch": 1160, "lr": 6.848866952393876e-05}
{"train_loss": 0.10830355435609818, "global_step": 103308, "epoch": 1160, "lr": 6.848813087667956e-05}
{"train_loss": 0.18313683569431305, "global_step": 103309, "epoch": 1160, "lr": 6.848759222693487e-05}
{"train_loss": 0.140219047665596, "global_step": 103310, "epoch": 1160, "lr": 6.848705357470475e-05}
{"train_loss": 0.17802797257900238, "global_step": 103311, "epoch": 1160, "lr": 6.848651491998926e-05}
{"train_loss": 0.17801019549369812, "global_step": 103312, "epoch": 1160, "lr": 6.848597626278848e-05}
{"train_loss": 0.22521242499351501, "global_step": 103313, "epoch": 1160, "lr": 6.848543760310251e-05}
{"train_loss": 0.14724260568618774, "global_step": 103314, "epoch": 1160, "lr": 6.848489894093138e-05}
{"train_loss": 0.11581602692604065, "global_step": 103315, "epoch": 1160, "lr": 6.848436027627518e-05}
{"train_loss": 0.1340811848640442, "global_step": 103316, "epoch": 1160, "lr": 6.848382160913399e-05}
{"train_loss": 0.11084175109863281, "global_step": 103317, "epoch": 1160, "lr": 6.848328293950787e-05}
{"train_loss": 0.12905412912368774, "global_step": 103318, "epoch": 1160, "lr": 6.84827442673969e-05}
{"train_loss": 0.17715182900428772, "global_step": 103319, "epoch": 1160, "lr": 6.848220559280115e-05}
{"train_loss": 0.1374640166759491, "global_step": 103320, "epoch": 1160, "lr": 6.84816669157207e-05}
{"train_loss": 0.1399107575416565, "global_step": 103321, "epoch": 1160, "lr": 6.84811282361556e-05}
{"train_loss": 0.15960074961185455, "global_step": 103322, "epoch": 1160, "lr": 6.848058955410595e-05}
{"train_loss": 0.10290437191724777, "global_step": 103323, "epoch": 1160, "lr": 6.848005086957181e-05}
{"train_loss": 0.1334059238433838, "global_step": 103324, "epoch": 1160, "lr": 6.847951218255324e-05}
{"train_loss": 0.11376284807920456, "global_step": 103325, "epoch": 1160, "lr": 6.847897349305035e-05}
{"train_loss": 0.1581019014120102, "global_step": 103326, "epoch": 1160, "lr": 6.847843480106316e-05}
{"train_loss": 0.12302715331315994, "global_step": 103327, "epoch": 1160, "lr": 6.847789610659179e-05}
{"train_loss": 0.15307997811711235, "global_step": 103328, "epoch": 1160, "lr": 6.847735740963627e-05, "val_loss": 4.723336219787598, "train_action_mse_error": 12.889518737792969}
{"train_loss": 0.1711859107017517, "global_step": 103329, "epoch": 1161, "lr": 6.847681871019671e-05}
{"train_loss": 0.08566585183143616, "global_step": 103330, "epoch": 1161, "lr": 6.847628000827316e-05}
{"train_loss": 0.15430432558059692, "global_step": 103331, "epoch": 1161, "lr": 6.847574130386572e-05}
{"train_loss": 0.14284227788448334, "global_step": 103332, "epoch": 1161, "lr": 6.847520259697442e-05}
{"train_loss": 0.19464315474033356, "global_step": 103333, "epoch": 1161, "lr": 6.847466388759935e-05}
{"train_loss": 0.10693352669477463, "global_step": 103334, "epoch": 1161, "lr": 6.84741251757406e-05}
{"train_loss": 0.226418599486351, "global_step": 103335, "epoch": 1161, "lr": 6.847358646139822e-05}
{"train_loss": 0.16359931230545044, "global_step": 103336, "epoch": 1161, "lr": 6.84730477445723e-05}
{"train_loss": 0.12406682968139648, "global_step": 103337, "epoch": 1161, "lr": 6.84725090252629e-05}
{"train_loss": 0.15568900108337402, "global_step": 103338, "epoch": 1161, "lr": 6.84719703034701e-05}
{"train_loss": 0.12591734528541565, "global_step": 103339, "epoch": 1161, "lr": 6.847143157919396e-05}
{"train_loss": 0.15777790546417236, "global_step": 103340, "epoch": 1161, "lr": 6.847089285243456e-05}
{"train_loss": 0.11421849578619003, "global_step": 103341, "epoch": 1161, "lr": 6.847035412319198e-05}
{"train_loss": 0.14437450468540192, "global_step": 103342, "epoch": 1161, "lr": 6.84698153914663e-05}
{"train_loss": 0.08690119534730911, "global_step": 103343, "epoch": 1161, "lr": 6.846927665725755e-05}
{"train_loss": 0.13035260140895844, "global_step": 103344, "epoch": 1161, "lr": 6.846873792056585e-05}
{"train_loss": 0.13220137357711792, "global_step": 103345, "epoch": 1161, "lr": 6.846819918139125e-05}
{"train_loss": 0.19108901917934418, "global_step": 103346, "epoch": 1161, "lr": 6.846766043973383e-05}
{"train_loss": 0.17920060455799103, "global_step": 103347, "epoch": 1161, "lr": 6.846712169559365e-05}
{"train_loss": 0.12818728387355804, "global_step": 103348, "epoch": 1161, "lr": 6.84665829489708e-05}
{"train_loss": 0.12407851219177246, "global_step": 103349, "epoch": 1161, "lr": 6.846604419986533e-05}
{"train_loss": 0.1333463042974472, "global_step": 103350, "epoch": 1161, "lr": 6.846550544827732e-05}
{"train_loss": 0.1535465121269226, "global_step": 103351, "epoch": 1161, "lr": 6.846496669420686e-05}
{"train_loss": 0.17201974987983704, "global_step": 103352, "epoch": 1161, "lr": 6.846442793765402e-05}
{"train_loss": 0.1688774973154068, "global_step": 103353, "epoch": 1161, "lr": 6.846388917861885e-05}
{"train_loss": 0.1197488084435463, "global_step": 103354, "epoch": 1161, "lr": 6.846335041710143e-05}
{"train_loss": 0.10596860945224762, "global_step": 103355, "epoch": 1161, "lr": 6.846281165310185e-05}
{"train_loss": 0.11662428826093674, "global_step": 103356, "epoch": 1161, "lr": 6.846227288662017e-05}
{"train_loss": 0.14980654418468475, "global_step": 103357, "epoch": 1161, "lr": 6.846173411765646e-05}
{"train_loss": 0.14665985107421875, "global_step": 103358, "epoch": 1161, "lr": 6.846119534621079e-05}
{"train_loss": 0.16141587495803833, "global_step": 103359, "epoch": 1161, "lr": 6.846065657228323e-05}
{"train_loss": 0.15428899228572845, "global_step": 103360, "epoch": 1161, "lr": 6.846011779587388e-05}
{"train_loss": 0.15366248786449432, "global_step": 103361, "epoch": 1161, "lr": 6.845957901698277e-05}
{"train_loss": 0.1488959938287735, "global_step": 103362, "epoch": 1161, "lr": 6.845904023561e-05}
{"train_loss": 0.19991543889045715, "global_step": 103363, "epoch": 1161, "lr": 6.845850145175565e-05}
{"train_loss": 0.12828782200813293, "global_step": 103364, "epoch": 1161, "lr": 6.845796266541978e-05}
{"train_loss": 0.117625392973423, "global_step": 103365, "epoch": 1161, "lr": 6.845742387660245e-05}
{"train_loss": 0.1609683781862259, "global_step": 103366, "epoch": 1161, "lr": 6.845688508530375e-05}
{"train_loss": 0.17296865582466125, "global_step": 103367, "epoch": 1161, "lr": 6.845634629152375e-05}
{"train_loss": 0.12051926553249359, "global_step": 103368, "epoch": 1161, "lr": 6.845580749526252e-05}
{"train_loss": 0.18537047505378723, "global_step": 103369, "epoch": 1161, "lr": 6.845526869652012e-05}
{"train_loss": 0.28718018531799316, "global_step": 103370, "epoch": 1161, "lr": 6.845472989529665e-05}
{"train_loss": 0.18509270250797272, "global_step": 103371, "epoch": 1161, "lr": 6.845419109159214e-05}
{"train_loss": 0.10210542380809784, "global_step": 103372, "epoch": 1161, "lr": 6.845365228540671e-05}
{"train_loss": 0.09766872972249985, "global_step": 103373, "epoch": 1161, "lr": 6.84531134767404e-05}
{"train_loss": 0.25459539890289307, "global_step": 103374, "epoch": 1161, "lr": 6.845257466559331e-05}
{"train_loss": 0.17516423761844635, "global_step": 103375, "epoch": 1161, "lr": 6.84520358519655e-05}
{"train_loss": 0.16482128202915192, "global_step": 103376, "epoch": 1161, "lr": 6.845149703585701e-05}
{"train_loss": 0.11359499394893646, "global_step": 103377, "epoch": 1161, "lr": 6.845095821726795e-05}
{"train_loss": 0.14745259284973145, "global_step": 103378, "epoch": 1161, "lr": 6.84504193961984e-05}
{"train_loss": 0.21763747930526733, "global_step": 103379, "epoch": 1161, "lr": 6.84498805726484e-05}
{"train_loss": 0.22626113891601562, "global_step": 103380, "epoch": 1161, "lr": 6.844934174661803e-05}
{"train_loss": 0.1795060783624649, "global_step": 103381, "epoch": 1161, "lr": 6.84488029181074e-05}
{"train_loss": 0.16118457913398743, "global_step": 103382, "epoch": 1161, "lr": 6.844826408711655e-05}
{"train_loss": 0.1483672708272934, "global_step": 103383, "epoch": 1161, "lr": 6.844772525364554e-05}
{"train_loss": 0.1531798541545868, "global_step": 103384, "epoch": 1161, "lr": 6.844718641769446e-05}
{"train_loss": 0.11103601008653641, "global_step": 103385, "epoch": 1161, "lr": 6.84466475792634e-05}
{"train_loss": 0.15953323245048523, "global_step": 103386, "epoch": 1161, "lr": 6.844610873835239e-05}
{"train_loss": 0.23788943886756897, "global_step": 103387, "epoch": 1161, "lr": 6.844556989496153e-05}
{"train_loss": 0.1518886536359787, "global_step": 103388, "epoch": 1161, "lr": 6.84450310490909e-05}
{"train_loss": 0.13145163655281067, "global_step": 103389, "epoch": 1161, "lr": 6.844449220074054e-05}
{"train_loss": 0.1726045310497284, "global_step": 103390, "epoch": 1161, "lr": 6.844395334991055e-05}
{"train_loss": 0.18193989992141724, "global_step": 103391, "epoch": 1161, "lr": 6.8443414496601e-05}
{"train_loss": 0.17862828075885773, "global_step": 103392, "epoch": 1161, "lr": 6.844287564081195e-05}
{"train_loss": 0.25497308373451233, "global_step": 103393, "epoch": 1161, "lr": 6.84423367825435e-05}
{"train_loss": 0.1325518637895584, "global_step": 103394, "epoch": 1161, "lr": 6.844179792179568e-05}
{"train_loss": 0.17286382615566254, "global_step": 103395, "epoch": 1161, "lr": 6.84412590585686e-05}
{"train_loss": 0.1438184529542923, "global_step": 103396, "epoch": 1161, "lr": 6.844072019286231e-05}
{"train_loss": 0.131831556558609, "global_step": 103397, "epoch": 1161, "lr": 6.844018132467689e-05}
{"train_loss": 0.15356987714767456, "global_step": 103398, "epoch": 1161, "lr": 6.84396424540124e-05}
{"train_loss": 0.22617891430854797, "global_step": 103399, "epoch": 1161, "lr": 6.843910358086895e-05}
{"train_loss": 0.1481454074382782, "global_step": 103400, "epoch": 1161, "lr": 6.843856470524656e-05}
{"train_loss": 0.18010500073432922, "global_step": 103401, "epoch": 1161, "lr": 6.843802582714535e-05}
{"train_loss": 0.08947477489709854, "global_step": 103402, "epoch": 1161, "lr": 6.843748694656538e-05}
{"train_loss": 0.19017019867897034, "global_step": 103403, "epoch": 1161, "lr": 6.843694806350671e-05}
{"train_loss": 0.1610962152481079, "global_step": 103404, "epoch": 1161, "lr": 6.843640917796941e-05}
{"train_loss": 0.1991863250732422, "global_step": 103405, "epoch": 1161, "lr": 6.843587028995355e-05}
{"train_loss": 0.1932995617389679, "global_step": 103406, "epoch": 1161, "lr": 6.843533139945923e-05}
{"train_loss": 0.19282278418540955, "global_step": 103407, "epoch": 1161, "lr": 6.843479250648649e-05}
{"train_loss": 0.19647175073623657, "global_step": 103408, "epoch": 1161, "lr": 6.843425361103543e-05}
{"train_loss": 0.15410169959068298, "global_step": 103409, "epoch": 1161, "lr": 6.84337147131061e-05}
{"train_loss": 0.19944006204605103, "global_step": 103410, "epoch": 1161, "lr": 6.843317581269857e-05}
{"train_loss": 0.06710981577634811, "global_step": 103411, "epoch": 1161, "lr": 6.843263690981295e-05}
{"train_loss": 0.18283997476100922, "global_step": 103412, "epoch": 1161, "lr": 6.843209800444928e-05}
{"train_loss": 0.08903060853481293, "global_step": 103413, "epoch": 1161, "lr": 6.843155909660762e-05}
{"train_loss": 0.17721053957939148, "global_step": 103414, "epoch": 1161, "lr": 6.843102018628808e-05}
{"train_loss": 0.11111600697040558, "global_step": 103415, "epoch": 1161, "lr": 6.843048127349071e-05}
{"train_loss": 0.21134741604328156, "global_step": 103416, "epoch": 1161, "lr": 6.842994235821558e-05}
{"train_loss": 0.15758047227779132, "global_step": 103417, "epoch": 1161, "lr": 6.842940344046276e-05, "val_loss": 4.868204593658447}
{"train_loss": 0.15715007483959198, "global_step": 103418, "epoch": 1162, "lr": 6.842886452023236e-05}
{"train_loss": 0.21477365493774414, "global_step": 103419, "epoch": 1162, "lr": 6.84283255975244e-05}
{"train_loss": 0.12388765811920166, "global_step": 103420, "epoch": 1162, "lr": 6.842778667233899e-05}
{"train_loss": 0.12081709504127502, "global_step": 103421, "epoch": 1162, "lr": 6.842724774467619e-05}
{"train_loss": 0.14437437057495117, "global_step": 103422, "epoch": 1162, "lr": 6.842670881453606e-05}
{"train_loss": 0.11926943063735962, "global_step": 103423, "epoch": 1162, "lr": 6.84261698819187e-05}
{"train_loss": 0.16386348009109497, "global_step": 103424, "epoch": 1162, "lr": 6.842563094682415e-05}
{"train_loss": 0.19264301657676697, "global_step": 103425, "epoch": 1162, "lr": 6.84250920092525e-05}
{"train_loss": 0.1758546084165573, "global_step": 103426, "epoch": 1162, "lr": 6.842455306920382e-05}
{"train_loss": 0.2626205086708069, "global_step": 103427, "epoch": 1162, "lr": 6.842401412667819e-05}
{"train_loss": 0.15989477932453156, "global_step": 103428, "epoch": 1162, "lr": 6.842347518167568e-05}
{"train_loss": 0.2261824607849121, "global_step": 103429, "epoch": 1162, "lr": 6.842293623419636e-05}
{"train_loss": 0.11907882243394852, "global_step": 103430, "epoch": 1162, "lr": 6.842239728424029e-05}
{"train_loss": 0.20298221707344055, "global_step": 103431, "epoch": 1162, "lr": 6.842185833180754e-05}
{"train_loss": 0.25339820981025696, "global_step": 103432, "epoch": 1162, "lr": 6.842131937689822e-05}
{"train_loss": 0.10916782170534134, "global_step": 103433, "epoch": 1162, "lr": 6.842078041951237e-05}
{"train_loss": 0.157434344291687, "global_step": 103434, "epoch": 1162, "lr": 6.842024145965007e-05}
{"train_loss": 0.1289752870798111, "global_step": 103435, "epoch": 1162, "lr": 6.841970249731138e-05}
{"train_loss": 0.1367611289024353, "global_step": 103436, "epoch": 1162, "lr": 6.84191635324964e-05}
{"train_loss": 0.16125109791755676, "global_step": 103437, "epoch": 1162, "lr": 6.841862456520519e-05}
{"train_loss": 0.13858789205551147, "global_step": 103438, "epoch": 1162, "lr": 6.841808559543781e-05}
{"train_loss": 0.20336271822452545, "global_step": 103439, "epoch": 1162, "lr": 6.841754662319435e-05}
{"train_loss": 0.10963898152112961, "global_step": 103440, "epoch": 1162, "lr": 6.841700764847489e-05}
{"train_loss": 0.1340530663728714, "global_step": 103441, "epoch": 1162, "lr": 6.841646867127948e-05}
{"train_loss": 0.1462169885635376, "global_step": 103442, "epoch": 1162, "lr": 6.841592969160819e-05}
{"train_loss": 0.12012969702482224, "global_step": 103443, "epoch": 1162, "lr": 6.84153907094611e-05}
{"train_loss": 0.15836147964000702, "global_step": 103444, "epoch": 1162, "lr": 6.84148517248383e-05}
{"train_loss": 0.13929393887519836, "global_step": 103445, "epoch": 1162, "lr": 6.841431273773984e-05}
{"train_loss": 0.19659484922885895, "global_step": 103446, "epoch": 1162, "lr": 6.84137737481658e-05}
{"train_loss": 0.142297625541687, "global_step": 103447, "epoch": 1162, "lr": 6.841323475611625e-05}
{"train_loss": 0.14371803402900696, "global_step": 103448, "epoch": 1162, "lr": 6.841269576159127e-05}
{"train_loss": 0.12702809274196625, "global_step": 103449, "epoch": 1162, "lr": 6.841215676459091e-05}
{"train_loss": 0.16989818215370178, "global_step": 103450, "epoch": 1162, "lr": 6.841161776511527e-05}
{"train_loss": 0.1728818118572235, "global_step": 103451, "epoch": 1162, "lr": 6.841107876316443e-05}
{"train_loss": 0.12563879787921906, "global_step": 103452, "epoch": 1162, "lr": 6.841053975873843e-05}
{"train_loss": 0.18194691836833954, "global_step": 103453, "epoch": 1162, "lr": 6.841000075183734e-05}
{"train_loss": 0.14412203431129456, "global_step": 103454, "epoch": 1162, "lr": 6.840946174246128e-05}
{"train_loss": 0.14024129509925842, "global_step": 103455, "epoch": 1162, "lr": 6.840892273061028e-05}
{"train_loss": 0.15091267228126526, "global_step": 103456, "epoch": 1162, "lr": 6.840838371628441e-05}
{"train_loss": 0.17058104276657104, "global_step": 103457, "epoch": 1162, "lr": 6.840784469948378e-05}
{"train_loss": 0.13531611859798431, "global_step": 103458, "epoch": 1162, "lr": 6.840730568020844e-05}
{"train_loss": 0.14358638226985931, "global_step": 103459, "epoch": 1162, "lr": 6.840676665845845e-05}
{"train_loss": 0.1441534012556076, "global_step": 103460, "epoch": 1162, "lr": 6.840622763423391e-05}
{"train_loss": 0.17446734011173248, "global_step": 103461, "epoch": 1162, "lr": 6.840568860753486e-05}
{"train_loss": 0.15017037093639374, "global_step": 103462, "epoch": 1162, "lr": 6.84051495783614e-05}
{"train_loss": 0.13217192888259888, "global_step": 103463, "epoch": 1162, "lr": 6.840461054671359e-05}
{"train_loss": 0.11241380125284195, "global_step": 103464, "epoch": 1162, "lr": 6.84040715125915e-05}
{"train_loss": 0.16742268204689026, "global_step": 103465, "epoch": 1162, "lr": 6.84035324759952e-05}
{"train_loss": 0.12844207882881165, "global_step": 103466, "epoch": 1162, "lr": 6.840299343692478e-05}
{"train_loss": 0.09359880536794662, "global_step": 103467, "epoch": 1162, "lr": 6.84024543953803e-05}
{"train_loss": 0.17576546967029572, "global_step": 103468, "epoch": 1162, "lr": 6.840191535136184e-05}
{"train_loss": 0.11689183861017227, "global_step": 103469, "epoch": 1162, "lr": 6.840137630486945e-05}
{"train_loss": 0.14035147428512573, "global_step": 103470, "epoch": 1162, "lr": 6.840083725590323e-05}
{"train_loss": 0.1566479504108429, "global_step": 103471, "epoch": 1162, "lr": 6.840029820446325e-05}
{"train_loss": 0.19707639515399933, "global_step": 103472, "epoch": 1162, "lr": 6.839975915054956e-05}
{"train_loss": 0.09897790849208832, "global_step": 103473, "epoch": 1162, "lr": 6.839922009416225e-05}
{"train_loss": 0.11084520816802979, "global_step": 103474, "epoch": 1162, "lr": 6.839868103530139e-05}
{"train_loss": 0.21250493824481964, "global_step": 103475, "epoch": 1162, "lr": 6.839814197396705e-05}
{"train_loss": 0.27571216225624084, "global_step": 103476, "epoch": 1162, "lr": 6.839760291015931e-05}
{"train_loss": 0.12073148041963577, "global_step": 103477, "epoch": 1162, "lr": 6.839706384387824e-05}
{"train_loss": 0.24614958465099335, "global_step": 103478, "epoch": 1162, "lr": 6.839652477512391e-05}
{"train_loss": 0.17334690690040588, "global_step": 103479, "epoch": 1162, "lr": 6.839598570389638e-05}
{"train_loss": 0.21422629058361053, "global_step": 103480, "epoch": 1162, "lr": 6.839544663019574e-05}
{"train_loss": 0.21976128220558167, "global_step": 103481, "epoch": 1162, "lr": 6.839490755402206e-05}
{"train_loss": 0.12313847243785858, "global_step": 103482, "epoch": 1162, "lr": 6.83943684753754e-05}
{"train_loss": 0.3116551339626312, "global_step": 103483, "epoch": 1162, "lr": 6.839382939425585e-05}
{"train_loss": 0.11445993185043335, "global_step": 103484, "epoch": 1162, "lr": 6.839329031066346e-05}
{"train_loss": 0.17261752486228943, "global_step": 103485, "epoch": 1162, "lr": 6.839275122459833e-05}
{"train_loss": 0.14097920060157776, "global_step": 103486, "epoch": 1162, "lr": 6.839221213606051e-05}
{"train_loss": 0.10060295462608337, "global_step": 103487, "epoch": 1162, "lr": 6.839167304505009e-05}
{"train_loss": 0.20886710286140442, "global_step": 103488, "epoch": 1162, "lr": 6.839113395156713e-05}
{"train_loss": 0.0901760533452034, "global_step": 103489, "epoch": 1162, "lr": 6.83905948556117e-05}
{"train_loss": 0.12786372005939484, "global_step": 103490, "epoch": 1162, "lr": 6.839005575718388e-05}
{"train_loss": 0.15640738606452942, "global_step": 103491, "epoch": 1162, "lr": 6.838951665628375e-05}
{"train_loss": 0.16440333425998688, "global_step": 103492, "epoch": 1162, "lr": 6.838897755291136e-05}
{"train_loss": 0.15918397903442383, "global_step": 103493, "epoch": 1162, "lr": 6.83884384470668e-05}
{"train_loss": 0.11290239542722702, "global_step": 103494, "epoch": 1162, "lr": 6.838789933875015e-05}
{"train_loss": 0.16691604256629944, "global_step": 103495, "epoch": 1162, "lr": 6.838736022796145e-05}
{"train_loss": 0.1705140322446823, "global_step": 103496, "epoch": 1162, "lr": 6.838682111470082e-05}
{"train_loss": 0.20707276463508606, "global_step": 103497, "epoch": 1162, "lr": 6.838628199896829e-05}
{"train_loss": 0.09733089059591293, "global_step": 103498, "epoch": 1162, "lr": 6.838574288076397e-05}
{"train_loss": 0.2022401988506317, "global_step": 103499, "epoch": 1162, "lr": 6.83852037600879e-05}
{"train_loss": 0.15442873537540436, "global_step": 103500, "epoch": 1162, "lr": 6.838466463694014e-05}
{"train_loss": 0.16782121360301971, "global_step": 103501, "epoch": 1162, "lr": 6.83841255113208e-05}
{"train_loss": 0.14991611242294312, "global_step": 103502, "epoch": 1162, "lr": 6.838358638322995e-05}
{"train_loss": 0.11934329569339752, "global_step": 103503, "epoch": 1162, "lr": 6.838304725266764e-05}
{"train_loss": 0.1591872125864029, "global_step": 103504, "epoch": 1162, "lr": 6.838250811963396e-05}
{"train_loss": 0.18431618809700012, "global_step": 103505, "epoch": 1162, "lr": 6.838196898412896e-05}
{"train_loss": 0.15872785976428663, "global_step": 103506, "epoch": 1162, "lr": 6.838142984615274e-05, "val_loss": 4.730828285217285}
{"train_loss": 0.23906515538692474, "global_step": 103507, "epoch": 1163, "lr": 6.838089070570536e-05}
{"train_loss": 0.17003872990608215, "global_step": 103508, "epoch": 1163, "lr": 6.838035156278689e-05}
{"train_loss": 0.11952892690896988, "global_step": 103509, "epoch": 1163, "lr": 6.837981241739741e-05}
{"train_loss": 0.08639612793922424, "global_step": 103510, "epoch": 1163, "lr": 6.8379273269537e-05}
{"train_loss": 0.1480676829814911, "global_step": 103511, "epoch": 1163, "lr": 6.837873411920571e-05}
{"train_loss": 0.12286461144685745, "global_step": 103512, "epoch": 1163, "lr": 6.83781949664036e-05}
{"train_loss": 0.11672917753458023, "global_step": 103513, "epoch": 1163, "lr": 6.83776558111308e-05}
{"train_loss": 0.19009223580360413, "global_step": 103514, "epoch": 1163, "lr": 6.837711665338733e-05}
{"train_loss": 0.09535682946443558, "global_step": 103515, "epoch": 1163, "lr": 6.837657749317329e-05}
{"train_loss": 0.1442185491323471, "global_step": 103516, "epoch": 1163, "lr": 6.837603833048874e-05}
{"train_loss": 0.19438603520393372, "global_step": 103517, "epoch": 1163, "lr": 6.837549916533375e-05}
{"train_loss": 0.13983310759067535, "global_step": 103518, "epoch": 1163, "lr": 6.837495999770841e-05}
{"train_loss": 0.1370008885860443, "global_step": 103519, "epoch": 1163, "lr": 6.837442082761277e-05}
{"train_loss": 0.13669665157794952, "global_step": 103520, "epoch": 1163, "lr": 6.837388165504692e-05}
{"train_loss": 0.1564718633890152, "global_step": 103521, "epoch": 1163, "lr": 6.837334248001092e-05}
{"train_loss": 0.13997605443000793, "global_step": 103522, "epoch": 1163, "lr": 6.837280330250485e-05}
{"train_loss": 0.13292363286018372, "global_step": 103523, "epoch": 1163, "lr": 6.837226412252878e-05}
{"train_loss": 0.13623736798763275, "global_step": 103524, "epoch": 1163, "lr": 6.837172494008279e-05}
{"train_loss": 0.1691867560148239, "global_step": 103525, "epoch": 1163, "lr": 6.837118575516692e-05}
{"train_loss": 0.20006635785102844, "global_step": 103526, "epoch": 1163, "lr": 6.837064656778129e-05}
{"train_loss": 0.12452330440282822, "global_step": 103527, "epoch": 1163, "lr": 6.837010737792595e-05}
{"train_loss": 0.08308330923318863, "global_step": 103528, "epoch": 1163, "lr": 6.836956818560097e-05}
{"train_loss": 0.1310851275920868, "global_step": 103529, "epoch": 1163, "lr": 6.836902899080643e-05}
{"train_loss": 0.16895943880081177, "global_step": 103530, "epoch": 1163, "lr": 6.836848979354239e-05}
{"train_loss": 0.11095910519361496, "global_step": 103531, "epoch": 1163, "lr": 6.836795059380894e-05}
{"train_loss": 0.15318766236305237, "global_step": 103532, "epoch": 1163, "lr": 6.836741139160613e-05}
{"train_loss": 0.1345915049314499, "global_step": 103533, "epoch": 1163, "lr": 6.836687218693405e-05}
{"train_loss": 0.17730756103992462, "global_step": 103534, "epoch": 1163, "lr": 6.836633297979278e-05}
{"train_loss": 0.09074230492115021, "global_step": 103535, "epoch": 1163, "lr": 6.836579377018237e-05}
{"train_loss": 0.08301112055778503, "global_step": 103536, "epoch": 1163, "lr": 6.83652545581029e-05}
{"train_loss": 0.20300976932048798, "global_step": 103537, "epoch": 1163, "lr": 6.836471534355446e-05}
{"train_loss": 0.2131337672472, "global_step": 103538, "epoch": 1163, "lr": 6.83641761265371e-05}
{"train_loss": 0.22324198484420776, "global_step": 103539, "epoch": 1163, "lr": 6.83636369070509e-05}
{"train_loss": 0.14917327463626862, "global_step": 103540, "epoch": 1163, "lr": 6.836309768509593e-05}
{"train_loss": 0.2267346978187561, "global_step": 103541, "epoch": 1163, "lr": 6.836255846067226e-05}
{"train_loss": 0.11064973473548889, "global_step": 103542, "epoch": 1163, "lr": 6.836201923377997e-05}
{"train_loss": 0.15785588324069977, "global_step": 103543, "epoch": 1163, "lr": 6.836148000441914e-05}
{"train_loss": 0.1755865514278412, "global_step": 103544, "epoch": 1163, "lr": 6.836094077258981e-05}
{"train_loss": 0.09393123537302017, "global_step": 103545, "epoch": 1163, "lr": 6.83604015382921e-05}
{"train_loss": 0.2041386216878891, "global_step": 103546, "epoch": 1163, "lr": 6.835986230152604e-05}
{"train_loss": 0.1561933010816574, "global_step": 103547, "epoch": 1163, "lr": 6.835932306229174e-05}
{"train_loss": 0.2574622929096222, "global_step": 103548, "epoch": 1163, "lr": 6.835878382058923e-05}
{"train_loss": 0.24546566605567932, "global_step": 103549, "epoch": 1163, "lr": 6.835824457641862e-05}
{"train_loss": 0.14616183936595917, "global_step": 103550, "epoch": 1163, "lr": 6.835770532977997e-05}
{"train_loss": 0.15036077797412872, "global_step": 103551, "epoch": 1163, "lr": 6.835716608067335e-05}
{"train_loss": 0.18941622972488403, "global_step": 103552, "epoch": 1163, "lr": 6.835662682909882e-05}
{"train_loss": 0.17146573960781097, "global_step": 103553, "epoch": 1163, "lr": 6.835608757505647e-05}
{"train_loss": 0.17349064350128174, "global_step": 103554, "epoch": 1163, "lr": 6.835554831854638e-05}
{"train_loss": 0.14725424349308014, "global_step": 103555, "epoch": 1163, "lr": 6.83550090595686e-05}
{"train_loss": 0.22348828613758087, "global_step": 103556, "epoch": 1163, "lr": 6.835446979812323e-05}
{"train_loss": 0.19532035291194916, "global_step": 103557, "epoch": 1163, "lr": 6.835393053421032e-05}
{"train_loss": 0.2144472450017929, "global_step": 103558, "epoch": 1163, "lr": 6.835339126782993e-05}
{"train_loss": 0.23100516200065613, "global_step": 103559, "epoch": 1163, "lr": 6.835285199898215e-05}
{"train_loss": 0.13154955208301544, "global_step": 103560, "epoch": 1163, "lr": 6.835231272766707e-05}
{"train_loss": 0.16418811678886414, "global_step": 103561, "epoch": 1163, "lr": 6.835177345388473e-05}
{"train_loss": 0.10406667739152908, "global_step": 103562, "epoch": 1163, "lr": 6.835123417763522e-05}
{"train_loss": 0.18419860303401947, "global_step": 103563, "epoch": 1163, "lr": 6.835069489891861e-05}
{"train_loss": 0.09936877340078354, "global_step": 103564, "epoch": 1163, "lr": 6.835015561773497e-05}
{"train_loss": 0.16885313391685486, "global_step": 103565, "epoch": 1163, "lr": 6.834961633408439e-05}
{"train_loss": 0.12425389885902405, "global_step": 103566, "epoch": 1163, "lr": 6.834907704796692e-05}
{"train_loss": 0.15457850694656372, "global_step": 103567, "epoch": 1163, "lr": 6.834853775938264e-05}
{"train_loss": 0.15244993567466736, "global_step": 103568, "epoch": 1163, "lr": 6.834799846833162e-05}
{"train_loss": 0.14034360647201538, "global_step": 103569, "epoch": 1163, "lr": 6.834745917481394e-05}
{"train_loss": 0.12966017425060272, "global_step": 103570, "epoch": 1163, "lr": 6.834691987882967e-05}
{"train_loss": 0.13749288022518158, "global_step": 103571, "epoch": 1163, "lr": 6.834638058037887e-05}
{"train_loss": 0.20862029492855072, "global_step": 103572, "epoch": 1163, "lr": 6.834584127946162e-05}
{"train_loss": 0.0980457291007042, "global_step": 103573, "epoch": 1163, "lr": 6.8345301976078e-05}
{"train_loss": 0.09604565799236298, "global_step": 103574, "epoch": 1163, "lr": 6.834476267022809e-05}
{"train_loss": 0.21385027468204498, "global_step": 103575, "epoch": 1163, "lr": 6.834422336191194e-05}
{"train_loss": 0.17253553867340088, "global_step": 103576, "epoch": 1163, "lr": 6.834368405112964e-05}
{"train_loss": 0.13815812766551971, "global_step": 103577, "epoch": 1163, "lr": 6.834314473788126e-05}
{"train_loss": 0.21497981250286102, "global_step": 103578, "epoch": 1163, "lr": 6.834260542216684e-05}
{"train_loss": 0.15663617849349976, "global_step": 103579, "epoch": 1163, "lr": 6.834206610398652e-05}
{"train_loss": 0.16753607988357544, "global_step": 103580, "epoch": 1163, "lr": 6.83415267833403e-05}
{"train_loss": 0.10568723827600479, "global_step": 103581, "epoch": 1163, "lr": 6.834098746022829e-05}
{"train_loss": 0.1343349665403366, "global_step": 103582, "epoch": 1163, "lr": 6.834044813465056e-05}
{"train_loss": 0.11826620250940323, "global_step": 103583, "epoch": 1163, "lr": 6.833990880660718e-05}
{"train_loss": 0.163467139005661, "global_step": 103584, "epoch": 1163, "lr": 6.833936947609822e-05}
{"train_loss": 0.15617269277572632, "global_step": 103585, "epoch": 1163, "lr": 6.833883014312377e-05}
{"train_loss": 0.17012706398963928, "global_step": 103586, "epoch": 1163, "lr": 6.833829080768388e-05}
{"train_loss": 0.11278901994228363, "global_step": 103587, "epoch": 1163, "lr": 6.833775146977864e-05}
{"train_loss": 0.09411024302244186, "global_step": 103588, "epoch": 1163, "lr": 6.83372121294081e-05}
{"train_loss": 0.1778281033039093, "global_step": 103589, "epoch": 1163, "lr": 6.833667278657235e-05}
{"train_loss": 0.14938212931156158, "global_step": 103590, "epoch": 1163, "lr": 6.833613344127145e-05}
{"train_loss": 0.1817460060119629, "global_step": 103591, "epoch": 1163, "lr": 6.833559409350549e-05}
{"train_loss": 0.12920458614826202, "global_step": 103592, "epoch": 1163, "lr": 6.833505474327453e-05}
{"train_loss": 0.1289805918931961, "global_step": 103593, "epoch": 1163, "lr": 6.833451539057864e-05}
{"train_loss": 0.12230508029460907, "global_step": 103594, "epoch": 1163, "lr": 6.83339760354179e-05}
{"train_loss": 0.15456653461697395, "global_step": 103595, "epoch": 1163, "lr": 6.83334366777924e-05, "val_loss": 4.917579650878906}
{"train_loss": 0.14406202733516693, "global_step": 103596, "epoch": 1164, "lr": 6.833289731770217e-05}
{"train_loss": 0.23136459290981293, "global_step": 103597, "epoch": 1164, "lr": 6.833235795514732e-05}
{"train_loss": 0.17917199432849884, "global_step": 103598, "epoch": 1164, "lr": 6.833181859012789e-05}
{"train_loss": 0.14887353777885437, "global_step": 103599, "epoch": 1164, "lr": 6.833127922264397e-05}
{"train_loss": 0.1358528584241867, "global_step": 103600, "epoch": 1164, "lr": 6.833073985269565e-05}
{"train_loss": 0.11192682385444641, "global_step": 103601, "epoch": 1164, "lr": 6.833020048028297e-05}
{"train_loss": 0.1411018967628479, "global_step": 103602, "epoch": 1164, "lr": 6.832966110540603e-05}
{"train_loss": 0.1745302826166153, "global_step": 103603, "epoch": 1164, "lr": 6.832912172806489e-05}
{"train_loss": 0.10581472516059875, "global_step": 103604, "epoch": 1164, "lr": 6.832858234825961e-05}
{"train_loss": 0.1634923666715622, "global_step": 103605, "epoch": 1164, "lr": 6.832804296599029e-05}
{"train_loss": 0.13242372870445251, "global_step": 103606, "epoch": 1164, "lr": 6.832750358125697e-05}
{"train_loss": 0.15411226451396942, "global_step": 103607, "epoch": 1164, "lr": 6.832696419405977e-05}
{"train_loss": 0.08116848021745682, "global_step": 103608, "epoch": 1164, "lr": 6.832642480439871e-05}
{"train_loss": 0.1476171761751175, "global_step": 103609, "epoch": 1164, "lr": 6.832588541227389e-05}
{"train_loss": 0.13280729949474335, "global_step": 103610, "epoch": 1164, "lr": 6.832534601768539e-05}
{"train_loss": 0.13052470982074738, "global_step": 103611, "epoch": 1164, "lr": 6.832480662063325e-05}
{"train_loss": 0.14106397330760956, "global_step": 103612, "epoch": 1164, "lr": 6.832426722111758e-05}
{"train_loss": 0.15058495104312897, "global_step": 103613, "epoch": 1164, "lr": 6.832372781913843e-05}
{"train_loss": 0.19219651818275452, "global_step": 103614, "epoch": 1164, "lr": 6.832318841469588e-05}
{"train_loss": 0.1128552258014679, "global_step": 103615, "epoch": 1164, "lr": 6.832264900779e-05}
{"train_loss": 0.11642530560493469, "global_step": 103616, "epoch": 1164, "lr": 6.832210959842085e-05}
{"train_loss": 0.17126648128032684, "global_step": 103617, "epoch": 1164, "lr": 6.832157018658854e-05}
{"train_loss": 0.18033625185489655, "global_step": 103618, "epoch": 1164, "lr": 6.832103077229309e-05}
{"train_loss": 0.17197445034980774, "global_step": 103619, "epoch": 1164, "lr": 6.832049135553463e-05}
{"train_loss": 0.14609229564666748, "global_step": 103620, "epoch": 1164, "lr": 6.831995193631318e-05}
{"train_loss": 0.061694271862506866, "global_step": 103621, "epoch": 1164, "lr": 6.831941251462884e-05}
{"train_loss": 0.12141704559326172, "global_step": 103622, "epoch": 1164, "lr": 6.831887309048169e-05}
{"train_loss": 0.17618131637573242, "global_step": 103623, "epoch": 1164, "lr": 6.831833366387178e-05}
{"train_loss": 0.18606775999069214, "global_step": 103624, "epoch": 1164, "lr": 6.83177942347992e-05}
{"train_loss": 0.1654079407453537, "global_step": 103625, "epoch": 1164, "lr": 6.831725480326401e-05}
{"train_loss": 0.13885031640529633, "global_step": 103626, "epoch": 1164, "lr": 6.83167153692663e-05}
{"train_loss": 0.1667591780424118, "global_step": 103627, "epoch": 1164, "lr": 6.831617593280611e-05}
{"train_loss": 0.1528281569480896, "global_step": 103628, "epoch": 1164, "lr": 6.831563649388354e-05}
{"train_loss": 0.1079569086432457, "global_step": 103629, "epoch": 1164, "lr": 6.831509705249867e-05}
{"train_loss": 0.20989902317523956, "global_step": 103630, "epoch": 1164, "lr": 6.831455760865155e-05}
{"train_loss": 0.19823618233203888, "global_step": 103631, "epoch": 1164, "lr": 6.831401816234226e-05}
{"train_loss": 0.20762503147125244, "global_step": 103632, "epoch": 1164, "lr": 6.831347871357088e-05}
{"train_loss": 0.14584633708000183, "global_step": 103633, "epoch": 1164, "lr": 6.831293926233747e-05}
{"train_loss": 0.09604955464601517, "global_step": 103634, "epoch": 1164, "lr": 6.83123998086421e-05}
{"train_loss": 0.21308691799640656, "global_step": 103635, "epoch": 1164, "lr": 6.831186035248487e-05}
{"train_loss": 0.16855505108833313, "global_step": 103636, "epoch": 1164, "lr": 6.831132089386583e-05}
{"train_loss": 0.13810613751411438, "global_step": 103637, "epoch": 1164, "lr": 6.831078143278504e-05}
{"train_loss": 0.15157391130924225, "global_step": 103638, "epoch": 1164, "lr": 6.83102419692426e-05}
{"train_loss": 0.13860651850700378, "global_step": 103639, "epoch": 1164, "lr": 6.830970250323858e-05}
{"train_loss": 0.12071452289819717, "global_step": 103640, "epoch": 1164, "lr": 6.830916303477302e-05}
{"train_loss": 0.11861109733581543, "global_step": 103641, "epoch": 1164, "lr": 6.830862356384604e-05}
{"train_loss": 0.12050601840019226, "global_step": 103642, "epoch": 1164, "lr": 6.830808409045767e-05}
{"train_loss": 0.13502861559391022, "global_step": 103643, "epoch": 1164, "lr": 6.8307544614608e-05}
{"train_loss": 0.19100360572338104, "global_step": 103644, "epoch": 1164, "lr": 6.830700513629712e-05}
{"train_loss": 0.1935858428478241, "global_step": 103645, "epoch": 1164, "lr": 6.830646565552508e-05}
{"train_loss": 0.16775749623775482, "global_step": 103646, "epoch": 1164, "lr": 6.830592617229195e-05}
{"train_loss": 0.14600901305675507, "global_step": 103647, "epoch": 1164, "lr": 6.830538668659783e-05}
{"train_loss": 0.13649345934391022, "global_step": 103648, "epoch": 1164, "lr": 6.830484719844276e-05}
{"train_loss": 0.10467632114887238, "global_step": 103649, "epoch": 1164, "lr": 6.830430770782683e-05}
{"train_loss": 0.22068218886852264, "global_step": 103650, "epoch": 1164, "lr": 6.830376821475011e-05}
{"train_loss": 0.12719318270683289, "global_step": 103651, "epoch": 1164, "lr": 6.830322871921266e-05}
{"train_loss": 0.12422147393226624, "global_step": 103652, "epoch": 1164, "lr": 6.830268922121459e-05}
{"train_loss": 0.11635380238294601, "global_step": 103653, "epoch": 1164, "lr": 6.830214972075593e-05}
{"train_loss": 0.16154515743255615, "global_step": 103654, "epoch": 1164, "lr": 6.830161021783676e-05}
{"train_loss": 0.14734461903572083, "global_step": 103655, "epoch": 1164, "lr": 6.830107071245717e-05}
{"train_loss": 0.22951821982860565, "global_step": 103656, "epoch": 1164, "lr": 6.830053120461723e-05}
{"train_loss": 0.16256177425384521, "global_step": 103657, "epoch": 1164, "lr": 6.8299991694317e-05}
{"train_loss": 0.22097967565059662, "global_step": 103658, "epoch": 1164, "lr": 6.829945218155655e-05}
{"train_loss": 0.16930747032165527, "global_step": 103659, "epoch": 1164, "lr": 6.8298912666336e-05}
{"train_loss": 0.16294841468334198, "global_step": 103660, "epoch": 1164, "lr": 6.829837314865534e-05}
{"train_loss": 0.1323239803314209, "global_step": 103661, "epoch": 1164, "lr": 6.82978336285147e-05}
{"train_loss": 0.14322640001773834, "global_step": 103662, "epoch": 1164, "lr": 6.829729410591414e-05}
{"train_loss": 0.13484609127044678, "global_step": 103663, "epoch": 1164, "lr": 6.829675458085375e-05}
{"train_loss": 0.1161675900220871, "global_step": 103664, "epoch": 1164, "lr": 6.829621505333356e-05}
{"train_loss": 0.18548797070980072, "global_step": 103665, "epoch": 1164, "lr": 6.829567552335368e-05}
{"train_loss": 0.10710635036230087, "global_step": 103666, "epoch": 1164, "lr": 6.829513599091416e-05}
{"train_loss": 0.10189343243837357, "global_step": 103667, "epoch": 1164, "lr": 6.829459645601509e-05}
{"train_loss": 0.1404489129781723, "global_step": 103668, "epoch": 1164, "lr": 6.829405691865654e-05}
{"train_loss": 0.16661664843559265, "global_step": 103669, "epoch": 1164, "lr": 6.829351737883856e-05}
{"train_loss": 0.17122577130794525, "global_step": 103670, "epoch": 1164, "lr": 6.829297783656125e-05}
{"train_loss": 0.15159115195274353, "global_step": 103671, "epoch": 1164, "lr": 6.829243829182468e-05}
{"train_loss": 0.14379996061325073, "global_step": 103672, "epoch": 1164, "lr": 6.829189874462889e-05}
{"train_loss": 0.20718054473400116, "global_step": 103673, "epoch": 1164, "lr": 6.8291359194974e-05}
{"train_loss": 0.13580437004566193, "global_step": 103674, "epoch": 1164, "lr": 6.829081964286005e-05}
{"train_loss": 0.11396043747663498, "global_step": 103675, "epoch": 1164, "lr": 6.829028008828713e-05}
{"train_loss": 0.12649686634540558, "global_step": 103676, "epoch": 1164, "lr": 6.828974053125529e-05}
{"train_loss": 0.1789034605026245, "global_step": 103677, "epoch": 1164, "lr": 6.828920097176463e-05}
{"train_loss": 0.17672818899154663, "global_step": 103678, "epoch": 1164, "lr": 6.82886614098152e-05}
{"train_loss": 0.1581437587738037, "global_step": 103679, "epoch": 1164, "lr": 6.828812184540709e-05}
{"train_loss": 0.14586973190307617, "global_step": 103680, "epoch": 1164, "lr": 6.828758227854037e-05}
{"train_loss": 0.14734485745429993, "global_step": 103681, "epoch": 1164, "lr": 6.82870427092151e-05}
{"train_loss": 0.10419981181621552, "global_step": 103682, "epoch": 1164, "lr": 6.828650313743137e-05}
{"train_loss": 0.17509107291698456, "global_step": 103683, "epoch": 1164, "lr": 6.828596356318923e-05}
{"train_loss": 0.15136356325296874, "global_step": 103684, "epoch": 1164, "lr": 6.828542398648875e-05, "val_loss": 4.875250816345215}
{"train_loss": 0.12615366280078888, "global_step": 103685, "epoch": 1165, "lr": 6.828488440733006e-05}
{"train_loss": 0.10586673021316528, "global_step": 103686, "epoch": 1165, "lr": 6.828434482571315e-05}
{"train_loss": 0.23886995017528534, "global_step": 103687, "epoch": 1165, "lr": 6.828380524163815e-05}
{"train_loss": 0.10211561620235443, "global_step": 103688, "epoch": 1165, "lr": 6.828326565510512e-05}
{"train_loss": 0.16039365530014038, "global_step": 103689, "epoch": 1165, "lr": 6.828272606611413e-05}
{"train_loss": 0.07664873450994492, "global_step": 103690, "epoch": 1165, "lr": 6.828218647466524e-05}
{"train_loss": 0.07390575855970383, "global_step": 103691, "epoch": 1165, "lr": 6.828164688075853e-05}
{"train_loss": 0.18171419203281403, "global_step": 103692, "epoch": 1165, "lr": 6.828110728439408e-05}
{"train_loss": 0.06477250903844833, "global_step": 103693, "epoch": 1165, "lr": 6.828056768557196e-05}
{"train_loss": 0.14733418822288513, "global_step": 103694, "epoch": 1165, "lr": 6.828002808429223e-05}
{"train_loss": 0.1515016108751297, "global_step": 103695, "epoch": 1165, "lr": 6.827948848055498e-05}
{"train_loss": 0.07317973673343658, "global_step": 103696, "epoch": 1165, "lr": 6.827894887436028e-05}
{"train_loss": 0.13492247462272644, "global_step": 103697, "epoch": 1165, "lr": 6.827840926570821e-05}
{"train_loss": 0.1521845906972885, "global_step": 103698, "epoch": 1165, "lr": 6.827786965459882e-05}
{"train_loss": 0.17775806784629822, "global_step": 103699, "epoch": 1165, "lr": 6.827733004103218e-05}
{"train_loss": 0.1819087117910385, "global_step": 103700, "epoch": 1165, "lr": 6.827679042500839e-05}
{"train_loss": 0.1406610757112503, "global_step": 103701, "epoch": 1165, "lr": 6.827625080652749e-05}
{"train_loss": 0.24822132289409637, "global_step": 103702, "epoch": 1165, "lr": 6.827571118558959e-05}
{"train_loss": 0.1898028403520584, "global_step": 103703, "epoch": 1165, "lr": 6.827517156219474e-05}
{"train_loss": 0.146652951836586, "global_step": 103704, "epoch": 1165, "lr": 6.827463193634302e-05}
{"train_loss": 0.15269263088703156, "global_step": 103705, "epoch": 1165, "lr": 6.827409230803448e-05}
{"train_loss": 0.10840430855751038, "global_step": 103706, "epoch": 1165, "lr": 6.827355267726923e-05}
{"train_loss": 0.14811240136623383, "global_step": 103707, "epoch": 1165, "lr": 6.827301304404732e-05}
{"train_loss": 0.08194630593061447, "global_step": 103708, "epoch": 1165, "lr": 6.827247340836882e-05}
{"train_loss": 0.19225502014160156, "global_step": 103709, "epoch": 1165, "lr": 6.827193377023381e-05}
{"train_loss": 0.13926240801811218, "global_step": 103710, "epoch": 1165, "lr": 6.827139412964236e-05}
{"train_loss": 0.12043517827987671, "global_step": 103711, "epoch": 1165, "lr": 6.827085448659455e-05}
{"train_loss": 0.17143534123897552, "global_step": 103712, "epoch": 1165, "lr": 6.827031484109045e-05}
{"train_loss": 0.10830609500408173, "global_step": 103713, "epoch": 1165, "lr": 6.826977519313011e-05}
{"train_loss": 0.19951696693897247, "global_step": 103714, "epoch": 1165, "lr": 6.826923554271364e-05}
{"train_loss": 0.15579617023468018, "global_step": 103715, "epoch": 1165, "lr": 6.826869588984109e-05}
{"train_loss": 0.1940927356481552, "global_step": 103716, "epoch": 1165, "lr": 6.826815623451253e-05}
{"train_loss": 0.1251952350139618, "global_step": 103717, "epoch": 1165, "lr": 6.826761657672804e-05}
{"train_loss": 0.22537310421466827, "global_step": 103718, "epoch": 1165, "lr": 6.82670769164877e-05}
{"train_loss": 0.17662179470062256, "global_step": 103719, "epoch": 1165, "lr": 6.826653725379156e-05}
{"train_loss": 0.15122120082378387, "global_step": 103720, "epoch": 1165, "lr": 6.826599758863972e-05}
{"train_loss": 0.19997267425060272, "global_step": 103721, "epoch": 1165, "lr": 6.826545792103223e-05}
{"train_loss": 0.17277392745018005, "global_step": 103722, "epoch": 1165, "lr": 6.826491825096918e-05}
{"train_loss": 0.1618630439043045, "global_step": 103723, "epoch": 1165, "lr": 6.826437857845063e-05}
{"train_loss": 0.1042027622461319, "global_step": 103724, "epoch": 1165, "lr": 6.826383890347667e-05}
{"train_loss": 0.10624489188194275, "global_step": 103725, "epoch": 1165, "lr": 6.826329922604734e-05}
{"train_loss": 0.12029245495796204, "global_step": 103726, "epoch": 1165, "lr": 6.826275954616273e-05}
{"train_loss": 0.16196489334106445, "global_step": 103727, "epoch": 1165, "lr": 6.826221986382293e-05}
{"train_loss": 0.17158934473991394, "global_step": 103728, "epoch": 1165, "lr": 6.826168017902799e-05}
{"train_loss": 0.15390844643115997, "global_step": 103729, "epoch": 1165, "lr": 6.826114049177799e-05}
{"train_loss": 0.23604241013526917, "global_step": 103730, "epoch": 1165, "lr": 6.826060080207301e-05}
{"train_loss": 0.14943918585777283, "global_step": 103731, "epoch": 1165, "lr": 6.82600611099131e-05}
{"train_loss": 0.13487720489501953, "global_step": 103732, "epoch": 1165, "lr": 6.825952141529836e-05}
{"train_loss": 0.1473008543252945, "global_step": 103733, "epoch": 1165, "lr": 6.825898171822884e-05}
{"train_loss": 0.08163432776927948, "global_step": 103734, "epoch": 1165, "lr": 6.825844201870462e-05}
{"train_loss": 0.2104760706424713, "global_step": 103735, "epoch": 1165, "lr": 6.825790231672578e-05}
{"train_loss": 0.20176343619823456, "global_step": 103736, "epoch": 1165, "lr": 6.82573626122924e-05}
{"train_loss": 0.22535111010074615, "global_step": 103737, "epoch": 1165, "lr": 6.825682290540452e-05}
{"train_loss": 0.14226846396923065, "global_step": 103738, "epoch": 1165, "lr": 6.825628319606225e-05}
{"train_loss": 0.2124994993209839, "global_step": 103739, "epoch": 1165, "lr": 6.825574348426564e-05}
{"train_loss": 0.1312398910522461, "global_step": 103740, "epoch": 1165, "lr": 6.825520377001477e-05}
{"train_loss": 0.11459968239068985, "global_step": 103741, "epoch": 1165, "lr": 6.82546640533097e-05}
{"train_loss": 0.15740542113780975, "global_step": 103742, "epoch": 1165, "lr": 6.825412433415052e-05}
{"train_loss": 0.15976333618164062, "global_step": 103743, "epoch": 1165, "lr": 6.825358461253728e-05}
{"train_loss": 0.23320789635181427, "global_step": 103744, "epoch": 1165, "lr": 6.825304488847008e-05}
{"train_loss": 0.23509855568408966, "global_step": 103745, "epoch": 1165, "lr": 6.8252505161949e-05}
{"train_loss": 0.20917552709579468, "global_step": 103746, "epoch": 1165, "lr": 6.825196543297408e-05}
{"train_loss": 0.2971096336841583, "global_step": 103747, "epoch": 1165, "lr": 6.825142570154539e-05}
{"train_loss": 0.17714956402778625, "global_step": 103748, "epoch": 1165, "lr": 6.825088596766304e-05}
{"train_loss": 0.15559907257556915, "global_step": 103749, "epoch": 1165, "lr": 6.825034623132706e-05}
{"train_loss": 0.2630784213542938, "global_step": 103750, "epoch": 1165, "lr": 6.824980649253756e-05}
{"train_loss": 0.18761631846427917, "global_step": 103751, "epoch": 1165, "lr": 6.82492667512946e-05}
{"train_loss": 0.21739302575588226, "global_step": 103752, "epoch": 1165, "lr": 6.824872700759822e-05}
{"train_loss": 0.13619565963745117, "global_step": 103753, "epoch": 1165, "lr": 6.824818726144855e-05}
{"train_loss": 0.18785454332828522, "global_step": 103754, "epoch": 1165, "lr": 6.824764751284563e-05}
{"train_loss": 0.21606428921222687, "global_step": 103755, "epoch": 1165, "lr": 6.824710776178952e-05}
{"train_loss": 0.12749367952346802, "global_step": 103756, "epoch": 1165, "lr": 6.824656800828033e-05}
{"train_loss": 0.13958030939102173, "global_step": 103757, "epoch": 1165, "lr": 6.824602825231808e-05}
{"train_loss": 0.14510497450828552, "global_step": 103758, "epoch": 1165, "lr": 6.82454884939029e-05}
{"train_loss": 0.14271622896194458, "global_step": 103759, "epoch": 1165, "lr": 6.824494873303483e-05}
{"train_loss": 0.21014845371246338, "global_step": 103760, "epoch": 1165, "lr": 6.824440896971395e-05}
{"train_loss": 0.18160368502140045, "global_step": 103761, "epoch": 1165, "lr": 6.824386920394033e-05}
{"train_loss": 0.23205478489398956, "global_step": 103762, "epoch": 1165, "lr": 6.824332943571404e-05}
{"train_loss": 0.17418642342090607, "global_step": 103763, "epoch": 1165, "lr": 6.824278966503516e-05}
{"train_loss": 0.11757589131593704, "global_step": 103764, "epoch": 1165, "lr": 6.824224989190376e-05}
{"train_loss": 0.07381916791200638, "global_step": 103765, "epoch": 1165, "lr": 6.824171011631993e-05}
{"train_loss": 0.21904441714286804, "global_step": 103766, "epoch": 1165, "lr": 6.82411703382837e-05}
{"train_loss": 0.21067526936531067, "global_step": 103767, "epoch": 1165, "lr": 6.824063055779517e-05}
{"train_loss": 0.11572261154651642, "global_step": 103768, "epoch": 1165, "lr": 6.824009077485441e-05}
{"train_loss": 0.19423851370811462, "global_step": 103769, "epoch": 1165, "lr": 6.82395509894615e-05}
{"train_loss": 0.15225346386432648, "global_step": 103770, "epoch": 1165, "lr": 6.823901120161649e-05}
{"train_loss": 0.22623541951179504, "global_step": 103771, "epoch": 1165, "lr": 6.823847141131948e-05}
{"train_loss": 0.2362350970506668, "global_step": 103772, "epoch": 1165, "lr": 6.823793161857053e-05}
{"train_loss": 0.1632708382238163, "global_step": 103773, "epoch": 1165, "lr": 6.82373918233697e-05, "val_loss": 4.556309223175049, "train_action_mse_error": 24.836536407470703}
{"train_loss": 0.1131460890173912, "global_step": 103774, "epoch": 1166, "lr": 6.82368520257171e-05}
{"train_loss": 0.12032036483287811, "global_step": 103775, "epoch": 1166, "lr": 6.823631222561277e-05}
{"train_loss": 0.09782381355762482, "global_step": 103776, "epoch": 1166, "lr": 6.823577242305678e-05}
{"train_loss": 0.16223694384098053, "global_step": 103777, "epoch": 1166, "lr": 6.82352326180492e-05}
{"train_loss": 0.1648300588130951, "global_step": 103778, "epoch": 1166, "lr": 6.823469281059014e-05}
{"train_loss": 0.11320113390684128, "global_step": 103779, "epoch": 1166, "lr": 6.823415300067963e-05}
{"train_loss": 0.19895966351032257, "global_step": 103780, "epoch": 1166, "lr": 6.823361318831776e-05}
{"train_loss": 0.2377845048904419, "global_step": 103781, "epoch": 1166, "lr": 6.823307337350461e-05}
{"train_loss": 0.11911913007497787, "global_step": 103782, "epoch": 1166, "lr": 6.823253355624025e-05}
{"train_loss": 0.11794732511043549, "global_step": 103783, "epoch": 1166, "lr": 6.823199373652475e-05}
{"train_loss": 0.2192218154668808, "global_step": 103784, "epoch": 1166, "lr": 6.823145391435817e-05}
{"train_loss": 0.16136382520198822, "global_step": 103785, "epoch": 1166, "lr": 6.82309140897406e-05}
{"train_loss": 0.14730216562747955, "global_step": 103786, "epoch": 1166, "lr": 6.823037426267211e-05}
{"train_loss": 0.17496395111083984, "global_step": 103787, "epoch": 1166, "lr": 6.822983443315276e-05}
{"train_loss": 0.25726351141929626, "global_step": 103788, "epoch": 1166, "lr": 6.822929460118264e-05}
{"train_loss": 0.18705394864082336, "global_step": 103789, "epoch": 1166, "lr": 6.822875476676181e-05}
{"train_loss": 0.15059159696102142, "global_step": 103790, "epoch": 1166, "lr": 6.822821492989034e-05}
{"train_loss": 0.14071258902549744, "global_step": 103791, "epoch": 1166, "lr": 6.822767509056831e-05}
{"train_loss": 0.2535493075847626, "global_step": 103792, "epoch": 1166, "lr": 6.82271352487958e-05}
{"train_loss": 0.18116573989391327, "global_step": 103793, "epoch": 1166, "lr": 6.822659540457287e-05}
{"train_loss": 0.11029350012540817, "global_step": 103794, "epoch": 1166, "lr": 6.82260555578996e-05}
{"train_loss": 0.22561855614185333, "global_step": 103795, "epoch": 1166, "lr": 6.822551570877605e-05}
{"train_loss": 0.15595847368240356, "global_step": 103796, "epoch": 1166, "lr": 6.82249758572023e-05}
{"train_loss": 0.20812030136585236, "global_step": 103797, "epoch": 1166, "lr": 6.822443600317843e-05}
{"train_loss": 0.1154547706246376, "global_step": 103798, "epoch": 1166, "lr": 6.822389614670449e-05}
{"train_loss": 0.1748446375131607, "global_step": 103799, "epoch": 1166, "lr": 6.822335628778058e-05}
{"train_loss": 0.14234709739685059, "global_step": 103800, "epoch": 1166, "lr": 6.822281642640677e-05}
{"train_loss": 0.1990039199590683, "global_step": 103801, "epoch": 1166, "lr": 6.822227656258312e-05}
{"train_loss": 0.14045411348342896, "global_step": 103802, "epoch": 1166, "lr": 6.822173669630972e-05}
{"train_loss": 0.1467628926038742, "global_step": 103803, "epoch": 1166, "lr": 6.822119682758661e-05}
{"train_loss": 0.20775818824768066, "global_step": 103804, "epoch": 1166, "lr": 6.822065695641389e-05}
{"train_loss": 0.16707418859004974, "global_step": 103805, "epoch": 1166, "lr": 6.822011708279163e-05}
{"train_loss": 0.1457703709602356, "global_step": 103806, "epoch": 1166, "lr": 6.821957720671989e-05}
{"train_loss": 0.11088430881500244, "global_step": 103807, "epoch": 1166, "lr": 6.821903732819875e-05}
{"train_loss": 0.20277635753154755, "global_step": 103808, "epoch": 1166, "lr": 6.821849744722828e-05}
{"train_loss": 0.15001390874385834, "global_step": 103809, "epoch": 1166, "lr": 6.821795756380856e-05}
{"train_loss": 0.12907017767429352, "global_step": 103810, "epoch": 1166, "lr": 6.821741767793965e-05}
{"train_loss": 0.17997005581855774, "global_step": 103811, "epoch": 1166, "lr": 6.821687778962164e-05}
{"train_loss": 0.21458843350410461, "global_step": 103812, "epoch": 1166, "lr": 6.821633789885459e-05}
{"train_loss": 0.15018907189369202, "global_step": 103813, "epoch": 1166, "lr": 6.821579800563858e-05}
{"train_loss": 0.16722992062568665, "global_step": 103814, "epoch": 1166, "lr": 6.821525810997368e-05}
{"train_loss": 0.18765141069889069, "global_step": 103815, "epoch": 1166, "lr": 6.821471821185995e-05}
{"train_loss": 0.16003406047821045, "global_step": 103816, "epoch": 1166, "lr": 6.821417831129746e-05}
{"train_loss": 0.19919337332248688, "global_step": 103817, "epoch": 1166, "lr": 6.821363840828632e-05}
{"train_loss": 0.1354152411222458, "global_step": 103818, "epoch": 1166, "lr": 6.821309850282656e-05}
{"train_loss": 0.1602097749710083, "global_step": 103819, "epoch": 1166, "lr": 6.821255859491829e-05}
{"train_loss": 0.13869957625865936, "global_step": 103820, "epoch": 1166, "lr": 6.821201868456155e-05}
{"train_loss": 0.13662566244602203, "global_step": 103821, "epoch": 1166, "lr": 6.821147877175644e-05}
{"train_loss": 0.18922580778598785, "global_step": 103822, "epoch": 1166, "lr": 6.821093885650301e-05}
{"train_loss": 0.14074289798736572, "global_step": 103823, "epoch": 1166, "lr": 6.821039893880135e-05}
{"train_loss": 0.1312144696712494, "global_step": 103824, "epoch": 1166, "lr": 6.820985901865151e-05}
{"train_loss": 0.13979196548461914, "global_step": 103825, "epoch": 1166, "lr": 6.820931909605359e-05}
{"train_loss": 0.16969650983810425, "global_step": 103826, "epoch": 1166, "lr": 6.820877917100763e-05}
{"train_loss": 0.17079880833625793, "global_step": 103827, "epoch": 1166, "lr": 6.820823924351373e-05}
{"train_loss": 0.11810965090990067, "global_step": 103828, "epoch": 1166, "lr": 6.820769931357195e-05}
{"train_loss": 0.15580463409423828, "global_step": 103829, "epoch": 1166, "lr": 6.820715938118238e-05}
{"train_loss": 0.19826816022396088, "global_step": 103830, "epoch": 1166, "lr": 6.820661944634508e-05}
{"train_loss": 0.19206005334854126, "global_step": 103831, "epoch": 1166, "lr": 6.82060795090601e-05}
{"train_loss": 0.2436804473400116, "global_step": 103832, "epoch": 1166, "lr": 6.820553956932756e-05}
{"train_loss": 0.11424720287322998, "global_step": 103833, "epoch": 1166, "lr": 6.82049996271475e-05}
{"train_loss": 0.15726402401924133, "global_step": 103834, "epoch": 1166, "lr": 6.820445968251998e-05}
{"train_loss": 0.1855112761259079, "global_step": 103835, "epoch": 1166, "lr": 6.820391973544511e-05}
{"train_loss": 0.11533032357692719, "global_step": 103836, "epoch": 1166, "lr": 6.820337978592294e-05}
{"train_loss": 0.13170841336250305, "global_step": 103837, "epoch": 1166, "lr": 6.820283983395354e-05}
{"train_loss": 0.15613532066345215, "global_step": 103838, "epoch": 1166, "lr": 6.820229987953701e-05}
{"train_loss": 0.09801936894655228, "global_step": 103839, "epoch": 1166, "lr": 6.820175992267339e-05}
{"train_loss": 0.15173523128032684, "global_step": 103840, "epoch": 1166, "lr": 6.820121996336277e-05}
{"train_loss": 0.13866831362247467, "global_step": 103841, "epoch": 1166, "lr": 6.820068000160523e-05}
{"train_loss": 0.19118693470954895, "global_step": 103842, "epoch": 1166, "lr": 6.82001400374008e-05}
{"train_loss": 0.1563023328781128, "global_step": 103843, "epoch": 1166, "lr": 6.819960007074962e-05}
{"train_loss": 0.16886413097381592, "global_step": 103844, "epoch": 1166, "lr": 6.819906010165168e-05}
{"train_loss": 0.11226588487625122, "global_step": 103845, "epoch": 1166, "lr": 6.819852013010713e-05}
{"train_loss": 0.12131784856319427, "global_step": 103846, "epoch": 1166, "lr": 6.819798015611602e-05}
{"train_loss": 0.22796079516410828, "global_step": 103847, "epoch": 1166, "lr": 6.819744017967839e-05}
{"train_loss": 0.16826847195625305, "global_step": 103848, "epoch": 1166, "lr": 6.819690020079434e-05}
{"train_loss": 0.17119371891021729, "global_step": 103849, "epoch": 1166, "lr": 6.819636021946394e-05}
{"train_loss": 0.2060764580965042, "global_step": 103850, "epoch": 1166, "lr": 6.819582023568727e-05}
{"train_loss": 0.21337756514549255, "global_step": 103851, "epoch": 1166, "lr": 6.819528024946438e-05}
{"train_loss": 0.12220481038093567, "global_step": 103852, "epoch": 1166, "lr": 6.819474026079535e-05}
{"train_loss": 0.12718307971954346, "global_step": 103853, "epoch": 1166, "lr": 6.819420026968028e-05}
{"train_loss": 0.17551064491271973, "global_step": 103854, "epoch": 1166, "lr": 6.819366027611921e-05}
{"train_loss": 0.14946207404136658, "global_step": 103855, "epoch": 1166, "lr": 6.819312028011222e-05}
{"train_loss": 0.09314024448394775, "global_step": 103856, "epoch": 1166, "lr": 6.819258028165941e-05}
{"train_loss": 0.13782060146331787, "global_step": 103857, "epoch": 1166, "lr": 6.819204028076081e-05}
{"train_loss": 0.15570992231369019, "global_step": 103858, "epoch": 1166, "lr": 6.819150027741651e-05}
{"train_loss": 0.193552166223526, "global_step": 103859, "epoch": 1166, "lr": 6.819096027162659e-05}
{"train_loss": 0.2040611207485199, "global_step": 103860, "epoch": 1166, "lr": 6.819042026339111e-05}
{"train_loss": 0.1288275569677353, "global_step": 103861, "epoch": 1166, "lr": 6.818988025271017e-05}
{"train_loss": 0.161389000881254, "global_step": 103862, "epoch": 1166, "lr": 6.818934023958382e-05, "val_loss": 4.88270902633667}
{"train_loss": 0.17798013985157013, "global_step": 103863, "epoch": 1167, "lr": 6.818880022401211e-05}
{"train_loss": 0.1495477259159088, "global_step": 103864, "epoch": 1167, "lr": 6.818826020599516e-05}
{"train_loss": 0.16026771068572998, "global_step": 103865, "epoch": 1167, "lr": 6.818772018553301e-05}
{"train_loss": 0.13280785083770752, "global_step": 103866, "epoch": 1167, "lr": 6.818718016262575e-05}
{"train_loss": 0.10378383845090866, "global_step": 103867, "epoch": 1167, "lr": 6.818664013727343e-05}
{"train_loss": 0.1621277779340744, "global_step": 103868, "epoch": 1167, "lr": 6.818610010947614e-05}
{"train_loss": 0.1939322054386139, "global_step": 103869, "epoch": 1167, "lr": 6.818556007923397e-05}
{"train_loss": 0.17571194469928741, "global_step": 103870, "epoch": 1167, "lr": 6.818502004654694e-05}
{"train_loss": 0.21748319268226624, "global_step": 103871, "epoch": 1167, "lr": 6.818448001141519e-05}
{"train_loss": 0.17293284833431244, "global_step": 103872, "epoch": 1167, "lr": 6.818393997383874e-05}
{"train_loss": 0.2140314280986786, "global_step": 103873, "epoch": 1167, "lr": 6.818339993381768e-05}
{"train_loss": 0.24915656447410583, "global_step": 103874, "epoch": 1167, "lr": 6.818285989135208e-05}
{"train_loss": 0.1512427031993866, "global_step": 103875, "epoch": 1167, "lr": 6.818231984644203e-05}
{"train_loss": 0.15327587723731995, "global_step": 103876, "epoch": 1167, "lr": 6.81817797990876e-05}
{"train_loss": 0.11150509864091873, "global_step": 103877, "epoch": 1167, "lr": 6.818123974928883e-05}
{"train_loss": 0.10234034061431885, "global_step": 103878, "epoch": 1167, "lr": 6.818069969704581e-05}
{"train_loss": 0.15896138548851013, "global_step": 103879, "epoch": 1167, "lr": 6.818015964235863e-05}
{"train_loss": 0.17676404118537903, "global_step": 103880, "epoch": 1167, "lr": 6.817961958522736e-05}
{"train_loss": 0.1690119057893753, "global_step": 103881, "epoch": 1167, "lr": 6.817907952565204e-05}
{"train_loss": 0.09248479455709457, "global_step": 103882, "epoch": 1167, "lr": 6.817853946363278e-05}
{"train_loss": 0.10531473904848099, "global_step": 103883, "epoch": 1167, "lr": 6.817799939916962e-05}
{"train_loss": 0.1612643301486969, "global_step": 103884, "epoch": 1167, "lr": 6.817745933226266e-05}
{"train_loss": 0.16234248876571655, "global_step": 103885, "epoch": 1167, "lr": 6.817691926291196e-05}
{"train_loss": 0.11700785905122757, "global_step": 103886, "epoch": 1167, "lr": 6.817637919111758e-05}
{"train_loss": 0.12972496449947357, "global_step": 103887, "epoch": 1167, "lr": 6.817583911687964e-05}
{"train_loss": 0.1174357458949089, "global_step": 103888, "epoch": 1167, "lr": 6.817529904019815e-05}
{"train_loss": 0.15567989647388458, "global_step": 103889, "epoch": 1167, "lr": 6.817475896107323e-05}
{"train_loss": 0.1968875229358673, "global_step": 103890, "epoch": 1167, "lr": 6.817421887950492e-05}
{"train_loss": 0.14690759778022766, "global_step": 103891, "epoch": 1167, "lr": 6.817367879549332e-05}
{"train_loss": 0.15440259873867035, "global_step": 103892, "epoch": 1167, "lr": 6.817313870903849e-05}
{"train_loss": 0.09045787155628204, "global_step": 103893, "epoch": 1167, "lr": 6.817259862014049e-05}
{"train_loss": 0.1404443234205246, "global_step": 103894, "epoch": 1167, "lr": 6.817205852879943e-05}
{"train_loss": 0.11954644322395325, "global_step": 103895, "epoch": 1167, "lr": 6.817151843501533e-05}
{"train_loss": 0.22668522596359253, "global_step": 103896, "epoch": 1167, "lr": 6.817097833878831e-05}
{"train_loss": 0.1327134668827057, "global_step": 103897, "epoch": 1167, "lr": 6.817043824011842e-05}
{"train_loss": 0.12553074955940247, "global_step": 103898, "epoch": 1167, "lr": 6.816989813900574e-05}
{"train_loss": 0.13856938481330872, "global_step": 103899, "epoch": 1167, "lr": 6.816935803545033e-05}
{"train_loss": 0.2075771987438202, "global_step": 103900, "epoch": 1167, "lr": 6.816881792945229e-05}
{"train_loss": 0.16314856708049774, "global_step": 103901, "epoch": 1167, "lr": 6.816827782101165e-05}
{"train_loss": 0.15019011497497559, "global_step": 103902, "epoch": 1167, "lr": 6.816773771012853e-05}
{"train_loss": 0.10980083793401718, "global_step": 103903, "epoch": 1167, "lr": 6.816719759680296e-05}
{"train_loss": 0.18378014862537384, "global_step": 103904, "epoch": 1167, "lr": 6.816665748103504e-05}
{"train_loss": 0.10657229274511337, "global_step": 103905, "epoch": 1167, "lr": 6.816611736282482e-05}
{"train_loss": 0.14947880804538727, "global_step": 103906, "epoch": 1167, "lr": 6.816557724217238e-05}
{"train_loss": 0.173424631357193, "global_step": 103907, "epoch": 1167, "lr": 6.816503711907783e-05}
{"train_loss": 0.19384486973285675, "global_step": 103908, "epoch": 1167, "lr": 6.816449699354119e-05}
{"train_loss": 0.2042534500360489, "global_step": 103909, "epoch": 1167, "lr": 6.816395686556257e-05}
{"train_loss": 0.14038726687431335, "global_step": 103910, "epoch": 1167, "lr": 6.816341673514202e-05}
{"train_loss": 0.18306398391723633, "global_step": 103911, "epoch": 1167, "lr": 6.816287660227962e-05}
{"train_loss": 0.21242506802082062, "global_step": 103912, "epoch": 1167, "lr": 6.816233646697545e-05}
{"train_loss": 0.1253475546836853, "global_step": 103913, "epoch": 1167, "lr": 6.816179632922955e-05}
{"train_loss": 0.11180832237005234, "global_step": 103914, "epoch": 1167, "lr": 6.816125618904204e-05}
{"train_loss": 0.10705923289060593, "global_step": 103915, "epoch": 1167, "lr": 6.816071604641297e-05}
{"train_loss": 0.16595058143138885, "global_step": 103916, "epoch": 1167, "lr": 6.816017590134241e-05}
{"train_loss": 0.14439934492111206, "global_step": 103917, "epoch": 1167, "lr": 6.815963575383044e-05}
{"train_loss": 0.3030219078063965, "global_step": 103918, "epoch": 1167, "lr": 6.815909560387712e-05}
{"train_loss": 0.1555139571428299, "global_step": 103919, "epoch": 1167, "lr": 6.815855545148256e-05}
{"train_loss": 0.15863671898841858, "global_step": 103920, "epoch": 1167, "lr": 6.815801529664676e-05}
{"train_loss": 0.16157092154026031, "global_step": 103921, "epoch": 1167, "lr": 6.815747513936987e-05}
{"train_loss": 0.1761460304260254, "global_step": 103922, "epoch": 1167, "lr": 6.81569349796519e-05}
{"train_loss": 0.15023384988307953, "global_step": 103923, "epoch": 1167, "lr": 6.815639481749297e-05}
{"train_loss": 0.11522522568702698, "global_step": 103924, "epoch": 1167, "lr": 6.815585465289312e-05}
{"train_loss": 0.11454558372497559, "global_step": 103925, "epoch": 1167, "lr": 6.815531448585244e-05}
{"train_loss": 0.15324273705482483, "global_step": 103926, "epoch": 1167, "lr": 6.8154774316371e-05}
{"train_loss": 0.1854637861251831, "global_step": 103927, "epoch": 1167, "lr": 6.815423414444888e-05}
{"train_loss": 0.16427086293697357, "global_step": 103928, "epoch": 1167, "lr": 6.815369397008616e-05}
{"train_loss": 0.10686099529266357, "global_step": 103929, "epoch": 1167, "lr": 6.815315379328286e-05}
{"train_loss": 0.2545507252216339, "global_step": 103930, "epoch": 1167, "lr": 6.81526136140391e-05}
{"train_loss": 0.12625402212142944, "global_step": 103931, "epoch": 1167, "lr": 6.815207343235497e-05}
{"train_loss": 0.19159738719463348, "global_step": 103932, "epoch": 1167, "lr": 6.815153324823049e-05}
{"train_loss": 0.16588883101940155, "global_step": 103933, "epoch": 1167, "lr": 6.815099306166577e-05}
{"train_loss": 0.16183103621006012, "global_step": 103934, "epoch": 1167, "lr": 6.815045287266086e-05}
{"train_loss": 0.17446351051330566, "global_step": 103935, "epoch": 1167, "lr": 6.814991268121584e-05}
{"train_loss": 0.1314358413219452, "global_step": 103936, "epoch": 1167, "lr": 6.81493724873308e-05}
{"train_loss": 0.1466626673936844, "global_step": 103937, "epoch": 1167, "lr": 6.814883229100578e-05}
{"train_loss": 0.12883836030960083, "global_step": 103938, "epoch": 1167, "lr": 6.81482920922409e-05}
{"train_loss": 0.13333137333393097, "global_step": 103939, "epoch": 1167, "lr": 6.814775189103618e-05}
{"train_loss": 0.234026238322258, "global_step": 103940, "epoch": 1167, "lr": 6.81472116873917e-05}
{"train_loss": 0.13759687542915344, "global_step": 103941, "epoch": 1167, "lr": 6.814667148130758e-05}
{"train_loss": 0.1592458337545395, "global_step": 103942, "epoch": 1167, "lr": 6.814613127278386e-05}
{"train_loss": 0.15140044689178467, "global_step": 103943, "epoch": 1167, "lr": 6.814559106182059e-05}
{"train_loss": 0.2810687720775604, "global_step": 103944, "epoch": 1167, "lr": 6.814505084841788e-05}
{"train_loss": 0.12915173172950745, "global_step": 103945, "epoch": 1167, "lr": 6.814451063257579e-05}
{"train_loss": 0.11200793087482452, "global_step": 103946, "epoch": 1167, "lr": 6.814397041429439e-05}
{"train_loss": 0.11914294213056564, "global_step": 103947, "epoch": 1167, "lr": 6.814343019357374e-05}
{"train_loss": 0.1449296921491623, "global_step": 103948, "epoch": 1167, "lr": 6.814288997041395e-05}
{"train_loss": 0.10712400823831558, "global_step": 103949, "epoch": 1167, "lr": 6.814234974481506e-05}
{"train_loss": 0.18396423757076263, "global_step": 103950, "epoch": 1167, "lr": 6.814180951677715e-05}
{"train_loss": 0.1564409098263537, "global_step": 103951, "epoch": 1167, "lr": 6.81412692863003e-05, "val_loss": 4.952485084533691}
{"train_loss": 0.14820462465286255, "global_step": 103952, "epoch": 1168, "lr": 6.814072905338458e-05}
{"train_loss": 0.16235604882240295, "global_step": 103953, "epoch": 1168, "lr": 6.814018881803005e-05}
{"train_loss": 0.1836276650428772, "global_step": 103954, "epoch": 1168, "lr": 6.81396485802368e-05}
{"train_loss": 0.15260359644889832, "global_step": 103955, "epoch": 1168, "lr": 6.81391083400049e-05}
{"train_loss": 0.14853547513484955, "global_step": 103956, "epoch": 1168, "lr": 6.813856809733441e-05}
{"train_loss": 0.1481674462556839, "global_step": 103957, "epoch": 1168, "lr": 6.813802785222543e-05}
{"train_loss": 0.13322317600250244, "global_step": 103958, "epoch": 1168, "lr": 6.813748760467799e-05}
{"train_loss": 0.171439990401268, "global_step": 103959, "epoch": 1168, "lr": 6.81369473546922e-05}
{"train_loss": 0.17631874978542328, "global_step": 103960, "epoch": 1168, "lr": 6.81364071022681e-05}
{"train_loss": 0.09263669699430466, "global_step": 103961, "epoch": 1168, "lr": 6.813586684740578e-05}
{"train_loss": 0.15824659168720245, "global_step": 103962, "epoch": 1168, "lr": 6.813532659010533e-05}
{"train_loss": 0.13529716432094574, "global_step": 103963, "epoch": 1168, "lr": 6.81347863303668e-05}
{"train_loss": 0.143624410033226, "global_step": 103964, "epoch": 1168, "lr": 6.813424606819026e-05}
{"train_loss": 0.13734768331050873, "global_step": 103965, "epoch": 1168, "lr": 6.813370580357582e-05}
{"train_loss": 0.19103898108005524, "global_step": 103966, "epoch": 1168, "lr": 6.813316553652349e-05}
{"train_loss": 0.1644129604101181, "global_step": 103967, "epoch": 1168, "lr": 6.81326252670334e-05}
{"train_loss": 0.10781771689653397, "global_step": 103968, "epoch": 1168, "lr": 6.813208499510559e-05}
{"train_loss": 0.08823272585868835, "global_step": 103969, "epoch": 1168, "lr": 6.813154472074014e-05}
{"train_loss": 0.21071743965148926, "global_step": 103970, "epoch": 1168, "lr": 6.813100444393712e-05}
{"train_loss": 0.11470703035593033, "global_step": 103971, "epoch": 1168, "lr": 6.813046416469663e-05}
{"train_loss": 0.1436695158481598, "global_step": 103972, "epoch": 1168, "lr": 6.812992388301871e-05}
{"train_loss": 0.16274328529834747, "global_step": 103973, "epoch": 1168, "lr": 6.812938359890344e-05}
{"train_loss": 0.2057257890701294, "global_step": 103974, "epoch": 1168, "lr": 6.81288433123509e-05}
{"train_loss": 0.13107085227966309, "global_step": 103975, "epoch": 1168, "lr": 6.812830302336115e-05}
{"train_loss": 0.17797300219535828, "global_step": 103976, "epoch": 1168, "lr": 6.812776273193426e-05}
{"train_loss": 0.1344936639070511, "global_step": 103977, "epoch": 1168, "lr": 6.812722243807034e-05}
{"train_loss": 0.21183040738105774, "global_step": 103978, "epoch": 1168, "lr": 6.812668214176943e-05}
{"train_loss": 0.16754363477230072, "global_step": 103979, "epoch": 1168, "lr": 6.812614184303159e-05}
{"train_loss": 0.1346236914396286, "global_step": 103980, "epoch": 1168, "lr": 6.812560154185692e-05}
{"train_loss": 0.1779407262802124, "global_step": 103981, "epoch": 1168, "lr": 6.812506123824549e-05}
{"train_loss": 0.11896228045225143, "global_step": 103982, "epoch": 1168, "lr": 6.812452093219736e-05}
{"train_loss": 0.14872877299785614, "global_step": 103983, "epoch": 1168, "lr": 6.812398062371261e-05}
{"train_loss": 0.17024463415145874, "global_step": 103984, "epoch": 1168, "lr": 6.812344031279133e-05}
{"train_loss": 0.13549859821796417, "global_step": 103985, "epoch": 1168, "lr": 6.812289999943354e-05}
{"train_loss": 0.11827866733074188, "global_step": 103986, "epoch": 1168, "lr": 6.812235968363937e-05}
{"train_loss": 0.18706175684928894, "global_step": 103987, "epoch": 1168, "lr": 6.812181936540886e-05}
{"train_loss": 0.11293753981590271, "global_step": 103988, "epoch": 1168, "lr": 6.81212790447421e-05}
{"train_loss": 0.14856037497520447, "global_step": 103989, "epoch": 1168, "lr": 6.812073872163915e-05}
{"train_loss": 0.11670613288879395, "global_step": 103990, "epoch": 1168, "lr": 6.812019839610009e-05}
{"train_loss": 0.1774444431066513, "global_step": 103991, "epoch": 1168, "lr": 6.811965806812499e-05}
{"train_loss": 0.15032504498958588, "global_step": 103992, "epoch": 1168, "lr": 6.811911773771392e-05}
{"train_loss": 0.10744668543338776, "global_step": 103993, "epoch": 1168, "lr": 6.811857740486696e-05}
{"train_loss": 0.1960098147392273, "global_step": 103994, "epoch": 1168, "lr": 6.811803706958418e-05}
{"train_loss": 0.10333077609539032, "global_step": 103995, "epoch": 1168, "lr": 6.811749673186565e-05}
{"train_loss": 0.15844163298606873, "global_step": 103996, "epoch": 1168, "lr": 6.811695639171144e-05}
{"train_loss": 0.15982654690742493, "global_step": 103997, "epoch": 1168, "lr": 6.811641604912162e-05}
{"train_loss": 0.10679440200328827, "global_step": 103998, "epoch": 1168, "lr": 6.811587570409628e-05}
{"train_loss": 0.20710457861423492, "global_step": 103999, "epoch": 1168, "lr": 6.811533535663547e-05}
{"train_loss": 0.15753349661827087, "global_step": 104000, "epoch": 1168, "lr": 6.811479500673928e-05}
{"train_loss": 0.1341465413570404, "global_step": 104001, "epoch": 1168, "lr": 6.811425465440777e-05}
{"train_loss": 0.12033203989267349, "global_step": 104002, "epoch": 1168, "lr": 6.811371429964103e-05}
{"train_loss": 0.10192014276981354, "global_step": 104003, "epoch": 1168, "lr": 6.811317394243912e-05}
{"train_loss": 0.25836116075515747, "global_step": 104004, "epoch": 1168, "lr": 6.811263358280211e-05}
{"train_loss": 0.22557899355888367, "global_step": 104005, "epoch": 1168, "lr": 6.811209322073007e-05}
{"train_loss": 0.19858425855636597, "global_step": 104006, "epoch": 1168, "lr": 6.811155285622311e-05}
{"train_loss": 0.11267996579408646, "global_step": 104007, "epoch": 1168, "lr": 6.811101248928124e-05}
{"train_loss": 0.11154433339834213, "global_step": 104008, "epoch": 1168, "lr": 6.811047211990458e-05}
{"train_loss": 0.17612119019031525, "global_step": 104009, "epoch": 1168, "lr": 6.810993174809318e-05}
{"train_loss": 0.20417965948581696, "global_step": 104010, "epoch": 1168, "lr": 6.810939137384713e-05}
{"train_loss": 0.1752643883228302, "global_step": 104011, "epoch": 1168, "lr": 6.810885099716649e-05}
{"train_loss": 0.1241089403629303, "global_step": 104012, "epoch": 1168, "lr": 6.810831061805133e-05}
{"train_loss": 0.15318459272384644, "global_step": 104013, "epoch": 1168, "lr": 6.810777023650173e-05}
{"train_loss": 0.22944211959838867, "global_step": 104014, "epoch": 1168, "lr": 6.810722985251776e-05}
{"train_loss": 0.1823098510503769, "global_step": 104015, "epoch": 1168, "lr": 6.81066894660995e-05}
{"train_loss": 0.1375640630722046, "global_step": 104016, "epoch": 1168, "lr": 6.810614907724702e-05}
{"train_loss": 0.14671528339385986, "global_step": 104017, "epoch": 1168, "lr": 6.810560868596037e-05}
{"train_loss": 0.15125349164009094, "global_step": 104018, "epoch": 1168, "lr": 6.810506829223966e-05}
{"train_loss": 0.1746583878993988, "global_step": 104019, "epoch": 1168, "lr": 6.810452789608493e-05}
{"train_loss": 0.22318847477436066, "global_step": 104020, "epoch": 1168, "lr": 6.810398749749629e-05}
{"train_loss": 0.19364477694034576, "global_step": 104021, "epoch": 1168, "lr": 6.810344709647376e-05}
{"train_loss": 0.16583792865276337, "global_step": 104022, "epoch": 1168, "lr": 6.810290669301745e-05}
{"train_loss": 0.1667255014181137, "global_step": 104023, "epoch": 1168, "lr": 6.810236628712744e-05}
{"train_loss": 0.18494147062301636, "global_step": 104024, "epoch": 1168, "lr": 6.810182587880379e-05}
{"train_loss": 0.15116940438747406, "global_step": 104025, "epoch": 1168, "lr": 6.810128546804654e-05}
{"train_loss": 0.1353539377450943, "global_step": 104026, "epoch": 1168, "lr": 6.810074505485583e-05}
{"train_loss": 0.10315673053264618, "global_step": 104027, "epoch": 1168, "lr": 6.810020463923167e-05}
{"train_loss": 0.199445903301239, "global_step": 104028, "epoch": 1168, "lr": 6.809966422117416e-05}
{"train_loss": 0.0897345170378685, "global_step": 104029, "epoch": 1168, "lr": 6.809912380068337e-05}
{"train_loss": 0.15063171088695526, "global_step": 104030, "epoch": 1168, "lr": 6.809858337775939e-05}
{"train_loss": 0.1827939748764038, "global_step": 104031, "epoch": 1168, "lr": 6.809804295240227e-05}
{"train_loss": 0.24690847098827362, "global_step": 104032, "epoch": 1168, "lr": 6.809750252461206e-05}
{"train_loss": 0.1769055724143982, "global_step": 104033, "epoch": 1168, "lr": 6.809696209438889e-05}
{"train_loss": 0.11508752405643463, "global_step": 104034, "epoch": 1168, "lr": 6.809642166173281e-05}
{"train_loss": 0.182334765791893, "global_step": 104035, "epoch": 1168, "lr": 6.809588122664388e-05}
{"train_loss": 0.11959072202444077, "global_step": 104036, "epoch": 1168, "lr": 6.809534078912216e-05}
{"train_loss": 0.12531720101833344, "global_step": 104037, "epoch": 1168, "lr": 6.809480034916776e-05}
{"train_loss": 0.12515561282634735, "global_step": 104038, "epoch": 1168, "lr": 6.809425990678074e-05}
{"train_loss": 0.13660167157649994, "global_step": 104039, "epoch": 1168, "lr": 6.809371946196116e-05}
{"train_loss": 0.1567246188106162, "global_step": 104040, "epoch": 1168, "lr": 6.80931790147091e-05, "val_loss": 4.873259544372559}
{"train_loss": 0.2514501214027405, "global_step": 104041, "epoch": 1169, "lr": 6.809263856502464e-05}
{"train_loss": 0.15345177054405212, "global_step": 104042, "epoch": 1169, "lr": 6.809209811290782e-05}
{"train_loss": 0.19176709651947021, "global_step": 104043, "epoch": 1169, "lr": 6.809155765835876e-05}
{"train_loss": 0.16585201025009155, "global_step": 104044, "epoch": 1169, "lr": 6.809101720137751e-05}
{"train_loss": 0.13756728172302246, "global_step": 104045, "epoch": 1169, "lr": 6.809047674196414e-05}
{"train_loss": 0.2747194468975067, "global_step": 104046, "epoch": 1169, "lr": 6.808993628011872e-05}
{"train_loss": 0.1181793212890625, "global_step": 104047, "epoch": 1169, "lr": 6.808939581584133e-05}
{"train_loss": 0.13624078035354614, "global_step": 104048, "epoch": 1169, "lr": 6.808885534913205e-05}
{"train_loss": 0.12265678495168686, "global_step": 104049, "epoch": 1169, "lr": 6.808831487999094e-05}
{"train_loss": 0.09030228853225708, "global_step": 104050, "epoch": 1169, "lr": 6.808777440841808e-05}
{"train_loss": 0.1190214455127716, "global_step": 104051, "epoch": 1169, "lr": 6.808723393441352e-05}
{"train_loss": 0.17079058289527893, "global_step": 104052, "epoch": 1169, "lr": 6.808669345797737e-05}
{"train_loss": 0.17981769144535065, "global_step": 104053, "epoch": 1169, "lr": 6.808615297910967e-05}
{"train_loss": 0.15089119970798492, "global_step": 104054, "epoch": 1169, "lr": 6.808561249781052e-05}
{"train_loss": 0.18349909782409668, "global_step": 104055, "epoch": 1169, "lr": 6.808507201407998e-05}
{"train_loss": 0.17889246344566345, "global_step": 104056, "epoch": 1169, "lr": 6.808453152791812e-05}
{"train_loss": 0.1471766084432602, "global_step": 104057, "epoch": 1169, "lr": 6.808399103932502e-05}
{"train_loss": 0.18237072229385376, "global_step": 104058, "epoch": 1169, "lr": 6.808345054830074e-05}
{"train_loss": 0.1188838928937912, "global_step": 104059, "epoch": 1169, "lr": 6.808291005484537e-05}
{"train_loss": 0.18348845839500427, "global_step": 104060, "epoch": 1169, "lr": 6.808236955895897e-05}
{"train_loss": 0.12574651837348938, "global_step": 104061, "epoch": 1169, "lr": 6.80818290606416e-05}
{"train_loss": 0.1041584387421608, "global_step": 104062, "epoch": 1169, "lr": 6.808128855989336e-05}
{"train_loss": 0.18692445755004883, "global_step": 104063, "epoch": 1169, "lr": 6.808074805671432e-05}
{"train_loss": 0.07814987748861313, "global_step": 104064, "epoch": 1169, "lr": 6.808020755110453e-05}
{"train_loss": 0.19416363537311554, "global_step": 104065, "epoch": 1169, "lr": 6.807966704306409e-05}
{"train_loss": 0.19956889748573303, "global_step": 104066, "epoch": 1169, "lr": 6.807912653259305e-05}
{"train_loss": 0.14627622067928314, "global_step": 104067, "epoch": 1169, "lr": 6.80785860196915e-05}
{"train_loss": 0.22115586698055267, "global_step": 104068, "epoch": 1169, "lr": 6.807804550435948e-05}
{"train_loss": 0.27382907271385193, "global_step": 104069, "epoch": 1169, "lr": 6.807750498659711e-05}
{"train_loss": 0.12725426256656647, "global_step": 104070, "epoch": 1169, "lr": 6.807696446640444e-05}
{"train_loss": 0.21106663346290588, "global_step": 104071, "epoch": 1169, "lr": 6.807642394378153e-05}
{"train_loss": 0.13413351774215698, "global_step": 104072, "epoch": 1169, "lr": 6.807588341872848e-05}
{"train_loss": 0.12404990196228027, "global_step": 104073, "epoch": 1169, "lr": 6.807534289124533e-05}
{"train_loss": 0.1587965339422226, "global_step": 104074, "epoch": 1169, "lr": 6.807480236133219e-05}
{"train_loss": 0.15038558840751648, "global_step": 104075, "epoch": 1169, "lr": 6.80742618289891e-05}
{"train_loss": 0.10595273971557617, "global_step": 104076, "epoch": 1169, "lr": 6.807372129421614e-05}
{"train_loss": 0.1508452147245407, "global_step": 104077, "epoch": 1169, "lr": 6.807318075701341e-05}
{"train_loss": 0.1501861810684204, "global_step": 104078, "epoch": 1169, "lr": 6.807264021738095e-05}
{"train_loss": 0.3033338487148285, "global_step": 104079, "epoch": 1169, "lr": 6.807209967531883e-05}
{"train_loss": 0.17618785798549652, "global_step": 104080, "epoch": 1169, "lr": 6.807155913082716e-05}
{"train_loss": 0.1387159824371338, "global_step": 104081, "epoch": 1169, "lr": 6.807101858390597e-05}
{"train_loss": 0.22716385126113892, "global_step": 104082, "epoch": 1169, "lr": 6.807047803455536e-05}
{"train_loss": 0.1535029411315918, "global_step": 104083, "epoch": 1169, "lr": 6.80699374827754e-05}
{"train_loss": 0.21559792757034302, "global_step": 104084, "epoch": 1169, "lr": 6.806939692856615e-05}
{"train_loss": 0.2245398908853531, "global_step": 104085, "epoch": 1169, "lr": 6.806885637192769e-05}
{"train_loss": 0.1393083930015564, "global_step": 104086, "epoch": 1169, "lr": 6.806831581286009e-05}
{"train_loss": 0.15117458999156952, "global_step": 104087, "epoch": 1169, "lr": 6.806777525136343e-05}
{"train_loss": 0.11929547786712646, "global_step": 104088, "epoch": 1169, "lr": 6.806723468743779e-05}
{"train_loss": 0.09587028622627258, "global_step": 104089, "epoch": 1169, "lr": 6.806669412108321e-05}
{"train_loss": 0.17464721202850342, "global_step": 104090, "epoch": 1169, "lr": 6.806615355229979e-05}
{"train_loss": 0.16830678284168243, "global_step": 104091, "epoch": 1169, "lr": 6.806561298108761e-05}
{"train_loss": 0.16271135210990906, "global_step": 104092, "epoch": 1169, "lr": 6.80650724074467e-05}
{"train_loss": 0.21046391129493713, "global_step": 104093, "epoch": 1169, "lr": 6.806453183137717e-05}
{"train_loss": 0.1211545392870903, "global_step": 104094, "epoch": 1169, "lr": 6.80639912528791e-05}
{"train_loss": 0.17077021300792694, "global_step": 104095, "epoch": 1169, "lr": 6.806345067195253e-05}
{"train_loss": 0.07519214600324631, "global_step": 104096, "epoch": 1169, "lr": 6.806291008859755e-05}
{"train_loss": 0.16900241374969482, "global_step": 104097, "epoch": 1169, "lr": 6.806236950281425e-05}
{"train_loss": 0.2154417634010315, "global_step": 104098, "epoch": 1169, "lr": 6.806182891460266e-05}
{"train_loss": 0.23728519678115845, "global_step": 104099, "epoch": 1169, "lr": 6.80612883239629e-05}
{"train_loss": 0.1497684270143509, "global_step": 104100, "epoch": 1169, "lr": 6.806074773089502e-05}
{"train_loss": 0.20809835195541382, "global_step": 104101, "epoch": 1169, "lr": 6.806020713539907e-05}
{"train_loss": 0.136193186044693, "global_step": 104102, "epoch": 1169, "lr": 6.805966653747515e-05}
{"train_loss": 0.12728923559188843, "global_step": 104103, "epoch": 1169, "lr": 6.805912593712334e-05}
{"train_loss": 0.11505201458930969, "global_step": 104104, "epoch": 1169, "lr": 6.805858533434369e-05}
{"train_loss": 0.24018459022045135, "global_step": 104105, "epoch": 1169, "lr": 6.805804472913629e-05}
{"train_loss": 0.14717665314674377, "global_step": 104106, "epoch": 1169, "lr": 6.80575041215012e-05}
{"train_loss": 0.24216774106025696, "global_step": 104107, "epoch": 1169, "lr": 6.805696351143851e-05}
{"train_loss": 0.1914324015378952, "global_step": 104108, "epoch": 1169, "lr": 6.805642289894828e-05}
{"train_loss": 0.09682606160640717, "global_step": 104109, "epoch": 1169, "lr": 6.805588228403058e-05}
{"train_loss": 0.12272803485393524, "global_step": 104110, "epoch": 1169, "lr": 6.805534166668547e-05}
{"train_loss": 0.1896696388721466, "global_step": 104111, "epoch": 1169, "lr": 6.805480104691307e-05}
{"train_loss": 0.18800942599773407, "global_step": 104112, "epoch": 1169, "lr": 6.80542604247134e-05}
{"train_loss": 0.22230641543865204, "global_step": 104113, "epoch": 1169, "lr": 6.805371980008657e-05}
{"train_loss": 0.18965166807174683, "global_step": 104114, "epoch": 1169, "lr": 6.805317917303263e-05}
{"train_loss": 0.09554916620254517, "global_step": 104115, "epoch": 1169, "lr": 6.805263854355166e-05}
{"train_loss": 0.15229906141757965, "global_step": 104116, "epoch": 1169, "lr": 6.805209791164374e-05}
{"train_loss": 0.11073221266269684, "global_step": 104117, "epoch": 1169, "lr": 6.805155727730892e-05}
{"train_loss": 0.13246463239192963, "global_step": 104118, "epoch": 1169, "lr": 6.805101664054729e-05}
{"train_loss": 0.1462327539920807, "global_step": 104119, "epoch": 1169, "lr": 6.805047600135894e-05}
{"train_loss": 0.1353060007095337, "global_step": 104120, "epoch": 1169, "lr": 6.80499353597439e-05}
{"train_loss": 0.1250896006822586, "global_step": 104121, "epoch": 1169, "lr": 6.804939471570227e-05}
{"train_loss": 0.2023714780807495, "global_step": 104122, "epoch": 1169, "lr": 6.804885406923414e-05}
{"train_loss": 0.2383977472782135, "global_step": 104123, "epoch": 1169, "lr": 6.804831342033954e-05}
{"train_loss": 0.1959412842988968, "global_step": 104124, "epoch": 1169, "lr": 6.804777276901856e-05}
{"train_loss": 0.17955031991004944, "global_step": 104125, "epoch": 1169, "lr": 6.804723211527129e-05}
{"train_loss": 0.2529762387275696, "global_step": 104126, "epoch": 1169, "lr": 6.80466914590978e-05}
{"train_loss": 0.16463835537433624, "global_step": 104127, "epoch": 1169, "lr": 6.804615080049814e-05}
{"train_loss": 0.14013037085533142, "global_step": 104128, "epoch": 1169, "lr": 6.80456101394724e-05}
{"train_loss": 0.16602145913946495, "global_step": 104129, "epoch": 1169, "lr": 6.804506947602063e-05, "val_loss": 4.713413715362549}
{"train_loss": 0.14191965758800507, "global_step": 104130, "epoch": 1170, "lr": 6.804452881014293e-05}
{"train_loss": 0.14119799435138702, "global_step": 104131, "epoch": 1170, "lr": 6.804398814183938e-05}
{"train_loss": 0.14798998832702637, "global_step": 104132, "epoch": 1170, "lr": 6.804344747111002e-05}
{"train_loss": 0.16731710731983185, "global_step": 104133, "epoch": 1170, "lr": 6.804290679795493e-05}
{"train_loss": 0.12154670804738998, "global_step": 104134, "epoch": 1170, "lr": 6.80423661223742e-05}
{"train_loss": 0.22444042563438416, "global_step": 104135, "epoch": 1170, "lr": 6.80418254443679e-05}
{"train_loss": 0.15416786074638367, "global_step": 104136, "epoch": 1170, "lr": 6.80412847639361e-05}
{"train_loss": 0.20580630004405975, "global_step": 104137, "epoch": 1170, "lr": 6.804074408107886e-05}
{"train_loss": 0.1625060886144638, "global_step": 104138, "epoch": 1170, "lr": 6.804020339579625e-05}
{"train_loss": 0.19784343242645264, "global_step": 104139, "epoch": 1170, "lr": 6.803966270808837e-05}
{"train_loss": 0.11856725066900253, "global_step": 104140, "epoch": 1170, "lr": 6.803912201795529e-05}
{"train_loss": 0.20760810375213623, "global_step": 104141, "epoch": 1170, "lr": 6.803858132539704e-05}
{"train_loss": 0.17247046530246735, "global_step": 104142, "epoch": 1170, "lr": 6.803804063041372e-05}
{"train_loss": 0.21756082773208618, "global_step": 104143, "epoch": 1170, "lr": 6.803749993300543e-05}
{"train_loss": 0.14057442545890808, "global_step": 104144, "epoch": 1170, "lr": 6.803695923317222e-05}
{"train_loss": 0.14360137283802032, "global_step": 104145, "epoch": 1170, "lr": 6.803641853091413e-05}
{"train_loss": 0.1293765753507614, "global_step": 104146, "epoch": 1170, "lr": 6.803587782623131e-05}
{"train_loss": 0.19708815217018127, "global_step": 104147, "epoch": 1170, "lr": 6.803533711912375e-05}
{"train_loss": 0.1281261146068573, "global_step": 104148, "epoch": 1170, "lr": 6.803479640959157e-05}
{"train_loss": 0.13224908709526062, "global_step": 104149, "epoch": 1170, "lr": 6.803425569763482e-05}
{"train_loss": 0.2202862948179245, "global_step": 104150, "epoch": 1170, "lr": 6.803371498325361e-05}
{"train_loss": 0.175113245844841, "global_step": 104151, "epoch": 1170, "lr": 6.803317426644797e-05}
{"train_loss": 0.1259647160768509, "global_step": 104152, "epoch": 1170, "lr": 6.803263354721799e-05}
{"train_loss": 0.13322463631629944, "global_step": 104153, "epoch": 1170, "lr": 6.803209282556374e-05}
{"train_loss": 0.16687360405921936, "global_step": 104154, "epoch": 1170, "lr": 6.80315521014853e-05}
{"train_loss": 0.17330527305603027, "global_step": 104155, "epoch": 1170, "lr": 6.803101137498273e-05}
{"train_loss": 0.12089641392230988, "global_step": 104156, "epoch": 1170, "lr": 6.803047064605611e-05}
{"train_loss": 0.09935766458511353, "global_step": 104157, "epoch": 1170, "lr": 6.802992991470552e-05}
{"train_loss": 0.12461566925048828, "global_step": 104158, "epoch": 1170, "lr": 6.802938918093103e-05}
{"train_loss": 0.1156652569770813, "global_step": 104159, "epoch": 1170, "lr": 6.80288484447327e-05}
{"train_loss": 0.21913793683052063, "global_step": 104160, "epoch": 1170, "lr": 6.802830770611061e-05}
{"train_loss": 0.17390653491020203, "global_step": 104161, "epoch": 1170, "lr": 6.802776696506483e-05}
{"train_loss": 0.21505402028560638, "global_step": 104162, "epoch": 1170, "lr": 6.802722622159546e-05}
{"train_loss": 0.15073607861995697, "global_step": 104163, "epoch": 1170, "lr": 6.802668547570253e-05}
{"train_loss": 0.14324605464935303, "global_step": 104164, "epoch": 1170, "lr": 6.802614472738614e-05}
{"train_loss": 0.2408095747232437, "global_step": 104165, "epoch": 1170, "lr": 6.802560397664635e-05}
{"train_loss": 0.0942019373178482, "global_step": 104166, "epoch": 1170, "lr": 6.802506322348324e-05}
{"train_loss": 0.09619816392660141, "global_step": 104167, "epoch": 1170, "lr": 6.802452246789688e-05}
{"train_loss": 0.21540731191635132, "global_step": 104168, "epoch": 1170, "lr": 6.802398170988733e-05}
{"train_loss": 0.07721513509750366, "global_step": 104169, "epoch": 1170, "lr": 6.802344094945468e-05}
{"train_loss": 0.23987697064876556, "global_step": 104170, "epoch": 1170, "lr": 6.8022900186599e-05}
{"train_loss": 0.11961743235588074, "global_step": 104171, "epoch": 1170, "lr": 6.802235942132036e-05}
{"train_loss": 0.11749695986509323, "global_step": 104172, "epoch": 1170, "lr": 6.802181865361886e-05}
{"train_loss": 0.20856983959674835, "global_step": 104173, "epoch": 1170, "lr": 6.802127788349451e-05}
{"train_loss": 0.20589730143547058, "global_step": 104174, "epoch": 1170, "lr": 6.802073711094742e-05}
{"train_loss": 0.1869993507862091, "global_step": 104175, "epoch": 1170, "lr": 6.80201963359777e-05}
{"train_loss": 0.17535336315631866, "global_step": 104176, "epoch": 1170, "lr": 6.801965555858535e-05}
{"train_loss": 0.09905347228050232, "global_step": 104177, "epoch": 1170, "lr": 6.801911477877047e-05}
{"train_loss": 0.16397473216056824, "global_step": 104178, "epoch": 1170, "lr": 6.801857399653316e-05}
{"train_loss": 0.14004455506801605, "global_step": 104179, "epoch": 1170, "lr": 6.801803321187345e-05}
{"train_loss": 0.14674043655395508, "global_step": 104180, "epoch": 1170, "lr": 6.801749242479146e-05}
{"train_loss": 0.12201450765132904, "global_step": 104181, "epoch": 1170, "lr": 6.801695163528724e-05}
{"train_loss": 0.1498422771692276, "global_step": 104182, "epoch": 1170, "lr": 6.801641084336084e-05}
{"train_loss": 0.13867364823818207, "global_step": 104183, "epoch": 1170, "lr": 6.801587004901236e-05}
{"train_loss": 0.18230539560317993, "global_step": 104184, "epoch": 1170, "lr": 6.801532925224189e-05}
{"train_loss": 0.13003991544246674, "global_step": 104185, "epoch": 1170, "lr": 6.801478845304945e-05}
{"train_loss": 0.1488252729177475, "global_step": 104186, "epoch": 1170, "lr": 6.801424765143514e-05}
{"train_loss": 0.1626497358083725, "global_step": 104187, "epoch": 1170, "lr": 6.801370684739905e-05}
{"train_loss": 0.15335306525230408, "global_step": 104188, "epoch": 1170, "lr": 6.801316604094124e-05}
{"train_loss": 0.1970994621515274, "global_step": 104189, "epoch": 1170, "lr": 6.801262523206176e-05}
{"train_loss": 0.07888133823871613, "global_step": 104190, "epoch": 1170, "lr": 6.801208442076072e-05}
{"train_loss": 0.11207355558872223, "global_step": 104191, "epoch": 1170, "lr": 6.801154360703816e-05}
{"train_loss": 0.17060646414756775, "global_step": 104192, "epoch": 1170, "lr": 6.801100279089418e-05}
{"train_loss": 0.1116311177611351, "global_step": 104193, "epoch": 1170, "lr": 6.801046197232882e-05}
{"train_loss": 0.14167751371860504, "global_step": 104194, "epoch": 1170, "lr": 6.80099211513422e-05}
{"train_loss": 0.12727566063404083, "global_step": 104195, "epoch": 1170, "lr": 6.800938032793434e-05}
{"train_loss": 0.17875587940216064, "global_step": 104196, "epoch": 1170, "lr": 6.800883950210535e-05}
{"train_loss": 0.2613951563835144, "global_step": 104197, "epoch": 1170, "lr": 6.800829867385529e-05}
{"train_loss": 0.08865953981876373, "global_step": 104198, "epoch": 1170, "lr": 6.800775784318423e-05}
{"train_loss": 0.18894414603710175, "global_step": 104199, "epoch": 1170, "lr": 6.800721701009226e-05}
{"train_loss": 0.17278556525707245, "global_step": 104200, "epoch": 1170, "lr": 6.800667617457942e-05}
{"train_loss": 0.12155457586050034, "global_step": 104201, "epoch": 1170, "lr": 6.800613533664582e-05}
{"train_loss": 0.1508520543575287, "global_step": 104202, "epoch": 1170, "lr": 6.800559449629149e-05}
{"train_loss": 0.16313119232654572, "global_step": 104203, "epoch": 1170, "lr": 6.800505365351655e-05}
{"train_loss": 0.14859648048877716, "global_step": 104204, "epoch": 1170, "lr": 6.800451280832106e-05}
{"train_loss": 0.15508127212524414, "global_step": 104205, "epoch": 1170, "lr": 6.800397196070506e-05}
{"train_loss": 0.16068622469902039, "global_step": 104206, "epoch": 1170, "lr": 6.800343111066865e-05}
{"train_loss": 0.12617166340351105, "global_step": 104207, "epoch": 1170, "lr": 6.800289025821191e-05}
{"train_loss": 0.12345824390649796, "global_step": 104208, "epoch": 1170, "lr": 6.800234940333488e-05}
{"train_loss": 0.1658417284488678, "global_step": 104209, "epoch": 1170, "lr": 6.800180854603767e-05}
{"train_loss": 0.17909741401672363, "global_step": 104210, "epoch": 1170, "lr": 6.800126768632034e-05}
{"train_loss": 0.18377801775932312, "global_step": 104211, "epoch": 1170, "lr": 6.800072682418295e-05}
{"train_loss": 0.1339595913887024, "global_step": 104212, "epoch": 1170, "lr": 6.800018595962557e-05}
{"train_loss": 0.12225361168384552, "global_step": 104213, "epoch": 1170, "lr": 6.79996450926483e-05}
{"train_loss": 0.17078471183776855, "global_step": 104214, "epoch": 1170, "lr": 6.79991042232512e-05}
{"train_loss": 0.11649373173713684, "global_step": 104215, "epoch": 1170, "lr": 6.799856335143432e-05}
{"train_loss": 0.13430878520011902, "global_step": 104216, "epoch": 1170, "lr": 6.799802247719776e-05}
{"train_loss": 0.14494279026985168, "global_step": 104217, "epoch": 1170, "lr": 6.799748160054158e-05}
{"train_loss": 0.15605450337857343, "global_step": 104218, "epoch": 1170, "lr": 6.799694072146587e-05, "val_loss": 4.763025760650635, "train_action_mse_error": 19.653823852539062}
{"train_loss": 0.12446130067110062, "global_step": 104219, "epoch": 1171, "lr": 6.799639983997071e-05}
{"train_loss": 0.1887659877538681, "global_step": 104220, "epoch": 1171, "lr": 6.799585895605612e-05}
{"train_loss": 0.12394051253795624, "global_step": 104221, "epoch": 1171, "lr": 6.799531806972223e-05}
{"train_loss": 0.18131162226200104, "global_step": 104222, "epoch": 1171, "lr": 6.799477718096908e-05}
{"train_loss": 0.20241476595401764, "global_step": 104223, "epoch": 1171, "lr": 6.799423628979676e-05}
{"train_loss": 0.11432939022779465, "global_step": 104224, "epoch": 1171, "lr": 6.799369539620531e-05}
{"train_loss": 0.12159664928913116, "global_step": 104225, "epoch": 1171, "lr": 6.799315450019485e-05}
{"train_loss": 0.19478759169578552, "global_step": 104226, "epoch": 1171, "lr": 6.799261360176541e-05}
{"train_loss": 0.10933829843997955, "global_step": 104227, "epoch": 1171, "lr": 6.79920727009171e-05}
{"train_loss": 0.22390688955783844, "global_step": 104228, "epoch": 1171, "lr": 6.799153179764996e-05}
{"train_loss": 0.15449723601341248, "global_step": 104229, "epoch": 1171, "lr": 6.79909908919641e-05}
{"train_loss": 0.19631826877593994, "global_step": 104230, "epoch": 1171, "lr": 6.799044998385955e-05}
{"train_loss": 0.14361044764518738, "global_step": 104231, "epoch": 1171, "lr": 6.79899090733364e-05}
{"train_loss": 0.19865524768829346, "global_step": 104232, "epoch": 1171, "lr": 6.798936816039474e-05}
{"train_loss": 0.108897864818573, "global_step": 104233, "epoch": 1171, "lr": 6.798882724503461e-05}
{"train_loss": 0.19729658961296082, "global_step": 104234, "epoch": 1171, "lr": 6.798828632725611e-05}
{"train_loss": 0.1596512794494629, "global_step": 104235, "epoch": 1171, "lr": 6.798774540705932e-05}
{"train_loss": 0.1411045491695404, "global_step": 104236, "epoch": 1171, "lr": 6.79872044844443e-05}
{"train_loss": 0.15413938462734222, "global_step": 104237, "epoch": 1171, "lr": 6.79866635594111e-05}
{"train_loss": 0.1975998729467392, "global_step": 104238, "epoch": 1171, "lr": 6.79861226319598e-05}
{"train_loss": 0.17735934257507324, "global_step": 104239, "epoch": 1171, "lr": 6.79855817020905e-05}
{"train_loss": 0.13139832019805908, "global_step": 104240, "epoch": 1171, "lr": 6.798504076980327e-05}
{"train_loss": 0.1389501541852951, "global_step": 104241, "epoch": 1171, "lr": 6.798449983509816e-05}
{"train_loss": 0.1492478847503662, "global_step": 104242, "epoch": 1171, "lr": 6.798395889797527e-05}
{"train_loss": 0.10012713819742203, "global_step": 104243, "epoch": 1171, "lr": 6.798341795843464e-05}
{"train_loss": 0.12102101743221283, "global_step": 104244, "epoch": 1171, "lr": 6.798287701647636e-05}
{"train_loss": 0.12281538546085358, "global_step": 104245, "epoch": 1171, "lr": 6.798233607210049e-05}
{"train_loss": 0.17914815247058868, "global_step": 104246, "epoch": 1171, "lr": 6.798179512530713e-05}
{"train_loss": 0.10607446730136871, "global_step": 104247, "epoch": 1171, "lr": 6.798125417609633e-05}
{"train_loss": 0.06730291247367859, "global_step": 104248, "epoch": 1171, "lr": 6.798071322446817e-05}
{"train_loss": 0.11579293012619019, "global_step": 104249, "epoch": 1171, "lr": 6.798017227042272e-05}
{"train_loss": 0.2453242540359497, "global_step": 104250, "epoch": 1171, "lr": 6.797963131396007e-05}
{"train_loss": 0.14506538212299347, "global_step": 104251, "epoch": 1171, "lr": 6.797909035508025e-05}
{"train_loss": 0.14025981724262238, "global_step": 104252, "epoch": 1171, "lr": 6.797854939378339e-05}
{"train_loss": 0.17659442126750946, "global_step": 104253, "epoch": 1171, "lr": 6.79780084300695e-05}
{"train_loss": 0.22147829830646515, "global_step": 104254, "epoch": 1171, "lr": 6.797746746393871e-05}
{"train_loss": 0.08568808436393738, "global_step": 104255, "epoch": 1171, "lr": 6.797692649539107e-05}
{"train_loss": 0.1688050925731659, "global_step": 104256, "epoch": 1171, "lr": 6.797638552442665e-05}
{"train_loss": 0.24617618322372437, "global_step": 104257, "epoch": 1171, "lr": 6.797584455104551e-05}
{"train_loss": 0.12538723647594452, "global_step": 104258, "epoch": 1171, "lr": 6.797530357524776e-05}
{"train_loss": 0.12869659066200256, "global_step": 104259, "epoch": 1171, "lr": 6.797476259703342e-05}
{"train_loss": 0.15692932903766632, "global_step": 104260, "epoch": 1171, "lr": 6.797422161640261e-05}
{"train_loss": 0.10443852096796036, "global_step": 104261, "epoch": 1171, "lr": 6.797368063335539e-05}
{"train_loss": 0.09635751694440842, "global_step": 104262, "epoch": 1171, "lr": 6.797313964789182e-05}
{"train_loss": 0.2886759340763092, "global_step": 104263, "epoch": 1171, "lr": 6.797259866001199e-05}
{"train_loss": 0.1951976865530014, "global_step": 104264, "epoch": 1171, "lr": 6.797205766971595e-05}
{"train_loss": 0.0825665220618248, "global_step": 104265, "epoch": 1171, "lr": 6.79715166770038e-05}
{"train_loss": 0.18546819686889648, "global_step": 104266, "epoch": 1171, "lr": 6.797097568187559e-05}
{"train_loss": 0.15300090610980988, "global_step": 104267, "epoch": 1171, "lr": 6.797043468433139e-05}
{"train_loss": 0.12751956284046173, "global_step": 104268, "epoch": 1171, "lr": 6.79698936843713e-05}
{"train_loss": 0.09780652821063995, "global_step": 104269, "epoch": 1171, "lr": 6.796935268199536e-05}
{"train_loss": 0.09599077701568604, "global_step": 104270, "epoch": 1171, "lr": 6.796881167720366e-05}
{"train_loss": 0.0884445458650589, "global_step": 104271, "epoch": 1171, "lr": 6.796827066999629e-05}
{"train_loss": 0.23543818295001984, "global_step": 104272, "epoch": 1171, "lr": 6.796772966037331e-05}
{"train_loss": 0.1404757797718048, "global_step": 104273, "epoch": 1171, "lr": 6.796718864833477e-05}
{"train_loss": 0.15954604744911194, "global_step": 104274, "epoch": 1171, "lr": 6.796664763388076e-05}
{"train_loss": 0.17750686407089233, "global_step": 104275, "epoch": 1171, "lr": 6.796610661701135e-05}
{"train_loss": 0.1352112591266632, "global_step": 104276, "epoch": 1171, "lr": 6.796556559772664e-05}
{"train_loss": 0.14225336909294128, "global_step": 104277, "epoch": 1171, "lr": 6.796502457602666e-05}
{"train_loss": 0.12244927883148193, "global_step": 104278, "epoch": 1171, "lr": 6.796448355191151e-05}
{"train_loss": 0.13366283476352692, "global_step": 104279, "epoch": 1171, "lr": 6.796394252538126e-05}
{"train_loss": 0.14960846304893494, "global_step": 104280, "epoch": 1171, "lr": 6.796340149643595e-05}
{"train_loss": 0.21295146644115448, "global_step": 104281, "epoch": 1171, "lr": 6.796286046507572e-05}
{"train_loss": 0.1726650595664978, "global_step": 104282, "epoch": 1171, "lr": 6.796231943130057e-05}
{"train_loss": 0.16040481626987457, "global_step": 104283, "epoch": 1171, "lr": 6.796177839511061e-05}
{"train_loss": 0.10728197544813156, "global_step": 104284, "epoch": 1171, "lr": 6.796123735650592e-05}
{"train_loss": 0.10280704498291016, "global_step": 104285, "epoch": 1171, "lr": 6.796069631548655e-05}
{"train_loss": 0.17144297063350677, "global_step": 104286, "epoch": 1171, "lr": 6.796015527205259e-05}
{"train_loss": 0.14846792817115784, "global_step": 104287, "epoch": 1171, "lr": 6.795961422620408e-05}
{"train_loss": 0.1138194352388382, "global_step": 104288, "epoch": 1171, "lr": 6.795907317794114e-05}
{"train_loss": 0.22242571413516998, "global_step": 104289, "epoch": 1171, "lr": 6.795853212726382e-05}
{"train_loss": 0.1804024577140808, "global_step": 104290, "epoch": 1171, "lr": 6.79579910741722e-05}
{"train_loss": 0.1432558000087738, "global_step": 104291, "epoch": 1171, "lr": 6.795745001866633e-05}
{"train_loss": 0.21126241981983185, "global_step": 104292, "epoch": 1171, "lr": 6.795690896074632e-05}
{"train_loss": 0.2650115489959717, "global_step": 104293, "epoch": 1171, "lr": 6.795636790041223e-05}
{"train_loss": 0.1507806032896042, "global_step": 104294, "epoch": 1171, "lr": 6.79558268376641e-05}
{"train_loss": 0.1682111918926239, "global_step": 104295, "epoch": 1171, "lr": 6.795528577250203e-05}
{"train_loss": 0.1760854572057724, "global_step": 104296, "epoch": 1171, "lr": 6.795474470492611e-05}
{"train_loss": 0.16486339271068573, "global_step": 104297, "epoch": 1171, "lr": 6.795420363493638e-05}
{"train_loss": 0.12190084159374237, "global_step": 104298, "epoch": 1171, "lr": 6.795366256253293e-05}
{"train_loss": 0.14450009167194366, "global_step": 104299, "epoch": 1171, "lr": 6.795312148771582e-05}
{"train_loss": 0.08831878751516342, "global_step": 104300, "epoch": 1171, "lr": 6.795258041048515e-05}
{"train_loss": 0.13128994405269623, "global_step": 104301, "epoch": 1171, "lr": 6.795203933084097e-05}
{"train_loss": 0.18714426457881927, "global_step": 104302, "epoch": 1171, "lr": 6.795149824878335e-05}
{"train_loss": 0.1490279585123062, "global_step": 104303, "epoch": 1171, "lr": 6.795095716431237e-05}
{"train_loss": 0.23281607031822205, "global_step": 104304, "epoch": 1171, "lr": 6.79504160774281e-05}
{"train_loss": 0.16956128180027008, "global_step": 104305, "epoch": 1171, "lr": 6.794987498813061e-05}
{"train_loss": 0.23870410025119781, "global_step": 104306, "epoch": 1171, "lr": 6.794933389641999e-05}
{"train_loss": 0.15619329940736965, "global_step": 104307, "epoch": 1171, "lr": 6.794879280229629e-05, "val_loss": 4.834965229034424}
{"train_loss": 0.25961387157440186, "global_step": 104308, "epoch": 1172, "lr": 6.794825170575961e-05}
{"train_loss": 0.21042366325855255, "global_step": 104309, "epoch": 1172, "lr": 6.794771060680999e-05}
{"train_loss": 0.12284787744283676, "global_step": 104310, "epoch": 1172, "lr": 6.794716950544753e-05}
{"train_loss": 0.13309484720230103, "global_step": 104311, "epoch": 1172, "lr": 6.794662840167228e-05}
{"train_loss": 0.1480574756860733, "global_step": 104312, "epoch": 1172, "lr": 6.794608729548433e-05}
{"train_loss": 0.14240306615829468, "global_step": 104313, "epoch": 1172, "lr": 6.794554618688375e-05}
{"train_loss": 0.15806305408477783, "global_step": 104314, "epoch": 1172, "lr": 6.794500507587062e-05}
{"train_loss": 0.08920332044363022, "global_step": 104315, "epoch": 1172, "lr": 6.794446396244499e-05}
{"train_loss": 0.16383427381515503, "global_step": 104316, "epoch": 1172, "lr": 6.794392284660695e-05}
{"train_loss": 0.13539844751358032, "global_step": 104317, "epoch": 1172, "lr": 6.794338172835656e-05}
{"train_loss": 0.16089773178100586, "global_step": 104318, "epoch": 1172, "lr": 6.794284060769391e-05}
{"train_loss": 0.11704801023006439, "global_step": 104319, "epoch": 1172, "lr": 6.794229948461905e-05}
{"train_loss": 0.12508966028690338, "global_step": 104320, "epoch": 1172, "lr": 6.79417583591321e-05}
{"train_loss": 0.18268652260303497, "global_step": 104321, "epoch": 1172, "lr": 6.794121723123308e-05}
{"train_loss": 0.2014315128326416, "global_step": 104322, "epoch": 1172, "lr": 6.794067610092208e-05}
{"train_loss": 0.17638710141181946, "global_step": 104323, "epoch": 1172, "lr": 6.794013496819916e-05}
{"train_loss": 0.19457121193408966, "global_step": 104324, "epoch": 1172, "lr": 6.793959383306442e-05}
{"train_loss": 0.19178512692451477, "global_step": 104325, "epoch": 1172, "lr": 6.793905269551791e-05}
{"train_loss": 0.2626873254776001, "global_step": 104326, "epoch": 1172, "lr": 6.793851155555973e-05}
{"train_loss": 0.15820488333702087, "global_step": 104327, "epoch": 1172, "lr": 6.793797041318993e-05}
{"train_loss": 0.1606186181306839, "global_step": 104328, "epoch": 1172, "lr": 6.793742926840859e-05}
{"train_loss": 0.1155499592423439, "global_step": 104329, "epoch": 1172, "lr": 6.793688812121578e-05}
{"train_loss": 0.10132605582475662, "global_step": 104330, "epoch": 1172, "lr": 6.793634697161158e-05}
{"train_loss": 0.19272452592849731, "global_step": 104331, "epoch": 1172, "lr": 6.793580581959605e-05}
{"train_loss": 0.1717677265405655, "global_step": 104332, "epoch": 1172, "lr": 6.793526466516926e-05}
{"train_loss": 0.17877288162708282, "global_step": 104333, "epoch": 1172, "lr": 6.793472350833131e-05}
{"train_loss": 0.14404870569705963, "global_step": 104334, "epoch": 1172, "lr": 6.793418234908225e-05}
{"train_loss": 0.1495279222726822, "global_step": 104335, "epoch": 1172, "lr": 6.793364118742214e-05}
{"train_loss": 0.20457279682159424, "global_step": 104336, "epoch": 1172, "lr": 6.79331000233511e-05}
{"train_loss": 0.13259512186050415, "global_step": 104337, "epoch": 1172, "lr": 6.793255885686916e-05}
{"train_loss": 0.15687957406044006, "global_step": 104338, "epoch": 1172, "lr": 6.79320176879764e-05}
{"train_loss": 0.12617212533950806, "global_step": 104339, "epoch": 1172, "lr": 6.79314765166729e-05}
{"train_loss": 0.1296258270740509, "global_step": 104340, "epoch": 1172, "lr": 6.793093534295875e-05}
{"train_loss": 0.21927979588508606, "global_step": 104341, "epoch": 1172, "lr": 6.793039416683399e-05}
{"train_loss": 0.09097582846879959, "global_step": 104342, "epoch": 1172, "lr": 6.79298529882987e-05}
{"train_loss": 0.1129714772105217, "global_step": 104343, "epoch": 1172, "lr": 6.792931180735296e-05}
{"train_loss": 0.16339895129203796, "global_step": 104344, "epoch": 1172, "lr": 6.792877062399686e-05}
{"train_loss": 0.16783803701400757, "global_step": 104345, "epoch": 1172, "lr": 6.792822943823044e-05}
{"train_loss": 0.16426751017570496, "global_step": 104346, "epoch": 1172, "lr": 6.792768825005379e-05}
{"train_loss": 0.107124462723732, "global_step": 104347, "epoch": 1172, "lr": 6.792714705946698e-05}
{"train_loss": 0.08097518980503082, "global_step": 104348, "epoch": 1172, "lr": 6.792660586647008e-05}
{"train_loss": 0.18926046788692474, "global_step": 104349, "epoch": 1172, "lr": 6.792606467106318e-05}
{"train_loss": 0.21229922771453857, "global_step": 104350, "epoch": 1172, "lr": 6.792552347324633e-05}
{"train_loss": 0.13667549192905426, "global_step": 104351, "epoch": 1172, "lr": 6.792498227301962e-05}
{"train_loss": 0.12344036996364594, "global_step": 104352, "epoch": 1172, "lr": 6.792444107038311e-05}
{"train_loss": 0.1384519636631012, "global_step": 104353, "epoch": 1172, "lr": 6.792389986533687e-05}
{"train_loss": 0.10413019359111786, "global_step": 104354, "epoch": 1172, "lr": 6.792335865788099e-05}
{"train_loss": 0.17475652694702148, "global_step": 104355, "epoch": 1172, "lr": 6.792281744801553e-05}
{"train_loss": 0.11083638668060303, "global_step": 104356, "epoch": 1172, "lr": 6.792227623574057e-05}
{"train_loss": 0.12557978928089142, "global_step": 104357, "epoch": 1172, "lr": 6.792173502105616e-05}
{"train_loss": 0.21268117427825928, "global_step": 104358, "epoch": 1172, "lr": 6.792119380396242e-05}
{"train_loss": 0.132592111825943, "global_step": 104359, "epoch": 1172, "lr": 6.792065258445937e-05}
{"train_loss": 0.17809903621673584, "global_step": 104360, "epoch": 1172, "lr": 6.792011136254711e-05}
{"train_loss": 0.14360304176807404, "global_step": 104361, "epoch": 1172, "lr": 6.791957013822572e-05}
{"train_loss": 0.20171909034252167, "global_step": 104362, "epoch": 1172, "lr": 6.791902891149525e-05}
{"train_loss": 0.12407581508159637, "global_step": 104363, "epoch": 1172, "lr": 6.79184876823558e-05}
{"train_loss": 0.10888306796550751, "global_step": 104364, "epoch": 1172, "lr": 6.791794645080741e-05}
{"train_loss": 0.272382527589798, "global_step": 104365, "epoch": 1172, "lr": 6.791740521685018e-05}
{"train_loss": 0.10424210876226425, "global_step": 104366, "epoch": 1172, "lr": 6.791686398048418e-05}
{"train_loss": 0.12319314479827881, "global_step": 104367, "epoch": 1172, "lr": 6.791632274170947e-05}
{"train_loss": 0.1850586086511612, "global_step": 104368, "epoch": 1172, "lr": 6.791578150052613e-05}
{"train_loss": 0.11439700424671173, "global_step": 104369, "epoch": 1172, "lr": 6.791524025693424e-05}
{"train_loss": 0.08371923118829727, "global_step": 104370, "epoch": 1172, "lr": 6.791469901093384e-05}
{"train_loss": 0.17397412657737732, "global_step": 104371, "epoch": 1172, "lr": 6.791415776252504e-05}
{"train_loss": 0.11700881272554398, "global_step": 104372, "epoch": 1172, "lr": 6.79136165117079e-05}
{"train_loss": 0.2071017026901245, "global_step": 104373, "epoch": 1172, "lr": 6.79130752584825e-05}
{"train_loss": 0.1887226402759552, "global_step": 104374, "epoch": 1172, "lr": 6.79125340028489e-05}
{"train_loss": 0.1465005874633789, "global_step": 104375, "epoch": 1172, "lr": 6.791199274480717e-05}
{"train_loss": 0.16496863961219788, "global_step": 104376, "epoch": 1172, "lr": 6.791145148435741e-05}
{"train_loss": 0.11779040843248367, "global_step": 104377, "epoch": 1172, "lr": 6.791091022149965e-05}
{"train_loss": 0.12041930854320526, "global_step": 104378, "epoch": 1172, "lr": 6.791036895623399e-05}
{"train_loss": 0.18516278266906738, "global_step": 104379, "epoch": 1172, "lr": 6.790982768856052e-05}
{"train_loss": 0.14381997287273407, "global_step": 104380, "epoch": 1172, "lr": 6.790928641847927e-05}
{"train_loss": 0.08281981199979782, "global_step": 104381, "epoch": 1172, "lr": 6.790874514599035e-05}
{"train_loss": 0.11029452830553055, "global_step": 104382, "epoch": 1172, "lr": 6.79082038710938e-05}
{"train_loss": 0.25180530548095703, "global_step": 104383, "epoch": 1172, "lr": 6.790766259378972e-05}
{"train_loss": 0.2537745535373688, "global_step": 104384, "epoch": 1172, "lr": 6.790712131407818e-05}
{"train_loss": 0.17969238758087158, "global_step": 104385, "epoch": 1172, "lr": 6.790658003195924e-05}
{"train_loss": 0.15456373989582062, "global_step": 104386, "epoch": 1172, "lr": 6.790603874743298e-05}
{"train_loss": 0.17820820212364197, "global_step": 104387, "epoch": 1172, "lr": 6.790549746049946e-05}
{"train_loss": 0.13577449321746826, "global_step": 104388, "epoch": 1172, "lr": 6.790495617115879e-05}
{"train_loss": 0.17862562835216522, "global_step": 104389, "epoch": 1172, "lr": 6.7904414879411e-05}
{"train_loss": 0.17951707541942596, "global_step": 104390, "epoch": 1172, "lr": 6.790387358525617e-05}
{"train_loss": 0.1702152043581009, "global_step": 104391, "epoch": 1172, "lr": 6.79033322886944e-05}
{"train_loss": 0.12294824421405792, "global_step": 104392, "epoch": 1172, "lr": 6.790279098972575e-05}
{"train_loss": 0.2332581877708435, "global_step": 104393, "epoch": 1172, "lr": 6.790224968835027e-05}
{"train_loss": 0.1658964604139328, "global_step": 104394, "epoch": 1172, "lr": 6.790170838456806e-05}
{"train_loss": 0.22009608149528503, "global_step": 104395, "epoch": 1172, "lr": 6.790116707837917e-05}
{"train_loss": 0.1576564640141605, "global_step": 104396, "epoch": 1172, "lr": 6.79006257697837e-05, "val_loss": 4.737851142883301}
{"train_loss": 0.1474751979112625, "global_step": 104397, "epoch": 1173, "lr": 6.79000844587817e-05}
{"train_loss": 0.12014665454626083, "global_step": 104398, "epoch": 1173, "lr": 6.789954314537326e-05}
{"train_loss": 0.18024352192878723, "global_step": 104399, "epoch": 1173, "lr": 6.789900182955845e-05}
{"train_loss": 0.16672147810459137, "global_step": 104400, "epoch": 1173, "lr": 6.789846051133732e-05}
{"train_loss": 0.2205812633037567, "global_step": 104401, "epoch": 1173, "lr": 6.789791919070997e-05}
{"train_loss": 0.15676318109035492, "global_step": 104402, "epoch": 1173, "lr": 6.789737786767646e-05}
{"train_loss": 0.18255995213985443, "global_step": 104403, "epoch": 1173, "lr": 6.789683654223688e-05}
{"train_loss": 0.22949936985969543, "global_step": 104404, "epoch": 1173, "lr": 6.789629521439126e-05}
{"train_loss": 0.23186129331588745, "global_step": 104405, "epoch": 1173, "lr": 6.789575388413973e-05}
{"train_loss": 0.09362909942865372, "global_step": 104406, "epoch": 1173, "lr": 6.789521255148233e-05}
{"train_loss": 0.1975211501121521, "global_step": 104407, "epoch": 1173, "lr": 6.789467121641911e-05}
{"train_loss": 0.15848426520824432, "global_step": 104408, "epoch": 1173, "lr": 6.78941298789502e-05}
{"train_loss": 0.17720438539981842, "global_step": 104409, "epoch": 1173, "lr": 6.789358853907562e-05}
{"train_loss": 0.15420962870121002, "global_step": 104410, "epoch": 1173, "lr": 6.789304719679549e-05}
{"train_loss": 0.12703001499176025, "global_step": 104411, "epoch": 1173, "lr": 6.789250585210983e-05}
{"train_loss": 0.18603631854057312, "global_step": 104412, "epoch": 1173, "lr": 6.789196450501876e-05}
{"train_loss": 0.11283854395151138, "global_step": 104413, "epoch": 1173, "lr": 6.789142315552233e-05}
{"train_loss": 0.16177988052368164, "global_step": 104414, "epoch": 1173, "lr": 6.78908818036206e-05}
{"train_loss": 0.1301603615283966, "global_step": 104415, "epoch": 1173, "lr": 6.789034044931368e-05}
{"train_loss": 0.1595298796892166, "global_step": 104416, "epoch": 1173, "lr": 6.788979909260162e-05}
{"train_loss": 0.1429435908794403, "global_step": 104417, "epoch": 1173, "lr": 6.788925773348448e-05}
{"train_loss": 0.160091370344162, "global_step": 104418, "epoch": 1173, "lr": 6.788871637196237e-05}
{"train_loss": 0.1467115730047226, "global_step": 104419, "epoch": 1173, "lr": 6.788817500803532e-05}
{"train_loss": 0.13691234588623047, "global_step": 104420, "epoch": 1173, "lr": 6.788763364170343e-05}
{"train_loss": 0.2961418330669403, "global_step": 104421, "epoch": 1173, "lr": 6.788709227296677e-05}
{"train_loss": 0.2301633507013321, "global_step": 104422, "epoch": 1173, "lr": 6.78865509018254e-05}
{"train_loss": 0.2158709615468979, "global_step": 104423, "epoch": 1173, "lr": 6.78860095282794e-05}
{"train_loss": 0.10934732854366302, "global_step": 104424, "epoch": 1173, "lr": 6.788546815232885e-05}
{"train_loss": 0.09584254026412964, "global_step": 104425, "epoch": 1173, "lr": 6.788492677397382e-05}
{"train_loss": 0.16341960430145264, "global_step": 104426, "epoch": 1173, "lr": 6.788438539321437e-05}
{"train_loss": 0.16608278453350067, "global_step": 104427, "epoch": 1173, "lr": 6.788384401005058e-05}
{"train_loss": 0.14566650986671448, "global_step": 104428, "epoch": 1173, "lr": 6.788330262448255e-05}
{"train_loss": 0.1069905161857605, "global_step": 104429, "epoch": 1173, "lr": 6.788276123651031e-05}
{"train_loss": 0.15150617063045502, "global_step": 104430, "epoch": 1173, "lr": 6.788221984613394e-05}
{"train_loss": 0.1998976618051529, "global_step": 104431, "epoch": 1173, "lr": 6.788167845335354e-05}
{"train_loss": 0.12956330180168152, "global_step": 104432, "epoch": 1173, "lr": 6.788113705816916e-05}
{"train_loss": 0.1801108568906784, "global_step": 104433, "epoch": 1173, "lr": 6.788059566058089e-05}
{"train_loss": 0.1889154464006424, "global_step": 104434, "epoch": 1173, "lr": 6.788005426058877e-05}
{"train_loss": 0.1337389200925827, "global_step": 104435, "epoch": 1173, "lr": 6.787951285819292e-05}
{"train_loss": 0.13112087547779083, "global_step": 104436, "epoch": 1173, "lr": 6.787897145339337e-05}
{"train_loss": 0.14022210240364075, "global_step": 104437, "epoch": 1173, "lr": 6.787843004619022e-05}
{"train_loss": 0.16579242050647736, "global_step": 104438, "epoch": 1173, "lr": 6.787788863658352e-05}
{"train_loss": 0.22081920504570007, "global_step": 104439, "epoch": 1173, "lr": 6.787734722457337e-05}
{"train_loss": 0.12682439386844635, "global_step": 104440, "epoch": 1173, "lr": 6.787680581015982e-05}
{"train_loss": 0.19169503450393677, "global_step": 104441, "epoch": 1173, "lr": 6.787626439334295e-05}
{"train_loss": 0.1253180354833603, "global_step": 104442, "epoch": 1173, "lr": 6.787572297412284e-05}
{"train_loss": 0.1662440001964569, "global_step": 104443, "epoch": 1173, "lr": 6.787518155249954e-05}
{"train_loss": 0.1946079134941101, "global_step": 104444, "epoch": 1173, "lr": 6.787464012847317e-05}
{"train_loss": 0.07779452949762344, "global_step": 104445, "epoch": 1173, "lr": 6.787409870204375e-05}
{"train_loss": 0.12891533970832825, "global_step": 104446, "epoch": 1173, "lr": 6.787355727321139e-05}
{"train_loss": 0.1277180165052414, "global_step": 104447, "epoch": 1173, "lr": 6.787301584197613e-05}
{"train_loss": 0.14796556532382965, "global_step": 104448, "epoch": 1173, "lr": 6.787247440833807e-05}
{"train_loss": 0.18234844505786896, "global_step": 104449, "epoch": 1173, "lr": 6.787193297229728e-05}
{"train_loss": 0.12194889038801193, "global_step": 104450, "epoch": 1173, "lr": 6.787139153385382e-05}
{"train_loss": 0.15117193758487701, "global_step": 104451, "epoch": 1173, "lr": 6.787085009300776e-05}
{"train_loss": 0.12359713762998581, "global_step": 104452, "epoch": 1173, "lr": 6.787030864975919e-05}
{"train_loss": 0.14722666144371033, "global_step": 104453, "epoch": 1173, "lr": 6.786976720410818e-05}
{"train_loss": 0.13988083600997925, "global_step": 104454, "epoch": 1173, "lr": 6.78692257560548e-05}
{"train_loss": 0.11901471018791199, "global_step": 104455, "epoch": 1173, "lr": 6.786868430559912e-05}
{"train_loss": 0.11684809625148773, "global_step": 104456, "epoch": 1173, "lr": 6.78681428527412e-05}
{"train_loss": 0.19257104396820068, "global_step": 104457, "epoch": 1173, "lr": 6.786760139748114e-05}
{"train_loss": 0.12228590995073318, "global_step": 104458, "epoch": 1173, "lr": 6.786705993981898e-05}
{"train_loss": 0.09799794852733612, "global_step": 104459, "epoch": 1173, "lr": 6.786651847975482e-05}
{"train_loss": 0.11087361723184586, "global_step": 104460, "epoch": 1173, "lr": 6.786597701728874e-05}
{"train_loss": 0.14124982059001923, "global_step": 104461, "epoch": 1173, "lr": 6.786543555242079e-05}
{"train_loss": 0.13049480319023132, "global_step": 104462, "epoch": 1173, "lr": 6.786489408515105e-05}
{"train_loss": 0.12945175170898438, "global_step": 104463, "epoch": 1173, "lr": 6.78643526154796e-05}
{"train_loss": 0.17542366683483124, "global_step": 104464, "epoch": 1173, "lr": 6.786381114340647e-05}
{"train_loss": 0.14053936302661896, "global_step": 104465, "epoch": 1173, "lr": 6.78632696689318e-05}
{"train_loss": 0.16075004637241364, "global_step": 104466, "epoch": 1173, "lr": 6.786272819205561e-05}
{"train_loss": 0.1490073949098587, "global_step": 104467, "epoch": 1173, "lr": 6.786218671277802e-05}
{"train_loss": 0.1447163224220276, "global_step": 104468, "epoch": 1173, "lr": 6.786164523109905e-05}
{"train_loss": 0.12262393534183502, "global_step": 104469, "epoch": 1173, "lr": 6.786110374701882e-05}
{"train_loss": 0.1718841791152954, "global_step": 104470, "epoch": 1173, "lr": 6.786056226053738e-05}
{"train_loss": 0.12568357586860657, "global_step": 104471, "epoch": 1173, "lr": 6.78600207716548e-05}
{"train_loss": 0.10479523986577988, "global_step": 104472, "epoch": 1173, "lr": 6.785947928037116e-05}
{"train_loss": 0.2138359397649765, "global_step": 104473, "epoch": 1173, "lr": 6.785893778668652e-05}
{"train_loss": 0.18745800852775574, "global_step": 104474, "epoch": 1173, "lr": 6.785839629060097e-05}
{"train_loss": 0.1531093269586563, "global_step": 104475, "epoch": 1173, "lr": 6.78578547921146e-05}
{"train_loss": 0.10380324721336365, "global_step": 104476, "epoch": 1173, "lr": 6.785731329122743e-05}
{"train_loss": 0.15243560075759888, "global_step": 104477, "epoch": 1173, "lr": 6.785677178793957e-05}
{"train_loss": 0.08894471079111099, "global_step": 104478, "epoch": 1173, "lr": 6.785623028225108e-05}
{"train_loss": 0.14013928174972534, "global_step": 104479, "epoch": 1173, "lr": 6.785568877416203e-05}
{"train_loss": 0.1479794979095459, "global_step": 104480, "epoch": 1173, "lr": 6.785514726367252e-05}
{"train_loss": 0.19426698982715607, "global_step": 104481, "epoch": 1173, "lr": 6.78546057507826e-05}
{"train_loss": 0.19653166830539703, "global_step": 104482, "epoch": 1173, "lr": 6.785406423549234e-05}
{"train_loss": 0.19122955203056335, "global_step": 104483, "epoch": 1173, "lr": 6.785352271780183e-05}
{"train_loss": 0.15469340980052948, "global_step": 104484, "epoch": 1173, "lr": 6.785298119771112e-05}
{"train_loss": 0.1536845743991016, "global_step": 104485, "epoch": 1173, "lr": 6.785243967522029e-05, "val_loss": 4.802351474761963}
{"train_loss": 0.1875377744436264, "global_step": 104486, "epoch": 1174, "lr": 6.785189815032943e-05}
{"train_loss": 0.23488783836364746, "global_step": 104487, "epoch": 1174, "lr": 6.785135662303859e-05}
{"train_loss": 0.12336120009422302, "global_step": 104488, "epoch": 1174, "lr": 6.785081509334785e-05}
{"train_loss": 0.14791953563690186, "global_step": 104489, "epoch": 1174, "lr": 6.78502735612573e-05}
{"train_loss": 0.1226453185081482, "global_step": 104490, "epoch": 1174, "lr": 6.784973202676698e-05}
{"train_loss": 0.19217625260353088, "global_step": 104491, "epoch": 1174, "lr": 6.7849190489877e-05}
{"train_loss": 0.13340242207050323, "global_step": 104492, "epoch": 1174, "lr": 6.78486489505874e-05}
{"train_loss": 0.15444110333919525, "global_step": 104493, "epoch": 1174, "lr": 6.784810740889828e-05}
{"train_loss": 0.20513877272605896, "global_step": 104494, "epoch": 1174, "lr": 6.784756586480968e-05}
{"train_loss": 0.16703324019908905, "global_step": 104495, "epoch": 1174, "lr": 6.78470243183217e-05}
{"train_loss": 0.11334038525819778, "global_step": 104496, "epoch": 1174, "lr": 6.784648276943441e-05}
{"train_loss": 0.11237860471010208, "global_step": 104497, "epoch": 1174, "lr": 6.784594121814788e-05}
{"train_loss": 0.16035233438014984, "global_step": 104498, "epoch": 1174, "lr": 6.784539966446217e-05}
{"train_loss": 0.15399572253227234, "global_step": 104499, "epoch": 1174, "lr": 6.784485810837737e-05}
{"train_loss": 0.09887268394231796, "global_step": 104500, "epoch": 1174, "lr": 6.784431654989355e-05}
{"train_loss": 0.10638032853603363, "global_step": 104501, "epoch": 1174, "lr": 6.784377498901077e-05}
{"train_loss": 0.12173610180616379, "global_step": 104502, "epoch": 1174, "lr": 6.784323342572913e-05}
{"train_loss": 0.15093889832496643, "global_step": 104503, "epoch": 1174, "lr": 6.784269186004867e-05}
{"train_loss": 0.21826432645320892, "global_step": 104504, "epoch": 1174, "lr": 6.784215029196948e-05}
{"train_loss": 0.1636265218257904, "global_step": 104505, "epoch": 1174, "lr": 6.78416087214916e-05}
{"train_loss": 0.13368743658065796, "global_step": 104506, "epoch": 1174, "lr": 6.784106714861518e-05}
{"train_loss": 0.13673223555088043, "global_step": 104507, "epoch": 1174, "lr": 6.784052557334023e-05}
{"train_loss": 0.0906701609492302, "global_step": 104508, "epoch": 1174, "lr": 6.783998399566685e-05}
{"train_loss": 0.1467655897140503, "global_step": 104509, "epoch": 1174, "lr": 6.783944241559508e-05}
{"train_loss": 0.20604932308197021, "global_step": 104510, "epoch": 1174, "lr": 6.783890083312502e-05}
{"train_loss": 0.18204531073570251, "global_step": 104511, "epoch": 1174, "lr": 6.783835924825675e-05}
{"train_loss": 0.15374986827373505, "global_step": 104512, "epoch": 1174, "lr": 6.783781766099032e-05}
{"train_loss": 0.24838627874851227, "global_step": 104513, "epoch": 1174, "lr": 6.783727607132581e-05}
{"train_loss": 0.1629655361175537, "global_step": 104514, "epoch": 1174, "lr": 6.78367344792633e-05}
{"train_loss": 0.12761405110359192, "global_step": 104515, "epoch": 1174, "lr": 6.783619288480288e-05}
{"train_loss": 0.13365532457828522, "global_step": 104516, "epoch": 1174, "lr": 6.783565128794457e-05}
{"train_loss": 0.13817188143730164, "global_step": 104517, "epoch": 1174, "lr": 6.783510968868847e-05}
{"train_loss": 0.10172624886035919, "global_step": 104518, "epoch": 1174, "lr": 6.783456808703468e-05}
{"train_loss": 0.09014801681041718, "global_step": 104519, "epoch": 1174, "lr": 6.783402648298323e-05}
{"train_loss": 0.1521499752998352, "global_step": 104520, "epoch": 1174, "lr": 6.783348487653424e-05}
{"train_loss": 0.09188003093004227, "global_step": 104521, "epoch": 1174, "lr": 6.783294326768774e-05}
{"train_loss": 0.16982533037662506, "global_step": 104522, "epoch": 1174, "lr": 6.78324016564438e-05}
{"train_loss": 0.14724747836589813, "global_step": 104523, "epoch": 1174, "lr": 6.783186004280253e-05}
{"train_loss": 0.14341898262500763, "global_step": 104524, "epoch": 1174, "lr": 6.783131842676398e-05}
{"train_loss": 0.13715380430221558, "global_step": 104525, "epoch": 1174, "lr": 6.783077680832823e-05}
{"train_loss": 0.1326395720243454, "global_step": 104526, "epoch": 1174, "lr": 6.783023518749534e-05}
{"train_loss": 0.17285552620887756, "global_step": 104527, "epoch": 1174, "lr": 6.78296935642654e-05}
{"train_loss": 0.12822914123535156, "global_step": 104528, "epoch": 1174, "lr": 6.782915193863847e-05}
{"train_loss": 0.15897154808044434, "global_step": 104529, "epoch": 1174, "lr": 6.782861031061464e-05}
{"train_loss": 0.12714925408363342, "global_step": 104530, "epoch": 1174, "lr": 6.782806868019397e-05}
{"train_loss": 0.15670980513095856, "global_step": 104531, "epoch": 1174, "lr": 6.782752704737653e-05}
{"train_loss": 0.1638312041759491, "global_step": 104532, "epoch": 1174, "lr": 6.782698541216239e-05}
{"train_loss": 0.13868968188762665, "global_step": 104533, "epoch": 1174, "lr": 6.782644377455162e-05}
{"train_loss": 0.12158818542957306, "global_step": 104534, "epoch": 1174, "lr": 6.782590213454432e-05}
{"train_loss": 0.08530044555664062, "global_step": 104535, "epoch": 1174, "lr": 6.782536049214054e-05}
{"train_loss": 0.1535828709602356, "global_step": 104536, "epoch": 1174, "lr": 6.782481884734035e-05}
{"train_loss": 0.12926223874092102, "global_step": 104537, "epoch": 1174, "lr": 6.782427720014384e-05}
{"train_loss": 0.24807539582252502, "global_step": 104538, "epoch": 1174, "lr": 6.782373555055107e-05}
{"train_loss": 0.11294631659984589, "global_step": 104539, "epoch": 1174, "lr": 6.782319389856211e-05}
{"train_loss": 0.15210884809494019, "global_step": 104540, "epoch": 1174, "lr": 6.782265224417705e-05}
{"train_loss": 0.17226286232471466, "global_step": 104541, "epoch": 1174, "lr": 6.782211058739593e-05}
{"train_loss": 0.16863618791103363, "global_step": 104542, "epoch": 1174, "lr": 6.782156892821885e-05}
{"train_loss": 0.2224033623933792, "global_step": 104543, "epoch": 1174, "lr": 6.78210272666459e-05}
{"train_loss": 0.22519688308238983, "global_step": 104544, "epoch": 1174, "lr": 6.782048560267711e-05}
{"train_loss": 0.10498012602329254, "global_step": 104545, "epoch": 1174, "lr": 6.781994393631259e-05}
{"train_loss": 0.1970563679933548, "global_step": 104546, "epoch": 1174, "lr": 6.781940226755238e-05}
{"train_loss": 0.15197093784809113, "global_step": 104547, "epoch": 1174, "lr": 6.781886059639659e-05}
{"train_loss": 0.13504475355148315, "global_step": 104548, "epoch": 1174, "lr": 6.781831892284525e-05}
{"train_loss": 0.11124259233474731, "global_step": 104549, "epoch": 1174, "lr": 6.781777724689847e-05}
{"train_loss": 0.20281170308589935, "global_step": 104550, "epoch": 1174, "lr": 6.78172355685563e-05}
{"train_loss": 0.11269064992666245, "global_step": 104551, "epoch": 1174, "lr": 6.78166938878188e-05}
{"train_loss": 0.1204148381948471, "global_step": 104552, "epoch": 1174, "lr": 6.781615220468609e-05}
{"train_loss": 0.23854388296604156, "global_step": 104553, "epoch": 1174, "lr": 6.78156105191582e-05}
{"train_loss": 0.16670183837413788, "global_step": 104554, "epoch": 1174, "lr": 6.781506883123523e-05}
{"train_loss": 0.1451377123594284, "global_step": 104555, "epoch": 1174, "lr": 6.781452714091723e-05}
{"train_loss": 0.18023034930229187, "global_step": 104556, "epoch": 1174, "lr": 6.781398544820429e-05}
{"train_loss": 0.1364278644323349, "global_step": 104557, "epoch": 1174, "lr": 6.781344375309647e-05}
{"train_loss": 0.1313445121049881, "global_step": 104558, "epoch": 1174, "lr": 6.781290205559386e-05}
{"train_loss": 0.17037901282310486, "global_step": 104559, "epoch": 1174, "lr": 6.781236035569652e-05}
{"train_loss": 0.18672537803649902, "global_step": 104560, "epoch": 1174, "lr": 6.78118186534045e-05}
{"train_loss": 0.12538601458072662, "global_step": 104561, "epoch": 1174, "lr": 6.781127694871792e-05}
{"train_loss": 0.111236073076725, "global_step": 104562, "epoch": 1174, "lr": 6.781073524163684e-05}
{"train_loss": 0.1473056674003601, "global_step": 104563, "epoch": 1174, "lr": 6.781019353216131e-05}
{"train_loss": 0.27828285098075867, "global_step": 104564, "epoch": 1174, "lr": 6.780965182029141e-05}
{"train_loss": 0.13558849692344666, "global_step": 104565, "epoch": 1174, "lr": 6.780911010602724e-05}
{"train_loss": 0.2052706778049469, "global_step": 104566, "epoch": 1174, "lr": 6.780856838936887e-05}
{"train_loss": 0.12319120019674301, "global_step": 104567, "epoch": 1174, "lr": 6.780802667031632e-05}
{"train_loss": 0.16110798716545105, "global_step": 104568, "epoch": 1174, "lr": 6.78074849488697e-05}
{"train_loss": 0.16024336218833923, "global_step": 104569, "epoch": 1174, "lr": 6.78069432250291e-05}
{"train_loss": 0.11238779872655869, "global_step": 104570, "epoch": 1174, "lr": 6.780640149879455e-05}
{"train_loss": 0.12075105309486389, "global_step": 104571, "epoch": 1174, "lr": 6.780585977016616e-05}
{"train_loss": 0.2021656185388565, "global_step": 104572, "epoch": 1174, "lr": 6.780531803914399e-05}
{"train_loss": 0.14214999973773956, "global_step": 104573, "epoch": 1174, "lr": 6.78047763057281e-05}
{"train_loss": 0.15173991350980287, "global_step": 104574, "epoch": 1174, "lr": 6.780423456991857e-05, "val_loss": 4.943373680114746}
{"train_loss": 0.1494375765323639, "global_step": 104575, "epoch": 1175, "lr": 6.78036928317155e-05}
{"train_loss": 0.17426516115665436, "global_step": 104576, "epoch": 1175, "lr": 6.780315109111893e-05}
{"train_loss": 0.1276106834411621, "global_step": 104577, "epoch": 1175, "lr": 6.780260934812895e-05}
{"train_loss": 0.13993802666664124, "global_step": 104578, "epoch": 1175, "lr": 6.780206760274561e-05}
{"train_loss": 0.09753734618425369, "global_step": 104579, "epoch": 1175, "lr": 6.780152585496901e-05}
{"train_loss": 0.14079345762729645, "global_step": 104580, "epoch": 1175, "lr": 6.78009841047992e-05}
{"train_loss": 0.1338961124420166, "global_step": 104581, "epoch": 1175, "lr": 6.780044235223628e-05}
{"train_loss": 0.10345587879419327, "global_step": 104582, "epoch": 1175, "lr": 6.779990059728031e-05}
{"train_loss": 0.17237979173660278, "global_step": 104583, "epoch": 1175, "lr": 6.779935883993135e-05}
{"train_loss": 0.13066253066062927, "global_step": 104584, "epoch": 1175, "lr": 6.77988170801895e-05}
{"train_loss": 0.1777203232049942, "global_step": 104585, "epoch": 1175, "lr": 6.77982753180548e-05}
{"train_loss": 0.1373571902513504, "global_step": 104586, "epoch": 1175, "lr": 6.779773355352734e-05}
{"train_loss": 0.16040782630443573, "global_step": 104587, "epoch": 1175, "lr": 6.77971917866072e-05}
{"train_loss": 0.2686093747615814, "global_step": 104588, "epoch": 1175, "lr": 6.779665001729444e-05}
{"train_loss": 0.16600672900676727, "global_step": 104589, "epoch": 1175, "lr": 6.779610824558913e-05}
{"train_loss": 0.1652509868144989, "global_step": 104590, "epoch": 1175, "lr": 6.779556647149136e-05}
{"train_loss": 0.10089384019374847, "global_step": 104591, "epoch": 1175, "lr": 6.779502469500118e-05}
{"train_loss": 0.1814059019088745, "global_step": 104592, "epoch": 1175, "lr": 6.779448291611868e-05}
{"train_loss": 0.1900143176317215, "global_step": 104593, "epoch": 1175, "lr": 6.779394113484395e-05}
{"train_loss": 0.15116481482982635, "global_step": 104594, "epoch": 1175, "lr": 6.779339935117701e-05}
{"train_loss": 0.1747329831123352, "global_step": 104595, "epoch": 1175, "lr": 6.779285756511799e-05}
{"train_loss": 0.14644503593444824, "global_step": 104596, "epoch": 1175, "lr": 6.779231577666692e-05}
{"train_loss": 0.13162904977798462, "global_step": 104597, "epoch": 1175, "lr": 6.779177398582388e-05}
{"train_loss": 0.10577438026666641, "global_step": 104598, "epoch": 1175, "lr": 6.779123219258897e-05}
{"train_loss": 0.19850613176822662, "global_step": 104599, "epoch": 1175, "lr": 6.779069039696224e-05}
{"train_loss": 0.12320216745138168, "global_step": 104600, "epoch": 1175, "lr": 6.779014859894378e-05}
{"train_loss": 0.17634344100952148, "global_step": 104601, "epoch": 1175, "lr": 6.778960679853365e-05}
{"train_loss": 0.15497088432312012, "global_step": 104602, "epoch": 1175, "lr": 6.778906499573191e-05}
{"train_loss": 0.12112824618816376, "global_step": 104603, "epoch": 1175, "lr": 6.778852319053866e-05}
{"train_loss": 0.17177638411521912, "global_step": 104604, "epoch": 1175, "lr": 6.778798138295395e-05}
{"train_loss": 0.22425611317157745, "global_step": 104605, "epoch": 1175, "lr": 6.778743957297789e-05}
{"train_loss": 0.23696239292621613, "global_step": 104606, "epoch": 1175, "lr": 6.778689776061049e-05}
{"train_loss": 0.1048804298043251, "global_step": 104607, "epoch": 1175, "lr": 6.778635594585187e-05}
{"train_loss": 0.14655329287052155, "global_step": 104608, "epoch": 1175, "lr": 6.778581412870209e-05}
{"train_loss": 0.13130047917366028, "global_step": 104609, "epoch": 1175, "lr": 6.778527230916122e-05}
{"train_loss": 0.20145781338214874, "global_step": 104610, "epoch": 1175, "lr": 6.778473048722935e-05}
{"train_loss": 0.12481018155813217, "global_step": 104611, "epoch": 1175, "lr": 6.778418866290655e-05}
{"train_loss": 0.2042836844921112, "global_step": 104612, "epoch": 1175, "lr": 6.778364683619286e-05}
{"train_loss": 0.19044768810272217, "global_step": 104613, "epoch": 1175, "lr": 6.778310500708838e-05}
{"train_loss": 0.20517054200172424, "global_step": 104614, "epoch": 1175, "lr": 6.778256317559318e-05}
{"train_loss": 0.10544309765100479, "global_step": 104615, "epoch": 1175, "lr": 6.778202134170733e-05}
{"train_loss": 0.17341987788677216, "global_step": 104616, "epoch": 1175, "lr": 6.77814795054309e-05}
{"train_loss": 0.18641917407512665, "global_step": 104617, "epoch": 1175, "lr": 6.778093766676397e-05}
{"train_loss": 0.16781660914421082, "global_step": 104618, "epoch": 1175, "lr": 6.778039582570661e-05}
{"train_loss": 0.13097764551639557, "global_step": 104619, "epoch": 1175, "lr": 6.777985398225891e-05}
{"train_loss": 0.21076451241970062, "global_step": 104620, "epoch": 1175, "lr": 6.77793121364209e-05}
{"train_loss": 0.0875333845615387, "global_step": 104621, "epoch": 1175, "lr": 6.777877028819269e-05}
{"train_loss": 0.2358887642621994, "global_step": 104622, "epoch": 1175, "lr": 6.777822843757433e-05}
{"train_loss": 0.24943584203720093, "global_step": 104623, "epoch": 1175, "lr": 6.777768658456592e-05}
{"train_loss": 0.09858962148427963, "global_step": 104624, "epoch": 1175, "lr": 6.777714472916752e-05}
{"train_loss": 0.12487863004207611, "global_step": 104625, "epoch": 1175, "lr": 6.777660287137918e-05}
{"train_loss": 0.18266421556472778, "global_step": 104626, "epoch": 1175, "lr": 6.7776061011201e-05}
{"train_loss": 0.19621263444423676, "global_step": 104627, "epoch": 1175, "lr": 6.777551914863305e-05}
{"train_loss": 0.20914024114608765, "global_step": 104628, "epoch": 1175, "lr": 6.777497728367539e-05}
{"train_loss": 0.1921030580997467, "global_step": 104629, "epoch": 1175, "lr": 6.77744354163281e-05}
{"train_loss": 0.2163492888212204, "global_step": 104630, "epoch": 1175, "lr": 6.777389354659127e-05}
{"train_loss": 0.16753605008125305, "global_step": 104631, "epoch": 1175, "lr": 6.777335167446494e-05}
{"train_loss": 0.10604213178157806, "global_step": 104632, "epoch": 1175, "lr": 6.77728097999492e-05}
{"train_loss": 0.20011235773563385, "global_step": 104633, "epoch": 1175, "lr": 6.777226792304413e-05}
{"train_loss": 0.21865035593509674, "global_step": 104634, "epoch": 1175, "lr": 6.777172604374979e-05}
{"train_loss": 0.19964708387851715, "global_step": 104635, "epoch": 1175, "lr": 6.777118416206627e-05}
{"train_loss": 0.17194031178951263, "global_step": 104636, "epoch": 1175, "lr": 6.777064227799362e-05}
{"train_loss": 0.1894097775220871, "global_step": 104637, "epoch": 1175, "lr": 6.777010039153192e-05}
{"train_loss": 0.16897335648536682, "global_step": 104638, "epoch": 1175, "lr": 6.776955850268126e-05}
{"train_loss": 0.1163957491517067, "global_step": 104639, "epoch": 1175, "lr": 6.776901661144168e-05}
{"train_loss": 0.20132657885551453, "global_step": 104640, "epoch": 1175, "lr": 6.77684747178133e-05}
{"train_loss": 0.13960644602775574, "global_step": 104641, "epoch": 1175, "lr": 6.776793282179613e-05}
{"train_loss": 0.2263542115688324, "global_step": 104642, "epoch": 1175, "lr": 6.77673909233903e-05}
{"train_loss": 0.17091040313243866, "global_step": 104643, "epoch": 1175, "lr": 6.776684902259587e-05}
{"train_loss": 0.1793467253446579, "global_step": 104644, "epoch": 1175, "lr": 6.77663071194129e-05}
{"train_loss": 0.17112267017364502, "global_step": 104645, "epoch": 1175, "lr": 6.776576521384144e-05}
{"train_loss": 0.14566577970981598, "global_step": 104646, "epoch": 1175, "lr": 6.776522330588162e-05}
{"train_loss": 0.17160971462726593, "global_step": 104647, "epoch": 1175, "lr": 6.776468139553347e-05}
{"train_loss": 0.18350987136363983, "global_step": 104648, "epoch": 1175, "lr": 6.776413948279707e-05}
{"train_loss": 0.16157832741737366, "global_step": 104649, "epoch": 1175, "lr": 6.77635975676725e-05}
{"train_loss": 0.22385068237781525, "global_step": 104650, "epoch": 1175, "lr": 6.776305565015984e-05}
{"train_loss": 0.17592261731624603, "global_step": 104651, "epoch": 1175, "lr": 6.776251373025913e-05}
{"train_loss": 0.10768245905637741, "global_step": 104652, "epoch": 1175, "lr": 6.77619718079705e-05}
{"train_loss": 0.11908160895109177, "global_step": 104653, "epoch": 1175, "lr": 6.776142988329396e-05}
{"train_loss": 0.12454365193843842, "global_step": 104654, "epoch": 1175, "lr": 6.776088795622964e-05}
{"train_loss": 0.20010702311992645, "global_step": 104655, "epoch": 1175, "lr": 6.776034602677756e-05}
{"train_loss": 0.1578102707862854, "global_step": 104656, "epoch": 1175, "lr": 6.775980409493782e-05}
{"train_loss": 0.23558706045150757, "global_step": 104657, "epoch": 1175, "lr": 6.775926216071051e-05}
{"train_loss": 0.17260827124118805, "global_step": 104658, "epoch": 1175, "lr": 6.775872022409568e-05}
{"train_loss": 0.1640777587890625, "global_step": 104659, "epoch": 1175, "lr": 6.775817828509339e-05}
{"train_loss": 0.20769758522510529, "global_step": 104660, "epoch": 1175, "lr": 6.775763634370373e-05}
{"train_loss": 0.17950166761875153, "global_step": 104661, "epoch": 1175, "lr": 6.77570943999268e-05}
{"train_loss": 0.16796648502349854, "global_step": 104662, "epoch": 1175, "lr": 6.775655245376262e-05}
{"train_loss": 0.16607842142327447, "global_step": 104663, "epoch": 1175, "lr": 6.77560105052113e-05, "val_loss": 4.9365715980529785, "train_action_mse_error": 16.903812408447266}
{"train_loss": 0.1505347639322281, "global_step": 104664, "epoch": 1176, "lr": 6.77554685542729e-05}
{"train_loss": 0.15750239789485931, "global_step": 104665, "epoch": 1176, "lr": 6.775492660094749e-05}
{"train_loss": 0.24578168988227844, "global_step": 104666, "epoch": 1176, "lr": 6.775438464523514e-05}
{"train_loss": 0.20038248598575592, "global_step": 104667, "epoch": 1176, "lr": 6.775384268713593e-05}
{"train_loss": 0.13239943981170654, "global_step": 104668, "epoch": 1176, "lr": 6.775330072664994e-05}
{"train_loss": 0.15283770859241486, "global_step": 104669, "epoch": 1176, "lr": 6.775275876377723e-05}
{"train_loss": 0.12783218920230865, "global_step": 104670, "epoch": 1176, "lr": 6.775221679851787e-05}
{"train_loss": 0.08753300458192825, "global_step": 104671, "epoch": 1176, "lr": 6.775167483087195e-05}
{"train_loss": 0.22485633194446564, "global_step": 104672, "epoch": 1176, "lr": 6.775113286083954e-05}
{"train_loss": 0.21042540669441223, "global_step": 104673, "epoch": 1176, "lr": 6.775059088842069e-05}
{"train_loss": 0.20132967829704285, "global_step": 104674, "epoch": 1176, "lr": 6.775004891361552e-05}
{"train_loss": 0.15486618876457214, "global_step": 104675, "epoch": 1176, "lr": 6.774950693642405e-05}
{"train_loss": 0.1893116682767868, "global_step": 104676, "epoch": 1176, "lr": 6.774896495684638e-05}
{"train_loss": 0.09286940842866898, "global_step": 104677, "epoch": 1176, "lr": 6.774842297488257e-05}
{"train_loss": 0.14943568408489227, "global_step": 104678, "epoch": 1176, "lr": 6.77478809905327e-05}
{"train_loss": 0.18557272851467133, "global_step": 104679, "epoch": 1176, "lr": 6.774733900379685e-05}
{"train_loss": 0.21005341410636902, "global_step": 104680, "epoch": 1176, "lr": 6.77467970146751e-05}
{"train_loss": 0.13609568774700165, "global_step": 104681, "epoch": 1176, "lr": 6.774625502316749e-05}
{"train_loss": 0.1739419847726822, "global_step": 104682, "epoch": 1176, "lr": 6.774571302927413e-05}
{"train_loss": 0.1788657009601593, "global_step": 104683, "epoch": 1176, "lr": 6.774517103299505e-05}
{"train_loss": 0.1684977263212204, "global_step": 104684, "epoch": 1176, "lr": 6.774462903433037e-05}
{"train_loss": 0.1452970951795578, "global_step": 104685, "epoch": 1176, "lr": 6.774408703328014e-05}
{"train_loss": 0.17156270146369934, "global_step": 104686, "epoch": 1176, "lr": 6.774354502984441e-05}
{"train_loss": 0.1530810445547104, "global_step": 104687, "epoch": 1176, "lr": 6.774300302402328e-05}
{"train_loss": 0.09622827917337418, "global_step": 104688, "epoch": 1176, "lr": 6.774246101581685e-05}
{"train_loss": 0.14517414569854736, "global_step": 104689, "epoch": 1176, "lr": 6.774191900522514e-05}
{"train_loss": 0.11570315808057785, "global_step": 104690, "epoch": 1176, "lr": 6.774137699224825e-05}
{"train_loss": 0.12105615437030792, "global_step": 104691, "epoch": 1176, "lr": 6.774083497688624e-05}
{"train_loss": 0.10304290801286697, "global_step": 104692, "epoch": 1176, "lr": 6.77402929591392e-05}
{"train_loss": 0.12803177535533905, "global_step": 104693, "epoch": 1176, "lr": 6.77397509390072e-05}
{"train_loss": 0.10803838819265366, "global_step": 104694, "epoch": 1176, "lr": 6.77392089164903e-05}
{"train_loss": 0.21553851664066315, "global_step": 104695, "epoch": 1176, "lr": 6.773866689158858e-05}
{"train_loss": 0.20089183747768402, "global_step": 104696, "epoch": 1176, "lr": 6.77381248643021e-05}
{"train_loss": 0.13645118474960327, "global_step": 104697, "epoch": 1176, "lr": 6.773758283463096e-05}
{"train_loss": 0.1708022952079773, "global_step": 104698, "epoch": 1176, "lr": 6.773704080257522e-05}
{"train_loss": 0.16746975481510162, "global_step": 104699, "epoch": 1176, "lr": 6.773649876813495e-05}
{"train_loss": 0.16502107679843903, "global_step": 104700, "epoch": 1176, "lr": 6.773595673131021e-05}
{"train_loss": 0.1250072568655014, "global_step": 104701, "epoch": 1176, "lr": 6.773541469210112e-05}
{"train_loss": 0.13530486822128296, "global_step": 104702, "epoch": 1176, "lr": 6.77348726505077e-05}
{"train_loss": 0.13122913241386414, "global_step": 104703, "epoch": 1176, "lr": 6.773433060653003e-05}
{"train_loss": 0.10000305622816086, "global_step": 104704, "epoch": 1176, "lr": 6.773378856016821e-05}
{"train_loss": 0.10462640970945358, "global_step": 104705, "epoch": 1176, "lr": 6.773324651142227e-05}
{"train_loss": 0.2011721283197403, "global_step": 104706, "epoch": 1176, "lr": 6.773270446029234e-05}
{"train_loss": 0.17970241606235504, "global_step": 104707, "epoch": 1176, "lr": 6.773216240677846e-05}
{"train_loss": 0.17410282790660858, "global_step": 104708, "epoch": 1176, "lr": 6.773162035088071e-05}
{"train_loss": 0.09937050938606262, "global_step": 104709, "epoch": 1176, "lr": 6.773107829259915e-05}
{"train_loss": 0.22981926798820496, "global_step": 104710, "epoch": 1176, "lr": 6.773053623193387e-05}
{"train_loss": 0.12925533950328827, "global_step": 104711, "epoch": 1176, "lr": 6.772999416888494e-05}
{"train_loss": 0.14704370498657227, "global_step": 104712, "epoch": 1176, "lr": 6.77294521034524e-05}
{"train_loss": 0.15153785049915314, "global_step": 104713, "epoch": 1176, "lr": 6.772891003563638e-05}
{"train_loss": 0.17193078994750977, "global_step": 104714, "epoch": 1176, "lr": 6.772836796543693e-05}
{"train_loss": 0.12918832898139954, "global_step": 104715, "epoch": 1176, "lr": 6.772782589285409e-05}
{"train_loss": 0.23303565382957458, "global_step": 104716, "epoch": 1176, "lr": 6.772728381788797e-05}
{"train_loss": 0.23309233784675598, "global_step": 104717, "epoch": 1176, "lr": 6.772674174053864e-05}
{"train_loss": 0.18541933596134186, "global_step": 104718, "epoch": 1176, "lr": 6.772619966080617e-05}
{"train_loss": 0.17514668405056, "global_step": 104719, "epoch": 1176, "lr": 6.772565757869062e-05}
{"train_loss": 0.1419382393360138, "global_step": 104720, "epoch": 1176, "lr": 6.772511549419207e-05}
{"train_loss": 0.12189242988824844, "global_step": 104721, "epoch": 1176, "lr": 6.772457340731061e-05}
{"train_loss": 0.17759272456169128, "global_step": 104722, "epoch": 1176, "lr": 6.772403131804628e-05}
{"train_loss": 0.17841938138008118, "global_step": 104723, "epoch": 1176, "lr": 6.772348922639917e-05}
{"train_loss": 0.2071065902709961, "global_step": 104724, "epoch": 1176, "lr": 6.772294713236935e-05}
{"train_loss": 0.13560593128204346, "global_step": 104725, "epoch": 1176, "lr": 6.772240503595691e-05}
{"train_loss": 0.21722237765789032, "global_step": 104726, "epoch": 1176, "lr": 6.77218629371619e-05}
{"train_loss": 0.15023356676101685, "global_step": 104727, "epoch": 1176, "lr": 6.772132083598439e-05}
{"train_loss": 0.15553227066993713, "global_step": 104728, "epoch": 1176, "lr": 6.772077873242449e-05}
{"train_loss": 0.1924835443496704, "global_step": 104729, "epoch": 1176, "lr": 6.772023662648222e-05}
{"train_loss": 0.14273251593112946, "global_step": 104730, "epoch": 1176, "lr": 6.77196945181577e-05}
{"train_loss": 0.22020038962364197, "global_step": 104731, "epoch": 1176, "lr": 6.771915240745096e-05}
{"train_loss": 0.1464371383190155, "global_step": 104732, "epoch": 1176, "lr": 6.771861029436213e-05}
{"train_loss": 0.13483385741710663, "global_step": 104733, "epoch": 1176, "lr": 6.771806817889122e-05}
{"train_loss": 0.12955614924430847, "global_step": 104734, "epoch": 1176, "lr": 6.771752606103835e-05}
{"train_loss": 0.19973154366016388, "global_step": 104735, "epoch": 1176, "lr": 6.771698394080356e-05}
{"train_loss": 0.2114381045103073, "global_step": 104736, "epoch": 1176, "lr": 6.771644181818695e-05}
{"train_loss": 0.20813286304473877, "global_step": 104737, "epoch": 1176, "lr": 6.771589969318857e-05}
{"train_loss": 0.1154903694987297, "global_step": 104738, "epoch": 1176, "lr": 6.771535756580853e-05}
{"train_loss": 0.12204763293266296, "global_step": 104739, "epoch": 1176, "lr": 6.771481543604685e-05}
{"train_loss": 0.2253379225730896, "global_step": 104740, "epoch": 1176, "lr": 6.771427330390365e-05}
{"train_loss": 0.2293425053358078, "global_step": 104741, "epoch": 1176, "lr": 6.771373116937894e-05}
{"train_loss": 0.12678316235542297, "global_step": 104742, "epoch": 1176, "lr": 6.771318903247287e-05}
{"train_loss": 0.12490574270486832, "global_step": 104743, "epoch": 1176, "lr": 6.771264689318547e-05}
{"train_loss": 0.14811056852340698, "global_step": 104744, "epoch": 1176, "lr": 6.771210475151682e-05}
{"train_loss": 0.1235010027885437, "global_step": 104745, "epoch": 1176, "lr": 6.771156260746698e-05}
{"train_loss": 0.11593465507030487, "global_step": 104746, "epoch": 1176, "lr": 6.771102046103605e-05}
{"train_loss": 0.16230422258377075, "global_step": 104747, "epoch": 1176, "lr": 6.771047831222408e-05}
{"train_loss": 0.11294974386692047, "global_step": 104748, "epoch": 1176, "lr": 6.770993616103117e-05}
{"train_loss": 0.11459621042013168, "global_step": 104749, "epoch": 1176, "lr": 6.770939400745737e-05}
{"train_loss": 0.20458127558231354, "global_step": 104750, "epoch": 1176, "lr": 6.770885185150276e-05}
{"train_loss": 0.25079965591430664, "global_step": 104751, "epoch": 1176, "lr": 6.770830969316739e-05}
{"train_loss": 0.1608225991216938, "global_step": 104752, "epoch": 1176, "lr": 6.770776753245137e-05, "val_loss": 4.844504356384277}
{"train_loss": 0.12224985659122467, "global_step": 104753, "epoch": 1177, "lr": 6.770722536935475e-05}
{"train_loss": 0.19057127833366394, "global_step": 104754, "epoch": 1177, "lr": 6.77066832038776e-05}
{"train_loss": 0.20355473458766937, "global_step": 104755, "epoch": 1177, "lr": 6.770614103602003e-05}
{"train_loss": 0.16722236573696136, "global_step": 104756, "epoch": 1177, "lr": 6.770559886578207e-05}
{"train_loss": 0.13827523589134216, "global_step": 104757, "epoch": 1177, "lr": 6.77050566931638e-05}
{"train_loss": 0.12032398581504822, "global_step": 104758, "epoch": 1177, "lr": 6.770451451816532e-05}
{"train_loss": 0.16985023021697998, "global_step": 104759, "epoch": 1177, "lr": 6.770397234078668e-05}
{"train_loss": 0.1919521987438202, "global_step": 104760, "epoch": 1177, "lr": 6.770343016102793e-05}
{"train_loss": 0.14657072722911835, "global_step": 104761, "epoch": 1177, "lr": 6.77028879788892e-05}
{"train_loss": 0.1803186982870102, "global_step": 104762, "epoch": 1177, "lr": 6.770234579437051e-05}
{"train_loss": 0.11057095974683762, "global_step": 104763, "epoch": 1177, "lr": 6.770180360747198e-05}
{"train_loss": 0.12107327580451965, "global_step": 104764, "epoch": 1177, "lr": 6.770126141819365e-05}
{"train_loss": 0.17459702491760254, "global_step": 104765, "epoch": 1177, "lr": 6.770071922653558e-05}
{"train_loss": 0.20493561029434204, "global_step": 104766, "epoch": 1177, "lr": 6.770017703249789e-05}
{"train_loss": 0.14670690894126892, "global_step": 104767, "epoch": 1177, "lr": 6.769963483608061e-05}
{"train_loss": 0.23385633528232574, "global_step": 104768, "epoch": 1177, "lr": 6.769909263728385e-05}
{"train_loss": 0.08623802661895752, "global_step": 104769, "epoch": 1177, "lr": 6.769855043610764e-05}
{"train_loss": 0.138116717338562, "global_step": 104770, "epoch": 1177, "lr": 6.769800823255208e-05}
{"train_loss": 0.1338282972574234, "global_step": 104771, "epoch": 1177, "lr": 6.769746602661725e-05}
{"train_loss": 0.2016907036304474, "global_step": 104772, "epoch": 1177, "lr": 6.76969238183032e-05}
{"train_loss": 0.15892302989959717, "global_step": 104773, "epoch": 1177, "lr": 6.769638160761002e-05}
{"train_loss": 0.17258785665035248, "global_step": 104774, "epoch": 1177, "lr": 6.769583939453779e-05}
{"train_loss": 0.12777140736579895, "global_step": 104775, "epoch": 1177, "lr": 6.769529717908655e-05}
{"train_loss": 0.1324242502450943, "global_step": 104776, "epoch": 1177, "lr": 6.76947549612564e-05}
{"train_loss": 0.13117843866348267, "global_step": 104777, "epoch": 1177, "lr": 6.76942127410474e-05}
{"train_loss": 0.15191876888275146, "global_step": 104778, "epoch": 1177, "lr": 6.769367051845963e-05}
{"train_loss": 0.16231226921081543, "global_step": 104779, "epoch": 1177, "lr": 6.769312829349317e-05}
{"train_loss": 0.17967021465301514, "global_step": 104780, "epoch": 1177, "lr": 6.769258606614807e-05}
{"train_loss": 0.11090289056301117, "global_step": 104781, "epoch": 1177, "lr": 6.769204383642444e-05}
{"train_loss": 0.1923312395811081, "global_step": 104782, "epoch": 1177, "lr": 6.769150160432231e-05}
{"train_loss": 0.11529253423213959, "global_step": 104783, "epoch": 1177, "lr": 6.76909593698418e-05}
{"train_loss": 0.12649717926979065, "global_step": 104784, "epoch": 1177, "lr": 6.769041713298292e-05}
{"train_loss": 0.183900848031044, "global_step": 104785, "epoch": 1177, "lr": 6.768987489374581e-05}
{"train_loss": 0.1973387897014618, "global_step": 104786, "epoch": 1177, "lr": 6.76893326521305e-05}
{"train_loss": 0.17608676850795746, "global_step": 104787, "epoch": 1177, "lr": 6.768879040813707e-05}
{"train_loss": 0.2227633148431778, "global_step": 104788, "epoch": 1177, "lr": 6.76882481617656e-05}
{"train_loss": 0.13362357020378113, "global_step": 104789, "epoch": 1177, "lr": 6.768770591301617e-05}
{"train_loss": 0.08878522366285324, "global_step": 104790, "epoch": 1177, "lr": 6.768716366188883e-05}
{"train_loss": 0.11388586461544037, "global_step": 104791, "epoch": 1177, "lr": 6.768662140838367e-05}
{"train_loss": 0.1287955641746521, "global_step": 104792, "epoch": 1177, "lr": 6.768607915250076e-05}
{"train_loss": 0.1159784272313118, "global_step": 104793, "epoch": 1177, "lr": 6.768553689424018e-05}
{"train_loss": 0.07473142445087433, "global_step": 104794, "epoch": 1177, "lr": 6.768499463360199e-05}
{"train_loss": 0.22354423999786377, "global_step": 104795, "epoch": 1177, "lr": 6.768445237058626e-05}
{"train_loss": 0.129658043384552, "global_step": 104796, "epoch": 1177, "lr": 6.768391010519308e-05}
{"train_loss": 0.2513178884983063, "global_step": 104797, "epoch": 1177, "lr": 6.76833678374225e-05}
{"train_loss": 0.15976865589618683, "global_step": 104798, "epoch": 1177, "lr": 6.768282556727464e-05}
{"train_loss": 0.12548647820949554, "global_step": 104799, "epoch": 1177, "lr": 6.768228329474951e-05}
{"train_loss": 0.15568403899669647, "global_step": 104800, "epoch": 1177, "lr": 6.768174101984721e-05}
{"train_loss": 0.15039831399917603, "global_step": 104801, "epoch": 1177, "lr": 6.768119874256783e-05}
{"train_loss": 0.1496814340353012, "global_step": 104802, "epoch": 1177, "lr": 6.768065646291142e-05}
{"train_loss": 0.22641269862651825, "global_step": 104803, "epoch": 1177, "lr": 6.768011418087808e-05}
{"train_loss": 0.047471266239881516, "global_step": 104804, "epoch": 1177, "lr": 6.767957189646785e-05}
{"train_loss": 0.13679246604442596, "global_step": 104805, "epoch": 1177, "lr": 6.76790296096808e-05}
{"train_loss": 0.13047382235527039, "global_step": 104806, "epoch": 1177, "lr": 6.767848732051705e-05}
{"train_loss": 0.15970756113529205, "global_step": 104807, "epoch": 1177, "lr": 6.767794502897662e-05}
{"train_loss": 0.26535874605178833, "global_step": 104808, "epoch": 1177, "lr": 6.767740273505961e-05}
{"train_loss": 0.11792884767055511, "global_step": 104809, "epoch": 1177, "lr": 6.76768604387661e-05}
{"train_loss": 0.1896050125360489, "global_step": 104810, "epoch": 1177, "lr": 6.767631814009614e-05}
{"train_loss": 0.05543680116534233, "global_step": 104811, "epoch": 1177, "lr": 6.767577583904983e-05}
{"train_loss": 0.12789003551006317, "global_step": 104812, "epoch": 1177, "lr": 6.767523353562722e-05}
{"train_loss": 0.1683511734008789, "global_step": 104813, "epoch": 1177, "lr": 6.767469122982838e-05}
{"train_loss": 0.15448646247386932, "global_step": 104814, "epoch": 1177, "lr": 6.76741489216534e-05}
{"train_loss": 0.16855603456497192, "global_step": 104815, "epoch": 1177, "lr": 6.767360661110235e-05}
{"train_loss": 0.24989843368530273, "global_step": 104816, "epoch": 1177, "lr": 6.767306429817529e-05}
{"train_loss": 0.1007549837231636, "global_step": 104817, "epoch": 1177, "lr": 6.76725219828723e-05}
{"train_loss": 0.10873327404260635, "global_step": 104818, "epoch": 1177, "lr": 6.767197966519347e-05}
{"train_loss": 0.13617661595344543, "global_step": 104819, "epoch": 1177, "lr": 6.767143734513886e-05}
{"train_loss": 0.17932192981243134, "global_step": 104820, "epoch": 1177, "lr": 6.767089502270853e-05}
{"train_loss": 0.1329554170370102, "global_step": 104821, "epoch": 1177, "lr": 6.767035269790256e-05}
{"train_loss": 0.14643840491771698, "global_step": 104822, "epoch": 1177, "lr": 6.766981037072103e-05}
{"train_loss": 0.135147362947464, "global_step": 104823, "epoch": 1177, "lr": 6.7669268041164e-05}
{"train_loss": 0.17486485838890076, "global_step": 104824, "epoch": 1177, "lr": 6.766872570923158e-05}
{"train_loss": 0.10679647326469421, "global_step": 104825, "epoch": 1177, "lr": 6.766818337492378e-05}
{"train_loss": 0.24790075421333313, "global_step": 104826, "epoch": 1177, "lr": 6.766764103824072e-05}
{"train_loss": 0.12841679155826569, "global_step": 104827, "epoch": 1177, "lr": 6.766709869918248e-05}
{"train_loss": 0.14913463592529297, "global_step": 104828, "epoch": 1177, "lr": 6.76665563577491e-05}
{"train_loss": 0.15362128615379333, "global_step": 104829, "epoch": 1177, "lr": 6.766601401394066e-05}
{"train_loss": 0.13612550497055054, "global_step": 104830, "epoch": 1177, "lr": 6.766547166775726e-05}
{"train_loss": 0.2523188292980194, "global_step": 104831, "epoch": 1177, "lr": 6.766492931919893e-05}
{"train_loss": 0.15925896167755127, "global_step": 104832, "epoch": 1177, "lr": 6.766438696826578e-05}
{"train_loss": 0.12190601229667664, "global_step": 104833, "epoch": 1177, "lr": 6.766384461495787e-05}
{"train_loss": 0.19991114735603333, "global_step": 104834, "epoch": 1177, "lr": 6.766330225927526e-05}
{"train_loss": 0.1854022741317749, "global_step": 104835, "epoch": 1177, "lr": 6.766275990121804e-05}
{"train_loss": 0.12333983927965164, "global_step": 104836, "epoch": 1177, "lr": 6.766221754078628e-05}
{"train_loss": 0.20850691199302673, "global_step": 104837, "epoch": 1177, "lr": 6.766167517798004e-05}
{"train_loss": 0.16021421551704407, "global_step": 104838, "epoch": 1177, "lr": 6.766113281279942e-05}
{"train_loss": 0.1341659128665924, "global_step": 104839, "epoch": 1177, "lr": 6.766059044524445e-05}
{"train_loss": 0.25394728779792786, "global_step": 104840, "epoch": 1177, "lr": 6.766004807531525e-05}
{"train_loss": 0.15562190634481024, "global_step": 104841, "epoch": 1177, "lr": 6.765950570301187e-05, "val_loss": 4.945606708526611}
{"train_loss": 0.11962904781103134, "global_step": 104842, "epoch": 1178, "lr": 6.765896332833437e-05}
{"train_loss": 0.14251938462257385, "global_step": 104843, "epoch": 1178, "lr": 6.765842095128286e-05}
{"train_loss": 0.137383371591568, "global_step": 104844, "epoch": 1178, "lr": 6.765787857185738e-05}
{"train_loss": 0.25044792890548706, "global_step": 104845, "epoch": 1178, "lr": 6.7657336190058e-05}
{"train_loss": 0.171993687748909, "global_step": 104846, "epoch": 1178, "lr": 6.765679380588484e-05}
{"train_loss": 0.17313386499881744, "global_step": 104847, "epoch": 1178, "lr": 6.765625141933791e-05}
{"train_loss": 0.13183212280273438, "global_step": 104848, "epoch": 1178, "lr": 6.765570903041731e-05}
{"train_loss": 0.11328595876693726, "global_step": 104849, "epoch": 1178, "lr": 6.765516663912313e-05}
{"train_loss": 0.13437364995479584, "global_step": 104850, "epoch": 1178, "lr": 6.765462424545542e-05}
{"train_loss": 0.19536635279655457, "global_step": 104851, "epoch": 1178, "lr": 6.765408184941426e-05}
{"train_loss": 0.1849762201309204, "global_step": 104852, "epoch": 1178, "lr": 6.765353945099974e-05}
{"train_loss": 0.14743061363697052, "global_step": 104853, "epoch": 1178, "lr": 6.76529970502119e-05}
{"train_loss": 0.23362597823143005, "global_step": 104854, "epoch": 1178, "lr": 6.765245464705084e-05}
{"train_loss": 0.12531153857707977, "global_step": 104855, "epoch": 1178, "lr": 6.765191224151661e-05}
{"train_loss": 0.10904679447412491, "global_step": 104856, "epoch": 1178, "lr": 6.765136983360931e-05}
{"train_loss": 0.1456892043352127, "global_step": 104857, "epoch": 1178, "lr": 6.7650827423329e-05}
{"train_loss": 0.06593441963195801, "global_step": 104858, "epoch": 1178, "lr": 6.765028501067572e-05}
{"train_loss": 0.11431978642940521, "global_step": 104859, "epoch": 1178, "lr": 6.76497425956496e-05}
{"train_loss": 0.1763603240251541, "global_step": 104860, "epoch": 1178, "lr": 6.764920017825069e-05}
{"train_loss": 0.11453869193792343, "global_step": 104861, "epoch": 1178, "lr": 6.764865775847906e-05}
{"train_loss": 0.1310972422361374, "global_step": 104862, "epoch": 1178, "lr": 6.764811533633478e-05}
{"train_loss": 0.08601921796798706, "global_step": 104863, "epoch": 1178, "lr": 6.764757291181792e-05}
{"train_loss": 0.17851804196834564, "global_step": 104864, "epoch": 1178, "lr": 6.764703048492856e-05}
{"train_loss": 0.19606074690818787, "global_step": 104865, "epoch": 1178, "lr": 6.764648805566679e-05}
{"train_loss": 0.06730689853429794, "global_step": 104866, "epoch": 1178, "lr": 6.764594562403263e-05}
{"train_loss": 0.14587005972862244, "global_step": 104867, "epoch": 1178, "lr": 6.764540319002621e-05}
{"train_loss": 0.18636326491832733, "global_step": 104868, "epoch": 1178, "lr": 6.764486075364758e-05}
{"train_loss": 0.1553882211446762, "global_step": 104869, "epoch": 1178, "lr": 6.76443183148968e-05}
{"train_loss": 0.1381094753742218, "global_step": 104870, "epoch": 1178, "lr": 6.764377587377398e-05}
{"train_loss": 0.08062101900577545, "global_step": 104871, "epoch": 1178, "lr": 6.764323343027916e-05}
{"train_loss": 0.1328650861978531, "global_step": 104872, "epoch": 1178, "lr": 6.764269098441242e-05}
{"train_loss": 0.14199334383010864, "global_step": 104873, "epoch": 1178, "lr": 6.764214853617383e-05}
{"train_loss": 0.2333538681268692, "global_step": 104874, "epoch": 1178, "lr": 6.764160608556346e-05}
{"train_loss": 0.17443248629570007, "global_step": 104875, "epoch": 1178, "lr": 6.764106363258141e-05}
{"train_loss": 0.09290848672389984, "global_step": 104876, "epoch": 1178, "lr": 6.764052117722773e-05}
{"train_loss": 0.1653517782688141, "global_step": 104877, "epoch": 1178, "lr": 6.763997871950249e-05}
{"train_loss": 0.1245911493897438, "global_step": 104878, "epoch": 1178, "lr": 6.763943625940579e-05}
{"train_loss": 0.14603658020496368, "global_step": 104879, "epoch": 1178, "lr": 6.763889379693766e-05}
{"train_loss": 0.23982112109661102, "global_step": 104880, "epoch": 1178, "lr": 6.76383513320982e-05}
{"train_loss": 0.16307947039604187, "global_step": 104881, "epoch": 1178, "lr": 6.763780886488748e-05}
{"train_loss": 0.17304520308971405, "global_step": 104882, "epoch": 1178, "lr": 6.763726639530557e-05}
{"train_loss": 0.11569604277610779, "global_step": 104883, "epoch": 1178, "lr": 6.763672392335257e-05}
{"train_loss": 0.17557096481323242, "global_step": 104884, "epoch": 1178, "lr": 6.763618144902849e-05}
{"train_loss": 0.15650707483291626, "global_step": 104885, "epoch": 1178, "lr": 6.763563897233346e-05}
{"train_loss": 0.1382681429386139, "global_step": 104886, "epoch": 1178, "lr": 6.763509649326752e-05}
{"train_loss": 0.18109145760536194, "global_step": 104887, "epoch": 1178, "lr": 6.763455401183077e-05}
{"train_loss": 0.19549554586410522, "global_step": 104888, "epoch": 1178, "lr": 6.763401152802326e-05}
{"train_loss": 0.22915786504745483, "global_step": 104889, "epoch": 1178, "lr": 6.763346904184509e-05}
{"train_loss": 0.09533129632472992, "global_step": 104890, "epoch": 1178, "lr": 6.763292655329629e-05}
{"train_loss": 0.14766815304756165, "global_step": 104891, "epoch": 1178, "lr": 6.763238406237697e-05}
{"train_loss": 0.1664659082889557, "global_step": 104892, "epoch": 1178, "lr": 6.76318415690872e-05}
{"train_loss": 0.1534658521413803, "global_step": 104893, "epoch": 1178, "lr": 6.763129907342702e-05}
{"train_loss": 0.161707803606987, "global_step": 104894, "epoch": 1178, "lr": 6.763075657539654e-05}
{"train_loss": 0.11325442045927048, "global_step": 104895, "epoch": 1178, "lr": 6.763021407499582e-05}
{"train_loss": 0.17981265485286713, "global_step": 104896, "epoch": 1178, "lr": 6.762967157222494e-05}
{"train_loss": 0.18947705626487732, "global_step": 104897, "epoch": 1178, "lr": 6.762912906708396e-05}
{"train_loss": 0.11022454500198364, "global_step": 104898, "epoch": 1178, "lr": 6.762858655957296e-05}
{"train_loss": 0.13887889683246613, "global_step": 104899, "epoch": 1178, "lr": 6.7628044049692e-05}
{"train_loss": 0.1933513730764389, "global_step": 104900, "epoch": 1178, "lr": 6.762750153744117e-05}
{"train_loss": 0.16788101196289062, "global_step": 104901, "epoch": 1178, "lr": 6.762695902282054e-05}
{"train_loss": 0.17948542535305023, "global_step": 104902, "epoch": 1178, "lr": 6.762641650583018e-05}
{"train_loss": 0.19547811150550842, "global_step": 104903, "epoch": 1178, "lr": 6.762587398647017e-05}
{"train_loss": 0.1726256012916565, "global_step": 104904, "epoch": 1178, "lr": 6.762533146474056e-05}
{"train_loss": 0.13927555084228516, "global_step": 104905, "epoch": 1178, "lr": 6.762478894064144e-05}
{"train_loss": 0.12085264176130295, "global_step": 104906, "epoch": 1178, "lr": 6.76242464141729e-05}
{"train_loss": 0.1715611219406128, "global_step": 104907, "epoch": 1178, "lr": 6.762370388533497e-05}
{"train_loss": 0.1105160266160965, "global_step": 104908, "epoch": 1178, "lr": 6.762316135412777e-05}
{"train_loss": 0.22590550780296326, "global_step": 104909, "epoch": 1178, "lr": 6.762261882055136e-05}
{"train_loss": 0.15185439586639404, "global_step": 104910, "epoch": 1178, "lr": 6.762207628460579e-05}
{"train_loss": 0.1632358282804489, "global_step": 104911, "epoch": 1178, "lr": 6.762153374629114e-05}
{"train_loss": 0.15090756118297577, "global_step": 104912, "epoch": 1178, "lr": 6.76209912056075e-05}
{"train_loss": 0.3079543709754944, "global_step": 104913, "epoch": 1178, "lr": 6.762044866255492e-05}
{"train_loss": 0.16580165922641754, "global_step": 104914, "epoch": 1178, "lr": 6.761990611713349e-05}
{"train_loss": 0.18600623309612274, "global_step": 104915, "epoch": 1178, "lr": 6.761936356934328e-05}
{"train_loss": 0.2554025650024414, "global_step": 104916, "epoch": 1178, "lr": 6.761882101918436e-05}
{"train_loss": 0.2361115664243698, "global_step": 104917, "epoch": 1178, "lr": 6.76182784666568e-05}
{"train_loss": 0.18340742588043213, "global_step": 104918, "epoch": 1178, "lr": 6.761773591176069e-05}
{"train_loss": 0.17927582561969757, "global_step": 104919, "epoch": 1178, "lr": 6.761719335449608e-05}
{"train_loss": 0.21252217888832092, "global_step": 104920, "epoch": 1178, "lr": 6.761665079486305e-05}
{"train_loss": 0.29611098766326904, "global_step": 104921, "epoch": 1178, "lr": 6.761610823286169e-05}
{"train_loss": 0.16691187024116516, "global_step": 104922, "epoch": 1178, "lr": 6.761556566849204e-05}
{"train_loss": 0.1594703197479248, "global_step": 104923, "epoch": 1178, "lr": 6.76150231017542e-05}
{"train_loss": 0.16908879578113556, "global_step": 104924, "epoch": 1178, "lr": 6.761448053264824e-05}
{"train_loss": 0.17551206052303314, "global_step": 104925, "epoch": 1178, "lr": 6.761393796117423e-05}
{"train_loss": 0.09141003340482712, "global_step": 104926, "epoch": 1178, "lr": 6.761339538733224e-05}
{"train_loss": 0.08860873430967331, "global_step": 104927, "epoch": 1178, "lr": 6.761285281112233e-05}
{"train_loss": 0.21733900904655457, "global_step": 104928, "epoch": 1178, "lr": 6.761231023254461e-05}
{"train_loss": 0.16580288112163544, "global_step": 104929, "epoch": 1178, "lr": 6.761176765159911e-05}
{"train_loss": 0.16156864978289337, "global_step": 104930, "epoch": 1178, "lr": 6.761122506828594e-05, "val_loss": 4.8324127197265625}
{"train_loss": 0.19996517896652222, "global_step": 104931, "epoch": 1179, "lr": 6.761068248260514e-05}
{"train_loss": 0.21318425238132477, "global_step": 104932, "epoch": 1179, "lr": 6.76101398945568e-05}
{"train_loss": 0.17776310443878174, "global_step": 104933, "epoch": 1179, "lr": 6.7609597304141e-05}
{"train_loss": 0.22298604249954224, "global_step": 104934, "epoch": 1179, "lr": 6.76090547113578e-05}
{"train_loss": 0.10007549077272415, "global_step": 104935, "epoch": 1179, "lr": 6.760851211620726e-05}
{"train_loss": 0.14269520342350006, "global_step": 104936, "epoch": 1179, "lr": 6.760796951868948e-05}
{"train_loss": 0.1544787883758545, "global_step": 104937, "epoch": 1179, "lr": 6.760742691880454e-05}
{"train_loss": 0.13744960725307465, "global_step": 104938, "epoch": 1179, "lr": 6.760688431655248e-05}
{"train_loss": 0.1712826043367386, "global_step": 104939, "epoch": 1179, "lr": 6.76063417119334e-05}
{"train_loss": 0.1560625582933426, "global_step": 104940, "epoch": 1179, "lr": 6.760579910494735e-05}
{"train_loss": 0.13742798566818237, "global_step": 104941, "epoch": 1179, "lr": 6.760525649559442e-05}
{"train_loss": 0.12094927579164505, "global_step": 104942, "epoch": 1179, "lr": 6.760471388387467e-05}
{"train_loss": 0.18824748694896698, "global_step": 104943, "epoch": 1179, "lr": 6.760417126978819e-05}
{"train_loss": 0.16613084077835083, "global_step": 104944, "epoch": 1179, "lr": 6.760362865333503e-05}
{"train_loss": 0.13767163455486298, "global_step": 104945, "epoch": 1179, "lr": 6.760308603451531e-05}
{"train_loss": 0.14204047620296478, "global_step": 104946, "epoch": 1179, "lr": 6.760254341332903e-05}
{"train_loss": 0.14673544466495514, "global_step": 104947, "epoch": 1179, "lr": 6.760200078977633e-05}
{"train_loss": 0.2186124175786972, "global_step": 104948, "epoch": 1179, "lr": 6.760145816385725e-05}
{"train_loss": 0.1297604888677597, "global_step": 104949, "epoch": 1179, "lr": 6.760091553557188e-05}
{"train_loss": 0.13858556747436523, "global_step": 104950, "epoch": 1179, "lr": 6.760037290492026e-05}
{"train_loss": 0.1489243358373642, "global_step": 104951, "epoch": 1179, "lr": 6.759983027190249e-05}
{"train_loss": 0.1668769121170044, "global_step": 104952, "epoch": 1179, "lr": 6.759928763651864e-05}
{"train_loss": 0.27962398529052734, "global_step": 104953, "epoch": 1179, "lr": 6.759874499876878e-05}
{"train_loss": 0.10439843684434891, "global_step": 104954, "epoch": 1179, "lr": 6.759820235865299e-05}
{"train_loss": 0.18332572281360626, "global_step": 104955, "epoch": 1179, "lr": 6.759765971617131e-05}
{"train_loss": 0.11269938200712204, "global_step": 104956, "epoch": 1179, "lr": 6.759711707132386e-05}
{"train_loss": 0.180527001619339, "global_step": 104957, "epoch": 1179, "lr": 6.759657442411069e-05}
{"train_loss": 0.1067429855465889, "global_step": 104958, "epoch": 1179, "lr": 6.75960317745319e-05}
{"train_loss": 0.2121001034975052, "global_step": 104959, "epoch": 1179, "lr": 6.759548912258749e-05}
{"train_loss": 0.17557379603385925, "global_step": 104960, "epoch": 1179, "lr": 6.75949464682776e-05}
{"train_loss": 0.18785639107227325, "global_step": 104961, "epoch": 1179, "lr": 6.759440381160228e-05}
{"train_loss": 0.13371838629245758, "global_step": 104962, "epoch": 1179, "lr": 6.759386115256163e-05}
{"train_loss": 0.17337457835674286, "global_step": 104963, "epoch": 1179, "lr": 6.759331849115566e-05}
{"train_loss": 0.17804642021656036, "global_step": 104964, "epoch": 1179, "lr": 6.759277582738453e-05}
{"train_loss": 0.1942383199930191, "global_step": 104965, "epoch": 1179, "lr": 6.759223316124823e-05}
{"train_loss": 0.2125670611858368, "global_step": 104966, "epoch": 1179, "lr": 6.759169049274688e-05}
{"train_loss": 0.14250117540359497, "global_step": 104967, "epoch": 1179, "lr": 6.759114782188056e-05}
{"train_loss": 0.20038236677646637, "global_step": 104968, "epoch": 1179, "lr": 6.75906051486493e-05}
{"train_loss": 0.15430445969104767, "global_step": 104969, "epoch": 1179, "lr": 6.75900624730532e-05}
{"train_loss": 0.15340803563594818, "global_step": 104970, "epoch": 1179, "lr": 6.758951979509235e-05}
{"train_loss": 0.3101487457752228, "global_step": 104971, "epoch": 1179, "lr": 6.758897711476679e-05}
{"train_loss": 0.18632355332374573, "global_step": 104972, "epoch": 1179, "lr": 6.758843443207659e-05}
{"train_loss": 0.13905008137226105, "global_step": 104973, "epoch": 1179, "lr": 6.758789174702188e-05}
{"train_loss": 0.14025871455669403, "global_step": 104974, "epoch": 1179, "lr": 6.758734905960265e-05}
{"train_loss": 0.18658629059791565, "global_step": 104975, "epoch": 1179, "lr": 6.758680636981903e-05}
{"train_loss": 0.1061129942536354, "global_step": 104976, "epoch": 1179, "lr": 6.758626367767107e-05}
{"train_loss": 0.22191546857357025, "global_step": 104977, "epoch": 1179, "lr": 6.758572098315886e-05}
{"train_loss": 0.15629680454730988, "global_step": 104978, "epoch": 1179, "lr": 6.758517828628246e-05}
{"train_loss": 0.17535701394081116, "global_step": 104979, "epoch": 1179, "lr": 6.758463558704196e-05}
{"train_loss": 0.1487554907798767, "global_step": 104980, "epoch": 1179, "lr": 6.758409288543741e-05}
{"train_loss": 0.13768959045410156, "global_step": 104981, "epoch": 1179, "lr": 6.75835501814689e-05}
{"train_loss": 0.19056198000907898, "global_step": 104982, "epoch": 1179, "lr": 6.75830074751365e-05}
{"train_loss": 0.16023209691047668, "global_step": 104983, "epoch": 1179, "lr": 6.758246476644026e-05}
{"train_loss": 0.14242178201675415, "global_step": 104984, "epoch": 1179, "lr": 6.758192205538029e-05}
{"train_loss": 0.10773657262325287, "global_step": 104985, "epoch": 1179, "lr": 6.758137934195664e-05}
{"train_loss": 0.22089731693267822, "global_step": 104986, "epoch": 1179, "lr": 6.75808366261694e-05}
{"train_loss": 0.1063460111618042, "global_step": 104987, "epoch": 1179, "lr": 6.758029390801861e-05}
{"train_loss": 0.18450979888439178, "global_step": 104988, "epoch": 1179, "lr": 6.757975118750437e-05}
{"train_loss": 0.14839060604572296, "global_step": 104989, "epoch": 1179, "lr": 6.757920846462676e-05}
{"train_loss": 0.1722220629453659, "global_step": 104990, "epoch": 1179, "lr": 6.757866573938582e-05}
{"train_loss": 0.23194676637649536, "global_step": 104991, "epoch": 1179, "lr": 6.757812301178165e-05}
{"train_loss": 0.10768641531467438, "global_step": 104992, "epoch": 1179, "lr": 6.75775802818143e-05}
{"train_loss": 0.14509645104408264, "global_step": 104993, "epoch": 1179, "lr": 6.757703754948387e-05}
{"train_loss": 0.14964090287685394, "global_step": 104994, "epoch": 1179, "lr": 6.757649481479044e-05}
{"train_loss": 0.15047147870063782, "global_step": 104995, "epoch": 1179, "lr": 6.757595207773404e-05}
{"train_loss": 0.18559050559997559, "global_step": 104996, "epoch": 1179, "lr": 6.757540933831477e-05}
{"train_loss": 0.15650531649589539, "global_step": 104997, "epoch": 1179, "lr": 6.757486659653272e-05}
{"train_loss": 0.13270774483680725, "global_step": 104998, "epoch": 1179, "lr": 6.757432385238793e-05}
{"train_loss": 0.15273703634738922, "global_step": 104999, "epoch": 1179, "lr": 6.757378110588047e-05}
{"train_loss": 0.12109728902578354, "global_step": 105000, "epoch": 1179, "lr": 6.757323835701046e-05}
{"train_loss": 0.15847322344779968, "global_step": 105001, "epoch": 1179, "lr": 6.757269560577793e-05}
{"train_loss": 0.11925480514764786, "global_step": 105002, "epoch": 1179, "lr": 6.757215285218296e-05}
{"train_loss": 0.19170443713665009, "global_step": 105003, "epoch": 1179, "lr": 6.757161009622564e-05}
{"train_loss": 0.16690149903297424, "global_step": 105004, "epoch": 1179, "lr": 6.757106733790601e-05}
{"train_loss": 0.1376270055770874, "global_step": 105005, "epoch": 1179, "lr": 6.757052457722418e-05}
{"train_loss": 0.15879487991333008, "global_step": 105006, "epoch": 1179, "lr": 6.756998181418021e-05}
{"train_loss": 0.14867183566093445, "global_step": 105007, "epoch": 1179, "lr": 6.756943904877417e-05}
{"train_loss": 0.10546061396598816, "global_step": 105008, "epoch": 1179, "lr": 6.756889628100613e-05}
{"train_loss": 0.13515357673168182, "global_step": 105009, "epoch": 1179, "lr": 6.756835351087616e-05}
{"train_loss": 0.135311558842659, "global_step": 105010, "epoch": 1179, "lr": 6.756781073838434e-05}
{"train_loss": 0.14687319099903107, "global_step": 105011, "epoch": 1179, "lr": 6.756726796353074e-05}
{"train_loss": 0.18280236423015594, "global_step": 105012, "epoch": 1179, "lr": 6.756672518631545e-05}
{"train_loss": 0.07484568655490875, "global_step": 105013, "epoch": 1179, "lr": 6.756618240673851e-05}
{"train_loss": 0.14310388267040253, "global_step": 105014, "epoch": 1179, "lr": 6.756563962480002e-05}
{"train_loss": 0.25920528173446655, "global_step": 105015, "epoch": 1179, "lr": 6.756509684050004e-05}
{"train_loss": 0.13123856484889984, "global_step": 105016, "epoch": 1179, "lr": 6.756455405383865e-05}
{"train_loss": 0.11220673471689224, "global_step": 105017, "epoch": 1179, "lr": 6.756401126481592e-05}
{"train_loss": 0.20010969042778015, "global_step": 105018, "epoch": 1179, "lr": 6.756346847343191e-05}
{"train_loss": 0.16080253588014773, "global_step": 105019, "epoch": 1179, "lr": 6.756292567968673e-05, "val_loss": 4.8428850173950195}
{"train_loss": 0.20275168120861053, "global_step": 105020, "epoch": 1180, "lr": 6.75623828835804e-05}
{"train_loss": 0.09685225039720535, "global_step": 105021, "epoch": 1180, "lr": 6.756184008511305e-05}
{"train_loss": 0.13393546640872955, "global_step": 105022, "epoch": 1180, "lr": 6.756129728428471e-05}
{"train_loss": 0.11551257222890854, "global_step": 105023, "epoch": 1180, "lr": 6.756075448109546e-05}
{"train_loss": 0.1040688082575798, "global_step": 105024, "epoch": 1180, "lr": 6.75602116755454e-05}
{"train_loss": 0.18699108064174652, "global_step": 105025, "epoch": 1180, "lr": 6.755966886763459e-05}
{"train_loss": 0.19831110537052155, "global_step": 105026, "epoch": 1180, "lr": 6.755912605736307e-05}
{"train_loss": 0.14291884005069733, "global_step": 105027, "epoch": 1180, "lr": 6.755858324473096e-05}
{"train_loss": 0.12731508910655975, "global_step": 105028, "epoch": 1180, "lr": 6.75580404297383e-05}
{"train_loss": 0.09930787235498428, "global_step": 105029, "epoch": 1180, "lr": 6.755749761238518e-05}
{"train_loss": 0.1424730271100998, "global_step": 105030, "epoch": 1180, "lr": 6.755695479267167e-05}
{"train_loss": 0.1315753161907196, "global_step": 105031, "epoch": 1180, "lr": 6.755641197059783e-05}
{"train_loss": 0.21644073724746704, "global_step": 105032, "epoch": 1180, "lr": 6.755586914616374e-05}
{"train_loss": 0.1281093806028366, "global_step": 105033, "epoch": 1180, "lr": 6.755532631936949e-05}
{"train_loss": 0.17871692776679993, "global_step": 105034, "epoch": 1180, "lr": 6.755478349021514e-05}
{"train_loss": 0.17101134359836578, "global_step": 105035, "epoch": 1180, "lr": 6.755424065870077e-05}
{"train_loss": 0.16261504590511322, "global_step": 105036, "epoch": 1180, "lr": 6.755369782482644e-05}
{"train_loss": 0.13890619575977325, "global_step": 105037, "epoch": 1180, "lr": 6.755315498859223e-05}
{"train_loss": 0.14859634637832642, "global_step": 105038, "epoch": 1180, "lr": 6.755261214999822e-05}
{"train_loss": 0.15767109394073486, "global_step": 105039, "epoch": 1180, "lr": 6.755206930904446e-05}
{"train_loss": 0.1738080531358719, "global_step": 105040, "epoch": 1180, "lr": 6.755152646573104e-05}
{"train_loss": 0.12469295412302017, "global_step": 105041, "epoch": 1180, "lr": 6.755098362005805e-05}
{"train_loss": 0.17234507203102112, "global_step": 105042, "epoch": 1180, "lr": 6.755044077202553e-05}
{"train_loss": 0.22410987317562103, "global_step": 105043, "epoch": 1180, "lr": 6.754989792163356e-05}
{"train_loss": 0.06104470044374466, "global_step": 105044, "epoch": 1180, "lr": 6.754935506888224e-05}
{"train_loss": 0.11875123530626297, "global_step": 105045, "epoch": 1180, "lr": 6.75488122137716e-05}
{"train_loss": 0.17996466159820557, "global_step": 105046, "epoch": 1180, "lr": 6.754826935630175e-05}
{"train_loss": 0.1234285905957222, "global_step": 105047, "epoch": 1180, "lr": 6.754772649647276e-05}
{"train_loss": 0.20389236509799957, "global_step": 105048, "epoch": 1180, "lr": 6.754718363428467e-05}
{"train_loss": 0.16504600644111633, "global_step": 105049, "epoch": 1180, "lr": 6.754664076973757e-05}
{"train_loss": 0.19426940381526947, "global_step": 105050, "epoch": 1180, "lr": 6.754609790283155e-05}
{"train_loss": 0.13088391721248627, "global_step": 105051, "epoch": 1180, "lr": 6.754555503356668e-05}
{"train_loss": 0.12957191467285156, "global_step": 105052, "epoch": 1180, "lr": 6.754501216194301e-05}
{"train_loss": 0.12601572275161743, "global_step": 105053, "epoch": 1180, "lr": 6.754446928796063e-05}
{"train_loss": 0.11414790898561478, "global_step": 105054, "epoch": 1180, "lr": 6.754392641161962e-05}
{"train_loss": 0.1747778058052063, "global_step": 105055, "epoch": 1180, "lr": 6.754338353292002e-05}
{"train_loss": 0.14743129909038544, "global_step": 105056, "epoch": 1180, "lr": 6.754284065186195e-05}
{"train_loss": 0.16590575873851776, "global_step": 105057, "epoch": 1180, "lr": 6.754229776844545e-05}
{"train_loss": 0.14031092822551727, "global_step": 105058, "epoch": 1180, "lr": 6.754175488267061e-05}
{"train_loss": 0.20882827043533325, "global_step": 105059, "epoch": 1180, "lr": 6.754121199453748e-05}
{"train_loss": 0.0979498028755188, "global_step": 105060, "epoch": 1180, "lr": 6.754066910404616e-05}
{"train_loss": 0.1518813520669937, "global_step": 105061, "epoch": 1180, "lr": 6.754012621119669e-05}
{"train_loss": 0.11618376523256302, "global_step": 105062, "epoch": 1180, "lr": 6.75395833159892e-05}
{"train_loss": 0.15617352724075317, "global_step": 105063, "epoch": 1180, "lr": 6.75390404184237e-05}
{"train_loss": 0.1366627812385559, "global_step": 105064, "epoch": 1180, "lr": 6.75384975185003e-05}
{"train_loss": 0.1125558540225029, "global_step": 105065, "epoch": 1180, "lr": 6.753795461621907e-05}
{"train_loss": 0.21954406797885895, "global_step": 105066, "epoch": 1180, "lr": 6.753741171158006e-05}
{"train_loss": 0.09406507760286331, "global_step": 105067, "epoch": 1180, "lr": 6.753686880458337e-05}
{"train_loss": 0.12314672023057938, "global_step": 105068, "epoch": 1180, "lr": 6.753632589522904e-05}
{"train_loss": 0.10891041159629822, "global_step": 105069, "epoch": 1180, "lr": 6.753578298351717e-05}
{"train_loss": 0.13453896343708038, "global_step": 105070, "epoch": 1180, "lr": 6.753524006944785e-05}
{"train_loss": 0.13100212812423706, "global_step": 105071, "epoch": 1180, "lr": 6.753469715302112e-05}
{"train_loss": 0.17360690236091614, "global_step": 105072, "epoch": 1180, "lr": 6.753415423423706e-05}
{"train_loss": 0.11902357637882233, "global_step": 105073, "epoch": 1180, "lr": 6.753361131309574e-05}
{"train_loss": 0.10791067034006119, "global_step": 105074, "epoch": 1180, "lr": 6.753306838959725e-05}
{"train_loss": 0.23329047858715057, "global_step": 105075, "epoch": 1180, "lr": 6.753252546374165e-05}
{"train_loss": 0.13463647663593292, "global_step": 105076, "epoch": 1180, "lr": 6.753198253552901e-05}
{"train_loss": 0.13768115639686584, "global_step": 105077, "epoch": 1180, "lr": 6.753143960495942e-05}
{"train_loss": 0.0848219096660614, "global_step": 105078, "epoch": 1180, "lr": 6.753089667203294e-05}
{"train_loss": 0.11544758081436157, "global_step": 105079, "epoch": 1180, "lr": 6.753035373674965e-05}
{"train_loss": 0.08797464519739151, "global_step": 105080, "epoch": 1180, "lr": 6.752981079910961e-05}
{"train_loss": 0.15125791728496552, "global_step": 105081, "epoch": 1180, "lr": 6.752926785911288e-05}
{"train_loss": 0.08408643305301666, "global_step": 105082, "epoch": 1180, "lr": 6.752872491675959e-05}
{"train_loss": 0.10319835692644119, "global_step": 105083, "epoch": 1180, "lr": 6.752818197204977e-05}
{"train_loss": 0.17780686914920807, "global_step": 105084, "epoch": 1180, "lr": 6.752763902498349e-05}
{"train_loss": 0.15444521605968475, "global_step": 105085, "epoch": 1180, "lr": 6.752709607556083e-05}
{"train_loss": 0.09869974851608276, "global_step": 105086, "epoch": 1180, "lr": 6.752655312378186e-05}
{"train_loss": 0.17162679135799408, "global_step": 105087, "epoch": 1180, "lr": 6.752601016964666e-05}
{"train_loss": 0.19700585305690765, "global_step": 105088, "epoch": 1180, "lr": 6.752546721315532e-05}
{"train_loss": 0.12918803095817566, "global_step": 105089, "epoch": 1180, "lr": 6.752492425430788e-05}
{"train_loss": 0.1250438094139099, "global_step": 105090, "epoch": 1180, "lr": 6.752438129310442e-05}
{"train_loss": 0.13408797979354858, "global_step": 105091, "epoch": 1180, "lr": 6.752383832954504e-05}
{"train_loss": 0.1747790277004242, "global_step": 105092, "epoch": 1180, "lr": 6.752329536362978e-05}
{"train_loss": 0.10738682001829147, "global_step": 105093, "epoch": 1180, "lr": 6.752275239535873e-05}
{"train_loss": 0.11344747990369797, "global_step": 105094, "epoch": 1180, "lr": 6.752220942473196e-05}
{"train_loss": 0.12905803322792053, "global_step": 105095, "epoch": 1180, "lr": 6.752166645174955e-05}
{"train_loss": 0.11934057623147964, "global_step": 105096, "epoch": 1180, "lr": 6.752112347641156e-05}
{"train_loss": 0.15257909893989563, "global_step": 105097, "epoch": 1180, "lr": 6.752058049871805e-05}
{"train_loss": 0.14429329335689545, "global_step": 105098, "epoch": 1180, "lr": 6.752003751866914e-05}
{"train_loss": 0.16639652848243713, "global_step": 105099, "epoch": 1180, "lr": 6.751949453626486e-05}
{"train_loss": 0.1049499660730362, "global_step": 105100, "epoch": 1180, "lr": 6.751895155150529e-05}
{"train_loss": 0.1609007567167282, "global_step": 105101, "epoch": 1180, "lr": 6.751840856439052e-05}
{"train_loss": 0.1631566286087036, "global_step": 105102, "epoch": 1180, "lr": 6.751786557492064e-05}
{"train_loss": 0.12374920397996902, "global_step": 105103, "epoch": 1180, "lr": 6.751732258309566e-05}
{"train_loss": 0.1394757479429245, "global_step": 105104, "epoch": 1180, "lr": 6.751677958891569e-05}
{"train_loss": 0.23548758029937744, "global_step": 105105, "epoch": 1180, "lr": 6.751623659238081e-05}
{"train_loss": 0.16010946035385132, "global_step": 105106, "epoch": 1180, "lr": 6.751569359349109e-05}
{"train_loss": 0.16564145684242249, "global_step": 105107, "epoch": 1180, "lr": 6.75151505922466e-05}
{"train_loss": 0.14440351841824778, "global_step": 105108, "epoch": 1180, "lr": 6.751460758864741e-05, "val_loss": 4.991012096405029, "train_action_mse_error": 19.351425170898438}
{"train_loss": 0.1386619508266449, "global_step": 105109, "epoch": 1181, "lr": 6.751406458269358e-05}
{"train_loss": 0.15933619439601898, "global_step": 105110, "epoch": 1181, "lr": 6.751352157438522e-05}
{"train_loss": 0.09343612939119339, "global_step": 105111, "epoch": 1181, "lr": 6.751297856372236e-05}
{"train_loss": 0.1166197806596756, "global_step": 105112, "epoch": 1181, "lr": 6.751243555070511e-05}
{"train_loss": 0.16275015473365784, "global_step": 105113, "epoch": 1181, "lr": 6.751189253533352e-05}
{"train_loss": 0.17034609615802765, "global_step": 105114, "epoch": 1181, "lr": 6.751134951760767e-05}
{"train_loss": 0.179601788520813, "global_step": 105115, "epoch": 1181, "lr": 6.751080649752764e-05}
{"train_loss": 0.09981308877468109, "global_step": 105116, "epoch": 1181, "lr": 6.751026347509348e-05}
{"train_loss": 0.1420782506465912, "global_step": 105117, "epoch": 1181, "lr": 6.75097204503053e-05}
{"train_loss": 0.17877919971942902, "global_step": 105118, "epoch": 1181, "lr": 6.750917742316312e-05}
{"train_loss": 0.1961439698934555, "global_step": 105119, "epoch": 1181, "lr": 6.750863439366707e-05}
{"train_loss": 0.19934210181236267, "global_step": 105120, "epoch": 1181, "lr": 6.750809136181719e-05}
{"train_loss": 0.17629146575927734, "global_step": 105121, "epoch": 1181, "lr": 6.750754832761356e-05}
{"train_loss": 0.14976342022418976, "global_step": 105122, "epoch": 1181, "lr": 6.750700529105625e-05}
{"train_loss": 0.1532302051782608, "global_step": 105123, "epoch": 1181, "lr": 6.750646225214535e-05}
{"train_loss": 0.15497446060180664, "global_step": 105124, "epoch": 1181, "lr": 6.75059192108809e-05}
{"train_loss": 0.12177097052335739, "global_step": 105125, "epoch": 1181, "lr": 6.7505376167263e-05}
{"train_loss": 0.21103546023368835, "global_step": 105126, "epoch": 1181, "lr": 6.750483312129172e-05}
{"train_loss": 0.1744970977306366, "global_step": 105127, "epoch": 1181, "lr": 6.750429007296712e-05}
{"train_loss": 0.14728707075119019, "global_step": 105128, "epoch": 1181, "lr": 6.75037470222893e-05}
{"train_loss": 0.17219117283821106, "global_step": 105129, "epoch": 1181, "lr": 6.750320396925829e-05}
{"train_loss": 0.17898960411548615, "global_step": 105130, "epoch": 1181, "lr": 6.750266091387421e-05}
{"train_loss": 0.12351005524396896, "global_step": 105131, "epoch": 1181, "lr": 6.75021178561371e-05}
{"train_loss": 0.1730223000049591, "global_step": 105132, "epoch": 1181, "lr": 6.750157479604704e-05}
{"train_loss": 0.12051042914390564, "global_step": 105133, "epoch": 1181, "lr": 6.750103173360411e-05}
{"train_loss": 0.08276229351758957, "global_step": 105134, "epoch": 1181, "lr": 6.750048866880838e-05}
{"train_loss": 0.1206827461719513, "global_step": 105135, "epoch": 1181, "lr": 6.749994560165993e-05}
{"train_loss": 0.15277495980262756, "global_step": 105136, "epoch": 1181, "lr": 6.749940253215882e-05}
{"train_loss": 0.16967502236366272, "global_step": 105137, "epoch": 1181, "lr": 6.749885946030511e-05}
{"train_loss": 0.11338865011930466, "global_step": 105138, "epoch": 1181, "lr": 6.749831638609893e-05}
{"train_loss": 0.14911112189292908, "global_step": 105139, "epoch": 1181, "lr": 6.749777330954028e-05}
{"train_loss": 0.11802610009908676, "global_step": 105140, "epoch": 1181, "lr": 6.749723023062929e-05}
{"train_loss": 0.12984099984169006, "global_step": 105141, "epoch": 1181, "lr": 6.7496687149366e-05}
{"train_loss": 0.19049127399921417, "global_step": 105142, "epoch": 1181, "lr": 6.749614406575049e-05}
{"train_loss": 0.1931585669517517, "global_step": 105143, "epoch": 1181, "lr": 6.749560097978284e-05}
{"train_loss": 0.14084605872631073, "global_step": 105144, "epoch": 1181, "lr": 6.749505789146314e-05}
{"train_loss": 0.11521898210048676, "global_step": 105145, "epoch": 1181, "lr": 6.74945148007914e-05}
{"train_loss": 0.1275947093963623, "global_step": 105146, "epoch": 1181, "lr": 6.749397170776777e-05}
{"train_loss": 0.1327589750289917, "global_step": 105147, "epoch": 1181, "lr": 6.749342861239228e-05}
{"train_loss": 0.08886457234621048, "global_step": 105148, "epoch": 1181, "lr": 6.749288551466502e-05}
{"train_loss": 0.17546673119068146, "global_step": 105149, "epoch": 1181, "lr": 6.749234241458603e-05}
{"train_loss": 0.18873770534992218, "global_step": 105150, "epoch": 1181, "lr": 6.749179931215544e-05}
{"train_loss": 0.06746666133403778, "global_step": 105151, "epoch": 1181, "lr": 6.749125620737326e-05}
{"train_loss": 0.17802006006240845, "global_step": 105152, "epoch": 1181, "lr": 6.749071310023962e-05}
{"train_loss": 0.10447748005390167, "global_step": 105153, "epoch": 1181, "lr": 6.749016999075455e-05}
{"train_loss": 0.07404221594333649, "global_step": 105154, "epoch": 1181, "lr": 6.748962687891815e-05}
{"train_loss": 0.1406412124633789, "global_step": 105155, "epoch": 1181, "lr": 6.748908376473047e-05}
{"train_loss": 0.13017019629478455, "global_step": 105156, "epoch": 1181, "lr": 6.74885406481916e-05}
{"train_loss": 0.1205938309431076, "global_step": 105157, "epoch": 1181, "lr": 6.748799752930162e-05}
{"train_loss": 0.1325390636920929, "global_step": 105158, "epoch": 1181, "lr": 6.748745440806058e-05}
{"train_loss": 0.10561263561248779, "global_step": 105159, "epoch": 1181, "lr": 6.748691128446856e-05}
{"train_loss": 0.20526930689811707, "global_step": 105160, "epoch": 1181, "lr": 6.748636815852565e-05}
{"train_loss": 0.13620525598526, "global_step": 105161, "epoch": 1181, "lr": 6.748582503023191e-05}
{"train_loss": 0.1388900876045227, "global_step": 105162, "epoch": 1181, "lr": 6.748528189958742e-05}
{"train_loss": 0.12181007862091064, "global_step": 105163, "epoch": 1181, "lr": 6.748473876659223e-05}
{"train_loss": 0.12756463885307312, "global_step": 105164, "epoch": 1181, "lr": 6.748419563124644e-05}
{"train_loss": 0.24232980608940125, "global_step": 105165, "epoch": 1181, "lr": 6.748365249355011e-05}
{"train_loss": 0.10302148759365082, "global_step": 105166, "epoch": 1181, "lr": 6.748310935350332e-05}
{"train_loss": 0.15707223117351532, "global_step": 105167, "epoch": 1181, "lr": 6.748256621110613e-05}
{"train_loss": 0.09603393077850342, "global_step": 105168, "epoch": 1181, "lr": 6.748202306635864e-05}
{"train_loss": 0.1760079562664032, "global_step": 105169, "epoch": 1181, "lr": 6.748147991926088e-05}
{"train_loss": 0.1029287800192833, "global_step": 105170, "epoch": 1181, "lr": 6.748093676981298e-05}
{"train_loss": 0.22559380531311035, "global_step": 105171, "epoch": 1181, "lr": 6.748039361801494e-05}
{"train_loss": 0.16211143136024475, "global_step": 105172, "epoch": 1181, "lr": 6.747985046386691e-05}
{"train_loss": 0.16662439703941345, "global_step": 105173, "epoch": 1181, "lr": 6.74793073073689e-05}
{"train_loss": 0.13150109350681305, "global_step": 105174, "epoch": 1181, "lr": 6.747876414852104e-05}
{"train_loss": 0.1573302447795868, "global_step": 105175, "epoch": 1181, "lr": 6.747822098732336e-05}
{"train_loss": 0.07009976357221603, "global_step": 105176, "epoch": 1181, "lr": 6.747767782377594e-05}
{"train_loss": 0.15317443013191223, "global_step": 105177, "epoch": 1181, "lr": 6.747713465787886e-05}
{"train_loss": 0.07767263054847717, "global_step": 105178, "epoch": 1181, "lr": 6.74765914896322e-05}
{"train_loss": 0.16746675968170166, "global_step": 105179, "epoch": 1181, "lr": 6.747604831903602e-05}
{"train_loss": 0.11677467823028564, "global_step": 105180, "epoch": 1181, "lr": 6.74755051460904e-05}
{"train_loss": 0.17442789673805237, "global_step": 105181, "epoch": 1181, "lr": 6.747496197079542e-05}
{"train_loss": 0.1275748312473297, "global_step": 105182, "epoch": 1181, "lr": 6.747441879315113e-05}
{"train_loss": 0.19621312618255615, "global_step": 105183, "epoch": 1181, "lr": 6.74738756131576e-05}
{"train_loss": 0.06906773149967194, "global_step": 105184, "epoch": 1181, "lr": 6.747333243081495e-05}
{"train_loss": 0.16874997317790985, "global_step": 105185, "epoch": 1181, "lr": 6.747278924612323e-05}
{"train_loss": 0.07070917636156082, "global_step": 105186, "epoch": 1181, "lr": 6.747224605908249e-05}
{"train_loss": 0.14529238641262054, "global_step": 105187, "epoch": 1181, "lr": 6.747170286969282e-05}
{"train_loss": 0.15876677632331848, "global_step": 105188, "epoch": 1181, "lr": 6.747115967795428e-05}
{"train_loss": 0.1557827740907669, "global_step": 105189, "epoch": 1181, "lr": 6.747061648386698e-05}
{"train_loss": 0.15873025357723236, "global_step": 105190, "epoch": 1181, "lr": 6.747007328743096e-05}
{"train_loss": 0.11548089981079102, "global_step": 105191, "epoch": 1181, "lr": 6.746953008864629e-05}
{"train_loss": 0.12316128611564636, "global_step": 105192, "epoch": 1181, "lr": 6.746898688751307e-05}
{"train_loss": 0.13678433001041412, "global_step": 105193, "epoch": 1181, "lr": 6.746844368403135e-05}
{"train_loss": 0.205550417304039, "global_step": 105194, "epoch": 1181, "lr": 6.746790047820122e-05}
{"train_loss": 0.17555364966392517, "global_step": 105195, "epoch": 1181, "lr": 6.746735727002274e-05}
{"train_loss": 0.15136171877384186, "global_step": 105196, "epoch": 1181, "lr": 6.746681405949598e-05}
{"train_loss": 0.14509713398606588, "global_step": 105197, "epoch": 1181, "lr": 6.746627084662102e-05, "val_loss": 5.018792152404785}
{"train_loss": 0.13921931385993958, "global_step": 105198, "epoch": 1182, "lr": 6.746572763139793e-05}
{"train_loss": 0.09838517755270004, "global_step": 105199, "epoch": 1182, "lr": 6.74651844138268e-05}
{"train_loss": 0.22810098528862, "global_step": 105200, "epoch": 1182, "lr": 6.746464119390768e-05}
{"train_loss": 0.1670011579990387, "global_step": 105201, "epoch": 1182, "lr": 6.746409797164065e-05}
{"train_loss": 0.07466582208871841, "global_step": 105202, "epoch": 1182, "lr": 6.74635547470258e-05}
{"train_loss": 0.2151944637298584, "global_step": 105203, "epoch": 1182, "lr": 6.746301152006316e-05}
{"train_loss": 0.11138326674699783, "global_step": 105204, "epoch": 1182, "lr": 6.746246829075284e-05}
{"train_loss": 0.11172455549240112, "global_step": 105205, "epoch": 1182, "lr": 6.746192505909493e-05}
{"train_loss": 0.13993020355701447, "global_step": 105206, "epoch": 1182, "lr": 6.746138182508945e-05}
{"train_loss": 0.17619960010051727, "global_step": 105207, "epoch": 1182, "lr": 6.746083858873649e-05}
{"train_loss": 0.1454041302204132, "global_step": 105208, "epoch": 1182, "lr": 6.746029535003616e-05}
{"train_loss": 0.12948131561279297, "global_step": 105209, "epoch": 1182, "lr": 6.74597521089885e-05}
{"train_loss": 0.16590847074985504, "global_step": 105210, "epoch": 1182, "lr": 6.745920886559358e-05}
{"train_loss": 0.19175665080547333, "global_step": 105211, "epoch": 1182, "lr": 6.745866561985149e-05}
{"train_loss": 0.12885168194770813, "global_step": 105212, "epoch": 1182, "lr": 6.745812237176229e-05}
{"train_loss": 0.1534295380115509, "global_step": 105213, "epoch": 1182, "lr": 6.745757912132606e-05}
{"train_loss": 0.20875763893127441, "global_step": 105214, "epoch": 1182, "lr": 6.745703586854288e-05}
{"train_loss": 0.19372783601284027, "global_step": 105215, "epoch": 1182, "lr": 6.74564926134128e-05}
{"train_loss": 0.20337337255477905, "global_step": 105216, "epoch": 1182, "lr": 6.745594935593592e-05}
{"train_loss": 0.1897905319929123, "global_step": 105217, "epoch": 1182, "lr": 6.745540609611229e-05}
{"train_loss": 0.12204700708389282, "global_step": 105218, "epoch": 1182, "lr": 6.7454862833942e-05}
{"train_loss": 0.10199488699436188, "global_step": 105219, "epoch": 1182, "lr": 6.74543195694251e-05}
{"train_loss": 0.11815195530653, "global_step": 105220, "epoch": 1182, "lr": 6.74537763025617e-05}
{"train_loss": 0.14466430246829987, "global_step": 105221, "epoch": 1182, "lr": 6.745323303335184e-05}
{"train_loss": 0.20466260612010956, "global_step": 105222, "epoch": 1182, "lr": 6.745268976179562e-05}
{"train_loss": 0.2350151538848877, "global_step": 105223, "epoch": 1182, "lr": 6.745214648789308e-05}
{"train_loss": 0.26732149720191956, "global_step": 105224, "epoch": 1182, "lr": 6.745160321164433e-05}
{"train_loss": 0.1433713138103485, "global_step": 105225, "epoch": 1182, "lr": 6.745105993304942e-05}
{"train_loss": 0.19579990208148956, "global_step": 105226, "epoch": 1182, "lr": 6.745051665210842e-05}
{"train_loss": 0.1253126859664917, "global_step": 105227, "epoch": 1182, "lr": 6.744997336882141e-05}
{"train_loss": 0.09981414675712585, "global_step": 105228, "epoch": 1182, "lr": 6.744943008318847e-05}
{"train_loss": 0.1564127802848816, "global_step": 105229, "epoch": 1182, "lr": 6.744888679520966e-05}
{"train_loss": 0.09663204103708267, "global_step": 105230, "epoch": 1182, "lr": 6.744834350488507e-05}
{"train_loss": 0.1448773294687271, "global_step": 105231, "epoch": 1182, "lr": 6.744780021221475e-05}
{"train_loss": 0.18903683125972748, "global_step": 105232, "epoch": 1182, "lr": 6.744725691719881e-05}
{"train_loss": 0.2517188489437103, "global_step": 105233, "epoch": 1182, "lr": 6.744671361983727e-05}
{"train_loss": 0.1271357238292694, "global_step": 105234, "epoch": 1182, "lr": 6.744617032013025e-05}
{"train_loss": 0.2568090856075287, "global_step": 105235, "epoch": 1182, "lr": 6.74456270180778e-05}
{"train_loss": 0.08738734573125839, "global_step": 105236, "epoch": 1182, "lr": 6.744508371368e-05}
{"train_loss": 0.1251644790172577, "global_step": 105237, "epoch": 1182, "lr": 6.744454040693692e-05}
{"train_loss": 0.191582590341568, "global_step": 105238, "epoch": 1182, "lr": 6.744399709784863e-05}
{"train_loss": 0.11401987820863724, "global_step": 105239, "epoch": 1182, "lr": 6.744345378641521e-05}
{"train_loss": 0.24517913162708282, "global_step": 105240, "epoch": 1182, "lr": 6.744291047263673e-05}
{"train_loss": 0.14342668652534485, "global_step": 105241, "epoch": 1182, "lr": 6.744236715651326e-05}
{"train_loss": 0.2150535136461258, "global_step": 105242, "epoch": 1182, "lr": 6.74418238380449e-05}
{"train_loss": 0.1659533679485321, "global_step": 105243, "epoch": 1182, "lr": 6.744128051723167e-05}
{"train_loss": 0.15751276910305023, "global_step": 105244, "epoch": 1182, "lr": 6.744073719407368e-05}
{"train_loss": 0.15309026837348938, "global_step": 105245, "epoch": 1182, "lr": 6.7440193868571e-05}
{"train_loss": 0.16908882558345795, "global_step": 105246, "epoch": 1182, "lr": 6.74396505407237e-05}
{"train_loss": 0.15421068668365479, "global_step": 105247, "epoch": 1182, "lr": 6.743910721053184e-05}
{"train_loss": 0.09030217677354813, "global_step": 105248, "epoch": 1182, "lr": 6.743856387799553e-05}
{"train_loss": 0.1537923663854599, "global_step": 105249, "epoch": 1182, "lr": 6.743802054311478e-05}
{"train_loss": 0.15193040668964386, "global_step": 105250, "epoch": 1182, "lr": 6.743747720588971e-05}
{"train_loss": 0.1326722651720047, "global_step": 105251, "epoch": 1182, "lr": 6.74369338663204e-05}
{"train_loss": 0.1856520026922226, "global_step": 105252, "epoch": 1182, "lr": 6.743639052440692e-05}
{"train_loss": 0.16058258712291718, "global_step": 105253, "epoch": 1182, "lr": 6.743584718014931e-05}
{"train_loss": 0.197720468044281, "global_step": 105254, "epoch": 1182, "lr": 6.743530383354766e-05}
{"train_loss": 0.148837149143219, "global_step": 105255, "epoch": 1182, "lr": 6.743476048460204e-05}
{"train_loss": 0.1500379741191864, "global_step": 105256, "epoch": 1182, "lr": 6.743421713331254e-05}
{"train_loss": 0.17585735023021698, "global_step": 105257, "epoch": 1182, "lr": 6.743367377967922e-05}
{"train_loss": 0.17068859934806824, "global_step": 105258, "epoch": 1182, "lr": 6.743313042370217e-05}
{"train_loss": 0.18406829237937927, "global_step": 105259, "epoch": 1182, "lr": 6.743258706538143e-05}
{"train_loss": 0.16220852732658386, "global_step": 105260, "epoch": 1182, "lr": 6.743204370471708e-05}
{"train_loss": 0.18424095213413239, "global_step": 105261, "epoch": 1182, "lr": 6.743150034170922e-05}
{"train_loss": 0.16465263068675995, "global_step": 105262, "epoch": 1182, "lr": 6.743095697635791e-05}
{"train_loss": 0.1436053216457367, "global_step": 105263, "epoch": 1182, "lr": 6.743041360866323e-05}
{"train_loss": 0.11680199205875397, "global_step": 105264, "epoch": 1182, "lr": 6.742987023862521e-05}
{"train_loss": 0.22012652456760406, "global_step": 105265, "epoch": 1182, "lr": 6.742932686624399e-05}
{"train_loss": 0.13827764987945557, "global_step": 105266, "epoch": 1182, "lr": 6.742878349151959e-05}
{"train_loss": 0.09973664581775665, "global_step": 105267, "epoch": 1182, "lr": 6.74282401144521e-05}
{"train_loss": 0.147556871175766, "global_step": 105268, "epoch": 1182, "lr": 6.74276967350416e-05}
{"train_loss": 0.1952153891324997, "global_step": 105269, "epoch": 1182, "lr": 6.742715335328816e-05}
{"train_loss": 0.1410120278596878, "global_step": 105270, "epoch": 1182, "lr": 6.742660996919186e-05}
{"train_loss": 0.1885116547346115, "global_step": 105271, "epoch": 1182, "lr": 6.742606658275277e-05}
{"train_loss": 0.12557540833950043, "global_step": 105272, "epoch": 1182, "lr": 6.742552319397093e-05}
{"train_loss": 0.12029365450143814, "global_step": 105273, "epoch": 1182, "lr": 6.742497980284647e-05}
{"train_loss": 0.16622449457645416, "global_step": 105274, "epoch": 1182, "lr": 6.742443640937942e-05}
{"train_loss": 0.18123067915439606, "global_step": 105275, "epoch": 1182, "lr": 6.742389301356988e-05}
{"train_loss": 0.1716664731502533, "global_step": 105276, "epoch": 1182, "lr": 6.742334961541788e-05}
{"train_loss": 0.1633450984954834, "global_step": 105277, "epoch": 1182, "lr": 6.742280621492355e-05}
{"train_loss": 0.15049980580806732, "global_step": 105278, "epoch": 1182, "lr": 6.742226281208692e-05}
{"train_loss": 0.1379982978105545, "global_step": 105279, "epoch": 1182, "lr": 6.742171940690809e-05}
{"train_loss": 0.14964602887630463, "global_step": 105280, "epoch": 1182, "lr": 6.742117599938712e-05}
{"train_loss": 0.2623479664325714, "global_step": 105281, "epoch": 1182, "lr": 6.742063258952409e-05}
{"train_loss": 0.125003382563591, "global_step": 105282, "epoch": 1182, "lr": 6.742008917731905e-05}
{"train_loss": 0.16657604277133942, "global_step": 105283, "epoch": 1182, "lr": 6.74195457627721e-05}
{"train_loss": 0.12826766073703766, "global_step": 105284, "epoch": 1182, "lr": 6.741900234588332e-05}
{"train_loss": 0.15770083665847778, "global_step": 105285, "epoch": 1182, "lr": 6.741845892665274e-05}
{"train_loss": 0.16041163572769487, "global_step": 105286, "epoch": 1182, "lr": 6.741791550508047e-05, "val_loss": 4.868538856506348}
{"train_loss": 0.1223069578409195, "global_step": 105287, "epoch": 1183, "lr": 6.741737208116659e-05}
{"train_loss": 0.1326901763677597, "global_step": 105288, "epoch": 1183, "lr": 6.741682865491113e-05}
{"train_loss": 0.18502803146839142, "global_step": 105289, "epoch": 1183, "lr": 6.741628522631421e-05}
{"train_loss": 0.16139759123325348, "global_step": 105290, "epoch": 1183, "lr": 6.741574179537589e-05}
{"train_loss": 0.25535693764686584, "global_step": 105291, "epoch": 1183, "lr": 6.741519836209622e-05}
{"train_loss": 0.1930134892463684, "global_step": 105292, "epoch": 1183, "lr": 6.741465492647529e-05}
{"train_loss": 0.11840192973613739, "global_step": 105293, "epoch": 1183, "lr": 6.741411148851318e-05}
{"train_loss": 0.1748863309621811, "global_step": 105294, "epoch": 1183, "lr": 6.741356804820995e-05}
{"train_loss": 0.16869190335273743, "global_step": 105295, "epoch": 1183, "lr": 6.741302460556567e-05}
{"train_loss": 0.17831332981586456, "global_step": 105296, "epoch": 1183, "lr": 6.741248116058044e-05}
{"train_loss": 0.11685461550951004, "global_step": 105297, "epoch": 1183, "lr": 6.74119377132543e-05}
{"train_loss": 0.1604926586151123, "global_step": 105298, "epoch": 1183, "lr": 6.741139426358734e-05}
{"train_loss": 0.1840890347957611, "global_step": 105299, "epoch": 1183, "lr": 6.741085081157963e-05}
{"train_loss": 0.13082261383533478, "global_step": 105300, "epoch": 1183, "lr": 6.741030735723124e-05}
{"train_loss": 0.2034672349691391, "global_step": 105301, "epoch": 1183, "lr": 6.740976390054226e-05}
{"train_loss": 0.10893348604440689, "global_step": 105302, "epoch": 1183, "lr": 6.740922044151273e-05}
{"train_loss": 0.13797952234745026, "global_step": 105303, "epoch": 1183, "lr": 6.740867698014275e-05}
{"train_loss": 0.07354986667633057, "global_step": 105304, "epoch": 1183, "lr": 6.740813351643238e-05}
{"train_loss": 0.122184157371521, "global_step": 105305, "epoch": 1183, "lr": 6.740759005038171e-05}
{"train_loss": 0.08240236341953278, "global_step": 105306, "epoch": 1183, "lr": 6.74070465819908e-05}
{"train_loss": 0.20062899589538574, "global_step": 105307, "epoch": 1183, "lr": 6.740650311125971e-05}
{"train_loss": 0.10195475071668625, "global_step": 105308, "epoch": 1183, "lr": 6.740595963818854e-05}
{"train_loss": 0.12542621791362762, "global_step": 105309, "epoch": 1183, "lr": 6.740541616277736e-05}
{"train_loss": 0.11967023462057114, "global_step": 105310, "epoch": 1183, "lr": 6.740487268502623e-05}
{"train_loss": 0.18583466112613678, "global_step": 105311, "epoch": 1183, "lr": 6.740432920493522e-05}
{"train_loss": 0.1804594099521637, "global_step": 105312, "epoch": 1183, "lr": 6.740378572250443e-05}
{"train_loss": 0.08616351336240768, "global_step": 105313, "epoch": 1183, "lr": 6.740324223773389e-05}
{"train_loss": 0.0708903968334198, "global_step": 105314, "epoch": 1183, "lr": 6.740269875062369e-05}
{"train_loss": 0.05895223468542099, "global_step": 105315, "epoch": 1183, "lr": 6.740215526117393e-05}
{"train_loss": 0.27736911177635193, "global_step": 105316, "epoch": 1183, "lr": 6.740161176938464e-05}
{"train_loss": 0.22864986956119537, "global_step": 105317, "epoch": 1183, "lr": 6.740106827525595e-05}
{"train_loss": 0.12531381845474243, "global_step": 105318, "epoch": 1183, "lr": 6.740052477878786e-05}
{"train_loss": 0.0815291628241539, "global_step": 105319, "epoch": 1183, "lr": 6.739998127998051e-05}
{"train_loss": 0.14559878408908844, "global_step": 105320, "epoch": 1183, "lr": 6.739943777883393e-05}
{"train_loss": 0.1647137701511383, "global_step": 105321, "epoch": 1183, "lr": 6.73988942753482e-05}
{"train_loss": 0.10108474642038345, "global_step": 105322, "epoch": 1183, "lr": 6.73983507695234e-05}
{"train_loss": 0.1522350162267685, "global_step": 105323, "epoch": 1183, "lr": 6.739780726135963e-05}
{"train_loss": 0.12688571214675903, "global_step": 105324, "epoch": 1183, "lr": 6.739726375085691e-05}
{"train_loss": 0.061011672019958496, "global_step": 105325, "epoch": 1183, "lr": 6.739672023801535e-05}
{"train_loss": 0.09970685839653015, "global_step": 105326, "epoch": 1183, "lr": 6.739617672283503e-05}
{"train_loss": 0.07314150780439377, "global_step": 105327, "epoch": 1183, "lr": 6.739563320531598e-05}
{"train_loss": 0.15946605801582336, "global_step": 105328, "epoch": 1183, "lr": 6.739508968545831e-05}
{"train_loss": 0.16077136993408203, "global_step": 105329, "epoch": 1183, "lr": 6.739454616326209e-05}
{"train_loss": 0.16875165700912476, "global_step": 105330, "epoch": 1183, "lr": 6.739400263872738e-05}
{"train_loss": 0.15931551158428192, "global_step": 105331, "epoch": 1183, "lr": 6.739345911185426e-05}
{"train_loss": 0.15604108572006226, "global_step": 105332, "epoch": 1183, "lr": 6.739291558264279e-05}
{"train_loss": 0.2516178786754608, "global_step": 105333, "epoch": 1183, "lr": 6.739237205109307e-05}
{"train_loss": 0.18737590312957764, "global_step": 105334, "epoch": 1183, "lr": 6.739182851720515e-05}
{"train_loss": 0.16816569864749908, "global_step": 105335, "epoch": 1183, "lr": 6.73912849809791e-05}
{"train_loss": 0.15710927546024323, "global_step": 105336, "epoch": 1183, "lr": 6.739074144241502e-05}
{"train_loss": 0.09611384570598602, "global_step": 105337, "epoch": 1183, "lr": 6.739019790151296e-05}
{"train_loss": 0.1558026224374771, "global_step": 105338, "epoch": 1183, "lr": 6.7389654358273e-05}
{"train_loss": 0.18728679418563843, "global_step": 105339, "epoch": 1183, "lr": 6.738911081269522e-05}
{"train_loss": 0.1013273373246193, "global_step": 105340, "epoch": 1183, "lr": 6.738856726477966e-05}
{"train_loss": 0.14452378451824188, "global_step": 105341, "epoch": 1183, "lr": 6.738802371452644e-05}
{"train_loss": 0.1279601901769638, "global_step": 105342, "epoch": 1183, "lr": 6.738748016193561e-05}
{"train_loss": 0.11250653117895126, "global_step": 105343, "epoch": 1183, "lr": 6.738693660700725e-05}
{"train_loss": 0.13768728077411652, "global_step": 105344, "epoch": 1183, "lr": 6.738639304974142e-05}
{"train_loss": 0.1495685875415802, "global_step": 105345, "epoch": 1183, "lr": 6.73858494901382e-05}
{"train_loss": 0.1743941754102707, "global_step": 105346, "epoch": 1183, "lr": 6.738530592819767e-05}
{"train_loss": 0.17526200413703918, "global_step": 105347, "epoch": 1183, "lr": 6.738476236391991e-05}
{"train_loss": 0.18976354598999023, "global_step": 105348, "epoch": 1183, "lr": 6.738421879730497e-05}
{"train_loss": 0.13808707892894745, "global_step": 105349, "epoch": 1183, "lr": 6.738367522835294e-05}
{"train_loss": 0.16303105652332306, "global_step": 105350, "epoch": 1183, "lr": 6.738313165706387e-05}
{"train_loss": 0.09844889491796494, "global_step": 105351, "epoch": 1183, "lr": 6.738258808343786e-05}
{"train_loss": 0.12162113934755325, "global_step": 105352, "epoch": 1183, "lr": 6.738204450747498e-05}
{"train_loss": 0.20622453093528748, "global_step": 105353, "epoch": 1183, "lr": 6.738150092917528e-05}
{"train_loss": 0.15226927399635315, "global_step": 105354, "epoch": 1183, "lr": 6.738095734853886e-05}
{"train_loss": 0.13155685365200043, "global_step": 105355, "epoch": 1183, "lr": 6.738041376556578e-05}
{"train_loss": 0.19251097738742828, "global_step": 105356, "epoch": 1183, "lr": 6.73798701802561e-05}
{"train_loss": 0.09852346032857895, "global_step": 105357, "epoch": 1183, "lr": 6.737932659260992e-05}
{"train_loss": 0.25404417514801025, "global_step": 105358, "epoch": 1183, "lr": 6.73787830026273e-05}
{"train_loss": 0.13713239133358002, "global_step": 105359, "epoch": 1183, "lr": 6.737823941030831e-05}
{"train_loss": 0.16091124713420868, "global_step": 105360, "epoch": 1183, "lr": 6.737769581565304e-05}
{"train_loss": 0.16168160736560822, "global_step": 105361, "epoch": 1183, "lr": 6.737715221866155e-05}
{"train_loss": 0.21946008503437042, "global_step": 105362, "epoch": 1183, "lr": 6.73766086193339e-05}
{"train_loss": 0.20033791661262512, "global_step": 105363, "epoch": 1183, "lr": 6.73760650176702e-05}
{"train_loss": 0.17206020653247833, "global_step": 105364, "epoch": 1183, "lr": 6.737552141367047e-05}
{"train_loss": 0.11337075382471085, "global_step": 105365, "epoch": 1183, "lr": 6.737497780733484e-05}
{"train_loss": 0.21047507226467133, "global_step": 105366, "epoch": 1183, "lr": 6.737443419866333e-05}
{"train_loss": 0.27656790614128113, "global_step": 105367, "epoch": 1183, "lr": 6.737389058765606e-05}
{"train_loss": 0.2955562472343445, "global_step": 105368, "epoch": 1183, "lr": 6.737334697431308e-05}
{"train_loss": 0.20990416407585144, "global_step": 105369, "epoch": 1183, "lr": 6.737280335863448e-05}
{"train_loss": 0.1780497282743454, "global_step": 105370, "epoch": 1183, "lr": 6.737225974062029e-05}
{"train_loss": 0.10020487010478973, "global_step": 105371, "epoch": 1183, "lr": 6.737171612027062e-05}
{"train_loss": 0.26859864592552185, "global_step": 105372, "epoch": 1183, "lr": 6.737117249758554e-05}
{"train_loss": 0.1976780742406845, "global_step": 105373, "epoch": 1183, "lr": 6.737062887256511e-05}
{"train_loss": 0.16039340198040009, "global_step": 105374, "epoch": 1183, "lr": 6.737008524520941e-05}
{"train_loss": 0.15621874849782902, "global_step": 105375, "epoch": 1183, "lr": 6.736954161551853e-05, "val_loss": 4.8849263191223145}
{"train_loss": 0.13303151726722717, "global_step": 105376, "epoch": 1184, "lr": 6.73689979834925e-05}
{"train_loss": 0.1806877851486206, "global_step": 105377, "epoch": 1184, "lr": 6.736845434913143e-05}
{"train_loss": 0.18977312743663788, "global_step": 105378, "epoch": 1184, "lr": 6.736791071243539e-05}
{"train_loss": 0.14339454472064972, "global_step": 105379, "epoch": 1184, "lr": 6.736736707340444e-05}
{"train_loss": 0.21841780841350555, "global_step": 105380, "epoch": 1184, "lr": 6.736682343203866e-05}
{"train_loss": 0.12882716953754425, "global_step": 105381, "epoch": 1184, "lr": 6.736627978833813e-05}
{"train_loss": 0.15268021821975708, "global_step": 105382, "epoch": 1184, "lr": 6.736573614230292e-05}
{"train_loss": 0.1361699402332306, "global_step": 105383, "epoch": 1184, "lr": 6.736519249393308e-05}
{"train_loss": 0.1775250881910324, "global_step": 105384, "epoch": 1184, "lr": 6.736464884322871e-05}
{"train_loss": 0.161201611161232, "global_step": 105385, "epoch": 1184, "lr": 6.736410519018988e-05}
{"train_loss": 0.1989172250032425, "global_step": 105386, "epoch": 1184, "lr": 6.736356153481665e-05}
{"train_loss": 0.11597858369350433, "global_step": 105387, "epoch": 1184, "lr": 6.736301787710911e-05}
{"train_loss": 0.09834424406290054, "global_step": 105388, "epoch": 1184, "lr": 6.736247421706732e-05}
{"train_loss": 0.1699589490890503, "global_step": 105389, "epoch": 1184, "lr": 6.736193055469135e-05}
{"train_loss": 0.15494316816329956, "global_step": 105390, "epoch": 1184, "lr": 6.736138688998128e-05}
{"train_loss": 0.17469701170921326, "global_step": 105391, "epoch": 1184, "lr": 6.73608432229372e-05}
{"train_loss": 0.24826054275035858, "global_step": 105392, "epoch": 1184, "lr": 6.736029955355914e-05}
{"train_loss": 0.14351296424865723, "global_step": 105393, "epoch": 1184, "lr": 6.735975588184722e-05}
{"train_loss": 0.11240076273679733, "global_step": 105394, "epoch": 1184, "lr": 6.735921220780148e-05}
{"train_loss": 0.10748844593763351, "global_step": 105395, "epoch": 1184, "lr": 6.735866853142201e-05}
{"train_loss": 0.16306936740875244, "global_step": 105396, "epoch": 1184, "lr": 6.735812485270887e-05}
{"train_loss": 0.15856032073497772, "global_step": 105397, "epoch": 1184, "lr": 6.735758117166216e-05}
{"train_loss": 0.14312686026096344, "global_step": 105398, "epoch": 1184, "lr": 6.735703748828193e-05}
{"train_loss": 0.0970330685377121, "global_step": 105399, "epoch": 1184, "lr": 6.735649380256825e-05}
{"train_loss": 0.1578529179096222, "global_step": 105400, "epoch": 1184, "lr": 6.73559501145212e-05}
{"train_loss": 0.09092552214860916, "global_step": 105401, "epoch": 1184, "lr": 6.735540642414086e-05}
{"train_loss": 0.19473405182361603, "global_step": 105402, "epoch": 1184, "lr": 6.735486273142729e-05}
{"train_loss": 0.14171695709228516, "global_step": 105403, "epoch": 1184, "lr": 6.735431903638059e-05}
{"train_loss": 0.25092580914497375, "global_step": 105404, "epoch": 1184, "lr": 6.73537753390008e-05}
{"train_loss": 0.19261759519577026, "global_step": 105405, "epoch": 1184, "lr": 6.7353231639288e-05}
{"train_loss": 0.14142441749572754, "global_step": 105406, "epoch": 1184, "lr": 6.735268793724228e-05}
{"train_loss": 0.18774181604385376, "global_step": 105407, "epoch": 1184, "lr": 6.73521442328637e-05}
{"train_loss": 0.18480421602725983, "global_step": 105408, "epoch": 1184, "lr": 6.735160052615234e-05}
{"train_loss": 0.09961505234241486, "global_step": 105409, "epoch": 1184, "lr": 6.735105681710826e-05}
{"train_loss": 0.13400103151798248, "global_step": 105410, "epoch": 1184, "lr": 6.735051310573154e-05}
{"train_loss": 0.10483987629413605, "global_step": 105411, "epoch": 1184, "lr": 6.734996939202226e-05}
{"train_loss": 0.1402103453874588, "global_step": 105412, "epoch": 1184, "lr": 6.734942567598048e-05}
{"train_loss": 0.2012207955121994, "global_step": 105413, "epoch": 1184, "lr": 6.734888195760628e-05}
{"train_loss": 0.09633397310972214, "global_step": 105414, "epoch": 1184, "lr": 6.734833823689974e-05}
{"train_loss": 0.19557739794254303, "global_step": 105415, "epoch": 1184, "lr": 6.734779451386094e-05}
{"train_loss": 0.19763725996017456, "global_step": 105416, "epoch": 1184, "lr": 6.734725078848993e-05}
{"train_loss": 0.14714811742305756, "global_step": 105417, "epoch": 1184, "lr": 6.73467070607868e-05}
{"train_loss": 0.14967793226242065, "global_step": 105418, "epoch": 1184, "lr": 6.73461633307516e-05}
{"train_loss": 0.19670559465885162, "global_step": 105419, "epoch": 1184, "lr": 6.734561959838444e-05}
{"train_loss": 0.12640692293643951, "global_step": 105420, "epoch": 1184, "lr": 6.734507586368536e-05}
{"train_loss": 0.19178220629692078, "global_step": 105421, "epoch": 1184, "lr": 6.734453212665445e-05}
{"train_loss": 0.13169047236442566, "global_step": 105422, "epoch": 1184, "lr": 6.734398838729177e-05}
{"train_loss": 0.17311501502990723, "global_step": 105423, "epoch": 1184, "lr": 6.734344464559741e-05}
{"train_loss": 0.15146422386169434, "global_step": 105424, "epoch": 1184, "lr": 6.734290090157144e-05}
{"train_loss": 0.10304249078035355, "global_step": 105425, "epoch": 1184, "lr": 6.734235715521393e-05}
{"train_loss": 0.16959722340106964, "global_step": 105426, "epoch": 1184, "lr": 6.734181340652495e-05}
{"train_loss": 0.2073829025030136, "global_step": 105427, "epoch": 1184, "lr": 6.734126965550457e-05}
{"train_loss": 0.12330776453018188, "global_step": 105428, "epoch": 1184, "lr": 6.734072590215287e-05}
{"train_loss": 0.18117085099220276, "global_step": 105429, "epoch": 1184, "lr": 6.734018214646992e-05}
{"train_loss": 0.14727704226970673, "global_step": 105430, "epoch": 1184, "lr": 6.733963838845579e-05}
{"train_loss": 0.11578208208084106, "global_step": 105431, "epoch": 1184, "lr": 6.733909462811055e-05}
{"train_loss": 0.14550898969173431, "global_step": 105432, "epoch": 1184, "lr": 6.73385508654343e-05}
{"train_loss": 0.18524129688739777, "global_step": 105433, "epoch": 1184, "lr": 6.733800710042707e-05}
{"train_loss": 0.13480950891971588, "global_step": 105434, "epoch": 1184, "lr": 6.733746333308897e-05}
{"train_loss": 0.17553363740444183, "global_step": 105435, "epoch": 1184, "lr": 6.733691956342006e-05}
{"train_loss": 0.09531212598085403, "global_step": 105436, "epoch": 1184, "lr": 6.733637579142041e-05}
{"train_loss": 0.15984971821308136, "global_step": 105437, "epoch": 1184, "lr": 6.733583201709009e-05}
{"train_loss": 0.1326923370361328, "global_step": 105438, "epoch": 1184, "lr": 6.73352882404292e-05}
{"train_loss": 0.20249487459659576, "global_step": 105439, "epoch": 1184, "lr": 6.733474446143777e-05}
{"train_loss": 0.1242411658167839, "global_step": 105440, "epoch": 1184, "lr": 6.73342006801159e-05}
{"train_loss": 0.18423010408878326, "global_step": 105441, "epoch": 1184, "lr": 6.733365689646366e-05}
{"train_loss": 0.16768145561218262, "global_step": 105442, "epoch": 1184, "lr": 6.733311311048112e-05}
{"train_loss": 0.2544049024581909, "global_step": 105443, "epoch": 1184, "lr": 6.733256932216836e-05}
{"train_loss": 0.15225280821323395, "global_step": 105444, "epoch": 1184, "lr": 6.733202553152545e-05}
{"train_loss": 0.1474134624004364, "global_step": 105445, "epoch": 1184, "lr": 6.733148173855245e-05}
{"train_loss": 0.13677231967449188, "global_step": 105446, "epoch": 1184, "lr": 6.733093794324946e-05}
{"train_loss": 0.14161129295825958, "global_step": 105447, "epoch": 1184, "lr": 6.73303941456165e-05}
{"train_loss": 0.21163803339004517, "global_step": 105448, "epoch": 1184, "lr": 6.73298503456537e-05}
{"train_loss": 0.1021781787276268, "global_step": 105449, "epoch": 1184, "lr": 6.732930654336112e-05}
{"train_loss": 0.15390148758888245, "global_step": 105450, "epoch": 1184, "lr": 6.732876273873882e-05}
{"train_loss": 0.09737471491098404, "global_step": 105451, "epoch": 1184, "lr": 6.732821893178687e-05}
{"train_loss": 0.15954041481018066, "global_step": 105452, "epoch": 1184, "lr": 6.732767512250537e-05}
{"train_loss": 0.18295709788799286, "global_step": 105453, "epoch": 1184, "lr": 6.732713131089437e-05}
{"train_loss": 0.23595671355724335, "global_step": 105454, "epoch": 1184, "lr": 6.732658749695396e-05}
{"train_loss": 0.12691432237625122, "global_step": 105455, "epoch": 1184, "lr": 6.732604368068418e-05}
{"train_loss": 0.1974906027317047, "global_step": 105456, "epoch": 1184, "lr": 6.732549986208512e-05}
{"train_loss": 0.12606088817119598, "global_step": 105457, "epoch": 1184, "lr": 6.732495604115687e-05}
{"train_loss": 0.2004418671131134, "global_step": 105458, "epoch": 1184, "lr": 6.732441221789948e-05}
{"train_loss": 0.17132221162319183, "global_step": 105459, "epoch": 1184, "lr": 6.732386839231306e-05}
{"train_loss": 0.10398755222558975, "global_step": 105460, "epoch": 1184, "lr": 6.732332456439765e-05}
{"train_loss": 0.18418990075588226, "global_step": 105461, "epoch": 1184, "lr": 6.732278073415332e-05}
{"train_loss": 0.17442156374454498, "global_step": 105462, "epoch": 1184, "lr": 6.732223690158014e-05}
{"train_loss": 0.1351490467786789, "global_step": 105463, "epoch": 1184, "lr": 6.732169306667823e-05}
{"train_loss": 0.15677243002345054, "global_step": 105464, "epoch": 1184, "lr": 6.732114922944761e-05, "val_loss": 4.903499126434326}
{"train_loss": 0.19699352979660034, "global_step": 105465, "epoch": 1185, "lr": 6.732060538988838e-05}
{"train_loss": 0.11140590906143188, "global_step": 105466, "epoch": 1185, "lr": 6.73200615480006e-05}
{"train_loss": 0.19370882213115692, "global_step": 105467, "epoch": 1185, "lr": 6.731951770378433e-05}
{"train_loss": 0.12256428599357605, "global_step": 105468, "epoch": 1185, "lr": 6.731897385723969e-05}
{"train_loss": 0.11905500292778015, "global_step": 105469, "epoch": 1185, "lr": 6.73184300083667e-05}
{"train_loss": 0.2536422312259674, "global_step": 105470, "epoch": 1185, "lr": 6.731788615716549e-05}
{"train_loss": 0.12283765524625778, "global_step": 105471, "epoch": 1185, "lr": 6.731734230363607e-05}
{"train_loss": 0.1550232619047165, "global_step": 105472, "epoch": 1185, "lr": 6.731679844777857e-05}
{"train_loss": 0.21684472262859344, "global_step": 105473, "epoch": 1185, "lr": 6.731625458959304e-05}
{"train_loss": 0.2283416986465454, "global_step": 105474, "epoch": 1185, "lr": 6.731571072907951e-05}
{"train_loss": 0.16252821683883667, "global_step": 105475, "epoch": 1185, "lr": 6.731516686623813e-05}
{"train_loss": 0.24712733924388885, "global_step": 105476, "epoch": 1185, "lr": 6.731462300106893e-05}
{"train_loss": 0.19296321272850037, "global_step": 105477, "epoch": 1185, "lr": 6.7314079133572e-05}
{"train_loss": 0.14034388959407806, "global_step": 105478, "epoch": 1185, "lr": 6.731353526374739e-05}
{"train_loss": 0.19599948823451996, "global_step": 105479, "epoch": 1185, "lr": 6.731299139159518e-05}
{"train_loss": 0.1518593728542328, "global_step": 105480, "epoch": 1185, "lr": 6.731244751711546e-05}
{"train_loss": 0.10931393504142761, "global_step": 105481, "epoch": 1185, "lr": 6.731190364030828e-05}
{"train_loss": 0.13512054085731506, "global_step": 105482, "epoch": 1185, "lr": 6.731135976117374e-05}
{"train_loss": 0.07762081176042557, "global_step": 105483, "epoch": 1185, "lr": 6.731081587971189e-05}
{"train_loss": 0.19964852929115295, "global_step": 105484, "epoch": 1185, "lr": 6.731027199592282e-05}
{"train_loss": 0.13850416243076324, "global_step": 105485, "epoch": 1185, "lr": 6.730972810980658e-05}
{"train_loss": 0.22887131571769714, "global_step": 105486, "epoch": 1185, "lr": 6.730918422136328e-05}
{"train_loss": 0.24696610867977142, "global_step": 105487, "epoch": 1185, "lr": 6.730864033059295e-05}
{"train_loss": 0.14611084759235382, "global_step": 105488, "epoch": 1185, "lr": 6.73080964374957e-05}
{"train_loss": 0.15786343812942505, "global_step": 105489, "epoch": 1185, "lr": 6.730755254207157e-05}
{"train_loss": 0.15063829720020294, "global_step": 105490, "epoch": 1185, "lr": 6.730700864432066e-05}
{"train_loss": 0.19850732386112213, "global_step": 105491, "epoch": 1185, "lr": 6.730646474424304e-05}
{"train_loss": 0.1497637927532196, "global_step": 105492, "epoch": 1185, "lr": 6.730592084183878e-05}
{"train_loss": 0.21409592032432556, "global_step": 105493, "epoch": 1185, "lr": 6.730537693710793e-05}
{"train_loss": 0.15622767806053162, "global_step": 105494, "epoch": 1185, "lr": 6.73048330300506e-05}
{"train_loss": 0.14795349538326263, "global_step": 105495, "epoch": 1185, "lr": 6.730428912066684e-05}
{"train_loss": 0.20329542458057404, "global_step": 105496, "epoch": 1185, "lr": 6.730374520895672e-05}
{"train_loss": 0.17361906170845032, "global_step": 105497, "epoch": 1185, "lr": 6.730320129492034e-05}
{"train_loss": 0.17313329875469208, "global_step": 105498, "epoch": 1185, "lr": 6.730265737855776e-05}
{"train_loss": 0.16958677768707275, "global_step": 105499, "epoch": 1185, "lr": 6.730211345986904e-05}
{"train_loss": 0.12062717229127884, "global_step": 105500, "epoch": 1185, "lr": 6.730156953885425e-05}
{"train_loss": 0.1644270271062851, "global_step": 105501, "epoch": 1185, "lr": 6.730102561551348e-05}
{"train_loss": 0.1819581836462021, "global_step": 105502, "epoch": 1185, "lr": 6.730048168984679e-05}
{"train_loss": 0.16426394879817963, "global_step": 105503, "epoch": 1185, "lr": 6.729993776185429e-05}
{"train_loss": 0.12531515955924988, "global_step": 105504, "epoch": 1185, "lr": 6.7299393831536e-05}
{"train_loss": 0.13749456405639648, "global_step": 105505, "epoch": 1185, "lr": 6.729884989889202e-05}
{"train_loss": 0.2492040991783142, "global_step": 105506, "epoch": 1185, "lr": 6.729830596392243e-05}
{"train_loss": 0.12052927166223526, "global_step": 105507, "epoch": 1185, "lr": 6.729776202662728e-05}
{"train_loss": 0.12089788913726807, "global_step": 105508, "epoch": 1185, "lr": 6.729721808700667e-05}
{"train_loss": 0.12011098861694336, "global_step": 105509, "epoch": 1185, "lr": 6.729667414506064e-05}
{"train_loss": 0.15463131666183472, "global_step": 105510, "epoch": 1185, "lr": 6.72961302007893e-05}
{"train_loss": 0.146200031042099, "global_step": 105511, "epoch": 1185, "lr": 6.72955862541927e-05}
{"train_loss": 0.1098259836435318, "global_step": 105512, "epoch": 1185, "lr": 6.729504230527093e-05}
{"train_loss": 0.21726547181606293, "global_step": 105513, "epoch": 1185, "lr": 6.729449835402405e-05}
{"train_loss": 0.17377345263957977, "global_step": 105514, "epoch": 1185, "lr": 6.729395440045214e-05}
{"train_loss": 0.1683739572763443, "global_step": 105515, "epoch": 1185, "lr": 6.729341044455525e-05}
{"train_loss": 0.13753409683704376, "global_step": 105516, "epoch": 1185, "lr": 6.729286648633348e-05}
{"train_loss": 0.15146629512310028, "global_step": 105517, "epoch": 1185, "lr": 6.72923225257869e-05}
{"train_loss": 0.169624924659729, "global_step": 105518, "epoch": 1185, "lr": 6.729177856291557e-05}
{"train_loss": 0.20767268538475037, "global_step": 105519, "epoch": 1185, "lr": 6.729123459771958e-05}
{"train_loss": 0.13096313178539276, "global_step": 105520, "epoch": 1185, "lr": 6.729069063019899e-05}
{"train_loss": 0.19736185669898987, "global_step": 105521, "epoch": 1185, "lr": 6.729014666035387e-05}
{"train_loss": 0.1433548927307129, "global_step": 105522, "epoch": 1185, "lr": 6.728960268818431e-05}
{"train_loss": 0.2332794964313507, "global_step": 105523, "epoch": 1185, "lr": 6.728905871369036e-05}
{"train_loss": 0.1746165156364441, "global_step": 105524, "epoch": 1185, "lr": 6.728851473687213e-05}
{"train_loss": 0.16317136585712433, "global_step": 105525, "epoch": 1185, "lr": 6.728797075772965e-05}
{"train_loss": 0.17308464646339417, "global_step": 105526, "epoch": 1185, "lr": 6.728742677626302e-05}
{"train_loss": 0.143464595079422, "global_step": 105527, "epoch": 1185, "lr": 6.728688279247231e-05}
{"train_loss": 0.22294017672538757, "global_step": 105528, "epoch": 1185, "lr": 6.728633880635758e-05}
{"train_loss": 0.28891947865486145, "global_step": 105529, "epoch": 1185, "lr": 6.728579481791892e-05}
{"train_loss": 0.06678738445043564, "global_step": 105530, "epoch": 1185, "lr": 6.72852508271564e-05}
{"train_loss": 0.16321656107902527, "global_step": 105531, "epoch": 1185, "lr": 6.728470683407008e-05}
{"train_loss": 0.062137309461832047, "global_step": 105532, "epoch": 1185, "lr": 6.728416283866004e-05}
{"train_loss": 0.2092069387435913, "global_step": 105533, "epoch": 1185, "lr": 6.728361884092635e-05}
{"train_loss": 0.12104055285453796, "global_step": 105534, "epoch": 1185, "lr": 6.728307484086911e-05}
{"train_loss": 0.09947903454303741, "global_step": 105535, "epoch": 1185, "lr": 6.728253083848836e-05}
{"train_loss": 0.14256013929843903, "global_step": 105536, "epoch": 1185, "lr": 6.728198683378417e-05}
{"train_loss": 0.1694260537624359, "global_step": 105537, "epoch": 1185, "lr": 6.728144282675665e-05}
{"train_loss": 0.14663651585578918, "global_step": 105538, "epoch": 1185, "lr": 6.728089881740584e-05}
{"train_loss": 0.119522325694561, "global_step": 105539, "epoch": 1185, "lr": 6.728035480573183e-05}
{"train_loss": 0.08766855299472809, "global_step": 105540, "epoch": 1185, "lr": 6.727981079173467e-05}
{"train_loss": 0.08152469992637634, "global_step": 105541, "epoch": 1185, "lr": 6.727926677541447e-05}
{"train_loss": 0.14887075126171112, "global_step": 105542, "epoch": 1185, "lr": 6.727872275677128e-05}
{"train_loss": 0.14468169212341309, "global_step": 105543, "epoch": 1185, "lr": 6.727817873580517e-05}
{"train_loss": 0.21246649324893951, "global_step": 105544, "epoch": 1185, "lr": 6.727763471251621e-05}
{"train_loss": 0.15200354158878326, "global_step": 105545, "epoch": 1185, "lr": 6.727709068690449e-05}
{"train_loss": 0.15216560661792755, "global_step": 105546, "epoch": 1185, "lr": 6.727654665897008e-05}
{"train_loss": 0.16312101483345032, "global_step": 105547, "epoch": 1185, "lr": 6.727600262871305e-05}
{"train_loss": 0.16350114345550537, "global_step": 105548, "epoch": 1185, "lr": 6.727545859613348e-05}
{"train_loss": 0.16127970814704895, "global_step": 105549, "epoch": 1185, "lr": 6.727491456123143e-05}
{"train_loss": 0.13599398732185364, "global_step": 105550, "epoch": 1185, "lr": 6.727437052400696e-05}
{"train_loss": 0.20631200075149536, "global_step": 105551, "epoch": 1185, "lr": 6.727382648446018e-05}
{"train_loss": 0.16458317637443542, "global_step": 105552, "epoch": 1185, "lr": 6.727328244259114e-05}
{"train_loss": 0.1628258674081122, "global_step": 105553, "epoch": 1185, "lr": 6.727273839839991e-05, "val_loss": 4.879124164581299, "train_action_mse_error": 26.937164306640625}
{"train_loss": 0.18780875205993652, "global_step": 105554, "epoch": 1186, "lr": 6.727219435188658e-05}
{"train_loss": 0.1586570292711258, "global_step": 105555, "epoch": 1186, "lr": 6.727165030305121e-05}
{"train_loss": 0.12280600517988205, "global_step": 105556, "epoch": 1186, "lr": 6.72711062518939e-05}
{"train_loss": 0.16720463335514069, "global_step": 105557, "epoch": 1186, "lr": 6.727056219841466e-05}
{"train_loss": 0.14503394067287445, "global_step": 105558, "epoch": 1186, "lr": 6.727001814261362e-05}
{"train_loss": 0.13812454044818878, "global_step": 105559, "epoch": 1186, "lr": 6.726947408449084e-05}
{"train_loss": 0.10227184742689133, "global_step": 105560, "epoch": 1186, "lr": 6.726893002404639e-05}
{"train_loss": 0.14455732703208923, "global_step": 105561, "epoch": 1186, "lr": 6.726838596128034e-05}
{"train_loss": 0.1495581418275833, "global_step": 105562, "epoch": 1186, "lr": 6.726784189619278e-05}
{"train_loss": 0.1647193729877472, "global_step": 105563, "epoch": 1186, "lr": 6.726729782878376e-05}
{"train_loss": 0.15009906888008118, "global_step": 105564, "epoch": 1186, "lr": 6.726675375905336e-05}
{"train_loss": 0.14465808868408203, "global_step": 105565, "epoch": 1186, "lr": 6.726620968700165e-05}
{"train_loss": 0.17763733863830566, "global_step": 105566, "epoch": 1186, "lr": 6.726566561262871e-05}
{"train_loss": 0.20034974813461304, "global_step": 105567, "epoch": 1186, "lr": 6.726512153593461e-05}
{"train_loss": 0.26303187012672424, "global_step": 105568, "epoch": 1186, "lr": 6.726457745691942e-05}
{"train_loss": 0.1244325041770935, "global_step": 105569, "epoch": 1186, "lr": 6.726403337558325e-05}
{"train_loss": 0.18051746487617493, "global_step": 105570, "epoch": 1186, "lr": 6.726348929192612e-05}
{"train_loss": 0.14126750826835632, "global_step": 105571, "epoch": 1186, "lr": 6.726294520594812e-05}
{"train_loss": 0.12965354323387146, "global_step": 105572, "epoch": 1186, "lr": 6.726240111764932e-05}
{"train_loss": 0.13204486668109894, "global_step": 105573, "epoch": 1186, "lr": 6.726185702702982e-05}
{"train_loss": 0.14629706740379333, "global_step": 105574, "epoch": 1186, "lr": 6.726131293408966e-05}
{"train_loss": 0.1711898297071457, "global_step": 105575, "epoch": 1186, "lr": 6.726076883882894e-05}
{"train_loss": 0.258849561214447, "global_step": 105576, "epoch": 1186, "lr": 6.72602247412477e-05}
{"train_loss": 0.15986815094947815, "global_step": 105577, "epoch": 1186, "lr": 6.725968064134605e-05}
{"train_loss": 0.11470525711774826, "global_step": 105578, "epoch": 1186, "lr": 6.725913653912403e-05}
{"train_loss": 0.13921289145946503, "global_step": 105579, "epoch": 1186, "lr": 6.725859243458175e-05}
{"train_loss": 0.16244570910930634, "global_step": 105580, "epoch": 1186, "lr": 6.725804832771925e-05}
{"train_loss": 0.16632933914661407, "global_step": 105581, "epoch": 1186, "lr": 6.725750421853663e-05}
{"train_loss": 0.210352823138237, "global_step": 105582, "epoch": 1186, "lr": 6.725696010703392e-05}
{"train_loss": 0.20491643249988556, "global_step": 105583, "epoch": 1186, "lr": 6.725641599321125e-05}
{"train_loss": 0.223056823015213, "global_step": 105584, "epoch": 1186, "lr": 6.725587187706865e-05}
{"train_loss": 0.25303107500076294, "global_step": 105585, "epoch": 1186, "lr": 6.725532775860622e-05}
{"train_loss": 0.12824979424476624, "global_step": 105586, "epoch": 1186, "lr": 6.725478363782401e-05}
{"train_loss": 0.15692958235740662, "global_step": 105587, "epoch": 1186, "lr": 6.725423951472212e-05}
{"train_loss": 0.14771991968154907, "global_step": 105588, "epoch": 1186, "lr": 6.72536953893006e-05}
{"train_loss": 0.21030360460281372, "global_step": 105589, "epoch": 1186, "lr": 6.725315126155953e-05}
{"train_loss": 0.15493875741958618, "global_step": 105590, "epoch": 1186, "lr": 6.725260713149896e-05}
{"train_loss": 0.20580938458442688, "global_step": 105591, "epoch": 1186, "lr": 6.725206299911902e-05}
{"train_loss": 0.08870391547679901, "global_step": 105592, "epoch": 1186, "lr": 6.725151886441974e-05}
{"train_loss": 0.12013162672519684, "global_step": 105593, "epoch": 1186, "lr": 6.72509747274012e-05}
{"train_loss": 0.17192243039608002, "global_step": 105594, "epoch": 1186, "lr": 6.725043058806348e-05}
{"train_loss": 0.15338343381881714, "global_step": 105595, "epoch": 1186, "lr": 6.724988644640667e-05}
{"train_loss": 0.10655274242162704, "global_step": 105596, "epoch": 1186, "lr": 6.724934230243079e-05}
{"train_loss": 0.12781260907649994, "global_step": 105597, "epoch": 1186, "lr": 6.724879815613596e-05}
{"train_loss": 0.14271752536296844, "global_step": 105598, "epoch": 1186, "lr": 6.724825400752224e-05}
{"train_loss": 0.13019342720508575, "global_step": 105599, "epoch": 1186, "lr": 6.72477098565897e-05}
{"train_loss": 0.11392571032047272, "global_step": 105600, "epoch": 1186, "lr": 6.724716570333842e-05}
{"train_loss": 0.08262753486633301, "global_step": 105601, "epoch": 1186, "lr": 6.724662154776847e-05}
{"train_loss": 0.13373106718063354, "global_step": 105602, "epoch": 1186, "lr": 6.724607738987992e-05}
{"train_loss": 0.12025951594114304, "global_step": 105603, "epoch": 1186, "lr": 6.724553322967283e-05}
{"train_loss": 0.1365339308977127, "global_step": 105604, "epoch": 1186, "lr": 6.72449890671473e-05}
{"train_loss": 0.16161686182022095, "global_step": 105605, "epoch": 1186, "lr": 6.724444490230339e-05}
{"train_loss": 0.17239387333393097, "global_step": 105606, "epoch": 1186, "lr": 6.724390073514118e-05}
{"train_loss": 0.12763360142707825, "global_step": 105607, "epoch": 1186, "lr": 6.724335656566074e-05}
{"train_loss": 0.19316120445728302, "global_step": 105608, "epoch": 1186, "lr": 6.724281239386213e-05}
{"train_loss": 0.12014573067426682, "global_step": 105609, "epoch": 1186, "lr": 6.724226821974544e-05}
{"train_loss": 0.10834646970033646, "global_step": 105610, "epoch": 1186, "lr": 6.724172404331074e-05}
{"train_loss": 0.1612475961446762, "global_step": 105611, "epoch": 1186, "lr": 6.724117986455809e-05}
{"train_loss": 0.1555211991071701, "global_step": 105612, "epoch": 1186, "lr": 6.724063568348758e-05}
{"train_loss": 0.19560515880584717, "global_step": 105613, "epoch": 1186, "lr": 6.724009150009927e-05}
{"train_loss": 0.11841025948524475, "global_step": 105614, "epoch": 1186, "lr": 6.723954731439325e-05}
{"train_loss": 0.1149367168545723, "global_step": 105615, "epoch": 1186, "lr": 6.723900312636959e-05}
{"train_loss": 0.11396512389183044, "global_step": 105616, "epoch": 1186, "lr": 6.723845893602835e-05}
{"train_loss": 0.22418197989463806, "global_step": 105617, "epoch": 1186, "lr": 6.72379147433696e-05}
{"train_loss": 0.17441347241401672, "global_step": 105618, "epoch": 1186, "lr": 6.723737054839344e-05}
{"train_loss": 0.10412219911813736, "global_step": 105619, "epoch": 1186, "lr": 6.72368263510999e-05}
{"train_loss": 0.20519492030143738, "global_step": 105620, "epoch": 1186, "lr": 6.72362821514891e-05}
{"train_loss": 0.1490764170885086, "global_step": 105621, "epoch": 1186, "lr": 6.723573794956108e-05}
{"train_loss": 0.09323912113904953, "global_step": 105622, "epoch": 1186, "lr": 6.723519374531592e-05}
{"train_loss": 0.13101528584957123, "global_step": 105623, "epoch": 1186, "lr": 6.723464953875372e-05}
{"train_loss": 0.14015057682991028, "global_step": 105624, "epoch": 1186, "lr": 6.723410532987452e-05}
{"train_loss": 0.15333305299282074, "global_step": 105625, "epoch": 1186, "lr": 6.72335611186784e-05}
{"train_loss": 0.11809822916984558, "global_step": 105626, "epoch": 1186, "lr": 6.723301690516544e-05}
{"train_loss": 0.10500548779964447, "global_step": 105627, "epoch": 1186, "lr": 6.723247268933571e-05}
{"train_loss": 0.1249968633055687, "global_step": 105628, "epoch": 1186, "lr": 6.723192847118929e-05}
{"train_loss": 0.12138494849205017, "global_step": 105629, "epoch": 1186, "lr": 6.723138425072624e-05}
{"train_loss": 0.20283737778663635, "global_step": 105630, "epoch": 1186, "lr": 6.723084002794664e-05}
{"train_loss": 0.14055407047271729, "global_step": 105631, "epoch": 1186, "lr": 6.723029580285056e-05}
{"train_loss": 0.1719394028186798, "global_step": 105632, "epoch": 1186, "lr": 6.722975157543809e-05}
{"train_loss": 0.1980723738670349, "global_step": 105633, "epoch": 1186, "lr": 6.72292073457093e-05}
{"train_loss": 0.15250521898269653, "global_step": 105634, "epoch": 1186, "lr": 6.722866311366424e-05}
{"train_loss": 0.18377402424812317, "global_step": 105635, "epoch": 1186, "lr": 6.722811887930298e-05}
{"train_loss": 0.11717169731855392, "global_step": 105636, "epoch": 1186, "lr": 6.722757464262564e-05}
{"train_loss": 0.09241876751184464, "global_step": 105637, "epoch": 1186, "lr": 6.722703040363225e-05}
{"train_loss": 0.15006665885448456, "global_step": 105638, "epoch": 1186, "lr": 6.722648616232288e-05}
{"train_loss": 0.11590734124183655, "global_step": 105639, "epoch": 1186, "lr": 6.722594191869761e-05}
{"train_loss": 0.10255841910839081, "global_step": 105640, "epoch": 1186, "lr": 6.722539767275656e-05}
{"train_loss": 0.09771330654621124, "global_step": 105641, "epoch": 1186, "lr": 6.722485342449975e-05}
{"train_loss": 0.15061999497453818, "global_step": 105642, "epoch": 1186, "lr": 6.722430917392726e-05, "val_loss": 4.9495391845703125}
{"train_loss": 0.1276719570159912, "global_step": 105643, "epoch": 1187, "lr": 6.722376492103918e-05}
{"train_loss": 0.1571745127439499, "global_step": 105644, "epoch": 1187, "lr": 6.722322066583557e-05}
{"train_loss": 0.11393708735704422, "global_step": 105645, "epoch": 1187, "lr": 6.722267640831653e-05}
{"train_loss": 0.1264055073261261, "global_step": 105646, "epoch": 1187, "lr": 6.722213214848207e-05}
{"train_loss": 0.14446137845516205, "global_step": 105647, "epoch": 1187, "lr": 6.722158788633233e-05}
{"train_loss": 0.15470612049102783, "global_step": 105648, "epoch": 1187, "lr": 6.722104362186735e-05}
{"train_loss": 0.14587584137916565, "global_step": 105649, "epoch": 1187, "lr": 6.722049935508721e-05}
{"train_loss": 0.21002620458602905, "global_step": 105650, "epoch": 1187, "lr": 6.721995508599198e-05}
{"train_loss": 0.10833057761192322, "global_step": 105651, "epoch": 1187, "lr": 6.721941081458176e-05}
{"train_loss": 0.1379496306180954, "global_step": 105652, "epoch": 1187, "lr": 6.721886654085657e-05}
{"train_loss": 0.11408200860023499, "global_step": 105653, "epoch": 1187, "lr": 6.721832226481655e-05}
{"train_loss": 0.1417287290096283, "global_step": 105654, "epoch": 1187, "lr": 6.72177779864617e-05}
{"train_loss": 0.14221087098121643, "global_step": 105655, "epoch": 1187, "lr": 6.721723370579215e-05}
{"train_loss": 0.21330039203166962, "global_step": 105656, "epoch": 1187, "lr": 6.721668942280795e-05}
{"train_loss": 0.11618469655513763, "global_step": 105657, "epoch": 1187, "lr": 6.721614513750916e-05}
{"train_loss": 0.08804637938737869, "global_step": 105658, "epoch": 1187, "lr": 6.721560084989589e-05}
{"train_loss": 0.13786499202251434, "global_step": 105659, "epoch": 1187, "lr": 6.721505655996817e-05}
{"train_loss": 0.14668521285057068, "global_step": 105660, "epoch": 1187, "lr": 6.721451226772613e-05}
{"train_loss": 0.1341685950756073, "global_step": 105661, "epoch": 1187, "lr": 6.721396797316978e-05}
{"train_loss": 0.14255324006080627, "global_step": 105662, "epoch": 1187, "lr": 6.721342367629923e-05}
{"train_loss": 0.12652406096458435, "global_step": 105663, "epoch": 1187, "lr": 6.721287937711454e-05}
{"train_loss": 0.10266552865505219, "global_step": 105664, "epoch": 1187, "lr": 6.721233507561579e-05}
{"train_loss": 0.17555968463420868, "global_step": 105665, "epoch": 1187, "lr": 6.721179077180306e-05}
{"train_loss": 0.22691522538661957, "global_step": 105666, "epoch": 1187, "lr": 6.72112464656764e-05}
{"train_loss": 0.11974365264177322, "global_step": 105667, "epoch": 1187, "lr": 6.721070215723591e-05}
{"train_loss": 0.16305160522460938, "global_step": 105668, "epoch": 1187, "lr": 6.721015784648164e-05}
{"train_loss": 0.1602097898721695, "global_step": 105669, "epoch": 1187, "lr": 6.720961353341369e-05}
{"train_loss": 0.15338453650474548, "global_step": 105670, "epoch": 1187, "lr": 6.720906921803209e-05}
{"train_loss": 0.14762763679027557, "global_step": 105671, "epoch": 1187, "lr": 6.720852490033697e-05}
{"train_loss": 0.14337240159511566, "global_step": 105672, "epoch": 1187, "lr": 6.720798058032838e-05}
{"train_loss": 0.1409711241722107, "global_step": 105673, "epoch": 1187, "lr": 6.720743625800636e-05}
{"train_loss": 0.17503571510314941, "global_step": 105674, "epoch": 1187, "lr": 6.720689193337101e-05}
{"train_loss": 0.21939878165721893, "global_step": 105675, "epoch": 1187, "lr": 6.720634760642241e-05}
{"train_loss": 0.24347442388534546, "global_step": 105676, "epoch": 1187, "lr": 6.720580327716064e-05}
{"train_loss": 0.1383907049894333, "global_step": 105677, "epoch": 1187, "lr": 6.720525894558575e-05}
{"train_loss": 0.12457840889692307, "global_step": 105678, "epoch": 1187, "lr": 6.720471461169782e-05}
{"train_loss": 0.1409282684326172, "global_step": 105679, "epoch": 1187, "lr": 6.720417027549692e-05}
{"train_loss": 0.31975293159484863, "global_step": 105680, "epoch": 1187, "lr": 6.720362593698314e-05}
{"train_loss": 0.16402016580104828, "global_step": 105681, "epoch": 1187, "lr": 6.720308159615654e-05}
{"train_loss": 0.17243829369544983, "global_step": 105682, "epoch": 1187, "lr": 6.720253725301719e-05}
{"train_loss": 0.10112319886684418, "global_step": 105683, "epoch": 1187, "lr": 6.720199290756516e-05}
{"train_loss": 0.14997880160808563, "global_step": 105684, "epoch": 1187, "lr": 6.720144855980055e-05}
{"train_loss": 0.09911142289638519, "global_step": 105685, "epoch": 1187, "lr": 6.72009042097234e-05}
{"train_loss": 0.13359279930591583, "global_step": 105686, "epoch": 1187, "lr": 6.720035985733381e-05}
{"train_loss": 0.1514088660478592, "global_step": 105687, "epoch": 1187, "lr": 6.719981550263183e-05}
{"train_loss": 0.13605433702468872, "global_step": 105688, "epoch": 1187, "lr": 6.719927114561756e-05}
{"train_loss": 0.20809833705425262, "global_step": 105689, "epoch": 1187, "lr": 6.719872678629104e-05}
{"train_loss": 0.1842963844537735, "global_step": 105690, "epoch": 1187, "lr": 6.719818242465237e-05}
{"train_loss": 0.15293553471565247, "global_step": 105691, "epoch": 1187, "lr": 6.719763806070162e-05}
{"train_loss": 0.19706480205059052, "global_step": 105692, "epoch": 1187, "lr": 6.719709369443886e-05}
{"train_loss": 0.1742006540298462, "global_step": 105693, "epoch": 1187, "lr": 6.719654932586416e-05}
{"train_loss": 0.16443680226802826, "global_step": 105694, "epoch": 1187, "lr": 6.719600495497758e-05}
{"train_loss": 0.16684451699256897, "global_step": 105695, "epoch": 1187, "lr": 6.719546058177921e-05}
{"train_loss": 0.07513631880283356, "global_step": 105696, "epoch": 1187, "lr": 6.719491620626912e-05}
{"train_loss": 0.1584290713071823, "global_step": 105697, "epoch": 1187, "lr": 6.71943718284474e-05}
{"train_loss": 0.17397239804267883, "global_step": 105698, "epoch": 1187, "lr": 6.719382744831408e-05}
{"train_loss": 0.1374773532152176, "global_step": 105699, "epoch": 1187, "lr": 6.719328306586927e-05}
{"train_loss": 0.186087504029274, "global_step": 105700, "epoch": 1187, "lr": 6.719273868111304e-05}
{"train_loss": 0.17524750530719757, "global_step": 105701, "epoch": 1187, "lr": 6.719219429404544e-05}
{"train_loss": 0.12050051242113113, "global_step": 105702, "epoch": 1187, "lr": 6.719164990466657e-05}
{"train_loss": 0.16303829848766327, "global_step": 105703, "epoch": 1187, "lr": 6.719110551297648e-05}
{"train_loss": 0.22317419946193695, "global_step": 105704, "epoch": 1187, "lr": 6.719056111897527e-05}
{"train_loss": 0.19422654807567596, "global_step": 105705, "epoch": 1187, "lr": 6.719001672266298e-05}
{"train_loss": 0.17665253579616547, "global_step": 105706, "epoch": 1187, "lr": 6.718947232403973e-05}
{"train_loss": 0.1603224128484726, "global_step": 105707, "epoch": 1187, "lr": 6.718892792310555e-05}
{"train_loss": 0.11901161819696426, "global_step": 105708, "epoch": 1187, "lr": 6.718838351986053e-05}
{"train_loss": 0.11113172769546509, "global_step": 105709, "epoch": 1187, "lr": 6.718783911430474e-05}
{"train_loss": 0.14084377884864807, "global_step": 105710, "epoch": 1187, "lr": 6.718729470643825e-05}
{"train_loss": 0.2570951282978058, "global_step": 105711, "epoch": 1187, "lr": 6.718675029626116e-05}
{"train_loss": 0.15518563985824585, "global_step": 105712, "epoch": 1187, "lr": 6.718620588377351e-05}
{"train_loss": 0.1487983763217926, "global_step": 105713, "epoch": 1187, "lr": 6.718566146897537e-05}
{"train_loss": 0.16469267010688782, "global_step": 105714, "epoch": 1187, "lr": 6.718511705186683e-05}
{"train_loss": 0.14358025789260864, "global_step": 105715, "epoch": 1187, "lr": 6.718457263244797e-05}
{"train_loss": 0.14120446145534515, "global_step": 105716, "epoch": 1187, "lr": 6.718402821071885e-05}
{"train_loss": 0.13139358162879944, "global_step": 105717, "epoch": 1187, "lr": 6.718348378667955e-05}
{"train_loss": 0.21434931457042694, "global_step": 105718, "epoch": 1187, "lr": 6.718293936033013e-05}
{"train_loss": 0.170632466673851, "global_step": 105719, "epoch": 1187, "lr": 6.718239493167068e-05}
{"train_loss": 0.12943382561206818, "global_step": 105720, "epoch": 1187, "lr": 6.718185050070127e-05}
{"train_loss": 0.2684197425842285, "global_step": 105721, "epoch": 1187, "lr": 6.718130606742194e-05}
{"train_loss": 0.16395902633666992, "global_step": 105722, "epoch": 1187, "lr": 6.718076163183283e-05}
{"train_loss": 0.16799065470695496, "global_step": 105723, "epoch": 1187, "lr": 6.718021719393397e-05}
{"train_loss": 0.12130588293075562, "global_step": 105724, "epoch": 1187, "lr": 6.717967275372544e-05}
{"train_loss": 0.1736387014389038, "global_step": 105725, "epoch": 1187, "lr": 6.71791283112073e-05}
{"train_loss": 0.1765630543231964, "global_step": 105726, "epoch": 1187, "lr": 6.717858386637963e-05}
{"train_loss": 0.17871077358722687, "global_step": 105727, "epoch": 1187, "lr": 6.717803941924253e-05}
{"train_loss": 0.16717574000358582, "global_step": 105728, "epoch": 1187, "lr": 6.717749496979604e-05}
{"train_loss": 0.16829806566238403, "global_step": 105729, "epoch": 1187, "lr": 6.717695051804025e-05}
{"train_loss": 0.221055805683136, "global_step": 105730, "epoch": 1187, "lr": 6.717640606397523e-05}
{"train_loss": 0.15848370934470316, "global_step": 105731, "epoch": 1187, "lr": 6.717586160760105e-05, "val_loss": 4.865218162536621}
{"train_loss": 0.14370115101337433, "global_step": 105732, "epoch": 1188, "lr": 6.717531714891778e-05}
{"train_loss": 0.16638679802417755, "global_step": 105733, "epoch": 1188, "lr": 6.71747726879255e-05}
{"train_loss": 0.1577511578798294, "global_step": 105734, "epoch": 1188, "lr": 6.717422822462428e-05}
{"train_loss": 0.11743807792663574, "global_step": 105735, "epoch": 1188, "lr": 6.71736837590142e-05}
{"train_loss": 0.18718641996383667, "global_step": 105736, "epoch": 1188, "lr": 6.717313929109532e-05}
{"train_loss": 0.16697010397911072, "global_step": 105737, "epoch": 1188, "lr": 6.717259482086772e-05}
{"train_loss": 0.1525806188583374, "global_step": 105738, "epoch": 1188, "lr": 6.717205034833149e-05}
{"train_loss": 0.1425679624080658, "global_step": 105739, "epoch": 1188, "lr": 6.717150587348665e-05}
{"train_loss": 0.1413085162639618, "global_step": 105740, "epoch": 1188, "lr": 6.717096139633335e-05}
{"train_loss": 0.21974384784698486, "global_step": 105741, "epoch": 1188, "lr": 6.71704169168716e-05}
{"train_loss": 0.2689914107322693, "global_step": 105742, "epoch": 1188, "lr": 6.71698724351015e-05}
{"train_loss": 0.245428204536438, "global_step": 105743, "epoch": 1188, "lr": 6.716932795102313e-05}
{"train_loss": 0.1495376080274582, "global_step": 105744, "epoch": 1188, "lr": 6.716878346463656e-05}
{"train_loss": 0.1656368523836136, "global_step": 105745, "epoch": 1188, "lr": 6.716823897594184e-05}
{"train_loss": 0.11410189419984818, "global_step": 105746, "epoch": 1188, "lr": 6.716769448493906e-05}
{"train_loss": 0.1643708199262619, "global_step": 105747, "epoch": 1188, "lr": 6.716714999162829e-05}
{"train_loss": 0.18107177317142487, "global_step": 105748, "epoch": 1188, "lr": 6.716660549600961e-05}
{"train_loss": 0.18739591538906097, "global_step": 105749, "epoch": 1188, "lr": 6.716606099808309e-05}
{"train_loss": 0.1311691254377365, "global_step": 105750, "epoch": 1188, "lr": 6.716551649784883e-05}
{"train_loss": 0.15390543639659882, "global_step": 105751, "epoch": 1188, "lr": 6.716497199530684e-05}
{"train_loss": 0.16267205774784088, "global_step": 105752, "epoch": 1188, "lr": 6.716442749045724e-05}
{"train_loss": 0.1567530333995819, "global_step": 105753, "epoch": 1188, "lr": 6.716388298330009e-05}
{"train_loss": 0.15628118813037872, "global_step": 105754, "epoch": 1188, "lr": 6.716333847383547e-05}
{"train_loss": 0.1632322371006012, "global_step": 105755, "epoch": 1188, "lr": 6.716279396206344e-05}
{"train_loss": 0.19853612780570984, "global_step": 105756, "epoch": 1188, "lr": 6.716224944798408e-05}
{"train_loss": 0.1707974225282669, "global_step": 105757, "epoch": 1188, "lr": 6.716170493159747e-05}
{"train_loss": 0.10708360373973846, "global_step": 105758, "epoch": 1188, "lr": 6.716116041290368e-05}
{"train_loss": 0.20017081499099731, "global_step": 105759, "epoch": 1188, "lr": 6.716061589190277e-05}
{"train_loss": 0.15445418655872345, "global_step": 105760, "epoch": 1188, "lr": 6.716007136859483e-05}
{"train_loss": 0.17974895238876343, "global_step": 105761, "epoch": 1188, "lr": 6.715952684297994e-05}
{"train_loss": 0.16781991720199585, "global_step": 105762, "epoch": 1188, "lr": 6.715898231505815e-05}
{"train_loss": 0.10798235237598419, "global_step": 105763, "epoch": 1188, "lr": 6.715843778482954e-05}
{"train_loss": 0.16445323824882507, "global_step": 105764, "epoch": 1188, "lr": 6.71578932522942e-05}
{"train_loss": 0.18029356002807617, "global_step": 105765, "epoch": 1188, "lr": 6.715734871745218e-05}
{"train_loss": 0.11825548857450485, "global_step": 105766, "epoch": 1188, "lr": 6.715680418030357e-05}
{"train_loss": 0.19680428504943848, "global_step": 105767, "epoch": 1188, "lr": 6.715625964084843e-05}
{"train_loss": 0.10256529599428177, "global_step": 105768, "epoch": 1188, "lr": 6.715571509908684e-05}
{"train_loss": 0.1326308250427246, "global_step": 105769, "epoch": 1188, "lr": 6.71551705550189e-05}
{"train_loss": 0.12965130805969238, "global_step": 105770, "epoch": 1188, "lr": 6.715462600864462e-05}
{"train_loss": 0.09918951988220215, "global_step": 105771, "epoch": 1188, "lr": 6.715408145996414e-05}
{"train_loss": 0.1570354700088501, "global_step": 105772, "epoch": 1188, "lr": 6.715353690897748e-05}
{"train_loss": 0.18367967009544373, "global_step": 105773, "epoch": 1188, "lr": 6.715299235568474e-05}
{"train_loss": 0.12062736600637436, "global_step": 105774, "epoch": 1188, "lr": 6.715244780008599e-05}
{"train_loss": 0.1774701029062271, "global_step": 105775, "epoch": 1188, "lr": 6.71519032421813e-05}
{"train_loss": 0.1845240294933319, "global_step": 105776, "epoch": 1188, "lr": 6.715135868197074e-05}
{"train_loss": 0.17791524529457092, "global_step": 105777, "epoch": 1188, "lr": 6.715081411945439e-05}
{"train_loss": 0.15523874759674072, "global_step": 105778, "epoch": 1188, "lr": 6.715026955463234e-05}
{"train_loss": 0.09068311750888824, "global_step": 105779, "epoch": 1188, "lr": 6.714972498750463e-05}
{"train_loss": 0.21243590116500854, "global_step": 105780, "epoch": 1188, "lr": 6.714918041807137e-05}
{"train_loss": 0.12527434527873993, "global_step": 105781, "epoch": 1188, "lr": 6.714863584633259e-05}
{"train_loss": 0.20491334795951843, "global_step": 105782, "epoch": 1188, "lr": 6.714809127228838e-05}
{"train_loss": 0.2255297303199768, "global_step": 105783, "epoch": 1188, "lr": 6.714754669593883e-05}
{"train_loss": 0.1763426959514618, "global_step": 105784, "epoch": 1188, "lr": 6.714700211728401e-05}
{"train_loss": 0.1187138557434082, "global_step": 105785, "epoch": 1188, "lr": 6.714645753632397e-05}
{"train_loss": 0.13317526876926422, "global_step": 105786, "epoch": 1188, "lr": 6.714591295305881e-05}
{"train_loss": 0.2176557183265686, "global_step": 105787, "epoch": 1188, "lr": 6.714536836748857e-05}
{"train_loss": 0.19497928023338318, "global_step": 105788, "epoch": 1188, "lr": 6.714482377961338e-05}
{"train_loss": 0.10532982647418976, "global_step": 105789, "epoch": 1188, "lr": 6.714427918943324e-05}
{"train_loss": 0.1788698434829712, "global_step": 105790, "epoch": 1188, "lr": 6.714373459694828e-05}
{"train_loss": 0.20529603958129883, "global_step": 105791, "epoch": 1188, "lr": 6.714319000215856e-05}
{"train_loss": 0.13175183534622192, "global_step": 105792, "epoch": 1188, "lr": 6.714264540506412e-05}
{"train_loss": 0.12737368047237396, "global_step": 105793, "epoch": 1188, "lr": 6.714210080566507e-05}
{"train_loss": 0.13478784263134003, "global_step": 105794, "epoch": 1188, "lr": 6.714155620396148e-05}
{"train_loss": 0.18328304588794708, "global_step": 105795, "epoch": 1188, "lr": 6.714101159995343e-05}
{"train_loss": 0.17993254959583282, "global_step": 105796, "epoch": 1188, "lr": 6.714046699364097e-05}
{"train_loss": 0.11833415180444717, "global_step": 105797, "epoch": 1188, "lr": 6.713992238502418e-05}
{"train_loss": 0.147724449634552, "global_step": 105798, "epoch": 1188, "lr": 6.713937777410313e-05}
{"train_loss": 0.14068818092346191, "global_step": 105799, "epoch": 1188, "lr": 6.71388331608779e-05}
{"train_loss": 0.14895787835121155, "global_step": 105800, "epoch": 1188, "lr": 6.713828854534858e-05}
{"train_loss": 0.13550978899002075, "global_step": 105801, "epoch": 1188, "lr": 6.71377439275152e-05}
{"train_loss": 0.2171715795993805, "global_step": 105802, "epoch": 1188, "lr": 6.713719930737788e-05}
{"train_loss": 0.1256342977285385, "global_step": 105803, "epoch": 1188, "lr": 6.713665468493666e-05}
{"train_loss": 0.13152728974819183, "global_step": 105804, "epoch": 1188, "lr": 6.713611006019163e-05}
{"train_loss": 0.07352408766746521, "global_step": 105805, "epoch": 1188, "lr": 6.713556543314288e-05}
{"train_loss": 0.1288725584745407, "global_step": 105806, "epoch": 1188, "lr": 6.713502080379044e-05}
{"train_loss": 0.11144359409809113, "global_step": 105807, "epoch": 1188, "lr": 6.71344761721344e-05}
{"train_loss": 0.10310952365398407, "global_step": 105808, "epoch": 1188, "lr": 6.713393153817487e-05}
{"train_loss": 0.13598568737506866, "global_step": 105809, "epoch": 1188, "lr": 6.713338690191186e-05}
{"train_loss": 0.1554967314004898, "global_step": 105810, "epoch": 1188, "lr": 6.713284226334549e-05}
{"train_loss": 0.16658367216587067, "global_step": 105811, "epoch": 1188, "lr": 6.71322976224758e-05}
{"train_loss": 0.1541745662689209, "global_step": 105812, "epoch": 1188, "lr": 6.713175297930288e-05}
{"train_loss": 0.19881583750247955, "global_step": 105813, "epoch": 1188, "lr": 6.713120833382683e-05}
{"train_loss": 0.16513285040855408, "global_step": 105814, "epoch": 1188, "lr": 6.713066368604768e-05}
{"train_loss": 0.17132645845413208, "global_step": 105815, "epoch": 1188, "lr": 6.713011903596552e-05}
{"train_loss": 0.2206917703151703, "global_step": 105816, "epoch": 1188, "lr": 6.712957438358044e-05}
{"train_loss": 0.12488620728254318, "global_step": 105817, "epoch": 1188, "lr": 6.712902972889248e-05}
{"train_loss": 0.20077724754810333, "global_step": 105818, "epoch": 1188, "lr": 6.712848507190175e-05}
{"train_loss": 0.14249145984649658, "global_step": 105819, "epoch": 1188, "lr": 6.712794041260828e-05}
{"train_loss": 0.15751470324028743, "global_step": 105820, "epoch": 1188, "lr": 6.712739575101218e-05, "val_loss": 4.909548282623291}
{"train_loss": 0.10411444306373596, "global_step": 105821, "epoch": 1189, "lr": 6.712685108711351e-05}
{"train_loss": 0.12390526384115219, "global_step": 105822, "epoch": 1189, "lr": 6.712630642091236e-05}
{"train_loss": 0.10523942857980728, "global_step": 105823, "epoch": 1189, "lr": 6.712576175240877e-05}
{"train_loss": 0.19329088926315308, "global_step": 105824, "epoch": 1189, "lr": 6.712521708160282e-05}
{"train_loss": 0.16102251410484314, "global_step": 105825, "epoch": 1189, "lr": 6.712467240849461e-05}
{"train_loss": 0.1513088047504425, "global_step": 105826, "epoch": 1189, "lr": 6.712412773308419e-05}
{"train_loss": 0.15594588220119476, "global_step": 105827, "epoch": 1189, "lr": 6.712358305537166e-05}
{"train_loss": 0.11291084438562393, "global_step": 105828, "epoch": 1189, "lr": 6.712303837535705e-05}
{"train_loss": 0.1947772204875946, "global_step": 105829, "epoch": 1189, "lr": 6.712249369304045e-05}
{"train_loss": 0.17451418936252594, "global_step": 105830, "epoch": 1189, "lr": 6.712194900842195e-05}
{"train_loss": 0.1312105655670166, "global_step": 105831, "epoch": 1189, "lr": 6.71214043215016e-05}
{"train_loss": 0.11304667592048645, "global_step": 105832, "epoch": 1189, "lr": 6.71208596322795e-05}
{"train_loss": 0.1562965214252472, "global_step": 105833, "epoch": 1189, "lr": 6.71203149407557e-05}
{"train_loss": 0.1922137439250946, "global_step": 105834, "epoch": 1189, "lr": 6.71197702469303e-05}
{"train_loss": 0.1455775499343872, "global_step": 105835, "epoch": 1189, "lr": 6.711922555080334e-05}
{"train_loss": 0.16191397607326508, "global_step": 105836, "epoch": 1189, "lr": 6.711868085237492e-05}
{"train_loss": 0.1898934692144394, "global_step": 105837, "epoch": 1189, "lr": 6.71181361516451e-05}
{"train_loss": 0.18942692875862122, "global_step": 105838, "epoch": 1189, "lr": 6.711759144861394e-05}
{"train_loss": 0.15438450872898102, "global_step": 105839, "epoch": 1189, "lr": 6.711704674328154e-05}
{"train_loss": 0.07930110394954681, "global_step": 105840, "epoch": 1189, "lr": 6.711650203564797e-05}
{"train_loss": 0.10865235328674316, "global_step": 105841, "epoch": 1189, "lr": 6.711595732571327e-05}
{"train_loss": 0.15447160601615906, "global_step": 105842, "epoch": 1189, "lr": 6.711541261347756e-05}
{"train_loss": 0.16989673674106598, "global_step": 105843, "epoch": 1189, "lr": 6.711486789894089e-05}
{"train_loss": 0.11692780256271362, "global_step": 105844, "epoch": 1189, "lr": 6.711432318210333e-05}
{"train_loss": 0.09433745592832565, "global_step": 105845, "epoch": 1189, "lr": 6.711377846296495e-05}
{"train_loss": 0.161892831325531, "global_step": 105846, "epoch": 1189, "lr": 6.711323374152586e-05}
{"train_loss": 0.09574980288743973, "global_step": 105847, "epoch": 1189, "lr": 6.711268901778608e-05}
{"train_loss": 0.19436267018318176, "global_step": 105848, "epoch": 1189, "lr": 6.711214429174571e-05}
{"train_loss": 0.14085423946380615, "global_step": 105849, "epoch": 1189, "lr": 6.711159956340481e-05}
{"train_loss": 0.14995862543582916, "global_step": 105850, "epoch": 1189, "lr": 6.711105483276348e-05}
{"train_loss": 0.18891257047653198, "global_step": 105851, "epoch": 1189, "lr": 6.711051009982177e-05}
{"train_loss": 0.181715726852417, "global_step": 105852, "epoch": 1189, "lr": 6.710996536457976e-05}
{"train_loss": 0.17294655740261078, "global_step": 105853, "epoch": 1189, "lr": 6.710942062703754e-05}
{"train_loss": 0.19119738042354584, "global_step": 105854, "epoch": 1189, "lr": 6.710887588719514e-05}
{"train_loss": 0.09825906902551651, "global_step": 105855, "epoch": 1189, "lr": 6.710833114505269e-05}
{"train_loss": 0.20553703606128693, "global_step": 105856, "epoch": 1189, "lr": 6.71077864006102e-05}
{"train_loss": 0.20208822190761566, "global_step": 105857, "epoch": 1189, "lr": 6.71072416538678e-05}
{"train_loss": 0.13980185985565186, "global_step": 105858, "epoch": 1189, "lr": 6.710669690482553e-05}
{"train_loss": 0.10626005381345749, "global_step": 105859, "epoch": 1189, "lr": 6.710615215348349e-05}
{"train_loss": 0.14527198672294617, "global_step": 105860, "epoch": 1189, "lr": 6.71056073998417e-05}
{"train_loss": 0.11414653062820435, "global_step": 105861, "epoch": 1189, "lr": 6.710506264390031e-05}
{"train_loss": 0.11364200711250305, "global_step": 105862, "epoch": 1189, "lr": 6.710451788565934e-05}
{"train_loss": 0.17837807536125183, "global_step": 105863, "epoch": 1189, "lr": 6.710397312511887e-05}
{"train_loss": 0.16611193120479584, "global_step": 105864, "epoch": 1189, "lr": 6.710342836227897e-05}
{"train_loss": 0.16690893471240997, "global_step": 105865, "epoch": 1189, "lr": 6.710288359713975e-05}
{"train_loss": 0.15623103082180023, "global_step": 105866, "epoch": 1189, "lr": 6.710233882970123e-05}
{"train_loss": 0.2558876574039459, "global_step": 105867, "epoch": 1189, "lr": 6.710179405996351e-05}
{"train_loss": 0.16613100469112396, "global_step": 105868, "epoch": 1189, "lr": 6.710124928792667e-05}
{"train_loss": 0.1445927768945694, "global_step": 105869, "epoch": 1189, "lr": 6.710070451359078e-05}
{"train_loss": 0.09629932045936584, "global_step": 105870, "epoch": 1189, "lr": 6.71001597369559e-05}
{"train_loss": 0.15058591961860657, "global_step": 105871, "epoch": 1189, "lr": 6.709961495802213e-05}
{"train_loss": 0.08237182348966599, "global_step": 105872, "epoch": 1189, "lr": 6.70990701767895e-05}
{"train_loss": 0.12909917533397675, "global_step": 105873, "epoch": 1189, "lr": 6.709852539325813e-05}
{"train_loss": 0.14635170996189117, "global_step": 105874, "epoch": 1189, "lr": 6.709798060742806e-05}
{"train_loss": 0.14436408877372742, "global_step": 105875, "epoch": 1189, "lr": 6.709743581929938e-05}
{"train_loss": 0.11466749012470245, "global_step": 105876, "epoch": 1189, "lr": 6.709689102887215e-05}
{"train_loss": 0.10883743315935135, "global_step": 105877, "epoch": 1189, "lr": 6.709634623614645e-05}
{"train_loss": 0.21769002079963684, "global_step": 105878, "epoch": 1189, "lr": 6.709580144112236e-05}
{"train_loss": 0.16919627785682678, "global_step": 105879, "epoch": 1189, "lr": 6.709525664379996e-05}
{"train_loss": 0.15800631046295166, "global_step": 105880, "epoch": 1189, "lr": 6.70947118441793e-05}
{"train_loss": 0.19609594345092773, "global_step": 105881, "epoch": 1189, "lr": 6.709416704226045e-05}
{"train_loss": 0.11633190512657166, "global_step": 105882, "epoch": 1189, "lr": 6.709362223804352e-05}
{"train_loss": 0.11766289174556732, "global_step": 105883, "epoch": 1189, "lr": 6.709307743152854e-05}
{"train_loss": 0.11603837460279465, "global_step": 105884, "epoch": 1189, "lr": 6.709253262271563e-05}
{"train_loss": 0.15603892505168915, "global_step": 105885, "epoch": 1189, "lr": 6.709198781160482e-05}
{"train_loss": 0.14040103554725647, "global_step": 105886, "epoch": 1189, "lr": 6.70914429981962e-05}
{"train_loss": 0.10456545650959015, "global_step": 105887, "epoch": 1189, "lr": 6.709089818248986e-05}
{"train_loss": 0.16845054924488068, "global_step": 105888, "epoch": 1189, "lr": 6.709035336448583e-05}
{"train_loss": 0.16024266183376312, "global_step": 105889, "epoch": 1189, "lr": 6.708980854418424e-05}
{"train_loss": 0.19328436255455017, "global_step": 105890, "epoch": 1189, "lr": 6.708926372158512e-05}
{"train_loss": 0.11997810006141663, "global_step": 105891, "epoch": 1189, "lr": 6.708871889668855e-05}
{"train_loss": 0.05559220910072327, "global_step": 105892, "epoch": 1189, "lr": 6.708817406949461e-05}
{"train_loss": 0.1555410921573639, "global_step": 105893, "epoch": 1189, "lr": 6.70876292400034e-05}
{"train_loss": 0.14577864110469818, "global_step": 105894, "epoch": 1189, "lr": 6.708708440821495e-05}
{"train_loss": 0.20789237320423126, "global_step": 105895, "epoch": 1189, "lr": 6.708653957412934e-05}
{"train_loss": 0.14940163493156433, "global_step": 105896, "epoch": 1189, "lr": 6.708599473774665e-05}
{"train_loss": 0.15811893343925476, "global_step": 105897, "epoch": 1189, "lr": 6.708544989906696e-05}
{"train_loss": 0.12714038789272308, "global_step": 105898, "epoch": 1189, "lr": 6.708490505809036e-05}
{"train_loss": 0.1580638587474823, "global_step": 105899, "epoch": 1189, "lr": 6.708436021481687e-05}
{"train_loss": 0.11987273395061493, "global_step": 105900, "epoch": 1189, "lr": 6.708381536924662e-05}
{"train_loss": 0.15973636507987976, "global_step": 105901, "epoch": 1189, "lr": 6.708327052137967e-05}
{"train_loss": 0.11636991798877716, "global_step": 105902, "epoch": 1189, "lr": 6.708272567121605e-05}
{"train_loss": 0.16716957092285156, "global_step": 105903, "epoch": 1189, "lr": 6.708218081875588e-05}
{"train_loss": 0.11483769863843918, "global_step": 105904, "epoch": 1189, "lr": 6.708163596399921e-05}
{"train_loss": 0.12355196475982666, "global_step": 105905, "epoch": 1189, "lr": 6.708109110694614e-05}
{"train_loss": 0.12573739886283875, "global_step": 105906, "epoch": 1189, "lr": 6.708054624759672e-05}
{"train_loss": 0.14497610926628113, "global_step": 105907, "epoch": 1189, "lr": 6.708000138595103e-05}
{"train_loss": 0.2071867436170578, "global_step": 105908, "epoch": 1189, "lr": 6.707945652200913e-05}
{"train_loss": 0.15036417502030897, "global_step": 105909, "epoch": 1189, "lr": 6.707891165577112e-05, "val_loss": 4.96620512008667}
{"train_loss": 0.09900233149528503, "global_step": 105910, "epoch": 1190, "lr": 6.707836678723704e-05}
{"train_loss": 0.17783281207084656, "global_step": 105911, "epoch": 1190, "lr": 6.7077821916407e-05}
{"train_loss": 0.21824386715888977, "global_step": 105912, "epoch": 1190, "lr": 6.707727704328105e-05}
{"train_loss": 0.17350506782531738, "global_step": 105913, "epoch": 1190, "lr": 6.707673216785927e-05}
{"train_loss": 0.17313840985298157, "global_step": 105914, "epoch": 1190, "lr": 6.707618729014172e-05}
{"train_loss": 0.1649826020002365, "global_step": 105915, "epoch": 1190, "lr": 6.707564241012849e-05}
{"train_loss": 0.1375402957201004, "global_step": 105916, "epoch": 1190, "lr": 6.707509752781966e-05}
{"train_loss": 0.1711951494216919, "global_step": 105917, "epoch": 1190, "lr": 6.707455264321528e-05}
{"train_loss": 0.18911239504814148, "global_step": 105918, "epoch": 1190, "lr": 6.707400775631544e-05}
{"train_loss": 0.17536577582359314, "global_step": 105919, "epoch": 1190, "lr": 6.70734628671202e-05}
{"train_loss": 0.12291666120290756, "global_step": 105920, "epoch": 1190, "lr": 6.707291797562964e-05}
{"train_loss": 0.11239131540060043, "global_step": 105921, "epoch": 1190, "lr": 6.707237308184384e-05}
{"train_loss": 0.1277303397655487, "global_step": 105922, "epoch": 1190, "lr": 6.707182818576287e-05}
{"train_loss": 0.2263222634792328, "global_step": 105923, "epoch": 1190, "lr": 6.707128328738682e-05}
{"train_loss": 0.18685491383075714, "global_step": 105924, "epoch": 1190, "lr": 6.707073838671571e-05}
{"train_loss": 0.16174302995204926, "global_step": 105925, "epoch": 1190, "lr": 6.707019348374966e-05}
{"train_loss": 0.1918061226606369, "global_step": 105926, "epoch": 1190, "lr": 6.706964857848872e-05}
{"train_loss": 0.27876368165016174, "global_step": 105927, "epoch": 1190, "lr": 6.706910367093298e-05}
{"train_loss": 0.13702984154224396, "global_step": 105928, "epoch": 1190, "lr": 6.706855876108252e-05}
{"train_loss": 0.16315554082393646, "global_step": 105929, "epoch": 1190, "lr": 6.70680138489374e-05}
{"train_loss": 0.1290794461965561, "global_step": 105930, "epoch": 1190, "lr": 6.706746893449768e-05}
{"train_loss": 0.14040973782539368, "global_step": 105931, "epoch": 1190, "lr": 6.706692401776344e-05}
{"train_loss": 0.147914856672287, "global_step": 105932, "epoch": 1190, "lr": 6.706637909873478e-05}
{"train_loss": 0.114533931016922, "global_step": 105933, "epoch": 1190, "lr": 6.706583417741175e-05}
{"train_loss": 0.16284531354904175, "global_step": 105934, "epoch": 1190, "lr": 6.706528925379441e-05}
{"train_loss": 0.21467342972755432, "global_step": 105935, "epoch": 1190, "lr": 6.706474432788286e-05}
{"train_loss": 0.09113400429487228, "global_step": 105936, "epoch": 1190, "lr": 6.706419939967717e-05}
{"train_loss": 0.2687075138092041, "global_step": 105937, "epoch": 1190, "lr": 6.706365446917741e-05}
{"train_loss": 0.16525274515151978, "global_step": 105938, "epoch": 1190, "lr": 6.706310953638364e-05}
{"train_loss": 0.16537009179592133, "global_step": 105939, "epoch": 1190, "lr": 6.706256460129594e-05}
{"train_loss": 0.2563786208629608, "global_step": 105940, "epoch": 1190, "lr": 6.70620196639144e-05}
{"train_loss": 0.157527893781662, "global_step": 105941, "epoch": 1190, "lr": 6.706147472423906e-05}
{"train_loss": 0.1602022796869278, "global_step": 105942, "epoch": 1190, "lr": 6.706092978227003e-05}
{"train_loss": 0.20008373260498047, "global_step": 105943, "epoch": 1190, "lr": 6.706038483800734e-05}
{"train_loss": 0.17893385887145996, "global_step": 105944, "epoch": 1190, "lr": 6.705983989145112e-05}
{"train_loss": 0.15801601111888885, "global_step": 105945, "epoch": 1190, "lr": 6.70592949426014e-05}
{"train_loss": 0.11856957525014877, "global_step": 105946, "epoch": 1190, "lr": 6.705874999145825e-05}
{"train_loss": 0.13751232624053955, "global_step": 105947, "epoch": 1190, "lr": 6.705820503802178e-05}
{"train_loss": 0.12967760860919952, "global_step": 105948, "epoch": 1190, "lr": 6.705766008229204e-05}
{"train_loss": 0.1801062971353531, "global_step": 105949, "epoch": 1190, "lr": 6.70571151242691e-05}
{"train_loss": 0.1454622745513916, "global_step": 105950, "epoch": 1190, "lr": 6.705657016395304e-05}
{"train_loss": 0.14708039164543152, "global_step": 105951, "epoch": 1190, "lr": 6.705602520134393e-05}
{"train_loss": 0.1602431982755661, "global_step": 105952, "epoch": 1190, "lr": 6.705548023644184e-05}
{"train_loss": 0.1352146863937378, "global_step": 105953, "epoch": 1190, "lr": 6.705493526924687e-05}
{"train_loss": 0.12805481255054474, "global_step": 105954, "epoch": 1190, "lr": 6.705439029975906e-05}
{"train_loss": 0.18281997740268707, "global_step": 105955, "epoch": 1190, "lr": 6.70538453279785e-05}
{"train_loss": 0.13089406490325928, "global_step": 105956, "epoch": 1190, "lr": 6.705330035390525e-05}
{"train_loss": 0.18479317426681519, "global_step": 105957, "epoch": 1190, "lr": 6.705275537753939e-05}
{"train_loss": 0.1493760645389557, "global_step": 105958, "epoch": 1190, "lr": 6.705221039888098e-05}
{"train_loss": 0.2429833561182022, "global_step": 105959, "epoch": 1190, "lr": 6.705166541793014e-05}
{"train_loss": 0.12344378978013992, "global_step": 105960, "epoch": 1190, "lr": 6.70511204346869e-05}
{"train_loss": 0.20322854816913605, "global_step": 105961, "epoch": 1190, "lr": 6.705057544915135e-05}
{"train_loss": 0.1716725379228592, "global_step": 105962, "epoch": 1190, "lr": 6.705003046132354e-05}
{"train_loss": 0.21059027314186096, "global_step": 105963, "epoch": 1190, "lr": 6.704948547120358e-05}
{"train_loss": 0.18495580554008484, "global_step": 105964, "epoch": 1190, "lr": 6.704894047879151e-05}
{"train_loss": 0.2131895124912262, "global_step": 105965, "epoch": 1190, "lr": 6.704839548408743e-05}
{"train_loss": 0.1711239516735077, "global_step": 105966, "epoch": 1190, "lr": 6.70478504870914e-05}
{"train_loss": 0.1558275818824768, "global_step": 105967, "epoch": 1190, "lr": 6.704730548780347e-05}
{"train_loss": 0.17162564396858215, "global_step": 105968, "epoch": 1190, "lr": 6.704676048622375e-05}
{"train_loss": 0.20423421263694763, "global_step": 105969, "epoch": 1190, "lr": 6.704621548235232e-05}
{"train_loss": 0.14154434204101562, "global_step": 105970, "epoch": 1190, "lr": 6.704567047618921e-05}
{"train_loss": 0.250198096036911, "global_step": 105971, "epoch": 1190, "lr": 6.704512546773454e-05}
{"train_loss": 0.14896009862422943, "global_step": 105972, "epoch": 1190, "lr": 6.704458045698835e-05}
{"train_loss": 0.10879966616630554, "global_step": 105973, "epoch": 1190, "lr": 6.704403544395073e-05}
{"train_loss": 0.13586559891700745, "global_step": 105974, "epoch": 1190, "lr": 6.704349042862173e-05}
{"train_loss": 0.30176040530204773, "global_step": 105975, "epoch": 1190, "lr": 6.704294541100145e-05}
{"train_loss": 0.1495000272989273, "global_step": 105976, "epoch": 1190, "lr": 6.704240039108995e-05}
{"train_loss": 0.1950952410697937, "global_step": 105977, "epoch": 1190, "lr": 6.704185536888732e-05}
{"train_loss": 0.17635232210159302, "global_step": 105978, "epoch": 1190, "lr": 6.704131034439361e-05}
{"train_loss": 0.13115723431110382, "global_step": 105979, "epoch": 1190, "lr": 6.704076531760891e-05}
{"train_loss": 0.1652558296918869, "global_step": 105980, "epoch": 1190, "lr": 6.704022028853329e-05}
{"train_loss": 0.14148013293743134, "global_step": 105981, "epoch": 1190, "lr": 6.70396752571668e-05}
{"train_loss": 0.13392619788646698, "global_step": 105982, "epoch": 1190, "lr": 6.703913022350955e-05}
{"train_loss": 0.12220937013626099, "global_step": 105983, "epoch": 1190, "lr": 6.70385851875616e-05}
{"train_loss": 0.19016392529010773, "global_step": 105984, "epoch": 1190, "lr": 6.7038040149323e-05}
{"train_loss": 0.16830039024353027, "global_step": 105985, "epoch": 1190, "lr": 6.703749510879386e-05}
{"train_loss": 0.17835968732833862, "global_step": 105986, "epoch": 1190, "lr": 6.703695006597423e-05}
{"train_loss": 0.12170238047838211, "global_step": 105987, "epoch": 1190, "lr": 6.703640502086419e-05}
{"train_loss": 0.1880890429019928, "global_step": 105988, "epoch": 1190, "lr": 6.703585997346382e-05}
{"train_loss": 0.11444246023893356, "global_step": 105989, "epoch": 1190, "lr": 6.703531492377317e-05}
{"train_loss": 0.16350243985652924, "global_step": 105990, "epoch": 1190, "lr": 6.703476987179236e-05}
{"train_loss": 0.14749117195606232, "global_step": 105991, "epoch": 1190, "lr": 6.70342248175214e-05}
{"train_loss": 0.10940396785736084, "global_step": 105992, "epoch": 1190, "lr": 6.70336797609604e-05}
{"train_loss": 0.14677639305591583, "global_step": 105993, "epoch": 1190, "lr": 6.703313470210945e-05}
{"train_loss": 0.17585061490535736, "global_step": 105994, "epoch": 1190, "lr": 6.703258964096858e-05}
{"train_loss": 0.09955809265375137, "global_step": 105995, "epoch": 1190, "lr": 6.70320445775379e-05}
{"train_loss": 0.1435810774564743, "global_step": 105996, "epoch": 1190, "lr": 6.703149951181748e-05}
{"train_loss": 0.2418140172958374, "global_step": 105997, "epoch": 1190, "lr": 6.703095444380737e-05}
{"train_loss": 0.16529679850915843, "global_step": 105998, "epoch": 1190, "lr": 6.703040937350765e-05, "val_loss": 4.799795627593994, "train_action_mse_error": 27.13351058959961}
{"train_loss": 0.20683187246322632, "global_step": 105999, "epoch": 1191, "lr": 6.70298643009184e-05}
{"train_loss": 0.1584416925907135, "global_step": 106000, "epoch": 1191, "lr": 6.70293192260397e-05}
{"train_loss": 0.11251392215490341, "global_step": 106001, "epoch": 1191, "lr": 6.70287741488716e-05}
{"train_loss": 0.11777222901582718, "global_step": 106002, "epoch": 1191, "lr": 6.702822906941422e-05}
{"train_loss": 0.13378030061721802, "global_step": 106003, "epoch": 1191, "lr": 6.702768398766758e-05}
{"train_loss": 0.16770276427268982, "global_step": 106004, "epoch": 1191, "lr": 6.702713890363177e-05}
{"train_loss": 0.1337805688381195, "global_step": 106005, "epoch": 1191, "lr": 6.702659381730688e-05}
{"train_loss": 0.16996406018733978, "global_step": 106006, "epoch": 1191, "lr": 6.702604872869297e-05}
{"train_loss": 0.15669257938861847, "global_step": 106007, "epoch": 1191, "lr": 6.702550363779011e-05}
{"train_loss": 0.11245684325695038, "global_step": 106008, "epoch": 1191, "lr": 6.702495854459838e-05}
{"train_loss": 0.16455617547035217, "global_step": 106009, "epoch": 1191, "lr": 6.702441344911787e-05}
{"train_loss": 0.16705676913261414, "global_step": 106010, "epoch": 1191, "lr": 6.702386835134862e-05}
{"train_loss": 0.1601492315530777, "global_step": 106011, "epoch": 1191, "lr": 6.702332325129071e-05}
{"train_loss": 0.08722101897001266, "global_step": 106012, "epoch": 1191, "lr": 6.702277814894422e-05}
{"train_loss": 0.2056809663772583, "global_step": 106013, "epoch": 1191, "lr": 6.702223304430923e-05}
{"train_loss": 0.09478170424699783, "global_step": 106014, "epoch": 1191, "lr": 6.702168793738582e-05}
{"train_loss": 0.12304499000310898, "global_step": 106015, "epoch": 1191, "lr": 6.702114282817405e-05}
{"train_loss": 0.1571093052625656, "global_step": 106016, "epoch": 1191, "lr": 6.7020597716674e-05}
{"train_loss": 0.1138392984867096, "global_step": 106017, "epoch": 1191, "lr": 6.702005260288572e-05}
{"train_loss": 0.2360004186630249, "global_step": 106018, "epoch": 1191, "lr": 6.70195074868093e-05}
{"train_loss": 0.23804685473442078, "global_step": 106019, "epoch": 1191, "lr": 6.701896236844482e-05}
{"train_loss": 0.12197601795196533, "global_step": 106020, "epoch": 1191, "lr": 6.701841724779236e-05}
{"train_loss": 0.1186014786362648, "global_step": 106021, "epoch": 1191, "lr": 6.701787212485198e-05}
{"train_loss": 0.15915580093860626, "global_step": 106022, "epoch": 1191, "lr": 6.701732699962375e-05}
{"train_loss": 0.1262742429971695, "global_step": 106023, "epoch": 1191, "lr": 6.701678187210773e-05}
{"train_loss": 0.15027272701263428, "global_step": 106024, "epoch": 1191, "lr": 6.701623674230403e-05}
{"train_loss": 0.08545868843793869, "global_step": 106025, "epoch": 1191, "lr": 6.701569161021271e-05}
{"train_loss": 0.13777486979961395, "global_step": 106026, "epoch": 1191, "lr": 6.701514647583382e-05}
{"train_loss": 0.20064367353916168, "global_step": 106027, "epoch": 1191, "lr": 6.701460133916746e-05}
{"train_loss": 0.1869879812002182, "global_step": 106028, "epoch": 1191, "lr": 6.701405620021369e-05}
{"train_loss": 0.16039720177650452, "global_step": 106029, "epoch": 1191, "lr": 6.701351105897258e-05}
{"train_loss": 0.16718077659606934, "global_step": 106030, "epoch": 1191, "lr": 6.701296591544421e-05}
{"train_loss": 0.16267919540405273, "global_step": 106031, "epoch": 1191, "lr": 6.701242076962868e-05}
{"train_loss": 0.1297304332256317, "global_step": 106032, "epoch": 1191, "lr": 6.701187562152602e-05}
{"train_loss": 0.17597487568855286, "global_step": 106033, "epoch": 1191, "lr": 6.701133047113632e-05}
{"train_loss": 0.1363077014684677, "global_step": 106034, "epoch": 1191, "lr": 6.701078531845967e-05}
{"train_loss": 0.1380545049905777, "global_step": 106035, "epoch": 1191, "lr": 6.701024016349612e-05}
{"train_loss": 0.09520401060581207, "global_step": 106036, "epoch": 1191, "lr": 6.700969500624575e-05}
{"train_loss": 0.11711163073778152, "global_step": 106037, "epoch": 1191, "lr": 6.700914984670862e-05}
{"train_loss": 0.0582134872674942, "global_step": 106038, "epoch": 1191, "lr": 6.700860468488483e-05}
{"train_loss": 0.1319725066423416, "global_step": 106039, "epoch": 1191, "lr": 6.700805952077444e-05}
{"train_loss": 0.21933187544345856, "global_step": 106040, "epoch": 1191, "lr": 6.700751435437751e-05}
{"train_loss": 0.19503411650657654, "global_step": 106041, "epoch": 1191, "lr": 6.700696918569413e-05}
{"train_loss": 0.10951140522956848, "global_step": 106042, "epoch": 1191, "lr": 6.700642401472439e-05}
{"train_loss": 0.13562817871570587, "global_step": 106043, "epoch": 1191, "lr": 6.700587884146832e-05}
{"train_loss": 0.14050185680389404, "global_step": 106044, "epoch": 1191, "lr": 6.700533366592604e-05}
{"train_loss": 0.08245328813791275, "global_step": 106045, "epoch": 1191, "lr": 6.700478848809759e-05}
{"train_loss": 0.12432211637496948, "global_step": 106046, "epoch": 1191, "lr": 6.700424330798305e-05}
{"train_loss": 0.12112013250589371, "global_step": 106047, "epoch": 1191, "lr": 6.700369812558249e-05}
{"train_loss": 0.24697650969028473, "global_step": 106048, "epoch": 1191, "lr": 6.7003152940896e-05}
{"train_loss": 0.1325419396162033, "global_step": 106049, "epoch": 1191, "lr": 6.700260775392364e-05}
{"train_loss": 0.18459419906139374, "global_step": 106050, "epoch": 1191, "lr": 6.700206256466548e-05}
{"train_loss": 0.14257612824440002, "global_step": 106051, "epoch": 1191, "lr": 6.700151737312162e-05}
{"train_loss": 0.1563257873058319, "global_step": 106052, "epoch": 1191, "lr": 6.700097217929208e-05}
{"train_loss": 0.1737108677625656, "global_step": 106053, "epoch": 1191, "lr": 6.7000426983177e-05}
{"train_loss": 0.10900121927261353, "global_step": 106054, "epoch": 1191, "lr": 6.699988178477642e-05}
{"train_loss": 0.13460002839565277, "global_step": 106055, "epoch": 1191, "lr": 6.69993365840904e-05}
{"train_loss": 0.10257436335086823, "global_step": 106056, "epoch": 1191, "lr": 6.699879138111903e-05}
{"train_loss": 0.1316346824169159, "global_step": 106057, "epoch": 1191, "lr": 6.699824617586237e-05}
{"train_loss": 0.15308859944343567, "global_step": 106058, "epoch": 1191, "lr": 6.699770096832051e-05}
{"train_loss": 0.18842589855194092, "global_step": 106059, "epoch": 1191, "lr": 6.699715575849353e-05}
{"train_loss": 0.14235110580921173, "global_step": 106060, "epoch": 1191, "lr": 6.699661054638147e-05}
{"train_loss": 0.06967967003583908, "global_step": 106061, "epoch": 1191, "lr": 6.699606533198443e-05}
{"train_loss": 0.051889825612306595, "global_step": 106062, "epoch": 1191, "lr": 6.699552011530246e-05}
{"train_loss": 0.1791524589061737, "global_step": 106063, "epoch": 1191, "lr": 6.699497489633567e-05}
{"train_loss": 0.1555081158876419, "global_step": 106064, "epoch": 1191, "lr": 6.699442967508412e-05}
{"train_loss": 0.16585305333137512, "global_step": 106065, "epoch": 1191, "lr": 6.699388445154784e-05}
{"train_loss": 0.12315084040164948, "global_step": 106066, "epoch": 1191, "lr": 6.699333922572696e-05}
{"train_loss": 0.17222872376441956, "global_step": 106067, "epoch": 1191, "lr": 6.699279399762154e-05}
{"train_loss": 0.15234875679016113, "global_step": 106068, "epoch": 1191, "lr": 6.699224876723165e-05}
{"train_loss": 0.10711181163787842, "global_step": 106069, "epoch": 1191, "lr": 6.699170353455734e-05}
{"train_loss": 0.12887611985206604, "global_step": 106070, "epoch": 1191, "lr": 6.699115829959871e-05}
{"train_loss": 0.09525072574615479, "global_step": 106071, "epoch": 1191, "lr": 6.699061306235584e-05}
{"train_loss": 0.09022238105535507, "global_step": 106072, "epoch": 1191, "lr": 6.699006782282877e-05}
{"train_loss": 0.07439716905355453, "global_step": 106073, "epoch": 1191, "lr": 6.698952258101761e-05}
{"train_loss": 0.13531257212162018, "global_step": 106074, "epoch": 1191, "lr": 6.698897733692241e-05}
{"train_loss": 0.12610428035259247, "global_step": 106075, "epoch": 1191, "lr": 6.698843209054324e-05}
{"train_loss": 0.2003318816423416, "global_step": 106076, "epoch": 1191, "lr": 6.698788684188019e-05}
{"train_loss": 0.20390838384628296, "global_step": 106077, "epoch": 1191, "lr": 6.698734159093332e-05}
{"train_loss": 0.11817396432161331, "global_step": 106078, "epoch": 1191, "lr": 6.698679633770271e-05}
{"train_loss": 0.17286857962608337, "global_step": 106079, "epoch": 1191, "lr": 6.698625108218843e-05}
{"train_loss": 0.20997262001037598, "global_step": 106080, "epoch": 1191, "lr": 6.698570582439056e-05}
{"train_loss": 0.1337275207042694, "global_step": 106081, "epoch": 1191, "lr": 6.698516056430917e-05}
{"train_loss": 0.1624232977628708, "global_step": 106082, "epoch": 1191, "lr": 6.698461530194435e-05}
{"train_loss": 0.10772781074047089, "global_step": 106083, "epoch": 1191, "lr": 6.698407003729612e-05}
{"train_loss": 0.16987358033657074, "global_step": 106084, "epoch": 1191, "lr": 6.69835247703646e-05}
{"train_loss": 0.1398160457611084, "global_step": 106085, "epoch": 1191, "lr": 6.698297950114986e-05}
{"train_loss": 0.19036778807640076, "global_step": 106086, "epoch": 1191, "lr": 6.698243422965195e-05}
{"train_loss": 0.14468238063240318, "global_step": 106087, "epoch": 1191, "lr": 6.698188895587099e-05, "val_loss": 5.070174694061279}
{"train_loss": 0.17348991334438324, "global_step": 106088, "epoch": 1192, "lr": 6.6981343679807e-05}
{"train_loss": 0.1461666077375412, "global_step": 106089, "epoch": 1192, "lr": 6.698079840146007e-05}
{"train_loss": 0.15754489600658417, "global_step": 106090, "epoch": 1192, "lr": 6.69802531208303e-05}
{"train_loss": 0.15061424672603607, "global_step": 106091, "epoch": 1192, "lr": 6.697970783791773e-05}
{"train_loss": 0.17371661961078644, "global_step": 106092, "epoch": 1192, "lr": 6.697916255272245e-05}
{"train_loss": 0.11935149878263474, "global_step": 106093, "epoch": 1192, "lr": 6.697861726524453e-05}
{"train_loss": 0.12111621350049973, "global_step": 106094, "epoch": 1192, "lr": 6.697807197548403e-05}
{"train_loss": 0.21219639480113983, "global_step": 106095, "epoch": 1192, "lr": 6.697752668344103e-05}
{"train_loss": 0.12802588939666748, "global_step": 106096, "epoch": 1192, "lr": 6.697698138911562e-05}
{"train_loss": 0.14188367128372192, "global_step": 106097, "epoch": 1192, "lr": 6.697643609250787e-05}
{"train_loss": 0.20821620523929596, "global_step": 106098, "epoch": 1192, "lr": 6.697589079361784e-05}
{"train_loss": 0.148225799202919, "global_step": 106099, "epoch": 1192, "lr": 6.69753454924456e-05}
{"train_loss": 0.14370131492614746, "global_step": 106100, "epoch": 1192, "lr": 6.697480018899123e-05}
{"train_loss": 0.12436734139919281, "global_step": 106101, "epoch": 1192, "lr": 6.697425488325481e-05}
{"train_loss": 0.20817266404628754, "global_step": 106102, "epoch": 1192, "lr": 6.69737095752364e-05}
{"train_loss": 0.13284385204315186, "global_step": 106103, "epoch": 1192, "lr": 6.69731642649361e-05}
{"train_loss": 0.11499680578708649, "global_step": 106104, "epoch": 1192, "lr": 6.697261895235395e-05}
{"train_loss": 0.16148829460144043, "global_step": 106105, "epoch": 1192, "lr": 6.697207363749004e-05}
{"train_loss": 0.2201491892337799, "global_step": 106106, "epoch": 1192, "lr": 6.697152832034444e-05}
{"train_loss": 0.2062034159898758, "global_step": 106107, "epoch": 1192, "lr": 6.697098300091723e-05}
{"train_loss": 0.18952928483486176, "global_step": 106108, "epoch": 1192, "lr": 6.697043767920849e-05}
{"train_loss": 0.13988353312015533, "global_step": 106109, "epoch": 1192, "lr": 6.696989235521827e-05}
{"train_loss": 0.15946194529533386, "global_step": 106110, "epoch": 1192, "lr": 6.696934702894665e-05}
{"train_loss": 0.09743747860193253, "global_step": 106111, "epoch": 1192, "lr": 6.69688017003937e-05}
{"train_loss": 0.18651308119297028, "global_step": 106112, "epoch": 1192, "lr": 6.696825636955953e-05}
{"train_loss": 0.15061211585998535, "global_step": 106113, "epoch": 1192, "lr": 6.696771103644417e-05}
{"train_loss": 0.2338789999485016, "global_step": 106114, "epoch": 1192, "lr": 6.69671657010477e-05}
{"train_loss": 0.13291381299495697, "global_step": 106115, "epoch": 1192, "lr": 6.696662036337022e-05}
{"train_loss": 0.14164143800735474, "global_step": 106116, "epoch": 1192, "lr": 6.696607502341177e-05}
{"train_loss": 0.1284077763557434, "global_step": 106117, "epoch": 1192, "lr": 6.696552968117245e-05}
{"train_loss": 0.2359381765127182, "global_step": 106118, "epoch": 1192, "lr": 6.69649843366523e-05}
{"train_loss": 0.22307713329792023, "global_step": 106119, "epoch": 1192, "lr": 6.696443898985143e-05}
{"train_loss": 0.09486842155456543, "global_step": 106120, "epoch": 1192, "lr": 6.696389364076989e-05}
{"train_loss": 0.1880568116903305, "global_step": 106121, "epoch": 1192, "lr": 6.696334828940777e-05}
{"train_loss": 0.13882115483283997, "global_step": 106122, "epoch": 1192, "lr": 6.696280293576512e-05}
{"train_loss": 0.1872064620256424, "global_step": 106123, "epoch": 1192, "lr": 6.696225757984205e-05}
{"train_loss": 0.17686030268669128, "global_step": 106124, "epoch": 1192, "lr": 6.69617122216386e-05}
{"train_loss": 0.2245907187461853, "global_step": 106125, "epoch": 1192, "lr": 6.696116686115484e-05}
{"train_loss": 0.13216575980186462, "global_step": 106126, "epoch": 1192, "lr": 6.696062149839089e-05}
{"train_loss": 0.2873991131782532, "global_step": 106127, "epoch": 1192, "lr": 6.696007613334676e-05}
{"train_loss": 0.25993770360946655, "global_step": 106128, "epoch": 1192, "lr": 6.695953076602257e-05}
{"train_loss": 0.18792462348937988, "global_step": 106129, "epoch": 1192, "lr": 6.695898539641838e-05}
{"train_loss": 0.2202228158712387, "global_step": 106130, "epoch": 1192, "lr": 6.695844002453426e-05}
{"train_loss": 0.1458074450492859, "global_step": 106131, "epoch": 1192, "lr": 6.69578946503703e-05}
{"train_loss": 0.09991011768579483, "global_step": 106132, "epoch": 1192, "lr": 6.695734927392653e-05}
{"train_loss": 0.1432335078716278, "global_step": 106133, "epoch": 1192, "lr": 6.695680389520305e-05}
{"train_loss": 0.20882442593574524, "global_step": 106134, "epoch": 1192, "lr": 6.695625851419996e-05}
{"train_loss": 0.14112615585327148, "global_step": 106135, "epoch": 1192, "lr": 6.695571313091729e-05}
{"train_loss": 0.12425929307937622, "global_step": 106136, "epoch": 1192, "lr": 6.695516774535513e-05}
{"train_loss": 0.19641423225402832, "global_step": 106137, "epoch": 1192, "lr": 6.695462235751356e-05}
{"train_loss": 0.1447412222623825, "global_step": 106138, "epoch": 1192, "lr": 6.695407696739263e-05}
{"train_loss": 0.1766406148672104, "global_step": 106139, "epoch": 1192, "lr": 6.695353157499245e-05}
{"train_loss": 0.07905425131320953, "global_step": 106140, "epoch": 1192, "lr": 6.695298618031308e-05}
{"train_loss": 0.11152134090662003, "global_step": 106141, "epoch": 1192, "lr": 6.695244078335456e-05}
{"train_loss": 0.13536852598190308, "global_step": 106142, "epoch": 1192, "lr": 6.695189538411702e-05}
{"train_loss": 0.18187496066093445, "global_step": 106143, "epoch": 1192, "lr": 6.695134998260048e-05}
{"train_loss": 0.16920673847198486, "global_step": 106144, "epoch": 1192, "lr": 6.695080457880506e-05}
{"train_loss": 0.250710129737854, "global_step": 106145, "epoch": 1192, "lr": 6.69502591727308e-05}
{"train_loss": 0.11131879687309265, "global_step": 106146, "epoch": 1192, "lr": 6.694971376437779e-05}
{"train_loss": 0.1856617033481598, "global_step": 106147, "epoch": 1192, "lr": 6.694916835374609e-05}
{"train_loss": 0.17702633142471313, "global_step": 106148, "epoch": 1192, "lr": 6.694862294083578e-05}
{"train_loss": 0.17244142293930054, "global_step": 106149, "epoch": 1192, "lr": 6.694807752564695e-05}
{"train_loss": 0.19710995256900787, "global_step": 106150, "epoch": 1192, "lr": 6.694753210817963e-05}
{"train_loss": 0.15030129253864288, "global_step": 106151, "epoch": 1192, "lr": 6.694698668843395e-05}
{"train_loss": 0.14093171060085297, "global_step": 106152, "epoch": 1192, "lr": 6.694644126640995e-05}
{"train_loss": 0.1830555498600006, "global_step": 106153, "epoch": 1192, "lr": 6.694589584210769e-05}
{"train_loss": 0.10336097329854965, "global_step": 106154, "epoch": 1192, "lr": 6.694535041552727e-05}
{"train_loss": 0.14604191482067108, "global_step": 106155, "epoch": 1192, "lr": 6.694480498666877e-05}
{"train_loss": 0.18098601698875427, "global_step": 106156, "epoch": 1192, "lr": 6.69442595555322e-05}
{"train_loss": 0.13146308064460754, "global_step": 106157, "epoch": 1192, "lr": 6.694371412211774e-05}
{"train_loss": 0.15887032449245453, "global_step": 106158, "epoch": 1192, "lr": 6.694316868642536e-05}
{"train_loss": 0.19763700664043427, "global_step": 106159, "epoch": 1192, "lr": 6.694262324845519e-05}
{"train_loss": 0.1826738566160202, "global_step": 106160, "epoch": 1192, "lr": 6.694207780820729e-05}
{"train_loss": 0.15568630397319794, "global_step": 106161, "epoch": 1192, "lr": 6.694153236568175e-05}
{"train_loss": 0.09819093346595764, "global_step": 106162, "epoch": 1192, "lr": 6.694098692087862e-05}
{"train_loss": 0.1554499715566635, "global_step": 106163, "epoch": 1192, "lr": 6.694044147379797e-05}
{"train_loss": 0.12722957134246826, "global_step": 106164, "epoch": 1192, "lr": 6.69398960244399e-05}
{"train_loss": 0.11341634392738342, "global_step": 106165, "epoch": 1192, "lr": 6.693935057280445e-05}
{"train_loss": 0.187583327293396, "global_step": 106166, "epoch": 1192, "lr": 6.693880511889172e-05}
{"train_loss": 0.21282172203063965, "global_step": 106167, "epoch": 1192, "lr": 6.693825966270177e-05}
{"train_loss": 0.10923837125301361, "global_step": 106168, "epoch": 1192, "lr": 6.693771420423469e-05}
{"train_loss": 0.08855762332677841, "global_step": 106169, "epoch": 1192, "lr": 6.693716874349053e-05}
{"train_loss": 0.1354210525751114, "global_step": 106170, "epoch": 1192, "lr": 6.693662328046939e-05}
{"train_loss": 0.13379795849323273, "global_step": 106171, "epoch": 1192, "lr": 6.693607781517131e-05}
{"train_loss": 0.16448292136192322, "global_step": 106172, "epoch": 1192, "lr": 6.693553234759638e-05}
{"train_loss": 0.14758935570716858, "global_step": 106173, "epoch": 1192, "lr": 6.693498687774469e-05}
{"train_loss": 0.14908196032047272, "global_step": 106174, "epoch": 1192, "lr": 6.693444140561626e-05}
{"train_loss": 0.1583547294139862, "global_step": 106175, "epoch": 1192, "lr": 6.693389593121124e-05}
{"train_loss": 0.16148578409063682, "global_step": 106176, "epoch": 1192, "lr": 6.693335045452964e-05, "val_loss": 4.955624580383301}
{"train_loss": 0.13153980672359467, "global_step": 106177, "epoch": 1193, "lr": 6.693280497557157e-05}
{"train_loss": 0.1159084215760231, "global_step": 106178, "epoch": 1193, "lr": 6.693225949433709e-05}
{"train_loss": 0.1447450965642929, "global_step": 106179, "epoch": 1193, "lr": 6.693171401082628e-05}
{"train_loss": 0.16841533780097961, "global_step": 106180, "epoch": 1193, "lr": 6.693116852503919e-05}
{"train_loss": 0.20340703427791595, "global_step": 106181, "epoch": 1193, "lr": 6.693062303697593e-05}
{"train_loss": 0.17702127993106842, "global_step": 106182, "epoch": 1193, "lr": 6.693007754663654e-05}
{"train_loss": 0.19910606741905212, "global_step": 106183, "epoch": 1193, "lr": 6.69295320540211e-05}
{"train_loss": 0.10060009360313416, "global_step": 106184, "epoch": 1193, "lr": 6.692898655912971e-05}
{"train_loss": 0.16118352115154266, "global_step": 106185, "epoch": 1193, "lr": 6.692844106196242e-05}
{"train_loss": 0.17714543640613556, "global_step": 106186, "epoch": 1193, "lr": 6.69278955625193e-05}
{"train_loss": 0.17420963943004608, "global_step": 106187, "epoch": 1193, "lr": 6.692735006080045e-05}
{"train_loss": 0.14268717169761658, "global_step": 106188, "epoch": 1193, "lr": 6.69268045568059e-05}
{"train_loss": 0.17757485806941986, "global_step": 106189, "epoch": 1193, "lr": 6.692625905053577e-05}
{"train_loss": 0.22097063064575195, "global_step": 106190, "epoch": 1193, "lr": 6.69257135419901e-05}
{"train_loss": 0.16619297862052917, "global_step": 106191, "epoch": 1193, "lr": 6.692516803116898e-05}
{"train_loss": 0.20309105515480042, "global_step": 106192, "epoch": 1193, "lr": 6.692462251807245e-05}
{"train_loss": 0.16293731331825256, "global_step": 106193, "epoch": 1193, "lr": 6.692407700270063e-05}
{"train_loss": 0.2150774896144867, "global_step": 106194, "epoch": 1193, "lr": 6.692353148505357e-05}
{"train_loss": 0.2081749141216278, "global_step": 106195, "epoch": 1193, "lr": 6.692298596513135e-05}
{"train_loss": 0.11977771669626236, "global_step": 106196, "epoch": 1193, "lr": 6.692244044293404e-05}
{"train_loss": 0.14666812121868134, "global_step": 106197, "epoch": 1193, "lr": 6.692189491846173e-05}
{"train_loss": 0.10844942182302475, "global_step": 106198, "epoch": 1193, "lr": 6.692134939171446e-05}
{"train_loss": 0.07586098462343216, "global_step": 106199, "epoch": 1193, "lr": 6.692080386269233e-05}
{"train_loss": 0.1431843489408493, "global_step": 106200, "epoch": 1193, "lr": 6.69202583313954e-05}
{"train_loss": 0.24565571546554565, "global_step": 106201, "epoch": 1193, "lr": 6.691971279782374e-05}
{"train_loss": 0.1436196118593216, "global_step": 106202, "epoch": 1193, "lr": 6.691916726197744e-05}
{"train_loss": 0.1481194794178009, "global_step": 106203, "epoch": 1193, "lr": 6.691862172385656e-05}
{"train_loss": 0.19632723927497864, "global_step": 106204, "epoch": 1193, "lr": 6.691807618346118e-05}
{"train_loss": 0.12368490546941757, "global_step": 106205, "epoch": 1193, "lr": 6.691753064079135e-05}
{"train_loss": 0.14515510201454163, "global_step": 106206, "epoch": 1193, "lr": 6.69169850958472e-05}
{"train_loss": 0.17373797297477722, "global_step": 106207, "epoch": 1193, "lr": 6.691643954862876e-05}
{"train_loss": 0.14169536530971527, "global_step": 106208, "epoch": 1193, "lr": 6.691589399913609e-05}
{"train_loss": 0.17346858978271484, "global_step": 106209, "epoch": 1193, "lr": 6.691534844736931e-05}
{"train_loss": 0.1410273164510727, "global_step": 106210, "epoch": 1193, "lr": 6.691480289332846e-05}
{"train_loss": 0.16048075258731842, "global_step": 106211, "epoch": 1193, "lr": 6.69142573370136e-05}
{"train_loss": 0.1409817785024643, "global_step": 106212, "epoch": 1193, "lr": 6.691371177842484e-05}
{"train_loss": 0.2038702368736267, "global_step": 106213, "epoch": 1193, "lr": 6.691316621756224e-05}
{"train_loss": 0.134752094745636, "global_step": 106214, "epoch": 1193, "lr": 6.691262065442586e-05}
{"train_loss": 0.14440983533859253, "global_step": 106215, "epoch": 1193, "lr": 6.69120750890158e-05}
{"train_loss": 0.19837544858455658, "global_step": 106216, "epoch": 1193, "lr": 6.69115295213321e-05}
{"train_loss": 0.16425903141498566, "global_step": 106217, "epoch": 1193, "lr": 6.691098395137487e-05}
{"train_loss": 0.1789984107017517, "global_step": 106218, "epoch": 1193, "lr": 6.691043837914416e-05}
{"train_loss": 0.13942751288414001, "global_step": 106219, "epoch": 1193, "lr": 6.690989280464004e-05}
{"train_loss": 0.12494119256734848, "global_step": 106220, "epoch": 1193, "lr": 6.69093472278626e-05}
{"train_loss": 0.1791699081659317, "global_step": 106221, "epoch": 1193, "lr": 6.69088016488119e-05}
{"train_loss": 0.12717784941196442, "global_step": 106222, "epoch": 1193, "lr": 6.6908256067488e-05}
{"train_loss": 0.11385171860456467, "global_step": 106223, "epoch": 1193, "lr": 6.690771048389101e-05}
{"train_loss": 0.12605583667755127, "global_step": 106224, "epoch": 1193, "lr": 6.690716489802098e-05}
{"train_loss": 0.24177952110767365, "global_step": 106225, "epoch": 1193, "lr": 6.6906619309878e-05}
{"train_loss": 0.21079224348068237, "global_step": 106226, "epoch": 1193, "lr": 6.690607371946212e-05}
{"train_loss": 0.16887050867080688, "global_step": 106227, "epoch": 1193, "lr": 6.690552812677342e-05}
{"train_loss": 0.22006477415561676, "global_step": 106228, "epoch": 1193, "lr": 6.690498253181198e-05}
{"train_loss": 0.11487369239330292, "global_step": 106229, "epoch": 1193, "lr": 6.690443693457787e-05}
{"train_loss": 0.11175472289323807, "global_step": 106230, "epoch": 1193, "lr": 6.690389133507117e-05}
{"train_loss": 0.14849834144115448, "global_step": 106231, "epoch": 1193, "lr": 6.690334573329195e-05}
{"train_loss": 0.16635183990001678, "global_step": 106232, "epoch": 1193, "lr": 6.690280012924028e-05}
{"train_loss": 0.12254549562931061, "global_step": 106233, "epoch": 1193, "lr": 6.690225452291623e-05}
{"train_loss": 0.16762502491474152, "global_step": 106234, "epoch": 1193, "lr": 6.690170891431988e-05}
{"train_loss": 0.10192270576953888, "global_step": 106235, "epoch": 1193, "lr": 6.69011633034513e-05}
{"train_loss": 0.1213098093867302, "global_step": 106236, "epoch": 1193, "lr": 6.690061769031056e-05}
{"train_loss": 0.1728963851928711, "global_step": 106237, "epoch": 1193, "lr": 6.690007207489774e-05}
{"train_loss": 0.13811329007148743, "global_step": 106238, "epoch": 1193, "lr": 6.689952645721292e-05}
{"train_loss": 0.12957924604415894, "global_step": 106239, "epoch": 1193, "lr": 6.689898083725616e-05}
{"train_loss": 0.1628807634115219, "global_step": 106240, "epoch": 1193, "lr": 6.689843521502755e-05}
{"train_loss": 0.13733907043933868, "global_step": 106241, "epoch": 1193, "lr": 6.689788959052713e-05}
{"train_loss": 0.13444533944129944, "global_step": 106242, "epoch": 1193, "lr": 6.6897343963755e-05}
{"train_loss": 0.2018120288848877, "global_step": 106243, "epoch": 1193, "lr": 6.689679833471124e-05}
{"train_loss": 0.1501755714416504, "global_step": 106244, "epoch": 1193, "lr": 6.68962527033959e-05}
{"train_loss": 0.11457795649766922, "global_step": 106245, "epoch": 1193, "lr": 6.689570706980906e-05}
{"train_loss": 0.13249272108078003, "global_step": 106246, "epoch": 1193, "lr": 6.689516143395082e-05}
{"train_loss": 0.18703871965408325, "global_step": 106247, "epoch": 1193, "lr": 6.689461579582121e-05}
{"train_loss": 0.16667047142982483, "global_step": 106248, "epoch": 1193, "lr": 6.689407015542033e-05}
{"train_loss": 0.1866270750761032, "global_step": 106249, "epoch": 1193, "lr": 6.689352451274826e-05}
{"train_loss": 0.23283052444458008, "global_step": 106250, "epoch": 1193, "lr": 6.689297886780505e-05}
{"train_loss": 0.18234673142433167, "global_step": 106251, "epoch": 1193, "lr": 6.689243322059077e-05}
{"train_loss": 0.11337478458881378, "global_step": 106252, "epoch": 1193, "lr": 6.689188757110554e-05}
{"train_loss": 0.13013949990272522, "global_step": 106253, "epoch": 1193, "lr": 6.689134191934938e-05}
{"train_loss": 0.14487624168395996, "global_step": 106254, "epoch": 1193, "lr": 6.68907962653224e-05}
{"train_loss": 0.15996041893959045, "global_step": 106255, "epoch": 1193, "lr": 6.689025060902464e-05}
{"train_loss": 0.19507676362991333, "global_step": 106256, "epoch": 1193, "lr": 6.68897049504562e-05}
{"train_loss": 0.17344234883785248, "global_step": 106257, "epoch": 1193, "lr": 6.688915928961716e-05}
{"train_loss": 0.07537127286195755, "global_step": 106258, "epoch": 1193, "lr": 6.688861362650756e-05}
{"train_loss": 0.09811541438102722, "global_step": 106259, "epoch": 1193, "lr": 6.688806796112749e-05}
{"train_loss": 0.14520105719566345, "global_step": 106260, "epoch": 1193, "lr": 6.688752229347703e-05}
{"train_loss": 0.1595894694328308, "global_step": 106261, "epoch": 1193, "lr": 6.688697662355625e-05}
{"train_loss": 0.18132749199867249, "global_step": 106262, "epoch": 1193, "lr": 6.688643095136522e-05}
{"train_loss": 0.19968417286872864, "global_step": 106263, "epoch": 1193, "lr": 6.688588527690402e-05}
{"train_loss": 0.20793145895004272, "global_step": 106264, "epoch": 1193, "lr": 6.688533960017271e-05}
{"train_loss": 0.15899242033784308, "global_step": 106265, "epoch": 1193, "lr": 6.688479392117139e-05, "val_loss": 4.972009181976318}
{"train_loss": 0.15824703872203827, "global_step": 106266, "epoch": 1194, "lr": 6.688424823990009e-05}
{"train_loss": 0.14362020790576935, "global_step": 106267, "epoch": 1194, "lr": 6.688370255635893e-05}
{"train_loss": 0.16172121465206146, "global_step": 106268, "epoch": 1194, "lr": 6.688315687054795e-05}
{"train_loss": 0.15591634809970856, "global_step": 106269, "epoch": 1194, "lr": 6.688261118246724e-05}
{"train_loss": 0.1726016402244568, "global_step": 106270, "epoch": 1194, "lr": 6.688206549211686e-05}
{"train_loss": 0.1599380075931549, "global_step": 106271, "epoch": 1194, "lr": 6.68815197994969e-05}
{"train_loss": 0.10303209722042084, "global_step": 106272, "epoch": 1194, "lr": 6.688097410460742e-05}
{"train_loss": 0.1545131951570511, "global_step": 106273, "epoch": 1194, "lr": 6.688042840744851e-05}
{"train_loss": 0.23836854100227356, "global_step": 106274, "epoch": 1194, "lr": 6.687988270802024e-05}
{"train_loss": 0.20856937766075134, "global_step": 106275, "epoch": 1194, "lr": 6.687933700632265e-05}
{"train_loss": 0.10868076235055923, "global_step": 106276, "epoch": 1194, "lr": 6.687879130235585e-05}
{"train_loss": 0.17066583037376404, "global_step": 106277, "epoch": 1194, "lr": 6.68782455961199e-05}
{"train_loss": 0.22648313641548157, "global_step": 106278, "epoch": 1194, "lr": 6.687769988761488e-05}
{"train_loss": 0.13272687792778015, "global_step": 106279, "epoch": 1194, "lr": 6.687715417684087e-05}
{"train_loss": 0.10124902427196503, "global_step": 106280, "epoch": 1194, "lr": 6.68766084637979e-05}
{"train_loss": 0.19466185569763184, "global_step": 106281, "epoch": 1194, "lr": 6.68760627484861e-05}
{"train_loss": 0.17938636243343353, "global_step": 106282, "epoch": 1194, "lr": 6.687551703090552e-05}
{"train_loss": 0.1343921273946762, "global_step": 106283, "epoch": 1194, "lr": 6.687497131105623e-05}
{"train_loss": 0.1888844072818756, "global_step": 106284, "epoch": 1194, "lr": 6.68744255889383e-05}
{"train_loss": 0.12951479852199554, "global_step": 106285, "epoch": 1194, "lr": 6.687387986455182e-05}
{"train_loss": 0.1677752137184143, "global_step": 106286, "epoch": 1194, "lr": 6.687333413789683e-05}
{"train_loss": 0.1864449679851532, "global_step": 106287, "epoch": 1194, "lr": 6.687278840897343e-05}
{"train_loss": 0.14339695870876312, "global_step": 106288, "epoch": 1194, "lr": 6.687224267778171e-05}
{"train_loss": 0.12516507506370544, "global_step": 106289, "epoch": 1194, "lr": 6.687169694432172e-05}
{"train_loss": 0.15452954173088074, "global_step": 106290, "epoch": 1194, "lr": 6.687115120859353e-05}
{"train_loss": 0.1930965930223465, "global_step": 106291, "epoch": 1194, "lr": 6.687060547059722e-05}
{"train_loss": 0.154828742146492, "global_step": 106292, "epoch": 1194, "lr": 6.687005973033285e-05}
{"train_loss": 0.17766602337360382, "global_step": 106293, "epoch": 1194, "lr": 6.686951398780051e-05}
{"train_loss": 0.14258092641830444, "global_step": 106294, "epoch": 1194, "lr": 6.686896824300029e-05}
{"train_loss": 0.1341911405324936, "global_step": 106295, "epoch": 1194, "lr": 6.686842249593221e-05}
{"train_loss": 0.14123433828353882, "global_step": 106296, "epoch": 1194, "lr": 6.686787674659641e-05}
{"train_loss": 0.15735751390457153, "global_step": 106297, "epoch": 1194, "lr": 6.686733099499291e-05}
{"train_loss": 0.11706554144620895, "global_step": 106298, "epoch": 1194, "lr": 6.68667852411218e-05}
{"train_loss": 0.23542222380638123, "global_step": 106299, "epoch": 1194, "lr": 6.686623948498317e-05}
{"train_loss": 0.12912572920322418, "global_step": 106300, "epoch": 1194, "lr": 6.686569372657707e-05}
{"train_loss": 0.14493562281131744, "global_step": 106301, "epoch": 1194, "lr": 6.686514796590359e-05}
{"train_loss": 0.1094619408249855, "global_step": 106302, "epoch": 1194, "lr": 6.686460220296279e-05}
{"train_loss": 0.20923079550266266, "global_step": 106303, "epoch": 1194, "lr": 6.686405643775475e-05}
{"train_loss": 0.1522606611251831, "global_step": 106304, "epoch": 1194, "lr": 6.686351067027955e-05}
{"train_loss": 0.15234175324440002, "global_step": 106305, "epoch": 1194, "lr": 6.686296490053725e-05}
{"train_loss": 0.13812929391860962, "global_step": 106306, "epoch": 1194, "lr": 6.686241912852793e-05}
{"train_loss": 0.221824511885643, "global_step": 106307, "epoch": 1194, "lr": 6.686187335425167e-05}
{"train_loss": 0.16202937066555023, "global_step": 106308, "epoch": 1194, "lr": 6.686132757770854e-05}
{"train_loss": 0.22161683440208435, "global_step": 106309, "epoch": 1194, "lr": 6.686078179889859e-05}
{"train_loss": 0.14618676900863647, "global_step": 106310, "epoch": 1194, "lr": 6.686023601782192e-05}
{"train_loss": 0.0930587500333786, "global_step": 106311, "epoch": 1194, "lr": 6.68596902344786e-05}
{"train_loss": 0.09852674603462219, "global_step": 106312, "epoch": 1194, "lr": 6.685914444886871e-05}
{"train_loss": 0.1558954268693924, "global_step": 106313, "epoch": 1194, "lr": 6.68585986609923e-05}
{"train_loss": 0.1395936757326126, "global_step": 106314, "epoch": 1194, "lr": 6.685805287084947e-05}
{"train_loss": 0.1561804562807083, "global_step": 106315, "epoch": 1194, "lr": 6.685750707844027e-05}
{"train_loss": 0.20669327676296234, "global_step": 106316, "epoch": 1194, "lr": 6.685696128376479e-05}
{"train_loss": 0.1862289011478424, "global_step": 106317, "epoch": 1194, "lr": 6.685641548682309e-05}
{"train_loss": 0.13364002108573914, "global_step": 106318, "epoch": 1194, "lr": 6.685586968761525e-05}
{"train_loss": 0.1520339995622635, "global_step": 106319, "epoch": 1194, "lr": 6.685532388614135e-05}
{"train_loss": 0.20507849752902985, "global_step": 106320, "epoch": 1194, "lr": 6.685477808240145e-05}
{"train_loss": 0.16120962798595428, "global_step": 106321, "epoch": 1194, "lr": 6.685423227639565e-05}
{"train_loss": 0.11884944885969162, "global_step": 106322, "epoch": 1194, "lr": 6.685368646812398e-05}
{"train_loss": 0.20267251133918762, "global_step": 106323, "epoch": 1194, "lr": 6.685314065758655e-05}
{"train_loss": 0.21635928750038147, "global_step": 106324, "epoch": 1194, "lr": 6.68525948447834e-05}
{"train_loss": 0.12929607927799225, "global_step": 106325, "epoch": 1194, "lr": 6.685204902971464e-05}
{"train_loss": 0.20177948474884033, "global_step": 106326, "epoch": 1194, "lr": 6.685150321238034e-05}
{"train_loss": 0.12049777060747147, "global_step": 106327, "epoch": 1194, "lr": 6.685095739278054e-05}
{"train_loss": 0.12992238998413086, "global_step": 106328, "epoch": 1194, "lr": 6.685041157091535e-05}
{"train_loss": 0.15639130771160126, "global_step": 106329, "epoch": 1194, "lr": 6.684986574678482e-05}
{"train_loss": 0.20561109483242035, "global_step": 106330, "epoch": 1194, "lr": 6.684931992038903e-05}
{"train_loss": 0.12667620182037354, "global_step": 106331, "epoch": 1194, "lr": 6.684877409172806e-05}
{"train_loss": 0.1348089873790741, "global_step": 106332, "epoch": 1194, "lr": 6.684822826080196e-05}
{"train_loss": 0.1772184520959854, "global_step": 106333, "epoch": 1194, "lr": 6.684768242761085e-05}
{"train_loss": 0.16630525887012482, "global_step": 106334, "epoch": 1194, "lr": 6.684713659215476e-05}
{"train_loss": 0.1378660798072815, "global_step": 106335, "epoch": 1194, "lr": 6.684659075443378e-05}
{"train_loss": 0.12246186286211014, "global_step": 106336, "epoch": 1194, "lr": 6.684604491444796e-05}
{"train_loss": 0.17489586770534515, "global_step": 106337, "epoch": 1194, "lr": 6.684549907219743e-05}
{"train_loss": 0.11718795448541641, "global_step": 106338, "epoch": 1194, "lr": 6.68449532276822e-05}
{"train_loss": 0.17189978063106537, "global_step": 106339, "epoch": 1194, "lr": 6.684440738090238e-05}
{"train_loss": 0.15986128151416779, "global_step": 106340, "epoch": 1194, "lr": 6.684386153185804e-05}
{"train_loss": 0.211149200797081, "global_step": 106341, "epoch": 1194, "lr": 6.684331568054924e-05}
{"train_loss": 0.11850441247224808, "global_step": 106342, "epoch": 1194, "lr": 6.684276982697609e-05}
{"train_loss": 0.1142408475279808, "global_step": 106343, "epoch": 1194, "lr": 6.68422239711386e-05}
{"train_loss": 0.1426052302122116, "global_step": 106344, "epoch": 1194, "lr": 6.684167811303689e-05}
{"train_loss": 0.16556893289089203, "global_step": 106345, "epoch": 1194, "lr": 6.6841132252671e-05}
{"train_loss": 0.17344270646572113, "global_step": 106346, "epoch": 1194, "lr": 6.684058639004107e-05}
{"train_loss": 0.223650723695755, "global_step": 106347, "epoch": 1194, "lr": 6.68400405251471e-05}
{"train_loss": 0.173991397023201, "global_step": 106348, "epoch": 1194, "lr": 6.68394946579892e-05}
{"train_loss": 0.2165149599313736, "global_step": 106349, "epoch": 1194, "lr": 6.683894878856742e-05}
{"train_loss": 0.11161276698112488, "global_step": 106350, "epoch": 1194, "lr": 6.683840291688186e-05}
{"train_loss": 0.1222841814160347, "global_step": 106351, "epoch": 1194, "lr": 6.683785704293259e-05}
{"train_loss": 0.12682101130485535, "global_step": 106352, "epoch": 1194, "lr": 6.683731116671968e-05}
{"train_loss": 0.15265415608882904, "global_step": 106353, "epoch": 1194, "lr": 6.683676528824317e-05}
{"train_loss": 0.15900146793783382, "global_step": 106354, "epoch": 1194, "lr": 6.683621940750317e-05, "val_loss": 4.936441898345947}
{"train_loss": 0.11227795481681824, "global_step": 106355, "epoch": 1195, "lr": 6.683567352449977e-05}
{"train_loss": 0.1194574162364006, "global_step": 106356, "epoch": 1195, "lr": 6.683512763923298e-05}
{"train_loss": 0.11069130152463913, "global_step": 106357, "epoch": 1195, "lr": 6.683458175170294e-05}
{"train_loss": 0.10926571488380432, "global_step": 106358, "epoch": 1195, "lr": 6.68340358619097e-05}
{"train_loss": 0.1170513704419136, "global_step": 106359, "epoch": 1195, "lr": 6.683348996985332e-05}
{"train_loss": 0.1775255799293518, "global_step": 106360, "epoch": 1195, "lr": 6.68329440755339e-05}
{"train_loss": 0.09702320396900177, "global_step": 106361, "epoch": 1195, "lr": 6.683239817895147e-05}
{"train_loss": 0.18983308970928192, "global_step": 106362, "epoch": 1195, "lr": 6.683185228010613e-05}
{"train_loss": 0.1126500591635704, "global_step": 106363, "epoch": 1195, "lr": 6.683130637899796e-05}
{"train_loss": 0.1568426936864853, "global_step": 106364, "epoch": 1195, "lr": 6.683076047562703e-05}
{"train_loss": 0.20712463557720184, "global_step": 106365, "epoch": 1195, "lr": 6.683021456999341e-05}
{"train_loss": 0.2163715809583664, "global_step": 106366, "epoch": 1195, "lr": 6.682966866209716e-05}
{"train_loss": 0.2524920105934143, "global_step": 106367, "epoch": 1195, "lr": 6.682912275193838e-05}
{"train_loss": 0.09499779343605042, "global_step": 106368, "epoch": 1195, "lr": 6.682857683951712e-05}
{"train_loss": 0.13129286468029022, "global_step": 106369, "epoch": 1195, "lr": 6.682803092483346e-05}
{"train_loss": 0.1816542148590088, "global_step": 106370, "epoch": 1195, "lr": 6.682748500788749e-05}
{"train_loss": 0.19803164899349213, "global_step": 106371, "epoch": 1195, "lr": 6.682693908867927e-05}
{"train_loss": 0.1620064228773117, "global_step": 106372, "epoch": 1195, "lr": 6.682639316720886e-05}
{"train_loss": 0.15333303809165955, "global_step": 106373, "epoch": 1195, "lr": 6.682584724347634e-05}
{"train_loss": 0.16581350564956665, "global_step": 106374, "epoch": 1195, "lr": 6.68253013174818e-05}
{"train_loss": 0.1802126169204712, "global_step": 106375, "epoch": 1195, "lr": 6.68247553892253e-05}
{"train_loss": 0.15154016017913818, "global_step": 106376, "epoch": 1195, "lr": 6.682420945870693e-05}
{"train_loss": 0.18114745616912842, "global_step": 106377, "epoch": 1195, "lr": 6.682366352592676e-05}
{"train_loss": 0.16671673953533173, "global_step": 106378, "epoch": 1195, "lr": 6.682311759088483e-05}
{"train_loss": 0.16722197830677032, "global_step": 106379, "epoch": 1195, "lr": 6.682257165358125e-05}
{"train_loss": 0.19439707696437836, "global_step": 106380, "epoch": 1195, "lr": 6.682202571401607e-05}
{"train_loss": 0.18517111241817474, "global_step": 106381, "epoch": 1195, "lr": 6.682147977218937e-05}
{"train_loss": 0.24061354994773865, "global_step": 106382, "epoch": 1195, "lr": 6.682093382810125e-05}
{"train_loss": 0.10599636286497116, "global_step": 106383, "epoch": 1195, "lr": 6.682038788175173e-05}
{"train_loss": 0.1708402782678604, "global_step": 106384, "epoch": 1195, "lr": 6.681984193314092e-05}
{"train_loss": 0.13444620370864868, "global_step": 106385, "epoch": 1195, "lr": 6.68192959822689e-05}
{"train_loss": 0.17212197184562683, "global_step": 106386, "epoch": 1195, "lr": 6.681875002913573e-05}
{"train_loss": 0.19521811604499817, "global_step": 106387, "epoch": 1195, "lr": 6.681820407374149e-05}
{"train_loss": 0.10052141547203064, "global_step": 106388, "epoch": 1195, "lr": 6.681765811608623e-05}
{"train_loss": 0.21911771595478058, "global_step": 106389, "epoch": 1195, "lr": 6.681711215617005e-05}
{"train_loss": 0.16473859548568726, "global_step": 106390, "epoch": 1195, "lr": 6.681656619399301e-05}
{"train_loss": 0.2070300281047821, "global_step": 106391, "epoch": 1195, "lr": 6.681602022955519e-05}
{"train_loss": 0.14358870685100555, "global_step": 106392, "epoch": 1195, "lr": 6.681547426285666e-05}
{"train_loss": 0.18068140745162964, "global_step": 106393, "epoch": 1195, "lr": 6.681492829389748e-05}
{"train_loss": 0.24103881418704987, "global_step": 106394, "epoch": 1195, "lr": 6.681438232267776e-05}
{"train_loss": 0.2446228712797165, "global_step": 106395, "epoch": 1195, "lr": 6.681383634919755e-05}
{"train_loss": 0.24184709787368774, "global_step": 106396, "epoch": 1195, "lr": 6.681329037345691e-05}
{"train_loss": 0.18628519773483276, "global_step": 106397, "epoch": 1195, "lr": 6.681274439545594e-05}
{"train_loss": 0.1549459844827652, "global_step": 106398, "epoch": 1195, "lr": 6.681219841519471e-05}
{"train_loss": 0.18478341400623322, "global_step": 106399, "epoch": 1195, "lr": 6.681165243267328e-05}
{"train_loss": 0.18251550197601318, "global_step": 106400, "epoch": 1195, "lr": 6.681110644789173e-05}
{"train_loss": 0.19015510380268097, "global_step": 106401, "epoch": 1195, "lr": 6.681056046085011e-05}
{"train_loss": 0.10344283282756805, "global_step": 106402, "epoch": 1195, "lr": 6.681001447154853e-05}
{"train_loss": 0.20741185545921326, "global_step": 106403, "epoch": 1195, "lr": 6.680946847998705e-05}
{"train_loss": 0.1655297428369522, "global_step": 106404, "epoch": 1195, "lr": 6.680892248616573e-05}
{"train_loss": 0.15490303933620453, "global_step": 106405, "epoch": 1195, "lr": 6.680837649008467e-05}
{"train_loss": 0.13876093924045563, "global_step": 106406, "epoch": 1195, "lr": 6.680783049174392e-05}
{"train_loss": 0.19059130549430847, "global_step": 106407, "epoch": 1195, "lr": 6.680728449114357e-05}
{"train_loss": 0.15032677352428436, "global_step": 106408, "epoch": 1195, "lr": 6.680673848828368e-05}
{"train_loss": 0.12114310264587402, "global_step": 106409, "epoch": 1195, "lr": 6.680619248316433e-05}
{"train_loss": 0.18485596776008606, "global_step": 106410, "epoch": 1195, "lr": 6.680564647578558e-05}
{"train_loss": 0.2131163775920868, "global_step": 106411, "epoch": 1195, "lr": 6.680510046614754e-05}
{"train_loss": 0.20615772902965546, "global_step": 106412, "epoch": 1195, "lr": 6.680455445425023e-05}
{"train_loss": 0.17770016193389893, "global_step": 106413, "epoch": 1195, "lr": 6.680400844009378e-05}
{"train_loss": 0.17373596131801605, "global_step": 106414, "epoch": 1195, "lr": 6.680346242367822e-05}
{"train_loss": 0.20013941824436188, "global_step": 106415, "epoch": 1195, "lr": 6.680291640500364e-05}
{"train_loss": 0.1526479423046112, "global_step": 106416, "epoch": 1195, "lr": 6.680237038407014e-05}
{"train_loss": 0.18920905888080597, "global_step": 106417, "epoch": 1195, "lr": 6.680182436087772e-05}
{"train_loss": 0.15055717527866364, "global_step": 106418, "epoch": 1195, "lr": 6.680127833542654e-05}
{"train_loss": 0.1308211237192154, "global_step": 106419, "epoch": 1195, "lr": 6.680073230771661e-05}
{"train_loss": 0.1910393238067627, "global_step": 106420, "epoch": 1195, "lr": 6.680018627774804e-05}
{"train_loss": 0.1599004864692688, "global_step": 106421, "epoch": 1195, "lr": 6.679964024552087e-05}
{"train_loss": 0.15221481025218964, "global_step": 106422, "epoch": 1195, "lr": 6.679909421103521e-05}
{"train_loss": 0.17568808794021606, "global_step": 106423, "epoch": 1195, "lr": 6.679854817429111e-05}
{"train_loss": 0.18362551927566528, "global_step": 106424, "epoch": 1195, "lr": 6.679800213528866e-05}
{"train_loss": 0.19497938454151154, "global_step": 106425, "epoch": 1195, "lr": 6.679745609402789e-05}
{"train_loss": 0.1680406630039215, "global_step": 106426, "epoch": 1195, "lr": 6.679691005050894e-05}
{"train_loss": 0.16396477818489075, "global_step": 106427, "epoch": 1195, "lr": 6.679636400473185e-05}
{"train_loss": 0.12664778530597687, "global_step": 106428, "epoch": 1195, "lr": 6.679581795669666e-05}
{"train_loss": 0.18816396594047546, "global_step": 106429, "epoch": 1195, "lr": 6.679527190640351e-05}
{"train_loss": 0.14552408456802368, "global_step": 106430, "epoch": 1195, "lr": 6.679472585385243e-05}
{"train_loss": 0.16856840252876282, "global_step": 106431, "epoch": 1195, "lr": 6.67941797990435e-05}
{"train_loss": 0.1500278264284134, "global_step": 106432, "epoch": 1195, "lr": 6.67936337419768e-05}
{"train_loss": 0.13813528418540955, "global_step": 106433, "epoch": 1195, "lr": 6.679308768265242e-05}
{"train_loss": 0.12352370470762253, "global_step": 106434, "epoch": 1195, "lr": 6.679254162107039e-05}
{"train_loss": 0.10282312333583832, "global_step": 106435, "epoch": 1195, "lr": 6.679199555723083e-05}
{"train_loss": 0.13902586698532104, "global_step": 106436, "epoch": 1195, "lr": 6.679144949113377e-05}
{"train_loss": 0.10182952135801315, "global_step": 106437, "epoch": 1195, "lr": 6.679090342277931e-05}
{"train_loss": 0.2235579788684845, "global_step": 106438, "epoch": 1195, "lr": 6.679035735216753e-05}
{"train_loss": 0.20382487773895264, "global_step": 106439, "epoch": 1195, "lr": 6.678981127929847e-05}
{"train_loss": 0.11333455890417099, "global_step": 106440, "epoch": 1195, "lr": 6.678926520417224e-05}
{"train_loss": 0.19107095897197723, "global_step": 106441, "epoch": 1195, "lr": 6.678871912678889e-05}
{"train_loss": 0.10834670811891556, "global_step": 106442, "epoch": 1195, "lr": 6.678817304714851e-05}
{"train_loss": 0.1663994679457686, "global_step": 106443, "epoch": 1195, "lr": 6.678762696525116e-05, "val_loss": 4.82243013381958, "train_action_mse_error": 14.012737274169922}
{"train_loss": 0.10973698645830154, "global_step": 106444, "epoch": 1196, "lr": 6.678708088109691e-05}
{"train_loss": 0.17352715134620667, "global_step": 106445, "epoch": 1196, "lr": 6.678653479468584e-05}
{"train_loss": 0.18955962359905243, "global_step": 106446, "epoch": 1196, "lr": 6.678598870601804e-05}
{"train_loss": 0.18545764684677124, "global_step": 106447, "epoch": 1196, "lr": 6.678544261509353e-05}
{"train_loss": 0.1531427502632141, "global_step": 106448, "epoch": 1196, "lr": 6.678489652191247e-05}
{"train_loss": 0.15455585718154907, "global_step": 106449, "epoch": 1196, "lr": 6.678435042647487e-05}
{"train_loss": 0.1606413573026657, "global_step": 106450, "epoch": 1196, "lr": 6.678380432878081e-05}
{"train_loss": 0.12506619095802307, "global_step": 106451, "epoch": 1196, "lr": 6.678325822883038e-05}
{"train_loss": 0.19188468158245087, "global_step": 106452, "epoch": 1196, "lr": 6.678271212662365e-05}
{"train_loss": 0.1723259538412094, "global_step": 106453, "epoch": 1196, "lr": 6.678216602216067e-05}
{"train_loss": 0.09797113388776779, "global_step": 106454, "epoch": 1196, "lr": 6.678161991544155e-05}
{"train_loss": 0.18053141236305237, "global_step": 106455, "epoch": 1196, "lr": 6.678107380646636e-05}
{"train_loss": 0.150333970785141, "global_step": 106456, "epoch": 1196, "lr": 6.678052769523513e-05}
{"train_loss": 0.12949268519878387, "global_step": 106457, "epoch": 1196, "lr": 6.677998158174798e-05}
{"train_loss": 0.20086146891117096, "global_step": 106458, "epoch": 1196, "lr": 6.677943546600496e-05}
{"train_loss": 0.08670545369386673, "global_step": 106459, "epoch": 1196, "lr": 6.677888934800614e-05}
{"train_loss": 0.1382240355014801, "global_step": 106460, "epoch": 1196, "lr": 6.677834322775162e-05}
{"train_loss": 0.21179194748401642, "global_step": 106461, "epoch": 1196, "lr": 6.677779710524143e-05}
{"train_loss": 0.08731112629175186, "global_step": 106462, "epoch": 1196, "lr": 6.67772509804757e-05}
{"train_loss": 0.13069957494735718, "global_step": 106463, "epoch": 1196, "lr": 6.677670485345446e-05}
{"train_loss": 0.13370287418365479, "global_step": 106464, "epoch": 1196, "lr": 6.677615872417779e-05}
{"train_loss": 0.15519413352012634, "global_step": 106465, "epoch": 1196, "lr": 6.677561259264577e-05}
{"train_loss": 0.15451659262180328, "global_step": 106466, "epoch": 1196, "lr": 6.677506645885849e-05}
{"train_loss": 0.15241111814975739, "global_step": 106467, "epoch": 1196, "lr": 6.6774520322816e-05}
{"train_loss": 0.11840950697660446, "global_step": 106468, "epoch": 1196, "lr": 6.677397418451837e-05}
{"train_loss": 0.1689469963312149, "global_step": 106469, "epoch": 1196, "lr": 6.677342804396569e-05}
{"train_loss": 0.19847503304481506, "global_step": 106470, "epoch": 1196, "lr": 6.677288190115802e-05}
{"train_loss": 0.14028017222881317, "global_step": 106471, "epoch": 1196, "lr": 6.677233575609546e-05}
{"train_loss": 0.14334984123706818, "global_step": 106472, "epoch": 1196, "lr": 6.677178960877805e-05}
{"train_loss": 0.17308220267295837, "global_step": 106473, "epoch": 1196, "lr": 6.677124345920587e-05}
{"train_loss": 0.19500397145748138, "global_step": 106474, "epoch": 1196, "lr": 6.677069730737902e-05}
{"train_loss": 0.14824287593364716, "global_step": 106475, "epoch": 1196, "lr": 6.677015115329755e-05}
{"train_loss": 0.17799940705299377, "global_step": 106476, "epoch": 1196, "lr": 6.676960499696152e-05}
{"train_loss": 0.13077759742736816, "global_step": 106477, "epoch": 1196, "lr": 6.676905883837104e-05}
{"train_loss": 0.21522289514541626, "global_step": 106478, "epoch": 1196, "lr": 6.676851267752615e-05}
{"train_loss": 0.10847508162260056, "global_step": 106479, "epoch": 1196, "lr": 6.676796651442694e-05}
{"train_loss": 0.11254279315471649, "global_step": 106480, "epoch": 1196, "lr": 6.67674203490735e-05}
{"train_loss": 0.09254497289657593, "global_step": 106481, "epoch": 1196, "lr": 6.676687418146587e-05}
{"train_loss": 0.1352836936712265, "global_step": 106482, "epoch": 1196, "lr": 6.676632801160414e-05}
{"train_loss": 0.19689537584781647, "global_step": 106483, "epoch": 1196, "lr": 6.676578183948836e-05}
{"train_loss": 0.1824616640806198, "global_step": 106484, "epoch": 1196, "lr": 6.676523566511865e-05}
{"train_loss": 0.1281173974275589, "global_step": 106485, "epoch": 1196, "lr": 6.676468948849505e-05}
{"train_loss": 0.14159110188484192, "global_step": 106486, "epoch": 1196, "lr": 6.676414330961766e-05}
{"train_loss": 0.2605496346950531, "global_step": 106487, "epoch": 1196, "lr": 6.67635971284865e-05}
{"train_loss": 0.1327756941318512, "global_step": 106488, "epoch": 1196, "lr": 6.676305094510172e-05}
{"train_loss": 0.18653535842895508, "global_step": 106489, "epoch": 1196, "lr": 6.676250475946333e-05}
{"train_loss": 0.148195281624794, "global_step": 106490, "epoch": 1196, "lr": 6.676195857157142e-05}
{"train_loss": 0.18777114152908325, "global_step": 106491, "epoch": 1196, "lr": 6.676141238142608e-05}
{"train_loss": 0.2018076777458191, "global_step": 106492, "epoch": 1196, "lr": 6.676086618902737e-05}
{"train_loss": 0.1748219132423401, "global_step": 106493, "epoch": 1196, "lr": 6.676031999437536e-05}
{"train_loss": 0.13719843327999115, "global_step": 106494, "epoch": 1196, "lr": 6.675977379747015e-05}
{"train_loss": 0.14522215723991394, "global_step": 106495, "epoch": 1196, "lr": 6.675922759831177e-05}
{"train_loss": 0.1460336148738861, "global_step": 106496, "epoch": 1196, "lr": 6.675868139690033e-05}
{"train_loss": 0.13870827853679657, "global_step": 106497, "epoch": 1196, "lr": 6.675813519323588e-05}
{"train_loss": 0.1659766286611557, "global_step": 106498, "epoch": 1196, "lr": 6.67575889873185e-05}
{"train_loss": 0.1314496397972107, "global_step": 106499, "epoch": 1196, "lr": 6.675704277914828e-05}
{"train_loss": 0.14559583365917206, "global_step": 106500, "epoch": 1196, "lr": 6.675649656872526e-05}
{"train_loss": 0.3135392367839813, "global_step": 106501, "epoch": 1196, "lr": 6.675595035604953e-05}
{"train_loss": 0.0702836886048317, "global_step": 106502, "epoch": 1196, "lr": 6.67554041411212e-05}
{"train_loss": 0.07346024364233017, "global_step": 106503, "epoch": 1196, "lr": 6.675485792394028e-05}
{"train_loss": 0.10480968654155731, "global_step": 106504, "epoch": 1196, "lr": 6.675431170450688e-05}
{"train_loss": 0.20188668370246887, "global_step": 106505, "epoch": 1196, "lr": 6.675376548282108e-05}
{"train_loss": 0.1750333458185196, "global_step": 106506, "epoch": 1196, "lr": 6.675321925888294e-05}
{"train_loss": 0.1860010176897049, "global_step": 106507, "epoch": 1196, "lr": 6.675267303269253e-05}
{"train_loss": 0.16320200264453888, "global_step": 106508, "epoch": 1196, "lr": 6.675212680424992e-05}
{"train_loss": 0.12717674672603607, "global_step": 106509, "epoch": 1196, "lr": 6.675158057355519e-05}
{"train_loss": 0.1735822856426239, "global_step": 106510, "epoch": 1196, "lr": 6.675103434060841e-05}
{"train_loss": 0.19080641865730286, "global_step": 106511, "epoch": 1196, "lr": 6.675048810540967e-05}
{"train_loss": 0.07688345015048981, "global_step": 106512, "epoch": 1196, "lr": 6.674994186795903e-05}
{"train_loss": 0.19196844100952148, "global_step": 106513, "epoch": 1196, "lr": 6.674939562825658e-05}
{"train_loss": 0.11209326237440109, "global_step": 106514, "epoch": 1196, "lr": 6.674884938630235e-05}
{"train_loss": 0.16983380913734436, "global_step": 106515, "epoch": 1196, "lr": 6.674830314209646e-05}
{"train_loss": 0.11083997040987015, "global_step": 106516, "epoch": 1196, "lr": 6.674775689563894e-05}
{"train_loss": 0.1409820318222046, "global_step": 106517, "epoch": 1196, "lr": 6.674721064692992e-05}
{"train_loss": 0.15337738394737244, "global_step": 106518, "epoch": 1196, "lr": 6.674666439596942e-05}
{"train_loss": 0.08345497399568558, "global_step": 106519, "epoch": 1196, "lr": 6.674611814275753e-05}
{"train_loss": 0.14766649901866913, "global_step": 106520, "epoch": 1196, "lr": 6.674557188729433e-05}
{"train_loss": 0.14513997733592987, "global_step": 106521, "epoch": 1196, "lr": 6.67450256295799e-05}
{"train_loss": 0.17946657538414001, "global_step": 106522, "epoch": 1196, "lr": 6.674447936961431e-05}
{"train_loss": 0.18528537452220917, "global_step": 106523, "epoch": 1196, "lr": 6.674393310739763e-05}
{"train_loss": 0.19237707555294037, "global_step": 106524, "epoch": 1196, "lr": 6.674338684292992e-05}
{"train_loss": 0.15619023144245148, "global_step": 106525, "epoch": 1196, "lr": 6.674284057621128e-05}
{"train_loss": 0.1610899418592453, "global_step": 106526, "epoch": 1196, "lr": 6.674229430724176e-05}
{"train_loss": 0.0951557382941246, "global_step": 106527, "epoch": 1196, "lr": 6.674174803602145e-05}
{"train_loss": 0.12871551513671875, "global_step": 106528, "epoch": 1196, "lr": 6.674120176255041e-05}
{"train_loss": 0.1738475114107132, "global_step": 106529, "epoch": 1196, "lr": 6.674065548682872e-05}
{"train_loss": 0.13989955186843872, "global_step": 106530, "epoch": 1196, "lr": 6.674010920885645e-05}
{"train_loss": 0.09422944486141205, "global_step": 106531, "epoch": 1196, "lr": 6.673956292863368e-05}
{"train_loss": 0.15281075641010586, "global_step": 106532, "epoch": 1196, "lr": 6.67390166461605e-05, "val_loss": 4.840752601623535}
{"train_loss": 0.10916004329919815, "global_step": 106533, "epoch": 1197, "lr": 6.673847036143693e-05}
{"train_loss": 0.1890154927968979, "global_step": 106534, "epoch": 1197, "lr": 6.673792407446309e-05}
{"train_loss": 0.15136852860450745, "global_step": 106535, "epoch": 1197, "lr": 6.673737778523904e-05}
{"train_loss": 0.17479342222213745, "global_step": 106536, "epoch": 1197, "lr": 6.673683149376487e-05}
{"train_loss": 0.1457158327102661, "global_step": 106537, "epoch": 1197, "lr": 6.673628520004059e-05}
{"train_loss": 0.21541224420070648, "global_step": 106538, "epoch": 1197, "lr": 6.673573890406635e-05}
{"train_loss": 0.14280137419700623, "global_step": 106539, "epoch": 1197, "lr": 6.67351926058422e-05}
{"train_loss": 0.12785811722278595, "global_step": 106540, "epoch": 1197, "lr": 6.673464630536821e-05}
{"train_loss": 0.1025194600224495, "global_step": 106541, "epoch": 1197, "lr": 6.673410000264445e-05}
{"train_loss": 0.08724046498537064, "global_step": 106542, "epoch": 1197, "lr": 6.673355369767096e-05}
{"train_loss": 0.20470482110977173, "global_step": 106543, "epoch": 1197, "lr": 6.673300739044789e-05}
{"train_loss": 0.1422818899154663, "global_step": 106544, "epoch": 1197, "lr": 6.673246108097525e-05}
{"train_loss": 0.18371771275997162, "global_step": 106545, "epoch": 1197, "lr": 6.673191476925314e-05}
{"train_loss": 0.22044099867343903, "global_step": 106546, "epoch": 1197, "lr": 6.673136845528163e-05}
{"train_loss": 0.2169947475194931, "global_step": 106547, "epoch": 1197, "lr": 6.673082213906079e-05}
{"train_loss": 0.15957340598106384, "global_step": 106548, "epoch": 1197, "lr": 6.67302758205907e-05}
{"train_loss": 0.11091500520706177, "global_step": 106549, "epoch": 1197, "lr": 6.672972949987142e-05}
{"train_loss": 0.15605565905570984, "global_step": 106550, "epoch": 1197, "lr": 6.672918317690303e-05}
{"train_loss": 0.09599251300096512, "global_step": 106551, "epoch": 1197, "lr": 6.67286368516856e-05}
{"train_loss": 0.098957359790802, "global_step": 106552, "epoch": 1197, "lr": 6.672809052421924e-05}
{"train_loss": 0.1545964926481247, "global_step": 106553, "epoch": 1197, "lr": 6.672754419450398e-05}
{"train_loss": 0.0808483213186264, "global_step": 106554, "epoch": 1197, "lr": 6.672699786253988e-05}
{"train_loss": 0.1279703676700592, "global_step": 106555, "epoch": 1197, "lr": 6.672645152832705e-05}
{"train_loss": 0.12460971623659134, "global_step": 106556, "epoch": 1197, "lr": 6.672590519186555e-05}
{"train_loss": 0.20352961122989655, "global_step": 106557, "epoch": 1197, "lr": 6.672535885315548e-05}
{"train_loss": 0.12169577926397324, "global_step": 106558, "epoch": 1197, "lr": 6.672481251219687e-05}
{"train_loss": 0.19582019746303558, "global_step": 106559, "epoch": 1197, "lr": 6.672426616898981e-05}
{"train_loss": 0.16679733991622925, "global_step": 106560, "epoch": 1197, "lr": 6.672371982353437e-05}
{"train_loss": 0.14260412752628326, "global_step": 106561, "epoch": 1197, "lr": 6.672317347583066e-05}
{"train_loss": 0.1491663157939911, "global_step": 106562, "epoch": 1197, "lr": 6.67226271258787e-05}
{"train_loss": 0.06695915013551712, "global_step": 106563, "epoch": 1197, "lr": 6.672208077367859e-05}
{"train_loss": 0.14168761670589447, "global_step": 106564, "epoch": 1197, "lr": 6.67215344192304e-05}
{"train_loss": 0.27140581607818604, "global_step": 106565, "epoch": 1197, "lr": 6.67209880625342e-05}
{"train_loss": 0.2143850028514862, "global_step": 106566, "epoch": 1197, "lr": 6.672044170359007e-05}
{"train_loss": 0.1319809854030609, "global_step": 106567, "epoch": 1197, "lr": 6.671989534239809e-05}
{"train_loss": 0.14562982320785522, "global_step": 106568, "epoch": 1197, "lr": 6.671934897895832e-05}
{"train_loss": 0.19239987432956696, "global_step": 106569, "epoch": 1197, "lr": 6.671880261327084e-05}
{"train_loss": 0.23149099946022034, "global_step": 106570, "epoch": 1197, "lr": 6.671825624533571e-05}
{"train_loss": 0.08568248897790909, "global_step": 106571, "epoch": 1197, "lr": 6.671770987515302e-05}
{"train_loss": 0.194064199924469, "global_step": 106572, "epoch": 1197, "lr": 6.671716350272284e-05}
{"train_loss": 0.12899604439735413, "global_step": 106573, "epoch": 1197, "lr": 6.671661712804523e-05}
{"train_loss": 0.18124240636825562, "global_step": 106574, "epoch": 1197, "lr": 6.671607075112027e-05}
{"train_loss": 0.14036256074905396, "global_step": 106575, "epoch": 1197, "lr": 6.671552437194806e-05}
{"train_loss": 0.16132934391498566, "global_step": 106576, "epoch": 1197, "lr": 6.671497799052863e-05}
{"train_loss": 0.15622937679290771, "global_step": 106577, "epoch": 1197, "lr": 6.671443160686209e-05}
{"train_loss": 0.09597494453191757, "global_step": 106578, "epoch": 1197, "lr": 6.67138852209485e-05}
{"train_loss": 0.1499825417995453, "global_step": 106579, "epoch": 1197, "lr": 6.671333883278791e-05}
{"train_loss": 0.1224873885512352, "global_step": 106580, "epoch": 1197, "lr": 6.671279244238044e-05}
{"train_loss": 0.20123900473117828, "global_step": 106581, "epoch": 1197, "lr": 6.671224604972613e-05}
{"train_loss": 0.11127126216888428, "global_step": 106582, "epoch": 1197, "lr": 6.671169965482506e-05}
{"train_loss": 0.11196320503950119, "global_step": 106583, "epoch": 1197, "lr": 6.671115325767732e-05}
{"train_loss": 0.24587322771549225, "global_step": 106584, "epoch": 1197, "lr": 6.671060685828295e-05}
{"train_loss": 0.13434773683547974, "global_step": 106585, "epoch": 1197, "lr": 6.671006045664206e-05}
{"train_loss": 0.22980083525180817, "global_step": 106586, "epoch": 1197, "lr": 6.67095140527547e-05}
{"train_loss": 0.15772698819637299, "global_step": 106587, "epoch": 1197, "lr": 6.670896764662094e-05}
{"train_loss": 0.11745032668113708, "global_step": 106588, "epoch": 1197, "lr": 6.670842123824088e-05}
{"train_loss": 0.20807264745235443, "global_step": 106589, "epoch": 1197, "lr": 6.670787482761456e-05}
{"train_loss": 0.12394888699054718, "global_step": 106590, "epoch": 1197, "lr": 6.670732841474208e-05}
{"train_loss": 0.15088362991809845, "global_step": 106591, "epoch": 1197, "lr": 6.67067819996235e-05}
{"train_loss": 0.17829088866710663, "global_step": 106592, "epoch": 1197, "lr": 6.670623558225888e-05}
{"train_loss": 0.21463415026664734, "global_step": 106593, "epoch": 1197, "lr": 6.670568916264833e-05}
{"train_loss": 0.12945935130119324, "global_step": 106594, "epoch": 1197, "lr": 6.670514274079189e-05}
{"train_loss": 0.139824777841568, "global_step": 106595, "epoch": 1197, "lr": 6.670459631668966e-05}
{"train_loss": 0.1341351419687271, "global_step": 106596, "epoch": 1197, "lr": 6.670404989034169e-05}
{"train_loss": 0.18199831247329712, "global_step": 106597, "epoch": 1197, "lr": 6.670350346174807e-05}
{"train_loss": 0.09901437908411026, "global_step": 106598, "epoch": 1197, "lr": 6.670295703090888e-05}
{"train_loss": 0.08769211173057556, "global_step": 106599, "epoch": 1197, "lr": 6.670241059782416e-05}
{"train_loss": 0.1626850813627243, "global_step": 106600, "epoch": 1197, "lr": 6.670186416249402e-05}
{"train_loss": 0.14924073219299316, "global_step": 106601, "epoch": 1197, "lr": 6.670131772491852e-05}
{"train_loss": 0.1384510099887848, "global_step": 106602, "epoch": 1197, "lr": 6.670077128509771e-05}
{"train_loss": 0.16944830119609833, "global_step": 106603, "epoch": 1197, "lr": 6.67002248430317e-05}
{"train_loss": 0.11911799013614655, "global_step": 106604, "epoch": 1197, "lr": 6.669967839872056e-05}
{"train_loss": 0.17970021069049835, "global_step": 106605, "epoch": 1197, "lr": 6.669913195216433e-05}
{"train_loss": 0.1595636010169983, "global_step": 106606, "epoch": 1197, "lr": 6.669858550336311e-05}
{"train_loss": 0.15300320088863373, "global_step": 106607, "epoch": 1197, "lr": 6.669803905231698e-05}
{"train_loss": 0.15092521905899048, "global_step": 106608, "epoch": 1197, "lr": 6.669749259902598e-05}
{"train_loss": 0.09960605949163437, "global_step": 106609, "epoch": 1197, "lr": 6.669694614349023e-05}
{"train_loss": 0.10528478771448135, "global_step": 106610, "epoch": 1197, "lr": 6.669639968570977e-05}
{"train_loss": 0.19173184037208557, "global_step": 106611, "epoch": 1197, "lr": 6.669585322568468e-05}
{"train_loss": 0.09495433419942856, "global_step": 106612, "epoch": 1197, "lr": 6.669530676341503e-05}
{"train_loss": 0.2198353409767151, "global_step": 106613, "epoch": 1197, "lr": 6.669476029890091e-05}
{"train_loss": 0.12054067105054855, "global_step": 106614, "epoch": 1197, "lr": 6.669421383214238e-05}
{"train_loss": 0.162603497505188, "global_step": 106615, "epoch": 1197, "lr": 6.669366736313952e-05}
{"train_loss": 0.13184891641139984, "global_step": 106616, "epoch": 1197, "lr": 6.66931208918924e-05}
{"train_loss": 0.1672648787498474, "global_step": 106617, "epoch": 1197, "lr": 6.669257441840109e-05}
{"train_loss": 0.18075601756572723, "global_step": 106618, "epoch": 1197, "lr": 6.669202794266566e-05}
{"train_loss": 0.0941464900970459, "global_step": 106619, "epoch": 1197, "lr": 6.66914814646862e-05}
{"train_loss": 0.23782044649124146, "global_step": 106620, "epoch": 1197, "lr": 6.669093498446278e-05}
{"train_loss": 0.15303592216432765, "global_step": 106621, "epoch": 1197, "lr": 6.669038850199545e-05, "val_loss": 4.919066905975342}
{"train_loss": 0.17489154636859894, "global_step": 106622, "epoch": 1198, "lr": 6.66898420172843e-05}
{"train_loss": 0.16623634099960327, "global_step": 106623, "epoch": 1198, "lr": 6.668929553032943e-05}
{"train_loss": 0.16592027246952057, "global_step": 106624, "epoch": 1198, "lr": 6.668874904113086e-05}
{"train_loss": 0.1900240182876587, "global_step": 106625, "epoch": 1198, "lr": 6.66882025496887e-05}
{"train_loss": 0.15308436751365662, "global_step": 106626, "epoch": 1198, "lr": 6.668765605600301e-05}
{"train_loss": 0.14640836417675018, "global_step": 106627, "epoch": 1198, "lr": 6.668710956007388e-05}
{"train_loss": 0.2065459042787552, "global_step": 106628, "epoch": 1198, "lr": 6.668656306190136e-05}
{"train_loss": 0.14672476053237915, "global_step": 106629, "epoch": 1198, "lr": 6.668601656148555e-05}
{"train_loss": 0.16503043472766876, "global_step": 106630, "epoch": 1198, "lr": 6.668547005882649e-05}
{"train_loss": 0.20954962074756622, "global_step": 106631, "epoch": 1198, "lr": 6.668492355392428e-05}
{"train_loss": 0.13375656306743622, "global_step": 106632, "epoch": 1198, "lr": 6.668437704677899e-05}
{"train_loss": 0.17909403145313263, "global_step": 106633, "epoch": 1198, "lr": 6.668383053739069e-05}
{"train_loss": 0.19810737669467926, "global_step": 106634, "epoch": 1198, "lr": 6.668328402575944e-05}
{"train_loss": 0.11901892721652985, "global_step": 106635, "epoch": 1198, "lr": 6.668273751188534e-05}
{"train_loss": 0.17033296823501587, "global_step": 106636, "epoch": 1198, "lr": 6.668219099576843e-05}
{"train_loss": 0.22747983038425446, "global_step": 106637, "epoch": 1198, "lr": 6.668164447740883e-05}
{"train_loss": 0.16517947614192963, "global_step": 106638, "epoch": 1198, "lr": 6.668109795680657e-05}
{"train_loss": 0.19123399257659912, "global_step": 106639, "epoch": 1198, "lr": 6.668055143396173e-05}
{"train_loss": 0.13190990686416626, "global_step": 106640, "epoch": 1198, "lr": 6.66800049088744e-05}
{"train_loss": 0.21407851576805115, "global_step": 106641, "epoch": 1198, "lr": 6.667945838154465e-05}
{"train_loss": 0.1678600311279297, "global_step": 106642, "epoch": 1198, "lr": 6.667891185197257e-05}
{"train_loss": 0.129848912358284, "global_step": 106643, "epoch": 1198, "lr": 6.667836532015819e-05}
{"train_loss": 0.12425492703914642, "global_step": 106644, "epoch": 1198, "lr": 6.667781878610161e-05}
{"train_loss": 0.1799691617488861, "global_step": 106645, "epoch": 1198, "lr": 6.66772722498029e-05}
{"train_loss": 0.15424831211566925, "global_step": 106646, "epoch": 1198, "lr": 6.667672571126213e-05}
{"train_loss": 0.20006562769412994, "global_step": 106647, "epoch": 1198, "lr": 6.66761791704794e-05}
{"train_loss": 0.12705238163471222, "global_step": 106648, "epoch": 1198, "lr": 6.667563262745473e-05}
{"train_loss": 0.30069994926452637, "global_step": 106649, "epoch": 1198, "lr": 6.667508608218825e-05}
{"train_loss": 0.1857278198003769, "global_step": 106650, "epoch": 1198, "lr": 6.667453953467999e-05}
{"train_loss": 0.18150953948497772, "global_step": 106651, "epoch": 1198, "lr": 6.667399298493004e-05}
{"train_loss": 0.1527814120054245, "global_step": 106652, "epoch": 1198, "lr": 6.667344643293849e-05}
{"train_loss": 0.19782990217208862, "global_step": 106653, "epoch": 1198, "lr": 6.66728998787054e-05}
{"train_loss": 0.25719040632247925, "global_step": 106654, "epoch": 1198, "lr": 6.667235332223084e-05}
{"train_loss": 0.15307818353176117, "global_step": 106655, "epoch": 1198, "lr": 6.667180676351487e-05}
{"train_loss": 0.19011929631233215, "global_step": 106656, "epoch": 1198, "lr": 6.667126020255758e-05}
{"train_loss": 0.1905701756477356, "global_step": 106657, "epoch": 1198, "lr": 6.667071363935906e-05}
{"train_loss": 0.20330661535263062, "global_step": 106658, "epoch": 1198, "lr": 6.667016707391935e-05}
{"train_loss": 0.20060111582279205, "global_step": 106659, "epoch": 1198, "lr": 6.666962050623855e-05}
{"train_loss": 0.22351595759391785, "global_step": 106660, "epoch": 1198, "lr": 6.666907393631671e-05}
{"train_loss": 0.1541316956281662, "global_step": 106661, "epoch": 1198, "lr": 6.666852736415393e-05}
{"train_loss": 0.12686094641685486, "global_step": 106662, "epoch": 1198, "lr": 6.666798078975026e-05}
{"train_loss": 0.13392125070095062, "global_step": 106663, "epoch": 1198, "lr": 6.666743421310578e-05}
{"train_loss": 0.2079298198223114, "global_step": 106664, "epoch": 1198, "lr": 6.666688763422058e-05}
{"train_loss": 0.20557203888893127, "global_step": 106665, "epoch": 1198, "lr": 6.666634105309472e-05}
{"train_loss": 0.17018000781536102, "global_step": 106666, "epoch": 1198, "lr": 6.666579446972826e-05}
{"train_loss": 0.17074085772037506, "global_step": 106667, "epoch": 1198, "lr": 6.66652478841213e-05}
{"train_loss": 0.1258348673582077, "global_step": 106668, "epoch": 1198, "lr": 6.666470129627388e-05}
{"train_loss": 0.21755188703536987, "global_step": 106669, "epoch": 1198, "lr": 6.666415470618612e-05}
{"train_loss": 0.12903767824172974, "global_step": 106670, "epoch": 1198, "lr": 6.666360811385805e-05}
{"train_loss": 0.14708295464515686, "global_step": 106671, "epoch": 1198, "lr": 6.666306151928977e-05}
{"train_loss": 0.21122972667217255, "global_step": 106672, "epoch": 1198, "lr": 6.666251492248134e-05}
{"train_loss": 0.11445525288581848, "global_step": 106673, "epoch": 1198, "lr": 6.666196832343283e-05}
{"train_loss": 0.17217373847961426, "global_step": 106674, "epoch": 1198, "lr": 6.666142172214434e-05}
{"train_loss": 0.11884742975234985, "global_step": 106675, "epoch": 1198, "lr": 6.666087511861592e-05}
{"train_loss": 0.13326433300971985, "global_step": 106676, "epoch": 1198, "lr": 6.666032851284764e-05}
{"train_loss": 0.08960127085447311, "global_step": 106677, "epoch": 1198, "lr": 6.66597819048396e-05}
{"train_loss": 0.1940203756093979, "global_step": 106678, "epoch": 1198, "lr": 6.665923529459182e-05}
{"train_loss": 0.18659411370754242, "global_step": 106679, "epoch": 1198, "lr": 6.665868868210442e-05}
{"train_loss": 0.1693558543920517, "global_step": 106680, "epoch": 1198, "lr": 6.66581420673775e-05}
{"train_loss": 0.14969247579574585, "global_step": 106681, "epoch": 1198, "lr": 6.665759545041106e-05}
{"train_loss": 0.18904642760753632, "global_step": 106682, "epoch": 1198, "lr": 6.665704883120522e-05}
{"train_loss": 0.17716354131698608, "global_step": 106683, "epoch": 1198, "lr": 6.665650220976003e-05}
{"train_loss": 0.084071584045887, "global_step": 106684, "epoch": 1198, "lr": 6.665595558607559e-05}
{"train_loss": 0.1533139944076538, "global_step": 106685, "epoch": 1198, "lr": 6.665540896015196e-05}
{"train_loss": 0.11248756945133209, "global_step": 106686, "epoch": 1198, "lr": 6.665486233198921e-05}
{"train_loss": 0.05283348262310028, "global_step": 106687, "epoch": 1198, "lr": 6.665431570158742e-05}
{"train_loss": 0.13788928091526031, "global_step": 106688, "epoch": 1198, "lr": 6.665376906894664e-05}
{"train_loss": 0.13008970022201538, "global_step": 106689, "epoch": 1198, "lr": 6.6653222434067e-05}
{"train_loss": 0.17852404713630676, "global_step": 106690, "epoch": 1198, "lr": 6.665267579694851e-05}
{"train_loss": 0.17595668137073517, "global_step": 106691, "epoch": 1198, "lr": 6.665212915759128e-05}
{"train_loss": 0.1343902051448822, "global_step": 106692, "epoch": 1198, "lr": 6.665158251599537e-05}
{"train_loss": 0.16260921955108643, "global_step": 106693, "epoch": 1198, "lr": 6.665103587216086e-05}
{"train_loss": 0.13612240552902222, "global_step": 106694, "epoch": 1198, "lr": 6.665048922608783e-05}
{"train_loss": 0.15350043773651123, "global_step": 106695, "epoch": 1198, "lr": 6.664994257777632e-05}
{"train_loss": 0.12423966079950333, "global_step": 106696, "epoch": 1198, "lr": 6.664939592722646e-05}
{"train_loss": 0.14393073320388794, "global_step": 106697, "epoch": 1198, "lr": 6.664884927443828e-05}
{"train_loss": 0.27014344930648804, "global_step": 106698, "epoch": 1198, "lr": 6.664830261941185e-05}
{"train_loss": 0.13609667122364044, "global_step": 106699, "epoch": 1198, "lr": 6.664775596214728e-05}
{"train_loss": 0.17958909273147583, "global_step": 106700, "epoch": 1198, "lr": 6.664720930264461e-05}
{"train_loss": 0.13766640424728394, "global_step": 106701, "epoch": 1198, "lr": 6.664666264090392e-05}
{"train_loss": 0.18975484371185303, "global_step": 106702, "epoch": 1198, "lr": 6.66461159769253e-05}
{"train_loss": 0.12896333634853363, "global_step": 106703, "epoch": 1198, "lr": 6.664556931070882e-05}
{"train_loss": 0.1441783308982849, "global_step": 106704, "epoch": 1198, "lr": 6.664502264225452e-05}
{"train_loss": 0.20589780807495117, "global_step": 106705, "epoch": 1198, "lr": 6.664447597156252e-05}
{"train_loss": 0.13529916107654572, "global_step": 106706, "epoch": 1198, "lr": 6.664392929863287e-05}
{"train_loss": 0.17368744313716888, "global_step": 106707, "epoch": 1198, "lr": 6.664338262346565e-05}
{"train_loss": 0.12769241631031036, "global_step": 106708, "epoch": 1198, "lr": 6.664283594606091e-05}
{"train_loss": 0.16639344394207, "global_step": 106709, "epoch": 1198, "lr": 6.664228926641876e-05}
{"train_loss": 0.16575852413190884, "global_step": 106710, "epoch": 1198, "lr": 6.664174258453926e-05, "val_loss": 5.008525371551514}
{"train_loss": 0.14934512972831726, "global_step": 106711, "epoch": 1199, "lr": 6.664119590042247e-05}
{"train_loss": 0.08876830339431763, "global_step": 106712, "epoch": 1199, "lr": 6.664064921406849e-05}
{"train_loss": 0.1151028499007225, "global_step": 106713, "epoch": 1199, "lr": 6.664010252547736e-05}
{"train_loss": 0.2449679970741272, "global_step": 106714, "epoch": 1199, "lr": 6.663955583464918e-05}
{"train_loss": 0.15442916750907898, "global_step": 106715, "epoch": 1199, "lr": 6.663900914158401e-05}
{"train_loss": 0.07376661151647568, "global_step": 106716, "epoch": 1199, "lr": 6.663846244628193e-05}
{"train_loss": 0.1517772525548935, "global_step": 106717, "epoch": 1199, "lr": 6.663791574874302e-05}
{"train_loss": 0.16289161145687103, "global_step": 106718, "epoch": 1199, "lr": 6.663736904896734e-05}
{"train_loss": 0.12290655821561813, "global_step": 106719, "epoch": 1199, "lr": 6.663682234695495e-05}
{"train_loss": 0.17612260580062866, "global_step": 106720, "epoch": 1199, "lr": 6.663627564270597e-05}
{"train_loss": 0.17592503130435944, "global_step": 106721, "epoch": 1199, "lr": 6.663572893622043e-05}
{"train_loss": 0.15008963644504547, "global_step": 106722, "epoch": 1199, "lr": 6.663518222749843e-05}
{"train_loss": 0.20286384224891663, "global_step": 106723, "epoch": 1199, "lr": 6.663463551654003e-05}
{"train_loss": 0.18390390276908875, "global_step": 106724, "epoch": 1199, "lr": 6.663408880334529e-05}
{"train_loss": 0.15751239657402039, "global_step": 106725, "epoch": 1199, "lr": 6.663354208791431e-05}
{"train_loss": 0.1538389027118683, "global_step": 106726, "epoch": 1199, "lr": 6.663299537024716e-05}
{"train_loss": 0.17134425044059753, "global_step": 106727, "epoch": 1199, "lr": 6.66324486503439e-05}
{"train_loss": 0.08050665259361267, "global_step": 106728, "epoch": 1199, "lr": 6.663190192820461e-05}
{"train_loss": 0.1374567598104477, "global_step": 106729, "epoch": 1199, "lr": 6.663135520382936e-05}
{"train_loss": 0.12922357022762299, "global_step": 106730, "epoch": 1199, "lr": 6.663080847721823e-05}
{"train_loss": 0.12105809897184372, "global_step": 106731, "epoch": 1199, "lr": 6.663026174837129e-05}
{"train_loss": 0.15095657110214233, "global_step": 106732, "epoch": 1199, "lr": 6.662971501728861e-05}
{"train_loss": 0.22305098176002502, "global_step": 106733, "epoch": 1199, "lr": 6.662916828397028e-05}
{"train_loss": 0.18854093551635742, "global_step": 106734, "epoch": 1199, "lr": 6.662862154841635e-05}
{"train_loss": 0.30698567628860474, "global_step": 106735, "epoch": 1199, "lr": 6.66280748106269e-05}
{"train_loss": 0.13636073470115662, "global_step": 106736, "epoch": 1199, "lr": 6.662752807060201e-05}
{"train_loss": 0.21521617472171783, "global_step": 106737, "epoch": 1199, "lr": 6.662698132834177e-05}
{"train_loss": 0.17688782513141632, "global_step": 106738, "epoch": 1199, "lr": 6.662643458384621e-05}
{"train_loss": 0.127635657787323, "global_step": 106739, "epoch": 1199, "lr": 6.662588783711544e-05}
{"train_loss": 0.15088097751140594, "global_step": 106740, "epoch": 1199, "lr": 6.662534108814954e-05}
{"train_loss": 0.13592970371246338, "global_step": 106741, "epoch": 1199, "lr": 6.662479433694854e-05}
{"train_loss": 0.150831401348114, "global_step": 106742, "epoch": 1199, "lr": 6.662424758351256e-05}
{"train_loss": 0.11844214797019958, "global_step": 106743, "epoch": 1199, "lr": 6.662370082784164e-05}
{"train_loss": 0.17543886601924896, "global_step": 106744, "epoch": 1199, "lr": 6.662315406993586e-05}
{"train_loss": 0.1642468273639679, "global_step": 106745, "epoch": 1199, "lr": 6.662260730979531e-05}
{"train_loss": 0.19430594146251678, "global_step": 106746, "epoch": 1199, "lr": 6.662206054742004e-05}
{"train_loss": 0.2203707993030548, "global_step": 106747, "epoch": 1199, "lr": 6.662151378281016e-05}
{"train_loss": 0.12211048603057861, "global_step": 106748, "epoch": 1199, "lr": 6.66209670159657e-05}
{"train_loss": 0.26454147696495056, "global_step": 106749, "epoch": 1199, "lr": 6.662042024688675e-05}
{"train_loss": 0.19846925139427185, "global_step": 106750, "epoch": 1199, "lr": 6.66198734755734e-05}
{"train_loss": 0.1808229237794876, "global_step": 106751, "epoch": 1199, "lr": 6.661932670202571e-05}
{"train_loss": 0.17242884635925293, "global_step": 106752, "epoch": 1199, "lr": 6.661877992624375e-05}
{"train_loss": 0.15213242173194885, "global_step": 106753, "epoch": 1199, "lr": 6.661823314822759e-05}
{"train_loss": 0.12764610350131989, "global_step": 106754, "epoch": 1199, "lr": 6.661768636797732e-05}
{"train_loss": 0.1120726615190506, "global_step": 106755, "epoch": 1199, "lr": 6.6617139585493e-05}
{"train_loss": 0.10248901695013046, "global_step": 106756, "epoch": 1199, "lr": 6.661659280077471e-05}
{"train_loss": 0.24510598182678223, "global_step": 106757, "epoch": 1199, "lr": 6.661604601382251e-05}
{"train_loss": 0.18793264031410217, "global_step": 106758, "epoch": 1199, "lr": 6.661549922463651e-05}
{"train_loss": 0.14059118926525116, "global_step": 106759, "epoch": 1199, "lr": 6.661495243321676e-05}
{"train_loss": 0.19988799095153809, "global_step": 106760, "epoch": 1199, "lr": 6.661440563956332e-05}
{"train_loss": 0.1372002512216568, "global_step": 106761, "epoch": 1199, "lr": 6.661385884367628e-05}
{"train_loss": 0.2139425277709961, "global_step": 106762, "epoch": 1199, "lr": 6.66133120455557e-05}
{"train_loss": 0.18738973140716553, "global_step": 106763, "epoch": 1199, "lr": 6.661276524520167e-05}
{"train_loss": 0.18828125298023224, "global_step": 106764, "epoch": 1199, "lr": 6.661221844261425e-05}
{"train_loss": 0.2198510318994522, "global_step": 106765, "epoch": 1199, "lr": 6.661167163779352e-05}
{"train_loss": 0.1204661950469017, "global_step": 106766, "epoch": 1199, "lr": 6.661112483073955e-05}
{"train_loss": 0.17194344103336334, "global_step": 106767, "epoch": 1199, "lr": 6.661057802145243e-05}
{"train_loss": 0.14168325066566467, "global_step": 106768, "epoch": 1199, "lr": 6.661003120993221e-05}
{"train_loss": 0.22372938692569733, "global_step": 106769, "epoch": 1199, "lr": 6.660948439617898e-05}
{"train_loss": 0.13207708299160004, "global_step": 106770, "epoch": 1199, "lr": 6.660893758019281e-05}
{"train_loss": 0.1525699496269226, "global_step": 106771, "epoch": 1199, "lr": 6.660839076197376e-05}
{"train_loss": 0.17350412905216217, "global_step": 106772, "epoch": 1199, "lr": 6.660784394152193e-05}
{"train_loss": 0.1340295523405075, "global_step": 106773, "epoch": 1199, "lr": 6.660729711883735e-05}
{"train_loss": 0.18654273450374603, "global_step": 106774, "epoch": 1199, "lr": 6.660675029392014e-05}
{"train_loss": 0.24516931176185608, "global_step": 106775, "epoch": 1199, "lr": 6.660620346677036e-05}
{"train_loss": 0.1607472449541092, "global_step": 106776, "epoch": 1199, "lr": 6.660565663738808e-05}
{"train_loss": 0.16381344199180603, "global_step": 106777, "epoch": 1199, "lr": 6.660510980577335e-05}
{"train_loss": 0.12527552247047424, "global_step": 106778, "epoch": 1199, "lr": 6.66045629719263e-05}
{"train_loss": 0.08881174772977829, "global_step": 106779, "epoch": 1199, "lr": 6.660401613584693e-05}
{"train_loss": 0.16099533438682556, "global_step": 106780, "epoch": 1199, "lr": 6.660346929753537e-05}
{"train_loss": 0.13674606382846832, "global_step": 106781, "epoch": 1199, "lr": 6.660292245699168e-05}
{"train_loss": 0.12288675457239151, "global_step": 106782, "epoch": 1199, "lr": 6.660237561421593e-05}
{"train_loss": 0.19476644694805145, "global_step": 106783, "epoch": 1199, "lr": 6.660182876920819e-05}
{"train_loss": 0.15169550478458405, "global_step": 106784, "epoch": 1199, "lr": 6.660128192196853e-05}
{"train_loss": 0.12102887034416199, "global_step": 106785, "epoch": 1199, "lr": 6.660073507249703e-05}
{"train_loss": 0.1937231570482254, "global_step": 106786, "epoch": 1199, "lr": 6.660018822079377e-05}
{"train_loss": 0.13281551003456116, "global_step": 106787, "epoch": 1199, "lr": 6.659964136685882e-05}
{"train_loss": 0.11323773115873337, "global_step": 106788, "epoch": 1199, "lr": 6.659909451069224e-05}
{"train_loss": 0.21496619284152985, "global_step": 106789, "epoch": 1199, "lr": 6.659854765229412e-05}
{"train_loss": 0.12146470695734024, "global_step": 106790, "epoch": 1199, "lr": 6.659800079166453e-05}
{"train_loss": 0.14351683855056763, "global_step": 106791, "epoch": 1199, "lr": 6.659745392880353e-05}
{"train_loss": 0.19213294982910156, "global_step": 106792, "epoch": 1199, "lr": 6.659690706371121e-05}
{"train_loss": 0.1949034482240677, "global_step": 106793, "epoch": 1199, "lr": 6.659636019638765e-05}
{"train_loss": 0.16068781912326813, "global_step": 106794, "epoch": 1199, "lr": 6.65958133268329e-05}
{"train_loss": 0.18623086810112, "global_step": 106795, "epoch": 1199, "lr": 6.659526645504704e-05}
{"train_loss": 0.1492859572172165, "global_step": 106796, "epoch": 1199, "lr": 6.659471958103016e-05}
{"train_loss": 0.18916752934455872, "global_step": 106797, "epoch": 1199, "lr": 6.659417270478233e-05}
{"train_loss": 0.14684560894966125, "global_step": 106798, "epoch": 1199, "lr": 6.659362582630361e-05}
{"train_loss": 0.16378400633844098, "global_step": 106799, "epoch": 1199, "lr": 6.659307894559408e-05, "val_loss": 4.8603010177612305}
{"train_loss": 0.24659927189350128, "global_step": 106800, "epoch": 1200, "lr": 6.65925320626538e-05}
{"train_loss": 0.19673755764961243, "global_step": 106801, "epoch": 1200, "lr": 6.659198517748286e-05}
{"train_loss": 0.11622575670480728, "global_step": 106802, "epoch": 1200, "lr": 6.659143829008134e-05}
{"train_loss": 0.08721223473548889, "global_step": 106803, "epoch": 1200, "lr": 6.65908914004493e-05}
{"train_loss": 0.1789420247077942, "global_step": 106804, "epoch": 1200, "lr": 6.659034450858682e-05}
{"train_loss": 0.16150110960006714, "global_step": 106805, "epoch": 1200, "lr": 6.658979761449395e-05}
{"train_loss": 0.10303859412670135, "global_step": 106806, "epoch": 1200, "lr": 6.658925071817079e-05}
{"train_loss": 0.15623314678668976, "global_step": 106807, "epoch": 1200, "lr": 6.658870381961742e-05}
{"train_loss": 0.18026840686798096, "global_step": 106808, "epoch": 1200, "lr": 6.65881569188339e-05}
{"train_loss": 0.16656461358070374, "global_step": 106809, "epoch": 1200, "lr": 6.658761001582029e-05}
{"train_loss": 0.14659219980239868, "global_step": 106810, "epoch": 1200, "lr": 6.658706311057669e-05}
{"train_loss": 0.27744996547698975, "global_step": 106811, "epoch": 1200, "lr": 6.658651620310316e-05}
{"train_loss": 0.1256324052810669, "global_step": 106812, "epoch": 1200, "lr": 6.658596929339978e-05}
{"train_loss": 0.17287129163742065, "global_step": 106813, "epoch": 1200, "lr": 6.658542238146661e-05}
{"train_loss": 0.11098182201385498, "global_step": 106814, "epoch": 1200, "lr": 6.658487546730372e-05}
{"train_loss": 0.13165518641471863, "global_step": 106815, "epoch": 1200, "lr": 6.658432855091122e-05}
{"train_loss": 0.13052234053611755, "global_step": 106816, "epoch": 1200, "lr": 6.658378163228915e-05}
{"train_loss": 0.31329023838043213, "global_step": 106817, "epoch": 1200, "lr": 6.658323471143762e-05}
{"train_loss": 0.16937211155891418, "global_step": 106818, "epoch": 1200, "lr": 6.658268778835663e-05}
{"train_loss": 0.19443964958190918, "global_step": 106819, "epoch": 1200, "lr": 6.658214086304633e-05}
{"train_loss": 0.10600204020738602, "global_step": 106820, "epoch": 1200, "lr": 6.658159393550675e-05}
{"train_loss": 0.1573420763015747, "global_step": 106821, "epoch": 1200, "lr": 6.658104700573797e-05}
{"train_loss": 0.14264653623104095, "global_step": 106822, "epoch": 1200, "lr": 6.658050007374008e-05}
{"train_loss": 0.2990037500858307, "global_step": 106823, "epoch": 1200, "lr": 6.657995313951316e-05}
{"train_loss": 0.15608452260494232, "global_step": 106824, "epoch": 1200, "lr": 6.657940620305725e-05}
{"train_loss": 0.21856677532196045, "global_step": 106825, "epoch": 1200, "lr": 6.657885926437244e-05}
{"train_loss": 0.16814933717250824, "global_step": 106826, "epoch": 1200, "lr": 6.65783123234588e-05}
{"train_loss": 0.19353355467319489, "global_step": 106827, "epoch": 1200, "lr": 6.657776538031641e-05}
{"train_loss": 0.20808719098567963, "global_step": 106828, "epoch": 1200, "lr": 6.657721843494534e-05}
{"train_loss": 0.12875652313232422, "global_step": 106829, "epoch": 1200, "lr": 6.657667148734567e-05}
{"train_loss": 0.19862808287143707, "global_step": 106830, "epoch": 1200, "lr": 6.657612453751747e-05}
{"train_loss": 0.150481179356575, "global_step": 106831, "epoch": 1200, "lr": 6.657557758546082e-05}
{"train_loss": 0.19312219321727753, "global_step": 106832, "epoch": 1200, "lr": 6.657503063117577e-05}
{"train_loss": 0.22632437944412231, "global_step": 106833, "epoch": 1200, "lr": 6.657448367466242e-05}
{"train_loss": 0.17101964354515076, "global_step": 106834, "epoch": 1200, "lr": 6.657393671592083e-05}
{"train_loss": 0.21683043241500854, "global_step": 106835, "epoch": 1200, "lr": 6.657338975495107e-05}
{"train_loss": 0.15783458948135376, "global_step": 106836, "epoch": 1200, "lr": 6.657284279175323e-05}
{"train_loss": 0.19778290390968323, "global_step": 106837, "epoch": 1200, "lr": 6.657229582632737e-05}
{"train_loss": 0.208047553896904, "global_step": 106838, "epoch": 1200, "lr": 6.657174885867357e-05}
{"train_loss": 0.23541994392871857, "global_step": 106839, "epoch": 1200, "lr": 6.65712018887919e-05}
{"train_loss": 0.1766665279865265, "global_step": 106840, "epoch": 1200, "lr": 6.657065491668242e-05}
{"train_loss": 0.19372814893722534, "global_step": 106841, "epoch": 1200, "lr": 6.657010794234523e-05}
{"train_loss": 0.12816843390464783, "global_step": 106842, "epoch": 1200, "lr": 6.656956096578038e-05}
{"train_loss": 0.12914219498634338, "global_step": 106843, "epoch": 1200, "lr": 6.656901398698797e-05}
{"train_loss": 0.0941624566912651, "global_step": 106844, "epoch": 1200, "lr": 6.656846700596805e-05}
{"train_loss": 0.09849182516336441, "global_step": 106845, "epoch": 1200, "lr": 6.65679200227207e-05}
{"train_loss": 0.15249687433242798, "global_step": 106846, "epoch": 1200, "lr": 6.656737303724598e-05}
{"train_loss": 0.14470231533050537, "global_step": 106847, "epoch": 1200, "lr": 6.6566826049544e-05}
{"train_loss": 0.13996797800064087, "global_step": 106848, "epoch": 1200, "lr": 6.656627905961481e-05}
{"train_loss": 0.22864994406700134, "global_step": 106849, "epoch": 1200, "lr": 6.656573206745849e-05}
{"train_loss": 0.1569916009902954, "global_step": 106850, "epoch": 1200, "lr": 6.65651850730751e-05}
{"train_loss": 0.186712384223938, "global_step": 106851, "epoch": 1200, "lr": 6.656463807646472e-05}
{"train_loss": 0.1232808381319046, "global_step": 106852, "epoch": 1200, "lr": 6.656409107762744e-05}
{"train_loss": 0.20241974294185638, "global_step": 106853, "epoch": 1200, "lr": 6.65635440765633e-05}
{"train_loss": 0.19395659863948822, "global_step": 106854, "epoch": 1200, "lr": 6.656299707327241e-05}
{"train_loss": 0.17222817242145538, "global_step": 106855, "epoch": 1200, "lr": 6.656245006775481e-05}
{"train_loss": 0.09283729642629623, "global_step": 106856, "epoch": 1200, "lr": 6.656190306001062e-05}
{"train_loss": 0.19649255275726318, "global_step": 106857, "epoch": 1200, "lr": 6.656135605003986e-05}
{"train_loss": 0.12798874080181122, "global_step": 106858, "epoch": 1200, "lr": 6.656080903784265e-05}
{"train_loss": 0.15574350953102112, "global_step": 106859, "epoch": 1200, "lr": 6.656026202341901e-05}
{"train_loss": 0.15356093645095825, "global_step": 106860, "epoch": 1200, "lr": 6.655971500676907e-05}
{"train_loss": 0.141424298286438, "global_step": 106861, "epoch": 1200, "lr": 6.655916798789285e-05}
{"train_loss": 0.15748903155326843, "global_step": 106862, "epoch": 1200, "lr": 6.655862096679047e-05}
{"train_loss": 0.1386549025774002, "global_step": 106863, "epoch": 1200, "lr": 6.6558073943462e-05}
{"train_loss": 0.13649824261665344, "global_step": 106864, "epoch": 1200, "lr": 6.655752691790747e-05}
{"train_loss": 0.17456823587417603, "global_step": 106865, "epoch": 1200, "lr": 6.6556979890127e-05}
{"train_loss": 0.0991504117846489, "global_step": 106866, "epoch": 1200, "lr": 6.655643286012064e-05}
{"train_loss": 0.16082733869552612, "global_step": 106867, "epoch": 1200, "lr": 6.655588582788846e-05}
{"train_loss": 0.21031980216503143, "global_step": 106868, "epoch": 1200, "lr": 6.655533879343055e-05}
{"train_loss": 0.2031428962945938, "global_step": 106869, "epoch": 1200, "lr": 6.6554791756747e-05}
{"train_loss": 0.18861618638038635, "global_step": 106870, "epoch": 1200, "lr": 6.655424471783782e-05}
{"train_loss": 0.127985417842865, "global_step": 106871, "epoch": 1200, "lr": 6.655369767670315e-05}
{"train_loss": 0.15910595655441284, "global_step": 106872, "epoch": 1200, "lr": 6.655315063334304e-05}
{"train_loss": 0.13582411408424377, "global_step": 106873, "epoch": 1200, "lr": 6.655260358775754e-05}
{"train_loss": 0.17712457478046417, "global_step": 106874, "epoch": 1200, "lr": 6.655205653994677e-05}
{"train_loss": 0.13235537707805634, "global_step": 106875, "epoch": 1200, "lr": 6.655150948991076e-05}
{"train_loss": 0.16158920526504517, "global_step": 106876, "epoch": 1200, "lr": 6.65509624376496e-05}
{"train_loss": 0.12518702447414398, "global_step": 106877, "epoch": 1200, "lr": 6.655041538316338e-05}
{"train_loss": 0.12655141949653625, "global_step": 106878, "epoch": 1200, "lr": 6.654986832645215e-05}
{"train_loss": 0.17793487012386322, "global_step": 106879, "epoch": 1200, "lr": 6.654932126751598e-05}
{"train_loss": 0.14850254356861115, "global_step": 106880, "epoch": 1200, "lr": 6.654877420635496e-05}
{"train_loss": 0.1408398449420929, "global_step": 106881, "epoch": 1200, "lr": 6.654822714296917e-05}
{"train_loss": 0.16224916279315948, "global_step": 106882, "epoch": 1200, "lr": 6.654768007735866e-05}
{"train_loss": 0.11588097363710403, "global_step": 106883, "epoch": 1200, "lr": 6.654713300952352e-05}
{"train_loss": 0.12477338314056396, "global_step": 106884, "epoch": 1200, "lr": 6.654658593946383e-05}
{"train_loss": 0.1690458506345749, "global_step": 106885, "epoch": 1200, "lr": 6.654603886717964e-05}
{"train_loss": 0.17222651839256287, "global_step": 106886, "epoch": 1200, "lr": 6.654549179267105e-05}
{"train_loss": 0.1552617996931076, "global_step": 106887, "epoch": 1200, "lr": 6.65449447159381e-05}
{"train_loss": 0.16508122483331167, "global_step": 106888, "epoch": 1200, "lr": 6.654439763698091e-05, "train/sim_max_reward_0": 0.49241375881335636, "train/sim_max_reward_1": 0.7810346468292322, "train/sim_max_reward_2": 0.30204814592604745, "train/sim_max_reward_3": 0.8754542982002294, "train/sim_max_reward_4": 0.9811834570306789, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.9464491736128404, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.38446644946456343, "test/sim_max_reward_4500004": 0.6724738468287768, "test/sim_max_reward_4500005": 0.991438601992475, "test/sim_max_reward_4500006": 0.9283186361114951, "test/sim_max_reward_4500007": 0.8621834199059103, "test/sim_max_reward_4500008": 0.18784107618070683, "test/sim_max_reward_4500009": 0.050501348474724025, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.23155985366144793, "test/sim_max_reward_4500014": 0.9862609625258407, "test/sim_max_reward_4500015": 0.9341836070013568, "test/sim_max_reward_4500016": 0.57209999124072, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.925306370995424, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9948365810140457, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6226134730882444, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9485756797812626, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6733914628697505, "test/sim_max_reward_4500029": 0.7245047305441162, "test/sim_max_reward_4500030": 0.9709873418693291, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.7101497642849915, "test/sim_max_reward_4500033": 0.8783053606274441, "test/sim_max_reward_4500034": 0.49523879387028463, "test/sim_max_reward_4500035": 0.4456490101834448, "test/sim_max_reward_4500036": 0.9988092635482466, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.9820623436830634, "test/sim_max_reward_4500039": 0.989418886654348, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6414668070219058, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.509791778359555, "test/sim_max_reward_4500044": 0.6163246041641022, "test/sim_max_reward_4500045": 0.6873486937742728, "test/sim_max_reward_4500046": 0.5074150236433242, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7386890511332572, "test/mean_score": 0.6413994587395604, "val_loss": 4.861605644226074, "train_action_mse_error": 42.874027252197266}
{"train_loss": 0.10270606726408005, "global_step": 106889, "epoch": 1201, "lr": 6.65438505557995e-05}
{"train_loss": 0.16443775594234467, "global_step": 106890, "epoch": 1201, "lr": 6.654330347239399e-05}
{"train_loss": 0.21358908712863922, "global_step": 106891, "epoch": 1201, "lr": 6.654275638676442e-05}
{"train_loss": 0.13924551010131836, "global_step": 106892, "epoch": 1201, "lr": 6.65422092989109e-05}
{"train_loss": 0.13676734268665314, "global_step": 106893, "epoch": 1201, "lr": 6.654166220883346e-05}
{"train_loss": 0.12491638213396072, "global_step": 106894, "epoch": 1201, "lr": 6.65411151165322e-05}
{"train_loss": 0.20747573673725128, "global_step": 106895, "epoch": 1201, "lr": 6.654056802200721e-05}
{"train_loss": 0.17433606088161469, "global_step": 106896, "epoch": 1201, "lr": 6.654002092525852e-05}
{"train_loss": 0.13499219715595245, "global_step": 106897, "epoch": 1201, "lr": 6.653947382628624e-05}
{"train_loss": 0.0836736336350441, "global_step": 106898, "epoch": 1201, "lr": 6.653892672509042e-05}
{"train_loss": 0.1607828289270401, "global_step": 106899, "epoch": 1201, "lr": 6.653837962167114e-05}
{"train_loss": 0.1560411900281906, "global_step": 106900, "epoch": 1201, "lr": 6.653783251602848e-05}
{"train_loss": 0.12956734001636505, "global_step": 106901, "epoch": 1201, "lr": 6.65372854081625e-05}
{"train_loss": 0.13567030429840088, "global_step": 106902, "epoch": 1201, "lr": 6.65367382980733e-05}
{"train_loss": 0.15825814008712769, "global_step": 106903, "epoch": 1201, "lr": 6.653619118576092e-05}
{"train_loss": 0.12856315076351166, "global_step": 106904, "epoch": 1201, "lr": 6.653564407122547e-05}
{"train_loss": 0.1567583978176117, "global_step": 106905, "epoch": 1201, "lr": 6.6535096954467e-05}
{"train_loss": 0.1696699559688568, "global_step": 106906, "epoch": 1201, "lr": 6.653454983548557e-05}
{"train_loss": 0.12287446111440659, "global_step": 106907, "epoch": 1201, "lr": 6.653400271428128e-05}
{"train_loss": 0.14843183755874634, "global_step": 106908, "epoch": 1201, "lr": 6.65334555908542e-05}
{"train_loss": 0.14221206307411194, "global_step": 106909, "epoch": 1201, "lr": 6.65329084652044e-05}
{"train_loss": 0.19860945641994476, "global_step": 106910, "epoch": 1201, "lr": 6.653236133733196e-05}
{"train_loss": 0.1570885181427002, "global_step": 106911, "epoch": 1201, "lr": 6.653181420723693e-05}
{"train_loss": 0.1109568327665329, "global_step": 106912, "epoch": 1201, "lr": 6.653126707491941e-05}
{"train_loss": 0.22198829054832458, "global_step": 106913, "epoch": 1201, "lr": 6.653071994037945e-05}
{"train_loss": 0.15467794239521027, "global_step": 106914, "epoch": 1201, "lr": 6.653017280361715e-05}
{"train_loss": 0.13576725125312805, "global_step": 106915, "epoch": 1201, "lr": 6.652962566463257e-05}
{"train_loss": 0.11820138990879059, "global_step": 106916, "epoch": 1201, "lr": 6.652907852342577e-05}
{"train_loss": 0.12916652858257294, "global_step": 106917, "epoch": 1201, "lr": 6.652853137999685e-05}
{"train_loss": 0.13204041123390198, "global_step": 106918, "epoch": 1201, "lr": 6.652798423434585e-05}
{"train_loss": 0.13163648545742035, "global_step": 106919, "epoch": 1201, "lr": 6.652743708647287e-05}
{"train_loss": 0.2075567990541458, "global_step": 106920, "epoch": 1201, "lr": 6.652688993637799e-05}
{"train_loss": 0.15160396695137024, "global_step": 106921, "epoch": 1201, "lr": 6.652634278406126e-05}
{"train_loss": 0.14008720219135284, "global_step": 106922, "epoch": 1201, "lr": 6.652579562952277e-05}
{"train_loss": 0.09184743463993073, "global_step": 106923, "epoch": 1201, "lr": 6.652524847276259e-05}
{"train_loss": 0.13580811023712158, "global_step": 106924, "epoch": 1201, "lr": 6.652470131378079e-05}
{"train_loss": 0.09765397757291794, "global_step": 106925, "epoch": 1201, "lr": 6.652415415257745e-05}
{"train_loss": 0.21157662570476532, "global_step": 106926, "epoch": 1201, "lr": 6.652360698915263e-05}
{"train_loss": 0.16500289738178253, "global_step": 106927, "epoch": 1201, "lr": 6.652305982350642e-05}
{"train_loss": 0.15653696656227112, "global_step": 106928, "epoch": 1201, "lr": 6.652251265563889e-05}
{"train_loss": 0.15461114048957825, "global_step": 106929, "epoch": 1201, "lr": 6.652196548555009e-05}
{"train_loss": 0.12538373470306396, "global_step": 106930, "epoch": 1201, "lr": 6.652141831324012e-05}
{"train_loss": 0.13981445133686066, "global_step": 106931, "epoch": 1201, "lr": 6.652087113870907e-05}
{"train_loss": 0.18817870318889618, "global_step": 106932, "epoch": 1201, "lr": 6.652032396195696e-05}
{"train_loss": 0.12491018325090408, "global_step": 106933, "epoch": 1201, "lr": 6.651977678298391e-05}
{"train_loss": 0.11534339934587479, "global_step": 106934, "epoch": 1201, "lr": 6.651922960178998e-05}
{"train_loss": 0.13167054951190948, "global_step": 106935, "epoch": 1201, "lr": 6.651868241837523e-05}
{"train_loss": 0.15488474071025848, "global_step": 106936, "epoch": 1201, "lr": 6.651813523273974e-05}
{"train_loss": 0.11477982997894287, "global_step": 106937, "epoch": 1201, "lr": 6.65175880448836e-05}
{"train_loss": 0.2514318525791168, "global_step": 106938, "epoch": 1201, "lr": 6.651704085480687e-05}
{"train_loss": 0.20287607610225677, "global_step": 106939, "epoch": 1201, "lr": 6.651649366250963e-05}
{"train_loss": 0.11783626675605774, "global_step": 106940, "epoch": 1201, "lr": 6.651594646799193e-05}
{"train_loss": 0.2162652313709259, "global_step": 106941, "epoch": 1201, "lr": 6.651539927125387e-05}
{"train_loss": 0.10432064533233643, "global_step": 106942, "epoch": 1201, "lr": 6.651485207229553e-05}
{"train_loss": 0.1304001361131668, "global_step": 106943, "epoch": 1201, "lr": 6.651430487111697e-05}
{"train_loss": 0.16452965140342712, "global_step": 106944, "epoch": 1201, "lr": 6.651375766771824e-05}
{"train_loss": 0.17470796406269073, "global_step": 106945, "epoch": 1201, "lr": 6.651321046209945e-05}
{"train_loss": 0.052881740033626556, "global_step": 106946, "epoch": 1201, "lr": 6.651266325426067e-05}
{"train_loss": 0.17443743348121643, "global_step": 106947, "epoch": 1201, "lr": 6.651211604420195e-05}
{"train_loss": 0.14284923672676086, "global_step": 106948, "epoch": 1201, "lr": 6.651156883192338e-05}
{"train_loss": 0.13466361165046692, "global_step": 106949, "epoch": 1201, "lr": 6.651102161742503e-05}
{"train_loss": 0.14419583976268768, "global_step": 106950, "epoch": 1201, "lr": 6.651047440070696e-05}
{"train_loss": 0.17710909247398376, "global_step": 106951, "epoch": 1201, "lr": 6.650992718176928e-05}
{"train_loss": 0.1650284379720688, "global_step": 106952, "epoch": 1201, "lr": 6.650937996061204e-05}
{"train_loss": 0.23877212405204773, "global_step": 106953, "epoch": 1201, "lr": 6.650883273723532e-05}
{"train_loss": 0.08900463581085205, "global_step": 106954, "epoch": 1201, "lr": 6.650828551163917e-05}
{"train_loss": 0.15357127785682678, "global_step": 106955, "epoch": 1201, "lr": 6.650773828382369e-05}
{"train_loss": 0.16912271082401276, "global_step": 106956, "epoch": 1201, "lr": 6.650719105378895e-05}
{"train_loss": 0.17852292954921722, "global_step": 106957, "epoch": 1201, "lr": 6.650664382153501e-05}
{"train_loss": 0.09285793453454971, "global_step": 106958, "epoch": 1201, "lr": 6.650609658706196e-05}
{"train_loss": 0.1728336364030838, "global_step": 106959, "epoch": 1201, "lr": 6.650554935036987e-05}
{"train_loss": 0.16286486387252808, "global_step": 106960, "epoch": 1201, "lr": 6.650500211145881e-05}
{"train_loss": 0.1567104160785675, "global_step": 106961, "epoch": 1201, "lr": 6.650445487032885e-05}
{"train_loss": 0.15707755088806152, "global_step": 106962, "epoch": 1201, "lr": 6.650390762698007e-05}
{"train_loss": 0.17111290991306305, "global_step": 106963, "epoch": 1201, "lr": 6.650336038141252e-05}
{"train_loss": 0.16564342379570007, "global_step": 106964, "epoch": 1201, "lr": 6.650281313362632e-05}
{"train_loss": 0.16422387957572937, "global_step": 106965, "epoch": 1201, "lr": 6.650226588362151e-05}
{"train_loss": 0.17986662685871124, "global_step": 106966, "epoch": 1201, "lr": 6.650171863139818e-05}
{"train_loss": 0.1466158926486969, "global_step": 106967, "epoch": 1201, "lr": 6.650117137695637e-05}
{"train_loss": 0.17058750987052917, "global_step": 106968, "epoch": 1201, "lr": 6.65006241202962e-05}
{"train_loss": 0.17174844443798065, "global_step": 106969, "epoch": 1201, "lr": 6.650007686141772e-05}
{"train_loss": 0.16265209019184113, "global_step": 106970, "epoch": 1201, "lr": 6.6499529600321e-05}
{"train_loss": 0.1235203891992569, "global_step": 106971, "epoch": 1201, "lr": 6.649898233700613e-05}
{"train_loss": 0.17060048878192902, "global_step": 106972, "epoch": 1201, "lr": 6.649843507147318e-05}
{"train_loss": 0.14717282354831696, "global_step": 106973, "epoch": 1201, "lr": 6.649788780372218e-05}
{"train_loss": 0.11048783361911774, "global_step": 106974, "epoch": 1201, "lr": 6.649734053375327e-05}
{"train_loss": 0.14199329912662506, "global_step": 106975, "epoch": 1201, "lr": 6.649679326156648e-05}
{"train_loss": 0.18479947745800018, "global_step": 106976, "epoch": 1201, "lr": 6.649624598716191e-05}
{"train_loss": 0.1511465614933646, "global_step": 106977, "epoch": 1201, "lr": 6.64956987105396e-05, "val_loss": 4.814877033233643}
{"train_loss": 0.15514229238033295, "global_step": 106978, "epoch": 1202, "lr": 6.649515143169966e-05}
{"train_loss": 0.13181394338607788, "global_step": 106979, "epoch": 1202, "lr": 6.649460415064214e-05}
{"train_loss": 0.1681618094444275, "global_step": 106980, "epoch": 1202, "lr": 6.649405686736714e-05}
{"train_loss": 0.11427424848079681, "global_step": 106981, "epoch": 1202, "lr": 6.64935095818747e-05}
{"train_loss": 0.2362796664237976, "global_step": 106982, "epoch": 1202, "lr": 6.649296229416491e-05}
{"train_loss": 0.25445231795310974, "global_step": 106983, "epoch": 1202, "lr": 6.649241500423784e-05}
{"train_loss": 0.20947618782520294, "global_step": 106984, "epoch": 1202, "lr": 6.649186771209357e-05}
{"train_loss": 0.1925734430551529, "global_step": 106985, "epoch": 1202, "lr": 6.649132041773217e-05}
{"train_loss": 0.12996384501457214, "global_step": 106986, "epoch": 1202, "lr": 6.649077312115372e-05}
{"train_loss": 0.17893464863300323, "global_step": 106987, "epoch": 1202, "lr": 6.649022582235827e-05}
{"train_loss": 0.15326477587223053, "global_step": 106988, "epoch": 1202, "lr": 6.648967852134593e-05}
{"train_loss": 0.1519973874092102, "global_step": 106989, "epoch": 1202, "lr": 6.648913121811673e-05}
{"train_loss": 0.19949811697006226, "global_step": 106990, "epoch": 1202, "lr": 6.648858391267078e-05}
{"train_loss": 0.16669046878814697, "global_step": 106991, "epoch": 1202, "lr": 6.648803660500815e-05}
{"train_loss": 0.10888199508190155, "global_step": 106992, "epoch": 1202, "lr": 6.648748929512888e-05}
{"train_loss": 0.2197311967611313, "global_step": 106993, "epoch": 1202, "lr": 6.64869419830331e-05}
{"train_loss": 0.14592371881008148, "global_step": 106994, "epoch": 1202, "lr": 6.648639466872081e-05}
{"train_loss": 0.12553495168685913, "global_step": 106995, "epoch": 1202, "lr": 6.648584735219215e-05}
{"train_loss": 0.11046890914440155, "global_step": 106996, "epoch": 1202, "lr": 6.648530003344717e-05}
{"train_loss": 0.13556019961833954, "global_step": 106997, "epoch": 1202, "lr": 6.648475271248594e-05}
{"train_loss": 0.12701217830181122, "global_step": 106998, "epoch": 1202, "lr": 6.648420538930854e-05}
{"train_loss": 0.18086645007133484, "global_step": 106999, "epoch": 1202, "lr": 6.648365806391504e-05}
{"train_loss": 0.10071417689323425, "global_step": 107000, "epoch": 1202, "lr": 6.648311073630551e-05}
{"train_loss": 0.19704362750053406, "global_step": 107001, "epoch": 1202, "lr": 6.648256340648003e-05}
{"train_loss": 0.1537018120288849, "global_step": 107002, "epoch": 1202, "lr": 6.648201607443866e-05}
{"train_loss": 0.09944108128547668, "global_step": 107003, "epoch": 1202, "lr": 6.648146874018148e-05}
{"train_loss": 0.19048379361629486, "global_step": 107004, "epoch": 1202, "lr": 6.648092140370858e-05}
{"train_loss": 0.16011613607406616, "global_step": 107005, "epoch": 1202, "lr": 6.648037406502002e-05}
{"train_loss": 0.14137214422225952, "global_step": 107006, "epoch": 1202, "lr": 6.647982672411588e-05}
{"train_loss": 0.25621747970581055, "global_step": 107007, "epoch": 1202, "lr": 6.647927938099621e-05}
{"train_loss": 0.17298467457294464, "global_step": 107008, "epoch": 1202, "lr": 6.647873203566112e-05}
{"train_loss": 0.17785713076591492, "global_step": 107009, "epoch": 1202, "lr": 6.647818468811065e-05}
{"train_loss": 0.10706809908151627, "global_step": 107010, "epoch": 1202, "lr": 6.64776373383449e-05}
{"train_loss": 0.08180505037307739, "global_step": 107011, "epoch": 1202, "lr": 6.647708998636392e-05}
{"train_loss": 0.18395401537418365, "global_step": 107012, "epoch": 1202, "lr": 6.647654263216782e-05}
{"train_loss": 0.10425533354282379, "global_step": 107013, "epoch": 1202, "lr": 6.647599527575663e-05}
{"train_loss": 0.1662106066942215, "global_step": 107014, "epoch": 1202, "lr": 6.647544791713045e-05}
{"train_loss": 0.05312997102737427, "global_step": 107015, "epoch": 1202, "lr": 6.647490055628934e-05}
{"train_loss": 0.16480188071727753, "global_step": 107016, "epoch": 1202, "lr": 6.647435319323338e-05}
{"train_loss": 0.19855213165283203, "global_step": 107017, "epoch": 1202, "lr": 6.647380582796265e-05}
{"train_loss": 0.14408928155899048, "global_step": 107018, "epoch": 1202, "lr": 6.647325846047722e-05}
{"train_loss": 0.13407889008522034, "global_step": 107019, "epoch": 1202, "lr": 6.647271109077716e-05}
{"train_loss": 0.11670331656932831, "global_step": 107020, "epoch": 1202, "lr": 6.647216371886255e-05}
{"train_loss": 0.14753730595111847, "global_step": 107021, "epoch": 1202, "lr": 6.647161634473345e-05}
{"train_loss": 0.15024471282958984, "global_step": 107022, "epoch": 1202, "lr": 6.647106896838994e-05}
{"train_loss": 0.1373274028301239, "global_step": 107023, "epoch": 1202, "lr": 6.647052158983211e-05}
{"train_loss": 0.13071309030056, "global_step": 107024, "epoch": 1202, "lr": 6.646997420906e-05}
{"train_loss": 0.17059068381786346, "global_step": 107025, "epoch": 1202, "lr": 6.646942682607372e-05}
{"train_loss": 0.17672935128211975, "global_step": 107026, "epoch": 1202, "lr": 6.64688794408733e-05}
{"train_loss": 0.16641077399253845, "global_step": 107027, "epoch": 1202, "lr": 6.646833205345885e-05}
{"train_loss": 0.19100287556648254, "global_step": 107028, "epoch": 1202, "lr": 6.646778466383045e-05}
{"train_loss": 0.1339302659034729, "global_step": 107029, "epoch": 1202, "lr": 6.646723727198815e-05}
{"train_loss": 0.21783705055713654, "global_step": 107030, "epoch": 1202, "lr": 6.646668987793203e-05}
{"train_loss": 0.16614830493927002, "global_step": 107031, "epoch": 1202, "lr": 6.646614248166217e-05}
{"train_loss": 0.17363600432872772, "global_step": 107032, "epoch": 1202, "lr": 6.646559508317862e-05}
{"train_loss": 0.13083630800247192, "global_step": 107033, "epoch": 1202, "lr": 6.646504768248148e-05}
{"train_loss": 0.21920306980609894, "global_step": 107034, "epoch": 1202, "lr": 6.646450027957081e-05}
{"train_loss": 0.13387103378772736, "global_step": 107035, "epoch": 1202, "lr": 6.64639528744467e-05}
{"train_loss": 0.11711156368255615, "global_step": 107036, "epoch": 1202, "lr": 6.64634054671092e-05}
{"train_loss": 0.1387479603290558, "global_step": 107037, "epoch": 1202, "lr": 6.646285805755841e-05}
{"train_loss": 0.14805205166339874, "global_step": 107038, "epoch": 1202, "lr": 6.646231064579438e-05}
{"train_loss": 0.19107194244861603, "global_step": 107039, "epoch": 1202, "lr": 6.64617632318172e-05}
{"train_loss": 0.18599411845207214, "global_step": 107040, "epoch": 1202, "lr": 6.646121581562692e-05}
{"train_loss": 0.1541590392589569, "global_step": 107041, "epoch": 1202, "lr": 6.646066839722365e-05}
{"train_loss": 0.1704913079738617, "global_step": 107042, "epoch": 1202, "lr": 6.646012097660743e-05}
{"train_loss": 0.21784964203834534, "global_step": 107043, "epoch": 1202, "lr": 6.645957355377836e-05}
{"train_loss": 0.22558173537254333, "global_step": 107044, "epoch": 1202, "lr": 6.645902612873649e-05}
{"train_loss": 0.18566174805164337, "global_step": 107045, "epoch": 1202, "lr": 6.645847870148192e-05}
{"train_loss": 0.09256155043840408, "global_step": 107046, "epoch": 1202, "lr": 6.64579312720147e-05}
{"train_loss": 0.15869086980819702, "global_step": 107047, "epoch": 1202, "lr": 6.64573838403349e-05}
{"train_loss": 0.2592496871948242, "global_step": 107048, "epoch": 1202, "lr": 6.645683640644262e-05}
{"train_loss": 0.14724832773208618, "global_step": 107049, "epoch": 1202, "lr": 6.645628897033791e-05}
{"train_loss": 0.11016581952571869, "global_step": 107050, "epoch": 1202, "lr": 6.645574153202086e-05}
{"train_loss": 0.2674923837184906, "global_step": 107051, "epoch": 1202, "lr": 6.645519409149152e-05}
{"train_loss": 0.12676174938678741, "global_step": 107052, "epoch": 1202, "lr": 6.645464664874999e-05}
{"train_loss": 0.1549054980278015, "global_step": 107053, "epoch": 1202, "lr": 6.645409920379634e-05}
{"train_loss": 0.09166178107261658, "global_step": 107054, "epoch": 1202, "lr": 6.645355175663063e-05}
{"train_loss": 0.16146020591259003, "global_step": 107055, "epoch": 1202, "lr": 6.645300430725294e-05}
{"train_loss": 0.22613759338855743, "global_step": 107056, "epoch": 1202, "lr": 6.645245685566335e-05}
{"train_loss": 0.12034662812948227, "global_step": 107057, "epoch": 1202, "lr": 6.645190940186193e-05}
{"train_loss": 0.1402161717414856, "global_step": 107058, "epoch": 1202, "lr": 6.645136194584873e-05}
{"train_loss": 0.17128624022006989, "global_step": 107059, "epoch": 1202, "lr": 6.645081448762387e-05}
{"train_loss": 0.1831693947315216, "global_step": 107060, "epoch": 1202, "lr": 6.645026702718738e-05}
{"train_loss": 0.2917206883430481, "global_step": 107061, "epoch": 1202, "lr": 6.644971956453937e-05}
{"train_loss": 0.1454908400774002, "global_step": 107062, "epoch": 1202, "lr": 6.644917209967988e-05}
{"train_loss": 0.1004156842827797, "global_step": 107063, "epoch": 1202, "lr": 6.644862463260901e-05}
{"train_loss": 0.1942930519580841, "global_step": 107064, "epoch": 1202, "lr": 6.644807716332682e-05}
{"train_loss": 0.20260706543922424, "global_step": 107065, "epoch": 1202, "lr": 6.644752969183339e-05}
{"train_loss": 0.1617979815836703, "global_step": 107066, "epoch": 1202, "lr": 6.644698221812878e-05, "val_loss": 4.8331475257873535}
{"train_loss": 0.19106824696063995, "global_step": 107067, "epoch": 1203, "lr": 6.644643474221308e-05}
{"train_loss": 0.18046550452709198, "global_step": 107068, "epoch": 1203, "lr": 6.644588726408634e-05}
{"train_loss": 0.17511571943759918, "global_step": 107069, "epoch": 1203, "lr": 6.644533978374867e-05}
{"train_loss": 0.12227518111467361, "global_step": 107070, "epoch": 1203, "lr": 6.644479230120013e-05}
{"train_loss": 0.15461145341396332, "global_step": 107071, "epoch": 1203, "lr": 6.644424481644078e-05}
{"train_loss": 0.11537887156009674, "global_step": 107072, "epoch": 1203, "lr": 6.64436973294707e-05}
{"train_loss": 0.16933239996433258, "global_step": 107073, "epoch": 1203, "lr": 6.644314984028997e-05}
{"train_loss": 0.1461724191904068, "global_step": 107074, "epoch": 1203, "lr": 6.644260234889865e-05}
{"train_loss": 0.14358703792095184, "global_step": 107075, "epoch": 1203, "lr": 6.644205485529683e-05}
{"train_loss": 0.14075441658496857, "global_step": 107076, "epoch": 1203, "lr": 6.644150735948458e-05}
{"train_loss": 0.17387856543064117, "global_step": 107077, "epoch": 1203, "lr": 6.644095986146195e-05}
{"train_loss": 0.11657707393169403, "global_step": 107078, "epoch": 1203, "lr": 6.644041236122906e-05}
{"train_loss": 0.12251967191696167, "global_step": 107079, "epoch": 1203, "lr": 6.643986485878594e-05}
{"train_loss": 0.14420035481452942, "global_step": 107080, "epoch": 1203, "lr": 6.64393173541327e-05}
{"train_loss": 0.10963553935289383, "global_step": 107081, "epoch": 1203, "lr": 6.643876984726936e-05}
{"train_loss": 0.14861491322517395, "global_step": 107082, "epoch": 1203, "lr": 6.643822233819605e-05}
{"train_loss": 0.10709413886070251, "global_step": 107083, "epoch": 1203, "lr": 6.643767482691282e-05}
{"train_loss": 0.15388429164886475, "global_step": 107084, "epoch": 1203, "lr": 6.643712731341975e-05}
{"train_loss": 0.18679207563400269, "global_step": 107085, "epoch": 1203, "lr": 6.643657979771689e-05}
{"train_loss": 0.0988360196352005, "global_step": 107086, "epoch": 1203, "lr": 6.643603227980434e-05}
{"train_loss": 0.12941859662532806, "global_step": 107087, "epoch": 1203, "lr": 6.643548475968218e-05}
{"train_loss": 0.17513100802898407, "global_step": 107088, "epoch": 1203, "lr": 6.643493723735045e-05}
{"train_loss": 0.18567045032978058, "global_step": 107089, "epoch": 1203, "lr": 6.643438971280926e-05}
{"train_loss": 0.18542037904262543, "global_step": 107090, "epoch": 1203, "lr": 6.643384218605866e-05}
{"train_loss": 0.13180720806121826, "global_step": 107091, "epoch": 1203, "lr": 6.643329465709872e-05}
{"train_loss": 0.12299976497888565, "global_step": 107092, "epoch": 1203, "lr": 6.643274712592954e-05}
{"train_loss": 0.214655339717865, "global_step": 107093, "epoch": 1203, "lr": 6.643219959255118e-05}
{"train_loss": 0.1490762084722519, "global_step": 107094, "epoch": 1203, "lr": 6.64316520569637e-05}
{"train_loss": 0.1718897670507431, "global_step": 107095, "epoch": 1203, "lr": 6.643110451916718e-05}
{"train_loss": 0.09734264016151428, "global_step": 107096, "epoch": 1203, "lr": 6.64305569791617e-05}
{"train_loss": 0.12991173565387726, "global_step": 107097, "epoch": 1203, "lr": 6.643000943694733e-05}
{"train_loss": 0.12874655425548553, "global_step": 107098, "epoch": 1203, "lr": 6.642946189252416e-05}
{"train_loss": 0.12406093627214432, "global_step": 107099, "epoch": 1203, "lr": 6.642891434589223e-05}
{"train_loss": 0.18780671060085297, "global_step": 107100, "epoch": 1203, "lr": 6.642836679705164e-05}
{"train_loss": 0.07344070822000504, "global_step": 107101, "epoch": 1203, "lr": 6.642781924600245e-05}
{"train_loss": 0.14318948984146118, "global_step": 107102, "epoch": 1203, "lr": 6.642727169274477e-05}
{"train_loss": 0.16966675221920013, "global_step": 107103, "epoch": 1203, "lr": 6.642672413727861e-05}
{"train_loss": 0.1667405664920807, "global_step": 107104, "epoch": 1203, "lr": 6.64261765796041e-05}
{"train_loss": 0.20024338364601135, "global_step": 107105, "epoch": 1203, "lr": 6.642562901972128e-05}
{"train_loss": 0.07253678888082504, "global_step": 107106, "epoch": 1203, "lr": 6.642508145763023e-05}
{"train_loss": 0.1614745855331421, "global_step": 107107, "epoch": 1203, "lr": 6.642453389333105e-05}
{"train_loss": 0.06027752533555031, "global_step": 107108, "epoch": 1203, "lr": 6.642398632682376e-05}
{"train_loss": 0.16326913237571716, "global_step": 107109, "epoch": 1203, "lr": 6.642343875810848e-05}
{"train_loss": 0.17872799932956696, "global_step": 107110, "epoch": 1203, "lr": 6.642289118718528e-05}
{"train_loss": 0.11313684284687042, "global_step": 107111, "epoch": 1203, "lr": 6.642234361405421e-05}
{"train_loss": 0.11731348186731339, "global_step": 107112, "epoch": 1203, "lr": 6.642179603871534e-05}
{"train_loss": 0.08421827107667923, "global_step": 107113, "epoch": 1203, "lr": 6.64212484611688e-05}
{"train_loss": 0.13116303086280823, "global_step": 107114, "epoch": 1203, "lr": 6.642070088141458e-05}
{"train_loss": 0.06780490279197693, "global_step": 107115, "epoch": 1203, "lr": 6.642015329945282e-05}
{"train_loss": 0.15347521007061005, "global_step": 107116, "epoch": 1203, "lr": 6.641960571528357e-05}
{"train_loss": 0.19655835628509521, "global_step": 107117, "epoch": 1203, "lr": 6.641905812890691e-05}
{"train_loss": 0.17919784784317017, "global_step": 107118, "epoch": 1203, "lr": 6.641851054032288e-05}
{"train_loss": 0.14014007151126862, "global_step": 107119, "epoch": 1203, "lr": 6.641796294953161e-05}
{"train_loss": 0.1299947202205658, "global_step": 107120, "epoch": 1203, "lr": 6.641741535653313e-05}
{"train_loss": 0.2839719355106354, "global_step": 107121, "epoch": 1203, "lr": 6.641686776132755e-05}
{"train_loss": 0.10513892024755478, "global_step": 107122, "epoch": 1203, "lr": 6.64163201639149e-05}
{"train_loss": 0.15859803557395935, "global_step": 107123, "epoch": 1203, "lr": 6.641577256429529e-05}
{"train_loss": 0.13559573888778687, "global_step": 107124, "epoch": 1203, "lr": 6.641522496246877e-05}
{"train_loss": 0.18377134203910828, "global_step": 107125, "epoch": 1203, "lr": 6.641467735843543e-05}
{"train_loss": 0.11389081925153732, "global_step": 107126, "epoch": 1203, "lr": 6.641412975219532e-05}
{"train_loss": 0.18000710010528564, "global_step": 107127, "epoch": 1203, "lr": 6.641358214374856e-05}
{"train_loss": 0.1076594889163971, "global_step": 107128, "epoch": 1203, "lr": 6.641303453309516e-05}
{"train_loss": 0.13037779927253723, "global_step": 107129, "epoch": 1203, "lr": 6.641248692023525e-05}
{"train_loss": 0.1855989694595337, "global_step": 107130, "epoch": 1203, "lr": 6.641193930516887e-05}
{"train_loss": 0.14967399835586548, "global_step": 107131, "epoch": 1203, "lr": 6.641139168789612e-05}
{"train_loss": 0.15613619983196259, "global_step": 107132, "epoch": 1203, "lr": 6.641084406841704e-05}
{"train_loss": 0.15974180400371552, "global_step": 107133, "epoch": 1203, "lr": 6.641029644673173e-05}
{"train_loss": 0.1576080173254013, "global_step": 107134, "epoch": 1203, "lr": 6.640974882284025e-05}
{"train_loss": 0.08681348711252213, "global_step": 107135, "epoch": 1203, "lr": 6.640920119674269e-05}
{"train_loss": 0.10576979070901871, "global_step": 107136, "epoch": 1203, "lr": 6.64086535684391e-05}
{"train_loss": 0.11007439345121384, "global_step": 107137, "epoch": 1203, "lr": 6.640810593792958e-05}
{"train_loss": 0.12543942034244537, "global_step": 107138, "epoch": 1203, "lr": 6.640755830521418e-05}
{"train_loss": 0.1882769614458084, "global_step": 107139, "epoch": 1203, "lr": 6.640701067029298e-05}
{"train_loss": 0.18222881853580475, "global_step": 107140, "epoch": 1203, "lr": 6.640646303316606e-05}
{"train_loss": 0.21489931643009186, "global_step": 107141, "epoch": 1203, "lr": 6.640591539383351e-05}
{"train_loss": 0.11055699735879898, "global_step": 107142, "epoch": 1203, "lr": 6.640536775229538e-05}
{"train_loss": 0.16871649026870728, "global_step": 107143, "epoch": 1203, "lr": 6.640482010855173e-05}
{"train_loss": 0.17319953441619873, "global_step": 107144, "epoch": 1203, "lr": 6.640427246260267e-05}
{"train_loss": 0.1967281699180603, "global_step": 107145, "epoch": 1203, "lr": 6.640372481444826e-05}
{"train_loss": 0.13231457769870758, "global_step": 107146, "epoch": 1203, "lr": 6.640317716408854e-05}
{"train_loss": 0.1594049334526062, "global_step": 107147, "epoch": 1203, "lr": 6.640262951152363e-05}
{"train_loss": 0.1425258070230484, "global_step": 107148, "epoch": 1203, "lr": 6.640208185675359e-05}
{"train_loss": 0.1106528639793396, "global_step": 107149, "epoch": 1203, "lr": 6.640153419977848e-05}
{"train_loss": 0.17048701643943787, "global_step": 107150, "epoch": 1203, "lr": 6.64009865405984e-05}
{"train_loss": 0.08951732516288757, "global_step": 107151, "epoch": 1203, "lr": 6.64004388792134e-05}
{"train_loss": 0.09097994118928909, "global_step": 107152, "epoch": 1203, "lr": 6.639989121562354e-05}
{"train_loss": 0.11929270625114441, "global_step": 107153, "epoch": 1203, "lr": 6.639934354982894e-05}
{"train_loss": 0.14762085676193237, "global_step": 107154, "epoch": 1203, "lr": 6.639879588182962e-05}
{"train_loss": 0.1452497775886166, "global_step": 107155, "epoch": 1203, "lr": 6.639824821162572e-05, "val_loss": 5.1343994140625}
{"train_loss": 0.09348376095294952, "global_step": 107156, "epoch": 1204, "lr": 6.639770053921725e-05}
{"train_loss": 0.16387531161308289, "global_step": 107157, "epoch": 1204, "lr": 6.639715286460431e-05}
{"train_loss": 0.15721474587917328, "global_step": 107158, "epoch": 1204, "lr": 6.639660518778699e-05}
{"train_loss": 0.14361916482448578, "global_step": 107159, "epoch": 1204, "lr": 6.639605750876532e-05}
{"train_loss": 0.16769057512283325, "global_step": 107160, "epoch": 1204, "lr": 6.639550982753944e-05}
{"train_loss": 0.13362933695316315, "global_step": 107161, "epoch": 1204, "lr": 6.639496214410935e-05}
{"train_loss": 0.16590018570423126, "global_step": 107162, "epoch": 1204, "lr": 6.639441445847516e-05}
{"train_loss": 0.13974185287952423, "global_step": 107163, "epoch": 1204, "lr": 6.639386677063695e-05}
{"train_loss": 0.15157686173915863, "global_step": 107164, "epoch": 1204, "lr": 6.639331908059479e-05}
{"train_loss": 0.12170221656560898, "global_step": 107165, "epoch": 1204, "lr": 6.639277138834873e-05}
{"train_loss": 0.0840831995010376, "global_step": 107166, "epoch": 1204, "lr": 6.639222369389886e-05}
{"train_loss": 0.08897712826728821, "global_step": 107167, "epoch": 1204, "lr": 6.639167599724528e-05}
{"train_loss": 0.13126638531684875, "global_step": 107168, "epoch": 1204, "lr": 6.639112829838802e-05}
{"train_loss": 0.14510728418827057, "global_step": 107169, "epoch": 1204, "lr": 6.639058059732718e-05}
{"train_loss": 0.11312466114759445, "global_step": 107170, "epoch": 1204, "lr": 6.639003289406282e-05}
{"train_loss": 0.13432680070400238, "global_step": 107171, "epoch": 1204, "lr": 6.638948518859502e-05}
{"train_loss": 0.13041862845420837, "global_step": 107172, "epoch": 1204, "lr": 6.638893748092385e-05}
{"train_loss": 0.17292501032352448, "global_step": 107173, "epoch": 1204, "lr": 6.63883897710494e-05}
{"train_loss": 0.20013055205345154, "global_step": 107174, "epoch": 1204, "lr": 6.638784205897173e-05}
{"train_loss": 0.16002394258975983, "global_step": 107175, "epoch": 1204, "lr": 6.63872943446909e-05}
{"train_loss": 0.18612128496170044, "global_step": 107176, "epoch": 1204, "lr": 6.638674662820701e-05}
{"train_loss": 0.14310264587402344, "global_step": 107177, "epoch": 1204, "lr": 6.638619890952012e-05}
{"train_loss": 0.12818388640880585, "global_step": 107178, "epoch": 1204, "lr": 6.638565118863029e-05}
{"train_loss": 0.22744148969650269, "global_step": 107179, "epoch": 1204, "lr": 6.638510346553765e-05}
{"train_loss": 0.10390733927488327, "global_step": 107180, "epoch": 1204, "lr": 6.63845557402422e-05}
{"train_loss": 0.16477610170841217, "global_step": 107181, "epoch": 1204, "lr": 6.638400801274405e-05}
{"train_loss": 0.10515329241752625, "global_step": 107182, "epoch": 1204, "lr": 6.638346028304328e-05}
{"train_loss": 0.1240730732679367, "global_step": 107183, "epoch": 1204, "lr": 6.638291255113993e-05}
{"train_loss": 0.10997568070888519, "global_step": 107184, "epoch": 1204, "lr": 6.638236481703411e-05}
{"train_loss": 0.11158329993486404, "global_step": 107185, "epoch": 1204, "lr": 6.63818170807259e-05}
{"train_loss": 0.1047540158033371, "global_step": 107186, "epoch": 1204, "lr": 6.638126934221533e-05}
{"train_loss": 0.10690302401781082, "global_step": 107187, "epoch": 1204, "lr": 6.638072160150251e-05}
{"train_loss": 0.13909417390823364, "global_step": 107188, "epoch": 1204, "lr": 6.638017385858749e-05}
{"train_loss": 0.16377265751361847, "global_step": 107189, "epoch": 1204, "lr": 6.637962611347036e-05}
{"train_loss": 0.08202832192182541, "global_step": 107190, "epoch": 1204, "lr": 6.637907836615118e-05}
{"train_loss": 0.16655750572681427, "global_step": 107191, "epoch": 1204, "lr": 6.637853061663004e-05}
{"train_loss": 0.20377999544143677, "global_step": 107192, "epoch": 1204, "lr": 6.637798286490702e-05}
{"train_loss": 0.12245327979326248, "global_step": 107193, "epoch": 1204, "lr": 6.637743511098216e-05}
{"train_loss": 0.09288053959608078, "global_step": 107194, "epoch": 1204, "lr": 6.637688735485557e-05}
{"train_loss": 0.15262553095817566, "global_step": 107195, "epoch": 1204, "lr": 6.63763395965273e-05}
{"train_loss": 0.17177274823188782, "global_step": 107196, "epoch": 1204, "lr": 6.637579183599742e-05}
{"train_loss": 0.18270093202590942, "global_step": 107197, "epoch": 1204, "lr": 6.637524407326603e-05}
{"train_loss": 0.16014686226844788, "global_step": 107198, "epoch": 1204, "lr": 6.637469630833319e-05}
{"train_loss": 0.1261451244354248, "global_step": 107199, "epoch": 1204, "lr": 6.637414854119897e-05}
{"train_loss": 0.1459828019142151, "global_step": 107200, "epoch": 1204, "lr": 6.637360077186345e-05}
{"train_loss": 0.15894779562950134, "global_step": 107201, "epoch": 1204, "lr": 6.637305300032668e-05}
{"train_loss": 0.1647711545228958, "global_step": 107202, "epoch": 1204, "lr": 6.637250522658876e-05}
{"train_loss": 0.15301984548568726, "global_step": 107203, "epoch": 1204, "lr": 6.637195745064976e-05}
{"train_loss": 0.11928107589483261, "global_step": 107204, "epoch": 1204, "lr": 6.637140967250975e-05}
{"train_loss": 0.13763262331485748, "global_step": 107205, "epoch": 1204, "lr": 6.63708618921688e-05}
{"train_loss": 0.2030094414949417, "global_step": 107206, "epoch": 1204, "lr": 6.637031410962698e-05}
{"train_loss": 0.1863948404788971, "global_step": 107207, "epoch": 1204, "lr": 6.636976632488438e-05}
{"train_loss": 0.13572095334529877, "global_step": 107208, "epoch": 1204, "lr": 6.636921853794107e-05}
{"train_loss": 0.1743151843547821, "global_step": 107209, "epoch": 1204, "lr": 6.636867074879711e-05}
{"train_loss": 0.15228049457073212, "global_step": 107210, "epoch": 1204, "lr": 6.636812295745257e-05}
{"train_loss": 0.09648799896240234, "global_step": 107211, "epoch": 1204, "lr": 6.636757516390756e-05}
{"train_loss": 0.1733170449733734, "global_step": 107212, "epoch": 1204, "lr": 6.636702736816212e-05}
{"train_loss": 0.10144653171300888, "global_step": 107213, "epoch": 1204, "lr": 6.636647957021632e-05}
{"train_loss": 0.14868377149105072, "global_step": 107214, "epoch": 1204, "lr": 6.636593177007027e-05}
{"train_loss": 0.15217123925685883, "global_step": 107215, "epoch": 1204, "lr": 6.6365383967724e-05}
{"train_loss": 0.18965604901313782, "global_step": 107216, "epoch": 1204, "lr": 6.636483616317761e-05}
{"train_loss": 0.15195629000663757, "global_step": 107217, "epoch": 1204, "lr": 6.636428835643116e-05}
{"train_loss": 0.1424369364976883, "global_step": 107218, "epoch": 1204, "lr": 6.636374054748476e-05}
{"train_loss": 0.16075538098812103, "global_step": 107219, "epoch": 1204, "lr": 6.636319273633842e-05}
{"train_loss": 0.1433166265487671, "global_step": 107220, "epoch": 1204, "lr": 6.636264492299226e-05}
{"train_loss": 0.11277367174625397, "global_step": 107221, "epoch": 1204, "lr": 6.636209710744635e-05}
{"train_loss": 0.16187280416488647, "global_step": 107222, "epoch": 1204, "lr": 6.636154928970076e-05}
{"train_loss": 0.16184118390083313, "global_step": 107223, "epoch": 1204, "lr": 6.636100146975553e-05}
{"train_loss": 0.13271570205688477, "global_step": 107224, "epoch": 1204, "lr": 6.636045364761079e-05}
{"train_loss": 0.135591059923172, "global_step": 107225, "epoch": 1204, "lr": 6.635990582326658e-05}
{"train_loss": 0.11647850275039673, "global_step": 107226, "epoch": 1204, "lr": 6.635935799672297e-05}
{"train_loss": 0.1606900840997696, "global_step": 107227, "epoch": 1204, "lr": 6.635881016798004e-05}
{"train_loss": 0.14035306870937347, "global_step": 107228, "epoch": 1204, "lr": 6.635826233703788e-05}
{"train_loss": 0.08496584743261337, "global_step": 107229, "epoch": 1204, "lr": 6.635771450389655e-05}
{"train_loss": 0.1645752191543579, "global_step": 107230, "epoch": 1204, "lr": 6.635716666855611e-05}
{"train_loss": 0.12982888519763947, "global_step": 107231, "epoch": 1204, "lr": 6.635661883101668e-05}
{"train_loss": 0.12904372811317444, "global_step": 107232, "epoch": 1204, "lr": 6.635607099127827e-05}
{"train_loss": 0.12014952301979065, "global_step": 107233, "epoch": 1204, "lr": 6.6355523149341e-05}
{"train_loss": 0.13855896890163422, "global_step": 107234, "epoch": 1204, "lr": 6.635497530520493e-05}
{"train_loss": 0.15274116396903992, "global_step": 107235, "epoch": 1204, "lr": 6.635442745887013e-05}
{"train_loss": 0.18977001309394836, "global_step": 107236, "epoch": 1204, "lr": 6.635387961033668e-05}
{"train_loss": 0.12839579582214355, "global_step": 107237, "epoch": 1204, "lr": 6.635333175960466e-05}
{"train_loss": 0.278898686170578, "global_step": 107238, "epoch": 1204, "lr": 6.635278390667412e-05}
{"train_loss": 0.07221669703722, "global_step": 107239, "epoch": 1204, "lr": 6.635223605154516e-05}
{"train_loss": 0.14177531003952026, "global_step": 107240, "epoch": 1204, "lr": 6.635168819421782e-05}
{"train_loss": 0.13706618547439575, "global_step": 107241, "epoch": 1204, "lr": 6.63511403346922e-05}
{"train_loss": 0.14997179806232452, "global_step": 107242, "epoch": 1204, "lr": 6.635059247296839e-05}
{"train_loss": 0.15413665771484375, "global_step": 107243, "epoch": 1204, "lr": 6.635004460904642e-05}
{"train_loss": 0.14400970048449013, "global_step": 107244, "epoch": 1204, "lr": 6.634949674292639e-05, "val_loss": 4.934575080871582}
{"train_loss": 0.16903473436832428, "global_step": 107245, "epoch": 1205, "lr": 6.634894887460836e-05}
{"train_loss": 0.1777976155281067, "global_step": 107246, "epoch": 1205, "lr": 6.634840100409244e-05}
{"train_loss": 0.18201333284378052, "global_step": 107247, "epoch": 1205, "lr": 6.634785313137866e-05}
{"train_loss": 0.13946092128753662, "global_step": 107248, "epoch": 1205, "lr": 6.634730525646711e-05}
{"train_loss": 0.1132272332906723, "global_step": 107249, "epoch": 1205, "lr": 6.634675737935786e-05}
{"train_loss": 0.18912899494171143, "global_step": 107250, "epoch": 1205, "lr": 6.634620950005099e-05}
{"train_loss": 0.10132031887769699, "global_step": 107251, "epoch": 1205, "lr": 6.634566161854657e-05}
{"train_loss": 0.14447848498821259, "global_step": 107252, "epoch": 1205, "lr": 6.634511373484469e-05}
{"train_loss": 0.13707007467746735, "global_step": 107253, "epoch": 1205, "lr": 6.634456584894539e-05}
{"train_loss": 0.19048619270324707, "global_step": 107254, "epoch": 1205, "lr": 6.634401796084876e-05}
{"train_loss": 0.1342911720275879, "global_step": 107255, "epoch": 1205, "lr": 6.634347007055491e-05}
{"train_loss": 0.17386499047279358, "global_step": 107256, "epoch": 1205, "lr": 6.634292217806385e-05}
{"train_loss": 0.1755618005990982, "global_step": 107257, "epoch": 1205, "lr": 6.634237428337569e-05}
{"train_loss": 0.140294149518013, "global_step": 107258, "epoch": 1205, "lr": 6.634182638649051e-05}
{"train_loss": 0.1382499486207962, "global_step": 107259, "epoch": 1205, "lr": 6.634127848740835e-05}
{"train_loss": 0.19654977321624756, "global_step": 107260, "epoch": 1205, "lr": 6.634073058612932e-05}
{"train_loss": 0.10351621359586716, "global_step": 107261, "epoch": 1205, "lr": 6.634018268265346e-05}
{"train_loss": 0.2034006267786026, "global_step": 107262, "epoch": 1205, "lr": 6.633963477698087e-05}
{"train_loss": 0.28906044363975525, "global_step": 107263, "epoch": 1205, "lr": 6.633908686911161e-05}
{"train_loss": 0.13865572214126587, "global_step": 107264, "epoch": 1205, "lr": 6.633853895904577e-05}
{"train_loss": 0.14199614524841309, "global_step": 107265, "epoch": 1205, "lr": 6.633799104678341e-05}
{"train_loss": 0.15869499742984772, "global_step": 107266, "epoch": 1205, "lr": 6.633744313232459e-05}
{"train_loss": 0.0885222777724266, "global_step": 107267, "epoch": 1205, "lr": 6.633689521566943e-05}
{"train_loss": 0.16259241104125977, "global_step": 107268, "epoch": 1205, "lr": 6.633634729681794e-05}
{"train_loss": 0.09398611634969711, "global_step": 107269, "epoch": 1205, "lr": 6.633579937577026e-05}
{"train_loss": 0.24219264090061188, "global_step": 107270, "epoch": 1205, "lr": 6.633525145252641e-05}
{"train_loss": 0.16466152667999268, "global_step": 107271, "epoch": 1205, "lr": 6.63347035270865e-05}
{"train_loss": 0.18284818530082703, "global_step": 107272, "epoch": 1205, "lr": 6.633415559945058e-05}
{"train_loss": 0.1775493621826172, "global_step": 107273, "epoch": 1205, "lr": 6.633360766961873e-05}
{"train_loss": 0.11842068284749985, "global_step": 107274, "epoch": 1205, "lr": 6.633305973759102e-05}
{"train_loss": 0.09371648728847504, "global_step": 107275, "epoch": 1205, "lr": 6.633251180336754e-05}
{"train_loss": 0.09744422137737274, "global_step": 107276, "epoch": 1205, "lr": 6.633196386694836e-05}
{"train_loss": 0.1629790961742401, "global_step": 107277, "epoch": 1205, "lr": 6.633141592833354e-05}
{"train_loss": 0.1439335197210312, "global_step": 107278, "epoch": 1205, "lr": 6.633086798752313e-05}
{"train_loss": 0.16177403926849365, "global_step": 107279, "epoch": 1205, "lr": 6.633032004451727e-05}
{"train_loss": 0.14101627469062805, "global_step": 107280, "epoch": 1205, "lr": 6.6329772099316e-05}
{"train_loss": 0.14133700728416443, "global_step": 107281, "epoch": 1205, "lr": 6.632922415191935e-05}
{"train_loss": 0.16695167124271393, "global_step": 107282, "epoch": 1205, "lr": 6.632867620232747e-05}
{"train_loss": 0.1423492729663849, "global_step": 107283, "epoch": 1205, "lr": 6.632812825054037e-05}
{"train_loss": 0.17350442707538605, "global_step": 107284, "epoch": 1205, "lr": 6.632758029655819e-05}
{"train_loss": 0.13266734778881073, "global_step": 107285, "epoch": 1205, "lr": 6.632703234038095e-05}
{"train_loss": 0.13958430290222168, "global_step": 107286, "epoch": 1205, "lr": 6.632648438200873e-05}
{"train_loss": 0.09556767344474792, "global_step": 107287, "epoch": 1205, "lr": 6.632593642144162e-05}
{"train_loss": 0.21798741817474365, "global_step": 107288, "epoch": 1205, "lr": 6.632538845867969e-05}
{"train_loss": 0.16444915533065796, "global_step": 107289, "epoch": 1205, "lr": 6.632484049372301e-05}
{"train_loss": 0.12778443098068237, "global_step": 107290, "epoch": 1205, "lr": 6.632429252657165e-05}
{"train_loss": 0.18015630543231964, "global_step": 107291, "epoch": 1205, "lr": 6.63237445572257e-05}
{"train_loss": 0.19473496079444885, "global_step": 107292, "epoch": 1205, "lr": 6.632319658568521e-05}
{"train_loss": 0.10885392874479294, "global_step": 107293, "epoch": 1205, "lr": 6.632264861195026e-05}
{"train_loss": 0.13705629110336304, "global_step": 107294, "epoch": 1205, "lr": 6.632210063602094e-05}
{"train_loss": 0.10785354673862457, "global_step": 107295, "epoch": 1205, "lr": 6.632155265789731e-05}
{"train_loss": 0.1339699625968933, "global_step": 107296, "epoch": 1205, "lr": 6.632100467757945e-05}
{"train_loss": 0.1705746054649353, "global_step": 107297, "epoch": 1205, "lr": 6.632045669506744e-05}
{"train_loss": 0.1723138988018036, "global_step": 107298, "epoch": 1205, "lr": 6.631990871036131e-05}
{"train_loss": 0.2626648247241974, "global_step": 107299, "epoch": 1205, "lr": 6.631936072346119e-05}
{"train_loss": 0.15789200365543365, "global_step": 107300, "epoch": 1205, "lr": 6.631881273436712e-05}
{"train_loss": 0.19361555576324463, "global_step": 107301, "epoch": 1205, "lr": 6.631826474307918e-05}
{"train_loss": 0.20527532696723938, "global_step": 107302, "epoch": 1205, "lr": 6.631771674959747e-05}
{"train_loss": 0.09202569723129272, "global_step": 107303, "epoch": 1205, "lr": 6.631716875392202e-05}
{"train_loss": 0.1734870970249176, "global_step": 107304, "epoch": 1205, "lr": 6.631662075605294e-05}
{"train_loss": 0.14952458441257477, "global_step": 107305, "epoch": 1205, "lr": 6.631607275599029e-05}
{"train_loss": 0.1322457492351532, "global_step": 107306, "epoch": 1205, "lr": 6.631552475373414e-05}
{"train_loss": 0.12334144115447998, "global_step": 107307, "epoch": 1205, "lr": 6.631497674928456e-05}
{"train_loss": 0.1260649859905243, "global_step": 107308, "epoch": 1205, "lr": 6.631442874264164e-05}
{"train_loss": 0.17305110394954681, "global_step": 107309, "epoch": 1205, "lr": 6.631388073380544e-05}
{"train_loss": 0.07596198469400406, "global_step": 107310, "epoch": 1205, "lr": 6.631333272277603e-05}
{"train_loss": 0.17655354738235474, "global_step": 107311, "epoch": 1205, "lr": 6.63127847095535e-05}
{"train_loss": 0.1793895959854126, "global_step": 107312, "epoch": 1205, "lr": 6.63122366941379e-05}
{"train_loss": 0.1635558158159256, "global_step": 107313, "epoch": 1205, "lr": 6.631168867652935e-05}
{"train_loss": 0.16600915789604187, "global_step": 107314, "epoch": 1205, "lr": 6.631114065672787e-05}
{"train_loss": 0.154727965593338, "global_step": 107315, "epoch": 1205, "lr": 6.631059263473357e-05}
{"train_loss": 0.15659783780574799, "global_step": 107316, "epoch": 1205, "lr": 6.631004461054649e-05}
{"train_loss": 0.17320473492145538, "global_step": 107317, "epoch": 1205, "lr": 6.630949658416674e-05}
{"train_loss": 0.2116960883140564, "global_step": 107318, "epoch": 1205, "lr": 6.630894855559436e-05}
{"train_loss": 0.17283400893211365, "global_step": 107319, "epoch": 1205, "lr": 6.630840052482945e-05}
{"train_loss": 0.18011058866977692, "global_step": 107320, "epoch": 1205, "lr": 6.630785249187208e-05}
{"train_loss": 0.09804969280958176, "global_step": 107321, "epoch": 1205, "lr": 6.630730445672231e-05}
{"train_loss": 0.16627243161201477, "global_step": 107322, "epoch": 1205, "lr": 6.630675641938023e-05}
{"train_loss": 0.2231428623199463, "global_step": 107323, "epoch": 1205, "lr": 6.630620837984589e-05}
{"train_loss": 0.1487557590007782, "global_step": 107324, "epoch": 1205, "lr": 6.630566033811939e-05}
{"train_loss": 0.09789400547742844, "global_step": 107325, "epoch": 1205, "lr": 6.630511229420079e-05}
{"train_loss": 0.09639491885900497, "global_step": 107326, "epoch": 1205, "lr": 6.630456424809017e-05}
{"train_loss": 0.18309295177459717, "global_step": 107327, "epoch": 1205, "lr": 6.63040161997876e-05}
{"train_loss": 0.11134360730648041, "global_step": 107328, "epoch": 1205, "lr": 6.630346814929314e-05}
{"train_loss": 0.0847449079155922, "global_step": 107329, "epoch": 1205, "lr": 6.63029200966069e-05}
{"train_loss": 0.1680293083190918, "global_step": 107330, "epoch": 1205, "lr": 6.630237204172892e-05}
{"train_loss": 0.15768709778785706, "global_step": 107331, "epoch": 1205, "lr": 6.630182398465929e-05}
{"train_loss": 0.16558009386062622, "global_step": 107332, "epoch": 1205, "lr": 6.630127592539806e-05}
{"train_loss": 0.1542069577936376, "global_step": 107333, "epoch": 1205, "lr": 6.630072786394534e-05, "val_loss": 4.896389961242676, "train_action_mse_error": 24.648136138916016}
{"train_loss": 0.21591275930404663, "global_step": 107334, "epoch": 1206, "lr": 6.630017980030119e-05}
{"train_loss": 0.19116199016571045, "global_step": 107335, "epoch": 1206, "lr": 6.629963173446567e-05}
{"train_loss": 0.1853671669960022, "global_step": 107336, "epoch": 1206, "lr": 6.629908366643886e-05}
{"train_loss": 0.13942615687847137, "global_step": 107337, "epoch": 1206, "lr": 6.629853559622084e-05}
{"train_loss": 0.12407732009887695, "global_step": 107338, "epoch": 1206, "lr": 6.629798752381167e-05}
{"train_loss": 0.10205964744091034, "global_step": 107339, "epoch": 1206, "lr": 6.629743944921146e-05}
{"train_loss": 0.19762340188026428, "global_step": 107340, "epoch": 1206, "lr": 6.629689137242023e-05}
{"train_loss": 0.1267857551574707, "global_step": 107341, "epoch": 1206, "lr": 6.62963432934381e-05}
{"train_loss": 0.1614353507757187, "global_step": 107342, "epoch": 1206, "lr": 6.629579521226512e-05}
{"train_loss": 0.21342821419239044, "global_step": 107343, "epoch": 1206, "lr": 6.629524712890137e-05}
{"train_loss": 0.15848909318447113, "global_step": 107344, "epoch": 1206, "lr": 6.629469904334692e-05}
{"train_loss": 0.2145189642906189, "global_step": 107345, "epoch": 1206, "lr": 6.629415095560186e-05}
{"train_loss": 0.319410115480423, "global_step": 107346, "epoch": 1206, "lr": 6.629360286566623e-05}
{"train_loss": 0.11123895645141602, "global_step": 107347, "epoch": 1206, "lr": 6.629305477354012e-05}
{"train_loss": 0.1262490302324295, "global_step": 107348, "epoch": 1206, "lr": 6.629250667922364e-05}
{"train_loss": 0.12267829477787018, "global_step": 107349, "epoch": 1206, "lr": 6.629195858271679e-05}
{"train_loss": 0.17507286369800568, "global_step": 107350, "epoch": 1206, "lr": 6.62914104840197e-05}
{"train_loss": 0.19330522418022156, "global_step": 107351, "epoch": 1206, "lr": 6.629086238313245e-05}
{"train_loss": 0.16965018212795258, "global_step": 107352, "epoch": 1206, "lr": 6.629031428005508e-05}
{"train_loss": 0.11314313858747482, "global_step": 107353, "epoch": 1206, "lr": 6.628976617478766e-05}
{"train_loss": 0.1615823358297348, "global_step": 107354, "epoch": 1206, "lr": 6.62892180673303e-05}
{"train_loss": 0.18157020211219788, "global_step": 107355, "epoch": 1206, "lr": 6.628866995768304e-05}
{"train_loss": 0.23609711229801178, "global_step": 107356, "epoch": 1206, "lr": 6.628812184584597e-05}
{"train_loss": 0.12104278057813644, "global_step": 107357, "epoch": 1206, "lr": 6.628757373181916e-05}
{"train_loss": 0.1555425226688385, "global_step": 107358, "epoch": 1206, "lr": 6.628702561560268e-05}
{"train_loss": 0.25163576006889343, "global_step": 107359, "epoch": 1206, "lr": 6.628647749719662e-05}
{"train_loss": 0.17985545098781586, "global_step": 107360, "epoch": 1206, "lr": 6.628592937660103e-05}
{"train_loss": 0.16555142402648926, "global_step": 107361, "epoch": 1206, "lr": 6.6285381253816e-05}
{"train_loss": 0.15325209498405457, "global_step": 107362, "epoch": 1206, "lr": 6.628483312884161e-05}
{"train_loss": 0.09747448563575745, "global_step": 107363, "epoch": 1206, "lr": 6.62842850016779e-05}
{"train_loss": 0.13247132301330566, "global_step": 107364, "epoch": 1206, "lr": 6.628373687232499e-05}
{"train_loss": 0.13272543251514435, "global_step": 107365, "epoch": 1206, "lr": 6.62831887407829e-05}
{"train_loss": 0.20703141391277313, "global_step": 107366, "epoch": 1206, "lr": 6.628264060705177e-05}
{"train_loss": 0.12074778228998184, "global_step": 107367, "epoch": 1206, "lr": 6.628209247113162e-05}
{"train_loss": 0.1410323530435562, "global_step": 107368, "epoch": 1206, "lr": 6.628154433302254e-05}
{"train_loss": 0.12832914292812347, "global_step": 107369, "epoch": 1206, "lr": 6.62809961927246e-05}
{"train_loss": 0.19363842904567719, "global_step": 107370, "epoch": 1206, "lr": 6.628044805023788e-05}
{"train_loss": 0.21107089519500732, "global_step": 107371, "epoch": 1206, "lr": 6.627989990556246e-05}
{"train_loss": 0.12485536932945251, "global_step": 107372, "epoch": 1206, "lr": 6.62793517586984e-05}
{"train_loss": 0.10400556027889252, "global_step": 107373, "epoch": 1206, "lr": 6.627880360964578e-05}
{"train_loss": 0.16365177929401398, "global_step": 107374, "epoch": 1206, "lr": 6.627825545840468e-05}
{"train_loss": 0.12993226945400238, "global_step": 107375, "epoch": 1206, "lr": 6.627770730497516e-05}
{"train_loss": 0.16224627196788788, "global_step": 107376, "epoch": 1206, "lr": 6.627715914935731e-05}
{"train_loss": 0.2338644415140152, "global_step": 107377, "epoch": 1206, "lr": 6.627661099155118e-05}
{"train_loss": 0.1824059784412384, "global_step": 107378, "epoch": 1206, "lr": 6.627606283155687e-05}
{"train_loss": 0.11564184725284576, "global_step": 107379, "epoch": 1206, "lr": 6.627551466937443e-05}
{"train_loss": 0.2036476731300354, "global_step": 107380, "epoch": 1206, "lr": 6.627496650500396e-05}
{"train_loss": 0.20304009318351746, "global_step": 107381, "epoch": 1206, "lr": 6.62744183384455e-05}
{"train_loss": 0.09432615339756012, "global_step": 107382, "epoch": 1206, "lr": 6.627387016969916e-05}
{"train_loss": 0.1705527901649475, "global_step": 107383, "epoch": 1206, "lr": 6.6273321998765e-05}
{"train_loss": 0.22743315994739532, "global_step": 107384, "epoch": 1206, "lr": 6.627277382564308e-05}
{"train_loss": 0.06358402222394943, "global_step": 107385, "epoch": 1206, "lr": 6.627222565033348e-05}
{"train_loss": 0.19416919350624084, "global_step": 107386, "epoch": 1206, "lr": 6.627167747283628e-05}
{"train_loss": 0.10725176334381104, "global_step": 107387, "epoch": 1206, "lr": 6.627112929315155e-05}
{"train_loss": 0.11408352106809616, "global_step": 107388, "epoch": 1206, "lr": 6.627058111127938e-05}
{"train_loss": 0.21836145222187042, "global_step": 107389, "epoch": 1206, "lr": 6.62700329272198e-05}
{"train_loss": 0.1744309961795807, "global_step": 107390, "epoch": 1206, "lr": 6.626948474097293e-05}
{"train_loss": 0.18141940236091614, "global_step": 107391, "epoch": 1206, "lr": 6.626893655253883e-05}
{"train_loss": 0.155021071434021, "global_step": 107392, "epoch": 1206, "lr": 6.626838836191757e-05}
{"train_loss": 0.09805788099765778, "global_step": 107393, "epoch": 1206, "lr": 6.62678401691092e-05}
{"train_loss": 0.14125362038612366, "global_step": 107394, "epoch": 1206, "lr": 6.626729197411385e-05}
{"train_loss": 0.19227732717990875, "global_step": 107395, "epoch": 1206, "lr": 6.626674377693156e-05}
{"train_loss": 0.17919784784317017, "global_step": 107396, "epoch": 1206, "lr": 6.626619557756237e-05}
{"train_loss": 0.14678116142749786, "global_step": 107397, "epoch": 1206, "lr": 6.626564737600643e-05}
{"train_loss": 0.16216398775577545, "global_step": 107398, "epoch": 1206, "lr": 6.626509917226373e-05}
{"train_loss": 0.1721648871898651, "global_step": 107399, "epoch": 1206, "lr": 6.626455096633442e-05}
{"train_loss": 0.15638959407806396, "global_step": 107400, "epoch": 1206, "lr": 6.626400275821853e-05}
{"train_loss": 0.1372312605381012, "global_step": 107401, "epoch": 1206, "lr": 6.626345454791614e-05}
{"train_loss": 0.16429607570171356, "global_step": 107402, "epoch": 1206, "lr": 6.626290633542732e-05}
{"train_loss": 0.12038511782884598, "global_step": 107403, "epoch": 1206, "lr": 6.626235812075215e-05}
{"train_loss": 0.21469859778881073, "global_step": 107404, "epoch": 1206, "lr": 6.626180990389072e-05}
{"train_loss": 0.1648930013179779, "global_step": 107405, "epoch": 1206, "lr": 6.626126168484307e-05}
{"train_loss": 0.1462014615535736, "global_step": 107406, "epoch": 1206, "lr": 6.62607134636093e-05}
{"train_loss": 0.08471985906362534, "global_step": 107407, "epoch": 1206, "lr": 6.626016524018948e-05}
{"train_loss": 0.14707204699516296, "global_step": 107408, "epoch": 1206, "lr": 6.625961701458368e-05}
{"train_loss": 0.19589072465896606, "global_step": 107409, "epoch": 1206, "lr": 6.625906878679197e-05}
{"train_loss": 0.10407766699790955, "global_step": 107410, "epoch": 1206, "lr": 6.625852055681441e-05}
{"train_loss": 0.14578118920326233, "global_step": 107411, "epoch": 1206, "lr": 6.625797232465112e-05}
{"train_loss": 0.1830649971961975, "global_step": 107412, "epoch": 1206, "lr": 6.625742409030213e-05}
{"train_loss": 0.14742180705070496, "global_step": 107413, "epoch": 1206, "lr": 6.625687585376752e-05}
{"train_loss": 0.12410079687833786, "global_step": 107414, "epoch": 1206, "lr": 6.625632761504738e-05}
{"train_loss": 0.19883854687213898, "global_step": 107415, "epoch": 1206, "lr": 6.625577937414176e-05}
{"train_loss": 0.15457527339458466, "global_step": 107416, "epoch": 1206, "lr": 6.625523113105076e-05}
{"train_loss": 0.16452372074127197, "global_step": 107417, "epoch": 1206, "lr": 6.625468288577445e-05}
{"train_loss": 0.16897687315940857, "global_step": 107418, "epoch": 1206, "lr": 6.625413463831287e-05}
{"train_loss": 0.1379760503768921, "global_step": 107419, "epoch": 1206, "lr": 6.625358638866614e-05}
{"train_loss": 0.1457664668560028, "global_step": 107420, "epoch": 1206, "lr": 6.625303813683433e-05}
{"train_loss": 0.10798130184412003, "global_step": 107421, "epoch": 1206, "lr": 6.625248988281747e-05}
{"train_loss": 0.15958841287353065, "global_step": 107422, "epoch": 1206, "lr": 6.625194162661567e-05, "val_loss": 4.764565944671631}
{"train_loss": 0.17101360857486725, "global_step": 107423, "epoch": 1207, "lr": 6.625139336822898e-05}
{"train_loss": 0.16467660665512085, "global_step": 107424, "epoch": 1207, "lr": 6.625084510765752e-05}
{"train_loss": 0.1460788995027542, "global_step": 107425, "epoch": 1207, "lr": 6.62502968449013e-05}
{"train_loss": 0.14867502450942993, "global_step": 107426, "epoch": 1207, "lr": 6.624974857996045e-05}
{"train_loss": 0.06454119831323624, "global_step": 107427, "epoch": 1207, "lr": 6.624920031283501e-05}
{"train_loss": 0.1692160815000534, "global_step": 107428, "epoch": 1207, "lr": 6.624865204352506e-05}
{"train_loss": 0.1354312151670456, "global_step": 107429, "epoch": 1207, "lr": 6.624810377203068e-05}
{"train_loss": 0.21486321091651917, "global_step": 107430, "epoch": 1207, "lr": 6.624755549835193e-05}
{"train_loss": 0.1654118299484253, "global_step": 107431, "epoch": 1207, "lr": 6.624700722248891e-05}
{"train_loss": 0.13511966168880463, "global_step": 107432, "epoch": 1207, "lr": 6.624645894444167e-05}
{"train_loss": 0.14867646992206573, "global_step": 107433, "epoch": 1207, "lr": 6.62459106642103e-05}
{"train_loss": 0.1463683396577835, "global_step": 107434, "epoch": 1207, "lr": 6.624536238179485e-05}
{"train_loss": 0.1628856509923935, "global_step": 107435, "epoch": 1207, "lr": 6.624481409719542e-05}
{"train_loss": 0.09757339209318161, "global_step": 107436, "epoch": 1207, "lr": 6.624426581041208e-05}
{"train_loss": 0.10488531738519669, "global_step": 107437, "epoch": 1207, "lr": 6.624371752144487e-05}
{"train_loss": 0.09278438240289688, "global_step": 107438, "epoch": 1207, "lr": 6.62431692302939e-05}
{"train_loss": 0.14675438404083252, "global_step": 107439, "epoch": 1207, "lr": 6.624262093695926e-05}
{"train_loss": 0.0960974469780922, "global_step": 107440, "epoch": 1207, "lr": 6.624207264144098e-05}
{"train_loss": 0.14611399173736572, "global_step": 107441, "epoch": 1207, "lr": 6.624152434373914e-05}
{"train_loss": 0.11121814697980881, "global_step": 107442, "epoch": 1207, "lr": 6.624097604385384e-05}
{"train_loss": 0.21458913385868073, "global_step": 107443, "epoch": 1207, "lr": 6.624042774178512e-05}
{"train_loss": 0.1742437332868576, "global_step": 107444, "epoch": 1207, "lr": 6.623987943753308e-05}
{"train_loss": 0.18250857293605804, "global_step": 107445, "epoch": 1207, "lr": 6.62393311310978e-05}
{"train_loss": 0.15482287108898163, "global_step": 107446, "epoch": 1207, "lr": 6.623878282247934e-05}
{"train_loss": 0.18980297446250916, "global_step": 107447, "epoch": 1207, "lr": 6.623823451167776e-05}
{"train_loss": 0.2219121754169464, "global_step": 107448, "epoch": 1207, "lr": 6.623768619869315e-05}
{"train_loss": 0.137207493185997, "global_step": 107449, "epoch": 1207, "lr": 6.623713788352558e-05}
{"train_loss": 0.1475883275270462, "global_step": 107450, "epoch": 1207, "lr": 6.623658956617513e-05}
{"train_loss": 0.1877995729446411, "global_step": 107451, "epoch": 1207, "lr": 6.623604124664187e-05}
{"train_loss": 0.14310963451862335, "global_step": 107452, "epoch": 1207, "lr": 6.623549292492587e-05}
{"train_loss": 0.12416625767946243, "global_step": 107453, "epoch": 1207, "lr": 6.62349446010272e-05}
{"train_loss": 0.20062527060508728, "global_step": 107454, "epoch": 1207, "lr": 6.623439627494595e-05}
{"train_loss": 0.13976064324378967, "global_step": 107455, "epoch": 1207, "lr": 6.623384794668218e-05}
{"train_loss": 0.17431074380874634, "global_step": 107456, "epoch": 1207, "lr": 6.623329961623597e-05}
{"train_loss": 0.19642643630504608, "global_step": 107457, "epoch": 1207, "lr": 6.623275128360737e-05}
{"train_loss": 0.24153849482536316, "global_step": 107458, "epoch": 1207, "lr": 6.623220294879649e-05}
{"train_loss": 0.1715770810842514, "global_step": 107459, "epoch": 1207, "lr": 6.62316546118034e-05}
{"train_loss": 0.13021546602249146, "global_step": 107460, "epoch": 1207, "lr": 6.623110627262815e-05}
{"train_loss": 0.17263269424438477, "global_step": 107461, "epoch": 1207, "lr": 6.623055793127083e-05}
{"train_loss": 0.1323947161436081, "global_step": 107462, "epoch": 1207, "lr": 6.62300095877315e-05}
{"train_loss": 0.12902036309242249, "global_step": 107463, "epoch": 1207, "lr": 6.622946124201022e-05}
{"train_loss": 0.06872233748435974, "global_step": 107464, "epoch": 1207, "lr": 6.622891289410713e-05}
{"train_loss": 0.10842956602573395, "global_step": 107465, "epoch": 1207, "lr": 6.622836454402225e-05}
{"train_loss": 0.1287536472082138, "global_step": 107466, "epoch": 1207, "lr": 6.622781619175566e-05}
{"train_loss": 0.10847979038953781, "global_step": 107467, "epoch": 1207, "lr": 6.622726783730744e-05}
{"train_loss": 0.2535523474216461, "global_step": 107468, "epoch": 1207, "lr": 6.622671948067766e-05}
{"train_loss": 0.13625681400299072, "global_step": 107469, "epoch": 1207, "lr": 6.622617112186639e-05}
{"train_loss": 0.15089592337608337, "global_step": 107470, "epoch": 1207, "lr": 6.622562276087372e-05}
{"train_loss": 0.21552909910678864, "global_step": 107471, "epoch": 1207, "lr": 6.62250743976997e-05}
{"train_loss": 0.1599099487066269, "global_step": 107472, "epoch": 1207, "lr": 6.622452603234441e-05}
{"train_loss": 0.08573011308908463, "global_step": 107473, "epoch": 1207, "lr": 6.622397766480795e-05}
{"train_loss": 0.06001840904355049, "global_step": 107474, "epoch": 1207, "lr": 6.622342929509037e-05}
{"train_loss": 0.10893523693084717, "global_step": 107475, "epoch": 1207, "lr": 6.622288092319174e-05}
{"train_loss": 0.19556204974651337, "global_step": 107476, "epoch": 1207, "lr": 6.622233254911214e-05}
{"train_loss": 0.11528200656175613, "global_step": 107477, "epoch": 1207, "lr": 6.622178417285166e-05}
{"train_loss": 0.174299955368042, "global_step": 107478, "epoch": 1207, "lr": 6.622123579441035e-05}
{"train_loss": 0.10773961991071701, "global_step": 107479, "epoch": 1207, "lr": 6.622068741378829e-05}
{"train_loss": 0.19109602272510529, "global_step": 107480, "epoch": 1207, "lr": 6.622013903098556e-05}
{"train_loss": 0.14878742396831512, "global_step": 107481, "epoch": 1207, "lr": 6.621959064600221e-05}
{"train_loss": 0.21955691277980804, "global_step": 107482, "epoch": 1207, "lr": 6.621904225883835e-05}
{"train_loss": 0.14057236909866333, "global_step": 107483, "epoch": 1207, "lr": 6.621849386949404e-05}
{"train_loss": 0.19766274094581604, "global_step": 107484, "epoch": 1207, "lr": 6.621794547796937e-05}
{"train_loss": 0.16057512164115906, "global_step": 107485, "epoch": 1207, "lr": 6.621739708426437e-05}
{"train_loss": 0.15283595025539398, "global_step": 107486, "epoch": 1207, "lr": 6.621684868837914e-05}
{"train_loss": 0.15510180592536926, "global_step": 107487, "epoch": 1207, "lr": 6.621630029031376e-05}
{"train_loss": 0.19603820145130157, "global_step": 107488, "epoch": 1207, "lr": 6.62157518900683e-05}
{"train_loss": 0.1246509924530983, "global_step": 107489, "epoch": 1207, "lr": 6.621520348764282e-05}
{"train_loss": 0.15721917152404785, "global_step": 107490, "epoch": 1207, "lr": 6.62146550830374e-05}
{"train_loss": 0.07016754895448685, "global_step": 107491, "epoch": 1207, "lr": 6.621410667625211e-05}
{"train_loss": 0.11382929235696793, "global_step": 107492, "epoch": 1207, "lr": 6.621355826728705e-05}
{"train_loss": 0.1550193727016449, "global_step": 107493, "epoch": 1207, "lr": 6.621300985614227e-05}
{"train_loss": 0.09120766818523407, "global_step": 107494, "epoch": 1207, "lr": 6.621246144281786e-05}
{"train_loss": 0.081545889377594, "global_step": 107495, "epoch": 1207, "lr": 6.621191302731386e-05}
{"train_loss": 0.1495516002178192, "global_step": 107496, "epoch": 1207, "lr": 6.621136460963038e-05}
{"train_loss": 0.09640882164239883, "global_step": 107497, "epoch": 1207, "lr": 6.621081618976748e-05}
{"train_loss": 0.09959694743156433, "global_step": 107498, "epoch": 1207, "lr": 6.621026776772522e-05}
{"train_loss": 0.13289356231689453, "global_step": 107499, "epoch": 1207, "lr": 6.620971934350368e-05}
{"train_loss": 0.15858998894691467, "global_step": 107500, "epoch": 1207, "lr": 6.620917091710297e-05}
{"train_loss": 0.15931464731693268, "global_step": 107501, "epoch": 1207, "lr": 6.62086224885231e-05}
{"train_loss": 0.20746290683746338, "global_step": 107502, "epoch": 1207, "lr": 6.620807405776421e-05}
{"train_loss": 0.11294426769018173, "global_step": 107503, "epoch": 1207, "lr": 6.620752562482635e-05}
{"train_loss": 0.11234898120164871, "global_step": 107504, "epoch": 1207, "lr": 6.620697718970956e-05}
{"train_loss": 0.20685653388500214, "global_step": 107505, "epoch": 1207, "lr": 6.620642875241394e-05}
{"train_loss": 0.16558775305747986, "global_step": 107506, "epoch": 1207, "lr": 6.620588031293958e-05}
{"train_loss": 0.16726168990135193, "global_step": 107507, "epoch": 1207, "lr": 6.620533187128652e-05}
{"train_loss": 0.1841810941696167, "global_step": 107508, "epoch": 1207, "lr": 6.620478342745486e-05}
{"train_loss": 0.1382140815258026, "global_step": 107509, "epoch": 1207, "lr": 6.620423498144465e-05}
{"train_loss": 0.1694921851158142, "global_step": 107510, "epoch": 1207, "lr": 6.620368653325599e-05}
{"train_loss": 0.14848714133494356, "global_step": 107511, "epoch": 1207, "lr": 6.620313808288894e-05, "val_loss": 5.038706302642822}
{"train_loss": 0.18535436689853668, "global_step": 107512, "epoch": 1208, "lr": 6.620258963034357e-05}
{"train_loss": 0.17350110411643982, "global_step": 107513, "epoch": 1208, "lr": 6.620204117561999e-05}
{"train_loss": 0.11668357253074646, "global_step": 107514, "epoch": 1208, "lr": 6.62014927187182e-05}
{"train_loss": 0.20146465301513672, "global_step": 107515, "epoch": 1208, "lr": 6.620094425963834e-05}
{"train_loss": 0.16721341013908386, "global_step": 107516, "epoch": 1208, "lr": 6.620039579838046e-05}
{"train_loss": 0.1256987601518631, "global_step": 107517, "epoch": 1208, "lr": 6.619984733494462e-05}
{"train_loss": 0.10660212486982346, "global_step": 107518, "epoch": 1208, "lr": 6.619929886933092e-05}
{"train_loss": 0.15743455290794373, "global_step": 107519, "epoch": 1208, "lr": 6.619875040153943e-05}
{"train_loss": 0.14540158212184906, "global_step": 107520, "epoch": 1208, "lr": 6.61982019315702e-05}
{"train_loss": 0.1657014638185501, "global_step": 107521, "epoch": 1208, "lr": 6.619765345942332e-05}
{"train_loss": 0.13179618120193481, "global_step": 107522, "epoch": 1208, "lr": 6.619710498509888e-05}
{"train_loss": 0.13354849815368652, "global_step": 107523, "epoch": 1208, "lr": 6.619655650859692e-05}
{"train_loss": 0.13099321722984314, "global_step": 107524, "epoch": 1208, "lr": 6.619600802991754e-05}
{"train_loss": 0.14941415190696716, "global_step": 107525, "epoch": 1208, "lr": 6.61954595490608e-05}
{"train_loss": 0.1619921326637268, "global_step": 107526, "epoch": 1208, "lr": 6.619491106602678e-05}
{"train_loss": 0.12890972197055817, "global_step": 107527, "epoch": 1208, "lr": 6.619436258081555e-05}
{"train_loss": 0.14535431563854218, "global_step": 107528, "epoch": 1208, "lr": 6.619381409342717e-05}
{"train_loss": 0.09173256158828735, "global_step": 107529, "epoch": 1208, "lr": 6.619326560386175e-05}
{"train_loss": 0.2355973720550537, "global_step": 107530, "epoch": 1208, "lr": 6.619271711211934e-05}
{"train_loss": 0.10918396711349487, "global_step": 107531, "epoch": 1208, "lr": 6.619216861820002e-05}
{"train_loss": 0.19854053854942322, "global_step": 107532, "epoch": 1208, "lr": 6.619162012210384e-05}
{"train_loss": 0.15549521148204803, "global_step": 107533, "epoch": 1208, "lr": 6.619107162383091e-05}
{"train_loss": 0.12346746772527695, "global_step": 107534, "epoch": 1208, "lr": 6.619052312338127e-05}
{"train_loss": 0.15886275470256805, "global_step": 107535, "epoch": 1208, "lr": 6.618997462075502e-05}
{"train_loss": 0.08911202102899551, "global_step": 107536, "epoch": 1208, "lr": 6.618942611595224e-05}
{"train_loss": 0.1142600029706955, "global_step": 107537, "epoch": 1208, "lr": 6.618887760897297e-05}
{"train_loss": 0.16877806186676025, "global_step": 107538, "epoch": 1208, "lr": 6.618832909981731e-05}
{"train_loss": 0.11965997517108917, "global_step": 107539, "epoch": 1208, "lr": 6.618778058848534e-05}
{"train_loss": 0.2826153039932251, "global_step": 107540, "epoch": 1208, "lr": 6.618723207497711e-05}
{"train_loss": 0.15607935190200806, "global_step": 107541, "epoch": 1208, "lr": 6.618668355929271e-05}
{"train_loss": 0.1954011470079422, "global_step": 107542, "epoch": 1208, "lr": 6.618613504143221e-05}
{"train_loss": 0.18562597036361694, "global_step": 107543, "epoch": 1208, "lr": 6.618558652139566e-05}
{"train_loss": 0.10167109221220016, "global_step": 107544, "epoch": 1208, "lr": 6.618503799918318e-05}
{"train_loss": 0.18110276758670807, "global_step": 107545, "epoch": 1208, "lr": 6.618448947479479e-05}
{"train_loss": 0.11875949800014496, "global_step": 107546, "epoch": 1208, "lr": 6.618394094823061e-05}
{"train_loss": 0.10323961824178696, "global_step": 107547, "epoch": 1208, "lr": 6.61833924194907e-05}
{"train_loss": 0.16675159335136414, "global_step": 107548, "epoch": 1208, "lr": 6.618284388857513e-05}
{"train_loss": 0.20872151851654053, "global_step": 107549, "epoch": 1208, "lr": 6.618229535548396e-05}
{"train_loss": 0.15810802578926086, "global_step": 107550, "epoch": 1208, "lr": 6.618174682021729e-05}
{"train_loss": 0.1354447901248932, "global_step": 107551, "epoch": 1208, "lr": 6.618119828277518e-05}
{"train_loss": 0.08619971573352814, "global_step": 107552, "epoch": 1208, "lr": 6.618064974315771e-05}
{"train_loss": 0.10155844688415527, "global_step": 107553, "epoch": 1208, "lr": 6.618010120136494e-05}
{"train_loss": 0.16864122450351715, "global_step": 107554, "epoch": 1208, "lr": 6.617955265739694e-05}
{"train_loss": 0.17789725959300995, "global_step": 107555, "epoch": 1208, "lr": 6.617900411125382e-05}
{"train_loss": 0.15040694177150726, "global_step": 107556, "epoch": 1208, "lr": 6.617845556293562e-05}
{"train_loss": 0.18803691864013672, "global_step": 107557, "epoch": 1208, "lr": 6.617790701244242e-05}
{"train_loss": 0.2123982012271881, "global_step": 107558, "epoch": 1208, "lr": 6.617735845977431e-05}
{"train_loss": 0.14516706764698029, "global_step": 107559, "epoch": 1208, "lr": 6.617680990493134e-05}
{"train_loss": 0.14999505877494812, "global_step": 107560, "epoch": 1208, "lr": 6.617626134791359e-05}
{"train_loss": 0.18522198498249054, "global_step": 107561, "epoch": 1208, "lr": 6.617571278872116e-05}
{"train_loss": 0.11706623435020447, "global_step": 107562, "epoch": 1208, "lr": 6.61751642273541e-05}
{"train_loss": 0.10895319283008575, "global_step": 107563, "epoch": 1208, "lr": 6.617461566381247e-05}
{"train_loss": 0.15958239138126373, "global_step": 107564, "epoch": 1208, "lr": 6.617406709809637e-05}
{"train_loss": 0.13945631682872772, "global_step": 107565, "epoch": 1208, "lr": 6.617351853020585e-05}
{"train_loss": 0.17911599576473236, "global_step": 107566, "epoch": 1208, "lr": 6.617296996014102e-05}
{"train_loss": 0.1853589117527008, "global_step": 107567, "epoch": 1208, "lr": 6.617242138790191e-05}
{"train_loss": 0.1243591234087944, "global_step": 107568, "epoch": 1208, "lr": 6.617187281348864e-05}
{"train_loss": 0.18694233894348145, "global_step": 107569, "epoch": 1208, "lr": 6.617132423690124e-05}
{"train_loss": 0.1304682046175003, "global_step": 107570, "epoch": 1208, "lr": 6.617077565813981e-05}
{"train_loss": 0.19151116907596588, "global_step": 107571, "epoch": 1208, "lr": 6.61702270772044e-05}
{"train_loss": 0.12887276709079742, "global_step": 107572, "epoch": 1208, "lr": 6.61696784940951e-05}
{"train_loss": 0.15303432941436768, "global_step": 107573, "epoch": 1208, "lr": 6.616912990881202e-05}
{"train_loss": 0.10142552107572556, "global_step": 107574, "epoch": 1208, "lr": 6.616858132135516e-05}
{"train_loss": 0.10403776913881302, "global_step": 107575, "epoch": 1208, "lr": 6.616803273172465e-05}
{"train_loss": 0.10081060975790024, "global_step": 107576, "epoch": 1208, "lr": 6.616748413992053e-05}
{"train_loss": 0.09271305054426193, "global_step": 107577, "epoch": 1208, "lr": 6.616693554594291e-05}
{"train_loss": 0.12420029193162918, "global_step": 107578, "epoch": 1208, "lr": 6.616638694979182e-05}
{"train_loss": 0.16931204497814178, "global_step": 107579, "epoch": 1208, "lr": 6.616583835146738e-05}
{"train_loss": 0.13849057257175446, "global_step": 107580, "epoch": 1208, "lr": 6.616528975096962e-05}
{"train_loss": 0.09477213025093079, "global_step": 107581, "epoch": 1208, "lr": 6.616474114829866e-05}
{"train_loss": 0.09046785533428192, "global_step": 107582, "epoch": 1208, "lr": 6.616419254345453e-05}
{"train_loss": 0.0893898531794548, "global_step": 107583, "epoch": 1208, "lr": 6.616364393643732e-05}
{"train_loss": 0.20243027806282043, "global_step": 107584, "epoch": 1208, "lr": 6.616309532724711e-05}
{"train_loss": 0.17089809477329254, "global_step": 107585, "epoch": 1208, "lr": 6.616254671588396e-05}
{"train_loss": 0.12418243288993835, "global_step": 107586, "epoch": 1208, "lr": 6.616199810234796e-05}
{"train_loss": 0.15431919693946838, "global_step": 107587, "epoch": 1208, "lr": 6.616144948663917e-05}
{"train_loss": 0.25378456711769104, "global_step": 107588, "epoch": 1208, "lr": 6.616090086875767e-05}
{"train_loss": 0.2027556300163269, "global_step": 107589, "epoch": 1208, "lr": 6.616035224870354e-05}
{"train_loss": 0.15755800902843475, "global_step": 107590, "epoch": 1208, "lr": 6.615980362647683e-05}
{"train_loss": 0.16776053607463837, "global_step": 107591, "epoch": 1208, "lr": 6.615925500207764e-05}
{"train_loss": 0.11151730269193649, "global_step": 107592, "epoch": 1208, "lr": 6.615870637550605e-05}
{"train_loss": 0.15155988931655884, "global_step": 107593, "epoch": 1208, "lr": 6.61581577467621e-05}
{"train_loss": 0.15186043083667755, "global_step": 107594, "epoch": 1208, "lr": 6.615760911584591e-05}
{"train_loss": 0.1476946622133255, "global_step": 107595, "epoch": 1208, "lr": 6.61570604827575e-05}
{"train_loss": 0.12668363749980927, "global_step": 107596, "epoch": 1208, "lr": 6.615651184749699e-05}
{"train_loss": 0.17749343812465668, "global_step": 107597, "epoch": 1208, "lr": 6.615596321006442e-05}
{"train_loss": 0.20496638119220734, "global_step": 107598, "epoch": 1208, "lr": 6.615541457045989e-05}
{"train_loss": 0.2554932236671448, "global_step": 107599, "epoch": 1208, "lr": 6.615486592868344e-05}
{"train_loss": 0.1509556005175194, "global_step": 107600, "epoch": 1208, "lr": 6.615431728473519e-05, "val_loss": 4.876830577850342}
{"train_loss": 0.19248805940151215, "global_step": 107601, "epoch": 1209, "lr": 6.615376863861517e-05}
{"train_loss": 0.1521296501159668, "global_step": 107602, "epoch": 1209, "lr": 6.615321999032349e-05}
{"train_loss": 0.16138562560081482, "global_step": 107603, "epoch": 1209, "lr": 6.61526713398602e-05}
{"train_loss": 0.10873495042324066, "global_step": 107604, "epoch": 1209, "lr": 6.615212268722537e-05}
{"train_loss": 0.14145348966121674, "global_step": 107605, "epoch": 1209, "lr": 6.615157403241909e-05}
{"train_loss": 0.16333262622356415, "global_step": 107606, "epoch": 1209, "lr": 6.615102537544142e-05}
{"train_loss": 0.09328188002109528, "global_step": 107607, "epoch": 1209, "lr": 6.615047671629245e-05}
{"train_loss": 0.10944761335849762, "global_step": 107608, "epoch": 1209, "lr": 6.614992805497225e-05}
{"train_loss": 0.2020028829574585, "global_step": 107609, "epoch": 1209, "lr": 6.614937939148087e-05}
{"train_loss": 0.20459449291229248, "global_step": 107610, "epoch": 1209, "lr": 6.614883072581842e-05}
{"train_loss": 0.12308720499277115, "global_step": 107611, "epoch": 1209, "lr": 6.614828205798495e-05}
{"train_loss": 0.1210121363401413, "global_step": 107612, "epoch": 1209, "lr": 6.614773338798054e-05}
{"train_loss": 0.120032399892807, "global_step": 107613, "epoch": 1209, "lr": 6.614718471580528e-05}
{"train_loss": 0.2644015848636627, "global_step": 107614, "epoch": 1209, "lr": 6.61466360414592e-05}
{"train_loss": 0.2185753583908081, "global_step": 107615, "epoch": 1209, "lr": 6.614608736494242e-05}
{"train_loss": 0.18292905390262604, "global_step": 107616, "epoch": 1209, "lr": 6.6145538686255e-05}
{"train_loss": 0.11165669560432434, "global_step": 107617, "epoch": 1209, "lr": 6.6144990005397e-05}
{"train_loss": 0.23348072171211243, "global_step": 107618, "epoch": 1209, "lr": 6.61444413223685e-05}
{"train_loss": 0.12925563752651215, "global_step": 107619, "epoch": 1209, "lr": 6.614389263716959e-05}
{"train_loss": 0.17381298542022705, "global_step": 107620, "epoch": 1209, "lr": 6.614334394980033e-05}
{"train_loss": 0.11147969216108322, "global_step": 107621, "epoch": 1209, "lr": 6.614279526026078e-05}
{"train_loss": 0.1090211272239685, "global_step": 107622, "epoch": 1209, "lr": 6.614224656855104e-05}
{"train_loss": 0.16639824211597443, "global_step": 107623, "epoch": 1209, "lr": 6.614169787467115e-05}
{"train_loss": 0.1068088635802269, "global_step": 107624, "epoch": 1209, "lr": 6.614114917862121e-05}
{"train_loss": 0.14265966415405273, "global_step": 107625, "epoch": 1209, "lr": 6.614060048040131e-05}
{"train_loss": 0.11799319088459015, "global_step": 107626, "epoch": 1209, "lr": 6.614005178001149e-05}
{"train_loss": 0.14019513130187988, "global_step": 107627, "epoch": 1209, "lr": 6.613950307745183e-05}
{"train_loss": 0.2588062286376953, "global_step": 107628, "epoch": 1209, "lr": 6.613895437272243e-05}
{"train_loss": 0.17956168949604034, "global_step": 107629, "epoch": 1209, "lr": 6.613840566582333e-05}
{"train_loss": 0.20110630989074707, "global_step": 107630, "epoch": 1209, "lr": 6.613785695675463e-05}
{"train_loss": 0.16589656472206116, "global_step": 107631, "epoch": 1209, "lr": 6.613730824551639e-05}
{"train_loss": 0.19158993661403656, "global_step": 107632, "epoch": 1209, "lr": 6.613675953210867e-05}
{"train_loss": 0.18628549575805664, "global_step": 107633, "epoch": 1209, "lr": 6.613621081653157e-05}
{"train_loss": 0.09299835562705994, "global_step": 107634, "epoch": 1209, "lr": 6.613566209878514e-05}
{"train_loss": 0.14842644333839417, "global_step": 107635, "epoch": 1209, "lr": 6.613511337886949e-05}
{"train_loss": 0.1328633725643158, "global_step": 107636, "epoch": 1209, "lr": 6.613456465678465e-05}
{"train_loss": 0.10413381457328796, "global_step": 107637, "epoch": 1209, "lr": 6.613401593253073e-05}
{"train_loss": 0.14771603047847748, "global_step": 107638, "epoch": 1209, "lr": 6.613346720610778e-05}
{"train_loss": 0.15341520309448242, "global_step": 107639, "epoch": 1209, "lr": 6.613291847751591e-05}
{"train_loss": 0.16793644428253174, "global_step": 107640, "epoch": 1209, "lr": 6.613236974675513e-05}
{"train_loss": 0.10300987958908081, "global_step": 107641, "epoch": 1209, "lr": 6.613182101382556e-05}
{"train_loss": 0.09992510080337524, "global_step": 107642, "epoch": 1209, "lr": 6.613127227872727e-05}
{"train_loss": 0.08463849872350693, "global_step": 107643, "epoch": 1209, "lr": 6.61307235414603e-05}
{"train_loss": 0.16851156949996948, "global_step": 107644, "epoch": 1209, "lr": 6.613017480202478e-05}
{"train_loss": 0.2064565122127533, "global_step": 107645, "epoch": 1209, "lr": 6.612962606042074e-05}
{"train_loss": 0.1118992492556572, "global_step": 107646, "epoch": 1209, "lr": 6.612907731664827e-05}
{"train_loss": 0.1727006733417511, "global_step": 107647, "epoch": 1209, "lr": 6.612852857070744e-05}
{"train_loss": 0.13453969359397888, "global_step": 107648, "epoch": 1209, "lr": 6.612797982259833e-05}
{"train_loss": 0.13288980722427368, "global_step": 107649, "epoch": 1209, "lr": 6.612743107232103e-05}
{"train_loss": 0.20316459238529205, "global_step": 107650, "epoch": 1209, "lr": 6.612688231987556e-05}
{"train_loss": 0.10358862578868866, "global_step": 107651, "epoch": 1209, "lr": 6.612633356526204e-05}
{"train_loss": 0.14395858347415924, "global_step": 107652, "epoch": 1209, "lr": 6.612578480848054e-05}
{"train_loss": 0.17206540703773499, "global_step": 107653, "epoch": 1209, "lr": 6.612523604953111e-05}
{"train_loss": 0.2297215461730957, "global_step": 107654, "epoch": 1209, "lr": 6.612468728841385e-05}
{"train_loss": 0.13858085870742798, "global_step": 107655, "epoch": 1209, "lr": 6.612413852512882e-05}
{"train_loss": 0.11776899546384811, "global_step": 107656, "epoch": 1209, "lr": 6.61235897596761e-05}
{"train_loss": 0.16399145126342773, "global_step": 107657, "epoch": 1209, "lr": 6.612304099205576e-05}
{"train_loss": 0.11669936031103134, "global_step": 107658, "epoch": 1209, "lr": 6.612249222226787e-05}
{"train_loss": 0.09588668495416641, "global_step": 107659, "epoch": 1209, "lr": 6.61219434503125e-05}
{"train_loss": 0.13417957723140717, "global_step": 107660, "epoch": 1209, "lr": 6.612139467618975e-05}
{"train_loss": 0.1455816626548767, "global_step": 107661, "epoch": 1209, "lr": 6.612084589989966e-05}
{"train_loss": 0.2270790934562683, "global_step": 107662, "epoch": 1209, "lr": 6.61202971214423e-05}
{"train_loss": 0.14516694843769073, "global_step": 107663, "epoch": 1209, "lr": 6.611974834081779e-05}
{"train_loss": 0.14818504452705383, "global_step": 107664, "epoch": 1209, "lr": 6.611919955802617e-05}
{"train_loss": 0.1811046302318573, "global_step": 107665, "epoch": 1209, "lr": 6.611865077306752e-05}
{"train_loss": 0.14814940094947815, "global_step": 107666, "epoch": 1209, "lr": 6.61181019859419e-05}
{"train_loss": 0.19740116596221924, "global_step": 107667, "epoch": 1209, "lr": 6.611755319664942e-05}
{"train_loss": 0.14672453701496124, "global_step": 107668, "epoch": 1209, "lr": 6.611700440519012e-05}
{"train_loss": 0.11050014197826385, "global_step": 107669, "epoch": 1209, "lr": 6.611645561156409e-05}
{"train_loss": 0.1432419717311859, "global_step": 107670, "epoch": 1209, "lr": 6.61159068157714e-05}
{"train_loss": 0.14384959638118744, "global_step": 107671, "epoch": 1209, "lr": 6.61153580178121e-05}
{"train_loss": 0.2301962971687317, "global_step": 107672, "epoch": 1209, "lr": 6.611480921768632e-05}
{"train_loss": 0.18575067818164825, "global_step": 107673, "epoch": 1209, "lr": 6.61142604153941e-05}
{"train_loss": 0.11739304661750793, "global_step": 107674, "epoch": 1209, "lr": 6.61137116109355e-05}
{"train_loss": 0.19849219918251038, "global_step": 107675, "epoch": 1209, "lr": 6.61131628043106e-05}
{"train_loss": 0.12969017028808594, "global_step": 107676, "epoch": 1209, "lr": 6.611261399551949e-05}
{"train_loss": 0.1674565076828003, "global_step": 107677, "epoch": 1209, "lr": 6.611206518456225e-05}
{"train_loss": 0.14458000659942627, "global_step": 107678, "epoch": 1209, "lr": 6.611151637143893e-05}
{"train_loss": 0.1812833696603775, "global_step": 107679, "epoch": 1209, "lr": 6.61109675561496e-05}
{"train_loss": 0.23053422570228577, "global_step": 107680, "epoch": 1209, "lr": 6.611041873869433e-05}
{"train_loss": 0.20002977550029755, "global_step": 107681, "epoch": 1209, "lr": 6.610986991907325e-05}
{"train_loss": 0.1622214913368225, "global_step": 107682, "epoch": 1209, "lr": 6.610932109728637e-05}
{"train_loss": 0.09985457360744476, "global_step": 107683, "epoch": 1209, "lr": 6.61087722733338e-05}
{"train_loss": 0.15114860236644745, "global_step": 107684, "epoch": 1209, "lr": 6.61082234472156e-05}
{"train_loss": 0.12321866303682327, "global_step": 107685, "epoch": 1209, "lr": 6.610767461893185e-05}
{"train_loss": 0.12268775701522827, "global_step": 107686, "epoch": 1209, "lr": 6.61071257884826e-05}
{"train_loss": 0.17867127060890198, "global_step": 107687, "epoch": 1209, "lr": 6.610657695586798e-05}
{"train_loss": 0.128696009516716, "global_step": 107688, "epoch": 1209, "lr": 6.6106028121088e-05}
{"train_loss": 0.15356237642215878, "global_step": 107689, "epoch": 1209, "lr": 6.610547928414276e-05, "val_loss": 4.884198188781738}
{"train_loss": 0.1553642749786377, "global_step": 107690, "epoch": 1210, "lr": 6.610493044503234e-05}
{"train_loss": 0.17827558517456055, "global_step": 107691, "epoch": 1210, "lr": 6.61043816037568e-05}
{"train_loss": 0.09992384910583496, "global_step": 107692, "epoch": 1210, "lr": 6.610383276031624e-05}
{"train_loss": 0.14530766010284424, "global_step": 107693, "epoch": 1210, "lr": 6.61032839147107e-05}
{"train_loss": 0.17228057980537415, "global_step": 107694, "epoch": 1210, "lr": 6.610273506694028e-05}
{"train_loss": 0.14633609354496002, "global_step": 107695, "epoch": 1210, "lr": 6.610218621700503e-05}
{"train_loss": 0.3019839823246002, "global_step": 107696, "epoch": 1210, "lr": 6.610163736490506e-05}
{"train_loss": 0.13378188014030457, "global_step": 107697, "epoch": 1210, "lr": 6.610108851064042e-05}
{"train_loss": 0.13900257647037506, "global_step": 107698, "epoch": 1210, "lr": 6.610053965421116e-05}
{"train_loss": 0.12027065455913544, "global_step": 107699, "epoch": 1210, "lr": 6.609999079561739e-05}
{"train_loss": 0.19679012894630432, "global_step": 107700, "epoch": 1210, "lr": 6.609944193485917e-05}
{"train_loss": 0.23729994893074036, "global_step": 107701, "epoch": 1210, "lr": 6.609889307193658e-05}
{"train_loss": 0.1828993707895279, "global_step": 107702, "epoch": 1210, "lr": 6.60983442068497e-05}
{"train_loss": 0.21887382864952087, "global_step": 107703, "epoch": 1210, "lr": 6.609779533959858e-05}
{"train_loss": 0.1047634705901146, "global_step": 107704, "epoch": 1210, "lr": 6.609724647018332e-05}
{"train_loss": 0.15002106130123138, "global_step": 107705, "epoch": 1210, "lr": 6.609669759860396e-05}
{"train_loss": 0.11040741205215454, "global_step": 107706, "epoch": 1210, "lr": 6.609614872486062e-05}
{"train_loss": 0.1493002325296402, "global_step": 107707, "epoch": 1210, "lr": 6.609559984895333e-05}
{"train_loss": 0.07102963328361511, "global_step": 107708, "epoch": 1210, "lr": 6.609505097088219e-05}
{"train_loss": 0.15438364446163177, "global_step": 107709, "epoch": 1210, "lr": 6.609450209064726e-05}
{"train_loss": 0.1406012922525406, "global_step": 107710, "epoch": 1210, "lr": 6.609395320824864e-05}
{"train_loss": 0.10778379440307617, "global_step": 107711, "epoch": 1210, "lr": 6.609340432368637e-05}
{"train_loss": 0.11423951387405396, "global_step": 107712, "epoch": 1210, "lr": 6.609285543696053e-05}
{"train_loss": 0.13898645341396332, "global_step": 107713, "epoch": 1210, "lr": 6.609230654807122e-05}
{"train_loss": 0.11618594825267792, "global_step": 107714, "epoch": 1210, "lr": 6.609175765701848e-05}
{"train_loss": 0.15040642023086548, "global_step": 107715, "epoch": 1210, "lr": 6.609120876380242e-05}
{"train_loss": 0.15541383624076843, "global_step": 107716, "epoch": 1210, "lr": 6.609065986842309e-05}
{"train_loss": 0.0942450612783432, "global_step": 107717, "epoch": 1210, "lr": 6.609011097088054e-05}
{"train_loss": 0.16307461261749268, "global_step": 107718, "epoch": 1210, "lr": 6.608956207117488e-05}
{"train_loss": 0.12981972098350525, "global_step": 107719, "epoch": 1210, "lr": 6.608901316930618e-05}
{"train_loss": 0.14737333357334137, "global_step": 107720, "epoch": 1210, "lr": 6.608846426527452e-05}
{"train_loss": 0.2083905041217804, "global_step": 107721, "epoch": 1210, "lr": 6.608791535907995e-05}
{"train_loss": 0.1187480241060257, "global_step": 107722, "epoch": 1210, "lr": 6.608736645072256e-05}
{"train_loss": 0.18768426775932312, "global_step": 107723, "epoch": 1210, "lr": 6.608681754020241e-05}
{"train_loss": 0.17561031877994537, "global_step": 107724, "epoch": 1210, "lr": 6.608626862751959e-05}
{"train_loss": 0.07217193394899368, "global_step": 107725, "epoch": 1210, "lr": 6.608571971267417e-05}
{"train_loss": 0.14000937342643738, "global_step": 107726, "epoch": 1210, "lr": 6.608517079566621e-05}
{"train_loss": 0.12988172471523285, "global_step": 107727, "epoch": 1210, "lr": 6.608462187649581e-05}
{"train_loss": 0.09325705468654633, "global_step": 107728, "epoch": 1210, "lr": 6.608407295516303e-05}
{"train_loss": 0.1914985477924347, "global_step": 107729, "epoch": 1210, "lr": 6.608352403166794e-05}
{"train_loss": 0.1763216257095337, "global_step": 107730, "epoch": 1210, "lr": 6.608297510601061e-05}
{"train_loss": 0.1342049390077591, "global_step": 107731, "epoch": 1210, "lr": 6.608242617819112e-05}
{"train_loss": 0.12580648064613342, "global_step": 107732, "epoch": 1210, "lr": 6.608187724820953e-05}
{"train_loss": 0.158760204911232, "global_step": 107733, "epoch": 1210, "lr": 6.608132831606595e-05}
{"train_loss": 0.2753356695175171, "global_step": 107734, "epoch": 1210, "lr": 6.608077938176043e-05}
{"train_loss": 0.14835351705551147, "global_step": 107735, "epoch": 1210, "lr": 6.608023044529303e-05}
{"train_loss": 0.24303342401981354, "global_step": 107736, "epoch": 1210, "lr": 6.607968150666385e-05}
{"train_loss": 0.15807513892650604, "global_step": 107737, "epoch": 1210, "lr": 6.607913256587295e-05}
{"train_loss": 0.11667829751968384, "global_step": 107738, "epoch": 1210, "lr": 6.60785836229204e-05}
{"train_loss": 0.11751066893339157, "global_step": 107739, "epoch": 1210, "lr": 6.60780346778063e-05}
{"train_loss": 0.18275506794452667, "global_step": 107740, "epoch": 1210, "lr": 6.607748573053069e-05}
{"train_loss": 0.17982035875320435, "global_step": 107741, "epoch": 1210, "lr": 6.607693678109365e-05}
{"train_loss": 0.10980567336082458, "global_step": 107742, "epoch": 1210, "lr": 6.607638782949528e-05}
{"train_loss": 0.12225423008203506, "global_step": 107743, "epoch": 1210, "lr": 6.607583887573563e-05}
{"train_loss": 0.17290937900543213, "global_step": 107744, "epoch": 1210, "lr": 6.607528991981478e-05}
{"train_loss": 0.17221960425376892, "global_step": 107745, "epoch": 1210, "lr": 6.60747409617328e-05}
{"train_loss": 0.08351676911115646, "global_step": 107746, "epoch": 1210, "lr": 6.607419200148977e-05}
{"train_loss": 0.21615776419639587, "global_step": 107747, "epoch": 1210, "lr": 6.607364303908575e-05}
{"train_loss": 0.09202268719673157, "global_step": 107748, "epoch": 1210, "lr": 6.607309407452084e-05}
{"train_loss": 0.09003204852342606, "global_step": 107749, "epoch": 1210, "lr": 6.607254510779508e-05}
{"train_loss": 0.17120830714702606, "global_step": 107750, "epoch": 1210, "lr": 6.607199613890858e-05}
{"train_loss": 0.1542840451002121, "global_step": 107751, "epoch": 1210, "lr": 6.607144716786139e-05}
{"train_loss": 0.23987290263175964, "global_step": 107752, "epoch": 1210, "lr": 6.607089819465359e-05}
{"train_loss": 0.2277858406305313, "global_step": 107753, "epoch": 1210, "lr": 6.607034921928525e-05}
{"train_loss": 0.11481870710849762, "global_step": 107754, "epoch": 1210, "lr": 6.606980024175645e-05}
{"train_loss": 0.18520691990852356, "global_step": 107755, "epoch": 1210, "lr": 6.606925126206727e-05}
{"train_loss": 0.12742021679878235, "global_step": 107756, "epoch": 1210, "lr": 6.606870228021775e-05}
{"train_loss": 0.11268089711666107, "global_step": 107757, "epoch": 1210, "lr": 6.606815329620801e-05}
{"train_loss": 0.09174297749996185, "global_step": 107758, "epoch": 1210, "lr": 6.60676043100381e-05}
{"train_loss": 0.20037049055099487, "global_step": 107759, "epoch": 1210, "lr": 6.606705532170809e-05}
{"train_loss": 0.1427668184041977, "global_step": 107760, "epoch": 1210, "lr": 6.606650633121807e-05}
{"train_loss": 0.09530231356620789, "global_step": 107761, "epoch": 1210, "lr": 6.606595733856809e-05}
{"train_loss": 0.10246508568525314, "global_step": 107762, "epoch": 1210, "lr": 6.606540834375825e-05}
{"train_loss": 0.18220409750938416, "global_step": 107763, "epoch": 1210, "lr": 6.60648593467886e-05}
{"train_loss": 0.14328745007514954, "global_step": 107764, "epoch": 1210, "lr": 6.606431034765924e-05}
{"train_loss": 0.16816872358322144, "global_step": 107765, "epoch": 1210, "lr": 6.606376134637022e-05}
{"train_loss": 0.17336556315422058, "global_step": 107766, "epoch": 1210, "lr": 6.606321234292163e-05}
{"train_loss": 0.14549219608306885, "global_step": 107767, "epoch": 1210, "lr": 6.606266333731354e-05}
{"train_loss": 0.1502663791179657, "global_step": 107768, "epoch": 1210, "lr": 6.606211432954601e-05}
{"train_loss": 0.13336646556854248, "global_step": 107769, "epoch": 1210, "lr": 6.606156531961912e-05}
{"train_loss": 0.17880915105342865, "global_step": 107770, "epoch": 1210, "lr": 6.606101630753296e-05}
{"train_loss": 0.11374764144420624, "global_step": 107771, "epoch": 1210, "lr": 6.606046729328759e-05}
{"train_loss": 0.15566861629486084, "global_step": 107772, "epoch": 1210, "lr": 6.60599182768831e-05}
{"train_loss": 0.16041891276836395, "global_step": 107773, "epoch": 1210, "lr": 6.605936925831953e-05}
{"train_loss": 0.1452123075723648, "global_step": 107774, "epoch": 1210, "lr": 6.605882023759697e-05}
{"train_loss": 0.22538557648658752, "global_step": 107775, "epoch": 1210, "lr": 6.60582712147155e-05}
{"train_loss": 0.11285820603370667, "global_step": 107776, "epoch": 1210, "lr": 6.60577221896752e-05}
{"train_loss": 0.12672710418701172, "global_step": 107777, "epoch": 1210, "lr": 6.605717316247613e-05}
{"train_loss": 0.15164314888501434, "global_step": 107778, "epoch": 1210, "lr": 6.605662413311838e-05, "val_loss": 4.852708339691162, "train_action_mse_error": 18.115169525146484}
{"train_loss": 0.12037524580955505, "global_step": 107779, "epoch": 1211, "lr": 6.605607510160199e-05}
{"train_loss": 0.08611997961997986, "global_step": 107780, "epoch": 1211, "lr": 6.605552606792707e-05}
{"train_loss": 0.11852797120809555, "global_step": 107781, "epoch": 1211, "lr": 6.605497703209367e-05}
{"train_loss": 0.20466209948062897, "global_step": 107782, "epoch": 1211, "lr": 6.605442799410188e-05}
{"train_loss": 0.09127562493085861, "global_step": 107783, "epoch": 1211, "lr": 6.605387895395179e-05}
{"train_loss": 0.1633080244064331, "global_step": 107784, "epoch": 1211, "lr": 6.605332991164343e-05}
{"train_loss": 0.12765873968601227, "global_step": 107785, "epoch": 1211, "lr": 6.605278086717689e-05}
{"train_loss": 0.1853802353143692, "global_step": 107786, "epoch": 1211, "lr": 6.605223182055225e-05}
{"train_loss": 0.1053171381354332, "global_step": 107787, "epoch": 1211, "lr": 6.605168277176959e-05}
{"train_loss": 0.0741317868232727, "global_step": 107788, "epoch": 1211, "lr": 6.605113372082899e-05}
{"train_loss": 0.25150546431541443, "global_step": 107789, "epoch": 1211, "lr": 6.60505846677305e-05}
{"train_loss": 0.09668458998203278, "global_step": 107790, "epoch": 1211, "lr": 6.60500356124742e-05}
{"train_loss": 0.11270824074745178, "global_step": 107791, "epoch": 1211, "lr": 6.604948655506018e-05}
{"train_loss": 0.18136745691299438, "global_step": 107792, "epoch": 1211, "lr": 6.604893749548848e-05}
{"train_loss": 0.08198806643486023, "global_step": 107793, "epoch": 1211, "lr": 6.604838843375921e-05}
{"train_loss": 0.19168192148208618, "global_step": 107794, "epoch": 1211, "lr": 6.604783936987242e-05}
{"train_loss": 0.10877573490142822, "global_step": 107795, "epoch": 1211, "lr": 6.604729030382822e-05}
{"train_loss": 0.207887664437294, "global_step": 107796, "epoch": 1211, "lr": 6.604674123562663e-05}
{"train_loss": 0.15365314483642578, "global_step": 107797, "epoch": 1211, "lr": 6.604619216526776e-05}
{"train_loss": 0.10750680416822433, "global_step": 107798, "epoch": 1211, "lr": 6.604564309275168e-05}
{"train_loss": 0.1822950690984726, "global_step": 107799, "epoch": 1211, "lr": 6.604509401807846e-05}
{"train_loss": 0.15083980560302734, "global_step": 107800, "epoch": 1211, "lr": 6.604454494124817e-05}
{"train_loss": 0.22671599686145782, "global_step": 107801, "epoch": 1211, "lr": 6.604399586226086e-05}
{"train_loss": 0.14174428582191467, "global_step": 107802, "epoch": 1211, "lr": 6.604344678111666e-05}
{"train_loss": 0.16746190190315247, "global_step": 107803, "epoch": 1211, "lr": 6.604289769781562e-05}
{"train_loss": 0.20034196972846985, "global_step": 107804, "epoch": 1211, "lr": 6.604234861235779e-05}
{"train_loss": 0.14535321295261383, "global_step": 107805, "epoch": 1211, "lr": 6.604179952474327e-05}
{"train_loss": 0.17192624509334564, "global_step": 107806, "epoch": 1211, "lr": 6.604125043497213e-05}
{"train_loss": 0.19382144510746002, "global_step": 107807, "epoch": 1211, "lr": 6.604070134304443e-05}
{"train_loss": 0.22795355319976807, "global_step": 107808, "epoch": 1211, "lr": 6.604015224896025e-05}
{"train_loss": 0.09411559253931046, "global_step": 107809, "epoch": 1211, "lr": 6.603960315271967e-05}
{"train_loss": 0.16219079494476318, "global_step": 107810, "epoch": 1211, "lr": 6.603905405432277e-05}
{"train_loss": 0.12555791437625885, "global_step": 107811, "epoch": 1211, "lr": 6.603850495376961e-05}
{"train_loss": 0.11660098284482956, "global_step": 107812, "epoch": 1211, "lr": 6.603795585106027e-05}
{"train_loss": 0.11374692618846893, "global_step": 107813, "epoch": 1211, "lr": 6.603740674619481e-05}
{"train_loss": 0.17487174272537231, "global_step": 107814, "epoch": 1211, "lr": 6.603685763917333e-05}
{"train_loss": 0.1727905124425888, "global_step": 107815, "epoch": 1211, "lr": 6.603630852999588e-05}
{"train_loss": 0.16486619412899017, "global_step": 107816, "epoch": 1211, "lr": 6.603575941866256e-05}
{"train_loss": 0.14182405173778534, "global_step": 107817, "epoch": 1211, "lr": 6.603521030517341e-05}
{"train_loss": 0.17856214940547943, "global_step": 107818, "epoch": 1211, "lr": 6.603466118952853e-05}
{"train_loss": 0.10294418781995773, "global_step": 107819, "epoch": 1211, "lr": 6.603411207172798e-05}
{"train_loss": 0.13727428019046783, "global_step": 107820, "epoch": 1211, "lr": 6.603356295177184e-05}
{"train_loss": 0.17389966547489166, "global_step": 107821, "epoch": 1211, "lr": 6.603301382966019e-05}
{"train_loss": 0.1647365242242813, "global_step": 107822, "epoch": 1211, "lr": 6.60324647053931e-05}
{"train_loss": 0.2462126910686493, "global_step": 107823, "epoch": 1211, "lr": 6.603191557897062e-05}
{"train_loss": 0.11540566384792328, "global_step": 107824, "epoch": 1211, "lr": 6.603136645039286e-05}
{"train_loss": 0.17393849790096283, "global_step": 107825, "epoch": 1211, "lr": 6.603081731965987e-05}
{"train_loss": 0.16522946953773499, "global_step": 107826, "epoch": 1211, "lr": 6.603026818677174e-05}
{"train_loss": 0.13105402886867523, "global_step": 107827, "epoch": 1211, "lr": 6.602971905172855e-05}
{"train_loss": 0.19851699471473694, "global_step": 107828, "epoch": 1211, "lr": 6.602916991453034e-05}
{"train_loss": 0.1729533076286316, "global_step": 107829, "epoch": 1211, "lr": 6.602862077517719e-05}
{"train_loss": 0.14028598368167877, "global_step": 107830, "epoch": 1211, "lr": 6.602807163366921e-05}
{"train_loss": 0.21736744046211243, "global_step": 107831, "epoch": 1211, "lr": 6.602752249000644e-05}
{"train_loss": 0.19114108383655548, "global_step": 107832, "epoch": 1211, "lr": 6.602697334418897e-05}
{"train_loss": 0.15427105128765106, "global_step": 107833, "epoch": 1211, "lr": 6.602642419621687e-05}
{"train_loss": 0.1802850067615509, "global_step": 107834, "epoch": 1211, "lr": 6.602587504609021e-05}
{"train_loss": 0.19801177084445953, "global_step": 107835, "epoch": 1211, "lr": 6.602532589380907e-05}
{"train_loss": 0.2100023627281189, "global_step": 107836, "epoch": 1211, "lr": 6.602477673937353e-05}
{"train_loss": 0.1508546620607376, "global_step": 107837, "epoch": 1211, "lr": 6.602422758278364e-05}
{"train_loss": 0.17454388737678528, "global_step": 107838, "epoch": 1211, "lr": 6.602367842403948e-05}
{"train_loss": 0.21542565524578094, "global_step": 107839, "epoch": 1211, "lr": 6.602312926314115e-05}
{"train_loss": 0.14713642001152039, "global_step": 107840, "epoch": 1211, "lr": 6.602258010008869e-05}
{"train_loss": 0.18016527593135834, "global_step": 107841, "epoch": 1211, "lr": 6.602203093488219e-05}
{"train_loss": 0.139531210064888, "global_step": 107842, "epoch": 1211, "lr": 6.602148176752175e-05}
{"train_loss": 0.1882009208202362, "global_step": 107843, "epoch": 1211, "lr": 6.602093259800739e-05}
{"train_loss": 0.17280712723731995, "global_step": 107844, "epoch": 1211, "lr": 6.60203834263392e-05}
{"train_loss": 0.08283941447734833, "global_step": 107845, "epoch": 1211, "lr": 6.601983425251728e-05}
{"train_loss": 0.1882353127002716, "global_step": 107846, "epoch": 1211, "lr": 6.60192850765417e-05}
{"train_loss": 0.18571637570858002, "global_step": 107847, "epoch": 1211, "lr": 6.601873589841252e-05}
{"train_loss": 0.18587537109851837, "global_step": 107848, "epoch": 1211, "lr": 6.60181867181298e-05}
{"train_loss": 0.2215479612350464, "global_step": 107849, "epoch": 1211, "lr": 6.601763753569365e-05}
{"train_loss": 0.11987128853797913, "global_step": 107850, "epoch": 1211, "lr": 6.60170883511041e-05}
{"train_loss": 0.18001390993595123, "global_step": 107851, "epoch": 1211, "lr": 6.601653916436127e-05}
{"train_loss": 0.16824975609779358, "global_step": 107852, "epoch": 1211, "lr": 6.601598997546521e-05}
{"train_loss": 0.11665906012058258, "global_step": 107853, "epoch": 1211, "lr": 6.601544078441598e-05}
{"train_loss": 0.15784718096256256, "global_step": 107854, "epoch": 1211, "lr": 6.601489159121368e-05}
{"train_loss": 0.1796105057001114, "global_step": 107855, "epoch": 1211, "lr": 6.601434239585837e-05}
{"train_loss": 0.10577542334794998, "global_step": 107856, "epoch": 1211, "lr": 6.601379319835012e-05}
{"train_loss": 0.15163590013980865, "global_step": 107857, "epoch": 1211, "lr": 6.601324399868902e-05}
{"train_loss": 0.22283387184143066, "global_step": 107858, "epoch": 1211, "lr": 6.601269479687514e-05}
{"train_loss": 0.15666162967681885, "global_step": 107859, "epoch": 1211, "lr": 6.601214559290854e-05}
{"train_loss": 0.2534179985523224, "global_step": 107860, "epoch": 1211, "lr": 6.60115963867893e-05}
{"train_loss": 0.13046027719974518, "global_step": 107861, "epoch": 1211, "lr": 6.60110471785175e-05}
{"train_loss": 0.1348894089460373, "global_step": 107862, "epoch": 1211, "lr": 6.601049796809319e-05}
{"train_loss": 0.1646801382303238, "global_step": 107863, "epoch": 1211, "lr": 6.600994875551649e-05}
{"train_loss": 0.10603566467761993, "global_step": 107864, "epoch": 1211, "lr": 6.600939954078744e-05}
{"train_loss": 0.11734144389629364, "global_step": 107865, "epoch": 1211, "lr": 6.600885032390612e-05}
{"train_loss": 0.1631479561328888, "global_step": 107866, "epoch": 1211, "lr": 6.600830110487261e-05}
{"train_loss": 0.15815787276860033, "global_step": 107867, "epoch": 1211, "lr": 6.600775188368698e-05, "val_loss": 4.825293064117432}
{"train_loss": 0.11923706531524658, "global_step": 107868, "epoch": 1212, "lr": 6.60072026603493e-05}
{"train_loss": 0.09740262478590012, "global_step": 107869, "epoch": 1212, "lr": 6.600665343485964e-05}
{"train_loss": 0.17970865964889526, "global_step": 107870, "epoch": 1212, "lr": 6.60061042072181e-05}
{"train_loss": 0.10229792445898056, "global_step": 107871, "epoch": 1212, "lr": 6.600555497742471e-05}
{"train_loss": 0.17792487144470215, "global_step": 107872, "epoch": 1212, "lr": 6.600500574547957e-05}
{"train_loss": 0.23083597421646118, "global_step": 107873, "epoch": 1212, "lr": 6.600445651138276e-05}
{"train_loss": 0.1763143390417099, "global_step": 107874, "epoch": 1212, "lr": 6.600390727513435e-05}
{"train_loss": 0.10101048648357391, "global_step": 107875, "epoch": 1212, "lr": 6.600335803673441e-05}
{"train_loss": 0.17523092031478882, "global_step": 107876, "epoch": 1212, "lr": 6.6002808796183e-05}
{"train_loss": 0.181596577167511, "global_step": 107877, "epoch": 1212, "lr": 6.600225955348023e-05}
{"train_loss": 0.1306658387184143, "global_step": 107878, "epoch": 1212, "lr": 6.600171030862613e-05}
{"train_loss": 0.12128197401762009, "global_step": 107879, "epoch": 1212, "lr": 6.60011610616208e-05}
{"train_loss": 0.1393665373325348, "global_step": 107880, "epoch": 1212, "lr": 6.60006118124643e-05}
{"train_loss": 0.12941744923591614, "global_step": 107881, "epoch": 1212, "lr": 6.600006256115673e-05}
{"train_loss": 0.1668064147233963, "global_step": 107882, "epoch": 1212, "lr": 6.599951330769813e-05}
{"train_loss": 0.17760254442691803, "global_step": 107883, "epoch": 1212, "lr": 6.599896405208861e-05}
{"train_loss": 0.1566634476184845, "global_step": 107884, "epoch": 1212, "lr": 6.599841479432823e-05}
{"train_loss": 0.17028523981571198, "global_step": 107885, "epoch": 1212, "lr": 6.599786553441705e-05}
{"train_loss": 0.162443146109581, "global_step": 107886, "epoch": 1212, "lr": 6.599731627235513e-05}
{"train_loss": 0.13889387249946594, "global_step": 107887, "epoch": 1212, "lr": 6.59967670081426e-05}
{"train_loss": 0.11185234785079956, "global_step": 107888, "epoch": 1212, "lr": 6.599621774177948e-05}
{"train_loss": 0.15414637327194214, "global_step": 107889, "epoch": 1212, "lr": 6.599566847326586e-05}
{"train_loss": 0.22576966881752014, "global_step": 107890, "epoch": 1212, "lr": 6.599511920260182e-05}
{"train_loss": 0.2231469303369522, "global_step": 107891, "epoch": 1212, "lr": 6.599456992978742e-05}
{"train_loss": 0.12894056737422943, "global_step": 107892, "epoch": 1212, "lr": 6.599402065482277e-05}
{"train_loss": 0.17767368257045746, "global_step": 107893, "epoch": 1212, "lr": 6.599347137770791e-05}
{"train_loss": 0.17301024496555328, "global_step": 107894, "epoch": 1212, "lr": 6.599292209844292e-05}
{"train_loss": 0.22105486690998077, "global_step": 107895, "epoch": 1212, "lr": 6.599237281702787e-05}
{"train_loss": 0.07136886566877365, "global_step": 107896, "epoch": 1212, "lr": 6.599182353346284e-05}
{"train_loss": 0.23733338713645935, "global_step": 107897, "epoch": 1212, "lr": 6.599127424774792e-05}
{"train_loss": 0.17639751732349396, "global_step": 107898, "epoch": 1212, "lr": 6.599072495988315e-05}
{"train_loss": 0.10928180813789368, "global_step": 107899, "epoch": 1212, "lr": 6.599017566986864e-05}
{"train_loss": 0.09784896671772003, "global_step": 107900, "epoch": 1212, "lr": 6.598962637770443e-05}
{"train_loss": 0.10823511332273483, "global_step": 107901, "epoch": 1212, "lr": 6.598907708339061e-05}
{"train_loss": 0.12938837707042694, "global_step": 107902, "epoch": 1212, "lr": 6.598852778692727e-05}
{"train_loss": 0.15860427916049957, "global_step": 107903, "epoch": 1212, "lr": 6.598797848831447e-05}
{"train_loss": 0.16091738641262054, "global_step": 107904, "epoch": 1212, "lr": 6.598742918755227e-05}
{"train_loss": 0.10077140480279922, "global_step": 107905, "epoch": 1212, "lr": 6.598687988464075e-05}
{"train_loss": 0.20223604142665863, "global_step": 107906, "epoch": 1212, "lr": 6.598633057957999e-05}
{"train_loss": 0.08176787197589874, "global_step": 107907, "epoch": 1212, "lr": 6.598578127237007e-05}
{"train_loss": 0.1657187044620514, "global_step": 107908, "epoch": 1212, "lr": 6.598523196301105e-05}
{"train_loss": 0.1551637053489685, "global_step": 107909, "epoch": 1212, "lr": 6.598468265150302e-05}
{"train_loss": 0.19841252267360687, "global_step": 107910, "epoch": 1212, "lr": 6.598413333784604e-05}
{"train_loss": 0.1248522624373436, "global_step": 107911, "epoch": 1212, "lr": 6.598358402204017e-05}
{"train_loss": 0.10981781780719757, "global_step": 107912, "epoch": 1212, "lr": 6.598303470408551e-05}
{"train_loss": 0.23210549354553223, "global_step": 107913, "epoch": 1212, "lr": 6.598248538398215e-05}
{"train_loss": 0.2103017419576645, "global_step": 107914, "epoch": 1212, "lr": 6.598193606173011e-05}
{"train_loss": 0.20682229101657867, "global_step": 107915, "epoch": 1212, "lr": 6.598138673732952e-05}
{"train_loss": 0.09776081889867783, "global_step": 107916, "epoch": 1212, "lr": 6.598083741078038e-05}
{"train_loss": 0.15298773348331451, "global_step": 107917, "epoch": 1212, "lr": 6.598028808208285e-05}
{"train_loss": 0.14920085668563843, "global_step": 107918, "epoch": 1212, "lr": 6.597973875123697e-05}
{"train_loss": 0.19530987739562988, "global_step": 107919, "epoch": 1212, "lr": 6.59791894182428e-05}
{"train_loss": 0.19777841866016388, "global_step": 107920, "epoch": 1212, "lr": 6.597864008310041e-05}
{"train_loss": 0.12650556862354279, "global_step": 107921, "epoch": 1212, "lr": 6.59780907458099e-05}
{"train_loss": 0.1205662339925766, "global_step": 107922, "epoch": 1212, "lr": 6.597754140637132e-05}
{"train_loss": 0.2260838896036148, "global_step": 107923, "epoch": 1212, "lr": 6.597699206478478e-05}
{"train_loss": 0.208753764629364, "global_step": 107924, "epoch": 1212, "lr": 6.59764427210503e-05}
{"train_loss": 0.10012776404619217, "global_step": 107925, "epoch": 1212, "lr": 6.5975893375168e-05}
{"train_loss": 0.08755101263523102, "global_step": 107926, "epoch": 1212, "lr": 6.597534402713793e-05}
{"train_loss": 0.1520610749721527, "global_step": 107927, "epoch": 1212, "lr": 6.597479467696016e-05}
{"train_loss": 0.19752399623394012, "global_step": 107928, "epoch": 1212, "lr": 6.597424532463478e-05}
{"train_loss": 0.18426711857318878, "global_step": 107929, "epoch": 1212, "lr": 6.597369597016186e-05}
{"train_loss": 0.22005519270896912, "global_step": 107930, "epoch": 1212, "lr": 6.597314661354147e-05}
{"train_loss": 0.137028306722641, "global_step": 107931, "epoch": 1212, "lr": 6.59725972547737e-05}
{"train_loss": 0.1429044008255005, "global_step": 107932, "epoch": 1212, "lr": 6.597204789385858e-05}
{"train_loss": 0.11700272560119629, "global_step": 107933, "epoch": 1212, "lr": 6.597149853079623e-05}
{"train_loss": 0.18673677742481232, "global_step": 107934, "epoch": 1212, "lr": 6.59709491655867e-05}
{"train_loss": 0.16260530054569244, "global_step": 107935, "epoch": 1212, "lr": 6.597039979823006e-05}
{"train_loss": 0.11473967880010605, "global_step": 107936, "epoch": 1212, "lr": 6.596985042872641e-05}
{"train_loss": 0.162916362285614, "global_step": 107937, "epoch": 1212, "lr": 6.596930105707581e-05}
{"train_loss": 0.08895602822303772, "global_step": 107938, "epoch": 1212, "lr": 6.596875168327833e-05}
{"train_loss": 0.1484299749135971, "global_step": 107939, "epoch": 1212, "lr": 6.596820230733404e-05}
{"train_loss": 0.07311450690031052, "global_step": 107940, "epoch": 1212, "lr": 6.596765292924304e-05}
{"train_loss": 0.20280949771404266, "global_step": 107941, "epoch": 1212, "lr": 6.596710354900536e-05}
{"train_loss": 0.1284915655851364, "global_step": 107942, "epoch": 1212, "lr": 6.596655416662109e-05}
{"train_loss": 0.25270870327949524, "global_step": 107943, "epoch": 1212, "lr": 6.596600478209036e-05}
{"train_loss": 0.12960194051265717, "global_step": 107944, "epoch": 1212, "lr": 6.596545539541316e-05}
{"train_loss": 0.1918717622756958, "global_step": 107945, "epoch": 1212, "lr": 6.59649060065896e-05}
{"train_loss": 0.12198007851839066, "global_step": 107946, "epoch": 1212, "lr": 6.596435661561975e-05}
{"train_loss": 0.14342303574085236, "global_step": 107947, "epoch": 1212, "lr": 6.596380722250369e-05}
{"train_loss": 0.17481103539466858, "global_step": 107948, "epoch": 1212, "lr": 6.596325782724148e-05}
{"train_loss": 0.11822400987148285, "global_step": 107949, "epoch": 1212, "lr": 6.596270842983323e-05}
{"train_loss": 0.13908648490905762, "global_step": 107950, "epoch": 1212, "lr": 6.596215903027898e-05}
{"train_loss": 0.19278523325920105, "global_step": 107951, "epoch": 1212, "lr": 6.59616096285788e-05}
{"train_loss": 0.11668268591165543, "global_step": 107952, "epoch": 1212, "lr": 6.596106022473277e-05}
{"train_loss": 0.11366348713636398, "global_step": 107953, "epoch": 1212, "lr": 6.596051081874098e-05}
{"train_loss": 0.20786921679973602, "global_step": 107954, "epoch": 1212, "lr": 6.59599614106035e-05}
{"train_loss": 0.17938938736915588, "global_step": 107955, "epoch": 1212, "lr": 6.59594120003204e-05}
{"train_loss": 0.15499948601374466, "global_step": 107956, "epoch": 1212, "lr": 6.595886258789175e-05, "val_loss": 4.85245943069458}
{"train_loss": 0.12101368606090546, "global_step": 107957, "epoch": 1213, "lr": 6.595831317331762e-05}
{"train_loss": 0.15480582416057587, "global_step": 107958, "epoch": 1213, "lr": 6.595776375659809e-05}
{"train_loss": 0.14084245264530182, "global_step": 107959, "epoch": 1213, "lr": 6.595721433773324e-05}
{"train_loss": 0.1311817765235901, "global_step": 107960, "epoch": 1213, "lr": 6.595666491672312e-05}
{"train_loss": 0.16832789778709412, "global_step": 107961, "epoch": 1213, "lr": 6.595611549356784e-05}
{"train_loss": 0.06882502138614655, "global_step": 107962, "epoch": 1213, "lr": 6.595556606826744e-05}
{"train_loss": 0.16644972562789917, "global_step": 107963, "epoch": 1213, "lr": 6.595501664082204e-05}
{"train_loss": 0.12094427645206451, "global_step": 107964, "epoch": 1213, "lr": 6.595446721123166e-05}
{"train_loss": 0.09550240635871887, "global_step": 107965, "epoch": 1213, "lr": 6.595391777949637e-05}
{"train_loss": 0.11815576255321503, "global_step": 107966, "epoch": 1213, "lr": 6.595336834561631e-05}
{"train_loss": 0.1984068751335144, "global_step": 107967, "epoch": 1213, "lr": 6.59528189095915e-05}
{"train_loss": 0.14361047744750977, "global_step": 107968, "epoch": 1213, "lr": 6.595226947142202e-05}
{"train_loss": 0.08121421188116074, "global_step": 107969, "epoch": 1213, "lr": 6.595172003110797e-05}
{"train_loss": 0.1492031365633011, "global_step": 107970, "epoch": 1213, "lr": 6.595117058864939e-05}
{"train_loss": 0.17230865359306335, "global_step": 107971, "epoch": 1213, "lr": 6.595062114404636e-05}
{"train_loss": 0.17341354489326477, "global_step": 107972, "epoch": 1213, "lr": 6.595007169729898e-05}
{"train_loss": 0.22191855311393738, "global_step": 107973, "epoch": 1213, "lr": 6.594952224840731e-05}
{"train_loss": 0.15010038018226624, "global_step": 107974, "epoch": 1213, "lr": 6.59489727973714e-05}
{"train_loss": 0.08561451733112335, "global_step": 107975, "epoch": 1213, "lr": 6.594842334419137e-05}
{"train_loss": 0.164811909198761, "global_step": 107976, "epoch": 1213, "lr": 6.594787388886725e-05}
{"train_loss": 0.14822626113891602, "global_step": 107977, "epoch": 1213, "lr": 6.594732443139916e-05}
{"train_loss": 0.20636285841464996, "global_step": 107978, "epoch": 1213, "lr": 6.594677497178712e-05}
{"train_loss": 0.11397389322519302, "global_step": 107979, "epoch": 1213, "lr": 6.594622551003123e-05}
{"train_loss": 0.15103408694267273, "global_step": 107980, "epoch": 1213, "lr": 6.594567604613157e-05}
{"train_loss": 0.13665664196014404, "global_step": 107981, "epoch": 1213, "lr": 6.594512658008823e-05}
{"train_loss": 0.14033064246177673, "global_step": 107982, "epoch": 1213, "lr": 6.594457711190125e-05}
{"train_loss": 0.14569993317127228, "global_step": 107983, "epoch": 1213, "lr": 6.59440276415707e-05}
{"train_loss": 0.14326421916484833, "global_step": 107984, "epoch": 1213, "lr": 6.594347816909669e-05}
{"train_loss": 0.12423650175333023, "global_step": 107985, "epoch": 1213, "lr": 6.594292869447926e-05}
{"train_loss": 0.10658317804336548, "global_step": 107986, "epoch": 1213, "lr": 6.594237921771849e-05}
{"train_loss": 0.15872864425182343, "global_step": 107987, "epoch": 1213, "lr": 6.594182973881449e-05}
{"train_loss": 0.1095232293009758, "global_step": 107988, "epoch": 1213, "lr": 6.594128025776728e-05}
{"train_loss": 0.10903698951005936, "global_step": 107989, "epoch": 1213, "lr": 6.594073077457696e-05}
{"train_loss": 0.19929635524749756, "global_step": 107990, "epoch": 1213, "lr": 6.594018128924361e-05}
{"train_loss": 0.1295844465494156, "global_step": 107991, "epoch": 1213, "lr": 6.593963180176728e-05}
{"train_loss": 0.17448747158050537, "global_step": 107992, "epoch": 1213, "lr": 6.59390823121481e-05}
{"train_loss": 0.1290968507528305, "global_step": 107993, "epoch": 1213, "lr": 6.593853282038607e-05}
{"train_loss": 0.09349486976861954, "global_step": 107994, "epoch": 1213, "lr": 6.59379833264813e-05}
{"train_loss": 0.1452759951353073, "global_step": 107995, "epoch": 1213, "lr": 6.593743383043389e-05}
{"train_loss": 0.1970680207014084, "global_step": 107996, "epoch": 1213, "lr": 6.593688433224387e-05}
{"train_loss": 0.16416709125041962, "global_step": 107997, "epoch": 1213, "lr": 6.593633483191133e-05}
{"train_loss": 0.20832127332687378, "global_step": 107998, "epoch": 1213, "lr": 6.593578532943634e-05}
{"train_loss": 0.1060766726732254, "global_step": 107999, "epoch": 1213, "lr": 6.593523582481898e-05}
{"train_loss": 0.18955734372138977, "global_step": 108000, "epoch": 1213, "lr": 6.593468631805932e-05}
{"train_loss": 0.10305624455213547, "global_step": 108001, "epoch": 1213, "lr": 6.593413680915745e-05}
{"train_loss": 0.17619626224040985, "global_step": 108002, "epoch": 1213, "lr": 6.593358729811341e-05}
{"train_loss": 0.17390072345733643, "global_step": 108003, "epoch": 1213, "lr": 6.593303778492731e-05}
{"train_loss": 0.13290970027446747, "global_step": 108004, "epoch": 1213, "lr": 6.59324882695992e-05}
{"train_loss": 0.13252024352550507, "global_step": 108005, "epoch": 1213, "lr": 6.593193875212915e-05}
{"train_loss": 0.19998522102832794, "global_step": 108006, "epoch": 1213, "lr": 6.593138923251726e-05}
{"train_loss": 0.1530209183692932, "global_step": 108007, "epoch": 1213, "lr": 6.593083971076358e-05}
{"train_loss": 0.19625437259674072, "global_step": 108008, "epoch": 1213, "lr": 6.593029018686819e-05}
{"train_loss": 0.1256067156791687, "global_step": 108009, "epoch": 1213, "lr": 6.592974066083117e-05}
{"train_loss": 0.08757065236568451, "global_step": 108010, "epoch": 1213, "lr": 6.59291911326526e-05}
{"train_loss": 0.13259705901145935, "global_step": 108011, "epoch": 1213, "lr": 6.592864160233254e-05}
{"train_loss": 0.12232664972543716, "global_step": 108012, "epoch": 1213, "lr": 6.592809206987108e-05}
{"train_loss": 0.18983152508735657, "global_step": 108013, "epoch": 1213, "lr": 6.592754253526825e-05}
{"train_loss": 0.18240876495838165, "global_step": 108014, "epoch": 1213, "lr": 6.592699299852418e-05}
{"train_loss": 0.13682377338409424, "global_step": 108015, "epoch": 1213, "lr": 6.592644345963892e-05}
{"train_loss": 0.12971793115139008, "global_step": 108016, "epoch": 1213, "lr": 6.592589391861253e-05}
{"train_loss": 0.18979312479496002, "global_step": 108017, "epoch": 1213, "lr": 6.59253443754451e-05}
{"train_loss": 0.09549910575151443, "global_step": 108018, "epoch": 1213, "lr": 6.592479483013671e-05}
{"train_loss": 0.12155155837535858, "global_step": 108019, "epoch": 1213, "lr": 6.592424528268743e-05}
{"train_loss": 0.12679390609264374, "global_step": 108020, "epoch": 1213, "lr": 6.592369573309734e-05}
{"train_loss": 0.2599010169506073, "global_step": 108021, "epoch": 1213, "lr": 6.592314618136648e-05}
{"train_loss": 0.24718739092350006, "global_step": 108022, "epoch": 1213, "lr": 6.592259662749495e-05}
{"train_loss": 0.1538509875535965, "global_step": 108023, "epoch": 1213, "lr": 6.592204707148282e-05}
{"train_loss": 0.1593850702047348, "global_step": 108024, "epoch": 1213, "lr": 6.592149751333017e-05}
{"train_loss": 0.12740546464920044, "global_step": 108025, "epoch": 1213, "lr": 6.592094795303706e-05}
{"train_loss": 0.11497554928064346, "global_step": 108026, "epoch": 1213, "lr": 6.592039839060358e-05}
{"train_loss": 0.18809369206428528, "global_step": 108027, "epoch": 1213, "lr": 6.59198488260298e-05}
{"train_loss": 0.1445993036031723, "global_step": 108028, "epoch": 1213, "lr": 6.591929925931578e-05}
{"train_loss": 0.17873890697956085, "global_step": 108029, "epoch": 1213, "lr": 6.591874969046161e-05}
{"train_loss": 0.19641074538230896, "global_step": 108030, "epoch": 1213, "lr": 6.591820011946737e-05}
{"train_loss": 0.1200294941663742, "global_step": 108031, "epoch": 1213, "lr": 6.59176505463331e-05}
{"train_loss": 0.10698195546865463, "global_step": 108032, "epoch": 1213, "lr": 6.59171009710589e-05}
{"train_loss": 0.12760037183761597, "global_step": 108033, "epoch": 1213, "lr": 6.591655139364485e-05}
{"train_loss": 0.1438073217868805, "global_step": 108034, "epoch": 1213, "lr": 6.591600181409102e-05}
{"train_loss": 0.15314824879169464, "global_step": 108035, "epoch": 1213, "lr": 6.591545223239747e-05}
{"train_loss": 0.1806788295507431, "global_step": 108036, "epoch": 1213, "lr": 6.591490264856428e-05}
{"train_loss": 0.1527591496706009, "global_step": 108037, "epoch": 1213, "lr": 6.591435306259151e-05}
{"train_loss": 0.16305097937583923, "global_step": 108038, "epoch": 1213, "lr": 6.591380347447927e-05}
{"train_loss": 0.180003359913826, "global_step": 108039, "epoch": 1213, "lr": 6.591325388422761e-05}
{"train_loss": 0.1818901151418686, "global_step": 108040, "epoch": 1213, "lr": 6.591270429183662e-05}
{"train_loss": 0.11368922144174576, "global_step": 108041, "epoch": 1213, "lr": 6.591215469730634e-05}
{"train_loss": 0.14750012755393982, "global_step": 108042, "epoch": 1213, "lr": 6.591160510063686e-05}
{"train_loss": 0.11707182973623276, "global_step": 108043, "epoch": 1213, "lr": 6.591105550182827e-05}
{"train_loss": 0.14294342696666718, "global_step": 108044, "epoch": 1213, "lr": 6.591050590088062e-05}
{"train_loss": 0.1489029432113251, "global_step": 108045, "epoch": 1213, "lr": 6.590995629779401e-05, "val_loss": 4.722592353820801}
{"train_loss": 0.19306786358356476, "global_step": 108046, "epoch": 1214, "lr": 6.590940669256849e-05}
{"train_loss": 0.12587760388851166, "global_step": 108047, "epoch": 1214, "lr": 6.590885708520415e-05}
{"train_loss": 0.10929616540670395, "global_step": 108048, "epoch": 1214, "lr": 6.590830747570108e-05}
{"train_loss": 0.15230582654476166, "global_step": 108049, "epoch": 1214, "lr": 6.59077578640593e-05}
{"train_loss": 0.1648271381855011, "global_step": 108050, "epoch": 1214, "lr": 6.590720825027892e-05}
{"train_loss": 0.11171411722898483, "global_step": 108051, "epoch": 1214, "lr": 6.590665863436003e-05}
{"train_loss": 0.10765538364648819, "global_step": 108052, "epoch": 1214, "lr": 6.590610901630267e-05}
{"train_loss": 0.1792803704738617, "global_step": 108053, "epoch": 1214, "lr": 6.590555939610694e-05}
{"train_loss": 0.15779981017112732, "global_step": 108054, "epoch": 1214, "lr": 6.590500977377287e-05}
{"train_loss": 0.12514282763004303, "global_step": 108055, "epoch": 1214, "lr": 6.590446014930058e-05}
{"train_loss": 0.15762126445770264, "global_step": 108056, "epoch": 1214, "lr": 6.590391052269015e-05}
{"train_loss": 0.1800326704978943, "global_step": 108057, "epoch": 1214, "lr": 6.590336089394161e-05}
{"train_loss": 0.16630782186985016, "global_step": 108058, "epoch": 1214, "lr": 6.590281126305506e-05}
{"train_loss": 0.1303759515285492, "global_step": 108059, "epoch": 1214, "lr": 6.59022616300306e-05}
{"train_loss": 0.18117262423038483, "global_step": 108060, "epoch": 1214, "lr": 6.590171199486825e-05}
{"train_loss": 0.2214532047510147, "global_step": 108061, "epoch": 1214, "lr": 6.59011623575681e-05}
{"train_loss": 0.1571386754512787, "global_step": 108062, "epoch": 1214, "lr": 6.590061271813022e-05}
{"train_loss": 0.17323359847068787, "global_step": 108063, "epoch": 1214, "lr": 6.590006307655472e-05}
{"train_loss": 0.15020106732845306, "global_step": 108064, "epoch": 1214, "lr": 6.589951343284165e-05}
{"train_loss": 0.14445534348487854, "global_step": 108065, "epoch": 1214, "lr": 6.589896378699109e-05}
{"train_loss": 0.10009218752384186, "global_step": 108066, "epoch": 1214, "lr": 6.589841413900309e-05}
{"train_loss": 0.06798595190048218, "global_step": 108067, "epoch": 1214, "lr": 6.589786448887776e-05}
{"train_loss": 0.11202040314674377, "global_step": 108068, "epoch": 1214, "lr": 6.589731483661515e-05}
{"train_loss": 0.09949801117181778, "global_step": 108069, "epoch": 1214, "lr": 6.589676518221534e-05}
{"train_loss": 0.1472359299659729, "global_step": 108070, "epoch": 1214, "lr": 6.58962155256784e-05}
{"train_loss": 0.11404884606599808, "global_step": 108071, "epoch": 1214, "lr": 6.58956658670044e-05}
{"train_loss": 0.2450612485408783, "global_step": 108072, "epoch": 1214, "lr": 6.589511620619344e-05}
{"train_loss": 0.1348821222782135, "global_step": 108073, "epoch": 1214, "lr": 6.589456654324556e-05}
{"train_loss": 0.21435046195983887, "global_step": 108074, "epoch": 1214, "lr": 6.589401687816086e-05}
{"train_loss": 0.13162505626678467, "global_step": 108075, "epoch": 1214, "lr": 6.58934672109394e-05}
{"train_loss": 0.15626758337020874, "global_step": 108076, "epoch": 1214, "lr": 6.589291754158125e-05}
{"train_loss": 0.09556567668914795, "global_step": 108077, "epoch": 1214, "lr": 6.58923678700865e-05}
{"train_loss": 0.1651098132133484, "global_step": 108078, "epoch": 1214, "lr": 6.589181819645522e-05}
{"train_loss": 0.11040551960468292, "global_step": 108079, "epoch": 1214, "lr": 6.589126852068745e-05}
{"train_loss": 0.1764003336429596, "global_step": 108080, "epoch": 1214, "lr": 6.589071884278332e-05}
{"train_loss": 0.17041349411010742, "global_step": 108081, "epoch": 1214, "lr": 6.589016916274286e-05}
{"train_loss": 0.1536441594362259, "global_step": 108082, "epoch": 1214, "lr": 6.588961948056617e-05}
{"train_loss": 0.20029966533184052, "global_step": 108083, "epoch": 1214, "lr": 6.588906979625332e-05}
{"train_loss": 0.17843763530254364, "global_step": 108084, "epoch": 1214, "lr": 6.588852010980438e-05}
{"train_loss": 0.1436508446931839, "global_step": 108085, "epoch": 1214, "lr": 6.58879704212194e-05}
{"train_loss": 0.1081613302230835, "global_step": 108086, "epoch": 1214, "lr": 6.58874207304985e-05}
{"train_loss": 0.1801435947418213, "global_step": 108087, "epoch": 1214, "lr": 6.588687103764171e-05}
{"train_loss": 0.13997036218643188, "global_step": 108088, "epoch": 1214, "lr": 6.588632134264914e-05}
{"train_loss": 0.1451662927865982, "global_step": 108089, "epoch": 1214, "lr": 6.588577164552085e-05}
{"train_loss": 0.0699479728937149, "global_step": 108090, "epoch": 1214, "lr": 6.588522194625689e-05}
{"train_loss": 0.15349885821342468, "global_step": 108091, "epoch": 1214, "lr": 6.588467224485738e-05}
{"train_loss": 0.09023518115282059, "global_step": 108092, "epoch": 1214, "lr": 6.588412254132234e-05}
{"train_loss": 0.2010815292596817, "global_step": 108093, "epoch": 1214, "lr": 6.58835728356519e-05}
{"train_loss": 0.1287163645029068, "global_step": 108094, "epoch": 1214, "lr": 6.58830231278461e-05}
{"train_loss": 0.16236324608325958, "global_step": 108095, "epoch": 1214, "lr": 6.588247341790502e-05}
{"train_loss": 0.09396621584892273, "global_step": 108096, "epoch": 1214, "lr": 6.588192370582873e-05}
{"train_loss": 0.11590341478586197, "global_step": 108097, "epoch": 1214, "lr": 6.588137399161731e-05}
{"train_loss": 0.09561465680599213, "global_step": 108098, "epoch": 1214, "lr": 6.588082427527084e-05}
{"train_loss": 0.14397390186786652, "global_step": 108099, "epoch": 1214, "lr": 6.588027455678939e-05}
{"train_loss": 0.21486824750900269, "global_step": 108100, "epoch": 1214, "lr": 6.587972483617301e-05}
{"train_loss": 0.11148451268672943, "global_step": 108101, "epoch": 1214, "lr": 6.587917511342182e-05}
{"train_loss": 0.15879128873348236, "global_step": 108102, "epoch": 1214, "lr": 6.587862538853585e-05}
{"train_loss": 0.2261340618133545, "global_step": 108103, "epoch": 1214, "lr": 6.58780756615152e-05}
{"train_loss": 0.12361785769462585, "global_step": 108104, "epoch": 1214, "lr": 6.587752593235994e-05}
{"train_loss": 0.1421511173248291, "global_step": 108105, "epoch": 1214, "lr": 6.587697620107014e-05}
{"train_loss": 0.20949070155620575, "global_step": 108106, "epoch": 1214, "lr": 6.587642646764587e-05}
{"train_loss": 0.13627654314041138, "global_step": 108107, "epoch": 1214, "lr": 6.58758767320872e-05}
{"train_loss": 0.18138615787029266, "global_step": 108108, "epoch": 1214, "lr": 6.587532699439422e-05}
{"train_loss": 0.12266015261411667, "global_step": 108109, "epoch": 1214, "lr": 6.5874777254567e-05}
{"train_loss": 0.10938149690628052, "global_step": 108110, "epoch": 1214, "lr": 6.58742275126056e-05}
{"train_loss": 0.0728982537984848, "global_step": 108111, "epoch": 1214, "lr": 6.587367776851011e-05}
{"train_loss": 0.1745097041130066, "global_step": 108112, "epoch": 1214, "lr": 6.58731280222806e-05}
{"train_loss": 0.06053357571363449, "global_step": 108113, "epoch": 1214, "lr": 6.587257827391712e-05}
{"train_loss": 0.1441432386636734, "global_step": 108114, "epoch": 1214, "lr": 6.58720285234198e-05}
{"train_loss": 0.226872518658638, "global_step": 108115, "epoch": 1214, "lr": 6.587147877078866e-05}
{"train_loss": 0.19011710584163666, "global_step": 108116, "epoch": 1214, "lr": 6.58709290160238e-05}
{"train_loss": 0.14991751313209534, "global_step": 108117, "epoch": 1214, "lr": 6.587037925912528e-05}
{"train_loss": 0.20184434950351715, "global_step": 108118, "epoch": 1214, "lr": 6.586982950009318e-05}
{"train_loss": 0.15121103823184967, "global_step": 108119, "epoch": 1214, "lr": 6.586927973892758e-05}
{"train_loss": 0.11356355249881744, "global_step": 108120, "epoch": 1214, "lr": 6.586872997562854e-05}
{"train_loss": 0.17667444050312042, "global_step": 108121, "epoch": 1214, "lr": 6.586818021019614e-05}
{"train_loss": 0.14269544184207916, "global_step": 108122, "epoch": 1214, "lr": 6.586763044263049e-05}
{"train_loss": 0.1376151144504547, "global_step": 108123, "epoch": 1214, "lr": 6.586708067293159e-05}
{"train_loss": 0.14879532158374786, "global_step": 108124, "epoch": 1214, "lr": 6.586653090109958e-05}
{"train_loss": 0.13762812316417694, "global_step": 108125, "epoch": 1214, "lr": 6.586598112713449e-05}
{"train_loss": 0.14459319412708282, "global_step": 108126, "epoch": 1214, "lr": 6.586543135103643e-05}
{"train_loss": 0.17399613559246063, "global_step": 108127, "epoch": 1214, "lr": 6.586488157280544e-05}
{"train_loss": 0.22959384322166443, "global_step": 108128, "epoch": 1214, "lr": 6.586433179244163e-05}
{"train_loss": 0.13936267793178558, "global_step": 108129, "epoch": 1214, "lr": 6.586378200994503e-05}
{"train_loss": 0.19772852957248688, "global_step": 108130, "epoch": 1214, "lr": 6.586323222531575e-05}
{"train_loss": 0.14278660714626312, "global_step": 108131, "epoch": 1214, "lr": 6.586268243855386e-05}
{"train_loss": 0.12107562273740768, "global_step": 108132, "epoch": 1214, "lr": 6.586213264965942e-05}
{"train_loss": 0.1205160915851593, "global_step": 108133, "epoch": 1214, "lr": 6.586158285863249e-05}
{"train_loss": 0.14944199648466003, "global_step": 108134, "epoch": 1214, "lr": 6.586103306547318e-05, "val_loss": 4.975642204284668}
{"train_loss": 0.19889196753501892, "global_step": 108135, "epoch": 1215, "lr": 6.586048327018155e-05}
{"train_loss": 0.1510978788137436, "global_step": 108136, "epoch": 1215, "lr": 6.585993347275768e-05}
{"train_loss": 0.15538519620895386, "global_step": 108137, "epoch": 1215, "lr": 6.585938367320162e-05}
{"train_loss": 0.2308758944272995, "global_step": 108138, "epoch": 1215, "lr": 6.585883387151346e-05}
{"train_loss": 0.13939926028251648, "global_step": 108139, "epoch": 1215, "lr": 6.585828406769328e-05}
{"train_loss": 0.1134849414229393, "global_step": 108140, "epoch": 1215, "lr": 6.585773426174115e-05}
{"train_loss": 0.17362375557422638, "global_step": 108141, "epoch": 1215, "lr": 6.585718445365713e-05}
{"train_loss": 0.17219553887844086, "global_step": 108142, "epoch": 1215, "lr": 6.58566346434413e-05}
{"train_loss": 0.11530838161706924, "global_step": 108143, "epoch": 1215, "lr": 6.585608483109376e-05}
{"train_loss": 0.21372871100902557, "global_step": 108144, "epoch": 1215, "lr": 6.585553501661455e-05}
{"train_loss": 0.1947277933359146, "global_step": 108145, "epoch": 1215, "lr": 6.585498520000376e-05}
{"train_loss": 0.13819707930088043, "global_step": 108146, "epoch": 1215, "lr": 6.585443538126145e-05}
{"train_loss": 0.15513834357261658, "global_step": 108147, "epoch": 1215, "lr": 6.585388556038771e-05}
{"train_loss": 0.19314563274383545, "global_step": 108148, "epoch": 1215, "lr": 6.58533357373826e-05}
{"train_loss": 0.22356016933918, "global_step": 108149, "epoch": 1215, "lr": 6.585278591224622e-05}
{"train_loss": 0.3836163282394409, "global_step": 108150, "epoch": 1215, "lr": 6.585223608497861e-05}
{"train_loss": 0.2482018619775772, "global_step": 108151, "epoch": 1215, "lr": 6.585168625557987e-05}
{"train_loss": 0.13649724423885345, "global_step": 108152, "epoch": 1215, "lr": 6.585113642405008e-05}
{"train_loss": 0.15797416865825653, "global_step": 108153, "epoch": 1215, "lr": 6.585058659038927e-05}
{"train_loss": 0.1272173970937729, "global_step": 108154, "epoch": 1215, "lr": 6.585003675459755e-05}
{"train_loss": 0.1627386063337326, "global_step": 108155, "epoch": 1215, "lr": 6.584948691667499e-05}
{"train_loss": 0.11094170808792114, "global_step": 108156, "epoch": 1215, "lr": 6.584893707662166e-05}
{"train_loss": 0.21081063151359558, "global_step": 108157, "epoch": 1215, "lr": 6.584838723443763e-05}
{"train_loss": 0.18959441781044006, "global_step": 108158, "epoch": 1215, "lr": 6.584783739012298e-05}
{"train_loss": 0.18387404084205627, "global_step": 108159, "epoch": 1215, "lr": 6.584728754367777e-05}
{"train_loss": 0.14002856612205505, "global_step": 108160, "epoch": 1215, "lr": 6.58467376951021e-05}
{"train_loss": 0.2315741777420044, "global_step": 108161, "epoch": 1215, "lr": 6.584618784439602e-05}
{"train_loss": 0.148528590798378, "global_step": 108162, "epoch": 1215, "lr": 6.584563799155962e-05}
{"train_loss": 0.18300245702266693, "global_step": 108163, "epoch": 1215, "lr": 6.584508813659297e-05}
{"train_loss": 0.1577582061290741, "global_step": 108164, "epoch": 1215, "lr": 6.584453827949613e-05}
{"train_loss": 0.14680059254169464, "global_step": 108165, "epoch": 1215, "lr": 6.58439884202692e-05}
{"train_loss": 0.13545207679271698, "global_step": 108166, "epoch": 1215, "lr": 6.584343855891223e-05}
{"train_loss": 0.13433565199375153, "global_step": 108167, "epoch": 1215, "lr": 6.58428886954253e-05}
{"train_loss": 0.27814409136772156, "global_step": 108168, "epoch": 1215, "lr": 6.584233882980849e-05}
{"train_loss": 0.15412268042564392, "global_step": 108169, "epoch": 1215, "lr": 6.584178896206187e-05}
{"train_loss": 0.15820297598838806, "global_step": 108170, "epoch": 1215, "lr": 6.584123909218551e-05}
{"train_loss": 0.19830012321472168, "global_step": 108171, "epoch": 1215, "lr": 6.58406892201795e-05}
{"train_loss": 0.17746858298778534, "global_step": 108172, "epoch": 1215, "lr": 6.584013934604389e-05}
{"train_loss": 0.1922488957643509, "global_step": 108173, "epoch": 1215, "lr": 6.583958946977878e-05}
{"train_loss": 0.17750318348407745, "global_step": 108174, "epoch": 1215, "lr": 6.583903959138423e-05}
{"train_loss": 0.14171290397644043, "global_step": 108175, "epoch": 1215, "lr": 6.583848971086028e-05}
{"train_loss": 0.09830991923809052, "global_step": 108176, "epoch": 1215, "lr": 6.583793982820708e-05}
{"train_loss": 0.22688941657543182, "global_step": 108177, "epoch": 1215, "lr": 6.583738994342464e-05}
{"train_loss": 0.17277301847934723, "global_step": 108178, "epoch": 1215, "lr": 6.583684005651306e-05}
{"train_loss": 0.14993053674697876, "global_step": 108179, "epoch": 1215, "lr": 6.583629016747242e-05}
{"train_loss": 0.19272541999816895, "global_step": 108180, "epoch": 1215, "lr": 6.583574027630278e-05}
{"train_loss": 0.15057231485843658, "global_step": 108181, "epoch": 1215, "lr": 6.583519038300421e-05}
{"train_loss": 0.12667888402938843, "global_step": 108182, "epoch": 1215, "lr": 6.583464048757679e-05}
{"train_loss": 0.22672294080257416, "global_step": 108183, "epoch": 1215, "lr": 6.58340905900206e-05}
{"train_loss": 0.2102455496788025, "global_step": 108184, "epoch": 1215, "lr": 6.583354069033572e-05}
{"train_loss": 0.10154429823160172, "global_step": 108185, "epoch": 1215, "lr": 6.583299078852219e-05}
{"train_loss": 0.11822930723428726, "global_step": 108186, "epoch": 1215, "lr": 6.583244088458012e-05}
{"train_loss": 0.2250804305076599, "global_step": 108187, "epoch": 1215, "lr": 6.583189097850959e-05}
{"train_loss": 0.1700802743434906, "global_step": 108188, "epoch": 1215, "lr": 6.583134107031061e-05}
{"train_loss": 0.14079709351062775, "global_step": 108189, "epoch": 1215, "lr": 6.583079115998334e-05}
{"train_loss": 0.20577365159988403, "global_step": 108190, "epoch": 1215, "lr": 6.58302412475278e-05}
{"train_loss": 0.19056202471256256, "global_step": 108191, "epoch": 1215, "lr": 6.582969133294408e-05}
{"train_loss": 0.15275083482265472, "global_step": 108192, "epoch": 1215, "lr": 6.582914141623225e-05}
{"train_loss": 0.11262892186641693, "global_step": 108193, "epoch": 1215, "lr": 6.582859149739237e-05}
{"train_loss": 0.19433654844760895, "global_step": 108194, "epoch": 1215, "lr": 6.582804157642455e-05}
{"train_loss": 0.24748975038528442, "global_step": 108195, "epoch": 1215, "lr": 6.582749165332884e-05}
{"train_loss": 0.13745619356632233, "global_step": 108196, "epoch": 1215, "lr": 6.58269417281053e-05}
{"train_loss": 0.15161101520061493, "global_step": 108197, "epoch": 1215, "lr": 6.582639180075403e-05}
{"train_loss": 0.15001749992370605, "global_step": 108198, "epoch": 1215, "lr": 6.582584187127511e-05}
{"train_loss": 0.2078545093536377, "global_step": 108199, "epoch": 1215, "lr": 6.582529193966858e-05}
{"train_loss": 0.13286355137825012, "global_step": 108200, "epoch": 1215, "lr": 6.582474200593453e-05}
{"train_loss": 0.13335926830768585, "global_step": 108201, "epoch": 1215, "lr": 6.582419207007305e-05}
{"train_loss": 0.17288705706596375, "global_step": 108202, "epoch": 1215, "lr": 6.58236421320842e-05}
{"train_loss": 0.16917450726032257, "global_step": 108203, "epoch": 1215, "lr": 6.582309219196805e-05}
{"train_loss": 0.13321447372436523, "global_step": 108204, "epoch": 1215, "lr": 6.582254224972468e-05}
{"train_loss": 0.15038983523845673, "global_step": 108205, "epoch": 1215, "lr": 6.582199230535416e-05}
{"train_loss": 0.20657294988632202, "global_step": 108206, "epoch": 1215, "lr": 6.582144235885654e-05}
{"train_loss": 0.24311740696430206, "global_step": 108207, "epoch": 1215, "lr": 6.582089241023195e-05}
{"train_loss": 0.12242865562438965, "global_step": 108208, "epoch": 1215, "lr": 6.582034245948043e-05}
{"train_loss": 0.1893230676651001, "global_step": 108209, "epoch": 1215, "lr": 6.581979250660207e-05}
{"train_loss": 0.12372628599405289, "global_step": 108210, "epoch": 1215, "lr": 6.58192425515969e-05}
{"train_loss": 0.11575867235660553, "global_step": 108211, "epoch": 1215, "lr": 6.581869259446504e-05}
{"train_loss": 0.13665176928043365, "global_step": 108212, "epoch": 1215, "lr": 6.581814263520656e-05}
{"train_loss": 0.17192821204662323, "global_step": 108213, "epoch": 1215, "lr": 6.581759267382152e-05}
{"train_loss": 0.21544596552848816, "global_step": 108214, "epoch": 1215, "lr": 6.581704271030999e-05}
{"train_loss": 0.19187644124031067, "global_step": 108215, "epoch": 1215, "lr": 6.581649274467206e-05}
{"train_loss": 0.1490224003791809, "global_step": 108216, "epoch": 1215, "lr": 6.581594277690778e-05}
{"train_loss": 0.1294233649969101, "global_step": 108217, "epoch": 1215, "lr": 6.581539280701725e-05}
{"train_loss": 0.18634605407714844, "global_step": 108218, "epoch": 1215, "lr": 6.581484283500054e-05}
{"train_loss": 0.16348570585250854, "global_step": 108219, "epoch": 1215, "lr": 6.58142928608577e-05}
{"train_loss": 0.150181382894516, "global_step": 108220, "epoch": 1215, "lr": 6.581374288458883e-05}
{"train_loss": 0.13524068892002106, "global_step": 108221, "epoch": 1215, "lr": 6.5813192906194e-05}
{"train_loss": 0.08451929688453674, "global_step": 108222, "epoch": 1215, "lr": 6.581264292567327e-05}
{"train_loss": 0.17016585681880458, "global_step": 108223, "epoch": 1215, "lr": 6.581209294302673e-05, "val_loss": 4.895596504211426, "train_action_mse_error": 35.52950668334961}
{"train_loss": 0.15516428649425507, "global_step": 108224, "epoch": 1216, "lr": 6.581154295825444e-05}
{"train_loss": 0.2208852469921112, "global_step": 108225, "epoch": 1216, "lr": 6.581099297135647e-05}
{"train_loss": 0.14371518790721893, "global_step": 108226, "epoch": 1216, "lr": 6.581044298233293e-05}
{"train_loss": 0.2422584593296051, "global_step": 108227, "epoch": 1216, "lr": 6.580989299118385e-05}
{"train_loss": 0.16051068902015686, "global_step": 108228, "epoch": 1216, "lr": 6.580934299790935e-05}
{"train_loss": 0.17401689291000366, "global_step": 108229, "epoch": 1216, "lr": 6.580879300250946e-05}
{"train_loss": 0.12063229084014893, "global_step": 108230, "epoch": 1216, "lr": 6.580824300498428e-05}
{"train_loss": 0.1345396786928177, "global_step": 108231, "epoch": 1216, "lr": 6.580769300533385e-05}
{"train_loss": 0.1462857574224472, "global_step": 108232, "epoch": 1216, "lr": 6.580714300355828e-05}
{"train_loss": 0.2681034803390503, "global_step": 108233, "epoch": 1216, "lr": 6.580659299965764e-05}
{"train_loss": 0.1444241851568222, "global_step": 108234, "epoch": 1216, "lr": 6.580604299363198e-05}
{"train_loss": 0.10519766807556152, "global_step": 108235, "epoch": 1216, "lr": 6.58054929854814e-05}
{"train_loss": 0.13502928614616394, "global_step": 108236, "epoch": 1216, "lr": 6.580494297520596e-05}
{"train_loss": 0.18517424166202545, "global_step": 108237, "epoch": 1216, "lr": 6.580439296280574e-05}
{"train_loss": 0.09851384907960892, "global_step": 108238, "epoch": 1216, "lr": 6.580384294828082e-05}
{"train_loss": 0.08480067551136017, "global_step": 108239, "epoch": 1216, "lr": 6.580329293163126e-05}
{"train_loss": 0.16430550813674927, "global_step": 108240, "epoch": 1216, "lr": 6.580274291285715e-05}
{"train_loss": 0.135307177901268, "global_step": 108241, "epoch": 1216, "lr": 6.580219289195853e-05}
{"train_loss": 0.19635659456253052, "global_step": 108242, "epoch": 1216, "lr": 6.580164286893551e-05}
{"train_loss": 0.15773741900920868, "global_step": 108243, "epoch": 1216, "lr": 6.580109284378815e-05}
{"train_loss": 0.282183974981308, "global_step": 108244, "epoch": 1216, "lr": 6.580054281651652e-05}
{"train_loss": 0.135864719748497, "global_step": 108245, "epoch": 1216, "lr": 6.579999278712071e-05}
{"train_loss": 0.12156081199645996, "global_step": 108246, "epoch": 1216, "lr": 6.579944275560078e-05}
{"train_loss": 0.10890273749828339, "global_step": 108247, "epoch": 1216, "lr": 6.579889272195682e-05}
{"train_loss": 0.10258153080940247, "global_step": 108248, "epoch": 1216, "lr": 6.579834268618888e-05}
{"train_loss": 0.17519564926624298, "global_step": 108249, "epoch": 1216, "lr": 6.579779264829704e-05}
{"train_loss": 0.23446886241436005, "global_step": 108250, "epoch": 1216, "lr": 6.57972426082814e-05}
{"train_loss": 0.16419166326522827, "global_step": 108251, "epoch": 1216, "lr": 6.579669256614199e-05}
{"train_loss": 0.14934472739696503, "global_step": 108252, "epoch": 1216, "lr": 6.57961425218789e-05}
{"train_loss": 0.14749786257743835, "global_step": 108253, "epoch": 1216, "lr": 6.579559247549224e-05}
{"train_loss": 0.19569961726665497, "global_step": 108254, "epoch": 1216, "lr": 6.579504242698204e-05}
{"train_loss": 0.14173787832260132, "global_step": 108255, "epoch": 1216, "lr": 6.57944923763484e-05}
{"train_loss": 0.12824974954128265, "global_step": 108256, "epoch": 1216, "lr": 6.579394232359138e-05}
{"train_loss": 0.17908534407615662, "global_step": 108257, "epoch": 1216, "lr": 6.579339226871104e-05}
{"train_loss": 0.11960400640964508, "global_step": 108258, "epoch": 1216, "lr": 6.579284221170749e-05}
{"train_loss": 0.1326780617237091, "global_step": 108259, "epoch": 1216, "lr": 6.579229215258077e-05}
{"train_loss": 0.1722184270620346, "global_step": 108260, "epoch": 1216, "lr": 6.579174209133098e-05}
{"train_loss": 0.11101847141981125, "global_step": 108261, "epoch": 1216, "lr": 6.579119202795816e-05}
{"train_loss": 0.10775363445281982, "global_step": 108262, "epoch": 1216, "lr": 6.579064196246243e-05}
{"train_loss": 0.18014171719551086, "global_step": 108263, "epoch": 1216, "lr": 6.579009189484383e-05}
{"train_loss": 0.19636580348014832, "global_step": 108264, "epoch": 1216, "lr": 6.578954182510245e-05}
{"train_loss": 0.12179628014564514, "global_step": 108265, "epoch": 1216, "lr": 6.578899175323836e-05}
{"train_loss": 0.16818445920944214, "global_step": 108266, "epoch": 1216, "lr": 6.578844167925165e-05}
{"train_loss": 0.13190121948719025, "global_step": 108267, "epoch": 1216, "lr": 6.578789160314235e-05}
{"train_loss": 0.17827686667442322, "global_step": 108268, "epoch": 1216, "lr": 6.578734152491058e-05}
{"train_loss": 0.1452382653951645, "global_step": 108269, "epoch": 1216, "lr": 6.578679144455637e-05}
{"train_loss": 0.12124182283878326, "global_step": 108270, "epoch": 1216, "lr": 6.578624136207984e-05}
{"train_loss": 0.1761402189731598, "global_step": 108271, "epoch": 1216, "lr": 6.578569127748103e-05}
{"train_loss": 0.1289840042591095, "global_step": 108272, "epoch": 1216, "lr": 6.578514119076002e-05}
{"train_loss": 0.19215428829193115, "global_step": 108273, "epoch": 1216, "lr": 6.57845911019169e-05}
{"train_loss": 0.14136847853660583, "global_step": 108274, "epoch": 1216, "lr": 6.578404101095173e-05}
{"train_loss": 0.13747599720954895, "global_step": 108275, "epoch": 1216, "lr": 6.57834909178646e-05}
{"train_loss": 0.11095923185348511, "global_step": 108276, "epoch": 1216, "lr": 6.578294082265556e-05}
{"train_loss": 0.15406598150730133, "global_step": 108277, "epoch": 1216, "lr": 6.578239072532471e-05}
{"train_loss": 0.15357908606529236, "global_step": 108278, "epoch": 1216, "lr": 6.578184062587209e-05}
{"train_loss": 0.09827080368995667, "global_step": 108279, "epoch": 1216, "lr": 6.57812905242978e-05}
{"train_loss": 0.15194523334503174, "global_step": 108280, "epoch": 1216, "lr": 6.57807404206019e-05}
{"train_loss": 0.15340712666511536, "global_step": 108281, "epoch": 1216, "lr": 6.578019031478449e-05}
{"train_loss": 0.1871383637189865, "global_step": 108282, "epoch": 1216, "lr": 6.577964020684562e-05}
{"train_loss": 0.17184877395629883, "global_step": 108283, "epoch": 1216, "lr": 6.577909009678537e-05}
{"train_loss": 0.07274595648050308, "global_step": 108284, "epoch": 1216, "lr": 6.57785399846038e-05}
{"train_loss": 0.0737689882516861, "global_step": 108285, "epoch": 1216, "lr": 6.577798987030102e-05}
{"train_loss": 0.18331748247146606, "global_step": 108286, "epoch": 1216, "lr": 6.577743975387707e-05}
{"train_loss": 0.08130014687776566, "global_step": 108287, "epoch": 1216, "lr": 6.577688963533204e-05}
{"train_loss": 0.14972297847270966, "global_step": 108288, "epoch": 1216, "lr": 6.5776339514666e-05}
{"train_loss": 0.12751318514347076, "global_step": 108289, "epoch": 1216, "lr": 6.577578939187903e-05}
{"train_loss": 0.147267147898674, "global_step": 108290, "epoch": 1216, "lr": 6.577523926697117e-05}
{"train_loss": 0.19236452877521515, "global_step": 108291, "epoch": 1216, "lr": 6.577468913994255e-05}
{"train_loss": 0.1350627988576889, "global_step": 108292, "epoch": 1216, "lr": 6.57741390107932e-05}
{"train_loss": 0.13651737570762634, "global_step": 108293, "epoch": 1216, "lr": 6.577358887952322e-05}
{"train_loss": 0.08437584340572357, "global_step": 108294, "epoch": 1216, "lr": 6.577303874613267e-05}
{"train_loss": 0.09383070468902588, "global_step": 108295, "epoch": 1216, "lr": 6.577248861062161e-05}
{"train_loss": 0.12810905277729034, "global_step": 108296, "epoch": 1216, "lr": 6.577193847299015e-05}
{"train_loss": 0.11465917527675629, "global_step": 108297, "epoch": 1216, "lr": 6.577138833323833e-05}
{"train_loss": 0.22413845360279083, "global_step": 108298, "epoch": 1216, "lr": 6.577083819136625e-05}
{"train_loss": 0.1030658558011055, "global_step": 108299, "epoch": 1216, "lr": 6.577028804737397e-05}
{"train_loss": 0.13281404972076416, "global_step": 108300, "epoch": 1216, "lr": 6.576973790126156e-05}
{"train_loss": 0.08725690841674805, "global_step": 108301, "epoch": 1216, "lr": 6.576918775302912e-05}
{"train_loss": 0.18548700213432312, "global_step": 108302, "epoch": 1216, "lr": 6.576863760267668e-05}
{"train_loss": 0.16816607117652893, "global_step": 108303, "epoch": 1216, "lr": 6.576808745020436e-05}
{"train_loss": 0.18825729191303253, "global_step": 108304, "epoch": 1216, "lr": 6.576753729561221e-05}
{"train_loss": 0.1728326827287674, "global_step": 108305, "epoch": 1216, "lr": 6.576698713890032e-05}
{"train_loss": 0.11659583449363708, "global_step": 108306, "epoch": 1216, "lr": 6.576643698006873e-05}
{"train_loss": 0.18438862264156342, "global_step": 108307, "epoch": 1216, "lr": 6.576588681911753e-05}
{"train_loss": 0.17497289180755615, "global_step": 108308, "epoch": 1216, "lr": 6.576533665604682e-05}
{"train_loss": 0.10056588798761368, "global_step": 108309, "epoch": 1216, "lr": 6.576478649085663e-05}
{"train_loss": 0.19388659298419952, "global_step": 108310, "epoch": 1216, "lr": 6.576423632354706e-05}
{"train_loss": 0.10494940727949142, "global_step": 108311, "epoch": 1216, "lr": 6.57636861541182e-05}
{"train_loss": 0.1486609462606773, "global_step": 108312, "epoch": 1216, "lr": 6.57631359825701e-05, "val_loss": 4.991037845611572}
{"train_loss": 0.15173009037971497, "global_step": 108313, "epoch": 1217, "lr": 6.576258580890282e-05}
{"train_loss": 0.13101372122764587, "global_step": 108314, "epoch": 1217, "lr": 6.576203563311646e-05}
{"train_loss": 0.12469235062599182, "global_step": 108315, "epoch": 1217, "lr": 6.576148545521107e-05}
{"train_loss": 0.14393942058086395, "global_step": 108316, "epoch": 1217, "lr": 6.576093527518676e-05}
{"train_loss": 0.16241276264190674, "global_step": 108317, "epoch": 1217, "lr": 6.57603850930436e-05}
{"train_loss": 0.10548204183578491, "global_step": 108318, "epoch": 1217, "lr": 6.575983490878163e-05}
{"train_loss": 0.15401391685009003, "global_step": 108319, "epoch": 1217, "lr": 6.575928472240093e-05}
{"train_loss": 0.13340018689632416, "global_step": 108320, "epoch": 1217, "lr": 6.575873453390161e-05}
{"train_loss": 0.10185351222753525, "global_step": 108321, "epoch": 1217, "lr": 6.575818434328371e-05}
{"train_loss": 0.12920290231704712, "global_step": 108322, "epoch": 1217, "lr": 6.575763415054731e-05}
{"train_loss": 0.17714273929595947, "global_step": 108323, "epoch": 1217, "lr": 6.57570839556925e-05}
{"train_loss": 0.10624903440475464, "global_step": 108324, "epoch": 1217, "lr": 6.575653375871934e-05}
{"train_loss": 0.1842377632856369, "global_step": 108325, "epoch": 1217, "lr": 6.57559835596279e-05}
{"train_loss": 0.15520407259464264, "global_step": 108326, "epoch": 1217, "lr": 6.575543335841828e-05}
{"train_loss": 0.14102742075920105, "global_step": 108327, "epoch": 1217, "lr": 6.57548831550905e-05}
{"train_loss": 0.12977369129657745, "global_step": 108328, "epoch": 1217, "lr": 6.575433294964468e-05}
{"train_loss": 0.08343124389648438, "global_step": 108329, "epoch": 1217, "lr": 6.575378274208089e-05}
{"train_loss": 0.12596042454242706, "global_step": 108330, "epoch": 1217, "lr": 6.575323253239918e-05}
{"train_loss": 0.15467751026153564, "global_step": 108331, "epoch": 1217, "lr": 6.575268232059966e-05}
{"train_loss": 0.1560632884502411, "global_step": 108332, "epoch": 1217, "lr": 6.575213210668238e-05}
{"train_loss": 0.17358562350273132, "global_step": 108333, "epoch": 1217, "lr": 6.57515818906474e-05}
{"train_loss": 0.10185980796813965, "global_step": 108334, "epoch": 1217, "lr": 6.575103167249483e-05}
{"train_loss": 0.22626470029354095, "global_step": 108335, "epoch": 1217, "lr": 6.575048145222471e-05}
{"train_loss": 0.21554799377918243, "global_step": 108336, "epoch": 1217, "lr": 6.574993122983715e-05}
{"train_loss": 0.16396863758563995, "global_step": 108337, "epoch": 1217, "lr": 6.57493810053322e-05}
{"train_loss": 0.11309502273797989, "global_step": 108338, "epoch": 1217, "lr": 6.574883077870993e-05}
{"train_loss": 0.17261014878749847, "global_step": 108339, "epoch": 1217, "lr": 6.574828054997042e-05}
{"train_loss": 0.16907581686973572, "global_step": 108340, "epoch": 1217, "lr": 6.574773031911374e-05}
{"train_loss": 0.12178916484117508, "global_step": 108341, "epoch": 1217, "lr": 6.574718008613998e-05}
{"train_loss": 0.1183258667588234, "global_step": 108342, "epoch": 1217, "lr": 6.57466298510492e-05}
{"train_loss": 0.105105921626091, "global_step": 108343, "epoch": 1217, "lr": 6.574607961384148e-05}
{"train_loss": 0.09158486127853394, "global_step": 108344, "epoch": 1217, "lr": 6.574552937451691e-05}
{"train_loss": 0.12722890079021454, "global_step": 108345, "epoch": 1217, "lr": 6.574497913307552e-05}
{"train_loss": 0.16915899515151978, "global_step": 108346, "epoch": 1217, "lr": 6.574442888951742e-05}
{"train_loss": 0.1654943823814392, "global_step": 108347, "epoch": 1217, "lr": 6.574387864384266e-05}
{"train_loss": 0.17579779028892517, "global_step": 108348, "epoch": 1217, "lr": 6.574332839605134e-05}
{"train_loss": 0.09377434849739075, "global_step": 108349, "epoch": 1217, "lr": 6.574277814614353e-05}
{"train_loss": 0.22541989386081696, "global_step": 108350, "epoch": 1217, "lr": 6.574222789411927e-05}
{"train_loss": 0.17516694962978363, "global_step": 108351, "epoch": 1217, "lr": 6.574167763997867e-05}
{"train_loss": 0.0941169485449791, "global_step": 108352, "epoch": 1217, "lr": 6.57411273837218e-05}
{"train_loss": 0.11610982567071915, "global_step": 108353, "epoch": 1217, "lr": 6.574057712534871e-05}
{"train_loss": 0.1622822880744934, "global_step": 108354, "epoch": 1217, "lr": 6.574002686485952e-05}
{"train_loss": 0.13182199001312256, "global_step": 108355, "epoch": 1217, "lr": 6.573947660225426e-05}
{"train_loss": 0.13878287374973297, "global_step": 108356, "epoch": 1217, "lr": 6.5738926337533e-05}
{"train_loss": 0.14662998914718628, "global_step": 108357, "epoch": 1217, "lr": 6.573837607069587e-05}
{"train_loss": 0.18615154922008514, "global_step": 108358, "epoch": 1217, "lr": 6.573782580174288e-05}
{"train_loss": 0.1261758804321289, "global_step": 108359, "epoch": 1217, "lr": 6.573727553067415e-05}
{"train_loss": 0.20342151820659637, "global_step": 108360, "epoch": 1217, "lr": 6.573672525748972e-05}
{"train_loss": 0.18926137685775757, "global_step": 108361, "epoch": 1217, "lr": 6.573617498218968e-05}
{"train_loss": 0.12369757145643234, "global_step": 108362, "epoch": 1217, "lr": 6.573562470477411e-05}
{"train_loss": 0.13134995102882385, "global_step": 108363, "epoch": 1217, "lr": 6.573507442524309e-05}
{"train_loss": 0.2061866819858551, "global_step": 108364, "epoch": 1217, "lr": 6.573452414359667e-05}
{"train_loss": 0.12445481866598129, "global_step": 108365, "epoch": 1217, "lr": 6.573397385983493e-05}
{"train_loss": 0.1379200667142868, "global_step": 108366, "epoch": 1217, "lr": 6.573342357395795e-05}
{"train_loss": 0.1218937560915947, "global_step": 108367, "epoch": 1217, "lr": 6.57328732859658e-05}
{"train_loss": 0.21410833299160004, "global_step": 108368, "epoch": 1217, "lr": 6.573232299585855e-05}
{"train_loss": 0.12177196145057678, "global_step": 108369, "epoch": 1217, "lr": 6.573177270363629e-05}
{"train_loss": 0.17813090980052948, "global_step": 108370, "epoch": 1217, "lr": 6.573122240929908e-05}
{"train_loss": 0.13379734754562378, "global_step": 108371, "epoch": 1217, "lr": 6.573067211284701e-05}
{"train_loss": 0.17999239265918732, "global_step": 108372, "epoch": 1217, "lr": 6.573012181428013e-05}
{"train_loss": 0.1485147625207901, "global_step": 108373, "epoch": 1217, "lr": 6.572957151359852e-05}
{"train_loss": 0.13826875388622284, "global_step": 108374, "epoch": 1217, "lr": 6.572902121080229e-05}
{"train_loss": 0.1575978845357895, "global_step": 108375, "epoch": 1217, "lr": 6.572847090589148e-05}
{"train_loss": 0.21077637374401093, "global_step": 108376, "epoch": 1217, "lr": 6.572792059886616e-05}
{"train_loss": 0.16810652613639832, "global_step": 108377, "epoch": 1217, "lr": 6.57273702897264e-05}
{"train_loss": 0.15754522383213043, "global_step": 108378, "epoch": 1217, "lr": 6.57268199784723e-05}
{"train_loss": 0.1693238765001297, "global_step": 108379, "epoch": 1217, "lr": 6.572626966510393e-05}
{"train_loss": 0.15473993122577667, "global_step": 108380, "epoch": 1217, "lr": 6.572571934962133e-05}
{"train_loss": 0.1668178141117096, "global_step": 108381, "epoch": 1217, "lr": 6.572516903202461e-05}
{"train_loss": 0.15049460530281067, "global_step": 108382, "epoch": 1217, "lr": 6.572461871231385e-05}
{"train_loss": 0.2487405687570572, "global_step": 108383, "epoch": 1217, "lr": 6.57240683904891e-05}
{"train_loss": 0.16384169459342957, "global_step": 108384, "epoch": 1217, "lr": 6.572351806655043e-05}
{"train_loss": 0.1603049784898758, "global_step": 108385, "epoch": 1217, "lr": 6.572296774049792e-05}
{"train_loss": 0.23249240219593048, "global_step": 108386, "epoch": 1217, "lr": 6.572241741233166e-05}
{"train_loss": 0.19260461628437042, "global_step": 108387, "epoch": 1217, "lr": 6.57218670820517e-05}
{"train_loss": 0.1017843559384346, "global_step": 108388, "epoch": 1217, "lr": 6.572131674965814e-05}
{"train_loss": 0.1552167385816574, "global_step": 108389, "epoch": 1217, "lr": 6.572076641515103e-05}
{"train_loss": 0.15594927966594696, "global_step": 108390, "epoch": 1217, "lr": 6.572021607853045e-05}
{"train_loss": 0.08126287907361984, "global_step": 108391, "epoch": 1217, "lr": 6.57196657397965e-05}
{"train_loss": 0.17596343159675598, "global_step": 108392, "epoch": 1217, "lr": 6.571911539894921e-05}
{"train_loss": 0.15076559782028198, "global_step": 108393, "epoch": 1217, "lr": 6.571856505598869e-05}
{"train_loss": 0.1775878369808197, "global_step": 108394, "epoch": 1217, "lr": 6.5718014710915e-05}
{"train_loss": 0.07954772561788559, "global_step": 108395, "epoch": 1217, "lr": 6.571746436372821e-05}
{"train_loss": 0.1252894550561905, "global_step": 108396, "epoch": 1217, "lr": 6.571691401442841e-05}
{"train_loss": 0.24483944475650787, "global_step": 108397, "epoch": 1217, "lr": 6.571636366301564e-05}
{"train_loss": 0.1397557258605957, "global_step": 108398, "epoch": 1217, "lr": 6.571581330949002e-05}
{"train_loss": 0.1798241138458252, "global_step": 108399, "epoch": 1217, "lr": 6.571526295385158e-05}
{"train_loss": 0.2944946587085724, "global_step": 108400, "epoch": 1217, "lr": 6.571471259610044e-05}
{"train_loss": 0.15375207498502197, "global_step": 108401, "epoch": 1217, "lr": 6.571416223623663e-05, "val_loss": 5.003830909729004}
{"train_loss": 0.177153542637825, "global_step": 108402, "epoch": 1218, "lr": 6.571361187426026e-05}
{"train_loss": 0.10360019654035568, "global_step": 108403, "epoch": 1218, "lr": 6.571306151017137e-05}
{"train_loss": 0.10935191065073013, "global_step": 108404, "epoch": 1218, "lr": 6.571251114397004e-05}
{"train_loss": 0.14047551155090332, "global_step": 108405, "epoch": 1218, "lr": 6.571196077565638e-05}
{"train_loss": 0.13089963793754578, "global_step": 108406, "epoch": 1218, "lr": 6.571141040523041e-05}
{"train_loss": 0.11273515224456787, "global_step": 108407, "epoch": 1218, "lr": 6.571086003269224e-05}
{"train_loss": 0.1553940325975418, "global_step": 108408, "epoch": 1218, "lr": 6.571030965804194e-05}
{"train_loss": 0.16396832466125488, "global_step": 108409, "epoch": 1218, "lr": 6.57097592812796e-05}
{"train_loss": 0.14170412719249725, "global_step": 108410, "epoch": 1218, "lr": 6.570920890240527e-05}
{"train_loss": 0.13098779320716858, "global_step": 108411, "epoch": 1218, "lr": 6.570865852141901e-05}
{"train_loss": 0.223582923412323, "global_step": 108412, "epoch": 1218, "lr": 6.570810813832092e-05}
{"train_loss": 0.16346345841884613, "global_step": 108413, "epoch": 1218, "lr": 6.570755775311106e-05}
{"train_loss": 0.13117410242557526, "global_step": 108414, "epoch": 1218, "lr": 6.570700736578952e-05}
{"train_loss": 0.15701594948768616, "global_step": 108415, "epoch": 1218, "lr": 6.570645697635636e-05}
{"train_loss": 0.15427452325820923, "global_step": 108416, "epoch": 1218, "lr": 6.570590658481165e-05}
{"train_loss": 0.21130810678005219, "global_step": 108417, "epoch": 1218, "lr": 6.570535619115549e-05}
{"train_loss": 0.15986379981040955, "global_step": 108418, "epoch": 1218, "lr": 6.570480579538793e-05}
{"train_loss": 0.09784848988056183, "global_step": 108419, "epoch": 1218, "lr": 6.570425539750904e-05}
{"train_loss": 0.19735600054264069, "global_step": 108420, "epoch": 1218, "lr": 6.57037049975189e-05}
{"train_loss": 0.2109161764383316, "global_step": 108421, "epoch": 1218, "lr": 6.570315459541762e-05}
{"train_loss": 0.106230229139328, "global_step": 108422, "epoch": 1218, "lr": 6.570260419120522e-05}
{"train_loss": 0.16694089770317078, "global_step": 108423, "epoch": 1218, "lr": 6.57020537848818e-05}
{"train_loss": 0.09755279868841171, "global_step": 108424, "epoch": 1218, "lr": 6.570150337644743e-05}
{"train_loss": 0.18096883594989777, "global_step": 108425, "epoch": 1218, "lr": 6.570095296590218e-05}
{"train_loss": 0.15779046714305878, "global_step": 108426, "epoch": 1218, "lr": 6.570040255324612e-05}
{"train_loss": 0.13807515799999237, "global_step": 108427, "epoch": 1218, "lr": 6.569985213847935e-05}
{"train_loss": 0.22047939896583557, "global_step": 108428, "epoch": 1218, "lr": 6.569930172160191e-05}
{"train_loss": 0.21179968118667603, "global_step": 108429, "epoch": 1218, "lr": 6.56987513026139e-05}
{"train_loss": 0.230507031083107, "global_step": 108430, "epoch": 1218, "lr": 6.56982008815154e-05}
{"train_loss": 0.17863640189170837, "global_step": 108431, "epoch": 1218, "lr": 6.569765045830644e-05}
{"train_loss": 0.09069734811782837, "global_step": 108432, "epoch": 1218, "lr": 6.569710003298713e-05}
{"train_loss": 0.12675026059150696, "global_step": 108433, "epoch": 1218, "lr": 6.569654960555756e-05}
{"train_loss": 0.18480999767780304, "global_step": 108434, "epoch": 1218, "lr": 6.569599917601775e-05}
{"train_loss": 0.18207144737243652, "global_step": 108435, "epoch": 1218, "lr": 6.569544874436781e-05}
{"train_loss": 0.17090217769145966, "global_step": 108436, "epoch": 1218, "lr": 6.569489831060782e-05}
{"train_loss": 0.17559406161308289, "global_step": 108437, "epoch": 1218, "lr": 6.569434787473784e-05}
{"train_loss": 0.0986715778708458, "global_step": 108438, "epoch": 1218, "lr": 6.569379743675794e-05}
{"train_loss": 0.11034597456455231, "global_step": 108439, "epoch": 1218, "lr": 6.56932469966682e-05}
{"train_loss": 0.07181403785943985, "global_step": 108440, "epoch": 1218, "lr": 6.569269655446871e-05}
{"train_loss": 0.14455397427082062, "global_step": 108441, "epoch": 1218, "lr": 6.569214611015951e-05}
{"train_loss": 0.19187282025814056, "global_step": 108442, "epoch": 1218, "lr": 6.56915956637407e-05}
{"train_loss": 0.16812664270401, "global_step": 108443, "epoch": 1218, "lr": 6.569104521521233e-05}
{"train_loss": 0.11520480364561081, "global_step": 108444, "epoch": 1218, "lr": 6.56904947645745e-05}
{"train_loss": 0.138772115111351, "global_step": 108445, "epoch": 1218, "lr": 6.568994431182727e-05}
{"train_loss": 0.12962017953395844, "global_step": 108446, "epoch": 1218, "lr": 6.568939385697073e-05}
{"train_loss": 0.12441279739141464, "global_step": 108447, "epoch": 1218, "lr": 6.568884340000494e-05}
{"train_loss": 0.1879831999540329, "global_step": 108448, "epoch": 1218, "lr": 6.568829294092998e-05}
{"train_loss": 0.16645163297653198, "global_step": 108449, "epoch": 1218, "lr": 6.568774247974592e-05}
{"train_loss": 0.12662984430789948, "global_step": 108450, "epoch": 1218, "lr": 6.568719201645281e-05}
{"train_loss": 0.19153130054473877, "global_step": 108451, "epoch": 1218, "lr": 6.568664155105078e-05}
{"train_loss": 0.17007958889007568, "global_step": 108452, "epoch": 1218, "lr": 6.568609108353986e-05}
{"train_loss": 0.1790579855442047, "global_step": 108453, "epoch": 1218, "lr": 6.568554061392013e-05}
{"train_loss": 0.14342767000198364, "global_step": 108454, "epoch": 1218, "lr": 6.568499014219167e-05}
{"train_loss": 0.1262737363576889, "global_step": 108455, "epoch": 1218, "lr": 6.568443966835457e-05}
{"train_loss": 0.18082422018051147, "global_step": 108456, "epoch": 1218, "lr": 6.568388919240887e-05}
{"train_loss": 0.12516796588897705, "global_step": 108457, "epoch": 1218, "lr": 6.568333871435467e-05}
{"train_loss": 0.11152847111225128, "global_step": 108458, "epoch": 1218, "lr": 6.568278823419204e-05}
{"train_loss": 0.14309348165988922, "global_step": 108459, "epoch": 1218, "lr": 6.568223775192105e-05}
{"train_loss": 0.1878783255815506, "global_step": 108460, "epoch": 1218, "lr": 6.568168726754177e-05}
{"train_loss": 0.16836518049240112, "global_step": 108461, "epoch": 1218, "lr": 6.568113678105428e-05}
{"train_loss": 0.12769150733947754, "global_step": 108462, "epoch": 1218, "lr": 6.568058629245865e-05}
{"train_loss": 0.14927895367145538, "global_step": 108463, "epoch": 1218, "lr": 6.568003580175496e-05}
{"train_loss": 0.10455414652824402, "global_step": 108464, "epoch": 1218, "lr": 6.567948530894328e-05}
{"train_loss": 0.12341473996639252, "global_step": 108465, "epoch": 1218, "lr": 6.567893481402368e-05}
{"train_loss": 0.17769652605056763, "global_step": 108466, "epoch": 1218, "lr": 6.567838431699624e-05}
{"train_loss": 0.18061819672584534, "global_step": 108467, "epoch": 1218, "lr": 6.567783381786103e-05}
{"train_loss": 0.11276517808437347, "global_step": 108468, "epoch": 1218, "lr": 6.567728331661813e-05}
{"train_loss": 0.15610076487064362, "global_step": 108469, "epoch": 1218, "lr": 6.567673281326761e-05}
{"train_loss": 0.16546529531478882, "global_step": 108470, "epoch": 1218, "lr": 6.567618230780954e-05}
{"train_loss": 0.14784856140613556, "global_step": 108471, "epoch": 1218, "lr": 6.567563180024401e-05}
{"train_loss": 0.21344546973705292, "global_step": 108472, "epoch": 1218, "lr": 6.567508129057107e-05}
{"train_loss": 0.118559330701828, "global_step": 108473, "epoch": 1218, "lr": 6.567453077879083e-05}
{"train_loss": 0.1001635268330574, "global_step": 108474, "epoch": 1218, "lr": 6.567398026490331e-05}
{"train_loss": 0.18848344683647156, "global_step": 108475, "epoch": 1218, "lr": 6.567342974890863e-05}
{"train_loss": 0.20579281449317932, "global_step": 108476, "epoch": 1218, "lr": 6.567287923080684e-05}
{"train_loss": 0.1081680953502655, "global_step": 108477, "epoch": 1218, "lr": 6.567232871059802e-05}
{"train_loss": 0.10320919752120972, "global_step": 108478, "epoch": 1218, "lr": 6.567177818828226e-05}
{"train_loss": 0.14697523415088654, "global_step": 108479, "epoch": 1218, "lr": 6.567122766385961e-05}
{"train_loss": 0.11071733385324478, "global_step": 108480, "epoch": 1218, "lr": 6.567067713733016e-05}
{"train_loss": 0.1301933079957962, "global_step": 108481, "epoch": 1218, "lr": 6.567012660869399e-05}
{"train_loss": 0.15913282334804535, "global_step": 108482, "epoch": 1218, "lr": 6.566957607795113e-05}
{"train_loss": 0.15642477571964264, "global_step": 108483, "epoch": 1218, "lr": 6.566902554510172e-05}
{"train_loss": 0.13282975554466248, "global_step": 108484, "epoch": 1218, "lr": 6.566847501014578e-05}
{"train_loss": 0.20918413996696472, "global_step": 108485, "epoch": 1218, "lr": 6.56679244730834e-05}
{"train_loss": 0.13154959678649902, "global_step": 108486, "epoch": 1218, "lr": 6.566737393391468e-05}
{"train_loss": 0.1160554587841034, "global_step": 108487, "epoch": 1218, "lr": 6.566682339263965e-05}
{"train_loss": 0.12732276320457458, "global_step": 108488, "epoch": 1218, "lr": 6.566627284925842e-05}
{"train_loss": 0.17187926173210144, "global_step": 108489, "epoch": 1218, "lr": 6.566572230377106e-05}
{"train_loss": 0.15085818986879307, "global_step": 108490, "epoch": 1218, "lr": 6.566517175617763e-05, "val_loss": 4.882105827331543}
{"train_loss": 0.16533006727695465, "global_step": 108491, "epoch": 1219, "lr": 6.56646212064782e-05}
{"train_loss": 0.14253513514995575, "global_step": 108492, "epoch": 1219, "lr": 6.566407065467285e-05}
{"train_loss": 0.1742600053548813, "global_step": 108493, "epoch": 1219, "lr": 6.566352010076165e-05}
{"train_loss": 0.11261879652738571, "global_step": 108494, "epoch": 1219, "lr": 6.56629695447447e-05}
{"train_loss": 0.11855514347553253, "global_step": 108495, "epoch": 1219, "lr": 6.566241898662204e-05}
{"train_loss": 0.17514564096927643, "global_step": 108496, "epoch": 1219, "lr": 6.566186842639376e-05}
{"train_loss": 0.11780530214309692, "global_step": 108497, "epoch": 1219, "lr": 6.566131786405995e-05}
{"train_loss": 0.2192116230726242, "global_step": 108498, "epoch": 1219, "lr": 6.566076729962066e-05}
{"train_loss": 0.12907055020332336, "global_step": 108499, "epoch": 1219, "lr": 6.566021673307596e-05}
{"train_loss": 0.17970815300941467, "global_step": 108500, "epoch": 1219, "lr": 6.565966616442594e-05}
{"train_loss": 0.09618156403303146, "global_step": 108501, "epoch": 1219, "lr": 6.565911559367067e-05}
{"train_loss": 0.16253438591957092, "global_step": 108502, "epoch": 1219, "lr": 6.565856502081022e-05}
{"train_loss": 0.13390693068504333, "global_step": 108503, "epoch": 1219, "lr": 6.565801444584467e-05}
{"train_loss": 0.11075184494256973, "global_step": 108504, "epoch": 1219, "lr": 6.565746386877408e-05}
{"train_loss": 0.1490636169910431, "global_step": 108505, "epoch": 1219, "lr": 6.565691328959855e-05}
{"train_loss": 0.16433706879615784, "global_step": 108506, "epoch": 1219, "lr": 6.565636270831813e-05}
{"train_loss": 0.11554080247879028, "global_step": 108507, "epoch": 1219, "lr": 6.56558121249329e-05}
{"train_loss": 0.11512552946805954, "global_step": 108508, "epoch": 1219, "lr": 6.565526153944294e-05}
{"train_loss": 0.18205459415912628, "global_step": 108509, "epoch": 1219, "lr": 6.565471095184833e-05}
{"train_loss": 0.21036802232265472, "global_step": 108510, "epoch": 1219, "lr": 6.565416036214913e-05}
{"train_loss": 0.09795554727315903, "global_step": 108511, "epoch": 1219, "lr": 6.565360977034543e-05}
{"train_loss": 0.12044335156679153, "global_step": 108512, "epoch": 1219, "lr": 6.565305917643727e-05}
{"train_loss": 0.18941085040569305, "global_step": 108513, "epoch": 1219, "lr": 6.565250858042475e-05}
{"train_loss": 0.21897898614406586, "global_step": 108514, "epoch": 1219, "lr": 6.565195798230795e-05}
{"train_loss": 0.1597863882780075, "global_step": 108515, "epoch": 1219, "lr": 6.565140738208694e-05}
{"train_loss": 0.1754741072654724, "global_step": 108516, "epoch": 1219, "lr": 6.565085677976179e-05}
{"train_loss": 0.1455756574869156, "global_step": 108517, "epoch": 1219, "lr": 6.565030617533256e-05}
{"train_loss": 0.15400488674640656, "global_step": 108518, "epoch": 1219, "lr": 6.564975556879933e-05}
{"train_loss": 0.07648587226867676, "global_step": 108519, "epoch": 1219, "lr": 6.56492049601622e-05}
{"train_loss": 0.19035769999027252, "global_step": 108520, "epoch": 1219, "lr": 6.564865434942122e-05}
{"train_loss": 0.11682263016700745, "global_step": 108521, "epoch": 1219, "lr": 6.564810373657646e-05}
{"train_loss": 0.10206147283315659, "global_step": 108522, "epoch": 1219, "lr": 6.564755312162803e-05}
{"train_loss": 0.11728081107139587, "global_step": 108523, "epoch": 1219, "lr": 6.564700250457595e-05}
{"train_loss": 0.12659429013729095, "global_step": 108524, "epoch": 1219, "lr": 6.564645188542032e-05}
{"train_loss": 0.2180705964565277, "global_step": 108525, "epoch": 1219, "lr": 6.564590126416124e-05}
{"train_loss": 0.16694095730781555, "global_step": 108526, "epoch": 1219, "lr": 6.564535064079873e-05}
{"train_loss": 0.15973061323165894, "global_step": 108527, "epoch": 1219, "lr": 6.564480001533291e-05}
{"train_loss": 0.1329246163368225, "global_step": 108528, "epoch": 1219, "lr": 6.564424938776383e-05}
{"train_loss": 0.1596078723669052, "global_step": 108529, "epoch": 1219, "lr": 6.564369875809157e-05}
{"train_loss": 0.12826403975486755, "global_step": 108530, "epoch": 1219, "lr": 6.564314812631622e-05}
{"train_loss": 0.12591713666915894, "global_step": 108531, "epoch": 1219, "lr": 6.564259749243783e-05}
{"train_loss": 0.1398020088672638, "global_step": 108532, "epoch": 1219, "lr": 6.564204685645648e-05}
{"train_loss": 0.162021666765213, "global_step": 108533, "epoch": 1219, "lr": 6.564149621837225e-05}
{"train_loss": 0.1763366013765335, "global_step": 108534, "epoch": 1219, "lr": 6.564094557818523e-05}
{"train_loss": 0.1691691130399704, "global_step": 108535, "epoch": 1219, "lr": 6.564039493589544e-05}
{"train_loss": 0.16908392310142517, "global_step": 108536, "epoch": 1219, "lr": 6.563984429150302e-05}
{"train_loss": 0.21134665608406067, "global_step": 108537, "epoch": 1219, "lr": 6.563929364500801e-05}
{"train_loss": 0.1327189803123474, "global_step": 108538, "epoch": 1219, "lr": 6.563874299641046e-05}
{"train_loss": 0.18454237282276154, "global_step": 108539, "epoch": 1219, "lr": 6.56381923457105e-05}
{"train_loss": 0.19231851398944855, "global_step": 108540, "epoch": 1219, "lr": 6.563764169290817e-05}
{"train_loss": 0.14014998078346252, "global_step": 108541, "epoch": 1219, "lr": 6.563709103800353e-05}
{"train_loss": 0.120571069419384, "global_step": 108542, "epoch": 1219, "lr": 6.56365403809967e-05}
{"train_loss": 0.10652799904346466, "global_step": 108543, "epoch": 1219, "lr": 6.563598972188772e-05}
{"train_loss": 0.10221383720636368, "global_step": 108544, "epoch": 1219, "lr": 6.563543906067667e-05}
{"train_loss": 0.139314666390419, "global_step": 108545, "epoch": 1219, "lr": 6.563488839736362e-05}
{"train_loss": 0.1789352148771286, "global_step": 108546, "epoch": 1219, "lr": 6.563433773194867e-05}
{"train_loss": 0.1643761843442917, "global_step": 108547, "epoch": 1219, "lr": 6.563378706443184e-05}
{"train_loss": 0.2828783094882965, "global_step": 108548, "epoch": 1219, "lr": 6.563323639481327e-05}
{"train_loss": 0.18728016316890717, "global_step": 108549, "epoch": 1219, "lr": 6.563268572309298e-05}
{"train_loss": 0.11249900609254837, "global_step": 108550, "epoch": 1219, "lr": 6.563213504927108e-05}
{"train_loss": 0.1264510154724121, "global_step": 108551, "epoch": 1219, "lr": 6.563158437334763e-05}
{"train_loss": 0.16757848858833313, "global_step": 108552, "epoch": 1219, "lr": 6.563103369532269e-05}
{"train_loss": 0.16784721612930298, "global_step": 108553, "epoch": 1219, "lr": 6.563048301519636e-05}
{"train_loss": 0.14667193591594696, "global_step": 108554, "epoch": 1219, "lr": 6.562993233296869e-05}
{"train_loss": 0.19428637623786926, "global_step": 108555, "epoch": 1219, "lr": 6.562938164863978e-05}
{"train_loss": 0.1449175626039505, "global_step": 108556, "epoch": 1219, "lr": 6.562883096220969e-05}
{"train_loss": 0.0905478447675705, "global_step": 108557, "epoch": 1219, "lr": 6.56282802736785e-05}
{"train_loss": 0.20172150433063507, "global_step": 108558, "epoch": 1219, "lr": 6.562772958304625e-05}
{"train_loss": 0.15539446473121643, "global_step": 108559, "epoch": 1219, "lr": 6.562717889031306e-05}
{"train_loss": 0.14811696112155914, "global_step": 108560, "epoch": 1219, "lr": 6.562662819547898e-05}
{"train_loss": 0.16888386011123657, "global_step": 108561, "epoch": 1219, "lr": 6.56260774985441e-05}
{"train_loss": 0.17167197167873383, "global_step": 108562, "epoch": 1219, "lr": 6.562552679950848e-05}
{"train_loss": 0.1546228677034378, "global_step": 108563, "epoch": 1219, "lr": 6.56249760983722e-05}
{"train_loss": 0.19515548646450043, "global_step": 108564, "epoch": 1219, "lr": 6.562442539513533e-05}
{"train_loss": 0.16370722651481628, "global_step": 108565, "epoch": 1219, "lr": 6.562387468979795e-05}
{"train_loss": 0.19542469084262848, "global_step": 108566, "epoch": 1219, "lr": 6.562332398236012e-05}
{"train_loss": 0.12208924442529678, "global_step": 108567, "epoch": 1219, "lr": 6.562277327282193e-05}
{"train_loss": 0.09110890328884125, "global_step": 108568, "epoch": 1219, "lr": 6.562222256118346e-05}
{"train_loss": 0.16712380945682526, "global_step": 108569, "epoch": 1219, "lr": 6.562167184744476e-05}
{"train_loss": 0.15432730317115784, "global_step": 108570, "epoch": 1219, "lr": 6.562112113160591e-05}
{"train_loss": 0.08095965534448624, "global_step": 108571, "epoch": 1219, "lr": 6.562057041366701e-05}
{"train_loss": 0.16996818780899048, "global_step": 108572, "epoch": 1219, "lr": 6.562001969362812e-05}
{"train_loss": 0.28390809893608093, "global_step": 108573, "epoch": 1219, "lr": 6.561946897148928e-05}
{"train_loss": 0.16091464459896088, "global_step": 108574, "epoch": 1219, "lr": 6.561891824725061e-05}
{"train_loss": 0.14400067925453186, "global_step": 108575, "epoch": 1219, "lr": 6.561836752091216e-05}
{"train_loss": 0.2729700207710266, "global_step": 108576, "epoch": 1219, "lr": 6.561781679247402e-05}
{"train_loss": 0.2107141613960266, "global_step": 108577, "epoch": 1219, "lr": 6.561726606193624e-05}
{"train_loss": 0.20037543773651123, "global_step": 108578, "epoch": 1219, "lr": 6.561671532929893e-05}
{"train_loss": 0.1564994196543533, "global_step": 108579, "epoch": 1219, "lr": 6.561616459456212e-05, "val_loss": 4.923838138580322}
{"train_loss": 0.1961575299501419, "global_step": 108580, "epoch": 1220, "lr": 6.561561385772592e-05}
{"train_loss": 0.2007419914007187, "global_step": 108581, "epoch": 1220, "lr": 6.56150631187904e-05}
{"train_loss": 0.0936555415391922, "global_step": 108582, "epoch": 1220, "lr": 6.56145123777556e-05}
{"train_loss": 0.1556743085384369, "global_step": 108583, "epoch": 1220, "lr": 6.561396163462165e-05}
{"train_loss": 0.09269063919782639, "global_step": 108584, "epoch": 1220, "lr": 6.561341088938858e-05}
{"train_loss": 0.2056705802679062, "global_step": 108585, "epoch": 1220, "lr": 6.561286014205647e-05}
{"train_loss": 0.060541268438100815, "global_step": 108586, "epoch": 1220, "lr": 6.561230939262539e-05}
{"train_loss": 0.11939660459756851, "global_step": 108587, "epoch": 1220, "lr": 6.561175864109544e-05}
{"train_loss": 0.14309871196746826, "global_step": 108588, "epoch": 1220, "lr": 6.561120788746669e-05}
{"train_loss": 0.18992307782173157, "global_step": 108589, "epoch": 1220, "lr": 6.561065713173918e-05}
{"train_loss": 0.2407957762479782, "global_step": 108590, "epoch": 1220, "lr": 6.561010637391303e-05}
{"train_loss": 0.12886790931224823, "global_step": 108591, "epoch": 1220, "lr": 6.560955561398828e-05}
{"train_loss": 0.20174235105514526, "global_step": 108592, "epoch": 1220, "lr": 6.560900485196502e-05}
{"train_loss": 0.18559707701206207, "global_step": 108593, "epoch": 1220, "lr": 6.560845408784332e-05}
{"train_loss": 0.1105884239077568, "global_step": 108594, "epoch": 1220, "lr": 6.560790332162326e-05}
{"train_loss": 0.18375536799430847, "global_step": 108595, "epoch": 1220, "lr": 6.560735255330489e-05}
{"train_loss": 0.23096489906311035, "global_step": 108596, "epoch": 1220, "lr": 6.560680178288831e-05}
{"train_loss": 0.21259541809558868, "global_step": 108597, "epoch": 1220, "lr": 6.560625101037359e-05}
{"train_loss": 0.15003524720668793, "global_step": 108598, "epoch": 1220, "lr": 6.56057002357608e-05}
{"train_loss": 0.13879762589931488, "global_step": 108599, "epoch": 1220, "lr": 6.560514945905001e-05}
{"train_loss": 0.1679195910692215, "global_step": 108600, "epoch": 1220, "lr": 6.560459868024129e-05}
{"train_loss": 0.10453590750694275, "global_step": 108601, "epoch": 1220, "lr": 6.560404789933473e-05}
{"train_loss": 0.11412377655506134, "global_step": 108602, "epoch": 1220, "lr": 6.560349711633039e-05}
{"train_loss": 0.14555533230304718, "global_step": 108603, "epoch": 1220, "lr": 6.560294633122836e-05}
{"train_loss": 0.12324286252260208, "global_step": 108604, "epoch": 1220, "lr": 6.56023955440287e-05}
{"train_loss": 0.1318969875574112, "global_step": 108605, "epoch": 1220, "lr": 6.560184475473148e-05}
{"train_loss": 0.09052920341491699, "global_step": 108606, "epoch": 1220, "lr": 6.560129396333678e-05}
{"train_loss": 0.14098864793777466, "global_step": 108607, "epoch": 1220, "lr": 6.56007431698447e-05}
{"train_loss": 0.14515632390975952, "global_step": 108608, "epoch": 1220, "lr": 6.560019237425526e-05}
{"train_loss": 0.17513854801654816, "global_step": 108609, "epoch": 1220, "lr": 6.55996415765686e-05}
{"train_loss": 0.18677014112472534, "global_step": 108610, "epoch": 1220, "lr": 6.559909077678474e-05}
{"train_loss": 0.11389081180095673, "global_step": 108611, "epoch": 1220, "lr": 6.559853997490376e-05}
{"train_loss": 0.10012712329626083, "global_step": 108612, "epoch": 1220, "lr": 6.559798917092577e-05}
{"train_loss": 0.14537227153778076, "global_step": 108613, "epoch": 1220, "lr": 6.55974383648508e-05}
{"train_loss": 0.184691920876503, "global_step": 108614, "epoch": 1220, "lr": 6.559688755667895e-05}
{"train_loss": 0.13525767624378204, "global_step": 108615, "epoch": 1220, "lr": 6.55963367464103e-05}
{"train_loss": 0.10029672831296921, "global_step": 108616, "epoch": 1220, "lr": 6.55957859340449e-05}
{"train_loss": 0.09326005727052689, "global_step": 108617, "epoch": 1220, "lr": 6.559523511958283e-05}
{"train_loss": 0.15891991555690765, "global_step": 108618, "epoch": 1220, "lr": 6.55946843030242e-05}
{"train_loss": 0.12670955061912537, "global_step": 108619, "epoch": 1220, "lr": 6.559413348436902e-05}
{"train_loss": 0.11147766560316086, "global_step": 108620, "epoch": 1220, "lr": 6.559358266361742e-05}
{"train_loss": 0.16209173202514648, "global_step": 108621, "epoch": 1220, "lr": 6.559303184076945e-05}
{"train_loss": 0.15898633003234863, "global_step": 108622, "epoch": 1220, "lr": 6.559248101582518e-05}
{"train_loss": 0.13070577383041382, "global_step": 108623, "epoch": 1220, "lr": 6.559193018878468e-05}
{"train_loss": 0.12108801305294037, "global_step": 108624, "epoch": 1220, "lr": 6.559137935964805e-05}
{"train_loss": 0.14518721401691437, "global_step": 108625, "epoch": 1220, "lr": 6.559082852841535e-05}
{"train_loss": 0.14056764543056488, "global_step": 108626, "epoch": 1220, "lr": 6.559027769508665e-05}
{"train_loss": 0.19699972867965698, "global_step": 108627, "epoch": 1220, "lr": 6.558972685966204e-05}
{"train_loss": 0.1126403883099556, "global_step": 108628, "epoch": 1220, "lr": 6.558917602214156e-05}
{"train_loss": 0.14722120761871338, "global_step": 108629, "epoch": 1220, "lr": 6.558862518252533e-05}
{"train_loss": 0.10120247304439545, "global_step": 108630, "epoch": 1220, "lr": 6.558807434081338e-05}
{"train_loss": 0.11587171256542206, "global_step": 108631, "epoch": 1220, "lr": 6.558752349700582e-05}
{"train_loss": 0.11755761504173279, "global_step": 108632, "epoch": 1220, "lr": 6.558697265110269e-05}
{"train_loss": 0.19030959904193878, "global_step": 108633, "epoch": 1220, "lr": 6.558642180310409e-05}
{"train_loss": 0.11583263427019119, "global_step": 108634, "epoch": 1220, "lr": 6.558587095301009e-05}
{"train_loss": 0.10226781666278839, "global_step": 108635, "epoch": 1220, "lr": 6.558532010082074e-05}
{"train_loss": 0.17238064110279083, "global_step": 108636, "epoch": 1220, "lr": 6.558476924653615e-05}
{"train_loss": 0.10353610664606094, "global_step": 108637, "epoch": 1220, "lr": 6.558421839015638e-05}
{"train_loss": 0.19675375521183014, "global_step": 108638, "epoch": 1220, "lr": 6.558366753168149e-05}
{"train_loss": 0.14616012573242188, "global_step": 108639, "epoch": 1220, "lr": 6.558311667111157e-05}
{"train_loss": 0.1407695710659027, "global_step": 108640, "epoch": 1220, "lr": 6.55825658084467e-05}
{"train_loss": 0.10430579632520676, "global_step": 108641, "epoch": 1220, "lr": 6.558201494368694e-05}
{"train_loss": 0.16636595129966736, "global_step": 108642, "epoch": 1220, "lr": 6.558146407683235e-05}
{"train_loss": 0.12524519860744476, "global_step": 108643, "epoch": 1220, "lr": 6.558091320788303e-05}
{"train_loss": 0.09186346083879471, "global_step": 108644, "epoch": 1220, "lr": 6.558036233683905e-05}
{"train_loss": 0.14627154171466827, "global_step": 108645, "epoch": 1220, "lr": 6.557981146370049e-05}
{"train_loss": 0.07459589838981628, "global_step": 108646, "epoch": 1220, "lr": 6.55792605884674e-05}
{"train_loss": 0.15842688083648682, "global_step": 108647, "epoch": 1220, "lr": 6.557870971113987e-05}
{"train_loss": 0.09802906960248947, "global_step": 108648, "epoch": 1220, "lr": 6.557815883171799e-05}
{"train_loss": 0.10651770234107971, "global_step": 108649, "epoch": 1220, "lr": 6.55776079502018e-05}
{"train_loss": 0.12877453863620758, "global_step": 108650, "epoch": 1220, "lr": 6.55770570665914e-05}
{"train_loss": 0.13463519513607025, "global_step": 108651, "epoch": 1220, "lr": 6.557650618088685e-05}
{"train_loss": 0.1404678076505661, "global_step": 108652, "epoch": 1220, "lr": 6.557595529308823e-05}
{"train_loss": 0.12937651574611664, "global_step": 108653, "epoch": 1220, "lr": 6.55754044031956e-05}
{"train_loss": 0.1087065041065216, "global_step": 108654, "epoch": 1220, "lr": 6.557485351120906e-05}
{"train_loss": 0.16201350092887878, "global_step": 108655, "epoch": 1220, "lr": 6.557430261712867e-05}
{"train_loss": 0.13674336671829224, "global_step": 108656, "epoch": 1220, "lr": 6.55737517209545e-05}
{"train_loss": 0.14395685493946075, "global_step": 108657, "epoch": 1220, "lr": 6.557320082268664e-05}
{"train_loss": 0.10358988493680954, "global_step": 108658, "epoch": 1220, "lr": 6.557264992232513e-05}
{"train_loss": 0.1915658414363861, "global_step": 108659, "epoch": 1220, "lr": 6.557209901987009e-05}
{"train_loss": 0.09737586230039597, "global_step": 108660, "epoch": 1220, "lr": 6.557154811532155e-05}
{"train_loss": 0.18551361560821533, "global_step": 108661, "epoch": 1220, "lr": 6.557099720867962e-05}
{"train_loss": 0.12710349261760712, "global_step": 108662, "epoch": 1220, "lr": 6.557044629994436e-05}
{"train_loss": 0.12742850184440613, "global_step": 108663, "epoch": 1220, "lr": 6.556989538911583e-05}
{"train_loss": 0.10812094062566757, "global_step": 108664, "epoch": 1220, "lr": 6.556934447619413e-05}
{"train_loss": 0.12813252210617065, "global_step": 108665, "epoch": 1220, "lr": 6.556879356117933e-05}
{"train_loss": 0.17144273221492767, "global_step": 108666, "epoch": 1220, "lr": 6.556824264407148e-05}
{"train_loss": 0.20934906601905823, "global_step": 108667, "epoch": 1220, "lr": 6.556769172487068e-05}
{"train_loss": 0.14169433599944864, "global_step": 108668, "epoch": 1220, "lr": 6.556714080357697e-05, "val_loss": 5.183381080627441, "train_action_mse_error": 26.142242431640625}
{"train_loss": 0.1423742026090622, "global_step": 108669, "epoch": 1221, "lr": 6.556658988019047e-05}
{"train_loss": 0.11422371864318848, "global_step": 108670, "epoch": 1221, "lr": 6.556603895471123e-05}
{"train_loss": 0.15209835767745972, "global_step": 108671, "epoch": 1221, "lr": 6.556548802713933e-05}
{"train_loss": 0.13122570514678955, "global_step": 108672, "epoch": 1221, "lr": 6.556493709747484e-05}
{"train_loss": 0.09466972947120667, "global_step": 108673, "epoch": 1221, "lr": 6.556438616571782e-05}
{"train_loss": 0.1574476808309555, "global_step": 108674, "epoch": 1221, "lr": 6.556383523186835e-05}
{"train_loss": 0.1652528941631317, "global_step": 108675, "epoch": 1221, "lr": 6.556328429592653e-05}
{"train_loss": 0.15907563269138336, "global_step": 108676, "epoch": 1221, "lr": 6.556273335789242e-05}
{"train_loss": 0.18652305006980896, "global_step": 108677, "epoch": 1221, "lr": 6.556218241776608e-05}
{"train_loss": 0.1493748128414154, "global_step": 108678, "epoch": 1221, "lr": 6.556163147554758e-05}
{"train_loss": 0.1880636215209961, "global_step": 108679, "epoch": 1221, "lr": 6.556108053123702e-05}
{"train_loss": 0.1296079307794571, "global_step": 108680, "epoch": 1221, "lr": 6.556052958483447e-05}
{"train_loss": 0.11982909590005875, "global_step": 108681, "epoch": 1221, "lr": 6.555997863634e-05}
{"train_loss": 0.10718624293804169, "global_step": 108682, "epoch": 1221, "lr": 6.555942768575367e-05}
{"train_loss": 0.14817912876605988, "global_step": 108683, "epoch": 1221, "lr": 6.555887673307557e-05}
{"train_loss": 0.0975150540471077, "global_step": 108684, "epoch": 1221, "lr": 6.555832577830578e-05}
{"train_loss": 0.09756358712911606, "global_step": 108685, "epoch": 1221, "lr": 6.555777482144435e-05}
{"train_loss": 0.10117072612047195, "global_step": 108686, "epoch": 1221, "lr": 6.555722386249135e-05}
{"train_loss": 0.1562330424785614, "global_step": 108687, "epoch": 1221, "lr": 6.55566729014469e-05}
{"train_loss": 0.12769344449043274, "global_step": 108688, "epoch": 1221, "lr": 6.555612193831105e-05}
{"train_loss": 0.195339635014534, "global_step": 108689, "epoch": 1221, "lr": 6.555557097308384e-05}
{"train_loss": 0.18781213462352753, "global_step": 108690, "epoch": 1221, "lr": 6.555502000576539e-05}
{"train_loss": 0.17186391353607178, "global_step": 108691, "epoch": 1221, "lr": 6.555446903635575e-05}
{"train_loss": 0.13818614184856415, "global_step": 108692, "epoch": 1221, "lr": 6.5553918064855e-05}
{"train_loss": 0.08109986782073975, "global_step": 108693, "epoch": 1221, "lr": 6.555336709126322e-05}
{"train_loss": 0.16692067682743073, "global_step": 108694, "epoch": 1221, "lr": 6.555281611558047e-05}
{"train_loss": 0.15426170825958252, "global_step": 108695, "epoch": 1221, "lr": 6.555226513780685e-05}
{"train_loss": 0.21435636281967163, "global_step": 108696, "epoch": 1221, "lr": 6.555171415794239e-05}
{"train_loss": 0.15283091366291046, "global_step": 108697, "epoch": 1221, "lr": 6.555116317598722e-05}
{"train_loss": 0.1397310197353363, "global_step": 108698, "epoch": 1221, "lr": 6.555061219194137e-05}
{"train_loss": 0.14124293625354767, "global_step": 108699, "epoch": 1221, "lr": 6.555006120580493e-05}
{"train_loss": 0.1823839396238327, "global_step": 108700, "epoch": 1221, "lr": 6.554951021757798e-05}
{"train_loss": 0.10339844971895218, "global_step": 108701, "epoch": 1221, "lr": 6.55489592272606e-05}
{"train_loss": 0.16221503913402557, "global_step": 108702, "epoch": 1221, "lr": 6.554840823485283e-05}
{"train_loss": 0.09511350840330124, "global_step": 108703, "epoch": 1221, "lr": 6.554785724035478e-05}
{"train_loss": 0.16765201091766357, "global_step": 108704, "epoch": 1221, "lr": 6.55473062437665e-05}
{"train_loss": 0.14571630954742432, "global_step": 108705, "epoch": 1221, "lr": 6.554675524508809e-05}
{"train_loss": 0.1768343299627304, "global_step": 108706, "epoch": 1221, "lr": 6.55462042443196e-05}
{"train_loss": 0.17286217212677002, "global_step": 108707, "epoch": 1221, "lr": 6.554565324146111e-05}
{"train_loss": 0.12251849472522736, "global_step": 108708, "epoch": 1221, "lr": 6.55451022365127e-05}
{"train_loss": 0.1717797815799713, "global_step": 108709, "epoch": 1221, "lr": 6.554455122947444e-05}
{"train_loss": 0.12140524387359619, "global_step": 108710, "epoch": 1221, "lr": 6.55440002203464e-05}
{"train_loss": 0.12406690418720245, "global_step": 108711, "epoch": 1221, "lr": 6.554344920912865e-05}
{"train_loss": 0.2038748413324356, "global_step": 108712, "epoch": 1221, "lr": 6.554289819582128e-05}
{"train_loss": 0.19619417190551758, "global_step": 108713, "epoch": 1221, "lr": 6.554234718042437e-05}
{"train_loss": 0.0797601193189621, "global_step": 108714, "epoch": 1221, "lr": 6.554179616293796e-05}
{"train_loss": 0.22740793228149414, "global_step": 108715, "epoch": 1221, "lr": 6.554124514336215e-05}
{"train_loss": 0.19368423521518707, "global_step": 108716, "epoch": 1221, "lr": 6.554069412169702e-05}
{"train_loss": 0.14049707353115082, "global_step": 108717, "epoch": 1221, "lr": 6.554014309794263e-05}
{"train_loss": 0.18078961968421936, "global_step": 108718, "epoch": 1221, "lr": 6.553959207209905e-05}
{"train_loss": 0.0854264497756958, "global_step": 108719, "epoch": 1221, "lr": 6.553904104416637e-05}
{"train_loss": 0.09994112700223923, "global_step": 108720, "epoch": 1221, "lr": 6.553849001414465e-05}
{"train_loss": 0.1464778482913971, "global_step": 108721, "epoch": 1221, "lr": 6.553793898203398e-05}
{"train_loss": 0.20865502953529358, "global_step": 108722, "epoch": 1221, "lr": 6.55373879478344e-05}
{"train_loss": 0.1504279524087906, "global_step": 108723, "epoch": 1221, "lr": 6.553683691154602e-05}
{"train_loss": 0.15157116949558258, "global_step": 108724, "epoch": 1221, "lr": 6.553628587316892e-05}
{"train_loss": 0.15877123177051544, "global_step": 108725, "epoch": 1221, "lr": 6.553573483270314e-05}
{"train_loss": 0.08342869579792023, "global_step": 108726, "epoch": 1221, "lr": 6.553518379014876e-05}
{"train_loss": 0.14541219174861908, "global_step": 108727, "epoch": 1221, "lr": 6.553463274550589e-05}
{"train_loss": 0.14926835894584656, "global_step": 108728, "epoch": 1221, "lr": 6.553408169877455e-05}
{"train_loss": 0.12424987554550171, "global_step": 108729, "epoch": 1221, "lr": 6.553353064995486e-05}
{"train_loss": 0.17727842926979065, "global_step": 108730, "epoch": 1221, "lr": 6.553297959904687e-05}
{"train_loss": 0.12293820083141327, "global_step": 108731, "epoch": 1221, "lr": 6.553242854605066e-05}
{"train_loss": 0.10625816881656647, "global_step": 108732, "epoch": 1221, "lr": 6.55318774909663e-05}
{"train_loss": 0.11293710023164749, "global_step": 108733, "epoch": 1221, "lr": 6.553132643379386e-05}
{"train_loss": 0.2172655612230301, "global_step": 108734, "epoch": 1221, "lr": 6.553077537453344e-05}
{"train_loss": 0.18399611115455627, "global_step": 108735, "epoch": 1221, "lr": 6.553022431318508e-05}
{"train_loss": 0.16942794620990753, "global_step": 108736, "epoch": 1221, "lr": 6.55296732497489e-05}
{"train_loss": 0.12544980645179749, "global_step": 108737, "epoch": 1221, "lr": 6.552912218422491e-05}
{"train_loss": 0.13960477709770203, "global_step": 108738, "epoch": 1221, "lr": 6.552857111661323e-05}
{"train_loss": 0.1457640826702118, "global_step": 108739, "epoch": 1221, "lr": 6.552802004691394e-05}
{"train_loss": 0.15428350865840912, "global_step": 108740, "epoch": 1221, "lr": 6.552746897512708e-05}
{"train_loss": 0.16201576590538025, "global_step": 108741, "epoch": 1221, "lr": 6.552691790125275e-05}
{"train_loss": 0.24431169033050537, "global_step": 108742, "epoch": 1221, "lr": 6.5526366825291e-05}
{"train_loss": 0.16819757223129272, "global_step": 108743, "epoch": 1221, "lr": 6.552581574724194e-05}
{"train_loss": 0.15603499114513397, "global_step": 108744, "epoch": 1221, "lr": 6.552526466710561e-05}
{"train_loss": 0.1462600827217102, "global_step": 108745, "epoch": 1221, "lr": 6.552471358488212e-05}
{"train_loss": 0.13071565330028534, "global_step": 108746, "epoch": 1221, "lr": 6.552416250057149e-05}
{"train_loss": 0.11233339458703995, "global_step": 108747, "epoch": 1221, "lr": 6.552361141417385e-05}
{"train_loss": 0.16734062135219574, "global_step": 108748, "epoch": 1221, "lr": 6.552306032568923e-05}
{"train_loss": 0.12953169643878937, "global_step": 108749, "epoch": 1221, "lr": 6.552250923511774e-05}
{"train_loss": 0.16632941365242004, "global_step": 108750, "epoch": 1221, "lr": 6.552195814245943e-05}
{"train_loss": 0.14803951978683472, "global_step": 108751, "epoch": 1221, "lr": 6.552140704771437e-05}
{"train_loss": 0.17952287197113037, "global_step": 108752, "epoch": 1221, "lr": 6.552085595088265e-05}
{"train_loss": 0.19195522367954254, "global_step": 108753, "epoch": 1221, "lr": 6.552030485196435e-05}
{"train_loss": 0.10336842387914658, "global_step": 108754, "epoch": 1221, "lr": 6.551975375095954e-05}
{"train_loss": 0.12478820979595184, "global_step": 108755, "epoch": 1221, "lr": 6.551920264786827e-05}
{"train_loss": 0.23991544544696808, "global_step": 108756, "epoch": 1221, "lr": 6.551865154269064e-05}
{"train_loss": 0.1497883992583564, "global_step": 108757, "epoch": 1221, "lr": 6.551810043542672e-05, "val_loss": 4.864487171173096}
{"train_loss": 0.15219585597515106, "global_step": 108758, "epoch": 1222, "lr": 6.551754932607657e-05}
{"train_loss": 0.1834680587053299, "global_step": 108759, "epoch": 1222, "lr": 6.551699821464029e-05}
{"train_loss": 0.19250881671905518, "global_step": 108760, "epoch": 1222, "lr": 6.551644710111794e-05}
{"train_loss": 0.0871199518442154, "global_step": 108761, "epoch": 1222, "lr": 6.551589598550958e-05}
{"train_loss": 0.1719948649406433, "global_step": 108762, "epoch": 1222, "lr": 6.55153448678153e-05}
{"train_loss": 0.12657466530799866, "global_step": 108763, "epoch": 1222, "lr": 6.551479374803518e-05}
{"train_loss": 0.16718804836273193, "global_step": 108764, "epoch": 1222, "lr": 6.551424262616929e-05}
{"train_loss": 0.1504133939743042, "global_step": 108765, "epoch": 1222, "lr": 6.551369150221769e-05}
{"train_loss": 0.11418740451335907, "global_step": 108766, "epoch": 1222, "lr": 6.551314037618045e-05}
{"train_loss": 0.11390751600265503, "global_step": 108767, "epoch": 1222, "lr": 6.551258924805767e-05}
{"train_loss": 0.13079078495502472, "global_step": 108768, "epoch": 1222, "lr": 6.55120381178494e-05}
{"train_loss": 0.17236371338367462, "global_step": 108769, "epoch": 1222, "lr": 6.551148698555574e-05}
{"train_loss": 0.13832685351371765, "global_step": 108770, "epoch": 1222, "lr": 6.551093585117673e-05}
{"train_loss": 0.20038510859012604, "global_step": 108771, "epoch": 1222, "lr": 6.55103847147125e-05}
{"train_loss": 0.14992986619472504, "global_step": 108772, "epoch": 1222, "lr": 6.550983357616305e-05}
{"train_loss": 0.16979505121707916, "global_step": 108773, "epoch": 1222, "lr": 6.550928243552852e-05}
{"train_loss": 0.171925887465477, "global_step": 108774, "epoch": 1222, "lr": 6.550873129280894e-05}
{"train_loss": 0.09828970581293106, "global_step": 108775, "epoch": 1222, "lr": 6.55081801480044e-05}
{"train_loss": 0.14565183222293854, "global_step": 108776, "epoch": 1222, "lr": 6.550762900111498e-05}
{"train_loss": 0.16377916932106018, "global_step": 108777, "epoch": 1222, "lr": 6.550707785214074e-05}
{"train_loss": 0.1565028578042984, "global_step": 108778, "epoch": 1222, "lr": 6.550652670108178e-05}
{"train_loss": 0.16688866913318634, "global_step": 108779, "epoch": 1222, "lr": 6.550597554793813e-05}
{"train_loss": 0.10819625854492188, "global_step": 108780, "epoch": 1222, "lr": 6.550542439270992e-05}
{"train_loss": 0.19334706664085388, "global_step": 108781, "epoch": 1222, "lr": 6.550487323539717e-05}
{"train_loss": 0.18388932943344116, "global_step": 108782, "epoch": 1222, "lr": 6.550432207599999e-05}
{"train_loss": 0.1430155485868454, "global_step": 108783, "epoch": 1222, "lr": 6.550377091451846e-05}
{"train_loss": 0.24918678402900696, "global_step": 108784, "epoch": 1222, "lr": 6.550321975095261e-05}
{"train_loss": 0.1560315042734146, "global_step": 108785, "epoch": 1222, "lr": 6.550266858530256e-05}
{"train_loss": 0.18992365896701813, "global_step": 108786, "epoch": 1222, "lr": 6.550211741756835e-05}
{"train_loss": 0.18550294637680054, "global_step": 108787, "epoch": 1222, "lr": 6.550156624775007e-05}
{"train_loss": 0.1393885463476181, "global_step": 108788, "epoch": 1222, "lr": 6.55010150758478e-05}
{"train_loss": 0.13029202818870544, "global_step": 108789, "epoch": 1222, "lr": 6.550046390186159e-05}
{"train_loss": 0.16537892818450928, "global_step": 108790, "epoch": 1222, "lr": 6.549991272579156e-05}
{"train_loss": 0.17479047179222107, "global_step": 108791, "epoch": 1222, "lr": 6.549936154763774e-05}
{"train_loss": 0.18883122503757477, "global_step": 108792, "epoch": 1222, "lr": 6.549881036740022e-05}
{"train_loss": 0.14299792051315308, "global_step": 108793, "epoch": 1222, "lr": 6.549825918507908e-05}
{"train_loss": 0.1836833655834198, "global_step": 108794, "epoch": 1222, "lr": 6.549770800067439e-05}
{"train_loss": 0.1276032030582428, "global_step": 108795, "epoch": 1222, "lr": 6.549715681418622e-05}
{"train_loss": 0.16667495667934418, "global_step": 108796, "epoch": 1222, "lr": 6.549660562561464e-05}
{"train_loss": 0.12806855142116547, "global_step": 108797, "epoch": 1222, "lr": 6.549605443495972e-05}
{"train_loss": 0.14069324731826782, "global_step": 108798, "epoch": 1222, "lr": 6.549550324222157e-05}
{"train_loss": 0.17235270142555237, "global_step": 108799, "epoch": 1222, "lr": 6.549495204740022e-05}
{"train_loss": 0.1703304797410965, "global_step": 108800, "epoch": 1222, "lr": 6.549440085049577e-05}
{"train_loss": 0.14816392958164215, "global_step": 108801, "epoch": 1222, "lr": 6.54938496515083e-05}
{"train_loss": 0.25147587060928345, "global_step": 108802, "epoch": 1222, "lr": 6.549329845043783e-05}
{"train_loss": 0.09338653087615967, "global_step": 108803, "epoch": 1222, "lr": 6.549274724728452e-05}
{"train_loss": 0.19748112559318542, "global_step": 108804, "epoch": 1222, "lr": 6.549219604204839e-05}
{"train_loss": 0.14013096690177917, "global_step": 108805, "epoch": 1222, "lr": 6.549164483472949e-05}
{"train_loss": 0.12545059621334076, "global_step": 108806, "epoch": 1222, "lr": 6.549109362532795e-05}
{"train_loss": 0.09994815289974213, "global_step": 108807, "epoch": 1222, "lr": 6.549054241384382e-05}
{"train_loss": 0.10180787742137909, "global_step": 108808, "epoch": 1222, "lr": 6.548999120027718e-05}
{"train_loss": 0.14034801721572876, "global_step": 108809, "epoch": 1222, "lr": 6.54894399846281e-05}
{"train_loss": 0.1747882068157196, "global_step": 108810, "epoch": 1222, "lr": 6.548888876689665e-05}
{"train_loss": 0.11343316733837128, "global_step": 108811, "epoch": 1222, "lr": 6.54883375470829e-05}
{"train_loss": 0.11154720932245255, "global_step": 108812, "epoch": 1222, "lr": 6.548778632518693e-05}
{"train_loss": 0.11944684386253357, "global_step": 108813, "epoch": 1222, "lr": 6.548723510120884e-05}
{"train_loss": 0.1392674595117569, "global_step": 108814, "epoch": 1222, "lr": 6.548668387514865e-05}
{"train_loss": 0.17753608524799347, "global_step": 108815, "epoch": 1222, "lr": 6.548613264700649e-05}
{"train_loss": 0.15683718025684357, "global_step": 108816, "epoch": 1222, "lr": 6.548558141678241e-05}
{"train_loss": 0.21554026007652283, "global_step": 108817, "epoch": 1222, "lr": 6.548503018447646e-05}
{"train_loss": 0.1345514953136444, "global_step": 108818, "epoch": 1222, "lr": 6.548447895008875e-05}
{"train_loss": 0.13280348479747772, "global_step": 108819, "epoch": 1222, "lr": 6.548392771361933e-05}
{"train_loss": 0.1439274400472641, "global_step": 108820, "epoch": 1222, "lr": 6.548337647506829e-05}
{"train_loss": 0.14193539321422577, "global_step": 108821, "epoch": 1222, "lr": 6.54828252344357e-05}
{"train_loss": 0.2053266167640686, "global_step": 108822, "epoch": 1222, "lr": 6.548227399172164e-05}
{"train_loss": 0.09700236469507217, "global_step": 108823, "epoch": 1222, "lr": 6.548172274692617e-05}
{"train_loss": 0.1721220165491104, "global_step": 108824, "epoch": 1222, "lr": 6.548117150004937e-05}
{"train_loss": 0.1202412024140358, "global_step": 108825, "epoch": 1222, "lr": 6.548062025109131e-05}
{"train_loss": 0.14078906178474426, "global_step": 108826, "epoch": 1222, "lr": 6.548006900005208e-05}
{"train_loss": 0.13797889649868011, "global_step": 108827, "epoch": 1222, "lr": 6.547951774693173e-05}
{"train_loss": 0.1655493974685669, "global_step": 108828, "epoch": 1222, "lr": 6.547896649173036e-05}
{"train_loss": 0.10822930186986923, "global_step": 108829, "epoch": 1222, "lr": 6.547841523444802e-05}
{"train_loss": 0.1981774866580963, "global_step": 108830, "epoch": 1222, "lr": 6.54778639750848e-05}
{"train_loss": 0.1584549993276596, "global_step": 108831, "epoch": 1222, "lr": 6.547731271364079e-05}
{"train_loss": 0.13537296652793884, "global_step": 108832, "epoch": 1222, "lr": 6.547676145011602e-05}
{"train_loss": 0.07889308780431747, "global_step": 108833, "epoch": 1222, "lr": 6.54762101845106e-05}
{"train_loss": 0.15108680725097656, "global_step": 108834, "epoch": 1222, "lr": 6.547565891682459e-05}
{"train_loss": 0.1317334920167923, "global_step": 108835, "epoch": 1222, "lr": 6.547510764705805e-05}
{"train_loss": 0.20398926734924316, "global_step": 108836, "epoch": 1222, "lr": 6.54745563752111e-05}
{"train_loss": 0.15872354805469513, "global_step": 108837, "epoch": 1222, "lr": 6.547400510128377e-05}
{"train_loss": 0.15694081783294678, "global_step": 108838, "epoch": 1222, "lr": 6.547345382527615e-05}
{"train_loss": 0.09810531884431839, "global_step": 108839, "epoch": 1222, "lr": 6.54729025471883e-05}
{"train_loss": 0.18185259401798248, "global_step": 108840, "epoch": 1222, "lr": 6.547235126702031e-05}
{"train_loss": 0.16248464584350586, "global_step": 108841, "epoch": 1222, "lr": 6.547179998477226e-05}
{"train_loss": 0.18720099329948425, "global_step": 108842, "epoch": 1222, "lr": 6.547124870044421e-05}
{"train_loss": 0.1218491643667221, "global_step": 108843, "epoch": 1222, "lr": 6.547069741403624e-05}
{"train_loss": 0.14172635972499847, "global_step": 108844, "epoch": 1222, "lr": 6.547014612554842e-05}
{"train_loss": 0.15492181479930878, "global_step": 108845, "epoch": 1222, "lr": 6.546959483498084e-05}
{"train_loss": 0.15302695258614724, "global_step": 108846, "epoch": 1222, "lr": 6.546904354233356e-05, "val_loss": 4.994750499725342}
{"train_loss": 0.11213550716638565, "global_step": 108847, "epoch": 1223, "lr": 6.546849224760665e-05}
{"train_loss": 0.09699225425720215, "global_step": 108848, "epoch": 1223, "lr": 6.546794095080019e-05}
{"train_loss": 0.24101759493350983, "global_step": 108849, "epoch": 1223, "lr": 6.546738965191425e-05}
{"train_loss": 0.24022622406482697, "global_step": 108850, "epoch": 1223, "lr": 6.546683835094891e-05}
{"train_loss": 0.10492703318595886, "global_step": 108851, "epoch": 1223, "lr": 6.546628704790424e-05}
{"train_loss": 0.1030915230512619, "global_step": 108852, "epoch": 1223, "lr": 6.546573574278032e-05}
{"train_loss": 0.12938353419303894, "global_step": 108853, "epoch": 1223, "lr": 6.546518443557723e-05}
{"train_loss": 0.11233270913362503, "global_step": 108854, "epoch": 1223, "lr": 6.546463312629503e-05}
{"train_loss": 0.12020401656627655, "global_step": 108855, "epoch": 1223, "lr": 6.546408181493378e-05}
{"train_loss": 0.13314189016819, "global_step": 108856, "epoch": 1223, "lr": 6.546353050149358e-05}
{"train_loss": 0.0910540372133255, "global_step": 108857, "epoch": 1223, "lr": 6.546297918597451e-05}
{"train_loss": 0.1979648619890213, "global_step": 108858, "epoch": 1223, "lr": 6.546242786837662e-05}
{"train_loss": 0.12476561218500137, "global_step": 108859, "epoch": 1223, "lr": 6.546187654870001e-05}
{"train_loss": 0.2384689450263977, "global_step": 108860, "epoch": 1223, "lr": 6.546132522694473e-05}
{"train_loss": 0.17632904648780823, "global_step": 108861, "epoch": 1223, "lr": 6.546077390311085e-05}
{"train_loss": 0.21611805260181427, "global_step": 108862, "epoch": 1223, "lr": 6.546022257719847e-05}
{"train_loss": 0.2161046415567398, "global_step": 108863, "epoch": 1223, "lr": 6.545967124920765e-05}
{"train_loss": 0.1853538602590561, "global_step": 108864, "epoch": 1223, "lr": 6.545911991913847e-05}
{"train_loss": 0.13569627702236176, "global_step": 108865, "epoch": 1223, "lr": 6.5458568586991e-05}
{"train_loss": 0.20022781193256378, "global_step": 108866, "epoch": 1223, "lr": 6.54580172527653e-05}
{"train_loss": 0.1547822803258896, "global_step": 108867, "epoch": 1223, "lr": 6.545746591646147e-05}
{"train_loss": 0.12051307410001755, "global_step": 108868, "epoch": 1223, "lr": 6.545691457807956e-05}
{"train_loss": 0.20438683032989502, "global_step": 108869, "epoch": 1223, "lr": 6.545636323761968e-05}
{"train_loss": 0.16489624977111816, "global_step": 108870, "epoch": 1223, "lr": 6.545581189508185e-05}
{"train_loss": 0.14766639471054077, "global_step": 108871, "epoch": 1223, "lr": 6.54552605504662e-05}
{"train_loss": 0.14271603524684906, "global_step": 108872, "epoch": 1223, "lr": 6.545470920377276e-05}
{"train_loss": 0.18881192803382874, "global_step": 108873, "epoch": 1223, "lr": 6.545415785500164e-05}
{"train_loss": 0.14557664096355438, "global_step": 108874, "epoch": 1223, "lr": 6.545360650415289e-05}
{"train_loss": 0.14306001365184784, "global_step": 108875, "epoch": 1223, "lr": 6.545305515122659e-05}
{"train_loss": 0.08765068650245667, "global_step": 108876, "epoch": 1223, "lr": 6.545250379622281e-05}
{"train_loss": 0.16721220314502716, "global_step": 108877, "epoch": 1223, "lr": 6.545195243914164e-05}
{"train_loss": 0.10388878732919693, "global_step": 108878, "epoch": 1223, "lr": 6.545140107998313e-05}
{"train_loss": 0.17294684052467346, "global_step": 108879, "epoch": 1223, "lr": 6.545084971874738e-05}
{"train_loss": 0.2266443520784378, "global_step": 108880, "epoch": 1223, "lr": 6.545029835543444e-05}
{"train_loss": 0.1418641358613968, "global_step": 108881, "epoch": 1223, "lr": 6.544974699004441e-05}
{"train_loss": 0.16959765553474426, "global_step": 108882, "epoch": 1223, "lr": 6.544919562257734e-05}
{"train_loss": 0.15746387839317322, "global_step": 108883, "epoch": 1223, "lr": 6.544864425303331e-05}
{"train_loss": 0.16226151585578918, "global_step": 108884, "epoch": 1223, "lr": 6.54480928814124e-05}
{"train_loss": 0.18105114996433258, "global_step": 108885, "epoch": 1223, "lr": 6.54475415077147e-05}
{"train_loss": 0.10474635660648346, "global_step": 108886, "epoch": 1223, "lr": 6.544699013194025e-05}
{"train_loss": 0.16481351852416992, "global_step": 108887, "epoch": 1223, "lr": 6.544643875408913e-05}
{"train_loss": 0.16272245347499847, "global_step": 108888, "epoch": 1223, "lr": 6.544588737416144e-05}
{"train_loss": 0.1322786659002304, "global_step": 108889, "epoch": 1223, "lr": 6.544533599215724e-05}
{"train_loss": 0.22193345427513123, "global_step": 108890, "epoch": 1223, "lr": 6.54447846080766e-05}
{"train_loss": 0.3054231107234955, "global_step": 108891, "epoch": 1223, "lr": 6.544423322191959e-05}
{"train_loss": 0.20998573303222656, "global_step": 108892, "epoch": 1223, "lr": 6.54436818336863e-05}
{"train_loss": 0.11112719774246216, "global_step": 108893, "epoch": 1223, "lr": 6.544313044337679e-05}
{"train_loss": 0.15024419128894806, "global_step": 108894, "epoch": 1223, "lr": 6.544257905099113e-05}
{"train_loss": 0.09224773198366165, "global_step": 108895, "epoch": 1223, "lr": 6.544202765652943e-05}
{"train_loss": 0.09908080101013184, "global_step": 108896, "epoch": 1223, "lr": 6.544147625999172e-05}
{"train_loss": 0.19744201004505157, "global_step": 108897, "epoch": 1223, "lr": 6.54409248613781e-05}
{"train_loss": 0.11094391345977783, "global_step": 108898, "epoch": 1223, "lr": 6.544037346068861e-05}
{"train_loss": 0.1662912517786026, "global_step": 108899, "epoch": 1223, "lr": 6.543982205792338e-05}
{"train_loss": 0.20467627048492432, "global_step": 108900, "epoch": 1223, "lr": 6.543927065308245e-05}
{"train_loss": 0.13866594433784485, "global_step": 108901, "epoch": 1223, "lr": 6.543871924616589e-05}
{"train_loss": 0.110521599650383, "global_step": 108902, "epoch": 1223, "lr": 6.543816783717378e-05}
{"train_loss": 0.15136384963989258, "global_step": 108903, "epoch": 1223, "lr": 6.543761642610621e-05}
{"train_loss": 0.1458987444639206, "global_step": 108904, "epoch": 1223, "lr": 6.543706501296322e-05}
{"train_loss": 0.16601094603538513, "global_step": 108905, "epoch": 1223, "lr": 6.543651359774491e-05}
{"train_loss": 0.13372419774532318, "global_step": 108906, "epoch": 1223, "lr": 6.543596218045137e-05}
{"train_loss": 0.1701282560825348, "global_step": 108907, "epoch": 1223, "lr": 6.543541076108264e-05}
{"train_loss": 0.08693783730268478, "global_step": 108908, "epoch": 1223, "lr": 6.543485933963879e-05}
{"train_loss": 0.11455267667770386, "global_step": 108909, "epoch": 1223, "lr": 6.543430791611994e-05}
{"train_loss": 0.1543954759836197, "global_step": 108910, "epoch": 1223, "lr": 6.543375649052613e-05}
{"train_loss": 0.148158997297287, "global_step": 108911, "epoch": 1223, "lr": 6.543320506285742e-05}
{"train_loss": 0.17875537276268005, "global_step": 108912, "epoch": 1223, "lr": 6.543265363311392e-05}
{"train_loss": 0.149369016289711, "global_step": 108913, "epoch": 1223, "lr": 6.543210220129568e-05}
{"train_loss": 0.19233748316764832, "global_step": 108914, "epoch": 1223, "lr": 6.543155076740279e-05}
{"train_loss": 0.12127917259931564, "global_step": 108915, "epoch": 1223, "lr": 6.54309993314353e-05}
{"train_loss": 0.301967591047287, "global_step": 108916, "epoch": 1223, "lr": 6.543044789339332e-05}
{"train_loss": 0.15853026509284973, "global_step": 108917, "epoch": 1223, "lr": 6.54298964532769e-05}
{"train_loss": 0.14607614278793335, "global_step": 108918, "epoch": 1223, "lr": 6.54293450110861e-05}
{"train_loss": 0.10465659946203232, "global_step": 108919, "epoch": 1223, "lr": 6.542879356682103e-05}
{"train_loss": 0.17744508385658264, "global_step": 108920, "epoch": 1223, "lr": 6.542824212048176e-05}
{"train_loss": 0.16117489337921143, "global_step": 108921, "epoch": 1223, "lr": 6.542769067206832e-05}
{"train_loss": 0.14616434276103973, "global_step": 108922, "epoch": 1223, "lr": 6.542713922158083e-05}
{"train_loss": 0.14384332299232483, "global_step": 108923, "epoch": 1223, "lr": 6.542658776901935e-05}
{"train_loss": 0.1426856964826584, "global_step": 108924, "epoch": 1223, "lr": 6.542603631438395e-05}
{"train_loss": 0.11372524499893188, "global_step": 108925, "epoch": 1223, "lr": 6.542548485767471e-05}
{"train_loss": 0.13430018723011017, "global_step": 108926, "epoch": 1223, "lr": 6.542493339889171e-05}
{"train_loss": 0.15416735410690308, "global_step": 108927, "epoch": 1223, "lr": 6.5424381938035e-05}
{"train_loss": 0.24922651052474976, "global_step": 108928, "epoch": 1223, "lr": 6.542383047510467e-05}
{"train_loss": 0.21313506364822388, "global_step": 108929, "epoch": 1223, "lr": 6.542327901010082e-05}
{"train_loss": 0.1271687000989914, "global_step": 108930, "epoch": 1223, "lr": 6.542272754302346e-05}
{"train_loss": 0.16226211190223694, "global_step": 108931, "epoch": 1223, "lr": 6.542217607387274e-05}
{"train_loss": 0.12830936908721924, "global_step": 108932, "epoch": 1223, "lr": 6.542162460264865e-05}
{"train_loss": 0.2211301475763321, "global_step": 108933, "epoch": 1223, "lr": 6.542107312935135e-05}
{"train_loss": 0.11840076744556427, "global_step": 108934, "epoch": 1223, "lr": 6.542052165398087e-05}
{"train_loss": 0.157900065937069, "global_step": 108935, "epoch": 1223, "lr": 6.541997017653728e-05, "val_loss": 4.918015480041504}
{"train_loss": 0.1764645278453827, "global_step": 108936, "epoch": 1224, "lr": 6.541941869702066e-05}
{"train_loss": 0.09118001163005829, "global_step": 108937, "epoch": 1224, "lr": 6.541886721543109e-05}
{"train_loss": 0.1518585979938507, "global_step": 108938, "epoch": 1224, "lr": 6.541831573176864e-05}
{"train_loss": 0.19376212358474731, "global_step": 108939, "epoch": 1224, "lr": 6.54177642460334e-05}
{"train_loss": 0.18538573384284973, "global_step": 108940, "epoch": 1224, "lr": 6.541721275822541e-05}
{"train_loss": 0.17699328064918518, "global_step": 108941, "epoch": 1224, "lr": 6.541666126834477e-05}
{"train_loss": 0.17323367297649384, "global_step": 108942, "epoch": 1224, "lr": 6.541610977639155e-05}
{"train_loss": 0.1780342161655426, "global_step": 108943, "epoch": 1224, "lr": 6.541555828236584e-05}
{"train_loss": 0.11962628364562988, "global_step": 108944, "epoch": 1224, "lr": 6.541500678626766e-05}
{"train_loss": 0.13320857286453247, "global_step": 108945, "epoch": 1224, "lr": 6.541445528809713e-05}
{"train_loss": 0.11182598024606705, "global_step": 108946, "epoch": 1224, "lr": 6.541390378785432e-05}
{"train_loss": 0.10394825786352158, "global_step": 108947, "epoch": 1224, "lr": 6.541335228553931e-05}
{"train_loss": 0.19069881737232208, "global_step": 108948, "epoch": 1224, "lr": 6.541280078115215e-05}
{"train_loss": 0.13021031022071838, "global_step": 108949, "epoch": 1224, "lr": 6.541224927469295e-05}
{"train_loss": 0.13919563591480255, "global_step": 108950, "epoch": 1224, "lr": 6.541169776616172e-05}
{"train_loss": 0.13194793462753296, "global_step": 108951, "epoch": 1224, "lr": 6.541114625555858e-05}
{"train_loss": 0.12090664356946945, "global_step": 108952, "epoch": 1224, "lr": 6.541059474288363e-05}
{"train_loss": 0.13971443474292755, "global_step": 108953, "epoch": 1224, "lr": 6.54100432281369e-05}
{"train_loss": 0.11929876357316971, "global_step": 108954, "epoch": 1224, "lr": 6.540949171131847e-05}
{"train_loss": 0.09042712301015854, "global_step": 108955, "epoch": 1224, "lr": 6.540894019242844e-05}
{"train_loss": 0.16215009987354279, "global_step": 108956, "epoch": 1224, "lr": 6.540838867146685e-05}
{"train_loss": 0.11920847743749619, "global_step": 108957, "epoch": 1224, "lr": 6.540783714843379e-05}
{"train_loss": 0.19867227971553802, "global_step": 108958, "epoch": 1224, "lr": 6.540728562332934e-05}
{"train_loss": 0.14747825264930725, "global_step": 108959, "epoch": 1224, "lr": 6.540673409615357e-05}
{"train_loss": 0.157352015376091, "global_step": 108960, "epoch": 1224, "lr": 6.540618256690654e-05}
{"train_loss": 0.12550513446331024, "global_step": 108961, "epoch": 1224, "lr": 6.540563103558834e-05}
{"train_loss": 0.1820652186870575, "global_step": 108962, "epoch": 1224, "lr": 6.540507950219906e-05}
{"train_loss": 0.14238250255584717, "global_step": 108963, "epoch": 1224, "lr": 6.540452796673873e-05}
{"train_loss": 0.17131134867668152, "global_step": 108964, "epoch": 1224, "lr": 6.540397642920746e-05}
{"train_loss": 0.1263955980539322, "global_step": 108965, "epoch": 1224, "lr": 6.540342488960531e-05}
{"train_loss": 0.12603428959846497, "global_step": 108966, "epoch": 1224, "lr": 6.540287334793235e-05}
{"train_loss": 0.16373683512210846, "global_step": 108967, "epoch": 1224, "lr": 6.540232180418868e-05}
{"train_loss": 0.09165052324533463, "global_step": 108968, "epoch": 1224, "lr": 6.540177025837434e-05}
{"train_loss": 0.14731453359127045, "global_step": 108969, "epoch": 1224, "lr": 6.540121871048942e-05}
{"train_loss": 0.12934111058712006, "global_step": 108970, "epoch": 1224, "lr": 6.5400667160534e-05}
{"train_loss": 0.11114712804555893, "global_step": 108971, "epoch": 1224, "lr": 6.540011560850815e-05}
{"train_loss": 0.14362573623657227, "global_step": 108972, "epoch": 1224, "lr": 6.539956405441194e-05}
{"train_loss": 0.1660728007555008, "global_step": 108973, "epoch": 1224, "lr": 6.539901249824546e-05}
{"train_loss": 0.12170916050672531, "global_step": 108974, "epoch": 1224, "lr": 6.539846094000875e-05}
{"train_loss": 0.14515820145606995, "global_step": 108975, "epoch": 1224, "lr": 6.539790937970191e-05}
{"train_loss": 0.13023462891578674, "global_step": 108976, "epoch": 1224, "lr": 6.539735781732502e-05}
{"train_loss": 0.15985476970672607, "global_step": 108977, "epoch": 1224, "lr": 6.539680625287812e-05}
{"train_loss": 0.10568080097436905, "global_step": 108978, "epoch": 1224, "lr": 6.539625468636132e-05}
{"train_loss": 0.1414753794670105, "global_step": 108979, "epoch": 1224, "lr": 6.539570311777469e-05}
{"train_loss": 0.19958357512950897, "global_step": 108980, "epoch": 1224, "lr": 6.539515154711829e-05}
{"train_loss": 0.17416459321975708, "global_step": 108981, "epoch": 1224, "lr": 6.53945999743922e-05}
{"train_loss": 0.17553141713142395, "global_step": 108982, "epoch": 1224, "lr": 6.539404839959648e-05}
{"train_loss": 0.14358283579349518, "global_step": 108983, "epoch": 1224, "lr": 6.539349682273123e-05}
{"train_loss": 0.14479093253612518, "global_step": 108984, "epoch": 1224, "lr": 6.539294524379651e-05}
{"train_loss": 0.16644394397735596, "global_step": 108985, "epoch": 1224, "lr": 6.53923936627924e-05}
{"train_loss": 0.20736661553382874, "global_step": 108986, "epoch": 1224, "lr": 6.539184207971895e-05}
{"train_loss": 0.1120784729719162, "global_step": 108987, "epoch": 1224, "lr": 6.539129049457627e-05}
{"train_loss": 0.1849343329668045, "global_step": 108988, "epoch": 1224, "lr": 6.539073890736442e-05}
{"train_loss": 0.05362025275826454, "global_step": 108989, "epoch": 1224, "lr": 6.539018731808346e-05}
{"train_loss": 0.14225557446479797, "global_step": 108990, "epoch": 1224, "lr": 6.53896357267335e-05}
{"train_loss": 0.14370238780975342, "global_step": 108991, "epoch": 1224, "lr": 6.53890841333146e-05}
{"train_loss": 0.11490768194198608, "global_step": 108992, "epoch": 1224, "lr": 6.53885325378268e-05}
{"train_loss": 0.1899230182170868, "global_step": 108993, "epoch": 1224, "lr": 6.53879809402702e-05}
{"train_loss": 0.13785526156425476, "global_step": 108994, "epoch": 1224, "lr": 6.538742934064488e-05}
{"train_loss": 0.2036634385585785, "global_step": 108995, "epoch": 1224, "lr": 6.53868777389509e-05}
{"train_loss": 0.08126044273376465, "global_step": 108996, "epoch": 1224, "lr": 6.538632613518834e-05}
{"train_loss": 0.15325939655303955, "global_step": 108997, "epoch": 1224, "lr": 6.538577452935729e-05}
{"train_loss": 0.20204445719718933, "global_step": 108998, "epoch": 1224, "lr": 6.53852229214578e-05}
{"train_loss": 0.20252269506454468, "global_step": 108999, "epoch": 1224, "lr": 6.538467131148997e-05}
{"train_loss": 0.13923394680023193, "global_step": 109000, "epoch": 1224, "lr": 6.538411969945384e-05}
{"train_loss": 0.11200923472642899, "global_step": 109001, "epoch": 1224, "lr": 6.538356808534951e-05}
{"train_loss": 0.18577218055725098, "global_step": 109002, "epoch": 1224, "lr": 6.538301646917704e-05}
{"train_loss": 0.1612500697374344, "global_step": 109003, "epoch": 1224, "lr": 6.538246485093652e-05}
{"train_loss": 0.17309723794460297, "global_step": 109004, "epoch": 1224, "lr": 6.538191323062802e-05}
{"train_loss": 0.12748634815216064, "global_step": 109005, "epoch": 1224, "lr": 6.538136160825158e-05}
{"train_loss": 0.18857504427433014, "global_step": 109006, "epoch": 1224, "lr": 6.538080998380732e-05}
{"train_loss": 0.1072152778506279, "global_step": 109007, "epoch": 1224, "lr": 6.538025835729532e-05}
{"train_loss": 0.16047897934913635, "global_step": 109008, "epoch": 1224, "lr": 6.53797067287156e-05}
{"train_loss": 0.16395847499370575, "global_step": 109009, "epoch": 1224, "lr": 6.537915509806828e-05}
{"train_loss": 0.1211218610405922, "global_step": 109010, "epoch": 1224, "lr": 6.537860346535342e-05}
{"train_loss": 0.11375067383050919, "global_step": 109011, "epoch": 1224, "lr": 6.537805183057111e-05}
{"train_loss": 0.11550534516572952, "global_step": 109012, "epoch": 1224, "lr": 6.537750019372138e-05}
{"train_loss": 0.17202942073345184, "global_step": 109013, "epoch": 1224, "lr": 6.537694855480436e-05}
{"train_loss": 0.1908838450908661, "global_step": 109014, "epoch": 1224, "lr": 6.537639691382007e-05}
{"train_loss": 0.17781665921211243, "global_step": 109015, "epoch": 1224, "lr": 6.537584527076862e-05}
{"train_loss": 0.11881666630506516, "global_step": 109016, "epoch": 1224, "lr": 6.537529362565008e-05}
{"train_loss": 0.2551236152648926, "global_step": 109017, "epoch": 1224, "lr": 6.537474197846451e-05}
{"train_loss": 0.10708216577768326, "global_step": 109018, "epoch": 1224, "lr": 6.5374190329212e-05}
{"train_loss": 0.13196121156215668, "global_step": 109019, "epoch": 1224, "lr": 6.537363867789262e-05}
{"train_loss": 0.16608910262584686, "global_step": 109020, "epoch": 1224, "lr": 6.537308702450644e-05}
{"train_loss": 0.18569691479206085, "global_step": 109021, "epoch": 1224, "lr": 6.537253536905353e-05}
{"train_loss": 0.18034516274929047, "global_step": 109022, "epoch": 1224, "lr": 6.537198371153398e-05}
{"train_loss": 0.1784329116344452, "global_step": 109023, "epoch": 1224, "lr": 6.537143205194783e-05}
{"train_loss": 0.14920220327343833, "global_step": 109024, "epoch": 1224, "lr": 6.537088039029519e-05, "val_loss": 4.897059917449951}
{"train_loss": 0.17488472163677216, "global_step": 109025, "epoch": 1225, "lr": 6.537032872657614e-05}
{"train_loss": 0.13385778665542603, "global_step": 109026, "epoch": 1225, "lr": 6.536977706079073e-05}
{"train_loss": 0.16269463300704956, "global_step": 109027, "epoch": 1225, "lr": 6.536922539293903e-05}
{"train_loss": 0.10505834221839905, "global_step": 109028, "epoch": 1225, "lr": 6.536867372302114e-05}
{"train_loss": 0.12929145991802216, "global_step": 109029, "epoch": 1225, "lr": 6.536812205103711e-05}
{"train_loss": 0.16730022430419922, "global_step": 109030, "epoch": 1225, "lr": 6.536757037698703e-05}
{"train_loss": 0.1723424345254898, "global_step": 109031, "epoch": 1225, "lr": 6.536701870087097e-05}
{"train_loss": 0.19361761212348938, "global_step": 109032, "epoch": 1225, "lr": 6.536646702268898e-05}
{"train_loss": 0.21682998538017273, "global_step": 109033, "epoch": 1225, "lr": 6.536591534244117e-05}
{"train_loss": 0.11881586164236069, "global_step": 109034, "epoch": 1225, "lr": 6.536536366012761e-05}
{"train_loss": 0.1083756759762764, "global_step": 109035, "epoch": 1225, "lr": 6.536481197574835e-05}
{"train_loss": 0.22445714473724365, "global_step": 109036, "epoch": 1225, "lr": 6.536426028930348e-05}
{"train_loss": 0.2915881872177124, "global_step": 109037, "epoch": 1225, "lr": 6.536370860079308e-05}
{"train_loss": 0.16538649797439575, "global_step": 109038, "epoch": 1225, "lr": 6.53631569102172e-05}
{"train_loss": 0.07303310930728912, "global_step": 109039, "epoch": 1225, "lr": 6.536260521757594e-05}
{"train_loss": 0.19628340005874634, "global_step": 109040, "epoch": 1225, "lr": 6.536205352286937e-05}
{"train_loss": 0.11775648593902588, "global_step": 109041, "epoch": 1225, "lr": 6.536150182609755e-05}
{"train_loss": 0.2011204957962036, "global_step": 109042, "epoch": 1225, "lr": 6.536095012726057e-05}
{"train_loss": 0.18528461456298828, "global_step": 109043, "epoch": 1225, "lr": 6.53603984263585e-05}
{"train_loss": 0.14813737571239471, "global_step": 109044, "epoch": 1225, "lr": 6.53598467233914e-05}
{"train_loss": 0.15982097387313843, "global_step": 109045, "epoch": 1225, "lr": 6.535929501835937e-05}
{"train_loss": 0.17846885323524475, "global_step": 109046, "epoch": 1225, "lr": 6.535874331126246e-05}
{"train_loss": 0.2652812898159027, "global_step": 109047, "epoch": 1225, "lr": 6.535819160210076e-05}
{"train_loss": 0.16512541472911835, "global_step": 109048, "epoch": 1225, "lr": 6.535763989087433e-05}
{"train_loss": 0.19197076559066772, "global_step": 109049, "epoch": 1225, "lr": 6.535708817758327e-05}
{"train_loss": 0.1890251189470291, "global_step": 109050, "epoch": 1225, "lr": 6.535653646222763e-05}
{"train_loss": 0.0944262221455574, "global_step": 109051, "epoch": 1225, "lr": 6.535598474480748e-05}
{"train_loss": 0.10835499316453934, "global_step": 109052, "epoch": 1225, "lr": 6.535543302532292e-05}
{"train_loss": 0.09662950783967972, "global_step": 109053, "epoch": 1225, "lr": 6.5354881303774e-05}
{"train_loss": 0.130606010556221, "global_step": 109054, "epoch": 1225, "lr": 6.53543295801608e-05}
{"train_loss": 0.1587166041135788, "global_step": 109055, "epoch": 1225, "lr": 6.53537778544834e-05}
{"train_loss": 0.13986732065677643, "global_step": 109056, "epoch": 1225, "lr": 6.535322612674188e-05}
{"train_loss": 0.14713917672634125, "global_step": 109057, "epoch": 1225, "lr": 6.535267439693629e-05}
{"train_loss": 0.1481849104166031, "global_step": 109058, "epoch": 1225, "lr": 6.535212266506674e-05}
{"train_loss": 0.14310984313488007, "global_step": 109059, "epoch": 1225, "lr": 6.535157093113325e-05}
{"train_loss": 0.23832620680332184, "global_step": 109060, "epoch": 1225, "lr": 6.535101919513595e-05}
{"train_loss": 0.11653097718954086, "global_step": 109061, "epoch": 1225, "lr": 6.535046745707491e-05}
{"train_loss": 0.24150238931179047, "global_step": 109062, "epoch": 1225, "lr": 6.534991571695016e-05}
{"train_loss": 0.1339573860168457, "global_step": 109063, "epoch": 1225, "lr": 6.534936397476182e-05}
{"train_loss": 0.21479371190071106, "global_step": 109064, "epoch": 1225, "lr": 6.534881223050993e-05}
{"train_loss": 0.18796654045581818, "global_step": 109065, "epoch": 1225, "lr": 6.53482604841946e-05}
{"train_loss": 0.14244107902050018, "global_step": 109066, "epoch": 1225, "lr": 6.534770873581585e-05}
{"train_loss": 0.11294932663440704, "global_step": 109067, "epoch": 1225, "lr": 6.534715698537382e-05}
{"train_loss": 0.1308353692293167, "global_step": 109068, "epoch": 1225, "lr": 6.534660523286855e-05}
{"train_loss": 0.1661023199558258, "global_step": 109069, "epoch": 1225, "lr": 6.53460534783001e-05}
{"train_loss": 0.1208234652876854, "global_step": 109070, "epoch": 1225, "lr": 6.534550172166857e-05}
{"train_loss": 0.15944558382034302, "global_step": 109071, "epoch": 1225, "lr": 6.534494996297401e-05}
{"train_loss": 0.16486915946006775, "global_step": 109072, "epoch": 1225, "lr": 6.534439820221651e-05}
{"train_loss": 0.13088367879390717, "global_step": 109073, "epoch": 1225, "lr": 6.534384643939615e-05}
{"train_loss": 0.14104631543159485, "global_step": 109074, "epoch": 1225, "lr": 6.534329467451301e-05}
{"train_loss": 0.13435642421245575, "global_step": 109075, "epoch": 1225, "lr": 6.534274290756714e-05}
{"train_loss": 0.17979322373867035, "global_step": 109076, "epoch": 1225, "lr": 6.534219113855863e-05}
{"train_loss": 0.1375834196805954, "global_step": 109077, "epoch": 1225, "lr": 6.534163936748753e-05}
{"train_loss": 0.1795758754014969, "global_step": 109078, "epoch": 1225, "lr": 6.534108759435394e-05}
{"train_loss": 0.17698664963245392, "global_step": 109079, "epoch": 1225, "lr": 6.534053581915793e-05}
{"train_loss": 0.1317301094532013, "global_step": 109080, "epoch": 1225, "lr": 6.533998404189958e-05}
{"train_loss": 0.1599263697862625, "global_step": 109081, "epoch": 1225, "lr": 6.533943226257895e-05}
{"train_loss": 0.10731078684329987, "global_step": 109082, "epoch": 1225, "lr": 6.533888048119614e-05}
{"train_loss": 0.10050497204065323, "global_step": 109083, "epoch": 1225, "lr": 6.533832869775118e-05}
{"train_loss": 0.19863343238830566, "global_step": 109084, "epoch": 1225, "lr": 6.533777691224417e-05}
{"train_loss": 0.12058340758085251, "global_step": 109085, "epoch": 1225, "lr": 6.53372251246752e-05}
{"train_loss": 0.1659468710422516, "global_step": 109086, "epoch": 1225, "lr": 6.533667333504431e-05}
{"train_loss": 0.20874513685703278, "global_step": 109087, "epoch": 1225, "lr": 6.533612154335161e-05}
{"train_loss": 0.17007093131542206, "global_step": 109088, "epoch": 1225, "lr": 6.533556974959713e-05}
{"train_loss": 0.10218169540166855, "global_step": 109089, "epoch": 1225, "lr": 6.533501795378098e-05}
{"train_loss": 0.10926232486963272, "global_step": 109090, "epoch": 1225, "lr": 6.533446615590323e-05}
{"train_loss": 0.1414855569601059, "global_step": 109091, "epoch": 1225, "lr": 6.533391435596396e-05}
{"train_loss": 0.1634993702173233, "global_step": 109092, "epoch": 1225, "lr": 6.53333625539632e-05}
{"train_loss": 0.12885534763336182, "global_step": 109093, "epoch": 1225, "lr": 6.533281074990107e-05}
{"train_loss": 0.12741267681121826, "global_step": 109094, "epoch": 1225, "lr": 6.533225894377764e-05}
{"train_loss": 0.19178378582000732, "global_step": 109095, "epoch": 1225, "lr": 6.533170713559296e-05}
{"train_loss": 0.15863963961601257, "global_step": 109096, "epoch": 1225, "lr": 6.533115532534712e-05}
{"train_loss": 0.23085825145244598, "global_step": 109097, "epoch": 1225, "lr": 6.53306035130402e-05}
{"train_loss": 0.1037030816078186, "global_step": 109098, "epoch": 1225, "lr": 6.533005169867227e-05}
{"train_loss": 0.17266932129859924, "global_step": 109099, "epoch": 1225, "lr": 6.53294998822434e-05}
{"train_loss": 0.1643001139163971, "global_step": 109100, "epoch": 1225, "lr": 6.532894806375366e-05}
{"train_loss": 0.20682662725448608, "global_step": 109101, "epoch": 1225, "lr": 6.532839624320313e-05}
{"train_loss": 0.1368536800146103, "global_step": 109102, "epoch": 1225, "lr": 6.532784442059189e-05}
{"train_loss": 0.11794184148311615, "global_step": 109103, "epoch": 1225, "lr": 6.532729259592e-05}
{"train_loss": 0.12470198422670364, "global_step": 109104, "epoch": 1225, "lr": 6.532674076918756e-05}
{"train_loss": 0.1528453230857849, "global_step": 109105, "epoch": 1225, "lr": 6.532618894039461e-05}
{"train_loss": 0.23447906970977783, "global_step": 109106, "epoch": 1225, "lr": 6.532563710954125e-05}
{"train_loss": 0.13866670429706573, "global_step": 109107, "epoch": 1225, "lr": 6.532508527662754e-05}
{"train_loss": 0.1501382738351822, "global_step": 109108, "epoch": 1225, "lr": 6.532453344165357e-05}
{"train_loss": 0.1776842474937439, "global_step": 109109, "epoch": 1225, "lr": 6.532398160461939e-05}
{"train_loss": 0.21910527348518372, "global_step": 109110, "epoch": 1225, "lr": 6.532342976552509e-05}
{"train_loss": 0.18237937986850739, "global_step": 109111, "epoch": 1225, "lr": 6.532287792437073e-05}
{"train_loss": 0.17155154049396515, "global_step": 109112, "epoch": 1225, "lr": 6.53223260811564e-05}
{"train_loss": 0.15947555306922184, "global_step": 109113, "epoch": 1225, "lr": 6.53217742358822e-05, "val_loss": 4.7698655128479, "train_action_mse_error": 15.320168495178223}
{"train_loss": 0.134231835603714, "global_step": 109114, "epoch": 1226, "lr": 6.532122238854814e-05}
{"train_loss": 0.13758254051208496, "global_step": 109115, "epoch": 1226, "lr": 6.532067053915433e-05}
{"train_loss": 0.15334707498550415, "global_step": 109116, "epoch": 1226, "lr": 6.532011868770085e-05}
{"train_loss": 0.13163122534751892, "global_step": 109117, "epoch": 1226, "lr": 6.531956683418776e-05}
{"train_loss": 0.190470352768898, "global_step": 109118, "epoch": 1226, "lr": 6.531901497861515e-05}
{"train_loss": 0.22020970284938812, "global_step": 109119, "epoch": 1226, "lr": 6.531846312098309e-05}
{"train_loss": 0.07787509262561798, "global_step": 109120, "epoch": 1226, "lr": 6.531791126129166e-05}
{"train_loss": 0.13381104171276093, "global_step": 109121, "epoch": 1226, "lr": 6.531735939954089e-05}
{"train_loss": 0.12125493586063385, "global_step": 109122, "epoch": 1226, "lr": 6.531680753573091e-05}
{"train_loss": 0.25760215520858765, "global_step": 109123, "epoch": 1226, "lr": 6.531625566986177e-05}
{"train_loss": 0.14154942333698273, "global_step": 109124, "epoch": 1226, "lr": 6.531570380193354e-05}
{"train_loss": 0.22539584338665009, "global_step": 109125, "epoch": 1226, "lr": 6.53151519319463e-05}
{"train_loss": 0.17461372911930084, "global_step": 109126, "epoch": 1226, "lr": 6.531460005990014e-05}
{"train_loss": 0.13299736380577087, "global_step": 109127, "epoch": 1226, "lr": 6.531404818579512e-05}
{"train_loss": 0.17938032746315002, "global_step": 109128, "epoch": 1226, "lr": 6.53134963096313e-05}
{"train_loss": 0.1611841470003128, "global_step": 109129, "epoch": 1226, "lr": 6.531294443140876e-05}
{"train_loss": 0.11009722948074341, "global_step": 109130, "epoch": 1226, "lr": 6.531239255112759e-05}
{"train_loss": 0.12342219799757004, "global_step": 109131, "epoch": 1226, "lr": 6.531184066878785e-05}
{"train_loss": 0.1398872435092926, "global_step": 109132, "epoch": 1226, "lr": 6.531128878438962e-05}
{"train_loss": 0.20188696682453156, "global_step": 109133, "epoch": 1226, "lr": 6.531073689793297e-05}
{"train_loss": 0.13939818739891052, "global_step": 109134, "epoch": 1226, "lr": 6.531018500941799e-05}
{"train_loss": 0.10433308035135269, "global_step": 109135, "epoch": 1226, "lr": 6.530963311884473e-05}
{"train_loss": 0.1355731040239334, "global_step": 109136, "epoch": 1226, "lr": 6.530908122621329e-05}
{"train_loss": 0.16571404039859772, "global_step": 109137, "epoch": 1226, "lr": 6.530852933152371e-05}
{"train_loss": 0.09650819003582001, "global_step": 109138, "epoch": 1226, "lr": 6.530797743477612e-05}
{"train_loss": 0.18459397554397583, "global_step": 109139, "epoch": 1226, "lr": 6.530742553597053e-05}
{"train_loss": 0.13347214460372925, "global_step": 109140, "epoch": 1226, "lr": 6.530687363510705e-05}
{"train_loss": 0.1805945634841919, "global_step": 109141, "epoch": 1226, "lr": 6.530632173218575e-05}
{"train_loss": 0.16508547961711884, "global_step": 109142, "epoch": 1226, "lr": 6.53057698272067e-05}
{"train_loss": 0.15024767816066742, "global_step": 109143, "epoch": 1226, "lr": 6.530521792016997e-05}
{"train_loss": 0.17983244359493256, "global_step": 109144, "epoch": 1226, "lr": 6.530466601107566e-05}
{"train_loss": 0.16272279620170593, "global_step": 109145, "epoch": 1226, "lr": 6.53041140999238e-05}
{"train_loss": 0.1663724035024643, "global_step": 109146, "epoch": 1226, "lr": 6.530356218671452e-05}
{"train_loss": 0.1847587525844574, "global_step": 109147, "epoch": 1226, "lr": 6.530301027144783e-05}
{"train_loss": 0.12594102323055267, "global_step": 109148, "epoch": 1226, "lr": 6.530245835412385e-05}
{"train_loss": 0.08550675213336945, "global_step": 109149, "epoch": 1226, "lr": 6.530190643474264e-05}
{"train_loss": 0.17130926251411438, "global_step": 109150, "epoch": 1226, "lr": 6.530135451330426e-05}
{"train_loss": 0.2110886424779892, "global_step": 109151, "epoch": 1226, "lr": 6.530080258980882e-05}
{"train_loss": 0.16300593316555023, "global_step": 109152, "epoch": 1226, "lr": 6.530025066425636e-05}
{"train_loss": 0.07967545092105865, "global_step": 109153, "epoch": 1226, "lr": 6.529969873664699e-05}
{"train_loss": 0.14982213079929352, "global_step": 109154, "epoch": 1226, "lr": 6.529914680698074e-05}
{"train_loss": 0.19762098789215088, "global_step": 109155, "epoch": 1226, "lr": 6.52985948752577e-05}
{"train_loss": 0.17617084085941315, "global_step": 109156, "epoch": 1226, "lr": 6.529804294147798e-05}
{"train_loss": 0.10558690130710602, "global_step": 109157, "epoch": 1226, "lr": 6.52974910056416e-05}
{"train_loss": 0.13805367052555084, "global_step": 109158, "epoch": 1226, "lr": 6.529693906774867e-05}
{"train_loss": 0.17973126471042633, "global_step": 109159, "epoch": 1226, "lr": 6.529638712779925e-05}
{"train_loss": 0.1380864977836609, "global_step": 109160, "epoch": 1226, "lr": 6.529583518579341e-05}
{"train_loss": 0.152092844247818, "global_step": 109161, "epoch": 1226, "lr": 6.529528324173125e-05}
{"train_loss": 0.20303694903850555, "global_step": 109162, "epoch": 1226, "lr": 6.529473129561281e-05}
{"train_loss": 0.15941300988197327, "global_step": 109163, "epoch": 1226, "lr": 6.529417934743817e-05}
{"train_loss": 0.1322263926267624, "global_step": 109164, "epoch": 1226, "lr": 6.529362739720743e-05}
{"train_loss": 0.0973614901304245, "global_step": 109165, "epoch": 1226, "lr": 6.529307544492067e-05}
{"train_loss": 0.18901467323303223, "global_step": 109166, "epoch": 1226, "lr": 6.529252349057792e-05}
{"train_loss": 0.15088149905204773, "global_step": 109167, "epoch": 1226, "lr": 6.529197153417927e-05}
{"train_loss": 0.2088809460401535, "global_step": 109168, "epoch": 1226, "lr": 6.52914195757248e-05}
{"train_loss": 0.13606815040111542, "global_step": 109169, "epoch": 1226, "lr": 6.52908676152146e-05}
{"train_loss": 0.14038386940956116, "global_step": 109170, "epoch": 1226, "lr": 6.529031565264872e-05}
{"train_loss": 0.12987732887268066, "global_step": 109171, "epoch": 1226, "lr": 6.528976368802724e-05}
{"train_loss": 0.14943799376487732, "global_step": 109172, "epoch": 1226, "lr": 6.528921172135023e-05}
{"train_loss": 0.13067220151424408, "global_step": 109173, "epoch": 1226, "lr": 6.528865975261779e-05}
{"train_loss": 0.16922971606254578, "global_step": 109174, "epoch": 1226, "lr": 6.528810778182997e-05}
{"train_loss": 0.18320050835609436, "global_step": 109175, "epoch": 1226, "lr": 6.528755580898685e-05}
{"train_loss": 0.19331306219100952, "global_step": 109176, "epoch": 1226, "lr": 6.52870038340885e-05}
{"train_loss": 0.13006539642810822, "global_step": 109177, "epoch": 1226, "lr": 6.5286451857135e-05}
{"train_loss": 0.09498348832130432, "global_step": 109178, "epoch": 1226, "lr": 6.528589987812644e-05}
{"train_loss": 0.08193018287420273, "global_step": 109179, "epoch": 1226, "lr": 6.528534789706285e-05}
{"train_loss": 0.1373380869626999, "global_step": 109180, "epoch": 1226, "lr": 6.528479591394434e-05}
{"train_loss": 0.18388643860816956, "global_step": 109181, "epoch": 1226, "lr": 6.528424392877099e-05}
{"train_loss": 0.13167737424373627, "global_step": 109182, "epoch": 1226, "lr": 6.528369194154285e-05}
{"train_loss": 0.11254143714904785, "global_step": 109183, "epoch": 1226, "lr": 6.528313995225999e-05}
{"train_loss": 0.11255624145269394, "global_step": 109184, "epoch": 1226, "lr": 6.528258796092252e-05}
{"train_loss": 0.11714357882738113, "global_step": 109185, "epoch": 1226, "lr": 6.528203596753048e-05}
{"train_loss": 0.187488853931427, "global_step": 109186, "epoch": 1226, "lr": 6.528148397208395e-05}
{"train_loss": 0.15983085334300995, "global_step": 109187, "epoch": 1226, "lr": 6.528093197458302e-05}
{"train_loss": 0.16818593442440033, "global_step": 109188, "epoch": 1226, "lr": 6.528037997502776e-05}
{"train_loss": 0.1979319155216217, "global_step": 109189, "epoch": 1226, "lr": 6.527982797341824e-05}
{"train_loss": 0.15073047578334808, "global_step": 109190, "epoch": 1226, "lr": 6.527927596975453e-05}
{"train_loss": 0.12343861162662506, "global_step": 109191, "epoch": 1226, "lr": 6.52787239640367e-05}
{"train_loss": 0.11401398479938507, "global_step": 109192, "epoch": 1226, "lr": 6.527817195626484e-05}
{"train_loss": 0.1538105010986328, "global_step": 109193, "epoch": 1226, "lr": 6.527761994643901e-05}
{"train_loss": 0.16513946652412415, "global_step": 109194, "epoch": 1226, "lr": 6.52770679345593e-05}
{"train_loss": 0.14546935260295868, "global_step": 109195, "epoch": 1226, "lr": 6.527651592062576e-05}
{"train_loss": 0.21215331554412842, "global_step": 109196, "epoch": 1226, "lr": 6.52759639046385e-05}
{"train_loss": 0.1793753057718277, "global_step": 109197, "epoch": 1226, "lr": 6.527541188659754e-05}
{"train_loss": 0.1135333925485611, "global_step": 109198, "epoch": 1226, "lr": 6.527485986650302e-05}
{"train_loss": 0.1395229697227478, "global_step": 109199, "epoch": 1226, "lr": 6.527430784435496e-05}
{"train_loss": 0.14459660649299622, "global_step": 109200, "epoch": 1226, "lr": 6.527375582015348e-05}
{"train_loss": 0.1190301701426506, "global_step": 109201, "epoch": 1226, "lr": 6.527320379389861e-05}
{"train_loss": 0.15067658774303586, "global_step": 109202, "epoch": 1226, "lr": 6.527265176559044e-05, "val_loss": 4.821403503417969}
{"train_loss": 0.15011993050575256, "global_step": 109203, "epoch": 1227, "lr": 6.527209973522905e-05}
{"train_loss": 0.12460233271121979, "global_step": 109204, "epoch": 1227, "lr": 6.527154770281452e-05}
{"train_loss": 0.12799367308616638, "global_step": 109205, "epoch": 1227, "lr": 6.527099566834691e-05}
{"train_loss": 0.1801375448703766, "global_step": 109206, "epoch": 1227, "lr": 6.52704436318263e-05}
{"train_loss": 0.23422744870185852, "global_step": 109207, "epoch": 1227, "lr": 6.526989159325279e-05}
{"train_loss": 0.14128179848194122, "global_step": 109208, "epoch": 1227, "lr": 6.52693395526264e-05}
{"train_loss": 0.15898747742176056, "global_step": 109209, "epoch": 1227, "lr": 6.526878750994724e-05}
{"train_loss": 0.1174623966217041, "global_step": 109210, "epoch": 1227, "lr": 6.526823546521537e-05}
{"train_loss": 0.13998442888259888, "global_step": 109211, "epoch": 1227, "lr": 6.52676834184309e-05}
{"train_loss": 0.10763173550367355, "global_step": 109212, "epoch": 1227, "lr": 6.526713136959384e-05}
{"train_loss": 0.14011968672275543, "global_step": 109213, "epoch": 1227, "lr": 6.526657931870432e-05}
{"train_loss": 0.12034489214420319, "global_step": 109214, "epoch": 1227, "lr": 6.52660272657624e-05}
{"train_loss": 0.11795003712177277, "global_step": 109215, "epoch": 1227, "lr": 6.526547521076814e-05}
{"train_loss": 0.12361259013414383, "global_step": 109216, "epoch": 1227, "lr": 6.526492315372163e-05}
{"train_loss": 0.1610897034406662, "global_step": 109217, "epoch": 1227, "lr": 6.526437109462292e-05}
{"train_loss": 0.14281080663204193, "global_step": 109218, "epoch": 1227, "lr": 6.526381903347212e-05}
{"train_loss": 0.17531850934028625, "global_step": 109219, "epoch": 1227, "lr": 6.526326697026927e-05}
{"train_loss": 0.14145569503307343, "global_step": 109220, "epoch": 1227, "lr": 6.526271490501447e-05}
{"train_loss": 0.12515512108802795, "global_step": 109221, "epoch": 1227, "lr": 6.526216283770778e-05}
{"train_loss": 0.15072615444660187, "global_step": 109222, "epoch": 1227, "lr": 6.526161076834929e-05}
{"train_loss": 0.15705303847789764, "global_step": 109223, "epoch": 1227, "lr": 6.526105869693904e-05}
{"train_loss": 0.1583939790725708, "global_step": 109224, "epoch": 1227, "lr": 6.526050662347715e-05}
{"train_loss": 0.18837293982505798, "global_step": 109225, "epoch": 1227, "lr": 6.525995454796365e-05}
{"train_loss": 0.1546442061662674, "global_step": 109226, "epoch": 1227, "lr": 6.525940247039866e-05}
{"train_loss": 0.12272553145885468, "global_step": 109227, "epoch": 1227, "lr": 6.525885039078221e-05}
{"train_loss": 0.13890443742275238, "global_step": 109228, "epoch": 1227, "lr": 6.52582983091144e-05}
{"train_loss": 0.1772361397743225, "global_step": 109229, "epoch": 1227, "lr": 6.52577462253953e-05}
{"train_loss": 0.16711550951004028, "global_step": 109230, "epoch": 1227, "lr": 6.525719413962496e-05}
{"train_loss": 0.2069454938173294, "global_step": 109231, "epoch": 1227, "lr": 6.52566420518035e-05}
{"train_loss": 0.1822514533996582, "global_step": 109232, "epoch": 1227, "lr": 6.525608996193097e-05}
{"train_loss": 0.14636845886707306, "global_step": 109233, "epoch": 1227, "lr": 6.525553787000743e-05}
{"train_loss": 0.12768206000328064, "global_step": 109234, "epoch": 1227, "lr": 6.525498577603298e-05}
{"train_loss": 0.1690444052219391, "global_step": 109235, "epoch": 1227, "lr": 6.525443368000769e-05}
{"train_loss": 0.14763624966144562, "global_step": 109236, "epoch": 1227, "lr": 6.525388158193162e-05}
{"train_loss": 0.14899154007434845, "global_step": 109237, "epoch": 1227, "lr": 6.525332948180484e-05}
{"train_loss": 0.12763403356075287, "global_step": 109238, "epoch": 1227, "lr": 6.525277737962744e-05}
{"train_loss": 0.13549351692199707, "global_step": 109239, "epoch": 1227, "lr": 6.525222527539951e-05}
{"train_loss": 0.2521325349807739, "global_step": 109240, "epoch": 1227, "lr": 6.525167316912109e-05}
{"train_loss": 0.16135510802268982, "global_step": 109241, "epoch": 1227, "lr": 6.525112106079227e-05}
{"train_loss": 0.09439130872488022, "global_step": 109242, "epoch": 1227, "lr": 6.525056895041311e-05}
{"train_loss": 0.09702261537313461, "global_step": 109243, "epoch": 1227, "lr": 6.525001683798371e-05}
{"train_loss": 0.1237545982003212, "global_step": 109244, "epoch": 1227, "lr": 6.524946472350413e-05}
{"train_loss": 0.15049050748348236, "global_step": 109245, "epoch": 1227, "lr": 6.524891260697444e-05}
{"train_loss": 0.14689978957176208, "global_step": 109246, "epoch": 1227, "lr": 6.524836048839473e-05}
{"train_loss": 0.182888925075531, "global_step": 109247, "epoch": 1227, "lr": 6.524780836776505e-05}
{"train_loss": 0.16980063915252686, "global_step": 109248, "epoch": 1227, "lr": 6.52472562450855e-05}
{"train_loss": 0.13990364968776703, "global_step": 109249, "epoch": 1227, "lr": 6.524670412035614e-05}
{"train_loss": 0.17089486122131348, "global_step": 109250, "epoch": 1227, "lr": 6.524615199357704e-05}
{"train_loss": 0.12304075807332993, "global_step": 109251, "epoch": 1227, "lr": 6.524559986474829e-05}
{"train_loss": 0.1809101700782776, "global_step": 109252, "epoch": 1227, "lr": 6.524504773386995e-05}
{"train_loss": 0.15725389122962952, "global_step": 109253, "epoch": 1227, "lr": 6.52444956009421e-05}
{"train_loss": 0.24593932926654816, "global_step": 109254, "epoch": 1227, "lr": 6.52439434659648e-05}
{"train_loss": 0.10732782632112503, "global_step": 109255, "epoch": 1227, "lr": 6.524339132893817e-05}
{"train_loss": 0.21613240242004395, "global_step": 109256, "epoch": 1227, "lr": 6.524283918986222e-05}
{"train_loss": 0.13198222219944, "global_step": 109257, "epoch": 1227, "lr": 6.524228704873707e-05}
{"train_loss": 0.1948215514421463, "global_step": 109258, "epoch": 1227, "lr": 6.524173490556278e-05}
{"train_loss": 0.11624330282211304, "global_step": 109259, "epoch": 1227, "lr": 6.524118276033942e-05}
{"train_loss": 0.1734035611152649, "global_step": 109260, "epoch": 1227, "lr": 6.524063061306707e-05}
{"train_loss": 0.21194663643836975, "global_step": 109261, "epoch": 1227, "lr": 6.52400784637458e-05}
{"train_loss": 0.11746355146169662, "global_step": 109262, "epoch": 1227, "lr": 6.52395263123757e-05}
{"train_loss": 0.16714119911193848, "global_step": 109263, "epoch": 1227, "lr": 6.523897415895682e-05}
{"train_loss": 0.13735979795455933, "global_step": 109264, "epoch": 1227, "lr": 6.523842200348926e-05}
{"train_loss": 0.13358905911445618, "global_step": 109265, "epoch": 1227, "lr": 6.523786984597307e-05}
{"train_loss": 0.19430766999721527, "global_step": 109266, "epoch": 1227, "lr": 6.523731768640833e-05}
{"train_loss": 0.11954857409000397, "global_step": 109267, "epoch": 1227, "lr": 6.523676552479512e-05}
{"train_loss": 0.24555456638336182, "global_step": 109268, "epoch": 1227, "lr": 6.52362133611335e-05}
{"train_loss": 0.1629529744386673, "global_step": 109269, "epoch": 1227, "lr": 6.523566119542358e-05}
{"train_loss": 0.18309181928634644, "global_step": 109270, "epoch": 1227, "lr": 6.52351090276654e-05}
{"train_loss": 0.13651663064956665, "global_step": 109271, "epoch": 1227, "lr": 6.523455685785904e-05}
{"train_loss": 0.12174566835165024, "global_step": 109272, "epoch": 1227, "lr": 6.523400468600459e-05}
{"train_loss": 0.10899829864501953, "global_step": 109273, "epoch": 1227, "lr": 6.52334525121021e-05}
{"train_loss": 0.17172035574913025, "global_step": 109274, "epoch": 1227, "lr": 6.523290033615168e-05}
{"train_loss": 0.18533402681350708, "global_step": 109275, "epoch": 1227, "lr": 6.523234815815338e-05}
{"train_loss": 0.16466203331947327, "global_step": 109276, "epoch": 1227, "lr": 6.523179597810725e-05}
{"train_loss": 0.14396823942661285, "global_step": 109277, "epoch": 1227, "lr": 6.523124379601342e-05}
{"train_loss": 0.1474543809890747, "global_step": 109278, "epoch": 1227, "lr": 6.523069161187192e-05}
{"train_loss": 0.2006387859582901, "global_step": 109279, "epoch": 1227, "lr": 6.523013942568283e-05}
{"train_loss": 0.1448928862810135, "global_step": 109280, "epoch": 1227, "lr": 6.522958723744626e-05}
{"train_loss": 0.12293682992458344, "global_step": 109281, "epoch": 1227, "lr": 6.522903504716223e-05}
{"train_loss": 0.09488459676504135, "global_step": 109282, "epoch": 1227, "lr": 6.522848285483086e-05}
{"train_loss": 0.1485203504562378, "global_step": 109283, "epoch": 1227, "lr": 6.52279306604522e-05}
{"train_loss": 0.13525982201099396, "global_step": 109284, "epoch": 1227, "lr": 6.522737846402633e-05}
{"train_loss": 0.09842658042907715, "global_step": 109285, "epoch": 1227, "lr": 6.522682626555333e-05}
{"train_loss": 0.12706676125526428, "global_step": 109286, "epoch": 1227, "lr": 6.522627406503327e-05}
{"train_loss": 0.17325963079929352, "global_step": 109287, "epoch": 1227, "lr": 6.52257218624662e-05}
{"train_loss": 0.16967111825942993, "global_step": 109288, "epoch": 1227, "lr": 6.522516965785225e-05}
{"train_loss": 0.20495066046714783, "global_step": 109289, "epoch": 1227, "lr": 6.522461745119144e-05}
{"train_loss": 0.15613192319869995, "global_step": 109290, "epoch": 1227, "lr": 6.522406524248387e-05}
{"train_loss": 0.15321755442726478, "global_step": 109291, "epoch": 1227, "lr": 6.522351303172962e-05, "val_loss": 4.881978511810303}
{"train_loss": 0.1465608775615692, "global_step": 109292, "epoch": 1228, "lr": 6.522296081892874e-05}
{"train_loss": 0.15725496411323547, "global_step": 109293, "epoch": 1228, "lr": 6.522240860408132e-05}
{"train_loss": 0.1706220656633377, "global_step": 109294, "epoch": 1228, "lr": 6.522185638718745e-05}
{"train_loss": 0.17074745893478394, "global_step": 109295, "epoch": 1228, "lr": 6.522130416824717e-05}
{"train_loss": 0.21192066371440887, "global_step": 109296, "epoch": 1228, "lr": 6.522075194726058e-05}
{"train_loss": 0.18554531037807465, "global_step": 109297, "epoch": 1228, "lr": 6.522019972422773e-05}
{"train_loss": 0.20831628143787384, "global_step": 109298, "epoch": 1228, "lr": 6.521964749914871e-05}
{"train_loss": 0.15250316262245178, "global_step": 109299, "epoch": 1228, "lr": 6.521909527202361e-05}
{"train_loss": 0.10179397463798523, "global_step": 109300, "epoch": 1228, "lr": 6.521854304285249e-05}
{"train_loss": 0.1680409163236618, "global_step": 109301, "epoch": 1228, "lr": 6.521799081163541e-05}
{"train_loss": 0.20378735661506653, "global_step": 109302, "epoch": 1228, "lr": 6.521743857837245e-05}
{"train_loss": 0.1111261174082756, "global_step": 109303, "epoch": 1228, "lr": 6.521688634306372e-05}
{"train_loss": 0.08160713315010071, "global_step": 109304, "epoch": 1228, "lr": 6.521633410570924e-05}
{"train_loss": 0.12366044521331787, "global_step": 109305, "epoch": 1228, "lr": 6.521578186630913e-05}
{"train_loss": 0.18879233300685883, "global_step": 109306, "epoch": 1228, "lr": 6.521522962486341e-05}
{"train_loss": 0.1430409997701645, "global_step": 109307, "epoch": 1228, "lr": 6.521467738137222e-05}
{"train_loss": 0.17563392221927643, "global_step": 109308, "epoch": 1228, "lr": 6.521412513583557e-05}
{"train_loss": 0.14725321531295776, "global_step": 109309, "epoch": 1228, "lr": 6.52135728882536e-05}
{"train_loss": 0.11513528972864151, "global_step": 109310, "epoch": 1228, "lr": 6.521302063862635e-05}
{"train_loss": 0.13612352311611176, "global_step": 109311, "epoch": 1228, "lr": 6.521246838695387e-05}
{"train_loss": 0.22378595173358917, "global_step": 109312, "epoch": 1228, "lr": 6.521191613323627e-05}
{"train_loss": 0.1889893114566803, "global_step": 109313, "epoch": 1228, "lr": 6.521136387747361e-05}
{"train_loss": 0.0974685475230217, "global_step": 109314, "epoch": 1228, "lr": 6.521081161966597e-05}
{"train_loss": 0.13200005888938904, "global_step": 109315, "epoch": 1228, "lr": 6.521025935981343e-05}
{"train_loss": 0.13133174180984497, "global_step": 109316, "epoch": 1228, "lr": 6.520970709791605e-05}
{"train_loss": 0.1245124414563179, "global_step": 109317, "epoch": 1228, "lr": 6.520915483397391e-05}
{"train_loss": 0.2654671370983124, "global_step": 109318, "epoch": 1228, "lr": 6.520860256798709e-05}
{"train_loss": 0.07855603098869324, "global_step": 109319, "epoch": 1228, "lr": 6.520805029995565e-05}
{"train_loss": 0.1922614872455597, "global_step": 109320, "epoch": 1228, "lr": 6.520749802987968e-05}
{"train_loss": 0.19409267604351044, "global_step": 109321, "epoch": 1228, "lr": 6.520694575775923e-05}
{"train_loss": 0.15450167655944824, "global_step": 109322, "epoch": 1228, "lr": 6.520639348359441e-05}
{"train_loss": 0.14288479089736938, "global_step": 109323, "epoch": 1228, "lr": 6.520584120738527e-05}
{"train_loss": 0.16416259109973907, "global_step": 109324, "epoch": 1228, "lr": 6.520528892913189e-05}
{"train_loss": 0.23996984958648682, "global_step": 109325, "epoch": 1228, "lr": 6.520473664883435e-05}
{"train_loss": 0.15419340133666992, "global_step": 109326, "epoch": 1228, "lr": 6.52041843664927e-05}
{"train_loss": 0.22442790865898132, "global_step": 109327, "epoch": 1228, "lr": 6.520363208210705e-05}
{"train_loss": 0.16226916015148163, "global_step": 109328, "epoch": 1228, "lr": 6.520307979567746e-05}
{"train_loss": 0.09054607897996902, "global_step": 109329, "epoch": 1228, "lr": 6.520252750720399e-05}
{"train_loss": 0.14996883273124695, "global_step": 109330, "epoch": 1228, "lr": 6.520197521668673e-05}
{"train_loss": 0.16452786326408386, "global_step": 109331, "epoch": 1228, "lr": 6.520142292412575e-05}
{"train_loss": 0.12445829808712006, "global_step": 109332, "epoch": 1228, "lr": 6.520087062952111e-05}
{"train_loss": 0.14217133820056915, "global_step": 109333, "epoch": 1228, "lr": 6.52003183328729e-05}
{"train_loss": 0.1613268107175827, "global_step": 109334, "epoch": 1228, "lr": 6.519976603418123e-05}
{"train_loss": 0.13030529022216797, "global_step": 109335, "epoch": 1228, "lr": 6.519921373344611e-05}
{"train_loss": 0.11983810365200043, "global_step": 109336, "epoch": 1228, "lr": 6.519866143066764e-05}
{"train_loss": 0.16503946483135223, "global_step": 109337, "epoch": 1228, "lr": 6.51981091258459e-05}
{"train_loss": 0.15374428033828735, "global_step": 109338, "epoch": 1228, "lr": 6.519755681898096e-05}
{"train_loss": 0.11101412028074265, "global_step": 109339, "epoch": 1228, "lr": 6.519700451007288e-05}
{"train_loss": 0.14172807335853577, "global_step": 109340, "epoch": 1228, "lr": 6.519645219912177e-05}
{"train_loss": 0.23426324129104614, "global_step": 109341, "epoch": 1228, "lr": 6.519589988612767e-05}
{"train_loss": 0.09891968220472336, "global_step": 109342, "epoch": 1228, "lr": 6.519534757109066e-05}
{"train_loss": 0.1257242113351822, "global_step": 109343, "epoch": 1228, "lr": 6.519479525401082e-05}
{"train_loss": 0.17461925745010376, "global_step": 109344, "epoch": 1228, "lr": 6.519424293488824e-05}
{"train_loss": 0.1760859340429306, "global_step": 109345, "epoch": 1228, "lr": 6.519369061372297e-05}
{"train_loss": 0.2138560712337494, "global_step": 109346, "epoch": 1228, "lr": 6.51931382905151e-05}
{"train_loss": 0.2432301640510559, "global_step": 109347, "epoch": 1228, "lr": 6.519258596526468e-05}
{"train_loss": 0.19345778226852417, "global_step": 109348, "epoch": 1228, "lr": 6.519203363797181e-05}
{"train_loss": 0.1360575556755066, "global_step": 109349, "epoch": 1228, "lr": 6.519148130863656e-05}
{"train_loss": 0.1636207550764084, "global_step": 109350, "epoch": 1228, "lr": 6.5190928977259e-05}
{"train_loss": 0.1408258080482483, "global_step": 109351, "epoch": 1228, "lr": 6.51903766438392e-05}
{"train_loss": 0.09213660657405853, "global_step": 109352, "epoch": 1228, "lr": 6.518982430837726e-05}
{"train_loss": 0.1410609483718872, "global_step": 109353, "epoch": 1228, "lr": 6.518927197087322e-05}
{"train_loss": 0.13196928799152374, "global_step": 109354, "epoch": 1228, "lr": 6.518871963132717e-05}
{"train_loss": 0.10035046190023422, "global_step": 109355, "epoch": 1228, "lr": 6.518816728973918e-05}
{"train_loss": 0.16661526262760162, "global_step": 109356, "epoch": 1228, "lr": 6.518761494610932e-05}
{"train_loss": 0.15946519374847412, "global_step": 109357, "epoch": 1228, "lr": 6.518706260043768e-05}
{"train_loss": 0.14633890986442566, "global_step": 109358, "epoch": 1228, "lr": 6.51865102527243e-05}
{"train_loss": 0.09419151395559311, "global_step": 109359, "epoch": 1228, "lr": 6.518595790296931e-05}
{"train_loss": 0.13042646646499634, "global_step": 109360, "epoch": 1228, "lr": 6.518540555117274e-05}
{"train_loss": 0.15366220474243164, "global_step": 109361, "epoch": 1228, "lr": 6.518485319733467e-05}
{"train_loss": 0.19860346615314484, "global_step": 109362, "epoch": 1228, "lr": 6.518430084145519e-05}
{"train_loss": 0.14493665099143982, "global_step": 109363, "epoch": 1228, "lr": 6.518374848353436e-05}
{"train_loss": 0.13721713423728943, "global_step": 109364, "epoch": 1228, "lr": 6.518319612357227e-05}
{"train_loss": 0.14123907685279846, "global_step": 109365, "epoch": 1228, "lr": 6.518264376156898e-05}
{"train_loss": 0.1353539377450943, "global_step": 109366, "epoch": 1228, "lr": 6.518209139752457e-05}
{"train_loss": 0.15832531452178955, "global_step": 109367, "epoch": 1228, "lr": 6.518153903143911e-05}
{"train_loss": 0.12377142906188965, "global_step": 109368, "epoch": 1228, "lr": 6.518098666331267e-05}
{"train_loss": 0.14973773062229156, "global_step": 109369, "epoch": 1228, "lr": 6.518043429314535e-05}
{"train_loss": 0.157973974943161, "global_step": 109370, "epoch": 1228, "lr": 6.517988192093718e-05}
{"train_loss": 0.2083887904882431, "global_step": 109371, "epoch": 1228, "lr": 6.517932954668829e-05}
{"train_loss": 0.16899117827415466, "global_step": 109372, "epoch": 1228, "lr": 6.517877717039869e-05}
{"train_loss": 0.17564859986305237, "global_step": 109373, "epoch": 1228, "lr": 6.517822479206854e-05}
{"train_loss": 0.0887976884841919, "global_step": 109374, "epoch": 1228, "lr": 6.517767241169782e-05}
{"train_loss": 0.15178528428077698, "global_step": 109375, "epoch": 1228, "lr": 6.517712002928666e-05}
{"train_loss": 0.15379321575164795, "global_step": 109376, "epoch": 1228, "lr": 6.517656764483512e-05}
{"train_loss": 0.20952388644218445, "global_step": 109377, "epoch": 1228, "lr": 6.517601525834327e-05}
{"train_loss": 0.17525842785835266, "global_step": 109378, "epoch": 1228, "lr": 6.517546286981119e-05}
{"train_loss": 0.14018775522708893, "global_step": 109379, "epoch": 1228, "lr": 6.517491047923896e-05}
{"train_loss": 0.15490458210867442, "global_step": 109380, "epoch": 1228, "lr": 6.517435808662664e-05, "val_loss": 4.8471479415893555}
{"train_loss": 0.17213302850723267, "global_step": 109381, "epoch": 1229, "lr": 6.517380569197433e-05}
{"train_loss": 0.072980597615242, "global_step": 109382, "epoch": 1229, "lr": 6.517325329528207e-05}
{"train_loss": 0.1099260076880455, "global_step": 109383, "epoch": 1229, "lr": 6.517270089654995e-05}
{"train_loss": 0.13745197653770447, "global_step": 109384, "epoch": 1229, "lr": 6.517214849577806e-05}
{"train_loss": 0.16616351902484894, "global_step": 109385, "epoch": 1229, "lr": 6.517159609296645e-05}
{"train_loss": 0.1541510820388794, "global_step": 109386, "epoch": 1229, "lr": 6.51710436881152e-05}
{"train_loss": 0.16397985816001892, "global_step": 109387, "epoch": 1229, "lr": 6.517049128122439e-05}
{"train_loss": 0.16753993928432465, "global_step": 109388, "epoch": 1229, "lr": 6.516993887229409e-05}
{"train_loss": 0.18829511106014252, "global_step": 109389, "epoch": 1229, "lr": 6.516938646132438e-05}
{"train_loss": 0.16916292905807495, "global_step": 109390, "epoch": 1229, "lr": 6.516883404831533e-05}
{"train_loss": 0.18472154438495636, "global_step": 109391, "epoch": 1229, "lr": 6.516828163326702e-05}
{"train_loss": 0.2002658247947693, "global_step": 109392, "epoch": 1229, "lr": 6.516772921617953e-05}
{"train_loss": 0.12682871520519257, "global_step": 109393, "epoch": 1229, "lr": 6.51671767970529e-05}
{"train_loss": 0.207724928855896, "global_step": 109394, "epoch": 1229, "lr": 6.516662437588725e-05}
{"train_loss": 0.1515982449054718, "global_step": 109395, "epoch": 1229, "lr": 6.516607195268261e-05}
{"train_loss": 0.24645747244358063, "global_step": 109396, "epoch": 1229, "lr": 6.51655195274391e-05}
{"train_loss": 0.1970498263835907, "global_step": 109397, "epoch": 1229, "lr": 6.516496710015675e-05}
{"train_loss": 0.1217082291841507, "global_step": 109398, "epoch": 1229, "lr": 6.516441467083568e-05}
{"train_loss": 0.10563714802265167, "global_step": 109399, "epoch": 1229, "lr": 6.51638622394759e-05}
{"train_loss": 0.11402885615825653, "global_step": 109400, "epoch": 1229, "lr": 6.516330980607756e-05}
{"train_loss": 0.12379608303308487, "global_step": 109401, "epoch": 1229, "lr": 6.516275737064069e-05}
{"train_loss": 0.09422285854816437, "global_step": 109402, "epoch": 1229, "lr": 6.516220493316536e-05}
{"train_loss": 0.1175459697842598, "global_step": 109403, "epoch": 1229, "lr": 6.516165249365166e-05}
{"train_loss": 0.09405364841222763, "global_step": 109404, "epoch": 1229, "lr": 6.516110005209966e-05}
{"train_loss": 0.096352219581604, "global_step": 109405, "epoch": 1229, "lr": 6.516054760850942e-05}
{"train_loss": 0.10331162065267563, "global_step": 109406, "epoch": 1229, "lr": 6.515999516288105e-05}
{"train_loss": 0.1938924342393875, "global_step": 109407, "epoch": 1229, "lr": 6.51594427152146e-05}
{"train_loss": 0.1705630123615265, "global_step": 109408, "epoch": 1229, "lr": 6.515889026551016e-05}
{"train_loss": 0.14122405648231506, "global_step": 109409, "epoch": 1229, "lr": 6.515833781376776e-05}
{"train_loss": 0.20117080211639404, "global_step": 109410, "epoch": 1229, "lr": 6.515778535998753e-05}
{"train_loss": 0.16058297455310822, "global_step": 109411, "epoch": 1229, "lr": 6.515723290416954e-05}
{"train_loss": 0.12621945142745972, "global_step": 109412, "epoch": 1229, "lr": 6.515668044631382e-05}
{"train_loss": 0.21494199335575104, "global_step": 109413, "epoch": 1229, "lr": 6.515612798642046e-05}
{"train_loss": 0.17583918571472168, "global_step": 109414, "epoch": 1229, "lr": 6.515557552448957e-05}
{"train_loss": 0.14290271699428558, "global_step": 109415, "epoch": 1229, "lr": 6.515502306052117e-05}
{"train_loss": 0.1980627030134201, "global_step": 109416, "epoch": 1229, "lr": 6.515447059451538e-05}
{"train_loss": 0.09699618816375732, "global_step": 109417, "epoch": 1229, "lr": 6.515391812647225e-05}
{"train_loss": 0.14515410363674164, "global_step": 109418, "epoch": 1229, "lr": 6.515336565639185e-05}
{"train_loss": 0.13085228204727173, "global_step": 109419, "epoch": 1229, "lr": 6.515281318427429e-05}
{"train_loss": 0.14183276891708374, "global_step": 109420, "epoch": 1229, "lr": 6.515226071011959e-05}
{"train_loss": 0.15583965182304382, "global_step": 109421, "epoch": 1229, "lr": 6.515170823392787e-05}
{"train_loss": 0.20602969825267792, "global_step": 109422, "epoch": 1229, "lr": 6.515115575569918e-05}
{"train_loss": 0.11376068741083145, "global_step": 109423, "epoch": 1229, "lr": 6.515060327543361e-05}
{"train_loss": 0.13223367929458618, "global_step": 109424, "epoch": 1229, "lr": 6.515005079313121e-05}
{"train_loss": 0.1932099163532257, "global_step": 109425, "epoch": 1229, "lr": 6.514949830879209e-05}
{"train_loss": 0.10415032505989075, "global_step": 109426, "epoch": 1229, "lr": 6.51489458224163e-05}
{"train_loss": 0.14617447555065155, "global_step": 109427, "epoch": 1229, "lr": 6.514839333400391e-05}
{"train_loss": 0.15631471574306488, "global_step": 109428, "epoch": 1229, "lr": 6.514784084355501e-05}
{"train_loss": 0.12921687960624695, "global_step": 109429, "epoch": 1229, "lr": 6.514728835106967e-05}
{"train_loss": 0.11288182437419891, "global_step": 109430, "epoch": 1229, "lr": 6.514673585654795e-05}
{"train_loss": 0.13452354073524475, "global_step": 109431, "epoch": 1229, "lr": 6.514618335998995e-05}
{"train_loss": 0.175808846950531, "global_step": 109432, "epoch": 1229, "lr": 6.514563086139573e-05}
{"train_loss": 0.16760706901550293, "global_step": 109433, "epoch": 1229, "lr": 6.514507836076536e-05}
{"train_loss": 0.1503620445728302, "global_step": 109434, "epoch": 1229, "lr": 6.514452585809891e-05}
{"train_loss": 0.14168387651443481, "global_step": 109435, "epoch": 1229, "lr": 6.514397335339648e-05}
{"train_loss": 0.11225404590368271, "global_step": 109436, "epoch": 1229, "lr": 6.514342084665812e-05}
{"train_loss": 0.09452193975448608, "global_step": 109437, "epoch": 1229, "lr": 6.514286833788392e-05}
{"train_loss": 0.1436966359615326, "global_step": 109438, "epoch": 1229, "lr": 6.514231582707393e-05}
{"train_loss": 0.18938207626342773, "global_step": 109439, "epoch": 1229, "lr": 6.514176331422824e-05}
{"train_loss": 0.1130804717540741, "global_step": 109440, "epoch": 1229, "lr": 6.514121079934692e-05}
{"train_loss": 0.11025926470756531, "global_step": 109441, "epoch": 1229, "lr": 6.514065828243006e-05}
{"train_loss": 0.1301250457763672, "global_step": 109442, "epoch": 1229, "lr": 6.514010576347772e-05}
{"train_loss": 0.15250222384929657, "global_step": 109443, "epoch": 1229, "lr": 6.513955324248997e-05}
{"train_loss": 0.17493057250976562, "global_step": 109444, "epoch": 1229, "lr": 6.513900071946691e-05}
{"train_loss": 0.12213743478059769, "global_step": 109445, "epoch": 1229, "lr": 6.513844819440859e-05}
{"train_loss": 0.1170383170247078, "global_step": 109446, "epoch": 1229, "lr": 6.513789566731509e-05}
{"train_loss": 0.10878271609544754, "global_step": 109447, "epoch": 1229, "lr": 6.513734313818648e-05}
{"train_loss": 0.10915857553482056, "global_step": 109448, "epoch": 1229, "lr": 6.513679060702284e-05}
{"train_loss": 0.12219323962926865, "global_step": 109449, "epoch": 1229, "lr": 6.513623807382424e-05}
{"train_loss": 0.14643597602844238, "global_step": 109450, "epoch": 1229, "lr": 6.513568553859077e-05}
{"train_loss": 0.10139457881450653, "global_step": 109451, "epoch": 1229, "lr": 6.513513300132249e-05}
{"train_loss": 0.22169223427772522, "global_step": 109452, "epoch": 1229, "lr": 6.513458046201945e-05}
{"train_loss": 0.13952909409999847, "global_step": 109453, "epoch": 1229, "lr": 6.513402792068177e-05}
{"train_loss": 0.10746172070503235, "global_step": 109454, "epoch": 1229, "lr": 6.513347537730951e-05}
{"train_loss": 0.07515732198953629, "global_step": 109455, "epoch": 1229, "lr": 6.513292283190273e-05}
{"train_loss": 0.17305564880371094, "global_step": 109456, "epoch": 1229, "lr": 6.513237028446153e-05}
{"train_loss": 0.17048798501491547, "global_step": 109457, "epoch": 1229, "lr": 6.513181773498594e-05}
{"train_loss": 0.26014837622642517, "global_step": 109458, "epoch": 1229, "lr": 6.513126518347606e-05}
{"train_loss": 0.1568172425031662, "global_step": 109459, "epoch": 1229, "lr": 6.513071262993197e-05}
{"train_loss": 0.13850678503513336, "global_step": 109460, "epoch": 1229, "lr": 6.513016007435375e-05}
{"train_loss": 0.15118689835071564, "global_step": 109461, "epoch": 1229, "lr": 6.512960751674148e-05}
{"train_loss": 0.12285176664590836, "global_step": 109462, "epoch": 1229, "lr": 6.51290549570952e-05}
{"train_loss": 0.18311655521392822, "global_step": 109463, "epoch": 1229, "lr": 6.512850239541499e-05}
{"train_loss": 0.16802214086055756, "global_step": 109464, "epoch": 1229, "lr": 6.512794983170096e-05}
{"train_loss": 0.181889146566391, "global_step": 109465, "epoch": 1229, "lr": 6.512739726595315e-05}
{"train_loss": 0.18701471388339996, "global_step": 109466, "epoch": 1229, "lr": 6.512684469817165e-05}
{"train_loss": 0.12490097433328629, "global_step": 109467, "epoch": 1229, "lr": 6.512629212835654e-05}
{"train_loss": 0.2197325974702835, "global_step": 109468, "epoch": 1229, "lr": 6.512573955650787e-05}
{"train_loss": 0.14866440801808004, "global_step": 109469, "epoch": 1229, "lr": 6.512518698262574e-05, "val_loss": 5.148468017578125}
{"train_loss": 0.11162558943033218, "global_step": 109470, "epoch": 1230, "lr": 6.512463440671021e-05}
{"train_loss": 0.1402900516986847, "global_step": 109471, "epoch": 1230, "lr": 6.512408182876135e-05}
{"train_loss": 0.10932508856058121, "global_step": 109472, "epoch": 1230, "lr": 6.512352924877925e-05}
{"train_loss": 0.1697196662425995, "global_step": 109473, "epoch": 1230, "lr": 6.512297666676397e-05}
{"train_loss": 0.23645268380641937, "global_step": 109474, "epoch": 1230, "lr": 6.512242408271558e-05}
{"train_loss": 0.13999700546264648, "global_step": 109475, "epoch": 1230, "lr": 6.512187149663418e-05}
{"train_loss": 0.15745176374912262, "global_step": 109476, "epoch": 1230, "lr": 6.51213189085198e-05}
{"train_loss": 0.1922110915184021, "global_step": 109477, "epoch": 1230, "lr": 6.512076631837258e-05}
{"train_loss": 0.18565990030765533, "global_step": 109478, "epoch": 1230, "lr": 6.512021372619255e-05}
{"train_loss": 0.20400069653987885, "global_step": 109479, "epoch": 1230, "lr": 6.511966113197978e-05}
{"train_loss": 0.17778760194778442, "global_step": 109480, "epoch": 1230, "lr": 6.511910853573435e-05}
{"train_loss": 0.10530462861061096, "global_step": 109481, "epoch": 1230, "lr": 6.511855593745636e-05}
{"train_loss": 0.12372046709060669, "global_step": 109482, "epoch": 1230, "lr": 6.511800333714585e-05}
{"train_loss": 0.1439155489206314, "global_step": 109483, "epoch": 1230, "lr": 6.511745073480292e-05}
{"train_loss": 0.17938852310180664, "global_step": 109484, "epoch": 1230, "lr": 6.511689813042763e-05}
{"train_loss": 0.19031716883182526, "global_step": 109485, "epoch": 1230, "lr": 6.511634552402004e-05}
{"train_loss": 0.12228261679410934, "global_step": 109486, "epoch": 1230, "lr": 6.511579291558027e-05}
{"train_loss": 0.18415674567222595, "global_step": 109487, "epoch": 1230, "lr": 6.511524030510835e-05}
{"train_loss": 0.1549016237258911, "global_step": 109488, "epoch": 1230, "lr": 6.511468769260437e-05}
{"train_loss": 0.1614648848772049, "global_step": 109489, "epoch": 1230, "lr": 6.511413507806841e-05}
{"train_loss": 0.17382365465164185, "global_step": 109490, "epoch": 1230, "lr": 6.511358246150054e-05}
{"train_loss": 0.12286339700222015, "global_step": 109491, "epoch": 1230, "lr": 6.511302984290083e-05}
{"train_loss": 0.13778212666511536, "global_step": 109492, "epoch": 1230, "lr": 6.511247722226936e-05}
{"train_loss": 0.22635391354560852, "global_step": 109493, "epoch": 1230, "lr": 6.51119245996062e-05}
{"train_loss": 0.13303714990615845, "global_step": 109494, "epoch": 1230, "lr": 6.511137197491141e-05}
{"train_loss": 0.11277401447296143, "global_step": 109495, "epoch": 1230, "lr": 6.511081934818509e-05}
{"train_loss": 0.17087523639202118, "global_step": 109496, "epoch": 1230, "lr": 6.51102667194273e-05}
{"train_loss": 0.08125926554203033, "global_step": 109497, "epoch": 1230, "lr": 6.510971408863813e-05}
{"train_loss": 0.2087278664112091, "global_step": 109498, "epoch": 1230, "lr": 6.510916145581764e-05}
{"train_loss": 0.15913712978363037, "global_step": 109499, "epoch": 1230, "lr": 6.51086088209659e-05}
{"train_loss": 0.1756914108991623, "global_step": 109500, "epoch": 1230, "lr": 6.510805618408301e-05}
{"train_loss": 0.17302002012729645, "global_step": 109501, "epoch": 1230, "lr": 6.510750354516901e-05}
{"train_loss": 0.1250002235174179, "global_step": 109502, "epoch": 1230, "lr": 6.510695090422401e-05}
{"train_loss": 0.13950493931770325, "global_step": 109503, "epoch": 1230, "lr": 6.510639826124805e-05}
{"train_loss": 0.19839678704738617, "global_step": 109504, "epoch": 1230, "lr": 6.510584561624121e-05}
{"train_loss": 0.09247054159641266, "global_step": 109505, "epoch": 1230, "lr": 6.510529296920358e-05}
{"train_loss": 0.1440744549036026, "global_step": 109506, "epoch": 1230, "lr": 6.510474032013522e-05}
{"train_loss": 0.11837988346815109, "global_step": 109507, "epoch": 1230, "lr": 6.510418766903623e-05}
{"train_loss": 0.12514349818229675, "global_step": 109508, "epoch": 1230, "lr": 6.510363501590667e-05}
{"train_loss": 0.14115625619888306, "global_step": 109509, "epoch": 1230, "lr": 6.51030823607466e-05}
{"train_loss": 0.20953258872032166, "global_step": 109510, "epoch": 1230, "lr": 6.51025297035561e-05}
{"train_loss": 0.16916485130786896, "global_step": 109511, "epoch": 1230, "lr": 6.510197704433525e-05}
{"train_loss": 0.10718899220228195, "global_step": 109512, "epoch": 1230, "lr": 6.510142438308413e-05}
{"train_loss": 0.2575514316558838, "global_step": 109513, "epoch": 1230, "lr": 6.51008717198028e-05}
{"train_loss": 0.17101579904556274, "global_step": 109514, "epoch": 1230, "lr": 6.510031905449133e-05}
{"train_loss": 0.15461423993110657, "global_step": 109515, "epoch": 1230, "lr": 6.509976638714984e-05}
{"train_loss": 0.20998264849185944, "global_step": 109516, "epoch": 1230, "lr": 6.509921371777835e-05}
{"train_loss": 0.12001491338014603, "global_step": 109517, "epoch": 1230, "lr": 6.509866104637695e-05}
{"train_loss": 0.19684000313282013, "global_step": 109518, "epoch": 1230, "lr": 6.509810837294571e-05}
{"train_loss": 0.16758881509304047, "global_step": 109519, "epoch": 1230, "lr": 6.509755569748472e-05}
{"train_loss": 0.10284075140953064, "global_step": 109520, "epoch": 1230, "lr": 6.509700301999406e-05}
{"train_loss": 0.17765097320079803, "global_step": 109521, "epoch": 1230, "lr": 6.509645034047379e-05}
{"train_loss": 0.163445383310318, "global_step": 109522, "epoch": 1230, "lr": 6.509589765892399e-05}
{"train_loss": 0.20722118020057678, "global_step": 109523, "epoch": 1230, "lr": 6.509534497534472e-05}
{"train_loss": 0.23933039605617523, "global_step": 109524, "epoch": 1230, "lr": 6.509479228973607e-05}
{"train_loss": 0.1701134294271469, "global_step": 109525, "epoch": 1230, "lr": 6.50942396020981e-05}
{"train_loss": 0.13666190207004547, "global_step": 109526, "epoch": 1230, "lr": 6.50936869124309e-05}
{"train_loss": 0.18858438730239868, "global_step": 109527, "epoch": 1230, "lr": 6.509313422073455e-05}
{"train_loss": 0.1390398144721985, "global_step": 109528, "epoch": 1230, "lr": 6.509258152700912e-05}
{"train_loss": 0.11431760340929031, "global_step": 109529, "epoch": 1230, "lr": 6.509202883125465e-05}
{"train_loss": 0.1376350373029709, "global_step": 109530, "epoch": 1230, "lr": 6.509147613347124e-05}
{"train_loss": 0.170507550239563, "global_step": 109531, "epoch": 1230, "lr": 6.509092343365897e-05}
{"train_loss": 0.1704990714788437, "global_step": 109532, "epoch": 1230, "lr": 6.509037073181791e-05}
{"train_loss": 0.16310057044029236, "global_step": 109533, "epoch": 1230, "lr": 6.508981802794814e-05}
{"train_loss": 0.1896742582321167, "global_step": 109534, "epoch": 1230, "lr": 6.508926532204972e-05}
{"train_loss": 0.16681285202503204, "global_step": 109535, "epoch": 1230, "lr": 6.508871261412274e-05}
{"train_loss": 0.15966859459877014, "global_step": 109536, "epoch": 1230, "lr": 6.508815990416726e-05}
{"train_loss": 0.2245890200138092, "global_step": 109537, "epoch": 1230, "lr": 6.508760719218337e-05}
{"train_loss": 0.11809588223695755, "global_step": 109538, "epoch": 1230, "lr": 6.508705447817113e-05}
{"train_loss": 0.2313324213027954, "global_step": 109539, "epoch": 1230, "lr": 6.508650176213062e-05}
{"train_loss": 0.1374252438545227, "global_step": 109540, "epoch": 1230, "lr": 6.508594904406192e-05}
{"train_loss": 0.1688733994960785, "global_step": 109541, "epoch": 1230, "lr": 6.508539632396508e-05}
{"train_loss": 0.13047443330287933, "global_step": 109542, "epoch": 1230, "lr": 6.50848436018402e-05}
{"train_loss": 0.16511130332946777, "global_step": 109543, "epoch": 1230, "lr": 6.508429087768735e-05}
{"train_loss": 0.15617212653160095, "global_step": 109544, "epoch": 1230, "lr": 6.50837381515066e-05}
{"train_loss": 0.1338200569152832, "global_step": 109545, "epoch": 1230, "lr": 6.508318542329803e-05}
{"train_loss": 0.13748076558113098, "global_step": 109546, "epoch": 1230, "lr": 6.508263269306169e-05}
{"train_loss": 0.20585566759109497, "global_step": 109547, "epoch": 1230, "lr": 6.50820799607977e-05}
{"train_loss": 0.1382794827222824, "global_step": 109548, "epoch": 1230, "lr": 6.50815272265061e-05}
{"train_loss": 0.15183314681053162, "global_step": 109549, "epoch": 1230, "lr": 6.508097449018696e-05}
{"train_loss": 0.23210205137729645, "global_step": 109550, "epoch": 1230, "lr": 6.508042175184037e-05}
{"train_loss": 0.13944797217845917, "global_step": 109551, "epoch": 1230, "lr": 6.507986901146642e-05}
{"train_loss": 0.15481603145599365, "global_step": 109552, "epoch": 1230, "lr": 6.507931626906514e-05}
{"train_loss": 0.1557261049747467, "global_step": 109553, "epoch": 1230, "lr": 6.507876352463664e-05}
{"train_loss": 0.14321772754192352, "global_step": 109554, "epoch": 1230, "lr": 6.507821077818099e-05}
{"train_loss": 0.08182020485401154, "global_step": 109555, "epoch": 1230, "lr": 6.507765802969825e-05}
{"train_loss": 0.2078806310892105, "global_step": 109556, "epoch": 1230, "lr": 6.507710527918851e-05}
{"train_loss": 0.14580632746219635, "global_step": 109557, "epoch": 1230, "lr": 6.507655252665183e-05}
{"train_loss": 0.1600314231401079, "global_step": 109558, "epoch": 1230, "lr": 6.50759997720883e-05, "val_loss": 5.021698951721191, "train_action_mse_error": 32.78429412841797}
{"train_loss": 0.12151327729225159, "global_step": 109559, "epoch": 1231, "lr": 6.507544701549798e-05}
{"train_loss": 0.12657728791236877, "global_step": 109560, "epoch": 1231, "lr": 6.507489425688096e-05}
{"train_loss": 0.19176657497882843, "global_step": 109561, "epoch": 1231, "lr": 6.507434149623729e-05}
{"train_loss": 0.1582024097442627, "global_step": 109562, "epoch": 1231, "lr": 6.507378873356707e-05}
{"train_loss": 0.12477948516607285, "global_step": 109563, "epoch": 1231, "lr": 6.507323596887034e-05}
{"train_loss": 0.11533752083778381, "global_step": 109564, "epoch": 1231, "lr": 6.507268320214722e-05}
{"train_loss": 0.12322112917900085, "global_step": 109565, "epoch": 1231, "lr": 6.507213043339776e-05}
{"train_loss": 0.2030104696750641, "global_step": 109566, "epoch": 1231, "lr": 6.507157766262204e-05}
{"train_loss": 0.1949017196893692, "global_step": 109567, "epoch": 1231, "lr": 6.50710248898201e-05}
{"train_loss": 0.13668279349803925, "global_step": 109568, "epoch": 1231, "lr": 6.507047211499208e-05}
{"train_loss": 0.2165382206439972, "global_step": 109569, "epoch": 1231, "lr": 6.5069919338138e-05}
{"train_loss": 0.13610586524009705, "global_step": 109570, "epoch": 1231, "lr": 6.506936655925796e-05}
{"train_loss": 0.15328381955623627, "global_step": 109571, "epoch": 1231, "lr": 6.506881377835202e-05}
{"train_loss": 0.18977931141853333, "global_step": 109572, "epoch": 1231, "lr": 6.506826099542027e-05}
{"train_loss": 0.13528890907764435, "global_step": 109573, "epoch": 1231, "lr": 6.506770821046277e-05}
{"train_loss": 0.11824612319469452, "global_step": 109574, "epoch": 1231, "lr": 6.50671554234796e-05}
{"train_loss": 0.09660642594099045, "global_step": 109575, "epoch": 1231, "lr": 6.506660263447084e-05}
{"train_loss": 0.1523509919643402, "global_step": 109576, "epoch": 1231, "lr": 6.506604984343655e-05}
{"train_loss": 0.16766105592250824, "global_step": 109577, "epoch": 1231, "lr": 6.506549705037682e-05}
{"train_loss": 0.1815565824508667, "global_step": 109578, "epoch": 1231, "lr": 6.506494425529171e-05}
{"train_loss": 0.13263441622257233, "global_step": 109579, "epoch": 1231, "lr": 6.506439145818131e-05}
{"train_loss": 0.18696428835391998, "global_step": 109580, "epoch": 1231, "lr": 6.506383865904569e-05}
{"train_loss": 0.1602587252855301, "global_step": 109581, "epoch": 1231, "lr": 6.506328585788491e-05}
{"train_loss": 0.19310134649276733, "global_step": 109582, "epoch": 1231, "lr": 6.506273305469906e-05}
{"train_loss": 0.1549682915210724, "global_step": 109583, "epoch": 1231, "lr": 6.506218024948821e-05}
{"train_loss": 0.14942343533039093, "global_step": 109584, "epoch": 1231, "lr": 6.506162744225244e-05}
{"train_loss": 0.13595455884933472, "global_step": 109585, "epoch": 1231, "lr": 6.506107463299181e-05}
{"train_loss": 0.17639368772506714, "global_step": 109586, "epoch": 1231, "lr": 6.50605218217064e-05}
{"train_loss": 0.17114247381687164, "global_step": 109587, "epoch": 1231, "lr": 6.505996900839628e-05}
{"train_loss": 0.1817740499973297, "global_step": 109588, "epoch": 1231, "lr": 6.505941619306156e-05}
{"train_loss": 0.19135373830795288, "global_step": 109589, "epoch": 1231, "lr": 6.505886337570226e-05}
{"train_loss": 0.13194558024406433, "global_step": 109590, "epoch": 1231, "lr": 6.505831055631848e-05}
{"train_loss": 0.17025363445281982, "global_step": 109591, "epoch": 1231, "lr": 6.50577577349103e-05}
{"train_loss": 0.13749828934669495, "global_step": 109592, "epoch": 1231, "lr": 6.505720491147777e-05}
{"train_loss": 0.17827917635440826, "global_step": 109593, "epoch": 1231, "lr": 6.505665208602101e-05}
{"train_loss": 0.18858660757541656, "global_step": 109594, "epoch": 1231, "lr": 6.505609925854005e-05}
{"train_loss": 0.14438113570213318, "global_step": 109595, "epoch": 1231, "lr": 6.505554642903499e-05}
{"train_loss": 0.19606125354766846, "global_step": 109596, "epoch": 1231, "lr": 6.505499359750588e-05}
{"train_loss": 0.1448306441307068, "global_step": 109597, "epoch": 1231, "lr": 6.505444076395283e-05}
{"train_loss": 0.18024881184101105, "global_step": 109598, "epoch": 1231, "lr": 6.505388792837587e-05}
{"train_loss": 0.14933957159519196, "global_step": 109599, "epoch": 1231, "lr": 6.505333509077513e-05}
{"train_loss": 0.19244499504566193, "global_step": 109600, "epoch": 1231, "lr": 6.505278225115063e-05}
{"train_loss": 0.09457506984472275, "global_step": 109601, "epoch": 1231, "lr": 6.505222940950248e-05}
{"train_loss": 0.1860002875328064, "global_step": 109602, "epoch": 1231, "lr": 6.505167656583073e-05}
{"train_loss": 0.12018056213855743, "global_step": 109603, "epoch": 1231, "lr": 6.505112372013547e-05}
{"train_loss": 0.1014954075217247, "global_step": 109604, "epoch": 1231, "lr": 6.505057087241676e-05}
{"train_loss": 0.1629187911748886, "global_step": 109605, "epoch": 1231, "lr": 6.50500180226747e-05}
{"train_loss": 0.15682683885097504, "global_step": 109606, "epoch": 1231, "lr": 6.504946517090933e-05}
{"train_loss": 0.1967187225818634, "global_step": 109607, "epoch": 1231, "lr": 6.504891231712075e-05}
{"train_loss": 0.24343779683113098, "global_step": 109608, "epoch": 1231, "lr": 6.504835946130904e-05}
{"train_loss": 0.18311233818531036, "global_step": 109609, "epoch": 1231, "lr": 6.504780660347425e-05}
{"train_loss": 0.19963930547237396, "global_step": 109610, "epoch": 1231, "lr": 6.504725374361647e-05}
{"train_loss": 0.1484859436750412, "global_step": 109611, "epoch": 1231, "lr": 6.504670088173577e-05}
{"train_loss": 0.13152290880680084, "global_step": 109612, "epoch": 1231, "lr": 6.504614801783222e-05}
{"train_loss": 0.24281422793865204, "global_step": 109613, "epoch": 1231, "lr": 6.50455951519059e-05}
{"train_loss": 0.1308314949274063, "global_step": 109614, "epoch": 1231, "lr": 6.504504228395689e-05}
{"train_loss": 0.12966224551200867, "global_step": 109615, "epoch": 1231, "lr": 6.504448941398526e-05}
{"train_loss": 0.1401788592338562, "global_step": 109616, "epoch": 1231, "lr": 6.504393654199107e-05}
{"train_loss": 0.14863629639148712, "global_step": 109617, "epoch": 1231, "lr": 6.50433836679744e-05}
{"train_loss": 0.12801170349121094, "global_step": 109618, "epoch": 1231, "lr": 6.504283079193535e-05}
{"train_loss": 0.1583268642425537, "global_step": 109619, "epoch": 1231, "lr": 6.504227791387397e-05}
{"train_loss": 0.13190312683582306, "global_step": 109620, "epoch": 1231, "lr": 6.504172503379033e-05}
{"train_loss": 0.2179870307445526, "global_step": 109621, "epoch": 1231, "lr": 6.504117215168451e-05}
{"train_loss": 0.16571199893951416, "global_step": 109622, "epoch": 1231, "lr": 6.50406192675566e-05}
{"train_loss": 0.1716906726360321, "global_step": 109623, "epoch": 1231, "lr": 6.504006638140667e-05}
{"train_loss": 0.09834679961204529, "global_step": 109624, "epoch": 1231, "lr": 6.503951349323476e-05}
{"train_loss": 0.15045370161533356, "global_step": 109625, "epoch": 1231, "lr": 6.503896060304098e-05}
{"train_loss": 0.09731709957122803, "global_step": 109626, "epoch": 1231, "lr": 6.503840771082541e-05}
{"train_loss": 0.11332801729440689, "global_step": 109627, "epoch": 1231, "lr": 6.50378548165881e-05}
{"train_loss": 0.16898319125175476, "global_step": 109628, "epoch": 1231, "lr": 6.503730192032913e-05}
{"train_loss": 0.20453591644763947, "global_step": 109629, "epoch": 1231, "lr": 6.503674902204859e-05}
{"train_loss": 0.16757835447788239, "global_step": 109630, "epoch": 1231, "lr": 6.503619612174653e-05}
{"train_loss": 0.1253974735736847, "global_step": 109631, "epoch": 1231, "lr": 6.503564321942304e-05}
{"train_loss": 0.14594750106334686, "global_step": 109632, "epoch": 1231, "lr": 6.503509031507819e-05}
{"train_loss": 0.1991913616657257, "global_step": 109633, "epoch": 1231, "lr": 6.503453740871206e-05}
{"train_loss": 0.10231559723615646, "global_step": 109634, "epoch": 1231, "lr": 6.503398450032471e-05}
{"train_loss": 0.16257208585739136, "global_step": 109635, "epoch": 1231, "lr": 6.503343158991623e-05}
{"train_loss": 0.10111857205629349, "global_step": 109636, "epoch": 1231, "lr": 6.503287867748668e-05}
{"train_loss": 0.14401723444461823, "global_step": 109637, "epoch": 1231, "lr": 6.503232576303616e-05}
{"train_loss": 0.1875591278076172, "global_step": 109638, "epoch": 1231, "lr": 6.503177284656472e-05}
{"train_loss": 0.28122448921203613, "global_step": 109639, "epoch": 1231, "lr": 6.503121992807244e-05}
{"train_loss": 0.16655902564525604, "global_step": 109640, "epoch": 1231, "lr": 6.50306670075594e-05}
{"train_loss": 0.09719596803188324, "global_step": 109641, "epoch": 1231, "lr": 6.503011408502567e-05}
{"train_loss": 0.19474636018276215, "global_step": 109642, "epoch": 1231, "lr": 6.502956116047131e-05}
{"train_loss": 0.12257590889930725, "global_step": 109643, "epoch": 1231, "lr": 6.502900823389641e-05}
{"train_loss": 0.16174429655075073, "global_step": 109644, "epoch": 1231, "lr": 6.502845530530106e-05}
{"train_loss": 0.12218040972948074, "global_step": 109645, "epoch": 1231, "lr": 6.50279023746853e-05}
{"train_loss": 0.10056313127279282, "global_step": 109646, "epoch": 1231, "lr": 6.502734944204923e-05}
{"train_loss": 0.15693482965900657, "global_step": 109647, "epoch": 1231, "lr": 6.502679650739292e-05, "val_loss": 4.939824104309082}
{"train_loss": 0.20768338441848755, "global_step": 109648, "epoch": 1232, "lr": 6.502624357071644e-05}
{"train_loss": 0.18590670824050903, "global_step": 109649, "epoch": 1232, "lr": 6.502569063201987e-05}
{"train_loss": 0.24371086061000824, "global_step": 109650, "epoch": 1232, "lr": 6.502513769130326e-05}
{"train_loss": 0.16934263706207275, "global_step": 109651, "epoch": 1232, "lr": 6.502458474856671e-05}
{"train_loss": 0.19939248263835907, "global_step": 109652, "epoch": 1232, "lr": 6.50240318038103e-05}
{"train_loss": 0.1656528264284134, "global_step": 109653, "epoch": 1232, "lr": 6.502347885703408e-05}
{"train_loss": 0.14459192752838135, "global_step": 109654, "epoch": 1232, "lr": 6.502292590823814e-05}
{"train_loss": 0.13589440286159515, "global_step": 109655, "epoch": 1232, "lr": 6.502237295742256e-05}
{"train_loss": 0.10854475945234299, "global_step": 109656, "epoch": 1232, "lr": 6.502182000458738e-05}
{"train_loss": 0.1177288293838501, "global_step": 109657, "epoch": 1232, "lr": 6.502126704973272e-05}
{"train_loss": 0.12873747944831848, "global_step": 109658, "epoch": 1232, "lr": 6.502071409285864e-05}
{"train_loss": 0.16902941465377808, "global_step": 109659, "epoch": 1232, "lr": 6.502016113396519e-05}
{"train_loss": 0.06664211302995682, "global_step": 109660, "epoch": 1232, "lr": 6.501960817305245e-05}
{"train_loss": 0.17746689915657043, "global_step": 109661, "epoch": 1232, "lr": 6.501905521012054e-05}
{"train_loss": 0.15209047496318817, "global_step": 109662, "epoch": 1232, "lr": 6.501850224516949e-05}
{"train_loss": 0.14017298817634583, "global_step": 109663, "epoch": 1232, "lr": 6.501794927819939e-05}
{"train_loss": 0.16442784667015076, "global_step": 109664, "epoch": 1232, "lr": 6.501739630921029e-05}
{"train_loss": 0.16614797711372375, "global_step": 109665, "epoch": 1232, "lr": 6.501684333820231e-05}
{"train_loss": 0.13732530176639557, "global_step": 109666, "epoch": 1232, "lr": 6.501629036517547e-05}
{"train_loss": 0.2049911469221115, "global_step": 109667, "epoch": 1232, "lr": 6.50157373901299e-05}
{"train_loss": 0.12914876639842987, "global_step": 109668, "epoch": 1232, "lr": 6.501518441306562e-05}
{"train_loss": 0.19192366302013397, "global_step": 109669, "epoch": 1232, "lr": 6.501463143398275e-05}
{"train_loss": 0.15174855291843414, "global_step": 109670, "epoch": 1232, "lr": 6.501407845288134e-05}
{"train_loss": 0.12491493672132492, "global_step": 109671, "epoch": 1232, "lr": 6.501352546976148e-05}
{"train_loss": 0.15218821167945862, "global_step": 109672, "epoch": 1232, "lr": 6.501297248462322e-05}
{"train_loss": 0.16383543610572815, "global_step": 109673, "epoch": 1232, "lr": 6.501241949746666e-05}
{"train_loss": 0.1455475091934204, "global_step": 109674, "epoch": 1232, "lr": 6.501186650829187e-05}
{"train_loss": 0.184776172041893, "global_step": 109675, "epoch": 1232, "lr": 6.50113135170989e-05}
{"train_loss": 0.146335631608963, "global_step": 109676, "epoch": 1232, "lr": 6.501076052388783e-05}
{"train_loss": 0.07967839390039444, "global_step": 109677, "epoch": 1232, "lr": 6.501020752865878e-05}
{"train_loss": 0.20810577273368835, "global_step": 109678, "epoch": 1232, "lr": 6.500965453141176e-05}
{"train_loss": 0.1839013397693634, "global_step": 109679, "epoch": 1232, "lr": 6.50091015321469e-05}
{"train_loss": 0.10553324967622757, "global_step": 109680, "epoch": 1232, "lr": 6.500854853086424e-05}
{"train_loss": 0.1359538584947586, "global_step": 109681, "epoch": 1232, "lr": 6.500799552756386e-05}
{"train_loss": 0.15500876307487488, "global_step": 109682, "epoch": 1232, "lr": 6.500744252224584e-05}
{"train_loss": 0.158306285738945, "global_step": 109683, "epoch": 1232, "lr": 6.500688951491025e-05}
{"train_loss": 0.17146120965480804, "global_step": 109684, "epoch": 1232, "lr": 6.500633650555718e-05}
{"train_loss": 0.1925189346075058, "global_step": 109685, "epoch": 1232, "lr": 6.500578349418668e-05}
{"train_loss": 0.17010672390460968, "global_step": 109686, "epoch": 1232, "lr": 6.500523048079883e-05}
{"train_loss": 0.23151196539402008, "global_step": 109687, "epoch": 1232, "lr": 6.500467746539371e-05}
{"train_loss": 0.12515795230865479, "global_step": 109688, "epoch": 1232, "lr": 6.500412444797142e-05}
{"train_loss": 0.1929740607738495, "global_step": 109689, "epoch": 1232, "lr": 6.500357142853198e-05}
{"train_loss": 0.12733379006385803, "global_step": 109690, "epoch": 1232, "lr": 6.50030184070755e-05}
{"train_loss": 0.13098064064979553, "global_step": 109691, "epoch": 1232, "lr": 6.500246538360204e-05}
{"train_loss": 0.1658637970685959, "global_step": 109692, "epoch": 1232, "lr": 6.500191235811169e-05}
{"train_loss": 0.14705653488636017, "global_step": 109693, "epoch": 1232, "lr": 6.500135933060452e-05}
{"train_loss": 0.12804622948169708, "global_step": 109694, "epoch": 1232, "lr": 6.500080630108058e-05}
{"train_loss": 0.07018725574016571, "global_step": 109695, "epoch": 1232, "lr": 6.500025326953998e-05}
{"train_loss": 0.15936505794525146, "global_step": 109696, "epoch": 1232, "lr": 6.499970023598278e-05}
{"train_loss": 0.13250517845153809, "global_step": 109697, "epoch": 1232, "lr": 6.499914720040907e-05}
{"train_loss": 0.10063862800598145, "global_step": 109698, "epoch": 1232, "lr": 6.499859416281888e-05}
{"train_loss": 0.16714619100093842, "global_step": 109699, "epoch": 1232, "lr": 6.499804112321232e-05}
{"train_loss": 0.17498771846294403, "global_step": 109700, "epoch": 1232, "lr": 6.499748808158946e-05}
{"train_loss": 0.15046870708465576, "global_step": 109701, "epoch": 1232, "lr": 6.499693503795036e-05}
{"train_loss": 0.10485528409481049, "global_step": 109702, "epoch": 1232, "lr": 6.499638199229512e-05}
{"train_loss": 0.11643890291452408, "global_step": 109703, "epoch": 1232, "lr": 6.499582894462378e-05}
{"train_loss": 0.13093242049217224, "global_step": 109704, "epoch": 1232, "lr": 6.499527589493645e-05}
{"train_loss": 0.08300091326236725, "global_step": 109705, "epoch": 1232, "lr": 6.499472284323318e-05}
{"train_loss": 0.1562427133321762, "global_step": 109706, "epoch": 1232, "lr": 6.499416978951405e-05}
{"train_loss": 0.2424565702676773, "global_step": 109707, "epoch": 1232, "lr": 6.499361673377915e-05}
{"train_loss": 0.09655975550413132, "global_step": 109708, "epoch": 1232, "lr": 6.499306367602853e-05}
{"train_loss": 0.1488671898841858, "global_step": 109709, "epoch": 1232, "lr": 6.499251061626228e-05}
{"train_loss": 0.1376095414161682, "global_step": 109710, "epoch": 1232, "lr": 6.499195755448047e-05}
{"train_loss": 0.18320493400096893, "global_step": 109711, "epoch": 1232, "lr": 6.499140449068316e-05}
{"train_loss": 0.21678422391414642, "global_step": 109712, "epoch": 1232, "lr": 6.499085142487045e-05}
{"train_loss": 0.09655711054801941, "global_step": 109713, "epoch": 1232, "lr": 6.49902983570424e-05}
{"train_loss": 0.19018042087554932, "global_step": 109714, "epoch": 1232, "lr": 6.498974528719909e-05}
{"train_loss": 0.07946762442588806, "global_step": 109715, "epoch": 1232, "lr": 6.498919221534059e-05}
{"train_loss": 0.09763927757740021, "global_step": 109716, "epoch": 1232, "lr": 6.498863914146697e-05}
{"train_loss": 0.221885547041893, "global_step": 109717, "epoch": 1232, "lr": 6.498808606557834e-05}
{"train_loss": 0.1774543672800064, "global_step": 109718, "epoch": 1232, "lr": 6.498753298767471e-05}
{"train_loss": 0.1432880163192749, "global_step": 109719, "epoch": 1232, "lr": 6.49869799077562e-05}
{"train_loss": 0.12501351535320282, "global_step": 109720, "epoch": 1232, "lr": 6.498642682582287e-05}
{"train_loss": 0.09687696397304535, "global_step": 109721, "epoch": 1232, "lr": 6.498587374187482e-05}
{"train_loss": 0.12791943550109863, "global_step": 109722, "epoch": 1232, "lr": 6.498532065591208e-05}
{"train_loss": 0.14398688077926636, "global_step": 109723, "epoch": 1232, "lr": 6.498476756793475e-05}
{"train_loss": 0.10728015005588531, "global_step": 109724, "epoch": 1232, "lr": 6.498421447794289e-05}
{"train_loss": 0.11309996247291565, "global_step": 109725, "epoch": 1232, "lr": 6.498366138593659e-05}
{"train_loss": 0.12873877584934235, "global_step": 109726, "epoch": 1232, "lr": 6.498310829191593e-05}
{"train_loss": 0.17797309160232544, "global_step": 109727, "epoch": 1232, "lr": 6.498255519588096e-05}
{"train_loss": 0.0949406623840332, "global_step": 109728, "epoch": 1232, "lr": 6.498200209783178e-05}
{"train_loss": 0.1342075765132904, "global_step": 109729, "epoch": 1232, "lr": 6.498144899776846e-05}
{"train_loss": 0.12379470467567444, "global_step": 109730, "epoch": 1232, "lr": 6.498089589569104e-05}
{"train_loss": 0.14275021851062775, "global_step": 109731, "epoch": 1232, "lr": 6.498034279159964e-05}
{"train_loss": 0.1546129584312439, "global_step": 109732, "epoch": 1232, "lr": 6.497978968549431e-05}
{"train_loss": 0.16094036400318146, "global_step": 109733, "epoch": 1232, "lr": 6.497923657737512e-05}
{"train_loss": 0.19162647426128387, "global_step": 109734, "epoch": 1232, "lr": 6.497868346724218e-05}
{"train_loss": 0.16025841236114502, "global_step": 109735, "epoch": 1232, "lr": 6.497813035509552e-05}
{"train_loss": 0.1504597289676077, "global_step": 109736, "epoch": 1232, "lr": 6.497757724093525e-05, "val_loss": 5.032329082489014}
{"train_loss": 0.1668263077735901, "global_step": 109737, "epoch": 1233, "lr": 6.49770241247614e-05}
{"train_loss": 0.10240995138883591, "global_step": 109738, "epoch": 1233, "lr": 6.497647100657408e-05}
{"train_loss": 0.14853309094905853, "global_step": 109739, "epoch": 1233, "lr": 6.497591788637339e-05}
{"train_loss": 0.18172766268253326, "global_step": 109740, "epoch": 1233, "lr": 6.497536476415933e-05}
{"train_loss": 0.15808910131454468, "global_step": 109741, "epoch": 1233, "lr": 6.497481163993203e-05}
{"train_loss": 0.11048545688390732, "global_step": 109742, "epoch": 1233, "lr": 6.497425851369154e-05}
{"train_loss": 0.13338932394981384, "global_step": 109743, "epoch": 1233, "lr": 6.497370538543796e-05}
{"train_loss": 0.16083523631095886, "global_step": 109744, "epoch": 1233, "lr": 6.497315225517134e-05}
{"train_loss": 0.20268207788467407, "global_step": 109745, "epoch": 1233, "lr": 6.497259912289175e-05}
{"train_loss": 0.17818938195705414, "global_step": 109746, "epoch": 1233, "lr": 6.49720459885993e-05}
{"train_loss": 0.08223924785852432, "global_step": 109747, "epoch": 1233, "lr": 6.497149285229402e-05}
{"train_loss": 0.23034098744392395, "global_step": 109748, "epoch": 1233, "lr": 6.497093971397601e-05}
{"train_loss": 0.15717749297618866, "global_step": 109749, "epoch": 1233, "lr": 6.497038657364534e-05}
{"train_loss": 0.1530960649251938, "global_step": 109750, "epoch": 1233, "lr": 6.496983343130207e-05}
{"train_loss": 0.10862681269645691, "global_step": 109751, "epoch": 1233, "lr": 6.496928028694632e-05}
{"train_loss": 0.2046796679496765, "global_step": 109752, "epoch": 1233, "lr": 6.496872714057812e-05}
{"train_loss": 0.17676541209220886, "global_step": 109753, "epoch": 1233, "lr": 6.496817399219754e-05}
{"train_loss": 0.13094815611839294, "global_step": 109754, "epoch": 1233, "lr": 6.49676208418047e-05}
{"train_loss": 0.140961691737175, "global_step": 109755, "epoch": 1233, "lr": 6.496706768939963e-05}
{"train_loss": 0.2184169739484787, "global_step": 109756, "epoch": 1233, "lr": 6.496651453498243e-05}
{"train_loss": 0.18050014972686768, "global_step": 109757, "epoch": 1233, "lr": 6.496596137855315e-05}
{"train_loss": 0.1770435869693756, "global_step": 109758, "epoch": 1233, "lr": 6.496540822011189e-05}
{"train_loss": 0.1328105628490448, "global_step": 109759, "epoch": 1233, "lr": 6.496485505965871e-05}
{"train_loss": 0.20606577396392822, "global_step": 109760, "epoch": 1233, "lr": 6.49643018971937e-05}
{"train_loss": 0.043142784386873245, "global_step": 109761, "epoch": 1233, "lr": 6.49637487327169e-05}
{"train_loss": 0.13270695507526398, "global_step": 109762, "epoch": 1233, "lr": 6.496319556622841e-05}
{"train_loss": 0.18650111556053162, "global_step": 109763, "epoch": 1233, "lr": 6.496264239772831e-05}
{"train_loss": 0.13993632793426514, "global_step": 109764, "epoch": 1233, "lr": 6.496208922721664e-05}
{"train_loss": 0.13649772107601166, "global_step": 109765, "epoch": 1233, "lr": 6.496153605469353e-05}
{"train_loss": 0.12026350945234299, "global_step": 109766, "epoch": 1233, "lr": 6.496098288015902e-05}
{"train_loss": 0.17587457597255707, "global_step": 109767, "epoch": 1233, "lr": 6.496042970361317e-05}
{"train_loss": 0.12463725358247757, "global_step": 109768, "epoch": 1233, "lr": 6.495987652505608e-05}
{"train_loss": 0.1875762790441513, "global_step": 109769, "epoch": 1233, "lr": 6.495932334448782e-05}
{"train_loss": 0.15546341240406036, "global_step": 109770, "epoch": 1233, "lr": 6.495877016190846e-05}
{"train_loss": 0.1701715588569641, "global_step": 109771, "epoch": 1233, "lr": 6.495821697731808e-05}
{"train_loss": 0.11331178992986679, "global_step": 109772, "epoch": 1233, "lr": 6.495766379071675e-05}
{"train_loss": 0.12292496860027313, "global_step": 109773, "epoch": 1233, "lr": 6.495711060210454e-05}
{"train_loss": 0.1733483374118805, "global_step": 109774, "epoch": 1233, "lr": 6.495655741148155e-05}
{"train_loss": 0.1101742535829544, "global_step": 109775, "epoch": 1233, "lr": 6.495600421884783e-05}
{"train_loss": 0.1936398595571518, "global_step": 109776, "epoch": 1233, "lr": 6.495545102420342e-05}
{"train_loss": 0.15928348898887634, "global_step": 109777, "epoch": 1233, "lr": 6.495489782754846e-05}
{"train_loss": 0.09762551635503769, "global_step": 109778, "epoch": 1233, "lr": 6.4954344628883e-05}
{"train_loss": 0.11813876032829285, "global_step": 109779, "epoch": 1233, "lr": 6.49537914282071e-05}
{"train_loss": 0.1207963302731514, "global_step": 109780, "epoch": 1233, "lr": 6.495323822552084e-05}
{"train_loss": 0.1757388710975647, "global_step": 109781, "epoch": 1233, "lr": 6.49526850208243e-05}
{"train_loss": 0.15186917781829834, "global_step": 109782, "epoch": 1233, "lr": 6.495213181411756e-05}
{"train_loss": 0.14907123148441315, "global_step": 109783, "epoch": 1233, "lr": 6.495157860540069e-05}
{"train_loss": 0.1615653932094574, "global_step": 109784, "epoch": 1233, "lr": 6.495102539467376e-05}
{"train_loss": 0.18136465549468994, "global_step": 109785, "epoch": 1233, "lr": 6.495047218193684e-05}
{"train_loss": 0.13866963982582092, "global_step": 109786, "epoch": 1233, "lr": 6.494991896719001e-05}
{"train_loss": 0.1901540607213974, "global_step": 109787, "epoch": 1233, "lr": 6.494936575043335e-05}
{"train_loss": 0.16735032200813293, "global_step": 109788, "epoch": 1233, "lr": 6.494881253166695e-05}
{"train_loss": 0.12011566758155823, "global_step": 109789, "epoch": 1233, "lr": 6.494825931089084e-05}
{"train_loss": 0.14189350605010986, "global_step": 109790, "epoch": 1233, "lr": 6.494770608810513e-05}
{"train_loss": 0.13982096314430237, "global_step": 109791, "epoch": 1233, "lr": 6.494715286330987e-05}
{"train_loss": 0.18852578103542328, "global_step": 109792, "epoch": 1233, "lr": 6.494659963650516e-05}
{"train_loss": 0.13529475033283234, "global_step": 109793, "epoch": 1233, "lr": 6.494604640769107e-05}
{"train_loss": 0.1154574528336525, "global_step": 109794, "epoch": 1233, "lr": 6.494549317686767e-05}
{"train_loss": 0.1409694254398346, "global_step": 109795, "epoch": 1233, "lr": 6.4944939944035e-05}
{"train_loss": 0.09366929531097412, "global_step": 109796, "epoch": 1233, "lr": 6.494438670919318e-05}
{"train_loss": 0.1180841475725174, "global_step": 109797, "epoch": 1233, "lr": 6.494383347234227e-05}
{"train_loss": 0.16177286207675934, "global_step": 109798, "epoch": 1233, "lr": 6.494328023348235e-05}
{"train_loss": 0.08728480339050293, "global_step": 109799, "epoch": 1233, "lr": 6.494272699261348e-05}
{"train_loss": 0.17131103575229645, "global_step": 109800, "epoch": 1233, "lr": 6.494217374973575e-05}
{"train_loss": 0.21306611597537994, "global_step": 109801, "epoch": 1233, "lr": 6.494162050484922e-05}
{"train_loss": 0.14341877400875092, "global_step": 109802, "epoch": 1233, "lr": 6.494106725795397e-05}
{"train_loss": 0.12975408136844635, "global_step": 109803, "epoch": 1233, "lr": 6.494051400905007e-05}
{"train_loss": 0.11973439902067184, "global_step": 109804, "epoch": 1233, "lr": 6.493996075813761e-05}
{"train_loss": 0.08155660331249237, "global_step": 109805, "epoch": 1233, "lr": 6.493940750521666e-05}
{"train_loss": 0.17625349760055542, "global_step": 109806, "epoch": 1233, "lr": 6.493885425028727e-05}
{"train_loss": 0.1325545459985733, "global_step": 109807, "epoch": 1233, "lr": 6.493830099334954e-05}
{"train_loss": 0.17545023560523987, "global_step": 109808, "epoch": 1233, "lr": 6.493774773440355e-05}
{"train_loss": 0.17867131531238556, "global_step": 109809, "epoch": 1233, "lr": 6.493719447344935e-05}
{"train_loss": 0.20477639138698578, "global_step": 109810, "epoch": 1233, "lr": 6.493664121048704e-05}
{"train_loss": 0.133090540766716, "global_step": 109811, "epoch": 1233, "lr": 6.493608794551666e-05}
{"train_loss": 0.16217944025993347, "global_step": 109812, "epoch": 1233, "lr": 6.493553467853833e-05}
{"train_loss": 0.14964479207992554, "global_step": 109813, "epoch": 1233, "lr": 6.493498140955208e-05}
{"train_loss": 0.14880315959453583, "global_step": 109814, "epoch": 1233, "lr": 6.493442813855801e-05}
{"train_loss": 0.18576213717460632, "global_step": 109815, "epoch": 1233, "lr": 6.49338748655562e-05}
{"train_loss": 0.12717394530773163, "global_step": 109816, "epoch": 1233, "lr": 6.49333215905467e-05}
{"train_loss": 0.08877671509981155, "global_step": 109817, "epoch": 1233, "lr": 6.493276831352959e-05}
{"train_loss": 0.1804286390542984, "global_step": 109818, "epoch": 1233, "lr": 6.493221503450495e-05}
{"train_loss": 0.22837066650390625, "global_step": 109819, "epoch": 1233, "lr": 6.493166175347287e-05}
{"train_loss": 0.12858060002326965, "global_step": 109820, "epoch": 1233, "lr": 6.49311084704334e-05}
{"train_loss": 0.22585955262184143, "global_step": 109821, "epoch": 1233, "lr": 6.493055518538663e-05}
{"train_loss": 0.2074514627456665, "global_step": 109822, "epoch": 1233, "lr": 6.493000189833262e-05}
{"train_loss": 0.13138549029827118, "global_step": 109823, "epoch": 1233, "lr": 6.492944860927146e-05}
{"train_loss": 0.14957201480865479, "global_step": 109824, "epoch": 1233, "lr": 6.492889531820321e-05}
{"train_loss": 0.15115020905485313, "global_step": 109825, "epoch": 1233, "lr": 6.492834202512796e-05, "val_loss": 5.090782642364502}
{"train_loss": 0.15313266217708588, "global_step": 109826, "epoch": 1234, "lr": 6.492778873004578e-05}
{"train_loss": 0.08112962543964386, "global_step": 109827, "epoch": 1234, "lr": 6.492723543295674e-05}
{"train_loss": 0.10231553763151169, "global_step": 109828, "epoch": 1234, "lr": 6.492668213386092e-05}
{"train_loss": 0.13179005682468414, "global_step": 109829, "epoch": 1234, "lr": 6.492612883275837e-05}
{"train_loss": 0.18763568997383118, "global_step": 109830, "epoch": 1234, "lr": 6.49255755296492e-05}
{"train_loss": 0.14182205498218536, "global_step": 109831, "epoch": 1234, "lr": 6.492502222453346e-05}
{"train_loss": 0.18797685205936432, "global_step": 109832, "epoch": 1234, "lr": 6.492446891741125e-05}
{"train_loss": 0.16112621128559113, "global_step": 109833, "epoch": 1234, "lr": 6.492391560828262e-05}
{"train_loss": 0.18821574747562408, "global_step": 109834, "epoch": 1234, "lr": 6.492336229714763e-05}
{"train_loss": 0.10804018378257751, "global_step": 109835, "epoch": 1234, "lr": 6.49228089840064e-05}
{"train_loss": 0.1135745495557785, "global_step": 109836, "epoch": 1234, "lr": 6.492225566885897e-05}
{"train_loss": 0.12661106884479523, "global_step": 109837, "epoch": 1234, "lr": 6.492170235170542e-05}
{"train_loss": 0.14724503457546234, "global_step": 109838, "epoch": 1234, "lr": 6.492114903254585e-05}
{"train_loss": 0.11518914252519608, "global_step": 109839, "epoch": 1234, "lr": 6.492059571138027e-05}
{"train_loss": 0.15172193944454193, "global_step": 109840, "epoch": 1234, "lr": 6.492004238820884e-05}
{"train_loss": 0.1124931275844574, "global_step": 109841, "epoch": 1234, "lr": 6.491948906303157e-05}
{"train_loss": 0.14571814239025116, "global_step": 109842, "epoch": 1234, "lr": 6.491893573584857e-05}
{"train_loss": 0.1279882937669754, "global_step": 109843, "epoch": 1234, "lr": 6.491838240665989e-05}
{"train_loss": 0.16197840869426727, "global_step": 109844, "epoch": 1234, "lr": 6.491782907546562e-05}
{"train_loss": 0.16504734754562378, "global_step": 109845, "epoch": 1234, "lr": 6.491727574226583e-05}
{"train_loss": 0.12612320482730865, "global_step": 109846, "epoch": 1234, "lr": 6.491672240706059e-05}
{"train_loss": 0.16207346320152283, "global_step": 109847, "epoch": 1234, "lr": 6.491616906984998e-05}
{"train_loss": 0.1224680244922638, "global_step": 109848, "epoch": 1234, "lr": 6.491561573063408e-05}
{"train_loss": 0.11257050931453705, "global_step": 109849, "epoch": 1234, "lr": 6.491506238941296e-05}
{"train_loss": 0.21840831637382507, "global_step": 109850, "epoch": 1234, "lr": 6.491450904618668e-05}
{"train_loss": 0.18553273379802704, "global_step": 109851, "epoch": 1234, "lr": 6.491395570095534e-05}
{"train_loss": 0.11943000555038452, "global_step": 109852, "epoch": 1234, "lr": 6.4913402353719e-05}
{"train_loss": 0.16916725039482117, "global_step": 109853, "epoch": 1234, "lr": 6.491284900447771e-05}
{"train_loss": 0.17783783376216888, "global_step": 109854, "epoch": 1234, "lr": 6.49122956532316e-05}
{"train_loss": 0.18896612524986267, "global_step": 109855, "epoch": 1234, "lr": 6.49117422999807e-05}
{"train_loss": 0.10872000455856323, "global_step": 109856, "epoch": 1234, "lr": 6.49111889447251e-05}
{"train_loss": 0.16800348460674286, "global_step": 109857, "epoch": 1234, "lr": 6.491063558746485e-05}
{"train_loss": 0.13822203874588013, "global_step": 109858, "epoch": 1234, "lr": 6.491008222820007e-05}
{"train_loss": 0.18272040784358978, "global_step": 109859, "epoch": 1234, "lr": 6.49095288669308e-05}
{"train_loss": 0.17354385554790497, "global_step": 109860, "epoch": 1234, "lr": 6.490897550365714e-05}
{"train_loss": 0.12368987500667572, "global_step": 109861, "epoch": 1234, "lr": 6.490842213837914e-05}
{"train_loss": 0.14506502449512482, "global_step": 109862, "epoch": 1234, "lr": 6.490786877109689e-05}
{"train_loss": 0.11475830525159836, "global_step": 109863, "epoch": 1234, "lr": 6.490731540181046e-05}
{"train_loss": 0.20601898431777954, "global_step": 109864, "epoch": 1234, "lr": 6.490676203051991e-05}
{"train_loss": 0.1231880784034729, "global_step": 109865, "epoch": 1234, "lr": 6.490620865722535e-05}
{"train_loss": 0.12555649876594543, "global_step": 109866, "epoch": 1234, "lr": 6.49056552819268e-05}
{"train_loss": 0.2574530839920044, "global_step": 109867, "epoch": 1234, "lr": 6.49051019046244e-05}
{"train_loss": 0.09093525260686874, "global_step": 109868, "epoch": 1234, "lr": 6.490454852531817e-05}
{"train_loss": 0.18092434108257294, "global_step": 109869, "epoch": 1234, "lr": 6.490399514400821e-05}
{"train_loss": 0.12549349665641785, "global_step": 109870, "epoch": 1234, "lr": 6.49034417606946e-05}
{"train_loss": 0.15220311284065247, "global_step": 109871, "epoch": 1234, "lr": 6.49028883753774e-05}
{"train_loss": 0.17199181020259857, "global_step": 109872, "epoch": 1234, "lr": 6.490233498805667e-05}
{"train_loss": 0.15102244913578033, "global_step": 109873, "epoch": 1234, "lr": 6.490178159873252e-05}
{"train_loss": 0.26958999037742615, "global_step": 109874, "epoch": 1234, "lr": 6.490122820740501e-05}
{"train_loss": 0.14032861590385437, "global_step": 109875, "epoch": 1234, "lr": 6.490067481407419e-05}
{"train_loss": 0.16333609819412231, "global_step": 109876, "epoch": 1234, "lr": 6.490012141874017e-05}
{"train_loss": 0.10923685878515244, "global_step": 109877, "epoch": 1234, "lr": 6.4899568021403e-05}
{"train_loss": 0.16832759976387024, "global_step": 109878, "epoch": 1234, "lr": 6.489901462206279e-05}
{"train_loss": 0.1252005398273468, "global_step": 109879, "epoch": 1234, "lr": 6.489846122071956e-05}
{"train_loss": 0.16518040001392365, "global_step": 109880, "epoch": 1234, "lr": 6.489790781737342e-05}
{"train_loss": 0.1553921103477478, "global_step": 109881, "epoch": 1234, "lr": 6.489735441202445e-05}
{"train_loss": 0.1869381219148636, "global_step": 109882, "epoch": 1234, "lr": 6.489680100467272e-05}
{"train_loss": 0.10044935345649719, "global_step": 109883, "epoch": 1234, "lr": 6.489624759531827e-05}
{"train_loss": 0.11148524284362793, "global_step": 109884, "epoch": 1234, "lr": 6.489569418396122e-05}
{"train_loss": 0.19208399951457977, "global_step": 109885, "epoch": 1234, "lr": 6.489514077060162e-05}
{"train_loss": 0.16153259575366974, "global_step": 109886, "epoch": 1234, "lr": 6.489458735523954e-05}
{"train_loss": 0.1368333399295807, "global_step": 109887, "epoch": 1234, "lr": 6.489403393787509e-05}
{"train_loss": 0.15298043191432953, "global_step": 109888, "epoch": 1234, "lr": 6.48934805185083e-05}
{"train_loss": 0.21799592673778534, "global_step": 109889, "epoch": 1234, "lr": 6.489292709713927e-05}
{"train_loss": 0.19067153334617615, "global_step": 109890, "epoch": 1234, "lr": 6.489237367376806e-05}
{"train_loss": 0.1717219054698944, "global_step": 109891, "epoch": 1234, "lr": 6.489182024839476e-05}
{"train_loss": 0.17086710035800934, "global_step": 109892, "epoch": 1234, "lr": 6.489126682101943e-05}
{"train_loss": 0.18890878558158875, "global_step": 109893, "epoch": 1234, "lr": 6.489071339164216e-05}
{"train_loss": 0.11506922543048859, "global_step": 109894, "epoch": 1234, "lr": 6.4890159960263e-05}
{"train_loss": 0.22107356786727905, "global_step": 109895, "epoch": 1234, "lr": 6.488960652688204e-05}
{"train_loss": 0.18219874799251556, "global_step": 109896, "epoch": 1234, "lr": 6.488905309149936e-05}
{"train_loss": 0.22826892137527466, "global_step": 109897, "epoch": 1234, "lr": 6.488849965411503e-05}
{"train_loss": 0.15826229751110077, "global_step": 109898, "epoch": 1234, "lr": 6.488794621472912e-05}
{"train_loss": 0.16949251294136047, "global_step": 109899, "epoch": 1234, "lr": 6.488739277334171e-05}
{"train_loss": 0.18543113768100739, "global_step": 109900, "epoch": 1234, "lr": 6.488683932995287e-05}
{"train_loss": 0.10557311028242111, "global_step": 109901, "epoch": 1234, "lr": 6.488628588456267e-05}
{"train_loss": 0.16967558860778809, "global_step": 109902, "epoch": 1234, "lr": 6.488573243717122e-05}
{"train_loss": 0.1390666961669922, "global_step": 109903, "epoch": 1234, "lr": 6.488517898777852e-05}
{"train_loss": 0.18744130432605743, "global_step": 109904, "epoch": 1234, "lr": 6.488462553638472e-05}
{"train_loss": 0.15941578149795532, "global_step": 109905, "epoch": 1234, "lr": 6.488407208298986e-05}
{"train_loss": 0.1716604083776474, "global_step": 109906, "epoch": 1234, "lr": 6.4883518627594e-05}
{"train_loss": 0.2871256172657013, "global_step": 109907, "epoch": 1234, "lr": 6.488296517019726e-05}
{"train_loss": 0.1757456660270691, "global_step": 109908, "epoch": 1234, "lr": 6.488241171079967e-05}
{"train_loss": 0.15214434266090393, "global_step": 109909, "epoch": 1234, "lr": 6.488185824940132e-05}
{"train_loss": 0.15507075190544128, "global_step": 109910, "epoch": 1234, "lr": 6.48813047860023e-05}
{"train_loss": 0.13102120161056519, "global_step": 109911, "epoch": 1234, "lr": 6.488075132060264e-05}
{"train_loss": 0.18231555819511414, "global_step": 109912, "epoch": 1234, "lr": 6.488019785320245e-05}
{"train_loss": 0.21535414457321167, "global_step": 109913, "epoch": 1234, "lr": 6.487964438380182e-05}
{"train_loss": 0.1585168955701121, "global_step": 109914, "epoch": 1234, "lr": 6.487909091240079e-05, "val_loss": 4.850747108459473}
{"train_loss": 0.07062727957963943, "global_step": 109915, "epoch": 1235, "lr": 6.487853743899946e-05}
{"train_loss": 0.14419721066951752, "global_step": 109916, "epoch": 1235, "lr": 6.487798396359788e-05}
{"train_loss": 0.14771394431591034, "global_step": 109917, "epoch": 1235, "lr": 6.487743048619614e-05}
{"train_loss": 0.16058026254177094, "global_step": 109918, "epoch": 1235, "lr": 6.487687700679431e-05}
{"train_loss": 0.15141107141971588, "global_step": 109919, "epoch": 1235, "lr": 6.487632352539246e-05}
{"train_loss": 0.1493290811777115, "global_step": 109920, "epoch": 1235, "lr": 6.487577004199068e-05}
{"train_loss": 0.11819784343242645, "global_step": 109921, "epoch": 1235, "lr": 6.487521655658903e-05}
{"train_loss": 0.17623314261436462, "global_step": 109922, "epoch": 1235, "lr": 6.487466306918759e-05}
{"train_loss": 0.21379289031028748, "global_step": 109923, "epoch": 1235, "lr": 6.487410957978642e-05}
{"train_loss": 0.09006338566541672, "global_step": 109924, "epoch": 1235, "lr": 6.487355608838563e-05}
{"train_loss": 0.10079793632030487, "global_step": 109925, "epoch": 1235, "lr": 6.487300259498525e-05}
{"train_loss": 0.15693852305412292, "global_step": 109926, "epoch": 1235, "lr": 6.487244909958539e-05}
{"train_loss": 0.18155048787593842, "global_step": 109927, "epoch": 1235, "lr": 6.48718956021861e-05}
{"train_loss": 0.13962429761886597, "global_step": 109928, "epoch": 1235, "lr": 6.487134210278749e-05}
{"train_loss": 0.08823523670434952, "global_step": 109929, "epoch": 1235, "lr": 6.487078860138959e-05}
{"train_loss": 0.11646125465631485, "global_step": 109930, "epoch": 1235, "lr": 6.487023509799249e-05}
{"train_loss": 0.19024226069450378, "global_step": 109931, "epoch": 1235, "lr": 6.486968159259627e-05}
{"train_loss": 0.18873924016952515, "global_step": 109932, "epoch": 1235, "lr": 6.4869128085201e-05}
{"train_loss": 0.1228378638625145, "global_step": 109933, "epoch": 1235, "lr": 6.486857457580676e-05}
{"train_loss": 0.1562396138906479, "global_step": 109934, "epoch": 1235, "lr": 6.486802106441363e-05}
{"train_loss": 0.17272193729877472, "global_step": 109935, "epoch": 1235, "lr": 6.486746755102166e-05}
{"train_loss": 0.14774908125400543, "global_step": 109936, "epoch": 1235, "lr": 6.486691403563095e-05}
{"train_loss": 0.1373160034418106, "global_step": 109937, "epoch": 1235, "lr": 6.486636051824156e-05}
{"train_loss": 0.14828546345233917, "global_step": 109938, "epoch": 1235, "lr": 6.486580699885359e-05}
{"train_loss": 0.1677020937204361, "global_step": 109939, "epoch": 1235, "lr": 6.486525347746707e-05}
{"train_loss": 0.17166903614997864, "global_step": 109940, "epoch": 1235, "lr": 6.48646999540821e-05}
{"train_loss": 0.10531612485647202, "global_step": 109941, "epoch": 1235, "lr": 6.486414642869876e-05}
{"train_loss": 0.20805008709430695, "global_step": 109942, "epoch": 1235, "lr": 6.486359290131712e-05}
{"train_loss": 0.08384913206100464, "global_step": 109943, "epoch": 1235, "lr": 6.486303937193723e-05}
{"train_loss": 0.11693915724754333, "global_step": 109944, "epoch": 1235, "lr": 6.48624858405592e-05}
{"train_loss": 0.1257370561361313, "global_step": 109945, "epoch": 1235, "lr": 6.486193230718309e-05}
{"train_loss": 0.11008433252573013, "global_step": 109946, "epoch": 1235, "lr": 6.486137877180898e-05}
{"train_loss": 0.1341673731803894, "global_step": 109947, "epoch": 1235, "lr": 6.486082523443692e-05}
{"train_loss": 0.1695777177810669, "global_step": 109948, "epoch": 1235, "lr": 6.486027169506704e-05}
{"train_loss": 0.21588753163814545, "global_step": 109949, "epoch": 1235, "lr": 6.485971815369933e-05}
{"train_loss": 0.13813501596450806, "global_step": 109950, "epoch": 1235, "lr": 6.485916461033395e-05}
{"train_loss": 0.20415502786636353, "global_step": 109951, "epoch": 1235, "lr": 6.485861106497092e-05}
{"train_loss": 0.08432970941066742, "global_step": 109952, "epoch": 1235, "lr": 6.485805751761033e-05}
{"train_loss": 0.19141355156898499, "global_step": 109953, "epoch": 1235, "lr": 6.485750396825225e-05}
{"train_loss": 0.12040939182043076, "global_step": 109954, "epoch": 1235, "lr": 6.485695041689677e-05}
{"train_loss": 0.17034253478050232, "global_step": 109955, "epoch": 1235, "lr": 6.485639686354396e-05}
{"train_loss": 0.21452657878398895, "global_step": 109956, "epoch": 1235, "lr": 6.485584330819388e-05}
{"train_loss": 0.22418174147605896, "global_step": 109957, "epoch": 1235, "lr": 6.485528975084661e-05}
{"train_loss": 0.13796155154705048, "global_step": 109958, "epoch": 1235, "lr": 6.485473619150224e-05}
{"train_loss": 0.19804628193378448, "global_step": 109959, "epoch": 1235, "lr": 6.48541826301608e-05}
{"train_loss": 0.17755721509456635, "global_step": 109960, "epoch": 1235, "lr": 6.485362906682243e-05}
{"train_loss": 0.1627510040998459, "global_step": 109961, "epoch": 1235, "lr": 6.485307550148715e-05}
{"train_loss": 0.1139126643538475, "global_step": 109962, "epoch": 1235, "lr": 6.485252193415507e-05}
{"train_loss": 0.07651199400424957, "global_step": 109963, "epoch": 1235, "lr": 6.485196836482625e-05}
{"train_loss": 0.30984050035476685, "global_step": 109964, "epoch": 1235, "lr": 6.485141479350076e-05}
{"train_loss": 0.15481507778167725, "global_step": 109965, "epoch": 1235, "lr": 6.485086122017867e-05}
{"train_loss": 0.1281922608613968, "global_step": 109966, "epoch": 1235, "lr": 6.485030764486006e-05}
{"train_loss": 0.14564277231693268, "global_step": 109967, "epoch": 1235, "lr": 6.484975406754503e-05}
{"train_loss": 0.13206112384796143, "global_step": 109968, "epoch": 1235, "lr": 6.484920048823362e-05}
{"train_loss": 0.13002698123455048, "global_step": 109969, "epoch": 1235, "lr": 6.484864690692591e-05}
{"train_loss": 0.19897383451461792, "global_step": 109970, "epoch": 1235, "lr": 6.484809332362198e-05}
{"train_loss": 0.16758964955806732, "global_step": 109971, "epoch": 1235, "lr": 6.48475397383219e-05}
{"train_loss": 0.183940589427948, "global_step": 109972, "epoch": 1235, "lr": 6.484698615102577e-05}
{"train_loss": 0.13206720352172852, "global_step": 109973, "epoch": 1235, "lr": 6.484643256173363e-05}
{"train_loss": 0.14889338612556458, "global_step": 109974, "epoch": 1235, "lr": 6.484587897044556e-05}
{"train_loss": 0.1765274852514267, "global_step": 109975, "epoch": 1235, "lr": 6.484532537716166e-05}
{"train_loss": 0.18552657961845398, "global_step": 109976, "epoch": 1235, "lr": 6.484477178188198e-05}
{"train_loss": 0.1322920024394989, "global_step": 109977, "epoch": 1235, "lr": 6.48442181846066e-05}
{"train_loss": 0.1175154596567154, "global_step": 109978, "epoch": 1235, "lr": 6.484366458533559e-05}
{"train_loss": 0.1608588993549347, "global_step": 109979, "epoch": 1235, "lr": 6.484311098406904e-05}
{"train_loss": 0.15070921182632446, "global_step": 109980, "epoch": 1235, "lr": 6.484255738080701e-05}
{"train_loss": 0.17262284457683563, "global_step": 109981, "epoch": 1235, "lr": 6.484200377554959e-05}
{"train_loss": 0.11976420879364014, "global_step": 109982, "epoch": 1235, "lr": 6.484145016829683e-05}
{"train_loss": 0.07377716898918152, "global_step": 109983, "epoch": 1235, "lr": 6.484089655904881e-05}
{"train_loss": 0.13042204082012177, "global_step": 109984, "epoch": 1235, "lr": 6.484034294780563e-05}
{"train_loss": 0.11503160744905472, "global_step": 109985, "epoch": 1235, "lr": 6.483978933456733e-05}
{"train_loss": 0.12760533392429352, "global_step": 109986, "epoch": 1235, "lr": 6.483923571933402e-05}
{"train_loss": 0.13995768129825592, "global_step": 109987, "epoch": 1235, "lr": 6.483868210210575e-05}
{"train_loss": 0.09816350042819977, "global_step": 109988, "epoch": 1235, "lr": 6.48381284828826e-05}
{"train_loss": 0.10303092002868652, "global_step": 109989, "epoch": 1235, "lr": 6.483757486166463e-05}
{"train_loss": 0.16847772896289825, "global_step": 109990, "epoch": 1235, "lr": 6.483702123845195e-05}
{"train_loss": 0.15454630553722382, "global_step": 109991, "epoch": 1235, "lr": 6.48364676132446e-05}
{"train_loss": 0.15132984519004822, "global_step": 109992, "epoch": 1235, "lr": 6.483591398604268e-05}
{"train_loss": 0.27513155341148376, "global_step": 109993, "epoch": 1235, "lr": 6.483536035684624e-05}
{"train_loss": 0.09307552874088287, "global_step": 109994, "epoch": 1235, "lr": 6.483480672565538e-05}
{"train_loss": 0.0774613618850708, "global_step": 109995, "epoch": 1235, "lr": 6.483425309247014e-05}
{"train_loss": 0.10745490342378616, "global_step": 109996, "epoch": 1235, "lr": 6.483369945729062e-05}
{"train_loss": 0.15710780024528503, "global_step": 109997, "epoch": 1235, "lr": 6.483314582011691e-05}
{"train_loss": 0.17123658955097198, "global_step": 109998, "epoch": 1235, "lr": 6.483259218094904e-05}
{"train_loss": 0.22923702001571655, "global_step": 109999, "epoch": 1235, "lr": 6.483203853978713e-05}
{"train_loss": 0.12378882616758347, "global_step": 110000, "epoch": 1235, "lr": 6.483148489663122e-05}
{"train_loss": 0.23073634505271912, "global_step": 110001, "epoch": 1235, "lr": 6.48309312514814e-05}
{"train_loss": 0.12631531059741974, "global_step": 110002, "epoch": 1235, "lr": 6.483037760433775e-05}
{"train_loss": 0.15026189637987802, "global_step": 110003, "epoch": 1235, "lr": 6.482982395520032e-05, "val_loss": 5.1191229820251465, "train_action_mse_error": 6.527529239654541}
{"train_loss": 0.1742289960384369, "global_step": 110004, "epoch": 1236, "lr": 6.482927030406922e-05}
{"train_loss": 0.18305741250514984, "global_step": 110005, "epoch": 1236, "lr": 6.48287166509445e-05}
{"train_loss": 0.16370469331741333, "global_step": 110006, "epoch": 1236, "lr": 6.482816299582622e-05}
{"train_loss": 0.05284591019153595, "global_step": 110007, "epoch": 1236, "lr": 6.48276093387145e-05}
{"train_loss": 0.17157983779907227, "global_step": 110008, "epoch": 1236, "lr": 6.482705567960937e-05}
{"train_loss": 0.19694997370243073, "global_step": 110009, "epoch": 1236, "lr": 6.482650201851094e-05}
{"train_loss": 0.18070271611213684, "global_step": 110010, "epoch": 1236, "lr": 6.482594835541925e-05}
{"train_loss": 0.1659867763519287, "global_step": 110011, "epoch": 1236, "lr": 6.48253946903344e-05}
{"train_loss": 0.1613645702600479, "global_step": 110012, "epoch": 1236, "lr": 6.482484102325646e-05}
{"train_loss": 0.1302613765001297, "global_step": 110013, "epoch": 1236, "lr": 6.48242873541855e-05}
{"train_loss": 0.17679721117019653, "global_step": 110014, "epoch": 1236, "lr": 6.482373368312161e-05}
{"train_loss": 0.15696316957473755, "global_step": 110015, "epoch": 1236, "lr": 6.482318001006482e-05}
{"train_loss": 0.21264491975307465, "global_step": 110016, "epoch": 1236, "lr": 6.482262633501525e-05}
{"train_loss": 0.12979938089847565, "global_step": 110017, "epoch": 1236, "lr": 6.482207265797295e-05}
{"train_loss": 0.13494355976581573, "global_step": 110018, "epoch": 1236, "lr": 6.482151897893801e-05}
{"train_loss": 0.11870822310447693, "global_step": 110019, "epoch": 1236, "lr": 6.48209652979105e-05}
{"train_loss": 0.17456622421741486, "global_step": 110020, "epoch": 1236, "lr": 6.482041161489048e-05}
{"train_loss": 0.1714434176683426, "global_step": 110021, "epoch": 1236, "lr": 6.481985792987805e-05}
{"train_loss": 0.17878457903862, "global_step": 110022, "epoch": 1236, "lr": 6.481930424287326e-05}
{"train_loss": 0.20869982242584229, "global_step": 110023, "epoch": 1236, "lr": 6.481875055387621e-05}
{"train_loss": 0.08441378176212311, "global_step": 110024, "epoch": 1236, "lr": 6.481819686288694e-05}
{"train_loss": 0.13305409252643585, "global_step": 110025, "epoch": 1236, "lr": 6.481764316990556e-05}
{"train_loss": 0.15058785676956177, "global_step": 110026, "epoch": 1236, "lr": 6.481708947493213e-05}
{"train_loss": 0.11996576189994812, "global_step": 110027, "epoch": 1236, "lr": 6.481653577796671e-05}
{"train_loss": 0.17815780639648438, "global_step": 110028, "epoch": 1236, "lr": 6.481598207900939e-05}
{"train_loss": 0.2326928824186325, "global_step": 110029, "epoch": 1236, "lr": 6.481542837806025e-05}
{"train_loss": 0.17452552914619446, "global_step": 110030, "epoch": 1236, "lr": 6.481487467511934e-05}
{"train_loss": 0.16225776076316833, "global_step": 110031, "epoch": 1236, "lr": 6.481432097018676e-05}
{"train_loss": 0.14185784757137299, "global_step": 110032, "epoch": 1236, "lr": 6.481376726326259e-05}
{"train_loss": 0.1187143623828888, "global_step": 110033, "epoch": 1236, "lr": 6.481321355434686e-05}
{"train_loss": 0.16670414805412292, "global_step": 110034, "epoch": 1236, "lr": 6.48126598434397e-05}
{"train_loss": 0.19926892220973969, "global_step": 110035, "epoch": 1236, "lr": 6.481210613054115e-05}
{"train_loss": 0.14505784213542938, "global_step": 110036, "epoch": 1236, "lr": 6.48115524156513e-05}
{"train_loss": 0.09678080677986145, "global_step": 110037, "epoch": 1236, "lr": 6.481099869877019e-05}
{"train_loss": 0.18313553929328918, "global_step": 110038, "epoch": 1236, "lr": 6.481044497989794e-05}
{"train_loss": 0.0941360667347908, "global_step": 110039, "epoch": 1236, "lr": 6.480989125903461e-05}
{"train_loss": 0.1641891449689865, "global_step": 110040, "epoch": 1236, "lr": 6.480933753618028e-05}
{"train_loss": 0.16953802108764648, "global_step": 110041, "epoch": 1236, "lr": 6.480878381133499e-05}
{"train_loss": 0.13955284655094147, "global_step": 110042, "epoch": 1236, "lr": 6.480823008449885e-05}
{"train_loss": 0.10561211407184601, "global_step": 110043, "epoch": 1236, "lr": 6.480767635567193e-05}
{"train_loss": 0.13986797630786896, "global_step": 110044, "epoch": 1236, "lr": 6.480712262485429e-05}
{"train_loss": 0.17896877229213715, "global_step": 110045, "epoch": 1236, "lr": 6.480656889204601e-05}
{"train_loss": 0.1739470213651657, "global_step": 110046, "epoch": 1236, "lr": 6.480601515724718e-05}
{"train_loss": 0.11729267984628677, "global_step": 110047, "epoch": 1236, "lr": 6.480546142045786e-05}
{"train_loss": 0.12683100998401642, "global_step": 110048, "epoch": 1236, "lr": 6.480490768167812e-05}
{"train_loss": 0.11740635335445404, "global_step": 110049, "epoch": 1236, "lr": 6.480435394090805e-05}
{"train_loss": 0.10585346817970276, "global_step": 110050, "epoch": 1236, "lr": 6.48038001981477e-05}
{"train_loss": 0.1464017778635025, "global_step": 110051, "epoch": 1236, "lr": 6.480324645339717e-05}
{"train_loss": 0.08097472041845322, "global_step": 110052, "epoch": 1236, "lr": 6.480269270665651e-05}
{"train_loss": 0.08725601434707642, "global_step": 110053, "epoch": 1236, "lr": 6.480213895792583e-05}
{"train_loss": 0.1521725207567215, "global_step": 110054, "epoch": 1236, "lr": 6.480158520720517e-05}
{"train_loss": 0.1634284406900406, "global_step": 110055, "epoch": 1236, "lr": 6.480103145449463e-05}
{"train_loss": 0.07400599122047424, "global_step": 110056, "epoch": 1236, "lr": 6.480047769979425e-05}
{"train_loss": 0.10083994269371033, "global_step": 110057, "epoch": 1236, "lr": 6.479992394310414e-05}
{"train_loss": 0.16971710324287415, "global_step": 110058, "epoch": 1236, "lr": 6.479937018442435e-05}
{"train_loss": 0.15747776627540588, "global_step": 110059, "epoch": 1236, "lr": 6.479881642375499e-05}
{"train_loss": 0.07714726775884628, "global_step": 110060, "epoch": 1236, "lr": 6.479826266109609e-05}
{"train_loss": 0.11990363150835037, "global_step": 110061, "epoch": 1236, "lr": 6.479770889644774e-05}
{"train_loss": 0.10390789061784744, "global_step": 110062, "epoch": 1236, "lr": 6.479715512981002e-05}
{"train_loss": 0.112861767411232, "global_step": 110063, "epoch": 1236, "lr": 6.479660136118301e-05}
{"train_loss": 0.10941187292337418, "global_step": 110064, "epoch": 1236, "lr": 6.479604759056679e-05}
{"train_loss": 0.21639712154865265, "global_step": 110065, "epoch": 1236, "lr": 6.47954938179614e-05}
{"train_loss": 0.18286705017089844, "global_step": 110066, "epoch": 1236, "lr": 6.479494004336695e-05}
{"train_loss": 0.15559527277946472, "global_step": 110067, "epoch": 1236, "lr": 6.47943862667835e-05}
{"train_loss": 0.16039679944515228, "global_step": 110068, "epoch": 1236, "lr": 6.47938324882111e-05}
{"train_loss": 0.21217046678066254, "global_step": 110069, "epoch": 1236, "lr": 6.479327870764987e-05}
{"train_loss": 0.14197993278503418, "global_step": 110070, "epoch": 1236, "lr": 6.479272492509987e-05}
{"train_loss": 0.0797462984919548, "global_step": 110071, "epoch": 1236, "lr": 6.479217114056115e-05}
{"train_loss": 0.06926431506872177, "global_step": 110072, "epoch": 1236, "lr": 6.479161735403382e-05}
{"train_loss": 0.11432720720767975, "global_step": 110073, "epoch": 1236, "lr": 6.479106356551793e-05}
{"train_loss": 0.14842620491981506, "global_step": 110074, "epoch": 1236, "lr": 6.479050977501356e-05}
{"train_loss": 0.14229658246040344, "global_step": 110075, "epoch": 1236, "lr": 6.478995598252077e-05}
{"train_loss": 0.17389772832393646, "global_step": 110076, "epoch": 1236, "lr": 6.478940218803969e-05}
{"train_loss": 0.2834920883178711, "global_step": 110077, "epoch": 1236, "lr": 6.478884839157032e-05}
{"train_loss": 0.18855494260787964, "global_step": 110078, "epoch": 1236, "lr": 6.478829459311279e-05}
{"train_loss": 0.14509694278240204, "global_step": 110079, "epoch": 1236, "lr": 6.478774079266716e-05}
{"train_loss": 0.2172272801399231, "global_step": 110080, "epoch": 1236, "lr": 6.478718699023348e-05}
{"train_loss": 0.12852990627288818, "global_step": 110081, "epoch": 1236, "lr": 6.478663318581186e-05}
{"train_loss": 0.0929800420999527, "global_step": 110082, "epoch": 1236, "lr": 6.478607937940234e-05}
{"train_loss": 0.11627762019634247, "global_step": 110083, "epoch": 1236, "lr": 6.478552557100502e-05}
{"train_loss": 0.1629994809627533, "global_step": 110084, "epoch": 1236, "lr": 6.478497176061998e-05}
{"train_loss": 0.17607763409614563, "global_step": 110085, "epoch": 1236, "lr": 6.478441794824725e-05}
{"train_loss": 0.1358884572982788, "global_step": 110086, "epoch": 1236, "lr": 6.478386413388696e-05}
{"train_loss": 0.14963826537132263, "global_step": 110087, "epoch": 1236, "lr": 6.478331031753914e-05}
{"train_loss": 0.146908700466156, "global_step": 110088, "epoch": 1236, "lr": 6.47827564992039e-05}
{"train_loss": 0.13112443685531616, "global_step": 110089, "epoch": 1236, "lr": 6.47822026788813e-05}
{"train_loss": 0.18659430742263794, "global_step": 110090, "epoch": 1236, "lr": 6.478164885657141e-05}
{"train_loss": 0.1796390414237976, "global_step": 110091, "epoch": 1236, "lr": 6.47810950322743e-05}
{"train_loss": 0.1480751739124234, "global_step": 110092, "epoch": 1236, "lr": 6.478054120599006e-05, "val_loss": 5.211056709289551}
{"train_loss": 0.15757286548614502, "global_step": 110093, "epoch": 1237, "lr": 6.477998737771873e-05}
{"train_loss": 0.2047358900308609, "global_step": 110094, "epoch": 1237, "lr": 6.477943354746044e-05}
{"train_loss": 0.18615593016147614, "global_step": 110095, "epoch": 1237, "lr": 6.477887971521522e-05}
{"train_loss": 0.1103987842798233, "global_step": 110096, "epoch": 1237, "lr": 6.477832588098318e-05}
{"train_loss": 0.16041557490825653, "global_step": 110097, "epoch": 1237, "lr": 6.477777204476436e-05}
{"train_loss": 0.15280349552631378, "global_step": 110098, "epoch": 1237, "lr": 6.477721820655887e-05}
{"train_loss": 0.10022778064012527, "global_step": 110099, "epoch": 1237, "lr": 6.477666436636673e-05}
{"train_loss": 0.13914835453033447, "global_step": 110100, "epoch": 1237, "lr": 6.477611052418806e-05}
{"train_loss": 0.11901302635669708, "global_step": 110101, "epoch": 1237, "lr": 6.477555668002294e-05}
{"train_loss": 0.10516392439603806, "global_step": 110102, "epoch": 1237, "lr": 6.47750028338714e-05}
{"train_loss": 0.17249396443367004, "global_step": 110103, "epoch": 1237, "lr": 6.477444898573355e-05}
{"train_loss": 0.16548407077789307, "global_step": 110104, "epoch": 1237, "lr": 6.477389513560945e-05}
{"train_loss": 0.1773405373096466, "global_step": 110105, "epoch": 1237, "lr": 6.47733412834992e-05}
{"train_loss": 0.18625950813293457, "global_step": 110106, "epoch": 1237, "lr": 6.477278742940283e-05}
{"train_loss": 0.09896289557218552, "global_step": 110107, "epoch": 1237, "lr": 6.477223357332046e-05}
{"train_loss": 0.2285115122795105, "global_step": 110108, "epoch": 1237, "lr": 6.477167971525214e-05}
{"train_loss": 0.18113820254802704, "global_step": 110109, "epoch": 1237, "lr": 6.477112585519793e-05}
{"train_loss": 0.14357097446918488, "global_step": 110110, "epoch": 1237, "lr": 6.477057199315794e-05}
{"train_loss": 0.11219310760498047, "global_step": 110111, "epoch": 1237, "lr": 6.47700181291322e-05}
{"train_loss": 0.17728303372859955, "global_step": 110112, "epoch": 1237, "lr": 6.476946426312083e-05}
{"train_loss": 0.1294587105512619, "global_step": 110113, "epoch": 1237, "lr": 6.47689103951239e-05}
{"train_loss": 0.10244058072566986, "global_step": 110114, "epoch": 1237, "lr": 6.476835652514145e-05}
{"train_loss": 0.13790877163410187, "global_step": 110115, "epoch": 1237, "lr": 6.476780265317357e-05}
{"train_loss": 0.168249249458313, "global_step": 110116, "epoch": 1237, "lr": 6.476724877922035e-05}
{"train_loss": 0.1325186789035797, "global_step": 110117, "epoch": 1237, "lr": 6.476669490328187e-05}
{"train_loss": 0.15539993345737457, "global_step": 110118, "epoch": 1237, "lr": 6.476614102535818e-05}
{"train_loss": 0.15297779440879822, "global_step": 110119, "epoch": 1237, "lr": 6.476558714544935e-05}
{"train_loss": 0.11870236694812775, "global_step": 110120, "epoch": 1237, "lr": 6.476503326355546e-05}
{"train_loss": 0.137968972325325, "global_step": 110121, "epoch": 1237, "lr": 6.47644793796766e-05}
{"train_loss": 0.17270714044570923, "global_step": 110122, "epoch": 1237, "lr": 6.476392549381284e-05}
{"train_loss": 0.14745451509952545, "global_step": 110123, "epoch": 1237, "lr": 6.476337160596425e-05}
{"train_loss": 0.12996628880500793, "global_step": 110124, "epoch": 1237, "lr": 6.476281771613089e-05}
{"train_loss": 0.13551700115203857, "global_step": 110125, "epoch": 1237, "lr": 6.476226382431287e-05}
{"train_loss": 0.09737230837345123, "global_step": 110126, "epoch": 1237, "lr": 6.476170993051022e-05}
{"train_loss": 0.09127848595380783, "global_step": 110127, "epoch": 1237, "lr": 6.476115603472306e-05}
{"train_loss": 0.17080800235271454, "global_step": 110128, "epoch": 1237, "lr": 6.476060213695145e-05}
{"train_loss": 0.11231010407209396, "global_step": 110129, "epoch": 1237, "lr": 6.476004823719543e-05}
{"train_loss": 0.1616179496049881, "global_step": 110130, "epoch": 1237, "lr": 6.47594943354551e-05}
{"train_loss": 0.1055387482047081, "global_step": 110131, "epoch": 1237, "lr": 6.475894043173056e-05}
{"train_loss": 0.14180825650691986, "global_step": 110132, "epoch": 1237, "lr": 6.475838652602185e-05}
{"train_loss": 0.17138482630252838, "global_step": 110133, "epoch": 1237, "lr": 6.475783261832905e-05}
{"train_loss": 0.10909689217805862, "global_step": 110134, "epoch": 1237, "lr": 6.475727870865224e-05}
{"train_loss": 0.23640809953212738, "global_step": 110135, "epoch": 1237, "lr": 6.47567247969915e-05}
{"train_loss": 0.16797344386577606, "global_step": 110136, "epoch": 1237, "lr": 6.475617088334692e-05}
{"train_loss": 0.12950310111045837, "global_step": 110137, "epoch": 1237, "lr": 6.475561696771852e-05}
{"train_loss": 0.1585482805967331, "global_step": 110138, "epoch": 1237, "lr": 6.475506305010641e-05}
{"train_loss": 0.06515906751155853, "global_step": 110139, "epoch": 1237, "lr": 6.475450913051067e-05}
{"train_loss": 0.0726035088300705, "global_step": 110140, "epoch": 1237, "lr": 6.475395520893136e-05}
{"train_loss": 0.16381430625915527, "global_step": 110141, "epoch": 1237, "lr": 6.475340128536856e-05}
{"train_loss": 0.09351962059736252, "global_step": 110142, "epoch": 1237, "lr": 6.475284735982236e-05}
{"train_loss": 0.1718941628932953, "global_step": 110143, "epoch": 1237, "lr": 6.475229343229279e-05}
{"train_loss": 0.13285773992538452, "global_step": 110144, "epoch": 1237, "lr": 6.475173950277998e-05}
{"train_loss": 0.20646092295646667, "global_step": 110145, "epoch": 1237, "lr": 6.475118557128397e-05}
{"train_loss": 0.11778996884822845, "global_step": 110146, "epoch": 1237, "lr": 6.475063163780485e-05}
{"train_loss": 0.18854667246341705, "global_step": 110147, "epoch": 1237, "lr": 6.475007770234268e-05}
{"train_loss": 0.15700004994869232, "global_step": 110148, "epoch": 1237, "lr": 6.474952376489753e-05}
{"train_loss": 0.14003050327301025, "global_step": 110149, "epoch": 1237, "lr": 6.47489698254695e-05}
{"train_loss": 0.13082267343997955, "global_step": 110150, "epoch": 1237, "lr": 6.474841588405865e-05}
{"train_loss": 0.14807787537574768, "global_step": 110151, "epoch": 1237, "lr": 6.474786194066505e-05}
{"train_loss": 0.1131318211555481, "global_step": 110152, "epoch": 1237, "lr": 6.474730799528879e-05}
{"train_loss": 0.13214221596717834, "global_step": 110153, "epoch": 1237, "lr": 6.474675404792992e-05}
{"train_loss": 0.14113746583461761, "global_step": 110154, "epoch": 1237, "lr": 6.474620009858854e-05}
{"train_loss": 0.12448760122060776, "global_step": 110155, "epoch": 1237, "lr": 6.474564614726473e-05}
{"train_loss": 0.18826282024383545, "global_step": 110156, "epoch": 1237, "lr": 6.474509219395853e-05}
{"train_loss": 0.13395826518535614, "global_step": 110157, "epoch": 1237, "lr": 6.474453823867001e-05}
{"train_loss": 0.17160145938396454, "global_step": 110158, "epoch": 1237, "lr": 6.474398428139929e-05}
{"train_loss": 0.1436711698770523, "global_step": 110159, "epoch": 1237, "lr": 6.474343032214642e-05}
{"train_loss": 0.2115740031003952, "global_step": 110160, "epoch": 1237, "lr": 6.474287636091148e-05}
{"train_loss": 0.19311898946762085, "global_step": 110161, "epoch": 1237, "lr": 6.474232239769453e-05}
{"train_loss": 0.15570038557052612, "global_step": 110162, "epoch": 1237, "lr": 6.474176843249566e-05}
{"train_loss": 0.18147888779640198, "global_step": 110163, "epoch": 1237, "lr": 6.474121446531494e-05}
{"train_loss": 0.12535321712493896, "global_step": 110164, "epoch": 1237, "lr": 6.474066049615243e-05}
{"train_loss": 0.25509771704673767, "global_step": 110165, "epoch": 1237, "lr": 6.474010652500824e-05}
{"train_loss": 0.26716384291648865, "global_step": 110166, "epoch": 1237, "lr": 6.47395525518824e-05}
{"train_loss": 0.12111228704452515, "global_step": 110167, "epoch": 1237, "lr": 6.473899857677501e-05}
{"train_loss": 0.14023910462856293, "global_step": 110168, "epoch": 1237, "lr": 6.473844459968614e-05}
{"train_loss": 0.15952861309051514, "global_step": 110169, "epoch": 1237, "lr": 6.473789062061589e-05}
{"train_loss": 0.17907103896141052, "global_step": 110170, "epoch": 1237, "lr": 6.473733663956429e-05}
{"train_loss": 0.16185884177684784, "global_step": 110171, "epoch": 1237, "lr": 6.473678265653143e-05}
{"train_loss": 0.1823335587978363, "global_step": 110172, "epoch": 1237, "lr": 6.47362286715174e-05}
{"train_loss": 0.13751356303691864, "global_step": 110173, "epoch": 1237, "lr": 6.473567468452226e-05}
{"train_loss": 0.20747752487659454, "global_step": 110174, "epoch": 1237, "lr": 6.47351206955461e-05}
{"train_loss": 0.1632923185825348, "global_step": 110175, "epoch": 1237, "lr": 6.473456670458897e-05}
{"train_loss": 0.12968061864376068, "global_step": 110176, "epoch": 1237, "lr": 6.473401271165095e-05}
{"train_loss": 0.13882333040237427, "global_step": 110177, "epoch": 1237, "lr": 6.473345871673215e-05}
{"train_loss": 0.1831514537334442, "global_step": 110178, "epoch": 1237, "lr": 6.47329047198326e-05}
{"train_loss": 0.11085329949855804, "global_step": 110179, "epoch": 1237, "lr": 6.473235072095238e-05}
{"train_loss": 0.12100096791982651, "global_step": 110180, "epoch": 1237, "lr": 6.473179672009158e-05}
{"train_loss": 0.14975135522277166, "global_step": 110181, "epoch": 1237, "lr": 6.473124271725028e-05, "val_loss": 5.069899559020996}
{"train_loss": 0.11570841073989868, "global_step": 110182, "epoch": 1238, "lr": 6.473068871242853e-05}
{"train_loss": 0.12436050921678543, "global_step": 110183, "epoch": 1238, "lr": 6.473013470562643e-05}
{"train_loss": 0.142655611038208, "global_step": 110184, "epoch": 1238, "lr": 6.472958069684403e-05}
{"train_loss": 0.16690008342266083, "global_step": 110185, "epoch": 1238, "lr": 6.472902668608141e-05}
{"train_loss": 0.15495313704013824, "global_step": 110186, "epoch": 1238, "lr": 6.472847267333868e-05}
{"train_loss": 0.2843933701515198, "global_step": 110187, "epoch": 1238, "lr": 6.472791865861587e-05}
{"train_loss": 0.13227249681949615, "global_step": 110188, "epoch": 1238, "lr": 6.472736464191306e-05}
{"train_loss": 0.16816484928131104, "global_step": 110189, "epoch": 1238, "lr": 6.472681062323035e-05}
{"train_loss": 0.15200725197792053, "global_step": 110190, "epoch": 1238, "lr": 6.472625660256781e-05}
{"train_loss": 0.1148800179362297, "global_step": 110191, "epoch": 1238, "lr": 6.47257025799255e-05}
{"train_loss": 0.11445339024066925, "global_step": 110192, "epoch": 1238, "lr": 6.472514855530349e-05}
{"train_loss": 0.1475229561328888, "global_step": 110193, "epoch": 1238, "lr": 6.472459452870186e-05}
{"train_loss": 0.16095265746116638, "global_step": 110194, "epoch": 1238, "lr": 6.472404050012071e-05}
{"train_loss": 0.1113470047712326, "global_step": 110195, "epoch": 1238, "lr": 6.472348646956008e-05}
{"train_loss": 0.13394564390182495, "global_step": 110196, "epoch": 1238, "lr": 6.472293243702005e-05}
{"train_loss": 0.18332520127296448, "global_step": 110197, "epoch": 1238, "lr": 6.472237840250071e-05}
{"train_loss": 0.23169340193271637, "global_step": 110198, "epoch": 1238, "lr": 6.472182436600212e-05}
{"train_loss": 0.14475686848163605, "global_step": 110199, "epoch": 1238, "lr": 6.472127032752436e-05}
{"train_loss": 0.14375197887420654, "global_step": 110200, "epoch": 1238, "lr": 6.472071628706748e-05}
{"train_loss": 0.1547452211380005, "global_step": 110201, "epoch": 1238, "lr": 6.47201622446316e-05}
{"train_loss": 0.1287211775779724, "global_step": 110202, "epoch": 1238, "lr": 6.471960820021678e-05}
{"train_loss": 0.1719280183315277, "global_step": 110203, "epoch": 1238, "lr": 6.471905415382308e-05}
{"train_loss": 0.14235949516296387, "global_step": 110204, "epoch": 1238, "lr": 6.47185001054506e-05}
{"train_loss": 0.07999682426452637, "global_step": 110205, "epoch": 1238, "lr": 6.471794605509938e-05}
{"train_loss": 0.1275937557220459, "global_step": 110206, "epoch": 1238, "lr": 6.471739200276952e-05}
{"train_loss": 0.1356893628835678, "global_step": 110207, "epoch": 1238, "lr": 6.471683794846107e-05}
{"train_loss": 0.1780601441860199, "global_step": 110208, "epoch": 1238, "lr": 6.471628389217414e-05}
{"train_loss": 0.19243913888931274, "global_step": 110209, "epoch": 1238, "lr": 6.471572983390877e-05}
{"train_loss": 0.16324006021022797, "global_step": 110210, "epoch": 1238, "lr": 6.471517577366507e-05}
{"train_loss": 0.14893651008605957, "global_step": 110211, "epoch": 1238, "lr": 6.471462171144307e-05}
{"train_loss": 0.12135373055934906, "global_step": 110212, "epoch": 1238, "lr": 6.471406764724289e-05}
{"train_loss": 0.13235822319984436, "global_step": 110213, "epoch": 1238, "lr": 6.471351358106457e-05}
{"train_loss": 0.23401717841625214, "global_step": 110214, "epoch": 1238, "lr": 6.47129595129082e-05}
{"train_loss": 0.1666197031736374, "global_step": 110215, "epoch": 1238, "lr": 6.471240544277386e-05}
{"train_loss": 0.1664329618215561, "global_step": 110216, "epoch": 1238, "lr": 6.47118513706616e-05}
{"train_loss": 0.16139596700668335, "global_step": 110217, "epoch": 1238, "lr": 6.471129729657153e-05}
{"train_loss": 0.18435779213905334, "global_step": 110218, "epoch": 1238, "lr": 6.471074322050368e-05}
{"train_loss": 0.1299096643924713, "global_step": 110219, "epoch": 1238, "lr": 6.471018914245816e-05}
{"train_loss": 0.10508269816637039, "global_step": 110220, "epoch": 1238, "lr": 6.470963506243503e-05}
{"train_loss": 0.09929786622524261, "global_step": 110221, "epoch": 1238, "lr": 6.470908098043437e-05}
{"train_loss": 0.09200676530599594, "global_step": 110222, "epoch": 1238, "lr": 6.470852689645628e-05}
{"train_loss": 0.12503309547901154, "global_step": 110223, "epoch": 1238, "lr": 6.470797281050078e-05}
{"train_loss": 0.18239444494247437, "global_step": 110224, "epoch": 1238, "lr": 6.470741872256799e-05}
{"train_loss": 0.11050095409154892, "global_step": 110225, "epoch": 1238, "lr": 6.470686463265795e-05}
{"train_loss": 0.2037852704524994, "global_step": 110226, "epoch": 1238, "lr": 6.470631054077076e-05}
{"train_loss": 0.15792958438396454, "global_step": 110227, "epoch": 1238, "lr": 6.470575644690648e-05}
{"train_loss": 0.10398948192596436, "global_step": 110228, "epoch": 1238, "lr": 6.47052023510652e-05}
{"train_loss": 0.11766326427459717, "global_step": 110229, "epoch": 1238, "lr": 6.470464825324698e-05}
{"train_loss": 0.16560545563697815, "global_step": 110230, "epoch": 1238, "lr": 6.47040941534519e-05}
{"train_loss": 0.15811821818351746, "global_step": 110231, "epoch": 1238, "lr": 6.470354005168003e-05}
{"train_loss": 0.17442743480205536, "global_step": 110232, "epoch": 1238, "lr": 6.470298594793144e-05}
{"train_loss": 0.1427890956401825, "global_step": 110233, "epoch": 1238, "lr": 6.470243184220625e-05}
{"train_loss": 0.12291912734508514, "global_step": 110234, "epoch": 1238, "lr": 6.470187773450447e-05}
{"train_loss": 0.1736549586057663, "global_step": 110235, "epoch": 1238, "lr": 6.47013236248262e-05}
{"train_loss": 0.12314676493406296, "global_step": 110236, "epoch": 1238, "lr": 6.470076951317153e-05}
{"train_loss": 0.14722475409507751, "global_step": 110237, "epoch": 1238, "lr": 6.47002153995405e-05}
{"train_loss": 0.1312759816646576, "global_step": 110238, "epoch": 1238, "lr": 6.46996612839332e-05}
{"train_loss": 0.1347409039735794, "global_step": 110239, "epoch": 1238, "lr": 6.469910716634973e-05}
{"train_loss": 0.19298210740089417, "global_step": 110240, "epoch": 1238, "lr": 6.469855304679014e-05}
{"train_loss": 0.11574308574199677, "global_step": 110241, "epoch": 1238, "lr": 6.46979989252545e-05}
{"train_loss": 0.10774310678243637, "global_step": 110242, "epoch": 1238, "lr": 6.46974448017429e-05}
{"train_loss": 0.1787828505039215, "global_step": 110243, "epoch": 1238, "lr": 6.469689067625541e-05}
{"train_loss": 0.14620675146579742, "global_step": 110244, "epoch": 1238, "lr": 6.469633654879211e-05}
{"train_loss": 0.14016622304916382, "global_step": 110245, "epoch": 1238, "lr": 6.469578241935305e-05}
{"train_loss": 0.15221357345581055, "global_step": 110246, "epoch": 1238, "lr": 6.469522828793832e-05}
{"train_loss": 0.2172173261642456, "global_step": 110247, "epoch": 1238, "lr": 6.469467415454802e-05}
{"train_loss": 0.1752891093492508, "global_step": 110248, "epoch": 1238, "lr": 6.469412001918217e-05}
{"train_loss": 0.10138168931007385, "global_step": 110249, "epoch": 1238, "lr": 6.46935658818409e-05}
{"train_loss": 0.27071791887283325, "global_step": 110250, "epoch": 1238, "lr": 6.469301174252424e-05}
{"train_loss": 0.2049071341753006, "global_step": 110251, "epoch": 1238, "lr": 6.46924576012323e-05}
{"train_loss": 0.1561596840620041, "global_step": 110252, "epoch": 1238, "lr": 6.469190345796515e-05}
{"train_loss": 0.1739547848701477, "global_step": 110253, "epoch": 1238, "lr": 6.469134931272281e-05}
{"train_loss": 0.12261100858449936, "global_step": 110254, "epoch": 1238, "lr": 6.469079516550543e-05}
{"train_loss": 0.22606632113456726, "global_step": 110255, "epoch": 1238, "lr": 6.469024101631303e-05}
{"train_loss": 0.15076807141304016, "global_step": 110256, "epoch": 1238, "lr": 6.468968686514572e-05}
{"train_loss": 0.09695012122392654, "global_step": 110257, "epoch": 1238, "lr": 6.468913271200355e-05}
{"train_loss": 0.098703533411026, "global_step": 110258, "epoch": 1238, "lr": 6.468857855688662e-05}
{"train_loss": 0.13743478059768677, "global_step": 110259, "epoch": 1238, "lr": 6.468802439979498e-05}
{"train_loss": 0.11594164371490479, "global_step": 110260, "epoch": 1238, "lr": 6.468747024072872e-05}
{"train_loss": 0.15843448042869568, "global_step": 110261, "epoch": 1238, "lr": 6.46869160796879e-05}
{"train_loss": 0.20776504278182983, "global_step": 110262, "epoch": 1238, "lr": 6.468636191667261e-05}
{"train_loss": 0.19589880108833313, "global_step": 110263, "epoch": 1238, "lr": 6.468580775168292e-05}
{"train_loss": 0.1653933972120285, "global_step": 110264, "epoch": 1238, "lr": 6.468525358471889e-05}
{"train_loss": 0.16629548370838165, "global_step": 110265, "epoch": 1238, "lr": 6.468469941578061e-05}
{"train_loss": 0.13862504065036774, "global_step": 110266, "epoch": 1238, "lr": 6.468414524486817e-05}
{"train_loss": 0.18419510126113892, "global_step": 110267, "epoch": 1238, "lr": 6.46835910719816e-05}
{"train_loss": 0.07027995586395264, "global_step": 110268, "epoch": 1238, "lr": 6.468303689712102e-05}
{"train_loss": 0.13030892610549927, "global_step": 110269, "epoch": 1238, "lr": 6.468248272028647e-05}
{"train_loss": 0.1518773442406333, "global_step": 110270, "epoch": 1238, "lr": 6.468192854147805e-05, "val_loss": 5.0566325187683105}
{"train_loss": 0.16979682445526123, "global_step": 110271, "epoch": 1239, "lr": 6.468137436069583e-05}
{"train_loss": 0.18382319808006287, "global_step": 110272, "epoch": 1239, "lr": 6.468082017793987e-05}
{"train_loss": 0.22044500708580017, "global_step": 110273, "epoch": 1239, "lr": 6.468026599321026e-05}
{"train_loss": 0.16273315250873566, "global_step": 110274, "epoch": 1239, "lr": 6.467971180650704e-05}
{"train_loss": 0.136186420917511, "global_step": 110275, "epoch": 1239, "lr": 6.467915761783032e-05}
{"train_loss": 0.14561530947685242, "global_step": 110276, "epoch": 1239, "lr": 6.467860342718018e-05}
{"train_loss": 0.13543759286403656, "global_step": 110277, "epoch": 1239, "lr": 6.467804923455668e-05}
{"train_loss": 0.189446359872818, "global_step": 110278, "epoch": 1239, "lr": 6.467749503995989e-05}
{"train_loss": 0.1315009742975235, "global_step": 110279, "epoch": 1239, "lr": 6.467694084338989e-05}
{"train_loss": 0.13320371508598328, "global_step": 110280, "epoch": 1239, "lr": 6.467638664484676e-05}
{"train_loss": 0.17792049050331116, "global_step": 110281, "epoch": 1239, "lr": 6.467583244433056e-05}
{"train_loss": 0.1793670654296875, "global_step": 110282, "epoch": 1239, "lr": 6.467527824184138e-05}
{"train_loss": 0.11609557271003723, "global_step": 110283, "epoch": 1239, "lr": 6.467472403737928e-05}
{"train_loss": 0.15560775995254517, "global_step": 110284, "epoch": 1239, "lr": 6.467416983094434e-05}
{"train_loss": 0.16435495018959045, "global_step": 110285, "epoch": 1239, "lr": 6.467361562253665e-05}
{"train_loss": 0.21558472514152527, "global_step": 110286, "epoch": 1239, "lr": 6.467306141215627e-05}
{"train_loss": 0.17095252871513367, "global_step": 110287, "epoch": 1239, "lr": 6.467250719980327e-05}
{"train_loss": 0.1380937099456787, "global_step": 110288, "epoch": 1239, "lr": 6.467195298547773e-05}
{"train_loss": 0.1558876484632492, "global_step": 110289, "epoch": 1239, "lr": 6.467139876917972e-05}
{"train_loss": 0.16911861300468445, "global_step": 110290, "epoch": 1239, "lr": 6.467084455090933e-05}
{"train_loss": 0.1422794610261917, "global_step": 110291, "epoch": 1239, "lr": 6.467029033066663e-05}
{"train_loss": 0.13053081929683685, "global_step": 110292, "epoch": 1239, "lr": 6.466973610845167e-05}
{"train_loss": 0.19172315299510956, "global_step": 110293, "epoch": 1239, "lr": 6.466918188426454e-05}
{"train_loss": 0.12406960129737854, "global_step": 110294, "epoch": 1239, "lr": 6.466862765810532e-05}
{"train_loss": 0.14611782133579254, "global_step": 110295, "epoch": 1239, "lr": 6.46680734299741e-05}
{"train_loss": 0.13326828181743622, "global_step": 110296, "epoch": 1239, "lr": 6.466751919987093e-05}
{"train_loss": 0.10990744084119797, "global_step": 110297, "epoch": 1239, "lr": 6.466696496779589e-05}
{"train_loss": 0.24587637186050415, "global_step": 110298, "epoch": 1239, "lr": 6.466641073374904e-05}
{"train_loss": 0.2674292027950287, "global_step": 110299, "epoch": 1239, "lr": 6.466585649773048e-05}
{"train_loss": 0.11981290578842163, "global_step": 110300, "epoch": 1239, "lr": 6.466530225974027e-05}
{"train_loss": 0.17960025370121002, "global_step": 110301, "epoch": 1239, "lr": 6.466474801977851e-05}
{"train_loss": 0.08572500944137573, "global_step": 110302, "epoch": 1239, "lr": 6.466419377784523e-05}
{"train_loss": 0.1690518856048584, "global_step": 110303, "epoch": 1239, "lr": 6.466363953394054e-05}
{"train_loss": 0.16597923636436462, "global_step": 110304, "epoch": 1239, "lr": 6.46630852880645e-05}
{"train_loss": 0.1744881421327591, "global_step": 110305, "epoch": 1239, "lr": 6.466253104021718e-05}
{"train_loss": 0.08847936987876892, "global_step": 110306, "epoch": 1239, "lr": 6.466197679039867e-05}
{"train_loss": 0.18045271933078766, "global_step": 110307, "epoch": 1239, "lr": 6.466142253860902e-05}
{"train_loss": 0.16071859002113342, "global_step": 110308, "epoch": 1239, "lr": 6.466086828484834e-05}
{"train_loss": 0.17455340921878815, "global_step": 110309, "epoch": 1239, "lr": 6.466031402911667e-05}
{"train_loss": 0.1787298619747162, "global_step": 110310, "epoch": 1239, "lr": 6.465975977141412e-05}
{"train_loss": 0.10747133195400238, "global_step": 110311, "epoch": 1239, "lr": 6.465920551174073e-05}
{"train_loss": 0.1868198812007904, "global_step": 110312, "epoch": 1239, "lr": 6.465865125009658e-05}
{"train_loss": 0.12828503549098969, "global_step": 110313, "epoch": 1239, "lr": 6.465809698648177e-05}
{"train_loss": 0.15735679864883423, "global_step": 110314, "epoch": 1239, "lr": 6.465754272089634e-05}
{"train_loss": 0.19958865642547607, "global_step": 110315, "epoch": 1239, "lr": 6.465698845334039e-05}
{"train_loss": 0.1481776237487793, "global_step": 110316, "epoch": 1239, "lr": 6.4656434183814e-05}
{"train_loss": 0.12411066144704819, "global_step": 110317, "epoch": 1239, "lr": 6.465587991231721e-05}
{"train_loss": 0.14980781078338623, "global_step": 110318, "epoch": 1239, "lr": 6.465532563885012e-05}
{"train_loss": 0.17681914567947388, "global_step": 110319, "epoch": 1239, "lr": 6.465477136341281e-05}
{"train_loss": 0.15188245475292206, "global_step": 110320, "epoch": 1239, "lr": 6.465421708600535e-05}
{"train_loss": 0.14835311472415924, "global_step": 110321, "epoch": 1239, "lr": 6.465366280662779e-05}
{"train_loss": 0.2160160392522812, "global_step": 110322, "epoch": 1239, "lr": 6.465310852528024e-05}
{"train_loss": 0.21712669730186462, "global_step": 110323, "epoch": 1239, "lr": 6.465255424196275e-05}
{"train_loss": 0.11873181909322739, "global_step": 110324, "epoch": 1239, "lr": 6.465199995667541e-05}
{"train_loss": 0.1187719851732254, "global_step": 110325, "epoch": 1239, "lr": 6.465144566941828e-05}
{"train_loss": 0.17198540270328522, "global_step": 110326, "epoch": 1239, "lr": 6.465089138019143e-05}
{"train_loss": 0.15593132376670837, "global_step": 110327, "epoch": 1239, "lr": 6.465033708899495e-05}
{"train_loss": 0.20411597192287445, "global_step": 110328, "epoch": 1239, "lr": 6.464978279582894e-05}
{"train_loss": 0.12226573377847672, "global_step": 110329, "epoch": 1239, "lr": 6.464922850069342e-05}
{"train_loss": 0.12685301899909973, "global_step": 110330, "epoch": 1239, "lr": 6.464867420358849e-05}
{"train_loss": 0.0894102230668068, "global_step": 110331, "epoch": 1239, "lr": 6.464811990451422e-05}
{"train_loss": 0.12609045207500458, "global_step": 110332, "epoch": 1239, "lr": 6.464756560347071e-05}
{"train_loss": 0.12326022982597351, "global_step": 110333, "epoch": 1239, "lr": 6.4647011300458e-05}
{"train_loss": 0.20931214094161987, "global_step": 110334, "epoch": 1239, "lr": 6.464645699547617e-05}
{"train_loss": 0.07453705370426178, "global_step": 110335, "epoch": 1239, "lr": 6.464590268852532e-05}
{"train_loss": 0.14300359785556793, "global_step": 110336, "epoch": 1239, "lr": 6.46453483796055e-05}
{"train_loss": 0.1070307120680809, "global_step": 110337, "epoch": 1239, "lr": 6.464479406871679e-05}
{"train_loss": 0.09583902359008789, "global_step": 110338, "epoch": 1239, "lr": 6.464423975585926e-05}
{"train_loss": 0.16649003326892853, "global_step": 110339, "epoch": 1239, "lr": 6.464368544103299e-05}
{"train_loss": 0.2183954119682312, "global_step": 110340, "epoch": 1239, "lr": 6.464313112423807e-05}
{"train_loss": 0.1448483169078827, "global_step": 110341, "epoch": 1239, "lr": 6.464257680547454e-05}
{"train_loss": 0.16432711482048035, "global_step": 110342, "epoch": 1239, "lr": 6.464202248474251e-05}
{"train_loss": 0.1164332702755928, "global_step": 110343, "epoch": 1239, "lr": 6.464146816204204e-05}
{"train_loss": 0.1253843456506729, "global_step": 110344, "epoch": 1239, "lr": 6.464091383737318e-05}
{"train_loss": 0.09705791622400284, "global_step": 110345, "epoch": 1239, "lr": 6.464035951073606e-05}
{"train_loss": 0.11563560366630554, "global_step": 110346, "epoch": 1239, "lr": 6.46398051821307e-05}
{"train_loss": 0.16576159000396729, "global_step": 110347, "epoch": 1239, "lr": 6.46392508515572e-05}
{"train_loss": 0.11988995224237442, "global_step": 110348, "epoch": 1239, "lr": 6.463869651901564e-05}
{"train_loss": 0.10839156061410904, "global_step": 110349, "epoch": 1239, "lr": 6.463814218450609e-05}
{"train_loss": 0.13959819078445435, "global_step": 110350, "epoch": 1239, "lr": 6.46375878480286e-05}
{"train_loss": 0.14770005643367767, "global_step": 110351, "epoch": 1239, "lr": 6.463703350958326e-05}
{"train_loss": 0.2282213717699051, "global_step": 110352, "epoch": 1239, "lr": 6.463647916917017e-05}
{"train_loss": 0.13416966795921326, "global_step": 110353, "epoch": 1239, "lr": 6.463592482678938e-05}
{"train_loss": 0.1586884707212448, "global_step": 110354, "epoch": 1239, "lr": 6.463537048244096e-05}
{"train_loss": 0.10857578366994858, "global_step": 110355, "epoch": 1239, "lr": 6.4634816136125e-05}
{"train_loss": 0.17539924383163452, "global_step": 110356, "epoch": 1239, "lr": 6.463426178784157e-05}
{"train_loss": 0.09248779714107513, "global_step": 110357, "epoch": 1239, "lr": 6.463370743759073e-05}
{"train_loss": 0.10102250427007675, "global_step": 110358, "epoch": 1239, "lr": 6.463315308537257e-05}
{"train_loss": 0.15165490548262436, "global_step": 110359, "epoch": 1239, "lr": 6.463259873118716e-05, "val_loss": 5.017479419708252}
{"train_loss": 0.19243945181369781, "global_step": 110360, "epoch": 1240, "lr": 6.463204437503459e-05}
{"train_loss": 0.17728127539157867, "global_step": 110361, "epoch": 1240, "lr": 6.463149001691491e-05}
{"train_loss": 0.14095208048820496, "global_step": 110362, "epoch": 1240, "lr": 6.46309356568282e-05}
{"train_loss": 0.13851402699947357, "global_step": 110363, "epoch": 1240, "lr": 6.463038129477455e-05}
{"train_loss": 0.12768647074699402, "global_step": 110364, "epoch": 1240, "lr": 6.462982693075401e-05}
{"train_loss": 0.17162063717842102, "global_step": 110365, "epoch": 1240, "lr": 6.462927256476668e-05}
{"train_loss": 0.1959601193666458, "global_step": 110366, "epoch": 1240, "lr": 6.462871819681261e-05}
{"train_loss": 0.15410834550857544, "global_step": 110367, "epoch": 1240, "lr": 6.462816382689189e-05}
{"train_loss": 0.1738264113664627, "global_step": 110368, "epoch": 1240, "lr": 6.462760945500461e-05}
{"train_loss": 0.18794593214988708, "global_step": 110369, "epoch": 1240, "lr": 6.462705508115081e-05}
{"train_loss": 0.14049136638641357, "global_step": 110370, "epoch": 1240, "lr": 6.462650070533058e-05}
{"train_loss": 0.174840047955513, "global_step": 110371, "epoch": 1240, "lr": 6.462594632754401e-05}
{"train_loss": 0.1480187326669693, "global_step": 110372, "epoch": 1240, "lr": 6.462539194779113e-05}
{"train_loss": 0.0972888320684433, "global_step": 110373, "epoch": 1240, "lr": 6.462483756607208e-05}
{"train_loss": 0.17356641590595245, "global_step": 110374, "epoch": 1240, "lr": 6.462428318238688e-05}
{"train_loss": 0.175653338432312, "global_step": 110375, "epoch": 1240, "lr": 6.462372879673562e-05}
{"train_loss": 0.1390431970357895, "global_step": 110376, "epoch": 1240, "lr": 6.462317440911838e-05}
{"train_loss": 0.25518980622291565, "global_step": 110377, "epoch": 1240, "lr": 6.462262001953524e-05}
{"train_loss": 0.17811936140060425, "global_step": 110378, "epoch": 1240, "lr": 6.462206562798626e-05}
{"train_loss": 0.12726597487926483, "global_step": 110379, "epoch": 1240, "lr": 6.462151123447154e-05}
{"train_loss": 0.160129576921463, "global_step": 110380, "epoch": 1240, "lr": 6.462095683899111e-05}
{"train_loss": 0.1664372831583023, "global_step": 110381, "epoch": 1240, "lr": 6.462040244154508e-05}
{"train_loss": 0.07200883328914642, "global_step": 110382, "epoch": 1240, "lr": 6.461984804213352e-05}
{"train_loss": 0.12295691668987274, "global_step": 110383, "epoch": 1240, "lr": 6.461929364075648e-05}
{"train_loss": 0.16136743128299713, "global_step": 110384, "epoch": 1240, "lr": 6.461873923741408e-05}
{"train_loss": 0.1406872570514679, "global_step": 110385, "epoch": 1240, "lr": 6.461818483210635e-05}
{"train_loss": 0.09163373708724976, "global_step": 110386, "epoch": 1240, "lr": 6.46176304248334e-05}
{"train_loss": 0.17532438039779663, "global_step": 110387, "epoch": 1240, "lr": 6.461707601559528e-05}
{"train_loss": 0.17465326189994812, "global_step": 110388, "epoch": 1240, "lr": 6.461652160439206e-05}
{"train_loss": 0.1681486964225769, "global_step": 110389, "epoch": 1240, "lr": 6.461596719122384e-05}
{"train_loss": 0.15285903215408325, "global_step": 110390, "epoch": 1240, "lr": 6.461541277609068e-05}
{"train_loss": 0.1464216560125351, "global_step": 110391, "epoch": 1240, "lr": 6.461485835899266e-05}
{"train_loss": 0.17527323961257935, "global_step": 110392, "epoch": 1240, "lr": 6.461430393992984e-05}
{"train_loss": 0.12839066982269287, "global_step": 110393, "epoch": 1240, "lr": 6.46137495189023e-05}
{"train_loss": 0.14399445056915283, "global_step": 110394, "epoch": 1240, "lr": 6.461319509591014e-05}
{"train_loss": 0.0992819294333458, "global_step": 110395, "epoch": 1240, "lr": 6.46126406709534e-05}
{"train_loss": 0.14598314464092255, "global_step": 110396, "epoch": 1240, "lr": 6.461208624403218e-05}
{"train_loss": 0.11211340129375458, "global_step": 110397, "epoch": 1240, "lr": 6.461153181514654e-05}
{"train_loss": 0.17165285348892212, "global_step": 110398, "epoch": 1240, "lr": 6.461097738429655e-05}
{"train_loss": 0.1932302713394165, "global_step": 110399, "epoch": 1240, "lr": 6.46104229514823e-05}
{"train_loss": 0.1007104367017746, "global_step": 110400, "epoch": 1240, "lr": 6.460986851670386e-05}
{"train_loss": 0.1528359353542328, "global_step": 110401, "epoch": 1240, "lr": 6.460931407996128e-05}
{"train_loss": 0.09094442427158356, "global_step": 110402, "epoch": 1240, "lr": 6.460875964125467e-05}
{"train_loss": 0.0992836281657219, "global_step": 110403, "epoch": 1240, "lr": 6.460820520058409e-05}
{"train_loss": 0.09409963339567184, "global_step": 110404, "epoch": 1240, "lr": 6.460765075794962e-05}
{"train_loss": 0.1066337376832962, "global_step": 110405, "epoch": 1240, "lr": 6.460709631335131e-05}
{"train_loss": 0.1511334329843521, "global_step": 110406, "epoch": 1240, "lr": 6.460654186678927e-05}
{"train_loss": 0.12651115655899048, "global_step": 110407, "epoch": 1240, "lr": 6.460598741826356e-05}
{"train_loss": 0.14039263129234314, "global_step": 110408, "epoch": 1240, "lr": 6.460543296777425e-05}
{"train_loss": 0.11929309368133545, "global_step": 110409, "epoch": 1240, "lr": 6.460487851532142e-05}
{"train_loss": 0.1323632150888443, "global_step": 110410, "epoch": 1240, "lr": 6.460432406090514e-05}
{"train_loss": 0.15308897197246552, "global_step": 110411, "epoch": 1240, "lr": 6.460376960452547e-05}
{"train_loss": 0.17446550726890564, "global_step": 110412, "epoch": 1240, "lr": 6.460321514618253e-05}
{"train_loss": 0.15445426106452942, "global_step": 110413, "epoch": 1240, "lr": 6.460266068587634e-05}
{"train_loss": 0.15218320488929749, "global_step": 110414, "epoch": 1240, "lr": 6.460210622360703e-05}
{"train_loss": 0.15987196564674377, "global_step": 110415, "epoch": 1240, "lr": 6.460155175937463e-05}
{"train_loss": 0.1397610604763031, "global_step": 110416, "epoch": 1240, "lr": 6.460099729317922e-05}
{"train_loss": 0.21352548897266388, "global_step": 110417, "epoch": 1240, "lr": 6.460044282502089e-05}
{"train_loss": 0.1813390851020813, "global_step": 110418, "epoch": 1240, "lr": 6.45998883548997e-05}
{"train_loss": 0.12543900310993195, "global_step": 110419, "epoch": 1240, "lr": 6.459933388281575e-05}
{"train_loss": 0.15534289181232452, "global_step": 110420, "epoch": 1240, "lr": 6.459877940876909e-05}
{"train_loss": 0.10781408101320267, "global_step": 110421, "epoch": 1240, "lr": 6.45982249327598e-05}
{"train_loss": 0.17810392379760742, "global_step": 110422, "epoch": 1240, "lr": 6.459767045478797e-05}
{"train_loss": 0.12294694781303406, "global_step": 110423, "epoch": 1240, "lr": 6.459711597485364e-05}
{"train_loss": 0.15216469764709473, "global_step": 110424, "epoch": 1240, "lr": 6.459656149295692e-05}
{"train_loss": 0.14971089363098145, "global_step": 110425, "epoch": 1240, "lr": 6.459600700909785e-05}
{"train_loss": 0.1920781284570694, "global_step": 110426, "epoch": 1240, "lr": 6.459545252327654e-05}
{"train_loss": 0.19286619126796722, "global_step": 110427, "epoch": 1240, "lr": 6.459489803549305e-05}
{"train_loss": 0.1597902923822403, "global_step": 110428, "epoch": 1240, "lr": 6.459434354574746e-05}
{"train_loss": 0.23685181140899658, "global_step": 110429, "epoch": 1240, "lr": 6.459378905403984e-05}
{"train_loss": 0.2070997804403305, "global_step": 110430, "epoch": 1240, "lr": 6.459323456037026e-05}
{"train_loss": 0.1240360364317894, "global_step": 110431, "epoch": 1240, "lr": 6.459268006473877e-05}
{"train_loss": 0.15866325795650482, "global_step": 110432, "epoch": 1240, "lr": 6.459212556714552e-05}
{"train_loss": 0.09533657878637314, "global_step": 110433, "epoch": 1240, "lr": 6.459157106759049e-05}
{"train_loss": 0.16093531250953674, "global_step": 110434, "epoch": 1240, "lr": 6.459101656607382e-05}
{"train_loss": 0.08221392333507538, "global_step": 110435, "epoch": 1240, "lr": 6.459046206259559e-05}
{"train_loss": 0.18808506429195404, "global_step": 110436, "epoch": 1240, "lr": 6.458990755715581e-05}
{"train_loss": 0.11284241825342178, "global_step": 110437, "epoch": 1240, "lr": 6.458935304975462e-05}
{"train_loss": 0.12187042832374573, "global_step": 110438, "epoch": 1240, "lr": 6.458879854039205e-05}
{"train_loss": 0.14070504903793335, "global_step": 110439, "epoch": 1240, "lr": 6.45882440290682e-05}
{"train_loss": 0.14278508722782135, "global_step": 110440, "epoch": 1240, "lr": 6.458768951578316e-05}
{"train_loss": 0.16394731402397156, "global_step": 110441, "epoch": 1240, "lr": 6.458713500053696e-05}
{"train_loss": 0.15139713883399963, "global_step": 110442, "epoch": 1240, "lr": 6.45865804833297e-05}
{"train_loss": 0.2018253058195114, "global_step": 110443, "epoch": 1240, "lr": 6.458602596416147e-05}
{"train_loss": 0.11835768818855286, "global_step": 110444, "epoch": 1240, "lr": 6.458547144303231e-05}
{"train_loss": 0.156032532453537, "global_step": 110445, "epoch": 1240, "lr": 6.458491691994231e-05}
{"train_loss": 0.0801621600985527, "global_step": 110446, "epoch": 1240, "lr": 6.458436239489155e-05}
{"train_loss": 0.11537805944681168, "global_step": 110447, "epoch": 1240, "lr": 6.45838078678801e-05}
{"train_loss": 0.14833158999681473, "global_step": 110448, "epoch": 1240, "lr": 6.458325333890804e-05, "val_loss": 4.999979496002197, "train_action_mse_error": 20.966625213623047}
{"train_loss": 0.11692846566438675, "global_step": 110449, "epoch": 1241, "lr": 6.458269880797543e-05}
{"train_loss": 0.14230363070964813, "global_step": 110450, "epoch": 1241, "lr": 6.458214427508236e-05}
{"train_loss": 0.09071899950504303, "global_step": 110451, "epoch": 1241, "lr": 6.45815897402289e-05}
{"train_loss": 0.13232064247131348, "global_step": 110452, "epoch": 1241, "lr": 6.458103520341512e-05}
{"train_loss": 0.13334247469902039, "global_step": 110453, "epoch": 1241, "lr": 6.458048066464109e-05}
{"train_loss": 0.10012005269527435, "global_step": 110454, "epoch": 1241, "lr": 6.45799261239069e-05}
{"train_loss": 0.11501540243625641, "global_step": 110455, "epoch": 1241, "lr": 6.457937158121261e-05}
{"train_loss": 0.2003941833972931, "global_step": 110456, "epoch": 1241, "lr": 6.45788170365583e-05}
{"train_loss": 0.16884809732437134, "global_step": 110457, "epoch": 1241, "lr": 6.457826248994405e-05}
{"train_loss": 0.22352080047130585, "global_step": 110458, "epoch": 1241, "lr": 6.457770794136993e-05}
{"train_loss": 0.1809709221124649, "global_step": 110459, "epoch": 1241, "lr": 6.457715339083602e-05}
{"train_loss": 0.1456795483827591, "global_step": 110460, "epoch": 1241, "lr": 6.45765988383424e-05}
{"train_loss": 0.11669103056192398, "global_step": 110461, "epoch": 1241, "lr": 6.457604428388912e-05}
{"train_loss": 0.11227136105298996, "global_step": 110462, "epoch": 1241, "lr": 6.457548972747627e-05}
{"train_loss": 0.13780295848846436, "global_step": 110463, "epoch": 1241, "lr": 6.457493516910392e-05}
{"train_loss": 0.15762172639369965, "global_step": 110464, "epoch": 1241, "lr": 6.457438060877216e-05}
{"train_loss": 0.17328307032585144, "global_step": 110465, "epoch": 1241, "lr": 6.457382604648103e-05}
{"train_loss": 0.11628755927085876, "global_step": 110466, "epoch": 1241, "lr": 6.457327148223065e-05}
{"train_loss": 0.0629243552684784, "global_step": 110467, "epoch": 1241, "lr": 6.457271691602106e-05}
{"train_loss": 0.12101195007562637, "global_step": 110468, "epoch": 1241, "lr": 6.457216234785236e-05}
{"train_loss": 0.18312117457389832, "global_step": 110469, "epoch": 1241, "lr": 6.457160777772458e-05}
{"train_loss": 0.11830612272024155, "global_step": 110470, "epoch": 1241, "lr": 6.457105320563784e-05}
{"train_loss": 0.09364954382181168, "global_step": 110471, "epoch": 1241, "lr": 6.45704986315922e-05}
{"train_loss": 0.181789830327034, "global_step": 110472, "epoch": 1241, "lr": 6.456994405558774e-05}
{"train_loss": 0.13792045414447784, "global_step": 110473, "epoch": 1241, "lr": 6.456938947762452e-05}
{"train_loss": 0.1364559531211853, "global_step": 110474, "epoch": 1241, "lr": 6.456883489770262e-05}
{"train_loss": 0.15119975805282593, "global_step": 110475, "epoch": 1241, "lr": 6.456828031582212e-05}
{"train_loss": 0.1394684612751007, "global_step": 110476, "epoch": 1241, "lr": 6.45677257319831e-05}
{"train_loss": 0.13646693527698517, "global_step": 110477, "epoch": 1241, "lr": 6.456717114618562e-05}
{"train_loss": 0.22743533551692963, "global_step": 110478, "epoch": 1241, "lr": 6.456661655842978e-05}
{"train_loss": 0.16580817103385925, "global_step": 110479, "epoch": 1241, "lr": 6.456606196871562e-05}
{"train_loss": 0.1552247852087021, "global_step": 110480, "epoch": 1241, "lr": 6.456550737704323e-05}
{"train_loss": 0.08628369867801666, "global_step": 110481, "epoch": 1241, "lr": 6.456495278341269e-05}
{"train_loss": 0.14750033617019653, "global_step": 110482, "epoch": 1241, "lr": 6.456439818782406e-05}
{"train_loss": 0.12553103268146515, "global_step": 110483, "epoch": 1241, "lr": 6.456384359027742e-05}
{"train_loss": 0.13977792859077454, "global_step": 110484, "epoch": 1241, "lr": 6.456328899077288e-05}
{"train_loss": 0.21308888494968414, "global_step": 110485, "epoch": 1241, "lr": 6.456273438931045e-05}
{"train_loss": 0.17652691900730133, "global_step": 110486, "epoch": 1241, "lr": 6.456217978589026e-05}
{"train_loss": 0.11626729369163513, "global_step": 110487, "epoch": 1241, "lr": 6.456162518051235e-05}
{"train_loss": 0.20329046249389648, "global_step": 110488, "epoch": 1241, "lr": 6.456107057317681e-05}
{"train_loss": 0.17405086755752563, "global_step": 110489, "epoch": 1241, "lr": 6.456051596388371e-05}
{"train_loss": 0.08304838091135025, "global_step": 110490, "epoch": 1241, "lr": 6.455996135263312e-05}
{"train_loss": 0.11549399048089981, "global_step": 110491, "epoch": 1241, "lr": 6.455940673942513e-05}
{"train_loss": 0.11938190460205078, "global_step": 110492, "epoch": 1241, "lr": 6.45588521242598e-05}
{"train_loss": 0.13073699176311493, "global_step": 110493, "epoch": 1241, "lr": 6.455829750713721e-05}
{"train_loss": 0.13281358778476715, "global_step": 110494, "epoch": 1241, "lr": 6.455774288805744e-05}
{"train_loss": 0.14768190681934357, "global_step": 110495, "epoch": 1241, "lr": 6.455718826702056e-05}
{"train_loss": 0.20168522000312805, "global_step": 110496, "epoch": 1241, "lr": 6.455663364402664e-05}
{"train_loss": 0.1557423174381256, "global_step": 110497, "epoch": 1241, "lr": 6.455607901907576e-05}
{"train_loss": 0.14944632351398468, "global_step": 110498, "epoch": 1241, "lr": 6.455552439216799e-05}
{"train_loss": 0.14809684455394745, "global_step": 110499, "epoch": 1241, "lr": 6.455496976330341e-05}
{"train_loss": 0.14866411685943604, "global_step": 110500, "epoch": 1241, "lr": 6.455441513248208e-05}
{"train_loss": 0.15466691553592682, "global_step": 110501, "epoch": 1241, "lr": 6.455386049970412e-05}
{"train_loss": 0.11327050626277924, "global_step": 110502, "epoch": 1241, "lr": 6.455330586496953e-05}
{"train_loss": 0.13800127804279327, "global_step": 110503, "epoch": 1241, "lr": 6.455275122827844e-05}
{"train_loss": 0.2035422921180725, "global_step": 110504, "epoch": 1241, "lr": 6.455219658963091e-05}
{"train_loss": 0.16672354936599731, "global_step": 110505, "epoch": 1241, "lr": 6.455164194902702e-05}
{"train_loss": 0.1419474333524704, "global_step": 110506, "epoch": 1241, "lr": 6.455108730646684e-05}
{"train_loss": 0.1268259584903717, "global_step": 110507, "epoch": 1241, "lr": 6.455053266195043e-05}
{"train_loss": 0.1540847271680832, "global_step": 110508, "epoch": 1241, "lr": 6.454997801547788e-05}
{"train_loss": 0.1414281576871872, "global_step": 110509, "epoch": 1241, "lr": 6.454942336704927e-05}
{"train_loss": 0.19943569600582123, "global_step": 110510, "epoch": 1241, "lr": 6.454886871666468e-05}
{"train_loss": 0.13486552238464355, "global_step": 110511, "epoch": 1241, "lr": 6.454831406432414e-05}
{"train_loss": 0.2088196724653244, "global_step": 110512, "epoch": 1241, "lr": 6.454775941002779e-05}
{"train_loss": 0.11818085610866547, "global_step": 110513, "epoch": 1241, "lr": 6.454720475377565e-05}
{"train_loss": 0.1568857729434967, "global_step": 110514, "epoch": 1241, "lr": 6.454665009556782e-05}
{"train_loss": 0.16088293492794037, "global_step": 110515, "epoch": 1241, "lr": 6.454609543540437e-05}
{"train_loss": 0.15148420631885529, "global_step": 110516, "epoch": 1241, "lr": 6.454554077328539e-05}
{"train_loss": 0.16833005845546722, "global_step": 110517, "epoch": 1241, "lr": 6.454498610921091e-05}
{"train_loss": 0.21544770896434784, "global_step": 110518, "epoch": 1241, "lr": 6.454443144318108e-05}
{"train_loss": 0.16142922639846802, "global_step": 110519, "epoch": 1241, "lr": 6.454387677519588e-05}
{"train_loss": 0.21687622368335724, "global_step": 110520, "epoch": 1241, "lr": 6.454332210525547e-05}
{"train_loss": 0.13595907390117645, "global_step": 110521, "epoch": 1241, "lr": 6.454276743335987e-05}
{"train_loss": 0.14665785431861877, "global_step": 110522, "epoch": 1241, "lr": 6.454221275950915e-05}
{"train_loss": 0.14771610498428345, "global_step": 110523, "epoch": 1241, "lr": 6.454165808370343e-05}
{"train_loss": 0.08832009881734848, "global_step": 110524, "epoch": 1241, "lr": 6.454110340594278e-05}
{"train_loss": 0.11512373387813568, "global_step": 110525, "epoch": 1241, "lr": 6.454054872622722e-05}
{"train_loss": 0.11935350298881531, "global_step": 110526, "epoch": 1241, "lr": 6.453999404455689e-05}
{"train_loss": 0.13582772016525269, "global_step": 110527, "epoch": 1241, "lr": 6.45394393609318e-05}
{"train_loss": 0.20112960040569305, "global_step": 110528, "epoch": 1241, "lr": 6.45388846753521e-05}
{"train_loss": 0.22138360142707825, "global_step": 110529, "epoch": 1241, "lr": 6.453832998781779e-05}
{"train_loss": 0.10250914096832275, "global_step": 110530, "epoch": 1241, "lr": 6.453777529832899e-05}
{"train_loss": 0.2635855972766876, "global_step": 110531, "epoch": 1241, "lr": 6.453722060688577e-05}
{"train_loss": 0.24844640493392944, "global_step": 110532, "epoch": 1241, "lr": 6.453666591348819e-05}
{"train_loss": 0.22326907515525818, "global_step": 110533, "epoch": 1241, "lr": 6.453611121813634e-05}
{"train_loss": 0.21905632317066193, "global_step": 110534, "epoch": 1241, "lr": 6.453555652083029e-05}
{"train_loss": 0.13030801713466644, "global_step": 110535, "epoch": 1241, "lr": 6.453500182157011e-05}
{"train_loss": 0.1337423026561737, "global_step": 110536, "epoch": 1241, "lr": 6.453444712035587e-05}
{"train_loss": 0.15153783333770346, "global_step": 110537, "epoch": 1241, "lr": 6.453389241718766e-05, "val_loss": 5.084967136383057}
{"train_loss": 0.19231946766376495, "global_step": 110538, "epoch": 1242, "lr": 6.453333771206553e-05}
{"train_loss": 0.15638194978237152, "global_step": 110539, "epoch": 1242, "lr": 6.453278300498958e-05}
{"train_loss": 0.19499610364437103, "global_step": 110540, "epoch": 1242, "lr": 6.453222829595988e-05}
{"train_loss": 0.1732090711593628, "global_step": 110541, "epoch": 1242, "lr": 6.453167358497649e-05}
{"train_loss": 0.12999624013900757, "global_step": 110542, "epoch": 1242, "lr": 6.45311188720395e-05}
{"train_loss": 0.19553793966770172, "global_step": 110543, "epoch": 1242, "lr": 6.453056415714897e-05}
{"train_loss": 0.13133123517036438, "global_step": 110544, "epoch": 1242, "lr": 6.453000944030499e-05}
{"train_loss": 0.2332053780555725, "global_step": 110545, "epoch": 1242, "lr": 6.452945472150762e-05}
{"train_loss": 0.14238937199115753, "global_step": 110546, "epoch": 1242, "lr": 6.452890000075695e-05}
{"train_loss": 0.1361120343208313, "global_step": 110547, "epoch": 1242, "lr": 6.452834527805303e-05}
{"train_loss": 0.21403761208057404, "global_step": 110548, "epoch": 1242, "lr": 6.452779055339595e-05}
{"train_loss": 0.18852539360523224, "global_step": 110549, "epoch": 1242, "lr": 6.452723582678581e-05}
{"train_loss": 0.1316002756357193, "global_step": 110550, "epoch": 1242, "lr": 6.452668109822265e-05}
{"train_loss": 0.15295863151550293, "global_step": 110551, "epoch": 1242, "lr": 6.452612636770654e-05}
{"train_loss": 0.12596864998340607, "global_step": 110552, "epoch": 1242, "lr": 6.45255716352376e-05}
{"train_loss": 0.21858075261116028, "global_step": 110553, "epoch": 1242, "lr": 6.452501690081585e-05}
{"train_loss": 0.12033829092979431, "global_step": 110554, "epoch": 1242, "lr": 6.45244621644414e-05}
{"train_loss": 0.19375228881835938, "global_step": 110555, "epoch": 1242, "lr": 6.452390742611429e-05}
{"train_loss": 0.3272631764411926, "global_step": 110556, "epoch": 1242, "lr": 6.452335268583466e-05}
{"train_loss": 0.2500974237918854, "global_step": 110557, "epoch": 1242, "lr": 6.452279794360253e-05}
{"train_loss": 0.1969124674797058, "global_step": 110558, "epoch": 1242, "lr": 6.452224319941795e-05}
{"train_loss": 0.1458054482936859, "global_step": 110559, "epoch": 1242, "lr": 6.452168845328107e-05}
{"train_loss": 0.1275118738412857, "global_step": 110560, "epoch": 1242, "lr": 6.452113370519191e-05}
{"train_loss": 0.10071200132369995, "global_step": 110561, "epoch": 1242, "lr": 6.452057895515056e-05}
{"train_loss": 0.1522638201713562, "global_step": 110562, "epoch": 1242, "lr": 6.452002420315711e-05}
{"train_loss": 0.14568907022476196, "global_step": 110563, "epoch": 1242, "lr": 6.45194694492116e-05}
{"train_loss": 0.19371230900287628, "global_step": 110564, "epoch": 1242, "lr": 6.451891469331413e-05}
{"train_loss": 0.16661104559898376, "global_step": 110565, "epoch": 1242, "lr": 6.451835993546476e-05}
{"train_loss": 0.12012929469347, "global_step": 110566, "epoch": 1242, "lr": 6.451780517566358e-05}
{"train_loss": 0.12414438277482986, "global_step": 110567, "epoch": 1242, "lr": 6.451725041391067e-05}
{"train_loss": 0.21202600002288818, "global_step": 110568, "epoch": 1242, "lr": 6.451669565020608e-05}
{"train_loss": 0.1626163125038147, "global_step": 110569, "epoch": 1242, "lr": 6.45161408845499e-05}
{"train_loss": 0.11368463933467865, "global_step": 110570, "epoch": 1242, "lr": 6.45155861169422e-05}
{"train_loss": 0.21111200749874115, "global_step": 110571, "epoch": 1242, "lr": 6.451503134738306e-05}
{"train_loss": 0.13281956315040588, "global_step": 110572, "epoch": 1242, "lr": 6.451447657587256e-05}
{"train_loss": 0.179006427526474, "global_step": 110573, "epoch": 1242, "lr": 6.451392180241075e-05}
{"train_loss": 0.20694975554943085, "global_step": 110574, "epoch": 1242, "lr": 6.451336702699774e-05}
{"train_loss": 0.16862566769123077, "global_step": 110575, "epoch": 1242, "lr": 6.451281224963357e-05}
{"train_loss": 0.16665910184383392, "global_step": 110576, "epoch": 1242, "lr": 6.451225747031834e-05}
{"train_loss": 0.11839757859706879, "global_step": 110577, "epoch": 1242, "lr": 6.451170268905211e-05}
{"train_loss": 0.1839793175458908, "global_step": 110578, "epoch": 1242, "lr": 6.451114790583495e-05}
{"train_loss": 0.1364126056432724, "global_step": 110579, "epoch": 1242, "lr": 6.451059312066694e-05}
{"train_loss": 0.1595880687236786, "global_step": 110580, "epoch": 1242, "lr": 6.451003833354816e-05}
{"train_loss": 0.13936203718185425, "global_step": 110581, "epoch": 1242, "lr": 6.45094835444787e-05}
{"train_loss": 0.15038277208805084, "global_step": 110582, "epoch": 1242, "lr": 6.45089287534586e-05}
{"train_loss": 0.10947321355342865, "global_step": 110583, "epoch": 1242, "lr": 6.450837396048794e-05}
{"train_loss": 0.17119023203849792, "global_step": 110584, "epoch": 1242, "lr": 6.450781916556682e-05}
{"train_loss": 0.18316248059272766, "global_step": 110585, "epoch": 1242, "lr": 6.45072643686953e-05}
{"train_loss": 0.1760638803243637, "global_step": 110586, "epoch": 1242, "lr": 6.450670956987346e-05}
{"train_loss": 0.33612191677093506, "global_step": 110587, "epoch": 1242, "lr": 6.450615476910137e-05}
{"train_loss": 0.16861577332019806, "global_step": 110588, "epoch": 1242, "lr": 6.450559996637909e-05}
{"train_loss": 0.11114334315061569, "global_step": 110589, "epoch": 1242, "lr": 6.450504516170672e-05}
{"train_loss": 0.23410286009311676, "global_step": 110590, "epoch": 1242, "lr": 6.450449035508433e-05}
{"train_loss": 0.1789456158876419, "global_step": 110591, "epoch": 1242, "lr": 6.450393554651198e-05}
{"train_loss": 0.11861728131771088, "global_step": 110592, "epoch": 1242, "lr": 6.450338073598975e-05}
{"train_loss": 0.1526317000389099, "global_step": 110593, "epoch": 1242, "lr": 6.450282592351772e-05}
{"train_loss": 0.08716181665658951, "global_step": 110594, "epoch": 1242, "lr": 6.450227110909597e-05}
{"train_loss": 0.12746132910251617, "global_step": 110595, "epoch": 1242, "lr": 6.450171629272457e-05}
{"train_loss": 0.165004700422287, "global_step": 110596, "epoch": 1242, "lr": 6.450116147440358e-05}
{"train_loss": 0.15369410812854767, "global_step": 110597, "epoch": 1242, "lr": 6.450060665413308e-05}
{"train_loss": 0.14918743073940277, "global_step": 110598, "epoch": 1242, "lr": 6.450005183191316e-05}
{"train_loss": 0.10607896745204926, "global_step": 110599, "epoch": 1242, "lr": 6.449949700774387e-05}
{"train_loss": 0.09117947518825531, "global_step": 110600, "epoch": 1242, "lr": 6.449894218162531e-05}
{"train_loss": 0.21100834012031555, "global_step": 110601, "epoch": 1242, "lr": 6.449838735355753e-05}
{"train_loss": 0.16005000472068787, "global_step": 110602, "epoch": 1242, "lr": 6.449783252354064e-05}
{"train_loss": 0.13633553683757782, "global_step": 110603, "epoch": 1242, "lr": 6.449727769157468e-05}
{"train_loss": 0.14628124237060547, "global_step": 110604, "epoch": 1242, "lr": 6.449672285765974e-05}
{"train_loss": 0.17531843483448029, "global_step": 110605, "epoch": 1242, "lr": 6.44961680217959e-05}
{"train_loss": 0.11775798350572586, "global_step": 110606, "epoch": 1242, "lr": 6.449561318398322e-05}
{"train_loss": 0.20786599814891815, "global_step": 110607, "epoch": 1242, "lr": 6.449505834422179e-05}
{"train_loss": 0.12269116938114166, "global_step": 110608, "epoch": 1242, "lr": 6.449450350251167e-05}
{"train_loss": 0.25871989130973816, "global_step": 110609, "epoch": 1242, "lr": 6.449394865885294e-05}
{"train_loss": 0.10910320281982422, "global_step": 110610, "epoch": 1242, "lr": 6.449339381324567e-05}
{"train_loss": 0.1727171540260315, "global_step": 110611, "epoch": 1242, "lr": 6.449283896568993e-05}
{"train_loss": 0.13261109590530396, "global_step": 110612, "epoch": 1242, "lr": 6.449228411618582e-05}
{"train_loss": 0.12561936676502228, "global_step": 110613, "epoch": 1242, "lr": 6.449172926473342e-05}
{"train_loss": 0.13090305030345917, "global_step": 110614, "epoch": 1242, "lr": 6.449117441133276e-05}
{"train_loss": 0.07231105118989944, "global_step": 110615, "epoch": 1242, "lr": 6.449061955598395e-05}
{"train_loss": 0.15351815521717072, "global_step": 110616, "epoch": 1242, "lr": 6.449006469868704e-05}
{"train_loss": 0.14361849427223206, "global_step": 110617, "epoch": 1242, "lr": 6.448950983944212e-05}
{"train_loss": 0.10436784476041794, "global_step": 110618, "epoch": 1242, "lr": 6.448895497824927e-05}
{"train_loss": 0.2096535563468933, "global_step": 110619, "epoch": 1242, "lr": 6.448840011510852e-05}
{"train_loss": 0.10513687133789062, "global_step": 110620, "epoch": 1242, "lr": 6.448784525002004e-05}
{"train_loss": 0.10767830908298492, "global_step": 110621, "epoch": 1242, "lr": 6.448729038298381e-05}
{"train_loss": 0.17472675442695618, "global_step": 110622, "epoch": 1242, "lr": 6.448673551399993e-05}
{"train_loss": 0.10715223848819733, "global_step": 110623, "epoch": 1242, "lr": 6.44861806430685e-05}
{"train_loss": 0.1936783790588379, "global_step": 110624, "epoch": 1242, "lr": 6.448562577018959e-05}
{"train_loss": 0.25859397649765015, "global_step": 110625, "epoch": 1242, "lr": 6.448507089536326e-05}
{"train_loss": 0.16156572764844038, "global_step": 110626, "epoch": 1242, "lr": 6.448451601858958e-05, "val_loss": 5.067280292510986}
{"train_loss": 0.14088816940784454, "global_step": 110627, "epoch": 1243, "lr": 6.448396113986863e-05}
{"train_loss": 0.18278323113918304, "global_step": 110628, "epoch": 1243, "lr": 6.44834062592005e-05}
{"train_loss": 0.13611695170402527, "global_step": 110629, "epoch": 1243, "lr": 6.448285137658525e-05}
{"train_loss": 0.19784529507160187, "global_step": 110630, "epoch": 1243, "lr": 6.448229649202295e-05}
{"train_loss": 0.11645306646823883, "global_step": 110631, "epoch": 1243, "lr": 6.448174160551369e-05}
{"train_loss": 0.17511585354804993, "global_step": 110632, "epoch": 1243, "lr": 6.448118671705754e-05}
{"train_loss": 0.1473318636417389, "global_step": 110633, "epoch": 1243, "lr": 6.448063182665456e-05}
{"train_loss": 0.10665159672498703, "global_step": 110634, "epoch": 1243, "lr": 6.448007693430484e-05}
{"train_loss": 0.11234097182750702, "global_step": 110635, "epoch": 1243, "lr": 6.447952204000846e-05}
{"train_loss": 0.09804702550172806, "global_step": 110636, "epoch": 1243, "lr": 6.447896714376548e-05}
{"train_loss": 0.17850513756275177, "global_step": 110637, "epoch": 1243, "lr": 6.447841224557598e-05}
{"train_loss": 0.1424679011106491, "global_step": 110638, "epoch": 1243, "lr": 6.447785734544e-05}
{"train_loss": 0.17368271946907043, "global_step": 110639, "epoch": 1243, "lr": 6.447730244335768e-05}
{"train_loss": 0.13378508388996124, "global_step": 110640, "epoch": 1243, "lr": 6.447674753932906e-05}
{"train_loss": 0.10588539391756058, "global_step": 110641, "epoch": 1243, "lr": 6.447619263335423e-05}
{"train_loss": 0.15974532067775726, "global_step": 110642, "epoch": 1243, "lr": 6.447563772543323e-05}
{"train_loss": 0.24698640406131744, "global_step": 110643, "epoch": 1243, "lr": 6.447508281556617e-05}
{"train_loss": 0.1764240711927414, "global_step": 110644, "epoch": 1243, "lr": 6.447452790375311e-05}
{"train_loss": 0.16900858283042908, "global_step": 110645, "epoch": 1243, "lr": 6.447397298999411e-05}
{"train_loss": 0.23171856999397278, "global_step": 110646, "epoch": 1243, "lr": 6.447341807428928e-05}
{"train_loss": 0.1186620369553566, "global_step": 110647, "epoch": 1243, "lr": 6.447286315663867e-05}
{"train_loss": 0.09771512448787689, "global_step": 110648, "epoch": 1243, "lr": 6.447230823704236e-05}
{"train_loss": 0.2668384909629822, "global_step": 110649, "epoch": 1243, "lr": 6.447175331550043e-05}
{"train_loss": 0.13482165336608887, "global_step": 110650, "epoch": 1243, "lr": 6.447119839201294e-05}
{"train_loss": 0.1720881164073944, "global_step": 110651, "epoch": 1243, "lr": 6.447064346657998e-05}
{"train_loss": 0.2200567126274109, "global_step": 110652, "epoch": 1243, "lr": 6.447008853920162e-05}
{"train_loss": 0.15980680286884308, "global_step": 110653, "epoch": 1243, "lr": 6.446953360987795e-05}
{"train_loss": 0.17258305847644806, "global_step": 110654, "epoch": 1243, "lr": 6.4468978678609e-05}
{"train_loss": 0.14991122484207153, "global_step": 110655, "epoch": 1243, "lr": 6.446842374539487e-05}
{"train_loss": 0.11170525848865509, "global_step": 110656, "epoch": 1243, "lr": 6.446786881023565e-05}
{"train_loss": 0.2037486433982849, "global_step": 110657, "epoch": 1243, "lr": 6.446731387313139e-05}
{"train_loss": 0.23527775704860687, "global_step": 110658, "epoch": 1243, "lr": 6.446675893408218e-05}
{"train_loss": 0.14383742213249207, "global_step": 110659, "epoch": 1243, "lr": 6.446620399308809e-05}
{"train_loss": 0.2132045328617096, "global_step": 110660, "epoch": 1243, "lr": 6.44656490501492e-05}
{"train_loss": 0.1259312480688095, "global_step": 110661, "epoch": 1243, "lr": 6.446509410526557e-05}
{"train_loss": 0.20363706350326538, "global_step": 110662, "epoch": 1243, "lr": 6.446453915843731e-05}
{"train_loss": 0.12012901902198792, "global_step": 110663, "epoch": 1243, "lr": 6.446398420966445e-05}
{"train_loss": 0.1508416086435318, "global_step": 110664, "epoch": 1243, "lr": 6.446342925894708e-05}
{"train_loss": 0.12049662321805954, "global_step": 110665, "epoch": 1243, "lr": 6.446287430628528e-05}
{"train_loss": 0.13246272504329681, "global_step": 110666, "epoch": 1243, "lr": 6.446231935167913e-05}
{"train_loss": 0.17987333238124847, "global_step": 110667, "epoch": 1243, "lr": 6.44617643951287e-05}
{"train_loss": 0.13936524093151093, "global_step": 110668, "epoch": 1243, "lr": 6.446120943663406e-05}
{"train_loss": 0.17419534921646118, "global_step": 110669, "epoch": 1243, "lr": 6.446065447619529e-05}
{"train_loss": 0.12215908616781235, "global_step": 110670, "epoch": 1243, "lr": 6.446009951381245e-05}
{"train_loss": 0.17323161661624908, "global_step": 110671, "epoch": 1243, "lr": 6.445954454948562e-05}
{"train_loss": 0.19713741540908813, "global_step": 110672, "epoch": 1243, "lr": 6.44589895832149e-05}
{"train_loss": 0.20302854478359222, "global_step": 110673, "epoch": 1243, "lr": 6.445843461500035e-05}
{"train_loss": 0.1822212189435959, "global_step": 110674, "epoch": 1243, "lr": 6.4457879644842e-05}
{"train_loss": 0.168935626745224, "global_step": 110675, "epoch": 1243, "lr": 6.445732467274e-05}
{"train_loss": 0.2261197566986084, "global_step": 110676, "epoch": 1243, "lr": 6.445676969869437e-05}
{"train_loss": 0.19352515041828156, "global_step": 110677, "epoch": 1243, "lr": 6.445621472270521e-05}
{"train_loss": 0.1864880919456482, "global_step": 110678, "epoch": 1243, "lr": 6.445565974477259e-05}
{"train_loss": 0.17412370443344116, "global_step": 110679, "epoch": 1243, "lr": 6.445510476489659e-05}
{"train_loss": 0.15254271030426025, "global_step": 110680, "epoch": 1243, "lr": 6.445454978307726e-05}
{"train_loss": 0.11433254182338715, "global_step": 110681, "epoch": 1243, "lr": 6.445399479931472e-05}
{"train_loss": 0.16886664927005768, "global_step": 110682, "epoch": 1243, "lr": 6.445343981360898e-05}
{"train_loss": 0.189313605427742, "global_step": 110683, "epoch": 1243, "lr": 6.445288482596018e-05}
{"train_loss": 0.18651247024536133, "global_step": 110684, "epoch": 1243, "lr": 6.445232983636835e-05}
{"train_loss": 0.161210298538208, "global_step": 110685, "epoch": 1243, "lr": 6.44517748448336e-05}
{"train_loss": 0.10215330123901367, "global_step": 110686, "epoch": 1243, "lr": 6.445121985135596e-05}
{"train_loss": 0.0897035226225853, "global_step": 110687, "epoch": 1243, "lr": 6.445066485593553e-05}
{"train_loss": 0.13927316665649414, "global_step": 110688, "epoch": 1243, "lr": 6.44501098585724e-05}
{"train_loss": 0.1570182740688324, "global_step": 110689, "epoch": 1243, "lr": 6.444955485926663e-05}
{"train_loss": 0.20226988196372986, "global_step": 110690, "epoch": 1243, "lr": 6.444899985801828e-05}
{"train_loss": 0.13565261662006378, "global_step": 110691, "epoch": 1243, "lr": 6.444844485482744e-05}
{"train_loss": 0.07632379233837128, "global_step": 110692, "epoch": 1243, "lr": 6.44478898496942e-05}
{"train_loss": 0.1215834692120552, "global_step": 110693, "epoch": 1243, "lr": 6.44473348426186e-05}
{"train_loss": 0.11070054769515991, "global_step": 110694, "epoch": 1243, "lr": 6.444677983360074e-05}
{"train_loss": 0.11546065658330917, "global_step": 110695, "epoch": 1243, "lr": 6.444622482264069e-05}
{"train_loss": 0.1343657523393631, "global_step": 110696, "epoch": 1243, "lr": 6.444566980973851e-05}
{"train_loss": 0.17748281359672546, "global_step": 110697, "epoch": 1243, "lr": 6.444511479489429e-05}
{"train_loss": 0.08829208463430405, "global_step": 110698, "epoch": 1243, "lr": 6.44445597781081e-05}
{"train_loss": 0.14603327214717865, "global_step": 110699, "epoch": 1243, "lr": 6.444400475938002e-05}
{"train_loss": 0.14255517721176147, "global_step": 110700, "epoch": 1243, "lr": 6.444344973871012e-05}
{"train_loss": 0.16750207543373108, "global_step": 110701, "epoch": 1243, "lr": 6.444289471609847e-05}
{"train_loss": 0.15207263827323914, "global_step": 110702, "epoch": 1243, "lr": 6.444233969154513e-05}
{"train_loss": 0.21039307117462158, "global_step": 110703, "epoch": 1243, "lr": 6.444178466505024e-05}
{"train_loss": 0.10630983114242554, "global_step": 110704, "epoch": 1243, "lr": 6.444122963661379e-05}
{"train_loss": 0.15175822377204895, "global_step": 110705, "epoch": 1243, "lr": 6.44406746062359e-05}
{"train_loss": 0.14429616928100586, "global_step": 110706, "epoch": 1243, "lr": 6.444011957391665e-05}
{"train_loss": 0.17164841294288635, "global_step": 110707, "epoch": 1243, "lr": 6.443956453965608e-05}
{"train_loss": 0.2419806867837906, "global_step": 110708, "epoch": 1243, "lr": 6.443900950345431e-05}
{"train_loss": 0.1318834275007248, "global_step": 110709, "epoch": 1243, "lr": 6.443845446531137e-05}
{"train_loss": 0.13103719055652618, "global_step": 110710, "epoch": 1243, "lr": 6.443789942522736e-05}
{"train_loss": 0.2191532701253891, "global_step": 110711, "epoch": 1243, "lr": 6.443734438320237e-05}
{"train_loss": 0.20474229753017426, "global_step": 110712, "epoch": 1243, "lr": 6.443678933923643e-05}
{"train_loss": 0.15813589096069336, "global_step": 110713, "epoch": 1243, "lr": 6.443623429332966e-05}
{"train_loss": 0.10410285741090775, "global_step": 110714, "epoch": 1243, "lr": 6.44356792454821e-05}
{"train_loss": 0.15804440645354517, "global_step": 110715, "epoch": 1243, "lr": 6.443512419569384e-05, "val_loss": 4.9523491859436035}
{"train_loss": 0.13214626908302307, "global_step": 110716, "epoch": 1244, "lr": 6.443456914396496e-05}
{"train_loss": 0.1121709942817688, "global_step": 110717, "epoch": 1244, "lr": 6.443401409029553e-05}
{"train_loss": 0.07250943779945374, "global_step": 110718, "epoch": 1244, "lr": 6.443345903468561e-05}
{"train_loss": 0.17262843251228333, "global_step": 110719, "epoch": 1244, "lr": 6.44329039771353e-05}
{"train_loss": 0.0992589220404625, "global_step": 110720, "epoch": 1244, "lr": 6.443234891764466e-05}
{"train_loss": 0.17298884689807892, "global_step": 110721, "epoch": 1244, "lr": 6.443179385621376e-05}
{"train_loss": 0.11746125668287277, "global_step": 110722, "epoch": 1244, "lr": 6.443123879284268e-05}
{"train_loss": 0.18548177182674408, "global_step": 110723, "epoch": 1244, "lr": 6.443068372753152e-05}
{"train_loss": 0.16860665380954742, "global_step": 110724, "epoch": 1244, "lr": 6.44301286602803e-05}
{"train_loss": 0.17958618700504303, "global_step": 110725, "epoch": 1244, "lr": 6.442957359108914e-05}
{"train_loss": 0.1587187945842743, "global_step": 110726, "epoch": 1244, "lr": 6.44290185199581e-05}
{"train_loss": 0.13871663808822632, "global_step": 110727, "epoch": 1244, "lr": 6.442846344688724e-05}
{"train_loss": 0.14431391656398773, "global_step": 110728, "epoch": 1244, "lr": 6.442790837187667e-05}
{"train_loss": 0.10107770562171936, "global_step": 110729, "epoch": 1244, "lr": 6.442735329492643e-05}
{"train_loss": 0.24534669518470764, "global_step": 110730, "epoch": 1244, "lr": 6.442679821603661e-05}
{"train_loss": 0.15512442588806152, "global_step": 110731, "epoch": 1244, "lr": 6.442624313520729e-05}
{"train_loss": 0.11695335060358047, "global_step": 110732, "epoch": 1244, "lr": 6.442568805243853e-05}
{"train_loss": 0.13911643624305725, "global_step": 110733, "epoch": 1244, "lr": 6.442513296773041e-05}
{"train_loss": 0.10830014944076538, "global_step": 110734, "epoch": 1244, "lr": 6.4424577881083e-05}
{"train_loss": 0.14280733466148376, "global_step": 110735, "epoch": 1244, "lr": 6.44240227924964e-05}
{"train_loss": 0.12714043259620667, "global_step": 110736, "epoch": 1244, "lr": 6.442346770197066e-05}
{"train_loss": 0.19846117496490479, "global_step": 110737, "epoch": 1244, "lr": 6.442291260950586e-05}
{"train_loss": 0.1418699026107788, "global_step": 110738, "epoch": 1244, "lr": 6.442235751510208e-05}
{"train_loss": 0.24956391751766205, "global_step": 110739, "epoch": 1244, "lr": 6.442180241875937e-05}
{"train_loss": 0.11600428819656372, "global_step": 110740, "epoch": 1244, "lr": 6.442124732047784e-05}
{"train_loss": 0.21228444576263428, "global_step": 110741, "epoch": 1244, "lr": 6.442069222025755e-05}
{"train_loss": 0.15170353651046753, "global_step": 110742, "epoch": 1244, "lr": 6.442013711809857e-05}
{"train_loss": 0.10628701746463776, "global_step": 110743, "epoch": 1244, "lr": 6.441958201400098e-05}
{"train_loss": 0.1792411357164383, "global_step": 110744, "epoch": 1244, "lr": 6.441902690796485e-05}
{"train_loss": 0.20293357968330383, "global_step": 110745, "epoch": 1244, "lr": 6.441847179999025e-05}
{"train_loss": 0.11059946566820145, "global_step": 110746, "epoch": 1244, "lr": 6.441791669007728e-05}
{"train_loss": 0.20495454967021942, "global_step": 110747, "epoch": 1244, "lr": 6.441736157822598e-05}
{"train_loss": 0.1831645369529724, "global_step": 110748, "epoch": 1244, "lr": 6.441680646443647e-05}
{"train_loss": 0.21837522089481354, "global_step": 110749, "epoch": 1244, "lr": 6.441625134870876e-05}
{"train_loss": 0.12546047568321228, "global_step": 110750, "epoch": 1244, "lr": 6.441569623104298e-05}
{"train_loss": 0.16921453177928925, "global_step": 110751, "epoch": 1244, "lr": 6.441514111143918e-05}
{"train_loss": 0.1444031000137329, "global_step": 110752, "epoch": 1244, "lr": 6.441458598989744e-05}
{"train_loss": 0.11753039807081223, "global_step": 110753, "epoch": 1244, "lr": 6.441403086641784e-05}
{"train_loss": 0.12661026418209076, "global_step": 110754, "epoch": 1244, "lr": 6.441347574100045e-05}
{"train_loss": 0.16299167275428772, "global_step": 110755, "epoch": 1244, "lr": 6.441292061364534e-05}
{"train_loss": 0.1506858468055725, "global_step": 110756, "epoch": 1244, "lr": 6.441236548435259e-05}
{"train_loss": 0.15343822538852692, "global_step": 110757, "epoch": 1244, "lr": 6.441181035312227e-05}
{"train_loss": 0.1993822157382965, "global_step": 110758, "epoch": 1244, "lr": 6.441125521995445e-05}
{"train_loss": 0.15324623882770538, "global_step": 110759, "epoch": 1244, "lr": 6.441070008484922e-05}
{"train_loss": 0.14138241112232208, "global_step": 110760, "epoch": 1244, "lr": 6.441014494780666e-05}
{"train_loss": 0.1549481898546219, "global_step": 110761, "epoch": 1244, "lr": 6.440958980882682e-05}
{"train_loss": 0.20993083715438843, "global_step": 110762, "epoch": 1244, "lr": 6.440903466790978e-05}
{"train_loss": 0.12509785592556, "global_step": 110763, "epoch": 1244, "lr": 6.440847952505563e-05}
{"train_loss": 0.20266732573509216, "global_step": 110764, "epoch": 1244, "lr": 6.440792438026444e-05}
{"train_loss": 0.2103678286075592, "global_step": 110765, "epoch": 1244, "lr": 6.440736923353626e-05}
{"train_loss": 0.11381030827760696, "global_step": 110766, "epoch": 1244, "lr": 6.44068140848712e-05}
{"train_loss": 0.14790493249893188, "global_step": 110767, "epoch": 1244, "lr": 6.440625893426932e-05}
{"train_loss": 0.14649735391139984, "global_step": 110768, "epoch": 1244, "lr": 6.440570378173069e-05}
{"train_loss": 0.08832783997058868, "global_step": 110769, "epoch": 1244, "lr": 6.440514862725539e-05}
{"train_loss": 0.21053247153759003, "global_step": 110770, "epoch": 1244, "lr": 6.44045934708435e-05}
{"train_loss": 0.17078666388988495, "global_step": 110771, "epoch": 1244, "lr": 6.440403831249508e-05}
{"train_loss": 0.0966309905052185, "global_step": 110772, "epoch": 1244, "lr": 6.44034831522102e-05}
{"train_loss": 0.12566021084785461, "global_step": 110773, "epoch": 1244, "lr": 6.440292798998897e-05}
{"train_loss": 0.11176174134016037, "global_step": 110774, "epoch": 1244, "lr": 6.440237282583143e-05}
{"train_loss": 0.261168509721756, "global_step": 110775, "epoch": 1244, "lr": 6.440181765973767e-05}
{"train_loss": 0.11798971891403198, "global_step": 110776, "epoch": 1244, "lr": 6.440126249170777e-05}
{"train_loss": 0.20380473136901855, "global_step": 110777, "epoch": 1244, "lr": 6.440070732174178e-05}
{"train_loss": 0.1718132495880127, "global_step": 110778, "epoch": 1244, "lr": 6.44001521498398e-05}
{"train_loss": 0.1705268770456314, "global_step": 110779, "epoch": 1244, "lr": 6.439959697600188e-05}
{"train_loss": 0.09624659270048141, "global_step": 110780, "epoch": 1244, "lr": 6.439904180022813e-05}
{"train_loss": 0.09785540401935577, "global_step": 110781, "epoch": 1244, "lr": 6.43984866225186e-05}
{"train_loss": 0.13794183731079102, "global_step": 110782, "epoch": 1244, "lr": 6.439793144287337e-05}
{"train_loss": 0.14819760620594025, "global_step": 110783, "epoch": 1244, "lr": 6.43973762612925e-05}
{"train_loss": 0.19429431855678558, "global_step": 110784, "epoch": 1244, "lr": 6.43968210777761e-05}
{"train_loss": 0.11301784217357635, "global_step": 110785, "epoch": 1244, "lr": 6.439626589232419e-05}
{"train_loss": 0.11403320729732513, "global_step": 110786, "epoch": 1244, "lr": 6.43957107049369e-05}
{"train_loss": 0.12811769545078278, "global_step": 110787, "epoch": 1244, "lr": 6.439515551561428e-05}
{"train_loss": 0.06550942361354828, "global_step": 110788, "epoch": 1244, "lr": 6.439460032435641e-05}
{"train_loss": 0.16495291888713837, "global_step": 110789, "epoch": 1244, "lr": 6.439404513116337e-05}
{"train_loss": 0.08817631006240845, "global_step": 110790, "epoch": 1244, "lr": 6.43934899360352e-05}
{"train_loss": 0.1889132708311081, "global_step": 110791, "epoch": 1244, "lr": 6.4392934738972e-05}
{"train_loss": 0.116836316883564, "global_step": 110792, "epoch": 1244, "lr": 6.439237953997387e-05}
{"train_loss": 0.17275169491767883, "global_step": 110793, "epoch": 1244, "lr": 6.439182433904085e-05}
{"train_loss": 0.12462764978408813, "global_step": 110794, "epoch": 1244, "lr": 6.439126913617302e-05}
{"train_loss": 0.13731345534324646, "global_step": 110795, "epoch": 1244, "lr": 6.439071393137047e-05}
{"train_loss": 0.16200420260429382, "global_step": 110796, "epoch": 1244, "lr": 6.439015872463326e-05}
{"train_loss": 0.15273970365524292, "global_step": 110797, "epoch": 1244, "lr": 6.438960351596147e-05}
{"train_loss": 0.11278387904167175, "global_step": 110798, "epoch": 1244, "lr": 6.438904830535517e-05}
{"train_loss": 0.1732264757156372, "global_step": 110799, "epoch": 1244, "lr": 6.438849309281443e-05}
{"train_loss": 0.1565820276737213, "global_step": 110800, "epoch": 1244, "lr": 6.438793787833934e-05}
{"train_loss": 0.19442537426948547, "global_step": 110801, "epoch": 1244, "lr": 6.438738266192997e-05}
{"train_loss": 0.1482359766960144, "global_step": 110802, "epoch": 1244, "lr": 6.438682744358638e-05}
{"train_loss": 0.10988917946815491, "global_step": 110803, "epoch": 1244, "lr": 6.438627222330868e-05}
{"train_loss": 0.1503363460134924, "global_step": 110804, "epoch": 1244, "lr": 6.43857170010969e-05, "val_loss": 5.1276984214782715}
{"train_loss": 0.15056473016738892, "global_step": 110805, "epoch": 1245, "lr": 6.438516177695114e-05}
{"train_loss": 0.10367944091558456, "global_step": 110806, "epoch": 1245, "lr": 6.438460655087147e-05}
{"train_loss": 0.10004565864801407, "global_step": 110807, "epoch": 1245, "lr": 6.438405132285798e-05}
{"train_loss": 0.11295273900032043, "global_step": 110808, "epoch": 1245, "lr": 6.438349609291071e-05}
{"train_loss": 0.15163584053516388, "global_step": 110809, "epoch": 1245, "lr": 6.438294086102976e-05}
{"train_loss": 0.1347983181476593, "global_step": 110810, "epoch": 1245, "lr": 6.438238562721519e-05}
{"train_loss": 0.19452576339244843, "global_step": 110811, "epoch": 1245, "lr": 6.43818303914671e-05}
{"train_loss": 0.12251016497612, "global_step": 110812, "epoch": 1245, "lr": 6.438127515378553e-05}
{"train_loss": 0.24342843890190125, "global_step": 110813, "epoch": 1245, "lr": 6.43807199141706e-05}
{"train_loss": 0.12379278987646103, "global_step": 110814, "epoch": 1245, "lr": 6.438016467262233e-05}
{"train_loss": 0.13265100121498108, "global_step": 110815, "epoch": 1245, "lr": 6.437960942914085e-05}
{"train_loss": 0.13093098998069763, "global_step": 110816, "epoch": 1245, "lr": 6.437905418372618e-05}
{"train_loss": 0.24412721395492554, "global_step": 110817, "epoch": 1245, "lr": 6.437849893637844e-05}
{"train_loss": 0.15137091279029846, "global_step": 110818, "epoch": 1245, "lr": 6.437794368709767e-05}
{"train_loss": 0.17259375751018524, "global_step": 110819, "epoch": 1245, "lr": 6.437738843588397e-05}
{"train_loss": 0.11245573312044144, "global_step": 110820, "epoch": 1245, "lr": 6.43768331827374e-05}
{"train_loss": 0.15525315701961517, "global_step": 110821, "epoch": 1245, "lr": 6.437627792765804e-05}
{"train_loss": 0.0586063377559185, "global_step": 110822, "epoch": 1245, "lr": 6.437572267064597e-05}
{"train_loss": 0.1827322393655777, "global_step": 110823, "epoch": 1245, "lr": 6.437516741170126e-05}
{"train_loss": 0.16356131434440613, "global_step": 110824, "epoch": 1245, "lr": 6.437461215082399e-05}
{"train_loss": 0.12210728973150253, "global_step": 110825, "epoch": 1245, "lr": 6.437405688801423e-05}
{"train_loss": 0.16630686819553375, "global_step": 110826, "epoch": 1245, "lr": 6.437350162327204e-05}
{"train_loss": 0.09519962221384048, "global_step": 110827, "epoch": 1245, "lr": 6.437294635659751e-05}
{"train_loss": 0.19154928624629974, "global_step": 110828, "epoch": 1245, "lr": 6.437239108799071e-05}
{"train_loss": 0.15667985379695892, "global_step": 110829, "epoch": 1245, "lr": 6.437183581745172e-05}
{"train_loss": 0.09452991187572479, "global_step": 110830, "epoch": 1245, "lr": 6.437128054498062e-05}
{"train_loss": 0.16468864679336548, "global_step": 110831, "epoch": 1245, "lr": 6.437072527057746e-05}
{"train_loss": 0.1494855433702469, "global_step": 110832, "epoch": 1245, "lr": 6.437016999424235e-05}
{"train_loss": 0.18252989649772644, "global_step": 110833, "epoch": 1245, "lr": 6.436961471597534e-05}
{"train_loss": 0.16943375766277313, "global_step": 110834, "epoch": 1245, "lr": 6.43690594357765e-05}
{"train_loss": 0.19482016563415527, "global_step": 110835, "epoch": 1245, "lr": 6.436850415364593e-05}
{"train_loss": 0.07336914539337158, "global_step": 110836, "epoch": 1245, "lr": 6.436794886958367e-05}
{"train_loss": 0.2003268152475357, "global_step": 110837, "epoch": 1245, "lr": 6.436739358358982e-05}
{"train_loss": 0.13423921167850494, "global_step": 110838, "epoch": 1245, "lr": 6.436683829566445e-05}
{"train_loss": 0.18989069759845734, "global_step": 110839, "epoch": 1245, "lr": 6.436628300580764e-05}
{"train_loss": 0.16425453126430511, "global_step": 110840, "epoch": 1245, "lr": 6.436572771401947e-05}
{"train_loss": 0.18549376726150513, "global_step": 110841, "epoch": 1245, "lr": 6.43651724203e-05}
{"train_loss": 0.16190940141677856, "global_step": 110842, "epoch": 1245, "lr": 6.43646171246493e-05}
{"train_loss": 0.1898939311504364, "global_step": 110843, "epoch": 1245, "lr": 6.436406182706744e-05}
{"train_loss": 0.17455363273620605, "global_step": 110844, "epoch": 1245, "lr": 6.436350652755451e-05}
{"train_loss": 0.15075835585594177, "global_step": 110845, "epoch": 1245, "lr": 6.43629512261106e-05}
{"train_loss": 0.12295123934745789, "global_step": 110846, "epoch": 1245, "lr": 6.436239592273575e-05}
{"train_loss": 0.1677730530500412, "global_step": 110847, "epoch": 1245, "lr": 6.436184061743006e-05}
{"train_loss": 0.13438402116298676, "global_step": 110848, "epoch": 1245, "lr": 6.436128531019359e-05}
{"train_loss": 0.10049175471067429, "global_step": 110849, "epoch": 1245, "lr": 6.436073000102643e-05}
{"train_loss": 0.09517188370227814, "global_step": 110850, "epoch": 1245, "lr": 6.436017468992863e-05}
{"train_loss": 0.09525737166404724, "global_step": 110851, "epoch": 1245, "lr": 6.43596193769003e-05}
{"train_loss": 0.137277290225029, "global_step": 110852, "epoch": 1245, "lr": 6.435906406194147e-05}
{"train_loss": 0.09625605493783951, "global_step": 110853, "epoch": 1245, "lr": 6.435850874505225e-05}
{"train_loss": 0.13283590972423553, "global_step": 110854, "epoch": 1245, "lr": 6.435795342623269e-05}
{"train_loss": 0.19124481081962585, "global_step": 110855, "epoch": 1245, "lr": 6.435739810548289e-05}
{"train_loss": 0.10787540674209595, "global_step": 110856, "epoch": 1245, "lr": 6.435684278280291e-05}
{"train_loss": 0.14071208238601685, "global_step": 110857, "epoch": 1245, "lr": 6.435628745819285e-05}
{"train_loss": 0.15586602687835693, "global_step": 110858, "epoch": 1245, "lr": 6.435573213165274e-05}
{"train_loss": 0.094243623316288, "global_step": 110859, "epoch": 1245, "lr": 6.435517680318267e-05}
{"train_loss": 0.14901626110076904, "global_step": 110860, "epoch": 1245, "lr": 6.435462147278274e-05}
{"train_loss": 0.15260323882102966, "global_step": 110861, "epoch": 1245, "lr": 6.4354066140453e-05}
{"train_loss": 0.09893191605806351, "global_step": 110862, "epoch": 1245, "lr": 6.435351080619354e-05}
{"train_loss": 0.1592901051044464, "global_step": 110863, "epoch": 1245, "lr": 6.435295547000441e-05}
{"train_loss": 0.24290868639945984, "global_step": 110864, "epoch": 1245, "lr": 6.435240013188571e-05}
{"train_loss": 0.14671960473060608, "global_step": 110865, "epoch": 1245, "lr": 6.435184479183748e-05}
{"train_loss": 0.10270741581916809, "global_step": 110866, "epoch": 1245, "lr": 6.435128944985985e-05}
{"train_loss": 0.10597061365842819, "global_step": 110867, "epoch": 1245, "lr": 6.435073410595284e-05}
{"train_loss": 0.18803364038467407, "global_step": 110868, "epoch": 1245, "lr": 6.435017876011658e-05}
{"train_loss": 0.15489259362220764, "global_step": 110869, "epoch": 1245, "lr": 6.43496234123511e-05}
{"train_loss": 0.1514967828989029, "global_step": 110870, "epoch": 1245, "lr": 6.434906806265647e-05}
{"train_loss": 0.06605569273233414, "global_step": 110871, "epoch": 1245, "lr": 6.43485127110328e-05}
{"train_loss": 0.16011545062065125, "global_step": 110872, "epoch": 1245, "lr": 6.434795735748014e-05}
{"train_loss": 0.2514714300632477, "global_step": 110873, "epoch": 1245, "lr": 6.434740200199856e-05}
{"train_loss": 0.15753288567066193, "global_step": 110874, "epoch": 1245, "lr": 6.434684664458816e-05}
{"train_loss": 0.18349964916706085, "global_step": 110875, "epoch": 1245, "lr": 6.434629128524901e-05}
{"train_loss": 0.1145825982093811, "global_step": 110876, "epoch": 1245, "lr": 6.434573592398116e-05}
{"train_loss": 0.07113796472549438, "global_step": 110877, "epoch": 1245, "lr": 6.434518056078472e-05}
{"train_loss": 0.18621011078357697, "global_step": 110878, "epoch": 1245, "lr": 6.434462519565973e-05}
{"train_loss": 0.08090303093194962, "global_step": 110879, "epoch": 1245, "lr": 6.434406982860628e-05}
{"train_loss": 0.17190337181091309, "global_step": 110880, "epoch": 1245, "lr": 6.434351445962447e-05}
{"train_loss": 0.21749505400657654, "global_step": 110881, "epoch": 1245, "lr": 6.434295908871433e-05}
{"train_loss": 0.14155343174934387, "global_step": 110882, "epoch": 1245, "lr": 6.434240371587594e-05}
{"train_loss": 0.12146161496639252, "global_step": 110883, "epoch": 1245, "lr": 6.434184834110941e-05}
{"train_loss": 0.16010136902332306, "global_step": 110884, "epoch": 1245, "lr": 6.434129296441478e-05}
{"train_loss": 0.19885356724262238, "global_step": 110885, "epoch": 1245, "lr": 6.434073758579215e-05}
{"train_loss": 0.18651559948921204, "global_step": 110886, "epoch": 1245, "lr": 6.434018220524156e-05}
{"train_loss": 0.11846490949392319, "global_step": 110887, "epoch": 1245, "lr": 6.433962682276314e-05}
{"train_loss": 0.14131960272789001, "global_step": 110888, "epoch": 1245, "lr": 6.433907143835691e-05}
{"train_loss": 0.09402479976415634, "global_step": 110889, "epoch": 1245, "lr": 6.433851605202297e-05}
{"train_loss": 0.15149323642253876, "global_step": 110890, "epoch": 1245, "lr": 6.433796066376138e-05}
{"train_loss": 0.2998853027820587, "global_step": 110891, "epoch": 1245, "lr": 6.433740527357225e-05}
{"train_loss": 0.17115795612335205, "global_step": 110892, "epoch": 1245, "lr": 6.43368498814556e-05}
{"train_loss": 0.1487156515627095, "global_step": 110893, "epoch": 1245, "lr": 6.433629448741156e-05, "val_loss": 5.101164817810059, "train_action_mse_error": 21.89933204650879}
{"train_loss": 0.12901680171489716, "global_step": 110894, "epoch": 1246, "lr": 6.433573909144018e-05}
{"train_loss": 0.1492592692375183, "global_step": 110895, "epoch": 1246, "lr": 6.433518369354152e-05}
{"train_loss": 0.16999657452106476, "global_step": 110896, "epoch": 1246, "lr": 6.433462829371568e-05}
{"train_loss": 0.17521226406097412, "global_step": 110897, "epoch": 1246, "lr": 6.433407289196271e-05}
{"train_loss": 0.11594036966562271, "global_step": 110898, "epoch": 1246, "lr": 6.433351748828272e-05}
{"train_loss": 0.11896560341119766, "global_step": 110899, "epoch": 1246, "lr": 6.433296208267577e-05}
{"train_loss": 0.104985810816288, "global_step": 110900, "epoch": 1246, "lr": 6.43324066751419e-05}
{"train_loss": 0.210773766040802, "global_step": 110901, "epoch": 1246, "lr": 6.433185126568123e-05}
{"train_loss": 0.19208350777626038, "global_step": 110902, "epoch": 1246, "lr": 6.43312958542938e-05}
{"train_loss": 0.10184833407402039, "global_step": 110903, "epoch": 1246, "lr": 6.433074044097973e-05}
{"train_loss": 0.11962847411632538, "global_step": 110904, "epoch": 1246, "lr": 6.433018502573904e-05}
{"train_loss": 0.17690996825695038, "global_step": 110905, "epoch": 1246, "lr": 6.432962960857185e-05}
{"train_loss": 0.2247556746006012, "global_step": 110906, "epoch": 1246, "lr": 6.43290741894782e-05}
{"train_loss": 0.101336270570755, "global_step": 110907, "epoch": 1246, "lr": 6.432851876845819e-05}
{"train_loss": 0.123605877161026, "global_step": 110908, "epoch": 1246, "lr": 6.432796334551189e-05}
{"train_loss": 0.15203827619552612, "global_step": 110909, "epoch": 1246, "lr": 6.432740792063936e-05}
{"train_loss": 0.18608257174491882, "global_step": 110910, "epoch": 1246, "lr": 6.432685249384068e-05}
{"train_loss": 0.1282544583082199, "global_step": 110911, "epoch": 1246, "lr": 6.432629706511592e-05}
{"train_loss": 0.20585815608501434, "global_step": 110912, "epoch": 1246, "lr": 6.432574163446519e-05}
{"train_loss": 0.16471795737743378, "global_step": 110913, "epoch": 1246, "lr": 6.432518620188853e-05}
{"train_loss": 0.23481307923793793, "global_step": 110914, "epoch": 1246, "lr": 6.432463076738601e-05}
{"train_loss": 0.14247392117977142, "global_step": 110915, "epoch": 1246, "lr": 6.432407533095774e-05}
{"train_loss": 0.07652547210454941, "global_step": 110916, "epoch": 1246, "lr": 6.432351989260377e-05}
{"train_loss": 0.18833646178245544, "global_step": 110917, "epoch": 1246, "lr": 6.432296445232418e-05}
{"train_loss": 0.12130267173051834, "global_step": 110918, "epoch": 1246, "lr": 6.432240901011903e-05}
{"train_loss": 0.08277259021997452, "global_step": 110919, "epoch": 1246, "lr": 6.432185356598841e-05}
{"train_loss": 0.14971446990966797, "global_step": 110920, "epoch": 1246, "lr": 6.43212981199324e-05}
{"train_loss": 0.14461642503738403, "global_step": 110921, "epoch": 1246, "lr": 6.432074267195104e-05}
{"train_loss": 0.15864473581314087, "global_step": 110922, "epoch": 1246, "lr": 6.432018722204446e-05}
{"train_loss": 0.13888496160507202, "global_step": 110923, "epoch": 1246, "lr": 6.43196317702127e-05}
{"train_loss": 0.1501011699438095, "global_step": 110924, "epoch": 1246, "lr": 6.431907631645583e-05}
{"train_loss": 0.14636771380901337, "global_step": 110925, "epoch": 1246, "lr": 6.431852086077393e-05}
{"train_loss": 0.11238875985145569, "global_step": 110926, "epoch": 1246, "lr": 6.431796540316709e-05}
{"train_loss": 0.11245241016149521, "global_step": 110927, "epoch": 1246, "lr": 6.431740994363538e-05}
{"train_loss": 0.17052973806858063, "global_step": 110928, "epoch": 1246, "lr": 6.431685448217883e-05}
{"train_loss": 0.20910201966762543, "global_step": 110929, "epoch": 1246, "lr": 6.431629901879758e-05}
{"train_loss": 0.0963553935289383, "global_step": 110930, "epoch": 1246, "lr": 6.43157435534917e-05}
{"train_loss": 0.14408910274505615, "global_step": 110931, "epoch": 1246, "lr": 6.431518808626123e-05}
{"train_loss": 0.11173002421855927, "global_step": 110932, "epoch": 1246, "lr": 6.431463261710624e-05}
{"train_loss": 0.1665843427181244, "global_step": 110933, "epoch": 1246, "lr": 6.431407714602684e-05}
{"train_loss": 0.07182725518941879, "global_step": 110934, "epoch": 1246, "lr": 6.431352167302306e-05}
{"train_loss": 0.20757795870304108, "global_step": 110935, "epoch": 1246, "lr": 6.431296619809503e-05}
{"train_loss": 0.1732284128665924, "global_step": 110936, "epoch": 1246, "lr": 6.43124107212428e-05}
{"train_loss": 0.10815552622079849, "global_step": 110937, "epoch": 1246, "lr": 6.431185524246642e-05}
{"train_loss": 0.17832492291927338, "global_step": 110938, "epoch": 1246, "lr": 6.4311299761766e-05}
{"train_loss": 0.10461518168449402, "global_step": 110939, "epoch": 1246, "lr": 6.431074427914161e-05}
{"train_loss": 0.11624801158905029, "global_step": 110940, "epoch": 1246, "lr": 6.43101887945933e-05}
{"train_loss": 0.1289307326078415, "global_step": 110941, "epoch": 1246, "lr": 6.430963330812115e-05}
{"train_loss": 0.19243597984313965, "global_step": 110942, "epoch": 1246, "lr": 6.430907781972526e-05}
{"train_loss": 0.13626737892627716, "global_step": 110943, "epoch": 1246, "lr": 6.430852232940567e-05}
{"train_loss": 0.1852446049451828, "global_step": 110944, "epoch": 1246, "lr": 6.430796683716249e-05}
{"train_loss": 0.15197254717350006, "global_step": 110945, "epoch": 1246, "lr": 6.430741134299578e-05}
{"train_loss": 0.14234775304794312, "global_step": 110946, "epoch": 1246, "lr": 6.430685584690558e-05}
{"train_loss": 0.17354267835617065, "global_step": 110947, "epoch": 1246, "lr": 6.430630034889204e-05}
{"train_loss": 0.11256594210863113, "global_step": 110948, "epoch": 1246, "lr": 6.430574484895518e-05}
{"train_loss": 0.26436737179756165, "global_step": 110949, "epoch": 1246, "lr": 6.430518934709508e-05}
{"train_loss": 0.18270380795001984, "global_step": 110950, "epoch": 1246, "lr": 6.430463384331182e-05}
{"train_loss": 0.1762513518333435, "global_step": 110951, "epoch": 1246, "lr": 6.430407833760549e-05}
{"train_loss": 0.1657032072544098, "global_step": 110952, "epoch": 1246, "lr": 6.430352282997614e-05}
{"train_loss": 0.07965389639139175, "global_step": 110953, "epoch": 1246, "lr": 6.430296732042386e-05}
{"train_loss": 0.14479497075080872, "global_step": 110954, "epoch": 1246, "lr": 6.430241180894871e-05}
{"train_loss": 0.16908463835716248, "global_step": 110955, "epoch": 1246, "lr": 6.430185629555079e-05}
{"train_loss": 0.17147493362426758, "global_step": 110956, "epoch": 1246, "lr": 6.430130078023016e-05}
{"train_loss": 0.08357637375593185, "global_step": 110957, "epoch": 1246, "lr": 6.43007452629869e-05}
{"train_loss": 0.23288670182228088, "global_step": 110958, "epoch": 1246, "lr": 6.430018974382106e-05}
{"train_loss": 0.11809443682432175, "global_step": 110959, "epoch": 1246, "lr": 6.429963422273274e-05}
{"train_loss": 0.1318841129541397, "global_step": 110960, "epoch": 1246, "lr": 6.429907869972202e-05}
{"train_loss": 0.14023834466934204, "global_step": 110961, "epoch": 1246, "lr": 6.429852317478895e-05}
{"train_loss": 0.13613685965538025, "global_step": 110962, "epoch": 1246, "lr": 6.429796764793361e-05}
{"train_loss": 0.18681518733501434, "global_step": 110963, "epoch": 1246, "lr": 6.42974121191561e-05}
{"train_loss": 0.17777025699615479, "global_step": 110964, "epoch": 1246, "lr": 6.429685658845645e-05}
{"train_loss": 0.13432419300079346, "global_step": 110965, "epoch": 1246, "lr": 6.42963010558348e-05}
{"train_loss": 0.1142880991101265, "global_step": 110966, "epoch": 1246, "lr": 6.429574552129117e-05}
{"train_loss": 0.11954334378242493, "global_step": 110967, "epoch": 1246, "lr": 6.429518998482565e-05}
{"train_loss": 0.10850533843040466, "global_step": 110968, "epoch": 1246, "lr": 6.42946344464383e-05}
{"train_loss": 0.10231121629476547, "global_step": 110969, "epoch": 1246, "lr": 6.429407890612923e-05}
{"train_loss": 0.13261428475379944, "global_step": 110970, "epoch": 1246, "lr": 6.429352336389849e-05}
{"train_loss": 0.13564865291118622, "global_step": 110971, "epoch": 1246, "lr": 6.429296781974617e-05}
{"train_loss": 0.1771775782108307, "global_step": 110972, "epoch": 1246, "lr": 6.429241227367231e-05}
{"train_loss": 0.15879610180854797, "global_step": 110973, "epoch": 1246, "lr": 6.429185672567704e-05}
{"train_loss": 0.1393568217754364, "global_step": 110974, "epoch": 1246, "lr": 6.429130117576038e-05}
{"train_loss": 0.14987851679325104, "global_step": 110975, "epoch": 1246, "lr": 6.429074562392245e-05}
{"train_loss": 0.13298466801643372, "global_step": 110976, "epoch": 1246, "lr": 6.429019007016329e-05}
{"train_loss": 0.0944204330444336, "global_step": 110977, "epoch": 1246, "lr": 6.4289634514483e-05}
{"train_loss": 0.12450136244297028, "global_step": 110978, "epoch": 1246, "lr": 6.428907895688164e-05}
{"train_loss": 0.18250024318695068, "global_step": 110979, "epoch": 1246, "lr": 6.428852339735927e-05}
{"train_loss": 0.12421580404043198, "global_step": 110980, "epoch": 1246, "lr": 6.4287967835916e-05}
{"train_loss": 0.0888039618730545, "global_step": 110981, "epoch": 1246, "lr": 6.428741227255188e-05}
{"train_loss": 0.14736446503842815, "global_step": 110982, "epoch": 1246, "lr": 6.428685670726698e-05, "val_loss": 4.958509922027588}
{"train_loss": 0.08930660039186478, "global_step": 110983, "epoch": 1247, "lr": 6.428630114006139e-05}
{"train_loss": 0.09028011560440063, "global_step": 110984, "epoch": 1247, "lr": 6.428574557093519e-05}
{"train_loss": 0.13582907617092133, "global_step": 110985, "epoch": 1247, "lr": 6.428518999988843e-05}
{"train_loss": 0.1412501037120819, "global_step": 110986, "epoch": 1247, "lr": 6.428463442692123e-05}
{"train_loss": 0.08584290742874146, "global_step": 110987, "epoch": 1247, "lr": 6.428407885203361e-05}
{"train_loss": 0.10576888918876648, "global_step": 110988, "epoch": 1247, "lr": 6.428352327522569e-05}
{"train_loss": 0.17559821903705597, "global_step": 110989, "epoch": 1247, "lr": 6.428296769649751e-05}
{"train_loss": 0.10764381289482117, "global_step": 110990, "epoch": 1247, "lr": 6.428241211584916e-05}
{"train_loss": 0.14261774718761444, "global_step": 110991, "epoch": 1247, "lr": 6.428185653328072e-05}
{"train_loss": 0.1925136148929596, "global_step": 110992, "epoch": 1247, "lr": 6.428130094879226e-05}
{"train_loss": 0.1649843156337738, "global_step": 110993, "epoch": 1247, "lr": 6.428074536238385e-05}
{"train_loss": 0.13783003389835358, "global_step": 110994, "epoch": 1247, "lr": 6.428018977405556e-05}
{"train_loss": 0.15351556241512299, "global_step": 110995, "epoch": 1247, "lr": 6.427963418380746e-05}
{"train_loss": 0.1723059117794037, "global_step": 110996, "epoch": 1247, "lr": 6.427907859163968e-05}
{"train_loss": 0.1360219120979309, "global_step": 110997, "epoch": 1247, "lr": 6.427852299755223e-05}
{"train_loss": 0.09213560819625854, "global_step": 110998, "epoch": 1247, "lr": 6.42779674015452e-05}
{"train_loss": 0.17428527772426605, "global_step": 110999, "epoch": 1247, "lr": 6.427741180361868e-05}
{"train_loss": 0.08832729607820511, "global_step": 111000, "epoch": 1247, "lr": 6.427685620377272e-05}
{"train_loss": 0.147182896733284, "global_step": 111001, "epoch": 1247, "lr": 6.427630060200742e-05}
{"train_loss": 0.16902793943881989, "global_step": 111002, "epoch": 1247, "lr": 6.427574499832286e-05}
{"train_loss": 0.14811260998249054, "global_step": 111003, "epoch": 1247, "lr": 6.427518939271907e-05}
{"train_loss": 0.19803008437156677, "global_step": 111004, "epoch": 1247, "lr": 6.427463378519617e-05}
{"train_loss": 0.13968895375728607, "global_step": 111005, "epoch": 1247, "lr": 6.427407817575423e-05}
{"train_loss": 0.22619785368442535, "global_step": 111006, "epoch": 1247, "lr": 6.42735225643933e-05}
{"train_loss": 0.16386136412620544, "global_step": 111007, "epoch": 1247, "lr": 6.427296695111348e-05}
{"train_loss": 0.3027457892894745, "global_step": 111008, "epoch": 1247, "lr": 6.427241133591482e-05}
{"train_loss": 0.16869914531707764, "global_step": 111009, "epoch": 1247, "lr": 6.427185571879741e-05}
{"train_loss": 0.1557457149028778, "global_step": 111010, "epoch": 1247, "lr": 6.427130009976135e-05}
{"train_loss": 0.09843602031469345, "global_step": 111011, "epoch": 1247, "lr": 6.427074447880665e-05}
{"train_loss": 0.06430625915527344, "global_step": 111012, "epoch": 1247, "lr": 6.427018885593345e-05}
{"train_loss": 0.1119086965918541, "global_step": 111013, "epoch": 1247, "lr": 6.426963323114178e-05}
{"train_loss": 0.1194566935300827, "global_step": 111014, "epoch": 1247, "lr": 6.426907760443174e-05}
{"train_loss": 0.13105064630508423, "global_step": 111015, "epoch": 1247, "lr": 6.42685219758034e-05}
{"train_loss": 0.1197093054652214, "global_step": 111016, "epoch": 1247, "lr": 6.426796634525683e-05}
{"train_loss": 0.15368333458900452, "global_step": 111017, "epoch": 1247, "lr": 6.42674107127921e-05}
{"train_loss": 0.14071787893772125, "global_step": 111018, "epoch": 1247, "lr": 6.42668550784093e-05}
{"train_loss": 0.14975430071353912, "global_step": 111019, "epoch": 1247, "lr": 6.426629944210846e-05}
{"train_loss": 0.12708675861358643, "global_step": 111020, "epoch": 1247, "lr": 6.426574380388972e-05}
{"train_loss": 0.13515055179595947, "global_step": 111021, "epoch": 1247, "lr": 6.426518816375313e-05}
{"train_loss": 0.08978095650672913, "global_step": 111022, "epoch": 1247, "lr": 6.426463252169875e-05}
{"train_loss": 0.17156018316745758, "global_step": 111023, "epoch": 1247, "lr": 6.426407687772666e-05}
{"train_loss": 0.2020275592803955, "global_step": 111024, "epoch": 1247, "lr": 6.426352123183695e-05}
{"train_loss": 0.1753423810005188, "global_step": 111025, "epoch": 1247, "lr": 6.426296558402967e-05}
{"train_loss": 0.0746719017624855, "global_step": 111026, "epoch": 1247, "lr": 6.426240993430493e-05}
{"train_loss": 0.13897906243801117, "global_step": 111027, "epoch": 1247, "lr": 6.426185428266276e-05}
{"train_loss": 0.13335740566253662, "global_step": 111028, "epoch": 1247, "lr": 6.426129862910326e-05}
{"train_loss": 0.13137482106685638, "global_step": 111029, "epoch": 1247, "lr": 6.426074297362652e-05}
{"train_loss": 0.14810165762901306, "global_step": 111030, "epoch": 1247, "lr": 6.426018731623258e-05}
{"train_loss": 0.18514502048492432, "global_step": 111031, "epoch": 1247, "lr": 6.425963165692153e-05}
{"train_loss": 0.15045978128910065, "global_step": 111032, "epoch": 1247, "lr": 6.425907599569346e-05}
{"train_loss": 0.1468486487865448, "global_step": 111033, "epoch": 1247, "lr": 6.425852033254842e-05}
{"train_loss": 0.17054244875907898, "global_step": 111034, "epoch": 1247, "lr": 6.425796466748651e-05}
{"train_loss": 0.12494058907032013, "global_step": 111035, "epoch": 1247, "lr": 6.425740900050778e-05}
{"train_loss": 0.17229732871055603, "global_step": 111036, "epoch": 1247, "lr": 6.425685333161232e-05}
{"train_loss": 0.18279072642326355, "global_step": 111037, "epoch": 1247, "lr": 6.425629766080019e-05}
{"train_loss": 0.09975354373455048, "global_step": 111038, "epoch": 1247, "lr": 6.425574198807148e-05}
{"train_loss": 0.1260860711336136, "global_step": 111039, "epoch": 1247, "lr": 6.425518631342625e-05}
{"train_loss": 0.12716105580329895, "global_step": 111040, "epoch": 1247, "lr": 6.425463063686459e-05}
{"train_loss": 0.0910564661026001, "global_step": 111041, "epoch": 1247, "lr": 6.425407495838658e-05}
{"train_loss": 0.12311147153377533, "global_step": 111042, "epoch": 1247, "lr": 6.425351927799227e-05}
{"train_loss": 0.11104623228311539, "global_step": 111043, "epoch": 1247, "lr": 6.425296359568176e-05}
{"train_loss": 0.10475257784128189, "global_step": 111044, "epoch": 1247, "lr": 6.425240791145511e-05}
{"train_loss": 0.14544545114040375, "global_step": 111045, "epoch": 1247, "lr": 6.425185222531238e-05}
{"train_loss": 0.18945461511611938, "global_step": 111046, "epoch": 1247, "lr": 6.425129653725367e-05}
{"train_loss": 0.10568661987781525, "global_step": 111047, "epoch": 1247, "lr": 6.425074084727906e-05}
{"train_loss": 0.17562201619148254, "global_step": 111048, "epoch": 1247, "lr": 6.42501851553886e-05}
{"train_loss": 0.14910276234149933, "global_step": 111049, "epoch": 1247, "lr": 6.424962946158239e-05}
{"train_loss": 0.12864938378334045, "global_step": 111050, "epoch": 1247, "lr": 6.424907376586047e-05}
{"train_loss": 0.10749533772468567, "global_step": 111051, "epoch": 1247, "lr": 6.424851806822295e-05}
{"train_loss": 0.1632499247789383, "global_step": 111052, "epoch": 1247, "lr": 6.424796236866989e-05}
{"train_loss": 0.14964842796325684, "global_step": 111053, "epoch": 1247, "lr": 6.424740666720136e-05}
{"train_loss": 0.12715260684490204, "global_step": 111054, "epoch": 1247, "lr": 6.424685096381745e-05}
{"train_loss": 0.1461271196603775, "global_step": 111055, "epoch": 1247, "lr": 6.42462952585182e-05}
{"train_loss": 0.15104983747005463, "global_step": 111056, "epoch": 1247, "lr": 6.424573955130373e-05}
{"train_loss": 0.07391779869794846, "global_step": 111057, "epoch": 1247, "lr": 6.424518384217409e-05}
{"train_loss": 0.1393425613641739, "global_step": 111058, "epoch": 1247, "lr": 6.424462813112936e-05}
{"train_loss": 0.1124851331114769, "global_step": 111059, "epoch": 1247, "lr": 6.424407241816961e-05}
{"train_loss": 0.13611681759357452, "global_step": 111060, "epoch": 1247, "lr": 6.424351670329493e-05}
{"train_loss": 0.15506093204021454, "global_step": 111061, "epoch": 1247, "lr": 6.424296098650536e-05}
{"train_loss": 0.13297705352306366, "global_step": 111062, "epoch": 1247, "lr": 6.4242405267801e-05}
{"train_loss": 0.1057053878903389, "global_step": 111063, "epoch": 1247, "lr": 6.424184954718195e-05}
{"train_loss": 0.1575537919998169, "global_step": 111064, "epoch": 1247, "lr": 6.424129382464823e-05}
{"train_loss": 0.15070873498916626, "global_step": 111065, "epoch": 1247, "lr": 6.424073810019996e-05}
{"train_loss": 0.15979579091072083, "global_step": 111066, "epoch": 1247, "lr": 6.424018237383719e-05}
{"train_loss": 0.17854580283164978, "global_step": 111067, "epoch": 1247, "lr": 6.423962664555999e-05}
{"train_loss": 0.13438796997070312, "global_step": 111068, "epoch": 1247, "lr": 6.423907091536846e-05}
{"train_loss": 0.15670396387577057, "global_step": 111069, "epoch": 1247, "lr": 6.423851518326265e-05}
{"train_loss": 0.14801427721977234, "global_step": 111070, "epoch": 1247, "lr": 6.423795944924264e-05}
{"train_loss": 0.1414385689945703, "global_step": 111071, "epoch": 1247, "lr": 6.423740371330853e-05, "val_loss": 5.095163822174072}
{"train_loss": 0.15418864786624908, "global_step": 111072, "epoch": 1248, "lr": 6.423684797546037e-05}
{"train_loss": 0.24345798790454865, "global_step": 111073, "epoch": 1248, "lr": 6.423629223569823e-05}
{"train_loss": 0.18679338693618774, "global_step": 111074, "epoch": 1248, "lr": 6.423573649402219e-05}
{"train_loss": 0.18299657106399536, "global_step": 111075, "epoch": 1248, "lr": 6.423518075043235e-05}
{"train_loss": 0.1385810226202011, "global_step": 111076, "epoch": 1248, "lr": 6.423462500492873e-05}
{"train_loss": 0.1639561504125595, "global_step": 111077, "epoch": 1248, "lr": 6.423406925751148e-05}
{"train_loss": 0.11209318786859512, "global_step": 111078, "epoch": 1248, "lr": 6.42335135081806e-05}
{"train_loss": 0.16963855922222137, "global_step": 111079, "epoch": 1248, "lr": 6.423295775693621e-05}
{"train_loss": 0.1414417177438736, "global_step": 111080, "epoch": 1248, "lr": 6.423240200377837e-05}
{"train_loss": 0.191228449344635, "global_step": 111081, "epoch": 1248, "lr": 6.423184624870715e-05}
{"train_loss": 0.24053114652633667, "global_step": 111082, "epoch": 1248, "lr": 6.423129049172265e-05}
{"train_loss": 0.10662536323070526, "global_step": 111083, "epoch": 1248, "lr": 6.42307347328249e-05}
{"train_loss": 0.13163535296916962, "global_step": 111084, "epoch": 1248, "lr": 6.423017897201402e-05}
{"train_loss": 0.12179701775312424, "global_step": 111085, "epoch": 1248, "lr": 6.422962320929007e-05}
{"train_loss": 0.15388736128807068, "global_step": 111086, "epoch": 1248, "lr": 6.422906744465312e-05}
{"train_loss": 0.1776018738746643, "global_step": 111087, "epoch": 1248, "lr": 6.422851167810323e-05}
{"train_loss": 0.10152651369571686, "global_step": 111088, "epoch": 1248, "lr": 6.42279559096405e-05}
{"train_loss": 0.11265020072460175, "global_step": 111089, "epoch": 1248, "lr": 6.422740013926499e-05}
{"train_loss": 0.09524639695882797, "global_step": 111090, "epoch": 1248, "lr": 6.422684436697679e-05}
{"train_loss": 0.08509473502635956, "global_step": 111091, "epoch": 1248, "lr": 6.422628859277596e-05}
{"train_loss": 0.1637675166130066, "global_step": 111092, "epoch": 1248, "lr": 6.422573281666258e-05}
{"train_loss": 0.16801932454109192, "global_step": 111093, "epoch": 1248, "lr": 6.422517703863671e-05}
{"train_loss": 0.15913593769073486, "global_step": 111094, "epoch": 1248, "lr": 6.422462125869844e-05}
{"train_loss": 0.12191610038280487, "global_step": 111095, "epoch": 1248, "lr": 6.422406547684786e-05}
{"train_loss": 0.2135554999113083, "global_step": 111096, "epoch": 1248, "lr": 6.422350969308502e-05}
{"train_loss": 0.15432274341583252, "global_step": 111097, "epoch": 1248, "lr": 6.422295390741e-05}
{"train_loss": 0.1559191644191742, "global_step": 111098, "epoch": 1248, "lr": 6.422239811982288e-05}
{"train_loss": 0.16381429135799408, "global_step": 111099, "epoch": 1248, "lr": 6.422184233032373e-05}
{"train_loss": 0.10867846012115479, "global_step": 111100, "epoch": 1248, "lr": 6.422128653891263e-05}
{"train_loss": 0.17899249494075775, "global_step": 111101, "epoch": 1248, "lr": 6.422073074558964e-05}
{"train_loss": 0.10172389447689056, "global_step": 111102, "epoch": 1248, "lr": 6.422017495035487e-05}
{"train_loss": 0.13114646077156067, "global_step": 111103, "epoch": 1248, "lr": 6.421961915320835e-05}
{"train_loss": 0.13650764524936676, "global_step": 111104, "epoch": 1248, "lr": 6.421906335415018e-05}
{"train_loss": 0.18925783038139343, "global_step": 111105, "epoch": 1248, "lr": 6.421850755318042e-05}
{"train_loss": 0.08173041045665741, "global_step": 111106, "epoch": 1248, "lr": 6.421795175029916e-05}
{"train_loss": 0.19654370844364166, "global_step": 111107, "epoch": 1248, "lr": 6.421739594550648e-05}
{"train_loss": 0.221385195851326, "global_step": 111108, "epoch": 1248, "lr": 6.421684013880245e-05}
{"train_loss": 0.11551857739686966, "global_step": 111109, "epoch": 1248, "lr": 6.421628433018713e-05}
{"train_loss": 0.12907426059246063, "global_step": 111110, "epoch": 1248, "lr": 6.42157285196606e-05}
{"train_loss": 0.2067698985338211, "global_step": 111111, "epoch": 1248, "lr": 6.421517270722293e-05}
{"train_loss": 0.11999913305044174, "global_step": 111112, "epoch": 1248, "lr": 6.421461689287422e-05}
{"train_loss": 0.16209585964679718, "global_step": 111113, "epoch": 1248, "lr": 6.421406107661453e-05}
{"train_loss": 0.15917284786701202, "global_step": 111114, "epoch": 1248, "lr": 6.421350525844392e-05}
{"train_loss": 0.21451927721500397, "global_step": 111115, "epoch": 1248, "lr": 6.421294943836248e-05}
{"train_loss": 0.12118567526340485, "global_step": 111116, "epoch": 1248, "lr": 6.42123936163703e-05}
{"train_loss": 0.12027554214000702, "global_step": 111117, "epoch": 1248, "lr": 6.421183779246741e-05}
{"train_loss": 0.13491681218147278, "global_step": 111118, "epoch": 1248, "lr": 6.421128196665392e-05}
{"train_loss": 0.1602056324481964, "global_step": 111119, "epoch": 1248, "lr": 6.42107261389299e-05}
{"train_loss": 0.17765389382839203, "global_step": 111120, "epoch": 1248, "lr": 6.421017030929543e-05}
{"train_loss": 0.08554554730653763, "global_step": 111121, "epoch": 1248, "lr": 6.420961447775056e-05}
{"train_loss": 0.20832328498363495, "global_step": 111122, "epoch": 1248, "lr": 6.42090586442954e-05}
{"train_loss": 0.11935406923294067, "global_step": 111123, "epoch": 1248, "lr": 6.420850280892998e-05}
{"train_loss": 0.18988609313964844, "global_step": 111124, "epoch": 1248, "lr": 6.420794697165442e-05}
{"train_loss": 0.08302110433578491, "global_step": 111125, "epoch": 1248, "lr": 6.420739113246877e-05}
{"train_loss": 0.10340863466262817, "global_step": 111126, "epoch": 1248, "lr": 6.42068352913731e-05}
{"train_loss": 0.18484905362129211, "global_step": 111127, "epoch": 1248, "lr": 6.42062794483675e-05}
{"train_loss": 0.13058869540691376, "global_step": 111128, "epoch": 1248, "lr": 6.420572360345205e-05}
{"train_loss": 0.2829345762729645, "global_step": 111129, "epoch": 1248, "lr": 6.42051677566268e-05}
{"train_loss": 0.11581472307443619, "global_step": 111130, "epoch": 1248, "lr": 6.420461190789184e-05}
{"train_loss": 0.16972512006759644, "global_step": 111131, "epoch": 1248, "lr": 6.420405605724725e-05}
{"train_loss": 0.15107233822345734, "global_step": 111132, "epoch": 1248, "lr": 6.420350020469309e-05}
{"train_loss": 0.1539432853460312, "global_step": 111133, "epoch": 1248, "lr": 6.420294435022945e-05}
{"train_loss": 0.17297832667827606, "global_step": 111134, "epoch": 1248, "lr": 6.420238849385638e-05}
{"train_loss": 0.089375339448452, "global_step": 111135, "epoch": 1248, "lr": 6.420183263557399e-05}
{"train_loss": 0.08468709141016006, "global_step": 111136, "epoch": 1248, "lr": 6.420127677538232e-05}
{"train_loss": 0.19840198755264282, "global_step": 111137, "epoch": 1248, "lr": 6.420072091328149e-05}
{"train_loss": 0.12540209293365479, "global_step": 111138, "epoch": 1248, "lr": 6.420016504927152e-05}
{"train_loss": 0.10444646328687668, "global_step": 111139, "epoch": 1248, "lr": 6.419960918335251e-05}
{"train_loss": 0.19523847103118896, "global_step": 111140, "epoch": 1248, "lr": 6.419905331552456e-05}
{"train_loss": 0.13246503472328186, "global_step": 111141, "epoch": 1248, "lr": 6.41984974457877e-05}
{"train_loss": 0.17225979268550873, "global_step": 111142, "epoch": 1248, "lr": 6.419794157414203e-05}
{"train_loss": 0.10529302060604095, "global_step": 111143, "epoch": 1248, "lr": 6.419738570058762e-05}
{"train_loss": 0.11395569890737534, "global_step": 111144, "epoch": 1248, "lr": 6.419682982512454e-05}
{"train_loss": 0.11623254418373108, "global_step": 111145, "epoch": 1248, "lr": 6.419627394775289e-05}
{"train_loss": 0.12038539350032806, "global_step": 111146, "epoch": 1248, "lr": 6.41957180684727e-05}
{"train_loss": 0.2154838591814041, "global_step": 111147, "epoch": 1248, "lr": 6.419516218728407e-05}
{"train_loss": 0.11255285143852234, "global_step": 111148, "epoch": 1248, "lr": 6.419460630418709e-05}
{"train_loss": 0.16940049827098846, "global_step": 111149, "epoch": 1248, "lr": 6.419405041918181e-05}
{"train_loss": 0.125541090965271, "global_step": 111150, "epoch": 1248, "lr": 6.41934945322683e-05}
{"train_loss": 0.1622726023197174, "global_step": 111151, "epoch": 1248, "lr": 6.419293864344667e-05}
{"train_loss": 0.19198647141456604, "global_step": 111152, "epoch": 1248, "lr": 6.419238275271696e-05}
{"train_loss": 0.15464237332344055, "global_step": 111153, "epoch": 1248, "lr": 6.419182686007925e-05}
{"train_loss": 0.2004735916852951, "global_step": 111154, "epoch": 1248, "lr": 6.419127096553363e-05}
{"train_loss": 0.16590392589569092, "global_step": 111155, "epoch": 1248, "lr": 6.419071506908018e-05}
{"train_loss": 0.11300394684076309, "global_step": 111156, "epoch": 1248, "lr": 6.419015917071894e-05}
{"train_loss": 0.23290470242500305, "global_step": 111157, "epoch": 1248, "lr": 6.418960327045001e-05}
{"train_loss": 0.159955695271492, "global_step": 111158, "epoch": 1248, "lr": 6.418904736827347e-05}
{"train_loss": 0.12441327422857285, "global_step": 111159, "epoch": 1248, "lr": 6.418849146418938e-05}
{"train_loss": 0.15194103213843335, "global_step": 111160, "epoch": 1248, "lr": 6.418793555819782e-05, "val_loss": 5.121303081512451}
{"train_loss": 0.12072666734457016, "global_step": 111161, "epoch": 1249, "lr": 6.418737965029886e-05}
{"train_loss": 0.18043673038482666, "global_step": 111162, "epoch": 1249, "lr": 6.418682374049259e-05}
{"train_loss": 0.10604820400476456, "global_step": 111163, "epoch": 1249, "lr": 6.418626782877908e-05}
{"train_loss": 0.16668309271335602, "global_step": 111164, "epoch": 1249, "lr": 6.418571191515837e-05}
{"train_loss": 0.16271063685417175, "global_step": 111165, "epoch": 1249, "lr": 6.41851559996306e-05}
{"train_loss": 0.12787631154060364, "global_step": 111166, "epoch": 1249, "lr": 6.418460008219579e-05}
{"train_loss": 0.26286303997039795, "global_step": 111167, "epoch": 1249, "lr": 6.418404416285405e-05}
{"train_loss": 0.1497720628976822, "global_step": 111168, "epoch": 1249, "lr": 6.418348824160541e-05}
{"train_loss": 0.19222620129585266, "global_step": 111169, "epoch": 1249, "lr": 6.418293231844998e-05}
{"train_loss": 0.1269461065530777, "global_step": 111170, "epoch": 1249, "lr": 6.418237639338784e-05}
{"train_loss": 0.22688938677310944, "global_step": 111171, "epoch": 1249, "lr": 6.418182046641905e-05}
{"train_loss": 0.16043955087661743, "global_step": 111172, "epoch": 1249, "lr": 6.418126453754367e-05}
{"train_loss": 0.16829036176204681, "global_step": 111173, "epoch": 1249, "lr": 6.418070860676182e-05}
{"train_loss": 0.12719547748565674, "global_step": 111174, "epoch": 1249, "lr": 6.418015267407352e-05}
{"train_loss": 0.0957595556974411, "global_step": 111175, "epoch": 1249, "lr": 6.417959673947889e-05}
{"train_loss": 0.1413431018590927, "global_step": 111176, "epoch": 1249, "lr": 6.417904080297798e-05}
{"train_loss": 0.13844461739063263, "global_step": 111177, "epoch": 1249, "lr": 6.417848486457086e-05}
{"train_loss": 0.12642988562583923, "global_step": 111178, "epoch": 1249, "lr": 6.417792892425763e-05}
{"train_loss": 0.23812519013881683, "global_step": 111179, "epoch": 1249, "lr": 6.417737298203835e-05}
{"train_loss": 0.17409542202949524, "global_step": 111180, "epoch": 1249, "lr": 6.417681703791309e-05}
{"train_loss": 0.21935313940048218, "global_step": 111181, "epoch": 1249, "lr": 6.417626109188192e-05}
{"train_loss": 0.1342564821243286, "global_step": 111182, "epoch": 1249, "lr": 6.417570514394493e-05}
{"train_loss": 0.10086771845817566, "global_step": 111183, "epoch": 1249, "lr": 6.41751491941022e-05}
{"train_loss": 0.13928821682929993, "global_step": 111184, "epoch": 1249, "lr": 6.41745932423538e-05}
{"train_loss": 0.06368890404701233, "global_step": 111185, "epoch": 1249, "lr": 6.417403728869978e-05}
{"train_loss": 0.13397255539894104, "global_step": 111186, "epoch": 1249, "lr": 6.417348133314025e-05}
{"train_loss": 0.18226748704910278, "global_step": 111187, "epoch": 1249, "lr": 6.417292537567525e-05}
{"train_loss": 0.12642906606197357, "global_step": 111188, "epoch": 1249, "lr": 6.417236941630489e-05}
{"train_loss": 0.1489543616771698, "global_step": 111189, "epoch": 1249, "lr": 6.417181345502921e-05}
{"train_loss": 0.1573367565870285, "global_step": 111190, "epoch": 1249, "lr": 6.417125749184831e-05}
{"train_loss": 0.129489928483963, "global_step": 111191, "epoch": 1249, "lr": 6.417070152676226e-05}
{"train_loss": 0.07145870476961136, "global_step": 111192, "epoch": 1249, "lr": 6.417014555977114e-05}
{"train_loss": 0.16968730092048645, "global_step": 111193, "epoch": 1249, "lr": 6.4169589590875e-05}
{"train_loss": 0.13394123315811157, "global_step": 111194, "epoch": 1249, "lr": 6.416903362007394e-05}
{"train_loss": 0.1348610371351242, "global_step": 111195, "epoch": 1249, "lr": 6.416847764736801e-05}
{"train_loss": 0.22885635495185852, "global_step": 111196, "epoch": 1249, "lr": 6.416792167275733e-05}
{"train_loss": 0.1434239000082016, "global_step": 111197, "epoch": 1249, "lr": 6.416736569624194e-05}
{"train_loss": 0.14910918474197388, "global_step": 111198, "epoch": 1249, "lr": 6.416680971782191e-05}
{"train_loss": 0.16823938488960266, "global_step": 111199, "epoch": 1249, "lr": 6.416625373749732e-05}
{"train_loss": 0.19008563458919525, "global_step": 111200, "epoch": 1249, "lr": 6.416569775526824e-05}
{"train_loss": 0.16821373999118805, "global_step": 111201, "epoch": 1249, "lr": 6.416514177113477e-05}
{"train_loss": 0.16293640434741974, "global_step": 111202, "epoch": 1249, "lr": 6.416458578509697e-05}
{"train_loss": 0.17399302124977112, "global_step": 111203, "epoch": 1249, "lr": 6.416402979715493e-05}
{"train_loss": 0.11975984275341034, "global_step": 111204, "epoch": 1249, "lr": 6.41634738073087e-05}
{"train_loss": 0.1450851857662201, "global_step": 111205, "epoch": 1249, "lr": 6.416291781555835e-05}
{"train_loss": 0.11299046874046326, "global_step": 111206, "epoch": 1249, "lr": 6.416236182190398e-05}
{"train_loss": 0.11441729217767715, "global_step": 111207, "epoch": 1249, "lr": 6.416180582634564e-05}
{"train_loss": 0.13817618787288666, "global_step": 111208, "epoch": 1249, "lr": 6.416124982888344e-05}
{"train_loss": 0.10327495634555817, "global_step": 111209, "epoch": 1249, "lr": 6.416069382951742e-05}
{"train_loss": 0.1290108859539032, "global_step": 111210, "epoch": 1249, "lr": 6.416013782824766e-05}
{"train_loss": 0.16992925107479095, "global_step": 111211, "epoch": 1249, "lr": 6.415958182507424e-05}
{"train_loss": 0.1377008706331253, "global_step": 111212, "epoch": 1249, "lr": 6.415902581999726e-05}
{"train_loss": 0.24192237854003906, "global_step": 111213, "epoch": 1249, "lr": 6.415846981301674e-05}
{"train_loss": 0.24031537771224976, "global_step": 111214, "epoch": 1249, "lr": 6.41579138041328e-05}
{"train_loss": 0.1913597732782364, "global_step": 111215, "epoch": 1249, "lr": 6.41573577933455e-05}
{"train_loss": 0.14568164944648743, "global_step": 111216, "epoch": 1249, "lr": 6.415680178065493e-05}
{"train_loss": 0.15623198449611664, "global_step": 111217, "epoch": 1249, "lr": 6.415624576606113e-05}
{"train_loss": 0.19386886060237885, "global_step": 111218, "epoch": 1249, "lr": 6.415568974956421e-05}
{"train_loss": 0.22243154048919678, "global_step": 111219, "epoch": 1249, "lr": 6.415513373116422e-05}
{"train_loss": 0.13572761416435242, "global_step": 111220, "epoch": 1249, "lr": 6.415457771086126e-05}
{"train_loss": 0.14793352782726288, "global_step": 111221, "epoch": 1249, "lr": 6.415402168865537e-05}
{"train_loss": 0.12327595800161362, "global_step": 111222, "epoch": 1249, "lr": 6.415346566454666e-05}
{"train_loss": 0.21911726891994476, "global_step": 111223, "epoch": 1249, "lr": 6.41529096385352e-05}
{"train_loss": 0.19764330983161926, "global_step": 111224, "epoch": 1249, "lr": 6.415235361062103e-05}
{"train_loss": 0.1583719402551651, "global_step": 111225, "epoch": 1249, "lr": 6.415179758080425e-05}
{"train_loss": 0.13258545100688934, "global_step": 111226, "epoch": 1249, "lr": 6.415124154908494e-05}
{"train_loss": 0.14895612001419067, "global_step": 111227, "epoch": 1249, "lr": 6.415068551546318e-05}
{"train_loss": 0.14938153326511383, "global_step": 111228, "epoch": 1249, "lr": 6.415012947993902e-05}
{"train_loss": 0.20751532912254333, "global_step": 111229, "epoch": 1249, "lr": 6.414957344251254e-05}
{"train_loss": 0.24032597243785858, "global_step": 111230, "epoch": 1249, "lr": 6.414901740318382e-05}
{"train_loss": 0.13451169431209564, "global_step": 111231, "epoch": 1249, "lr": 6.414846136195295e-05}
{"train_loss": 0.13075479865074158, "global_step": 111232, "epoch": 1249, "lr": 6.414790531881999e-05}
{"train_loss": 0.2157505601644516, "global_step": 111233, "epoch": 1249, "lr": 6.414734927378502e-05}
{"train_loss": 0.11741450428962708, "global_step": 111234, "epoch": 1249, "lr": 6.41467932268481e-05}
{"train_loss": 0.1400493085384369, "global_step": 111235, "epoch": 1249, "lr": 6.414623717800932e-05}
{"train_loss": 0.13727429509162903, "global_step": 111236, "epoch": 1249, "lr": 6.414568112726875e-05}
{"train_loss": 0.13221478462219238, "global_step": 111237, "epoch": 1249, "lr": 6.414512507462647e-05}
{"train_loss": 0.11795707046985626, "global_step": 111238, "epoch": 1249, "lr": 6.414456902008255e-05}
{"train_loss": 0.14666901528835297, "global_step": 111239, "epoch": 1249, "lr": 6.414401296363706e-05}
{"train_loss": 0.18283747136592865, "global_step": 111240, "epoch": 1249, "lr": 6.414345690529009e-05}
{"train_loss": 0.09322090446949005, "global_step": 111241, "epoch": 1249, "lr": 6.41429008450417e-05}
{"train_loss": 0.17903761565685272, "global_step": 111242, "epoch": 1249, "lr": 6.414234478289197e-05}
{"train_loss": 0.08121223747730255, "global_step": 111243, "epoch": 1249, "lr": 6.414178871884098e-05}
{"train_loss": 0.15610042214393616, "global_step": 111244, "epoch": 1249, "lr": 6.414123265288878e-05}
{"train_loss": 0.08766753226518631, "global_step": 111245, "epoch": 1249, "lr": 6.414067658503548e-05}
{"train_loss": 0.15643569827079773, "global_step": 111246, "epoch": 1249, "lr": 6.414012051528114e-05}
{"train_loss": 0.14646168053150177, "global_step": 111247, "epoch": 1249, "lr": 6.413956444362581e-05}
{"train_loss": 0.13333134353160858, "global_step": 111248, "epoch": 1249, "lr": 6.413900837006962e-05}
{"train_loss": 0.15368224680423737, "global_step": 111249, "epoch": 1249, "lr": 6.41384522946126e-05, "val_loss": 4.931221008300781}
{"train_loss": 0.1186535432934761, "global_step": 111250, "epoch": 1250, "lr": 6.413789621725483e-05}
{"train_loss": 0.09642848372459412, "global_step": 111251, "epoch": 1250, "lr": 6.41373401379964e-05}
{"train_loss": 0.1197146400809288, "global_step": 111252, "epoch": 1250, "lr": 6.413678405683736e-05}
{"train_loss": 0.19100642204284668, "global_step": 111253, "epoch": 1250, "lr": 6.413622797377782e-05}
{"train_loss": 0.22256335616111755, "global_step": 111254, "epoch": 1250, "lr": 6.413567188881781e-05}
{"train_loss": 0.1297985464334488, "global_step": 111255, "epoch": 1250, "lr": 6.413511580195746e-05}
{"train_loss": 0.1357221007347107, "global_step": 111256, "epoch": 1250, "lr": 6.41345597131968e-05}
{"train_loss": 0.16829848289489746, "global_step": 111257, "epoch": 1250, "lr": 6.413400362253594e-05}
{"train_loss": 0.16969189047813416, "global_step": 111258, "epoch": 1250, "lr": 6.413344752997492e-05}
{"train_loss": 0.13176535069942474, "global_step": 111259, "epoch": 1250, "lr": 6.413289143551384e-05}
{"train_loss": 0.15918287634849548, "global_step": 111260, "epoch": 1250, "lr": 6.413233533915276e-05}
{"train_loss": 0.20983994007110596, "global_step": 111261, "epoch": 1250, "lr": 6.413177924089176e-05}
{"train_loss": 0.14066484570503235, "global_step": 111262, "epoch": 1250, "lr": 6.413122314073092e-05}
{"train_loss": 0.1998487114906311, "global_step": 111263, "epoch": 1250, "lr": 6.41306670386703e-05}
{"train_loss": 0.06412346661090851, "global_step": 111264, "epoch": 1250, "lr": 6.413011093471e-05}
{"train_loss": 0.18107187747955322, "global_step": 111265, "epoch": 1250, "lr": 6.412955482885005e-05}
{"train_loss": 0.18834063410758972, "global_step": 111266, "epoch": 1250, "lr": 6.412899872109059e-05}
{"train_loss": 0.13625046610832214, "global_step": 111267, "epoch": 1250, "lr": 6.412844261143163e-05}
{"train_loss": 0.07685433328151703, "global_step": 111268, "epoch": 1250, "lr": 6.412788649987328e-05}
{"train_loss": 0.1406795233488083, "global_step": 111269, "epoch": 1250, "lr": 6.41273303864156e-05}
{"train_loss": 0.15540732443332672, "global_step": 111270, "epoch": 1250, "lr": 6.412677427105869e-05}
{"train_loss": 0.16632303595542908, "global_step": 111271, "epoch": 1250, "lr": 6.41262181538026e-05}
{"train_loss": 0.14871038496494293, "global_step": 111272, "epoch": 1250, "lr": 6.41256620346474e-05}
{"train_loss": 0.1420375555753708, "global_step": 111273, "epoch": 1250, "lr": 6.412510591359319e-05}
{"train_loss": 0.1213659718632698, "global_step": 111274, "epoch": 1250, "lr": 6.412454979064003e-05}
{"train_loss": 0.14316977560520172, "global_step": 111275, "epoch": 1250, "lr": 6.4123993665788e-05}
{"train_loss": 0.18385858833789825, "global_step": 111276, "epoch": 1250, "lr": 6.412343753903716e-05}
{"train_loss": 0.10227089375257492, "global_step": 111277, "epoch": 1250, "lr": 6.412288141038761e-05}
{"train_loss": 0.1372712254524231, "global_step": 111278, "epoch": 1250, "lr": 6.412232527983941e-05}
{"train_loss": 0.15025046467781067, "global_step": 111279, "epoch": 1250, "lr": 6.412176914739262e-05}
{"train_loss": 0.13447333872318268, "global_step": 111280, "epoch": 1250, "lr": 6.412121301304734e-05}
{"train_loss": 0.09743632376194, "global_step": 111281, "epoch": 1250, "lr": 6.412065687680365e-05}
{"train_loss": 0.13107380270957947, "global_step": 111282, "epoch": 1250, "lr": 6.41201007386616e-05}
{"train_loss": 0.10546659678220749, "global_step": 111283, "epoch": 1250, "lr": 6.411954459862126e-05}
{"train_loss": 0.09544550627470016, "global_step": 111284, "epoch": 1250, "lr": 6.411898845668273e-05}
{"train_loss": 0.13063903152942657, "global_step": 111285, "epoch": 1250, "lr": 6.411843231284607e-05}
{"train_loss": 0.11493313312530518, "global_step": 111286, "epoch": 1250, "lr": 6.411787616711135e-05}
{"train_loss": 0.13772839307785034, "global_step": 111287, "epoch": 1250, "lr": 6.411732001947867e-05}
{"train_loss": 0.1930626928806305, "global_step": 111288, "epoch": 1250, "lr": 6.41167638699481e-05}
{"train_loss": 0.19931967556476593, "global_step": 111289, "epoch": 1250, "lr": 6.411620771851968e-05}
{"train_loss": 0.20634347200393677, "global_step": 111290, "epoch": 1250, "lr": 6.411565156519351e-05}
{"train_loss": 0.19142022728919983, "global_step": 111291, "epoch": 1250, "lr": 6.411509540996966e-05}
{"train_loss": 0.17540380358695984, "global_step": 111292, "epoch": 1250, "lr": 6.411453925284822e-05}
{"train_loss": 0.11558330804109573, "global_step": 111293, "epoch": 1250, "lr": 6.411398309382924e-05}
{"train_loss": 0.18728645145893097, "global_step": 111294, "epoch": 1250, "lr": 6.411342693291282e-05}
{"train_loss": 0.1402992308139801, "global_step": 111295, "epoch": 1250, "lr": 6.411287077009901e-05}
{"train_loss": 0.11347123235464096, "global_step": 111296, "epoch": 1250, "lr": 6.41123146053879e-05}
{"train_loss": 0.15107759833335876, "global_step": 111297, "epoch": 1250, "lr": 6.411175843877956e-05}
{"train_loss": 0.15859240293502808, "global_step": 111298, "epoch": 1250, "lr": 6.411120227027407e-05}
{"train_loss": 0.14558623731136322, "global_step": 111299, "epoch": 1250, "lr": 6.41106460998715e-05}
{"train_loss": 0.1062096580862999, "global_step": 111300, "epoch": 1250, "lr": 6.411008992757195e-05}
{"train_loss": 0.19630558788776398, "global_step": 111301, "epoch": 1250, "lr": 6.410953375337543e-05}
{"train_loss": 0.11341835558414459, "global_step": 111302, "epoch": 1250, "lr": 6.410897757728208e-05}
{"train_loss": 0.11228646337985992, "global_step": 111303, "epoch": 1250, "lr": 6.410842139929194e-05}
{"train_loss": 0.09796842932701111, "global_step": 111304, "epoch": 1250, "lr": 6.41078652194051e-05}
{"train_loss": 0.19053202867507935, "global_step": 111305, "epoch": 1250, "lr": 6.410730903762163e-05}
{"train_loss": 0.17618928849697113, "global_step": 111306, "epoch": 1250, "lr": 6.41067528539416e-05}
{"train_loss": 0.1569666862487793, "global_step": 111307, "epoch": 1250, "lr": 6.410619666836509e-05}
{"train_loss": 0.21071405708789825, "global_step": 111308, "epoch": 1250, "lr": 6.410564048089217e-05}
{"train_loss": 0.08957225829362869, "global_step": 111309, "epoch": 1250, "lr": 6.410508429152291e-05}
{"train_loss": 0.15012063086032867, "global_step": 111310, "epoch": 1250, "lr": 6.410452810025741e-05}
{"train_loss": 0.13774162530899048, "global_step": 111311, "epoch": 1250, "lr": 6.410397190709572e-05}
{"train_loss": 0.19793733954429626, "global_step": 111312, "epoch": 1250, "lr": 6.410341571203793e-05}
{"train_loss": 0.15121915936470032, "global_step": 111313, "epoch": 1250, "lr": 6.41028595150841e-05}
{"train_loss": 0.10434463620185852, "global_step": 111314, "epoch": 1250, "lr": 6.410230331623432e-05}
{"train_loss": 0.1253584921360016, "global_step": 111315, "epoch": 1250, "lr": 6.410174711548866e-05}
{"train_loss": 0.08298656344413757, "global_step": 111316, "epoch": 1250, "lr": 6.410119091284718e-05}
{"train_loss": 0.17934954166412354, "global_step": 111317, "epoch": 1250, "lr": 6.410063470830997e-05}
{"train_loss": 0.10929784923791885, "global_step": 111318, "epoch": 1250, "lr": 6.41000785018771e-05}
{"train_loss": 0.15954045951366425, "global_step": 111319, "epoch": 1250, "lr": 6.409952229354866e-05}
{"train_loss": 0.18855127692222595, "global_step": 111320, "epoch": 1250, "lr": 6.40989660833247e-05}
{"train_loss": 0.057656414806842804, "global_step": 111321, "epoch": 1250, "lr": 6.40984098712053e-05}
{"train_loss": 0.09010564535856247, "global_step": 111322, "epoch": 1250, "lr": 6.409785365719056e-05}
{"train_loss": 0.11176762729883194, "global_step": 111323, "epoch": 1250, "lr": 6.409729744128053e-05}
{"train_loss": 0.246497243642807, "global_step": 111324, "epoch": 1250, "lr": 6.409674122347526e-05}
{"train_loss": 0.10975956171751022, "global_step": 111325, "epoch": 1250, "lr": 6.409618500377489e-05}
{"train_loss": 0.0971662849187851, "global_step": 111326, "epoch": 1250, "lr": 6.409562878217946e-05}
{"train_loss": 0.1206561028957367, "global_step": 111327, "epoch": 1250, "lr": 6.409507255868901e-05}
{"train_loss": 0.18162816762924194, "global_step": 111328, "epoch": 1250, "lr": 6.409451633330367e-05}
{"train_loss": 0.13115671277046204, "global_step": 111329, "epoch": 1250, "lr": 6.409396010602349e-05}
{"train_loss": 0.1390378475189209, "global_step": 111330, "epoch": 1250, "lr": 6.409340387684856e-05}
{"train_loss": 0.1619044989347458, "global_step": 111331, "epoch": 1250, "lr": 6.409284764577895e-05}
{"train_loss": 0.14958621561527252, "global_step": 111332, "epoch": 1250, "lr": 6.409229141281469e-05}
{"train_loss": 0.1433737725019455, "global_step": 111333, "epoch": 1250, "lr": 6.409173517795593e-05}
{"train_loss": 0.1438247710466385, "global_step": 111334, "epoch": 1250, "lr": 6.40911789412027e-05}
{"train_loss": 0.1290518343448639, "global_step": 111335, "epoch": 1250, "lr": 6.409062270255508e-05}
{"train_loss": 0.12070298939943314, "global_step": 111336, "epoch": 1250, "lr": 6.409006646201314e-05}
{"train_loss": 0.09374161809682846, "global_step": 111337, "epoch": 1250, "lr": 6.408951021957697e-05}
{"train_loss": 0.14347403714161241, "global_step": 111338, "epoch": 1250, "lr": 6.408895397524664e-05, "train/sim_max_reward_0": 0.698002691431898, "train/sim_max_reward_1": 0.4791398327837469, "train/sim_max_reward_2": 0.6230830732409569, "train/sim_max_reward_3": 0.6151359953218903, "train/sim_max_reward_4": 0.960205741659299, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.6081229217802193, "test/sim_max_reward_4500001": 0.969253239546594, "test/sim_max_reward_4500002": 0.9531466438056473, "test/sim_max_reward_4500003": 0.3623671665731306, "test/sim_max_reward_4500004": 0.5455684409285134, "test/sim_max_reward_4500005": 0.16972206692703512, "test/sim_max_reward_4500006": 0.9849350705379535, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.049190520496595724, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.9169727959622378, "test/sim_max_reward_4500013": 0.5696741996547635, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9793475140560481, "test/sim_max_reward_4500016": 0.6373978574073039, "test/sim_max_reward_4500017": 0.5401568439765414, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9621825838144397, "test/sim_max_reward_4500020": 0.6022950828776991, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.7057730307410509, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.3915257225388699, "test/sim_max_reward_4500026": 0.9890845733131939, "test/sim_max_reward_4500027": 0.9507325452271775, "test/sim_max_reward_4500028": 0.440101770956634, "test/sim_max_reward_4500029": 0.921127658594022, "test/sim_max_reward_4500030": 0.910104290073447, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6635445670293009, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.9990346544409886, "test/sim_max_reward_4500035": 0.5361619963881707, "test/sim_max_reward_4500036": 0.9942423009375985, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.6271583712739184, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.8752351551646174, "test/sim_max_reward_4500042": 0.13976507848023675, "test/sim_max_reward_4500043": 0.23144023282051907, "test/sim_max_reward_4500044": 0.5070876270854452, "test/sim_max_reward_4500045": 0.7000126712922069, "test/sim_max_reward_4500046": 0.4383160415258085, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9985468245297595, "test/sim_max_reward_4500049": 0.9766951846317702, "train/mean_score": 0.7292612224062985, "test/mean_score": 0.6769204649077893, "val_loss": 5.162447929382324, "train_action_mse_error": 23.272321701049805}
{"train_loss": 0.1456638127565384, "global_step": 111339, "epoch": 1251, "lr": 6.408839772902223e-05}
{"train_loss": 0.14985038340091705, "global_step": 111340, "epoch": 1251, "lr": 6.40878414809038e-05}
{"train_loss": 0.09567604213953018, "global_step": 111341, "epoch": 1251, "lr": 6.408728523089142e-05}
{"train_loss": 0.17040574550628662, "global_step": 111342, "epoch": 1251, "lr": 6.408672897898519e-05}
{"train_loss": 0.1487867534160614, "global_step": 111343, "epoch": 1251, "lr": 6.408617272518517e-05}
{"train_loss": 0.1539769470691681, "global_step": 111344, "epoch": 1251, "lr": 6.408561646949143e-05}
{"train_loss": 0.14067989587783813, "global_step": 111345, "epoch": 1251, "lr": 6.408506021190403e-05}
{"train_loss": 0.1939959079027176, "global_step": 111346, "epoch": 1251, "lr": 6.408450395242309e-05}
{"train_loss": 0.13564035296440125, "global_step": 111347, "epoch": 1251, "lr": 6.408394769104865e-05}
{"train_loss": 0.2008364498615265, "global_step": 111348, "epoch": 1251, "lr": 6.40833914277808e-05}
{"train_loss": 0.10920913517475128, "global_step": 111349, "epoch": 1251, "lr": 6.408283516261962e-05}
{"train_loss": 0.1542663723230362, "global_step": 111350, "epoch": 1251, "lr": 6.408227889556516e-05}
{"train_loss": 0.1440809816122055, "global_step": 111351, "epoch": 1251, "lr": 6.408172262661752e-05}
{"train_loss": 0.13103461265563965, "global_step": 111352, "epoch": 1251, "lr": 6.408116635577675e-05}
{"train_loss": 0.19170516729354858, "global_step": 111353, "epoch": 1251, "lr": 6.408061008304295e-05}
{"train_loss": 0.08628164231777191, "global_step": 111354, "epoch": 1251, "lr": 6.408005380841619e-05}
{"train_loss": 0.1424359381198883, "global_step": 111355, "epoch": 1251, "lr": 6.407949753189652e-05}
{"train_loss": 0.19801494479179382, "global_step": 111356, "epoch": 1251, "lr": 6.407894125348404e-05}
{"train_loss": 0.20218802988529205, "global_step": 111357, "epoch": 1251, "lr": 6.407838497317882e-05}
{"train_loss": 0.13617530465126038, "global_step": 111358, "epoch": 1251, "lr": 6.407782869098095e-05}
{"train_loss": 0.19370673596858978, "global_step": 111359, "epoch": 1251, "lr": 6.407727240689046e-05}
{"train_loss": 0.09513027220964432, "global_step": 111360, "epoch": 1251, "lr": 6.407671612090746e-05}
{"train_loss": 0.10797034204006195, "global_step": 111361, "epoch": 1251, "lr": 6.407615983303204e-05}
{"train_loss": 0.2089300900697708, "global_step": 111362, "epoch": 1251, "lr": 6.407560354326422e-05}
{"train_loss": 0.18370401859283447, "global_step": 111363, "epoch": 1251, "lr": 6.407504725160412e-05}
{"train_loss": 0.1720958799123764, "global_step": 111364, "epoch": 1251, "lr": 6.407449095805178e-05}
{"train_loss": 0.12388737499713898, "global_step": 111365, "epoch": 1251, "lr": 6.407393466260733e-05}
{"train_loss": 0.12018921971321106, "global_step": 111366, "epoch": 1251, "lr": 6.407337836527079e-05}
{"train_loss": 0.13587726652622223, "global_step": 111367, "epoch": 1251, "lr": 6.407282206604227e-05}
{"train_loss": 0.17268270254135132, "global_step": 111368, "epoch": 1251, "lr": 6.407226576492181e-05}
{"train_loss": 0.19570687413215637, "global_step": 111369, "epoch": 1251, "lr": 6.407170946190953e-05}
{"train_loss": 0.14217087626457214, "global_step": 111370, "epoch": 1251, "lr": 6.407115315700547e-05}
{"train_loss": 0.12115959078073502, "global_step": 111371, "epoch": 1251, "lr": 6.40705968502097e-05}
{"train_loss": 0.07486061751842499, "global_step": 111372, "epoch": 1251, "lr": 6.407004054152233e-05}
{"train_loss": 0.1764373630285263, "global_step": 111373, "epoch": 1251, "lr": 6.406948423094341e-05}
{"train_loss": 0.1253504902124405, "global_step": 111374, "epoch": 1251, "lr": 6.406892791847303e-05}
{"train_loss": 0.24703112244606018, "global_step": 111375, "epoch": 1251, "lr": 6.406837160411124e-05}
{"train_loss": 0.17089977860450745, "global_step": 111376, "epoch": 1251, "lr": 6.406781528785812e-05}
{"train_loss": 0.0809529721736908, "global_step": 111377, "epoch": 1251, "lr": 6.406725896971376e-05}
{"train_loss": 0.1895403265953064, "global_step": 111378, "epoch": 1251, "lr": 6.406670264967825e-05}
{"train_loss": 0.1594119369983673, "global_step": 111379, "epoch": 1251, "lr": 6.406614632775163e-05}
{"train_loss": 0.1211346983909607, "global_step": 111380, "epoch": 1251, "lr": 6.406559000393398e-05}
{"train_loss": 0.12012264132499695, "global_step": 111381, "epoch": 1251, "lr": 6.406503367822537e-05}
{"train_loss": 0.16280223429203033, "global_step": 111382, "epoch": 1251, "lr": 6.406447735062591e-05}
{"train_loss": 0.19344256818294525, "global_step": 111383, "epoch": 1251, "lr": 6.406392102113565e-05}
{"train_loss": 0.2142188549041748, "global_step": 111384, "epoch": 1251, "lr": 6.406336468975466e-05}
{"train_loss": 0.14946220815181732, "global_step": 111385, "epoch": 1251, "lr": 6.406280835648303e-05}
{"train_loss": 0.16746921837329865, "global_step": 111386, "epoch": 1251, "lr": 6.406225202132082e-05}
{"train_loss": 0.19019964337348938, "global_step": 111387, "epoch": 1251, "lr": 6.406169568426812e-05}
{"train_loss": 0.11134769767522812, "global_step": 111388, "epoch": 1251, "lr": 6.406113934532499e-05}
{"train_loss": 0.14932042360305786, "global_step": 111389, "epoch": 1251, "lr": 6.40605830044915e-05}
{"train_loss": 0.21707724034786224, "global_step": 111390, "epoch": 1251, "lr": 6.406002666176776e-05}
{"train_loss": 0.11615432053804398, "global_step": 111391, "epoch": 1251, "lr": 6.40594703171538e-05}
{"train_loss": 0.23841269314289093, "global_step": 111392, "epoch": 1251, "lr": 6.405891397064972e-05}
{"train_loss": 0.22557146847248077, "global_step": 111393, "epoch": 1251, "lr": 6.405835762225561e-05}
{"train_loss": 0.1386955976486206, "global_step": 111394, "epoch": 1251, "lr": 6.40578012719715e-05}
{"train_loss": 0.17684870958328247, "global_step": 111395, "epoch": 1251, "lr": 6.40572449197975e-05}
{"train_loss": 0.16520072519779205, "global_step": 111396, "epoch": 1251, "lr": 6.405668856573368e-05}
{"train_loss": 0.198750838637352, "global_step": 111397, "epoch": 1251, "lr": 6.405613220978012e-05}
{"train_loss": 0.17988421022891998, "global_step": 111398, "epoch": 1251, "lr": 6.405557585193686e-05}
{"train_loss": 0.10935764759778976, "global_step": 111399, "epoch": 1251, "lr": 6.4055019492204e-05}
{"train_loss": 0.14966894686222076, "global_step": 111400, "epoch": 1251, "lr": 6.405446313058162e-05}
{"train_loss": 0.16304515302181244, "global_step": 111401, "epoch": 1251, "lr": 6.405390676706979e-05}
{"train_loss": 0.16311883926391602, "global_step": 111402, "epoch": 1251, "lr": 6.405335040166858e-05}
{"train_loss": 0.19695594906806946, "global_step": 111403, "epoch": 1251, "lr": 6.405279403437809e-05}
{"train_loss": 0.08904565870761871, "global_step": 111404, "epoch": 1251, "lr": 6.405223766519834e-05}
{"train_loss": 0.23927557468414307, "global_step": 111405, "epoch": 1251, "lr": 6.405168129412946e-05}
{"train_loss": 0.18624000251293182, "global_step": 111406, "epoch": 1251, "lr": 6.405112492117148e-05}
{"train_loss": 0.20338355004787445, "global_step": 111407, "epoch": 1251, "lr": 6.405056854632453e-05}
{"train_loss": 0.12952466309070587, "global_step": 111408, "epoch": 1251, "lr": 6.405001216958863e-05}
{"train_loss": 0.1359536051750183, "global_step": 111409, "epoch": 1251, "lr": 6.404945579096389e-05}
{"train_loss": 0.1494644284248352, "global_step": 111410, "epoch": 1251, "lr": 6.404889941045035e-05}
{"train_loss": 0.12186184525489807, "global_step": 111411, "epoch": 1251, "lr": 6.404834302804814e-05}
{"train_loss": 0.16990630328655243, "global_step": 111412, "epoch": 1251, "lr": 6.404778664375727e-05}
{"train_loss": 0.18789014220237732, "global_step": 111413, "epoch": 1251, "lr": 6.404723025757786e-05}
{"train_loss": 0.1672573834657669, "global_step": 111414, "epoch": 1251, "lr": 6.404667386950997e-05}
{"train_loss": 0.16085408627986908, "global_step": 111415, "epoch": 1251, "lr": 6.404611747955369e-05}
{"train_loss": 0.1496901512145996, "global_step": 111416, "epoch": 1251, "lr": 6.404556108770908e-05}
{"train_loss": 0.13205768167972565, "global_step": 111417, "epoch": 1251, "lr": 6.404500469397619e-05}
{"train_loss": 0.22662317752838135, "global_step": 111418, "epoch": 1251, "lr": 6.404444829835513e-05}
{"train_loss": 0.15727318823337555, "global_step": 111419, "epoch": 1251, "lr": 6.404389190084597e-05}
{"train_loss": 0.14947709441184998, "global_step": 111420, "epoch": 1251, "lr": 6.404333550144878e-05}
{"train_loss": 0.15438146889209747, "global_step": 111421, "epoch": 1251, "lr": 6.404277910016364e-05}
{"train_loss": 0.12117461115121841, "global_step": 111422, "epoch": 1251, "lr": 6.40422226969906e-05}
{"train_loss": 0.16335003077983856, "global_step": 111423, "epoch": 1251, "lr": 6.404166629192977e-05}
{"train_loss": 0.13476087152957916, "global_step": 111424, "epoch": 1251, "lr": 6.404110988498119e-05}
{"train_loss": 0.13416296243667603, "global_step": 111425, "epoch": 1251, "lr": 6.404055347614498e-05}
{"train_loss": 0.24887526035308838, "global_step": 111426, "epoch": 1251, "lr": 6.403999706542118e-05}
{"train_loss": 0.15840503832932268, "global_step": 111427, "epoch": 1251, "lr": 6.403944065280987e-05, "val_loss": 5.025576591491699}
{"train_loss": 0.12929494678974152, "global_step": 111428, "epoch": 1252, "lr": 6.403888423831112e-05}
{"train_loss": 0.14081057906150818, "global_step": 111429, "epoch": 1252, "lr": 6.403832782192504e-05}
{"train_loss": 0.25479236245155334, "global_step": 111430, "epoch": 1252, "lr": 6.403777140365166e-05}
{"train_loss": 0.1332949697971344, "global_step": 111431, "epoch": 1252, "lr": 6.403721498349106e-05}
{"train_loss": 0.2353678196668625, "global_step": 111432, "epoch": 1252, "lr": 6.403665856144336e-05}
{"train_loss": 0.1356562376022339, "global_step": 111433, "epoch": 1252, "lr": 6.403610213750857e-05}
{"train_loss": 0.0989084467291832, "global_step": 111434, "epoch": 1252, "lr": 6.403554571168682e-05}
{"train_loss": 0.09451024234294891, "global_step": 111435, "epoch": 1252, "lr": 6.403498928397816e-05}
{"train_loss": 0.179372638463974, "global_step": 111436, "epoch": 1252, "lr": 6.403443285438266e-05}
{"train_loss": 0.14495022594928741, "global_step": 111437, "epoch": 1252, "lr": 6.40338764229004e-05}
{"train_loss": 0.1939958930015564, "global_step": 111438, "epoch": 1252, "lr": 6.403331998953144e-05}
{"train_loss": 0.18035255372524261, "global_step": 111439, "epoch": 1252, "lr": 6.40327635542759e-05}
{"train_loss": 0.13671106100082397, "global_step": 111440, "epoch": 1252, "lr": 6.403220711713381e-05}
{"train_loss": 0.12499385327100754, "global_step": 111441, "epoch": 1252, "lr": 6.403165067810527e-05}
{"train_loss": 0.10433105379343033, "global_step": 111442, "epoch": 1252, "lr": 6.403109423719034e-05}
{"train_loss": 0.1681826114654541, "global_step": 111443, "epoch": 1252, "lr": 6.40305377943891e-05}
{"train_loss": 0.16231873631477356, "global_step": 111444, "epoch": 1252, "lr": 6.402998134970163e-05}
{"train_loss": 0.23779840767383575, "global_step": 111445, "epoch": 1252, "lr": 6.4029424903128e-05}
{"train_loss": 0.0865216851234436, "global_step": 111446, "epoch": 1252, "lr": 6.402886845466828e-05}
{"train_loss": 0.0887836292386055, "global_step": 111447, "epoch": 1252, "lr": 6.402831200432254e-05}
{"train_loss": 0.12811748683452606, "global_step": 111448, "epoch": 1252, "lr": 6.402775555209088e-05}
{"train_loss": 0.16831612586975098, "global_step": 111449, "epoch": 1252, "lr": 6.402719909797335e-05}
{"train_loss": 0.14089128375053406, "global_step": 111450, "epoch": 1252, "lr": 6.402664264197005e-05}
{"train_loss": 0.20146581530570984, "global_step": 111451, "epoch": 1252, "lr": 6.402608618408103e-05}
{"train_loss": 0.11646275967359543, "global_step": 111452, "epoch": 1252, "lr": 6.402552972430637e-05}
{"train_loss": 0.11698669195175171, "global_step": 111453, "epoch": 1252, "lr": 6.402497326264614e-05}
{"train_loss": 0.16947872936725616, "global_step": 111454, "epoch": 1252, "lr": 6.402441679910044e-05}
{"train_loss": 0.11294928938150406, "global_step": 111455, "epoch": 1252, "lr": 6.402386033366932e-05}
{"train_loss": 0.13075964152812958, "global_step": 111456, "epoch": 1252, "lr": 6.402330386635286e-05}
{"train_loss": 0.19712428748607635, "global_step": 111457, "epoch": 1252, "lr": 6.402274739715114e-05}
{"train_loss": 0.1714685708284378, "global_step": 111458, "epoch": 1252, "lr": 6.402219092606424e-05}
{"train_loss": 0.1498689502477646, "global_step": 111459, "epoch": 1252, "lr": 6.402163445309221e-05}
{"train_loss": 0.1428806185722351, "global_step": 111460, "epoch": 1252, "lr": 6.402107797823517e-05}
{"train_loss": 0.11745871603488922, "global_step": 111461, "epoch": 1252, "lr": 6.402052150149315e-05}
{"train_loss": 0.1709236204624176, "global_step": 111462, "epoch": 1252, "lr": 6.401996502286624e-05}
{"train_loss": 0.12504048645496368, "global_step": 111463, "epoch": 1252, "lr": 6.401940854235453e-05}
{"train_loss": 0.15053170919418335, "global_step": 111464, "epoch": 1252, "lr": 6.401885205995805e-05}
{"train_loss": 0.15078863501548767, "global_step": 111465, "epoch": 1252, "lr": 6.401829557567694e-05}
{"train_loss": 0.13678380846977234, "global_step": 111466, "epoch": 1252, "lr": 6.401773908951123e-05}
{"train_loss": 0.13917841017246246, "global_step": 111467, "epoch": 1252, "lr": 6.4017182601461e-05}
{"train_loss": 0.1715695708990097, "global_step": 111468, "epoch": 1252, "lr": 6.401662611152633e-05}
{"train_loss": 0.16249732673168182, "global_step": 111469, "epoch": 1252, "lr": 6.40160696197073e-05}
{"train_loss": 0.11166084557771683, "global_step": 111470, "epoch": 1252, "lr": 6.401551312600398e-05}
{"train_loss": 0.0735112652182579, "global_step": 111471, "epoch": 1252, "lr": 6.401495663041644e-05}
{"train_loss": 0.09911943972110748, "global_step": 111472, "epoch": 1252, "lr": 6.401440013294478e-05}
{"train_loss": 0.10420144349336624, "global_step": 111473, "epoch": 1252, "lr": 6.401384363358904e-05}
{"train_loss": 0.16352103650569916, "global_step": 111474, "epoch": 1252, "lr": 6.40132871323493e-05}
{"train_loss": 0.1801372468471527, "global_step": 111475, "epoch": 1252, "lr": 6.401273062922566e-05}
{"train_loss": 0.1088382750749588, "global_step": 111476, "epoch": 1252, "lr": 6.401217412421816e-05}
{"train_loss": 0.21888913214206696, "global_step": 111477, "epoch": 1252, "lr": 6.401161761732692e-05}
{"train_loss": 0.1896052360534668, "global_step": 111478, "epoch": 1252, "lr": 6.401106110855197e-05}
{"train_loss": 0.07552853226661682, "global_step": 111479, "epoch": 1252, "lr": 6.401050459789341e-05}
{"train_loss": 0.10261262208223343, "global_step": 111480, "epoch": 1252, "lr": 6.40099480853513e-05}
{"train_loss": 0.13225676119327545, "global_step": 111481, "epoch": 1252, "lr": 6.400939157092574e-05}
{"train_loss": 0.1484936773777008, "global_step": 111482, "epoch": 1252, "lr": 6.400883505461678e-05}
{"train_loss": 0.1452123373746872, "global_step": 111483, "epoch": 1252, "lr": 6.40082785364245e-05}
{"train_loss": 0.19336625933647156, "global_step": 111484, "epoch": 1252, "lr": 6.400772201634898e-05}
{"train_loss": 0.18050436675548553, "global_step": 111485, "epoch": 1252, "lr": 6.400716549439029e-05}
{"train_loss": 0.1233171820640564, "global_step": 111486, "epoch": 1252, "lr": 6.40066089705485e-05}
{"train_loss": 0.10266140848398209, "global_step": 111487, "epoch": 1252, "lr": 6.400605244482372e-05}
{"train_loss": 0.08744754642248154, "global_step": 111488, "epoch": 1252, "lr": 6.400549591721596e-05}
{"train_loss": 0.08615374565124512, "global_step": 111489, "epoch": 1252, "lr": 6.400493938772536e-05}
{"train_loss": 0.16888532042503357, "global_step": 111490, "epoch": 1252, "lr": 6.400438285635195e-05}
{"train_loss": 0.10334376245737076, "global_step": 111491, "epoch": 1252, "lr": 6.400382632309582e-05}
{"train_loss": 0.10838967561721802, "global_step": 111492, "epoch": 1252, "lr": 6.400326978795706e-05}
{"train_loss": 0.11598248779773712, "global_step": 111493, "epoch": 1252, "lr": 6.400271325093573e-05}
{"train_loss": 0.1347939819097519, "global_step": 111494, "epoch": 1252, "lr": 6.40021567120319e-05}
{"train_loss": 0.11683713644742966, "global_step": 111495, "epoch": 1252, "lr": 6.400160017124565e-05}
{"train_loss": 0.17709657549858093, "global_step": 111496, "epoch": 1252, "lr": 6.400104362857707e-05}
{"train_loss": 0.1911797672510147, "global_step": 111497, "epoch": 1252, "lr": 6.40004870840262e-05}
{"train_loss": 0.08135712891817093, "global_step": 111498, "epoch": 1252, "lr": 6.399993053759314e-05}
{"train_loss": 0.09471163153648376, "global_step": 111499, "epoch": 1252, "lr": 6.399937398927796e-05}
{"train_loss": 0.06595215201377869, "global_step": 111500, "epoch": 1252, "lr": 6.399881743908073e-05}
{"train_loss": 0.14454004168510437, "global_step": 111501, "epoch": 1252, "lr": 6.399826088700154e-05}
{"train_loss": 0.13800205290317535, "global_step": 111502, "epoch": 1252, "lr": 6.399770433304046e-05}
{"train_loss": 0.1330290287733078, "global_step": 111503, "epoch": 1252, "lr": 6.399714777719755e-05}
{"train_loss": 0.11563722789287567, "global_step": 111504, "epoch": 1252, "lr": 6.39965912194729e-05}
{"train_loss": 0.11049443483352661, "global_step": 111505, "epoch": 1252, "lr": 6.399603465986657e-05}
{"train_loss": 0.19283294677734375, "global_step": 111506, "epoch": 1252, "lr": 6.399547809837865e-05}
{"train_loss": 0.16406665742397308, "global_step": 111507, "epoch": 1252, "lr": 6.399492153500922e-05}
{"train_loss": 0.2097187638282776, "global_step": 111508, "epoch": 1252, "lr": 6.399436496975832e-05}
{"train_loss": 0.14172610640525818, "global_step": 111509, "epoch": 1252, "lr": 6.399380840262606e-05}
{"train_loss": 0.1947331428527832, "global_step": 111510, "epoch": 1252, "lr": 6.39932518336125e-05}
{"train_loss": 0.14665517210960388, "global_step": 111511, "epoch": 1252, "lr": 6.399269526271771e-05}
{"train_loss": 0.14296600222587585, "global_step": 111512, "epoch": 1252, "lr": 6.399213868994178e-05}
{"train_loss": 0.12077142298221588, "global_step": 111513, "epoch": 1252, "lr": 6.399158211528479e-05}
{"train_loss": 0.19687791168689728, "global_step": 111514, "epoch": 1252, "lr": 6.399102553874678e-05}
{"train_loss": 0.1161050871014595, "global_step": 111515, "epoch": 1252, "lr": 6.399046896032786e-05}
{"train_loss": 0.14426993302414926, "global_step": 111516, "epoch": 1252, "lr": 6.39899123800281e-05, "val_loss": 5.193129539489746}
{"train_loss": 0.11855234950780869, "global_step": 111517, "epoch": 1253, "lr": 6.398935579784755e-05}
{"train_loss": 0.21257582306861877, "global_step": 111518, "epoch": 1253, "lr": 6.398879921378631e-05}
{"train_loss": 0.19640520215034485, "global_step": 111519, "epoch": 1253, "lr": 6.398824262784444e-05}
{"train_loss": 0.11055140942335129, "global_step": 111520, "epoch": 1253, "lr": 6.398768604002203e-05}
{"train_loss": 0.13468700647354126, "global_step": 111521, "epoch": 1253, "lr": 6.398712945031915e-05}
{"train_loss": 0.11989860236644745, "global_step": 111522, "epoch": 1253, "lr": 6.398657285873586e-05}
{"train_loss": 0.16871130466461182, "global_step": 111523, "epoch": 1253, "lr": 6.398601626527226e-05}
{"train_loss": 0.15245135128498077, "global_step": 111524, "epoch": 1253, "lr": 6.398545966992839e-05}
{"train_loss": 0.0744682177901268, "global_step": 111525, "epoch": 1253, "lr": 6.398490307270435e-05}
{"train_loss": 0.09079142659902573, "global_step": 111526, "epoch": 1253, "lr": 6.398434647360022e-05}
{"train_loss": 0.19459353387355804, "global_step": 111527, "epoch": 1253, "lr": 6.398378987261606e-05}
{"train_loss": 0.16091541945934296, "global_step": 111528, "epoch": 1253, "lr": 6.398323326975196e-05}
{"train_loss": 0.149327352643013, "global_step": 111529, "epoch": 1253, "lr": 6.398267666500797e-05}
{"train_loss": 0.19902484118938446, "global_step": 111530, "epoch": 1253, "lr": 6.398212005838417e-05}
{"train_loss": 0.10258475691080093, "global_step": 111531, "epoch": 1253, "lr": 6.398156344988066e-05}
{"train_loss": 0.11164729297161102, "global_step": 111532, "epoch": 1253, "lr": 6.39810068394975e-05}
{"train_loss": 0.07516706734895706, "global_step": 111533, "epoch": 1253, "lr": 6.398045022723477e-05}
{"train_loss": 0.09440881013870239, "global_step": 111534, "epoch": 1253, "lr": 6.397989361309253e-05}
{"train_loss": 0.16455817222595215, "global_step": 111535, "epoch": 1253, "lr": 6.397933699707087e-05}
{"train_loss": 0.08485350757837296, "global_step": 111536, "epoch": 1253, "lr": 6.397878037916984e-05}
{"train_loss": 0.1577901840209961, "global_step": 111537, "epoch": 1253, "lr": 6.397822375938956e-05}
{"train_loss": 0.12142299115657806, "global_step": 111538, "epoch": 1253, "lr": 6.397766713773006e-05}
{"train_loss": 0.19113796949386597, "global_step": 111539, "epoch": 1253, "lr": 6.397711051419144e-05}
{"train_loss": 0.17438168823719025, "global_step": 111540, "epoch": 1253, "lr": 6.397655388877377e-05}
{"train_loss": 0.1765587329864502, "global_step": 111541, "epoch": 1253, "lr": 6.397599726147712e-05}
{"train_loss": 0.09726451337337494, "global_step": 111542, "epoch": 1253, "lr": 6.397544063230157e-05}
{"train_loss": 0.1234474629163742, "global_step": 111543, "epoch": 1253, "lr": 6.397488400124718e-05}
{"train_loss": 0.1409163624048233, "global_step": 111544, "epoch": 1253, "lr": 6.397432736831405e-05}
{"train_loss": 0.09794687479734421, "global_step": 111545, "epoch": 1253, "lr": 6.397377073350223e-05}
{"train_loss": 0.1436241716146469, "global_step": 111546, "epoch": 1253, "lr": 6.39732140968118e-05}
{"train_loss": 0.17077228426933289, "global_step": 111547, "epoch": 1253, "lr": 6.397265745824287e-05}
{"train_loss": 0.2277262657880783, "global_step": 111548, "epoch": 1253, "lr": 6.397210081779548e-05}
{"train_loss": 0.13542547821998596, "global_step": 111549, "epoch": 1253, "lr": 6.39715441754697e-05}
{"train_loss": 0.11193013936281204, "global_step": 111550, "epoch": 1253, "lr": 6.397098753126562e-05}
{"train_loss": 0.13679872453212738, "global_step": 111551, "epoch": 1253, "lr": 6.397043088518331e-05}
{"train_loss": 0.212848499417305, "global_step": 111552, "epoch": 1253, "lr": 6.396987423722285e-05}
{"train_loss": 0.12202753126621246, "global_step": 111553, "epoch": 1253, "lr": 6.39693175873843e-05}
{"train_loss": 0.14714285731315613, "global_step": 111554, "epoch": 1253, "lr": 6.396876093566776e-05}
{"train_loss": 0.16757820546627045, "global_step": 111555, "epoch": 1253, "lr": 6.396820428207328e-05}
{"train_loss": 0.18717508018016815, "global_step": 111556, "epoch": 1253, "lr": 6.396764762660094e-05}
{"train_loss": 0.16324637830257416, "global_step": 111557, "epoch": 1253, "lr": 6.396709096925084e-05}
{"train_loss": 0.14543846249580383, "global_step": 111558, "epoch": 1253, "lr": 6.396653431002301e-05}
{"train_loss": 0.22318601608276367, "global_step": 111559, "epoch": 1253, "lr": 6.396597764891758e-05}
{"train_loss": 0.15865813195705414, "global_step": 111560, "epoch": 1253, "lr": 6.396542098593458e-05}
{"train_loss": 0.2007395327091217, "global_step": 111561, "epoch": 1253, "lr": 6.396486432107409e-05}
{"train_loss": 0.12390372157096863, "global_step": 111562, "epoch": 1253, "lr": 6.39643076543362e-05}
{"train_loss": 0.1271475851535797, "global_step": 111563, "epoch": 1253, "lr": 6.396375098572097e-05}
{"train_loss": 0.13263902068138123, "global_step": 111564, "epoch": 1253, "lr": 6.39631943152285e-05}
{"train_loss": 0.20093733072280884, "global_step": 111565, "epoch": 1253, "lr": 6.396263764285884e-05}
{"train_loss": 0.13197366893291473, "global_step": 111566, "epoch": 1253, "lr": 6.396208096861209e-05}
{"train_loss": 0.11255178600549698, "global_step": 111567, "epoch": 1253, "lr": 6.39615242924883e-05}
{"train_loss": 0.1527443379163742, "global_step": 111568, "epoch": 1253, "lr": 6.396096761448754e-05}
{"train_loss": 0.23064593970775604, "global_step": 111569, "epoch": 1253, "lr": 6.39604109346099e-05}
{"train_loss": 0.14951208233833313, "global_step": 111570, "epoch": 1253, "lr": 6.395985425285547e-05}
{"train_loss": 0.08451743423938751, "global_step": 111571, "epoch": 1253, "lr": 6.39592975692243e-05}
{"train_loss": 0.17372459173202515, "global_step": 111572, "epoch": 1253, "lr": 6.395874088371646e-05}
{"train_loss": 0.183877095580101, "global_step": 111573, "epoch": 1253, "lr": 6.395818419633206e-05}
{"train_loss": 0.16092365980148315, "global_step": 111574, "epoch": 1253, "lr": 6.395762750707112e-05}
{"train_loss": 0.12319125980138779, "global_step": 111575, "epoch": 1253, "lr": 6.395707081593378e-05}
{"train_loss": 0.09212338179349899, "global_step": 111576, "epoch": 1253, "lr": 6.395651412292007e-05}
{"train_loss": 0.12572430074214935, "global_step": 111577, "epoch": 1253, "lr": 6.395595742803007e-05}
{"train_loss": 0.1484907865524292, "global_step": 111578, "epoch": 1253, "lr": 6.395540073126385e-05}
{"train_loss": 0.15093886852264404, "global_step": 111579, "epoch": 1253, "lr": 6.395484403262154e-05}
{"train_loss": 0.104751355946064, "global_step": 111580, "epoch": 1253, "lr": 6.395428733210313e-05}
{"train_loss": 0.07608991861343384, "global_step": 111581, "epoch": 1253, "lr": 6.395373062970874e-05}
{"train_loss": 0.17715314030647278, "global_step": 111582, "epoch": 1253, "lr": 6.395317392543845e-05}
{"train_loss": 0.10317336022853851, "global_step": 111583, "epoch": 1253, "lr": 6.395261721929233e-05}
{"train_loss": 0.135903999209404, "global_step": 111584, "epoch": 1253, "lr": 6.395206051127045e-05}
{"train_loss": 0.13521763682365417, "global_step": 111585, "epoch": 1253, "lr": 6.395150380137289e-05}
{"train_loss": 0.12644879519939423, "global_step": 111586, "epoch": 1253, "lr": 6.395094708959971e-05}
{"train_loss": 0.13072657585144043, "global_step": 111587, "epoch": 1253, "lr": 6.3950390375951e-05}
{"train_loss": 0.15140658617019653, "global_step": 111588, "epoch": 1253, "lr": 6.394983366042683e-05}
{"train_loss": 0.17904454469680786, "global_step": 111589, "epoch": 1253, "lr": 6.394927694302727e-05}
{"train_loss": 0.17988541722297668, "global_step": 111590, "epoch": 1253, "lr": 6.394872022375241e-05}
{"train_loss": 0.1762259304523468, "global_step": 111591, "epoch": 1253, "lr": 6.39481635026023e-05}
{"train_loss": 0.189215287566185, "global_step": 111592, "epoch": 1253, "lr": 6.394760677957704e-05}
{"train_loss": 0.1066124364733696, "global_step": 111593, "epoch": 1253, "lr": 6.394705005467669e-05}
{"train_loss": 0.1317511349916458, "global_step": 111594, "epoch": 1253, "lr": 6.394649332790133e-05}
{"train_loss": 0.18316686153411865, "global_step": 111595, "epoch": 1253, "lr": 6.394593659925105e-05}
{"train_loss": 0.1766519695520401, "global_step": 111596, "epoch": 1253, "lr": 6.394537986872588e-05}
{"train_loss": 0.1561266928911209, "global_step": 111597, "epoch": 1253, "lr": 6.394482313632594e-05}
{"train_loss": 0.22107475996017456, "global_step": 111598, "epoch": 1253, "lr": 6.39442664020513e-05}
{"train_loss": 0.14260238409042358, "global_step": 111599, "epoch": 1253, "lr": 6.394370966590198e-05}
{"train_loss": 0.1610736846923828, "global_step": 111600, "epoch": 1253, "lr": 6.394315292787813e-05}
{"train_loss": 0.18225136399269104, "global_step": 111601, "epoch": 1253, "lr": 6.39425961879798e-05}
{"train_loss": 0.10614404082298279, "global_step": 111602, "epoch": 1253, "lr": 6.394203944620705e-05}
{"train_loss": 0.16005924344062805, "global_step": 111603, "epoch": 1253, "lr": 6.394148270255995e-05}
{"train_loss": 0.14986418187618256, "global_step": 111604, "epoch": 1253, "lr": 6.39409259570386e-05}
{"train_loss": 0.14749182961630017, "global_step": 111605, "epoch": 1253, "lr": 6.394036920964308e-05, "val_loss": 5.043538570404053}
{"train_loss": 0.10492420941591263, "global_step": 111606, "epoch": 1254, "lr": 6.393981246037343e-05}
{"train_loss": 0.17460636794567108, "global_step": 111607, "epoch": 1254, "lr": 6.393925570922974e-05}
{"train_loss": 0.11659453064203262, "global_step": 111608, "epoch": 1254, "lr": 6.393869895621208e-05}
{"train_loss": 0.13645131886005402, "global_step": 111609, "epoch": 1254, "lr": 6.393814220132054e-05}
{"train_loss": 0.10263326019048691, "global_step": 111610, "epoch": 1254, "lr": 6.393758544455519e-05}
{"train_loss": 0.1313735395669937, "global_step": 111611, "epoch": 1254, "lr": 6.393702868591609e-05}
{"train_loss": 0.1648675799369812, "global_step": 111612, "epoch": 1254, "lr": 6.393647192540334e-05}
{"train_loss": 0.1317884474992752, "global_step": 111613, "epoch": 1254, "lr": 6.393591516301699e-05}
{"train_loss": 0.10997480899095535, "global_step": 111614, "epoch": 1254, "lr": 6.393535839875713e-05}
{"train_loss": 0.19469749927520752, "global_step": 111615, "epoch": 1254, "lr": 6.393480163262384e-05}
{"train_loss": 0.10283081978559494, "global_step": 111616, "epoch": 1254, "lr": 6.393424486461718e-05}
{"train_loss": 0.11624710261821747, "global_step": 111617, "epoch": 1254, "lr": 6.393368809473721e-05}
{"train_loss": 0.09755383431911469, "global_step": 111618, "epoch": 1254, "lr": 6.393313132298403e-05}
{"train_loss": 0.12523795664310455, "global_step": 111619, "epoch": 1254, "lr": 6.393257454935774e-05}
{"train_loss": 0.13537262380123138, "global_step": 111620, "epoch": 1254, "lr": 6.393201777385836e-05}
{"train_loss": 0.24093405902385712, "global_step": 111621, "epoch": 1254, "lr": 6.393146099648602e-05}
{"train_loss": 0.14877906441688538, "global_step": 111622, "epoch": 1254, "lr": 6.393090421724073e-05}
{"train_loss": 0.13357585668563843, "global_step": 111623, "epoch": 1254, "lr": 6.393034743612262e-05}
{"train_loss": 0.13194714486598969, "global_step": 111624, "epoch": 1254, "lr": 6.392979065313176e-05}
{"train_loss": 0.10537080466747284, "global_step": 111625, "epoch": 1254, "lr": 6.392923386826817e-05}
{"train_loss": 0.1993500292301178, "global_step": 111626, "epoch": 1254, "lr": 6.392867708153198e-05}
{"train_loss": 0.22408539056777954, "global_step": 111627, "epoch": 1254, "lr": 6.392812029292325e-05}
{"train_loss": 0.1261695772409439, "global_step": 111628, "epoch": 1254, "lr": 6.392756350244207e-05}
{"train_loss": 0.19353340566158295, "global_step": 111629, "epoch": 1254, "lr": 6.392700671008847e-05}
{"train_loss": 0.12954846024513245, "global_step": 111630, "epoch": 1254, "lr": 6.392644991586258e-05}
{"train_loss": 0.12968949973583221, "global_step": 111631, "epoch": 1254, "lr": 6.392589311976443e-05}
{"train_loss": 0.08230828493833542, "global_step": 111632, "epoch": 1254, "lr": 6.392533632179411e-05}
{"train_loss": 0.19481679797172546, "global_step": 111633, "epoch": 1254, "lr": 6.392477952195171e-05}
{"train_loss": 0.19295431673526764, "global_step": 111634, "epoch": 1254, "lr": 6.39242227202373e-05}
{"train_loss": 0.11683934181928635, "global_step": 111635, "epoch": 1254, "lr": 6.392366591665093e-05}
{"train_loss": 0.16113466024398804, "global_step": 111636, "epoch": 1254, "lr": 6.39231091111927e-05}
{"train_loss": 0.18062125146389008, "global_step": 111637, "epoch": 1254, "lr": 6.392255230386268e-05}
{"train_loss": 0.14564263820648193, "global_step": 111638, "epoch": 1254, "lr": 6.392199549466095e-05}
{"train_loss": 0.13817234337329865, "global_step": 111639, "epoch": 1254, "lr": 6.392143868358757e-05}
{"train_loss": 0.09738945215940475, "global_step": 111640, "epoch": 1254, "lr": 6.392088187064262e-05}
{"train_loss": 0.1505356878042221, "global_step": 111641, "epoch": 1254, "lr": 6.392032505582617e-05}
{"train_loss": 0.1326552778482437, "global_step": 111642, "epoch": 1254, "lr": 6.391976823913832e-05}
{"train_loss": 0.16414707899093628, "global_step": 111643, "epoch": 1254, "lr": 6.391921142057913e-05}
{"train_loss": 0.15864570438861847, "global_step": 111644, "epoch": 1254, "lr": 6.391865460014866e-05}
{"train_loss": 0.15085425972938538, "global_step": 111645, "epoch": 1254, "lr": 6.391809777784698e-05}
{"train_loss": 0.13305726647377014, "global_step": 111646, "epoch": 1254, "lr": 6.391754095367421e-05}
{"train_loss": 0.14405673742294312, "global_step": 111647, "epoch": 1254, "lr": 6.391698412763037e-05}
{"train_loss": 0.14195337891578674, "global_step": 111648, "epoch": 1254, "lr": 6.391642729971559e-05}
{"train_loss": 0.12687133252620697, "global_step": 111649, "epoch": 1254, "lr": 6.391587046992989e-05}
{"train_loss": 0.19943130016326904, "global_step": 111650, "epoch": 1254, "lr": 6.39153136382734e-05}
{"train_loss": 0.12702567875385284, "global_step": 111651, "epoch": 1254, "lr": 6.391475680474613e-05}
{"train_loss": 0.13034561276435852, "global_step": 111652, "epoch": 1254, "lr": 6.391419996934821e-05}
{"train_loss": 0.12957701086997986, "global_step": 111653, "epoch": 1254, "lr": 6.391364313207969e-05}
{"train_loss": 0.11196194589138031, "global_step": 111654, "epoch": 1254, "lr": 6.391308629294064e-05}
{"train_loss": 0.14333418011665344, "global_step": 111655, "epoch": 1254, "lr": 6.391252945193116e-05}
{"train_loss": 0.10848188400268555, "global_step": 111656, "epoch": 1254, "lr": 6.391197260905131e-05}
{"train_loss": 0.14307108521461487, "global_step": 111657, "epoch": 1254, "lr": 6.391141576430117e-05}
{"train_loss": 0.13249625265598297, "global_step": 111658, "epoch": 1254, "lr": 6.391085891768081e-05}
{"train_loss": 0.22235502302646637, "global_step": 111659, "epoch": 1254, "lr": 6.391030206919029e-05}
{"train_loss": 0.10947620868682861, "global_step": 111660, "epoch": 1254, "lr": 6.390974521882971e-05}
{"train_loss": 0.14686231315135956, "global_step": 111661, "epoch": 1254, "lr": 6.390918836659912e-05}
{"train_loss": 0.15651938319206238, "global_step": 111662, "epoch": 1254, "lr": 6.390863151249862e-05}
{"train_loss": 0.13581609725952148, "global_step": 111663, "epoch": 1254, "lr": 6.39080746565283e-05}
{"train_loss": 0.08507722616195679, "global_step": 111664, "epoch": 1254, "lr": 6.390751779868817e-05}
{"train_loss": 0.17825402319431305, "global_step": 111665, "epoch": 1254, "lr": 6.390696093897837e-05}
{"train_loss": 0.09915637969970703, "global_step": 111666, "epoch": 1254, "lr": 6.390640407739892e-05}
{"train_loss": 0.22454313933849335, "global_step": 111667, "epoch": 1254, "lr": 6.390584721394995e-05}
{"train_loss": 0.15853077173233032, "global_step": 111668, "epoch": 1254, "lr": 6.39052903486315e-05}
{"train_loss": 0.15002447366714478, "global_step": 111669, "epoch": 1254, "lr": 6.390473348144365e-05}
{"train_loss": 0.13183090090751648, "global_step": 111670, "epoch": 1254, "lr": 6.390417661238648e-05}
{"train_loss": 0.16881011426448822, "global_step": 111671, "epoch": 1254, "lr": 6.390361974146006e-05}
{"train_loss": 0.1268898844718933, "global_step": 111672, "epoch": 1254, "lr": 6.390306286866446e-05}
{"train_loss": 0.16375912725925446, "global_step": 111673, "epoch": 1254, "lr": 6.390250599399976e-05}
{"train_loss": 0.12060563266277313, "global_step": 111674, "epoch": 1254, "lr": 6.390194911746605e-05}
{"train_loss": 0.16086600720882416, "global_step": 111675, "epoch": 1254, "lr": 6.39013922390634e-05}
{"train_loss": 0.14142723381519318, "global_step": 111676, "epoch": 1254, "lr": 6.390083535879187e-05}
{"train_loss": 0.15870638191699982, "global_step": 111677, "epoch": 1254, "lr": 6.390027847665154e-05}
{"train_loss": 0.1195550262928009, "global_step": 111678, "epoch": 1254, "lr": 6.389972159264248e-05}
{"train_loss": 0.12511326372623444, "global_step": 111679, "epoch": 1254, "lr": 6.389916470676478e-05}
{"train_loss": 0.13726811110973358, "global_step": 111680, "epoch": 1254, "lr": 6.389860781901851e-05}
{"train_loss": 0.12605907022953033, "global_step": 111681, "epoch": 1254, "lr": 6.389805092940374e-05}
{"train_loss": 0.17529603838920593, "global_step": 111682, "epoch": 1254, "lr": 6.389749403792056e-05}
{"train_loss": 0.13664771616458893, "global_step": 111683, "epoch": 1254, "lr": 6.389693714456902e-05}
{"train_loss": 0.10371451079845428, "global_step": 111684, "epoch": 1254, "lr": 6.389638024934919e-05}
{"train_loss": 0.16044966876506805, "global_step": 111685, "epoch": 1254, "lr": 6.389582335226119e-05}
{"train_loss": 0.1441221982240677, "global_step": 111686, "epoch": 1254, "lr": 6.389526645330504e-05}
{"train_loss": 0.2034386843442917, "global_step": 111687, "epoch": 1254, "lr": 6.389470955248085e-05}
{"train_loss": 0.08175268024206161, "global_step": 111688, "epoch": 1254, "lr": 6.389415264978869e-05}
{"train_loss": 0.1366601437330246, "global_step": 111689, "epoch": 1254, "lr": 6.389359574522861e-05}
{"train_loss": 0.14472529292106628, "global_step": 111690, "epoch": 1254, "lr": 6.389303883880073e-05}
{"train_loss": 0.1252995729446411, "global_step": 111691, "epoch": 1254, "lr": 6.389248193050508e-05}
{"train_loss": 0.12624213099479675, "global_step": 111692, "epoch": 1254, "lr": 6.389192502034177e-05}
{"train_loss": 0.15921446681022644, "global_step": 111693, "epoch": 1254, "lr": 6.389136810831086e-05}
{"train_loss": 0.14337413473410554, "global_step": 111694, "epoch": 1254, "lr": 6.389081119441242e-05, "val_loss": 5.235759735107422}
{"train_loss": 0.173181414604187, "global_step": 111695, "epoch": 1255, "lr": 6.389025427864653e-05}
{"train_loss": 0.13733966648578644, "global_step": 111696, "epoch": 1255, "lr": 6.388969736101327e-05}
{"train_loss": 0.14515556395053864, "global_step": 111697, "epoch": 1255, "lr": 6.38891404415127e-05}
{"train_loss": 0.10526126623153687, "global_step": 111698, "epoch": 1255, "lr": 6.388858352014492e-05}
{"train_loss": 0.20310334861278534, "global_step": 111699, "epoch": 1255, "lr": 6.388802659690997e-05}
{"train_loss": 0.15471646189689636, "global_step": 111700, "epoch": 1255, "lr": 6.388746967180795e-05}
{"train_loss": 0.10983248054981232, "global_step": 111701, "epoch": 1255, "lr": 6.388691274483896e-05}
{"train_loss": 0.1580182909965515, "global_step": 111702, "epoch": 1255, "lr": 6.388635581600301e-05}
{"train_loss": 0.09182355552911758, "global_step": 111703, "epoch": 1255, "lr": 6.388579888530021e-05}
{"train_loss": 0.2835833430290222, "global_step": 111704, "epoch": 1255, "lr": 6.388524195273064e-05}
{"train_loss": 0.1385071724653244, "global_step": 111705, "epoch": 1255, "lr": 6.388468501829437e-05}
{"train_loss": 0.07319086045026779, "global_step": 111706, "epoch": 1255, "lr": 6.388412808199146e-05}
{"train_loss": 0.21520943939685822, "global_step": 111707, "epoch": 1255, "lr": 6.388357114382201e-05}
{"train_loss": 0.14736244082450867, "global_step": 111708, "epoch": 1255, "lr": 6.388301420378607e-05}
{"train_loss": 0.1552392989397049, "global_step": 111709, "epoch": 1255, "lr": 6.388245726188375e-05}
{"train_loss": 0.13115231692790985, "global_step": 111710, "epoch": 1255, "lr": 6.38819003181151e-05}
{"train_loss": 0.14211952686309814, "global_step": 111711, "epoch": 1255, "lr": 6.388134337248017e-05}
{"train_loss": 0.12205774337053299, "global_step": 111712, "epoch": 1255, "lr": 6.388078642497908e-05}
{"train_loss": 0.17939415574073792, "global_step": 111713, "epoch": 1255, "lr": 6.38802294756119e-05}
{"train_loss": 0.20005211234092712, "global_step": 111714, "epoch": 1255, "lr": 6.387967252437869e-05}
{"train_loss": 0.18392813205718994, "global_step": 111715, "epoch": 1255, "lr": 6.387911557127952e-05}
{"train_loss": 0.1525917798280716, "global_step": 111716, "epoch": 1255, "lr": 6.387855861631446e-05}
{"train_loss": 0.14865140616893768, "global_step": 111717, "epoch": 1255, "lr": 6.387800165948361e-05}
{"train_loss": 0.07298705726861954, "global_step": 111718, "epoch": 1255, "lr": 6.387744470078704e-05}
{"train_loss": 0.2120460718870163, "global_step": 111719, "epoch": 1255, "lr": 6.38768877402248e-05}
{"train_loss": 0.1834966391324997, "global_step": 111720, "epoch": 1255, "lr": 6.3876330777797e-05}
{"train_loss": 0.1269838809967041, "global_step": 111721, "epoch": 1255, "lr": 6.387577381350369e-05}
{"train_loss": 0.19401773810386658, "global_step": 111722, "epoch": 1255, "lr": 6.387521684734494e-05}
{"train_loss": 0.1175151914358139, "global_step": 111723, "epoch": 1255, "lr": 6.387465987932084e-05}
{"train_loss": 0.08553445339202881, "global_step": 111724, "epoch": 1255, "lr": 6.387410290943148e-05}
{"train_loss": 0.11288445442914963, "global_step": 111725, "epoch": 1255, "lr": 6.387354593767689e-05}
{"train_loss": 0.17665790021419525, "global_step": 111726, "epoch": 1255, "lr": 6.387298896405717e-05}
{"train_loss": 0.17017319798469543, "global_step": 111727, "epoch": 1255, "lr": 6.38724319885724e-05}
{"train_loss": 0.10775545239448547, "global_step": 111728, "epoch": 1255, "lr": 6.387187501122267e-05}
{"train_loss": 0.1511622667312622, "global_step": 111729, "epoch": 1255, "lr": 6.387131803200802e-05}
{"train_loss": 0.15136146545410156, "global_step": 111730, "epoch": 1255, "lr": 6.387076105092855e-05}
{"train_loss": 0.12727844715118408, "global_step": 111731, "epoch": 1255, "lr": 6.387020406798431e-05}
{"train_loss": 0.15411776304244995, "global_step": 111732, "epoch": 1255, "lr": 6.38696470831754e-05}
{"train_loss": 0.193729966878891, "global_step": 111733, "epoch": 1255, "lr": 6.386909009650188e-05}
{"train_loss": 0.19404000043869019, "global_step": 111734, "epoch": 1255, "lr": 6.386853310796384e-05}
{"train_loss": 0.1262756735086441, "global_step": 111735, "epoch": 1255, "lr": 6.386797611756134e-05}
{"train_loss": 0.10964358597993851, "global_step": 111736, "epoch": 1255, "lr": 6.386741912529445e-05}
{"train_loss": 0.15873843431472778, "global_step": 111737, "epoch": 1255, "lr": 6.386686213116327e-05}
{"train_loss": 0.12126365303993225, "global_step": 111738, "epoch": 1255, "lr": 6.386630513516785e-05}
{"train_loss": 0.1736670732498169, "global_step": 111739, "epoch": 1255, "lr": 6.386574813730826e-05}
{"train_loss": 0.11542531102895737, "global_step": 111740, "epoch": 1255, "lr": 6.386519113758463e-05}
{"train_loss": 0.19681565463542938, "global_step": 111741, "epoch": 1255, "lr": 6.386463413599697e-05}
{"train_loss": 0.13931037485599518, "global_step": 111742, "epoch": 1255, "lr": 6.386407713254537e-05}
{"train_loss": 0.1313491314649582, "global_step": 111743, "epoch": 1255, "lr": 6.386352012722991e-05}
{"train_loss": 0.10749845951795578, "global_step": 111744, "epoch": 1255, "lr": 6.386296312005068e-05}
{"train_loss": 0.25761324167251587, "global_step": 111745, "epoch": 1255, "lr": 6.386240611100774e-05}
{"train_loss": 0.12718719244003296, "global_step": 111746, "epoch": 1255, "lr": 6.386184910010117e-05}
{"train_loss": 0.2114764004945755, "global_step": 111747, "epoch": 1255, "lr": 6.386129208733105e-05}
{"train_loss": 0.18490487337112427, "global_step": 111748, "epoch": 1255, "lr": 6.386073507269746e-05}
{"train_loss": 0.18478624522686005, "global_step": 111749, "epoch": 1255, "lr": 6.386017805620044e-05}
{"train_loss": 0.1670716255903244, "global_step": 111750, "epoch": 1255, "lr": 6.385962103784009e-05}
{"train_loss": 0.0773337334394455, "global_step": 111751, "epoch": 1255, "lr": 6.385906401761648e-05}
{"train_loss": 0.11594324558973312, "global_step": 111752, "epoch": 1255, "lr": 6.38585069955297e-05}
{"train_loss": 0.1688096821308136, "global_step": 111753, "epoch": 1255, "lr": 6.385794997157982e-05}
{"train_loss": 0.19647224247455597, "global_step": 111754, "epoch": 1255, "lr": 6.385739294576689e-05}
{"train_loss": 0.11085187643766403, "global_step": 111755, "epoch": 1255, "lr": 6.3856835918091e-05}
{"train_loss": 0.11695410311222076, "global_step": 111756, "epoch": 1255, "lr": 6.385627888855223e-05}
{"train_loss": 0.10916436463594437, "global_step": 111757, "epoch": 1255, "lr": 6.385572185715068e-05}
{"train_loss": 0.09522828459739685, "global_step": 111758, "epoch": 1255, "lr": 6.385516482388636e-05}
{"train_loss": 0.22478953003883362, "global_step": 111759, "epoch": 1255, "lr": 6.385460778875941e-05}
{"train_loss": 0.2648470103740692, "global_step": 111760, "epoch": 1255, "lr": 6.385405075176987e-05}
{"train_loss": 0.1549389660358429, "global_step": 111761, "epoch": 1255, "lr": 6.385349371291781e-05}
{"train_loss": 0.17698843777179718, "global_step": 111762, "epoch": 1255, "lr": 6.385293667220331e-05}
{"train_loss": 0.16581740975379944, "global_step": 111763, "epoch": 1255, "lr": 6.385237962962647e-05}
{"train_loss": 0.1472025364637375, "global_step": 111764, "epoch": 1255, "lr": 6.385182258518734e-05}
{"train_loss": 0.10922233760356903, "global_step": 111765, "epoch": 1255, "lr": 6.3851265538886e-05}
{"train_loss": 0.10230404138565063, "global_step": 111766, "epoch": 1255, "lr": 6.385070849072252e-05}
{"train_loss": 0.12880466878414154, "global_step": 111767, "epoch": 1255, "lr": 6.3850151440697e-05}
{"train_loss": 0.14149828255176544, "global_step": 111768, "epoch": 1255, "lr": 6.384959438880947e-05}
{"train_loss": 0.1552807092666626, "global_step": 111769, "epoch": 1255, "lr": 6.384903733506004e-05}
{"train_loss": 0.1992553472518921, "global_step": 111770, "epoch": 1255, "lr": 6.384848027944878e-05}
{"train_loss": 0.1214071661233902, "global_step": 111771, "epoch": 1255, "lr": 6.384792322197577e-05}
{"train_loss": 0.17252688109874725, "global_step": 111772, "epoch": 1255, "lr": 6.384736616264106e-05}
{"train_loss": 0.09349352866411209, "global_step": 111773, "epoch": 1255, "lr": 6.384680910144476e-05}
{"train_loss": 0.1187913566827774, "global_step": 111774, "epoch": 1255, "lr": 6.384625203838692e-05}
{"train_loss": 0.11928851902484894, "global_step": 111775, "epoch": 1255, "lr": 6.38456949734676e-05}
{"train_loss": 0.09956707060337067, "global_step": 111776, "epoch": 1255, "lr": 6.384513790668691e-05}
{"train_loss": 0.11381105333566666, "global_step": 111777, "epoch": 1255, "lr": 6.38445808380449e-05}
{"train_loss": 0.1480596959590912, "global_step": 111778, "epoch": 1255, "lr": 6.384402376754167e-05}
{"train_loss": 0.18425306677818298, "global_step": 111779, "epoch": 1255, "lr": 6.384346669517729e-05}
{"train_loss": 0.13741804659366608, "global_step": 111780, "epoch": 1255, "lr": 6.38429096209518e-05}
{"train_loss": 0.14155875146389008, "global_step": 111781, "epoch": 1255, "lr": 6.38423525448653e-05}
{"train_loss": 0.11173302680253983, "global_step": 111782, "epoch": 1255, "lr": 6.384179546691787e-05}
{"train_loss": 0.14916374308339667, "global_step": 111783, "epoch": 1255, "lr": 6.384123838710959e-05, "val_loss": 5.188758373260498, "train_action_mse_error": 7.450920104980469}
{"train_loss": 0.11556128412485123, "global_step": 111784, "epoch": 1256, "lr": 6.38406813054405e-05}
{"train_loss": 0.15013808012008667, "global_step": 111785, "epoch": 1256, "lr": 6.384012422191072e-05}
{"train_loss": 0.08977435529232025, "global_step": 111786, "epoch": 1256, "lr": 6.383956713652029e-05}
{"train_loss": 0.1000591591000557, "global_step": 111787, "epoch": 1256, "lr": 6.383901004926932e-05}
{"train_loss": 0.12645253539085388, "global_step": 111788, "epoch": 1256, "lr": 6.383845296015785e-05}
{"train_loss": 0.12464972585439682, "global_step": 111789, "epoch": 1256, "lr": 6.383789586918596e-05}
{"train_loss": 0.13844645023345947, "global_step": 111790, "epoch": 1256, "lr": 6.383733877635374e-05}
{"train_loss": 0.10799524933099747, "global_step": 111791, "epoch": 1256, "lr": 6.383678168166126e-05}
{"train_loss": 0.12351567298173904, "global_step": 111792, "epoch": 1256, "lr": 6.383622458510861e-05}
{"train_loss": 0.18709836900234222, "global_step": 111793, "epoch": 1256, "lr": 6.383566748669583e-05}
{"train_loss": 0.1322280615568161, "global_step": 111794, "epoch": 1256, "lr": 6.3835110386423e-05}
{"train_loss": 0.13235141336917877, "global_step": 111795, "epoch": 1256, "lr": 6.383455328429024e-05}
{"train_loss": 0.08929000794887543, "global_step": 111796, "epoch": 1256, "lr": 6.383399618029758e-05}
{"train_loss": 0.14516368508338928, "global_step": 111797, "epoch": 1256, "lr": 6.38334390744451e-05}
{"train_loss": 0.20427340269088745, "global_step": 111798, "epoch": 1256, "lr": 6.383288196673289e-05}
{"train_loss": 0.11566013842821121, "global_step": 111799, "epoch": 1256, "lr": 6.383232485716101e-05}
{"train_loss": 0.0715428963303566, "global_step": 111800, "epoch": 1256, "lr": 6.383176774572955e-05}
{"train_loss": 0.07093339413404465, "global_step": 111801, "epoch": 1256, "lr": 6.383121063243857e-05}
{"train_loss": 0.1556754857301712, "global_step": 111802, "epoch": 1256, "lr": 6.383065351728815e-05}
{"train_loss": 0.06970416009426117, "global_step": 111803, "epoch": 1256, "lr": 6.383009640027838e-05}
{"train_loss": 0.1971246600151062, "global_step": 111804, "epoch": 1256, "lr": 6.382953928140931e-05}
{"train_loss": 0.11472181230783463, "global_step": 111805, "epoch": 1256, "lr": 6.382898216068104e-05}
{"train_loss": 0.22777314484119415, "global_step": 111806, "epoch": 1256, "lr": 6.382842503809361e-05}
{"train_loss": 0.20804844796657562, "global_step": 111807, "epoch": 1256, "lr": 6.382786791364714e-05}
{"train_loss": 0.22072896361351013, "global_step": 111808, "epoch": 1256, "lr": 6.382731078734167e-05}
{"train_loss": 0.1235976591706276, "global_step": 111809, "epoch": 1256, "lr": 6.38267536591773e-05}
{"train_loss": 0.13996471464633942, "global_step": 111810, "epoch": 1256, "lr": 6.382619652915408e-05}
{"train_loss": 0.14630579948425293, "global_step": 111811, "epoch": 1256, "lr": 6.38256393972721e-05}
{"train_loss": 0.13363149762153625, "global_step": 111812, "epoch": 1256, "lr": 6.382508226353143e-05}
{"train_loss": 0.19876611232757568, "global_step": 111813, "epoch": 1256, "lr": 6.382452512793215e-05}
{"train_loss": 0.17608192563056946, "global_step": 111814, "epoch": 1256, "lr": 6.382396799047432e-05}
{"train_loss": 0.16063150763511658, "global_step": 111815, "epoch": 1256, "lr": 6.382341085115804e-05}
{"train_loss": 0.16621659696102142, "global_step": 111816, "epoch": 1256, "lr": 6.382285370998335e-05}
{"train_loss": 0.10864223539829254, "global_step": 111817, "epoch": 1256, "lr": 6.382229656695037e-05}
{"train_loss": 0.08987021446228027, "global_step": 111818, "epoch": 1256, "lr": 6.382173942205913e-05}
{"train_loss": 0.1075512245297432, "global_step": 111819, "epoch": 1256, "lr": 6.382118227530974e-05}
{"train_loss": 0.15397819876670837, "global_step": 111820, "epoch": 1256, "lr": 6.382062512670226e-05}
{"train_loss": 0.1532861292362213, "global_step": 111821, "epoch": 1256, "lr": 6.382006797623676e-05}
{"train_loss": 0.10614541918039322, "global_step": 111822, "epoch": 1256, "lr": 6.381951082391333e-05}
{"train_loss": 0.28339245915412903, "global_step": 111823, "epoch": 1256, "lr": 6.381895366973203e-05}
{"train_loss": 0.19456054270267487, "global_step": 111824, "epoch": 1256, "lr": 6.381839651369292e-05}
{"train_loss": 0.1368435174226761, "global_step": 111825, "epoch": 1256, "lr": 6.381783935579613e-05}
{"train_loss": 0.18307273089885712, "global_step": 111826, "epoch": 1256, "lr": 6.381728219604167e-05}
{"train_loss": 0.13420912623405457, "global_step": 111827, "epoch": 1256, "lr": 6.381672503442966e-05}
{"train_loss": 0.051603589206933975, "global_step": 111828, "epoch": 1256, "lr": 6.381616787096016e-05}
{"train_loss": 0.10700689256191254, "global_step": 111829, "epoch": 1256, "lr": 6.381561070563324e-05}
{"train_loss": 0.17175599932670593, "global_step": 111830, "epoch": 1256, "lr": 6.381505353844898e-05}
{"train_loss": 0.13202737271785736, "global_step": 111831, "epoch": 1256, "lr": 6.381449636940746e-05}
{"train_loss": 0.1824292093515396, "global_step": 111832, "epoch": 1256, "lr": 6.381393919850874e-05}
{"train_loss": 0.14642108976840973, "global_step": 111833, "epoch": 1256, "lr": 6.38133820257529e-05}
{"train_loss": 0.10146862268447876, "global_step": 111834, "epoch": 1256, "lr": 6.381282485114005e-05}
{"train_loss": 0.16431964933872223, "global_step": 111835, "epoch": 1256, "lr": 6.38122676746702e-05}
{"train_loss": 0.20740728080272675, "global_step": 111836, "epoch": 1256, "lr": 6.381171049634347e-05}
{"train_loss": 0.14535295963287354, "global_step": 111837, "epoch": 1256, "lr": 6.381115331615993e-05}
{"train_loss": 0.21540716290473938, "global_step": 111838, "epoch": 1256, "lr": 6.381059613411963e-05}
{"train_loss": 0.2072104513645172, "global_step": 111839, "epoch": 1256, "lr": 6.381003895022269e-05}
{"train_loss": 0.13264745473861694, "global_step": 111840, "epoch": 1256, "lr": 6.380948176446915e-05}
{"train_loss": 0.22744089365005493, "global_step": 111841, "epoch": 1256, "lr": 6.380892457685908e-05}
{"train_loss": 0.14286920428276062, "global_step": 111842, "epoch": 1256, "lr": 6.380836738739257e-05}
{"train_loss": 0.09839729964733124, "global_step": 111843, "epoch": 1256, "lr": 6.380781019606972e-05}
{"train_loss": 0.15108630061149597, "global_step": 111844, "epoch": 1256, "lr": 6.380725300289055e-05}
{"train_loss": 0.2246372401714325, "global_step": 111845, "epoch": 1256, "lr": 6.380669580785517e-05}
{"train_loss": 0.1416252702474594, "global_step": 111846, "epoch": 1256, "lr": 6.380613861096365e-05}
{"train_loss": 0.09842447191476822, "global_step": 111847, "epoch": 1256, "lr": 6.380558141221606e-05}
{"train_loss": 0.2175471931695938, "global_step": 111848, "epoch": 1256, "lr": 6.38050242116125e-05}
{"train_loss": 0.09263568371534348, "global_step": 111849, "epoch": 1256, "lr": 6.380446700915299e-05}
{"train_loss": 0.11866642534732819, "global_step": 111850, "epoch": 1256, "lr": 6.380390980483765e-05}
{"train_loss": 0.12398319691419601, "global_step": 111851, "epoch": 1256, "lr": 6.380335259866656e-05}
{"train_loss": 0.24599140882492065, "global_step": 111852, "epoch": 1256, "lr": 6.380279539063976e-05}
{"train_loss": 0.16323818266391754, "global_step": 111853, "epoch": 1256, "lr": 6.380223818075734e-05}
{"train_loss": 0.18827633559703827, "global_step": 111854, "epoch": 1256, "lr": 6.380168096901937e-05}
{"train_loss": 0.09301789849996567, "global_step": 111855, "epoch": 1256, "lr": 6.380112375542595e-05}
{"train_loss": 0.12044549733400345, "global_step": 111856, "epoch": 1256, "lr": 6.380056653997714e-05}
{"train_loss": 0.1384650617837906, "global_step": 111857, "epoch": 1256, "lr": 6.3800009322673e-05}
{"train_loss": 0.25972697138786316, "global_step": 111858, "epoch": 1256, "lr": 6.37994521035136e-05}
{"train_loss": 0.15803523361682892, "global_step": 111859, "epoch": 1256, "lr": 6.379889488249908e-05}
{"train_loss": 0.1917751282453537, "global_step": 111860, "epoch": 1256, "lr": 6.379833765962942e-05}
{"train_loss": 0.135994091629982, "global_step": 111861, "epoch": 1256, "lr": 6.379778043490477e-05}
{"train_loss": 0.19855710864067078, "global_step": 111862, "epoch": 1256, "lr": 6.379722320832517e-05}
{"train_loss": 0.13004453480243683, "global_step": 111863, "epoch": 1256, "lr": 6.379666597989068e-05}
{"train_loss": 0.1832408756017685, "global_step": 111864, "epoch": 1256, "lr": 6.379610874960143e-05}
{"train_loss": 0.11112271249294281, "global_step": 111865, "epoch": 1256, "lr": 6.379555151745744e-05}
{"train_loss": 0.18557503819465637, "global_step": 111866, "epoch": 1256, "lr": 6.379499428345881e-05}
{"train_loss": 0.16953003406524658, "global_step": 111867, "epoch": 1256, "lr": 6.379443704760562e-05}
{"train_loss": 0.1211295872926712, "global_step": 111868, "epoch": 1256, "lr": 6.379387980989793e-05}
{"train_loss": 0.15327155590057373, "global_step": 111869, "epoch": 1256, "lr": 6.379332257033582e-05}
{"train_loss": 0.22116383910179138, "global_step": 111870, "epoch": 1256, "lr": 6.379276532891937e-05}
{"train_loss": 0.10900088399648666, "global_step": 111871, "epoch": 1256, "lr": 6.379220808564864e-05}
{"train_loss": 0.14943841112296233, "global_step": 111872, "epoch": 1256, "lr": 6.379165084052371e-05, "val_loss": 4.956985950469971}
{"train_loss": 0.1482568383216858, "global_step": 111873, "epoch": 1257, "lr": 6.37910935935447e-05}
{"train_loss": 0.11446931958198547, "global_step": 111874, "epoch": 1257, "lr": 6.37905363447116e-05}
{"train_loss": 0.20145638287067413, "global_step": 111875, "epoch": 1257, "lr": 6.378997909402456e-05}
{"train_loss": 0.15292644500732422, "global_step": 111876, "epoch": 1257, "lr": 6.378942184148361e-05}
{"train_loss": 0.1532689481973648, "global_step": 111877, "epoch": 1257, "lr": 6.378886458708885e-05}
{"train_loss": 0.12298175692558289, "global_step": 111878, "epoch": 1257, "lr": 6.378830733084034e-05}
{"train_loss": 0.14138378202915192, "global_step": 111879, "epoch": 1257, "lr": 6.378775007273816e-05}
{"train_loss": 0.11087395995855331, "global_step": 111880, "epoch": 1257, "lr": 6.378719281278239e-05}
{"train_loss": 0.10141651332378387, "global_step": 111881, "epoch": 1257, "lr": 6.378663555097308e-05}
{"train_loss": 0.19807633757591248, "global_step": 111882, "epoch": 1257, "lr": 6.378607828731035e-05}
{"train_loss": 0.15937772393226624, "global_step": 111883, "epoch": 1257, "lr": 6.378552102179423e-05}
{"train_loss": 0.0590500570833683, "global_step": 111884, "epoch": 1257, "lr": 6.378496375442484e-05}
{"train_loss": 0.10191350430250168, "global_step": 111885, "epoch": 1257, "lr": 6.378440648520221e-05}
{"train_loss": 0.0911860242486, "global_step": 111886, "epoch": 1257, "lr": 6.378384921412644e-05}
{"train_loss": 0.135691836476326, "global_step": 111887, "epoch": 1257, "lr": 6.37832919411976e-05}
{"train_loss": 0.1546238362789154, "global_step": 111888, "epoch": 1257, "lr": 6.378273466641575e-05}
{"train_loss": 0.17962568998336792, "global_step": 111889, "epoch": 1257, "lr": 6.3782177389781e-05}
{"train_loss": 0.13095973432064056, "global_step": 111890, "epoch": 1257, "lr": 6.37816201112934e-05}
{"train_loss": 0.15744072198867798, "global_step": 111891, "epoch": 1257, "lr": 6.378106283095302e-05}
{"train_loss": 0.21299025416374207, "global_step": 111892, "epoch": 1257, "lr": 6.378050554875996e-05}
{"train_loss": 0.11074638366699219, "global_step": 111893, "epoch": 1257, "lr": 6.377994826471426e-05}
{"train_loss": 0.11696561425924301, "global_step": 111894, "epoch": 1257, "lr": 6.377939097881603e-05}
{"train_loss": 0.22879861295223236, "global_step": 111895, "epoch": 1257, "lr": 6.377883369106532e-05}
{"train_loss": 0.1553134173154831, "global_step": 111896, "epoch": 1257, "lr": 6.377827640146221e-05}
{"train_loss": 0.1851079761981964, "global_step": 111897, "epoch": 1257, "lr": 6.377771911000678e-05}
{"train_loss": 0.12928031384944916, "global_step": 111898, "epoch": 1257, "lr": 6.377716181669911e-05}
{"train_loss": 0.1616508513689041, "global_step": 111899, "epoch": 1257, "lr": 6.377660452153928e-05}
{"train_loss": 0.12927313148975372, "global_step": 111900, "epoch": 1257, "lr": 6.377604722452733e-05}
{"train_loss": 0.1515362709760666, "global_step": 111901, "epoch": 1257, "lr": 6.377548992566337e-05}
{"train_loss": 0.20823577046394348, "global_step": 111902, "epoch": 1257, "lr": 6.377493262494746e-05}
{"train_loss": 0.2611636519432068, "global_step": 111903, "epoch": 1257, "lr": 6.377437532237969e-05}
{"train_loss": 0.25785720348358154, "global_step": 111904, "epoch": 1257, "lr": 6.37738180179601e-05}
{"train_loss": 0.1011238694190979, "global_step": 111905, "epoch": 1257, "lr": 6.377326071168881e-05}
{"train_loss": 0.19453604519367218, "global_step": 111906, "epoch": 1257, "lr": 6.377270340356586e-05}
{"train_loss": 0.09131546318531036, "global_step": 111907, "epoch": 1257, "lr": 6.377214609359135e-05}
{"train_loss": 0.11299265921115875, "global_step": 111908, "epoch": 1257, "lr": 6.377158878176531e-05}
{"train_loss": 0.12623438239097595, "global_step": 111909, "epoch": 1257, "lr": 6.377103146808788e-05}
{"train_loss": 0.15627391636371613, "global_step": 111910, "epoch": 1257, "lr": 6.37704741525591e-05}
{"train_loss": 0.13983401656150818, "global_step": 111911, "epoch": 1257, "lr": 6.376991683517904e-05}
{"train_loss": 0.11802331358194351, "global_step": 111912, "epoch": 1257, "lr": 6.376935951594779e-05}
{"train_loss": 0.17039376497268677, "global_step": 111913, "epoch": 1257, "lr": 6.37688021948654e-05}
{"train_loss": 0.14261052012443542, "global_step": 111914, "epoch": 1257, "lr": 6.376824487193198e-05}
{"train_loss": 0.13328184187412262, "global_step": 111915, "epoch": 1257, "lr": 6.37676875471476e-05}
{"train_loss": 0.14076995849609375, "global_step": 111916, "epoch": 1257, "lr": 6.376713022051229e-05}
{"train_loss": 0.16788680851459503, "global_step": 111917, "epoch": 1257, "lr": 6.376657289202616e-05}
{"train_loss": 0.1325651854276657, "global_step": 111918, "epoch": 1257, "lr": 6.37660155616893e-05}
{"train_loss": 0.11920742690563202, "global_step": 111919, "epoch": 1257, "lr": 6.376545822950176e-05}
{"train_loss": 0.08723890036344528, "global_step": 111920, "epoch": 1257, "lr": 6.376490089546362e-05}
{"train_loss": 0.13595807552337646, "global_step": 111921, "epoch": 1257, "lr": 6.376434355957496e-05}
{"train_loss": 0.12445327639579773, "global_step": 111922, "epoch": 1257, "lr": 6.376378622183585e-05}
{"train_loss": 0.1716955602169037, "global_step": 111923, "epoch": 1257, "lr": 6.376322888224637e-05}
{"train_loss": 0.15952439606189728, "global_step": 111924, "epoch": 1257, "lr": 6.37626715408066e-05}
{"train_loss": 0.16764473915100098, "global_step": 111925, "epoch": 1257, "lr": 6.376211419751659e-05}
{"train_loss": 0.16810105741024017, "global_step": 111926, "epoch": 1257, "lr": 6.376155685237643e-05}
{"train_loss": 0.08955003321170807, "global_step": 111927, "epoch": 1257, "lr": 6.37609995053862e-05}
{"train_loss": 0.07643570750951767, "global_step": 111928, "epoch": 1257, "lr": 6.376044215654598e-05}
{"train_loss": 0.12774094939231873, "global_step": 111929, "epoch": 1257, "lr": 6.375988480585585e-05}
{"train_loss": 0.11903556436300278, "global_step": 111930, "epoch": 1257, "lr": 6.375932745331584e-05}
{"train_loss": 0.15732041001319885, "global_step": 111931, "epoch": 1257, "lr": 6.375877009892607e-05}
{"train_loss": 0.14078949391841888, "global_step": 111932, "epoch": 1257, "lr": 6.37582127426866e-05}
{"train_loss": 0.14294745028018951, "global_step": 111933, "epoch": 1257, "lr": 6.375765538459751e-05}
{"train_loss": 0.11596881598234177, "global_step": 111934, "epoch": 1257, "lr": 6.375709802465886e-05}
{"train_loss": 0.11515714973211288, "global_step": 111935, "epoch": 1257, "lr": 6.375654066287075e-05}
{"train_loss": 0.15692996978759766, "global_step": 111936, "epoch": 1257, "lr": 6.375598329923322e-05}
{"train_loss": 0.14794884622097015, "global_step": 111937, "epoch": 1257, "lr": 6.375542593374638e-05}
{"train_loss": 0.09576538950204849, "global_step": 111938, "epoch": 1257, "lr": 6.375486856641028e-05}
{"train_loss": 0.1666509211063385, "global_step": 111939, "epoch": 1257, "lr": 6.375431119722503e-05}
{"train_loss": 0.21890002489089966, "global_step": 111940, "epoch": 1257, "lr": 6.375375382619066e-05}
{"train_loss": 0.11770381033420563, "global_step": 111941, "epoch": 1257, "lr": 6.375319645330727e-05}
{"train_loss": 0.16281630098819733, "global_step": 111942, "epoch": 1257, "lr": 6.375263907857494e-05}
{"train_loss": 0.11017652601003647, "global_step": 111943, "epoch": 1257, "lr": 6.375208170199372e-05}
{"train_loss": 0.11153355985879898, "global_step": 111944, "epoch": 1257, "lr": 6.37515243235637e-05}
{"train_loss": 0.125926211476326, "global_step": 111945, "epoch": 1257, "lr": 6.375096694328497e-05}
{"train_loss": 0.14683254063129425, "global_step": 111946, "epoch": 1257, "lr": 6.375040956115758e-05}
{"train_loss": 0.09655796736478806, "global_step": 111947, "epoch": 1257, "lr": 6.37498521771816e-05}
{"train_loss": 0.12175369262695312, "global_step": 111948, "epoch": 1257, "lr": 6.374929479135717e-05}
{"train_loss": 0.12529301643371582, "global_step": 111949, "epoch": 1257, "lr": 6.374873740368427e-05}
{"train_loss": 0.21733805537223816, "global_step": 111950, "epoch": 1257, "lr": 6.374818001416303e-05}
{"train_loss": 0.13101401925086975, "global_step": 111951, "epoch": 1257, "lr": 6.374762262279352e-05}
{"train_loss": 0.14929577708244324, "global_step": 111952, "epoch": 1257, "lr": 6.374706522957581e-05}
{"train_loss": 0.11496379226446152, "global_step": 111953, "epoch": 1257, "lr": 6.374650783450995e-05}
{"train_loss": 0.23267914354801178, "global_step": 111954, "epoch": 1257, "lr": 6.374595043759608e-05}
{"train_loss": 0.10820018500089645, "global_step": 111955, "epoch": 1257, "lr": 6.37453930388342e-05}
{"train_loss": 0.16804452240467072, "global_step": 111956, "epoch": 1257, "lr": 6.374483563822445e-05}
{"train_loss": 0.15105494856834412, "global_step": 111957, "epoch": 1257, "lr": 6.374427823576686e-05}
{"train_loss": 0.116226427257061, "global_step": 111958, "epoch": 1257, "lr": 6.374372083146151e-05}
{"train_loss": 0.23989813029766083, "global_step": 111959, "epoch": 1257, "lr": 6.37431634253085e-05}
{"train_loss": 0.17005400359630585, "global_step": 111960, "epoch": 1257, "lr": 6.374260601730788e-05}
{"train_loss": 0.14525314407904497, "global_step": 111961, "epoch": 1257, "lr": 6.374204860745973e-05, "val_loss": 5.342825412750244}
{"train_loss": 0.15502485632896423, "global_step": 111962, "epoch": 1258, "lr": 6.374149119576413e-05}
{"train_loss": 0.15678545832633972, "global_step": 111963, "epoch": 1258, "lr": 6.374093378222115e-05}
{"train_loss": 0.2318883240222931, "global_step": 111964, "epoch": 1258, "lr": 6.374037636683087e-05}
{"train_loss": 0.1964719444513321, "global_step": 111965, "epoch": 1258, "lr": 6.373981894959338e-05}
{"train_loss": 0.08981753140687943, "global_step": 111966, "epoch": 1258, "lr": 6.373926153050874e-05}
{"train_loss": 0.11735504120588303, "global_step": 111967, "epoch": 1258, "lr": 6.373870410957702e-05}
{"train_loss": 0.244611456990242, "global_step": 111968, "epoch": 1258, "lr": 6.373814668679829e-05}
{"train_loss": 0.13322827219963074, "global_step": 111969, "epoch": 1258, "lr": 6.373758926217264e-05}
{"train_loss": 0.22253434360027313, "global_step": 111970, "epoch": 1258, "lr": 6.373703183570014e-05}
{"train_loss": 0.1136547178030014, "global_step": 111971, "epoch": 1258, "lr": 6.373647440738085e-05}
{"train_loss": 0.22272264957427979, "global_step": 111972, "epoch": 1258, "lr": 6.373591697721488e-05}
{"train_loss": 0.14117518067359924, "global_step": 111973, "epoch": 1258, "lr": 6.373535954520226e-05}
{"train_loss": 0.11652960628271103, "global_step": 111974, "epoch": 1258, "lr": 6.37348021113431e-05}
{"train_loss": 0.12259261310100555, "global_step": 111975, "epoch": 1258, "lr": 6.373424467563745e-05}
{"train_loss": 0.23072561621665955, "global_step": 111976, "epoch": 1258, "lr": 6.373368723808542e-05}
{"train_loss": 0.12303634732961655, "global_step": 111977, "epoch": 1258, "lr": 6.373312979868705e-05}
{"train_loss": 0.26142406463623047, "global_step": 111978, "epoch": 1258, "lr": 6.373257235744244e-05}
{"train_loss": 0.152694433927536, "global_step": 111979, "epoch": 1258, "lr": 6.373201491435163e-05}
{"train_loss": 0.09604814648628235, "global_step": 111980, "epoch": 1258, "lr": 6.373145746941473e-05}
{"train_loss": 0.22619637846946716, "global_step": 111981, "epoch": 1258, "lr": 6.37309000226318e-05}
{"train_loss": 0.14179441332817078, "global_step": 111982, "epoch": 1258, "lr": 6.373034257400293e-05}
{"train_loss": 0.13828667998313904, "global_step": 111983, "epoch": 1258, "lr": 6.372978512352817e-05}
{"train_loss": 0.19708846509456635, "global_step": 111984, "epoch": 1258, "lr": 6.372922767120761e-05}
{"train_loss": 0.16434180736541748, "global_step": 111985, "epoch": 1258, "lr": 6.372867021704133e-05}
{"train_loss": 0.18388856947422028, "global_step": 111986, "epoch": 1258, "lr": 6.37281127610294e-05}
{"train_loss": 0.14175638556480408, "global_step": 111987, "epoch": 1258, "lr": 6.372755530317189e-05}
{"train_loss": 0.14209504425525665, "global_step": 111988, "epoch": 1258, "lr": 6.372699784346887e-05}
{"train_loss": 0.14331360161304474, "global_step": 111989, "epoch": 1258, "lr": 6.372644038192044e-05}
{"train_loss": 0.12139030545949936, "global_step": 111990, "epoch": 1258, "lr": 6.372588291852664e-05}
{"train_loss": 0.15886351466178894, "global_step": 111991, "epoch": 1258, "lr": 6.372532545328756e-05}
{"train_loss": 0.1469973772764206, "global_step": 111992, "epoch": 1258, "lr": 6.372476798620328e-05}
{"train_loss": 0.15144659578800201, "global_step": 111993, "epoch": 1258, "lr": 6.372421051727389e-05}
{"train_loss": 0.2436995506286621, "global_step": 111994, "epoch": 1258, "lr": 6.372365304649942e-05}
{"train_loss": 0.14376294612884521, "global_step": 111995, "epoch": 1258, "lr": 6.372309557387999e-05}
{"train_loss": 0.16788837313652039, "global_step": 111996, "epoch": 1258, "lr": 6.372253809941565e-05}
{"train_loss": 0.24118371307849884, "global_step": 111997, "epoch": 1258, "lr": 6.372198062310648e-05}
{"train_loss": 0.13250434398651123, "global_step": 111998, "epoch": 1258, "lr": 6.372142314495256e-05}
{"train_loss": 0.14529788494110107, "global_step": 111999, "epoch": 1258, "lr": 6.372086566495395e-05}
{"train_loss": 0.21141240000724792, "global_step": 112000, "epoch": 1258, "lr": 6.372030818311075e-05}
{"train_loss": 0.1353369653224945, "global_step": 112001, "epoch": 1258, "lr": 6.371975069942303e-05}
{"train_loss": 0.16474759578704834, "global_step": 112002, "epoch": 1258, "lr": 6.371919321389085e-05}
{"train_loss": 0.20257316529750824, "global_step": 112003, "epoch": 1258, "lr": 6.371863572651428e-05}
{"train_loss": 0.08733144402503967, "global_step": 112004, "epoch": 1258, "lr": 6.371807823729342e-05}
{"train_loss": 0.15121756494045258, "global_step": 112005, "epoch": 1258, "lr": 6.371752074622834e-05}
{"train_loss": 0.15960381925106049, "global_step": 112006, "epoch": 1258, "lr": 6.371696325331909e-05}
{"train_loss": 0.12777695059776306, "global_step": 112007, "epoch": 1258, "lr": 6.371640575856577e-05}
{"train_loss": 0.13833925127983093, "global_step": 112008, "epoch": 1258, "lr": 6.371584826196845e-05}
{"train_loss": 0.16033196449279785, "global_step": 112009, "epoch": 1258, "lr": 6.37152907635272e-05}
{"train_loss": 0.07820987701416016, "global_step": 112010, "epoch": 1258, "lr": 6.371473326324207e-05}
{"train_loss": 0.1606520563364029, "global_step": 112011, "epoch": 1258, "lr": 6.371417576111319e-05}
{"train_loss": 0.0870845839381218, "global_step": 112012, "epoch": 1258, "lr": 6.37136182571406e-05}
{"train_loss": 0.15978169441223145, "global_step": 112013, "epoch": 1258, "lr": 6.371306075132437e-05}
{"train_loss": 0.12900136411190033, "global_step": 112014, "epoch": 1258, "lr": 6.37125032436646e-05}
{"train_loss": 0.11229310929775238, "global_step": 112015, "epoch": 1258, "lr": 6.371194573416135e-05}
{"train_loss": 0.13702496886253357, "global_step": 112016, "epoch": 1258, "lr": 6.371138822281468e-05}
{"train_loss": 0.11161839962005615, "global_step": 112017, "epoch": 1258, "lr": 6.371083070962469e-05}
{"train_loss": 0.14587539434432983, "global_step": 112018, "epoch": 1258, "lr": 6.371027319459145e-05}
{"train_loss": 0.17008879780769348, "global_step": 112019, "epoch": 1258, "lr": 6.370971567771503e-05}
{"train_loss": 0.19212956726551056, "global_step": 112020, "epoch": 1258, "lr": 6.370915815899551e-05}
{"train_loss": 0.10911533981561661, "global_step": 112021, "epoch": 1258, "lr": 6.370860063843296e-05}
{"train_loss": 0.18767058849334717, "global_step": 112022, "epoch": 1258, "lr": 6.370804311602745e-05}
{"train_loss": 0.1505284309387207, "global_step": 112023, "epoch": 1258, "lr": 6.370748559177906e-05}
{"train_loss": 0.18695105612277985, "global_step": 112024, "epoch": 1258, "lr": 6.370692806568786e-05}
{"train_loss": 0.1948232501745224, "global_step": 112025, "epoch": 1258, "lr": 6.370637053775395e-05}
{"train_loss": 0.1271049827337265, "global_step": 112026, "epoch": 1258, "lr": 6.370581300797737e-05}
{"train_loss": 0.16516074538230896, "global_step": 112027, "epoch": 1258, "lr": 6.370525547635822e-05}
{"train_loss": 0.13237573206424713, "global_step": 112028, "epoch": 1258, "lr": 6.370469794289658e-05}
{"train_loss": 0.09969460964202881, "global_step": 112029, "epoch": 1258, "lr": 6.370414040759247e-05}
{"train_loss": 0.15398482978343964, "global_step": 112030, "epoch": 1258, "lr": 6.370358287044604e-05}
{"train_loss": 0.17360392212867737, "global_step": 112031, "epoch": 1258, "lr": 6.37030253314573e-05}
{"train_loss": 0.10045105218887329, "global_step": 112032, "epoch": 1258, "lr": 6.370246779062638e-05}
{"train_loss": 0.17800599336624146, "global_step": 112033, "epoch": 1258, "lr": 6.370191024795332e-05}
{"train_loss": 0.16076231002807617, "global_step": 112034, "epoch": 1258, "lr": 6.370135270343821e-05}
{"train_loss": 0.1059303879737854, "global_step": 112035, "epoch": 1258, "lr": 6.37007951570811e-05}
{"train_loss": 0.13992765545845032, "global_step": 112036, "epoch": 1258, "lr": 6.37002376088821e-05}
{"train_loss": 0.15716421604156494, "global_step": 112037, "epoch": 1258, "lr": 6.369968005884129e-05}
{"train_loss": 0.17509494721889496, "global_step": 112038, "epoch": 1258, "lr": 6.369912250695871e-05}
{"train_loss": 0.1467377096414566, "global_step": 112039, "epoch": 1258, "lr": 6.369856495323443e-05}
{"train_loss": 0.1705906242132187, "global_step": 112040, "epoch": 1258, "lr": 6.369800739766856e-05}
{"train_loss": 0.14504043757915497, "global_step": 112041, "epoch": 1258, "lr": 6.369744984026117e-05}
{"train_loss": 0.0949566513299942, "global_step": 112042, "epoch": 1258, "lr": 6.369689228101232e-05}
{"train_loss": 0.20683135092258453, "global_step": 112043, "epoch": 1258, "lr": 6.369633471992209e-05}
{"train_loss": 0.19476696848869324, "global_step": 112044, "epoch": 1258, "lr": 6.369577715699055e-05}
{"train_loss": 0.18677900731563568, "global_step": 112045, "epoch": 1258, "lr": 6.369521959221779e-05}
{"train_loss": 0.18943823873996735, "global_step": 112046, "epoch": 1258, "lr": 6.369466202560387e-05}
{"train_loss": 0.131158709526062, "global_step": 112047, "epoch": 1258, "lr": 6.369410445714887e-05}
{"train_loss": 0.20935489237308502, "global_step": 112048, "epoch": 1258, "lr": 6.369354688685285e-05}
{"train_loss": 0.1952904611825943, "global_step": 112049, "epoch": 1258, "lr": 6.369298931471591e-05}
{"train_loss": 0.15657149539904647, "global_step": 112050, "epoch": 1258, "lr": 6.369243174073812e-05, "val_loss": 4.972029685974121}
{"train_loss": 0.13640570640563965, "global_step": 112051, "epoch": 1259, "lr": 6.369187416491954e-05}
{"train_loss": 0.11271187663078308, "global_step": 112052, "epoch": 1259, "lr": 6.369131658726027e-05}
{"train_loss": 0.1496141254901886, "global_step": 112053, "epoch": 1259, "lr": 6.369075900776034e-05}
{"train_loss": 0.08780015259981155, "global_step": 112054, "epoch": 1259, "lr": 6.369020142641986e-05}
{"train_loss": 0.18433870375156403, "global_step": 112055, "epoch": 1259, "lr": 6.368964384323892e-05}
{"train_loss": 0.1606355905532837, "global_step": 112056, "epoch": 1259, "lr": 6.368908625821756e-05}
{"train_loss": 0.14441661536693573, "global_step": 112057, "epoch": 1259, "lr": 6.368852867135587e-05}
{"train_loss": 0.13308168947696686, "global_step": 112058, "epoch": 1259, "lr": 6.368797108265393e-05}
{"train_loss": 0.17967846989631653, "global_step": 112059, "epoch": 1259, "lr": 6.368741349211181e-05}
{"train_loss": 0.13488978147506714, "global_step": 112060, "epoch": 1259, "lr": 6.368685589972958e-05}
{"train_loss": 0.12572041153907776, "global_step": 112061, "epoch": 1259, "lr": 6.368629830550732e-05}
{"train_loss": 0.16733762621879578, "global_step": 112062, "epoch": 1259, "lr": 6.368574070944511e-05}
{"train_loss": 0.13276541233062744, "global_step": 112063, "epoch": 1259, "lr": 6.368518311154302e-05}
{"train_loss": 0.11766409128904343, "global_step": 112064, "epoch": 1259, "lr": 6.368462551180112e-05}
{"train_loss": 0.14734981954097748, "global_step": 112065, "epoch": 1259, "lr": 6.368406791021948e-05}
{"train_loss": 0.1688401699066162, "global_step": 112066, "epoch": 1259, "lr": 6.36835103067982e-05}
{"train_loss": 0.14342282712459564, "global_step": 112067, "epoch": 1259, "lr": 6.368295270153732e-05}
{"train_loss": 0.08001549541950226, "global_step": 112068, "epoch": 1259, "lr": 6.368239509443695e-05}
{"train_loss": 0.10054436326026917, "global_step": 112069, "epoch": 1259, "lr": 6.368183748549715e-05}
{"train_loss": 0.17029674351215363, "global_step": 112070, "epoch": 1259, "lr": 6.368127987471797e-05}
{"train_loss": 0.2303917407989502, "global_step": 112071, "epoch": 1259, "lr": 6.368072226209951e-05}
{"train_loss": 0.1289290487766266, "global_step": 112072, "epoch": 1259, "lr": 6.368016464764186e-05}
{"train_loss": 0.12537676095962524, "global_step": 112073, "epoch": 1259, "lr": 6.367960703134508e-05}
{"train_loss": 0.2154386043548584, "global_step": 112074, "epoch": 1259, "lr": 6.367904941320924e-05}
{"train_loss": 0.03771762549877167, "global_step": 112075, "epoch": 1259, "lr": 6.367849179323441e-05}
{"train_loss": 0.1747785359621048, "global_step": 112076, "epoch": 1259, "lr": 6.36779341714207e-05}
{"train_loss": 0.14999397099018097, "global_step": 112077, "epoch": 1259, "lr": 6.367737654776813e-05}
{"train_loss": 0.11600268632173538, "global_step": 112078, "epoch": 1259, "lr": 6.367681892227681e-05}
{"train_loss": 0.1173805296421051, "global_step": 112079, "epoch": 1259, "lr": 6.367626129494681e-05}
{"train_loss": 0.17294089496135712, "global_step": 112080, "epoch": 1259, "lr": 6.367570366577822e-05}
{"train_loss": 0.20953011512756348, "global_step": 112081, "epoch": 1259, "lr": 6.367514603477107e-05}
{"train_loss": 0.17973238229751587, "global_step": 112082, "epoch": 1259, "lr": 6.367458840192548e-05}
{"train_loss": 0.1275530457496643, "global_step": 112083, "epoch": 1259, "lr": 6.367403076724152e-05}
{"train_loss": 0.13186313211917877, "global_step": 112084, "epoch": 1259, "lr": 6.367347313071924e-05}
{"train_loss": 0.15427356958389282, "global_step": 112085, "epoch": 1259, "lr": 6.367291549235872e-05}
{"train_loss": 0.09602826833724976, "global_step": 112086, "epoch": 1259, "lr": 6.367235785216006e-05}
{"train_loss": 0.1196114718914032, "global_step": 112087, "epoch": 1259, "lr": 6.367180021012331e-05}
{"train_loss": 0.1816251128911972, "global_step": 112088, "epoch": 1259, "lr": 6.367124256624855e-05}
{"train_loss": 0.16032764315605164, "global_step": 112089, "epoch": 1259, "lr": 6.367068492053585e-05}
{"train_loss": 0.22095134854316711, "global_step": 112090, "epoch": 1259, "lr": 6.367012727298529e-05}
{"train_loss": 0.11817745119333267, "global_step": 112091, "epoch": 1259, "lr": 6.366956962359696e-05}
{"train_loss": 0.1433703899383545, "global_step": 112092, "epoch": 1259, "lr": 6.366901197237093e-05}
{"train_loss": 0.16413921117782593, "global_step": 112093, "epoch": 1259, "lr": 6.366845431930726e-05}
{"train_loss": 0.14707405865192413, "global_step": 112094, "epoch": 1259, "lr": 6.366789666440604e-05}
{"train_loss": 0.14022697508335114, "global_step": 112095, "epoch": 1259, "lr": 6.366733900766733e-05}
{"train_loss": 0.1350545585155487, "global_step": 112096, "epoch": 1259, "lr": 6.36667813490912e-05}
{"train_loss": 0.1063348576426506, "global_step": 112097, "epoch": 1259, "lr": 6.366622368867777e-05}
{"train_loss": 0.137528195977211, "global_step": 112098, "epoch": 1259, "lr": 6.366566602642707e-05}
{"train_loss": 0.1343812346458435, "global_step": 112099, "epoch": 1259, "lr": 6.366510836233917e-05}
{"train_loss": 0.12381842732429504, "global_step": 112100, "epoch": 1259, "lr": 6.366455069641419e-05}
{"train_loss": 0.10142387449741364, "global_step": 112101, "epoch": 1259, "lr": 6.366399302865216e-05}
{"train_loss": 0.10906126350164413, "global_step": 112102, "epoch": 1259, "lr": 6.366343535905319e-05}
{"train_loss": 0.14997053146362305, "global_step": 112103, "epoch": 1259, "lr": 6.366287768761732e-05}
{"train_loss": 0.13437679409980774, "global_step": 112104, "epoch": 1259, "lr": 6.366232001434464e-05}
{"train_loss": 0.14191897213459015, "global_step": 112105, "epoch": 1259, "lr": 6.366176233923525e-05}
{"train_loss": 0.15275391936302185, "global_step": 112106, "epoch": 1259, "lr": 6.366120466228918e-05}
{"train_loss": 0.1007452979683876, "global_step": 112107, "epoch": 1259, "lr": 6.366064698350651e-05}
{"train_loss": 0.17475388944149017, "global_step": 112108, "epoch": 1259, "lr": 6.366008930288736e-05}
{"train_loss": 0.1674986183643341, "global_step": 112109, "epoch": 1259, "lr": 6.365953162043178e-05}
{"train_loss": 0.15895144641399384, "global_step": 112110, "epoch": 1259, "lr": 6.365897393613983e-05}
{"train_loss": 0.1536206156015396, "global_step": 112111, "epoch": 1259, "lr": 6.365841625001161e-05}
{"train_loss": 0.12076390534639359, "global_step": 112112, "epoch": 1259, "lr": 6.365785856204719e-05}
{"train_loss": 0.09064427763223648, "global_step": 112113, "epoch": 1259, "lr": 6.365730087224661e-05}
{"train_loss": 0.18905971944332123, "global_step": 112114, "epoch": 1259, "lr": 6.365674318060999e-05}
{"train_loss": 0.09715424478054047, "global_step": 112115, "epoch": 1259, "lr": 6.365618548713737e-05}
{"train_loss": 0.1775902509689331, "global_step": 112116, "epoch": 1259, "lr": 6.365562779182887e-05}
{"train_loss": 0.1447855830192566, "global_step": 112117, "epoch": 1259, "lr": 6.365507009468452e-05}
{"train_loss": 0.1585133969783783, "global_step": 112118, "epoch": 1259, "lr": 6.365451239570441e-05}
{"train_loss": 0.10733935236930847, "global_step": 112119, "epoch": 1259, "lr": 6.365395469488862e-05}
{"train_loss": 0.16591840982437134, "global_step": 112120, "epoch": 1259, "lr": 6.365339699223723e-05}
{"train_loss": 0.14231853187084198, "global_step": 112121, "epoch": 1259, "lr": 6.36528392877503e-05}
{"train_loss": 0.1422279179096222, "global_step": 112122, "epoch": 1259, "lr": 6.365228158142793e-05}
{"train_loss": 0.16451047360897064, "global_step": 112123, "epoch": 1259, "lr": 6.365172387327017e-05}
{"train_loss": 0.14784184098243713, "global_step": 112124, "epoch": 1259, "lr": 6.365116616327708e-05}
{"train_loss": 0.1514360010623932, "global_step": 112125, "epoch": 1259, "lr": 6.365060845144877e-05}
{"train_loss": 0.12523537874221802, "global_step": 112126, "epoch": 1259, "lr": 6.36500507377853e-05}
{"train_loss": 0.10479176789522171, "global_step": 112127, "epoch": 1259, "lr": 6.364949302228675e-05}
{"train_loss": 0.15121573209762573, "global_step": 112128, "epoch": 1259, "lr": 6.36489353049532e-05}
{"train_loss": 0.1318839192390442, "global_step": 112129, "epoch": 1259, "lr": 6.36483775857847e-05}
{"train_loss": 0.09073273092508316, "global_step": 112130, "epoch": 1259, "lr": 6.364781986478134e-05}
{"train_loss": 0.15065358579158783, "global_step": 112131, "epoch": 1259, "lr": 6.364726214194321e-05}
{"train_loss": 0.1971229463815689, "global_step": 112132, "epoch": 1259, "lr": 6.364670441727037e-05}
{"train_loss": 0.11085924506187439, "global_step": 112133, "epoch": 1259, "lr": 6.36461466907629e-05}
{"train_loss": 0.17101116478443146, "global_step": 112134, "epoch": 1259, "lr": 6.364558896242087e-05}
{"train_loss": 0.14537763595581055, "global_step": 112135, "epoch": 1259, "lr": 6.364503123224436e-05}
{"train_loss": 0.13619519770145416, "global_step": 112136, "epoch": 1259, "lr": 6.364447350023344e-05}
{"train_loss": 0.12867121398448944, "global_step": 112137, "epoch": 1259, "lr": 6.364391576638818e-05}
{"train_loss": 0.17060641944408417, "global_step": 112138, "epoch": 1259, "lr": 6.364335803070866e-05}
{"train_loss": 0.1421493109357491, "global_step": 112139, "epoch": 1259, "lr": 6.364280029319497e-05, "val_loss": 5.020614147186279}
{"train_loss": 0.19844627380371094, "global_step": 112140, "epoch": 1260, "lr": 6.364224255384716e-05}
{"train_loss": 0.16995377838611603, "global_step": 112141, "epoch": 1260, "lr": 6.364168481266534e-05}
{"train_loss": 0.17953377962112427, "global_step": 112142, "epoch": 1260, "lr": 6.364112706964953e-05}
{"train_loss": 0.11257356405258179, "global_step": 112143, "epoch": 1260, "lr": 6.364056932479985e-05}
{"train_loss": 0.10683877766132355, "global_step": 112144, "epoch": 1260, "lr": 6.364001157811635e-05}
{"train_loss": 0.15085510909557343, "global_step": 112145, "epoch": 1260, "lr": 6.363945382959914e-05}
{"train_loss": 0.11039917171001434, "global_step": 112146, "epoch": 1260, "lr": 6.363889607924825e-05}
{"train_loss": 0.16188620030879974, "global_step": 112147, "epoch": 1260, "lr": 6.363833832706378e-05}
{"train_loss": 0.1960713416337967, "global_step": 112148, "epoch": 1260, "lr": 6.363778057304582e-05}
{"train_loss": 0.12373841553926468, "global_step": 112149, "epoch": 1260, "lr": 6.363722281719442e-05}
{"train_loss": 0.13262836635112762, "global_step": 112150, "epoch": 1260, "lr": 6.363666505950965e-05}
{"train_loss": 0.10790057480335236, "global_step": 112151, "epoch": 1260, "lr": 6.36361072999916e-05}
{"train_loss": 0.1518109142780304, "global_step": 112152, "epoch": 1260, "lr": 6.363554953864035e-05}
{"train_loss": 0.15627999603748322, "global_step": 112153, "epoch": 1260, "lr": 6.363499177545596e-05}
{"train_loss": 0.15734080970287323, "global_step": 112154, "epoch": 1260, "lr": 6.363443401043852e-05}
{"train_loss": 0.17552974820137024, "global_step": 112155, "epoch": 1260, "lr": 6.363387624358807e-05}
{"train_loss": 0.1872774362564087, "global_step": 112156, "epoch": 1260, "lr": 6.363331847490475e-05}
{"train_loss": 0.21088100969791412, "global_step": 112157, "epoch": 1260, "lr": 6.363276070438858e-05}
{"train_loss": 0.17249546945095062, "global_step": 112158, "epoch": 1260, "lr": 6.363220293203964e-05}
{"train_loss": 0.23660406470298767, "global_step": 112159, "epoch": 1260, "lr": 6.363164515785802e-05}
{"train_loss": 0.09566731750965118, "global_step": 112160, "epoch": 1260, "lr": 6.363108738184381e-05}
{"train_loss": 0.11879665404558182, "global_step": 112161, "epoch": 1260, "lr": 6.363052960399705e-05}
{"train_loss": 0.22100476920604706, "global_step": 112162, "epoch": 1260, "lr": 6.362997182431782e-05}
{"train_loss": 0.16298843920230865, "global_step": 112163, "epoch": 1260, "lr": 6.362941404280622e-05}
{"train_loss": 0.19170483946800232, "global_step": 112164, "epoch": 1260, "lr": 6.362885625946232e-05}
{"train_loss": 0.2165345996618271, "global_step": 112165, "epoch": 1260, "lr": 6.362829847428617e-05}
{"train_loss": 0.19218067824840546, "global_step": 112166, "epoch": 1260, "lr": 6.362774068727787e-05}
{"train_loss": 0.1321885734796524, "global_step": 112167, "epoch": 1260, "lr": 6.36271828984375e-05}
{"train_loss": 0.16651323437690735, "global_step": 112168, "epoch": 1260, "lr": 6.36266251077651e-05}
{"train_loss": 0.15589576959609985, "global_step": 112169, "epoch": 1260, "lr": 6.362606731526078e-05}
{"train_loss": 0.120688796043396, "global_step": 112170, "epoch": 1260, "lr": 6.362550952092459e-05}
{"train_loss": 0.1773880124092102, "global_step": 112171, "epoch": 1260, "lr": 6.362495172475662e-05}
{"train_loss": 0.20887403190135956, "global_step": 112172, "epoch": 1260, "lr": 6.362439392675694e-05}
{"train_loss": 0.12635868787765503, "global_step": 112173, "epoch": 1260, "lr": 6.362383612692564e-05}
{"train_loss": 0.10801582038402557, "global_step": 112174, "epoch": 1260, "lr": 6.362327832526276e-05}
{"train_loss": 0.11037875711917877, "global_step": 112175, "epoch": 1260, "lr": 6.362272052176841e-05}
{"train_loss": 0.11149238049983978, "global_step": 112176, "epoch": 1260, "lr": 6.362216271644265e-05}
{"train_loss": 0.14093583822250366, "global_step": 112177, "epoch": 1260, "lr": 6.362160490928556e-05}
{"train_loss": 0.13870221376419067, "global_step": 112178, "epoch": 1260, "lr": 6.36210471002972e-05}
{"train_loss": 0.23088818788528442, "global_step": 112179, "epoch": 1260, "lr": 6.362048928947766e-05}
{"train_loss": 0.21289950609207153, "global_step": 112180, "epoch": 1260, "lr": 6.361993147682701e-05}
{"train_loss": 0.15032362937927246, "global_step": 112181, "epoch": 1260, "lr": 6.361937366234534e-05}
{"train_loss": 0.14273180067539215, "global_step": 112182, "epoch": 1260, "lr": 6.36188158460327e-05}
{"train_loss": 0.1377626210451126, "global_step": 112183, "epoch": 1260, "lr": 6.361825802788916e-05}
{"train_loss": 0.1371532380580902, "global_step": 112184, "epoch": 1260, "lr": 6.361770020791484e-05}
{"train_loss": 0.12518243491649628, "global_step": 112185, "epoch": 1260, "lr": 6.361714238610978e-05}
{"train_loss": 0.12597818672657013, "global_step": 112186, "epoch": 1260, "lr": 6.361658456247405e-05}
{"train_loss": 0.09064189344644547, "global_step": 112187, "epoch": 1260, "lr": 6.361602673700774e-05}
{"train_loss": 0.17349594831466675, "global_step": 112188, "epoch": 1260, "lr": 6.361546890971093e-05}
{"train_loss": 0.16951262950897217, "global_step": 112189, "epoch": 1260, "lr": 6.361491108058368e-05}
{"train_loss": 0.12871624529361725, "global_step": 112190, "epoch": 1260, "lr": 6.361435324962606e-05}
{"train_loss": 0.14682792127132416, "global_step": 112191, "epoch": 1260, "lr": 6.361379541683816e-05}
{"train_loss": 0.12445862591266632, "global_step": 112192, "epoch": 1260, "lr": 6.361323758222007e-05}
{"train_loss": 0.12749852240085602, "global_step": 112193, "epoch": 1260, "lr": 6.361267974577184e-05}
{"train_loss": 0.2231384813785553, "global_step": 112194, "epoch": 1260, "lr": 6.361212190749353e-05}
{"train_loss": 0.14174117147922516, "global_step": 112195, "epoch": 1260, "lr": 6.361156406738526e-05}
{"train_loss": 0.0885055661201477, "global_step": 112196, "epoch": 1260, "lr": 6.361100622544706e-05}
{"train_loss": 0.11604249477386475, "global_step": 112197, "epoch": 1260, "lr": 6.361044838167904e-05}
{"train_loss": 0.28889352083206177, "global_step": 112198, "epoch": 1260, "lr": 6.360989053608126e-05}
{"train_loss": 0.13769520819187164, "global_step": 112199, "epoch": 1260, "lr": 6.36093326886538e-05}
{"train_loss": 0.11901958286762238, "global_step": 112200, "epoch": 1260, "lr": 6.360877483939672e-05}
{"train_loss": 0.10857854783535004, "global_step": 112201, "epoch": 1260, "lr": 6.360821698831012e-05}
{"train_loss": 0.14334453642368317, "global_step": 112202, "epoch": 1260, "lr": 6.360765913539405e-05}
{"train_loss": 0.11186298727989197, "global_step": 112203, "epoch": 1260, "lr": 6.36071012806486e-05}
{"train_loss": 0.13443167507648468, "global_step": 112204, "epoch": 1260, "lr": 6.360654342407383e-05}
{"train_loss": 0.241167351603508, "global_step": 112205, "epoch": 1260, "lr": 6.360598556566985e-05}
{"train_loss": 0.1669239103794098, "global_step": 112206, "epoch": 1260, "lr": 6.36054277054367e-05}
{"train_loss": 0.1440003514289856, "global_step": 112207, "epoch": 1260, "lr": 6.360486984337446e-05}
{"train_loss": 0.20004673302173615, "global_step": 112208, "epoch": 1260, "lr": 6.36043119794832e-05}
{"train_loss": 0.14330421388149261, "global_step": 112209, "epoch": 1260, "lr": 6.360375411376302e-05}
{"train_loss": 0.19963224232196808, "global_step": 112210, "epoch": 1260, "lr": 6.360319624621399e-05}
{"train_loss": 0.14001666009426117, "global_step": 112211, "epoch": 1260, "lr": 6.360263837683618e-05}
{"train_loss": 0.12777012586593628, "global_step": 112212, "epoch": 1260, "lr": 6.360208050562963e-05}
{"train_loss": 0.1758609414100647, "global_step": 112213, "epoch": 1260, "lr": 6.360152263259449e-05}
{"train_loss": 0.12287987768650055, "global_step": 112214, "epoch": 1260, "lr": 6.360096475773075e-05}
{"train_loss": 0.14962881803512573, "global_step": 112215, "epoch": 1260, "lr": 6.360040688103855e-05}
{"train_loss": 0.13176928460597992, "global_step": 112216, "epoch": 1260, "lr": 6.359984900251793e-05}
{"train_loss": 0.18555164337158203, "global_step": 112217, "epoch": 1260, "lr": 6.359929112216899e-05}
{"train_loss": 0.14303189516067505, "global_step": 112218, "epoch": 1260, "lr": 6.359873323999176e-05}
{"train_loss": 0.13258901238441467, "global_step": 112219, "epoch": 1260, "lr": 6.359817535598637e-05}
{"train_loss": 0.2021639049053192, "global_step": 112220, "epoch": 1260, "lr": 6.359761747015287e-05}
{"train_loss": 0.13357456028461456, "global_step": 112221, "epoch": 1260, "lr": 6.359705958249133e-05}
{"train_loss": 0.147123321890831, "global_step": 112222, "epoch": 1260, "lr": 6.359650169300184e-05}
{"train_loss": 0.1933497190475464, "global_step": 112223, "epoch": 1260, "lr": 6.359594380168446e-05}
{"train_loss": 0.17534470558166504, "global_step": 112224, "epoch": 1260, "lr": 6.359538590853927e-05}
{"train_loss": 0.18745173513889313, "global_step": 112225, "epoch": 1260, "lr": 6.359482801356635e-05}
{"train_loss": 0.09187348186969757, "global_step": 112226, "epoch": 1260, "lr": 6.359427011676578e-05}
{"train_loss": 0.1340712308883667, "global_step": 112227, "epoch": 1260, "lr": 6.35937122181376e-05}
{"train_loss": 0.15598875727881206, "global_step": 112228, "epoch": 1260, "lr": 6.359315431768194e-05, "val_loss": 5.055338382720947, "train_action_mse_error": 21.466428756713867}
{"train_loss": 0.23727449774742126, "global_step": 112229, "epoch": 1261, "lr": 6.359259641539882e-05}
{"train_loss": 0.12947188317775726, "global_step": 112230, "epoch": 1261, "lr": 6.359203851128835e-05}
{"train_loss": 0.11563456803560257, "global_step": 112231, "epoch": 1261, "lr": 6.359148060535061e-05}
{"train_loss": 0.12833121418952942, "global_step": 112232, "epoch": 1261, "lr": 6.359092269758565e-05}
{"train_loss": 0.13127027451992035, "global_step": 112233, "epoch": 1261, "lr": 6.359036478799355e-05}
{"train_loss": 0.11872892826795578, "global_step": 112234, "epoch": 1261, "lr": 6.35898068765744e-05}
{"train_loss": 0.12232472747564316, "global_step": 112235, "epoch": 1261, "lr": 6.358924896332827e-05}
{"train_loss": 0.12871673703193665, "global_step": 112236, "epoch": 1261, "lr": 6.358869104825523e-05}
{"train_loss": 0.1182796061038971, "global_step": 112237, "epoch": 1261, "lr": 6.358813313135535e-05}
{"train_loss": 0.09807013720273972, "global_step": 112238, "epoch": 1261, "lr": 6.358757521262871e-05}
{"train_loss": 0.28493547439575195, "global_step": 112239, "epoch": 1261, "lr": 6.358701729207538e-05}
{"train_loss": 0.200135737657547, "global_step": 112240, "epoch": 1261, "lr": 6.358645936969545e-05}
{"train_loss": 0.15898661315441132, "global_step": 112241, "epoch": 1261, "lr": 6.358590144548898e-05}
{"train_loss": 0.19135387241840363, "global_step": 112242, "epoch": 1261, "lr": 6.358534351945607e-05}
{"train_loss": 0.25465622544288635, "global_step": 112243, "epoch": 1261, "lr": 6.358478559159675e-05}
{"train_loss": 0.16054536402225494, "global_step": 112244, "epoch": 1261, "lr": 6.358422766191114e-05}
{"train_loss": 0.13034769892692566, "global_step": 112245, "epoch": 1261, "lr": 6.358366973039929e-05}
{"train_loss": 0.18914270401000977, "global_step": 112246, "epoch": 1261, "lr": 6.358311179706128e-05}
{"train_loss": 0.17340759932994843, "global_step": 112247, "epoch": 1261, "lr": 6.358255386189718e-05}
{"train_loss": 0.21613375842571259, "global_step": 112248, "epoch": 1261, "lr": 6.358199592490708e-05}
{"train_loss": 0.1980053186416626, "global_step": 112249, "epoch": 1261, "lr": 6.358143798609104e-05}
{"train_loss": 0.13982948660850525, "global_step": 112250, "epoch": 1261, "lr": 6.358088004544915e-05}
{"train_loss": 0.20319226384162903, "global_step": 112251, "epoch": 1261, "lr": 6.358032210298146e-05}
{"train_loss": 0.20918035507202148, "global_step": 112252, "epoch": 1261, "lr": 6.357976415868809e-05}
{"train_loss": 0.16246193647384644, "global_step": 112253, "epoch": 1261, "lr": 6.357920621256907e-05}
{"train_loss": 0.15033285319805145, "global_step": 112254, "epoch": 1261, "lr": 6.357864826462448e-05}
{"train_loss": 0.10690862685441971, "global_step": 112255, "epoch": 1261, "lr": 6.357809031485442e-05}
{"train_loss": 0.2468566596508026, "global_step": 112256, "epoch": 1261, "lr": 6.357753236325895e-05}
{"train_loss": 0.18071457743644714, "global_step": 112257, "epoch": 1261, "lr": 6.357697440983814e-05}
{"train_loss": 0.15656150877475739, "global_step": 112258, "epoch": 1261, "lr": 6.357641645459209e-05}
{"train_loss": 0.1378653645515442, "global_step": 112259, "epoch": 1261, "lr": 6.357585849752083e-05}
{"train_loss": 0.0868389829993248, "global_step": 112260, "epoch": 1261, "lr": 6.357530053862448e-05}
{"train_loss": 0.13618040084838867, "global_step": 112261, "epoch": 1261, "lr": 6.35747425779031e-05}
{"train_loss": 0.3610449731349945, "global_step": 112262, "epoch": 1261, "lr": 6.357418461535673e-05}
{"train_loss": 0.14297690987586975, "global_step": 112263, "epoch": 1261, "lr": 6.35736266509855e-05}
{"train_loss": 0.1346796154975891, "global_step": 112264, "epoch": 1261, "lr": 6.357306868478947e-05}
{"train_loss": 0.1493026465177536, "global_step": 112265, "epoch": 1261, "lr": 6.357251071676869e-05}
{"train_loss": 0.22210334241390228, "global_step": 112266, "epoch": 1261, "lr": 6.357195274692326e-05}
{"train_loss": 0.15716227889060974, "global_step": 112267, "epoch": 1261, "lr": 6.357139477525325e-05}
{"train_loss": 0.1378449946641922, "global_step": 112268, "epoch": 1261, "lr": 6.357083680175872e-05}
{"train_loss": 0.15728536248207092, "global_step": 112269, "epoch": 1261, "lr": 6.357027882643975e-05}
{"train_loss": 0.14994923770427704, "global_step": 112270, "epoch": 1261, "lr": 6.356972084929642e-05}
{"train_loss": 0.12372159957885742, "global_step": 112271, "epoch": 1261, "lr": 6.356916287032884e-05}
{"train_loss": 0.10423719882965088, "global_step": 112272, "epoch": 1261, "lr": 6.356860488953702e-05}
{"train_loss": 0.12079676985740662, "global_step": 112273, "epoch": 1261, "lr": 6.356804690692108e-05}
{"train_loss": 0.16915132105350494, "global_step": 112274, "epoch": 1261, "lr": 6.356748892248108e-05}
{"train_loss": 0.1010885015130043, "global_step": 112275, "epoch": 1261, "lr": 6.356693093621709e-05}
{"train_loss": 0.16305235028266907, "global_step": 112276, "epoch": 1261, "lr": 6.356637294812919e-05}
{"train_loss": 0.14406870305538177, "global_step": 112277, "epoch": 1261, "lr": 6.356581495821748e-05}
{"train_loss": 0.21364018321037292, "global_step": 112278, "epoch": 1261, "lr": 6.3565256966482e-05}
{"train_loss": 0.167066290974617, "global_step": 112279, "epoch": 1261, "lr": 6.356469897292281e-05}
{"train_loss": 0.09292152523994446, "global_step": 112280, "epoch": 1261, "lr": 6.356414097754003e-05}
{"train_loss": 0.08475270122289658, "global_step": 112281, "epoch": 1261, "lr": 6.356358298033373e-05}
{"train_loss": 0.16632041335105896, "global_step": 112282, "epoch": 1261, "lr": 6.356302498130396e-05}
{"train_loss": 0.1825728714466095, "global_step": 112283, "epoch": 1261, "lr": 6.35624669804508e-05}
{"train_loss": 0.19084444642066956, "global_step": 112284, "epoch": 1261, "lr": 6.356190897777434e-05}
{"train_loss": 0.20428994297981262, "global_step": 112285, "epoch": 1261, "lr": 6.356135097327465e-05}
{"train_loss": 0.08692122995853424, "global_step": 112286, "epoch": 1261, "lr": 6.35607929669518e-05}
{"train_loss": 0.18079766631126404, "global_step": 112287, "epoch": 1261, "lr": 6.356023495880585e-05}
{"train_loss": 0.14232680201530457, "global_step": 112288, "epoch": 1261, "lr": 6.35596769488369e-05}
{"train_loss": 0.21807461977005005, "global_step": 112289, "epoch": 1261, "lr": 6.355911893704502e-05}
{"train_loss": 0.19528920948505402, "global_step": 112290, "epoch": 1261, "lr": 6.35585609234303e-05}
{"train_loss": 0.10453976690769196, "global_step": 112291, "epoch": 1261, "lr": 6.355800290799278e-05}
{"train_loss": 0.24696843326091766, "global_step": 112292, "epoch": 1261, "lr": 6.355744489073255e-05}
{"train_loss": 0.20309098064899445, "global_step": 112293, "epoch": 1261, "lr": 6.355688687164967e-05}
{"train_loss": 0.1801385134458542, "global_step": 112294, "epoch": 1261, "lr": 6.355632885074426e-05}
{"train_loss": 0.2204660028219223, "global_step": 112295, "epoch": 1261, "lr": 6.355577082801636e-05}
{"train_loss": 0.13261407613754272, "global_step": 112296, "epoch": 1261, "lr": 6.355521280346605e-05}
{"train_loss": 0.0971137285232544, "global_step": 112297, "epoch": 1261, "lr": 6.355465477709341e-05}
{"train_loss": 0.15394996106624603, "global_step": 112298, "epoch": 1261, "lr": 6.355409674889849e-05}
{"train_loss": 0.11114764213562012, "global_step": 112299, "epoch": 1261, "lr": 6.355353871888141e-05}
{"train_loss": 0.14627906680107117, "global_step": 112300, "epoch": 1261, "lr": 6.355298068704222e-05}
{"train_loss": 0.0822036936879158, "global_step": 112301, "epoch": 1261, "lr": 6.3552422653381e-05}
{"train_loss": 0.14795318245887756, "global_step": 112302, "epoch": 1261, "lr": 6.355186461789781e-05}
{"train_loss": 0.19366703927516937, "global_step": 112303, "epoch": 1261, "lr": 6.355130658059274e-05}
{"train_loss": 0.27015721797943115, "global_step": 112304, "epoch": 1261, "lr": 6.355074854146587e-05}
{"train_loss": 0.16593879461288452, "global_step": 112305, "epoch": 1261, "lr": 6.355019050051726e-05}
{"train_loss": 0.16131357848644257, "global_step": 112306, "epoch": 1261, "lr": 6.354963245774699e-05}
{"train_loss": 0.1367708295583725, "global_step": 112307, "epoch": 1261, "lr": 6.354907441315513e-05}
{"train_loss": 0.11316471546888351, "global_step": 112308, "epoch": 1261, "lr": 6.354851636674177e-05}
{"train_loss": 0.19778649508953094, "global_step": 112309, "epoch": 1261, "lr": 6.354795831850698e-05}
{"train_loss": 0.1413784772157669, "global_step": 112310, "epoch": 1261, "lr": 6.354740026845085e-05}
{"train_loss": 0.10951430350542068, "global_step": 112311, "epoch": 1261, "lr": 6.354684221657341e-05}
{"train_loss": 0.15569216012954712, "global_step": 112312, "epoch": 1261, "lr": 6.354628416287478e-05}
{"train_loss": 0.1207236796617508, "global_step": 112313, "epoch": 1261, "lr": 6.3545726107355e-05}
{"train_loss": 0.07637424021959305, "global_step": 112314, "epoch": 1261, "lr": 6.354516805001419e-05}
{"train_loss": 0.1682853400707245, "global_step": 112315, "epoch": 1261, "lr": 6.354460999085237e-05}
{"train_loss": 0.20932947099208832, "global_step": 112316, "epoch": 1261, "lr": 6.354405192986964e-05}
{"train_loss": 0.16003027307183554, "global_step": 112317, "epoch": 1261, "lr": 6.35434938670661e-05, "val_loss": 4.803701877593994}
{"train_loss": 0.1914469301700592, "global_step": 112318, "epoch": 1262, "lr": 6.354293580244179e-05}
{"train_loss": 0.14627504348754883, "global_step": 112319, "epoch": 1262, "lr": 6.354237773599679e-05}
{"train_loss": 0.08743714541196823, "global_step": 112320, "epoch": 1262, "lr": 6.354181966773118e-05}
{"train_loss": 0.16970621049404144, "global_step": 112321, "epoch": 1262, "lr": 6.354126159764506e-05}
{"train_loss": 0.11746273934841156, "global_step": 112322, "epoch": 1262, "lr": 6.354070352573847e-05}
{"train_loss": 0.2067287266254425, "global_step": 112323, "epoch": 1262, "lr": 6.35401454520115e-05}
{"train_loss": 0.11643245071172714, "global_step": 112324, "epoch": 1262, "lr": 6.353958737646422e-05}
{"train_loss": 0.12023928761482239, "global_step": 112325, "epoch": 1262, "lr": 6.353902929909668e-05}
{"train_loss": 0.09286850690841675, "global_step": 112326, "epoch": 1262, "lr": 6.353847121990901e-05}
{"train_loss": 0.17192073166370392, "global_step": 112327, "epoch": 1262, "lr": 6.353791313890126e-05}
{"train_loss": 0.1916273683309555, "global_step": 112328, "epoch": 1262, "lr": 6.35373550560735e-05}
{"train_loss": 0.18051934242248535, "global_step": 112329, "epoch": 1262, "lr": 6.353679697142582e-05}
{"train_loss": 0.13664685189723969, "global_step": 112330, "epoch": 1262, "lr": 6.353623888495827e-05}
{"train_loss": 0.15445999801158905, "global_step": 112331, "epoch": 1262, "lr": 6.353568079667094e-05}
{"train_loss": 0.11321988701820374, "global_step": 112332, "epoch": 1262, "lr": 6.35351227065639e-05}
{"train_loss": 0.14547312259674072, "global_step": 112333, "epoch": 1262, "lr": 6.353456461463722e-05}
{"train_loss": 0.16209909319877625, "global_step": 112334, "epoch": 1262, "lr": 6.353400652089098e-05}
{"train_loss": 0.13259170949459076, "global_step": 112335, "epoch": 1262, "lr": 6.353344842532528e-05}
{"train_loss": 0.11363992840051651, "global_step": 112336, "epoch": 1262, "lr": 6.353289032794016e-05}
{"train_loss": 0.15355367958545685, "global_step": 112337, "epoch": 1262, "lr": 6.353233222873572e-05}
{"train_loss": 0.13305918872356415, "global_step": 112338, "epoch": 1262, "lr": 6.3531774127712e-05}
{"train_loss": 0.0773671343922615, "global_step": 112339, "epoch": 1262, "lr": 6.353121602486911e-05}
{"train_loss": 0.18345437943935394, "global_step": 112340, "epoch": 1262, "lr": 6.353065792020712e-05}
{"train_loss": 0.13872818648815155, "global_step": 112341, "epoch": 1262, "lr": 6.353009981372609e-05}
{"train_loss": 0.10647287964820862, "global_step": 112342, "epoch": 1262, "lr": 6.35295417054261e-05}
{"train_loss": 0.08484378457069397, "global_step": 112343, "epoch": 1262, "lr": 6.352898359530722e-05}
{"train_loss": 0.20374412834644318, "global_step": 112344, "epoch": 1262, "lr": 6.352842548336954e-05}
{"train_loss": 0.09723329544067383, "global_step": 112345, "epoch": 1262, "lr": 6.352786736961314e-05}
{"train_loss": 0.1463036984205246, "global_step": 112346, "epoch": 1262, "lr": 6.352730925403807e-05}
{"train_loss": 0.13260848820209503, "global_step": 112347, "epoch": 1262, "lr": 6.352675113664442e-05}
{"train_loss": 0.1793816238641739, "global_step": 112348, "epoch": 1262, "lr": 6.352619301743228e-05}
{"train_loss": 0.12473171204328537, "global_step": 112349, "epoch": 1262, "lr": 6.35256348964017e-05}
{"train_loss": 0.17263996601104736, "global_step": 112350, "epoch": 1262, "lr": 6.352507677355276e-05}
{"train_loss": 0.174890398979187, "global_step": 112351, "epoch": 1262, "lr": 6.352451864888553e-05}
{"train_loss": 0.14594846963882446, "global_step": 112352, "epoch": 1262, "lr": 6.352396052240011e-05}
{"train_loss": 0.13166846334934235, "global_step": 112353, "epoch": 1262, "lr": 6.352340239409654e-05}
{"train_loss": 0.1203315481543541, "global_step": 112354, "epoch": 1262, "lr": 6.352284426397493e-05}
{"train_loss": 0.11969496309757233, "global_step": 112355, "epoch": 1262, "lr": 6.352228613203532e-05}
{"train_loss": 0.12120693922042847, "global_step": 112356, "epoch": 1262, "lr": 6.352172799827781e-05}
{"train_loss": 0.14881835877895355, "global_step": 112357, "epoch": 1262, "lr": 6.352116986270248e-05}
{"train_loss": 0.09821107238531113, "global_step": 112358, "epoch": 1262, "lr": 6.352061172530939e-05}
{"train_loss": 0.11012545228004456, "global_step": 112359, "epoch": 1262, "lr": 6.35200535860986e-05}
{"train_loss": 0.17710086703300476, "global_step": 112360, "epoch": 1262, "lr": 6.351949544507024e-05}
{"train_loss": 0.21900522708892822, "global_step": 112361, "epoch": 1262, "lr": 6.35189373022243e-05}
{"train_loss": 0.1271422803401947, "global_step": 112362, "epoch": 1262, "lr": 6.351837915756094e-05}
{"train_loss": 0.19453902542591095, "global_step": 112363, "epoch": 1262, "lr": 6.351782101108017e-05}
{"train_loss": 0.17333315312862396, "global_step": 112364, "epoch": 1262, "lr": 6.351726286278213e-05}
{"train_loss": 0.11871613562107086, "global_step": 112365, "epoch": 1262, "lr": 6.351670471266684e-05}
{"train_loss": 0.2569649815559387, "global_step": 112366, "epoch": 1262, "lr": 6.35161465607344e-05}
{"train_loss": 0.21159698069095612, "global_step": 112367, "epoch": 1262, "lr": 6.351558840698486e-05}
{"train_loss": 0.1259203851222992, "global_step": 112368, "epoch": 1262, "lr": 6.351503025141835e-05}
{"train_loss": 0.1625358909368515, "global_step": 112369, "epoch": 1262, "lr": 6.351447209403489e-05}
{"train_loss": 0.14495496451854706, "global_step": 112370, "epoch": 1262, "lr": 6.351391393483455e-05}
{"train_loss": 0.17090071737766266, "global_step": 112371, "epoch": 1262, "lr": 6.351335577381746e-05}
{"train_loss": 0.12362460792064667, "global_step": 112372, "epoch": 1262, "lr": 6.351279761098366e-05}
{"train_loss": 0.1388818323612213, "global_step": 112373, "epoch": 1262, "lr": 6.351223944633322e-05}
{"train_loss": 0.16010336577892303, "global_step": 112374, "epoch": 1262, "lr": 6.351168127986622e-05}
{"train_loss": 0.1319410502910614, "global_step": 112375, "epoch": 1262, "lr": 6.351112311158275e-05}
{"train_loss": 0.26867833733558655, "global_step": 112376, "epoch": 1262, "lr": 6.351056494148287e-05}
{"train_loss": 0.1203460618853569, "global_step": 112377, "epoch": 1262, "lr": 6.351000676956666e-05}
{"train_loss": 0.18576154112815857, "global_step": 112378, "epoch": 1262, "lr": 6.350944859583419e-05}
{"train_loss": 0.1557643562555313, "global_step": 112379, "epoch": 1262, "lr": 6.350889042028553e-05}
{"train_loss": 0.14670057594776154, "global_step": 112380, "epoch": 1262, "lr": 6.350833224292077e-05}
{"train_loss": 0.1601928174495697, "global_step": 112381, "epoch": 1262, "lr": 6.350777406374e-05}
{"train_loss": 0.22501909732818604, "global_step": 112382, "epoch": 1262, "lr": 6.350721588274326e-05}
{"train_loss": 0.13193130493164062, "global_step": 112383, "epoch": 1262, "lr": 6.350665769993063e-05}
{"train_loss": 0.18404771387577057, "global_step": 112384, "epoch": 1262, "lr": 6.35060995153022e-05}
{"train_loss": 0.1903286576271057, "global_step": 112385, "epoch": 1262, "lr": 6.350554132885805e-05}
{"train_loss": 0.09813524782657623, "global_step": 112386, "epoch": 1262, "lr": 6.350498314059825e-05}
{"train_loss": 0.13776086270809174, "global_step": 112387, "epoch": 1262, "lr": 6.350442495052287e-05}
{"train_loss": 0.15853208303451538, "global_step": 112388, "epoch": 1262, "lr": 6.350386675863197e-05}
{"train_loss": 0.1215285211801529, "global_step": 112389, "epoch": 1262, "lr": 6.350330856492564e-05}
{"train_loss": 0.1811363697052002, "global_step": 112390, "epoch": 1262, "lr": 6.350275036940396e-05}
{"train_loss": 0.12781190872192383, "global_step": 112391, "epoch": 1262, "lr": 6.3502192172067e-05}
{"train_loss": 0.10462909191846848, "global_step": 112392, "epoch": 1262, "lr": 6.350163397291483e-05}
{"train_loss": 0.10321597009897232, "global_step": 112393, "epoch": 1262, "lr": 6.350107577194753e-05}
{"train_loss": 0.18525925278663635, "global_step": 112394, "epoch": 1262, "lr": 6.350051756916518e-05}
{"train_loss": 0.25361770391464233, "global_step": 112395, "epoch": 1262, "lr": 6.349995936456785e-05}
{"train_loss": 0.13310502469539642, "global_step": 112396, "epoch": 1262, "lr": 6.34994011581556e-05}
{"train_loss": 0.19392560422420502, "global_step": 112397, "epoch": 1262, "lr": 6.349884294992852e-05}
{"train_loss": 0.16754117608070374, "global_step": 112398, "epoch": 1262, "lr": 6.34982847398867e-05}
{"train_loss": 0.13452987372875214, "global_step": 112399, "epoch": 1262, "lr": 6.34977265280302e-05}
{"train_loss": 0.1294127106666565, "global_step": 112400, "epoch": 1262, "lr": 6.34971683143591e-05}
{"train_loss": 0.1488877385854721, "global_step": 112401, "epoch": 1262, "lr": 6.349661009887346e-05}
{"train_loss": 0.11385399848222733, "global_step": 112402, "epoch": 1262, "lr": 6.349605188157337e-05}
{"train_loss": 0.18215495347976685, "global_step": 112403, "epoch": 1262, "lr": 6.34954936624589e-05}
{"train_loss": 0.16355250775814056, "global_step": 112404, "epoch": 1262, "lr": 6.349493544153012e-05}
{"train_loss": 0.09485206007957458, "global_step": 112405, "epoch": 1262, "lr": 6.349437721878712e-05}
{"train_loss": 0.15063086662734493, "global_step": 112406, "epoch": 1262, "lr": 6.349381899422995e-05, "val_loss": 4.969929218292236}
{"train_loss": 0.17780765891075134, "global_step": 112407, "epoch": 1263, "lr": 6.349326076785872e-05}
{"train_loss": 0.15659941732883453, "global_step": 112408, "epoch": 1263, "lr": 6.349270253967347e-05}
{"train_loss": 0.19626955687999725, "global_step": 112409, "epoch": 1263, "lr": 6.349214430967431e-05}
{"train_loss": 0.18511590361595154, "global_step": 112410, "epoch": 1263, "lr": 6.349158607786127e-05}
{"train_loss": 0.13560788333415985, "global_step": 112411, "epoch": 1263, "lr": 6.349102784423447e-05}
{"train_loss": 0.12340312451124191, "global_step": 112412, "epoch": 1263, "lr": 6.349046960879395e-05}
{"train_loss": 0.12327449023723602, "global_step": 112413, "epoch": 1263, "lr": 6.348991137153981e-05}
{"train_loss": 0.10846984386444092, "global_step": 112414, "epoch": 1263, "lr": 6.348935313247212e-05}
{"train_loss": 0.15738947689533234, "global_step": 112415, "epoch": 1263, "lr": 6.348879489159094e-05}
{"train_loss": 0.15528079867362976, "global_step": 112416, "epoch": 1263, "lr": 6.348823664889636e-05}
{"train_loss": 0.14631159603595734, "global_step": 112417, "epoch": 1263, "lr": 6.348767840438845e-05}
{"train_loss": 0.12483438849449158, "global_step": 112418, "epoch": 1263, "lr": 6.348712015806729e-05}
{"train_loss": 0.1625141203403473, "global_step": 112419, "epoch": 1263, "lr": 6.348656190993295e-05}
{"train_loss": 0.1103154793381691, "global_step": 112420, "epoch": 1263, "lr": 6.34860036599855e-05}
{"train_loss": 0.24736186861991882, "global_step": 112421, "epoch": 1263, "lr": 6.348544540822503e-05}
{"train_loss": 0.15515664219856262, "global_step": 112422, "epoch": 1263, "lr": 6.348488715465161e-05}
{"train_loss": 0.18189142644405365, "global_step": 112423, "epoch": 1263, "lr": 6.34843288992653e-05}
{"train_loss": 0.12576770782470703, "global_step": 112424, "epoch": 1263, "lr": 6.348377064206618e-05}
{"train_loss": 0.1293179839849472, "global_step": 112425, "epoch": 1263, "lr": 6.348321238305436e-05}
{"train_loss": 0.1260165572166443, "global_step": 112426, "epoch": 1263, "lr": 6.348265412222987e-05}
{"train_loss": 0.1049165353178978, "global_step": 112427, "epoch": 1263, "lr": 6.34820958595928e-05}
{"train_loss": 0.2192019522190094, "global_step": 112428, "epoch": 1263, "lr": 6.348153759514323e-05}
{"train_loss": 0.19080626964569092, "global_step": 112429, "epoch": 1263, "lr": 6.348097932888122e-05}
{"train_loss": 0.14216303825378418, "global_step": 112430, "epoch": 1263, "lr": 6.348042106080688e-05}
{"train_loss": 0.14621256291866302, "global_step": 112431, "epoch": 1263, "lr": 6.347986279092024e-05}
{"train_loss": 0.08747559040784836, "global_step": 112432, "epoch": 1263, "lr": 6.347930451922141e-05}
{"train_loss": 0.1825784593820572, "global_step": 112433, "epoch": 1263, "lr": 6.347874624571044e-05}
{"train_loss": 0.19983750581741333, "global_step": 112434, "epoch": 1263, "lr": 6.347818797038742e-05}
{"train_loss": 0.1217598170042038, "global_step": 112435, "epoch": 1263, "lr": 6.347762969325242e-05}
{"train_loss": 0.10154026001691818, "global_step": 112436, "epoch": 1263, "lr": 6.347707141430553e-05}
{"train_loss": 0.17277799546718597, "global_step": 112437, "epoch": 1263, "lr": 6.347651313354679e-05}
{"train_loss": 0.13816654682159424, "global_step": 112438, "epoch": 1263, "lr": 6.34759548509763e-05}
{"train_loss": 0.11188435554504395, "global_step": 112439, "epoch": 1263, "lr": 6.347539656659415e-05}
{"train_loss": 0.12958700954914093, "global_step": 112440, "epoch": 1263, "lr": 6.34748382804004e-05}
{"train_loss": 0.14913657307624817, "global_step": 112441, "epoch": 1263, "lr": 6.34742799923951e-05}
{"train_loss": 0.2023526132106781, "global_step": 112442, "epoch": 1263, "lr": 6.347372170257836e-05}
{"train_loss": 0.18189473450183868, "global_step": 112443, "epoch": 1263, "lr": 6.347316341095024e-05}
{"train_loss": 0.1448877453804016, "global_step": 112444, "epoch": 1263, "lr": 6.347260511751082e-05}
{"train_loss": 0.2221192717552185, "global_step": 112445, "epoch": 1263, "lr": 6.347204682226017e-05}
{"train_loss": 0.14268669486045837, "global_step": 112446, "epoch": 1263, "lr": 6.347148852519838e-05}
{"train_loss": 0.14822977781295776, "global_step": 112447, "epoch": 1263, "lr": 6.347093022632551e-05}
{"train_loss": 0.15617480874061584, "global_step": 112448, "epoch": 1263, "lr": 6.347037192564163e-05}
{"train_loss": 0.11275631189346313, "global_step": 112449, "epoch": 1263, "lr": 6.346981362314681e-05}
{"train_loss": 0.21513256430625916, "global_step": 112450, "epoch": 1263, "lr": 6.346925531884115e-05}
{"train_loss": 0.08623558282852173, "global_step": 112451, "epoch": 1263, "lr": 6.346869701272471e-05}
{"train_loss": 0.13052916526794434, "global_step": 112452, "epoch": 1263, "lr": 6.346813870479756e-05}
{"train_loss": 0.1399986892938614, "global_step": 112453, "epoch": 1263, "lr": 6.34675803950598e-05}
{"train_loss": 0.09386667609214783, "global_step": 112454, "epoch": 1263, "lr": 6.346702208351147e-05}
{"train_loss": 0.11260611563920975, "global_step": 112455, "epoch": 1263, "lr": 6.346646377015268e-05}
{"train_loss": 0.10884779691696167, "global_step": 112456, "epoch": 1263, "lr": 6.346590545498348e-05}
{"train_loss": 0.13950784504413605, "global_step": 112457, "epoch": 1263, "lr": 6.346534713800396e-05}
{"train_loss": 0.21367289125919342, "global_step": 112458, "epoch": 1263, "lr": 6.346478881921419e-05}
{"train_loss": 0.1761535108089447, "global_step": 112459, "epoch": 1263, "lr": 6.346423049861423e-05}
{"train_loss": 0.13818581402301788, "global_step": 112460, "epoch": 1263, "lr": 6.346367217620417e-05}
{"train_loss": 0.15666992962360382, "global_step": 112461, "epoch": 1263, "lr": 6.346311385198409e-05}
{"train_loss": 0.1354457437992096, "global_step": 112462, "epoch": 1263, "lr": 6.346255552595406e-05}
{"train_loss": 0.16692322492599487, "global_step": 112463, "epoch": 1263, "lr": 6.346199719811414e-05}
{"train_loss": 0.2263680100440979, "global_step": 112464, "epoch": 1263, "lr": 6.346143886846443e-05}
{"train_loss": 0.12999506294727325, "global_step": 112465, "epoch": 1263, "lr": 6.346088053700501e-05}
{"train_loss": 0.20254874229431152, "global_step": 112466, "epoch": 1263, "lr": 6.346032220373592e-05}
{"train_loss": 0.205397829413414, "global_step": 112467, "epoch": 1263, "lr": 6.345976386865726e-05}
{"train_loss": 0.1266433745622635, "global_step": 112468, "epoch": 1263, "lr": 6.345920553176908e-05}
{"train_loss": 0.08189921826124191, "global_step": 112469, "epoch": 1263, "lr": 6.345864719307149e-05}
{"train_loss": 0.18431635200977325, "global_step": 112470, "epoch": 1263, "lr": 6.345808885256454e-05}
{"train_loss": 0.1429382860660553, "global_step": 112471, "epoch": 1263, "lr": 6.345753051024832e-05}
{"train_loss": 0.13220250606536865, "global_step": 112472, "epoch": 1263, "lr": 6.34569721661229e-05}
{"train_loss": 0.12403128296136856, "global_step": 112473, "epoch": 1263, "lr": 6.345641382018836e-05}
{"train_loss": 0.10904628038406372, "global_step": 112474, "epoch": 1263, "lr": 6.345585547244475e-05}
{"train_loss": 0.18397964537143707, "global_step": 112475, "epoch": 1263, "lr": 6.345529712289218e-05}
{"train_loss": 0.07260066270828247, "global_step": 112476, "epoch": 1263, "lr": 6.34547387715307e-05}
{"train_loss": 0.1314038783311844, "global_step": 112477, "epoch": 1263, "lr": 6.34541804183604e-05}
{"train_loss": 0.11085265129804611, "global_step": 112478, "epoch": 1263, "lr": 6.345362206338136e-05}
{"train_loss": 0.1962192952632904, "global_step": 112479, "epoch": 1263, "lr": 6.345306370659363e-05}
{"train_loss": 0.14852796494960785, "global_step": 112480, "epoch": 1263, "lr": 6.345250534799731e-05}
{"train_loss": 0.13395848870277405, "global_step": 112481, "epoch": 1263, "lr": 6.345194698759244e-05}
{"train_loss": 0.14304111897945404, "global_step": 112482, "epoch": 1263, "lr": 6.345138862537915e-05}
{"train_loss": 0.11770838499069214, "global_step": 112483, "epoch": 1263, "lr": 6.345083026135748e-05}
{"train_loss": 0.17526523768901825, "global_step": 112484, "epoch": 1263, "lr": 6.34502718955275e-05}
{"train_loss": 0.0833909884095192, "global_step": 112485, "epoch": 1263, "lr": 6.344971352788932e-05}
{"train_loss": 0.11762270331382751, "global_step": 112486, "epoch": 1263, "lr": 6.344915515844295e-05}
{"train_loss": 0.15836210548877716, "global_step": 112487, "epoch": 1263, "lr": 6.344859678718853e-05}
{"train_loss": 0.11413753032684326, "global_step": 112488, "epoch": 1263, "lr": 6.34480384141261e-05}
{"train_loss": 0.16009365022182465, "global_step": 112489, "epoch": 1263, "lr": 6.344748003925573e-05}
{"train_loss": 0.1630258858203888, "global_step": 112490, "epoch": 1263, "lr": 6.344692166257753e-05}
{"train_loss": 0.1853538602590561, "global_step": 112491, "epoch": 1263, "lr": 6.344636328409155e-05}
{"train_loss": 0.14881546795368195, "global_step": 112492, "epoch": 1263, "lr": 6.344580490379788e-05}
{"train_loss": 0.12122992426156998, "global_step": 112493, "epoch": 1263, "lr": 6.344524652169658e-05}
{"train_loss": 0.13028141856193542, "global_step": 112494, "epoch": 1263, "lr": 6.344468813778772e-05}
{"train_loss": 0.1467172522725684, "global_step": 112495, "epoch": 1263, "lr": 6.34441297520714e-05, "val_loss": 5.028831481933594}
{"train_loss": 0.20380021631717682, "global_step": 112496, "epoch": 1264, "lr": 6.344357136454768e-05}
{"train_loss": 0.16667696833610535, "global_step": 112497, "epoch": 1264, "lr": 6.344301297521663e-05}
{"train_loss": 0.14221060276031494, "global_step": 112498, "epoch": 1264, "lr": 6.344245458407834e-05}
{"train_loss": 0.19392581284046173, "global_step": 112499, "epoch": 1264, "lr": 6.344189619113285e-05}
{"train_loss": 0.12695161998271942, "global_step": 112500, "epoch": 1264, "lr": 6.344133779638028e-05}
{"train_loss": 0.10251562297344208, "global_step": 112501, "epoch": 1264, "lr": 6.34407793998207e-05}
{"train_loss": 0.1680741310119629, "global_step": 112502, "epoch": 1264, "lr": 6.344022100145415e-05}
{"train_loss": 0.18207506835460663, "global_step": 112503, "epoch": 1264, "lr": 6.343966260128074e-05}
{"train_loss": 0.13395372033119202, "global_step": 112504, "epoch": 1264, "lr": 6.343910419930053e-05}
{"train_loss": 0.1793074756860733, "global_step": 112505, "epoch": 1264, "lr": 6.343854579551359e-05}
{"train_loss": 0.09961152821779251, "global_step": 112506, "epoch": 1264, "lr": 6.343798738991999e-05}
{"train_loss": 0.17747816443443298, "global_step": 112507, "epoch": 1264, "lr": 6.343742898251982e-05}
{"train_loss": 0.10075926780700684, "global_step": 112508, "epoch": 1264, "lr": 6.343687057331315e-05}
{"train_loss": 0.12128088623285294, "global_step": 112509, "epoch": 1264, "lr": 6.343631216230007e-05}
{"train_loss": 0.16092273592948914, "global_step": 112510, "epoch": 1264, "lr": 6.343575374948064e-05}
{"train_loss": 0.15435494482517242, "global_step": 112511, "epoch": 1264, "lr": 6.343519533485493e-05}
{"train_loss": 0.1166800782084465, "global_step": 112512, "epoch": 1264, "lr": 6.343463691842303e-05}
{"train_loss": 0.11800923198461533, "global_step": 112513, "epoch": 1264, "lr": 6.3434078500185e-05}
{"train_loss": 0.1372869908809662, "global_step": 112514, "epoch": 1264, "lr": 6.343352008014091e-05}
{"train_loss": 0.1264573335647583, "global_step": 112515, "epoch": 1264, "lr": 6.343296165829087e-05}
{"train_loss": 0.1058492362499237, "global_step": 112516, "epoch": 1264, "lr": 6.343240323463493e-05}
{"train_loss": 0.11825752258300781, "global_step": 112517, "epoch": 1264, "lr": 6.343184480917316e-05}
{"train_loss": 0.12243400514125824, "global_step": 112518, "epoch": 1264, "lr": 6.343128638190564e-05}
{"train_loss": 0.1630682349205017, "global_step": 112519, "epoch": 1264, "lr": 6.343072795283245e-05}
{"train_loss": 0.13384459912776947, "global_step": 112520, "epoch": 1264, "lr": 6.343016952195367e-05}
{"train_loss": 0.2512372136116028, "global_step": 112521, "epoch": 1264, "lr": 6.342961108926935e-05}
{"train_loss": 0.1984601765871048, "global_step": 112522, "epoch": 1264, "lr": 6.342905265477959e-05}
{"train_loss": 0.11684190481901169, "global_step": 112523, "epoch": 1264, "lr": 6.342849421848447e-05}
{"train_loss": 0.1401074081659317, "global_step": 112524, "epoch": 1264, "lr": 6.342793578038402e-05}
{"train_loss": 0.2247515767812729, "global_step": 112525, "epoch": 1264, "lr": 6.342737734047838e-05}
{"train_loss": 0.08260104060173035, "global_step": 112526, "epoch": 1264, "lr": 6.342681889876757e-05}
{"train_loss": 0.13397596776485443, "global_step": 112527, "epoch": 1264, "lr": 6.342626045525171e-05}
{"train_loss": 0.14890843629837036, "global_step": 112528, "epoch": 1264, "lr": 6.342570200993084e-05}
{"train_loss": 0.15199585258960724, "global_step": 112529, "epoch": 1264, "lr": 6.342514356280506e-05}
{"train_loss": 0.0756717398762703, "global_step": 112530, "epoch": 1264, "lr": 6.34245851138744e-05}
{"train_loss": 0.19863034784793854, "global_step": 112531, "epoch": 1264, "lr": 6.342402666313902e-05}
{"train_loss": 0.07920922338962555, "global_step": 112532, "epoch": 1264, "lr": 6.34234682105989e-05}
{"train_loss": 0.23962560296058655, "global_step": 112533, "epoch": 1264, "lr": 6.342290975625418e-05}
{"train_loss": 0.1329156458377838, "global_step": 112534, "epoch": 1264, "lr": 6.34223513001049e-05}
{"train_loss": 0.09930047392845154, "global_step": 112535, "epoch": 1264, "lr": 6.342179284215116e-05}
{"train_loss": 0.11089752614498138, "global_step": 112536, "epoch": 1264, "lr": 6.342123438239303e-05}
{"train_loss": 0.15580081939697266, "global_step": 112537, "epoch": 1264, "lr": 6.342067592083056e-05}
{"train_loss": 0.18426890671253204, "global_step": 112538, "epoch": 1264, "lr": 6.342011745746384e-05}
{"train_loss": 0.15237578749656677, "global_step": 112539, "epoch": 1264, "lr": 6.341955899229298e-05}
{"train_loss": 0.12173564732074738, "global_step": 112540, "epoch": 1264, "lr": 6.3419000525318e-05}
{"train_loss": 0.18208932876586914, "global_step": 112541, "epoch": 1264, "lr": 6.341844205653899e-05}
{"train_loss": 0.15291570127010345, "global_step": 112542, "epoch": 1264, "lr": 6.341788358595606e-05}
{"train_loss": 0.16595840454101562, "global_step": 112543, "epoch": 1264, "lr": 6.341732511356924e-05}
{"train_loss": 0.12227825075387955, "global_step": 112544, "epoch": 1264, "lr": 6.341676663937864e-05}
{"train_loss": 0.20338140428066254, "global_step": 112545, "epoch": 1264, "lr": 6.34162081633843e-05}
{"train_loss": 0.05771266669034958, "global_step": 112546, "epoch": 1264, "lr": 6.341564968558632e-05}
{"train_loss": 0.10821740329265594, "global_step": 112547, "epoch": 1264, "lr": 6.341509120598478e-05}
{"train_loss": 0.12671974301338196, "global_step": 112548, "epoch": 1264, "lr": 6.341453272457974e-05}
{"train_loss": 0.12142025679349899, "global_step": 112549, "epoch": 1264, "lr": 6.341397424137126e-05}
{"train_loss": 0.13144725561141968, "global_step": 112550, "epoch": 1264, "lr": 6.341341575635946e-05}
{"train_loss": 0.11325683444738388, "global_step": 112551, "epoch": 1264, "lr": 6.341285726954438e-05}
{"train_loss": 0.11147971451282501, "global_step": 112552, "epoch": 1264, "lr": 6.34122987809261e-05}
{"train_loss": 0.13268828392028809, "global_step": 112553, "epoch": 1264, "lr": 6.341174029050471e-05}
{"train_loss": 0.20815534889698029, "global_step": 112554, "epoch": 1264, "lr": 6.341118179828027e-05}
{"train_loss": 0.11855557560920715, "global_step": 112555, "epoch": 1264, "lr": 6.341062330425287e-05}
{"train_loss": 0.09408695995807648, "global_step": 112556, "epoch": 1264, "lr": 6.341006480842256e-05}
{"train_loss": 0.18047724664211273, "global_step": 112557, "epoch": 1264, "lr": 6.340950631078944e-05}
{"train_loss": 0.09778156131505966, "global_step": 112558, "epoch": 1264, "lr": 6.340894781135357e-05}
{"train_loss": 0.1029198095202446, "global_step": 112559, "epoch": 1264, "lr": 6.340838931011503e-05}
{"train_loss": 0.14074240624904633, "global_step": 112560, "epoch": 1264, "lr": 6.34078308070739e-05}
{"train_loss": 0.09519937634468079, "global_step": 112561, "epoch": 1264, "lr": 6.340727230223024e-05}
{"train_loss": 0.13671770691871643, "global_step": 112562, "epoch": 1264, "lr": 6.340671379558416e-05}
{"train_loss": 0.10452884435653687, "global_step": 112563, "epoch": 1264, "lr": 6.34061552871357e-05}
{"train_loss": 0.14154358208179474, "global_step": 112564, "epoch": 1264, "lr": 6.340559677688493e-05}
{"train_loss": 0.09434310346841812, "global_step": 112565, "epoch": 1264, "lr": 6.340503826483196e-05}
{"train_loss": 0.19029101729393005, "global_step": 112566, "epoch": 1264, "lr": 6.340447975097682e-05}
{"train_loss": 0.1396513134241104, "global_step": 112567, "epoch": 1264, "lr": 6.340392123531964e-05}
{"train_loss": 0.11119059473276138, "global_step": 112568, "epoch": 1264, "lr": 6.340336271786044e-05}
{"train_loss": 0.13538257777690887, "global_step": 112569, "epoch": 1264, "lr": 6.340280419859936e-05}
{"train_loss": 0.12971559166908264, "global_step": 112570, "epoch": 1264, "lr": 6.34022456775364e-05}
{"train_loss": 0.16659361124038696, "global_step": 112571, "epoch": 1264, "lr": 6.340168715467169e-05}
{"train_loss": 0.18142682313919067, "global_step": 112572, "epoch": 1264, "lr": 6.340112863000529e-05}
{"train_loss": 0.17372289299964905, "global_step": 112573, "epoch": 1264, "lr": 6.340057010353725e-05}
{"train_loss": 0.17372481524944305, "global_step": 112574, "epoch": 1264, "lr": 6.340001157526766e-05}
{"train_loss": 0.16890709102153778, "global_step": 112575, "epoch": 1264, "lr": 6.339945304519663e-05}
{"train_loss": 0.14784611761569977, "global_step": 112576, "epoch": 1264, "lr": 6.33988945133242e-05}
{"train_loss": 0.1535973995923996, "global_step": 112577, "epoch": 1264, "lr": 6.339833597965046e-05}
{"train_loss": 0.18952220678329468, "global_step": 112578, "epoch": 1264, "lr": 6.339777744417546e-05}
{"train_loss": 0.09278060495853424, "global_step": 112579, "epoch": 1264, "lr": 6.33972189068993e-05}
{"train_loss": 0.10495913028717041, "global_step": 112580, "epoch": 1264, "lr": 6.339666036782204e-05}
{"train_loss": 0.17389732599258423, "global_step": 112581, "epoch": 1264, "lr": 6.339610182694378e-05}
{"train_loss": 0.06492312252521515, "global_step": 112582, "epoch": 1264, "lr": 6.339554328426455e-05}
{"train_loss": 0.09129831194877625, "global_step": 112583, "epoch": 1264, "lr": 6.339498473978446e-05}
{"train_loss": 0.1408367585600092, "global_step": 112584, "epoch": 1264, "lr": 6.33944261935036e-05, "val_loss": 5.173287391662598}
{"train_loss": 0.14722798764705658, "global_step": 112585, "epoch": 1265, "lr": 6.339386764542202e-05}
{"train_loss": 0.10432641208171844, "global_step": 112586, "epoch": 1265, "lr": 6.339330909553978e-05}
{"train_loss": 0.16059571504592896, "global_step": 112587, "epoch": 1265, "lr": 6.339275054385697e-05}
{"train_loss": 0.15978877246379852, "global_step": 112588, "epoch": 1265, "lr": 6.339219199037368e-05}
{"train_loss": 0.07804255932569504, "global_step": 112589, "epoch": 1265, "lr": 6.339163343508997e-05}
{"train_loss": 0.14425316452980042, "global_step": 112590, "epoch": 1265, "lr": 6.339107487800591e-05}
{"train_loss": 0.17552797496318817, "global_step": 112591, "epoch": 1265, "lr": 6.33905163191216e-05}
{"train_loss": 0.1913459300994873, "global_step": 112592, "epoch": 1265, "lr": 6.338995775843708e-05}
{"train_loss": 0.1325041949748993, "global_step": 112593, "epoch": 1265, "lr": 6.338939919595246e-05}
{"train_loss": 0.17161929607391357, "global_step": 112594, "epoch": 1265, "lr": 6.338884063166779e-05}
{"train_loss": 0.17814312875270844, "global_step": 112595, "epoch": 1265, "lr": 6.338828206558315e-05}
{"train_loss": 0.16170842945575714, "global_step": 112596, "epoch": 1265, "lr": 6.338772349769862e-05}
{"train_loss": 0.11861054599285126, "global_step": 112597, "epoch": 1265, "lr": 6.338716492801429e-05}
{"train_loss": 0.16151291131973267, "global_step": 112598, "epoch": 1265, "lr": 6.33866063565302e-05}
{"train_loss": 0.15272265672683716, "global_step": 112599, "epoch": 1265, "lr": 6.338604778324645e-05}
{"train_loss": 0.08462373167276382, "global_step": 112600, "epoch": 1265, "lr": 6.33854892081631e-05}
{"train_loss": 0.15104296803474426, "global_step": 112601, "epoch": 1265, "lr": 6.338493063128025e-05}
{"train_loss": 0.23847877979278564, "global_step": 112602, "epoch": 1265, "lr": 6.338437205259795e-05}
{"train_loss": 0.09443853795528412, "global_step": 112603, "epoch": 1265, "lr": 6.338381347211629e-05}
{"train_loss": 0.16628605127334595, "global_step": 112604, "epoch": 1265, "lr": 6.338325488983533e-05}
{"train_loss": 0.19898997247219086, "global_step": 112605, "epoch": 1265, "lr": 6.338269630575516e-05}
{"train_loss": 0.14299336075782776, "global_step": 112606, "epoch": 1265, "lr": 6.338213771987585e-05}
{"train_loss": 0.12335637211799622, "global_step": 112607, "epoch": 1265, "lr": 6.338157913219747e-05}
{"train_loss": 0.15733109414577484, "global_step": 112608, "epoch": 1265, "lr": 6.338102054272011e-05}
{"train_loss": 0.09638486802577972, "global_step": 112609, "epoch": 1265, "lr": 6.338046195144382e-05}
{"train_loss": 0.14551259577274323, "global_step": 112610, "epoch": 1265, "lr": 6.33799033583687e-05}
{"train_loss": 0.118849977850914, "global_step": 112611, "epoch": 1265, "lr": 6.337934476349483e-05}
{"train_loss": 0.09125005453824997, "global_step": 112612, "epoch": 1265, "lr": 6.337878616682225e-05}
{"train_loss": 0.136122465133667, "global_step": 112613, "epoch": 1265, "lr": 6.337822756835106e-05}
{"train_loss": 0.22171472012996674, "global_step": 112614, "epoch": 1265, "lr": 6.337766896808133e-05}
{"train_loss": 0.16257593035697937, "global_step": 112615, "epoch": 1265, "lr": 6.337711036601312e-05}
{"train_loss": 0.1366729438304901, "global_step": 112616, "epoch": 1265, "lr": 6.337655176214655e-05}
{"train_loss": 0.1637328863143921, "global_step": 112617, "epoch": 1265, "lr": 6.337599315648165e-05}
{"train_loss": 0.19272026419639587, "global_step": 112618, "epoch": 1265, "lr": 6.337543454901851e-05}
{"train_loss": 0.10690410435199738, "global_step": 112619, "epoch": 1265, "lr": 6.337487593975722e-05}
{"train_loss": 0.13606658577919006, "global_step": 112620, "epoch": 1265, "lr": 6.337431732869782e-05}
{"train_loss": 0.12964677810668945, "global_step": 112621, "epoch": 1265, "lr": 6.337375871584042e-05}
{"train_loss": 0.13072049617767334, "global_step": 112622, "epoch": 1265, "lr": 6.337320010118507e-05}
{"train_loss": 0.15228313207626343, "global_step": 112623, "epoch": 1265, "lr": 6.337264148473186e-05}
{"train_loss": 0.135331928730011, "global_step": 112624, "epoch": 1265, "lr": 6.337208286648086e-05}
{"train_loss": 0.17226822674274445, "global_step": 112625, "epoch": 1265, "lr": 6.337152424643215e-05}
{"train_loss": 0.16086165606975555, "global_step": 112626, "epoch": 1265, "lr": 6.337096562458581e-05}
{"train_loss": 0.12203880399465561, "global_step": 112627, "epoch": 1265, "lr": 6.33704070009419e-05}
{"train_loss": 0.15735304355621338, "global_step": 112628, "epoch": 1265, "lr": 6.33698483755005e-05}
{"train_loss": 0.19013461470603943, "global_step": 112629, "epoch": 1265, "lr": 6.336928974826169e-05}
{"train_loss": 0.15437787771224976, "global_step": 112630, "epoch": 1265, "lr": 6.336873111922554e-05}
{"train_loss": 0.17042848467826843, "global_step": 112631, "epoch": 1265, "lr": 6.336817248839212e-05}
{"train_loss": 0.1383063793182373, "global_step": 112632, "epoch": 1265, "lr": 6.336761385576152e-05}
{"train_loss": 0.1367940753698349, "global_step": 112633, "epoch": 1265, "lr": 6.33670552213338e-05}
{"train_loss": 0.14641621708869934, "global_step": 112634, "epoch": 1265, "lr": 6.336649658510904e-05}
{"train_loss": 0.10972047597169876, "global_step": 112635, "epoch": 1265, "lr": 6.336593794708734e-05}
{"train_loss": 0.16097933053970337, "global_step": 112636, "epoch": 1265, "lr": 6.336537930726874e-05}
{"train_loss": 0.17913132905960083, "global_step": 112637, "epoch": 1265, "lr": 6.336482066565333e-05}
{"train_loss": 0.1010904312133789, "global_step": 112638, "epoch": 1265, "lr": 6.336426202224117e-05}
{"train_loss": 0.13244280219078064, "global_step": 112639, "epoch": 1265, "lr": 6.336370337703236e-05}
{"train_loss": 0.13591043651103973, "global_step": 112640, "epoch": 1265, "lr": 6.336314473002696e-05}
{"train_loss": 0.15281634032726288, "global_step": 112641, "epoch": 1265, "lr": 6.336258608122504e-05}
{"train_loss": 0.09877678006887436, "global_step": 112642, "epoch": 1265, "lr": 6.336202743062669e-05}
{"train_loss": 0.1504894495010376, "global_step": 112643, "epoch": 1265, "lr": 6.336146877823198e-05}
{"train_loss": 0.1224871277809143, "global_step": 112644, "epoch": 1265, "lr": 6.336091012404098e-05}
{"train_loss": 0.16944456100463867, "global_step": 112645, "epoch": 1265, "lr": 6.336035146805377e-05}
{"train_loss": 0.12675011157989502, "global_step": 112646, "epoch": 1265, "lr": 6.335979281027042e-05}
{"train_loss": 0.12284577637910843, "global_step": 112647, "epoch": 1265, "lr": 6.335923415069102e-05}
{"train_loss": 0.12401946634054184, "global_step": 112648, "epoch": 1265, "lr": 6.335867548931562e-05}
{"train_loss": 0.14240817725658417, "global_step": 112649, "epoch": 1265, "lr": 6.335811682614432e-05}
{"train_loss": 0.1821407526731491, "global_step": 112650, "epoch": 1265, "lr": 6.335755816117719e-05}
{"train_loss": 0.10664204508066177, "global_step": 112651, "epoch": 1265, "lr": 6.335699949441427e-05}
{"train_loss": 0.15244698524475098, "global_step": 112652, "epoch": 1265, "lr": 6.335644082585569e-05}
{"train_loss": 0.14808520674705505, "global_step": 112653, "epoch": 1265, "lr": 6.33558821555015e-05}
{"train_loss": 0.15090090036392212, "global_step": 112654, "epoch": 1265, "lr": 6.335532348335176e-05}
{"train_loss": 0.11073355376720428, "global_step": 112655, "epoch": 1265, "lr": 6.335476480940656e-05}
{"train_loss": 0.14579418301582336, "global_step": 112656, "epoch": 1265, "lr": 6.335420613366598e-05}
{"train_loss": 0.07687016576528549, "global_step": 112657, "epoch": 1265, "lr": 6.33536474561301e-05}
{"train_loss": 0.1736004501581192, "global_step": 112658, "epoch": 1265, "lr": 6.335308877679897e-05}
{"train_loss": 0.09471940249204636, "global_step": 112659, "epoch": 1265, "lr": 6.335253009567269e-05}
{"train_loss": 0.10758485645055771, "global_step": 112660, "epoch": 1265, "lr": 6.335197141275131e-05}
{"train_loss": 0.14488482475280762, "global_step": 112661, "epoch": 1265, "lr": 6.335141272803495e-05}
{"train_loss": 0.15945866703987122, "global_step": 112662, "epoch": 1265, "lr": 6.335085404152363e-05}
{"train_loss": 0.19686511158943176, "global_step": 112663, "epoch": 1265, "lr": 6.335029535321746e-05}
{"train_loss": 0.12302585691213608, "global_step": 112664, "epoch": 1265, "lr": 6.33497366631165e-05}
{"train_loss": 0.12957657873630524, "global_step": 112665, "epoch": 1265, "lr": 6.334917797122083e-05}
{"train_loss": 0.07981489598751068, "global_step": 112666, "epoch": 1265, "lr": 6.334861927753053e-05}
{"train_loss": 0.1926618367433548, "global_step": 112667, "epoch": 1265, "lr": 6.334806058204566e-05}
{"train_loss": 0.1034255400300026, "global_step": 112668, "epoch": 1265, "lr": 6.334750188476632e-05}
{"train_loss": 0.08909474313259125, "global_step": 112669, "epoch": 1265, "lr": 6.334694318569257e-05}
{"train_loss": 0.11192132532596588, "global_step": 112670, "epoch": 1265, "lr": 6.334638448482446e-05}
{"train_loss": 0.11644230782985687, "global_step": 112671, "epoch": 1265, "lr": 6.334582578216212e-05}
{"train_loss": 0.08502518385648727, "global_step": 112672, "epoch": 1265, "lr": 6.334526707770561e-05}
{"train_loss": 0.14098069666141874, "global_step": 112673, "epoch": 1265, "lr": 6.334470837145497e-05, "val_loss": 5.115821838378906, "train_action_mse_error": 6.7414655685424805}
{"train_loss": 0.14389988780021667, "global_step": 112674, "epoch": 1266, "lr": 6.33441496634103e-05}
{"train_loss": 0.1433938592672348, "global_step": 112675, "epoch": 1266, "lr": 6.334359095357165e-05}
{"train_loss": 0.12233461439609528, "global_step": 112676, "epoch": 1266, "lr": 6.334303224193915e-05}
{"train_loss": 0.1305142045021057, "global_step": 112677, "epoch": 1266, "lr": 6.334247352851282e-05}
{"train_loss": 0.18511144816875458, "global_step": 112678, "epoch": 1266, "lr": 6.334191481329277e-05}
{"train_loss": 0.16418878734111786, "global_step": 112679, "epoch": 1266, "lr": 6.334135609627905e-05}
{"train_loss": 0.20397718250751495, "global_step": 112680, "epoch": 1266, "lr": 6.334079737747176e-05}
{"train_loss": 0.18979762494564056, "global_step": 112681, "epoch": 1266, "lr": 6.334023865687096e-05}
{"train_loss": 0.12927663326263428, "global_step": 112682, "epoch": 1266, "lr": 6.333967993447672e-05}
{"train_loss": 0.16206374764442444, "global_step": 112683, "epoch": 1266, "lr": 6.333912121028913e-05}
{"train_loss": 0.12562626600265503, "global_step": 112684, "epoch": 1266, "lr": 6.333856248430824e-05}
{"train_loss": 0.1674424707889557, "global_step": 112685, "epoch": 1266, "lr": 6.333800375653417e-05}
{"train_loss": 0.09409670531749725, "global_step": 112686, "epoch": 1266, "lr": 6.333744502696696e-05}
{"train_loss": 0.1295892298221588, "global_step": 112687, "epoch": 1266, "lr": 6.33368862956067e-05}
{"train_loss": 0.15222449600696564, "global_step": 112688, "epoch": 1266, "lr": 6.333632756245342e-05}
{"train_loss": 0.16365234553813934, "global_step": 112689, "epoch": 1266, "lr": 6.333576882750726e-05}
{"train_loss": 0.19189631938934326, "global_step": 112690, "epoch": 1266, "lr": 6.333521009076827e-05}
{"train_loss": 0.11998986452817917, "global_step": 112691, "epoch": 1266, "lr": 6.333465135223652e-05}
{"train_loss": 0.16012492775917053, "global_step": 112692, "epoch": 1266, "lr": 6.33340926119121e-05}
{"train_loss": 0.17048609256744385, "global_step": 112693, "epoch": 1266, "lr": 6.333353386979506e-05}
{"train_loss": 0.11440014839172363, "global_step": 112694, "epoch": 1266, "lr": 6.333297512588551e-05}
{"train_loss": 0.15588945150375366, "global_step": 112695, "epoch": 1266, "lr": 6.333241638018347e-05}
{"train_loss": 0.15644586086273193, "global_step": 112696, "epoch": 1266, "lr": 6.333185763268909e-05}
{"train_loss": 0.1868007928133011, "global_step": 112697, "epoch": 1266, "lr": 6.333129888340237e-05}
{"train_loss": 0.1113472729921341, "global_step": 112698, "epoch": 1266, "lr": 6.333074013232344e-05}
{"train_loss": 0.1085510402917862, "global_step": 112699, "epoch": 1266, "lr": 6.333018137945233e-05}
{"train_loss": 0.21628330647945404, "global_step": 112700, "epoch": 1266, "lr": 6.332962262478916e-05}
{"train_loss": 0.1028679758310318, "global_step": 112701, "epoch": 1266, "lr": 6.332906386833398e-05}
{"train_loss": 0.07806470990180969, "global_step": 112702, "epoch": 1266, "lr": 6.332850511008687e-05}
{"train_loss": 0.17419658601284027, "global_step": 112703, "epoch": 1266, "lr": 6.332794635004791e-05}
{"train_loss": 0.19427649676799774, "global_step": 112704, "epoch": 1266, "lr": 6.332738758821716e-05}
{"train_loss": 0.15266719460487366, "global_step": 112705, "epoch": 1266, "lr": 6.332682882459471e-05}
{"train_loss": 0.1542932689189911, "global_step": 112706, "epoch": 1266, "lr": 6.332627005918061e-05}
{"train_loss": 0.14105625450611115, "global_step": 112707, "epoch": 1266, "lr": 6.332571129197498e-05}
{"train_loss": 0.10379774868488312, "global_step": 112708, "epoch": 1266, "lr": 6.332515252297786e-05}
{"train_loss": 0.24161480367183685, "global_step": 112709, "epoch": 1266, "lr": 6.332459375218935e-05}
{"train_loss": 0.10924328118562698, "global_step": 112710, "epoch": 1266, "lr": 6.33240349796095e-05}
{"train_loss": 0.10917145013809204, "global_step": 112711, "epoch": 1266, "lr": 6.332347620523841e-05}
{"train_loss": 0.05190879851579666, "global_step": 112712, "epoch": 1266, "lr": 6.332291742907612e-05}
{"train_loss": 0.1487983614206314, "global_step": 112713, "epoch": 1266, "lr": 6.332235865112274e-05}
{"train_loss": 0.14912641048431396, "global_step": 112714, "epoch": 1266, "lr": 6.332179987137833e-05}
{"train_loss": 0.16788339614868164, "global_step": 112715, "epoch": 1266, "lr": 6.332124108984297e-05}
{"train_loss": 0.15685325860977173, "global_step": 112716, "epoch": 1266, "lr": 6.332068230651672e-05}
{"train_loss": 0.1590811014175415, "global_step": 112717, "epoch": 1266, "lr": 6.332012352139966e-05}
{"train_loss": 0.16627396643161774, "global_step": 112718, "epoch": 1266, "lr": 6.331956473449189e-05}
{"train_loss": 0.099631167948246, "global_step": 112719, "epoch": 1266, "lr": 6.331900594579345e-05}
{"train_loss": 0.16852658987045288, "global_step": 112720, "epoch": 1266, "lr": 6.331844715530445e-05}
{"train_loss": 0.1333322674036026, "global_step": 112721, "epoch": 1266, "lr": 6.331788836302494e-05}
{"train_loss": 0.1165992021560669, "global_step": 112722, "epoch": 1266, "lr": 6.3317329568955e-05}
{"train_loss": 0.16424722969532013, "global_step": 112723, "epoch": 1266, "lr": 6.33167707730947e-05}
{"train_loss": 0.138885959982872, "global_step": 112724, "epoch": 1266, "lr": 6.331621197544413e-05}
{"train_loss": 0.1256653219461441, "global_step": 112725, "epoch": 1266, "lr": 6.331565317600336e-05}
{"train_loss": 0.13170559704303741, "global_step": 112726, "epoch": 1266, "lr": 6.331509437477245e-05}
{"train_loss": 0.11227957904338837, "global_step": 112727, "epoch": 1266, "lr": 6.33145355717515e-05}
{"train_loss": 0.14436933398246765, "global_step": 112728, "epoch": 1266, "lr": 6.331397676694057e-05}
{"train_loss": 0.12615998089313507, "global_step": 112729, "epoch": 1266, "lr": 6.331341796033974e-05}
{"train_loss": 0.13618719577789307, "global_step": 112730, "epoch": 1266, "lr": 6.33128591519491e-05}
{"train_loss": 0.11042470484972, "global_step": 112731, "epoch": 1266, "lr": 6.331230034176869e-05}
{"train_loss": 0.0919971838593483, "global_step": 112732, "epoch": 1266, "lr": 6.33117415297986e-05}
{"train_loss": 0.15425491333007812, "global_step": 112733, "epoch": 1266, "lr": 6.331118271603892e-05}
{"train_loss": 0.15798261761665344, "global_step": 112734, "epoch": 1266, "lr": 6.33106239004897e-05}
{"train_loss": 0.17449398338794708, "global_step": 112735, "epoch": 1266, "lr": 6.331006508315104e-05}
{"train_loss": 0.10953699052333832, "global_step": 112736, "epoch": 1266, "lr": 6.3309506264023e-05}
{"train_loss": 0.19107483327388763, "global_step": 112737, "epoch": 1266, "lr": 6.330894744310566e-05}
{"train_loss": 0.1554233431816101, "global_step": 112738, "epoch": 1266, "lr": 6.33083886203991e-05}
{"train_loss": 0.1601911187171936, "global_step": 112739, "epoch": 1266, "lr": 6.330782979590337e-05}
{"train_loss": 0.14946484565734863, "global_step": 112740, "epoch": 1266, "lr": 6.330727096961858e-05}
{"train_loss": 0.11776790022850037, "global_step": 112741, "epoch": 1266, "lr": 6.330671214154478e-05}
{"train_loss": 0.12660357356071472, "global_step": 112742, "epoch": 1266, "lr": 6.330615331168205e-05}
{"train_loss": 0.08650198578834534, "global_step": 112743, "epoch": 1266, "lr": 6.330559448003049e-05}
{"train_loss": 0.08262982964515686, "global_step": 112744, "epoch": 1266, "lr": 6.330503564659015e-05}
{"train_loss": 0.08529815822839737, "global_step": 112745, "epoch": 1266, "lr": 6.330447681136111e-05}
{"train_loss": 0.2510213255882263, "global_step": 112746, "epoch": 1266, "lr": 6.330391797434343e-05}
{"train_loss": 0.27083635330200195, "global_step": 112747, "epoch": 1266, "lr": 6.330335913553723e-05}
{"train_loss": 0.09557249397039413, "global_step": 112748, "epoch": 1266, "lr": 6.330280029494252e-05}
{"train_loss": 0.12245842814445496, "global_step": 112749, "epoch": 1266, "lr": 6.330224145255943e-05}
{"train_loss": 0.13998501002788544, "global_step": 112750, "epoch": 1266, "lr": 6.330168260838802e-05}
{"train_loss": 0.17287710309028625, "global_step": 112751, "epoch": 1266, "lr": 6.330112376242836e-05}
{"train_loss": 0.14171145856380463, "global_step": 112752, "epoch": 1266, "lr": 6.330056491468053e-05}
{"train_loss": 0.08598826825618744, "global_step": 112753, "epoch": 1266, "lr": 6.33000060651446e-05}
{"train_loss": 0.1315479576587677, "global_step": 112754, "epoch": 1266, "lr": 6.329944721382061e-05}
{"train_loss": 0.138461634516716, "global_step": 112755, "epoch": 1266, "lr": 6.32988883607087e-05}
{"train_loss": 0.1029198095202446, "global_step": 112756, "epoch": 1266, "lr": 6.329832950580893e-05}
{"train_loss": 0.12206847220659256, "global_step": 112757, "epoch": 1266, "lr": 6.329777064912134e-05}
{"train_loss": 0.17717088758945465, "global_step": 112758, "epoch": 1266, "lr": 6.329721179064604e-05}
{"train_loss": 0.09040318429470062, "global_step": 112759, "epoch": 1266, "lr": 6.32966529303831e-05}
{"train_loss": 0.25476208329200745, "global_step": 112760, "epoch": 1266, "lr": 6.329609406833255e-05}
{"train_loss": 0.2332700788974762, "global_step": 112761, "epoch": 1266, "lr": 6.329553520449453e-05}
{"train_loss": 0.14492521744765594, "global_step": 112762, "epoch": 1266, "lr": 6.329497633886908e-05, "val_loss": 5.003910541534424}
{"train_loss": 0.16562215983867645, "global_step": 112763, "epoch": 1267, "lr": 6.329441747145628e-05}
{"train_loss": 0.1175975650548935, "global_step": 112764, "epoch": 1267, "lr": 6.329385860225622e-05}
{"train_loss": 0.10278144478797913, "global_step": 112765, "epoch": 1267, "lr": 6.329329973126896e-05}
{"train_loss": 0.1576707810163498, "global_step": 112766, "epoch": 1267, "lr": 6.329274085849456e-05}
{"train_loss": 0.15254442393779755, "global_step": 112767, "epoch": 1267, "lr": 6.329218198393313e-05}
{"train_loss": 0.20652705430984497, "global_step": 112768, "epoch": 1267, "lr": 6.329162310758474e-05}
{"train_loss": 0.14508599042892456, "global_step": 112769, "epoch": 1267, "lr": 6.329106422944944e-05}
{"train_loss": 0.11872966587543488, "global_step": 112770, "epoch": 1267, "lr": 6.329050534952732e-05}
{"train_loss": 0.1193215548992157, "global_step": 112771, "epoch": 1267, "lr": 6.328994646781845e-05}
{"train_loss": 0.1520790308713913, "global_step": 112772, "epoch": 1267, "lr": 6.32893875843229e-05}
{"train_loss": 0.1796974092721939, "global_step": 112773, "epoch": 1267, "lr": 6.328882869904077e-05}
{"train_loss": 0.1832246035337448, "global_step": 112774, "epoch": 1267, "lr": 6.328826981197212e-05}
{"train_loss": 0.17028796672821045, "global_step": 112775, "epoch": 1267, "lr": 6.3287710923117e-05}
{"train_loss": 0.1303594410419464, "global_step": 112776, "epoch": 1267, "lr": 6.328715203247552e-05}
{"train_loss": 0.13191711902618408, "global_step": 112777, "epoch": 1267, "lr": 6.328659314004777e-05}
{"train_loss": 0.16553620994091034, "global_step": 112778, "epoch": 1267, "lr": 6.328603424583375e-05}
{"train_loss": 0.13341711461544037, "global_step": 112779, "epoch": 1267, "lr": 6.328547534983362e-05}
{"train_loss": 0.14687885344028473, "global_step": 112780, "epoch": 1267, "lr": 6.328491645204741e-05}
{"train_loss": 0.08530909568071365, "global_step": 112781, "epoch": 1267, "lr": 6.328435755247522e-05}
{"train_loss": 0.18706545233726501, "global_step": 112782, "epoch": 1267, "lr": 6.328379865111709e-05}
{"train_loss": 0.15530167520046234, "global_step": 112783, "epoch": 1267, "lr": 6.328323974797313e-05}
{"train_loss": 0.1768537014722824, "global_step": 112784, "epoch": 1267, "lr": 6.32826808430434e-05}
{"train_loss": 0.11766248196363449, "global_step": 112785, "epoch": 1267, "lr": 6.328212193632796e-05}
{"train_loss": 0.1676153987646103, "global_step": 112786, "epoch": 1267, "lr": 6.328156302782691e-05}
{"train_loss": 0.16496112942695618, "global_step": 112787, "epoch": 1267, "lr": 6.328100411754033e-05}
{"train_loss": 0.1510465443134308, "global_step": 112788, "epoch": 1267, "lr": 6.328044520546827e-05}
{"train_loss": 0.1237887516617775, "global_step": 112789, "epoch": 1267, "lr": 6.327988629161081e-05}
{"train_loss": 0.2473927140235901, "global_step": 112790, "epoch": 1267, "lr": 6.327932737596804e-05}
{"train_loss": 0.20210938155651093, "global_step": 112791, "epoch": 1267, "lr": 6.327876845854002e-05}
{"train_loss": 0.1027580127120018, "global_step": 112792, "epoch": 1267, "lr": 6.327820953932682e-05}
{"train_loss": 0.14021411538124084, "global_step": 112793, "epoch": 1267, "lr": 6.327765061832854e-05}
{"train_loss": 0.11243472248315811, "global_step": 112794, "epoch": 1267, "lr": 6.327709169554524e-05}
{"train_loss": 0.12772005796432495, "global_step": 112795, "epoch": 1267, "lr": 6.327653277097699e-05}
{"train_loss": 0.16452784836292267, "global_step": 112796, "epoch": 1267, "lr": 6.327597384462389e-05}
{"train_loss": 0.13270123302936554, "global_step": 112797, "epoch": 1267, "lr": 6.327541491648597e-05}
{"train_loss": 0.1521703153848648, "global_step": 112798, "epoch": 1267, "lr": 6.327485598656335e-05}
{"train_loss": 0.17438702285289764, "global_step": 112799, "epoch": 1267, "lr": 6.327429705485609e-05}
{"train_loss": 0.12450294941663742, "global_step": 112800, "epoch": 1267, "lr": 6.327373812136424e-05}
{"train_loss": 0.16500407457351685, "global_step": 112801, "epoch": 1267, "lr": 6.327317918608793e-05}
{"train_loss": 0.1104012206196785, "global_step": 112802, "epoch": 1267, "lr": 6.327262024902719e-05}
{"train_loss": 0.16415630280971527, "global_step": 112803, "epoch": 1267, "lr": 6.32720613101821e-05}
{"train_loss": 0.09575363248586655, "global_step": 112804, "epoch": 1267, "lr": 6.327150236955275e-05}
{"train_loss": 0.1832658052444458, "global_step": 112805, "epoch": 1267, "lr": 6.327094342713921e-05}
{"train_loss": 0.09796252846717834, "global_step": 112806, "epoch": 1267, "lr": 6.327038448294155e-05}
{"train_loss": 0.14828883111476898, "global_step": 112807, "epoch": 1267, "lr": 6.326982553695984e-05}
{"train_loss": 0.10343705117702484, "global_step": 112808, "epoch": 1267, "lr": 6.326926658919418e-05}
{"train_loss": 0.1606488823890686, "global_step": 112809, "epoch": 1267, "lr": 6.326870763964463e-05}
{"train_loss": 0.24146050214767456, "global_step": 112810, "epoch": 1267, "lr": 6.326814868831126e-05}
{"train_loss": 0.12029349058866501, "global_step": 112811, "epoch": 1267, "lr": 6.326758973519414e-05}
{"train_loss": 0.199252650141716, "global_step": 112812, "epoch": 1267, "lr": 6.326703078029337e-05}
{"train_loss": 0.1082543283700943, "global_step": 112813, "epoch": 1267, "lr": 6.326647182360899e-05}
{"train_loss": 0.2013172209262848, "global_step": 112814, "epoch": 1267, "lr": 6.32659128651411e-05}
{"train_loss": 0.18827523291110992, "global_step": 112815, "epoch": 1267, "lr": 6.326535390488977e-05}
{"train_loss": 0.16766570508480072, "global_step": 112816, "epoch": 1267, "lr": 6.326479494285508e-05}
{"train_loss": 0.13060306012630463, "global_step": 112817, "epoch": 1267, "lr": 6.32642359790371e-05}
{"train_loss": 0.16417202353477478, "global_step": 112818, "epoch": 1267, "lr": 6.32636770134359e-05}
{"train_loss": 0.11998232454061508, "global_step": 112819, "epoch": 1267, "lr": 6.326311804605157e-05}
{"train_loss": 0.1684054434299469, "global_step": 112820, "epoch": 1267, "lr": 6.326255907688416e-05}
{"train_loss": 0.12809239327907562, "global_step": 112821, "epoch": 1267, "lr": 6.326200010593377e-05}
{"train_loss": 0.16001027822494507, "global_step": 112822, "epoch": 1267, "lr": 6.326144113320047e-05}
{"train_loss": 0.1593860685825348, "global_step": 112823, "epoch": 1267, "lr": 6.326088215868432e-05}
{"train_loss": 0.09863914549350739, "global_step": 112824, "epoch": 1267, "lr": 6.326032318238542e-05}
{"train_loss": 0.09598098695278168, "global_step": 112825, "epoch": 1267, "lr": 6.325976420430381e-05}
{"train_loss": 0.17477773129940033, "global_step": 112826, "epoch": 1267, "lr": 6.325920522443961e-05}
{"train_loss": 0.16242417693138123, "global_step": 112827, "epoch": 1267, "lr": 6.325864624279288e-05}
{"train_loss": 0.16906693577766418, "global_step": 112828, "epoch": 1267, "lr": 6.325808725936366e-05}
{"train_loss": 0.16110867261886597, "global_step": 112829, "epoch": 1267, "lr": 6.325752827415205e-05}
{"train_loss": 0.14501264691352844, "global_step": 112830, "epoch": 1267, "lr": 6.325696928715814e-05}
{"train_loss": 0.1714840829372406, "global_step": 112831, "epoch": 1267, "lr": 6.3256410298382e-05}
{"train_loss": 0.12448113411664963, "global_step": 112832, "epoch": 1267, "lr": 6.325585130782368e-05}
{"train_loss": 0.19200722873210907, "global_step": 112833, "epoch": 1267, "lr": 6.325529231548326e-05}
{"train_loss": 0.10765822976827621, "global_step": 112834, "epoch": 1267, "lr": 6.325473332136084e-05}
{"train_loss": 0.11794054508209229, "global_step": 112835, "epoch": 1267, "lr": 6.325417432545651e-05}
{"train_loss": 0.15709188580513, "global_step": 112836, "epoch": 1267, "lr": 6.325361532777029e-05}
{"train_loss": 0.08811372518539429, "global_step": 112837, "epoch": 1267, "lr": 6.325305632830228e-05}
{"train_loss": 0.23350422084331512, "global_step": 112838, "epoch": 1267, "lr": 6.325249732705257e-05}
{"train_loss": 0.3483376204967499, "global_step": 112839, "epoch": 1267, "lr": 6.325193832402122e-05}
{"train_loss": 0.18090160191059113, "global_step": 112840, "epoch": 1267, "lr": 6.325137931920831e-05}
{"train_loss": 0.11015774309635162, "global_step": 112841, "epoch": 1267, "lr": 6.325082031261392e-05}
{"train_loss": 0.15485256910324097, "global_step": 112842, "epoch": 1267, "lr": 6.32502613042381e-05}
{"train_loss": 0.13507269322872162, "global_step": 112843, "epoch": 1267, "lr": 6.324970229408095e-05}
{"train_loss": 0.15433527529239655, "global_step": 112844, "epoch": 1267, "lr": 6.324914328214256e-05}
{"train_loss": 0.14287018775939941, "global_step": 112845, "epoch": 1267, "lr": 6.324858426842296e-05}
{"train_loss": 0.10999514907598495, "global_step": 112846, "epoch": 1267, "lr": 6.324802525292227e-05}
{"train_loss": 0.2775753438472748, "global_step": 112847, "epoch": 1267, "lr": 6.324746623564053e-05}
{"train_loss": 0.14128953218460083, "global_step": 112848, "epoch": 1267, "lr": 6.324690721657783e-05}
{"train_loss": 0.10496029257774353, "global_step": 112849, "epoch": 1267, "lr": 6.324634819573425e-05}
{"train_loss": 0.11977515369653702, "global_step": 112850, "epoch": 1267, "lr": 6.324578917310986e-05}
{"train_loss": 0.15121622207794297, "global_step": 112851, "epoch": 1267, "lr": 6.324523014870471e-05, "val_loss": 4.8448309898376465}
{"train_loss": 0.12322324514389038, "global_step": 112852, "epoch": 1268, "lr": 6.324467112251893e-05}
{"train_loss": 0.2466421276330948, "global_step": 112853, "epoch": 1268, "lr": 6.324411209455256e-05}
{"train_loss": 0.1352566033601761, "global_step": 112854, "epoch": 1268, "lr": 6.324355306480567e-05}
{"train_loss": 0.12249411642551422, "global_step": 112855, "epoch": 1268, "lr": 6.324299403327835e-05}
{"train_loss": 0.17011360824108124, "global_step": 112856, "epoch": 1268, "lr": 6.324243499997068e-05}
{"train_loss": 0.11789701879024506, "global_step": 112857, "epoch": 1268, "lr": 6.324187596488271e-05}
{"train_loss": 0.17167046666145325, "global_step": 112858, "epoch": 1268, "lr": 6.324131692801455e-05}
{"train_loss": 0.21177750825881958, "global_step": 112859, "epoch": 1268, "lr": 6.324075788936625e-05}
{"train_loss": 0.2220146358013153, "global_step": 112860, "epoch": 1268, "lr": 6.324019884893788e-05}
{"train_loss": 0.12329272925853729, "global_step": 112861, "epoch": 1268, "lr": 6.323963980672955e-05}
{"train_loss": 0.16323424875736237, "global_step": 112862, "epoch": 1268, "lr": 6.323908076274128e-05}
{"train_loss": 0.17562377452850342, "global_step": 112863, "epoch": 1268, "lr": 6.323852171697319e-05}
{"train_loss": 0.1766670197248459, "global_step": 112864, "epoch": 1268, "lr": 6.323796266942537e-05}
{"train_loss": 0.1860613077878952, "global_step": 112865, "epoch": 1268, "lr": 6.323740362009784e-05}
{"train_loss": 0.12840135395526886, "global_step": 112866, "epoch": 1268, "lr": 6.32368445689907e-05}
{"train_loss": 0.14496414363384247, "global_step": 112867, "epoch": 1268, "lr": 6.323628551610405e-05}
{"train_loss": 0.16915836930274963, "global_step": 112868, "epoch": 1268, "lr": 6.323572646143792e-05}
{"train_loss": 0.12373277544975281, "global_step": 112869, "epoch": 1268, "lr": 6.32351674049924e-05}
{"train_loss": 0.10799930989742279, "global_step": 112870, "epoch": 1268, "lr": 6.32346083467676e-05}
{"train_loss": 0.16677607595920563, "global_step": 112871, "epoch": 1268, "lr": 6.323404928676355e-05}
{"train_loss": 0.15244828164577484, "global_step": 112872, "epoch": 1268, "lr": 6.323349022498036e-05}
{"train_loss": 0.18317455053329468, "global_step": 112873, "epoch": 1268, "lr": 6.323293116141807e-05}
{"train_loss": 0.16358497738838196, "global_step": 112874, "epoch": 1268, "lr": 6.323237209607678e-05}
{"train_loss": 0.09868939965963364, "global_step": 112875, "epoch": 1268, "lr": 6.323181302895656e-05}
{"train_loss": 0.11033379286527634, "global_step": 112876, "epoch": 1268, "lr": 6.323125396005748e-05}
{"train_loss": 0.1862391084432602, "global_step": 112877, "epoch": 1268, "lr": 6.323069488937963e-05}
{"train_loss": 0.13735726475715637, "global_step": 112878, "epoch": 1268, "lr": 6.323013581692306e-05}
{"train_loss": 0.10302139073610306, "global_step": 112879, "epoch": 1268, "lr": 6.322957674268787e-05}
{"train_loss": 0.1657656878232956, "global_step": 112880, "epoch": 1268, "lr": 6.322901766667411e-05}
{"train_loss": 0.14718811213970184, "global_step": 112881, "epoch": 1268, "lr": 6.322845858888189e-05}
{"train_loss": 0.12310782074928284, "global_step": 112882, "epoch": 1268, "lr": 6.322789950931126e-05}
{"train_loss": 0.12700998783111572, "global_step": 112883, "epoch": 1268, "lr": 6.32273404279623e-05}
{"train_loss": 0.19614073634147644, "global_step": 112884, "epoch": 1268, "lr": 6.322678134483509e-05}
{"train_loss": 0.11865445226430893, "global_step": 112885, "epoch": 1268, "lr": 6.322622225992969e-05}
{"train_loss": 0.17990706861019135, "global_step": 112886, "epoch": 1268, "lr": 6.322566317324618e-05}
{"train_loss": 0.1589657962322235, "global_step": 112887, "epoch": 1268, "lr": 6.322510408478465e-05}
{"train_loss": 0.12784966826438904, "global_step": 112888, "epoch": 1268, "lr": 6.322454499454517e-05}
{"train_loss": 0.1122145876288414, "global_step": 112889, "epoch": 1268, "lr": 6.32239859025278e-05}
{"train_loss": 0.1471412032842636, "global_step": 112890, "epoch": 1268, "lr": 6.322342680873263e-05}
{"train_loss": 0.1732340157032013, "global_step": 112891, "epoch": 1268, "lr": 6.322286771315973e-05}
{"train_loss": 0.18262679874897003, "global_step": 112892, "epoch": 1268, "lr": 6.322230861580918e-05}
{"train_loss": 0.15580850839614868, "global_step": 112893, "epoch": 1268, "lr": 6.322174951668104e-05}
{"train_loss": 0.2429436594247818, "global_step": 112894, "epoch": 1268, "lr": 6.322119041577542e-05}
{"train_loss": 0.16656015813350677, "global_step": 112895, "epoch": 1268, "lr": 6.322063131309237e-05}
{"train_loss": 0.144400492310524, "global_step": 112896, "epoch": 1268, "lr": 6.322007220863196e-05}
{"train_loss": 0.1234428808093071, "global_step": 112897, "epoch": 1268, "lr": 6.321951310239426e-05}
{"train_loss": 0.18108414113521576, "global_step": 112898, "epoch": 1268, "lr": 6.321895399437937e-05}
{"train_loss": 0.1545684039592743, "global_step": 112899, "epoch": 1268, "lr": 6.321839488458737e-05}
{"train_loss": 0.1917676031589508, "global_step": 112900, "epoch": 1268, "lr": 6.321783577301829e-05}
{"train_loss": 0.2235090136528015, "global_step": 112901, "epoch": 1268, "lr": 6.321727665967224e-05}
{"train_loss": 0.07180279493331909, "global_step": 112902, "epoch": 1268, "lr": 6.321671754454929e-05}
{"train_loss": 0.06582329422235489, "global_step": 112903, "epoch": 1268, "lr": 6.321615842764951e-05}
{"train_loss": 0.08142346143722534, "global_step": 112904, "epoch": 1268, "lr": 6.3215599308973e-05}
{"train_loss": 0.11803510040044785, "global_step": 112905, "epoch": 1268, "lr": 6.321504018851979e-05}
{"train_loss": 0.08401601016521454, "global_step": 112906, "epoch": 1268, "lr": 6.321448106628997e-05}
{"train_loss": 0.13524402678012848, "global_step": 112907, "epoch": 1268, "lr": 6.321392194228364e-05}
{"train_loss": 0.1528032124042511, "global_step": 112908, "epoch": 1268, "lr": 6.321336281650086e-05}
{"train_loss": 0.126103013753891, "global_step": 112909, "epoch": 1268, "lr": 6.32128036889417e-05}
{"train_loss": 0.10185800492763519, "global_step": 112910, "epoch": 1268, "lr": 6.321224455960625e-05}
{"train_loss": 0.08861548453569412, "global_step": 112911, "epoch": 1268, "lr": 6.321168542849456e-05}
{"train_loss": 0.14203745126724243, "global_step": 112912, "epoch": 1268, "lr": 6.321112629560673e-05}
{"train_loss": 0.10702626407146454, "global_step": 112913, "epoch": 1268, "lr": 6.321056716094283e-05}
{"train_loss": 0.182691752910614, "global_step": 112914, "epoch": 1268, "lr": 6.321000802450292e-05}
{"train_loss": 0.1742909699678421, "global_step": 112915, "epoch": 1268, "lr": 6.320944888628708e-05}
{"train_loss": 0.20045140385627747, "global_step": 112916, "epoch": 1268, "lr": 6.32088897462954e-05}
{"train_loss": 0.1522306352853775, "global_step": 112917, "epoch": 1268, "lr": 6.320833060452795e-05}
{"train_loss": 0.17455223202705383, "global_step": 112918, "epoch": 1268, "lr": 6.320777146098478e-05}
{"train_loss": 0.14052680134773254, "global_step": 112919, "epoch": 1268, "lr": 6.3207212315666e-05}
{"train_loss": 0.12615403532981873, "global_step": 112920, "epoch": 1268, "lr": 6.320665316857168e-05}
{"train_loss": 0.1651442050933838, "global_step": 112921, "epoch": 1268, "lr": 6.320609401970187e-05}
{"train_loss": 0.15205487608909607, "global_step": 112922, "epoch": 1268, "lr": 6.320553486905668e-05}
{"train_loss": 0.09372812509536743, "global_step": 112923, "epoch": 1268, "lr": 6.320497571663616e-05}
{"train_loss": 0.14697232842445374, "global_step": 112924, "epoch": 1268, "lr": 6.32044165624404e-05}
{"train_loss": 0.1489042341709137, "global_step": 112925, "epoch": 1268, "lr": 6.320385740646944e-05}
{"train_loss": 0.14446058869361877, "global_step": 112926, "epoch": 1268, "lr": 6.32032982487234e-05}
{"train_loss": 0.18928882479667664, "global_step": 112927, "epoch": 1268, "lr": 6.320273908920235e-05}
{"train_loss": 0.11030931770801544, "global_step": 112928, "epoch": 1268, "lr": 6.320217992790634e-05}
{"train_loss": 0.10830418020486832, "global_step": 112929, "epoch": 1268, "lr": 6.320162076483544e-05}
{"train_loss": 0.12652705609798431, "global_step": 112930, "epoch": 1268, "lr": 6.320106159998978e-05}
{"train_loss": 0.12947770953178406, "global_step": 112931, "epoch": 1268, "lr": 6.320050243336937e-05}
{"train_loss": 0.17770040035247803, "global_step": 112932, "epoch": 1268, "lr": 6.319994326497431e-05}
{"train_loss": 0.19825974106788635, "global_step": 112933, "epoch": 1268, "lr": 6.31993840948047e-05}
{"train_loss": 0.09571505337953568, "global_step": 112934, "epoch": 1268, "lr": 6.319882492286057e-05}
{"train_loss": 0.18084438145160675, "global_step": 112935, "epoch": 1268, "lr": 6.319826574914205e-05}
{"train_loss": 0.20008940994739532, "global_step": 112936, "epoch": 1268, "lr": 6.319770657364916e-05}
{"train_loss": 0.17501652240753174, "global_step": 112937, "epoch": 1268, "lr": 6.3197147396382e-05}
{"train_loss": 0.11844192445278168, "global_step": 112938, "epoch": 1268, "lr": 6.319658821734066e-05}
{"train_loss": 0.1342315375804901, "global_step": 112939, "epoch": 1268, "lr": 6.319602903652519e-05}
{"train_loss": 0.14867014638828427, "global_step": 112940, "epoch": 1268, "lr": 6.319546985393567e-05, "val_loss": 5.082790851593018}
{"train_loss": 0.1947019100189209, "global_step": 112941, "epoch": 1269, "lr": 6.319491066957219e-05}
{"train_loss": 0.16399355232715607, "global_step": 112942, "epoch": 1269, "lr": 6.31943514834348e-05}
{"train_loss": 0.13516706228256226, "global_step": 112943, "epoch": 1269, "lr": 6.31937922955236e-05}
{"train_loss": 0.12966300547122955, "global_step": 112944, "epoch": 1269, "lr": 6.319323310583866e-05}
{"train_loss": 0.12164632230997086, "global_step": 112945, "epoch": 1269, "lr": 6.319267391438004e-05}
{"train_loss": 0.1721256971359253, "global_step": 112946, "epoch": 1269, "lr": 6.319211472114784e-05}
{"train_loss": 0.14308592677116394, "global_step": 112947, "epoch": 1269, "lr": 6.319155552614211e-05}
{"train_loss": 0.1318829357624054, "global_step": 112948, "epoch": 1269, "lr": 6.319099632936293e-05}
{"train_loss": 0.13056713342666626, "global_step": 112949, "epoch": 1269, "lr": 6.319043713081039e-05}
{"train_loss": 0.1413215696811676, "global_step": 112950, "epoch": 1269, "lr": 6.318987793048457e-05}
{"train_loss": 0.15513688325881958, "global_step": 112951, "epoch": 1269, "lr": 6.31893187283855e-05}
{"train_loss": 0.08892198652029037, "global_step": 112952, "epoch": 1269, "lr": 6.31887595245133e-05}
{"train_loss": 0.09807556122541428, "global_step": 112953, "epoch": 1269, "lr": 6.318820031886805e-05}
{"train_loss": 0.13372012972831726, "global_step": 112954, "epoch": 1269, "lr": 6.31876411114498e-05}
{"train_loss": 0.066081702709198, "global_step": 112955, "epoch": 1269, "lr": 6.318708190225862e-05}
{"train_loss": 0.07831110060214996, "global_step": 112956, "epoch": 1269, "lr": 6.31865226912946e-05}
{"train_loss": 0.1350574493408203, "global_step": 112957, "epoch": 1269, "lr": 6.318596347855781e-05}
{"train_loss": 0.16079860925674438, "global_step": 112958, "epoch": 1269, "lr": 6.318540426404834e-05}
{"train_loss": 0.1322653889656067, "global_step": 112959, "epoch": 1269, "lr": 6.318484504776626e-05}
{"train_loss": 0.14514081180095673, "global_step": 112960, "epoch": 1269, "lr": 6.318428582971162e-05}
{"train_loss": 0.15653817355632782, "global_step": 112961, "epoch": 1269, "lr": 6.318372660988452e-05}
{"train_loss": 0.16983363032341003, "global_step": 112962, "epoch": 1269, "lr": 6.318316738828503e-05}
{"train_loss": 0.07579819858074188, "global_step": 112963, "epoch": 1269, "lr": 6.318260816491321e-05}
{"train_loss": 0.18053020536899567, "global_step": 112964, "epoch": 1269, "lr": 6.318204893976917e-05}
{"train_loss": 0.1466062068939209, "global_step": 112965, "epoch": 1269, "lr": 6.318148971285296e-05}
{"train_loss": 0.17111481726169586, "global_step": 112966, "epoch": 1269, "lr": 6.318093048416464e-05}
{"train_loss": 0.1552063524723053, "global_step": 112967, "epoch": 1269, "lr": 6.318037125370433e-05}
{"train_loss": 0.13258224725723267, "global_step": 112968, "epoch": 1269, "lr": 6.317981202147206e-05}
{"train_loss": 0.08423864096403122, "global_step": 112969, "epoch": 1269, "lr": 6.317925278746793e-05}
{"train_loss": 0.20429879426956177, "global_step": 112970, "epoch": 1269, "lr": 6.317869355169203e-05}
{"train_loss": 0.18099960684776306, "global_step": 112971, "epoch": 1269, "lr": 6.317813431414438e-05}
{"train_loss": 0.13276569545269012, "global_step": 112972, "epoch": 1269, "lr": 6.317757507482513e-05}
{"train_loss": 0.13268119096755981, "global_step": 112973, "epoch": 1269, "lr": 6.317701583373429e-05}
{"train_loss": 0.15558822453022003, "global_step": 112974, "epoch": 1269, "lr": 6.317645659087197e-05}
{"train_loss": 0.1630009561777115, "global_step": 112975, "epoch": 1269, "lr": 6.317589734623823e-05}
{"train_loss": 0.20715229213237762, "global_step": 112976, "epoch": 1269, "lr": 6.317533809983314e-05}
{"train_loss": 0.13868741691112518, "global_step": 112977, "epoch": 1269, "lr": 6.317477885165681e-05}
{"train_loss": 0.07523594796657562, "global_step": 112978, "epoch": 1269, "lr": 6.317421960170928e-05}
{"train_loss": 0.10022246837615967, "global_step": 112979, "epoch": 1269, "lr": 6.317366034999063e-05}
{"train_loss": 0.12847304344177246, "global_step": 112980, "epoch": 1269, "lr": 6.317310109650095e-05}
{"train_loss": 0.15740326046943665, "global_step": 112981, "epoch": 1269, "lr": 6.31725418412403e-05}
{"train_loss": 0.08905824273824692, "global_step": 112982, "epoch": 1269, "lr": 6.317198258420878e-05}
{"train_loss": 0.129720076918602, "global_step": 112983, "epoch": 1269, "lr": 6.317142332540645e-05}
{"train_loss": 0.14341874420642853, "global_step": 112984, "epoch": 1269, "lr": 6.317086406483336e-05}
{"train_loss": 0.14194345474243164, "global_step": 112985, "epoch": 1269, "lr": 6.317030480248961e-05}
{"train_loss": 0.0812724232673645, "global_step": 112986, "epoch": 1269, "lr": 6.316974553837529e-05}
{"train_loss": 0.07530337572097778, "global_step": 112987, "epoch": 1269, "lr": 6.316918627249044e-05}
{"train_loss": 0.13547304272651672, "global_step": 112988, "epoch": 1269, "lr": 6.316862700483517e-05}
{"train_loss": 0.12934903800487518, "global_step": 112989, "epoch": 1269, "lr": 6.316806773540953e-05}
{"train_loss": 0.13033527135849, "global_step": 112990, "epoch": 1269, "lr": 6.31675084642136e-05}
{"train_loss": 0.17302048206329346, "global_step": 112991, "epoch": 1269, "lr": 6.316694919124746e-05}
{"train_loss": 0.08486764132976532, "global_step": 112992, "epoch": 1269, "lr": 6.316638991651119e-05}
{"train_loss": 0.17436663806438446, "global_step": 112993, "epoch": 1269, "lr": 6.316583064000485e-05}
{"train_loss": 0.11987987905740738, "global_step": 112994, "epoch": 1269, "lr": 6.316527136172853e-05}
{"train_loss": 0.17088565230369568, "global_step": 112995, "epoch": 1269, "lr": 6.316471208168231e-05}
{"train_loss": 0.17957225441932678, "global_step": 112996, "epoch": 1269, "lr": 6.316415279986625e-05}
{"train_loss": 0.13513484597206116, "global_step": 112997, "epoch": 1269, "lr": 6.316359351628043e-05}
{"train_loss": 0.1629234105348587, "global_step": 112998, "epoch": 1269, "lr": 6.316303423092492e-05}
{"train_loss": 0.08906426280736923, "global_step": 112999, "epoch": 1269, "lr": 6.31624749437998e-05}
{"train_loss": 0.1416734755039215, "global_step": 113000, "epoch": 1269, "lr": 6.316191565490514e-05}
{"train_loss": 0.1910831481218338, "global_step": 113001, "epoch": 1269, "lr": 6.316135636424103e-05}
{"train_loss": 0.13760314881801605, "global_step": 113002, "epoch": 1269, "lr": 6.316079707180754e-05}
{"train_loss": 0.18550734221935272, "global_step": 113003, "epoch": 1269, "lr": 6.316023777760473e-05}
{"train_loss": 0.11900490522384644, "global_step": 113004, "epoch": 1269, "lr": 6.315967848163269e-05}
{"train_loss": 0.19962044060230255, "global_step": 113005, "epoch": 1269, "lr": 6.31591191838915e-05}
{"train_loss": 0.18084216117858887, "global_step": 113006, "epoch": 1269, "lr": 6.315855988438121e-05}
{"train_loss": 0.15237551927566528, "global_step": 113007, "epoch": 1269, "lr": 6.315800058310193e-05}
{"train_loss": 0.14567303657531738, "global_step": 113008, "epoch": 1269, "lr": 6.315744128005371e-05}
{"train_loss": 0.10325886309146881, "global_step": 113009, "epoch": 1269, "lr": 6.315688197523664e-05}
{"train_loss": 0.109778992831707, "global_step": 113010, "epoch": 1269, "lr": 6.315632266865077e-05}
{"train_loss": 0.09894845634698868, "global_step": 113011, "epoch": 1269, "lr": 6.31557633602962e-05}
{"train_loss": 0.10788445919752121, "global_step": 113012, "epoch": 1269, "lr": 6.3155204050173e-05}
{"train_loss": 0.11911269277334213, "global_step": 113013, "epoch": 1269, "lr": 6.315464473828125e-05}
{"train_loss": 0.16831478476524353, "global_step": 113014, "epoch": 1269, "lr": 6.3154085424621e-05}
{"train_loss": 0.1607547551393509, "global_step": 113015, "epoch": 1269, "lr": 6.315352610919237e-05}
{"train_loss": 0.21777497231960297, "global_step": 113016, "epoch": 1269, "lr": 6.315296679199539e-05}
{"train_loss": 0.08312121033668518, "global_step": 113017, "epoch": 1269, "lr": 6.315240747303015e-05}
{"train_loss": 0.2151292860507965, "global_step": 113018, "epoch": 1269, "lr": 6.315184815229675e-05}
{"train_loss": 0.08663300424814224, "global_step": 113019, "epoch": 1269, "lr": 6.315128882979522e-05}
{"train_loss": 0.10315143316984177, "global_step": 113020, "epoch": 1269, "lr": 6.315072950552568e-05}
{"train_loss": 0.20533998310565948, "global_step": 113021, "epoch": 1269, "lr": 6.315017017948817e-05}
{"train_loss": 0.10601384192705154, "global_step": 113022, "epoch": 1269, "lr": 6.314961085168278e-05}
{"train_loss": 0.16956989467144012, "global_step": 113023, "epoch": 1269, "lr": 6.31490515221096e-05}
{"train_loss": 0.2020336538553238, "global_step": 113024, "epoch": 1269, "lr": 6.314849219076868e-05}
{"train_loss": 0.2344932109117508, "global_step": 113025, "epoch": 1269, "lr": 6.314793285766011e-05}
{"train_loss": 0.168698251247406, "global_step": 113026, "epoch": 1269, "lr": 6.314737352278395e-05}
{"train_loss": 0.08736000210046768, "global_step": 113027, "epoch": 1269, "lr": 6.31468141861403e-05}
{"train_loss": 0.12833085656166077, "global_step": 113028, "epoch": 1269, "lr": 6.314625484772922e-05}
{"train_loss": 0.14062937399309672, "global_step": 113029, "epoch": 1269, "lr": 6.314569550755077e-05, "val_loss": 5.174012184143066}
{"train_loss": 0.17823049426078796, "global_step": 113030, "epoch": 1270, "lr": 6.314513616560506e-05}
{"train_loss": 0.16024988889694214, "global_step": 113031, "epoch": 1270, "lr": 6.314457682189213e-05}
{"train_loss": 0.12589120864868164, "global_step": 113032, "epoch": 1270, "lr": 6.314401747641207e-05}
{"train_loss": 0.17935021221637726, "global_step": 113033, "epoch": 1270, "lr": 6.314345812916496e-05}
{"train_loss": 0.2051653265953064, "global_step": 113034, "epoch": 1270, "lr": 6.314289878015089e-05}
{"train_loss": 0.0928020030260086, "global_step": 113035, "epoch": 1270, "lr": 6.314233942936992e-05}
{"train_loss": 0.11775241792201996, "global_step": 113036, "epoch": 1270, "lr": 6.31417800768221e-05}
{"train_loss": 0.14709126949310303, "global_step": 113037, "epoch": 1270, "lr": 6.314122072250754e-05}
{"train_loss": 0.1782313585281372, "global_step": 113038, "epoch": 1270, "lr": 6.31406613664263e-05}
{"train_loss": 0.19893479347229004, "global_step": 113039, "epoch": 1270, "lr": 6.314010200857846e-05}
{"train_loss": 0.17131561040878296, "global_step": 113040, "epoch": 1270, "lr": 6.31395426489641e-05}
{"train_loss": 0.17120470106601715, "global_step": 113041, "epoch": 1270, "lr": 6.313898328758328e-05}
{"train_loss": 0.12375329434871674, "global_step": 113042, "epoch": 1270, "lr": 6.313842392443609e-05}
{"train_loss": 0.16997747123241425, "global_step": 113043, "epoch": 1270, "lr": 6.313786455952258e-05}
{"train_loss": 0.12744732201099396, "global_step": 113044, "epoch": 1270, "lr": 6.313730519284287e-05}
{"train_loss": 0.17985710501670837, "global_step": 113045, "epoch": 1270, "lr": 6.313674582439699e-05}
{"train_loss": 0.22032350301742554, "global_step": 113046, "epoch": 1270, "lr": 6.313618645418505e-05}
{"train_loss": 0.15443682670593262, "global_step": 113047, "epoch": 1270, "lr": 6.31356270822071e-05}
{"train_loss": 0.2532203793525696, "global_step": 113048, "epoch": 1270, "lr": 6.313506770846322e-05}
{"train_loss": 0.16265584528446198, "global_step": 113049, "epoch": 1270, "lr": 6.31345083329535e-05}
{"train_loss": 0.19186601042747498, "global_step": 113050, "epoch": 1270, "lr": 6.3133948955678e-05}
{"train_loss": 0.14322909712791443, "global_step": 113051, "epoch": 1270, "lr": 6.313338957663679e-05}
{"train_loss": 0.15995906293392181, "global_step": 113052, "epoch": 1270, "lr": 6.313283019582997e-05}
{"train_loss": 0.12172989547252655, "global_step": 113053, "epoch": 1270, "lr": 6.31322708132576e-05}
{"train_loss": 0.15462365746498108, "global_step": 113054, "epoch": 1270, "lr": 6.313171142891977e-05}
{"train_loss": 0.19464631378650665, "global_step": 113055, "epoch": 1270, "lr": 6.313115204281653e-05}
{"train_loss": 0.14203132688999176, "global_step": 113056, "epoch": 1270, "lr": 6.313059265494796e-05}
{"train_loss": 0.10181103646755219, "global_step": 113057, "epoch": 1270, "lr": 6.313003326531414e-05}
{"train_loss": 0.11692080646753311, "global_step": 113058, "epoch": 1270, "lr": 6.312947387391516e-05}
{"train_loss": 0.170917809009552, "global_step": 113059, "epoch": 1270, "lr": 6.312891448075108e-05}
{"train_loss": 0.06341404467821121, "global_step": 113060, "epoch": 1270, "lr": 6.312835508582197e-05}
{"train_loss": 0.1340191513299942, "global_step": 113061, "epoch": 1270, "lr": 6.312779568912793e-05}
{"train_loss": 0.07149910926818848, "global_step": 113062, "epoch": 1270, "lr": 6.312723629066899e-05}
{"train_loss": 0.14183004200458527, "global_step": 113063, "epoch": 1270, "lr": 6.312667689044527e-05}
{"train_loss": 0.16395962238311768, "global_step": 113064, "epoch": 1270, "lr": 6.312611748845682e-05}
{"train_loss": 0.19967293739318848, "global_step": 113065, "epoch": 1270, "lr": 6.312555808470372e-05}
{"train_loss": 0.15622363984584808, "global_step": 113066, "epoch": 1270, "lr": 6.312499867918607e-05}
{"train_loss": 0.12339604645967484, "global_step": 113067, "epoch": 1270, "lr": 6.312443927190391e-05}
{"train_loss": 0.06921914219856262, "global_step": 113068, "epoch": 1270, "lr": 6.312387986285734e-05}
{"train_loss": 0.15451094508171082, "global_step": 113069, "epoch": 1270, "lr": 6.31233204520464e-05}
{"train_loss": 0.21735244989395142, "global_step": 113070, "epoch": 1270, "lr": 6.31227610394712e-05}
{"train_loss": 0.12175699323415756, "global_step": 113071, "epoch": 1270, "lr": 6.312220162513182e-05}
{"train_loss": 0.14702491462230682, "global_step": 113072, "epoch": 1270, "lr": 6.31216422090283e-05}
{"train_loss": 0.1366846114397049, "global_step": 113073, "epoch": 1270, "lr": 6.312108279116076e-05}
{"train_loss": 0.1100459024310112, "global_step": 113074, "epoch": 1270, "lr": 6.312052337152923e-05}
{"train_loss": 0.17625494301319122, "global_step": 113075, "epoch": 1270, "lr": 6.311996395013382e-05}
{"train_loss": 0.17224612832069397, "global_step": 113076, "epoch": 1270, "lr": 6.311940452697457e-05}
{"train_loss": 0.1797225922346115, "global_step": 113077, "epoch": 1270, "lr": 6.31188451020516e-05}
{"train_loss": 0.09723208099603653, "global_step": 113078, "epoch": 1270, "lr": 6.311828567536493e-05}
{"train_loss": 0.13211007416248322, "global_step": 113079, "epoch": 1270, "lr": 6.311772624691468e-05}
{"train_loss": 0.21627676486968994, "global_step": 113080, "epoch": 1270, "lr": 6.311716681670093e-05}
{"train_loss": 0.2761828601360321, "global_step": 113081, "epoch": 1270, "lr": 6.311660738472372e-05}
{"train_loss": 0.13320216536521912, "global_step": 113082, "epoch": 1270, "lr": 6.311604795098313e-05}
{"train_loss": 0.12551331520080566, "global_step": 113083, "epoch": 1270, "lr": 6.311548851547927e-05}
{"train_loss": 0.18130750954151154, "global_step": 113084, "epoch": 1270, "lr": 6.31149290782122e-05}
{"train_loss": 0.14845840632915497, "global_step": 113085, "epoch": 1270, "lr": 6.311436963918195e-05}
{"train_loss": 0.06531669199466705, "global_step": 113086, "epoch": 1270, "lr": 6.311381019838866e-05}
{"train_loss": 0.13002333045005798, "global_step": 113087, "epoch": 1270, "lr": 6.311325075583236e-05}
{"train_loss": 0.15610677003860474, "global_step": 113088, "epoch": 1270, "lr": 6.311269131151316e-05}
{"train_loss": 0.08042273670434952, "global_step": 113089, "epoch": 1270, "lr": 6.311213186543112e-05}
{"train_loss": 0.1230931282043457, "global_step": 113090, "epoch": 1270, "lr": 6.311157241758631e-05}
{"train_loss": 0.12886127829551697, "global_step": 113091, "epoch": 1270, "lr": 6.31110129679788e-05}
{"train_loss": 0.10613560676574707, "global_step": 113092, "epoch": 1270, "lr": 6.311045351660868e-05}
{"train_loss": 0.14713065326213837, "global_step": 113093, "epoch": 1270, "lr": 6.310989406347604e-05}
{"train_loss": 0.18059347569942474, "global_step": 113094, "epoch": 1270, "lr": 6.310933460858091e-05}
{"train_loss": 0.15953943133354187, "global_step": 113095, "epoch": 1270, "lr": 6.31087751519234e-05}
{"train_loss": 0.08984996378421783, "global_step": 113096, "epoch": 1270, "lr": 6.310821569350356e-05}
{"train_loss": 0.13398218154907227, "global_step": 113097, "epoch": 1270, "lr": 6.310765623332149e-05}
{"train_loss": 0.11965300142765045, "global_step": 113098, "epoch": 1270, "lr": 6.310709677137727e-05}
{"train_loss": 0.22557906806468964, "global_step": 113099, "epoch": 1270, "lr": 6.310653730767093e-05}
{"train_loss": 0.12620596587657928, "global_step": 113100, "epoch": 1270, "lr": 6.31059778422026e-05}
{"train_loss": 0.16115544736385345, "global_step": 113101, "epoch": 1270, "lr": 6.310541837497232e-05}
{"train_loss": 0.21172146499156952, "global_step": 113102, "epoch": 1270, "lr": 6.310485890598018e-05}
{"train_loss": 0.14886122941970825, "global_step": 113103, "epoch": 1270, "lr": 6.310429943522625e-05}
{"train_loss": 0.16441626846790314, "global_step": 113104, "epoch": 1270, "lr": 6.310373996271061e-05}
{"train_loss": 0.09556616842746735, "global_step": 113105, "epoch": 1270, "lr": 6.310318048843332e-05}
{"train_loss": 0.11943276226520538, "global_step": 113106, "epoch": 1270, "lr": 6.310262101239447e-05}
{"train_loss": 0.13198313117027283, "global_step": 113107, "epoch": 1270, "lr": 6.310206153459413e-05}
{"train_loss": 0.20093485713005066, "global_step": 113108, "epoch": 1270, "lr": 6.310150205503239e-05}
{"train_loss": 0.1435903012752533, "global_step": 113109, "epoch": 1270, "lr": 6.310094257370931e-05}
{"train_loss": 0.19089126586914062, "global_step": 113110, "epoch": 1270, "lr": 6.310038309062497e-05}
{"train_loss": 0.2500307261943817, "global_step": 113111, "epoch": 1270, "lr": 6.309982360577944e-05}
{"train_loss": 0.1273675560951233, "global_step": 113112, "epoch": 1270, "lr": 6.309926411917281e-05}
{"train_loss": 0.07993391156196594, "global_step": 113113, "epoch": 1270, "lr": 6.309870463080513e-05}
{"train_loss": 0.17029641568660736, "global_step": 113114, "epoch": 1270, "lr": 6.30981451406765e-05}
{"train_loss": 0.15467485785484314, "global_step": 113115, "epoch": 1270, "lr": 6.309758564878696e-05}
{"train_loss": 0.14509642124176025, "global_step": 113116, "epoch": 1270, "lr": 6.309702615513663e-05}
{"train_loss": 0.12786436080932617, "global_step": 113117, "epoch": 1270, "lr": 6.309646665972554e-05}
{"train_loss": 0.15091704007949722, "global_step": 113118, "epoch": 1270, "lr": 6.309590716255382e-05, "val_loss": 5.2224884033203125, "train_action_mse_error": 20.650074005126953}
{"train_loss": 0.10304951667785645, "global_step": 113119, "epoch": 1271, "lr": 6.30953476636215e-05}
{"train_loss": 0.20385877788066864, "global_step": 113120, "epoch": 1271, "lr": 6.309478816292867e-05}
{"train_loss": 0.15792438387870789, "global_step": 113121, "epoch": 1271, "lr": 6.309422866047542e-05}
{"train_loss": 0.18687213957309723, "global_step": 113122, "epoch": 1271, "lr": 6.309366915626179e-05}
{"train_loss": 0.15409715473651886, "global_step": 113123, "epoch": 1271, "lr": 6.309310965028787e-05}
{"train_loss": 0.1155206561088562, "global_step": 113124, "epoch": 1271, "lr": 6.309255014255376e-05}
{"train_loss": 0.15401725471019745, "global_step": 113125, "epoch": 1271, "lr": 6.309199063305952e-05}
{"train_loss": 0.1982371062040329, "global_step": 113126, "epoch": 1271, "lr": 6.309143112180522e-05}
{"train_loss": 0.1768786758184433, "global_step": 113127, "epoch": 1271, "lr": 6.309087160879093e-05}
{"train_loss": 0.15086233615875244, "global_step": 113128, "epoch": 1271, "lr": 6.309031209401672e-05}
{"train_loss": 0.15689215064048767, "global_step": 113129, "epoch": 1271, "lr": 6.308975257748269e-05}
{"train_loss": 0.25904250144958496, "global_step": 113130, "epoch": 1271, "lr": 6.30891930591889e-05}
{"train_loss": 0.09572021663188934, "global_step": 113131, "epoch": 1271, "lr": 6.308863353913545e-05}
{"train_loss": 0.11103152483701706, "global_step": 113132, "epoch": 1271, "lr": 6.308807401732236e-05}
{"train_loss": 0.141134113073349, "global_step": 113133, "epoch": 1271, "lr": 6.308751449374976e-05}
{"train_loss": 0.11435453593730927, "global_step": 113134, "epoch": 1271, "lr": 6.308695496841768e-05}
{"train_loss": 0.18723151087760925, "global_step": 113135, "epoch": 1271, "lr": 6.308639544132624e-05}
{"train_loss": 0.15174049139022827, "global_step": 113136, "epoch": 1271, "lr": 6.308583591247548e-05}
{"train_loss": 0.11427339911460876, "global_step": 113137, "epoch": 1271, "lr": 6.30852763818655e-05}
{"train_loss": 0.22942744195461273, "global_step": 113138, "epoch": 1271, "lr": 6.308471684949636e-05}
{"train_loss": 0.16050292551517487, "global_step": 113139, "epoch": 1271, "lr": 6.308415731536814e-05}
{"train_loss": 0.13946236670017242, "global_step": 113140, "epoch": 1271, "lr": 6.30835977794809e-05}
{"train_loss": 0.12527987360954285, "global_step": 113141, "epoch": 1271, "lr": 6.308303824183473e-05}
{"train_loss": 0.13454288244247437, "global_step": 113142, "epoch": 1271, "lr": 6.308247870242971e-05}
{"train_loss": 0.171425461769104, "global_step": 113143, "epoch": 1271, "lr": 6.308191916126593e-05}
{"train_loss": 0.15954752266407013, "global_step": 113144, "epoch": 1271, "lr": 6.308135961834343e-05}
{"train_loss": 0.24381406605243683, "global_step": 113145, "epoch": 1271, "lr": 6.30808000736623e-05}
{"train_loss": 0.13611125946044922, "global_step": 113146, "epoch": 1271, "lr": 6.308024052722261e-05}
{"train_loss": 0.1642870008945465, "global_step": 113147, "epoch": 1271, "lr": 6.307968097902446e-05}
{"train_loss": 0.12360209971666336, "global_step": 113148, "epoch": 1271, "lr": 6.307912142906789e-05}
{"train_loss": 0.15556451678276062, "global_step": 113149, "epoch": 1271, "lr": 6.307856187735299e-05}
{"train_loss": 0.08664527535438538, "global_step": 113150, "epoch": 1271, "lr": 6.307800232387985e-05}
{"train_loss": 0.13040940463542938, "global_step": 113151, "epoch": 1271, "lr": 6.307744276864853e-05}
{"train_loss": 0.22198618948459625, "global_step": 113152, "epoch": 1271, "lr": 6.307688321165911e-05}
{"train_loss": 0.2227017730474472, "global_step": 113153, "epoch": 1271, "lr": 6.307632365291164e-05}
{"train_loss": 0.15800690650939941, "global_step": 113154, "epoch": 1271, "lr": 6.307576409240625e-05}
{"train_loss": 0.21659718453884125, "global_step": 113155, "epoch": 1271, "lr": 6.307520453014296e-05}
{"train_loss": 0.07446790486574173, "global_step": 113156, "epoch": 1271, "lr": 6.307464496612188e-05}
{"train_loss": 0.20947737991809845, "global_step": 113157, "epoch": 1271, "lr": 6.307408540034306e-05}
{"train_loss": 0.11770404875278473, "global_step": 113158, "epoch": 1271, "lr": 6.30735258328066e-05}
{"train_loss": 0.16833902895450592, "global_step": 113159, "epoch": 1271, "lr": 6.307296626351256e-05}
{"train_loss": 0.1307452768087387, "global_step": 113160, "epoch": 1271, "lr": 6.307240669246101e-05}
{"train_loss": 0.14800800383090973, "global_step": 113161, "epoch": 1271, "lr": 6.307184711965203e-05}
{"train_loss": 0.1531362384557724, "global_step": 113162, "epoch": 1271, "lr": 6.307128754508572e-05}
{"train_loss": 0.23681214451789856, "global_step": 113163, "epoch": 1271, "lr": 6.307072796876214e-05}
{"train_loss": 0.17993196845054626, "global_step": 113164, "epoch": 1271, "lr": 6.307016839068134e-05}
{"train_loss": 0.12262141704559326, "global_step": 113165, "epoch": 1271, "lr": 6.306960881084343e-05}
{"train_loss": 0.1652452051639557, "global_step": 113166, "epoch": 1271, "lr": 6.306904922924846e-05}
{"train_loss": 0.1720830202102661, "global_step": 113167, "epoch": 1271, "lr": 6.306848964589651e-05}
{"train_loss": 0.14261436462402344, "global_step": 113168, "epoch": 1271, "lr": 6.306793006078767e-05}
{"train_loss": 0.17784826457500458, "global_step": 113169, "epoch": 1271, "lr": 6.3067370473922e-05}
{"train_loss": 0.11914052814245224, "global_step": 113170, "epoch": 1271, "lr": 6.30668108852996e-05}
{"train_loss": 0.17455069720745087, "global_step": 113171, "epoch": 1271, "lr": 6.306625129492052e-05}
{"train_loss": 0.15166006982326508, "global_step": 113172, "epoch": 1271, "lr": 6.306569170278483e-05}
{"train_loss": 0.1740517020225525, "global_step": 113173, "epoch": 1271, "lr": 6.306513210889263e-05}
{"train_loss": 0.20053265988826752, "global_step": 113174, "epoch": 1271, "lr": 6.306457251324397e-05}
{"train_loss": 0.18202587962150574, "global_step": 113175, "epoch": 1271, "lr": 6.306401291583895e-05}
{"train_loss": 0.1905519962310791, "global_step": 113176, "epoch": 1271, "lr": 6.306345331667761e-05}
{"train_loss": 0.16674350202083588, "global_step": 113177, "epoch": 1271, "lr": 6.306289371576006e-05}
{"train_loss": 0.11875827610492706, "global_step": 113178, "epoch": 1271, "lr": 6.306233411308636e-05}
{"train_loss": 0.1369219571352005, "global_step": 113179, "epoch": 1271, "lr": 6.30617745086566e-05}
{"train_loss": 0.1299755722284317, "global_step": 113180, "epoch": 1271, "lr": 6.306121490247083e-05}
{"train_loss": 0.22455978393554688, "global_step": 113181, "epoch": 1271, "lr": 6.306065529452915e-05}
{"train_loss": 0.11235372722148895, "global_step": 113182, "epoch": 1271, "lr": 6.306009568483162e-05}
{"train_loss": 0.1385427713394165, "global_step": 113183, "epoch": 1271, "lr": 6.305953607337832e-05}
{"train_loss": 0.13603506982326508, "global_step": 113184, "epoch": 1271, "lr": 6.305897646016932e-05}
{"train_loss": 0.23633617162704468, "global_step": 113185, "epoch": 1271, "lr": 6.305841684520471e-05}
{"train_loss": 0.2072993665933609, "global_step": 113186, "epoch": 1271, "lr": 6.305785722848454e-05}
{"train_loss": 0.154727503657341, "global_step": 113187, "epoch": 1271, "lr": 6.305729761000892e-05}
{"train_loss": 0.16074053943157196, "global_step": 113188, "epoch": 1271, "lr": 6.305673798977789e-05}
{"train_loss": 0.14785021543502808, "global_step": 113189, "epoch": 1271, "lr": 6.305617836779157e-05}
{"train_loss": 0.13867777585983276, "global_step": 113190, "epoch": 1271, "lr": 6.305561874404997e-05}
{"train_loss": 0.2135792374610901, "global_step": 113191, "epoch": 1271, "lr": 6.30550591185532e-05}
{"train_loss": 0.1296052187681198, "global_step": 113192, "epoch": 1271, "lr": 6.305449949130136e-05}
{"train_loss": 0.1274423897266388, "global_step": 113193, "epoch": 1271, "lr": 6.305393986229448e-05}
{"train_loss": 0.15209290385246277, "global_step": 113194, "epoch": 1271, "lr": 6.305338023153268e-05}
{"train_loss": 0.26645028591156006, "global_step": 113195, "epoch": 1271, "lr": 6.305282059901599e-05}
{"train_loss": 0.12731163203716278, "global_step": 113196, "epoch": 1271, "lr": 6.305226096474452e-05}
{"train_loss": 0.1972687691450119, "global_step": 113197, "epoch": 1271, "lr": 6.305170132871833e-05}
{"train_loss": 0.14774876832962036, "global_step": 113198, "epoch": 1271, "lr": 6.305114169093748e-05}
{"train_loss": 0.10309858620166779, "global_step": 113199, "epoch": 1271, "lr": 6.305058205140209e-05}
{"train_loss": 0.11983291059732437, "global_step": 113200, "epoch": 1271, "lr": 6.305002241011219e-05}
{"train_loss": 0.18040254712104797, "global_step": 113201, "epoch": 1271, "lr": 6.30494627670679e-05}
{"train_loss": 0.16789206862449646, "global_step": 113202, "epoch": 1271, "lr": 6.304890312226924e-05}
{"train_loss": 0.15233753621578217, "global_step": 113203, "epoch": 1271, "lr": 6.304834347571633e-05}
{"train_loss": 0.187274768948555, "global_step": 113204, "epoch": 1271, "lr": 6.304778382740922e-05}
{"train_loss": 0.09574903547763824, "global_step": 113205, "epoch": 1271, "lr": 6.304722417734801e-05}
{"train_loss": 0.14731349050998688, "global_step": 113206, "epoch": 1271, "lr": 6.304666452553274e-05}
{"train_loss": 0.15998715957563914, "global_step": 113207, "epoch": 1271, "lr": 6.304610487196352e-05, "val_loss": 4.973904609680176}
{"train_loss": 0.08129937201738358, "global_step": 113208, "epoch": 1272, "lr": 6.30455452166404e-05}
{"train_loss": 0.11958140879869461, "global_step": 113209, "epoch": 1272, "lr": 6.304498555956349e-05}
{"train_loss": 0.11703605949878693, "global_step": 113210, "epoch": 1272, "lr": 6.30444259007328e-05}
{"train_loss": 0.10692892223596573, "global_step": 113211, "epoch": 1272, "lr": 6.304386624014847e-05}
{"train_loss": 0.18410547077655792, "global_step": 113212, "epoch": 1272, "lr": 6.304330657781054e-05}
{"train_loss": 0.14147844910621643, "global_step": 113213, "epoch": 1272, "lr": 6.304274691371912e-05}
{"train_loss": 0.16627861559391022, "global_step": 113214, "epoch": 1272, "lr": 6.304218724787423e-05}
{"train_loss": 0.12471028417348862, "global_step": 113215, "epoch": 1272, "lr": 6.304162758027599e-05}
{"train_loss": 0.06668133288621902, "global_step": 113216, "epoch": 1272, "lr": 6.304106791092446e-05}
{"train_loss": 0.12406138330698013, "global_step": 113217, "epoch": 1272, "lr": 6.304050823981973e-05}
{"train_loss": 0.08689176291227341, "global_step": 113218, "epoch": 1272, "lr": 6.303994856696185e-05}
{"train_loss": 0.1647031605243683, "global_step": 113219, "epoch": 1272, "lr": 6.303938889235092e-05}
{"train_loss": 0.17334216833114624, "global_step": 113220, "epoch": 1272, "lr": 6.303882921598699e-05}
{"train_loss": 0.1711423695087433, "global_step": 113221, "epoch": 1272, "lr": 6.303826953787015e-05}
{"train_loss": 0.13280756771564484, "global_step": 113222, "epoch": 1272, "lr": 6.303770985800048e-05}
{"train_loss": 0.3080487847328186, "global_step": 113223, "epoch": 1272, "lr": 6.303715017637805e-05}
{"train_loss": 0.19154956936836243, "global_step": 113224, "epoch": 1272, "lr": 6.303659049300292e-05}
{"train_loss": 0.13216310739517212, "global_step": 113225, "epoch": 1272, "lr": 6.30360308078752e-05}
{"train_loss": 0.16488279402256012, "global_step": 113226, "epoch": 1272, "lr": 6.303547112099493e-05}
{"train_loss": 0.13196979463100433, "global_step": 113227, "epoch": 1272, "lr": 6.303491143236219e-05}
{"train_loss": 0.12863579392433167, "global_step": 113228, "epoch": 1272, "lr": 6.303435174197709e-05}
{"train_loss": 0.15375158190727234, "global_step": 113229, "epoch": 1272, "lr": 6.303379204983967e-05}
{"train_loss": 0.15132752060890198, "global_step": 113230, "epoch": 1272, "lr": 6.303323235595002e-05}
{"train_loss": 0.13392046093940735, "global_step": 113231, "epoch": 1272, "lr": 6.303267266030821e-05}
{"train_loss": 0.14157657325267792, "global_step": 113232, "epoch": 1272, "lr": 6.303211296291431e-05}
{"train_loss": 0.10643524676561356, "global_step": 113233, "epoch": 1272, "lr": 6.303155326376839e-05}
{"train_loss": 0.15633000433444977, "global_step": 113234, "epoch": 1272, "lr": 6.303099356287054e-05}
{"train_loss": 0.1312837153673172, "global_step": 113235, "epoch": 1272, "lr": 6.303043386022085e-05}
{"train_loss": 0.14576087892055511, "global_step": 113236, "epoch": 1272, "lr": 6.302987415581937e-05}
{"train_loss": 0.1567882001399994, "global_step": 113237, "epoch": 1272, "lr": 6.302931444966617e-05}
{"train_loss": 0.09750471264123917, "global_step": 113238, "epoch": 1272, "lr": 6.302875474176135e-05}
{"train_loss": 0.09185830503702164, "global_step": 113239, "epoch": 1272, "lr": 6.302819503210498e-05}
{"train_loss": 0.14486059546470642, "global_step": 113240, "epoch": 1272, "lr": 6.302763532069711e-05}
{"train_loss": 0.117976114153862, "global_step": 113241, "epoch": 1272, "lr": 6.302707560753784e-05}
{"train_loss": 0.1373802274465561, "global_step": 113242, "epoch": 1272, "lr": 6.302651589262724e-05}
{"train_loss": 0.17603005468845367, "global_step": 113243, "epoch": 1272, "lr": 6.302595617596538e-05}
{"train_loss": 0.14749036729335785, "global_step": 113244, "epoch": 1272, "lr": 6.302539645755235e-05}
{"train_loss": 0.12277930974960327, "global_step": 113245, "epoch": 1272, "lr": 6.302483673738821e-05}
{"train_loss": 0.15329909324645996, "global_step": 113246, "epoch": 1272, "lr": 6.302427701547303e-05}
{"train_loss": 0.137212872505188, "global_step": 113247, "epoch": 1272, "lr": 6.302371729180692e-05}
{"train_loss": 0.14783446490764618, "global_step": 113248, "epoch": 1272, "lr": 6.30231575663899e-05}
{"train_loss": 0.17476382851600647, "global_step": 113249, "epoch": 1272, "lr": 6.302259783922209e-05}
{"train_loss": 0.14814366400241852, "global_step": 113250, "epoch": 1272, "lr": 6.302203811030354e-05}
{"train_loss": 0.17551738023757935, "global_step": 113251, "epoch": 1272, "lr": 6.302147837963433e-05}
{"train_loss": 0.08379977196455002, "global_step": 113252, "epoch": 1272, "lr": 6.302091864721455e-05}
{"train_loss": 0.13041889667510986, "global_step": 113253, "epoch": 1272, "lr": 6.302035891304428e-05}
{"train_loss": 0.07924281805753708, "global_step": 113254, "epoch": 1272, "lr": 6.301979917712357e-05}
{"train_loss": 0.16610488295555115, "global_step": 113255, "epoch": 1272, "lr": 6.30192394394525e-05}
{"train_loss": 0.10709874331951141, "global_step": 113256, "epoch": 1272, "lr": 6.301867970003117e-05}
{"train_loss": 0.1507570445537567, "global_step": 113257, "epoch": 1272, "lr": 6.301811995885962e-05}
{"train_loss": 0.10075399279594421, "global_step": 113258, "epoch": 1272, "lr": 6.301756021593795e-05}
{"train_loss": 0.08820843696594238, "global_step": 113259, "epoch": 1272, "lr": 6.301700047126622e-05}
{"train_loss": 0.1771109402179718, "global_step": 113260, "epoch": 1272, "lr": 6.301644072484453e-05}
{"train_loss": 0.10917805135250092, "global_step": 113261, "epoch": 1272, "lr": 6.301588097667292e-05}
{"train_loss": 0.12280242145061493, "global_step": 113262, "epoch": 1272, "lr": 6.301532122675149e-05}
{"train_loss": 0.11996199190616608, "global_step": 113263, "epoch": 1272, "lr": 6.301476147508031e-05}
{"train_loss": 0.165963813662529, "global_step": 113264, "epoch": 1272, "lr": 6.301420172165947e-05}
{"train_loss": 0.26328861713409424, "global_step": 113265, "epoch": 1272, "lr": 6.3013641966489e-05}
{"train_loss": 0.12612615525722504, "global_step": 113266, "epoch": 1272, "lr": 6.301308220956901e-05}
{"train_loss": 0.15336015820503235, "global_step": 113267, "epoch": 1272, "lr": 6.301252245089959e-05}
{"train_loss": 0.15293021500110626, "global_step": 113268, "epoch": 1272, "lr": 6.301196269048077e-05}
{"train_loss": 0.15250298380851746, "global_step": 113269, "epoch": 1272, "lr": 6.301140292831267e-05}
{"train_loss": 0.10545636713504791, "global_step": 113270, "epoch": 1272, "lr": 6.301084316439534e-05}
{"train_loss": 0.12253665179014206, "global_step": 113271, "epoch": 1272, "lr": 6.301028339872885e-05}
{"train_loss": 0.14176644384860992, "global_step": 113272, "epoch": 1272, "lr": 6.30097236313133e-05}
{"train_loss": 0.12200098484754562, "global_step": 113273, "epoch": 1272, "lr": 6.300916386214875e-05}
{"train_loss": 0.13794907927513123, "global_step": 113274, "epoch": 1272, "lr": 6.300860409123526e-05}
{"train_loss": 0.10391002148389816, "global_step": 113275, "epoch": 1272, "lr": 6.300804431857294e-05}
{"train_loss": 0.16763702034950256, "global_step": 113276, "epoch": 1272, "lr": 6.300748454416185e-05}
{"train_loss": 0.18397259712219238, "global_step": 113277, "epoch": 1272, "lr": 6.300692476800206e-05}
{"train_loss": 0.12831172347068787, "global_step": 113278, "epoch": 1272, "lr": 6.300636499009362e-05}
{"train_loss": 0.07899511605501175, "global_step": 113279, "epoch": 1272, "lr": 6.300580521043667e-05}
{"train_loss": 0.11173325031995773, "global_step": 113280, "epoch": 1272, "lr": 6.300524542903122e-05}
{"train_loss": 0.15666760504245758, "global_step": 113281, "epoch": 1272, "lr": 6.300468564587738e-05}
{"train_loss": 0.09219024330377579, "global_step": 113282, "epoch": 1272, "lr": 6.300412586097523e-05}
{"train_loss": 0.11919105052947998, "global_step": 113283, "epoch": 1272, "lr": 6.300356607432483e-05}
{"train_loss": 0.22572310268878937, "global_step": 113284, "epoch": 1272, "lr": 6.300300628592625e-05}
{"train_loss": 0.1876859962940216, "global_step": 113285, "epoch": 1272, "lr": 6.300244649577958e-05}
{"train_loss": 0.1112336665391922, "global_step": 113286, "epoch": 1272, "lr": 6.300188670388489e-05}
{"train_loss": 0.13541167974472046, "global_step": 113287, "epoch": 1272, "lr": 6.300132691024226e-05}
{"train_loss": 0.22394827008247375, "global_step": 113288, "epoch": 1272, "lr": 6.300076711485174e-05}
{"train_loss": 0.11347062140703201, "global_step": 113289, "epoch": 1272, "lr": 6.300020731771345e-05}
{"train_loss": 0.09015380591154099, "global_step": 113290, "epoch": 1272, "lr": 6.299964751882743e-05}
{"train_loss": 0.12241709977388382, "global_step": 113291, "epoch": 1272, "lr": 6.299908771819376e-05}
{"train_loss": 0.12569889426231384, "global_step": 113292, "epoch": 1272, "lr": 6.299852791581253e-05}
{"train_loss": 0.12513859570026398, "global_step": 113293, "epoch": 1272, "lr": 6.299796811168382e-05}
{"train_loss": 0.11032424122095108, "global_step": 113294, "epoch": 1272, "lr": 6.299740830580767e-05}
{"train_loss": 0.0667206272482872, "global_step": 113295, "epoch": 1272, "lr": 6.299684849818418e-05}
{"train_loss": 0.137096656591035, "global_step": 113296, "epoch": 1272, "lr": 6.299628868881342e-05, "val_loss": 5.165168285369873}
{"train_loss": 0.08755040913820267, "global_step": 113297, "epoch": 1273, "lr": 6.299572887769547e-05}
{"train_loss": 0.14477254450321198, "global_step": 113298, "epoch": 1273, "lr": 6.29951690648304e-05}
{"train_loss": 0.2057422399520874, "global_step": 113299, "epoch": 1273, "lr": 6.29946092502183e-05}
{"train_loss": 0.2219548523426056, "global_step": 113300, "epoch": 1273, "lr": 6.299404943385923e-05}
{"train_loss": 0.15923826396465302, "global_step": 113301, "epoch": 1273, "lr": 6.299348961575327e-05}
{"train_loss": 0.14728949964046478, "global_step": 113302, "epoch": 1273, "lr": 6.299292979590048e-05}
{"train_loss": 0.1586790382862091, "global_step": 113303, "epoch": 1273, "lr": 6.299236997430096e-05}
{"train_loss": 0.08405847102403641, "global_step": 113304, "epoch": 1273, "lr": 6.299181015095477e-05}
{"train_loss": 0.15049929916858673, "global_step": 113305, "epoch": 1273, "lr": 6.299125032586199e-05}
{"train_loss": 0.12210965901613235, "global_step": 113306, "epoch": 1273, "lr": 6.299069049902268e-05}
{"train_loss": 0.1025499477982521, "global_step": 113307, "epoch": 1273, "lr": 6.299013067043696e-05}
{"train_loss": 0.1536119133234024, "global_step": 113308, "epoch": 1273, "lr": 6.298957084010486e-05}
{"train_loss": 0.14695753157138824, "global_step": 113309, "epoch": 1273, "lr": 6.298901100802647e-05}
{"train_loss": 0.16308626532554626, "global_step": 113310, "epoch": 1273, "lr": 6.298845117420185e-05}
{"train_loss": 0.2296663373708725, "global_step": 113311, "epoch": 1273, "lr": 6.298789133863112e-05}
{"train_loss": 0.1629079431295395, "global_step": 113312, "epoch": 1273, "lr": 6.298733150131431e-05}
{"train_loss": 0.12838900089263916, "global_step": 113313, "epoch": 1273, "lr": 6.298677166225152e-05}
{"train_loss": 0.1623431295156479, "global_step": 113314, "epoch": 1273, "lr": 6.298621182144281e-05}
{"train_loss": 0.12267067283391953, "global_step": 113315, "epoch": 1273, "lr": 6.298565197888826e-05}
{"train_loss": 0.21401874721050262, "global_step": 113316, "epoch": 1273, "lr": 6.298509213458796e-05}
{"train_loss": 0.1065976619720459, "global_step": 113317, "epoch": 1273, "lr": 6.298453228854197e-05}
{"train_loss": 0.1488812118768692, "global_step": 113318, "epoch": 1273, "lr": 6.298397244075035e-05}
{"train_loss": 0.1368577778339386, "global_step": 113319, "epoch": 1273, "lr": 6.29834125912132e-05}
{"train_loss": 0.1693960279226303, "global_step": 113320, "epoch": 1273, "lr": 6.29828527399306e-05}
{"train_loss": 0.125828817486763, "global_step": 113321, "epoch": 1273, "lr": 6.298229288690261e-05}
{"train_loss": 0.2043709009885788, "global_step": 113322, "epoch": 1273, "lr": 6.298173303212931e-05}
{"train_loss": 0.10019984096288681, "global_step": 113323, "epoch": 1273, "lr": 6.298117317561076e-05}
{"train_loss": 0.1579190194606781, "global_step": 113324, "epoch": 1273, "lr": 6.298061331734707e-05}
{"train_loss": 0.20240890979766846, "global_step": 113325, "epoch": 1273, "lr": 6.298005345733828e-05}
{"train_loss": 0.11693217605352402, "global_step": 113326, "epoch": 1273, "lr": 6.297949359558448e-05}
{"train_loss": 0.14039577543735504, "global_step": 113327, "epoch": 1273, "lr": 6.297893373208576e-05}
{"train_loss": 0.14315129816532135, "global_step": 113328, "epoch": 1273, "lr": 6.297837386684217e-05}
{"train_loss": 0.15843908488750458, "global_step": 113329, "epoch": 1273, "lr": 6.29778139998538e-05}
{"train_loss": 0.18342001736164093, "global_step": 113330, "epoch": 1273, "lr": 6.29772541311207e-05}
{"train_loss": 0.11910055577754974, "global_step": 113331, "epoch": 1273, "lr": 6.2976694260643e-05}
{"train_loss": 0.1986357867717743, "global_step": 113332, "epoch": 1273, "lr": 6.297613438842072e-05}
{"train_loss": 0.14431726932525635, "global_step": 113333, "epoch": 1273, "lr": 6.297557451445397e-05}
{"train_loss": 0.10951943695545197, "global_step": 113334, "epoch": 1273, "lr": 6.29750146387428e-05}
{"train_loss": 0.10299383848905563, "global_step": 113335, "epoch": 1273, "lr": 6.29744547612873e-05}
{"train_loss": 0.24452552199363708, "global_step": 113336, "epoch": 1273, "lr": 6.297389488208755e-05}
{"train_loss": 0.11075779050588608, "global_step": 113337, "epoch": 1273, "lr": 6.29733350011436e-05}
{"train_loss": 0.19337540864944458, "global_step": 113338, "epoch": 1273, "lr": 6.297277511845556e-05}
{"train_loss": 0.17942559719085693, "global_step": 113339, "epoch": 1273, "lr": 6.29722152340235e-05}
{"train_loss": 0.19973082840442657, "global_step": 113340, "epoch": 1273, "lr": 6.297165534784746e-05}
{"train_loss": 0.15843799710273743, "global_step": 113341, "epoch": 1273, "lr": 6.297109545992754e-05}
{"train_loss": 0.16023395955562592, "global_step": 113342, "epoch": 1273, "lr": 6.297053557026383e-05}
{"train_loss": 0.1299429088830948, "global_step": 113343, "epoch": 1273, "lr": 6.296997567885638e-05}
{"train_loss": 0.1668245792388916, "global_step": 113344, "epoch": 1273, "lr": 6.296941578570528e-05}
{"train_loss": 0.13100390136241913, "global_step": 113345, "epoch": 1273, "lr": 6.296885589081059e-05}
{"train_loss": 0.18318670988082886, "global_step": 113346, "epoch": 1273, "lr": 6.29682959941724e-05}
{"train_loss": 0.1443820297718048, "global_step": 113347, "epoch": 1273, "lr": 6.296773609579079e-05}
{"train_loss": 0.08301397413015366, "global_step": 113348, "epoch": 1273, "lr": 6.296717619566581e-05}
{"train_loss": 0.09894269704818726, "global_step": 113349, "epoch": 1273, "lr": 6.296661629379756e-05}
{"train_loss": 0.22436493635177612, "global_step": 113350, "epoch": 1273, "lr": 6.29660563901861e-05}
{"train_loss": 0.1614367812871933, "global_step": 113351, "epoch": 1273, "lr": 6.296549648483152e-05}
{"train_loss": 0.17524811625480652, "global_step": 113352, "epoch": 1273, "lr": 6.29649365777339e-05}
{"train_loss": 0.09434874355792999, "global_step": 113353, "epoch": 1273, "lr": 6.296437666889328e-05}
{"train_loss": 0.18008007109165192, "global_step": 113354, "epoch": 1273, "lr": 6.296381675830976e-05}
{"train_loss": 0.09505623579025269, "global_step": 113355, "epoch": 1273, "lr": 6.296325684598342e-05}
{"train_loss": 0.1615590751171112, "global_step": 113356, "epoch": 1273, "lr": 6.296269693191432e-05}
{"train_loss": 0.10427775979042053, "global_step": 113357, "epoch": 1273, "lr": 6.296213701610255e-05}
{"train_loss": 0.08007411658763885, "global_step": 113358, "epoch": 1273, "lr": 6.296157709854817e-05}
{"train_loss": 0.1357192099094391, "global_step": 113359, "epoch": 1273, "lr": 6.296101717925128e-05}
{"train_loss": 0.1354268342256546, "global_step": 113360, "epoch": 1273, "lr": 6.296045725821193e-05}
{"train_loss": 0.1483573615550995, "global_step": 113361, "epoch": 1273, "lr": 6.295989733543021e-05}
{"train_loss": 0.17527063190937042, "global_step": 113362, "epoch": 1273, "lr": 6.295933741090619e-05}
{"train_loss": 0.2417207658290863, "global_step": 113363, "epoch": 1273, "lr": 6.295877748463993e-05}
{"train_loss": 0.1401229202747345, "global_step": 113364, "epoch": 1273, "lr": 6.295821755663154e-05}
{"train_loss": 0.11744467169046402, "global_step": 113365, "epoch": 1273, "lr": 6.295765762688107e-05}
{"train_loss": 0.14616429805755615, "global_step": 113366, "epoch": 1273, "lr": 6.295709769538859e-05}
{"train_loss": 0.11168764531612396, "global_step": 113367, "epoch": 1273, "lr": 6.295653776215418e-05}
{"train_loss": 0.18036474287509918, "global_step": 113368, "epoch": 1273, "lr": 6.295597782717795e-05}
{"train_loss": 0.19122973084449768, "global_step": 113369, "epoch": 1273, "lr": 6.295541789045992e-05}
{"train_loss": 0.20959968864917755, "global_step": 113370, "epoch": 1273, "lr": 6.29548579520002e-05}
{"train_loss": 0.08088532835245132, "global_step": 113371, "epoch": 1273, "lr": 6.295429801179887e-05}
{"train_loss": 0.15813513100147247, "global_step": 113372, "epoch": 1273, "lr": 6.295373806985598e-05}
{"train_loss": 0.1736946552991867, "global_step": 113373, "epoch": 1273, "lr": 6.295317812617161e-05}
{"train_loss": 0.09130433946847916, "global_step": 113374, "epoch": 1273, "lr": 6.295261818074587e-05}
{"train_loss": 0.14558517932891846, "global_step": 113375, "epoch": 1273, "lr": 6.29520582335788e-05}
{"train_loss": 0.16688646376132965, "global_step": 113376, "epoch": 1273, "lr": 6.295149828467046e-05}
{"train_loss": 0.1869572252035141, "global_step": 113377, "epoch": 1273, "lr": 6.295093833402097e-05}
{"train_loss": 0.16513657569885254, "global_step": 113378, "epoch": 1273, "lr": 6.295037838163038e-05}
{"train_loss": 0.1623852401971817, "global_step": 113379, "epoch": 1273, "lr": 6.294981842749877e-05}
{"train_loss": 0.11957143247127533, "global_step": 113380, "epoch": 1273, "lr": 6.29492584716262e-05}
{"train_loss": 0.1475684940814972, "global_step": 113381, "epoch": 1273, "lr": 6.294869851401278e-05}
{"train_loss": 0.1736903190612793, "global_step": 113382, "epoch": 1273, "lr": 6.294813855465856e-05}
{"train_loss": 0.176726296544075, "global_step": 113383, "epoch": 1273, "lr": 6.294757859356361e-05}
{"train_loss": 0.15666313469409943, "global_step": 113384, "epoch": 1273, "lr": 6.294701863072803e-05}
{"train_loss": 0.15080885630979968, "global_step": 113385, "epoch": 1273, "lr": 6.294645866615187e-05, "val_loss": 5.047123908996582}
{"train_loss": 0.18991370499134064, "global_step": 113386, "epoch": 1274, "lr": 6.294589869983523e-05}
{"train_loss": 0.14238013327121735, "global_step": 113387, "epoch": 1274, "lr": 6.294533873177817e-05}
{"train_loss": 0.18711096048355103, "global_step": 113388, "epoch": 1274, "lr": 6.294477876198075e-05}
{"train_loss": 0.14936934411525726, "global_step": 113389, "epoch": 1274, "lr": 6.294421879044308e-05}
{"train_loss": 0.14773370325565338, "global_step": 113390, "epoch": 1274, "lr": 6.294365881716521e-05}
{"train_loss": 0.16674788296222687, "global_step": 113391, "epoch": 1274, "lr": 6.294309884214722e-05}
{"train_loss": 0.11303343623876572, "global_step": 113392, "epoch": 1274, "lr": 6.29425388653892e-05}
{"train_loss": 0.10671822726726532, "global_step": 113393, "epoch": 1274, "lr": 6.29419788868912e-05}
{"train_loss": 0.09102542698383331, "global_step": 113394, "epoch": 1274, "lr": 6.294141890665331e-05}
{"train_loss": 0.14227434992790222, "global_step": 113395, "epoch": 1274, "lr": 6.294085892467561e-05}
{"train_loss": 0.15618522465229034, "global_step": 113396, "epoch": 1274, "lr": 6.294029894095816e-05}
{"train_loss": 0.22394487261772156, "global_step": 113397, "epoch": 1274, "lr": 6.293973895550107e-05}
{"train_loss": 0.14824682474136353, "global_step": 113398, "epoch": 1274, "lr": 6.293917896830436e-05}
{"train_loss": 0.22608421742916107, "global_step": 113399, "epoch": 1274, "lr": 6.293861897936814e-05}
{"train_loss": 0.19182710349559784, "global_step": 113400, "epoch": 1274, "lr": 6.29380589886925e-05}
{"train_loss": 0.08784720301628113, "global_step": 113401, "epoch": 1274, "lr": 6.293749899627748e-05}
{"train_loss": 0.13989372551441193, "global_step": 113402, "epoch": 1274, "lr": 6.293693900212318e-05}
{"train_loss": 0.2013813704252243, "global_step": 113403, "epoch": 1274, "lr": 6.293637900622965e-05}
{"train_loss": 0.23448505997657776, "global_step": 113404, "epoch": 1274, "lr": 6.293581900859699e-05}
{"train_loss": 0.11343280225992203, "global_step": 113405, "epoch": 1274, "lr": 6.293525900922526e-05}
{"train_loss": 0.17921343445777893, "global_step": 113406, "epoch": 1274, "lr": 6.293469900811457e-05}
{"train_loss": 0.13440297544002533, "global_step": 113407, "epoch": 1274, "lr": 6.293413900526493e-05}
{"train_loss": 0.23327450454235077, "global_step": 113408, "epoch": 1274, "lr": 6.293357900067648e-05}
{"train_loss": 0.09010346233844757, "global_step": 113409, "epoch": 1274, "lr": 6.293301899434925e-05}
{"train_loss": 0.13348126411437988, "global_step": 113410, "epoch": 1274, "lr": 6.293245898628335e-05}
{"train_loss": 0.13403743505477905, "global_step": 113411, "epoch": 1274, "lr": 6.293189897647881e-05}
{"train_loss": 0.154179185628891, "global_step": 113412, "epoch": 1274, "lr": 6.293133896493578e-05}
{"train_loss": 0.1342683583498001, "global_step": 113413, "epoch": 1274, "lr": 6.293077895165425e-05}
{"train_loss": 0.22620216012001038, "global_step": 113414, "epoch": 1274, "lr": 6.293021893663433e-05}
{"train_loss": 0.1913214921951294, "global_step": 113415, "epoch": 1274, "lr": 6.292965891987612e-05}
{"train_loss": 0.20046836137771606, "global_step": 113416, "epoch": 1274, "lr": 6.292909890137969e-05}
{"train_loss": 0.1326598823070526, "global_step": 113417, "epoch": 1274, "lr": 6.292853888114509e-05}
{"train_loss": 0.1312384456396103, "global_step": 113418, "epoch": 1274, "lr": 6.292797885917239e-05}
{"train_loss": 0.10955904424190521, "global_step": 113419, "epoch": 1274, "lr": 6.292741883546168e-05}
{"train_loss": 0.143856942653656, "global_step": 113420, "epoch": 1274, "lr": 6.292685881001305e-05}
{"train_loss": 0.16353461146354675, "global_step": 113421, "epoch": 1274, "lr": 6.292629878282655e-05}
{"train_loss": 0.1714223325252533, "global_step": 113422, "epoch": 1274, "lr": 6.292573875390227e-05}
{"train_loss": 0.16707727313041687, "global_step": 113423, "epoch": 1274, "lr": 6.292517872324028e-05}
{"train_loss": 0.1570960134267807, "global_step": 113424, "epoch": 1274, "lr": 6.292461869084067e-05}
{"train_loss": 0.20062948763370514, "global_step": 113425, "epoch": 1274, "lr": 6.29240586567035e-05}
{"train_loss": 0.12389200180768967, "global_step": 113426, "epoch": 1274, "lr": 6.292349862082885e-05}
{"train_loss": 0.1409098356962204, "global_step": 113427, "epoch": 1274, "lr": 6.292293858321677e-05}
{"train_loss": 0.11202654987573624, "global_step": 113428, "epoch": 1274, "lr": 6.292237854386737e-05}
{"train_loss": 0.14552085101604462, "global_step": 113429, "epoch": 1274, "lr": 6.292181850278073e-05}
{"train_loss": 0.1364705115556717, "global_step": 113430, "epoch": 1274, "lr": 6.292125845995689e-05}
{"train_loss": 0.18615932762622833, "global_step": 113431, "epoch": 1274, "lr": 6.292069841539595e-05}
{"train_loss": 0.13986776769161224, "global_step": 113432, "epoch": 1274, "lr": 6.292013836909797e-05}
{"train_loss": 0.10046451538801193, "global_step": 113433, "epoch": 1274, "lr": 6.291957832106305e-05}
{"train_loss": 0.14791597425937653, "global_step": 113434, "epoch": 1274, "lr": 6.291901827129124e-05}
{"train_loss": 0.18588340282440186, "global_step": 113435, "epoch": 1274, "lr": 6.291845821978264e-05}
{"train_loss": 0.16444294154644012, "global_step": 113436, "epoch": 1274, "lr": 6.29178981665373e-05}
{"train_loss": 0.18698468804359436, "global_step": 113437, "epoch": 1274, "lr": 6.291733811155531e-05}
{"train_loss": 0.14152610301971436, "global_step": 113438, "epoch": 1274, "lr": 6.291677805483675e-05}
{"train_loss": 0.1552099734544754, "global_step": 113439, "epoch": 1274, "lr": 6.291621799638167e-05}
{"train_loss": 0.12182456254959106, "global_step": 113440, "epoch": 1274, "lr": 6.291565793619017e-05}
{"train_loss": 0.12289135158061981, "global_step": 113441, "epoch": 1274, "lr": 6.291509787426231e-05}
{"train_loss": 0.28124314546585083, "global_step": 113442, "epoch": 1274, "lr": 6.291453781059817e-05}
{"train_loss": 0.16163018345832825, "global_step": 113443, "epoch": 1274, "lr": 6.291397774519783e-05}
{"train_loss": 0.22567914426326752, "global_step": 113444, "epoch": 1274, "lr": 6.291341767806138e-05}
{"train_loss": 0.09004870057106018, "global_step": 113445, "epoch": 1274, "lr": 6.291285760918887e-05}
{"train_loss": 0.1859203726053238, "global_step": 113446, "epoch": 1274, "lr": 6.291229753858037e-05}
{"train_loss": 0.15563951432704926, "global_step": 113447, "epoch": 1274, "lr": 6.291173746623598e-05}
{"train_loss": 0.18732720613479614, "global_step": 113448, "epoch": 1274, "lr": 6.291117739215577e-05}
{"train_loss": 0.1509973257780075, "global_step": 113449, "epoch": 1274, "lr": 6.29106173163398e-05}
{"train_loss": 0.17239847779273987, "global_step": 113450, "epoch": 1274, "lr": 6.291005723878815e-05}
{"train_loss": 0.126103937625885, "global_step": 113451, "epoch": 1274, "lr": 6.290949715950091e-05}
{"train_loss": 0.16634219884872437, "global_step": 113452, "epoch": 1274, "lr": 6.290893707847814e-05}
{"train_loss": 0.0954897329211235, "global_step": 113453, "epoch": 1274, "lr": 6.290837699571992e-05}
{"train_loss": 0.22305184602737427, "global_step": 113454, "epoch": 1274, "lr": 6.290781691122634e-05}
{"train_loss": 0.12898282706737518, "global_step": 113455, "epoch": 1274, "lr": 6.290725682499745e-05}
{"train_loss": 0.10189628601074219, "global_step": 113456, "epoch": 1274, "lr": 6.290669673703334e-05}
{"train_loss": 0.11853812634944916, "global_step": 113457, "epoch": 1274, "lr": 6.290613664733408e-05}
{"train_loss": 0.14090164005756378, "global_step": 113458, "epoch": 1274, "lr": 6.290557655589975e-05}
{"train_loss": 0.13380062580108643, "global_step": 113459, "epoch": 1274, "lr": 6.290501646273041e-05}
{"train_loss": 0.1564154028892517, "global_step": 113460, "epoch": 1274, "lr": 6.290445636782617e-05}
{"train_loss": 0.14481107890605927, "global_step": 113461, "epoch": 1274, "lr": 6.290389627118707e-05}
{"train_loss": 0.19320650398731232, "global_step": 113462, "epoch": 1274, "lr": 6.29033361728132e-05}
{"train_loss": 0.1518871784210205, "global_step": 113463, "epoch": 1274, "lr": 6.290277607270462e-05}
{"train_loss": 0.11595948040485382, "global_step": 113464, "epoch": 1274, "lr": 6.290221597086144e-05}
{"train_loss": 0.13257892429828644, "global_step": 113465, "epoch": 1274, "lr": 6.29016558672837e-05}
{"train_loss": 0.2102198600769043, "global_step": 113466, "epoch": 1274, "lr": 6.29010957619715e-05}
{"train_loss": 0.16890132427215576, "global_step": 113467, "epoch": 1274, "lr": 6.290053565492491e-05}
{"train_loss": 0.22396373748779297, "global_step": 113468, "epoch": 1274, "lr": 6.289997554614397e-05}
{"train_loss": 0.20825085043907166, "global_step": 113469, "epoch": 1274, "lr": 6.28994154356288e-05}
{"train_loss": 0.09304554760456085, "global_step": 113470, "epoch": 1274, "lr": 6.289885532337949e-05}
{"train_loss": 0.14705127477645874, "global_step": 113471, "epoch": 1274, "lr": 6.289829520939605e-05}
{"train_loss": 0.12162281572818756, "global_step": 113472, "epoch": 1274, "lr": 6.28977350936786e-05}
{"train_loss": 0.21673128008842468, "global_step": 113473, "epoch": 1274, "lr": 6.289717497622721e-05}
{"train_loss": 0.15580740035249946, "global_step": 113474, "epoch": 1274, "lr": 6.289661485704196e-05, "val_loss": 5.2021989822387695}
{"train_loss": 0.12652121484279633, "global_step": 113475, "epoch": 1275, "lr": 6.289605473612292e-05}
{"train_loss": 0.11584127694368362, "global_step": 113476, "epoch": 1275, "lr": 6.289549461347015e-05}
{"train_loss": 0.17522291839122772, "global_step": 113477, "epoch": 1275, "lr": 6.289493448908374e-05}
{"train_loss": 0.11163528263568878, "global_step": 113478, "epoch": 1275, "lr": 6.289437436296378e-05}
{"train_loss": 0.14847862720489502, "global_step": 113479, "epoch": 1275, "lr": 6.289381423511031e-05}
{"train_loss": 0.1421741098165512, "global_step": 113480, "epoch": 1275, "lr": 6.289325410552343e-05}
{"train_loss": 0.18006986379623413, "global_step": 113481, "epoch": 1275, "lr": 6.28926939742032e-05}
{"train_loss": 0.1526835858821869, "global_step": 113482, "epoch": 1275, "lr": 6.28921338411497e-05}
{"train_loss": 0.20517510175704956, "global_step": 113483, "epoch": 1275, "lr": 6.289157370636303e-05}
{"train_loss": 0.07416415214538574, "global_step": 113484, "epoch": 1275, "lr": 6.289101356984324e-05}
{"train_loss": 0.1697089672088623, "global_step": 113485, "epoch": 1275, "lr": 6.28904534315904e-05}
{"train_loss": 0.14847879111766815, "global_step": 113486, "epoch": 1275, "lr": 6.288989329160459e-05}
{"train_loss": 0.11955016851425171, "global_step": 113487, "epoch": 1275, "lr": 6.28893331498859e-05}
{"train_loss": 0.12251654267311096, "global_step": 113488, "epoch": 1275, "lr": 6.28887730064344e-05}
{"train_loss": 0.0975949615240097, "global_step": 113489, "epoch": 1275, "lr": 6.288821286125016e-05}
{"train_loss": 0.16060501337051392, "global_step": 113490, "epoch": 1275, "lr": 6.288765271433326e-05}
{"train_loss": 0.18136416375637054, "global_step": 113491, "epoch": 1275, "lr": 6.288709256568377e-05}
{"train_loss": 0.1758640557527542, "global_step": 113492, "epoch": 1275, "lr": 6.288653241530175e-05}
{"train_loss": 0.14895446598529816, "global_step": 113493, "epoch": 1275, "lr": 6.288597226318732e-05}
{"train_loss": 0.11485079675912857, "global_step": 113494, "epoch": 1275, "lr": 6.288541210934054e-05}
{"train_loss": 0.0947406068444252, "global_step": 113495, "epoch": 1275, "lr": 6.288485195376143e-05}
{"train_loss": 0.08671271055936813, "global_step": 113496, "epoch": 1275, "lr": 6.288429179645015e-05}
{"train_loss": 0.15517185628414154, "global_step": 113497, "epoch": 1275, "lr": 6.28837316374067e-05}
{"train_loss": 0.17886807024478912, "global_step": 113498, "epoch": 1275, "lr": 6.288317147663121e-05}
{"train_loss": 0.15542489290237427, "global_step": 113499, "epoch": 1275, "lr": 6.288261131412372e-05}
{"train_loss": 0.10536385327577591, "global_step": 113500, "epoch": 1275, "lr": 6.288205114988432e-05}
{"train_loss": 0.153909832239151, "global_step": 113501, "epoch": 1275, "lr": 6.288149098391311e-05}
{"train_loss": 0.09258625656366348, "global_step": 113502, "epoch": 1275, "lr": 6.288093081621011e-05}
{"train_loss": 0.13021093606948853, "global_step": 113503, "epoch": 1275, "lr": 6.288037064677545e-05}
{"train_loss": 0.13233107328414917, "global_step": 113504, "epoch": 1275, "lr": 6.287981047560916e-05}
{"train_loss": 0.16195838153362274, "global_step": 113505, "epoch": 1275, "lr": 6.287925030271134e-05}
{"train_loss": 0.180869922041893, "global_step": 113506, "epoch": 1275, "lr": 6.287869012808207e-05}
{"train_loss": 0.19743062555789948, "global_step": 113507, "epoch": 1275, "lr": 6.287812995172142e-05}
{"train_loss": 0.14474181830883026, "global_step": 113508, "epoch": 1275, "lr": 6.287756977362946e-05}
{"train_loss": 0.12005811184644699, "global_step": 113509, "epoch": 1275, "lr": 6.287700959380626e-05}
{"train_loss": 0.16358037292957306, "global_step": 113510, "epoch": 1275, "lr": 6.28764494122519e-05}
{"train_loss": 0.13439153134822845, "global_step": 113511, "epoch": 1275, "lr": 6.287588922896647e-05}
{"train_loss": 0.20673003792762756, "global_step": 113512, "epoch": 1275, "lr": 6.287532904395004e-05}
{"train_loss": 0.14146539568901062, "global_step": 113513, "epoch": 1275, "lr": 6.287476885720267e-05}
{"train_loss": 0.18020577728748322, "global_step": 113514, "epoch": 1275, "lr": 6.287420866872444e-05}
{"train_loss": 0.20213618874549866, "global_step": 113515, "epoch": 1275, "lr": 6.287364847851544e-05}
{"train_loss": 0.10450971871614456, "global_step": 113516, "epoch": 1275, "lr": 6.287308828657573e-05}
{"train_loss": 0.08738395571708679, "global_step": 113517, "epoch": 1275, "lr": 6.287252809290539e-05}
{"train_loss": 0.15511874854564667, "global_step": 113518, "epoch": 1275, "lr": 6.28719678975045e-05}
{"train_loss": 0.13654816150665283, "global_step": 113519, "epoch": 1275, "lr": 6.287140770037312e-05}
{"train_loss": 0.13073702156543732, "global_step": 113520, "epoch": 1275, "lr": 6.287084750151134e-05}
{"train_loss": 0.13046158850193024, "global_step": 113521, "epoch": 1275, "lr": 6.287028730091923e-05}
{"train_loss": 0.12346377968788147, "global_step": 113522, "epoch": 1275, "lr": 6.286972709859686e-05}
{"train_loss": 0.1628894805908203, "global_step": 113523, "epoch": 1275, "lr": 6.286916689454432e-05}
{"train_loss": 0.1424417495727539, "global_step": 113524, "epoch": 1275, "lr": 6.286860668876168e-05}
{"train_loss": 0.10786833614110947, "global_step": 113525, "epoch": 1275, "lr": 6.286804648124901e-05}
{"train_loss": 0.1193254217505455, "global_step": 113526, "epoch": 1275, "lr": 6.286748627200638e-05}
{"train_loss": 0.12731748819351196, "global_step": 113527, "epoch": 1275, "lr": 6.286692606103389e-05}
{"train_loss": 0.14032134413719177, "global_step": 113528, "epoch": 1275, "lr": 6.286636584833159e-05}
{"train_loss": 0.21351540088653564, "global_step": 113529, "epoch": 1275, "lr": 6.286580563389956e-05}
{"train_loss": 0.12326876074075699, "global_step": 113530, "epoch": 1275, "lr": 6.286524541773788e-05}
{"train_loss": 0.11291717737913132, "global_step": 113531, "epoch": 1275, "lr": 6.286468519984664e-05}
{"train_loss": 0.10004884004592896, "global_step": 113532, "epoch": 1275, "lr": 6.286412498022589e-05}
{"train_loss": 0.16186852753162384, "global_step": 113533, "epoch": 1275, "lr": 6.286356475887572e-05}
{"train_loss": 0.16943500936031342, "global_step": 113534, "epoch": 1275, "lr": 6.28630045357962e-05}
{"train_loss": 0.1791549026966095, "global_step": 113535, "epoch": 1275, "lr": 6.28624443109874e-05}
{"train_loss": 0.12341879308223724, "global_step": 113536, "epoch": 1275, "lr": 6.28618840844494e-05}
{"train_loss": 0.16853077709674835, "global_step": 113537, "epoch": 1275, "lr": 6.286132385618228e-05}
{"train_loss": 0.1188453659415245, "global_step": 113538, "epoch": 1275, "lr": 6.28607636261861e-05}
{"train_loss": 0.0986352264881134, "global_step": 113539, "epoch": 1275, "lr": 6.286020339446096e-05}
{"train_loss": 0.06969841569662094, "global_step": 113540, "epoch": 1275, "lr": 6.28596431610069e-05}
{"train_loss": 0.13832756876945496, "global_step": 113541, "epoch": 1275, "lr": 6.285908292582405e-05}
{"train_loss": 0.09425697475671768, "global_step": 113542, "epoch": 1275, "lr": 6.285852268891243e-05}
{"train_loss": 0.18264788389205933, "global_step": 113543, "epoch": 1275, "lr": 6.285796245027216e-05}
{"train_loss": 0.10232000797986984, "global_step": 113544, "epoch": 1275, "lr": 6.285740220990327e-05}
{"train_loss": 0.12482355535030365, "global_step": 113545, "epoch": 1275, "lr": 6.285684196780588e-05}
{"train_loss": 0.13022080063819885, "global_step": 113546, "epoch": 1275, "lr": 6.285628172398002e-05}
{"train_loss": 0.15568970143795013, "global_step": 113547, "epoch": 1275, "lr": 6.285572147842581e-05}
{"train_loss": 0.07525280863046646, "global_step": 113548, "epoch": 1275, "lr": 6.28551612311433e-05}
{"train_loss": 0.22697217762470245, "global_step": 113549, "epoch": 1275, "lr": 6.285460098213257e-05}
{"train_loss": 0.16118474304676056, "global_step": 113550, "epoch": 1275, "lr": 6.28540407313937e-05}
{"train_loss": 0.13476242125034332, "global_step": 113551, "epoch": 1275, "lr": 6.285348047892675e-05}
{"train_loss": 0.28437379002571106, "global_step": 113552, "epoch": 1275, "lr": 6.285292022473182e-05}
{"train_loss": 0.10563024133443832, "global_step": 113553, "epoch": 1275, "lr": 6.285235996880897e-05}
{"train_loss": 0.27861708402633667, "global_step": 113554, "epoch": 1275, "lr": 6.285179971115827e-05}
{"train_loss": 0.1530882567167282, "global_step": 113555, "epoch": 1275, "lr": 6.285123945177981e-05}
{"train_loss": 0.1485559642314911, "global_step": 113556, "epoch": 1275, "lr": 6.285067919067364e-05}
{"train_loss": 0.17228646576404572, "global_step": 113557, "epoch": 1275, "lr": 6.285011892783988e-05}
{"train_loss": 0.18275079131126404, "global_step": 113558, "epoch": 1275, "lr": 6.284955866327856e-05}
{"train_loss": 0.21089673042297363, "global_step": 113559, "epoch": 1275, "lr": 6.284899839698976e-05}
{"train_loss": 0.14270077645778656, "global_step": 113560, "epoch": 1275, "lr": 6.284843812897358e-05}
{"train_loss": 0.20220842957496643, "global_step": 113561, "epoch": 1275, "lr": 6.284787785923008e-05}
{"train_loss": 0.13195008039474487, "global_step": 113562, "epoch": 1275, "lr": 6.284731758775934e-05}
{"train_loss": 0.1459153460317783, "global_step": 113563, "epoch": 1275, "lr": 6.284675731456146e-05, "val_loss": 5.182064056396484, "train_action_mse_error": 8.987443923950195}
{"train_loss": 0.18716377019882202, "global_step": 113564, "epoch": 1276, "lr": 6.284619703963647e-05}
{"train_loss": 0.1714673936367035, "global_step": 113565, "epoch": 1276, "lr": 6.284563676298446e-05}
{"train_loss": 0.14733730256557465, "global_step": 113566, "epoch": 1276, "lr": 6.284507648460553e-05}
{"train_loss": 0.10238523036241531, "global_step": 113567, "epoch": 1276, "lr": 6.284451620449971e-05}
{"train_loss": 0.23639804124832153, "global_step": 113568, "epoch": 1276, "lr": 6.284395592266712e-05}
{"train_loss": 0.10666748881340027, "global_step": 113569, "epoch": 1276, "lr": 6.284339563910781e-05}
{"train_loss": 0.11876866966485977, "global_step": 113570, "epoch": 1276, "lr": 6.284283535382187e-05}
{"train_loss": 0.1978650689125061, "global_step": 113571, "epoch": 1276, "lr": 6.284227506680937e-05}
{"train_loss": 0.10776305198669434, "global_step": 113572, "epoch": 1276, "lr": 6.284171477807037e-05}
{"train_loss": 0.1601220667362213, "global_step": 113573, "epoch": 1276, "lr": 6.284115448760496e-05}
{"train_loss": 0.1867784559726715, "global_step": 113574, "epoch": 1276, "lr": 6.284059419541321e-05}
{"train_loss": 0.12178349494934082, "global_step": 113575, "epoch": 1276, "lr": 6.28400339014952e-05}
{"train_loss": 0.16984084248542786, "global_step": 113576, "epoch": 1276, "lr": 6.283947360585102e-05}
{"train_loss": 0.1716652512550354, "global_step": 113577, "epoch": 1276, "lr": 6.28389133084807e-05}
{"train_loss": 0.15567202866077423, "global_step": 113578, "epoch": 1276, "lr": 6.283835300938436e-05}
{"train_loss": 0.1659848541021347, "global_step": 113579, "epoch": 1276, "lr": 6.283779270856208e-05}
{"train_loss": 0.16453370451927185, "global_step": 113580, "epoch": 1276, "lr": 6.283723240601389e-05}
{"train_loss": 0.1226106807589531, "global_step": 113581, "epoch": 1276, "lr": 6.28366721017399e-05}
{"train_loss": 0.14256219565868378, "global_step": 113582, "epoch": 1276, "lr": 6.283611179574017e-05}
{"train_loss": 0.15602222084999084, "global_step": 113583, "epoch": 1276, "lr": 6.283555148801479e-05}
{"train_loss": 0.17260690033435822, "global_step": 113584, "epoch": 1276, "lr": 6.283499117856381e-05}
{"train_loss": 0.10823589563369751, "global_step": 113585, "epoch": 1276, "lr": 6.283443086738734e-05}
{"train_loss": 0.11102811247110367, "global_step": 113586, "epoch": 1276, "lr": 6.283387055448543e-05}
{"train_loss": 0.17319531738758087, "global_step": 113587, "epoch": 1276, "lr": 6.283331023985816e-05}
{"train_loss": 0.12551811337471008, "global_step": 113588, "epoch": 1276, "lr": 6.28327499235056e-05}
{"train_loss": 0.136136993765831, "global_step": 113589, "epoch": 1276, "lr": 6.283218960542785e-05}
{"train_loss": 0.1374329775571823, "global_step": 113590, "epoch": 1276, "lr": 6.283162928562499e-05}
{"train_loss": 0.14460408687591553, "global_step": 113591, "epoch": 1276, "lr": 6.283106896409703e-05}
{"train_loss": 0.15723997354507446, "global_step": 113592, "epoch": 1276, "lr": 6.28305086408441e-05}
{"train_loss": 0.12445518374443054, "global_step": 113593, "epoch": 1276, "lr": 6.282994831586628e-05}
{"train_loss": 0.16657498478889465, "global_step": 113594, "epoch": 1276, "lr": 6.282938798916363e-05}
{"train_loss": 0.12454798817634583, "global_step": 113595, "epoch": 1276, "lr": 6.282882766073619e-05}
{"train_loss": 0.14450235664844513, "global_step": 113596, "epoch": 1276, "lr": 6.282826733058409e-05}
{"train_loss": 0.11337030678987503, "global_step": 113597, "epoch": 1276, "lr": 6.28277069987074e-05}
{"train_loss": 0.14096274971961975, "global_step": 113598, "epoch": 1276, "lr": 6.282714666510617e-05}
{"train_loss": 0.18057690560817719, "global_step": 113599, "epoch": 1276, "lr": 6.282658632978049e-05}
{"train_loss": 0.0948587954044342, "global_step": 113600, "epoch": 1276, "lr": 6.282602599273043e-05}
{"train_loss": 0.13515302538871765, "global_step": 113601, "epoch": 1276, "lr": 6.282546565395606e-05}
{"train_loss": 0.16033461689949036, "global_step": 113602, "epoch": 1276, "lr": 6.282490531345747e-05}
{"train_loss": 0.12480462342500687, "global_step": 113603, "epoch": 1276, "lr": 6.282434497123473e-05}
{"train_loss": 0.12311693280935287, "global_step": 113604, "epoch": 1276, "lr": 6.282378462728792e-05}
{"train_loss": 0.2506210207939148, "global_step": 113605, "epoch": 1276, "lr": 6.282322428161709e-05}
{"train_loss": 0.1401461660861969, "global_step": 113606, "epoch": 1276, "lr": 6.282266393422235e-05}
{"train_loss": 0.11542443931102753, "global_step": 113607, "epoch": 1276, "lr": 6.282210358510374e-05}
{"train_loss": 0.1958622932434082, "global_step": 113608, "epoch": 1276, "lr": 6.282154323426137e-05}
{"train_loss": 0.20074422657489777, "global_step": 113609, "epoch": 1276, "lr": 6.28209828816953e-05}
{"train_loss": 0.14038869738578796, "global_step": 113610, "epoch": 1276, "lr": 6.282042252740561e-05}
{"train_loss": 0.14032907783985138, "global_step": 113611, "epoch": 1276, "lr": 6.281986217139237e-05}
{"train_loss": 0.12464845180511475, "global_step": 113612, "epoch": 1276, "lr": 6.281930181365565e-05}
{"train_loss": 0.16395455598831177, "global_step": 113613, "epoch": 1276, "lr": 6.28187414541955e-05}
{"train_loss": 0.14660397171974182, "global_step": 113614, "epoch": 1276, "lr": 6.281818109301208e-05}
{"train_loss": 0.1996356099843979, "global_step": 113615, "epoch": 1276, "lr": 6.281762073010537e-05}
{"train_loss": 0.12934990227222443, "global_step": 113616, "epoch": 1276, "lr": 6.281706036547551e-05}
{"train_loss": 0.12455099821090698, "global_step": 113617, "epoch": 1276, "lr": 6.281649999912255e-05}
{"train_loss": 0.1729924976825714, "global_step": 113618, "epoch": 1276, "lr": 6.281593963104656e-05}
{"train_loss": 0.08050267398357391, "global_step": 113619, "epoch": 1276, "lr": 6.281537926124763e-05}
{"train_loss": 0.11917496472597122, "global_step": 113620, "epoch": 1276, "lr": 6.281481888972583e-05}
{"train_loss": 0.11103025078773499, "global_step": 113621, "epoch": 1276, "lr": 6.281425851648122e-05}
{"train_loss": 0.13443772494792938, "global_step": 113622, "epoch": 1276, "lr": 6.28136981415139e-05}
{"train_loss": 0.1394396275281906, "global_step": 113623, "epoch": 1276, "lr": 6.281313776482393e-05}
{"train_loss": 0.17298609018325806, "global_step": 113624, "epoch": 1276, "lr": 6.28125773864114e-05}
{"train_loss": 0.10080306231975555, "global_step": 113625, "epoch": 1276, "lr": 6.281201700627637e-05}
{"train_loss": 0.17111635208129883, "global_step": 113626, "epoch": 1276, "lr": 6.281145662441892e-05}
{"train_loss": 0.09771525859832764, "global_step": 113627, "epoch": 1276, "lr": 6.281089624083913e-05}
{"train_loss": 0.22320297360420227, "global_step": 113628, "epoch": 1276, "lr": 6.281033585553706e-05}
{"train_loss": 0.1532049924135208, "global_step": 113629, "epoch": 1276, "lr": 6.280977546851281e-05}
{"train_loss": 0.10282675921916962, "global_step": 113630, "epoch": 1276, "lr": 6.280921507976643e-05}
{"train_loss": 0.18755240738391876, "global_step": 113631, "epoch": 1276, "lr": 6.280865468929801e-05}
{"train_loss": 0.1380046308040619, "global_step": 113632, "epoch": 1276, "lr": 6.280809429710762e-05}
{"train_loss": 0.1127173900604248, "global_step": 113633, "epoch": 1276, "lr": 6.280753390319533e-05}
{"train_loss": 0.1442929208278656, "global_step": 113634, "epoch": 1276, "lr": 6.280697350756123e-05}
{"train_loss": 0.10462361574172974, "global_step": 113635, "epoch": 1276, "lr": 6.280641311020541e-05}
{"train_loss": 0.09445497393608093, "global_step": 113636, "epoch": 1276, "lr": 6.280585271112791e-05}
{"train_loss": 0.16209398210048676, "global_step": 113637, "epoch": 1276, "lr": 6.280529231032881e-05}
{"train_loss": 0.16693183779716492, "global_step": 113638, "epoch": 1276, "lr": 6.28047319078082e-05}
{"train_loss": 0.13961054384708405, "global_step": 113639, "epoch": 1276, "lr": 6.280417150356615e-05}
{"train_loss": 0.11988598108291626, "global_step": 113640, "epoch": 1276, "lr": 6.280361109760274e-05}
{"train_loss": 0.09813404828310013, "global_step": 113641, "epoch": 1276, "lr": 6.280305068991803e-05}
{"train_loss": 0.10128563642501831, "global_step": 113642, "epoch": 1276, "lr": 6.280249028051211e-05}
{"train_loss": 0.12482395023107529, "global_step": 113643, "epoch": 1276, "lr": 6.280192986938506e-05}
{"train_loss": 0.17193754017353058, "global_step": 113644, "epoch": 1276, "lr": 6.280136945653693e-05}
{"train_loss": 0.15379688143730164, "global_step": 113645, "epoch": 1276, "lr": 6.280080904196783e-05}
{"train_loss": 0.11220293492078781, "global_step": 113646, "epoch": 1276, "lr": 6.280024862567782e-05}
{"train_loss": 0.07045186311006546, "global_step": 113647, "epoch": 1276, "lr": 6.279968820766695e-05}
{"train_loss": 0.12835080921649933, "global_step": 113648, "epoch": 1276, "lr": 6.279912778793535e-05}
{"train_loss": 0.12867668271064758, "global_step": 113649, "epoch": 1276, "lr": 6.279856736648302e-05}
{"train_loss": 0.1582203358411789, "global_step": 113650, "epoch": 1276, "lr": 6.279800694331011e-05}
{"train_loss": 0.11564791947603226, "global_step": 113651, "epoch": 1276, "lr": 6.279744651841666e-05}
{"train_loss": 0.14337893341029628, "global_step": 113652, "epoch": 1276, "lr": 6.279688609180274e-05, "val_loss": 5.363185405731201}
{"train_loss": 0.11525584757328033, "global_step": 113653, "epoch": 1277, "lr": 6.279632566346844e-05}
{"train_loss": 0.10631594061851501, "global_step": 113654, "epoch": 1277, "lr": 6.279576523341382e-05}
{"train_loss": 0.08692079037427902, "global_step": 113655, "epoch": 1277, "lr": 6.279520480163898e-05}
{"train_loss": 0.10030219703912735, "global_step": 113656, "epoch": 1277, "lr": 6.279464436814398e-05}
{"train_loss": 0.13348723948001862, "global_step": 113657, "epoch": 1277, "lr": 6.27940839329289e-05}
{"train_loss": 0.1498422771692276, "global_step": 113658, "epoch": 1277, "lr": 6.27935234959938e-05}
{"train_loss": 0.12689772248268127, "global_step": 113659, "epoch": 1277, "lr": 6.279296305733876e-05}
{"train_loss": 0.1370663344860077, "global_step": 113660, "epoch": 1277, "lr": 6.279240261696387e-05}
{"train_loss": 0.11582016199827194, "global_step": 113661, "epoch": 1277, "lr": 6.27918421748692e-05}
{"train_loss": 0.11731790006160736, "global_step": 113662, "epoch": 1277, "lr": 6.279128173105483e-05}
{"train_loss": 0.17353267967700958, "global_step": 113663, "epoch": 1277, "lr": 6.279072128552081e-05}
{"train_loss": 0.15016311407089233, "global_step": 113664, "epoch": 1277, "lr": 6.279016083826725e-05}
{"train_loss": 0.12141358107328415, "global_step": 113665, "epoch": 1277, "lr": 6.27896003892942e-05}
{"train_loss": 0.12890610098838806, "global_step": 113666, "epoch": 1277, "lr": 6.278903993860174e-05}
{"train_loss": 0.15248645842075348, "global_step": 113667, "epoch": 1277, "lr": 6.278847948618997e-05}
{"train_loss": 0.12828470766544342, "global_step": 113668, "epoch": 1277, "lr": 6.278791903205892e-05}
{"train_loss": 0.14779041707515717, "global_step": 113669, "epoch": 1277, "lr": 6.27873585762087e-05}
{"train_loss": 0.17146006226539612, "global_step": 113670, "epoch": 1277, "lr": 6.278679811863938e-05}
{"train_loss": 0.21733999252319336, "global_step": 113671, "epoch": 1277, "lr": 6.278623765935104e-05}
{"train_loss": 0.14275233447551727, "global_step": 113672, "epoch": 1277, "lr": 6.278567719834372e-05}
{"train_loss": 0.13964731991291046, "global_step": 113673, "epoch": 1277, "lr": 6.278511673561753e-05}
{"train_loss": 0.21748368442058563, "global_step": 113674, "epoch": 1277, "lr": 6.278455627117254e-05}
{"train_loss": 0.14471423625946045, "global_step": 113675, "epoch": 1277, "lr": 6.278399580500883e-05}
{"train_loss": 0.10561888664960861, "global_step": 113676, "epoch": 1277, "lr": 6.278343533712647e-05}
{"train_loss": 0.13589197397232056, "global_step": 113677, "epoch": 1277, "lr": 6.278287486752552e-05}
{"train_loss": 0.1240052655339241, "global_step": 113678, "epoch": 1277, "lr": 6.278231439620606e-05}
{"train_loss": 0.1449219286441803, "global_step": 113679, "epoch": 1277, "lr": 6.278175392316819e-05}
{"train_loss": 0.038484856486320496, "global_step": 113680, "epoch": 1277, "lr": 6.278119344841197e-05}
{"train_loss": 0.14457131922245026, "global_step": 113681, "epoch": 1277, "lr": 6.278063297193747e-05}
{"train_loss": 0.10129217058420181, "global_step": 113682, "epoch": 1277, "lr": 6.278007249374476e-05}
{"train_loss": 0.0823662132024765, "global_step": 113683, "epoch": 1277, "lr": 6.277951201383394e-05}
{"train_loss": 0.06814558058977127, "global_step": 113684, "epoch": 1277, "lr": 6.277895153220507e-05}
{"train_loss": 0.08604220300912857, "global_step": 113685, "epoch": 1277, "lr": 6.277839104885821e-05}
{"train_loss": 0.14692261815071106, "global_step": 113686, "epoch": 1277, "lr": 6.277783056379347e-05}
{"train_loss": 0.08706076443195343, "global_step": 113687, "epoch": 1277, "lr": 6.27772700770109e-05}
{"train_loss": 0.1316724419593811, "global_step": 113688, "epoch": 1277, "lr": 6.277670958851057e-05}
{"train_loss": 0.15542548894882202, "global_step": 113689, "epoch": 1277, "lr": 6.277614909829257e-05}
{"train_loss": 0.17388181388378143, "global_step": 113690, "epoch": 1277, "lr": 6.2775588606357e-05}
{"train_loss": 0.24311110377311707, "global_step": 113691, "epoch": 1277, "lr": 6.277502811270388e-05}
{"train_loss": 0.1711348444223404, "global_step": 113692, "epoch": 1277, "lr": 6.277446761733332e-05}
{"train_loss": 0.13642075657844543, "global_step": 113693, "epoch": 1277, "lr": 6.277390712024538e-05}
{"train_loss": 0.11147874593734741, "global_step": 113694, "epoch": 1277, "lr": 6.277334662144016e-05}
{"train_loss": 0.16631931066513062, "global_step": 113695, "epoch": 1277, "lr": 6.27727861209177e-05}
{"train_loss": 0.11978846788406372, "global_step": 113696, "epoch": 1277, "lr": 6.277222561867812e-05}
{"train_loss": 0.20530056953430176, "global_step": 113697, "epoch": 1277, "lr": 6.277166511472146e-05}
{"train_loss": 0.15498940646648407, "global_step": 113698, "epoch": 1277, "lr": 6.27711046090478e-05}
{"train_loss": 0.1744995415210724, "global_step": 113699, "epoch": 1277, "lr": 6.277054410165721e-05}
{"train_loss": 0.1679108738899231, "global_step": 113700, "epoch": 1277, "lr": 6.27699835925498e-05}
{"train_loss": 0.1624433547258377, "global_step": 113701, "epoch": 1277, "lr": 6.27694230817256e-05}
{"train_loss": 0.15710663795471191, "global_step": 113702, "epoch": 1277, "lr": 6.276886256918472e-05}
{"train_loss": 0.1992034763097763, "global_step": 113703, "epoch": 1277, "lr": 6.276830205492722e-05}
{"train_loss": 0.13439425826072693, "global_step": 113704, "epoch": 1277, "lr": 6.276774153895317e-05}
{"train_loss": 0.1344059705734253, "global_step": 113705, "epoch": 1277, "lr": 6.276718102126267e-05}
{"train_loss": 0.1007458046078682, "global_step": 113706, "epoch": 1277, "lr": 6.276662050185575e-05}
{"train_loss": 0.10019470006227493, "global_step": 113707, "epoch": 1277, "lr": 6.276605998073254e-05}
{"train_loss": 0.12005364894866943, "global_step": 113708, "epoch": 1277, "lr": 6.276549945789307e-05}
{"train_loss": 0.13601505756378174, "global_step": 113709, "epoch": 1277, "lr": 6.276493893333745e-05}
{"train_loss": 0.17400816082954407, "global_step": 113710, "epoch": 1277, "lr": 6.276437840706573e-05}
{"train_loss": 0.1570688635110855, "global_step": 113711, "epoch": 1277, "lr": 6.2763817879078e-05}
{"train_loss": 0.09313469380140305, "global_step": 113712, "epoch": 1277, "lr": 6.276325734937432e-05}
{"train_loss": 0.11755392700433731, "global_step": 113713, "epoch": 1277, "lr": 6.27626968179548e-05}
{"train_loss": 0.12840671837329865, "global_step": 113714, "epoch": 1277, "lr": 6.276213628481946e-05}
{"train_loss": 0.16503050923347473, "global_step": 113715, "epoch": 1277, "lr": 6.276157574996843e-05}
{"train_loss": 0.13477154076099396, "global_step": 113716, "epoch": 1277, "lr": 6.276101521340175e-05}
{"train_loss": 0.1340140700340271, "global_step": 113717, "epoch": 1277, "lr": 6.276045467511951e-05}
{"train_loss": 0.1266738772392273, "global_step": 113718, "epoch": 1277, "lr": 6.275989413512178e-05}
{"train_loss": 0.14261367917060852, "global_step": 113719, "epoch": 1277, "lr": 6.275933359340865e-05}
{"train_loss": 0.06797448545694351, "global_step": 113720, "epoch": 1277, "lr": 6.275877304998017e-05}
{"train_loss": 0.14292487502098083, "global_step": 113721, "epoch": 1277, "lr": 6.275821250483643e-05}
{"train_loss": 0.2390764355659485, "global_step": 113722, "epoch": 1277, "lr": 6.27576519579775e-05}
{"train_loss": 0.22870442271232605, "global_step": 113723, "epoch": 1277, "lr": 6.275709140940347e-05}
{"train_loss": 0.11945844441652298, "global_step": 113724, "epoch": 1277, "lr": 6.275653085911439e-05}
{"train_loss": 0.13147543370723724, "global_step": 113725, "epoch": 1277, "lr": 6.275597030711035e-05}
{"train_loss": 0.15092439949512482, "global_step": 113726, "epoch": 1277, "lr": 6.275540975339145e-05}
{"train_loss": 0.1317427158355713, "global_step": 113727, "epoch": 1277, "lr": 6.275484919795772e-05}
{"train_loss": 0.1969165802001953, "global_step": 113728, "epoch": 1277, "lr": 6.275428864080925e-05}
{"train_loss": 0.18349996209144592, "global_step": 113729, "epoch": 1277, "lr": 6.275372808194613e-05}
{"train_loss": 0.11130504310131073, "global_step": 113730, "epoch": 1277, "lr": 6.275316752136844e-05}
{"train_loss": 0.21973402798175812, "global_step": 113731, "epoch": 1277, "lr": 6.275260695907623e-05}
{"train_loss": 0.1802009642124176, "global_step": 113732, "epoch": 1277, "lr": 6.275204639506959e-05}
{"train_loss": 0.1947617530822754, "global_step": 113733, "epoch": 1277, "lr": 6.275148582934859e-05}
{"train_loss": 0.15160420536994934, "global_step": 113734, "epoch": 1277, "lr": 6.27509252619133e-05}
{"train_loss": 0.13136577606201172, "global_step": 113735, "epoch": 1277, "lr": 6.275036469276382e-05}
{"train_loss": 0.14732329547405243, "global_step": 113736, "epoch": 1277, "lr": 6.27498041219002e-05}
{"train_loss": 0.18346023559570312, "global_step": 113737, "epoch": 1277, "lr": 6.274924354932252e-05}
{"train_loss": 0.17434388399124146, "global_step": 113738, "epoch": 1277, "lr": 6.274868297503088e-05}
{"train_loss": 0.08093589544296265, "global_step": 113739, "epoch": 1277, "lr": 6.274812239902531e-05}
{"train_loss": 0.12491835653781891, "global_step": 113740, "epoch": 1277, "lr": 6.274756182130592e-05}
{"train_loss": 0.14150097036964437, "global_step": 113741, "epoch": 1277, "lr": 6.274700124187279e-05, "val_loss": 5.230738639831543}
{"train_loss": 0.11496401578187943, "global_step": 113742, "epoch": 1278, "lr": 6.274644066072595e-05}
{"train_loss": 0.1491086781024933, "global_step": 113743, "epoch": 1278, "lr": 6.274588007786552e-05}
{"train_loss": 0.1215234026312828, "global_step": 113744, "epoch": 1278, "lr": 6.274531949329157e-05}
{"train_loss": 0.09990408271551132, "global_step": 113745, "epoch": 1278, "lr": 6.274475890700417e-05}
{"train_loss": 0.20622500777244568, "global_step": 113746, "epoch": 1278, "lr": 6.274419831900337e-05}
{"train_loss": 0.12924101948738098, "global_step": 113747, "epoch": 1278, "lr": 6.274363772928929e-05}
{"train_loss": 0.13859327137470245, "global_step": 113748, "epoch": 1278, "lr": 6.274307713786198e-05}
{"train_loss": 0.197963684797287, "global_step": 113749, "epoch": 1278, "lr": 6.274251654472152e-05}
{"train_loss": 0.16506485641002655, "global_step": 113750, "epoch": 1278, "lr": 6.274195594986797e-05}
{"train_loss": 0.15736472606658936, "global_step": 113751, "epoch": 1278, "lr": 6.274139535330144e-05}
{"train_loss": 0.11862664669752121, "global_step": 113752, "epoch": 1278, "lr": 6.274083475502195e-05}
{"train_loss": 0.10189052671194077, "global_step": 113753, "epoch": 1278, "lr": 6.274027415502964e-05}
{"train_loss": 0.1928916871547699, "global_step": 113754, "epoch": 1278, "lr": 6.273971355332455e-05}
{"train_loss": 0.14336846768856049, "global_step": 113755, "epoch": 1278, "lr": 6.273915294990676e-05}
{"train_loss": 0.1452743113040924, "global_step": 113756, "epoch": 1278, "lr": 6.273859234477635e-05}
{"train_loss": 0.1358373463153839, "global_step": 113757, "epoch": 1278, "lr": 6.273803173793338e-05}
{"train_loss": 0.16749712824821472, "global_step": 113758, "epoch": 1278, "lr": 6.273747112937793e-05}
{"train_loss": 0.20223715901374817, "global_step": 113759, "epoch": 1278, "lr": 6.273691051911012e-05}
{"train_loss": 0.11835120618343353, "global_step": 113760, "epoch": 1278, "lr": 6.273634990712995e-05}
{"train_loss": 0.1408642679452896, "global_step": 113761, "epoch": 1278, "lr": 6.273578929343755e-05}
{"train_loss": 0.2386426031589508, "global_step": 113762, "epoch": 1278, "lr": 6.273522867803297e-05}
{"train_loss": 0.11057928204536438, "global_step": 113763, "epoch": 1278, "lr": 6.27346680609163e-05}
{"train_loss": 0.1470608413219452, "global_step": 113764, "epoch": 1278, "lr": 6.27341074420876e-05}
{"train_loss": 0.15530185401439667, "global_step": 113765, "epoch": 1278, "lr": 6.273354682154697e-05}
{"train_loss": 0.11301694065332413, "global_step": 113766, "epoch": 1278, "lr": 6.273298619929446e-05}
{"train_loss": 0.14258889853954315, "global_step": 113767, "epoch": 1278, "lr": 6.273242557533017e-05}
{"train_loss": 0.0975164994597435, "global_step": 113768, "epoch": 1278, "lr": 6.273186494965414e-05}
{"train_loss": 0.19164502620697021, "global_step": 113769, "epoch": 1278, "lr": 6.273130432226648e-05}
{"train_loss": 0.13908104598522186, "global_step": 113770, "epoch": 1278, "lr": 6.273074369316723e-05}
{"train_loss": 0.08735156059265137, "global_step": 113771, "epoch": 1278, "lr": 6.273018306235652e-05}
{"train_loss": 0.15477652847766876, "global_step": 113772, "epoch": 1278, "lr": 6.272962242983436e-05}
{"train_loss": 0.11171407252550125, "global_step": 113773, "epoch": 1278, "lr": 6.272906179560088e-05}
{"train_loss": 0.09399980306625366, "global_step": 113774, "epoch": 1278, "lr": 6.272850115965612e-05}
{"train_loss": 0.07568059116601944, "global_step": 113775, "epoch": 1278, "lr": 6.272794052200018e-05}
{"train_loss": 0.17708539962768555, "global_step": 113776, "epoch": 1278, "lr": 6.272737988263309e-05}
{"train_loss": 0.08664912730455399, "global_step": 113777, "epoch": 1278, "lr": 6.272681924155499e-05}
{"train_loss": 0.08428226411342621, "global_step": 113778, "epoch": 1278, "lr": 6.272625859876593e-05}
{"train_loss": 0.1278993785381317, "global_step": 113779, "epoch": 1278, "lr": 6.272569795426598e-05}
{"train_loss": 0.15117080509662628, "global_step": 113780, "epoch": 1278, "lr": 6.272513730805519e-05}
{"train_loss": 0.10093487054109573, "global_step": 113781, "epoch": 1278, "lr": 6.272457666013367e-05}
{"train_loss": 0.1474912017583847, "global_step": 113782, "epoch": 1278, "lr": 6.27240160105015e-05}
{"train_loss": 0.16584299504756927, "global_step": 113783, "epoch": 1278, "lr": 6.272345535915872e-05}
{"train_loss": 0.22478361427783966, "global_step": 113784, "epoch": 1278, "lr": 6.272289470610544e-05}
{"train_loss": 0.13925553858280182, "global_step": 113785, "epoch": 1278, "lr": 6.272233405134172e-05}
{"train_loss": 0.1570049524307251, "global_step": 113786, "epoch": 1278, "lr": 6.272177339486765e-05}
{"train_loss": 0.16132789850234985, "global_step": 113787, "epoch": 1278, "lr": 6.272121273668326e-05}
{"train_loss": 0.06273465603590012, "global_step": 113788, "epoch": 1278, "lr": 6.272065207678868e-05}
{"train_loss": 0.2316562980413437, "global_step": 113789, "epoch": 1278, "lr": 6.272009141518397e-05}
{"train_loss": 0.1554017961025238, "global_step": 113790, "epoch": 1278, "lr": 6.271953075186918e-05}
{"train_loss": 0.09199851751327515, "global_step": 113791, "epoch": 1278, "lr": 6.271897008684442e-05}
{"train_loss": 0.12535460293293, "global_step": 113792, "epoch": 1278, "lr": 6.271840942010973e-05}
{"train_loss": 0.10676673799753189, "global_step": 113793, "epoch": 1278, "lr": 6.271784875166523e-05}
{"train_loss": 0.076772041618824, "global_step": 113794, "epoch": 1278, "lr": 6.271728808151093e-05}
{"train_loss": 0.21940216422080994, "global_step": 113795, "epoch": 1278, "lr": 6.271672740964698e-05}
{"train_loss": 0.11645416915416718, "global_step": 113796, "epoch": 1278, "lr": 6.271616673607342e-05}
{"train_loss": 0.2781371772289276, "global_step": 113797, "epoch": 1278, "lr": 6.271560606079032e-05}
{"train_loss": 0.1437063366174698, "global_step": 113798, "epoch": 1278, "lr": 6.271504538379778e-05}
{"train_loss": 0.15836770832538605, "global_step": 113799, "epoch": 1278, "lr": 6.271448470509583e-05}
{"train_loss": 0.1535000503063202, "global_step": 113800, "epoch": 1278, "lr": 6.271392402468459e-05}
{"train_loss": 0.11560872197151184, "global_step": 113801, "epoch": 1278, "lr": 6.27133633425641e-05}
{"train_loss": 0.05191701278090477, "global_step": 113802, "epoch": 1278, "lr": 6.271280265873447e-05}
{"train_loss": 0.13632483780384064, "global_step": 113803, "epoch": 1278, "lr": 6.271224197319574e-05}
{"train_loss": 0.16493359208106995, "global_step": 113804, "epoch": 1278, "lr": 6.271168128594802e-05}
{"train_loss": 0.24803929030895233, "global_step": 113805, "epoch": 1278, "lr": 6.271112059699137e-05}
{"train_loss": 0.08808209002017975, "global_step": 113806, "epoch": 1278, "lr": 6.271055990632587e-05}
{"train_loss": 0.18550506234169006, "global_step": 113807, "epoch": 1278, "lr": 6.270999921395157e-05}
{"train_loss": 0.17105278372764587, "global_step": 113808, "epoch": 1278, "lr": 6.270943851986858e-05}
{"train_loss": 0.13439378142356873, "global_step": 113809, "epoch": 1278, "lr": 6.270887782407695e-05}
{"train_loss": 0.12300672382116318, "global_step": 113810, "epoch": 1278, "lr": 6.270831712657677e-05}
{"train_loss": 0.12014070898294449, "global_step": 113811, "epoch": 1278, "lr": 6.270775642736812e-05}
{"train_loss": 0.15772922337055206, "global_step": 113812, "epoch": 1278, "lr": 6.270719572645106e-05}
{"train_loss": 0.1329057514667511, "global_step": 113813, "epoch": 1278, "lr": 6.270663502382567e-05}
{"train_loss": 0.11222092807292938, "global_step": 113814, "epoch": 1278, "lr": 6.270607431949203e-05}
{"train_loss": 0.14755332469940186, "global_step": 113815, "epoch": 1278, "lr": 6.270551361345021e-05}
{"train_loss": 0.19688045978546143, "global_step": 113816, "epoch": 1278, "lr": 6.27049529057003e-05}
{"train_loss": 0.13872869312763214, "global_step": 113817, "epoch": 1278, "lr": 6.270439219624236e-05}
{"train_loss": 0.12274650484323502, "global_step": 113818, "epoch": 1278, "lr": 6.270383148507647e-05}
{"train_loss": 0.1529073417186737, "global_step": 113819, "epoch": 1278, "lr": 6.27032707722027e-05}
{"train_loss": 0.13536480069160461, "global_step": 113820, "epoch": 1278, "lr": 6.270271005762112e-05}
{"train_loss": 0.11811351776123047, "global_step": 113821, "epoch": 1278, "lr": 6.270214934133185e-05}
{"train_loss": 0.3338508605957031, "global_step": 113822, "epoch": 1278, "lr": 6.27015886233349e-05}
{"train_loss": 0.11695186048746109, "global_step": 113823, "epoch": 1278, "lr": 6.270102790363038e-05}
{"train_loss": 0.08867684751749039, "global_step": 113824, "epoch": 1278, "lr": 6.270046718221836e-05}
{"train_loss": 0.14369052648544312, "global_step": 113825, "epoch": 1278, "lr": 6.269990645909893e-05}
{"train_loss": 0.11081996560096741, "global_step": 113826, "epoch": 1278, "lr": 6.269934573427215e-05}
{"train_loss": 0.16790089011192322, "global_step": 113827, "epoch": 1278, "lr": 6.26987850077381e-05}
{"train_loss": 0.1979905068874359, "global_step": 113828, "epoch": 1278, "lr": 6.269822427949683e-05}
{"train_loss": 0.09743926674127579, "global_step": 113829, "epoch": 1278, "lr": 6.269766354954846e-05}
{"train_loss": 0.14366335543186476, "global_step": 113830, "epoch": 1278, "lr": 6.269710281789303e-05, "val_loss": 5.204899311065674}
{"train_loss": 0.14463402330875397, "global_step": 113831, "epoch": 1279, "lr": 6.269654208453064e-05}
{"train_loss": 0.20898538827896118, "global_step": 113832, "epoch": 1279, "lr": 6.269598134946134e-05}
{"train_loss": 0.10507366806268692, "global_step": 113833, "epoch": 1279, "lr": 6.269542061268523e-05}
{"train_loss": 0.22109736502170563, "global_step": 113834, "epoch": 1279, "lr": 6.269485987420238e-05}
{"train_loss": 0.1577368527650833, "global_step": 113835, "epoch": 1279, "lr": 6.269429913401285e-05}
{"train_loss": 0.1245289221405983, "global_step": 113836, "epoch": 1279, "lr": 6.269373839211676e-05}
{"train_loss": 0.12617294490337372, "global_step": 113837, "epoch": 1279, "lr": 6.269317764851412e-05}
{"train_loss": 0.1430022120475769, "global_step": 113838, "epoch": 1279, "lr": 6.269261690320505e-05}
{"train_loss": 0.17827552556991577, "global_step": 113839, "epoch": 1279, "lr": 6.26920561561896e-05}
{"train_loss": 0.16664688289165497, "global_step": 113840, "epoch": 1279, "lr": 6.269149540746786e-05}
{"train_loss": 0.09493265300989151, "global_step": 113841, "epoch": 1279, "lr": 6.269093465703991e-05}
{"train_loss": 0.14470013976097107, "global_step": 113842, "epoch": 1279, "lr": 6.269037390490581e-05}
{"train_loss": 0.185056671500206, "global_step": 113843, "epoch": 1279, "lr": 6.268981315106566e-05}
{"train_loss": 0.11901021003723145, "global_step": 113844, "epoch": 1279, "lr": 6.26892523955195e-05}
{"train_loss": 0.17095711827278137, "global_step": 113845, "epoch": 1279, "lr": 6.268869163826744e-05}
{"train_loss": 0.1549956202507019, "global_step": 113846, "epoch": 1279, "lr": 6.268813087930953e-05}
{"train_loss": 0.166494220495224, "global_step": 113847, "epoch": 1279, "lr": 6.268757011864586e-05}
{"train_loss": 0.11997497826814651, "global_step": 113848, "epoch": 1279, "lr": 6.26870093562765e-05}
{"train_loss": 0.1268521249294281, "global_step": 113849, "epoch": 1279, "lr": 6.268644859220151e-05}
{"train_loss": 0.28380948305130005, "global_step": 113850, "epoch": 1279, "lr": 6.2685887826421e-05}
{"train_loss": 0.189419224858284, "global_step": 113851, "epoch": 1279, "lr": 6.268532705893502e-05}
{"train_loss": 0.17434875667095184, "global_step": 113852, "epoch": 1279, "lr": 6.268476628974365e-05}
{"train_loss": 0.13254612684249878, "global_step": 113853, "epoch": 1279, "lr": 6.268420551884699e-05}
{"train_loss": 0.1681804060935974, "global_step": 113854, "epoch": 1279, "lr": 6.268364474624506e-05}
{"train_loss": 0.22044122219085693, "global_step": 113855, "epoch": 1279, "lr": 6.2683083971938e-05}
{"train_loss": 0.058883219957351685, "global_step": 113856, "epoch": 1279, "lr": 6.268252319592584e-05}
{"train_loss": 0.16206157207489014, "global_step": 113857, "epoch": 1279, "lr": 6.268196241820868e-05}
{"train_loss": 0.15265339612960815, "global_step": 113858, "epoch": 1279, "lr": 6.268140163878658e-05}
{"train_loss": 0.2076481729745865, "global_step": 113859, "epoch": 1279, "lr": 6.26808408576596e-05}
{"train_loss": 0.13556893169879913, "global_step": 113860, "epoch": 1279, "lr": 6.268028007482785e-05}
{"train_loss": 0.07168709486722946, "global_step": 113861, "epoch": 1279, "lr": 6.267971929029139e-05}
{"train_loss": 0.1418066918849945, "global_step": 113862, "epoch": 1279, "lr": 6.26791585040503e-05}
{"train_loss": 0.10991458594799042, "global_step": 113863, "epoch": 1279, "lr": 6.267859771610465e-05}
{"train_loss": 0.17196717858314514, "global_step": 113864, "epoch": 1279, "lr": 6.267803692645451e-05}
{"train_loss": 0.13943271338939667, "global_step": 113865, "epoch": 1279, "lr": 6.267747613509997e-05}
{"train_loss": 0.1238672211766243, "global_step": 113866, "epoch": 1279, "lr": 6.26769153420411e-05}
{"train_loss": 0.21150261163711548, "global_step": 113867, "epoch": 1279, "lr": 6.267635454727796e-05}
{"train_loss": 0.20526616275310516, "global_step": 113868, "epoch": 1279, "lr": 6.267579375081064e-05}
{"train_loss": 0.22279421985149384, "global_step": 113869, "epoch": 1279, "lr": 6.267523295263921e-05}
{"train_loss": 0.09488683938980103, "global_step": 113870, "epoch": 1279, "lr": 6.267467215276378e-05}
{"train_loss": 0.12843933701515198, "global_step": 113871, "epoch": 1279, "lr": 6.267411135118437e-05}
{"train_loss": 0.14265257120132446, "global_step": 113872, "epoch": 1279, "lr": 6.267355054790108e-05}
{"train_loss": 0.15874095261096954, "global_step": 113873, "epoch": 1279, "lr": 6.2672989742914e-05}
{"train_loss": 0.15919558703899384, "global_step": 113874, "epoch": 1279, "lr": 6.267242893622318e-05}
{"train_loss": 0.19827251136302948, "global_step": 113875, "epoch": 1279, "lr": 6.267186812782872e-05}
{"train_loss": 0.17405271530151367, "global_step": 113876, "epoch": 1279, "lr": 6.267130731773068e-05}
{"train_loss": 0.20194247364997864, "global_step": 113877, "epoch": 1279, "lr": 6.267074650592913e-05}
{"train_loss": 0.21409063041210175, "global_step": 113878, "epoch": 1279, "lr": 6.267018569242413e-05}
{"train_loss": 0.12067142874002457, "global_step": 113879, "epoch": 1279, "lr": 6.26696248772158e-05}
{"train_loss": 0.14345702528953552, "global_step": 113880, "epoch": 1279, "lr": 6.26690640603042e-05}
{"train_loss": 0.10649921745061874, "global_step": 113881, "epoch": 1279, "lr": 6.266850324168939e-05}
{"train_loss": 0.20888762176036835, "global_step": 113882, "epoch": 1279, "lr": 6.266794242137145e-05}
{"train_loss": 0.08329098671674728, "global_step": 113883, "epoch": 1279, "lr": 6.266738159935047e-05}
{"train_loss": 0.21031858026981354, "global_step": 113884, "epoch": 1279, "lr": 6.266682077562651e-05}
{"train_loss": 0.0808010920882225, "global_step": 113885, "epoch": 1279, "lr": 6.266625995019963e-05}
{"train_loss": 0.14474181830883026, "global_step": 113886, "epoch": 1279, "lr": 6.266569912306995e-05}
{"train_loss": 0.18539029359817505, "global_step": 113887, "epoch": 1279, "lr": 6.266513829423752e-05}
{"train_loss": 0.1512092649936676, "global_step": 113888, "epoch": 1279, "lr": 6.266457746370241e-05}
{"train_loss": 0.12976475059986115, "global_step": 113889, "epoch": 1279, "lr": 6.26640166314647e-05}
{"train_loss": 0.18739749491214752, "global_step": 113890, "epoch": 1279, "lr": 6.266345579752449e-05}
{"train_loss": 0.2019229233264923, "global_step": 113891, "epoch": 1279, "lr": 6.26628949618818e-05}
{"train_loss": 0.23292028903961182, "global_step": 113892, "epoch": 1279, "lr": 6.266233412453676e-05}
{"train_loss": 0.16345451772212982, "global_step": 113893, "epoch": 1279, "lr": 6.266177328548941e-05}
{"train_loss": 0.12252963334321976, "global_step": 113894, "epoch": 1279, "lr": 6.266121244473985e-05}
{"train_loss": 0.10676596313714981, "global_step": 113895, "epoch": 1279, "lr": 6.266065160228814e-05}
{"train_loss": 0.11061761528253555, "global_step": 113896, "epoch": 1279, "lr": 6.266009075813437e-05}
{"train_loss": 0.0892498567700386, "global_step": 113897, "epoch": 1279, "lr": 6.265952991227859e-05}
{"train_loss": 0.1565704345703125, "global_step": 113898, "epoch": 1279, "lr": 6.265896906472089e-05}
{"train_loss": 0.14808832108974457, "global_step": 113899, "epoch": 1279, "lr": 6.265840821546136e-05}
{"train_loss": 0.13415279984474182, "global_step": 113900, "epoch": 1279, "lr": 6.265784736450004e-05}
{"train_loss": 0.11486650258302689, "global_step": 113901, "epoch": 1279, "lr": 6.265728651183705e-05}
{"train_loss": 0.1618984490633011, "global_step": 113902, "epoch": 1279, "lr": 6.265672565747243e-05}
{"train_loss": 0.1647202968597412, "global_step": 113903, "epoch": 1279, "lr": 6.265616480140626e-05}
{"train_loss": 0.12345684319734573, "global_step": 113904, "epoch": 1279, "lr": 6.265560394363863e-05}
{"train_loss": 0.15498299896717072, "global_step": 113905, "epoch": 1279, "lr": 6.26550430841696e-05}
{"train_loss": 0.16719955205917358, "global_step": 113906, "epoch": 1279, "lr": 6.265448222299928e-05}
{"train_loss": 0.10216757655143738, "global_step": 113907, "epoch": 1279, "lr": 6.26539213601277e-05}
{"train_loss": 0.1956029236316681, "global_step": 113908, "epoch": 1279, "lr": 6.265336049555494e-05}
{"train_loss": 0.15056994557380676, "global_step": 113909, "epoch": 1279, "lr": 6.265279962928112e-05}
{"train_loss": 0.11629638075828552, "global_step": 113910, "epoch": 1279, "lr": 6.265223876130627e-05}
{"train_loss": 0.1187925785779953, "global_step": 113911, "epoch": 1279, "lr": 6.265167789163049e-05}
{"train_loss": 0.1023741140961647, "global_step": 113912, "epoch": 1279, "lr": 6.265111702025383e-05}
{"train_loss": 0.11257780343294144, "global_step": 113913, "epoch": 1279, "lr": 6.26505561471764e-05}
{"train_loss": 0.0880441665649414, "global_step": 113914, "epoch": 1279, "lr": 6.264999527239826e-05}
{"train_loss": 0.13196273148059845, "global_step": 113915, "epoch": 1279, "lr": 6.264943439591947e-05}
{"train_loss": 0.1398865282535553, "global_step": 113916, "epoch": 1279, "lr": 6.264887351774012e-05}
{"train_loss": 0.12149349600076675, "global_step": 113917, "epoch": 1279, "lr": 6.264831263786028e-05}
{"train_loss": 0.14607961475849152, "global_step": 113918, "epoch": 1279, "lr": 6.264775175628001e-05}
{"train_loss": 0.1511115451039893, "global_step": 113919, "epoch": 1279, "lr": 6.264719087299944e-05, "val_loss": 5.130423069000244}
{"train_loss": 0.13911320269107819, "global_step": 113920, "epoch": 1280, "lr": 6.264662998801858e-05}
{"train_loss": 0.1945430338382721, "global_step": 113921, "epoch": 1280, "lr": 6.264606910133755e-05}
{"train_loss": 0.1366695761680603, "global_step": 113922, "epoch": 1280, "lr": 6.26455082129564e-05}
{"train_loss": 0.13417573273181915, "global_step": 113923, "epoch": 1280, "lr": 6.264494732287522e-05}
{"train_loss": 0.1570376604795456, "global_step": 113924, "epoch": 1280, "lr": 6.264438643109408e-05}
{"train_loss": 0.10678211599588394, "global_step": 113925, "epoch": 1280, "lr": 6.264382553761307e-05}
{"train_loss": 0.1351655274629593, "global_step": 113926, "epoch": 1280, "lr": 6.264326464243224e-05}
{"train_loss": 0.24122095108032227, "global_step": 113927, "epoch": 1280, "lr": 6.264270374555168e-05}
{"train_loss": 0.11310430616140366, "global_step": 113928, "epoch": 1280, "lr": 6.264214284697145e-05}
{"train_loss": 0.1007952019572258, "global_step": 113929, "epoch": 1280, "lr": 6.264158194669165e-05}
{"train_loss": 0.07719981670379639, "global_step": 113930, "epoch": 1280, "lr": 6.264102104471236e-05}
{"train_loss": 0.16841313242912292, "global_step": 113931, "epoch": 1280, "lr": 6.264046014103362e-05}
{"train_loss": 0.1496390849351883, "global_step": 113932, "epoch": 1280, "lr": 6.263989923565552e-05}
{"train_loss": 0.13930313289165497, "global_step": 113933, "epoch": 1280, "lr": 6.263933832857816e-05}
{"train_loss": 0.1491735726594925, "global_step": 113934, "epoch": 1280, "lr": 6.263877741980158e-05}
{"train_loss": 0.1236945316195488, "global_step": 113935, "epoch": 1280, "lr": 6.263821650932587e-05}
{"train_loss": 0.19714805483818054, "global_step": 113936, "epoch": 1280, "lr": 6.26376555971511e-05}
{"train_loss": 0.1389712244272232, "global_step": 113937, "epoch": 1280, "lr": 6.263709468327736e-05}
{"train_loss": 0.1335432231426239, "global_step": 113938, "epoch": 1280, "lr": 6.263653376770472e-05}
{"train_loss": 0.1636982262134552, "global_step": 113939, "epoch": 1280, "lr": 6.263597285043324e-05}
{"train_loss": 0.19091834127902985, "global_step": 113940, "epoch": 1280, "lr": 6.2635411931463e-05}
{"train_loss": 0.12317407876253128, "global_step": 113941, "epoch": 1280, "lr": 6.26348510107941e-05}
{"train_loss": 0.1475946456193924, "global_step": 113942, "epoch": 1280, "lr": 6.263429008842661e-05}
{"train_loss": 0.18618351221084595, "global_step": 113943, "epoch": 1280, "lr": 6.263372916436057e-05}
{"train_loss": 0.15549029409885406, "global_step": 113944, "epoch": 1280, "lr": 6.263316823859609e-05}
{"train_loss": 0.1644408404827118, "global_step": 113945, "epoch": 1280, "lr": 6.263260731113323e-05}
{"train_loss": 0.13598765432834625, "global_step": 113946, "epoch": 1280, "lr": 6.263204638197208e-05}
{"train_loss": 0.09441611170768738, "global_step": 113947, "epoch": 1280, "lr": 6.263148545111269e-05}
{"train_loss": 0.09491448849439621, "global_step": 113948, "epoch": 1280, "lr": 6.263092451855515e-05}
{"train_loss": 0.16353411972522736, "global_step": 113949, "epoch": 1280, "lr": 6.263036358429955e-05}
{"train_loss": 0.15274624526500702, "global_step": 113950, "epoch": 1280, "lr": 6.262980264834595e-05}
{"train_loss": 0.095136858522892, "global_step": 113951, "epoch": 1280, "lr": 6.262924171069441e-05}
{"train_loss": 0.13021749258041382, "global_step": 113952, "epoch": 1280, "lr": 6.262868077134506e-05}
{"train_loss": 0.12439167499542236, "global_step": 113953, "epoch": 1280, "lr": 6.262811983029792e-05}
{"train_loss": 0.12628379464149475, "global_step": 113954, "epoch": 1280, "lr": 6.262755888755307e-05}
{"train_loss": 0.1309249848127365, "global_step": 113955, "epoch": 1280, "lr": 6.262699794311061e-05}
{"train_loss": 0.1515093296766281, "global_step": 113956, "epoch": 1280, "lr": 6.26264369969706e-05}
{"train_loss": 0.1459800899028778, "global_step": 113957, "epoch": 1280, "lr": 6.262587604913312e-05}
{"train_loss": 0.22205336391925812, "global_step": 113958, "epoch": 1280, "lr": 6.262531509959822e-05}
{"train_loss": 0.16870346665382385, "global_step": 113959, "epoch": 1280, "lr": 6.262475414836603e-05}
{"train_loss": 0.1853286623954773, "global_step": 113960, "epoch": 1280, "lr": 6.262419319543659e-05}
{"train_loss": 0.0953250527381897, "global_step": 113961, "epoch": 1280, "lr": 6.262363224080999e-05}
{"train_loss": 0.1310754120349884, "global_step": 113962, "epoch": 1280, "lr": 6.262307128448628e-05}
{"train_loss": 0.1543412059545517, "global_step": 113963, "epoch": 1280, "lr": 6.262251032646556e-05}
{"train_loss": 0.14880290627479553, "global_step": 113964, "epoch": 1280, "lr": 6.26219493667479e-05}
{"train_loss": 0.1314760148525238, "global_step": 113965, "epoch": 1280, "lr": 6.262138840533336e-05}
{"train_loss": 0.13644292950630188, "global_step": 113966, "epoch": 1280, "lr": 6.262082744222205e-05}
{"train_loss": 0.1255641132593155, "global_step": 113967, "epoch": 1280, "lr": 6.262026647741401e-05}
{"train_loss": 0.23334923386573792, "global_step": 113968, "epoch": 1280, "lr": 6.261970551090931e-05}
{"train_loss": 0.20352396368980408, "global_step": 113969, "epoch": 1280, "lr": 6.261914454270808e-05}
{"train_loss": 0.11584676802158356, "global_step": 113970, "epoch": 1280, "lr": 6.261858357281034e-05}
{"train_loss": 0.19711047410964966, "global_step": 113971, "epoch": 1280, "lr": 6.261802260121619e-05}
{"train_loss": 0.116303950548172, "global_step": 113972, "epoch": 1280, "lr": 6.26174616279257e-05}
{"train_loss": 0.15671314299106598, "global_step": 113973, "epoch": 1280, "lr": 6.261690065293896e-05}
{"train_loss": 0.1229247972369194, "global_step": 113974, "epoch": 1280, "lr": 6.261633967625602e-05}
{"train_loss": 0.15903793275356293, "global_step": 113975, "epoch": 1280, "lr": 6.261577869787695e-05}
{"train_loss": 0.10335958749055862, "global_step": 113976, "epoch": 1280, "lr": 6.261521771780186e-05}
{"train_loss": 0.09703382104635239, "global_step": 113977, "epoch": 1280, "lr": 6.26146567360308e-05}
{"train_loss": 0.17264851927757263, "global_step": 113978, "epoch": 1280, "lr": 6.261409575256386e-05}
{"train_loss": 0.1386696994304657, "global_step": 113979, "epoch": 1280, "lr": 6.26135347674011e-05}
{"train_loss": 0.17812390625476837, "global_step": 113980, "epoch": 1280, "lr": 6.261297378054261e-05}
{"train_loss": 0.09445532411336899, "global_step": 113981, "epoch": 1280, "lr": 6.261241279198846e-05}
{"train_loss": 0.19479075074195862, "global_step": 113982, "epoch": 1280, "lr": 6.261185180173874e-05}
{"train_loss": 0.14069119095802307, "global_step": 113983, "epoch": 1280, "lr": 6.261129080979348e-05}
{"train_loss": 0.17828142642974854, "global_step": 113984, "epoch": 1280, "lr": 6.261072981615281e-05}
{"train_loss": 0.10034695267677307, "global_step": 113985, "epoch": 1280, "lr": 6.261016882081677e-05}
{"train_loss": 0.14081016182899475, "global_step": 113986, "epoch": 1280, "lr": 6.260960782378544e-05}
{"train_loss": 0.11379510909318924, "global_step": 113987, "epoch": 1280, "lr": 6.260904682505891e-05}
{"train_loss": 0.1002328023314476, "global_step": 113988, "epoch": 1280, "lr": 6.260848582463725e-05}
{"train_loss": 0.1403755247592926, "global_step": 113989, "epoch": 1280, "lr": 6.260792482252053e-05}
{"train_loss": 0.12384083122015, "global_step": 113990, "epoch": 1280, "lr": 6.260736381870883e-05}
{"train_loss": 0.17748144268989563, "global_step": 113991, "epoch": 1280, "lr": 6.260680281320223e-05}
{"train_loss": 0.10496729612350464, "global_step": 113992, "epoch": 1280, "lr": 6.26062418060008e-05}
{"train_loss": 0.12123561650514603, "global_step": 113993, "epoch": 1280, "lr": 6.26056807971046e-05}
{"train_loss": 0.20029182732105255, "global_step": 113994, "epoch": 1280, "lr": 6.260511978651372e-05}
{"train_loss": 0.15340162813663483, "global_step": 113995, "epoch": 1280, "lr": 6.260455877422823e-05}
{"train_loss": 0.12426065653562546, "global_step": 113996, "epoch": 1280, "lr": 6.260399776024823e-05}
{"train_loss": 0.10103420913219452, "global_step": 113997, "epoch": 1280, "lr": 6.260343674457376e-05}
{"train_loss": 0.09525872766971588, "global_step": 113998, "epoch": 1280, "lr": 6.260287572720493e-05}
{"train_loss": 0.1579914093017578, "global_step": 113999, "epoch": 1280, "lr": 6.260231470814177e-05}
{"train_loss": 0.1430986374616623, "global_step": 114000, "epoch": 1280, "lr": 6.260175368738441e-05}
{"train_loss": 0.14019106328487396, "global_step": 114001, "epoch": 1280, "lr": 6.260119266493288e-05}
{"train_loss": 0.14836065471172333, "global_step": 114002, "epoch": 1280, "lr": 6.260063164078727e-05}
{"train_loss": 0.11020007729530334, "global_step": 114003, "epoch": 1280, "lr": 6.260007061494768e-05}
{"train_loss": 0.11516450345516205, "global_step": 114004, "epoch": 1280, "lr": 6.259950958741415e-05}
{"train_loss": 0.1349593698978424, "global_step": 114005, "epoch": 1280, "lr": 6.259894855818678e-05}
{"train_loss": 0.18351569771766663, "global_step": 114006, "epoch": 1280, "lr": 6.259838752726561e-05}
{"train_loss": 0.07614556699991226, "global_step": 114007, "epoch": 1280, "lr": 6.259782649465076e-05}
{"train_loss": 0.1417573243379593, "global_step": 114008, "epoch": 1280, "lr": 6.259726546034229e-05, "val_loss": 5.163588047027588, "train_action_mse_error": 15.68441104888916}
{"train_loss": 0.16497085988521576, "global_step": 114009, "epoch": 1281, "lr": 6.259670442434025e-05}
{"train_loss": 0.11680780351161957, "global_step": 114010, "epoch": 1281, "lr": 6.259614338664477e-05}
{"train_loss": 0.155523419380188, "global_step": 114011, "epoch": 1281, "lr": 6.259558234725587e-05}
{"train_loss": 0.18586203455924988, "global_step": 114012, "epoch": 1281, "lr": 6.259502130617364e-05}
{"train_loss": 0.17050370573997498, "global_step": 114013, "epoch": 1281, "lr": 6.259446026339819e-05}
{"train_loss": 0.1764303296804428, "global_step": 114014, "epoch": 1281, "lr": 6.259389921892954e-05}
{"train_loss": 0.11637898534536362, "global_step": 114015, "epoch": 1281, "lr": 6.25933381727678e-05}
{"train_loss": 0.18802018463611603, "global_step": 114016, "epoch": 1281, "lr": 6.259277712491305e-05}
{"train_loss": 0.13844695687294006, "global_step": 114017, "epoch": 1281, "lr": 6.259221607536534e-05}
{"train_loss": 0.24608837068080902, "global_step": 114018, "epoch": 1281, "lr": 6.259165502412477e-05}
{"train_loss": 0.10296165943145752, "global_step": 114019, "epoch": 1281, "lr": 6.259109397119141e-05}
{"train_loss": 0.11878068000078201, "global_step": 114020, "epoch": 1281, "lr": 6.259053291656532e-05}
{"train_loss": 0.12252694368362427, "global_step": 114021, "epoch": 1281, "lr": 6.25899718602466e-05}
{"train_loss": 0.13257038593292236, "global_step": 114022, "epoch": 1281, "lr": 6.25894108022353e-05}
{"train_loss": 0.15375304222106934, "global_step": 114023, "epoch": 1281, "lr": 6.258884974253151e-05}
{"train_loss": 0.15080587565898895, "global_step": 114024, "epoch": 1281, "lr": 6.25882886811353e-05}
{"train_loss": 0.16160224378108978, "global_step": 114025, "epoch": 1281, "lr": 6.258772761804675e-05}
{"train_loss": 0.14618496596813202, "global_step": 114026, "epoch": 1281, "lr": 6.258716655326593e-05}
{"train_loss": 0.168459951877594, "global_step": 114027, "epoch": 1281, "lr": 6.258660548679292e-05}
{"train_loss": 0.17271621525287628, "global_step": 114028, "epoch": 1281, "lr": 6.25860444186278e-05}
{"train_loss": 0.14388930797576904, "global_step": 114029, "epoch": 1281, "lr": 6.258548334877065e-05}
{"train_loss": 0.11930423229932785, "global_step": 114030, "epoch": 1281, "lr": 6.258492227722151e-05}
{"train_loss": 0.15955941379070282, "global_step": 114031, "epoch": 1281, "lr": 6.25843612039805e-05}
{"train_loss": 0.1398734301328659, "global_step": 114032, "epoch": 1281, "lr": 6.258380012904766e-05}
{"train_loss": 0.17291319370269775, "global_step": 114033, "epoch": 1281, "lr": 6.258323905242309e-05}
{"train_loss": 0.14688479900360107, "global_step": 114034, "epoch": 1281, "lr": 6.258267797410686e-05}
{"train_loss": 0.1611299067735672, "global_step": 114035, "epoch": 1281, "lr": 6.258211689409905e-05}
{"train_loss": 0.14612740278244019, "global_step": 114036, "epoch": 1281, "lr": 6.25815558123997e-05}
{"train_loss": 0.21291251480579376, "global_step": 114037, "epoch": 1281, "lr": 6.258099472900893e-05}
{"train_loss": 0.07941050082445145, "global_step": 114038, "epoch": 1281, "lr": 6.258043364392681e-05}
{"train_loss": 0.17416329681873322, "global_step": 114039, "epoch": 1281, "lr": 6.25798725571534e-05}
{"train_loss": 0.1999821811914444, "global_step": 114040, "epoch": 1281, "lr": 6.257931146868876e-05}
{"train_loss": 0.1800020933151245, "global_step": 114041, "epoch": 1281, "lr": 6.2578750378533e-05}
{"train_loss": 0.26033541560173035, "global_step": 114042, "epoch": 1281, "lr": 6.257818928668619e-05}
{"train_loss": 0.180268794298172, "global_step": 114043, "epoch": 1281, "lr": 6.257762819314837e-05}
{"train_loss": 0.15055415034294128, "global_step": 114044, "epoch": 1281, "lr": 6.257706709791966e-05}
{"train_loss": 0.19171583652496338, "global_step": 114045, "epoch": 1281, "lr": 6.257650600100011e-05}
{"train_loss": 0.15320956707000732, "global_step": 114046, "epoch": 1281, "lr": 6.25759449023898e-05}
{"train_loss": 0.1629388928413391, "global_step": 114047, "epoch": 1281, "lr": 6.257538380208881e-05}
{"train_loss": 0.11609049141407013, "global_step": 114048, "epoch": 1281, "lr": 6.257482270009722e-05}
{"train_loss": 0.2107347846031189, "global_step": 114049, "epoch": 1281, "lr": 6.257426159641509e-05}
{"train_loss": 0.12024488300085068, "global_step": 114050, "epoch": 1281, "lr": 6.25737004910425e-05}
{"train_loss": 0.1712512969970703, "global_step": 114051, "epoch": 1281, "lr": 6.257313938397954e-05}
{"train_loss": 0.2013818919658661, "global_step": 114052, "epoch": 1281, "lr": 6.257257827522627e-05}
{"train_loss": 0.17750442028045654, "global_step": 114053, "epoch": 1281, "lr": 6.257201716478278e-05}
{"train_loss": 0.07738116383552551, "global_step": 114054, "epoch": 1281, "lr": 6.257145605264912e-05}
{"train_loss": 0.08343707025051117, "global_step": 114055, "epoch": 1281, "lr": 6.257089493882538e-05}
{"train_loss": 0.12844866514205933, "global_step": 114056, "epoch": 1281, "lr": 6.257033382331165e-05}
{"train_loss": 0.14380541443824768, "global_step": 114057, "epoch": 1281, "lr": 6.256977270610798e-05}
{"train_loss": 0.12038988620042801, "global_step": 114058, "epoch": 1281, "lr": 6.256921158721445e-05}
{"train_loss": 0.12155594676733017, "global_step": 114059, "epoch": 1281, "lr": 6.256865046663116e-05}
{"train_loss": 0.219700887799263, "global_step": 114060, "epoch": 1281, "lr": 6.256808934435817e-05}
{"train_loss": 0.11878328770399094, "global_step": 114061, "epoch": 1281, "lr": 6.256752822039554e-05}
{"train_loss": 0.13042205572128296, "global_step": 114062, "epoch": 1281, "lr": 6.256696709474337e-05}
{"train_loss": 0.10106541216373444, "global_step": 114063, "epoch": 1281, "lr": 6.256640596740171e-05}
{"train_loss": 0.17874284088611603, "global_step": 114064, "epoch": 1281, "lr": 6.256584483837067e-05}
{"train_loss": 0.1724090576171875, "global_step": 114065, "epoch": 1281, "lr": 6.256528370765029e-05}
{"train_loss": 0.16682860255241394, "global_step": 114066, "epoch": 1281, "lr": 6.256472257524067e-05}
{"train_loss": 0.13405479490756989, "global_step": 114067, "epoch": 1281, "lr": 6.256416144114187e-05}
{"train_loss": 0.16850316524505615, "global_step": 114068, "epoch": 1281, "lr": 6.256360030535397e-05}
{"train_loss": 0.09965433180332184, "global_step": 114069, "epoch": 1281, "lr": 6.256303916787706e-05}
{"train_loss": 0.14015646278858185, "global_step": 114070, "epoch": 1281, "lr": 6.25624780287112e-05}
{"train_loss": 0.11654017120599747, "global_step": 114071, "epoch": 1281, "lr": 6.256191688785644e-05}
{"train_loss": 0.14499352872371674, "global_step": 114072, "epoch": 1281, "lr": 6.256135574531291e-05}
{"train_loss": 0.14726275205612183, "global_step": 114073, "epoch": 1281, "lr": 6.256079460108065e-05}
{"train_loss": 0.12485801428556442, "global_step": 114074, "epoch": 1281, "lr": 6.256023345515976e-05}
{"train_loss": 0.14552554488182068, "global_step": 114075, "epoch": 1281, "lr": 6.255967230755027e-05}
{"train_loss": 0.16620194911956787, "global_step": 114076, "epoch": 1281, "lr": 6.255911115825232e-05}
{"train_loss": 0.1838916838169098, "global_step": 114077, "epoch": 1281, "lr": 6.255855000726592e-05}
{"train_loss": 0.10529424250125885, "global_step": 114078, "epoch": 1281, "lr": 6.255798885459119e-05}
{"train_loss": 0.12997035682201385, "global_step": 114079, "epoch": 1281, "lr": 6.255742770022818e-05}
{"train_loss": 0.19029659032821655, "global_step": 114080, "epoch": 1281, "lr": 6.2556866544177e-05}
{"train_loss": 0.13304297626018524, "global_step": 114081, "epoch": 1281, "lr": 6.255630538643768e-05}
{"train_loss": 0.12266316264867783, "global_step": 114082, "epoch": 1281, "lr": 6.255574422701034e-05}
{"train_loss": 0.14740127325057983, "global_step": 114083, "epoch": 1281, "lr": 6.255518306589502e-05}
{"train_loss": 0.07483460009098053, "global_step": 114084, "epoch": 1281, "lr": 6.25546219030918e-05}
{"train_loss": 0.18583393096923828, "global_step": 114085, "epoch": 1281, "lr": 6.255406073860078e-05}
{"train_loss": 0.12265089899301529, "global_step": 114086, "epoch": 1281, "lr": 6.255349957242201e-05}
{"train_loss": 0.09002252668142319, "global_step": 114087, "epoch": 1281, "lr": 6.255293840455558e-05}
{"train_loss": 0.10127472877502441, "global_step": 114088, "epoch": 1281, "lr": 6.255237723500156e-05}
{"train_loss": 0.21089458465576172, "global_step": 114089, "epoch": 1281, "lr": 6.255181606376003e-05}
{"train_loss": 0.09990208595991135, "global_step": 114090, "epoch": 1281, "lr": 6.255125489083105e-05}
{"train_loss": 0.10220842808485031, "global_step": 114091, "epoch": 1281, "lr": 6.255069371621471e-05}
{"train_loss": 0.1828339695930481, "global_step": 114092, "epoch": 1281, "lr": 6.255013253991108e-05}
{"train_loss": 0.12632443010807037, "global_step": 114093, "epoch": 1281, "lr": 6.254957136192026e-05}
{"train_loss": 0.09815404564142227, "global_step": 114094, "epoch": 1281, "lr": 6.254901018224227e-05}
{"train_loss": 0.15266738831996918, "global_step": 114095, "epoch": 1281, "lr": 6.254844900087723e-05}
{"train_loss": 0.15749095380306244, "global_step": 114096, "epoch": 1281, "lr": 6.254788781782521e-05}
{"train_loss": 0.14901075855399784, "global_step": 114097, "epoch": 1281, "lr": 6.254732663308628e-05, "val_loss": 5.277527332305908}
{"train_loss": 0.09199229627847672, "global_step": 114098, "epoch": 1282, "lr": 6.254676544666052e-05}
{"train_loss": 0.1789301484823227, "global_step": 114099, "epoch": 1282, "lr": 6.254620425854799e-05}
{"train_loss": 0.14408525824546814, "global_step": 114100, "epoch": 1282, "lr": 6.254564306874877e-05}
{"train_loss": 0.1896672248840332, "global_step": 114101, "epoch": 1282, "lr": 6.254508187726295e-05}
{"train_loss": 0.14036062359809875, "global_step": 114102, "epoch": 1282, "lr": 6.25445206840906e-05}
{"train_loss": 0.1259257197380066, "global_step": 114103, "epoch": 1282, "lr": 6.254395948923179e-05}
{"train_loss": 0.1153203472495079, "global_step": 114104, "epoch": 1282, "lr": 6.254339829268659e-05}
{"train_loss": 0.17725041508674622, "global_step": 114105, "epoch": 1282, "lr": 6.25428370944551e-05}
{"train_loss": 0.11638175696134567, "global_step": 114106, "epoch": 1282, "lr": 6.254227589453736e-05}
{"train_loss": 0.13669390976428986, "global_step": 114107, "epoch": 1282, "lr": 6.254171469293348e-05}
{"train_loss": 0.10632812976837158, "global_step": 114108, "epoch": 1282, "lr": 6.25411534896435e-05}
{"train_loss": 0.12837563455104828, "global_step": 114109, "epoch": 1282, "lr": 6.254059228466753e-05}
{"train_loss": 0.11805666238069534, "global_step": 114110, "epoch": 1282, "lr": 6.254003107800563e-05}
{"train_loss": 0.2011139988899231, "global_step": 114111, "epoch": 1282, "lr": 6.253946986965786e-05}
{"train_loss": 0.14121799170970917, "global_step": 114112, "epoch": 1282, "lr": 6.253890865962433e-05}
{"train_loss": 0.0925828069448471, "global_step": 114113, "epoch": 1282, "lr": 6.253834744790507e-05}
{"train_loss": 0.13825048506259918, "global_step": 114114, "epoch": 1282, "lr": 6.253778623450022e-05}
{"train_loss": 0.13105739653110504, "global_step": 114115, "epoch": 1282, "lr": 6.253722501940979e-05}
{"train_loss": 0.2766996920108795, "global_step": 114116, "epoch": 1282, "lr": 6.253666380263388e-05}
{"train_loss": 0.11442946642637253, "global_step": 114117, "epoch": 1282, "lr": 6.25361025841726e-05}
{"train_loss": 0.14470988512039185, "global_step": 114118, "epoch": 1282, "lr": 6.253554136402597e-05}
{"train_loss": 0.1925629824399948, "global_step": 114119, "epoch": 1282, "lr": 6.253498014219409e-05}
{"train_loss": 0.11737686395645142, "global_step": 114120, "epoch": 1282, "lr": 6.253441891867703e-05}
{"train_loss": 0.12149691581726074, "global_step": 114121, "epoch": 1282, "lr": 6.253385769347488e-05}
{"train_loss": 0.08351919800043106, "global_step": 114122, "epoch": 1282, "lr": 6.25332964665877e-05}
{"train_loss": 0.1488918960094452, "global_step": 114123, "epoch": 1282, "lr": 6.253273523801557e-05}
{"train_loss": 0.1204659715294838, "global_step": 114124, "epoch": 1282, "lr": 6.253217400775856e-05}
{"train_loss": 0.10510209947824478, "global_step": 114125, "epoch": 1282, "lr": 6.253161277581678e-05}
{"train_loss": 0.10896287113428116, "global_step": 114126, "epoch": 1282, "lr": 6.253105154219025e-05}
{"train_loss": 0.12936025857925415, "global_step": 114127, "epoch": 1282, "lr": 6.253049030687908e-05}
{"train_loss": 0.13864779472351074, "global_step": 114128, "epoch": 1282, "lr": 6.252992906988335e-05}
{"train_loss": 0.15108612179756165, "global_step": 114129, "epoch": 1282, "lr": 6.252936783120312e-05}
{"train_loss": 0.1568501889705658, "global_step": 114130, "epoch": 1282, "lr": 6.252880659083847e-05}
{"train_loss": 0.1804593801498413, "global_step": 114131, "epoch": 1282, "lr": 6.252824534878944e-05}
{"train_loss": 0.20337003469467163, "global_step": 114132, "epoch": 1282, "lr": 6.252768410505618e-05}
{"train_loss": 0.13999061286449432, "global_step": 114133, "epoch": 1282, "lr": 6.252712285963872e-05}
{"train_loss": 0.18415941298007965, "global_step": 114134, "epoch": 1282, "lr": 6.252656161253713e-05}
{"train_loss": 0.11619819700717926, "global_step": 114135, "epoch": 1282, "lr": 6.25260003637515e-05}
{"train_loss": 0.177883580327034, "global_step": 114136, "epoch": 1282, "lr": 6.25254391132819e-05}
{"train_loss": 0.15503890812397003, "global_step": 114137, "epoch": 1282, "lr": 6.252487786112842e-05}
{"train_loss": 0.10513249784708023, "global_step": 114138, "epoch": 1282, "lr": 6.252431660729111e-05}
{"train_loss": 0.13482151925563812, "global_step": 114139, "epoch": 1282, "lr": 6.252375535177005e-05}
{"train_loss": 0.2491922676563263, "global_step": 114140, "epoch": 1282, "lr": 6.252319409456534e-05}
{"train_loss": 0.08998966962099075, "global_step": 114141, "epoch": 1282, "lr": 6.252263283567705e-05}
{"train_loss": 0.17591464519500732, "global_step": 114142, "epoch": 1282, "lr": 6.252207157510523e-05}
{"train_loss": 0.20248639583587646, "global_step": 114143, "epoch": 1282, "lr": 6.252151031284997e-05}
{"train_loss": 0.1840621680021286, "global_step": 114144, "epoch": 1282, "lr": 6.252094904891134e-05}
{"train_loss": 0.13513992726802826, "global_step": 114145, "epoch": 1282, "lr": 6.252038778328942e-05}
{"train_loss": 0.21495793759822845, "global_step": 114146, "epoch": 1282, "lr": 6.251982651598429e-05}
{"train_loss": 0.08083358407020569, "global_step": 114147, "epoch": 1282, "lr": 6.251926524699603e-05}
{"train_loss": 0.11812098324298859, "global_step": 114148, "epoch": 1282, "lr": 6.251870397632472e-05}
{"train_loss": 0.24242031574249268, "global_step": 114149, "epoch": 1282, "lr": 6.251814270397039e-05}
{"train_loss": 0.20165219902992249, "global_step": 114150, "epoch": 1282, "lr": 6.251758142993316e-05}
{"train_loss": 0.19474057853221893, "global_step": 114151, "epoch": 1282, "lr": 6.251702015421311e-05}
{"train_loss": 0.18643662333488464, "global_step": 114152, "epoch": 1282, "lr": 6.251645887681028e-05}
{"train_loss": 0.14954254031181335, "global_step": 114153, "epoch": 1282, "lr": 6.251589759772476e-05}
{"train_loss": 0.1250571310520172, "global_step": 114154, "epoch": 1282, "lr": 6.251533631695666e-05}
{"train_loss": 0.11333253234624863, "global_step": 114155, "epoch": 1282, "lr": 6.251477503450599e-05}
{"train_loss": 0.1507011502981186, "global_step": 114156, "epoch": 1282, "lr": 6.251421375037288e-05}
{"train_loss": 0.19783391058444977, "global_step": 114157, "epoch": 1282, "lr": 6.251365246455737e-05}
{"train_loss": 0.2275371104478836, "global_step": 114158, "epoch": 1282, "lr": 6.251309117705957e-05}
{"train_loss": 0.1580590158700943, "global_step": 114159, "epoch": 1282, "lr": 6.251252988787953e-05}
{"train_loss": 0.12256897240877151, "global_step": 114160, "epoch": 1282, "lr": 6.251196859701734e-05}
{"train_loss": 0.21072448790073395, "global_step": 114161, "epoch": 1282, "lr": 6.251140730447307e-05}
{"train_loss": 0.21495237946510315, "global_step": 114162, "epoch": 1282, "lr": 6.25108460102468e-05}
{"train_loss": 0.20764750242233276, "global_step": 114163, "epoch": 1282, "lr": 6.251028471433857e-05}
{"train_loss": 0.13120022416114807, "global_step": 114164, "epoch": 1282, "lr": 6.250972341674851e-05}
{"train_loss": 0.15412823855876923, "global_step": 114165, "epoch": 1282, "lr": 6.250916211747666e-05}
{"train_loss": 0.19465188682079315, "global_step": 114166, "epoch": 1282, "lr": 6.250860081652311e-05}
{"train_loss": 0.2323986440896988, "global_step": 114167, "epoch": 1282, "lr": 6.250803951388792e-05}
{"train_loss": 0.11599043756723404, "global_step": 114168, "epoch": 1282, "lr": 6.250747820957119e-05}
{"train_loss": 0.1385604292154312, "global_step": 114169, "epoch": 1282, "lr": 6.250691690357298e-05}
{"train_loss": 0.2527690529823303, "global_step": 114170, "epoch": 1282, "lr": 6.250635559589338e-05}
{"train_loss": 0.11722148954868317, "global_step": 114171, "epoch": 1282, "lr": 6.250579428653243e-05}
{"train_loss": 0.12895174324512482, "global_step": 114172, "epoch": 1282, "lr": 6.250523297549023e-05}
{"train_loss": 0.20486138761043549, "global_step": 114173, "epoch": 1282, "lr": 6.250467166276685e-05}
{"train_loss": 0.15043357014656067, "global_step": 114174, "epoch": 1282, "lr": 6.250411034836239e-05}
{"train_loss": 0.09965484589338303, "global_step": 114175, "epoch": 1282, "lr": 6.250354903227688e-05}
{"train_loss": 0.14406710863113403, "global_step": 114176, "epoch": 1282, "lr": 6.250298771451045e-05}
{"train_loss": 0.11464464664459229, "global_step": 114177, "epoch": 1282, "lr": 6.250242639506312e-05}
{"train_loss": 0.16551002860069275, "global_step": 114178, "epoch": 1282, "lr": 6.2501865073935e-05}
{"train_loss": 0.06243428587913513, "global_step": 114179, "epoch": 1282, "lr": 6.250130375112617e-05}
{"train_loss": 0.10147039592266083, "global_step": 114180, "epoch": 1282, "lr": 6.250074242663668e-05}
{"train_loss": 0.15767496824264526, "global_step": 114181, "epoch": 1282, "lr": 6.250018110046661e-05}
{"train_loss": 0.13910441100597382, "global_step": 114182, "epoch": 1282, "lr": 6.249961977261605e-05}
{"train_loss": 0.16692565381526947, "global_step": 114183, "epoch": 1282, "lr": 6.249905844308507e-05}
{"train_loss": 0.15040212869644165, "global_step": 114184, "epoch": 1282, "lr": 6.249849711187374e-05}
{"train_loss": 0.1272127330303192, "global_step": 114185, "epoch": 1282, "lr": 6.249793577898216e-05}
{"train_loss": 0.1515286820155851, "global_step": 114186, "epoch": 1282, "lr": 6.249737444441036e-05, "val_loss": 5.112233638763428}
{"train_loss": 0.15583188831806183, "global_step": 114187, "epoch": 1283, "lr": 6.249681310815845e-05}
{"train_loss": 0.1635701209306717, "global_step": 114188, "epoch": 1283, "lr": 6.249625177022649e-05}
{"train_loss": 0.16746321320533752, "global_step": 114189, "epoch": 1283, "lr": 6.249569043061457e-05}
{"train_loss": 0.10703843832015991, "global_step": 114190, "epoch": 1283, "lr": 6.249512908932274e-05}
{"train_loss": 0.11913364380598068, "global_step": 114191, "epoch": 1283, "lr": 6.24945677463511e-05}
{"train_loss": 0.14214980602264404, "global_step": 114192, "epoch": 1283, "lr": 6.249400640169973e-05}
{"train_loss": 0.10097935050725937, "global_step": 114193, "epoch": 1283, "lr": 6.249344505536868e-05}
{"train_loss": 0.18057101964950562, "global_step": 114194, "epoch": 1283, "lr": 6.249288370735803e-05}
{"train_loss": 0.23142756521701813, "global_step": 114195, "epoch": 1283, "lr": 6.249232235766787e-05}
{"train_loss": 0.09375245869159698, "global_step": 114196, "epoch": 1283, "lr": 6.249176100629828e-05}
{"train_loss": 0.11273732781410217, "global_step": 114197, "epoch": 1283, "lr": 6.249119965324933e-05}
{"train_loss": 0.13138246536254883, "global_step": 114198, "epoch": 1283, "lr": 6.249063829852107e-05}
{"train_loss": 0.17751233279705048, "global_step": 114199, "epoch": 1283, "lr": 6.249007694211363e-05}
{"train_loss": 0.11481941491365433, "global_step": 114200, "epoch": 1283, "lr": 6.248951558402703e-05}
{"train_loss": 0.17374292016029358, "global_step": 114201, "epoch": 1283, "lr": 6.248895422426136e-05}
{"train_loss": 0.12673377990722656, "global_step": 114202, "epoch": 1283, "lr": 6.248839286281672e-05}
{"train_loss": 0.18348802626132965, "global_step": 114203, "epoch": 1283, "lr": 6.248783149969315e-05}
{"train_loss": 0.15522335469722748, "global_step": 114204, "epoch": 1283, "lr": 6.248727013489075e-05}
{"train_loss": 0.09982138127088547, "global_step": 114205, "epoch": 1283, "lr": 6.248670876840959e-05}
{"train_loss": 0.1186177209019661, "global_step": 114206, "epoch": 1283, "lr": 6.248614740024975e-05}
{"train_loss": 0.1799430549144745, "global_step": 114207, "epoch": 1283, "lr": 6.248558603041129e-05}
{"train_loss": 0.08768154680728912, "global_step": 114208, "epoch": 1283, "lr": 6.248502465889429e-05}
{"train_loss": 0.102147676050663, "global_step": 114209, "epoch": 1283, "lr": 6.248446328569883e-05}
{"train_loss": 0.15002219378948212, "global_step": 114210, "epoch": 1283, "lr": 6.2483901910825e-05}
{"train_loss": 0.12634366750717163, "global_step": 114211, "epoch": 1283, "lr": 6.248334053427285e-05}
{"train_loss": 0.16268017888069153, "global_step": 114212, "epoch": 1283, "lr": 6.248277915604246e-05}
{"train_loss": 0.18740470707416534, "global_step": 114213, "epoch": 1283, "lr": 6.248221777613393e-05}
{"train_loss": 0.12202725559473038, "global_step": 114214, "epoch": 1283, "lr": 6.248165639454732e-05}
{"train_loss": 0.17125841975212097, "global_step": 114215, "epoch": 1283, "lr": 6.248109501128268e-05}
{"train_loss": 0.12214301526546478, "global_step": 114216, "epoch": 1283, "lr": 6.248053362634014e-05}
{"train_loss": 0.1959785670042038, "global_step": 114217, "epoch": 1283, "lr": 6.247997223971972e-05}
{"train_loss": 0.17546938359737396, "global_step": 114218, "epoch": 1283, "lr": 6.247941085142155e-05}
{"train_loss": 0.23721154034137726, "global_step": 114219, "epoch": 1283, "lr": 6.247884946144565e-05}
{"train_loss": 0.2041279524564743, "global_step": 114220, "epoch": 1283, "lr": 6.247828806979212e-05}
{"train_loss": 0.18947316706180573, "global_step": 114221, "epoch": 1283, "lr": 6.247772667646105e-05}
{"train_loss": 0.14712285995483398, "global_step": 114222, "epoch": 1283, "lr": 6.247716528145248e-05}
{"train_loss": 0.21741892397403717, "global_step": 114223, "epoch": 1283, "lr": 6.247660388476652e-05}
{"train_loss": 0.0977669358253479, "global_step": 114224, "epoch": 1283, "lr": 6.247604248640325e-05}
{"train_loss": 0.1692030429840088, "global_step": 114225, "epoch": 1283, "lr": 6.24754810863627e-05}
{"train_loss": 0.13947586715221405, "global_step": 114226, "epoch": 1283, "lr": 6.247491968464498e-05}
{"train_loss": 0.21308794617652893, "global_step": 114227, "epoch": 1283, "lr": 6.247435828125017e-05}
{"train_loss": 0.09909684956073761, "global_step": 114228, "epoch": 1283, "lr": 6.247379687617834e-05}
{"train_loss": 0.18310295045375824, "global_step": 114229, "epoch": 1283, "lr": 6.247323546942954e-05}
{"train_loss": 0.10520973056554794, "global_step": 114230, "epoch": 1283, "lr": 6.247267406100387e-05}
{"train_loss": 0.09450249373912811, "global_step": 114231, "epoch": 1283, "lr": 6.24721126509014e-05}
{"train_loss": 0.13200674951076508, "global_step": 114232, "epoch": 1283, "lr": 6.247155123912221e-05}
{"train_loss": 0.08379335701465607, "global_step": 114233, "epoch": 1283, "lr": 6.247098982566637e-05}
{"train_loss": 0.1317436248064041, "global_step": 114234, "epoch": 1283, "lr": 6.247042841053397e-05}
{"train_loss": 0.12039054930210114, "global_step": 114235, "epoch": 1283, "lr": 6.246986699372505e-05}
{"train_loss": 0.11341413855552673, "global_step": 114236, "epoch": 1283, "lr": 6.246930557523973e-05}
{"train_loss": 0.11146610975265503, "global_step": 114237, "epoch": 1283, "lr": 6.246874415507807e-05}
{"train_loss": 0.1860446184873581, "global_step": 114238, "epoch": 1283, "lr": 6.246818273324012e-05}
{"train_loss": 0.12408456206321716, "global_step": 114239, "epoch": 1283, "lr": 6.246762130972598e-05}
{"train_loss": 0.14509132504463196, "global_step": 114240, "epoch": 1283, "lr": 6.246705988453572e-05}
{"train_loss": 0.1776336282491684, "global_step": 114241, "epoch": 1283, "lr": 6.246649845766942e-05}
{"train_loss": 0.15262244641780853, "global_step": 114242, "epoch": 1283, "lr": 6.246593702912715e-05}
{"train_loss": 0.1645234227180481, "global_step": 114243, "epoch": 1283, "lr": 6.246537559890898e-05}
{"train_loss": 0.1874324381351471, "global_step": 114244, "epoch": 1283, "lr": 6.246481416701499e-05}
{"train_loss": 0.1423775851726532, "global_step": 114245, "epoch": 1283, "lr": 6.246425273344527e-05}
{"train_loss": 0.13787004351615906, "global_step": 114246, "epoch": 1283, "lr": 6.246369129819988e-05}
{"train_loss": 0.2354116588830948, "global_step": 114247, "epoch": 1283, "lr": 6.246312986127889e-05}
{"train_loss": 0.18865932524204254, "global_step": 114248, "epoch": 1283, "lr": 6.246256842268239e-05}
{"train_loss": 0.14962267875671387, "global_step": 114249, "epoch": 1283, "lr": 6.246200698241044e-05}
{"train_loss": 0.24863803386688232, "global_step": 114250, "epoch": 1283, "lr": 6.246144554046314e-05}
{"train_loss": 0.1390671730041504, "global_step": 114251, "epoch": 1283, "lr": 6.246088409684054e-05}
{"train_loss": 0.16763611137866974, "global_step": 114252, "epoch": 1283, "lr": 6.246032265154274e-05}
{"train_loss": 0.08304077386856079, "global_step": 114253, "epoch": 1283, "lr": 6.245976120456979e-05}
{"train_loss": 0.17265762388706207, "global_step": 114254, "epoch": 1283, "lr": 6.245919975592178e-05}
{"train_loss": 0.15129005908966064, "global_step": 114255, "epoch": 1283, "lr": 6.245863830559878e-05}
{"train_loss": 0.11209280788898468, "global_step": 114256, "epoch": 1283, "lr": 6.245807685360087e-05}
{"train_loss": 0.0610770657658577, "global_step": 114257, "epoch": 1283, "lr": 6.245751539992814e-05}
{"train_loss": 0.15023811161518097, "global_step": 114258, "epoch": 1283, "lr": 6.245695394458062e-05}
{"train_loss": 0.13257859647274017, "global_step": 114259, "epoch": 1283, "lr": 6.245639248755843e-05}
{"train_loss": 0.09867885708808899, "global_step": 114260, "epoch": 1283, "lr": 6.245583102886164e-05}
{"train_loss": 0.10652992874383926, "global_step": 114261, "epoch": 1283, "lr": 6.245526956849029e-05}
{"train_loss": 0.1841263771057129, "global_step": 114262, "epoch": 1283, "lr": 6.245470810644449e-05}
{"train_loss": 0.16109806299209595, "global_step": 114263, "epoch": 1283, "lr": 6.245414664272431e-05}
{"train_loss": 0.16985107958316803, "global_step": 114264, "epoch": 1283, "lr": 6.245358517732982e-05}
{"train_loss": 0.16390907764434814, "global_step": 114265, "epoch": 1283, "lr": 6.245302371026111e-05}
{"train_loss": 0.2479243278503418, "global_step": 114266, "epoch": 1283, "lr": 6.245246224151822e-05}
{"train_loss": 0.14777985215187073, "global_step": 114267, "epoch": 1283, "lr": 6.245190077110125e-05}
{"train_loss": 0.14454452693462372, "global_step": 114268, "epoch": 1283, "lr": 6.245133929901028e-05}
{"train_loss": 0.1201595887541771, "global_step": 114269, "epoch": 1283, "lr": 6.245077782524539e-05}
{"train_loss": 0.0780087560415268, "global_step": 114270, "epoch": 1283, "lr": 6.245021634980663e-05}
{"train_loss": 0.1867457926273346, "global_step": 114271, "epoch": 1283, "lr": 6.24496548726941e-05}
{"train_loss": 0.13273431360721588, "global_step": 114272, "epoch": 1283, "lr": 6.244909339390787e-05}
{"train_loss": 0.17540502548217773, "global_step": 114273, "epoch": 1283, "lr": 6.244853191344799e-05}
{"train_loss": 0.15905128419399261, "global_step": 114274, "epoch": 1283, "lr": 6.244797043131458e-05}
{"train_loss": 0.14929553640357565, "global_step": 114275, "epoch": 1283, "lr": 6.244740894750767e-05, "val_loss": 5.217708110809326}
{"train_loss": 0.11574582755565643, "global_step": 114276, "epoch": 1284, "lr": 6.244684746202738e-05}
{"train_loss": 0.12982787191867828, "global_step": 114277, "epoch": 1284, "lr": 6.244628597487376e-05}
{"train_loss": 0.14452582597732544, "global_step": 114278, "epoch": 1284, "lr": 6.244572448604688e-05}
{"train_loss": 0.11711009591817856, "global_step": 114279, "epoch": 1284, "lr": 6.244516299554682e-05}
{"train_loss": 0.1090385690331459, "global_step": 114280, "epoch": 1284, "lr": 6.244460150337367e-05}
{"train_loss": 0.1458621323108673, "global_step": 114281, "epoch": 1284, "lr": 6.24440400095275e-05}
{"train_loss": 0.15727083384990692, "global_step": 114282, "epoch": 1284, "lr": 6.244347851400836e-05}
{"train_loss": 0.10066495090723038, "global_step": 114283, "epoch": 1284, "lr": 6.244291701681636e-05}
{"train_loss": 0.07727079838514328, "global_step": 114284, "epoch": 1284, "lr": 6.244235551795154e-05}
{"train_loss": 0.04999212175607681, "global_step": 114285, "epoch": 1284, "lr": 6.244179401741403e-05}
{"train_loss": 0.11802612990140915, "global_step": 114286, "epoch": 1284, "lr": 6.244123251520383e-05}
{"train_loss": 0.11573956161737442, "global_step": 114287, "epoch": 1284, "lr": 6.244067101132111e-05}
{"train_loss": 0.12233109772205353, "global_step": 114288, "epoch": 1284, "lr": 6.244010950576587e-05}
{"train_loss": 0.11083056777715683, "global_step": 114289, "epoch": 1284, "lr": 6.243954799853819e-05}
{"train_loss": 0.14504729211330414, "global_step": 114290, "epoch": 1284, "lr": 6.243898648963819e-05}
{"train_loss": 0.11801697313785553, "global_step": 114291, "epoch": 1284, "lr": 6.243842497906591e-05}
{"train_loss": 0.12981592118740082, "global_step": 114292, "epoch": 1284, "lr": 6.243786346682144e-05}
{"train_loss": 0.17867732048034668, "global_step": 114293, "epoch": 1284, "lr": 6.243730195290484e-05}
{"train_loss": 0.09971515834331512, "global_step": 114294, "epoch": 1284, "lr": 6.24367404373162e-05}
{"train_loss": 0.05417425557971001, "global_step": 114295, "epoch": 1284, "lr": 6.243617892005561e-05}
{"train_loss": 0.12156262248754501, "global_step": 114296, "epoch": 1284, "lr": 6.243561740112313e-05}
{"train_loss": 0.20651254057884216, "global_step": 114297, "epoch": 1284, "lr": 6.243505588051881e-05}
{"train_loss": 0.14676007628440857, "global_step": 114298, "epoch": 1284, "lr": 6.243449435824276e-05}
{"train_loss": 0.1841920167207718, "global_step": 114299, "epoch": 1284, "lr": 6.243393283429503e-05}
{"train_loss": 0.12263261526823044, "global_step": 114300, "epoch": 1284, "lr": 6.243337130867572e-05}
{"train_loss": 0.1291515976190567, "global_step": 114301, "epoch": 1284, "lr": 6.243280978138489e-05}
{"train_loss": 0.08772441744804382, "global_step": 114302, "epoch": 1284, "lr": 6.243224825242261e-05}
{"train_loss": 0.13057461380958557, "global_step": 114303, "epoch": 1284, "lr": 6.243168672178898e-05}
{"train_loss": 0.17010162770748138, "global_step": 114304, "epoch": 1284, "lr": 6.243112518948406e-05}
{"train_loss": 0.1635274589061737, "global_step": 114305, "epoch": 1284, "lr": 6.243056365550793e-05}
{"train_loss": 0.15199100971221924, "global_step": 114306, "epoch": 1284, "lr": 6.243000211986065e-05}
{"train_loss": 0.12186429649591446, "global_step": 114307, "epoch": 1284, "lr": 6.242944058254231e-05}
{"train_loss": 0.15513163805007935, "global_step": 114308, "epoch": 1284, "lr": 6.2428879043553e-05}
{"train_loss": 0.16515128314495087, "global_step": 114309, "epoch": 1284, "lr": 6.242831750289277e-05}
{"train_loss": 0.10807398706674576, "global_step": 114310, "epoch": 1284, "lr": 6.242775596056169e-05}
{"train_loss": 0.17544177174568176, "global_step": 114311, "epoch": 1284, "lr": 6.242719441655988e-05}
{"train_loss": 0.09337805211544037, "global_step": 114312, "epoch": 1284, "lr": 6.242663287088736e-05}
{"train_loss": 0.136933833360672, "global_step": 114313, "epoch": 1284, "lr": 6.242607132354423e-05}
{"train_loss": 0.11777281016111374, "global_step": 114314, "epoch": 1284, "lr": 6.242550977453058e-05}
{"train_loss": 0.1658349484205246, "global_step": 114315, "epoch": 1284, "lr": 6.242494822384649e-05}
{"train_loss": 0.11789318919181824, "global_step": 114316, "epoch": 1284, "lr": 6.242438667149198e-05}
{"train_loss": 0.11706375330686569, "global_step": 114317, "epoch": 1284, "lr": 6.242382511746719e-05}
{"train_loss": 0.18343202769756317, "global_step": 114318, "epoch": 1284, "lr": 6.242326356177213e-05}
{"train_loss": 0.10082076489925385, "global_step": 114319, "epoch": 1284, "lr": 6.242270200440695e-05}
{"train_loss": 0.11224188655614853, "global_step": 114320, "epoch": 1284, "lr": 6.242214044537167e-05}
{"train_loss": 0.10509812831878662, "global_step": 114321, "epoch": 1284, "lr": 6.242157888466639e-05}
{"train_loss": 0.13809847831726074, "global_step": 114322, "epoch": 1284, "lr": 6.242101732229118e-05}
{"train_loss": 0.133797749876976, "global_step": 114323, "epoch": 1284, "lr": 6.242045575824611e-05}
{"train_loss": 0.13418249785900116, "global_step": 114324, "epoch": 1284, "lr": 6.241989419253127e-05}
{"train_loss": 0.173100084066391, "global_step": 114325, "epoch": 1284, "lr": 6.241933262514672e-05}
{"train_loss": 0.17236021161079407, "global_step": 114326, "epoch": 1284, "lr": 6.241877105609255e-05}
{"train_loss": 0.2035272717475891, "global_step": 114327, "epoch": 1284, "lr": 6.241820948536883e-05}
{"train_loss": 0.16947227716445923, "global_step": 114328, "epoch": 1284, "lr": 6.241764791297561e-05}
{"train_loss": 0.16483929753303528, "global_step": 114329, "epoch": 1284, "lr": 6.241708633891302e-05}
{"train_loss": 0.08437007665634155, "global_step": 114330, "epoch": 1284, "lr": 6.241652476318109e-05}
{"train_loss": 0.09756988286972046, "global_step": 114331, "epoch": 1284, "lr": 6.241596318577992e-05}
{"train_loss": 0.186641126871109, "global_step": 114332, "epoch": 1284, "lr": 6.241540160670956e-05}
{"train_loss": 0.13642092049121857, "global_step": 114333, "epoch": 1284, "lr": 6.24148400259701e-05}
{"train_loss": 0.12261106818914413, "global_step": 114334, "epoch": 1284, "lr": 6.241427844356163e-05}
{"train_loss": 0.11765587329864502, "global_step": 114335, "epoch": 1284, "lr": 6.241371685948421e-05}
{"train_loss": 0.0955594927072525, "global_step": 114336, "epoch": 1284, "lr": 6.24131552737379e-05}
{"train_loss": 0.14668947458267212, "global_step": 114337, "epoch": 1284, "lr": 6.241259368632281e-05}
{"train_loss": 0.08943688124418259, "global_step": 114338, "epoch": 1284, "lr": 6.241203209723898e-05}
{"train_loss": 0.1153084859251976, "global_step": 114339, "epoch": 1284, "lr": 6.241147050648651e-05}
{"train_loss": 0.0952327772974968, "global_step": 114340, "epoch": 1284, "lr": 6.241090891406547e-05}
{"train_loss": 0.18132402002811432, "global_step": 114341, "epoch": 1284, "lr": 6.241034731997594e-05}
{"train_loss": 0.16651275753974915, "global_step": 114342, "epoch": 1284, "lr": 6.240978572421798e-05}
{"train_loss": 0.17474760115146637, "global_step": 114343, "epoch": 1284, "lr": 6.240922412679169e-05}
{"train_loss": 0.13125495612621307, "global_step": 114344, "epoch": 1284, "lr": 6.240866252769712e-05}
{"train_loss": 0.1109650507569313, "global_step": 114345, "epoch": 1284, "lr": 6.240810092693436e-05}
{"train_loss": 0.10923387855291367, "global_step": 114346, "epoch": 1284, "lr": 6.24075393245035e-05}
{"train_loss": 0.17362704873085022, "global_step": 114347, "epoch": 1284, "lr": 6.240697772040457e-05}
{"train_loss": 0.08072245121002197, "global_step": 114348, "epoch": 1284, "lr": 6.240641611463769e-05}
{"train_loss": 0.1331661343574524, "global_step": 114349, "epoch": 1284, "lr": 6.240585450720291e-05}
{"train_loss": 0.09434068202972412, "global_step": 114350, "epoch": 1284, "lr": 6.240529289810031e-05}
{"train_loss": 0.07959141582250595, "global_step": 114351, "epoch": 1284, "lr": 6.240473128732997e-05}
{"train_loss": 0.10533653944730759, "global_step": 114352, "epoch": 1284, "lr": 6.240416967489198e-05}
{"train_loss": 0.13813810050487518, "global_step": 114353, "epoch": 1284, "lr": 6.240360806078639e-05}
{"train_loss": 0.1635408103466034, "global_step": 114354, "epoch": 1284, "lr": 6.24030464450133e-05}
{"train_loss": 0.13361907005310059, "global_step": 114355, "epoch": 1284, "lr": 6.240248482757276e-05}
{"train_loss": 0.11928537487983704, "global_step": 114356, "epoch": 1284, "lr": 6.240192320846485e-05}
{"train_loss": 0.14742375910282135, "global_step": 114357, "epoch": 1284, "lr": 6.240136158768965e-05}
{"train_loss": 0.10809322446584702, "global_step": 114358, "epoch": 1284, "lr": 6.240079996524724e-05}
{"train_loss": 0.09014317393302917, "global_step": 114359, "epoch": 1284, "lr": 6.24002383411377e-05}
{"train_loss": 0.16544659435749054, "global_step": 114360, "epoch": 1284, "lr": 6.239967671536111e-05}
{"train_loss": 0.14807939529418945, "global_step": 114361, "epoch": 1284, "lr": 6.239911508791753e-05}
{"train_loss": 0.15387828648090363, "global_step": 114362, "epoch": 1284, "lr": 6.239855345880703e-05}
{"train_loss": 0.20802903175354004, "global_step": 114363, "epoch": 1284, "lr": 6.23979918280297e-05}
{"train_loss": 0.1321643399639746, "global_step": 114364, "epoch": 1284, "lr": 6.239743019558562e-05, "val_loss": 5.347290992736816}
{"train_loss": 0.13187357783317566, "global_step": 114365, "epoch": 1285, "lr": 6.239686856147484e-05}
{"train_loss": 0.06920666992664337, "global_step": 114366, "epoch": 1285, "lr": 6.239630692569745e-05}
{"train_loss": 0.09928617626428604, "global_step": 114367, "epoch": 1285, "lr": 6.239574528825354e-05}
{"train_loss": 0.19925764203071594, "global_step": 114368, "epoch": 1285, "lr": 6.239518364914318e-05}
{"train_loss": 0.1899348646402359, "global_step": 114369, "epoch": 1285, "lr": 6.239462200836642e-05}
{"train_loss": 0.18825297057628632, "global_step": 114370, "epoch": 1285, "lr": 6.239406036592337e-05}
{"train_loss": 0.24499677121639252, "global_step": 114371, "epoch": 1285, "lr": 6.239349872181409e-05}
{"train_loss": 0.07124573737382889, "global_step": 114372, "epoch": 1285, "lr": 6.239293707603864e-05}
{"train_loss": 0.1749318242073059, "global_step": 114373, "epoch": 1285, "lr": 6.239237542859713e-05}
{"train_loss": 0.13272352516651154, "global_step": 114374, "epoch": 1285, "lr": 6.23918137794896e-05}
{"train_loss": 0.12656795978546143, "global_step": 114375, "epoch": 1285, "lr": 6.239125212871613e-05}
{"train_loss": 0.15730595588684082, "global_step": 114376, "epoch": 1285, "lr": 6.239069047627684e-05}
{"train_loss": 0.19653435051441193, "global_step": 114377, "epoch": 1285, "lr": 6.239012882217175e-05}
{"train_loss": 0.09477386623620987, "global_step": 114378, "epoch": 1285, "lr": 6.238956716640096e-05}
{"train_loss": 0.1045733317732811, "global_step": 114379, "epoch": 1285, "lr": 6.238900550896454e-05}
{"train_loss": 0.15278039872646332, "global_step": 114380, "epoch": 1285, "lr": 6.238844384986258e-05}
{"train_loss": 0.11207381635904312, "global_step": 114381, "epoch": 1285, "lr": 6.238788218909514e-05}
{"train_loss": 0.15163667500019073, "global_step": 114382, "epoch": 1285, "lr": 6.238732052666232e-05}
{"train_loss": 0.21195363998413086, "global_step": 114383, "epoch": 1285, "lr": 6.238675886256416e-05}
{"train_loss": 0.0730060338973999, "global_step": 114384, "epoch": 1285, "lr": 6.238619719680074e-05}
{"train_loss": 0.1396966129541397, "global_step": 114385, "epoch": 1285, "lr": 6.238563552937215e-05}
{"train_loss": 0.11179269105195999, "global_step": 114386, "epoch": 1285, "lr": 6.238507386027848e-05}
{"train_loss": 0.17147254943847656, "global_step": 114387, "epoch": 1285, "lr": 6.238451218951978e-05}
{"train_loss": 0.15763267874717712, "global_step": 114388, "epoch": 1285, "lr": 6.238395051709612e-05}
{"train_loss": 0.13764794170856476, "global_step": 114389, "epoch": 1285, "lr": 6.23833888430076e-05}
{"train_loss": 0.16289706528186798, "global_step": 114390, "epoch": 1285, "lr": 6.238282716725429e-05}
{"train_loss": 0.13147711753845215, "global_step": 114391, "epoch": 1285, "lr": 6.238226548983624e-05}
{"train_loss": 0.12478647381067276, "global_step": 114392, "epoch": 1285, "lr": 6.238170381075357e-05}
{"train_loss": 0.13885042071342468, "global_step": 114393, "epoch": 1285, "lr": 6.23811421300063e-05}
{"train_loss": 0.15152260661125183, "global_step": 114394, "epoch": 1285, "lr": 6.238058044759457e-05}
{"train_loss": 0.20159068703651428, "global_step": 114395, "epoch": 1285, "lr": 6.23800187635184e-05}
{"train_loss": 0.09676764160394669, "global_step": 114396, "epoch": 1285, "lr": 6.237945707777791e-05}
{"train_loss": 0.11599838733673096, "global_step": 114397, "epoch": 1285, "lr": 6.237889539037312e-05}
{"train_loss": 0.14711898565292358, "global_step": 114398, "epoch": 1285, "lr": 6.237833370130416e-05}
{"train_loss": 0.18960310518741608, "global_step": 114399, "epoch": 1285, "lr": 6.237777201057107e-05}
{"train_loss": 0.07176295667886734, "global_step": 114400, "epoch": 1285, "lr": 6.237721031817396e-05}
{"train_loss": 0.17151540517807007, "global_step": 114401, "epoch": 1285, "lr": 6.237664862411288e-05}
{"train_loss": 0.16359958052635193, "global_step": 114402, "epoch": 1285, "lr": 6.23760869283879e-05}
{"train_loss": 0.19121153652668, "global_step": 114403, "epoch": 1285, "lr": 6.237552523099911e-05}
{"train_loss": 0.18619166314601898, "global_step": 114404, "epoch": 1285, "lr": 6.237496353194658e-05}
{"train_loss": 0.1358536034822464, "global_step": 114405, "epoch": 1285, "lr": 6.23744018312304e-05}
{"train_loss": 0.1367071121931076, "global_step": 114406, "epoch": 1285, "lr": 6.237384012885061e-05}
{"train_loss": 0.08766156435012817, "global_step": 114407, "epoch": 1285, "lr": 6.237327842480732e-05}
{"train_loss": 0.10849504172801971, "global_step": 114408, "epoch": 1285, "lr": 6.237271671910059e-05}
{"train_loss": 0.12208858132362366, "global_step": 114409, "epoch": 1285, "lr": 6.237215501173051e-05}
{"train_loss": 0.11216122657060623, "global_step": 114410, "epoch": 1285, "lr": 6.237159330269712e-05}
{"train_loss": 0.07616662979125977, "global_step": 114411, "epoch": 1285, "lr": 6.237103159200056e-05}
{"train_loss": 0.15714389085769653, "global_step": 114412, "epoch": 1285, "lr": 6.237046987964083e-05}
{"train_loss": 0.08997362852096558, "global_step": 114413, "epoch": 1285, "lr": 6.236990816561805e-05}
{"train_loss": 0.16334718465805054, "global_step": 114414, "epoch": 1285, "lr": 6.236934644993229e-05}
{"train_loss": 0.09199060499668121, "global_step": 114415, "epoch": 1285, "lr": 6.236878473258361e-05}
{"train_loss": 0.12597492337226868, "global_step": 114416, "epoch": 1285, "lr": 6.236822301357212e-05}
{"train_loss": 0.11957880109548569, "global_step": 114417, "epoch": 1285, "lr": 6.236766129289786e-05}
{"train_loss": 0.16611887514591217, "global_step": 114418, "epoch": 1285, "lr": 6.236709957056092e-05}
{"train_loss": 0.17405717074871063, "global_step": 114419, "epoch": 1285, "lr": 6.236653784656137e-05}
{"train_loss": 0.14103440940380096, "global_step": 114420, "epoch": 1285, "lr": 6.236597612089931e-05}
{"train_loss": 0.1884247064590454, "global_step": 114421, "epoch": 1285, "lr": 6.236541439357477e-05}
{"train_loss": 0.1582992672920227, "global_step": 114422, "epoch": 1285, "lr": 6.236485266458786e-05}
{"train_loss": 0.1172841489315033, "global_step": 114423, "epoch": 1285, "lr": 6.236429093393866e-05}
{"train_loss": 0.20572982728481293, "global_step": 114424, "epoch": 1285, "lr": 6.236372920162721e-05}
{"train_loss": 0.16414760053157806, "global_step": 114425, "epoch": 1285, "lr": 6.236316746765362e-05}
{"train_loss": 0.17375731468200684, "global_step": 114426, "epoch": 1285, "lr": 6.236260573201795e-05}
{"train_loss": 0.11357969790697098, "global_step": 114427, "epoch": 1285, "lr": 6.236204399472027e-05}
{"train_loss": 0.20709800720214844, "global_step": 114428, "epoch": 1285, "lr": 6.236148225576067e-05}
{"train_loss": 0.19832035899162292, "global_step": 114429, "epoch": 1285, "lr": 6.236092051513923e-05}
{"train_loss": 0.16264916956424713, "global_step": 114430, "epoch": 1285, "lr": 6.2360358772856e-05}
{"train_loss": 0.1574941724538803, "global_step": 114431, "epoch": 1285, "lr": 6.235979702891108e-05}
{"train_loss": 0.08883850276470184, "global_step": 114432, "epoch": 1285, "lr": 6.235923528330453e-05}
{"train_loss": 0.13878338038921356, "global_step": 114433, "epoch": 1285, "lr": 6.235867353603643e-05}
{"train_loss": 0.13988424837589264, "global_step": 114434, "epoch": 1285, "lr": 6.235811178710686e-05}
{"train_loss": 0.0995531752705574, "global_step": 114435, "epoch": 1285, "lr": 6.23575500365159e-05}
{"train_loss": 0.18436847627162933, "global_step": 114436, "epoch": 1285, "lr": 6.235698828426361e-05}
{"train_loss": 0.1102156788110733, "global_step": 114437, "epoch": 1285, "lr": 6.235642653035007e-05}
{"train_loss": 0.10543008893728256, "global_step": 114438, "epoch": 1285, "lr": 6.235586477477536e-05}
{"train_loss": 0.0995568260550499, "global_step": 114439, "epoch": 1285, "lr": 6.235530301753956e-05}
{"train_loss": 0.16722643375396729, "global_step": 114440, "epoch": 1285, "lr": 6.235474125864274e-05}
{"train_loss": 0.12459982186555862, "global_step": 114441, "epoch": 1285, "lr": 6.235417949808498e-05}
{"train_loss": 0.17827287316322327, "global_step": 114442, "epoch": 1285, "lr": 6.235361773586634e-05}
{"train_loss": 0.11632904410362244, "global_step": 114443, "epoch": 1285, "lr": 6.235305597198692e-05}
{"train_loss": 0.13056129217147827, "global_step": 114444, "epoch": 1285, "lr": 6.235249420644676e-05}
{"train_loss": 0.13223809003829956, "global_step": 114445, "epoch": 1285, "lr": 6.235193243924597e-05}
{"train_loss": 0.1288660764694214, "global_step": 114446, "epoch": 1285, "lr": 6.235137067038462e-05}
{"train_loss": 0.11608774214982986, "global_step": 114447, "epoch": 1285, "lr": 6.235080889986278e-05}
{"train_loss": 0.12590579688549042, "global_step": 114448, "epoch": 1285, "lr": 6.235024712768051e-05}
{"train_loss": 0.14000964164733887, "global_step": 114449, "epoch": 1285, "lr": 6.234968535383791e-05}
{"train_loss": 0.13796387612819672, "global_step": 114450, "epoch": 1285, "lr": 6.234912357833505e-05}
{"train_loss": 0.1255531758069992, "global_step": 114451, "epoch": 1285, "lr": 6.2348561801172e-05}
{"train_loss": 0.15964852273464203, "global_step": 114452, "epoch": 1285, "lr": 6.234800002234882e-05}
{"train_loss": 0.14152078338888255, "global_step": 114453, "epoch": 1285, "lr": 6.234743824186561e-05, "val_loss": 5.436328887939453, "train_action_mse_error": 44.72978591918945}
{"train_loss": 0.1459864377975464, "global_step": 114454, "epoch": 1286, "lr": 6.234687645972244e-05}
{"train_loss": 0.10654241591691971, "global_step": 114455, "epoch": 1286, "lr": 6.234631467591938e-05}
{"train_loss": 0.15954551100730896, "global_step": 114456, "epoch": 1286, "lr": 6.234575289045652e-05}
{"train_loss": 0.1600964218378067, "global_step": 114457, "epoch": 1286, "lr": 6.234519110333391e-05}
{"train_loss": 0.14577458798885345, "global_step": 114458, "epoch": 1286, "lr": 6.234462931455165e-05}
{"train_loss": 0.14357997477054596, "global_step": 114459, "epoch": 1286, "lr": 6.23440675241098e-05}
{"train_loss": 0.07927192747592926, "global_step": 114460, "epoch": 1286, "lr": 6.234350573200843e-05}
{"train_loss": 0.15153270959854126, "global_step": 114461, "epoch": 1286, "lr": 6.234294393824764e-05}
{"train_loss": 0.17290988564491272, "global_step": 114462, "epoch": 1286, "lr": 6.23423821428275e-05}
{"train_loss": 0.17170651257038116, "global_step": 114463, "epoch": 1286, "lr": 6.234182034574806e-05}
{"train_loss": 0.18384826183319092, "global_step": 114464, "epoch": 1286, "lr": 6.234125854700941e-05}
{"train_loss": 0.17535780370235443, "global_step": 114465, "epoch": 1286, "lr": 6.234069674661163e-05}
{"train_loss": 0.12922725081443787, "global_step": 114466, "epoch": 1286, "lr": 6.23401349445548e-05}
{"train_loss": 0.07248517870903015, "global_step": 114467, "epoch": 1286, "lr": 6.233957314083899e-05}
{"train_loss": 0.14036928117275238, "global_step": 114468, "epoch": 1286, "lr": 6.233901133546428e-05}
{"train_loss": 0.13479021191596985, "global_step": 114469, "epoch": 1286, "lr": 6.233844952843073e-05}
{"train_loss": 0.1564090996980667, "global_step": 114470, "epoch": 1286, "lr": 6.233788771973843e-05}
{"train_loss": 0.13796484470367432, "global_step": 114471, "epoch": 1286, "lr": 6.233732590938746e-05}
{"train_loss": 0.11439288407564163, "global_step": 114472, "epoch": 1286, "lr": 6.233676409737788e-05}
{"train_loss": 0.16930067539215088, "global_step": 114473, "epoch": 1286, "lr": 6.233620228370977e-05}
{"train_loss": 0.12173724919557571, "global_step": 114474, "epoch": 1286, "lr": 6.233564046838321e-05}
{"train_loss": 0.14623087644577026, "global_step": 114475, "epoch": 1286, "lr": 6.233507865139827e-05}
{"train_loss": 0.15147192776203156, "global_step": 114476, "epoch": 1286, "lr": 6.233451683275502e-05}
{"train_loss": 0.14381395280361176, "global_step": 114477, "epoch": 1286, "lr": 6.233395501245356e-05}
{"train_loss": 0.10228186100721359, "global_step": 114478, "epoch": 1286, "lr": 6.233339319049395e-05}
{"train_loss": 0.13021789491176605, "global_step": 114479, "epoch": 1286, "lr": 6.233283136687626e-05}
{"train_loss": 0.12014283984899521, "global_step": 114480, "epoch": 1286, "lr": 6.233226954160057e-05}
{"train_loss": 0.1471240520477295, "global_step": 114481, "epoch": 1286, "lr": 6.233170771466695e-05}
{"train_loss": 0.11406690627336502, "global_step": 114482, "epoch": 1286, "lr": 6.23311458860755e-05}
{"train_loss": 0.13057473301887512, "global_step": 114483, "epoch": 1286, "lr": 6.233058405582627e-05}
{"train_loss": 0.19632545113563538, "global_step": 114484, "epoch": 1286, "lr": 6.233002222391934e-05}
{"train_loss": 0.12029372155666351, "global_step": 114485, "epoch": 1286, "lr": 6.232946039035479e-05}
{"train_loss": 0.12431720644235611, "global_step": 114486, "epoch": 1286, "lr": 6.23288985551327e-05}
{"train_loss": 0.10197748243808746, "global_step": 114487, "epoch": 1286, "lr": 6.232833671825314e-05}
{"train_loss": 0.17819128930568695, "global_step": 114488, "epoch": 1286, "lr": 6.232777487971617e-05}
{"train_loss": 0.11700357496738434, "global_step": 114489, "epoch": 1286, "lr": 6.23272130395219e-05}
{"train_loss": 0.1889074295759201, "global_step": 114490, "epoch": 1286, "lr": 6.232665119767037e-05}
{"train_loss": 0.20654164254665375, "global_step": 114491, "epoch": 1286, "lr": 6.232608935416168e-05}
{"train_loss": 0.1514720469713211, "global_step": 114492, "epoch": 1286, "lr": 6.23255275089959e-05}
{"train_loss": 0.12317278236150742, "global_step": 114493, "epoch": 1286, "lr": 6.232496566217309e-05}
{"train_loss": 0.1874556690454483, "global_step": 114494, "epoch": 1286, "lr": 6.232440381369335e-05}
{"train_loss": 0.09885744005441666, "global_step": 114495, "epoch": 1286, "lr": 6.232384196355676e-05}
{"train_loss": 0.12129838019609451, "global_step": 114496, "epoch": 1286, "lr": 6.232328011176337e-05}
{"train_loss": 0.1747157722711563, "global_step": 114497, "epoch": 1286, "lr": 6.232271825831324e-05}
{"train_loss": 0.147873654961586, "global_step": 114498, "epoch": 1286, "lr": 6.232215640320648e-05}
{"train_loss": 0.15384337306022644, "global_step": 114499, "epoch": 1286, "lr": 6.232159454644318e-05}
{"train_loss": 0.1606169193983078, "global_step": 114500, "epoch": 1286, "lr": 6.232103268802338e-05}
{"train_loss": 0.1328074187040329, "global_step": 114501, "epoch": 1286, "lr": 6.232047082794716e-05}
{"train_loss": 0.11249952018260956, "global_step": 114502, "epoch": 1286, "lr": 6.231990896621459e-05}
{"train_loss": 0.1390148252248764, "global_step": 114503, "epoch": 1286, "lr": 6.231934710282578e-05}
{"train_loss": 0.16497857868671417, "global_step": 114504, "epoch": 1286, "lr": 6.231878523778079e-05}
{"train_loss": 0.2317308485507965, "global_step": 114505, "epoch": 1286, "lr": 6.231822337107967e-05}
{"train_loss": 0.19086915254592896, "global_step": 114506, "epoch": 1286, "lr": 6.231766150272253e-05}
{"train_loss": 0.17470720410346985, "global_step": 114507, "epoch": 1286, "lr": 6.231709963270942e-05}
{"train_loss": 0.2143428474664688, "global_step": 114508, "epoch": 1286, "lr": 6.231653776104044e-05}
{"train_loss": 0.14733825623989105, "global_step": 114509, "epoch": 1286, "lr": 6.231597588771564e-05}
{"train_loss": 0.1293841153383255, "global_step": 114510, "epoch": 1286, "lr": 6.231541401273511e-05}
{"train_loss": 0.1550936996936798, "global_step": 114511, "epoch": 1286, "lr": 6.231485213609892e-05}
{"train_loss": 0.1385965794324875, "global_step": 114512, "epoch": 1286, "lr": 6.231429025780716e-05}
{"train_loss": 0.06433067470788956, "global_step": 114513, "epoch": 1286, "lr": 6.231372837785987e-05}
{"train_loss": 0.19532349705696106, "global_step": 114514, "epoch": 1286, "lr": 6.231316649625717e-05}
{"train_loss": 0.16419929265975952, "global_step": 114515, "epoch": 1286, "lr": 6.231260461299911e-05}
{"train_loss": 0.18817691504955292, "global_step": 114516, "epoch": 1286, "lr": 6.231204272808577e-05}
{"train_loss": 0.12889446318149567, "global_step": 114517, "epoch": 1286, "lr": 6.231148084151722e-05}
{"train_loss": 0.10794073343276978, "global_step": 114518, "epoch": 1286, "lr": 6.231091895329355e-05}
{"train_loss": 0.1574644297361374, "global_step": 114519, "epoch": 1286, "lr": 6.231035706341481e-05}
{"train_loss": 0.12788790464401245, "global_step": 114520, "epoch": 1286, "lr": 6.23097951718811e-05}
{"train_loss": 0.1275198608636856, "global_step": 114521, "epoch": 1286, "lr": 6.23092332786925e-05}
{"train_loss": 0.14333383738994598, "global_step": 114522, "epoch": 1286, "lr": 6.230867138384907e-05}
{"train_loss": 0.11543089151382446, "global_step": 114523, "epoch": 1286, "lr": 6.230810948735088e-05}
{"train_loss": 0.23295262455940247, "global_step": 114524, "epoch": 1286, "lr": 6.230754758919804e-05}
{"train_loss": 0.18478454649448395, "global_step": 114525, "epoch": 1286, "lr": 6.230698568939058e-05}
{"train_loss": 0.2011786848306656, "global_step": 114526, "epoch": 1286, "lr": 6.230642378792859e-05}
{"train_loss": 0.11507126688957214, "global_step": 114527, "epoch": 1286, "lr": 6.230586188481216e-05}
{"train_loss": 0.1365712732076645, "global_step": 114528, "epoch": 1286, "lr": 6.230529998004136e-05}
{"train_loss": 0.10838451981544495, "global_step": 114529, "epoch": 1286, "lr": 6.230473807361625e-05}
{"train_loss": 0.11399558186531067, "global_step": 114530, "epoch": 1286, "lr": 6.230417616553692e-05}
{"train_loss": 0.07651980221271515, "global_step": 114531, "epoch": 1286, "lr": 6.230361425580345e-05}
{"train_loss": 0.12847554683685303, "global_step": 114532, "epoch": 1286, "lr": 6.230305234441591e-05}
{"train_loss": 0.13401833176612854, "global_step": 114533, "epoch": 1286, "lr": 6.230249043137437e-05}
{"train_loss": 0.18222685158252716, "global_step": 114534, "epoch": 1286, "lr": 6.230192851667892e-05}
{"train_loss": 0.1282774806022644, "global_step": 114535, "epoch": 1286, "lr": 6.230136660032961e-05}
{"train_loss": 0.11674162745475769, "global_step": 114536, "epoch": 1286, "lr": 6.230080468232655e-05}
{"train_loss": 0.13758282363414764, "global_step": 114537, "epoch": 1286, "lr": 6.230024276266979e-05}
{"train_loss": 0.10988353192806244, "global_step": 114538, "epoch": 1286, "lr": 6.229968084135939e-05}
{"train_loss": 0.15576398372650146, "global_step": 114539, "epoch": 1286, "lr": 6.229911891839546e-05}
{"train_loss": 0.14136391878128052, "global_step": 114540, "epoch": 1286, "lr": 6.229855699377808e-05}
{"train_loss": 0.09225861728191376, "global_step": 114541, "epoch": 1286, "lr": 6.229799506750729e-05}
{"train_loss": 0.1440034258399117, "global_step": 114542, "epoch": 1286, "lr": 6.22974331395832e-05, "val_loss": 5.062535762786865}
{"train_loss": 0.17078474164009094, "global_step": 114543, "epoch": 1287, "lr": 6.229687121000587e-05}
{"train_loss": 0.11193099617958069, "global_step": 114544, "epoch": 1287, "lr": 6.229630927877536e-05}
{"train_loss": 0.15128062665462494, "global_step": 114545, "epoch": 1287, "lr": 6.229574734589178e-05}
{"train_loss": 0.14595560729503632, "global_step": 114546, "epoch": 1287, "lr": 6.229518541135517e-05}
{"train_loss": 0.09856393188238144, "global_step": 114547, "epoch": 1287, "lr": 6.229462347516563e-05}
{"train_loss": 0.15197831392288208, "global_step": 114548, "epoch": 1287, "lr": 6.229406153732322e-05}
{"train_loss": 0.18361519277095795, "global_step": 114549, "epoch": 1287, "lr": 6.229349959782804e-05}
{"train_loss": 0.22528459131717682, "global_step": 114550, "epoch": 1287, "lr": 6.229293765668014e-05}
{"train_loss": 0.16210199892520905, "global_step": 114551, "epoch": 1287, "lr": 6.22923757138796e-05}
{"train_loss": 0.1778830736875534, "global_step": 114552, "epoch": 1287, "lr": 6.22918137694265e-05}
{"train_loss": 0.09606720507144928, "global_step": 114553, "epoch": 1287, "lr": 6.229125182332092e-05}
{"train_loss": 0.1361798495054245, "global_step": 114554, "epoch": 1287, "lr": 6.229068987556293e-05}
{"train_loss": 0.15173569321632385, "global_step": 114555, "epoch": 1287, "lr": 6.229012792615261e-05}
{"train_loss": 0.16196121275424957, "global_step": 114556, "epoch": 1287, "lr": 6.228956597509001e-05}
{"train_loss": 0.18208447098731995, "global_step": 114557, "epoch": 1287, "lr": 6.228900402237525e-05}
{"train_loss": 0.15839873254299164, "global_step": 114558, "epoch": 1287, "lr": 6.228844206800838e-05}
{"train_loss": 0.12417153269052505, "global_step": 114559, "epoch": 1287, "lr": 6.228788011198947e-05}
{"train_loss": 0.14629709720611572, "global_step": 114560, "epoch": 1287, "lr": 6.228731815431861e-05}
{"train_loss": 0.16249892115592957, "global_step": 114561, "epoch": 1287, "lr": 6.228675619499588e-05}
{"train_loss": 0.1351352334022522, "global_step": 114562, "epoch": 1287, "lr": 6.228619423402134e-05}
{"train_loss": 0.161842480301857, "global_step": 114563, "epoch": 1287, "lr": 6.228563227139506e-05}
{"train_loss": 0.16395996510982513, "global_step": 114564, "epoch": 1287, "lr": 6.228507030711715e-05}
{"train_loss": 0.17246393859386444, "global_step": 114565, "epoch": 1287, "lr": 6.228450834118764e-05}
{"train_loss": 0.1144900918006897, "global_step": 114566, "epoch": 1287, "lr": 6.228394637360664e-05}
{"train_loss": 0.17992714047431946, "global_step": 114567, "epoch": 1287, "lr": 6.22833844043742e-05}
{"train_loss": 0.19477878510951996, "global_step": 114568, "epoch": 1287, "lr": 6.228282243349042e-05}
{"train_loss": 0.20476403832435608, "global_step": 114569, "epoch": 1287, "lr": 6.228226046095536e-05}
{"train_loss": 0.12979407608509064, "global_step": 114570, "epoch": 1287, "lr": 6.22816984867691e-05}
{"train_loss": 0.1690538376569748, "global_step": 114571, "epoch": 1287, "lr": 6.228113651093172e-05}
{"train_loss": 0.08527253568172455, "global_step": 114572, "epoch": 1287, "lr": 6.228057453344327e-05}
{"train_loss": 0.16063785552978516, "global_step": 114573, "epoch": 1287, "lr": 6.228001255430386e-05}
{"train_loss": 0.14812767505645752, "global_step": 114574, "epoch": 1287, "lr": 6.227945057351356e-05}
{"train_loss": 0.18134349584579468, "global_step": 114575, "epoch": 1287, "lr": 6.227888859107243e-05}
{"train_loss": 0.09016096591949463, "global_step": 114576, "epoch": 1287, "lr": 6.227832660698056e-05}
{"train_loss": 0.20062881708145142, "global_step": 114577, "epoch": 1287, "lr": 6.2277764621238e-05}
{"train_loss": 0.17892295122146606, "global_step": 114578, "epoch": 1287, "lr": 6.227720263384486e-05}
{"train_loss": 0.08358082920312881, "global_step": 114579, "epoch": 1287, "lr": 6.22766406448012e-05}
{"train_loss": 0.19906572997570038, "global_step": 114580, "epoch": 1287, "lr": 6.22760786541071e-05}
{"train_loss": 0.2324158251285553, "global_step": 114581, "epoch": 1287, "lr": 6.227551666176262e-05}
{"train_loss": 0.19669300317764282, "global_step": 114582, "epoch": 1287, "lr": 6.227495466776785e-05}
{"train_loss": 0.14844979345798492, "global_step": 114583, "epoch": 1287, "lr": 6.227439267212287e-05}
{"train_loss": 0.16131004691123962, "global_step": 114584, "epoch": 1287, "lr": 6.227383067482774e-05}
{"train_loss": 0.06878933310508728, "global_step": 114585, "epoch": 1287, "lr": 6.227326867588253e-05}
{"train_loss": 0.16425369679927826, "global_step": 114586, "epoch": 1287, "lr": 6.227270667528735e-05}
{"train_loss": 0.12318162620067596, "global_step": 114587, "epoch": 1287, "lr": 6.227214467304224e-05}
{"train_loss": 0.18964436650276184, "global_step": 114588, "epoch": 1287, "lr": 6.227158266914729e-05}
{"train_loss": 0.13426204025745392, "global_step": 114589, "epoch": 1287, "lr": 6.227102066360258e-05}
{"train_loss": 0.12719303369522095, "global_step": 114590, "epoch": 1287, "lr": 6.22704586564082e-05}
{"train_loss": 0.172255277633667, "global_step": 114591, "epoch": 1287, "lr": 6.226989664756418e-05}
{"train_loss": 0.09648695588111877, "global_step": 114592, "epoch": 1287, "lr": 6.226933463707063e-05}
{"train_loss": 0.1044543907046318, "global_step": 114593, "epoch": 1287, "lr": 6.226877262492761e-05}
{"train_loss": 0.18928708136081696, "global_step": 114594, "epoch": 1287, "lr": 6.22682106111352e-05}
{"train_loss": 0.19098292291164398, "global_step": 114595, "epoch": 1287, "lr": 6.226764859569349e-05}
{"train_loss": 0.15914352238178253, "global_step": 114596, "epoch": 1287, "lr": 6.226708657860254e-05}
{"train_loss": 0.15998703241348267, "global_step": 114597, "epoch": 1287, "lr": 6.226652455986244e-05}
{"train_loss": 0.20115242898464203, "global_step": 114598, "epoch": 1287, "lr": 6.226596253947324e-05}
{"train_loss": 0.1193503588438034, "global_step": 114599, "epoch": 1287, "lr": 6.226540051743504e-05}
{"train_loss": 0.11204352229833603, "global_step": 114600, "epoch": 1287, "lr": 6.226483849374792e-05}
{"train_loss": 0.16178403794765472, "global_step": 114601, "epoch": 1287, "lr": 6.226427646841192e-05}
{"train_loss": 0.13009870052337646, "global_step": 114602, "epoch": 1287, "lr": 6.226371444142716e-05}
{"train_loss": 0.17370977997779846, "global_step": 114603, "epoch": 1287, "lr": 6.226315241279367e-05}
{"train_loss": 0.1760360449552536, "global_step": 114604, "epoch": 1287, "lr": 6.226259038251155e-05}
{"train_loss": 0.09940848499536514, "global_step": 114605, "epoch": 1287, "lr": 6.226202835058088e-05}
{"train_loss": 0.19031745195388794, "global_step": 114606, "epoch": 1287, "lr": 6.226146631700174e-05}
{"train_loss": 0.2893539369106293, "global_step": 114607, "epoch": 1287, "lr": 6.226090428177418e-05}
{"train_loss": 0.11334136873483658, "global_step": 114608, "epoch": 1287, "lr": 6.226034224489829e-05}
{"train_loss": 0.12773580849170685, "global_step": 114609, "epoch": 1287, "lr": 6.225978020637416e-05}
{"train_loss": 0.1671133190393448, "global_step": 114610, "epoch": 1287, "lr": 6.225921816620183e-05}
{"train_loss": 0.12889106571674347, "global_step": 114611, "epoch": 1287, "lr": 6.225865612438142e-05}
{"train_loss": 0.13371485471725464, "global_step": 114612, "epoch": 1287, "lr": 6.225809408091297e-05}
{"train_loss": 0.11641425639390945, "global_step": 114613, "epoch": 1287, "lr": 6.225753203579657e-05}
{"train_loss": 0.1041468158364296, "global_step": 114614, "epoch": 1287, "lr": 6.22569699890323e-05}
{"train_loss": 0.11142284423112869, "global_step": 114615, "epoch": 1287, "lr": 6.225640794062023e-05}
{"train_loss": 0.14230644702911377, "global_step": 114616, "epoch": 1287, "lr": 6.225584589056044e-05}
{"train_loss": 0.1956544816493988, "global_step": 114617, "epoch": 1287, "lr": 6.225528383885298e-05}
{"train_loss": 0.150112584233284, "global_step": 114618, "epoch": 1287, "lr": 6.225472178549796e-05}
{"train_loss": 0.1266687512397766, "global_step": 114619, "epoch": 1287, "lr": 6.225415973049546e-05}
{"train_loss": 0.15206219255924225, "global_step": 114620, "epoch": 1287, "lr": 6.225359767384552e-05}
{"train_loss": 0.17476913332939148, "global_step": 114621, "epoch": 1287, "lr": 6.225303561554823e-05}
{"train_loss": 0.12029004096984863, "global_step": 114622, "epoch": 1287, "lr": 6.225247355560368e-05}
{"train_loss": 0.1536199003458023, "global_step": 114623, "epoch": 1287, "lr": 6.225191149401193e-05}
{"train_loss": 0.19214951992034912, "global_step": 114624, "epoch": 1287, "lr": 6.225134943077305e-05}
{"train_loss": 0.17602099478244781, "global_step": 114625, "epoch": 1287, "lr": 6.225078736588712e-05}
{"train_loss": 0.11584692448377609, "global_step": 114626, "epoch": 1287, "lr": 6.225022529935424e-05}
{"train_loss": 0.14536631107330322, "global_step": 114627, "epoch": 1287, "lr": 6.224966323117445e-05}
{"train_loss": 0.17096789181232452, "global_step": 114628, "epoch": 1287, "lr": 6.224910116134787e-05}
{"train_loss": 0.1472533792257309, "global_step": 114629, "epoch": 1287, "lr": 6.224853908987451e-05}
{"train_loss": 0.0730995312333107, "global_step": 114630, "epoch": 1287, "lr": 6.22479770167545e-05}
{"train_loss": 0.1519655828060729, "global_step": 114631, "epoch": 1287, "lr": 6.22474149419879e-05, "val_loss": 5.017487525939941}
{"train_loss": 0.16891659796237946, "global_step": 114632, "epoch": 1288, "lr": 6.22468528655748e-05}
{"train_loss": 0.16161401569843292, "global_step": 114633, "epoch": 1288, "lr": 6.224629078751524e-05}
{"train_loss": 0.1020515039563179, "global_step": 114634, "epoch": 1288, "lr": 6.224572870780932e-05}
{"train_loss": 0.1536301225423813, "global_step": 114635, "epoch": 1288, "lr": 6.224516662645711e-05}
{"train_loss": 0.09879831969738007, "global_step": 114636, "epoch": 1288, "lr": 6.224460454345868e-05}
{"train_loss": 0.13852737843990326, "global_step": 114637, "epoch": 1288, "lr": 6.224404245881412e-05}
{"train_loss": 0.09376240521669388, "global_step": 114638, "epoch": 1288, "lr": 6.22434803725235e-05}
{"train_loss": 0.15998324751853943, "global_step": 114639, "epoch": 1288, "lr": 6.22429182845869e-05}
{"train_loss": 0.14713223278522491, "global_step": 114640, "epoch": 1288, "lr": 6.224235619500438e-05}
{"train_loss": 0.1609882265329361, "global_step": 114641, "epoch": 1288, "lr": 6.224179410377602e-05}
{"train_loss": 0.09418074041604996, "global_step": 114642, "epoch": 1288, "lr": 6.224123201090192e-05}
{"train_loss": 0.13166798651218414, "global_step": 114643, "epoch": 1288, "lr": 6.22406699163821e-05}
{"train_loss": 0.12120918184518814, "global_step": 114644, "epoch": 1288, "lr": 6.22401078202167e-05}
{"train_loss": 0.13705669343471527, "global_step": 114645, "epoch": 1288, "lr": 6.223954572240577e-05}
{"train_loss": 0.14013029634952545, "global_step": 114646, "epoch": 1288, "lr": 6.223898362294936e-05}
{"train_loss": 0.20658323168754578, "global_step": 114647, "epoch": 1288, "lr": 6.223842152184757e-05}
{"train_loss": 0.11948491632938385, "global_step": 114648, "epoch": 1288, "lr": 6.223785941910048e-05}
{"train_loss": 0.10865077376365662, "global_step": 114649, "epoch": 1288, "lr": 6.223729731470817e-05}
{"train_loss": 0.09144344180822372, "global_step": 114650, "epoch": 1288, "lr": 6.223673520867068e-05}
{"train_loss": 0.16131402552127838, "global_step": 114651, "epoch": 1288, "lr": 6.223617310098812e-05}
{"train_loss": 0.1673806607723236, "global_step": 114652, "epoch": 1288, "lr": 6.223561099166058e-05}
{"train_loss": 0.1442309319972992, "global_step": 114653, "epoch": 1288, "lr": 6.223504888068809e-05}
{"train_loss": 0.14393025636672974, "global_step": 114654, "epoch": 1288, "lr": 6.223448676807074e-05}
{"train_loss": 0.10250595957040787, "global_step": 114655, "epoch": 1288, "lr": 6.223392465380862e-05}
{"train_loss": 0.09744511544704437, "global_step": 114656, "epoch": 1288, "lr": 6.223336253790181e-05}
{"train_loss": 0.11694850772619247, "global_step": 114657, "epoch": 1288, "lr": 6.223280042035036e-05}
{"train_loss": 0.18480181694030762, "global_step": 114658, "epoch": 1288, "lr": 6.223223830115437e-05}
{"train_loss": 0.12834210693836212, "global_step": 114659, "epoch": 1288, "lr": 6.22316761803139e-05}
{"train_loss": 0.20276618003845215, "global_step": 114660, "epoch": 1288, "lr": 6.223111405782903e-05}
{"train_loss": 0.17662487924098969, "global_step": 114661, "epoch": 1288, "lr": 6.223055193369983e-05}
{"train_loss": 0.11774005740880966, "global_step": 114662, "epoch": 1288, "lr": 6.222998980792638e-05}
{"train_loss": 0.10768405348062515, "global_step": 114663, "epoch": 1288, "lr": 6.222942768050875e-05}
{"train_loss": 0.06615354120731354, "global_step": 114664, "epoch": 1288, "lr": 6.222886555144703e-05}
{"train_loss": 0.1447473019361496, "global_step": 114665, "epoch": 1288, "lr": 6.222830342074128e-05}
{"train_loss": 0.14164790511131287, "global_step": 114666, "epoch": 1288, "lr": 6.22277412883916e-05}
{"train_loss": 0.16775201261043549, "global_step": 114667, "epoch": 1288, "lr": 6.222717915439805e-05}
{"train_loss": 0.1719810515642166, "global_step": 114668, "epoch": 1288, "lr": 6.22266170187607e-05}
{"train_loss": 0.22968387603759766, "global_step": 114669, "epoch": 1288, "lr": 6.222605488147962e-05}
{"train_loss": 0.11236122250556946, "global_step": 114670, "epoch": 1288, "lr": 6.222549274255491e-05}
{"train_loss": 0.11608335375785828, "global_step": 114671, "epoch": 1288, "lr": 6.222493060198663e-05}
{"train_loss": 0.12027733027935028, "global_step": 114672, "epoch": 1288, "lr": 6.222436845977485e-05}
{"train_loss": 0.20488229393959045, "global_step": 114673, "epoch": 1288, "lr": 6.222380631591965e-05}
{"train_loss": 0.2895627021789551, "global_step": 114674, "epoch": 1288, "lr": 6.222324417042112e-05}
{"train_loss": 0.13272801041603088, "global_step": 114675, "epoch": 1288, "lr": 6.222268202327933e-05}
{"train_loss": 0.16102059185504913, "global_step": 114676, "epoch": 1288, "lr": 6.222211987449434e-05}
{"train_loss": 0.12915274500846863, "global_step": 114677, "epoch": 1288, "lr": 6.222155772406624e-05}
{"train_loss": 0.1670161336660385, "global_step": 114678, "epoch": 1288, "lr": 6.22209955719951e-05}
{"train_loss": 0.19547758996486664, "global_step": 114679, "epoch": 1288, "lr": 6.222043341828097e-05}
{"train_loss": 0.11745905131101608, "global_step": 114680, "epoch": 1288, "lr": 6.221987126292398e-05}
{"train_loss": 0.10376658290624619, "global_step": 114681, "epoch": 1288, "lr": 6.221930910592416e-05}
{"train_loss": 0.07760999351739883, "global_step": 114682, "epoch": 1288, "lr": 6.221874694728161e-05}
{"train_loss": 0.09617628902196884, "global_step": 114683, "epoch": 1288, "lr": 6.22181847869964e-05}
{"train_loss": 0.11578210443258286, "global_step": 114684, "epoch": 1288, "lr": 6.22176226250686e-05}
{"train_loss": 0.20562073588371277, "global_step": 114685, "epoch": 1288, "lr": 6.221706046149829e-05}
{"train_loss": 0.1801346093416214, "global_step": 114686, "epoch": 1288, "lr": 6.221649829628556e-05}
{"train_loss": 0.18408404290676117, "global_step": 114687, "epoch": 1288, "lr": 6.221593612943044e-05}
{"train_loss": 0.16714900732040405, "global_step": 114688, "epoch": 1288, "lr": 6.221537396093306e-05}
{"train_loss": 0.11010729521512985, "global_step": 114689, "epoch": 1288, "lr": 6.221481179079347e-05}
{"train_loss": 0.13924425840377808, "global_step": 114690, "epoch": 1288, "lr": 6.221424961901174e-05}
{"train_loss": 0.11729256063699722, "global_step": 114691, "epoch": 1288, "lr": 6.221368744558796e-05}
{"train_loss": 0.17707428336143494, "global_step": 114692, "epoch": 1288, "lr": 6.22131252705222e-05}
{"train_loss": 0.16855397820472717, "global_step": 114693, "epoch": 1288, "lr": 6.221256309381453e-05}
{"train_loss": 0.15111777186393738, "global_step": 114694, "epoch": 1288, "lr": 6.221200091546502e-05}
{"train_loss": 0.09947928786277771, "global_step": 114695, "epoch": 1288, "lr": 6.221143873547376e-05}
{"train_loss": 0.17179647088050842, "global_step": 114696, "epoch": 1288, "lr": 6.221087655384084e-05}
{"train_loss": 0.2078338861465454, "global_step": 114697, "epoch": 1288, "lr": 6.22103143705663e-05}
{"train_loss": 0.09285672008991241, "global_step": 114698, "epoch": 1288, "lr": 6.220975218565023e-05}
{"train_loss": 0.1133611723780632, "global_step": 114699, "epoch": 1288, "lr": 6.220918999909273e-05}
{"train_loss": 0.1632545441389084, "global_step": 114700, "epoch": 1288, "lr": 6.220862781089383e-05}
{"train_loss": 0.09077642112970352, "global_step": 114701, "epoch": 1288, "lr": 6.220806562105363e-05}
{"train_loss": 0.10313478857278824, "global_step": 114702, "epoch": 1288, "lr": 6.22075034295722e-05}
{"train_loss": 0.11783835291862488, "global_step": 114703, "epoch": 1288, "lr": 6.220694123644963e-05}
{"train_loss": 0.15781374275684357, "global_step": 114704, "epoch": 1288, "lr": 6.220637904168598e-05}
{"train_loss": 0.0973689928650856, "global_step": 114705, "epoch": 1288, "lr": 6.220581684528133e-05}
{"train_loss": 0.1429479867219925, "global_step": 114706, "epoch": 1288, "lr": 6.220525464723577e-05}
{"train_loss": 0.11018263548612595, "global_step": 114707, "epoch": 1288, "lr": 6.220469244754936e-05}
{"train_loss": 0.13037125766277313, "global_step": 114708, "epoch": 1288, "lr": 6.220413024622218e-05}
{"train_loss": 0.10022424161434174, "global_step": 114709, "epoch": 1288, "lr": 6.22035680432543e-05}
{"train_loss": 0.2070481926202774, "global_step": 114710, "epoch": 1288, "lr": 6.220300583864578e-05}
{"train_loss": 0.17258039116859436, "global_step": 114711, "epoch": 1288, "lr": 6.220244363239673e-05}
{"train_loss": 0.127385213971138, "global_step": 114712, "epoch": 1288, "lr": 6.22018814245072e-05}
{"train_loss": 0.1182645633816719, "global_step": 114713, "epoch": 1288, "lr": 6.22013192149773e-05}
{"train_loss": 0.07495440542697906, "global_step": 114714, "epoch": 1288, "lr": 6.220075700380706e-05}
{"train_loss": 0.15415078401565552, "global_step": 114715, "epoch": 1288, "lr": 6.220019479099659e-05}
{"train_loss": 0.18475162982940674, "global_step": 114716, "epoch": 1288, "lr": 6.219963257654595e-05}
{"train_loss": 0.18312986195087433, "global_step": 114717, "epoch": 1288, "lr": 6.219907036045522e-05}
{"train_loss": 0.1448020190000534, "global_step": 114718, "epoch": 1288, "lr": 6.219850814272445e-05}
{"train_loss": 0.20930586755275726, "global_step": 114719, "epoch": 1288, "lr": 6.219794592335377e-05}
{"train_loss": 0.14289947523829644, "global_step": 114720, "epoch": 1288, "lr": 6.219738370234319e-05, "val_loss": 5.036830425262451}
{"train_loss": 0.13885411620140076, "global_step": 114721, "epoch": 1289, "lr": 6.219682147969285e-05}
{"train_loss": 0.1380913257598877, "global_step": 114722, "epoch": 1289, "lr": 6.219625925540279e-05}
{"train_loss": 0.1495000123977661, "global_step": 114723, "epoch": 1289, "lr": 6.219569702947309e-05}
{"train_loss": 0.12077438831329346, "global_step": 114724, "epoch": 1289, "lr": 6.219513480190382e-05}
{"train_loss": 0.17697477340698242, "global_step": 114725, "epoch": 1289, "lr": 6.219457257269508e-05}
{"train_loss": 0.11459089815616608, "global_step": 114726, "epoch": 1289, "lr": 6.219401034184691e-05}
{"train_loss": 0.1043655276298523, "global_step": 114727, "epoch": 1289, "lr": 6.219344810935941e-05}
{"train_loss": 0.10191035270690918, "global_step": 114728, "epoch": 1289, "lr": 6.219288587523266e-05}
{"train_loss": 0.12230885028839111, "global_step": 114729, "epoch": 1289, "lr": 6.219232363946671e-05}
{"train_loss": 0.1404748260974884, "global_step": 114730, "epoch": 1289, "lr": 6.219176140206165e-05}
{"train_loss": 0.11626376956701279, "global_step": 114731, "epoch": 1289, "lr": 6.219119916301757e-05}
{"train_loss": 0.22532843053340912, "global_step": 114732, "epoch": 1289, "lr": 6.219063692233453e-05}
{"train_loss": 0.12316076457500458, "global_step": 114733, "epoch": 1289, "lr": 6.21900746800126e-05}
{"train_loss": 0.15717698633670807, "global_step": 114734, "epoch": 1289, "lr": 6.218951243605187e-05}
{"train_loss": 0.09598847478628159, "global_step": 114735, "epoch": 1289, "lr": 6.218895019045241e-05}
{"train_loss": 0.11879429966211319, "global_step": 114736, "epoch": 1289, "lr": 6.218838794321429e-05}
{"train_loss": 0.07776336371898651, "global_step": 114737, "epoch": 1289, "lr": 6.218782569433759e-05}
{"train_loss": 0.11118151992559433, "global_step": 114738, "epoch": 1289, "lr": 6.218726344382237e-05}
{"train_loss": 0.136116623878479, "global_step": 114739, "epoch": 1289, "lr": 6.218670119166874e-05}
{"train_loss": 0.16117769479751587, "global_step": 114740, "epoch": 1289, "lr": 6.218613893787675e-05}
{"train_loss": 0.09793784469366074, "global_step": 114741, "epoch": 1289, "lr": 6.218557668244648e-05}
{"train_loss": 0.10856294631958008, "global_step": 114742, "epoch": 1289, "lr": 6.218501442537801e-05}
{"train_loss": 0.11158420145511627, "global_step": 114743, "epoch": 1289, "lr": 6.218445216667141e-05}
{"train_loss": 0.1663939505815506, "global_step": 114744, "epoch": 1289, "lr": 6.218388990632678e-05}
{"train_loss": 0.14447788894176483, "global_step": 114745, "epoch": 1289, "lr": 6.218332764434415e-05}
{"train_loss": 0.1513771116733551, "global_step": 114746, "epoch": 1289, "lr": 6.218276538072363e-05}
{"train_loss": 0.14916639029979706, "global_step": 114747, "epoch": 1289, "lr": 6.218220311546528e-05}
{"train_loss": 0.15145933628082275, "global_step": 114748, "epoch": 1289, "lr": 6.218164084856919e-05}
{"train_loss": 0.12904798984527588, "global_step": 114749, "epoch": 1289, "lr": 6.218107858003542e-05}
{"train_loss": 0.11559898406267166, "global_step": 114750, "epoch": 1289, "lr": 6.218051630986406e-05}
{"train_loss": 0.15111784636974335, "global_step": 114751, "epoch": 1289, "lr": 6.217995403805517e-05}
{"train_loss": 0.12727110087871552, "global_step": 114752, "epoch": 1289, "lr": 6.217939176460883e-05}
{"train_loss": 0.11648552864789963, "global_step": 114753, "epoch": 1289, "lr": 6.217882948952512e-05}
{"train_loss": 0.11802434921264648, "global_step": 114754, "epoch": 1289, "lr": 6.217826721280413e-05}
{"train_loss": 0.12996141612529755, "global_step": 114755, "epoch": 1289, "lr": 6.217770493444591e-05}
{"train_loss": 0.09081283211708069, "global_step": 114756, "epoch": 1289, "lr": 6.217714265445053e-05}
{"train_loss": 0.11118191480636597, "global_step": 114757, "epoch": 1289, "lr": 6.21765803728181e-05}
{"train_loss": 0.19742953777313232, "global_step": 114758, "epoch": 1289, "lr": 6.217601808954868e-05}
{"train_loss": 0.1170431524515152, "global_step": 114759, "epoch": 1289, "lr": 6.217545580464234e-05}
{"train_loss": 0.26566481590270996, "global_step": 114760, "epoch": 1289, "lr": 6.217489351809917e-05}
{"train_loss": 0.18153150379657745, "global_step": 114761, "epoch": 1289, "lr": 6.217433122991922e-05}
{"train_loss": 0.12697531282901764, "global_step": 114762, "epoch": 1289, "lr": 6.217376894010257e-05}
{"train_loss": 0.1751362681388855, "global_step": 114763, "epoch": 1289, "lr": 6.217320664864931e-05}
{"train_loss": 0.11824659258127213, "global_step": 114764, "epoch": 1289, "lr": 6.217264435555952e-05}
{"train_loss": 0.1129981130361557, "global_step": 114765, "epoch": 1289, "lr": 6.217208206083325e-05}
{"train_loss": 0.1365986168384552, "global_step": 114766, "epoch": 1289, "lr": 6.217151976447062e-05}
{"train_loss": 0.15416574478149414, "global_step": 114767, "epoch": 1289, "lr": 6.217095746647166e-05}
{"train_loss": 0.11159083247184753, "global_step": 114768, "epoch": 1289, "lr": 6.217039516683646e-05}
{"train_loss": 0.13823619484901428, "global_step": 114769, "epoch": 1289, "lr": 6.21698328655651e-05}
{"train_loss": 0.1669546365737915, "global_step": 114770, "epoch": 1289, "lr": 6.216927056265767e-05}
{"train_loss": 0.10657817125320435, "global_step": 114771, "epoch": 1289, "lr": 6.216870825811422e-05}
{"train_loss": 0.13157568871974945, "global_step": 114772, "epoch": 1289, "lr": 6.216814595193483e-05}
{"train_loss": 0.10888762027025223, "global_step": 114773, "epoch": 1289, "lr": 6.216758364411958e-05}
{"train_loss": 0.12317845225334167, "global_step": 114774, "epoch": 1289, "lr": 6.216702133466855e-05}
{"train_loss": 0.13509075343608856, "global_step": 114775, "epoch": 1289, "lr": 6.216645902358181e-05}
{"train_loss": 0.1747487634420395, "global_step": 114776, "epoch": 1289, "lr": 6.216589671085944e-05}
{"train_loss": 0.20575787127017975, "global_step": 114777, "epoch": 1289, "lr": 6.216533439650152e-05}
{"train_loss": 0.11732681095600128, "global_step": 114778, "epoch": 1289, "lr": 6.216477208050811e-05}
{"train_loss": 0.1579924076795578, "global_step": 114779, "epoch": 1289, "lr": 6.21642097628793e-05}
{"train_loss": 0.11046024411916733, "global_step": 114780, "epoch": 1289, "lr": 6.216364744361515e-05}
{"train_loss": 0.11718571186065674, "global_step": 114781, "epoch": 1289, "lr": 6.216308512271576e-05}
{"train_loss": 0.11038083583116531, "global_step": 114782, "epoch": 1289, "lr": 6.216252280018119e-05}
{"train_loss": 0.1298520267009735, "global_step": 114783, "epoch": 1289, "lr": 6.216196047601151e-05}
{"train_loss": 0.10214004665613174, "global_step": 114784, "epoch": 1289, "lr": 6.21613981502068e-05}
{"train_loss": 0.11024501919746399, "global_step": 114785, "epoch": 1289, "lr": 6.216083582276715e-05}
{"train_loss": 0.20834943652153015, "global_step": 114786, "epoch": 1289, "lr": 6.216027349369262e-05}
{"train_loss": 0.11669272929430008, "global_step": 114787, "epoch": 1289, "lr": 6.215971116298328e-05}
{"train_loss": 0.10560145229101181, "global_step": 114788, "epoch": 1289, "lr": 6.215914883063922e-05}
{"train_loss": 0.15785858035087585, "global_step": 114789, "epoch": 1289, "lr": 6.215858649666052e-05}
{"train_loss": 0.14862960577011108, "global_step": 114790, "epoch": 1289, "lr": 6.215802416104724e-05}
{"train_loss": 0.0936008021235466, "global_step": 114791, "epoch": 1289, "lr": 6.215746182379945e-05}
{"train_loss": 0.17119210958480835, "global_step": 114792, "epoch": 1289, "lr": 6.215689948491725e-05}
{"train_loss": 0.16097451746463776, "global_step": 114793, "epoch": 1289, "lr": 6.215633714440071e-05}
{"train_loss": 0.17751947045326233, "global_step": 114794, "epoch": 1289, "lr": 6.215577480224988e-05}
{"train_loss": 0.0938389003276825, "global_step": 114795, "epoch": 1289, "lr": 6.215521245846487e-05}
{"train_loss": 0.14752939343452454, "global_step": 114796, "epoch": 1289, "lr": 6.215465011304573e-05}
{"train_loss": 0.11651276797056198, "global_step": 114797, "epoch": 1289, "lr": 6.215408776599255e-05}
{"train_loss": 0.09840922057628632, "global_step": 114798, "epoch": 1289, "lr": 6.215352541730541e-05}
{"train_loss": 0.14440183341503143, "global_step": 114799, "epoch": 1289, "lr": 6.215296306698437e-05}
{"train_loss": 0.14458702504634857, "global_step": 114800, "epoch": 1289, "lr": 6.21524007150295e-05}
{"train_loss": 0.12504078447818756, "global_step": 114801, "epoch": 1289, "lr": 6.215183836144089e-05}
{"train_loss": 0.1113002672791481, "global_step": 114802, "epoch": 1289, "lr": 6.215127600621863e-05}
{"train_loss": 0.08635051548480988, "global_step": 114803, "epoch": 1289, "lr": 6.215071364936277e-05}
{"train_loss": 0.11668901890516281, "global_step": 114804, "epoch": 1289, "lr": 6.215015129087339e-05}
{"train_loss": 0.0989321693778038, "global_step": 114805, "epoch": 1289, "lr": 6.214958893075058e-05}
{"train_loss": 0.13921642303466797, "global_step": 114806, "epoch": 1289, "lr": 6.214902656899439e-05}
{"train_loss": 0.14458853006362915, "global_step": 114807, "epoch": 1289, "lr": 6.214846420560493e-05}
{"train_loss": 0.12285338342189789, "global_step": 114808, "epoch": 1289, "lr": 6.214790184058225e-05}
{"train_loss": 0.1340766665473413, "global_step": 114809, "epoch": 1289, "lr": 6.214733947392644e-05, "val_loss": 5.318305969238281}
{"train_loss": 0.08663288503885269, "global_step": 114810, "epoch": 1290, "lr": 6.214677710563755e-05}
{"train_loss": 0.13821449875831604, "global_step": 114811, "epoch": 1290, "lr": 6.214621473571568e-05}
{"train_loss": 0.1370428204536438, "global_step": 114812, "epoch": 1290, "lr": 6.214565236416091e-05}
{"train_loss": 0.21437981724739075, "global_step": 114813, "epoch": 1290, "lr": 6.214508999097329e-05}
{"train_loss": 0.12407004833221436, "global_step": 114814, "epoch": 1290, "lr": 6.214452761615292e-05}
{"train_loss": 0.13871556520462036, "global_step": 114815, "epoch": 1290, "lr": 6.214396523969987e-05}
{"train_loss": 0.1619691550731659, "global_step": 114816, "epoch": 1290, "lr": 6.21434028616142e-05}
{"train_loss": 0.1721583902835846, "global_step": 114817, "epoch": 1290, "lr": 6.214284048189601e-05}
{"train_loss": 0.17273150384426117, "global_step": 114818, "epoch": 1290, "lr": 6.214227810054536e-05}
{"train_loss": 0.13515682518482208, "global_step": 114819, "epoch": 1290, "lr": 6.214171571756232e-05}
{"train_loss": 0.08931872248649597, "global_step": 114820, "epoch": 1290, "lr": 6.2141153332947e-05}
{"train_loss": 0.1292475014925003, "global_step": 114821, "epoch": 1290, "lr": 6.214059094669942e-05}
{"train_loss": 0.23707224428653717, "global_step": 114822, "epoch": 1290, "lr": 6.21400285588197e-05}
{"train_loss": 0.12221432477235794, "global_step": 114823, "epoch": 1290, "lr": 6.213946616930789e-05}
{"train_loss": 0.0972842201590538, "global_step": 114824, "epoch": 1290, "lr": 6.213890377816408e-05}
{"train_loss": 0.15779174864292145, "global_step": 114825, "epoch": 1290, "lr": 6.213834138538835e-05}
{"train_loss": 0.13663436472415924, "global_step": 114826, "epoch": 1290, "lr": 6.213777899098076e-05}
{"train_loss": 0.19296367466449738, "global_step": 114827, "epoch": 1290, "lr": 6.213721659494141e-05}
{"train_loss": 0.135410875082016, "global_step": 114828, "epoch": 1290, "lr": 6.213665419727033e-05}
{"train_loss": 0.18466393649578094, "global_step": 114829, "epoch": 1290, "lr": 6.213609179796764e-05}
{"train_loss": 0.16752029955387115, "global_step": 114830, "epoch": 1290, "lr": 6.213552939703341e-05}
{"train_loss": 0.1327919214963913, "global_step": 114831, "epoch": 1290, "lr": 6.21349669944677e-05}
{"train_loss": 0.08321597427129745, "global_step": 114832, "epoch": 1290, "lr": 6.213440459027059e-05}
{"train_loss": 0.1316329836845398, "global_step": 114833, "epoch": 1290, "lr": 6.213384218444216e-05}
{"train_loss": 0.12206453084945679, "global_step": 114834, "epoch": 1290, "lr": 6.213327977698247e-05}
{"train_loss": 0.15999546647071838, "global_step": 114835, "epoch": 1290, "lr": 6.213271736789162e-05}
{"train_loss": 0.1146094799041748, "global_step": 114836, "epoch": 1290, "lr": 6.213215495716968e-05}
{"train_loss": 0.16528506577014923, "global_step": 114837, "epoch": 1290, "lr": 6.213159254481671e-05}
{"train_loss": 0.13641296327114105, "global_step": 114838, "epoch": 1290, "lr": 6.213103013083278e-05}
{"train_loss": 0.0864354819059372, "global_step": 114839, "epoch": 1290, "lr": 6.213046771521801e-05}
{"train_loss": 0.25145748257637024, "global_step": 114840, "epoch": 1290, "lr": 6.212990529797244e-05}
{"train_loss": 0.07281582802534103, "global_step": 114841, "epoch": 1290, "lr": 6.212934287909614e-05}
{"train_loss": 0.11222470551729202, "global_step": 114842, "epoch": 1290, "lr": 6.212878045858921e-05}
{"train_loss": 0.10166454315185547, "global_step": 114843, "epoch": 1290, "lr": 6.21282180364517e-05}
{"train_loss": 0.0765741690993309, "global_step": 114844, "epoch": 1290, "lr": 6.212765561268371e-05}
{"train_loss": 0.0982855036854744, "global_step": 114845, "epoch": 1290, "lr": 6.21270931872853e-05}
{"train_loss": 0.13826212286949158, "global_step": 114846, "epoch": 1290, "lr": 6.212653076025655e-05}
{"train_loss": 0.11301961541175842, "global_step": 114847, "epoch": 1290, "lr": 6.212596833159754e-05}
{"train_loss": 0.1755383461713791, "global_step": 114848, "epoch": 1290, "lr": 6.212540590130833e-05}
{"train_loss": 0.19483420252799988, "global_step": 114849, "epoch": 1290, "lr": 6.212484346938902e-05}
{"train_loss": 0.13080178201198578, "global_step": 114850, "epoch": 1290, "lr": 6.212428103583968e-05}
{"train_loss": 0.11764610558748245, "global_step": 114851, "epoch": 1290, "lr": 6.212371860066036e-05}
{"train_loss": 0.17024962604045868, "global_step": 114852, "epoch": 1290, "lr": 6.212315616385117e-05}
{"train_loss": 0.1586633026599884, "global_step": 114853, "epoch": 1290, "lr": 6.212259372541215e-05}
{"train_loss": 0.1484430730342865, "global_step": 114854, "epoch": 1290, "lr": 6.21220312853434e-05}
{"train_loss": 0.1417980045080185, "global_step": 114855, "epoch": 1290, "lr": 6.212146884364499e-05}
{"train_loss": 0.17343443632125854, "global_step": 114856, "epoch": 1290, "lr": 6.212090640031701e-05}
{"train_loss": 0.24823899567127228, "global_step": 114857, "epoch": 1290, "lr": 6.212034395535951e-05}
{"train_loss": 0.23320631682872772, "global_step": 114858, "epoch": 1290, "lr": 6.211978150877258e-05}
{"train_loss": 0.10543844848871231, "global_step": 114859, "epoch": 1290, "lr": 6.21192190605563e-05}
{"train_loss": 0.1747177243232727, "global_step": 114860, "epoch": 1290, "lr": 6.211865661071073e-05}
{"train_loss": 0.07988335192203522, "global_step": 114861, "epoch": 1290, "lr": 6.211809415923595e-05}
{"train_loss": 0.16349317133426666, "global_step": 114862, "epoch": 1290, "lr": 6.211753170613205e-05}
{"train_loss": 0.21738030016422272, "global_step": 114863, "epoch": 1290, "lr": 6.211696925139908e-05}
{"train_loss": 0.09131324291229248, "global_step": 114864, "epoch": 1290, "lr": 6.211640679503715e-05}
{"train_loss": 0.0780816376209259, "global_step": 114865, "epoch": 1290, "lr": 6.211584433704631e-05}
{"train_loss": 0.21881113946437836, "global_step": 114866, "epoch": 1290, "lr": 6.211528187742663e-05}
{"train_loss": 0.0988050028681755, "global_step": 114867, "epoch": 1290, "lr": 6.211471941617822e-05}
{"train_loss": 0.07181820273399353, "global_step": 114868, "epoch": 1290, "lr": 6.211415695330113e-05}
{"train_loss": 0.1741124838590622, "global_step": 114869, "epoch": 1290, "lr": 6.211359448879542e-05}
{"train_loss": 0.11286132037639618, "global_step": 114870, "epoch": 1290, "lr": 6.211303202266121e-05}
{"train_loss": 0.12971609830856323, "global_step": 114871, "epoch": 1290, "lr": 6.211246955489853e-05}
{"train_loss": 0.09584760665893555, "global_step": 114872, "epoch": 1290, "lr": 6.21119070855075e-05}
{"train_loss": 0.14219191670417786, "global_step": 114873, "epoch": 1290, "lr": 6.211134461448814e-05}
{"train_loss": 0.13166451454162598, "global_step": 114874, "epoch": 1290, "lr": 6.211078214184058e-05}
{"train_loss": 0.16806742548942566, "global_step": 114875, "epoch": 1290, "lr": 6.211021966756486e-05}
{"train_loss": 0.08390840888023376, "global_step": 114876, "epoch": 1290, "lr": 6.210965719166109e-05}
{"train_loss": 0.1801055371761322, "global_step": 114877, "epoch": 1290, "lr": 6.21090947141293e-05}
{"train_loss": 0.0988234207034111, "global_step": 114878, "epoch": 1290, "lr": 6.21085322349696e-05}
{"train_loss": 0.1742437779903412, "global_step": 114879, "epoch": 1290, "lr": 6.210796975418205e-05}
{"train_loss": 0.1120949387550354, "global_step": 114880, "epoch": 1290, "lr": 6.210740727176673e-05}
{"train_loss": 0.12331196665763855, "global_step": 114881, "epoch": 1290, "lr": 6.210684478772373e-05}
{"train_loss": 0.1476387083530426, "global_step": 114882, "epoch": 1290, "lr": 6.21062823020531e-05}
{"train_loss": 0.12024234235286713, "global_step": 114883, "epoch": 1290, "lr": 6.210571981475492e-05}
{"train_loss": 0.11574959754943848, "global_step": 114884, "epoch": 1290, "lr": 6.210515732582929e-05}
{"train_loss": 0.07725263386964798, "global_step": 114885, "epoch": 1290, "lr": 6.210459483527628e-05}
{"train_loss": 0.1912171095609665, "global_step": 114886, "epoch": 1290, "lr": 6.210403234309595e-05}
{"train_loss": 0.171794593334198, "global_step": 114887, "epoch": 1290, "lr": 6.210346984928835e-05}
{"train_loss": 0.14825287461280823, "global_step": 114888, "epoch": 1290, "lr": 6.210290735385361e-05}
{"train_loss": 0.18156391382217407, "global_step": 114889, "epoch": 1290, "lr": 6.210234485679179e-05}
{"train_loss": 0.10453411936759949, "global_step": 114890, "epoch": 1290, "lr": 6.210178235810294e-05}
{"train_loss": 0.1514371782541275, "global_step": 114891, "epoch": 1290, "lr": 6.210121985778715e-05}
{"train_loss": 0.1494416743516922, "global_step": 114892, "epoch": 1290, "lr": 6.210065735584451e-05}
{"train_loss": 0.17895087599754333, "global_step": 114893, "epoch": 1290, "lr": 6.210009485227509e-05}
{"train_loss": 0.16208525002002716, "global_step": 114894, "epoch": 1290, "lr": 6.209953234707895e-05}
{"train_loss": 0.15759623050689697, "global_step": 114895, "epoch": 1290, "lr": 6.209896984025617e-05}
{"train_loss": 0.10950057208538055, "global_step": 114896, "epoch": 1290, "lr": 6.209840733180683e-05}
{"train_loss": 0.12883885204792023, "global_step": 114897, "epoch": 1290, "lr": 6.209784482173102e-05}
{"train_loss": 0.14151160544558858, "global_step": 114898, "epoch": 1290, "lr": 6.209728231002879e-05, "val_loss": 5.091740131378174, "train_action_mse_error": 30.51873016357422}
{"train_loss": 0.16366730630397797, "global_step": 114899, "epoch": 1291, "lr": 6.209671979670024e-05}
{"train_loss": 0.15600523352622986, "global_step": 114900, "epoch": 1291, "lr": 6.209615728174543e-05}
{"train_loss": 0.21319735050201416, "global_step": 114901, "epoch": 1291, "lr": 6.209559476516443e-05}
{"train_loss": 0.11817470192909241, "global_step": 114902, "epoch": 1291, "lr": 6.209503224695732e-05}
{"train_loss": 0.13318361341953278, "global_step": 114903, "epoch": 1291, "lr": 6.20944697271242e-05}
{"train_loss": 0.13929463922977448, "global_step": 114904, "epoch": 1291, "lr": 6.209390720566511e-05}
{"train_loss": 0.1421421319246292, "global_step": 114905, "epoch": 1291, "lr": 6.209334468258016e-05}
{"train_loss": 0.14065498113632202, "global_step": 114906, "epoch": 1291, "lr": 6.209278215786938e-05}
{"train_loss": 0.14502471685409546, "global_step": 114907, "epoch": 1291, "lr": 6.20922196315329e-05}
{"train_loss": 0.0945456326007843, "global_step": 114908, "epoch": 1291, "lr": 6.209165710357076e-05}
{"train_loss": 0.12888601422309875, "global_step": 114909, "epoch": 1291, "lr": 6.209109457398303e-05}
{"train_loss": 0.12819381058216095, "global_step": 114910, "epoch": 1291, "lr": 6.209053204276982e-05}
{"train_loss": 0.25440698862075806, "global_step": 114911, "epoch": 1291, "lr": 6.208996950993118e-05}
{"train_loss": 0.13558922708034515, "global_step": 114912, "epoch": 1291, "lr": 6.208940697546718e-05}
{"train_loss": 0.13741634786128998, "global_step": 114913, "epoch": 1291, "lr": 6.208884443937791e-05}
{"train_loss": 0.1522492617368698, "global_step": 114914, "epoch": 1291, "lr": 6.208828190166345e-05}
{"train_loss": 0.18044893443584442, "global_step": 114915, "epoch": 1291, "lr": 6.208771936232386e-05}
{"train_loss": 0.05015748366713524, "global_step": 114916, "epoch": 1291, "lr": 6.208715682135923e-05}
{"train_loss": 0.26425701379776, "global_step": 114917, "epoch": 1291, "lr": 6.208659427876963e-05}
{"train_loss": 0.16133421659469604, "global_step": 114918, "epoch": 1291, "lr": 6.208603173455513e-05}
{"train_loss": 0.12794356048107147, "global_step": 114919, "epoch": 1291, "lr": 6.208546918871579e-05}
{"train_loss": 0.1274634599685669, "global_step": 114920, "epoch": 1291, "lr": 6.208490664125172e-05}
{"train_loss": 0.1895969659090042, "global_step": 114921, "epoch": 1291, "lr": 6.2084344092163e-05}
{"train_loss": 0.14618560671806335, "global_step": 114922, "epoch": 1291, "lr": 6.208378154144968e-05}
{"train_loss": 0.15868067741394043, "global_step": 114923, "epoch": 1291, "lr": 6.208321898911185e-05}
{"train_loss": 0.18605156242847443, "global_step": 114924, "epoch": 1291, "lr": 6.208265643514957e-05}
{"train_loss": 0.21610499918460846, "global_step": 114925, "epoch": 1291, "lr": 6.208209387956292e-05}
{"train_loss": 0.1534920334815979, "global_step": 114926, "epoch": 1291, "lr": 6.208153132235198e-05}
{"train_loss": 0.13460902869701385, "global_step": 114927, "epoch": 1291, "lr": 6.208096876351682e-05}
{"train_loss": 0.16320641338825226, "global_step": 114928, "epoch": 1291, "lr": 6.208040620305753e-05}
{"train_loss": 0.16007600724697113, "global_step": 114929, "epoch": 1291, "lr": 6.207984364097418e-05}
{"train_loss": 0.12903785705566406, "global_step": 114930, "epoch": 1291, "lr": 6.207928107726684e-05}
{"train_loss": 0.13791991770267487, "global_step": 114931, "epoch": 1291, "lr": 6.207871851193557e-05}
{"train_loss": 0.311177134513855, "global_step": 114932, "epoch": 1291, "lr": 6.207815594498049e-05}
{"train_loss": 0.15603791177272797, "global_step": 114933, "epoch": 1291, "lr": 6.207759337640164e-05}
{"train_loss": 0.18988746404647827, "global_step": 114934, "epoch": 1291, "lr": 6.207703080619909e-05}
{"train_loss": 0.13211503624916077, "global_step": 114935, "epoch": 1291, "lr": 6.207646823437294e-05}
{"train_loss": 0.17188847064971924, "global_step": 114936, "epoch": 1291, "lr": 6.207590566092327e-05}
{"train_loss": 0.098894402384758, "global_step": 114937, "epoch": 1291, "lr": 6.207534308585013e-05}
{"train_loss": 0.13940124213695526, "global_step": 114938, "epoch": 1291, "lr": 6.207478050915358e-05}
{"train_loss": 0.17950554192066193, "global_step": 114939, "epoch": 1291, "lr": 6.207421793083375e-05}
{"train_loss": 0.15909279882907867, "global_step": 114940, "epoch": 1291, "lr": 6.207365535089069e-05}
{"train_loss": 0.2638890743255615, "global_step": 114941, "epoch": 1291, "lr": 6.207309276932448e-05}
{"train_loss": 0.11539724469184875, "global_step": 114942, "epoch": 1291, "lr": 6.207253018613519e-05}
{"train_loss": 0.17071624100208282, "global_step": 114943, "epoch": 1291, "lr": 6.207196760132289e-05}
{"train_loss": 0.19275853037834167, "global_step": 114944, "epoch": 1291, "lr": 6.207140501488766e-05}
{"train_loss": 0.0855112373828888, "global_step": 114945, "epoch": 1291, "lr": 6.207084242682958e-05}
{"train_loss": 0.11016670614480972, "global_step": 114946, "epoch": 1291, "lr": 6.207027983714871e-05}
{"train_loss": 0.11875328421592712, "global_step": 114947, "epoch": 1291, "lr": 6.206971724584515e-05}
{"train_loss": 0.15038323402404785, "global_step": 114948, "epoch": 1291, "lr": 6.206915465291896e-05}
{"train_loss": 0.04466209188103676, "global_step": 114949, "epoch": 1291, "lr": 6.206859205837023e-05}
{"train_loss": 0.09452323615550995, "global_step": 114950, "epoch": 1291, "lr": 6.2068029462199e-05}
{"train_loss": 0.18609745800495148, "global_step": 114951, "epoch": 1291, "lr": 6.206746686440538e-05}
{"train_loss": 0.21959537267684937, "global_step": 114952, "epoch": 1291, "lr": 6.206690426498945e-05}
{"train_loss": 0.10918154567480087, "global_step": 114953, "epoch": 1291, "lr": 6.206634166395127e-05}
{"train_loss": 0.10329020023345947, "global_step": 114954, "epoch": 1291, "lr": 6.20657790612909e-05}
{"train_loss": 0.08711883425712585, "global_step": 114955, "epoch": 1291, "lr": 6.206521645700844e-05}
{"train_loss": 0.17604035139083862, "global_step": 114956, "epoch": 1291, "lr": 6.206465385110396e-05}
{"train_loss": 0.10062098503112793, "global_step": 114957, "epoch": 1291, "lr": 6.206409124357752e-05}
{"train_loss": 0.15598328411579132, "global_step": 114958, "epoch": 1291, "lr": 6.206352863442923e-05}
{"train_loss": 0.11376588046550751, "global_step": 114959, "epoch": 1291, "lr": 6.206296602365915e-05}
{"train_loss": 0.16040444374084473, "global_step": 114960, "epoch": 1291, "lr": 6.206240341126734e-05}
{"train_loss": 0.04571203142404556, "global_step": 114961, "epoch": 1291, "lr": 6.206184079725388e-05}
{"train_loss": 0.13428324460983276, "global_step": 114962, "epoch": 1291, "lr": 6.206127818161888e-05}
{"train_loss": 0.132962167263031, "global_step": 114963, "epoch": 1291, "lr": 6.206071556436237e-05}
{"train_loss": 0.1491990089416504, "global_step": 114964, "epoch": 1291, "lr": 6.206015294548446e-05}
{"train_loss": 0.13647525012493134, "global_step": 114965, "epoch": 1291, "lr": 6.205959032498518e-05}
{"train_loss": 0.13561594486236572, "global_step": 114966, "epoch": 1291, "lr": 6.205902770286464e-05}
{"train_loss": 0.08463636785745621, "global_step": 114967, "epoch": 1291, "lr": 6.205846507912293e-05}
{"train_loss": 0.1532628983259201, "global_step": 114968, "epoch": 1291, "lr": 6.20579024537601e-05}
{"train_loss": 0.1081833466887474, "global_step": 114969, "epoch": 1291, "lr": 6.205733982677622e-05}
{"train_loss": 0.17074573040008545, "global_step": 114970, "epoch": 1291, "lr": 6.205677719817138e-05}
{"train_loss": 0.1692357361316681, "global_step": 114971, "epoch": 1291, "lr": 6.205621456794568e-05}
{"train_loss": 0.1624400019645691, "global_step": 114972, "epoch": 1291, "lr": 6.205565193609914e-05}
{"train_loss": 0.1569708287715912, "global_step": 114973, "epoch": 1291, "lr": 6.205508930263186e-05}
{"train_loss": 0.11569642275571823, "global_step": 114974, "epoch": 1291, "lr": 6.205452666754393e-05}
{"train_loss": 0.1266581416130066, "global_step": 114975, "epoch": 1291, "lr": 6.205396403083541e-05}
{"train_loss": 0.20131906867027283, "global_step": 114976, "epoch": 1291, "lr": 6.205340139250639e-05}
{"train_loss": 0.13950885832309723, "global_step": 114977, "epoch": 1291, "lr": 6.205283875255695e-05}
{"train_loss": 0.1682663857936859, "global_step": 114978, "epoch": 1291, "lr": 6.205227611098712e-05}
{"train_loss": 0.12861648201942444, "global_step": 114979, "epoch": 1291, "lr": 6.205171346779703e-05}
{"train_loss": 0.12464774399995804, "global_step": 114980, "epoch": 1291, "lr": 6.205115082298674e-05}
{"train_loss": 0.1858457326889038, "global_step": 114981, "epoch": 1291, "lr": 6.205058817655632e-05}
{"train_loss": 0.08647671341896057, "global_step": 114982, "epoch": 1291, "lr": 6.205002552850584e-05}
{"train_loss": 0.17577724158763885, "global_step": 114983, "epoch": 1291, "lr": 6.204946287883537e-05}
{"train_loss": 0.22543883323669434, "global_step": 114984, "epoch": 1291, "lr": 6.204890022754501e-05}
{"train_loss": 0.15839742124080658, "global_step": 114985, "epoch": 1291, "lr": 6.204833757463482e-05}
{"train_loss": 0.13660097122192383, "global_step": 114986, "epoch": 1291, "lr": 6.204777492010487e-05}
{"train_loss": 0.14870921522378922, "global_step": 114987, "epoch": 1291, "lr": 6.204721226395524e-05, "val_loss": 5.235447406768799}
{"train_loss": 0.13869524002075195, "global_step": 114988, "epoch": 1292, "lr": 6.204664960618603e-05}
{"train_loss": 0.16017627716064453, "global_step": 114989, "epoch": 1292, "lr": 6.204608694679728e-05}
{"train_loss": 0.10801883041858673, "global_step": 114990, "epoch": 1292, "lr": 6.204552428578908e-05}
{"train_loss": 0.13100701570510864, "global_step": 114991, "epoch": 1292, "lr": 6.204496162316152e-05}
{"train_loss": 0.14880706369876862, "global_step": 114992, "epoch": 1292, "lr": 6.204439895891463e-05}
{"train_loss": 0.1304912567138672, "global_step": 114993, "epoch": 1292, "lr": 6.204383629304853e-05}
{"train_loss": 0.1914127767086029, "global_step": 114994, "epoch": 1292, "lr": 6.204327362556329e-05}
{"train_loss": 0.19682523608207703, "global_step": 114995, "epoch": 1292, "lr": 6.204271095645899e-05}
{"train_loss": 0.13518264889717102, "global_step": 114996, "epoch": 1292, "lr": 6.204214828573567e-05}
{"train_loss": 0.08351032435894012, "global_step": 114997, "epoch": 1292, "lr": 6.204158561339346e-05}
{"train_loss": 0.1434798687696457, "global_step": 114998, "epoch": 1292, "lr": 6.204102293943237e-05}
{"train_loss": 0.1622157096862793, "global_step": 114999, "epoch": 1292, "lr": 6.204046026385254e-05}
{"train_loss": 0.14799495041370392, "global_step": 115000, "epoch": 1292, "lr": 6.2039897586654e-05}
{"train_loss": 0.12304593622684479, "global_step": 115001, "epoch": 1292, "lr": 6.203933490783686e-05}
{"train_loss": 0.16692180931568146, "global_step": 115002, "epoch": 1292, "lr": 6.203877222740117e-05}
{"train_loss": 0.15444177389144897, "global_step": 115003, "epoch": 1292, "lr": 6.203820954534702e-05}
{"train_loss": 0.133775532245636, "global_step": 115004, "epoch": 1292, "lr": 6.203764686167446e-05}
{"train_loss": 0.13881425559520721, "global_step": 115005, "epoch": 1292, "lr": 6.20370841763836e-05}
{"train_loss": 0.1239108145236969, "global_step": 115006, "epoch": 1292, "lr": 6.20365214894745e-05}
{"train_loss": 0.09660199284553528, "global_step": 115007, "epoch": 1292, "lr": 6.203595880094722e-05}
{"train_loss": 0.1408604383468628, "global_step": 115008, "epoch": 1292, "lr": 6.203539611080187e-05}
{"train_loss": 0.10859577357769012, "global_step": 115009, "epoch": 1292, "lr": 6.20348334190385e-05}
{"train_loss": 0.0872420221567154, "global_step": 115010, "epoch": 1292, "lr": 6.203427072565719e-05}
{"train_loss": 0.12131340801715851, "global_step": 115011, "epoch": 1292, "lr": 6.203370803065802e-05}
{"train_loss": 0.08441399782896042, "global_step": 115012, "epoch": 1292, "lr": 6.203314533404107e-05}
{"train_loss": 0.14705051481723785, "global_step": 115013, "epoch": 1292, "lr": 6.203258263580641e-05}
{"train_loss": 0.1999041885137558, "global_step": 115014, "epoch": 1292, "lr": 6.203201993595412e-05}
{"train_loss": 0.19821681082248688, "global_step": 115015, "epoch": 1292, "lr": 6.203145723448427e-05}
{"train_loss": 0.17282137274742126, "global_step": 115016, "epoch": 1292, "lr": 6.203089453139693e-05}
{"train_loss": 0.18333397805690765, "global_step": 115017, "epoch": 1292, "lr": 6.20303318266922e-05}
{"train_loss": 0.2557443678379059, "global_step": 115018, "epoch": 1292, "lr": 6.202976912037012e-05}
{"train_loss": 0.16123944520950317, "global_step": 115019, "epoch": 1292, "lr": 6.202920641243078e-05}
{"train_loss": 0.10253817588090897, "global_step": 115020, "epoch": 1292, "lr": 6.202864370287428e-05}
{"train_loss": 0.16234499216079712, "global_step": 115021, "epoch": 1292, "lr": 6.202808099170067e-05}
{"train_loss": 0.16230933368206024, "global_step": 115022, "epoch": 1292, "lr": 6.202751827891002e-05}
{"train_loss": 0.18876858055591583, "global_step": 115023, "epoch": 1292, "lr": 6.202695556450244e-05}
{"train_loss": 0.11545203626155853, "global_step": 115024, "epoch": 1292, "lr": 6.202639284847796e-05}
{"train_loss": 0.14162170886993408, "global_step": 115025, "epoch": 1292, "lr": 6.202583013083669e-05}
{"train_loss": 0.11830338835716248, "global_step": 115026, "epoch": 1292, "lr": 6.202526741157868e-05}
{"train_loss": 0.1801120489835739, "global_step": 115027, "epoch": 1292, "lr": 6.202470469070402e-05}
{"train_loss": 0.13635805249214172, "global_step": 115028, "epoch": 1292, "lr": 6.20241419682128e-05}
{"train_loss": 0.1501138061285019, "global_step": 115029, "epoch": 1292, "lr": 6.202357924410506e-05}
{"train_loss": 0.19304639101028442, "global_step": 115030, "epoch": 1292, "lr": 6.20230165183809e-05}
{"train_loss": 0.14908066391944885, "global_step": 115031, "epoch": 1292, "lr": 6.202245379104041e-05}
{"train_loss": 0.11017359048128128, "global_step": 115032, "epoch": 1292, "lr": 6.202189106208364e-05}
{"train_loss": 0.15803834795951843, "global_step": 115033, "epoch": 1292, "lr": 6.202132833151067e-05}
{"train_loss": 0.11059052497148514, "global_step": 115034, "epoch": 1292, "lr": 6.202076559932159e-05}
{"train_loss": 0.16057094931602478, "global_step": 115035, "epoch": 1292, "lr": 6.202020286551644e-05}
{"train_loss": 0.1501021683216095, "global_step": 115036, "epoch": 1292, "lr": 6.201964013009534e-05}
{"train_loss": 0.13145175576210022, "global_step": 115037, "epoch": 1292, "lr": 6.201907739305834e-05}
{"train_loss": 0.15577204525470734, "global_step": 115038, "epoch": 1292, "lr": 6.201851465440552e-05}
{"train_loss": 0.13310909271240234, "global_step": 115039, "epoch": 1292, "lr": 6.201795191413696e-05}
{"train_loss": 0.152662992477417, "global_step": 115040, "epoch": 1292, "lr": 6.201738917225275e-05}
{"train_loss": 0.12182106822729111, "global_step": 115041, "epoch": 1292, "lr": 6.201682642875293e-05}
{"train_loss": 0.16186271607875824, "global_step": 115042, "epoch": 1292, "lr": 6.20162636836376e-05}
{"train_loss": 0.14020183682441711, "global_step": 115043, "epoch": 1292, "lr": 6.201570093690682e-05}
{"train_loss": 0.09838380664587021, "global_step": 115044, "epoch": 1292, "lr": 6.201513818856069e-05}
{"train_loss": 0.157430961728096, "global_step": 115045, "epoch": 1292, "lr": 6.201457543859926e-05}
{"train_loss": 0.18671540915966034, "global_step": 115046, "epoch": 1292, "lr": 6.201401268702261e-05}
{"train_loss": 0.18681561946868896, "global_step": 115047, "epoch": 1292, "lr": 6.201344993383082e-05}
{"train_loss": 0.10514118522405624, "global_step": 115048, "epoch": 1292, "lr": 6.201288717902399e-05}
{"train_loss": 0.10879482328891754, "global_step": 115049, "epoch": 1292, "lr": 6.201232442260215e-05}
{"train_loss": 0.18488968908786774, "global_step": 115050, "epoch": 1292, "lr": 6.201176166456542e-05}
{"train_loss": 0.1328865885734558, "global_step": 115051, "epoch": 1292, "lr": 6.201119890491385e-05}
{"train_loss": 0.12625740468502045, "global_step": 115052, "epoch": 1292, "lr": 6.201063614364752e-05}
{"train_loss": 0.14207923412322998, "global_step": 115053, "epoch": 1292, "lr": 6.20100733807665e-05}
{"train_loss": 0.16674737632274628, "global_step": 115054, "epoch": 1292, "lr": 6.200951061627087e-05}
{"train_loss": 0.2115868330001831, "global_step": 115055, "epoch": 1292, "lr": 6.200894785016071e-05}
{"train_loss": 0.1561887413263321, "global_step": 115056, "epoch": 1292, "lr": 6.20083850824361e-05}
{"train_loss": 0.21373067796230316, "global_step": 115057, "epoch": 1292, "lr": 6.200782231309711e-05}
{"train_loss": 0.06948111951351166, "global_step": 115058, "epoch": 1292, "lr": 6.200725954214382e-05}
{"train_loss": 0.18856143951416016, "global_step": 115059, "epoch": 1292, "lr": 6.20066967695763e-05}
{"train_loss": 0.1714085042476654, "global_step": 115060, "epoch": 1292, "lr": 6.200613399539462e-05}
{"train_loss": 0.1642768383026123, "global_step": 115061, "epoch": 1292, "lr": 6.200557121959884e-05}
{"train_loss": 0.18099890649318695, "global_step": 115062, "epoch": 1292, "lr": 6.200500844218909e-05}
{"train_loss": 0.1817978322505951, "global_step": 115063, "epoch": 1292, "lr": 6.20044456631654e-05}
{"train_loss": 0.17286156117916107, "global_step": 115064, "epoch": 1292, "lr": 6.200388288252785e-05}
{"train_loss": 0.16143561899662018, "global_step": 115065, "epoch": 1292, "lr": 6.200332010027654e-05}
{"train_loss": 0.1811693012714386, "global_step": 115066, "epoch": 1292, "lr": 6.20027573164115e-05}
{"train_loss": 0.18617524206638336, "global_step": 115067, "epoch": 1292, "lr": 6.200219453093287e-05}
{"train_loss": 0.13610319793224335, "global_step": 115068, "epoch": 1292, "lr": 6.200163174384068e-05}
{"train_loss": 0.16140344738960266, "global_step": 115069, "epoch": 1292, "lr": 6.200106895513501e-05}
{"train_loss": 0.15740446746349335, "global_step": 115070, "epoch": 1292, "lr": 6.200050616481596e-05}
{"train_loss": 0.08271025866270065, "global_step": 115071, "epoch": 1292, "lr": 6.199994337288357e-05}
{"train_loss": 0.12421047687530518, "global_step": 115072, "epoch": 1292, "lr": 6.199938057933794e-05}
{"train_loss": 0.14565394818782806, "global_step": 115073, "epoch": 1292, "lr": 6.199881778417913e-05}
{"train_loss": 0.16408322751522064, "global_step": 115074, "epoch": 1292, "lr": 6.199825498740723e-05}
{"train_loss": 0.15032103657722473, "global_step": 115075, "epoch": 1292, "lr": 6.199769218902232e-05}
{"train_loss": 0.14850149613417937, "global_step": 115076, "epoch": 1292, "lr": 6.199712938902446e-05, "val_loss": 4.950962066650391}
{"train_loss": 0.1549552083015442, "global_step": 115077, "epoch": 1293, "lr": 6.199656658741374e-05}
{"train_loss": 0.13247621059417725, "global_step": 115078, "epoch": 1293, "lr": 6.199600378419024e-05}
{"train_loss": 0.12669190764427185, "global_step": 115079, "epoch": 1293, "lr": 6.1995440979354e-05}
{"train_loss": 0.08490824699401855, "global_step": 115080, "epoch": 1293, "lr": 6.199487817290513e-05}
{"train_loss": 0.11589419096708298, "global_step": 115081, "epoch": 1293, "lr": 6.199431536484369e-05}
{"train_loss": 0.13229331374168396, "global_step": 115082, "epoch": 1293, "lr": 6.199375255516977e-05}
{"train_loss": 0.12066887319087982, "global_step": 115083, "epoch": 1293, "lr": 6.199318974388341e-05}
{"train_loss": 0.11277364194393158, "global_step": 115084, "epoch": 1293, "lr": 6.199262693098472e-05}
{"train_loss": 0.10863540321588516, "global_step": 115085, "epoch": 1293, "lr": 6.199206411647378e-05}
{"train_loss": 0.1742807924747467, "global_step": 115086, "epoch": 1293, "lr": 6.199150130035065e-05}
{"train_loss": 0.15354062616825104, "global_step": 115087, "epoch": 1293, "lr": 6.19909384826154e-05}
{"train_loss": 0.08739497512578964, "global_step": 115088, "epoch": 1293, "lr": 6.199037566326813e-05}
{"train_loss": 0.1381671279668808, "global_step": 115089, "epoch": 1293, "lr": 6.198981284230889e-05}
{"train_loss": 0.13277500867843628, "global_step": 115090, "epoch": 1293, "lr": 6.198925001973776e-05}
{"train_loss": 0.1756514012813568, "global_step": 115091, "epoch": 1293, "lr": 6.198868719555484e-05}
{"train_loss": 0.10378087311983109, "global_step": 115092, "epoch": 1293, "lr": 6.198812436976017e-05}
{"train_loss": 0.17677415907382965, "global_step": 115093, "epoch": 1293, "lr": 6.198756154235385e-05}
{"train_loss": 0.11924318224191666, "global_step": 115094, "epoch": 1293, "lr": 6.198699871333595e-05}
{"train_loss": 0.15276674926280975, "global_step": 115095, "epoch": 1293, "lr": 6.198643588270654e-05}
{"train_loss": 0.14586606621742249, "global_step": 115096, "epoch": 1293, "lr": 6.19858730504657e-05}
{"train_loss": 0.1415230631828308, "global_step": 115097, "epoch": 1293, "lr": 6.198531021661352e-05}
{"train_loss": 0.142720565199852, "global_step": 115098, "epoch": 1293, "lr": 6.198474738115006e-05}
{"train_loss": 0.11185117810964584, "global_step": 115099, "epoch": 1293, "lr": 6.198418454407537e-05}
{"train_loss": 0.1649932563304901, "global_step": 115100, "epoch": 1293, "lr": 6.198362170538957e-05}
{"train_loss": 0.14060194790363312, "global_step": 115101, "epoch": 1293, "lr": 6.19830588650927e-05}
{"train_loss": 0.08790677040815353, "global_step": 115102, "epoch": 1293, "lr": 6.198249602318488e-05}
{"train_loss": 0.10908378660678864, "global_step": 115103, "epoch": 1293, "lr": 6.198193317966614e-05}
{"train_loss": 0.1200345903635025, "global_step": 115104, "epoch": 1293, "lr": 6.19813703345366e-05}
{"train_loss": 0.13234928250312805, "global_step": 115105, "epoch": 1293, "lr": 6.198080748779629e-05}
{"train_loss": 0.08121241629123688, "global_step": 115106, "epoch": 1293, "lr": 6.198024463944531e-05}
{"train_loss": 0.20376640558242798, "global_step": 115107, "epoch": 1293, "lr": 6.197968178948374e-05}
{"train_loss": 0.14735791087150574, "global_step": 115108, "epoch": 1293, "lr": 6.197911893791163e-05}
{"train_loss": 0.15176430344581604, "global_step": 115109, "epoch": 1293, "lr": 6.19785560847291e-05}
{"train_loss": 0.16416329145431519, "global_step": 115110, "epoch": 1293, "lr": 6.197799322993617e-05}
{"train_loss": 0.12084899097681046, "global_step": 115111, "epoch": 1293, "lr": 6.197743037353296e-05}
{"train_loss": 0.10644684731960297, "global_step": 115112, "epoch": 1293, "lr": 6.197686751551953e-05}
{"train_loss": 0.1332583874464035, "global_step": 115113, "epoch": 1293, "lr": 6.197630465589596e-05}
{"train_loss": 0.1462680697441101, "global_step": 115114, "epoch": 1293, "lr": 6.19757417946623e-05}
{"train_loss": 0.14285285770893097, "global_step": 115115, "epoch": 1293, "lr": 6.197517893181868e-05}
{"train_loss": 0.07369059324264526, "global_step": 115116, "epoch": 1293, "lr": 6.197461606736512e-05}
{"train_loss": 0.16518919169902802, "global_step": 115117, "epoch": 1293, "lr": 6.197405320130174e-05}
{"train_loss": 0.13907822966575623, "global_step": 115118, "epoch": 1293, "lr": 6.197349033362858e-05}
{"train_loss": 0.1379147171974182, "global_step": 115119, "epoch": 1293, "lr": 6.197292746434572e-05}
{"train_loss": 0.1061452180147171, "global_step": 115120, "epoch": 1293, "lr": 6.197236459345326e-05}
{"train_loss": 0.1453445553779602, "global_step": 115121, "epoch": 1293, "lr": 6.197180172095126e-05}
{"train_loss": 0.11723999679088593, "global_step": 115122, "epoch": 1293, "lr": 6.19712388468398e-05}
{"train_loss": 0.17016911506652832, "global_step": 115123, "epoch": 1293, "lr": 6.197067597111894e-05}
{"train_loss": 0.19462566077709198, "global_step": 115124, "epoch": 1293, "lr": 6.197011309378879e-05}
{"train_loss": 0.2168230563402176, "global_step": 115125, "epoch": 1293, "lr": 6.196955021484938e-05}
{"train_loss": 0.10040415078401566, "global_step": 115126, "epoch": 1293, "lr": 6.196898733430082e-05}
{"train_loss": 0.0702400803565979, "global_step": 115127, "epoch": 1293, "lr": 6.196842445214318e-05}
{"train_loss": 0.17134903371334076, "global_step": 115128, "epoch": 1293, "lr": 6.196786156837652e-05}
{"train_loss": 0.18925082683563232, "global_step": 115129, "epoch": 1293, "lr": 6.196729868300094e-05}
{"train_loss": 0.13504864275455475, "global_step": 115130, "epoch": 1293, "lr": 6.19667357960165e-05}
{"train_loss": 0.10185673087835312, "global_step": 115131, "epoch": 1293, "lr": 6.196617290742328e-05}
{"train_loss": 0.1223701536655426, "global_step": 115132, "epoch": 1293, "lr": 6.196561001722134e-05}
{"train_loss": 0.12511292099952698, "global_step": 115133, "epoch": 1293, "lr": 6.196504712541079e-05}
{"train_loss": 0.1806417852640152, "global_step": 115134, "epoch": 1293, "lr": 6.196448423199167e-05}
{"train_loss": 0.12343893945217133, "global_step": 115135, "epoch": 1293, "lr": 6.196392133696407e-05}
{"train_loss": 0.138410285115242, "global_step": 115136, "epoch": 1293, "lr": 6.19633584403281e-05}
{"train_loss": 0.1232600137591362, "global_step": 115137, "epoch": 1293, "lr": 6.196279554208375e-05}
{"train_loss": 0.09117846190929413, "global_step": 115138, "epoch": 1293, "lr": 6.196223264223118e-05}
{"train_loss": 0.13390646874904633, "global_step": 115139, "epoch": 1293, "lr": 6.196166974077042e-05}
{"train_loss": 0.14005079865455627, "global_step": 115140, "epoch": 1293, "lr": 6.196110683770157e-05}
{"train_loss": 0.1327597200870514, "global_step": 115141, "epoch": 1293, "lr": 6.196054393302469e-05}
{"train_loss": 0.10872472077608109, "global_step": 115142, "epoch": 1293, "lr": 6.195998102673986e-05}
{"train_loss": 0.1490563005208969, "global_step": 115143, "epoch": 1293, "lr": 6.195941811884716e-05}
{"train_loss": 0.13187813758850098, "global_step": 115144, "epoch": 1293, "lr": 6.195885520934665e-05}
{"train_loss": 0.09761150181293488, "global_step": 115145, "epoch": 1293, "lr": 6.195829229823843e-05}
{"train_loss": 0.10265439003705978, "global_step": 115146, "epoch": 1293, "lr": 6.195772938552255e-05}
{"train_loss": 0.07839396595954895, "global_step": 115147, "epoch": 1293, "lr": 6.19571664711991e-05}
{"train_loss": 0.15580803155899048, "global_step": 115148, "epoch": 1293, "lr": 6.195660355526818e-05}
{"train_loss": 0.16261987388134003, "global_step": 115149, "epoch": 1293, "lr": 6.195604063772982e-05}
{"train_loss": 0.1306387037038803, "global_step": 115150, "epoch": 1293, "lr": 6.195547771858412e-05}
{"train_loss": 0.11068127304315567, "global_step": 115151, "epoch": 1293, "lr": 6.195491479783115e-05}
{"train_loss": 0.14499320089817047, "global_step": 115152, "epoch": 1293, "lr": 6.195435187547099e-05}
{"train_loss": 0.14142818748950958, "global_step": 115153, "epoch": 1293, "lr": 6.195378895150372e-05}
{"train_loss": 0.1041012555360794, "global_step": 115154, "epoch": 1293, "lr": 6.19532260259294e-05}
{"train_loss": 0.1476784646511078, "global_step": 115155, "epoch": 1293, "lr": 6.19526630987481e-05}
{"train_loss": 0.15545429289340973, "global_step": 115156, "epoch": 1293, "lr": 6.195210016995994e-05}
{"train_loss": 0.19873839616775513, "global_step": 115157, "epoch": 1293, "lr": 6.195153723956495e-05}
{"train_loss": 0.13193422555923462, "global_step": 115158, "epoch": 1293, "lr": 6.195097430756322e-05}
{"train_loss": 0.09007204324007034, "global_step": 115159, "epoch": 1293, "lr": 6.195041137395483e-05}
{"train_loss": 0.12477196753025055, "global_step": 115160, "epoch": 1293, "lr": 6.194984843873985e-05}
{"train_loss": 0.19262917339801788, "global_step": 115161, "epoch": 1293, "lr": 6.194928550191836e-05}
{"train_loss": 0.12383124232292175, "global_step": 115162, "epoch": 1293, "lr": 6.194872256349044e-05}
{"train_loss": 0.17859579622745514, "global_step": 115163, "epoch": 1293, "lr": 6.194815962345616e-05}
{"train_loss": 0.11696023494005203, "global_step": 115164, "epoch": 1293, "lr": 6.194759668181559e-05}
{"train_loss": 0.13505977626596943, "global_step": 115165, "epoch": 1293, "lr": 6.194703373856882e-05, "val_loss": 5.341933250427246}
{"train_loss": 0.14692237973213196, "global_step": 115166, "epoch": 1294, "lr": 6.194647079371591e-05}
{"train_loss": 0.23041260242462158, "global_step": 115167, "epoch": 1294, "lr": 6.194590784725693e-05}
{"train_loss": 0.11415456235408783, "global_step": 115168, "epoch": 1294, "lr": 6.194534489919199e-05}
{"train_loss": 0.1103837862610817, "global_step": 115169, "epoch": 1294, "lr": 6.194478194952114e-05}
{"train_loss": 0.13884644210338593, "global_step": 115170, "epoch": 1294, "lr": 6.194421899824444e-05}
{"train_loss": 0.13368576765060425, "global_step": 115171, "epoch": 1294, "lr": 6.194365604536201e-05}
{"train_loss": 0.09779954701662064, "global_step": 115172, "epoch": 1294, "lr": 6.194309309087388e-05}
{"train_loss": 0.07468672096729279, "global_step": 115173, "epoch": 1294, "lr": 6.194253013478016e-05}
{"train_loss": 0.16096727550029755, "global_step": 115174, "epoch": 1294, "lr": 6.19419671770809e-05}
{"train_loss": 0.11172810196876526, "global_step": 115175, "epoch": 1294, "lr": 6.194140421777622e-05}
{"train_loss": 0.17120903730392456, "global_step": 115176, "epoch": 1294, "lr": 6.194084125686614e-05}
{"train_loss": 0.1519821435213089, "global_step": 115177, "epoch": 1294, "lr": 6.194027829435075e-05}
{"train_loss": 0.13902492821216583, "global_step": 115178, "epoch": 1294, "lr": 6.193971533023014e-05}
{"train_loss": 0.15362487733364105, "global_step": 115179, "epoch": 1294, "lr": 6.19391523645044e-05}
{"train_loss": 0.2887934148311615, "global_step": 115180, "epoch": 1294, "lr": 6.193858939717357e-05}
{"train_loss": 0.15453419089317322, "global_step": 115181, "epoch": 1294, "lr": 6.193802642823773e-05}
{"train_loss": 0.1631065458059311, "global_step": 115182, "epoch": 1294, "lr": 6.1937463457697e-05}
{"train_loss": 0.10590814799070358, "global_step": 115183, "epoch": 1294, "lr": 6.19369004855514e-05}
{"train_loss": 0.15024638175964355, "global_step": 115184, "epoch": 1294, "lr": 6.193633751180103e-05}
{"train_loss": 0.12649887800216675, "global_step": 115185, "epoch": 1294, "lr": 6.193577453644598e-05}
{"train_loss": 0.10004321485757828, "global_step": 115186, "epoch": 1294, "lr": 6.193521155948629e-05}
{"train_loss": 0.1350083351135254, "global_step": 115187, "epoch": 1294, "lr": 6.193464858092208e-05}
{"train_loss": 0.12314505875110626, "global_step": 115188, "epoch": 1294, "lr": 6.19340856007534e-05}
{"train_loss": 0.18597175180912018, "global_step": 115189, "epoch": 1294, "lr": 6.19335226189803e-05}
{"train_loss": 0.12824195623397827, "global_step": 115190, "epoch": 1294, "lr": 6.19329596356029e-05}
{"train_loss": 0.16157443821430206, "global_step": 115191, "epoch": 1294, "lr": 6.193239665062126e-05}
{"train_loss": 0.19258075952529907, "global_step": 115192, "epoch": 1294, "lr": 6.193183366403546e-05}
{"train_loss": 0.08962172269821167, "global_step": 115193, "epoch": 1294, "lr": 6.193127067584556e-05}
{"train_loss": 0.12031575292348862, "global_step": 115194, "epoch": 1294, "lr": 6.193070768605164e-05}
{"train_loss": 0.19322121143341064, "global_step": 115195, "epoch": 1294, "lr": 6.19301446946538e-05}
{"train_loss": 0.17880608141422272, "global_step": 115196, "epoch": 1294, "lr": 6.19295817016521e-05}
{"train_loss": 0.11782698333263397, "global_step": 115197, "epoch": 1294, "lr": 6.19290187070466e-05}
{"train_loss": 0.11167895048856735, "global_step": 115198, "epoch": 1294, "lr": 6.192845571083739e-05}
{"train_loss": 0.08255098015069962, "global_step": 115199, "epoch": 1294, "lr": 6.192789271302455e-05}
{"train_loss": 0.09391994774341583, "global_step": 115200, "epoch": 1294, "lr": 6.192732971360815e-05}
{"train_loss": 0.11016526073217392, "global_step": 115201, "epoch": 1294, "lr": 6.192676671258826e-05}
{"train_loss": 0.13902577757835388, "global_step": 115202, "epoch": 1294, "lr": 6.192620370996496e-05}
{"train_loss": 0.13017617166042328, "global_step": 115203, "epoch": 1294, "lr": 6.192564070573833e-05}
{"train_loss": 0.155369833111763, "global_step": 115204, "epoch": 1294, "lr": 6.192507769990844e-05}
{"train_loss": 0.08851540833711624, "global_step": 115205, "epoch": 1294, "lr": 6.192451469247538e-05}
{"train_loss": 0.09418074786663055, "global_step": 115206, "epoch": 1294, "lr": 6.19239516834392e-05}
{"train_loss": 0.1673532873392105, "global_step": 115207, "epoch": 1294, "lr": 6.19233886728e-05}
{"train_loss": 0.11209607869386673, "global_step": 115208, "epoch": 1294, "lr": 6.192282566055785e-05}
{"train_loss": 0.1149572879076004, "global_step": 115209, "epoch": 1294, "lr": 6.192226264671282e-05}
{"train_loss": 0.16707341372966766, "global_step": 115210, "epoch": 1294, "lr": 6.192169963126497e-05}
{"train_loss": 0.18388910591602325, "global_step": 115211, "epoch": 1294, "lr": 6.19211366142144e-05}
{"train_loss": 0.08735085278749466, "global_step": 115212, "epoch": 1294, "lr": 6.192057359556119e-05}
{"train_loss": 0.09711845964193344, "global_step": 115213, "epoch": 1294, "lr": 6.192001057530539e-05}
{"train_loss": 0.11479553580284119, "global_step": 115214, "epoch": 1294, "lr": 6.191944755344709e-05}
{"train_loss": 0.08803486824035645, "global_step": 115215, "epoch": 1294, "lr": 6.191888452998639e-05}
{"train_loss": 0.10958854854106903, "global_step": 115216, "epoch": 1294, "lr": 6.191832150492331e-05}
{"train_loss": 0.24123190343379974, "global_step": 115217, "epoch": 1294, "lr": 6.191775847825798e-05}
{"train_loss": 0.1470782458782196, "global_step": 115218, "epoch": 1294, "lr": 6.191719544999044e-05}
{"train_loss": 0.13574562966823578, "global_step": 115219, "epoch": 1294, "lr": 6.191663242012078e-05}
{"train_loss": 0.1699296534061432, "global_step": 115220, "epoch": 1294, "lr": 6.191606938864908e-05}
{"train_loss": 0.1442146897315979, "global_step": 115221, "epoch": 1294, "lr": 6.191550635557541e-05}
{"train_loss": 0.13125118613243103, "global_step": 115222, "epoch": 1294, "lr": 6.191494332089983e-05}
{"train_loss": 0.1867091804742813, "global_step": 115223, "epoch": 1294, "lr": 6.191438028462244e-05}
{"train_loss": 0.16808360815048218, "global_step": 115224, "epoch": 1294, "lr": 6.191381724674332e-05}
{"train_loss": 0.191791832447052, "global_step": 115225, "epoch": 1294, "lr": 6.191325420726251e-05}
{"train_loss": 0.13147850334644318, "global_step": 115226, "epoch": 1294, "lr": 6.191269116618013e-05}
{"train_loss": 0.17666935920715332, "global_step": 115227, "epoch": 1294, "lr": 6.191212812349621e-05}
{"train_loss": 0.21061158180236816, "global_step": 115228, "epoch": 1294, "lr": 6.191156507921088e-05}
{"train_loss": 0.14258988201618195, "global_step": 115229, "epoch": 1294, "lr": 6.191100203332416e-05}
{"train_loss": 0.19323725998401642, "global_step": 115230, "epoch": 1294, "lr": 6.191043898583616e-05}
{"train_loss": 0.18679477274417877, "global_step": 115231, "epoch": 1294, "lr": 6.190987593674694e-05}
{"train_loss": 0.1751609444618225, "global_step": 115232, "epoch": 1294, "lr": 6.190931288605659e-05}
{"train_loss": 0.18413566052913666, "global_step": 115233, "epoch": 1294, "lr": 6.190874983376518e-05}
{"train_loss": 0.14942273497581482, "global_step": 115234, "epoch": 1294, "lr": 6.190818677987278e-05}
{"train_loss": 0.15744659304618835, "global_step": 115235, "epoch": 1294, "lr": 6.190762372437947e-05}
{"train_loss": 0.17616726458072662, "global_step": 115236, "epoch": 1294, "lr": 6.190706066728533e-05}
{"train_loss": 0.10918352007865906, "global_step": 115237, "epoch": 1294, "lr": 6.190649760859041e-05}
{"train_loss": 0.12136458605527878, "global_step": 115238, "epoch": 1294, "lr": 6.190593454829483e-05}
{"train_loss": 0.10306118428707123, "global_step": 115239, "epoch": 1294, "lr": 6.190537148639863e-05}
{"train_loss": 0.10059763491153717, "global_step": 115240, "epoch": 1294, "lr": 6.19048084229019e-05}
{"train_loss": 0.15078303217887878, "global_step": 115241, "epoch": 1294, "lr": 6.190424535780471e-05}
{"train_loss": 0.12210837006568909, "global_step": 115242, "epoch": 1294, "lr": 6.190368229110715e-05}
{"train_loss": 0.18546444177627563, "global_step": 115243, "epoch": 1294, "lr": 6.190311922280928e-05}
{"train_loss": 0.14644835889339447, "global_step": 115244, "epoch": 1294, "lr": 6.190255615291119e-05}
{"train_loss": 0.15530924499034882, "global_step": 115245, "epoch": 1294, "lr": 6.190199308141292e-05}
{"train_loss": 0.16268981993198395, "global_step": 115246, "epoch": 1294, "lr": 6.190143000831459e-05}
{"train_loss": 0.219221293926239, "global_step": 115247, "epoch": 1294, "lr": 6.190086693361626e-05}
{"train_loss": 0.0892409011721611, "global_step": 115248, "epoch": 1294, "lr": 6.1900303857318e-05}
{"train_loss": 0.09023946523666382, "global_step": 115249, "epoch": 1294, "lr": 6.189974077941989e-05}
{"train_loss": 0.1505444496870041, "global_step": 115250, "epoch": 1294, "lr": 6.189917769992201e-05}
{"train_loss": 0.19656576216220856, "global_step": 115251, "epoch": 1294, "lr": 6.189861461882442e-05}
{"train_loss": 0.12955497205257416, "global_step": 115252, "epoch": 1294, "lr": 6.189805153612721e-05}
{"train_loss": 0.24227900803089142, "global_step": 115253, "epoch": 1294, "lr": 6.189748845183045e-05}
{"train_loss": 0.1449912724535117, "global_step": 115254, "epoch": 1294, "lr": 6.189692536593422e-05, "val_loss": 5.226573467254639}
{"train_loss": 0.13896793127059937, "global_step": 115255, "epoch": 1295, "lr": 6.18963622784386e-05}
{"train_loss": 0.19868984818458557, "global_step": 115256, "epoch": 1295, "lr": 6.189579918934365e-05}
{"train_loss": 0.17828433215618134, "global_step": 115257, "epoch": 1295, "lr": 6.189523609864945e-05}
{"train_loss": 0.1173788458108902, "global_step": 115258, "epoch": 1295, "lr": 6.18946730063561e-05}
{"train_loss": 0.18326354026794434, "global_step": 115259, "epoch": 1295, "lr": 6.189410991246364e-05}
{"train_loss": 0.07591743022203445, "global_step": 115260, "epoch": 1295, "lr": 6.189354681697217e-05}
{"train_loss": 0.18100789189338684, "global_step": 115261, "epoch": 1295, "lr": 6.189298371988175e-05}
{"train_loss": 0.1277405023574829, "global_step": 115262, "epoch": 1295, "lr": 6.189242062119247e-05}
{"train_loss": 0.1630600094795227, "global_step": 115263, "epoch": 1295, "lr": 6.189185752090439e-05}
{"train_loss": 0.187967449426651, "global_step": 115264, "epoch": 1295, "lr": 6.18912944190176e-05}
{"train_loss": 0.1814325600862503, "global_step": 115265, "epoch": 1295, "lr": 6.189073131553217e-05}
{"train_loss": 0.08511807769536972, "global_step": 115266, "epoch": 1295, "lr": 6.189016821044818e-05}
{"train_loss": 0.0984995886683464, "global_step": 115267, "epoch": 1295, "lr": 6.188960510376571e-05}
{"train_loss": 0.20922519266605377, "global_step": 115268, "epoch": 1295, "lr": 6.188904199548481e-05}
{"train_loss": 0.1155746653676033, "global_step": 115269, "epoch": 1295, "lr": 6.188847888560558e-05}
{"train_loss": 0.14804984629154205, "global_step": 115270, "epoch": 1295, "lr": 6.188791577412809e-05}
{"train_loss": 0.1896885335445404, "global_step": 115271, "epoch": 1295, "lr": 6.18873526610524e-05}
{"train_loss": 0.1411598026752472, "global_step": 115272, "epoch": 1295, "lr": 6.188678954637863e-05}
{"train_loss": 0.12840566039085388, "global_step": 115273, "epoch": 1295, "lr": 6.188622643010681e-05}
{"train_loss": 0.11209622770547867, "global_step": 115274, "epoch": 1295, "lr": 6.188566331223703e-05}
{"train_loss": 0.08826752007007599, "global_step": 115275, "epoch": 1295, "lr": 6.188510019276937e-05}
{"train_loss": 0.14381243288516998, "global_step": 115276, "epoch": 1295, "lr": 6.18845370717039e-05}
{"train_loss": 0.14908252656459808, "global_step": 115277, "epoch": 1295, "lr": 6.188397394904071e-05}
{"train_loss": 0.20217469334602356, "global_step": 115278, "epoch": 1295, "lr": 6.188341082477985e-05}
{"train_loss": 0.20411527156829834, "global_step": 115279, "epoch": 1295, "lr": 6.188284769892142e-05}
{"train_loss": 0.129095658659935, "global_step": 115280, "epoch": 1295, "lr": 6.188228457146547e-05}
{"train_loss": 0.16523876786231995, "global_step": 115281, "epoch": 1295, "lr": 6.188172144241212e-05}
{"train_loss": 0.09351692348718643, "global_step": 115282, "epoch": 1295, "lr": 6.188115831176138e-05}
{"train_loss": 0.12979386746883392, "global_step": 115283, "epoch": 1295, "lr": 6.188059517951339e-05}
{"train_loss": 0.11284798383712769, "global_step": 115284, "epoch": 1295, "lr": 6.188003204566819e-05}
{"train_loss": 0.11581488698720932, "global_step": 115285, "epoch": 1295, "lr": 6.187946891022587e-05}
{"train_loss": 0.11498871445655823, "global_step": 115286, "epoch": 1295, "lr": 6.18789057731865e-05}
{"train_loss": 0.0968368723988533, "global_step": 115287, "epoch": 1295, "lr": 6.187834263455016e-05}
{"train_loss": 0.07915294915437698, "global_step": 115288, "epoch": 1295, "lr": 6.187777949431692e-05}
{"train_loss": 0.09011370688676834, "global_step": 115289, "epoch": 1295, "lr": 6.187721635248685e-05}
{"train_loss": 0.11977630853652954, "global_step": 115290, "epoch": 1295, "lr": 6.187665320906003e-05}
{"train_loss": 0.14367464184761047, "global_step": 115291, "epoch": 1295, "lr": 6.187609006403654e-05}
{"train_loss": 0.10584133863449097, "global_step": 115292, "epoch": 1295, "lr": 6.187552691741646e-05}
{"train_loss": 0.13091020286083221, "global_step": 115293, "epoch": 1295, "lr": 6.187496376919986e-05}
{"train_loss": 0.13307517766952515, "global_step": 115294, "epoch": 1295, "lr": 6.187440061938682e-05}
{"train_loss": 0.11844895780086517, "global_step": 115295, "epoch": 1295, "lr": 6.187383746797741e-05}
{"train_loss": 0.09350104629993439, "global_step": 115296, "epoch": 1295, "lr": 6.18732743149717e-05}
{"train_loss": 0.11602221429347992, "global_step": 115297, "epoch": 1295, "lr": 6.187271116036976e-05}
{"train_loss": 0.1427631825208664, "global_step": 115298, "epoch": 1295, "lr": 6.187214800417171e-05}
{"train_loss": 0.20452502369880676, "global_step": 115299, "epoch": 1295, "lr": 6.187158484637758e-05}
{"train_loss": 0.14455077052116394, "global_step": 115300, "epoch": 1295, "lr": 6.187102168698744e-05}
{"train_loss": 0.12205126136541367, "global_step": 115301, "epoch": 1295, "lr": 6.18704585260014e-05}
{"train_loss": 0.17108123004436493, "global_step": 115302, "epoch": 1295, "lr": 6.186989536341954e-05}
{"train_loss": 0.08832599222660065, "global_step": 115303, "epoch": 1295, "lr": 6.18693321992419e-05}
{"train_loss": 0.11503031104803085, "global_step": 115304, "epoch": 1295, "lr": 6.186876903346858e-05}
{"train_loss": 0.15602439641952515, "global_step": 115305, "epoch": 1295, "lr": 6.186820586609965e-05}
{"train_loss": 0.07553398609161377, "global_step": 115306, "epoch": 1295, "lr": 6.186764269713519e-05}
{"train_loss": 0.15295617282390594, "global_step": 115307, "epoch": 1295, "lr": 6.186707952657526e-05}
{"train_loss": 0.11113498359918594, "global_step": 115308, "epoch": 1295, "lr": 6.186651635441994e-05}
{"train_loss": 0.21691128611564636, "global_step": 115309, "epoch": 1295, "lr": 6.186595318066933e-05}
{"train_loss": 0.13316088914871216, "global_step": 115310, "epoch": 1295, "lr": 6.186539000532346e-05}
{"train_loss": 0.07801298797130585, "global_step": 115311, "epoch": 1295, "lr": 6.186482682838246e-05}
{"train_loss": 0.1578644961118698, "global_step": 115312, "epoch": 1295, "lr": 6.186426364984636e-05}
{"train_loss": 0.10210374742746353, "global_step": 115313, "epoch": 1295, "lr": 6.186370046971526e-05}
{"train_loss": 0.15561318397521973, "global_step": 115314, "epoch": 1295, "lr": 6.186313728798923e-05}
{"train_loss": 0.16533806920051575, "global_step": 115315, "epoch": 1295, "lr": 6.186257410466834e-05}
{"train_loss": 0.12935996055603027, "global_step": 115316, "epoch": 1295, "lr": 6.186201091975268e-05}
{"train_loss": 0.17458538711071014, "global_step": 115317, "epoch": 1295, "lr": 6.186144773324232e-05}
{"train_loss": 0.11237997561693192, "global_step": 115318, "epoch": 1295, "lr": 6.186088454513733e-05}
{"train_loss": 0.22007380425930023, "global_step": 115319, "epoch": 1295, "lr": 6.186032135543776e-05}
{"train_loss": 0.21858438849449158, "global_step": 115320, "epoch": 1295, "lr": 6.185975816414375e-05}
{"train_loss": 0.1557098776102066, "global_step": 115321, "epoch": 1295, "lr": 6.185919497125533e-05}
{"train_loss": 0.12460192292928696, "global_step": 115322, "epoch": 1295, "lr": 6.18586317767726e-05}
{"train_loss": 0.14217710494995117, "global_step": 115323, "epoch": 1295, "lr": 6.18580685806956e-05}
{"train_loss": 0.11262066662311554, "global_step": 115324, "epoch": 1295, "lr": 6.185750538302444e-05}
{"train_loss": 0.0652717873454094, "global_step": 115325, "epoch": 1295, "lr": 6.185694218375918e-05}
{"train_loss": 0.07658006995916367, "global_step": 115326, "epoch": 1295, "lr": 6.18563789828999e-05}
{"train_loss": 0.1206776350736618, "global_step": 115327, "epoch": 1295, "lr": 6.185581578044666e-05}
{"train_loss": 0.13404704630374908, "global_step": 115328, "epoch": 1295, "lr": 6.185525257639957e-05}
{"train_loss": 0.13219581544399261, "global_step": 115329, "epoch": 1295, "lr": 6.185468937075867e-05}
{"train_loss": 0.15708786249160767, "global_step": 115330, "epoch": 1295, "lr": 6.185412616352407e-05}
{"train_loss": 0.15537191927433014, "global_step": 115331, "epoch": 1295, "lr": 6.185356295469582e-05}
{"train_loss": 0.09663465619087219, "global_step": 115332, "epoch": 1295, "lr": 6.1852999744274e-05}
{"train_loss": 0.23052696883678436, "global_step": 115333, "epoch": 1295, "lr": 6.185243653225868e-05}
{"train_loss": 0.211443692445755, "global_step": 115334, "epoch": 1295, "lr": 6.185187331864996e-05}
{"train_loss": 0.1494610458612442, "global_step": 115335, "epoch": 1295, "lr": 6.185131010344789e-05}
{"train_loss": 0.18333208560943604, "global_step": 115336, "epoch": 1295, "lr": 6.185074688665255e-05}
{"train_loss": 0.11672534793615341, "global_step": 115337, "epoch": 1295, "lr": 6.185018366826403e-05}
{"train_loss": 0.2308255434036255, "global_step": 115338, "epoch": 1295, "lr": 6.184962044828239e-05}
{"train_loss": 0.11220480501651764, "global_step": 115339, "epoch": 1295, "lr": 6.184905722670772e-05}
{"train_loss": 0.13193579018115997, "global_step": 115340, "epoch": 1295, "lr": 6.18484940035401e-05}
{"train_loss": 0.2199496328830719, "global_step": 115341, "epoch": 1295, "lr": 6.184793077877958e-05}
{"train_loss": 0.21248990297317505, "global_step": 115342, "epoch": 1295, "lr": 6.184736755242624e-05}
{"train_loss": 0.142091384178467, "global_step": 115343, "epoch": 1295, "lr": 6.184680432448018e-05, "val_loss": 5.144461631774902, "train_action_mse_error": 20.9277286529541}
{"train_loss": 0.19988347589969635, "global_step": 115344, "epoch": 1296, "lr": 6.184624109494147e-05}
{"train_loss": 0.16551704704761505, "global_step": 115345, "epoch": 1296, "lr": 6.184567786381017e-05}
{"train_loss": 0.14783725142478943, "global_step": 115346, "epoch": 1296, "lr": 6.184511463108636e-05}
{"train_loss": 0.13202692568302155, "global_step": 115347, "epoch": 1296, "lr": 6.184455139677012e-05}
{"train_loss": 0.14956359565258026, "global_step": 115348, "epoch": 1296, "lr": 6.184398816086152e-05}
{"train_loss": 0.11873514950275421, "global_step": 115349, "epoch": 1296, "lr": 6.184342492336066e-05}
{"train_loss": 0.10460890829563141, "global_step": 115350, "epoch": 1296, "lr": 6.184286168426757e-05}
{"train_loss": 0.11797425150871277, "global_step": 115351, "epoch": 1296, "lr": 6.184229844358236e-05}
{"train_loss": 0.12259585410356522, "global_step": 115352, "epoch": 1296, "lr": 6.18417352013051e-05}
{"train_loss": 0.14999277889728546, "global_step": 115353, "epoch": 1296, "lr": 6.184117195743586e-05}
{"train_loss": 0.14707526564598083, "global_step": 115354, "epoch": 1296, "lr": 6.184060871197472e-05}
{"train_loss": 0.17740686237812042, "global_step": 115355, "epoch": 1296, "lr": 6.184004546492175e-05}
{"train_loss": 0.2140449583530426, "global_step": 115356, "epoch": 1296, "lr": 6.183948221627703e-05}
{"train_loss": 0.09799064695835114, "global_step": 115357, "epoch": 1296, "lr": 6.183891896604063e-05}
{"train_loss": 0.12645186483860016, "global_step": 115358, "epoch": 1296, "lr": 6.183835571421266e-05}
{"train_loss": 0.10754958540201187, "global_step": 115359, "epoch": 1296, "lr": 6.183779246079314e-05}
{"train_loss": 0.1374310404062271, "global_step": 115360, "epoch": 1296, "lr": 6.183722920578219e-05}
{"train_loss": 0.14183592796325684, "global_step": 115361, "epoch": 1296, "lr": 6.183666594917986e-05}
{"train_loss": 0.16624484956264496, "global_step": 115362, "epoch": 1296, "lr": 6.183610269098623e-05}
{"train_loss": 0.09723643958568573, "global_step": 115363, "epoch": 1296, "lr": 6.18355394312014e-05}
{"train_loss": 0.11177285015583038, "global_step": 115364, "epoch": 1296, "lr": 6.183497616982541e-05}
{"train_loss": 0.14779940247535706, "global_step": 115365, "epoch": 1296, "lr": 6.183441290685837e-05}
{"train_loss": 0.11719691008329391, "global_step": 115366, "epoch": 1296, "lr": 6.183384964230031e-05}
{"train_loss": 0.19115574657917023, "global_step": 115367, "epoch": 1296, "lr": 6.183328637615135e-05}
{"train_loss": 0.13708212971687317, "global_step": 115368, "epoch": 1296, "lr": 6.183272310841154e-05}
{"train_loss": 0.14018678665161133, "global_step": 115369, "epoch": 1296, "lr": 6.183215983908098e-05}
{"train_loss": 0.11695926636457443, "global_step": 115370, "epoch": 1296, "lr": 6.183159656815971e-05}
{"train_loss": 0.13427864015102386, "global_step": 115371, "epoch": 1296, "lr": 6.183103329564784e-05}
{"train_loss": 0.2253027707338333, "global_step": 115372, "epoch": 1296, "lr": 6.183047002154542e-05}
{"train_loss": 0.1727658361196518, "global_step": 115373, "epoch": 1296, "lr": 6.182990674585252e-05}
{"train_loss": 0.15797804296016693, "global_step": 115374, "epoch": 1296, "lr": 6.182934346856927e-05}
{"train_loss": 0.12537549436092377, "global_step": 115375, "epoch": 1296, "lr": 6.182878018969569e-05}
{"train_loss": 0.2017957717180252, "global_step": 115376, "epoch": 1296, "lr": 6.182821690923189e-05}
{"train_loss": 0.12264399230480194, "global_step": 115377, "epoch": 1296, "lr": 6.182765362717791e-05}
{"train_loss": 0.1312623918056488, "global_step": 115378, "epoch": 1296, "lr": 6.182709034353387e-05}
{"train_loss": 0.1364106684923172, "global_step": 115379, "epoch": 1296, "lr": 6.18265270582998e-05}
{"train_loss": 0.15307186543941498, "global_step": 115380, "epoch": 1296, "lr": 6.182596377147582e-05}
{"train_loss": 0.14186346530914307, "global_step": 115381, "epoch": 1296, "lr": 6.182540048306195e-05}
{"train_loss": 0.19943879544734955, "global_step": 115382, "epoch": 1296, "lr": 6.182483719305834e-05}
{"train_loss": 0.1908067911863327, "global_step": 115383, "epoch": 1296, "lr": 6.1824273901465e-05}
{"train_loss": 0.12022089958190918, "global_step": 115384, "epoch": 1296, "lr": 6.182371060828204e-05}
{"train_loss": 0.12538295984268188, "global_step": 115385, "epoch": 1296, "lr": 6.182314731350952e-05}
{"train_loss": 0.124076709151268, "global_step": 115386, "epoch": 1296, "lr": 6.182258401714753e-05}
{"train_loss": 0.14923310279846191, "global_step": 115387, "epoch": 1296, "lr": 6.182202071919611e-05}
{"train_loss": 0.1351630687713623, "global_step": 115388, "epoch": 1296, "lr": 6.182145741965539e-05}
{"train_loss": 0.16808538138866425, "global_step": 115389, "epoch": 1296, "lr": 6.182089411852542e-05}
{"train_loss": 0.23261719942092896, "global_step": 115390, "epoch": 1296, "lr": 6.182033081580627e-05}
{"train_loss": 0.13231942057609558, "global_step": 115391, "epoch": 1296, "lr": 6.1819767511498e-05}
{"train_loss": 0.2081497609615326, "global_step": 115392, "epoch": 1296, "lr": 6.181920420560072e-05}
{"train_loss": 0.12321320921182632, "global_step": 115393, "epoch": 1296, "lr": 6.181864089811451e-05}
{"train_loss": 0.1687900573015213, "global_step": 115394, "epoch": 1296, "lr": 6.181807758903941e-05}
{"train_loss": 0.15124186873435974, "global_step": 115395, "epoch": 1296, "lr": 6.181751427837552e-05}
{"train_loss": 0.13483288884162903, "global_step": 115396, "epoch": 1296, "lr": 6.18169509661229e-05}
{"train_loss": 0.10814268887042999, "global_step": 115397, "epoch": 1296, "lr": 6.181638765228165e-05}
{"train_loss": 0.13600251078605652, "global_step": 115398, "epoch": 1296, "lr": 6.181582433685182e-05}
{"train_loss": 0.11188312619924545, "global_step": 115399, "epoch": 1296, "lr": 6.18152610198335e-05}
{"train_loss": 0.12230765074491501, "global_step": 115400, "epoch": 1296, "lr": 6.181469770122676e-05}
{"train_loss": 0.2042117565870285, "global_step": 115401, "epoch": 1296, "lr": 6.181413438103168e-05}
{"train_loss": 0.1523718386888504, "global_step": 115402, "epoch": 1296, "lr": 6.181357105924834e-05}
{"train_loss": 0.19514986872673035, "global_step": 115403, "epoch": 1296, "lr": 6.18130077358768e-05}
{"train_loss": 0.13883060216903687, "global_step": 115404, "epoch": 1296, "lr": 6.181244441091716e-05}
{"train_loss": 0.1323477178812027, "global_step": 115405, "epoch": 1296, "lr": 6.181188108436947e-05}
{"train_loss": 0.14735233783721924, "global_step": 115406, "epoch": 1296, "lr": 6.181131775623381e-05}
{"train_loss": 0.22906295955181122, "global_step": 115407, "epoch": 1296, "lr": 6.181075442651027e-05}
{"train_loss": 0.12052325904369354, "global_step": 115408, "epoch": 1296, "lr": 6.181019109519892e-05}
{"train_loss": 0.1729414016008377, "global_step": 115409, "epoch": 1296, "lr": 6.18096277622998e-05}
{"train_loss": 0.1438501626253128, "global_step": 115410, "epoch": 1296, "lr": 6.180906442781305e-05}
{"train_loss": 0.08194050192832947, "global_step": 115411, "epoch": 1296, "lr": 6.180850109173873e-05}
{"train_loss": 0.11446699500083923, "global_step": 115412, "epoch": 1296, "lr": 6.180793775407688e-05}
{"train_loss": 0.1641363650560379, "global_step": 115413, "epoch": 1296, "lr": 6.180737441482759e-05}
{"train_loss": 0.1579585075378418, "global_step": 115414, "epoch": 1296, "lr": 6.180681107399095e-05}
{"train_loss": 0.2133583277463913, "global_step": 115415, "epoch": 1296, "lr": 6.180624773156704e-05}
{"train_loss": 0.1456669121980667, "global_step": 115416, "epoch": 1296, "lr": 6.180568438755591e-05}
{"train_loss": 0.15761002898216248, "global_step": 115417, "epoch": 1296, "lr": 6.180512104195766e-05}
{"train_loss": 0.11280959099531174, "global_step": 115418, "epoch": 1296, "lr": 6.180455769477234e-05}
{"train_loss": 0.15702906250953674, "global_step": 115419, "epoch": 1296, "lr": 6.180399434600006e-05}
{"train_loss": 0.12613359093666077, "global_step": 115420, "epoch": 1296, "lr": 6.180343099564087e-05}
{"train_loss": 0.1698392927646637, "global_step": 115421, "epoch": 1296, "lr": 6.180286764369486e-05}
{"train_loss": 0.12973454594612122, "global_step": 115422, "epoch": 1296, "lr": 6.18023042901621e-05}
{"train_loss": 0.14224186539649963, "global_step": 115423, "epoch": 1296, "lr": 6.180174093504265e-05}
{"train_loss": 0.09568561613559723, "global_step": 115424, "epoch": 1296, "lr": 6.18011775783366e-05}
{"train_loss": 0.1621558666229248, "global_step": 115425, "epoch": 1296, "lr": 6.180061422004403e-05}
{"train_loss": 0.18740060925483704, "global_step": 115426, "epoch": 1296, "lr": 6.180005086016501e-05}
{"train_loss": 0.11977693438529968, "global_step": 115427, "epoch": 1296, "lr": 6.179948749869961e-05}
{"train_loss": 0.14046460390090942, "global_step": 115428, "epoch": 1296, "lr": 6.179892413564792e-05}
{"train_loss": 0.10868195444345474, "global_step": 115429, "epoch": 1296, "lr": 6.179836077101e-05}
{"train_loss": 0.1969534158706665, "global_step": 115430, "epoch": 1296, "lr": 6.179779740478595e-05}
{"train_loss": 0.18534722924232483, "global_step": 115431, "epoch": 1296, "lr": 6.179723403697581e-05}
{"train_loss": 0.14791697372546356, "global_step": 115432, "epoch": 1296, "lr": 6.179667066757969e-05, "val_loss": 5.167534828186035}
{"train_loss": 0.18218666315078735, "global_step": 115433, "epoch": 1297, "lr": 6.179610729659766e-05}
{"train_loss": 0.14825057983398438, "global_step": 115434, "epoch": 1297, "lr": 6.179554392402977e-05}
{"train_loss": 0.14439433813095093, "global_step": 115435, "epoch": 1297, "lr": 6.179498054987613e-05}
{"train_loss": 0.08502950519323349, "global_step": 115436, "epoch": 1297, "lr": 6.179441717413677e-05}
{"train_loss": 0.17470508813858032, "global_step": 115437, "epoch": 1297, "lr": 6.179385379681182e-05}
{"train_loss": 0.09271150082349777, "global_step": 115438, "epoch": 1297, "lr": 6.179329041790133e-05}
{"train_loss": 0.15857656300067902, "global_step": 115439, "epoch": 1297, "lr": 6.179272703740536e-05}
{"train_loss": 0.1311095803976059, "global_step": 115440, "epoch": 1297, "lr": 6.179216365532401e-05}
{"train_loss": 0.1531161367893219, "global_step": 115441, "epoch": 1297, "lr": 6.179160027165735e-05}
{"train_loss": 0.11165634542703629, "global_step": 115442, "epoch": 1297, "lr": 6.179103688640546e-05}
{"train_loss": 0.15149149298667908, "global_step": 115443, "epoch": 1297, "lr": 6.17904734995684e-05}
{"train_loss": 0.07353907823562622, "global_step": 115444, "epoch": 1297, "lr": 6.178991011114625e-05}
{"train_loss": 0.1628718227148056, "global_step": 115445, "epoch": 1297, "lr": 6.178934672113909e-05}
{"train_loss": 0.1006283387541771, "global_step": 115446, "epoch": 1297, "lr": 6.178878332954699e-05}
{"train_loss": 0.08651841431856155, "global_step": 115447, "epoch": 1297, "lr": 6.178821993637005e-05}
{"train_loss": 0.15516872704029083, "global_step": 115448, "epoch": 1297, "lr": 6.178765654160832e-05}
{"train_loss": 0.12037330865859985, "global_step": 115449, "epoch": 1297, "lr": 6.178709314526187e-05}
{"train_loss": 0.11617834866046906, "global_step": 115450, "epoch": 1297, "lr": 6.178652974733081e-05}
{"train_loss": 0.17828097939491272, "global_step": 115451, "epoch": 1297, "lr": 6.17859663478152e-05}
{"train_loss": 0.1405622363090515, "global_step": 115452, "epoch": 1297, "lr": 6.178540294671509e-05}
{"train_loss": 0.15666566789150238, "global_step": 115453, "epoch": 1297, "lr": 6.178483954403059e-05}
{"train_loss": 0.16501063108444214, "global_step": 115454, "epoch": 1297, "lr": 6.178427613976175e-05}
{"train_loss": 0.1387535184621811, "global_step": 115455, "epoch": 1297, "lr": 6.178371273390866e-05}
{"train_loss": 0.10298623144626617, "global_step": 115456, "epoch": 1297, "lr": 6.17831493264714e-05}
{"train_loss": 0.144308403134346, "global_step": 115457, "epoch": 1297, "lr": 6.178258591745003e-05}
{"train_loss": 0.1512787640094757, "global_step": 115458, "epoch": 1297, "lr": 6.178202250684464e-05}
{"train_loss": 0.09456854313611984, "global_step": 115459, "epoch": 1297, "lr": 6.17814590946553e-05}
{"train_loss": 0.1440144181251526, "global_step": 115460, "epoch": 1297, "lr": 6.17808956808821e-05}
{"train_loss": 0.174346461892128, "global_step": 115461, "epoch": 1297, "lr": 6.178033226552509e-05}
{"train_loss": 0.08015664666891098, "global_step": 115462, "epoch": 1297, "lr": 6.177976884858435e-05}
{"train_loss": 0.1377953439950943, "global_step": 115463, "epoch": 1297, "lr": 6.177920543005997e-05}
{"train_loss": 0.11755619943141937, "global_step": 115464, "epoch": 1297, "lr": 6.177864200995201e-05}
{"train_loss": 0.17270641028881073, "global_step": 115465, "epoch": 1297, "lr": 6.177807858826056e-05}
{"train_loss": 0.15711243450641632, "global_step": 115466, "epoch": 1297, "lr": 6.17775151649857e-05}
{"train_loss": 0.17476201057434082, "global_step": 115467, "epoch": 1297, "lr": 6.17769517401275e-05}
{"train_loss": 0.14964565634727478, "global_step": 115468, "epoch": 1297, "lr": 6.177638831368603e-05}
{"train_loss": 0.17458593845367432, "global_step": 115469, "epoch": 1297, "lr": 6.177582488566136e-05}
{"train_loss": 0.1273285448551178, "global_step": 115470, "epoch": 1297, "lr": 6.177526145605358e-05}
{"train_loss": 0.11266637593507767, "global_step": 115471, "epoch": 1297, "lr": 6.177469802486276e-05}
{"train_loss": 0.152348130941391, "global_step": 115472, "epoch": 1297, "lr": 6.177413459208896e-05}
{"train_loss": 0.1007915511727333, "global_step": 115473, "epoch": 1297, "lr": 6.17735711577323e-05}
{"train_loss": 0.16496655344963074, "global_step": 115474, "epoch": 1297, "lr": 6.17730077217928e-05}
{"train_loss": 0.12128433585166931, "global_step": 115475, "epoch": 1297, "lr": 6.177244428427056e-05}
{"train_loss": 0.13808873295783997, "global_step": 115476, "epoch": 1297, "lr": 6.177188084516569e-05}
{"train_loss": 0.07127637416124344, "global_step": 115477, "epoch": 1297, "lr": 6.177131740447822e-05}
{"train_loss": 0.0729716494679451, "global_step": 115478, "epoch": 1297, "lr": 6.177075396220823e-05}
{"train_loss": 0.1260770708322525, "global_step": 115479, "epoch": 1297, "lr": 6.177019051835582e-05}
{"train_loss": 0.21032439172267914, "global_step": 115480, "epoch": 1297, "lr": 6.176962707292104e-05}
{"train_loss": 0.15811371803283691, "global_step": 115481, "epoch": 1297, "lr": 6.176906362590398e-05}
{"train_loss": 0.12947426736354828, "global_step": 115482, "epoch": 1297, "lr": 6.17685001773047e-05}
{"train_loss": 0.13050174713134766, "global_step": 115483, "epoch": 1297, "lr": 6.17679367271233e-05}
{"train_loss": 0.13026845455169678, "global_step": 115484, "epoch": 1297, "lr": 6.176737327535985e-05}
{"train_loss": 0.08201635628938675, "global_step": 115485, "epoch": 1297, "lr": 6.176680982201441e-05}
{"train_loss": 0.12564659118652344, "global_step": 115486, "epoch": 1297, "lr": 6.176624636708707e-05}
{"train_loss": 0.16204147040843964, "global_step": 115487, "epoch": 1297, "lr": 6.17656829105779e-05}
{"train_loss": 0.11936906725168228, "global_step": 115488, "epoch": 1297, "lr": 6.176511945248698e-05}
{"train_loss": 0.15069237351417542, "global_step": 115489, "epoch": 1297, "lr": 6.176455599281438e-05}
{"train_loss": 0.09533865749835968, "global_step": 115490, "epoch": 1297, "lr": 6.176399253156019e-05}
{"train_loss": 0.09090648591518402, "global_step": 115491, "epoch": 1297, "lr": 6.176342906872446e-05}
{"train_loss": 0.13883478939533234, "global_step": 115492, "epoch": 1297, "lr": 6.176286560430729e-05}
{"train_loss": 0.15334433317184448, "global_step": 115493, "epoch": 1297, "lr": 6.176230213830875e-05}
{"train_loss": 0.12338794022798538, "global_step": 115494, "epoch": 1297, "lr": 6.176173867072889e-05}
{"train_loss": 0.114476777613163, "global_step": 115495, "epoch": 1297, "lr": 6.176117520156783e-05}
{"train_loss": 0.21354657411575317, "global_step": 115496, "epoch": 1297, "lr": 6.176061173082561e-05}
{"train_loss": 0.08364348858594894, "global_step": 115497, "epoch": 1297, "lr": 6.176004825850233e-05}
{"train_loss": 0.09957750886678696, "global_step": 115498, "epoch": 1297, "lr": 6.175948478459806e-05}
{"train_loss": 0.1710755079984665, "global_step": 115499, "epoch": 1297, "lr": 6.175892130911286e-05}
{"train_loss": 0.18426913022994995, "global_step": 115500, "epoch": 1297, "lr": 6.17583578320468e-05}
{"train_loss": 0.09358358383178711, "global_step": 115501, "epoch": 1297, "lr": 6.17577943534e-05}
{"train_loss": 0.10175049304962158, "global_step": 115502, "epoch": 1297, "lr": 6.17572308731725e-05}
{"train_loss": 0.1329227238893509, "global_step": 115503, "epoch": 1297, "lr": 6.175666739136438e-05}
{"train_loss": 0.21975064277648926, "global_step": 115504, "epoch": 1297, "lr": 6.175610390797571e-05}
{"train_loss": 0.19271694123744965, "global_step": 115505, "epoch": 1297, "lr": 6.175554042300659e-05}
{"train_loss": 0.10353314876556396, "global_step": 115506, "epoch": 1297, "lr": 6.175497693645707e-05}
{"train_loss": 0.1408827155828476, "global_step": 115507, "epoch": 1297, "lr": 6.175441344832724e-05}
{"train_loss": 0.1493944376707077, "global_step": 115508, "epoch": 1297, "lr": 6.175384995861718e-05}
{"train_loss": 0.12506353855133057, "global_step": 115509, "epoch": 1297, "lr": 6.175328646732695e-05}
{"train_loss": 0.15224166214466095, "global_step": 115510, "epoch": 1297, "lr": 6.175272297445663e-05}
{"train_loss": 0.15869058668613434, "global_step": 115511, "epoch": 1297, "lr": 6.175215948000631e-05}
{"train_loss": 0.1777370274066925, "global_step": 115512, "epoch": 1297, "lr": 6.175159598397605e-05}
{"train_loss": 0.14170299470424652, "global_step": 115513, "epoch": 1297, "lr": 6.175103248636593e-05}
{"train_loss": 0.10436344891786575, "global_step": 115514, "epoch": 1297, "lr": 6.175046898717604e-05}
{"train_loss": 0.16877466440200806, "global_step": 115515, "epoch": 1297, "lr": 6.174990548640643e-05}
{"train_loss": 0.3030889332294464, "global_step": 115516, "epoch": 1297, "lr": 6.17493419840572e-05}
{"train_loss": 0.1942216008901596, "global_step": 115517, "epoch": 1297, "lr": 6.17487784801284e-05}
{"train_loss": 0.24910968542099, "global_step": 115518, "epoch": 1297, "lr": 6.174821497462013e-05}
{"train_loss": 0.12133702635765076, "global_step": 115519, "epoch": 1297, "lr": 6.174765146753244e-05}
{"train_loss": 0.2247011363506317, "global_step": 115520, "epoch": 1297, "lr": 6.174708795886544e-05}
{"train_loss": 0.1409803649012962, "global_step": 115521, "epoch": 1297, "lr": 6.174652444861916e-05, "val_loss": 5.15382719039917}
{"train_loss": 0.16060014069080353, "global_step": 115522, "epoch": 1298, "lr": 6.174596093679373e-05}
{"train_loss": 0.08309773355722427, "global_step": 115523, "epoch": 1298, "lr": 6.17453974233892e-05}
{"train_loss": 0.1775631606578827, "global_step": 115524, "epoch": 1298, "lr": 6.174483390840564e-05}
{"train_loss": 0.1522539108991623, "global_step": 115525, "epoch": 1298, "lr": 6.174427039184312e-05}
{"train_loss": 0.1391160488128662, "global_step": 115526, "epoch": 1298, "lr": 6.174370687370172e-05}
{"train_loss": 0.11754114180803299, "global_step": 115527, "epoch": 1298, "lr": 6.174314335398154e-05}
{"train_loss": 0.12157226353883743, "global_step": 115528, "epoch": 1298, "lr": 6.174257983268263e-05}
{"train_loss": 0.20088814198970795, "global_step": 115529, "epoch": 1298, "lr": 6.174201630980508e-05}
{"train_loss": 0.17167744040489197, "global_step": 115530, "epoch": 1298, "lr": 6.174145278534895e-05}
{"train_loss": 0.20708656311035156, "global_step": 115531, "epoch": 1298, "lr": 6.174088925931433e-05}
{"train_loss": 0.1173376515507698, "global_step": 115532, "epoch": 1298, "lr": 6.174032573170128e-05}
{"train_loss": 0.12079987674951553, "global_step": 115533, "epoch": 1298, "lr": 6.17397622025099e-05}
{"train_loss": 0.10652732104063034, "global_step": 115534, "epoch": 1298, "lr": 6.173919867174025e-05}
{"train_loss": 0.15148116648197174, "global_step": 115535, "epoch": 1298, "lr": 6.17386351393924e-05}
{"train_loss": 0.1324857771396637, "global_step": 115536, "epoch": 1298, "lr": 6.173807160546644e-05}
{"train_loss": 0.11987493932247162, "global_step": 115537, "epoch": 1298, "lr": 6.173750806996244e-05}
{"train_loss": 0.15185806155204773, "global_step": 115538, "epoch": 1298, "lr": 6.173694453288046e-05}
{"train_loss": 0.12385587394237518, "global_step": 115539, "epoch": 1298, "lr": 6.173638099422061e-05}
{"train_loss": 0.08298569917678833, "global_step": 115540, "epoch": 1298, "lr": 6.173581745398294e-05}
{"train_loss": 0.17407536506652832, "global_step": 115541, "epoch": 1298, "lr": 6.173525391216754e-05}
{"train_loss": 0.14437335729599, "global_step": 115542, "epoch": 1298, "lr": 6.173469036877447e-05}
{"train_loss": 0.17387093603610992, "global_step": 115543, "epoch": 1298, "lr": 6.173412682380381e-05}
{"train_loss": 0.16961660981178284, "global_step": 115544, "epoch": 1298, "lr": 6.173356327725565e-05}
{"train_loss": 0.16377976536750793, "global_step": 115545, "epoch": 1298, "lr": 6.173299972913004e-05}
{"train_loss": 0.10913238674402237, "global_step": 115546, "epoch": 1298, "lr": 6.173243617942707e-05}
{"train_loss": 0.19360128045082092, "global_step": 115547, "epoch": 1298, "lr": 6.173187262814683e-05}
{"train_loss": 0.19863133132457733, "global_step": 115548, "epoch": 1298, "lr": 6.173130907528939e-05}
{"train_loss": 0.16760824620723724, "global_step": 115549, "epoch": 1298, "lr": 6.17307455208548e-05}
{"train_loss": 0.16701067984104156, "global_step": 115550, "epoch": 1298, "lr": 6.173018196484317e-05}
{"train_loss": 0.09971167147159576, "global_step": 115551, "epoch": 1298, "lr": 6.172961840725455e-05}
{"train_loss": 0.11111341416835785, "global_step": 115552, "epoch": 1298, "lr": 6.172905484808905e-05}
{"train_loss": 0.13084907829761505, "global_step": 115553, "epoch": 1298, "lr": 6.172849128734669e-05}
{"train_loss": 0.11355067044496536, "global_step": 115554, "epoch": 1298, "lr": 6.172792772502758e-05}
{"train_loss": 0.17040793597698212, "global_step": 115555, "epoch": 1298, "lr": 6.17273641611318e-05}
{"train_loss": 0.1480942964553833, "global_step": 115556, "epoch": 1298, "lr": 6.172680059565943e-05}
{"train_loss": 0.08518006652593613, "global_step": 115557, "epoch": 1298, "lr": 6.172623702861052e-05}
{"train_loss": 0.15525707602500916, "global_step": 115558, "epoch": 1298, "lr": 6.172567345998516e-05}
{"train_loss": 0.13292497396469116, "global_step": 115559, "epoch": 1298, "lr": 6.172510988978344e-05}
{"train_loss": 0.09069530665874481, "global_step": 115560, "epoch": 1298, "lr": 6.172454631800541e-05}
{"train_loss": 0.144674152135849, "global_step": 115561, "epoch": 1298, "lr": 6.172398274465115e-05}
{"train_loss": 0.22501209378242493, "global_step": 115562, "epoch": 1298, "lr": 6.172341916972076e-05}
{"train_loss": 0.11864598840475082, "global_step": 115563, "epoch": 1298, "lr": 6.17228555932143e-05}
{"train_loss": 0.12890343368053436, "global_step": 115564, "epoch": 1298, "lr": 6.172229201513183e-05}
{"train_loss": 0.1468050628900528, "global_step": 115565, "epoch": 1298, "lr": 6.172172843547344e-05}
{"train_loss": 0.13382959365844727, "global_step": 115566, "epoch": 1298, "lr": 6.172116485423923e-05}
{"train_loss": 0.08416938781738281, "global_step": 115567, "epoch": 1298, "lr": 6.172060127142924e-05}
{"train_loss": 0.07286258786916733, "global_step": 115568, "epoch": 1298, "lr": 6.172003768704355e-05}
{"train_loss": 0.0930858701467514, "global_step": 115569, "epoch": 1298, "lr": 6.171947410108224e-05}
{"train_loss": 0.17044825851917267, "global_step": 115570, "epoch": 1298, "lr": 6.17189105135454e-05}
{"train_loss": 0.14244745671749115, "global_step": 115571, "epoch": 1298, "lr": 6.17183469244331e-05}
{"train_loss": 0.09132874011993408, "global_step": 115572, "epoch": 1298, "lr": 6.171778333374541e-05}
{"train_loss": 0.09377656131982803, "global_step": 115573, "epoch": 1298, "lr": 6.17172197414824e-05}
{"train_loss": 0.1749611496925354, "global_step": 115574, "epoch": 1298, "lr": 6.171665614764414e-05}
{"train_loss": 0.14173907041549683, "global_step": 115575, "epoch": 1298, "lr": 6.171609255223075e-05}
{"train_loss": 0.1505824774503708, "global_step": 115576, "epoch": 1298, "lr": 6.171552895524225e-05}
{"train_loss": 0.15920546650886536, "global_step": 115577, "epoch": 1298, "lr": 6.171496535667875e-05}
{"train_loss": 0.19722311198711395, "global_step": 115578, "epoch": 1298, "lr": 6.171440175654031e-05}
{"train_loss": 0.1483028680086136, "global_step": 115579, "epoch": 1298, "lr": 6.171383815482701e-05}
{"train_loss": 0.11507191509008408, "global_step": 115580, "epoch": 1298, "lr": 6.171327455153893e-05}
{"train_loss": 0.08902624994516373, "global_step": 115581, "epoch": 1298, "lr": 6.171271094667614e-05}
{"train_loss": 0.10715007781982422, "global_step": 115582, "epoch": 1298, "lr": 6.171214734023873e-05}
{"train_loss": 0.15743155777454376, "global_step": 115583, "epoch": 1298, "lr": 6.171158373222676e-05}
{"train_loss": 0.16357633471488953, "global_step": 115584, "epoch": 1298, "lr": 6.171102012264031e-05}
{"train_loss": 0.13908825814723969, "global_step": 115585, "epoch": 1298, "lr": 6.171045651147945e-05}
{"train_loss": 0.12657447159290314, "global_step": 115586, "epoch": 1298, "lr": 6.170989289874427e-05}
{"train_loss": 0.20108802616596222, "global_step": 115587, "epoch": 1298, "lr": 6.170932928443483e-05}
{"train_loss": 0.1498689502477646, "global_step": 115588, "epoch": 1298, "lr": 6.170876566855121e-05}
{"train_loss": 0.12360583245754242, "global_step": 115589, "epoch": 1298, "lr": 6.17082020510935e-05}
{"train_loss": 0.1368187665939331, "global_step": 115590, "epoch": 1298, "lr": 6.170763843206177e-05}
{"train_loss": 0.11023520678281784, "global_step": 115591, "epoch": 1298, "lr": 6.170707481145607e-05}
{"train_loss": 0.1584322452545166, "global_step": 115592, "epoch": 1298, "lr": 6.170651118927652e-05}
{"train_loss": 0.1345694661140442, "global_step": 115593, "epoch": 1298, "lr": 6.170594756552315e-05}
{"train_loss": 0.20080670714378357, "global_step": 115594, "epoch": 1298, "lr": 6.170538394019608e-05}
{"train_loss": 0.19507743418216705, "global_step": 115595, "epoch": 1298, "lr": 6.170482031329534e-05}
{"train_loss": 0.20260371267795563, "global_step": 115596, "epoch": 1298, "lr": 6.170425668482105e-05}
{"train_loss": 0.12000176310539246, "global_step": 115597, "epoch": 1298, "lr": 6.170369305477325e-05}
{"train_loss": 0.09770487248897552, "global_step": 115598, "epoch": 1298, "lr": 6.170312942315203e-05}
{"train_loss": 0.06038609892129898, "global_step": 115599, "epoch": 1298, "lr": 6.170256578995749e-05}
{"train_loss": 0.18307529389858246, "global_step": 115600, "epoch": 1298, "lr": 6.170200215518967e-05}
{"train_loss": 0.11069054901599884, "global_step": 115601, "epoch": 1298, "lr": 6.170143851884866e-05}
{"train_loss": 0.17589537799358368, "global_step": 115602, "epoch": 1298, "lr": 6.170087488093453e-05}
{"train_loss": 0.11957397311925888, "global_step": 115603, "epoch": 1298, "lr": 6.170031124144735e-05}
{"train_loss": 0.11427050083875656, "global_step": 115604, "epoch": 1298, "lr": 6.169974760038722e-05}
{"train_loss": 0.15737153589725494, "global_step": 115605, "epoch": 1298, "lr": 6.169918395775419e-05}
{"train_loss": 0.08457550406455994, "global_step": 115606, "epoch": 1298, "lr": 6.169862031354836e-05}
{"train_loss": 0.11554449796676636, "global_step": 115607, "epoch": 1298, "lr": 6.169805666776979e-05}
{"train_loss": 0.17117756605148315, "global_step": 115608, "epoch": 1298, "lr": 6.169749302041855e-05}
{"train_loss": 0.16209083795547485, "global_step": 115609, "epoch": 1298, "lr": 6.169692937149473e-05}
{"train_loss": 0.14100723671779203, "global_step": 115610, "epoch": 1298, "lr": 6.169636572099841e-05, "val_loss": 5.303521156311035}
{"train_loss": 0.18616780638694763, "global_step": 115611, "epoch": 1299, "lr": 6.169580206892964e-05}
{"train_loss": 0.11891940981149673, "global_step": 115612, "epoch": 1299, "lr": 6.169523841528853e-05}
{"train_loss": 0.15304571390151978, "global_step": 115613, "epoch": 1299, "lr": 6.169467476007512e-05}
{"train_loss": 0.1371915191411972, "global_step": 115614, "epoch": 1299, "lr": 6.169411110328951e-05}
{"train_loss": 0.1133347749710083, "global_step": 115615, "epoch": 1299, "lr": 6.169354744493177e-05}
{"train_loss": 0.12453664839267731, "global_step": 115616, "epoch": 1299, "lr": 6.169298378500198e-05}
{"train_loss": 0.1121722161769867, "global_step": 115617, "epoch": 1299, "lr": 6.16924201235002e-05}
{"train_loss": 0.12636025249958038, "global_step": 115618, "epoch": 1299, "lr": 6.169185646042653e-05}
{"train_loss": 0.15548500418663025, "global_step": 115619, "epoch": 1299, "lr": 6.169129279578103e-05}
{"train_loss": 0.1289159059524536, "global_step": 115620, "epoch": 1299, "lr": 6.169072912956377e-05}
{"train_loss": 0.18512453138828278, "global_step": 115621, "epoch": 1299, "lr": 6.169016546177484e-05}
{"train_loss": 0.16328376531600952, "global_step": 115622, "epoch": 1299, "lr": 6.168960179241431e-05}
{"train_loss": 0.14710956811904907, "global_step": 115623, "epoch": 1299, "lr": 6.168903812148226e-05}
{"train_loss": 0.09593719244003296, "global_step": 115624, "epoch": 1299, "lr": 6.168847444897874e-05}
{"train_loss": 0.16825270652770996, "global_step": 115625, "epoch": 1299, "lr": 6.168791077490387e-05}
{"train_loss": 0.10241169482469559, "global_step": 115626, "epoch": 1299, "lr": 6.168734709925768e-05}
{"train_loss": 0.1604645699262619, "global_step": 115627, "epoch": 1299, "lr": 6.168678342204029e-05}
{"train_loss": 0.17801445722579956, "global_step": 115628, "epoch": 1299, "lr": 6.168621974325174e-05}
{"train_loss": 0.16646257042884827, "global_step": 115629, "epoch": 1299, "lr": 6.168565606289214e-05}
{"train_loss": 0.1694553792476654, "global_step": 115630, "epoch": 1299, "lr": 6.168509238096153e-05}
{"train_loss": 0.12013083696365356, "global_step": 115631, "epoch": 1299, "lr": 6.168452869746001e-05}
{"train_loss": 0.15146182477474213, "global_step": 115632, "epoch": 1299, "lr": 6.168396501238765e-05}
{"train_loss": 0.06918850541114807, "global_step": 115633, "epoch": 1299, "lr": 6.16834013257445e-05}
{"train_loss": 0.14468754827976227, "global_step": 115634, "epoch": 1299, "lr": 6.168283763753068e-05}
{"train_loss": 0.10843117535114288, "global_step": 115635, "epoch": 1299, "lr": 6.168227394774624e-05}
{"train_loss": 0.11955710500478745, "global_step": 115636, "epoch": 1299, "lr": 6.168171025639127e-05}
{"train_loss": 0.12070389091968536, "global_step": 115637, "epoch": 1299, "lr": 6.168114656346583e-05}
{"train_loss": 0.14306020736694336, "global_step": 115638, "epoch": 1299, "lr": 6.168058286896999e-05}
{"train_loss": 0.13004620373249054, "global_step": 115639, "epoch": 1299, "lr": 6.168001917290384e-05}
{"train_loss": 0.12122587114572525, "global_step": 115640, "epoch": 1299, "lr": 6.167945547526746e-05}
{"train_loss": 0.10855924338102341, "global_step": 115641, "epoch": 1299, "lr": 6.167889177606093e-05}
{"train_loss": 0.10937467962503433, "global_step": 115642, "epoch": 1299, "lr": 6.167832807528431e-05}
{"train_loss": 0.1894524246454239, "global_step": 115643, "epoch": 1299, "lr": 6.167776437293768e-05}
{"train_loss": 0.1232653483748436, "global_step": 115644, "epoch": 1299, "lr": 6.167720066902111e-05}
{"train_loss": 0.15144751965999603, "global_step": 115645, "epoch": 1299, "lr": 6.167663696353468e-05}
{"train_loss": 0.1354135125875473, "global_step": 115646, "epoch": 1299, "lr": 6.167607325647846e-05}
{"train_loss": 0.16442270576953888, "global_step": 115647, "epoch": 1299, "lr": 6.167550954785256e-05}
{"train_loss": 0.17945744097232819, "global_step": 115648, "epoch": 1299, "lr": 6.167494583765702e-05}
{"train_loss": 0.10449046641588211, "global_step": 115649, "epoch": 1299, "lr": 6.167438212589193e-05}
{"train_loss": 0.1825590580701828, "global_step": 115650, "epoch": 1299, "lr": 6.167381841255736e-05}
{"train_loss": 0.12295060604810715, "global_step": 115651, "epoch": 1299, "lr": 6.167325469765338e-05}
{"train_loss": 0.12251512706279755, "global_step": 115652, "epoch": 1299, "lr": 6.167269098118009e-05}
{"train_loss": 0.12955676019191742, "global_step": 115653, "epoch": 1299, "lr": 6.167212726313753e-05}
{"train_loss": 0.16374123096466064, "global_step": 115654, "epoch": 1299, "lr": 6.16715635435258e-05}
{"train_loss": 0.15533575415611267, "global_step": 115655, "epoch": 1299, "lr": 6.167099982234496e-05}
{"train_loss": 0.1969967931509018, "global_step": 115656, "epoch": 1299, "lr": 6.167043609959511e-05}
{"train_loss": 0.16028539836406708, "global_step": 115657, "epoch": 1299, "lr": 6.166987237527631e-05}
{"train_loss": 0.11128932982683182, "global_step": 115658, "epoch": 1299, "lr": 6.166930864938865e-05}
{"train_loss": 0.1269829273223877, "global_step": 115659, "epoch": 1299, "lr": 6.166874492193217e-05}
{"train_loss": 0.1730276495218277, "global_step": 115660, "epoch": 1299, "lr": 6.1668181192907e-05}
{"train_loss": 0.15108038485050201, "global_step": 115661, "epoch": 1299, "lr": 6.166761746231315e-05}
{"train_loss": 0.2004743367433548, "global_step": 115662, "epoch": 1299, "lr": 6.166705373015076e-05}
{"train_loss": 0.1534649133682251, "global_step": 115663, "epoch": 1299, "lr": 6.166648999641985e-05}
{"train_loss": 0.12762656807899475, "global_step": 115664, "epoch": 1299, "lr": 6.166592626112054e-05}
{"train_loss": 0.09673430025577545, "global_step": 115665, "epoch": 1299, "lr": 6.166536252425287e-05}
{"train_loss": 0.15636996924877167, "global_step": 115666, "epoch": 1299, "lr": 6.166479878581696e-05}
{"train_loss": 0.07934270054101944, "global_step": 115667, "epoch": 1299, "lr": 6.166423504581286e-05}
{"train_loss": 0.20573630928993225, "global_step": 115668, "epoch": 1299, "lr": 6.166367130424064e-05}
{"train_loss": 0.16843339800834656, "global_step": 115669, "epoch": 1299, "lr": 6.166310756110038e-05}
{"train_loss": 0.1219768300652504, "global_step": 115670, "epoch": 1299, "lr": 6.166254381639215e-05}
{"train_loss": 0.12204241007566452, "global_step": 115671, "epoch": 1299, "lr": 6.166198007011604e-05}
{"train_loss": 0.14630165696144104, "global_step": 115672, "epoch": 1299, "lr": 6.166141632227211e-05}
{"train_loss": 0.09128402173519135, "global_step": 115673, "epoch": 1299, "lr": 6.166085257286044e-05}
{"train_loss": 0.0979902595281601, "global_step": 115674, "epoch": 1299, "lr": 6.166028882188113e-05}
{"train_loss": 0.16192516684532166, "global_step": 115675, "epoch": 1299, "lr": 6.165972506933423e-05}
{"train_loss": 0.1331082284450531, "global_step": 115676, "epoch": 1299, "lr": 6.165916131521982e-05}
{"train_loss": 0.11391753703355789, "global_step": 115677, "epoch": 1299, "lr": 6.165859755953798e-05}
{"train_loss": 0.15074200928211212, "global_step": 115678, "epoch": 1299, "lr": 6.165803380228878e-05}
{"train_loss": 0.2170676440000534, "global_step": 115679, "epoch": 1299, "lr": 6.16574700434723e-05}
{"train_loss": 0.12480086833238602, "global_step": 115680, "epoch": 1299, "lr": 6.165690628308862e-05}
{"train_loss": 0.1604190617799759, "global_step": 115681, "epoch": 1299, "lr": 6.165634252113779e-05}
{"train_loss": 0.13161492347717285, "global_step": 115682, "epoch": 1299, "lr": 6.165577875761993e-05}
{"train_loss": 0.10307015478610992, "global_step": 115683, "epoch": 1299, "lr": 6.165521499253508e-05}
{"train_loss": 0.14972278475761414, "global_step": 115684, "epoch": 1299, "lr": 6.165465122588335e-05}
{"train_loss": 0.16412244737148285, "global_step": 115685, "epoch": 1299, "lr": 6.165408745766477e-05}
{"train_loss": 0.13335637748241425, "global_step": 115686, "epoch": 1299, "lr": 6.165352368787944e-05}
{"train_loss": 0.11782792210578918, "global_step": 115687, "epoch": 1299, "lr": 6.165295991652746e-05}
{"train_loss": 0.16516214609146118, "global_step": 115688, "epoch": 1299, "lr": 6.165239614360887e-05}
{"train_loss": 0.12404073774814606, "global_step": 115689, "epoch": 1299, "lr": 6.165183236912375e-05}
{"train_loss": 0.11165256798267365, "global_step": 115690, "epoch": 1299, "lr": 6.165126859307219e-05}
{"train_loss": 0.126651331782341, "global_step": 115691, "epoch": 1299, "lr": 6.165070481545426e-05}
{"train_loss": 0.1552247703075409, "global_step": 115692, "epoch": 1299, "lr": 6.165014103627003e-05}
{"train_loss": 0.11425653100013733, "global_step": 115693, "epoch": 1299, "lr": 6.164957725551958e-05}
{"train_loss": 0.1712750792503357, "global_step": 115694, "epoch": 1299, "lr": 6.164901347320299e-05}
{"train_loss": 0.16894248127937317, "global_step": 115695, "epoch": 1299, "lr": 6.164844968932033e-05}
{"train_loss": 0.13424621522426605, "global_step": 115696, "epoch": 1299, "lr": 6.164788590387167e-05}
{"train_loss": 0.1404428482055664, "global_step": 115697, "epoch": 1299, "lr": 6.164732211685711e-05}
{"train_loss": 0.08583303540945053, "global_step": 115698, "epoch": 1299, "lr": 6.16467583282767e-05}
{"train_loss": 0.1408713414762797, "global_step": 115699, "epoch": 1299, "lr": 6.164619453813052e-05, "val_loss": 5.285342693328857}
{"train_loss": 0.1697394996881485, "global_step": 115700, "epoch": 1300, "lr": 6.164563074641865e-05}
{"train_loss": 0.13419249653816223, "global_step": 115701, "epoch": 1300, "lr": 6.164506695314118e-05}
{"train_loss": 0.17230089008808136, "global_step": 115702, "epoch": 1300, "lr": 6.164450315829816e-05}
{"train_loss": 0.1674308329820633, "global_step": 115703, "epoch": 1300, "lr": 6.164393936188968e-05}
{"train_loss": 0.1786017268896103, "global_step": 115704, "epoch": 1300, "lr": 6.164337556391583e-05}
{"train_loss": 0.1684465855360031, "global_step": 115705, "epoch": 1300, "lr": 6.164281176437666e-05}
{"train_loss": 0.10238048434257507, "global_step": 115706, "epoch": 1300, "lr": 6.164224796327226e-05}
{"train_loss": 0.132537841796875, "global_step": 115707, "epoch": 1300, "lr": 6.16416841606027e-05}
{"train_loss": 0.09295154362916946, "global_step": 115708, "epoch": 1300, "lr": 6.164112035636805e-05}
{"train_loss": 0.15987688302993774, "global_step": 115709, "epoch": 1300, "lr": 6.16405565505684e-05}
{"train_loss": 0.15662747621536255, "global_step": 115710, "epoch": 1300, "lr": 6.163999274320382e-05}
{"train_loss": 0.13478946685791016, "global_step": 115711, "epoch": 1300, "lr": 6.163942893427438e-05}
{"train_loss": 0.13883152604103088, "global_step": 115712, "epoch": 1300, "lr": 6.163886512378017e-05}
{"train_loss": 0.15163050591945648, "global_step": 115713, "epoch": 1300, "lr": 6.163830131172124e-05}
{"train_loss": 0.11000657826662064, "global_step": 115714, "epoch": 1300, "lr": 6.163773749809769e-05}
{"train_loss": 0.13235944509506226, "global_step": 115715, "epoch": 1300, "lr": 6.163717368290959e-05}
{"train_loss": 0.1186385527253151, "global_step": 115716, "epoch": 1300, "lr": 6.163660986615701e-05}
{"train_loss": 0.167458638548851, "global_step": 115717, "epoch": 1300, "lr": 6.163604604784004e-05}
{"train_loss": 0.12514697015285492, "global_step": 115718, "epoch": 1300, "lr": 6.163548222795872e-05}
{"train_loss": 0.1788177192211151, "global_step": 115719, "epoch": 1300, "lr": 6.163491840651317e-05}
{"train_loss": 0.16923430562019348, "global_step": 115720, "epoch": 1300, "lr": 6.163435458350346e-05}
{"train_loss": 0.1968071460723877, "global_step": 115721, "epoch": 1300, "lr": 6.163379075892962e-05}
{"train_loss": 0.1372799128293991, "global_step": 115722, "epoch": 1300, "lr": 6.163322693279178e-05}
{"train_loss": 0.14510247111320496, "global_step": 115723, "epoch": 1300, "lr": 6.163266310509e-05}
{"train_loss": 0.08445677906274796, "global_step": 115724, "epoch": 1300, "lr": 6.163209927582435e-05}
{"train_loss": 0.16501250863075256, "global_step": 115725, "epoch": 1300, "lr": 6.16315354449949e-05}
{"train_loss": 0.07110441476106644, "global_step": 115726, "epoch": 1300, "lr": 6.163097161260173e-05}
{"train_loss": 0.11689459532499313, "global_step": 115727, "epoch": 1300, "lr": 6.163040777864492e-05}
{"train_loss": 0.13150694966316223, "global_step": 115728, "epoch": 1300, "lr": 6.162984394312453e-05}
{"train_loss": 0.10055731981992722, "global_step": 115729, "epoch": 1300, "lr": 6.162928010604066e-05}
{"train_loss": 0.21009783446788788, "global_step": 115730, "epoch": 1300, "lr": 6.162871626739338e-05}
{"train_loss": 0.12240061163902283, "global_step": 115731, "epoch": 1300, "lr": 6.162815242718276e-05}
{"train_loss": 0.1270405799150467, "global_step": 115732, "epoch": 1300, "lr": 6.162758858540887e-05}
{"train_loss": 0.13639721274375916, "global_step": 115733, "epoch": 1300, "lr": 6.162702474207179e-05}
{"train_loss": 0.19036239385604858, "global_step": 115734, "epoch": 1300, "lr": 6.16264608971716e-05}
{"train_loss": 0.09457382559776306, "global_step": 115735, "epoch": 1300, "lr": 6.16258970507084e-05}
{"train_loss": 0.12520000338554382, "global_step": 115736, "epoch": 1300, "lr": 6.16253332026822e-05}
{"train_loss": 0.13943061232566833, "global_step": 115737, "epoch": 1300, "lr": 6.162476935309312e-05}
{"train_loss": 0.19771724939346313, "global_step": 115738, "epoch": 1300, "lr": 6.162420550194126e-05}
{"train_loss": 0.10524782538414001, "global_step": 115739, "epoch": 1300, "lr": 6.162364164922665e-05}
{"train_loss": 0.1495550274848938, "global_step": 115740, "epoch": 1300, "lr": 6.162307779494939e-05}
{"train_loss": 0.14217932522296906, "global_step": 115741, "epoch": 1300, "lr": 6.162251393910953e-05}
{"train_loss": 0.13695909082889557, "global_step": 115742, "epoch": 1300, "lr": 6.162195008170718e-05}
{"train_loss": 0.12215989083051682, "global_step": 115743, "epoch": 1300, "lr": 6.162138622274241e-05}
{"train_loss": 0.19990558922290802, "global_step": 115744, "epoch": 1300, "lr": 6.162082236221527e-05}
{"train_loss": 0.16246965527534485, "global_step": 115745, "epoch": 1300, "lr": 6.162025850012586e-05}
{"train_loss": 0.14257174730300903, "global_step": 115746, "epoch": 1300, "lr": 6.161969463647424e-05}
{"train_loss": 0.12396194785833359, "global_step": 115747, "epoch": 1300, "lr": 6.161913077126051e-05}
{"train_loss": 0.24301183223724365, "global_step": 115748, "epoch": 1300, "lr": 6.161856690448472e-05}
{"train_loss": 0.1395954042673111, "global_step": 115749, "epoch": 1300, "lr": 6.161800303614695e-05}
{"train_loss": 0.18536220490932465, "global_step": 115750, "epoch": 1300, "lr": 6.161743916624728e-05}
{"train_loss": 0.11837489902973175, "global_step": 115751, "epoch": 1300, "lr": 6.161687529478579e-05}
{"train_loss": 0.11549971997737885, "global_step": 115752, "epoch": 1300, "lr": 6.161631142176256e-05}
{"train_loss": 0.11515432596206665, "global_step": 115753, "epoch": 1300, "lr": 6.161574754717766e-05}
{"train_loss": 0.10369084775447845, "global_step": 115754, "epoch": 1300, "lr": 6.161518367103114e-05}
{"train_loss": 0.23032979667186737, "global_step": 115755, "epoch": 1300, "lr": 6.161461979332312e-05}
{"train_loss": 0.13174286484718323, "global_step": 115756, "epoch": 1300, "lr": 6.161405591405364e-05}
{"train_loss": 0.12670613825321198, "global_step": 115757, "epoch": 1300, "lr": 6.161349203322282e-05}
{"train_loss": 0.07656314969062805, "global_step": 115758, "epoch": 1300, "lr": 6.161292815083068e-05}
{"train_loss": 0.13027413189411163, "global_step": 115759, "epoch": 1300, "lr": 6.161236426687734e-05}
{"train_loss": 0.14051905274391174, "global_step": 115760, "epoch": 1300, "lr": 6.161180038136286e-05}
{"train_loss": 0.1606741100549698, "global_step": 115761, "epoch": 1300, "lr": 6.16112364942873e-05}
{"train_loss": 0.10640144348144531, "global_step": 115762, "epoch": 1300, "lr": 6.161067260565077e-05}
{"train_loss": 0.11231332272291183, "global_step": 115763, "epoch": 1300, "lr": 6.161010871545332e-05}
{"train_loss": 0.17161110043525696, "global_step": 115764, "epoch": 1300, "lr": 6.160954482369503e-05}
{"train_loss": 0.15683545172214508, "global_step": 115765, "epoch": 1300, "lr": 6.160898093037599e-05}
{"train_loss": 0.12497099488973618, "global_step": 115766, "epoch": 1300, "lr": 6.160841703549625e-05}
{"train_loss": 0.11953794956207275, "global_step": 115767, "epoch": 1300, "lr": 6.160785313905591e-05}
{"train_loss": 0.1351630836725235, "global_step": 115768, "epoch": 1300, "lr": 6.160728924105501e-05}
{"train_loss": 0.12247667461633682, "global_step": 115769, "epoch": 1300, "lr": 6.160672534149367e-05}
{"train_loss": 0.13511651754379272, "global_step": 115770, "epoch": 1300, "lr": 6.160616144037194e-05}
{"train_loss": 0.2331947684288025, "global_step": 115771, "epoch": 1300, "lr": 6.160559753768992e-05}
{"train_loss": 0.08874290436506271, "global_step": 115772, "epoch": 1300, "lr": 6.160503363344765e-05}
{"train_loss": 0.14631657302379608, "global_step": 115773, "epoch": 1300, "lr": 6.160446972764524e-05}
{"train_loss": 0.07591396570205688, "global_step": 115774, "epoch": 1300, "lr": 6.160390582028274e-05}
{"train_loss": 0.16211432218551636, "global_step": 115775, "epoch": 1300, "lr": 6.160334191136024e-05}
{"train_loss": 0.11170247197151184, "global_step": 115776, "epoch": 1300, "lr": 6.16027780008778e-05}
{"train_loss": 0.13338102400302887, "global_step": 115777, "epoch": 1300, "lr": 6.160221408883552e-05}
{"train_loss": 0.13155163824558258, "global_step": 115778, "epoch": 1300, "lr": 6.160165017523347e-05}
{"train_loss": 0.13643434643745422, "global_step": 115779, "epoch": 1300, "lr": 6.16010862600717e-05}
{"train_loss": 0.21591012179851532, "global_step": 115780, "epoch": 1300, "lr": 6.160052234335033e-05}
{"train_loss": 0.12284404039382935, "global_step": 115781, "epoch": 1300, "lr": 6.159995842506939e-05}
{"train_loss": 0.13596545159816742, "global_step": 115782, "epoch": 1300, "lr": 6.159939450522899e-05}
{"train_loss": 0.14235587418079376, "global_step": 115783, "epoch": 1300, "lr": 6.15988305838292e-05}
{"train_loss": 0.1380833387374878, "global_step": 115784, "epoch": 1300, "lr": 6.159826666087008e-05}
{"train_loss": 0.10773246735334396, "global_step": 115785, "epoch": 1300, "lr": 6.159770273635171e-05}
{"train_loss": 0.1737266629934311, "global_step": 115786, "epoch": 1300, "lr": 6.159713881027417e-05}
{"train_loss": 0.12597961723804474, "global_step": 115787, "epoch": 1300, "lr": 6.159657488263754e-05}
{"train_loss": 0.14087558755379045, "global_step": 115788, "epoch": 1300, "lr": 6.15960109534419e-05, "train/sim_max_reward_0": 0.5360665410448233, "train/sim_max_reward_1": 0.9990774001514343, "train/sim_max_reward_2": 0.5469660889794425, "train/sim_max_reward_3": 0.3246123616094556, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9667467702517634, "test/sim_max_reward_4500000": 0.5596710259707786, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9922944099183225, "test/sim_max_reward_4500003": 0.4543891928426298, "test/sim_max_reward_4500004": 0.9510757196112405, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9525412335572508, "test/sim_max_reward_4500008": 0.4589936045937111, "test/sim_max_reward_4500009": 0.05041881506812176, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9649779655015523, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.7141314093929295, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9645120479281603, "test/sim_max_reward_4500016": 0.5029619475202893, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9739589330141291, "test/sim_max_reward_4500020": 0.9508337642525904, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9509935194610847, "test/sim_max_reward_4500023": 0.6299620868547453, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.0, "test/sim_max_reward_4500026": 0.6182482362854625, "test/sim_max_reward_4500027": 0.9838507141191245, "test/sim_max_reward_4500028": 0.49540094070757046, "test/sim_max_reward_4500029": 0.6989590817502572, "test/sim_max_reward_4500030": 0.9967783146952663, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.7110085575360237, "test/sim_max_reward_4500033": 0.9745966320518524, "test/sim_max_reward_4500034": 0.8618578474591183, "test/sim_max_reward_4500035": 0.5483279712453184, "test/sim_max_reward_4500036": 0.9898827645866334, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.671198400486707, "test/sim_max_reward_4500039": 0.954052218085209, "test/sim_max_reward_4500040": 0.9319660320922792, "test/sim_max_reward_4500041": 0.6937975311723201, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.25254352918398254, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.7762525581846141, "test/sim_max_reward_4500046": 0.4003379981891733, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.8316411286040778, "test/sim_max_reward_4500049": 0.7093062270429067, "train/mean_score": 0.7289115270061531, "test/mean_score": 0.6234344471793086, "val_loss": 5.186110496520996, "train_action_mse_error": 22.123336791992188}
{"train_loss": 0.16533899307250977, "global_step": 115789, "epoch": 1301, "lr": 6.159544702268731e-05}
{"train_loss": 0.11341048032045364, "global_step": 115790, "epoch": 1301, "lr": 6.159488309037384e-05}
{"train_loss": 0.1771238148212433, "global_step": 115791, "epoch": 1301, "lr": 6.159431915650158e-05}
{"train_loss": 0.11060483753681183, "global_step": 115792, "epoch": 1301, "lr": 6.159375522107061e-05}
{"train_loss": 0.113868847489357, "global_step": 115793, "epoch": 1301, "lr": 6.159319128408101e-05}
{"train_loss": 0.1009434387087822, "global_step": 115794, "epoch": 1301, "lr": 6.159262734553285e-05}
{"train_loss": 0.0870283767580986, "global_step": 115795, "epoch": 1301, "lr": 6.159206340542619e-05}
{"train_loss": 0.0902014970779419, "global_step": 115796, "epoch": 1301, "lr": 6.159149946376112e-05}
{"train_loss": 0.16946035623550415, "global_step": 115797, "epoch": 1301, "lr": 6.159093552053772e-05}
{"train_loss": 0.08866845816373825, "global_step": 115798, "epoch": 1301, "lr": 6.159037157575607e-05}
{"train_loss": 0.1401517242193222, "global_step": 115799, "epoch": 1301, "lr": 6.158980762941622e-05}
{"train_loss": 0.06454024463891983, "global_step": 115800, "epoch": 1301, "lr": 6.158924368151826e-05}
{"train_loss": 0.1302710920572281, "global_step": 115801, "epoch": 1301, "lr": 6.158867973206227e-05}
{"train_loss": 0.12239176034927368, "global_step": 115802, "epoch": 1301, "lr": 6.158811578104833e-05}
{"train_loss": 0.1672566682100296, "global_step": 115803, "epoch": 1301, "lr": 6.158755182847651e-05}
{"train_loss": 0.15705792605876923, "global_step": 115804, "epoch": 1301, "lr": 6.158698787434689e-05}
{"train_loss": 0.16893836855888367, "global_step": 115805, "epoch": 1301, "lr": 6.158642391865952e-05}
{"train_loss": 0.24156619608402252, "global_step": 115806, "epoch": 1301, "lr": 6.158585996141451e-05}
{"train_loss": 0.13809828460216522, "global_step": 115807, "epoch": 1301, "lr": 6.158529600261192e-05}
{"train_loss": 0.20703965425491333, "global_step": 115808, "epoch": 1301, "lr": 6.158473204225182e-05}
{"train_loss": 0.10102056711912155, "global_step": 115809, "epoch": 1301, "lr": 6.15841680803343e-05}
{"train_loss": 0.1443060040473938, "global_step": 115810, "epoch": 1301, "lr": 6.158360411685944e-05}
{"train_loss": 0.16292117536067963, "global_step": 115811, "epoch": 1301, "lr": 6.15830401518273e-05}
{"train_loss": 0.1805487871170044, "global_step": 115812, "epoch": 1301, "lr": 6.158247618523796e-05}
{"train_loss": 0.09693969786167145, "global_step": 115813, "epoch": 1301, "lr": 6.158191221709149e-05}
{"train_loss": 0.10292141139507294, "global_step": 115814, "epoch": 1301, "lr": 6.158134824738798e-05}
{"train_loss": 0.1702914834022522, "global_step": 115815, "epoch": 1301, "lr": 6.15807842761275e-05}
{"train_loss": 0.1478733867406845, "global_step": 115816, "epoch": 1301, "lr": 6.158022030331012e-05}
{"train_loss": 0.18671660125255585, "global_step": 115817, "epoch": 1301, "lr": 6.157965632893592e-05}
{"train_loss": 0.15660308301448822, "global_step": 115818, "epoch": 1301, "lr": 6.157909235300499e-05}
{"train_loss": 0.09207393229007721, "global_step": 115819, "epoch": 1301, "lr": 6.157852837551737e-05}
{"train_loss": 0.13581582903862, "global_step": 115820, "epoch": 1301, "lr": 6.157796439647317e-05}
{"train_loss": 0.15877588093280792, "global_step": 115821, "epoch": 1301, "lr": 6.157740041587245e-05}
{"train_loss": 0.21544533967971802, "global_step": 115822, "epoch": 1301, "lr": 6.157683643371531e-05}
{"train_loss": 0.09877604991197586, "global_step": 115823, "epoch": 1301, "lr": 6.157627245000179e-05}
{"train_loss": 0.15283796191215515, "global_step": 115824, "epoch": 1301, "lr": 6.157570846473198e-05}
{"train_loss": 0.12221886217594147, "global_step": 115825, "epoch": 1301, "lr": 6.157514447790595e-05}
{"train_loss": 0.15966638922691345, "global_step": 115826, "epoch": 1301, "lr": 6.15745804895238e-05}
{"train_loss": 0.12740467488765717, "global_step": 115827, "epoch": 1301, "lr": 6.157401649958557e-05}
{"train_loss": 0.09449493139982224, "global_step": 115828, "epoch": 1301, "lr": 6.157345250809137e-05}
{"train_loss": 0.1148248165845871, "global_step": 115829, "epoch": 1301, "lr": 6.157288851504124e-05}
{"train_loss": 0.12049925327301025, "global_step": 115830, "epoch": 1301, "lr": 6.15723245204353e-05}
{"train_loss": 0.1377692073583603, "global_step": 115831, "epoch": 1301, "lr": 6.157176052427359e-05}
{"train_loss": 0.1403343826532364, "global_step": 115832, "epoch": 1301, "lr": 6.15711965265562e-05}
{"train_loss": 0.13578203320503235, "global_step": 115833, "epoch": 1301, "lr": 6.15706325272832e-05}
{"train_loss": 0.18765601515769958, "global_step": 115834, "epoch": 1301, "lr": 6.157006852645467e-05}
{"train_loss": 0.13486674427986145, "global_step": 115835, "epoch": 1301, "lr": 6.156950452407068e-05}
{"train_loss": 0.1442820280790329, "global_step": 115836, "epoch": 1301, "lr": 6.156894052013133e-05}
{"train_loss": 0.17166483402252197, "global_step": 115837, "epoch": 1301, "lr": 6.156837651463667e-05}
{"train_loss": 0.15461796522140503, "global_step": 115838, "epoch": 1301, "lr": 6.156781250758678e-05}
{"train_loss": 0.15758980810642242, "global_step": 115839, "epoch": 1301, "lr": 6.156724849898173e-05}
{"train_loss": 0.15723569691181183, "global_step": 115840, "epoch": 1301, "lr": 6.15666844888216e-05}
{"train_loss": 0.15532724559307098, "global_step": 115841, "epoch": 1301, "lr": 6.15661204771065e-05}
{"train_loss": 0.16468344628810883, "global_step": 115842, "epoch": 1301, "lr": 6.156555646383646e-05}
{"train_loss": 0.1264125555753708, "global_step": 115843, "epoch": 1301, "lr": 6.156499244901157e-05}
{"train_loss": 0.1499292254447937, "global_step": 115844, "epoch": 1301, "lr": 6.156442843263191e-05}
{"train_loss": 0.10432979464530945, "global_step": 115845, "epoch": 1301, "lr": 6.156386441469754e-05}
{"train_loss": 0.12969541549682617, "global_step": 115846, "epoch": 1301, "lr": 6.156330039520857e-05}
{"train_loss": 0.16204552352428436, "global_step": 115847, "epoch": 1301, "lr": 6.156273637416505e-05}
{"train_loss": 0.13405585289001465, "global_step": 115848, "epoch": 1301, "lr": 6.156217235156705e-05}
{"train_loss": 0.1300341635942459, "global_step": 115849, "epoch": 1301, "lr": 6.156160832741466e-05}
{"train_loss": 0.18639183044433594, "global_step": 115850, "epoch": 1301, "lr": 6.156104430170796e-05}
{"train_loss": 0.12259622663259506, "global_step": 115851, "epoch": 1301, "lr": 6.156048027444702e-05}
{"train_loss": 0.08605597913265228, "global_step": 115852, "epoch": 1301, "lr": 6.15599162456319e-05}
{"train_loss": 0.13057690858840942, "global_step": 115853, "epoch": 1301, "lr": 6.15593522152627e-05}
{"train_loss": 0.1273086667060852, "global_step": 115854, "epoch": 1301, "lr": 6.155878818333949e-05}
{"train_loss": 0.1995871216058731, "global_step": 115855, "epoch": 1301, "lr": 6.155822414986234e-05}
{"train_loss": 0.10958515852689743, "global_step": 115856, "epoch": 1301, "lr": 6.155766011483132e-05}
{"train_loss": 0.1794952154159546, "global_step": 115857, "epoch": 1301, "lr": 6.155709607824652e-05}
{"train_loss": 0.14059263467788696, "global_step": 115858, "epoch": 1301, "lr": 6.1556532040108e-05}
{"train_loss": 0.12226952612400055, "global_step": 115859, "epoch": 1301, "lr": 6.155596800041587e-05}
{"train_loss": 0.15065521001815796, "global_step": 115860, "epoch": 1301, "lr": 6.155540395917015e-05}
{"train_loss": 0.1626751720905304, "global_step": 115861, "epoch": 1301, "lr": 6.155483991637098e-05}
{"train_loss": 0.13795123994350433, "global_step": 115862, "epoch": 1301, "lr": 6.155427587201837e-05}
{"train_loss": 0.1527034491300583, "global_step": 115863, "epoch": 1301, "lr": 6.155371182611244e-05}
{"train_loss": 0.12059523165225983, "global_step": 115864, "epoch": 1301, "lr": 6.155314777865326e-05}
{"train_loss": 0.1213059201836586, "global_step": 115865, "epoch": 1301, "lr": 6.155258372964089e-05}
{"train_loss": 0.12807227671146393, "global_step": 115866, "epoch": 1301, "lr": 6.155201967907542e-05}
{"train_loss": 0.14339324831962585, "global_step": 115867, "epoch": 1301, "lr": 6.155145562695692e-05}
{"train_loss": 0.11337705701589584, "global_step": 115868, "epoch": 1301, "lr": 6.155089157328549e-05}
{"train_loss": 0.22578942775726318, "global_step": 115869, "epoch": 1301, "lr": 6.155032751806115e-05}
{"train_loss": 0.08819989114999771, "global_step": 115870, "epoch": 1301, "lr": 6.154976346128402e-05}
{"train_loss": 0.11715760082006454, "global_step": 115871, "epoch": 1301, "lr": 6.154919940295418e-05}
{"train_loss": 0.16916076838970184, "global_step": 115872, "epoch": 1301, "lr": 6.154863534307168e-05}
{"train_loss": 0.11318562924861908, "global_step": 115873, "epoch": 1301, "lr": 6.15480712816366e-05}
{"train_loss": 0.08115489035844803, "global_step": 115874, "epoch": 1301, "lr": 6.154750721864904e-05}
{"train_loss": 0.18251222372055054, "global_step": 115875, "epoch": 1301, "lr": 6.154694315410904e-05}
{"train_loss": 0.21648278832435608, "global_step": 115876, "epoch": 1301, "lr": 6.15463790880167e-05}
{"train_loss": 0.1403750239798192, "global_step": 115877, "epoch": 1301, "lr": 6.15458150203721e-05, "val_loss": 5.435366153717041}
{"train_loss": 0.1633598804473877, "global_step": 115878, "epoch": 1302, "lr": 6.15452509511753e-05}
{"train_loss": 0.23346762359142303, "global_step": 115879, "epoch": 1302, "lr": 6.154468688042638e-05}
{"train_loss": 0.08399473875761032, "global_step": 115880, "epoch": 1302, "lr": 6.154412280812541e-05}
{"train_loss": 0.1329195648431778, "global_step": 115881, "epoch": 1302, "lr": 6.154355873427248e-05}
{"train_loss": 0.08533782511949539, "global_step": 115882, "epoch": 1302, "lr": 6.154299465886765e-05}
{"train_loss": 0.18320858478546143, "global_step": 115883, "epoch": 1302, "lr": 6.154243058191103e-05}
{"train_loss": 0.1831585019826889, "global_step": 115884, "epoch": 1302, "lr": 6.154186650340264e-05}
{"train_loss": 0.14896155893802643, "global_step": 115885, "epoch": 1302, "lr": 6.154130242334262e-05}
{"train_loss": 0.13217495381832123, "global_step": 115886, "epoch": 1302, "lr": 6.154073834173099e-05}
{"train_loss": 0.11915122717618942, "global_step": 115887, "epoch": 1302, "lr": 6.154017425856785e-05}
{"train_loss": 0.15788668394088745, "global_step": 115888, "epoch": 1302, "lr": 6.153961017385328e-05}
{"train_loss": 0.16958661377429962, "global_step": 115889, "epoch": 1302, "lr": 6.153904608758735e-05}
{"train_loss": 0.14303463697433472, "global_step": 115890, "epoch": 1302, "lr": 6.153848199977013e-05}
{"train_loss": 0.17175143957138062, "global_step": 115891, "epoch": 1302, "lr": 6.15379179104017e-05}
{"train_loss": 0.13616731762886047, "global_step": 115892, "epoch": 1302, "lr": 6.153735381948214e-05}
{"train_loss": 0.14785930514335632, "global_step": 115893, "epoch": 1302, "lr": 6.153678972701153e-05}
{"train_loss": 0.11028265953063965, "global_step": 115894, "epoch": 1302, "lr": 6.153622563298994e-05}
{"train_loss": 0.08441620320081711, "global_step": 115895, "epoch": 1302, "lr": 6.153566153741744e-05}
{"train_loss": 0.20356906950473785, "global_step": 115896, "epoch": 1302, "lr": 6.153509744029411e-05}
{"train_loss": 0.24576517939567566, "global_step": 115897, "epoch": 1302, "lr": 6.153453334162002e-05}
{"train_loss": 0.18786948919296265, "global_step": 115898, "epoch": 1302, "lr": 6.153396924139527e-05}
{"train_loss": 0.10753294825553894, "global_step": 115899, "epoch": 1302, "lr": 6.153340513961991e-05}
{"train_loss": 0.20994363725185394, "global_step": 115900, "epoch": 1302, "lr": 6.153284103629401e-05}
{"train_loss": 0.13747651875019073, "global_step": 115901, "epoch": 1302, "lr": 6.153227693141769e-05}
{"train_loss": 0.16624970734119415, "global_step": 115902, "epoch": 1302, "lr": 6.153171282499097e-05}
{"train_loss": 0.11655905097723007, "global_step": 115903, "epoch": 1302, "lr": 6.153114871701397e-05}
{"train_loss": 0.12123807519674301, "global_step": 115904, "epoch": 1302, "lr": 6.153058460748673e-05}
{"train_loss": 0.2042870670557022, "global_step": 115905, "epoch": 1302, "lr": 6.153002049640935e-05}
{"train_loss": 0.13214018940925598, "global_step": 115906, "epoch": 1302, "lr": 6.152945638378191e-05}
{"train_loss": 0.22620433568954468, "global_step": 115907, "epoch": 1302, "lr": 6.152889226960446e-05}
{"train_loss": 0.23984394967556, "global_step": 115908, "epoch": 1302, "lr": 6.15283281538771e-05}
{"train_loss": 0.13219545781612396, "global_step": 115909, "epoch": 1302, "lr": 6.15277640365999e-05}
{"train_loss": 0.12782986462116241, "global_step": 115910, "epoch": 1302, "lr": 6.152719991777292e-05}
{"train_loss": 0.13327038288116455, "global_step": 115911, "epoch": 1302, "lr": 6.152663579739626e-05}
{"train_loss": 0.15133310854434967, "global_step": 115912, "epoch": 1302, "lr": 6.152607167546998e-05}
{"train_loss": 0.13120001554489136, "global_step": 115913, "epoch": 1302, "lr": 6.152550755199416e-05}
{"train_loss": 0.16856537759304047, "global_step": 115914, "epoch": 1302, "lr": 6.152494342696887e-05}
{"train_loss": 0.20488779246807098, "global_step": 115915, "epoch": 1302, "lr": 6.152437930039421e-05}
{"train_loss": 0.1449689418077469, "global_step": 115916, "epoch": 1302, "lr": 6.152381517227022e-05}
{"train_loss": 0.15743613243103027, "global_step": 115917, "epoch": 1302, "lr": 6.1523251042597e-05}
{"train_loss": 0.22461432218551636, "global_step": 115918, "epoch": 1302, "lr": 6.152268691137462e-05}
{"train_loss": 0.13195253908634186, "global_step": 115919, "epoch": 1302, "lr": 6.152212277860315e-05}
{"train_loss": 0.1043136939406395, "global_step": 115920, "epoch": 1302, "lr": 6.152155864428267e-05}
{"train_loss": 0.08003439009189606, "global_step": 115921, "epoch": 1302, "lr": 6.152099450841326e-05}
{"train_loss": 0.18364718556404114, "global_step": 115922, "epoch": 1302, "lr": 6.152043037099499e-05}
{"train_loss": 0.10713166743516922, "global_step": 115923, "epoch": 1302, "lr": 6.151986623202795e-05}
{"train_loss": 0.14971871674060822, "global_step": 115924, "epoch": 1302, "lr": 6.15193020915122e-05}
{"train_loss": 0.10112115740776062, "global_step": 115925, "epoch": 1302, "lr": 6.151873794944781e-05}
{"train_loss": 0.06971000134944916, "global_step": 115926, "epoch": 1302, "lr": 6.151817380583488e-05}
{"train_loss": 0.07668834924697876, "global_step": 115927, "epoch": 1302, "lr": 6.151760966067347e-05}
{"train_loss": 0.15536558628082275, "global_step": 115928, "epoch": 1302, "lr": 6.151704551396364e-05}
{"train_loss": 0.16102732717990875, "global_step": 115929, "epoch": 1302, "lr": 6.15164813657055e-05}
{"train_loss": 0.1216479018330574, "global_step": 115930, "epoch": 1302, "lr": 6.151591721589911e-05}
{"train_loss": 0.20309625566005707, "global_step": 115931, "epoch": 1302, "lr": 6.151535306454454e-05}
{"train_loss": 0.17355386912822723, "global_step": 115932, "epoch": 1302, "lr": 6.151478891164187e-05}
{"train_loss": 0.12498922646045685, "global_step": 115933, "epoch": 1302, "lr": 6.151422475719119e-05}
{"train_loss": 0.13622838258743286, "global_step": 115934, "epoch": 1302, "lr": 6.151366060119254e-05}
{"train_loss": 0.11158688366413116, "global_step": 115935, "epoch": 1302, "lr": 6.151309644364603e-05}
{"train_loss": 0.17559610307216644, "global_step": 115936, "epoch": 1302, "lr": 6.151253228455174e-05}
{"train_loss": 0.10895539075136185, "global_step": 115937, "epoch": 1302, "lr": 6.151196812390971e-05}
{"train_loss": 0.1353691816329956, "global_step": 115938, "epoch": 1302, "lr": 6.151140396172003e-05}
{"train_loss": 0.1562035232782364, "global_step": 115939, "epoch": 1302, "lr": 6.15108397979828e-05}
{"train_loss": 0.1558408886194229, "global_step": 115940, "epoch": 1302, "lr": 6.151027563269807e-05}
{"train_loss": 0.1789245307445526, "global_step": 115941, "epoch": 1302, "lr": 6.150971146586593e-05}
{"train_loss": 0.19252222776412964, "global_step": 115942, "epoch": 1302, "lr": 6.150914729748645e-05}
{"train_loss": 0.16558153927326202, "global_step": 115943, "epoch": 1302, "lr": 6.150858312755969e-05}
{"train_loss": 0.12796324491500854, "global_step": 115944, "epoch": 1302, "lr": 6.150801895608576e-05}
{"train_loss": 0.16172432899475098, "global_step": 115945, "epoch": 1302, "lr": 6.150745478306472e-05}
{"train_loss": 0.1465824842453003, "global_step": 115946, "epoch": 1302, "lr": 6.150689060849661e-05}
{"train_loss": 0.11739086359739304, "global_step": 115947, "epoch": 1302, "lr": 6.150632643238156e-05}
{"train_loss": 0.1205279603600502, "global_step": 115948, "epoch": 1302, "lr": 6.150576225471963e-05}
{"train_loss": 0.12827804684638977, "global_step": 115949, "epoch": 1302, "lr": 6.15051980755109e-05}
{"train_loss": 0.11602970957756042, "global_step": 115950, "epoch": 1302, "lr": 6.150463389475541e-05}
{"train_loss": 0.16414980590343475, "global_step": 115951, "epoch": 1302, "lr": 6.150406971245328e-05}
{"train_loss": 0.19884267449378967, "global_step": 115952, "epoch": 1302, "lr": 6.150350552860457e-05}
{"train_loss": 0.13398458063602448, "global_step": 115953, "epoch": 1302, "lr": 6.150294134320934e-05}
{"train_loss": 0.17838940024375916, "global_step": 115954, "epoch": 1302, "lr": 6.150237715626769e-05}
{"train_loss": 0.13330784440040588, "global_step": 115955, "epoch": 1302, "lr": 6.150181296777968e-05}
{"train_loss": 0.15830303728580475, "global_step": 115956, "epoch": 1302, "lr": 6.150124877774539e-05}
{"train_loss": 0.10067695379257202, "global_step": 115957, "epoch": 1302, "lr": 6.15006845861649e-05}
{"train_loss": 0.15543988347053528, "global_step": 115958, "epoch": 1302, "lr": 6.150012039303829e-05}
{"train_loss": 0.17627303302288055, "global_step": 115959, "epoch": 1302, "lr": 6.149955619836562e-05}
{"train_loss": 0.09690596163272858, "global_step": 115960, "epoch": 1302, "lr": 6.149899200214697e-05}
{"train_loss": 0.12033910304307938, "global_step": 115961, "epoch": 1302, "lr": 6.149842780438243e-05}
{"train_loss": 0.21607905626296997, "global_step": 115962, "epoch": 1302, "lr": 6.149786360507207e-05}
{"train_loss": 0.10475325584411621, "global_step": 115963, "epoch": 1302, "lr": 6.149729940421596e-05}
{"train_loss": 0.15604384243488312, "global_step": 115964, "epoch": 1302, "lr": 6.149673520181416e-05}
{"train_loss": 0.15395988523960114, "global_step": 115965, "epoch": 1302, "lr": 6.149617099786678e-05}
{"train_loss": 0.14953023203638163, "global_step": 115966, "epoch": 1302, "lr": 6.149560679237388e-05, "val_loss": 5.1078386306762695}
{"train_loss": 0.19477516412734985, "global_step": 115967, "epoch": 1303, "lr": 6.149504258533553e-05}
{"train_loss": 0.11707113683223724, "global_step": 115968, "epoch": 1303, "lr": 6.149447837675183e-05}
{"train_loss": 0.09084448218345642, "global_step": 115969, "epoch": 1303, "lr": 6.149391416662283e-05}
{"train_loss": 0.14357991516590118, "global_step": 115970, "epoch": 1303, "lr": 6.14933499549486e-05}
{"train_loss": 0.16723133623600006, "global_step": 115971, "epoch": 1303, "lr": 6.149278574172924e-05}
{"train_loss": 0.1151643618941307, "global_step": 115972, "epoch": 1303, "lr": 6.149222152696481e-05}
{"train_loss": 0.1191556304693222, "global_step": 115973, "epoch": 1303, "lr": 6.14916573106554e-05}
{"train_loss": 0.12243921309709549, "global_step": 115974, "epoch": 1303, "lr": 6.149109309280106e-05}
{"train_loss": 0.18500743806362152, "global_step": 115975, "epoch": 1303, "lr": 6.14905288734019e-05}
{"train_loss": 0.1326233148574829, "global_step": 115976, "epoch": 1303, "lr": 6.148996465245796e-05}
{"train_loss": 0.17456743121147156, "global_step": 115977, "epoch": 1303, "lr": 6.148940042996934e-05}
{"train_loss": 0.12397423386573792, "global_step": 115978, "epoch": 1303, "lr": 6.148883620593612e-05}
{"train_loss": 0.10094821453094482, "global_step": 115979, "epoch": 1303, "lr": 6.148827198035835e-05}
{"train_loss": 0.10409433394670486, "global_step": 115980, "epoch": 1303, "lr": 6.148770775323615e-05}
{"train_loss": 0.14357924461364746, "global_step": 115981, "epoch": 1303, "lr": 6.148714352456955e-05}
{"train_loss": 0.1365617960691452, "global_step": 115982, "epoch": 1303, "lr": 6.148657929435863e-05}
{"train_loss": 0.15147145092487335, "global_step": 115983, "epoch": 1303, "lr": 6.14860150626035e-05}
{"train_loss": 0.12825936079025269, "global_step": 115984, "epoch": 1303, "lr": 6.148545082930421e-05}
{"train_loss": 0.12138555943965912, "global_step": 115985, "epoch": 1303, "lr": 6.148488659446085e-05}
{"train_loss": 0.15747414529323578, "global_step": 115986, "epoch": 1303, "lr": 6.148432235807347e-05}
{"train_loss": 0.07184876501560211, "global_step": 115987, "epoch": 1303, "lr": 6.148375812014216e-05}
{"train_loss": 0.11955264955759048, "global_step": 115988, "epoch": 1303, "lr": 6.148319388066701e-05}
{"train_loss": 0.10511273145675659, "global_step": 115989, "epoch": 1303, "lr": 6.14826296396481e-05}
{"train_loss": 0.14918549358844757, "global_step": 115990, "epoch": 1303, "lr": 6.148206539708546e-05}
{"train_loss": 0.08616622537374496, "global_step": 115991, "epoch": 1303, "lr": 6.148150115297921e-05}
{"train_loss": 0.18279051780700684, "global_step": 115992, "epoch": 1303, "lr": 6.148093690732942e-05}
{"train_loss": 0.10869172215461731, "global_step": 115993, "epoch": 1303, "lr": 6.148037266013615e-05}
{"train_loss": 0.10071083158254623, "global_step": 115994, "epoch": 1303, "lr": 6.147980841139948e-05}
{"train_loss": 0.1278563290834427, "global_step": 115995, "epoch": 1303, "lr": 6.147924416111949e-05}
{"train_loss": 0.09721730649471283, "global_step": 115996, "epoch": 1303, "lr": 6.147867990929626e-05}
{"train_loss": 0.20501552522182465, "global_step": 115997, "epoch": 1303, "lr": 6.147811565592986e-05}
{"train_loss": 0.14564745128154755, "global_step": 115998, "epoch": 1303, "lr": 6.147755140102036e-05}
{"train_loss": 0.15190571546554565, "global_step": 115999, "epoch": 1303, "lr": 6.147698714456785e-05}
{"train_loss": 0.15087579190731049, "global_step": 116000, "epoch": 1303, "lr": 6.14764228865724e-05}
{"train_loss": 0.15865390002727509, "global_step": 116001, "epoch": 1303, "lr": 6.147585862703409e-05}
{"train_loss": 0.12794257700443268, "global_step": 116002, "epoch": 1303, "lr": 6.147529436595299e-05}
{"train_loss": 0.12410520017147064, "global_step": 116003, "epoch": 1303, "lr": 6.147473010332915e-05}
{"train_loss": 0.21741899847984314, "global_step": 116004, "epoch": 1303, "lr": 6.14741658391627e-05}
{"train_loss": 0.0942511186003685, "global_step": 116005, "epoch": 1303, "lr": 6.147360157345369e-05}
{"train_loss": 0.2120535522699356, "global_step": 116006, "epoch": 1303, "lr": 6.147303730620218e-05}
{"train_loss": 0.08966188877820969, "global_step": 116007, "epoch": 1303, "lr": 6.147247303740827e-05}
{"train_loss": 0.11879081279039383, "global_step": 116008, "epoch": 1303, "lr": 6.1471908767072e-05}
{"train_loss": 0.15167337656021118, "global_step": 116009, "epoch": 1303, "lr": 6.14713444951935e-05}
{"train_loss": 0.22034011781215668, "global_step": 116010, "epoch": 1303, "lr": 6.147078022177283e-05}
{"train_loss": 0.20609432458877563, "global_step": 116011, "epoch": 1303, "lr": 6.147021594681003e-05}
{"train_loss": 0.1548071801662445, "global_step": 116012, "epoch": 1303, "lr": 6.14696516703052e-05}
{"train_loss": 0.1520242691040039, "global_step": 116013, "epoch": 1303, "lr": 6.146908739225843e-05}
{"train_loss": 0.10683140903711319, "global_step": 116014, "epoch": 1303, "lr": 6.146852311266976e-05}
{"train_loss": 0.12001565843820572, "global_step": 116015, "epoch": 1303, "lr": 6.14679588315393e-05}
{"train_loss": 0.11417386680841446, "global_step": 116016, "epoch": 1303, "lr": 6.14673945488671e-05}
{"train_loss": 0.2327803373336792, "global_step": 116017, "epoch": 1303, "lr": 6.146683026465326e-05}
{"train_loss": 0.09994178265333176, "global_step": 116018, "epoch": 1303, "lr": 6.146626597889785e-05}
{"train_loss": 0.14716191589832306, "global_step": 116019, "epoch": 1303, "lr": 6.146570169160092e-05}
{"train_loss": 0.1674814373254776, "global_step": 116020, "epoch": 1303, "lr": 6.14651374027626e-05}
{"train_loss": 0.08491431176662445, "global_step": 116021, "epoch": 1303, "lr": 6.14645731123829e-05}
{"train_loss": 0.18567167222499847, "global_step": 116022, "epoch": 1303, "lr": 6.146400882046193e-05}
{"train_loss": 0.1469050496816635, "global_step": 116023, "epoch": 1303, "lr": 6.146344452699978e-05}
{"train_loss": 0.16603465378284454, "global_step": 116024, "epoch": 1303, "lr": 6.14628802319965e-05}
{"train_loss": 0.20003312826156616, "global_step": 116025, "epoch": 1303, "lr": 6.146231593545219e-05}
{"train_loss": 0.11289585381746292, "global_step": 116026, "epoch": 1303, "lr": 6.146175163736688e-05}
{"train_loss": 0.07523331791162491, "global_step": 116027, "epoch": 1303, "lr": 6.14611873377407e-05}
{"train_loss": 0.15193982422351837, "global_step": 116028, "epoch": 1303, "lr": 6.146062303657369e-05}
{"train_loss": 0.2524157464504242, "global_step": 116029, "epoch": 1303, "lr": 6.146005873386595e-05}
{"train_loss": 0.1593044251203537, "global_step": 116030, "epoch": 1303, "lr": 6.145949442961756e-05}
{"train_loss": 0.24649950861930847, "global_step": 116031, "epoch": 1303, "lr": 6.145893012382856e-05}
{"train_loss": 0.22249114513397217, "global_step": 116032, "epoch": 1303, "lr": 6.145836581649905e-05}
{"train_loss": 0.13179507851600647, "global_step": 116033, "epoch": 1303, "lr": 6.14578015076291e-05}
{"train_loss": 0.19394002854824066, "global_step": 116034, "epoch": 1303, "lr": 6.14572371972188e-05}
{"train_loss": 0.1783818006515503, "global_step": 116035, "epoch": 1303, "lr": 6.145667288526819e-05}
{"train_loss": 0.2092580497264862, "global_step": 116036, "epoch": 1303, "lr": 6.14561085717774e-05}
{"train_loss": 0.17744548618793488, "global_step": 116037, "epoch": 1303, "lr": 6.145554425674646e-05}
{"train_loss": 0.15938802063465118, "global_step": 116038, "epoch": 1303, "lr": 6.145497994017546e-05}
{"train_loss": 0.1444217562675476, "global_step": 116039, "epoch": 1303, "lr": 6.145441562206447e-05}
{"train_loss": 0.16762511432170868, "global_step": 116040, "epoch": 1303, "lr": 6.14538513024136e-05}
{"train_loss": 0.1007661297917366, "global_step": 116041, "epoch": 1303, "lr": 6.145328698122287e-05}
{"train_loss": 0.1310148686170578, "global_step": 116042, "epoch": 1303, "lr": 6.145272265849241e-05}
{"train_loss": 0.10364492982625961, "global_step": 116043, "epoch": 1303, "lr": 6.145215833422225e-05}
{"train_loss": 0.14699618518352509, "global_step": 116044, "epoch": 1303, "lr": 6.145159400841248e-05}
{"train_loss": 0.16728803515434265, "global_step": 116045, "epoch": 1303, "lr": 6.14510296810632e-05}
{"train_loss": 0.13524478673934937, "global_step": 116046, "epoch": 1303, "lr": 6.145046535217447e-05}
{"train_loss": 0.16682815551757812, "global_step": 116047, "epoch": 1303, "lr": 6.144990102174636e-05}
{"train_loss": 0.15616929531097412, "global_step": 116048, "epoch": 1303, "lr": 6.144933668977895e-05}
{"train_loss": 0.20208200812339783, "global_step": 116049, "epoch": 1303, "lr": 6.144877235627233e-05}
{"train_loss": 0.12222592532634735, "global_step": 116050, "epoch": 1303, "lr": 6.144820802122654e-05}
{"train_loss": 0.20835432410240173, "global_step": 116051, "epoch": 1303, "lr": 6.14476436846417e-05}
{"train_loss": 0.07324907928705215, "global_step": 116052, "epoch": 1303, "lr": 6.144707934651785e-05}
{"train_loss": 0.09815722703933716, "global_step": 116053, "epoch": 1303, "lr": 6.144651500685508e-05}
{"train_loss": 0.11016714572906494, "global_step": 116054, "epoch": 1303, "lr": 6.144595066565348e-05}
{"train_loss": 0.14612019514099936, "global_step": 116055, "epoch": 1303, "lr": 6.144538632291309e-05, "val_loss": 5.055438995361328}
{"train_loss": 0.17987275123596191, "global_step": 116056, "epoch": 1304, "lr": 6.144482197863402e-05}
{"train_loss": 0.09215529263019562, "global_step": 116057, "epoch": 1304, "lr": 6.144425763281633e-05}
{"train_loss": 0.10542862862348557, "global_step": 116058, "epoch": 1304, "lr": 6.144369328546009e-05}
{"train_loss": 0.12870047986507416, "global_step": 116059, "epoch": 1304, "lr": 6.144312893656539e-05}
{"train_loss": 0.10213667899370193, "global_step": 116060, "epoch": 1304, "lr": 6.144256458613231e-05}
{"train_loss": 0.12757349014282227, "global_step": 116061, "epoch": 1304, "lr": 6.144200023416092e-05}
{"train_loss": 0.17101262509822845, "global_step": 116062, "epoch": 1304, "lr": 6.144143588065127e-05}
{"train_loss": 0.19200927019119263, "global_step": 116063, "epoch": 1304, "lr": 6.144087152560346e-05}
{"train_loss": 0.10021589696407318, "global_step": 116064, "epoch": 1304, "lr": 6.144030716901758e-05}
{"train_loss": 0.15974345803260803, "global_step": 116065, "epoch": 1304, "lr": 6.143974281089368e-05}
{"train_loss": 0.13024397194385529, "global_step": 116066, "epoch": 1304, "lr": 6.143917845123187e-05}
{"train_loss": 0.09536336362361908, "global_step": 116067, "epoch": 1304, "lr": 6.143861409003218e-05}
{"train_loss": 0.15928220748901367, "global_step": 116068, "epoch": 1304, "lr": 6.143804972729471e-05}
{"train_loss": 0.24085094034671783, "global_step": 116069, "epoch": 1304, "lr": 6.143748536301954e-05}
{"train_loss": 0.1939854621887207, "global_step": 116070, "epoch": 1304, "lr": 6.143692099720672e-05}
{"train_loss": 0.17867256700992584, "global_step": 116071, "epoch": 1304, "lr": 6.143635662985637e-05}
{"train_loss": 0.12680500745773315, "global_step": 116072, "epoch": 1304, "lr": 6.143579226096854e-05}
{"train_loss": 0.14823755621910095, "global_step": 116073, "epoch": 1304, "lr": 6.143522789054329e-05}
{"train_loss": 0.14806048572063446, "global_step": 116074, "epoch": 1304, "lr": 6.143466351858073e-05}
{"train_loss": 0.14359255135059357, "global_step": 116075, "epoch": 1304, "lr": 6.143409914508091e-05}
{"train_loss": 0.23125684261322021, "global_step": 116076, "epoch": 1304, "lr": 6.143353477004391e-05}
{"train_loss": 0.08299334347248077, "global_step": 116077, "epoch": 1304, "lr": 6.143297039346981e-05}
{"train_loss": 0.1634146273136139, "global_step": 116078, "epoch": 1304, "lr": 6.14324060153587e-05}
{"train_loss": 0.22566691040992737, "global_step": 116079, "epoch": 1304, "lr": 6.143184163571063e-05}
{"train_loss": 0.17164185643196106, "global_step": 116080, "epoch": 1304, "lr": 6.143127725452568e-05}
{"train_loss": 0.26646578311920166, "global_step": 116081, "epoch": 1304, "lr": 6.143071287180396e-05}
{"train_loss": 0.16409017145633698, "global_step": 116082, "epoch": 1304, "lr": 6.143014848754549e-05}
{"train_loss": 0.17466415464878082, "global_step": 116083, "epoch": 1304, "lr": 6.142958410175039e-05}
{"train_loss": 0.12131798267364502, "global_step": 116084, "epoch": 1304, "lr": 6.142901971441873e-05}
{"train_loss": 0.18272149562835693, "global_step": 116085, "epoch": 1304, "lr": 6.142845532555057e-05}
{"train_loss": 0.16252562403678894, "global_step": 116086, "epoch": 1304, "lr": 6.1427890935146e-05}
{"train_loss": 0.1460028439760208, "global_step": 116087, "epoch": 1304, "lr": 6.14273265432051e-05}
{"train_loss": 0.11405805498361588, "global_step": 116088, "epoch": 1304, "lr": 6.142676214972791e-05}
{"train_loss": 0.0860367864370346, "global_step": 116089, "epoch": 1304, "lr": 6.142619775471455e-05}
{"train_loss": 0.1451137363910675, "global_step": 116090, "epoch": 1304, "lr": 6.142563335816507e-05}
{"train_loss": 0.14276093244552612, "global_step": 116091, "epoch": 1304, "lr": 6.142506896007955e-05}
{"train_loss": 0.10896241664886475, "global_step": 116092, "epoch": 1304, "lr": 6.142450456045806e-05}
{"train_loss": 0.1445825845003128, "global_step": 116093, "epoch": 1304, "lr": 6.14239401593007e-05}
{"train_loss": 0.16320930421352386, "global_step": 116094, "epoch": 1304, "lr": 6.142337575660752e-05}
{"train_loss": 0.0586458295583725, "global_step": 116095, "epoch": 1304, "lr": 6.142281135237862e-05}
{"train_loss": 0.15743166208267212, "global_step": 116096, "epoch": 1304, "lr": 6.142224694661405e-05}
{"train_loss": 0.13206788897514343, "global_step": 116097, "epoch": 1304, "lr": 6.142168253931389e-05}
{"train_loss": 0.1330944448709488, "global_step": 116098, "epoch": 1304, "lr": 6.142111813047824e-05}
{"train_loss": 0.19549833238124847, "global_step": 116099, "epoch": 1304, "lr": 6.142055372010715e-05}
{"train_loss": 0.16475576162338257, "global_step": 116100, "epoch": 1304, "lr": 6.141998930820071e-05}
{"train_loss": 0.09951463341712952, "global_step": 116101, "epoch": 1304, "lr": 6.141942489475899e-05}
{"train_loss": 0.13248905539512634, "global_step": 116102, "epoch": 1304, "lr": 6.141886047978207e-05}
{"train_loss": 0.14750218391418457, "global_step": 116103, "epoch": 1304, "lr": 6.141829606327004e-05}
{"train_loss": 0.060710564255714417, "global_step": 116104, "epoch": 1304, "lr": 6.141773164522293e-05}
{"train_loss": 0.14652594923973083, "global_step": 116105, "epoch": 1304, "lr": 6.141716722564086e-05}
{"train_loss": 0.1358473151922226, "global_step": 116106, "epoch": 1304, "lr": 6.141660280452388e-05}
{"train_loss": 0.1133875697851181, "global_step": 116107, "epoch": 1304, "lr": 6.14160383818721e-05}
{"train_loss": 0.1118982806801796, "global_step": 116108, "epoch": 1304, "lr": 6.141547395768557e-05}
{"train_loss": 0.10536554455757141, "global_step": 116109, "epoch": 1304, "lr": 6.141490953196434e-05}
{"train_loss": 0.13499963283538818, "global_step": 116110, "epoch": 1304, "lr": 6.141434510470854e-05}
{"train_loss": 0.1048722043633461, "global_step": 116111, "epoch": 1304, "lr": 6.141378067591821e-05}
{"train_loss": 0.13785621523857117, "global_step": 116112, "epoch": 1304, "lr": 6.141321624559344e-05}
{"train_loss": 0.17523974180221558, "global_step": 116113, "epoch": 1304, "lr": 6.14126518137343e-05}
{"train_loss": 0.08700498938560486, "global_step": 116114, "epoch": 1304, "lr": 6.141208738034087e-05}
{"train_loss": 0.11282232403755188, "global_step": 116115, "epoch": 1304, "lr": 6.141152294541322e-05}
{"train_loss": 0.1741691529750824, "global_step": 116116, "epoch": 1304, "lr": 6.141095850895143e-05}
{"train_loss": 0.11101929098367691, "global_step": 116117, "epoch": 1304, "lr": 6.141039407095556e-05}
{"train_loss": 0.11502772569656372, "global_step": 116118, "epoch": 1304, "lr": 6.140982963142573e-05}
{"train_loss": 0.1282026767730713, "global_step": 116119, "epoch": 1304, "lr": 6.140926519036197e-05}
{"train_loss": 0.18765674531459808, "global_step": 116120, "epoch": 1304, "lr": 6.140870074776439e-05}
{"train_loss": 0.12456069141626358, "global_step": 116121, "epoch": 1304, "lr": 6.140813630363305e-05}
{"train_loss": 0.18916001915931702, "global_step": 116122, "epoch": 1304, "lr": 6.1407571857968e-05}
{"train_loss": 0.12235761433839798, "global_step": 116123, "epoch": 1304, "lr": 6.140700741076936e-05}
{"train_loss": 0.10762462764978409, "global_step": 116124, "epoch": 1304, "lr": 6.140644296203718e-05}
{"train_loss": 0.14176149666309357, "global_step": 116125, "epoch": 1304, "lr": 6.140587851177155e-05}
{"train_loss": 0.15034830570220947, "global_step": 116126, "epoch": 1304, "lr": 6.140531405997254e-05}
{"train_loss": 0.1860595941543579, "global_step": 116127, "epoch": 1304, "lr": 6.140474960664023e-05}
{"train_loss": 0.13190239667892456, "global_step": 116128, "epoch": 1304, "lr": 6.140418515177467e-05}
{"train_loss": 0.14790451526641846, "global_step": 116129, "epoch": 1304, "lr": 6.140362069537598e-05}
{"train_loss": 0.09417454153299332, "global_step": 116130, "epoch": 1304, "lr": 6.14030562374442e-05}
{"train_loss": 0.12357965111732483, "global_step": 116131, "epoch": 1304, "lr": 6.14024917779794e-05}
{"train_loss": 0.11571697890758514, "global_step": 116132, "epoch": 1304, "lr": 6.140192731698168e-05}
{"train_loss": 0.10904818028211594, "global_step": 116133, "epoch": 1304, "lr": 6.140136285445113e-05}
{"train_loss": 0.06497718393802643, "global_step": 116134, "epoch": 1304, "lr": 6.140079839038779e-05}
{"train_loss": 0.17597419023513794, "global_step": 116135, "epoch": 1304, "lr": 6.140023392479176e-05}
{"train_loss": 0.12626877427101135, "global_step": 116136, "epoch": 1304, "lr": 6.13996694576631e-05}
{"train_loss": 0.15304015576839447, "global_step": 116137, "epoch": 1304, "lr": 6.13991049890019e-05}
{"train_loss": 0.1406637728214264, "global_step": 116138, "epoch": 1304, "lr": 6.139854051880822e-05}
{"train_loss": 0.18128474056720734, "global_step": 116139, "epoch": 1304, "lr": 6.139797604708216e-05}
{"train_loss": 0.16780580580234528, "global_step": 116140, "epoch": 1304, "lr": 6.139741157382378e-05}
{"train_loss": 0.1486092507839203, "global_step": 116141, "epoch": 1304, "lr": 6.139684709903314e-05}
{"train_loss": 0.23326119780540466, "global_step": 116142, "epoch": 1304, "lr": 6.139628262271034e-05}
{"train_loss": 0.1828611195087433, "global_step": 116143, "epoch": 1304, "lr": 6.139571814485545e-05}
{"train_loss": 0.1431614546461052, "global_step": 116144, "epoch": 1304, "lr": 6.139515366546855e-05, "val_loss": 5.2185773849487305}
{"train_loss": 0.17939196527004242, "global_step": 116145, "epoch": 1305, "lr": 6.139458918454972e-05}
{"train_loss": 0.10688083618879318, "global_step": 116146, "epoch": 1305, "lr": 6.139402470209901e-05}
{"train_loss": 0.08475848287343979, "global_step": 116147, "epoch": 1305, "lr": 6.139346021811652e-05}
{"train_loss": 0.10394656658172607, "global_step": 116148, "epoch": 1305, "lr": 6.13928957326023e-05}
{"train_loss": 0.10241872072219849, "global_step": 116149, "epoch": 1305, "lr": 6.139233124555647e-05}
{"train_loss": 0.16185323894023895, "global_step": 116150, "epoch": 1305, "lr": 6.139176675697906e-05}
{"train_loss": 0.21988603472709656, "global_step": 116151, "epoch": 1305, "lr": 6.139120226687019e-05}
{"train_loss": 0.12843090295791626, "global_step": 116152, "epoch": 1305, "lr": 6.139063777522989e-05}
{"train_loss": 0.12339925765991211, "global_step": 116153, "epoch": 1305, "lr": 6.139007328205824e-05}
{"train_loss": 0.11674442142248154, "global_step": 116154, "epoch": 1305, "lr": 6.138950878735536e-05}
{"train_loss": 0.1420363038778305, "global_step": 116155, "epoch": 1305, "lr": 6.138894429112129e-05}
{"train_loss": 0.13443949818611145, "global_step": 116156, "epoch": 1305, "lr": 6.138837979335614e-05}
{"train_loss": 0.13648267090320587, "global_step": 116157, "epoch": 1305, "lr": 6.138781529405994e-05}
{"train_loss": 0.0907568484544754, "global_step": 116158, "epoch": 1305, "lr": 6.138725079323278e-05}
{"train_loss": 0.12439759820699692, "global_step": 116159, "epoch": 1305, "lr": 6.138668629087475e-05}
{"train_loss": 0.17446137964725494, "global_step": 116160, "epoch": 1305, "lr": 6.138612178698593e-05}
{"train_loss": 0.11891146004199982, "global_step": 116161, "epoch": 1305, "lr": 6.138555728156636e-05}
{"train_loss": 0.1821482628583908, "global_step": 116162, "epoch": 1305, "lr": 6.138499277461617e-05}
{"train_loss": 0.12126114964485168, "global_step": 116163, "epoch": 1305, "lr": 6.138442826613538e-05}
{"train_loss": 0.1199563667178154, "global_step": 116164, "epoch": 1305, "lr": 6.138386375612411e-05}
{"train_loss": 0.10648664087057114, "global_step": 116165, "epoch": 1305, "lr": 6.138329924458242e-05}
{"train_loss": 0.1320231556892395, "global_step": 116166, "epoch": 1305, "lr": 6.138273473151038e-05}
{"train_loss": 0.10053868591785431, "global_step": 116167, "epoch": 1305, "lr": 6.138217021690806e-05}
{"train_loss": 0.11841509491205215, "global_step": 116168, "epoch": 1305, "lr": 6.138160570077556e-05}
{"train_loss": 0.1391219198703766, "global_step": 116169, "epoch": 1305, "lr": 6.138104118311294e-05}
{"train_loss": 0.18795330822467804, "global_step": 116170, "epoch": 1305, "lr": 6.138047666392027e-05}
{"train_loss": 0.09756457060575485, "global_step": 116171, "epoch": 1305, "lr": 6.137991214319763e-05}
{"train_loss": 0.11684524267911911, "global_step": 116172, "epoch": 1305, "lr": 6.13793476209451e-05}
{"train_loss": 0.17078855633735657, "global_step": 116173, "epoch": 1305, "lr": 6.137878309716277e-05}
{"train_loss": 0.12916597723960876, "global_step": 116174, "epoch": 1305, "lr": 6.137821857185069e-05}
{"train_loss": 0.09902704507112503, "global_step": 116175, "epoch": 1305, "lr": 6.137765404500895e-05}
{"train_loss": 0.07243503630161285, "global_step": 116176, "epoch": 1305, "lr": 6.137708951663762e-05}
{"train_loss": 0.13677287101745605, "global_step": 116177, "epoch": 1305, "lr": 6.137652498673678e-05}
{"train_loss": 0.10920034348964691, "global_step": 116178, "epoch": 1305, "lr": 6.137596045530652e-05}
{"train_loss": 0.1173279881477356, "global_step": 116179, "epoch": 1305, "lr": 6.137539592234688e-05}
{"train_loss": 0.10227206349372864, "global_step": 116180, "epoch": 1305, "lr": 6.137483138785796e-05}
{"train_loss": 0.18542617559432983, "global_step": 116181, "epoch": 1305, "lr": 6.137426685183983e-05}
{"train_loss": 0.10021334141492844, "global_step": 116182, "epoch": 1305, "lr": 6.137370231429257e-05}
{"train_loss": 0.15173178911209106, "global_step": 116183, "epoch": 1305, "lr": 6.137313777521627e-05}
{"train_loss": 0.1224268302321434, "global_step": 116184, "epoch": 1305, "lr": 6.137257323461099e-05}
{"train_loss": 0.18434834480285645, "global_step": 116185, "epoch": 1305, "lr": 6.13720086924768e-05}
{"train_loss": 0.11862249672412872, "global_step": 116186, "epoch": 1305, "lr": 6.137144414881378e-05}
{"train_loss": 0.11550586670637131, "global_step": 116187, "epoch": 1305, "lr": 6.1370879603622e-05}
{"train_loss": 0.15582981705665588, "global_step": 116188, "epoch": 1305, "lr": 6.137031505690155e-05}
{"train_loss": 0.10403093695640564, "global_step": 116189, "epoch": 1305, "lr": 6.136975050865249e-05}
{"train_loss": 0.1147259846329689, "global_step": 116190, "epoch": 1305, "lr": 6.136918595887491e-05}
{"train_loss": 0.15456092357635498, "global_step": 116191, "epoch": 1305, "lr": 6.136862140756889e-05}
{"train_loss": 0.07666119933128357, "global_step": 116192, "epoch": 1305, "lr": 6.13680568547345e-05}
{"train_loss": 0.11727456003427505, "global_step": 116193, "epoch": 1305, "lr": 6.13674923003718e-05}
{"train_loss": 0.1565917283296585, "global_step": 116194, "epoch": 1305, "lr": 6.13669277444809e-05}
{"train_loss": 0.15801222622394562, "global_step": 116195, "epoch": 1305, "lr": 6.136636318706184e-05}
{"train_loss": 0.13873791694641113, "global_step": 116196, "epoch": 1305, "lr": 6.13657986281147e-05}
{"train_loss": 0.25072309374809265, "global_step": 116197, "epoch": 1305, "lr": 6.136523406763959e-05}
{"train_loss": 0.12534929811954498, "global_step": 116198, "epoch": 1305, "lr": 6.136466950563655e-05}
{"train_loss": 0.14720261096954346, "global_step": 116199, "epoch": 1305, "lr": 6.136410494210567e-05}
{"train_loss": 0.20540408790111542, "global_step": 116200, "epoch": 1305, "lr": 6.136354037704703e-05}
{"train_loss": 0.16902640461921692, "global_step": 116201, "epoch": 1305, "lr": 6.136297581046069e-05}
{"train_loss": 0.14965292811393738, "global_step": 116202, "epoch": 1305, "lr": 6.136241124234674e-05}
{"train_loss": 0.1447528600692749, "global_step": 116203, "epoch": 1305, "lr": 6.136184667270526e-05}
{"train_loss": 0.15026593208312988, "global_step": 116204, "epoch": 1305, "lr": 6.136128210153632e-05}
{"train_loss": 0.22702957689762115, "global_step": 116205, "epoch": 1305, "lr": 6.136071752883998e-05}
{"train_loss": 0.13789606094360352, "global_step": 116206, "epoch": 1305, "lr": 6.136015295461635e-05}
{"train_loss": 0.11015112698078156, "global_step": 116207, "epoch": 1305, "lr": 6.135958837886548e-05}
{"train_loss": 0.133814737200737, "global_step": 116208, "epoch": 1305, "lr": 6.135902380158742e-05}
{"train_loss": 0.15074563026428223, "global_step": 116209, "epoch": 1305, "lr": 6.135845922278231e-05}
{"train_loss": 0.1112796887755394, "global_step": 116210, "epoch": 1305, "lr": 6.135789464245019e-05}
{"train_loss": 0.08271341770887375, "global_step": 116211, "epoch": 1305, "lr": 6.135733006059112e-05}
{"train_loss": 0.13783064484596252, "global_step": 116212, "epoch": 1305, "lr": 6.135676547720521e-05}
{"train_loss": 0.1019168347120285, "global_step": 116213, "epoch": 1305, "lr": 6.135620089229252e-05}
{"train_loss": 0.08698530495166779, "global_step": 116214, "epoch": 1305, "lr": 6.135563630585314e-05}
{"train_loss": 0.18593177199363708, "global_step": 116215, "epoch": 1305, "lr": 6.135507171788712e-05}
{"train_loss": 0.10942129790782928, "global_step": 116216, "epoch": 1305, "lr": 6.135450712839456e-05}
{"train_loss": 0.1459013819694519, "global_step": 116217, "epoch": 1305, "lr": 6.135394253737551e-05}
{"train_loss": 0.24844242632389069, "global_step": 116218, "epoch": 1305, "lr": 6.135337794483008e-05}
{"train_loss": 0.10876242816448212, "global_step": 116219, "epoch": 1305, "lr": 6.13528133507583e-05}
{"train_loss": 0.14301683008670807, "global_step": 116220, "epoch": 1305, "lr": 6.135224875516029e-05}
{"train_loss": 0.14743532240390778, "global_step": 116221, "epoch": 1305, "lr": 6.135168415803611e-05}
{"train_loss": 0.1439732164144516, "global_step": 116222, "epoch": 1305, "lr": 6.135111955938583e-05}
{"train_loss": 0.16872499883174896, "global_step": 116223, "epoch": 1305, "lr": 6.135055495920954e-05}
{"train_loss": 0.08653125166893005, "global_step": 116224, "epoch": 1305, "lr": 6.134999035750729e-05}
{"train_loss": 0.1146450862288475, "global_step": 116225, "epoch": 1305, "lr": 6.134942575427917e-05}
{"train_loss": 0.1490531861782074, "global_step": 116226, "epoch": 1305, "lr": 6.134886114952527e-05}
{"train_loss": 0.18735750019550323, "global_step": 116227, "epoch": 1305, "lr": 6.134829654324564e-05}
{"train_loss": 0.15688027441501617, "global_step": 116228, "epoch": 1305, "lr": 6.134773193544039e-05}
{"train_loss": 0.16197822988033295, "global_step": 116229, "epoch": 1305, "lr": 6.134716732610957e-05}
{"train_loss": 0.12118706852197647, "global_step": 116230, "epoch": 1305, "lr": 6.134660271525325e-05}
{"train_loss": 0.1567341387271881, "global_step": 116231, "epoch": 1305, "lr": 6.134603810287153e-05}
{"train_loss": 0.0905856043100357, "global_step": 116232, "epoch": 1305, "lr": 6.134547348896447e-05}
{"train_loss": 0.13609911359093163, "global_step": 116233, "epoch": 1305, "lr": 6.134490887353215e-05, "val_loss": 5.293549060821533, "train_action_mse_error": 20.201541900634766}
{"train_loss": 0.11141600459814072, "global_step": 116234, "epoch": 1306, "lr": 6.134434425657464e-05}
{"train_loss": 0.11206988245248795, "global_step": 116235, "epoch": 1306, "lr": 6.134377963809201e-05}
{"train_loss": 0.1740289330482483, "global_step": 116236, "epoch": 1306, "lr": 6.134321501808438e-05}
{"train_loss": 0.172714501619339, "global_step": 116237, "epoch": 1306, "lr": 6.134265039655176e-05}
{"train_loss": 0.1673387885093689, "global_step": 116238, "epoch": 1306, "lr": 6.134208577349427e-05}
{"train_loss": 0.10960143804550171, "global_step": 116239, "epoch": 1306, "lr": 6.134152114891198e-05}
{"train_loss": 0.1683177500963211, "global_step": 116240, "epoch": 1306, "lr": 6.134095652280496e-05}
{"train_loss": 0.19524025917053223, "global_step": 116241, "epoch": 1306, "lr": 6.134039189517329e-05}
{"train_loss": 0.19335120916366577, "global_step": 116242, "epoch": 1306, "lr": 6.133982726601703e-05}
{"train_loss": 0.19642527401447296, "global_step": 116243, "epoch": 1306, "lr": 6.133926263533629e-05}
{"train_loss": 0.11455459892749786, "global_step": 116244, "epoch": 1306, "lr": 6.133869800313111e-05}
{"train_loss": 0.12488581240177155, "global_step": 116245, "epoch": 1306, "lr": 6.133813336940157e-05}
{"train_loss": 0.10175497829914093, "global_step": 116246, "epoch": 1306, "lr": 6.133756873414777e-05}
{"train_loss": 0.0993027314543724, "global_step": 116247, "epoch": 1306, "lr": 6.133700409736978e-05}
{"train_loss": 0.16539953649044037, "global_step": 116248, "epoch": 1306, "lr": 6.133643945906766e-05}
{"train_loss": 0.16057182848453522, "global_step": 116249, "epoch": 1306, "lr": 6.133587481924148e-05}
{"train_loss": 0.17174412310123444, "global_step": 116250, "epoch": 1306, "lr": 6.133531017789134e-05}
{"train_loss": 0.12469181418418884, "global_step": 116251, "epoch": 1306, "lr": 6.133474553501731e-05}
{"train_loss": 0.1649525910615921, "global_step": 116252, "epoch": 1306, "lr": 6.133418089061947e-05}
{"train_loss": 0.17049430310726166, "global_step": 116253, "epoch": 1306, "lr": 6.133361624469787e-05}
{"train_loss": 0.10175690054893494, "global_step": 116254, "epoch": 1306, "lr": 6.13330515972526e-05}
{"train_loss": 0.16521859169006348, "global_step": 116255, "epoch": 1306, "lr": 6.133248694828375e-05}
{"train_loss": 0.19944721460342407, "global_step": 116256, "epoch": 1306, "lr": 6.133192229779139e-05}
{"train_loss": 0.26478713750839233, "global_step": 116257, "epoch": 1306, "lr": 6.133135764577558e-05}
{"train_loss": 0.11922068893909454, "global_step": 116258, "epoch": 1306, "lr": 6.133079299223642e-05}
{"train_loss": 0.17961542308330536, "global_step": 116259, "epoch": 1306, "lr": 6.133022833717395e-05}
{"train_loss": 0.13085344433784485, "global_step": 116260, "epoch": 1306, "lr": 6.132966368058829e-05}
{"train_loss": 0.09284906089305878, "global_step": 116261, "epoch": 1306, "lr": 6.13290990224795e-05}
{"train_loss": 0.11945407837629318, "global_step": 116262, "epoch": 1306, "lr": 6.132853436284764e-05}
{"train_loss": 0.12046672403812408, "global_step": 116263, "epoch": 1306, "lr": 6.13279697016928e-05}
{"train_loss": 0.11754658073186874, "global_step": 116264, "epoch": 1306, "lr": 6.132740503901504e-05}
{"train_loss": 0.11810796707868576, "global_step": 116265, "epoch": 1306, "lr": 6.132684037481446e-05}
{"train_loss": 0.17861612141132355, "global_step": 116266, "epoch": 1306, "lr": 6.132627570909113e-05}
{"train_loss": 0.16095460951328278, "global_step": 116267, "epoch": 1306, "lr": 6.13257110418451e-05}
{"train_loss": 0.1272851526737213, "global_step": 116268, "epoch": 1306, "lr": 6.132514637307649e-05}
{"train_loss": 0.14862458407878876, "global_step": 116269, "epoch": 1306, "lr": 6.132458170278532e-05}
{"train_loss": 0.17368319630622864, "global_step": 116270, "epoch": 1306, "lr": 6.132401703097172e-05}
{"train_loss": 0.09854038059711456, "global_step": 116271, "epoch": 1306, "lr": 6.132345235763575e-05}
{"train_loss": 0.13410644233226776, "global_step": 116272, "epoch": 1306, "lr": 6.132288768277747e-05}
{"train_loss": 0.1890673041343689, "global_step": 116273, "epoch": 1306, "lr": 6.132232300639696e-05}
{"train_loss": 0.22089701890945435, "global_step": 116274, "epoch": 1306, "lr": 6.132175832849431e-05}
{"train_loss": 0.10723669826984406, "global_step": 116275, "epoch": 1306, "lr": 6.13211936490696e-05}
{"train_loss": 0.12864305078983307, "global_step": 116276, "epoch": 1306, "lr": 6.132062896812287e-05}
{"train_loss": 0.11332787573337555, "global_step": 116277, "epoch": 1306, "lr": 6.132006428565424e-05}
{"train_loss": 0.11533938348293304, "global_step": 116278, "epoch": 1306, "lr": 6.131949960166374e-05}
{"train_loss": 0.29684239625930786, "global_step": 116279, "epoch": 1306, "lr": 6.131893491615149e-05}
{"train_loss": 0.15190069377422333, "global_step": 116280, "epoch": 1306, "lr": 6.131837022911754e-05}
{"train_loss": 0.1382044404745102, "global_step": 116281, "epoch": 1306, "lr": 6.131780554056198e-05}
{"train_loss": 0.1554451733827591, "global_step": 116282, "epoch": 1306, "lr": 6.131724085048488e-05}
{"train_loss": 0.1714363694190979, "global_step": 116283, "epoch": 1306, "lr": 6.131667615888631e-05}
{"train_loss": 0.07544378191232681, "global_step": 116284, "epoch": 1306, "lr": 6.131611146576635e-05}
{"train_loss": 0.11097977310419083, "global_step": 116285, "epoch": 1306, "lr": 6.131554677112507e-05}
{"train_loss": 0.10571946203708649, "global_step": 116286, "epoch": 1306, "lr": 6.131498207496257e-05}
{"train_loss": 0.15055397152900696, "global_step": 116287, "epoch": 1306, "lr": 6.13144173772789e-05}
{"train_loss": 0.18305538594722748, "global_step": 116288, "epoch": 1306, "lr": 6.131385267807414e-05}
{"train_loss": 0.14964857697486877, "global_step": 116289, "epoch": 1306, "lr": 6.131328797734838e-05}
{"train_loss": 0.09333784878253937, "global_step": 116290, "epoch": 1306, "lr": 6.131272327510167e-05}
{"train_loss": 0.07686247676610947, "global_step": 116291, "epoch": 1306, "lr": 6.131215857133412e-05}
{"train_loss": 0.12304557114839554, "global_step": 116292, "epoch": 1306, "lr": 6.131159386604579e-05}
{"train_loss": 0.2136884331703186, "global_step": 116293, "epoch": 1306, "lr": 6.131102915923675e-05}
{"train_loss": 0.11865898221731186, "global_step": 116294, "epoch": 1306, "lr": 6.131046445090706e-05}
{"train_loss": 0.15741634368896484, "global_step": 116295, "epoch": 1306, "lr": 6.130989974105682e-05}
{"train_loss": 0.13494156301021576, "global_step": 116296, "epoch": 1306, "lr": 6.130933502968611e-05}
{"train_loss": 0.10190730541944504, "global_step": 116297, "epoch": 1306, "lr": 6.130877031679501e-05}
{"train_loss": 0.13423438370227814, "global_step": 116298, "epoch": 1306, "lr": 6.130820560238358e-05}
{"train_loss": 0.16022989153862, "global_step": 116299, "epoch": 1306, "lr": 6.130764088645189e-05}
{"train_loss": 0.10003717988729477, "global_step": 116300, "epoch": 1306, "lr": 6.130707616900004e-05}
{"train_loss": 0.21418388187885284, "global_step": 116301, "epoch": 1306, "lr": 6.130651145002807e-05}
{"train_loss": 0.12600882351398468, "global_step": 116302, "epoch": 1306, "lr": 6.130594672953609e-05}
{"train_loss": 0.10966107994318008, "global_step": 116303, "epoch": 1306, "lr": 6.130538200752416e-05}
{"train_loss": 0.11695516854524612, "global_step": 116304, "epoch": 1306, "lr": 6.130481728399236e-05}
{"train_loss": 0.13875143229961395, "global_step": 116305, "epoch": 1306, "lr": 6.130425255894077e-05}
{"train_loss": 0.09393379837274551, "global_step": 116306, "epoch": 1306, "lr": 6.130368783236946e-05}
{"train_loss": 0.13722360134124756, "global_step": 116307, "epoch": 1306, "lr": 6.13031231042785e-05}
{"train_loss": 0.1567440927028656, "global_step": 116308, "epoch": 1306, "lr": 6.130255837466797e-05}
{"train_loss": 0.08903513848781586, "global_step": 116309, "epoch": 1306, "lr": 6.130199364353796e-05}
{"train_loss": 0.08560562878847122, "global_step": 116310, "epoch": 1306, "lr": 6.130142891088854e-05}
{"train_loss": 0.14670905470848083, "global_step": 116311, "epoch": 1306, "lr": 6.130086417671977e-05}
{"train_loss": 0.11460825055837631, "global_step": 116312, "epoch": 1306, "lr": 6.130029944103173e-05}
{"train_loss": 0.13840535283088684, "global_step": 116313, "epoch": 1306, "lr": 6.129973470382451e-05}
{"train_loss": 0.13424207270145416, "global_step": 116314, "epoch": 1306, "lr": 6.129916996509819e-05}
{"train_loss": 0.10867518186569214, "global_step": 116315, "epoch": 1306, "lr": 6.129860522485281e-05}
{"train_loss": 0.0698687881231308, "global_step": 116316, "epoch": 1306, "lr": 6.129804048308847e-05}
{"train_loss": 0.0903397873044014, "global_step": 116317, "epoch": 1306, "lr": 6.129747573980526e-05}
{"train_loss": 0.10652370005846024, "global_step": 116318, "epoch": 1306, "lr": 6.129691099500326e-05}
{"train_loss": 0.11726690828800201, "global_step": 116319, "epoch": 1306, "lr": 6.129634624868251e-05}
{"train_loss": 0.110081247985363, "global_step": 116320, "epoch": 1306, "lr": 6.12957815008431e-05}
{"train_loss": 0.17325501143932343, "global_step": 116321, "epoch": 1306, "lr": 6.12952167514851e-05}
{"train_loss": 0.14036070974020476, "global_step": 116322, "epoch": 1306, "lr": 6.129465200060861e-05, "val_loss": 5.22802734375}
{"train_loss": 0.10792331397533417, "global_step": 116323, "epoch": 1307, "lr": 6.12940872482137e-05}
{"train_loss": 0.11331336200237274, "global_step": 116324, "epoch": 1307, "lr": 6.129352249430043e-05}
{"train_loss": 0.09279782325029373, "global_step": 116325, "epoch": 1307, "lr": 6.129295773886887e-05}
{"train_loss": 0.1564313918352127, "global_step": 116326, "epoch": 1307, "lr": 6.129239298191912e-05}
{"train_loss": 0.1171642541885376, "global_step": 116327, "epoch": 1307, "lr": 6.129182822345125e-05}
{"train_loss": 0.11819999665021896, "global_step": 116328, "epoch": 1307, "lr": 6.129126346346532e-05}
{"train_loss": 0.06613953411579132, "global_step": 116329, "epoch": 1307, "lr": 6.129069870196142e-05}
{"train_loss": 0.12979182600975037, "global_step": 116330, "epoch": 1307, "lr": 6.129013393893964e-05}
{"train_loss": 0.15485495328903198, "global_step": 116331, "epoch": 1307, "lr": 6.128956917440002e-05}
{"train_loss": 0.08397524058818817, "global_step": 116332, "epoch": 1307, "lr": 6.128900440834267e-05}
{"train_loss": 0.11999747157096863, "global_step": 116333, "epoch": 1307, "lr": 6.128843964076764e-05}
{"train_loss": 0.18798679113388062, "global_step": 116334, "epoch": 1307, "lr": 6.128787487167502e-05}
{"train_loss": 0.15918835997581482, "global_step": 116335, "epoch": 1307, "lr": 6.128731010106487e-05}
{"train_loss": 0.14869365096092224, "global_step": 116336, "epoch": 1307, "lr": 6.12867453289373e-05}
{"train_loss": 0.11116280406713486, "global_step": 116337, "epoch": 1307, "lr": 6.128618055529237e-05}
{"train_loss": 0.13312536478042603, "global_step": 116338, "epoch": 1307, "lr": 6.128561578013013e-05}
{"train_loss": 0.12079550325870514, "global_step": 116339, "epoch": 1307, "lr": 6.128505100345068e-05}
{"train_loss": 0.0669635459780693, "global_step": 116340, "epoch": 1307, "lr": 6.128448622525409e-05}
{"train_loss": 0.15111364424228668, "global_step": 116341, "epoch": 1307, "lr": 6.128392144554045e-05}
{"train_loss": 0.17834624648094177, "global_step": 116342, "epoch": 1307, "lr": 6.128335666430982e-05}
{"train_loss": 0.13968762755393982, "global_step": 116343, "epoch": 1307, "lr": 6.128279188156228e-05}
{"train_loss": 0.12222100794315338, "global_step": 116344, "epoch": 1307, "lr": 6.128222709729789e-05}
{"train_loss": 0.16619428992271423, "global_step": 116345, "epoch": 1307, "lr": 6.128166231151676e-05}
{"train_loss": 0.1335938423871994, "global_step": 116346, "epoch": 1307, "lr": 6.128109752421895e-05}
{"train_loss": 0.13536760210990906, "global_step": 116347, "epoch": 1307, "lr": 6.128053273540453e-05}
{"train_loss": 0.12154152989387512, "global_step": 116348, "epoch": 1307, "lr": 6.127996794507357e-05}
{"train_loss": 0.18062950670719147, "global_step": 116349, "epoch": 1307, "lr": 6.127940315322617e-05}
{"train_loss": 0.09460084140300751, "global_step": 116350, "epoch": 1307, "lr": 6.127883835986239e-05}
{"train_loss": 0.11324890702962875, "global_step": 116351, "epoch": 1307, "lr": 6.12782735649823e-05}
{"train_loss": 0.14716783165931702, "global_step": 116352, "epoch": 1307, "lr": 6.127770876858598e-05}
{"train_loss": 0.14849820733070374, "global_step": 116353, "epoch": 1307, "lr": 6.127714397067352e-05}
{"train_loss": 0.19475151598453522, "global_step": 116354, "epoch": 1307, "lr": 6.127657917124498e-05}
{"train_loss": 0.15601156651973724, "global_step": 116355, "epoch": 1307, "lr": 6.127601437030046e-05}
{"train_loss": 0.12175232917070389, "global_step": 116356, "epoch": 1307, "lr": 6.127544956784e-05}
{"train_loss": 0.12926709651947021, "global_step": 116357, "epoch": 1307, "lr": 6.127488476386368e-05}
{"train_loss": 0.1453118920326233, "global_step": 116358, "epoch": 1307, "lr": 6.12743199583716e-05}
{"train_loss": 0.1007567048072815, "global_step": 116359, "epoch": 1307, "lr": 6.127375515136384e-05}
{"train_loss": 0.16153384745121002, "global_step": 116360, "epoch": 1307, "lr": 6.127319034284044e-05}
{"train_loss": 0.10177670419216156, "global_step": 116361, "epoch": 1307, "lr": 6.12726255328015e-05}
{"train_loss": 0.12197980284690857, "global_step": 116362, "epoch": 1307, "lr": 6.127206072124711e-05}
{"train_loss": 0.16483131051063538, "global_step": 116363, "epoch": 1307, "lr": 6.127149590817732e-05}
{"train_loss": 0.13080547749996185, "global_step": 116364, "epoch": 1307, "lr": 6.127093109359222e-05}
{"train_loss": 0.1370892971754074, "global_step": 116365, "epoch": 1307, "lr": 6.127036627749186e-05}
{"train_loss": 0.10031943768262863, "global_step": 116366, "epoch": 1307, "lr": 6.126980145987636e-05}
{"train_loss": 0.1172109916806221, "global_step": 116367, "epoch": 1307, "lr": 6.126923664074575e-05}
{"train_loss": 0.13990813493728638, "global_step": 116368, "epoch": 1307, "lr": 6.126867182010014e-05}
{"train_loss": 0.1235949844121933, "global_step": 116369, "epoch": 1307, "lr": 6.12681069979396e-05}
{"train_loss": 0.08923699706792831, "global_step": 116370, "epoch": 1307, "lr": 6.12675421742642e-05}
{"train_loss": 0.1658434271812439, "global_step": 116371, "epoch": 1307, "lr": 6.1266977349074e-05}
{"train_loss": 0.15112416446208954, "global_step": 116372, "epoch": 1307, "lr": 6.126641252236911e-05}
{"train_loss": 0.10088221728801727, "global_step": 116373, "epoch": 1307, "lr": 6.126584769414959e-05}
{"train_loss": 0.18419112265110016, "global_step": 116374, "epoch": 1307, "lr": 6.12652828644155e-05}
{"train_loss": 0.20729337632656097, "global_step": 116375, "epoch": 1307, "lr": 6.126471803316694e-05}
{"train_loss": 0.13969308137893677, "global_step": 116376, "epoch": 1307, "lr": 6.126415320040397e-05}
{"train_loss": 0.24419978260993958, "global_step": 116377, "epoch": 1307, "lr": 6.126358836612668e-05}
{"train_loss": 0.12950347363948822, "global_step": 116378, "epoch": 1307, "lr": 6.126302353033512e-05}
{"train_loss": 0.12028715759515762, "global_step": 116379, "epoch": 1307, "lr": 6.126245869302941e-05}
{"train_loss": 0.13271114230155945, "global_step": 116380, "epoch": 1307, "lr": 6.126189385420958e-05}
{"train_loss": 0.18311472237110138, "global_step": 116381, "epoch": 1307, "lr": 6.126132901387574e-05}
{"train_loss": 0.1684751957654953, "global_step": 116382, "epoch": 1307, "lr": 6.126076417202794e-05}
{"train_loss": 0.15224066376686096, "global_step": 116383, "epoch": 1307, "lr": 6.126019932866626e-05}
{"train_loss": 0.17163145542144775, "global_step": 116384, "epoch": 1307, "lr": 6.12596344837908e-05}
{"train_loss": 0.10322941839694977, "global_step": 116385, "epoch": 1307, "lr": 6.125906963740162e-05}
{"train_loss": 0.16244661808013916, "global_step": 116386, "epoch": 1307, "lr": 6.125850478949878e-05}
{"train_loss": 0.13067291676998138, "global_step": 116387, "epoch": 1307, "lr": 6.125793994008239e-05}
{"train_loss": 0.14748665690422058, "global_step": 116388, "epoch": 1307, "lr": 6.125737508915249e-05}
{"train_loss": 0.08514373749494553, "global_step": 116389, "epoch": 1307, "lr": 6.125681023670917e-05}
{"train_loss": 0.11969562619924545, "global_step": 116390, "epoch": 1307, "lr": 6.125624538275251e-05}
{"train_loss": 0.10932362824678421, "global_step": 116391, "epoch": 1307, "lr": 6.125568052728259e-05}
{"train_loss": 0.1435139924287796, "global_step": 116392, "epoch": 1307, "lr": 6.12551156702995e-05}
{"train_loss": 0.1407230943441391, "global_step": 116393, "epoch": 1307, "lr": 6.125455081180329e-05}
{"train_loss": 0.1326780468225479, "global_step": 116394, "epoch": 1307, "lr": 6.125398595179404e-05}
{"train_loss": 0.18686097860336304, "global_step": 116395, "epoch": 1307, "lr": 6.125342109027183e-05}
{"train_loss": 0.19753772020339966, "global_step": 116396, "epoch": 1307, "lr": 6.125285622723672e-05}
{"train_loss": 0.19183479249477386, "global_step": 116397, "epoch": 1307, "lr": 6.12522913626888e-05}
{"train_loss": 0.14684274792671204, "global_step": 116398, "epoch": 1307, "lr": 6.125172649662816e-05}
{"train_loss": 0.10315920412540436, "global_step": 116399, "epoch": 1307, "lr": 6.125116162905487e-05}
{"train_loss": 0.10697394609451294, "global_step": 116400, "epoch": 1307, "lr": 6.125059675996898e-05}
{"train_loss": 0.13625258207321167, "global_step": 116401, "epoch": 1307, "lr": 6.125003188937059e-05}
{"train_loss": 0.10003425180912018, "global_step": 116402, "epoch": 1307, "lr": 6.124946701725977e-05}
{"train_loss": 0.07656040787696838, "global_step": 116403, "epoch": 1307, "lr": 6.12489021436366e-05}
{"train_loss": 0.07899478077888489, "global_step": 116404, "epoch": 1307, "lr": 6.124833726850114e-05}
{"train_loss": 0.15943992137908936, "global_step": 116405, "epoch": 1307, "lr": 6.12477723918535e-05}
{"train_loss": 0.15531489253044128, "global_step": 116406, "epoch": 1307, "lr": 6.124720751369372e-05}
{"train_loss": 0.1809016615152359, "global_step": 116407, "epoch": 1307, "lr": 6.124664263402187e-05}
{"train_loss": 0.15150047838687897, "global_step": 116408, "epoch": 1307, "lr": 6.124607775283807e-05}
{"train_loss": 0.11179817467927933, "global_step": 116409, "epoch": 1307, "lr": 6.124551287014238e-05}
{"train_loss": 0.17884708940982819, "global_step": 116410, "epoch": 1307, "lr": 6.124494798593485e-05}
{"train_loss": 0.1372359689701809, "global_step": 116411, "epoch": 1307, "lr": 6.12443831002156e-05, "val_loss": 5.272582530975342}
{"train_loss": 0.14341950416564941, "global_step": 116412, "epoch": 1308, "lr": 6.124381821298467e-05}
{"train_loss": 0.11571969836950302, "global_step": 116413, "epoch": 1308, "lr": 6.124325332424213e-05}
{"train_loss": 0.12343154102563858, "global_step": 116414, "epoch": 1308, "lr": 6.124268843398807e-05}
{"train_loss": 0.14588738977909088, "global_step": 116415, "epoch": 1308, "lr": 6.124212354222259e-05}
{"train_loss": 0.13522563874721527, "global_step": 116416, "epoch": 1308, "lr": 6.124155864894571e-05}
{"train_loss": 0.1673262119293213, "global_step": 116417, "epoch": 1308, "lr": 6.124099375415757e-05}
{"train_loss": 0.13245750963687897, "global_step": 116418, "epoch": 1308, "lr": 6.12404288578582e-05}
{"train_loss": 0.12091805785894394, "global_step": 116419, "epoch": 1308, "lr": 6.12398639600477e-05}
{"train_loss": 0.12642422318458557, "global_step": 116420, "epoch": 1308, "lr": 6.123929906072614e-05}
{"train_loss": 0.11931083351373672, "global_step": 116421, "epoch": 1308, "lr": 6.123873415989358e-05}
{"train_loss": 0.12162647396326065, "global_step": 116422, "epoch": 1308, "lr": 6.123816925755011e-05}
{"train_loss": 0.1720779538154602, "global_step": 116423, "epoch": 1308, "lr": 6.123760435369581e-05}
{"train_loss": 0.2008165568113327, "global_step": 116424, "epoch": 1308, "lr": 6.123703944833076e-05}
{"train_loss": 0.09745577722787857, "global_step": 116425, "epoch": 1308, "lr": 6.123647454145501e-05}
{"train_loss": 0.12299598008394241, "global_step": 116426, "epoch": 1308, "lr": 6.123590963306865e-05}
{"train_loss": 0.1032194048166275, "global_step": 116427, "epoch": 1308, "lr": 6.123534472317177e-05}
{"train_loss": 0.09910929948091507, "global_step": 116428, "epoch": 1308, "lr": 6.123477981176444e-05}
{"train_loss": 0.1282297521829605, "global_step": 116429, "epoch": 1308, "lr": 6.123421489884673e-05}
{"train_loss": 0.11806976050138474, "global_step": 116430, "epoch": 1308, "lr": 6.12336499844187e-05}
{"train_loss": 0.1047120913863182, "global_step": 116431, "epoch": 1308, "lr": 6.123308506848047e-05}
{"train_loss": 0.1415855586528778, "global_step": 116432, "epoch": 1308, "lr": 6.123252015103208e-05}
{"train_loss": 0.1461191326379776, "global_step": 116433, "epoch": 1308, "lr": 6.123195523207361e-05}
{"train_loss": 0.104267418384552, "global_step": 116434, "epoch": 1308, "lr": 6.123139031160514e-05}
{"train_loss": 0.1619078814983368, "global_step": 116435, "epoch": 1308, "lr": 6.123082538962674e-05}
{"train_loss": 0.1615331619977951, "global_step": 116436, "epoch": 1308, "lr": 6.123026046613851e-05}
{"train_loss": 0.10373036563396454, "global_step": 116437, "epoch": 1308, "lr": 6.122969554114049e-05}
{"train_loss": 0.09592787176370621, "global_step": 116438, "epoch": 1308, "lr": 6.122913061463278e-05}
{"train_loss": 0.1368562877178192, "global_step": 116439, "epoch": 1308, "lr": 6.122856568661545e-05}
{"train_loss": 0.11974931508302689, "global_step": 116440, "epoch": 1308, "lr": 6.122800075708858e-05}
{"train_loss": 0.13504116237163544, "global_step": 116441, "epoch": 1308, "lr": 6.122743582605224e-05}
{"train_loss": 0.14861738681793213, "global_step": 116442, "epoch": 1308, "lr": 6.122687089350651e-05}
{"train_loss": 0.12997402250766754, "global_step": 116443, "epoch": 1308, "lr": 6.122630595945145e-05}
{"train_loss": 0.12018557637929916, "global_step": 116444, "epoch": 1308, "lr": 6.122574102388716e-05}
{"train_loss": 0.13316582143306732, "global_step": 116445, "epoch": 1308, "lr": 6.12251760868137e-05}
{"train_loss": 0.10332891345024109, "global_step": 116446, "epoch": 1308, "lr": 6.122461114823117e-05}
{"train_loss": 0.18441787362098694, "global_step": 116447, "epoch": 1308, "lr": 6.122404620813962e-05}
{"train_loss": 0.10497965663671494, "global_step": 116448, "epoch": 1308, "lr": 6.122348126653912e-05}
{"train_loss": 0.12168172001838684, "global_step": 116449, "epoch": 1308, "lr": 6.122291632342978e-05}
{"train_loss": 0.18399108946323395, "global_step": 116450, "epoch": 1308, "lr": 6.122235137881164e-05}
{"train_loss": 0.19507014751434326, "global_step": 116451, "epoch": 1308, "lr": 6.122178643268482e-05}
{"train_loss": 0.11571233719587326, "global_step": 116452, "epoch": 1308, "lr": 6.122122148504934e-05}
{"train_loss": 0.1409415453672409, "global_step": 116453, "epoch": 1308, "lr": 6.122065653590531e-05}
{"train_loss": 0.07759210467338562, "global_step": 116454, "epoch": 1308, "lr": 6.12200915852528e-05}
{"train_loss": 0.1555226892232895, "global_step": 116455, "epoch": 1308, "lr": 6.121952663309187e-05}
{"train_loss": 0.12417513877153397, "global_step": 116456, "epoch": 1308, "lr": 6.121896167942264e-05}
{"train_loss": 0.22011882066726685, "global_step": 116457, "epoch": 1308, "lr": 6.121839672424514e-05}
{"train_loss": 0.12232162058353424, "global_step": 116458, "epoch": 1308, "lr": 6.121783176755946e-05}
{"train_loss": 0.1395023912191391, "global_step": 116459, "epoch": 1308, "lr": 6.121726680936569e-05}
{"train_loss": 0.1600523293018341, "global_step": 116460, "epoch": 1308, "lr": 6.121670184966388e-05}
{"train_loss": 0.11644668132066727, "global_step": 116461, "epoch": 1308, "lr": 6.121613688845414e-05}
{"train_loss": 0.11218763887882233, "global_step": 116462, "epoch": 1308, "lr": 6.121557192573651e-05}
{"train_loss": 0.17717455327510834, "global_step": 116463, "epoch": 1308, "lr": 6.12150069615111e-05}
{"train_loss": 0.1285414695739746, "global_step": 116464, "epoch": 1308, "lr": 6.121444199577795e-05}
{"train_loss": 0.0932174101471901, "global_step": 116465, "epoch": 1308, "lr": 6.121387702853717e-05}
{"train_loss": 0.17216423153877258, "global_step": 116466, "epoch": 1308, "lr": 6.121331205978882e-05}
{"train_loss": 0.15091191232204437, "global_step": 116467, "epoch": 1308, "lr": 6.121274708953297e-05}
{"train_loss": 0.07293220609426498, "global_step": 116468, "epoch": 1308, "lr": 6.12121821177697e-05}
{"train_loss": 0.16415394842624664, "global_step": 116469, "epoch": 1308, "lr": 6.121161714449911e-05}
{"train_loss": 0.18210910260677338, "global_step": 116470, "epoch": 1308, "lr": 6.121105216972124e-05}
{"train_loss": 0.18706785142421722, "global_step": 116471, "epoch": 1308, "lr": 6.121048719343618e-05}
{"train_loss": 0.15538306534290314, "global_step": 116472, "epoch": 1308, "lr": 6.120992221564401e-05}
{"train_loss": 0.18615645170211792, "global_step": 116473, "epoch": 1308, "lr": 6.12093572363448e-05}
{"train_loss": 0.09859400242567062, "global_step": 116474, "epoch": 1308, "lr": 6.120879225553863e-05}
{"train_loss": 0.1829604208469391, "global_step": 116475, "epoch": 1308, "lr": 6.120822727322558e-05}
{"train_loss": 0.15319223701953888, "global_step": 116476, "epoch": 1308, "lr": 6.12076622894057e-05}
{"train_loss": 0.16532392799854279, "global_step": 116477, "epoch": 1308, "lr": 6.12070973040791e-05}
{"train_loss": 0.12341254949569702, "global_step": 116478, "epoch": 1308, "lr": 6.120653231724585e-05}
{"train_loss": 0.20097748935222626, "global_step": 116479, "epoch": 1308, "lr": 6.1205967328906e-05}
{"train_loss": 0.1343742161989212, "global_step": 116480, "epoch": 1308, "lr": 6.120540233905965e-05}
{"train_loss": 0.1227601170539856, "global_step": 116481, "epoch": 1308, "lr": 6.120483734770688e-05}
{"train_loss": 0.08769448101520538, "global_step": 116482, "epoch": 1308, "lr": 6.120427235484775e-05}
{"train_loss": 0.11186793446540833, "global_step": 116483, "epoch": 1308, "lr": 6.120370736048234e-05}
{"train_loss": 0.12550684809684753, "global_step": 116484, "epoch": 1308, "lr": 6.120314236461073e-05}
{"train_loss": 0.16043908894062042, "global_step": 116485, "epoch": 1308, "lr": 6.120257736723299e-05}
{"train_loss": 0.17372570931911469, "global_step": 116486, "epoch": 1308, "lr": 6.120201236834921e-05}
{"train_loss": 0.08412957936525345, "global_step": 116487, "epoch": 1308, "lr": 6.120144736795945e-05}
{"train_loss": 0.20302028954029083, "global_step": 116488, "epoch": 1308, "lr": 6.12008823660638e-05}
{"train_loss": 0.14861749112606049, "global_step": 116489, "epoch": 1308, "lr": 6.120031736266232e-05}
{"train_loss": 0.14186051487922668, "global_step": 116490, "epoch": 1308, "lr": 6.11997523577551e-05}
{"train_loss": 0.09654387831687927, "global_step": 116491, "epoch": 1308, "lr": 6.119918735134221e-05}
{"train_loss": 0.09908567368984222, "global_step": 116492, "epoch": 1308, "lr": 6.119862234342372e-05}
{"train_loss": 0.13845255970954895, "global_step": 116493, "epoch": 1308, "lr": 6.119805733399971e-05}
{"train_loss": 0.09010616689920425, "global_step": 116494, "epoch": 1308, "lr": 6.119749232307025e-05}
{"train_loss": 0.19649970531463623, "global_step": 116495, "epoch": 1308, "lr": 6.119692731063544e-05}
{"train_loss": 0.058382559567689896, "global_step": 116496, "epoch": 1308, "lr": 6.119636229669534e-05}
{"train_loss": 0.1324739009141922, "global_step": 116497, "epoch": 1308, "lr": 6.119579728125003e-05}
{"train_loss": 0.10075405985116959, "global_step": 116498, "epoch": 1308, "lr": 6.119523226429955e-05}
{"train_loss": 0.14078930020332336, "global_step": 116499, "epoch": 1308, "lr": 6.119466724584402e-05}
{"train_loss": 0.13504359042376615, "global_step": 116500, "epoch": 1308, "lr": 6.119410222588352e-05, "val_loss": 5.243465900421143}
{"train_loss": 0.1622658371925354, "global_step": 116501, "epoch": 1309, "lr": 6.11935372044181e-05}
{"train_loss": 0.12932510673999786, "global_step": 116502, "epoch": 1309, "lr": 6.119297218144783e-05}
{"train_loss": 0.16938628256320953, "global_step": 116503, "epoch": 1309, "lr": 6.119240715697283e-05}
{"train_loss": 0.103809654712677, "global_step": 116504, "epoch": 1309, "lr": 6.119184213099313e-05}
{"train_loss": 0.14475111663341522, "global_step": 116505, "epoch": 1309, "lr": 6.119127710350883e-05}
{"train_loss": 0.15439732372760773, "global_step": 116506, "epoch": 1309, "lr": 6.119071207452e-05}
{"train_loss": 0.15985167026519775, "global_step": 116507, "epoch": 1309, "lr": 6.119014704402671e-05}
{"train_loss": 0.07373639196157455, "global_step": 116508, "epoch": 1309, "lr": 6.118958201202905e-05}
{"train_loss": 0.12936219573020935, "global_step": 116509, "epoch": 1309, "lr": 6.118901697852708e-05}
{"train_loss": 0.15701524913311005, "global_step": 116510, "epoch": 1309, "lr": 6.118845194352088e-05}
{"train_loss": 0.08826841413974762, "global_step": 116511, "epoch": 1309, "lr": 6.118788690701054e-05}
{"train_loss": 0.12894096970558167, "global_step": 116512, "epoch": 1309, "lr": 6.118732186899612e-05}
{"train_loss": 0.1686737984418869, "global_step": 116513, "epoch": 1309, "lr": 6.11867568294777e-05}
{"train_loss": 0.11432506144046783, "global_step": 116514, "epoch": 1309, "lr": 6.118619178845535e-05}
{"train_loss": 0.22955967485904694, "global_step": 116515, "epoch": 1309, "lr": 6.118562674592915e-05}
{"train_loss": 0.1655222773551941, "global_step": 116516, "epoch": 1309, "lr": 6.11850617018992e-05}
{"train_loss": 0.16350452601909637, "global_step": 116517, "epoch": 1309, "lr": 6.118449665636553e-05}
{"train_loss": 0.09379976242780685, "global_step": 116518, "epoch": 1309, "lr": 6.118393160932826e-05}
{"train_loss": 0.15054814517498016, "global_step": 116519, "epoch": 1309, "lr": 6.118336656078742e-05}
{"train_loss": 0.14803862571716309, "global_step": 116520, "epoch": 1309, "lr": 6.118280151074314e-05}
{"train_loss": 0.2721847593784332, "global_step": 116521, "epoch": 1309, "lr": 6.118223645919546e-05}
{"train_loss": 0.1793523132801056, "global_step": 116522, "epoch": 1309, "lr": 6.118167140614446e-05}
{"train_loss": 0.10894646495580673, "global_step": 116523, "epoch": 1309, "lr": 6.118110635159024e-05}
{"train_loss": 0.17576001584529877, "global_step": 116524, "epoch": 1309, "lr": 6.118054129553283e-05}
{"train_loss": 0.13230039179325104, "global_step": 116525, "epoch": 1309, "lr": 6.117997623797234e-05}
{"train_loss": 0.12822335958480835, "global_step": 116526, "epoch": 1309, "lr": 6.117941117890885e-05}
{"train_loss": 0.1687256097793579, "global_step": 116527, "epoch": 1309, "lr": 6.117884611834244e-05}
{"train_loss": 0.18823964893817902, "global_step": 116528, "epoch": 1309, "lr": 6.117828105627315e-05}
{"train_loss": 0.18313130736351013, "global_step": 116529, "epoch": 1309, "lr": 6.117771599270107e-05}
{"train_loss": 0.15135809779167175, "global_step": 116530, "epoch": 1309, "lr": 6.117715092762629e-05}
{"train_loss": 0.12222184240818024, "global_step": 116531, "epoch": 1309, "lr": 6.117658586104887e-05}
{"train_loss": 0.18277305364608765, "global_step": 116532, "epoch": 1309, "lr": 6.11760207929689e-05}
{"train_loss": 0.18907476961612701, "global_step": 116533, "epoch": 1309, "lr": 6.117545572338645e-05}
{"train_loss": 0.17357759177684784, "global_step": 116534, "epoch": 1309, "lr": 6.117489065230158e-05}
{"train_loss": 0.13601256906986237, "global_step": 116535, "epoch": 1309, "lr": 6.11743255797144e-05}
{"train_loss": 0.15030576288700104, "global_step": 116536, "epoch": 1309, "lr": 6.117376050562497e-05}
{"train_loss": 0.19718751311302185, "global_step": 116537, "epoch": 1309, "lr": 6.117319543003335e-05}
{"train_loss": 0.1249285414814949, "global_step": 116538, "epoch": 1309, "lr": 6.117263035293964e-05}
{"train_loss": 0.1390133649110794, "global_step": 116539, "epoch": 1309, "lr": 6.11720652743439e-05}
{"train_loss": 0.12315422296524048, "global_step": 116540, "epoch": 1309, "lr": 6.117150019424623e-05}
{"train_loss": 0.12142223864793777, "global_step": 116541, "epoch": 1309, "lr": 6.117093511264667e-05}
{"train_loss": 0.07557404041290283, "global_step": 116542, "epoch": 1309, "lr": 6.117037002954532e-05}
{"train_loss": 0.16413220763206482, "global_step": 116543, "epoch": 1309, "lr": 6.116980494494225e-05}
{"train_loss": 0.1300039291381836, "global_step": 116544, "epoch": 1309, "lr": 6.116923985883754e-05}
{"train_loss": 0.15631374716758728, "global_step": 116545, "epoch": 1309, "lr": 6.116867477123125e-05}
{"train_loss": 0.15333576500415802, "global_step": 116546, "epoch": 1309, "lr": 6.116810968212347e-05}
{"train_loss": 0.1838364452123642, "global_step": 116547, "epoch": 1309, "lr": 6.11675445915143e-05}
{"train_loss": 0.15648707747459412, "global_step": 116548, "epoch": 1309, "lr": 6.116697949940376e-05}
{"train_loss": 0.09966175258159637, "global_step": 116549, "epoch": 1309, "lr": 6.116641440579197e-05}
{"train_loss": 0.07273798435926437, "global_step": 116550, "epoch": 1309, "lr": 6.1165849310679e-05}
{"train_loss": 0.2070942521095276, "global_step": 116551, "epoch": 1309, "lr": 6.116528421406488e-05}
{"train_loss": 0.09890282154083252, "global_step": 116552, "epoch": 1309, "lr": 6.116471911594976e-05}
{"train_loss": 0.08604716509580612, "global_step": 116553, "epoch": 1309, "lr": 6.116415401633366e-05}
{"train_loss": 0.09816554188728333, "global_step": 116554, "epoch": 1309, "lr": 6.116358891521668e-05}
{"train_loss": 0.12941159307956696, "global_step": 116555, "epoch": 1309, "lr": 6.11630238125989e-05}
{"train_loss": 0.26179930567741394, "global_step": 116556, "epoch": 1309, "lr": 6.116245870848037e-05}
{"train_loss": 0.09271426498889923, "global_step": 116557, "epoch": 1309, "lr": 6.11618936028612e-05}
{"train_loss": 0.18992502987384796, "global_step": 116558, "epoch": 1309, "lr": 6.116132849574144e-05}
{"train_loss": 0.17280679941177368, "global_step": 116559, "epoch": 1309, "lr": 6.11607633871212e-05}
{"train_loss": 0.252925843000412, "global_step": 116560, "epoch": 1309, "lr": 6.116019827700051e-05}
{"train_loss": 0.13456380367279053, "global_step": 116561, "epoch": 1309, "lr": 6.115963316537946e-05}
{"train_loss": 0.1144004538655281, "global_step": 116562, "epoch": 1309, "lr": 6.115906805225814e-05}
{"train_loss": 0.11225105822086334, "global_step": 116563, "epoch": 1309, "lr": 6.115850293763662e-05}
{"train_loss": 0.16004545986652374, "global_step": 116564, "epoch": 1309, "lr": 6.115793782151498e-05}
{"train_loss": 0.232864648103714, "global_step": 116565, "epoch": 1309, "lr": 6.115737270389328e-05}
{"train_loss": 0.2569578289985657, "global_step": 116566, "epoch": 1309, "lr": 6.115680758477162e-05}
{"train_loss": 0.2023428976535797, "global_step": 116567, "epoch": 1309, "lr": 6.115624246415007e-05}
{"train_loss": 0.14041583240032196, "global_step": 116568, "epoch": 1309, "lr": 6.115567734202868e-05}
{"train_loss": 0.07201972603797913, "global_step": 116569, "epoch": 1309, "lr": 6.115511221840756e-05}
{"train_loss": 0.13919149339199066, "global_step": 116570, "epoch": 1309, "lr": 6.115454709328675e-05}
{"train_loss": 0.06469305604696274, "global_step": 116571, "epoch": 1309, "lr": 6.115398196666636e-05}
{"train_loss": 0.09956090152263641, "global_step": 116572, "epoch": 1309, "lr": 6.115341683854646e-05}
{"train_loss": 0.1621631383895874, "global_step": 116573, "epoch": 1309, "lr": 6.11528517089271e-05}
{"train_loss": 0.1630220264196396, "global_step": 116574, "epoch": 1309, "lr": 6.11522865778084e-05}
{"train_loss": 0.18047088384628296, "global_step": 116575, "epoch": 1309, "lr": 6.11517214451904e-05}
{"train_loss": 0.14597682654857635, "global_step": 116576, "epoch": 1309, "lr": 6.115115631107319e-05}
{"train_loss": 0.11986343562602997, "global_step": 116577, "epoch": 1309, "lr": 6.115059117545682e-05}
{"train_loss": 0.06609603762626648, "global_step": 116578, "epoch": 1309, "lr": 6.115002603834143e-05}
{"train_loss": 0.19487547874450684, "global_step": 116579, "epoch": 1309, "lr": 6.114946089972701e-05}
{"train_loss": 0.17659799754619598, "global_step": 116580, "epoch": 1309, "lr": 6.11488957596137e-05}
{"train_loss": 0.18045705556869507, "global_step": 116581, "epoch": 1309, "lr": 6.114833061800158e-05}
{"train_loss": 0.08160997182130814, "global_step": 116582, "epoch": 1309, "lr": 6.114776547489068e-05}
{"train_loss": 0.11792673915624619, "global_step": 116583, "epoch": 1309, "lr": 6.114720033028111e-05}
{"train_loss": 0.16038212180137634, "global_step": 116584, "epoch": 1309, "lr": 6.114663518417292e-05}
{"train_loss": 0.16418619453907013, "global_step": 116585, "epoch": 1309, "lr": 6.11460700365662e-05}
{"train_loss": 0.12286490201950073, "global_step": 116586, "epoch": 1309, "lr": 6.114550488746105e-05}
{"train_loss": 0.13527920842170715, "global_step": 116587, "epoch": 1309, "lr": 6.114493973685751e-05}
{"train_loss": 0.20965580642223358, "global_step": 116588, "epoch": 1309, "lr": 6.114437458475567e-05}
{"train_loss": 0.14955874314803755, "global_step": 116589, "epoch": 1309, "lr": 6.114380943115559e-05, "val_loss": 5.227383613586426}
{"train_loss": 0.12803857028484344, "global_step": 116590, "epoch": 1310, "lr": 6.114324427605737e-05}
{"train_loss": 0.1488305926322937, "global_step": 116591, "epoch": 1310, "lr": 6.114267911946107e-05}
{"train_loss": 0.09288053214550018, "global_step": 116592, "epoch": 1310, "lr": 6.114211396136678e-05}
{"train_loss": 0.15132160484790802, "global_step": 116593, "epoch": 1310, "lr": 6.114154880177458e-05}
{"train_loss": 0.13344377279281616, "global_step": 116594, "epoch": 1310, "lr": 6.114098364068452e-05}
{"train_loss": 0.18335524201393127, "global_step": 116595, "epoch": 1310, "lr": 6.11404184780967e-05}
{"train_loss": 0.18781593441963196, "global_step": 116596, "epoch": 1310, "lr": 6.113985331401118e-05}
{"train_loss": 0.1535891443490982, "global_step": 116597, "epoch": 1310, "lr": 6.113928814842804e-05}
{"train_loss": 0.1051168292760849, "global_step": 116598, "epoch": 1310, "lr": 6.113872298134735e-05}
{"train_loss": 0.24241425096988678, "global_step": 116599, "epoch": 1310, "lr": 6.113815781276921e-05}
{"train_loss": 0.15580548346042633, "global_step": 116600, "epoch": 1310, "lr": 6.113759264269368e-05}
{"train_loss": 0.1729004681110382, "global_step": 116601, "epoch": 1310, "lr": 6.113702747112083e-05}
{"train_loss": 0.1979151964187622, "global_step": 116602, "epoch": 1310, "lr": 6.113646229805074e-05}
{"train_loss": 0.1415453553199768, "global_step": 116603, "epoch": 1310, "lr": 6.11358971234835e-05}
{"train_loss": 0.1704915165901184, "global_step": 116604, "epoch": 1310, "lr": 6.113533194741917e-05}
{"train_loss": 0.1630285382270813, "global_step": 116605, "epoch": 1310, "lr": 6.113476676985783e-05}
{"train_loss": 0.14003504812717438, "global_step": 116606, "epoch": 1310, "lr": 6.113420159079957e-05}
{"train_loss": 0.13314573466777802, "global_step": 116607, "epoch": 1310, "lr": 6.113363641024442e-05}
{"train_loss": 0.1278727799654007, "global_step": 116608, "epoch": 1310, "lr": 6.11330712281925e-05}
{"train_loss": 0.12205787003040314, "global_step": 116609, "epoch": 1310, "lr": 6.113250604464389e-05}
{"train_loss": 0.15716974437236786, "global_step": 116610, "epoch": 1310, "lr": 6.113194085959864e-05}
{"train_loss": 0.26442331075668335, "global_step": 116611, "epoch": 1310, "lr": 6.113137567305684e-05}
{"train_loss": 0.15752863883972168, "global_step": 116612, "epoch": 1310, "lr": 6.113081048501855e-05}
{"train_loss": 0.21844597160816193, "global_step": 116613, "epoch": 1310, "lr": 6.113024529548387e-05}
{"train_loss": 0.16382364928722382, "global_step": 116614, "epoch": 1310, "lr": 6.112968010445286e-05}
{"train_loss": 0.10250917822122574, "global_step": 116615, "epoch": 1310, "lr": 6.11291149119256e-05}
{"train_loss": 0.11234822869300842, "global_step": 116616, "epoch": 1310, "lr": 6.112854971790217e-05}
{"train_loss": 0.18602755665779114, "global_step": 116617, "epoch": 1310, "lr": 6.112798452238263e-05}
{"train_loss": 0.16739468276500702, "global_step": 116618, "epoch": 1310, "lr": 6.112741932536707e-05}
{"train_loss": 0.13821184635162354, "global_step": 116619, "epoch": 1310, "lr": 6.112685412685558e-05}
{"train_loss": 0.16418729722499847, "global_step": 116620, "epoch": 1310, "lr": 6.112628892684821e-05}
{"train_loss": 0.10542973130941391, "global_step": 116621, "epoch": 1310, "lr": 6.112572372534506e-05}
{"train_loss": 0.17422638833522797, "global_step": 116622, "epoch": 1310, "lr": 6.112515852234616e-05}
{"train_loss": 0.2090393304824829, "global_step": 116623, "epoch": 1310, "lr": 6.112459331785164e-05}
{"train_loss": 0.09207817912101746, "global_step": 116624, "epoch": 1310, "lr": 6.112402811186156e-05}
{"train_loss": 0.16583408415317535, "global_step": 116625, "epoch": 1310, "lr": 6.112346290437598e-05}
{"train_loss": 0.12880386412143707, "global_step": 116626, "epoch": 1310, "lr": 6.112289769539499e-05}
{"train_loss": 0.13534122705459595, "global_step": 116627, "epoch": 1310, "lr": 6.112233248491865e-05}
{"train_loss": 0.143376424908638, "global_step": 116628, "epoch": 1310, "lr": 6.112176727294707e-05}
{"train_loss": 0.13369330763816833, "global_step": 116629, "epoch": 1310, "lr": 6.112120205948029e-05}
{"train_loss": 0.16273993253707886, "global_step": 116630, "epoch": 1310, "lr": 6.11206368445184e-05}
{"train_loss": 0.15739087760448456, "global_step": 116631, "epoch": 1310, "lr": 6.112007162806147e-05}
{"train_loss": 0.12263750284910202, "global_step": 116632, "epoch": 1310, "lr": 6.111950641010959e-05}
{"train_loss": 0.16390864551067352, "global_step": 116633, "epoch": 1310, "lr": 6.111894119066282e-05}
{"train_loss": 0.11056169122457504, "global_step": 116634, "epoch": 1310, "lr": 6.111837596972126e-05}
{"train_loss": 0.14810970425605774, "global_step": 116635, "epoch": 1310, "lr": 6.111781074728496e-05}
{"train_loss": 0.1389036476612091, "global_step": 116636, "epoch": 1310, "lr": 6.1117245523354e-05}
{"train_loss": 0.16582265496253967, "global_step": 116637, "epoch": 1310, "lr": 6.111668029792848e-05}
{"train_loss": 0.21330060064792633, "global_step": 116638, "epoch": 1310, "lr": 6.111611507100844e-05}
{"train_loss": 0.19357024133205414, "global_step": 116639, "epoch": 1310, "lr": 6.111554984259398e-05}
{"train_loss": 0.14995166659355164, "global_step": 116640, "epoch": 1310, "lr": 6.111498461268516e-05}
{"train_loss": 0.1392359733581543, "global_step": 116641, "epoch": 1310, "lr": 6.111441938128209e-05}
{"train_loss": 0.21293221414089203, "global_step": 116642, "epoch": 1310, "lr": 6.11138541483848e-05}
{"train_loss": 0.17607291042804718, "global_step": 116643, "epoch": 1310, "lr": 6.111328891399339e-05}
{"train_loss": 0.1796373724937439, "global_step": 116644, "epoch": 1310, "lr": 6.111272367810795e-05}
{"train_loss": 0.13293635845184326, "global_step": 116645, "epoch": 1310, "lr": 6.111215844072853e-05}
{"train_loss": 0.14278563857078552, "global_step": 116646, "epoch": 1310, "lr": 6.111159320185521e-05}
{"train_loss": 0.20149847865104675, "global_step": 116647, "epoch": 1310, "lr": 6.111102796148808e-05}
{"train_loss": 0.15310245752334595, "global_step": 116648, "epoch": 1310, "lr": 6.11104627196272e-05}
{"train_loss": 0.07568495720624924, "global_step": 116649, "epoch": 1310, "lr": 6.110989747627266e-05}
{"train_loss": 0.16499873995780945, "global_step": 116650, "epoch": 1310, "lr": 6.110933223142453e-05}
{"train_loss": 0.1357220709323883, "global_step": 116651, "epoch": 1310, "lr": 6.110876698508288e-05}
{"train_loss": 0.15999771654605865, "global_step": 116652, "epoch": 1310, "lr": 6.11082017372478e-05}
{"train_loss": 0.18444301187992096, "global_step": 116653, "epoch": 1310, "lr": 6.110763648791935e-05}
{"train_loss": 0.16435514390468597, "global_step": 116654, "epoch": 1310, "lr": 6.110707123709761e-05}
{"train_loss": 0.17336340248584747, "global_step": 116655, "epoch": 1310, "lr": 6.110650598478267e-05}
{"train_loss": 0.1485963761806488, "global_step": 116656, "epoch": 1310, "lr": 6.110594073097458e-05}
{"train_loss": 0.22648322582244873, "global_step": 116657, "epoch": 1310, "lr": 6.110537547567345e-05}
{"train_loss": 0.13383811712265015, "global_step": 116658, "epoch": 1310, "lr": 6.110481021887934e-05}
{"train_loss": 0.12930279970169067, "global_step": 116659, "epoch": 1310, "lr": 6.11042449605923e-05}
{"train_loss": 0.22065244615077972, "global_step": 116660, "epoch": 1310, "lr": 6.110367970081245e-05}
{"train_loss": 0.1570768803358078, "global_step": 116661, "epoch": 1310, "lr": 6.110311443953985e-05}
{"train_loss": 0.20761626958847046, "global_step": 116662, "epoch": 1310, "lr": 6.110254917677454e-05}
{"train_loss": 0.122073695063591, "global_step": 116663, "epoch": 1310, "lr": 6.110198391251666e-05}
{"train_loss": 0.11778585612773895, "global_step": 116664, "epoch": 1310, "lr": 6.110141864676624e-05}
{"train_loss": 0.1495535522699356, "global_step": 116665, "epoch": 1310, "lr": 6.110085337952337e-05}
{"train_loss": 0.17130793631076813, "global_step": 116666, "epoch": 1310, "lr": 6.110028811078812e-05}
{"train_loss": 0.0964052826166153, "global_step": 116667, "epoch": 1310, "lr": 6.109972284056059e-05}
{"train_loss": 0.1218048557639122, "global_step": 116668, "epoch": 1310, "lr": 6.109915756884082e-05}
{"train_loss": 0.1004161611199379, "global_step": 116669, "epoch": 1310, "lr": 6.109859229562891e-05}
{"train_loss": 0.10298682749271393, "global_step": 116670, "epoch": 1310, "lr": 6.109802702092493e-05}
{"train_loss": 0.12465380877256393, "global_step": 116671, "epoch": 1310, "lr": 6.109746174472896e-05}
{"train_loss": 0.09200002998113632, "global_step": 116672, "epoch": 1310, "lr": 6.109689646704106e-05}
{"train_loss": 0.10258341580629349, "global_step": 116673, "epoch": 1310, "lr": 6.109633118786133e-05}
{"train_loss": 0.1607038825750351, "global_step": 116674, "epoch": 1310, "lr": 6.109576590718983e-05}
{"train_loss": 0.1514994502067566, "global_step": 116675, "epoch": 1310, "lr": 6.109520062502664e-05}
{"train_loss": 0.17499297857284546, "global_step": 116676, "epoch": 1310, "lr": 6.109463534137183e-05}
{"train_loss": 0.07377222180366516, "global_step": 116677, "epoch": 1310, "lr": 6.109407005622548e-05}
{"train_loss": 0.15248309829261866, "global_step": 116678, "epoch": 1310, "lr": 6.109350476958768e-05, "val_loss": 5.023972511291504, "train_action_mse_error": 28.035072326660156}
{"train_loss": 0.1333400309085846, "global_step": 116679, "epoch": 1311, "lr": 6.109293948145848e-05}
{"train_loss": 0.12703444063663483, "global_step": 116680, "epoch": 1311, "lr": 6.109237419183797e-05}
{"train_loss": 0.15377946197986603, "global_step": 116681, "epoch": 1311, "lr": 6.109180890072623e-05}
{"train_loss": 0.11842845380306244, "global_step": 116682, "epoch": 1311, "lr": 6.109124360812334e-05}
{"train_loss": 0.06165822967886925, "global_step": 116683, "epoch": 1311, "lr": 6.109067831402935e-05}
{"train_loss": 0.099025197327137, "global_step": 116684, "epoch": 1311, "lr": 6.109011301844436e-05}
{"train_loss": 0.14951978623867035, "global_step": 116685, "epoch": 1311, "lr": 6.108954772136845e-05}
{"train_loss": 0.24060577154159546, "global_step": 116686, "epoch": 1311, "lr": 6.108898242280167e-05}
{"train_loss": 0.12460262328386307, "global_step": 116687, "epoch": 1311, "lr": 6.108841712274412e-05}
{"train_loss": 0.11350205540657043, "global_step": 116688, "epoch": 1311, "lr": 6.108785182119588e-05}
{"train_loss": 0.24577301740646362, "global_step": 116689, "epoch": 1311, "lr": 6.1087286518157e-05}
{"train_loss": 0.11093152314424515, "global_step": 116690, "epoch": 1311, "lr": 6.108672121362756e-05}
{"train_loss": 0.17067959904670715, "global_step": 116691, "epoch": 1311, "lr": 6.108615590760766e-05}
{"train_loss": 0.14310911297798157, "global_step": 116692, "epoch": 1311, "lr": 6.108559060009735e-05}
{"train_loss": 0.1015922874212265, "global_step": 116693, "epoch": 1311, "lr": 6.108502529109672e-05}
{"train_loss": 0.12563179433345795, "global_step": 116694, "epoch": 1311, "lr": 6.108445998060585e-05}
{"train_loss": 0.14851701259613037, "global_step": 116695, "epoch": 1311, "lr": 6.108389466862481e-05}
{"train_loss": 0.24488534033298492, "global_step": 116696, "epoch": 1311, "lr": 6.108332935515368e-05}
{"train_loss": 0.15188603103160858, "global_step": 116697, "epoch": 1311, "lr": 6.10827640401925e-05}
{"train_loss": 0.14781871438026428, "global_step": 116698, "epoch": 1311, "lr": 6.108219872374141e-05}
{"train_loss": 0.13137033581733704, "global_step": 116699, "epoch": 1311, "lr": 6.108163340580046e-05}
{"train_loss": 0.14139370620250702, "global_step": 116700, "epoch": 1311, "lr": 6.108106808636969e-05}
{"train_loss": 0.09840364754199982, "global_step": 116701, "epoch": 1311, "lr": 6.108050276544922e-05}
{"train_loss": 0.07843958586454391, "global_step": 116702, "epoch": 1311, "lr": 6.107993744303912e-05}
{"train_loss": 0.12686192989349365, "global_step": 116703, "epoch": 1311, "lr": 6.107937211913945e-05}
{"train_loss": 0.17913557589054108, "global_step": 116704, "epoch": 1311, "lr": 6.107880679375028e-05}
{"train_loss": 0.13219435513019562, "global_step": 116705, "epoch": 1311, "lr": 6.107824146687171e-05}
{"train_loss": 0.15476639568805695, "global_step": 116706, "epoch": 1311, "lr": 6.107767613850381e-05}
{"train_loss": 0.14935584366321564, "global_step": 116707, "epoch": 1311, "lr": 6.107711080864665e-05}
{"train_loss": 0.12163122743368149, "global_step": 116708, "epoch": 1311, "lr": 6.10765454773003e-05}
{"train_loss": 0.26950445771217346, "global_step": 116709, "epoch": 1311, "lr": 6.107598014446485e-05}
{"train_loss": 0.09287866204977036, "global_step": 116710, "epoch": 1311, "lr": 6.107541481014037e-05}
{"train_loss": 0.1996287703514099, "global_step": 116711, "epoch": 1311, "lr": 6.107484947432693e-05}
{"train_loss": 0.16672678291797638, "global_step": 116712, "epoch": 1311, "lr": 6.107428413702462e-05}
{"train_loss": 0.09613444656133652, "global_step": 116713, "epoch": 1311, "lr": 6.10737187982335e-05}
{"train_loss": 0.14138101041316986, "global_step": 116714, "epoch": 1311, "lr": 6.107315345795366e-05}
{"train_loss": 0.1169101744890213, "global_step": 116715, "epoch": 1311, "lr": 6.107258811618517e-05}
{"train_loss": 0.17561833560466766, "global_step": 116716, "epoch": 1311, "lr": 6.107202277292809e-05}
{"train_loss": 0.127462238073349, "global_step": 116717, "epoch": 1311, "lr": 6.107145742818253e-05}
{"train_loss": 0.2373826503753662, "global_step": 116718, "epoch": 1311, "lr": 6.107089208194854e-05}
{"train_loss": 0.1553466022014618, "global_step": 116719, "epoch": 1311, "lr": 6.10703267342262e-05}
{"train_loss": 0.1402333378791809, "global_step": 116720, "epoch": 1311, "lr": 6.10697613850156e-05}
{"train_loss": 0.18247416615486145, "global_step": 116721, "epoch": 1311, "lr": 6.10691960343168e-05}
{"train_loss": 0.12705065310001373, "global_step": 116722, "epoch": 1311, "lr": 6.106863068212989e-05}
{"train_loss": 0.15361124277114868, "global_step": 116723, "epoch": 1311, "lr": 6.106806532845492e-05}
{"train_loss": 0.1333809792995453, "global_step": 116724, "epoch": 1311, "lr": 6.106749997329199e-05}
{"train_loss": 0.13855093717575073, "global_step": 116725, "epoch": 1311, "lr": 6.106693461664117e-05}
{"train_loss": 0.1206934005022049, "global_step": 116726, "epoch": 1311, "lr": 6.106636925850253e-05}
{"train_loss": 0.13089598715305328, "global_step": 116727, "epoch": 1311, "lr": 6.106580389887616e-05}
{"train_loss": 0.1407952755689621, "global_step": 116728, "epoch": 1311, "lr": 6.106523853776213e-05}
{"train_loss": 0.14713218808174133, "global_step": 116729, "epoch": 1311, "lr": 6.106467317516051e-05}
{"train_loss": 0.15791688859462738, "global_step": 116730, "epoch": 1311, "lr": 6.106410781107137e-05}
{"train_loss": 0.15367592871189117, "global_step": 116731, "epoch": 1311, "lr": 6.106354244549481e-05}
{"train_loss": 0.21797290444374084, "global_step": 116732, "epoch": 1311, "lr": 6.106297707843088e-05}
{"train_loss": 0.17321091890335083, "global_step": 116733, "epoch": 1311, "lr": 6.106241170987967e-05}
{"train_loss": 0.12896253168582916, "global_step": 116734, "epoch": 1311, "lr": 6.106184633984126e-05}
{"train_loss": 0.10227399319410324, "global_step": 116735, "epoch": 1311, "lr": 6.106128096831572e-05}
{"train_loss": 0.11758057773113251, "global_step": 116736, "epoch": 1311, "lr": 6.106071559530312e-05}
{"train_loss": 0.19412048161029816, "global_step": 116737, "epoch": 1311, "lr": 6.106015022080354e-05}
{"train_loss": 0.1503629833459854, "global_step": 116738, "epoch": 1311, "lr": 6.105958484481707e-05}
{"train_loss": 0.14242514967918396, "global_step": 116739, "epoch": 1311, "lr": 6.105901946734377e-05}
{"train_loss": 0.2269260436296463, "global_step": 116740, "epoch": 1311, "lr": 6.105845408838371e-05}
{"train_loss": 0.08038496971130371, "global_step": 116741, "epoch": 1311, "lr": 6.105788870793698e-05}
{"train_loss": 0.20811034739017487, "global_step": 116742, "epoch": 1311, "lr": 6.105732332600365e-05}
{"train_loss": 0.15608856081962585, "global_step": 116743, "epoch": 1311, "lr": 6.10567579425838e-05}
{"train_loss": 0.21464207768440247, "global_step": 116744, "epoch": 1311, "lr": 6.105619255767749e-05}
{"train_loss": 0.12882249057292938, "global_step": 116745, "epoch": 1311, "lr": 6.105562717128484e-05}
{"train_loss": 0.14950726926326752, "global_step": 116746, "epoch": 1311, "lr": 6.105506178340587e-05}
{"train_loss": 0.11070849746465683, "global_step": 116747, "epoch": 1311, "lr": 6.105449639404069e-05}
{"train_loss": 0.1502387821674347, "global_step": 116748, "epoch": 1311, "lr": 6.105393100318936e-05}
{"train_loss": 0.16931197047233582, "global_step": 116749, "epoch": 1311, "lr": 6.105336561085198e-05}
{"train_loss": 0.08826282620429993, "global_step": 116750, "epoch": 1311, "lr": 6.105280021702861e-05}
{"train_loss": 0.20967626571655273, "global_step": 116751, "epoch": 1311, "lr": 6.105223482171932e-05}
{"train_loss": 0.11085398495197296, "global_step": 116752, "epoch": 1311, "lr": 6.105166942492417e-05}
{"train_loss": 0.17255480587482452, "global_step": 116753, "epoch": 1311, "lr": 6.105110402664327e-05}
{"train_loss": 0.13446131348609924, "global_step": 116754, "epoch": 1311, "lr": 6.105053862687671e-05}
{"train_loss": 0.17280733585357666, "global_step": 116755, "epoch": 1311, "lr": 6.104997322562451e-05}
{"train_loss": 0.14006513357162476, "global_step": 116756, "epoch": 1311, "lr": 6.10494078228868e-05}
{"train_loss": 0.12576952576637268, "global_step": 116757, "epoch": 1311, "lr": 6.104884241866361e-05}
{"train_loss": 0.09510304033756256, "global_step": 116758, "epoch": 1311, "lr": 6.104827701295504e-05}
{"train_loss": 0.16450124979019165, "global_step": 116759, "epoch": 1311, "lr": 6.104771160576117e-05}
{"train_loss": 0.15228693187236786, "global_step": 116760, "epoch": 1311, "lr": 6.104714619708207e-05}
{"train_loss": 0.1872383952140808, "global_step": 116761, "epoch": 1311, "lr": 6.104658078691782e-05}
{"train_loss": 0.12165823578834534, "global_step": 116762, "epoch": 1311, "lr": 6.104601537526848e-05}
{"train_loss": 0.10714025795459747, "global_step": 116763, "epoch": 1311, "lr": 6.104544996213414e-05}
{"train_loss": 0.11460768431425095, "global_step": 116764, "epoch": 1311, "lr": 6.104488454751489e-05}
{"train_loss": 0.20325900614261627, "global_step": 116765, "epoch": 1311, "lr": 6.104431913141077e-05}
{"train_loss": 0.2097959667444229, "global_step": 116766, "epoch": 1311, "lr": 6.104375371382188e-05}
{"train_loss": 0.14796897321102326, "global_step": 116767, "epoch": 1311, "lr": 6.10431882947483e-05, "val_loss": 5.153107166290283}
{"train_loss": 0.16031354665756226, "global_step": 116768, "epoch": 1312, "lr": 6.10426228741901e-05}
{"train_loss": 0.15255896747112274, "global_step": 116769, "epoch": 1312, "lr": 6.104205745214734e-05}
{"train_loss": 0.16287240386009216, "global_step": 116770, "epoch": 1312, "lr": 6.10414920286201e-05}
{"train_loss": 0.18105055391788483, "global_step": 116771, "epoch": 1312, "lr": 6.10409266036085e-05}
{"train_loss": 0.20025363564491272, "global_step": 116772, "epoch": 1312, "lr": 6.104036117711256e-05}
{"train_loss": 0.13159652054309845, "global_step": 116773, "epoch": 1312, "lr": 6.10397957491324e-05}
{"train_loss": 0.16613557934761047, "global_step": 116774, "epoch": 1312, "lr": 6.103923031966805e-05}
{"train_loss": 0.14020602405071259, "global_step": 116775, "epoch": 1312, "lr": 6.103866488871962e-05}
{"train_loss": 0.14513002336025238, "global_step": 116776, "epoch": 1312, "lr": 6.103809945628718e-05}
{"train_loss": 0.1839466243982315, "global_step": 116777, "epoch": 1312, "lr": 6.103753402237079e-05}
{"train_loss": 0.10993128269910812, "global_step": 116778, "epoch": 1312, "lr": 6.103696858697055e-05}
{"train_loss": 0.16639259457588196, "global_step": 116779, "epoch": 1312, "lr": 6.103640315008652e-05}
{"train_loss": 0.0802314355969429, "global_step": 116780, "epoch": 1312, "lr": 6.103583771171877e-05}
{"train_loss": 0.1893664449453354, "global_step": 116781, "epoch": 1312, "lr": 6.103527227186739e-05}
{"train_loss": 0.13321185111999512, "global_step": 116782, "epoch": 1312, "lr": 6.1034706830532464e-05}
{"train_loss": 0.13069754838943481, "global_step": 116783, "epoch": 1312, "lr": 6.103414138771405e-05}
{"train_loss": 0.18718616664409637, "global_step": 116784, "epoch": 1312, "lr": 6.103357594341222e-05}
{"train_loss": 0.17249825596809387, "global_step": 116785, "epoch": 1312, "lr": 6.1033010497627076e-05}
{"train_loss": 0.12406239658594131, "global_step": 116786, "epoch": 1312, "lr": 6.103244505035867e-05}
{"train_loss": 0.12431357055902481, "global_step": 116787, "epoch": 1312, "lr": 6.103187960160709e-05}
{"train_loss": 0.10906459391117096, "global_step": 116788, "epoch": 1312, "lr": 6.1031314151372395e-05}
{"train_loss": 0.12044436484575272, "global_step": 116789, "epoch": 1312, "lr": 6.1030748699654686e-05}
{"train_loss": 0.18919605016708374, "global_step": 116790, "epoch": 1312, "lr": 6.103018324645402e-05}
{"train_loss": 0.10350248217582703, "global_step": 116791, "epoch": 1312, "lr": 6.102961779177049e-05}
{"train_loss": 0.13320212066173553, "global_step": 116792, "epoch": 1312, "lr": 6.102905233560417e-05}
{"train_loss": 0.13649983704090118, "global_step": 116793, "epoch": 1312, "lr": 6.1028486877955113e-05}
{"train_loss": 0.17496150732040405, "global_step": 116794, "epoch": 1312, "lr": 6.102792141882342e-05}
{"train_loss": 0.07324503362178802, "global_step": 116795, "epoch": 1312, "lr": 6.1027355958209154e-05}
{"train_loss": 0.11939971148967743, "global_step": 116796, "epoch": 1312, "lr": 6.1026790496112384e-05}
{"train_loss": 0.18140771985054016, "global_step": 116797, "epoch": 1312, "lr": 6.102622503253321e-05}
{"train_loss": 0.16265125572681427, "global_step": 116798, "epoch": 1312, "lr": 6.102565956747169e-05}
{"train_loss": 0.15975899994373322, "global_step": 116799, "epoch": 1312, "lr": 6.10250941009279e-05}
{"train_loss": 0.14086803793907166, "global_step": 116800, "epoch": 1312, "lr": 6.102452863290192e-05}
{"train_loss": 0.10852298140525818, "global_step": 116801, "epoch": 1312, "lr": 6.102396316339384e-05}
{"train_loss": 0.12954314053058624, "global_step": 116802, "epoch": 1312, "lr": 6.10233976924037e-05}
{"train_loss": 0.09737172722816467, "global_step": 116803, "epoch": 1312, "lr": 6.1022832219931626e-05}
{"train_loss": 0.19029834866523743, "global_step": 116804, "epoch": 1312, "lr": 6.1022266745977655e-05}
{"train_loss": 0.1372394859790802, "global_step": 116805, "epoch": 1312, "lr": 6.1021701270541864e-05}
{"train_loss": 0.18065138161182404, "global_step": 116806, "epoch": 1312, "lr": 6.102113579362434e-05}
{"train_loss": 0.17792218923568726, "global_step": 116807, "epoch": 1312, "lr": 6.102057031522517e-05}
{"train_loss": 0.20580364763736725, "global_step": 116808, "epoch": 1312, "lr": 6.1020004835344414e-05}
{"train_loss": 0.20745818316936493, "global_step": 116809, "epoch": 1312, "lr": 6.101943935398216e-05}
{"train_loss": 0.13954484462738037, "global_step": 116810, "epoch": 1312, "lr": 6.101887387113846e-05}
{"train_loss": 0.18120461702346802, "global_step": 116811, "epoch": 1312, "lr": 6.101830838681342e-05}
{"train_loss": 0.088782899081707, "global_step": 116812, "epoch": 1312, "lr": 6.1017742901007105e-05}
{"train_loss": 0.15769454836845398, "global_step": 116813, "epoch": 1312, "lr": 6.101717741371958e-05}
{"train_loss": 0.18943870067596436, "global_step": 116814, "epoch": 1312, "lr": 6.101661192495094e-05}
{"train_loss": 0.10066279768943787, "global_step": 116815, "epoch": 1312, "lr": 6.1016046434701245e-05}
{"train_loss": 0.12708233296871185, "global_step": 116816, "epoch": 1312, "lr": 6.101548094297057e-05}
{"train_loss": 0.12010113149881363, "global_step": 116817, "epoch": 1312, "lr": 6.1014915449759e-05}
{"train_loss": 0.13998830318450928, "global_step": 116818, "epoch": 1312, "lr": 6.1014349955066616e-05}
{"train_loss": 0.11800660938024521, "global_step": 116819, "epoch": 1312, "lr": 6.101378445889349e-05}
{"train_loss": 0.15801432728767395, "global_step": 116820, "epoch": 1312, "lr": 6.1013218961239683e-05}
{"train_loss": 0.11924701184034348, "global_step": 116821, "epoch": 1312, "lr": 6.1012653462105286e-05}
{"train_loss": 0.12861588597297668, "global_step": 116822, "epoch": 1312, "lr": 6.1012087961490384e-05}
{"train_loss": 0.13404475152492523, "global_step": 116823, "epoch": 1312, "lr": 6.101152245939502e-05}
{"train_loss": 0.15433955192565918, "global_step": 116824, "epoch": 1312, "lr": 6.10109569558193e-05}
{"train_loss": 0.13068708777427673, "global_step": 116825, "epoch": 1312, "lr": 6.101039145076329e-05}
{"train_loss": 0.10168422758579254, "global_step": 116826, "epoch": 1312, "lr": 6.100982594422707e-05}
{"train_loss": 0.12735897302627563, "global_step": 116827, "epoch": 1312, "lr": 6.1009260436210715e-05}
{"train_loss": 0.12029970437288284, "global_step": 116828, "epoch": 1312, "lr": 6.100869492671429e-05}
{"train_loss": 0.22915899753570557, "global_step": 116829, "epoch": 1312, "lr": 6.1008129415737894e-05}
{"train_loss": 0.22260494530200958, "global_step": 116830, "epoch": 1312, "lr": 6.100756390328159e-05}
{"train_loss": 0.20628294348716736, "global_step": 116831, "epoch": 1312, "lr": 6.100699838934544e-05}
{"train_loss": 0.10939180850982666, "global_step": 116832, "epoch": 1312, "lr": 6.100643287392955e-05}
{"train_loss": 0.142660990357399, "global_step": 116833, "epoch": 1312, "lr": 6.100586735703396e-05}
{"train_loss": 0.1711978018283844, "global_step": 116834, "epoch": 1312, "lr": 6.100530183865878e-05}
{"train_loss": 0.20454135537147522, "global_step": 116835, "epoch": 1312, "lr": 6.1004736318804066e-05}
{"train_loss": 0.0703137144446373, "global_step": 116836, "epoch": 1312, "lr": 6.10041707974699e-05}
{"train_loss": 0.07506197690963745, "global_step": 116837, "epoch": 1312, "lr": 6.100360527465635e-05}
{"train_loss": 0.21411117911338806, "global_step": 116838, "epoch": 1312, "lr": 6.100303975036351e-05}
{"train_loss": 0.11008097231388092, "global_step": 116839, "epoch": 1312, "lr": 6.100247422459144e-05}
{"train_loss": 0.13246606290340424, "global_step": 116840, "epoch": 1312, "lr": 6.1001908697340214e-05}
{"train_loss": 0.1230640783905983, "global_step": 116841, "epoch": 1312, "lr": 6.1001343168609924e-05}
{"train_loss": 0.17071785032749176, "global_step": 116842, "epoch": 1312, "lr": 6.100077763840064e-05}
{"train_loss": 0.19471776485443115, "global_step": 116843, "epoch": 1312, "lr": 6.100021210671243e-05}
{"train_loss": 0.23687762022018433, "global_step": 116844, "epoch": 1312, "lr": 6.099964657354538e-05}
{"train_loss": 0.17632853984832764, "global_step": 116845, "epoch": 1312, "lr": 6.099908103889955e-05}
{"train_loss": 0.14662806689739227, "global_step": 116846, "epoch": 1312, "lr": 6.0998515502775044e-05}
{"train_loss": 0.12697733938694, "global_step": 116847, "epoch": 1312, "lr": 6.099794996517191e-05}
{"train_loss": 0.16324014961719513, "global_step": 116848, "epoch": 1312, "lr": 6.0997384426090244e-05}
{"train_loss": 0.09740041196346283, "global_step": 116849, "epoch": 1312, "lr": 6.099681888553011e-05}
{"train_loss": 0.0722825676202774, "global_step": 116850, "epoch": 1312, "lr": 6.099625334349158e-05}
{"train_loss": 0.08431199938058853, "global_step": 116851, "epoch": 1312, "lr": 6.0995687799974765e-05}
{"train_loss": 0.13893863558769226, "global_step": 116852, "epoch": 1312, "lr": 6.09951222549797e-05}
{"train_loss": 0.18344026803970337, "global_step": 116853, "epoch": 1312, "lr": 6.0994556708506455e-05}
{"train_loss": 0.1801137924194336, "global_step": 116854, "epoch": 1312, "lr": 6.0993991160555144e-05}
{"train_loss": 0.09995292872190475, "global_step": 116855, "epoch": 1312, "lr": 6.099342561112582e-05}
{"train_loss": 0.14668731441658533, "global_step": 116856, "epoch": 1312, "lr": 6.099286006021857e-05, "val_loss": 5.1683349609375}
{"train_loss": 0.11361026763916016, "global_step": 116857, "epoch": 1313, "lr": 6.099229450783346e-05}
{"train_loss": 0.11297213286161423, "global_step": 116858, "epoch": 1313, "lr": 6.099172895397056e-05}
{"train_loss": 0.1180596798658371, "global_step": 116859, "epoch": 1313, "lr": 6.0991163398629967e-05}
{"train_loss": 0.11925271153450012, "global_step": 116860, "epoch": 1313, "lr": 6.0990597841811745e-05}
{"train_loss": 0.14711372554302216, "global_step": 116861, "epoch": 1313, "lr": 6.0990032283515966e-05}
{"train_loss": 0.23103301227092743, "global_step": 116862, "epoch": 1313, "lr": 6.0989466723742704e-05}
{"train_loss": 0.1882101446390152, "global_step": 116863, "epoch": 1313, "lr": 6.098890116249206e-05}
{"train_loss": 0.11663586646318436, "global_step": 116864, "epoch": 1313, "lr": 6.098833559976409e-05}
{"train_loss": 0.14008301496505737, "global_step": 116865, "epoch": 1313, "lr": 6.098777003555887e-05}
{"train_loss": 0.1815275251865387, "global_step": 116866, "epoch": 1313, "lr": 6.0987204469876466e-05}
{"train_loss": 0.17979119718074799, "global_step": 116867, "epoch": 1313, "lr": 6.098663890271699e-05}
{"train_loss": 0.1136704832315445, "global_step": 116868, "epoch": 1313, "lr": 6.0986073334080475e-05}
{"train_loss": 0.15895651280879974, "global_step": 116869, "epoch": 1313, "lr": 6.098550776396702e-05}
{"train_loss": 0.14908073842525482, "global_step": 116870, "epoch": 1313, "lr": 6.09849421923767e-05}
{"train_loss": 0.1329997032880783, "global_step": 116871, "epoch": 1313, "lr": 6.098437661930959e-05}
{"train_loss": 0.11428972333669662, "global_step": 116872, "epoch": 1313, "lr": 6.098381104476577e-05}
{"train_loss": 0.2029619812965393, "global_step": 116873, "epoch": 1313, "lr": 6.09832454687453e-05}
{"train_loss": 0.0987834706902504, "global_step": 116874, "epoch": 1313, "lr": 6.0982679891248274e-05}
{"train_loss": 0.07056212425231934, "global_step": 116875, "epoch": 1313, "lr": 6.0982114312274754e-05}
{"train_loss": 0.13277000188827515, "global_step": 116876, "epoch": 1313, "lr": 6.098154873182482e-05}
{"train_loss": 0.17706258594989777, "global_step": 116877, "epoch": 1313, "lr": 6.098098314989855e-05}
{"train_loss": 0.15046468377113342, "global_step": 116878, "epoch": 1313, "lr": 6.098041756649603e-05}
{"train_loss": 0.15992547571659088, "global_step": 116879, "epoch": 1313, "lr": 6.097985198161731e-05}
{"train_loss": 0.07828915119171143, "global_step": 116880, "epoch": 1313, "lr": 6.097928639526249e-05}
{"train_loss": 0.20100697875022888, "global_step": 116881, "epoch": 1313, "lr": 6.097872080743166e-05}
{"train_loss": 0.12456046044826508, "global_step": 116882, "epoch": 1313, "lr": 6.097815521812484e-05}
{"train_loss": 0.09685780107975006, "global_step": 116883, "epoch": 1313, "lr": 6.0977589627342166e-05}
{"train_loss": 0.18179044127464294, "global_step": 116884, "epoch": 1313, "lr": 6.0977024035083686e-05}
{"train_loss": 0.1656113564968109, "global_step": 116885, "epoch": 1313, "lr": 6.097645844134947e-05}
{"train_loss": 0.12780380249023438, "global_step": 116886, "epoch": 1313, "lr": 6.097589284613962e-05}
{"train_loss": 0.06572149693965912, "global_step": 116887, "epoch": 1313, "lr": 6.097532724945417e-05}
{"train_loss": 0.12185927480459213, "global_step": 116888, "epoch": 1313, "lr": 6.0974761651293246e-05}
{"train_loss": 0.1384286731481552, "global_step": 116889, "epoch": 1313, "lr": 6.0974196051656886e-05}
{"train_loss": 0.15534305572509766, "global_step": 116890, "epoch": 1313, "lr": 6.0973630450545184e-05}
{"train_loss": 0.14066046476364136, "global_step": 116891, "epoch": 1313, "lr": 6.097306484795821e-05}
{"train_loss": 0.20181560516357422, "global_step": 116892, "epoch": 1313, "lr": 6.097249924389604e-05}
{"train_loss": 0.18128980696201324, "global_step": 116893, "epoch": 1313, "lr": 6.0971933638358746e-05}
{"train_loss": 0.1151212602853775, "global_step": 116894, "epoch": 1313, "lr": 6.097136803134641e-05}
{"train_loss": 0.06523791700601578, "global_step": 116895, "epoch": 1313, "lr": 6.0970802422859116e-05}
{"train_loss": 0.11290449649095535, "global_step": 116896, "epoch": 1313, "lr": 6.097023681289693e-05}
{"train_loss": 0.1096443235874176, "global_step": 116897, "epoch": 1313, "lr": 6.096967120145991e-05}
{"train_loss": 0.09089881926774979, "global_step": 116898, "epoch": 1313, "lr": 6.096910558854816e-05}
{"train_loss": 0.11427278071641922, "global_step": 116899, "epoch": 1313, "lr": 6.096853997416176e-05}
{"train_loss": 0.11835959553718567, "global_step": 116900, "epoch": 1313, "lr": 6.096797435830076e-05}
{"train_loss": 0.1920623630285263, "global_step": 116901, "epoch": 1313, "lr": 6.096740874096526e-05}
{"train_loss": 0.175388365983963, "global_step": 116902, "epoch": 1313, "lr": 6.096684312215533e-05}
{"train_loss": 0.1563129723072052, "global_step": 116903, "epoch": 1313, "lr": 6.096627750187103e-05}
{"train_loss": 0.1336161345243454, "global_step": 116904, "epoch": 1313, "lr": 6.096571188011245e-05}
{"train_loss": 0.07722701132297516, "global_step": 116905, "epoch": 1313, "lr": 6.096514625687967e-05}
{"train_loss": 0.14748965203762054, "global_step": 116906, "epoch": 1313, "lr": 6.0964580632172755e-05}
{"train_loss": 0.1170160099864006, "global_step": 116907, "epoch": 1313, "lr": 6.0964015005991784e-05}
{"train_loss": 0.1557450294494629, "global_step": 116908, "epoch": 1313, "lr": 6.0963449378336834e-05}
{"train_loss": 0.0976090207695961, "global_step": 116909, "epoch": 1313, "lr": 6.0962883749208e-05}
{"train_loss": 0.13648968935012817, "global_step": 116910, "epoch": 1313, "lr": 6.096231811860532e-05}
{"train_loss": 0.1090477779507637, "global_step": 116911, "epoch": 1313, "lr": 6.0961752486528886e-05}
{"train_loss": 0.2129979431629181, "global_step": 116912, "epoch": 1313, "lr": 6.096118685297879e-05}
{"train_loss": 0.13953271508216858, "global_step": 116913, "epoch": 1313, "lr": 6.0960621217955094e-05}
{"train_loss": 0.18513637781143188, "global_step": 116914, "epoch": 1313, "lr": 6.096005558145788e-05}
{"train_loss": 0.19402095675468445, "global_step": 116915, "epoch": 1313, "lr": 6.09594899434872e-05}
{"train_loss": 0.10634855180978775, "global_step": 116916, "epoch": 1313, "lr": 6.095892430404316e-05}
{"train_loss": 0.1832250952720642, "global_step": 116917, "epoch": 1313, "lr": 6.095835866312584e-05}
{"train_loss": 0.16058653593063354, "global_step": 116918, "epoch": 1313, "lr": 6.095779302073529e-05}
{"train_loss": 0.1719168722629547, "global_step": 116919, "epoch": 1313, "lr": 6.095722737687161e-05}
{"train_loss": 0.14753592014312744, "global_step": 116920, "epoch": 1313, "lr": 6.095666173153485e-05}
{"train_loss": 0.09650254249572754, "global_step": 116921, "epoch": 1313, "lr": 6.09560960847251e-05}
{"train_loss": 0.13983936607837677, "global_step": 116922, "epoch": 1313, "lr": 6.0955530436442445e-05}
{"train_loss": 0.1975223422050476, "global_step": 116923, "epoch": 1313, "lr": 6.0954964786686956e-05}
{"train_loss": 0.14894844591617584, "global_step": 116924, "epoch": 1313, "lr": 6.095439913545869e-05}
{"train_loss": 0.14424623548984528, "global_step": 116925, "epoch": 1313, "lr": 6.095383348275775e-05}
{"train_loss": 0.19905054569244385, "global_step": 116926, "epoch": 1313, "lr": 6.0953267828584205e-05}
{"train_loss": 0.11821715533733368, "global_step": 116927, "epoch": 1313, "lr": 6.0952702172938116e-05}
{"train_loss": 0.13621953129768372, "global_step": 116928, "epoch": 1313, "lr": 6.095213651581959e-05}
{"train_loss": 0.21689781546592712, "global_step": 116929, "epoch": 1313, "lr": 6.095157085722867e-05}
{"train_loss": 0.12882429361343384, "global_step": 116930, "epoch": 1313, "lr": 6.095100519716543e-05}
{"train_loss": 0.0787210687994957, "global_step": 116931, "epoch": 1313, "lr": 6.0950439535629977e-05}
{"train_loss": 0.1356021761894226, "global_step": 116932, "epoch": 1313, "lr": 6.0949873872622365e-05}
{"train_loss": 0.16100849211215973, "global_step": 116933, "epoch": 1313, "lr": 6.0949308208142664e-05}
{"train_loss": 0.10248835384845734, "global_step": 116934, "epoch": 1313, "lr": 6.094874254219098e-05}
{"train_loss": 0.13810212910175323, "global_step": 116935, "epoch": 1313, "lr": 6.094817687476737e-05}
{"train_loss": 0.1649540364742279, "global_step": 116936, "epoch": 1313, "lr": 6.09476112058719e-05}
{"train_loss": 0.18551866710186005, "global_step": 116937, "epoch": 1313, "lr": 6.0947045535504666e-05}
{"train_loss": 0.0921357199549675, "global_step": 116938, "epoch": 1313, "lr": 6.0946479863665726e-05}
{"train_loss": 0.11881557106971741, "global_step": 116939, "epoch": 1313, "lr": 6.094591419035518e-05}
{"train_loss": 0.15217037498950958, "global_step": 116940, "epoch": 1313, "lr": 6.0945348515573084e-05}
{"train_loss": 0.08779096603393555, "global_step": 116941, "epoch": 1313, "lr": 6.094478283931952e-05}
{"train_loss": 0.11268971860408783, "global_step": 116942, "epoch": 1313, "lr": 6.094421716159455e-05}
{"train_loss": 0.1403043568134308, "global_step": 116943, "epoch": 1313, "lr": 6.094365148239828e-05}
{"train_loss": 0.1459738165140152, "global_step": 116944, "epoch": 1313, "lr": 6.094308580173076e-05}
{"train_loss": 0.14094311810946197, "global_step": 116945, "epoch": 1313, "lr": 6.094252011959208e-05, "val_loss": 5.22089147567749}
{"train_loss": 0.13602012395858765, "global_step": 116946, "epoch": 1314, "lr": 6.094195443598232e-05}
{"train_loss": 0.07721563428640366, "global_step": 116947, "epoch": 1314, "lr": 6.094138875090153e-05}
{"train_loss": 0.1196884959936142, "global_step": 116948, "epoch": 1314, "lr": 6.0940823064349825e-05}
{"train_loss": 0.1602877527475357, "global_step": 116949, "epoch": 1314, "lr": 6.094025737632725e-05}
{"train_loss": 0.0994914174079895, "global_step": 116950, "epoch": 1314, "lr": 6.0939691686833874e-05}
{"train_loss": 0.13957548141479492, "global_step": 116951, "epoch": 1314, "lr": 6.093912599586979e-05}
{"train_loss": 0.15954282879829407, "global_step": 116952, "epoch": 1314, "lr": 6.0938560303435096e-05}
{"train_loss": 0.16435754299163818, "global_step": 116953, "epoch": 1314, "lr": 6.093799460952984e-05}
{"train_loss": 0.19934453070163727, "global_step": 116954, "epoch": 1314, "lr": 6.0937428914154095e-05}
{"train_loss": 0.07517925649881363, "global_step": 116955, "epoch": 1314, "lr": 6.093686321730796e-05}
{"train_loss": 0.16459187865257263, "global_step": 116956, "epoch": 1314, "lr": 6.093629751899148e-05}
{"train_loss": 0.14812488853931427, "global_step": 116957, "epoch": 1314, "lr": 6.0935731819204755e-05}
{"train_loss": 0.12632758915424347, "global_step": 116958, "epoch": 1314, "lr": 6.093516611794786e-05}
{"train_loss": 0.1433652937412262, "global_step": 116959, "epoch": 1314, "lr": 6.0934600415220854e-05}
{"train_loss": 0.16051506996154785, "global_step": 116960, "epoch": 1314, "lr": 6.093403471102383e-05}
{"train_loss": 0.12081870436668396, "global_step": 116961, "epoch": 1314, "lr": 6.093346900535687e-05}
{"train_loss": 0.2075381875038147, "global_step": 116962, "epoch": 1314, "lr": 6.093290329822002e-05}
{"train_loss": 0.13833478093147278, "global_step": 116963, "epoch": 1314, "lr": 6.093233758961339e-05}
{"train_loss": 0.14028997719287872, "global_step": 116964, "epoch": 1314, "lr": 6.093177187953702e-05}
{"train_loss": 0.11803491413593292, "global_step": 116965, "epoch": 1314, "lr": 6.093120616799103e-05}
{"train_loss": 0.10899537801742554, "global_step": 116966, "epoch": 1314, "lr": 6.0930640454975465e-05}
{"train_loss": 0.10454975068569183, "global_step": 116967, "epoch": 1314, "lr": 6.0930074740490416e-05}
{"train_loss": 0.12077543139457703, "global_step": 116968, "epoch": 1314, "lr": 6.0929509024535934e-05}
{"train_loss": 0.15033692121505737, "global_step": 116969, "epoch": 1314, "lr": 6.092894330711212e-05}
{"train_loss": 0.10342775285243988, "global_step": 116970, "epoch": 1314, "lr": 6.092837758821903e-05}
{"train_loss": 0.06233643367886543, "global_step": 116971, "epoch": 1314, "lr": 6.0927811867856764e-05}
{"train_loss": 0.127945676445961, "global_step": 116972, "epoch": 1314, "lr": 6.09272461460254e-05}
{"train_loss": 0.1332242786884308, "global_step": 116973, "epoch": 1314, "lr": 6.092668042272498e-05}
{"train_loss": 0.1647319495677948, "global_step": 116974, "epoch": 1314, "lr": 6.0926114697955614e-05}
{"train_loss": 0.21871787309646606, "global_step": 116975, "epoch": 1314, "lr": 6.092554897171736e-05}
{"train_loss": 0.1470794379711151, "global_step": 116976, "epoch": 1314, "lr": 6.09249832440103e-05}
{"train_loss": 0.12981459498405457, "global_step": 116977, "epoch": 1314, "lr": 6.092441751483451e-05}
{"train_loss": 0.12614698708057404, "global_step": 116978, "epoch": 1314, "lr": 6.092385178419007e-05}
{"train_loss": 0.1126890704035759, "global_step": 116979, "epoch": 1314, "lr": 6.0923286052077044e-05}
{"train_loss": 0.14000600576400757, "global_step": 116980, "epoch": 1314, "lr": 6.092272031849552e-05}
{"train_loss": 0.2283870428800583, "global_step": 116981, "epoch": 1314, "lr": 6.092215458344557e-05}
{"train_loss": 0.08637166768312454, "global_step": 116982, "epoch": 1314, "lr": 6.0921588846927255e-05}
{"train_loss": 0.06447935104370117, "global_step": 116983, "epoch": 1314, "lr": 6.092102310894068e-05}
{"train_loss": 0.143141508102417, "global_step": 116984, "epoch": 1314, "lr": 6.0920457369485905e-05}
{"train_loss": 0.20515358448028564, "global_step": 116985, "epoch": 1314, "lr": 6.0919891628563006e-05}
{"train_loss": 0.1942978948354721, "global_step": 116986, "epoch": 1314, "lr": 6.091932588617206e-05}
{"train_loss": 0.09700478613376617, "global_step": 116987, "epoch": 1314, "lr": 6.091876014231316e-05}
{"train_loss": 0.1697438508272171, "global_step": 116988, "epoch": 1314, "lr": 6.091819439698634e-05}
{"train_loss": 0.061963390558958054, "global_step": 116989, "epoch": 1314, "lr": 6.0917628650191704e-05}
{"train_loss": 0.16740109026432037, "global_step": 116990, "epoch": 1314, "lr": 6.091706290192933e-05}
{"train_loss": 0.11997134238481522, "global_step": 116991, "epoch": 1314, "lr": 6.09164971521993e-05}
{"train_loss": 0.15432208776474, "global_step": 116992, "epoch": 1314, "lr": 6.091593140100167e-05}
{"train_loss": 0.14330746233463287, "global_step": 116993, "epoch": 1314, "lr": 6.091536564833653e-05}
{"train_loss": 0.10542105883359909, "global_step": 116994, "epoch": 1314, "lr": 6.0914799894203953e-05}
{"train_loss": 0.12747536599636078, "global_step": 116995, "epoch": 1314, "lr": 6.091423413860402e-05}
{"train_loss": 0.147039532661438, "global_step": 116996, "epoch": 1314, "lr": 6.0913668381536794e-05}
{"train_loss": 0.22121693193912506, "global_step": 116997, "epoch": 1314, "lr": 6.091310262300236e-05}
{"train_loss": 0.14645938575267792, "global_step": 116998, "epoch": 1314, "lr": 6.091253686300078e-05}
{"train_loss": 0.12019681930541992, "global_step": 116999, "epoch": 1314, "lr": 6.091197110153216e-05}
{"train_loss": 0.09902627021074295, "global_step": 117000, "epoch": 1314, "lr": 6.091140533859655e-05}
{"train_loss": 0.18668782711029053, "global_step": 117001, "epoch": 1314, "lr": 6.091083957419403e-05}
{"train_loss": 0.1326044797897339, "global_step": 117002, "epoch": 1314, "lr": 6.091027380832469e-05}
{"train_loss": 0.11014559119939804, "global_step": 117003, "epoch": 1314, "lr": 6.09097080409886e-05}
{"train_loss": 0.13410675525665283, "global_step": 117004, "epoch": 1314, "lr": 6.090914227218583e-05}
{"train_loss": 0.09435287863016129, "global_step": 117005, "epoch": 1314, "lr": 6.090857650191645e-05}
{"train_loss": 0.09649301320314407, "global_step": 117006, "epoch": 1314, "lr": 6.090801073018055e-05}
{"train_loss": 0.13203808665275574, "global_step": 117007, "epoch": 1314, "lr": 6.0907444956978196e-05}
{"train_loss": 0.07804358005523682, "global_step": 117008, "epoch": 1314, "lr": 6.0906879182309483e-05}
{"train_loss": 0.15439224243164062, "global_step": 117009, "epoch": 1314, "lr": 6.090631340617446e-05}
{"train_loss": 0.14515872299671173, "global_step": 117010, "epoch": 1314, "lr": 6.090574762857322e-05}
{"train_loss": 0.07712321728467941, "global_step": 117011, "epoch": 1314, "lr": 6.090518184950583e-05}
{"train_loss": 0.1286606639623642, "global_step": 117012, "epoch": 1314, "lr": 6.0904616068972385e-05}
{"train_loss": 0.12803435325622559, "global_step": 117013, "epoch": 1314, "lr": 6.0904050286972935e-05}
{"train_loss": 0.19819240272045135, "global_step": 117014, "epoch": 1314, "lr": 6.090348450350757e-05}
{"train_loss": 0.07981663197278976, "global_step": 117015, "epoch": 1314, "lr": 6.0902918718576374e-05}
{"train_loss": 0.16779571771621704, "global_step": 117016, "epoch": 1314, "lr": 6.0902352932179406e-05}
{"train_loss": 0.2406107634305954, "global_step": 117017, "epoch": 1314, "lr": 6.0901787144316746e-05}
{"train_loss": 0.10968533903360367, "global_step": 117018, "epoch": 1314, "lr": 6.090122135498848e-05}
{"train_loss": 0.10593331605195999, "global_step": 117019, "epoch": 1314, "lr": 6.0900655564194675e-05}
{"train_loss": 0.17367234826087952, "global_step": 117020, "epoch": 1314, "lr": 6.0900089771935406e-05}
{"train_loss": 0.13670270144939423, "global_step": 117021, "epoch": 1314, "lr": 6.089952397821076e-05}
{"train_loss": 0.10309264063835144, "global_step": 117022, "epoch": 1314, "lr": 6.0898958183020793e-05}
{"train_loss": 0.12523987889289856, "global_step": 117023, "epoch": 1314, "lr": 6.089839238636561e-05}
{"train_loss": 0.08317061513662338, "global_step": 117024, "epoch": 1314, "lr": 6.089782658824527e-05}
{"train_loss": 0.16344718635082245, "global_step": 117025, "epoch": 1314, "lr": 6.0897260788659835e-05}
{"train_loss": 0.15345075726509094, "global_step": 117026, "epoch": 1314, "lr": 6.08966949876094e-05}
{"train_loss": 0.13401959836483002, "global_step": 117027, "epoch": 1314, "lr": 6.089612918509404e-05}
{"train_loss": 0.09375478327274323, "global_step": 117028, "epoch": 1314, "lr": 6.089556338111384e-05}
{"train_loss": 0.1274971216917038, "global_step": 117029, "epoch": 1314, "lr": 6.089499757566886e-05}
{"train_loss": 0.24725379049777985, "global_step": 117030, "epoch": 1314, "lr": 6.089443176875918e-05}
{"train_loss": 0.13379202783107758, "global_step": 117031, "epoch": 1314, "lr": 6.089386596038487e-05}
{"train_loss": 0.15458229184150696, "global_step": 117032, "epoch": 1314, "lr": 6.089330015054601e-05}
{"train_loss": 0.12015055119991302, "global_step": 117033, "epoch": 1314, "lr": 6.089273433924267e-05}
{"train_loss": 0.13608380750323948, "global_step": 117034, "epoch": 1314, "lr": 6.089216852647496e-05, "val_loss": 5.375277519226074}
{"train_loss": 0.11093774437904358, "global_step": 117035, "epoch": 1315, "lr": 6.089160271224291e-05}
{"train_loss": 0.15007562935352325, "global_step": 117036, "epoch": 1315, "lr": 6.0891036896546626e-05}
{"train_loss": 0.10207002609968185, "global_step": 117037, "epoch": 1315, "lr": 6.089047107938617e-05}
{"train_loss": 0.07257773727178574, "global_step": 117038, "epoch": 1315, "lr": 6.0889905260761636e-05}
{"train_loss": 0.20977617800235748, "global_step": 117039, "epoch": 1315, "lr": 6.088933944067308e-05}
{"train_loss": 0.17186535894870758, "global_step": 117040, "epoch": 1315, "lr": 6.088877361912058e-05}
{"train_loss": 0.12687502801418304, "global_step": 117041, "epoch": 1315, "lr": 6.0888207796104214e-05}
{"train_loss": 0.17631521821022034, "global_step": 117042, "epoch": 1315, "lr": 6.088764197162406e-05}
{"train_loss": 0.14645560085773468, "global_step": 117043, "epoch": 1315, "lr": 6.088707614568021e-05}
{"train_loss": 0.08958269655704498, "global_step": 117044, "epoch": 1315, "lr": 6.088651031827271e-05}
{"train_loss": 0.16343683004379272, "global_step": 117045, "epoch": 1315, "lr": 6.088594448940167e-05}
{"train_loss": 0.1825772225856781, "global_step": 117046, "epoch": 1315, "lr": 6.088537865906714e-05}
{"train_loss": 0.1708674281835556, "global_step": 117047, "epoch": 1315, "lr": 6.088481282726919e-05}
{"train_loss": 0.19042614102363586, "global_step": 117048, "epoch": 1315, "lr": 6.0884246994007924e-05}
{"train_loss": 0.13925185799598694, "global_step": 117049, "epoch": 1315, "lr": 6.08836811592834e-05}
{"train_loss": 0.14855386316776276, "global_step": 117050, "epoch": 1315, "lr": 6.08831153230957e-05}
{"train_loss": 0.0943012461066246, "global_step": 117051, "epoch": 1315, "lr": 6.088254948544488e-05}
{"train_loss": 0.19155296683311462, "global_step": 117052, "epoch": 1315, "lr": 6.0881983646331056e-05}
{"train_loss": 0.1998642086982727, "global_step": 117053, "epoch": 1315, "lr": 6.088141780575427e-05}
{"train_loss": 0.09430902451276779, "global_step": 117054, "epoch": 1315, "lr": 6.088085196371462e-05}
{"train_loss": 0.11722928285598755, "global_step": 117055, "epoch": 1315, "lr": 6.088028612021216e-05}
{"train_loss": 0.11224304139614105, "global_step": 117056, "epoch": 1315, "lr": 6.087972027524699e-05}
{"train_loss": 0.12468969076871872, "global_step": 117057, "epoch": 1315, "lr": 6.087915442881916e-05}
{"train_loss": 0.10203315317630768, "global_step": 117058, "epoch": 1315, "lr": 6.0878588580928776e-05}
{"train_loss": 0.15632353723049164, "global_step": 117059, "epoch": 1315, "lr": 6.087802273157589e-05}
{"train_loss": 0.17784398794174194, "global_step": 117060, "epoch": 1315, "lr": 6.087745688076059e-05}
{"train_loss": 0.12251283973455429, "global_step": 117061, "epoch": 1315, "lr": 6.087689102848294e-05}
{"train_loss": 0.1300782710313797, "global_step": 117062, "epoch": 1315, "lr": 6.087632517474303e-05}
{"train_loss": 0.1285366415977478, "global_step": 117063, "epoch": 1315, "lr": 6.0875759319540935e-05}
{"train_loss": 0.10846886783838272, "global_step": 117064, "epoch": 1315, "lr": 6.087519346287672e-05}
{"train_loss": 0.1061880812048912, "global_step": 117065, "epoch": 1315, "lr": 6.087462760475048e-05}
{"train_loss": 0.09795106202363968, "global_step": 117066, "epoch": 1315, "lr": 6.087406174516227e-05}
{"train_loss": 0.10542092472314835, "global_step": 117067, "epoch": 1315, "lr": 6.087349588411217e-05}
{"train_loss": 0.10067327320575714, "global_step": 117068, "epoch": 1315, "lr": 6.087293002160027e-05}
{"train_loss": 0.11049661040306091, "global_step": 117069, "epoch": 1315, "lr": 6.087236415762663e-05}
{"train_loss": 0.1274104118347168, "global_step": 117070, "epoch": 1315, "lr": 6.087179829219134e-05}
{"train_loss": 0.1199495941400528, "global_step": 117071, "epoch": 1315, "lr": 6.0871232425294465e-05}
{"train_loss": 0.19473278522491455, "global_step": 117072, "epoch": 1315, "lr": 6.087066655693608e-05}
{"train_loss": 0.18295736610889435, "global_step": 117073, "epoch": 1315, "lr": 6.087010068711627e-05}
{"train_loss": 0.1294938027858734, "global_step": 117074, "epoch": 1315, "lr": 6.0869534815835115e-05}
{"train_loss": 0.10285884886980057, "global_step": 117075, "epoch": 1315, "lr": 6.086896894309267e-05}
{"train_loss": 0.16376134753227234, "global_step": 117076, "epoch": 1315, "lr": 6.086840306888904e-05}
{"train_loss": 0.1307600438594818, "global_step": 117077, "epoch": 1315, "lr": 6.0867837193224276e-05}
{"train_loss": 0.16230596601963043, "global_step": 117078, "epoch": 1315, "lr": 6.086727131609846e-05}
{"train_loss": 0.10898457467556, "global_step": 117079, "epoch": 1315, "lr": 6.0866705437511685e-05}
{"train_loss": 0.1419636458158493, "global_step": 117080, "epoch": 1315, "lr": 6.086613955746401e-05}
{"train_loss": 0.165970116853714, "global_step": 117081, "epoch": 1315, "lr": 6.086557367595551e-05}
{"train_loss": 0.1584295928478241, "global_step": 117082, "epoch": 1315, "lr": 6.086500779298626e-05}
{"train_loss": 0.10367333889007568, "global_step": 117083, "epoch": 1315, "lr": 6.086444190855636e-05}
{"train_loss": 0.11606809496879578, "global_step": 117084, "epoch": 1315, "lr": 6.0863876022665856e-05}
{"train_loss": 0.13129304349422455, "global_step": 117085, "epoch": 1315, "lr": 6.086331013531483e-05}
{"train_loss": 0.13695326447486877, "global_step": 117086, "epoch": 1315, "lr": 6.0862744246503375e-05}
{"train_loss": 0.14930060505867004, "global_step": 117087, "epoch": 1315, "lr": 6.086217835623156e-05}
{"train_loss": 0.13633643090724945, "global_step": 117088, "epoch": 1315, "lr": 6.086161246449945e-05}
{"train_loss": 0.10576146841049194, "global_step": 117089, "epoch": 1315, "lr": 6.086104657130713e-05}
{"train_loss": 0.16770996153354645, "global_step": 117090, "epoch": 1315, "lr": 6.086048067665467e-05}
{"train_loss": 0.14174367487430573, "global_step": 117091, "epoch": 1315, "lr": 6.085991478054216e-05}
{"train_loss": 0.12275709211826324, "global_step": 117092, "epoch": 1315, "lr": 6.0859348882969655e-05}
{"train_loss": 0.15661892294883728, "global_step": 117093, "epoch": 1315, "lr": 6.0858782983937254e-05}
{"train_loss": 0.15790368616580963, "global_step": 117094, "epoch": 1315, "lr": 6.085821708344502e-05}
{"train_loss": 0.18824537098407745, "global_step": 117095, "epoch": 1315, "lr": 6.085765118149302e-05}
{"train_loss": 0.15827204287052155, "global_step": 117096, "epoch": 1315, "lr": 6.085708527808136e-05}
{"train_loss": 0.08251691609621048, "global_step": 117097, "epoch": 1315, "lr": 6.0856519373210084e-05}
{"train_loss": 0.14268407225608826, "global_step": 117098, "epoch": 1315, "lr": 6.085595346687929e-05}
{"train_loss": 0.2301427572965622, "global_step": 117099, "epoch": 1315, "lr": 6.085538755908904e-05}
{"train_loss": 0.1685691922903061, "global_step": 117100, "epoch": 1315, "lr": 6.085482164983941e-05}
{"train_loss": 0.16188931465148926, "global_step": 117101, "epoch": 1315, "lr": 6.085425573913049e-05}
{"train_loss": 0.14131861925125122, "global_step": 117102, "epoch": 1315, "lr": 6.0853689826962345e-05}
{"train_loss": 0.12378964573144913, "global_step": 117103, "epoch": 1315, "lr": 6.085312391333505e-05}
{"train_loss": 0.17066176235675812, "global_step": 117104, "epoch": 1315, "lr": 6.0852557998248694e-05}
{"train_loss": 0.23173019289970398, "global_step": 117105, "epoch": 1315, "lr": 6.085199208170333e-05}
{"train_loss": 0.1238468810915947, "global_step": 117106, "epoch": 1315, "lr": 6.085142616369907e-05}
{"train_loss": 0.1429847776889801, "global_step": 117107, "epoch": 1315, "lr": 6.0850860244235954e-05}
{"train_loss": 0.12184469401836395, "global_step": 117108, "epoch": 1315, "lr": 6.0850294323314075e-05}
{"train_loss": 0.17673031985759735, "global_step": 117109, "epoch": 1315, "lr": 6.0849728400933496e-05}
{"train_loss": 0.12124989181756973, "global_step": 117110, "epoch": 1315, "lr": 6.084916247709431e-05}
{"train_loss": 0.08407697826623917, "global_step": 117111, "epoch": 1315, "lr": 6.0848596551796586e-05}
{"train_loss": 0.13351646065711975, "global_step": 117112, "epoch": 1315, "lr": 6.084803062504041e-05}
{"train_loss": 0.10617600381374359, "global_step": 117113, "epoch": 1315, "lr": 6.084746469682583e-05}
{"train_loss": 0.17522504925727844, "global_step": 117114, "epoch": 1315, "lr": 6.0846898767152946e-05}
{"train_loss": 0.11393643170595169, "global_step": 117115, "epoch": 1315, "lr": 6.084633283602183e-05}
{"train_loss": 0.10032426565885544, "global_step": 117116, "epoch": 1315, "lr": 6.0845766903432554e-05}
{"train_loss": 0.13705506920814514, "global_step": 117117, "epoch": 1315, "lr": 6.084520096938522e-05}
{"train_loss": 0.09189356118440628, "global_step": 117118, "epoch": 1315, "lr": 6.084463503387986e-05}
{"train_loss": 0.18912316858768463, "global_step": 117119, "epoch": 1315, "lr": 6.084406909691658e-05}
{"train_loss": 0.2014761120080948, "global_step": 117120, "epoch": 1315, "lr": 6.084350315849544e-05}
{"train_loss": 0.0954006090760231, "global_step": 117121, "epoch": 1315, "lr": 6.084293721861652e-05}
{"train_loss": 0.10844117403030396, "global_step": 117122, "epoch": 1315, "lr": 6.084237127727991e-05}
{"train_loss": 0.13995514893799685, "global_step": 117123, "epoch": 1315, "lr": 6.084180533448567e-05, "val_loss": 5.293726444244385, "train_action_mse_error": 36.64570617675781}
{"train_loss": 0.14617548882961273, "global_step": 117124, "epoch": 1316, "lr": 6.084123939023388e-05}
{"train_loss": 0.10787025094032288, "global_step": 117125, "epoch": 1316, "lr": 6.084067344452462e-05}
{"train_loss": 0.11685840040445328, "global_step": 117126, "epoch": 1316, "lr": 6.084010749735796e-05}
{"train_loss": 0.1518138200044632, "global_step": 117127, "epoch": 1316, "lr": 6.083954154873398e-05}
{"train_loss": 0.13394100964069366, "global_step": 117128, "epoch": 1316, "lr": 6.0838975598652756e-05}
{"train_loss": 0.08480019867420197, "global_step": 117129, "epoch": 1316, "lr": 6.083840964711437e-05}
{"train_loss": 0.124659463763237, "global_step": 117130, "epoch": 1316, "lr": 6.083784369411889e-05}
{"train_loss": 0.05927887558937073, "global_step": 117131, "epoch": 1316, "lr": 6.083727773966639e-05}
{"train_loss": 0.1560167670249939, "global_step": 117132, "epoch": 1316, "lr": 6.0836711783756953e-05}
{"train_loss": 0.1278940737247467, "global_step": 117133, "epoch": 1316, "lr": 6.0836145826390634e-05}
{"train_loss": 0.2416418492794037, "global_step": 117134, "epoch": 1316, "lr": 6.083557986756755e-05}
{"train_loss": 0.0800451934337616, "global_step": 117135, "epoch": 1316, "lr": 6.083501390728774e-05}
{"train_loss": 0.14889273047447205, "global_step": 117136, "epoch": 1316, "lr": 6.083444794555132e-05}
{"train_loss": 0.2083144336938858, "global_step": 117137, "epoch": 1316, "lr": 6.083388198235832e-05}
{"train_loss": 0.12282063066959381, "global_step": 117138, "epoch": 1316, "lr": 6.083331601770884e-05}
{"train_loss": 0.1387118399143219, "global_step": 117139, "epoch": 1316, "lr": 6.083275005160295e-05}
{"train_loss": 0.18883530795574188, "global_step": 117140, "epoch": 1316, "lr": 6.083218408404073e-05}
{"train_loss": 0.20639316737651825, "global_step": 117141, "epoch": 1316, "lr": 6.083161811502226e-05}
{"train_loss": 0.09831593930721283, "global_step": 117142, "epoch": 1316, "lr": 6.0831052144547605e-05}
{"train_loss": 0.14865604043006897, "global_step": 117143, "epoch": 1316, "lr": 6.083048617261685e-05}
{"train_loss": 0.14345240592956543, "global_step": 117144, "epoch": 1316, "lr": 6.0829920199230064e-05}
{"train_loss": 0.13100504875183105, "global_step": 117145, "epoch": 1316, "lr": 6.0829354224387334e-05}
{"train_loss": 0.06342825293540955, "global_step": 117146, "epoch": 1316, "lr": 6.082878824808873e-05}
{"train_loss": 0.1246795728802681, "global_step": 117147, "epoch": 1316, "lr": 6.082822227033432e-05}
{"train_loss": 0.10996407270431519, "global_step": 117148, "epoch": 1316, "lr": 6.0827656291124194e-05}
{"train_loss": 0.20406360924243927, "global_step": 117149, "epoch": 1316, "lr": 6.082709031045841e-05}
{"train_loss": 0.20243003964424133, "global_step": 117150, "epoch": 1316, "lr": 6.0826524328337065e-05}
{"train_loss": 0.1797582358121872, "global_step": 117151, "epoch": 1316, "lr": 6.082595834476021e-05}
{"train_loss": 0.1621427834033966, "global_step": 117152, "epoch": 1316, "lr": 6.082539235972796e-05}
{"train_loss": 0.1638088822364807, "global_step": 117153, "epoch": 1316, "lr": 6.0824826373240364e-05}
{"train_loss": 0.1755393147468567, "global_step": 117154, "epoch": 1316, "lr": 6.082426038529749e-05}
{"train_loss": 0.15675583481788635, "global_step": 117155, "epoch": 1316, "lr": 6.082369439589945e-05}
{"train_loss": 0.18825897574424744, "global_step": 117156, "epoch": 1316, "lr": 6.082312840504627e-05}
{"train_loss": 0.24758657813072205, "global_step": 117157, "epoch": 1316, "lr": 6.082256241273807e-05}
{"train_loss": 0.124339759349823, "global_step": 117158, "epoch": 1316, "lr": 6.0821996418974904e-05}
{"train_loss": 0.18326126039028168, "global_step": 117159, "epoch": 1316, "lr": 6.0821430423756843e-05}
{"train_loss": 0.13572083413600922, "global_step": 117160, "epoch": 1316, "lr": 6.082086442708398e-05}
{"train_loss": 0.12171396613121033, "global_step": 117161, "epoch": 1316, "lr": 6.082029842895638e-05}
{"train_loss": 0.16401909291744232, "global_step": 117162, "epoch": 1316, "lr": 6.081973242937413e-05}
{"train_loss": 0.12334738671779633, "global_step": 117163, "epoch": 1316, "lr": 6.081916642833729e-05}
{"train_loss": 0.15855351090431213, "global_step": 117164, "epoch": 1316, "lr": 6.081860042584595e-05}
{"train_loss": 0.1601363718509674, "global_step": 117165, "epoch": 1316, "lr": 6.0818034421900185e-05}
{"train_loss": 0.15833136439323425, "global_step": 117166, "epoch": 1316, "lr": 6.081746841650006e-05}
{"train_loss": 0.1189970150589943, "global_step": 117167, "epoch": 1316, "lr": 6.081690240964566e-05}
{"train_loss": 0.1334679275751114, "global_step": 117168, "epoch": 1316, "lr": 6.081633640133706e-05}
{"train_loss": 0.13952983915805817, "global_step": 117169, "epoch": 1316, "lr": 6.0815770391574325e-05}
{"train_loss": 0.16964450478553772, "global_step": 117170, "epoch": 1316, "lr": 6.081520438035755e-05}
{"train_loss": 0.1885022222995758, "global_step": 117171, "epoch": 1316, "lr": 6.081463836768679e-05}
{"train_loss": 0.12331832945346832, "global_step": 117172, "epoch": 1316, "lr": 6.081407235356216e-05}
{"train_loss": 0.12027513980865479, "global_step": 117173, "epoch": 1316, "lr": 6.0813506337983696e-05}
{"train_loss": 0.16523829102516174, "global_step": 117174, "epoch": 1316, "lr": 6.081294032095148e-05}
{"train_loss": 0.1335456371307373, "global_step": 117175, "epoch": 1316, "lr": 6.081237430246562e-05}
{"train_loss": 0.09506721794605255, "global_step": 117176, "epoch": 1316, "lr": 6.081180828252616e-05}
{"train_loss": 0.16590923070907593, "global_step": 117177, "epoch": 1316, "lr": 6.081124226113317e-05}
{"train_loss": 0.11294091492891312, "global_step": 117178, "epoch": 1316, "lr": 6.081067623828676e-05}
{"train_loss": 0.1540837585926056, "global_step": 117179, "epoch": 1316, "lr": 6.081011021398697e-05}
{"train_loss": 0.20801039040088654, "global_step": 117180, "epoch": 1316, "lr": 6.080954418823389e-05}
{"train_loss": 0.16847211122512817, "global_step": 117181, "epoch": 1316, "lr": 6.080897816102761e-05}
{"train_loss": 0.12308615446090698, "global_step": 117182, "epoch": 1316, "lr": 6.080841213236819e-05}
{"train_loss": 0.1360045075416565, "global_step": 117183, "epoch": 1316, "lr": 6.0807846102255705e-05}
{"train_loss": 0.11765644699335098, "global_step": 117184, "epoch": 1316, "lr": 6.080728007069024e-05}
{"train_loss": 0.16200563311576843, "global_step": 117185, "epoch": 1316, "lr": 6.080671403767187e-05}
{"train_loss": 0.1794847995042801, "global_step": 117186, "epoch": 1316, "lr": 6.080614800320068e-05}
{"train_loss": 0.12728606164455414, "global_step": 117187, "epoch": 1316, "lr": 6.08055819672767e-05}
{"train_loss": 0.11289943754673004, "global_step": 117188, "epoch": 1316, "lr": 6.080501592990008e-05}
{"train_loss": 0.2009381502866745, "global_step": 117189, "epoch": 1316, "lr": 6.080444989107084e-05}
{"train_loss": 0.1976214200258255, "global_step": 117190, "epoch": 1316, "lr": 6.080388385078908e-05}
{"train_loss": 0.20450520515441895, "global_step": 117191, "epoch": 1316, "lr": 6.080331780905486e-05}
{"train_loss": 0.13176047801971436, "global_step": 117192, "epoch": 1316, "lr": 6.080275176586827e-05}
{"train_loss": 0.19553326070308685, "global_step": 117193, "epoch": 1316, "lr": 6.0802185721229375e-05}
{"train_loss": 0.1500083953142166, "global_step": 117194, "epoch": 1316, "lr": 6.0801619675138286e-05}
{"train_loss": 0.12839969992637634, "global_step": 117195, "epoch": 1316, "lr": 6.080105362759503e-05}
{"train_loss": 0.14476151764392853, "global_step": 117196, "epoch": 1316, "lr": 6.080048757859971e-05}
{"train_loss": 0.14671590924263, "global_step": 117197, "epoch": 1316, "lr": 6.079992152815238e-05}
{"train_loss": 0.12041524797677994, "global_step": 117198, "epoch": 1316, "lr": 6.079935547625315e-05}
{"train_loss": 0.13919684290885925, "global_step": 117199, "epoch": 1316, "lr": 6.079878942290207e-05}
{"train_loss": 0.1724538505077362, "global_step": 117200, "epoch": 1316, "lr": 6.079822336809923e-05}
{"train_loss": 0.11894021183252335, "global_step": 117201, "epoch": 1316, "lr": 6.079765731184469e-05}
{"train_loss": 0.11457249522209167, "global_step": 117202, "epoch": 1316, "lr": 6.079709125413854e-05}
{"train_loss": 0.2074246108531952, "global_step": 117203, "epoch": 1316, "lr": 6.079652519498086e-05}
{"train_loss": 0.12271832674741745, "global_step": 117204, "epoch": 1316, "lr": 6.0795959134371714e-05}
{"train_loss": 0.13810506463050842, "global_step": 117205, "epoch": 1316, "lr": 6.079539307231117e-05}
{"train_loss": 0.11922325193881989, "global_step": 117206, "epoch": 1316, "lr": 6.079482700879933e-05}
{"train_loss": 0.12943165004253387, "global_step": 117207, "epoch": 1316, "lr": 6.0794260943836256e-05}
{"train_loss": 0.10364046692848206, "global_step": 117208, "epoch": 1316, "lr": 6.079369487742203e-05}
{"train_loss": 0.08370272815227509, "global_step": 117209, "epoch": 1316, "lr": 6.079312880955672e-05}
{"train_loss": 0.1491422802209854, "global_step": 117210, "epoch": 1316, "lr": 6.07925627402404e-05}
{"train_loss": 0.21817196905612946, "global_step": 117211, "epoch": 1316, "lr": 6.079199666947316e-05}
{"train_loss": 0.14681589151366373, "global_step": 117212, "epoch": 1316, "lr": 6.079143059725506e-05, "val_loss": 5.1311163902282715}
{"train_loss": 0.13448208570480347, "global_step": 117213, "epoch": 1317, "lr": 6.07908645235862e-05}
{"train_loss": 0.1571575403213501, "global_step": 117214, "epoch": 1317, "lr": 6.079029844846663e-05}
{"train_loss": 0.1409316062927246, "global_step": 117215, "epoch": 1317, "lr": 6.0789732371896435e-05}
{"train_loss": 0.12448839098215103, "global_step": 117216, "epoch": 1317, "lr": 6.078916629387569e-05}
{"train_loss": 0.13848038017749786, "global_step": 117217, "epoch": 1317, "lr": 6.0788600214404465e-05}
{"train_loss": 0.08171871304512024, "global_step": 117218, "epoch": 1317, "lr": 6.0788034133482866e-05}
{"train_loss": 0.10479670017957687, "global_step": 117219, "epoch": 1317, "lr": 6.078746805111093e-05}
{"train_loss": 0.1436101347208023, "global_step": 117220, "epoch": 1317, "lr": 6.0786901967288744e-05}
{"train_loss": 0.09753338247537613, "global_step": 117221, "epoch": 1317, "lr": 6.078633588201641e-05}
{"train_loss": 0.10473974049091339, "global_step": 117222, "epoch": 1317, "lr": 6.0785769795293964e-05}
{"train_loss": 0.13880953192710876, "global_step": 117223, "epoch": 1317, "lr": 6.078520370712152e-05}
{"train_loss": 0.1757657825946808, "global_step": 117224, "epoch": 1317, "lr": 6.0784637617499116e-05}
{"train_loss": 0.15640854835510254, "global_step": 117225, "epoch": 1317, "lr": 6.0784071526426864e-05}
{"train_loss": 0.16160207986831665, "global_step": 117226, "epoch": 1317, "lr": 6.078350543390483e-05}
{"train_loss": 0.10456126183271408, "global_step": 117227, "epoch": 1317, "lr": 6.0782939339933075e-05}
{"train_loss": 0.13232706487178802, "global_step": 117228, "epoch": 1317, "lr": 6.0782373244511695e-05}
{"train_loss": 0.12256909906864166, "global_step": 117229, "epoch": 1317, "lr": 6.078180714764076e-05}
{"train_loss": 0.10075508058071136, "global_step": 117230, "epoch": 1317, "lr": 6.078124104932033e-05}
{"train_loss": 0.07140105217695236, "global_step": 117231, "epoch": 1317, "lr": 6.07806749495505e-05}
{"train_loss": 0.15495911240577698, "global_step": 117232, "epoch": 1317, "lr": 6.0780108848331326e-05}
{"train_loss": 0.18944653868675232, "global_step": 117233, "epoch": 1317, "lr": 6.077954274566292e-05}
{"train_loss": 0.15553154051303864, "global_step": 117234, "epoch": 1317, "lr": 6.0778976641545326e-05}
{"train_loss": 0.13015961647033691, "global_step": 117235, "epoch": 1317, "lr": 6.077841053597864e-05}
{"train_loss": 0.18904203176498413, "global_step": 117236, "epoch": 1317, "lr": 6.07778444289629e-05}
{"train_loss": 0.14183896780014038, "global_step": 117237, "epoch": 1317, "lr": 6.077727832049823e-05}
{"train_loss": 0.15116865932941437, "global_step": 117238, "epoch": 1317, "lr": 6.0776712210584686e-05}
{"train_loss": 0.16669678688049316, "global_step": 117239, "epoch": 1317, "lr": 6.077614609922234e-05}
{"train_loss": 0.13770227134227753, "global_step": 117240, "epoch": 1317, "lr": 6.077557998641128e-05}
{"train_loss": 0.12588363885879517, "global_step": 117241, "epoch": 1317, "lr": 6.077501387215156e-05}
{"train_loss": 0.14904868602752686, "global_step": 117242, "epoch": 1317, "lr": 6.0774447756443276e-05}
{"train_loss": 0.1848205327987671, "global_step": 117243, "epoch": 1317, "lr": 6.07738816392865e-05}
{"train_loss": 0.14221012592315674, "global_step": 117244, "epoch": 1317, "lr": 6.077331552068131e-05}
{"train_loss": 0.20155714452266693, "global_step": 117245, "epoch": 1317, "lr": 6.077274940062778e-05}
{"train_loss": 0.158938467502594, "global_step": 117246, "epoch": 1317, "lr": 6.077218327912598e-05}
{"train_loss": 0.2347198873758316, "global_step": 117247, "epoch": 1317, "lr": 6.0771617156175995e-05}
{"train_loss": 0.10268555581569672, "global_step": 117248, "epoch": 1317, "lr": 6.07710510317779e-05}
{"train_loss": 0.09874178469181061, "global_step": 117249, "epoch": 1317, "lr": 6.077048490593177e-05}
{"train_loss": 0.11313723772764206, "global_step": 117250, "epoch": 1317, "lr": 6.076991877863767e-05}
{"train_loss": 0.07233265042304993, "global_step": 117251, "epoch": 1317, "lr": 6.0769352649895695e-05}
{"train_loss": 0.23716259002685547, "global_step": 117252, "epoch": 1317, "lr": 6.076878651970591e-05}
{"train_loss": 0.22722823917865753, "global_step": 117253, "epoch": 1317, "lr": 6.0768220388068396e-05}
{"train_loss": 0.2064175009727478, "global_step": 117254, "epoch": 1317, "lr": 6.076765425498322e-05}
{"train_loss": 0.20987387001514435, "global_step": 117255, "epoch": 1317, "lr": 6.076708812045047e-05}
{"train_loss": 0.10802658647298813, "global_step": 117256, "epoch": 1317, "lr": 6.07665219844702e-05}
{"train_loss": 0.11283241212368011, "global_step": 117257, "epoch": 1317, "lr": 6.076595584704252e-05}
{"train_loss": 0.15189388394355774, "global_step": 117258, "epoch": 1317, "lr": 6.0765389708167475e-05}
{"train_loss": 0.15928387641906738, "global_step": 117259, "epoch": 1317, "lr": 6.076482356784516e-05}
{"train_loss": 0.21480347216129303, "global_step": 117260, "epoch": 1317, "lr": 6.0764257426075644e-05}
{"train_loss": 0.1073552742600441, "global_step": 117261, "epoch": 1317, "lr": 6.0763691282859e-05}
{"train_loss": 0.14834094047546387, "global_step": 117262, "epoch": 1317, "lr": 6.076312513819532e-05}
{"train_loss": 0.17862726747989655, "global_step": 117263, "epoch": 1317, "lr": 6.0762558992084675e-05}
{"train_loss": 0.11968978494405746, "global_step": 117264, "epoch": 1317, "lr": 6.0761992844527116e-05}
{"train_loss": 0.22000020742416382, "global_step": 117265, "epoch": 1317, "lr": 6.076142669552275e-05}
{"train_loss": 0.141087144613266, "global_step": 117266, "epoch": 1317, "lr": 6.076086054507164e-05}
{"train_loss": 0.22200386226177216, "global_step": 117267, "epoch": 1317, "lr": 6.076029439317387e-05}
{"train_loss": 0.09620092809200287, "global_step": 117268, "epoch": 1317, "lr": 6.0759728239829495e-05}
{"train_loss": 0.22383172810077667, "global_step": 117269, "epoch": 1317, "lr": 6.075916208503861e-05}
{"train_loss": 0.16217820346355438, "global_step": 117270, "epoch": 1317, "lr": 6.07585959288013e-05}
{"train_loss": 0.10184624046087265, "global_step": 117271, "epoch": 1317, "lr": 6.075802977111762e-05}
{"train_loss": 0.12377402186393738, "global_step": 117272, "epoch": 1317, "lr": 6.0757463611987655e-05}
{"train_loss": 0.18094338476657867, "global_step": 117273, "epoch": 1317, "lr": 6.075689745141148e-05}
{"train_loss": 0.20963634550571442, "global_step": 117274, "epoch": 1317, "lr": 6.075633128938917e-05}
{"train_loss": 0.10818290710449219, "global_step": 117275, "epoch": 1317, "lr": 6.0755765125920795e-05}
{"train_loss": 0.1494469791650772, "global_step": 117276, "epoch": 1317, "lr": 6.0755198961006447e-05}
{"train_loss": 0.13759514689445496, "global_step": 117277, "epoch": 1317, "lr": 6.075463279464619e-05}
{"train_loss": 0.15211142599582672, "global_step": 117278, "epoch": 1317, "lr": 6.075406662684009e-05}
{"train_loss": 0.0837215930223465, "global_step": 117279, "epoch": 1317, "lr": 6.075350045758825e-05}
{"train_loss": 0.13243690133094788, "global_step": 117280, "epoch": 1317, "lr": 6.075293428689074e-05}
{"train_loss": 0.12433953583240509, "global_step": 117281, "epoch": 1317, "lr": 6.075236811474762e-05}
{"train_loss": 0.1208006739616394, "global_step": 117282, "epoch": 1317, "lr": 6.075180194115898e-05}
{"train_loss": 0.13851721584796906, "global_step": 117283, "epoch": 1317, "lr": 6.075123576612489e-05}
{"train_loss": 0.19586297869682312, "global_step": 117284, "epoch": 1317, "lr": 6.075066958964543e-05}
{"train_loss": 0.1276690810918808, "global_step": 117285, "epoch": 1317, "lr": 6.075010341172066e-05}
{"train_loss": 0.15139353275299072, "global_step": 117286, "epoch": 1317, "lr": 6.0749537232350685e-05}
{"train_loss": 0.15902887284755707, "global_step": 117287, "epoch": 1317, "lr": 6.074897105153556e-05}
{"train_loss": 0.1477811634540558, "global_step": 117288, "epoch": 1317, "lr": 6.074840486927537e-05}
{"train_loss": 0.140069380402565, "global_step": 117289, "epoch": 1317, "lr": 6.074783868557018e-05}
{"train_loss": 0.14404968917369843, "global_step": 117290, "epoch": 1317, "lr": 6.074727250042008e-05}
{"train_loss": 0.11263579875230789, "global_step": 117291, "epoch": 1317, "lr": 6.0746706313825153e-05}
{"train_loss": 0.11549251526594162, "global_step": 117292, "epoch": 1317, "lr": 6.0746140125785443e-05}
{"train_loss": 0.08750269562005997, "global_step": 117293, "epoch": 1317, "lr": 6.074557393630106e-05}
{"train_loss": 0.2146313488483429, "global_step": 117294, "epoch": 1317, "lr": 6.0745007745372054e-05}
{"train_loss": 0.1471901386976242, "global_step": 117295, "epoch": 1317, "lr": 6.0744441552998524e-05}
{"train_loss": 0.13553114235401154, "global_step": 117296, "epoch": 1317, "lr": 6.074387535918051e-05}
{"train_loss": 0.15128572285175323, "global_step": 117297, "epoch": 1317, "lr": 6.074330916391813e-05}
{"train_loss": 0.2789883613586426, "global_step": 117298, "epoch": 1317, "lr": 6.0742742967211444e-05}
{"train_loss": 0.21053948998451233, "global_step": 117299, "epoch": 1317, "lr": 6.074217676906052e-05}
{"train_loss": 0.19662998616695404, "global_step": 117300, "epoch": 1317, "lr": 6.074161056946544e-05}
{"train_loss": 0.1487097197704101, "global_step": 117301, "epoch": 1317, "lr": 6.074104436842629e-05, "val_loss": 5.183917045593262}
{"train_loss": 0.12298235297203064, "global_step": 117302, "epoch": 1318, "lr": 6.074047816594315e-05}
{"train_loss": 0.1321447342634201, "global_step": 117303, "epoch": 1318, "lr": 6.073991196201606e-05}
{"train_loss": 0.10272455215454102, "global_step": 117304, "epoch": 1318, "lr": 6.0739345756645126e-05}
{"train_loss": 0.11336595565080643, "global_step": 117305, "epoch": 1318, "lr": 6.073877954983043e-05}
{"train_loss": 0.17590393126010895, "global_step": 117306, "epoch": 1318, "lr": 6.073821334157202e-05}
{"train_loss": 0.1344669759273529, "global_step": 117307, "epoch": 1318, "lr": 6.073764713187e-05}
{"train_loss": 0.1320161372423172, "global_step": 117308, "epoch": 1318, "lr": 6.073708092072443e-05}
{"train_loss": 0.1271829605102539, "global_step": 117309, "epoch": 1318, "lr": 6.073651470813538e-05}
{"train_loss": 0.14293888211250305, "global_step": 117310, "epoch": 1318, "lr": 6.073594849410297e-05}
{"train_loss": 0.12083294242620468, "global_step": 117311, "epoch": 1318, "lr": 6.073538227862722e-05}
{"train_loss": 0.13563889265060425, "global_step": 117312, "epoch": 1318, "lr": 6.0734816061708224e-05}
{"train_loss": 0.09841251373291016, "global_step": 117313, "epoch": 1318, "lr": 6.0734249843346056e-05}
{"train_loss": 0.15901236236095428, "global_step": 117314, "epoch": 1318, "lr": 6.07336836235408e-05}
{"train_loss": 0.165372833609581, "global_step": 117315, "epoch": 1318, "lr": 6.073311740229255e-05}
{"train_loss": 0.1517876535654068, "global_step": 117316, "epoch": 1318, "lr": 6.073255117960135e-05}
{"train_loss": 0.1111777126789093, "global_step": 117317, "epoch": 1318, "lr": 6.073198495546729e-05}
{"train_loss": 0.12937308847904205, "global_step": 117318, "epoch": 1318, "lr": 6.0731418729890454e-05}
{"train_loss": 0.14644242823123932, "global_step": 117319, "epoch": 1318, "lr": 6.073085250287089e-05}
{"train_loss": 0.1047673299908638, "global_step": 117320, "epoch": 1318, "lr": 6.073028627440872e-05}
{"train_loss": 0.14553216099739075, "global_step": 117321, "epoch": 1318, "lr": 6.072972004450398e-05}
{"train_loss": 0.1354985535144806, "global_step": 117322, "epoch": 1318, "lr": 6.072915381315676e-05}
{"train_loss": 0.10917019098997116, "global_step": 117323, "epoch": 1318, "lr": 6.072858758036713e-05}
{"train_loss": 0.14547033607959747, "global_step": 117324, "epoch": 1318, "lr": 6.072802134613519e-05}
{"train_loss": 0.22704532742500305, "global_step": 117325, "epoch": 1318, "lr": 6.0727455110460985e-05}
{"train_loss": 0.18963827192783356, "global_step": 117326, "epoch": 1318, "lr": 6.07268888733446e-05}
{"train_loss": 0.17450688779354095, "global_step": 117327, "epoch": 1318, "lr": 6.072632263478612e-05}
{"train_loss": 0.12902776896953583, "global_step": 117328, "epoch": 1318, "lr": 6.0725756394785624e-05}
{"train_loss": 0.19275203347206116, "global_step": 117329, "epoch": 1318, "lr": 6.072519015334319e-05}
{"train_loss": 0.12596294283866882, "global_step": 117330, "epoch": 1318, "lr": 6.072462391045886e-05}
{"train_loss": 0.20127524435520172, "global_step": 117331, "epoch": 1318, "lr": 6.0724057666132764e-05}
{"train_loss": 0.14073610305786133, "global_step": 117332, "epoch": 1318, "lr": 6.0723491420364906e-05}
{"train_loss": 0.0804286003112793, "global_step": 117333, "epoch": 1318, "lr": 6.072292517315543e-05}
{"train_loss": 0.14352941513061523, "global_step": 117334, "epoch": 1318, "lr": 6.0722358924504395e-05}
{"train_loss": 0.16385184228420258, "global_step": 117335, "epoch": 1318, "lr": 6.072179267441186e-05}
{"train_loss": 0.14559867978096008, "global_step": 117336, "epoch": 1318, "lr": 6.072122642287792e-05}
{"train_loss": 0.08652856945991516, "global_step": 117337, "epoch": 1318, "lr": 6.072066016990263e-05}
{"train_loss": 0.12592951953411102, "global_step": 117338, "epoch": 1318, "lr": 6.072009391548607e-05}
{"train_loss": 0.09931755065917969, "global_step": 117339, "epoch": 1318, "lr": 6.071952765962835e-05}
{"train_loss": 0.14805810153484344, "global_step": 117340, "epoch": 1318, "lr": 6.071896140232951e-05}
{"train_loss": 0.09307830780744553, "global_step": 117341, "epoch": 1318, "lr": 6.0718395143589626e-05}
{"train_loss": 0.17402265965938568, "global_step": 117342, "epoch": 1318, "lr": 6.071782888340879e-05}
{"train_loss": 0.12099429219961166, "global_step": 117343, "epoch": 1318, "lr": 6.071726262178707e-05}
{"train_loss": 0.17018906772136688, "global_step": 117344, "epoch": 1318, "lr": 6.0716696358724555e-05}
{"train_loss": 0.17833290994167328, "global_step": 117345, "epoch": 1318, "lr": 6.0716130094221304e-05}
{"train_loss": 0.19241958856582642, "global_step": 117346, "epoch": 1318, "lr": 6.07155638282774e-05}
{"train_loss": 0.2043195366859436, "global_step": 117347, "epoch": 1318, "lr": 6.0714997560892914e-05}
{"train_loss": 0.12598365545272827, "global_step": 117348, "epoch": 1318, "lr": 6.071443129206794e-05}
{"train_loss": 0.13267499208450317, "global_step": 117349, "epoch": 1318, "lr": 6.071386502180254e-05}
{"train_loss": 0.1789979487657547, "global_step": 117350, "epoch": 1318, "lr": 6.071329875009679e-05}
{"train_loss": 0.19281704723834991, "global_step": 117351, "epoch": 1318, "lr": 6.0712732476950754e-05}
{"train_loss": 0.11795584112405777, "global_step": 117352, "epoch": 1318, "lr": 6.071216620236453e-05}
{"train_loss": 0.12730994820594788, "global_step": 117353, "epoch": 1318, "lr": 6.071159992633818e-05}
{"train_loss": 0.1744045466184616, "global_step": 117354, "epoch": 1318, "lr": 6.07110336488718e-05}
{"train_loss": 0.2390178143978119, "global_step": 117355, "epoch": 1318, "lr": 6.071046736996544e-05}
{"train_loss": 0.13774831593036652, "global_step": 117356, "epoch": 1318, "lr": 6.070990108961919e-05}
{"train_loss": 0.07983105629682541, "global_step": 117357, "epoch": 1318, "lr": 6.070933480783313e-05}
{"train_loss": 0.1306244432926178, "global_step": 117358, "epoch": 1318, "lr": 6.070876852460733e-05}
{"train_loss": 0.22119417786598206, "global_step": 117359, "epoch": 1318, "lr": 6.070820223994186e-05}
{"train_loss": 0.16931599378585815, "global_step": 117360, "epoch": 1318, "lr": 6.0707635953836806e-05}
{"train_loss": 0.16790860891342163, "global_step": 117361, "epoch": 1318, "lr": 6.0707069666292246e-05}
{"train_loss": 0.11875135451555252, "global_step": 117362, "epoch": 1318, "lr": 6.0706503377308245e-05}
{"train_loss": 0.16114217042922974, "global_step": 117363, "epoch": 1318, "lr": 6.070593708688489e-05}
{"train_loss": 0.1486520916223526, "global_step": 117364, "epoch": 1318, "lr": 6.0705370795022246e-05}
{"train_loss": 0.10338915884494781, "global_step": 117365, "epoch": 1318, "lr": 6.07048045017204e-05}
{"train_loss": 0.18027706444263458, "global_step": 117366, "epoch": 1318, "lr": 6.0704238206979424e-05}
{"train_loss": 0.08276568353176117, "global_step": 117367, "epoch": 1318, "lr": 6.0703671910799383e-05}
{"train_loss": 0.07658848166465759, "global_step": 117368, "epoch": 1318, "lr": 6.070310561318039e-05}
{"train_loss": 0.10524974763393402, "global_step": 117369, "epoch": 1318, "lr": 6.0702539314122464e-05}
{"train_loss": 0.13835112750530243, "global_step": 117370, "epoch": 1318, "lr": 6.0701973013625726e-05}
{"train_loss": 0.11975561827421188, "global_step": 117371, "epoch": 1318, "lr": 6.070140671169023e-05}
{"train_loss": 0.14694152772426605, "global_step": 117372, "epoch": 1318, "lr": 6.070084040831607e-05}
{"train_loss": 0.11520587652921677, "global_step": 117373, "epoch": 1318, "lr": 6.070027410350332e-05}
{"train_loss": 0.11215675622224808, "global_step": 117374, "epoch": 1318, "lr": 6.069970779725204e-05}
{"train_loss": 0.14515024423599243, "global_step": 117375, "epoch": 1318, "lr": 6.0699141489562304e-05}
{"train_loss": 0.24593108892440796, "global_step": 117376, "epoch": 1318, "lr": 6.069857518043421e-05}
{"train_loss": 0.07570639252662659, "global_step": 117377, "epoch": 1318, "lr": 6.069800886986782e-05}
{"train_loss": 0.1636289656162262, "global_step": 117378, "epoch": 1318, "lr": 6.069744255786322e-05}
{"train_loss": 0.1547376960515976, "global_step": 117379, "epoch": 1318, "lr": 6.0696876244420463e-05}
{"train_loss": 0.19114072620868683, "global_step": 117380, "epoch": 1318, "lr": 6.0696309929539663e-05}
{"train_loss": 0.15921033918857574, "global_step": 117381, "epoch": 1318, "lr": 6.069574361322086e-05}
{"train_loss": 0.18354305624961853, "global_step": 117382, "epoch": 1318, "lr": 6.0695177295464145e-05}
{"train_loss": 0.11702224612236023, "global_step": 117383, "epoch": 1318, "lr": 6.0694610976269596e-05}
{"train_loss": 0.14135906100273132, "global_step": 117384, "epoch": 1318, "lr": 6.069404465563729e-05}
{"train_loss": 0.12380937486886978, "global_step": 117385, "epoch": 1318, "lr": 6.06934783335673e-05}
{"train_loss": 0.15012632310390472, "global_step": 117386, "epoch": 1318, "lr": 6.06929120100597e-05}
{"train_loss": 0.08892292529344559, "global_step": 117387, "epoch": 1318, "lr": 6.0692345685114574e-05}
{"train_loss": 0.13696885108947754, "global_step": 117388, "epoch": 1318, "lr": 6.069177935873198e-05}
{"train_loss": 0.12482266873121262, "global_step": 117389, "epoch": 1318, "lr": 6.069121303091201e-05}
{"train_loss": 0.1430248828249031, "global_step": 117390, "epoch": 1318, "lr": 6.069064670165474e-05, "val_loss": 5.197277545928955}
{"train_loss": 0.099138043820858, "global_step": 117391, "epoch": 1319, "lr": 6.069008037096024e-05}
{"train_loss": 0.11948823183774948, "global_step": 117392, "epoch": 1319, "lr": 6.06895140388286e-05}
{"train_loss": 0.10273223370313644, "global_step": 117393, "epoch": 1319, "lr": 6.068894770525987e-05}
{"train_loss": 0.16643661260604858, "global_step": 117394, "epoch": 1319, "lr": 6.0688381370254155e-05}
{"train_loss": 0.09800957143306732, "global_step": 117395, "epoch": 1319, "lr": 6.0687815033811515e-05}
{"train_loss": 0.1142735630273819, "global_step": 117396, "epoch": 1319, "lr": 6.068724869593202e-05}
{"train_loss": 0.13369089365005493, "global_step": 117397, "epoch": 1319, "lr": 6.068668235661575e-05}
{"train_loss": 0.13151949644088745, "global_step": 117398, "epoch": 1319, "lr": 6.0686116015862805e-05}
{"train_loss": 0.12816549837589264, "global_step": 117399, "epoch": 1319, "lr": 6.068554967367322e-05}
{"train_loss": 0.09883187711238861, "global_step": 117400, "epoch": 1319, "lr": 6.0684983330047106e-05}
{"train_loss": 0.17798258364200592, "global_step": 117401, "epoch": 1319, "lr": 6.068441698498453e-05}
{"train_loss": 0.13780245184898376, "global_step": 117402, "epoch": 1319, "lr": 6.068385063848556e-05}
{"train_loss": 0.15112978219985962, "global_step": 117403, "epoch": 1319, "lr": 6.068328429055028e-05}
{"train_loss": 0.11669089645147324, "global_step": 117404, "epoch": 1319, "lr": 6.068271794117875e-05}
{"train_loss": 0.125472292304039, "global_step": 117405, "epoch": 1319, "lr": 6.0682151590371074e-05}
{"train_loss": 0.118382029235363, "global_step": 117406, "epoch": 1319, "lr": 6.0681585238127305e-05}
{"train_loss": 0.13304632902145386, "global_step": 117407, "epoch": 1319, "lr": 6.0681018884447524e-05}
{"train_loss": 0.1668950319290161, "global_step": 117408, "epoch": 1319, "lr": 6.068045252933181e-05}
{"train_loss": 0.09764839708805084, "global_step": 117409, "epoch": 1319, "lr": 6.0679886172780244e-05}
{"train_loss": 0.1373600959777832, "global_step": 117410, "epoch": 1319, "lr": 6.06793198147929e-05}
{"train_loss": 0.17497870326042175, "global_step": 117411, "epoch": 1319, "lr": 6.067875345536985e-05}
{"train_loss": 0.2302410900592804, "global_step": 117412, "epoch": 1319, "lr": 6.067818709451117e-05}
{"train_loss": 0.1606818586587906, "global_step": 117413, "epoch": 1319, "lr": 6.0677620732216945e-05}
{"train_loss": 0.13088618218898773, "global_step": 117414, "epoch": 1319, "lr": 6.0677054368487226e-05}
{"train_loss": 0.21569788455963135, "global_step": 117415, "epoch": 1319, "lr": 6.067648800332212e-05}
{"train_loss": 0.19280794262886047, "global_step": 117416, "epoch": 1319, "lr": 6.067592163672169e-05}
{"train_loss": 0.09053721278905869, "global_step": 117417, "epoch": 1319, "lr": 6.0675355268686015e-05}
{"train_loss": 0.09894496947526932, "global_step": 117418, "epoch": 1319, "lr": 6.067478889921516e-05}
{"train_loss": 0.0812888815999031, "global_step": 117419, "epoch": 1319, "lr": 6.0674222528309224e-05}
{"train_loss": 0.17152146995067596, "global_step": 117420, "epoch": 1319, "lr": 6.067365615596825e-05}
{"train_loss": 0.15878264605998993, "global_step": 117421, "epoch": 1319, "lr": 6.067308978219235e-05}
{"train_loss": 0.16662684082984924, "global_step": 117422, "epoch": 1319, "lr": 6.067252340698157e-05}
{"train_loss": 0.11486697196960449, "global_step": 117423, "epoch": 1319, "lr": 6.0671957030336004e-05}
{"train_loss": 0.10601670295000076, "global_step": 117424, "epoch": 1319, "lr": 6.067139065225573e-05}
{"train_loss": 0.20527881383895874, "global_step": 117425, "epoch": 1319, "lr": 6.067082427274081e-05}
{"train_loss": 0.14060497283935547, "global_step": 117426, "epoch": 1319, "lr": 6.0670257891791323e-05}
{"train_loss": 0.14552119374275208, "global_step": 117427, "epoch": 1319, "lr": 6.0669691509407364e-05}
{"train_loss": 0.10793965309858322, "global_step": 117428, "epoch": 1319, "lr": 6.0669125125588986e-05}
{"train_loss": 0.20210862159729004, "global_step": 117429, "epoch": 1319, "lr": 6.066855874033628e-05}
{"train_loss": 0.17000608146190643, "global_step": 117430, "epoch": 1319, "lr": 6.066799235364931e-05}
{"train_loss": 0.14141525328159332, "global_step": 117431, "epoch": 1319, "lr": 6.066742596552816e-05}
{"train_loss": 0.08234698325395584, "global_step": 117432, "epoch": 1319, "lr": 6.06668595759729e-05}
{"train_loss": 0.17321838438510895, "global_step": 117433, "epoch": 1319, "lr": 6.066629318498363e-05}
{"train_loss": 0.10432717204093933, "global_step": 117434, "epoch": 1319, "lr": 6.066572679256038e-05}
{"train_loss": 0.18890154361724854, "global_step": 117435, "epoch": 1319, "lr": 6.066516039870327e-05}
{"train_loss": 0.18055687844753265, "global_step": 117436, "epoch": 1319, "lr": 6.0664594003412356e-05}
{"train_loss": 0.11003924161195755, "global_step": 117437, "epoch": 1319, "lr": 6.0664027606687716e-05}
{"train_loss": 0.19256716966629028, "global_step": 117438, "epoch": 1319, "lr": 6.066346120852943e-05}
{"train_loss": 0.16318951547145844, "global_step": 117439, "epoch": 1319, "lr": 6.066289480893757e-05}
{"train_loss": 0.19904902577400208, "global_step": 117440, "epoch": 1319, "lr": 6.06623284079122e-05}
{"train_loss": 0.11041883379220963, "global_step": 117441, "epoch": 1319, "lr": 6.066176200545344e-05}
{"train_loss": 0.16981282830238342, "global_step": 117442, "epoch": 1319, "lr": 6.0661195601561316e-05}
{"train_loss": 0.1443970948457718, "global_step": 117443, "epoch": 1319, "lr": 6.066062919623592e-05}
{"train_loss": 0.13590343296527863, "global_step": 117444, "epoch": 1319, "lr": 6.0660062789477345e-05}
{"train_loss": 0.1324613094329834, "global_step": 117445, "epoch": 1319, "lr": 6.065949638128564e-05}
{"train_loss": 0.125529482960701, "global_step": 117446, "epoch": 1319, "lr": 6.065892997166092e-05}
{"train_loss": 0.17363736033439636, "global_step": 117447, "epoch": 1319, "lr": 6.0658363560603214e-05}
{"train_loss": 0.12328571081161499, "global_step": 117448, "epoch": 1319, "lr": 6.065779714811263e-05}
{"train_loss": 0.10921324789524078, "global_step": 117449, "epoch": 1319, "lr": 6.065723073418923e-05}
{"train_loss": 0.17890320718288422, "global_step": 117450, "epoch": 1319, "lr": 6.06566643188331e-05}
{"train_loss": 0.15154345333576202, "global_step": 117451, "epoch": 1319, "lr": 6.0656097902044304e-05}
{"train_loss": 0.18759696185588837, "global_step": 117452, "epoch": 1319, "lr": 6.0655531483822936e-05}
{"train_loss": 0.11558282375335693, "global_step": 117453, "epoch": 1319, "lr": 6.065496506416905e-05}
{"train_loss": 0.13723213970661163, "global_step": 117454, "epoch": 1319, "lr": 6.0654398643082756e-05}
{"train_loss": 0.16632500290870667, "global_step": 117455, "epoch": 1319, "lr": 6.0653832220564086e-05}
{"train_loss": 0.132902592420578, "global_step": 117456, "epoch": 1319, "lr": 6.0653265796613143e-05}
{"train_loss": 0.16392894089221954, "global_step": 117457, "epoch": 1319, "lr": 6.0652699371230004e-05}
{"train_loss": 0.1492433249950409, "global_step": 117458, "epoch": 1319, "lr": 6.065213294441474e-05}
{"train_loss": 0.1253224015235901, "global_step": 117459, "epoch": 1319, "lr": 6.0651566516167405e-05}
{"train_loss": 0.16610941290855408, "global_step": 117460, "epoch": 1319, "lr": 6.065100008648813e-05}
{"train_loss": 0.20477250218391418, "global_step": 117461, "epoch": 1319, "lr": 6.065043365537694e-05}
{"train_loss": 0.16294614970684052, "global_step": 117462, "epoch": 1319, "lr": 6.064986722283393e-05}
{"train_loss": 0.17730645835399628, "global_step": 117463, "epoch": 1319, "lr": 6.0649300788859174e-05}
{"train_loss": 0.08978641033172607, "global_step": 117464, "epoch": 1319, "lr": 6.0648734353452755e-05}
{"train_loss": 0.12056705355644226, "global_step": 117465, "epoch": 1319, "lr": 6.064816791661474e-05}
{"train_loss": 0.09799753129482269, "global_step": 117466, "epoch": 1319, "lr": 6.064760147834521e-05}
{"train_loss": 0.17354780435562134, "global_step": 117467, "epoch": 1319, "lr": 6.064703503864424e-05}
{"train_loss": 0.1386643946170807, "global_step": 117468, "epoch": 1319, "lr": 6.0646468597511897e-05}
{"train_loss": 0.1244715005159378, "global_step": 117469, "epoch": 1319, "lr": 6.064590215494827e-05}
{"train_loss": 0.10743071883916855, "global_step": 117470, "epoch": 1319, "lr": 6.064533571095344e-05}
{"train_loss": 0.11477858573198318, "global_step": 117471, "epoch": 1319, "lr": 6.064476926552747e-05}
{"train_loss": 0.15402312576770782, "global_step": 117472, "epoch": 1319, "lr": 6.064420281867044e-05}
{"train_loss": 0.17059184610843658, "global_step": 117473, "epoch": 1319, "lr": 6.064363637038243e-05}
{"train_loss": 0.1375378966331482, "global_step": 117474, "epoch": 1319, "lr": 6.0643069920663506e-05}
{"train_loss": 0.10561982542276382, "global_step": 117475, "epoch": 1319, "lr": 6.0642503469513746e-05}
{"train_loss": 0.13180634379386902, "global_step": 117476, "epoch": 1319, "lr": 6.0641937016933247e-05}
{"train_loss": 0.16580873727798462, "global_step": 117477, "epoch": 1319, "lr": 6.064137056292206e-05}
{"train_loss": 0.14202839136123657, "global_step": 117478, "epoch": 1319, "lr": 6.0640804107480256e-05}
{"train_loss": 0.1429144073738141, "global_step": 117479, "epoch": 1319, "lr": 6.0640237650607955e-05, "val_loss": 5.093148708343506}
{"train_loss": 0.17322862148284912, "global_step": 117480, "epoch": 1320, "lr": 6.063967119230519e-05}
{"train_loss": 0.13841046392917633, "global_step": 117481, "epoch": 1320, "lr": 6.063910473257205e-05}
{"train_loss": 0.19276148080825806, "global_step": 117482, "epoch": 1320, "lr": 6.063853827140862e-05}
{"train_loss": 0.17058369517326355, "global_step": 117483, "epoch": 1320, "lr": 6.0637971808814955e-05}
{"train_loss": 0.11311789602041245, "global_step": 117484, "epoch": 1320, "lr": 6.0637405344791155e-05}
{"train_loss": 0.17171162366867065, "global_step": 117485, "epoch": 1320, "lr": 6.063683887933728e-05}
{"train_loss": 0.15729920566082, "global_step": 117486, "epoch": 1320, "lr": 6.063627241245341e-05}
{"train_loss": 0.11994096636772156, "global_step": 117487, "epoch": 1320, "lr": 6.063570594413962e-05}
{"train_loss": 0.10224300622940063, "global_step": 117488, "epoch": 1320, "lr": 6.0635139474396e-05}
{"train_loss": 0.18023096024990082, "global_step": 117489, "epoch": 1320, "lr": 6.0634573003222605e-05}
{"train_loss": 0.2208372801542282, "global_step": 117490, "epoch": 1320, "lr": 6.0634006530619526e-05}
{"train_loss": 0.1685788631439209, "global_step": 117491, "epoch": 1320, "lr": 6.063344005658683e-05}
{"train_loss": 0.13537822663784027, "global_step": 117492, "epoch": 1320, "lr": 6.0632873581124605e-05}
{"train_loss": 0.16703513264656067, "global_step": 117493, "epoch": 1320, "lr": 6.0632307104232925e-05}
{"train_loss": 0.14997513592243195, "global_step": 117494, "epoch": 1320, "lr": 6.0631740625911846e-05}
{"train_loss": 0.06767752766609192, "global_step": 117495, "epoch": 1320, "lr": 6.063117414616146e-05}
{"train_loss": 0.1470309942960739, "global_step": 117496, "epoch": 1320, "lr": 6.0630607664981844e-05}
{"train_loss": 0.17916014790534973, "global_step": 117497, "epoch": 1320, "lr": 6.063004118237308e-05}
{"train_loss": 0.12347579002380371, "global_step": 117498, "epoch": 1320, "lr": 6.0629474698335244e-05}
{"train_loss": 0.17527255415916443, "global_step": 117499, "epoch": 1320, "lr": 6.0628908212868385e-05}
{"train_loss": 0.21536479890346527, "global_step": 117500, "epoch": 1320, "lr": 6.0628341725972616e-05}
{"train_loss": 0.09041806310415268, "global_step": 117501, "epoch": 1320, "lr": 6.0627775237647976e-05}
{"train_loss": 0.1384415626525879, "global_step": 117502, "epoch": 1320, "lr": 6.062720874789458e-05}
{"train_loss": 0.19404840469360352, "global_step": 117503, "epoch": 1320, "lr": 6.062664225671247e-05}
{"train_loss": 0.12911461293697357, "global_step": 117504, "epoch": 1320, "lr": 6.062607576410175e-05}
{"train_loss": 0.17905235290527344, "global_step": 117505, "epoch": 1320, "lr": 6.062550927006248e-05}
{"train_loss": 0.13793426752090454, "global_step": 117506, "epoch": 1320, "lr": 6.062494277459474e-05}
{"train_loss": 0.11985859274864197, "global_step": 117507, "epoch": 1320, "lr": 6.06243762776986e-05}
{"train_loss": 0.17369812726974487, "global_step": 117508, "epoch": 1320, "lr": 6.062380977937415e-05}
{"train_loss": 0.09881674498319626, "global_step": 117509, "epoch": 1320, "lr": 6.0623243279621456e-05}
{"train_loss": 0.22364404797554016, "global_step": 117510, "epoch": 1320, "lr": 6.062267677844059e-05}
{"train_loss": 0.16941751539707184, "global_step": 117511, "epoch": 1320, "lr": 6.062211027583164e-05}
{"train_loss": 0.13834156095981598, "global_step": 117512, "epoch": 1320, "lr": 6.0621543771794676e-05}
{"train_loss": 0.08812697976827621, "global_step": 117513, "epoch": 1320, "lr": 6.062097726632978e-05}
{"train_loss": 0.11508723348379135, "global_step": 117514, "epoch": 1320, "lr": 6.062041075943701e-05}
{"train_loss": 0.09398922324180603, "global_step": 117515, "epoch": 1320, "lr": 6.061984425111646e-05}
{"train_loss": 0.1241738349199295, "global_step": 117516, "epoch": 1320, "lr": 6.061927774136821e-05}
{"train_loss": 0.11694128811359406, "global_step": 117517, "epoch": 1320, "lr": 6.0618711230192316e-05}
{"train_loss": 0.1729278415441513, "global_step": 117518, "epoch": 1320, "lr": 6.0618144717588886e-05}
{"train_loss": 0.17695210874080658, "global_step": 117519, "epoch": 1320, "lr": 6.0617578203557955e-05}
{"train_loss": 0.11076999455690384, "global_step": 117520, "epoch": 1320, "lr": 6.061701168809963e-05}
{"train_loss": 0.08635034412145615, "global_step": 117521, "epoch": 1320, "lr": 6.0616445171213976e-05}
{"train_loss": 0.08591251820325851, "global_step": 117522, "epoch": 1320, "lr": 6.061587865290107e-05}
{"train_loss": 0.11748558282852173, "global_step": 117523, "epoch": 1320, "lr": 6.061531213316099e-05}
{"train_loss": 0.13939522206783295, "global_step": 117524, "epoch": 1320, "lr": 6.0614745611993804e-05}
{"train_loss": 0.07294684648513794, "global_step": 117525, "epoch": 1320, "lr": 6.06141790893996e-05}
{"train_loss": 0.12927554547786713, "global_step": 117526, "epoch": 1320, "lr": 6.061361256537845e-05}
{"train_loss": 0.12105832993984222, "global_step": 117527, "epoch": 1320, "lr": 6.0613046039930424e-05}
{"train_loss": 0.13026902079582214, "global_step": 117528, "epoch": 1320, "lr": 6.061247951305561e-05}
{"train_loss": 0.17998266220092773, "global_step": 117529, "epoch": 1320, "lr": 6.061191298475407e-05}
{"train_loss": 0.09151006489992142, "global_step": 117530, "epoch": 1320, "lr": 6.061134645502589e-05}
{"train_loss": 0.15262171626091003, "global_step": 117531, "epoch": 1320, "lr": 6.061077992387114e-05}
{"train_loss": 0.12708961963653564, "global_step": 117532, "epoch": 1320, "lr": 6.061021339128989e-05}
{"train_loss": 0.15728934109210968, "global_step": 117533, "epoch": 1320, "lr": 6.060964685728225e-05}
{"train_loss": 0.13498112559318542, "global_step": 117534, "epoch": 1320, "lr": 6.060908032184827e-05}
{"train_loss": 0.1958097517490387, "global_step": 117535, "epoch": 1320, "lr": 6.060851378498802e-05}
{"train_loss": 0.1049620509147644, "global_step": 117536, "epoch": 1320, "lr": 6.060794724670158e-05}
{"train_loss": 0.16214090585708618, "global_step": 117537, "epoch": 1320, "lr": 6.060738070698905e-05}
{"train_loss": 0.09455712139606476, "global_step": 117538, "epoch": 1320, "lr": 6.0606814165850476e-05}
{"train_loss": 0.12051595002412796, "global_step": 117539, "epoch": 1320, "lr": 6.060624762328594e-05}
{"train_loss": 0.11365474760532379, "global_step": 117540, "epoch": 1320, "lr": 6.060568107929553e-05}
{"train_loss": 0.08294802159070969, "global_step": 117541, "epoch": 1320, "lr": 6.0605114533879315e-05}
{"train_loss": 0.13419575989246368, "global_step": 117542, "epoch": 1320, "lr": 6.0604547987037364e-05}
{"train_loss": 0.16221930086612701, "global_step": 117543, "epoch": 1320, "lr": 6.060398143876976e-05}
{"train_loss": 0.1653006374835968, "global_step": 117544, "epoch": 1320, "lr": 6.0603414889076595e-05}
{"train_loss": 0.13201728463172913, "global_step": 117545, "epoch": 1320, "lr": 6.060284833795792e-05}
{"train_loss": 0.13380463421344757, "global_step": 117546, "epoch": 1320, "lr": 6.0602281785413817e-05}
{"train_loss": 0.0939221903681755, "global_step": 117547, "epoch": 1320, "lr": 6.060171523144438e-05}
{"train_loss": 0.20059464871883392, "global_step": 117548, "epoch": 1320, "lr": 6.060114867604966e-05}
{"train_loss": 0.1655772179365158, "global_step": 117549, "epoch": 1320, "lr": 6.060058211922974e-05}
{"train_loss": 0.1806129664182663, "global_step": 117550, "epoch": 1320, "lr": 6.06000155609847e-05}
{"train_loss": 0.16701550781726837, "global_step": 117551, "epoch": 1320, "lr": 6.0599449001314624e-05}
{"train_loss": 0.09392539411783218, "global_step": 117552, "epoch": 1320, "lr": 6.059888244021958e-05}
{"train_loss": 0.11062246561050415, "global_step": 117553, "epoch": 1320, "lr": 6.059831587769965e-05}
{"train_loss": 0.16676700115203857, "global_step": 117554, "epoch": 1320, "lr": 6.059774931375491e-05}
{"train_loss": 0.10448088496923447, "global_step": 117555, "epoch": 1320, "lr": 6.059718274838543e-05}
{"train_loss": 0.1110699474811554, "global_step": 117556, "epoch": 1320, "lr": 6.059661618159129e-05}
{"train_loss": 0.14470939338207245, "global_step": 117557, "epoch": 1320, "lr": 6.0596049613372554e-05}
{"train_loss": 0.07491624355316162, "global_step": 117558, "epoch": 1320, "lr": 6.0595483043729316e-05}
{"train_loss": 0.12004776298999786, "global_step": 117559, "epoch": 1320, "lr": 6.0594916472661646e-05}
{"train_loss": 0.10805736482143402, "global_step": 117560, "epoch": 1320, "lr": 6.059434990016961e-05}
{"train_loss": 0.17783483862876892, "global_step": 117561, "epoch": 1320, "lr": 6.0593783326253296e-05}
{"train_loss": 0.22993095219135284, "global_step": 117562, "epoch": 1320, "lr": 6.059321675091277e-05}
{"train_loss": 0.11024140566587448, "global_step": 117563, "epoch": 1320, "lr": 6.059265017414813e-05}
{"train_loss": 0.1431034654378891, "global_step": 117564, "epoch": 1320, "lr": 6.0592083595959424e-05}
{"train_loss": 0.11295119673013687, "global_step": 117565, "epoch": 1320, "lr": 6.0591517016346744e-05}
{"train_loss": 0.12972961366176605, "global_step": 117566, "epoch": 1320, "lr": 6.059095043531017e-05}
{"train_loss": 0.14495407044887543, "global_step": 117567, "epoch": 1320, "lr": 6.059038385284976e-05}
{"train_loss": 0.14021830913725863, "global_step": 117568, "epoch": 1320, "lr": 6.058981726896561e-05, "val_loss": 5.231708526611328, "train_action_mse_error": 22.85940170288086}
{"train_loss": 0.1434985101222992, "global_step": 117569, "epoch": 1321, "lr": 6.0589250683657784e-05}
{"train_loss": 0.11476006358861923, "global_step": 117570, "epoch": 1321, "lr": 6.0588684096926374e-05}
{"train_loss": 0.14868667721748352, "global_step": 117571, "epoch": 1321, "lr": 6.058811750877143e-05}
{"train_loss": 0.1684233844280243, "global_step": 117572, "epoch": 1321, "lr": 6.0587550919193045e-05}
{"train_loss": 0.1330450475215912, "global_step": 117573, "epoch": 1321, "lr": 6.05869843281913e-05}
{"train_loss": 0.11386995017528534, "global_step": 117574, "epoch": 1321, "lr": 6.0586417735766265e-05}
{"train_loss": 0.15386179089546204, "global_step": 117575, "epoch": 1321, "lr": 6.058585114191802e-05}
{"train_loss": 0.22583776712417603, "global_step": 117576, "epoch": 1321, "lr": 6.0585284546646625e-05}
{"train_loss": 0.16093945503234863, "global_step": 117577, "epoch": 1321, "lr": 6.0584717949952176e-05}
{"train_loss": 0.15013746917247772, "global_step": 117578, "epoch": 1321, "lr": 6.058415135183474e-05}
{"train_loss": 0.09383003413677216, "global_step": 117579, "epoch": 1321, "lr": 6.058358475229439e-05}
{"train_loss": 0.13163799047470093, "global_step": 117580, "epoch": 1321, "lr": 6.058301815133119e-05}
{"train_loss": 0.0930418148636818, "global_step": 117581, "epoch": 1321, "lr": 6.0582451548945266e-05}
{"train_loss": 0.09593018144369125, "global_step": 117582, "epoch": 1321, "lr": 6.0581884945136635e-05}
{"train_loss": 0.11708462238311768, "global_step": 117583, "epoch": 1321, "lr": 6.05813183399054e-05}
{"train_loss": 0.1722511351108551, "global_step": 117584, "epoch": 1321, "lr": 6.058075173325164e-05}
{"train_loss": 0.13645975291728973, "global_step": 117585, "epoch": 1321, "lr": 6.058018512517543e-05}
{"train_loss": 0.09061770141124725, "global_step": 117586, "epoch": 1321, "lr": 6.057961851567683e-05}
{"train_loss": 0.1955711543560028, "global_step": 117587, "epoch": 1321, "lr": 6.057905190475595e-05}
{"train_loss": 0.18383309245109558, "global_step": 117588, "epoch": 1321, "lr": 6.0578485292412824e-05}
{"train_loss": 0.09978395700454712, "global_step": 117589, "epoch": 1321, "lr": 6.057791867864756e-05}
{"train_loss": 0.17446660995483398, "global_step": 117590, "epoch": 1321, "lr": 6.0577352063460226e-05}
{"train_loss": 0.13396379351615906, "global_step": 117591, "epoch": 1321, "lr": 6.05767854468509e-05}
{"train_loss": 0.22732067108154297, "global_step": 117592, "epoch": 1321, "lr": 6.057621882881965e-05}
{"train_loss": 0.14080430567264557, "global_step": 117593, "epoch": 1321, "lr": 6.057565220936655e-05}
{"train_loss": 0.11592812836170197, "global_step": 117594, "epoch": 1321, "lr": 6.0575085588491686e-05}
{"train_loss": 0.13302281498908997, "global_step": 117595, "epoch": 1321, "lr": 6.057451896619515e-05}
{"train_loss": 0.1336001455783844, "global_step": 117596, "epoch": 1321, "lr": 6.057395234247698e-05}
{"train_loss": 0.11607908457517624, "global_step": 117597, "epoch": 1321, "lr": 6.057338571733727e-05}
{"train_loss": 0.08372577279806137, "global_step": 117598, "epoch": 1321, "lr": 6.05728190907761e-05}
{"train_loss": 0.10644686967134476, "global_step": 117599, "epoch": 1321, "lr": 6.0572252462793545e-05}
{"train_loss": 0.1433098018169403, "global_step": 117600, "epoch": 1321, "lr": 6.057168583338968e-05}
{"train_loss": 0.17930111289024353, "global_step": 117601, "epoch": 1321, "lr": 6.0571119202564574e-05}
{"train_loss": 0.20794489979743958, "global_step": 117602, "epoch": 1321, "lr": 6.0570552570318316e-05}
{"train_loss": 0.12969844043254852, "global_step": 117603, "epoch": 1321, "lr": 6.0569985936650975e-05}
{"train_loss": 0.13817258179187775, "global_step": 117604, "epoch": 1321, "lr": 6.056941930156262e-05}
{"train_loss": 0.2149357795715332, "global_step": 117605, "epoch": 1321, "lr": 6.056885266505333e-05}
{"train_loss": 0.11605312675237656, "global_step": 117606, "epoch": 1321, "lr": 6.05682860271232e-05}
{"train_loss": 0.07978074252605438, "global_step": 117607, "epoch": 1321, "lr": 6.056771938777229e-05}
{"train_loss": 0.2441982924938202, "global_step": 117608, "epoch": 1321, "lr": 6.056715274700069e-05}
{"train_loss": 0.10784183442592621, "global_step": 117609, "epoch": 1321, "lr": 6.0566586104808454e-05}
{"train_loss": 0.1390402764081955, "global_step": 117610, "epoch": 1321, "lr": 6.0566019461195665e-05}
{"train_loss": 0.13786868751049042, "global_step": 117611, "epoch": 1321, "lr": 6.0565452816162415e-05}
{"train_loss": 0.13334135711193085, "global_step": 117612, "epoch": 1321, "lr": 6.056488616970876e-05}
{"train_loss": 0.13710300624370575, "global_step": 117613, "epoch": 1321, "lr": 6.056431952183479e-05}
{"train_loss": 0.08779837191104889, "global_step": 117614, "epoch": 1321, "lr": 6.056375287254058e-05}
{"train_loss": 0.15273205935955048, "global_step": 117615, "epoch": 1321, "lr": 6.05631862218262e-05}
{"train_loss": 0.12548178434371948, "global_step": 117616, "epoch": 1321, "lr": 6.056261956969173e-05}
{"train_loss": 0.0915026068687439, "global_step": 117617, "epoch": 1321, "lr": 6.056205291613723e-05}
{"train_loss": 0.12452836334705353, "global_step": 117618, "epoch": 1321, "lr": 6.0561486261162803e-05}
{"train_loss": 0.09919194877147675, "global_step": 117619, "epoch": 1321, "lr": 6.0560919604768504e-05}
{"train_loss": 0.08046270161867142, "global_step": 117620, "epoch": 1321, "lr": 6.056035294695442e-05}
{"train_loss": 0.1756523996591568, "global_step": 117621, "epoch": 1321, "lr": 6.055978628772063e-05}
{"train_loss": 0.1581530123949051, "global_step": 117622, "epoch": 1321, "lr": 6.055921962706721e-05}
{"train_loss": 0.06748262792825699, "global_step": 117623, "epoch": 1321, "lr": 6.0558652964994214e-05}
{"train_loss": 0.1219850704073906, "global_step": 117624, "epoch": 1321, "lr": 6.055808630150175e-05}
{"train_loss": 0.18545573949813843, "global_step": 117625, "epoch": 1321, "lr": 6.055751963658988e-05}
{"train_loss": 0.10268283635377884, "global_step": 117626, "epoch": 1321, "lr": 6.0556952970258675e-05}
{"train_loss": 0.13993467390537262, "global_step": 117627, "epoch": 1321, "lr": 6.055638630250822e-05}
{"train_loss": 0.09930186718702316, "global_step": 117628, "epoch": 1321, "lr": 6.055581963333858e-05}
{"train_loss": 0.08744452148675919, "global_step": 117629, "epoch": 1321, "lr": 6.055525296274985e-05}
{"train_loss": 0.23623548448085785, "global_step": 117630, "epoch": 1321, "lr": 6.0554686290742094e-05}
{"train_loss": 0.1398390680551529, "global_step": 117631, "epoch": 1321, "lr": 6.0554119617315385e-05}
{"train_loss": 0.14837360382080078, "global_step": 117632, "epoch": 1321, "lr": 6.055355294246981e-05}
{"train_loss": 0.08591385185718536, "global_step": 117633, "epoch": 1321, "lr": 6.055298626620542e-05}
{"train_loss": 0.09723395854234695, "global_step": 117634, "epoch": 1321, "lr": 6.055241958852233e-05}
{"train_loss": 0.1605948954820633, "global_step": 117635, "epoch": 1321, "lr": 6.055185290942059e-05}
{"train_loss": 0.1557832509279251, "global_step": 117636, "epoch": 1321, "lr": 6.0551286228900285e-05}
{"train_loss": 0.18186457455158234, "global_step": 117637, "epoch": 1321, "lr": 6.0550719546961485e-05}
{"train_loss": 0.19716280698776245, "global_step": 117638, "epoch": 1321, "lr": 6.055015286360427e-05}
{"train_loss": 0.08234824985265732, "global_step": 117639, "epoch": 1321, "lr": 6.054958617882871e-05}
{"train_loss": 0.10843586176633835, "global_step": 117640, "epoch": 1321, "lr": 6.054901949263488e-05}
{"train_loss": 0.13552235066890717, "global_step": 117641, "epoch": 1321, "lr": 6.054845280502287e-05}
{"train_loss": 0.1264478713274002, "global_step": 117642, "epoch": 1321, "lr": 6.0547886115992744e-05}
{"train_loss": 0.10700160264968872, "global_step": 117643, "epoch": 1321, "lr": 6.054731942554459e-05}
{"train_loss": 0.1612706333398819, "global_step": 117644, "epoch": 1321, "lr": 6.054675273367849e-05}
{"train_loss": 0.102225661277771, "global_step": 117645, "epoch": 1321, "lr": 6.0546186040394495e-05}
{"train_loss": 0.1234532967209816, "global_step": 117646, "epoch": 1321, "lr": 6.0545619345692696e-05}
{"train_loss": 0.09490819275379181, "global_step": 117647, "epoch": 1321, "lr": 6.054505264957316e-05}
{"train_loss": 0.12805761396884918, "global_step": 117648, "epoch": 1321, "lr": 6.054448595203598e-05}
{"train_loss": 0.09607120603322983, "global_step": 117649, "epoch": 1321, "lr": 6.0543919253081214e-05}
{"train_loss": 0.12395329773426056, "global_step": 117650, "epoch": 1321, "lr": 6.054335255270896e-05}
{"train_loss": 0.09748353064060211, "global_step": 117651, "epoch": 1321, "lr": 6.054278585091926e-05}
{"train_loss": 0.10300196707248688, "global_step": 117652, "epoch": 1321, "lr": 6.054221914771222e-05}
{"train_loss": 0.1275227665901184, "global_step": 117653, "epoch": 1321, "lr": 6.054165244308793e-05}
{"train_loss": 0.108998142182827, "global_step": 117654, "epoch": 1321, "lr": 6.054108573704642e-05}
{"train_loss": 0.13930928707122803, "global_step": 117655, "epoch": 1321, "lr": 6.0540519029587795e-05}
{"train_loss": 0.11418740451335907, "global_step": 117656, "epoch": 1321, "lr": 6.053995232071212e-05}
{"train_loss": 0.13440940110536104, "global_step": 117657, "epoch": 1321, "lr": 6.0539385610419475e-05, "val_loss": 5.346961975097656}
{"train_loss": 0.16368547081947327, "global_step": 117658, "epoch": 1322, "lr": 6.0538818898709946e-05}
{"train_loss": 0.11287626624107361, "global_step": 117659, "epoch": 1322, "lr": 6.053825218558359e-05}
{"train_loss": 0.11579620093107224, "global_step": 117660, "epoch": 1322, "lr": 6.0537685471040504e-05}
{"train_loss": 0.0691157728433609, "global_step": 117661, "epoch": 1322, "lr": 6.053711875508075e-05}
{"train_loss": 0.10691629350185394, "global_step": 117662, "epoch": 1322, "lr": 6.0536552037704406e-05}
{"train_loss": 0.15462571382522583, "global_step": 117663, "epoch": 1322, "lr": 6.053598531891156e-05}
{"train_loss": 0.19224393367767334, "global_step": 117664, "epoch": 1322, "lr": 6.053541859870227e-05}
{"train_loss": 0.2042481154203415, "global_step": 117665, "epoch": 1322, "lr": 6.053485187707664e-05}
{"train_loss": 0.09394114464521408, "global_step": 117666, "epoch": 1322, "lr": 6.053428515403471e-05}
{"train_loss": 0.1716853380203247, "global_step": 117667, "epoch": 1322, "lr": 6.053371842957658e-05}
{"train_loss": 0.09205430001020432, "global_step": 117668, "epoch": 1322, "lr": 6.053315170370232e-05}
{"train_loss": 0.13617509603500366, "global_step": 117669, "epoch": 1322, "lr": 6.053258497641201e-05}
{"train_loss": 0.09586812555789948, "global_step": 117670, "epoch": 1322, "lr": 6.053201824770571e-05}
{"train_loss": 0.206558495759964, "global_step": 117671, "epoch": 1322, "lr": 6.0531451517583515e-05}
{"train_loss": 0.12084656208753586, "global_step": 117672, "epoch": 1322, "lr": 6.0530884786045514e-05}
{"train_loss": 0.13216744363307953, "global_step": 117673, "epoch": 1322, "lr": 6.053031805309174e-05}
{"train_loss": 0.09311043471097946, "global_step": 117674, "epoch": 1322, "lr": 6.052975131872229e-05}
{"train_loss": 0.21406254172325134, "global_step": 117675, "epoch": 1322, "lr": 6.052918458293726e-05}
{"train_loss": 0.16043858230113983, "global_step": 117676, "epoch": 1322, "lr": 6.0528617845736704e-05}
{"train_loss": 0.16652944684028625, "global_step": 117677, "epoch": 1322, "lr": 6.0528051107120686e-05}
{"train_loss": 0.11987027525901794, "global_step": 117678, "epoch": 1322, "lr": 6.0527484367089316e-05}
{"train_loss": 0.12009037286043167, "global_step": 117679, "epoch": 1322, "lr": 6.0526917625642656e-05}
{"train_loss": 0.11309052258729935, "global_step": 117680, "epoch": 1322, "lr": 6.052635088278078e-05}
{"train_loss": 0.1241847574710846, "global_step": 117681, "epoch": 1322, "lr": 6.052578413850376e-05}
{"train_loss": 0.12141160666942596, "global_step": 117682, "epoch": 1322, "lr": 6.052521739281167e-05}
{"train_loss": 0.12856417894363403, "global_step": 117683, "epoch": 1322, "lr": 6.05246506457046e-05}
{"train_loss": 0.10419928282499313, "global_step": 117684, "epoch": 1322, "lr": 6.052408389718263e-05}
{"train_loss": 0.15957742929458618, "global_step": 117685, "epoch": 1322, "lr": 6.052351714724581e-05}
{"train_loss": 0.10198084264993668, "global_step": 117686, "epoch": 1322, "lr": 6.052295039589424e-05}
{"train_loss": 0.1007901206612587, "global_step": 117687, "epoch": 1322, "lr": 6.052238364312799e-05}
{"train_loss": 0.1316557377576828, "global_step": 117688, "epoch": 1322, "lr": 6.052181688894712e-05}
{"train_loss": 0.1439969539642334, "global_step": 117689, "epoch": 1322, "lr": 6.052125013335173e-05}
{"train_loss": 0.12129645049571991, "global_step": 117690, "epoch": 1322, "lr": 6.052068337634188e-05}
{"train_loss": 0.12429585307836533, "global_step": 117691, "epoch": 1322, "lr": 6.052011661791766e-05}
{"train_loss": 0.11864400655031204, "global_step": 117692, "epoch": 1322, "lr": 6.051954985807915e-05}
{"train_loss": 0.12225572019815445, "global_step": 117693, "epoch": 1322, "lr": 6.05189830968264e-05}
{"train_loss": 0.11376757174730301, "global_step": 117694, "epoch": 1322, "lr": 6.05184163341595e-05}
{"train_loss": 0.14940455555915833, "global_step": 117695, "epoch": 1322, "lr": 6.051784957007851e-05}
{"train_loss": 0.13692449033260345, "global_step": 117696, "epoch": 1322, "lr": 6.051728280458354e-05}
{"train_loss": 0.1608906239271164, "global_step": 117697, "epoch": 1322, "lr": 6.0516716037674657e-05}
{"train_loss": 0.1399623155593872, "global_step": 117698, "epoch": 1322, "lr": 6.051614926935192e-05}
{"train_loss": 0.11063957214355469, "global_step": 117699, "epoch": 1322, "lr": 6.051558249961542e-05}
{"train_loss": 0.1323256641626358, "global_step": 117700, "epoch": 1322, "lr": 6.051501572846523e-05}
{"train_loss": 0.12354275584220886, "global_step": 117701, "epoch": 1322, "lr": 6.051444895590142e-05}
{"train_loss": 0.19970069825649261, "global_step": 117702, "epoch": 1322, "lr": 6.0513882181924064e-05}
{"train_loss": 0.11515884846448898, "global_step": 117703, "epoch": 1322, "lr": 6.0513315406533256e-05}
{"train_loss": 0.1834549754858017, "global_step": 117704, "epoch": 1322, "lr": 6.0512748629729056e-05}
{"train_loss": 0.11681868880987167, "global_step": 117705, "epoch": 1322, "lr": 6.051218185151154e-05}
{"train_loss": 0.18960991501808167, "global_step": 117706, "epoch": 1322, "lr": 6.05116150718808e-05}
{"train_loss": 0.1170530617237091, "global_step": 117707, "epoch": 1322, "lr": 6.051104829083689e-05}
{"train_loss": 0.1279832422733307, "global_step": 117708, "epoch": 1322, "lr": 6.051048150837991e-05}
{"train_loss": 0.14580972492694855, "global_step": 117709, "epoch": 1322, "lr": 6.050991472450992e-05}
{"train_loss": 0.13011428713798523, "global_step": 117710, "epoch": 1322, "lr": 6.0509347939227004e-05}
{"train_loss": 0.10952784866094589, "global_step": 117711, "epoch": 1322, "lr": 6.050878115253123e-05}
{"train_loss": 0.10889985412359238, "global_step": 117712, "epoch": 1322, "lr": 6.0508214364422674e-05}
{"train_loss": 0.1373974233865738, "global_step": 117713, "epoch": 1322, "lr": 6.050764757490142e-05}
{"train_loss": 0.07101646810770035, "global_step": 117714, "epoch": 1322, "lr": 6.050708078396753e-05}
{"train_loss": 0.08364919573068619, "global_step": 117715, "epoch": 1322, "lr": 6.050651399162111e-05}
{"train_loss": 0.14369800686836243, "global_step": 117716, "epoch": 1322, "lr": 6.05059471978622e-05}
{"train_loss": 0.1082635447382927, "global_step": 117717, "epoch": 1322, "lr": 6.0505380402690905e-05}
{"train_loss": 0.13254909217357635, "global_step": 117718, "epoch": 1322, "lr": 6.050481360610729e-05}
{"train_loss": 0.1357947438955307, "global_step": 117719, "epoch": 1322, "lr": 6.050424680811142e-05}
{"train_loss": 0.13180547952651978, "global_step": 117720, "epoch": 1322, "lr": 6.050368000870339e-05}
{"train_loss": 0.09751677513122559, "global_step": 117721, "epoch": 1322, "lr": 6.050311320788327e-05}
{"train_loss": 0.12605300545692444, "global_step": 117722, "epoch": 1322, "lr": 6.050254640565113e-05}
{"train_loss": 0.15288835763931274, "global_step": 117723, "epoch": 1322, "lr": 6.050197960200705e-05}
{"train_loss": 0.16366875171661377, "global_step": 117724, "epoch": 1322, "lr": 6.050141279695112e-05}
{"train_loss": 0.08508820831775665, "global_step": 117725, "epoch": 1322, "lr": 6.050084599048339e-05}
{"train_loss": 0.1841292381286621, "global_step": 117726, "epoch": 1322, "lr": 6.050027918260395e-05}
{"train_loss": 0.09949927777051926, "global_step": 117727, "epoch": 1322, "lr": 6.049971237331288e-05}
{"train_loss": 0.11425010859966278, "global_step": 117728, "epoch": 1322, "lr": 6.049914556261025e-05}
{"train_loss": 0.16379013657569885, "global_step": 117729, "epoch": 1322, "lr": 6.049857875049614e-05}
{"train_loss": 0.14871861040592194, "global_step": 117730, "epoch": 1322, "lr": 6.049801193697062e-05}
{"train_loss": 0.09987009316682816, "global_step": 117731, "epoch": 1322, "lr": 6.049744512203378e-05}
{"train_loss": 0.1497771292924881, "global_step": 117732, "epoch": 1322, "lr": 6.0496878305685666e-05}
{"train_loss": 0.17266923189163208, "global_step": 117733, "epoch": 1322, "lr": 6.0496311487926386e-05}
{"train_loss": 0.14865578711032867, "global_step": 117734, "epoch": 1322, "lr": 6.0495744668756e-05}
{"train_loss": 0.16845649480819702, "global_step": 117735, "epoch": 1322, "lr": 6.04951778481746e-05}
{"train_loss": 0.17867973446846008, "global_step": 117736, "epoch": 1322, "lr": 6.049461102618225e-05}
{"train_loss": 0.08203169703483582, "global_step": 117737, "epoch": 1322, "lr": 6.0494044202779024e-05}
{"train_loss": 0.14347003400325775, "global_step": 117738, "epoch": 1322, "lr": 6.049347737796499e-05}
{"train_loss": 0.19417007267475128, "global_step": 117739, "epoch": 1322, "lr": 6.0492910551740245e-05}
{"train_loss": 0.14300397038459778, "global_step": 117740, "epoch": 1322, "lr": 6.049234372410486e-05}
{"train_loss": 0.15200437605381012, "global_step": 117741, "epoch": 1322, "lr": 6.0491776895058906e-05}
{"train_loss": 0.14289012551307678, "global_step": 117742, "epoch": 1322, "lr": 6.049121006460245e-05}
{"train_loss": 0.1427212655544281, "global_step": 117743, "epoch": 1322, "lr": 6.04906432327356e-05}
{"train_loss": 0.13409298658370972, "global_step": 117744, "epoch": 1322, "lr": 6.049007639945839e-05}
{"train_loss": 0.22130759060382843, "global_step": 117745, "epoch": 1322, "lr": 6.048950956477093e-05}
{"train_loss": 0.13608719634540964, "global_step": 117746, "epoch": 1322, "lr": 6.0488942728673284e-05, "val_loss": 5.33844518661499}
{"train_loss": 0.2662997543811798, "global_step": 117747, "epoch": 1323, "lr": 6.048837589116552e-05}
{"train_loss": 0.06792188435792923, "global_step": 117748, "epoch": 1323, "lr": 6.048780905224772e-05}
{"train_loss": 0.17753832042217255, "global_step": 117749, "epoch": 1323, "lr": 6.048724221191997e-05}
{"train_loss": 0.15777520835399628, "global_step": 117750, "epoch": 1323, "lr": 6.0486675370182336e-05}
{"train_loss": 0.20046314597129822, "global_step": 117751, "epoch": 1323, "lr": 6.048610852703489e-05}
{"train_loss": 0.1813572347164154, "global_step": 117752, "epoch": 1323, "lr": 6.048554168247772e-05}
{"train_loss": 0.29143938422203064, "global_step": 117753, "epoch": 1323, "lr": 6.04849748365109e-05}
{"train_loss": 0.164589062333107, "global_step": 117754, "epoch": 1323, "lr": 6.0484407989134493e-05}
{"train_loss": 0.1373598873615265, "global_step": 117755, "epoch": 1323, "lr": 6.048384114034859e-05}
{"train_loss": 0.17846103012561798, "global_step": 117756, "epoch": 1323, "lr": 6.048327429015327e-05}
{"train_loss": 0.2449249029159546, "global_step": 117757, "epoch": 1323, "lr": 6.048270743854859e-05}
{"train_loss": 0.12107458710670471, "global_step": 117758, "epoch": 1323, "lr": 6.048214058553465e-05}
{"train_loss": 0.210446298122406, "global_step": 117759, "epoch": 1323, "lr": 6.048157373111151e-05}
{"train_loss": 0.11484617739915848, "global_step": 117760, "epoch": 1323, "lr": 6.048100687527924e-05}
{"train_loss": 0.17752188444137573, "global_step": 117761, "epoch": 1323, "lr": 6.048044001803793e-05}
{"train_loss": 0.06620139628648758, "global_step": 117762, "epoch": 1323, "lr": 6.0479873159387656e-05}
{"train_loss": 0.11360663175582886, "global_step": 117763, "epoch": 1323, "lr": 6.04793062993285e-05}
{"train_loss": 0.16678160429000854, "global_step": 117764, "epoch": 1323, "lr": 6.047873943786051e-05}
{"train_loss": 0.14485415816307068, "global_step": 117765, "epoch": 1323, "lr": 6.0478172574983796e-05}
{"train_loss": 0.12921513617038727, "global_step": 117766, "epoch": 1323, "lr": 6.047760571069841e-05}
{"train_loss": 0.14110955595970154, "global_step": 117767, "epoch": 1323, "lr": 6.047703884500444e-05}
{"train_loss": 0.12305381894111633, "global_step": 117768, "epoch": 1323, "lr": 6.047647197790196e-05}
{"train_loss": 0.13316583633422852, "global_step": 117769, "epoch": 1323, "lr": 6.0475905109391054e-05}
{"train_loss": 0.25036925077438354, "global_step": 117770, "epoch": 1323, "lr": 6.047533823947178e-05}
{"train_loss": 0.11610080301761627, "global_step": 117771, "epoch": 1323, "lr": 6.0474771368144236e-05}
{"train_loss": 0.14503827691078186, "global_step": 117772, "epoch": 1323, "lr": 6.047420449540847e-05}
{"train_loss": 0.14919255673885345, "global_step": 117773, "epoch": 1323, "lr": 6.047363762126459e-05}
{"train_loss": 0.2509124279022217, "global_step": 117774, "epoch": 1323, "lr": 6.047307074571265e-05}
{"train_loss": 0.09873361140489578, "global_step": 117775, "epoch": 1323, "lr": 6.047250386875274e-05}
{"train_loss": 0.14794395864009857, "global_step": 117776, "epoch": 1323, "lr": 6.0471936990384916e-05}
{"train_loss": 0.15063829720020294, "global_step": 117777, "epoch": 1323, "lr": 6.0471370110609285e-05}
{"train_loss": 0.17229346930980682, "global_step": 117778, "epoch": 1323, "lr": 6.0470803229425886e-05}
{"train_loss": 0.17545470595359802, "global_step": 117779, "epoch": 1323, "lr": 6.047023634683483e-05}
{"train_loss": 0.105579674243927, "global_step": 117780, "epoch": 1323, "lr": 6.0469669462836167e-05}
{"train_loss": 0.1200837790966034, "global_step": 117781, "epoch": 1323, "lr": 6.046910257742999e-05}
{"train_loss": 0.16621248424053192, "global_step": 117782, "epoch": 1323, "lr": 6.046853569061638e-05}
{"train_loss": 0.14887812733650208, "global_step": 117783, "epoch": 1323, "lr": 6.046796880239539e-05}
{"train_loss": 0.09050599485635757, "global_step": 117784, "epoch": 1323, "lr": 6.046740191276712e-05}
{"train_loss": 0.2011858969926834, "global_step": 117785, "epoch": 1323, "lr": 6.0466835021731624e-05}
{"train_loss": 0.2107192575931549, "global_step": 117786, "epoch": 1323, "lr": 6.0466268129288995e-05}
{"train_loss": 0.10721147060394287, "global_step": 117787, "epoch": 1323, "lr": 6.04657012354393e-05}
{"train_loss": 0.15382377803325653, "global_step": 117788, "epoch": 1323, "lr": 6.046513434018262e-05}
{"train_loss": 0.13095316290855408, "global_step": 117789, "epoch": 1323, "lr": 6.0464567443519035e-05}
{"train_loss": 0.1128690242767334, "global_step": 117790, "epoch": 1323, "lr": 6.046400054544862e-05}
{"train_loss": 0.1039842814207077, "global_step": 117791, "epoch": 1323, "lr": 6.0463433645971434e-05}
{"train_loss": 0.11113978177309036, "global_step": 117792, "epoch": 1323, "lr": 6.046286674508758e-05}
{"train_loss": 0.12590675055980682, "global_step": 117793, "epoch": 1323, "lr": 6.046229984279711e-05}
{"train_loss": 0.1723344922065735, "global_step": 117794, "epoch": 1323, "lr": 6.0461732939100114e-05}
{"train_loss": 0.11778619885444641, "global_step": 117795, "epoch": 1323, "lr": 6.0461166033996674e-05}
{"train_loss": 0.12593498826026917, "global_step": 117796, "epoch": 1323, "lr": 6.046059912748685e-05}
{"train_loss": 0.12794159352779388, "global_step": 117797, "epoch": 1323, "lr": 6.046003221957073e-05}
{"train_loss": 0.10898011177778244, "global_step": 117798, "epoch": 1323, "lr": 6.045946531024837e-05}
{"train_loss": 0.11688406765460968, "global_step": 117799, "epoch": 1323, "lr": 6.0458898399519894e-05}
{"train_loss": 0.10383796691894531, "global_step": 117800, "epoch": 1323, "lr": 6.045833148738532e-05}
{"train_loss": 0.11916831880807877, "global_step": 117801, "epoch": 1323, "lr": 6.045776457384477e-05}
{"train_loss": 0.12399497628211975, "global_step": 117802, "epoch": 1323, "lr": 6.0457197658898276e-05}
{"train_loss": 0.16361966729164124, "global_step": 117803, "epoch": 1323, "lr": 6.045663074254596e-05}
{"train_loss": 0.1058996394276619, "global_step": 117804, "epoch": 1323, "lr": 6.045606382478788e-05}
{"train_loss": 0.14763405919075012, "global_step": 117805, "epoch": 1323, "lr": 6.04554969056241e-05}
{"train_loss": 0.11966005712747574, "global_step": 117806, "epoch": 1323, "lr": 6.04549299850547e-05}
{"train_loss": 0.08911576122045517, "global_step": 117807, "epoch": 1323, "lr": 6.0454363063079764e-05}
{"train_loss": 0.12970231473445892, "global_step": 117808, "epoch": 1323, "lr": 6.045379613969938e-05}
{"train_loss": 0.11846480518579483, "global_step": 117809, "epoch": 1323, "lr": 6.0453229214913596e-05}
{"train_loss": 0.12104260176420212, "global_step": 117810, "epoch": 1323, "lr": 6.045266228872252e-05}
{"train_loss": 0.12693668901920319, "global_step": 117811, "epoch": 1323, "lr": 6.045209536112619e-05}
{"train_loss": 0.11017192155122757, "global_step": 117812, "epoch": 1323, "lr": 6.045152843212471e-05}
{"train_loss": 0.13735993206501007, "global_step": 117813, "epoch": 1323, "lr": 6.045096150171816e-05}
{"train_loss": 0.09048303961753845, "global_step": 117814, "epoch": 1323, "lr": 6.04503945699066e-05}
{"train_loss": 0.15653863549232483, "global_step": 117815, "epoch": 1323, "lr": 6.04498276366901e-05}
{"train_loss": 0.24604542553424835, "global_step": 117816, "epoch": 1323, "lr": 6.044926070206877e-05}
{"train_loss": 0.1657562404870987, "global_step": 117817, "epoch": 1323, "lr": 6.044869376604264e-05}
{"train_loss": 0.11463702470064163, "global_step": 117818, "epoch": 1323, "lr": 6.044812682861183e-05}
{"train_loss": 0.05880739167332649, "global_step": 117819, "epoch": 1323, "lr": 6.0447559889776386e-05}
{"train_loss": 0.20117419958114624, "global_step": 117820, "epoch": 1323, "lr": 6.04469929495364e-05}
{"train_loss": 0.14022672176361084, "global_step": 117821, "epoch": 1323, "lr": 6.0446426007891944e-05}
{"train_loss": 0.1338285505771637, "global_step": 117822, "epoch": 1323, "lr": 6.044585906484308e-05}
{"train_loss": 0.1273311972618103, "global_step": 117823, "epoch": 1323, "lr": 6.044529212038992e-05}
{"train_loss": 0.1969873309135437, "global_step": 117824, "epoch": 1323, "lr": 6.0444725174532503e-05}
{"train_loss": 0.10172925144433975, "global_step": 117825, "epoch": 1323, "lr": 6.044415822727092e-05}
{"train_loss": 0.08186596632003784, "global_step": 117826, "epoch": 1323, "lr": 6.044359127860525e-05}
{"train_loss": 0.15785472095012665, "global_step": 117827, "epoch": 1323, "lr": 6.044302432853557e-05}
{"train_loss": 0.09895187616348267, "global_step": 117828, "epoch": 1323, "lr": 6.0442457377061946e-05}
{"train_loss": 0.19444185495376587, "global_step": 117829, "epoch": 1323, "lr": 6.0441890424184465e-05}
{"train_loss": 0.2310829758644104, "global_step": 117830, "epoch": 1323, "lr": 6.04413234699032e-05}
{"train_loss": 0.09764751046895981, "global_step": 117831, "epoch": 1323, "lr": 6.0440756514218224e-05}
{"train_loss": 0.08242571353912354, "global_step": 117832, "epoch": 1323, "lr": 6.0440189557129613e-05}
{"train_loss": 0.21902644634246826, "global_step": 117833, "epoch": 1323, "lr": 6.043962259863745e-05}
{"train_loss": 0.1378825604915619, "global_step": 117834, "epoch": 1323, "lr": 6.0439055638741805e-05}
{"train_loss": 0.14575726091024582, "global_step": 117835, "epoch": 1323, "lr": 6.043848867744275e-05, "val_loss": 5.1668853759765625}
{"train_loss": 0.13446210324764252, "global_step": 117836, "epoch": 1324, "lr": 6.043792171474038e-05}
{"train_loss": 0.10410355776548386, "global_step": 117837, "epoch": 1324, "lr": 6.0437354750634756e-05}
{"train_loss": 0.14334921538829803, "global_step": 117838, "epoch": 1324, "lr": 6.043678778512595e-05}
{"train_loss": 0.13620266318321228, "global_step": 117839, "epoch": 1324, "lr": 6.043622081821405e-05}
{"train_loss": 0.10111077129840851, "global_step": 117840, "epoch": 1324, "lr": 6.043565384989911e-05}
{"train_loss": 0.11207728087902069, "global_step": 117841, "epoch": 1324, "lr": 6.043508688018126e-05}
{"train_loss": 0.15737134218215942, "global_step": 117842, "epoch": 1324, "lr": 6.043451990906052e-05}
{"train_loss": 0.14226257801055908, "global_step": 117843, "epoch": 1324, "lr": 6.043395293653699e-05}
{"train_loss": 0.22832009196281433, "global_step": 117844, "epoch": 1324, "lr": 6.043338596261073e-05}
{"train_loss": 0.17827658355236053, "global_step": 117845, "epoch": 1324, "lr": 6.0432818987281835e-05}
{"train_loss": 0.12536770105361938, "global_step": 117846, "epoch": 1324, "lr": 6.0432252010550374e-05}
{"train_loss": 0.0820617601275444, "global_step": 117847, "epoch": 1324, "lr": 6.043168503241643e-05}
{"train_loss": 0.15839466452598572, "global_step": 117848, "epoch": 1324, "lr": 6.0431118052880064e-05}
{"train_loss": 0.12079164385795593, "global_step": 117849, "epoch": 1324, "lr": 6.0430551071941364e-05}
{"train_loss": 0.08261194825172424, "global_step": 117850, "epoch": 1324, "lr": 6.042998408960041e-05}
{"train_loss": 0.17611023783683777, "global_step": 117851, "epoch": 1324, "lr": 6.0429417105857266e-05}
{"train_loss": 0.12808860838413239, "global_step": 117852, "epoch": 1324, "lr": 6.0428850120712024e-05}
{"train_loss": 0.13610179722309113, "global_step": 117853, "epoch": 1324, "lr": 6.042828313416473e-05}
{"train_loss": 0.13039951026439667, "global_step": 117854, "epoch": 1324, "lr": 6.04277161462155e-05}
{"train_loss": 0.11216127127408981, "global_step": 117855, "epoch": 1324, "lr": 6.042714915686438e-05}
{"train_loss": 0.15807393193244934, "global_step": 117856, "epoch": 1324, "lr": 6.0426582166111454e-05}
{"train_loss": 0.1413908451795578, "global_step": 117857, "epoch": 1324, "lr": 6.0426015173956805e-05}
{"train_loss": 0.11370697617530823, "global_step": 117858, "epoch": 1324, "lr": 6.042544818040051e-05}
{"train_loss": 0.1719593107700348, "global_step": 117859, "epoch": 1324, "lr": 6.042488118544263e-05}
{"train_loss": 0.09728437662124634, "global_step": 117860, "epoch": 1324, "lr": 6.042431418908326e-05}
{"train_loss": 0.19564618170261383, "global_step": 117861, "epoch": 1324, "lr": 6.042374719132248e-05}
{"train_loss": 0.13335414230823517, "global_step": 117862, "epoch": 1324, "lr": 6.042318019216035e-05}
{"train_loss": 0.09223797917366028, "global_step": 117863, "epoch": 1324, "lr": 6.042261319159693e-05}
{"train_loss": 0.14966414868831635, "global_step": 117864, "epoch": 1324, "lr": 6.042204618963233e-05}
{"train_loss": 0.09366697072982788, "global_step": 117865, "epoch": 1324, "lr": 6.042147918626662e-05}
{"train_loss": 0.17631089687347412, "global_step": 117866, "epoch": 1324, "lr": 6.042091218149986e-05}
{"train_loss": 0.1077689677476883, "global_step": 117867, "epoch": 1324, "lr": 6.0420345175332125e-05}
{"train_loss": 0.08616756647825241, "global_step": 117868, "epoch": 1324, "lr": 6.041977816776352e-05}
{"train_loss": 0.10575765371322632, "global_step": 117869, "epoch": 1324, "lr": 6.0419211158794096e-05}
{"train_loss": 0.13650377094745636, "global_step": 117870, "epoch": 1324, "lr": 6.041864414842393e-05}
{"train_loss": 0.2620055377483368, "global_step": 117871, "epoch": 1324, "lr": 6.041807713665312e-05}
{"train_loss": 0.21969297528266907, "global_step": 117872, "epoch": 1324, "lr": 6.041751012348171e-05}
{"train_loss": 0.13010212779045105, "global_step": 117873, "epoch": 1324, "lr": 6.0416943108909794e-05}
{"train_loss": 0.19112606346607208, "global_step": 117874, "epoch": 1324, "lr": 6.041637609293745e-05}
{"train_loss": 0.15410183370113373, "global_step": 117875, "epoch": 1324, "lr": 6.0415809075564754e-05}
{"train_loss": 0.16845490038394928, "global_step": 117876, "epoch": 1324, "lr": 6.041524205679179e-05}
{"train_loss": 0.1346336305141449, "global_step": 117877, "epoch": 1324, "lr": 6.04146750366186e-05}
{"train_loss": 0.10838805884122849, "global_step": 117878, "epoch": 1324, "lr": 6.041410801504529e-05}
{"train_loss": 0.19581377506256104, "global_step": 117879, "epoch": 1324, "lr": 6.041354099207194e-05}
{"train_loss": 0.17356617748737335, "global_step": 117880, "epoch": 1324, "lr": 6.041297396769863e-05}
{"train_loss": 0.11102060228586197, "global_step": 117881, "epoch": 1324, "lr": 6.0412406941925405e-05}
{"train_loss": 0.0884563997387886, "global_step": 117882, "epoch": 1324, "lr": 6.0411839914752354e-05}
{"train_loss": 0.16341903805732727, "global_step": 117883, "epoch": 1324, "lr": 6.0411272886179556e-05}
{"train_loss": 0.10247724503278732, "global_step": 117884, "epoch": 1324, "lr": 6.041070585620711e-05}
{"train_loss": 0.09785226732492447, "global_step": 117885, "epoch": 1324, "lr": 6.041013882483505e-05}
{"train_loss": 0.1560806781053543, "global_step": 117886, "epoch": 1324, "lr": 6.040957179206348e-05}
{"train_loss": 0.1545514166355133, "global_step": 117887, "epoch": 1324, "lr": 6.0409004757892475e-05}
{"train_loss": 0.1236625537276268, "global_step": 117888, "epoch": 1324, "lr": 6.0408437722322095e-05}
{"train_loss": 0.17523424327373505, "global_step": 117889, "epoch": 1324, "lr": 6.040787068535244e-05}
{"train_loss": 0.1347295343875885, "global_step": 117890, "epoch": 1324, "lr": 6.0407303646983574e-05}
{"train_loss": 0.11200718581676483, "global_step": 117891, "epoch": 1324, "lr": 6.040673660721556e-05}
{"train_loss": 0.16462823748588562, "global_step": 117892, "epoch": 1324, "lr": 6.0406169566048496e-05}
{"train_loss": 0.16061802208423615, "global_step": 117893, "epoch": 1324, "lr": 6.040560252348245e-05}
{"train_loss": 0.2023323029279709, "global_step": 117894, "epoch": 1324, "lr": 6.040503547951749e-05}
{"train_loss": 0.1663362830877304, "global_step": 117895, "epoch": 1324, "lr": 6.04044684341537e-05}
{"train_loss": 0.2818413972854614, "global_step": 117896, "epoch": 1324, "lr": 6.040390138739116e-05}
{"train_loss": 0.09340416640043259, "global_step": 117897, "epoch": 1324, "lr": 6.0403334339229944e-05}
{"train_loss": 0.2005147486925125, "global_step": 117898, "epoch": 1324, "lr": 6.040276728967012e-05}
{"train_loss": 0.12109924852848053, "global_step": 117899, "epoch": 1324, "lr": 6.040220023871179e-05}
{"train_loss": 0.13105888664722443, "global_step": 117900, "epoch": 1324, "lr": 6.0401633186355e-05}
{"train_loss": 0.1435067057609558, "global_step": 117901, "epoch": 1324, "lr": 6.040106613259984e-05}
{"train_loss": 0.11952415108680725, "global_step": 117902, "epoch": 1324, "lr": 6.040049907744638e-05}
{"train_loss": 0.15706588327884674, "global_step": 117903, "epoch": 1324, "lr": 6.039993202089469e-05}
{"train_loss": 0.20505157113075256, "global_step": 117904, "epoch": 1324, "lr": 6.039936496294487e-05}
{"train_loss": 0.14015617966651917, "global_step": 117905, "epoch": 1324, "lr": 6.0398797903596974e-05}
{"train_loss": 0.1756436675786972, "global_step": 117906, "epoch": 1324, "lr": 6.039823084285109e-05}
{"train_loss": 0.1494317352771759, "global_step": 117907, "epoch": 1324, "lr": 6.039766378070729e-05}
{"train_loss": 0.1628534495830536, "global_step": 117908, "epoch": 1324, "lr": 6.0397096717165646e-05}
{"train_loss": 0.12642963230609894, "global_step": 117909, "epoch": 1324, "lr": 6.039652965222624e-05}
{"train_loss": 0.11884437501430511, "global_step": 117910, "epoch": 1324, "lr": 6.039596258588915e-05}
{"train_loss": 0.09355907887220383, "global_step": 117911, "epoch": 1324, "lr": 6.039539551815445e-05}
{"train_loss": 0.14499935507774353, "global_step": 117912, "epoch": 1324, "lr": 6.0394828449022224e-05}
{"train_loss": 0.1834426373243332, "global_step": 117913, "epoch": 1324, "lr": 6.0394261378492524e-05}
{"train_loss": 0.0843915045261383, "global_step": 117914, "epoch": 1324, "lr": 6.039369430656545e-05}
{"train_loss": 0.08467071503400803, "global_step": 117915, "epoch": 1324, "lr": 6.039312723324106e-05}
{"train_loss": 0.1409313827753067, "global_step": 117916, "epoch": 1324, "lr": 6.0392560158519464e-05}
{"train_loss": 0.11168059706687927, "global_step": 117917, "epoch": 1324, "lr": 6.03919930824007e-05}
{"train_loss": 0.14138351380825043, "global_step": 117918, "epoch": 1324, "lr": 6.0391426004884856e-05}
{"train_loss": 0.1252530962228775, "global_step": 117919, "epoch": 1324, "lr": 6.039085892597203e-05}
{"train_loss": 0.1400950700044632, "global_step": 117920, "epoch": 1324, "lr": 6.0390291845662264e-05}
{"train_loss": 0.10748720169067383, "global_step": 117921, "epoch": 1324, "lr": 6.038972476395566e-05}
{"train_loss": 0.15700308978557587, "global_step": 117922, "epoch": 1324, "lr": 6.038915768085227e-05}
{"train_loss": 0.11093559116125107, "global_step": 117923, "epoch": 1324, "lr": 6.038859059635219e-05}
{"train_loss": 0.1412750883551126, "global_step": 117924, "epoch": 1324, "lr": 6.0388023510455496e-05, "val_loss": 5.175838470458984}
{"train_loss": 0.12714330852031708, "global_step": 117925, "epoch": 1325, "lr": 6.038745642316225e-05}
{"train_loss": 0.15666650235652924, "global_step": 117926, "epoch": 1325, "lr": 6.038688933447254e-05}
{"train_loss": 0.1187872439622879, "global_step": 117927, "epoch": 1325, "lr": 6.038632224438644e-05}
{"train_loss": 0.10720407962799072, "global_step": 117928, "epoch": 1325, "lr": 6.0385755152904034e-05}
{"train_loss": 0.16665354371070862, "global_step": 117929, "epoch": 1325, "lr": 6.0385188060025376e-05}
{"train_loss": 0.10297031700611115, "global_step": 117930, "epoch": 1325, "lr": 6.038462096575056e-05}
{"train_loss": 0.125472292304039, "global_step": 117931, "epoch": 1325, "lr": 6.038405387007965e-05}
{"train_loss": 0.13683941960334778, "global_step": 117932, "epoch": 1325, "lr": 6.038348677301274e-05}
{"train_loss": 0.06187191605567932, "global_step": 117933, "epoch": 1325, "lr": 6.03829196745499e-05}
{"train_loss": 0.14566075801849365, "global_step": 117934, "epoch": 1325, "lr": 6.03823525746912e-05}
{"train_loss": 0.1367105394601822, "global_step": 117935, "epoch": 1325, "lr": 6.0381785473436724e-05}
{"train_loss": 0.18769235908985138, "global_step": 117936, "epoch": 1325, "lr": 6.038121837078654e-05}
{"train_loss": 0.12027883529663086, "global_step": 117937, "epoch": 1325, "lr": 6.0380651266740726e-05}
{"train_loss": 0.1284671574831009, "global_step": 117938, "epoch": 1325, "lr": 6.0380084161299364e-05}
{"train_loss": 0.1389489471912384, "global_step": 117939, "epoch": 1325, "lr": 6.037951705446253e-05}
{"train_loss": 0.1332908570766449, "global_step": 117940, "epoch": 1325, "lr": 6.037894994623029e-05}
{"train_loss": 0.1369820386171341, "global_step": 117941, "epoch": 1325, "lr": 6.0378382836602735e-05}
{"train_loss": 0.12838250398635864, "global_step": 117942, "epoch": 1325, "lr": 6.037781572557991e-05}
{"train_loss": 0.10747598111629486, "global_step": 117943, "epoch": 1325, "lr": 6.037724861316193e-05}
{"train_loss": 0.11120717972517014, "global_step": 117944, "epoch": 1325, "lr": 6.0376681499348855e-05}
{"train_loss": 0.08800782263278961, "global_step": 117945, "epoch": 1325, "lr": 6.037611438414077e-05}
{"train_loss": 0.19958820939064026, "global_step": 117946, "epoch": 1325, "lr": 6.037554726753772e-05}
{"train_loss": 0.0957396999001503, "global_step": 117947, "epoch": 1325, "lr": 6.037498014953983e-05}
{"train_loss": 0.18188796937465668, "global_step": 117948, "epoch": 1325, "lr": 6.0374413030147134e-05}
{"train_loss": 0.19340281188488007, "global_step": 117949, "epoch": 1325, "lr": 6.0373845909359716e-05}
{"train_loss": 0.17661148309707642, "global_step": 117950, "epoch": 1325, "lr": 6.0373278787177676e-05}
{"train_loss": 0.1670299470424652, "global_step": 117951, "epoch": 1325, "lr": 6.0372711663601067e-05}
{"train_loss": 0.08918430656194687, "global_step": 117952, "epoch": 1325, "lr": 6.0372144538629984e-05}
{"train_loss": 0.2045523226261139, "global_step": 117953, "epoch": 1325, "lr": 6.037157741226449e-05}
{"train_loss": 0.15926815569400787, "global_step": 117954, "epoch": 1325, "lr": 6.037101028450466e-05}
{"train_loss": 0.10695250332355499, "global_step": 117955, "epoch": 1325, "lr": 6.037044315535058e-05}
{"train_loss": 0.14287979900836945, "global_step": 117956, "epoch": 1325, "lr": 6.036987602480231e-05}
{"train_loss": 0.1005907952785492, "global_step": 117957, "epoch": 1325, "lr": 6.036930889285996e-05}
{"train_loss": 0.16992022097110748, "global_step": 117958, "epoch": 1325, "lr": 6.0368741759523564e-05}
{"train_loss": 0.08508273959159851, "global_step": 117959, "epoch": 1325, "lr": 6.0368174624793204e-05}
{"train_loss": 0.10945650935173035, "global_step": 117960, "epoch": 1325, "lr": 6.036760748866899e-05}
{"train_loss": 0.11305482685565948, "global_step": 117961, "epoch": 1325, "lr": 6.036704035115097e-05}
{"train_loss": 0.13712266087532043, "global_step": 117962, "epoch": 1325, "lr": 6.036647321223923e-05}
{"train_loss": 0.13126623630523682, "global_step": 117963, "epoch": 1325, "lr": 6.036590607193384e-05}
{"train_loss": 0.16203729808330536, "global_step": 117964, "epoch": 1325, "lr": 6.036533893023488e-05}
{"train_loss": 0.12548795342445374, "global_step": 117965, "epoch": 1325, "lr": 6.0364771787142435e-05}
{"train_loss": 0.09032699465751648, "global_step": 117966, "epoch": 1325, "lr": 6.036420464265656e-05}
{"train_loss": 0.11196659505367279, "global_step": 117967, "epoch": 1325, "lr": 6.036363749677735e-05}
{"train_loss": 0.22530069947242737, "global_step": 117968, "epoch": 1325, "lr": 6.036307034950488e-05}
{"train_loss": 0.1583147644996643, "global_step": 117969, "epoch": 1325, "lr": 6.0362503200839215e-05}
{"train_loss": 0.1447906643152237, "global_step": 117970, "epoch": 1325, "lr": 6.0361936050780446e-05}
{"train_loss": 0.11649510264396667, "global_step": 117971, "epoch": 1325, "lr": 6.036136889932863e-05}
{"train_loss": 0.10483702272176743, "global_step": 117972, "epoch": 1325, "lr": 6.0360801746483864e-05}
{"train_loss": 0.09049730002880096, "global_step": 117973, "epoch": 1325, "lr": 6.03602345922462e-05}
{"train_loss": 0.15409281849861145, "global_step": 117974, "epoch": 1325, "lr": 6.0359667436615744e-05}
{"train_loss": 0.0816076323390007, "global_step": 117975, "epoch": 1325, "lr": 6.035910027959255e-05}
{"train_loss": 0.14489758014678955, "global_step": 117976, "epoch": 1325, "lr": 6.035853312117671e-05}
{"train_loss": 0.19405144453048706, "global_step": 117977, "epoch": 1325, "lr": 6.035796596136829e-05}
{"train_loss": 0.14133474230766296, "global_step": 117978, "epoch": 1325, "lr": 6.035739880016737e-05}
{"train_loss": 0.11907047778367996, "global_step": 117979, "epoch": 1325, "lr": 6.035683163757401e-05}
{"train_loss": 0.0727895200252533, "global_step": 117980, "epoch": 1325, "lr": 6.035626447358831e-05}
{"train_loss": 0.06859996169805527, "global_step": 117981, "epoch": 1325, "lr": 6.035569730821033e-05}
{"train_loss": 0.14495903253555298, "global_step": 117982, "epoch": 1325, "lr": 6.035513014144015e-05}
{"train_loss": 0.11323723942041397, "global_step": 117983, "epoch": 1325, "lr": 6.0354562973277864e-05}
{"train_loss": 0.14085789024829865, "global_step": 117984, "epoch": 1325, "lr": 6.035399580372352e-05}
{"train_loss": 0.20687004923820496, "global_step": 117985, "epoch": 1325, "lr": 6.035342863277721e-05}
{"train_loss": 0.17945842444896698, "global_step": 117986, "epoch": 1325, "lr": 6.0352861460439e-05}
{"train_loss": 0.1222759485244751, "global_step": 117987, "epoch": 1325, "lr": 6.035229428670899e-05}
{"train_loss": 0.21504703164100647, "global_step": 117988, "epoch": 1325, "lr": 6.0351727111587234e-05}
{"train_loss": 0.19001708924770355, "global_step": 117989, "epoch": 1325, "lr": 6.035115993507381e-05}
{"train_loss": 0.13026268780231476, "global_step": 117990, "epoch": 1325, "lr": 6.03505927571688e-05}
{"train_loss": 0.09600944817066193, "global_step": 117991, "epoch": 1325, "lr": 6.035002557787228e-05}
{"train_loss": 0.11781103163957596, "global_step": 117992, "epoch": 1325, "lr": 6.0349458397184334e-05}
{"train_loss": 0.15661469101905823, "global_step": 117993, "epoch": 1325, "lr": 6.0348891215105016e-05}
{"train_loss": 0.15021459758281708, "global_step": 117994, "epoch": 1325, "lr": 6.0348324031634416e-05}
{"train_loss": 0.19917422533035278, "global_step": 117995, "epoch": 1325, "lr": 6.034775684677263e-05}
{"train_loss": 0.16880980134010315, "global_step": 117996, "epoch": 1325, "lr": 6.034718966051971e-05}
{"train_loss": 0.12345526367425919, "global_step": 117997, "epoch": 1325, "lr": 6.034662247287572e-05}
{"train_loss": 0.12009122967720032, "global_step": 117998, "epoch": 1325, "lr": 6.0346055283840766e-05}
{"train_loss": 0.1348264217376709, "global_step": 117999, "epoch": 1325, "lr": 6.03454880934149e-05}
{"train_loss": 0.16842782497406006, "global_step": 118000, "epoch": 1325, "lr": 6.0344920901598214e-05}
{"train_loss": 0.11730358749628067, "global_step": 118001, "epoch": 1325, "lr": 6.034435370839078e-05}
{"train_loss": 0.18764138221740723, "global_step": 118002, "epoch": 1325, "lr": 6.034378651379267e-05}
{"train_loss": 0.16589920222759247, "global_step": 118003, "epoch": 1325, "lr": 6.034321931780397e-05}
{"train_loss": 0.11595483124256134, "global_step": 118004, "epoch": 1325, "lr": 6.034265212042474e-05}
{"train_loss": 0.14064764976501465, "global_step": 118005, "epoch": 1325, "lr": 6.0342084921655075e-05}
{"train_loss": 0.182830810546875, "global_step": 118006, "epoch": 1325, "lr": 6.034151772149503e-05}
{"train_loss": 0.11329568177461624, "global_step": 118007, "epoch": 1325, "lr": 6.0340950519944704e-05}
{"train_loss": 0.09109866619110107, "global_step": 118008, "epoch": 1325, "lr": 6.034038331700417e-05}
{"train_loss": 0.14094167947769165, "global_step": 118009, "epoch": 1325, "lr": 6.0339816112673486e-05}
{"train_loss": 0.13945116102695465, "global_step": 118010, "epoch": 1325, "lr": 6.033924890695275e-05}
{"train_loss": 0.12369171530008316, "global_step": 118011, "epoch": 1325, "lr": 6.0338681699842026e-05}
{"train_loss": 0.23719827830791473, "global_step": 118012, "epoch": 1325, "lr": 6.033811449134139e-05}
{"train_loss": 0.13809727693206808, "global_step": 118013, "epoch": 1325, "lr": 6.033754728145093e-05, "val_loss": 5.263148307800293, "train_action_mse_error": 20.147661209106445}
{"train_loss": 0.1932433396577835, "global_step": 118014, "epoch": 1326, "lr": 6.033698007017069e-05}
{"train_loss": 0.15800030529499054, "global_step": 118015, "epoch": 1326, "lr": 6.033641285750079e-05}
{"train_loss": 0.12749534845352173, "global_step": 118016, "epoch": 1326, "lr": 6.033584564344128e-05}
{"train_loss": 0.11628025025129318, "global_step": 118017, "epoch": 1326, "lr": 6.033527842799225e-05}
{"train_loss": 0.0688977986574173, "global_step": 118018, "epoch": 1326, "lr": 6.033471121115375e-05}
{"train_loss": 0.07546357065439224, "global_step": 118019, "epoch": 1326, "lr": 6.0334143992925875e-05}
{"train_loss": 0.13472388684749603, "global_step": 118020, "epoch": 1326, "lr": 6.0333576773308706e-05}
{"train_loss": 0.05928942188620567, "global_step": 118021, "epoch": 1326, "lr": 6.033300955230232e-05}
{"train_loss": 0.07390353828668594, "global_step": 118022, "epoch": 1326, "lr": 6.033244232990677e-05}
{"train_loss": 0.17590227723121643, "global_step": 118023, "epoch": 1326, "lr": 6.033187510612216e-05}
{"train_loss": 0.16683582961559296, "global_step": 118024, "epoch": 1326, "lr": 6.033130788094855e-05}
{"train_loss": 0.11602228879928589, "global_step": 118025, "epoch": 1326, "lr": 6.033074065438603e-05}
{"train_loss": 0.10055235028266907, "global_step": 118026, "epoch": 1326, "lr": 6.0330173426434665e-05}
{"train_loss": 0.13010407984256744, "global_step": 118027, "epoch": 1326, "lr": 6.0329606197094533e-05}
{"train_loss": 0.17990104854106903, "global_step": 118028, "epoch": 1326, "lr": 6.0329038966365715e-05}
{"train_loss": 0.1774984896183014, "global_step": 118029, "epoch": 1326, "lr": 6.032847173424827e-05}
{"train_loss": 0.1130731999874115, "global_step": 118030, "epoch": 1326, "lr": 6.0327904500742304e-05}
{"train_loss": 0.13720469176769257, "global_step": 118031, "epoch": 1326, "lr": 6.032733726584787e-05}
{"train_loss": 0.21472737193107605, "global_step": 118032, "epoch": 1326, "lr": 6.032677002956505e-05}
{"train_loss": 0.12415079027414322, "global_step": 118033, "epoch": 1326, "lr": 6.0326202791893925e-05}
{"train_loss": 0.14606723189353943, "global_step": 118034, "epoch": 1326, "lr": 6.032563555283457e-05}
{"train_loss": 0.1146308034658432, "global_step": 118035, "epoch": 1326, "lr": 6.032506831238707e-05}
{"train_loss": 0.22505655884742737, "global_step": 118036, "epoch": 1326, "lr": 6.032450107055147e-05}
{"train_loss": 0.10048693418502808, "global_step": 118037, "epoch": 1326, "lr": 6.032393382732788e-05}
{"train_loss": 0.10597316920757294, "global_step": 118038, "epoch": 1326, "lr": 6.0323366582716356e-05}
{"train_loss": 0.08277998864650726, "global_step": 118039, "epoch": 1326, "lr": 6.0322799336716984e-05}
{"train_loss": 0.17918172478675842, "global_step": 118040, "epoch": 1326, "lr": 6.032223208932982e-05}
{"train_loss": 0.12738028168678284, "global_step": 118041, "epoch": 1326, "lr": 6.032166484055497e-05}
{"train_loss": 0.09987809509038925, "global_step": 118042, "epoch": 1326, "lr": 6.032109759039251e-05}
{"train_loss": 0.17240840196609497, "global_step": 118043, "epoch": 1326, "lr": 6.032053033884249e-05}
{"train_loss": 0.13024020195007324, "global_step": 118044, "epoch": 1326, "lr": 6.031996308590501e-05}
{"train_loss": 0.16567833721637726, "global_step": 118045, "epoch": 1326, "lr": 6.031939583158013e-05}
{"train_loss": 0.09427110105752945, "global_step": 118046, "epoch": 1326, "lr": 6.0318828575867925e-05}
{"train_loss": 0.12763048708438873, "global_step": 118047, "epoch": 1326, "lr": 6.03182613187685e-05}
{"train_loss": 0.19924227893352509, "global_step": 118048, "epoch": 1326, "lr": 6.0317694060281895e-05}
{"train_loss": 0.14727067947387695, "global_step": 118049, "epoch": 1326, "lr": 6.0317126800408205e-05}
{"train_loss": 0.13467924296855927, "global_step": 118050, "epoch": 1326, "lr": 6.031655953914751e-05}
{"train_loss": 0.09667917340993881, "global_step": 118051, "epoch": 1326, "lr": 6.0315992276499867e-05}
{"train_loss": 0.14931054413318634, "global_step": 118052, "epoch": 1326, "lr": 6.0315425012465374e-05}
{"train_loss": 0.10137338191270828, "global_step": 118053, "epoch": 1326, "lr": 6.031485774704409e-05}
{"train_loss": 0.15297859907150269, "global_step": 118054, "epoch": 1326, "lr": 6.031429048023612e-05}
{"train_loss": 0.1481531709432602, "global_step": 118055, "epoch": 1326, "lr": 6.0313723212041506e-05}
{"train_loss": 0.20462749898433685, "global_step": 118056, "epoch": 1326, "lr": 6.0313155942460344e-05}
{"train_loss": 0.12043686956167221, "global_step": 118057, "epoch": 1326, "lr": 6.0312588671492684e-05}
{"train_loss": 0.1570572704076767, "global_step": 118058, "epoch": 1326, "lr": 6.031202139913863e-05}
{"train_loss": 0.16464145481586456, "global_step": 118059, "epoch": 1326, "lr": 6.031145412539825e-05}
{"train_loss": 0.11234475672245026, "global_step": 118060, "epoch": 1326, "lr": 6.031088685027162e-05}
{"train_loss": 0.16716989874839783, "global_step": 118061, "epoch": 1326, "lr": 6.0310319573758824e-05}
{"train_loss": 0.12374734878540039, "global_step": 118062, "epoch": 1326, "lr": 6.030975229585993e-05}
{"train_loss": 0.12349396198987961, "global_step": 118063, "epoch": 1326, "lr": 6.030918501657502e-05}
{"train_loss": 0.15696042776107788, "global_step": 118064, "epoch": 1326, "lr": 6.030861773590416e-05}
{"train_loss": 0.1745014786720276, "global_step": 118065, "epoch": 1326, "lr": 6.030805045384743e-05}
{"train_loss": 0.12578506767749786, "global_step": 118066, "epoch": 1326, "lr": 6.030748317040491e-05}
{"train_loss": 0.11425328254699707, "global_step": 118067, "epoch": 1326, "lr": 6.030691588557668e-05}
{"train_loss": 0.13887912034988403, "global_step": 118068, "epoch": 1326, "lr": 6.03063485993628e-05}
{"train_loss": 0.08615657687187195, "global_step": 118069, "epoch": 1326, "lr": 6.030578131176337e-05}
{"train_loss": 0.18113137781620026, "global_step": 118070, "epoch": 1326, "lr": 6.0305214022778435e-05}
{"train_loss": 0.1125035360455513, "global_step": 118071, "epoch": 1326, "lr": 6.0304646732408096e-05}
{"train_loss": 0.11619424819946289, "global_step": 118072, "epoch": 1326, "lr": 6.030407944065244e-05}
{"train_loss": 0.12749320268630981, "global_step": 118073, "epoch": 1326, "lr": 6.030351214751152e-05}
{"train_loss": 0.18396899104118347, "global_step": 118074, "epoch": 1326, "lr": 6.030294485298541e-05}
{"train_loss": 0.09576451778411865, "global_step": 118075, "epoch": 1326, "lr": 6.0302377557074194e-05}
{"train_loss": 0.08064266294240952, "global_step": 118076, "epoch": 1326, "lr": 6.0301810259777944e-05}
{"train_loss": 0.20664873719215393, "global_step": 118077, "epoch": 1326, "lr": 6.030124296109675e-05}
{"train_loss": 0.1125970408320427, "global_step": 118078, "epoch": 1326, "lr": 6.030067566103068e-05}
{"train_loss": 0.18488116562366486, "global_step": 118079, "epoch": 1326, "lr": 6.03001083595798e-05}
{"train_loss": 0.140959233045578, "global_step": 118080, "epoch": 1326, "lr": 6.02995410567442e-05}
{"train_loss": 0.1286168098449707, "global_step": 118081, "epoch": 1326, "lr": 6.029897375252396e-05}
{"train_loss": 0.11452080309391022, "global_step": 118082, "epoch": 1326, "lr": 6.0298406446919155e-05}
{"train_loss": 0.1479441076517105, "global_step": 118083, "epoch": 1326, "lr": 6.029783913992983e-05}
{"train_loss": 0.10441184788942337, "global_step": 118084, "epoch": 1326, "lr": 6.0297271831556114e-05}
{"train_loss": 0.19658581912517548, "global_step": 118085, "epoch": 1326, "lr": 6.029670452179803e-05}
{"train_loss": 0.12292049080133438, "global_step": 118086, "epoch": 1326, "lr": 6.02961372106557e-05}
{"train_loss": 0.12943735718727112, "global_step": 118087, "epoch": 1326, "lr": 6.029556989812917e-05}
{"train_loss": 0.1334356665611267, "global_step": 118088, "epoch": 1326, "lr": 6.0295002584218515e-05}
{"train_loss": 0.12813012301921844, "global_step": 118089, "epoch": 1326, "lr": 6.0294435268923846e-05}
{"train_loss": 0.12572860717773438, "global_step": 118090, "epoch": 1326, "lr": 6.02938679522452e-05}
{"train_loss": 0.12700878083705902, "global_step": 118091, "epoch": 1326, "lr": 6.029330063418267e-05}
{"train_loss": 0.14076842367649078, "global_step": 118092, "epoch": 1326, "lr": 6.029273331473634e-05}
{"train_loss": 0.12742879986763, "global_step": 118093, "epoch": 1326, "lr": 6.029216599390628e-05}
{"train_loss": 0.12773418426513672, "global_step": 118094, "epoch": 1326, "lr": 6.029159867169254e-05}
{"train_loss": 0.12866075336933136, "global_step": 118095, "epoch": 1326, "lr": 6.029103134809524e-05}
{"train_loss": 0.13711929321289062, "global_step": 118096, "epoch": 1326, "lr": 6.0290464023114434e-05}
{"train_loss": 0.21152083575725555, "global_step": 118097, "epoch": 1326, "lr": 6.028989669675019e-05}
{"train_loss": 0.13902083039283752, "global_step": 118098, "epoch": 1326, "lr": 6.028932936900261e-05}
{"train_loss": 0.12004462629556656, "global_step": 118099, "epoch": 1326, "lr": 6.028876203987175e-05}
{"train_loss": 0.09897226840257645, "global_step": 118100, "epoch": 1326, "lr": 6.028819470935769e-05}
{"train_loss": 0.10553877055644989, "global_step": 118101, "epoch": 1326, "lr": 6.0287627377460496e-05}
{"train_loss": 0.13662651570492915, "global_step": 118102, "epoch": 1326, "lr": 6.0287060044180274e-05, "val_loss": 5.249800682067871}
{"train_loss": 0.08669386804103851, "global_step": 118103, "epoch": 1327, "lr": 6.028649270951707e-05}
{"train_loss": 0.17334133386611938, "global_step": 118104, "epoch": 1327, "lr": 6.028592537347099e-05}
{"train_loss": 0.05171789228916168, "global_step": 118105, "epoch": 1327, "lr": 6.0285358036042074e-05}
{"train_loss": 0.17932486534118652, "global_step": 118106, "epoch": 1327, "lr": 6.028479069723042e-05}
{"train_loss": 0.18498541414737701, "global_step": 118107, "epoch": 1327, "lr": 6.028422335703611e-05}
{"train_loss": 0.12956342101097107, "global_step": 118108, "epoch": 1327, "lr": 6.02836560154592e-05}
{"train_loss": 0.24237114191055298, "global_step": 118109, "epoch": 1327, "lr": 6.028308867249979e-05}
{"train_loss": 0.17266124486923218, "global_step": 118110, "epoch": 1327, "lr": 6.028252132815794e-05}
{"train_loss": 0.08867117017507553, "global_step": 118111, "epoch": 1327, "lr": 6.0281953982433727e-05}
{"train_loss": 0.13796469569206238, "global_step": 118112, "epoch": 1327, "lr": 6.0281386635327244e-05}
{"train_loss": 0.1296280175447464, "global_step": 118113, "epoch": 1327, "lr": 6.028081928683853e-05}
{"train_loss": 0.128570094704628, "global_step": 118114, "epoch": 1327, "lr": 6.028025193696769e-05}
{"train_loss": 0.10941404104232788, "global_step": 118115, "epoch": 1327, "lr": 6.02796845857148e-05}
{"train_loss": 0.18084712326526642, "global_step": 118116, "epoch": 1327, "lr": 6.027911723307994e-05}
{"train_loss": 0.10822980105876923, "global_step": 118117, "epoch": 1327, "lr": 6.0278549879063176e-05}
{"train_loss": 0.11553680896759033, "global_step": 118118, "epoch": 1327, "lr": 6.027798252366458e-05}
{"train_loss": 0.14918813109397888, "global_step": 118119, "epoch": 1327, "lr": 6.0277415166884235e-05}
{"train_loss": 0.1987657994031906, "global_step": 118120, "epoch": 1327, "lr": 6.027684780872222e-05}
{"train_loss": 0.11178672313690186, "global_step": 118121, "epoch": 1327, "lr": 6.02762804491786e-05}
{"train_loss": 0.19799791276454926, "global_step": 118122, "epoch": 1327, "lr": 6.027571308825347e-05}
{"train_loss": 0.09294986724853516, "global_step": 118123, "epoch": 1327, "lr": 6.027514572594689e-05}
{"train_loss": 0.2063913196325302, "global_step": 118124, "epoch": 1327, "lr": 6.0274578362258936e-05}
{"train_loss": 0.21219193935394287, "global_step": 118125, "epoch": 1327, "lr": 6.02740109971897e-05}
{"train_loss": 0.09233777225017548, "global_step": 118126, "epoch": 1327, "lr": 6.0273443630739244e-05}
{"train_loss": 0.13469254970550537, "global_step": 118127, "epoch": 1327, "lr": 6.0272876262907654e-05}
{"train_loss": 0.10826268047094345, "global_step": 118128, "epoch": 1327, "lr": 6.0272308893694994e-05}
{"train_loss": 0.11827850341796875, "global_step": 118129, "epoch": 1327, "lr": 6.027174152310136e-05}
{"train_loss": 0.11106356978416443, "global_step": 118130, "epoch": 1327, "lr": 6.027117415112681e-05}
{"train_loss": 0.15554726123809814, "global_step": 118131, "epoch": 1327, "lr": 6.0270606777771424e-05}
{"train_loss": 0.13729363679885864, "global_step": 118132, "epoch": 1327, "lr": 6.0270039403035275e-05}
{"train_loss": 0.1327374130487442, "global_step": 118133, "epoch": 1327, "lr": 6.0269472026918436e-05}
{"train_loss": 0.14080429077148438, "global_step": 118134, "epoch": 1327, "lr": 6.026890464942101e-05}
{"train_loss": 0.12115722894668579, "global_step": 118135, "epoch": 1327, "lr": 6.026833727054305e-05}
{"train_loss": 0.09153541177511215, "global_step": 118136, "epoch": 1327, "lr": 6.026776989028463e-05}
{"train_loss": 0.22271504998207092, "global_step": 118137, "epoch": 1327, "lr": 6.026720250864585e-05}
{"train_loss": 0.09829843044281006, "global_step": 118138, "epoch": 1327, "lr": 6.0266635125626755e-05}
{"train_loss": 0.09291946887969971, "global_step": 118139, "epoch": 1327, "lr": 6.026606774122744e-05}
{"train_loss": 0.09944972395896912, "global_step": 118140, "epoch": 1327, "lr": 6.026550035544798e-05}
{"train_loss": 0.12476643174886703, "global_step": 118141, "epoch": 1327, "lr": 6.0264932968288445e-05}
{"train_loss": 0.20058099925518036, "global_step": 118142, "epoch": 1327, "lr": 6.026436557974892e-05}
{"train_loss": 0.18771836161613464, "global_step": 118143, "epoch": 1327, "lr": 6.026379818982947e-05}
{"train_loss": 0.14949429035186768, "global_step": 118144, "epoch": 1327, "lr": 6.0263230798530165e-05}
{"train_loss": 0.1715298593044281, "global_step": 118145, "epoch": 1327, "lr": 6.0262663405851114e-05}
{"train_loss": 0.1450636088848114, "global_step": 118146, "epoch": 1327, "lr": 6.026209601179237e-05}
{"train_loss": 0.052029285579919815, "global_step": 118147, "epoch": 1327, "lr": 6.0261528616354016e-05}
{"train_loss": 0.11637841165065765, "global_step": 118148, "epoch": 1327, "lr": 6.0260961219536105e-05}
{"train_loss": 0.12533338367938995, "global_step": 118149, "epoch": 1327, "lr": 6.026039382133874e-05}
{"train_loss": 0.23580186069011688, "global_step": 118150, "epoch": 1327, "lr": 6.0259826421762e-05}
{"train_loss": 0.11886432021856308, "global_step": 118151, "epoch": 1327, "lr": 6.025925902080596e-05}
{"train_loss": 0.07564662396907806, "global_step": 118152, "epoch": 1327, "lr": 6.025869161847066e-05}
{"train_loss": 0.15241609513759613, "global_step": 118153, "epoch": 1327, "lr": 6.025812421475623e-05}
{"train_loss": 0.08378791809082031, "global_step": 118154, "epoch": 1327, "lr": 6.0257556809662704e-05}
{"train_loss": 0.14304785430431366, "global_step": 118155, "epoch": 1327, "lr": 6.0256989403190176e-05}
{"train_loss": 0.16008210182189941, "global_step": 118156, "epoch": 1327, "lr": 6.025642199533873e-05}
{"train_loss": 0.18887858092784882, "global_step": 118157, "epoch": 1327, "lr": 6.0255854586108427e-05}
{"train_loss": 0.17358233034610748, "global_step": 118158, "epoch": 1327, "lr": 6.0255287175499355e-05}
{"train_loss": 0.11766927689313889, "global_step": 118159, "epoch": 1327, "lr": 6.025471976351157e-05}
{"train_loss": 0.1868470311164856, "global_step": 118160, "epoch": 1327, "lr": 6.025415235014518e-05}
{"train_loss": 0.15083938837051392, "global_step": 118161, "epoch": 1327, "lr": 6.0253584935400234e-05}
{"train_loss": 0.20214514434337616, "global_step": 118162, "epoch": 1327, "lr": 6.025301751927682e-05}
{"train_loss": 0.10688232630491257, "global_step": 118163, "epoch": 1327, "lr": 6.0252450101775016e-05}
{"train_loss": 0.08946894854307175, "global_step": 118164, "epoch": 1327, "lr": 6.025188268289489e-05}
{"train_loss": 0.08142576366662979, "global_step": 118165, "epoch": 1327, "lr": 6.025131526263652e-05}
{"train_loss": 0.12232597917318344, "global_step": 118166, "epoch": 1327, "lr": 6.0250747841e-05}
{"train_loss": 0.1481667160987854, "global_step": 118167, "epoch": 1327, "lr": 6.025018041798538e-05}
{"train_loss": 0.09908445924520493, "global_step": 118168, "epoch": 1327, "lr": 6.0249612993592754e-05}
{"train_loss": 0.1596364676952362, "global_step": 118169, "epoch": 1327, "lr": 6.0249045567822185e-05}
{"train_loss": 0.13167165219783783, "global_step": 118170, "epoch": 1327, "lr": 6.024847814067376e-05}
{"train_loss": 0.0865209549665451, "global_step": 118171, "epoch": 1327, "lr": 6.024791071214756e-05}
{"train_loss": 0.12946100533008575, "global_step": 118172, "epoch": 1327, "lr": 6.024734328224364e-05}
{"train_loss": 0.14921189844608307, "global_step": 118173, "epoch": 1327, "lr": 6.02467758509621e-05}
{"train_loss": 0.09624384343624115, "global_step": 118174, "epoch": 1327, "lr": 6.024620841830301e-05}
{"train_loss": 0.12504056096076965, "global_step": 118175, "epoch": 1327, "lr": 6.024564098426643e-05}
{"train_loss": 0.17932984232902527, "global_step": 118176, "epoch": 1327, "lr": 6.024507354885245e-05}
{"train_loss": 0.15540923178195953, "global_step": 118177, "epoch": 1327, "lr": 6.024450611206115e-05}
{"train_loss": 0.15547972917556763, "global_step": 118178, "epoch": 1327, "lr": 6.0243938673892594e-05}
{"train_loss": 0.09154540300369263, "global_step": 118179, "epoch": 1327, "lr": 6.024337123434688e-05}
{"train_loss": 0.15248416364192963, "global_step": 118180, "epoch": 1327, "lr": 6.024280379342405e-05}
{"train_loss": 0.08178092539310455, "global_step": 118181, "epoch": 1327, "lr": 6.0242236351124216e-05}
{"train_loss": 0.20621207356452942, "global_step": 118182, "epoch": 1327, "lr": 6.024166890744743e-05}
{"train_loss": 0.13388429582118988, "global_step": 118183, "epoch": 1327, "lr": 6.024110146239378e-05}
{"train_loss": 0.1459571272134781, "global_step": 118184, "epoch": 1327, "lr": 6.024053401596333e-05}
{"train_loss": 0.12163595855236053, "global_step": 118185, "epoch": 1327, "lr": 6.023996656815618e-05}
{"train_loss": 0.11534301936626434, "global_step": 118186, "epoch": 1327, "lr": 6.0239399118972386e-05}
{"train_loss": 0.11622385680675507, "global_step": 118187, "epoch": 1327, "lr": 6.023883166841202e-05}
{"train_loss": 0.11280382424592972, "global_step": 118188, "epoch": 1327, "lr": 6.0238264216475184e-05}
{"train_loss": 0.12348341941833496, "global_step": 118189, "epoch": 1327, "lr": 6.023769676316192e-05}
{"train_loss": 0.12851887941360474, "global_step": 118190, "epoch": 1327, "lr": 6.023712930847234e-05}
{"train_loss": 0.13756788115990295, "global_step": 118191, "epoch": 1327, "lr": 6.023656185240649e-05, "val_loss": 5.169865608215332}
{"train_loss": 0.16417232155799866, "global_step": 118192, "epoch": 1328, "lr": 6.023599439496447e-05}
{"train_loss": 0.14406892657279968, "global_step": 118193, "epoch": 1328, "lr": 6.0235426936146336e-05}
{"train_loss": 0.14277106523513794, "global_step": 118194, "epoch": 1328, "lr": 6.023485947595218e-05}
{"train_loss": 0.14233475923538208, "global_step": 118195, "epoch": 1328, "lr": 6.023429201438206e-05}
{"train_loss": 0.15584789216518402, "global_step": 118196, "epoch": 1328, "lr": 6.023372455143608e-05}
{"train_loss": 0.12134774029254913, "global_step": 118197, "epoch": 1328, "lr": 6.023315708711429e-05}
{"train_loss": 0.09838160872459412, "global_step": 118198, "epoch": 1328, "lr": 6.023258962141678e-05}
{"train_loss": 0.13592125475406647, "global_step": 118199, "epoch": 1328, "lr": 6.0232022154343615e-05}
{"train_loss": 0.24367296695709229, "global_step": 118200, "epoch": 1328, "lr": 6.0231454685894885e-05}
{"train_loss": 0.17232027649879456, "global_step": 118201, "epoch": 1328, "lr": 6.023088721607066e-05}
{"train_loss": 0.14373038709163666, "global_step": 118202, "epoch": 1328, "lr": 6.023031974487102e-05}
{"train_loss": 0.1648373156785965, "global_step": 118203, "epoch": 1328, "lr": 6.022975227229603e-05}
{"train_loss": 0.06599371135234833, "global_step": 118204, "epoch": 1328, "lr": 6.02291847983458e-05}
{"train_loss": 0.11104191839694977, "global_step": 118205, "epoch": 1328, "lr": 6.0228617323020366e-05}
{"train_loss": 0.13136078417301178, "global_step": 118206, "epoch": 1328, "lr": 6.0228049846319814e-05}
{"train_loss": 0.14505375921726227, "global_step": 118207, "epoch": 1328, "lr": 6.022748236824422e-05}
{"train_loss": 0.22399123013019562, "global_step": 118208, "epoch": 1328, "lr": 6.022691488879367e-05}
{"train_loss": 0.1731371134519577, "global_step": 118209, "epoch": 1328, "lr": 6.022634740796824e-05}
{"train_loss": 0.12266132980585098, "global_step": 118210, "epoch": 1328, "lr": 6.022577992576801e-05}
{"train_loss": 0.13388079404830933, "global_step": 118211, "epoch": 1328, "lr": 6.0225212442193035e-05}
{"train_loss": 0.20695559680461884, "global_step": 118212, "epoch": 1328, "lr": 6.0224644957243406e-05}
{"train_loss": 0.18524913489818573, "global_step": 118213, "epoch": 1328, "lr": 6.0224077470919204e-05}
{"train_loss": 0.1878172755241394, "global_step": 118214, "epoch": 1328, "lr": 6.022350998322049e-05}
{"train_loss": 0.10276516526937485, "global_step": 118215, "epoch": 1328, "lr": 6.0222942494147363e-05}
{"train_loss": 0.13886049389839172, "global_step": 118216, "epoch": 1328, "lr": 6.022237500369988e-05}
{"train_loss": 0.15036112070083618, "global_step": 118217, "epoch": 1328, "lr": 6.022180751187812e-05}
{"train_loss": 0.09689022600650787, "global_step": 118218, "epoch": 1328, "lr": 6.022124001868217e-05}
{"train_loss": 0.09318030625581741, "global_step": 118219, "epoch": 1328, "lr": 6.02206725241121e-05}
{"train_loss": 0.14357294142246246, "global_step": 118220, "epoch": 1328, "lr": 6.0220105028167976e-05}
{"train_loss": 0.10672980546951294, "global_step": 118221, "epoch": 1328, "lr": 6.0219537530849876e-05}
{"train_loss": 0.12100423872470856, "global_step": 118222, "epoch": 1328, "lr": 6.0218970032157896e-05}
{"train_loss": 0.1729431301355362, "global_step": 118223, "epoch": 1328, "lr": 6.02184025320921e-05}
{"train_loss": 0.14141765236854553, "global_step": 118224, "epoch": 1328, "lr": 6.021783503065257e-05}
{"train_loss": 0.12114111334085464, "global_step": 118225, "epoch": 1328, "lr": 6.021726752783937e-05}
{"train_loss": 0.10832535475492477, "global_step": 118226, "epoch": 1328, "lr": 6.0216700023652584e-05}
{"train_loss": 0.12071892619132996, "global_step": 118227, "epoch": 1328, "lr": 6.0216132518092285e-05}
{"train_loss": 0.14379079639911652, "global_step": 118228, "epoch": 1328, "lr": 6.0215565011158545e-05}
{"train_loss": 0.09140526503324509, "global_step": 118229, "epoch": 1328, "lr": 6.0214997502851454e-05}
{"train_loss": 0.15727677941322327, "global_step": 118230, "epoch": 1328, "lr": 6.0214429993171084e-05}
{"train_loss": 0.16494762897491455, "global_step": 118231, "epoch": 1328, "lr": 6.021386248211751e-05}
{"train_loss": 0.1321469396352768, "global_step": 118232, "epoch": 1328, "lr": 6.0213294969690804e-05}
{"train_loss": 0.08150031417608261, "global_step": 118233, "epoch": 1328, "lr": 6.0212727455891036e-05}
{"train_loss": 0.16174566745758057, "global_step": 118234, "epoch": 1328, "lr": 6.02121599407183e-05}
{"train_loss": 0.14195884764194489, "global_step": 118235, "epoch": 1328, "lr": 6.021159242417267e-05}
{"train_loss": 0.09194870293140411, "global_step": 118236, "epoch": 1328, "lr": 6.02110249062542e-05}
{"train_loss": 0.12056691944599152, "global_step": 118237, "epoch": 1328, "lr": 6.0210457386962995e-05}
{"train_loss": 0.13887378573417664, "global_step": 118238, "epoch": 1328, "lr": 6.020988986629911e-05}
{"train_loss": 0.0935441255569458, "global_step": 118239, "epoch": 1328, "lr": 6.020932234426263e-05}
{"train_loss": 0.12593385577201843, "global_step": 118240, "epoch": 1328, "lr": 6.020875482085363e-05}
{"train_loss": 0.1410403996706009, "global_step": 118241, "epoch": 1328, "lr": 6.020818729607218e-05}
{"train_loss": 0.15121792256832123, "global_step": 118242, "epoch": 1328, "lr": 6.02076197699184e-05}
{"train_loss": 0.12469404935836792, "global_step": 118243, "epoch": 1328, "lr": 6.0207052242392305e-05}
{"train_loss": 0.12531018257141113, "global_step": 118244, "epoch": 1328, "lr": 6.020648471349399e-05}
{"train_loss": 0.17944197356700897, "global_step": 118245, "epoch": 1328, "lr": 6.0205917183223546e-05}
{"train_loss": 0.20356646180152893, "global_step": 118246, "epoch": 1328, "lr": 6.0205349651581035e-05}
{"train_loss": 0.18033407628536224, "global_step": 118247, "epoch": 1328, "lr": 6.020478211856655e-05}
{"train_loss": 0.16003090143203735, "global_step": 118248, "epoch": 1328, "lr": 6.020421458418014e-05}
{"train_loss": 0.15797409415245056, "global_step": 118249, "epoch": 1328, "lr": 6.020364704842191e-05}
{"train_loss": 0.11575597524642944, "global_step": 118250, "epoch": 1328, "lr": 6.0203079511291925e-05}
{"train_loss": 0.14338630437850952, "global_step": 118251, "epoch": 1328, "lr": 6.020251197279026e-05}
{"train_loss": 0.13235603272914886, "global_step": 118252, "epoch": 1328, "lr": 6.020194443291698e-05}
{"train_loss": 0.12413468956947327, "global_step": 118253, "epoch": 1328, "lr": 6.020137689167218e-05}
{"train_loss": 0.10356263816356659, "global_step": 118254, "epoch": 1328, "lr": 6.0200809349055934e-05}
{"train_loss": 0.14934971928596497, "global_step": 118255, "epoch": 1328, "lr": 6.020024180506831e-05}
{"train_loss": 0.17114003002643585, "global_step": 118256, "epoch": 1328, "lr": 6.019967425970938e-05}
{"train_loss": 0.14990678429603577, "global_step": 118257, "epoch": 1328, "lr": 6.019910671297924e-05}
{"train_loss": 0.11002080142498016, "global_step": 118258, "epoch": 1328, "lr": 6.019853916487794e-05}
{"train_loss": 0.14982527494430542, "global_step": 118259, "epoch": 1328, "lr": 6.019797161540558e-05}
{"train_loss": 0.12310841679573059, "global_step": 118260, "epoch": 1328, "lr": 6.0197404064562225e-05}
{"train_loss": 0.1453145295381546, "global_step": 118261, "epoch": 1328, "lr": 6.019683651234797e-05}
{"train_loss": 0.11461471766233444, "global_step": 118262, "epoch": 1328, "lr": 6.019626895876286e-05}
{"train_loss": 0.11179835349321365, "global_step": 118263, "epoch": 1328, "lr": 6.019570140380699e-05}
{"train_loss": 0.10228931903839111, "global_step": 118264, "epoch": 1328, "lr": 6.019513384748043e-05}
{"train_loss": 0.08592252433300018, "global_step": 118265, "epoch": 1328, "lr": 6.019456628978326e-05}
{"train_loss": 0.1444314569234848, "global_step": 118266, "epoch": 1328, "lr": 6.019399873071555e-05}
{"train_loss": 0.1512436717748642, "global_step": 118267, "epoch": 1328, "lr": 6.0193431170277384e-05}
{"train_loss": 0.13494634628295898, "global_step": 118268, "epoch": 1328, "lr": 6.019286360846883e-05}
{"train_loss": 0.08456873893737793, "global_step": 118269, "epoch": 1328, "lr": 6.0192296045289976e-05}
{"train_loss": 0.1234184056520462, "global_step": 118270, "epoch": 1328, "lr": 6.019172848074089e-05}
{"train_loss": 0.12437049299478531, "global_step": 118271, "epoch": 1328, "lr": 6.0191160914821656e-05}
{"train_loss": 0.08889798820018768, "global_step": 118272, "epoch": 1328, "lr": 6.019059334753233e-05}
{"train_loss": 0.1351112723350525, "global_step": 118273, "epoch": 1328, "lr": 6.019002577887302e-05}
{"train_loss": 0.15677398443222046, "global_step": 118274, "epoch": 1328, "lr": 6.0189458208843775e-05}
{"train_loss": 0.18312977254390717, "global_step": 118275, "epoch": 1328, "lr": 6.018889063744468e-05}
{"train_loss": 0.15838178992271423, "global_step": 118276, "epoch": 1328, "lr": 6.018832306467581e-05}
{"train_loss": 0.15775063633918762, "global_step": 118277, "epoch": 1328, "lr": 6.018775549053725e-05}
{"train_loss": 0.13603028655052185, "global_step": 118278, "epoch": 1328, "lr": 6.0187187915029076e-05}
{"train_loss": 0.11036499589681625, "global_step": 118279, "epoch": 1328, "lr": 6.018662033815135e-05}
{"train_loss": 0.13760115951299667, "global_step": 118280, "epoch": 1328, "lr": 6.018605275990417e-05, "val_loss": 5.343395709991455}
{"train_loss": 0.09285371005535126, "global_step": 118281, "epoch": 1329, "lr": 6.0185485180287596e-05}
{"train_loss": 0.12392506003379822, "global_step": 118282, "epoch": 1329, "lr": 6.0184917599301696e-05}
{"train_loss": 0.08141742646694183, "global_step": 118283, "epoch": 1329, "lr": 6.0184350016946564e-05}
{"train_loss": 0.10195418447256088, "global_step": 118284, "epoch": 1329, "lr": 6.018378243322227e-05}
{"train_loss": 0.20471327006816864, "global_step": 118285, "epoch": 1329, "lr": 6.01832148481289e-05}
{"train_loss": 0.16169458627700806, "global_step": 118286, "epoch": 1329, "lr": 6.01826472616665e-05}
{"train_loss": 0.1745709925889969, "global_step": 118287, "epoch": 1329, "lr": 6.018207967383519e-05}
{"train_loss": 0.1015571728348732, "global_step": 118288, "epoch": 1329, "lr": 6.0181512084635e-05}
{"train_loss": 0.15271295607089996, "global_step": 118289, "epoch": 1329, "lr": 6.018094449406605e-05}
{"train_loss": 0.156773179769516, "global_step": 118290, "epoch": 1329, "lr": 6.0180376902128387e-05}
{"train_loss": 0.15786415338516235, "global_step": 118291, "epoch": 1329, "lr": 6.01798093088221e-05}
{"train_loss": 0.11448132246732712, "global_step": 118292, "epoch": 1329, "lr": 6.017924171414726e-05}
{"train_loss": 0.0937558263540268, "global_step": 118293, "epoch": 1329, "lr": 6.017867411810394e-05}
{"train_loss": 0.1010218933224678, "global_step": 118294, "epoch": 1329, "lr": 6.017810652069222e-05}
{"train_loss": 0.15681597590446472, "global_step": 118295, "epoch": 1329, "lr": 6.017753892191218e-05}
{"train_loss": 0.07817980647087097, "global_step": 118296, "epoch": 1329, "lr": 6.01769713217639e-05}
{"train_loss": 0.13029366731643677, "global_step": 118297, "epoch": 1329, "lr": 6.017640372024744e-05}
{"train_loss": 0.17416085302829742, "global_step": 118298, "epoch": 1329, "lr": 6.0175836117362906e-05}
{"train_loss": 0.13611358404159546, "global_step": 118299, "epoch": 1329, "lr": 6.017526851311034e-05}
{"train_loss": 0.14834238588809967, "global_step": 118300, "epoch": 1329, "lr": 6.017470090748985e-05}
{"train_loss": 0.19550305604934692, "global_step": 118301, "epoch": 1329, "lr": 6.017413330050148e-05}
{"train_loss": 0.1896095722913742, "global_step": 118302, "epoch": 1329, "lr": 6.0173565692145316e-05}
{"train_loss": 0.08599496632814407, "global_step": 118303, "epoch": 1329, "lr": 6.017299808242145e-05}
{"train_loss": 0.06350374221801758, "global_step": 118304, "epoch": 1329, "lr": 6.0172430471329954e-05}
{"train_loss": 0.12027270346879959, "global_step": 118305, "epoch": 1329, "lr": 6.0171862858870885e-05}
{"train_loss": 0.14475645124912262, "global_step": 118306, "epoch": 1329, "lr": 6.0171295245044345e-05}
{"train_loss": 0.1185254454612732, "global_step": 118307, "epoch": 1329, "lr": 6.017072762985039e-05}
{"train_loss": 0.1198706328868866, "global_step": 118308, "epoch": 1329, "lr": 6.01701600132891e-05}
{"train_loss": 0.1549277901649475, "global_step": 118309, "epoch": 1329, "lr": 6.0169592395360575e-05}
{"train_loss": 0.06996126472949982, "global_step": 118310, "epoch": 1329, "lr": 6.016902477606485e-05}
{"train_loss": 0.10179876536130905, "global_step": 118311, "epoch": 1329, "lr": 6.016845715540204e-05}
{"train_loss": 0.16348783671855927, "global_step": 118312, "epoch": 1329, "lr": 6.016788953337218e-05}
{"train_loss": 0.12318820506334305, "global_step": 118313, "epoch": 1329, "lr": 6.0167321909975396e-05}
{"train_loss": 0.043582502752542496, "global_step": 118314, "epoch": 1329, "lr": 6.016675428521174e-05}
{"train_loss": 0.1888059824705124, "global_step": 118315, "epoch": 1329, "lr": 6.016618665908127e-05}
{"train_loss": 0.10024337470531464, "global_step": 118316, "epoch": 1329, "lr": 6.0165619031584094e-05}
{"train_loss": 0.16262270510196686, "global_step": 118317, "epoch": 1329, "lr": 6.0165051402720273e-05}
{"train_loss": 0.13877931237220764, "global_step": 118318, "epoch": 1329, "lr": 6.016448377248988e-05}
{"train_loss": 0.1446683406829834, "global_step": 118319, "epoch": 1329, "lr": 6.0163916140893006e-05}
{"train_loss": 0.2451484203338623, "global_step": 118320, "epoch": 1329, "lr": 6.0163348507929716e-05}
{"train_loss": 0.11161546409130096, "global_step": 118321, "epoch": 1329, "lr": 6.016278087360008e-05}
{"train_loss": 0.13874128460884094, "global_step": 118322, "epoch": 1329, "lr": 6.0162213237904186e-05}
{"train_loss": 0.18174369633197784, "global_step": 118323, "epoch": 1329, "lr": 6.016164560084211e-05}
{"train_loss": 0.13296379148960114, "global_step": 118324, "epoch": 1329, "lr": 6.016107796241391e-05}
{"train_loss": 0.1603357493877411, "global_step": 118325, "epoch": 1329, "lr": 6.016051032261969e-05}
{"train_loss": 0.11873799562454224, "global_step": 118326, "epoch": 1329, "lr": 6.0159942681459515e-05}
{"train_loss": 0.1396651268005371, "global_step": 118327, "epoch": 1329, "lr": 6.015937503893345e-05}
{"train_loss": 0.1830206960439682, "global_step": 118328, "epoch": 1329, "lr": 6.015880739504158e-05}
{"train_loss": 0.16732144355773926, "global_step": 118329, "epoch": 1329, "lr": 6.0158239749783986e-05}
{"train_loss": 0.1798626333475113, "global_step": 118330, "epoch": 1329, "lr": 6.015767210316073e-05}
{"train_loss": 0.1807556003332138, "global_step": 118331, "epoch": 1329, "lr": 6.015710445517191e-05}
{"train_loss": 0.1653376966714859, "global_step": 118332, "epoch": 1329, "lr": 6.0156536805817596e-05}
{"train_loss": 0.18475578725337982, "global_step": 118333, "epoch": 1329, "lr": 6.015596915509786e-05}
{"train_loss": 0.10892440378665924, "global_step": 118334, "epoch": 1329, "lr": 6.0155401503012765e-05}
{"train_loss": 0.1709708869457245, "global_step": 118335, "epoch": 1329, "lr": 6.015483384956241e-05}
{"train_loss": 0.11729642748832703, "global_step": 118336, "epoch": 1329, "lr": 6.0154266194746855e-05}
{"train_loss": 0.18394161760807037, "global_step": 118337, "epoch": 1329, "lr": 6.015369853856618e-05}
{"train_loss": 0.1091093197464943, "global_step": 118338, "epoch": 1329, "lr": 6.0153130881020484e-05}
{"train_loss": 0.1730569452047348, "global_step": 118339, "epoch": 1329, "lr": 6.0152563222109816e-05}
{"train_loss": 0.1582547277212143, "global_step": 118340, "epoch": 1329, "lr": 6.015199556183425e-05}
{"train_loss": 0.08392176032066345, "global_step": 118341, "epoch": 1329, "lr": 6.015142790019388e-05}
{"train_loss": 0.20141273736953735, "global_step": 118342, "epoch": 1329, "lr": 6.0150860237188766e-05}
{"train_loss": 0.1611454337835312, "global_step": 118343, "epoch": 1329, "lr": 6.0150292572819e-05}
{"train_loss": 0.09288974851369858, "global_step": 118344, "epoch": 1329, "lr": 6.0149724907084646e-05}
{"train_loss": 0.18113268911838531, "global_step": 118345, "epoch": 1329, "lr": 6.0149157239985785e-05}
{"train_loss": 0.17006216943264008, "global_step": 118346, "epoch": 1329, "lr": 6.0148589571522496e-05}
{"train_loss": 0.16002358496189117, "global_step": 118347, "epoch": 1329, "lr": 6.014802190169485e-05}
{"train_loss": 0.15170860290527344, "global_step": 118348, "epoch": 1329, "lr": 6.014745423050292e-05}
{"train_loss": 0.09358525276184082, "global_step": 118349, "epoch": 1329, "lr": 6.014688655794679e-05}
{"train_loss": 0.13109081983566284, "global_step": 118350, "epoch": 1329, "lr": 6.014631888402654e-05}
{"train_loss": 0.14887958765029907, "global_step": 118351, "epoch": 1329, "lr": 6.014575120874224e-05}
{"train_loss": 0.11576610058546066, "global_step": 118352, "epoch": 1329, "lr": 6.0145183532093976e-05}
{"train_loss": 0.22780048847198486, "global_step": 118353, "epoch": 1329, "lr": 6.014461585408181e-05}
{"train_loss": 0.17023396492004395, "global_step": 118354, "epoch": 1329, "lr": 6.014404817470583e-05}
{"train_loss": 0.20431746542453766, "global_step": 118355, "epoch": 1329, "lr": 6.01434804939661e-05}
{"train_loss": 0.08168265223503113, "global_step": 118356, "epoch": 1329, "lr": 6.014291281186269e-05}
{"train_loss": 0.11117740720510483, "global_step": 118357, "epoch": 1329, "lr": 6.0142345128395705e-05}
{"train_loss": 0.19754858314990997, "global_step": 118358, "epoch": 1329, "lr": 6.014177744356522e-05}
{"train_loss": 0.102780282497406, "global_step": 118359, "epoch": 1329, "lr": 6.0141209757371275e-05}
{"train_loss": 0.16717392206192017, "global_step": 118360, "epoch": 1329, "lr": 6.0140642069813966e-05}
{"train_loss": 0.15302613377571106, "global_step": 118361, "epoch": 1329, "lr": 6.0140074380893375e-05}
{"train_loss": 0.11934521049261093, "global_step": 118362, "epoch": 1329, "lr": 6.013950669060958e-05}
{"train_loss": 0.11107431352138519, "global_step": 118363, "epoch": 1329, "lr": 6.013893899896264e-05}
{"train_loss": 0.11462809890508652, "global_step": 118364, "epoch": 1329, "lr": 6.013837130595266e-05}
{"train_loss": 0.10894543677568436, "global_step": 118365, "epoch": 1329, "lr": 6.013780361157969e-05}
{"train_loss": 0.13053977489471436, "global_step": 118366, "epoch": 1329, "lr": 6.013723591584382e-05}
{"train_loss": 0.14303681254386902, "global_step": 118367, "epoch": 1329, "lr": 6.013666821874511e-05}
{"train_loss": 0.13389834761619568, "global_step": 118368, "epoch": 1329, "lr": 6.013610052028365e-05}
{"train_loss": 0.14132774424519431, "global_step": 118369, "epoch": 1329, "lr": 6.013553282045952e-05, "val_loss": 5.31579065322876}
{"train_loss": 0.18300457298755646, "global_step": 118370, "epoch": 1330, "lr": 6.013496511927279e-05}
{"train_loss": 0.0952678993344307, "global_step": 118371, "epoch": 1330, "lr": 6.0134397416723544e-05}
{"train_loss": 0.09499683976173401, "global_step": 118372, "epoch": 1330, "lr": 6.013382971281185e-05}
{"train_loss": 0.15951992571353912, "global_step": 118373, "epoch": 1330, "lr": 6.0133262007537785e-05}
{"train_loss": 0.20427103340625763, "global_step": 118374, "epoch": 1330, "lr": 6.0132694300901426e-05}
{"train_loss": 0.1323196440935135, "global_step": 118375, "epoch": 1330, "lr": 6.013212659290285e-05}
{"train_loss": 0.12827211618423462, "global_step": 118376, "epoch": 1330, "lr": 6.0131558883542124e-05}
{"train_loss": 0.21171528100967407, "global_step": 118377, "epoch": 1330, "lr": 6.013099117281935e-05}
{"train_loss": 0.08242078870534897, "global_step": 118378, "epoch": 1330, "lr": 6.0130423460734585e-05}
{"train_loss": 0.12398400157690048, "global_step": 118379, "epoch": 1330, "lr": 6.012985574728791e-05}
{"train_loss": 0.09837247431278229, "global_step": 118380, "epoch": 1330, "lr": 6.012928803247939e-05}
{"train_loss": 0.2369718998670578, "global_step": 118381, "epoch": 1330, "lr": 6.012872031630912e-05}
{"train_loss": 0.08163657039403915, "global_step": 118382, "epoch": 1330, "lr": 6.012815259877715e-05}
{"train_loss": 0.11959923058748245, "global_step": 118383, "epoch": 1330, "lr": 6.012758487988359e-05}
{"train_loss": 0.13322001695632935, "global_step": 118384, "epoch": 1330, "lr": 6.01270171596285e-05}
{"train_loss": 0.18503980338573456, "global_step": 118385, "epoch": 1330, "lr": 6.012644943801193e-05}
{"train_loss": 0.15802909433841705, "global_step": 118386, "epoch": 1330, "lr": 6.0125881715034e-05}
{"train_loss": 0.10264459997415543, "global_step": 118387, "epoch": 1330, "lr": 6.0125313990694764e-05}
{"train_loss": 0.13362756371498108, "global_step": 118388, "epoch": 1330, "lr": 6.0124746264994315e-05}
{"train_loss": 0.13333867490291595, "global_step": 118389, "epoch": 1330, "lr": 6.0124178537932704e-05}
{"train_loss": 0.18470227718353271, "global_step": 118390, "epoch": 1330, "lr": 6.0123610809510034e-05}
{"train_loss": 0.11793061345815659, "global_step": 118391, "epoch": 1330, "lr": 6.012304307972636e-05}
{"train_loss": 0.14052757620811462, "global_step": 118392, "epoch": 1330, "lr": 6.012247534858178e-05}
{"train_loss": 0.1916162222623825, "global_step": 118393, "epoch": 1330, "lr": 6.012190761607633e-05}
{"train_loss": 0.0953962579369545, "global_step": 118394, "epoch": 1330, "lr": 6.012133988221013e-05}
{"train_loss": 0.11172931641340256, "global_step": 118395, "epoch": 1330, "lr": 6.012077214698324e-05}
{"train_loss": 0.10869713872671127, "global_step": 118396, "epoch": 1330, "lr": 6.012020441039573e-05}
{"train_loss": 0.1489698737859726, "global_step": 118397, "epoch": 1330, "lr": 6.011963667244769e-05}
{"train_loss": 0.11274034529924393, "global_step": 118398, "epoch": 1330, "lr": 6.011906893313919e-05}
{"train_loss": 0.11981910467147827, "global_step": 118399, "epoch": 1330, "lr": 6.011850119247029e-05}
{"train_loss": 0.1630239486694336, "global_step": 118400, "epoch": 1330, "lr": 6.011793345044109e-05}
{"train_loss": 0.20493726432323456, "global_step": 118401, "epoch": 1330, "lr": 6.011736570705165e-05}
{"train_loss": 0.15463455021381378, "global_step": 118402, "epoch": 1330, "lr": 6.011679796230206e-05}
{"train_loss": 0.1367688626050949, "global_step": 118403, "epoch": 1330, "lr": 6.011623021619238e-05}
{"train_loss": 0.15539732575416565, "global_step": 118404, "epoch": 1330, "lr": 6.01156624687227e-05}
{"train_loss": 0.12645447254180908, "global_step": 118405, "epoch": 1330, "lr": 6.011509471989309e-05}
{"train_loss": 0.11410629004240036, "global_step": 118406, "epoch": 1330, "lr": 6.0114526969703646e-05}
{"train_loss": 0.11239949613809586, "global_step": 118407, "epoch": 1330, "lr": 6.0113959218154415e-05}
{"train_loss": 0.13330502808094025, "global_step": 118408, "epoch": 1330, "lr": 6.0113391465245486e-05}
{"train_loss": 0.17670825123786926, "global_step": 118409, "epoch": 1330, "lr": 6.011282371097694e-05}
{"train_loss": 0.1258499175310135, "global_step": 118410, "epoch": 1330, "lr": 6.011225595534884e-05}
{"train_loss": 0.10840112715959549, "global_step": 118411, "epoch": 1330, "lr": 6.011168819836127e-05}
{"train_loss": 0.11571629345417023, "global_step": 118412, "epoch": 1330, "lr": 6.011112044001431e-05}
{"train_loss": 0.13069875538349152, "global_step": 118413, "epoch": 1330, "lr": 6.011055268030804e-05}
{"train_loss": 0.12135587632656097, "global_step": 118414, "epoch": 1330, "lr": 6.010998491924252e-05}
{"train_loss": 0.1623881459236145, "global_step": 118415, "epoch": 1330, "lr": 6.010941715681784e-05}
{"train_loss": 0.16188566386699677, "global_step": 118416, "epoch": 1330, "lr": 6.0108849393034086e-05}
{"train_loss": 0.11436622589826584, "global_step": 118417, "epoch": 1330, "lr": 6.0108281627891304e-05}
{"train_loss": 0.10414686799049377, "global_step": 118418, "epoch": 1330, "lr": 6.010771386138959e-05}
{"train_loss": 0.1533229500055313, "global_step": 118419, "epoch": 1330, "lr": 6.010714609352901e-05}
{"train_loss": 0.11628586798906326, "global_step": 118420, "epoch": 1330, "lr": 6.010657832430966e-05}
{"train_loss": 0.1776646077632904, "global_step": 118421, "epoch": 1330, "lr": 6.0106010553731585e-05}
{"train_loss": 0.11151403933763504, "global_step": 118422, "epoch": 1330, "lr": 6.0105442781794894e-05}
{"train_loss": 0.07294801622629166, "global_step": 118423, "epoch": 1330, "lr": 6.0104875008499636e-05}
{"train_loss": 0.11517452448606491, "global_step": 118424, "epoch": 1330, "lr": 6.010430723384593e-05}
{"train_loss": 0.0989457443356514, "global_step": 118425, "epoch": 1330, "lr": 6.0103739457833795e-05}
{"train_loss": 0.1854618936777115, "global_step": 118426, "epoch": 1330, "lr": 6.010317168046335e-05}
{"train_loss": 0.0913819670677185, "global_step": 118427, "epoch": 1330, "lr": 6.0102603901734644e-05}
{"train_loss": 0.13417986035346985, "global_step": 118428, "epoch": 1330, "lr": 6.0102036121647775e-05}
{"train_loss": 0.10783304274082184, "global_step": 118429, "epoch": 1330, "lr": 6.0101468340202814e-05}
{"train_loss": 0.1610933244228363, "global_step": 118430, "epoch": 1330, "lr": 6.010090055739983e-05}
{"train_loss": 0.142014741897583, "global_step": 118431, "epoch": 1330, "lr": 6.01003327732389e-05}
{"train_loss": 0.11717052757740021, "global_step": 118432, "epoch": 1330, "lr": 6.009976498772011e-05}
{"train_loss": 0.09947729110717773, "global_step": 118433, "epoch": 1330, "lr": 6.009919720084351e-05}
{"train_loss": 0.11696089059114456, "global_step": 118434, "epoch": 1330, "lr": 6.009862941260922e-05}
{"train_loss": 0.1366512030363083, "global_step": 118435, "epoch": 1330, "lr": 6.009806162301729e-05}
{"train_loss": 0.1439286321401596, "global_step": 118436, "epoch": 1330, "lr": 6.00974938320678e-05}
{"train_loss": 0.11484476923942566, "global_step": 118437, "epoch": 1330, "lr": 6.009692603976081e-05}
{"train_loss": 0.10075898468494415, "global_step": 118438, "epoch": 1330, "lr": 6.009635824609643e-05}
{"train_loss": 0.13913512229919434, "global_step": 118439, "epoch": 1330, "lr": 6.0095790451074685e-05}
{"train_loss": 0.11690288782119751, "global_step": 118440, "epoch": 1330, "lr": 6.009522265469571e-05}
{"train_loss": 0.15074709057807922, "global_step": 118441, "epoch": 1330, "lr": 6.009465485695956e-05}
{"train_loss": 0.0959627702832222, "global_step": 118442, "epoch": 1330, "lr": 6.009408705786629e-05}
{"train_loss": 0.09180814027786255, "global_step": 118443, "epoch": 1330, "lr": 6.0093519257416e-05}
{"train_loss": 0.10817597061395645, "global_step": 118444, "epoch": 1330, "lr": 6.009295145560876e-05}
{"train_loss": 0.15211926400661469, "global_step": 118445, "epoch": 1330, "lr": 6.009238365244464e-05}
{"train_loss": 0.1885843724012375, "global_step": 118446, "epoch": 1330, "lr": 6.009181584792373e-05}
{"train_loss": 0.17340297996997833, "global_step": 118447, "epoch": 1330, "lr": 6.00912480420461e-05}
{"train_loss": 0.17175041139125824, "global_step": 118448, "epoch": 1330, "lr": 6.009068023481181e-05}
{"train_loss": 0.13139988481998444, "global_step": 118449, "epoch": 1330, "lr": 6.009011242622097e-05}
{"train_loss": 0.11479414999485016, "global_step": 118450, "epoch": 1330, "lr": 6.008954461627363e-05}
{"train_loss": 0.1683199554681778, "global_step": 118451, "epoch": 1330, "lr": 6.0088976804969875e-05}
{"train_loss": 0.15899868309497833, "global_step": 118452, "epoch": 1330, "lr": 6.008840899230978e-05}
{"train_loss": 0.11960893124341965, "global_step": 118453, "epoch": 1330, "lr": 6.008784117829341e-05}
{"train_loss": 0.09366177022457123, "global_step": 118454, "epoch": 1330, "lr": 6.008727336292086e-05}
{"train_loss": 0.1415892392396927, "global_step": 118455, "epoch": 1330, "lr": 6.008670554619221e-05}
{"train_loss": 0.09588280320167542, "global_step": 118456, "epoch": 1330, "lr": 6.008613772810752e-05}
{"train_loss": 0.15236935019493103, "global_step": 118457, "epoch": 1330, "lr": 6.008556990866686e-05}
{"train_loss": 0.1341442937931318, "global_step": 118458, "epoch": 1330, "lr": 6.008500208787032e-05, "val_loss": 5.3204264640808105, "train_action_mse_error": 19.22600555419922}
{"train_loss": 0.13164114952087402, "global_step": 118459, "epoch": 1331, "lr": 6.008443426571798e-05}
{"train_loss": 0.12480685114860535, "global_step": 118460, "epoch": 1331, "lr": 6.008386644220991e-05}
{"train_loss": 0.17445512115955353, "global_step": 118461, "epoch": 1331, "lr": 6.00832986173462e-05}
{"train_loss": 0.17010141909122467, "global_step": 118462, "epoch": 1331, "lr": 6.00827307911269e-05}
{"train_loss": 0.11693350970745087, "global_step": 118463, "epoch": 1331, "lr": 6.00821629635521e-05}
{"train_loss": 0.2042807787656784, "global_step": 118464, "epoch": 1331, "lr": 6.008159513462187e-05}
{"train_loss": 0.09059665352106094, "global_step": 118465, "epoch": 1331, "lr": 6.00810273043363e-05}
{"train_loss": 0.1363084465265274, "global_step": 118466, "epoch": 1331, "lr": 6.008045947269546e-05}
{"train_loss": 0.12494338303804398, "global_step": 118467, "epoch": 1331, "lr": 6.0079891639699423e-05}
{"train_loss": 0.12458135932683945, "global_step": 118468, "epoch": 1331, "lr": 6.0079323805348276e-05}
{"train_loss": 0.13447205722332, "global_step": 118469, "epoch": 1331, "lr": 6.007875596964207e-05}
{"train_loss": 0.12664100527763367, "global_step": 118470, "epoch": 1331, "lr": 6.0078188132580915e-05}
{"train_loss": 0.10500369966030121, "global_step": 118471, "epoch": 1331, "lr": 6.007762029416486e-05}
{"train_loss": 0.10251660645008087, "global_step": 118472, "epoch": 1331, "lr": 6.007705245439399e-05}
{"train_loss": 0.16466738283634186, "global_step": 118473, "epoch": 1331, "lr": 6.007648461326839e-05}
{"train_loss": 0.2908518612384796, "global_step": 118474, "epoch": 1331, "lr": 6.0075916770788146e-05}
{"train_loss": 0.11167830973863602, "global_step": 118475, "epoch": 1331, "lr": 6.007534892695328e-05}
{"train_loss": 0.13773486018180847, "global_step": 118476, "epoch": 1331, "lr": 6.007478108176392e-05}
{"train_loss": 0.0889570564031601, "global_step": 118477, "epoch": 1331, "lr": 6.007421323522013e-05}
{"train_loss": 0.15694570541381836, "global_step": 118478, "epoch": 1331, "lr": 6.0073645387322e-05}
{"train_loss": 0.09710000455379486, "global_step": 118479, "epoch": 1331, "lr": 6.007307753806958e-05}
{"train_loss": 0.07607638090848923, "global_step": 118480, "epoch": 1331, "lr": 6.0072509687462954e-05}
{"train_loss": 0.1603546440601349, "global_step": 118481, "epoch": 1331, "lr": 6.00719418355022e-05}
{"train_loss": 0.08192849904298782, "global_step": 118482, "epoch": 1331, "lr": 6.007137398218741e-05}
{"train_loss": 0.149606391787529, "global_step": 118483, "epoch": 1331, "lr": 6.007080612751863e-05}
{"train_loss": 0.07986702769994736, "global_step": 118484, "epoch": 1331, "lr": 6.007023827149596e-05}
{"train_loss": 0.12114202976226807, "global_step": 118485, "epoch": 1331, "lr": 6.006967041411947e-05}
{"train_loss": 0.20464859902858734, "global_step": 118486, "epoch": 1331, "lr": 6.0069102555389233e-05}
{"train_loss": 0.12602931261062622, "global_step": 118487, "epoch": 1331, "lr": 6.0068534695305334e-05}
{"train_loss": 0.07929246127605438, "global_step": 118488, "epoch": 1331, "lr": 6.006796683386784e-05}
{"train_loss": 0.130776509642601, "global_step": 118489, "epoch": 1331, "lr": 6.0067398971076824e-05}
{"train_loss": 0.15257540345191956, "global_step": 118490, "epoch": 1331, "lr": 6.006683110693238e-05}
{"train_loss": 0.15184451639652252, "global_step": 118491, "epoch": 1331, "lr": 6.006626324143456e-05}
{"train_loss": 0.13979217410087585, "global_step": 118492, "epoch": 1331, "lr": 6.0065695374583465e-05}
{"train_loss": 0.14809629321098328, "global_step": 118493, "epoch": 1331, "lr": 6.006512750637916e-05}
{"train_loss": 0.13193614780902863, "global_step": 118494, "epoch": 1331, "lr": 6.006455963682171e-05}
{"train_loss": 0.1043303832411766, "global_step": 118495, "epoch": 1331, "lr": 6.006399176591121e-05}
{"train_loss": 0.13035835325717926, "global_step": 118496, "epoch": 1331, "lr": 6.006342389364773e-05}
{"train_loss": 0.08264872431755066, "global_step": 118497, "epoch": 1331, "lr": 6.0062856020031343e-05}
{"train_loss": 0.16041415929794312, "global_step": 118498, "epoch": 1331, "lr": 6.006228814506213e-05}
{"train_loss": 0.1569175273180008, "global_step": 118499, "epoch": 1331, "lr": 6.0061720268740165e-05}
{"train_loss": 0.16078095138072968, "global_step": 118500, "epoch": 1331, "lr": 6.0061152391065526e-05}
{"train_loss": 0.09616287797689438, "global_step": 118501, "epoch": 1331, "lr": 6.006058451203829e-05}
{"train_loss": 0.17505204677581787, "global_step": 118502, "epoch": 1331, "lr": 6.0060016631658525e-05}
{"train_loss": 0.21534688770771027, "global_step": 118503, "epoch": 1331, "lr": 6.0059448749926306e-05}
{"train_loss": 0.10840313136577606, "global_step": 118504, "epoch": 1331, "lr": 6.005888086684173e-05}
{"train_loss": 0.1132877916097641, "global_step": 118505, "epoch": 1331, "lr": 6.005831298240485e-05}
{"train_loss": 0.09133557230234146, "global_step": 118506, "epoch": 1331, "lr": 6.005774509661576e-05}
{"train_loss": 0.13369086384773254, "global_step": 118507, "epoch": 1331, "lr": 6.005717720947452e-05}
{"train_loss": 0.12703149020671844, "global_step": 118508, "epoch": 1331, "lr": 6.0056609320981226e-05}
{"train_loss": 0.17014585435390472, "global_step": 118509, "epoch": 1331, "lr": 6.005604143113594e-05}
{"train_loss": 0.10539638996124268, "global_step": 118510, "epoch": 1331, "lr": 6.005547353993875e-05}
{"train_loss": 0.06614965200424194, "global_step": 118511, "epoch": 1331, "lr": 6.00549056473897e-05}
{"train_loss": 0.15745985507965088, "global_step": 118512, "epoch": 1331, "lr": 6.005433775348891e-05}
{"train_loss": 0.1249072477221489, "global_step": 118513, "epoch": 1331, "lr": 6.005376985823643e-05}
{"train_loss": 0.07525108009576797, "global_step": 118514, "epoch": 1331, "lr": 6.005320196163236e-05}
{"train_loss": 0.15534649789333344, "global_step": 118515, "epoch": 1331, "lr": 6.005263406367674e-05}
{"train_loss": 0.16519732773303986, "global_step": 118516, "epoch": 1331, "lr": 6.005206616436967e-05}
{"train_loss": 0.14963985979557037, "global_step": 118517, "epoch": 1331, "lr": 6.005149826371121e-05}
{"train_loss": 0.09579741954803467, "global_step": 118518, "epoch": 1331, "lr": 6.005093036170147e-05}
{"train_loss": 0.14005142450332642, "global_step": 118519, "epoch": 1331, "lr": 6.0050362458340505e-05}
{"train_loss": 0.11169355362653732, "global_step": 118520, "epoch": 1331, "lr": 6.004979455362838e-05}
{"train_loss": 0.16718725860118866, "global_step": 118521, "epoch": 1331, "lr": 6.0049226647565184e-05}
{"train_loss": 0.1647641509771347, "global_step": 118522, "epoch": 1331, "lr": 6.004865874015099e-05}
{"train_loss": 0.14696499705314636, "global_step": 118523, "epoch": 1331, "lr": 6.004809083138588e-05}
{"train_loss": 0.15517418086528778, "global_step": 118524, "epoch": 1331, "lr": 6.004752292126993e-05}
{"train_loss": 0.102805495262146, "global_step": 118525, "epoch": 1331, "lr": 6.004695500980321e-05}
{"train_loss": 0.09600766748189926, "global_step": 118526, "epoch": 1331, "lr": 6.004638709698579e-05}
{"train_loss": 0.12688151001930237, "global_step": 118527, "epoch": 1331, "lr": 6.004581918281776e-05}
{"train_loss": 0.13093699514865875, "global_step": 118528, "epoch": 1331, "lr": 6.004525126729921e-05}
{"train_loss": 0.1687411665916443, "global_step": 118529, "epoch": 1331, "lr": 6.004468335043017e-05}
{"train_loss": 0.15367397665977478, "global_step": 118530, "epoch": 1331, "lr": 6.004411543221077e-05}
{"train_loss": 0.09789425879716873, "global_step": 118531, "epoch": 1331, "lr": 6.0043547512641054e-05}
{"train_loss": 0.11619589477777481, "global_step": 118532, "epoch": 1331, "lr": 6.004297959172109e-05}
{"train_loss": 0.12375851720571518, "global_step": 118533, "epoch": 1331, "lr": 6.0042411669450995e-05}
{"train_loss": 0.21303194761276245, "global_step": 118534, "epoch": 1331, "lr": 6.004184374583081e-05}
{"train_loss": 0.2729836404323578, "global_step": 118535, "epoch": 1331, "lr": 6.004127582086061e-05}
{"train_loss": 0.10363466292619705, "global_step": 118536, "epoch": 1331, "lr": 6.004070789454049e-05}
{"train_loss": 0.12012864649295807, "global_step": 118537, "epoch": 1331, "lr": 6.0040139966870526e-05}
{"train_loss": 0.1293327808380127, "global_step": 118538, "epoch": 1331, "lr": 6.0039572037850776e-05}
{"train_loss": 0.17231567203998566, "global_step": 118539, "epoch": 1331, "lr": 6.0039004107481335e-05}
{"train_loss": 0.15612202882766724, "global_step": 118540, "epoch": 1331, "lr": 6.003843617576227e-05}
{"train_loss": 0.17184384167194366, "global_step": 118541, "epoch": 1331, "lr": 6.0037868242693665e-05}
{"train_loss": 0.22870205342769623, "global_step": 118542, "epoch": 1331, "lr": 6.003730030827559e-05}
{"train_loss": 0.12956665456295013, "global_step": 118543, "epoch": 1331, "lr": 6.003673237250812e-05}
{"train_loss": 0.15533193945884705, "global_step": 118544, "epoch": 1331, "lr": 6.003616443539133e-05}
{"train_loss": 0.15630729496479034, "global_step": 118545, "epoch": 1331, "lr": 6.0035596496925296e-05}
{"train_loss": 0.09960932284593582, "global_step": 118546, "epoch": 1331, "lr": 6.0035028557110104e-05}
{"train_loss": 0.13661733738492043, "global_step": 118547, "epoch": 1331, "lr": 6.003446061594583e-05, "val_loss": 5.335994243621826}
{"train_loss": 0.07765769213438034, "global_step": 118548, "epoch": 1332, "lr": 6.003389267343255e-05}
{"train_loss": 0.12886084616184235, "global_step": 118549, "epoch": 1332, "lr": 6.003332472957032e-05}
{"train_loss": 0.20324978232383728, "global_step": 118550, "epoch": 1332, "lr": 6.0032756784359237e-05}
{"train_loss": 0.11080022156238556, "global_step": 118551, "epoch": 1332, "lr": 6.003218883779937e-05}
{"train_loss": 0.21236269176006317, "global_step": 118552, "epoch": 1332, "lr": 6.0031620889890806e-05}
{"train_loss": 0.19843566417694092, "global_step": 118553, "epoch": 1332, "lr": 6.00310529406336e-05}
{"train_loss": 0.16445159912109375, "global_step": 118554, "epoch": 1332, "lr": 6.003048499002786e-05}
{"train_loss": 0.12111659348011017, "global_step": 118555, "epoch": 1332, "lr": 6.002991703807362e-05}
{"train_loss": 0.11226465553045273, "global_step": 118556, "epoch": 1332, "lr": 6.0029349084771e-05}
{"train_loss": 0.1544751524925232, "global_step": 118557, "epoch": 1332, "lr": 6.002878113012005e-05}
{"train_loss": 0.1260748654603958, "global_step": 118558, "epoch": 1332, "lr": 6.0028213174120854e-05}
{"train_loss": 0.1124441921710968, "global_step": 118559, "epoch": 1332, "lr": 6.002764521677349e-05}
{"train_loss": 0.2037627398967743, "global_step": 118560, "epoch": 1332, "lr": 6.0027077258078014e-05}
{"train_loss": 0.1530573070049286, "global_step": 118561, "epoch": 1332, "lr": 6.002650929803453e-05}
{"train_loss": 0.14401084184646606, "global_step": 118562, "epoch": 1332, "lr": 6.002594133664311e-05}
{"train_loss": 0.08395091444253922, "global_step": 118563, "epoch": 1332, "lr": 6.002537337390381e-05}
{"train_loss": 0.15177536010742188, "global_step": 118564, "epoch": 1332, "lr": 6.002480540981674e-05}
{"train_loss": 0.20206891000270844, "global_step": 118565, "epoch": 1332, "lr": 6.0024237444381946e-05}
{"train_loss": 0.08626550436019897, "global_step": 118566, "epoch": 1332, "lr": 6.002366947759952e-05}
{"train_loss": 0.1352376639842987, "global_step": 118567, "epoch": 1332, "lr": 6.002310150946953e-05}
{"train_loss": 0.11155930161476135, "global_step": 118568, "epoch": 1332, "lr": 6.0022533539992056e-05}
{"train_loss": 0.12396885454654694, "global_step": 118569, "epoch": 1332, "lr": 6.002196556916717e-05}
{"train_loss": 0.21504971385002136, "global_step": 118570, "epoch": 1332, "lr": 6.002139759699497e-05}
{"train_loss": 0.09681235253810883, "global_step": 118571, "epoch": 1332, "lr": 6.002082962347549e-05}
{"train_loss": 0.14913144707679749, "global_step": 118572, "epoch": 1332, "lr": 6.002026164860886e-05}
{"train_loss": 0.13141192495822906, "global_step": 118573, "epoch": 1332, "lr": 6.001969367239511e-05}
{"train_loss": 0.21553336083889008, "global_step": 118574, "epoch": 1332, "lr": 6.001912569483433e-05}
{"train_loss": 0.17137129604816437, "global_step": 118575, "epoch": 1332, "lr": 6.001855771592662e-05}
{"train_loss": 0.15558214485645294, "global_step": 118576, "epoch": 1332, "lr": 6.001798973567201e-05}
{"train_loss": 0.1365618258714676, "global_step": 118577, "epoch": 1332, "lr": 6.0017421754070624e-05}
{"train_loss": 0.08265340328216553, "global_step": 118578, "epoch": 1332, "lr": 6.0016853771122515e-05}
{"train_loss": 0.15294194221496582, "global_step": 118579, "epoch": 1332, "lr": 6.001628578682775e-05}
{"train_loss": 0.10547731071710587, "global_step": 118580, "epoch": 1332, "lr": 6.001571780118643e-05}
{"train_loss": 0.11135976761579514, "global_step": 118581, "epoch": 1332, "lr": 6.0015149814198625e-05}
{"train_loss": 0.1631559431552887, "global_step": 118582, "epoch": 1332, "lr": 6.001458182586439e-05}
{"train_loss": 0.24310241639614105, "global_step": 118583, "epoch": 1332, "lr": 6.0014013836183825e-05}
{"train_loss": 0.2022515833377838, "global_step": 118584, "epoch": 1332, "lr": 6.001344584515698e-05}
{"train_loss": 0.18433046340942383, "global_step": 118585, "epoch": 1332, "lr": 6.0012877852783975e-05}
{"train_loss": 0.15302503108978271, "global_step": 118586, "epoch": 1332, "lr": 6.001230985906485e-05}
{"train_loss": 0.12356457859277725, "global_step": 118587, "epoch": 1332, "lr": 6.0011741863999694e-05}
{"train_loss": 0.16680465638637543, "global_step": 118588, "epoch": 1332, "lr": 6.001117386758858e-05}
{"train_loss": 0.12966524064540863, "global_step": 118589, "epoch": 1332, "lr": 6.001060586983158e-05}
{"train_loss": 0.10678084194660187, "global_step": 118590, "epoch": 1332, "lr": 6.0010037870728786e-05}
{"train_loss": 0.1069856584072113, "global_step": 118591, "epoch": 1332, "lr": 6.000946987028025e-05}
{"train_loss": 0.11758183687925339, "global_step": 118592, "epoch": 1332, "lr": 6.000890186848608e-05}
{"train_loss": 0.09107465296983719, "global_step": 118593, "epoch": 1332, "lr": 6.0008333865346324e-05}
{"train_loss": 0.26600682735443115, "global_step": 118594, "epoch": 1332, "lr": 6.0007765860861074e-05}
{"train_loss": 0.17921148240566254, "global_step": 118595, "epoch": 1332, "lr": 6.0007197855030404e-05}
{"train_loss": 0.22574582695960999, "global_step": 118596, "epoch": 1332, "lr": 6.000662984785439e-05}
{"train_loss": 0.09249156713485718, "global_step": 118597, "epoch": 1332, "lr": 6.0006061839333094e-05}
{"train_loss": 0.17131803929805756, "global_step": 118598, "epoch": 1332, "lr": 6.0005493829466616e-05}
{"train_loss": 0.17653794586658478, "global_step": 118599, "epoch": 1332, "lr": 6.000492581825502e-05}
{"train_loss": 0.1378878653049469, "global_step": 118600, "epoch": 1332, "lr": 6.0004357805698376e-05}
{"train_loss": 0.10699767619371414, "global_step": 118601, "epoch": 1332, "lr": 6.000378979179677e-05}
{"train_loss": 0.13509665429592133, "global_step": 118602, "epoch": 1332, "lr": 6.000322177655028e-05}
{"train_loss": 0.1520155966281891, "global_step": 118603, "epoch": 1332, "lr": 6.000265375995897e-05}
{"train_loss": 0.08808986842632294, "global_step": 118604, "epoch": 1332, "lr": 6.000208574202294e-05}
{"train_loss": 0.1703011691570282, "global_step": 118605, "epoch": 1332, "lr": 6.000151772274225e-05}
{"train_loss": 0.14370213449001312, "global_step": 118606, "epoch": 1332, "lr": 6.0000949702116985e-05}
{"train_loss": 0.13548113405704498, "global_step": 118607, "epoch": 1332, "lr": 6.0000381680147186e-05}
{"train_loss": 0.09337521344423294, "global_step": 118608, "epoch": 1332, "lr": 5.999981365683298e-05}
{"train_loss": 0.15755456686019897, "global_step": 118609, "epoch": 1332, "lr": 5.999924563217442e-05}
{"train_loss": 0.08769048750400543, "global_step": 118610, "epoch": 1332, "lr": 5.999867760617157e-05}
{"train_loss": 0.17188161611557007, "global_step": 118611, "epoch": 1332, "lr": 5.9998109578824534e-05}
{"train_loss": 0.13932037353515625, "global_step": 118612, "epoch": 1332, "lr": 5.9997541550133366e-05}
{"train_loss": 0.18878193199634552, "global_step": 118613, "epoch": 1332, "lr": 5.999697352009815e-05}
{"train_loss": 0.2062154859304428, "global_step": 118614, "epoch": 1332, "lr": 5.999640548871896e-05}
{"train_loss": 0.09908653050661087, "global_step": 118615, "epoch": 1332, "lr": 5.9995837455995885e-05}
{"train_loss": 0.14896920323371887, "global_step": 118616, "epoch": 1332, "lr": 5.999526942192899e-05}
{"train_loss": 0.12007397413253784, "global_step": 118617, "epoch": 1332, "lr": 5.9994701386518345e-05}
{"train_loss": 0.12153424322605133, "global_step": 118618, "epoch": 1332, "lr": 5.999413334976404e-05}
{"train_loss": 0.17920902371406555, "global_step": 118619, "epoch": 1332, "lr": 5.9993565311666144e-05}
{"train_loss": 0.08929352462291718, "global_step": 118620, "epoch": 1332, "lr": 5.999299727222474e-05}
{"train_loss": 0.07594065368175507, "global_step": 118621, "epoch": 1332, "lr": 5.9992429231439885e-05}
{"train_loss": 0.15585197508335114, "global_step": 118622, "epoch": 1332, "lr": 5.999186118931167e-05}
{"train_loss": 0.11770730465650558, "global_step": 118623, "epoch": 1332, "lr": 5.999129314584019e-05}
{"train_loss": 0.16984973847866058, "global_step": 118624, "epoch": 1332, "lr": 5.99907251010255e-05}
{"train_loss": 0.13828225433826447, "global_step": 118625, "epoch": 1332, "lr": 5.9990157054867666e-05}
{"train_loss": 0.08270531892776489, "global_step": 118626, "epoch": 1332, "lr": 5.9989589007366783e-05}
{"train_loss": 0.14837516844272614, "global_step": 118627, "epoch": 1332, "lr": 5.9989020958522926e-05}
{"train_loss": 0.089505635201931, "global_step": 118628, "epoch": 1332, "lr": 5.9988452908336156e-05}
{"train_loss": 0.0755145251750946, "global_step": 118629, "epoch": 1332, "lr": 5.9987884856806574e-05}
{"train_loss": 0.11092622578144073, "global_step": 118630, "epoch": 1332, "lr": 5.9987316803934235e-05}
{"train_loss": 0.13032251596450806, "global_step": 118631, "epoch": 1332, "lr": 5.998674874971923e-05}
{"train_loss": 0.1380169689655304, "global_step": 118632, "epoch": 1332, "lr": 5.998618069416161e-05}
{"train_loss": 0.13849279284477234, "global_step": 118633, "epoch": 1332, "lr": 5.998561263726148e-05}
{"train_loss": 0.19781236350536346, "global_step": 118634, "epoch": 1332, "lr": 5.998504457901891e-05}
{"train_loss": 0.13596951961517334, "global_step": 118635, "epoch": 1332, "lr": 5.9984476519433965e-05}
{"train_loss": 0.1428108751941263, "global_step": 118636, "epoch": 1332, "lr": 5.998390845850673e-05, "val_loss": 5.32334566116333}
{"train_loss": 0.1784020960330963, "global_step": 118637, "epoch": 1333, "lr": 5.9983340396237284e-05}
{"train_loss": 0.09364941716194153, "global_step": 118638, "epoch": 1333, "lr": 5.9982772332625706e-05}
{"train_loss": 0.19633840024471283, "global_step": 118639, "epoch": 1333, "lr": 5.998220426767205e-05}
{"train_loss": 0.16195805370807648, "global_step": 118640, "epoch": 1333, "lr": 5.998163620137641e-05}
{"train_loss": 0.12175501883029938, "global_step": 118641, "epoch": 1333, "lr": 5.998106813373887e-05}
{"train_loss": 0.08562752604484558, "global_step": 118642, "epoch": 1333, "lr": 5.998050006475949e-05}
{"train_loss": 0.11185453832149506, "global_step": 118643, "epoch": 1333, "lr": 5.9979931994438374e-05}
{"train_loss": 0.14369972050189972, "global_step": 118644, "epoch": 1333, "lr": 5.9979363922775564e-05}
{"train_loss": 0.07653564214706421, "global_step": 118645, "epoch": 1333, "lr": 5.997879584977115e-05}
{"train_loss": 0.20904338359832764, "global_step": 118646, "epoch": 1333, "lr": 5.99782277754252e-05}
{"train_loss": 0.10510803759098053, "global_step": 118647, "epoch": 1333, "lr": 5.997765969973781e-05}
{"train_loss": 0.13352371752262115, "global_step": 118648, "epoch": 1333, "lr": 5.997709162270904e-05}
{"train_loss": 0.07956081628799438, "global_step": 118649, "epoch": 1333, "lr": 5.997652354433897e-05}
{"train_loss": 0.14537687599658966, "global_step": 118650, "epoch": 1333, "lr": 5.9975955464627685e-05}
{"train_loss": 0.12206237018108368, "global_step": 118651, "epoch": 1333, "lr": 5.9975387383575244e-05}
{"train_loss": 0.16346850991249084, "global_step": 118652, "epoch": 1333, "lr": 5.997481930118174e-05}
{"train_loss": 0.13131015002727509, "global_step": 118653, "epoch": 1333, "lr": 5.9974251217447254e-05}
{"train_loss": 0.11616423726081848, "global_step": 118654, "epoch": 1333, "lr": 5.997368313237184e-05}
{"train_loss": 0.09188896417617798, "global_step": 118655, "epoch": 1333, "lr": 5.9973115045955574e-05}
{"train_loss": 0.22349223494529724, "global_step": 118656, "epoch": 1333, "lr": 5.997254695819856e-05}
{"train_loss": 0.11352511495351791, "global_step": 118657, "epoch": 1333, "lr": 5.997197886910084e-05}
{"train_loss": 0.11298809945583344, "global_step": 118658, "epoch": 1333, "lr": 5.997141077866253e-05}
{"train_loss": 0.11360936611890793, "global_step": 118659, "epoch": 1333, "lr": 5.9970842686883676e-05}
{"train_loss": 0.17115160822868347, "global_step": 118660, "epoch": 1333, "lr": 5.9970274593764376e-05}
{"train_loss": 0.11032657325267792, "global_step": 118661, "epoch": 1333, "lr": 5.996970649930468e-05}
{"train_loss": 0.15124432742595673, "global_step": 118662, "epoch": 1333, "lr": 5.996913840350469e-05}
{"train_loss": 0.14426520466804504, "global_step": 118663, "epoch": 1333, "lr": 5.996857030636447e-05}
{"train_loss": 0.12204135954380035, "global_step": 118664, "epoch": 1333, "lr": 5.996800220788409e-05}
{"train_loss": 0.21466544270515442, "global_step": 118665, "epoch": 1333, "lr": 5.9967434108063634e-05}
{"train_loss": 0.11464545875787735, "global_step": 118666, "epoch": 1333, "lr": 5.996686600690318e-05}
{"train_loss": 0.14875824749469757, "global_step": 118667, "epoch": 1333, "lr": 5.99662979044028e-05}
{"train_loss": 0.151840478181839, "global_step": 118668, "epoch": 1333, "lr": 5.996572980056258e-05}
{"train_loss": 0.1815667301416397, "global_step": 118669, "epoch": 1333, "lr": 5.9965161695382574e-05}
{"train_loss": 0.17036548256874084, "global_step": 118670, "epoch": 1333, "lr": 5.9964593588862895e-05}
{"train_loss": 0.10255610942840576, "global_step": 118671, "epoch": 1333, "lr": 5.996402548100358e-05}
{"train_loss": 0.16655191779136658, "global_step": 118672, "epoch": 1333, "lr": 5.996345737180473e-05}
{"train_loss": 0.1591445952653885, "global_step": 118673, "epoch": 1333, "lr": 5.996288926126641e-05}
{"train_loss": 0.1937810778617859, "global_step": 118674, "epoch": 1333, "lr": 5.996232114938871e-05}
{"train_loss": 0.14725199341773987, "global_step": 118675, "epoch": 1333, "lr": 5.996175303617167e-05}
{"train_loss": 0.09655600786209106, "global_step": 118676, "epoch": 1333, "lr": 5.996118492161542e-05}
{"train_loss": 0.16687218844890594, "global_step": 118677, "epoch": 1333, "lr": 5.996061680572001e-05}
{"train_loss": 0.12214410305023193, "global_step": 118678, "epoch": 1333, "lr": 5.9960048688485517e-05}
{"train_loss": 0.15096844732761383, "global_step": 118679, "epoch": 1333, "lr": 5.995948056991201e-05}
{"train_loss": 0.13777142763137817, "global_step": 118680, "epoch": 1333, "lr": 5.9958912449999574e-05}
{"train_loss": 0.08056806027889252, "global_step": 118681, "epoch": 1333, "lr": 5.9958344328748285e-05}
{"train_loss": 0.181122288107872, "global_step": 118682, "epoch": 1333, "lr": 5.995777620615822e-05}
{"train_loss": 0.09219399094581604, "global_step": 118683, "epoch": 1333, "lr": 5.9957208082229455e-05}
{"train_loss": 0.15865682065486908, "global_step": 118684, "epoch": 1333, "lr": 5.995663995696206e-05}
{"train_loss": 0.15514886379241943, "global_step": 118685, "epoch": 1333, "lr": 5.995607183035612e-05}
{"train_loss": 0.13158583641052246, "global_step": 118686, "epoch": 1333, "lr": 5.9955503702411694e-05}
{"train_loss": 0.1632869988679886, "global_step": 118687, "epoch": 1333, "lr": 5.995493557312889e-05}
{"train_loss": 0.1543673723936081, "global_step": 118688, "epoch": 1333, "lr": 5.995436744250775e-05}
{"train_loss": 0.11095137894153595, "global_step": 118689, "epoch": 1333, "lr": 5.995379931054837e-05}
{"train_loss": 0.17949232459068298, "global_step": 118690, "epoch": 1333, "lr": 5.995323117725083e-05}
{"train_loss": 0.191680446267128, "global_step": 118691, "epoch": 1333, "lr": 5.9952663042615195e-05}
{"train_loss": 0.09705250710248947, "global_step": 118692, "epoch": 1333, "lr": 5.995209490664155e-05}
{"train_loss": 0.13613134622573853, "global_step": 118693, "epoch": 1333, "lr": 5.9951526769329954e-05}
{"train_loss": 0.09014927595853806, "global_step": 118694, "epoch": 1333, "lr": 5.9950958630680506e-05}
{"train_loss": 0.12459894269704819, "global_step": 118695, "epoch": 1333, "lr": 5.995039049069329e-05}
{"train_loss": 0.10686787962913513, "global_step": 118696, "epoch": 1333, "lr": 5.994982234936835e-05}
{"train_loss": 0.10465723276138306, "global_step": 118697, "epoch": 1333, "lr": 5.994925420670576e-05}
{"train_loss": 0.13684247434139252, "global_step": 118698, "epoch": 1333, "lr": 5.994868606270564e-05}
{"train_loss": 0.13402803242206573, "global_step": 118699, "epoch": 1333, "lr": 5.994811791736803e-05}
{"train_loss": 0.1687973290681839, "global_step": 118700, "epoch": 1333, "lr": 5.9947549770693014e-05}
{"train_loss": 0.10895431786775589, "global_step": 118701, "epoch": 1333, "lr": 5.994698162268069e-05}
{"train_loss": 0.14682994782924652, "global_step": 118702, "epoch": 1333, "lr": 5.99464134733311e-05}
{"train_loss": 0.14390775561332703, "global_step": 118703, "epoch": 1333, "lr": 5.994584532264433e-05}
{"train_loss": 0.11737259477376938, "global_step": 118704, "epoch": 1333, "lr": 5.9945277170620476e-05}
{"train_loss": 0.07490304112434387, "global_step": 118705, "epoch": 1333, "lr": 5.994470901725959e-05}
{"train_loss": 0.2098996788263321, "global_step": 118706, "epoch": 1333, "lr": 5.994414086256177e-05}
{"train_loss": 0.12063480168581009, "global_step": 118707, "epoch": 1333, "lr": 5.994357270652707e-05}
{"train_loss": 0.20080338418483734, "global_step": 118708, "epoch": 1333, "lr": 5.994300454915558e-05}
{"train_loss": 0.16275092959403992, "global_step": 118709, "epoch": 1333, "lr": 5.9942436390447375e-05}
{"train_loss": 0.11171644181013107, "global_step": 118710, "epoch": 1333, "lr": 5.9941868230402545e-05}
{"train_loss": 0.08830371499061584, "global_step": 118711, "epoch": 1333, "lr": 5.9941300069021125e-05}
{"train_loss": 0.14567598700523376, "global_step": 118712, "epoch": 1333, "lr": 5.994073190630323e-05}
{"train_loss": 0.1349140703678131, "global_step": 118713, "epoch": 1333, "lr": 5.9940163742248925e-05}
{"train_loss": 0.11565090715885162, "global_step": 118714, "epoch": 1333, "lr": 5.993959557685829e-05}
{"train_loss": 0.1095556765794754, "global_step": 118715, "epoch": 1333, "lr": 5.99390274101314e-05}
{"train_loss": 0.10723517835140228, "global_step": 118716, "epoch": 1333, "lr": 5.9938459242068325e-05}
{"train_loss": 0.12969106435775757, "global_step": 118717, "epoch": 1333, "lr": 5.993789107266914e-05}
{"train_loss": 0.1223001778125763, "global_step": 118718, "epoch": 1333, "lr": 5.9937322901933936e-05}
{"train_loss": 0.11560197174549103, "global_step": 118719, "epoch": 1333, "lr": 5.993675472986278e-05}
{"train_loss": 0.1093282401561737, "global_step": 118720, "epoch": 1333, "lr": 5.993618655645573e-05}
{"train_loss": 0.1503506302833557, "global_step": 118721, "epoch": 1333, "lr": 5.993561838171291e-05}
{"train_loss": 0.1347944289445877, "global_step": 118722, "epoch": 1333, "lr": 5.9935050205634356e-05}
{"train_loss": 0.2010194957256317, "global_step": 118723, "epoch": 1333, "lr": 5.9934482028220154e-05}
{"train_loss": 0.2015460580587387, "global_step": 118724, "epoch": 1333, "lr": 5.993391384947038e-05}
{"train_loss": 0.137529528375422, "global_step": 118725, "epoch": 1333, "lr": 5.993334566938511e-05, "val_loss": 5.14486837387085}
{"train_loss": 0.1684957593679428, "global_step": 118726, "epoch": 1334, "lr": 5.993277748796442e-05}
{"train_loss": 0.08659385144710541, "global_step": 118727, "epoch": 1334, "lr": 5.99322093052084e-05}
{"train_loss": 0.13370324671268463, "global_step": 118728, "epoch": 1334, "lr": 5.993164112111711e-05}
{"train_loss": 0.1250227391719818, "global_step": 118729, "epoch": 1334, "lr": 5.9931072935690615e-05}
{"train_loss": 0.14551502466201782, "global_step": 118730, "epoch": 1334, "lr": 5.993050474892903e-05}
{"train_loss": 0.11520086228847504, "global_step": 118731, "epoch": 1334, "lr": 5.99299365608324e-05}
{"train_loss": 0.11967097967863083, "global_step": 118732, "epoch": 1334, "lr": 5.9929368371400815e-05}
{"train_loss": 0.15142285823822021, "global_step": 118733, "epoch": 1334, "lr": 5.992880018063435e-05}
{"train_loss": 0.18295539915561676, "global_step": 118734, "epoch": 1334, "lr": 5.992823198853308e-05}
{"train_loss": 0.12855549156665802, "global_step": 118735, "epoch": 1334, "lr": 5.992766379509708e-05}
{"train_loss": 0.15971754491329193, "global_step": 118736, "epoch": 1334, "lr": 5.992709560032642e-05}
{"train_loss": 0.160162091255188, "global_step": 118737, "epoch": 1334, "lr": 5.9926527404221175e-05}
{"train_loss": 0.1116170883178711, "global_step": 118738, "epoch": 1334, "lr": 5.992595920678145e-05}
{"train_loss": 0.08633431792259216, "global_step": 118739, "epoch": 1334, "lr": 5.992539100800729e-05}
{"train_loss": 0.07673126459121704, "global_step": 118740, "epoch": 1334, "lr": 5.99248228078988e-05}
{"train_loss": 0.14609891176223755, "global_step": 118741, "epoch": 1334, "lr": 5.992425460645602e-05}
{"train_loss": 0.18182949721813202, "global_step": 118742, "epoch": 1334, "lr": 5.992368640367905e-05}
{"train_loss": 0.10170332342386246, "global_step": 118743, "epoch": 1334, "lr": 5.9923118199567954e-05}
{"train_loss": 0.14411978423595428, "global_step": 118744, "epoch": 1334, "lr": 5.992254999412282e-05}
{"train_loss": 0.11585022509098053, "global_step": 118745, "epoch": 1334, "lr": 5.9921981787343716e-05}
{"train_loss": 0.10175195336341858, "global_step": 118746, "epoch": 1334, "lr": 5.9921413579230736e-05}
{"train_loss": 0.14348357915878296, "global_step": 118747, "epoch": 1334, "lr": 5.992084536978393e-05}
{"train_loss": 0.15845242142677307, "global_step": 118748, "epoch": 1334, "lr": 5.992027715900338e-05}
{"train_loss": 0.09235816448926926, "global_step": 118749, "epoch": 1334, "lr": 5.9919708946889184e-05}
{"train_loss": 0.1102861762046814, "global_step": 118750, "epoch": 1334, "lr": 5.99191407334414e-05}
{"train_loss": 0.10148852318525314, "global_step": 118751, "epoch": 1334, "lr": 5.991857251866011e-05}
{"train_loss": 0.1383785456418991, "global_step": 118752, "epoch": 1334, "lr": 5.9918004302545385e-05}
{"train_loss": 0.08429784327745438, "global_step": 118753, "epoch": 1334, "lr": 5.991743608509731e-05}
{"train_loss": 0.12027911096811295, "global_step": 118754, "epoch": 1334, "lr": 5.991686786631595e-05}
{"train_loss": 0.11946837604045868, "global_step": 118755, "epoch": 1334, "lr": 5.99162996462014e-05}
{"train_loss": 0.10865437984466553, "global_step": 118756, "epoch": 1334, "lr": 5.991573142475371e-05}
{"train_loss": 0.10038145631551743, "global_step": 118757, "epoch": 1334, "lr": 5.991516320197298e-05}
{"train_loss": 0.14649344980716705, "global_step": 118758, "epoch": 1334, "lr": 5.9914594977859274e-05}
{"train_loss": 0.09694478660821915, "global_step": 118759, "epoch": 1334, "lr": 5.9914026752412686e-05}
{"train_loss": 0.11174669861793518, "global_step": 118760, "epoch": 1334, "lr": 5.991345852563326e-05}
{"train_loss": 0.1767318695783615, "global_step": 118761, "epoch": 1334, "lr": 5.9912890297521095e-05}
{"train_loss": 0.11217361688613892, "global_step": 118762, "epoch": 1334, "lr": 5.991232206807627e-05}
{"train_loss": 0.11267837882041931, "global_step": 118763, "epoch": 1334, "lr": 5.991175383729885e-05}
{"train_loss": 0.08469817787408829, "global_step": 118764, "epoch": 1334, "lr": 5.99111856051889e-05}
{"train_loss": 0.24032172560691833, "global_step": 118765, "epoch": 1334, "lr": 5.9910617371746526e-05}
{"train_loss": 0.1516406089067459, "global_step": 118766, "epoch": 1334, "lr": 5.991004913697178e-05}
{"train_loss": 0.12221375107765198, "global_step": 118767, "epoch": 1334, "lr": 5.9909480900864756e-05}
{"train_loss": 0.09290466457605362, "global_step": 118768, "epoch": 1334, "lr": 5.990891266342553e-05}
{"train_loss": 0.1537979692220688, "global_step": 118769, "epoch": 1334, "lr": 5.990834442465416e-05}
{"train_loss": 0.14261773228645325, "global_step": 118770, "epoch": 1334, "lr": 5.990777618455074e-05}
{"train_loss": 0.1394180804491043, "global_step": 118771, "epoch": 1334, "lr": 5.990720794311534e-05}
{"train_loss": 0.18370822072029114, "global_step": 118772, "epoch": 1334, "lr": 5.9906639700348044e-05}
{"train_loss": 0.13059335947036743, "global_step": 118773, "epoch": 1334, "lr": 5.9906071456248916e-05}
{"train_loss": 0.21390961110591888, "global_step": 118774, "epoch": 1334, "lr": 5.990550321081802e-05}
{"train_loss": 0.12776391208171844, "global_step": 118775, "epoch": 1334, "lr": 5.990493496405547e-05}
{"train_loss": 0.1405748724937439, "global_step": 118776, "epoch": 1334, "lr": 5.9904366715961316e-05}
{"train_loss": 0.1347261667251587, "global_step": 118777, "epoch": 1334, "lr": 5.9903798466535646e-05}
{"train_loss": 0.16939979791641235, "global_step": 118778, "epoch": 1334, "lr": 5.990323021577853e-05}
{"train_loss": 0.165345698595047, "global_step": 118779, "epoch": 1334, "lr": 5.990266196369004e-05}
{"train_loss": 0.142371267080307, "global_step": 118780, "epoch": 1334, "lr": 5.990209371027027e-05}
{"train_loss": 0.08216413855552673, "global_step": 118781, "epoch": 1334, "lr": 5.9901525455519266e-05}
{"train_loss": 0.19361448287963867, "global_step": 118782, "epoch": 1334, "lr": 5.990095719943714e-05}
{"train_loss": 0.138388529419899, "global_step": 118783, "epoch": 1334, "lr": 5.990038894202393e-05}
{"train_loss": 0.20092153549194336, "global_step": 118784, "epoch": 1334, "lr": 5.989982068327974e-05}
{"train_loss": 0.183808833360672, "global_step": 118785, "epoch": 1334, "lr": 5.989925242320464e-05}
{"train_loss": 0.1296350210905075, "global_step": 118786, "epoch": 1334, "lr": 5.9898684161798715e-05}
{"train_loss": 0.15426293015480042, "global_step": 118787, "epoch": 1334, "lr": 5.989811589906202e-05}
{"train_loss": 0.11866620928049088, "global_step": 118788, "epoch": 1334, "lr": 5.989754763499466e-05}
{"train_loss": 0.13463303446769714, "global_step": 118789, "epoch": 1334, "lr": 5.989697936959668e-05}
{"train_loss": 0.12726782262325287, "global_step": 118790, "epoch": 1334, "lr": 5.989641110286819e-05}
{"train_loss": 0.10859628021717072, "global_step": 118791, "epoch": 1334, "lr": 5.989584283480924e-05}
{"train_loss": 0.17573362588882446, "global_step": 118792, "epoch": 1334, "lr": 5.9895274565419915e-05}
{"train_loss": 0.12705793976783752, "global_step": 118793, "epoch": 1334, "lr": 5.989470629470029e-05}
{"train_loss": 0.22578053176403046, "global_step": 118794, "epoch": 1334, "lr": 5.9894138022650424e-05}
{"train_loss": 0.12800443172454834, "global_step": 118795, "epoch": 1334, "lr": 5.989356974927044e-05}
{"train_loss": 0.09934186935424805, "global_step": 118796, "epoch": 1334, "lr": 5.989300147456037e-05}
{"train_loss": 0.13135385513305664, "global_step": 118797, "epoch": 1334, "lr": 5.989243319852032e-05}
{"train_loss": 0.13898028433322906, "global_step": 118798, "epoch": 1334, "lr": 5.989186492115034e-05}
{"train_loss": 0.20531553030014038, "global_step": 118799, "epoch": 1334, "lr": 5.9891296642450526e-05}
{"train_loss": 0.13423137366771698, "global_step": 118800, "epoch": 1334, "lr": 5.9890728362420945e-05}
{"train_loss": 0.16036218404769897, "global_step": 118801, "epoch": 1334, "lr": 5.989016008106167e-05}
{"train_loss": 0.14263816177845, "global_step": 118802, "epoch": 1334, "lr": 5.9889591798372794e-05}
{"train_loss": 0.14831703901290894, "global_step": 118803, "epoch": 1334, "lr": 5.988902351435437e-05}
{"train_loss": 0.21310794353485107, "global_step": 118804, "epoch": 1334, "lr": 5.98884552290065e-05}
{"train_loss": 0.1705087274312973, "global_step": 118805, "epoch": 1334, "lr": 5.9887886942329244e-05}
{"train_loss": 0.13056710362434387, "global_step": 118806, "epoch": 1334, "lr": 5.988731865432268e-05}
{"train_loss": 0.20373108983039856, "global_step": 118807, "epoch": 1334, "lr": 5.988675036498689e-05}
{"train_loss": 0.13600647449493408, "global_step": 118808, "epoch": 1334, "lr": 5.988618207432194e-05}
{"train_loss": 0.1286449134349823, "global_step": 118809, "epoch": 1334, "lr": 5.988561378232792e-05}
{"train_loss": 0.11964800208806992, "global_step": 118810, "epoch": 1334, "lr": 5.98850454890049e-05}
{"train_loss": 0.1702887862920761, "global_step": 118811, "epoch": 1334, "lr": 5.988447719435295e-05}
{"train_loss": 0.14513131976127625, "global_step": 118812, "epoch": 1334, "lr": 5.9883908898372156e-05}
{"train_loss": 0.07943982630968094, "global_step": 118813, "epoch": 1334, "lr": 5.98833406010626e-05}
{"train_loss": 0.13814519789446605, "global_step": 118814, "epoch": 1334, "lr": 5.988277230242434e-05, "val_loss": 5.244865417480469}
{"train_loss": 0.17609556019306183, "global_step": 118815, "epoch": 1335, "lr": 5.988220400245745e-05}
{"train_loss": 0.11455871164798737, "global_step": 118816, "epoch": 1335, "lr": 5.988163570116203e-05}
{"train_loss": 0.1314319670200348, "global_step": 118817, "epoch": 1335, "lr": 5.988106739853816e-05}
{"train_loss": 0.08980914950370789, "global_step": 118818, "epoch": 1335, "lr": 5.9880499094585887e-05}
{"train_loss": 0.14258837699890137, "global_step": 118819, "epoch": 1335, "lr": 5.9879930789305305e-05}
{"train_loss": 0.13678205013275146, "global_step": 118820, "epoch": 1335, "lr": 5.987936248269647e-05}
{"train_loss": 0.16616396605968475, "global_step": 118821, "epoch": 1335, "lr": 5.987879417475949e-05}
{"train_loss": 0.10895857959985733, "global_step": 118822, "epoch": 1335, "lr": 5.987822586549442e-05}
{"train_loss": 0.1296764463186264, "global_step": 118823, "epoch": 1335, "lr": 5.9877657554901347e-05}
{"train_loss": 0.1551429033279419, "global_step": 118824, "epoch": 1335, "lr": 5.9877089242980345e-05}
{"train_loss": 0.09002990275621414, "global_step": 118825, "epoch": 1335, "lr": 5.987652092973148e-05}
{"train_loss": 0.17119479179382324, "global_step": 118826, "epoch": 1335, "lr": 5.987595261515485e-05}
{"train_loss": 0.11581984162330627, "global_step": 118827, "epoch": 1335, "lr": 5.9875384299250515e-05}
{"train_loss": 0.22788405418395996, "global_step": 118828, "epoch": 1335, "lr": 5.987481598201855e-05}
{"train_loss": 0.15913163125514984, "global_step": 118829, "epoch": 1335, "lr": 5.987424766345904e-05}
{"train_loss": 0.09327305853366852, "global_step": 118830, "epoch": 1335, "lr": 5.987367934357205e-05}
{"train_loss": 0.14897401630878448, "global_step": 118831, "epoch": 1335, "lr": 5.9873111022357676e-05}
{"train_loss": 0.12785394489765167, "global_step": 118832, "epoch": 1335, "lr": 5.987254269981597e-05}
{"train_loss": 0.10924617201089859, "global_step": 118833, "epoch": 1335, "lr": 5.9871974375947036e-05}
{"train_loss": 0.18761475384235382, "global_step": 118834, "epoch": 1335, "lr": 5.987140605075092e-05}
{"train_loss": 0.16647349298000336, "global_step": 118835, "epoch": 1335, "lr": 5.987083772422772e-05}
{"train_loss": 0.17718735337257385, "global_step": 118836, "epoch": 1335, "lr": 5.987026939637751e-05}
{"train_loss": 0.0973043367266655, "global_step": 118837, "epoch": 1335, "lr": 5.986970106720038e-05}
{"train_loss": 0.14648890495300293, "global_step": 118838, "epoch": 1335, "lr": 5.9869132736696355e-05}
{"train_loss": 0.1138501688838005, "global_step": 118839, "epoch": 1335, "lr": 5.986856440486556e-05}
{"train_loss": 0.212769016623497, "global_step": 118840, "epoch": 1335, "lr": 5.986799607170805e-05}
{"train_loss": 0.1186179593205452, "global_step": 118841, "epoch": 1335, "lr": 5.986742773722391e-05}
{"train_loss": 0.19308294355869293, "global_step": 118842, "epoch": 1335, "lr": 5.986685940141323e-05}
{"train_loss": 0.16131478548049927, "global_step": 118843, "epoch": 1335, "lr": 5.986629106427606e-05}
{"train_loss": 0.11537892371416092, "global_step": 118844, "epoch": 1335, "lr": 5.986572272581249e-05}
{"train_loss": 0.07827277481555939, "global_step": 118845, "epoch": 1335, "lr": 5.986515438602258e-05}
{"train_loss": 0.13889309763908386, "global_step": 118846, "epoch": 1335, "lr": 5.986458604490643e-05}
{"train_loss": 0.18224772810935974, "global_step": 118847, "epoch": 1335, "lr": 5.9864017702464116e-05}
{"train_loss": 0.11437676101922989, "global_step": 118848, "epoch": 1335, "lr": 5.98634493586957e-05}
{"train_loss": 0.1450652778148651, "global_step": 118849, "epoch": 1335, "lr": 5.9862881013601256e-05}
{"train_loss": 0.08080357313156128, "global_step": 118850, "epoch": 1335, "lr": 5.9862312667180873e-05}
{"train_loss": 0.2730019688606262, "global_step": 118851, "epoch": 1335, "lr": 5.986174431943462e-05}
{"train_loss": 0.13763204216957092, "global_step": 118852, "epoch": 1335, "lr": 5.9861175970362585e-05}
{"train_loss": 0.12020153552293777, "global_step": 118853, "epoch": 1335, "lr": 5.986060761996483e-05}
{"train_loss": 0.10862461477518082, "global_step": 118854, "epoch": 1335, "lr": 5.986003926824143e-05}
{"train_loss": 0.15227441489696503, "global_step": 118855, "epoch": 1335, "lr": 5.985947091519247e-05}
{"train_loss": 0.09024512767791748, "global_step": 118856, "epoch": 1335, "lr": 5.985890256081803e-05}
{"train_loss": 0.16474561393260956, "global_step": 118857, "epoch": 1335, "lr": 5.985833420511817e-05}
{"train_loss": 0.14326684176921844, "global_step": 118858, "epoch": 1335, "lr": 5.9857765848092984e-05}
{"train_loss": 0.12322511523962021, "global_step": 118859, "epoch": 1335, "lr": 5.985719748974254e-05}
{"train_loss": 0.15556421875953674, "global_step": 118860, "epoch": 1335, "lr": 5.985662913006692e-05}
{"train_loss": 0.19466683268547058, "global_step": 118861, "epoch": 1335, "lr": 5.98560607690662e-05}
{"train_loss": 0.16320201754570007, "global_step": 118862, "epoch": 1335, "lr": 5.9855492406740445e-05}
{"train_loss": 0.1218501552939415, "global_step": 118863, "epoch": 1335, "lr": 5.9854924043089746e-05}
{"train_loss": 0.1159711554646492, "global_step": 118864, "epoch": 1335, "lr": 5.985435567811416e-05}
{"train_loss": 0.14873278141021729, "global_step": 118865, "epoch": 1335, "lr": 5.985378731181378e-05}
{"train_loss": 0.1075446605682373, "global_step": 118866, "epoch": 1335, "lr": 5.9853218944188684e-05}
{"train_loss": 0.14744995534420013, "global_step": 118867, "epoch": 1335, "lr": 5.985265057523895e-05}
{"train_loss": 0.17966219782829285, "global_step": 118868, "epoch": 1335, "lr": 5.985208220496463e-05}
{"train_loss": 0.14556804299354553, "global_step": 118869, "epoch": 1335, "lr": 5.985151383336582e-05}
{"train_loss": 0.11992732435464859, "global_step": 118870, "epoch": 1335, "lr": 5.985094546044261e-05}
{"train_loss": 0.14049026370048523, "global_step": 118871, "epoch": 1335, "lr": 5.9850377086195044e-05}
{"train_loss": 0.14837601780891418, "global_step": 118872, "epoch": 1335, "lr": 5.9849808710623224e-05}
{"train_loss": 0.14855152368545532, "global_step": 118873, "epoch": 1335, "lr": 5.9849240333727206e-05}
{"train_loss": 0.08962707221508026, "global_step": 118874, "epoch": 1335, "lr": 5.98486719555071e-05}
{"train_loss": 0.13275080919265747, "global_step": 118875, "epoch": 1335, "lr": 5.984810357596293e-05}
{"train_loss": 0.1658804714679718, "global_step": 118876, "epoch": 1335, "lr": 5.984753519509483e-05}
{"train_loss": 0.12238114327192307, "global_step": 118877, "epoch": 1335, "lr": 5.984696681290283e-05}
{"train_loss": 0.14139656722545624, "global_step": 118878, "epoch": 1335, "lr": 5.984639842938704e-05}
{"train_loss": 0.14910902082920074, "global_step": 118879, "epoch": 1335, "lr": 5.9845830044547513e-05}
{"train_loss": 0.10941526293754578, "global_step": 118880, "epoch": 1335, "lr": 5.984526165838433e-05}
{"train_loss": 0.07116910815238953, "global_step": 118881, "epoch": 1335, "lr": 5.9844693270897576e-05}
{"train_loss": 0.10821288079023361, "global_step": 118882, "epoch": 1335, "lr": 5.984412488208733e-05}
{"train_loss": 0.15434888005256653, "global_step": 118883, "epoch": 1335, "lr": 5.9843556491953656e-05}
{"train_loss": 0.1168016791343689, "global_step": 118884, "epoch": 1335, "lr": 5.9842988100496634e-05}
{"train_loss": 0.11512675881385803, "global_step": 118885, "epoch": 1335, "lr": 5.984241970771634e-05}
{"train_loss": 0.13250088691711426, "global_step": 118886, "epoch": 1335, "lr": 5.9841851313612864e-05}
{"train_loss": 0.16923294961452484, "global_step": 118887, "epoch": 1335, "lr": 5.984128291818626e-05}
{"train_loss": 0.0883827954530716, "global_step": 118888, "epoch": 1335, "lr": 5.9840714521436616e-05}
{"train_loss": 0.12160637974739075, "global_step": 118889, "epoch": 1335, "lr": 5.9840146123364015e-05}
{"train_loss": 0.13792182505130768, "global_step": 118890, "epoch": 1335, "lr": 5.983957772396852e-05}
{"train_loss": 0.10896069556474686, "global_step": 118891, "epoch": 1335, "lr": 5.9839009323250214e-05}
{"train_loss": 0.08845911175012589, "global_step": 118892, "epoch": 1335, "lr": 5.983844092120917e-05}
{"train_loss": 0.09615381807088852, "global_step": 118893, "epoch": 1335, "lr": 5.983787251784547e-05}
{"train_loss": 0.14571312069892883, "global_step": 118894, "epoch": 1335, "lr": 5.98373041131592e-05}
{"train_loss": 0.17553576827049255, "global_step": 118895, "epoch": 1335, "lr": 5.98367357071504e-05}
{"train_loss": 0.14773160219192505, "global_step": 118896, "epoch": 1335, "lr": 5.98361672998192e-05}
{"train_loss": 0.09310448169708252, "global_step": 118897, "epoch": 1335, "lr": 5.9835598891165635e-05}
{"train_loss": 0.10299692302942276, "global_step": 118898, "epoch": 1335, "lr": 5.983503048118978e-05}
{"train_loss": 0.1288820505142212, "global_step": 118899, "epoch": 1335, "lr": 5.983446206989174e-05}
{"train_loss": 0.07059256732463837, "global_step": 118900, "epoch": 1335, "lr": 5.9833893657271564e-05}
{"train_loss": 0.14191949367523193, "global_step": 118901, "epoch": 1335, "lr": 5.983332524332936e-05}
{"train_loss": 0.07911331206560135, "global_step": 118902, "epoch": 1335, "lr": 5.9832756828065176e-05}
{"train_loss": 0.1359418278162399, "global_step": 118903, "epoch": 1335, "lr": 5.98321884114791e-05, "val_loss": 5.247967720031738, "train_action_mse_error": 19.96169090270996}
{"train_loss": 0.1371736228466034, "global_step": 118904, "epoch": 1336, "lr": 5.9831619993571196e-05}
{"train_loss": 0.2070866823196411, "global_step": 118905, "epoch": 1336, "lr": 5.9831051574341555e-05}
{"train_loss": 0.09370069950819016, "global_step": 118906, "epoch": 1336, "lr": 5.983048315379025e-05}
{"train_loss": 0.14640410244464874, "global_step": 118907, "epoch": 1336, "lr": 5.982991473191736e-05}
{"train_loss": 0.13785459101200104, "global_step": 118908, "epoch": 1336, "lr": 5.9829346308722956e-05}
{"train_loss": 0.16178636252880096, "global_step": 118909, "epoch": 1336, "lr": 5.982877788420711e-05}
{"train_loss": 0.06688304245471954, "global_step": 118910, "epoch": 1336, "lr": 5.982820945836991e-05}
{"train_loss": 0.14663484692573547, "global_step": 118911, "epoch": 1336, "lr": 5.982764103121143e-05}
{"train_loss": 0.07687237113714218, "global_step": 118912, "epoch": 1336, "lr": 5.982707260273174e-05}
{"train_loss": 0.10039979219436646, "global_step": 118913, "epoch": 1336, "lr": 5.982650417293092e-05}
{"train_loss": 0.17301732301712036, "global_step": 118914, "epoch": 1336, "lr": 5.982593574180905e-05}
{"train_loss": 0.13827458024024963, "global_step": 118915, "epoch": 1336, "lr": 5.9825367309366195e-05}
{"train_loss": 0.16090765595436096, "global_step": 118916, "epoch": 1336, "lr": 5.982479887560244e-05}
{"train_loss": 0.13666434586048126, "global_step": 118917, "epoch": 1336, "lr": 5.982423044051787e-05}
{"train_loss": 0.1207692101597786, "global_step": 118918, "epoch": 1336, "lr": 5.9823662004112545e-05}
{"train_loss": 0.1362193077802658, "global_step": 118919, "epoch": 1336, "lr": 5.9823093566386545e-05}
{"train_loss": 0.24530497193336487, "global_step": 118920, "epoch": 1336, "lr": 5.982252512733996e-05}
{"train_loss": 0.19094540178775787, "global_step": 118921, "epoch": 1336, "lr": 5.982195668697285e-05}
{"train_loss": 0.15371523797512054, "global_step": 118922, "epoch": 1336, "lr": 5.9821388245285295e-05}
{"train_loss": 0.08625280112028122, "global_step": 118923, "epoch": 1336, "lr": 5.982081980227737e-05}
{"train_loss": 0.13545766472816467, "global_step": 118924, "epoch": 1336, "lr": 5.9820251357949164e-05}
{"train_loss": 0.10704353451728821, "global_step": 118925, "epoch": 1336, "lr": 5.981968291230074e-05}
{"train_loss": 0.10895775258541107, "global_step": 118926, "epoch": 1336, "lr": 5.9819114465332196e-05}
{"train_loss": 0.06499926745891571, "global_step": 118927, "epoch": 1336, "lr": 5.981854601704357e-05}
{"train_loss": 0.1409609168767929, "global_step": 118928, "epoch": 1336, "lr": 5.981797756743497e-05}
{"train_loss": 0.14679402112960815, "global_step": 118929, "epoch": 1336, "lr": 5.981740911650646e-05}
{"train_loss": 0.09829000383615494, "global_step": 118930, "epoch": 1336, "lr": 5.981684066425812e-05}
{"train_loss": 0.1660911589860916, "global_step": 118931, "epoch": 1336, "lr": 5.981627221069003e-05}
{"train_loss": 0.10987875610589981, "global_step": 118932, "epoch": 1336, "lr": 5.981570375580225e-05}
{"train_loss": 0.09306370466947556, "global_step": 118933, "epoch": 1336, "lr": 5.981513529959488e-05}
{"train_loss": 0.15410327911376953, "global_step": 118934, "epoch": 1336, "lr": 5.9814566842067974e-05}
{"train_loss": 0.14535823464393616, "global_step": 118935, "epoch": 1336, "lr": 5.9813998383221634e-05}
{"train_loss": 0.10481829941272736, "global_step": 118936, "epoch": 1336, "lr": 5.981342992305591e-05}
{"train_loss": 0.11800231039524078, "global_step": 118937, "epoch": 1336, "lr": 5.981286146157089e-05}
{"train_loss": 0.15662957727909088, "global_step": 118938, "epoch": 1336, "lr": 5.981229299876666e-05}
{"train_loss": 0.10540904104709625, "global_step": 118939, "epoch": 1336, "lr": 5.9811724534643274e-05}
{"train_loss": 0.14959314465522766, "global_step": 118940, "epoch": 1336, "lr": 5.9811156069200824e-05}
{"train_loss": 0.15940813720226288, "global_step": 118941, "epoch": 1336, "lr": 5.9810587602439394e-05}
{"train_loss": 0.13825026154518127, "global_step": 118942, "epoch": 1336, "lr": 5.981001913435904e-05}
{"train_loss": 0.1332533359527588, "global_step": 118943, "epoch": 1336, "lr": 5.9809450664959855e-05}
{"train_loss": 0.08433452993631363, "global_step": 118944, "epoch": 1336, "lr": 5.98088821942419e-05}
{"train_loss": 0.15459974110126495, "global_step": 118945, "epoch": 1336, "lr": 5.9808313722205265e-05}
{"train_loss": 0.06474832445383072, "global_step": 118946, "epoch": 1336, "lr": 5.980774524885002e-05}
{"train_loss": 0.1216466948390007, "global_step": 118947, "epoch": 1336, "lr": 5.980717677417624e-05}
{"train_loss": 0.15155911445617676, "global_step": 118948, "epoch": 1336, "lr": 5.9806608298184017e-05}
{"train_loss": 0.08909120410680771, "global_step": 118949, "epoch": 1336, "lr": 5.9806039820873415e-05}
{"train_loss": 0.1691902130842209, "global_step": 118950, "epoch": 1336, "lr": 5.98054713422445e-05}
{"train_loss": 0.1411428451538086, "global_step": 118951, "epoch": 1336, "lr": 5.980490286229736e-05}
{"train_loss": 0.15000271797180176, "global_step": 118952, "epoch": 1336, "lr": 5.980433438103208e-05}
{"train_loss": 0.154093936085701, "global_step": 118953, "epoch": 1336, "lr": 5.980376589844872e-05}
{"train_loss": 0.13323497772216797, "global_step": 118954, "epoch": 1336, "lr": 5.9803197414547366e-05}
{"train_loss": 0.10875321179628372, "global_step": 118955, "epoch": 1336, "lr": 5.9802628929328086e-05}
{"train_loss": 0.09766612946987152, "global_step": 118956, "epoch": 1336, "lr": 5.9802060442790964e-05}
{"train_loss": 0.1422342211008072, "global_step": 118957, "epoch": 1336, "lr": 5.980149195493607e-05}
{"train_loss": 0.1578817516565323, "global_step": 118958, "epoch": 1336, "lr": 5.980092346576349e-05}
{"train_loss": 0.15522314608097076, "global_step": 118959, "epoch": 1336, "lr": 5.98003549752733e-05}
{"train_loss": 0.12586471438407898, "global_step": 118960, "epoch": 1336, "lr": 5.979978648346557e-05}
{"train_loss": 0.1079302430152893, "global_step": 118961, "epoch": 1336, "lr": 5.979921799034037e-05}
{"train_loss": 0.15581370890140533, "global_step": 118962, "epoch": 1336, "lr": 5.9798649495897785e-05}
{"train_loss": 0.17976371943950653, "global_step": 118963, "epoch": 1336, "lr": 5.97980810001379e-05}
{"train_loss": 0.11008918285369873, "global_step": 118964, "epoch": 1336, "lr": 5.979751250306078e-05}
{"train_loss": 0.16886743903160095, "global_step": 118965, "epoch": 1336, "lr": 5.979694400466649e-05}
{"train_loss": 0.14199939370155334, "global_step": 118966, "epoch": 1336, "lr": 5.979637550495513e-05}
{"train_loss": 0.25795552134513855, "global_step": 118967, "epoch": 1336, "lr": 5.9795807003926784e-05}
{"train_loss": 0.10756541788578033, "global_step": 118968, "epoch": 1336, "lr": 5.9795238501581494e-05}
{"train_loss": 0.16491207480430603, "global_step": 118969, "epoch": 1336, "lr": 5.979466999791935e-05}
{"train_loss": 0.18448315560817719, "global_step": 118970, "epoch": 1336, "lr": 5.9794101492940446e-05}
{"train_loss": 0.18128150701522827, "global_step": 118971, "epoch": 1336, "lr": 5.979353298664483e-05}
{"train_loss": 0.15193584561347961, "global_step": 118972, "epoch": 1336, "lr": 5.9792964479032595e-05}
{"train_loss": 0.1409500688314438, "global_step": 118973, "epoch": 1336, "lr": 5.979239597010382e-05}
{"train_loss": 0.06158507615327835, "global_step": 118974, "epoch": 1336, "lr": 5.9791827459858584e-05}
{"train_loss": 0.16305464506149292, "global_step": 118975, "epoch": 1336, "lr": 5.9791258948296945e-05}
{"train_loss": 0.1347096860408783, "global_step": 118976, "epoch": 1336, "lr": 5.9790690435419e-05}
{"train_loss": 0.10819321870803833, "global_step": 118977, "epoch": 1336, "lr": 5.9790121921224807e-05}
{"train_loss": 0.14219501614570618, "global_step": 118978, "epoch": 1336, "lr": 5.978955340571445e-05}
{"train_loss": 0.2024935930967331, "global_step": 118979, "epoch": 1336, "lr": 5.978898488888801e-05}
{"train_loss": 0.13826830685138702, "global_step": 118980, "epoch": 1336, "lr": 5.978841637074556e-05}
{"train_loss": 0.13165396451950073, "global_step": 118981, "epoch": 1336, "lr": 5.978784785128718e-05}
{"train_loss": 0.06572176516056061, "global_step": 118982, "epoch": 1336, "lr": 5.978727933051293e-05}
{"train_loss": 0.13385768234729767, "global_step": 118983, "epoch": 1336, "lr": 5.978671080842291e-05}
{"train_loss": 0.19201244413852692, "global_step": 118984, "epoch": 1336, "lr": 5.978614228501719e-05}
{"train_loss": 0.25835251808166504, "global_step": 118985, "epoch": 1336, "lr": 5.978557376029583e-05}
{"train_loss": 0.16579893231391907, "global_step": 118986, "epoch": 1336, "lr": 5.978500523425894e-05}
{"train_loss": 0.1610310673713684, "global_step": 118987, "epoch": 1336, "lr": 5.978443670690657e-05}
{"train_loss": 0.09964703023433685, "global_step": 118988, "epoch": 1336, "lr": 5.9783868178238786e-05}
{"train_loss": 0.10617664456367493, "global_step": 118989, "epoch": 1336, "lr": 5.97832996482557e-05}
{"train_loss": 0.11931509524583817, "global_step": 118990, "epoch": 1336, "lr": 5.978273111695736e-05}
{"train_loss": 0.16081048548221588, "global_step": 118991, "epoch": 1336, "lr": 5.978216258434385e-05}
{"train_loss": 0.1382488754023327, "global_step": 118992, "epoch": 1336, "lr": 5.978159405041523e-05, "val_loss": 5.262979984283447}
{"train_loss": 0.1534844934940338, "global_step": 118993, "epoch": 1337, "lr": 5.9781025515171616e-05}
{"train_loss": 0.1296604573726654, "global_step": 118994, "epoch": 1337, "lr": 5.9780456978613066e-05}
{"train_loss": 0.09855028986930847, "global_step": 118995, "epoch": 1337, "lr": 5.977988844073964e-05}
{"train_loss": 0.09904634952545166, "global_step": 118996, "epoch": 1337, "lr": 5.9779319901551434e-05}
{"train_loss": 0.20744799077510834, "global_step": 118997, "epoch": 1337, "lr": 5.9778751361048516e-05}
{"train_loss": 0.1015106588602066, "global_step": 118998, "epoch": 1337, "lr": 5.9778182819230955e-05}
{"train_loss": 0.17061758041381836, "global_step": 118999, "epoch": 1337, "lr": 5.9777614276098846e-05}
{"train_loss": 0.18419377505779266, "global_step": 119000, "epoch": 1337, "lr": 5.977704573165225e-05}
{"train_loss": 0.12486366182565689, "global_step": 119001, "epoch": 1337, "lr": 5.9776477185891256e-05}
{"train_loss": 0.1309150755405426, "global_step": 119002, "epoch": 1337, "lr": 5.977590863881591e-05}
{"train_loss": 0.1824112981557846, "global_step": 119003, "epoch": 1337, "lr": 5.977534009042634e-05}
{"train_loss": 0.14965392649173737, "global_step": 119004, "epoch": 1337, "lr": 5.977477154072259e-05}
{"train_loss": 0.21133872866630554, "global_step": 119005, "epoch": 1337, "lr": 5.9774202989704744e-05}
{"train_loss": 0.11971477419137955, "global_step": 119006, "epoch": 1337, "lr": 5.9773634437372875e-05}
{"train_loss": 0.12203896045684814, "global_step": 119007, "epoch": 1337, "lr": 5.9773065883727054e-05}
{"train_loss": 0.14078909158706665, "global_step": 119008, "epoch": 1337, "lr": 5.9772497328767364e-05}
{"train_loss": 0.15234723687171936, "global_step": 119009, "epoch": 1337, "lr": 5.977192877249388e-05}
{"train_loss": 0.18636688590049744, "global_step": 119010, "epoch": 1337, "lr": 5.977136021490668e-05}
{"train_loss": 0.09949588775634766, "global_step": 119011, "epoch": 1337, "lr": 5.9770791656005844e-05}
{"train_loss": 0.13115617632865906, "global_step": 119012, "epoch": 1337, "lr": 5.9770223095791446e-05}
{"train_loss": 0.13415074348449707, "global_step": 119013, "epoch": 1337, "lr": 5.976965453426355e-05}
{"train_loss": 0.08293467015028, "global_step": 119014, "epoch": 1337, "lr": 5.9769085971422245e-05}
{"train_loss": 0.18889585137367249, "global_step": 119015, "epoch": 1337, "lr": 5.976851740726761e-05}
{"train_loss": 0.12247584760189056, "global_step": 119016, "epoch": 1337, "lr": 5.976794884179971e-05}
{"train_loss": 0.21089890599250793, "global_step": 119017, "epoch": 1337, "lr": 5.976738027501865e-05}
{"train_loss": 0.14836838841438293, "global_step": 119018, "epoch": 1337, "lr": 5.9766811706924464e-05}
{"train_loss": 0.13386201858520508, "global_step": 119019, "epoch": 1337, "lr": 5.976624313751725e-05}
{"train_loss": 0.1600547581911087, "global_step": 119020, "epoch": 1337, "lr": 5.976567456679708e-05}
{"train_loss": 0.23682574927806854, "global_step": 119021, "epoch": 1337, "lr": 5.976510599476404e-05}
{"train_loss": 0.1149156242609024, "global_step": 119022, "epoch": 1337, "lr": 5.97645374214182e-05}
{"train_loss": 0.09004352241754532, "global_step": 119023, "epoch": 1337, "lr": 5.9763968846759645e-05}
{"train_loss": 0.08395706117153168, "global_step": 119024, "epoch": 1337, "lr": 5.976340027078844e-05}
{"train_loss": 0.09945018589496613, "global_step": 119025, "epoch": 1337, "lr": 5.9762831693504676e-05}
{"train_loss": 0.14548183977603912, "global_step": 119026, "epoch": 1337, "lr": 5.97622631149084e-05}
{"train_loss": 0.12452138960361481, "global_step": 119027, "epoch": 1337, "lr": 5.9761694534999716e-05}
{"train_loss": 0.15712526440620422, "global_step": 119028, "epoch": 1337, "lr": 5.9761125953778676e-05}
{"train_loss": 0.22398781776428223, "global_step": 119029, "epoch": 1337, "lr": 5.976055737124539e-05}
{"train_loss": 0.12830854952335358, "global_step": 119030, "epoch": 1337, "lr": 5.97599887873999e-05}
{"train_loss": 0.06259040534496307, "global_step": 119031, "epoch": 1337, "lr": 5.9759420202242313e-05}
{"train_loss": 0.11909732222557068, "global_step": 119032, "epoch": 1337, "lr": 5.975885161577268e-05}
{"train_loss": 0.10310576111078262, "global_step": 119033, "epoch": 1337, "lr": 5.975828302799109e-05}
{"train_loss": 0.1894497573375702, "global_step": 119034, "epoch": 1337, "lr": 5.975771443889763e-05}
{"train_loss": 0.15232299268245697, "global_step": 119035, "epoch": 1337, "lr": 5.9757145848492345e-05}
{"train_loss": 0.09366905689239502, "global_step": 119036, "epoch": 1337, "lr": 5.975657725677535e-05}
{"train_loss": 0.1950433999300003, "global_step": 119037, "epoch": 1337, "lr": 5.975600866374669e-05}
{"train_loss": 0.15947961807250977, "global_step": 119038, "epoch": 1337, "lr": 5.9755440069406445e-05}
{"train_loss": 0.19387111067771912, "global_step": 119039, "epoch": 1337, "lr": 5.97548714737547e-05}
{"train_loss": 0.11327908933162689, "global_step": 119040, "epoch": 1337, "lr": 5.975430287679154e-05}
{"train_loss": 0.14417245984077454, "global_step": 119041, "epoch": 1337, "lr": 5.975373427851705e-05}
{"train_loss": 0.1444135159254074, "global_step": 119042, "epoch": 1337, "lr": 5.975316567893127e-05}
{"train_loss": 0.11213549971580505, "global_step": 119043, "epoch": 1337, "lr": 5.97525970780343e-05}
{"train_loss": 0.13260972499847412, "global_step": 119044, "epoch": 1337, "lr": 5.975202847582623e-05}
{"train_loss": 0.10476203262805939, "global_step": 119045, "epoch": 1337, "lr": 5.975145987230709e-05}
{"train_loss": 0.1668364554643631, "global_step": 119046, "epoch": 1337, "lr": 5.9750891267477005e-05}
{"train_loss": 0.12024721503257751, "global_step": 119047, "epoch": 1337, "lr": 5.975032266133602e-05}
{"train_loss": 0.10652703791856766, "global_step": 119048, "epoch": 1337, "lr": 5.9749754053884234e-05}
{"train_loss": 0.09127415716648102, "global_step": 119049, "epoch": 1337, "lr": 5.974918544512169e-05}
{"train_loss": 0.1822814792394638, "global_step": 119050, "epoch": 1337, "lr": 5.974861683504851e-05}
{"train_loss": 0.14227701723575592, "global_step": 119051, "epoch": 1337, "lr": 5.974804822366473e-05}
{"train_loss": 0.06704489886760712, "global_step": 119052, "epoch": 1337, "lr": 5.974747961097046e-05}
{"train_loss": 0.11009687930345535, "global_step": 119053, "epoch": 1337, "lr": 5.974691099696576e-05}
{"train_loss": 0.1652786284685135, "global_step": 119054, "epoch": 1337, "lr": 5.974634238165069e-05}
{"train_loss": 0.1647009551525116, "global_step": 119055, "epoch": 1337, "lr": 5.974577376502535e-05}
{"train_loss": 0.13065004348754883, "global_step": 119056, "epoch": 1337, "lr": 5.97452051470898e-05}
{"train_loss": 0.22354978322982788, "global_step": 119057, "epoch": 1337, "lr": 5.974463652784414e-05}
{"train_loss": 0.18372118473052979, "global_step": 119058, "epoch": 1337, "lr": 5.9744067907288424e-05}
{"train_loss": 0.07177726924419403, "global_step": 119059, "epoch": 1337, "lr": 5.9743499285422746e-05}
{"train_loss": 0.12402909994125366, "global_step": 119060, "epoch": 1337, "lr": 5.9742930662247176e-05}
{"train_loss": 0.1018751934170723, "global_step": 119061, "epoch": 1337, "lr": 5.974236203776178e-05}
{"train_loss": 0.12362270802259445, "global_step": 119062, "epoch": 1337, "lr": 5.974179341196664e-05}
{"train_loss": 0.12708568572998047, "global_step": 119063, "epoch": 1337, "lr": 5.9741224784861836e-05}
{"train_loss": 0.10708028823137283, "global_step": 119064, "epoch": 1337, "lr": 5.974065615644746e-05}
{"train_loss": 0.11308836191892624, "global_step": 119065, "epoch": 1337, "lr": 5.9740087526723554e-05}
{"train_loss": 0.12547007203102112, "global_step": 119066, "epoch": 1337, "lr": 5.9739518895690215e-05}
{"train_loss": 0.0693400427699089, "global_step": 119067, "epoch": 1337, "lr": 5.973895026334751e-05}
{"train_loss": 0.08952486515045166, "global_step": 119068, "epoch": 1337, "lr": 5.973838162969554e-05}
{"train_loss": 0.14968819916248322, "global_step": 119069, "epoch": 1337, "lr": 5.973781299473435e-05}
{"train_loss": 0.11162591725587845, "global_step": 119070, "epoch": 1337, "lr": 5.9737244358464026e-05}
{"train_loss": 0.10983706265687943, "global_step": 119071, "epoch": 1337, "lr": 5.973667572088466e-05}
{"train_loss": 0.12280865758657455, "global_step": 119072, "epoch": 1337, "lr": 5.9736107081996304e-05}
{"train_loss": 0.12165086716413498, "global_step": 119073, "epoch": 1337, "lr": 5.973553844179906e-05}
{"train_loss": 0.184828519821167, "global_step": 119074, "epoch": 1337, "lr": 5.973496980029297e-05}
{"train_loss": 0.09597083926200867, "global_step": 119075, "epoch": 1337, "lr": 5.973440115747815e-05}
{"train_loss": 0.2541149854660034, "global_step": 119076, "epoch": 1337, "lr": 5.9733832513354646e-05}
{"train_loss": 0.09388986974954605, "global_step": 119077, "epoch": 1337, "lr": 5.973326386792256e-05}
{"train_loss": 0.17353558540344238, "global_step": 119078, "epoch": 1337, "lr": 5.973269522118196e-05}
{"train_loss": 0.13756267726421356, "global_step": 119079, "epoch": 1337, "lr": 5.97321265731329e-05}
{"train_loss": 0.1262577772140503, "global_step": 119080, "epoch": 1337, "lr": 5.9731557923775486e-05}
{"train_loss": 0.13704438197813676, "global_step": 119081, "epoch": 1337, "lr": 5.9730989273109786e-05, "val_loss": 5.446777820587158}
{"train_loss": 0.11892254650592804, "global_step": 119082, "epoch": 1338, "lr": 5.973042062113586e-05}
{"train_loss": 0.1507832109928131, "global_step": 119083, "epoch": 1338, "lr": 5.972985196785381e-05}
{"train_loss": 0.09594327211380005, "global_step": 119084, "epoch": 1338, "lr": 5.9729283313263695e-05}
{"train_loss": 0.1009739562869072, "global_step": 119085, "epoch": 1338, "lr": 5.97287146573656e-05}
{"train_loss": 0.11227995157241821, "global_step": 119086, "epoch": 1338, "lr": 5.972814600015959e-05}
{"train_loss": 0.13233137130737305, "global_step": 119087, "epoch": 1338, "lr": 5.9727577341645756e-05}
{"train_loss": 0.11743079870939255, "global_step": 119088, "epoch": 1338, "lr": 5.972700868182417e-05}
{"train_loss": 0.12597006559371948, "global_step": 119089, "epoch": 1338, "lr": 5.972644002069489e-05}
{"train_loss": 0.19638873636722565, "global_step": 119090, "epoch": 1338, "lr": 5.9725871358258024e-05}
{"train_loss": 0.08854718506336212, "global_step": 119091, "epoch": 1338, "lr": 5.972530269451363e-05}
{"train_loss": 0.08795896172523499, "global_step": 119092, "epoch": 1338, "lr": 5.972473402946177e-05}
{"train_loss": 0.1132006123661995, "global_step": 119093, "epoch": 1338, "lr": 5.972416536310256e-05}
{"train_loss": 0.21161901950836182, "global_step": 119094, "epoch": 1338, "lr": 5.9723596695436046e-05}
{"train_loss": 0.17184218764305115, "global_step": 119095, "epoch": 1338, "lr": 5.9723028026462326e-05}
{"train_loss": 0.1557117998600006, "global_step": 119096, "epoch": 1338, "lr": 5.972245935618145e-05}
{"train_loss": 0.17523011565208435, "global_step": 119097, "epoch": 1338, "lr": 5.972189068459351e-05}
{"train_loss": 0.13909730315208435, "global_step": 119098, "epoch": 1338, "lr": 5.9721322011698576e-05}
{"train_loss": 0.12656287848949432, "global_step": 119099, "epoch": 1338, "lr": 5.972075333749674e-05}
{"train_loss": 0.19671565294265747, "global_step": 119100, "epoch": 1338, "lr": 5.972018466198806e-05}
{"train_loss": 0.10772456228733063, "global_step": 119101, "epoch": 1338, "lr": 5.9719615985172614e-05}
{"train_loss": 0.1187509074807167, "global_step": 119102, "epoch": 1338, "lr": 5.971904730705051e-05}
{"train_loss": 0.14300774037837982, "global_step": 119103, "epoch": 1338, "lr": 5.9718478627621775e-05}
{"train_loss": 0.13495545089244843, "global_step": 119104, "epoch": 1338, "lr": 5.971790994688652e-05}
{"train_loss": 0.18216484785079956, "global_step": 119105, "epoch": 1338, "lr": 5.9717341264844804e-05}
{"train_loss": 0.11617498099803925, "global_step": 119106, "epoch": 1338, "lr": 5.9716772581496704e-05}
{"train_loss": 0.16486302018165588, "global_step": 119107, "epoch": 1338, "lr": 5.971620389684232e-05}
{"train_loss": 0.11082930117845535, "global_step": 119108, "epoch": 1338, "lr": 5.971563521088169e-05}
{"train_loss": 0.18761324882507324, "global_step": 119109, "epoch": 1338, "lr": 5.971506652361492e-05}
{"train_loss": 0.10099398344755173, "global_step": 119110, "epoch": 1338, "lr": 5.971449783504207e-05}
{"train_loss": 0.13717146217823029, "global_step": 119111, "epoch": 1338, "lr": 5.971392914516323e-05}
{"train_loss": 0.11294705420732498, "global_step": 119112, "epoch": 1338, "lr": 5.9713360453978476e-05}
{"train_loss": 0.09156035631895065, "global_step": 119113, "epoch": 1338, "lr": 5.9712791761487875e-05}
{"train_loss": 0.09913696348667145, "global_step": 119114, "epoch": 1338, "lr": 5.971222306769151e-05}
{"train_loss": 0.14580591022968292, "global_step": 119115, "epoch": 1338, "lr": 5.971165437258945e-05}
{"train_loss": 0.12379613518714905, "global_step": 119116, "epoch": 1338, "lr": 5.971108567618178e-05}
{"train_loss": 0.14984175562858582, "global_step": 119117, "epoch": 1338, "lr": 5.9710516978468575e-05}
{"train_loss": 0.08220069855451584, "global_step": 119118, "epoch": 1338, "lr": 5.970994827944991e-05}
{"train_loss": 0.18823979794979095, "global_step": 119119, "epoch": 1338, "lr": 5.9709379579125856e-05}
{"train_loss": 0.15592102706432343, "global_step": 119120, "epoch": 1338, "lr": 5.970881087749649e-05}
{"train_loss": 0.1303042769432068, "global_step": 119121, "epoch": 1338, "lr": 5.97082421745619e-05}
{"train_loss": 0.15652737021446228, "global_step": 119122, "epoch": 1338, "lr": 5.970767347032217e-05}
{"train_loss": 0.15579578280448914, "global_step": 119123, "epoch": 1338, "lr": 5.970710476477734e-05}
{"train_loss": 0.1901908963918686, "global_step": 119124, "epoch": 1338, "lr": 5.970653605792752e-05}
{"train_loss": 0.15099361538887024, "global_step": 119125, "epoch": 1338, "lr": 5.970596734977276e-05}
{"train_loss": 0.1079552099108696, "global_step": 119126, "epoch": 1338, "lr": 5.970539864031317e-05}
{"train_loss": 0.1903461068868637, "global_step": 119127, "epoch": 1338, "lr": 5.970482992954879e-05}
{"train_loss": 0.10068491846323013, "global_step": 119128, "epoch": 1338, "lr": 5.970426121747972e-05}
{"train_loss": 0.1267392784357071, "global_step": 119129, "epoch": 1338, "lr": 5.970369250410602e-05}
{"train_loss": 0.1798604279756546, "global_step": 119130, "epoch": 1338, "lr": 5.970312378942779e-05}
{"train_loss": 0.17579637467861176, "global_step": 119131, "epoch": 1338, "lr": 5.9702555073445085e-05}
{"train_loss": 0.15706393122673035, "global_step": 119132, "epoch": 1338, "lr": 5.9701986356158e-05}
{"train_loss": 0.1695503145456314, "global_step": 119133, "epoch": 1338, "lr": 5.97014176375666e-05}
{"train_loss": 0.25886204838752747, "global_step": 119134, "epoch": 1338, "lr": 5.970084891767096e-05}
{"train_loss": 0.18770629167556763, "global_step": 119135, "epoch": 1338, "lr": 5.9700280196471156e-05}
{"train_loss": 0.07555701583623886, "global_step": 119136, "epoch": 1338, "lr": 5.969971147396727e-05}
{"train_loss": 0.13987411558628082, "global_step": 119137, "epoch": 1338, "lr": 5.9699142750159375e-05}
{"train_loss": 0.13894836604595184, "global_step": 119138, "epoch": 1338, "lr": 5.969857402504755e-05}
{"train_loss": 0.15613257884979248, "global_step": 119139, "epoch": 1338, "lr": 5.969800529863188e-05}
{"train_loss": 0.1081448420882225, "global_step": 119140, "epoch": 1338, "lr": 5.969743657091241e-05}
{"train_loss": 0.08224137127399445, "global_step": 119141, "epoch": 1338, "lr": 5.969686784188927e-05}
{"train_loss": 0.0980193242430687, "global_step": 119142, "epoch": 1338, "lr": 5.969629911156248e-05}
{"train_loss": 0.09486285597085953, "global_step": 119143, "epoch": 1338, "lr": 5.969573037993215e-05}
{"train_loss": 0.1337290108203888, "global_step": 119144, "epoch": 1338, "lr": 5.9695161646998335e-05}
{"train_loss": 0.054145414382219315, "global_step": 119145, "epoch": 1338, "lr": 5.9694592912761126e-05}
{"train_loss": 0.20806899666786194, "global_step": 119146, "epoch": 1338, "lr": 5.969402417722061e-05}
{"train_loss": 0.1685364991426468, "global_step": 119147, "epoch": 1338, "lr": 5.9693455440376824e-05}
{"train_loss": 0.10323552042245865, "global_step": 119148, "epoch": 1338, "lr": 5.969288670222989e-05}
{"train_loss": 0.1235417053103447, "global_step": 119149, "epoch": 1338, "lr": 5.969231796277986e-05}
{"train_loss": 0.16409605741500854, "global_step": 119150, "epoch": 1338, "lr": 5.969174922202683e-05}
{"train_loss": 0.10349856317043304, "global_step": 119151, "epoch": 1338, "lr": 5.969118047997084e-05}
{"train_loss": 0.15845100581645966, "global_step": 119152, "epoch": 1338, "lr": 5.9690611736612015e-05}
{"train_loss": 0.22184470295906067, "global_step": 119153, "epoch": 1338, "lr": 5.969004299195039e-05}
{"train_loss": 0.17907516658306122, "global_step": 119154, "epoch": 1338, "lr": 5.968947424598605e-05}
{"train_loss": 0.14941267669200897, "global_step": 119155, "epoch": 1338, "lr": 5.968890549871909e-05}
{"train_loss": 0.1533702164888382, "global_step": 119156, "epoch": 1338, "lr": 5.968833675014958e-05}
{"train_loss": 0.20696061849594116, "global_step": 119157, "epoch": 1338, "lr": 5.9687768000277577e-05}
{"train_loss": 0.0939449891448021, "global_step": 119158, "epoch": 1338, "lr": 5.968719924910318e-05}
{"train_loss": 0.20776242017745972, "global_step": 119159, "epoch": 1338, "lr": 5.968663049662645e-05}
{"train_loss": 0.06780640780925751, "global_step": 119160, "epoch": 1338, "lr": 5.9686061742847486e-05}
{"train_loss": 0.18351083993911743, "global_step": 119161, "epoch": 1338, "lr": 5.968549298776635e-05}
{"train_loss": 0.14786545932292938, "global_step": 119162, "epoch": 1338, "lr": 5.9684924231383096e-05}
{"train_loss": 0.11474748700857162, "global_step": 119163, "epoch": 1338, "lr": 5.968435547369784e-05}
{"train_loss": 0.13937810063362122, "global_step": 119164, "epoch": 1338, "lr": 5.968378671471062e-05}
{"train_loss": 0.090225450694561, "global_step": 119165, "epoch": 1338, "lr": 5.9683217954421544e-05}
{"train_loss": 0.17000530660152435, "global_step": 119166, "epoch": 1338, "lr": 5.968264919283067e-05}
{"train_loss": 0.17817005515098572, "global_step": 119167, "epoch": 1338, "lr": 5.96820804299381e-05}
{"train_loss": 0.16123610734939575, "global_step": 119168, "epoch": 1338, "lr": 5.9681511665743886e-05}
{"train_loss": 0.16959920525550842, "global_step": 119169, "epoch": 1338, "lr": 5.9680942900248105e-05}
{"train_loss": 0.1421624737258038, "global_step": 119170, "epoch": 1338, "lr": 5.9680374133450835e-05, "val_loss": 5.431158542633057}
{"train_loss": 0.15720133483409882, "global_step": 119171, "epoch": 1339, "lr": 5.967980536535217e-05}
{"train_loss": 0.1726774275302887, "global_step": 119172, "epoch": 1339, "lr": 5.967923659595216e-05}
{"train_loss": 0.22105982899665833, "global_step": 119173, "epoch": 1339, "lr": 5.967866782525091e-05}
{"train_loss": 0.09502477943897247, "global_step": 119174, "epoch": 1339, "lr": 5.967809905324847e-05}
{"train_loss": 0.17539606988430023, "global_step": 119175, "epoch": 1339, "lr": 5.9677530279944926e-05}
{"train_loss": 0.13765108585357666, "global_step": 119176, "epoch": 1339, "lr": 5.967696150534037e-05}
{"train_loss": 0.22100278735160828, "global_step": 119177, "epoch": 1339, "lr": 5.967639272943485e-05}
{"train_loss": 0.06762515753507614, "global_step": 119178, "epoch": 1339, "lr": 5.967582395222846e-05}
{"train_loss": 0.0746464878320694, "global_step": 119179, "epoch": 1339, "lr": 5.9675255173721265e-05}
{"train_loss": 0.1555095762014389, "global_step": 119180, "epoch": 1339, "lr": 5.967468639391338e-05}
{"train_loss": 0.1575269103050232, "global_step": 119181, "epoch": 1339, "lr": 5.967411761280482e-05}
{"train_loss": 0.1780187487602234, "global_step": 119182, "epoch": 1339, "lr": 5.967354883039571e-05}
{"train_loss": 0.2263847142457962, "global_step": 119183, "epoch": 1339, "lr": 5.9672980046686086e-05}
{"train_loss": 0.0997542068362236, "global_step": 119184, "epoch": 1339, "lr": 5.9672411261676075e-05}
{"train_loss": 0.1172308474779129, "global_step": 119185, "epoch": 1339, "lr": 5.9671842475365713e-05}
{"train_loss": 0.1246771290898323, "global_step": 119186, "epoch": 1339, "lr": 5.967127368775509e-05}
{"train_loss": 0.09604276716709137, "global_step": 119187, "epoch": 1339, "lr": 5.967070489884429e-05}
{"train_loss": 0.21773342788219452, "global_step": 119188, "epoch": 1339, "lr": 5.967013610863338e-05}
{"train_loss": 0.06292825937271118, "global_step": 119189, "epoch": 1339, "lr": 5.966956731712243e-05}
{"train_loss": 0.1259479969739914, "global_step": 119190, "epoch": 1339, "lr": 5.9668998524311526e-05}
{"train_loss": 0.19849272072315216, "global_step": 119191, "epoch": 1339, "lr": 5.966842973020075e-05}
{"train_loss": 0.16858474910259247, "global_step": 119192, "epoch": 1339, "lr": 5.9667860934790164e-05}
{"train_loss": 0.1428869366645813, "global_step": 119193, "epoch": 1339, "lr": 5.966729213807986e-05}
{"train_loss": 0.14713922142982483, "global_step": 119194, "epoch": 1339, "lr": 5.96667233400699e-05}
{"train_loss": 0.07339942455291748, "global_step": 119195, "epoch": 1339, "lr": 5.9666154540760364e-05}
{"train_loss": 0.16549283266067505, "global_step": 119196, "epoch": 1339, "lr": 5.966558574015134e-05}
{"train_loss": 0.13383184373378754, "global_step": 119197, "epoch": 1339, "lr": 5.966501693824289e-05}
{"train_loss": 0.18278904259204865, "global_step": 119198, "epoch": 1339, "lr": 5.96644481350351e-05}
{"train_loss": 0.13210266828536987, "global_step": 119199, "epoch": 1339, "lr": 5.966387933052805e-05}
{"train_loss": 0.11616832762956619, "global_step": 119200, "epoch": 1339, "lr": 5.96633105247218e-05}
{"train_loss": 0.15339310467243195, "global_step": 119201, "epoch": 1339, "lr": 5.966274171761642e-05}
{"train_loss": 0.14519090950489044, "global_step": 119202, "epoch": 1339, "lr": 5.966217290921202e-05}
{"train_loss": 0.14643746614456177, "global_step": 119203, "epoch": 1339, "lr": 5.966160409950866e-05}
{"train_loss": 0.164943665266037, "global_step": 119204, "epoch": 1339, "lr": 5.966103528850641e-05}
{"train_loss": 0.09614486247301102, "global_step": 119205, "epoch": 1339, "lr": 5.9660466476205355e-05}
{"train_loss": 0.12873491644859314, "global_step": 119206, "epoch": 1339, "lr": 5.965989766260557e-05}
{"train_loss": 0.1136355847120285, "global_step": 119207, "epoch": 1339, "lr": 5.965932884770712e-05}
{"train_loss": 0.09705618023872375, "global_step": 119208, "epoch": 1339, "lr": 5.9658760031510096e-05}
{"train_loss": 0.1432444453239441, "global_step": 119209, "epoch": 1339, "lr": 5.965819121401457e-05}
{"train_loss": 0.1064368188381195, "global_step": 119210, "epoch": 1339, "lr": 5.9657622395220616e-05}
{"train_loss": 0.18832279741764069, "global_step": 119211, "epoch": 1339, "lr": 5.965705357512832e-05}
{"train_loss": 0.10246618837118149, "global_step": 119212, "epoch": 1339, "lr": 5.965648475373774e-05}
{"train_loss": 0.09566213935613632, "global_step": 119213, "epoch": 1339, "lr": 5.9655915931048966e-05}
{"train_loss": 0.13769996166229248, "global_step": 119214, "epoch": 1339, "lr": 5.965534710706208e-05}
{"train_loss": 0.14623495936393738, "global_step": 119215, "epoch": 1339, "lr": 5.965477828177715e-05}
{"train_loss": 0.09071356803178787, "global_step": 119216, "epoch": 1339, "lr": 5.9654209455194246e-05}
{"train_loss": 0.06795690208673477, "global_step": 119217, "epoch": 1339, "lr": 5.9653640627313445e-05}
{"train_loss": 0.12390518188476562, "global_step": 119218, "epoch": 1339, "lr": 5.9653071798134854e-05}
{"train_loss": 0.14868071675300598, "global_step": 119219, "epoch": 1339, "lr": 5.96525029676585e-05}
{"train_loss": 0.1109071597456932, "global_step": 119220, "epoch": 1339, "lr": 5.965193413588449e-05}
{"train_loss": 0.198205828666687, "global_step": 119221, "epoch": 1339, "lr": 5.9651365302812903e-05}
{"train_loss": 0.1302606463432312, "global_step": 119222, "epoch": 1339, "lr": 5.965079646844379e-05}
{"train_loss": 0.20510834455490112, "global_step": 119223, "epoch": 1339, "lr": 5.965022763277726e-05}
{"train_loss": 0.13360868394374847, "global_step": 119224, "epoch": 1339, "lr": 5.964965879581337e-05}
{"train_loss": 0.10314659774303436, "global_step": 119225, "epoch": 1339, "lr": 5.964908995755221e-05}
{"train_loss": 0.2006157636642456, "global_step": 119226, "epoch": 1339, "lr": 5.964852111799384e-05}
{"train_loss": 0.17949259281158447, "global_step": 119227, "epoch": 1339, "lr": 5.964795227713834e-05}
{"train_loss": 0.18584008514881134, "global_step": 119228, "epoch": 1339, "lr": 5.9647383434985795e-05}
{"train_loss": 0.1279514878988266, "global_step": 119229, "epoch": 1339, "lr": 5.9646814591536285e-05}
{"train_loss": 0.12288759648799896, "global_step": 119230, "epoch": 1339, "lr": 5.9646245746789865e-05}
{"train_loss": 0.19048874080181122, "global_step": 119231, "epoch": 1339, "lr": 5.9645676900746625e-05}
{"train_loss": 0.08682088553905487, "global_step": 119232, "epoch": 1339, "lr": 5.964510805340665e-05}
{"train_loss": 0.17266149818897247, "global_step": 119233, "epoch": 1339, "lr": 5.964453920477e-05}
{"train_loss": 0.14591100811958313, "global_step": 119234, "epoch": 1339, "lr": 5.9643970354836764e-05}
{"train_loss": 0.15281498432159424, "global_step": 119235, "epoch": 1339, "lr": 5.964340150360701e-05}
{"train_loss": 0.14766162633895874, "global_step": 119236, "epoch": 1339, "lr": 5.9642832651080825e-05}
{"train_loss": 0.16644538938999176, "global_step": 119237, "epoch": 1339, "lr": 5.964226379725828e-05}
{"train_loss": 0.19590754806995392, "global_step": 119238, "epoch": 1339, "lr": 5.9641694942139445e-05}
{"train_loss": 0.11817145347595215, "global_step": 119239, "epoch": 1339, "lr": 5.9641126085724395e-05}
{"train_loss": 0.09817704558372498, "global_step": 119240, "epoch": 1339, "lr": 5.964055722801323e-05}
{"train_loss": 0.13673189282417297, "global_step": 119241, "epoch": 1339, "lr": 5.963998836900599e-05}
{"train_loss": 0.1544693112373352, "global_step": 119242, "epoch": 1339, "lr": 5.963941950870279e-05}
{"train_loss": 0.1503460854291916, "global_step": 119243, "epoch": 1339, "lr": 5.9638850647103674e-05}
{"train_loss": 0.08146291971206665, "global_step": 119244, "epoch": 1339, "lr": 5.963828178420874e-05}
{"train_loss": 0.16296790540218353, "global_step": 119245, "epoch": 1339, "lr": 5.963771292001805e-05}
{"train_loss": 0.10628162324428558, "global_step": 119246, "epoch": 1339, "lr": 5.963714405453169e-05}
{"train_loss": 0.1570291966199875, "global_step": 119247, "epoch": 1339, "lr": 5.9636575187749734e-05}
{"train_loss": 0.06311918795108795, "global_step": 119248, "epoch": 1339, "lr": 5.963600631967225e-05}
{"train_loss": 0.1568765789270401, "global_step": 119249, "epoch": 1339, "lr": 5.963543745029933e-05}
{"train_loss": 0.12468355149030685, "global_step": 119250, "epoch": 1339, "lr": 5.963486857963104e-05}
{"train_loss": 0.11632294207811356, "global_step": 119251, "epoch": 1339, "lr": 5.963429970766746e-05}
{"train_loss": 0.10925580561161041, "global_step": 119252, "epoch": 1339, "lr": 5.963373083440867e-05}
{"train_loss": 0.1389528512954712, "global_step": 119253, "epoch": 1339, "lr": 5.9633161959854744e-05}
{"train_loss": 0.1505652666091919, "global_step": 119254, "epoch": 1339, "lr": 5.963259308400575e-05}
{"train_loss": 0.1471748948097229, "global_step": 119255, "epoch": 1339, "lr": 5.963202420686177e-05}
{"train_loss": 0.1892838329076767, "global_step": 119256, "epoch": 1339, "lr": 5.9631455328422894e-05}
{"train_loss": 0.11522045731544495, "global_step": 119257, "epoch": 1339, "lr": 5.963088644868917e-05}
{"train_loss": 0.10178055614233017, "global_step": 119258, "epoch": 1339, "lr": 5.9630317567660696e-05}
{"train_loss": 0.13912865762295348, "global_step": 119259, "epoch": 1339, "lr": 5.962974868533755e-05, "val_loss": 5.19875431060791}
{"train_loss": 0.12678369879722595, "global_step": 119260, "epoch": 1340, "lr": 5.962917980171979e-05}
{"train_loss": 0.16187365353107452, "global_step": 119261, "epoch": 1340, "lr": 5.962861091680752e-05}
{"train_loss": 0.10765034705400467, "global_step": 119262, "epoch": 1340, "lr": 5.9628042030600794e-05}
{"train_loss": 0.12850239872932434, "global_step": 119263, "epoch": 1340, "lr": 5.962747314309969e-05}
{"train_loss": 0.14427563548088074, "global_step": 119264, "epoch": 1340, "lr": 5.9626904254304296e-05}
{"train_loss": 0.15145088732242584, "global_step": 119265, "epoch": 1340, "lr": 5.962633536421467e-05}
{"train_loss": 0.11423113197088242, "global_step": 119266, "epoch": 1340, "lr": 5.9625766472830916e-05}
{"train_loss": 0.10975835472345352, "global_step": 119267, "epoch": 1340, "lr": 5.962519758015308e-05}
{"train_loss": 0.1853678673505783, "global_step": 119268, "epoch": 1340, "lr": 5.962462868618127e-05}
{"train_loss": 0.10735603421926498, "global_step": 119269, "epoch": 1340, "lr": 5.962405979091553e-05}
{"train_loss": 0.173542782664299, "global_step": 119270, "epoch": 1340, "lr": 5.962349089435597e-05}
{"train_loss": 0.11847380548715591, "global_step": 119271, "epoch": 1340, "lr": 5.9622921996502636e-05}
{"train_loss": 0.195090651512146, "global_step": 119272, "epoch": 1340, "lr": 5.962235309735562e-05}
{"train_loss": 0.12160839885473251, "global_step": 119273, "epoch": 1340, "lr": 5.9621784196914996e-05}
{"train_loss": 0.20085154473781586, "global_step": 119274, "epoch": 1340, "lr": 5.9621215295180844e-05}
{"train_loss": 0.21455688774585724, "global_step": 119275, "epoch": 1340, "lr": 5.962064639215323e-05}
{"train_loss": 0.16053983569145203, "global_step": 119276, "epoch": 1340, "lr": 5.962007748783224e-05}
{"train_loss": 0.10459521412849426, "global_step": 119277, "epoch": 1340, "lr": 5.961950858221795e-05}
{"train_loss": 0.10934492200613022, "global_step": 119278, "epoch": 1340, "lr": 5.9618939675310434e-05}
{"train_loss": 0.12120769172906876, "global_step": 119279, "epoch": 1340, "lr": 5.961837076710977e-05}
{"train_loss": 0.151679128408432, "global_step": 119280, "epoch": 1340, "lr": 5.961780185761603e-05}
{"train_loss": 0.11086452752351761, "global_step": 119281, "epoch": 1340, "lr": 5.9617232946829295e-05}
{"train_loss": 0.14203986525535583, "global_step": 119282, "epoch": 1340, "lr": 5.961666403474964e-05}
{"train_loss": 0.16837336122989655, "global_step": 119283, "epoch": 1340, "lr": 5.961609512137714e-05}
{"train_loss": 0.14419278502464294, "global_step": 119284, "epoch": 1340, "lr": 5.961552620671187e-05}
{"train_loss": 0.11444516479969025, "global_step": 119285, "epoch": 1340, "lr": 5.9614957290753925e-05}
{"train_loss": 0.12352430820465088, "global_step": 119286, "epoch": 1340, "lr": 5.961438837350335e-05}
{"train_loss": 0.1674480438232422, "global_step": 119287, "epoch": 1340, "lr": 5.961381945496024e-05}
{"train_loss": 0.15946762263774872, "global_step": 119288, "epoch": 1340, "lr": 5.9613250535124676e-05}
{"train_loss": 0.15482541918754578, "global_step": 119289, "epoch": 1340, "lr": 5.961268161399672e-05}
{"train_loss": 0.20256944000720978, "global_step": 119290, "epoch": 1340, "lr": 5.961211269157647e-05}
{"train_loss": 0.10053408890962601, "global_step": 119291, "epoch": 1340, "lr": 5.961154376786398e-05}
{"train_loss": 0.19039005041122437, "global_step": 119292, "epoch": 1340, "lr": 5.9610974842859336e-05}
{"train_loss": 0.16593152284622192, "global_step": 119293, "epoch": 1340, "lr": 5.9610405916562604e-05}
{"train_loss": 0.1639646291732788, "global_step": 119294, "epoch": 1340, "lr": 5.9609836988973875e-05}
{"train_loss": 0.151215597987175, "global_step": 119295, "epoch": 1340, "lr": 5.9609268060093225e-05}
{"train_loss": 0.20091690123081207, "global_step": 119296, "epoch": 1340, "lr": 5.9608699129920734e-05}
{"train_loss": 0.13646677136421204, "global_step": 119297, "epoch": 1340, "lr": 5.960813019845646e-05}
{"train_loss": 0.12521740794181824, "global_step": 119298, "epoch": 1340, "lr": 5.960756126570048e-05}
{"train_loss": 0.16018633544445038, "global_step": 119299, "epoch": 1340, "lr": 5.96069923316529e-05}
{"train_loss": 0.1271786391735077, "global_step": 119300, "epoch": 1340, "lr": 5.960642339631377e-05}
{"train_loss": 0.14625075459480286, "global_step": 119301, "epoch": 1340, "lr": 5.9605854459683167e-05}
{"train_loss": 0.15104305744171143, "global_step": 119302, "epoch": 1340, "lr": 5.960528552176118e-05}
{"train_loss": 0.15530019998550415, "global_step": 119303, "epoch": 1340, "lr": 5.9604716582547884e-05}
{"train_loss": 0.20796087384223938, "global_step": 119304, "epoch": 1340, "lr": 5.960414764204334e-05}
{"train_loss": 0.10031464695930481, "global_step": 119305, "epoch": 1340, "lr": 5.960357870024764e-05}
{"train_loss": 0.1837085336446762, "global_step": 119306, "epoch": 1340, "lr": 5.960300975716086e-05}
{"train_loss": 0.10768245160579681, "global_step": 119307, "epoch": 1340, "lr": 5.960244081278307e-05}
{"train_loss": 0.110283762216568, "global_step": 119308, "epoch": 1340, "lr": 5.960187186711435e-05}
{"train_loss": 0.07682563364505768, "global_step": 119309, "epoch": 1340, "lr": 5.960130292015477e-05}
{"train_loss": 0.1808328926563263, "global_step": 119310, "epoch": 1340, "lr": 5.960073397190441e-05}
{"train_loss": 0.21480967104434967, "global_step": 119311, "epoch": 1340, "lr": 5.9600165022363354e-05}
{"train_loss": 0.12488381564617157, "global_step": 119312, "epoch": 1340, "lr": 5.959959607153167e-05}
{"train_loss": 0.22713494300842285, "global_step": 119313, "epoch": 1340, "lr": 5.9599027119409443e-05}
{"train_loss": 0.08216889947652817, "global_step": 119314, "epoch": 1340, "lr": 5.959845816599674e-05}
{"train_loss": 0.13406114280223846, "global_step": 119315, "epoch": 1340, "lr": 5.959788921129365e-05}
{"train_loss": 0.1230560690164566, "global_step": 119316, "epoch": 1340, "lr": 5.9597320255300224e-05}
{"train_loss": 0.14351870119571686, "global_step": 119317, "epoch": 1340, "lr": 5.959675129801656e-05}
{"train_loss": 0.13073961436748505, "global_step": 119318, "epoch": 1340, "lr": 5.959618233944274e-05}
{"train_loss": 0.130842924118042, "global_step": 119319, "epoch": 1340, "lr": 5.959561337957882e-05}
{"train_loss": 0.1531025469303131, "global_step": 119320, "epoch": 1340, "lr": 5.959504441842489e-05}
{"train_loss": 0.18410056829452515, "global_step": 119321, "epoch": 1340, "lr": 5.959447545598102e-05}
{"train_loss": 0.1791837215423584, "global_step": 119322, "epoch": 1340, "lr": 5.95939064922473e-05}
{"train_loss": 0.1492493599653244, "global_step": 119323, "epoch": 1340, "lr": 5.959333752722379e-05}
{"train_loss": 0.05097724869847298, "global_step": 119324, "epoch": 1340, "lr": 5.959276856091057e-05}
{"train_loss": 0.09976869821548462, "global_step": 119325, "epoch": 1340, "lr": 5.959219959330772e-05}
{"train_loss": 0.12428317219018936, "global_step": 119326, "epoch": 1340, "lr": 5.959163062441532e-05}
{"train_loss": 0.10012085735797882, "global_step": 119327, "epoch": 1340, "lr": 5.9591061654233435e-05}
{"train_loss": 0.13657796382904053, "global_step": 119328, "epoch": 1340, "lr": 5.9590492682762145e-05}
{"train_loss": 0.117639921605587, "global_step": 119329, "epoch": 1340, "lr": 5.958992371000155e-05}
{"train_loss": 0.08308949321508408, "global_step": 119330, "epoch": 1340, "lr": 5.9589354735951695e-05}
{"train_loss": 0.14142097532749176, "global_step": 119331, "epoch": 1340, "lr": 5.958878576061266e-05}
{"train_loss": 0.10901914536952972, "global_step": 119332, "epoch": 1340, "lr": 5.958821678398454e-05}
{"train_loss": 0.1273803561925888, "global_step": 119333, "epoch": 1340, "lr": 5.9587647806067405e-05}
{"train_loss": 0.14132675528526306, "global_step": 119334, "epoch": 1340, "lr": 5.958707882686132e-05}
{"train_loss": 0.14821910858154297, "global_step": 119335, "epoch": 1340, "lr": 5.958650984636637e-05}
{"train_loss": 0.11024665087461472, "global_step": 119336, "epoch": 1340, "lr": 5.9585940864582634e-05}
{"train_loss": 0.11417930573225021, "global_step": 119337, "epoch": 1340, "lr": 5.958537188151018e-05}
{"train_loss": 0.1919688582420349, "global_step": 119338, "epoch": 1340, "lr": 5.958480289714909e-05}
{"train_loss": 0.09290861338376999, "global_step": 119339, "epoch": 1340, "lr": 5.958423391149944e-05}
{"train_loss": 0.17590218782424927, "global_step": 119340, "epoch": 1340, "lr": 5.958366492456131e-05}
{"train_loss": 0.22144261002540588, "global_step": 119341, "epoch": 1340, "lr": 5.958309593633478e-05}
{"train_loss": 0.1092468723654747, "global_step": 119342, "epoch": 1340, "lr": 5.9582526946819905e-05}
{"train_loss": 0.13085928559303284, "global_step": 119343, "epoch": 1340, "lr": 5.9581957956016776e-05}
{"train_loss": 0.13450099527835846, "global_step": 119344, "epoch": 1340, "lr": 5.958138896392548e-05}
{"train_loss": 0.15195636451244354, "global_step": 119345, "epoch": 1340, "lr": 5.9580819970546074e-05}
{"train_loss": 0.28125131130218506, "global_step": 119346, "epoch": 1340, "lr": 5.958025097587864e-05}
{"train_loss": 0.1300940066576004, "global_step": 119347, "epoch": 1340, "lr": 5.9579681979923265e-05}
{"train_loss": 0.14455454909567084, "global_step": 119348, "epoch": 1340, "lr": 5.957911298268002e-05, "val_loss": 5.481131553649902, "train_action_mse_error": 13.196493148803711}
{"train_loss": 0.10191014409065247, "global_step": 119349, "epoch": 1341, "lr": 5.957854398414899e-05}
{"train_loss": 0.11161109060049057, "global_step": 119350, "epoch": 1341, "lr": 5.957797498433023e-05}
{"train_loss": 0.14738395810127258, "global_step": 119351, "epoch": 1341, "lr": 5.9577405983223834e-05}
{"train_loss": 0.12440270185470581, "global_step": 119352, "epoch": 1341, "lr": 5.9576836980829865e-05}
{"train_loss": 0.07820005714893341, "global_step": 119353, "epoch": 1341, "lr": 5.957626797714841e-05}
{"train_loss": 0.15985049307346344, "global_step": 119354, "epoch": 1341, "lr": 5.957569897217954e-05}
{"train_loss": 0.13000565767288208, "global_step": 119355, "epoch": 1341, "lr": 5.9575129965923336e-05}
{"train_loss": 0.16856564581394196, "global_step": 119356, "epoch": 1341, "lr": 5.9574560958379874e-05}
{"train_loss": 0.18687263131141663, "global_step": 119357, "epoch": 1341, "lr": 5.957399194954922e-05}
{"train_loss": 0.10243827849626541, "global_step": 119358, "epoch": 1341, "lr": 5.957342293943148e-05}
{"train_loss": 0.09290071576833725, "global_step": 119359, "epoch": 1341, "lr": 5.95728539280267e-05}
{"train_loss": 0.10070837289094925, "global_step": 119360, "epoch": 1341, "lr": 5.9572284915334955e-05}
{"train_loss": 0.10254938900470734, "global_step": 119361, "epoch": 1341, "lr": 5.957171590135634e-05}
{"train_loss": 0.11268650740385056, "global_step": 119362, "epoch": 1341, "lr": 5.9571146886090926e-05}
{"train_loss": 0.10854543000459671, "global_step": 119363, "epoch": 1341, "lr": 5.957057786953879e-05}
{"train_loss": 0.13367952406406403, "global_step": 119364, "epoch": 1341, "lr": 5.9570008851699986e-05}
{"train_loss": 0.1726229190826416, "global_step": 119365, "epoch": 1341, "lr": 5.9569439832574636e-05}
{"train_loss": 0.20126214623451233, "global_step": 119366, "epoch": 1341, "lr": 5.956887081216278e-05}
{"train_loss": 0.1271006166934967, "global_step": 119367, "epoch": 1341, "lr": 5.956830179046451e-05}
{"train_loss": 0.12019135057926178, "global_step": 119368, "epoch": 1341, "lr": 5.9567732767479914e-05}
{"train_loss": 0.11608339101076126, "global_step": 119369, "epoch": 1341, "lr": 5.956716374320903e-05}
{"train_loss": 0.06691048294305801, "global_step": 119370, "epoch": 1341, "lr": 5.956659471765197e-05}
{"train_loss": 0.11427117139101028, "global_step": 119371, "epoch": 1341, "lr": 5.95660256908088e-05}
{"train_loss": 0.1589169204235077, "global_step": 119372, "epoch": 1341, "lr": 5.9565456662679584e-05}
{"train_loss": 0.2564195990562439, "global_step": 119373, "epoch": 1341, "lr": 5.956488763326441e-05}
{"train_loss": 0.1509825587272644, "global_step": 119374, "epoch": 1341, "lr": 5.956431860256336e-05}
{"train_loss": 0.14136627316474915, "global_step": 119375, "epoch": 1341, "lr": 5.95637495705765e-05}
{"train_loss": 0.09508725255727768, "global_step": 119376, "epoch": 1341, "lr": 5.956318053730391e-05}
{"train_loss": 0.13435505330562592, "global_step": 119377, "epoch": 1341, "lr": 5.956261150274567e-05}
{"train_loss": 0.12419264018535614, "global_step": 119378, "epoch": 1341, "lr": 5.9562042466901845e-05}
{"train_loss": 0.18822067975997925, "global_step": 119379, "epoch": 1341, "lr": 5.956147342977253e-05}
{"train_loss": 0.15694165229797363, "global_step": 119380, "epoch": 1341, "lr": 5.956090439135779e-05}
{"train_loss": 0.09810872375965118, "global_step": 119381, "epoch": 1341, "lr": 5.956033535165769e-05}
{"train_loss": 0.1863914579153061, "global_step": 119382, "epoch": 1341, "lr": 5.955976631067234e-05}
{"train_loss": 0.13217613101005554, "global_step": 119383, "epoch": 1341, "lr": 5.9559197268401765e-05}
{"train_loss": 0.11118366569280624, "global_step": 119384, "epoch": 1341, "lr": 5.95586282248461e-05}
{"train_loss": 0.12126919627189636, "global_step": 119385, "epoch": 1341, "lr": 5.955805918000539e-05}
{"train_loss": 0.14290618896484375, "global_step": 119386, "epoch": 1341, "lr": 5.955749013387971e-05}
{"train_loss": 0.1338782161474228, "global_step": 119387, "epoch": 1341, "lr": 5.955692108646915e-05}
{"train_loss": 0.12369292974472046, "global_step": 119388, "epoch": 1341, "lr": 5.955635203777378e-05}
{"train_loss": 0.13207755982875824, "global_step": 119389, "epoch": 1341, "lr": 5.955578298779366e-05}
{"train_loss": 0.13282744586467743, "global_step": 119390, "epoch": 1341, "lr": 5.9555213936528894e-05}
{"train_loss": 0.17151913046836853, "global_step": 119391, "epoch": 1341, "lr": 5.9554644883979536e-05}
{"train_loss": 0.11204759776592255, "global_step": 119392, "epoch": 1341, "lr": 5.955407583014568e-05}
{"train_loss": 0.1250845342874527, "global_step": 119393, "epoch": 1341, "lr": 5.955350677502739e-05}
{"train_loss": 0.11645901948213577, "global_step": 119394, "epoch": 1341, "lr": 5.9552937718624756e-05}
{"train_loss": 0.1323978453874588, "global_step": 119395, "epoch": 1341, "lr": 5.955236866093784e-05}
{"train_loss": 0.1131691262125969, "global_step": 119396, "epoch": 1341, "lr": 5.955179960196672e-05}
{"train_loss": 0.09188497066497803, "global_step": 119397, "epoch": 1341, "lr": 5.955123054171148e-05}
{"train_loss": 0.16322264075279236, "global_step": 119398, "epoch": 1341, "lr": 5.9550661480172196e-05}
{"train_loss": 0.14876115322113037, "global_step": 119399, "epoch": 1341, "lr": 5.9550092417348945e-05}
{"train_loss": 0.2033306509256363, "global_step": 119400, "epoch": 1341, "lr": 5.95495233532418e-05}
{"train_loss": 0.13241970539093018, "global_step": 119401, "epoch": 1341, "lr": 5.9548954287850814e-05}
{"train_loss": 0.10636886954307556, "global_step": 119402, "epoch": 1341, "lr": 5.9548385221176104e-05}
{"train_loss": 0.07881955802440643, "global_step": 119403, "epoch": 1341, "lr": 5.954781615321774e-05}
{"train_loss": 0.10189356654882431, "global_step": 119404, "epoch": 1341, "lr": 5.954724708397579e-05}
{"train_loss": 0.13177096843719482, "global_step": 119405, "epoch": 1341, "lr": 5.954667801345031e-05}
{"train_loss": 0.16397300362586975, "global_step": 119406, "epoch": 1341, "lr": 5.954610894164141e-05}
{"train_loss": 0.16526256501674652, "global_step": 119407, "epoch": 1341, "lr": 5.954553986854916e-05}
{"train_loss": 0.13416844606399536, "global_step": 119408, "epoch": 1341, "lr": 5.954497079417362e-05}
{"train_loss": 0.16506525874137878, "global_step": 119409, "epoch": 1341, "lr": 5.9544401718514876e-05}
{"train_loss": 0.17819514870643616, "global_step": 119410, "epoch": 1341, "lr": 5.9543832641572995e-05}
{"train_loss": 0.11676299571990967, "global_step": 119411, "epoch": 1341, "lr": 5.954326356334806e-05}
{"train_loss": 0.17424547672271729, "global_step": 119412, "epoch": 1341, "lr": 5.954269448384015e-05}
{"train_loss": 0.1776585876941681, "global_step": 119413, "epoch": 1341, "lr": 5.9542125403049354e-05}
{"train_loss": 0.10007765144109726, "global_step": 119414, "epoch": 1341, "lr": 5.9541556320975735e-05}
{"train_loss": 0.133445605635643, "global_step": 119415, "epoch": 1341, "lr": 5.954098723761936e-05}
{"train_loss": 0.16804999113082886, "global_step": 119416, "epoch": 1341, "lr": 5.954041815298033e-05}
{"train_loss": 0.14829103648662567, "global_step": 119417, "epoch": 1341, "lr": 5.9539849067058686e-05}
{"train_loss": 0.22719639539718628, "global_step": 119418, "epoch": 1341, "lr": 5.9539279979854534e-05}
{"train_loss": 0.14910051226615906, "global_step": 119419, "epoch": 1341, "lr": 5.953871089136793e-05}
{"train_loss": 0.1276271939277649, "global_step": 119420, "epoch": 1341, "lr": 5.953814180159898e-05}
{"train_loss": 0.13857248425483704, "global_step": 119421, "epoch": 1341, "lr": 5.953757271054774e-05}
{"train_loss": 0.08079854398965836, "global_step": 119422, "epoch": 1341, "lr": 5.953700361821429e-05}
{"train_loss": 0.1028369665145874, "global_step": 119423, "epoch": 1341, "lr": 5.95364345245987e-05}
{"train_loss": 0.11665230989456177, "global_step": 119424, "epoch": 1341, "lr": 5.9535865429701066e-05}
{"train_loss": 0.24871531128883362, "global_step": 119425, "epoch": 1341, "lr": 5.953529633352144e-05}
{"train_loss": 0.14850397408008575, "global_step": 119426, "epoch": 1341, "lr": 5.953472723605993e-05}
{"train_loss": 0.1630672812461853, "global_step": 119427, "epoch": 1341, "lr": 5.953415813731658e-05}
{"train_loss": 0.13764214515686035, "global_step": 119428, "epoch": 1341, "lr": 5.953358903729147e-05}
{"train_loss": 0.16147321462631226, "global_step": 119429, "epoch": 1341, "lr": 5.953301993598468e-05}
{"train_loss": 0.12094010412693024, "global_step": 119430, "epoch": 1341, "lr": 5.95324508333963e-05}
{"train_loss": 0.08450067788362503, "global_step": 119431, "epoch": 1341, "lr": 5.9531881729526404e-05}
{"train_loss": 0.20220670104026794, "global_step": 119432, "epoch": 1341, "lr": 5.953131262437506e-05}
{"train_loss": 0.1515214592218399, "global_step": 119433, "epoch": 1341, "lr": 5.9530743517942345e-05}
{"train_loss": 0.10368774086236954, "global_step": 119434, "epoch": 1341, "lr": 5.953017441022833e-05}
{"train_loss": 0.14285238087177277, "global_step": 119435, "epoch": 1341, "lr": 5.952960530123311e-05}
{"train_loss": 0.09132511913776398, "global_step": 119436, "epoch": 1341, "lr": 5.9529036190956745e-05}
{"train_loss": 0.1377348621909538, "global_step": 119437, "epoch": 1341, "lr": 5.952846707939931e-05, "val_loss": 5.132782936096191}
{"train_loss": 0.13097307085990906, "global_step": 119438, "epoch": 1342, "lr": 5.95278979665609e-05}
{"train_loss": 0.09304860979318619, "global_step": 119439, "epoch": 1342, "lr": 5.9527328852441576e-05}
{"train_loss": 0.11990173906087875, "global_step": 119440, "epoch": 1342, "lr": 5.9526759737041425e-05}
{"train_loss": 0.16623514890670776, "global_step": 119441, "epoch": 1342, "lr": 5.952619062036051e-05}
{"train_loss": 0.15897731482982635, "global_step": 119442, "epoch": 1342, "lr": 5.952562150239892e-05}
{"train_loss": 0.14533092081546783, "global_step": 119443, "epoch": 1342, "lr": 5.952505238315672e-05}
{"train_loss": 0.16397401690483093, "global_step": 119444, "epoch": 1342, "lr": 5.9524483262634e-05}
{"train_loss": 0.11960678547620773, "global_step": 119445, "epoch": 1342, "lr": 5.952391414083084e-05}
{"train_loss": 0.11050307750701904, "global_step": 119446, "epoch": 1342, "lr": 5.952334501774729e-05}
{"train_loss": 0.08870167285203934, "global_step": 119447, "epoch": 1342, "lr": 5.952277589338344e-05}
{"train_loss": 0.1554097831249237, "global_step": 119448, "epoch": 1342, "lr": 5.9522206767739375e-05}
{"train_loss": 0.1337583214044571, "global_step": 119449, "epoch": 1342, "lr": 5.9521637640815164e-05}
{"train_loss": 0.13709059357643127, "global_step": 119450, "epoch": 1342, "lr": 5.952106851261089e-05}
{"train_loss": 0.13161684572696686, "global_step": 119451, "epoch": 1342, "lr": 5.952049938312661e-05}
{"train_loss": 0.08016438782215118, "global_step": 119452, "epoch": 1342, "lr": 5.951993025236243e-05}
{"train_loss": 0.16941618919372559, "global_step": 119453, "epoch": 1342, "lr": 5.9519361120318406e-05}
{"train_loss": 0.25074151158332825, "global_step": 119454, "epoch": 1342, "lr": 5.951879198699462e-05}
{"train_loss": 0.1724962443113327, "global_step": 119455, "epoch": 1342, "lr": 5.951822285239114e-05}
{"train_loss": 0.13769793510437012, "global_step": 119456, "epoch": 1342, "lr": 5.951765371650806e-05}
{"train_loss": 0.14472700655460358, "global_step": 119457, "epoch": 1342, "lr": 5.951708457934544e-05}
{"train_loss": 0.22423988580703735, "global_step": 119458, "epoch": 1342, "lr": 5.9516515440903376e-05}
{"train_loss": 0.09325168281793594, "global_step": 119459, "epoch": 1342, "lr": 5.951594630118193e-05}
{"train_loss": 0.09337446838617325, "global_step": 119460, "epoch": 1342, "lr": 5.9515377160181174e-05}
{"train_loss": 0.145116925239563, "global_step": 119461, "epoch": 1342, "lr": 5.951480801790119e-05}
{"train_loss": 0.1389668583869934, "global_step": 119462, "epoch": 1342, "lr": 5.9514238874342066e-05}
{"train_loss": 0.11767695844173431, "global_step": 119463, "epoch": 1342, "lr": 5.951366972950386e-05}
{"train_loss": 0.1335894614458084, "global_step": 119464, "epoch": 1342, "lr": 5.951310058338666e-05}
{"train_loss": 0.13898487389087677, "global_step": 119465, "epoch": 1342, "lr": 5.951253143599056e-05}
{"train_loss": 0.17198136448860168, "global_step": 119466, "epoch": 1342, "lr": 5.951196228731559e-05}
{"train_loss": 0.15229177474975586, "global_step": 119467, "epoch": 1342, "lr": 5.951139313736186e-05}
{"train_loss": 0.18724490702152252, "global_step": 119468, "epoch": 1342, "lr": 5.951082398612944e-05}
{"train_loss": 0.2218451052904129, "global_step": 119469, "epoch": 1342, "lr": 5.9510254833618404e-05}
{"train_loss": 0.13367155194282532, "global_step": 119470, "epoch": 1342, "lr": 5.950968567982883e-05}
{"train_loss": 0.21866080164909363, "global_step": 119471, "epoch": 1342, "lr": 5.950911652476079e-05}
{"train_loss": 0.1915161907672882, "global_step": 119472, "epoch": 1342, "lr": 5.950854736841437e-05}
{"train_loss": 0.1786022037267685, "global_step": 119473, "epoch": 1342, "lr": 5.950797821078964e-05}
{"train_loss": 0.14772279560565948, "global_step": 119474, "epoch": 1342, "lr": 5.9507409051886674e-05}
{"train_loss": 0.17227286100387573, "global_step": 119475, "epoch": 1342, "lr": 5.950683989170556e-05}
{"train_loss": 0.10284590721130371, "global_step": 119476, "epoch": 1342, "lr": 5.950627073024636e-05}
{"train_loss": 0.18405532836914062, "global_step": 119477, "epoch": 1342, "lr": 5.9505701567509166e-05}
{"train_loss": 0.18323393166065216, "global_step": 119478, "epoch": 1342, "lr": 5.9505132403494046e-05}
{"train_loss": 0.11333193629980087, "global_step": 119479, "epoch": 1342, "lr": 5.950456323820107e-05}
{"train_loss": 0.15357349812984467, "global_step": 119480, "epoch": 1342, "lr": 5.950399407163033e-05}
{"train_loss": 0.16803936660289764, "global_step": 119481, "epoch": 1342, "lr": 5.950342490378189e-05}
{"train_loss": 0.14955197274684906, "global_step": 119482, "epoch": 1342, "lr": 5.9502855734655825e-05}
{"train_loss": 0.1486043632030487, "global_step": 119483, "epoch": 1342, "lr": 5.950228656425223e-05}
{"train_loss": 0.12400945276021957, "global_step": 119484, "epoch": 1342, "lr": 5.950171739257117e-05}
{"train_loss": 0.2212313562631607, "global_step": 119485, "epoch": 1342, "lr": 5.9501148219612716e-05}
{"train_loss": 0.14078818261623383, "global_step": 119486, "epoch": 1342, "lr": 5.950057904537694e-05}
{"train_loss": 0.1512811928987503, "global_step": 119487, "epoch": 1342, "lr": 5.950000986986393e-05}
{"train_loss": 0.09241096675395966, "global_step": 119488, "epoch": 1342, "lr": 5.949944069307375e-05}
{"train_loss": 0.1266462504863739, "global_step": 119489, "epoch": 1342, "lr": 5.9498871515006496e-05}
{"train_loss": 0.11742807179689407, "global_step": 119490, "epoch": 1342, "lr": 5.9498302335662235e-05}
{"train_loss": 0.11891698837280273, "global_step": 119491, "epoch": 1342, "lr": 5.9497733155041034e-05}
{"train_loss": 0.16329209506511688, "global_step": 119492, "epoch": 1342, "lr": 5.9497163973142976e-05}
{"train_loss": 0.08142245560884476, "global_step": 119493, "epoch": 1342, "lr": 5.949659478996816e-05}
{"train_loss": 0.10870827734470367, "global_step": 119494, "epoch": 1342, "lr": 5.9496025605516636e-05}
{"train_loss": 0.11015757918357849, "global_step": 119495, "epoch": 1342, "lr": 5.949545641978848e-05}
{"train_loss": 0.14179830253124237, "global_step": 119496, "epoch": 1342, "lr": 5.949488723278378e-05}
{"train_loss": 0.21327762305736542, "global_step": 119497, "epoch": 1342, "lr": 5.949431804450261e-05}
{"train_loss": 0.18242336809635162, "global_step": 119498, "epoch": 1342, "lr": 5.949374885494504e-05}
{"train_loss": 0.17149026691913605, "global_step": 119499, "epoch": 1342, "lr": 5.949317966411115e-05}
{"train_loss": 0.14559301733970642, "global_step": 119500, "epoch": 1342, "lr": 5.949261047200102e-05}
{"train_loss": 0.15175867080688477, "global_step": 119501, "epoch": 1342, "lr": 5.949204127861473e-05}
{"train_loss": 0.11931359022855759, "global_step": 119502, "epoch": 1342, "lr": 5.949147208395235e-05}
{"train_loss": 0.10152755677700043, "global_step": 119503, "epoch": 1342, "lr": 5.9490902888013955e-05}
{"train_loss": 0.11802566796541214, "global_step": 119504, "epoch": 1342, "lr": 5.949033369079963e-05}
{"train_loss": 0.10110349953174591, "global_step": 119505, "epoch": 1342, "lr": 5.948976449230943e-05}
{"train_loss": 0.09192751348018646, "global_step": 119506, "epoch": 1342, "lr": 5.9489195292543454e-05}
{"train_loss": 0.1425623744726181, "global_step": 119507, "epoch": 1342, "lr": 5.9488626091501774e-05}
{"train_loss": 0.17771200835704803, "global_step": 119508, "epoch": 1342, "lr": 5.948805688918446e-05}
{"train_loss": 0.134750097990036, "global_step": 119509, "epoch": 1342, "lr": 5.9487487685591605e-05}
{"train_loss": 0.16523315012454987, "global_step": 119510, "epoch": 1342, "lr": 5.948691848072324e-05}
{"train_loss": 0.08060961216688156, "global_step": 119511, "epoch": 1342, "lr": 5.94863492745795e-05}
{"train_loss": 0.24547182023525238, "global_step": 119512, "epoch": 1342, "lr": 5.9485780067160434e-05}
{"train_loss": 0.11036170274019241, "global_step": 119513, "epoch": 1342, "lr": 5.9485210858466115e-05}
{"train_loss": 0.11719147115945816, "global_step": 119514, "epoch": 1342, "lr": 5.948464164849663e-05}
{"train_loss": 0.09302252531051636, "global_step": 119515, "epoch": 1342, "lr": 5.948407243725205e-05}
{"train_loss": 0.1014111340045929, "global_step": 119516, "epoch": 1342, "lr": 5.948350322473245e-05}
{"train_loss": 0.1421511322259903, "global_step": 119517, "epoch": 1342, "lr": 5.9482934010937905e-05}
{"train_loss": 0.11151619255542755, "global_step": 119518, "epoch": 1342, "lr": 5.9482364795868505e-05}
{"train_loss": 0.13876290619373322, "global_step": 119519, "epoch": 1342, "lr": 5.948179557952431e-05}
{"train_loss": 0.10673092305660248, "global_step": 119520, "epoch": 1342, "lr": 5.9481226361905405e-05}
{"train_loss": 0.04408659040927887, "global_step": 119521, "epoch": 1342, "lr": 5.948065714301187e-05}
{"train_loss": 0.08141960203647614, "global_step": 119522, "epoch": 1342, "lr": 5.948008792284376e-05}
{"train_loss": 0.09863963723182678, "global_step": 119523, "epoch": 1342, "lr": 5.947951870140119e-05}
{"train_loss": 0.15293554961681366, "global_step": 119524, "epoch": 1342, "lr": 5.947894947868421e-05}
{"train_loss": 0.17903034389019012, "global_step": 119525, "epoch": 1342, "lr": 5.947838025469289e-05}
{"train_loss": 0.14072275957029856, "global_step": 119526, "epoch": 1342, "lr": 5.947781102942732e-05, "val_loss": 5.380631446838379}
{"train_loss": 0.12762971222400665, "global_step": 119527, "epoch": 1343, "lr": 5.947724180288758e-05}
{"train_loss": 0.10355976223945618, "global_step": 119528, "epoch": 1343, "lr": 5.9476672575073723e-05}
{"train_loss": 0.11944255232810974, "global_step": 119529, "epoch": 1343, "lr": 5.9476103345985853e-05}
{"train_loss": 0.13033434748649597, "global_step": 119530, "epoch": 1343, "lr": 5.9475534115624046e-05}
{"train_loss": 0.10422702878713608, "global_step": 119531, "epoch": 1343, "lr": 5.947496488398836e-05}
{"train_loss": 0.16209785640239716, "global_step": 119532, "epoch": 1343, "lr": 5.9474395651078875e-05}
{"train_loss": 0.11774758249521255, "global_step": 119533, "epoch": 1343, "lr": 5.947382641689568e-05}
{"train_loss": 0.12222112715244293, "global_step": 119534, "epoch": 1343, "lr": 5.947325718143885e-05}
{"train_loss": 0.1150052472949028, "global_step": 119535, "epoch": 1343, "lr": 5.947268794470844e-05}
{"train_loss": 0.12607456743717194, "global_step": 119536, "epoch": 1343, "lr": 5.9472118706704564e-05}
{"train_loss": 0.14162953197956085, "global_step": 119537, "epoch": 1343, "lr": 5.947154946742725e-05}
{"train_loss": 0.15579232573509216, "global_step": 119538, "epoch": 1343, "lr": 5.947098022687663e-05}
{"train_loss": 0.13831475377082825, "global_step": 119539, "epoch": 1343, "lr": 5.947041098505274e-05}
{"train_loss": 0.10622357577085495, "global_step": 119540, "epoch": 1343, "lr": 5.946984174195567e-05}
{"train_loss": 0.13915933668613434, "global_step": 119541, "epoch": 1343, "lr": 5.946927249758549e-05}
{"train_loss": 0.1345239132642746, "global_step": 119542, "epoch": 1343, "lr": 5.946870325194229e-05}
{"train_loss": 0.1425136923789978, "global_step": 119543, "epoch": 1343, "lr": 5.946813400502613e-05}
{"train_loss": 0.15315291285514832, "global_step": 119544, "epoch": 1343, "lr": 5.94675647568371e-05}
{"train_loss": 0.12991566956043243, "global_step": 119545, "epoch": 1343, "lr": 5.946699550737528e-05}
{"train_loss": 0.15277892351150513, "global_step": 119546, "epoch": 1343, "lr": 5.946642625664072e-05}
{"train_loss": 0.09251941740512848, "global_step": 119547, "epoch": 1343, "lr": 5.946585700463352e-05}
{"train_loss": 0.11980869621038437, "global_step": 119548, "epoch": 1343, "lr": 5.946528775135375e-05}
{"train_loss": 0.14552830159664154, "global_step": 119549, "epoch": 1343, "lr": 5.946471849680149e-05}
{"train_loss": 0.08889046311378479, "global_step": 119550, "epoch": 1343, "lr": 5.946414924097681e-05}
{"train_loss": 0.1230895072221756, "global_step": 119551, "epoch": 1343, "lr": 5.9463579983879794e-05}
{"train_loss": 0.13106438517570496, "global_step": 119552, "epoch": 1343, "lr": 5.9463010725510516e-05}
{"train_loss": 0.10291114449501038, "global_step": 119553, "epoch": 1343, "lr": 5.946244146586904e-05}
{"train_loss": 0.13627201318740845, "global_step": 119554, "epoch": 1343, "lr": 5.946187220495547e-05}
{"train_loss": 0.09863655269145966, "global_step": 119555, "epoch": 1343, "lr": 5.946130294276987e-05}
{"train_loss": 0.1323128342628479, "global_step": 119556, "epoch": 1343, "lr": 5.9460733679312294e-05}
{"train_loss": 0.12296660244464874, "global_step": 119557, "epoch": 1343, "lr": 5.9460164414582856e-05}
{"train_loss": 0.12879133224487305, "global_step": 119558, "epoch": 1343, "lr": 5.94595951485816e-05}
{"train_loss": 0.11983651667833328, "global_step": 119559, "epoch": 1343, "lr": 5.9459025881308626e-05}
{"train_loss": 0.06568038463592529, "global_step": 119560, "epoch": 1343, "lr": 5.9458456612764e-05}
{"train_loss": 0.14269030094146729, "global_step": 119561, "epoch": 1343, "lr": 5.9457887342947814e-05}
{"train_loss": 0.16710606217384338, "global_step": 119562, "epoch": 1343, "lr": 5.9457318071860114e-05}
{"train_loss": 0.10670789331197739, "global_step": 119563, "epoch": 1343, "lr": 5.9456748799500996e-05}
{"train_loss": 0.13885636627674103, "global_step": 119564, "epoch": 1343, "lr": 5.9456179525870516e-05}
{"train_loss": 0.1477525681257248, "global_step": 119565, "epoch": 1343, "lr": 5.945561025096878e-05}
{"train_loss": 0.14909186959266663, "global_step": 119566, "epoch": 1343, "lr": 5.945504097479586e-05}
{"train_loss": 0.0734439268708229, "global_step": 119567, "epoch": 1343, "lr": 5.945447169735182e-05}
{"train_loss": 0.0938488245010376, "global_step": 119568, "epoch": 1343, "lr": 5.945390241863674e-05}
{"train_loss": 0.14937885105609894, "global_step": 119569, "epoch": 1343, "lr": 5.94533331386507e-05}
{"train_loss": 0.12701690196990967, "global_step": 119570, "epoch": 1343, "lr": 5.945276385739378e-05}
{"train_loss": 0.08042021095752716, "global_step": 119571, "epoch": 1343, "lr": 5.945219457486604e-05}
{"train_loss": 0.10075348615646362, "global_step": 119572, "epoch": 1343, "lr": 5.945162529106758e-05}
{"train_loss": 0.1015094667673111, "global_step": 119573, "epoch": 1343, "lr": 5.9451056005998454e-05}
{"train_loss": 0.08288554102182388, "global_step": 119574, "epoch": 1343, "lr": 5.9450486719658746e-05}
{"train_loss": 0.09441208839416504, "global_step": 119575, "epoch": 1343, "lr": 5.944991743204855e-05}
{"train_loss": 0.12552712857723236, "global_step": 119576, "epoch": 1343, "lr": 5.944934814316791e-05}
{"train_loss": 0.11792725324630737, "global_step": 119577, "epoch": 1343, "lr": 5.944877885301694e-05}
{"train_loss": 0.09473469108343124, "global_step": 119578, "epoch": 1343, "lr": 5.944820956159568e-05}
{"train_loss": 0.1536041498184204, "global_step": 119579, "epoch": 1343, "lr": 5.9447640268904235e-05}
{"train_loss": 0.10311958938837051, "global_step": 119580, "epoch": 1343, "lr": 5.944707097494267e-05}
{"train_loss": 0.10345761477947235, "global_step": 119581, "epoch": 1343, "lr": 5.944650167971105e-05}
{"train_loss": 0.10245922207832336, "global_step": 119582, "epoch": 1343, "lr": 5.944593238320947e-05}
{"train_loss": 0.06498182564973831, "global_step": 119583, "epoch": 1343, "lr": 5.9445363085438e-05}
{"train_loss": 0.11636755615472794, "global_step": 119584, "epoch": 1343, "lr": 5.9444793786396716e-05}
{"train_loss": 0.1299036294221878, "global_step": 119585, "epoch": 1343, "lr": 5.944422448608569e-05}
{"train_loss": 0.14447826147079468, "global_step": 119586, "epoch": 1343, "lr": 5.9443655184505e-05}
{"train_loss": 0.12355218082666397, "global_step": 119587, "epoch": 1343, "lr": 5.9443085881654734e-05}
{"train_loss": 0.12024668604135513, "global_step": 119588, "epoch": 1343, "lr": 5.944251657753496e-05}
{"train_loss": 0.10068386793136597, "global_step": 119589, "epoch": 1343, "lr": 5.944194727214575e-05}
{"train_loss": 0.09103494137525558, "global_step": 119590, "epoch": 1343, "lr": 5.944137796548719e-05}
{"train_loss": 0.15312141180038452, "global_step": 119591, "epoch": 1343, "lr": 5.9440808657559345e-05}
{"train_loss": 0.17562510073184967, "global_step": 119592, "epoch": 1343, "lr": 5.944023934836232e-05}
{"train_loss": 0.11460153758525848, "global_step": 119593, "epoch": 1343, "lr": 5.943967003789614e-05}
{"train_loss": 0.10521502792835236, "global_step": 119594, "epoch": 1343, "lr": 5.9439100726160924e-05}
{"train_loss": 0.1253732442855835, "global_step": 119595, "epoch": 1343, "lr": 5.943853141315674e-05}
{"train_loss": 0.10634902864694595, "global_step": 119596, "epoch": 1343, "lr": 5.943796209888366e-05}
{"train_loss": 0.11422505229711533, "global_step": 119597, "epoch": 1343, "lr": 5.943739278334175e-05}
{"train_loss": 0.06091314181685448, "global_step": 119598, "epoch": 1343, "lr": 5.9436823466531113e-05}
{"train_loss": 0.18142642080783844, "global_step": 119599, "epoch": 1343, "lr": 5.94362541484518e-05}
{"train_loss": 0.1469530463218689, "global_step": 119600, "epoch": 1343, "lr": 5.9435684829103896e-05}
{"train_loss": 0.1408914029598236, "global_step": 119601, "epoch": 1343, "lr": 5.943511550848748e-05}
{"train_loss": 0.08374523371458054, "global_step": 119602, "epoch": 1343, "lr": 5.943454618660263e-05}
{"train_loss": 0.08980826288461685, "global_step": 119603, "epoch": 1343, "lr": 5.943397686344942e-05}
{"train_loss": 0.1516733020544052, "global_step": 119604, "epoch": 1343, "lr": 5.943340753902794e-05}
{"train_loss": 0.09156981855630875, "global_step": 119605, "epoch": 1343, "lr": 5.943283821333824e-05}
{"train_loss": 0.22318539023399353, "global_step": 119606, "epoch": 1343, "lr": 5.9432268886380415e-05}
{"train_loss": 0.11871207505464554, "global_step": 119607, "epoch": 1343, "lr": 5.943169955815453e-05}
{"train_loss": 0.14446455240249634, "global_step": 119608, "epoch": 1343, "lr": 5.943113022866067e-05}
{"train_loss": 0.08145423233509064, "global_step": 119609, "epoch": 1343, "lr": 5.943056089789891e-05}
{"train_loss": 0.15410448610782623, "global_step": 119610, "epoch": 1343, "lr": 5.9429991565869326e-05}
{"train_loss": 0.12156561017036438, "global_step": 119611, "epoch": 1343, "lr": 5.9429422232571995e-05}
{"train_loss": 0.23445887863636017, "global_step": 119612, "epoch": 1343, "lr": 5.9428852898006995e-05}
{"train_loss": 0.1187170147895813, "global_step": 119613, "epoch": 1343, "lr": 5.94282835621744e-05}
{"train_loss": 0.14596740901470184, "global_step": 119614, "epoch": 1343, "lr": 5.9427714225074284e-05}
{"train_loss": 0.12338121340013622, "global_step": 119615, "epoch": 1343, "lr": 5.942714488670673e-05, "val_loss": 5.589865684509277}
{"train_loss": 0.0897948294878006, "global_step": 119616, "epoch": 1344, "lr": 5.9426575547071804e-05}
{"train_loss": 0.10646171867847443, "global_step": 119617, "epoch": 1344, "lr": 5.94260062061696e-05}
{"train_loss": 0.23116594552993774, "global_step": 119618, "epoch": 1344, "lr": 5.942543686400018e-05}
{"train_loss": 0.1322391927242279, "global_step": 119619, "epoch": 1344, "lr": 5.942486752056363e-05}
{"train_loss": 0.06839492917060852, "global_step": 119620, "epoch": 1344, "lr": 5.942429817586002e-05}
{"train_loss": 0.16175617277622223, "global_step": 119621, "epoch": 1344, "lr": 5.942372882988942e-05}
{"train_loss": 0.0944732204079628, "global_step": 119622, "epoch": 1344, "lr": 5.9423159482651915e-05}
{"train_loss": 0.08424800634384155, "global_step": 119623, "epoch": 1344, "lr": 5.942259013414759e-05}
{"train_loss": 0.17426688969135284, "global_step": 119624, "epoch": 1344, "lr": 5.9422020784376507e-05}
{"train_loss": 0.17616422474384308, "global_step": 119625, "epoch": 1344, "lr": 5.942145143333875e-05}
{"train_loss": 0.11319392919540405, "global_step": 119626, "epoch": 1344, "lr": 5.942088208103439e-05}
{"train_loss": 0.15101398527622223, "global_step": 119627, "epoch": 1344, "lr": 5.942031272746351e-05}
{"train_loss": 0.13863788545131683, "global_step": 119628, "epoch": 1344, "lr": 5.941974337262618e-05}
{"train_loss": 0.1098313182592392, "global_step": 119629, "epoch": 1344, "lr": 5.9419174016522486e-05}
{"train_loss": 0.11299512535333633, "global_step": 119630, "epoch": 1344, "lr": 5.94186046591525e-05}
{"train_loss": 0.22314484417438507, "global_step": 119631, "epoch": 1344, "lr": 5.9418035300516294e-05}
{"train_loss": 0.14127600193023682, "global_step": 119632, "epoch": 1344, "lr": 5.941746594061395e-05}
{"train_loss": 0.1411789059638977, "global_step": 119633, "epoch": 1344, "lr": 5.9416896579445535e-05}
{"train_loss": 0.11586917191743851, "global_step": 119634, "epoch": 1344, "lr": 5.941632721701114e-05}
{"train_loss": 0.20050475001335144, "global_step": 119635, "epoch": 1344, "lr": 5.941575785331084e-05}
{"train_loss": 0.06738467514514923, "global_step": 119636, "epoch": 1344, "lr": 5.941518848834469e-05}
{"train_loss": 0.07317796349525452, "global_step": 119637, "epoch": 1344, "lr": 5.9414619122112793e-05}
{"train_loss": 0.19895224273204803, "global_step": 119638, "epoch": 1344, "lr": 5.941404975461522e-05}
{"train_loss": 0.1541353315114975, "global_step": 119639, "epoch": 1344, "lr": 5.941348038585204e-05}
{"train_loss": 0.1646048128604889, "global_step": 119640, "epoch": 1344, "lr": 5.9412911015823326e-05}
{"train_loss": 0.12316838651895523, "global_step": 119641, "epoch": 1344, "lr": 5.9412341644529154e-05}
{"train_loss": 0.1570856273174286, "global_step": 119642, "epoch": 1344, "lr": 5.9411772271969624e-05}
{"train_loss": 0.19612520933151245, "global_step": 119643, "epoch": 1344, "lr": 5.941120289814479e-05}
{"train_loss": 0.12717364728450775, "global_step": 119644, "epoch": 1344, "lr": 5.941063352305474e-05}
{"train_loss": 0.12933450937271118, "global_step": 119645, "epoch": 1344, "lr": 5.941006414669953e-05}
{"train_loss": 0.19183006882667542, "global_step": 119646, "epoch": 1344, "lr": 5.940949476907926e-05}
{"train_loss": 0.17019854485988617, "global_step": 119647, "epoch": 1344, "lr": 5.9408925390194e-05}
{"train_loss": 0.26585811376571655, "global_step": 119648, "epoch": 1344, "lr": 5.940835601004382e-05}
{"train_loss": 0.13096538186073303, "global_step": 119649, "epoch": 1344, "lr": 5.94077866286288e-05}
{"train_loss": 0.13292695581912994, "global_step": 119650, "epoch": 1344, "lr": 5.940721724594903e-05}
{"train_loss": 0.14314791560173035, "global_step": 119651, "epoch": 1344, "lr": 5.940664786200456e-05}
{"train_loss": 0.22189126908779144, "global_step": 119652, "epoch": 1344, "lr": 5.940607847679549e-05}
{"train_loss": 0.11319654434919357, "global_step": 119653, "epoch": 1344, "lr": 5.9405509090321885e-05}
{"train_loss": 0.15480588376522064, "global_step": 119654, "epoch": 1344, "lr": 5.940493970258383e-05}
{"train_loss": 0.1487150639295578, "global_step": 119655, "epoch": 1344, "lr": 5.940437031358138e-05}
{"train_loss": 0.08875672519207001, "global_step": 119656, "epoch": 1344, "lr": 5.940380092331464e-05}
{"train_loss": 0.19118517637252808, "global_step": 119657, "epoch": 1344, "lr": 5.940323153178367e-05}
{"train_loss": 0.18088054656982422, "global_step": 119658, "epoch": 1344, "lr": 5.9402662138988555e-05}
{"train_loss": 0.1594601571559906, "global_step": 119659, "epoch": 1344, "lr": 5.940209274492936e-05}
{"train_loss": 0.10893802344799042, "global_step": 119660, "epoch": 1344, "lr": 5.940152334960617e-05}
{"train_loss": 0.16337160766124725, "global_step": 119661, "epoch": 1344, "lr": 5.940095395301906e-05}
{"train_loss": 0.1584305763244629, "global_step": 119662, "epoch": 1344, "lr": 5.94003845551681e-05}
{"train_loss": 0.18999135494232178, "global_step": 119663, "epoch": 1344, "lr": 5.939981515605339e-05}
{"train_loss": 0.20522287487983704, "global_step": 119664, "epoch": 1344, "lr": 5.939924575567497e-05}
{"train_loss": 0.11605624854564667, "global_step": 119665, "epoch": 1344, "lr": 5.9398676354032943e-05}
{"train_loss": 0.12128832191228867, "global_step": 119666, "epoch": 1344, "lr": 5.9398106951127373e-05}
{"train_loss": 0.17250902950763702, "global_step": 119667, "epoch": 1344, "lr": 5.939753754695836e-05}
{"train_loss": 0.11851579695940018, "global_step": 119668, "epoch": 1344, "lr": 5.9396968141525944e-05}
{"train_loss": 0.11398526281118393, "global_step": 119669, "epoch": 1344, "lr": 5.939639873483023e-05}
{"train_loss": 0.1170903667807579, "global_step": 119670, "epoch": 1344, "lr": 5.9395829326871286e-05}
{"train_loss": 0.07732716202735901, "global_step": 119671, "epoch": 1344, "lr": 5.939525991764918e-05}
{"train_loss": 0.14423175156116486, "global_step": 119672, "epoch": 1344, "lr": 5.9394690507163994e-05}
{"train_loss": 0.17353758215904236, "global_step": 119673, "epoch": 1344, "lr": 5.939412109541582e-05}
{"train_loss": 0.1254376620054245, "global_step": 119674, "epoch": 1344, "lr": 5.9393551682404714e-05}
{"train_loss": 0.11074724048376083, "global_step": 119675, "epoch": 1344, "lr": 5.939298226813075e-05}
{"train_loss": 0.13524793088436127, "global_step": 119676, "epoch": 1344, "lr": 5.939241285259403e-05}
{"train_loss": 0.09131766110658646, "global_step": 119677, "epoch": 1344, "lr": 5.93918434357946e-05}
{"train_loss": 0.14112576842308044, "global_step": 119678, "epoch": 1344, "lr": 5.939127401773256e-05}
{"train_loss": 0.08251213282346725, "global_step": 119679, "epoch": 1344, "lr": 5.939070459840798e-05}
{"train_loss": 0.20267024636268616, "global_step": 119680, "epoch": 1344, "lr": 5.939013517782093e-05}
{"train_loss": 0.0887855663895607, "global_step": 119681, "epoch": 1344, "lr": 5.938956575597149e-05}
{"train_loss": 0.15022480487823486, "global_step": 119682, "epoch": 1344, "lr": 5.9388996332859736e-05}
{"train_loss": 0.13744807243347168, "global_step": 119683, "epoch": 1344, "lr": 5.9388426908485736e-05}
{"train_loss": 0.1897074431180954, "global_step": 119684, "epoch": 1344, "lr": 5.93878574828496e-05}
{"train_loss": 0.09007851034402847, "global_step": 119685, "epoch": 1344, "lr": 5.9387288055951365e-05}
{"train_loss": 0.19316989183425903, "global_step": 119686, "epoch": 1344, "lr": 5.938671862779113e-05}
{"train_loss": 0.17161191999912262, "global_step": 119687, "epoch": 1344, "lr": 5.9386149198368954e-05}
{"train_loss": 0.10314059257507324, "global_step": 119688, "epoch": 1344, "lr": 5.938557976768494e-05}
{"train_loss": 0.09589587152004242, "global_step": 119689, "epoch": 1344, "lr": 5.938501033573914e-05}
{"train_loss": 0.1146969124674797, "global_step": 119690, "epoch": 1344, "lr": 5.938444090253165e-05}
{"train_loss": 0.09656874090433121, "global_step": 119691, "epoch": 1344, "lr": 5.938387146806251e-05}
{"train_loss": 0.1850178986787796, "global_step": 119692, "epoch": 1344, "lr": 5.9383302032331854e-05}
{"train_loss": 0.1522672325372696, "global_step": 119693, "epoch": 1344, "lr": 5.938273259533972e-05}
{"train_loss": 0.13158006966114044, "global_step": 119694, "epoch": 1344, "lr": 5.938216315708618e-05}
{"train_loss": 0.10747232288122177, "global_step": 119695, "epoch": 1344, "lr": 5.938159371757133e-05}
{"train_loss": 0.17674966156482697, "global_step": 119696, "epoch": 1344, "lr": 5.938102427679524e-05}
{"train_loss": 0.1885068118572235, "global_step": 119697, "epoch": 1344, "lr": 5.938045483475798e-05}
{"train_loss": 0.1335282325744629, "global_step": 119698, "epoch": 1344, "lr": 5.937988539145965e-05}
{"train_loss": 0.12327733635902405, "global_step": 119699, "epoch": 1344, "lr": 5.9379315946900283e-05}
{"train_loss": 0.1881769746541977, "global_step": 119700, "epoch": 1344, "lr": 5.937874650108e-05}
{"train_loss": 0.11249783635139465, "global_step": 119701, "epoch": 1344, "lr": 5.937817705399885e-05}
{"train_loss": 0.11067817360162735, "global_step": 119702, "epoch": 1344, "lr": 5.9377607605656914e-05}
{"train_loss": 0.14387933909893036, "global_step": 119703, "epoch": 1344, "lr": 5.937703815605429e-05}
{"train_loss": 0.1414366788911016, "global_step": 119704, "epoch": 1344, "lr": 5.937646870519102e-05, "val_loss": 5.335773468017578}
{"train_loss": 0.11075080931186676, "global_step": 119705, "epoch": 1345, "lr": 5.93758992530672e-05}
{"train_loss": 0.11780058592557907, "global_step": 119706, "epoch": 1345, "lr": 5.937532979968291e-05}
{"train_loss": 0.2053554654121399, "global_step": 119707, "epoch": 1345, "lr": 5.9374760345038224e-05}
{"train_loss": 0.07978922873735428, "global_step": 119708, "epoch": 1345, "lr": 5.9374190889133206e-05}
{"train_loss": 0.21847158670425415, "global_step": 119709, "epoch": 1345, "lr": 5.937362143196794e-05}
{"train_loss": 0.12124200910329819, "global_step": 119710, "epoch": 1345, "lr": 5.937305197354252e-05}
{"train_loss": 0.12311586737632751, "global_step": 119711, "epoch": 1345, "lr": 5.937248251385701e-05}
{"train_loss": 0.20213682949543, "global_step": 119712, "epoch": 1345, "lr": 5.9371913052911474e-05}
{"train_loss": 0.15536849200725555, "global_step": 119713, "epoch": 1345, "lr": 5.9371343590705995e-05}
{"train_loss": 0.09402075409889221, "global_step": 119714, "epoch": 1345, "lr": 5.937077412724066e-05}
{"train_loss": 0.07105070352554321, "global_step": 119715, "epoch": 1345, "lr": 5.937020466251554e-05}
{"train_loss": 0.19406577944755554, "global_step": 119716, "epoch": 1345, "lr": 5.93696351965307e-05}
{"train_loss": 0.10109000653028488, "global_step": 119717, "epoch": 1345, "lr": 5.9369065729286245e-05}
{"train_loss": 0.11405174434185028, "global_step": 119718, "epoch": 1345, "lr": 5.9368496260782215e-05}
{"train_loss": 0.12993954122066498, "global_step": 119719, "epoch": 1345, "lr": 5.9367926791018715e-05}
{"train_loss": 0.1719253957271576, "global_step": 119720, "epoch": 1345, "lr": 5.9367357319995806e-05}
{"train_loss": 0.13233791291713715, "global_step": 119721, "epoch": 1345, "lr": 5.936678784771357e-05}
{"train_loss": 0.15027916431427002, "global_step": 119722, "epoch": 1345, "lr": 5.936621837417209e-05}
{"train_loss": 0.14734973013401031, "global_step": 119723, "epoch": 1345, "lr": 5.936564889937143e-05}
{"train_loss": 0.09875285625457764, "global_step": 119724, "epoch": 1345, "lr": 5.936507942331166e-05}
{"train_loss": 0.09929506480693817, "global_step": 119725, "epoch": 1345, "lr": 5.93645099459929e-05}
{"train_loss": 0.10969725251197815, "global_step": 119726, "epoch": 1345, "lr": 5.936394046741517e-05}
{"train_loss": 0.1139066070318222, "global_step": 119727, "epoch": 1345, "lr": 5.936337098757857e-05}
{"train_loss": 0.13463455438613892, "global_step": 119728, "epoch": 1345, "lr": 5.936280150648319e-05}
{"train_loss": 0.1375742107629776, "global_step": 119729, "epoch": 1345, "lr": 5.9362232024129095e-05}
{"train_loss": 0.09884147346019745, "global_step": 119730, "epoch": 1345, "lr": 5.936166254051638e-05}
{"train_loss": 0.08734390884637833, "global_step": 119731, "epoch": 1345, "lr": 5.936109305564508e-05}
{"train_loss": 0.27465999126434326, "global_step": 119732, "epoch": 1345, "lr": 5.9360523569515306e-05}
{"train_loss": 0.07882273942232132, "global_step": 119733, "epoch": 1345, "lr": 5.9359954082127124e-05}
{"train_loss": 0.08665665984153748, "global_step": 119734, "epoch": 1345, "lr": 5.93593845934806e-05}
{"train_loss": 0.10560587793588638, "global_step": 119735, "epoch": 1345, "lr": 5.9358815103575825e-05}
{"train_loss": 0.1136229932308197, "global_step": 119736, "epoch": 1345, "lr": 5.935824561241288e-05}
{"train_loss": 0.08204396069049835, "global_step": 119737, "epoch": 1345, "lr": 5.935767611999182e-05}
{"train_loss": 0.12929725646972656, "global_step": 119738, "epoch": 1345, "lr": 5.935710662631274e-05}
{"train_loss": 0.1237495169043541, "global_step": 119739, "epoch": 1345, "lr": 5.935653713137571e-05}
{"train_loss": 0.11301719397306442, "global_step": 119740, "epoch": 1345, "lr": 5.935596763518081e-05}
{"train_loss": 0.16042013466358185, "global_step": 119741, "epoch": 1345, "lr": 5.935539813772811e-05}
{"train_loss": 0.10009607672691345, "global_step": 119742, "epoch": 1345, "lr": 5.935482863901769e-05}
{"train_loss": 0.19553378224372864, "global_step": 119743, "epoch": 1345, "lr": 5.9354259139049636e-05}
{"train_loss": 0.14584855735301971, "global_step": 119744, "epoch": 1345, "lr": 5.935368963782402e-05}
{"train_loss": 0.11930890381336212, "global_step": 119745, "epoch": 1345, "lr": 5.935312013534089e-05}
{"train_loss": 0.11402572691440582, "global_step": 119746, "epoch": 1345, "lr": 5.935255063160036e-05}
{"train_loss": 0.13229289650917053, "global_step": 119747, "epoch": 1345, "lr": 5.9351981126602496e-05}
{"train_loss": 0.22635237872600555, "global_step": 119748, "epoch": 1345, "lr": 5.9351411620347367e-05}
{"train_loss": 0.13814246654510498, "global_step": 119749, "epoch": 1345, "lr": 5.935084211283506e-05}
{"train_loss": 0.1060226708650589, "global_step": 119750, "epoch": 1345, "lr": 5.935027260406566e-05}
{"train_loss": 0.16825172305107117, "global_step": 119751, "epoch": 1345, "lr": 5.934970309403921e-05}
{"train_loss": 0.11449895799160004, "global_step": 119752, "epoch": 1345, "lr": 5.934913358275581e-05}
{"train_loss": 0.09536591917276382, "global_step": 119753, "epoch": 1345, "lr": 5.934856407021554e-05}
{"train_loss": 0.0777951329946518, "global_step": 119754, "epoch": 1345, "lr": 5.934799455641846e-05}
{"train_loss": 0.11278719455003738, "global_step": 119755, "epoch": 1345, "lr": 5.9347425041364666e-05}
{"train_loss": 0.16267423331737518, "global_step": 119756, "epoch": 1345, "lr": 5.9346855525054226e-05}
{"train_loss": 0.16452297568321228, "global_step": 119757, "epoch": 1345, "lr": 5.9346286007487195e-05}
{"train_loss": 0.14727963507175446, "global_step": 119758, "epoch": 1345, "lr": 5.9345716488663696e-05}
{"train_loss": 0.1580362170934677, "global_step": 119759, "epoch": 1345, "lr": 5.934514696858376e-05}
{"train_loss": 0.1353973001241684, "global_step": 119760, "epoch": 1345, "lr": 5.934457744724748e-05}
{"train_loss": 0.1239614337682724, "global_step": 119761, "epoch": 1345, "lr": 5.934400792465496e-05}
{"train_loss": 0.12854517996311188, "global_step": 119762, "epoch": 1345, "lr": 5.934343840080623e-05}
{"train_loss": 0.15706489980220795, "global_step": 119763, "epoch": 1345, "lr": 5.93428688757014e-05}
{"train_loss": 0.13513104617595673, "global_step": 119764, "epoch": 1345, "lr": 5.934229934934051e-05}
{"train_loss": 0.13773109018802643, "global_step": 119765, "epoch": 1345, "lr": 5.934172982172369e-05}
{"train_loss": 0.06904658675193787, "global_step": 119766, "epoch": 1345, "lr": 5.934116029285097e-05}
{"train_loss": 0.19786842167377472, "global_step": 119767, "epoch": 1345, "lr": 5.934059076272247e-05}
{"train_loss": 0.16950690746307373, "global_step": 119768, "epoch": 1345, "lr": 5.934002123133822e-05}
{"train_loss": 0.13152949512004852, "global_step": 119769, "epoch": 1345, "lr": 5.933945169869833e-05}
{"train_loss": 0.11288753151893616, "global_step": 119770, "epoch": 1345, "lr": 5.933888216480287e-05}
{"train_loss": 0.22828355431556702, "global_step": 119771, "epoch": 1345, "lr": 5.9338312629651896e-05}
{"train_loss": 0.1391546130180359, "global_step": 119772, "epoch": 1345, "lr": 5.9337743093245504e-05}
{"train_loss": 0.11525692790746689, "global_step": 119773, "epoch": 1345, "lr": 5.933717355558377e-05}
{"train_loss": 0.06841851770877838, "global_step": 119774, "epoch": 1345, "lr": 5.9336604016666764e-05}
{"train_loss": 0.09452763199806213, "global_step": 119775, "epoch": 1345, "lr": 5.9336034476494565e-05}
{"train_loss": 0.1252422034740448, "global_step": 119776, "epoch": 1345, "lr": 5.933546493506724e-05}
{"train_loss": 0.14216813445091248, "global_step": 119777, "epoch": 1345, "lr": 5.9334895392384895e-05}
{"train_loss": 0.1827743500471115, "global_step": 119778, "epoch": 1345, "lr": 5.933432584844757e-05}
{"train_loss": 0.10886827856302261, "global_step": 119779, "epoch": 1345, "lr": 5.933375630325537e-05}
{"train_loss": 0.08423297107219696, "global_step": 119780, "epoch": 1345, "lr": 5.933318675680836e-05}
{"train_loss": 0.08021683245897293, "global_step": 119781, "epoch": 1345, "lr": 5.933261720910661e-05}
{"train_loss": 0.10469874739646912, "global_step": 119782, "epoch": 1345, "lr": 5.9332047660150193e-05}
{"train_loss": 0.1938827782869339, "global_step": 119783, "epoch": 1345, "lr": 5.933147810993921e-05}
{"train_loss": 0.13835224509239197, "global_step": 119784, "epoch": 1345, "lr": 5.933090855847372e-05}
{"train_loss": 0.12440682202577591, "global_step": 119785, "epoch": 1345, "lr": 5.933033900575381e-05}
{"train_loss": 0.13924510776996613, "global_step": 119786, "epoch": 1345, "lr": 5.9329769451779535e-05}
{"train_loss": 0.13261519372463226, "global_step": 119787, "epoch": 1345, "lr": 5.9329199896551e-05}
{"train_loss": 0.16614916920661926, "global_step": 119788, "epoch": 1345, "lr": 5.932863034006828e-05}
{"train_loss": 0.18406057357788086, "global_step": 119789, "epoch": 1345, "lr": 5.932806078233142e-05}
{"train_loss": 0.15490764379501343, "global_step": 119790, "epoch": 1345, "lr": 5.9327491223340524e-05}
{"train_loss": 0.15621958673000336, "global_step": 119791, "epoch": 1345, "lr": 5.932692166309566e-05}
{"train_loss": 0.10083958506584167, "global_step": 119792, "epoch": 1345, "lr": 5.9326352101596896e-05}
{"train_loss": 0.1329339619600371, "global_step": 119793, "epoch": 1345, "lr": 5.9325782538844324e-05, "val_loss": 5.270358085632324, "train_action_mse_error": 19.77326774597168}
{"train_loss": 0.1378212869167328, "global_step": 119794, "epoch": 1346, "lr": 5.932521297483802e-05}
{"train_loss": 0.16665993630886078, "global_step": 119795, "epoch": 1346, "lr": 5.932464340957804e-05}
{"train_loss": 0.18046621978282928, "global_step": 119796, "epoch": 1346, "lr": 5.9324073843064486e-05}
{"train_loss": 0.20931756496429443, "global_step": 119797, "epoch": 1346, "lr": 5.932350427529742e-05}
{"train_loss": 0.1259099841117859, "global_step": 119798, "epoch": 1346, "lr": 5.932293470627692e-05}
{"train_loss": 0.11689022928476334, "global_step": 119799, "epoch": 1346, "lr": 5.932236513600306e-05}
{"train_loss": 0.07591160386800766, "global_step": 119800, "epoch": 1346, "lr": 5.932179556447592e-05}
{"train_loss": 0.12116476893424988, "global_step": 119801, "epoch": 1346, "lr": 5.9321225991695596e-05}
{"train_loss": 0.14400742948055267, "global_step": 119802, "epoch": 1346, "lr": 5.932065641766214e-05}
{"train_loss": 0.1370546519756317, "global_step": 119803, "epoch": 1346, "lr": 5.932008684237563e-05}
{"train_loss": 0.08773574233055115, "global_step": 119804, "epoch": 1346, "lr": 5.9319517265836154e-05}
{"train_loss": 0.17948295176029205, "global_step": 119805, "epoch": 1346, "lr": 5.931894768804378e-05}
{"train_loss": 0.15757057070732117, "global_step": 119806, "epoch": 1346, "lr": 5.9318378108998596e-05}
{"train_loss": 0.08982866257429123, "global_step": 119807, "epoch": 1346, "lr": 5.931780852870066e-05}
{"train_loss": 0.11952969431877136, "global_step": 119808, "epoch": 1346, "lr": 5.9317238947150066e-05}
{"train_loss": 0.11081000417470932, "global_step": 119809, "epoch": 1346, "lr": 5.9316669364346875e-05}
{"train_loss": 0.14026349782943726, "global_step": 119810, "epoch": 1346, "lr": 5.931609978029118e-05}
{"train_loss": 0.1162119060754776, "global_step": 119811, "epoch": 1346, "lr": 5.931553019498304e-05}
{"train_loss": 0.14179733395576477, "global_step": 119812, "epoch": 1346, "lr": 5.931496060842254e-05}
{"train_loss": 0.1394401639699936, "global_step": 119813, "epoch": 1346, "lr": 5.931439102060976e-05}
{"train_loss": 0.1350741684436798, "global_step": 119814, "epoch": 1346, "lr": 5.931382143154478e-05}
{"train_loss": 0.13771787285804749, "global_step": 119815, "epoch": 1346, "lr": 5.9313251841227654e-05}
{"train_loss": 0.12965479493141174, "global_step": 119816, "epoch": 1346, "lr": 5.931268224965849e-05}
{"train_loss": 0.12689679861068726, "global_step": 119817, "epoch": 1346, "lr": 5.931211265683734e-05}
{"train_loss": 0.14764688909053802, "global_step": 119818, "epoch": 1346, "lr": 5.931154306276429e-05}
{"train_loss": 0.09379091113805771, "global_step": 119819, "epoch": 1346, "lr": 5.931097346743941e-05}
{"train_loss": 0.1681058704853058, "global_step": 119820, "epoch": 1346, "lr": 5.93104038708628e-05}
{"train_loss": 0.1323809176683426, "global_step": 119821, "epoch": 1346, "lr": 5.930983427303452e-05}
{"train_loss": 0.22698871791362762, "global_step": 119822, "epoch": 1346, "lr": 5.930926467395463e-05}
{"train_loss": 0.1269230991601944, "global_step": 119823, "epoch": 1346, "lr": 5.930869507362323e-05}
{"train_loss": 0.13007664680480957, "global_step": 119824, "epoch": 1346, "lr": 5.93081254720404e-05}
{"train_loss": 0.13742318749427795, "global_step": 119825, "epoch": 1346, "lr": 5.930755586920619e-05}
{"train_loss": 0.11592152714729309, "global_step": 119826, "epoch": 1346, "lr": 5.9306986265120704e-05}
{"train_loss": 0.1504134237766266, "global_step": 119827, "epoch": 1346, "lr": 5.930641665978401e-05}
{"train_loss": 0.09893075376749039, "global_step": 119828, "epoch": 1346, "lr": 5.9305847053196175e-05}
{"train_loss": 0.10503105819225311, "global_step": 119829, "epoch": 1346, "lr": 5.9305277445357286e-05}
{"train_loss": 0.10551044344902039, "global_step": 119830, "epoch": 1346, "lr": 5.93047078362674e-05}
{"train_loss": 0.07141038775444031, "global_step": 119831, "epoch": 1346, "lr": 5.930413822592663e-05}
{"train_loss": 0.1500505805015564, "global_step": 119832, "epoch": 1346, "lr": 5.9303568614335024e-05}
{"train_loss": 0.1448313593864441, "global_step": 119833, "epoch": 1346, "lr": 5.9302999001492664e-05}
{"train_loss": 0.1247585192322731, "global_step": 119834, "epoch": 1346, "lr": 5.930242938739963e-05}
{"train_loss": 0.10678726434707642, "global_step": 119835, "epoch": 1346, "lr": 5.9301859772056e-05}
{"train_loss": 0.09094826877117157, "global_step": 119836, "epoch": 1346, "lr": 5.930129015546184e-05}
{"train_loss": 0.11824823915958405, "global_step": 119837, "epoch": 1346, "lr": 5.930072053761724e-05}
{"train_loss": 0.10093216598033905, "global_step": 119838, "epoch": 1346, "lr": 5.930015091852228e-05}
{"train_loss": 0.16060473024845123, "global_step": 119839, "epoch": 1346, "lr": 5.929958129817701e-05}
{"train_loss": 0.1427556574344635, "global_step": 119840, "epoch": 1346, "lr": 5.929901167658154e-05}
{"train_loss": 0.1440112143754959, "global_step": 119841, "epoch": 1346, "lr": 5.929844205373592e-05}
{"train_loss": 0.13494403660297394, "global_step": 119842, "epoch": 1346, "lr": 5.9297872429640254e-05}
{"train_loss": 0.14052794873714447, "global_step": 119843, "epoch": 1346, "lr": 5.9297302804294597e-05}
{"train_loss": 0.12114416807889938, "global_step": 119844, "epoch": 1346, "lr": 5.929673317769903e-05}
{"train_loss": 0.10148295015096664, "global_step": 119845, "epoch": 1346, "lr": 5.929616354985362e-05}
{"train_loss": 0.11226529628038406, "global_step": 119846, "epoch": 1346, "lr": 5.9295593920758474e-05}
{"train_loss": 0.11562065035104752, "global_step": 119847, "epoch": 1346, "lr": 5.9295024290413635e-05}
{"train_loss": 0.13828592002391815, "global_step": 119848, "epoch": 1346, "lr": 5.92944546588192e-05}
{"train_loss": 0.22055459022521973, "global_step": 119849, "epoch": 1346, "lr": 5.929388502597524e-05}
{"train_loss": 0.0898648276925087, "global_step": 119850, "epoch": 1346, "lr": 5.9293315391881824e-05}
{"train_loss": 0.12505106627941132, "global_step": 119851, "epoch": 1346, "lr": 5.929274575653902e-05}
{"train_loss": 0.21397395431995392, "global_step": 119852, "epoch": 1346, "lr": 5.929217611994694e-05}
{"train_loss": 0.12195592373609543, "global_step": 119853, "epoch": 1346, "lr": 5.9291606482105644e-05}
{"train_loss": 0.17690306901931763, "global_step": 119854, "epoch": 1346, "lr": 5.929103684301518e-05}
{"train_loss": 0.11660096049308777, "global_step": 119855, "epoch": 1346, "lr": 5.929046720267567e-05}
{"train_loss": 0.10850349068641663, "global_step": 119856, "epoch": 1346, "lr": 5.928989756108716e-05}
{"train_loss": 0.17156285047531128, "global_step": 119857, "epoch": 1346, "lr": 5.928932791824975e-05}
{"train_loss": 0.1533936709165573, "global_step": 119858, "epoch": 1346, "lr": 5.928875827416349e-05}
{"train_loss": 0.09434259682893753, "global_step": 119859, "epoch": 1346, "lr": 5.928818862882848e-05}
{"train_loss": 0.136631578207016, "global_step": 119860, "epoch": 1346, "lr": 5.9287618982244777e-05}
{"train_loss": 0.1608719378709793, "global_step": 119861, "epoch": 1346, "lr": 5.928704933441247e-05}
{"train_loss": 0.11113172769546509, "global_step": 119862, "epoch": 1346, "lr": 5.928647968533163e-05}
{"train_loss": 0.14263373613357544, "global_step": 119863, "epoch": 1346, "lr": 5.9285910035002344e-05}
{"train_loss": 0.11252711713314056, "global_step": 119864, "epoch": 1346, "lr": 5.928534038342467e-05}
{"train_loss": 0.12489545345306396, "global_step": 119865, "epoch": 1346, "lr": 5.9284770730598704e-05}
{"train_loss": 0.1370915025472641, "global_step": 119866, "epoch": 1346, "lr": 5.928420107652453e-05}
{"train_loss": 0.07554053515195847, "global_step": 119867, "epoch": 1346, "lr": 5.928363142120218e-05}
{"train_loss": 0.17897194623947144, "global_step": 119868, "epoch": 1346, "lr": 5.928306176463176e-05}
{"train_loss": 0.17569956183433533, "global_step": 119869, "epoch": 1346, "lr": 5.928249210681336e-05}
{"train_loss": 0.12237545847892761, "global_step": 119870, "epoch": 1346, "lr": 5.928192244774704e-05}
{"train_loss": 0.08052188903093338, "global_step": 119871, "epoch": 1346, "lr": 5.928135278743286e-05}
{"train_loss": 0.10533963143825531, "global_step": 119872, "epoch": 1346, "lr": 5.928078312587093e-05}
{"train_loss": 0.1334644854068756, "global_step": 119873, "epoch": 1346, "lr": 5.928021346306131e-05}
{"train_loss": 0.16177353262901306, "global_step": 119874, "epoch": 1346, "lr": 5.9279643799004084e-05}
{"train_loss": 0.18335871398448944, "global_step": 119875, "epoch": 1346, "lr": 5.927907413369931e-05}
{"train_loss": 0.17077676951885223, "global_step": 119876, "epoch": 1346, "lr": 5.927850446714709e-05}
{"train_loss": 0.18809549510478973, "global_step": 119877, "epoch": 1346, "lr": 5.9277934799347476e-05}
{"train_loss": 0.13575468957424164, "global_step": 119878, "epoch": 1346, "lr": 5.9277365130300576e-05}
{"train_loss": 0.05350593477487564, "global_step": 119879, "epoch": 1346, "lr": 5.927679546000643e-05}
{"train_loss": 0.10011885315179825, "global_step": 119880, "epoch": 1346, "lr": 5.927622578846515e-05}
{"train_loss": 0.14162683486938477, "global_step": 119881, "epoch": 1346, "lr": 5.927565611567677e-05}
{"train_loss": 0.13261218117863943, "global_step": 119882, "epoch": 1346, "lr": 5.9275086441641406e-05, "val_loss": 5.336644649505615}
{"train_loss": 0.1841856688261032, "global_step": 119883, "epoch": 1347, "lr": 5.9274516766359124e-05}
{"train_loss": 0.1507209986448288, "global_step": 119884, "epoch": 1347, "lr": 5.927394708982998e-05}
{"train_loss": 0.12915368378162384, "global_step": 119885, "epoch": 1347, "lr": 5.927337741205409e-05}
{"train_loss": 0.20356246829032898, "global_step": 119886, "epoch": 1347, "lr": 5.92728077330315e-05}
{"train_loss": 0.1704765111207962, "global_step": 119887, "epoch": 1347, "lr": 5.927223805276229e-05}
{"train_loss": 0.13381311297416687, "global_step": 119888, "epoch": 1347, "lr": 5.9271668371246534e-05}
{"train_loss": 0.13526467978954315, "global_step": 119889, "epoch": 1347, "lr": 5.9271098688484314e-05}
{"train_loss": 0.07597754895687103, "global_step": 119890, "epoch": 1347, "lr": 5.9270529004475716e-05}
{"train_loss": 0.09627272933721542, "global_step": 119891, "epoch": 1347, "lr": 5.926995931922079e-05}
{"train_loss": 0.1250317096710205, "global_step": 119892, "epoch": 1347, "lr": 5.9269389632719655e-05}
{"train_loss": 0.11136644333600998, "global_step": 119893, "epoch": 1347, "lr": 5.9268819944972354e-05}
{"train_loss": 0.13319621980190277, "global_step": 119894, "epoch": 1347, "lr": 5.9268250255978974e-05}
{"train_loss": 0.12902696430683136, "global_step": 119895, "epoch": 1347, "lr": 5.9267680565739594e-05}
{"train_loss": 0.12108970433473587, "global_step": 119896, "epoch": 1347, "lr": 5.9267110874254283e-05}
{"train_loss": 0.16316701471805573, "global_step": 119897, "epoch": 1347, "lr": 5.926654118152313e-05}
{"train_loss": 0.11194772273302078, "global_step": 119898, "epoch": 1347, "lr": 5.926597148754619e-05}
{"train_loss": 0.0981682687997818, "global_step": 119899, "epoch": 1347, "lr": 5.926540179232356e-05}
{"train_loss": 0.1668056696653366, "global_step": 119900, "epoch": 1347, "lr": 5.926483209585531e-05}
{"train_loss": 0.10937362164258957, "global_step": 119901, "epoch": 1347, "lr": 5.926426239814151e-05}
{"train_loss": 0.11674314737319946, "global_step": 119902, "epoch": 1347, "lr": 5.926369269918226e-05}
{"train_loss": 0.15890946984291077, "global_step": 119903, "epoch": 1347, "lr": 5.9263122998977606e-05}
{"train_loss": 0.12045726925134659, "global_step": 119904, "epoch": 1347, "lr": 5.926255329752764e-05}
{"train_loss": 0.11776147782802582, "global_step": 119905, "epoch": 1347, "lr": 5.926198359483245e-05}
{"train_loss": 0.14510633051395416, "global_step": 119906, "epoch": 1347, "lr": 5.926141389089208e-05}
{"train_loss": 0.14378567039966583, "global_step": 119907, "epoch": 1347, "lr": 5.9260844185706635e-05}
{"train_loss": 0.16835936903953552, "global_step": 119908, "epoch": 1347, "lr": 5.926027447927618e-05}
{"train_loss": 0.1252494752407074, "global_step": 119909, "epoch": 1347, "lr": 5.9259704771600776e-05}
{"train_loss": 0.09817295521497726, "global_step": 119910, "epoch": 1347, "lr": 5.925913506268054e-05}
{"train_loss": 0.154695063829422, "global_step": 119911, "epoch": 1347, "lr": 5.9258565352515515e-05}
{"train_loss": 0.15206025540828705, "global_step": 119912, "epoch": 1347, "lr": 5.92579956411058e-05}
{"train_loss": 0.11893496662378311, "global_step": 119913, "epoch": 1347, "lr": 5.925742592845145e-05}
{"train_loss": 0.1320459097623825, "global_step": 119914, "epoch": 1347, "lr": 5.9256856214552556e-05}
{"train_loss": 0.27951058745384216, "global_step": 119915, "epoch": 1347, "lr": 5.9256286499409195e-05}
{"train_loss": 0.14087000489234924, "global_step": 119916, "epoch": 1347, "lr": 5.925571678302144e-05}
{"train_loss": 0.08421701192855835, "global_step": 119917, "epoch": 1347, "lr": 5.925514706538936e-05}
{"train_loss": 0.20047682523727417, "global_step": 119918, "epoch": 1347, "lr": 5.925457734651304e-05}
{"train_loss": 0.13500477373600006, "global_step": 119919, "epoch": 1347, "lr": 5.925400762639255e-05}
{"train_loss": 0.1290573626756668, "global_step": 119920, "epoch": 1347, "lr": 5.925343790502798e-05}
{"train_loss": 0.1080462858080864, "global_step": 119921, "epoch": 1347, "lr": 5.9252868182419396e-05}
{"train_loss": 0.156447634100914, "global_step": 119922, "epoch": 1347, "lr": 5.9252298458566876e-05}
{"train_loss": 0.09853190928697586, "global_step": 119923, "epoch": 1347, "lr": 5.925172873347049e-05}
{"train_loss": 0.1495269536972046, "global_step": 119924, "epoch": 1347, "lr": 5.9251159007130344e-05}
{"train_loss": 0.1857251673936844, "global_step": 119925, "epoch": 1347, "lr": 5.9250589279546475e-05}
{"train_loss": 0.13949361443519592, "global_step": 119926, "epoch": 1347, "lr": 5.9250019550718985e-05}
{"train_loss": 0.09009827673435211, "global_step": 119927, "epoch": 1347, "lr": 5.924944982064793e-05}
{"train_loss": 0.12141264975070953, "global_step": 119928, "epoch": 1347, "lr": 5.92488800893334e-05}
{"train_loss": 0.10947389900684357, "global_step": 119929, "epoch": 1347, "lr": 5.924831035677547e-05}
{"train_loss": 0.14597001671791077, "global_step": 119930, "epoch": 1347, "lr": 5.924774062297422e-05}
{"train_loss": 0.19770020246505737, "global_step": 119931, "epoch": 1347, "lr": 5.924717088792974e-05}
{"train_loss": 0.10673226416110992, "global_step": 119932, "epoch": 1347, "lr": 5.924660115164208e-05}
{"train_loss": 0.09157127887010574, "global_step": 119933, "epoch": 1347, "lr": 5.9246031414111324e-05}
{"train_loss": 0.12484141439199448, "global_step": 119934, "epoch": 1347, "lr": 5.924546167533755e-05}
{"train_loss": 0.20580905675888062, "global_step": 119935, "epoch": 1347, "lr": 5.9244891935320846e-05}
{"train_loss": 0.15368588268756866, "global_step": 119936, "epoch": 1347, "lr": 5.9244322194061264e-05}
{"train_loss": 0.16168628633022308, "global_step": 119937, "epoch": 1347, "lr": 5.924375245155891e-05}
{"train_loss": 0.07046017795801163, "global_step": 119938, "epoch": 1347, "lr": 5.924318270781384e-05}
{"train_loss": 0.11149770021438599, "global_step": 119939, "epoch": 1347, "lr": 5.924261296282614e-05}
{"train_loss": 0.08644126355648041, "global_step": 119940, "epoch": 1347, "lr": 5.9242043216595876e-05}
{"train_loss": 0.09396176785230637, "global_step": 119941, "epoch": 1347, "lr": 5.9241473469123145e-05}
{"train_loss": 0.13570326566696167, "global_step": 119942, "epoch": 1347, "lr": 5.924090372040799e-05}
{"train_loss": 0.220753476023674, "global_step": 119943, "epoch": 1347, "lr": 5.924033397045053e-05}
{"train_loss": 0.08657483011484146, "global_step": 119944, "epoch": 1347, "lr": 5.9239764219250816e-05}
{"train_loss": 0.09954521059989929, "global_step": 119945, "epoch": 1347, "lr": 5.9239194466808914e-05}
{"train_loss": 0.1875431388616562, "global_step": 119946, "epoch": 1347, "lr": 5.9238624713124924e-05}
{"train_loss": 0.15977834165096283, "global_step": 119947, "epoch": 1347, "lr": 5.9238054958198905e-05}
{"train_loss": 0.11924079060554504, "global_step": 119948, "epoch": 1347, "lr": 5.923748520203095e-05}
{"train_loss": 0.17643091082572937, "global_step": 119949, "epoch": 1347, "lr": 5.9236915444621136e-05}
{"train_loss": 0.11328057199716568, "global_step": 119950, "epoch": 1347, "lr": 5.923634568596952e-05}
{"train_loss": 0.12409712374210358, "global_step": 119951, "epoch": 1347, "lr": 5.923577592607619e-05}
{"train_loss": 0.18325157463550568, "global_step": 119952, "epoch": 1347, "lr": 5.923520616494122e-05}
{"train_loss": 0.08243096619844437, "global_step": 119953, "epoch": 1347, "lr": 5.92346364025647e-05}
{"train_loss": 0.1069662794470787, "global_step": 119954, "epoch": 1347, "lr": 5.9234066638946696e-05}
{"train_loss": 0.11511704325675964, "global_step": 119955, "epoch": 1347, "lr": 5.923349687408727e-05}
{"train_loss": 0.09422025829553604, "global_step": 119956, "epoch": 1347, "lr": 5.923292710798652e-05}
{"train_loss": 0.12559325993061066, "global_step": 119957, "epoch": 1347, "lr": 5.923235734064452e-05}
{"train_loss": 0.17551155388355255, "global_step": 119958, "epoch": 1347, "lr": 5.9231787572061335e-05}
{"train_loss": 0.1364704966545105, "global_step": 119959, "epoch": 1347, "lr": 5.923121780223706e-05}
{"train_loss": 0.08023154735565186, "global_step": 119960, "epoch": 1347, "lr": 5.9230648031171754e-05}
{"train_loss": 0.13323082029819489, "global_step": 119961, "epoch": 1347, "lr": 5.9230078258865494e-05}
{"train_loss": 0.15926530957221985, "global_step": 119962, "epoch": 1347, "lr": 5.922950848531838e-05}
{"train_loss": 0.14277581870555878, "global_step": 119963, "epoch": 1347, "lr": 5.9228938710530446e-05}
{"train_loss": 0.16153335571289062, "global_step": 119964, "epoch": 1347, "lr": 5.922836893450181e-05}
{"train_loss": 0.10109038650989532, "global_step": 119965, "epoch": 1347, "lr": 5.922779915723252e-05}
{"train_loss": 0.13666319847106934, "global_step": 119966, "epoch": 1347, "lr": 5.922722937872267e-05}
{"train_loss": 0.1789974570274353, "global_step": 119967, "epoch": 1347, "lr": 5.9226659598972336e-05}
{"train_loss": 0.13682985305786133, "global_step": 119968, "epoch": 1347, "lr": 5.922608981798159e-05}
{"train_loss": 0.08633694797754288, "global_step": 119969, "epoch": 1347, "lr": 5.92255200357505e-05}
{"train_loss": 0.13072219491004944, "global_step": 119970, "epoch": 1347, "lr": 5.922495025227917e-05}
{"train_loss": 0.13462757578726564, "global_step": 119971, "epoch": 1347, "lr": 5.922438046756764e-05, "val_loss": 5.568353176116943}
{"train_loss": 0.1171744093298912, "global_step": 119972, "epoch": 1348, "lr": 5.922381068161602e-05}
{"train_loss": 0.16341179609298706, "global_step": 119973, "epoch": 1348, "lr": 5.922324089442436e-05}
{"train_loss": 0.15873752534389496, "global_step": 119974, "epoch": 1348, "lr": 5.922267110599274e-05}
{"train_loss": 0.12283971905708313, "global_step": 119975, "epoch": 1348, "lr": 5.922210131632127e-05}
{"train_loss": 0.16529254615306854, "global_step": 119976, "epoch": 1348, "lr": 5.922153152540998e-05}
{"train_loss": 0.09860765933990479, "global_step": 119977, "epoch": 1348, "lr": 5.9220961733258975e-05}
{"train_loss": 0.19767418503761292, "global_step": 119978, "epoch": 1348, "lr": 5.9220391939868316e-05}
{"train_loss": 0.18792204558849335, "global_step": 119979, "epoch": 1348, "lr": 5.921982214523809e-05}
{"train_loss": 0.08996724337339401, "global_step": 119980, "epoch": 1348, "lr": 5.921925234936837e-05}
{"train_loss": 0.1414257287979126, "global_step": 119981, "epoch": 1348, "lr": 5.921868255225924e-05}
{"train_loss": 0.21180453896522522, "global_step": 119982, "epoch": 1348, "lr": 5.921811275391077e-05}
{"train_loss": 0.11122008413076401, "global_step": 119983, "epoch": 1348, "lr": 5.9217542954323035e-05}
{"train_loss": 0.16720478236675262, "global_step": 119984, "epoch": 1348, "lr": 5.921697315349611e-05}
{"train_loss": 0.13594049215316772, "global_step": 119985, "epoch": 1348, "lr": 5.9216403351430086e-05}
{"train_loss": 0.12096734344959259, "global_step": 119986, "epoch": 1348, "lr": 5.921583354812502e-05}
{"train_loss": 0.21560843288898468, "global_step": 119987, "epoch": 1348, "lr": 5.9215263743581006e-05}
{"train_loss": 0.1945071965456009, "global_step": 119988, "epoch": 1348, "lr": 5.9214693937798106e-05}
{"train_loss": 0.12258054316043854, "global_step": 119989, "epoch": 1348, "lr": 5.9214124130776403e-05}
{"train_loss": 0.13682661950588226, "global_step": 119990, "epoch": 1348, "lr": 5.921355432251597e-05}
{"train_loss": 0.10461878776550293, "global_step": 119991, "epoch": 1348, "lr": 5.921298451301689e-05}
{"train_loss": 0.10078472644090652, "global_step": 119992, "epoch": 1348, "lr": 5.921241470227924e-05}
{"train_loss": 0.18521901965141296, "global_step": 119993, "epoch": 1348, "lr": 5.921184489030309e-05}
{"train_loss": 0.09104703366756439, "global_step": 119994, "epoch": 1348, "lr": 5.9211275077088514e-05}
{"train_loss": 0.12456835061311722, "global_step": 119995, "epoch": 1348, "lr": 5.92107052626356e-05}
{"train_loss": 0.10615000873804092, "global_step": 119996, "epoch": 1348, "lr": 5.921013544694443e-05}
{"train_loss": 0.11881610751152039, "global_step": 119997, "epoch": 1348, "lr": 5.920956563001505e-05}
{"train_loss": 0.11862466484308243, "global_step": 119998, "epoch": 1348, "lr": 5.9208995811847566e-05}
{"train_loss": 0.1776357889175415, "global_step": 119999, "epoch": 1348, "lr": 5.920842599244204e-05}
{"train_loss": 0.14312800765037537, "global_step": 120000, "epoch": 1348, "lr": 5.920785617179856e-05}
{"train_loss": 0.183624267578125, "global_step": 120001, "epoch": 1348, "lr": 5.920728634991719e-05}
{"train_loss": 0.15829713642597198, "global_step": 120002, "epoch": 1348, "lr": 5.9206716526798014e-05}
{"train_loss": 0.1283741146326065, "global_step": 120003, "epoch": 1348, "lr": 5.920614670244111e-05}
{"train_loss": 0.1551743745803833, "global_step": 120004, "epoch": 1348, "lr": 5.9205576876846544e-05}
{"train_loss": 0.17614629864692688, "global_step": 120005, "epoch": 1348, "lr": 5.9205007050014415e-05}
{"train_loss": 0.13620637357234955, "global_step": 120006, "epoch": 1348, "lr": 5.920443722194478e-05}
{"train_loss": 0.09356822818517685, "global_step": 120007, "epoch": 1348, "lr": 5.920386739263771e-05}
{"train_loss": 0.1495688408613205, "global_step": 120008, "epoch": 1348, "lr": 5.9203297562093294e-05}
{"train_loss": 0.11840344220399857, "global_step": 120009, "epoch": 1348, "lr": 5.920272773031161e-05}
{"train_loss": 0.13164468109607697, "global_step": 120010, "epoch": 1348, "lr": 5.920215789729274e-05}
{"train_loss": 0.13297373056411743, "global_step": 120011, "epoch": 1348, "lr": 5.920158806303675e-05}
{"train_loss": 0.1864701807498932, "global_step": 120012, "epoch": 1348, "lr": 5.920101822754371e-05}
{"train_loss": 0.15800216794013977, "global_step": 120013, "epoch": 1348, "lr": 5.920044839081371e-05}
{"train_loss": 0.13808849453926086, "global_step": 120014, "epoch": 1348, "lr": 5.919987855284682e-05}
{"train_loss": 0.09912119060754776, "global_step": 120015, "epoch": 1348, "lr": 5.919930871364312e-05}
{"train_loss": 0.12473544478416443, "global_step": 120016, "epoch": 1348, "lr": 5.9198738873202684e-05}
{"train_loss": 0.17009802162647247, "global_step": 120017, "epoch": 1348, "lr": 5.91981690315256e-05}
{"train_loss": 0.1371103674173355, "global_step": 120018, "epoch": 1348, "lr": 5.919759918861192e-05}
{"train_loss": 0.10462763160467148, "global_step": 120019, "epoch": 1348, "lr": 5.919702934446174e-05}
{"train_loss": 0.16735075414180756, "global_step": 120020, "epoch": 1348, "lr": 5.919645949907513e-05}
{"train_loss": 0.12169279158115387, "global_step": 120021, "epoch": 1348, "lr": 5.9195889652452176e-05}
{"train_loss": 0.20967258512973785, "global_step": 120022, "epoch": 1348, "lr": 5.919531980459293e-05}
{"train_loss": 0.12317650765180588, "global_step": 120023, "epoch": 1348, "lr": 5.91947499554975e-05}
{"train_loss": 0.12189162522554398, "global_step": 120024, "epoch": 1348, "lr": 5.919418010516594e-05}
{"train_loss": 0.13372354209423065, "global_step": 120025, "epoch": 1348, "lr": 5.919361025359834e-05}
{"train_loss": 0.20888833701610565, "global_step": 120026, "epoch": 1348, "lr": 5.9193040400794764e-05}
{"train_loss": 0.09836145490407944, "global_step": 120027, "epoch": 1348, "lr": 5.9192470546755306e-05}
{"train_loss": 0.1723165512084961, "global_step": 120028, "epoch": 1348, "lr": 5.919190069148003e-05}
{"train_loss": 0.16590113937854767, "global_step": 120029, "epoch": 1348, "lr": 5.919133083496901e-05}
{"train_loss": 0.11523698270320892, "global_step": 120030, "epoch": 1348, "lr": 5.9190760977222324e-05}
{"train_loss": 0.13978923857212067, "global_step": 120031, "epoch": 1348, "lr": 5.9190191118240056e-05}
{"train_loss": 0.13290181756019592, "global_step": 120032, "epoch": 1348, "lr": 5.9189621258022286e-05}
{"train_loss": 0.19409745931625366, "global_step": 120033, "epoch": 1348, "lr": 5.9189051396569074e-05}
{"train_loss": 0.14748769998550415, "global_step": 120034, "epoch": 1348, "lr": 5.9188481533880515e-05}
{"train_loss": 0.18217313289642334, "global_step": 120035, "epoch": 1348, "lr": 5.918791166995668e-05}
{"train_loss": 0.1299813985824585, "global_step": 120036, "epoch": 1348, "lr": 5.918734180479762e-05}
{"train_loss": 0.13191576302051544, "global_step": 120037, "epoch": 1348, "lr": 5.918677193840345e-05}
{"train_loss": 0.1406993269920349, "global_step": 120038, "epoch": 1348, "lr": 5.9186202070774236e-05}
{"train_loss": 0.18546412885189056, "global_step": 120039, "epoch": 1348, "lr": 5.9185632201910034e-05}
{"train_loss": 0.16737930476665497, "global_step": 120040, "epoch": 1348, "lr": 5.918506233181095e-05}
{"train_loss": 0.19073915481567383, "global_step": 120041, "epoch": 1348, "lr": 5.9184492460477034e-05}
{"train_loss": 0.1758076250553131, "global_step": 120042, "epoch": 1348, "lr": 5.918392258790838e-05}
{"train_loss": 0.13194459676742554, "global_step": 120043, "epoch": 1348, "lr": 5.918335271410507e-05}
{"train_loss": 0.12645196914672852, "global_step": 120044, "epoch": 1348, "lr": 5.9182782839067154e-05}
{"train_loss": 0.1324935108423233, "global_step": 120045, "epoch": 1348, "lr": 5.9182212962794735e-05}
{"train_loss": 0.13296255469322205, "global_step": 120046, "epoch": 1348, "lr": 5.9181643085287877e-05}
{"train_loss": 0.13261952996253967, "global_step": 120047, "epoch": 1348, "lr": 5.9181073206546654e-05}
{"train_loss": 0.16216666996479034, "global_step": 120048, "epoch": 1348, "lr": 5.918050332657115e-05}
{"train_loss": 0.10552803426980972, "global_step": 120049, "epoch": 1348, "lr": 5.9179933445361436e-05}
{"train_loss": 0.1174195185303688, "global_step": 120050, "epoch": 1348, "lr": 5.91793635629176e-05}
{"train_loss": 0.10785044729709625, "global_step": 120051, "epoch": 1348, "lr": 5.917879367923971e-05}
{"train_loss": 0.16042792797088623, "global_step": 120052, "epoch": 1348, "lr": 5.917822379432784e-05}
{"train_loss": 0.15613442659378052, "global_step": 120053, "epoch": 1348, "lr": 5.9177653908182064e-05}
{"train_loss": 0.1493224948644638, "global_step": 120054, "epoch": 1348, "lr": 5.917708402080247e-05}
{"train_loss": 0.17755252122879028, "global_step": 120055, "epoch": 1348, "lr": 5.917651413218913e-05}
{"train_loss": 0.14372093975543976, "global_step": 120056, "epoch": 1348, "lr": 5.917594424234212e-05}
{"train_loss": 0.20368877053260803, "global_step": 120057, "epoch": 1348, "lr": 5.917537435126151e-05}
{"train_loss": 0.19882002472877502, "global_step": 120058, "epoch": 1348, "lr": 5.9174804458947386e-05}
{"train_loss": 0.07347077876329422, "global_step": 120059, "epoch": 1348, "lr": 5.9174234565399825e-05}
{"train_loss": 0.14437957258706682, "global_step": 120060, "epoch": 1348, "lr": 5.917366467061889e-05, "val_loss": 5.153839111328125}
{"train_loss": 0.19997084140777588, "global_step": 120061, "epoch": 1349, "lr": 5.917309477460468e-05}
{"train_loss": 0.1285836100578308, "global_step": 120062, "epoch": 1349, "lr": 5.917252487735725e-05}
{"train_loss": 0.1517535299062729, "global_step": 120063, "epoch": 1349, "lr": 5.91719549788767e-05}
{"train_loss": 0.09461386501789093, "global_step": 120064, "epoch": 1349, "lr": 5.9171385079163075e-05}
{"train_loss": 0.12809790670871735, "global_step": 120065, "epoch": 1349, "lr": 5.917081517821648e-05}
{"train_loss": 0.17614689469337463, "global_step": 120066, "epoch": 1349, "lr": 5.917024527603697e-05}
{"train_loss": 0.11606777459383011, "global_step": 120067, "epoch": 1349, "lr": 5.916967537262464e-05}
{"train_loss": 0.12604589760303497, "global_step": 120068, "epoch": 1349, "lr": 5.916910546797956e-05}
{"train_loss": 0.18015748262405396, "global_step": 120069, "epoch": 1349, "lr": 5.91685355621018e-05}
{"train_loss": 0.1428873986005783, "global_step": 120070, "epoch": 1349, "lr": 5.916796565499145e-05}
{"train_loss": 0.09501008689403534, "global_step": 120071, "epoch": 1349, "lr": 5.916739574664858e-05}
{"train_loss": 0.053101636469364166, "global_step": 120072, "epoch": 1349, "lr": 5.916682583707325e-05}
{"train_loss": 0.1957179456949234, "global_step": 120073, "epoch": 1349, "lr": 5.916625592626557e-05}
{"train_loss": 0.15822282433509827, "global_step": 120074, "epoch": 1349, "lr": 5.916568601422558e-05}
{"train_loss": 0.13968032598495483, "global_step": 120075, "epoch": 1349, "lr": 5.916511610095339e-05}
{"train_loss": 0.1665208637714386, "global_step": 120076, "epoch": 1349, "lr": 5.916454618644907e-05}
{"train_loss": 0.13255932927131653, "global_step": 120077, "epoch": 1349, "lr": 5.9163976270712664e-05}
{"train_loss": 0.13693924248218536, "global_step": 120078, "epoch": 1349, "lr": 5.916340635374429e-05}
{"train_loss": 0.10393059998750687, "global_step": 120079, "epoch": 1349, "lr": 5.9162836435544e-05}
{"train_loss": 0.16251900792121887, "global_step": 120080, "epoch": 1349, "lr": 5.916226651611189e-05}
{"train_loss": 0.1327972263097763, "global_step": 120081, "epoch": 1349, "lr": 5.9161696595448016e-05}
{"train_loss": 0.1786900758743286, "global_step": 120082, "epoch": 1349, "lr": 5.916112667355246e-05}
{"train_loss": 0.1535770744085312, "global_step": 120083, "epoch": 1349, "lr": 5.91605567504253e-05}
{"train_loss": 0.1786317676305771, "global_step": 120084, "epoch": 1349, "lr": 5.915998682606663e-05}
{"train_loss": 0.17747023701667786, "global_step": 120085, "epoch": 1349, "lr": 5.9159416900476496e-05}
{"train_loss": 0.1445005238056183, "global_step": 120086, "epoch": 1349, "lr": 5.9158846973654994e-05}
{"train_loss": 0.15052036941051483, "global_step": 120087, "epoch": 1349, "lr": 5.91582770456022e-05}
{"train_loss": 0.14030808210372925, "global_step": 120088, "epoch": 1349, "lr": 5.9157707116318186e-05}
{"train_loss": 0.13495051860809326, "global_step": 120089, "epoch": 1349, "lr": 5.915713718580304e-05}
{"train_loss": 0.1045076996088028, "global_step": 120090, "epoch": 1349, "lr": 5.915656725405679e-05}
{"train_loss": 0.1766897588968277, "global_step": 120091, "epoch": 1349, "lr": 5.9155997321079595e-05}
{"train_loss": 0.08462579548358917, "global_step": 120092, "epoch": 1349, "lr": 5.9155427386871463e-05}
{"train_loss": 0.12203111499547958, "global_step": 120093, "epoch": 1349, "lr": 5.915485745143252e-05}
{"train_loss": 0.21063217520713806, "global_step": 120094, "epoch": 1349, "lr": 5.9154287514762795e-05}
{"train_loss": 0.13047552108764648, "global_step": 120095, "epoch": 1349, "lr": 5.91537175768624e-05}
{"train_loss": 0.10258075594902039, "global_step": 120096, "epoch": 1349, "lr": 5.91531476377314e-05}
{"train_loss": 0.1120014563202858, "global_step": 120097, "epoch": 1349, "lr": 5.915257769736986e-05}
{"train_loss": 0.20299001038074493, "global_step": 120098, "epoch": 1349, "lr": 5.915200775577787e-05}
{"train_loss": 0.12861154973506927, "global_step": 120099, "epoch": 1349, "lr": 5.915143781295551e-05}
{"train_loss": 0.13010208308696747, "global_step": 120100, "epoch": 1349, "lr": 5.915086786890285e-05}
{"train_loss": 0.18745028972625732, "global_step": 120101, "epoch": 1349, "lr": 5.9150297923619966e-05}
{"train_loss": 0.11646082997322083, "global_step": 120102, "epoch": 1349, "lr": 5.914972797710694e-05}
{"train_loss": 0.09195268154144287, "global_step": 120103, "epoch": 1349, "lr": 5.914915802936384e-05}
{"train_loss": 0.1716783493757248, "global_step": 120104, "epoch": 1349, "lr": 5.9148588080390744e-05}
{"train_loss": 0.1698356717824936, "global_step": 120105, "epoch": 1349, "lr": 5.914801813018773e-05}
{"train_loss": 0.09832614660263062, "global_step": 120106, "epoch": 1349, "lr": 5.9147448178754886e-05}
{"train_loss": 0.1315413862466812, "global_step": 120107, "epoch": 1349, "lr": 5.914687822609227e-05}
{"train_loss": 0.13751281797885895, "global_step": 120108, "epoch": 1349, "lr": 5.9146308272199955e-05}
{"train_loss": 0.11461088806390762, "global_step": 120109, "epoch": 1349, "lr": 5.914573831707806e-05}
{"train_loss": 0.1186489462852478, "global_step": 120110, "epoch": 1349, "lr": 5.914516836072661e-05}
{"train_loss": 0.18337516486644745, "global_step": 120111, "epoch": 1349, "lr": 5.9144598403145715e-05}
{"train_loss": 0.06993860751390457, "global_step": 120112, "epoch": 1349, "lr": 5.914402844433545e-05}
{"train_loss": 0.15948346257209778, "global_step": 120113, "epoch": 1349, "lr": 5.9143458484295864e-05}
{"train_loss": 0.12131942063570023, "global_step": 120114, "epoch": 1349, "lr": 5.914288852302706e-05}
{"train_loss": 0.10956969112157822, "global_step": 120115, "epoch": 1349, "lr": 5.914231856052911e-05}
{"train_loss": 0.11238524317741394, "global_step": 120116, "epoch": 1349, "lr": 5.914174859680208e-05}
{"train_loss": 0.07422010600566864, "global_step": 120117, "epoch": 1349, "lr": 5.914117863184605e-05}
{"train_loss": 0.1141120195388794, "global_step": 120118, "epoch": 1349, "lr": 5.91406086656611e-05}
{"train_loss": 0.10595765709877014, "global_step": 120119, "epoch": 1349, "lr": 5.914003869824731e-05}
{"train_loss": 0.19184145331382751, "global_step": 120120, "epoch": 1349, "lr": 5.913946872960476e-05}
{"train_loss": 0.0962502658367157, "global_step": 120121, "epoch": 1349, "lr": 5.913889875973351e-05}
{"train_loss": 0.14273163676261902, "global_step": 120122, "epoch": 1349, "lr": 5.913832878863364e-05}
{"train_loss": 0.13409988582134247, "global_step": 120123, "epoch": 1349, "lr": 5.9137758816305255e-05}
{"train_loss": 0.11961391568183899, "global_step": 120124, "epoch": 1349, "lr": 5.913718884274839e-05}
{"train_loss": 0.061184920370578766, "global_step": 120125, "epoch": 1349, "lr": 5.913661886796315e-05}
{"train_loss": 0.2129908949136734, "global_step": 120126, "epoch": 1349, "lr": 5.913604889194959e-05}
{"train_loss": 0.10680454224348068, "global_step": 120127, "epoch": 1349, "lr": 5.9135478914707806e-05}
{"train_loss": 0.11368701606988907, "global_step": 120128, "epoch": 1349, "lr": 5.913490893623789e-05}
{"train_loss": 0.11931656301021576, "global_step": 120129, "epoch": 1349, "lr": 5.913433895653987e-05}
{"train_loss": 0.16709710657596588, "global_step": 120130, "epoch": 1349, "lr": 5.913376897561387e-05}
{"train_loss": 0.1010817438364029, "global_step": 120131, "epoch": 1349, "lr": 5.913319899345994e-05}
{"train_loss": 0.1284283995628357, "global_step": 120132, "epoch": 1349, "lr": 5.913262901007816e-05}
{"train_loss": 0.12219934910535812, "global_step": 120133, "epoch": 1349, "lr": 5.91320590254686e-05}
{"train_loss": 0.16295404732227325, "global_step": 120134, "epoch": 1349, "lr": 5.913148903963136e-05}
{"train_loss": 0.1392366886138916, "global_step": 120135, "epoch": 1349, "lr": 5.913091905256649e-05}
{"train_loss": 0.13643653690814972, "global_step": 120136, "epoch": 1349, "lr": 5.913034906427409e-05}
{"train_loss": 0.1457214653491974, "global_step": 120137, "epoch": 1349, "lr": 5.912977907475422e-05}
{"train_loss": 0.12382009625434875, "global_step": 120138, "epoch": 1349, "lr": 5.912920908400697e-05}
{"train_loss": 0.15101301670074463, "global_step": 120139, "epoch": 1349, "lr": 5.91286390920324e-05}
{"train_loss": 0.08839350938796997, "global_step": 120140, "epoch": 1349, "lr": 5.9128069098830594e-05}
{"train_loss": 0.11096775531768799, "global_step": 120141, "epoch": 1349, "lr": 5.912749910440164e-05}
{"train_loss": 0.1040332242846489, "global_step": 120142, "epoch": 1349, "lr": 5.91269291087456e-05}
{"train_loss": 0.14758288860321045, "global_step": 120143, "epoch": 1349, "lr": 5.912635911186255e-05}
{"train_loss": 0.09517513960599899, "global_step": 120144, "epoch": 1349, "lr": 5.912578911375258e-05}
{"train_loss": 0.11474845558404922, "global_step": 120145, "epoch": 1349, "lr": 5.912521911441575e-05}
{"train_loss": 0.10806048661470413, "global_step": 120146, "epoch": 1349, "lr": 5.912464911385215e-05}
{"train_loss": 0.1425447165966034, "global_step": 120147, "epoch": 1349, "lr": 5.912407911206186e-05}
{"train_loss": 0.0739477127790451, "global_step": 120148, "epoch": 1349, "lr": 5.9123509109044937e-05}
{"train_loss": 0.1333027939950482, "global_step": 120149, "epoch": 1349, "lr": 5.912293910480148e-05, "val_loss": 5.178282737731934}
{"train_loss": 0.14570680260658264, "global_step": 120150, "epoch": 1350, "lr": 5.9122369099331544e-05}
{"train_loss": 0.14167098701000214, "global_step": 120151, "epoch": 1350, "lr": 5.912179909263524e-05}
{"train_loss": 0.1227646917104721, "global_step": 120152, "epoch": 1350, "lr": 5.9121229084712604e-05}
{"train_loss": 0.1553230732679367, "global_step": 120153, "epoch": 1350, "lr": 5.9120659075563734e-05}
{"train_loss": 0.085489422082901, "global_step": 120154, "epoch": 1350, "lr": 5.9120089065188686e-05}
{"train_loss": 0.13735178112983704, "global_step": 120155, "epoch": 1350, "lr": 5.911951905358757e-05}
{"train_loss": 0.1032525897026062, "global_step": 120156, "epoch": 1350, "lr": 5.9118949040760453e-05}
{"train_loss": 0.1519840657711029, "global_step": 120157, "epoch": 1350, "lr": 5.9118379026707385e-05}
{"train_loss": 0.15925678610801697, "global_step": 120158, "epoch": 1350, "lr": 5.911780901142847e-05}
{"train_loss": 0.1606438159942627, "global_step": 120159, "epoch": 1350, "lr": 5.911723899492378e-05}
{"train_loss": 0.10855425149202347, "global_step": 120160, "epoch": 1350, "lr": 5.911666897719339e-05}
{"train_loss": 0.19055749475955963, "global_step": 120161, "epoch": 1350, "lr": 5.9116098958237364e-05}
{"train_loss": 0.1345900595188141, "global_step": 120162, "epoch": 1350, "lr": 5.9115528938055796e-05}
{"train_loss": 0.10031616687774658, "global_step": 120163, "epoch": 1350, "lr": 5.9114958916648746e-05}
{"train_loss": 0.1420791745185852, "global_step": 120164, "epoch": 1350, "lr": 5.911438889401632e-05}
{"train_loss": 0.16354387998580933, "global_step": 120165, "epoch": 1350, "lr": 5.9113818870158565e-05}
{"train_loss": 0.09019766002893448, "global_step": 120166, "epoch": 1350, "lr": 5.911324884507556e-05}
{"train_loss": 0.1469576060771942, "global_step": 120167, "epoch": 1350, "lr": 5.911267881876741e-05}
{"train_loss": 0.15482312440872192, "global_step": 120168, "epoch": 1350, "lr": 5.911210879123416e-05}
{"train_loss": 0.16359692811965942, "global_step": 120169, "epoch": 1350, "lr": 5.911153876247589e-05}
{"train_loss": 0.18377919495105743, "global_step": 120170, "epoch": 1350, "lr": 5.911096873249271e-05}
{"train_loss": 0.1687459498643875, "global_step": 120171, "epoch": 1350, "lr": 5.9110398701284656e-05}
{"train_loss": 0.1580791473388672, "global_step": 120172, "epoch": 1350, "lr": 5.910982866885182e-05}
{"train_loss": 0.11513236165046692, "global_step": 120173, "epoch": 1350, "lr": 5.9109258635194275e-05}
{"train_loss": 0.0894051045179367, "global_step": 120174, "epoch": 1350, "lr": 5.9108688600312103e-05}
{"train_loss": 0.2185007929801941, "global_step": 120175, "epoch": 1350, "lr": 5.910811856420539e-05}
{"train_loss": 0.1277683973312378, "global_step": 120176, "epoch": 1350, "lr": 5.910754852687418e-05}
{"train_loss": 0.1379048079252243, "global_step": 120177, "epoch": 1350, "lr": 5.910697848831859e-05}
{"train_loss": 0.12013726681470871, "global_step": 120178, "epoch": 1350, "lr": 5.910640844853866e-05}
{"train_loss": 0.08541006594896317, "global_step": 120179, "epoch": 1350, "lr": 5.9105838407534496e-05}
{"train_loss": 0.16203546524047852, "global_step": 120180, "epoch": 1350, "lr": 5.9105268365306164e-05}
{"train_loss": 0.1329839676618576, "global_step": 120181, "epoch": 1350, "lr": 5.910469832185373e-05}
{"train_loss": 0.11020395904779434, "global_step": 120182, "epoch": 1350, "lr": 5.910412827717729e-05}
{"train_loss": 0.11625150591135025, "global_step": 120183, "epoch": 1350, "lr": 5.91035582312769e-05}
{"train_loss": 0.08613793551921844, "global_step": 120184, "epoch": 1350, "lr": 5.9102988184152654e-05}
{"train_loss": 0.20857055485248566, "global_step": 120185, "epoch": 1350, "lr": 5.9102418135804624e-05}
{"train_loss": 0.19945067167282104, "global_step": 120186, "epoch": 1350, "lr": 5.910184808623289e-05}
{"train_loss": 0.1206807866692543, "global_step": 120187, "epoch": 1350, "lr": 5.910127803543751e-05}
{"train_loss": 0.16171330213546753, "global_step": 120188, "epoch": 1350, "lr": 5.910070798341859e-05}
{"train_loss": 0.0996829941868782, "global_step": 120189, "epoch": 1350, "lr": 5.910013793017619e-05}
{"train_loss": 0.15940797328948975, "global_step": 120190, "epoch": 1350, "lr": 5.909956787571038e-05}
{"train_loss": 0.07440710812807083, "global_step": 120191, "epoch": 1350, "lr": 5.909899782002125e-05}
{"train_loss": 0.14438696205615997, "global_step": 120192, "epoch": 1350, "lr": 5.9098427763108866e-05}
{"train_loss": 0.11047805100679398, "global_step": 120193, "epoch": 1350, "lr": 5.90978577049733e-05}
{"train_loss": 0.15721522271633148, "global_step": 120194, "epoch": 1350, "lr": 5.909728764561465e-05}
{"train_loss": 0.12570783495903015, "global_step": 120195, "epoch": 1350, "lr": 5.909671758503298e-05}
{"train_loss": 0.1293569803237915, "global_step": 120196, "epoch": 1350, "lr": 5.9096147523228366e-05}
{"train_loss": 0.116154745221138, "global_step": 120197, "epoch": 1350, "lr": 5.909557746020088e-05}
{"train_loss": 0.13390450179576874, "global_step": 120198, "epoch": 1350, "lr": 5.909500739595061e-05}
{"train_loss": 0.12328147888183594, "global_step": 120199, "epoch": 1350, "lr": 5.909443733047761e-05}
{"train_loss": 0.11617159098386765, "global_step": 120200, "epoch": 1350, "lr": 5.909386726378199e-05}
{"train_loss": 0.12074513733386993, "global_step": 120201, "epoch": 1350, "lr": 5.909329719586382e-05}
{"train_loss": 0.15766052901744843, "global_step": 120202, "epoch": 1350, "lr": 5.909272712672315e-05}
{"train_loss": 0.1560564935207367, "global_step": 120203, "epoch": 1350, "lr": 5.9092157056360074e-05}
{"train_loss": 0.15571971237659454, "global_step": 120204, "epoch": 1350, "lr": 5.9091586984774684e-05}
{"train_loss": 0.15495702624320984, "global_step": 120205, "epoch": 1350, "lr": 5.909101691196703e-05}
{"train_loss": 0.17725816369056702, "global_step": 120206, "epoch": 1350, "lr": 5.9090446837937194e-05}
{"train_loss": 0.11421669274568558, "global_step": 120207, "epoch": 1350, "lr": 5.908987676268527e-05}
{"train_loss": 0.16856510937213898, "global_step": 120208, "epoch": 1350, "lr": 5.908930668621132e-05}
{"train_loss": 0.08666038513183594, "global_step": 120209, "epoch": 1350, "lr": 5.908873660851543e-05}
{"train_loss": 0.11154039204120636, "global_step": 120210, "epoch": 1350, "lr": 5.908816652959767e-05}
{"train_loss": 0.1178550124168396, "global_step": 120211, "epoch": 1350, "lr": 5.908759644945811e-05}
{"train_loss": 0.1019473522901535, "global_step": 120212, "epoch": 1350, "lr": 5.908702636809682e-05}
{"train_loss": 0.1600983887910843, "global_step": 120213, "epoch": 1350, "lr": 5.9086456285513906e-05}
{"train_loss": 0.15089556574821472, "global_step": 120214, "epoch": 1350, "lr": 5.908588620170943e-05}
{"train_loss": 0.12166247516870499, "global_step": 120215, "epoch": 1350, "lr": 5.9085316116683456e-05}
{"train_loss": 0.17612431943416595, "global_step": 120216, "epoch": 1350, "lr": 5.9084746030436075e-05}
{"train_loss": 0.07679511606693268, "global_step": 120217, "epoch": 1350, "lr": 5.9084175942967354e-05}
{"train_loss": 0.15456843376159668, "global_step": 120218, "epoch": 1350, "lr": 5.908360585427739e-05}
{"train_loss": 0.20128296315670013, "global_step": 120219, "epoch": 1350, "lr": 5.9083035764366244e-05}
{"train_loss": 0.17007209360599518, "global_step": 120220, "epoch": 1350, "lr": 5.908246567323399e-05}
{"train_loss": 0.0691085234284401, "global_step": 120221, "epoch": 1350, "lr": 5.908189558088071e-05}
{"train_loss": 0.17637278139591217, "global_step": 120222, "epoch": 1350, "lr": 5.9081325487306486e-05}
{"train_loss": 0.13733389973640442, "global_step": 120223, "epoch": 1350, "lr": 5.908075539251138e-05}
{"train_loss": 0.1107558086514473, "global_step": 120224, "epoch": 1350, "lr": 5.9080185296495484e-05}
{"train_loss": 0.17990174889564514, "global_step": 120225, "epoch": 1350, "lr": 5.907961519925887e-05}
{"train_loss": 0.18102231621742249, "global_step": 120226, "epoch": 1350, "lr": 5.90790451008016e-05}
{"train_loss": 0.184364452958107, "global_step": 120227, "epoch": 1350, "lr": 5.907847500112377e-05}
{"train_loss": 0.22556480765342712, "global_step": 120228, "epoch": 1350, "lr": 5.907790490022547e-05}
{"train_loss": 0.13642925024032593, "global_step": 120229, "epoch": 1350, "lr": 5.907733479810673e-05}
{"train_loss": 0.141569584608078, "global_step": 120230, "epoch": 1350, "lr": 5.907676469476766e-05}
{"train_loss": 0.1807480901479721, "global_step": 120231, "epoch": 1350, "lr": 5.907619459020832e-05}
{"train_loss": 0.2039160579442978, "global_step": 120232, "epoch": 1350, "lr": 5.9075624484428815e-05}
{"train_loss": 0.1417502909898758, "global_step": 120233, "epoch": 1350, "lr": 5.907505437742918e-05}
{"train_loss": 0.15059241652488708, "global_step": 120234, "epoch": 1350, "lr": 5.907448426920954e-05}
{"train_loss": 0.09586899727582932, "global_step": 120235, "epoch": 1350, "lr": 5.907391415976992e-05}
{"train_loss": 0.17519991099834442, "global_step": 120236, "epoch": 1350, "lr": 5.907334404911044e-05}
{"train_loss": 0.11674585938453674, "global_step": 120237, "epoch": 1350, "lr": 5.907277393723115e-05}
{"train_loss": 0.14023362376381843, "global_step": 120238, "epoch": 1350, "lr": 5.9072203824132146e-05, "train/sim_max_reward_0": 0.5340370928192109, "train/sim_max_reward_1": 0.9796840613168111, "train/sim_max_reward_2": 1.0, "train/sim_max_reward_3": 0.45833499784968074, "train/sim_max_reward_4": 0.9784930162355946, "train/sim_max_reward_5": 0.33386964771960886, "test/sim_max_reward_4500000": 1.0, "test/sim_max_reward_4500001": 0.9741425940829718, "test/sim_max_reward_4500002": 0.17583976048937675, "test/sim_max_reward_4500003": 0.6020712833288832, "test/sim_max_reward_4500004": 0.6519099446586547, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.7871658878446952, "test/sim_max_reward_4500008": 0.1565854547966244, "test/sim_max_reward_4500009": 0.327613815583706, "test/sim_max_reward_4500010": 1.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.9475785610978964, "test/sim_max_reward_4500014": 0.9895204717500059, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.28902983757288764, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.39496762743265906, "test/sim_max_reward_4500020": 0.010082264581501797, "test/sim_max_reward_4500021": 0.9846378543379383, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.9015821298084987, "test/sim_max_reward_4500024": 0.6411456891314573, "test/sim_max_reward_4500025": 0.9578344025881931, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.4307690522421939, "test/sim_max_reward_4500029": 0.37744652836584996, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.5227719737991691, "test/sim_max_reward_4500032": 0.3827557143975154, "test/sim_max_reward_4500033": 0.9995111986920343, "test/sim_max_reward_4500034": 0.6359441055460548, "test/sim_max_reward_4500035": 0.5315225115799976, "test/sim_max_reward_4500036": 0.9724839866561196, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.7186670995464663, "test/sim_max_reward_4500039": 0.9845671683899392, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.45964359448100806, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.47356135525831977, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.7589777944730652, "test/sim_max_reward_4500049": 0.49795802704374537, "train/mean_score": 0.7140698026568177, "test/mean_score": 0.6507657537911485, "val_loss": 5.308679103851318, "train_action_mse_error": 46.24003601074219}
{"train_loss": 0.10309368371963501, "global_step": 120239, "epoch": 1351, "lr": 5.907163370981349e-05}
{"train_loss": 0.11501086503267288, "global_step": 120240, "epoch": 1351, "lr": 5.907106359427527e-05}
{"train_loss": 0.12175629287958145, "global_step": 120241, "epoch": 1351, "lr": 5.907049347751754e-05}
{"train_loss": 0.16447488963603973, "global_step": 120242, "epoch": 1351, "lr": 5.9069923359540405e-05}
{"train_loss": 0.19786103069782257, "global_step": 120243, "epoch": 1351, "lr": 5.906935324034393e-05}
{"train_loss": 0.12792257964611053, "global_step": 120244, "epoch": 1351, "lr": 5.906878311992818e-05}
{"train_loss": 0.2202518731355667, "global_step": 120245, "epoch": 1351, "lr": 5.906821299829326e-05}
{"train_loss": 0.161026269197464, "global_step": 120246, "epoch": 1351, "lr": 5.906764287543922e-05}
{"train_loss": 0.12303090840578079, "global_step": 120247, "epoch": 1351, "lr": 5.906707275136616e-05}
{"train_loss": 0.14309050142765045, "global_step": 120248, "epoch": 1351, "lr": 5.906650262607413e-05}
{"train_loss": 0.15684041380882263, "global_step": 120249, "epoch": 1351, "lr": 5.9065932499563214e-05}
{"train_loss": 0.13062581419944763, "global_step": 120250, "epoch": 1351, "lr": 5.90653623718335e-05}
{"train_loss": 0.15858644247055054, "global_step": 120251, "epoch": 1351, "lr": 5.906479224288506e-05}
{"train_loss": 0.13107310235500336, "global_step": 120252, "epoch": 1351, "lr": 5.906422211271796e-05}
{"train_loss": 0.20497353374958038, "global_step": 120253, "epoch": 1351, "lr": 5.906365198133228e-05}
{"train_loss": 0.20977680385112762, "global_step": 120254, "epoch": 1351, "lr": 5.906308184872812e-05}
{"train_loss": 0.19508881866931915, "global_step": 120255, "epoch": 1351, "lr": 5.9062511714905534e-05}
{"train_loss": 0.15407899022102356, "global_step": 120256, "epoch": 1351, "lr": 5.90619415798646e-05}
{"train_loss": 0.18162038922309875, "global_step": 120257, "epoch": 1351, "lr": 5.9061371443605396e-05}
{"train_loss": 0.1045464277267456, "global_step": 120258, "epoch": 1351, "lr": 5.906080130612801e-05}
{"train_loss": 0.10953591018915176, "global_step": 120259, "epoch": 1351, "lr": 5.9060231167432514e-05}
{"train_loss": 0.20264260470867157, "global_step": 120260, "epoch": 1351, "lr": 5.905966102751896e-05}
{"train_loss": 0.1156994104385376, "global_step": 120261, "epoch": 1351, "lr": 5.905909088638746e-05}
{"train_loss": 0.13572072982788086, "global_step": 120262, "epoch": 1351, "lr": 5.905852074403807e-05}
{"train_loss": 0.1570224165916443, "global_step": 120263, "epoch": 1351, "lr": 5.905795060047088e-05}
{"train_loss": 0.14278604090213776, "global_step": 120264, "epoch": 1351, "lr": 5.905738045568595e-05}
{"train_loss": 0.15839572250843048, "global_step": 120265, "epoch": 1351, "lr": 5.905681030968338e-05}
{"train_loss": 0.16656984388828278, "global_step": 120266, "epoch": 1351, "lr": 5.905624016246322e-05}
{"train_loss": 0.08709652721881866, "global_step": 120267, "epoch": 1351, "lr": 5.905567001402557e-05}
{"train_loss": 0.1355752795934677, "global_step": 120268, "epoch": 1351, "lr": 5.9055099864370486e-05}
{"train_loss": 0.17334042489528656, "global_step": 120269, "epoch": 1351, "lr": 5.9054529713498055e-05}
{"train_loss": 0.15687048435211182, "global_step": 120270, "epoch": 1351, "lr": 5.905395956140836e-05}
{"train_loss": 0.14430543780326843, "global_step": 120271, "epoch": 1351, "lr": 5.905338940810146e-05}
{"train_loss": 0.09588669240474701, "global_step": 120272, "epoch": 1351, "lr": 5.905281925357743e-05}
{"train_loss": 0.0954824686050415, "global_step": 120273, "epoch": 1351, "lr": 5.9052249097836375e-05}
{"train_loss": 0.18988288938999176, "global_step": 120274, "epoch": 1351, "lr": 5.905167894087835e-05}
{"train_loss": 0.12861524522304535, "global_step": 120275, "epoch": 1351, "lr": 5.905110878270345e-05}
{"train_loss": 0.10640622675418854, "global_step": 120276, "epoch": 1351, "lr": 5.905053862331172e-05}
{"train_loss": 0.15811702609062195, "global_step": 120277, "epoch": 1351, "lr": 5.904996846270328e-05}
{"train_loss": 0.15290087461471558, "global_step": 120278, "epoch": 1351, "lr": 5.9049398300878156e-05}
{"train_loss": 0.15464282035827637, "global_step": 120279, "epoch": 1351, "lr": 5.9048828137836465e-05}
{"train_loss": 0.15453395247459412, "global_step": 120280, "epoch": 1351, "lr": 5.904825797357827e-05}
{"train_loss": 0.11552762985229492, "global_step": 120281, "epoch": 1351, "lr": 5.9047687808103636e-05}
{"train_loss": 0.14129886031150818, "global_step": 120282, "epoch": 1351, "lr": 5.904711764141266e-05}
{"train_loss": 0.15376566350460052, "global_step": 120283, "epoch": 1351, "lr": 5.904654747350541e-05}
{"train_loss": 0.11330898106098175, "global_step": 120284, "epoch": 1351, "lr": 5.904597730438196e-05}
{"train_loss": 0.14495441317558289, "global_step": 120285, "epoch": 1351, "lr": 5.904540713404238e-05}
{"train_loss": 0.12276210635900497, "global_step": 120286, "epoch": 1351, "lr": 5.904483696248678e-05}
{"train_loss": 0.17306344211101532, "global_step": 120287, "epoch": 1351, "lr": 5.90442667897152e-05}
{"train_loss": 0.11535850912332535, "global_step": 120288, "epoch": 1351, "lr": 5.904369661572772e-05}
{"train_loss": 0.14070743322372437, "global_step": 120289, "epoch": 1351, "lr": 5.904312644052443e-05}
{"train_loss": 0.1233324483036995, "global_step": 120290, "epoch": 1351, "lr": 5.90425562641054e-05}
{"train_loss": 0.1323387175798416, "global_step": 120291, "epoch": 1351, "lr": 5.9041986086470703e-05}
{"train_loss": 0.14641836285591125, "global_step": 120292, "epoch": 1351, "lr": 5.904141590762042e-05}
{"train_loss": 0.1551099568605423, "global_step": 120293, "epoch": 1351, "lr": 5.9040845727554636e-05}
{"train_loss": 0.11539685726165771, "global_step": 120294, "epoch": 1351, "lr": 5.9040275546273424e-05}
{"train_loss": 0.0961870327591896, "global_step": 120295, "epoch": 1351, "lr": 5.9039705363776856e-05}
{"train_loss": 0.11341848969459534, "global_step": 120296, "epoch": 1351, "lr": 5.9039135180065006e-05}
{"train_loss": 0.15136489272117615, "global_step": 120297, "epoch": 1351, "lr": 5.9038564995137955e-05}
{"train_loss": 0.11173292994499207, "global_step": 120298, "epoch": 1351, "lr": 5.903799480899579e-05}
{"train_loss": 0.08875062316656113, "global_step": 120299, "epoch": 1351, "lr": 5.903742462163856e-05}
{"train_loss": 0.11245917528867722, "global_step": 120300, "epoch": 1351, "lr": 5.903685443306637e-05}
{"train_loss": 0.11787138879299164, "global_step": 120301, "epoch": 1351, "lr": 5.9036284243279286e-05}
{"train_loss": 0.13725045323371887, "global_step": 120302, "epoch": 1351, "lr": 5.903571405227737e-05}
{"train_loss": 0.12387725710868835, "global_step": 120303, "epoch": 1351, "lr": 5.9035143860060725e-05}
{"train_loss": 0.1146729439496994, "global_step": 120304, "epoch": 1351, "lr": 5.903457366662941e-05}
{"train_loss": 0.1467227339744568, "global_step": 120305, "epoch": 1351, "lr": 5.9034003471983515e-05}
{"train_loss": 0.12393359839916229, "global_step": 120306, "epoch": 1351, "lr": 5.9033433276123105e-05}
{"train_loss": 0.16237770020961761, "global_step": 120307, "epoch": 1351, "lr": 5.903286307904825e-05}
{"train_loss": 0.13179364800453186, "global_step": 120308, "epoch": 1351, "lr": 5.903229288075903e-05}
{"train_loss": 0.1281120628118515, "global_step": 120309, "epoch": 1351, "lr": 5.9031722681255555e-05}
{"train_loss": 0.09153196960687637, "global_step": 120310, "epoch": 1351, "lr": 5.903115248053786e-05}
{"train_loss": 0.1649126410484314, "global_step": 120311, "epoch": 1351, "lr": 5.903058227860603e-05}
{"train_loss": 0.14328637719154358, "global_step": 120312, "epoch": 1351, "lr": 5.903001207546016e-05}
{"train_loss": 0.13427968323230743, "global_step": 120313, "epoch": 1351, "lr": 5.902944187110031e-05}
{"train_loss": 0.1521403193473816, "global_step": 120314, "epoch": 1351, "lr": 5.902887166552656e-05}
{"train_loss": 0.12281258404254913, "global_step": 120315, "epoch": 1351, "lr": 5.9028301458738985e-05}
{"train_loss": 0.1402309685945511, "global_step": 120316, "epoch": 1351, "lr": 5.902773125073767e-05}
{"train_loss": 0.19989222288131714, "global_step": 120317, "epoch": 1351, "lr": 5.9027161041522696e-05}
{"train_loss": 0.07372675836086273, "global_step": 120318, "epoch": 1351, "lr": 5.902659083109412e-05}
{"train_loss": 0.1722216159105301, "global_step": 120319, "epoch": 1351, "lr": 5.902602061945203e-05}
{"train_loss": 0.12564672529697418, "global_step": 120320, "epoch": 1351, "lr": 5.9025450406596505e-05}
{"train_loss": 0.11961254477500916, "global_step": 120321, "epoch": 1351, "lr": 5.902488019252761e-05}
{"train_loss": 0.15051446855068207, "global_step": 120322, "epoch": 1351, "lr": 5.902430997724544e-05}
{"train_loss": 0.13099338114261627, "global_step": 120323, "epoch": 1351, "lr": 5.902373976075005e-05}
{"train_loss": 0.1354604959487915, "global_step": 120324, "epoch": 1351, "lr": 5.902316954304153e-05}
{"train_loss": 0.08333031833171844, "global_step": 120325, "epoch": 1351, "lr": 5.902259932411997e-05}
{"train_loss": 0.13728740811347961, "global_step": 120326, "epoch": 1351, "lr": 5.902202910398542e-05}
{"train_loss": 0.13913051725438472, "global_step": 120327, "epoch": 1351, "lr": 5.902145888263797e-05, "val_loss": 5.483218193054199}
{"train_loss": 0.12336815148591995, "global_step": 120328, "epoch": 1352, "lr": 5.902088866007769e-05}
{"train_loss": 0.23524364829063416, "global_step": 120329, "epoch": 1352, "lr": 5.902031843630467e-05}
{"train_loss": 0.12797406315803528, "global_step": 120330, "epoch": 1352, "lr": 5.901974821131897e-05}
{"train_loss": 0.061421725898981094, "global_step": 120331, "epoch": 1352, "lr": 5.901917798512068e-05}
{"train_loss": 0.1572878211736679, "global_step": 120332, "epoch": 1352, "lr": 5.9018607757709864e-05}
{"train_loss": 0.19472308456897736, "global_step": 120333, "epoch": 1352, "lr": 5.901803752908661e-05}
{"train_loss": 0.1321852058172226, "global_step": 120334, "epoch": 1352, "lr": 5.901746729925099e-05}
{"train_loss": 0.15903763473033905, "global_step": 120335, "epoch": 1352, "lr": 5.901689706820309e-05}
{"train_loss": 0.11235193908214569, "global_step": 120336, "epoch": 1352, "lr": 5.901632683594297e-05}
{"train_loss": 0.18320131301879883, "global_step": 120337, "epoch": 1352, "lr": 5.9015756602470715e-05}
{"train_loss": 0.1734405755996704, "global_step": 120338, "epoch": 1352, "lr": 5.90151863677864e-05}
{"train_loss": 0.17353472113609314, "global_step": 120339, "epoch": 1352, "lr": 5.901461613189011e-05}
{"train_loss": 0.18088693916797638, "global_step": 120340, "epoch": 1352, "lr": 5.901404589478191e-05}
{"train_loss": 0.10452380776405334, "global_step": 120341, "epoch": 1352, "lr": 5.901347565646188e-05}
{"train_loss": 0.1265188455581665, "global_step": 120342, "epoch": 1352, "lr": 5.901290541693011e-05}
{"train_loss": 0.26460808515548706, "global_step": 120343, "epoch": 1352, "lr": 5.901233517618665e-05}
{"train_loss": 0.17924101650714874, "global_step": 120344, "epoch": 1352, "lr": 5.901176493423159e-05}
{"train_loss": 0.12223780155181885, "global_step": 120345, "epoch": 1352, "lr": 5.901119469106502e-05}
{"train_loss": 0.09769483655691147, "global_step": 120346, "epoch": 1352, "lr": 5.9010624446687005e-05}
{"train_loss": 0.12294274568557739, "global_step": 120347, "epoch": 1352, "lr": 5.901005420109761e-05}
{"train_loss": 0.10252353549003601, "global_step": 120348, "epoch": 1352, "lr": 5.900948395429693e-05}
{"train_loss": 0.12475377321243286, "global_step": 120349, "epoch": 1352, "lr": 5.900891370628503e-05}
{"train_loss": 0.12720325589179993, "global_step": 120350, "epoch": 1352, "lr": 5.9008343457062e-05}
{"train_loss": 0.1410752534866333, "global_step": 120351, "epoch": 1352, "lr": 5.90077732066279e-05}
{"train_loss": 0.10084569454193115, "global_step": 120352, "epoch": 1352, "lr": 5.900720295498282e-05}
{"train_loss": 0.1066470518708229, "global_step": 120353, "epoch": 1352, "lr": 5.900663270212683e-05}
{"train_loss": 0.09254513680934906, "global_step": 120354, "epoch": 1352, "lr": 5.9006062448060006e-05}
{"train_loss": 0.13259659707546234, "global_step": 120355, "epoch": 1352, "lr": 5.900549219278243e-05}
{"train_loss": 0.19611859321594238, "global_step": 120356, "epoch": 1352, "lr": 5.9004921936294166e-05}
{"train_loss": 0.13714230060577393, "global_step": 120357, "epoch": 1352, "lr": 5.90043516785953e-05}
{"train_loss": 0.1429043561220169, "global_step": 120358, "epoch": 1352, "lr": 5.900378141968592e-05}
{"train_loss": 0.13005690276622772, "global_step": 120359, "epoch": 1352, "lr": 5.900321115956609e-05}
{"train_loss": 0.15597935020923615, "global_step": 120360, "epoch": 1352, "lr": 5.9002640898235885e-05}
{"train_loss": 0.09153205901384354, "global_step": 120361, "epoch": 1352, "lr": 5.900207063569538e-05}
{"train_loss": 0.13500618934631348, "global_step": 120362, "epoch": 1352, "lr": 5.9001500371944663e-05}
{"train_loss": 0.15969927608966827, "global_step": 120363, "epoch": 1352, "lr": 5.90009301069838e-05}
{"train_loss": 0.16007661819458008, "global_step": 120364, "epoch": 1352, "lr": 5.900035984081288e-05}
{"train_loss": 0.12745651602745056, "global_step": 120365, "epoch": 1352, "lr": 5.899978957343196e-05}
{"train_loss": 0.17329098284244537, "global_step": 120366, "epoch": 1352, "lr": 5.8999219304841135e-05}
{"train_loss": 0.14362196624279022, "global_step": 120367, "epoch": 1352, "lr": 5.8998649035040474e-05}
{"train_loss": 0.12323755770921707, "global_step": 120368, "epoch": 1352, "lr": 5.8998078764030054e-05}
{"train_loss": 0.15583999454975128, "global_step": 120369, "epoch": 1352, "lr": 5.899750849180995e-05}
{"train_loss": 0.11980722844600677, "global_step": 120370, "epoch": 1352, "lr": 5.899693821838024e-05}
{"train_loss": 0.10802771151065826, "global_step": 120371, "epoch": 1352, "lr": 5.8996367943741e-05}
{"train_loss": 0.17062260210514069, "global_step": 120372, "epoch": 1352, "lr": 5.8995797667892316e-05}
{"train_loss": 0.10762973129749298, "global_step": 120373, "epoch": 1352, "lr": 5.899522739083425e-05}
{"train_loss": 0.2363928258419037, "global_step": 120374, "epoch": 1352, "lr": 5.8994657112566886e-05}
{"train_loss": 0.18092840909957886, "global_step": 120375, "epoch": 1352, "lr": 5.8994086833090304e-05}
{"train_loss": 0.13295356929302216, "global_step": 120376, "epoch": 1352, "lr": 5.899351655240457e-05}
{"train_loss": 0.1107955351471901, "global_step": 120377, "epoch": 1352, "lr": 5.8992946270509765e-05}
{"train_loss": 0.11013025045394897, "global_step": 120378, "epoch": 1352, "lr": 5.899237598740598e-05}
{"train_loss": 0.17190973460674286, "global_step": 120379, "epoch": 1352, "lr": 5.899180570309327e-05}
{"train_loss": 0.11554425954818726, "global_step": 120380, "epoch": 1352, "lr": 5.899123541757172e-05}
{"train_loss": 0.09141378104686737, "global_step": 120381, "epoch": 1352, "lr": 5.899066513084142e-05}
{"train_loss": 0.17480529844760895, "global_step": 120382, "epoch": 1352, "lr": 5.899009484290242e-05}
{"train_loss": 0.13506917655467987, "global_step": 120383, "epoch": 1352, "lr": 5.898952455375482e-05}
{"train_loss": 0.12041133642196655, "global_step": 120384, "epoch": 1352, "lr": 5.8988954263398686e-05}
{"train_loss": 0.12508058547973633, "global_step": 120385, "epoch": 1352, "lr": 5.8988383971834103e-05}
{"train_loss": 0.08106252551078796, "global_step": 120386, "epoch": 1352, "lr": 5.898781367906113e-05}
{"train_loss": 0.10297859460115433, "global_step": 120387, "epoch": 1352, "lr": 5.898724338507986e-05}
{"train_loss": 0.13412246108055115, "global_step": 120388, "epoch": 1352, "lr": 5.8986673089890364e-05}
{"train_loss": 0.11755815893411636, "global_step": 120389, "epoch": 1352, "lr": 5.8986102793492726e-05}
{"train_loss": 0.11316687613725662, "global_step": 120390, "epoch": 1352, "lr": 5.898553249588701e-05}
{"train_loss": 0.15400785207748413, "global_step": 120391, "epoch": 1352, "lr": 5.89849621970733e-05}
{"train_loss": 0.12510131299495697, "global_step": 120392, "epoch": 1352, "lr": 5.898439189705167e-05}
{"train_loss": 0.22226956486701965, "global_step": 120393, "epoch": 1352, "lr": 5.89838215958222e-05}
{"train_loss": 0.13147100806236267, "global_step": 120394, "epoch": 1352, "lr": 5.898325129338495e-05}
{"train_loss": 0.1236199140548706, "global_step": 120395, "epoch": 1352, "lr": 5.8982680989740034e-05}
{"train_loss": 0.11232484877109528, "global_step": 120396, "epoch": 1352, "lr": 5.898211068488749e-05}
{"train_loss": 0.25967150926589966, "global_step": 120397, "epoch": 1352, "lr": 5.898154037882743e-05}
{"train_loss": 0.16858170926570892, "global_step": 120398, "epoch": 1352, "lr": 5.898097007155989e-05}
{"train_loss": 0.06798234581947327, "global_step": 120399, "epoch": 1352, "lr": 5.898039976308498e-05}
{"train_loss": 0.13651159405708313, "global_step": 120400, "epoch": 1352, "lr": 5.897982945340276e-05}
{"train_loss": 0.11368201673030853, "global_step": 120401, "epoch": 1352, "lr": 5.897925914251331e-05}
{"train_loss": 0.15803959965705872, "global_step": 120402, "epoch": 1352, "lr": 5.897868883041671e-05}
{"train_loss": 0.09922505915164948, "global_step": 120403, "epoch": 1352, "lr": 5.897811851711304e-05}
{"train_loss": 0.1722593903541565, "global_step": 120404, "epoch": 1352, "lr": 5.897754820260236e-05}
{"train_loss": 0.1877167522907257, "global_step": 120405, "epoch": 1352, "lr": 5.897697788688477e-05}
{"train_loss": 0.13665561378002167, "global_step": 120406, "epoch": 1352, "lr": 5.8976407569960324e-05}
{"train_loss": 0.167324036359787, "global_step": 120407, "epoch": 1352, "lr": 5.8975837251829116e-05}
{"train_loss": 0.08234068751335144, "global_step": 120408, "epoch": 1352, "lr": 5.8975266932491224e-05}
{"train_loss": 0.11043548583984375, "global_step": 120409, "epoch": 1352, "lr": 5.8974696611946703e-05}
{"train_loss": 0.08910495042800903, "global_step": 120410, "epoch": 1352, "lr": 5.8974126290195656e-05}
{"train_loss": 0.13001534342765808, "global_step": 120411, "epoch": 1352, "lr": 5.897355596723813e-05}
{"train_loss": 0.15657025575637817, "global_step": 120412, "epoch": 1352, "lr": 5.897298564307423e-05}
{"train_loss": 0.09593792259693146, "global_step": 120413, "epoch": 1352, "lr": 5.897241531770402e-05}
{"train_loss": 0.1291482299566269, "global_step": 120414, "epoch": 1352, "lr": 5.897184499112758e-05}
{"train_loss": 0.13285917043685913, "global_step": 120415, "epoch": 1352, "lr": 5.897127466334498e-05}
{"train_loss": 0.13866827306285334, "global_step": 120416, "epoch": 1352, "lr": 5.89707043343563e-05, "val_loss": 5.346920967102051}
{"train_loss": 0.08779340237379074, "global_step": 120417, "epoch": 1353, "lr": 5.897013400416163e-05}
{"train_loss": 0.1787692904472351, "global_step": 120418, "epoch": 1353, "lr": 5.896956367276102e-05}
{"train_loss": 0.1713874638080597, "global_step": 120419, "epoch": 1353, "lr": 5.8968993340154566e-05}
{"train_loss": 0.07188989222049713, "global_step": 120420, "epoch": 1353, "lr": 5.8968423006342345e-05}
{"train_loss": 0.10222139209508896, "global_step": 120421, "epoch": 1353, "lr": 5.896785267132443e-05}
{"train_loss": 0.1385643035173416, "global_step": 120422, "epoch": 1353, "lr": 5.8967282335100896e-05}
{"train_loss": 0.16054487228393555, "global_step": 120423, "epoch": 1353, "lr": 5.896671199767181e-05}
{"train_loss": 0.124263234436512, "global_step": 120424, "epoch": 1353, "lr": 5.896614165903727e-05}
{"train_loss": 0.18712857365608215, "global_step": 120425, "epoch": 1353, "lr": 5.896557131919733e-05}
{"train_loss": 0.11587969958782196, "global_step": 120426, "epoch": 1353, "lr": 5.89650009781521e-05}
{"train_loss": 0.09612870961427689, "global_step": 120427, "epoch": 1353, "lr": 5.896443063590161e-05}
{"train_loss": 0.13036802411079407, "global_step": 120428, "epoch": 1353, "lr": 5.896386029244597e-05}
{"train_loss": 0.10070814192295074, "global_step": 120429, "epoch": 1353, "lr": 5.896328994778526e-05}
{"train_loss": 0.1624055802822113, "global_step": 120430, "epoch": 1353, "lr": 5.8962719601919524e-05}
{"train_loss": 0.117405965924263, "global_step": 120431, "epoch": 1353, "lr": 5.896214925484887e-05}
{"train_loss": 0.14270463585853577, "global_step": 120432, "epoch": 1353, "lr": 5.896157890657337e-05}
{"train_loss": 0.16012835502624512, "global_step": 120433, "epoch": 1353, "lr": 5.896100855709309e-05}
{"train_loss": 0.13118378818035126, "global_step": 120434, "epoch": 1353, "lr": 5.896043820640811e-05}
{"train_loss": 0.21790899336338043, "global_step": 120435, "epoch": 1353, "lr": 5.8959867854518503e-05}
{"train_loss": 0.16722898185253143, "global_step": 120436, "epoch": 1353, "lr": 5.895929750142437e-05}
{"train_loss": 0.11187732219696045, "global_step": 120437, "epoch": 1353, "lr": 5.8958727147125756e-05}
{"train_loss": 0.08893298357725143, "global_step": 120438, "epoch": 1353, "lr": 5.8958156791622756e-05}
{"train_loss": 0.21410484611988068, "global_step": 120439, "epoch": 1353, "lr": 5.895758643491544e-05}
{"train_loss": 0.08621755242347717, "global_step": 120440, "epoch": 1353, "lr": 5.895701607700387e-05}
{"train_loss": 0.10664515942335129, "global_step": 120441, "epoch": 1353, "lr": 5.895644571788815e-05}
{"train_loss": 0.06499942392110825, "global_step": 120442, "epoch": 1353, "lr": 5.8955875357568346e-05}
{"train_loss": 0.09078631550073624, "global_step": 120443, "epoch": 1353, "lr": 5.8955304996044534e-05}
{"train_loss": 0.08969252556562424, "global_step": 120444, "epoch": 1353, "lr": 5.895473463331679e-05}
{"train_loss": 0.12997522950172424, "global_step": 120445, "epoch": 1353, "lr": 5.8954164269385195e-05}
{"train_loss": 0.1377236694097519, "global_step": 120446, "epoch": 1353, "lr": 5.895359390424982e-05}
{"train_loss": 0.2201840579509735, "global_step": 120447, "epoch": 1353, "lr": 5.895302353791073e-05}
{"train_loss": 0.13093286752700806, "global_step": 120448, "epoch": 1353, "lr": 5.8952453170368026e-05}
{"train_loss": 0.10402980446815491, "global_step": 120449, "epoch": 1353, "lr": 5.8951882801621774e-05}
{"train_loss": 0.07595297694206238, "global_step": 120450, "epoch": 1353, "lr": 5.8951312431672044e-05}
{"train_loss": 0.13819648325443268, "global_step": 120451, "epoch": 1353, "lr": 5.8950742060518925e-05}
{"train_loss": 0.11812137812376022, "global_step": 120452, "epoch": 1353, "lr": 5.895017168816249e-05}
{"train_loss": 0.10651898384094238, "global_step": 120453, "epoch": 1353, "lr": 5.894960131460281e-05}
{"train_loss": 0.09177882224321365, "global_step": 120454, "epoch": 1353, "lr": 5.8949030939839965e-05}
{"train_loss": 0.11461837589740753, "global_step": 120455, "epoch": 1353, "lr": 5.8948460563874045e-05}
{"train_loss": 0.16668571531772614, "global_step": 120456, "epoch": 1353, "lr": 5.89478901867051e-05}
{"train_loss": 0.18381455540657043, "global_step": 120457, "epoch": 1353, "lr": 5.894731980833322e-05}
{"train_loss": 0.11811176687479019, "global_step": 120458, "epoch": 1353, "lr": 5.8946749428758486e-05}
{"train_loss": 0.12061445415019989, "global_step": 120459, "epoch": 1353, "lr": 5.8946179047980965e-05}
{"train_loss": 0.1080360859632492, "global_step": 120460, "epoch": 1353, "lr": 5.894560866600074e-05}
{"train_loss": 0.21873892843723297, "global_step": 120461, "epoch": 1353, "lr": 5.894503828281789e-05}
{"train_loss": 0.11570762097835541, "global_step": 120462, "epoch": 1353, "lr": 5.8944467898432486e-05}
{"train_loss": 0.18113836646080017, "global_step": 120463, "epoch": 1353, "lr": 5.894389751284462e-05}
{"train_loss": 0.10958722233772278, "global_step": 120464, "epoch": 1353, "lr": 5.8943327126054334e-05}
{"train_loss": 0.10836026817560196, "global_step": 120465, "epoch": 1353, "lr": 5.8942756738061746e-05}
{"train_loss": 0.11425741016864777, "global_step": 120466, "epoch": 1353, "lr": 5.89421863488669e-05}
{"train_loss": 0.1479390263557434, "global_step": 120467, "epoch": 1353, "lr": 5.8941615958469895e-05}
{"train_loss": 0.11045820266008377, "global_step": 120468, "epoch": 1353, "lr": 5.89410455668708e-05}
{"train_loss": 0.15404294431209564, "global_step": 120469, "epoch": 1353, "lr": 5.894047517406969e-05}
{"train_loss": 0.0882902443408966, "global_step": 120470, "epoch": 1353, "lr": 5.8939904780066636e-05}
{"train_loss": 0.08374622464179993, "global_step": 120471, "epoch": 1353, "lr": 5.893933438486171e-05}
{"train_loss": 0.08933375775814056, "global_step": 120472, "epoch": 1353, "lr": 5.8938763988455016e-05}
{"train_loss": 0.153468057513237, "global_step": 120473, "epoch": 1353, "lr": 5.8938193590846604e-05}
{"train_loss": 0.11115893721580505, "global_step": 120474, "epoch": 1353, "lr": 5.893762319203657e-05}
{"train_loss": 0.15342658758163452, "global_step": 120475, "epoch": 1353, "lr": 5.8937052792024985e-05}
{"train_loss": 0.13612189888954163, "global_step": 120476, "epoch": 1353, "lr": 5.893648239081192e-05}
{"train_loss": 0.121515654027462, "global_step": 120477, "epoch": 1353, "lr": 5.893591198839745e-05}
{"train_loss": 0.13967126607894897, "global_step": 120478, "epoch": 1353, "lr": 5.893534158478166e-05}
{"train_loss": 0.14482846856117249, "global_step": 120479, "epoch": 1353, "lr": 5.893477117996461e-05}
{"train_loss": 0.11430109292268753, "global_step": 120480, "epoch": 1353, "lr": 5.893420077394639e-05}
{"train_loss": 0.08503729104995728, "global_step": 120481, "epoch": 1353, "lr": 5.8933630366727085e-05}
{"train_loss": 0.14134110510349274, "global_step": 120482, "epoch": 1353, "lr": 5.893305995830676e-05}
{"train_loss": 0.08587981015443802, "global_step": 120483, "epoch": 1353, "lr": 5.89324895486855e-05}
{"train_loss": 0.08244311064481735, "global_step": 120484, "epoch": 1353, "lr": 5.893191913786337e-05}
{"train_loss": 0.13811545073986053, "global_step": 120485, "epoch": 1353, "lr": 5.893134872584044e-05}
{"train_loss": 0.12452776730060577, "global_step": 120486, "epoch": 1353, "lr": 5.8930778312616816e-05}
{"train_loss": 0.19362451136112213, "global_step": 120487, "epoch": 1353, "lr": 5.8930207898192544e-05}
{"train_loss": 0.17804744839668274, "global_step": 120488, "epoch": 1353, "lr": 5.892963748256772e-05}
{"train_loss": 0.10666411370038986, "global_step": 120489, "epoch": 1353, "lr": 5.8929067065742416e-05}
{"train_loss": 0.11881563067436218, "global_step": 120490, "epoch": 1353, "lr": 5.8928496647716715e-05}
{"train_loss": 0.18429702520370483, "global_step": 120491, "epoch": 1353, "lr": 5.892792622849067e-05}
{"train_loss": 0.16452156007289886, "global_step": 120492, "epoch": 1353, "lr": 5.892735580806439e-05}
{"train_loss": 0.1950550228357315, "global_step": 120493, "epoch": 1353, "lr": 5.892678538643793e-05}
{"train_loss": 0.15620991587638855, "global_step": 120494, "epoch": 1353, "lr": 5.892621496361138e-05}
{"train_loss": 0.11543495953083038, "global_step": 120495, "epoch": 1353, "lr": 5.89256445395848e-05}
{"train_loss": 0.13833749294281006, "global_step": 120496, "epoch": 1353, "lr": 5.892507411435828e-05}
{"train_loss": 0.11805754899978638, "global_step": 120497, "epoch": 1353, "lr": 5.89245036879319e-05}
{"train_loss": 0.13673914968967438, "global_step": 120498, "epoch": 1353, "lr": 5.892393326030572e-05}
{"train_loss": 0.12235397845506668, "global_step": 120499, "epoch": 1353, "lr": 5.8923362831479825e-05}
{"train_loss": 0.15825508534908295, "global_step": 120500, "epoch": 1353, "lr": 5.892279240145428e-05}
{"train_loss": 0.06923162937164307, "global_step": 120501, "epoch": 1353, "lr": 5.8922221970229194e-05}
{"train_loss": 0.07967333495616913, "global_step": 120502, "epoch": 1353, "lr": 5.8921651537804625e-05}
{"train_loss": 0.18367527425289154, "global_step": 120503, "epoch": 1353, "lr": 5.892108110418064e-05}
{"train_loss": 0.10726828873157501, "global_step": 120504, "epoch": 1353, "lr": 5.892051066935732e-05}
{"train_loss": 0.13025425474965172, "global_step": 120505, "epoch": 1353, "lr": 5.8919940233334755e-05, "val_loss": 5.556197166442871}
{"train_loss": 0.14401529729366302, "global_step": 120506, "epoch": 1354, "lr": 5.891936979611301e-05}
{"train_loss": 0.10150956362485886, "global_step": 120507, "epoch": 1354, "lr": 5.8918799357692154e-05}
{"train_loss": 0.09983189404010773, "global_step": 120508, "epoch": 1354, "lr": 5.8918228918072284e-05}
{"train_loss": 0.11296886205673218, "global_step": 120509, "epoch": 1354, "lr": 5.891765847725347e-05}
{"train_loss": 0.15937648713588715, "global_step": 120510, "epoch": 1354, "lr": 5.891708803523578e-05}
{"train_loss": 0.08899751305580139, "global_step": 120511, "epoch": 1354, "lr": 5.8916517592019296e-05}
{"train_loss": 0.23344363272190094, "global_step": 120512, "epoch": 1354, "lr": 5.89159471476041e-05}
{"train_loss": 0.13461744785308838, "global_step": 120513, "epoch": 1354, "lr": 5.8915376701990274e-05}
{"train_loss": 0.16839616000652313, "global_step": 120514, "epoch": 1354, "lr": 5.891480625517787e-05}
{"train_loss": 0.1408454179763794, "global_step": 120515, "epoch": 1354, "lr": 5.891423580716697e-05}
{"train_loss": 0.08151308447122574, "global_step": 120516, "epoch": 1354, "lr": 5.8913665357957684e-05}
{"train_loss": 0.19799773395061493, "global_step": 120517, "epoch": 1354, "lr": 5.891309490755005e-05}
{"train_loss": 0.20579062402248383, "global_step": 120518, "epoch": 1354, "lr": 5.891252445594415e-05}
{"train_loss": 0.11735128611326218, "global_step": 120519, "epoch": 1354, "lr": 5.891195400314008e-05}
{"train_loss": 0.10165955871343613, "global_step": 120520, "epoch": 1354, "lr": 5.891138354913791e-05}
{"train_loss": 0.1261167973279953, "global_step": 120521, "epoch": 1354, "lr": 5.891081309393771e-05}
{"train_loss": 0.10002259165048599, "global_step": 120522, "epoch": 1354, "lr": 5.891024263753956e-05}
{"train_loss": 0.1490817815065384, "global_step": 120523, "epoch": 1354, "lr": 5.890967217994353e-05}
{"train_loss": 0.15484344959259033, "global_step": 120524, "epoch": 1354, "lr": 5.890910172114971e-05}
{"train_loss": 0.11748362332582474, "global_step": 120525, "epoch": 1354, "lr": 5.890853126115816e-05}
{"train_loss": 0.12380866706371307, "global_step": 120526, "epoch": 1354, "lr": 5.8907960799968974e-05}
{"train_loss": 0.15184052288532257, "global_step": 120527, "epoch": 1354, "lr": 5.8907390337582213e-05}
{"train_loss": 0.09606661647558212, "global_step": 120528, "epoch": 1354, "lr": 5.890681987399798e-05}
{"train_loss": 0.0829160138964653, "global_step": 120529, "epoch": 1354, "lr": 5.890624940921632e-05}
{"train_loss": 0.14138616621494293, "global_step": 120530, "epoch": 1354, "lr": 5.890567894323733e-05}
{"train_loss": 0.1328025460243225, "global_step": 120531, "epoch": 1354, "lr": 5.890510847606108e-05}
{"train_loss": 0.1280001550912857, "global_step": 120532, "epoch": 1354, "lr": 5.8904538007687646e-05}
{"train_loss": 0.16961441934108734, "global_step": 120533, "epoch": 1354, "lr": 5.890396753811711e-05}
{"train_loss": 0.15197362005710602, "global_step": 120534, "epoch": 1354, "lr": 5.890339706734954e-05}
{"train_loss": 0.08290188014507294, "global_step": 120535, "epoch": 1354, "lr": 5.890282659538501e-05}
{"train_loss": 0.13584832847118378, "global_step": 120536, "epoch": 1354, "lr": 5.890225612222361e-05}
{"train_loss": 0.16555115580558777, "global_step": 120537, "epoch": 1354, "lr": 5.890168564786541e-05}
{"train_loss": 0.2039184868335724, "global_step": 120538, "epoch": 1354, "lr": 5.890111517231048e-05}
{"train_loss": 0.11305960267782211, "global_step": 120539, "epoch": 1354, "lr": 5.890054469555891e-05}
{"train_loss": 0.17199735343456268, "global_step": 120540, "epoch": 1354, "lr": 5.889997421761078e-05}
{"train_loss": 0.1995292603969574, "global_step": 120541, "epoch": 1354, "lr": 5.889940373846614e-05}
{"train_loss": 0.14657512307167053, "global_step": 120542, "epoch": 1354, "lr": 5.889883325812509e-05}
{"train_loss": 0.17484448850154877, "global_step": 120543, "epoch": 1354, "lr": 5.88982627765877e-05}
{"train_loss": 0.06650158017873764, "global_step": 120544, "epoch": 1354, "lr": 5.8897692293854034e-05}
{"train_loss": 0.15153737366199493, "global_step": 120545, "epoch": 1354, "lr": 5.8897121809924196e-05}
{"train_loss": 0.11563868820667267, "global_step": 120546, "epoch": 1354, "lr": 5.889655132479825e-05}
{"train_loss": 0.1593472957611084, "global_step": 120547, "epoch": 1354, "lr": 5.889598083847626e-05}
{"train_loss": 0.171588733792305, "global_step": 120548, "epoch": 1354, "lr": 5.889541035095832e-05}
{"train_loss": 0.1360575407743454, "global_step": 120549, "epoch": 1354, "lr": 5.889483986224451e-05}
{"train_loss": 0.19690126180648804, "global_step": 120550, "epoch": 1354, "lr": 5.8894269372334885e-05}
{"train_loss": 0.197379469871521, "global_step": 120551, "epoch": 1354, "lr": 5.889369888122954e-05}
{"train_loss": 0.13362543284893036, "global_step": 120552, "epoch": 1354, "lr": 5.889312838892855e-05}
{"train_loss": 0.20941779017448425, "global_step": 120553, "epoch": 1354, "lr": 5.8892557895431985e-05}
{"train_loss": 0.13921450078487396, "global_step": 120554, "epoch": 1354, "lr": 5.889198740073992e-05}
{"train_loss": 0.15311966836452484, "global_step": 120555, "epoch": 1354, "lr": 5.889141690485244e-05}
{"train_loss": 0.15686917304992676, "global_step": 120556, "epoch": 1354, "lr": 5.8890846407769604e-05}
{"train_loss": 0.05096953734755516, "global_step": 120557, "epoch": 1354, "lr": 5.889027590949152e-05}
{"train_loss": 0.13168202340602875, "global_step": 120558, "epoch": 1354, "lr": 5.888970541001824e-05}
{"train_loss": 0.17467957735061646, "global_step": 120559, "epoch": 1354, "lr": 5.8889134909349844e-05}
{"train_loss": 0.19478516280651093, "global_step": 120560, "epoch": 1354, "lr": 5.8888564407486415e-05}
{"train_loss": 0.16565845906734467, "global_step": 120561, "epoch": 1354, "lr": 5.888799390442803e-05}
{"train_loss": 0.08095482736825943, "global_step": 120562, "epoch": 1354, "lr": 5.8887423400174744e-05}
{"train_loss": 0.13222776353359222, "global_step": 120563, "epoch": 1354, "lr": 5.8886852894726664e-05}
{"train_loss": 0.12586936354637146, "global_step": 120564, "epoch": 1354, "lr": 5.8886282388083855e-05}
{"train_loss": 0.12721006572246552, "global_step": 120565, "epoch": 1354, "lr": 5.88857118802464e-05}
{"train_loss": 0.16287659108638763, "global_step": 120566, "epoch": 1354, "lr": 5.8885141371214366e-05}
{"train_loss": 0.16033530235290527, "global_step": 120567, "epoch": 1354, "lr": 5.888457086098783e-05}
{"train_loss": 0.15169309079647064, "global_step": 120568, "epoch": 1354, "lr": 5.888400034956687e-05}
{"train_loss": 0.13650362193584442, "global_step": 120569, "epoch": 1354, "lr": 5.888342983695157e-05}
{"train_loss": 0.11519316583871841, "global_step": 120570, "epoch": 1354, "lr": 5.8882859323142005e-05}
{"train_loss": 0.1691960245370865, "global_step": 120571, "epoch": 1354, "lr": 5.8882288808138256e-05}
{"train_loss": 0.11818160116672516, "global_step": 120572, "epoch": 1354, "lr": 5.8881718291940366e-05}
{"train_loss": 0.17404457926750183, "global_step": 120573, "epoch": 1354, "lr": 5.888114777454845e-05}
{"train_loss": 0.12200772762298584, "global_step": 120574, "epoch": 1354, "lr": 5.888057725596258e-05}
{"train_loss": 0.1521744728088379, "global_step": 120575, "epoch": 1354, "lr": 5.888000673618281e-05}
{"train_loss": 0.11345624923706055, "global_step": 120576, "epoch": 1354, "lr": 5.887943621520924e-05}
{"train_loss": 0.18193687498569489, "global_step": 120577, "epoch": 1354, "lr": 5.887886569304194e-05}
{"train_loss": 0.15628232061862946, "global_step": 120578, "epoch": 1354, "lr": 5.8878295169680975e-05}
{"train_loss": 0.13599635660648346, "global_step": 120579, "epoch": 1354, "lr": 5.887772464512644e-05}
{"train_loss": 0.10613732784986496, "global_step": 120580, "epoch": 1354, "lr": 5.8877154119378384e-05}
{"train_loss": 0.1721414476633072, "global_step": 120581, "epoch": 1354, "lr": 5.887658359243692e-05}
{"train_loss": 0.1784992814064026, "global_step": 120582, "epoch": 1354, "lr": 5.8876013064302105e-05}
{"train_loss": 0.1825425624847412, "global_step": 120583, "epoch": 1354, "lr": 5.8875442534974026e-05}
{"train_loss": 0.15641997754573822, "global_step": 120584, "epoch": 1354, "lr": 5.887487200445274e-05}
{"train_loss": 0.299903929233551, "global_step": 120585, "epoch": 1354, "lr": 5.887430147273833e-05}
{"train_loss": 0.17502060532569885, "global_step": 120586, "epoch": 1354, "lr": 5.8873730939830896e-05}
{"train_loss": 0.16990941762924194, "global_step": 120587, "epoch": 1354, "lr": 5.88731604057305e-05}
{"train_loss": 0.11104436218738556, "global_step": 120588, "epoch": 1354, "lr": 5.88725898704372e-05}
{"train_loss": 0.1538575291633606, "global_step": 120589, "epoch": 1354, "lr": 5.887201933395109e-05}
{"train_loss": 0.06809328496456146, "global_step": 120590, "epoch": 1354, "lr": 5.8871448796272267e-05}
{"train_loss": 0.121588796377182, "global_step": 120591, "epoch": 1354, "lr": 5.887087825740076e-05}
{"train_loss": 0.10788176953792572, "global_step": 120592, "epoch": 1354, "lr": 5.887030771733668e-05}
{"train_loss": 0.07598444074392319, "global_step": 120593, "epoch": 1354, "lr": 5.8869737176080095e-05}
{"train_loss": 0.14226974484123542, "global_step": 120594, "epoch": 1354, "lr": 5.886916663363109e-05, "val_loss": 5.3316144943237305}
{"train_loss": 0.14764046669006348, "global_step": 120595, "epoch": 1355, "lr": 5.8868596089989725e-05}
{"train_loss": 0.15770754218101501, "global_step": 120596, "epoch": 1355, "lr": 5.886802554515608e-05}
{"train_loss": 0.13284516334533691, "global_step": 120597, "epoch": 1355, "lr": 5.886745499913024e-05}
{"train_loss": 0.1344437450170517, "global_step": 120598, "epoch": 1355, "lr": 5.886688445191228e-05}
{"train_loss": 0.15949296951293945, "global_step": 120599, "epoch": 1355, "lr": 5.886631390350228e-05}
{"train_loss": 0.15481527149677277, "global_step": 120600, "epoch": 1355, "lr": 5.8865743353900305e-05}
{"train_loss": 0.09753847122192383, "global_step": 120601, "epoch": 1355, "lr": 5.886517280310644e-05}
{"train_loss": 0.12247883528470993, "global_step": 120602, "epoch": 1355, "lr": 5.886460225112076e-05}
{"train_loss": 0.15196645259857178, "global_step": 120603, "epoch": 1355, "lr": 5.886403169794335e-05}
{"train_loss": 0.1600457727909088, "global_step": 120604, "epoch": 1355, "lr": 5.886346114357427e-05}
{"train_loss": 0.11209751665592194, "global_step": 120605, "epoch": 1355, "lr": 5.886289058801362e-05}
{"train_loss": 0.1539001166820526, "global_step": 120606, "epoch": 1355, "lr": 5.8862320031261454e-05}
{"train_loss": 0.09292557835578918, "global_step": 120607, "epoch": 1355, "lr": 5.886174947331785e-05}
{"train_loss": 0.07399246841669083, "global_step": 120608, "epoch": 1355, "lr": 5.8861178914182905e-05}
{"train_loss": 0.08373294770717621, "global_step": 120609, "epoch": 1355, "lr": 5.886060835385667e-05}
{"train_loss": 0.13485342264175415, "global_step": 120610, "epoch": 1355, "lr": 5.8860037792339254e-05}
{"train_loss": 0.08994501084089279, "global_step": 120611, "epoch": 1355, "lr": 5.885946722963069e-05}
{"train_loss": 0.11428090184926987, "global_step": 120612, "epoch": 1355, "lr": 5.88588966657311e-05}
{"train_loss": 0.11947283893823624, "global_step": 120613, "epoch": 1355, "lr": 5.885832610064053e-05}
{"train_loss": 0.11570967733860016, "global_step": 120614, "epoch": 1355, "lr": 5.885775553435906e-05}
{"train_loss": 0.12498855590820312, "global_step": 120615, "epoch": 1355, "lr": 5.885718496688677e-05}
{"train_loss": 0.15347538888454437, "global_step": 120616, "epoch": 1355, "lr": 5.8856614398223744e-05}
{"train_loss": 0.19517818093299866, "global_step": 120617, "epoch": 1355, "lr": 5.8856043828370045e-05}
{"train_loss": 0.14423538744449615, "global_step": 120618, "epoch": 1355, "lr": 5.8855473257325776e-05}
{"train_loss": 0.13705787062644958, "global_step": 120619, "epoch": 1355, "lr": 5.8854902685090985e-05}
{"train_loss": 0.1323939710855484, "global_step": 120620, "epoch": 1355, "lr": 5.885433211166577e-05}
{"train_loss": 0.22973208129405975, "global_step": 120621, "epoch": 1355, "lr": 5.885376153705019e-05}
{"train_loss": 0.14778782427310944, "global_step": 120622, "epoch": 1355, "lr": 5.885319096124433e-05}
{"train_loss": 0.07109340280294418, "global_step": 120623, "epoch": 1355, "lr": 5.885262038424827e-05}
{"train_loss": 0.18318182229995728, "global_step": 120624, "epoch": 1355, "lr": 5.8852049806062085e-05}
{"train_loss": 0.17389242351055145, "global_step": 120625, "epoch": 1355, "lr": 5.8851479226685844e-05}
{"train_loss": 0.1595369577407837, "global_step": 120626, "epoch": 1355, "lr": 5.885090864611964e-05}
{"train_loss": 0.1300221085548401, "global_step": 120627, "epoch": 1355, "lr": 5.885033806436353e-05}
{"train_loss": 0.15122079849243164, "global_step": 120628, "epoch": 1355, "lr": 5.88497674814176e-05}
{"train_loss": 0.13969972729682922, "global_step": 120629, "epoch": 1355, "lr": 5.8849196897281945e-05}
{"train_loss": 0.12123188376426697, "global_step": 120630, "epoch": 1355, "lr": 5.88486263119566e-05}
{"train_loss": 0.15266607701778412, "global_step": 120631, "epoch": 1355, "lr": 5.884805572544167e-05}
{"train_loss": 0.09214165806770325, "global_step": 120632, "epoch": 1355, "lr": 5.8847485137737235e-05}
{"train_loss": 0.11185342073440552, "global_step": 120633, "epoch": 1355, "lr": 5.884691454884336e-05}
{"train_loss": 0.1118650734424591, "global_step": 120634, "epoch": 1355, "lr": 5.884634395876011e-05}
{"train_loss": 0.11000469326972961, "global_step": 120635, "epoch": 1355, "lr": 5.8845773367487586e-05}
{"train_loss": 0.16648320853710175, "global_step": 120636, "epoch": 1355, "lr": 5.884520277502585e-05}
{"train_loss": 0.11051742732524872, "global_step": 120637, "epoch": 1355, "lr": 5.8844632181375006e-05}
{"train_loss": 0.14112979173660278, "global_step": 120638, "epoch": 1355, "lr": 5.8844061586535094e-05}
{"train_loss": 0.14402234554290771, "global_step": 120639, "epoch": 1355, "lr": 5.884349099050621e-05}
{"train_loss": 0.12364377081394196, "global_step": 120640, "epoch": 1355, "lr": 5.884292039328842e-05}
{"train_loss": 0.1148235946893692, "global_step": 120641, "epoch": 1355, "lr": 5.884234979488181e-05}
{"train_loss": 0.178073450922966, "global_step": 120642, "epoch": 1355, "lr": 5.884177919528646e-05}
{"train_loss": 0.14283986389636993, "global_step": 120643, "epoch": 1355, "lr": 5.884120859450243e-05}
{"train_loss": 0.11252892017364502, "global_step": 120644, "epoch": 1355, "lr": 5.884063799252981e-05}
{"train_loss": 0.14680200815200806, "global_step": 120645, "epoch": 1355, "lr": 5.884006738936868e-05}
{"train_loss": 0.17719662189483643, "global_step": 120646, "epoch": 1355, "lr": 5.883949678501911e-05}
{"train_loss": 0.09686227142810822, "global_step": 120647, "epoch": 1355, "lr": 5.883892617948118e-05}
{"train_loss": 0.18482322990894318, "global_step": 120648, "epoch": 1355, "lr": 5.8838355572754964e-05}
{"train_loss": 0.08378979563713074, "global_step": 120649, "epoch": 1355, "lr": 5.8837784964840536e-05}
{"train_loss": 0.10468020290136337, "global_step": 120650, "epoch": 1355, "lr": 5.8837214355737966e-05}
{"train_loss": 0.122125044465065, "global_step": 120651, "epoch": 1355, "lr": 5.8836643745447354e-05}
{"train_loss": 0.138962060213089, "global_step": 120652, "epoch": 1355, "lr": 5.883607313396876e-05}
{"train_loss": 0.16614623367786407, "global_step": 120653, "epoch": 1355, "lr": 5.883550252130224e-05}
{"train_loss": 0.186594620347023, "global_step": 120654, "epoch": 1355, "lr": 5.8834931907447924e-05}
{"train_loss": 0.07876506447792053, "global_step": 120655, "epoch": 1355, "lr": 5.883436129240585e-05}
{"train_loss": 0.09075575321912766, "global_step": 120656, "epoch": 1355, "lr": 5.883379067617611e-05}
{"train_loss": 0.0524788424372673, "global_step": 120657, "epoch": 1355, "lr": 5.883322005875876e-05}
{"train_loss": 0.13965006172657013, "global_step": 120658, "epoch": 1355, "lr": 5.88326494401539e-05}
{"train_loss": 0.1456596404314041, "global_step": 120659, "epoch": 1355, "lr": 5.883207882036159e-05}
{"train_loss": 0.1465902328491211, "global_step": 120660, "epoch": 1355, "lr": 5.883150819938192e-05}
{"train_loss": 0.12083413451910019, "global_step": 120661, "epoch": 1355, "lr": 5.883093757721497e-05}
{"train_loss": 0.1651880145072937, "global_step": 120662, "epoch": 1355, "lr": 5.88303669538608e-05}
{"train_loss": 0.17069938778877258, "global_step": 120663, "epoch": 1355, "lr": 5.88297963293195e-05}
{"train_loss": 0.1403150111436844, "global_step": 120664, "epoch": 1355, "lr": 5.882922570359114e-05}
{"train_loss": 0.12842229008674622, "global_step": 120665, "epoch": 1355, "lr": 5.88286550766758e-05}
{"train_loss": 0.13613106310367584, "global_step": 120666, "epoch": 1355, "lr": 5.882808444857355e-05}
{"train_loss": 0.1410919427871704, "global_step": 120667, "epoch": 1355, "lr": 5.882751381928446e-05}
{"train_loss": 0.06849876046180725, "global_step": 120668, "epoch": 1355, "lr": 5.882694318880865e-05}
{"train_loss": 0.057677265256643295, "global_step": 120669, "epoch": 1355, "lr": 5.8826372557146146e-05}
{"train_loss": 0.11958891153335571, "global_step": 120670, "epoch": 1355, "lr": 5.882580192429704e-05}
{"train_loss": 0.15883472561836243, "global_step": 120671, "epoch": 1355, "lr": 5.8825231290261405e-05}
{"train_loss": 0.17490847408771515, "global_step": 120672, "epoch": 1355, "lr": 5.882466065503934e-05}
{"train_loss": 0.10607808083295822, "global_step": 120673, "epoch": 1355, "lr": 5.882409001863091e-05}
{"train_loss": 0.16807083785533905, "global_step": 120674, "epoch": 1355, "lr": 5.8823519381036174e-05}
{"train_loss": 0.14162255823612213, "global_step": 120675, "epoch": 1355, "lr": 5.8822948742255236e-05}
{"train_loss": 0.08921243250370026, "global_step": 120676, "epoch": 1355, "lr": 5.8822378102288154e-05}
{"train_loss": 0.14783436059951782, "global_step": 120677, "epoch": 1355, "lr": 5.8821807461135015e-05}
{"train_loss": 0.1079879105091095, "global_step": 120678, "epoch": 1355, "lr": 5.882123681879589e-05}
{"train_loss": 0.1221599355340004, "global_step": 120679, "epoch": 1355, "lr": 5.882066617527085e-05}
{"train_loss": 0.11717311292886734, "global_step": 120680, "epoch": 1355, "lr": 5.882009553055998e-05}
{"train_loss": 0.17990396916866302, "global_step": 120681, "epoch": 1355, "lr": 5.881952488466337e-05}
{"train_loss": 0.09284738451242447, "global_step": 120682, "epoch": 1355, "lr": 5.881895423758107e-05}
{"train_loss": 0.13189318713344886, "global_step": 120683, "epoch": 1355, "lr": 5.881838358931318e-05, "val_loss": 5.426088809967041, "train_action_mse_error": 23.329753875732422}
{"train_loss": 0.14295494556427002, "global_step": 120684, "epoch": 1356, "lr": 5.881781293985975e-05}
{"train_loss": 0.08326148241758347, "global_step": 120685, "epoch": 1356, "lr": 5.8817242289220875e-05}
{"train_loss": 0.16996543109416962, "global_step": 120686, "epoch": 1356, "lr": 5.881667163739664e-05}
{"train_loss": 0.15687979757785797, "global_step": 120687, "epoch": 1356, "lr": 5.881610098438711e-05}
{"train_loss": 0.13687272369861603, "global_step": 120688, "epoch": 1356, "lr": 5.881553033019236e-05}
{"train_loss": 0.24869844317436218, "global_step": 120689, "epoch": 1356, "lr": 5.8814959674812467e-05}
{"train_loss": 0.07540886849164963, "global_step": 120690, "epoch": 1356, "lr": 5.881438901824751e-05}
{"train_loss": 0.12029000371694565, "global_step": 120691, "epoch": 1356, "lr": 5.8813818360497566e-05}
{"train_loss": 0.1336347758769989, "global_step": 120692, "epoch": 1356, "lr": 5.881324770156271e-05}
{"train_loss": 0.16537947952747345, "global_step": 120693, "epoch": 1356, "lr": 5.881267704144302e-05}
{"train_loss": 0.17391934990882874, "global_step": 120694, "epoch": 1356, "lr": 5.881210638013858e-05}
{"train_loss": 0.09335290640592575, "global_step": 120695, "epoch": 1356, "lr": 5.8811535717649444e-05}
{"train_loss": 0.10300126671791077, "global_step": 120696, "epoch": 1356, "lr": 5.881096505397572e-05}
{"train_loss": 0.07934586703777313, "global_step": 120697, "epoch": 1356, "lr": 5.881039438911746e-05}
{"train_loss": 0.09986098110675812, "global_step": 120698, "epoch": 1356, "lr": 5.880982372307476e-05}
{"train_loss": 0.17010509967803955, "global_step": 120699, "epoch": 1356, "lr": 5.880925305584768e-05}
{"train_loss": 0.12515786290168762, "global_step": 120700, "epoch": 1356, "lr": 5.880868238743631e-05}
{"train_loss": 0.1177956759929657, "global_step": 120701, "epoch": 1356, "lr": 5.880811171784071e-05}
{"train_loss": 0.09480424225330353, "global_step": 120702, "epoch": 1356, "lr": 5.8807541047060976e-05}
{"train_loss": 0.11568647623062134, "global_step": 120703, "epoch": 1356, "lr": 5.880697037509717e-05}
{"train_loss": 0.12183394283056259, "global_step": 120704, "epoch": 1356, "lr": 5.880639970194937e-05}
{"train_loss": 0.1898360401391983, "global_step": 120705, "epoch": 1356, "lr": 5.880582902761766e-05}
{"train_loss": 0.17333337664604187, "global_step": 120706, "epoch": 1356, "lr": 5.880525835210213e-05}
{"train_loss": 0.1666479855775833, "global_step": 120707, "epoch": 1356, "lr": 5.8804687675402816e-05}
{"train_loss": 0.12756630778312683, "global_step": 120708, "epoch": 1356, "lr": 5.880411699751983e-05}
{"train_loss": 0.13727855682373047, "global_step": 120709, "epoch": 1356, "lr": 5.880354631845324e-05}
{"train_loss": 0.08514350652694702, "global_step": 120710, "epoch": 1356, "lr": 5.880297563820312e-05}
{"train_loss": 0.13453924655914307, "global_step": 120711, "epoch": 1356, "lr": 5.8802404956769544e-05}
{"train_loss": 0.09347120672464371, "global_step": 120712, "epoch": 1356, "lr": 5.880183427415258e-05}
{"train_loss": 0.09543427079916, "global_step": 120713, "epoch": 1356, "lr": 5.8801263590352326e-05}
{"train_loss": 0.21984297037124634, "global_step": 120714, "epoch": 1356, "lr": 5.880069290536886e-05}
{"train_loss": 0.1462010145187378, "global_step": 120715, "epoch": 1356, "lr": 5.880012221920223e-05}
{"train_loss": 0.11294747143983841, "global_step": 120716, "epoch": 1356, "lr": 5.879955153185255e-05}
{"train_loss": 0.16744065284729004, "global_step": 120717, "epoch": 1356, "lr": 5.8798980843319876e-05}
{"train_loss": 0.147208109498024, "global_step": 120718, "epoch": 1356, "lr": 5.879841015360428e-05}
{"train_loss": 0.10035146772861481, "global_step": 120719, "epoch": 1356, "lr": 5.879783946270584e-05}
{"train_loss": 0.14207710325717926, "global_step": 120720, "epoch": 1356, "lr": 5.879726877062464e-05}
{"train_loss": 0.0876893475651741, "global_step": 120721, "epoch": 1356, "lr": 5.879669807736076e-05}
{"train_loss": 0.13514944911003113, "global_step": 120722, "epoch": 1356, "lr": 5.879612738291426e-05}
{"train_loss": 0.08593998104333878, "global_step": 120723, "epoch": 1356, "lr": 5.879555668728523e-05}
{"train_loss": 0.17250104248523712, "global_step": 120724, "epoch": 1356, "lr": 5.879498599047375e-05}
{"train_loss": 0.14610686898231506, "global_step": 120725, "epoch": 1356, "lr": 5.8794415292479896e-05}
{"train_loss": 0.13942524790763855, "global_step": 120726, "epoch": 1356, "lr": 5.879384459330374e-05}
{"train_loss": 0.17019185423851013, "global_step": 120727, "epoch": 1356, "lr": 5.8793273892945356e-05}
{"train_loss": 0.13687677681446075, "global_step": 120728, "epoch": 1356, "lr": 5.879270319140482e-05}
{"train_loss": 0.11809010803699493, "global_step": 120729, "epoch": 1356, "lr": 5.8792132488682205e-05}
{"train_loss": 0.12023092061281204, "global_step": 120730, "epoch": 1356, "lr": 5.8791561784777606e-05}
{"train_loss": 0.09199768304824829, "global_step": 120731, "epoch": 1356, "lr": 5.8790991079691084e-05}
{"train_loss": 0.08717223256826401, "global_step": 120732, "epoch": 1356, "lr": 5.879042037342273e-05}
{"train_loss": 0.1274932622909546, "global_step": 120733, "epoch": 1356, "lr": 5.87898496659726e-05}
{"train_loss": 0.1657382845878601, "global_step": 120734, "epoch": 1356, "lr": 5.878927895734079e-05}
{"train_loss": 0.11048534512519836, "global_step": 120735, "epoch": 1356, "lr": 5.878870824752736e-05}
{"train_loss": 0.1549217253923416, "global_step": 120736, "epoch": 1356, "lr": 5.8788137536532395e-05}
{"train_loss": 0.13141684234142303, "global_step": 120737, "epoch": 1356, "lr": 5.878756682435598e-05}
{"train_loss": 0.16857747733592987, "global_step": 120738, "epoch": 1356, "lr": 5.878699611099818e-05}
{"train_loss": 0.09546684473752975, "global_step": 120739, "epoch": 1356, "lr": 5.8786425396459075e-05}
{"train_loss": 0.18529966473579407, "global_step": 120740, "epoch": 1356, "lr": 5.878585468073874e-05}
{"train_loss": 0.15732435882091522, "global_step": 120741, "epoch": 1356, "lr": 5.8785283963837254e-05}
{"train_loss": 0.103240467607975, "global_step": 120742, "epoch": 1356, "lr": 5.8784713245754706e-05}
{"train_loss": 0.13551533222198486, "global_step": 120743, "epoch": 1356, "lr": 5.878414252649115e-05}
{"train_loss": 0.133962020277977, "global_step": 120744, "epoch": 1356, "lr": 5.878357180604668e-05}
{"train_loss": 0.1580476462841034, "global_step": 120745, "epoch": 1356, "lr": 5.8783001084421354e-05}
{"train_loss": 0.09368797391653061, "global_step": 120746, "epoch": 1356, "lr": 5.878243036161526e-05}
{"train_loss": 0.13190346956253052, "global_step": 120747, "epoch": 1356, "lr": 5.878185963762849e-05}
{"train_loss": 0.1429443359375, "global_step": 120748, "epoch": 1356, "lr": 5.87812889124611e-05}
{"train_loss": 0.13125738501548767, "global_step": 120749, "epoch": 1356, "lr": 5.878071818611316e-05}
{"train_loss": 0.14014793932437897, "global_step": 120750, "epoch": 1356, "lr": 5.8780147458584774e-05}
{"train_loss": 0.08453727513551712, "global_step": 120751, "epoch": 1356, "lr": 5.8779576729875995e-05}
{"train_loss": 0.12691718339920044, "global_step": 120752, "epoch": 1356, "lr": 5.877900599998691e-05}
{"train_loss": 0.15611553192138672, "global_step": 120753, "epoch": 1356, "lr": 5.877843526891761e-05}
{"train_loss": 0.18091221153736115, "global_step": 120754, "epoch": 1356, "lr": 5.8777864536668136e-05}
{"train_loss": 0.13720563054084778, "global_step": 120755, "epoch": 1356, "lr": 5.87772938032386e-05}
{"train_loss": 0.10699313879013062, "global_step": 120756, "epoch": 1356, "lr": 5.877672306862906e-05}
{"train_loss": 0.16960880160331726, "global_step": 120757, "epoch": 1356, "lr": 5.8776152332839594e-05}
{"train_loss": 0.0995604544878006, "global_step": 120758, "epoch": 1356, "lr": 5.877558159587028e-05}
{"train_loss": 0.18696270883083344, "global_step": 120759, "epoch": 1356, "lr": 5.87750108577212e-05}
{"train_loss": 0.13135987520217896, "global_step": 120760, "epoch": 1356, "lr": 5.8774440118392425e-05}
{"train_loss": 0.08970695734024048, "global_step": 120761, "epoch": 1356, "lr": 5.877386937788403e-05}
{"train_loss": 0.13565655052661896, "global_step": 120762, "epoch": 1356, "lr": 5.8773298636196096e-05}
{"train_loss": 0.15876449644565582, "global_step": 120763, "epoch": 1356, "lr": 5.8772727893328706e-05}
{"train_loss": 0.211420938372612, "global_step": 120764, "epoch": 1356, "lr": 5.877215714928193e-05}
{"train_loss": 0.14095854759216309, "global_step": 120765, "epoch": 1356, "lr": 5.877158640405583e-05}
{"train_loss": 0.14705659449100494, "global_step": 120766, "epoch": 1356, "lr": 5.877101565765051e-05}
{"train_loss": 0.15961015224456787, "global_step": 120767, "epoch": 1356, "lr": 5.8770444910066025e-05}
{"train_loss": 0.19091863930225372, "global_step": 120768, "epoch": 1356, "lr": 5.876987416130246e-05}
{"train_loss": 0.136823832988739, "global_step": 120769, "epoch": 1356, "lr": 5.876930341135991e-05}
{"train_loss": 0.14073580503463745, "global_step": 120770, "epoch": 1356, "lr": 5.876873266023841e-05}
{"train_loss": 0.10252659767866135, "global_step": 120771, "epoch": 1356, "lr": 5.8768161907938076e-05}
{"train_loss": 0.1351255296488826, "global_step": 120772, "epoch": 1356, "lr": 5.876759115445897e-05, "val_loss": 5.412673473358154}
{"train_loss": 0.1659068614244461, "global_step": 120773, "epoch": 1357, "lr": 5.876702039980117e-05}
{"train_loss": 0.20480872690677643, "global_step": 120774, "epoch": 1357, "lr": 5.876644964396474e-05}
{"train_loss": 0.11228777468204498, "global_step": 120775, "epoch": 1357, "lr": 5.8765878886949774e-05}
{"train_loss": 0.17901413142681122, "global_step": 120776, "epoch": 1357, "lr": 5.8765308128756346e-05}
{"train_loss": 0.12782622873783112, "global_step": 120777, "epoch": 1357, "lr": 5.876473736938453e-05}
{"train_loss": 0.26522278785705566, "global_step": 120778, "epoch": 1357, "lr": 5.876416660883439e-05}
{"train_loss": 0.1270724982023239, "global_step": 120779, "epoch": 1357, "lr": 5.876359584710602e-05}
{"train_loss": 0.11328580975532532, "global_step": 120780, "epoch": 1357, "lr": 5.876302508419949e-05}
{"train_loss": 0.1548028290271759, "global_step": 120781, "epoch": 1357, "lr": 5.876245432011488e-05}
{"train_loss": 0.19641296565532684, "global_step": 120782, "epoch": 1357, "lr": 5.876188355485227e-05}
{"train_loss": 0.20966339111328125, "global_step": 120783, "epoch": 1357, "lr": 5.8761312788411724e-05}
{"train_loss": 0.08776705712080002, "global_step": 120784, "epoch": 1357, "lr": 5.8760742020793334e-05}
{"train_loss": 0.12248718738555908, "global_step": 120785, "epoch": 1357, "lr": 5.876017125199717e-05}
{"train_loss": 0.18344362080097198, "global_step": 120786, "epoch": 1357, "lr": 5.8759600482023295e-05}
{"train_loss": 0.0915413498878479, "global_step": 120787, "epoch": 1357, "lr": 5.87590297108718e-05}
{"train_loss": 0.10806147754192352, "global_step": 120788, "epoch": 1357, "lr": 5.875845893854277e-05}
{"train_loss": 0.08111944794654846, "global_step": 120789, "epoch": 1357, "lr": 5.875788816503627e-05}
{"train_loss": 0.1988789588212967, "global_step": 120790, "epoch": 1357, "lr": 5.8757317390352376e-05}
{"train_loss": 0.10952111333608627, "global_step": 120791, "epoch": 1357, "lr": 5.875674661449116e-05}
{"train_loss": 0.08996224403381348, "global_step": 120792, "epoch": 1357, "lr": 5.875617583745272e-05}
{"train_loss": 0.1375056356191635, "global_step": 120793, "epoch": 1357, "lr": 5.875560505923712e-05}
{"train_loss": 0.2271854728460312, "global_step": 120794, "epoch": 1357, "lr": 5.875503427984442e-05}
{"train_loss": 0.12183007597923279, "global_step": 120795, "epoch": 1357, "lr": 5.875446349927473e-05}
{"train_loss": 0.14897440373897552, "global_step": 120796, "epoch": 1357, "lr": 5.875389271752809e-05}
{"train_loss": 0.12590919435024261, "global_step": 120797, "epoch": 1357, "lr": 5.875332193460461e-05}
{"train_loss": 0.11763568222522736, "global_step": 120798, "epoch": 1357, "lr": 5.875275115050435e-05}
{"train_loss": 0.09564574807882309, "global_step": 120799, "epoch": 1357, "lr": 5.875218036522738e-05}
{"train_loss": 0.1649399846792221, "global_step": 120800, "epoch": 1357, "lr": 5.875160957877379e-05}
{"train_loss": 0.15476560592651367, "global_step": 120801, "epoch": 1357, "lr": 5.8751038791143665e-05}
{"train_loss": 0.1520889699459076, "global_step": 120802, "epoch": 1357, "lr": 5.875046800233706e-05}
{"train_loss": 0.08768763393163681, "global_step": 120803, "epoch": 1357, "lr": 5.874989721235406e-05}
{"train_loss": 0.20083707571029663, "global_step": 120804, "epoch": 1357, "lr": 5.874932642119475e-05}
{"train_loss": 0.12524601817131042, "global_step": 120805, "epoch": 1357, "lr": 5.874875562885919e-05}
{"train_loss": 0.10246143490076065, "global_step": 120806, "epoch": 1357, "lr": 5.8748184835347474e-05}
{"train_loss": 0.1803128868341446, "global_step": 120807, "epoch": 1357, "lr": 5.874761404065967e-05}
{"train_loss": 0.10465560853481293, "global_step": 120808, "epoch": 1357, "lr": 5.874704324479585e-05}
{"train_loss": 0.10909996181726456, "global_step": 120809, "epoch": 1357, "lr": 5.8746472447756104e-05}
{"train_loss": 0.1197635754942894, "global_step": 120810, "epoch": 1357, "lr": 5.874590164954049e-05}
{"train_loss": 0.1399819403886795, "global_step": 120811, "epoch": 1357, "lr": 5.8745330850149104e-05}
{"train_loss": 0.05899139121174812, "global_step": 120812, "epoch": 1357, "lr": 5.874476004958203e-05}
{"train_loss": 0.09362199902534485, "global_step": 120813, "epoch": 1357, "lr": 5.874418924783931e-05}
{"train_loss": 0.15245771408081055, "global_step": 120814, "epoch": 1357, "lr": 5.8743618444921044e-05}
{"train_loss": 0.17812077701091766, "global_step": 120815, "epoch": 1357, "lr": 5.87430476408273e-05}
{"train_loss": 0.11188234388828278, "global_step": 120816, "epoch": 1357, "lr": 5.874247683555816e-05}
{"train_loss": 0.10878191143274307, "global_step": 120817, "epoch": 1357, "lr": 5.874190602911371e-05}
{"train_loss": 0.1904991716146469, "global_step": 120818, "epoch": 1357, "lr": 5.874133522149402e-05}
{"train_loss": 0.07675253599882126, "global_step": 120819, "epoch": 1357, "lr": 5.874076441269916e-05}
{"train_loss": 0.17553184926509857, "global_step": 120820, "epoch": 1357, "lr": 5.874019360272921e-05}
{"train_loss": 0.11830524355173111, "global_step": 120821, "epoch": 1357, "lr": 5.873962279158425e-05}
{"train_loss": 0.12409593909978867, "global_step": 120822, "epoch": 1357, "lr": 5.8739051979264346e-05}
{"train_loss": 0.1286056786775589, "global_step": 120823, "epoch": 1357, "lr": 5.873848116576959e-05}
{"train_loss": 0.12987379729747772, "global_step": 120824, "epoch": 1357, "lr": 5.873791035110004e-05}
{"train_loss": 0.08028544485569, "global_step": 120825, "epoch": 1357, "lr": 5.87373395352558e-05}
{"train_loss": 0.0770668014883995, "global_step": 120826, "epoch": 1357, "lr": 5.873676871823692e-05}
{"train_loss": 0.08300923556089401, "global_step": 120827, "epoch": 1357, "lr": 5.87361979000435e-05}
{"train_loss": 0.1269456446170807, "global_step": 120828, "epoch": 1357, "lr": 5.873562708067559e-05}
{"train_loss": 0.10341612249612808, "global_step": 120829, "epoch": 1357, "lr": 5.873505626013329e-05}
{"train_loss": 0.18277499079704285, "global_step": 120830, "epoch": 1357, "lr": 5.873448543841666e-05}
{"train_loss": 0.10938664525747299, "global_step": 120831, "epoch": 1357, "lr": 5.87339146155258e-05}
{"train_loss": 0.10970789939165115, "global_step": 120832, "epoch": 1357, "lr": 5.873334379146076e-05}
{"train_loss": 0.15608759224414825, "global_step": 120833, "epoch": 1357, "lr": 5.873277296622164e-05}
{"train_loss": 0.19228516519069672, "global_step": 120834, "epoch": 1357, "lr": 5.873220213980848e-05}
{"train_loss": 0.14837929606437683, "global_step": 120835, "epoch": 1357, "lr": 5.87316313122214e-05}
{"train_loss": 0.15084096789360046, "global_step": 120836, "epoch": 1357, "lr": 5.8731060483460455e-05}
{"train_loss": 0.1452982872724533, "global_step": 120837, "epoch": 1357, "lr": 5.873048965352574e-05}
{"train_loss": 0.17873400449752808, "global_step": 120838, "epoch": 1357, "lr": 5.872991882241731e-05}
{"train_loss": 0.09044166654348373, "global_step": 120839, "epoch": 1357, "lr": 5.872934799013524e-05}
{"train_loss": 0.1423603594303131, "global_step": 120840, "epoch": 1357, "lr": 5.872877715667962e-05}
{"train_loss": 0.11350278556346893, "global_step": 120841, "epoch": 1357, "lr": 5.872820632205051e-05}
{"train_loss": 0.18536575138568878, "global_step": 120842, "epoch": 1357, "lr": 5.8727635486248024e-05}
{"train_loss": 0.23177467286586761, "global_step": 120843, "epoch": 1357, "lr": 5.87270646492722e-05}
{"train_loss": 0.1090661883354187, "global_step": 120844, "epoch": 1357, "lr": 5.872649381112313e-05}
{"train_loss": 0.13571946322917938, "global_step": 120845, "epoch": 1357, "lr": 5.872592297180089e-05}
{"train_loss": 0.11972961574792862, "global_step": 120846, "epoch": 1357, "lr": 5.8725352131305544e-05}
{"train_loss": 0.14691494405269623, "global_step": 120847, "epoch": 1357, "lr": 5.8724781289637196e-05}
{"train_loss": 0.17557752132415771, "global_step": 120848, "epoch": 1357, "lr": 5.8724210446795904e-05}
{"train_loss": 0.10444856435060501, "global_step": 120849, "epoch": 1357, "lr": 5.872363960278174e-05}
{"train_loss": 0.15486940741539001, "global_step": 120850, "epoch": 1357, "lr": 5.87230687575948e-05}
{"train_loss": 0.0889817401766777, "global_step": 120851, "epoch": 1357, "lr": 5.872249791123514e-05}
{"train_loss": 0.15280300378799438, "global_step": 120852, "epoch": 1357, "lr": 5.872192706370284e-05}
{"train_loss": 0.1540050506591797, "global_step": 120853, "epoch": 1357, "lr": 5.8721356214998005e-05}
{"train_loss": 0.11581636220216751, "global_step": 120854, "epoch": 1357, "lr": 5.872078536512068e-05}
{"train_loss": 0.1006835326552391, "global_step": 120855, "epoch": 1357, "lr": 5.872021451407095e-05}
{"train_loss": 0.14089427888393402, "global_step": 120856, "epoch": 1357, "lr": 5.87196436618489e-05}
{"train_loss": 0.11441324651241302, "global_step": 120857, "epoch": 1357, "lr": 5.87190728084546e-05}
{"train_loss": 0.1340949386358261, "global_step": 120858, "epoch": 1357, "lr": 5.871850195388812e-05}
{"train_loss": 0.1794644296169281, "global_step": 120859, "epoch": 1357, "lr": 5.871793109814955e-05}
{"train_loss": 0.12509861588478088, "global_step": 120860, "epoch": 1357, "lr": 5.871736024123895e-05}
{"train_loss": 0.13726450172200633, "global_step": 120861, "epoch": 1357, "lr": 5.8716789383156414e-05, "val_loss": 5.280655860900879}
{"train_loss": 0.15120016038417816, "global_step": 120862, "epoch": 1358, "lr": 5.871621852390201e-05}
{"train_loss": 0.1926005482673645, "global_step": 120863, "epoch": 1358, "lr": 5.8715647663475825e-05}
{"train_loss": 0.13005034625530243, "global_step": 120864, "epoch": 1358, "lr": 5.871507680187791e-05}
{"train_loss": 0.1338859498500824, "global_step": 120865, "epoch": 1358, "lr": 5.871450593910838e-05}
{"train_loss": 0.19075599312782288, "global_step": 120866, "epoch": 1358, "lr": 5.871393507516728e-05}
{"train_loss": 0.1273414045572281, "global_step": 120867, "epoch": 1358, "lr": 5.871336421005469e-05}
{"train_loss": 0.08901748806238174, "global_step": 120868, "epoch": 1358, "lr": 5.871279334377071e-05}
{"train_loss": 0.14745740592479706, "global_step": 120869, "epoch": 1358, "lr": 5.8712222476315395e-05}
{"train_loss": 0.08619150519371033, "global_step": 120870, "epoch": 1358, "lr": 5.8711651607688814e-05}
{"train_loss": 0.1301375776529312, "global_step": 120871, "epoch": 1358, "lr": 5.8711080737891064e-05}
{"train_loss": 0.1291275918483734, "global_step": 120872, "epoch": 1358, "lr": 5.871050986692223e-05}
{"train_loss": 0.14699652791023254, "global_step": 120873, "epoch": 1358, "lr": 5.8709938994782364e-05}
{"train_loss": 0.11502503603696823, "global_step": 120874, "epoch": 1358, "lr": 5.8709368121471555e-05}
{"train_loss": 0.15110793709754944, "global_step": 120875, "epoch": 1358, "lr": 5.870879724698989e-05}
{"train_loss": 0.1427629441022873, "global_step": 120876, "epoch": 1358, "lr": 5.8708226371337414e-05}
{"train_loss": 0.160787895321846, "global_step": 120877, "epoch": 1358, "lr": 5.8707655494514224e-05}
{"train_loss": 0.13317637145519257, "global_step": 120878, "epoch": 1358, "lr": 5.87070846165204e-05}
{"train_loss": 0.08484271168708801, "global_step": 120879, "epoch": 1358, "lr": 5.870651373735603e-05}
{"train_loss": 0.14390979707241058, "global_step": 120880, "epoch": 1358, "lr": 5.870594285702116e-05}
{"train_loss": 0.18622265756130219, "global_step": 120881, "epoch": 1358, "lr": 5.8705371975515874e-05}
{"train_loss": 0.1388874500989914, "global_step": 120882, "epoch": 1358, "lr": 5.870480109284027e-05}
{"train_loss": 0.13712121546268463, "global_step": 120883, "epoch": 1358, "lr": 5.8704230208994404e-05}
{"train_loss": 0.11979775130748749, "global_step": 120884, "epoch": 1358, "lr": 5.8703659323978365e-05}
{"train_loss": 0.16971519589424133, "global_step": 120885, "epoch": 1358, "lr": 5.870308843779223e-05}
{"train_loss": 0.08112990856170654, "global_step": 120886, "epoch": 1358, "lr": 5.870251755043606e-05}
{"train_loss": 0.17725712060928345, "global_step": 120887, "epoch": 1358, "lr": 5.8701946661909946e-05}
{"train_loss": 0.08562926203012466, "global_step": 120888, "epoch": 1358, "lr": 5.870137577221395e-05}
{"train_loss": 0.12065035849809647, "global_step": 120889, "epoch": 1358, "lr": 5.870080488134818e-05}
{"train_loss": 0.1287384033203125, "global_step": 120890, "epoch": 1358, "lr": 5.870023398931268e-05}
{"train_loss": 0.15650087594985962, "global_step": 120891, "epoch": 1358, "lr": 5.869966309610755e-05}
{"train_loss": 0.13074226677417755, "global_step": 120892, "epoch": 1358, "lr": 5.869909220173285e-05}
{"train_loss": 0.056406158953905106, "global_step": 120893, "epoch": 1358, "lr": 5.869852130618867e-05}
{"train_loss": 0.10231808573007584, "global_step": 120894, "epoch": 1358, "lr": 5.8697950409475076e-05}
{"train_loss": 0.13089627027511597, "global_step": 120895, "epoch": 1358, "lr": 5.869737951159215e-05}
{"train_loss": 0.0783759355545044, "global_step": 120896, "epoch": 1358, "lr": 5.8696808612539965e-05}
{"train_loss": 0.07071524858474731, "global_step": 120897, "epoch": 1358, "lr": 5.86962377123186e-05}
{"train_loss": 0.09066768735647202, "global_step": 120898, "epoch": 1358, "lr": 5.869566681092813e-05}
{"train_loss": 0.07451261579990387, "global_step": 120899, "epoch": 1358, "lr": 5.8695095908368635e-05}
{"train_loss": 0.10925213992595673, "global_step": 120900, "epoch": 1358, "lr": 5.86945250046402e-05}
{"train_loss": 0.14358501136302948, "global_step": 120901, "epoch": 1358, "lr": 5.869395409974288e-05}
{"train_loss": 0.17696735262870789, "global_step": 120902, "epoch": 1358, "lr": 5.869338319367676e-05}
{"train_loss": 0.11977067589759827, "global_step": 120903, "epoch": 1358, "lr": 5.869281228644193e-05}
{"train_loss": 0.09523408114910126, "global_step": 120904, "epoch": 1358, "lr": 5.869224137803846e-05}
{"train_loss": 0.12553922832012177, "global_step": 120905, "epoch": 1358, "lr": 5.869167046846642e-05}
{"train_loss": 0.11722734570503235, "global_step": 120906, "epoch": 1358, "lr": 5.8691099557725885e-05}
{"train_loss": 0.10600539296865463, "global_step": 120907, "epoch": 1358, "lr": 5.8690528645816937e-05}
{"train_loss": 0.1511951982975006, "global_step": 120908, "epoch": 1358, "lr": 5.868995773273965e-05}
{"train_loss": 0.11990507692098618, "global_step": 120909, "epoch": 1358, "lr": 5.868938681849412e-05}
{"train_loss": 0.11892063170671463, "global_step": 120910, "epoch": 1358, "lr": 5.86888159030804e-05}
{"train_loss": 0.17756180465221405, "global_step": 120911, "epoch": 1358, "lr": 5.868824498649857e-05}
{"train_loss": 0.1319384127855301, "global_step": 120912, "epoch": 1358, "lr": 5.868767406874871e-05}
{"train_loss": 0.07794458419084549, "global_step": 120913, "epoch": 1358, "lr": 5.868710314983091e-05}
{"train_loss": 0.10698172450065613, "global_step": 120914, "epoch": 1358, "lr": 5.868653222974524e-05}
{"train_loss": 0.08437082171440125, "global_step": 120915, "epoch": 1358, "lr": 5.868596130849176e-05}
{"train_loss": 0.12217026948928833, "global_step": 120916, "epoch": 1358, "lr": 5.868539038607056e-05}
{"train_loss": 0.11102386564016342, "global_step": 120917, "epoch": 1358, "lr": 5.868481946248171e-05}
{"train_loss": 0.13312487304210663, "global_step": 120918, "epoch": 1358, "lr": 5.8684248537725295e-05}
{"train_loss": 0.16931864619255066, "global_step": 120919, "epoch": 1358, "lr": 5.8683677611801394e-05}
{"train_loss": 0.0914105474948883, "global_step": 120920, "epoch": 1358, "lr": 5.868310668471008e-05}
{"train_loss": 0.17947126924991608, "global_step": 120921, "epoch": 1358, "lr": 5.8682535756451415e-05}
{"train_loss": 0.17126406729221344, "global_step": 120922, "epoch": 1358, "lr": 5.8681964827025493e-05}
{"train_loss": 0.16503643989562988, "global_step": 120923, "epoch": 1358, "lr": 5.868139389643239e-05}
{"train_loss": 0.12567558884620667, "global_step": 120924, "epoch": 1358, "lr": 5.8680822964672186e-05}
{"train_loss": 0.1452578753232956, "global_step": 120925, "epoch": 1358, "lr": 5.868025203174493e-05}
{"train_loss": 0.12969107925891876, "global_step": 120926, "epoch": 1358, "lr": 5.8679681097650744e-05}
{"train_loss": 0.14898203313350677, "global_step": 120927, "epoch": 1358, "lr": 5.867911016238966e-05}
{"train_loss": 0.08340627700090408, "global_step": 120928, "epoch": 1358, "lr": 5.8678539225961784e-05}
{"train_loss": 0.11175186932086945, "global_step": 120929, "epoch": 1358, "lr": 5.867796828836719e-05}
{"train_loss": 0.11740729957818985, "global_step": 120930, "epoch": 1358, "lr": 5.8677397349605944e-05}
{"train_loss": 0.2104375660419464, "global_step": 120931, "epoch": 1358, "lr": 5.867682640967813e-05}
{"train_loss": 0.10393281280994415, "global_step": 120932, "epoch": 1358, "lr": 5.867625546858382e-05}
{"train_loss": 0.12603527307510376, "global_step": 120933, "epoch": 1358, "lr": 5.8675684526323106e-05}
{"train_loss": 0.06945188343524933, "global_step": 120934, "epoch": 1358, "lr": 5.867511358289605e-05}
{"train_loss": 0.13835692405700684, "global_step": 120935, "epoch": 1358, "lr": 5.867454263830271e-05}
{"train_loss": 0.13989712297916412, "global_step": 120936, "epoch": 1358, "lr": 5.867397169254319e-05}
{"train_loss": 0.1426774263381958, "global_step": 120937, "epoch": 1358, "lr": 5.867340074561758e-05}
{"train_loss": 0.2130012959241867, "global_step": 120938, "epoch": 1358, "lr": 5.8672829797525916e-05}
{"train_loss": 0.1756494641304016, "global_step": 120939, "epoch": 1358, "lr": 5.86722588482683e-05}
{"train_loss": 0.1685594767332077, "global_step": 120940, "epoch": 1358, "lr": 5.867168789784481e-05}
{"train_loss": 0.1360408514738083, "global_step": 120941, "epoch": 1358, "lr": 5.867111694625551e-05}
{"train_loss": 0.185677170753479, "global_step": 120942, "epoch": 1358, "lr": 5.867054599350049e-05}
{"train_loss": 0.14619667828083038, "global_step": 120943, "epoch": 1358, "lr": 5.866997503957981e-05}
{"train_loss": 0.17088791728019714, "global_step": 120944, "epoch": 1358, "lr": 5.8669404084493565e-05}
{"train_loss": 0.1953110247850418, "global_step": 120945, "epoch": 1358, "lr": 5.866883312824183e-05}
{"train_loss": 0.19738255441188812, "global_step": 120946, "epoch": 1358, "lr": 5.866826217082466e-05}
{"train_loss": 0.11910101026296616, "global_step": 120947, "epoch": 1358, "lr": 5.866769121224216e-05}
{"train_loss": 0.172551229596138, "global_step": 120948, "epoch": 1358, "lr": 5.8667120252494403e-05}
{"train_loss": 0.1447008103132248, "global_step": 120949, "epoch": 1358, "lr": 5.866654929158144e-05}
{"train_loss": 0.1330502282283949, "global_step": 120950, "epoch": 1358, "lr": 5.866597832950338e-05, "val_loss": 5.417006969451904}
{"train_loss": 0.10526340454816818, "global_step": 120951, "epoch": 1359, "lr": 5.8665407366260274e-05}
{"train_loss": 0.1128026619553566, "global_step": 120952, "epoch": 1359, "lr": 5.866483640185221e-05}
{"train_loss": 0.11272621899843216, "global_step": 120953, "epoch": 1359, "lr": 5.866426543627927e-05}
{"train_loss": 0.13669271767139435, "global_step": 120954, "epoch": 1359, "lr": 5.866369446954153e-05}
{"train_loss": 0.12542718648910522, "global_step": 120955, "epoch": 1359, "lr": 5.866312350163906e-05}
{"train_loss": 0.13344526290893555, "global_step": 120956, "epoch": 1359, "lr": 5.866255253257194e-05}
{"train_loss": 0.08560197055339813, "global_step": 120957, "epoch": 1359, "lr": 5.866198156234023e-05}
{"train_loss": 0.1596030592918396, "global_step": 120958, "epoch": 1359, "lr": 5.866141059094403e-05}
{"train_loss": 0.11908530443906784, "global_step": 120959, "epoch": 1359, "lr": 5.866083961838341e-05}
{"train_loss": 0.09670382738113403, "global_step": 120960, "epoch": 1359, "lr": 5.8660268644658444e-05}
{"train_loss": 0.13357481360435486, "global_step": 120961, "epoch": 1359, "lr": 5.865969766976921e-05}
{"train_loss": 0.0835253968834877, "global_step": 120962, "epoch": 1359, "lr": 5.865912669371578e-05}
{"train_loss": 0.17988023161888123, "global_step": 120963, "epoch": 1359, "lr": 5.865855571649824e-05}
{"train_loss": 0.19787172973155975, "global_step": 120964, "epoch": 1359, "lr": 5.865798473811667e-05}
{"train_loss": 0.11570291221141815, "global_step": 120965, "epoch": 1359, "lr": 5.865741375857113e-05}
{"train_loss": 0.15080563724040985, "global_step": 120966, "epoch": 1359, "lr": 5.865684277786172e-05}
{"train_loss": 0.12705066800117493, "global_step": 120967, "epoch": 1359, "lr": 5.8656271795988496e-05}
{"train_loss": 0.16669315099716187, "global_step": 120968, "epoch": 1359, "lr": 5.865570081295154e-05}
{"train_loss": 0.12022212147712708, "global_step": 120969, "epoch": 1359, "lr": 5.865512982875093e-05}
{"train_loss": 0.10141736268997192, "global_step": 120970, "epoch": 1359, "lr": 5.865455884338674e-05}
{"train_loss": 0.1615331768989563, "global_step": 120971, "epoch": 1359, "lr": 5.865398785685906e-05}
{"train_loss": 0.10641366988420486, "global_step": 120972, "epoch": 1359, "lr": 5.865341686916795e-05}
{"train_loss": 0.16890741884708405, "global_step": 120973, "epoch": 1359, "lr": 5.86528458803135e-05}
{"train_loss": 0.18115706741809845, "global_step": 120974, "epoch": 1359, "lr": 5.8652274890295775e-05}
{"train_loss": 0.15380823612213135, "global_step": 120975, "epoch": 1359, "lr": 5.865170389911486e-05}
{"train_loss": 0.16106463968753815, "global_step": 120976, "epoch": 1359, "lr": 5.8651132906770826e-05}
{"train_loss": 0.11362336575984955, "global_step": 120977, "epoch": 1359, "lr": 5.8650561913263745e-05}
{"train_loss": 0.18630297482013702, "global_step": 120978, "epoch": 1359, "lr": 5.8649990918593715e-05}
{"train_loss": 0.18672414124011993, "global_step": 120979, "epoch": 1359, "lr": 5.864941992276078e-05}
{"train_loss": 0.10560053586959839, "global_step": 120980, "epoch": 1359, "lr": 5.864884892576506e-05}
{"train_loss": 0.13963116705417633, "global_step": 120981, "epoch": 1359, "lr": 5.8648277927606586e-05}
{"train_loss": 0.16960151493549347, "global_step": 120982, "epoch": 1359, "lr": 5.864770692828547e-05}
{"train_loss": 0.13533423840999603, "global_step": 120983, "epoch": 1359, "lr": 5.864713592780178e-05}
{"train_loss": 0.17491315305233002, "global_step": 120984, "epoch": 1359, "lr": 5.8646564926155565e-05}
{"train_loss": 0.22836142778396606, "global_step": 120985, "epoch": 1359, "lr": 5.8645993923346945e-05}
{"train_loss": 0.13894739747047424, "global_step": 120986, "epoch": 1359, "lr": 5.864542291937597e-05}
{"train_loss": 0.1524960994720459, "global_step": 120987, "epoch": 1359, "lr": 5.8644851914242725e-05}
{"train_loss": 0.11837723106145859, "global_step": 120988, "epoch": 1359, "lr": 5.864428090794729e-05}
{"train_loss": 0.10221993178129196, "global_step": 120989, "epoch": 1359, "lr": 5.864370990048973e-05}
{"train_loss": 0.16161255538463593, "global_step": 120990, "epoch": 1359, "lr": 5.864313889187013e-05}
{"train_loss": 0.10933870077133179, "global_step": 120991, "epoch": 1359, "lr": 5.8642567882088574e-05}
{"train_loss": 0.15755963325500488, "global_step": 120992, "epoch": 1359, "lr": 5.8641996871145123e-05}
{"train_loss": 0.20725204050540924, "global_step": 120993, "epoch": 1359, "lr": 5.8641425859039855e-05}
{"train_loss": 0.1260208934545517, "global_step": 120994, "epoch": 1359, "lr": 5.8640854845772864e-05}
{"train_loss": 0.21131499111652374, "global_step": 120995, "epoch": 1359, "lr": 5.864028383134421e-05}
{"train_loss": 0.12161705642938614, "global_step": 120996, "epoch": 1359, "lr": 5.8639712815753965e-05}
{"train_loss": 0.14038673043251038, "global_step": 120997, "epoch": 1359, "lr": 5.863914179900222e-05}
{"train_loss": 0.090591199696064, "global_step": 120998, "epoch": 1359, "lr": 5.863857078108905e-05}
{"train_loss": 0.17029888927936554, "global_step": 120999, "epoch": 1359, "lr": 5.863799976201453e-05}
{"train_loss": 0.18227440118789673, "global_step": 121000, "epoch": 1359, "lr": 5.863742874177874e-05}
{"train_loss": 0.18700970709323883, "global_step": 121001, "epoch": 1359, "lr": 5.863685772038175e-05}
{"train_loss": 0.13661228120326996, "global_step": 121002, "epoch": 1359, "lr": 5.8636286697823636e-05}
{"train_loss": 0.09463847428560257, "global_step": 121003, "epoch": 1359, "lr": 5.863571567410448e-05}
{"train_loss": 0.09450072050094604, "global_step": 121004, "epoch": 1359, "lr": 5.863514464922437e-05}
{"train_loss": 0.17318202555179596, "global_step": 121005, "epoch": 1359, "lr": 5.8634573623183365e-05}
{"train_loss": 0.13817323744297028, "global_step": 121006, "epoch": 1359, "lr": 5.863400259598154e-05}
{"train_loss": 0.1153271496295929, "global_step": 121007, "epoch": 1359, "lr": 5.863343156761898e-05}
{"train_loss": 0.09758300334215164, "global_step": 121008, "epoch": 1359, "lr": 5.8632860538095755e-05}
{"train_loss": 0.17343828082084656, "global_step": 121009, "epoch": 1359, "lr": 5.863228950741195e-05}
{"train_loss": 0.19862109422683716, "global_step": 121010, "epoch": 1359, "lr": 5.863171847556766e-05}
{"train_loss": 0.13658809661865234, "global_step": 121011, "epoch": 1359, "lr": 5.863114744256293e-05}
{"train_loss": 0.15189163386821747, "global_step": 121012, "epoch": 1359, "lr": 5.863057640839783e-05}
{"train_loss": 0.10721779614686966, "global_step": 121013, "epoch": 1359, "lr": 5.863000537307246e-05}
{"train_loss": 0.15218591690063477, "global_step": 121014, "epoch": 1359, "lr": 5.8629434336586896e-05}
{"train_loss": 0.21120283007621765, "global_step": 121015, "epoch": 1359, "lr": 5.8628863298941206e-05}
{"train_loss": 0.14503918588161469, "global_step": 121016, "epoch": 1359, "lr": 5.8628292260135474e-05}
{"train_loss": 0.15764161944389343, "global_step": 121017, "epoch": 1359, "lr": 5.8627721220169774e-05}
{"train_loss": 0.09637950360774994, "global_step": 121018, "epoch": 1359, "lr": 5.8627150179044174e-05}
{"train_loss": 0.12140804529190063, "global_step": 121019, "epoch": 1359, "lr": 5.8626579136758776e-05}
{"train_loss": 0.19890472292900085, "global_step": 121020, "epoch": 1359, "lr": 5.862600809331362e-05}
{"train_loss": 0.25888386368751526, "global_step": 121021, "epoch": 1359, "lr": 5.862543704870882e-05}
{"train_loss": 0.15561072528362274, "global_step": 121022, "epoch": 1359, "lr": 5.8624866002944424e-05}
{"train_loss": 0.13951966166496277, "global_step": 121023, "epoch": 1359, "lr": 5.862429495602052e-05}
{"train_loss": 0.1777292937040329, "global_step": 121024, "epoch": 1359, "lr": 5.8623723907937186e-05}
{"train_loss": 0.08910991996526718, "global_step": 121025, "epoch": 1359, "lr": 5.86231528586945e-05}
{"train_loss": 0.13318891823291779, "global_step": 121026, "epoch": 1359, "lr": 5.862258180829253e-05}
{"train_loss": 0.18743015825748444, "global_step": 121027, "epoch": 1359, "lr": 5.862201075673137e-05}
{"train_loss": 0.14311730861663818, "global_step": 121028, "epoch": 1359, "lr": 5.862143970401108e-05}
{"train_loss": 0.149459570646286, "global_step": 121029, "epoch": 1359, "lr": 5.862086865013174e-05}
{"train_loss": 0.11274567991495132, "global_step": 121030, "epoch": 1359, "lr": 5.862029759509344e-05}
{"train_loss": 0.13648420572280884, "global_step": 121031, "epoch": 1359, "lr": 5.8619726538896234e-05}
{"train_loss": 0.18291738629341125, "global_step": 121032, "epoch": 1359, "lr": 5.861915548154022e-05}
{"train_loss": 0.16109038889408112, "global_step": 121033, "epoch": 1359, "lr": 5.8618584423025456e-05}
{"train_loss": 0.1967002898454666, "global_step": 121034, "epoch": 1359, "lr": 5.861801336335202e-05}
{"train_loss": 0.09407199919223785, "global_step": 121035, "epoch": 1359, "lr": 5.861744230252002e-05}
{"train_loss": 0.12065651267766953, "global_step": 121036, "epoch": 1359, "lr": 5.861687124052949e-05}
{"train_loss": 0.16688190400600433, "global_step": 121037, "epoch": 1359, "lr": 5.8616300177380534e-05}
{"train_loss": 0.1119384616613388, "global_step": 121038, "epoch": 1359, "lr": 5.861572911307323e-05}
{"train_loss": 0.14414151339383607, "global_step": 121039, "epoch": 1359, "lr": 5.861515804760764e-05, "val_loss": 5.40269136428833}
{"train_loss": 0.150577574968338, "global_step": 121040, "epoch": 1360, "lr": 5.861458698098384e-05}
{"train_loss": 0.11102749407291412, "global_step": 121041, "epoch": 1360, "lr": 5.861401591320193e-05}
{"train_loss": 0.13319657742977142, "global_step": 121042, "epoch": 1360, "lr": 5.8613444844261954e-05}
{"train_loss": 0.10882169008255005, "global_step": 121043, "epoch": 1360, "lr": 5.861287377416401e-05}
{"train_loss": 0.12584643065929413, "global_step": 121044, "epoch": 1360, "lr": 5.8612302702908184e-05}
{"train_loss": 0.13106192648410797, "global_step": 121045, "epoch": 1360, "lr": 5.861173163049453e-05}
{"train_loss": 0.16066274046897888, "global_step": 121046, "epoch": 1360, "lr": 5.861116055692313e-05}
{"train_loss": 0.1471867561340332, "global_step": 121047, "epoch": 1360, "lr": 5.861058948219406e-05}
{"train_loss": 0.09382949024438858, "global_step": 121048, "epoch": 1360, "lr": 5.86100184063074e-05}
{"train_loss": 0.17808401584625244, "global_step": 121049, "epoch": 1360, "lr": 5.860944732926326e-05}
{"train_loss": 0.14504565298557281, "global_step": 121050, "epoch": 1360, "lr": 5.860887625106165e-05}
{"train_loss": 0.16278238594532013, "global_step": 121051, "epoch": 1360, "lr": 5.860830517170269e-05}
{"train_loss": 0.12373597174882889, "global_step": 121052, "epoch": 1360, "lr": 5.860773409118644e-05}
{"train_loss": 0.10267187654972076, "global_step": 121053, "epoch": 1360, "lr": 5.860716300951299e-05}
{"train_loss": 0.1414719671010971, "global_step": 121054, "epoch": 1360, "lr": 5.860659192668242e-05}
{"train_loss": 0.16240845620632172, "global_step": 121055, "epoch": 1360, "lr": 5.86060208426948e-05}
{"train_loss": 0.12851649522781372, "global_step": 121056, "epoch": 1360, "lr": 5.860544975755019e-05}
{"train_loss": 0.13553839921951294, "global_step": 121057, "epoch": 1360, "lr": 5.86048786712487e-05}
{"train_loss": 0.08913742005825043, "global_step": 121058, "epoch": 1360, "lr": 5.860430758379038e-05}
{"train_loss": 0.10329415649175644, "global_step": 121059, "epoch": 1360, "lr": 5.860373649517531e-05}
{"train_loss": 0.11808934062719345, "global_step": 121060, "epoch": 1360, "lr": 5.860316540540358e-05}
{"train_loss": 0.14804811775684357, "global_step": 121061, "epoch": 1360, "lr": 5.8602594314475256e-05}
{"train_loss": 0.1446283459663391, "global_step": 121062, "epoch": 1360, "lr": 5.860202322239042e-05}
{"train_loss": 0.19131846725940704, "global_step": 121063, "epoch": 1360, "lr": 5.860145212914916e-05}
{"train_loss": 0.13767190277576447, "global_step": 121064, "epoch": 1360, "lr": 5.8600881034751524e-05}
{"train_loss": 0.15518537163734436, "global_step": 121065, "epoch": 1360, "lr": 5.8600309939197606e-05}
{"train_loss": 0.1429271697998047, "global_step": 121066, "epoch": 1360, "lr": 5.859973884248748e-05}
{"train_loss": 0.08252720534801483, "global_step": 121067, "epoch": 1360, "lr": 5.859916774462123e-05}
{"train_loss": 0.1894654929637909, "global_step": 121068, "epoch": 1360, "lr": 5.8598596645598925e-05}
{"train_loss": 0.083274707198143, "global_step": 121069, "epoch": 1360, "lr": 5.859802554542064e-05}
{"train_loss": 0.13821235299110413, "global_step": 121070, "epoch": 1360, "lr": 5.8597454444086455e-05}
{"train_loss": 0.16211116313934326, "global_step": 121071, "epoch": 1360, "lr": 5.8596883341596456e-05}
{"train_loss": 0.11230577528476715, "global_step": 121072, "epoch": 1360, "lr": 5.8596312237950704e-05}
{"train_loss": 0.07877462357282639, "global_step": 121073, "epoch": 1360, "lr": 5.8595741133149274e-05}
{"train_loss": 0.2089829444885254, "global_step": 121074, "epoch": 1360, "lr": 5.859517002719226e-05}
{"train_loss": 0.1900891661643982, "global_step": 121075, "epoch": 1360, "lr": 5.8594598920079724e-05}
{"train_loss": 0.18037375807762146, "global_step": 121076, "epoch": 1360, "lr": 5.859402781181176e-05}
{"train_loss": 0.0912308394908905, "global_step": 121077, "epoch": 1360, "lr": 5.8593456702388424e-05}
{"train_loss": 0.11685005575418472, "global_step": 121078, "epoch": 1360, "lr": 5.859288559180981e-05}
{"train_loss": 0.18164345622062683, "global_step": 121079, "epoch": 1360, "lr": 5.8592314480075984e-05}
{"train_loss": 0.08366856724023819, "global_step": 121080, "epoch": 1360, "lr": 5.859174336718702e-05}
{"train_loss": 0.08249950408935547, "global_step": 121081, "epoch": 1360, "lr": 5.859117225314301e-05}
{"train_loss": 0.11365202069282532, "global_step": 121082, "epoch": 1360, "lr": 5.859060113794402e-05}
{"train_loss": 0.10314194113016129, "global_step": 121083, "epoch": 1360, "lr": 5.859003002159014e-05}
{"train_loss": 0.10451566427946091, "global_step": 121084, "epoch": 1360, "lr": 5.858945890408142e-05}
{"train_loss": 0.09340840578079224, "global_step": 121085, "epoch": 1360, "lr": 5.858888778541794e-05}
{"train_loss": 0.22608967125415802, "global_step": 121086, "epoch": 1360, "lr": 5.8588316665599804e-05}
{"train_loss": 0.19760538637638092, "global_step": 121087, "epoch": 1360, "lr": 5.858774554462708e-05}
{"train_loss": 0.15948952734470367, "global_step": 121088, "epoch": 1360, "lr": 5.8587174422499833e-05}
{"train_loss": 0.15991850197315216, "global_step": 121089, "epoch": 1360, "lr": 5.858660329921814e-05}
{"train_loss": 0.24698425829410553, "global_step": 121090, "epoch": 1360, "lr": 5.858603217478209e-05}
{"train_loss": 0.12244770675897598, "global_step": 121091, "epoch": 1360, "lr": 5.858546104919175e-05}
{"train_loss": 0.13135862350463867, "global_step": 121092, "epoch": 1360, "lr": 5.8584889922447196e-05}
{"train_loss": 0.15089260041713715, "global_step": 121093, "epoch": 1360, "lr": 5.858431879454851e-05}
{"train_loss": 0.15326862037181854, "global_step": 121094, "epoch": 1360, "lr": 5.858374766549577e-05}
{"train_loss": 0.08772505074739456, "global_step": 121095, "epoch": 1360, "lr": 5.8583176535289043e-05}
{"train_loss": 0.11794698238372803, "global_step": 121096, "epoch": 1360, "lr": 5.8582605403928414e-05}
{"train_loss": 0.13342319428920746, "global_step": 121097, "epoch": 1360, "lr": 5.8582034271413956e-05}
{"train_loss": 0.10771606862545013, "global_step": 121098, "epoch": 1360, "lr": 5.858146313774575e-05}
{"train_loss": 0.09899993240833282, "global_step": 121099, "epoch": 1360, "lr": 5.858089200292388e-05}
{"train_loss": 0.05652552470564842, "global_step": 121100, "epoch": 1360, "lr": 5.85803208669484e-05}
{"train_loss": 0.1515522003173828, "global_step": 121101, "epoch": 1360, "lr": 5.857974972981941e-05}
{"train_loss": 0.08226518332958221, "global_step": 121102, "epoch": 1360, "lr": 5.8579178591536975e-05}
{"train_loss": 0.1689702272415161, "global_step": 121103, "epoch": 1360, "lr": 5.8578607452101174e-05}
{"train_loss": 0.14993081986904144, "global_step": 121104, "epoch": 1360, "lr": 5.8578036311512086e-05}
{"train_loss": 0.11197800189256668, "global_step": 121105, "epoch": 1360, "lr": 5.857746516976979e-05}
{"train_loss": 0.10424699634313583, "global_step": 121106, "epoch": 1360, "lr": 5.857689402687434e-05}
{"train_loss": 0.1486901044845581, "global_step": 121107, "epoch": 1360, "lr": 5.857632288282585e-05}
{"train_loss": 0.12021446228027344, "global_step": 121108, "epoch": 1360, "lr": 5.857575173762437e-05}
{"train_loss": 0.11195184290409088, "global_step": 121109, "epoch": 1360, "lr": 5.8575180591269987e-05}
{"train_loss": 0.09095192700624466, "global_step": 121110, "epoch": 1360, "lr": 5.8574609443762775e-05}
{"train_loss": 0.11124566942453384, "global_step": 121111, "epoch": 1360, "lr": 5.8574038295102805e-05}
{"train_loss": 0.09977786988019943, "global_step": 121112, "epoch": 1360, "lr": 5.857346714529016e-05}
{"train_loss": 0.15581431984901428, "global_step": 121113, "epoch": 1360, "lr": 5.8572895994324925e-05}
{"train_loss": 0.11976472288370132, "global_step": 121114, "epoch": 1360, "lr": 5.857232484220717e-05}
{"train_loss": 0.12323286384344101, "global_step": 121115, "epoch": 1360, "lr": 5.857175368893697e-05}
{"train_loss": 0.09191691875457764, "global_step": 121116, "epoch": 1360, "lr": 5.857118253451439e-05}
{"train_loss": 0.10029936581850052, "global_step": 121117, "epoch": 1360, "lr": 5.857061137893953e-05}
{"train_loss": 0.1211063340306282, "global_step": 121118, "epoch": 1360, "lr": 5.857004022221245e-05}
{"train_loss": 0.1314571052789688, "global_step": 121119, "epoch": 1360, "lr": 5.856946906433324e-05}
{"train_loss": 0.11257502436637878, "global_step": 121120, "epoch": 1360, "lr": 5.856889790530197e-05}
{"train_loss": 0.1973721832036972, "global_step": 121121, "epoch": 1360, "lr": 5.8568326745118706e-05}
{"train_loss": 0.05859089642763138, "global_step": 121122, "epoch": 1360, "lr": 5.856775558378353e-05}
{"train_loss": 0.17921409010887146, "global_step": 121123, "epoch": 1360, "lr": 5.8567184421296536e-05}
{"train_loss": 0.1378636658191681, "global_step": 121124, "epoch": 1360, "lr": 5.8566613257657785e-05}
{"train_loss": 0.11841075867414474, "global_step": 121125, "epoch": 1360, "lr": 5.856604209286737e-05}
{"train_loss": 0.08093884587287903, "global_step": 121126, "epoch": 1360, "lr": 5.856547092692534e-05}
{"train_loss": 0.124872125685215, "global_step": 121127, "epoch": 1360, "lr": 5.8564899759831796e-05}
{"train_loss": 0.13055887303492997, "global_step": 121128, "epoch": 1360, "lr": 5.856432859158679e-05, "val_loss": 5.460469722747803, "train_action_mse_error": 29.428359985351562}
{"train_loss": 0.17711907625198364, "global_step": 121129, "epoch": 1361, "lr": 5.8563757422190426e-05}
{"train_loss": 0.15396326780319214, "global_step": 121130, "epoch": 1361, "lr": 5.8563186251642766e-05}
{"train_loss": 0.12823906540870667, "global_step": 121131, "epoch": 1361, "lr": 5.856261507994389e-05}
{"train_loss": 0.1503048837184906, "global_step": 121132, "epoch": 1361, "lr": 5.8562043907093875e-05}
{"train_loss": 0.1095258817076683, "global_step": 121133, "epoch": 1361, "lr": 5.85614727330928e-05}
{"train_loss": 0.12458837777376175, "global_step": 121134, "epoch": 1361, "lr": 5.856090155794074e-05}
{"train_loss": 0.16460035741329193, "global_step": 121135, "epoch": 1361, "lr": 5.856033038163777e-05}
{"train_loss": 0.11079700291156769, "global_step": 121136, "epoch": 1361, "lr": 5.855975920418396e-05}
{"train_loss": 0.05962912365794182, "global_step": 121137, "epoch": 1361, "lr": 5.8559188025579406e-05}
{"train_loss": 0.17726321518421173, "global_step": 121138, "epoch": 1361, "lr": 5.8558616845824166e-05}
{"train_loss": 0.09215369075536728, "global_step": 121139, "epoch": 1361, "lr": 5.855804566491832e-05}
{"train_loss": 0.1477002054452896, "global_step": 121140, "epoch": 1361, "lr": 5.8557474482861965e-05}
{"train_loss": 0.13165529072284698, "global_step": 121141, "epoch": 1361, "lr": 5.8556903299655154e-05}
{"train_loss": 0.12066441029310226, "global_step": 121142, "epoch": 1361, "lr": 5.855633211529798e-05}
{"train_loss": 0.1468629240989685, "global_step": 121143, "epoch": 1361, "lr": 5.8555760929790495e-05}
{"train_loss": 0.20653371512889862, "global_step": 121144, "epoch": 1361, "lr": 5.85551897431328e-05}
{"train_loss": 0.23665229976177216, "global_step": 121145, "epoch": 1361, "lr": 5.855461855532496e-05}
{"train_loss": 0.22173701226711273, "global_step": 121146, "epoch": 1361, "lr": 5.8554047366367056e-05}
{"train_loss": 0.12270490825176239, "global_step": 121147, "epoch": 1361, "lr": 5.855347617625917e-05}
{"train_loss": 0.14520539343357086, "global_step": 121148, "epoch": 1361, "lr": 5.855290498500138e-05}
{"train_loss": 0.12602302432060242, "global_step": 121149, "epoch": 1361, "lr": 5.855233379259374e-05}
{"train_loss": 0.11764365434646606, "global_step": 121150, "epoch": 1361, "lr": 5.855176259903635e-05}
{"train_loss": 0.09937036782503128, "global_step": 121151, "epoch": 1361, "lr": 5.8551191404329285e-05}
{"train_loss": 0.14240232110023499, "global_step": 121152, "epoch": 1361, "lr": 5.8550620208472606e-05}
{"train_loss": 0.17426574230194092, "global_step": 121153, "epoch": 1361, "lr": 5.8550049011466415e-05}
{"train_loss": 0.11068610846996307, "global_step": 121154, "epoch": 1361, "lr": 5.854947781331076e-05}
{"train_loss": 0.09140610694885254, "global_step": 121155, "epoch": 1361, "lr": 5.854890661400574e-05}
{"train_loss": 0.1896928995847702, "global_step": 121156, "epoch": 1361, "lr": 5.854833541355143e-05}
{"train_loss": 0.126891627907753, "global_step": 121157, "epoch": 1361, "lr": 5.8547764211947874e-05}
{"train_loss": 0.08302487432956696, "global_step": 121158, "epoch": 1361, "lr": 5.85471930091952e-05}
{"train_loss": 0.11627224832773209, "global_step": 121159, "epoch": 1361, "lr": 5.854662180529346e-05}
{"train_loss": 0.14389446377754211, "global_step": 121160, "epoch": 1361, "lr": 5.854605060024272e-05}
{"train_loss": 0.09051226079463959, "global_step": 121161, "epoch": 1361, "lr": 5.854547939404307e-05}
{"train_loss": 0.12933525443077087, "global_step": 121162, "epoch": 1361, "lr": 5.854490818669458e-05}
{"train_loss": 0.12753137946128845, "global_step": 121163, "epoch": 1361, "lr": 5.854433697819735e-05}
{"train_loss": 0.08817856758832932, "global_step": 121164, "epoch": 1361, "lr": 5.854376576855142e-05}
{"train_loss": 0.1132841408252716, "global_step": 121165, "epoch": 1361, "lr": 5.854319455775689e-05}
{"train_loss": 0.12619879841804504, "global_step": 121166, "epoch": 1361, "lr": 5.854262334581383e-05}
{"train_loss": 0.1763433963060379, "global_step": 121167, "epoch": 1361, "lr": 5.8542052132722316e-05}
{"train_loss": 0.19579936563968658, "global_step": 121168, "epoch": 1361, "lr": 5.8541480918482427e-05}
{"train_loss": 0.08781325817108154, "global_step": 121169, "epoch": 1361, "lr": 5.854090970309425e-05}
{"train_loss": 0.1003195196390152, "global_step": 121170, "epoch": 1361, "lr": 5.8540338486557846e-05}
{"train_loss": 0.23320117592811584, "global_step": 121171, "epoch": 1361, "lr": 5.853976726887329e-05}
{"train_loss": 0.12032028287649155, "global_step": 121172, "epoch": 1361, "lr": 5.853919605004068e-05}
{"train_loss": 0.15529963374137878, "global_step": 121173, "epoch": 1361, "lr": 5.8538624830060075e-05}
{"train_loss": 0.17057479918003082, "global_step": 121174, "epoch": 1361, "lr": 5.853805360893155e-05}
{"train_loss": 0.14929942786693573, "global_step": 121175, "epoch": 1361, "lr": 5.8537482386655196e-05}
{"train_loss": 0.09746378660202026, "global_step": 121176, "epoch": 1361, "lr": 5.853691116323108e-05}
{"train_loss": 0.10899724811315536, "global_step": 121177, "epoch": 1361, "lr": 5.853633993865927e-05}
{"train_loss": 0.11591903865337372, "global_step": 121178, "epoch": 1361, "lr": 5.853576871293986e-05}
{"train_loss": 0.09806592762470245, "global_step": 121179, "epoch": 1361, "lr": 5.8535197486072926e-05}
{"train_loss": 0.12201287597417831, "global_step": 121180, "epoch": 1361, "lr": 5.853462625805853e-05}
{"train_loss": 0.16065704822540283, "global_step": 121181, "epoch": 1361, "lr": 5.853405502889676e-05}
{"train_loss": 0.15945123136043549, "global_step": 121182, "epoch": 1361, "lr": 5.8533483798587695e-05}
{"train_loss": 0.13202659785747528, "global_step": 121183, "epoch": 1361, "lr": 5.85329125671314e-05}
{"train_loss": 0.164016455411911, "global_step": 121184, "epoch": 1361, "lr": 5.8532341334527965e-05}
{"train_loss": 0.12771107256412506, "global_step": 121185, "epoch": 1361, "lr": 5.853177010077746e-05}
{"train_loss": 0.117996446788311, "global_step": 121186, "epoch": 1361, "lr": 5.853119886587995e-05}
{"train_loss": 0.15574227273464203, "global_step": 121187, "epoch": 1361, "lr": 5.8530627629835545e-05}
{"train_loss": 0.11316473037004471, "global_step": 121188, "epoch": 1361, "lr": 5.853005639264428e-05}
{"train_loss": 0.16852690279483795, "global_step": 121189, "epoch": 1361, "lr": 5.852948515430628e-05}
{"train_loss": 0.1866520494222641, "global_step": 121190, "epoch": 1361, "lr": 5.8528913914821574e-05}
{"train_loss": 0.08832577615976334, "global_step": 121191, "epoch": 1361, "lr": 5.8528342674190265e-05}
{"train_loss": 0.1270221322774887, "global_step": 121192, "epoch": 1361, "lr": 5.852777143241241e-05}
{"train_loss": 0.22247181832790375, "global_step": 121193, "epoch": 1361, "lr": 5.852720018948812e-05}
{"train_loss": 0.12454766035079956, "global_step": 121194, "epoch": 1361, "lr": 5.852662894541745e-05}
{"train_loss": 0.17036275565624237, "global_step": 121195, "epoch": 1361, "lr": 5.852605770020047e-05}
{"train_loss": 0.11633168160915375, "global_step": 121196, "epoch": 1361, "lr": 5.852548645383727e-05}
{"train_loss": 0.16035442054271698, "global_step": 121197, "epoch": 1361, "lr": 5.8524915206327916e-05}
{"train_loss": 0.16976304352283478, "global_step": 121198, "epoch": 1361, "lr": 5.852434395767249e-05}
{"train_loss": 0.1023206040263176, "global_step": 121199, "epoch": 1361, "lr": 5.8523772707871086e-05}
{"train_loss": 0.11803176999092102, "global_step": 121200, "epoch": 1361, "lr": 5.8523201456923756e-05}
{"train_loss": 0.16759520769119263, "global_step": 121201, "epoch": 1361, "lr": 5.852263020483059e-05}
{"train_loss": 0.11761177331209183, "global_step": 121202, "epoch": 1361, "lr": 5.8522058951591655e-05}
{"train_loss": 0.16430969536304474, "global_step": 121203, "epoch": 1361, "lr": 5.8521487697207025e-05}
{"train_loss": 0.21884000301361084, "global_step": 121204, "epoch": 1361, "lr": 5.85209164416768e-05}
{"train_loss": 0.1132502555847168, "global_step": 121205, "epoch": 1361, "lr": 5.852034518500103e-05}
{"train_loss": 0.1845192313194275, "global_step": 121206, "epoch": 1361, "lr": 5.851977392717981e-05}
{"train_loss": 0.15512718260288239, "global_step": 121207, "epoch": 1361, "lr": 5.8519202668213205e-05}
{"train_loss": 0.10192862898111343, "global_step": 121208, "epoch": 1361, "lr": 5.85186314081013e-05}
{"train_loss": 0.13782133162021637, "global_step": 121209, "epoch": 1361, "lr": 5.851806014684417e-05}
{"train_loss": 0.11098197102546692, "global_step": 121210, "epoch": 1361, "lr": 5.8517488884441883e-05}
{"train_loss": 0.20135365426540375, "global_step": 121211, "epoch": 1361, "lr": 5.851691762089453e-05}
{"train_loss": 0.20236723124980927, "global_step": 121212, "epoch": 1361, "lr": 5.851634635620219e-05}
{"train_loss": 0.11674489080905914, "global_step": 121213, "epoch": 1361, "lr": 5.851577509036492e-05}
{"train_loss": 0.1055912896990776, "global_step": 121214, "epoch": 1361, "lr": 5.8515203823382805e-05}
{"train_loss": 0.11616014689207077, "global_step": 121215, "epoch": 1361, "lr": 5.851463255525592e-05}
{"train_loss": 0.2872580289840698, "global_step": 121216, "epoch": 1361, "lr": 5.851406128598435e-05}
{"train_loss": 0.1422134850048617, "global_step": 121217, "epoch": 1361, "lr": 5.851349001556818e-05, "val_loss": 5.4098992347717285}
{"train_loss": 0.13223813474178314, "global_step": 121218, "epoch": 1362, "lr": 5.8512918744007486e-05}
{"train_loss": 0.10689602047204971, "global_step": 121219, "epoch": 1362, "lr": 5.8512347471302306e-05}
{"train_loss": 0.1365669071674347, "global_step": 121220, "epoch": 1362, "lr": 5.8511776197452764e-05}
{"train_loss": 0.08901719003915787, "global_step": 121221, "epoch": 1362, "lr": 5.851120492245891e-05}
{"train_loss": 0.11134354770183563, "global_step": 121222, "epoch": 1362, "lr": 5.8510633646320825e-05}
{"train_loss": 0.1523442417383194, "global_step": 121223, "epoch": 1362, "lr": 5.8510062369038584e-05}
{"train_loss": 0.1191655695438385, "global_step": 121224, "epoch": 1362, "lr": 5.850949109061228e-05}
{"train_loss": 0.12615132331848145, "global_step": 121225, "epoch": 1362, "lr": 5.8508919811041986e-05}
{"train_loss": 0.12459858506917953, "global_step": 121226, "epoch": 1362, "lr": 5.850834853032775e-05}
{"train_loss": 0.12672510743141174, "global_step": 121227, "epoch": 1362, "lr": 5.850777724846969e-05}
{"train_loss": 0.15618547797203064, "global_step": 121228, "epoch": 1362, "lr": 5.850720596546785e-05}
{"train_loss": 0.14249545335769653, "global_step": 121229, "epoch": 1362, "lr": 5.850663468132233e-05}
{"train_loss": 0.1100970208644867, "global_step": 121230, "epoch": 1362, "lr": 5.8506063396033186e-05}
{"train_loss": 0.10148358345031738, "global_step": 121231, "epoch": 1362, "lr": 5.85054921096005e-05}
{"train_loss": 0.17681188881397247, "global_step": 121232, "epoch": 1362, "lr": 5.8504920822024365e-05}
{"train_loss": 0.15149232745170593, "global_step": 121233, "epoch": 1362, "lr": 5.8504349533304834e-05}
{"train_loss": 0.13063399493694305, "global_step": 121234, "epoch": 1362, "lr": 5.8503778243442006e-05}
{"train_loss": 0.11191011965274811, "global_step": 121235, "epoch": 1362, "lr": 5.850320695243596e-05}
{"train_loss": 0.09505113214254379, "global_step": 121236, "epoch": 1362, "lr": 5.8502635660286753e-05}
{"train_loss": 0.13048772513866425, "global_step": 121237, "epoch": 1362, "lr": 5.850206436699447e-05}
{"train_loss": 0.14434637129306793, "global_step": 121238, "epoch": 1362, "lr": 5.85014930725592e-05}
{"train_loss": 0.12238886952400208, "global_step": 121239, "epoch": 1362, "lr": 5.850092177698099e-05}
{"train_loss": 0.09086766839027405, "global_step": 121240, "epoch": 1362, "lr": 5.8500350480259946e-05}
{"train_loss": 0.13202878832817078, "global_step": 121241, "epoch": 1362, "lr": 5.849977918239613e-05}
{"train_loss": 0.12638774514198303, "global_step": 121242, "epoch": 1362, "lr": 5.849920788338962e-05}
{"train_loss": 0.14774732291698456, "global_step": 121243, "epoch": 1362, "lr": 5.8498636583240495e-05}
{"train_loss": 0.16759918630123138, "global_step": 121244, "epoch": 1362, "lr": 5.8498065281948834e-05}
{"train_loss": 0.10851924121379852, "global_step": 121245, "epoch": 1362, "lr": 5.849749397951471e-05}
{"train_loss": 0.10998571664094925, "global_step": 121246, "epoch": 1362, "lr": 5.849692267593821e-05}
{"train_loss": 0.1105613186955452, "global_step": 121247, "epoch": 1362, "lr": 5.849635137121938e-05}
{"train_loss": 0.07931632548570633, "global_step": 121248, "epoch": 1362, "lr": 5.849578006535834e-05}
{"train_loss": 0.14885035157203674, "global_step": 121249, "epoch": 1362, "lr": 5.8495208758355134e-05}
{"train_loss": 0.20909813046455383, "global_step": 121250, "epoch": 1362, "lr": 5.849463745020985e-05}
{"train_loss": 0.10395825654268265, "global_step": 121251, "epoch": 1362, "lr": 5.8494066140922565e-05}
{"train_loss": 0.15547260642051697, "global_step": 121252, "epoch": 1362, "lr": 5.849349483049337e-05}
{"train_loss": 0.19162188470363617, "global_step": 121253, "epoch": 1362, "lr": 5.8492923518922314e-05}
{"train_loss": 0.07615984231233597, "global_step": 121254, "epoch": 1362, "lr": 5.84923522062095e-05}
{"train_loss": 0.16004744172096252, "global_step": 121255, "epoch": 1362, "lr": 5.849178089235498e-05}
{"train_loss": 0.13533847033977509, "global_step": 121256, "epoch": 1362, "lr": 5.849120957735885e-05}
{"train_loss": 0.15977637469768524, "global_step": 121257, "epoch": 1362, "lr": 5.8490638261221196e-05}
{"train_loss": 0.09592977166175842, "global_step": 121258, "epoch": 1362, "lr": 5.8490066943942066e-05}
{"train_loss": 0.16895830631256104, "global_step": 121259, "epoch": 1362, "lr": 5.848949562552155e-05}
{"train_loss": 0.10261091589927673, "global_step": 121260, "epoch": 1362, "lr": 5.8488924305959726e-05}
{"train_loss": 0.1502261906862259, "global_step": 121261, "epoch": 1362, "lr": 5.8488352985256665e-05}
{"train_loss": 0.269913911819458, "global_step": 121262, "epoch": 1362, "lr": 5.848778166341244e-05}
{"train_loss": 0.10475589334964752, "global_step": 121263, "epoch": 1362, "lr": 5.8487210340427155e-05}
{"train_loss": 0.12002982199192047, "global_step": 121264, "epoch": 1362, "lr": 5.848663901630086e-05}
{"train_loss": 0.1930229365825653, "global_step": 121265, "epoch": 1362, "lr": 5.848606769103364e-05}
{"train_loss": 0.12511728703975677, "global_step": 121266, "epoch": 1362, "lr": 5.848549636462557e-05}
{"train_loss": 0.0918632447719574, "global_step": 121267, "epoch": 1362, "lr": 5.8484925037076734e-05}
{"train_loss": 0.1482088267803192, "global_step": 121268, "epoch": 1362, "lr": 5.8484353708387196e-05}
{"train_loss": 0.13156862556934357, "global_step": 121269, "epoch": 1362, "lr": 5.848378237855703e-05}
{"train_loss": 0.15800608694553375, "global_step": 121270, "epoch": 1362, "lr": 5.848321104758634e-05}
{"train_loss": 0.11822448670864105, "global_step": 121271, "epoch": 1362, "lr": 5.848263971547517e-05}
{"train_loss": 0.22281669080257416, "global_step": 121272, "epoch": 1362, "lr": 5.848206838222363e-05}
{"train_loss": 0.10829656571149826, "global_step": 121273, "epoch": 1362, "lr": 5.8481497047831765e-05}
{"train_loss": 0.14301970601081848, "global_step": 121274, "epoch": 1362, "lr": 5.8480925712299685e-05}
{"train_loss": 0.20315194129943848, "global_step": 121275, "epoch": 1362, "lr": 5.848035437562743e-05}
{"train_loss": 0.15813088417053223, "global_step": 121276, "epoch": 1362, "lr": 5.847978303781512e-05}
{"train_loss": 0.10968875139951706, "global_step": 121277, "epoch": 1362, "lr": 5.847921169886279e-05}
{"train_loss": 0.13357917964458466, "global_step": 121278, "epoch": 1362, "lr": 5.847864035877053e-05}
{"train_loss": 0.1681472212076187, "global_step": 121279, "epoch": 1362, "lr": 5.847806901753842e-05}
{"train_loss": 0.12444306910037994, "global_step": 121280, "epoch": 1362, "lr": 5.8477497675166536e-05}
{"train_loss": 0.16265356540679932, "global_step": 121281, "epoch": 1362, "lr": 5.847692633165496e-05}
{"train_loss": 0.20851381123065948, "global_step": 121282, "epoch": 1362, "lr": 5.847635498700377e-05}
{"train_loss": 0.1719369888305664, "global_step": 121283, "epoch": 1362, "lr": 5.847578364121302e-05}
{"train_loss": 0.12124793976545334, "global_step": 121284, "epoch": 1362, "lr": 5.847521229428282e-05}
{"train_loss": 0.11325442045927048, "global_step": 121285, "epoch": 1362, "lr": 5.847464094621322e-05}
{"train_loss": 0.11766093224287033, "global_step": 121286, "epoch": 1362, "lr": 5.8474069597004324e-05}
{"train_loss": 0.08976687490940094, "global_step": 121287, "epoch": 1362, "lr": 5.847349824665618e-05}
{"train_loss": 0.19085608422756195, "global_step": 121288, "epoch": 1362, "lr": 5.8472926895168866e-05}
{"train_loss": 0.11882522702217102, "global_step": 121289, "epoch": 1362, "lr": 5.847235554254248e-05}
{"train_loss": 0.11215193569660187, "global_step": 121290, "epoch": 1362, "lr": 5.84717841887771e-05}
{"train_loss": 0.09105069190263748, "global_step": 121291, "epoch": 1362, "lr": 5.847121283387278e-05}
{"train_loss": 0.23172900080680847, "global_step": 121292, "epoch": 1362, "lr": 5.847064147782961e-05}
{"train_loss": 0.1782991588115692, "global_step": 121293, "epoch": 1362, "lr": 5.8470070120647665e-05}
{"train_loss": 0.11025363206863403, "global_step": 121294, "epoch": 1362, "lr": 5.846949876232704e-05}
{"train_loss": 0.12124732881784439, "global_step": 121295, "epoch": 1362, "lr": 5.846892740286778e-05}
{"train_loss": 0.0889243483543396, "global_step": 121296, "epoch": 1362, "lr": 5.846835604226998e-05}
{"train_loss": 0.07325167208909988, "global_step": 121297, "epoch": 1362, "lr": 5.846778468053371e-05}
{"train_loss": 0.1557956337928772, "global_step": 121298, "epoch": 1362, "lr": 5.846721331765905e-05}
{"train_loss": 0.19193941354751587, "global_step": 121299, "epoch": 1362, "lr": 5.8466641953646076e-05}
{"train_loss": 0.1969776749610901, "global_step": 121300, "epoch": 1362, "lr": 5.846607058849486e-05}
{"train_loss": 0.1409939080476761, "global_step": 121301, "epoch": 1362, "lr": 5.846549922220549e-05}
{"train_loss": 0.1049867570400238, "global_step": 121302, "epoch": 1362, "lr": 5.846492785477804e-05}
{"train_loss": 0.12429830431938171, "global_step": 121303, "epoch": 1362, "lr": 5.846435648621258e-05}
{"train_loss": 0.185970738530159, "global_step": 121304, "epoch": 1362, "lr": 5.8463785116509185e-05}
{"train_loss": 0.09802636504173279, "global_step": 121305, "epoch": 1362, "lr": 5.846321374566795e-05}
{"train_loss": 0.13690696757161216, "global_step": 121306, "epoch": 1362, "lr": 5.84626423736889e-05, "val_loss": 5.281510829925537}
{"train_loss": 0.1060459166765213, "global_step": 121307, "epoch": 1363, "lr": 5.846207100057219e-05}
{"train_loss": 0.11585038155317307, "global_step": 121308, "epoch": 1363, "lr": 5.846149962631785e-05}
{"train_loss": 0.13828299939632416, "global_step": 121309, "epoch": 1363, "lr": 5.846092825092596e-05}
{"train_loss": 0.10923746973276138, "global_step": 121310, "epoch": 1363, "lr": 5.8460356874396604e-05}
{"train_loss": 0.0818701758980751, "global_step": 121311, "epoch": 1363, "lr": 5.845978549672986e-05}
{"train_loss": 0.12551845610141754, "global_step": 121312, "epoch": 1363, "lr": 5.84592141179258e-05}
{"train_loss": 0.19523824751377106, "global_step": 121313, "epoch": 1363, "lr": 5.8458642737984494e-05}
{"train_loss": 0.1406686156988144, "global_step": 121314, "epoch": 1363, "lr": 5.845807135690603e-05}
{"train_loss": 0.21416768431663513, "global_step": 121315, "epoch": 1363, "lr": 5.84574999746905e-05}
{"train_loss": 0.08220375329256058, "global_step": 121316, "epoch": 1363, "lr": 5.845692859133794e-05}
{"train_loss": 0.1104128360748291, "global_step": 121317, "epoch": 1363, "lr": 5.8456357206848455e-05}
{"train_loss": 0.12659531831741333, "global_step": 121318, "epoch": 1363, "lr": 5.845578582122212e-05}
{"train_loss": 0.1313147395849228, "global_step": 121319, "epoch": 1363, "lr": 5.8455214434459e-05}
{"train_loss": 0.11803337931632996, "global_step": 121320, "epoch": 1363, "lr": 5.845464304655919e-05}
{"train_loss": 0.1465638279914856, "global_step": 121321, "epoch": 1363, "lr": 5.845407165752275e-05}
{"train_loss": 0.16573543846607208, "global_step": 121322, "epoch": 1363, "lr": 5.845350026734976e-05}
{"train_loss": 0.16862770915031433, "global_step": 121323, "epoch": 1363, "lr": 5.84529288760403e-05}
{"train_loss": 0.20315657556056976, "global_step": 121324, "epoch": 1363, "lr": 5.845235748359443e-05}
{"train_loss": 0.17879121005535126, "global_step": 121325, "epoch": 1363, "lr": 5.845178609001227e-05}
{"train_loss": 0.19003434479236603, "global_step": 121326, "epoch": 1363, "lr": 5.8451214695293865e-05}
{"train_loss": 0.17589950561523438, "global_step": 121327, "epoch": 1363, "lr": 5.845064329943929e-05}
{"train_loss": 0.13433469831943512, "global_step": 121328, "epoch": 1363, "lr": 5.845007190244863e-05}
{"train_loss": 0.1476413458585739, "global_step": 121329, "epoch": 1363, "lr": 5.844950050432198e-05}
{"train_loss": 0.0944695994257927, "global_step": 121330, "epoch": 1363, "lr": 5.844892910505937e-05}
{"train_loss": 0.216363325715065, "global_step": 121331, "epoch": 1363, "lr": 5.8448357704660915e-05}
{"train_loss": 0.08478519320487976, "global_step": 121332, "epoch": 1363, "lr": 5.8447786303126686e-05}
{"train_loss": 0.14695195853710175, "global_step": 121333, "epoch": 1363, "lr": 5.844721490045676e-05}
{"train_loss": 0.17601853609085083, "global_step": 121334, "epoch": 1363, "lr": 5.844664349665121e-05}
{"train_loss": 0.15075147151947021, "global_step": 121335, "epoch": 1363, "lr": 5.84460720917101e-05}
{"train_loss": 0.13988341391086578, "global_step": 121336, "epoch": 1363, "lr": 5.8445500685633524e-05}
{"train_loss": 0.1390596330165863, "global_step": 121337, "epoch": 1363, "lr": 5.8444929278421545e-05}
{"train_loss": 0.13870611786842346, "global_step": 121338, "epoch": 1363, "lr": 5.844435787007425e-05}
{"train_loss": 0.13779348134994507, "global_step": 121339, "epoch": 1363, "lr": 5.844378646059172e-05}
{"train_loss": 0.1325920969247818, "global_step": 121340, "epoch": 1363, "lr": 5.844321504997403e-05}
{"train_loss": 0.202494278550148, "global_step": 121341, "epoch": 1363, "lr": 5.844264363822125e-05}
{"train_loss": 0.11794646084308624, "global_step": 121342, "epoch": 1363, "lr": 5.844207222533344e-05}
{"train_loss": 0.13338840007781982, "global_step": 121343, "epoch": 1363, "lr": 5.8441500811310714e-05}
{"train_loss": 0.1531117558479309, "global_step": 121344, "epoch": 1363, "lr": 5.844092939615313e-05}
{"train_loss": 0.14010116457939148, "global_step": 121345, "epoch": 1363, "lr": 5.844035797986076e-05}
{"train_loss": 0.15725140273571014, "global_step": 121346, "epoch": 1363, "lr": 5.8439786562433696e-05}
{"train_loss": 0.14509253203868866, "global_step": 121347, "epoch": 1363, "lr": 5.8439215143872004e-05}
{"train_loss": 0.12867538630962372, "global_step": 121348, "epoch": 1363, "lr": 5.8438643724175756e-05}
{"train_loss": 0.18405666947364807, "global_step": 121349, "epoch": 1363, "lr": 5.843807230334504e-05}
{"train_loss": 0.09743223339319229, "global_step": 121350, "epoch": 1363, "lr": 5.843750088137993e-05}
{"train_loss": 0.13196395337581635, "global_step": 121351, "epoch": 1363, "lr": 5.84369294582805e-05}
{"train_loss": 0.1197417825460434, "global_step": 121352, "epoch": 1363, "lr": 5.843635803404682e-05}
{"train_loss": 0.08277330547571182, "global_step": 121353, "epoch": 1363, "lr": 5.8435786608678976e-05}
{"train_loss": 0.1240423247218132, "global_step": 121354, "epoch": 1363, "lr": 5.843521518217706e-05}
{"train_loss": 0.1322987973690033, "global_step": 121355, "epoch": 1363, "lr": 5.843464375454112e-05}
{"train_loss": 0.16892200708389282, "global_step": 121356, "epoch": 1363, "lr": 5.843407232577125e-05}
{"train_loss": 0.20452980697155, "global_step": 121357, "epoch": 1363, "lr": 5.843350089586751e-05}
{"train_loss": 0.09656849503517151, "global_step": 121358, "epoch": 1363, "lr": 5.843292946483e-05}
{"train_loss": 0.13047535717487335, "global_step": 121359, "epoch": 1363, "lr": 5.843235803265878e-05}
{"train_loss": 0.1484854817390442, "global_step": 121360, "epoch": 1363, "lr": 5.843178659935392e-05}
{"train_loss": 0.19807979464530945, "global_step": 121361, "epoch": 1363, "lr": 5.843121516491552e-05}
{"train_loss": 0.11927193403244019, "global_step": 121362, "epoch": 1363, "lr": 5.843064372934365e-05}
{"train_loss": 0.12210535258054733, "global_step": 121363, "epoch": 1363, "lr": 5.843007229263838e-05}
{"train_loss": 0.08386200666427612, "global_step": 121364, "epoch": 1363, "lr": 5.842950085479979e-05}
{"train_loss": 0.1090708076953888, "global_step": 121365, "epoch": 1363, "lr": 5.842892941582796e-05}
{"train_loss": 0.10788343846797943, "global_step": 121366, "epoch": 1363, "lr": 5.842835797572296e-05}
{"train_loss": 0.21004872024059296, "global_step": 121367, "epoch": 1363, "lr": 5.842778653448486e-05}
{"train_loss": 0.22835767269134521, "global_step": 121368, "epoch": 1363, "lr": 5.8427215092113765e-05}
{"train_loss": 0.12804028391838074, "global_step": 121369, "epoch": 1363, "lr": 5.8426643648609725e-05}
{"train_loss": 0.13619916141033173, "global_step": 121370, "epoch": 1363, "lr": 5.842607220397283e-05}
{"train_loss": 0.11922629922628403, "global_step": 121371, "epoch": 1363, "lr": 5.842550075820314e-05}
{"train_loss": 0.1317155808210373, "global_step": 121372, "epoch": 1363, "lr": 5.8424929311300745e-05}
{"train_loss": 0.20138977468013763, "global_step": 121373, "epoch": 1363, "lr": 5.842435786326574e-05}
{"train_loss": 0.1964614987373352, "global_step": 121374, "epoch": 1363, "lr": 5.842378641409817e-05}
{"train_loss": 0.1060180515050888, "global_step": 121375, "epoch": 1363, "lr": 5.8423214963798125e-05}
{"train_loss": 0.07462844252586365, "global_step": 121376, "epoch": 1363, "lr": 5.842264351236568e-05}
{"train_loss": 0.08943527936935425, "global_step": 121377, "epoch": 1363, "lr": 5.842207205980093e-05}
{"train_loss": 0.10807856172323227, "global_step": 121378, "epoch": 1363, "lr": 5.8421500606103895e-05}
{"train_loss": 0.15394146740436554, "global_step": 121379, "epoch": 1363, "lr": 5.842092915127472e-05}
{"train_loss": 0.0496654249727726, "global_step": 121380, "epoch": 1363, "lr": 5.8420357695313446e-05}
{"train_loss": 0.1035977378487587, "global_step": 121381, "epoch": 1363, "lr": 5.8419786238220164e-05}
{"train_loss": 0.09705904126167297, "global_step": 121382, "epoch": 1363, "lr": 5.841921477999495e-05}
{"train_loss": 0.10794266313314438, "global_step": 121383, "epoch": 1363, "lr": 5.8418643320637865e-05}
{"train_loss": 0.14832784235477448, "global_step": 121384, "epoch": 1363, "lr": 5.841807186014899e-05}
{"train_loss": 0.13068419694900513, "global_step": 121385, "epoch": 1363, "lr": 5.841750039852843e-05}
{"train_loss": 0.12144879996776581, "global_step": 121386, "epoch": 1363, "lr": 5.841692893577622e-05}
{"train_loss": 0.13228164613246918, "global_step": 121387, "epoch": 1363, "lr": 5.841635747189246e-05}
{"train_loss": 0.15472552180290222, "global_step": 121388, "epoch": 1363, "lr": 5.841578600687723e-05}
{"train_loss": 0.1569935530424118, "global_step": 121389, "epoch": 1363, "lr": 5.841521454073059e-05}
{"train_loss": 0.18333354592323303, "global_step": 121390, "epoch": 1363, "lr": 5.8414643073452636e-05}
{"train_loss": 0.14776302874088287, "global_step": 121391, "epoch": 1363, "lr": 5.841407160504343e-05}
{"train_loss": 0.111051045358181, "global_step": 121392, "epoch": 1363, "lr": 5.8413500135503065e-05}
{"train_loss": 0.09647572785615921, "global_step": 121393, "epoch": 1363, "lr": 5.84129286648316e-05}
{"train_loss": 0.22624504566192627, "global_step": 121394, "epoch": 1363, "lr": 5.8412357193029124e-05}
{"train_loss": 0.13900002213508894, "global_step": 121395, "epoch": 1363, "lr": 5.84117857200957e-05, "val_loss": 5.4166259765625}
{"train_loss": 0.17050018906593323, "global_step": 121396, "epoch": 1364, "lr": 5.841121424603141e-05}
{"train_loss": 0.13951432704925537, "global_step": 121397, "epoch": 1364, "lr": 5.8410642770836344e-05}
{"train_loss": 0.09109195321798325, "global_step": 121398, "epoch": 1364, "lr": 5.841007129451057e-05}
{"train_loss": 0.12090732902288437, "global_step": 121399, "epoch": 1364, "lr": 5.840949981705416e-05}
{"train_loss": 0.09952379763126373, "global_step": 121400, "epoch": 1364, "lr": 5.8408928338467195e-05}
{"train_loss": 0.15512330830097198, "global_step": 121401, "epoch": 1364, "lr": 5.840835685874976e-05}
{"train_loss": 0.11451449990272522, "global_step": 121402, "epoch": 1364, "lr": 5.840778537790191e-05}
{"train_loss": 0.07826616615056992, "global_step": 121403, "epoch": 1364, "lr": 5.840721389592375e-05}
{"train_loss": 0.1068841814994812, "global_step": 121404, "epoch": 1364, "lr": 5.8406642412815336e-05}
{"train_loss": 0.12124009430408478, "global_step": 121405, "epoch": 1364, "lr": 5.840607092857675e-05}
{"train_loss": 0.24198855459690094, "global_step": 121406, "epoch": 1364, "lr": 5.840549944320808e-05}
{"train_loss": 0.18233850598335266, "global_step": 121407, "epoch": 1364, "lr": 5.8404927956709376e-05}
{"train_loss": 0.14115072786808014, "global_step": 121408, "epoch": 1364, "lr": 5.8404356469080747e-05}
{"train_loss": 0.12324430793523788, "global_step": 121409, "epoch": 1364, "lr": 5.840378498032225e-05}
{"train_loss": 0.10500679910182953, "global_step": 121410, "epoch": 1364, "lr": 5.840321349043396e-05}
{"train_loss": 0.10110257565975189, "global_step": 121411, "epoch": 1364, "lr": 5.840264199941598e-05}
{"train_loss": 0.11511122435331345, "global_step": 121412, "epoch": 1364, "lr": 5.840207050726836e-05}
{"train_loss": 0.12419421970844269, "global_step": 121413, "epoch": 1364, "lr": 5.840149901399118e-05}
{"train_loss": 0.1944343000650406, "global_step": 121414, "epoch": 1364, "lr": 5.840092751958452e-05}
{"train_loss": 0.17272040247917175, "global_step": 121415, "epoch": 1364, "lr": 5.840035602404844e-05}
{"train_loss": 0.08152079582214355, "global_step": 121416, "epoch": 1364, "lr": 5.839978452738305e-05}
{"train_loss": 0.17636598646640778, "global_step": 121417, "epoch": 1364, "lr": 5.839921302958841e-05}
{"train_loss": 0.24383676052093506, "global_step": 121418, "epoch": 1364, "lr": 5.83986415306646e-05}
{"train_loss": 0.13215559720993042, "global_step": 121419, "epoch": 1364, "lr": 5.839807003061169e-05}
{"train_loss": 0.13213799893856049, "global_step": 121420, "epoch": 1364, "lr": 5.839749852942977e-05}
{"train_loss": 0.12001225352287292, "global_step": 121421, "epoch": 1364, "lr": 5.839692702711891e-05}
{"train_loss": 0.15391793847084045, "global_step": 121422, "epoch": 1364, "lr": 5.839635552367918e-05}
{"train_loss": 0.1348588764667511, "global_step": 121423, "epoch": 1364, "lr": 5.8395784019110666e-05}
{"train_loss": 0.17815013229846954, "global_step": 121424, "epoch": 1364, "lr": 5.8395212513413435e-05}
{"train_loss": 0.16779057681560516, "global_step": 121425, "epoch": 1364, "lr": 5.839464100658758e-05}
{"train_loss": 0.10393869131803513, "global_step": 121426, "epoch": 1364, "lr": 5.839406949863315e-05}
{"train_loss": 0.14532756805419922, "global_step": 121427, "epoch": 1364, "lr": 5.839349798955026e-05}
{"train_loss": 0.08912716805934906, "global_step": 121428, "epoch": 1364, "lr": 5.839292647933896e-05}
{"train_loss": 0.12060233950614929, "global_step": 121429, "epoch": 1364, "lr": 5.839235496799933e-05}
{"train_loss": 0.10707419365644455, "global_step": 121430, "epoch": 1364, "lr": 5.8391783455531445e-05}
{"train_loss": 0.16132505238056183, "global_step": 121431, "epoch": 1364, "lr": 5.839121194193541e-05}
{"train_loss": 0.056507792323827744, "global_step": 121432, "epoch": 1364, "lr": 5.8390640427211264e-05}
{"train_loss": 0.11124919354915619, "global_step": 121433, "epoch": 1364, "lr": 5.839006891135908e-05}
{"train_loss": 0.07049813121557236, "global_step": 121434, "epoch": 1364, "lr": 5.838949739437898e-05}
{"train_loss": 0.14421498775482178, "global_step": 121435, "epoch": 1364, "lr": 5.8388925876270994e-05}
{"train_loss": 0.22500039637088776, "global_step": 121436, "epoch": 1364, "lr": 5.8388354357035235e-05}
{"train_loss": 0.14712105691432953, "global_step": 121437, "epoch": 1364, "lr": 5.838778283667176e-05}
{"train_loss": 0.1167217418551445, "global_step": 121438, "epoch": 1364, "lr": 5.838721131518065e-05}
{"train_loss": 0.08357666432857513, "global_step": 121439, "epoch": 1364, "lr": 5.838663979256197e-05}
{"train_loss": 0.09446673095226288, "global_step": 121440, "epoch": 1364, "lr": 5.838606826881582e-05}
{"train_loss": 0.12097807228565216, "global_step": 121441, "epoch": 1364, "lr": 5.838549674394227e-05}
{"train_loss": 0.1300337016582489, "global_step": 121442, "epoch": 1364, "lr": 5.838492521794139e-05}
{"train_loss": 0.12890516221523285, "global_step": 121443, "epoch": 1364, "lr": 5.838435369081325e-05}
{"train_loss": 0.12485021352767944, "global_step": 121444, "epoch": 1364, "lr": 5.838378216255794e-05}
{"train_loss": 0.08867081254720688, "global_step": 121445, "epoch": 1364, "lr": 5.8383210633175533e-05}
{"train_loss": 0.19347019493579865, "global_step": 121446, "epoch": 1364, "lr": 5.8382639102666105e-05}
{"train_loss": 0.05635648965835571, "global_step": 121447, "epoch": 1364, "lr": 5.838206757102973e-05}
{"train_loss": 0.18195419013500214, "global_step": 121448, "epoch": 1364, "lr": 5.83814960382665e-05}
{"train_loss": 0.07874710857868195, "global_step": 121449, "epoch": 1364, "lr": 5.838092450437647e-05}
{"train_loss": 0.10261543840169907, "global_step": 121450, "epoch": 1364, "lr": 5.838035296935974e-05}
{"train_loss": 0.2192220538854599, "global_step": 121451, "epoch": 1364, "lr": 5.837978143321635e-05}
{"train_loss": 0.14968472719192505, "global_step": 121452, "epoch": 1364, "lr": 5.837920989594642e-05}
{"train_loss": 0.131151020526886, "global_step": 121453, "epoch": 1364, "lr": 5.8378638357549996e-05}
{"train_loss": 0.11515536904335022, "global_step": 121454, "epoch": 1364, "lr": 5.8378066818027174e-05}
{"train_loss": 0.1718512773513794, "global_step": 121455, "epoch": 1364, "lr": 5.837749527737802e-05}
{"train_loss": 0.17239142954349518, "global_step": 121456, "epoch": 1364, "lr": 5.8376923735602615e-05}
{"train_loss": 0.16043245792388916, "global_step": 121457, "epoch": 1364, "lr": 5.837635219270103e-05}
{"train_loss": 0.14951519668102264, "global_step": 121458, "epoch": 1364, "lr": 5.837578064867335e-05}
{"train_loss": 0.13211651146411896, "global_step": 121459, "epoch": 1364, "lr": 5.837520910351966e-05}
{"train_loss": 0.24995097517967224, "global_step": 121460, "epoch": 1364, "lr": 5.837463755724001e-05}
{"train_loss": 0.07252676784992218, "global_step": 121461, "epoch": 1364, "lr": 5.8374066009834494e-05}
{"train_loss": 0.06729651242494583, "global_step": 121462, "epoch": 1364, "lr": 5.837349446130319e-05}
{"train_loss": 0.25541678071022034, "global_step": 121463, "epoch": 1364, "lr": 5.8372922911646164e-05}
{"train_loss": 0.17428141832351685, "global_step": 121464, "epoch": 1364, "lr": 5.837235136086352e-05}
{"train_loss": 0.1924443393945694, "global_step": 121465, "epoch": 1364, "lr": 5.83717798089553e-05}
{"train_loss": 0.08217895776033401, "global_step": 121466, "epoch": 1364, "lr": 5.8371208255921596e-05}
{"train_loss": 0.12536276876926422, "global_step": 121467, "epoch": 1364, "lr": 5.8370636701762485e-05}
{"train_loss": 0.09791743755340576, "global_step": 121468, "epoch": 1364, "lr": 5.837006514647805e-05}
{"train_loss": 0.09299660474061966, "global_step": 121469, "epoch": 1364, "lr": 5.836949359006836e-05}
{"train_loss": 0.10668367147445679, "global_step": 121470, "epoch": 1364, "lr": 5.836892203253349e-05}
{"train_loss": 0.12221411615610123, "global_step": 121471, "epoch": 1364, "lr": 5.836835047387352e-05}
{"train_loss": 0.11126932501792908, "global_step": 121472, "epoch": 1364, "lr": 5.8367778914088534e-05}
{"train_loss": 0.12412417680025101, "global_step": 121473, "epoch": 1364, "lr": 5.83672073531786e-05}
{"train_loss": 0.1988600343465805, "global_step": 121474, "epoch": 1364, "lr": 5.836663579114379e-05}
{"train_loss": 0.12591469287872314, "global_step": 121475, "epoch": 1364, "lr": 5.8366064227984195e-05}
{"train_loss": 0.13363361358642578, "global_step": 121476, "epoch": 1364, "lr": 5.836549266369988e-05}
{"train_loss": 0.10786104202270508, "global_step": 121477, "epoch": 1364, "lr": 5.8364921098290935e-05}
{"train_loss": 0.1432664692401886, "global_step": 121478, "epoch": 1364, "lr": 5.8364349531757425e-05}
{"train_loss": 0.10936007648706436, "global_step": 121479, "epoch": 1364, "lr": 5.836377796409942e-05}
{"train_loss": 0.1512385606765747, "global_step": 121480, "epoch": 1364, "lr": 5.836320639531702e-05}
{"train_loss": 0.16577765345573425, "global_step": 121481, "epoch": 1364, "lr": 5.836263482541028e-05}
{"train_loss": 0.19344253838062286, "global_step": 121482, "epoch": 1364, "lr": 5.836206325437928e-05}
{"train_loss": 0.13148696720600128, "global_step": 121483, "epoch": 1364, "lr": 5.836149168222412e-05}
{"train_loss": 0.13572538505946652, "global_step": 121484, "epoch": 1364, "lr": 5.836092010894485e-05, "val_loss": 5.566750526428223}
{"train_loss": 0.13878610730171204, "global_step": 121485, "epoch": 1365, "lr": 5.836034853454156e-05}
{"train_loss": 0.12472481280565262, "global_step": 121486, "epoch": 1365, "lr": 5.8359776959014326e-05}
{"train_loss": 0.12494152039289474, "global_step": 121487, "epoch": 1365, "lr": 5.835920538236321e-05}
{"train_loss": 0.08946402370929718, "global_step": 121488, "epoch": 1365, "lr": 5.83586338045883e-05}
{"train_loss": 0.1473822295665741, "global_step": 121489, "epoch": 1365, "lr": 5.835806222568969e-05}
{"train_loss": 0.13494278490543365, "global_step": 121490, "epoch": 1365, "lr": 5.835749064566742e-05}
{"train_loss": 0.1022319346666336, "global_step": 121491, "epoch": 1365, "lr": 5.835691906452161e-05}
{"train_loss": 0.10510117560625076, "global_step": 121492, "epoch": 1365, "lr": 5.83563474822523e-05}
{"train_loss": 0.10564575344324112, "global_step": 121493, "epoch": 1365, "lr": 5.835577589885958e-05}
{"train_loss": 0.1663166582584381, "global_step": 121494, "epoch": 1365, "lr": 5.8355204314343526e-05}
{"train_loss": 0.17168593406677246, "global_step": 121495, "epoch": 1365, "lr": 5.8354632728704226e-05}
{"train_loss": 0.19859148561954498, "global_step": 121496, "epoch": 1365, "lr": 5.835406114194174e-05}
{"train_loss": 0.1424000859260559, "global_step": 121497, "epoch": 1365, "lr": 5.835348955405615e-05}
{"train_loss": 0.1497073769569397, "global_step": 121498, "epoch": 1365, "lr": 5.835291796504755e-05}
{"train_loss": 0.18410098552703857, "global_step": 121499, "epoch": 1365, "lr": 5.835234637491599e-05}
{"train_loss": 0.19230806827545166, "global_step": 121500, "epoch": 1365, "lr": 5.835177478366156e-05}
{"train_loss": 0.192861869931221, "global_step": 121501, "epoch": 1365, "lr": 5.835120319128433e-05}
{"train_loss": 0.23207205533981323, "global_step": 121502, "epoch": 1365, "lr": 5.83506315977844e-05}
{"train_loss": 0.10360638797283173, "global_step": 121503, "epoch": 1365, "lr": 5.835006000316181e-05}
{"train_loss": 0.12600640952587128, "global_step": 121504, "epoch": 1365, "lr": 5.8349488407416676e-05}
{"train_loss": 0.15571525692939758, "global_step": 121505, "epoch": 1365, "lr": 5.834891681054904e-05}
{"train_loss": 0.10588140040636063, "global_step": 121506, "epoch": 1365, "lr": 5.8348345212558996e-05}
{"train_loss": 0.09148110449314117, "global_step": 121507, "epoch": 1365, "lr": 5.834777361344662e-05}
{"train_loss": 0.07728155702352524, "global_step": 121508, "epoch": 1365, "lr": 5.834720201321199e-05}
{"train_loss": 0.16783632338047028, "global_step": 121509, "epoch": 1365, "lr": 5.834663041185517e-05}
{"train_loss": 0.1030113622546196, "global_step": 121510, "epoch": 1365, "lr": 5.834605880937626e-05}
{"train_loss": 0.12319044768810272, "global_step": 121511, "epoch": 1365, "lr": 5.8345487205775315e-05}
{"train_loss": 0.07814887911081314, "global_step": 121512, "epoch": 1365, "lr": 5.834491560105243e-05}
{"train_loss": 0.1066550612449646, "global_step": 121513, "epoch": 1365, "lr": 5.8344343995207674e-05}
{"train_loss": 0.14826500415802002, "global_step": 121514, "epoch": 1365, "lr": 5.8343772388241114e-05}
{"train_loss": 0.1311999261379242, "global_step": 121515, "epoch": 1365, "lr": 5.834320078015284e-05}
{"train_loss": 0.10217650979757309, "global_step": 121516, "epoch": 1365, "lr": 5.8342629170942916e-05}
{"train_loss": 0.15135401487350464, "global_step": 121517, "epoch": 1365, "lr": 5.834205756061144e-05}
{"train_loss": 0.2209109216928482, "global_step": 121518, "epoch": 1365, "lr": 5.834148594915847e-05}
{"train_loss": 0.06300404667854309, "global_step": 121519, "epoch": 1365, "lr": 5.834091433658409e-05}
{"train_loss": 0.1609087884426117, "global_step": 121520, "epoch": 1365, "lr": 5.834034272288837e-05}
{"train_loss": 0.1530076563358307, "global_step": 121521, "epoch": 1365, "lr": 5.83397711080714e-05}
{"train_loss": 0.11127781122922897, "global_step": 121522, "epoch": 1365, "lr": 5.8339199492133247e-05}
{"train_loss": 0.2418799102306366, "global_step": 121523, "epoch": 1365, "lr": 5.833862787507399e-05}
{"train_loss": 0.0946992039680481, "global_step": 121524, "epoch": 1365, "lr": 5.833805625689371e-05}
{"train_loss": 0.1729089915752411, "global_step": 121525, "epoch": 1365, "lr": 5.8337484637592475e-05}
{"train_loss": 0.1127355769276619, "global_step": 121526, "epoch": 1365, "lr": 5.833691301717037e-05}
{"train_loss": 0.12379468232393265, "global_step": 121527, "epoch": 1365, "lr": 5.833634139562747e-05}
{"train_loss": 0.18380533158779144, "global_step": 121528, "epoch": 1365, "lr": 5.8335769772963846e-05}
{"train_loss": 0.10578018426895142, "global_step": 121529, "epoch": 1365, "lr": 5.833519814917958e-05}
{"train_loss": 0.11177077144384384, "global_step": 121530, "epoch": 1365, "lr": 5.833462652427475e-05}
{"train_loss": 0.1787468045949936, "global_step": 121531, "epoch": 1365, "lr": 5.8334054898249427e-05}
{"train_loss": 0.15484122931957245, "global_step": 121532, "epoch": 1365, "lr": 5.8333483271103696e-05}
{"train_loss": 0.2128971964120865, "global_step": 121533, "epoch": 1365, "lr": 5.8332911642837626e-05}
{"train_loss": 0.19742456078529358, "global_step": 121534, "epoch": 1365, "lr": 5.83323400134513e-05}
{"train_loss": 0.13674475252628326, "global_step": 121535, "epoch": 1365, "lr": 5.83317683829448e-05}
{"train_loss": 0.09298869967460632, "global_step": 121536, "epoch": 1365, "lr": 5.833119675131819e-05}
{"train_loss": 0.16459286212921143, "global_step": 121537, "epoch": 1365, "lr": 5.833062511857155e-05}
{"train_loss": 0.11747953295707703, "global_step": 121538, "epoch": 1365, "lr": 5.833005348470495e-05}
{"train_loss": 0.1402743011713028, "global_step": 121539, "epoch": 1365, "lr": 5.832948184971848e-05}
{"train_loss": 0.11899809539318085, "global_step": 121540, "epoch": 1365, "lr": 5.832891021361222e-05}
{"train_loss": 0.11298613995313644, "global_step": 121541, "epoch": 1365, "lr": 5.832833857638624e-05}
{"train_loss": 0.1798846274614334, "global_step": 121542, "epoch": 1365, "lr": 5.8327766938040614e-05}
{"train_loss": 0.13211895525455475, "global_step": 121543, "epoch": 1365, "lr": 5.832719529857542e-05}
{"train_loss": 0.16379320621490479, "global_step": 121544, "epoch": 1365, "lr": 5.832662365799073e-05}
{"train_loss": 0.14100822806358337, "global_step": 121545, "epoch": 1365, "lr": 5.8326052016286633e-05}
{"train_loss": 0.10757743567228317, "global_step": 121546, "epoch": 1365, "lr": 5.8325480373463204e-05}
{"train_loss": 0.0934586375951767, "global_step": 121547, "epoch": 1365, "lr": 5.832490872952051e-05}
{"train_loss": 0.1418991982936859, "global_step": 121548, "epoch": 1365, "lr": 5.832433708445864e-05}
{"train_loss": 0.12381964176893234, "global_step": 121549, "epoch": 1365, "lr": 5.8323765438277666e-05}
{"train_loss": 0.1870516538619995, "global_step": 121550, "epoch": 1365, "lr": 5.832319379097765e-05}
{"train_loss": 0.15121839940547943, "global_step": 121551, "epoch": 1365, "lr": 5.832262214255869e-05}
{"train_loss": 0.10467281192541122, "global_step": 121552, "epoch": 1365, "lr": 5.832205049302085e-05}
{"train_loss": 0.11878810822963715, "global_step": 121553, "epoch": 1365, "lr": 5.832147884236422e-05}
{"train_loss": 0.21401987969875336, "global_step": 121554, "epoch": 1365, "lr": 5.832090719058887e-05}
{"train_loss": 0.09888168424367905, "global_step": 121555, "epoch": 1365, "lr": 5.832033553769487e-05}
{"train_loss": 0.1577015072107315, "global_step": 121556, "epoch": 1365, "lr": 5.83197638836823e-05}
{"train_loss": 0.1456156075000763, "global_step": 121557, "epoch": 1365, "lr": 5.831919222855125e-05}
{"train_loss": 0.11228697746992111, "global_step": 121558, "epoch": 1365, "lr": 5.8318620572301764e-05}
{"train_loss": 0.11648302525281906, "global_step": 121559, "epoch": 1365, "lr": 5.831804891493395e-05}
{"train_loss": 0.14118549227714539, "global_step": 121560, "epoch": 1365, "lr": 5.8317477256447874e-05}
{"train_loss": 0.12051533162593842, "global_step": 121561, "epoch": 1365, "lr": 5.831690559684363e-05}
{"train_loss": 0.2003978043794632, "global_step": 121562, "epoch": 1365, "lr": 5.831633393612127e-05}
{"train_loss": 0.12161564081907272, "global_step": 121563, "epoch": 1365, "lr": 5.831576227428088e-05}
{"train_loss": 0.1874953657388687, "global_step": 121564, "epoch": 1365, "lr": 5.8315190611322536e-05}
{"train_loss": 0.1805974394083023, "global_step": 121565, "epoch": 1365, "lr": 5.8314618947246325e-05}
{"train_loss": 0.16583117842674255, "global_step": 121566, "epoch": 1365, "lr": 5.8314047282052295e-05}
{"train_loss": 0.1057809442281723, "global_step": 121567, "epoch": 1365, "lr": 5.831347561574057e-05}
{"train_loss": 0.11608625203371048, "global_step": 121568, "epoch": 1365, "lr": 5.831290394831118e-05}
{"train_loss": 0.14656275510787964, "global_step": 121569, "epoch": 1365, "lr": 5.831233227976422e-05}
{"train_loss": 0.2728402316570282, "global_step": 121570, "epoch": 1365, "lr": 5.831176061009978e-05}
{"train_loss": 0.10237901657819748, "global_step": 121571, "epoch": 1365, "lr": 5.831118893931792e-05}
{"train_loss": 0.10877716541290283, "global_step": 121572, "epoch": 1365, "lr": 5.831061726741872e-05}
{"train_loss": 0.14075882423124955, "global_step": 121573, "epoch": 1365, "lr": 5.831004559440225e-05, "val_loss": 5.428382873535156, "train_action_mse_error": 6.740265846252441}
{"train_loss": 0.1279398500919342, "global_step": 121574, "epoch": 1366, "lr": 5.830947392026861e-05}
{"train_loss": 0.12479002773761749, "global_step": 121575, "epoch": 1366, "lr": 5.830890224501786e-05}
{"train_loss": 0.08597707748413086, "global_step": 121576, "epoch": 1366, "lr": 5.830833056865008e-05}
{"train_loss": 0.16435769200325012, "global_step": 121577, "epoch": 1366, "lr": 5.8307758891165345e-05}
{"train_loss": 0.1575278639793396, "global_step": 121578, "epoch": 1366, "lr": 5.830718721256372e-05}
{"train_loss": 0.10631788522005081, "global_step": 121579, "epoch": 1366, "lr": 5.8306615532845296e-05}
{"train_loss": 0.14270703494548798, "global_step": 121580, "epoch": 1366, "lr": 5.830604385201017e-05}
{"train_loss": 0.13648417592048645, "global_step": 121581, "epoch": 1366, "lr": 5.8305472170058394e-05}
{"train_loss": 0.06290902197360992, "global_step": 121582, "epoch": 1366, "lr": 5.8304900486990044e-05}
{"train_loss": 0.10155165940523148, "global_step": 121583, "epoch": 1366, "lr": 5.8304328802805196e-05}
{"train_loss": 0.08993983268737793, "global_step": 121584, "epoch": 1366, "lr": 5.830375711750393e-05}
{"train_loss": 0.14626114070415497, "global_step": 121585, "epoch": 1366, "lr": 5.830318543108634e-05}
{"train_loss": 0.11062733829021454, "global_step": 121586, "epoch": 1366, "lr": 5.830261374355247e-05}
{"train_loss": 0.12364037334918976, "global_step": 121587, "epoch": 1366, "lr": 5.8302042054902426e-05}
{"train_loss": 0.14940950274467468, "global_step": 121588, "epoch": 1366, "lr": 5.830147036513627e-05}
{"train_loss": 0.13719893991947174, "global_step": 121589, "epoch": 1366, "lr": 5.830089867425408e-05}
{"train_loss": 0.1542576104402542, "global_step": 121590, "epoch": 1366, "lr": 5.8300326982255936e-05}
{"train_loss": 0.12555797398090363, "global_step": 121591, "epoch": 1366, "lr": 5.8299755289141924e-05}
{"train_loss": 0.15192003548145294, "global_step": 121592, "epoch": 1366, "lr": 5.82991835949121e-05}
{"train_loss": 0.14099128544330597, "global_step": 121593, "epoch": 1366, "lr": 5.8298611899566555e-05}
{"train_loss": 0.13146157562732697, "global_step": 121594, "epoch": 1366, "lr": 5.829804020310536e-05}
{"train_loss": 0.13054868578910828, "global_step": 121595, "epoch": 1366, "lr": 5.82974685055286e-05}
{"train_loss": 0.12264735251665115, "global_step": 121596, "epoch": 1366, "lr": 5.829689680683633e-05}
{"train_loss": 0.11432809382677078, "global_step": 121597, "epoch": 1366, "lr": 5.829632510702866e-05}
{"train_loss": 0.13079744577407837, "global_step": 121598, "epoch": 1366, "lr": 5.829575340610565e-05}
{"train_loss": 0.12022393941879272, "global_step": 121599, "epoch": 1366, "lr": 5.829518170406737e-05}
{"train_loss": 0.08252143114805222, "global_step": 121600, "epoch": 1366, "lr": 5.829461000091392e-05}
{"train_loss": 0.12833374738693237, "global_step": 121601, "epoch": 1366, "lr": 5.829403829664535e-05}
{"train_loss": 0.1410088688135147, "global_step": 121602, "epoch": 1366, "lr": 5.829346659126175e-05}
{"train_loss": 0.1588931679725647, "global_step": 121603, "epoch": 1366, "lr": 5.829289488476319e-05}
{"train_loss": 0.12670572102069855, "global_step": 121604, "epoch": 1366, "lr": 5.829232317714975e-05}
{"train_loss": 0.10402915626764297, "global_step": 121605, "epoch": 1366, "lr": 5.829175146842152e-05}
{"train_loss": 0.10737492144107819, "global_step": 121606, "epoch": 1366, "lr": 5.8291179758578555e-05}
{"train_loss": 0.15958453714847565, "global_step": 121607, "epoch": 1366, "lr": 5.829060804762094e-05}
{"train_loss": 0.07143144309520721, "global_step": 121608, "epoch": 1366, "lr": 5.829003633554876e-05}
{"train_loss": 0.15261423587799072, "global_step": 121609, "epoch": 1366, "lr": 5.828946462236208e-05}
{"train_loss": 0.06843828409910202, "global_step": 121610, "epoch": 1366, "lr": 5.828889290806099e-05}
{"train_loss": 0.13241063058376312, "global_step": 121611, "epoch": 1366, "lr": 5.8288321192645555e-05}
{"train_loss": 0.13139866292476654, "global_step": 121612, "epoch": 1366, "lr": 5.828774947611585e-05}
{"train_loss": 0.17096766829490662, "global_step": 121613, "epoch": 1366, "lr": 5.828717775847197e-05}
{"train_loss": 0.19361896812915802, "global_step": 121614, "epoch": 1366, "lr": 5.828660603971396e-05}
{"train_loss": 0.1930459886789322, "global_step": 121615, "epoch": 1366, "lr": 5.828603431984193e-05}
{"train_loss": 0.1922418624162674, "global_step": 121616, "epoch": 1366, "lr": 5.8285462598855945e-05}
{"train_loss": 0.14484019577503204, "global_step": 121617, "epoch": 1366, "lr": 5.828489087675608e-05}
{"train_loss": 0.1352461725473404, "global_step": 121618, "epoch": 1366, "lr": 5.8284319153542413e-05}
{"train_loss": 0.1739356517791748, "global_step": 121619, "epoch": 1366, "lr": 5.828374742921503e-05}
{"train_loss": 0.11710755527019501, "global_step": 121620, "epoch": 1366, "lr": 5.828317570377398e-05}
{"train_loss": 0.1451520025730133, "global_step": 121621, "epoch": 1366, "lr": 5.8282603977219375e-05}
{"train_loss": 0.141129732131958, "global_step": 121622, "epoch": 1366, "lr": 5.8282032249551264e-05}
{"train_loss": 0.1134045347571373, "global_step": 121623, "epoch": 1366, "lr": 5.828146052076974e-05}
{"train_loss": 0.16040843725204468, "global_step": 121624, "epoch": 1366, "lr": 5.8280888790874866e-05}
{"train_loss": 0.08030425012111664, "global_step": 121625, "epoch": 1366, "lr": 5.828031705986673e-05}
{"train_loss": 0.16520270705223083, "global_step": 121626, "epoch": 1366, "lr": 5.8279745327745406e-05}
{"train_loss": 0.12246948480606079, "global_step": 121627, "epoch": 1366, "lr": 5.827917359451098e-05}
{"train_loss": 0.13698942959308624, "global_step": 121628, "epoch": 1366, "lr": 5.827860186016352e-05}
{"train_loss": 0.16765566170215607, "global_step": 121629, "epoch": 1366, "lr": 5.8278030124703086e-05}
{"train_loss": 0.11026709526777267, "global_step": 121630, "epoch": 1366, "lr": 5.827745838812979e-05}
{"train_loss": 0.15533393621444702, "global_step": 121631, "epoch": 1366, "lr": 5.827688665044369e-05}
{"train_loss": 0.12244030833244324, "global_step": 121632, "epoch": 1366, "lr": 5.827631491164484e-05}
{"train_loss": 0.17700330913066864, "global_step": 121633, "epoch": 1366, "lr": 5.8275743171733354e-05}
{"train_loss": 0.17818337678909302, "global_step": 121634, "epoch": 1366, "lr": 5.8275171430709296e-05}
{"train_loss": 0.09081335365772247, "global_step": 121635, "epoch": 1366, "lr": 5.827459968857274e-05}
{"train_loss": 0.12187483161687851, "global_step": 121636, "epoch": 1366, "lr": 5.8274027945323774e-05}
{"train_loss": 0.13259854912757874, "global_step": 121637, "epoch": 1366, "lr": 5.827345620096246e-05}
{"train_loss": 0.10585791617631912, "global_step": 121638, "epoch": 1366, "lr": 5.827288445548888e-05}
{"train_loss": 0.0979907363653183, "global_step": 121639, "epoch": 1366, "lr": 5.827231270890312e-05}
{"train_loss": 0.14559444785118103, "global_step": 121640, "epoch": 1366, "lr": 5.8271740961205246e-05}
{"train_loss": 0.11634958535432816, "global_step": 121641, "epoch": 1366, "lr": 5.827116921239533e-05}
{"train_loss": 0.17457138001918793, "global_step": 121642, "epoch": 1366, "lr": 5.8270597462473466e-05}
{"train_loss": 0.12298348546028137, "global_step": 121643, "epoch": 1366, "lr": 5.8270025711439714e-05}
{"train_loss": 0.10916367173194885, "global_step": 121644, "epoch": 1366, "lr": 5.8269453959294154e-05}
{"train_loss": 0.2032577246427536, "global_step": 121645, "epoch": 1366, "lr": 5.8268882206036866e-05}
{"train_loss": 0.13485465943813324, "global_step": 121646, "epoch": 1366, "lr": 5.826831045166794e-05}
{"train_loss": 0.1360141932964325, "global_step": 121647, "epoch": 1366, "lr": 5.8267738696187426e-05}
{"train_loss": 0.09328710287809372, "global_step": 121648, "epoch": 1366, "lr": 5.826716693959542e-05}
{"train_loss": 0.20803730189800262, "global_step": 121649, "epoch": 1366, "lr": 5.826659518189199e-05}
{"train_loss": 0.11945793777704239, "global_step": 121650, "epoch": 1366, "lr": 5.826602342307722e-05}
{"train_loss": 0.23617976903915405, "global_step": 121651, "epoch": 1366, "lr": 5.826545166315118e-05}
{"train_loss": 0.1432560533285141, "global_step": 121652, "epoch": 1366, "lr": 5.826487990211397e-05}
{"train_loss": 0.14290368556976318, "global_step": 121653, "epoch": 1366, "lr": 5.826430813996563e-05}
{"train_loss": 0.06514769792556763, "global_step": 121654, "epoch": 1366, "lr": 5.8263736376706256e-05}
{"train_loss": 0.15573374927043915, "global_step": 121655, "epoch": 1366, "lr": 5.826316461233593e-05}
{"train_loss": 0.13242904841899872, "global_step": 121656, "epoch": 1366, "lr": 5.8262592846854716e-05}
{"train_loss": 0.1110035628080368, "global_step": 121657, "epoch": 1366, "lr": 5.826202108026271e-05}
{"train_loss": 0.14022132754325867, "global_step": 121658, "epoch": 1366, "lr": 5.826144931255997e-05}
{"train_loss": 0.1902468502521515, "global_step": 121659, "epoch": 1366, "lr": 5.826087754374657e-05}
{"train_loss": 0.10361530631780624, "global_step": 121660, "epoch": 1366, "lr": 5.82603057738226e-05}
{"train_loss": 0.19537712633609772, "global_step": 121661, "epoch": 1366, "lr": 5.8259734002788135e-05}
{"train_loss": 0.13485279587212573, "global_step": 121662, "epoch": 1366, "lr": 5.825916223064325e-05, "val_loss": 5.5478692054748535}
{"train_loss": 0.09310771524906158, "global_step": 121663, "epoch": 1367, "lr": 5.825859045738802e-05}
{"train_loss": 0.14400944113731384, "global_step": 121664, "epoch": 1367, "lr": 5.825801868302252e-05}
{"train_loss": 0.18079310655593872, "global_step": 121665, "epoch": 1367, "lr": 5.825744690754683e-05}
{"train_loss": 0.22999373078346252, "global_step": 121666, "epoch": 1367, "lr": 5.825687513096103e-05}
{"train_loss": 0.13662324845790863, "global_step": 121667, "epoch": 1367, "lr": 5.8256303353265196e-05}
{"train_loss": 0.15117913484573364, "global_step": 121668, "epoch": 1367, "lr": 5.82557315744594e-05}
{"train_loss": 0.1772778332233429, "global_step": 121669, "epoch": 1367, "lr": 5.825515979454371e-05}
{"train_loss": 0.0738028958439827, "global_step": 121670, "epoch": 1367, "lr": 5.825458801351823e-05}
{"train_loss": 0.07815816253423691, "global_step": 121671, "epoch": 1367, "lr": 5.825401623138301e-05}
{"train_loss": 0.08750785142183304, "global_step": 121672, "epoch": 1367, "lr": 5.825344444813815e-05}
{"train_loss": 0.15568038821220398, "global_step": 121673, "epoch": 1367, "lr": 5.825287266378371e-05}
{"train_loss": 0.09554325044155121, "global_step": 121674, "epoch": 1367, "lr": 5.825230087831977e-05}
{"train_loss": 0.10093273967504501, "global_step": 121675, "epoch": 1367, "lr": 5.825172909174641e-05}
{"train_loss": 0.15793904662132263, "global_step": 121676, "epoch": 1367, "lr": 5.825115730406371e-05}
{"train_loss": 0.1452951729297638, "global_step": 121677, "epoch": 1367, "lr": 5.825058551527174e-05}
{"train_loss": 0.15658026933670044, "global_step": 121678, "epoch": 1367, "lr": 5.825001372537058e-05}
{"train_loss": 0.13621856272220612, "global_step": 121679, "epoch": 1367, "lr": 5.8249441934360307e-05}
{"train_loss": 0.11743645370006561, "global_step": 121680, "epoch": 1367, "lr": 5.8248870142241e-05}
{"train_loss": 0.11626047641038895, "global_step": 121681, "epoch": 1367, "lr": 5.824829834901272e-05}
{"train_loss": 0.13877463340759277, "global_step": 121682, "epoch": 1367, "lr": 5.824772655467556e-05}
{"train_loss": 0.111290343105793, "global_step": 121683, "epoch": 1367, "lr": 5.824715475922961e-05}
{"train_loss": 0.12626062333583832, "global_step": 121684, "epoch": 1367, "lr": 5.824658296267491e-05}
{"train_loss": 0.13854415714740753, "global_step": 121685, "epoch": 1367, "lr": 5.824601116501156e-05}
{"train_loss": 0.17599695920944214, "global_step": 121686, "epoch": 1367, "lr": 5.824543936623964e-05}
{"train_loss": 0.1485491245985031, "global_step": 121687, "epoch": 1367, "lr": 5.824486756635922e-05}
{"train_loss": 0.14391210675239563, "global_step": 121688, "epoch": 1367, "lr": 5.824429576537037e-05}
{"train_loss": 0.10503709316253662, "global_step": 121689, "epoch": 1367, "lr": 5.824372396327319e-05}
{"train_loss": 0.09362977743148804, "global_step": 121690, "epoch": 1367, "lr": 5.824315216006774e-05}
{"train_loss": 0.13698984682559967, "global_step": 121691, "epoch": 1367, "lr": 5.8242580355754084e-05}
{"train_loss": 0.0946681797504425, "global_step": 121692, "epoch": 1367, "lr": 5.8242008550332326e-05}
{"train_loss": 0.1443299502134323, "global_step": 121693, "epoch": 1367, "lr": 5.8241436743802534e-05}
{"train_loss": 0.18602438271045685, "global_step": 121694, "epoch": 1367, "lr": 5.8240864936164776e-05}
{"train_loss": 0.1527142971754074, "global_step": 121695, "epoch": 1367, "lr": 5.824029312741913e-05}
{"train_loss": 0.10426592826843262, "global_step": 121696, "epoch": 1367, "lr": 5.823972131756568e-05}
{"train_loss": 0.1552150398492813, "global_step": 121697, "epoch": 1367, "lr": 5.823914950660452e-05}
{"train_loss": 0.14302988350391388, "global_step": 121698, "epoch": 1367, "lr": 5.8238577694535675e-05}
{"train_loss": 0.07845228165388107, "global_step": 121699, "epoch": 1367, "lr": 5.823800588135927e-05}
{"train_loss": 0.16991256177425385, "global_step": 121700, "epoch": 1367, "lr": 5.823743406707536e-05}
{"train_loss": 0.0944206640124321, "global_step": 121701, "epoch": 1367, "lr": 5.823686225168403e-05}
{"train_loss": 0.1665351688861847, "global_step": 121702, "epoch": 1367, "lr": 5.823629043518536e-05}
{"train_loss": 0.1333632916212082, "global_step": 121703, "epoch": 1367, "lr": 5.823571861757942e-05}
{"train_loss": 0.1345852166414261, "global_step": 121704, "epoch": 1367, "lr": 5.823514679886627e-05}
{"train_loss": 0.18803180754184723, "global_step": 121705, "epoch": 1367, "lr": 5.823457497904601e-05}
{"train_loss": 0.17954160273075104, "global_step": 121706, "epoch": 1367, "lr": 5.8234003158118723e-05}
{"train_loss": 0.15128566324710846, "global_step": 121707, "epoch": 1367, "lr": 5.823343133608447e-05}
{"train_loss": 0.1792726218700409, "global_step": 121708, "epoch": 1367, "lr": 5.823285951294333e-05}
{"train_loss": 0.16601844131946564, "global_step": 121709, "epoch": 1367, "lr": 5.8232287688695384e-05}
{"train_loss": 0.15573285520076752, "global_step": 121710, "epoch": 1367, "lr": 5.823171586334071e-05}
{"train_loss": 0.20445336401462555, "global_step": 121711, "epoch": 1367, "lr": 5.823114403687938e-05}
{"train_loss": 0.10240539908409119, "global_step": 121712, "epoch": 1367, "lr": 5.823057220931147e-05}
{"train_loss": 0.20103196799755096, "global_step": 121713, "epoch": 1367, "lr": 5.823000038063706e-05}
{"train_loss": 0.18032647669315338, "global_step": 121714, "epoch": 1367, "lr": 5.8229428550856234e-05}
{"train_loss": 0.18401525914669037, "global_step": 121715, "epoch": 1367, "lr": 5.822885671996906e-05}
{"train_loss": 0.1626931130886078, "global_step": 121716, "epoch": 1367, "lr": 5.822828488797563e-05}
{"train_loss": 0.19361138343811035, "global_step": 121717, "epoch": 1367, "lr": 5.8227713054875985e-05}
{"train_loss": 0.17647898197174072, "global_step": 121718, "epoch": 1367, "lr": 5.822714122067023e-05}
{"train_loss": 0.09678363800048828, "global_step": 121719, "epoch": 1367, "lr": 5.8226569385358443e-05}
{"train_loss": 0.14140239357948303, "global_step": 121720, "epoch": 1367, "lr": 5.8225997548940694e-05}
{"train_loss": 0.1733810305595398, "global_step": 121721, "epoch": 1367, "lr": 5.8225425711417045e-05}
{"train_loss": 0.17687059938907623, "global_step": 121722, "epoch": 1367, "lr": 5.82248538727876e-05}
{"train_loss": 0.18043434619903564, "global_step": 121723, "epoch": 1367, "lr": 5.822428203305241e-05}
{"train_loss": 0.1420016884803772, "global_step": 121724, "epoch": 1367, "lr": 5.822371019221158e-05}
{"train_loss": 0.12109661102294922, "global_step": 121725, "epoch": 1367, "lr": 5.822313835026517e-05}
{"train_loss": 0.1448196917772293, "global_step": 121726, "epoch": 1367, "lr": 5.8222566507213264e-05}
{"train_loss": 0.1519085019826889, "global_step": 121727, "epoch": 1367, "lr": 5.8221994663055924e-05}
{"train_loss": 0.16863249242305756, "global_step": 121728, "epoch": 1367, "lr": 5.822142281779324e-05}
{"train_loss": 0.1337980479001999, "global_step": 121729, "epoch": 1367, "lr": 5.822085097142529e-05}
{"train_loss": 0.13515867292881012, "global_step": 121730, "epoch": 1367, "lr": 5.8220279123952137e-05}
{"train_loss": 0.12464215606451035, "global_step": 121731, "epoch": 1367, "lr": 5.821970727537388e-05}
{"train_loss": 0.1817876249551773, "global_step": 121732, "epoch": 1367, "lr": 5.821913542569057e-05}
{"train_loss": 0.10760223865509033, "global_step": 121733, "epoch": 1367, "lr": 5.8218563574902306e-05}
{"train_loss": 0.14461007714271545, "global_step": 121734, "epoch": 1367, "lr": 5.8217991723009156e-05}
{"train_loss": 0.22696706652641296, "global_step": 121735, "epoch": 1367, "lr": 5.821741987001121e-05}
{"train_loss": 0.16547203063964844, "global_step": 121736, "epoch": 1367, "lr": 5.821684801590852e-05}
{"train_loss": 0.08993767201900482, "global_step": 121737, "epoch": 1367, "lr": 5.821627616070118e-05}
{"train_loss": 0.1787969172000885, "global_step": 121738, "epoch": 1367, "lr": 5.821570430438925e-05}
{"train_loss": 0.15346308052539825, "global_step": 121739, "epoch": 1367, "lr": 5.8215132446972834e-05}
{"train_loss": 0.17142528295516968, "global_step": 121740, "epoch": 1367, "lr": 5.821456058845197e-05}
{"train_loss": 0.11433829367160797, "global_step": 121741, "epoch": 1367, "lr": 5.821398872882677e-05}
{"train_loss": 0.16732102632522583, "global_step": 121742, "epoch": 1367, "lr": 5.8213416868097306e-05}
{"train_loss": 0.1187431812286377, "global_step": 121743, "epoch": 1367, "lr": 5.821284500626364e-05}
{"train_loss": 0.14943799376487732, "global_step": 121744, "epoch": 1367, "lr": 5.821227314332586e-05}
{"train_loss": 0.110868439078331, "global_step": 121745, "epoch": 1367, "lr": 5.8211701279284035e-05}
{"train_loss": 0.13919048011302948, "global_step": 121746, "epoch": 1367, "lr": 5.8211129414138255e-05}
{"train_loss": 0.11734526604413986, "global_step": 121747, "epoch": 1367, "lr": 5.821055754788859e-05}
{"train_loss": 0.10531996935606003, "global_step": 121748, "epoch": 1367, "lr": 5.8209985680535115e-05}
{"train_loss": 0.1280198097229004, "global_step": 121749, "epoch": 1367, "lr": 5.8209413812077904e-05}
{"train_loss": 0.1999269276857376, "global_step": 121750, "epoch": 1367, "lr": 5.8208841942517036e-05}
{"train_loss": 0.14427424774745876, "global_step": 121751, "epoch": 1367, "lr": 5.8208270071852586e-05, "val_loss": 5.303862571716309}
{"train_loss": 0.08663348108530045, "global_step": 121752, "epoch": 1368, "lr": 5.820769820008465e-05}
{"train_loss": 0.13474655151367188, "global_step": 121753, "epoch": 1368, "lr": 5.820712632721328e-05}
{"train_loss": 0.16739025712013245, "global_step": 121754, "epoch": 1368, "lr": 5.820655445323856e-05}
{"train_loss": 0.0881054550409317, "global_step": 121755, "epoch": 1368, "lr": 5.8205982578160575e-05}
{"train_loss": 0.10568893700838089, "global_step": 121756, "epoch": 1368, "lr": 5.820541070197939e-05}
{"train_loss": 0.20688115060329437, "global_step": 121757, "epoch": 1368, "lr": 5.820483882469508e-05}
{"train_loss": 0.09778500348329544, "global_step": 121758, "epoch": 1368, "lr": 5.820426694630774e-05}
{"train_loss": 0.11438356339931488, "global_step": 121759, "epoch": 1368, "lr": 5.8203695066817435e-05}
{"train_loss": 0.15000630915164948, "global_step": 121760, "epoch": 1368, "lr": 5.820312318622423e-05}
{"train_loss": 0.20009052753448486, "global_step": 121761, "epoch": 1368, "lr": 5.820255130452823e-05}
{"train_loss": 0.0895313248038292, "global_step": 121762, "epoch": 1368, "lr": 5.820197942172949e-05}
{"train_loss": 0.09330052137374878, "global_step": 121763, "epoch": 1368, "lr": 5.8201407537828104e-05}
{"train_loss": 0.09508621692657471, "global_step": 121764, "epoch": 1368, "lr": 5.8200835652824124e-05}
{"train_loss": 0.12920764088630676, "global_step": 121765, "epoch": 1368, "lr": 5.8200263766717654e-05}
{"train_loss": 0.18820995092391968, "global_step": 121766, "epoch": 1368, "lr": 5.819969187950875e-05}
{"train_loss": 0.08827698230743408, "global_step": 121767, "epoch": 1368, "lr": 5.81991199911975e-05}
{"train_loss": 0.12075557559728622, "global_step": 121768, "epoch": 1368, "lr": 5.819854810178399e-05}
{"train_loss": 0.1829342097043991, "global_step": 121769, "epoch": 1368, "lr": 5.819797621126827e-05}
{"train_loss": 0.13847529888153076, "global_step": 121770, "epoch": 1368, "lr": 5.819740431965044e-05}
{"train_loss": 0.1184401661157608, "global_step": 121771, "epoch": 1368, "lr": 5.819683242693056e-05}
{"train_loss": 0.12836572527885437, "global_step": 121772, "epoch": 1368, "lr": 5.819626053310873e-05}
{"train_loss": 0.13502627611160278, "global_step": 121773, "epoch": 1368, "lr": 5.8195688638185e-05}
{"train_loss": 0.11237745732069016, "global_step": 121774, "epoch": 1368, "lr": 5.819511674215947e-05}
{"train_loss": 0.12450064718723297, "global_step": 121775, "epoch": 1368, "lr": 5.819454484503221e-05}
{"train_loss": 0.15224838256835938, "global_step": 121776, "epoch": 1368, "lr": 5.819397294680328e-05}
{"train_loss": 0.1421707570552826, "global_step": 121777, "epoch": 1368, "lr": 5.819340104747276e-05}
{"train_loss": 0.09428329765796661, "global_step": 121778, "epoch": 1368, "lr": 5.819282914704075e-05}
{"train_loss": 0.2154337763786316, "global_step": 121779, "epoch": 1368, "lr": 5.819225724550732e-05}
{"train_loss": 0.10225602984428406, "global_step": 121780, "epoch": 1368, "lr": 5.819168534287254e-05}
{"train_loss": 0.14088186621665955, "global_step": 121781, "epoch": 1368, "lr": 5.819111343913648e-05}
{"train_loss": 0.07068797945976257, "global_step": 121782, "epoch": 1368, "lr": 5.8190541534299225e-05}
{"train_loss": 0.13881799578666687, "global_step": 121783, "epoch": 1368, "lr": 5.8189969628360854e-05}
{"train_loss": 0.16192196309566498, "global_step": 121784, "epoch": 1368, "lr": 5.8189397721321434e-05}
{"train_loss": 0.07489912211894989, "global_step": 121785, "epoch": 1368, "lr": 5.818882581318106e-05}
{"train_loss": 0.12676823139190674, "global_step": 121786, "epoch": 1368, "lr": 5.8188253903939803e-05}
{"train_loss": 0.13726916909217834, "global_step": 121787, "epoch": 1368, "lr": 5.818768199359772e-05}
{"train_loss": 0.18182386457920074, "global_step": 121788, "epoch": 1368, "lr": 5.8187110082154916e-05}
{"train_loss": 0.10098671168088913, "global_step": 121789, "epoch": 1368, "lr": 5.818653816961145e-05}
{"train_loss": 0.11945363879203796, "global_step": 121790, "epoch": 1368, "lr": 5.8185966255967395e-05}
{"train_loss": 0.13165411353111267, "global_step": 121791, "epoch": 1368, "lr": 5.8185394341222856e-05}
{"train_loss": 0.13957151770591736, "global_step": 121792, "epoch": 1368, "lr": 5.818482242537787e-05}
{"train_loss": 0.18547461926937103, "global_step": 121793, "epoch": 1368, "lr": 5.818425050843256e-05}
{"train_loss": 0.125631183385849, "global_step": 121794, "epoch": 1368, "lr": 5.818367859038696e-05}
{"train_loss": 0.11508959531784058, "global_step": 121795, "epoch": 1368, "lr": 5.8183106671241164e-05}
{"train_loss": 0.1624603122472763, "global_step": 121796, "epoch": 1368, "lr": 5.818253475099524e-05}
{"train_loss": 0.07377626746892929, "global_step": 121797, "epoch": 1368, "lr": 5.8181962829649286e-05}
{"train_loss": 0.20199395716190338, "global_step": 121798, "epoch": 1368, "lr": 5.818139090720336e-05}
{"train_loss": 0.18500550091266632, "global_step": 121799, "epoch": 1368, "lr": 5.8180818983657545e-05}
{"train_loss": 0.12800168991088867, "global_step": 121800, "epoch": 1368, "lr": 5.818024705901193e-05}
{"train_loss": 0.1170191541314125, "global_step": 121801, "epoch": 1368, "lr": 5.817967513326657e-05}
{"train_loss": 0.09641347825527191, "global_step": 121802, "epoch": 1368, "lr": 5.817910320642156e-05}
{"train_loss": 0.13829141855239868, "global_step": 121803, "epoch": 1368, "lr": 5.8178531278476966e-05}
{"train_loss": 0.15075437724590302, "global_step": 121804, "epoch": 1368, "lr": 5.8177959349432863e-05}
{"train_loss": 0.11491470038890839, "global_step": 121805, "epoch": 1368, "lr": 5.817738741928934e-05}
{"train_loss": 0.11750920116901398, "global_step": 121806, "epoch": 1368, "lr": 5.8176815488046475e-05}
{"train_loss": 0.16992934048175812, "global_step": 121807, "epoch": 1368, "lr": 5.8176243555704325e-05}
{"train_loss": 0.14546163380146027, "global_step": 121808, "epoch": 1368, "lr": 5.817567162226298e-05}
{"train_loss": 0.1222316175699234, "global_step": 121809, "epoch": 1368, "lr": 5.817509968772251e-05}
{"train_loss": 0.12679792940616608, "global_step": 121810, "epoch": 1368, "lr": 5.817452775208301e-05}
{"train_loss": 0.17210592329502106, "global_step": 121811, "epoch": 1368, "lr": 5.8173955815344536e-05}
{"train_loss": 0.10498608648777008, "global_step": 121812, "epoch": 1368, "lr": 5.817338387750718e-05}
{"train_loss": 0.1258346289396286, "global_step": 121813, "epoch": 1368, "lr": 5.8172811938571015e-05}
{"train_loss": 0.19906039535999298, "global_step": 121814, "epoch": 1368, "lr": 5.817223999853609e-05}
{"train_loss": 0.14736610651016235, "global_step": 121815, "epoch": 1368, "lr": 5.817166805740253e-05}
{"train_loss": 0.10189206153154373, "global_step": 121816, "epoch": 1368, "lr": 5.8171096115170384e-05}
{"train_loss": 0.1919434368610382, "global_step": 121817, "epoch": 1368, "lr": 5.8170524171839736e-05}
{"train_loss": 0.2133980393409729, "global_step": 121818, "epoch": 1368, "lr": 5.816995222741066e-05}
{"train_loss": 0.1318691074848175, "global_step": 121819, "epoch": 1368, "lr": 5.816938028188323e-05}
{"train_loss": 0.09879600256681442, "global_step": 121820, "epoch": 1368, "lr": 5.816880833525753e-05}
{"train_loss": 0.07856570184230804, "global_step": 121821, "epoch": 1368, "lr": 5.816823638753363e-05}
{"train_loss": 0.11989568173885345, "global_step": 121822, "epoch": 1368, "lr": 5.816766443871161e-05}
{"train_loss": 0.13685427606105804, "global_step": 121823, "epoch": 1368, "lr": 5.8167092488791556e-05}
{"train_loss": 0.12078652530908585, "global_step": 121824, "epoch": 1368, "lr": 5.8166520537773515e-05}
{"train_loss": 0.16193093359470367, "global_step": 121825, "epoch": 1368, "lr": 5.8165948585657604e-05}
{"train_loss": 0.13443413376808167, "global_step": 121826, "epoch": 1368, "lr": 5.816537663244387e-05}
{"train_loss": 0.15571564435958862, "global_step": 121827, "epoch": 1368, "lr": 5.816480467813241e-05}
{"train_loss": 0.1158471629023552, "global_step": 121828, "epoch": 1368, "lr": 5.816423272272329e-05}
{"train_loss": 0.07698624581098557, "global_step": 121829, "epoch": 1368, "lr": 5.8163660766216576e-05}
{"train_loss": 0.22224946320056915, "global_step": 121830, "epoch": 1368, "lr": 5.816308880861238e-05}
{"train_loss": 0.1339416652917862, "global_step": 121831, "epoch": 1368, "lr": 5.816251684991073e-05}
{"train_loss": 0.11492232978343964, "global_step": 121832, "epoch": 1368, "lr": 5.816194489011174e-05}
{"train_loss": 0.07943951338529587, "global_step": 121833, "epoch": 1368, "lr": 5.8161372929215484e-05}
{"train_loss": 0.15924802422523499, "global_step": 121834, "epoch": 1368, "lr": 5.816080096722203e-05}
{"train_loss": 0.10970059037208557, "global_step": 121835, "epoch": 1368, "lr": 5.816022900413145e-05}
{"train_loss": 0.11576373130083084, "global_step": 121836, "epoch": 1368, "lr": 5.8159657039943825e-05}
{"train_loss": 0.12385398149490356, "global_step": 121837, "epoch": 1368, "lr": 5.8159085074659235e-05}
{"train_loss": 0.1107020378112793, "global_step": 121838, "epoch": 1368, "lr": 5.815851310827776e-05}
{"train_loss": 0.15624085068702698, "global_step": 121839, "epoch": 1368, "lr": 5.815794114079947e-05}
{"train_loss": 0.13310575317800716, "global_step": 121840, "epoch": 1368, "lr": 5.815736917222444e-05, "val_loss": 5.544785499572754}
{"train_loss": 0.08695528656244278, "global_step": 121841, "epoch": 1369, "lr": 5.815679720255275e-05}
{"train_loss": 0.16152434051036835, "global_step": 121842, "epoch": 1369, "lr": 5.815622523178448e-05}
{"train_loss": 0.17622563242912292, "global_step": 121843, "epoch": 1369, "lr": 5.8155653259919704e-05}
{"train_loss": 0.17686061561107635, "global_step": 121844, "epoch": 1369, "lr": 5.815508128695851e-05}
{"train_loss": 0.12729652225971222, "global_step": 121845, "epoch": 1369, "lr": 5.815450931290096e-05}
{"train_loss": 0.11751429736614227, "global_step": 121846, "epoch": 1369, "lr": 5.815393733774713e-05}
{"train_loss": 0.1937468945980072, "global_step": 121847, "epoch": 1369, "lr": 5.81533653614971e-05}
{"train_loss": 0.1442786008119583, "global_step": 121848, "epoch": 1369, "lr": 5.815279338415096e-05}
{"train_loss": 0.08636057376861572, "global_step": 121849, "epoch": 1369, "lr": 5.815222140570877e-05}
{"train_loss": 0.15590637922286987, "global_step": 121850, "epoch": 1369, "lr": 5.8151649426170616e-05}
{"train_loss": 0.17851823568344116, "global_step": 121851, "epoch": 1369, "lr": 5.8151077445536576e-05}
{"train_loss": 0.1714271456003189, "global_step": 121852, "epoch": 1369, "lr": 5.815050546380671e-05}
{"train_loss": 0.08920471370220184, "global_step": 121853, "epoch": 1369, "lr": 5.814993348098112e-05}
{"train_loss": 0.09923982620239258, "global_step": 121854, "epoch": 1369, "lr": 5.814936149705986e-05}
{"train_loss": 0.17250925302505493, "global_step": 121855, "epoch": 1369, "lr": 5.8148789512043036e-05}
{"train_loss": 0.14187584817409515, "global_step": 121856, "epoch": 1369, "lr": 5.814821752593069e-05}
{"train_loss": 0.11668815463781357, "global_step": 121857, "epoch": 1369, "lr": 5.814764553872292e-05}
{"train_loss": 0.13643017411231995, "global_step": 121858, "epoch": 1369, "lr": 5.8147073550419794e-05}
{"train_loss": 0.1000276729464531, "global_step": 121859, "epoch": 1369, "lr": 5.8146501561021396e-05}
{"train_loss": 0.2284795492887497, "global_step": 121860, "epoch": 1369, "lr": 5.8145929570527804e-05}
{"train_loss": 0.10421883314847946, "global_step": 121861, "epoch": 1369, "lr": 5.814535757893909e-05}
{"train_loss": 0.05862567573785782, "global_step": 121862, "epoch": 1369, "lr": 5.814478558625533e-05}
{"train_loss": 0.17202064394950867, "global_step": 121863, "epoch": 1369, "lr": 5.81442135924766e-05}
{"train_loss": 0.1914878785610199, "global_step": 121864, "epoch": 1369, "lr": 5.814364159760299e-05}
{"train_loss": 0.15207774937152863, "global_step": 121865, "epoch": 1369, "lr": 5.814306960163456e-05}
{"train_loss": 0.1743609756231308, "global_step": 121866, "epoch": 1369, "lr": 5.814249760457139e-05}
{"train_loss": 0.19515761733055115, "global_step": 121867, "epoch": 1369, "lr": 5.814192560641356e-05}
{"train_loss": 0.14671193063259125, "global_step": 121868, "epoch": 1369, "lr": 5.8141353607161155e-05}
{"train_loss": 0.1700851023197174, "global_step": 121869, "epoch": 1369, "lr": 5.8140781606814246e-05}
{"train_loss": 0.15254762768745422, "global_step": 121870, "epoch": 1369, "lr": 5.8140209605372896e-05}
{"train_loss": 0.14107881486415863, "global_step": 121871, "epoch": 1369, "lr": 5.813963760283721e-05}
{"train_loss": 0.10382889211177826, "global_step": 121872, "epoch": 1369, "lr": 5.813906559920723e-05}
{"train_loss": 0.08142542839050293, "global_step": 121873, "epoch": 1369, "lr": 5.813849359448307e-05}
{"train_loss": 0.1350422352552414, "global_step": 121874, "epoch": 1369, "lr": 5.813792158866478e-05}
{"train_loss": 0.13580001890659332, "global_step": 121875, "epoch": 1369, "lr": 5.8137349581752444e-05}
{"train_loss": 0.1092776358127594, "global_step": 121876, "epoch": 1369, "lr": 5.8136777573746135e-05}
{"train_loss": 0.1287669688463211, "global_step": 121877, "epoch": 1369, "lr": 5.8136205564645954e-05}
{"train_loss": 0.06758005172014236, "global_step": 121878, "epoch": 1369, "lr": 5.813563355445194e-05}
{"train_loss": 0.19544163346290588, "global_step": 121879, "epoch": 1369, "lr": 5.8135061543164206e-05}
{"train_loss": 0.1763230711221695, "global_step": 121880, "epoch": 1369, "lr": 5.8134489530782795e-05}
{"train_loss": 0.10835561901330948, "global_step": 121881, "epoch": 1369, "lr": 5.813391751730781e-05}
{"train_loss": 0.2088281512260437, "global_step": 121882, "epoch": 1369, "lr": 5.813334550273932e-05}
{"train_loss": 0.1329379677772522, "global_step": 121883, "epoch": 1369, "lr": 5.81327734870774e-05}
{"train_loss": 0.18234263360500336, "global_step": 121884, "epoch": 1369, "lr": 5.813220147032212e-05}
{"train_loss": 0.11384410411119461, "global_step": 121885, "epoch": 1369, "lr": 5.813162945247358e-05}
{"train_loss": 0.13404420018196106, "global_step": 121886, "epoch": 1369, "lr": 5.813105743353183e-05}
{"train_loss": 0.1429779827594757, "global_step": 121887, "epoch": 1369, "lr": 5.813048541349696e-05}
{"train_loss": 0.11717389523983002, "global_step": 121888, "epoch": 1369, "lr": 5.812991339236905e-05}
{"train_loss": 0.1428266167640686, "global_step": 121889, "epoch": 1369, "lr": 5.812934137014817e-05}
{"train_loss": 0.12761607766151428, "global_step": 121890, "epoch": 1369, "lr": 5.81287693468344e-05}
{"train_loss": 0.1435174196958542, "global_step": 121891, "epoch": 1369, "lr": 5.8128197322427816e-05}
{"train_loss": 0.19163213670253754, "global_step": 121892, "epoch": 1369, "lr": 5.812762529692849e-05}
{"train_loss": 0.1107456162571907, "global_step": 121893, "epoch": 1369, "lr": 5.812705327033651e-05}
{"train_loss": 0.06548864394426346, "global_step": 121894, "epoch": 1369, "lr": 5.812648124265194e-05}
{"train_loss": 0.11659125983715057, "global_step": 121895, "epoch": 1369, "lr": 5.812590921387486e-05}
{"train_loss": 0.13225306570529938, "global_step": 121896, "epoch": 1369, "lr": 5.812533718400537e-05}
{"train_loss": 0.1531711369752884, "global_step": 121897, "epoch": 1369, "lr": 5.812476515304351e-05}
{"train_loss": 0.10718315839767456, "global_step": 121898, "epoch": 1369, "lr": 5.8124193120989376e-05}
{"train_loss": 0.11227921396493912, "global_step": 121899, "epoch": 1369, "lr": 5.812362108784305e-05}
{"train_loss": 0.13149462640285492, "global_step": 121900, "epoch": 1369, "lr": 5.812304905360461e-05}
{"train_loss": 0.13392765820026398, "global_step": 121901, "epoch": 1369, "lr": 5.8122477018274114e-05}
{"train_loss": 0.11814418435096741, "global_step": 121902, "epoch": 1369, "lr": 5.812190498185164e-05}
{"train_loss": 0.15223762392997742, "global_step": 121903, "epoch": 1369, "lr": 5.812133294433729e-05}
{"train_loss": 0.07910064607858658, "global_step": 121904, "epoch": 1369, "lr": 5.812076090573112e-05}
{"train_loss": 0.10984108597040176, "global_step": 121905, "epoch": 1369, "lr": 5.812018886603321e-05}
{"train_loss": 0.0815839096903801, "global_step": 121906, "epoch": 1369, "lr": 5.811961682524365e-05}
{"train_loss": 0.146936297416687, "global_step": 121907, "epoch": 1369, "lr": 5.8119044783362494e-05}
{"train_loss": 0.15963272750377655, "global_step": 121908, "epoch": 1369, "lr": 5.811847274038984e-05}
{"train_loss": 0.10685889422893524, "global_step": 121909, "epoch": 1369, "lr": 5.811790069632576e-05}
{"train_loss": 0.058632805943489075, "global_step": 121910, "epoch": 1369, "lr": 5.811732865117031e-05}
{"train_loss": 0.1099465861916542, "global_step": 121911, "epoch": 1369, "lr": 5.8116756604923594e-05}
{"train_loss": 0.1060190424323082, "global_step": 121912, "epoch": 1369, "lr": 5.811618455758568e-05}
{"train_loss": 0.14458335936069489, "global_step": 121913, "epoch": 1369, "lr": 5.811561250915665e-05}
{"train_loss": 0.09668941050767899, "global_step": 121914, "epoch": 1369, "lr": 5.811504045963656e-05}
{"train_loss": 0.15811119973659515, "global_step": 121915, "epoch": 1369, "lr": 5.8114468409025515e-05}
{"train_loss": 0.1310114562511444, "global_step": 121916, "epoch": 1369, "lr": 5.8113896357323575e-05}
{"train_loss": 0.15365241467952728, "global_step": 121917, "epoch": 1369, "lr": 5.8113324304530825e-05}
{"train_loss": 0.17683349549770355, "global_step": 121918, "epoch": 1369, "lr": 5.811275225064733e-05}
{"train_loss": 0.18551015853881836, "global_step": 121919, "epoch": 1369, "lr": 5.811218019567317e-05}
{"train_loss": 0.13507872819900513, "global_step": 121920, "epoch": 1369, "lr": 5.8111608139608434e-05}
{"train_loss": 0.12240017205476761, "global_step": 121921, "epoch": 1369, "lr": 5.811103608245319e-05}
{"train_loss": 0.16735616326332092, "global_step": 121922, "epoch": 1369, "lr": 5.811046402420751e-05}
{"train_loss": 0.16404466331005096, "global_step": 121923, "epoch": 1369, "lr": 5.8109891964871486e-05}
{"train_loss": 0.12547315657138824, "global_step": 121924, "epoch": 1369, "lr": 5.810931990444518e-05}
{"train_loss": 0.13823282718658447, "global_step": 121925, "epoch": 1369, "lr": 5.8108747842928676e-05}
{"train_loss": 0.07198990881443024, "global_step": 121926, "epoch": 1369, "lr": 5.810817578032206e-05}
{"train_loss": 0.14122149348258972, "global_step": 121927, "epoch": 1369, "lr": 5.810760371662538e-05}
{"train_loss": 0.15748457610607147, "global_step": 121928, "epoch": 1369, "lr": 5.810703165183875e-05}
{"train_loss": 0.1360015839003445, "global_step": 121929, "epoch": 1369, "lr": 5.810645958596221e-05, "val_loss": 5.5060858726501465}
{"train_loss": 0.1637050360441208, "global_step": 121930, "epoch": 1370, "lr": 5.810588751899587e-05}
{"train_loss": 0.1253018081188202, "global_step": 121931, "epoch": 1370, "lr": 5.810531545093979e-05}
{"train_loss": 0.09786190092563629, "global_step": 121932, "epoch": 1370, "lr": 5.810474338179405e-05}
{"train_loss": 0.11627542227506638, "global_step": 121933, "epoch": 1370, "lr": 5.810417131155872e-05}
{"train_loss": 0.2072588950395584, "global_step": 121934, "epoch": 1370, "lr": 5.810359924023387e-05}
{"train_loss": 0.1120903268456459, "global_step": 121935, "epoch": 1370, "lr": 5.810302716781961e-05}
{"train_loss": 0.13763529062271118, "global_step": 121936, "epoch": 1370, "lr": 5.8102455094316e-05}
{"train_loss": 0.1352321058511734, "global_step": 121937, "epoch": 1370, "lr": 5.81018830197231e-05}
{"train_loss": 0.11490700393915176, "global_step": 121938, "epoch": 1370, "lr": 5.810131094404101e-05}
{"train_loss": 0.1944933980703354, "global_step": 121939, "epoch": 1370, "lr": 5.8100738867269786e-05}
{"train_loss": 0.11554794758558273, "global_step": 121940, "epoch": 1370, "lr": 5.810016678940953e-05}
{"train_loss": 0.13583438098430634, "global_step": 121941, "epoch": 1370, "lr": 5.8099594710460295e-05}
{"train_loss": 0.14154867827892303, "global_step": 121942, "epoch": 1370, "lr": 5.809902263042216e-05}
{"train_loss": 0.14107374846935272, "global_step": 121943, "epoch": 1370, "lr": 5.809845054929524e-05}
{"train_loss": 0.07859956473112106, "global_step": 121944, "epoch": 1370, "lr": 5.809787846707956e-05}
{"train_loss": 0.13535743951797485, "global_step": 121945, "epoch": 1370, "lr": 5.809730638377523e-05}
{"train_loss": 0.1475379765033722, "global_step": 121946, "epoch": 1370, "lr": 5.8096734299382306e-05}
{"train_loss": 0.10566171258687973, "global_step": 121947, "epoch": 1370, "lr": 5.8096162213900875e-05}
{"train_loss": 0.15316173434257507, "global_step": 121948, "epoch": 1370, "lr": 5.809559012733101e-05}
{"train_loss": 0.1579761803150177, "global_step": 121949, "epoch": 1370, "lr": 5.809501803967281e-05}
{"train_loss": 0.09371387213468552, "global_step": 121950, "epoch": 1370, "lr": 5.8094445950926316e-05}
{"train_loss": 0.16043530404567719, "global_step": 121951, "epoch": 1370, "lr": 5.809387386109163e-05}
{"train_loss": 0.17781922221183777, "global_step": 121952, "epoch": 1370, "lr": 5.809330177016882e-05}
{"train_loss": 0.12137053161859512, "global_step": 121953, "epoch": 1370, "lr": 5.809272967815796e-05}
{"train_loss": 0.12061386555433273, "global_step": 121954, "epoch": 1370, "lr": 5.8092157585059136e-05}
{"train_loss": 0.1061294674873352, "global_step": 121955, "epoch": 1370, "lr": 5.809158549087242e-05}
{"train_loss": 0.15312272310256958, "global_step": 121956, "epoch": 1370, "lr": 5.809101339559788e-05}
{"train_loss": 0.13383224606513977, "global_step": 121957, "epoch": 1370, "lr": 5.809044129923562e-05}
{"train_loss": 0.12952294945716858, "global_step": 121958, "epoch": 1370, "lr": 5.808986920178569e-05}
{"train_loss": 0.1510268747806549, "global_step": 121959, "epoch": 1370, "lr": 5.808929710324815e-05}
{"train_loss": 0.08816450834274292, "global_step": 121960, "epoch": 1370, "lr": 5.808872500362313e-05}
{"train_loss": 0.11757750064134598, "global_step": 121961, "epoch": 1370, "lr": 5.808815290291067e-05}
{"train_loss": 0.17072615027427673, "global_step": 121962, "epoch": 1370, "lr": 5.808758080111089e-05}
{"train_loss": 0.12353986501693726, "global_step": 121963, "epoch": 1370, "lr": 5.80870086982238e-05}
{"train_loss": 0.12135813385248184, "global_step": 121964, "epoch": 1370, "lr": 5.8086436594249515e-05}
{"train_loss": 0.1648847609758377, "global_step": 121965, "epoch": 1370, "lr": 5.808586448918811e-05}
{"train_loss": 0.11674443632364273, "global_step": 121966, "epoch": 1370, "lr": 5.808529238303966e-05}
{"train_loss": 0.13735848665237427, "global_step": 121967, "epoch": 1370, "lr": 5.808472027580424e-05}
{"train_loss": 0.09409444779157639, "global_step": 121968, "epoch": 1370, "lr": 5.808414816748192e-05}
{"train_loss": 0.09792283177375793, "global_step": 121969, "epoch": 1370, "lr": 5.80835760580728e-05}
{"train_loss": 0.0880153626203537, "global_step": 121970, "epoch": 1370, "lr": 5.808300394757693e-05}
{"train_loss": 0.06681154668331146, "global_step": 121971, "epoch": 1370, "lr": 5.8082431835994396e-05}
{"train_loss": 0.17555789649486542, "global_step": 121972, "epoch": 1370, "lr": 5.808185972332528e-05}
{"train_loss": 0.13372640311717987, "global_step": 121973, "epoch": 1370, "lr": 5.8081287609569666e-05}
{"train_loss": 0.13367173075675964, "global_step": 121974, "epoch": 1370, "lr": 5.808071549472761e-05}
{"train_loss": 0.12268312275409698, "global_step": 121975, "epoch": 1370, "lr": 5.808014337879921e-05}
{"train_loss": 0.15157227218151093, "global_step": 121976, "epoch": 1370, "lr": 5.8079571261784524e-05}
{"train_loss": 0.12053411453962326, "global_step": 121977, "epoch": 1370, "lr": 5.807899914368363e-05}
{"train_loss": 0.10334045439958572, "global_step": 121978, "epoch": 1370, "lr": 5.807842702449663e-05}
{"train_loss": 0.09628910571336746, "global_step": 121979, "epoch": 1370, "lr": 5.8077854904223584e-05}
{"train_loss": 0.13195836544036865, "global_step": 121980, "epoch": 1370, "lr": 5.807728278286456e-05}
{"train_loss": 0.1828349232673645, "global_step": 121981, "epoch": 1370, "lr": 5.8076710660419656e-05}
{"train_loss": 0.11469778418540955, "global_step": 121982, "epoch": 1370, "lr": 5.8076138536888935e-05}
{"train_loss": 0.11038006842136383, "global_step": 121983, "epoch": 1370, "lr": 5.807556641227247e-05}
{"train_loss": 0.13303762674331665, "global_step": 121984, "epoch": 1370, "lr": 5.807499428657034e-05}
{"train_loss": 0.07895258069038391, "global_step": 121985, "epoch": 1370, "lr": 5.807442215978263e-05}
{"train_loss": 0.15200965106487274, "global_step": 121986, "epoch": 1370, "lr": 5.807385003190941e-05}
{"train_loss": 0.17070499062538147, "global_step": 121987, "epoch": 1370, "lr": 5.8073277902950765e-05}
{"train_loss": 0.09170518070459366, "global_step": 121988, "epoch": 1370, "lr": 5.8072705772906754e-05}
{"train_loss": 0.14985904097557068, "global_step": 121989, "epoch": 1370, "lr": 5.807213364177747e-05}
{"train_loss": 0.11387692391872406, "global_step": 121990, "epoch": 1370, "lr": 5.807156150956299e-05}
{"train_loss": 0.1724545806646347, "global_step": 121991, "epoch": 1370, "lr": 5.807098937626338e-05}
{"train_loss": 0.09792307764291763, "global_step": 121992, "epoch": 1370, "lr": 5.807041724187874e-05}
{"train_loss": 0.10573221743106842, "global_step": 121993, "epoch": 1370, "lr": 5.806984510640912e-05}
{"train_loss": 0.09552580863237381, "global_step": 121994, "epoch": 1370, "lr": 5.80692729698546e-05}
{"train_loss": 0.11074813455343246, "global_step": 121995, "epoch": 1370, "lr": 5.806870083221527e-05}
{"train_loss": 0.13274024426937103, "global_step": 121996, "epoch": 1370, "lr": 5.806812869349121e-05}
{"train_loss": 0.13886593282222748, "global_step": 121997, "epoch": 1370, "lr": 5.806755655368248e-05}
{"train_loss": 0.1104910597205162, "global_step": 121998, "epoch": 1370, "lr": 5.806698441278916e-05}
{"train_loss": 0.16381977498531342, "global_step": 121999, "epoch": 1370, "lr": 5.8066412270811344e-05}
{"train_loss": 0.18159353733062744, "global_step": 122000, "epoch": 1370, "lr": 5.8065840127749094e-05}
{"train_loss": 0.17283505201339722, "global_step": 122001, "epoch": 1370, "lr": 5.80652679836025e-05}
{"train_loss": 0.15647974610328674, "global_step": 122002, "epoch": 1370, "lr": 5.806469583837162e-05}
{"train_loss": 0.15449614822864532, "global_step": 122003, "epoch": 1370, "lr": 5.8064123692056535e-05}
{"train_loss": 0.09573466330766678, "global_step": 122004, "epoch": 1370, "lr": 5.8063551544657334e-05}
{"train_loss": 0.1002902165055275, "global_step": 122005, "epoch": 1370, "lr": 5.8062979396174086e-05}
{"train_loss": 0.18191547691822052, "global_step": 122006, "epoch": 1370, "lr": 5.806240724660687e-05}
{"train_loss": 0.20842228829860687, "global_step": 122007, "epoch": 1370, "lr": 5.806183509595575e-05}
{"train_loss": 0.07869713753461838, "global_step": 122008, "epoch": 1370, "lr": 5.806126294422083e-05}
{"train_loss": 0.09343991428613663, "global_step": 122009, "epoch": 1370, "lr": 5.806069079140216e-05}
{"train_loss": 0.10577589273452759, "global_step": 122010, "epoch": 1370, "lr": 5.8060118637499825e-05}
{"train_loss": 0.2186034470796585, "global_step": 122011, "epoch": 1370, "lr": 5.805954648251392e-05}
{"train_loss": 0.1419048309326172, "global_step": 122012, "epoch": 1370, "lr": 5.80589743264445e-05}
{"train_loss": 0.14204718172550201, "global_step": 122013, "epoch": 1370, "lr": 5.8058402169291635e-05}
{"train_loss": 0.21425975859165192, "global_step": 122014, "epoch": 1370, "lr": 5.8057830011055434e-05}
{"train_loss": 0.13568849861621857, "global_step": 122015, "epoch": 1370, "lr": 5.8057257851735945e-05}
{"train_loss": 0.15232519805431366, "global_step": 122016, "epoch": 1370, "lr": 5.805668569133327e-05}
{"train_loss": 0.08650653064250946, "global_step": 122017, "epoch": 1370, "lr": 5.8056113529847464e-05}
{"train_loss": 0.1322376944878128, "global_step": 122018, "epoch": 1370, "lr": 5.805554136727861e-05, "val_loss": 5.514476299285889, "train_action_mse_error": 27.019651412963867}
{"train_loss": 0.17005015909671783, "global_step": 122019, "epoch": 1371, "lr": 5.805496920362678e-05}
{"train_loss": 0.19652876257896423, "global_step": 122020, "epoch": 1371, "lr": 5.805439703889208e-05}
{"train_loss": 0.1589733511209488, "global_step": 122021, "epoch": 1371, "lr": 5.805382487307456e-05}
{"train_loss": 0.17462028563022614, "global_step": 122022, "epoch": 1371, "lr": 5.8053252706174286e-05}
{"train_loss": 0.12377426773309708, "global_step": 122023, "epoch": 1371, "lr": 5.8052680538191364e-05}
{"train_loss": 0.1408867985010147, "global_step": 122024, "epoch": 1371, "lr": 5.805210836912585e-05}
{"train_loss": 0.1559857428073883, "global_step": 122025, "epoch": 1371, "lr": 5.805153619897783e-05}
{"train_loss": 0.128880113363266, "global_step": 122026, "epoch": 1371, "lr": 5.805096402774737e-05}
{"train_loss": 0.13570493459701538, "global_step": 122027, "epoch": 1371, "lr": 5.805039185543456e-05}
{"train_loss": 0.14650484919548035, "global_step": 122028, "epoch": 1371, "lr": 5.804981968203949e-05}
{"train_loss": 0.0470796637237072, "global_step": 122029, "epoch": 1371, "lr": 5.80492475075622e-05}
{"train_loss": 0.1928870975971222, "global_step": 122030, "epoch": 1371, "lr": 5.804867533200279e-05}
{"train_loss": 0.11226417124271393, "global_step": 122031, "epoch": 1371, "lr": 5.8048103155361325e-05}
{"train_loss": 0.1744617372751236, "global_step": 122032, "epoch": 1371, "lr": 5.80475309776379e-05}
{"train_loss": 0.11349877715110779, "global_step": 122033, "epoch": 1371, "lr": 5.8046958798832585e-05}
{"train_loss": 0.15944428741931915, "global_step": 122034, "epoch": 1371, "lr": 5.8046386618945456e-05}
{"train_loss": 0.17995209991931915, "global_step": 122035, "epoch": 1371, "lr": 5.8045814437976584e-05}
{"train_loss": 0.10162919759750366, "global_step": 122036, "epoch": 1371, "lr": 5.804524225592606e-05}
{"train_loss": 0.1342790424823761, "global_step": 122037, "epoch": 1371, "lr": 5.804467007279394e-05}
{"train_loss": 0.09394620358943939, "global_step": 122038, "epoch": 1371, "lr": 5.804409788858032e-05}
{"train_loss": 0.14313361048698425, "global_step": 122039, "epoch": 1371, "lr": 5.8043525703285265e-05}
{"train_loss": 0.12288573384284973, "global_step": 122040, "epoch": 1371, "lr": 5.804295351690886e-05}
{"train_loss": 0.09760936349630356, "global_step": 122041, "epoch": 1371, "lr": 5.804238132945118e-05}
{"train_loss": 0.0841873437166214, "global_step": 122042, "epoch": 1371, "lr": 5.8041809140912295e-05}
{"train_loss": 0.11538881063461304, "global_step": 122043, "epoch": 1371, "lr": 5.804123695129229e-05}
{"train_loss": 0.1480932980775833, "global_step": 122044, "epoch": 1371, "lr": 5.8040664760591225e-05}
{"train_loss": 0.15552537143230438, "global_step": 122045, "epoch": 1371, "lr": 5.80400925688092e-05}
{"train_loss": 0.12531229853630066, "global_step": 122046, "epoch": 1371, "lr": 5.803952037594629e-05}
{"train_loss": 0.12515732645988464, "global_step": 122047, "epoch": 1371, "lr": 5.803894818200255e-05}
{"train_loss": 0.09871874004602432, "global_step": 122048, "epoch": 1371, "lr": 5.803837598697808e-05}
{"train_loss": 0.14272144436836243, "global_step": 122049, "epoch": 1371, "lr": 5.8037803790872945e-05}
{"train_loss": 0.1348821073770523, "global_step": 122050, "epoch": 1371, "lr": 5.803723159368723e-05}
{"train_loss": 0.10198833793401718, "global_step": 122051, "epoch": 1371, "lr": 5.8036659395421e-05}
{"train_loss": 0.10183310508728027, "global_step": 122052, "epoch": 1371, "lr": 5.803608719607434e-05}
{"train_loss": 0.09261835366487503, "global_step": 122053, "epoch": 1371, "lr": 5.803551499564732e-05}
{"train_loss": 0.1637706160545349, "global_step": 122054, "epoch": 1371, "lr": 5.803494279414003e-05}
{"train_loss": 0.10037679970264435, "global_step": 122055, "epoch": 1371, "lr": 5.8034370591552556e-05}
{"train_loss": 0.08981052786111832, "global_step": 122056, "epoch": 1371, "lr": 5.803379838788494e-05}
{"train_loss": 0.1252547651529312, "global_step": 122057, "epoch": 1371, "lr": 5.803322618313728e-05}
{"train_loss": 0.2153467983007431, "global_step": 122058, "epoch": 1371, "lr": 5.803265397730966e-05}
{"train_loss": 0.14115706086158752, "global_step": 122059, "epoch": 1371, "lr": 5.8032081770402145e-05}
{"train_loss": 0.13665014505386353, "global_step": 122060, "epoch": 1371, "lr": 5.803150956241481e-05}
{"train_loss": 0.15472310781478882, "global_step": 122061, "epoch": 1371, "lr": 5.8030937353347744e-05}
{"train_loss": 0.2195473164319992, "global_step": 122062, "epoch": 1371, "lr": 5.8030365143201006e-05}
{"train_loss": 0.11472208797931671, "global_step": 122063, "epoch": 1371, "lr": 5.8029792931974695e-05}
{"train_loss": 0.18210053443908691, "global_step": 122064, "epoch": 1371, "lr": 5.802922071966886e-05}
{"train_loss": 0.1120452880859375, "global_step": 122065, "epoch": 1371, "lr": 5.80286485062836e-05}
{"train_loss": 0.12334860116243362, "global_step": 122066, "epoch": 1371, "lr": 5.802807629181899e-05}
{"train_loss": 0.07484013587236404, "global_step": 122067, "epoch": 1371, "lr": 5.802750407627511e-05}
{"train_loss": 0.13641005754470825, "global_step": 122068, "epoch": 1371, "lr": 5.8026931859652e-05}
{"train_loss": 0.1430247724056244, "global_step": 122069, "epoch": 1371, "lr": 5.802635964194979e-05}
{"train_loss": 0.16537635028362274, "global_step": 122070, "epoch": 1371, "lr": 5.802578742316853e-05}
{"train_loss": 0.09419246017932892, "global_step": 122071, "epoch": 1371, "lr": 5.8025215203308306e-05}
{"train_loss": 0.16983051598072052, "global_step": 122072, "epoch": 1371, "lr": 5.802464298236918e-05}
{"train_loss": 0.1789589673280716, "global_step": 122073, "epoch": 1371, "lr": 5.8024070760351245e-05}
{"train_loss": 0.06579770892858505, "global_step": 122074, "epoch": 1371, "lr": 5.802349853725457e-05}
{"train_loss": 0.13564744591712952, "global_step": 122075, "epoch": 1371, "lr": 5.802292631307924e-05}
{"train_loss": 0.123317651450634, "global_step": 122076, "epoch": 1371, "lr": 5.802235408782531e-05}
{"train_loss": 0.19340276718139648, "global_step": 122077, "epoch": 1371, "lr": 5.8021781861492886e-05}
{"train_loss": 0.08610237389802933, "global_step": 122078, "epoch": 1371, "lr": 5.802120963408204e-05}
{"train_loss": 0.18546296656131744, "global_step": 122079, "epoch": 1371, "lr": 5.802063740559283e-05}
{"train_loss": 0.14725038409233093, "global_step": 122080, "epoch": 1371, "lr": 5.802006517602534e-05}
{"train_loss": 0.12836776673793793, "global_step": 122081, "epoch": 1371, "lr": 5.801949294537965e-05}
{"train_loss": 0.16054798662662506, "global_step": 122082, "epoch": 1371, "lr": 5.8018920713655835e-05}
{"train_loss": 0.09917128831148148, "global_step": 122083, "epoch": 1371, "lr": 5.801834848085398e-05}
{"train_loss": 0.11651219427585602, "global_step": 122084, "epoch": 1371, "lr": 5.801777624697415e-05}
{"train_loss": 0.2280747890472412, "global_step": 122085, "epoch": 1371, "lr": 5.8017204012016424e-05}
{"train_loss": 0.2631721794605255, "global_step": 122086, "epoch": 1371, "lr": 5.801663177598088e-05}
{"train_loss": 0.1162375807762146, "global_step": 122087, "epoch": 1371, "lr": 5.8016059538867605e-05}
{"train_loss": 0.06745196133852005, "global_step": 122088, "epoch": 1371, "lr": 5.801548730067668e-05}
{"train_loss": 0.1953490674495697, "global_step": 122089, "epoch": 1371, "lr": 5.801491506140815e-05}
{"train_loss": 0.12008263170719147, "global_step": 122090, "epoch": 1371, "lr": 5.801434282106213e-05}
{"train_loss": 0.1934053748846054, "global_step": 122091, "epoch": 1371, "lr": 5.8013770579638674e-05}
{"train_loss": 0.12357855588197708, "global_step": 122092, "epoch": 1371, "lr": 5.801319833713785e-05}
{"train_loss": 0.24780066311359406, "global_step": 122093, "epoch": 1371, "lr": 5.801262609355976e-05}
{"train_loss": 0.13642914593219757, "global_step": 122094, "epoch": 1371, "lr": 5.801205384890448e-05}
{"train_loss": 0.08516661077737808, "global_step": 122095, "epoch": 1371, "lr": 5.801148160317207e-05}
{"train_loss": 0.11529184132814407, "global_step": 122096, "epoch": 1371, "lr": 5.8010909356362606e-05}
{"train_loss": 0.11472304165363312, "global_step": 122097, "epoch": 1371, "lr": 5.801033710847617e-05}
{"train_loss": 0.07723163068294525, "global_step": 122098, "epoch": 1371, "lr": 5.800976485951287e-05}
{"train_loss": 0.1655912697315216, "global_step": 122099, "epoch": 1371, "lr": 5.800919260947273e-05}
{"train_loss": 0.1478744000196457, "global_step": 122100, "epoch": 1371, "lr": 5.8008620358355856e-05}
{"train_loss": 0.11556148529052734, "global_step": 122101, "epoch": 1371, "lr": 5.800804810616233e-05}
{"train_loss": 0.10529637336730957, "global_step": 122102, "epoch": 1371, "lr": 5.8007475852892214e-05}
{"train_loss": 0.1314222514629364, "global_step": 122103, "epoch": 1371, "lr": 5.800690359854558e-05}
{"train_loss": 0.13948394358158112, "global_step": 122104, "epoch": 1371, "lr": 5.800633134312251e-05}
{"train_loss": 0.17043009400367737, "global_step": 122105, "epoch": 1371, "lr": 5.80057590866231e-05}
{"train_loss": 0.1592109352350235, "global_step": 122106, "epoch": 1371, "lr": 5.800518682904742e-05}
{"train_loss": 0.1383894488968876, "global_step": 122107, "epoch": 1371, "lr": 5.800461457039553e-05, "val_loss": 5.398449420928955}
{"train_loss": 0.14825496077537537, "global_step": 122108, "epoch": 1372, "lr": 5.800404231066752e-05}
{"train_loss": 0.1527312695980072, "global_step": 122109, "epoch": 1372, "lr": 5.800347004986346e-05}
{"train_loss": 0.17220564186573029, "global_step": 122110, "epoch": 1372, "lr": 5.800289778798343e-05}
{"train_loss": 0.10690098255872726, "global_step": 122111, "epoch": 1372, "lr": 5.8002325525027525e-05}
{"train_loss": 0.12421510368585587, "global_step": 122112, "epoch": 1372, "lr": 5.800175326099578e-05}
{"train_loss": 0.11619719862937927, "global_step": 122113, "epoch": 1372, "lr": 5.800118099588832e-05}
{"train_loss": 0.11240926384925842, "global_step": 122114, "epoch": 1372, "lr": 5.8000608729705186e-05}
{"train_loss": 0.11275307089090347, "global_step": 122115, "epoch": 1372, "lr": 5.800003646244647e-05}
{"train_loss": 0.11074404418468475, "global_step": 122116, "epoch": 1372, "lr": 5.799946419411224e-05}
{"train_loss": 0.17629264295101166, "global_step": 122117, "epoch": 1372, "lr": 5.79988919247026e-05}
{"train_loss": 0.18278858065605164, "global_step": 122118, "epoch": 1372, "lr": 5.7998319654217595e-05}
{"train_loss": 0.10831774026155472, "global_step": 122119, "epoch": 1372, "lr": 5.799774738265731e-05}
{"train_loss": 0.13622203469276428, "global_step": 122120, "epoch": 1372, "lr": 5.799717511002183e-05}
{"train_loss": 0.12917302548885345, "global_step": 122121, "epoch": 1372, "lr": 5.799660283631122e-05}
{"train_loss": 0.09775840491056442, "global_step": 122122, "epoch": 1372, "lr": 5.799603056152556e-05}
{"train_loss": 0.14258989691734314, "global_step": 122123, "epoch": 1372, "lr": 5.799545828566494e-05}
{"train_loss": 0.09566663950681686, "global_step": 122124, "epoch": 1372, "lr": 5.7994886008729424e-05}
{"train_loss": 0.20335368812084198, "global_step": 122125, "epoch": 1372, "lr": 5.79943137307191e-05}
{"train_loss": 0.14785680174827576, "global_step": 122126, "epoch": 1372, "lr": 5.799374145163403e-05}
{"train_loss": 0.1400025188922882, "global_step": 122127, "epoch": 1372, "lr": 5.7993169171474306e-05}
{"train_loss": 0.14023303985595703, "global_step": 122128, "epoch": 1372, "lr": 5.799259689023999e-05}
{"train_loss": 0.08542997390031815, "global_step": 122129, "epoch": 1372, "lr": 5.7992024607931174e-05}
{"train_loss": 0.16158099472522736, "global_step": 122130, "epoch": 1372, "lr": 5.799145232454792e-05}
{"train_loss": 0.1783214658498764, "global_step": 122131, "epoch": 1372, "lr": 5.7990880040090326e-05}
{"train_loss": 0.2016984224319458, "global_step": 122132, "epoch": 1372, "lr": 5.799030775455845e-05}
{"train_loss": 0.1955108493566513, "global_step": 122133, "epoch": 1372, "lr": 5.7989735467952365e-05}
{"train_loss": 0.12480320781469345, "global_step": 122134, "epoch": 1372, "lr": 5.798916318027217e-05}
{"train_loss": 0.12583507597446442, "global_step": 122135, "epoch": 1372, "lr": 5.798859089151791e-05}
{"train_loss": 0.08250030875205994, "global_step": 122136, "epoch": 1372, "lr": 5.798801860168972e-05}
{"train_loss": 0.1948775202035904, "global_step": 122137, "epoch": 1372, "lr": 5.798744631078761e-05}
{"train_loss": 0.11909382045269012, "global_step": 122138, "epoch": 1372, "lr": 5.798687401881169e-05}
{"train_loss": 0.20616906881332397, "global_step": 122139, "epoch": 1372, "lr": 5.798630172576203e-05}
{"train_loss": 0.13855859637260437, "global_step": 122140, "epoch": 1372, "lr": 5.798572943163869e-05}
{"train_loss": 0.13813145458698273, "global_step": 122141, "epoch": 1372, "lr": 5.798515713644179e-05}
{"train_loss": 0.1707247942686081, "global_step": 122142, "epoch": 1372, "lr": 5.7984584840171374e-05}
{"train_loss": 0.13405542075634003, "global_step": 122143, "epoch": 1372, "lr": 5.7984012542827534e-05}
{"train_loss": 0.2158765196800232, "global_step": 122144, "epoch": 1372, "lr": 5.798344024441035e-05}
{"train_loss": 0.14723211526870728, "global_step": 122145, "epoch": 1372, "lr": 5.798286794491987e-05}
{"train_loss": 0.10520126670598984, "global_step": 122146, "epoch": 1372, "lr": 5.79822956443562e-05}
{"train_loss": 0.19776253402233124, "global_step": 122147, "epoch": 1372, "lr": 5.798172334271941e-05}
{"train_loss": 0.16280770301818848, "global_step": 122148, "epoch": 1372, "lr": 5.798115104000957e-05}
{"train_loss": 0.12292003631591797, "global_step": 122149, "epoch": 1372, "lr": 5.7980578736226765e-05}
{"train_loss": 0.07873446494340897, "global_step": 122150, "epoch": 1372, "lr": 5.798000643137107e-05}
{"train_loss": 0.17526960372924805, "global_step": 122151, "epoch": 1372, "lr": 5.797943412544255e-05}
{"train_loss": 0.1322157233953476, "global_step": 122152, "epoch": 1372, "lr": 5.79788618184413e-05}
{"train_loss": 0.12928470969200134, "global_step": 122153, "epoch": 1372, "lr": 5.7978289510367387e-05}
{"train_loss": 0.12999901175498962, "global_step": 122154, "epoch": 1372, "lr": 5.7977717201220895e-05}
{"train_loss": 0.09916878491640091, "global_step": 122155, "epoch": 1372, "lr": 5.79771448910019e-05}
{"train_loss": 0.11389926075935364, "global_step": 122156, "epoch": 1372, "lr": 5.7976572579710474e-05}
{"train_loss": 0.10802300274372101, "global_step": 122157, "epoch": 1372, "lr": 5.797600026734669e-05}
{"train_loss": 0.11962033063173294, "global_step": 122158, "epoch": 1372, "lr": 5.7975427953910624e-05}
{"train_loss": 0.10848216712474823, "global_step": 122159, "epoch": 1372, "lr": 5.797485563940236e-05}
{"train_loss": 0.1490955948829651, "global_step": 122160, "epoch": 1372, "lr": 5.7974283323821975e-05}
{"train_loss": 0.19938303530216217, "global_step": 122161, "epoch": 1372, "lr": 5.797371100716955e-05}
{"train_loss": 0.13229243457317352, "global_step": 122162, "epoch": 1372, "lr": 5.7973138689445164e-05}
{"train_loss": 0.14166657626628876, "global_step": 122163, "epoch": 1372, "lr": 5.797256637064887e-05}
{"train_loss": 0.15225744247436523, "global_step": 122164, "epoch": 1372, "lr": 5.7971994050780774e-05}
{"train_loss": 0.1474980115890503, "global_step": 122165, "epoch": 1372, "lr": 5.7971421729840934e-05}
{"train_loss": 0.11610350012779236, "global_step": 122166, "epoch": 1372, "lr": 5.797084940782943e-05}
{"train_loss": 0.1355152279138565, "global_step": 122167, "epoch": 1372, "lr": 5.797027708474635e-05}
{"train_loss": 0.10509733110666275, "global_step": 122168, "epoch": 1372, "lr": 5.796970476059176e-05}
{"train_loss": 0.12712958455085754, "global_step": 122169, "epoch": 1372, "lr": 5.796913243536575e-05}
{"train_loss": 0.12950584292411804, "global_step": 122170, "epoch": 1372, "lr": 5.796856010906837e-05}
{"train_loss": 0.1809702068567276, "global_step": 122171, "epoch": 1372, "lr": 5.7967987781699726e-05}
{"train_loss": 0.12354876846075058, "global_step": 122172, "epoch": 1372, "lr": 5.796741545325988e-05}
{"train_loss": 0.24093161523342133, "global_step": 122173, "epoch": 1372, "lr": 5.796684312374892e-05}
{"train_loss": 0.12330388277769089, "global_step": 122174, "epoch": 1372, "lr": 5.79662707931669e-05}
{"train_loss": 0.13758131861686707, "global_step": 122175, "epoch": 1372, "lr": 5.796569846151393e-05}
{"train_loss": 0.18680424988269806, "global_step": 122176, "epoch": 1372, "lr": 5.796512612879006e-05}
{"train_loss": 0.11672594398260117, "global_step": 122177, "epoch": 1372, "lr": 5.796455379499536e-05}
{"train_loss": 0.1771690547466278, "global_step": 122178, "epoch": 1372, "lr": 5.796398146012994e-05}
{"train_loss": 0.16969101130962372, "global_step": 122179, "epoch": 1372, "lr": 5.7963409124193854e-05}
{"train_loss": 0.16245801746845245, "global_step": 122180, "epoch": 1372, "lr": 5.796283678718719e-05}
{"train_loss": 0.18950778245925903, "global_step": 122181, "epoch": 1372, "lr": 5.7962264449110015e-05}
{"train_loss": 0.22629129886627197, "global_step": 122182, "epoch": 1372, "lr": 5.796169210996241e-05}
{"train_loss": 0.12878504395484924, "global_step": 122183, "epoch": 1372, "lr": 5.796111976974445e-05}
{"train_loss": 0.1495327651500702, "global_step": 122184, "epoch": 1372, "lr": 5.7960547428456224e-05}
{"train_loss": 0.11438827961683273, "global_step": 122185, "epoch": 1372, "lr": 5.7959975086097795e-05}
{"train_loss": 0.08033082634210587, "global_step": 122186, "epoch": 1372, "lr": 5.7959402742669244e-05}
{"train_loss": 0.09609044343233109, "global_step": 122187, "epoch": 1372, "lr": 5.7958830398170646e-05}
{"train_loss": 0.1279885321855545, "global_step": 122188, "epoch": 1372, "lr": 5.795825805260209e-05}
{"train_loss": 0.12076777219772339, "global_step": 122189, "epoch": 1372, "lr": 5.7957685705963636e-05}
{"train_loss": 0.15364892780780792, "global_step": 122190, "epoch": 1372, "lr": 5.795711335825537e-05}
{"train_loss": 0.17701052129268646, "global_step": 122191, "epoch": 1372, "lr": 5.795654100947736e-05}
{"train_loss": 0.12041351944208145, "global_step": 122192, "epoch": 1372, "lr": 5.79559686596297e-05}
{"train_loss": 0.11369931697845459, "global_step": 122193, "epoch": 1372, "lr": 5.7955396308712454e-05}
{"train_loss": 0.18891549110412598, "global_step": 122194, "epoch": 1372, "lr": 5.79548239567257e-05}
{"train_loss": 0.15232770144939423, "global_step": 122195, "epoch": 1372, "lr": 5.795425160366952e-05}
{"train_loss": 0.14291653974672383, "global_step": 122196, "epoch": 1372, "lr": 5.795367924954398e-05, "val_loss": 5.210574626922607}
{"train_loss": 0.12713627517223358, "global_step": 122197, "epoch": 1373, "lr": 5.7953106894349165e-05}
{"train_loss": 0.20730724930763245, "global_step": 122198, "epoch": 1373, "lr": 5.795253453808516e-05}
{"train_loss": 0.13843125104904175, "global_step": 122199, "epoch": 1373, "lr": 5.7951962180752026e-05}
{"train_loss": 0.11367625743150711, "global_step": 122200, "epoch": 1373, "lr": 5.795138982234986e-05}
{"train_loss": 0.10124863684177399, "global_step": 122201, "epoch": 1373, "lr": 5.795081746287871e-05}
{"train_loss": 0.10980574041604996, "global_step": 122202, "epoch": 1373, "lr": 5.7950245102338686e-05}
{"train_loss": 0.12895134091377258, "global_step": 122203, "epoch": 1373, "lr": 5.794967274072983e-05}
{"train_loss": 0.08105701953172684, "global_step": 122204, "epoch": 1373, "lr": 5.7949100378052255e-05}
{"train_loss": 0.07935119420289993, "global_step": 122205, "epoch": 1373, "lr": 5.794852801430601e-05}
{"train_loss": 0.11413837969303131, "global_step": 122206, "epoch": 1373, "lr": 5.794795564949118e-05}
{"train_loss": 0.15460167825222015, "global_step": 122207, "epoch": 1373, "lr": 5.794738328360786e-05}
{"train_loss": 0.12868428230285645, "global_step": 122208, "epoch": 1373, "lr": 5.7946810916656094e-05}
{"train_loss": 0.09454365074634552, "global_step": 122209, "epoch": 1373, "lr": 5.794623854863599e-05}
{"train_loss": 0.1203136220574379, "global_step": 122210, "epoch": 1373, "lr": 5.79456661795476e-05}
{"train_loss": 0.172601580619812, "global_step": 122211, "epoch": 1373, "lr": 5.7945093809391016e-05}
{"train_loss": 0.13300639390945435, "global_step": 122212, "epoch": 1373, "lr": 5.794452143816631e-05}
{"train_loss": 0.08975091576576233, "global_step": 122213, "epoch": 1373, "lr": 5.794394906587356e-05}
{"train_loss": 0.10601677000522614, "global_step": 122214, "epoch": 1373, "lr": 5.794337669251284e-05}
{"train_loss": 0.16283223032951355, "global_step": 122215, "epoch": 1373, "lr": 5.794280431808423e-05}
{"train_loss": 0.1767270267009735, "global_step": 122216, "epoch": 1373, "lr": 5.794223194258782e-05}
{"train_loss": 0.11383126676082611, "global_step": 122217, "epoch": 1373, "lr": 5.7941659566023664e-05}
{"train_loss": 0.13487696647644043, "global_step": 122218, "epoch": 1373, "lr": 5.794108718839184e-05}
{"train_loss": 0.1690741926431656, "global_step": 122219, "epoch": 1373, "lr": 5.7940514809692445e-05}
{"train_loss": 0.1332886964082718, "global_step": 122220, "epoch": 1373, "lr": 5.7939942429925544e-05}
{"train_loss": 0.09804743528366089, "global_step": 122221, "epoch": 1373, "lr": 5.793937004909121e-05}
{"train_loss": 0.12546072900295258, "global_step": 122222, "epoch": 1373, "lr": 5.793879766718952e-05}
{"train_loss": 0.13798995316028595, "global_step": 122223, "epoch": 1373, "lr": 5.793822528422057e-05}
{"train_loss": 0.17767499387264252, "global_step": 122224, "epoch": 1373, "lr": 5.7937652900184415e-05}
{"train_loss": 0.09215734899044037, "global_step": 122225, "epoch": 1373, "lr": 5.7937080515081145e-05}
{"train_loss": 0.0682636946439743, "global_step": 122226, "epoch": 1373, "lr": 5.793650812891082e-05}
{"train_loss": 0.1014147400856018, "global_step": 122227, "epoch": 1373, "lr": 5.793593574167353e-05}
{"train_loss": 0.07621891796588898, "global_step": 122228, "epoch": 1373, "lr": 5.793536335336935e-05}
{"train_loss": 0.1554584503173828, "global_step": 122229, "epoch": 1373, "lr": 5.7934790963998364e-05}
{"train_loss": 0.08050376921892166, "global_step": 122230, "epoch": 1373, "lr": 5.7934218573560636e-05}
{"train_loss": 0.0890842080116272, "global_step": 122231, "epoch": 1373, "lr": 5.793364618205626e-05}
{"train_loss": 0.09908653050661087, "global_step": 122232, "epoch": 1373, "lr": 5.7933073789485294e-05}
{"train_loss": 0.10236304253339767, "global_step": 122233, "epoch": 1373, "lr": 5.793250139584783e-05}
{"train_loss": 0.14553174376487732, "global_step": 122234, "epoch": 1373, "lr": 5.793192900114393e-05}
{"train_loss": 0.10828256607055664, "global_step": 122235, "epoch": 1373, "lr": 5.793135660537367e-05}
{"train_loss": 0.0642501637339592, "global_step": 122236, "epoch": 1373, "lr": 5.793078420853715e-05}
{"train_loss": 0.1832161843776703, "global_step": 122237, "epoch": 1373, "lr": 5.793021181063443e-05}
{"train_loss": 0.13262327015399933, "global_step": 122238, "epoch": 1373, "lr": 5.792963941166558e-05}
{"train_loss": 0.13049015402793884, "global_step": 122239, "epoch": 1373, "lr": 5.79290670116307e-05}
{"train_loss": 0.08076268434524536, "global_step": 122240, "epoch": 1373, "lr": 5.7928494610529845e-05}
{"train_loss": 0.15092453360557556, "global_step": 122241, "epoch": 1373, "lr": 5.7927922208363103e-05}
{"train_loss": 0.1462664157152176, "global_step": 122242, "epoch": 1373, "lr": 5.792734980513055e-05}
{"train_loss": 0.14195826649665833, "global_step": 122243, "epoch": 1373, "lr": 5.792677740083225e-05}
{"train_loss": 0.13439960777759552, "global_step": 122244, "epoch": 1373, "lr": 5.79262049954683e-05}
{"train_loss": 0.15539899468421936, "global_step": 122245, "epoch": 1373, "lr": 5.792563258903878e-05}
{"train_loss": 0.13678887486457825, "global_step": 122246, "epoch": 1373, "lr": 5.792506018154375e-05}
{"train_loss": 0.13716156780719757, "global_step": 122247, "epoch": 1373, "lr": 5.792448777298328e-05}
{"train_loss": 0.15625934302806854, "global_step": 122248, "epoch": 1373, "lr": 5.792391536335746e-05}
{"train_loss": 0.11367753148078918, "global_step": 122249, "epoch": 1373, "lr": 5.792334295266638e-05}
{"train_loss": 0.07437712699174881, "global_step": 122250, "epoch": 1373, "lr": 5.7922770540910095e-05}
{"train_loss": 0.10338965803384781, "global_step": 122251, "epoch": 1373, "lr": 5.792219812808869e-05}
{"train_loss": 0.13866554200649261, "global_step": 122252, "epoch": 1373, "lr": 5.792162571420224e-05}
{"train_loss": 0.13167037069797516, "global_step": 122253, "epoch": 1373, "lr": 5.792105329925084e-05}
{"train_loss": 0.13368700444698334, "global_step": 122254, "epoch": 1373, "lr": 5.7920480883234526e-05}
{"train_loss": 0.11022624373435974, "global_step": 122255, "epoch": 1373, "lr": 5.791990846615342e-05}
{"train_loss": 0.19504685699939728, "global_step": 122256, "epoch": 1373, "lr": 5.791933604800757e-05}
{"train_loss": 0.2083207368850708, "global_step": 122257, "epoch": 1373, "lr": 5.7918763628797054e-05}
{"train_loss": 0.13254772126674652, "global_step": 122258, "epoch": 1373, "lr": 5.791819120852197e-05}
{"train_loss": 0.12454245239496231, "global_step": 122259, "epoch": 1373, "lr": 5.7917618787182384e-05}
{"train_loss": 0.13179907202720642, "global_step": 122260, "epoch": 1373, "lr": 5.791704636477836e-05}
{"train_loss": 0.10924039781093597, "global_step": 122261, "epoch": 1373, "lr": 5.791647394130999e-05}
{"train_loss": 0.1295085847377777, "global_step": 122262, "epoch": 1373, "lr": 5.791590151677734e-05}
{"train_loss": 0.09563899040222168, "global_step": 122263, "epoch": 1373, "lr": 5.7915329091180505e-05}
{"train_loss": 0.07019798457622528, "global_step": 122264, "epoch": 1373, "lr": 5.791475666451954e-05}
{"train_loss": 0.12790639698505402, "global_step": 122265, "epoch": 1373, "lr": 5.791418423679453e-05}
{"train_loss": 0.10107626020908356, "global_step": 122266, "epoch": 1373, "lr": 5.791361180800556e-05}
{"train_loss": 0.1024012342095375, "global_step": 122267, "epoch": 1373, "lr": 5.791303937815271e-05}
{"train_loss": 0.13631528615951538, "global_step": 122268, "epoch": 1373, "lr": 5.791246694723604e-05}
{"train_loss": 0.08680835366249084, "global_step": 122269, "epoch": 1373, "lr": 5.791189451525564e-05}
{"train_loss": 0.22938241064548492, "global_step": 122270, "epoch": 1373, "lr": 5.791132208221157e-05}
{"train_loss": 0.24321803450584412, "global_step": 122271, "epoch": 1373, "lr": 5.7910749648103934e-05}
{"train_loss": 0.12000259757041931, "global_step": 122272, "epoch": 1373, "lr": 5.791017721293279e-05}
{"train_loss": 0.13932156562805176, "global_step": 122273, "epoch": 1373, "lr": 5.790960477669822e-05}
{"train_loss": 0.2117396891117096, "global_step": 122274, "epoch": 1373, "lr": 5.790903233940029e-05}
{"train_loss": 0.0648648664355278, "global_step": 122275, "epoch": 1373, "lr": 5.79084599010391e-05}
{"train_loss": 0.11382035911083221, "global_step": 122276, "epoch": 1373, "lr": 5.790788746161471e-05}
{"train_loss": 0.09181994199752808, "global_step": 122277, "epoch": 1373, "lr": 5.79073150211272e-05}
{"train_loss": 0.11134935915470123, "global_step": 122278, "epoch": 1373, "lr": 5.790674257957664e-05}
{"train_loss": 0.10496771335601807, "global_step": 122279, "epoch": 1373, "lr": 5.790617013696312e-05}
{"train_loss": 0.1464514434337616, "global_step": 122280, "epoch": 1373, "lr": 5.790559769328672e-05}
{"train_loss": 0.13475190103054047, "global_step": 122281, "epoch": 1373, "lr": 5.79050252485475e-05}
{"train_loss": 0.1041022539138794, "global_step": 122282, "epoch": 1373, "lr": 5.790445280274555e-05}
{"train_loss": 0.09999730437994003, "global_step": 122283, "epoch": 1373, "lr": 5.790388035588093e-05}
{"train_loss": 0.17688807845115662, "global_step": 122284, "epoch": 1373, "lr": 5.790330790795375e-05}
{"train_loss": 0.12623176280032383, "global_step": 122285, "epoch": 1373, "lr": 5.790273545896404e-05, "val_loss": 5.478426456451416}
{"train_loss": 0.1327141523361206, "global_step": 122286, "epoch": 1374, "lr": 5.790216300891193e-05}
{"train_loss": 0.16673077642917633, "global_step": 122287, "epoch": 1374, "lr": 5.790159055779747e-05}
{"train_loss": 0.1137503981590271, "global_step": 122288, "epoch": 1374, "lr": 5.790101810562072e-05}
{"train_loss": 0.18482105433940887, "global_step": 122289, "epoch": 1374, "lr": 5.790044565238178e-05}
{"train_loss": 0.17475834488868713, "global_step": 122290, "epoch": 1374, "lr": 5.789987319808073e-05}
{"train_loss": 0.11281788349151611, "global_step": 122291, "epoch": 1374, "lr": 5.789930074271763e-05}
{"train_loss": 0.1460670381784439, "global_step": 122292, "epoch": 1374, "lr": 5.789872828629257e-05}
{"train_loss": 0.23211121559143066, "global_step": 122293, "epoch": 1374, "lr": 5.789815582880562e-05}
{"train_loss": 0.16410639882087708, "global_step": 122294, "epoch": 1374, "lr": 5.789758337025686e-05}
{"train_loss": 0.11351726204156876, "global_step": 122295, "epoch": 1374, "lr": 5.789701091064637e-05}
{"train_loss": 0.11242968589067459, "global_step": 122296, "epoch": 1374, "lr": 5.789643844997421e-05}
{"train_loss": 0.1805959790945053, "global_step": 122297, "epoch": 1374, "lr": 5.789586598824048e-05}
{"train_loss": 0.1488344669342041, "global_step": 122298, "epoch": 1374, "lr": 5.789529352544525e-05}
{"train_loss": 0.10951022803783417, "global_step": 122299, "epoch": 1374, "lr": 5.789472106158858e-05}
{"train_loss": 0.10465231537818909, "global_step": 122300, "epoch": 1374, "lr": 5.789414859667057e-05}
{"train_loss": 0.11675294488668442, "global_step": 122301, "epoch": 1374, "lr": 5.7893576130691294e-05}
{"train_loss": 0.133217453956604, "global_step": 122302, "epoch": 1374, "lr": 5.789300366365083e-05}
{"train_loss": 0.1587788164615631, "global_step": 122303, "epoch": 1374, "lr": 5.7892431195549224e-05}
{"train_loss": 0.1330137699842453, "global_step": 122304, "epoch": 1374, "lr": 5.789185872638658e-05}
{"train_loss": 0.11998621374368668, "global_step": 122305, "epoch": 1374, "lr": 5.789128625616299e-05}
{"train_loss": 0.1382780224084854, "global_step": 122306, "epoch": 1374, "lr": 5.78907137848785e-05}
{"train_loss": 0.15784941613674164, "global_step": 122307, "epoch": 1374, "lr": 5.789014131253321e-05}
{"train_loss": 0.08164039999246597, "global_step": 122308, "epoch": 1374, "lr": 5.788956883912717e-05}
{"train_loss": 0.12108568102121353, "global_step": 122309, "epoch": 1374, "lr": 5.788899636466049e-05}
{"train_loss": 0.12713652849197388, "global_step": 122310, "epoch": 1374, "lr": 5.788842388913321e-05}
{"train_loss": 0.0726725235581398, "global_step": 122311, "epoch": 1374, "lr": 5.788785141254544e-05}
{"train_loss": 0.11156304180622101, "global_step": 122312, "epoch": 1374, "lr": 5.788727893489725e-05}
{"train_loss": 0.09593359380960464, "global_step": 122313, "epoch": 1374, "lr": 5.788670645618871e-05}
{"train_loss": 0.17208266258239746, "global_step": 122314, "epoch": 1374, "lr": 5.78861339764199e-05}
{"train_loss": 0.12991659343242645, "global_step": 122315, "epoch": 1374, "lr": 5.788556149559089e-05}
{"train_loss": 0.12319032847881317, "global_step": 122316, "epoch": 1374, "lr": 5.788498901370176e-05}
{"train_loss": 0.13992822170257568, "global_step": 122317, "epoch": 1374, "lr": 5.7884416530752594e-05}
{"train_loss": 0.0691918283700943, "global_step": 122318, "epoch": 1374, "lr": 5.788384404674346e-05}
{"train_loss": 0.12331994622945786, "global_step": 122319, "epoch": 1374, "lr": 5.788327156167445e-05}
{"train_loss": 0.16633422672748566, "global_step": 122320, "epoch": 1374, "lr": 5.788269907554562e-05}
{"train_loss": 0.21997512876987457, "global_step": 122321, "epoch": 1374, "lr": 5.788212658835706e-05}
{"train_loss": 0.1458924412727356, "global_step": 122322, "epoch": 1374, "lr": 5.7881554100108824e-05}
{"train_loss": 0.07047227025032043, "global_step": 122323, "epoch": 1374, "lr": 5.7880981610801033e-05}
{"train_loss": 0.12069565057754517, "global_step": 122324, "epoch": 1374, "lr": 5.788040912043374e-05}
{"train_loss": 0.12879988551139832, "global_step": 122325, "epoch": 1374, "lr": 5.7879836629007025e-05}
{"train_loss": 0.18098390102386475, "global_step": 122326, "epoch": 1374, "lr": 5.787926413652095e-05}
{"train_loss": 0.13942769169807434, "global_step": 122327, "epoch": 1374, "lr": 5.7878691642975616e-05}
{"train_loss": 0.1068931519985199, "global_step": 122328, "epoch": 1374, "lr": 5.787811914837108e-05}
{"train_loss": 0.13354848325252533, "global_step": 122329, "epoch": 1374, "lr": 5.787754665270742e-05}
{"train_loss": 0.10532789677381516, "global_step": 122330, "epoch": 1374, "lr": 5.787697415598473e-05}
{"train_loss": 0.19743257761001587, "global_step": 122331, "epoch": 1374, "lr": 5.787640165820307e-05}
{"train_loss": 0.1253482848405838, "global_step": 122332, "epoch": 1374, "lr": 5.787582915936254e-05}
{"train_loss": 0.11032233387231827, "global_step": 122333, "epoch": 1374, "lr": 5.7875256659463184e-05}
{"train_loss": 0.1824975460767746, "global_step": 122334, "epoch": 1374, "lr": 5.78746841585051e-05}
{"train_loss": 0.13650298118591309, "global_step": 122335, "epoch": 1374, "lr": 5.787411165648836e-05}
{"train_loss": 0.11531282961368561, "global_step": 122336, "epoch": 1374, "lr": 5.7873539153413036e-05}
{"train_loss": 0.12649235129356384, "global_step": 122337, "epoch": 1374, "lr": 5.787296664927923e-05}
{"train_loss": 0.09411371499300003, "global_step": 122338, "epoch": 1374, "lr": 5.787239414408698e-05}
{"train_loss": 0.17762790620326996, "global_step": 122339, "epoch": 1374, "lr": 5.787182163783639e-05}
{"train_loss": 0.11937855929136276, "global_step": 122340, "epoch": 1374, "lr": 5.7871249130527516e-05}
{"train_loss": 0.13205157220363617, "global_step": 122341, "epoch": 1374, "lr": 5.787067662216046e-05}
{"train_loss": 0.06549108773469925, "global_step": 122342, "epoch": 1374, "lr": 5.787010411273529e-05}
{"train_loss": 0.15229466557502747, "global_step": 122343, "epoch": 1374, "lr": 5.786953160225208e-05}
{"train_loss": 0.17249850928783417, "global_step": 122344, "epoch": 1374, "lr": 5.786895909071092e-05}
{"train_loss": 0.07494745403528214, "global_step": 122345, "epoch": 1374, "lr": 5.7868386578111863e-05}
{"train_loss": 0.17431305348873138, "global_step": 122346, "epoch": 1374, "lr": 5.786781406445499e-05}
{"train_loss": 0.21571069955825806, "global_step": 122347, "epoch": 1374, "lr": 5.7867241549740394e-05}
{"train_loss": 0.1514315903186798, "global_step": 122348, "epoch": 1374, "lr": 5.786666903396814e-05}
{"train_loss": 0.20871491730213165, "global_step": 122349, "epoch": 1374, "lr": 5.7866096517138314e-05}
{"train_loss": 0.14115868508815765, "global_step": 122350, "epoch": 1374, "lr": 5.786552399925098e-05}
{"train_loss": 0.16357535123825073, "global_step": 122351, "epoch": 1374, "lr": 5.786495148030622e-05}
{"train_loss": 0.16106964647769928, "global_step": 122352, "epoch": 1374, "lr": 5.786437896030412e-05}
{"train_loss": 0.17743133008480072, "global_step": 122353, "epoch": 1374, "lr": 5.786380643924474e-05}
{"train_loss": 0.10195048153400421, "global_step": 122354, "epoch": 1374, "lr": 5.786323391712818e-05}
{"train_loss": 0.12756995856761932, "global_step": 122355, "epoch": 1374, "lr": 5.78626613939545e-05}
{"train_loss": 0.11898384988307953, "global_step": 122356, "epoch": 1374, "lr": 5.786208886972377e-05}
{"train_loss": 0.15201109647750854, "global_step": 122357, "epoch": 1374, "lr": 5.7861516344436094e-05}
{"train_loss": 0.1642773151397705, "global_step": 122358, "epoch": 1374, "lr": 5.7860943818091517e-05}
{"train_loss": 0.1475401520729065, "global_step": 122359, "epoch": 1374, "lr": 5.786037129069014e-05}
{"train_loss": 0.14738701283931732, "global_step": 122360, "epoch": 1374, "lr": 5.785979876223203e-05}
{"train_loss": 0.22657211124897003, "global_step": 122361, "epoch": 1374, "lr": 5.785922623271727e-05}
{"train_loss": 0.1282554715871811, "global_step": 122362, "epoch": 1374, "lr": 5.785865370214593e-05}
{"train_loss": 0.13725870847702026, "global_step": 122363, "epoch": 1374, "lr": 5.7858081170518095e-05}
{"train_loss": 0.14642509818077087, "global_step": 122364, "epoch": 1374, "lr": 5.7857508637833836e-05}
{"train_loss": 0.12323355674743652, "global_step": 122365, "epoch": 1374, "lr": 5.785693610409323e-05}
{"train_loss": 0.09622613340616226, "global_step": 122366, "epoch": 1374, "lr": 5.7856363569296356e-05}
{"train_loss": 0.09242274612188339, "global_step": 122367, "epoch": 1374, "lr": 5.7855791033443285e-05}
{"train_loss": 0.1862756460905075, "global_step": 122368, "epoch": 1374, "lr": 5.7855218496534106e-05}
{"train_loss": 0.13667401671409607, "global_step": 122369, "epoch": 1374, "lr": 5.785464595856887e-05}
{"train_loss": 0.21465225517749786, "global_step": 122370, "epoch": 1374, "lr": 5.7854073419547696e-05}
{"train_loss": 0.14051952958106995, "global_step": 122371, "epoch": 1374, "lr": 5.785350087947062e-05}
{"train_loss": 0.14219343662261963, "global_step": 122372, "epoch": 1374, "lr": 5.785292833833774e-05}
{"train_loss": 0.0831666812300682, "global_step": 122373, "epoch": 1374, "lr": 5.7852355796149136e-05}
{"train_loss": 0.13875854777151278, "global_step": 122374, "epoch": 1374, "lr": 5.7851783252904876e-05, "val_loss": 5.263754844665527}
{"train_loss": 0.12472562491893768, "global_step": 122375, "epoch": 1375, "lr": 5.785121070860504e-05}
{"train_loss": 0.15599903464317322, "global_step": 122376, "epoch": 1375, "lr": 5.785063816324969e-05}
{"train_loss": 0.16614361107349396, "global_step": 122377, "epoch": 1375, "lr": 5.785006561683893e-05}
{"train_loss": 0.07360518723726273, "global_step": 122378, "epoch": 1375, "lr": 5.784949306937283e-05}
{"train_loss": 0.1902661919593811, "global_step": 122379, "epoch": 1375, "lr": 5.7848920520851444e-05}
{"train_loss": 0.23714080452919006, "global_step": 122380, "epoch": 1375, "lr": 5.784834797127489e-05}
{"train_loss": 0.18416234850883484, "global_step": 122381, "epoch": 1375, "lr": 5.78477754206432e-05}
{"train_loss": 0.07088477909564972, "global_step": 122382, "epoch": 1375, "lr": 5.784720286895649e-05}
{"train_loss": 0.1438445895910263, "global_step": 122383, "epoch": 1375, "lr": 5.7846630316214826e-05}
{"train_loss": 0.1842341125011444, "global_step": 122384, "epoch": 1375, "lr": 5.784605776241826e-05}
{"train_loss": 0.07380803674459457, "global_step": 122385, "epoch": 1375, "lr": 5.784548520756689e-05}
{"train_loss": 0.1299983263015747, "global_step": 122386, "epoch": 1375, "lr": 5.7844912651660785e-05}
{"train_loss": 0.15334537625312805, "global_step": 122387, "epoch": 1375, "lr": 5.784434009470003e-05}
{"train_loss": 0.09194869548082352, "global_step": 122388, "epoch": 1375, "lr": 5.78437675366847e-05}
{"train_loss": 0.15244729816913605, "global_step": 122389, "epoch": 1375, "lr": 5.784319497761488e-05}
{"train_loss": 0.12988658249378204, "global_step": 122390, "epoch": 1375, "lr": 5.7842622417490635e-05}
{"train_loss": 0.11476447433233261, "global_step": 122391, "epoch": 1375, "lr": 5.784204985631203e-05}
{"train_loss": 0.11750370264053345, "global_step": 122392, "epoch": 1375, "lr": 5.784147729407917e-05}
{"train_loss": 0.1535538285970688, "global_step": 122393, "epoch": 1375, "lr": 5.7840904730792124e-05}
{"train_loss": 0.14297549426555634, "global_step": 122394, "epoch": 1375, "lr": 5.7840332166450936e-05}
{"train_loss": 0.119663767516613, "global_step": 122395, "epoch": 1375, "lr": 5.783975960105574e-05}
{"train_loss": 0.1603390872478485, "global_step": 122396, "epoch": 1375, "lr": 5.7839187034606566e-05}
{"train_loss": 0.11361104249954224, "global_step": 122397, "epoch": 1375, "lr": 5.7838614467103526e-05}
{"train_loss": 0.12194081395864487, "global_step": 122398, "epoch": 1375, "lr": 5.783804189854667e-05}
{"train_loss": 0.11653023213148117, "global_step": 122399, "epoch": 1375, "lr": 5.783746932893609e-05}
{"train_loss": 0.12944510579109192, "global_step": 122400, "epoch": 1375, "lr": 5.7836896758271854e-05}
{"train_loss": 0.09625378996133804, "global_step": 122401, "epoch": 1375, "lr": 5.7836324186554046e-05}
{"train_loss": 0.14105209708213806, "global_step": 122402, "epoch": 1375, "lr": 5.783575161378274e-05}
{"train_loss": 0.12206953018903732, "global_step": 122403, "epoch": 1375, "lr": 5.783517903995801e-05}
{"train_loss": 0.115303635597229, "global_step": 122404, "epoch": 1375, "lr": 5.7834606465079947e-05}
{"train_loss": 0.1559569537639618, "global_step": 122405, "epoch": 1375, "lr": 5.78340338891486e-05}
{"train_loss": 0.16672860085964203, "global_step": 122406, "epoch": 1375, "lr": 5.783346131216406e-05}
{"train_loss": 0.10239829123020172, "global_step": 122407, "epoch": 1375, "lr": 5.7832888734126424e-05}
{"train_loss": 0.10092195123434067, "global_step": 122408, "epoch": 1375, "lr": 5.7832316155035736e-05}
{"train_loss": 0.15802234411239624, "global_step": 122409, "epoch": 1375, "lr": 5.783174357489209e-05}
{"train_loss": 0.16968488693237305, "global_step": 122410, "epoch": 1375, "lr": 5.783117099369557e-05}
{"train_loss": 0.1556778997182846, "global_step": 122411, "epoch": 1375, "lr": 5.783059841144624e-05}
{"train_loss": 0.1148098036646843, "global_step": 122412, "epoch": 1375, "lr": 5.7830025828144176e-05}
{"train_loss": 0.10802683234214783, "global_step": 122413, "epoch": 1375, "lr": 5.782945324378947e-05}
{"train_loss": 0.08883205056190491, "global_step": 122414, "epoch": 1375, "lr": 5.782888065838218e-05}
{"train_loss": 0.10538935661315918, "global_step": 122415, "epoch": 1375, "lr": 5.78283080719224e-05}
{"train_loss": 0.10471590608358383, "global_step": 122416, "epoch": 1375, "lr": 5.782773548441019e-05}
{"train_loss": 0.13627272844314575, "global_step": 122417, "epoch": 1375, "lr": 5.782716289584566e-05}
{"train_loss": 0.16269642114639282, "global_step": 122418, "epoch": 1375, "lr": 5.782659030622884e-05}
{"train_loss": 0.09582003206014633, "global_step": 122419, "epoch": 1375, "lr": 5.782601771555983e-05}
{"train_loss": 0.16293641924858093, "global_step": 122420, "epoch": 1375, "lr": 5.782544512383872e-05}
{"train_loss": 0.11511528491973877, "global_step": 122421, "epoch": 1375, "lr": 5.7824872531065574e-05}
{"train_loss": 0.04823581129312515, "global_step": 122422, "epoch": 1375, "lr": 5.782429993724048e-05}
{"train_loss": 0.11770236492156982, "global_step": 122423, "epoch": 1375, "lr": 5.7823727342363474e-05}
{"train_loss": 0.17289763689041138, "global_step": 122424, "epoch": 1375, "lr": 5.782315474643468e-05}
{"train_loss": 0.07306472212076187, "global_step": 122425, "epoch": 1375, "lr": 5.782258214945416e-05}
{"train_loss": 0.16576305031776428, "global_step": 122426, "epoch": 1375, "lr": 5.782200955142199e-05}
{"train_loss": 0.11563347280025482, "global_step": 122427, "epoch": 1375, "lr": 5.782143695233824e-05}
{"train_loss": 0.17241188883781433, "global_step": 122428, "epoch": 1375, "lr": 5.7820864352203e-05}
{"train_loss": 0.07055675238370895, "global_step": 122429, "epoch": 1375, "lr": 5.782029175101633e-05}
{"train_loss": 0.13310304284095764, "global_step": 122430, "epoch": 1375, "lr": 5.781971914877832e-05}
{"train_loss": 0.13369442522525787, "global_step": 122431, "epoch": 1375, "lr": 5.781914654548903e-05}
{"train_loss": 0.2274252474308014, "global_step": 122432, "epoch": 1375, "lr": 5.781857394114857e-05}
{"train_loss": 0.1585356891155243, "global_step": 122433, "epoch": 1375, "lr": 5.7818001335757e-05}
{"train_loss": 0.1094677746295929, "global_step": 122434, "epoch": 1375, "lr": 5.781742872931439e-05}
{"train_loss": 0.18594831228256226, "global_step": 122435, "epoch": 1375, "lr": 5.781685612182082e-05}
{"train_loss": 0.12000445276498795, "global_step": 122436, "epoch": 1375, "lr": 5.781628351327637e-05}
{"train_loss": 0.1368161141872406, "global_step": 122437, "epoch": 1375, "lr": 5.781571090368112e-05}
{"train_loss": 0.11095017194747925, "global_step": 122438, "epoch": 1375, "lr": 5.781513829303513e-05}
{"train_loss": 0.08401578664779663, "global_step": 122439, "epoch": 1375, "lr": 5.781456568133851e-05}
{"train_loss": 0.13319356739521027, "global_step": 122440, "epoch": 1375, "lr": 5.78139930685913e-05}
{"train_loss": 0.1325090080499649, "global_step": 122441, "epoch": 1375, "lr": 5.781342045479361e-05}
{"train_loss": 0.14011017978191376, "global_step": 122442, "epoch": 1375, "lr": 5.7812847839945496e-05}
{"train_loss": 0.09599218517541885, "global_step": 122443, "epoch": 1375, "lr": 5.781227522404704e-05}
{"train_loss": 0.20149126648902893, "global_step": 122444, "epoch": 1375, "lr": 5.781170260709831e-05}
{"train_loss": 0.09605482220649719, "global_step": 122445, "epoch": 1375, "lr": 5.78111299890994e-05}
{"train_loss": 0.1419745236635208, "global_step": 122446, "epoch": 1375, "lr": 5.7810557370050366e-05}
{"train_loss": 0.0924450010061264, "global_step": 122447, "epoch": 1375, "lr": 5.780998474995131e-05}
{"train_loss": 0.17528200149536133, "global_step": 122448, "epoch": 1375, "lr": 5.78094121288023e-05}
{"train_loss": 0.13703559339046478, "global_step": 122449, "epoch": 1375, "lr": 5.780883950660339e-05}
{"train_loss": 0.07979629933834076, "global_step": 122450, "epoch": 1375, "lr": 5.780826688335469e-05}
{"train_loss": 0.1307574063539505, "global_step": 122451, "epoch": 1375, "lr": 5.7807694259056254e-05}
{"train_loss": 0.13989557325839996, "global_step": 122452, "epoch": 1375, "lr": 5.780712163370818e-05}
{"train_loss": 0.1543014794588089, "global_step": 122453, "epoch": 1375, "lr": 5.780654900731054e-05}
{"train_loss": 0.07586093991994858, "global_step": 122454, "epoch": 1375, "lr": 5.7805976379863394e-05}
{"train_loss": 0.19175049662590027, "global_step": 122455, "epoch": 1375, "lr": 5.7805403751366826e-05}
{"train_loss": 0.1687006801366806, "global_step": 122456, "epoch": 1375, "lr": 5.7804831121820926e-05}
{"train_loss": 0.13724514842033386, "global_step": 122457, "epoch": 1375, "lr": 5.780425849122576e-05}
{"train_loss": 0.14802898466587067, "global_step": 122458, "epoch": 1375, "lr": 5.7803685859581404e-05}
{"train_loss": 0.10700434446334839, "global_step": 122459, "epoch": 1375, "lr": 5.780311322688794e-05}
{"train_loss": 0.10719554126262665, "global_step": 122460, "epoch": 1375, "lr": 5.780254059314545e-05}
{"train_loss": 0.1256156861782074, "global_step": 122461, "epoch": 1375, "lr": 5.7801967958354e-05}
{"train_loss": 0.14873917400836945, "global_step": 122462, "epoch": 1375, "lr": 5.7801395322513654e-05}
{"train_loss": 0.1329505971810791, "global_step": 122463, "epoch": 1375, "lr": 5.780082268562452e-05, "val_loss": 5.412432670593262, "train_action_mse_error": 7.796274662017822}
{"train_loss": 0.1853668987751007, "global_step": 122464, "epoch": 1376, "lr": 5.780025004768665e-05}
{"train_loss": 0.17144450545310974, "global_step": 122465, "epoch": 1376, "lr": 5.7799677408700145e-05}
{"train_loss": 0.16133670508861542, "global_step": 122466, "epoch": 1376, "lr": 5.779910476866506e-05}
{"train_loss": 0.10786714404821396, "global_step": 122467, "epoch": 1376, "lr": 5.779853212758147e-05}
{"train_loss": 0.15782955288887024, "global_step": 122468, "epoch": 1376, "lr": 5.779795948544948e-05}
{"train_loss": 0.20862792432308197, "global_step": 122469, "epoch": 1376, "lr": 5.7797386842269144e-05}
{"train_loss": 0.14115053415298462, "global_step": 122470, "epoch": 1376, "lr": 5.779681419804054e-05}
{"train_loss": 0.15225227177143097, "global_step": 122471, "epoch": 1376, "lr": 5.779624155276375e-05}
{"train_loss": 0.09337559342384338, "global_step": 122472, "epoch": 1376, "lr": 5.779566890643886e-05}
{"train_loss": 0.13662150502204895, "global_step": 122473, "epoch": 1376, "lr": 5.779509625906593e-05}
{"train_loss": 0.13663476705551147, "global_step": 122474, "epoch": 1376, "lr": 5.7794523610645045e-05}
{"train_loss": 0.11049094051122665, "global_step": 122475, "epoch": 1376, "lr": 5.779395096117628e-05}
{"train_loss": 0.19166983664035797, "global_step": 122476, "epoch": 1376, "lr": 5.7793378310659716e-05}
{"train_loss": 0.14437289535999298, "global_step": 122477, "epoch": 1376, "lr": 5.7792805659095425e-05}
{"train_loss": 0.14812302589416504, "global_step": 122478, "epoch": 1376, "lr": 5.7792233006483484e-05}
{"train_loss": 0.10166340321302414, "global_step": 122479, "epoch": 1376, "lr": 5.779166035282399e-05}
{"train_loss": 0.14337562024593353, "global_step": 122480, "epoch": 1376, "lr": 5.779108769811699e-05}
{"train_loss": 0.0958530455827713, "global_step": 122481, "epoch": 1376, "lr": 5.779051504236257e-05}
{"train_loss": 0.20108327269554138, "global_step": 122482, "epoch": 1376, "lr": 5.7789942385560814e-05}
{"train_loss": 0.1377122700214386, "global_step": 122483, "epoch": 1376, "lr": 5.77893697277118e-05}
{"train_loss": 0.06398593634366989, "global_step": 122484, "epoch": 1376, "lr": 5.778879706881558e-05}
{"train_loss": 0.08154088258743286, "global_step": 122485, "epoch": 1376, "lr": 5.778822440887226e-05}
{"train_loss": 0.15648755431175232, "global_step": 122486, "epoch": 1376, "lr": 5.778765174788191e-05}
{"train_loss": 0.1495237499475479, "global_step": 122487, "epoch": 1376, "lr": 5.778707908584461e-05}
{"train_loss": 0.1485472023487091, "global_step": 122488, "epoch": 1376, "lr": 5.778650642276042e-05}
{"train_loss": 0.1260947436094284, "global_step": 122489, "epoch": 1376, "lr": 5.778593375862944e-05}
{"train_loss": 0.11750263720750809, "global_step": 122490, "epoch": 1376, "lr": 5.778536109345174e-05}
{"train_loss": 0.17124807834625244, "global_step": 122491, "epoch": 1376, "lr": 5.778478842722739e-05}
{"train_loss": 0.17161911725997925, "global_step": 122492, "epoch": 1376, "lr": 5.778421575995647e-05}
{"train_loss": 0.17530903220176697, "global_step": 122493, "epoch": 1376, "lr": 5.778364309163905e-05}
{"train_loss": 0.05546291917562485, "global_step": 122494, "epoch": 1376, "lr": 5.778307042227522e-05}
{"train_loss": 0.09891841560602188, "global_step": 122495, "epoch": 1376, "lr": 5.778249775186505e-05}
{"train_loss": 0.1632089614868164, "global_step": 122496, "epoch": 1376, "lr": 5.778192508040862e-05}
{"train_loss": 0.13637050986289978, "global_step": 122497, "epoch": 1376, "lr": 5.7781352407906e-05}
{"train_loss": 0.11478082835674286, "global_step": 122498, "epoch": 1376, "lr": 5.7780779734357274e-05}
{"train_loss": 0.085189089179039, "global_step": 122499, "epoch": 1376, "lr": 5.778020705976254e-05}
{"train_loss": 0.10376328974962234, "global_step": 122500, "epoch": 1376, "lr": 5.7779634384121816e-05}
{"train_loss": 0.14515937864780426, "global_step": 122501, "epoch": 1376, "lr": 5.777906170743523e-05}
{"train_loss": 0.14451317489147186, "global_step": 122502, "epoch": 1376, "lr": 5.777848902970284e-05}
{"train_loss": 0.09223511070013046, "global_step": 122503, "epoch": 1376, "lr": 5.777791635092473e-05}
{"train_loss": 0.11126350611448288, "global_step": 122504, "epoch": 1376, "lr": 5.777734367110097e-05}
{"train_loss": 0.1639634370803833, "global_step": 122505, "epoch": 1376, "lr": 5.777677099023164e-05}
{"train_loss": 0.1285325139760971, "global_step": 122506, "epoch": 1376, "lr": 5.7776198308316834e-05}
{"train_loss": 0.14551404118537903, "global_step": 122507, "epoch": 1376, "lr": 5.77756256253566e-05}
{"train_loss": 0.12120987474918365, "global_step": 122508, "epoch": 1376, "lr": 5.7775052941351025e-05}
{"train_loss": 0.1306547224521637, "global_step": 122509, "epoch": 1376, "lr": 5.777448025630019e-05}
{"train_loss": 0.14015479385852814, "global_step": 122510, "epoch": 1376, "lr": 5.777390757020418e-05}
{"train_loss": 0.16367964446544647, "global_step": 122511, "epoch": 1376, "lr": 5.7773334883063057e-05}
{"train_loss": 0.12362269312143326, "global_step": 122512, "epoch": 1376, "lr": 5.7772762194876904e-05}
{"train_loss": 0.1702391803264618, "global_step": 122513, "epoch": 1376, "lr": 5.7772189505645803e-05}
{"train_loss": 0.15142707526683807, "global_step": 122514, "epoch": 1376, "lr": 5.777161681536981e-05}
{"train_loss": 0.12468413263559341, "global_step": 122515, "epoch": 1376, "lr": 5.777104412404903e-05}
{"train_loss": 0.1433277279138565, "global_step": 122516, "epoch": 1376, "lr": 5.7770471431683524e-05}
{"train_loss": 0.10045366734266281, "global_step": 122517, "epoch": 1376, "lr": 5.776989873827339e-05}
{"train_loss": 0.11291014403104782, "global_step": 122518, "epoch": 1376, "lr": 5.7769326043818675e-05}
{"train_loss": 0.14457368850708008, "global_step": 122519, "epoch": 1376, "lr": 5.776875334831947e-05}
{"train_loss": 0.15392261743545532, "global_step": 122520, "epoch": 1376, "lr": 5.7768180651775846e-05}
{"train_loss": 0.12710349261760712, "global_step": 122521, "epoch": 1376, "lr": 5.776760795418787e-05}
{"train_loss": 0.09629783034324646, "global_step": 122522, "epoch": 1376, "lr": 5.7767035255555654e-05}
{"train_loss": 0.1719186156988144, "global_step": 122523, "epoch": 1376, "lr": 5.7766462555879256e-05}
{"train_loss": 0.1533159464597702, "global_step": 122524, "epoch": 1376, "lr": 5.7765889855158736e-05}
{"train_loss": 0.15231873095035553, "global_step": 122525, "epoch": 1376, "lr": 5.7765317153394205e-05}
{"train_loss": 0.11532348394393921, "global_step": 122526, "epoch": 1376, "lr": 5.7764744450585715e-05}
{"train_loss": 0.08934451639652252, "global_step": 122527, "epoch": 1376, "lr": 5.7764171746733356e-05}
{"train_loss": 0.138372540473938, "global_step": 122528, "epoch": 1376, "lr": 5.776359904183719e-05}
{"train_loss": 0.1707194745540619, "global_step": 122529, "epoch": 1376, "lr": 5.77630263358973e-05}
{"train_loss": 0.08731423318386078, "global_step": 122530, "epoch": 1376, "lr": 5.776245362891377e-05}
{"train_loss": 0.14365151524543762, "global_step": 122531, "epoch": 1376, "lr": 5.776188092088668e-05}
{"train_loss": 0.11980018019676208, "global_step": 122532, "epoch": 1376, "lr": 5.7761308211816086e-05}
{"train_loss": 0.1478966772556305, "global_step": 122533, "epoch": 1376, "lr": 5.776073550170209e-05}
{"train_loss": 0.18753927946090698, "global_step": 122534, "epoch": 1376, "lr": 5.7760162790544746e-05}
{"train_loss": 0.10476735234260559, "global_step": 122535, "epoch": 1376, "lr": 5.775959007834415e-05}
{"train_loss": 0.0814099833369255, "global_step": 122536, "epoch": 1376, "lr": 5.775901736510038e-05}
{"train_loss": 0.1341642290353775, "global_step": 122537, "epoch": 1376, "lr": 5.77584446508135e-05}
{"train_loss": 0.12472832947969437, "global_step": 122538, "epoch": 1376, "lr": 5.775787193548359e-05}
{"train_loss": 0.0943983718752861, "global_step": 122539, "epoch": 1376, "lr": 5.7757299219110714e-05}
{"train_loss": 0.08595424890518188, "global_step": 122540, "epoch": 1376, "lr": 5.775672650169498e-05}
{"train_loss": 0.11418958008289337, "global_step": 122541, "epoch": 1376, "lr": 5.7756153783236444e-05}
{"train_loss": 0.1849076896905899, "global_step": 122542, "epoch": 1376, "lr": 5.7755581063735185e-05}
{"train_loss": 0.15479230880737305, "global_step": 122543, "epoch": 1376, "lr": 5.775500834319128e-05}
{"train_loss": 0.14269782602787018, "global_step": 122544, "epoch": 1376, "lr": 5.775443562160483e-05}
{"train_loss": 0.15179450809955597, "global_step": 122545, "epoch": 1376, "lr": 5.775386289897587e-05}
{"train_loss": 0.16034986078739166, "global_step": 122546, "epoch": 1376, "lr": 5.775329017530451e-05}
{"train_loss": 0.11261504888534546, "global_step": 122547, "epoch": 1376, "lr": 5.77527174505908e-05}
{"train_loss": 0.11347034573554993, "global_step": 122548, "epoch": 1376, "lr": 5.7752144724834844e-05}
{"train_loss": 0.10199066251516342, "global_step": 122549, "epoch": 1376, "lr": 5.77515719980367e-05}
{"train_loss": 0.0981893241405487, "global_step": 122550, "epoch": 1376, "lr": 5.775099927019645e-05}
{"train_loss": 0.1488044559955597, "global_step": 122551, "epoch": 1376, "lr": 5.775042654131417e-05}
{"train_loss": 0.13353599172629668, "global_step": 122552, "epoch": 1376, "lr": 5.7749853811389956e-05, "val_loss": 5.56937837600708}
{"train_loss": 0.13753288984298706, "global_step": 122553, "epoch": 1377, "lr": 5.7749281080423855e-05}
{"train_loss": 0.16354621946811676, "global_step": 122554, "epoch": 1377, "lr": 5.774870834841597e-05}
{"train_loss": 0.13584330677986145, "global_step": 122555, "epoch": 1377, "lr": 5.774813561536635e-05}
{"train_loss": 0.05903171747922897, "global_step": 122556, "epoch": 1377, "lr": 5.77475628812751e-05}
{"train_loss": 0.13133744895458221, "global_step": 122557, "epoch": 1377, "lr": 5.774699014614229e-05}
{"train_loss": 0.15392014384269714, "global_step": 122558, "epoch": 1377, "lr": 5.7746417409967965e-05}
{"train_loss": 0.07342265546321869, "global_step": 122559, "epoch": 1377, "lr": 5.7745844672752245e-05}
{"train_loss": 0.18045803904533386, "global_step": 122560, "epoch": 1377, "lr": 5.7745271934495196e-05}
{"train_loss": 0.1116335391998291, "global_step": 122561, "epoch": 1377, "lr": 5.774469919519687e-05}
{"train_loss": 0.10009578615427017, "global_step": 122562, "epoch": 1377, "lr": 5.774412645485738e-05}
{"train_loss": 0.13819877803325653, "global_step": 122563, "epoch": 1377, "lr": 5.774355371347678e-05}
{"train_loss": 0.12564243376255035, "global_step": 122564, "epoch": 1377, "lr": 5.7742980971055164e-05}
{"train_loss": 0.1116647943854332, "global_step": 122565, "epoch": 1377, "lr": 5.774240822759259e-05}
{"train_loss": 0.10008395463228226, "global_step": 122566, "epoch": 1377, "lr": 5.774183548308915e-05}
{"train_loss": 0.10582304000854492, "global_step": 122567, "epoch": 1377, "lr": 5.774126273754491e-05}
{"train_loss": 0.11162041872739792, "global_step": 122568, "epoch": 1377, "lr": 5.774068999095995e-05}
{"train_loss": 0.07931672036647797, "global_step": 122569, "epoch": 1377, "lr": 5.7740117243334344e-05}
{"train_loss": 0.09150099754333496, "global_step": 122570, "epoch": 1377, "lr": 5.773954449466818e-05}
{"train_loss": 0.19860593974590302, "global_step": 122571, "epoch": 1377, "lr": 5.773897174496153e-05}
{"train_loss": 0.14516764879226685, "global_step": 122572, "epoch": 1377, "lr": 5.7738398994214474e-05}
{"train_loss": 0.13443785905838013, "global_step": 122573, "epoch": 1377, "lr": 5.7737826242427076e-05}
{"train_loss": 0.12930220365524292, "global_step": 122574, "epoch": 1377, "lr": 5.773725348959942e-05}
{"train_loss": 0.1523972600698471, "global_step": 122575, "epoch": 1377, "lr": 5.7736680735731596e-05}
{"train_loss": 0.15108206868171692, "global_step": 122576, "epoch": 1377, "lr": 5.773610798082366e-05}
{"train_loss": 0.09395959228277206, "global_step": 122577, "epoch": 1377, "lr": 5.773553522487569e-05}
{"train_loss": 0.13476774096488953, "global_step": 122578, "epoch": 1377, "lr": 5.7734962467887784e-05}
{"train_loss": 0.12246661633253098, "global_step": 122579, "epoch": 1377, "lr": 5.773438970986002e-05}
{"train_loss": 0.1143440455198288, "global_step": 122580, "epoch": 1377, "lr": 5.773381695079244e-05}
{"train_loss": 0.12430339306592941, "global_step": 122581, "epoch": 1377, "lr": 5.773324419068514e-05}
{"train_loss": 0.14788515865802765, "global_step": 122582, "epoch": 1377, "lr": 5.773267142953822e-05}
{"train_loss": 0.14282938838005066, "global_step": 122583, "epoch": 1377, "lr": 5.7732098667351716e-05}
{"train_loss": 0.08509508520364761, "global_step": 122584, "epoch": 1377, "lr": 5.7731525904125735e-05}
{"train_loss": 0.1644553244113922, "global_step": 122585, "epoch": 1377, "lr": 5.7730953139860344e-05}
{"train_loss": 0.1515883207321167, "global_step": 122586, "epoch": 1377, "lr": 5.773038037455563e-05}
{"train_loss": 0.09926068037748337, "global_step": 122587, "epoch": 1377, "lr": 5.772980760821165e-05}
{"train_loss": 0.11405336111783981, "global_step": 122588, "epoch": 1377, "lr": 5.772923484082848e-05}
{"train_loss": 0.14393413066864014, "global_step": 122589, "epoch": 1377, "lr": 5.772866207240624e-05}
{"train_loss": 0.1332748830318451, "global_step": 122590, "epoch": 1377, "lr": 5.772808930294494e-05}
{"train_loss": 0.12949039041996002, "global_step": 122591, "epoch": 1377, "lr": 5.772751653244472e-05}
{"train_loss": 0.10815802961587906, "global_step": 122592, "epoch": 1377, "lr": 5.7726943760905616e-05}
{"train_loss": 0.08653530478477478, "global_step": 122593, "epoch": 1377, "lr": 5.772637098832773e-05}
{"train_loss": 0.13666272163391113, "global_step": 122594, "epoch": 1377, "lr": 5.772579821471111e-05}
{"train_loss": 0.06128882244229317, "global_step": 122595, "epoch": 1377, "lr": 5.7725225440055865e-05}
{"train_loss": 0.060830555856227875, "global_step": 122596, "epoch": 1377, "lr": 5.772465266436205e-05}
{"train_loss": 0.18994106352329254, "global_step": 122597, "epoch": 1377, "lr": 5.7724079887629753e-05}
{"train_loss": 0.07722637802362442, "global_step": 122598, "epoch": 1377, "lr": 5.7723507109859045e-05}
{"train_loss": 0.19230815768241882, "global_step": 122599, "epoch": 1377, "lr": 5.772293433105e-05}
{"train_loss": 0.10940278321504593, "global_step": 122600, "epoch": 1377, "lr": 5.772236155120272e-05}
{"train_loss": 0.1063631922006607, "global_step": 122601, "epoch": 1377, "lr": 5.772178877031724e-05}
{"train_loss": 0.12286209315061569, "global_step": 122602, "epoch": 1377, "lr": 5.772121598839367e-05}
{"train_loss": 0.10452772676944733, "global_step": 122603, "epoch": 1377, "lr": 5.772064320543209e-05}
{"train_loss": 0.12933921813964844, "global_step": 122604, "epoch": 1377, "lr": 5.772007042143255e-05}
{"train_loss": 0.11473998427391052, "global_step": 122605, "epoch": 1377, "lr": 5.771949763639514e-05}
{"train_loss": 0.09080580621957779, "global_step": 122606, "epoch": 1377, "lr": 5.7718924850319936e-05}
{"train_loss": 0.08038599789142609, "global_step": 122607, "epoch": 1377, "lr": 5.7718352063207016e-05}
{"train_loss": 0.14491935074329376, "global_step": 122608, "epoch": 1377, "lr": 5.771777927505646e-05}
{"train_loss": 0.10924210399389267, "global_step": 122609, "epoch": 1377, "lr": 5.7717206485868344e-05}
{"train_loss": 0.18892420828342438, "global_step": 122610, "epoch": 1377, "lr": 5.7716633695642755e-05}
{"train_loss": 0.09893660247325897, "global_step": 122611, "epoch": 1377, "lr": 5.7716060904379746e-05}
{"train_loss": 0.13502384722232819, "global_step": 122612, "epoch": 1377, "lr": 5.77154881120794e-05}
{"train_loss": 0.15033113956451416, "global_step": 122613, "epoch": 1377, "lr": 5.771491531874181e-05}
{"train_loss": 0.09901638329029083, "global_step": 122614, "epoch": 1377, "lr": 5.771434252436704e-05}
{"train_loss": 0.07900669425725937, "global_step": 122615, "epoch": 1377, "lr": 5.771376972895517e-05}
{"train_loss": 0.1663224697113037, "global_step": 122616, "epoch": 1377, "lr": 5.771319693250629e-05}
{"train_loss": 0.10404525697231293, "global_step": 122617, "epoch": 1377, "lr": 5.771262413502045e-05}
{"train_loss": 0.1545620709657669, "global_step": 122618, "epoch": 1377, "lr": 5.771205133649774e-05}
{"train_loss": 0.1739870011806488, "global_step": 122619, "epoch": 1377, "lr": 5.7711478536938256e-05}
{"train_loss": 0.09708551317453384, "global_step": 122620, "epoch": 1377, "lr": 5.771090573634205e-05}
{"train_loss": 0.09361957758665085, "global_step": 122621, "epoch": 1377, "lr": 5.77103329347092e-05}
{"train_loss": 0.12437597662210464, "global_step": 122622, "epoch": 1377, "lr": 5.770976013203979e-05}
{"train_loss": 0.21529410779476166, "global_step": 122623, "epoch": 1377, "lr": 5.77091873283339e-05}
{"train_loss": 0.1607835441827774, "global_step": 122624, "epoch": 1377, "lr": 5.770861452359161e-05}
{"train_loss": 0.16291247308254242, "global_step": 122625, "epoch": 1377, "lr": 5.770804171781298e-05}
{"train_loss": 0.10186357796192169, "global_step": 122626, "epoch": 1377, "lr": 5.77074689109981e-05}
{"train_loss": 0.14746306836605072, "global_step": 122627, "epoch": 1377, "lr": 5.770689610314705e-05}
{"train_loss": 0.1120992973446846, "global_step": 122628, "epoch": 1377, "lr": 5.77063232942599e-05}
{"train_loss": 0.12432608753442764, "global_step": 122629, "epoch": 1377, "lr": 5.770575048433673e-05}
{"train_loss": 0.14670394361019135, "global_step": 122630, "epoch": 1377, "lr": 5.770517767337761e-05}
{"train_loss": 0.12524330615997314, "global_step": 122631, "epoch": 1377, "lr": 5.770460486138263e-05}
{"train_loss": 0.16398684680461884, "global_step": 122632, "epoch": 1377, "lr": 5.7704032048351855e-05}
{"train_loss": 0.13928815722465515, "global_step": 122633, "epoch": 1377, "lr": 5.7703459234285365e-05}
{"train_loss": 0.07248638570308685, "global_step": 122634, "epoch": 1377, "lr": 5.7702886419183244e-05}
{"train_loss": 0.1534273326396942, "global_step": 122635, "epoch": 1377, "lr": 5.7702313603045554e-05}
{"train_loss": 0.124705970287323, "global_step": 122636, "epoch": 1377, "lr": 5.770174078587239e-05}
{"train_loss": 0.12671592831611633, "global_step": 122637, "epoch": 1377, "lr": 5.770116796766383e-05}
{"train_loss": 0.1015617623925209, "global_step": 122638, "epoch": 1377, "lr": 5.770059514841992e-05}
{"train_loss": 0.10535562038421631, "global_step": 122639, "epoch": 1377, "lr": 5.770002232814078e-05}
{"train_loss": 0.13975974917411804, "global_step": 122640, "epoch": 1377, "lr": 5.769944950682645e-05}
{"train_loss": 0.12483861268068967, "global_step": 122641, "epoch": 1377, "lr": 5.769887668447703e-05, "val_loss": 5.530478477478027}
{"train_loss": 0.1390244960784912, "global_step": 122642, "epoch": 1378, "lr": 5.7698303861092594e-05}
{"train_loss": 0.1341167390346527, "global_step": 122643, "epoch": 1378, "lr": 5.7697731036673206e-05}
{"train_loss": 0.12969399988651276, "global_step": 122644, "epoch": 1378, "lr": 5.769715821121896e-05}
{"train_loss": 0.12317603826522827, "global_step": 122645, "epoch": 1378, "lr": 5.769658538472992e-05}
{"train_loss": 0.11868982762098312, "global_step": 122646, "epoch": 1378, "lr": 5.769601255720617e-05}
{"train_loss": 0.11565734446048737, "global_step": 122647, "epoch": 1378, "lr": 5.7695439728647784e-05}
{"train_loss": 0.16588261723518372, "global_step": 122648, "epoch": 1378, "lr": 5.7694866899054833e-05}
{"train_loss": 0.12997274100780487, "global_step": 122649, "epoch": 1378, "lr": 5.769429406842741e-05}
{"train_loss": 0.10888651013374329, "global_step": 122650, "epoch": 1378, "lr": 5.769372123676559e-05}
{"train_loss": 0.08797944337129593, "global_step": 122651, "epoch": 1378, "lr": 5.7693148404069434e-05}
{"train_loss": 0.10562124103307724, "global_step": 122652, "epoch": 1378, "lr": 5.7692575570339026e-05}
{"train_loss": 0.10164695233106613, "global_step": 122653, "epoch": 1378, "lr": 5.769200273557444e-05}
{"train_loss": 0.14375187456607819, "global_step": 122654, "epoch": 1378, "lr": 5.7691429899775765e-05}
{"train_loss": 0.12552309036254883, "global_step": 122655, "epoch": 1378, "lr": 5.769085706294307e-05}
{"train_loss": 0.11074839532375336, "global_step": 122656, "epoch": 1378, "lr": 5.769028422507643e-05}
{"train_loss": 0.17760069668293, "global_step": 122657, "epoch": 1378, "lr": 5.7689711386175935e-05}
{"train_loss": 0.17496193945407867, "global_step": 122658, "epoch": 1378, "lr": 5.768913854624164e-05}
{"train_loss": 0.17292644083499908, "global_step": 122659, "epoch": 1378, "lr": 5.768856570527365e-05}
{"train_loss": 0.11064862459897995, "global_step": 122660, "epoch": 1378, "lr": 5.768799286327201e-05}
{"train_loss": 0.17180965840816498, "global_step": 122661, "epoch": 1378, "lr": 5.768742002023682e-05}
{"train_loss": 0.1244458332657814, "global_step": 122662, "epoch": 1378, "lr": 5.768684717616815e-05}
{"train_loss": 0.13247427344322205, "global_step": 122663, "epoch": 1378, "lr": 5.768627433106607e-05}
{"train_loss": 0.1343390792608261, "global_step": 122664, "epoch": 1378, "lr": 5.7685701484930664e-05}
{"train_loss": 0.12137963622808456, "global_step": 122665, "epoch": 1378, "lr": 5.7685128637762024e-05}
{"train_loss": 0.11983244866132736, "global_step": 122666, "epoch": 1378, "lr": 5.768455578956019e-05}
{"train_loss": 0.17354810237884521, "global_step": 122667, "epoch": 1378, "lr": 5.768398294032529e-05}
{"train_loss": 0.19952549040317535, "global_step": 122668, "epoch": 1378, "lr": 5.7683410090057354e-05}
{"train_loss": 0.12332961708307266, "global_step": 122669, "epoch": 1378, "lr": 5.768283723875648e-05}
{"train_loss": 0.13213306665420532, "global_step": 122670, "epoch": 1378, "lr": 5.7682264386422736e-05}
{"train_loss": 0.08046787232160568, "global_step": 122671, "epoch": 1378, "lr": 5.768169153305621e-05}
{"train_loss": 0.1621677279472351, "global_step": 122672, "epoch": 1378, "lr": 5.7681118678656975e-05}
{"train_loss": 0.08153393864631653, "global_step": 122673, "epoch": 1378, "lr": 5.768054582322511e-05}
{"train_loss": 0.16742922365665436, "global_step": 122674, "epoch": 1378, "lr": 5.767997296676069e-05}
{"train_loss": 0.11057864874601364, "global_step": 122675, "epoch": 1378, "lr": 5.767940010926378e-05}
{"train_loss": 0.11221620440483093, "global_step": 122676, "epoch": 1378, "lr": 5.767882725073448e-05}
{"train_loss": 0.16663557291030884, "global_step": 122677, "epoch": 1378, "lr": 5.7678254391172846e-05}
{"train_loss": 0.2732251286506653, "global_step": 122678, "epoch": 1378, "lr": 5.767768153057898e-05}
{"train_loss": 0.17060960829257965, "global_step": 122679, "epoch": 1378, "lr": 5.767710866895293e-05}
{"train_loss": 0.13730476796627045, "global_step": 122680, "epoch": 1378, "lr": 5.7676535806294794e-05}
{"train_loss": 0.15733392536640167, "global_step": 122681, "epoch": 1378, "lr": 5.7675962942604634e-05}
{"train_loss": 0.12019769102334976, "global_step": 122682, "epoch": 1378, "lr": 5.767539007788254e-05}
{"train_loss": 0.24639543890953064, "global_step": 122683, "epoch": 1378, "lr": 5.767481721212857e-05}
{"train_loss": 0.17769287526607513, "global_step": 122684, "epoch": 1378, "lr": 5.767424434534283e-05}
{"train_loss": 0.18212942779064178, "global_step": 122685, "epoch": 1378, "lr": 5.767367147752536e-05}
{"train_loss": 0.13323237001895905, "global_step": 122686, "epoch": 1378, "lr": 5.767309860867628e-05}
{"train_loss": 0.10420040041208267, "global_step": 122687, "epoch": 1378, "lr": 5.7672525738795644e-05}
{"train_loss": 0.1491168886423111, "global_step": 122688, "epoch": 1378, "lr": 5.767195286788354e-05}
{"train_loss": 0.18115456402301788, "global_step": 122689, "epoch": 1378, "lr": 5.767137999594001e-05}
{"train_loss": 0.1810285896062851, "global_step": 122690, "epoch": 1378, "lr": 5.767080712296518e-05}
{"train_loss": 0.1631326824426651, "global_step": 122691, "epoch": 1378, "lr": 5.7670234248959086e-05}
{"train_loss": 0.221286341547966, "global_step": 122692, "epoch": 1378, "lr": 5.766966137392182e-05}
{"train_loss": 0.14086145162582397, "global_step": 122693, "epoch": 1378, "lr": 5.7669088497853475e-05}
{"train_loss": 0.18081305921077728, "global_step": 122694, "epoch": 1378, "lr": 5.7668515620754106e-05}
{"train_loss": 0.22735659778118134, "global_step": 122695, "epoch": 1378, "lr": 5.76679427426238e-05}
{"train_loss": 0.11920036375522614, "global_step": 122696, "epoch": 1378, "lr": 5.766736986346264e-05}
{"train_loss": 0.16518521308898926, "global_step": 122697, "epoch": 1378, "lr": 5.7666796983270686e-05}
{"train_loss": 0.15198194980621338, "global_step": 122698, "epoch": 1378, "lr": 5.766622410204803e-05}
{"train_loss": 0.1152338981628418, "global_step": 122699, "epoch": 1378, "lr": 5.766565121979475e-05}
{"train_loss": 0.12213432043790817, "global_step": 122700, "epoch": 1378, "lr": 5.76650783365109e-05}
{"train_loss": 0.09304334968328476, "global_step": 122701, "epoch": 1378, "lr": 5.766450545219658e-05}
{"train_loss": 0.19058024883270264, "global_step": 122702, "epoch": 1378, "lr": 5.766393256685187e-05}
{"train_loss": 0.0844874456524849, "global_step": 122703, "epoch": 1378, "lr": 5.7663359680476815e-05}
{"train_loss": 0.13446512818336487, "global_step": 122704, "epoch": 1378, "lr": 5.7662786793071536e-05}
{"train_loss": 0.15123900771141052, "global_step": 122705, "epoch": 1378, "lr": 5.766221390463609e-05}
{"train_loss": 0.15348346531391144, "global_step": 122706, "epoch": 1378, "lr": 5.766164101517055e-05}
{"train_loss": 0.1013205349445343, "global_step": 122707, "epoch": 1378, "lr": 5.766106812467499e-05}
{"train_loss": 0.10933896154165268, "global_step": 122708, "epoch": 1378, "lr": 5.766049523314949e-05}
{"train_loss": 0.1818348467350006, "global_step": 122709, "epoch": 1378, "lr": 5.7659922340594144e-05}
{"train_loss": 0.12469763308763504, "global_step": 122710, "epoch": 1378, "lr": 5.765934944700899e-05}
{"train_loss": 0.07577086985111237, "global_step": 122711, "epoch": 1378, "lr": 5.7658776552394155e-05}
{"train_loss": 0.13427267968654633, "global_step": 122712, "epoch": 1378, "lr": 5.765820365674968e-05}
{"train_loss": 0.1286492794752121, "global_step": 122713, "epoch": 1378, "lr": 5.7657630760075654e-05}
{"train_loss": 0.18374618887901306, "global_step": 122714, "epoch": 1378, "lr": 5.7657057862372155e-05}
{"train_loss": 0.14004869759082794, "global_step": 122715, "epoch": 1378, "lr": 5.7656484963639246e-05}
{"train_loss": 0.17686651647090912, "global_step": 122716, "epoch": 1378, "lr": 5.765591206387703e-05}
{"train_loss": 0.1686818152666092, "global_step": 122717, "epoch": 1378, "lr": 5.7655339163085566e-05}
{"train_loss": 0.2066519856452942, "global_step": 122718, "epoch": 1378, "lr": 5.765476626126494e-05}
{"train_loss": 0.2600703537464142, "global_step": 122719, "epoch": 1378, "lr": 5.765419335841521e-05}
{"train_loss": 0.06425495445728302, "global_step": 122720, "epoch": 1378, "lr": 5.7653620454536475e-05}
{"train_loss": 0.1443827599287033, "global_step": 122721, "epoch": 1378, "lr": 5.765304754962879e-05}
{"train_loss": 0.15930034220218658, "global_step": 122722, "epoch": 1378, "lr": 5.7652474643692266e-05}
{"train_loss": 0.11273069679737091, "global_step": 122723, "epoch": 1378, "lr": 5.7651901736726956e-05}
{"train_loss": 0.13908104598522186, "global_step": 122724, "epoch": 1378, "lr": 5.7651328828732934e-05}
{"train_loss": 0.14450453221797943, "global_step": 122725, "epoch": 1378, "lr": 5.7650755919710296e-05}
{"train_loss": 0.18806196749210358, "global_step": 122726, "epoch": 1378, "lr": 5.765018300965911e-05}
{"train_loss": 0.08549975603818893, "global_step": 122727, "epoch": 1378, "lr": 5.7649610098579434e-05}
{"train_loss": 0.1092284694314003, "global_step": 122728, "epoch": 1378, "lr": 5.764903718647137e-05}
{"train_loss": 0.12921175360679626, "global_step": 122729, "epoch": 1378, "lr": 5.764846427333498e-05}
{"train_loss": 0.14404365453827248, "global_step": 122730, "epoch": 1378, "lr": 5.764789135917036e-05, "val_loss": 5.462221622467041}
{"train_loss": 0.12080223858356476, "global_step": 122731, "epoch": 1379, "lr": 5.764731844397756e-05}
{"train_loss": 0.18439283967018127, "global_step": 122732, "epoch": 1379, "lr": 5.764674552775667e-05}
{"train_loss": 0.17018629610538483, "global_step": 122733, "epoch": 1379, "lr": 5.764617261050778e-05}
{"train_loss": 0.09139472991228104, "global_step": 122734, "epoch": 1379, "lr": 5.7645599692230954e-05}
{"train_loss": 0.09120769053697586, "global_step": 122735, "epoch": 1379, "lr": 5.764502677292627e-05}
{"train_loss": 0.14291304349899292, "global_step": 122736, "epoch": 1379, "lr": 5.764445385259379e-05}
{"train_loss": 0.08628594875335693, "global_step": 122737, "epoch": 1379, "lr": 5.764388093123363e-05}
{"train_loss": 0.10536795109510422, "global_step": 122738, "epoch": 1379, "lr": 5.764330800884582e-05}
{"train_loss": 0.15643702447414398, "global_step": 122739, "epoch": 1379, "lr": 5.764273508543047e-05}
{"train_loss": 0.2124149054288864, "global_step": 122740, "epoch": 1379, "lr": 5.764216216098764e-05}
{"train_loss": 0.10639716684818268, "global_step": 122741, "epoch": 1379, "lr": 5.764158923551743e-05}
{"train_loss": 0.1741785854101181, "global_step": 122742, "epoch": 1379, "lr": 5.764101630901989e-05}
{"train_loss": 0.17259959876537323, "global_step": 122743, "epoch": 1379, "lr": 5.764044338149511e-05}
{"train_loss": 0.12572582066059113, "global_step": 122744, "epoch": 1379, "lr": 5.763987045294318e-05}
{"train_loss": 0.15710613131523132, "global_step": 122745, "epoch": 1379, "lr": 5.7639297523364156e-05}
{"train_loss": 0.1731090545654297, "global_step": 122746, "epoch": 1379, "lr": 5.763872459275812e-05}
{"train_loss": 0.13200020790100098, "global_step": 122747, "epoch": 1379, "lr": 5.763815166112515e-05}
{"train_loss": 0.09978267550468445, "global_step": 122748, "epoch": 1379, "lr": 5.7637578728465314e-05}
{"train_loss": 0.07524240016937256, "global_step": 122749, "epoch": 1379, "lr": 5.763700579477871e-05}
{"train_loss": 0.09006239473819733, "global_step": 122750, "epoch": 1379, "lr": 5.7636432860065407e-05}
{"train_loss": 0.21578148007392883, "global_step": 122751, "epoch": 1379, "lr": 5.7635859924325465e-05}
{"train_loss": 0.19331689178943634, "global_step": 122752, "epoch": 1379, "lr": 5.763528698755898e-05}
{"train_loss": 0.14373064041137695, "global_step": 122753, "epoch": 1379, "lr": 5.763471404976602e-05}
{"train_loss": 0.1260654330253601, "global_step": 122754, "epoch": 1379, "lr": 5.763414111094667e-05}
{"train_loss": 0.18135273456573486, "global_step": 122755, "epoch": 1379, "lr": 5.763356817110102e-05}
{"train_loss": 0.1890103667974472, "global_step": 122756, "epoch": 1379, "lr": 5.7632995230229106e-05}
{"train_loss": 0.08152100443840027, "global_step": 122757, "epoch": 1379, "lr": 5.7632422288331026e-05}
{"train_loss": 0.2170644998550415, "global_step": 122758, "epoch": 1379, "lr": 5.763184934540686e-05}
{"train_loss": 0.10263831168413162, "global_step": 122759, "epoch": 1379, "lr": 5.76312764014567e-05}
{"train_loss": 0.14740732312202454, "global_step": 122760, "epoch": 1379, "lr": 5.763070345648061e-05}
{"train_loss": 0.11425377428531647, "global_step": 122761, "epoch": 1379, "lr": 5.7630130510478654e-05}
{"train_loss": 0.11035041511058807, "global_step": 122762, "epoch": 1379, "lr": 5.762955756345092e-05}
{"train_loss": 0.09588541835546494, "global_step": 122763, "epoch": 1379, "lr": 5.7628984615397486e-05}
{"train_loss": 0.12191396206617355, "global_step": 122764, "epoch": 1379, "lr": 5.762841166631845e-05}
{"train_loss": 0.14554256200790405, "global_step": 122765, "epoch": 1379, "lr": 5.7627838716213846e-05}
{"train_loss": 0.0931953638792038, "global_step": 122766, "epoch": 1379, "lr": 5.762726576508377e-05}
{"train_loss": 0.1429290771484375, "global_step": 122767, "epoch": 1379, "lr": 5.7626692812928304e-05}
{"train_loss": 0.21366451680660248, "global_step": 122768, "epoch": 1379, "lr": 5.762611985974752e-05}
{"train_loss": 0.10278400033712387, "global_step": 122769, "epoch": 1379, "lr": 5.76255469055415e-05}
{"train_loss": 0.1583930253982544, "global_step": 122770, "epoch": 1379, "lr": 5.7624973950310324e-05}
{"train_loss": 0.1065944954752922, "global_step": 122771, "epoch": 1379, "lr": 5.762440099405405e-05}
{"train_loss": 0.20634648203849792, "global_step": 122772, "epoch": 1379, "lr": 5.7623828036772775e-05}
{"train_loss": 0.15031111240386963, "global_step": 122773, "epoch": 1379, "lr": 5.7623255078466575e-05}
{"train_loss": 0.15463843941688538, "global_step": 122774, "epoch": 1379, "lr": 5.762268211913552e-05}
{"train_loss": 0.1266724169254303, "global_step": 122775, "epoch": 1379, "lr": 5.762210915877967e-05}
{"train_loss": 0.07980446517467499, "global_step": 122776, "epoch": 1379, "lr": 5.7621536197399127e-05}
{"train_loss": 0.11156065762042999, "global_step": 122777, "epoch": 1379, "lr": 5.762096323499397e-05}
{"train_loss": 0.13246206939220428, "global_step": 122778, "epoch": 1379, "lr": 5.762039027156427e-05}
{"train_loss": 0.08532410860061646, "global_step": 122779, "epoch": 1379, "lr": 5.7619817307110104e-05}
{"train_loss": 0.09766064584255219, "global_step": 122780, "epoch": 1379, "lr": 5.761924434163153e-05}
{"train_loss": 0.14258727431297302, "global_step": 122781, "epoch": 1379, "lr": 5.761867137512865e-05}
{"train_loss": 0.1734580546617508, "global_step": 122782, "epoch": 1379, "lr": 5.761809840760154e-05}
{"train_loss": 0.12246999889612198, "global_step": 122783, "epoch": 1379, "lr": 5.7617525439050255e-05}
{"train_loss": 0.16339151561260223, "global_step": 122784, "epoch": 1379, "lr": 5.761695246947492e-05}
{"train_loss": 0.10856576263904572, "global_step": 122785, "epoch": 1379, "lr": 5.761637949887555e-05}
{"train_loss": 0.18397951126098633, "global_step": 122786, "epoch": 1379, "lr": 5.761580652725225e-05}
{"train_loss": 0.12808938324451447, "global_step": 122787, "epoch": 1379, "lr": 5.761523355460511e-05}
{"train_loss": 0.12856774032115936, "global_step": 122788, "epoch": 1379, "lr": 5.761466058093418e-05}
{"train_loss": 0.12836958467960358, "global_step": 122789, "epoch": 1379, "lr": 5.7614087606239565e-05}
{"train_loss": 0.1794845461845398, "global_step": 122790, "epoch": 1379, "lr": 5.7613514630521316e-05}
{"train_loss": 0.17268329858779907, "global_step": 122791, "epoch": 1379, "lr": 5.761294165377953e-05}
{"train_loss": 0.17623524367809296, "global_step": 122792, "epoch": 1379, "lr": 5.7612368676014275e-05}
{"train_loss": 0.07956421375274658, "global_step": 122793, "epoch": 1379, "lr": 5.7611795697225625e-05}
{"train_loss": 0.12632009387016296, "global_step": 122794, "epoch": 1379, "lr": 5.761122271741366e-05}
{"train_loss": 0.0871361643075943, "global_step": 122795, "epoch": 1379, "lr": 5.761064973657847e-05}
{"train_loss": 0.08449051529169083, "global_step": 122796, "epoch": 1379, "lr": 5.761007675472011e-05}
{"train_loss": 0.1179036796092987, "global_step": 122797, "epoch": 1379, "lr": 5.760950377183868e-05}
{"train_loss": 0.11273892968893051, "global_step": 122798, "epoch": 1379, "lr": 5.760893078793424e-05}
{"train_loss": 0.1437346637248993, "global_step": 122799, "epoch": 1379, "lr": 5.7608357803006876e-05}
{"train_loss": 0.14978249371051788, "global_step": 122800, "epoch": 1379, "lr": 5.760778481705665e-05}
{"train_loss": 0.1253667026758194, "global_step": 122801, "epoch": 1379, "lr": 5.7607211830083665e-05}
{"train_loss": 0.13732635974884033, "global_step": 122802, "epoch": 1379, "lr": 5.7606638842087964e-05}
{"train_loss": 0.18011826276779175, "global_step": 122803, "epoch": 1379, "lr": 5.760606585306967e-05}
{"train_loss": 0.09926120936870575, "global_step": 122804, "epoch": 1379, "lr": 5.760549286302882e-05}
{"train_loss": 0.06794282048940659, "global_step": 122805, "epoch": 1379, "lr": 5.7604919871965504e-05}
{"train_loss": 0.15419062972068787, "global_step": 122806, "epoch": 1379, "lr": 5.7604346879879786e-05}
{"train_loss": 0.1867283135652542, "global_step": 122807, "epoch": 1379, "lr": 5.7603773886771774e-05}
{"train_loss": 0.05337660387158394, "global_step": 122808, "epoch": 1379, "lr": 5.7603200892641515e-05}
{"train_loss": 0.14683397114276886, "global_step": 122809, "epoch": 1379, "lr": 5.7602627897489104e-05}
{"train_loss": 0.12393828481435776, "global_step": 122810, "epoch": 1379, "lr": 5.7602054901314616e-05}
{"train_loss": 0.1740543246269226, "global_step": 122811, "epoch": 1379, "lr": 5.760148190411812e-05}
{"train_loss": 0.19350244104862213, "global_step": 122812, "epoch": 1379, "lr": 5.7600908905899685e-05}
{"train_loss": 0.13959215581417084, "global_step": 122813, "epoch": 1379, "lr": 5.760033590665942e-05}
{"train_loss": 0.1371130645275116, "global_step": 122814, "epoch": 1379, "lr": 5.759976290639737e-05}
{"train_loss": 0.12195689976215363, "global_step": 122815, "epoch": 1379, "lr": 5.759918990511363e-05}
{"train_loss": 0.17270946502685547, "global_step": 122816, "epoch": 1379, "lr": 5.7598616902808276e-05}
{"train_loss": 0.11284578591585159, "global_step": 122817, "epoch": 1379, "lr": 5.759804389948137e-05}
{"train_loss": 0.13710875809192657, "global_step": 122818, "epoch": 1379, "lr": 5.759747089513301e-05}
{"train_loss": 0.13679915486594266, "global_step": 122819, "epoch": 1379, "lr": 5.7596897889763255e-05, "val_loss": 5.601475238800049}
{"train_loss": 0.10925500094890594, "global_step": 122820, "epoch": 1380, "lr": 5.75963248833722e-05}
{"train_loss": 0.0920158177614212, "global_step": 122821, "epoch": 1380, "lr": 5.75957518759599e-05}
{"train_loss": 0.13962653279304504, "global_step": 122822, "epoch": 1380, "lr": 5.7595178867526456e-05}
{"train_loss": 0.09025885909795761, "global_step": 122823, "epoch": 1380, "lr": 5.759460585807193e-05}
{"train_loss": 0.12406259775161743, "global_step": 122824, "epoch": 1380, "lr": 5.759403284759639e-05}
{"train_loss": 0.10817144066095352, "global_step": 122825, "epoch": 1380, "lr": 5.759345983609994e-05}
{"train_loss": 0.10735134035348892, "global_step": 122826, "epoch": 1380, "lr": 5.7592886823582634e-05}
{"train_loss": 0.11475492268800735, "global_step": 122827, "epoch": 1380, "lr": 5.759231381004455e-05}
{"train_loss": 0.1899610310792923, "global_step": 122828, "epoch": 1380, "lr": 5.7591740795485784e-05}
{"train_loss": 0.14155924320220947, "global_step": 122829, "epoch": 1380, "lr": 5.7591167779906394e-05}
{"train_loss": 0.15590894222259521, "global_step": 122830, "epoch": 1380, "lr": 5.759059476330646e-05}
{"train_loss": 0.14763468503952026, "global_step": 122831, "epoch": 1380, "lr": 5.7590021745686065e-05}
{"train_loss": 0.06595184653997421, "global_step": 122832, "epoch": 1380, "lr": 5.7589448727045295e-05}
{"train_loss": 0.08602705597877502, "global_step": 122833, "epoch": 1380, "lr": 5.75888757073842e-05}
{"train_loss": 0.139539897441864, "global_step": 122834, "epoch": 1380, "lr": 5.758830268670289e-05}
{"train_loss": 0.11679202318191528, "global_step": 122835, "epoch": 1380, "lr": 5.758772966500141e-05}
{"train_loss": 0.11176767945289612, "global_step": 122836, "epoch": 1380, "lr": 5.758715664227986e-05}
{"train_loss": 0.1551634669303894, "global_step": 122837, "epoch": 1380, "lr": 5.7586583618538304e-05}
{"train_loss": 0.12108615785837173, "global_step": 122838, "epoch": 1380, "lr": 5.758601059377684e-05}
{"train_loss": 0.14143037796020508, "global_step": 122839, "epoch": 1380, "lr": 5.7585437567995514e-05}
{"train_loss": 0.11062704771757126, "global_step": 122840, "epoch": 1380, "lr": 5.758486454119443e-05}
{"train_loss": 0.11451201140880585, "global_step": 122841, "epoch": 1380, "lr": 5.758429151337363e-05}
{"train_loss": 0.10717090219259262, "global_step": 122842, "epoch": 1380, "lr": 5.7583718484533256e-05}
{"train_loss": 0.130787193775177, "global_step": 122843, "epoch": 1380, "lr": 5.758314545467332e-05}
{"train_loss": 0.1432763636112213, "global_step": 122844, "epoch": 1380, "lr": 5.758257242379391e-05}
{"train_loss": 0.15137100219726562, "global_step": 122845, "epoch": 1380, "lr": 5.7581999391895124e-05}
{"train_loss": 0.11036687344312668, "global_step": 122846, "epoch": 1380, "lr": 5.7581426358977044e-05}
{"train_loss": 0.11481201648712158, "global_step": 122847, "epoch": 1380, "lr": 5.758085332503972e-05}
{"train_loss": 0.14466793835163116, "global_step": 122848, "epoch": 1380, "lr": 5.758028029008323e-05}
{"train_loss": 0.1895478516817093, "global_step": 122849, "epoch": 1380, "lr": 5.757970725410768e-05}
{"train_loss": 0.08729545772075653, "global_step": 122850, "epoch": 1380, "lr": 5.757913421711313e-05}
{"train_loss": 0.07440107315778732, "global_step": 122851, "epoch": 1380, "lr": 5.757856117909965e-05}
{"train_loss": 0.12067665904760361, "global_step": 122852, "epoch": 1380, "lr": 5.757798814006734e-05}
{"train_loss": 0.1381755918264389, "global_step": 122853, "epoch": 1380, "lr": 5.757741510001625e-05}
{"train_loss": 0.14573100209236145, "global_step": 122854, "epoch": 1380, "lr": 5.7576842058946475e-05}
{"train_loss": 0.16507631540298462, "global_step": 122855, "epoch": 1380, "lr": 5.7576269016858085e-05}
{"train_loss": 0.16291339695453644, "global_step": 122856, "epoch": 1380, "lr": 5.7575695973751154e-05}
{"train_loss": 0.14161084592342377, "global_step": 122857, "epoch": 1380, "lr": 5.757512292962577e-05}
{"train_loss": 0.14081786572933197, "global_step": 122858, "epoch": 1380, "lr": 5.7574549884482e-05}
{"train_loss": 0.16931602358818054, "global_step": 122859, "epoch": 1380, "lr": 5.757397683831992e-05}
{"train_loss": 0.142648845911026, "global_step": 122860, "epoch": 1380, "lr": 5.757340379113961e-05}
{"train_loss": 0.14215490221977234, "global_step": 122861, "epoch": 1380, "lr": 5.7572830742941164e-05}
{"train_loss": 0.17652103304862976, "global_step": 122862, "epoch": 1380, "lr": 5.757225769372463e-05}
{"train_loss": 0.12468907237052917, "global_step": 122863, "epoch": 1380, "lr": 5.757168464349011e-05}
{"train_loss": 0.20282700657844543, "global_step": 122864, "epoch": 1380, "lr": 5.7571111592237656e-05}
{"train_loss": 0.1153668686747551, "global_step": 122865, "epoch": 1380, "lr": 5.7570538539967354e-05}
{"train_loss": 0.08371875435113907, "global_step": 122866, "epoch": 1380, "lr": 5.756996548667929e-05}
{"train_loss": 0.14873293042182922, "global_step": 122867, "epoch": 1380, "lr": 5.7569392432373534e-05}
{"train_loss": 0.1450570523738861, "global_step": 122868, "epoch": 1380, "lr": 5.756881937705017e-05}
{"train_loss": 0.11547669023275375, "global_step": 122869, "epoch": 1380, "lr": 5.7568246320709274e-05}
{"train_loss": 0.11088035255670547, "global_step": 122870, "epoch": 1380, "lr": 5.756767326335092e-05}
{"train_loss": 0.10350941121578217, "global_step": 122871, "epoch": 1380, "lr": 5.7567100204975176e-05}
{"train_loss": 0.1389002501964569, "global_step": 122872, "epoch": 1380, "lr": 5.756652714558214e-05}
{"train_loss": 0.18533647060394287, "global_step": 122873, "epoch": 1380, "lr": 5.756595408517187e-05}
{"train_loss": 0.14680911600589752, "global_step": 122874, "epoch": 1380, "lr": 5.7565381023744445e-05}
{"train_loss": 0.1883874237537384, "global_step": 122875, "epoch": 1380, "lr": 5.756480796129995e-05}
{"train_loss": 0.15996165573596954, "global_step": 122876, "epoch": 1380, "lr": 5.756423489783845e-05}
{"train_loss": 0.09452423453330994, "global_step": 122877, "epoch": 1380, "lr": 5.756366183336005e-05}
{"train_loss": 0.10500363260507584, "global_step": 122878, "epoch": 1380, "lr": 5.75630887678648e-05}
{"train_loss": 0.1908828765153885, "global_step": 122879, "epoch": 1380, "lr": 5.7562515701352785e-05}
{"train_loss": 0.20872245728969574, "global_step": 122880, "epoch": 1380, "lr": 5.7561942633824085e-05}
{"train_loss": 0.13221067190170288, "global_step": 122881, "epoch": 1380, "lr": 5.756136956527876e-05}
{"train_loss": 0.16914843022823334, "global_step": 122882, "epoch": 1380, "lr": 5.756079649571692e-05}
{"train_loss": 0.09128479659557343, "global_step": 122883, "epoch": 1380, "lr": 5.756022342513862e-05}
{"train_loss": 0.09746769070625305, "global_step": 122884, "epoch": 1380, "lr": 5.7559650353543916e-05}
{"train_loss": 0.10424331575632095, "global_step": 122885, "epoch": 1380, "lr": 5.755907728093293e-05}
{"train_loss": 0.08174688369035721, "global_step": 122886, "epoch": 1380, "lr": 5.755850420730571e-05}
{"train_loss": 0.09837200492620468, "global_step": 122887, "epoch": 1380, "lr": 5.7557931132662346e-05}
{"train_loss": 0.11818821728229523, "global_step": 122888, "epoch": 1380, "lr": 5.755735805700291e-05}
{"train_loss": 0.15373344719409943, "global_step": 122889, "epoch": 1380, "lr": 5.755678498032748e-05}
{"train_loss": 0.11275129020214081, "global_step": 122890, "epoch": 1380, "lr": 5.755621190263614e-05}
{"train_loss": 0.15811578929424286, "global_step": 122891, "epoch": 1380, "lr": 5.755563882392895e-05}
{"train_loss": 0.2631470561027527, "global_step": 122892, "epoch": 1380, "lr": 5.755506574420599e-05}
{"train_loss": 0.09284484386444092, "global_step": 122893, "epoch": 1380, "lr": 5.755449266346735e-05}
{"train_loss": 0.16701197624206543, "global_step": 122894, "epoch": 1380, "lr": 5.75539195817131e-05}
{"train_loss": 0.09429874271154404, "global_step": 122895, "epoch": 1380, "lr": 5.755334649894333e-05}
{"train_loss": 0.11316568404436111, "global_step": 122896, "epoch": 1380, "lr": 5.755277341515809e-05}
{"train_loss": 0.12506403028964996, "global_step": 122897, "epoch": 1380, "lr": 5.7552200330357474e-05}
{"train_loss": 0.13162845373153687, "global_step": 122898, "epoch": 1380, "lr": 5.755162724454156e-05}
{"train_loss": 0.18984195590019226, "global_step": 122899, "epoch": 1380, "lr": 5.755105415771041e-05}
{"train_loss": 0.13746711611747742, "global_step": 122900, "epoch": 1380, "lr": 5.755048106986414e-05}
{"train_loss": 0.1108902096748352, "global_step": 122901, "epoch": 1380, "lr": 5.7549907981002784e-05}
{"train_loss": 0.07289975881576538, "global_step": 122902, "epoch": 1380, "lr": 5.754933489112643e-05}
{"train_loss": 0.14875759184360504, "global_step": 122903, "epoch": 1380, "lr": 5.754876180023516e-05}
{"train_loss": 0.10755939781665802, "global_step": 122904, "epoch": 1380, "lr": 5.754818870832905e-05}
{"train_loss": 0.12217367440462112, "global_step": 122905, "epoch": 1380, "lr": 5.754761561540818e-05}
{"train_loss": 0.13552510738372803, "global_step": 122906, "epoch": 1380, "lr": 5.754704252147263e-05}
{"train_loss": 0.10272635519504547, "global_step": 122907, "epoch": 1380, "lr": 5.754646942652246e-05}
{"train_loss": 0.13078569055775577, "global_step": 122908, "epoch": 1380, "lr": 5.754589633055778e-05, "val_loss": 5.3861517906188965, "train_action_mse_error": 28.725238800048828}
{"train_loss": 0.10479090362787247, "global_step": 122909, "epoch": 1381, "lr": 5.754532323357864e-05}
{"train_loss": 0.10407283902168274, "global_step": 122910, "epoch": 1381, "lr": 5.7544750135585114e-05}
{"train_loss": 0.1421283781528473, "global_step": 122911, "epoch": 1381, "lr": 5.7544177036577296e-05}
{"train_loss": 0.12222456187009811, "global_step": 122912, "epoch": 1381, "lr": 5.754360393655525e-05}
{"train_loss": 0.10038376599550247, "global_step": 122913, "epoch": 1381, "lr": 5.754303083551906e-05}
{"train_loss": 0.12650910019874573, "global_step": 122914, "epoch": 1381, "lr": 5.75424577334688e-05}
{"train_loss": 0.06723382323980331, "global_step": 122915, "epoch": 1381, "lr": 5.754188463040455e-05}
{"train_loss": 0.14501869678497314, "global_step": 122916, "epoch": 1381, "lr": 5.754131152632638e-05}
{"train_loss": 0.09195589274168015, "global_step": 122917, "epoch": 1381, "lr": 5.7540738421234384e-05}
{"train_loss": 0.1492534726858139, "global_step": 122918, "epoch": 1381, "lr": 5.7540165315128626e-05}
{"train_loss": 0.16204442083835602, "global_step": 122919, "epoch": 1381, "lr": 5.7539592208009186e-05}
{"train_loss": 0.07843592017889023, "global_step": 122920, "epoch": 1381, "lr": 5.7539019099876126e-05}
{"train_loss": 0.1820172369480133, "global_step": 122921, "epoch": 1381, "lr": 5.7538445990729553e-05}
{"train_loss": 0.17816463112831116, "global_step": 122922, "epoch": 1381, "lr": 5.753787288056951e-05}
{"train_loss": 0.17094023525714874, "global_step": 122923, "epoch": 1381, "lr": 5.75372997693961e-05}
{"train_loss": 0.1713915765285492, "global_step": 122924, "epoch": 1381, "lr": 5.753672665720939e-05}
{"train_loss": 0.09460711479187012, "global_step": 122925, "epoch": 1381, "lr": 5.7536153544009476e-05}
{"train_loss": 0.1633862406015396, "global_step": 122926, "epoch": 1381, "lr": 5.75355804297964e-05}
{"train_loss": 0.12272339314222336, "global_step": 122927, "epoch": 1381, "lr": 5.7535007314570266e-05}
{"train_loss": 0.07327549904584885, "global_step": 122928, "epoch": 1381, "lr": 5.7534434198331134e-05}
{"train_loss": 0.1189759224653244, "global_step": 122929, "epoch": 1381, "lr": 5.75338610810791e-05}
{"train_loss": 0.09967844188213348, "global_step": 122930, "epoch": 1381, "lr": 5.753328796281422e-05}
{"train_loss": 0.09848345816135406, "global_step": 122931, "epoch": 1381, "lr": 5.753271484353658e-05}
{"train_loss": 0.09172327816486359, "global_step": 122932, "epoch": 1381, "lr": 5.7532141723246254e-05}
{"train_loss": 0.1042761504650116, "global_step": 122933, "epoch": 1381, "lr": 5.753156860194334e-05}
{"train_loss": 0.0968414843082428, "global_step": 122934, "epoch": 1381, "lr": 5.753099547962789e-05}
{"train_loss": 0.08438573777675629, "global_step": 122935, "epoch": 1381, "lr": 5.75304223563e-05}
{"train_loss": 0.10139148682355881, "global_step": 122936, "epoch": 1381, "lr": 5.752984923195972e-05}
{"train_loss": 0.19517137110233307, "global_step": 122937, "epoch": 1381, "lr": 5.7529276106607146e-05}
{"train_loss": 0.09561044722795486, "global_step": 122938, "epoch": 1381, "lr": 5.752870298024238e-05}
{"train_loss": 0.08788719773292542, "global_step": 122939, "epoch": 1381, "lr": 5.752812985286543e-05}
{"train_loss": 0.1208319291472435, "global_step": 122940, "epoch": 1381, "lr": 5.752755672447643e-05}
{"train_loss": 0.16648253798484802, "global_step": 122941, "epoch": 1381, "lr": 5.752698359507545e-05}
{"train_loss": 0.084353007376194, "global_step": 122942, "epoch": 1381, "lr": 5.752641046466255e-05}
{"train_loss": 0.16162918508052826, "global_step": 122943, "epoch": 1381, "lr": 5.752583733323782e-05}
{"train_loss": 0.12657757103443146, "global_step": 122944, "epoch": 1381, "lr": 5.7525264200801335e-05}
{"train_loss": 0.13177061080932617, "global_step": 122945, "epoch": 1381, "lr": 5.7524691067353163e-05}
{"train_loss": 0.17620369791984558, "global_step": 122946, "epoch": 1381, "lr": 5.75241179328934e-05}
{"train_loss": 0.13453367352485657, "global_step": 122947, "epoch": 1381, "lr": 5.7523544797422105e-05}
{"train_loss": 0.11182497441768646, "global_step": 122948, "epoch": 1381, "lr": 5.752297166093935e-05}
{"train_loss": 0.12083294987678528, "global_step": 122949, "epoch": 1381, "lr": 5.752239852344523e-05}
{"train_loss": 0.1292046159505844, "global_step": 122950, "epoch": 1381, "lr": 5.752182538493982e-05}
{"train_loss": 0.07381221652030945, "global_step": 122951, "epoch": 1381, "lr": 5.75212522454232e-05}
{"train_loss": 0.19549153745174408, "global_step": 122952, "epoch": 1381, "lr": 5.7520679104895424e-05}
{"train_loss": 0.13444910943508148, "global_step": 122953, "epoch": 1381, "lr": 5.7520105963356584e-05}
{"train_loss": 0.16900478303432465, "global_step": 122954, "epoch": 1381, "lr": 5.7519532820806765e-05}
{"train_loss": 0.12192992866039276, "global_step": 122955, "epoch": 1381, "lr": 5.751895967724603e-05}
{"train_loss": 0.1348317563533783, "global_step": 122956, "epoch": 1381, "lr": 5.751838653267447e-05}
{"train_loss": 0.12180525064468384, "global_step": 122957, "epoch": 1381, "lr": 5.7517813387092155e-05}
{"train_loss": 0.10921300202608109, "global_step": 122958, "epoch": 1381, "lr": 5.751724024049916e-05}
{"train_loss": 0.14935162663459778, "global_step": 122959, "epoch": 1381, "lr": 5.751666709289556e-05}
{"train_loss": 0.13922253251075745, "global_step": 122960, "epoch": 1381, "lr": 5.7516093944281434e-05}
{"train_loss": 0.09532543271780014, "global_step": 122961, "epoch": 1381, "lr": 5.751552079465687e-05}
{"train_loss": 0.12186530977487564, "global_step": 122962, "epoch": 1381, "lr": 5.7514947644021934e-05}
{"train_loss": 0.07878303527832031, "global_step": 122963, "epoch": 1381, "lr": 5.7514374492376696e-05}
{"train_loss": 0.12430819869041443, "global_step": 122964, "epoch": 1381, "lr": 5.751380133972125e-05}
{"train_loss": 0.11795860528945923, "global_step": 122965, "epoch": 1381, "lr": 5.751322818605566e-05}
{"train_loss": 0.08674489706754684, "global_step": 122966, "epoch": 1381, "lr": 5.7512655031380015e-05}
{"train_loss": 0.16019733250141144, "global_step": 122967, "epoch": 1381, "lr": 5.7512081875694376e-05}
{"train_loss": 0.11536677181720734, "global_step": 122968, "epoch": 1381, "lr": 5.751150871899883e-05}
{"train_loss": 0.13547584414482117, "global_step": 122969, "epoch": 1381, "lr": 5.751093556129346e-05}
{"train_loss": 0.1803223043680191, "global_step": 122970, "epoch": 1381, "lr": 5.7510362402578334e-05}
{"train_loss": 0.07760949432849884, "global_step": 122971, "epoch": 1381, "lr": 5.750978924285353e-05}
{"train_loss": 0.11583424359560013, "global_step": 122972, "epoch": 1381, "lr": 5.7509216082119125e-05}
{"train_loss": 0.13049717247486115, "global_step": 122973, "epoch": 1381, "lr": 5.75086429203752e-05}
{"train_loss": 0.11785238236188889, "global_step": 122974, "epoch": 1381, "lr": 5.750806975762183e-05}
{"train_loss": 0.12048713862895966, "global_step": 122975, "epoch": 1381, "lr": 5.750749659385909e-05}
{"train_loss": 0.16273042559623718, "global_step": 122976, "epoch": 1381, "lr": 5.7506923429087056e-05}
{"train_loss": 0.1567232757806778, "global_step": 122977, "epoch": 1381, "lr": 5.750635026330581e-05}
{"train_loss": 0.09051574021577835, "global_step": 122978, "epoch": 1381, "lr": 5.7505777096515425e-05}
{"train_loss": 0.0929642990231514, "global_step": 122979, "epoch": 1381, "lr": 5.750520392871598e-05}
{"train_loss": 0.10599582642316818, "global_step": 122980, "epoch": 1381, "lr": 5.750463075990755e-05}
{"train_loss": 0.23702768981456757, "global_step": 122981, "epoch": 1381, "lr": 5.750405759009022e-05}
{"train_loss": 0.17073649168014526, "global_step": 122982, "epoch": 1381, "lr": 5.750348441926406e-05}
{"train_loss": 0.1420179158449173, "global_step": 122983, "epoch": 1381, "lr": 5.750291124742915e-05}
{"train_loss": 0.16589786112308502, "global_step": 122984, "epoch": 1381, "lr": 5.750233807458556e-05}
{"train_loss": 0.15169450640678406, "global_step": 122985, "epoch": 1381, "lr": 5.7501764900733366e-05}
{"train_loss": 0.10567592829465866, "global_step": 122986, "epoch": 1381, "lr": 5.750119172587266e-05}
{"train_loss": 0.1702519804239273, "global_step": 122987, "epoch": 1381, "lr": 5.7500618550003516e-05}
{"train_loss": 0.14014489948749542, "global_step": 122988, "epoch": 1381, "lr": 5.7500045373126e-05}
{"train_loss": 0.0966549962759018, "global_step": 122989, "epoch": 1381, "lr": 5.749947219524019e-05}
{"train_loss": 0.16623808443546295, "global_step": 122990, "epoch": 1381, "lr": 5.749889901634617e-05}
{"train_loss": 0.18832142651081085, "global_step": 122991, "epoch": 1381, "lr": 5.749832583644401e-05}
{"train_loss": 0.13199402391910553, "global_step": 122992, "epoch": 1381, "lr": 5.74977526555338e-05}
{"train_loss": 0.1360624134540558, "global_step": 122993, "epoch": 1381, "lr": 5.7497179473615604e-05}
{"train_loss": 0.08797232806682587, "global_step": 122994, "epoch": 1381, "lr": 5.74966062906895e-05}
{"train_loss": 0.12992453575134277, "global_step": 122995, "epoch": 1381, "lr": 5.749603310675558e-05}
{"train_loss": 0.10564586520195007, "global_step": 122996, "epoch": 1381, "lr": 5.74954599218139e-05}
{"train_loss": 0.12805031742272752, "global_step": 122997, "epoch": 1381, "lr": 5.749488673586454e-05, "val_loss": 5.587945461273193}
{"train_loss": 0.16502192616462708, "global_step": 122998, "epoch": 1382, "lr": 5.74943135489076e-05}
{"train_loss": 0.20186105370521545, "global_step": 122999, "epoch": 1382, "lr": 5.7493740360943125e-05}
{"train_loss": 0.10239620506763458, "global_step": 123000, "epoch": 1382, "lr": 5.749316717197122e-05}
{"train_loss": 0.1861555129289627, "global_step": 123001, "epoch": 1382, "lr": 5.749259398199195e-05}
{"train_loss": 0.14990480244159698, "global_step": 123002, "epoch": 1382, "lr": 5.749202079100538e-05}
{"train_loss": 0.11254668235778809, "global_step": 123003, "epoch": 1382, "lr": 5.7491447599011616e-05}
{"train_loss": 0.16407649219036102, "global_step": 123004, "epoch": 1382, "lr": 5.7490874406010706e-05}
{"train_loss": 0.14476504921913147, "global_step": 123005, "epoch": 1382, "lr": 5.749030121200274e-05}
{"train_loss": 0.12598033249378204, "global_step": 123006, "epoch": 1382, "lr": 5.74897280169878e-05}
{"train_loss": 0.1944969892501831, "global_step": 123007, "epoch": 1382, "lr": 5.748915482096596e-05}
{"train_loss": 0.10525926947593689, "global_step": 123008, "epoch": 1382, "lr": 5.748858162393729e-05}
{"train_loss": 0.08669400960206985, "global_step": 123009, "epoch": 1382, "lr": 5.748800842590187e-05}
{"train_loss": 0.1621326357126236, "global_step": 123010, "epoch": 1382, "lr": 5.7487435226859774e-05}
{"train_loss": 0.19475360214710236, "global_step": 123011, "epoch": 1382, "lr": 5.7486862026811096e-05}
{"train_loss": 0.14557455480098724, "global_step": 123012, "epoch": 1382, "lr": 5.7486288825755886e-05}
{"train_loss": 0.1275295913219452, "global_step": 123013, "epoch": 1382, "lr": 5.748571562369425e-05}
{"train_loss": 0.14522705972194672, "global_step": 123014, "epoch": 1382, "lr": 5.748514242062624e-05}
{"train_loss": 0.1067783311009407, "global_step": 123015, "epoch": 1382, "lr": 5.7484569216551946e-05}
{"train_loss": 0.11746633797883987, "global_step": 123016, "epoch": 1382, "lr": 5.7483996011471445e-05}
{"train_loss": 0.13508836925029755, "global_step": 123017, "epoch": 1382, "lr": 5.748342280538481e-05}
{"train_loss": 0.17686673998832703, "global_step": 123018, "epoch": 1382, "lr": 5.748284959829212e-05}
{"train_loss": 0.1654486358165741, "global_step": 123019, "epoch": 1382, "lr": 5.748227639019346e-05}
{"train_loss": 0.12305337935686111, "global_step": 123020, "epoch": 1382, "lr": 5.7481703181088885e-05}
{"train_loss": 0.14535433053970337, "global_step": 123021, "epoch": 1382, "lr": 5.74811299709785e-05}
{"train_loss": 0.22401732206344604, "global_step": 123022, "epoch": 1382, "lr": 5.748055675986236e-05}
{"train_loss": 0.1224849671125412, "global_step": 123023, "epoch": 1382, "lr": 5.7479983547740556e-05}
{"train_loss": 0.19669511914253235, "global_step": 123024, "epoch": 1382, "lr": 5.747941033461315e-05}
{"train_loss": 0.16298766434192657, "global_step": 123025, "epoch": 1382, "lr": 5.7478837120480236e-05}
{"train_loss": 0.13156253099441528, "global_step": 123026, "epoch": 1382, "lr": 5.747826390534189e-05}
{"train_loss": 0.11145014315843582, "global_step": 123027, "epoch": 1382, "lr": 5.747769068919817e-05}
{"train_loss": 0.10302683711051941, "global_step": 123028, "epoch": 1382, "lr": 5.747711747204917e-05}
{"train_loss": 0.13240130245685577, "global_step": 123029, "epoch": 1382, "lr": 5.7476544253894946e-05}
{"train_loss": 0.20575548708438873, "global_step": 123030, "epoch": 1382, "lr": 5.747597103473562e-05}
{"train_loss": 0.12867984175682068, "global_step": 123031, "epoch": 1382, "lr": 5.747539781457123e-05}
{"train_loss": 0.09362917393445969, "global_step": 123032, "epoch": 1382, "lr": 5.747482459340187e-05}
{"train_loss": 0.12190457433462143, "global_step": 123033, "epoch": 1382, "lr": 5.7474251371227604e-05}
{"train_loss": 0.07844128459692001, "global_step": 123034, "epoch": 1382, "lr": 5.7473678148048516e-05}
{"train_loss": 0.17450091242790222, "global_step": 123035, "epoch": 1382, "lr": 5.7473104923864685e-05}
{"train_loss": 0.07258353382349014, "global_step": 123036, "epoch": 1382, "lr": 5.747253169867618e-05}
{"train_loss": 0.1400556117296219, "global_step": 123037, "epoch": 1382, "lr": 5.747195847248309e-05}
{"train_loss": 0.1321018785238266, "global_step": 123038, "epoch": 1382, "lr": 5.747138524528548e-05}
{"train_loss": 0.14389269053936005, "global_step": 123039, "epoch": 1382, "lr": 5.747081201708344e-05}
{"train_loss": 0.15596367418766022, "global_step": 123040, "epoch": 1382, "lr": 5.747023878787704e-05}
{"train_loss": 0.09108022600412369, "global_step": 123041, "epoch": 1382, "lr": 5.7469665557666354e-05}
{"train_loss": 0.15885473787784576, "global_step": 123042, "epoch": 1382, "lr": 5.7469092326451465e-05}
{"train_loss": 0.12143369764089584, "global_step": 123043, "epoch": 1382, "lr": 5.7468519094232455e-05}
{"train_loss": 0.1467088758945465, "global_step": 123044, "epoch": 1382, "lr": 5.746794586100937e-05}
{"train_loss": 0.16002146899700165, "global_step": 123045, "epoch": 1382, "lr": 5.746737262678233e-05}
{"train_loss": 0.1230204775929451, "global_step": 123046, "epoch": 1382, "lr": 5.7466799391551396e-05}
{"train_loss": 0.13196627795696259, "global_step": 123047, "epoch": 1382, "lr": 5.746622615531663e-05}
{"train_loss": 0.13656698167324066, "global_step": 123048, "epoch": 1382, "lr": 5.746565291807812e-05}
{"train_loss": 0.1527773141860962, "global_step": 123049, "epoch": 1382, "lr": 5.7465079679835964e-05}
{"train_loss": 0.14508134126663208, "global_step": 123050, "epoch": 1382, "lr": 5.7464506440590206e-05}
{"train_loss": 0.10827469825744629, "global_step": 123051, "epoch": 1382, "lr": 5.7463933200340925e-05}
{"train_loss": 0.16559481620788574, "global_step": 123052, "epoch": 1382, "lr": 5.746335995908823e-05}
{"train_loss": 0.1909073293209076, "global_step": 123053, "epoch": 1382, "lr": 5.746278671683216e-05}
{"train_loss": 0.09991137683391571, "global_step": 123054, "epoch": 1382, "lr": 5.746221347357281e-05}
{"train_loss": 0.14789190888404846, "global_step": 123055, "epoch": 1382, "lr": 5.7461640229310265e-05}
{"train_loss": 0.12746809422969818, "global_step": 123056, "epoch": 1382, "lr": 5.746106698404459e-05}
{"train_loss": 0.1027299165725708, "global_step": 123057, "epoch": 1382, "lr": 5.7460493737775866e-05}
{"train_loss": 0.12818409502506256, "global_step": 123058, "epoch": 1382, "lr": 5.745992049050417e-05}
{"train_loss": 0.11853092908859253, "global_step": 123059, "epoch": 1382, "lr": 5.745934724222958e-05}
{"train_loss": 0.12354343384504318, "global_step": 123060, "epoch": 1382, "lr": 5.7458773992952166e-05}
{"train_loss": 0.09685420989990234, "global_step": 123061, "epoch": 1382, "lr": 5.745820074267202e-05}
{"train_loss": 0.12956316769123077, "global_step": 123062, "epoch": 1382, "lr": 5.74576274913892e-05}
{"train_loss": 0.12344587594270706, "global_step": 123063, "epoch": 1382, "lr": 5.7457054239103804e-05}
{"train_loss": 0.1327022910118103, "global_step": 123064, "epoch": 1382, "lr": 5.745648098581589e-05}
{"train_loss": 0.1781461536884308, "global_step": 123065, "epoch": 1382, "lr": 5.745590773152554e-05}
{"train_loss": 0.06864038854837418, "global_step": 123066, "epoch": 1382, "lr": 5.745533447623283e-05}
{"train_loss": 0.18945105373859406, "global_step": 123067, "epoch": 1382, "lr": 5.745476121993785e-05}
{"train_loss": 0.08908439427614212, "global_step": 123068, "epoch": 1382, "lr": 5.7454187962640673e-05}
{"train_loss": 0.15180081129074097, "global_step": 123069, "epoch": 1382, "lr": 5.7453614704341364e-05}
{"train_loss": 0.17880988121032715, "global_step": 123070, "epoch": 1382, "lr": 5.7453041445040014e-05}
{"train_loss": 0.09439460188150406, "global_step": 123071, "epoch": 1382, "lr": 5.7452468184736685e-05}
{"train_loss": 0.10078413784503937, "global_step": 123072, "epoch": 1382, "lr": 5.745189492343147e-05}
{"train_loss": 0.12810629606246948, "global_step": 123073, "epoch": 1382, "lr": 5.7451321661124436e-05}
{"train_loss": 0.16766442358493805, "global_step": 123074, "epoch": 1382, "lr": 5.745074839781566e-05}
{"train_loss": 0.12843166291713715, "global_step": 123075, "epoch": 1382, "lr": 5.7450175133505226e-05}
{"train_loss": 0.11154575645923615, "global_step": 123076, "epoch": 1382, "lr": 5.7449601868193195e-05}
{"train_loss": 0.1548689305782318, "global_step": 123077, "epoch": 1382, "lr": 5.744902860187966e-05}
{"train_loss": 0.11175099015235901, "global_step": 123078, "epoch": 1382, "lr": 5.744845533456471e-05}
{"train_loss": 0.11612341552972794, "global_step": 123079, "epoch": 1382, "lr": 5.744788206624839e-05}
{"train_loss": 0.11133895069360733, "global_step": 123080, "epoch": 1382, "lr": 5.7447308796930796e-05}
{"train_loss": 0.16041339933872223, "global_step": 123081, "epoch": 1382, "lr": 5.744673552661201e-05}
{"train_loss": 0.17851321399211884, "global_step": 123082, "epoch": 1382, "lr": 5.744616225529209e-05}
{"train_loss": 0.16637906432151794, "global_step": 123083, "epoch": 1382, "lr": 5.7445588982971135e-05}
{"train_loss": 0.13453292846679688, "global_step": 123084, "epoch": 1382, "lr": 5.744501570964919e-05}
{"train_loss": 0.13494457304477692, "global_step": 123085, "epoch": 1382, "lr": 5.744444243532636e-05}
{"train_loss": 0.137868809314926, "global_step": 123086, "epoch": 1382, "lr": 5.744386916000273e-05, "val_loss": 5.295257568359375}
{"train_loss": 0.1822098046541214, "global_step": 123087, "epoch": 1383, "lr": 5.744329588367836e-05}
{"train_loss": 0.15379762649536133, "global_step": 123088, "epoch": 1383, "lr": 5.7442722606353324e-05}
{"train_loss": 0.07854810357093811, "global_step": 123089, "epoch": 1383, "lr": 5.7442149328027716e-05}
{"train_loss": 0.10480894148349762, "global_step": 123090, "epoch": 1383, "lr": 5.744157604870159e-05}
{"train_loss": 0.18099676072597504, "global_step": 123091, "epoch": 1383, "lr": 5.744100276837504e-05}
{"train_loss": 0.06623964756727219, "global_step": 123092, "epoch": 1383, "lr": 5.744042948704813e-05}
{"train_loss": 0.10534175485372543, "global_step": 123093, "epoch": 1383, "lr": 5.743985620472096e-05}
{"train_loss": 0.13340698182582855, "global_step": 123094, "epoch": 1383, "lr": 5.743928292139358e-05}
{"train_loss": 0.1422097533941269, "global_step": 123095, "epoch": 1383, "lr": 5.743870963706608e-05}
{"train_loss": 0.1327666938304901, "global_step": 123096, "epoch": 1383, "lr": 5.7438136351738536e-05}
{"train_loss": 0.14973662793636322, "global_step": 123097, "epoch": 1383, "lr": 5.743756306541103e-05}
{"train_loss": 0.15357060730457306, "global_step": 123098, "epoch": 1383, "lr": 5.743698977808363e-05}
{"train_loss": 0.10174731910228729, "global_step": 123099, "epoch": 1383, "lr": 5.743641648975642e-05}
{"train_loss": 0.1094910055398941, "global_step": 123100, "epoch": 1383, "lr": 5.743584320042947e-05}
{"train_loss": 0.15646040439605713, "global_step": 123101, "epoch": 1383, "lr": 5.743526991010286e-05}
{"train_loss": 0.14483201503753662, "global_step": 123102, "epoch": 1383, "lr": 5.743469661877666e-05}
{"train_loss": 0.09107711166143417, "global_step": 123103, "epoch": 1383, "lr": 5.743412332645097e-05}
{"train_loss": 0.15219374001026154, "global_step": 123104, "epoch": 1383, "lr": 5.743355003312585e-05}
{"train_loss": 0.1152815967798233, "global_step": 123105, "epoch": 1383, "lr": 5.743297673880138e-05}
{"train_loss": 0.08823511749505997, "global_step": 123106, "epoch": 1383, "lr": 5.743240344347764e-05}
{"train_loss": 0.14366333186626434, "global_step": 123107, "epoch": 1383, "lr": 5.743183014715472e-05}
{"train_loss": 0.1888352632522583, "global_step": 123108, "epoch": 1383, "lr": 5.743125684983266e-05}
{"train_loss": 0.19004125893115997, "global_step": 123109, "epoch": 1383, "lr": 5.7430683551511554e-05}
{"train_loss": 0.15156829357147217, "global_step": 123110, "epoch": 1383, "lr": 5.7430110252191495e-05}
{"train_loss": 0.1218247264623642, "global_step": 123111, "epoch": 1383, "lr": 5.742953695187254e-05}
{"train_loss": 0.1303659975528717, "global_step": 123112, "epoch": 1383, "lr": 5.742896365055478e-05}
{"train_loss": 0.08239396661520004, "global_step": 123113, "epoch": 1383, "lr": 5.7428390348238284e-05}
{"train_loss": 0.13685844838619232, "global_step": 123114, "epoch": 1383, "lr": 5.742781704492314e-05}
{"train_loss": 0.12722301483154297, "global_step": 123115, "epoch": 1383, "lr": 5.7427243740609406e-05}
{"train_loss": 0.11288554966449738, "global_step": 123116, "epoch": 1383, "lr": 5.742667043529717e-05}
{"train_loss": 0.10852619260549545, "global_step": 123117, "epoch": 1383, "lr": 5.742609712898651e-05}
{"train_loss": 0.10420089960098267, "global_step": 123118, "epoch": 1383, "lr": 5.7425523821677494e-05}
{"train_loss": 0.13348272442817688, "global_step": 123119, "epoch": 1383, "lr": 5.7424950513370227e-05}
{"train_loss": 0.2289310097694397, "global_step": 123120, "epoch": 1383, "lr": 5.742437720406474e-05}
{"train_loss": 0.21278947591781616, "global_step": 123121, "epoch": 1383, "lr": 5.742380389376114e-05}
{"train_loss": 0.1555878221988678, "global_step": 123122, "epoch": 1383, "lr": 5.742323058245952e-05}
{"train_loss": 0.15505655109882355, "global_step": 123123, "epoch": 1383, "lr": 5.7422657270159916e-05}
{"train_loss": 0.12239497154951096, "global_step": 123124, "epoch": 1383, "lr": 5.742208395686244e-05}
{"train_loss": 0.16739897429943085, "global_step": 123125, "epoch": 1383, "lr": 5.7421510642567156e-05}
{"train_loss": 0.09032129496335983, "global_step": 123126, "epoch": 1383, "lr": 5.742093732727415e-05}
{"train_loss": 0.1498126983642578, "global_step": 123127, "epoch": 1383, "lr": 5.7420364010983474e-05}
{"train_loss": 0.1448027789592743, "global_step": 123128, "epoch": 1383, "lr": 5.741979069369522e-05}
{"train_loss": 0.1313631385564804, "global_step": 123129, "epoch": 1383, "lr": 5.7419217375409475e-05}
{"train_loss": 0.12122642993927002, "global_step": 123130, "epoch": 1383, "lr": 5.7418644056126293e-05}
{"train_loss": 0.15144512057304382, "global_step": 123131, "epoch": 1383, "lr": 5.7418070735845775e-05}
{"train_loss": 0.10973857343196869, "global_step": 123132, "epoch": 1383, "lr": 5.7417497414567986e-05}
{"train_loss": 0.1013045608997345, "global_step": 123133, "epoch": 1383, "lr": 5.7416924092293e-05}
{"train_loss": 0.1533481776714325, "global_step": 123134, "epoch": 1383, "lr": 5.74163507690209e-05}
{"train_loss": 0.08742351830005646, "global_step": 123135, "epoch": 1383, "lr": 5.741577744475177e-05}
{"train_loss": 0.07746799290180206, "global_step": 123136, "epoch": 1383, "lr": 5.741520411948567e-05}
{"train_loss": 0.14359746873378754, "global_step": 123137, "epoch": 1383, "lr": 5.741463079322269e-05}
{"train_loss": 0.1877744197845459, "global_step": 123138, "epoch": 1383, "lr": 5.741405746596289e-05}
{"train_loss": 0.14699825644493103, "global_step": 123139, "epoch": 1383, "lr": 5.741348413770637e-05}
{"train_loss": 0.11652795970439911, "global_step": 123140, "epoch": 1383, "lr": 5.74129108084532e-05}
{"train_loss": 0.18808187544345856, "global_step": 123141, "epoch": 1383, "lr": 5.741233747820346e-05}
{"train_loss": 0.11328352242708206, "global_step": 123142, "epoch": 1383, "lr": 5.7411764146957214e-05}
{"train_loss": 0.13211502134799957, "global_step": 123143, "epoch": 1383, "lr": 5.741119081471456e-05}
{"train_loss": 0.12832137942314148, "global_step": 123144, "epoch": 1383, "lr": 5.7410617481475545e-05}
{"train_loss": 0.12861979007720947, "global_step": 123145, "epoch": 1383, "lr": 5.7410044147240274e-05}
{"train_loss": 0.20724979043006897, "global_step": 123146, "epoch": 1383, "lr": 5.740947081200882e-05}
{"train_loss": 0.16436287760734558, "global_step": 123147, "epoch": 1383, "lr": 5.740889747578123e-05}
{"train_loss": 0.11519601941108704, "global_step": 123148, "epoch": 1383, "lr": 5.740832413855761e-05}
{"train_loss": 0.11729222536087036, "global_step": 123149, "epoch": 1383, "lr": 5.7407750800338046e-05}
{"train_loss": 0.1529499590396881, "global_step": 123150, "epoch": 1383, "lr": 5.740717746112259e-05}
{"train_loss": 0.149332657456398, "global_step": 123151, "epoch": 1383, "lr": 5.740660412091132e-05}
{"train_loss": 0.1572081446647644, "global_step": 123152, "epoch": 1383, "lr": 5.740603077970433e-05}
{"train_loss": 0.13708727061748505, "global_step": 123153, "epoch": 1383, "lr": 5.74054574375017e-05}
{"train_loss": 0.13891705870628357, "global_step": 123154, "epoch": 1383, "lr": 5.740488409430349e-05}
{"train_loss": 0.13916891813278198, "global_step": 123155, "epoch": 1383, "lr": 5.740431075010978e-05}
{"train_loss": 0.16977658867835999, "global_step": 123156, "epoch": 1383, "lr": 5.7403737404920646e-05}
{"train_loss": 0.09935811161994934, "global_step": 123157, "epoch": 1383, "lr": 5.740316405873617e-05}
{"train_loss": 0.07553353160619736, "global_step": 123158, "epoch": 1383, "lr": 5.7402590711556435e-05}
{"train_loss": 0.08869269490242004, "global_step": 123159, "epoch": 1383, "lr": 5.740201736338151e-05}
{"train_loss": 0.18975645303726196, "global_step": 123160, "epoch": 1383, "lr": 5.740144401421148e-05}
{"train_loss": 0.13589659333229065, "global_step": 123161, "epoch": 1383, "lr": 5.74008706640464e-05}
{"train_loss": 0.1475859135389328, "global_step": 123162, "epoch": 1383, "lr": 5.7400297312886386e-05}
{"train_loss": 0.12662304937839508, "global_step": 123163, "epoch": 1383, "lr": 5.7399723960731476e-05}
{"train_loss": 0.11433415114879608, "global_step": 123164, "epoch": 1383, "lr": 5.739915060758176e-05}
{"train_loss": 0.1594097912311554, "global_step": 123165, "epoch": 1383, "lr": 5.739857725343735e-05}
{"train_loss": 0.12756115198135376, "global_step": 123166, "epoch": 1383, "lr": 5.739800389829826e-05}
{"train_loss": 0.09301573783159256, "global_step": 123167, "epoch": 1383, "lr": 5.739743054216461e-05}
{"train_loss": 0.18399359285831451, "global_step": 123168, "epoch": 1383, "lr": 5.7396857185036456e-05}
{"train_loss": 0.1844892054796219, "global_step": 123169, "epoch": 1383, "lr": 5.73962838269139e-05}
{"train_loss": 0.13130448758602142, "global_step": 123170, "epoch": 1383, "lr": 5.7395710467796995e-05}
{"train_loss": 0.10516881942749023, "global_step": 123171, "epoch": 1383, "lr": 5.739513710768582e-05}
{"train_loss": 0.23328937590122223, "global_step": 123172, "epoch": 1383, "lr": 5.7394563746580464e-05}
{"train_loss": 0.16031549870967865, "global_step": 123173, "epoch": 1383, "lr": 5.7393990384481e-05}
{"train_loss": 0.12883011996746063, "global_step": 123174, "epoch": 1383, "lr": 5.73934170213875e-05}
{"train_loss": 0.13667944726649295, "global_step": 123175, "epoch": 1383, "lr": 5.739284365730004e-05, "val_loss": 5.419949054718018}
{"train_loss": 0.10082855075597763, "global_step": 123176, "epoch": 1384, "lr": 5.739227029221872e-05}
{"train_loss": 0.10648063570261002, "global_step": 123177, "epoch": 1384, "lr": 5.739169692614359e-05}
{"train_loss": 0.14745895564556122, "global_step": 123178, "epoch": 1384, "lr": 5.7391123559074735e-05}
{"train_loss": 0.097258061170578, "global_step": 123179, "epoch": 1384, "lr": 5.739055019101224e-05}
{"train_loss": 0.1111428290605545, "global_step": 123180, "epoch": 1384, "lr": 5.738997682195617e-05}
{"train_loss": 0.1403624266386032, "global_step": 123181, "epoch": 1384, "lr": 5.738940345190661e-05}
{"train_loss": 0.13825763761997223, "global_step": 123182, "epoch": 1384, "lr": 5.738883008086363e-05}
{"train_loss": 0.04774654656648636, "global_step": 123183, "epoch": 1384, "lr": 5.7388256708827325e-05}
{"train_loss": 0.10687937587499619, "global_step": 123184, "epoch": 1384, "lr": 5.738768333579775e-05}
{"train_loss": 0.1784535050392151, "global_step": 123185, "epoch": 1384, "lr": 5.7387109961775e-05}
{"train_loss": 0.1453816443681717, "global_step": 123186, "epoch": 1384, "lr": 5.738653658675913e-05}
{"train_loss": 0.15768620371818542, "global_step": 123187, "epoch": 1384, "lr": 5.7385963210750235e-05}
{"train_loss": 0.11974626779556274, "global_step": 123188, "epoch": 1384, "lr": 5.738538983374839e-05}
{"train_loss": 0.12951909005641937, "global_step": 123189, "epoch": 1384, "lr": 5.7384816455753664e-05}
{"train_loss": 0.14856696128845215, "global_step": 123190, "epoch": 1384, "lr": 5.7384243076766145e-05}
{"train_loss": 0.09572270512580872, "global_step": 123191, "epoch": 1384, "lr": 5.7383669696785904e-05}
{"train_loss": 0.17611174285411835, "global_step": 123192, "epoch": 1384, "lr": 5.7383096315813014e-05}
{"train_loss": 0.16773855686187744, "global_step": 123193, "epoch": 1384, "lr": 5.738252293384754e-05}
{"train_loss": 0.19155073165893555, "global_step": 123194, "epoch": 1384, "lr": 5.73819495508896e-05}
{"train_loss": 0.1400475800037384, "global_step": 123195, "epoch": 1384, "lr": 5.738137616693924e-05}
{"train_loss": 0.0907466933131218, "global_step": 123196, "epoch": 1384, "lr": 5.738080278199655e-05}
{"train_loss": 0.09809298813343048, "global_step": 123197, "epoch": 1384, "lr": 5.738022939606159e-05}
{"train_loss": 0.11610402911901474, "global_step": 123198, "epoch": 1384, "lr": 5.737965600913445e-05}
{"train_loss": 0.11838873475790024, "global_step": 123199, "epoch": 1384, "lr": 5.737908262121521e-05}
{"train_loss": 0.12035498768091202, "global_step": 123200, "epoch": 1384, "lr": 5.737850923230395e-05}
{"train_loss": 0.1199548989534378, "global_step": 123201, "epoch": 1384, "lr": 5.737793584240072e-05}
{"train_loss": 0.13121207058429718, "global_step": 123202, "epoch": 1384, "lr": 5.7377362451505634e-05}
{"train_loss": 0.11429399251937866, "global_step": 123203, "epoch": 1384, "lr": 5.737678905961874e-05}
{"train_loss": 0.10382018983364105, "global_step": 123204, "epoch": 1384, "lr": 5.7376215666740144e-05}
{"train_loss": 0.10337532311677933, "global_step": 123205, "epoch": 1384, "lr": 5.737564227286989e-05}
{"train_loss": 0.15594209730625153, "global_step": 123206, "epoch": 1384, "lr": 5.737506887800807e-05}
{"train_loss": 0.10456990450620651, "global_step": 123207, "epoch": 1384, "lr": 5.737449548215478e-05}
{"train_loss": 0.07049127668142319, "global_step": 123208, "epoch": 1384, "lr": 5.737392208531006e-05}
{"train_loss": 0.11075624078512192, "global_step": 123209, "epoch": 1384, "lr": 5.7373348687474e-05}
{"train_loss": 0.11419674009084702, "global_step": 123210, "epoch": 1384, "lr": 5.737277528864671e-05}
{"train_loss": 0.1670229136943817, "global_step": 123211, "epoch": 1384, "lr": 5.7372201888828205e-05}
{"train_loss": 0.16541273891925812, "global_step": 123212, "epoch": 1384, "lr": 5.737162848801862e-05}
{"train_loss": 0.08594688028097153, "global_step": 123213, "epoch": 1384, "lr": 5.737105508621801e-05}
{"train_loss": 0.10874100774526596, "global_step": 123214, "epoch": 1384, "lr": 5.737048168342645e-05}
{"train_loss": 0.1587192714214325, "global_step": 123215, "epoch": 1384, "lr": 5.7369908279644014e-05}
{"train_loss": 0.1015113815665245, "global_step": 123216, "epoch": 1384, "lr": 5.7369334874870786e-05}
{"train_loss": 0.11279334127902985, "global_step": 123217, "epoch": 1384, "lr": 5.7368761469106835e-05}
{"train_loss": 0.12946966290473938, "global_step": 123218, "epoch": 1384, "lr": 5.736818806235225e-05}
{"train_loss": 0.2248317301273346, "global_step": 123219, "epoch": 1384, "lr": 5.736761465460709e-05}
{"train_loss": 0.17836801707744598, "global_step": 123220, "epoch": 1384, "lr": 5.7367041245871466e-05}
{"train_loss": 0.11922551691532135, "global_step": 123221, "epoch": 1384, "lr": 5.736646783614542e-05}
{"train_loss": 0.1423441767692566, "global_step": 123222, "epoch": 1384, "lr": 5.736589442542904e-05}
{"train_loss": 0.22048652172088623, "global_step": 123223, "epoch": 1384, "lr": 5.736532101372243e-05}
{"train_loss": 0.18848955631256104, "global_step": 123224, "epoch": 1384, "lr": 5.7364747601025616e-05}
{"train_loss": 0.14103415608406067, "global_step": 123225, "epoch": 1384, "lr": 5.73641741873387e-05}
{"train_loss": 0.10653374344110489, "global_step": 123226, "epoch": 1384, "lr": 5.7363600772661775e-05}
{"train_loss": 0.1185525506734848, "global_step": 123227, "epoch": 1384, "lr": 5.7363027356994904e-05}
{"train_loss": 0.1416666954755783, "global_step": 123228, "epoch": 1384, "lr": 5.736245394033814e-05}
{"train_loss": 0.180044487118721, "global_step": 123229, "epoch": 1384, "lr": 5.73618805226916e-05}
{"train_loss": 0.10381793230772018, "global_step": 123230, "epoch": 1384, "lr": 5.736130710405534e-05}
{"train_loss": 0.09974044561386108, "global_step": 123231, "epoch": 1384, "lr": 5.736073368442945e-05}
{"train_loss": 0.20443393290042877, "global_step": 123232, "epoch": 1384, "lr": 5.7360160263814e-05}
{"train_loss": 0.16030572354793549, "global_step": 123233, "epoch": 1384, "lr": 5.7359586842209055e-05}
{"train_loss": 0.14701445400714874, "global_step": 123234, "epoch": 1384, "lr": 5.7359013419614714e-05}
{"train_loss": 0.22043590247631073, "global_step": 123235, "epoch": 1384, "lr": 5.7358439996031047e-05}
{"train_loss": 0.1703672558069229, "global_step": 123236, "epoch": 1384, "lr": 5.735786657145812e-05}
{"train_loss": 0.12038001418113708, "global_step": 123237, "epoch": 1384, "lr": 5.735729314589602e-05}
{"train_loss": 0.1227966845035553, "global_step": 123238, "epoch": 1384, "lr": 5.735671971934481e-05}
{"train_loss": 0.07365099340677261, "global_step": 123239, "epoch": 1384, "lr": 5.735614629180459e-05}
{"train_loss": 0.13077262043952942, "global_step": 123240, "epoch": 1384, "lr": 5.735557286327542e-05}
{"train_loss": 0.09447889775037766, "global_step": 123241, "epoch": 1384, "lr": 5.735499943375739e-05}
{"train_loss": 0.09259238094091415, "global_step": 123242, "epoch": 1384, "lr": 5.735442600325057e-05}
{"train_loss": 0.1279619187116623, "global_step": 123243, "epoch": 1384, "lr": 5.735385257175504e-05}
{"train_loss": 0.1009189561009407, "global_step": 123244, "epoch": 1384, "lr": 5.7353279139270866e-05}
{"train_loss": 0.1863117665052414, "global_step": 123245, "epoch": 1384, "lr": 5.7352705705798136e-05}
{"train_loss": 0.08722056448459625, "global_step": 123246, "epoch": 1384, "lr": 5.735213227133693e-05}
{"train_loss": 0.10787853598594666, "global_step": 123247, "epoch": 1384, "lr": 5.7351558835887295e-05}
{"train_loss": 0.1261453926563263, "global_step": 123248, "epoch": 1384, "lr": 5.735098539944935e-05}
{"train_loss": 0.10516410320997238, "global_step": 123249, "epoch": 1384, "lr": 5.735041196202317e-05}
{"train_loss": 0.1154872477054596, "global_step": 123250, "epoch": 1384, "lr": 5.7349838523608787e-05}
{"train_loss": 0.06649075448513031, "global_step": 123251, "epoch": 1384, "lr": 5.7349265084206324e-05}
{"train_loss": 0.1406969279050827, "global_step": 123252, "epoch": 1384, "lr": 5.734869164381584e-05}
{"train_loss": 0.09807738661766052, "global_step": 123253, "epoch": 1384, "lr": 5.7348118202437415e-05}
{"train_loss": 0.1829589307308197, "global_step": 123254, "epoch": 1384, "lr": 5.7347544760071124e-05}
{"train_loss": 0.11568991839885712, "global_step": 123255, "epoch": 1384, "lr": 5.734697131671705e-05}
{"train_loss": 0.21913443505764008, "global_step": 123256, "epoch": 1384, "lr": 5.734639787237526e-05}
{"train_loss": 0.2034393846988678, "global_step": 123257, "epoch": 1384, "lr": 5.734582442704584e-05}
{"train_loss": 0.06062628701329231, "global_step": 123258, "epoch": 1384, "lr": 5.734525098072886e-05}
{"train_loss": 0.11175676435232162, "global_step": 123259, "epoch": 1384, "lr": 5.7344677533424396e-05}
{"train_loss": 0.07129526138305664, "global_step": 123260, "epoch": 1384, "lr": 5.734410408513255e-05}
{"train_loss": 0.1897318810224533, "global_step": 123261, "epoch": 1384, "lr": 5.734353063585335e-05}
{"train_loss": 0.14831706881523132, "global_step": 123262, "epoch": 1384, "lr": 5.734295718558693e-05}
{"train_loss": 0.09109731763601303, "global_step": 123263, "epoch": 1384, "lr": 5.7342383734333325e-05}
{"train_loss": 0.13039126335067697, "global_step": 123264, "epoch": 1384, "lr": 5.734181028209262e-05, "val_loss": 5.344840049743652}
{"train_loss": 0.1508762687444687, "global_step": 123265, "epoch": 1385, "lr": 5.7341236828864894e-05}
{"train_loss": 0.0876106321811676, "global_step": 123266, "epoch": 1385, "lr": 5.734066337465024e-05}
{"train_loss": 0.0764913558959961, "global_step": 123267, "epoch": 1385, "lr": 5.734008991944873e-05}
{"train_loss": 0.1092914268374443, "global_step": 123268, "epoch": 1385, "lr": 5.733951646326041e-05}
{"train_loss": 0.15538817644119263, "global_step": 123269, "epoch": 1385, "lr": 5.73389430060854e-05}
{"train_loss": 0.1800750494003296, "global_step": 123270, "epoch": 1385, "lr": 5.7338369547923755e-05}
{"train_loss": 0.14401525259017944, "global_step": 123271, "epoch": 1385, "lr": 5.7337796088775566e-05}
{"train_loss": 0.08136306703090668, "global_step": 123272, "epoch": 1385, "lr": 5.733722262864089e-05}
{"train_loss": 0.10485678166151047, "global_step": 123273, "epoch": 1385, "lr": 5.733664916751981e-05}
{"train_loss": 0.08016616106033325, "global_step": 123274, "epoch": 1385, "lr": 5.7336075705412406e-05}
{"train_loss": 0.09917157143354416, "global_step": 123275, "epoch": 1385, "lr": 5.7335502242318764e-05}
{"train_loss": 0.12154172360897064, "global_step": 123276, "epoch": 1385, "lr": 5.733492877823895e-05}
{"train_loss": 0.1622563898563385, "global_step": 123277, "epoch": 1385, "lr": 5.733435531317304e-05}
{"train_loss": 0.07099363207817078, "global_step": 123278, "epoch": 1385, "lr": 5.7333781847121125e-05}
{"train_loss": 0.10068517923355103, "global_step": 123279, "epoch": 1385, "lr": 5.733320838008326e-05}
{"train_loss": 0.10716478526592255, "global_step": 123280, "epoch": 1385, "lr": 5.7332634912059545e-05}
{"train_loss": 0.07611315697431564, "global_step": 123281, "epoch": 1385, "lr": 5.7332061443050056e-05}
{"train_loss": 0.11478053778409958, "global_step": 123282, "epoch": 1385, "lr": 5.7331487973054844e-05}
{"train_loss": 0.16438636183738708, "global_step": 123283, "epoch": 1385, "lr": 5.7330914502074e-05}
{"train_loss": 0.06325330585241318, "global_step": 123284, "epoch": 1385, "lr": 5.7330341030107604e-05}
{"train_loss": 0.13334639370441437, "global_step": 123285, "epoch": 1385, "lr": 5.732976755715573e-05}
{"train_loss": 0.1256083995103836, "global_step": 123286, "epoch": 1385, "lr": 5.732919408321848e-05}
{"train_loss": 0.0673089474439621, "global_step": 123287, "epoch": 1385, "lr": 5.732862060829589e-05}
{"train_loss": 0.13516093790531158, "global_step": 123288, "epoch": 1385, "lr": 5.7328047132388064e-05}
{"train_loss": 0.15538711845874786, "global_step": 123289, "epoch": 1385, "lr": 5.732747365549507e-05}
{"train_loss": 0.07513775676488876, "global_step": 123290, "epoch": 1385, "lr": 5.7326900177617e-05}
{"train_loss": 0.19620351493358612, "global_step": 123291, "epoch": 1385, "lr": 5.73263266987539e-05}
{"train_loss": 0.12318994849920273, "global_step": 123292, "epoch": 1385, "lr": 5.7325753218905864e-05}
{"train_loss": 0.10979598760604858, "global_step": 123293, "epoch": 1385, "lr": 5.7325179738072974e-05}
{"train_loss": 0.1309790462255478, "global_step": 123294, "epoch": 1385, "lr": 5.7324606256255296e-05}
{"train_loss": 0.17455995082855225, "global_step": 123295, "epoch": 1385, "lr": 5.7324032773452926e-05}
{"train_loss": 0.16345569491386414, "global_step": 123296, "epoch": 1385, "lr": 5.7323459289665924e-05}
{"train_loss": 0.17542211711406708, "global_step": 123297, "epoch": 1385, "lr": 5.732288580489438e-05}
{"train_loss": 0.07432674616575241, "global_step": 123298, "epoch": 1385, "lr": 5.732231231913835e-05}
{"train_loss": 0.08269863575696945, "global_step": 123299, "epoch": 1385, "lr": 5.732173883239793e-05}
{"train_loss": 0.08240443468093872, "global_step": 123300, "epoch": 1385, "lr": 5.732116534467319e-05}
{"train_loss": 0.12851937115192413, "global_step": 123301, "epoch": 1385, "lr": 5.732059185596422e-05}
{"train_loss": 0.17915187776088715, "global_step": 123302, "epoch": 1385, "lr": 5.732001836627107e-05}
{"train_loss": 0.11916681379079819, "global_step": 123303, "epoch": 1385, "lr": 5.731944487559384e-05}
{"train_loss": 0.11628369987010956, "global_step": 123304, "epoch": 1385, "lr": 5.7318871383932596e-05}
{"train_loss": 0.13621273636817932, "global_step": 123305, "epoch": 1385, "lr": 5.731829789128742e-05}
{"train_loss": 0.13856284320354462, "global_step": 123306, "epoch": 1385, "lr": 5.731772439765839e-05}
{"train_loss": 0.14069132506847382, "global_step": 123307, "epoch": 1385, "lr": 5.731715090304558e-05}
{"train_loss": 0.17251086235046387, "global_step": 123308, "epoch": 1385, "lr": 5.731657740744907e-05}
{"train_loss": 0.14134465157985687, "global_step": 123309, "epoch": 1385, "lr": 5.7316003910868934e-05}
{"train_loss": 0.16301977634429932, "global_step": 123310, "epoch": 1385, "lr": 5.7315430413305246e-05}
{"train_loss": 0.10002944618463516, "global_step": 123311, "epoch": 1385, "lr": 5.73148569147581e-05}
{"train_loss": 0.16763228178024292, "global_step": 123312, "epoch": 1385, "lr": 5.731428341522755e-05}
{"train_loss": 0.07124071568250656, "global_step": 123313, "epoch": 1385, "lr": 5.731370991471369e-05}
{"train_loss": 0.11663995683193207, "global_step": 123314, "epoch": 1385, "lr": 5.7313136413216584e-05}
{"train_loss": 0.079861119389534, "global_step": 123315, "epoch": 1385, "lr": 5.7312562910736323e-05}
{"train_loss": 0.12296776473522186, "global_step": 123316, "epoch": 1385, "lr": 5.731198940727297e-05}
{"train_loss": 0.1282171905040741, "global_step": 123317, "epoch": 1385, "lr": 5.731141590282662e-05}
{"train_loss": 0.08777839690446854, "global_step": 123318, "epoch": 1385, "lr": 5.731084239739733e-05}
{"train_loss": 0.1113453060388565, "global_step": 123319, "epoch": 1385, "lr": 5.731026889098519e-05}
{"train_loss": 0.162739560008049, "global_step": 123320, "epoch": 1385, "lr": 5.7309695383590265e-05}
{"train_loss": 0.13520953059196472, "global_step": 123321, "epoch": 1385, "lr": 5.730912187521264e-05}
{"train_loss": 0.08797609806060791, "global_step": 123322, "epoch": 1385, "lr": 5.73085483658524e-05}
{"train_loss": 0.16515645384788513, "global_step": 123323, "epoch": 1385, "lr": 5.7307974855509616e-05}
{"train_loss": 0.2028326541185379, "global_step": 123324, "epoch": 1385, "lr": 5.7307401344184366e-05}
{"train_loss": 0.12759555876255035, "global_step": 123325, "epoch": 1385, "lr": 5.730682783187672e-05}
{"train_loss": 0.09350895136594772, "global_step": 123326, "epoch": 1385, "lr": 5.730625431858676e-05}
{"train_loss": 0.09872901439666748, "global_step": 123327, "epoch": 1385, "lr": 5.730568080431456e-05}
{"train_loss": 0.15566591918468475, "global_step": 123328, "epoch": 1385, "lr": 5.730510728906021e-05}
{"train_loss": 0.1276676505804062, "global_step": 123329, "epoch": 1385, "lr": 5.730453377282378e-05}
{"train_loss": 0.0863746702671051, "global_step": 123330, "epoch": 1385, "lr": 5.730396025560533e-05}
{"train_loss": 0.1637575328350067, "global_step": 123331, "epoch": 1385, "lr": 5.730338673740496e-05}
{"train_loss": 0.14427219331264496, "global_step": 123332, "epoch": 1385, "lr": 5.7302813218222737e-05}
{"train_loss": 0.17195305228233337, "global_step": 123333, "epoch": 1385, "lr": 5.730223969805873e-05}
{"train_loss": 0.09119880944490433, "global_step": 123334, "epoch": 1385, "lr": 5.730166617691304e-05}
{"train_loss": 0.09647520631551743, "global_step": 123335, "epoch": 1385, "lr": 5.7301092654785726e-05}
{"train_loss": 0.13028934597969055, "global_step": 123336, "epoch": 1385, "lr": 5.7300519131676865e-05}
{"train_loss": 0.14332349598407745, "global_step": 123337, "epoch": 1385, "lr": 5.7299945607586545e-05}
{"train_loss": 0.13865378499031067, "global_step": 123338, "epoch": 1385, "lr": 5.729937208251483e-05}
{"train_loss": 0.0965191125869751, "global_step": 123339, "epoch": 1385, "lr": 5.729879855646181e-05}
{"train_loss": 0.19870434701442719, "global_step": 123340, "epoch": 1385, "lr": 5.729822502942755e-05}
{"train_loss": 0.22786365449428558, "global_step": 123341, "epoch": 1385, "lr": 5.729765150141214e-05}
{"train_loss": 0.18511153757572174, "global_step": 123342, "epoch": 1385, "lr": 5.7297077972415644e-05}
{"train_loss": 0.07382538169622421, "global_step": 123343, "epoch": 1385, "lr": 5.7296504442438146e-05}
{"train_loss": 0.07932213693857193, "global_step": 123344, "epoch": 1385, "lr": 5.7295930911479724e-05}
{"train_loss": 0.15057258307933807, "global_step": 123345, "epoch": 1385, "lr": 5.729535737954045e-05}
{"train_loss": 0.1558462381362915, "global_step": 123346, "epoch": 1385, "lr": 5.72947838466204e-05}
{"train_loss": 0.23472943902015686, "global_step": 123347, "epoch": 1385, "lr": 5.729421031271966e-05}
{"train_loss": 0.1180536225438118, "global_step": 123348, "epoch": 1385, "lr": 5.72936367778383e-05}
{"train_loss": 0.13308095932006836, "global_step": 123349, "epoch": 1385, "lr": 5.729306324197641e-05}
{"train_loss": 0.09110798686742783, "global_step": 123350, "epoch": 1385, "lr": 5.7292489705134035e-05}
{"train_loss": 0.152354434132576, "global_step": 123351, "epoch": 1385, "lr": 5.729191616731129e-05}
{"train_loss": 0.16327926516532898, "global_step": 123352, "epoch": 1385, "lr": 5.729134262850824e-05}
{"train_loss": 0.12741362973210518, "global_step": 123353, "epoch": 1385, "lr": 5.729076908872494e-05, "val_loss": 5.569582462310791, "train_action_mse_error": 5.040574550628662}
{"train_loss": 0.1301947683095932, "global_step": 123354, "epoch": 1386, "lr": 5.72901955479615e-05}
{"train_loss": 0.12245740741491318, "global_step": 123355, "epoch": 1386, "lr": 5.728962200621797e-05}
{"train_loss": 0.1111958771944046, "global_step": 123356, "epoch": 1386, "lr": 5.7289048463494455e-05}
{"train_loss": 0.1933165341615677, "global_step": 123357, "epoch": 1386, "lr": 5.7288474919791005e-05}
{"train_loss": 0.10042248666286469, "global_step": 123358, "epoch": 1386, "lr": 5.7287901375107714e-05}
{"train_loss": 0.1322602927684784, "global_step": 123359, "epoch": 1386, "lr": 5.728732782944465e-05}
{"train_loss": 0.1242305189371109, "global_step": 123360, "epoch": 1386, "lr": 5.728675428280189e-05}
{"train_loss": 0.12468387186527252, "global_step": 123361, "epoch": 1386, "lr": 5.7286180735179515e-05}
{"train_loss": 0.0837826281785965, "global_step": 123362, "epoch": 1386, "lr": 5.7285607186577606e-05}
{"train_loss": 0.14035531878471375, "global_step": 123363, "epoch": 1386, "lr": 5.7285033636996235e-05}
{"train_loss": 0.1429436355829239, "global_step": 123364, "epoch": 1386, "lr": 5.7284460086435485e-05}
{"train_loss": 0.07712344080209732, "global_step": 123365, "epoch": 1386, "lr": 5.728388653489543e-05}
{"train_loss": 0.1438724547624588, "global_step": 123366, "epoch": 1386, "lr": 5.728331298237613e-05}
{"train_loss": 0.1293887197971344, "global_step": 123367, "epoch": 1386, "lr": 5.7282739428877684e-05}
{"train_loss": 0.12596239149570465, "global_step": 123368, "epoch": 1386, "lr": 5.7282165874400164e-05}
{"train_loss": 0.130088210105896, "global_step": 123369, "epoch": 1386, "lr": 5.7281592318943645e-05}
{"train_loss": 0.14928020536899567, "global_step": 123370, "epoch": 1386, "lr": 5.728101876250821e-05}
{"train_loss": 0.16413791477680206, "global_step": 123371, "epoch": 1386, "lr": 5.728044520509392e-05}
{"train_loss": 0.08179159462451935, "global_step": 123372, "epoch": 1386, "lr": 5.727987164670088e-05}
{"train_loss": 0.13144485652446747, "global_step": 123373, "epoch": 1386, "lr": 5.727929808732914e-05}
{"train_loss": 0.08569755405187607, "global_step": 123374, "epoch": 1386, "lr": 5.727872452697878e-05}
{"train_loss": 0.12699395418167114, "global_step": 123375, "epoch": 1386, "lr": 5.727815096564989e-05}
{"train_loss": 0.14172276854515076, "global_step": 123376, "epoch": 1386, "lr": 5.727757740334254e-05}
{"train_loss": 0.17800873517990112, "global_step": 123377, "epoch": 1386, "lr": 5.7277003840056806e-05}
{"train_loss": 0.13075929880142212, "global_step": 123378, "epoch": 1386, "lr": 5.7276430275792784e-05}
{"train_loss": 0.1126192957162857, "global_step": 123379, "epoch": 1386, "lr": 5.727585671055051e-05}
{"train_loss": 0.12054891884326935, "global_step": 123380, "epoch": 1386, "lr": 5.727528314433011e-05}
{"train_loss": 0.09756750613451004, "global_step": 123381, "epoch": 1386, "lr": 5.727470957713161e-05}
{"train_loss": 0.10022648423910141, "global_step": 123382, "epoch": 1386, "lr": 5.727413600895514e-05}
{"train_loss": 0.13405445218086243, "global_step": 123383, "epoch": 1386, "lr": 5.727356243980073e-05}
{"train_loss": 0.13363999128341675, "global_step": 123384, "epoch": 1386, "lr": 5.72729888696685e-05}
{"train_loss": 0.10032021254301071, "global_step": 123385, "epoch": 1386, "lr": 5.727241529855849e-05}
{"train_loss": 0.08888792991638184, "global_step": 123386, "epoch": 1386, "lr": 5.72718417264708e-05}
{"train_loss": 0.17165985703468323, "global_step": 123387, "epoch": 1386, "lr": 5.727126815340549e-05}
{"train_loss": 0.07998436689376831, "global_step": 123388, "epoch": 1386, "lr": 5.727069457936265e-05}
{"train_loss": 0.2394072413444519, "global_step": 123389, "epoch": 1386, "lr": 5.727012100434236e-05}
{"train_loss": 0.07389869540929794, "global_step": 123390, "epoch": 1386, "lr": 5.726954742834468e-05}
{"train_loss": 0.0798647403717041, "global_step": 123391, "epoch": 1386, "lr": 5.726897385136971e-05}
{"train_loss": 0.07549500465393066, "global_step": 123392, "epoch": 1386, "lr": 5.726840027341751e-05}
{"train_loss": 0.15928475558757782, "global_step": 123393, "epoch": 1386, "lr": 5.726782669448816e-05}
{"train_loss": 0.08651428669691086, "global_step": 123394, "epoch": 1386, "lr": 5.726725311458174e-05}
{"train_loss": 0.1517271101474762, "global_step": 123395, "epoch": 1386, "lr": 5.726667953369833e-05}
{"train_loss": 0.12446745485067368, "global_step": 123396, "epoch": 1386, "lr": 5.7266105951838e-05}
{"train_loss": 0.10668279230594635, "global_step": 123397, "epoch": 1386, "lr": 5.726553236900083e-05}
{"train_loss": 0.14850081503391266, "global_step": 123398, "epoch": 1386, "lr": 5.7264958785186904e-05}
{"train_loss": 0.14938387274742126, "global_step": 123399, "epoch": 1386, "lr": 5.7264385200396285e-05}
{"train_loss": 0.15055546164512634, "global_step": 123400, "epoch": 1386, "lr": 5.726381161462906e-05}
{"train_loss": 0.17682112753391266, "global_step": 123401, "epoch": 1386, "lr": 5.7263238027885315e-05}
{"train_loss": 0.15801140666007996, "global_step": 123402, "epoch": 1386, "lr": 5.72626644401651e-05}
{"train_loss": 0.08610611408948898, "global_step": 123403, "epoch": 1386, "lr": 5.726209085146851e-05}
{"train_loss": 0.13301724195480347, "global_step": 123404, "epoch": 1386, "lr": 5.7261517261795624e-05}
{"train_loss": 0.16121530532836914, "global_step": 123405, "epoch": 1386, "lr": 5.726094367114652e-05}
{"train_loss": 0.1474926769733429, "global_step": 123406, "epoch": 1386, "lr": 5.7260370079521265e-05}
{"train_loss": 0.097638338804245, "global_step": 123407, "epoch": 1386, "lr": 5.7259796486919945e-05}
{"train_loss": 0.09280440211296082, "global_step": 123408, "epoch": 1386, "lr": 5.725922289334263e-05}
{"train_loss": 0.10519742965698242, "global_step": 123409, "epoch": 1386, "lr": 5.725864929878941e-05}
{"train_loss": 0.12819728255271912, "global_step": 123410, "epoch": 1386, "lr": 5.725807570326034e-05}
{"train_loss": 0.155836284160614, "global_step": 123411, "epoch": 1386, "lr": 5.7257502106755524e-05}
{"train_loss": 0.10647515952587128, "global_step": 123412, "epoch": 1386, "lr": 5.725692850927502e-05}
{"train_loss": 0.13325515389442444, "global_step": 123413, "epoch": 1386, "lr": 5.725635491081891e-05}
{"train_loss": 0.1201719269156456, "global_step": 123414, "epoch": 1386, "lr": 5.725578131138727e-05}
{"train_loss": 0.13824553787708282, "global_step": 123415, "epoch": 1386, "lr": 5.725520771098018e-05}
{"train_loss": 0.11459857225418091, "global_step": 123416, "epoch": 1386, "lr": 5.725463410959771e-05}
{"train_loss": 0.08428589254617691, "global_step": 123417, "epoch": 1386, "lr": 5.725406050723995e-05}
{"train_loss": 0.09314113110303879, "global_step": 123418, "epoch": 1386, "lr": 5.725348690390696e-05}
{"train_loss": 0.15478844940662384, "global_step": 123419, "epoch": 1386, "lr": 5.7252913299598844e-05}
{"train_loss": 0.09348028898239136, "global_step": 123420, "epoch": 1386, "lr": 5.725233969431565e-05}
{"train_loss": 0.10719182342290878, "global_step": 123421, "epoch": 1386, "lr": 5.725176608805747e-05}
{"train_loss": 0.13769052922725677, "global_step": 123422, "epoch": 1386, "lr": 5.725119248082438e-05}
{"train_loss": 0.13202965259552002, "global_step": 123423, "epoch": 1386, "lr": 5.7250618872616445e-05}
{"train_loss": 0.08592347055673599, "global_step": 123424, "epoch": 1386, "lr": 5.725004526343376e-05}
{"train_loss": 0.09503205865621567, "global_step": 123425, "epoch": 1386, "lr": 5.72494716532764e-05}
{"train_loss": 0.10514913499355316, "global_step": 123426, "epoch": 1386, "lr": 5.724889804214444e-05}
{"train_loss": 0.20049028098583221, "global_step": 123427, "epoch": 1386, "lr": 5.724832443003795e-05}
{"train_loss": 0.13454832136631012, "global_step": 123428, "epoch": 1386, "lr": 5.7247750816957004e-05}
{"train_loss": 0.10306722670793533, "global_step": 123429, "epoch": 1386, "lr": 5.724717720290168e-05}
{"train_loss": 0.08307787030935287, "global_step": 123430, "epoch": 1386, "lr": 5.7246603587872073e-05}
{"train_loss": 0.11240866035223007, "global_step": 123431, "epoch": 1386, "lr": 5.7246029971868256e-05}
{"train_loss": 0.10269060730934143, "global_step": 123432, "epoch": 1386, "lr": 5.724545635489029e-05}
{"train_loss": 0.21841807663440704, "global_step": 123433, "epoch": 1386, "lr": 5.724488273693826e-05}
{"train_loss": 0.16854850947856903, "global_step": 123434, "epoch": 1386, "lr": 5.724430911801224e-05}
{"train_loss": 0.1800282895565033, "global_step": 123435, "epoch": 1386, "lr": 5.724373549811232e-05}
{"train_loss": 0.11604197323322296, "global_step": 123436, "epoch": 1386, "lr": 5.724316187723856e-05}
{"train_loss": 0.1681929975748062, "global_step": 123437, "epoch": 1386, "lr": 5.724258825539105e-05}
{"train_loss": 0.13571766018867493, "global_step": 123438, "epoch": 1386, "lr": 5.724201463256986e-05}
{"train_loss": 0.16440744698047638, "global_step": 123439, "epoch": 1386, "lr": 5.7241441008775074e-05}
{"train_loss": 0.10767646878957748, "global_step": 123440, "epoch": 1386, "lr": 5.7240867384006756e-05}
{"train_loss": 0.12158757448196411, "global_step": 123441, "epoch": 1386, "lr": 5.724029375826501e-05}
{"train_loss": 0.12708564470993, "global_step": 123442, "epoch": 1386, "lr": 5.7239720131549876e-05, "val_loss": 5.472323894500732}
{"train_loss": 0.12210113555192947, "global_step": 123443, "epoch": 1387, "lr": 5.723914650386145e-05}
{"train_loss": 0.1238447055220604, "global_step": 123444, "epoch": 1387, "lr": 5.7238572875199816e-05}
{"train_loss": 0.07537343353033066, "global_step": 123445, "epoch": 1387, "lr": 5.723799924556504e-05}
{"train_loss": 0.15171648561954498, "global_step": 123446, "epoch": 1387, "lr": 5.7237425614957206e-05}
{"train_loss": 0.11402671784162521, "global_step": 123447, "epoch": 1387, "lr": 5.7236851983376385e-05}
{"train_loss": 0.09160210192203522, "global_step": 123448, "epoch": 1387, "lr": 5.7236278350822655e-05}
{"train_loss": 0.130652517080307, "global_step": 123449, "epoch": 1387, "lr": 5.72357047172961e-05}
{"train_loss": 0.10502566397190094, "global_step": 123450, "epoch": 1387, "lr": 5.723513108279681e-05}
{"train_loss": 0.050915222615003586, "global_step": 123451, "epoch": 1387, "lr": 5.723455744732482e-05}
{"train_loss": 0.1681506484746933, "global_step": 123452, "epoch": 1387, "lr": 5.723398381088024e-05}
{"train_loss": 0.12689849734306335, "global_step": 123453, "epoch": 1387, "lr": 5.723341017346314e-05}
{"train_loss": 0.1589086502790451, "global_step": 123454, "epoch": 1387, "lr": 5.7232836535073597e-05}
{"train_loss": 0.08301553875207901, "global_step": 123455, "epoch": 1387, "lr": 5.723226289571169e-05}
{"train_loss": 0.13868242502212524, "global_step": 123456, "epoch": 1387, "lr": 5.7231689255377485e-05}
{"train_loss": 0.12222068011760712, "global_step": 123457, "epoch": 1387, "lr": 5.723111561407107e-05}
{"train_loss": 0.08758174628019333, "global_step": 123458, "epoch": 1387, "lr": 5.723054197179253e-05}
{"train_loss": 0.06995537877082825, "global_step": 123459, "epoch": 1387, "lr": 5.7229968328541913e-05}
{"train_loss": 0.0773022249341011, "global_step": 123460, "epoch": 1387, "lr": 5.7229394684319336e-05}
{"train_loss": 0.17834675312042236, "global_step": 123461, "epoch": 1387, "lr": 5.722882103912484e-05}
{"train_loss": 0.12879571318626404, "global_step": 123462, "epoch": 1387, "lr": 5.7228247392958524e-05}
{"train_loss": 0.12604519724845886, "global_step": 123463, "epoch": 1387, "lr": 5.722767374582045e-05}
{"train_loss": 0.14830045402050018, "global_step": 123464, "epoch": 1387, "lr": 5.7227100097710705e-05}
{"train_loss": 0.08248240500688553, "global_step": 123465, "epoch": 1387, "lr": 5.722652644862937e-05}
{"train_loss": 0.10536131262779236, "global_step": 123466, "epoch": 1387, "lr": 5.722595279857651e-05}
{"train_loss": 0.07181185483932495, "global_step": 123467, "epoch": 1387, "lr": 5.7225379147552216e-05}
{"train_loss": 0.13695351779460907, "global_step": 123468, "epoch": 1387, "lr": 5.722480549555655e-05}
{"train_loss": 0.18009977042675018, "global_step": 123469, "epoch": 1387, "lr": 5.7224231842589614e-05}
{"train_loss": 0.16114313900470734, "global_step": 123470, "epoch": 1387, "lr": 5.722365818865146e-05}
{"train_loss": 0.14753326773643494, "global_step": 123471, "epoch": 1387, "lr": 5.7223084533742166e-05}
{"train_loss": 0.19895902276039124, "global_step": 123472, "epoch": 1387, "lr": 5.7222510877861827e-05}
{"train_loss": 0.15198838710784912, "global_step": 123473, "epoch": 1387, "lr": 5.72219372210105e-05}
{"train_loss": 0.20447774231433868, "global_step": 123474, "epoch": 1387, "lr": 5.722136356318827e-05}
{"train_loss": 0.20126263797283173, "global_step": 123475, "epoch": 1387, "lr": 5.722078990439522e-05}
{"train_loss": 0.15477994084358215, "global_step": 123476, "epoch": 1387, "lr": 5.722021624463142e-05}
{"train_loss": 0.13987289369106293, "global_step": 123477, "epoch": 1387, "lr": 5.7219642583896956e-05}
{"train_loss": 0.13622571527957916, "global_step": 123478, "epoch": 1387, "lr": 5.72190689221919e-05}
{"train_loss": 0.1195124164223671, "global_step": 123479, "epoch": 1387, "lr": 5.721849525951633e-05}
{"train_loss": 0.16652831435203552, "global_step": 123480, "epoch": 1387, "lr": 5.721792159587031e-05}
{"train_loss": 0.142923966050148, "global_step": 123481, "epoch": 1387, "lr": 5.721734793125393e-05}
{"train_loss": 0.1777443140745163, "global_step": 123482, "epoch": 1387, "lr": 5.721677426566727e-05}
{"train_loss": 0.14049652218818665, "global_step": 123483, "epoch": 1387, "lr": 5.721620059911039e-05}
{"train_loss": 0.09700456261634827, "global_step": 123484, "epoch": 1387, "lr": 5.721562693158339e-05}
{"train_loss": 0.21751084923744202, "global_step": 123485, "epoch": 1387, "lr": 5.721505326308634e-05}
{"train_loss": 0.14977246522903442, "global_step": 123486, "epoch": 1387, "lr": 5.7214479593619306e-05}
{"train_loss": 0.13417533040046692, "global_step": 123487, "epoch": 1387, "lr": 5.7213905923182386e-05}
{"train_loss": 0.12882685661315918, "global_step": 123488, "epoch": 1387, "lr": 5.721333225177563e-05}
{"train_loss": 0.13837160170078278, "global_step": 123489, "epoch": 1387, "lr": 5.721275857939915e-05}
{"train_loss": 0.14941208064556122, "global_step": 123490, "epoch": 1387, "lr": 5.721218490605299e-05}
{"train_loss": 0.07037336379289627, "global_step": 123491, "epoch": 1387, "lr": 5.721161123173724e-05}
{"train_loss": 0.13172248005867004, "global_step": 123492, "epoch": 1387, "lr": 5.721103755645197e-05}
{"train_loss": 0.12021681666374207, "global_step": 123493, "epoch": 1387, "lr": 5.721046388019727e-05}
{"train_loss": 0.11736662685871124, "global_step": 123494, "epoch": 1387, "lr": 5.720989020297321e-05}
{"train_loss": 0.22012993693351746, "global_step": 123495, "epoch": 1387, "lr": 5.720931652477988e-05}
{"train_loss": 0.14282463490962982, "global_step": 123496, "epoch": 1387, "lr": 5.720874284561733e-05}
{"train_loss": 0.17877155542373657, "global_step": 123497, "epoch": 1387, "lr": 5.720816916548566e-05}
{"train_loss": 0.12511098384857178, "global_step": 123498, "epoch": 1387, "lr": 5.7207595484384936e-05}
{"train_loss": 0.11162998527288437, "global_step": 123499, "epoch": 1387, "lr": 5.720702180231524e-05}
{"train_loss": 0.11231976002454758, "global_step": 123500, "epoch": 1387, "lr": 5.7206448119276645e-05}
{"train_loss": 0.15172308683395386, "global_step": 123501, "epoch": 1387, "lr": 5.720587443526922e-05}
{"train_loss": 0.13556401431560516, "global_step": 123502, "epoch": 1387, "lr": 5.720530075029307e-05}
{"train_loss": 0.1469610184431076, "global_step": 123503, "epoch": 1387, "lr": 5.720472706434825e-05}
{"train_loss": 0.13575252890586853, "global_step": 123504, "epoch": 1387, "lr": 5.720415337743484e-05}
{"train_loss": 0.14525346457958221, "global_step": 123505, "epoch": 1387, "lr": 5.720357968955292e-05}
{"train_loss": 0.1516176015138626, "global_step": 123506, "epoch": 1387, "lr": 5.7203006000702574e-05}
{"train_loss": 0.211464062333107, "global_step": 123507, "epoch": 1387, "lr": 5.720243231088387e-05}
{"train_loss": 0.10399103164672852, "global_step": 123508, "epoch": 1387, "lr": 5.720185862009688e-05}
{"train_loss": 0.16730238497257233, "global_step": 123509, "epoch": 1387, "lr": 5.72012849283417e-05}
{"train_loss": 0.17310181260108948, "global_step": 123510, "epoch": 1387, "lr": 5.720071123561839e-05}
{"train_loss": 0.09628652781248093, "global_step": 123511, "epoch": 1387, "lr": 5.720013754192702e-05}
{"train_loss": 0.14417852461338043, "global_step": 123512, "epoch": 1387, "lr": 5.719956384726769e-05}
{"train_loss": 0.17066006362438202, "global_step": 123513, "epoch": 1387, "lr": 5.7198990151640465e-05}
{"train_loss": 0.11306696385145187, "global_step": 123514, "epoch": 1387, "lr": 5.719841645504542e-05}
{"train_loss": 0.16310295462608337, "global_step": 123515, "epoch": 1387, "lr": 5.719784275748264e-05}
{"train_loss": 0.10916408151388168, "global_step": 123516, "epoch": 1387, "lr": 5.7197269058952196e-05}
{"train_loss": 0.10753008723258972, "global_step": 123517, "epoch": 1387, "lr": 5.7196695359454166e-05}
{"train_loss": 0.06026238948106766, "global_step": 123518, "epoch": 1387, "lr": 5.7196121658988635e-05}
{"train_loss": 0.06840836256742477, "global_step": 123519, "epoch": 1387, "lr": 5.719554795755565e-05}
{"train_loss": 0.20009763538837433, "global_step": 123520, "epoch": 1387, "lr": 5.719497425515533e-05}
{"train_loss": 0.18059071898460388, "global_step": 123521, "epoch": 1387, "lr": 5.719440055178773e-05}
{"train_loss": 0.10033073276281357, "global_step": 123522, "epoch": 1387, "lr": 5.719382684745294e-05}
{"train_loss": 0.15463876724243164, "global_step": 123523, "epoch": 1387, "lr": 5.719325314215101e-05}
{"train_loss": 0.1278582662343979, "global_step": 123524, "epoch": 1387, "lr": 5.7192679435882045e-05}
{"train_loss": 0.12937507033348083, "global_step": 123525, "epoch": 1387, "lr": 5.719210572864612e-05}
{"train_loss": 0.17558865249156952, "global_step": 123526, "epoch": 1387, "lr": 5.719153202044329e-05}
{"train_loss": 0.15752851963043213, "global_step": 123527, "epoch": 1387, "lr": 5.719095831127366e-05}
{"train_loss": 0.17088750004768372, "global_step": 123528, "epoch": 1387, "lr": 5.719038460113728e-05}
{"train_loss": 0.14684823155403137, "global_step": 123529, "epoch": 1387, "lr": 5.7189810890034255e-05}
{"train_loss": 0.17337927222251892, "global_step": 123530, "epoch": 1387, "lr": 5.718923717796464e-05}
{"train_loss": 0.13706768358523927, "global_step": 123531, "epoch": 1387, "lr": 5.718866346492851e-05, "val_loss": 5.365103721618652}
{"train_loss": 0.10217499732971191, "global_step": 123532, "epoch": 1388, "lr": 5.7188089750925965e-05}
{"train_loss": 0.1099851131439209, "global_step": 123533, "epoch": 1388, "lr": 5.718751603595707e-05}
{"train_loss": 0.13130392134189606, "global_step": 123534, "epoch": 1388, "lr": 5.718694232002188e-05}
{"train_loss": 0.09103011339902878, "global_step": 123535, "epoch": 1388, "lr": 5.718636860312051e-05}
{"train_loss": 0.08607305586338043, "global_step": 123536, "epoch": 1388, "lr": 5.718579488525302e-05}
{"train_loss": 0.1025644987821579, "global_step": 123537, "epoch": 1388, "lr": 5.718522116641948e-05}
{"train_loss": 0.11873770505189896, "global_step": 123538, "epoch": 1388, "lr": 5.718464744661998e-05}
{"train_loss": 0.15761692821979523, "global_step": 123539, "epoch": 1388, "lr": 5.718407372585459e-05}
{"train_loss": 0.12812703847885132, "global_step": 123540, "epoch": 1388, "lr": 5.718350000412339e-05}
{"train_loss": 0.1170869767665863, "global_step": 123541, "epoch": 1388, "lr": 5.7182926281426455e-05}
{"train_loss": 0.1119765117764473, "global_step": 123542, "epoch": 1388, "lr": 5.718235255776386e-05}
{"train_loss": 0.10207866132259369, "global_step": 123543, "epoch": 1388, "lr": 5.718177883313569e-05}
{"train_loss": 0.14216843247413635, "global_step": 123544, "epoch": 1388, "lr": 5.718120510754202e-05}
{"train_loss": 0.14919063448905945, "global_step": 123545, "epoch": 1388, "lr": 5.718063138098293e-05}
{"train_loss": 0.09137546271085739, "global_step": 123546, "epoch": 1388, "lr": 5.7180057653458476e-05}
{"train_loss": 0.12417049705982208, "global_step": 123547, "epoch": 1388, "lr": 5.717948392496877e-05}
{"train_loss": 0.13961677253246307, "global_step": 123548, "epoch": 1388, "lr": 5.7178910195513856e-05}
{"train_loss": 0.09769203513860703, "global_step": 123549, "epoch": 1388, "lr": 5.717833646509383e-05}
{"train_loss": 0.09742309153079987, "global_step": 123550, "epoch": 1388, "lr": 5.7177762733708764e-05}
{"train_loss": 0.1731928288936615, "global_step": 123551, "epoch": 1388, "lr": 5.717718900135873e-05}
{"train_loss": 0.19490714371204376, "global_step": 123552, "epoch": 1388, "lr": 5.717661526804381e-05}
{"train_loss": 0.12475858628749847, "global_step": 123553, "epoch": 1388, "lr": 5.717604153376409e-05}
{"train_loss": 0.12798422574996948, "global_step": 123554, "epoch": 1388, "lr": 5.7175467798519635e-05}
{"train_loss": 0.08949116617441177, "global_step": 123555, "epoch": 1388, "lr": 5.7174894062310526e-05}
{"train_loss": 0.11310472339391708, "global_step": 123556, "epoch": 1388, "lr": 5.717432032513682e-05}
{"train_loss": 0.08555866777896881, "global_step": 123557, "epoch": 1388, "lr": 5.7173746586998634e-05}
{"train_loss": 0.14217624068260193, "global_step": 123558, "epoch": 1388, "lr": 5.7173172847896027e-05}
{"train_loss": 0.18086954951286316, "global_step": 123559, "epoch": 1388, "lr": 5.717259910782906e-05}
{"train_loss": 0.1213194876909256, "global_step": 123560, "epoch": 1388, "lr": 5.717202536679784e-05}
{"train_loss": 0.14512676000595093, "global_step": 123561, "epoch": 1388, "lr": 5.717145162480242e-05}
{"train_loss": 0.10008233040571213, "global_step": 123562, "epoch": 1388, "lr": 5.71708778818429e-05}
{"train_loss": 0.12260864675045013, "global_step": 123563, "epoch": 1388, "lr": 5.717030413791933e-05}
{"train_loss": 0.11233189702033997, "global_step": 123564, "epoch": 1388, "lr": 5.7169730393031796e-05}
{"train_loss": 0.16189053654670715, "global_step": 123565, "epoch": 1388, "lr": 5.716915664718039e-05}
{"train_loss": 0.1417200267314911, "global_step": 123566, "epoch": 1388, "lr": 5.716858290036519e-05}
{"train_loss": 0.13871066272258759, "global_step": 123567, "epoch": 1388, "lr": 5.7168009152586247e-05}
{"train_loss": 0.1635606586933136, "global_step": 123568, "epoch": 1388, "lr": 5.716743540384364e-05}
{"train_loss": 0.13673178851604462, "global_step": 123569, "epoch": 1388, "lr": 5.716686165413747e-05}
{"train_loss": 0.11327698826789856, "global_step": 123570, "epoch": 1388, "lr": 5.7166287903467804e-05}
{"train_loss": 0.19273383915424347, "global_step": 123571, "epoch": 1388, "lr": 5.7165714151834714e-05}
{"train_loss": 0.11248476803302765, "global_step": 123572, "epoch": 1388, "lr": 5.716514039923828e-05}
{"train_loss": 0.12265487015247345, "global_step": 123573, "epoch": 1388, "lr": 5.7164566645678584e-05}
{"train_loss": 0.1673480123281479, "global_step": 123574, "epoch": 1388, "lr": 5.7163992891155695e-05}
{"train_loss": 0.16864332556724548, "global_step": 123575, "epoch": 1388, "lr": 5.7163419135669695e-05}
{"train_loss": 0.11329402774572372, "global_step": 123576, "epoch": 1388, "lr": 5.7162845379220666e-05}
{"train_loss": 0.12924614548683167, "global_step": 123577, "epoch": 1388, "lr": 5.716227162180867e-05}
{"train_loss": 0.18135201930999756, "global_step": 123578, "epoch": 1388, "lr": 5.7161697863433806e-05}
{"train_loss": 0.07681235671043396, "global_step": 123579, "epoch": 1388, "lr": 5.716112410409613e-05}
{"train_loss": 0.15971167385578156, "global_step": 123580, "epoch": 1388, "lr": 5.716055034379574e-05}
{"train_loss": 0.10511383414268494, "global_step": 123581, "epoch": 1388, "lr": 5.7159976582532696e-05}
{"train_loss": 0.10983823239803314, "global_step": 123582, "epoch": 1388, "lr": 5.715940282030707e-05}
{"train_loss": 0.11665567755699158, "global_step": 123583, "epoch": 1388, "lr": 5.7158829057118954e-05}
{"train_loss": 0.15293796360492706, "global_step": 123584, "epoch": 1388, "lr": 5.715825529296843e-05}
{"train_loss": 0.10574766993522644, "global_step": 123585, "epoch": 1388, "lr": 5.715768152785557e-05}
{"train_loss": 0.12342990189790726, "global_step": 123586, "epoch": 1388, "lr": 5.715710776178044e-05}
{"train_loss": 0.1913204789161682, "global_step": 123587, "epoch": 1388, "lr": 5.7156533994743124e-05}
{"train_loss": 0.10177060216665268, "global_step": 123588, "epoch": 1388, "lr": 5.7155960226743697e-05}
{"train_loss": 0.12617887556552887, "global_step": 123589, "epoch": 1388, "lr": 5.7155386457782236e-05}
{"train_loss": 0.135624960064888, "global_step": 123590, "epoch": 1388, "lr": 5.715481268785883e-05}
{"train_loss": 0.15175116062164307, "global_step": 123591, "epoch": 1388, "lr": 5.7154238916973535e-05}
{"train_loss": 0.1111486479640007, "global_step": 123592, "epoch": 1388, "lr": 5.7153665145126436e-05}
{"train_loss": 0.1319378912448883, "global_step": 123593, "epoch": 1388, "lr": 5.715309137231761e-05}
{"train_loss": 0.14213505387306213, "global_step": 123594, "epoch": 1388, "lr": 5.715251759854715e-05}
{"train_loss": 0.10545176267623901, "global_step": 123595, "epoch": 1388, "lr": 5.7151943823815125e-05}
{"train_loss": 0.13968366384506226, "global_step": 123596, "epoch": 1388, "lr": 5.715137004812161e-05}
{"train_loss": 0.0812288224697113, "global_step": 123597, "epoch": 1388, "lr": 5.7150796271466664e-05}
{"train_loss": 0.1293821632862091, "global_step": 123598, "epoch": 1388, "lr": 5.7150222493850404e-05}
{"train_loss": 0.14967553317546844, "global_step": 123599, "epoch": 1388, "lr": 5.714964871527286e-05}
{"train_loss": 0.11847859621047974, "global_step": 123600, "epoch": 1388, "lr": 5.714907493573415e-05}
{"train_loss": 0.09955427795648575, "global_step": 123601, "epoch": 1388, "lr": 5.7148501155234325e-05}
{"train_loss": 0.1437591016292572, "global_step": 123602, "epoch": 1388, "lr": 5.714792737377348e-05}
{"train_loss": 0.06988056004047394, "global_step": 123603, "epoch": 1388, "lr": 5.7147353591351674e-05}
{"train_loss": 0.11558646708726883, "global_step": 123604, "epoch": 1388, "lr": 5.7146779807968996e-05}
{"train_loss": 0.1006481796503067, "global_step": 123605, "epoch": 1388, "lr": 5.7146206023625526e-05}
{"train_loss": 0.14472994208335876, "global_step": 123606, "epoch": 1388, "lr": 5.714563223832134e-05}
{"train_loss": 0.1874578297138214, "global_step": 123607, "epoch": 1388, "lr": 5.71450584520565e-05}
{"train_loss": 0.17655351758003235, "global_step": 123608, "epoch": 1388, "lr": 5.7144484664831096e-05}
{"train_loss": 0.09351096302270889, "global_step": 123609, "epoch": 1388, "lr": 5.71439108766452e-05}
{"train_loss": 0.06144650653004646, "global_step": 123610, "epoch": 1388, "lr": 5.714333708749888e-05}
{"train_loss": 0.13480617105960846, "global_step": 123611, "epoch": 1388, "lr": 5.714276329739224e-05}
{"train_loss": 0.11305554211139679, "global_step": 123612, "epoch": 1388, "lr": 5.714218950632535e-05}
{"train_loss": 0.11254700273275375, "global_step": 123613, "epoch": 1388, "lr": 5.7141615714298266e-05}
{"train_loss": 0.1358419507741928, "global_step": 123614, "epoch": 1388, "lr": 5.714104192131109e-05}
{"train_loss": 0.10719998180866241, "global_step": 123615, "epoch": 1388, "lr": 5.714046812736389e-05}
{"train_loss": 0.10134707391262054, "global_step": 123616, "epoch": 1388, "lr": 5.7139894332456734e-05}
{"train_loss": 0.07337357848882675, "global_step": 123617, "epoch": 1388, "lr": 5.7139320536589705e-05}
{"train_loss": 0.10897329449653625, "global_step": 123618, "epoch": 1388, "lr": 5.713874673976288e-05}
{"train_loss": 0.2270534336566925, "global_step": 123619, "epoch": 1388, "lr": 5.713817294197634e-05}
{"train_loss": 0.1267953256793906, "global_step": 123620, "epoch": 1388, "lr": 5.713759914323016e-05, "val_loss": 5.397703647613525}
{"train_loss": 0.1173822432756424, "global_step": 123621, "epoch": 1389, "lr": 5.713702534352442e-05}
{"train_loss": 0.11429423838853836, "global_step": 123622, "epoch": 1389, "lr": 5.713645154285919e-05}
{"train_loss": 0.09205322712659836, "global_step": 123623, "epoch": 1389, "lr": 5.7135877741234554e-05}
{"train_loss": 0.15221095085144043, "global_step": 123624, "epoch": 1389, "lr": 5.7135303938650595e-05}
{"train_loss": 0.12490715831518173, "global_step": 123625, "epoch": 1389, "lr": 5.7134730135107375e-05}
{"train_loss": 0.07028607279062271, "global_step": 123626, "epoch": 1389, "lr": 5.713415633060497e-05}
{"train_loss": 0.11930856108665466, "global_step": 123627, "epoch": 1389, "lr": 5.713358252514347e-05}
{"train_loss": 0.24050936102867126, "global_step": 123628, "epoch": 1389, "lr": 5.713300871872294e-05}
{"train_loss": 0.1118709146976471, "global_step": 123629, "epoch": 1389, "lr": 5.713243491134346e-05}
{"train_loss": 0.05597245693206787, "global_step": 123630, "epoch": 1389, "lr": 5.713186110300512e-05}
{"train_loss": 0.12974387407302856, "global_step": 123631, "epoch": 1389, "lr": 5.7131287293707994e-05}
{"train_loss": 0.12358696758747101, "global_step": 123632, "epoch": 1389, "lr": 5.7130713483452146e-05}
{"train_loss": 0.12586769461631775, "global_step": 123633, "epoch": 1389, "lr": 5.713013967223766e-05}
{"train_loss": 0.15676802396774292, "global_step": 123634, "epoch": 1389, "lr": 5.712956586006462e-05}
{"train_loss": 0.09497638791799545, "global_step": 123635, "epoch": 1389, "lr": 5.7128992046933085e-05}
{"train_loss": 0.15000464022159576, "global_step": 123636, "epoch": 1389, "lr": 5.712841823284315e-05}
{"train_loss": 0.11296543478965759, "global_step": 123637, "epoch": 1389, "lr": 5.712784441779489e-05}
{"train_loss": 0.0726875439286232, "global_step": 123638, "epoch": 1389, "lr": 5.7127270601788377e-05}
{"train_loss": 0.13114199042320251, "global_step": 123639, "epoch": 1389, "lr": 5.712669678482369e-05}
{"train_loss": 0.20081906020641327, "global_step": 123640, "epoch": 1389, "lr": 5.71261229669009e-05}
{"train_loss": 0.061714813113212585, "global_step": 123641, "epoch": 1389, "lr": 5.7125549148020086e-05}
{"train_loss": 0.1531757116317749, "global_step": 123642, "epoch": 1389, "lr": 5.712497532818134e-05}
{"train_loss": 0.11728571355342865, "global_step": 123643, "epoch": 1389, "lr": 5.712440150738473e-05}
{"train_loss": 0.1866370141506195, "global_step": 123644, "epoch": 1389, "lr": 5.712382768563033e-05}
{"train_loss": 0.0850023552775383, "global_step": 123645, "epoch": 1389, "lr": 5.712325386291821e-05}
{"train_loss": 0.1003590077161789, "global_step": 123646, "epoch": 1389, "lr": 5.712268003924846e-05}
{"train_loss": 0.1322767287492752, "global_step": 123647, "epoch": 1389, "lr": 5.712210621462114e-05}
{"train_loss": 0.08673612028360367, "global_step": 123648, "epoch": 1389, "lr": 5.712153238903635e-05}
{"train_loss": 0.1405978947877884, "global_step": 123649, "epoch": 1389, "lr": 5.712095856249415e-05}
{"train_loss": 0.08458053320646286, "global_step": 123650, "epoch": 1389, "lr": 5.712038473499464e-05}
{"train_loss": 0.15027180314064026, "global_step": 123651, "epoch": 1389, "lr": 5.7119810906537875e-05}
{"train_loss": 0.14434093236923218, "global_step": 123652, "epoch": 1389, "lr": 5.711923707712393e-05}
{"train_loss": 0.13823005557060242, "global_step": 123653, "epoch": 1389, "lr": 5.711866324675289e-05}
{"train_loss": 0.2458914816379547, "global_step": 123654, "epoch": 1389, "lr": 5.7118089415424844e-05}
{"train_loss": 0.12489797919988632, "global_step": 123655, "epoch": 1389, "lr": 5.7117515583139844e-05}
{"train_loss": 0.13257499039173126, "global_step": 123656, "epoch": 1389, "lr": 5.711694174989799e-05}
{"train_loss": 0.1867865025997162, "global_step": 123657, "epoch": 1389, "lr": 5.711636791569935e-05}
{"train_loss": 0.16722415387630463, "global_step": 123658, "epoch": 1389, "lr": 5.711579408054401e-05}
{"train_loss": 0.11296073347330093, "global_step": 123659, "epoch": 1389, "lr": 5.7115220244432024e-05}
{"train_loss": 0.2054668515920639, "global_step": 123660, "epoch": 1389, "lr": 5.7114646407363484e-05}
{"train_loss": 0.22332213819026947, "global_step": 123661, "epoch": 1389, "lr": 5.7114072569338473e-05}
{"train_loss": 0.08110412210226059, "global_step": 123662, "epoch": 1389, "lr": 5.711349873035705e-05}
{"train_loss": 0.28971171379089355, "global_step": 123663, "epoch": 1389, "lr": 5.711292489041933e-05}
{"train_loss": 0.18126720190048218, "global_step": 123664, "epoch": 1389, "lr": 5.7112351049525323e-05}
{"train_loss": 0.12140868604183197, "global_step": 123665, "epoch": 1389, "lr": 5.7111777207675173e-05}
{"train_loss": 0.18136607110500336, "global_step": 123666, "epoch": 1389, "lr": 5.711120336486894e-05}
{"train_loss": 0.08921880275011063, "global_step": 123667, "epoch": 1389, "lr": 5.711062952110667e-05}
{"train_loss": 0.1027647852897644, "global_step": 123668, "epoch": 1389, "lr": 5.7110055676388474e-05}
{"train_loss": 0.1187380701303482, "global_step": 123669, "epoch": 1389, "lr": 5.710948183071442e-05}
{"train_loss": 0.15602411329746246, "global_step": 123670, "epoch": 1389, "lr": 5.710890798408458e-05}
{"train_loss": 0.13118603825569153, "global_step": 123671, "epoch": 1389, "lr": 5.710833413649903e-05}
{"train_loss": 0.12717686593532562, "global_step": 123672, "epoch": 1389, "lr": 5.7107760287957856e-05}
{"train_loss": 0.08804494887590408, "global_step": 123673, "epoch": 1389, "lr": 5.710718643846113e-05}
{"train_loss": 0.13723216950893402, "global_step": 123674, "epoch": 1389, "lr": 5.7106612588008936e-05}
{"train_loss": 0.12378581613302231, "global_step": 123675, "epoch": 1389, "lr": 5.710603873660133e-05}
{"train_loss": 0.12346228212118149, "global_step": 123676, "epoch": 1389, "lr": 5.7105464884238414e-05}
{"train_loss": 0.14388994872570038, "global_step": 123677, "epoch": 1389, "lr": 5.710489103092025e-05}
{"train_loss": 0.06520203500986099, "global_step": 123678, "epoch": 1389, "lr": 5.710431717664691e-05}
{"train_loss": 0.20129477977752686, "global_step": 123679, "epoch": 1389, "lr": 5.71037433214185e-05}
{"train_loss": 0.14087991416454315, "global_step": 123680, "epoch": 1389, "lr": 5.7103169465235077e-05}
{"train_loss": 0.1568327248096466, "global_step": 123681, "epoch": 1389, "lr": 5.710259560809671e-05}
{"train_loss": 0.16093294322490692, "global_step": 123682, "epoch": 1389, "lr": 5.71020217500035e-05}
{"train_loss": 0.04629397392272949, "global_step": 123683, "epoch": 1389, "lr": 5.7101447890955485e-05}
{"train_loss": 0.12683679163455963, "global_step": 123684, "epoch": 1389, "lr": 5.7100874030952776e-05}
{"train_loss": 0.17960312962532043, "global_step": 123685, "epoch": 1389, "lr": 5.7100300169995446e-05}
{"train_loss": 0.213380828499794, "global_step": 123686, "epoch": 1389, "lr": 5.7099726308083565e-05}
{"train_loss": 0.0928942933678627, "global_step": 123687, "epoch": 1389, "lr": 5.709915244521722e-05}
{"train_loss": 0.1459522843360901, "global_step": 123688, "epoch": 1389, "lr": 5.709857858139648e-05}
{"train_loss": 0.10070977360010147, "global_step": 123689, "epoch": 1389, "lr": 5.709800471662141e-05}
{"train_loss": 0.1464058756828308, "global_step": 123690, "epoch": 1389, "lr": 5.7097430850892106e-05}
{"train_loss": 0.11711017042398453, "global_step": 123691, "epoch": 1389, "lr": 5.7096856984208644e-05}
{"train_loss": 0.16311900317668915, "global_step": 123692, "epoch": 1389, "lr": 5.709628311657109e-05}
{"train_loss": 0.12968716025352478, "global_step": 123693, "epoch": 1389, "lr": 5.7095709247979535e-05}
{"train_loss": 0.07478110492229462, "global_step": 123694, "epoch": 1389, "lr": 5.709513537843404e-05}
{"train_loss": 0.08920341730117798, "global_step": 123695, "epoch": 1389, "lr": 5.7094561507934685e-05}
{"train_loss": 0.1185581386089325, "global_step": 123696, "epoch": 1389, "lr": 5.7093987636481573e-05}
{"train_loss": 0.20975688099861145, "global_step": 123697, "epoch": 1389, "lr": 5.7093413764074744e-05}
{"train_loss": 0.13049238920211792, "global_step": 123698, "epoch": 1389, "lr": 5.709283989071429e-05}
{"train_loss": 0.16172067821025848, "global_step": 123699, "epoch": 1389, "lr": 5.7092266016400295e-05}
{"train_loss": 0.18951858580112457, "global_step": 123700, "epoch": 1389, "lr": 5.709169214113284e-05}
{"train_loss": 0.1355620175600052, "global_step": 123701, "epoch": 1389, "lr": 5.709111826491198e-05}
{"train_loss": 0.1080063134431839, "global_step": 123702, "epoch": 1389, "lr": 5.709054438773782e-05}
{"train_loss": 0.2047366499900818, "global_step": 123703, "epoch": 1389, "lr": 5.7089970509610415e-05}
{"train_loss": 0.17363302409648895, "global_step": 123704, "epoch": 1389, "lr": 5.708939663052985e-05}
{"train_loss": 0.14876237511634827, "global_step": 123705, "epoch": 1389, "lr": 5.70888227504962e-05}
{"train_loss": 0.11122389137744904, "global_step": 123706, "epoch": 1389, "lr": 5.708824886950954e-05}
{"train_loss": 0.09903863817453384, "global_step": 123707, "epoch": 1389, "lr": 5.708767498756996e-05}
{"train_loss": 0.13622614741325378, "global_step": 123708, "epoch": 1389, "lr": 5.708710110467752e-05}
{"train_loss": 0.13576345154073802, "global_step": 123709, "epoch": 1389, "lr": 5.708652722083232e-05, "val_loss": 5.3891167640686035}
{"train_loss": 0.20459964871406555, "global_step": 123710, "epoch": 1390, "lr": 5.7085953336034416e-05}
{"train_loss": 0.1348707675933838, "global_step": 123711, "epoch": 1390, "lr": 5.7085379450283884e-05}
{"train_loss": 0.15829865634441376, "global_step": 123712, "epoch": 1390, "lr": 5.708480556358081e-05}
{"train_loss": 0.17593073844909668, "global_step": 123713, "epoch": 1390, "lr": 5.708423167592528e-05}
{"train_loss": 0.1527310609817505, "global_step": 123714, "epoch": 1390, "lr": 5.7083657787317354e-05}
{"train_loss": 0.12424612045288086, "global_step": 123715, "epoch": 1390, "lr": 5.7083083897757125e-05}
{"train_loss": 0.10573995858430862, "global_step": 123716, "epoch": 1390, "lr": 5.708251000724465e-05}
{"train_loss": 0.17186255753040314, "global_step": 123717, "epoch": 1390, "lr": 5.708193611578003e-05}
{"train_loss": 0.09697411954402924, "global_step": 123718, "epoch": 1390, "lr": 5.708136222336332e-05}
{"train_loss": 0.14112718403339386, "global_step": 123719, "epoch": 1390, "lr": 5.7080788329994615e-05}
{"train_loss": 0.1646261364221573, "global_step": 123720, "epoch": 1390, "lr": 5.708021443567397e-05}
{"train_loss": 0.16838674247264862, "global_step": 123721, "epoch": 1390, "lr": 5.707964054040149e-05}
{"train_loss": 0.09195122122764587, "global_step": 123722, "epoch": 1390, "lr": 5.707906664417724e-05}
{"train_loss": 0.0839439183473587, "global_step": 123723, "epoch": 1390, "lr": 5.707849274700129e-05}
{"train_loss": 0.2171342968940735, "global_step": 123724, "epoch": 1390, "lr": 5.7077918848873726e-05}
{"train_loss": 0.1398812234401703, "global_step": 123725, "epoch": 1390, "lr": 5.707734494979462e-05}
{"train_loss": 0.16192996501922607, "global_step": 123726, "epoch": 1390, "lr": 5.7076771049764056e-05}
{"train_loss": 0.14206063747406006, "global_step": 123727, "epoch": 1390, "lr": 5.7076197148782095e-05}
{"train_loss": 0.13147245347499847, "global_step": 123728, "epoch": 1390, "lr": 5.707562324684883e-05}
{"train_loss": 0.11565414071083069, "global_step": 123729, "epoch": 1390, "lr": 5.707504934396434e-05}
{"train_loss": 0.10830807685852051, "global_step": 123730, "epoch": 1390, "lr": 5.7074475440128696e-05}
{"train_loss": 0.10737273097038269, "global_step": 123731, "epoch": 1390, "lr": 5.7073901535341965e-05}
{"train_loss": 0.13690924644470215, "global_step": 123732, "epoch": 1390, "lr": 5.707332762960425e-05}
{"train_loss": 0.1276315301656723, "global_step": 123733, "epoch": 1390, "lr": 5.7072753722915595e-05}
{"train_loss": 0.07442890107631683, "global_step": 123734, "epoch": 1390, "lr": 5.7072179815276105e-05}
{"train_loss": 0.1830291450023651, "global_step": 123735, "epoch": 1390, "lr": 5.7071605906685844e-05}
{"train_loss": 0.1431654989719391, "global_step": 123736, "epoch": 1390, "lr": 5.7071031997144886e-05}
{"train_loss": 0.12667928636074066, "global_step": 123737, "epoch": 1390, "lr": 5.707045808665332e-05}
{"train_loss": 0.13153937458992004, "global_step": 123738, "epoch": 1390, "lr": 5.706988417521122e-05}
{"train_loss": 0.09812520444393158, "global_step": 123739, "epoch": 1390, "lr": 5.706931026281866e-05}
{"train_loss": 0.13431276381015778, "global_step": 123740, "epoch": 1390, "lr": 5.706873634947571e-05}
{"train_loss": 0.11461155861616135, "global_step": 123741, "epoch": 1390, "lr": 5.706816243518245e-05}
{"train_loss": 0.14842742681503296, "global_step": 123742, "epoch": 1390, "lr": 5.7067588519938973e-05}
{"train_loss": 0.09513051807880402, "global_step": 123743, "epoch": 1390, "lr": 5.706701460374534e-05}
{"train_loss": 0.1522453874349594, "global_step": 123744, "epoch": 1390, "lr": 5.706644068660164e-05}
{"train_loss": 0.10580474883317947, "global_step": 123745, "epoch": 1390, "lr": 5.706586676850794e-05}
{"train_loss": 0.11156389117240906, "global_step": 123746, "epoch": 1390, "lr": 5.7065292849464313e-05}
{"train_loss": 0.15201394259929657, "global_step": 123747, "epoch": 1390, "lr": 5.706471892947085e-05}
{"train_loss": 0.16022643446922302, "global_step": 123748, "epoch": 1390, "lr": 5.706414500852763e-05}
{"train_loss": 0.1894528865814209, "global_step": 123749, "epoch": 1390, "lr": 5.706357108663471e-05}
{"train_loss": 0.10474489629268646, "global_step": 123750, "epoch": 1390, "lr": 5.7062997163792175e-05}
{"train_loss": 0.18323318660259247, "global_step": 123751, "epoch": 1390, "lr": 5.7062423240000105e-05}
{"train_loss": 0.1153460219502449, "global_step": 123752, "epoch": 1390, "lr": 5.70618493152586e-05}
{"train_loss": 0.14425623416900635, "global_step": 123753, "epoch": 1390, "lr": 5.706127538956769e-05}
{"train_loss": 0.08022566884756088, "global_step": 123754, "epoch": 1390, "lr": 5.706070146292749e-05}
{"train_loss": 0.0760190412402153, "global_step": 123755, "epoch": 1390, "lr": 5.706012753533807e-05}
{"train_loss": 0.13201242685317993, "global_step": 123756, "epoch": 1390, "lr": 5.7059553606799485e-05}
{"train_loss": 0.12738005816936493, "global_step": 123757, "epoch": 1390, "lr": 5.7058979677311844e-05}
{"train_loss": 0.17868638038635254, "global_step": 123758, "epoch": 1390, "lr": 5.705840574687521e-05}
{"train_loss": 0.11677781492471695, "global_step": 123759, "epoch": 1390, "lr": 5.705783181548965e-05}
{"train_loss": 0.08931318670511246, "global_step": 123760, "epoch": 1390, "lr": 5.705725788315526e-05}
{"train_loss": 0.14580702781677246, "global_step": 123761, "epoch": 1390, "lr": 5.705668394987209e-05}
{"train_loss": 0.05962279066443443, "global_step": 123762, "epoch": 1390, "lr": 5.705611001564025e-05}
{"train_loss": 0.13310158252716064, "global_step": 123763, "epoch": 1390, "lr": 5.70555360804598e-05}
{"train_loss": 0.12048235535621643, "global_step": 123764, "epoch": 1390, "lr": 5.705496214433083e-05}
{"train_loss": 0.09078680723905563, "global_step": 123765, "epoch": 1390, "lr": 5.7054388207253385e-05}
{"train_loss": 0.11924007534980774, "global_step": 123766, "epoch": 1390, "lr": 5.7053814269227577e-05}
{"train_loss": 0.18066567182540894, "global_step": 123767, "epoch": 1390, "lr": 5.7053240330253464e-05}
{"train_loss": 0.1393277645111084, "global_step": 123768, "epoch": 1390, "lr": 5.7052666390331136e-05}
{"train_loss": 0.10480628162622452, "global_step": 123769, "epoch": 1390, "lr": 5.7052092449460646e-05}
{"train_loss": 0.09814120084047318, "global_step": 123770, "epoch": 1390, "lr": 5.705151850764211e-05}
{"train_loss": 0.06897908449172974, "global_step": 123771, "epoch": 1390, "lr": 5.7050944564875566e-05}
{"train_loss": 0.12303218990564346, "global_step": 123772, "epoch": 1390, "lr": 5.705037062116112e-05}
{"train_loss": 0.17500701546669006, "global_step": 123773, "epoch": 1390, "lr": 5.704979667649882e-05}
{"train_loss": 0.12639547884464264, "global_step": 123774, "epoch": 1390, "lr": 5.704922273088878e-05}
{"train_loss": 0.11414764821529388, "global_step": 123775, "epoch": 1390, "lr": 5.704864878433106e-05}
{"train_loss": 0.14950543642044067, "global_step": 123776, "epoch": 1390, "lr": 5.704807483682573e-05}
{"train_loss": 0.18041090667247772, "global_step": 123777, "epoch": 1390, "lr": 5.7047500888372863e-05}
{"train_loss": 0.15674901008605957, "global_step": 123778, "epoch": 1390, "lr": 5.7046926938972555e-05}
{"train_loss": 0.1992742121219635, "global_step": 123779, "epoch": 1390, "lr": 5.7046352988624864e-05}
{"train_loss": 0.1955481618642807, "global_step": 123780, "epoch": 1390, "lr": 5.7045779037329885e-05}
{"train_loss": 0.1869518905878067, "global_step": 123781, "epoch": 1390, "lr": 5.7045205085087686e-05}
{"train_loss": 0.11248172074556351, "global_step": 123782, "epoch": 1390, "lr": 5.704463113189834e-05}
{"train_loss": 0.15105444192886353, "global_step": 123783, "epoch": 1390, "lr": 5.704405717776193e-05}
{"train_loss": 0.12758494913578033, "global_step": 123784, "epoch": 1390, "lr": 5.704348322267854e-05}
{"train_loss": 0.10530312359333038, "global_step": 123785, "epoch": 1390, "lr": 5.7042909266648235e-05}
{"train_loss": 0.1269090622663498, "global_step": 123786, "epoch": 1390, "lr": 5.704233530967109e-05}
{"train_loss": 0.14946281909942627, "global_step": 123787, "epoch": 1390, "lr": 5.70417613517472e-05}
{"train_loss": 0.1495715230703354, "global_step": 123788, "epoch": 1390, "lr": 5.7041187392876625e-05}
{"train_loss": 0.16786310076713562, "global_step": 123789, "epoch": 1390, "lr": 5.704061343305945e-05}
{"train_loss": 0.060136038810014725, "global_step": 123790, "epoch": 1390, "lr": 5.704003947229575e-05}
{"train_loss": 0.12625114619731903, "global_step": 123791, "epoch": 1390, "lr": 5.7039465510585585e-05}
{"train_loss": 0.1191575899720192, "global_step": 123792, "epoch": 1390, "lr": 5.7038891547929075e-05}
{"train_loss": 0.10811690986156464, "global_step": 123793, "epoch": 1390, "lr": 5.7038317584326274e-05}
{"train_loss": 0.10253588110208511, "global_step": 123794, "epoch": 1390, "lr": 5.703774361977725e-05}
{"train_loss": 0.144722580909729, "global_step": 123795, "epoch": 1390, "lr": 5.703716965428209e-05}
{"train_loss": 0.110440194606781, "global_step": 123796, "epoch": 1390, "lr": 5.703659568784085e-05}
{"train_loss": 0.15715479850769043, "global_step": 123797, "epoch": 1390, "lr": 5.7036021720453636e-05}
{"train_loss": 0.13352138699775332, "global_step": 123798, "epoch": 1390, "lr": 5.703544775212052e-05, "val_loss": 5.390257358551025, "train_action_mse_error": 22.56914520263672}
{"train_loss": 0.135844886302948, "global_step": 123799, "epoch": 1391, "lr": 5.703487378284157e-05}
{"train_loss": 0.18103551864624023, "global_step": 123800, "epoch": 1391, "lr": 5.703429981261687e-05}
{"train_loss": 0.18634247779846191, "global_step": 123801, "epoch": 1391, "lr": 5.70337258414465e-05}
{"train_loss": 0.08166208118200302, "global_step": 123802, "epoch": 1391, "lr": 5.7033151869330514e-05}
{"train_loss": 0.2162967026233673, "global_step": 123803, "epoch": 1391, "lr": 5.703257789626901e-05}
{"train_loss": 0.14952261745929718, "global_step": 123804, "epoch": 1391, "lr": 5.703200392226207e-05}
{"train_loss": 0.15333342552185059, "global_step": 123805, "epoch": 1391, "lr": 5.703142994730975e-05}
{"train_loss": 0.14317938685417175, "global_step": 123806, "epoch": 1391, "lr": 5.703085597141215e-05}
{"train_loss": 0.11466524749994278, "global_step": 123807, "epoch": 1391, "lr": 5.7030281994569346e-05}
{"train_loss": 0.14398862421512604, "global_step": 123808, "epoch": 1391, "lr": 5.7029708016781406e-05}
{"train_loss": 0.15226660668849945, "global_step": 123809, "epoch": 1391, "lr": 5.7029134038048394e-05}
{"train_loss": 0.13757416605949402, "global_step": 123810, "epoch": 1391, "lr": 5.7028560058370384e-05}
{"train_loss": 0.13472570478916168, "global_step": 123811, "epoch": 1391, "lr": 5.70279860777475e-05}
{"train_loss": 0.08742399513721466, "global_step": 123812, "epoch": 1391, "lr": 5.7027412096179787e-05}
{"train_loss": 0.13754066824913025, "global_step": 123813, "epoch": 1391, "lr": 5.7026838113667313e-05}
{"train_loss": 0.10052476823329926, "global_step": 123814, "epoch": 1391, "lr": 5.7026264130210175e-05}
{"train_loss": 0.11669663339853287, "global_step": 123815, "epoch": 1391, "lr": 5.702569014580843e-05}
{"train_loss": 0.16394826769828796, "global_step": 123816, "epoch": 1391, "lr": 5.702511616046218e-05}
{"train_loss": 0.12344402819871902, "global_step": 123817, "epoch": 1391, "lr": 5.702454217417148e-05}
{"train_loss": 0.06672652065753937, "global_step": 123818, "epoch": 1391, "lr": 5.702396818693642e-05}
{"train_loss": 0.08892896771430969, "global_step": 123819, "epoch": 1391, "lr": 5.7023394198757064e-05}
{"train_loss": 0.15669049322605133, "global_step": 123820, "epoch": 1391, "lr": 5.702282020963351e-05}
{"train_loss": 0.1238967776298523, "global_step": 123821, "epoch": 1391, "lr": 5.702224621956581e-05}
{"train_loss": 0.11266997456550598, "global_step": 123822, "epoch": 1391, "lr": 5.702167222855407e-05}
{"train_loss": 0.10985352098941803, "global_step": 123823, "epoch": 1391, "lr": 5.7021098236598335e-05}
{"train_loss": 0.09724561125040054, "global_step": 123824, "epoch": 1391, "lr": 5.702052424369872e-05}
{"train_loss": 0.13555677235126495, "global_step": 123825, "epoch": 1391, "lr": 5.701995024985527e-05}
{"train_loss": 0.10166580975055695, "global_step": 123826, "epoch": 1391, "lr": 5.701937625506807e-05}
{"train_loss": 0.12368014454841614, "global_step": 123827, "epoch": 1391, "lr": 5.70188022593372e-05}
{"train_loss": 0.12349265813827515, "global_step": 123828, "epoch": 1391, "lr": 5.7018228262662734e-05}
{"train_loss": 0.13584843277931213, "global_step": 123829, "epoch": 1391, "lr": 5.701765426504475e-05}
{"train_loss": 0.12441364675760269, "global_step": 123830, "epoch": 1391, "lr": 5.701708026648335e-05}
{"train_loss": 0.12430565059185028, "global_step": 123831, "epoch": 1391, "lr": 5.701650626697857e-05}
{"train_loss": 0.12297359853982925, "global_step": 123832, "epoch": 1391, "lr": 5.701593226653052e-05}
{"train_loss": 0.1739349365234375, "global_step": 123833, "epoch": 1391, "lr": 5.7015358265139264e-05}
{"train_loss": 0.10029640793800354, "global_step": 123834, "epoch": 1391, "lr": 5.7014784262804865e-05}
{"train_loss": 0.11147766560316086, "global_step": 123835, "epoch": 1391, "lr": 5.701421025952742e-05}
{"train_loss": 0.12001406401395798, "global_step": 123836, "epoch": 1391, "lr": 5.7013636255307004e-05}
{"train_loss": 0.10839977115392685, "global_step": 123837, "epoch": 1391, "lr": 5.701306225014369e-05}
{"train_loss": 0.10927240550518036, "global_step": 123838, "epoch": 1391, "lr": 5.701248824403754e-05}
{"train_loss": 0.15149380266666412, "global_step": 123839, "epoch": 1391, "lr": 5.701191423698866e-05}
{"train_loss": 0.1529831737279892, "global_step": 123840, "epoch": 1391, "lr": 5.701134022899711e-05}
{"train_loss": 0.14101050794124603, "global_step": 123841, "epoch": 1391, "lr": 5.7010766220062975e-05}
{"train_loss": 0.15590718388557434, "global_step": 123842, "epoch": 1391, "lr": 5.7010192210186316e-05}
{"train_loss": 0.08057428151369095, "global_step": 123843, "epoch": 1391, "lr": 5.700961819936724e-05}
{"train_loss": 0.14401257038116455, "global_step": 123844, "epoch": 1391, "lr": 5.70090441876058e-05}
{"train_loss": 0.08297598361968994, "global_step": 123845, "epoch": 1391, "lr": 5.700847017490207e-05}
{"train_loss": 0.08634571731090546, "global_step": 123846, "epoch": 1391, "lr": 5.7007896161256136e-05}
{"train_loss": 0.21927903592586517, "global_step": 123847, "epoch": 1391, "lr": 5.7007322146668084e-05}
{"train_loss": 0.17224687337875366, "global_step": 123848, "epoch": 1391, "lr": 5.7006748131137975e-05}
{"train_loss": 0.145528644323349, "global_step": 123849, "epoch": 1391, "lr": 5.70061741146659e-05}
{"train_loss": 0.1399817317724228, "global_step": 123850, "epoch": 1391, "lr": 5.7005600097251934e-05}
{"train_loss": 0.07556728273630142, "global_step": 123851, "epoch": 1391, "lr": 5.7005026078896164e-05}
{"train_loss": 0.12039695680141449, "global_step": 123852, "epoch": 1391, "lr": 5.7004452059598636e-05}
{"train_loss": 0.14113086462020874, "global_step": 123853, "epoch": 1391, "lr": 5.700387803935945e-05}
{"train_loss": 0.13186272978782654, "global_step": 123854, "epoch": 1391, "lr": 5.700330401817867e-05}
{"train_loss": 0.10138452053070068, "global_step": 123855, "epoch": 1391, "lr": 5.700272999605638e-05}
{"train_loss": 0.18401086330413818, "global_step": 123856, "epoch": 1391, "lr": 5.700215597299267e-05}
{"train_loss": 0.11019518971443176, "global_step": 123857, "epoch": 1391, "lr": 5.7001581948987594e-05}
{"train_loss": 0.154548779129982, "global_step": 123858, "epoch": 1391, "lr": 5.700100792404125e-05}
{"train_loss": 0.15653672814369202, "global_step": 123859, "epoch": 1391, "lr": 5.700043389815369e-05}
{"train_loss": 0.0654396191239357, "global_step": 123860, "epoch": 1391, "lr": 5.699985987132502e-05}
{"train_loss": 0.15664757788181305, "global_step": 123861, "epoch": 1391, "lr": 5.699928584355531e-05}
{"train_loss": 0.14878013730049133, "global_step": 123862, "epoch": 1391, "lr": 5.699871181484462e-05}
{"train_loss": 0.11854513734579086, "global_step": 123863, "epoch": 1391, "lr": 5.699813778519304e-05}
{"train_loss": 0.1824755072593689, "global_step": 123864, "epoch": 1391, "lr": 5.6997563754600636e-05}
{"train_loss": 0.17744578421115875, "global_step": 123865, "epoch": 1391, "lr": 5.699698972306751e-05}
{"train_loss": 0.16815702617168427, "global_step": 123866, "epoch": 1391, "lr": 5.699641569059372e-05}
{"train_loss": 0.1351819783449173, "global_step": 123867, "epoch": 1391, "lr": 5.699584165717935e-05}
{"train_loss": 0.10988866537809372, "global_step": 123868, "epoch": 1391, "lr": 5.699526762282448e-05}
{"train_loss": 0.1279355138540268, "global_step": 123869, "epoch": 1391, "lr": 5.699469358752917e-05}
{"train_loss": 0.09090154618024826, "global_step": 123870, "epoch": 1391, "lr": 5.6994119551293524e-05}
{"train_loss": 0.10739488899707794, "global_step": 123871, "epoch": 1391, "lr": 5.699354551411758e-05}
{"train_loss": 0.14098243415355682, "global_step": 123872, "epoch": 1391, "lr": 5.699297147600147e-05}
{"train_loss": 0.1609974354505539, "global_step": 123873, "epoch": 1391, "lr": 5.699239743694521e-05}
{"train_loss": 0.04617660492658615, "global_step": 123874, "epoch": 1391, "lr": 5.699182339694892e-05}
{"train_loss": 0.22659344971179962, "global_step": 123875, "epoch": 1391, "lr": 5.699124935601267e-05}
{"train_loss": 0.24617603421211243, "global_step": 123876, "epoch": 1391, "lr": 5.699067531413652e-05}
{"train_loss": 0.11024811863899231, "global_step": 123877, "epoch": 1391, "lr": 5.699010127132056e-05}
{"train_loss": 0.1802120804786682, "global_step": 123878, "epoch": 1391, "lr": 5.6989527227564875e-05}
{"train_loss": 0.29273536801338196, "global_step": 123879, "epoch": 1391, "lr": 5.698895318286953e-05}
{"train_loss": 0.15230906009674072, "global_step": 123880, "epoch": 1391, "lr": 5.6988379137234605e-05}
{"train_loss": 0.11935143917798996, "global_step": 123881, "epoch": 1391, "lr": 5.6987805090660176e-05}
{"train_loss": 0.15083087980747223, "global_step": 123882, "epoch": 1391, "lr": 5.6987231043146314e-05}
{"train_loss": 0.13703900575637817, "global_step": 123883, "epoch": 1391, "lr": 5.698665699469311e-05}
{"train_loss": 0.11044252663850784, "global_step": 123884, "epoch": 1391, "lr": 5.698608294530063e-05}
{"train_loss": 0.1219579353928566, "global_step": 123885, "epoch": 1391, "lr": 5.698550889496896e-05}
{"train_loss": 0.2669370174407959, "global_step": 123886, "epoch": 1391, "lr": 5.698493484369819e-05}
{"train_loss": 0.13762695199987862, "global_step": 123887, "epoch": 1391, "lr": 5.698436079148837e-05, "val_loss": 5.443454265594482}
{"train_loss": 0.11648000776767731, "global_step": 123888, "epoch": 1392, "lr": 5.698378673833958e-05}
{"train_loss": 0.16666989028453827, "global_step": 123889, "epoch": 1392, "lr": 5.698321268425191e-05}
{"train_loss": 0.1560722142457962, "global_step": 123890, "epoch": 1392, "lr": 5.698263862922545e-05}
{"train_loss": 0.12921220064163208, "global_step": 123891, "epoch": 1392, "lr": 5.698206457326024e-05}
{"train_loss": 0.12021175026893616, "global_step": 123892, "epoch": 1392, "lr": 5.6981490516356376e-05}
{"train_loss": 0.23371131718158722, "global_step": 123893, "epoch": 1392, "lr": 5.6980916458513936e-05}
{"train_loss": 0.13047455251216888, "global_step": 123894, "epoch": 1392, "lr": 5.6980342399733e-05}
{"train_loss": 0.18040627241134644, "global_step": 123895, "epoch": 1392, "lr": 5.697976834001364e-05}
{"train_loss": 0.10556476563215256, "global_step": 123896, "epoch": 1392, "lr": 5.697919427935594e-05}
{"train_loss": 0.11213786154985428, "global_step": 123897, "epoch": 1392, "lr": 5.697862021775997e-05}
{"train_loss": 0.12079453468322754, "global_step": 123898, "epoch": 1392, "lr": 5.697804615522581e-05}
{"train_loss": 0.15851199626922607, "global_step": 123899, "epoch": 1392, "lr": 5.6977472091753534e-05}
{"train_loss": 0.1546027809381485, "global_step": 123900, "epoch": 1392, "lr": 5.697689802734321e-05}
{"train_loss": 0.14138026535511017, "global_step": 123901, "epoch": 1392, "lr": 5.697632396199494e-05}
{"train_loss": 0.13602440059185028, "global_step": 123902, "epoch": 1392, "lr": 5.69757498957088e-05}
{"train_loss": 0.10038209706544876, "global_step": 123903, "epoch": 1392, "lr": 5.697517582848484e-05}
{"train_loss": 0.136821448802948, "global_step": 123904, "epoch": 1392, "lr": 5.697460176032315e-05}
{"train_loss": 0.13424476981163025, "global_step": 123905, "epoch": 1392, "lr": 5.697402769122382e-05}
{"train_loss": 0.16165371239185333, "global_step": 123906, "epoch": 1392, "lr": 5.697345362118691e-05}
{"train_loss": 0.19468051195144653, "global_step": 123907, "epoch": 1392, "lr": 5.697287955021251e-05}
{"train_loss": 0.11394619941711426, "global_step": 123908, "epoch": 1392, "lr": 5.697230547830069e-05}
{"train_loss": 0.17992474138736725, "global_step": 123909, "epoch": 1392, "lr": 5.697173140545153e-05}
{"train_loss": 0.19812752306461334, "global_step": 123910, "epoch": 1392, "lr": 5.697115733166511e-05}
{"train_loss": 0.17132475972175598, "global_step": 123911, "epoch": 1392, "lr": 5.697058325694149e-05}
{"train_loss": 0.13983076810836792, "global_step": 123912, "epoch": 1392, "lr": 5.697000918128077e-05}
{"train_loss": 0.16349494457244873, "global_step": 123913, "epoch": 1392, "lr": 5.696943510468301e-05}
{"train_loss": 0.18825247883796692, "global_step": 123914, "epoch": 1392, "lr": 5.69688610271483e-05}
{"train_loss": 0.14336268603801727, "global_step": 123915, "epoch": 1392, "lr": 5.696828694867671e-05}
{"train_loss": 0.1244114339351654, "global_step": 123916, "epoch": 1392, "lr": 5.696771286926832e-05}
{"train_loss": 0.11586734652519226, "global_step": 123917, "epoch": 1392, "lr": 5.69671387889232e-05}
{"train_loss": 0.15529756247997284, "global_step": 123918, "epoch": 1392, "lr": 5.696656470764144e-05}
{"train_loss": 0.14164990186691284, "global_step": 123919, "epoch": 1392, "lr": 5.6965990625423096e-05}
{"train_loss": 0.11514008045196533, "global_step": 123920, "epoch": 1392, "lr": 5.696541654226827e-05}
{"train_loss": 0.11277066171169281, "global_step": 123921, "epoch": 1392, "lr": 5.696484245817704e-05}
{"train_loss": 0.09074600785970688, "global_step": 123922, "epoch": 1392, "lr": 5.6964268373149454e-05}
{"train_loss": 0.10261472314596176, "global_step": 123923, "epoch": 1392, "lr": 5.696369428718561e-05}
{"train_loss": 0.15772688388824463, "global_step": 123924, "epoch": 1392, "lr": 5.6963120200285594e-05}
{"train_loss": 0.08237794041633606, "global_step": 123925, "epoch": 1392, "lr": 5.696254611244945e-05}
{"train_loss": 0.1337675005197525, "global_step": 123926, "epoch": 1392, "lr": 5.69619720236773e-05}
{"train_loss": 0.07622890174388885, "global_step": 123927, "epoch": 1392, "lr": 5.696139793396918e-05}
{"train_loss": 0.15798932313919067, "global_step": 123928, "epoch": 1392, "lr": 5.696082384332521e-05}
{"train_loss": 0.18835994601249695, "global_step": 123929, "epoch": 1392, "lr": 5.696024975174543e-05}
{"train_loss": 0.17795392870903015, "global_step": 123930, "epoch": 1392, "lr": 5.6959675659229915e-05}
{"train_loss": 0.21337014436721802, "global_step": 123931, "epoch": 1392, "lr": 5.695910156577877e-05}
{"train_loss": 0.08767323195934296, "global_step": 123932, "epoch": 1392, "lr": 5.695852747139205e-05}
{"train_loss": 0.12970638275146484, "global_step": 123933, "epoch": 1392, "lr": 5.695795337606985e-05}
{"train_loss": 0.10135827958583832, "global_step": 123934, "epoch": 1392, "lr": 5.6957379279812225e-05}
{"train_loss": 0.10976747423410416, "global_step": 123935, "epoch": 1392, "lr": 5.695680518261928e-05}
{"train_loss": 0.1732136607170105, "global_step": 123936, "epoch": 1392, "lr": 5.6956231084491076e-05}
{"train_loss": 0.11151362955570221, "global_step": 123937, "epoch": 1392, "lr": 5.6955656985427676e-05}
{"train_loss": 0.20302878320217133, "global_step": 123938, "epoch": 1392, "lr": 5.695508288542918e-05}
{"train_loss": 0.14793045818805695, "global_step": 123939, "epoch": 1392, "lr": 5.6954508784495665e-05}
{"train_loss": 0.1194583922624588, "global_step": 123940, "epoch": 1392, "lr": 5.695393468262718e-05}
{"train_loss": 0.09140091389417648, "global_step": 123941, "epoch": 1392, "lr": 5.6953360579823854e-05}
{"train_loss": 0.10009735822677612, "global_step": 123942, "epoch": 1392, "lr": 5.695278647608572e-05}
{"train_loss": 0.12479425221681595, "global_step": 123943, "epoch": 1392, "lr": 5.695221237141286e-05}
{"train_loss": 0.051297109574079514, "global_step": 123944, "epoch": 1392, "lr": 5.695163826580537e-05}
{"train_loss": 0.15316671133041382, "global_step": 123945, "epoch": 1392, "lr": 5.6951064159263315e-05}
{"train_loss": 0.07765619456768036, "global_step": 123946, "epoch": 1392, "lr": 5.695049005178678e-05}
{"train_loss": 0.14037226140499115, "global_step": 123947, "epoch": 1392, "lr": 5.694991594337582e-05}
{"train_loss": 0.1968873143196106, "global_step": 123948, "epoch": 1392, "lr": 5.694934183403056e-05}
{"train_loss": 0.1481996327638626, "global_step": 123949, "epoch": 1392, "lr": 5.694876772375102e-05}
{"train_loss": 0.15379849076271057, "global_step": 123950, "epoch": 1392, "lr": 5.69481936125373e-05}
{"train_loss": 0.15913011133670807, "global_step": 123951, "epoch": 1392, "lr": 5.694761950038949e-05}
{"train_loss": 0.15441583096981049, "global_step": 123952, "epoch": 1392, "lr": 5.6947045387307664e-05}
{"train_loss": 0.18771296739578247, "global_step": 123953, "epoch": 1392, "lr": 5.694647127329188e-05}
{"train_loss": 0.06572599709033966, "global_step": 123954, "epoch": 1392, "lr": 5.694589715834223e-05}
{"train_loss": 0.13376548886299133, "global_step": 123955, "epoch": 1392, "lr": 5.6945323042458775e-05}
{"train_loss": 0.13727843761444092, "global_step": 123956, "epoch": 1392, "lr": 5.6944748925641625e-05}
{"train_loss": 0.15125349164009094, "global_step": 123957, "epoch": 1392, "lr": 5.694417480789083e-05}
{"train_loss": 0.12261107563972473, "global_step": 123958, "epoch": 1392, "lr": 5.694360068920648e-05}
{"train_loss": 0.16216079890727997, "global_step": 123959, "epoch": 1392, "lr": 5.6943026569588645e-05}
{"train_loss": 0.17586436867713928, "global_step": 123960, "epoch": 1392, "lr": 5.694245244903741e-05}
{"train_loss": 0.21521513164043427, "global_step": 123961, "epoch": 1392, "lr": 5.694187832755284e-05}
{"train_loss": 0.14529913663864136, "global_step": 123962, "epoch": 1392, "lr": 5.6941304205135016e-05}
{"train_loss": 0.24042361974716187, "global_step": 123963, "epoch": 1392, "lr": 5.694073008178403e-05}
{"train_loss": 0.1924155205488205, "global_step": 123964, "epoch": 1392, "lr": 5.694015595749994e-05}
{"train_loss": 0.18281513452529907, "global_step": 123965, "epoch": 1392, "lr": 5.693958183228283e-05}
{"train_loss": 0.10805868357419968, "global_step": 123966, "epoch": 1392, "lr": 5.693900770613279e-05}
{"train_loss": 0.14668487012386322, "global_step": 123967, "epoch": 1392, "lr": 5.693843357904988e-05}
{"train_loss": 0.09280064702033997, "global_step": 123968, "epoch": 1392, "lr": 5.693785945103418e-05}
{"train_loss": 0.10903454571962357, "global_step": 123969, "epoch": 1392, "lr": 5.693728532208576e-05}
{"train_loss": 0.13133856654167175, "global_step": 123970, "epoch": 1392, "lr": 5.6936711192204715e-05}
{"train_loss": 0.2019743025302887, "global_step": 123971, "epoch": 1392, "lr": 5.693613706139112e-05}
{"train_loss": 0.11267205327749252, "global_step": 123972, "epoch": 1392, "lr": 5.693556292964504e-05}
{"train_loss": 0.20987734198570251, "global_step": 123973, "epoch": 1392, "lr": 5.693498879696654e-05}
{"train_loss": 0.16287283599376678, "global_step": 123974, "epoch": 1392, "lr": 5.693441466335573e-05}
{"train_loss": 0.10600452870130539, "global_step": 123975, "epoch": 1392, "lr": 5.6933840528812674e-05}
{"train_loss": 0.14257230128297646, "global_step": 123976, "epoch": 1392, "lr": 5.693326639333745e-05, "val_loss": 5.258948802947998}
{"train_loss": 0.15826815366744995, "global_step": 123977, "epoch": 1393, "lr": 5.693269225693012e-05}
{"train_loss": 0.12034738063812256, "global_step": 123978, "epoch": 1393, "lr": 5.6932118119590785e-05}
{"train_loss": 0.16041408479213715, "global_step": 123979, "epoch": 1393, "lr": 5.693154398131951e-05}
{"train_loss": 0.1259702444076538, "global_step": 123980, "epoch": 1393, "lr": 5.693096984211638e-05}
{"train_loss": 0.09391564130783081, "global_step": 123981, "epoch": 1393, "lr": 5.6930395701981455e-05}
{"train_loss": 0.1018441691994667, "global_step": 123982, "epoch": 1393, "lr": 5.692982156091483e-05}
{"train_loss": 0.10461243987083435, "global_step": 123983, "epoch": 1393, "lr": 5.692924741891656e-05}
{"train_loss": 0.11790682375431061, "global_step": 123984, "epoch": 1393, "lr": 5.692867327598677e-05}
{"train_loss": 0.1484333723783493, "global_step": 123985, "epoch": 1393, "lr": 5.692809913212547e-05}
{"train_loss": 0.10676978528499603, "global_step": 123986, "epoch": 1393, "lr": 5.6927524987332793e-05}
{"train_loss": 0.17212972044944763, "global_step": 123987, "epoch": 1393, "lr": 5.6926950841608794e-05}
{"train_loss": 0.14594458043575287, "global_step": 123988, "epoch": 1393, "lr": 5.692637669495354e-05}
{"train_loss": 0.13064643740653992, "global_step": 123989, "epoch": 1393, "lr": 5.6925802547367126e-05}
{"train_loss": 0.1316995620727539, "global_step": 123990, "epoch": 1393, "lr": 5.692522839884962e-05}
{"train_loss": 0.19426681101322174, "global_step": 123991, "epoch": 1393, "lr": 5.692465424940109e-05}
{"train_loss": 0.13672052323818207, "global_step": 123992, "epoch": 1393, "lr": 5.692408009902164e-05}
{"train_loss": 0.14219637215137482, "global_step": 123993, "epoch": 1393, "lr": 5.692350594771132e-05}
{"train_loss": 0.1409851312637329, "global_step": 123994, "epoch": 1393, "lr": 5.6922931795470225e-05}
{"train_loss": 0.0981978327035904, "global_step": 123995, "epoch": 1393, "lr": 5.692235764229843e-05}
{"train_loss": 0.22669020295143127, "global_step": 123996, "epoch": 1393, "lr": 5.6921783488196e-05}
{"train_loss": 0.11336699873209, "global_step": 123997, "epoch": 1393, "lr": 5.6921209333163026e-05}
{"train_loss": 0.21027609705924988, "global_step": 123998, "epoch": 1393, "lr": 5.692063517719959e-05}
{"train_loss": 0.19081078469753265, "global_step": 123999, "epoch": 1393, "lr": 5.6920061020305735e-05}
{"train_loss": 0.15128661692142487, "global_step": 124000, "epoch": 1393, "lr": 5.6919486862481586e-05}
{"train_loss": 0.12222104519605637, "global_step": 124001, "epoch": 1393, "lr": 5.691891270372718e-05}
{"train_loss": 0.11551889032125473, "global_step": 124002, "epoch": 1393, "lr": 5.691833854404262e-05}
{"train_loss": 0.12567688524723053, "global_step": 124003, "epoch": 1393, "lr": 5.691776438342798e-05}
{"train_loss": 0.15612679719924927, "global_step": 124004, "epoch": 1393, "lr": 5.6917190221883307e-05}
{"train_loss": 0.14558114111423492, "global_step": 124005, "epoch": 1393, "lr": 5.691661605940872e-05}
{"train_loss": 0.2392074167728424, "global_step": 124006, "epoch": 1393, "lr": 5.691604189600429e-05}
{"train_loss": 0.18487712740898132, "global_step": 124007, "epoch": 1393, "lr": 5.691546773167007e-05}
{"train_loss": 0.1925179660320282, "global_step": 124008, "epoch": 1393, "lr": 5.6914893566406135e-05}
{"train_loss": 0.19574254751205444, "global_step": 124009, "epoch": 1393, "lr": 5.691431940021258e-05}
{"train_loss": 0.11775766313076019, "global_step": 124010, "epoch": 1393, "lr": 5.691374523308949e-05}
{"train_loss": 0.12578636407852173, "global_step": 124011, "epoch": 1393, "lr": 5.691317106503693e-05}
{"train_loss": 0.1242145523428917, "global_step": 124012, "epoch": 1393, "lr": 5.691259689605498e-05}
{"train_loss": 0.10979859530925751, "global_step": 124013, "epoch": 1393, "lr": 5.691202272614371e-05}
{"train_loss": 0.0983746200799942, "global_step": 124014, "epoch": 1393, "lr": 5.69114485553032e-05}
{"train_loss": 0.1549994945526123, "global_step": 124015, "epoch": 1393, "lr": 5.6910874383533543e-05}
{"train_loss": 0.1824652999639511, "global_step": 124016, "epoch": 1393, "lr": 5.691030021083479e-05}
{"train_loss": 0.22712019085884094, "global_step": 124017, "epoch": 1393, "lr": 5.6909726037207046e-05}
{"train_loss": 0.16431713104248047, "global_step": 124018, "epoch": 1393, "lr": 5.690915186265036e-05}
{"train_loss": 0.18364226818084717, "global_step": 124019, "epoch": 1393, "lr": 5.690857768716482e-05}
{"train_loss": 0.189292773604393, "global_step": 124020, "epoch": 1393, "lr": 5.690800351075052e-05}
{"train_loss": 0.11335477977991104, "global_step": 124021, "epoch": 1393, "lr": 5.690742933340751e-05}
{"train_loss": 0.1435069739818573, "global_step": 124022, "epoch": 1393, "lr": 5.69068551551359e-05}
{"train_loss": 0.1370954066514969, "global_step": 124023, "epoch": 1393, "lr": 5.6906280975935735e-05}
{"train_loss": 0.08153508603572845, "global_step": 124024, "epoch": 1393, "lr": 5.69057067958071e-05}
{"train_loss": 0.20675261318683624, "global_step": 124025, "epoch": 1393, "lr": 5.690513261475009e-05}
{"train_loss": 0.11882489174604416, "global_step": 124026, "epoch": 1393, "lr": 5.690455843276477e-05}
{"train_loss": 0.12290183454751968, "global_step": 124027, "epoch": 1393, "lr": 5.6903984249851184e-05}
{"train_loss": 0.11362022161483765, "global_step": 124028, "epoch": 1393, "lr": 5.690341006600946e-05}
{"train_loss": 0.17138983309268951, "global_step": 124029, "epoch": 1393, "lr": 5.690283588123967e-05}
{"train_loss": 0.13021622598171234, "global_step": 124030, "epoch": 1393, "lr": 5.690226169554187e-05}
{"train_loss": 0.11424829065799713, "global_step": 124031, "epoch": 1393, "lr": 5.690168750891613e-05}
{"train_loss": 0.161404550075531, "global_step": 124032, "epoch": 1393, "lr": 5.6901113321362564e-05}
{"train_loss": 0.10616499185562134, "global_step": 124033, "epoch": 1393, "lr": 5.6900539132881216e-05}
{"train_loss": 0.1078406423330307, "global_step": 124034, "epoch": 1393, "lr": 5.6899964943472184e-05}
{"train_loss": 0.09245635569095612, "global_step": 124035, "epoch": 1393, "lr": 5.689939075313553e-05}
{"train_loss": 0.10215961188077927, "global_step": 124036, "epoch": 1393, "lr": 5.6898816561871325e-05}
{"train_loss": 0.13630607724189758, "global_step": 124037, "epoch": 1393, "lr": 5.6898242369679674e-05}
{"train_loss": 0.11745452135801315, "global_step": 124038, "epoch": 1393, "lr": 5.689766817656064e-05}
{"train_loss": 0.13103485107421875, "global_step": 124039, "epoch": 1393, "lr": 5.6897093982514284e-05}
{"train_loss": 0.06845757365226746, "global_step": 124040, "epoch": 1393, "lr": 5.689651978754071e-05}
{"train_loss": 0.20867668092250824, "global_step": 124041, "epoch": 1393, "lr": 5.689594559163998e-05}
{"train_loss": 0.1603982150554657, "global_step": 124042, "epoch": 1393, "lr": 5.689537139481217e-05}
{"train_loss": 0.13532975316047668, "global_step": 124043, "epoch": 1393, "lr": 5.689479719705737e-05}
{"train_loss": 0.12517483532428741, "global_step": 124044, "epoch": 1393, "lr": 5.6894222998375646e-05}
{"train_loss": 0.15968884527683258, "global_step": 124045, "epoch": 1393, "lr": 5.689364879876708e-05}
{"train_loss": 0.15499886870384216, "global_step": 124046, "epoch": 1393, "lr": 5.689307459823173e-05}
{"train_loss": 0.10074737668037415, "global_step": 124047, "epoch": 1393, "lr": 5.689250039676971e-05}
{"train_loss": 0.10510993748903275, "global_step": 124048, "epoch": 1393, "lr": 5.6891926194381066e-05}
{"train_loss": 0.1266673356294632, "global_step": 124049, "epoch": 1393, "lr": 5.689135199106589e-05}
{"train_loss": 0.07358288764953613, "global_step": 124050, "epoch": 1393, "lr": 5.689077778682426e-05}
{"train_loss": 0.14927929639816284, "global_step": 124051, "epoch": 1393, "lr": 5.6890203581656245e-05}
{"train_loss": 0.09501063823699951, "global_step": 124052, "epoch": 1393, "lr": 5.6889629375561924e-05}
{"train_loss": 0.1990503966808319, "global_step": 124053, "epoch": 1393, "lr": 5.688905516854137e-05}
{"train_loss": 0.06793978065252304, "global_step": 124054, "epoch": 1393, "lr": 5.688848096059468e-05}
{"train_loss": 0.20969974994659424, "global_step": 124055, "epoch": 1393, "lr": 5.688790675172191e-05}
{"train_loss": 0.10149098187685013, "global_step": 124056, "epoch": 1393, "lr": 5.6887332541923156e-05}
{"train_loss": 0.07112860679626465, "global_step": 124057, "epoch": 1393, "lr": 5.6886758331198465e-05}
{"train_loss": 0.13254083693027496, "global_step": 124058, "epoch": 1393, "lr": 5.688618411954795e-05}
{"train_loss": 0.1978301852941513, "global_step": 124059, "epoch": 1393, "lr": 5.688560990697166e-05}
{"train_loss": 0.24143266677856445, "global_step": 124060, "epoch": 1393, "lr": 5.688503569346969e-05}
{"train_loss": 0.0678742527961731, "global_step": 124061, "epoch": 1393, "lr": 5.6884461479042106e-05}
{"train_loss": 0.17080195248126984, "global_step": 124062, "epoch": 1393, "lr": 5.6883887263689e-05}
{"train_loss": 0.11478833854198456, "global_step": 124063, "epoch": 1393, "lr": 5.688331304741043e-05}
{"train_loss": 0.18067079782485962, "global_step": 124064, "epoch": 1393, "lr": 5.688273883020648e-05}
{"train_loss": 0.1414887827266468, "global_step": 124065, "epoch": 1393, "lr": 5.688216461207724e-05, "val_loss": 5.369476795196533}
{"train_loss": 0.19376835227012634, "global_step": 124066, "epoch": 1394, "lr": 5.688159039302278e-05}
{"train_loss": 0.11402694135904312, "global_step": 124067, "epoch": 1394, "lr": 5.6881016173043156e-05}
{"train_loss": 0.1355443149805069, "global_step": 124068, "epoch": 1394, "lr": 5.688044195213848e-05}
{"train_loss": 0.2181241363286972, "global_step": 124069, "epoch": 1394, "lr": 5.68798677303088e-05}
{"train_loss": 0.1664891093969345, "global_step": 124070, "epoch": 1394, "lr": 5.687929350755421e-05}
{"train_loss": 0.21683384478092194, "global_step": 124071, "epoch": 1394, "lr": 5.6878719283874785e-05}
{"train_loss": 0.1560254842042923, "global_step": 124072, "epoch": 1394, "lr": 5.6878145059270605e-05}
{"train_loss": 0.14648663997650146, "global_step": 124073, "epoch": 1394, "lr": 5.687757083374173e-05}
{"train_loss": 0.15014806389808655, "global_step": 124074, "epoch": 1394, "lr": 5.687699660728826e-05}
{"train_loss": 0.13747596740722656, "global_step": 124075, "epoch": 1394, "lr": 5.687642237991025e-05}
{"train_loss": 0.10567549616098404, "global_step": 124076, "epoch": 1394, "lr": 5.6875848151607805e-05}
{"train_loss": 0.11879511922597885, "global_step": 124077, "epoch": 1394, "lr": 5.687527392238098e-05}
{"train_loss": 0.10957600921392441, "global_step": 124078, "epoch": 1394, "lr": 5.6874699692229846e-05}
{"train_loss": 0.17260149121284485, "global_step": 124079, "epoch": 1394, "lr": 5.6874125461154504e-05}
{"train_loss": 0.13602419197559357, "global_step": 124080, "epoch": 1394, "lr": 5.687355122915502e-05}
{"train_loss": 0.16623787581920624, "global_step": 124081, "epoch": 1394, "lr": 5.687297699623146e-05}
{"train_loss": 0.12955601513385773, "global_step": 124082, "epoch": 1394, "lr": 5.6872402762383924e-05}
{"train_loss": 0.15253491699695587, "global_step": 124083, "epoch": 1394, "lr": 5.687182852761248e-05}
{"train_loss": 0.15356089174747467, "global_step": 124084, "epoch": 1394, "lr": 5.68712542919172e-05}
{"train_loss": 0.14724743366241455, "global_step": 124085, "epoch": 1394, "lr": 5.6870680055298156e-05}
{"train_loss": 0.11699604988098145, "global_step": 124086, "epoch": 1394, "lr": 5.687010581775544e-05}
{"train_loss": 0.11702536046504974, "global_step": 124087, "epoch": 1394, "lr": 5.686953157928911e-05}
{"train_loss": 0.15249939262866974, "global_step": 124088, "epoch": 1394, "lr": 5.686895733989927e-05}
{"train_loss": 0.10096056759357452, "global_step": 124089, "epoch": 1394, "lr": 5.6868383099585975e-05}
{"train_loss": 0.1776362955570221, "global_step": 124090, "epoch": 1394, "lr": 5.68678088583493e-05}
{"train_loss": 0.12458405643701553, "global_step": 124091, "epoch": 1394, "lr": 5.686723461618935e-05}
{"train_loss": 0.18911050260066986, "global_step": 124092, "epoch": 1394, "lr": 5.6866660373106175e-05}
{"train_loss": 0.10498397797346115, "global_step": 124093, "epoch": 1394, "lr": 5.686608612909986e-05}
{"train_loss": 0.130994975566864, "global_step": 124094, "epoch": 1394, "lr": 5.6865511884170486e-05}
{"train_loss": 0.13136820495128632, "global_step": 124095, "epoch": 1394, "lr": 5.6864937638318126e-05}
{"train_loss": 0.10315564274787903, "global_step": 124096, "epoch": 1394, "lr": 5.6864363391542866e-05}
{"train_loss": 0.17507050931453705, "global_step": 124097, "epoch": 1394, "lr": 5.686378914384477e-05}
{"train_loss": 0.17593632638454437, "global_step": 124098, "epoch": 1394, "lr": 5.6863214895223924e-05}
{"train_loss": 0.10493801534175873, "global_step": 124099, "epoch": 1394, "lr": 5.68626406456804e-05}
{"train_loss": 0.2236042320728302, "global_step": 124100, "epoch": 1394, "lr": 5.6862066395214276e-05}
{"train_loss": 0.17392848432064056, "global_step": 124101, "epoch": 1394, "lr": 5.6861492143825634e-05}
{"train_loss": 0.16380879282951355, "global_step": 124102, "epoch": 1394, "lr": 5.686091789151454e-05}
{"train_loss": 0.12708964943885803, "global_step": 124103, "epoch": 1394, "lr": 5.686034363828109e-05}
{"train_loss": 0.10480982810258865, "global_step": 124104, "epoch": 1394, "lr": 5.685976938412535e-05}
{"train_loss": 0.11995495855808258, "global_step": 124105, "epoch": 1394, "lr": 5.685919512904739e-05}
{"train_loss": 0.11008453369140625, "global_step": 124106, "epoch": 1394, "lr": 5.6858620873047294e-05}
{"train_loss": 0.12581171095371246, "global_step": 124107, "epoch": 1394, "lr": 5.685804661612515e-05}
{"train_loss": 0.1745748221874237, "global_step": 124108, "epoch": 1394, "lr": 5.685747235828102e-05}
{"train_loss": 0.20200419425964355, "global_step": 124109, "epoch": 1394, "lr": 5.685689809951499e-05}
{"train_loss": 0.17982813715934753, "global_step": 124110, "epoch": 1394, "lr": 5.685632383982713e-05}
{"train_loss": 0.12913165986537933, "global_step": 124111, "epoch": 1394, "lr": 5.685574957921752e-05}
{"train_loss": 0.19413280487060547, "global_step": 124112, "epoch": 1394, "lr": 5.6855175317686235e-05}
{"train_loss": 0.1089031994342804, "global_step": 124113, "epoch": 1394, "lr": 5.685460105523336e-05}
{"train_loss": 0.10710050165653229, "global_step": 124114, "epoch": 1394, "lr": 5.685402679185896e-05}
{"train_loss": 0.17258210480213165, "global_step": 124115, "epoch": 1394, "lr": 5.685345252756313e-05}
{"train_loss": 0.12576600909233093, "global_step": 124116, "epoch": 1394, "lr": 5.685287826234593e-05}
{"train_loss": 0.14500871300697327, "global_step": 124117, "epoch": 1394, "lr": 5.685230399620744e-05}
{"train_loss": 0.1561930775642395, "global_step": 124118, "epoch": 1394, "lr": 5.685172972914776e-05}
{"train_loss": 0.1124780923128128, "global_step": 124119, "epoch": 1394, "lr": 5.685115546116694e-05}
{"train_loss": 0.11680228263139725, "global_step": 124120, "epoch": 1394, "lr": 5.685058119226506e-05}
{"train_loss": 0.09593971818685532, "global_step": 124121, "epoch": 1394, "lr": 5.68500069224422e-05}
{"train_loss": 0.16636866331100464, "global_step": 124122, "epoch": 1394, "lr": 5.6849432651698455e-05}
{"train_loss": 0.14614377915859222, "global_step": 124123, "epoch": 1394, "lr": 5.684885838003386e-05}
{"train_loss": 0.13554836809635162, "global_step": 124124, "epoch": 1394, "lr": 5.684828410744855e-05}
{"train_loss": 0.15049579739570618, "global_step": 124125, "epoch": 1394, "lr": 5.6847709833942554e-05}
{"train_loss": 0.11553578823804855, "global_step": 124126, "epoch": 1394, "lr": 5.684713555951597e-05}
{"train_loss": 0.17586344480514526, "global_step": 124127, "epoch": 1394, "lr": 5.684656128416887e-05}
{"train_loss": 0.12728506326675415, "global_step": 124128, "epoch": 1394, "lr": 5.684598700790135e-05}
{"train_loss": 0.1697111874818802, "global_step": 124129, "epoch": 1394, "lr": 5.684541273071345e-05}
{"train_loss": 0.14665210247039795, "global_step": 124130, "epoch": 1394, "lr": 5.684483845260528e-05}
{"train_loss": 0.12655708193778992, "global_step": 124131, "epoch": 1394, "lr": 5.68442641735769e-05}
{"train_loss": 0.14374031126499176, "global_step": 124132, "epoch": 1394, "lr": 5.684368989362838e-05}
{"train_loss": 0.12122994661331177, "global_step": 124133, "epoch": 1394, "lr": 5.6843115612759835e-05}
{"train_loss": 0.1591467559337616, "global_step": 124134, "epoch": 1394, "lr": 5.68425413309713e-05}
{"train_loss": 0.09129872173070908, "global_step": 124135, "epoch": 1394, "lr": 5.6841967048262865e-05}
{"train_loss": 0.16553717851638794, "global_step": 124136, "epoch": 1394, "lr": 5.6841392764634625e-05}
{"train_loss": 0.13142743706703186, "global_step": 124137, "epoch": 1394, "lr": 5.6840818480086635e-05}
{"train_loss": 0.20376010239124298, "global_step": 124138, "epoch": 1394, "lr": 5.684024419461898e-05}
{"train_loss": 0.09886544197797775, "global_step": 124139, "epoch": 1394, "lr": 5.683966990823173e-05}
{"train_loss": 0.15289618074893951, "global_step": 124140, "epoch": 1394, "lr": 5.6839095620924985e-05}
{"train_loss": 0.13094541430473328, "global_step": 124141, "epoch": 1394, "lr": 5.68385213326988e-05}
{"train_loss": 0.13447865843772888, "global_step": 124142, "epoch": 1394, "lr": 5.683794704355325e-05}
{"train_loss": 0.10954021662473679, "global_step": 124143, "epoch": 1394, "lr": 5.683737275348844e-05}
{"train_loss": 0.10623309761285782, "global_step": 124144, "epoch": 1394, "lr": 5.6836798462504416e-05}
{"train_loss": 0.10413842648267746, "global_step": 124145, "epoch": 1394, "lr": 5.6836224170601274e-05}
{"train_loss": 0.12544173002243042, "global_step": 124146, "epoch": 1394, "lr": 5.683564987777909e-05}
{"train_loss": 0.18175438046455383, "global_step": 124147, "epoch": 1394, "lr": 5.6835075584037924e-05}
{"train_loss": 0.15708163380622864, "global_step": 124148, "epoch": 1394, "lr": 5.6834501289377865e-05}
{"train_loss": 0.12299206852912903, "global_step": 124149, "epoch": 1394, "lr": 5.683392699379899e-05}
{"train_loss": 0.17485108971595764, "global_step": 124150, "epoch": 1394, "lr": 5.683335269730139e-05}
{"train_loss": 0.1240706592798233, "global_step": 124151, "epoch": 1394, "lr": 5.6832778399885124e-05}
{"train_loss": 0.1434321403503418, "global_step": 124152, "epoch": 1394, "lr": 5.683220410155027e-05}
{"train_loss": 0.11521024256944656, "global_step": 124153, "epoch": 1394, "lr": 5.68316298022969e-05}
{"train_loss": 0.14190447037474493, "global_step": 124154, "epoch": 1394, "lr": 5.683105550212511e-05, "val_loss": 5.154327392578125}
{"train_loss": 0.1339157670736313, "global_step": 124155, "epoch": 1395, "lr": 5.683048120103498e-05}
{"train_loss": 0.18865594267845154, "global_step": 124156, "epoch": 1395, "lr": 5.6829906899026566e-05}
{"train_loss": 0.12229569256305695, "global_step": 124157, "epoch": 1395, "lr": 5.682933259609995e-05}
{"train_loss": 0.1477656364440918, "global_step": 124158, "epoch": 1395, "lr": 5.682875829225521e-05}
{"train_loss": 0.12324219942092896, "global_step": 124159, "epoch": 1395, "lr": 5.6828183987492436e-05}
{"train_loss": 0.12230613827705383, "global_step": 124160, "epoch": 1395, "lr": 5.6827609681811686e-05}
{"train_loss": 0.10071392357349396, "global_step": 124161, "epoch": 1395, "lr": 5.682703537521306e-05}
{"train_loss": 0.15613603591918945, "global_step": 124162, "epoch": 1395, "lr": 5.682646106769661e-05}
{"train_loss": 0.19661250710487366, "global_step": 124163, "epoch": 1395, "lr": 5.6825886759262434e-05}
{"train_loss": 0.1725069284439087, "global_step": 124164, "epoch": 1395, "lr": 5.682531244991061e-05}
{"train_loss": 0.09518714994192123, "global_step": 124165, "epoch": 1395, "lr": 5.682473813964119e-05}
{"train_loss": 0.10756710171699524, "global_step": 124166, "epoch": 1395, "lr": 5.682416382845427e-05}
{"train_loss": 0.052558135241270065, "global_step": 124167, "epoch": 1395, "lr": 5.6823589516349915e-05}
{"train_loss": 0.10098563879728317, "global_step": 124168, "epoch": 1395, "lr": 5.682301520332822e-05}
{"train_loss": 0.09819388389587402, "global_step": 124169, "epoch": 1395, "lr": 5.682244088938926e-05}
{"train_loss": 0.18195660412311554, "global_step": 124170, "epoch": 1395, "lr": 5.68218665745331e-05}
{"train_loss": 0.10353752970695496, "global_step": 124171, "epoch": 1395, "lr": 5.6821292258759826e-05}
{"train_loss": 0.1281827837228775, "global_step": 124172, "epoch": 1395, "lr": 5.68207179420695e-05}
{"train_loss": 0.15738436579704285, "global_step": 124173, "epoch": 1395, "lr": 5.682014362446223e-05}
{"train_loss": 0.13051067292690277, "global_step": 124174, "epoch": 1395, "lr": 5.681956930593807e-05}
{"train_loss": 0.10839157551527023, "global_step": 124175, "epoch": 1395, "lr": 5.681899498649711e-05}
{"train_loss": 0.07772921025753021, "global_step": 124176, "epoch": 1395, "lr": 5.6818420666139405e-05}
{"train_loss": 0.13906146585941315, "global_step": 124177, "epoch": 1395, "lr": 5.6817846344865056e-05}
{"train_loss": 0.12577897310256958, "global_step": 124178, "epoch": 1395, "lr": 5.6817272022674116e-05}
{"train_loss": 0.14220179617404938, "global_step": 124179, "epoch": 1395, "lr": 5.681669769956669e-05}
{"train_loss": 0.1322583556175232, "global_step": 124180, "epoch": 1395, "lr": 5.681612337554284e-05}
{"train_loss": 0.10699544847011566, "global_step": 124181, "epoch": 1395, "lr": 5.681554905060265e-05}
{"train_loss": 0.15109050273895264, "global_step": 124182, "epoch": 1395, "lr": 5.681497472474617e-05}
{"train_loss": 0.2133457213640213, "global_step": 124183, "epoch": 1395, "lr": 5.681440039797352e-05}
{"train_loss": 0.08660392463207245, "global_step": 124184, "epoch": 1395, "lr": 5.681382607028476e-05}
{"train_loss": 0.11384306102991104, "global_step": 124185, "epoch": 1395, "lr": 5.681325174167995e-05}
{"train_loss": 0.11723002791404724, "global_step": 124186, "epoch": 1395, "lr": 5.681267741215919e-05}
{"train_loss": 0.1471376121044159, "global_step": 124187, "epoch": 1395, "lr": 5.6812103081722544e-05}
{"train_loss": 0.15435373783111572, "global_step": 124188, "epoch": 1395, "lr": 5.68115287503701e-05}
{"train_loss": 0.058629248291254044, "global_step": 124189, "epoch": 1395, "lr": 5.6810954418101923e-05}
{"train_loss": 0.1185368001461029, "global_step": 124190, "epoch": 1395, "lr": 5.68103800849181e-05}
{"train_loss": 0.15230384469032288, "global_step": 124191, "epoch": 1395, "lr": 5.680980575081869e-05}
{"train_loss": 0.13933828473091125, "global_step": 124192, "epoch": 1395, "lr": 5.68092314158038e-05}
{"train_loss": 0.088035948574543, "global_step": 124193, "epoch": 1395, "lr": 5.6808657079873485e-05}
{"train_loss": 0.23676972091197968, "global_step": 124194, "epoch": 1395, "lr": 5.6808082743027845e-05}
{"train_loss": 0.14452198147773743, "global_step": 124195, "epoch": 1395, "lr": 5.680750840526693e-05}
{"train_loss": 0.1360468566417694, "global_step": 124196, "epoch": 1395, "lr": 5.6806934066590836e-05}
{"train_loss": 0.180436909198761, "global_step": 124197, "epoch": 1395, "lr": 5.6806359726999624e-05}
{"train_loss": 0.14532603323459625, "global_step": 124198, "epoch": 1395, "lr": 5.6805785386493385e-05}
{"train_loss": 0.13844962418079376, "global_step": 124199, "epoch": 1395, "lr": 5.680521104507218e-05}
{"train_loss": 0.06989635527133942, "global_step": 124200, "epoch": 1395, "lr": 5.68046367027361e-05}
{"train_loss": 0.15930049121379852, "global_step": 124201, "epoch": 1395, "lr": 5.680406235948523e-05}
{"train_loss": 0.2055872678756714, "global_step": 124202, "epoch": 1395, "lr": 5.680348801531963e-05}
{"train_loss": 0.121497742831707, "global_step": 124203, "epoch": 1395, "lr": 5.680291367023938e-05}
{"train_loss": 0.1246255561709404, "global_step": 124204, "epoch": 1395, "lr": 5.680233932424457e-05}
{"train_loss": 0.12372615933418274, "global_step": 124205, "epoch": 1395, "lr": 5.680176497733526e-05}
{"train_loss": 0.11145991832017899, "global_step": 124206, "epoch": 1395, "lr": 5.6801190629511546e-05}
{"train_loss": 0.11533598601818085, "global_step": 124207, "epoch": 1395, "lr": 5.680061628077349e-05}
{"train_loss": 0.14769834280014038, "global_step": 124208, "epoch": 1395, "lr": 5.6800041931121164e-05}
{"train_loss": 0.1896085888147354, "global_step": 124209, "epoch": 1395, "lr": 5.6799467580554654e-05}
{"train_loss": 0.1983989030122757, "global_step": 124210, "epoch": 1395, "lr": 5.679889322907405e-05}
{"train_loss": 0.1191987544298172, "global_step": 124211, "epoch": 1395, "lr": 5.6798318876679416e-05}
{"train_loss": 0.13088220357894897, "global_step": 124212, "epoch": 1395, "lr": 5.679774452337083e-05}
{"train_loss": 0.15595242381095886, "global_step": 124213, "epoch": 1395, "lr": 5.6797170169148384e-05}
{"train_loss": 0.1662740856409073, "global_step": 124214, "epoch": 1395, "lr": 5.6796595814012124e-05}
{"train_loss": 0.094635508954525, "global_step": 124215, "epoch": 1395, "lr": 5.6796021457962146e-05}
{"train_loss": 0.08312106877565384, "global_step": 124216, "epoch": 1395, "lr": 5.679544710099853e-05}
{"train_loss": 0.14176157116889954, "global_step": 124217, "epoch": 1395, "lr": 5.6794872743121355e-05}
{"train_loss": 0.12502668797969818, "global_step": 124218, "epoch": 1395, "lr": 5.679429838433068e-05}
{"train_loss": 0.17262786626815796, "global_step": 124219, "epoch": 1395, "lr": 5.679372402462659e-05}
{"train_loss": 0.1423371285200119, "global_step": 124220, "epoch": 1395, "lr": 5.679314966400918e-05}
{"train_loss": 0.15298642218112946, "global_step": 124221, "epoch": 1395, "lr": 5.6792575302478515e-05}
{"train_loss": 0.09698283672332764, "global_step": 124222, "epoch": 1395, "lr": 5.679200094003466e-05}
{"train_loss": 0.09853581339120865, "global_step": 124223, "epoch": 1395, "lr": 5.679142657667771e-05}
{"train_loss": 0.10695341229438782, "global_step": 124224, "epoch": 1395, "lr": 5.679085221240774e-05}
{"train_loss": 0.1463450938463211, "global_step": 124225, "epoch": 1395, "lr": 5.679027784722482e-05}
{"train_loss": 0.09054090827703476, "global_step": 124226, "epoch": 1395, "lr": 5.6789703481129016e-05}
{"train_loss": 0.19003108143806458, "global_step": 124227, "epoch": 1395, "lr": 5.678912911412042e-05}
{"train_loss": 0.13894103467464447, "global_step": 124228, "epoch": 1395, "lr": 5.678855474619912e-05}
{"train_loss": 0.12175674736499786, "global_step": 124229, "epoch": 1395, "lr": 5.6787980377365176e-05}
{"train_loss": 0.1418096125125885, "global_step": 124230, "epoch": 1395, "lr": 5.678740600761868e-05}
{"train_loss": 0.12945252656936646, "global_step": 124231, "epoch": 1395, "lr": 5.678683163695969e-05}
{"train_loss": 0.10112608224153519, "global_step": 124232, "epoch": 1395, "lr": 5.67862572653883e-05}
{"train_loss": 0.0887540727853775, "global_step": 124233, "epoch": 1395, "lr": 5.678568289290459e-05}
{"train_loss": 0.12889143824577332, "global_step": 124234, "epoch": 1395, "lr": 5.678510851950861e-05}
{"train_loss": 0.10897792130708694, "global_step": 124235, "epoch": 1395, "lr": 5.6784534145200466e-05}
{"train_loss": 0.09160184860229492, "global_step": 124236, "epoch": 1395, "lr": 5.678395976998022e-05}
{"train_loss": 0.124354287981987, "global_step": 124237, "epoch": 1395, "lr": 5.6783385393847954e-05}
{"train_loss": 0.1699281930923462, "global_step": 124238, "epoch": 1395, "lr": 5.6782811016803736e-05}
{"train_loss": 0.14907175302505493, "global_step": 124239, "epoch": 1395, "lr": 5.678223663884766e-05}
{"train_loss": 0.17739027738571167, "global_step": 124240, "epoch": 1395, "lr": 5.67816622599798e-05}
{"train_loss": 0.11709211766719818, "global_step": 124241, "epoch": 1395, "lr": 5.678108788020021e-05}
{"train_loss": 0.11810656636953354, "global_step": 124242, "epoch": 1395, "lr": 5.6780513499509e-05}
{"train_loss": 0.13169934763834718, "global_step": 124243, "epoch": 1395, "lr": 5.6779939117906236e-05, "val_loss": 5.330090045928955, "train_action_mse_error": 23.213890075683594}
{"train_loss": 0.11806152760982513, "global_step": 124244, "epoch": 1396, "lr": 5.677936473539198e-05}
{"train_loss": 0.13087671995162964, "global_step": 124245, "epoch": 1396, "lr": 5.677879035196632e-05}
{"train_loss": 0.14549052715301514, "global_step": 124246, "epoch": 1396, "lr": 5.677821596762934e-05}
{"train_loss": 0.14383551478385925, "global_step": 124247, "epoch": 1396, "lr": 5.67776415823811e-05}
{"train_loss": 0.13382069766521454, "global_step": 124248, "epoch": 1396, "lr": 5.677706719622171e-05}
{"train_loss": 0.13128015398979187, "global_step": 124249, "epoch": 1396, "lr": 5.6776492809151214e-05}
{"train_loss": 0.09893006086349487, "global_step": 124250, "epoch": 1396, "lr": 5.677591842116971e-05}
{"train_loss": 0.14005543291568756, "global_step": 124251, "epoch": 1396, "lr": 5.677534403227725e-05}
{"train_loss": 0.17922598123550415, "global_step": 124252, "epoch": 1396, "lr": 5.6774769642473954e-05}
{"train_loss": 0.11709515005350113, "global_step": 124253, "epoch": 1396, "lr": 5.677419525175985e-05}
{"train_loss": 0.13057801127433777, "global_step": 124254, "epoch": 1396, "lr": 5.6773620860135045e-05}
{"train_loss": 0.13596494495868683, "global_step": 124255, "epoch": 1396, "lr": 5.677304646759961e-05}
{"train_loss": 0.14573979377746582, "global_step": 124256, "epoch": 1396, "lr": 5.6772472074153624e-05}
{"train_loss": 0.1643722504377365, "global_step": 124257, "epoch": 1396, "lr": 5.677189767979716e-05}
{"train_loss": 0.1004086285829544, "global_step": 124258, "epoch": 1396, "lr": 5.6771323284530286e-05}
{"train_loss": 0.11214989423751831, "global_step": 124259, "epoch": 1396, "lr": 5.67707488883531e-05}
{"train_loss": 0.13066938519477844, "global_step": 124260, "epoch": 1396, "lr": 5.677017449126567e-05}
{"train_loss": 0.12035100907087326, "global_step": 124261, "epoch": 1396, "lr": 5.676960009326807e-05}
{"train_loss": 0.08957797288894653, "global_step": 124262, "epoch": 1396, "lr": 5.676902569436038e-05}
{"train_loss": 0.1853591501712799, "global_step": 124263, "epoch": 1396, "lr": 5.6768451294542666e-05}
{"train_loss": 0.07228021323680878, "global_step": 124264, "epoch": 1396, "lr": 5.676787689381503e-05}
{"train_loss": 0.15054000914096832, "global_step": 124265, "epoch": 1396, "lr": 5.6767302492177533e-05}
{"train_loss": 0.09840822219848633, "global_step": 124266, "epoch": 1396, "lr": 5.676672808963026e-05}
{"train_loss": 0.1080498993396759, "global_step": 124267, "epoch": 1396, "lr": 5.6766153686173276e-05}
{"train_loss": 0.1490027904510498, "global_step": 124268, "epoch": 1396, "lr": 5.676557928180667e-05}
{"train_loss": 0.125367671251297, "global_step": 124269, "epoch": 1396, "lr": 5.6765004876530504e-05}
{"train_loss": 0.13918499648571014, "global_step": 124270, "epoch": 1396, "lr": 5.676443047034488e-05}
{"train_loss": 0.11190960556268692, "global_step": 124271, "epoch": 1396, "lr": 5.6763856063249854e-05}
{"train_loss": 0.11671362072229385, "global_step": 124272, "epoch": 1396, "lr": 5.676328165524551e-05}
{"train_loss": 0.07509156316518784, "global_step": 124273, "epoch": 1396, "lr": 5.676270724633192e-05}
{"train_loss": 0.07137887924909592, "global_step": 124274, "epoch": 1396, "lr": 5.676213283650918e-05}
{"train_loss": 0.14569303393363953, "global_step": 124275, "epoch": 1396, "lr": 5.676155842577735e-05}
{"train_loss": 0.14453382790088654, "global_step": 124276, "epoch": 1396, "lr": 5.6760984014136496e-05}
{"train_loss": 0.09113247692584991, "global_step": 124277, "epoch": 1396, "lr": 5.676040960158673e-05}
{"train_loss": 0.14039602875709534, "global_step": 124278, "epoch": 1396, "lr": 5.67598351881281e-05}
{"train_loss": 0.12548844516277313, "global_step": 124279, "epoch": 1396, "lr": 5.675926077376068e-05}
{"train_loss": 0.18068213760852814, "global_step": 124280, "epoch": 1396, "lr": 5.675868635848458e-05}
{"train_loss": 0.10544871538877487, "global_step": 124281, "epoch": 1396, "lr": 5.675811194229983e-05}
{"train_loss": 0.1969127357006073, "global_step": 124282, "epoch": 1396, "lr": 5.675753752520656e-05}
{"train_loss": 0.1094980239868164, "global_step": 124283, "epoch": 1396, "lr": 5.675696310720481e-05}
{"train_loss": 0.12755653262138367, "global_step": 124284, "epoch": 1396, "lr": 5.675638868829467e-05}
{"train_loss": 0.09376966208219528, "global_step": 124285, "epoch": 1396, "lr": 5.675581426847623e-05}
{"train_loss": 0.10739695280790329, "global_step": 124286, "epoch": 1396, "lr": 5.675523984774953e-05}
{"train_loss": 0.174078106880188, "global_step": 124287, "epoch": 1396, "lr": 5.675466542611468e-05}
{"train_loss": 0.16227801144123077, "global_step": 124288, "epoch": 1396, "lr": 5.675409100357175e-05}
{"train_loss": 0.06041717156767845, "global_step": 124289, "epoch": 1396, "lr": 5.675351658012081e-05}
{"train_loss": 0.13666865229606628, "global_step": 124290, "epoch": 1396, "lr": 5.675294215576196e-05}
{"train_loss": 0.13128268718719482, "global_step": 124291, "epoch": 1396, "lr": 5.6752367730495246e-05}
{"train_loss": 0.08265549689531326, "global_step": 124292, "epoch": 1396, "lr": 5.6751793304320766e-05}
{"train_loss": 0.11925040930509567, "global_step": 124293, "epoch": 1396, "lr": 5.675121887723858e-05}
{"train_loss": 0.0754946619272232, "global_step": 124294, "epoch": 1396, "lr": 5.675064444924878e-05}
{"train_loss": 0.16353575885295868, "global_step": 124295, "epoch": 1396, "lr": 5.675007002035143e-05}
{"train_loss": 0.14370407164096832, "global_step": 124296, "epoch": 1396, "lr": 5.674949559054662e-05}
{"train_loss": 0.1305045336484909, "global_step": 124297, "epoch": 1396, "lr": 5.6748921159834425e-05}
{"train_loss": 0.11688030511140823, "global_step": 124298, "epoch": 1396, "lr": 5.674834672821492e-05}
{"train_loss": 0.11095156520605087, "global_step": 124299, "epoch": 1396, "lr": 5.674777229568816e-05}
{"train_loss": 0.13356095552444458, "global_step": 124300, "epoch": 1396, "lr": 5.674719786225427e-05}
{"train_loss": 0.05790242925286293, "global_step": 124301, "epoch": 1396, "lr": 5.674662342791329e-05}
{"train_loss": 0.12147822231054306, "global_step": 124302, "epoch": 1396, "lr": 5.6746048992665314e-05}
{"train_loss": 0.1942797601222992, "global_step": 124303, "epoch": 1396, "lr": 5.6745474556510414e-05}
{"train_loss": 0.13269849121570587, "global_step": 124304, "epoch": 1396, "lr": 5.6744900119448674e-05}
{"train_loss": 0.1297108381986618, "global_step": 124305, "epoch": 1396, "lr": 5.674432568148015e-05}
{"train_loss": 0.11859538406133652, "global_step": 124306, "epoch": 1396, "lr": 5.674375124260495e-05}
{"train_loss": 0.1161661148071289, "global_step": 124307, "epoch": 1396, "lr": 5.6743176802823116e-05}
{"train_loss": 0.13188740611076355, "global_step": 124308, "epoch": 1396, "lr": 5.6742602362134754e-05}
{"train_loss": 0.09494604915380478, "global_step": 124309, "epoch": 1396, "lr": 5.674202792053993e-05}
{"train_loss": 0.07114766538143158, "global_step": 124310, "epoch": 1396, "lr": 5.674145347803874e-05}
{"train_loss": 0.1334710717201233, "global_step": 124311, "epoch": 1396, "lr": 5.6740879034631233e-05}
{"train_loss": 0.0783713310956955, "global_step": 124312, "epoch": 1396, "lr": 5.674030459031748e-05}
{"train_loss": 0.09925980120897293, "global_step": 124313, "epoch": 1396, "lr": 5.6739730145097595e-05}
{"train_loss": 0.15689383447170258, "global_step": 124314, "epoch": 1396, "lr": 5.6739155698971626e-05}
{"train_loss": 0.14313530921936035, "global_step": 124315, "epoch": 1396, "lr": 5.673858125193966e-05}
{"train_loss": 0.1304633468389511, "global_step": 124316, "epoch": 1396, "lr": 5.673800680400178e-05}
{"train_loss": 0.07743645459413528, "global_step": 124317, "epoch": 1396, "lr": 5.6737432355158046e-05}
{"train_loss": 0.13440142571926117, "global_step": 124318, "epoch": 1396, "lr": 5.6736857905408545e-05}
{"train_loss": 0.13410082459449768, "global_step": 124319, "epoch": 1396, "lr": 5.6736283454753356e-05}
{"train_loss": 0.07728505879640579, "global_step": 124320, "epoch": 1396, "lr": 5.673570900319256e-05}
{"train_loss": 0.06535457819700241, "global_step": 124321, "epoch": 1396, "lr": 5.673513455072623e-05}
{"train_loss": 0.09142771363258362, "global_step": 124322, "epoch": 1396, "lr": 5.673456009735445e-05}
{"train_loss": 0.12747621536254883, "global_step": 124323, "epoch": 1396, "lr": 5.6733985643077295e-05}
{"train_loss": 0.17301644384860992, "global_step": 124324, "epoch": 1396, "lr": 5.673341118789482e-05}
{"train_loss": 0.10338010638952255, "global_step": 124325, "epoch": 1396, "lr": 5.673283673180714e-05}
{"train_loss": 0.13006477057933807, "global_step": 124326, "epoch": 1396, "lr": 5.67322622748143e-05}
{"train_loss": 0.11668713390827179, "global_step": 124327, "epoch": 1396, "lr": 5.673168781691638e-05}
{"train_loss": 0.08713936805725098, "global_step": 124328, "epoch": 1396, "lr": 5.673111335811349e-05}
{"train_loss": 0.09328759461641312, "global_step": 124329, "epoch": 1396, "lr": 5.673053889840567e-05}
{"train_loss": 0.11018535494804382, "global_step": 124330, "epoch": 1396, "lr": 5.672996443779301e-05}
{"train_loss": 0.15959079563617706, "global_step": 124331, "epoch": 1396, "lr": 5.6729389976275595e-05}
{"train_loss": 0.1221465023883273, "global_step": 124332, "epoch": 1396, "lr": 5.672881551385349e-05, "val_loss": 5.578023433685303}
{"train_loss": 0.12871724367141724, "global_step": 124333, "epoch": 1397, "lr": 5.672824105052678e-05}
{"train_loss": 0.09839753806591034, "global_step": 124334, "epoch": 1397, "lr": 5.6727666586295534e-05}
{"train_loss": 0.13592971861362457, "global_step": 124335, "epoch": 1397, "lr": 5.672709212115984e-05}
{"train_loss": 0.10519874095916748, "global_step": 124336, "epoch": 1397, "lr": 5.672651765511976e-05}
{"train_loss": 0.1300775408744812, "global_step": 124337, "epoch": 1397, "lr": 5.672594318817539e-05}
{"train_loss": 0.1619301736354828, "global_step": 124338, "epoch": 1397, "lr": 5.672536872032681e-05}
{"train_loss": 0.12470950931310654, "global_step": 124339, "epoch": 1397, "lr": 5.672479425157408e-05}
{"train_loss": 0.16536305844783783, "global_step": 124340, "epoch": 1397, "lr": 5.6724219781917276e-05}
{"train_loss": 0.1547788679599762, "global_step": 124341, "epoch": 1397, "lr": 5.672364531135649e-05}
{"train_loss": 0.07400931417942047, "global_step": 124342, "epoch": 1397, "lr": 5.672307083989179e-05}
{"train_loss": 0.11950870603322983, "global_step": 124343, "epoch": 1397, "lr": 5.6722496367523246e-05}
{"train_loss": 0.05921746417880058, "global_step": 124344, "epoch": 1397, "lr": 5.672192189425095e-05}
{"train_loss": 0.10412603616714478, "global_step": 124345, "epoch": 1397, "lr": 5.6721347420074976e-05}
{"train_loss": 0.12251189351081848, "global_step": 124346, "epoch": 1397, "lr": 5.67207729449954e-05}
{"train_loss": 0.1725897639989853, "global_step": 124347, "epoch": 1397, "lr": 5.67201984690123e-05}
{"train_loss": 0.08550147712230682, "global_step": 124348, "epoch": 1397, "lr": 5.671962399212575e-05}
{"train_loss": 0.14620789885520935, "global_step": 124349, "epoch": 1397, "lr": 5.671904951433583e-05}
{"train_loss": 0.15025120973587036, "global_step": 124350, "epoch": 1397, "lr": 5.671847503564262e-05}
{"train_loss": 0.13076072931289673, "global_step": 124351, "epoch": 1397, "lr": 5.671790055604618e-05}
{"train_loss": 0.10612563043832779, "global_step": 124352, "epoch": 1397, "lr": 5.671732607554661e-05}
{"train_loss": 0.09656226634979248, "global_step": 124353, "epoch": 1397, "lr": 5.671675159414397e-05}
{"train_loss": 0.12214326858520508, "global_step": 124354, "epoch": 1397, "lr": 5.671617711183833e-05}
{"train_loss": 0.11504346877336502, "global_step": 124355, "epoch": 1397, "lr": 5.6715602628629806e-05}
{"train_loss": 0.11957195401191711, "global_step": 124356, "epoch": 1397, "lr": 5.6715028144518435e-05}
{"train_loss": 0.16757898032665253, "global_step": 124357, "epoch": 1397, "lr": 5.671445365950432e-05}
{"train_loss": 0.14476026594638824, "global_step": 124358, "epoch": 1397, "lr": 5.671387917358754e-05}
{"train_loss": 0.1726992279291153, "global_step": 124359, "epoch": 1397, "lr": 5.671330468676814e-05}
{"train_loss": 0.08413577079772949, "global_step": 124360, "epoch": 1397, "lr": 5.671273019904623e-05}
{"train_loss": 0.056264471262693405, "global_step": 124361, "epoch": 1397, "lr": 5.671215571042187e-05}
{"train_loss": 0.17125701904296875, "global_step": 124362, "epoch": 1397, "lr": 5.6711581220895136e-05}
{"train_loss": 0.1759875863790512, "global_step": 124363, "epoch": 1397, "lr": 5.671100673046612e-05}
{"train_loss": 0.1281951665878296, "global_step": 124364, "epoch": 1397, "lr": 5.67104322391349e-05}
{"train_loss": 0.14375907182693481, "global_step": 124365, "epoch": 1397, "lr": 5.670985774690153e-05}
{"train_loss": 0.16422656178474426, "global_step": 124366, "epoch": 1397, "lr": 5.670928325376611e-05}
{"train_loss": 0.1784263402223587, "global_step": 124367, "epoch": 1397, "lr": 5.67087087597287e-05}
{"train_loss": 0.09852202236652374, "global_step": 124368, "epoch": 1397, "lr": 5.6708134264789404e-05}
{"train_loss": 0.12265019118785858, "global_step": 124369, "epoch": 1397, "lr": 5.670755976894827e-05}
{"train_loss": 0.05851462855935097, "global_step": 124370, "epoch": 1397, "lr": 5.6706985272205384e-05}
{"train_loss": 0.09945932030677795, "global_step": 124371, "epoch": 1397, "lr": 5.6706410774560826e-05}
{"train_loss": 0.1031370609998703, "global_step": 124372, "epoch": 1397, "lr": 5.6705836276014665e-05}
{"train_loss": 0.13350602984428406, "global_step": 124373, "epoch": 1397, "lr": 5.670526177656699e-05}
{"train_loss": 0.17627176642417908, "global_step": 124374, "epoch": 1397, "lr": 5.670468727621788e-05}
{"train_loss": 0.15161669254302979, "global_step": 124375, "epoch": 1397, "lr": 5.67041127749674e-05}
{"train_loss": 0.12505963444709778, "global_step": 124376, "epoch": 1397, "lr": 5.6703538272815645e-05}
{"train_loss": 0.11217404156923294, "global_step": 124377, "epoch": 1397, "lr": 5.670296376976267e-05}
{"train_loss": 0.16216900944709778, "global_step": 124378, "epoch": 1397, "lr": 5.6702389265808566e-05}
{"train_loss": 0.13953128457069397, "global_step": 124379, "epoch": 1397, "lr": 5.670181476095341e-05}
{"train_loss": 0.05318481847643852, "global_step": 124380, "epoch": 1397, "lr": 5.670124025519727e-05}
{"train_loss": 0.15549321472644806, "global_step": 124381, "epoch": 1397, "lr": 5.670066574854025e-05}
{"train_loss": 0.11925377696752548, "global_step": 124382, "epoch": 1397, "lr": 5.670009124098239e-05}
{"train_loss": 0.07686743140220642, "global_step": 124383, "epoch": 1397, "lr": 5.669951673252378e-05}
{"train_loss": 0.1351889669895172, "global_step": 124384, "epoch": 1397, "lr": 5.6698942223164506e-05}
{"train_loss": 0.14153556525707245, "global_step": 124385, "epoch": 1397, "lr": 5.669836771290465e-05}
{"train_loss": 0.1018642783164978, "global_step": 124386, "epoch": 1397, "lr": 5.669779320174427e-05}
{"train_loss": 0.15729855000972748, "global_step": 124387, "epoch": 1397, "lr": 5.6697218689683464e-05}
{"train_loss": 0.16804511845111847, "global_step": 124388, "epoch": 1397, "lr": 5.66966441767223e-05}
{"train_loss": 0.11881902813911438, "global_step": 124389, "epoch": 1397, "lr": 5.669606966286085e-05}
{"train_loss": 0.11702542752027512, "global_step": 124390, "epoch": 1397, "lr": 5.669549514809918e-05}
{"train_loss": 0.13638123869895935, "global_step": 124391, "epoch": 1397, "lr": 5.669492063243739e-05}
{"train_loss": 0.08393865823745728, "global_step": 124392, "epoch": 1397, "lr": 5.669434611587555e-05}
{"train_loss": 0.15495692193508148, "global_step": 124393, "epoch": 1397, "lr": 5.669377159841374e-05}
{"train_loss": 0.1306867152452469, "global_step": 124394, "epoch": 1397, "lr": 5.6693197080052026e-05}
{"train_loss": 0.134856179356575, "global_step": 124395, "epoch": 1397, "lr": 5.669262256079051e-05}
{"train_loss": 0.1552746444940567, "global_step": 124396, "epoch": 1397, "lr": 5.669204804062924e-05}
{"train_loss": 0.07614631950855255, "global_step": 124397, "epoch": 1397, "lr": 5.6691473519568306e-05}
{"train_loss": 0.10235558450222015, "global_step": 124398, "epoch": 1397, "lr": 5.669089899760779e-05}
{"train_loss": 0.08278578519821167, "global_step": 124399, "epoch": 1397, "lr": 5.669032447474776e-05}
{"train_loss": 0.12615159153938293, "global_step": 124400, "epoch": 1397, "lr": 5.66897499509883e-05}
{"train_loss": 0.1411113291978836, "global_step": 124401, "epoch": 1397, "lr": 5.668917542632949e-05}
{"train_loss": 0.17096030712127686, "global_step": 124402, "epoch": 1397, "lr": 5.668860090077139e-05}
{"train_loss": 0.22502271831035614, "global_step": 124403, "epoch": 1397, "lr": 5.668802637431409e-05}
{"train_loss": 0.16747508943080902, "global_step": 124404, "epoch": 1397, "lr": 5.668745184695767e-05}
{"train_loss": 0.17242015898227692, "global_step": 124405, "epoch": 1397, "lr": 5.6686877318702216e-05}
{"train_loss": 0.12490971386432648, "global_step": 124406, "epoch": 1397, "lr": 5.6686302789547774e-05}
{"train_loss": 0.13792206346988678, "global_step": 124407, "epoch": 1397, "lr": 5.668572825949446e-05}
{"train_loss": 0.09853346645832062, "global_step": 124408, "epoch": 1397, "lr": 5.66851537285423e-05}
{"train_loss": 0.10225573182106018, "global_step": 124409, "epoch": 1397, "lr": 5.668457919669144e-05}
{"train_loss": 0.10906210541725159, "global_step": 124410, "epoch": 1397, "lr": 5.668400466394189e-05}
{"train_loss": 0.11032313108444214, "global_step": 124411, "epoch": 1397, "lr": 5.668343013029377e-05}
{"train_loss": 0.14892128109931946, "global_step": 124412, "epoch": 1397, "lr": 5.668285559574714e-05}
{"train_loss": 0.15317925810813904, "global_step": 124413, "epoch": 1397, "lr": 5.668228106030209e-05}
{"train_loss": 0.08158783614635468, "global_step": 124414, "epoch": 1397, "lr": 5.668170652395868e-05}
{"train_loss": 0.11908657103776932, "global_step": 124415, "epoch": 1397, "lr": 5.6681131986716995e-05}
{"train_loss": 0.09972183406352997, "global_step": 124416, "epoch": 1397, "lr": 5.668055744857712e-05}
{"train_loss": 0.1036272794008255, "global_step": 124417, "epoch": 1397, "lr": 5.667998290953912e-05}
{"train_loss": 0.11791907995939255, "global_step": 124418, "epoch": 1397, "lr": 5.6679408369603084e-05}
{"train_loss": 0.032018695026636124, "global_step": 124419, "epoch": 1397, "lr": 5.667883382876909e-05}
{"train_loss": 0.1284872144460678, "global_step": 124420, "epoch": 1397, "lr": 5.667825928703718e-05}
{"train_loss": 0.125381287857053, "global_step": 124421, "epoch": 1397, "lr": 5.667768474440749e-05, "val_loss": 5.558475971221924}
{"train_loss": 0.1464633345603943, "global_step": 124422, "epoch": 1398, "lr": 5.6677110200880044e-05}
{"train_loss": 0.11565355211496353, "global_step": 124423, "epoch": 1398, "lr": 5.667653565645495e-05}
{"train_loss": 0.10781816393136978, "global_step": 124424, "epoch": 1398, "lr": 5.667596111113228e-05}
{"train_loss": 0.12749072909355164, "global_step": 124425, "epoch": 1398, "lr": 5.6675386564912115e-05}
{"train_loss": 0.19050493836402893, "global_step": 124426, "epoch": 1398, "lr": 5.667481201779451e-05}
{"train_loss": 0.12743498384952545, "global_step": 124427, "epoch": 1398, "lr": 5.6674237469779565e-05}
{"train_loss": 0.12466475367546082, "global_step": 124428, "epoch": 1398, "lr": 5.6673662920867354e-05}
{"train_loss": 0.16890650987625122, "global_step": 124429, "epoch": 1398, "lr": 5.6673088371057946e-05}
{"train_loss": 0.08131663501262665, "global_step": 124430, "epoch": 1398, "lr": 5.667251382035142e-05}
{"train_loss": 0.09491200745105743, "global_step": 124431, "epoch": 1398, "lr": 5.667193926874786e-05}
{"train_loss": 0.255341500043869, "global_step": 124432, "epoch": 1398, "lr": 5.667136471624734e-05}
{"train_loss": 0.12131059914827347, "global_step": 124433, "epoch": 1398, "lr": 5.6670790162849943e-05}
{"train_loss": 0.10901772975921631, "global_step": 124434, "epoch": 1398, "lr": 5.667021560855572e-05}
{"train_loss": 0.16438709199428558, "global_step": 124435, "epoch": 1398, "lr": 5.6669641053364786e-05}
{"train_loss": 0.10791584849357605, "global_step": 124436, "epoch": 1398, "lr": 5.666906649727719e-05}
{"train_loss": 0.1139369085431099, "global_step": 124437, "epoch": 1398, "lr": 5.666849194029302e-05}
{"train_loss": 0.13158361613750458, "global_step": 124438, "epoch": 1398, "lr": 5.666791738241236e-05}
{"train_loss": 0.2173086553812027, "global_step": 124439, "epoch": 1398, "lr": 5.6667342823635275e-05}
{"train_loss": 0.1448439210653305, "global_step": 124440, "epoch": 1398, "lr": 5.666676826396184e-05}
{"train_loss": 0.14825810492038727, "global_step": 124441, "epoch": 1398, "lr": 5.6666193703392155e-05}
{"train_loss": 0.1554172784090042, "global_step": 124442, "epoch": 1398, "lr": 5.666561914192627e-05}
{"train_loss": 0.13687056303024292, "global_step": 124443, "epoch": 1398, "lr": 5.666504457956428e-05}
{"train_loss": 0.12329031527042389, "global_step": 124444, "epoch": 1398, "lr": 5.666447001630625e-05}
{"train_loss": 0.10866458714008331, "global_step": 124445, "epoch": 1398, "lr": 5.666389545215227e-05}
{"train_loss": 0.12155382335186005, "global_step": 124446, "epoch": 1398, "lr": 5.66633208871024e-05}
{"train_loss": 0.2016708254814148, "global_step": 124447, "epoch": 1398, "lr": 5.666274632115674e-05}
{"train_loss": 0.1082616075873375, "global_step": 124448, "epoch": 1398, "lr": 5.666217175431534e-05}
{"train_loss": 0.11149720847606659, "global_step": 124449, "epoch": 1398, "lr": 5.666159718657831e-05}
{"train_loss": 0.14646080136299133, "global_step": 124450, "epoch": 1398, "lr": 5.66610226179457e-05}
{"train_loss": 0.160490021109581, "global_step": 124451, "epoch": 1398, "lr": 5.666044804841759e-05}
{"train_loss": 0.11826308816671371, "global_step": 124452, "epoch": 1398, "lr": 5.665987347799408e-05}
{"train_loss": 0.1281207799911499, "global_step": 124453, "epoch": 1398, "lr": 5.6659298906675216e-05}
{"train_loss": 0.11086241155862808, "global_step": 124454, "epoch": 1398, "lr": 5.66587243344611e-05}
{"train_loss": 0.08150965720415115, "global_step": 124455, "epoch": 1398, "lr": 5.6658149761351797e-05}
{"train_loss": 0.13729159533977509, "global_step": 124456, "epoch": 1398, "lr": 5.665757518734739e-05}
{"train_loss": 0.14977990090847015, "global_step": 124457, "epoch": 1398, "lr": 5.665700061244794e-05}
{"train_loss": 0.1633700579404831, "global_step": 124458, "epoch": 1398, "lr": 5.6656426036653555e-05}
{"train_loss": 0.10862622410058975, "global_step": 124459, "epoch": 1398, "lr": 5.665585145996428e-05}
{"train_loss": 0.1852836012840271, "global_step": 124460, "epoch": 1398, "lr": 5.6655276882380215e-05}
{"train_loss": 0.11736178398132324, "global_step": 124461, "epoch": 1398, "lr": 5.665470230390143e-05}
{"train_loss": 0.16949932277202606, "global_step": 124462, "epoch": 1398, "lr": 5.6654127724528006e-05}
{"train_loss": 0.15850745141506195, "global_step": 124463, "epoch": 1398, "lr": 5.665355314426e-05}
{"train_loss": 0.11597598344087601, "global_step": 124464, "epoch": 1398, "lr": 5.6652978563097516e-05}
{"train_loss": 0.14112971723079681, "global_step": 124465, "epoch": 1398, "lr": 5.6652403981040615e-05}
{"train_loss": 0.13068228960037231, "global_step": 124466, "epoch": 1398, "lr": 5.665182939808938e-05}
{"train_loss": 0.13141930103302002, "global_step": 124467, "epoch": 1398, "lr": 5.6651254814243894e-05}
{"train_loss": 0.12254834920167923, "global_step": 124468, "epoch": 1398, "lr": 5.665068022950423e-05}
{"train_loss": 0.12162721902132034, "global_step": 124469, "epoch": 1398, "lr": 5.665010564387046e-05}
{"train_loss": 0.09420247375965118, "global_step": 124470, "epoch": 1398, "lr": 5.664953105734265e-05}
{"train_loss": 0.14928238093852997, "global_step": 124471, "epoch": 1398, "lr": 5.66489564699209e-05}
{"train_loss": 0.11438349634408951, "global_step": 124472, "epoch": 1398, "lr": 5.664838188160528e-05}
{"train_loss": 0.09499844163656235, "global_step": 124473, "epoch": 1398, "lr": 5.664780729239587e-05}
{"train_loss": 0.1908903568983078, "global_step": 124474, "epoch": 1398, "lr": 5.664723270229274e-05}
{"train_loss": 0.09590159356594086, "global_step": 124475, "epoch": 1398, "lr": 5.664665811129597e-05}
{"train_loss": 0.11886822432279587, "global_step": 124476, "epoch": 1398, "lr": 5.6646083519405645e-05}
{"train_loss": 0.13933251798152924, "global_step": 124477, "epoch": 1398, "lr": 5.6645508926621823e-05}
{"train_loss": 0.15884830057621002, "global_step": 124478, "epoch": 1398, "lr": 5.66449343329446e-05}
{"train_loss": 0.14685390889644623, "global_step": 124479, "epoch": 1398, "lr": 5.6644359738374054e-05}
{"train_loss": 0.07945001125335693, "global_step": 124480, "epoch": 1398, "lr": 5.664378514291025e-05}
{"train_loss": 0.1497882455587387, "global_step": 124481, "epoch": 1398, "lr": 5.664321054655326e-05}
{"train_loss": 0.13042885065078735, "global_step": 124482, "epoch": 1398, "lr": 5.664263594930318e-05}
{"train_loss": 0.13739266991615295, "global_step": 124483, "epoch": 1398, "lr": 5.664206135116008e-05}
{"train_loss": 0.19122953712940216, "global_step": 124484, "epoch": 1398, "lr": 5.664148675212403e-05}
{"train_loss": 0.10869357734918594, "global_step": 124485, "epoch": 1398, "lr": 5.664091215219511e-05}
{"train_loss": 0.16527390480041504, "global_step": 124486, "epoch": 1398, "lr": 5.6640337551373415e-05}
{"train_loss": 0.11484397202730179, "global_step": 124487, "epoch": 1398, "lr": 5.6639762949658994e-05}
{"train_loss": 0.12272495776414871, "global_step": 124488, "epoch": 1398, "lr": 5.663918834705194e-05}
{"train_loss": 0.13576549291610718, "global_step": 124489, "epoch": 1398, "lr": 5.663861374355234e-05}
{"train_loss": 0.08848985284566879, "global_step": 124490, "epoch": 1398, "lr": 5.663803913916025e-05}
{"train_loss": 0.07379822432994843, "global_step": 124491, "epoch": 1398, "lr": 5.663746453387575e-05}
{"train_loss": 0.19845892488956451, "global_step": 124492, "epoch": 1398, "lr": 5.6636889927698934e-05}
{"train_loss": 0.1038837879896164, "global_step": 124493, "epoch": 1398, "lr": 5.663631532062986e-05}
{"train_loss": 0.23223276436328888, "global_step": 124494, "epoch": 1398, "lr": 5.663574071266862e-05}
{"train_loss": 0.14571231603622437, "global_step": 124495, "epoch": 1398, "lr": 5.663516610381528e-05}
{"train_loss": 0.09674469381570816, "global_step": 124496, "epoch": 1398, "lr": 5.663459149406993e-05}
{"train_loss": 0.15780368447303772, "global_step": 124497, "epoch": 1398, "lr": 5.663401688343264e-05}
{"train_loss": 0.16374941170215607, "global_step": 124498, "epoch": 1398, "lr": 5.663344227190348e-05}
{"train_loss": 0.14430318772792816, "global_step": 124499, "epoch": 1398, "lr": 5.663286765948255e-05}
{"train_loss": 0.16212865710258484, "global_step": 124500, "epoch": 1398, "lr": 5.6632293046169907e-05}
{"train_loss": 0.18473775684833527, "global_step": 124501, "epoch": 1398, "lr": 5.663171843196562e-05}
{"train_loss": 0.07436425238847733, "global_step": 124502, "epoch": 1398, "lr": 5.6631143816869784e-05}
{"train_loss": 0.1164790689945221, "global_step": 124503, "epoch": 1398, "lr": 5.663056920088248e-05}
{"train_loss": 0.08081311732530594, "global_step": 124504, "epoch": 1398, "lr": 5.662999458400377e-05}
{"train_loss": 0.12410874664783478, "global_step": 124505, "epoch": 1398, "lr": 5.6629419966233735e-05}
{"train_loss": 0.12919244170188904, "global_step": 124506, "epoch": 1398, "lr": 5.6628845347572454e-05}
{"train_loss": 0.10833442211151123, "global_step": 124507, "epoch": 1398, "lr": 5.662827072802002e-05}
{"train_loss": 0.17915624380111694, "global_step": 124508, "epoch": 1398, "lr": 5.662769610757648e-05}
{"train_loss": 0.1091749370098114, "global_step": 124509, "epoch": 1398, "lr": 5.662712148624194e-05}
{"train_loss": 0.1352105529120799, "global_step": 124510, "epoch": 1398, "lr": 5.662654686401645e-05, "val_loss": 5.652052879333496}
{"train_loss": 0.08296773582696915, "global_step": 124511, "epoch": 1399, "lr": 5.6625972240900114e-05}
{"train_loss": 0.1443391740322113, "global_step": 124512, "epoch": 1399, "lr": 5.6625397616893e-05}
{"train_loss": 0.1564101278781891, "global_step": 124513, "epoch": 1399, "lr": 5.662482299199516e-05}
{"train_loss": 0.16671092808246613, "global_step": 124514, "epoch": 1399, "lr": 5.662424836620671e-05}
{"train_loss": 0.1613711714744568, "global_step": 124515, "epoch": 1399, "lr": 5.662367373952772e-05}
{"train_loss": 0.0953042209148407, "global_step": 124516, "epoch": 1399, "lr": 5.662309911195825e-05}
{"train_loss": 0.1377297192811966, "global_step": 124517, "epoch": 1399, "lr": 5.662252448349836e-05}
{"train_loss": 0.10610831528902054, "global_step": 124518, "epoch": 1399, "lr": 5.662194985414818e-05}
{"train_loss": 0.10065404325723648, "global_step": 124519, "epoch": 1399, "lr": 5.662137522390776e-05}
{"train_loss": 0.1499457210302353, "global_step": 124520, "epoch": 1399, "lr": 5.662080059277717e-05}
{"train_loss": 0.14858177304267883, "global_step": 124521, "epoch": 1399, "lr": 5.662022596075649e-05}
{"train_loss": 0.1468980312347412, "global_step": 124522, "epoch": 1399, "lr": 5.661965132784581e-05}
{"train_loss": 0.1065213680267334, "global_step": 124523, "epoch": 1399, "lr": 5.661907669404519e-05}
{"train_loss": 0.10333548486232758, "global_step": 124524, "epoch": 1399, "lr": 5.6618502059354715e-05}
{"train_loss": 0.16756615042686462, "global_step": 124525, "epoch": 1399, "lr": 5.661792742377447e-05}
{"train_loss": 0.1454029083251953, "global_step": 124526, "epoch": 1399, "lr": 5.661735278730452e-05}
{"train_loss": 0.20505109429359436, "global_step": 124527, "epoch": 1399, "lr": 5.661677814994496e-05}
{"train_loss": 0.11862996220588684, "global_step": 124528, "epoch": 1399, "lr": 5.661620351169584e-05}
{"train_loss": 0.1095634400844574, "global_step": 124529, "epoch": 1399, "lr": 5.661562887255726e-05}
{"train_loss": 0.15712273120880127, "global_step": 124530, "epoch": 1399, "lr": 5.661505423252929e-05}
{"train_loss": 0.16499340534210205, "global_step": 124531, "epoch": 1399, "lr": 5.6614479591612e-05}
{"train_loss": 0.10417275875806808, "global_step": 124532, "epoch": 1399, "lr": 5.6613904949805475e-05}
{"train_loss": 0.10312600433826447, "global_step": 124533, "epoch": 1399, "lr": 5.661333030710979e-05}
{"train_loss": 0.16277599334716797, "global_step": 124534, "epoch": 1399, "lr": 5.661275566352503e-05}
{"train_loss": 0.2117447853088379, "global_step": 124535, "epoch": 1399, "lr": 5.6612181019051256e-05}
{"train_loss": 0.1330399066209793, "global_step": 124536, "epoch": 1399, "lr": 5.6611606373688556e-05}
{"train_loss": 0.13794207572937012, "global_step": 124537, "epoch": 1399, "lr": 5.661103172743703e-05}
{"train_loss": 0.1602899581193924, "global_step": 124538, "epoch": 1399, "lr": 5.661045708029671e-05}
{"train_loss": 0.15006530284881592, "global_step": 124539, "epoch": 1399, "lr": 5.660988243226769e-05}
{"train_loss": 0.12531517446041107, "global_step": 124540, "epoch": 1399, "lr": 5.660930778335006e-05}
{"train_loss": 0.12396819144487381, "global_step": 124541, "epoch": 1399, "lr": 5.660873313354389e-05}
{"train_loss": 0.055189915001392365, "global_step": 124542, "epoch": 1399, "lr": 5.660815848284925e-05}
{"train_loss": 0.15616494417190552, "global_step": 124543, "epoch": 1399, "lr": 5.660758383126623e-05}
{"train_loss": 0.12524719536304474, "global_step": 124544, "epoch": 1399, "lr": 5.66070091787949e-05}
{"train_loss": 0.0942705050110817, "global_step": 124545, "epoch": 1399, "lr": 5.6606434525435334e-05}
{"train_loss": 0.1793593019247055, "global_step": 124546, "epoch": 1399, "lr": 5.660585987118762e-05}
{"train_loss": 0.1500738263130188, "global_step": 124547, "epoch": 1399, "lr": 5.660528521605183e-05}
{"train_loss": 0.11197453737258911, "global_step": 124548, "epoch": 1399, "lr": 5.660471056002803e-05}
{"train_loss": 0.07918909192085266, "global_step": 124549, "epoch": 1399, "lr": 5.660413590311632e-05}
{"train_loss": 0.14071838557720184, "global_step": 124550, "epoch": 1399, "lr": 5.660356124531675e-05}
{"train_loss": 0.17962490022182465, "global_step": 124551, "epoch": 1399, "lr": 5.6602986586629413e-05}
{"train_loss": 0.1612827032804489, "global_step": 124552, "epoch": 1399, "lr": 5.66024119270544e-05}
{"train_loss": 0.189758762717247, "global_step": 124553, "epoch": 1399, "lr": 5.6601837266591754e-05}
{"train_loss": 0.17159907519817352, "global_step": 124554, "epoch": 1399, "lr": 5.660126260524158e-05}
{"train_loss": 0.15257470309734344, "global_step": 124555, "epoch": 1399, "lr": 5.660068794300395e-05}
{"train_loss": 0.12085795402526855, "global_step": 124556, "epoch": 1399, "lr": 5.660011327987894e-05}
{"train_loss": 0.16488724946975708, "global_step": 124557, "epoch": 1399, "lr": 5.659953861586662e-05}
{"train_loss": 0.09604190289974213, "global_step": 124558, "epoch": 1399, "lr": 5.659896395096708e-05}
{"train_loss": 0.0920746773481369, "global_step": 124559, "epoch": 1399, "lr": 5.6598389285180374e-05}
{"train_loss": 0.09573428332805634, "global_step": 124560, "epoch": 1399, "lr": 5.65978146185066e-05}
{"train_loss": 0.11195296794176102, "global_step": 124561, "epoch": 1399, "lr": 5.6597239950945846e-05}
{"train_loss": 0.16315585374832153, "global_step": 124562, "epoch": 1399, "lr": 5.6596665282498164e-05}
{"train_loss": 0.09622502326965332, "global_step": 124563, "epoch": 1399, "lr": 5.659609061316363e-05}
{"train_loss": 0.13400262594223022, "global_step": 124564, "epoch": 1399, "lr": 5.6595515942942344e-05}
{"train_loss": 0.07529285550117493, "global_step": 124565, "epoch": 1399, "lr": 5.659494127183437e-05}
{"train_loss": 0.13413752615451813, "global_step": 124566, "epoch": 1399, "lr": 5.659436659983977e-05}
{"train_loss": 0.11719931662082672, "global_step": 124567, "epoch": 1399, "lr": 5.659379192695866e-05}
{"train_loss": 0.09576336294412613, "global_step": 124568, "epoch": 1399, "lr": 5.659321725319109e-05}
{"train_loss": 0.12496225535869598, "global_step": 124569, "epoch": 1399, "lr": 5.6592642578537136e-05}
{"train_loss": 0.13715806603431702, "global_step": 124570, "epoch": 1399, "lr": 5.6592067902996895e-05}
{"train_loss": 0.15874451398849487, "global_step": 124571, "epoch": 1399, "lr": 5.659149322657041e-05}
{"train_loss": 0.06166666001081467, "global_step": 124572, "epoch": 1399, "lr": 5.659091854925778e-05}
{"train_loss": 0.14602236449718475, "global_step": 124573, "epoch": 1399, "lr": 5.6590343871059095e-05}
{"train_loss": 0.13531939685344696, "global_step": 124574, "epoch": 1399, "lr": 5.658976919197442e-05}
{"train_loss": 0.16218678653240204, "global_step": 124575, "epoch": 1399, "lr": 5.658919451200382e-05}
{"train_loss": 0.1886037439107895, "global_step": 124576, "epoch": 1399, "lr": 5.658861983114741e-05}
{"train_loss": 0.17201241850852966, "global_step": 124577, "epoch": 1399, "lr": 5.658804514940521e-05}
{"train_loss": 0.1501149833202362, "global_step": 124578, "epoch": 1399, "lr": 5.658747046677734e-05}
{"train_loss": 0.11950551718473434, "global_step": 124579, "epoch": 1399, "lr": 5.658689578326386e-05}
{"train_loss": 0.15679161250591278, "global_step": 124580, "epoch": 1399, "lr": 5.658632109886485e-05}
{"train_loss": 0.1902560591697693, "global_step": 124581, "epoch": 1399, "lr": 5.65857464135804e-05}
{"train_loss": 0.11315155774354935, "global_step": 124582, "epoch": 1399, "lr": 5.658517172741056e-05}
{"train_loss": 0.15627406537532806, "global_step": 124583, "epoch": 1399, "lr": 5.6584597040355433e-05}
{"train_loss": 0.1947077512741089, "global_step": 124584, "epoch": 1399, "lr": 5.658402235241509e-05}
{"train_loss": 0.2149921953678131, "global_step": 124585, "epoch": 1399, "lr": 5.6583447663589605e-05}
{"train_loss": 0.09642106294631958, "global_step": 124586, "epoch": 1399, "lr": 5.658287297387905e-05}
{"train_loss": 0.07529390603303909, "global_step": 124587, "epoch": 1399, "lr": 5.658229828328352e-05}
{"train_loss": 0.2502098083496094, "global_step": 124588, "epoch": 1399, "lr": 5.658172359180307e-05}
{"train_loss": 0.12424073368310928, "global_step": 124589, "epoch": 1399, "lr": 5.658114889943779e-05}
{"train_loss": 0.18265226483345032, "global_step": 124590, "epoch": 1399, "lr": 5.658057420618774e-05}
{"train_loss": 0.12723645567893982, "global_step": 124591, "epoch": 1399, "lr": 5.657999951205303e-05}
{"train_loss": 0.09970581531524658, "global_step": 124592, "epoch": 1399, "lr": 5.657942481703371e-05}
{"train_loss": 0.13877423107624054, "global_step": 124593, "epoch": 1399, "lr": 5.657885012112988e-05}
{"train_loss": 0.2266031950712204, "global_step": 124594, "epoch": 1399, "lr": 5.65782754243416e-05}
{"train_loss": 0.15839126706123352, "global_step": 124595, "epoch": 1399, "lr": 5.657770072666895e-05}
{"train_loss": 0.10493925213813782, "global_step": 124596, "epoch": 1399, "lr": 5.6577126028112e-05}
{"train_loss": 0.2720664441585541, "global_step": 124597, "epoch": 1399, "lr": 5.657655132867084e-05}
{"train_loss": 0.12846162915229797, "global_step": 124598, "epoch": 1399, "lr": 5.657597662834555e-05}
{"train_loss": 0.1406574264336168, "global_step": 124599, "epoch": 1399, "lr": 5.65754019271362e-05, "val_loss": 5.368959903717041}
{"train_loss": 0.14948879182338715, "global_step": 124600, "epoch": 1400, "lr": 5.657482722504286e-05}
{"train_loss": 0.14622405171394348, "global_step": 124601, "epoch": 1400, "lr": 5.6574252522065606e-05}
{"train_loss": 0.09265992045402527, "global_step": 124602, "epoch": 1400, "lr": 5.657367781820454e-05}
{"train_loss": 0.16517497599124908, "global_step": 124603, "epoch": 1400, "lr": 5.6573103113459715e-05}
{"train_loss": 0.07991485297679901, "global_step": 124604, "epoch": 1400, "lr": 5.657252840783123e-05}
{"train_loss": 0.11891192942857742, "global_step": 124605, "epoch": 1400, "lr": 5.657195370131913e-05}
{"train_loss": 0.08900368958711624, "global_step": 124606, "epoch": 1400, "lr": 5.6571378993923515e-05}
{"train_loss": 0.16206172108650208, "global_step": 124607, "epoch": 1400, "lr": 5.657080428564446e-05}
{"train_loss": 0.1617366075515747, "global_step": 124608, "epoch": 1400, "lr": 5.6570229576482034e-05}
{"train_loss": 0.12569481134414673, "global_step": 124609, "epoch": 1400, "lr": 5.6569654866436326e-05}
{"train_loss": 0.13843044638633728, "global_step": 124610, "epoch": 1400, "lr": 5.6569080155507415e-05}
{"train_loss": 0.13687127828598022, "global_step": 124611, "epoch": 1400, "lr": 5.6568505443695366e-05}
{"train_loss": 0.1299828439950943, "global_step": 124612, "epoch": 1400, "lr": 5.656793073100026e-05}
{"train_loss": 0.16450314223766327, "global_step": 124613, "epoch": 1400, "lr": 5.6567356017422176e-05}
{"train_loss": 0.13428837060928345, "global_step": 124614, "epoch": 1400, "lr": 5.6566781302961205e-05}
{"train_loss": 0.09300434589385986, "global_step": 124615, "epoch": 1400, "lr": 5.65662065876174e-05}
{"train_loss": 0.1452547162771225, "global_step": 124616, "epoch": 1400, "lr": 5.656563187139085e-05}
{"train_loss": 0.10915937274694443, "global_step": 124617, "epoch": 1400, "lr": 5.656505715428162e-05}
{"train_loss": 0.16950972378253937, "global_step": 124618, "epoch": 1400, "lr": 5.656448243628981e-05}
{"train_loss": 0.16883084177970886, "global_step": 124619, "epoch": 1400, "lr": 5.6563907717415474e-05}
{"train_loss": 0.12189514935016632, "global_step": 124620, "epoch": 1400, "lr": 5.656333299765871e-05}
{"train_loss": 0.10965681821107864, "global_step": 124621, "epoch": 1400, "lr": 5.656275827701959e-05}
{"train_loss": 0.09742411971092224, "global_step": 124622, "epoch": 1400, "lr": 5.656218355549817e-05}
{"train_loss": 0.10756082087755203, "global_step": 124623, "epoch": 1400, "lr": 5.656160883309456e-05}
{"train_loss": 0.12484720349311829, "global_step": 124624, "epoch": 1400, "lr": 5.656103410980882e-05}
{"train_loss": 0.16476035118103027, "global_step": 124625, "epoch": 1400, "lr": 5.656045938564102e-05}
{"train_loss": 0.10853248834609985, "global_step": 124626, "epoch": 1400, "lr": 5.655988466059125e-05}
{"train_loss": 0.07194861024618149, "global_step": 124627, "epoch": 1400, "lr": 5.655930993465958e-05}
{"train_loss": 0.12950071692466736, "global_step": 124628, "epoch": 1400, "lr": 5.655873520784609e-05}
{"train_loss": 0.12416468560695648, "global_step": 124629, "epoch": 1400, "lr": 5.6558160480150866e-05}
{"train_loss": 0.13337425887584686, "global_step": 124630, "epoch": 1400, "lr": 5.655758575157397e-05}
{"train_loss": 0.05932176485657692, "global_step": 124631, "epoch": 1400, "lr": 5.65570110221155e-05}
{"train_loss": 0.14697235822677612, "global_step": 124632, "epoch": 1400, "lr": 5.655643629177552e-05}
{"train_loss": 0.08653046190738678, "global_step": 124633, "epoch": 1400, "lr": 5.655586156055409e-05}
{"train_loss": 0.14367926120758057, "global_step": 124634, "epoch": 1400, "lr": 5.655528682845133e-05}
{"train_loss": 0.10868500173091888, "global_step": 124635, "epoch": 1400, "lr": 5.655471209546726e-05}
{"train_loss": 0.09947056323289871, "global_step": 124636, "epoch": 1400, "lr": 5.655413736160201e-05}
{"train_loss": 0.08006462454795837, "global_step": 124637, "epoch": 1400, "lr": 5.6553562626855626e-05}
{"train_loss": 0.0923304334282875, "global_step": 124638, "epoch": 1400, "lr": 5.65529878912282e-05}
{"train_loss": 0.14719943702220917, "global_step": 124639, "epoch": 1400, "lr": 5.655241315471981e-05}
{"train_loss": 0.11352334916591644, "global_step": 124640, "epoch": 1400, "lr": 5.655183841733052e-05}
{"train_loss": 0.14732910692691803, "global_step": 124641, "epoch": 1400, "lr": 5.655126367906042e-05}
{"train_loss": 0.0900026261806488, "global_step": 124642, "epoch": 1400, "lr": 5.655068893990957e-05}
{"train_loss": 0.16083909571170807, "global_step": 124643, "epoch": 1400, "lr": 5.655011419987807e-05}
{"train_loss": 0.1019665077328682, "global_step": 124644, "epoch": 1400, "lr": 5.654953945896597e-05}
{"train_loss": 0.13143739104270935, "global_step": 124645, "epoch": 1400, "lr": 5.654896471717339e-05}
{"train_loss": 0.1364610195159912, "global_step": 124646, "epoch": 1400, "lr": 5.654838997450037e-05}
{"train_loss": 0.1878960281610489, "global_step": 124647, "epoch": 1400, "lr": 5.6547815230947e-05}
{"train_loss": 0.10195377469062805, "global_step": 124648, "epoch": 1400, "lr": 5.654724048651335e-05}
{"train_loss": 0.10813282430171967, "global_step": 124649, "epoch": 1400, "lr": 5.654666574119951e-05}
{"train_loss": 0.12228000909090042, "global_step": 124650, "epoch": 1400, "lr": 5.6546090995005554e-05}
{"train_loss": 0.17648586630821228, "global_step": 124651, "epoch": 1400, "lr": 5.654551624793155e-05}
{"train_loss": 0.1420736312866211, "global_step": 124652, "epoch": 1400, "lr": 5.654494149997759e-05}
{"train_loss": 0.1299685686826706, "global_step": 124653, "epoch": 1400, "lr": 5.654436675114374e-05}
{"train_loss": 0.1094200536608696, "global_step": 124654, "epoch": 1400, "lr": 5.654379200143007e-05}
{"train_loss": 0.11182061582803726, "global_step": 124655, "epoch": 1400, "lr": 5.6543217250836675e-05}
{"train_loss": 0.17919695377349854, "global_step": 124656, "epoch": 1400, "lr": 5.654264249936363e-05}
{"train_loss": 0.10417753458023071, "global_step": 124657, "epoch": 1400, "lr": 5.6542067747011e-05}
{"train_loss": 0.09905766695737839, "global_step": 124658, "epoch": 1400, "lr": 5.6541492993778856e-05}
{"train_loss": 0.12878157198429108, "global_step": 124659, "epoch": 1400, "lr": 5.65409182396673e-05}
{"train_loss": 0.10850030183792114, "global_step": 124660, "epoch": 1400, "lr": 5.6540343484676395e-05}
{"train_loss": 0.12732283771038055, "global_step": 124661, "epoch": 1400, "lr": 5.653976872880622e-05}
{"train_loss": 0.1056605651974678, "global_step": 124662, "epoch": 1400, "lr": 5.653919397205685e-05}
{"train_loss": 0.14988212287425995, "global_step": 124663, "epoch": 1400, "lr": 5.653861921442837e-05}
{"train_loss": 0.15356417000293732, "global_step": 124664, "epoch": 1400, "lr": 5.6538044455920844e-05}
{"train_loss": 0.10192683339118958, "global_step": 124665, "epoch": 1400, "lr": 5.653746969653437e-05}
{"train_loss": 0.1137186661362648, "global_step": 124666, "epoch": 1400, "lr": 5.6536894936269e-05}
{"train_loss": 0.09052417427301407, "global_step": 124667, "epoch": 1400, "lr": 5.653632017512483e-05}
{"train_loss": 0.08028334379196167, "global_step": 124668, "epoch": 1400, "lr": 5.653574541310194e-05}
{"train_loss": 0.10544289648532867, "global_step": 124669, "epoch": 1400, "lr": 5.65351706502004e-05}
{"train_loss": 0.1566389501094818, "global_step": 124670, "epoch": 1400, "lr": 5.653459588642027e-05}
{"train_loss": 0.09697379916906357, "global_step": 124671, "epoch": 1400, "lr": 5.653402112176165e-05}
{"train_loss": 0.07212797552347183, "global_step": 124672, "epoch": 1400, "lr": 5.653344635622463e-05}
{"train_loss": 0.18730847537517548, "global_step": 124673, "epoch": 1400, "lr": 5.6532871589809244e-05}
{"train_loss": 0.096409372985363, "global_step": 124674, "epoch": 1400, "lr": 5.6532296822515595e-05}
{"train_loss": 0.08605507016181946, "global_step": 124675, "epoch": 1400, "lr": 5.653172205434377e-05}
{"train_loss": 0.11550768464803696, "global_step": 124676, "epoch": 1400, "lr": 5.653114728529383e-05}
{"train_loss": 0.12270715087652206, "global_step": 124677, "epoch": 1400, "lr": 5.653057251536584e-05}
{"train_loss": 0.1550443172454834, "global_step": 124678, "epoch": 1400, "lr": 5.652999774455991e-05}
{"train_loss": 0.08135826885700226, "global_step": 124679, "epoch": 1400, "lr": 5.65294229728761e-05}
{"train_loss": 0.18470297753810883, "global_step": 124680, "epoch": 1400, "lr": 5.6528848200314474e-05}
{"train_loss": 0.08756779134273529, "global_step": 124681, "epoch": 1400, "lr": 5.652827342687515e-05}
{"train_loss": 0.17080916464328766, "global_step": 124682, "epoch": 1400, "lr": 5.652769865255816e-05}
{"train_loss": 0.14109134674072266, "global_step": 124683, "epoch": 1400, "lr": 5.652712387736361e-05}
{"train_loss": 0.0980541855096817, "global_step": 124684, "epoch": 1400, "lr": 5.6526549101291556e-05}
{"train_loss": 0.13737189769744873, "global_step": 124685, "epoch": 1400, "lr": 5.65259743243421e-05}
{"train_loss": 0.11319973319768906, "global_step": 124686, "epoch": 1400, "lr": 5.6525399546515303e-05}
{"train_loss": 0.12995697557926178, "global_step": 124687, "epoch": 1400, "lr": 5.652482476781125e-05}
{"train_loss": 0.12300884522748796, "global_step": 124688, "epoch": 1400, "lr": 5.652424998823e-05, "train/sim_max_reward_0": 0.5587320097860378, "train/sim_max_reward_1": 0.5408635011640792, "train/sim_max_reward_2": 0.6510291973190911, "train/sim_max_reward_3": 0.40130454671293236, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9522994031836177, "test/sim_max_reward_4500000": 0.45946868459634316, "test/sim_max_reward_4500001": 0.9722444967177378, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.37837115657506026, "test/sim_max_reward_4500004": 0.5246327021701495, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.9824047998568898, "test/sim_max_reward_4500007": 0.9769165499946927, "test/sim_max_reward_4500008": 0.1598760403301653, "test/sim_max_reward_4500009": 0.05056182158312409, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9802487619652922, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.6669090077815429, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9169600479319834, "test/sim_max_reward_4500016": 1.0, "test/sim_max_reward_4500017": 0.5358955494220053, "test/sim_max_reward_4500018": 0.9848976631998322, "test/sim_max_reward_4500019": 0.9339117725061905, "test/sim_max_reward_4500020": 0.5966265032105327, "test/sim_max_reward_4500021": 0.9909974503987125, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6993214411393424, "test/sim_max_reward_4500024": 0.508594932952002, "test/sim_max_reward_4500025": 0.8967311424542604, "test/sim_max_reward_4500026": 0.9707776521195428, "test/sim_max_reward_4500027": 0.9634540758564281, "test/sim_max_reward_4500028": 1.0, "test/sim_max_reward_4500029": 0.06741864442302238, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.5048209620694144, "test/sim_max_reward_4500032": 0.7252977301372711, "test/sim_max_reward_4500033": 0.9409286149950459, "test/sim_max_reward_4500034": 0.93543877276815, "test/sim_max_reward_4500035": 0.920715023370917, "test/sim_max_reward_4500036": 0.1589943047833549, "test/sim_max_reward_4500037": 0.9534200813451587, "test/sim_max_reward_4500038": 0.7885563004046132, "test/sim_max_reward_4500039": 0.9779607649947315, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.5104824508722183, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9265213671893981, "test/sim_max_reward_4500044": 0.02030809375640375, "test/sim_max_reward_4500045": 0.945812989350023, "test/sim_max_reward_4500046": 0.669356836505904, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9470449836914961, "test/sim_max_reward_4500049": 0.9269512918708543, "train/mean_score": 0.684038109694293, "test/mean_score": 0.7013966293057962, "val_loss": 5.482161521911621, "train_action_mse_error": 22.868303298950195}
{"train_loss": 0.10063544660806656, "global_step": 124689, "epoch": 1401, "lr": 5.652367520777167e-05}
{"train_loss": 0.11139777302742004, "global_step": 124690, "epoch": 1401, "lr": 5.652310042643628e-05}
{"train_loss": 0.10132373124361038, "global_step": 124691, "epoch": 1401, "lr": 5.6522525644223954e-05}
{"train_loss": 0.1012313961982727, "global_step": 124692, "epoch": 1401, "lr": 5.652195086113477e-05}
{"train_loss": 0.17813077569007874, "global_step": 124693, "epoch": 1401, "lr": 5.6521376077168766e-05}
{"train_loss": 0.15621110796928406, "global_step": 124694, "epoch": 1401, "lr": 5.652080129232605e-05}
{"train_loss": 0.15692205727100372, "global_step": 124695, "epoch": 1401, "lr": 5.652022650660669e-05}
{"train_loss": 0.09382221847772598, "global_step": 124696, "epoch": 1401, "lr": 5.651965172001077e-05}
{"train_loss": 0.1693263053894043, "global_step": 124697, "epoch": 1401, "lr": 5.651907693253835e-05}
{"train_loss": 0.09598229080438614, "global_step": 124698, "epoch": 1401, "lr": 5.6518502144189534e-05}
{"train_loss": 0.209312304854393, "global_step": 124699, "epoch": 1401, "lr": 5.651792735496436e-05}
{"train_loss": 0.09082429856061935, "global_step": 124700, "epoch": 1401, "lr": 5.6517352564862946e-05}
{"train_loss": 0.13487981259822845, "global_step": 124701, "epoch": 1401, "lr": 5.6516777773885355e-05}
{"train_loss": 0.16356931626796722, "global_step": 124702, "epoch": 1401, "lr": 5.651620298203166e-05}
{"train_loss": 0.14063838124275208, "global_step": 124703, "epoch": 1401, "lr": 5.651562818930195e-05}
{"train_loss": 0.07626403868198395, "global_step": 124704, "epoch": 1401, "lr": 5.6515053395696295e-05}
{"train_loss": 0.22531449794769287, "global_step": 124705, "epoch": 1401, "lr": 5.651447860121476e-05}
{"train_loss": 0.16971494257450104, "global_step": 124706, "epoch": 1401, "lr": 5.651390380585744e-05}
{"train_loss": 0.08666839450597763, "global_step": 124707, "epoch": 1401, "lr": 5.6513329009624394e-05}
{"train_loss": 0.16760389506816864, "global_step": 124708, "epoch": 1401, "lr": 5.6512754212515714e-05}
{"train_loss": 0.07050621509552002, "global_step": 124709, "epoch": 1401, "lr": 5.651217941453147e-05}
{"train_loss": 0.17691273987293243, "global_step": 124710, "epoch": 1401, "lr": 5.651160461567175e-05}
{"train_loss": 0.04237213730812073, "global_step": 124711, "epoch": 1401, "lr": 5.651102981593665e-05}
{"train_loss": 0.08312385529279709, "global_step": 124712, "epoch": 1401, "lr": 5.651045501532619e-05}
{"train_loss": 0.19612672924995422, "global_step": 124713, "epoch": 1401, "lr": 5.6509880213840485e-05}
{"train_loss": 0.0718228667974472, "global_step": 124714, "epoch": 1401, "lr": 5.65093054114796e-05}
{"train_loss": 0.1171235591173172, "global_step": 124715, "epoch": 1401, "lr": 5.650873060824362e-05}
{"train_loss": 0.13330213725566864, "global_step": 124716, "epoch": 1401, "lr": 5.6508155804132625e-05}
{"train_loss": 0.1480729728937149, "global_step": 124717, "epoch": 1401, "lr": 5.650758099914667e-05}
{"train_loss": 0.08872527629137039, "global_step": 124718, "epoch": 1401, "lr": 5.6507006193285874e-05}
{"train_loss": 0.17327815294265747, "global_step": 124719, "epoch": 1401, "lr": 5.650643138655027e-05}
{"train_loss": 0.14287397265434265, "global_step": 124720, "epoch": 1401, "lr": 5.6505856578939976e-05}
{"train_loss": 0.10923817753791809, "global_step": 124721, "epoch": 1401, "lr": 5.650528177045504e-05}
{"train_loss": 0.13143320381641388, "global_step": 124722, "epoch": 1401, "lr": 5.6504706961095554e-05}
{"train_loss": 0.1237112507224083, "global_step": 124723, "epoch": 1401, "lr": 5.6504132150861576e-05}
{"train_loss": 0.18298457562923431, "global_step": 124724, "epoch": 1401, "lr": 5.650355733975321e-05}
{"train_loss": 0.09470120817422867, "global_step": 124725, "epoch": 1401, "lr": 5.650298252777051e-05}
{"train_loss": 0.13807480037212372, "global_step": 124726, "epoch": 1401, "lr": 5.650240771491358e-05}
{"train_loss": 0.17725680768489838, "global_step": 124727, "epoch": 1401, "lr": 5.650183290118246e-05}
{"train_loss": 0.1262052357196808, "global_step": 124728, "epoch": 1401, "lr": 5.6501258086577256e-05}
{"train_loss": 0.21574443578720093, "global_step": 124729, "epoch": 1401, "lr": 5.6500683271098044e-05}
{"train_loss": 0.09942533075809479, "global_step": 124730, "epoch": 1401, "lr": 5.6500108454744884e-05}
{"train_loss": 0.2113056480884552, "global_step": 124731, "epoch": 1401, "lr": 5.649953363751788e-05}
{"train_loss": 0.22961241006851196, "global_step": 124732, "epoch": 1401, "lr": 5.649895881941708e-05}
{"train_loss": 0.08473777770996094, "global_step": 124733, "epoch": 1401, "lr": 5.6498384000442575e-05}
{"train_loss": 0.12370829284191132, "global_step": 124734, "epoch": 1401, "lr": 5.649780918059444e-05}
{"train_loss": 0.10142425447702408, "global_step": 124735, "epoch": 1401, "lr": 5.649723435987274e-05}
{"train_loss": 0.13879507780075073, "global_step": 124736, "epoch": 1401, "lr": 5.6496659538277585e-05}
{"train_loss": 0.12661349773406982, "global_step": 124737, "epoch": 1401, "lr": 5.649608471580903e-05}
{"train_loss": 0.13255690038204193, "global_step": 124738, "epoch": 1401, "lr": 5.649550989246716e-05}
{"train_loss": 0.20102939009666443, "global_step": 124739, "epoch": 1401, "lr": 5.6494935068252034e-05}
{"train_loss": 0.17239254713058472, "global_step": 124740, "epoch": 1401, "lr": 5.6494360243163755e-05}
{"train_loss": 0.14423242211341858, "global_step": 124741, "epoch": 1401, "lr": 5.649378541720238e-05}
{"train_loss": 0.13460785150527954, "global_step": 124742, "epoch": 1401, "lr": 5.649321059036801e-05}
{"train_loss": 0.09469576925039291, "global_step": 124743, "epoch": 1401, "lr": 5.64926357626607e-05}
{"train_loss": 0.10816963016986847, "global_step": 124744, "epoch": 1401, "lr": 5.649206093408053e-05}
{"train_loss": 0.10214139521121979, "global_step": 124745, "epoch": 1401, "lr": 5.649148610462758e-05}
{"train_loss": 0.15021584928035736, "global_step": 124746, "epoch": 1401, "lr": 5.6490911274301925e-05}
{"train_loss": 0.18380051851272583, "global_step": 124747, "epoch": 1401, "lr": 5.649033644310365e-05}
{"train_loss": 0.08812089264392853, "global_step": 124748, "epoch": 1401, "lr": 5.648976161103283e-05}
{"train_loss": 0.2983700931072235, "global_step": 124749, "epoch": 1401, "lr": 5.6489186778089544e-05}
{"train_loss": 0.1323307454586029, "global_step": 124750, "epoch": 1401, "lr": 5.6488611944273885e-05}
{"train_loss": 0.11720813810825348, "global_step": 124751, "epoch": 1401, "lr": 5.648803710958588e-05}
{"train_loss": 0.14490976929664612, "global_step": 124752, "epoch": 1401, "lr": 5.648746227402565e-05}
{"train_loss": 0.0923488587141037, "global_step": 124753, "epoch": 1401, "lr": 5.648688743759325e-05}
{"train_loss": 0.130569726228714, "global_step": 124754, "epoch": 1401, "lr": 5.648631260028877e-05}
{"train_loss": 0.13271315395832062, "global_step": 124755, "epoch": 1401, "lr": 5.648573776211229e-05}
{"train_loss": 0.17165257036685944, "global_step": 124756, "epoch": 1401, "lr": 5.648516292306388e-05}
{"train_loss": 0.12231308221817017, "global_step": 124757, "epoch": 1401, "lr": 5.648458808314362e-05}
{"train_loss": 0.18765144050121307, "global_step": 124758, "epoch": 1401, "lr": 5.648401324235159e-05}
{"train_loss": 0.1610831320285797, "global_step": 124759, "epoch": 1401, "lr": 5.648343840068785e-05}
{"train_loss": 0.11911971867084503, "global_step": 124760, "epoch": 1401, "lr": 5.6482863558152496e-05}
{"train_loss": 0.11555325984954834, "global_step": 124761, "epoch": 1401, "lr": 5.648228871474561e-05}
{"train_loss": 0.12890999019145966, "global_step": 124762, "epoch": 1401, "lr": 5.648171387046726e-05}
{"train_loss": 0.15499553084373474, "global_step": 124763, "epoch": 1401, "lr": 5.648113902531751e-05}
{"train_loss": 0.08023900538682938, "global_step": 124764, "epoch": 1401, "lr": 5.648056417929645e-05}
{"train_loss": 0.21577313542366028, "global_step": 124765, "epoch": 1401, "lr": 5.647998933240416e-05}
{"train_loss": 0.19432806968688965, "global_step": 124766, "epoch": 1401, "lr": 5.647941448464073e-05}
{"train_loss": 0.14463184773921967, "global_step": 124767, "epoch": 1401, "lr": 5.64788396360062e-05}
{"train_loss": 0.1929704248905182, "global_step": 124768, "epoch": 1401, "lr": 5.647826478650068e-05}
{"train_loss": 0.15062320232391357, "global_step": 124769, "epoch": 1401, "lr": 5.6477689936124245e-05}
{"train_loss": 0.12491367012262344, "global_step": 124770, "epoch": 1401, "lr": 5.647711508487694e-05}
{"train_loss": 0.09935534000396729, "global_step": 124771, "epoch": 1401, "lr": 5.647654023275888e-05}
{"train_loss": 0.11764931678771973, "global_step": 124772, "epoch": 1401, "lr": 5.6475965379770126e-05}
{"train_loss": 0.13376757502555847, "global_step": 124773, "epoch": 1401, "lr": 5.6475390525910754e-05}
{"train_loss": 0.16808973252773285, "global_step": 124774, "epoch": 1401, "lr": 5.647481567118085e-05}
{"train_loss": 0.11404290795326233, "global_step": 124775, "epoch": 1401, "lr": 5.6474240815580484e-05}
{"train_loss": 0.14974789321422577, "global_step": 124776, "epoch": 1401, "lr": 5.6473665959109737e-05}
{"train_loss": 0.13871483956829886, "global_step": 124777, "epoch": 1401, "lr": 5.647309110176868e-05, "val_loss": 5.264760494232178}
{"train_loss": 0.20576734840869904, "global_step": 124778, "epoch": 1402, "lr": 5.6472516243557394e-05}
{"train_loss": 0.13324858248233795, "global_step": 124779, "epoch": 1402, "lr": 5.6471941384475955e-05}
{"train_loss": 0.14526841044425964, "global_step": 124780, "epoch": 1402, "lr": 5.647136652452445e-05}
{"train_loss": 0.14563025534152985, "global_step": 124781, "epoch": 1402, "lr": 5.647079166370295e-05}
{"train_loss": 0.06196245178580284, "global_step": 124782, "epoch": 1402, "lr": 5.647021680201153e-05}
{"train_loss": 0.13105273246765137, "global_step": 124783, "epoch": 1402, "lr": 5.646964193945027e-05}
{"train_loss": 0.16085346043109894, "global_step": 124784, "epoch": 1402, "lr": 5.6469067076019236e-05}
{"train_loss": 0.11462952196598053, "global_step": 124785, "epoch": 1402, "lr": 5.6468492211718516e-05}
{"train_loss": 0.10091184824705124, "global_step": 124786, "epoch": 1402, "lr": 5.646791734654819e-05}
{"train_loss": 0.11773121356964111, "global_step": 124787, "epoch": 1402, "lr": 5.6467342480508335e-05}
{"train_loss": 0.11986730992794037, "global_step": 124788, "epoch": 1402, "lr": 5.6466767613599025e-05}
{"train_loss": 0.048717427998781204, "global_step": 124789, "epoch": 1402, "lr": 5.6466192745820326e-05}
{"train_loss": 0.09650733321905136, "global_step": 124790, "epoch": 1402, "lr": 5.646561787717234e-05}
{"train_loss": 0.14182622730731964, "global_step": 124791, "epoch": 1402, "lr": 5.646504300765512e-05}
{"train_loss": 0.200332373380661, "global_step": 124792, "epoch": 1402, "lr": 5.646446813726876e-05}
{"train_loss": 0.07644211500883102, "global_step": 124793, "epoch": 1402, "lr": 5.646389326601332e-05}
{"train_loss": 0.20540790259838104, "global_step": 124794, "epoch": 1402, "lr": 5.6463318393888895e-05}
{"train_loss": 0.09791114181280136, "global_step": 124795, "epoch": 1402, "lr": 5.6462743520895564e-05}
{"train_loss": 0.13103502988815308, "global_step": 124796, "epoch": 1402, "lr": 5.6462168647033387e-05}
{"train_loss": 0.12305250018835068, "global_step": 124797, "epoch": 1402, "lr": 5.646159377230246e-05}
{"train_loss": 0.22239640355110168, "global_step": 124798, "epoch": 1402, "lr": 5.646101889670285e-05}
{"train_loss": 0.09403719007968903, "global_step": 124799, "epoch": 1402, "lr": 5.6460444020234614e-05}
{"train_loss": 0.09971693903207779, "global_step": 124800, "epoch": 1402, "lr": 5.645986914289787e-05}
{"train_loss": 0.16208182275295258, "global_step": 124801, "epoch": 1402, "lr": 5.645929426469267e-05}
{"train_loss": 0.11234121769666672, "global_step": 124802, "epoch": 1402, "lr": 5.6458719385619095e-05}
{"train_loss": 0.10180201381444931, "global_step": 124803, "epoch": 1402, "lr": 5.645814450567723e-05}
{"train_loss": 0.1069144681096077, "global_step": 124804, "epoch": 1402, "lr": 5.645756962486714e-05}
{"train_loss": 0.11702047288417816, "global_step": 124805, "epoch": 1402, "lr": 5.6456994743188905e-05}
{"train_loss": 0.10976140946149826, "global_step": 124806, "epoch": 1402, "lr": 5.645641986064261e-05}
{"train_loss": 0.12483824789524078, "global_step": 124807, "epoch": 1402, "lr": 5.6455844977228324e-05}
{"train_loss": 0.16168862581253052, "global_step": 124808, "epoch": 1402, "lr": 5.645527009294613e-05}
{"train_loss": 0.14383256435394287, "global_step": 124809, "epoch": 1402, "lr": 5.6454695207796115e-05}
{"train_loss": 0.1284654587507248, "global_step": 124810, "epoch": 1402, "lr": 5.645412032177832e-05}
{"train_loss": 0.07931096851825714, "global_step": 124811, "epoch": 1402, "lr": 5.645354543489287e-05}
{"train_loss": 0.08279174566268921, "global_step": 124812, "epoch": 1402, "lr": 5.6452970547139815e-05}
{"train_loss": 0.07696971297264099, "global_step": 124813, "epoch": 1402, "lr": 5.645239565851923e-05}
{"train_loss": 0.09626274555921555, "global_step": 124814, "epoch": 1402, "lr": 5.64518207690312e-05}
{"train_loss": 0.1416957676410675, "global_step": 124815, "epoch": 1402, "lr": 5.645124587867582e-05}
{"train_loss": 0.15120896697044373, "global_step": 124816, "epoch": 1402, "lr": 5.645067098745312e-05}
{"train_loss": 0.07656726986169815, "global_step": 124817, "epoch": 1402, "lr": 5.6450096095363224e-05}
{"train_loss": 0.1128888800740242, "global_step": 124818, "epoch": 1402, "lr": 5.6449521202406186e-05}
{"train_loss": 0.11978898197412491, "global_step": 124819, "epoch": 1402, "lr": 5.64489463085821e-05}
{"train_loss": 0.07245651632547379, "global_step": 124820, "epoch": 1402, "lr": 5.644837141389101e-05}
{"train_loss": 0.1475772261619568, "global_step": 124821, "epoch": 1402, "lr": 5.644779651833303e-05}
{"train_loss": 0.08944889903068542, "global_step": 124822, "epoch": 1402, "lr": 5.644722162190822e-05}
{"train_loss": 0.12270952016115189, "global_step": 124823, "epoch": 1402, "lr": 5.644664672461666e-05}
{"train_loss": 0.1382725089788437, "global_step": 124824, "epoch": 1402, "lr": 5.6446071826458424e-05}
{"train_loss": 0.14966601133346558, "global_step": 124825, "epoch": 1402, "lr": 5.6445496927433595e-05}
{"train_loss": 0.1283029466867447, "global_step": 124826, "epoch": 1402, "lr": 5.644492202754224e-05}
{"train_loss": 0.1761224865913391, "global_step": 124827, "epoch": 1402, "lr": 5.644434712678445e-05}
{"train_loss": 0.12012872099876404, "global_step": 124828, "epoch": 1402, "lr": 5.64437722251603e-05}
{"train_loss": 0.20026682317256927, "global_step": 124829, "epoch": 1402, "lr": 5.644319732266985e-05}
{"train_loss": 0.12347989529371262, "global_step": 124830, "epoch": 1402, "lr": 5.64426224193132e-05}
{"train_loss": 0.12489297240972519, "global_step": 124831, "epoch": 1402, "lr": 5.644204751509041e-05}
{"train_loss": 0.07393290847539902, "global_step": 124832, "epoch": 1402, "lr": 5.644147261000158e-05}
{"train_loss": 0.11110217869281769, "global_step": 124833, "epoch": 1402, "lr": 5.644089770404677e-05}
{"train_loss": 0.08128920942544937, "global_step": 124834, "epoch": 1402, "lr": 5.644032279722605e-05}
{"train_loss": 0.14338909089565277, "global_step": 124835, "epoch": 1402, "lr": 5.6439747889539517e-05}
{"train_loss": 0.1541285216808319, "global_step": 124836, "epoch": 1402, "lr": 5.6439172980987234e-05}
{"train_loss": 0.11821233481168747, "global_step": 124837, "epoch": 1402, "lr": 5.643859807156928e-05}
{"train_loss": 0.07343042641878128, "global_step": 124838, "epoch": 1402, "lr": 5.643802316128573e-05}
{"train_loss": 0.11792915314435959, "global_step": 124839, "epoch": 1402, "lr": 5.6437448250136684e-05}
{"train_loss": 0.13108031451702118, "global_step": 124840, "epoch": 1402, "lr": 5.6436873338122184e-05}
{"train_loss": 0.1004612147808075, "global_step": 124841, "epoch": 1402, "lr": 5.643629842524233e-05}
{"train_loss": 0.14560534060001373, "global_step": 124842, "epoch": 1402, "lr": 5.64357235114972e-05}
{"train_loss": 0.11374400556087494, "global_step": 124843, "epoch": 1402, "lr": 5.6435148596886854e-05}
{"train_loss": 0.13569997251033783, "global_step": 124844, "epoch": 1402, "lr": 5.64345736814114e-05}
{"train_loss": 0.10515468567609787, "global_step": 124845, "epoch": 1402, "lr": 5.643399876507087e-05}
{"train_loss": 0.1134595274925232, "global_step": 124846, "epoch": 1402, "lr": 5.643342384786539e-05}
{"train_loss": 0.16312620043754578, "global_step": 124847, "epoch": 1402, "lr": 5.6432848929795e-05}
{"train_loss": 0.12300404906272888, "global_step": 124848, "epoch": 1402, "lr": 5.6432274010859797e-05}
{"train_loss": 0.12166202813386917, "global_step": 124849, "epoch": 1402, "lr": 5.643169909105985e-05}
{"train_loss": 0.14623497426509857, "global_step": 124850, "epoch": 1402, "lr": 5.643112417039524e-05}
{"train_loss": 0.14068694412708282, "global_step": 124851, "epoch": 1402, "lr": 5.643054924886606e-05}
{"train_loss": 0.14402072131633759, "global_step": 124852, "epoch": 1402, "lr": 5.642997432647236e-05}
{"train_loss": 0.14466078579425812, "global_step": 124853, "epoch": 1402, "lr": 5.642939940321422e-05}
{"train_loss": 0.09123726934194565, "global_step": 124854, "epoch": 1402, "lr": 5.642882447909174e-05}
{"train_loss": 0.14142119884490967, "global_step": 124855, "epoch": 1402, "lr": 5.642824955410497e-05}
{"train_loss": 0.09334566444158554, "global_step": 124856, "epoch": 1402, "lr": 5.6427674628254004e-05}
{"train_loss": 0.07932329177856445, "global_step": 124857, "epoch": 1402, "lr": 5.642709970153891e-05}
{"train_loss": 0.10763757675886154, "global_step": 124858, "epoch": 1402, "lr": 5.642652477395978e-05}
{"train_loss": 0.18847458064556122, "global_step": 124859, "epoch": 1402, "lr": 5.642594984551668e-05}
{"train_loss": 0.08477521687746048, "global_step": 124860, "epoch": 1402, "lr": 5.6425374916209693e-05}
{"train_loss": 0.20476996898651123, "global_step": 124861, "epoch": 1402, "lr": 5.642479998603889e-05}
{"train_loss": 0.17861393094062805, "global_step": 124862, "epoch": 1402, "lr": 5.6424225055004344e-05}
{"train_loss": 0.11761204153299332, "global_step": 124863, "epoch": 1402, "lr": 5.6423650123106145e-05}
{"train_loss": 0.1981593668460846, "global_step": 124864, "epoch": 1402, "lr": 5.6423075190344367e-05}
{"train_loss": 0.16999602317810059, "global_step": 124865, "epoch": 1402, "lr": 5.642250025671908e-05}
{"train_loss": 0.12794347793868419, "global_step": 124866, "epoch": 1402, "lr": 5.6421925322230365e-05, "val_loss": 5.481879234313965}
{"train_loss": 0.21919895708560944, "global_step": 124867, "epoch": 1403, "lr": 5.6421350386878306e-05}
{"train_loss": 0.14902114868164062, "global_step": 124868, "epoch": 1403, "lr": 5.642077545066298e-05}
{"train_loss": 0.12189635634422302, "global_step": 124869, "epoch": 1403, "lr": 5.642020051358444e-05}
{"train_loss": 0.07578926533460617, "global_step": 124870, "epoch": 1403, "lr": 5.64196255756428e-05}
{"train_loss": 0.11833488941192627, "global_step": 124871, "epoch": 1403, "lr": 5.641905063683811e-05}
{"train_loss": 0.09053356945514679, "global_step": 124872, "epoch": 1403, "lr": 5.641847569717046e-05}
{"train_loss": 0.20003055036067963, "global_step": 124873, "epoch": 1403, "lr": 5.6417900756639916e-05}
{"train_loss": 0.1381111741065979, "global_step": 124874, "epoch": 1403, "lr": 5.641732581524658e-05}
{"train_loss": 0.22494752705097198, "global_step": 124875, "epoch": 1403, "lr": 5.64167508729905e-05}
{"train_loss": 0.10541671514511108, "global_step": 124876, "epoch": 1403, "lr": 5.6416175929871775e-05}
{"train_loss": 0.12358919531106949, "global_step": 124877, "epoch": 1403, "lr": 5.641560098589047e-05}
{"train_loss": 0.14473113417625427, "global_step": 124878, "epoch": 1403, "lr": 5.641502604104666e-05}
{"train_loss": 0.1580827385187149, "global_step": 124879, "epoch": 1403, "lr": 5.6414451095340435e-05}
{"train_loss": 0.11840574443340302, "global_step": 124880, "epoch": 1403, "lr": 5.6413876148771864e-05}
{"train_loss": 0.11005866527557373, "global_step": 124881, "epoch": 1403, "lr": 5.641330120134102e-05}
{"train_loss": 0.12369376420974731, "global_step": 124882, "epoch": 1403, "lr": 5.641272625304799e-05}
{"train_loss": 0.12315899133682251, "global_step": 124883, "epoch": 1403, "lr": 5.6412151303892844e-05}
{"train_loss": 0.1312989890575409, "global_step": 124884, "epoch": 1403, "lr": 5.6411576353875676e-05}
{"train_loss": 0.12460962682962418, "global_step": 124885, "epoch": 1403, "lr": 5.641100140299653e-05}
{"train_loss": 0.07771169394254684, "global_step": 124886, "epoch": 1403, "lr": 5.641042645125552e-05}
{"train_loss": 0.1245436817407608, "global_step": 124887, "epoch": 1403, "lr": 5.640985149865269e-05}
{"train_loss": 0.10843177139759064, "global_step": 124888, "epoch": 1403, "lr": 5.640927654518814e-05}
{"train_loss": 0.08129376918077469, "global_step": 124889, "epoch": 1403, "lr": 5.6408701590861944e-05}
{"train_loss": 0.1401381939649582, "global_step": 124890, "epoch": 1403, "lr": 5.6408126635674175e-05}
{"train_loss": 0.10043644160032272, "global_step": 124891, "epoch": 1403, "lr": 5.640755167962491e-05}
{"train_loss": 0.07310856133699417, "global_step": 124892, "epoch": 1403, "lr": 5.640697672271423e-05}
{"train_loss": 0.13008683919906616, "global_step": 124893, "epoch": 1403, "lr": 5.6406401764942206e-05}
{"train_loss": 0.1654365062713623, "global_step": 124894, "epoch": 1403, "lr": 5.6405826806308924e-05}
{"train_loss": 0.12753649055957794, "global_step": 124895, "epoch": 1403, "lr": 5.640525184681445e-05}
{"train_loss": 0.12818565964698792, "global_step": 124896, "epoch": 1403, "lr": 5.640467688645888e-05}
{"train_loss": 0.23281437158584595, "global_step": 124897, "epoch": 1403, "lr": 5.640410192524228e-05}
{"train_loss": 0.10657252371311188, "global_step": 124898, "epoch": 1403, "lr": 5.64035269631647e-05}
{"train_loss": 0.1509653478860855, "global_step": 124899, "epoch": 1403, "lr": 5.640295200022627e-05}
{"train_loss": 0.13984255492687225, "global_step": 124900, "epoch": 1403, "lr": 5.640237703642704e-05}
{"train_loss": 0.12768131494522095, "global_step": 124901, "epoch": 1403, "lr": 5.640180207176709e-05}
{"train_loss": 0.1818673461675644, "global_step": 124902, "epoch": 1403, "lr": 5.640122710624648e-05}
{"train_loss": 0.10243356972932816, "global_step": 124903, "epoch": 1403, "lr": 5.640065213986532e-05}
{"train_loss": 0.17925375699996948, "global_step": 124904, "epoch": 1403, "lr": 5.6400077172623654e-05}
{"train_loss": 0.06692729890346527, "global_step": 124905, "epoch": 1403, "lr": 5.639950220452158e-05}
{"train_loss": 0.17188461124897003, "global_step": 124906, "epoch": 1403, "lr": 5.639892723555917e-05}
{"train_loss": 0.16966216266155243, "global_step": 124907, "epoch": 1403, "lr": 5.639835226573651e-05}
{"train_loss": 0.09354215115308762, "global_step": 124908, "epoch": 1403, "lr": 5.639777729505367e-05}
{"train_loss": 0.1568448394536972, "global_step": 124909, "epoch": 1403, "lr": 5.6397202323510724e-05}
{"train_loss": 0.12533070147037506, "global_step": 124910, "epoch": 1403, "lr": 5.6396627351107756e-05}
{"train_loss": 0.14115865528583527, "global_step": 124911, "epoch": 1403, "lr": 5.639605237784483e-05}
{"train_loss": 0.148747980594635, "global_step": 124912, "epoch": 1403, "lr": 5.639547740372204e-05}
{"train_loss": 0.14999927580356598, "global_step": 124913, "epoch": 1403, "lr": 5.639490242873945e-05}
{"train_loss": 0.07888102531433105, "global_step": 124914, "epoch": 1403, "lr": 5.639432745289714e-05}
{"train_loss": 0.18017643690109253, "global_step": 124915, "epoch": 1403, "lr": 5.6393752476195205e-05}
{"train_loss": 0.15653182566165924, "global_step": 124916, "epoch": 1403, "lr": 5.6393177498633684e-05}
{"train_loss": 0.09465043246746063, "global_step": 124917, "epoch": 1403, "lr": 5.63926025202127e-05}
{"train_loss": 0.14475688338279724, "global_step": 124918, "epoch": 1403, "lr": 5.63920275409323e-05}
{"train_loss": 0.15062610805034637, "global_step": 124919, "epoch": 1403, "lr": 5.639145256079258e-05}
{"train_loss": 0.17969708144664764, "global_step": 124920, "epoch": 1403, "lr": 5.63908775797936e-05}
{"train_loss": 0.12193036079406738, "global_step": 124921, "epoch": 1403, "lr": 5.639030259793545e-05}
{"train_loss": 0.0902458056807518, "global_step": 124922, "epoch": 1403, "lr": 5.638972761521819e-05}
{"train_loss": 0.12005792558193207, "global_step": 124923, "epoch": 1403, "lr": 5.6389152631641916e-05}
{"train_loss": 0.09325265139341354, "global_step": 124924, "epoch": 1403, "lr": 5.6388577647206695e-05}
{"train_loss": 0.13879041373729706, "global_step": 124925, "epoch": 1403, "lr": 5.638800266191261e-05}
{"train_loss": 0.1399562507867813, "global_step": 124926, "epoch": 1403, "lr": 5.638742767575974e-05}
{"train_loss": 0.18005993962287903, "global_step": 124927, "epoch": 1403, "lr": 5.638685268874815e-05}
{"train_loss": 0.16747678816318512, "global_step": 124928, "epoch": 1403, "lr": 5.638627770087793e-05}
{"train_loss": 0.15819112956523895, "global_step": 124929, "epoch": 1403, "lr": 5.638570271214916e-05}
{"train_loss": 0.18502052128314972, "global_step": 124930, "epoch": 1403, "lr": 5.638512772256189e-05}
{"train_loss": 0.18924538791179657, "global_step": 124931, "epoch": 1403, "lr": 5.6384552732116236e-05}
{"train_loss": 0.19601377844810486, "global_step": 124932, "epoch": 1403, "lr": 5.638397774081224e-05}
{"train_loss": 0.13798516988754272, "global_step": 124933, "epoch": 1403, "lr": 5.6383402748650015e-05}
{"train_loss": 0.13374826312065125, "global_step": 124934, "epoch": 1403, "lr": 5.6382827755629594e-05}
{"train_loss": 0.13648498058319092, "global_step": 124935, "epoch": 1403, "lr": 5.6382252761751094e-05}
{"train_loss": 0.08855708688497543, "global_step": 124936, "epoch": 1403, "lr": 5.6381677767014585e-05}
{"train_loss": 0.12458854168653488, "global_step": 124937, "epoch": 1403, "lr": 5.638110277142014e-05}
{"train_loss": 0.13642475008964539, "global_step": 124938, "epoch": 1403, "lr": 5.6380527774967825e-05}
{"train_loss": 0.1271195262670517, "global_step": 124939, "epoch": 1403, "lr": 5.637995277765773e-05}
{"train_loss": 0.12834839522838593, "global_step": 124940, "epoch": 1403, "lr": 5.6379377779489925e-05}
{"train_loss": 0.1394331008195877, "global_step": 124941, "epoch": 1403, "lr": 5.637880278046449e-05}
{"train_loss": 0.2295704185962677, "global_step": 124942, "epoch": 1403, "lr": 5.637822778058151e-05}
{"train_loss": 0.1288823038339615, "global_step": 124943, "epoch": 1403, "lr": 5.637765277984104e-05}
{"train_loss": 0.11032256484031677, "global_step": 124944, "epoch": 1403, "lr": 5.637707777824318e-05}
{"train_loss": 0.1764085590839386, "global_step": 124945, "epoch": 1403, "lr": 5.6376502775788e-05}
{"train_loss": 0.09670176357030869, "global_step": 124946, "epoch": 1403, "lr": 5.637592777247558e-05}
{"train_loss": 0.09528812766075134, "global_step": 124947, "epoch": 1403, "lr": 5.637535276830599e-05}
{"train_loss": 0.131506085395813, "global_step": 124948, "epoch": 1403, "lr": 5.637477776327931e-05}
{"train_loss": 0.1266835331916809, "global_step": 124949, "epoch": 1403, "lr": 5.637420275739562e-05}
{"train_loss": 0.1198156476020813, "global_step": 124950, "epoch": 1403, "lr": 5.6373627750655e-05}
{"train_loss": 0.13513721525669098, "global_step": 124951, "epoch": 1403, "lr": 5.637305274305752e-05}
{"train_loss": 0.09602285921573639, "global_step": 124952, "epoch": 1403, "lr": 5.6372477734603255e-05}
{"train_loss": 0.19459563493728638, "global_step": 124953, "epoch": 1403, "lr": 5.637190272529229e-05}
{"train_loss": 0.14118368923664093, "global_step": 124954, "epoch": 1403, "lr": 5.63713277151247e-05}
{"train_loss": 0.1362236842847942, "global_step": 124955, "epoch": 1403, "lr": 5.637075270410057e-05, "val_loss": 5.572772026062012}
{"train_loss": 0.12806452810764313, "global_step": 124956, "epoch": 1404, "lr": 5.637017769221997e-05}
{"train_loss": 0.09169954061508179, "global_step": 124957, "epoch": 1404, "lr": 5.636960267948298e-05}
{"train_loss": 0.08801998198032379, "global_step": 124958, "epoch": 1404, "lr": 5.6369027665889675e-05}
{"train_loss": 0.15397605299949646, "global_step": 124959, "epoch": 1404, "lr": 5.636845265144013e-05}
{"train_loss": 0.1368950456380844, "global_step": 124960, "epoch": 1404, "lr": 5.636787763613441e-05}
{"train_loss": 0.1635720133781433, "global_step": 124961, "epoch": 1404, "lr": 5.636730261997262e-05}
{"train_loss": 0.19160667061805725, "global_step": 124962, "epoch": 1404, "lr": 5.636672760295483e-05}
{"train_loss": 0.12589408457279205, "global_step": 124963, "epoch": 1404, "lr": 5.636615258508109e-05}
{"train_loss": 0.11358990520238876, "global_step": 124964, "epoch": 1404, "lr": 5.6365577566351516e-05}
{"train_loss": 0.09072074294090271, "global_step": 124965, "epoch": 1404, "lr": 5.6365002546766156e-05}
{"train_loss": 0.1057276725769043, "global_step": 124966, "epoch": 1404, "lr": 5.6364427526325103e-05}
{"train_loss": 0.15376847982406616, "global_step": 124967, "epoch": 1404, "lr": 5.636385250502844e-05}
{"train_loss": 0.1588340401649475, "global_step": 124968, "epoch": 1404, "lr": 5.636327748287622e-05}
{"train_loss": 0.14575740694999695, "global_step": 124969, "epoch": 1404, "lr": 5.6362702459868546e-05}
{"train_loss": 0.12341626733541489, "global_step": 124970, "epoch": 1404, "lr": 5.6362127436005474e-05}
{"train_loss": 0.12887656688690186, "global_step": 124971, "epoch": 1404, "lr": 5.636155241128709e-05}
{"train_loss": 0.1360069215297699, "global_step": 124972, "epoch": 1404, "lr": 5.6360977385713476e-05}
{"train_loss": 0.11355467140674591, "global_step": 124973, "epoch": 1404, "lr": 5.6360402359284714e-05}
{"train_loss": 0.12039743363857269, "global_step": 124974, "epoch": 1404, "lr": 5.635982733200087e-05}
{"train_loss": 0.09312672913074493, "global_step": 124975, "epoch": 1404, "lr": 5.635925230386202e-05}
{"train_loss": 0.09646038711071014, "global_step": 124976, "epoch": 1404, "lr": 5.635867727486825e-05}
{"train_loss": 0.0955791100859642, "global_step": 124977, "epoch": 1404, "lr": 5.635810224501965e-05}
{"train_loss": 0.13081791996955872, "global_step": 124978, "epoch": 1404, "lr": 5.635752721431626e-05}
{"train_loss": 0.1838204562664032, "global_step": 124979, "epoch": 1404, "lr": 5.6356952182758185e-05}
{"train_loss": 0.16403080523014069, "global_step": 124980, "epoch": 1404, "lr": 5.6356377150345494e-05}
{"train_loss": 0.1482563316822052, "global_step": 124981, "epoch": 1404, "lr": 5.635580211707826e-05}
{"train_loss": 0.14557085931301117, "global_step": 124982, "epoch": 1404, "lr": 5.6355227082956577e-05}
{"train_loss": 0.1559169739484787, "global_step": 124983, "epoch": 1404, "lr": 5.635465204798049e-05}
{"train_loss": 0.10919641703367233, "global_step": 124984, "epoch": 1404, "lr": 5.635407701215012e-05}
{"train_loss": 0.10194623470306396, "global_step": 124985, "epoch": 1404, "lr": 5.6353501975465515e-05}
{"train_loss": 0.12511244416236877, "global_step": 124986, "epoch": 1404, "lr": 5.635292693792675e-05}
{"train_loss": 0.10156908631324768, "global_step": 124987, "epoch": 1404, "lr": 5.6352351899533925e-05}
{"train_loss": 0.1655649095773697, "global_step": 124988, "epoch": 1404, "lr": 5.63517768602871e-05}
{"train_loss": 0.14548933506011963, "global_step": 124989, "epoch": 1404, "lr": 5.635120182018634e-05}
{"train_loss": 0.1740422546863556, "global_step": 124990, "epoch": 1404, "lr": 5.6350626779231754e-05}
{"train_loss": 0.08662892878055573, "global_step": 124991, "epoch": 1404, "lr": 5.63500517374234e-05}
{"train_loss": 0.1264113187789917, "global_step": 124992, "epoch": 1404, "lr": 5.634947669476136e-05}
{"train_loss": 0.1924632489681244, "global_step": 124993, "epoch": 1404, "lr": 5.6348901651245714e-05}
{"train_loss": 0.12823455035686493, "global_step": 124994, "epoch": 1404, "lr": 5.634832660687652e-05}
{"train_loss": 0.1051500216126442, "global_step": 124995, "epoch": 1404, "lr": 5.6347751561653895e-05}
{"train_loss": 0.11706222593784332, "global_step": 124996, "epoch": 1404, "lr": 5.6347176515577885e-05}
{"train_loss": 0.0678819864988327, "global_step": 124997, "epoch": 1404, "lr": 5.6346601468648574e-05}
{"train_loss": 0.11331596225500107, "global_step": 124998, "epoch": 1404, "lr": 5.6346026420866036e-05}
{"train_loss": 0.11005309224128723, "global_step": 124999, "epoch": 1404, "lr": 5.6345451372230354e-05}
{"train_loss": 0.1823417991399765, "global_step": 125000, "epoch": 1404, "lr": 5.6344876322741594e-05}
{"train_loss": 0.1535940170288086, "global_step": 125001, "epoch": 1404, "lr": 5.6344301272399844e-05}
{"train_loss": 0.1475193351507187, "global_step": 125002, "epoch": 1404, "lr": 5.6343726221205194e-05}
{"train_loss": 0.21443402767181396, "global_step": 125003, "epoch": 1404, "lr": 5.6343151169157696e-05}
{"train_loss": 0.1434246301651001, "global_step": 125004, "epoch": 1404, "lr": 5.634257611625744e-05}
{"train_loss": 0.1842706799507141, "global_step": 125005, "epoch": 1404, "lr": 5.634200106250451e-05}
{"train_loss": 0.1917860060930252, "global_step": 125006, "epoch": 1404, "lr": 5.634142600789897e-05}
{"train_loss": 0.13010330498218536, "global_step": 125007, "epoch": 1404, "lr": 5.634085095244088e-05}
{"train_loss": 0.14434783160686493, "global_step": 125008, "epoch": 1404, "lr": 5.634027589613037e-05}
{"train_loss": 0.09352897852659225, "global_step": 125009, "epoch": 1404, "lr": 5.633970083896748e-05}
{"train_loss": 0.24647517502307892, "global_step": 125010, "epoch": 1404, "lr": 5.633912578095229e-05}
{"train_loss": 0.2033318728208542, "global_step": 125011, "epoch": 1404, "lr": 5.633855072208488e-05}
{"train_loss": 0.15479516983032227, "global_step": 125012, "epoch": 1404, "lr": 5.633797566236534e-05}
{"train_loss": 0.12170731276273727, "global_step": 125013, "epoch": 1404, "lr": 5.6337400601793735e-05}
{"train_loss": 0.12382570654153824, "global_step": 125014, "epoch": 1404, "lr": 5.633682554037013e-05}
{"train_loss": 0.12844541668891907, "global_step": 125015, "epoch": 1404, "lr": 5.633625047809464e-05}
{"train_loss": 0.10349735617637634, "global_step": 125016, "epoch": 1404, "lr": 5.6335675414967304e-05}
{"train_loss": 0.11002825945615768, "global_step": 125017, "epoch": 1404, "lr": 5.6335100350988216e-05}
{"train_loss": 0.17333820462226868, "global_step": 125018, "epoch": 1404, "lr": 5.6334525286157446e-05}
{"train_loss": 0.16070663928985596, "global_step": 125019, "epoch": 1404, "lr": 5.6333950220475076e-05}
{"train_loss": 0.18676359951496124, "global_step": 125020, "epoch": 1404, "lr": 5.6333375153941194e-05}
{"train_loss": 0.1106308251619339, "global_step": 125021, "epoch": 1404, "lr": 5.633280008655585e-05}
{"train_loss": 0.11003240197896957, "global_step": 125022, "epoch": 1404, "lr": 5.633222501831915e-05}
{"train_loss": 0.10296545922756195, "global_step": 125023, "epoch": 1404, "lr": 5.633164994923116e-05}
{"train_loss": 0.08942264318466187, "global_step": 125024, "epoch": 1404, "lr": 5.633107487929194e-05}
{"train_loss": 0.1771797388792038, "global_step": 125025, "epoch": 1404, "lr": 5.633049980850159e-05}
{"train_loss": 0.12369295954704285, "global_step": 125026, "epoch": 1404, "lr": 5.6329924736860195e-05}
{"train_loss": 0.14229463040828705, "global_step": 125027, "epoch": 1404, "lr": 5.6329349664367816e-05}
{"train_loss": 0.08719733357429504, "global_step": 125028, "epoch": 1404, "lr": 5.632877459102453e-05}
{"train_loss": 0.0977705791592598, "global_step": 125029, "epoch": 1404, "lr": 5.632819951683042e-05}
{"train_loss": 0.1381661593914032, "global_step": 125030, "epoch": 1404, "lr": 5.6327624441785556e-05}
{"train_loss": 0.18924929201602936, "global_step": 125031, "epoch": 1404, "lr": 5.6327049365890015e-05}
{"train_loss": 0.10530531406402588, "global_step": 125032, "epoch": 1404, "lr": 5.6326474289143884e-05}
{"train_loss": 0.13598795235157013, "global_step": 125033, "epoch": 1404, "lr": 5.632589921154724e-05}
{"train_loss": 0.09837786853313446, "global_step": 125034, "epoch": 1404, "lr": 5.6325324133100155e-05}
{"train_loss": 0.13076196610927582, "global_step": 125035, "epoch": 1404, "lr": 5.632474905380272e-05}
{"train_loss": 0.08975589275360107, "global_step": 125036, "epoch": 1404, "lr": 5.632417397365498e-05}
{"train_loss": 0.10984832793474197, "global_step": 125037, "epoch": 1404, "lr": 5.6323598892657034e-05}
{"train_loss": 0.12058261036872864, "global_step": 125038, "epoch": 1404, "lr": 5.632302381080897e-05}
{"train_loss": 0.15831679105758667, "global_step": 125039, "epoch": 1404, "lr": 5.632244872811084e-05}
{"train_loss": 0.12551181018352509, "global_step": 125040, "epoch": 1404, "lr": 5.632187364456274e-05}
{"train_loss": 0.08798004686832428, "global_step": 125041, "epoch": 1404, "lr": 5.6321298560164735e-05}
{"train_loss": 0.13715852797031403, "global_step": 125042, "epoch": 1404, "lr": 5.6320723474916914e-05}
{"train_loss": 0.1554529368877411, "global_step": 125043, "epoch": 1404, "lr": 5.6320148388819335e-05}
{"train_loss": 0.13396437357315857, "global_step": 125044, "epoch": 1404, "lr": 5.6319573301872095e-05, "val_loss": 5.502062797546387}
{"train_loss": 0.12866495549678802, "global_step": 125045, "epoch": 1405, "lr": 5.631899821407528e-05}
{"train_loss": 0.10737858712673187, "global_step": 125046, "epoch": 1405, "lr": 5.6318423125428946e-05}
{"train_loss": 0.08243998140096664, "global_step": 125047, "epoch": 1405, "lr": 5.631784803593317e-05}
{"train_loss": 0.14713816344738007, "global_step": 125048, "epoch": 1405, "lr": 5.631727294558804e-05}
{"train_loss": 0.17464274168014526, "global_step": 125049, "epoch": 1405, "lr": 5.631669785439363e-05}
{"train_loss": 0.13992807269096375, "global_step": 125050, "epoch": 1405, "lr": 5.631612276235002e-05}
{"train_loss": 0.12402358651161194, "global_step": 125051, "epoch": 1405, "lr": 5.631554766945728e-05}
{"train_loss": 0.16428235173225403, "global_step": 125052, "epoch": 1405, "lr": 5.6314972575715496e-05}
{"train_loss": 0.15547879040241241, "global_step": 125053, "epoch": 1405, "lr": 5.631439748112474e-05}
{"train_loss": 0.1282678097486496, "global_step": 125054, "epoch": 1405, "lr": 5.63138223856851e-05}
{"train_loss": 0.13202165067195892, "global_step": 125055, "epoch": 1405, "lr": 5.631324728939663e-05}
{"train_loss": 0.15763334929943085, "global_step": 125056, "epoch": 1405, "lr": 5.631267219225942e-05}
{"train_loss": 0.14204178750514984, "global_step": 125057, "epoch": 1405, "lr": 5.631209709427355e-05}
{"train_loss": 0.14086830615997314, "global_step": 125058, "epoch": 1405, "lr": 5.63115219954391e-05}
{"train_loss": 0.21379020810127258, "global_step": 125059, "epoch": 1405, "lr": 5.631094689575614e-05}
{"train_loss": 0.21127134561538696, "global_step": 125060, "epoch": 1405, "lr": 5.631037179522475e-05}
{"train_loss": 0.11949091404676437, "global_step": 125061, "epoch": 1405, "lr": 5.6309796693844996e-05}
{"train_loss": 0.09048289805650711, "global_step": 125062, "epoch": 1405, "lr": 5.6309221591616976e-05}
{"train_loss": 0.17417798936367035, "global_step": 125063, "epoch": 1405, "lr": 5.6308646488540764e-05}
{"train_loss": 0.15321996808052063, "global_step": 125064, "epoch": 1405, "lr": 5.630807138461642e-05}
{"train_loss": 0.1730770468711853, "global_step": 125065, "epoch": 1405, "lr": 5.630749627984404e-05}
{"train_loss": 0.16074569523334503, "global_step": 125066, "epoch": 1405, "lr": 5.6306921174223705e-05}
{"train_loss": 0.0826127752661705, "global_step": 125067, "epoch": 1405, "lr": 5.6306346067755466e-05}
{"train_loss": 0.0984422117471695, "global_step": 125068, "epoch": 1405, "lr": 5.630577096043942e-05}
{"train_loss": 0.09102153033018112, "global_step": 125069, "epoch": 1405, "lr": 5.630519585227564e-05}
{"train_loss": 0.1493179202079773, "global_step": 125070, "epoch": 1405, "lr": 5.6304620743264206e-05}
{"train_loss": 0.10215633362531662, "global_step": 125071, "epoch": 1405, "lr": 5.630404563340519e-05}
{"train_loss": 0.12287857383489609, "global_step": 125072, "epoch": 1405, "lr": 5.6303470522698663e-05}
{"train_loss": 0.12228022515773773, "global_step": 125073, "epoch": 1405, "lr": 5.630289541114474e-05}
{"train_loss": 0.15438255667686462, "global_step": 125074, "epoch": 1405, "lr": 5.630232029874345e-05}
{"train_loss": 0.2060411125421524, "global_step": 125075, "epoch": 1405, "lr": 5.630174518549488e-05}
{"train_loss": 0.17381645739078522, "global_step": 125076, "epoch": 1405, "lr": 5.630117007139913e-05}
{"train_loss": 0.13175968825817108, "global_step": 125077, "epoch": 1405, "lr": 5.6300594956456266e-05}
{"train_loss": 0.09943056106567383, "global_step": 125078, "epoch": 1405, "lr": 5.6300019840666355e-05}
{"train_loss": 0.1157732829451561, "global_step": 125079, "epoch": 1405, "lr": 5.6299444724029484e-05}
{"train_loss": 0.12023597955703735, "global_step": 125080, "epoch": 1405, "lr": 5.6298869606545724e-05}
{"train_loss": 0.11767052859067917, "global_step": 125081, "epoch": 1405, "lr": 5.629829448821517e-05}
{"train_loss": 0.13527673482894897, "global_step": 125082, "epoch": 1405, "lr": 5.629771936903788e-05}
{"train_loss": 0.10882457345724106, "global_step": 125083, "epoch": 1405, "lr": 5.6297144249013945e-05}
{"train_loss": 0.11387979239225388, "global_step": 125084, "epoch": 1405, "lr": 5.629656912814343e-05}
{"train_loss": 0.1593531221151352, "global_step": 125085, "epoch": 1405, "lr": 5.629599400642642e-05}
{"train_loss": 0.13394013047218323, "global_step": 125086, "epoch": 1405, "lr": 5.629541888386298e-05}
{"train_loss": 0.1484752744436264, "global_step": 125087, "epoch": 1405, "lr": 5.6294843760453216e-05}
{"train_loss": 0.08568257838487625, "global_step": 125088, "epoch": 1405, "lr": 5.6294268636197177e-05}
{"train_loss": 0.15884698927402496, "global_step": 125089, "epoch": 1405, "lr": 5.629369351109496e-05}
{"train_loss": 0.06849716603755951, "global_step": 125090, "epoch": 1405, "lr": 5.6293118385146614e-05}
{"train_loss": 0.11249198764562607, "global_step": 125091, "epoch": 1405, "lr": 5.6292543258352246e-05}
{"train_loss": 0.13704974949359894, "global_step": 125092, "epoch": 1405, "lr": 5.6291968130711924e-05}
{"train_loss": 0.16273929178714752, "global_step": 125093, "epoch": 1405, "lr": 5.6291393002225734e-05}
{"train_loss": 0.061541128903627396, "global_step": 125094, "epoch": 1405, "lr": 5.629081787289373e-05}
{"train_loss": 0.11229077726602554, "global_step": 125095, "epoch": 1405, "lr": 5.6290242742716e-05}
{"train_loss": 0.15937593579292297, "global_step": 125096, "epoch": 1405, "lr": 5.628966761169263e-05}
{"train_loss": 0.22340092062950134, "global_step": 125097, "epoch": 1405, "lr": 5.628909247982369e-05}
{"train_loss": 0.20150260627269745, "global_step": 125098, "epoch": 1405, "lr": 5.628851734710925e-05}
{"train_loss": 0.12547743320465088, "global_step": 125099, "epoch": 1405, "lr": 5.6287942213549395e-05}
{"train_loss": 0.09074901044368744, "global_step": 125100, "epoch": 1405, "lr": 5.628736707914421e-05}
{"train_loss": 0.0997006893157959, "global_step": 125101, "epoch": 1405, "lr": 5.628679194389377e-05}
{"train_loss": 0.12278695404529572, "global_step": 125102, "epoch": 1405, "lr": 5.628621680779813e-05}
{"train_loss": 0.16654032468795776, "global_step": 125103, "epoch": 1405, "lr": 5.62856416708574e-05}
{"train_loss": 0.0952489897608757, "global_step": 125104, "epoch": 1405, "lr": 5.628506653307164e-05}
{"train_loss": 0.08845507353544235, "global_step": 125105, "epoch": 1405, "lr": 5.628449139444093e-05}
{"train_loss": 0.07721250504255295, "global_step": 125106, "epoch": 1405, "lr": 5.628391625496534e-05}
{"train_loss": 0.15926595032215118, "global_step": 125107, "epoch": 1405, "lr": 5.628334111464496e-05}
{"train_loss": 0.1608281433582306, "global_step": 125108, "epoch": 1405, "lr": 5.6282765973479855e-05}
{"train_loss": 0.129466250538826, "global_step": 125109, "epoch": 1405, "lr": 5.628219083147011e-05}
{"train_loss": 0.11382349580526352, "global_step": 125110, "epoch": 1405, "lr": 5.6281615688615805e-05}
{"train_loss": 0.18130095303058624, "global_step": 125111, "epoch": 1405, "lr": 5.628104054491702e-05}
{"train_loss": 0.13995146751403809, "global_step": 125112, "epoch": 1405, "lr": 5.628046540037383e-05}
{"train_loss": 0.1221214160323143, "global_step": 125113, "epoch": 1405, "lr": 5.62798902549863e-05}
{"train_loss": 0.14679352939128876, "global_step": 125114, "epoch": 1405, "lr": 5.627931510875451e-05}
{"train_loss": 0.09327752888202667, "global_step": 125115, "epoch": 1405, "lr": 5.627873996167854e-05}
{"train_loss": 0.10975741595029831, "global_step": 125116, "epoch": 1405, "lr": 5.627816481375847e-05}
{"train_loss": 0.11806214600801468, "global_step": 125117, "epoch": 1405, "lr": 5.627758966499438e-05}
{"train_loss": 0.12699180841445923, "global_step": 125118, "epoch": 1405, "lr": 5.627701451538634e-05}
{"train_loss": 0.1434115320444107, "global_step": 125119, "epoch": 1405, "lr": 5.6276439364934444e-05}
{"train_loss": 0.15501852333545685, "global_step": 125120, "epoch": 1405, "lr": 5.6275864213638754e-05}
{"train_loss": 0.1852550059556961, "global_step": 125121, "epoch": 1405, "lr": 5.6275289061499345e-05}
{"train_loss": 0.1170726865530014, "global_step": 125122, "epoch": 1405, "lr": 5.62747139085163e-05}
{"train_loss": 0.11928502470254898, "global_step": 125123, "epoch": 1405, "lr": 5.62741387546897e-05}
{"train_loss": 0.16556954383850098, "global_step": 125124, "epoch": 1405, "lr": 5.627356360001962e-05}
{"train_loss": 0.22150832414627075, "global_step": 125125, "epoch": 1405, "lr": 5.627298844450613e-05}
{"train_loss": 0.13507413864135742, "global_step": 125126, "epoch": 1405, "lr": 5.6272413288149315e-05}
{"train_loss": 0.1372310221195221, "global_step": 125127, "epoch": 1405, "lr": 5.627183813094925e-05}
{"train_loss": 0.12664130330085754, "global_step": 125128, "epoch": 1405, "lr": 5.6271262972906015e-05}
{"train_loss": 0.09947537630796432, "global_step": 125129, "epoch": 1405, "lr": 5.627068781401968e-05}
{"train_loss": 0.11473879963159561, "global_step": 125130, "epoch": 1405, "lr": 5.6270112654290344e-05}
{"train_loss": 0.16700203716754913, "global_step": 125131, "epoch": 1405, "lr": 5.626953749371805e-05}
{"train_loss": 0.1455731838941574, "global_step": 125132, "epoch": 1405, "lr": 5.626896233230291e-05}
{"train_loss": 0.1361362649902199, "global_step": 125133, "epoch": 1405, "lr": 5.626838717004498e-05, "val_loss": 5.411198616027832, "train_action_mse_error": 8.362442016601562}
{"train_loss": 0.09664875268936157, "global_step": 125134, "epoch": 1406, "lr": 5.626781200694432e-05}
{"train_loss": 0.057599227875471115, "global_step": 125135, "epoch": 1406, "lr": 5.626723684300105e-05}
{"train_loss": 0.13382728397846222, "global_step": 125136, "epoch": 1406, "lr": 5.6266661678215216e-05}
{"train_loss": 0.1556505560874939, "global_step": 125137, "epoch": 1406, "lr": 5.6266086512586916e-05}
{"train_loss": 0.12450257688760757, "global_step": 125138, "epoch": 1406, "lr": 5.6265511346116204e-05}
{"train_loss": 0.12009022384881973, "global_step": 125139, "epoch": 1406, "lr": 5.6264936178803176e-05}
{"train_loss": 0.11144354939460754, "global_step": 125140, "epoch": 1406, "lr": 5.626436101064791e-05}
{"train_loss": 0.10077309608459473, "global_step": 125141, "epoch": 1406, "lr": 5.6263785841650474e-05}
{"train_loss": 0.13965661823749542, "global_step": 125142, "epoch": 1406, "lr": 5.626321067181095e-05}
{"train_loss": 0.15823814272880554, "global_step": 125143, "epoch": 1406, "lr": 5.62626355011294e-05}
{"train_loss": 0.1736089587211609, "global_step": 125144, "epoch": 1406, "lr": 5.6262060329605926e-05}
{"train_loss": 0.1661282479763031, "global_step": 125145, "epoch": 1406, "lr": 5.62614851572406e-05}
{"train_loss": 0.13144539296627045, "global_step": 125146, "epoch": 1406, "lr": 5.6260909984033486e-05}
{"train_loss": 0.12271358072757721, "global_step": 125147, "epoch": 1406, "lr": 5.626033480998467e-05}
{"train_loss": 0.11019081622362137, "global_step": 125148, "epoch": 1406, "lr": 5.625975963509423e-05}
{"train_loss": 0.12304447591304779, "global_step": 125149, "epoch": 1406, "lr": 5.625918445936224e-05}
{"train_loss": 0.1259869486093521, "global_step": 125150, "epoch": 1406, "lr": 5.6258609282788775e-05}
{"train_loss": 0.16464337706565857, "global_step": 125151, "epoch": 1406, "lr": 5.6258034105373937e-05}
{"train_loss": 0.19334864616394043, "global_step": 125152, "epoch": 1406, "lr": 5.625745892711776e-05}
{"train_loss": 0.14410947263240814, "global_step": 125153, "epoch": 1406, "lr": 5.625688374802034e-05}
{"train_loss": 0.11814937740564346, "global_step": 125154, "epoch": 1406, "lr": 5.625630856808177e-05}
{"train_loss": 0.2158861607313156, "global_step": 125155, "epoch": 1406, "lr": 5.625573338730211e-05}
{"train_loss": 0.11739164590835571, "global_step": 125156, "epoch": 1406, "lr": 5.6255158205681455e-05}
{"train_loss": 0.07429218292236328, "global_step": 125157, "epoch": 1406, "lr": 5.6254583023219865e-05}
{"train_loss": 0.11424202471971512, "global_step": 125158, "epoch": 1406, "lr": 5.625400783991741e-05}
{"train_loss": 0.13165931403636932, "global_step": 125159, "epoch": 1406, "lr": 5.6253432655774195e-05}
{"train_loss": 0.10051668435335159, "global_step": 125160, "epoch": 1406, "lr": 5.625285747079028e-05}
{"train_loss": 0.10992433875799179, "global_step": 125161, "epoch": 1406, "lr": 5.6252282284965744e-05}
{"train_loss": 0.1629759669303894, "global_step": 125162, "epoch": 1406, "lr": 5.6251707098300655e-05}
{"train_loss": 0.1213446632027626, "global_step": 125163, "epoch": 1406, "lr": 5.6251131910795106e-05}
{"train_loss": 0.12263207137584686, "global_step": 125164, "epoch": 1406, "lr": 5.6250556722449166e-05}
{"train_loss": 0.15716944634914398, "global_step": 125165, "epoch": 1406, "lr": 5.624998153326292e-05}
{"train_loss": 0.1528865098953247, "global_step": 125166, "epoch": 1406, "lr": 5.6249406343236446e-05}
{"train_loss": 0.1620478481054306, "global_step": 125167, "epoch": 1406, "lr": 5.624883115236981e-05}
{"train_loss": 0.13729380071163177, "global_step": 125168, "epoch": 1406, "lr": 5.6248255960663086e-05}
{"train_loss": 0.10795341432094574, "global_step": 125169, "epoch": 1406, "lr": 5.624768076811637e-05}
{"train_loss": 0.12448664754629135, "global_step": 125170, "epoch": 1406, "lr": 5.624710557472973e-05}
{"train_loss": 0.14773158729076385, "global_step": 125171, "epoch": 1406, "lr": 5.624653038050324e-05}
{"train_loss": 0.12019521743059158, "global_step": 125172, "epoch": 1406, "lr": 5.624595518543697e-05}
{"train_loss": 0.11665941774845123, "global_step": 125173, "epoch": 1406, "lr": 5.624537998953102e-05}
{"train_loss": 0.12024028599262238, "global_step": 125174, "epoch": 1406, "lr": 5.624480479278545e-05}
{"train_loss": 0.11153291165828705, "global_step": 125175, "epoch": 1406, "lr": 5.624422959520035e-05}
{"train_loss": 0.1936318427324295, "global_step": 125176, "epoch": 1406, "lr": 5.624365439677578e-05}
{"train_loss": 0.19024142622947693, "global_step": 125177, "epoch": 1406, "lr": 5.6243079197511827e-05}
{"train_loss": 0.11912195384502411, "global_step": 125178, "epoch": 1406, "lr": 5.624250399740858e-05}
{"train_loss": 0.15113501250743866, "global_step": 125179, "epoch": 1406, "lr": 5.6241928796466094e-05}
{"train_loss": 0.14457355439662933, "global_step": 125180, "epoch": 1406, "lr": 5.6241353594684456e-05}
{"train_loss": 0.19021663069725037, "global_step": 125181, "epoch": 1406, "lr": 5.624077839206374e-05}
{"train_loss": 0.12169866263866425, "global_step": 125182, "epoch": 1406, "lr": 5.6240203188604034e-05}
{"train_loss": 0.1369808465242386, "global_step": 125183, "epoch": 1406, "lr": 5.623962798430541e-05}
{"train_loss": 0.14520959556102753, "global_step": 125184, "epoch": 1406, "lr": 5.6239052779167946e-05}
{"train_loss": 0.13084785640239716, "global_step": 125185, "epoch": 1406, "lr": 5.623847757319172e-05}
{"train_loss": 0.1338767409324646, "global_step": 125186, "epoch": 1406, "lr": 5.6237902366376794e-05}
{"train_loss": 0.11709336936473846, "global_step": 125187, "epoch": 1406, "lr": 5.6237327158723266e-05}
{"train_loss": 0.1209288015961647, "global_step": 125188, "epoch": 1406, "lr": 5.62367519502312e-05}
{"train_loss": 0.1703912764787674, "global_step": 125189, "epoch": 1406, "lr": 5.623617674090069e-05}
{"train_loss": 0.1269935667514801, "global_step": 125190, "epoch": 1406, "lr": 5.6235601530731793e-05}
{"train_loss": 0.1407037377357483, "global_step": 125191, "epoch": 1406, "lr": 5.623502631972459e-05}
{"train_loss": 0.181630477309227, "global_step": 125192, "epoch": 1406, "lr": 5.6234451107879184e-05}
{"train_loss": 0.1478208750486374, "global_step": 125193, "epoch": 1406, "lr": 5.623387589519561e-05}
{"train_loss": 0.1445159912109375, "global_step": 125194, "epoch": 1406, "lr": 5.6233300681673985e-05}
{"train_loss": 0.1659495234489441, "global_step": 125195, "epoch": 1406, "lr": 5.623272546731435e-05}
{"train_loss": 0.1410229206085205, "global_step": 125196, "epoch": 1406, "lr": 5.623215025211681e-05}
{"train_loss": 0.22357502579689026, "global_step": 125197, "epoch": 1406, "lr": 5.623157503608144e-05}
{"train_loss": 0.14111371338367462, "global_step": 125198, "epoch": 1406, "lr": 5.623099981920831e-05}
{"train_loss": 0.08449933677911758, "global_step": 125199, "epoch": 1406, "lr": 5.623042460149749e-05}
{"train_loss": 0.20499955117702484, "global_step": 125200, "epoch": 1406, "lr": 5.622984938294906e-05}
{"train_loss": 0.10938915610313416, "global_step": 125201, "epoch": 1406, "lr": 5.6229274163563115e-05}
{"train_loss": 0.18416860699653625, "global_step": 125202, "epoch": 1406, "lr": 5.622869894333972e-05}
{"train_loss": 0.11473413556814194, "global_step": 125203, "epoch": 1406, "lr": 5.622812372227895e-05}
{"train_loss": 0.10401871055364609, "global_step": 125204, "epoch": 1406, "lr": 5.622754850038088e-05}
{"train_loss": 0.10215449333190918, "global_step": 125205, "epoch": 1406, "lr": 5.62269732776456e-05}
{"train_loss": 0.09437213093042374, "global_step": 125206, "epoch": 1406, "lr": 5.622639805407317e-05}
{"train_loss": 0.09664082527160645, "global_step": 125207, "epoch": 1406, "lr": 5.622582282966369e-05}
{"train_loss": 0.14021718502044678, "global_step": 125208, "epoch": 1406, "lr": 5.6225247604417206e-05}
{"train_loss": 0.2294875681400299, "global_step": 125209, "epoch": 1406, "lr": 5.622467237833382e-05}
{"train_loss": 0.17660462856292725, "global_step": 125210, "epoch": 1406, "lr": 5.622409715141361e-05}
{"train_loss": 0.14925618469715118, "global_step": 125211, "epoch": 1406, "lr": 5.6223521923656644e-05}
{"train_loss": 0.13333678245544434, "global_step": 125212, "epoch": 1406, "lr": 5.6222946695063e-05}
{"train_loss": 0.12497926503419876, "global_step": 125213, "epoch": 1406, "lr": 5.6222371465632754e-05}
{"train_loss": 0.1519210785627365, "global_step": 125214, "epoch": 1406, "lr": 5.622179623536599e-05}
{"train_loss": 0.10355032235383987, "global_step": 125215, "epoch": 1406, "lr": 5.622122100426277e-05}
{"train_loss": 0.10989683121442795, "global_step": 125216, "epoch": 1406, "lr": 5.6220645772323196e-05}
{"train_loss": 0.1449127495288849, "global_step": 125217, "epoch": 1406, "lr": 5.622007053954732e-05}
{"train_loss": 0.21559672057628632, "global_step": 125218, "epoch": 1406, "lr": 5.621949530593524e-05}
{"train_loss": 0.18246705830097198, "global_step": 125219, "epoch": 1406, "lr": 5.6218920071487026e-05}
{"train_loss": 0.07377363741397858, "global_step": 125220, "epoch": 1406, "lr": 5.6218344836202754e-05}
{"train_loss": 0.09936211258172989, "global_step": 125221, "epoch": 1406, "lr": 5.6217769600082495e-05}
{"train_loss": 0.13697581240132953, "global_step": 125222, "epoch": 1406, "lr": 5.6217194363126336e-05, "val_loss": 5.392117023468018}
{"train_loss": 0.10391601920127869, "global_step": 125223, "epoch": 1407, "lr": 5.621661912533436e-05}
{"train_loss": 0.1319645792245865, "global_step": 125224, "epoch": 1407, "lr": 5.6216043886706625e-05}
{"train_loss": 0.08116891980171204, "global_step": 125225, "epoch": 1407, "lr": 5.621546864724322e-05}
{"train_loss": 0.15109732747077942, "global_step": 125226, "epoch": 1407, "lr": 5.621489340694422e-05}
{"train_loss": 0.17344486713409424, "global_step": 125227, "epoch": 1407, "lr": 5.62143181658097e-05}
{"train_loss": 0.15010887384414673, "global_step": 125228, "epoch": 1407, "lr": 5.621374292383975e-05}
{"train_loss": 0.18018880486488342, "global_step": 125229, "epoch": 1407, "lr": 5.621316768103443e-05}
{"train_loss": 0.0895216315984726, "global_step": 125230, "epoch": 1407, "lr": 5.621259243739383e-05}
{"train_loss": 0.2661900222301483, "global_step": 125231, "epoch": 1407, "lr": 5.621201719291802e-05}
{"train_loss": 0.18358924984931946, "global_step": 125232, "epoch": 1407, "lr": 5.621144194760708e-05}
{"train_loss": 0.12708152830600739, "global_step": 125233, "epoch": 1407, "lr": 5.6210866701461094e-05}
{"train_loss": 0.13316039741039276, "global_step": 125234, "epoch": 1407, "lr": 5.6210291454480124e-05}
{"train_loss": 0.13501982390880585, "global_step": 125235, "epoch": 1407, "lr": 5.6209716206664264e-05}
{"train_loss": 0.13760872185230255, "global_step": 125236, "epoch": 1407, "lr": 5.6209140958013575e-05}
{"train_loss": 0.17457681894302368, "global_step": 125237, "epoch": 1407, "lr": 5.6208565708528146e-05}
{"train_loss": 0.08960682898759842, "global_step": 125238, "epoch": 1407, "lr": 5.620799045820805e-05}
{"train_loss": 0.16012486815452576, "global_step": 125239, "epoch": 1407, "lr": 5.620741520705337e-05}
{"train_loss": 0.08040833473205566, "global_step": 125240, "epoch": 1407, "lr": 5.620683995506417e-05}
{"train_loss": 0.08087529242038727, "global_step": 125241, "epoch": 1407, "lr": 5.620626470224054e-05}
{"train_loss": 0.19412721693515778, "global_step": 125242, "epoch": 1407, "lr": 5.6205689448582556e-05}
{"train_loss": 0.13521063327789307, "global_step": 125243, "epoch": 1407, "lr": 5.620511419409029e-05}
{"train_loss": 0.13663619756698608, "global_step": 125244, "epoch": 1407, "lr": 5.620453893876383e-05}
{"train_loss": 0.1802654266357422, "global_step": 125245, "epoch": 1407, "lr": 5.620396368260323e-05}
{"train_loss": 0.13901835680007935, "global_step": 125246, "epoch": 1407, "lr": 5.6203388425608596e-05}
{"train_loss": 0.1364584118127823, "global_step": 125247, "epoch": 1407, "lr": 5.620281316777999e-05}
{"train_loss": 0.18011100590229034, "global_step": 125248, "epoch": 1407, "lr": 5.6202237909117484e-05}
{"train_loss": 0.06907172501087189, "global_step": 125249, "epoch": 1407, "lr": 5.620166264962117e-05}
{"train_loss": 0.1133018285036087, "global_step": 125250, "epoch": 1407, "lr": 5.6201087389291115e-05}
{"train_loss": 0.08144895732402802, "global_step": 125251, "epoch": 1407, "lr": 5.6200512128127406e-05}
{"train_loss": 0.13030508160591125, "global_step": 125252, "epoch": 1407, "lr": 5.61999368661301e-05}
{"train_loss": 0.14078935980796814, "global_step": 125253, "epoch": 1407, "lr": 5.619936160329931e-05}
{"train_loss": 0.18344850838184357, "global_step": 125254, "epoch": 1407, "lr": 5.619878633963507e-05}
{"train_loss": 0.1252693384885788, "global_step": 125255, "epoch": 1407, "lr": 5.6198211075137484e-05}
{"train_loss": 0.09521615505218506, "global_step": 125256, "epoch": 1407, "lr": 5.619763580980662e-05}
{"train_loss": 0.222435861825943, "global_step": 125257, "epoch": 1407, "lr": 5.6197060543642566e-05}
{"train_loss": 0.12456312775611877, "global_step": 125258, "epoch": 1407, "lr": 5.6196485276645394e-05}
{"train_loss": 0.11556785553693771, "global_step": 125259, "epoch": 1407, "lr": 5.6195910008815176e-05}
{"train_loss": 0.14475373923778534, "global_step": 125260, "epoch": 1407, "lr": 5.6195334740152006e-05}
{"train_loss": 0.1346348375082016, "global_step": 125261, "epoch": 1407, "lr": 5.6194759470655924e-05}
{"train_loss": 0.11510618031024933, "global_step": 125262, "epoch": 1407, "lr": 5.619418420032706e-05}
{"train_loss": 0.125568225979805, "global_step": 125263, "epoch": 1407, "lr": 5.619360892916544e-05}
{"train_loss": 0.19662784039974213, "global_step": 125264, "epoch": 1407, "lr": 5.619303365717118e-05}
{"train_loss": 0.11885193735361099, "global_step": 125265, "epoch": 1407, "lr": 5.619245838434434e-05}
{"train_loss": 0.14363977313041687, "global_step": 125266, "epoch": 1407, "lr": 5.6191883110684994e-05}
{"train_loss": 0.08152595907449722, "global_step": 125267, "epoch": 1407, "lr": 5.619130783619323e-05}
{"train_loss": 0.15048663318157196, "global_step": 125268, "epoch": 1407, "lr": 5.6190732560869105e-05}
{"train_loss": 0.09664034843444824, "global_step": 125269, "epoch": 1407, "lr": 5.619015728471273e-05}
{"train_loss": 0.09649825096130371, "global_step": 125270, "epoch": 1407, "lr": 5.618958200772416e-05}
{"train_loss": 0.14613084495067596, "global_step": 125271, "epoch": 1407, "lr": 5.618900672990347e-05}
{"train_loss": 0.13830827176570892, "global_step": 125272, "epoch": 1407, "lr": 5.618843145125074e-05}
{"train_loss": 0.15443135797977448, "global_step": 125273, "epoch": 1407, "lr": 5.6187856171766075e-05}
{"train_loss": 0.16877180337905884, "global_step": 125274, "epoch": 1407, "lr": 5.61872808914495e-05}
{"train_loss": 0.1660587042570114, "global_step": 125275, "epoch": 1407, "lr": 5.618670561030114e-05}
{"train_loss": 0.15932567417621613, "global_step": 125276, "epoch": 1407, "lr": 5.618613032832104e-05}
{"train_loss": 0.09905531257390976, "global_step": 125277, "epoch": 1407, "lr": 5.6185555045509295e-05}
{"train_loss": 0.14932683110237122, "global_step": 125278, "epoch": 1407, "lr": 5.6184979761865974e-05}
{"train_loss": 0.1894085705280304, "global_step": 125279, "epoch": 1407, "lr": 5.618440447739115e-05}
{"train_loss": 0.08722590655088425, "global_step": 125280, "epoch": 1407, "lr": 5.618382919208492e-05}
{"train_loss": 0.19613221287727356, "global_step": 125281, "epoch": 1407, "lr": 5.618325390594736e-05}
{"train_loss": 0.10669442266225815, "global_step": 125282, "epoch": 1407, "lr": 5.618267861897852e-05}
{"train_loss": 0.08329055458307266, "global_step": 125283, "epoch": 1407, "lr": 5.6182103331178505e-05}
{"train_loss": 0.08558771759271622, "global_step": 125284, "epoch": 1407, "lr": 5.618152804254738e-05}
{"train_loss": 0.10308254510164261, "global_step": 125285, "epoch": 1407, "lr": 5.6180952753085215e-05}
{"train_loss": 0.12550005316734314, "global_step": 125286, "epoch": 1407, "lr": 5.61803774627921e-05}
{"train_loss": 0.14578697085380554, "global_step": 125287, "epoch": 1407, "lr": 5.617980217166812e-05}
{"train_loss": 0.1269831359386444, "global_step": 125288, "epoch": 1407, "lr": 5.617922687971332e-05}
{"train_loss": 0.10310012847185135, "global_step": 125289, "epoch": 1407, "lr": 5.617865158692781e-05}
{"train_loss": 0.10696368664503098, "global_step": 125290, "epoch": 1407, "lr": 5.617807629331166e-05}
{"train_loss": 0.1174144595861435, "global_step": 125291, "epoch": 1407, "lr": 5.617750099886494e-05}
{"train_loss": 0.05913882330060005, "global_step": 125292, "epoch": 1407, "lr": 5.6176925703587724e-05}
{"train_loss": 0.11413837224245071, "global_step": 125293, "epoch": 1407, "lr": 5.61763504074801e-05}
{"train_loss": 0.13113851845264435, "global_step": 125294, "epoch": 1407, "lr": 5.617577511054214e-05}
{"train_loss": 0.11856570094823837, "global_step": 125295, "epoch": 1407, "lr": 5.617519981277393e-05}
{"train_loss": 0.09098564088344574, "global_step": 125296, "epoch": 1407, "lr": 5.617462451417552e-05}
{"train_loss": 0.25195741653442383, "global_step": 125297, "epoch": 1407, "lr": 5.617404921474702e-05}
{"train_loss": 0.13783852756023407, "global_step": 125298, "epoch": 1407, "lr": 5.617347391448848e-05}
{"train_loss": 0.10160840302705765, "global_step": 125299, "epoch": 1407, "lr": 5.617289861340002e-05}
{"train_loss": 0.12339125573635101, "global_step": 125300, "epoch": 1407, "lr": 5.6172323311481676e-05}
{"train_loss": 0.11988437175750732, "global_step": 125301, "epoch": 1407, "lr": 5.617174800873354e-05}
{"train_loss": 0.17396216094493866, "global_step": 125302, "epoch": 1407, "lr": 5.6171172705155686e-05}
{"train_loss": 0.06914863735437393, "global_step": 125303, "epoch": 1407, "lr": 5.617059740074819e-05}
{"train_loss": 0.14865583181381226, "global_step": 125304, "epoch": 1407, "lr": 5.617002209551113e-05}
{"train_loss": 0.0989566221833229, "global_step": 125305, "epoch": 1407, "lr": 5.616944678944459e-05}
{"train_loss": 0.14234334230422974, "global_step": 125306, "epoch": 1407, "lr": 5.6168871482548643e-05}
{"train_loss": 0.14380839467048645, "global_step": 125307, "epoch": 1407, "lr": 5.616829617482338e-05}
{"train_loss": 0.118229940533638, "global_step": 125308, "epoch": 1407, "lr": 5.6167720866268844e-05}
{"train_loss": 0.10011865198612213, "global_step": 125309, "epoch": 1407, "lr": 5.6167145556885134e-05}
{"train_loss": 0.12333866953849792, "global_step": 125310, "epoch": 1407, "lr": 5.6166570246672335e-05}
{"train_loss": 0.13273674731006782, "global_step": 125311, "epoch": 1407, "lr": 5.6165994935630514e-05, "val_loss": 5.611295223236084}
{"train_loss": 0.23046347498893738, "global_step": 125312, "epoch": 1408, "lr": 5.616541962375975e-05}
{"train_loss": 0.1161004826426506, "global_step": 125313, "epoch": 1408, "lr": 5.616484431106012e-05}
{"train_loss": 0.09993802011013031, "global_step": 125314, "epoch": 1408, "lr": 5.61642689975317e-05}
{"train_loss": 0.19701147079467773, "global_step": 125315, "epoch": 1408, "lr": 5.616369368317456e-05}
{"train_loss": 0.15454541146755219, "global_step": 125316, "epoch": 1408, "lr": 5.6163118367988795e-05}
{"train_loss": 0.1429368555545807, "global_step": 125317, "epoch": 1408, "lr": 5.6162543051974484e-05}
{"train_loss": 0.17143896222114563, "global_step": 125318, "epoch": 1408, "lr": 5.616196773513168e-05}
{"train_loss": 0.12899814546108246, "global_step": 125319, "epoch": 1408, "lr": 5.616139241746048e-05}
{"train_loss": 0.14599891006946564, "global_step": 125320, "epoch": 1408, "lr": 5.616081709896097e-05}
{"train_loss": 0.12285081297159195, "global_step": 125321, "epoch": 1408, "lr": 5.6160241779633204e-05}
{"train_loss": 0.1739448457956314, "global_step": 125322, "epoch": 1408, "lr": 5.615966645947727e-05}
{"train_loss": 0.166864275932312, "global_step": 125323, "epoch": 1408, "lr": 5.615909113849324e-05}
{"train_loss": 0.1661035567522049, "global_step": 125324, "epoch": 1408, "lr": 5.615851581668119e-05}
{"train_loss": 0.13087689876556396, "global_step": 125325, "epoch": 1408, "lr": 5.6157940494041205e-05}
{"train_loss": 0.09384564310312271, "global_step": 125326, "epoch": 1408, "lr": 5.6157365170573363e-05}
{"train_loss": 0.14193980395793915, "global_step": 125327, "epoch": 1408, "lr": 5.615678984627773e-05}
{"train_loss": 0.051060065627098083, "global_step": 125328, "epoch": 1408, "lr": 5.61562145211544e-05}
{"train_loss": 0.12162189930677414, "global_step": 125329, "epoch": 1408, "lr": 5.615563919520345e-05}
{"train_loss": 0.13948363065719604, "global_step": 125330, "epoch": 1408, "lr": 5.6155063868424926e-05}
{"train_loss": 0.10576607286930084, "global_step": 125331, "epoch": 1408, "lr": 5.615448854081895e-05}
{"train_loss": 0.09127721190452576, "global_step": 125332, "epoch": 1408, "lr": 5.615391321238557e-05}
{"train_loss": 0.1667148619890213, "global_step": 125333, "epoch": 1408, "lr": 5.615333788312487e-05}
{"train_loss": 0.16800768673419952, "global_step": 125334, "epoch": 1408, "lr": 5.615276255303692e-05}
{"train_loss": 0.10158269107341766, "global_step": 125335, "epoch": 1408, "lr": 5.615218722212182e-05}
{"train_loss": 0.156264990568161, "global_step": 125336, "epoch": 1408, "lr": 5.615161189037963e-05}
{"train_loss": 0.09539079666137695, "global_step": 125337, "epoch": 1408, "lr": 5.6151036557810434e-05}
{"train_loss": 0.13190607726573944, "global_step": 125338, "epoch": 1408, "lr": 5.6150461224414295e-05}
{"train_loss": 0.10451620072126389, "global_step": 125339, "epoch": 1408, "lr": 5.6149885890191324e-05}
{"train_loss": 0.11070875078439713, "global_step": 125340, "epoch": 1408, "lr": 5.6149310555141553e-05}
{"train_loss": 0.09876624494791031, "global_step": 125341, "epoch": 1408, "lr": 5.614873521926509e-05}
{"train_loss": 0.12642036378383636, "global_step": 125342, "epoch": 1408, "lr": 5.6148159882562015e-05}
{"train_loss": 0.07492756098508835, "global_step": 125343, "epoch": 1408, "lr": 5.614758454503238e-05}
{"train_loss": 0.09929373115301132, "global_step": 125344, "epoch": 1408, "lr": 5.6147009206676284e-05}
{"train_loss": 0.1465943455696106, "global_step": 125345, "epoch": 1408, "lr": 5.61464338674938e-05}
{"train_loss": 0.16492949426174164, "global_step": 125346, "epoch": 1408, "lr": 5.6145858527484996e-05}
{"train_loss": 0.1525530070066452, "global_step": 125347, "epoch": 1408, "lr": 5.614528318664996e-05}
{"train_loss": 0.18470266461372375, "global_step": 125348, "epoch": 1408, "lr": 5.6144707844988756e-05}
{"train_loss": 0.11192124336957932, "global_step": 125349, "epoch": 1408, "lr": 5.614413250250148e-05}
{"train_loss": 0.10327166318893433, "global_step": 125350, "epoch": 1408, "lr": 5.61435571591882e-05}
{"train_loss": 0.12105248868465424, "global_step": 125351, "epoch": 1408, "lr": 5.614298181504899e-05}
{"train_loss": 0.1592094451189041, "global_step": 125352, "epoch": 1408, "lr": 5.614240647008393e-05}
{"train_loss": 0.08815281838178635, "global_step": 125353, "epoch": 1408, "lr": 5.6141831124293096e-05}
{"train_loss": 0.10527230054140091, "global_step": 125354, "epoch": 1408, "lr": 5.614125577767657e-05}
{"train_loss": 0.13284732401371002, "global_step": 125355, "epoch": 1408, "lr": 5.6140680430234436e-05}
{"train_loss": 0.14814861118793488, "global_step": 125356, "epoch": 1408, "lr": 5.614010508196676e-05}
{"train_loss": 0.20636165142059326, "global_step": 125357, "epoch": 1408, "lr": 5.613952973287362e-05}
{"train_loss": 0.08641868829727173, "global_step": 125358, "epoch": 1408, "lr": 5.61389543829551e-05}
{"train_loss": 0.20941397547721863, "global_step": 125359, "epoch": 1408, "lr": 5.613837903221126e-05}
{"train_loss": 0.08499565720558167, "global_step": 125360, "epoch": 1408, "lr": 5.6137803680642206e-05}
{"train_loss": 0.13905739784240723, "global_step": 125361, "epoch": 1408, "lr": 5.613722832824799e-05}
{"train_loss": 0.11941255629062653, "global_step": 125362, "epoch": 1408, "lr": 5.6136652975028695e-05}
{"train_loss": 0.12416136264801025, "global_step": 125363, "epoch": 1408, "lr": 5.613607762098441e-05}
{"train_loss": 0.13415327668190002, "global_step": 125364, "epoch": 1408, "lr": 5.613550226611519e-05}
{"train_loss": 0.1695459932088852, "global_step": 125365, "epoch": 1408, "lr": 5.613492691042114e-05}
{"train_loss": 0.12627053260803223, "global_step": 125366, "epoch": 1408, "lr": 5.613435155390232e-05}
{"train_loss": 0.11774152517318726, "global_step": 125367, "epoch": 1408, "lr": 5.613377619655881e-05}
{"train_loss": 0.08930598944425583, "global_step": 125368, "epoch": 1408, "lr": 5.613320083839069e-05}
{"train_loss": 0.14559140801429749, "global_step": 125369, "epoch": 1408, "lr": 5.6132625479398036e-05}
{"train_loss": 0.17645983397960663, "global_step": 125370, "epoch": 1408, "lr": 5.6132050119580914e-05}
{"train_loss": 0.11918899416923523, "global_step": 125371, "epoch": 1408, "lr": 5.613147475893942e-05}
{"train_loss": 0.14553742110729218, "global_step": 125372, "epoch": 1408, "lr": 5.6130899397473625e-05}
{"train_loss": 0.2027634084224701, "global_step": 125373, "epoch": 1408, "lr": 5.61303240351836e-05}
{"train_loss": 0.21429726481437683, "global_step": 125374, "epoch": 1408, "lr": 5.612974867206944e-05}
{"train_loss": 0.12431313842535019, "global_step": 125375, "epoch": 1408, "lr": 5.61291733081312e-05}
{"train_loss": 0.17378564178943634, "global_step": 125376, "epoch": 1408, "lr": 5.6128597943368974e-05}
{"train_loss": 0.14859439432621002, "global_step": 125377, "epoch": 1408, "lr": 5.612802257778283e-05}
{"train_loss": 0.08648965507745743, "global_step": 125378, "epoch": 1408, "lr": 5.6127447211372867e-05}
{"train_loss": 0.15463615953922272, "global_step": 125379, "epoch": 1408, "lr": 5.612687184413912e-05}
{"train_loss": 0.13717685639858246, "global_step": 125380, "epoch": 1408, "lr": 5.61262964760817e-05}
{"train_loss": 0.16505354642868042, "global_step": 125381, "epoch": 1408, "lr": 5.612572110720067e-05}
{"train_loss": 0.12868814170360565, "global_step": 125382, "epoch": 1408, "lr": 5.6125145737496095e-05}
{"train_loss": 0.07827314734458923, "global_step": 125383, "epoch": 1408, "lr": 5.61245703669681e-05}
{"train_loss": 0.12550878524780273, "global_step": 125384, "epoch": 1408, "lr": 5.612399499561671e-05}
{"train_loss": 0.1263052523136139, "global_step": 125385, "epoch": 1408, "lr": 5.612341962344203e-05}
{"train_loss": 0.14015403389930725, "global_step": 125386, "epoch": 1408, "lr": 5.6122844250444127e-05}
{"train_loss": 0.12191904336214066, "global_step": 125387, "epoch": 1408, "lr": 5.612226887662308e-05}
{"train_loss": 0.14895012974739075, "global_step": 125388, "epoch": 1408, "lr": 5.612169350197897e-05}
{"train_loss": 0.10748950392007828, "global_step": 125389, "epoch": 1408, "lr": 5.6121118126511875e-05}
{"train_loss": 0.14245696365833282, "global_step": 125390, "epoch": 1408, "lr": 5.612054275022187e-05}
{"train_loss": 0.1314735859632492, "global_step": 125391, "epoch": 1408, "lr": 5.6119967373109036e-05}
{"train_loss": 0.1262526661157608, "global_step": 125392, "epoch": 1408, "lr": 5.6119391995173445e-05}
{"train_loss": 0.11661326140165329, "global_step": 125393, "epoch": 1408, "lr": 5.611881661641517e-05}
{"train_loss": 0.1274632066488266, "global_step": 125394, "epoch": 1408, "lr": 5.61182412368343e-05}
{"train_loss": 0.11859605461359024, "global_step": 125395, "epoch": 1408, "lr": 5.611766585643091e-05}
{"train_loss": 0.08513686805963516, "global_step": 125396, "epoch": 1408, "lr": 5.611709047520508e-05}
{"train_loss": 0.1512891799211502, "global_step": 125397, "epoch": 1408, "lr": 5.6116515093156885e-05}
{"train_loss": 0.15657256543636322, "global_step": 125398, "epoch": 1408, "lr": 5.611593971028639e-05}
{"train_loss": 0.13157488405704498, "global_step": 125399, "epoch": 1408, "lr": 5.6115364326593675e-05}
{"train_loss": 0.13383453774653123, "global_step": 125400, "epoch": 1408, "lr": 5.6114788942078833e-05, "val_loss": 5.514127254486084}
{"train_loss": 0.12301396578550339, "global_step": 125401, "epoch": 1409, "lr": 5.611421355674192e-05}
{"train_loss": 0.09007056802511215, "global_step": 125402, "epoch": 1409, "lr": 5.611363817058304e-05}
{"train_loss": 0.06734565645456314, "global_step": 125403, "epoch": 1409, "lr": 5.611306278360224e-05}
{"train_loss": 0.16355878114700317, "global_step": 125404, "epoch": 1409, "lr": 5.611248739579963e-05}
{"train_loss": 0.14011432230472565, "global_step": 125405, "epoch": 1409, "lr": 5.611191200717526e-05}
{"train_loss": 0.12520892918109894, "global_step": 125406, "epoch": 1409, "lr": 5.6111336617729213e-05}
{"train_loss": 0.14008109271526337, "global_step": 125407, "epoch": 1409, "lr": 5.611076122746158e-05}
{"train_loss": 0.14578109979629517, "global_step": 125408, "epoch": 1409, "lr": 5.611018583637243e-05}
{"train_loss": 0.11820296943187714, "global_step": 125409, "epoch": 1409, "lr": 5.610961044446183e-05}
{"train_loss": 0.08006387203931808, "global_step": 125410, "epoch": 1409, "lr": 5.6109035051729884e-05}
{"train_loss": 0.1079474687576294, "global_step": 125411, "epoch": 1409, "lr": 5.6108459658176646e-05}
{"train_loss": 0.08216357231140137, "global_step": 125412, "epoch": 1409, "lr": 5.61078842638022e-05}
{"train_loss": 0.09145662933588028, "global_step": 125413, "epoch": 1409, "lr": 5.6107308868606625e-05}
{"train_loss": 0.1192849650979042, "global_step": 125414, "epoch": 1409, "lr": 5.610673347258999e-05}
{"train_loss": 0.09115070104598999, "global_step": 125415, "epoch": 1409, "lr": 5.610615807575238e-05}
{"train_loss": 0.06559156626462936, "global_step": 125416, "epoch": 1409, "lr": 5.6105582678093884e-05}
{"train_loss": 0.12111718952655792, "global_step": 125417, "epoch": 1409, "lr": 5.6105007279614566e-05}
{"train_loss": 0.13981065154075623, "global_step": 125418, "epoch": 1409, "lr": 5.610443188031449e-05}
{"train_loss": 0.1035151332616806, "global_step": 125419, "epoch": 1409, "lr": 5.610385648019375e-05}
{"train_loss": 0.1292867362499237, "global_step": 125420, "epoch": 1409, "lr": 5.6103281079252424e-05}
{"train_loss": 0.1189136952161789, "global_step": 125421, "epoch": 1409, "lr": 5.610270567749058e-05}
{"train_loss": 0.19753175973892212, "global_step": 125422, "epoch": 1409, "lr": 5.610213027490831e-05}
{"train_loss": 0.13363079726696014, "global_step": 125423, "epoch": 1409, "lr": 5.610155487150569e-05}
{"train_loss": 0.14105738699436188, "global_step": 125424, "epoch": 1409, "lr": 5.610097946728277e-05}
{"train_loss": 0.09006024152040482, "global_step": 125425, "epoch": 1409, "lr": 5.6100404062239656e-05}
{"train_loss": 0.2045615315437317, "global_step": 125426, "epoch": 1409, "lr": 5.609982865637642e-05}
{"train_loss": 0.12951703369617462, "global_step": 125427, "epoch": 1409, "lr": 5.609925324969314e-05}
{"train_loss": 0.14960233867168427, "global_step": 125428, "epoch": 1409, "lr": 5.6098677842189886e-05}
{"train_loss": 0.10189925134181976, "global_step": 125429, "epoch": 1409, "lr": 5.609810243386674e-05}
{"train_loss": 0.08900918811559677, "global_step": 125430, "epoch": 1409, "lr": 5.609752702472377e-05}
{"train_loss": 0.10983862727880478, "global_step": 125431, "epoch": 1409, "lr": 5.609695161476107e-05}
{"train_loss": 0.08344219624996185, "global_step": 125432, "epoch": 1409, "lr": 5.6096376203978715e-05}
{"train_loss": 0.13856270909309387, "global_step": 125433, "epoch": 1409, "lr": 5.609580079237678e-05}
{"train_loss": 0.15673190355300903, "global_step": 125434, "epoch": 1409, "lr": 5.609522537995532e-05}
{"train_loss": 0.09635385870933533, "global_step": 125435, "epoch": 1409, "lr": 5.609464996671444e-05}
{"train_loss": 0.09352249652147293, "global_step": 125436, "epoch": 1409, "lr": 5.609407455265422e-05}
{"train_loss": 0.10211092233657837, "global_step": 125437, "epoch": 1409, "lr": 5.6093499137774705e-05}
{"train_loss": 0.14264188706874847, "global_step": 125438, "epoch": 1409, "lr": 5.609292372207601e-05}
{"train_loss": 0.17618350684642792, "global_step": 125439, "epoch": 1409, "lr": 5.609234830555819e-05}
{"train_loss": 0.06040068715810776, "global_step": 125440, "epoch": 1409, "lr": 5.609177288822133e-05}
{"train_loss": 0.09791848808526993, "global_step": 125441, "epoch": 1409, "lr": 5.60911974700655e-05}
{"train_loss": 0.09848473221063614, "global_step": 125442, "epoch": 1409, "lr": 5.6090622051090766e-05}
{"train_loss": 0.12615130841732025, "global_step": 125443, "epoch": 1409, "lr": 5.6090046631297246e-05}
{"train_loss": 0.18875914812088013, "global_step": 125444, "epoch": 1409, "lr": 5.6089471210684985e-05}
{"train_loss": 0.084870345890522, "global_step": 125445, "epoch": 1409, "lr": 5.608889578925408e-05}
{"train_loss": 0.1381572037935257, "global_step": 125446, "epoch": 1409, "lr": 5.608832036700459e-05}
{"train_loss": 0.21554943919181824, "global_step": 125447, "epoch": 1409, "lr": 5.6087744943936595e-05}
{"train_loss": 0.09281296283006668, "global_step": 125448, "epoch": 1409, "lr": 5.608716952005018e-05}
{"train_loss": 0.18351970613002777, "global_step": 125449, "epoch": 1409, "lr": 5.6086594095345426e-05}
{"train_loss": 0.08068053424358368, "global_step": 125450, "epoch": 1409, "lr": 5.608601866982239e-05}
{"train_loss": 0.16028261184692383, "global_step": 125451, "epoch": 1409, "lr": 5.6085443243481173e-05}
{"train_loss": 0.15241727232933044, "global_step": 125452, "epoch": 1409, "lr": 5.608486781632184e-05}
{"train_loss": 0.1317214071750641, "global_step": 125453, "epoch": 1409, "lr": 5.608429238834447e-05}
{"train_loss": 0.17141225934028625, "global_step": 125454, "epoch": 1409, "lr": 5.608371695954915e-05}
{"train_loss": 0.1041380912065506, "global_step": 125455, "epoch": 1409, "lr": 5.608314152993594e-05}
{"train_loss": 0.14305105805397034, "global_step": 125456, "epoch": 1409, "lr": 5.608256609950493e-05}
{"train_loss": 0.08276063948869705, "global_step": 125457, "epoch": 1409, "lr": 5.608199066825619e-05}
{"train_loss": 0.17606841027736664, "global_step": 125458, "epoch": 1409, "lr": 5.608141523618979e-05}
{"train_loss": 0.1527308076620102, "global_step": 125459, "epoch": 1409, "lr": 5.608083980330584e-05}
{"train_loss": 0.16176360845565796, "global_step": 125460, "epoch": 1409, "lr": 5.6080264369604386e-05}
{"train_loss": 0.1093493327498436, "global_step": 125461, "epoch": 1409, "lr": 5.6079688935085497e-05}
{"train_loss": 0.11687398701906204, "global_step": 125462, "epoch": 1409, "lr": 5.607911349974928e-05}
{"train_loss": 0.1297532021999359, "global_step": 125463, "epoch": 1409, "lr": 5.6078538063595795e-05}
{"train_loss": 0.1408304125070572, "global_step": 125464, "epoch": 1409, "lr": 5.607796262662514e-05}
{"train_loss": 0.1785358488559723, "global_step": 125465, "epoch": 1409, "lr": 5.607738718883737e-05}
{"train_loss": 0.15621516108512878, "global_step": 125466, "epoch": 1409, "lr": 5.607681175023256e-05}
{"train_loss": 0.15445901453495026, "global_step": 125467, "epoch": 1409, "lr": 5.6076236310810814e-05}
{"train_loss": 0.11159487813711166, "global_step": 125468, "epoch": 1409, "lr": 5.6075660870572176e-05}
{"train_loss": 0.14336290955543518, "global_step": 125469, "epoch": 1409, "lr": 5.607508542951675e-05}
{"train_loss": 0.11107563227415085, "global_step": 125470, "epoch": 1409, "lr": 5.6074509987644595e-05}
{"train_loss": 0.16843895614147186, "global_step": 125471, "epoch": 1409, "lr": 5.607393454495581e-05}
{"train_loss": 0.21799308061599731, "global_step": 125472, "epoch": 1409, "lr": 5.607335910145044e-05}
{"train_loss": 0.12195313721895218, "global_step": 125473, "epoch": 1409, "lr": 5.607278365712859e-05}
{"train_loss": 0.12193916738033295, "global_step": 125474, "epoch": 1409, "lr": 5.607220821199034e-05}
{"train_loss": 0.22793899476528168, "global_step": 125475, "epoch": 1409, "lr": 5.6071632766035755e-05}
{"train_loss": 0.09985841810703278, "global_step": 125476, "epoch": 1409, "lr": 5.60710573192649e-05}
{"train_loss": 0.11395370960235596, "global_step": 125477, "epoch": 1409, "lr": 5.607048187167786e-05}
{"train_loss": 0.11837612092494965, "global_step": 125478, "epoch": 1409, "lr": 5.606990642327473e-05}
{"train_loss": 0.07820656150579453, "global_step": 125479, "epoch": 1409, "lr": 5.606933097405556e-05}
{"train_loss": 0.1481965184211731, "global_step": 125480, "epoch": 1409, "lr": 5.606875552402045e-05}
{"train_loss": 0.1099507063627243, "global_step": 125481, "epoch": 1409, "lr": 5.6068180073169476e-05}
{"train_loss": 0.11182479560375214, "global_step": 125482, "epoch": 1409, "lr": 5.60676046215027e-05}
{"train_loss": 0.12158045172691345, "global_step": 125483, "epoch": 1409, "lr": 5.606702916902021e-05}
{"train_loss": 0.15085619688034058, "global_step": 125484, "epoch": 1409, "lr": 5.606645371572209e-05}
{"train_loss": 0.17094916105270386, "global_step": 125485, "epoch": 1409, "lr": 5.6065878261608404e-05}
{"train_loss": 0.17227137088775635, "global_step": 125486, "epoch": 1409, "lr": 5.606530280667923e-05}
{"train_loss": 0.18830189108848572, "global_step": 125487, "epoch": 1409, "lr": 5.606472735093467e-05}
{"train_loss": 0.1445290446281433, "global_step": 125488, "epoch": 1409, "lr": 5.606415189437476e-05}
{"train_loss": 0.12947169749924306, "global_step": 125489, "epoch": 1409, "lr": 5.606357643699961e-05, "val_loss": 5.451012134552002}
{"train_loss": 0.13590900599956512, "global_step": 125490, "epoch": 1410, "lr": 5.606300097880928e-05}
{"train_loss": 0.24000750482082367, "global_step": 125491, "epoch": 1410, "lr": 5.6062425519803854e-05}
{"train_loss": 0.13374745845794678, "global_step": 125492, "epoch": 1410, "lr": 5.606185005998341e-05}
{"train_loss": 0.09643815457820892, "global_step": 125493, "epoch": 1410, "lr": 5.6061274599348025e-05}
{"train_loss": 0.13835348188877106, "global_step": 125494, "epoch": 1410, "lr": 5.606069913789778e-05}
{"train_loss": 0.07086040079593658, "global_step": 125495, "epoch": 1410, "lr": 5.6060123675632746e-05}
{"train_loss": 0.14754261076450348, "global_step": 125496, "epoch": 1410, "lr": 5.605954821255299e-05}
{"train_loss": 0.17912906408309937, "global_step": 125497, "epoch": 1410, "lr": 5.605897274865861e-05}
{"train_loss": 0.14174547791481018, "global_step": 125498, "epoch": 1410, "lr": 5.605839728394967e-05}
{"train_loss": 0.10355015099048615, "global_step": 125499, "epoch": 1410, "lr": 5.605782181842626e-05}
{"train_loss": 0.10701192170381546, "global_step": 125500, "epoch": 1410, "lr": 5.6057246352088443e-05}
{"train_loss": 0.13857261836528778, "global_step": 125501, "epoch": 1410, "lr": 5.6056670884936316e-05}
{"train_loss": 0.12250145524740219, "global_step": 125502, "epoch": 1410, "lr": 5.6056095416969925e-05}
{"train_loss": 0.16329902410507202, "global_step": 125503, "epoch": 1410, "lr": 5.605551994818938e-05}
{"train_loss": 0.13751962780952454, "global_step": 125504, "epoch": 1410, "lr": 5.605494447859474e-05}
{"train_loss": 0.15400376915931702, "global_step": 125505, "epoch": 1410, "lr": 5.6054369008186084e-05}
{"train_loss": 0.12683701515197754, "global_step": 125506, "epoch": 1410, "lr": 5.6053793536963495e-05}
{"train_loss": 0.13842220604419708, "global_step": 125507, "epoch": 1410, "lr": 5.605321806492705e-05}
{"train_loss": 0.1242206022143364, "global_step": 125508, "epoch": 1410, "lr": 5.6052642592076824e-05}
{"train_loss": 0.10817346721887589, "global_step": 125509, "epoch": 1410, "lr": 5.6052067118412876e-05}
{"train_loss": 0.1373320072889328, "global_step": 125510, "epoch": 1410, "lr": 5.6051491643935316e-05}
{"train_loss": 0.15074865520000458, "global_step": 125511, "epoch": 1410, "lr": 5.6050916168644205e-05}
{"train_loss": 0.07949014753103256, "global_step": 125512, "epoch": 1410, "lr": 5.6050340692539624e-05}
{"train_loss": 0.13537487387657166, "global_step": 125513, "epoch": 1410, "lr": 5.604976521562167e-05}
{"train_loss": 0.11762233078479767, "global_step": 125514, "epoch": 1410, "lr": 5.6049189737890375e-05}
{"train_loss": 0.13825446367263794, "global_step": 125515, "epoch": 1410, "lr": 5.6048614259345824e-05}
{"train_loss": 0.10815133899450302, "global_step": 125516, "epoch": 1410, "lr": 5.6048038779988135e-05}
{"train_loss": 0.10769936442375183, "global_step": 125517, "epoch": 1410, "lr": 5.604746329981735e-05}
{"train_loss": 0.15114258229732513, "global_step": 125518, "epoch": 1410, "lr": 5.604688781883356e-05}
{"train_loss": 0.15095104277133942, "global_step": 125519, "epoch": 1410, "lr": 5.6046312337036834e-05}
{"train_loss": 0.0956512838602066, "global_step": 125520, "epoch": 1410, "lr": 5.604573685442726e-05}
{"train_loss": 0.1561167687177658, "global_step": 125521, "epoch": 1410, "lr": 5.604516137100492e-05}
{"train_loss": 0.14282391965389252, "global_step": 125522, "epoch": 1410, "lr": 5.604458588676987e-05}
{"train_loss": 0.11464580148458481, "global_step": 125523, "epoch": 1410, "lr": 5.60440104017222e-05}
{"train_loss": 0.13832831382751465, "global_step": 125524, "epoch": 1410, "lr": 5.604343491586199e-05}
{"train_loss": 0.10600131750106812, "global_step": 125525, "epoch": 1410, "lr": 5.6042859429189324e-05}
{"train_loss": 0.07853294163942337, "global_step": 125526, "epoch": 1410, "lr": 5.604228394170425e-05}
{"train_loss": 0.09241365641355515, "global_step": 125527, "epoch": 1410, "lr": 5.6041708453406875e-05}
{"train_loss": 0.11392315477132797, "global_step": 125528, "epoch": 1410, "lr": 5.604113296429726e-05}
{"train_loss": 0.061163436621427536, "global_step": 125529, "epoch": 1410, "lr": 5.60405574743755e-05}
{"train_loss": 0.15697608888149261, "global_step": 125530, "epoch": 1410, "lr": 5.603998198364164e-05}
{"train_loss": 0.10900779813528061, "global_step": 125531, "epoch": 1410, "lr": 5.6039406492095806e-05}
{"train_loss": 0.10579320043325424, "global_step": 125532, "epoch": 1410, "lr": 5.6038830999738024e-05}
{"train_loss": 0.16674181818962097, "global_step": 125533, "epoch": 1410, "lr": 5.6038255506568405e-05}
{"train_loss": 0.150570347905159, "global_step": 125534, "epoch": 1410, "lr": 5.603768001258701e-05}
{"train_loss": 0.12267681211233139, "global_step": 125535, "epoch": 1410, "lr": 5.603710451779393e-05}
{"train_loss": 0.1358291506767273, "global_step": 125536, "epoch": 1410, "lr": 5.603652902218923e-05}
{"train_loss": 0.17936377227306366, "global_step": 125537, "epoch": 1410, "lr": 5.6035953525772996e-05}
{"train_loss": 0.11550099402666092, "global_step": 125538, "epoch": 1410, "lr": 5.6035378028545304e-05}
{"train_loss": 0.15025970339775085, "global_step": 125539, "epoch": 1410, "lr": 5.6034802530506215e-05}
{"train_loss": 0.13875652849674225, "global_step": 125540, "epoch": 1410, "lr": 5.603422703165584e-05}
{"train_loss": 0.07107554376125336, "global_step": 125541, "epoch": 1410, "lr": 5.6033651531994226e-05}
{"train_loss": 0.11271631717681885, "global_step": 125542, "epoch": 1410, "lr": 5.603307603152146e-05}
{"train_loss": 0.17122413218021393, "global_step": 125543, "epoch": 1410, "lr": 5.603250053023762e-05}
{"train_loss": 0.13270363211631775, "global_step": 125544, "epoch": 1410, "lr": 5.603192502814278e-05}
{"train_loss": 0.225411519408226, "global_step": 125545, "epoch": 1410, "lr": 5.6031349525237034e-05}
{"train_loss": 0.09724915772676468, "global_step": 125546, "epoch": 1410, "lr": 5.603077402152044e-05}
{"train_loss": 0.1448393613100052, "global_step": 125547, "epoch": 1410, "lr": 5.6030198516993084e-05}
{"train_loss": 0.11870583891868591, "global_step": 125548, "epoch": 1410, "lr": 5.602962301165503e-05}
{"train_loss": 0.1251576840877533, "global_step": 125549, "epoch": 1410, "lr": 5.602904750550638e-05}
{"train_loss": 0.11892364919185638, "global_step": 125550, "epoch": 1410, "lr": 5.6028471998547194e-05}
{"train_loss": 0.2951450049877167, "global_step": 125551, "epoch": 1410, "lr": 5.602789649077755e-05}
{"train_loss": 0.11518511176109314, "global_step": 125552, "epoch": 1410, "lr": 5.602732098219754e-05}
{"train_loss": 0.17828623950481415, "global_step": 125553, "epoch": 1410, "lr": 5.602674547280722e-05}
{"train_loss": 0.11540897190570831, "global_step": 125554, "epoch": 1410, "lr": 5.602616996260668e-05}
{"train_loss": 0.10912435501813889, "global_step": 125555, "epoch": 1410, "lr": 5.6025594451595986e-05}
{"train_loss": 0.12291087210178375, "global_step": 125556, "epoch": 1410, "lr": 5.602501893977523e-05}
{"train_loss": 0.09819935262203217, "global_step": 125557, "epoch": 1410, "lr": 5.602444342714449e-05}
{"train_loss": 0.10723119974136353, "global_step": 125558, "epoch": 1410, "lr": 5.602386791370383e-05}
{"train_loss": 0.0914866030216217, "global_step": 125559, "epoch": 1410, "lr": 5.602329239945334e-05}
{"train_loss": 0.10438661277294159, "global_step": 125560, "epoch": 1410, "lr": 5.602271688439309e-05}
{"train_loss": 0.14718176424503326, "global_step": 125561, "epoch": 1410, "lr": 5.6022141368523163e-05}
{"train_loss": 0.15591277182102203, "global_step": 125562, "epoch": 1410, "lr": 5.602156585184363e-05}
{"train_loss": 0.07285087555646896, "global_step": 125563, "epoch": 1410, "lr": 5.6020990334354564e-05}
{"train_loss": 0.09685281664133072, "global_step": 125564, "epoch": 1410, "lr": 5.602041481605605e-05}
{"train_loss": 0.09014160186052322, "global_step": 125565, "epoch": 1410, "lr": 5.601983929694817e-05}
{"train_loss": 0.2253159135580063, "global_step": 125566, "epoch": 1410, "lr": 5.601926377703099e-05}
{"train_loss": 0.1942799836397171, "global_step": 125567, "epoch": 1410, "lr": 5.601868825630461e-05}
{"train_loss": 0.15098106861114502, "global_step": 125568, "epoch": 1410, "lr": 5.601811273476908e-05}
{"train_loss": 0.09912934899330139, "global_step": 125569, "epoch": 1410, "lr": 5.6017537212424484e-05}
{"train_loss": 0.1413124054670334, "global_step": 125570, "epoch": 1410, "lr": 5.601696168927091e-05}
{"train_loss": 0.13957944512367249, "global_step": 125571, "epoch": 1410, "lr": 5.6016386165308423e-05}
{"train_loss": 0.13475558161735535, "global_step": 125572, "epoch": 1410, "lr": 5.60158106405371e-05}
{"train_loss": 0.14006859064102173, "global_step": 125573, "epoch": 1410, "lr": 5.601523511495704e-05}
{"train_loss": 0.13372518122196198, "global_step": 125574, "epoch": 1410, "lr": 5.6014659588568296e-05}
{"train_loss": 0.10754404962062836, "global_step": 125575, "epoch": 1410, "lr": 5.6014084061370965e-05}
{"train_loss": 0.1615336537361145, "global_step": 125576, "epoch": 1410, "lr": 5.60135085333651e-05}
{"train_loss": 0.13537167012691498, "global_step": 125577, "epoch": 1410, "lr": 5.6012933004550794e-05}
{"train_loss": 0.13166594593210168, "global_step": 125578, "epoch": 1410, "lr": 5.601235747492812e-05, "val_loss": 5.423599720001221, "train_action_mse_error": 20.19169044494629}
{"train_loss": 0.10220007598400116, "global_step": 125579, "epoch": 1411, "lr": 5.6011781944497165e-05}
{"train_loss": 0.10084091871976852, "global_step": 125580, "epoch": 1411, "lr": 5.6011206413257996e-05}
{"train_loss": 0.07814921438694, "global_step": 125581, "epoch": 1411, "lr": 5.6010630881210704e-05}
{"train_loss": 0.10108207911252975, "global_step": 125582, "epoch": 1411, "lr": 5.601005534835535e-05}
{"train_loss": 0.13101613521575928, "global_step": 125583, "epoch": 1411, "lr": 5.600947981469201e-05}
{"train_loss": 0.1036984771490097, "global_step": 125584, "epoch": 1411, "lr": 5.6008904280220775e-05}
{"train_loss": 0.11178653687238693, "global_step": 125585, "epoch": 1411, "lr": 5.6008328744941706e-05}
{"train_loss": 0.05889665335416794, "global_step": 125586, "epoch": 1411, "lr": 5.600775320885491e-05}
{"train_loss": 0.1630859076976776, "global_step": 125587, "epoch": 1411, "lr": 5.600717767196043e-05}
{"train_loss": 0.07755441963672638, "global_step": 125588, "epoch": 1411, "lr": 5.600660213425836e-05}
{"train_loss": 0.12977996468544006, "global_step": 125589, "epoch": 1411, "lr": 5.6006026595748775e-05}
{"train_loss": 0.09789072722196579, "global_step": 125590, "epoch": 1411, "lr": 5.600545105643176e-05}
{"train_loss": 0.16723845899105072, "global_step": 125591, "epoch": 1411, "lr": 5.6004875516307376e-05}
{"train_loss": 0.19081157445907593, "global_step": 125592, "epoch": 1411, "lr": 5.600429997537572e-05}
{"train_loss": 0.13712027668952942, "global_step": 125593, "epoch": 1411, "lr": 5.600372443363685e-05}
{"train_loss": 0.1049601286649704, "global_step": 125594, "epoch": 1411, "lr": 5.600314889109085e-05}
{"train_loss": 0.15356147289276123, "global_step": 125595, "epoch": 1411, "lr": 5.600257334773781e-05}
{"train_loss": 0.09837690740823746, "global_step": 125596, "epoch": 1411, "lr": 5.600199780357779e-05}
{"train_loss": 0.059711311012506485, "global_step": 125597, "epoch": 1411, "lr": 5.600142225861088e-05}
{"train_loss": 0.11832620203495026, "global_step": 125598, "epoch": 1411, "lr": 5.600084671283714e-05}
{"train_loss": 0.1056591048836708, "global_step": 125599, "epoch": 1411, "lr": 5.600027116625668e-05}
{"train_loss": 0.11150125414133072, "global_step": 125600, "epoch": 1411, "lr": 5.5999695618869544e-05}
{"train_loss": 0.16441549360752106, "global_step": 125601, "epoch": 1411, "lr": 5.599912007067582e-05}
{"train_loss": 0.047578468918800354, "global_step": 125602, "epoch": 1411, "lr": 5.599854452167559e-05}
{"train_loss": 0.07066340744495392, "global_step": 125603, "epoch": 1411, "lr": 5.599796897186893e-05}
{"train_loss": 0.07981091737747192, "global_step": 125604, "epoch": 1411, "lr": 5.599739342125592e-05}
{"train_loss": 0.1396343857049942, "global_step": 125605, "epoch": 1411, "lr": 5.5996817869836625e-05}
{"train_loss": 0.15072359144687653, "global_step": 125606, "epoch": 1411, "lr": 5.599624231761114e-05}
{"train_loss": 0.13856980204582214, "global_step": 125607, "epoch": 1411, "lr": 5.5995666764579525e-05}
{"train_loss": 0.26553893089294434, "global_step": 125608, "epoch": 1411, "lr": 5.599509121074187e-05}
{"train_loss": 0.11133857071399689, "global_step": 125609, "epoch": 1411, "lr": 5.5994515656098245e-05}
{"train_loss": 0.06524965167045593, "global_step": 125610, "epoch": 1411, "lr": 5.5993940100648726e-05}
{"train_loss": 0.10068129748106003, "global_step": 125611, "epoch": 1411, "lr": 5.59933645443934e-05}
{"train_loss": 0.15950293838977814, "global_step": 125612, "epoch": 1411, "lr": 5.5992788987332345e-05}
{"train_loss": 0.1464921236038208, "global_step": 125613, "epoch": 1411, "lr": 5.5992213429465624e-05}
{"train_loss": 0.1634475439786911, "global_step": 125614, "epoch": 1411, "lr": 5.5991637870793335e-05}
{"train_loss": 0.12618488073349, "global_step": 125615, "epoch": 1411, "lr": 5.5991062311315524e-05}
{"train_loss": 0.11364080756902695, "global_step": 125616, "epoch": 1411, "lr": 5.59904867510323e-05}
{"train_loss": 0.11195851862430573, "global_step": 125617, "epoch": 1411, "lr": 5.598991118994373e-05}
{"train_loss": 0.09398122131824493, "global_step": 125618, "epoch": 1411, "lr": 5.598933562804989e-05}
{"train_loss": 0.10041549056768417, "global_step": 125619, "epoch": 1411, "lr": 5.5988760065350854e-05}
{"train_loss": 0.09573045372962952, "global_step": 125620, "epoch": 1411, "lr": 5.5988184501846696e-05}
{"train_loss": 0.1551527976989746, "global_step": 125621, "epoch": 1411, "lr": 5.5987608937537504e-05}
{"train_loss": 0.04959610849618912, "global_step": 125622, "epoch": 1411, "lr": 5.598703337242336e-05}
{"train_loss": 0.13035354018211365, "global_step": 125623, "epoch": 1411, "lr": 5.5986457806504324e-05}
{"train_loss": 0.13457445800304413, "global_step": 125624, "epoch": 1411, "lr": 5.598588223978047e-05}
{"train_loss": 0.1835893988609314, "global_step": 125625, "epoch": 1411, "lr": 5.5985306672251904e-05}
{"train_loss": 0.08496953547000885, "global_step": 125626, "epoch": 1411, "lr": 5.598473110391869e-05}
{"train_loss": 0.14371922612190247, "global_step": 125627, "epoch": 1411, "lr": 5.5984155534780894e-05}
{"train_loss": 0.1518886536359787, "global_step": 125628, "epoch": 1411, "lr": 5.5983579964838594e-05}
{"train_loss": 0.05881531164050102, "global_step": 125629, "epoch": 1411, "lr": 5.598300439409188e-05}
{"train_loss": 0.14802786707878113, "global_step": 125630, "epoch": 1411, "lr": 5.5982428822540825e-05}
{"train_loss": 0.11655790358781815, "global_step": 125631, "epoch": 1411, "lr": 5.5981853250185504e-05}
{"train_loss": 0.07615829259157181, "global_step": 125632, "epoch": 1411, "lr": 5.5981277677026004e-05}
{"train_loss": 0.07729227840900421, "global_step": 125633, "epoch": 1411, "lr": 5.598070210306238e-05}
{"train_loss": 0.07447849959135056, "global_step": 125634, "epoch": 1411, "lr": 5.598012652829473e-05}
{"train_loss": 0.16838212311267853, "global_step": 125635, "epoch": 1411, "lr": 5.597955095272313e-05}
{"train_loss": 0.17533758282661438, "global_step": 125636, "epoch": 1411, "lr": 5.597897537634764e-05}
{"train_loss": 0.0783705860376358, "global_step": 125637, "epoch": 1411, "lr": 5.5978399799168365e-05}
{"train_loss": 0.11430168896913528, "global_step": 125638, "epoch": 1411, "lr": 5.597782422118536e-05}
{"train_loss": 0.13922342658042908, "global_step": 125639, "epoch": 1411, "lr": 5.5977248642398706e-05}
{"train_loss": 0.1474931240081787, "global_step": 125640, "epoch": 1411, "lr": 5.597667306280849e-05}
{"train_loss": 0.07819188386201859, "global_step": 125641, "epoch": 1411, "lr": 5.597609748241478e-05}
{"train_loss": 0.15405488014221191, "global_step": 125642, "epoch": 1411, "lr": 5.597552190121764e-05}
{"train_loss": 0.1321636587381363, "global_step": 125643, "epoch": 1411, "lr": 5.597494631921719e-05}
{"train_loss": 0.10669206827878952, "global_step": 125644, "epoch": 1411, "lr": 5.5974370736413475e-05}
{"train_loss": 0.16090641915798187, "global_step": 125645, "epoch": 1411, "lr": 5.5973795152806576e-05}
{"train_loss": 0.13097618520259857, "global_step": 125646, "epoch": 1411, "lr": 5.5973219568396574e-05}
{"train_loss": 0.09610159695148468, "global_step": 125647, "epoch": 1411, "lr": 5.597264398318355e-05}
{"train_loss": 0.17858093976974487, "global_step": 125648, "epoch": 1411, "lr": 5.597206839716757e-05}
{"train_loss": 0.057876236736774445, "global_step": 125649, "epoch": 1411, "lr": 5.597149281034871e-05}
{"train_loss": 0.18035726249217987, "global_step": 125650, "epoch": 1411, "lr": 5.597091722272707e-05}
{"train_loss": 0.152326300740242, "global_step": 125651, "epoch": 1411, "lr": 5.5970341634302703e-05}
{"train_loss": 0.16527287662029266, "global_step": 125652, "epoch": 1411, "lr": 5.59697660450757e-05}
{"train_loss": 0.13090072572231293, "global_step": 125653, "epoch": 1411, "lr": 5.596919045504614e-05}
{"train_loss": 0.08988439291715622, "global_step": 125654, "epoch": 1411, "lr": 5.5968614864214085e-05}
{"train_loss": 0.10257120430469513, "global_step": 125655, "epoch": 1411, "lr": 5.596803927257963e-05}
{"train_loss": 0.07789480686187744, "global_step": 125656, "epoch": 1411, "lr": 5.596746368014284e-05}
{"train_loss": 0.0864105150103569, "global_step": 125657, "epoch": 1411, "lr": 5.59668880869038e-05}
{"train_loss": 0.10413981229066849, "global_step": 125658, "epoch": 1411, "lr": 5.596631249286259e-05}
{"train_loss": 0.17687083780765533, "global_step": 125659, "epoch": 1411, "lr": 5.596573689801927e-05}
{"train_loss": 0.11639988422393799, "global_step": 125660, "epoch": 1411, "lr": 5.596516130237393e-05}
{"train_loss": 0.13406579196453094, "global_step": 125661, "epoch": 1411, "lr": 5.596458570592665e-05}
{"train_loss": 0.11187001317739487, "global_step": 125662, "epoch": 1411, "lr": 5.5964010108677514e-05}
{"train_loss": 0.1208663135766983, "global_step": 125663, "epoch": 1411, "lr": 5.596343451062659e-05}
{"train_loss": 0.12336001545190811, "global_step": 125664, "epoch": 1411, "lr": 5.596285891177395e-05}
{"train_loss": 0.11544272303581238, "global_step": 125665, "epoch": 1411, "lr": 5.5962283312119665e-05}
{"train_loss": 0.1231488287448883, "global_step": 125666, "epoch": 1411, "lr": 5.596170771166384e-05}
{"train_loss": 0.1206518838579735, "global_step": 125667, "epoch": 1411, "lr": 5.596113211040652e-05, "val_loss": 5.612776756286621}
{"train_loss": 0.10557117313146591, "global_step": 125668, "epoch": 1412, "lr": 5.596055650834782e-05}
{"train_loss": 0.10441810637712479, "global_step": 125669, "epoch": 1412, "lr": 5.595998090548777e-05}
{"train_loss": 0.19790063798427582, "global_step": 125670, "epoch": 1412, "lr": 5.5959405301826495e-05}
{"train_loss": 0.10561997443437576, "global_step": 125671, "epoch": 1412, "lr": 5.595882969736404e-05}
{"train_loss": 0.12348060309886932, "global_step": 125672, "epoch": 1412, "lr": 5.595825409210049e-05}
{"train_loss": 0.17521092295646667, "global_step": 125673, "epoch": 1412, "lr": 5.5957678486035936e-05}
{"train_loss": 0.11970361322164536, "global_step": 125674, "epoch": 1412, "lr": 5.595710287917044e-05}
{"train_loss": 0.11681288480758667, "global_step": 125675, "epoch": 1412, "lr": 5.595652727150408e-05}
{"train_loss": 0.16331319510936737, "global_step": 125676, "epoch": 1412, "lr": 5.5955951663036944e-05}
{"train_loss": 0.15748882293701172, "global_step": 125677, "epoch": 1412, "lr": 5.595537605376909e-05}
{"train_loss": 0.14222511649131775, "global_step": 125678, "epoch": 1412, "lr": 5.5954800443700606e-05}
{"train_loss": 0.17670968174934387, "global_step": 125679, "epoch": 1412, "lr": 5.595422483283159e-05}
{"train_loss": 0.22014473378658295, "global_step": 125680, "epoch": 1412, "lr": 5.59536492211621e-05}
{"train_loss": 0.14233452081680298, "global_step": 125681, "epoch": 1412, "lr": 5.59530736086922e-05}
{"train_loss": 0.09286046773195267, "global_step": 125682, "epoch": 1412, "lr": 5.5952497995422e-05}
{"train_loss": 0.14434970915317535, "global_step": 125683, "epoch": 1412, "lr": 5.595192238135155e-05}
{"train_loss": 0.19847099483013153, "global_step": 125684, "epoch": 1412, "lr": 5.595134676648094e-05}
{"train_loss": 0.07692527025938034, "global_step": 125685, "epoch": 1412, "lr": 5.5950771150810246e-05}
{"train_loss": 0.11652418971061707, "global_step": 125686, "epoch": 1412, "lr": 5.595019553433953e-05}
{"train_loss": 0.14555583894252777, "global_step": 125687, "epoch": 1412, "lr": 5.5949619917068886e-05}
{"train_loss": 0.10048265755176544, "global_step": 125688, "epoch": 1412, "lr": 5.59490442989984e-05}
{"train_loss": 0.08996853232383728, "global_step": 125689, "epoch": 1412, "lr": 5.594846868012813e-05}
{"train_loss": 0.21312576532363892, "global_step": 125690, "epoch": 1412, "lr": 5.594789306045816e-05}
{"train_loss": 0.09767105430364609, "global_step": 125691, "epoch": 1412, "lr": 5.5947317439988566e-05}
{"train_loss": 0.14769825339317322, "global_step": 125692, "epoch": 1412, "lr": 5.594674181871943e-05}
{"train_loss": 0.15923811495304108, "global_step": 125693, "epoch": 1412, "lr": 5.594616619665083e-05}
{"train_loss": 0.08381245285272598, "global_step": 125694, "epoch": 1412, "lr": 5.594559057378284e-05}
{"train_loss": 0.12716896831989288, "global_step": 125695, "epoch": 1412, "lr": 5.594501495011553e-05}
{"train_loss": 0.11989975720643997, "global_step": 125696, "epoch": 1412, "lr": 5.594443932564898e-05}
{"train_loss": 0.14526627957820892, "global_step": 125697, "epoch": 1412, "lr": 5.5943863700383284e-05}
{"train_loss": 0.08399638533592224, "global_step": 125698, "epoch": 1412, "lr": 5.594328807431851e-05}
{"train_loss": 0.16900603473186493, "global_step": 125699, "epoch": 1412, "lr": 5.5942712447454724e-05}
{"train_loss": 0.12332594394683838, "global_step": 125700, "epoch": 1412, "lr": 5.594213681979202e-05}
{"train_loss": 0.1442672312259674, "global_step": 125701, "epoch": 1412, "lr": 5.5941561191330485e-05}
{"train_loss": 0.11863134801387787, "global_step": 125702, "epoch": 1412, "lr": 5.5940985562070157e-05}
{"train_loss": 0.09999732673168182, "global_step": 125703, "epoch": 1412, "lr": 5.594040993201115e-05}
{"train_loss": 0.1772313416004181, "global_step": 125704, "epoch": 1412, "lr": 5.5939834301153514e-05}
{"train_loss": 0.19428910315036774, "global_step": 125705, "epoch": 1412, "lr": 5.593925866949734e-05}
{"train_loss": 0.10640024393796921, "global_step": 125706, "epoch": 1412, "lr": 5.5938683037042716e-05}
{"train_loss": 0.24635902047157288, "global_step": 125707, "epoch": 1412, "lr": 5.59381074037897e-05}
{"train_loss": 0.05184948071837425, "global_step": 125708, "epoch": 1412, "lr": 5.593753176973838e-05}
{"train_loss": 0.13526418805122375, "global_step": 125709, "epoch": 1412, "lr": 5.593695613488883e-05}
{"train_loss": 0.17665599286556244, "global_step": 125710, "epoch": 1412, "lr": 5.5936380499241125e-05}
{"train_loss": 0.13857696950435638, "global_step": 125711, "epoch": 1412, "lr": 5.593580486279536e-05}
{"train_loss": 0.11900845915079117, "global_step": 125712, "epoch": 1412, "lr": 5.593522922555159e-05}
{"train_loss": 0.1380159854888916, "global_step": 125713, "epoch": 1412, "lr": 5.59346535875099e-05}
{"train_loss": 0.10637354105710983, "global_step": 125714, "epoch": 1412, "lr": 5.593407794867036e-05}
{"train_loss": 0.15603575110435486, "global_step": 125715, "epoch": 1412, "lr": 5.593350230903306e-05}
{"train_loss": 0.18433068692684174, "global_step": 125716, "epoch": 1412, "lr": 5.5932926668598076e-05}
{"train_loss": 0.1810079962015152, "global_step": 125717, "epoch": 1412, "lr": 5.593235102736548e-05}
{"train_loss": 0.133877232670784, "global_step": 125718, "epoch": 1412, "lr": 5.593177538533535e-05}
{"train_loss": 0.15336644649505615, "global_step": 125719, "epoch": 1412, "lr": 5.593119974250777e-05}
{"train_loss": 0.12030737847089767, "global_step": 125720, "epoch": 1412, "lr": 5.593062409888281e-05}
{"train_loss": 0.13172194361686707, "global_step": 125721, "epoch": 1412, "lr": 5.593004845446057e-05}
{"train_loss": 0.1277889907360077, "global_step": 125722, "epoch": 1412, "lr": 5.592947280924109e-05}
{"train_loss": 0.16902142763137817, "global_step": 125723, "epoch": 1412, "lr": 5.592889716322446e-05}
{"train_loss": 0.1756627857685089, "global_step": 125724, "epoch": 1412, "lr": 5.592832151641076e-05}
{"train_loss": 0.08485781401395798, "global_step": 125725, "epoch": 1412, "lr": 5.5927745868800074e-05}
{"train_loss": 0.1370178461074829, "global_step": 125726, "epoch": 1412, "lr": 5.592717022039248e-05}
{"train_loss": 0.13996924459934235, "global_step": 125727, "epoch": 1412, "lr": 5.592659457118805e-05}
{"train_loss": 0.10472625494003296, "global_step": 125728, "epoch": 1412, "lr": 5.592601892118685e-05}
{"train_loss": 0.15743517875671387, "global_step": 125729, "epoch": 1412, "lr": 5.592544327038898e-05}
{"train_loss": 0.19010616838932037, "global_step": 125730, "epoch": 1412, "lr": 5.5924867618794506e-05}
{"train_loss": 0.0960998386144638, "global_step": 125731, "epoch": 1412, "lr": 5.5924291966403496e-05}
{"train_loss": 0.1036776527762413, "global_step": 125732, "epoch": 1412, "lr": 5.5923716313216045e-05}
{"train_loss": 0.1355731189250946, "global_step": 125733, "epoch": 1412, "lr": 5.592314065923222e-05}
{"train_loss": 0.1274958997964859, "global_step": 125734, "epoch": 1412, "lr": 5.5922565004452096e-05}
{"train_loss": 0.07357022166252136, "global_step": 125735, "epoch": 1412, "lr": 5.592198934887577e-05}
{"train_loss": 0.11479085683822632, "global_step": 125736, "epoch": 1412, "lr": 5.592141369250329e-05}
{"train_loss": 0.11275988817214966, "global_step": 125737, "epoch": 1412, "lr": 5.592083803533476e-05}
{"train_loss": 0.10834576189517975, "global_step": 125738, "epoch": 1412, "lr": 5.592026237737024e-05}
{"train_loss": 0.0508078895509243, "global_step": 125739, "epoch": 1412, "lr": 5.591968671860981e-05}
{"train_loss": 0.13479381799697876, "global_step": 125740, "epoch": 1412, "lr": 5.5919111059053566e-05}
{"train_loss": 0.10948711633682251, "global_step": 125741, "epoch": 1412, "lr": 5.591853539870155e-05}
{"train_loss": 0.15366517007350922, "global_step": 125742, "epoch": 1412, "lr": 5.591795973755387e-05}
{"train_loss": 0.13305912911891937, "global_step": 125743, "epoch": 1412, "lr": 5.591738407561059e-05}
{"train_loss": 0.04312952980399132, "global_step": 125744, "epoch": 1412, "lr": 5.591680841287179e-05}
{"train_loss": 0.15560294687747955, "global_step": 125745, "epoch": 1412, "lr": 5.5916232749337535e-05}
{"train_loss": 0.11247795075178146, "global_step": 125746, "epoch": 1412, "lr": 5.5915657085007935e-05}
{"train_loss": 0.12660466134548187, "global_step": 125747, "epoch": 1412, "lr": 5.591508141988303e-05}
{"train_loss": 0.17627665400505066, "global_step": 125748, "epoch": 1412, "lr": 5.5914505753962934e-05}
{"train_loss": 0.1068815067410469, "global_step": 125749, "epoch": 1412, "lr": 5.591393008724768e-05}
{"train_loss": 0.15920701622962952, "global_step": 125750, "epoch": 1412, "lr": 5.5913354419737385e-05}
{"train_loss": 0.1632760763168335, "global_step": 125751, "epoch": 1412, "lr": 5.5912778751432104e-05}
{"train_loss": 0.1469457596540451, "global_step": 125752, "epoch": 1412, "lr": 5.5912203082331926e-05}
{"train_loss": 0.09464719891548157, "global_step": 125753, "epoch": 1412, "lr": 5.591162741243692e-05}
{"train_loss": 0.1674502044916153, "global_step": 125754, "epoch": 1412, "lr": 5.591105174174718e-05}
{"train_loss": 0.1165003851056099, "global_step": 125755, "epoch": 1412, "lr": 5.5910476070262766e-05}
{"train_loss": 0.13388127461075783, "global_step": 125756, "epoch": 1412, "lr": 5.590990039798375e-05, "val_loss": 5.352419376373291}
{"train_loss": 0.11951905488967896, "global_step": 125757, "epoch": 1413, "lr": 5.5909324724910226e-05}
{"train_loss": 0.16229389607906342, "global_step": 125758, "epoch": 1413, "lr": 5.5908749051042276e-05}
{"train_loss": 0.09644781798124313, "global_step": 125759, "epoch": 1413, "lr": 5.590817337637997e-05}
{"train_loss": 0.13186155259609222, "global_step": 125760, "epoch": 1413, "lr": 5.590759770092337e-05}
{"train_loss": 0.10891766101121902, "global_step": 125761, "epoch": 1413, "lr": 5.5907022024672564e-05}
{"train_loss": 0.15208376944065094, "global_step": 125762, "epoch": 1413, "lr": 5.590644634762763e-05}
{"train_loss": 0.13751676678657532, "global_step": 125763, "epoch": 1413, "lr": 5.5905870669788654e-05}
{"train_loss": 0.0886128693819046, "global_step": 125764, "epoch": 1413, "lr": 5.5905294991155697e-05}
{"train_loss": 0.18652242422103882, "global_step": 125765, "epoch": 1413, "lr": 5.5904719311728845e-05}
{"train_loss": 0.14745479822158813, "global_step": 125766, "epoch": 1413, "lr": 5.590414363150819e-05}
{"train_loss": 0.1717793494462967, "global_step": 125767, "epoch": 1413, "lr": 5.590356795049379e-05}
{"train_loss": 0.1836371272802353, "global_step": 125768, "epoch": 1413, "lr": 5.590299226868572e-05}
{"train_loss": 0.09990371763706207, "global_step": 125769, "epoch": 1413, "lr": 5.590241658608405e-05}
{"train_loss": 0.12707866728305817, "global_step": 125770, "epoch": 1413, "lr": 5.5901840902688884e-05}
{"train_loss": 0.14436514675617218, "global_step": 125771, "epoch": 1413, "lr": 5.59012652185003e-05}
{"train_loss": 0.10443931818008423, "global_step": 125772, "epoch": 1413, "lr": 5.590068953351835e-05}
{"train_loss": 0.12172544747591019, "global_step": 125773, "epoch": 1413, "lr": 5.590011384774313e-05}
{"train_loss": 0.14160607755184174, "global_step": 125774, "epoch": 1413, "lr": 5.5899538161174716e-05}
{"train_loss": 0.11821683496236801, "global_step": 125775, "epoch": 1413, "lr": 5.5898962473813165e-05}
{"train_loss": 0.11896557360887527, "global_step": 125776, "epoch": 1413, "lr": 5.589838678565859e-05}
{"train_loss": 0.19257234036922455, "global_step": 125777, "epoch": 1413, "lr": 5.589781109671104e-05}
{"train_loss": 0.0968330129981041, "global_step": 125778, "epoch": 1413, "lr": 5.589723540697059e-05}
{"train_loss": 0.10423852503299713, "global_step": 125779, "epoch": 1413, "lr": 5.589665971643736e-05}
{"train_loss": 0.13225236535072327, "global_step": 125780, "epoch": 1413, "lr": 5.5896084025111374e-05}
{"train_loss": 0.060755226761102676, "global_step": 125781, "epoch": 1413, "lr": 5.589550833299273e-05}
{"train_loss": 0.20028351247310638, "global_step": 125782, "epoch": 1413, "lr": 5.589493264008151e-05}
{"train_loss": 0.13316042721271515, "global_step": 125783, "epoch": 1413, "lr": 5.5894356946377794e-05}
{"train_loss": 0.13877612352371216, "global_step": 125784, "epoch": 1413, "lr": 5.589378125188164e-05}
{"train_loss": 0.1381194293498993, "global_step": 125785, "epoch": 1413, "lr": 5.5893205556593145e-05}
{"train_loss": 0.15101788938045502, "global_step": 125786, "epoch": 1413, "lr": 5.589262986051238e-05}
{"train_loss": 0.13328078389167786, "global_step": 125787, "epoch": 1413, "lr": 5.589205416363942e-05}
{"train_loss": 0.1228988990187645, "global_step": 125788, "epoch": 1413, "lr": 5.5891478465974344e-05}
{"train_loss": 0.09601166844367981, "global_step": 125789, "epoch": 1413, "lr": 5.589090276751724e-05}
{"train_loss": 0.0613885298371315, "global_step": 125790, "epoch": 1413, "lr": 5.589032706826817e-05}
{"train_loss": 0.2758634388446808, "global_step": 125791, "epoch": 1413, "lr": 5.588975136822722e-05}
{"train_loss": 0.054699305444955826, "global_step": 125792, "epoch": 1413, "lr": 5.588917566739447e-05}
{"train_loss": 0.07615198194980621, "global_step": 125793, "epoch": 1413, "lr": 5.588859996576998e-05}
{"train_loss": 0.056607555598020554, "global_step": 125794, "epoch": 1413, "lr": 5.588802426335384e-05}
{"train_loss": 0.11257849633693695, "global_step": 125795, "epoch": 1413, "lr": 5.588744856014614e-05}
{"train_loss": 0.07514294236898422, "global_step": 125796, "epoch": 1413, "lr": 5.588687285614693e-05}
{"train_loss": 0.08742011338472366, "global_step": 125797, "epoch": 1413, "lr": 5.5886297151356316e-05}
{"train_loss": 0.09787074476480484, "global_step": 125798, "epoch": 1413, "lr": 5.588572144577436e-05}
{"train_loss": 0.13191765546798706, "global_step": 125799, "epoch": 1413, "lr": 5.5885145739401136e-05}
{"train_loss": 0.13065685331821442, "global_step": 125800, "epoch": 1413, "lr": 5.5884570032236713e-05}
{"train_loss": 0.19889725744724274, "global_step": 125801, "epoch": 1413, "lr": 5.5883994324281196e-05}
{"train_loss": 0.11700937151908875, "global_step": 125802, "epoch": 1413, "lr": 5.5883418615534645e-05}
{"train_loss": 0.15209874510765076, "global_step": 125803, "epoch": 1413, "lr": 5.588284290599715e-05}
{"train_loss": 0.12724685668945312, "global_step": 125804, "epoch": 1413, "lr": 5.588226719566876e-05}
{"train_loss": 0.09473828226327896, "global_step": 125805, "epoch": 1413, "lr": 5.588169148454957e-05}
{"train_loss": 0.16411317884922028, "global_step": 125806, "epoch": 1413, "lr": 5.588111577263966e-05}
{"train_loss": 0.17497606575489044, "global_step": 125807, "epoch": 1413, "lr": 5.588054005993911e-05}
{"train_loss": 0.1607092320919037, "global_step": 125808, "epoch": 1413, "lr": 5.5879964346448e-05}
{"train_loss": 0.1698639690876007, "global_step": 125809, "epoch": 1413, "lr": 5.58793886321664e-05}
{"train_loss": 0.1097538098692894, "global_step": 125810, "epoch": 1413, "lr": 5.587881291709438e-05}
{"train_loss": 0.09769085794687271, "global_step": 125811, "epoch": 1413, "lr": 5.5878237201232034e-05}
{"train_loss": 0.09708191454410553, "global_step": 125812, "epoch": 1413, "lr": 5.5877661484579426e-05}
{"train_loss": 0.12222019582986832, "global_step": 125813, "epoch": 1413, "lr": 5.5877085767136625e-05}
{"train_loss": 0.1409219652414322, "global_step": 125814, "epoch": 1413, "lr": 5.587651004890374e-05}
{"train_loss": 0.08471641689538956, "global_step": 125815, "epoch": 1413, "lr": 5.587593432988083e-05}
{"train_loss": 0.134355828166008, "global_step": 125816, "epoch": 1413, "lr": 5.587535861006796e-05}
{"train_loss": 0.133756622672081, "global_step": 125817, "epoch": 1413, "lr": 5.587478288946524e-05}
{"train_loss": 0.11079657822847366, "global_step": 125818, "epoch": 1413, "lr": 5.587420716807271e-05}
{"train_loss": 0.1693829894065857, "global_step": 125819, "epoch": 1413, "lr": 5.587363144589046e-05}
{"train_loss": 0.11296667903661728, "global_step": 125820, "epoch": 1413, "lr": 5.587305572291859e-05}
{"train_loss": 0.081464484333992, "global_step": 125821, "epoch": 1413, "lr": 5.5872479999157144e-05}
{"train_loss": 0.13867615163326263, "global_step": 125822, "epoch": 1413, "lr": 5.5871904274606225e-05}
{"train_loss": 0.17818911373615265, "global_step": 125823, "epoch": 1413, "lr": 5.587132854926588e-05}
{"train_loss": 0.10529188811779022, "global_step": 125824, "epoch": 1413, "lr": 5.587075282313622e-05}
{"train_loss": 0.09106896072626114, "global_step": 125825, "epoch": 1413, "lr": 5.58701770962173e-05}
{"train_loss": 0.1768735945224762, "global_step": 125826, "epoch": 1413, "lr": 5.5869601368509226e-05}
{"train_loss": 0.13588811457157135, "global_step": 125827, "epoch": 1413, "lr": 5.586902564001204e-05}
{"train_loss": 0.15395906567573547, "global_step": 125828, "epoch": 1413, "lr": 5.5868449910725836e-05}
{"train_loss": 0.12695857882499695, "global_step": 125829, "epoch": 1413, "lr": 5.5867874180650704e-05}
{"train_loss": 0.11979226768016815, "global_step": 125830, "epoch": 1413, "lr": 5.58672984497867e-05}
{"train_loss": 0.15102022886276245, "global_step": 125831, "epoch": 1413, "lr": 5.5866722718133904e-05}
{"train_loss": 0.1232246533036232, "global_step": 125832, "epoch": 1413, "lr": 5.58661469856924e-05}
{"train_loss": 0.12578478455543518, "global_step": 125833, "epoch": 1413, "lr": 5.5865571252462266e-05}
{"train_loss": 0.10330630093812943, "global_step": 125834, "epoch": 1413, "lr": 5.586499551844357e-05}
{"train_loss": 0.18233734369277954, "global_step": 125835, "epoch": 1413, "lr": 5.586441978363641e-05}
{"train_loss": 0.1423998922109604, "global_step": 125836, "epoch": 1413, "lr": 5.586384404804085e-05}
{"train_loss": 0.06984680145978928, "global_step": 125837, "epoch": 1413, "lr": 5.586326831165697e-05}
{"train_loss": 0.11390145123004913, "global_step": 125838, "epoch": 1413, "lr": 5.5862692574484844e-05}
{"train_loss": 0.07491028308868408, "global_step": 125839, "epoch": 1413, "lr": 5.586211683652453e-05}
{"train_loss": 0.1693253517150879, "global_step": 125840, "epoch": 1413, "lr": 5.586154109777615e-05}
{"train_loss": 0.1419251561164856, "global_step": 125841, "epoch": 1413, "lr": 5.586096535823975e-05}
{"train_loss": 0.12615439295768738, "global_step": 125842, "epoch": 1413, "lr": 5.58603896179154e-05}
{"train_loss": 0.09437359124422073, "global_step": 125843, "epoch": 1413, "lr": 5.58598138768032e-05}
{"train_loss": 0.08123756945133209, "global_step": 125844, "epoch": 1413, "lr": 5.5859238134903223e-05}
{"train_loss": 0.12652319339051676, "global_step": 125845, "epoch": 1413, "lr": 5.585866239221555e-05, "val_loss": 5.5119123458862305}
{"train_loss": 0.10473573952913284, "global_step": 125846, "epoch": 1414, "lr": 5.585808664874024e-05}
{"train_loss": 0.15886865556240082, "global_step": 125847, "epoch": 1414, "lr": 5.5857510904477386e-05}
{"train_loss": 0.19049201905727386, "global_step": 125848, "epoch": 1414, "lr": 5.585693515942706e-05}
{"train_loss": 0.11049126088619232, "global_step": 125849, "epoch": 1414, "lr": 5.5856359413589346e-05}
{"train_loss": 0.08783046156167984, "global_step": 125850, "epoch": 1414, "lr": 5.58557836669643e-05}
{"train_loss": 0.16178904473781586, "global_step": 125851, "epoch": 1414, "lr": 5.585520791955203e-05}
{"train_loss": 0.05857245624065399, "global_step": 125852, "epoch": 1414, "lr": 5.5854632171352597e-05}
{"train_loss": 0.10199889540672302, "global_step": 125853, "epoch": 1414, "lr": 5.585405642236607e-05}
{"train_loss": 0.1137298047542572, "global_step": 125854, "epoch": 1414, "lr": 5.585348067259255e-05}
{"train_loss": 0.16997483372688293, "global_step": 125855, "epoch": 1414, "lr": 5.5852904922032103e-05}
{"train_loss": 0.1404646635055542, "global_step": 125856, "epoch": 1414, "lr": 5.58523291706848e-05}
{"train_loss": 0.2098124772310257, "global_step": 125857, "epoch": 1414, "lr": 5.5851753418550724e-05}
{"train_loss": 0.1377781629562378, "global_step": 125858, "epoch": 1414, "lr": 5.585117766562995e-05}
{"train_loss": 0.14636385440826416, "global_step": 125859, "epoch": 1414, "lr": 5.585060191192255e-05}
{"train_loss": 0.11048951745033264, "global_step": 125860, "epoch": 1414, "lr": 5.58500261574286e-05}
{"train_loss": 0.10230224579572678, "global_step": 125861, "epoch": 1414, "lr": 5.5849450402148195e-05}
{"train_loss": 0.15088434517383575, "global_step": 125862, "epoch": 1414, "lr": 5.58488746460814e-05}
{"train_loss": 0.14059147238731384, "global_step": 125863, "epoch": 1414, "lr": 5.58482988892283e-05}
{"train_loss": 0.18219713866710663, "global_step": 125864, "epoch": 1414, "lr": 5.584772313158897e-05}
{"train_loss": 0.17659707367420197, "global_step": 125865, "epoch": 1414, "lr": 5.584714737316348e-05}
{"train_loss": 0.2140611708164215, "global_step": 125866, "epoch": 1414, "lr": 5.584657161395191e-05}
{"train_loss": 0.1848302185535431, "global_step": 125867, "epoch": 1414, "lr": 5.5845995853954345e-05}
{"train_loss": 0.11927662044763565, "global_step": 125868, "epoch": 1414, "lr": 5.584542009317087e-05}
{"train_loss": 0.08407249301671982, "global_step": 125869, "epoch": 1414, "lr": 5.584484433160152e-05}
{"train_loss": 0.20971161127090454, "global_step": 125870, "epoch": 1414, "lr": 5.5844268569246414e-05}
{"train_loss": 0.1506861448287964, "global_step": 125871, "epoch": 1414, "lr": 5.584369280610563e-05}
{"train_loss": 0.15252156555652618, "global_step": 125872, "epoch": 1414, "lr": 5.584311704217922e-05}
{"train_loss": 0.10280726104974747, "global_step": 125873, "epoch": 1414, "lr": 5.584254127746728e-05}
{"train_loss": 0.19285117089748383, "global_step": 125874, "epoch": 1414, "lr": 5.584196551196987e-05}
{"train_loss": 0.09267845004796982, "global_step": 125875, "epoch": 1414, "lr": 5.58413897456871e-05}
{"train_loss": 0.1673116236925125, "global_step": 125876, "epoch": 1414, "lr": 5.584081397861902e-05}
{"train_loss": 0.19895432889461517, "global_step": 125877, "epoch": 1414, "lr": 5.58402382107657e-05}
{"train_loss": 0.16862019896507263, "global_step": 125878, "epoch": 1414, "lr": 5.583966244212724e-05}
{"train_loss": 0.15133656561374664, "global_step": 125879, "epoch": 1414, "lr": 5.58390866727037e-05}
{"train_loss": 0.1543128341436386, "global_step": 125880, "epoch": 1414, "lr": 5.583851090249518e-05}
{"train_loss": 0.1592545360326767, "global_step": 125881, "epoch": 1414, "lr": 5.583793513150173e-05}
{"train_loss": 0.17398427426815033, "global_step": 125882, "epoch": 1414, "lr": 5.583735935972345e-05}
{"train_loss": 0.0960807204246521, "global_step": 125883, "epoch": 1414, "lr": 5.58367835871604e-05}
{"train_loss": 0.10393088310956955, "global_step": 125884, "epoch": 1414, "lr": 5.583620781381268e-05}
{"train_loss": 0.1080603078007698, "global_step": 125885, "epoch": 1414, "lr": 5.583563203968034e-05}
{"train_loss": 0.13306277990341187, "global_step": 125886, "epoch": 1414, "lr": 5.5835056264763475e-05}
{"train_loss": 0.0908222645521164, "global_step": 125887, "epoch": 1414, "lr": 5.583448048906216e-05}
{"train_loss": 0.11392061412334442, "global_step": 125888, "epoch": 1414, "lr": 5.583390471257647e-05}
{"train_loss": 0.23244015872478485, "global_step": 125889, "epoch": 1414, "lr": 5.5833328935306475e-05}
{"train_loss": 0.15168681740760803, "global_step": 125890, "epoch": 1414, "lr": 5.583275315725227e-05}
{"train_loss": 0.13207904994487762, "global_step": 125891, "epoch": 1414, "lr": 5.583217737841392e-05}
{"train_loss": 0.13029199838638306, "global_step": 125892, "epoch": 1414, "lr": 5.5831601598791496e-05}
{"train_loss": 0.134173184633255, "global_step": 125893, "epoch": 1414, "lr": 5.58310258183851e-05}
{"train_loss": 0.11612343043088913, "global_step": 125894, "epoch": 1414, "lr": 5.583045003719478e-05}
{"train_loss": 0.08306258916854858, "global_step": 125895, "epoch": 1414, "lr": 5.582987425522064e-05}
{"train_loss": 0.09933198243379593, "global_step": 125896, "epoch": 1414, "lr": 5.582929847246272e-05}
{"train_loss": 0.09643509984016418, "global_step": 125897, "epoch": 1414, "lr": 5.5828722688921146e-05}
{"train_loss": 0.15552254021167755, "global_step": 125898, "epoch": 1414, "lr": 5.582814690459596e-05}
{"train_loss": 0.1644282191991806, "global_step": 125899, "epoch": 1414, "lr": 5.582757111948725e-05}
{"train_loss": 0.12823736667633057, "global_step": 125900, "epoch": 1414, "lr": 5.58269953335951e-05}
{"train_loss": 0.203972727060318, "global_step": 125901, "epoch": 1414, "lr": 5.582641954691958e-05}
{"train_loss": 0.12195366621017456, "global_step": 125902, "epoch": 1414, "lr": 5.582584375946077e-05}
{"train_loss": 0.13472311198711395, "global_step": 125903, "epoch": 1414, "lr": 5.5825267971218743e-05}
{"train_loss": 0.13968142867088318, "global_step": 125904, "epoch": 1414, "lr": 5.582469218219358e-05}
{"train_loss": 0.11223740875720978, "global_step": 125905, "epoch": 1414, "lr": 5.582411639238536e-05}
{"train_loss": 0.17414920032024384, "global_step": 125906, "epoch": 1414, "lr": 5.5823540601794156e-05}
{"train_loss": 0.08353941142559052, "global_step": 125907, "epoch": 1414, "lr": 5.5822964810420055e-05}
{"train_loss": 0.11992981284856796, "global_step": 125908, "epoch": 1414, "lr": 5.582238901826312e-05}
{"train_loss": 0.131662517786026, "global_step": 125909, "epoch": 1414, "lr": 5.582181322532344e-05}
{"train_loss": 0.09857427328824997, "global_step": 125910, "epoch": 1414, "lr": 5.582123743160108e-05}
{"train_loss": 0.13897156715393066, "global_step": 125911, "epoch": 1414, "lr": 5.582066163709614e-05}
{"train_loss": 0.09154798835515976, "global_step": 125912, "epoch": 1414, "lr": 5.582008584180867e-05}
{"train_loss": 0.22017720341682434, "global_step": 125913, "epoch": 1414, "lr": 5.5819510045738767e-05}
{"train_loss": 0.08435461670160294, "global_step": 125914, "epoch": 1414, "lr": 5.58189342488865e-05}
{"train_loss": 0.2174672782421112, "global_step": 125915, "epoch": 1414, "lr": 5.581835845125195e-05}
{"train_loss": 0.16206663846969604, "global_step": 125916, "epoch": 1414, "lr": 5.581778265283519e-05}
{"train_loss": 0.1344151347875595, "global_step": 125917, "epoch": 1414, "lr": 5.5817206853636295e-05}
{"train_loss": 0.18081925809383392, "global_step": 125918, "epoch": 1414, "lr": 5.581663105365536e-05}
{"train_loss": 0.12207651883363724, "global_step": 125919, "epoch": 1414, "lr": 5.581605525289244e-05}
{"train_loss": 0.10455309599637985, "global_step": 125920, "epoch": 1414, "lr": 5.581547945134763e-05}
{"train_loss": 0.12007781118154526, "global_step": 125921, "epoch": 1414, "lr": 5.5814903649020986e-05}
{"train_loss": 0.138197660446167, "global_step": 125922, "epoch": 1414, "lr": 5.581432784591262e-05}
{"train_loss": 0.1345757395029068, "global_step": 125923, "epoch": 1414, "lr": 5.581375204202257e-05}
{"train_loss": 0.13055190443992615, "global_step": 125924, "epoch": 1414, "lr": 5.581317623735094e-05}
{"train_loss": 0.13513164222240448, "global_step": 125925, "epoch": 1414, "lr": 5.58126004318978e-05}
{"train_loss": 0.22384488582611084, "global_step": 125926, "epoch": 1414, "lr": 5.581202462566322e-05}
{"train_loss": 0.08205471187829971, "global_step": 125927, "epoch": 1414, "lr": 5.581144881864729e-05}
{"train_loss": 0.15712913870811462, "global_step": 125928, "epoch": 1414, "lr": 5.581087301085008e-05}
{"train_loss": 0.11215692013502121, "global_step": 125929, "epoch": 1414, "lr": 5.5810297202271666e-05}
{"train_loss": 0.12307582050561905, "global_step": 125930, "epoch": 1414, "lr": 5.580972139291214e-05}
{"train_loss": 0.11305232346057892, "global_step": 125931, "epoch": 1414, "lr": 5.580914558277156e-05}
{"train_loss": 0.12686623632907867, "global_step": 125932, "epoch": 1414, "lr": 5.5808569771850006e-05}
{"train_loss": 0.14587228000164032, "global_step": 125933, "epoch": 1414, "lr": 5.580799396014756e-05}
{"train_loss": 0.13846347974927237, "global_step": 125934, "epoch": 1414, "lr": 5.580741814766432e-05, "val_loss": 5.399753093719482}
{"train_loss": 0.11660467088222504, "global_step": 125935, "epoch": 1415, "lr": 5.5806842334400324e-05}
{"train_loss": 0.12742334604263306, "global_step": 125936, "epoch": 1415, "lr": 5.5806266520355665e-05}
{"train_loss": 0.1069854274392128, "global_step": 125937, "epoch": 1415, "lr": 5.5805690705530435e-05}
{"train_loss": 0.14975100755691528, "global_step": 125938, "epoch": 1415, "lr": 5.58051148899247e-05}
{"train_loss": 0.10465365648269653, "global_step": 125939, "epoch": 1415, "lr": 5.580453907353853e-05}
{"train_loss": 0.12345894426107407, "global_step": 125940, "epoch": 1415, "lr": 5.5803963256372026e-05}
{"train_loss": 0.12428838759660721, "global_step": 125941, "epoch": 1415, "lr": 5.5803387438425236e-05}
{"train_loss": 0.16774533689022064, "global_step": 125942, "epoch": 1415, "lr": 5.580281161969826e-05}
{"train_loss": 0.11381995677947998, "global_step": 125943, "epoch": 1415, "lr": 5.580223580019116e-05}
{"train_loss": 0.08756033331155777, "global_step": 125944, "epoch": 1415, "lr": 5.5801659979904017e-05}
{"train_loss": 0.08708806335926056, "global_step": 125945, "epoch": 1415, "lr": 5.5801084158836926e-05}
{"train_loss": 0.07625514268875122, "global_step": 125946, "epoch": 1415, "lr": 5.580050833698994e-05}
{"train_loss": 0.11172821372747421, "global_step": 125947, "epoch": 1415, "lr": 5.579993251436315e-05}
{"train_loss": 0.15466827154159546, "global_step": 125948, "epoch": 1415, "lr": 5.579935669095663e-05}
{"train_loss": 0.13267721235752106, "global_step": 125949, "epoch": 1415, "lr": 5.579878086677045e-05}
{"train_loss": 0.16457979381084442, "global_step": 125950, "epoch": 1415, "lr": 5.57982050418047e-05}
{"train_loss": 0.12303844094276428, "global_step": 125951, "epoch": 1415, "lr": 5.5797629216059445e-05}
{"train_loss": 0.1264973282814026, "global_step": 125952, "epoch": 1415, "lr": 5.579705338953478e-05}
{"train_loss": 0.15029415488243103, "global_step": 125953, "epoch": 1415, "lr": 5.579647756223077e-05}
{"train_loss": 0.0921274796128273, "global_step": 125954, "epoch": 1415, "lr": 5.579590173414748e-05}
{"train_loss": 0.13885283470153809, "global_step": 125955, "epoch": 1415, "lr": 5.579532590528502e-05}
{"train_loss": 0.09182094782590866, "global_step": 125956, "epoch": 1415, "lr": 5.5794750075643444e-05}
{"train_loss": 0.07647763937711716, "global_step": 125957, "epoch": 1415, "lr": 5.579417424522283e-05}
{"train_loss": 0.1151665672659874, "global_step": 125958, "epoch": 1415, "lr": 5.579359841402326e-05}
{"train_loss": 0.15445713698863983, "global_step": 125959, "epoch": 1415, "lr": 5.5793022582044815e-05}
{"train_loss": 0.14040473103523254, "global_step": 125960, "epoch": 1415, "lr": 5.579244674928757e-05}
{"train_loss": 0.05697585642337799, "global_step": 125961, "epoch": 1415, "lr": 5.57918709157516e-05}
{"train_loss": 0.12528300285339355, "global_step": 125962, "epoch": 1415, "lr": 5.5791295081436975e-05}
{"train_loss": 0.17682740092277527, "global_step": 125963, "epoch": 1415, "lr": 5.579071924634379e-05}
{"train_loss": 0.10371682047843933, "global_step": 125964, "epoch": 1415, "lr": 5.5790143410472116e-05}
{"train_loss": 0.15035128593444824, "global_step": 125965, "epoch": 1415, "lr": 5.5789567573822025e-05}
{"train_loss": 0.15806685388088226, "global_step": 125966, "epoch": 1415, "lr": 5.578899173639359e-05}
{"train_loss": 0.10794396698474884, "global_step": 125967, "epoch": 1415, "lr": 5.578841589818691e-05}
{"train_loss": 0.06625036150217056, "global_step": 125968, "epoch": 1415, "lr": 5.578784005920204e-05}
{"train_loss": 0.17161746323108673, "global_step": 125969, "epoch": 1415, "lr": 5.578726421943906e-05}
{"train_loss": 0.09249518811702728, "global_step": 125970, "epoch": 1415, "lr": 5.578668837889806e-05}
{"train_loss": 0.1894775629043579, "global_step": 125971, "epoch": 1415, "lr": 5.5786112537579107e-05}
{"train_loss": 0.17872561514377594, "global_step": 125972, "epoch": 1415, "lr": 5.578553669548229e-05}
{"train_loss": 0.18067069351673126, "global_step": 125973, "epoch": 1415, "lr": 5.5784960852607657e-05}
{"train_loss": 0.10563299804925919, "global_step": 125974, "epoch": 1415, "lr": 5.5784385008955333e-05}
{"train_loss": 0.12382012605667114, "global_step": 125975, "epoch": 1415, "lr": 5.5783809164525346e-05}
{"train_loss": 0.12905152142047882, "global_step": 125976, "epoch": 1415, "lr": 5.578323331931781e-05}
{"train_loss": 0.12882930040359497, "global_step": 125977, "epoch": 1415, "lr": 5.578265747333279e-05}
{"train_loss": 0.1409519463777542, "global_step": 125978, "epoch": 1415, "lr": 5.578208162657036e-05}
{"train_loss": 0.1249082013964653, "global_step": 125979, "epoch": 1415, "lr": 5.5781505779030596e-05}
{"train_loss": 0.12241829931735992, "global_step": 125980, "epoch": 1415, "lr": 5.5780929930713574e-05}
{"train_loss": 0.14654351770877838, "global_step": 125981, "epoch": 1415, "lr": 5.578035408161939e-05}
{"train_loss": 0.15827696025371552, "global_step": 125982, "epoch": 1415, "lr": 5.577977823174809e-05}
{"train_loss": 0.17698988318443298, "global_step": 125983, "epoch": 1415, "lr": 5.577920238109978e-05}
{"train_loss": 0.12422685325145721, "global_step": 125984, "epoch": 1415, "lr": 5.577862652967453e-05}
{"train_loss": 0.15630002319812775, "global_step": 125985, "epoch": 1415, "lr": 5.577805067747242e-05}
{"train_loss": 0.10954194515943527, "global_step": 125986, "epoch": 1415, "lr": 5.5777474824493504e-05}
{"train_loss": 0.27779701352119446, "global_step": 125987, "epoch": 1415, "lr": 5.577689897073788e-05}
{"train_loss": 0.12466979771852493, "global_step": 125988, "epoch": 1415, "lr": 5.577632311620563e-05}
{"train_loss": 0.09951837360858917, "global_step": 125989, "epoch": 1415, "lr": 5.5775747260896824e-05}
{"train_loss": 0.1390034556388855, "global_step": 125990, "epoch": 1415, "lr": 5.5775171404811535e-05}
{"train_loss": 0.16684018075466156, "global_step": 125991, "epoch": 1415, "lr": 5.577459554794985e-05}
{"train_loss": 0.1383228898048401, "global_step": 125992, "epoch": 1415, "lr": 5.577401969031183e-05}
{"train_loss": 0.1000504344701767, "global_step": 125993, "epoch": 1415, "lr": 5.5773443831897574e-05}
{"train_loss": 0.10963837057352066, "global_step": 125994, "epoch": 1415, "lr": 5.577286797270714e-05}
{"train_loss": 0.16855724155902863, "global_step": 125995, "epoch": 1415, "lr": 5.577229211274062e-05}
{"train_loss": 0.11139379441738129, "global_step": 125996, "epoch": 1415, "lr": 5.5771716251998086e-05}
{"train_loss": 0.08857779949903488, "global_step": 125997, "epoch": 1415, "lr": 5.5771140390479615e-05}
{"train_loss": 0.11637246608734131, "global_step": 125998, "epoch": 1415, "lr": 5.577056452818529e-05}
{"train_loss": 0.13503454625606537, "global_step": 125999, "epoch": 1415, "lr": 5.576998866511517e-05}
{"train_loss": 0.1217828094959259, "global_step": 126000, "epoch": 1415, "lr": 5.576941280126935e-05}
{"train_loss": 0.10309183597564697, "global_step": 126001, "epoch": 1415, "lr": 5.576883693664791e-05}
{"train_loss": 0.1952505111694336, "global_step": 126002, "epoch": 1415, "lr": 5.576826107125091e-05}
{"train_loss": 0.12143289297819138, "global_step": 126003, "epoch": 1415, "lr": 5.5767685205078444e-05}
{"train_loss": 0.0708988830447197, "global_step": 126004, "epoch": 1415, "lr": 5.576710933813059e-05}
{"train_loss": 0.18787820637226105, "global_step": 126005, "epoch": 1415, "lr": 5.576653347040739e-05}
{"train_loss": 0.09741325676441193, "global_step": 126006, "epoch": 1415, "lr": 5.576595760190898e-05}
{"train_loss": 0.15849114954471588, "global_step": 126007, "epoch": 1415, "lr": 5.57653817326354e-05}
{"train_loss": 0.1957293152809143, "global_step": 126008, "epoch": 1415, "lr": 5.5764805862586734e-05}
{"train_loss": 0.15247631072998047, "global_step": 126009, "epoch": 1415, "lr": 5.576422999176305e-05}
{"train_loss": 0.15535375475883484, "global_step": 126010, "epoch": 1415, "lr": 5.576365412016444e-05}
{"train_loss": 0.13131849467754364, "global_step": 126011, "epoch": 1415, "lr": 5.576307824779099e-05}
{"train_loss": 0.166682630777359, "global_step": 126012, "epoch": 1415, "lr": 5.576250237464275e-05}
{"train_loss": 0.09179382771253586, "global_step": 126013, "epoch": 1415, "lr": 5.576192650071982e-05}
{"train_loss": 0.15399274230003357, "global_step": 126014, "epoch": 1415, "lr": 5.5761350626022266e-05}
{"train_loss": 0.18904972076416016, "global_step": 126015, "epoch": 1415, "lr": 5.5760774750550174e-05}
{"train_loss": 0.1099032610654831, "global_step": 126016, "epoch": 1415, "lr": 5.57601988743036e-05}
{"train_loss": 0.13515029847621918, "global_step": 126017, "epoch": 1415, "lr": 5.5759622997282654e-05}
{"train_loss": 0.1899803876876831, "global_step": 126018, "epoch": 1415, "lr": 5.5759047119487395e-05}
{"train_loss": 0.18307697772979736, "global_step": 126019, "epoch": 1415, "lr": 5.57584712409179e-05}
{"train_loss": 0.1383434236049652, "global_step": 126020, "epoch": 1415, "lr": 5.5757895361574245e-05}
{"train_loss": 0.15193811058998108, "global_step": 126021, "epoch": 1415, "lr": 5.57573194814565e-05}
{"train_loss": 0.09520678222179413, "global_step": 126022, "epoch": 1415, "lr": 5.5756743600564776e-05}
{"train_loss": 0.1341877853435077, "global_step": 126023, "epoch": 1415, "lr": 5.575616771889911e-05, "val_loss": 5.584216117858887, "train_action_mse_error": 40.31779098510742}
{"train_loss": 0.1523306965827942, "global_step": 126024, "epoch": 1416, "lr": 5.57555918364596e-05}
{"train_loss": 0.08469144254922867, "global_step": 126025, "epoch": 1416, "lr": 5.575501595324634e-05}
{"train_loss": 0.11732576787471771, "global_step": 126026, "epoch": 1416, "lr": 5.5754440069259376e-05}
{"train_loss": 0.1743936538696289, "global_step": 126027, "epoch": 1416, "lr": 5.5753864184498795e-05}
{"train_loss": 0.18134120106697083, "global_step": 126028, "epoch": 1416, "lr": 5.575328829896468e-05}
{"train_loss": 0.09777059406042099, "global_step": 126029, "epoch": 1416, "lr": 5.57527124126571e-05}
{"train_loss": 0.17125160992145538, "global_step": 126030, "epoch": 1416, "lr": 5.575213652557615e-05}
{"train_loss": 0.1560778021812439, "global_step": 126031, "epoch": 1416, "lr": 5.575156063772189e-05}
{"train_loss": 0.11473822593688965, "global_step": 126032, "epoch": 1416, "lr": 5.5750984749094395e-05}
{"train_loss": 0.10921566188335419, "global_step": 126033, "epoch": 1416, "lr": 5.575040885969376e-05}
{"train_loss": 0.11794287711381912, "global_step": 126034, "epoch": 1416, "lr": 5.5749832969520044e-05}
{"train_loss": 0.13628901541233063, "global_step": 126035, "epoch": 1416, "lr": 5.574925707857334e-05}
{"train_loss": 0.1694582849740982, "global_step": 126036, "epoch": 1416, "lr": 5.574868118685371e-05}
{"train_loss": 0.2170657366514206, "global_step": 126037, "epoch": 1416, "lr": 5.574810529436125e-05}
{"train_loss": 0.06330960243940353, "global_step": 126038, "epoch": 1416, "lr": 5.574752940109602e-05}
{"train_loss": 0.1863783448934555, "global_step": 126039, "epoch": 1416, "lr": 5.574695350705811e-05}
{"train_loss": 0.1459057480096817, "global_step": 126040, "epoch": 1416, "lr": 5.5746377612247594e-05}
{"train_loss": 0.0918414294719696, "global_step": 126041, "epoch": 1416, "lr": 5.574580171666453e-05}
{"train_loss": 0.1367599070072174, "global_step": 126042, "epoch": 1416, "lr": 5.574522582030902e-05}
{"train_loss": 0.14581778645515442, "global_step": 126043, "epoch": 1416, "lr": 5.5744649923181144e-05}
{"train_loss": 0.11886077374219894, "global_step": 126044, "epoch": 1416, "lr": 5.574407402528098e-05}
{"train_loss": 0.13490402698516846, "global_step": 126045, "epoch": 1416, "lr": 5.5743498126608576e-05}
{"train_loss": 0.14693188667297363, "global_step": 126046, "epoch": 1416, "lr": 5.574292222716404e-05}
{"train_loss": 0.12932828068733215, "global_step": 126047, "epoch": 1416, "lr": 5.574234632694743e-05}
{"train_loss": 0.08150599896907806, "global_step": 126048, "epoch": 1416, "lr": 5.574177042595883e-05}
{"train_loss": 0.13877151906490326, "global_step": 126049, "epoch": 1416, "lr": 5.5741194524198324e-05}
{"train_loss": 0.16940635442733765, "global_step": 126050, "epoch": 1416, "lr": 5.574061862166598e-05}
{"train_loss": 0.07010099291801453, "global_step": 126051, "epoch": 1416, "lr": 5.5740042718361886e-05}
{"train_loss": 0.09757121652364731, "global_step": 126052, "epoch": 1416, "lr": 5.573946681428611e-05}
{"train_loss": 0.13495977222919464, "global_step": 126053, "epoch": 1416, "lr": 5.573889090943873e-05}
{"train_loss": 0.07437840104103088, "global_step": 126054, "epoch": 1416, "lr": 5.573831500381983e-05}
{"train_loss": 0.1600167453289032, "global_step": 126055, "epoch": 1416, "lr": 5.5737739097429484e-05}
{"train_loss": 0.15239228308200836, "global_step": 126056, "epoch": 1416, "lr": 5.573716319026777e-05}
{"train_loss": 0.112035371363163, "global_step": 126057, "epoch": 1416, "lr": 5.5736587282334766e-05}
{"train_loss": 0.1325669139623642, "global_step": 126058, "epoch": 1416, "lr": 5.5736011373630536e-05}
{"train_loss": 0.09606295078992844, "global_step": 126059, "epoch": 1416, "lr": 5.5735435464155174e-05}
{"train_loss": 0.16524292528629303, "global_step": 126060, "epoch": 1416, "lr": 5.573485955390875e-05}
{"train_loss": 0.1130484789609909, "global_step": 126061, "epoch": 1416, "lr": 5.573428364289135e-05}
{"train_loss": 0.12983892858028412, "global_step": 126062, "epoch": 1416, "lr": 5.573370773110305e-05}
{"train_loss": 0.1465209573507309, "global_step": 126063, "epoch": 1416, "lr": 5.573313181854392e-05}
{"train_loss": 0.10966282337903976, "global_step": 126064, "epoch": 1416, "lr": 5.573255590521405e-05}
{"train_loss": 0.11588802188634872, "global_step": 126065, "epoch": 1416, "lr": 5.573197999111349e-05}
{"train_loss": 0.11594387143850327, "global_step": 126066, "epoch": 1416, "lr": 5.573140407624234e-05}
{"train_loss": 0.12686790525913239, "global_step": 126067, "epoch": 1416, "lr": 5.573082816060067e-05}
{"train_loss": 0.11212044209241867, "global_step": 126068, "epoch": 1416, "lr": 5.5730252244188566e-05}
{"train_loss": 0.12077351659536362, "global_step": 126069, "epoch": 1416, "lr": 5.57296763270061e-05}
{"train_loss": 0.12949597835540771, "global_step": 126070, "epoch": 1416, "lr": 5.572910040905335e-05}
{"train_loss": 0.15058036148548126, "global_step": 126071, "epoch": 1416, "lr": 5.572852449033038e-05}
{"train_loss": 0.11309029906988144, "global_step": 126072, "epoch": 1416, "lr": 5.5727948570837286e-05}
{"train_loss": 0.15690484642982483, "global_step": 126073, "epoch": 1416, "lr": 5.572737265057415e-05}
{"train_loss": 0.0845421627163887, "global_step": 126074, "epoch": 1416, "lr": 5.572679672954102e-05}
{"train_loss": 0.09042783081531525, "global_step": 126075, "epoch": 1416, "lr": 5.572622080773801e-05}
{"train_loss": 0.11172378063201904, "global_step": 126076, "epoch": 1416, "lr": 5.5725644885165165e-05}
{"train_loss": 0.16009774804115295, "global_step": 126077, "epoch": 1416, "lr": 5.5725068961822566e-05}
{"train_loss": 0.08726299554109573, "global_step": 126078, "epoch": 1416, "lr": 5.5724493037710326e-05}
{"train_loss": 0.10865563154220581, "global_step": 126079, "epoch": 1416, "lr": 5.572391711282849e-05}
{"train_loss": 0.1604236662387848, "global_step": 126080, "epoch": 1416, "lr": 5.5723341187177145e-05}
{"train_loss": 0.13853545486927032, "global_step": 126081, "epoch": 1416, "lr": 5.572276526075636e-05}
{"train_loss": 0.10104915499687195, "global_step": 126082, "epoch": 1416, "lr": 5.572218933356622e-05}
{"train_loss": 0.11188175529241562, "global_step": 126083, "epoch": 1416, "lr": 5.572161340560682e-05}
{"train_loss": 0.13476961851119995, "global_step": 126084, "epoch": 1416, "lr": 5.57210374768782e-05}
{"train_loss": 0.09904491901397705, "global_step": 126085, "epoch": 1416, "lr": 5.572046154738046e-05}
{"train_loss": 0.14350195229053497, "global_step": 126086, "epoch": 1416, "lr": 5.5719885617113674e-05}
{"train_loss": 0.17694194614887238, "global_step": 126087, "epoch": 1416, "lr": 5.571930968607792e-05}
{"train_loss": 0.13260440528392792, "global_step": 126088, "epoch": 1416, "lr": 5.571873375427328e-05}
{"train_loss": 0.12607857584953308, "global_step": 126089, "epoch": 1416, "lr": 5.571815782169981e-05}
{"train_loss": 0.058783918619155884, "global_step": 126090, "epoch": 1416, "lr": 5.5717581888357616e-05}
{"train_loss": 0.16522398591041565, "global_step": 126091, "epoch": 1416, "lr": 5.571700595424676e-05}
{"train_loss": 0.11752446740865707, "global_step": 126092, "epoch": 1416, "lr": 5.5716430019367316e-05}
{"train_loss": 0.08856207132339478, "global_step": 126093, "epoch": 1416, "lr": 5.5715854083719375e-05}
{"train_loss": 0.18211452662944794, "global_step": 126094, "epoch": 1416, "lr": 5.5715278147303005e-05}
{"train_loss": 0.10483933985233307, "global_step": 126095, "epoch": 1416, "lr": 5.571470221011829e-05}
{"train_loss": 0.07560349255800247, "global_step": 126096, "epoch": 1416, "lr": 5.571412627216528e-05}
{"train_loss": 0.20520026981830597, "global_step": 126097, "epoch": 1416, "lr": 5.57135503334441e-05}
{"train_loss": 0.10674852132797241, "global_step": 126098, "epoch": 1416, "lr": 5.5712974393954796e-05}
{"train_loss": 0.1274910867214203, "global_step": 126099, "epoch": 1416, "lr": 5.571239845369746e-05}
{"train_loss": 0.20748066902160645, "global_step": 126100, "epoch": 1416, "lr": 5.571182251267215e-05}
{"train_loss": 0.16020146012306213, "global_step": 126101, "epoch": 1416, "lr": 5.571124657087896e-05}
{"train_loss": 0.09588129818439484, "global_step": 126102, "epoch": 1416, "lr": 5.5710670628317974e-05}
{"train_loss": 0.13353703916072845, "global_step": 126103, "epoch": 1416, "lr": 5.571009468498925e-05}
{"train_loss": 0.14888475835323334, "global_step": 126104, "epoch": 1416, "lr": 5.570951874089286e-05}
{"train_loss": 0.09157964587211609, "global_step": 126105, "epoch": 1416, "lr": 5.5708942796028905e-05}
{"train_loss": 0.1615300178527832, "global_step": 126106, "epoch": 1416, "lr": 5.5708366850397454e-05}
{"train_loss": 0.1735079139471054, "global_step": 126107, "epoch": 1416, "lr": 5.570779090399858e-05}
{"train_loss": 0.11260508745908737, "global_step": 126108, "epoch": 1416, "lr": 5.5707214956832364e-05}
{"train_loss": 0.057272568345069885, "global_step": 126109, "epoch": 1416, "lr": 5.5706639008898874e-05}
{"train_loss": 0.05305054411292076, "global_step": 126110, "epoch": 1416, "lr": 5.570606306019821e-05}
{"train_loss": 0.09519460797309875, "global_step": 126111, "epoch": 1416, "lr": 5.570548711073042e-05}
{"train_loss": 0.12719983762402212, "global_step": 126112, "epoch": 1416, "lr": 5.570491116049561e-05, "val_loss": 5.595500946044922}
{"train_loss": 0.11774801462888718, "global_step": 126113, "epoch": 1417, "lr": 5.5704335209493844e-05}
{"train_loss": 0.10971172153949738, "global_step": 126114, "epoch": 1417, "lr": 5.5703759257725185e-05}
{"train_loss": 0.0992635190486908, "global_step": 126115, "epoch": 1417, "lr": 5.570318330518973e-05}
{"train_loss": 0.12924523651599884, "global_step": 126116, "epoch": 1417, "lr": 5.5702607351887556e-05}
{"train_loss": 0.1325354427099228, "global_step": 126117, "epoch": 1417, "lr": 5.570203139781874e-05}
{"train_loss": 0.185713991522789, "global_step": 126118, "epoch": 1417, "lr": 5.570145544298335e-05}
{"train_loss": 0.18296615779399872, "global_step": 126119, "epoch": 1417, "lr": 5.5700879487381474e-05}
{"train_loss": 0.06863870471715927, "global_step": 126120, "epoch": 1417, "lr": 5.570030353101319e-05}
{"train_loss": 0.08531415462493896, "global_step": 126121, "epoch": 1417, "lr": 5.5699727573878556e-05}
{"train_loss": 0.17423495650291443, "global_step": 126122, "epoch": 1417, "lr": 5.5699151615977677e-05}
{"train_loss": 0.08006083220243454, "global_step": 126123, "epoch": 1417, "lr": 5.569857565731061e-05}
{"train_loss": 0.20018059015274048, "global_step": 126124, "epoch": 1417, "lr": 5.5697999697877437e-05}
{"train_loss": 0.0934046059846878, "global_step": 126125, "epoch": 1417, "lr": 5.569742373767823e-05}
{"train_loss": 0.11307545006275177, "global_step": 126126, "epoch": 1417, "lr": 5.569684777671309e-05}
{"train_loss": 0.12118074297904968, "global_step": 126127, "epoch": 1417, "lr": 5.569627181498207e-05}
{"train_loss": 0.15221020579338074, "global_step": 126128, "epoch": 1417, "lr": 5.569569585248525e-05}
{"train_loss": 0.15349480509757996, "global_step": 126129, "epoch": 1417, "lr": 5.569511988922272e-05}
{"train_loss": 0.0765896812081337, "global_step": 126130, "epoch": 1417, "lr": 5.569454392519455e-05}
{"train_loss": 0.11555489897727966, "global_step": 126131, "epoch": 1417, "lr": 5.5693967960400815e-05}
{"train_loss": 0.10831668227910995, "global_step": 126132, "epoch": 1417, "lr": 5.569339199484159e-05}
{"train_loss": 0.16582180559635162, "global_step": 126133, "epoch": 1417, "lr": 5.569281602851696e-05}
{"train_loss": 0.11449822038412094, "global_step": 126134, "epoch": 1417, "lr": 5.5692240061427003e-05}
{"train_loss": 0.15065480768680573, "global_step": 126135, "epoch": 1417, "lr": 5.5691664093571807e-05}
{"train_loss": 0.1550201028585434, "global_step": 126136, "epoch": 1417, "lr": 5.569108812495142e-05}
{"train_loss": 0.16663525998592377, "global_step": 126137, "epoch": 1417, "lr": 5.569051215556595e-05}
{"train_loss": 0.09407878667116165, "global_step": 126138, "epoch": 1417, "lr": 5.568993618541544e-05}
{"train_loss": 0.1433647871017456, "global_step": 126139, "epoch": 1417, "lr": 5.5689360214499994e-05}
{"train_loss": 0.11819852888584137, "global_step": 126140, "epoch": 1417, "lr": 5.568878424281969e-05}
{"train_loss": 0.13039378821849823, "global_step": 126141, "epoch": 1417, "lr": 5.5688208270374606e-05}
{"train_loss": 0.1255066990852356, "global_step": 126142, "epoch": 1417, "lr": 5.5687632297164805e-05}
{"train_loss": 0.12787947058677673, "global_step": 126143, "epoch": 1417, "lr": 5.5687056323190365e-05}
{"train_loss": 0.1061193197965622, "global_step": 126144, "epoch": 1417, "lr": 5.568648034845138e-05}
{"train_loss": 0.08243872225284576, "global_step": 126145, "epoch": 1417, "lr": 5.5685904372947894e-05}
{"train_loss": 0.09062309563159943, "global_step": 126146, "epoch": 1417, "lr": 5.568532839668003e-05}
{"train_loss": 0.13073663413524628, "global_step": 126147, "epoch": 1417, "lr": 5.568475241964783e-05}
{"train_loss": 0.12724068760871887, "global_step": 126148, "epoch": 1417, "lr": 5.568417644185139e-05}
{"train_loss": 0.13483786582946777, "global_step": 126149, "epoch": 1417, "lr": 5.5683600463290786e-05}
{"train_loss": 0.10906215012073517, "global_step": 126150, "epoch": 1417, "lr": 5.568302448396607e-05}
{"train_loss": 0.1047094315290451, "global_step": 126151, "epoch": 1417, "lr": 5.568244850387736e-05}
{"train_loss": 0.12532536685466766, "global_step": 126152, "epoch": 1417, "lr": 5.5681872523024715e-05}
{"train_loss": 0.12244473397731781, "global_step": 126153, "epoch": 1417, "lr": 5.56812965414082e-05}
{"train_loss": 0.1347888559103012, "global_step": 126154, "epoch": 1417, "lr": 5.568072055902791e-05}
{"train_loss": 0.19300596415996552, "global_step": 126155, "epoch": 1417, "lr": 5.568014457588392e-05}
{"train_loss": 0.07884901016950607, "global_step": 126156, "epoch": 1417, "lr": 5.5679568591976304e-05}
{"train_loss": 0.16458824276924133, "global_step": 126157, "epoch": 1417, "lr": 5.567899260730514e-05}
{"train_loss": 0.06856212764978409, "global_step": 126158, "epoch": 1417, "lr": 5.56784166218705e-05}
{"train_loss": 0.14294128119945526, "global_step": 126159, "epoch": 1417, "lr": 5.5677840635672464e-05}
{"train_loss": 0.18134787678718567, "global_step": 126160, "epoch": 1417, "lr": 5.5677264648711124e-05}
{"train_loss": 0.10471972823143005, "global_step": 126161, "epoch": 1417, "lr": 5.567668866098654e-05}
{"train_loss": 0.11632890999317169, "global_step": 126162, "epoch": 1417, "lr": 5.567611267249879e-05}
{"train_loss": 0.1889299899339676, "global_step": 126163, "epoch": 1417, "lr": 5.567553668324795e-05}
{"train_loss": 0.16451101005077362, "global_step": 126164, "epoch": 1417, "lr": 5.5674960693234114e-05}
{"train_loss": 0.08808161318302155, "global_step": 126165, "epoch": 1417, "lr": 5.567438470245735e-05}
{"train_loss": 0.14068745076656342, "global_step": 126166, "epoch": 1417, "lr": 5.567380871091773e-05}
{"train_loss": 0.16515548527240753, "global_step": 126167, "epoch": 1417, "lr": 5.5673232718615334e-05}
{"train_loss": 0.07281671464443207, "global_step": 126168, "epoch": 1417, "lr": 5.567265672555022e-05}
{"train_loss": 0.11583933979272842, "global_step": 126169, "epoch": 1417, "lr": 5.567208073172251e-05}
{"train_loss": 0.0849795714020729, "global_step": 126170, "epoch": 1417, "lr": 5.567150473713226e-05}
{"train_loss": 0.2103336751461029, "global_step": 126171, "epoch": 1417, "lr": 5.5670928741779535e-05}
{"train_loss": 0.2318342924118042, "global_step": 126172, "epoch": 1417, "lr": 5.567035274566443e-05}
{"train_loss": 0.11284049600362778, "global_step": 126173, "epoch": 1417, "lr": 5.566977674878702e-05}
{"train_loss": 0.09430555999279022, "global_step": 126174, "epoch": 1417, "lr": 5.5669200751147366e-05}
{"train_loss": 0.06988491117954254, "global_step": 126175, "epoch": 1417, "lr": 5.5668624752745566e-05}
{"train_loss": 0.13488377630710602, "global_step": 126176, "epoch": 1417, "lr": 5.566804875358168e-05}
{"train_loss": 0.16347724199295044, "global_step": 126177, "epoch": 1417, "lr": 5.56674727536558e-05}
{"train_loss": 0.12942169606685638, "global_step": 126178, "epoch": 1417, "lr": 5.5666896752968003e-05}
{"train_loss": 0.13548091053962708, "global_step": 126179, "epoch": 1417, "lr": 5.5666320751518355e-05}
{"train_loss": 0.2149273157119751, "global_step": 126180, "epoch": 1417, "lr": 5.5665744749306946e-05}
{"train_loss": 0.12357229739427567, "global_step": 126181, "epoch": 1417, "lr": 5.566516874633384e-05}
{"train_loss": 0.16974392533302307, "global_step": 126182, "epoch": 1417, "lr": 5.566459274259912e-05}
{"train_loss": 0.14533264935016632, "global_step": 126183, "epoch": 1417, "lr": 5.566401673810286e-05}
{"train_loss": 0.1656288206577301, "global_step": 126184, "epoch": 1417, "lr": 5.566344073284515e-05}
{"train_loss": 0.07188481837511063, "global_step": 126185, "epoch": 1417, "lr": 5.566286472682605e-05}
{"train_loss": 0.13456028699874878, "global_step": 126186, "epoch": 1417, "lr": 5.5662288720045644e-05}
{"train_loss": 0.14146210253238678, "global_step": 126187, "epoch": 1417, "lr": 5.566171271250402e-05}
{"train_loss": 0.23527783155441284, "global_step": 126188, "epoch": 1417, "lr": 5.5661136704201244e-05}
{"train_loss": 0.10965332388877869, "global_step": 126189, "epoch": 1417, "lr": 5.5660560695137406e-05}
{"train_loss": 0.22714348137378693, "global_step": 126190, "epoch": 1417, "lr": 5.5659984685312574e-05}
{"train_loss": 0.15487153828144073, "global_step": 126191, "epoch": 1417, "lr": 5.565940867472682e-05}
{"train_loss": 0.11076603084802628, "global_step": 126192, "epoch": 1417, "lr": 5.565883266338024e-05}
{"train_loss": 0.08403795957565308, "global_step": 126193, "epoch": 1417, "lr": 5.565825665127288e-05}
{"train_loss": 0.11943309754133224, "global_step": 126194, "epoch": 1417, "lr": 5.565768063840484e-05}
{"train_loss": 0.1536959856748581, "global_step": 126195, "epoch": 1417, "lr": 5.56571046247762e-05}
{"train_loss": 0.09079930186271667, "global_step": 126196, "epoch": 1417, "lr": 5.565652861038704e-05}
{"train_loss": 0.15442615747451782, "global_step": 126197, "epoch": 1417, "lr": 5.5655952595237414e-05}
{"train_loss": 0.12995046377182007, "global_step": 126198, "epoch": 1417, "lr": 5.565537657932742e-05}
{"train_loss": 0.14752277731895447, "global_step": 126199, "epoch": 1417, "lr": 5.565480056265714e-05}
{"train_loss": 0.1062697023153305, "global_step": 126200, "epoch": 1417, "lr": 5.565422454522663e-05}
{"train_loss": 0.1318954246097736, "global_step": 126201, "epoch": 1417, "lr": 5.565364852703597e-05, "val_loss": 5.471813678741455}
{"train_loss": 0.1371019035577774, "global_step": 126202, "epoch": 1418, "lr": 5.565307250808526e-05}
{"train_loss": 0.06207041069865227, "global_step": 126203, "epoch": 1418, "lr": 5.5652496488374564e-05}
{"train_loss": 0.12214658409357071, "global_step": 126204, "epoch": 1418, "lr": 5.565192046790393e-05}
{"train_loss": 0.0931473970413208, "global_step": 126205, "epoch": 1418, "lr": 5.5651344446673494e-05}
{"train_loss": 0.10715819150209427, "global_step": 126206, "epoch": 1418, "lr": 5.5650768424683284e-05}
{"train_loss": 0.07318780571222305, "global_step": 126207, "epoch": 1418, "lr": 5.565019240193341e-05}
{"train_loss": 0.113321952521801, "global_step": 126208, "epoch": 1418, "lr": 5.564961637842394e-05}
{"train_loss": 0.11036416888237, "global_step": 126209, "epoch": 1418, "lr": 5.564904035415495e-05}
{"train_loss": 0.1511053442955017, "global_step": 126210, "epoch": 1418, "lr": 5.5648464329126506e-05}
{"train_loss": 0.17998045682907104, "global_step": 126211, "epoch": 1418, "lr": 5.5647888303338694e-05}
{"train_loss": 0.10760024935007095, "global_step": 126212, "epoch": 1418, "lr": 5.5647312276791595e-05}
{"train_loss": 0.1616334468126297, "global_step": 126213, "epoch": 1418, "lr": 5.564673624948529e-05}
{"train_loss": 0.09952739626169205, "global_step": 126214, "epoch": 1418, "lr": 5.564616022141984e-05}
{"train_loss": 0.11459264159202576, "global_step": 126215, "epoch": 1418, "lr": 5.5645584192595334e-05}
{"train_loss": 0.11544707417488098, "global_step": 126216, "epoch": 1418, "lr": 5.564500816301186e-05}
{"train_loss": 0.21894371509552002, "global_step": 126217, "epoch": 1418, "lr": 5.564443213266947e-05}
{"train_loss": 0.09532038122415543, "global_step": 126218, "epoch": 1418, "lr": 5.5643856101568274e-05}
{"train_loss": 0.0970105305314064, "global_step": 126219, "epoch": 1418, "lr": 5.564328006970831e-05}
{"train_loss": 0.11072780936956406, "global_step": 126220, "epoch": 1418, "lr": 5.564270403708969e-05}
{"train_loss": 0.14346368610858917, "global_step": 126221, "epoch": 1418, "lr": 5.5642128003712454e-05}
{"train_loss": 0.12015034258365631, "global_step": 126222, "epoch": 1418, "lr": 5.564155196957672e-05}
{"train_loss": 0.12399258464574814, "global_step": 126223, "epoch": 1418, "lr": 5.564097593468254e-05}
{"train_loss": 0.15516535937786102, "global_step": 126224, "epoch": 1418, "lr": 5.5640399899030003e-05}
{"train_loss": 0.08780233561992645, "global_step": 126225, "epoch": 1418, "lr": 5.563982386261919e-05}
{"train_loss": 0.12641525268554688, "global_step": 126226, "epoch": 1418, "lr": 5.563924782545016e-05}
{"train_loss": 0.08084671944379807, "global_step": 126227, "epoch": 1418, "lr": 5.563867178752301e-05}
{"train_loss": 0.09979582577943802, "global_step": 126228, "epoch": 1418, "lr": 5.5638095748837805e-05}
{"train_loss": 0.13646772503852844, "global_step": 126229, "epoch": 1418, "lr": 5.5637519709394636e-05}
{"train_loss": 0.1331169754266739, "global_step": 126230, "epoch": 1418, "lr": 5.563694366919356e-05}
{"train_loss": 0.14797748625278473, "global_step": 126231, "epoch": 1418, "lr": 5.563636762823468e-05}
{"train_loss": 0.1048840582370758, "global_step": 126232, "epoch": 1418, "lr": 5.563579158651804e-05}
{"train_loss": 0.16544321179389954, "global_step": 126233, "epoch": 1418, "lr": 5.563521554404375e-05}
{"train_loss": 0.08772072196006775, "global_step": 126234, "epoch": 1418, "lr": 5.5634639500811866e-05}
{"train_loss": 0.08311665803194046, "global_step": 126235, "epoch": 1418, "lr": 5.563406345682248e-05}
{"train_loss": 0.09626451879739761, "global_step": 126236, "epoch": 1418, "lr": 5.563348741207566e-05}
{"train_loss": 0.1068667620420456, "global_step": 126237, "epoch": 1418, "lr": 5.563291136657148e-05}
{"train_loss": 0.13506080210208893, "global_step": 126238, "epoch": 1418, "lr": 5.5632335320310044e-05}
{"train_loss": 0.12316831201314926, "global_step": 126239, "epoch": 1418, "lr": 5.563175927329139e-05}
{"train_loss": 0.15621471405029297, "global_step": 126240, "epoch": 1418, "lr": 5.563118322551563e-05}
{"train_loss": 0.1596919596195221, "global_step": 126241, "epoch": 1418, "lr": 5.56306071769828e-05}
{"train_loss": 0.058427613228559494, "global_step": 126242, "epoch": 1418, "lr": 5.5630031127693015e-05}
{"train_loss": 0.12224602699279785, "global_step": 126243, "epoch": 1418, "lr": 5.5629455077646355e-05}
{"train_loss": 0.0977025255560875, "global_step": 126244, "epoch": 1418, "lr": 5.5628879026842875e-05}
{"train_loss": 0.14753690361976624, "global_step": 126245, "epoch": 1418, "lr": 5.5628302975282656e-05}
{"train_loss": 0.10139581561088562, "global_step": 126246, "epoch": 1418, "lr": 5.5627726922965786e-05}
{"train_loss": 0.15781566500663757, "global_step": 126247, "epoch": 1418, "lr": 5.562715086989233e-05}
{"train_loss": 0.1860923320055008, "global_step": 126248, "epoch": 1418, "lr": 5.562657481606237e-05}
{"train_loss": 0.17109274864196777, "global_step": 126249, "epoch": 1418, "lr": 5.562599876147599e-05}
{"train_loss": 0.20977769792079926, "global_step": 126250, "epoch": 1418, "lr": 5.562542270613327e-05}
{"train_loss": 0.09615671634674072, "global_step": 126251, "epoch": 1418, "lr": 5.562484665003428e-05}
{"train_loss": 0.2547342777252197, "global_step": 126252, "epoch": 1418, "lr": 5.56242705931791e-05}
{"train_loss": 0.1615801900625229, "global_step": 126253, "epoch": 1418, "lr": 5.56236945355678e-05}
{"train_loss": 0.14189413189888, "global_step": 126254, "epoch": 1418, "lr": 5.5623118477200455e-05}
{"train_loss": 0.15566343069076538, "global_step": 126255, "epoch": 1418, "lr": 5.562254241807716e-05}
{"train_loss": 0.11735183000564575, "global_step": 126256, "epoch": 1418, "lr": 5.562196635819797e-05}
{"train_loss": 0.136521115899086, "global_step": 126257, "epoch": 1418, "lr": 5.5621390297563e-05}
{"train_loss": 0.08587834239006042, "global_step": 126258, "epoch": 1418, "lr": 5.5620814236172294e-05}
{"train_loss": 0.11936833709478378, "global_step": 126259, "epoch": 1418, "lr": 5.562023817402592e-05}
{"train_loss": 0.08987174928188324, "global_step": 126260, "epoch": 1418, "lr": 5.561966211112398e-05}
{"train_loss": 0.14250758290290833, "global_step": 126261, "epoch": 1418, "lr": 5.561908604746655e-05}
{"train_loss": 0.11383888870477676, "global_step": 126262, "epoch": 1418, "lr": 5.56185099830537e-05}
{"train_loss": 0.11808554083108902, "global_step": 126263, "epoch": 1418, "lr": 5.561793391788551e-05}
{"train_loss": 0.11357763409614563, "global_step": 126264, "epoch": 1418, "lr": 5.561735785196206e-05}
{"train_loss": 0.1532396376132965, "global_step": 126265, "epoch": 1418, "lr": 5.5616781785283426e-05}
{"train_loss": 0.09318656474351883, "global_step": 126266, "epoch": 1418, "lr": 5.561620571784969e-05}
{"train_loss": 0.18829438090324402, "global_step": 126267, "epoch": 1418, "lr": 5.5615629649660904e-05}
{"train_loss": 0.1566624641418457, "global_step": 126268, "epoch": 1418, "lr": 5.561505358071718e-05}
{"train_loss": 0.07445473968982697, "global_step": 126269, "epoch": 1418, "lr": 5.561447751101858e-05}
{"train_loss": 0.14447064697742462, "global_step": 126270, "epoch": 1418, "lr": 5.5613901440565176e-05}
{"train_loss": 0.09707123786211014, "global_step": 126271, "epoch": 1418, "lr": 5.561332536935706e-05}
{"train_loss": 0.15001647174358368, "global_step": 126272, "epoch": 1418, "lr": 5.5612749297394295e-05}
{"train_loss": 0.15783631801605225, "global_step": 126273, "epoch": 1418, "lr": 5.561217322467697e-05}
{"train_loss": 0.16894559562206268, "global_step": 126274, "epoch": 1418, "lr": 5.561159715120514e-05}
{"train_loss": 0.09135337173938751, "global_step": 126275, "epoch": 1418, "lr": 5.561102107697892e-05}
{"train_loss": 0.1271069496870041, "global_step": 126276, "epoch": 1418, "lr": 5.561044500199837e-05}
{"train_loss": 0.0960288792848587, "global_step": 126277, "epoch": 1418, "lr": 5.560986892626353e-05}
{"train_loss": 0.09828677773475647, "global_step": 126278, "epoch": 1418, "lr": 5.560929284977454e-05}
{"train_loss": 0.14042532444000244, "global_step": 126279, "epoch": 1418, "lr": 5.560871677253143e-05}
{"train_loss": 0.15006373822689056, "global_step": 126280, "epoch": 1418, "lr": 5.560814069453432e-05}
{"train_loss": 0.18057997524738312, "global_step": 126281, "epoch": 1418, "lr": 5.560756461578325e-05}
{"train_loss": 0.1937129646539688, "global_step": 126282, "epoch": 1418, "lr": 5.5606988536278316e-05}
{"train_loss": 0.18249095976352692, "global_step": 126283, "epoch": 1418, "lr": 5.5606412456019584e-05}
{"train_loss": 0.09197775274515152, "global_step": 126284, "epoch": 1418, "lr": 5.5605836375007135e-05}
{"train_loss": 0.1614464819431305, "global_step": 126285, "epoch": 1418, "lr": 5.560526029324107e-05}
{"train_loss": 0.08457483351230621, "global_step": 126286, "epoch": 1418, "lr": 5.5604684210721424e-05}
{"train_loss": 0.14088955521583557, "global_step": 126287, "epoch": 1418, "lr": 5.5604108127448305e-05}
{"train_loss": 0.12936913967132568, "global_step": 126288, "epoch": 1418, "lr": 5.560353204342178e-05}
{"train_loss": 0.09768054634332657, "global_step": 126289, "epoch": 1418, "lr": 5.560295595864193e-05}
{"train_loss": 0.12933865976467562, "global_step": 126290, "epoch": 1418, "lr": 5.5602379873108834e-05, "val_loss": 5.530022144317627}
{"train_loss": 0.1257513463497162, "global_step": 126291, "epoch": 1419, "lr": 5.560180378682257e-05}
{"train_loss": 0.14304931461811066, "global_step": 126292, "epoch": 1419, "lr": 5.56012276997832e-05}
{"train_loss": 0.0945817306637764, "global_step": 126293, "epoch": 1419, "lr": 5.5600651611990815e-05}
{"train_loss": 0.11668368428945541, "global_step": 126294, "epoch": 1419, "lr": 5.5600075523445506e-05}
{"train_loss": 0.11863894760608673, "global_step": 126295, "epoch": 1419, "lr": 5.559949943414732e-05}
{"train_loss": 0.08870863914489746, "global_step": 126296, "epoch": 1419, "lr": 5.559892334409636e-05}
{"train_loss": 0.11104732006788254, "global_step": 126297, "epoch": 1419, "lr": 5.559834725329268e-05}
{"train_loss": 0.16387315094470978, "global_step": 126298, "epoch": 1419, "lr": 5.559777116173638e-05}
{"train_loss": 0.12458760291337967, "global_step": 126299, "epoch": 1419, "lr": 5.559719506942752e-05}
{"train_loss": 0.1290145218372345, "global_step": 126300, "epoch": 1419, "lr": 5.559661897636619e-05}
{"train_loss": 0.11232709884643555, "global_step": 126301, "epoch": 1419, "lr": 5.5596042882552457e-05}
{"train_loss": 0.05557818338274956, "global_step": 126302, "epoch": 1419, "lr": 5.5595466787986416e-05}
{"train_loss": 0.13410337269306183, "global_step": 126303, "epoch": 1419, "lr": 5.559489069266812e-05}
{"train_loss": 0.0900462344288826, "global_step": 126304, "epoch": 1419, "lr": 5.559431459659768e-05}
{"train_loss": 0.19803334772586823, "global_step": 126305, "epoch": 1419, "lr": 5.559373849977513e-05}
{"train_loss": 0.15636053681373596, "global_step": 126306, "epoch": 1419, "lr": 5.559316240220058e-05}
{"train_loss": 0.08095964789390564, "global_step": 126307, "epoch": 1419, "lr": 5.5592586303874095e-05}
{"train_loss": 0.08973361551761627, "global_step": 126308, "epoch": 1419, "lr": 5.5592010204795754e-05}
{"train_loss": 0.11331207305192947, "global_step": 126309, "epoch": 1419, "lr": 5.5591434104965635e-05}
{"train_loss": 0.09200350195169449, "global_step": 126310, "epoch": 1419, "lr": 5.559085800438382e-05}
{"train_loss": 0.08021798729896545, "global_step": 126311, "epoch": 1419, "lr": 5.559028190305038e-05}
{"train_loss": 0.11504333466291428, "global_step": 126312, "epoch": 1419, "lr": 5.5589705800965395e-05}
{"train_loss": 0.08790109306573868, "global_step": 126313, "epoch": 1419, "lr": 5.558912969812894e-05}
{"train_loss": 0.11624972522258759, "global_step": 126314, "epoch": 1419, "lr": 5.5588553594541096e-05}
{"train_loss": 0.13225029408931732, "global_step": 126315, "epoch": 1419, "lr": 5.558797749020195e-05}
{"train_loss": 0.194258451461792, "global_step": 126316, "epoch": 1419, "lr": 5.558740138511155e-05}
{"train_loss": 0.14070190489292145, "global_step": 126317, "epoch": 1419, "lr": 5.5586825279270005e-05}
{"train_loss": 0.07020741701126099, "global_step": 126318, "epoch": 1419, "lr": 5.558624917267737e-05}
{"train_loss": 0.11654256284236908, "global_step": 126319, "epoch": 1419, "lr": 5.558567306533373e-05}
{"train_loss": 0.13629038631916046, "global_step": 126320, "epoch": 1419, "lr": 5.558509695723918e-05}
{"train_loss": 0.10818947851657867, "global_step": 126321, "epoch": 1419, "lr": 5.5584520848393774e-05}
{"train_loss": 0.14942674338817596, "global_step": 126322, "epoch": 1419, "lr": 5.558394473879759e-05}
{"train_loss": 0.1353544294834137, "global_step": 126323, "epoch": 1419, "lr": 5.558336862845073e-05}
{"train_loss": 0.17357808351516724, "global_step": 126324, "epoch": 1419, "lr": 5.5582792517353234e-05}
{"train_loss": 0.15244920551776886, "global_step": 126325, "epoch": 1419, "lr": 5.558221640550521e-05}
{"train_loss": 0.1519734263420105, "global_step": 126326, "epoch": 1419, "lr": 5.558164029290672e-05}
{"train_loss": 0.18092145025730133, "global_step": 126327, "epoch": 1419, "lr": 5.558106417955785e-05}
{"train_loss": 0.10276514291763306, "global_step": 126328, "epoch": 1419, "lr": 5.558048806545868e-05}
{"train_loss": 0.10960801690816879, "global_step": 126329, "epoch": 1419, "lr": 5.557991195060928e-05}
{"train_loss": 0.08455845713615417, "global_step": 126330, "epoch": 1419, "lr": 5.5579335835009715e-05}
{"train_loss": 0.1553608626127243, "global_step": 126331, "epoch": 1419, "lr": 5.557875971866009e-05}
{"train_loss": 0.12652406096458435, "global_step": 126332, "epoch": 1419, "lr": 5.5578183601560464e-05}
{"train_loss": 0.23840942978858948, "global_step": 126333, "epoch": 1419, "lr": 5.557760748371091e-05}
{"train_loss": 0.15929047763347626, "global_step": 126334, "epoch": 1419, "lr": 5.557703136511153e-05}
{"train_loss": 0.13297003507614136, "global_step": 126335, "epoch": 1419, "lr": 5.5576455245762384e-05}
{"train_loss": 0.11494063585996628, "global_step": 126336, "epoch": 1419, "lr": 5.5575879125663545e-05}
{"train_loss": 0.16525374352931976, "global_step": 126337, "epoch": 1419, "lr": 5.55753030048151e-05}
{"train_loss": 0.13161399960517883, "global_step": 126338, "epoch": 1419, "lr": 5.5574726883217124e-05}
{"train_loss": 0.12546716630458832, "global_step": 126339, "epoch": 1419, "lr": 5.557415076086969e-05}
{"train_loss": 0.10387679189443588, "global_step": 126340, "epoch": 1419, "lr": 5.557357463777288e-05}
{"train_loss": 0.08395812660455704, "global_step": 126341, "epoch": 1419, "lr": 5.557299851392678e-05}
{"train_loss": 0.09287340939044952, "global_step": 126342, "epoch": 1419, "lr": 5.557242238933145e-05}
{"train_loss": 0.14695118367671967, "global_step": 126343, "epoch": 1419, "lr": 5.557184626398698e-05}
{"train_loss": 0.18261674046516418, "global_step": 126344, "epoch": 1419, "lr": 5.557127013789344e-05}
{"train_loss": 0.12362895160913467, "global_step": 126345, "epoch": 1419, "lr": 5.5570694011050903e-05}
{"train_loss": 0.1568853259086609, "global_step": 126346, "epoch": 1419, "lr": 5.557011788345947e-05}
{"train_loss": 0.11465609818696976, "global_step": 126347, "epoch": 1419, "lr": 5.556954175511919e-05}
{"train_loss": 0.12238916754722595, "global_step": 126348, "epoch": 1419, "lr": 5.556896562603016e-05}
{"train_loss": 0.2075386941432953, "global_step": 126349, "epoch": 1419, "lr": 5.5568389496192454e-05}
{"train_loss": 0.174984872341156, "global_step": 126350, "epoch": 1419, "lr": 5.5567813365606134e-05}
{"train_loss": 0.10574805736541748, "global_step": 126351, "epoch": 1419, "lr": 5.5567237234271295e-05}
{"train_loss": 0.156902015209198, "global_step": 126352, "epoch": 1419, "lr": 5.556666110218801e-05}
{"train_loss": 0.1942668855190277, "global_step": 126353, "epoch": 1419, "lr": 5.5566084969356356e-05}
{"train_loss": 0.1188008040189743, "global_step": 126354, "epoch": 1419, "lr": 5.5565508835776405e-05}
{"train_loss": 0.15710744261741638, "global_step": 126355, "epoch": 1419, "lr": 5.5564932701448246e-05}
{"train_loss": 0.1389947235584259, "global_step": 126356, "epoch": 1419, "lr": 5.556435656637195e-05}
{"train_loss": 0.22726692259311676, "global_step": 126357, "epoch": 1419, "lr": 5.556378043054758e-05}
{"train_loss": 0.11418920010328293, "global_step": 126358, "epoch": 1419, "lr": 5.556320429397525e-05}
{"train_loss": 0.1695406436920166, "global_step": 126359, "epoch": 1419, "lr": 5.5562628156654995e-05}
{"train_loss": 0.1104850172996521, "global_step": 126360, "epoch": 1419, "lr": 5.556205201858692e-05}
{"train_loss": 0.11402900516986847, "global_step": 126361, "epoch": 1419, "lr": 5.5561475879771094e-05}
{"train_loss": 0.14545203745365143, "global_step": 126362, "epoch": 1419, "lr": 5.5560899740207594e-05}
{"train_loss": 0.11355588585138321, "global_step": 126363, "epoch": 1419, "lr": 5.55603235998965e-05}
{"train_loss": 0.12317440658807755, "global_step": 126364, "epoch": 1419, "lr": 5.5559747458837894e-05}
{"train_loss": 0.15087443590164185, "global_step": 126365, "epoch": 1419, "lr": 5.5559171317031844e-05}
{"train_loss": 0.11986535042524338, "global_step": 126366, "epoch": 1419, "lr": 5.555859517447843e-05}
{"train_loss": 0.09295192360877991, "global_step": 126367, "epoch": 1419, "lr": 5.5558019031177735e-05}
{"train_loss": 0.20692668855190277, "global_step": 126368, "epoch": 1419, "lr": 5.5557442887129826e-05}
{"train_loss": 0.10172387957572937, "global_step": 126369, "epoch": 1419, "lr": 5.555686674233479e-05}
{"train_loss": 0.18696630001068115, "global_step": 126370, "epoch": 1419, "lr": 5.555629059679269e-05}
{"train_loss": 0.17190863192081451, "global_step": 126371, "epoch": 1419, "lr": 5.555571445050363e-05}
{"train_loss": 0.21611301600933075, "global_step": 126372, "epoch": 1419, "lr": 5.555513830346767e-05}
{"train_loss": 0.1624995470046997, "global_step": 126373, "epoch": 1419, "lr": 5.555456215568489e-05}
{"train_loss": 0.1689448058605194, "global_step": 126374, "epoch": 1419, "lr": 5.5553986007155354e-05}
{"train_loss": 0.09980568289756775, "global_step": 126375, "epoch": 1419, "lr": 5.555340985787916e-05}
{"train_loss": 0.1701257973909378, "global_step": 126376, "epoch": 1419, "lr": 5.555283370785638e-05}
{"train_loss": 0.1390552520751953, "global_step": 126377, "epoch": 1419, "lr": 5.555225755708709e-05}
{"train_loss": 0.14477868378162384, "global_step": 126378, "epoch": 1419, "lr": 5.555168140557136e-05}
{"train_loss": 0.13459593907333492, "global_step": 126379, "epoch": 1419, "lr": 5.555110525330928e-05, "val_loss": 5.429080009460449}
{"train_loss": 0.15329648554325104, "global_step": 126380, "epoch": 1420, "lr": 5.555052910030092e-05}
{"train_loss": 0.10550948232412338, "global_step": 126381, "epoch": 1420, "lr": 5.554995294654636e-05}
{"train_loss": 0.08418891578912735, "global_step": 126382, "epoch": 1420, "lr": 5.554937679204567e-05}
{"train_loss": 0.11684166640043259, "global_step": 126383, "epoch": 1420, "lr": 5.5548800636798935e-05}
{"train_loss": 0.13809187710285187, "global_step": 126384, "epoch": 1420, "lr": 5.554822448080624e-05}
{"train_loss": 0.11229556053876877, "global_step": 126385, "epoch": 1420, "lr": 5.5547648324067644e-05}
{"train_loss": 0.10798215866088867, "global_step": 126386, "epoch": 1420, "lr": 5.554707216658324e-05}
{"train_loss": 0.10346945375204086, "global_step": 126387, "epoch": 1420, "lr": 5.554649600835311e-05}
{"train_loss": 0.17686446011066437, "global_step": 126388, "epoch": 1420, "lr": 5.55459198493773e-05}
{"train_loss": 0.16447235643863678, "global_step": 126389, "epoch": 1420, "lr": 5.554534368965593e-05}
{"train_loss": 0.12344392389059067, "global_step": 126390, "epoch": 1420, "lr": 5.5544767529189045e-05}
{"train_loss": 0.16245225071907043, "global_step": 126391, "epoch": 1420, "lr": 5.5544191367976724e-05}
{"train_loss": 0.1823553740978241, "global_step": 126392, "epoch": 1420, "lr": 5.5543615206019076e-05}
{"train_loss": 0.10411344468593597, "global_step": 126393, "epoch": 1420, "lr": 5.554303904331613e-05}
{"train_loss": 0.07746845483779907, "global_step": 126394, "epoch": 1420, "lr": 5.554246287986802e-05}
{"train_loss": 0.10389745980501175, "global_step": 126395, "epoch": 1420, "lr": 5.554188671567477e-05}
{"train_loss": 0.170648455619812, "global_step": 126396, "epoch": 1420, "lr": 5.554131055073649e-05}
{"train_loss": 0.19544440507888794, "global_step": 126397, "epoch": 1420, "lr": 5.554073438505325e-05}
{"train_loss": 0.1835375875234604, "global_step": 126398, "epoch": 1420, "lr": 5.554015821862512e-05}
{"train_loss": 0.24449311196804047, "global_step": 126399, "epoch": 1420, "lr": 5.5539582051452186e-05}
{"train_loss": 0.24018746614456177, "global_step": 126400, "epoch": 1420, "lr": 5.553900588353452e-05}
{"train_loss": 0.16237212717533112, "global_step": 126401, "epoch": 1420, "lr": 5.553842971487221e-05}
{"train_loss": 0.11835413426160812, "global_step": 126402, "epoch": 1420, "lr": 5.5537853545465325e-05}
{"train_loss": 0.1467493623495102, "global_step": 126403, "epoch": 1420, "lr": 5.5537277375313934e-05}
{"train_loss": 0.11003994941711426, "global_step": 126404, "epoch": 1420, "lr": 5.5536701204418116e-05}
{"train_loss": 0.1631237119436264, "global_step": 126405, "epoch": 1420, "lr": 5.5536125032777966e-05}
{"train_loss": 0.16577906906604767, "global_step": 126406, "epoch": 1420, "lr": 5.553554886039356e-05}
{"train_loss": 0.12567870318889618, "global_step": 126407, "epoch": 1420, "lr": 5.553497268726496e-05}
{"train_loss": 0.1392306685447693, "global_step": 126408, "epoch": 1420, "lr": 5.5534396513392264e-05}
{"train_loss": 0.12620973587036133, "global_step": 126409, "epoch": 1420, "lr": 5.5533820338775525e-05}
{"train_loss": 0.1602175533771515, "global_step": 126410, "epoch": 1420, "lr": 5.553324416341482e-05}
{"train_loss": 0.13460907340049744, "global_step": 126411, "epoch": 1420, "lr": 5.553266798731025e-05}
{"train_loss": 0.059234123677015305, "global_step": 126412, "epoch": 1420, "lr": 5.553209181046187e-05}
{"train_loss": 0.19451405107975006, "global_step": 126413, "epoch": 1420, "lr": 5.553151563286978e-05}
{"train_loss": 0.1677122861146927, "global_step": 126414, "epoch": 1420, "lr": 5.5530939454534035e-05}
{"train_loss": 0.13906081020832062, "global_step": 126415, "epoch": 1420, "lr": 5.553036327545473e-05}
{"train_loss": 0.13033431768417358, "global_step": 126416, "epoch": 1420, "lr": 5.552978709563193e-05}
{"train_loss": 0.13531550765037537, "global_step": 126417, "epoch": 1420, "lr": 5.552921091506572e-05}
{"train_loss": 0.1902705281972885, "global_step": 126418, "epoch": 1420, "lr": 5.552863473375618e-05}
{"train_loss": 0.19771531224250793, "global_step": 126419, "epoch": 1420, "lr": 5.5528058551703377e-05}
{"train_loss": 0.08724746853113174, "global_step": 126420, "epoch": 1420, "lr": 5.552748236890739e-05}
{"train_loss": 0.1325138807296753, "global_step": 126421, "epoch": 1420, "lr": 5.552690618536831e-05}
{"train_loss": 0.12496684491634369, "global_step": 126422, "epoch": 1420, "lr": 5.5526330001086194e-05}
{"train_loss": 0.15510782599449158, "global_step": 126423, "epoch": 1420, "lr": 5.5525753816061134e-05}
{"train_loss": 0.11926253139972687, "global_step": 126424, "epoch": 1420, "lr": 5.55251776302932e-05}
{"train_loss": 0.08603175729513168, "global_step": 126425, "epoch": 1420, "lr": 5.5524601443782485e-05}
{"train_loss": 0.10238299518823624, "global_step": 126426, "epoch": 1420, "lr": 5.552402525652906e-05}
{"train_loss": 0.11198198050260544, "global_step": 126427, "epoch": 1420, "lr": 5.552344906853299e-05}
{"train_loss": 0.11613311618566513, "global_step": 126428, "epoch": 1420, "lr": 5.552287287979435e-05}
{"train_loss": 0.1847960352897644, "global_step": 126429, "epoch": 1420, "lr": 5.5522296690313235e-05}
{"train_loss": 0.19146116077899933, "global_step": 126430, "epoch": 1420, "lr": 5.5521720500089704e-05}
{"train_loss": 0.06525979936122894, "global_step": 126431, "epoch": 1420, "lr": 5.5521144309123854e-05}
{"train_loss": 0.16148191690444946, "global_step": 126432, "epoch": 1420, "lr": 5.552056811741575e-05}
{"train_loss": 0.14218996465206146, "global_step": 126433, "epoch": 1420, "lr": 5.551999192496547e-05}
{"train_loss": 0.15457358956336975, "global_step": 126434, "epoch": 1420, "lr": 5.5519415731773105e-05}
{"train_loss": 0.12948331236839294, "global_step": 126435, "epoch": 1420, "lr": 5.5518839537838716e-05}
{"train_loss": 0.11980631202459335, "global_step": 126436, "epoch": 1420, "lr": 5.5518263343162394e-05}
{"train_loss": 0.13402624428272247, "global_step": 126437, "epoch": 1420, "lr": 5.55176871477442e-05}
{"train_loss": 0.09075018763542175, "global_step": 126438, "epoch": 1420, "lr": 5.5517110951584214e-05}
{"train_loss": 0.14963386952877045, "global_step": 126439, "epoch": 1420, "lr": 5.551653475468253e-05}
{"train_loss": 0.1674453169107437, "global_step": 126440, "epoch": 1420, "lr": 5.5515958557039204e-05}
{"train_loss": 0.12165074050426483, "global_step": 126441, "epoch": 1420, "lr": 5.5515382358654334e-05}
{"train_loss": 0.11520251631736755, "global_step": 126442, "epoch": 1420, "lr": 5.5514806159527985e-05}
{"train_loss": 0.11780601739883423, "global_step": 126443, "epoch": 1420, "lr": 5.5514229959660246e-05}
{"train_loss": 0.1180901974439621, "global_step": 126444, "epoch": 1420, "lr": 5.5513653759051176e-05}
{"train_loss": 0.09759571403265, "global_step": 126445, "epoch": 1420, "lr": 5.551307755770088e-05}
{"train_loss": 0.13463646173477173, "global_step": 126446, "epoch": 1420, "lr": 5.5512501355609405e-05}
{"train_loss": 0.17405402660369873, "global_step": 126447, "epoch": 1420, "lr": 5.5511925152776834e-05}
{"train_loss": 0.10303574800491333, "global_step": 126448, "epoch": 1420, "lr": 5.551134894920327e-05}
{"train_loss": 0.23975524306297302, "global_step": 126449, "epoch": 1420, "lr": 5.551077274488875e-05}
{"train_loss": 0.12441319972276688, "global_step": 126450, "epoch": 1420, "lr": 5.5510196539833394e-05}
{"train_loss": 0.1350778192281723, "global_step": 126451, "epoch": 1420, "lr": 5.550962033403724e-05}
{"train_loss": 0.20837561786174774, "global_step": 126452, "epoch": 1420, "lr": 5.55090441275004e-05}
{"train_loss": 0.06861729919910431, "global_step": 126453, "epoch": 1420, "lr": 5.550846792022293e-05}
{"train_loss": 0.16830284893512726, "global_step": 126454, "epoch": 1420, "lr": 5.5507891712204905e-05}
{"train_loss": 0.23783424496650696, "global_step": 126455, "epoch": 1420, "lr": 5.550731550344642e-05}
{"train_loss": 0.1364600956439972, "global_step": 126456, "epoch": 1420, "lr": 5.550673929394755e-05}
{"train_loss": 0.08532240241765976, "global_step": 126457, "epoch": 1420, "lr": 5.550616308370836e-05}
{"train_loss": 0.15382911264896393, "global_step": 126458, "epoch": 1420, "lr": 5.550558687272892e-05}
{"train_loss": 0.12624885141849518, "global_step": 126459, "epoch": 1420, "lr": 5.550501066100934e-05}
{"train_loss": 0.10598161816596985, "global_step": 126460, "epoch": 1420, "lr": 5.550443444854967e-05}
{"train_loss": 0.13352222740650177, "global_step": 126461, "epoch": 1420, "lr": 5.550385823535e-05}
{"train_loss": 0.12188149243593216, "global_step": 126462, "epoch": 1420, "lr": 5.5503282021410396e-05}
{"train_loss": 0.15292853116989136, "global_step": 126463, "epoch": 1420, "lr": 5.550270580673095e-05}
{"train_loss": 0.11811869591474533, "global_step": 126464, "epoch": 1420, "lr": 5.550212959131173e-05}
{"train_loss": 0.15683245658874512, "global_step": 126465, "epoch": 1420, "lr": 5.550155337515284e-05}
{"train_loss": 0.13550516963005066, "global_step": 126466, "epoch": 1420, "lr": 5.5500977158254296e-05}
{"train_loss": 0.10937312990427017, "global_step": 126467, "epoch": 1420, "lr": 5.550040094061624e-05}
{"train_loss": 0.1394311564273379, "global_step": 126468, "epoch": 1420, "lr": 5.54998247222387e-05, "val_loss": 5.440546989440918, "train_action_mse_error": 20.520776748657227}
{"train_loss": 0.09391909837722778, "global_step": 126469, "epoch": 1421, "lr": 5.549924850312178e-05}
{"train_loss": 0.1509614735841751, "global_step": 126470, "epoch": 1421, "lr": 5.5498672283265566e-05}
{"train_loss": 0.11040487885475159, "global_step": 126471, "epoch": 1421, "lr": 5.549809606267011e-05}
{"train_loss": 0.10145387798547745, "global_step": 126472, "epoch": 1421, "lr": 5.549751984133551e-05}
{"train_loss": 0.21199454367160797, "global_step": 126473, "epoch": 1421, "lr": 5.5496943619261835e-05}
{"train_loss": 0.12190491706132889, "global_step": 126474, "epoch": 1421, "lr": 5.549636739644916e-05}
{"train_loss": 0.1146390438079834, "global_step": 126475, "epoch": 1421, "lr": 5.549579117289756e-05}
{"train_loss": 0.13972288370132446, "global_step": 126476, "epoch": 1421, "lr": 5.549521494860712e-05}
{"train_loss": 0.10839127004146576, "global_step": 126477, "epoch": 1421, "lr": 5.549463872357791e-05}
{"train_loss": 0.1416722685098648, "global_step": 126478, "epoch": 1421, "lr": 5.5494062497810015e-05}
{"train_loss": 0.10533926635980606, "global_step": 126479, "epoch": 1421, "lr": 5.5493486271303516e-05}
{"train_loss": 0.15507502853870392, "global_step": 126480, "epoch": 1421, "lr": 5.5492910044058486e-05}
{"train_loss": 0.07200375944375992, "global_step": 126481, "epoch": 1421, "lr": 5.5492333816075005e-05}
{"train_loss": 0.12394922226667404, "global_step": 126482, "epoch": 1421, "lr": 5.5491757587353136e-05}
{"train_loss": 0.12116939574480057, "global_step": 126483, "epoch": 1421, "lr": 5.549118135789297e-05}
{"train_loss": 0.13252557814121246, "global_step": 126484, "epoch": 1421, "lr": 5.5490605127694595e-05}
{"train_loss": 0.12579526007175446, "global_step": 126485, "epoch": 1421, "lr": 5.549002889675807e-05}
{"train_loss": 0.14881782233715057, "global_step": 126486, "epoch": 1421, "lr": 5.5489452665083474e-05}
{"train_loss": 0.23262570798397064, "global_step": 126487, "epoch": 1421, "lr": 5.548887643267088e-05}
{"train_loss": 0.1474871039390564, "global_step": 126488, "epoch": 1421, "lr": 5.548830019952038e-05}
{"train_loss": 0.09674092382192612, "global_step": 126489, "epoch": 1421, "lr": 5.5487723965632044e-05}
{"train_loss": 0.15980492532253265, "global_step": 126490, "epoch": 1421, "lr": 5.548714773100595e-05}
{"train_loss": 0.1786171793937683, "global_step": 126491, "epoch": 1421, "lr": 5.548657149564218e-05}
{"train_loss": 0.12587197124958038, "global_step": 126492, "epoch": 1421, "lr": 5.54859952595408e-05}
{"train_loss": 0.17329762876033783, "global_step": 126493, "epoch": 1421, "lr": 5.54854190227019e-05}
{"train_loss": 0.11169665306806564, "global_step": 126494, "epoch": 1421, "lr": 5.548484278512555e-05}
{"train_loss": 0.11941179633140564, "global_step": 126495, "epoch": 1421, "lr": 5.548426654681182e-05}
{"train_loss": 0.1288134753704071, "global_step": 126496, "epoch": 1421, "lr": 5.548369030776081e-05}
{"train_loss": 0.17828066647052765, "global_step": 126497, "epoch": 1421, "lr": 5.5483114067972574e-05}
{"train_loss": 0.1287337988615036, "global_step": 126498, "epoch": 1421, "lr": 5.548253782744721e-05}
{"train_loss": 0.11131730675697327, "global_step": 126499, "epoch": 1421, "lr": 5.5481961586184796e-05}
{"train_loss": 0.1695496141910553, "global_step": 126500, "epoch": 1421, "lr": 5.548138534418538e-05}
{"train_loss": 0.1528678834438324, "global_step": 126501, "epoch": 1421, "lr": 5.548080910144906e-05}
{"train_loss": 0.13895094394683838, "global_step": 126502, "epoch": 1421, "lr": 5.548023285797592e-05}
{"train_loss": 0.09935690462589264, "global_step": 126503, "epoch": 1421, "lr": 5.547965661376604e-05}
{"train_loss": 0.05188819020986557, "global_step": 126504, "epoch": 1421, "lr": 5.547908036881948e-05}
{"train_loss": 0.1219639927148819, "global_step": 126505, "epoch": 1421, "lr": 5.547850412313631e-05}
{"train_loss": 0.1336827427148819, "global_step": 126506, "epoch": 1421, "lr": 5.547792787671664e-05}
{"train_loss": 0.1225847527384758, "global_step": 126507, "epoch": 1421, "lr": 5.5477351629560515e-05}
{"train_loss": 0.0655522271990776, "global_step": 126508, "epoch": 1421, "lr": 5.5476775381668034e-05}
{"train_loss": 0.12572060525417328, "global_step": 126509, "epoch": 1421, "lr": 5.5476199133039264e-05}
{"train_loss": 0.1358749121427536, "global_step": 126510, "epoch": 1421, "lr": 5.5475622883674286e-05}
{"train_loss": 0.13122667372226715, "global_step": 126511, "epoch": 1421, "lr": 5.547504663357318e-05}
{"train_loss": 0.10622159391641617, "global_step": 126512, "epoch": 1421, "lr": 5.547447038273602e-05}
{"train_loss": 0.0683165043592453, "global_step": 126513, "epoch": 1421, "lr": 5.547389413116287e-05}
{"train_loss": 0.10406613349914551, "global_step": 126514, "epoch": 1421, "lr": 5.547331787885384e-05}
{"train_loss": 0.12322620302438736, "global_step": 126515, "epoch": 1421, "lr": 5.5472741625808975e-05}
{"train_loss": 0.05409315228462219, "global_step": 126516, "epoch": 1421, "lr": 5.547216537202838e-05}
{"train_loss": 0.10455744713544846, "global_step": 126517, "epoch": 1421, "lr": 5.5471589117512116e-05}
{"train_loss": 0.08761784434318542, "global_step": 126518, "epoch": 1421, "lr": 5.5471012862260266e-05}
{"train_loss": 0.11741682887077332, "global_step": 126519, "epoch": 1421, "lr": 5.5470436606272894e-05}
{"train_loss": 0.12030866742134094, "global_step": 126520, "epoch": 1421, "lr": 5.5469860349550094e-05}
{"train_loss": 0.1307055652141571, "global_step": 126521, "epoch": 1421, "lr": 5.546928409209195e-05}
{"train_loss": 0.10446123033761978, "global_step": 126522, "epoch": 1421, "lr": 5.54687078338985e-05}
{"train_loss": 0.13467811048030853, "global_step": 126523, "epoch": 1421, "lr": 5.5468131574969885e-05}
{"train_loss": 0.15119311213493347, "global_step": 126524, "epoch": 1421, "lr": 5.546755531530613e-05}
{"train_loss": 0.13289035856723785, "global_step": 126525, "epoch": 1421, "lr": 5.546697905490732e-05}
{"train_loss": 0.09766410291194916, "global_step": 126526, "epoch": 1421, "lr": 5.5466402793773555e-05}
{"train_loss": 0.08237326145172119, "global_step": 126527, "epoch": 1421, "lr": 5.546582653190489e-05}
{"train_loss": 0.13911576569080353, "global_step": 126528, "epoch": 1421, "lr": 5.54652502693014e-05}
{"train_loss": 0.19612357020378113, "global_step": 126529, "epoch": 1421, "lr": 5.546467400596319e-05}
{"train_loss": 0.1530052125453949, "global_step": 126530, "epoch": 1421, "lr": 5.546409774189032e-05}
{"train_loss": 0.15375681221485138, "global_step": 126531, "epoch": 1421, "lr": 5.546352147708286e-05}
{"train_loss": 0.1341368556022644, "global_step": 126532, "epoch": 1421, "lr": 5.54629452115409e-05}
{"train_loss": 0.1263863891363144, "global_step": 126533, "epoch": 1421, "lr": 5.546236894526451e-05}
{"train_loss": 0.17754197120666504, "global_step": 126534, "epoch": 1421, "lr": 5.5461792678253775e-05}
{"train_loss": 0.09335113316774368, "global_step": 126535, "epoch": 1421, "lr": 5.546121641050878e-05}
{"train_loss": 0.12977011501789093, "global_step": 126536, "epoch": 1421, "lr": 5.5460640142029575e-05}
{"train_loss": 0.1386430412530899, "global_step": 126537, "epoch": 1421, "lr": 5.546006387281626e-05}
{"train_loss": 0.08577236533164978, "global_step": 126538, "epoch": 1421, "lr": 5.545948760286891e-05}
{"train_loss": 0.11792442202568054, "global_step": 126539, "epoch": 1421, "lr": 5.545891133218759e-05}
{"train_loss": 0.09914889186620712, "global_step": 126540, "epoch": 1421, "lr": 5.545833506077239e-05}
{"train_loss": 0.14755938947200775, "global_step": 126541, "epoch": 1421, "lr": 5.545775878862338e-05}
{"train_loss": 0.15769235789775848, "global_step": 126542, "epoch": 1421, "lr": 5.545718251574066e-05}
{"train_loss": 0.14751768112182617, "global_step": 126543, "epoch": 1421, "lr": 5.5456606242124276e-05}
{"train_loss": 0.0683993250131607, "global_step": 126544, "epoch": 1421, "lr": 5.5456029967774314e-05}
{"train_loss": 0.21633803844451904, "global_step": 126545, "epoch": 1421, "lr": 5.545545369269085e-05}
{"train_loss": 0.1724657565355301, "global_step": 126546, "epoch": 1421, "lr": 5.5454877416873974e-05}
{"train_loss": 0.1523745059967041, "global_step": 126547, "epoch": 1421, "lr": 5.545430114032375e-05}
{"train_loss": 0.1345253586769104, "global_step": 126548, "epoch": 1421, "lr": 5.545372486304027e-05}
{"train_loss": 0.10144913196563721, "global_step": 126549, "epoch": 1421, "lr": 5.5453148585023584e-05}
{"train_loss": 0.12060818076133728, "global_step": 126550, "epoch": 1421, "lr": 5.545257230627381e-05}
{"train_loss": 0.1293734759092331, "global_step": 126551, "epoch": 1421, "lr": 5.5451996026791e-05}
{"train_loss": 0.1632445901632309, "global_step": 126552, "epoch": 1421, "lr": 5.545141974657523e-05}
{"train_loss": 0.14765609800815582, "global_step": 126553, "epoch": 1421, "lr": 5.5450843465626587e-05}
{"train_loss": 0.07213035225868225, "global_step": 126554, "epoch": 1421, "lr": 5.545026718394515e-05}
{"train_loss": 0.09870102256536484, "global_step": 126555, "epoch": 1421, "lr": 5.5449690901530984e-05}
{"train_loss": 0.14220602810382843, "global_step": 126556, "epoch": 1421, "lr": 5.544911461838417e-05}
{"train_loss": 0.12823603293868932, "global_step": 126557, "epoch": 1421, "lr": 5.54485383345048e-05, "val_loss": 5.422457695007324}
{"train_loss": 0.10905937105417252, "global_step": 126558, "epoch": 1422, "lr": 5.5447962049892935e-05}
{"train_loss": 0.18602731823921204, "global_step": 126559, "epoch": 1422, "lr": 5.5447385764548655e-05}
{"train_loss": 0.09247469902038574, "global_step": 126560, "epoch": 1422, "lr": 5.5446809478472036e-05}
{"train_loss": 0.1467026025056839, "global_step": 126561, "epoch": 1422, "lr": 5.5446233191663174e-05}
{"train_loss": 0.16080208122730255, "global_step": 126562, "epoch": 1422, "lr": 5.544565690412213e-05}
{"train_loss": 0.08648818731307983, "global_step": 126563, "epoch": 1422, "lr": 5.544508061584898e-05}
{"train_loss": 0.10077040642499924, "global_step": 126564, "epoch": 1422, "lr": 5.54445043268438e-05}
{"train_loss": 0.10970243811607361, "global_step": 126565, "epoch": 1422, "lr": 5.544392803710667e-05}
{"train_loss": 0.11651390045881271, "global_step": 126566, "epoch": 1422, "lr": 5.544335174663769e-05}
{"train_loss": 0.12522760033607483, "global_step": 126567, "epoch": 1422, "lr": 5.544277545543689e-05}
{"train_loss": 0.09216385334730148, "global_step": 126568, "epoch": 1422, "lr": 5.5442199163504396e-05}
{"train_loss": 0.08272462338209152, "global_step": 126569, "epoch": 1422, "lr": 5.5441622870840246e-05}
{"train_loss": 0.17213329672813416, "global_step": 126570, "epoch": 1422, "lr": 5.544104657744456e-05}
{"train_loss": 0.16368351876735687, "global_step": 126571, "epoch": 1422, "lr": 5.544047028331737e-05}
{"train_loss": 0.20964565873146057, "global_step": 126572, "epoch": 1422, "lr": 5.543989398845879e-05}
{"train_loss": 0.1224842295050621, "global_step": 126573, "epoch": 1422, "lr": 5.5439317692868874e-05}
{"train_loss": 0.09045093506574631, "global_step": 126574, "epoch": 1422, "lr": 5.543874139654771e-05}
{"train_loss": 0.16715937852859497, "global_step": 126575, "epoch": 1422, "lr": 5.5438165099495375e-05}
{"train_loss": 0.07127537578344345, "global_step": 126576, "epoch": 1422, "lr": 5.543758880171195e-05}
{"train_loss": 0.11347030848264694, "global_step": 126577, "epoch": 1422, "lr": 5.5437012503197504e-05}
{"train_loss": 0.16290844976902008, "global_step": 126578, "epoch": 1422, "lr": 5.543643620395211e-05}
{"train_loss": 0.037601519376039505, "global_step": 126579, "epoch": 1422, "lr": 5.543585990397586e-05}
{"train_loss": 0.10314509272575378, "global_step": 126580, "epoch": 1422, "lr": 5.5435283603268826e-05}
{"train_loss": 0.14818458259105682, "global_step": 126581, "epoch": 1422, "lr": 5.54347073018311e-05}
{"train_loss": 0.1323094666004181, "global_step": 126582, "epoch": 1422, "lr": 5.543413099966273e-05}
{"train_loss": 0.0822865217924118, "global_step": 126583, "epoch": 1422, "lr": 5.54335546967638e-05}
{"train_loss": 0.09984999895095825, "global_step": 126584, "epoch": 1422, "lr": 5.5432978393134394e-05}
{"train_loss": 0.09658017754554749, "global_step": 126585, "epoch": 1422, "lr": 5.543240208877458e-05}
{"train_loss": 0.06584358215332031, "global_step": 126586, "epoch": 1422, "lr": 5.5431825783684466e-05}
{"train_loss": 0.17375457286834717, "global_step": 126587, "epoch": 1422, "lr": 5.543124947786411e-05}
{"train_loss": 0.12578323483467102, "global_step": 126588, "epoch": 1422, "lr": 5.543067317131358e-05}
{"train_loss": 0.11604882776737213, "global_step": 126589, "epoch": 1422, "lr": 5.5430096864032965e-05}
{"train_loss": 0.1353611797094345, "global_step": 126590, "epoch": 1422, "lr": 5.542952055602234e-05}
{"train_loss": 0.08648903667926788, "global_step": 126591, "epoch": 1422, "lr": 5.5428944247281776e-05}
{"train_loss": 0.10074920207262039, "global_step": 126592, "epoch": 1422, "lr": 5.542836793781137e-05}
{"train_loss": 0.10506259649991989, "global_step": 126593, "epoch": 1422, "lr": 5.5427791627611183e-05}
{"train_loss": 0.15873417258262634, "global_step": 126594, "epoch": 1422, "lr": 5.542721531668129e-05}
{"train_loss": 0.11862508952617645, "global_step": 126595, "epoch": 1422, "lr": 5.5426639005021776e-05}
{"train_loss": 0.04954202473163605, "global_step": 126596, "epoch": 1422, "lr": 5.542606269263271e-05}
{"train_loss": 0.1326836347579956, "global_step": 126597, "epoch": 1422, "lr": 5.542548637951418e-05}
{"train_loss": 0.13706085085868835, "global_step": 126598, "epoch": 1422, "lr": 5.542491006566627e-05}
{"train_loss": 0.1569429486989975, "global_step": 126599, "epoch": 1422, "lr": 5.542433375108903e-05}
{"train_loss": 0.12279672920703888, "global_step": 126600, "epoch": 1422, "lr": 5.542375743578257e-05}
{"train_loss": 0.11920825392007828, "global_step": 126601, "epoch": 1422, "lr": 5.542318111974695e-05}
{"train_loss": 0.12699362635612488, "global_step": 126602, "epoch": 1422, "lr": 5.5422604802982244e-05}
{"train_loss": 0.2046814113855362, "global_step": 126603, "epoch": 1422, "lr": 5.5422028485488544e-05}
{"train_loss": 0.0790473073720932, "global_step": 126604, "epoch": 1422, "lr": 5.5421452167265886e-05}
{"train_loss": 0.15313191711902618, "global_step": 126605, "epoch": 1422, "lr": 5.5420875848314415e-05}
{"train_loss": 0.19553999602794647, "global_step": 126606, "epoch": 1422, "lr": 5.5420299528634156e-05}
{"train_loss": 0.19114740192890167, "global_step": 126607, "epoch": 1422, "lr": 5.5419723208225205e-05}
{"train_loss": 0.15874990820884705, "global_step": 126608, "epoch": 1422, "lr": 5.541914688708765e-05}
{"train_loss": 0.11966101825237274, "global_step": 126609, "epoch": 1422, "lr": 5.541857056522155e-05}
{"train_loss": 0.13446541130542755, "global_step": 126610, "epoch": 1422, "lr": 5.541799424262698e-05}
{"train_loss": 0.12197338044643402, "global_step": 126611, "epoch": 1422, "lr": 5.541741791930404e-05}
{"train_loss": 0.11673729121685028, "global_step": 126612, "epoch": 1422, "lr": 5.541684159525279e-05}
{"train_loss": 0.1517428606748581, "global_step": 126613, "epoch": 1422, "lr": 5.541626527047331e-05}
{"train_loss": 0.12345190346240997, "global_step": 126614, "epoch": 1422, "lr": 5.541568894496568e-05}
{"train_loss": 0.1038941740989685, "global_step": 126615, "epoch": 1422, "lr": 5.541511261872998e-05}
{"train_loss": 0.1814688742160797, "global_step": 126616, "epoch": 1422, "lr": 5.541453629176629e-05}
{"train_loss": 0.13335087895393372, "global_step": 126617, "epoch": 1422, "lr": 5.5413959964074655e-05}
{"train_loss": 0.1504238396883011, "global_step": 126618, "epoch": 1422, "lr": 5.5413383635655205e-05}
{"train_loss": 0.15748636424541473, "global_step": 126619, "epoch": 1422, "lr": 5.5412807306508e-05}
{"train_loss": 0.12719854712486267, "global_step": 126620, "epoch": 1422, "lr": 5.541223097663308e-05}
{"train_loss": 0.10727836191654205, "global_step": 126621, "epoch": 1422, "lr": 5.541165464603056e-05}
{"train_loss": 0.17881928384304047, "global_step": 126622, "epoch": 1422, "lr": 5.541107831470052e-05}
{"train_loss": 0.1303326040506363, "global_step": 126623, "epoch": 1422, "lr": 5.5410501982643014e-05}
{"train_loss": 0.11943218111991882, "global_step": 126624, "epoch": 1422, "lr": 5.540992564985813e-05}
{"train_loss": 0.109943687915802, "global_step": 126625, "epoch": 1422, "lr": 5.540934931634596e-05}
{"train_loss": 0.12557412683963776, "global_step": 126626, "epoch": 1422, "lr": 5.540877298210656e-05}
{"train_loss": 0.10253101587295532, "global_step": 126627, "epoch": 1422, "lr": 5.540819664714002e-05}
{"train_loss": 0.11479447036981583, "global_step": 126628, "epoch": 1422, "lr": 5.5407620311446414e-05}
{"train_loss": 0.1442442387342453, "global_step": 126629, "epoch": 1422, "lr": 5.5407043975025817e-05}
{"train_loss": 0.10936420410871506, "global_step": 126630, "epoch": 1422, "lr": 5.540646763787831e-05}
{"train_loss": 0.09826091676950455, "global_step": 126631, "epoch": 1422, "lr": 5.540589130000398e-05}
{"train_loss": 0.153241366147995, "global_step": 126632, "epoch": 1422, "lr": 5.5405314961402874e-05}
{"train_loss": 0.18141856789588928, "global_step": 126633, "epoch": 1422, "lr": 5.5404738622075104e-05}
{"train_loss": 0.2031038999557495, "global_step": 126634, "epoch": 1422, "lr": 5.540416228202072e-05}
{"train_loss": 0.08137861639261246, "global_step": 126635, "epoch": 1422, "lr": 5.5403585941239835e-05}
{"train_loss": 0.11870388686656952, "global_step": 126636, "epoch": 1422, "lr": 5.540300959973248e-05}
{"train_loss": 0.094245545566082, "global_step": 126637, "epoch": 1422, "lr": 5.540243325749876e-05}
{"train_loss": 0.14336030185222626, "global_step": 126638, "epoch": 1422, "lr": 5.5401856914538765e-05}
{"train_loss": 0.21572259068489075, "global_step": 126639, "epoch": 1422, "lr": 5.5401280570852555e-05}
{"train_loss": 0.1485803723335266, "global_step": 126640, "epoch": 1422, "lr": 5.540070422644018e-05}
{"train_loss": 0.11490508913993835, "global_step": 126641, "epoch": 1422, "lr": 5.5400127881301765e-05}
{"train_loss": 0.14208045601844788, "global_step": 126642, "epoch": 1422, "lr": 5.539955153543738e-05}
{"train_loss": 0.09823846817016602, "global_step": 126643, "epoch": 1422, "lr": 5.539897518884707e-05}
{"train_loss": 0.19488130509853363, "global_step": 126644, "epoch": 1422, "lr": 5.539839884153095e-05}
{"train_loss": 0.10631559044122696, "global_step": 126645, "epoch": 1422, "lr": 5.539782249348907e-05}
{"train_loss": 0.12886526548628058, "global_step": 126646, "epoch": 1422, "lr": 5.5397246144721527e-05, "val_loss": 5.326028347015381}
{"train_loss": 0.10479690879583359, "global_step": 126647, "epoch": 1423, "lr": 5.5396669795228385e-05}
{"train_loss": 0.18838615715503693, "global_step": 126648, "epoch": 1423, "lr": 5.539609344500972e-05}
{"train_loss": 0.18490734696388245, "global_step": 126649, "epoch": 1423, "lr": 5.539551709406563e-05}
{"train_loss": 0.17155100405216217, "global_step": 126650, "epoch": 1423, "lr": 5.539494074239617e-05}
{"train_loss": 0.11465068906545639, "global_step": 126651, "epoch": 1423, "lr": 5.539436439000143e-05}
{"train_loss": 0.1421259492635727, "global_step": 126652, "epoch": 1423, "lr": 5.539378803688148e-05}
{"train_loss": 0.06688147783279419, "global_step": 126653, "epoch": 1423, "lr": 5.5393211683036405e-05}
{"train_loss": 0.1529237926006317, "global_step": 126654, "epoch": 1423, "lr": 5.539263532846628e-05}
{"train_loss": 0.13303032517433167, "global_step": 126655, "epoch": 1423, "lr": 5.539205897317117e-05}
{"train_loss": 0.1995304375886917, "global_step": 126656, "epoch": 1423, "lr": 5.539148261715117e-05}
{"train_loss": 0.12784795463085175, "global_step": 126657, "epoch": 1423, "lr": 5.539090626040635e-05}
{"train_loss": 0.11890415847301483, "global_step": 126658, "epoch": 1423, "lr": 5.5390329902936786e-05}
{"train_loss": 0.12099672853946686, "global_step": 126659, "epoch": 1423, "lr": 5.5389753544742576e-05}
{"train_loss": 0.13537409901618958, "global_step": 126660, "epoch": 1423, "lr": 5.538917718582376e-05}
{"train_loss": 0.18569612503051758, "global_step": 126661, "epoch": 1423, "lr": 5.538860082618044e-05}
{"train_loss": 0.14750954508781433, "global_step": 126662, "epoch": 1423, "lr": 5.538802446581268e-05}
{"train_loss": 0.18456701934337616, "global_step": 126663, "epoch": 1423, "lr": 5.5387448104720576e-05}
{"train_loss": 0.10704458504915237, "global_step": 126664, "epoch": 1423, "lr": 5.538687174290419e-05}
{"train_loss": 0.12006903439760208, "global_step": 126665, "epoch": 1423, "lr": 5.538629538036361e-05}
{"train_loss": 0.08611655235290527, "global_step": 126666, "epoch": 1423, "lr": 5.5385719017098904e-05}
{"train_loss": 0.14507082104682922, "global_step": 126667, "epoch": 1423, "lr": 5.5385142653110156e-05}
{"train_loss": 0.15138369798660278, "global_step": 126668, "epoch": 1423, "lr": 5.5384566288397444e-05}
{"train_loss": 0.06619734317064285, "global_step": 126669, "epoch": 1423, "lr": 5.538398992296083e-05}
{"train_loss": 0.10190986096858978, "global_step": 126670, "epoch": 1423, "lr": 5.5383413556800426e-05}
{"train_loss": 0.03384128212928772, "global_step": 126671, "epoch": 1423, "lr": 5.538283718991627e-05}
{"train_loss": 0.14123642444610596, "global_step": 126672, "epoch": 1423, "lr": 5.5382260822308466e-05}
{"train_loss": 0.11267868429422379, "global_step": 126673, "epoch": 1423, "lr": 5.538168445397708e-05}
{"train_loss": 0.18356220424175262, "global_step": 126674, "epoch": 1423, "lr": 5.538110808492218e-05}
{"train_loss": 0.10185439139604568, "global_step": 126675, "epoch": 1423, "lr": 5.538053171514388e-05}
{"train_loss": 0.08718781918287277, "global_step": 126676, "epoch": 1423, "lr": 5.537995534464221e-05}
{"train_loss": 0.12443695217370987, "global_step": 126677, "epoch": 1423, "lr": 5.5379378973417283e-05}
{"train_loss": 0.09900706261396408, "global_step": 126678, "epoch": 1423, "lr": 5.537880260146916e-05}
{"train_loss": 0.13389013707637787, "global_step": 126679, "epoch": 1423, "lr": 5.5378226228797926e-05}
{"train_loss": 0.10757621377706528, "global_step": 126680, "epoch": 1423, "lr": 5.5377649855403656e-05}
{"train_loss": 0.0863700807094574, "global_step": 126681, "epoch": 1423, "lr": 5.5377073481286426e-05}
{"train_loss": 0.11411960422992706, "global_step": 126682, "epoch": 1423, "lr": 5.5376497106446304e-05}
{"train_loss": 0.16199859976768494, "global_step": 126683, "epoch": 1423, "lr": 5.537592073088339e-05}
{"train_loss": 0.05003173649311066, "global_step": 126684, "epoch": 1423, "lr": 5.5375344354597745e-05}
{"train_loss": 0.16936898231506348, "global_step": 126685, "epoch": 1423, "lr": 5.537476797758945e-05}
{"train_loss": 0.10765755921602249, "global_step": 126686, "epoch": 1423, "lr": 5.537419159985858e-05}
{"train_loss": 0.1134408488869667, "global_step": 126687, "epoch": 1423, "lr": 5.537361522140522e-05}
{"train_loss": 0.08652178943157196, "global_step": 126688, "epoch": 1423, "lr": 5.537303884222944e-05}
{"train_loss": 0.09360497444868088, "global_step": 126689, "epoch": 1423, "lr": 5.537246246233132e-05}
{"train_loss": 0.08946536481380463, "global_step": 126690, "epoch": 1423, "lr": 5.537188608171093e-05}
{"train_loss": 0.14359790086746216, "global_step": 126691, "epoch": 1423, "lr": 5.537130970036837e-05}
{"train_loss": 0.16187112033367157, "global_step": 126692, "epoch": 1423, "lr": 5.5370733318303704e-05}
{"train_loss": 0.18986783921718597, "global_step": 126693, "epoch": 1423, "lr": 5.537015693551699e-05}
{"train_loss": 0.18636605143547058, "global_step": 126694, "epoch": 1423, "lr": 5.536958055200834e-05}
{"train_loss": 0.16476525366306305, "global_step": 126695, "epoch": 1423, "lr": 5.5369004167777815e-05}
{"train_loss": 0.13231557607650757, "global_step": 126696, "epoch": 1423, "lr": 5.536842778282548e-05}
{"train_loss": 0.15292364358901978, "global_step": 126697, "epoch": 1423, "lr": 5.5367851397151436e-05}
{"train_loss": 0.16428028047084808, "global_step": 126698, "epoch": 1423, "lr": 5.5367275010755746e-05}
{"train_loss": 0.15874634683132172, "global_step": 126699, "epoch": 1423, "lr": 5.5366698623638505e-05}
{"train_loss": 0.09309166669845581, "global_step": 126700, "epoch": 1423, "lr": 5.5366122235799754e-05}
{"train_loss": 0.09662680327892303, "global_step": 126701, "epoch": 1423, "lr": 5.536554584723961e-05}
{"train_loss": 0.1423734873533249, "global_step": 126702, "epoch": 1423, "lr": 5.5364969457958125e-05}
{"train_loss": 0.1882857382297516, "global_step": 126703, "epoch": 1423, "lr": 5.536439306795539e-05}
{"train_loss": 0.19995370507240295, "global_step": 126704, "epoch": 1423, "lr": 5.5363816677231475e-05}
{"train_loss": 0.11751142889261246, "global_step": 126705, "epoch": 1423, "lr": 5.5363240285786465e-05}
{"train_loss": 0.1084861233830452, "global_step": 126706, "epoch": 1423, "lr": 5.536266389362043e-05}
{"train_loss": 0.09773053973913193, "global_step": 126707, "epoch": 1423, "lr": 5.536208750073345e-05}
{"train_loss": 0.14996059238910675, "global_step": 126708, "epoch": 1423, "lr": 5.53615111071256e-05}
{"train_loss": 0.11279958486557007, "global_step": 126709, "epoch": 1423, "lr": 5.536093471279696e-05}
{"train_loss": 0.11813076585531235, "global_step": 126710, "epoch": 1423, "lr": 5.536035831774761e-05}
{"train_loss": 0.14181040227413177, "global_step": 126711, "epoch": 1423, "lr": 5.535978192197763e-05}
{"train_loss": 0.10208313167095184, "global_step": 126712, "epoch": 1423, "lr": 5.535920552548708e-05}
{"train_loss": 0.09996378421783447, "global_step": 126713, "epoch": 1423, "lr": 5.535862912827606e-05}
{"train_loss": 0.1477334350347519, "global_step": 126714, "epoch": 1423, "lr": 5.5358052730344635e-05}
{"train_loss": 0.19043318927288055, "global_step": 126715, "epoch": 1423, "lr": 5.535747633169288e-05}
{"train_loss": 0.09376087784767151, "global_step": 126716, "epoch": 1423, "lr": 5.5356899932320894e-05}
{"train_loss": 0.18219494819641113, "global_step": 126717, "epoch": 1423, "lr": 5.535632353222873e-05}
{"train_loss": 0.11068272590637207, "global_step": 126718, "epoch": 1423, "lr": 5.5355747131416455e-05}
{"train_loss": 0.09850342571735382, "global_step": 126719, "epoch": 1423, "lr": 5.535517072988419e-05}
{"train_loss": 0.14129233360290527, "global_step": 126720, "epoch": 1423, "lr": 5.535459432763198e-05}
{"train_loss": 0.10740876197814941, "global_step": 126721, "epoch": 1423, "lr": 5.5354017924659904e-05}
{"train_loss": 0.13968758285045624, "global_step": 126722, "epoch": 1423, "lr": 5.535344152096805e-05}
{"train_loss": 0.14224986732006073, "global_step": 126723, "epoch": 1423, "lr": 5.53528651165565e-05}
{"train_loss": 0.1856251358985901, "global_step": 126724, "epoch": 1423, "lr": 5.5352288711425306e-05}
{"train_loss": 0.12141461670398712, "global_step": 126725, "epoch": 1423, "lr": 5.535171230557456e-05}
{"train_loss": 0.11507052928209305, "global_step": 126726, "epoch": 1423, "lr": 5.5351135899004356e-05}
{"train_loss": 0.11481114476919174, "global_step": 126727, "epoch": 1423, "lr": 5.535055949171476e-05}
{"train_loss": 0.10963356494903564, "global_step": 126728, "epoch": 1423, "lr": 5.534998308370584e-05}
{"train_loss": 0.15930795669555664, "global_step": 126729, "epoch": 1423, "lr": 5.534940667497768e-05}
{"train_loss": 0.09297527372837067, "global_step": 126730, "epoch": 1423, "lr": 5.534883026553035e-05}
{"train_loss": 0.14580564200878143, "global_step": 126731, "epoch": 1423, "lr": 5.534825385536394e-05}
{"train_loss": 0.06597745418548584, "global_step": 126732, "epoch": 1423, "lr": 5.5347677444478515e-05}
{"train_loss": 0.07595294713973999, "global_step": 126733, "epoch": 1423, "lr": 5.534710103287417e-05}
{"train_loss": 0.0968102365732193, "global_step": 126734, "epoch": 1423, "lr": 5.5346524620550976e-05}
{"train_loss": 0.1282404518864128, "global_step": 126735, "epoch": 1423, "lr": 5.5345948207509004e-05, "val_loss": 5.599087715148926}
{"train_loss": 0.11155657470226288, "global_step": 126736, "epoch": 1424, "lr": 5.534537179374832e-05}
{"train_loss": 0.09452175348997116, "global_step": 126737, "epoch": 1424, "lr": 5.534479537926903e-05}
{"train_loss": 0.13435551524162292, "global_step": 126738, "epoch": 1424, "lr": 5.53442189640712e-05}
{"train_loss": 0.09451611340045929, "global_step": 126739, "epoch": 1424, "lr": 5.53436425481549e-05}
{"train_loss": 0.06036390736699104, "global_step": 126740, "epoch": 1424, "lr": 5.534306613152022e-05}
{"train_loss": 0.09570189565420151, "global_step": 126741, "epoch": 1424, "lr": 5.534248971416722e-05}
{"train_loss": 0.11295238137245178, "global_step": 126742, "epoch": 1424, "lr": 5.5341913296095995e-05}
{"train_loss": 0.08182597905397415, "global_step": 126743, "epoch": 1424, "lr": 5.5341336877306606e-05}
{"train_loss": 0.13312865793704987, "global_step": 126744, "epoch": 1424, "lr": 5.5340760457799144e-05}
{"train_loss": 0.1400294303894043, "global_step": 126745, "epoch": 1424, "lr": 5.5340184037573676e-05}
{"train_loss": 0.1708584874868393, "global_step": 126746, "epoch": 1424, "lr": 5.5339607616630296e-05}
{"train_loss": 0.058521904051303864, "global_step": 126747, "epoch": 1424, "lr": 5.533903119496906e-05}
{"train_loss": 0.11642578989267349, "global_step": 126748, "epoch": 1424, "lr": 5.533845477259007e-05}
{"train_loss": 0.15447744727134705, "global_step": 126749, "epoch": 1424, "lr": 5.533787834949337e-05}
{"train_loss": 0.1017792746424675, "global_step": 126750, "epoch": 1424, "lr": 5.5337301925679074e-05}
{"train_loss": 0.0944661796092987, "global_step": 126751, "epoch": 1424, "lr": 5.5336725501147236e-05}
{"train_loss": 0.09907262027263641, "global_step": 126752, "epoch": 1424, "lr": 5.5336149075897934e-05}
{"train_loss": 0.16167320311069489, "global_step": 126753, "epoch": 1424, "lr": 5.533557264993127e-05}
{"train_loss": 0.11421039700508118, "global_step": 126754, "epoch": 1424, "lr": 5.533499622324729e-05}
{"train_loss": 0.10349860042333603, "global_step": 126755, "epoch": 1424, "lr": 5.5334419795846084e-05}
{"train_loss": 0.09428013861179352, "global_step": 126756, "epoch": 1424, "lr": 5.533384336772773e-05}
{"train_loss": 0.1444806307554245, "global_step": 126757, "epoch": 1424, "lr": 5.5333266938892316e-05}
{"train_loss": 0.137907013297081, "global_step": 126758, "epoch": 1424, "lr": 5.5332690509339904e-05}
{"train_loss": 0.0843043252825737, "global_step": 126759, "epoch": 1424, "lr": 5.5332114079070576e-05}
{"train_loss": 0.09945701062679291, "global_step": 126760, "epoch": 1424, "lr": 5.533153764808441e-05}
{"train_loss": 0.19648170471191406, "global_step": 126761, "epoch": 1424, "lr": 5.533096121638148e-05}
{"train_loss": 0.2204980105161667, "global_step": 126762, "epoch": 1424, "lr": 5.5330384783961875e-05}
{"train_loss": 0.1529712975025177, "global_step": 126763, "epoch": 1424, "lr": 5.532980835082566e-05}
{"train_loss": 0.07771908491849899, "global_step": 126764, "epoch": 1424, "lr": 5.5329231916972915e-05}
{"train_loss": 0.0831885114312172, "global_step": 126765, "epoch": 1424, "lr": 5.532865548240372e-05}
{"train_loss": 0.16533227264881134, "global_step": 126766, "epoch": 1424, "lr": 5.5328079047118165e-05}
{"train_loss": 0.11746878176927567, "global_step": 126767, "epoch": 1424, "lr": 5.532750261111629e-05}
{"train_loss": 0.175960972905159, "global_step": 126768, "epoch": 1424, "lr": 5.532692617439821e-05}
{"train_loss": 0.09619763493537903, "global_step": 126769, "epoch": 1424, "lr": 5.5326349736964e-05}
{"train_loss": 0.16594187915325165, "global_step": 126770, "epoch": 1424, "lr": 5.532577329881372e-05}
{"train_loss": 0.05184946581721306, "global_step": 126771, "epoch": 1424, "lr": 5.532519685994746e-05}
{"train_loss": 0.1521686166524887, "global_step": 126772, "epoch": 1424, "lr": 5.532462042036528e-05}
{"train_loss": 0.1615009605884552, "global_step": 126773, "epoch": 1424, "lr": 5.5324043980067276e-05}
{"train_loss": 0.08633750677108765, "global_step": 126774, "epoch": 1424, "lr": 5.532346753905352e-05}
{"train_loss": 0.15295185148715973, "global_step": 126775, "epoch": 1424, "lr": 5.532289109732408e-05}
{"train_loss": 0.13030177354812622, "global_step": 126776, "epoch": 1424, "lr": 5.5322314654879046e-05}
{"train_loss": 0.13258473575115204, "global_step": 126777, "epoch": 1424, "lr": 5.53217382117185e-05}
{"train_loss": 0.09846068918704987, "global_step": 126778, "epoch": 1424, "lr": 5.532116176784251e-05}
{"train_loss": 0.11584093421697617, "global_step": 126779, "epoch": 1424, "lr": 5.532058532325115e-05}
{"train_loss": 0.13395272195339203, "global_step": 126780, "epoch": 1424, "lr": 5.532000887794451e-05}
{"train_loss": 0.07915196567773819, "global_step": 126781, "epoch": 1424, "lr": 5.5319432431922644e-05}
{"train_loss": 0.1289672553539276, "global_step": 126782, "epoch": 1424, "lr": 5.531885598518566e-05}
{"train_loss": 0.10239008069038391, "global_step": 126783, "epoch": 1424, "lr": 5.5318279537733606e-05}
{"train_loss": 0.1914924830198288, "global_step": 126784, "epoch": 1424, "lr": 5.531770308956659e-05}
{"train_loss": 0.1272294819355011, "global_step": 126785, "epoch": 1424, "lr": 5.5317126640684646e-05}
{"train_loss": 0.12970899045467377, "global_step": 126786, "epoch": 1424, "lr": 5.5316550191087904e-05}
{"train_loss": 0.07525462657213211, "global_step": 126787, "epoch": 1424, "lr": 5.531597374077641e-05}
{"train_loss": 0.14427736401557922, "global_step": 126788, "epoch": 1424, "lr": 5.531539728975026e-05}
{"train_loss": 0.1608773022890091, "global_step": 126789, "epoch": 1424, "lr": 5.531482083800951e-05}
{"train_loss": 0.1536794751882553, "global_step": 126790, "epoch": 1424, "lr": 5.5314244385554246e-05}
{"train_loss": 0.17510569095611572, "global_step": 126791, "epoch": 1424, "lr": 5.531366793238455e-05}
{"train_loss": 0.21871577203273773, "global_step": 126792, "epoch": 1424, "lr": 5.531309147850049e-05}
{"train_loss": 0.1589365154504776, "global_step": 126793, "epoch": 1424, "lr": 5.5312515023902156e-05}
{"train_loss": 0.13376539945602417, "global_step": 126794, "epoch": 1424, "lr": 5.5311938568589614e-05}
{"train_loss": 0.08936478197574615, "global_step": 126795, "epoch": 1424, "lr": 5.531136211256295e-05}
{"train_loss": 0.1410096138715744, "global_step": 126796, "epoch": 1424, "lr": 5.531078565582224e-05}
{"train_loss": 0.11256483942270279, "global_step": 126797, "epoch": 1424, "lr": 5.5310209198367556e-05}
{"train_loss": 0.15571998059749603, "global_step": 126798, "epoch": 1424, "lr": 5.530963274019898e-05}
{"train_loss": 0.09065496921539307, "global_step": 126799, "epoch": 1424, "lr": 5.5309056281316584e-05}
{"train_loss": 0.1305016279220581, "global_step": 126800, "epoch": 1424, "lr": 5.530847982172045e-05}
{"train_loss": 0.08278248459100723, "global_step": 126801, "epoch": 1424, "lr": 5.5307903361410664e-05}
{"train_loss": 0.11797338724136353, "global_step": 126802, "epoch": 1424, "lr": 5.530732690038729e-05}
{"train_loss": 0.12108953297138214, "global_step": 126803, "epoch": 1424, "lr": 5.53067504386504e-05}
{"train_loss": 0.10560683906078339, "global_step": 126804, "epoch": 1424, "lr": 5.530617397620009e-05}
{"train_loss": 0.13044437766075134, "global_step": 126805, "epoch": 1424, "lr": 5.530559751303643e-05}
{"train_loss": 0.1262897551059723, "global_step": 126806, "epoch": 1424, "lr": 5.530502104915951e-05}
{"train_loss": 0.14281609654426575, "global_step": 126807, "epoch": 1424, "lr": 5.530444458456938e-05}
{"train_loss": 0.11671794950962067, "global_step": 126808, "epoch": 1424, "lr": 5.530386811926615e-05}
{"train_loss": 0.15072639286518097, "global_step": 126809, "epoch": 1424, "lr": 5.5303291653249866e-05}
{"train_loss": 0.10288289934396744, "global_step": 126810, "epoch": 1424, "lr": 5.530271518652063e-05}
{"train_loss": 0.11548386514186859, "global_step": 126811, "epoch": 1424, "lr": 5.530213871907849e-05}
{"train_loss": 0.16124264895915985, "global_step": 126812, "epoch": 1424, "lr": 5.530156225092356e-05}
{"train_loss": 0.13435599207878113, "global_step": 126813, "epoch": 1424, "lr": 5.530098578205588e-05}
{"train_loss": 0.23537465929985046, "global_step": 126814, "epoch": 1424, "lr": 5.530040931247555e-05}
{"train_loss": 0.11755243688821793, "global_step": 126815, "epoch": 1424, "lr": 5.529983284218265e-05}
{"train_loss": 0.10951586812734604, "global_step": 126816, "epoch": 1424, "lr": 5.5299256371177265e-05}
{"train_loss": 0.07076682150363922, "global_step": 126817, "epoch": 1424, "lr": 5.529867989945945e-05}
{"train_loss": 0.19230015575885773, "global_step": 126818, "epoch": 1424, "lr": 5.529810342702928e-05}
{"train_loss": 0.13439063727855682, "global_step": 126819, "epoch": 1424, "lr": 5.529752695388686e-05}
{"train_loss": 0.08141325414180756, "global_step": 126820, "epoch": 1424, "lr": 5.529695048003225e-05}
{"train_loss": 0.17491082847118378, "global_step": 126821, "epoch": 1424, "lr": 5.5296374005465514e-05}
{"train_loss": 0.07301128655672073, "global_step": 126822, "epoch": 1424, "lr": 5.529579753018675e-05}
{"train_loss": 0.1295844465494156, "global_step": 126823, "epoch": 1424, "lr": 5.5295221054196044e-05}
{"train_loss": 0.12613253194964333, "global_step": 126824, "epoch": 1424, "lr": 5.529464457749345e-05, "val_loss": 5.478015422821045}
{"train_loss": 0.15392319858074188, "global_step": 126825, "epoch": 1425, "lr": 5.5294068100079066e-05}
{"train_loss": 0.1462739109992981, "global_step": 126826, "epoch": 1425, "lr": 5.529349162195295e-05}
{"train_loss": 0.1319625973701477, "global_step": 126827, "epoch": 1425, "lr": 5.5292915143115196e-05}
{"train_loss": 0.21903438866138458, "global_step": 126828, "epoch": 1425, "lr": 5.5292338663565865e-05}
{"train_loss": 0.11599630117416382, "global_step": 126829, "epoch": 1425, "lr": 5.529176218330505e-05}
{"train_loss": 0.11199876666069031, "global_step": 126830, "epoch": 1425, "lr": 5.529118570233283e-05}
{"train_loss": 0.10617385804653168, "global_step": 126831, "epoch": 1425, "lr": 5.529060922064926e-05}
{"train_loss": 0.09419272094964981, "global_step": 126832, "epoch": 1425, "lr": 5.529003273825444e-05}
{"train_loss": 0.08105047047138214, "global_step": 126833, "epoch": 1425, "lr": 5.5289456255148434e-05}
{"train_loss": 0.16858452558517456, "global_step": 126834, "epoch": 1425, "lr": 5.5288879771331323e-05}
{"train_loss": 0.10704290121793747, "global_step": 126835, "epoch": 1425, "lr": 5.5288303286803186e-05}
{"train_loss": 0.10646675527095795, "global_step": 126836, "epoch": 1425, "lr": 5.528772680156411e-05}
{"train_loss": 0.08678577095270157, "global_step": 126837, "epoch": 1425, "lr": 5.5287150315614164e-05}
{"train_loss": 0.12226077914237976, "global_step": 126838, "epoch": 1425, "lr": 5.528657382895343e-05}
{"train_loss": 0.13702715933322906, "global_step": 126839, "epoch": 1425, "lr": 5.528599734158195e-05}
{"train_loss": 0.1003219485282898, "global_step": 126840, "epoch": 1425, "lr": 5.528542085349986e-05}
{"train_loss": 0.12891273200511932, "global_step": 126841, "epoch": 1425, "lr": 5.52848443647072e-05}
{"train_loss": 0.08797896653413773, "global_step": 126842, "epoch": 1425, "lr": 5.5284267875204055e-05}
{"train_loss": 0.11600145697593689, "global_step": 126843, "epoch": 1425, "lr": 5.528369138499051e-05}
{"train_loss": 0.13742266595363617, "global_step": 126844, "epoch": 1425, "lr": 5.528311489406663e-05}
{"train_loss": 0.06322196125984192, "global_step": 126845, "epoch": 1425, "lr": 5.528253840243252e-05}
{"train_loss": 0.1298171430826187, "global_step": 126846, "epoch": 1425, "lr": 5.528196191008823e-05}
{"train_loss": 0.09866464138031006, "global_step": 126847, "epoch": 1425, "lr": 5.528138541703384e-05}
{"train_loss": 0.19564864039421082, "global_step": 126848, "epoch": 1425, "lr": 5.5280808923269426e-05}
{"train_loss": 0.06824798882007599, "global_step": 126849, "epoch": 1425, "lr": 5.5280232428795075e-05}
{"train_loss": 0.0957726314663887, "global_step": 126850, "epoch": 1425, "lr": 5.5279655933610864e-05}
{"train_loss": 0.1326349675655365, "global_step": 126851, "epoch": 1425, "lr": 5.5279079437716866e-05}
{"train_loss": 0.14550013840198517, "global_step": 126852, "epoch": 1425, "lr": 5.5278502941113156e-05}
{"train_loss": 0.12549687922000885, "global_step": 126853, "epoch": 1425, "lr": 5.527792644379982e-05}
{"train_loss": 0.09298699349164963, "global_step": 126854, "epoch": 1425, "lr": 5.5277349945776925e-05}
{"train_loss": 0.07863909006118774, "global_step": 126855, "epoch": 1425, "lr": 5.527677344704456e-05}
{"train_loss": 0.1029704213142395, "global_step": 126856, "epoch": 1425, "lr": 5.5276196947602796e-05}
{"train_loss": 0.1436670869588852, "global_step": 126857, "epoch": 1425, "lr": 5.527562044745171e-05}
{"train_loss": 0.15525905787944794, "global_step": 126858, "epoch": 1425, "lr": 5.5275043946591377e-05}
{"train_loss": 0.13242867588996887, "global_step": 126859, "epoch": 1425, "lr": 5.527446744502189e-05}
{"train_loss": 0.14891035854816437, "global_step": 126860, "epoch": 1425, "lr": 5.527389094274331e-05}
{"train_loss": 0.13303205370903015, "global_step": 126861, "epoch": 1425, "lr": 5.527331443975572e-05}
{"train_loss": 0.12201928347349167, "global_step": 126862, "epoch": 1425, "lr": 5.527273793605919e-05}
{"train_loss": 0.11656918376684189, "global_step": 126863, "epoch": 1425, "lr": 5.527216143165381e-05}
{"train_loss": 0.13464978337287903, "global_step": 126864, "epoch": 1425, "lr": 5.5271584926539665e-05}
{"train_loss": 0.14873255789279938, "global_step": 126865, "epoch": 1425, "lr": 5.527100842071681e-05}
{"train_loss": 0.14519210159778595, "global_step": 126866, "epoch": 1425, "lr": 5.5270431914185337e-05}
{"train_loss": 0.141714945435524, "global_step": 126867, "epoch": 1425, "lr": 5.526985540694531e-05}
{"train_loss": 0.20549249649047852, "global_step": 126868, "epoch": 1425, "lr": 5.526927889899682e-05}
{"train_loss": 0.14726343750953674, "global_step": 126869, "epoch": 1425, "lr": 5.5268702390339935e-05}
{"train_loss": 0.12564493715763092, "global_step": 126870, "epoch": 1425, "lr": 5.526812588097474e-05}
{"train_loss": 0.10992088168859482, "global_step": 126871, "epoch": 1425, "lr": 5.526754937090131e-05}
{"train_loss": 0.08900830894708633, "global_step": 126872, "epoch": 1425, "lr": 5.526697286011972e-05}
{"train_loss": 0.07287729531526566, "global_step": 126873, "epoch": 1425, "lr": 5.526639634863006e-05}
{"train_loss": 0.10975828766822815, "global_step": 126874, "epoch": 1425, "lr": 5.5265819836432385e-05}
{"train_loss": 0.10711919516324997, "global_step": 126875, "epoch": 1425, "lr": 5.526524332352678e-05}
{"train_loss": 0.13178595900535583, "global_step": 126876, "epoch": 1425, "lr": 5.526466680991332e-05}
{"train_loss": 0.0738687664270401, "global_step": 126877, "epoch": 1425, "lr": 5.526409029559212e-05}
{"train_loss": 0.15270712971687317, "global_step": 126878, "epoch": 1425, "lr": 5.52635137805632e-05}
{"train_loss": 0.14241306483745575, "global_step": 126879, "epoch": 1425, "lr": 5.5262937264826684e-05}
{"train_loss": 0.1352677047252655, "global_step": 126880, "epoch": 1425, "lr": 5.5262360748382623e-05}
{"train_loss": 0.12234484404325485, "global_step": 126881, "epoch": 1425, "lr": 5.5261784231231094e-05}
{"train_loss": 0.18389464914798737, "global_step": 126882, "epoch": 1425, "lr": 5.5261207713372196e-05}
{"train_loss": 0.12247344851493835, "global_step": 126883, "epoch": 1425, "lr": 5.526063119480599e-05}
{"train_loss": 0.10802574455738068, "global_step": 126884, "epoch": 1425, "lr": 5.5260054675532544e-05}
{"train_loss": 0.13101285696029663, "global_step": 126885, "epoch": 1425, "lr": 5.525947815555197e-05}
{"train_loss": 0.10399332642555237, "global_step": 126886, "epoch": 1425, "lr": 5.5258901634864316e-05}
{"train_loss": 0.12788939476013184, "global_step": 126887, "epoch": 1425, "lr": 5.525832511346965e-05}
{"train_loss": 0.16010300815105438, "global_step": 126888, "epoch": 1425, "lr": 5.525774859136808e-05}
{"train_loss": 0.08363182842731476, "global_step": 126889, "epoch": 1425, "lr": 5.5257172068559674e-05}
{"train_loss": 0.10215774923563004, "global_step": 126890, "epoch": 1425, "lr": 5.52565955450445e-05}
{"train_loss": 0.18366998434066772, "global_step": 126891, "epoch": 1425, "lr": 5.525601902082264e-05}
{"train_loss": 0.15596599876880646, "global_step": 126892, "epoch": 1425, "lr": 5.525544249589417e-05}
{"train_loss": 0.13831746578216553, "global_step": 126893, "epoch": 1425, "lr": 5.5254865970259174e-05}
{"train_loss": 0.13431470096111298, "global_step": 126894, "epoch": 1425, "lr": 5.52542894439177e-05}
{"train_loss": 0.11806298047304153, "global_step": 126895, "epoch": 1425, "lr": 5.525371291686988e-05}
{"train_loss": 0.13080140948295593, "global_step": 126896, "epoch": 1425, "lr": 5.5253136389115755e-05}
{"train_loss": 0.1388997882604599, "global_step": 126897, "epoch": 1425, "lr": 5.525255986065541e-05}
{"train_loss": 0.11256171017885208, "global_step": 126898, "epoch": 1425, "lr": 5.525198333148892e-05}
{"train_loss": 0.12904663383960724, "global_step": 126899, "epoch": 1425, "lr": 5.525140680161637e-05}
{"train_loss": 0.09925422072410583, "global_step": 126900, "epoch": 1425, "lr": 5.5250830271037836e-05}
{"train_loss": 0.10030673444271088, "global_step": 126901, "epoch": 1425, "lr": 5.525025373975338e-05}
{"train_loss": 0.18305356800556183, "global_step": 126902, "epoch": 1425, "lr": 5.52496772077631e-05}
{"train_loss": 0.12692585587501526, "global_step": 126903, "epoch": 1425, "lr": 5.524910067506706e-05}
{"train_loss": 0.1986478567123413, "global_step": 126904, "epoch": 1425, "lr": 5.5248524141665356e-05}
{"train_loss": 0.102199986577034, "global_step": 126905, "epoch": 1425, "lr": 5.5247947607558036e-05}
{"train_loss": 0.17325972020626068, "global_step": 126906, "epoch": 1425, "lr": 5.5247371072745205e-05}
{"train_loss": 0.18056154251098633, "global_step": 126907, "epoch": 1425, "lr": 5.524679453722692e-05}
{"train_loss": 0.09197260439395905, "global_step": 126908, "epoch": 1425, "lr": 5.524621800100327e-05}
{"train_loss": 0.13253796100616455, "global_step": 126909, "epoch": 1425, "lr": 5.5245641464074326e-05}
{"train_loss": 0.10759172588586807, "global_step": 126910, "epoch": 1425, "lr": 5.524506492644017e-05}
{"train_loss": 0.20421329140663147, "global_step": 126911, "epoch": 1425, "lr": 5.524448838810088e-05}
{"train_loss": 0.14516030251979828, "global_step": 126912, "epoch": 1425, "lr": 5.524391184905653e-05}
{"train_loss": 0.12787716952937372, "global_step": 126913, "epoch": 1425, "lr": 5.5243335309307195e-05, "val_loss": 5.6996026039123535, "train_action_mse_error": 22.31241226196289}
{"train_loss": 0.11039943248033524, "global_step": 126914, "epoch": 1426, "lr": 5.5242758768852976e-05}
{"train_loss": 0.26212066411972046, "global_step": 126915, "epoch": 1426, "lr": 5.524218222769392e-05}
{"train_loss": 0.2202955186367035, "global_step": 126916, "epoch": 1426, "lr": 5.524160568583012e-05}
{"train_loss": 0.1677897572517395, "global_step": 126917, "epoch": 1426, "lr": 5.5241029143261645e-05}
{"train_loss": 0.08397891372442245, "global_step": 126918, "epoch": 1426, "lr": 5.524045259998858e-05}
{"train_loss": 0.12690600752830505, "global_step": 126919, "epoch": 1426, "lr": 5.523987605601101e-05}
{"train_loss": 0.12292949855327606, "global_step": 126920, "epoch": 1426, "lr": 5.523929951132899e-05}
{"train_loss": 0.15369288623332977, "global_step": 126921, "epoch": 1426, "lr": 5.5238722965942614e-05}
{"train_loss": 0.1477365642786026, "global_step": 126922, "epoch": 1426, "lr": 5.523814641985195e-05}
{"train_loss": 0.18103930354118347, "global_step": 126923, "epoch": 1426, "lr": 5.5237569873057096e-05}
{"train_loss": 0.10828467458486557, "global_step": 126924, "epoch": 1426, "lr": 5.52369933255581e-05}
{"train_loss": 0.17209920287132263, "global_step": 126925, "epoch": 1426, "lr": 5.523641677735507e-05}
{"train_loss": 0.11452459543943405, "global_step": 126926, "epoch": 1426, "lr": 5.523584022844805e-05}
{"train_loss": 0.2187618911266327, "global_step": 126927, "epoch": 1426, "lr": 5.523526367883715e-05}
{"train_loss": 0.08309879899024963, "global_step": 126928, "epoch": 1426, "lr": 5.523468712852242e-05}
{"train_loss": 0.1514420062303543, "global_step": 126929, "epoch": 1426, "lr": 5.523411057750396e-05}
{"train_loss": 0.09892717003822327, "global_step": 126930, "epoch": 1426, "lr": 5.523353402578182e-05}
{"train_loss": 0.11392176151275635, "global_step": 126931, "epoch": 1426, "lr": 5.523295747335611e-05}
{"train_loss": 0.08200618624687195, "global_step": 126932, "epoch": 1426, "lr": 5.523238092022689e-05}
{"train_loss": 0.18495534360408783, "global_step": 126933, "epoch": 1426, "lr": 5.523180436639424e-05}
{"train_loss": 0.13979686796665192, "global_step": 126934, "epoch": 1426, "lr": 5.523122781185823e-05}
{"train_loss": 0.1270613968372345, "global_step": 126935, "epoch": 1426, "lr": 5.523065125661896e-05}
{"train_loss": 0.2061154842376709, "global_step": 126936, "epoch": 1426, "lr": 5.523007470067648e-05}
{"train_loss": 0.19792670011520386, "global_step": 126937, "epoch": 1426, "lr": 5.522949814403089e-05}
{"train_loss": 0.17830748856067657, "global_step": 126938, "epoch": 1426, "lr": 5.5228921586682246e-05}
{"train_loss": 0.14915195107460022, "global_step": 126939, "epoch": 1426, "lr": 5.522834502863066e-05}
{"train_loss": 0.1820623129606247, "global_step": 126940, "epoch": 1426, "lr": 5.5227768469876165e-05}
{"train_loss": 0.09508444368839264, "global_step": 126941, "epoch": 1426, "lr": 5.522719191041886e-05}
{"train_loss": 0.14983847737312317, "global_step": 126942, "epoch": 1426, "lr": 5.5226615350258835e-05}
{"train_loss": 0.10943663865327835, "global_step": 126943, "epoch": 1426, "lr": 5.5226038789396164e-05}
{"train_loss": 0.12410077452659607, "global_step": 126944, "epoch": 1426, "lr": 5.5225462227830895e-05}
{"train_loss": 0.12767118215560913, "global_step": 126945, "epoch": 1426, "lr": 5.5224885665563144e-05}
{"train_loss": 0.15037529170513153, "global_step": 126946, "epoch": 1426, "lr": 5.522430910259295e-05}
{"train_loss": 0.10448656976222992, "global_step": 126947, "epoch": 1426, "lr": 5.522373253892042e-05}
{"train_loss": 0.14293482899665833, "global_step": 126948, "epoch": 1426, "lr": 5.522315597454561e-05}
{"train_loss": 0.17327278852462769, "global_step": 126949, "epoch": 1426, "lr": 5.5222579409468633e-05}
{"train_loss": 0.15080468356609344, "global_step": 126950, "epoch": 1426, "lr": 5.522200284368954e-05}
{"train_loss": 0.1725330352783203, "global_step": 126951, "epoch": 1426, "lr": 5.52214262772084e-05}
{"train_loss": 0.1759561449289322, "global_step": 126952, "epoch": 1426, "lr": 5.522084971002531e-05}
{"train_loss": 0.1177607923746109, "global_step": 126953, "epoch": 1426, "lr": 5.5220273142140345e-05}
{"train_loss": 0.14218240976333618, "global_step": 126954, "epoch": 1426, "lr": 5.521969657355357e-05}
{"train_loss": 0.13080783188343048, "global_step": 126955, "epoch": 1426, "lr": 5.521912000426508e-05}
{"train_loss": 0.1661091446876526, "global_step": 126956, "epoch": 1426, "lr": 5.521854343427493e-05}
{"train_loss": 0.10312716662883759, "global_step": 126957, "epoch": 1426, "lr": 5.521796686358322e-05}
{"train_loss": 0.09085904061794281, "global_step": 126958, "epoch": 1426, "lr": 5.5217390292190016e-05}
{"train_loss": 0.11000359803438187, "global_step": 126959, "epoch": 1426, "lr": 5.5216813720095396e-05}
{"train_loss": 0.26070070266723633, "global_step": 126960, "epoch": 1426, "lr": 5.5216237147299446e-05}
{"train_loss": 0.17228153347969055, "global_step": 126961, "epoch": 1426, "lr": 5.521566057380222e-05}
{"train_loss": 0.14644628763198853, "global_step": 126962, "epoch": 1426, "lr": 5.5215083999603835e-05}
{"train_loss": 0.1804581731557846, "global_step": 126963, "epoch": 1426, "lr": 5.521450742470433e-05}
{"train_loss": 0.2761329710483551, "global_step": 126964, "epoch": 1426, "lr": 5.52139308491038e-05}
{"train_loss": 0.16957847774028778, "global_step": 126965, "epoch": 1426, "lr": 5.521335427280232e-05}
{"train_loss": 0.1538027971982956, "global_step": 126966, "epoch": 1426, "lr": 5.521277769579997e-05}
{"train_loss": 0.15828241407871246, "global_step": 126967, "epoch": 1426, "lr": 5.521220111809682e-05}
{"train_loss": 0.09828978776931763, "global_step": 126968, "epoch": 1426, "lr": 5.5211624539692955e-05}
{"train_loss": 0.17012740671634674, "global_step": 126969, "epoch": 1426, "lr": 5.521104796058846e-05}
{"train_loss": 0.14759311079978943, "global_step": 126970, "epoch": 1426, "lr": 5.5210471380783394e-05}
{"train_loss": 0.15686365962028503, "global_step": 126971, "epoch": 1426, "lr": 5.520989480027785e-05}
{"train_loss": 0.18810883164405823, "global_step": 126972, "epoch": 1426, "lr": 5.5209318219071895e-05}
{"train_loss": 0.2058967500925064, "global_step": 126973, "epoch": 1426, "lr": 5.5208741637165604e-05}
{"train_loss": 0.20254673063755035, "global_step": 126974, "epoch": 1426, "lr": 5.520816505455907e-05}
{"train_loss": 0.10096018761396408, "global_step": 126975, "epoch": 1426, "lr": 5.520758847125236e-05}
{"train_loss": 0.17709307372570038, "global_step": 126976, "epoch": 1426, "lr": 5.520701188724555e-05}
{"train_loss": 0.15776698291301727, "global_step": 126977, "epoch": 1426, "lr": 5.5206435302538725e-05}
{"train_loss": 0.16724537312984467, "global_step": 126978, "epoch": 1426, "lr": 5.5205858717131966e-05}
{"train_loss": 0.11358944326639175, "global_step": 126979, "epoch": 1426, "lr": 5.5205282131025327e-05}
{"train_loss": 0.14559033513069153, "global_step": 126980, "epoch": 1426, "lr": 5.5204705544218916e-05}
{"train_loss": 0.17184217274188995, "global_step": 126981, "epoch": 1426, "lr": 5.520412895671279e-05}
{"train_loss": 0.14543883502483368, "global_step": 126982, "epoch": 1426, "lr": 5.520355236850704e-05}
{"train_loss": 0.19485336542129517, "global_step": 126983, "epoch": 1426, "lr": 5.5202975779601716e-05}
{"train_loss": 0.14512357115745544, "global_step": 126984, "epoch": 1426, "lr": 5.520239918999693e-05}
{"train_loss": 0.14167219400405884, "global_step": 126985, "epoch": 1426, "lr": 5.520182259969273e-05}
{"train_loss": 0.10890486091375351, "global_step": 126986, "epoch": 1426, "lr": 5.5201246008689225e-05}
{"train_loss": 0.19805744290351868, "global_step": 126987, "epoch": 1426, "lr": 5.520066941698646e-05}
{"train_loss": 0.1426846981048584, "global_step": 126988, "epoch": 1426, "lr": 5.520009282458454e-05}
{"train_loss": 0.12778019905090332, "global_step": 126989, "epoch": 1426, "lr": 5.519951623148353e-05}
{"train_loss": 0.16958709061145782, "global_step": 126990, "epoch": 1426, "lr": 5.5198939637683514e-05}
{"train_loss": 0.13220472633838654, "global_step": 126991, "epoch": 1426, "lr": 5.5198363043184554e-05}
{"train_loss": 0.10697683691978455, "global_step": 126992, "epoch": 1426, "lr": 5.519778644798673e-05}
{"train_loss": 0.11998467892408371, "global_step": 126993, "epoch": 1426, "lr": 5.5197209852090146e-05}
{"train_loss": 0.11854556947946548, "global_step": 126994, "epoch": 1426, "lr": 5.519663325549485e-05}
{"train_loss": 0.11456262320280075, "global_step": 126995, "epoch": 1426, "lr": 5.5196056658200936e-05}
{"train_loss": 0.13362503051757812, "global_step": 126996, "epoch": 1426, "lr": 5.519548006020847e-05}
{"train_loss": 0.13436703383922577, "global_step": 126997, "epoch": 1426, "lr": 5.519490346151753e-05}
{"train_loss": 0.1917782574892044, "global_step": 126998, "epoch": 1426, "lr": 5.519432686212821e-05}
{"train_loss": 0.17622485756874084, "global_step": 126999, "epoch": 1426, "lr": 5.519375026204057e-05}
{"train_loss": 0.16518402099609375, "global_step": 127000, "epoch": 1426, "lr": 5.519317366125469e-05}
{"train_loss": 0.24057868123054504, "global_step": 127001, "epoch": 1426, "lr": 5.5192597059770666e-05}
{"train_loss": 0.15187754892231373, "global_step": 127002, "epoch": 1426, "lr": 5.5192020457588554e-05, "val_loss": 5.455399513244629}
{"train_loss": 0.15798190236091614, "global_step": 127003, "epoch": 1427, "lr": 5.519144385470843e-05}
{"train_loss": 0.14521916210651398, "global_step": 127004, "epoch": 1427, "lr": 5.5190867251130385e-05}
{"train_loss": 0.0984707847237587, "global_step": 127005, "epoch": 1427, "lr": 5.519029064685448e-05}
{"train_loss": 0.13951821625232697, "global_step": 127006, "epoch": 1427, "lr": 5.518971404188082e-05}
{"train_loss": 0.1625799983739853, "global_step": 127007, "epoch": 1427, "lr": 5.5189137436209457e-05}
{"train_loss": 0.12235843390226364, "global_step": 127008, "epoch": 1427, "lr": 5.518856082984049e-05}
{"train_loss": 0.16878648102283478, "global_step": 127009, "epoch": 1427, "lr": 5.518798422277397e-05}
{"train_loss": 0.12086629867553711, "global_step": 127010, "epoch": 1427, "lr": 5.518740761500999e-05}
{"train_loss": 0.1420278251171112, "global_step": 127011, "epoch": 1427, "lr": 5.518683100654863e-05}
{"train_loss": 0.10766396671533585, "global_step": 127012, "epoch": 1427, "lr": 5.518625439738997e-05}
{"train_loss": 0.16106146574020386, "global_step": 127013, "epoch": 1427, "lr": 5.518567778753407e-05}
{"train_loss": 0.09249322861433029, "global_step": 127014, "epoch": 1427, "lr": 5.5185101176981026e-05}
{"train_loss": 0.09705916047096252, "global_step": 127015, "epoch": 1427, "lr": 5.518452456573091e-05}
{"train_loss": 0.08271001279354095, "global_step": 127016, "epoch": 1427, "lr": 5.5183947953783785e-05}
{"train_loss": 0.14894840121269226, "global_step": 127017, "epoch": 1427, "lr": 5.518337134113975e-05}
{"train_loss": 0.11943329870700836, "global_step": 127018, "epoch": 1427, "lr": 5.5182794727798884e-05}
{"train_loss": 0.12100828438997269, "global_step": 127019, "epoch": 1427, "lr": 5.518221811376124e-05}
{"train_loss": 0.1300906538963318, "global_step": 127020, "epoch": 1427, "lr": 5.518164149902694e-05}
{"train_loss": 0.12653985619544983, "global_step": 127021, "epoch": 1427, "lr": 5.518106488359599e-05}
{"train_loss": 0.12718838453292847, "global_step": 127022, "epoch": 1427, "lr": 5.518048826746852e-05}
{"train_loss": 0.14285558462142944, "global_step": 127023, "epoch": 1427, "lr": 5.51799116506446e-05}
{"train_loss": 0.12671025097370148, "global_step": 127024, "epoch": 1427, "lr": 5.517933503312431e-05}
{"train_loss": 0.16580301523208618, "global_step": 127025, "epoch": 1427, "lr": 5.517875841490772e-05}
{"train_loss": 0.11953897029161453, "global_step": 127026, "epoch": 1427, "lr": 5.51781817959949e-05}
{"train_loss": 0.12633320689201355, "global_step": 127027, "epoch": 1427, "lr": 5.517760517638595e-05}
{"train_loss": 0.13224761188030243, "global_step": 127028, "epoch": 1427, "lr": 5.5177028556080925e-05}
{"train_loss": 0.09110157936811447, "global_step": 127029, "epoch": 1427, "lr": 5.517645193507991e-05}
{"train_loss": 0.1318758875131607, "global_step": 127030, "epoch": 1427, "lr": 5.5175875313383e-05}
{"train_loss": 0.12383594363927841, "global_step": 127031, "epoch": 1427, "lr": 5.517529869099024e-05}
{"train_loss": 0.0978533923625946, "global_step": 127032, "epoch": 1427, "lr": 5.517472206790173e-05}
{"train_loss": 0.1211208924651146, "global_step": 127033, "epoch": 1427, "lr": 5.517414544411753e-05}
{"train_loss": 0.11238056421279907, "global_step": 127034, "epoch": 1427, "lr": 5.517356881963774e-05}
{"train_loss": 0.1148643046617508, "global_step": 127035, "epoch": 1427, "lr": 5.5172992194462434e-05}
{"train_loss": 0.1012563407421112, "global_step": 127036, "epoch": 1427, "lr": 5.517241556859167e-05}
{"train_loss": 0.08763641864061356, "global_step": 127037, "epoch": 1427, "lr": 5.517183894202555e-05}
{"train_loss": 0.09640064090490341, "global_step": 127038, "epoch": 1427, "lr": 5.5171262314764125e-05}
{"train_loss": 0.0997895821928978, "global_step": 127039, "epoch": 1427, "lr": 5.51706856868075e-05}
{"train_loss": 0.16110078990459442, "global_step": 127040, "epoch": 1427, "lr": 5.517010905815573e-05}
{"train_loss": 0.11527351289987564, "global_step": 127041, "epoch": 1427, "lr": 5.51695324288089e-05}
{"train_loss": 0.11962486058473587, "global_step": 127042, "epoch": 1427, "lr": 5.5168955798767096e-05}
{"train_loss": 0.12405635416507721, "global_step": 127043, "epoch": 1427, "lr": 5.516837916803039e-05}
{"train_loss": 0.11078700423240662, "global_step": 127044, "epoch": 1427, "lr": 5.516780253659886e-05}
{"train_loss": 0.14326395094394684, "global_step": 127045, "epoch": 1427, "lr": 5.516722590447257e-05}
{"train_loss": 0.09243817627429962, "global_step": 127046, "epoch": 1427, "lr": 5.516664927165161e-05}
{"train_loss": 0.11346188187599182, "global_step": 127047, "epoch": 1427, "lr": 5.5166072638136066e-05}
{"train_loss": 0.1536453813314438, "global_step": 127048, "epoch": 1427, "lr": 5.516549600392601e-05}
{"train_loss": 0.10404369980096817, "global_step": 127049, "epoch": 1427, "lr": 5.516491936902151e-05}
{"train_loss": 0.09998708963394165, "global_step": 127050, "epoch": 1427, "lr": 5.5164342733422645e-05}
{"train_loss": 0.13686825335025787, "global_step": 127051, "epoch": 1427, "lr": 5.516376609712951e-05}
{"train_loss": 0.12413196265697479, "global_step": 127052, "epoch": 1427, "lr": 5.516318946014216e-05}
{"train_loss": 0.13951817154884338, "global_step": 127053, "epoch": 1427, "lr": 5.516261282246068e-05}
{"train_loss": 0.09869793057441711, "global_step": 127054, "epoch": 1427, "lr": 5.516203618408515e-05}
{"train_loss": 0.10190733522176743, "global_step": 127055, "epoch": 1427, "lr": 5.516145954501566e-05}
{"train_loss": 0.15322083234786987, "global_step": 127056, "epoch": 1427, "lr": 5.5160882905252266e-05}
{"train_loss": 0.05111799016594887, "global_step": 127057, "epoch": 1427, "lr": 5.5160306264795046e-05}
{"train_loss": 0.11164820194244385, "global_step": 127058, "epoch": 1427, "lr": 5.515972962364409e-05}
{"train_loss": 0.171795055270195, "global_step": 127059, "epoch": 1427, "lr": 5.5159152981799475e-05}
{"train_loss": 0.1251082420349121, "global_step": 127060, "epoch": 1427, "lr": 5.515857633926128e-05}
{"train_loss": 0.14839884638786316, "global_step": 127061, "epoch": 1427, "lr": 5.515799969602957e-05}
{"train_loss": 0.11071722209453583, "global_step": 127062, "epoch": 1427, "lr": 5.515742305210443e-05}
{"train_loss": 0.10052303969860077, "global_step": 127063, "epoch": 1427, "lr": 5.515684640748594e-05}
{"train_loss": 0.1177971363067627, "global_step": 127064, "epoch": 1427, "lr": 5.515626976217417e-05}
{"train_loss": 0.19415637850761414, "global_step": 127065, "epoch": 1427, "lr": 5.5155693116169205e-05}
{"train_loss": 0.08252064138650894, "global_step": 127066, "epoch": 1427, "lr": 5.515511646947112e-05}
{"train_loss": 0.0918872132897377, "global_step": 127067, "epoch": 1427, "lr": 5.5154539822079996e-05}
{"train_loss": 0.10889917612075806, "global_step": 127068, "epoch": 1427, "lr": 5.51539631739959e-05}
{"train_loss": 0.13317257165908813, "global_step": 127069, "epoch": 1427, "lr": 5.515338652521893e-05}
{"train_loss": 0.15970821678638458, "global_step": 127070, "epoch": 1427, "lr": 5.5152809875749136e-05}
{"train_loss": 0.061949487775564194, "global_step": 127071, "epoch": 1427, "lr": 5.515223322558661e-05}
{"train_loss": 0.10430095344781876, "global_step": 127072, "epoch": 1427, "lr": 5.5151656574731434e-05}
{"train_loss": 0.14767971634864807, "global_step": 127073, "epoch": 1427, "lr": 5.515107992318368e-05}
{"train_loss": 0.12868110835552216, "global_step": 127074, "epoch": 1427, "lr": 5.515050327094343e-05}
{"train_loss": 0.14537090063095093, "global_step": 127075, "epoch": 1427, "lr": 5.514992661801075e-05}
{"train_loss": 0.1720133125782013, "global_step": 127076, "epoch": 1427, "lr": 5.5149349964385735e-05}
{"train_loss": 0.09702073037624359, "global_step": 127077, "epoch": 1427, "lr": 5.5148773310068434e-05}
{"train_loss": 0.0705452412366867, "global_step": 127078, "epoch": 1427, "lr": 5.514819665505896e-05}
{"train_loss": 0.11683162301778793, "global_step": 127079, "epoch": 1427, "lr": 5.514761999935737e-05}
{"train_loss": 0.11190847307443619, "global_step": 127080, "epoch": 1427, "lr": 5.5147043342963745e-05}
{"train_loss": 0.1354326605796814, "global_step": 127081, "epoch": 1427, "lr": 5.5146466685878165e-05}
{"train_loss": 0.06915803253650665, "global_step": 127082, "epoch": 1427, "lr": 5.51458900281007e-05}
{"train_loss": 0.10651645064353943, "global_step": 127083, "epoch": 1427, "lr": 5.5145313369631433e-05}
{"train_loss": 0.09598180651664734, "global_step": 127084, "epoch": 1427, "lr": 5.5144736710470446e-05}
{"train_loss": 0.15544405579566956, "global_step": 127085, "epoch": 1427, "lr": 5.514416005061781e-05}
{"train_loss": 0.0645357295870781, "global_step": 127086, "epoch": 1427, "lr": 5.5143583390073606e-05}
{"train_loss": 0.17749999463558197, "global_step": 127087, "epoch": 1427, "lr": 5.5143006728837895e-05}
{"train_loss": 0.09622308611869812, "global_step": 127088, "epoch": 1427, "lr": 5.514243006691079e-05}
{"train_loss": 0.10297437757253647, "global_step": 127089, "epoch": 1427, "lr": 5.514185340429234e-05}
{"train_loss": 0.13633699715137482, "global_step": 127090, "epoch": 1427, "lr": 5.514127674098263e-05}
{"train_loss": 0.12155327678061603, "global_step": 127091, "epoch": 1427, "lr": 5.5140700076981734e-05, "val_loss": 5.65210485458374}
{"train_loss": 0.11583212018013, "global_step": 127092, "epoch": 1428, "lr": 5.5140123412289744e-05}
{"train_loss": 0.08482620865106583, "global_step": 127093, "epoch": 1428, "lr": 5.513954674690672e-05}
{"train_loss": 0.1578291952610016, "global_step": 127094, "epoch": 1428, "lr": 5.513897008083274e-05}
{"train_loss": 0.17019475996494293, "global_step": 127095, "epoch": 1428, "lr": 5.5138393414067904e-05}
{"train_loss": 0.11995167285203934, "global_step": 127096, "epoch": 1428, "lr": 5.513781674661226e-05}
{"train_loss": 0.17830975353717804, "global_step": 127097, "epoch": 1428, "lr": 5.51372400784659e-05}
{"train_loss": 0.08593510091304779, "global_step": 127098, "epoch": 1428, "lr": 5.513666340962891e-05}
{"train_loss": 0.11874519288539886, "global_step": 127099, "epoch": 1428, "lr": 5.513608674010136e-05}
{"train_loss": 0.11729360371828079, "global_step": 127100, "epoch": 1428, "lr": 5.513551006988331e-05}
{"train_loss": 0.19143198430538177, "global_step": 127101, "epoch": 1428, "lr": 5.513493339897486e-05}
{"train_loss": 0.14925406873226166, "global_step": 127102, "epoch": 1428, "lr": 5.513435672737609e-05}
{"train_loss": 0.11857787519693375, "global_step": 127103, "epoch": 1428, "lr": 5.513378005508706e-05}
{"train_loss": 0.14284393191337585, "global_step": 127104, "epoch": 1428, "lr": 5.5133203382107856e-05}
{"train_loss": 0.11303571611642838, "global_step": 127105, "epoch": 1428, "lr": 5.513262670843856e-05}
{"train_loss": 0.10562151670455933, "global_step": 127106, "epoch": 1428, "lr": 5.513205003407924e-05}
{"train_loss": 0.20492996275424957, "global_step": 127107, "epoch": 1428, "lr": 5.5131473359029983e-05}
{"train_loss": 0.1263091117143631, "global_step": 127108, "epoch": 1428, "lr": 5.5130896683290853e-05}
{"train_loss": 0.14937201142311096, "global_step": 127109, "epoch": 1428, "lr": 5.513032000686195e-05}
{"train_loss": 0.17290490865707397, "global_step": 127110, "epoch": 1428, "lr": 5.512974332974332e-05}
{"train_loss": 0.20891918241977692, "global_step": 127111, "epoch": 1428, "lr": 5.512916665193507e-05}
{"train_loss": 0.19352568686008453, "global_step": 127112, "epoch": 1428, "lr": 5.512858997343727e-05}
{"train_loss": 0.16022446751594543, "global_step": 127113, "epoch": 1428, "lr": 5.512801329424999e-05}
{"train_loss": 0.06836986541748047, "global_step": 127114, "epoch": 1428, "lr": 5.512743661437331e-05}
{"train_loss": 0.23574508726596832, "global_step": 127115, "epoch": 1428, "lr": 5.512685993380731e-05}
{"train_loss": 0.11667586117982864, "global_step": 127116, "epoch": 1428, "lr": 5.5126283252552056e-05}
{"train_loss": 0.11953220516443253, "global_step": 127117, "epoch": 1428, "lr": 5.512570657060765e-05}
{"train_loss": 0.1390989124774933, "global_step": 127118, "epoch": 1428, "lr": 5.512512988797415e-05}
{"train_loss": 0.08089888840913773, "global_step": 127119, "epoch": 1428, "lr": 5.512455320465164e-05}
{"train_loss": 0.17088061571121216, "global_step": 127120, "epoch": 1428, "lr": 5.512397652064019e-05}
{"train_loss": 0.13325409591197968, "global_step": 127121, "epoch": 1428, "lr": 5.512339983593989e-05}
{"train_loss": 0.09987442940473557, "global_step": 127122, "epoch": 1428, "lr": 5.5122823150550806e-05}
{"train_loss": 0.10633977502584457, "global_step": 127123, "epoch": 1428, "lr": 5.512224646447303e-05}
{"train_loss": 0.1481650173664093, "global_step": 127124, "epoch": 1428, "lr": 5.512166977770662e-05}
{"train_loss": 0.12870462238788605, "global_step": 127125, "epoch": 1428, "lr": 5.5121093090251676e-05}
{"train_loss": 0.0743081271648407, "global_step": 127126, "epoch": 1428, "lr": 5.5120516402108243e-05}
{"train_loss": 0.15230122208595276, "global_step": 127127, "epoch": 1428, "lr": 5.511993971327644e-05}
{"train_loss": 0.16504572331905365, "global_step": 127128, "epoch": 1428, "lr": 5.511936302375631e-05}
{"train_loss": 0.15761317312717438, "global_step": 127129, "epoch": 1428, "lr": 5.5118786333547954e-05}
{"train_loss": 0.11667507141828537, "global_step": 127130, "epoch": 1428, "lr": 5.511820964265142e-05}
{"train_loss": 0.1315983086824417, "global_step": 127131, "epoch": 1428, "lr": 5.511763295106683e-05}
{"train_loss": 0.20308294892311096, "global_step": 127132, "epoch": 1428, "lr": 5.511705625879423e-05}
{"train_loss": 0.1393662691116333, "global_step": 127133, "epoch": 1428, "lr": 5.51164795658337e-05}
{"train_loss": 0.12546144425868988, "global_step": 127134, "epoch": 1428, "lr": 5.511590287218533e-05}
{"train_loss": 0.08227947354316711, "global_step": 127135, "epoch": 1428, "lr": 5.511532617784918e-05}
{"train_loss": 0.16860125958919525, "global_step": 127136, "epoch": 1428, "lr": 5.511474948282533e-05}
{"train_loss": 0.11304900795221329, "global_step": 127137, "epoch": 1428, "lr": 5.511417278711388e-05}
{"train_loss": 0.13649676740169525, "global_step": 127138, "epoch": 1428, "lr": 5.5113596090714894e-05}
{"train_loss": 0.18880034983158112, "global_step": 127139, "epoch": 1428, "lr": 5.5113019393628426e-05}
{"train_loss": 0.0845755860209465, "global_step": 127140, "epoch": 1428, "lr": 5.5112442695854594e-05}
{"train_loss": 0.09074386954307556, "global_step": 127141, "epoch": 1428, "lr": 5.511186599739344e-05}
{"train_loss": 0.16565312445163727, "global_step": 127142, "epoch": 1428, "lr": 5.511128929824507e-05}
{"train_loss": 0.14410561323165894, "global_step": 127143, "epoch": 1428, "lr": 5.511071259840955e-05}
{"train_loss": 0.12917564809322357, "global_step": 127144, "epoch": 1428, "lr": 5.5110135897886964e-05}
{"train_loss": 0.07738766819238663, "global_step": 127145, "epoch": 1428, "lr": 5.5109559196677364e-05}
{"train_loss": 0.10509608685970306, "global_step": 127146, "epoch": 1428, "lr": 5.510898249478086e-05}
{"train_loss": 0.07763363420963287, "global_step": 127147, "epoch": 1428, "lr": 5.51084057921975e-05}
{"train_loss": 0.12196244299411774, "global_step": 127148, "epoch": 1428, "lr": 5.5107829088927374e-05}
{"train_loss": 0.1684802621603012, "global_step": 127149, "epoch": 1428, "lr": 5.510725238497059e-05}
{"train_loss": 0.1569334715604782, "global_step": 127150, "epoch": 1428, "lr": 5.510667568032718e-05}
{"train_loss": 0.10971838980913162, "global_step": 127151, "epoch": 1428, "lr": 5.5106098974997256e-05}
{"train_loss": 0.1750241369009018, "global_step": 127152, "epoch": 1428, "lr": 5.510552226898086e-05}
{"train_loss": 0.1459895819425583, "global_step": 127153, "epoch": 1428, "lr": 5.51049455622781e-05}
{"train_loss": 0.1270584613084793, "global_step": 127154, "epoch": 1428, "lr": 5.5104368854889035e-05}
{"train_loss": 0.3156810998916626, "global_step": 127155, "epoch": 1428, "lr": 5.510379214681375e-05}
{"train_loss": 0.11936893314123154, "global_step": 127156, "epoch": 1428, "lr": 5.5103215438052324e-05}
{"train_loss": 0.1811840832233429, "global_step": 127157, "epoch": 1428, "lr": 5.510263872860484e-05}
{"train_loss": 0.11235535889863968, "global_step": 127158, "epoch": 1428, "lr": 5.5102062018471354e-05}
{"train_loss": 0.11533617973327637, "global_step": 127159, "epoch": 1428, "lr": 5.510148530765197e-05}
{"train_loss": 0.13201554119586945, "global_step": 127160, "epoch": 1428, "lr": 5.510090859614675e-05}
{"train_loss": 0.13863110542297363, "global_step": 127161, "epoch": 1428, "lr": 5.510033188395578e-05}
{"train_loss": 0.1297169327735901, "global_step": 127162, "epoch": 1428, "lr": 5.509975517107913e-05}
{"train_loss": 0.0847194567322731, "global_step": 127163, "epoch": 1428, "lr": 5.5099178457516866e-05}
{"train_loss": 0.12515856325626373, "global_step": 127164, "epoch": 1428, "lr": 5.50986017432691e-05}
{"train_loss": 0.12860342860221863, "global_step": 127165, "epoch": 1428, "lr": 5.5098025028335876e-05}
{"train_loss": 0.09811773896217346, "global_step": 127166, "epoch": 1428, "lr": 5.509744831271728e-05}
{"train_loss": 0.1281955987215042, "global_step": 127167, "epoch": 1428, "lr": 5.50968715964134e-05}
{"train_loss": 0.11251600831747055, "global_step": 127168, "epoch": 1428, "lr": 5.5096294879424314e-05}
{"train_loss": 0.1762862354516983, "global_step": 127169, "epoch": 1428, "lr": 5.509571816175009e-05}
{"train_loss": 0.14630097150802612, "global_step": 127170, "epoch": 1428, "lr": 5.509514144339082e-05}
{"train_loss": 0.14232276380062103, "global_step": 127171, "epoch": 1428, "lr": 5.5094564724346555e-05}
{"train_loss": 0.11118066310882568, "global_step": 127172, "epoch": 1428, "lr": 5.509398800461739e-05}
{"train_loss": 0.1174931451678276, "global_step": 127173, "epoch": 1428, "lr": 5.50934112842034e-05}
{"train_loss": 0.20534977316856384, "global_step": 127174, "epoch": 1428, "lr": 5.5092834563104666e-05}
{"train_loss": 0.09039820730686188, "global_step": 127175, "epoch": 1428, "lr": 5.509225784132127e-05}
{"train_loss": 0.0956529900431633, "global_step": 127176, "epoch": 1428, "lr": 5.5091681118853265e-05}
{"train_loss": 0.11459990590810776, "global_step": 127177, "epoch": 1428, "lr": 5.5091104395700755e-05}
{"train_loss": 0.16889578104019165, "global_step": 127178, "epoch": 1428, "lr": 5.5090527671863814e-05}
{"train_loss": 0.16111166775226593, "global_step": 127179, "epoch": 1428, "lr": 5.5089950947342505e-05}
{"train_loss": 0.13634098321199417, "global_step": 127180, "epoch": 1428, "lr": 5.5089374222136915e-05, "val_loss": 5.614368915557861}
{"train_loss": 0.12022195011377335, "global_step": 127181, "epoch": 1429, "lr": 5.5088797496247126e-05}
{"train_loss": 0.16315293312072754, "global_step": 127182, "epoch": 1429, "lr": 5.5088220769673205e-05}
{"train_loss": 0.0912594348192215, "global_step": 127183, "epoch": 1429, "lr": 5.5087644042415234e-05}
{"train_loss": 0.0699225664138794, "global_step": 127184, "epoch": 1429, "lr": 5.508706731447329e-05}
{"train_loss": 0.08283142000436783, "global_step": 127185, "epoch": 1429, "lr": 5.508649058584745e-05}
{"train_loss": 0.18746240437030792, "global_step": 127186, "epoch": 1429, "lr": 5.50859138565378e-05}
{"train_loss": 0.146164670586586, "global_step": 127187, "epoch": 1429, "lr": 5.50853371265444e-05}
{"train_loss": 0.149562269449234, "global_step": 127188, "epoch": 1429, "lr": 5.5084760395867355e-05}
{"train_loss": 0.10000795125961304, "global_step": 127189, "epoch": 1429, "lr": 5.508418366450673e-05}
{"train_loss": 0.11226578056812286, "global_step": 127190, "epoch": 1429, "lr": 5.508360693246258e-05}
{"train_loss": 0.12955047190189362, "global_step": 127191, "epoch": 1429, "lr": 5.508303019973501e-05}
{"train_loss": 0.0804075300693512, "global_step": 127192, "epoch": 1429, "lr": 5.508245346632408e-05}
{"train_loss": 0.10113094002008438, "global_step": 127193, "epoch": 1429, "lr": 5.508187673222989e-05}
{"train_loss": 0.09481490403413773, "global_step": 127194, "epoch": 1429, "lr": 5.5081299997452495e-05}
{"train_loss": 0.15583620965480804, "global_step": 127195, "epoch": 1429, "lr": 5.5080723261991985e-05}
{"train_loss": 0.11404398083686829, "global_step": 127196, "epoch": 1429, "lr": 5.508014652584843e-05}
{"train_loss": 0.10255157202482224, "global_step": 127197, "epoch": 1429, "lr": 5.507956978902191e-05}
{"train_loss": 0.10191694647073746, "global_step": 127198, "epoch": 1429, "lr": 5.5078993051512505e-05}
{"train_loss": 0.07318881899118423, "global_step": 127199, "epoch": 1429, "lr": 5.5078416313320294e-05}
{"train_loss": 0.11358901858329773, "global_step": 127200, "epoch": 1429, "lr": 5.507783957444535e-05}
{"train_loss": 0.08438598364591599, "global_step": 127201, "epoch": 1429, "lr": 5.507726283488775e-05}
{"train_loss": 0.21349257230758667, "global_step": 127202, "epoch": 1429, "lr": 5.5076686094647566e-05}
{"train_loss": 0.09336576610803604, "global_step": 127203, "epoch": 1429, "lr": 5.507610935372489e-05}
{"train_loss": 0.12145262211561203, "global_step": 127204, "epoch": 1429, "lr": 5.50755326121198e-05}
{"train_loss": 0.15101154148578644, "global_step": 127205, "epoch": 1429, "lr": 5.5074955869832364e-05}
{"train_loss": 0.08119188994169235, "global_step": 127206, "epoch": 1429, "lr": 5.5074379126862664e-05}
{"train_loss": 0.11895088851451874, "global_step": 127207, "epoch": 1429, "lr": 5.507380238321077e-05}
{"train_loss": 0.07565458863973618, "global_step": 127208, "epoch": 1429, "lr": 5.507322563887677e-05}
{"train_loss": 0.1393262892961502, "global_step": 127209, "epoch": 1429, "lr": 5.507264889386075e-05}
{"train_loss": 0.10935593396425247, "global_step": 127210, "epoch": 1429, "lr": 5.5072072148162747e-05}
{"train_loss": 0.10381989926099777, "global_step": 127211, "epoch": 1429, "lr": 5.507149540178288e-05}
{"train_loss": 0.13401050865650177, "global_step": 127212, "epoch": 1429, "lr": 5.5070918654721216e-05}
{"train_loss": 0.13645103573799133, "global_step": 127213, "epoch": 1429, "lr": 5.507034190697782e-05}
{"train_loss": 0.14119140803813934, "global_step": 127214, "epoch": 1429, "lr": 5.506976515855278e-05}
{"train_loss": 0.13066911697387695, "global_step": 127215, "epoch": 1429, "lr": 5.506918840944617e-05}
{"train_loss": 0.11399167031049728, "global_step": 127216, "epoch": 1429, "lr": 5.506861165965808e-05}
{"train_loss": 0.08122755587100983, "global_step": 127217, "epoch": 1429, "lr": 5.506803490918856e-05}
{"train_loss": 0.0834621787071228, "global_step": 127218, "epoch": 1429, "lr": 5.506745815803771e-05}
{"train_loss": 0.1292150318622589, "global_step": 127219, "epoch": 1429, "lr": 5.506688140620561e-05}
{"train_loss": 0.09337815642356873, "global_step": 127220, "epoch": 1429, "lr": 5.506630465369231e-05}
{"train_loss": 0.10339213907718658, "global_step": 127221, "epoch": 1429, "lr": 5.5065727900497935e-05}
{"train_loss": 0.13762332499027252, "global_step": 127222, "epoch": 1429, "lr": 5.5065151146622516e-05}
{"train_loss": 0.13737370073795319, "global_step": 127223, "epoch": 1429, "lr": 5.506457439206615e-05}
{"train_loss": 0.12007777392864227, "global_step": 127224, "epoch": 1429, "lr": 5.5063997636828926e-05}
{"train_loss": 0.11180802434682846, "global_step": 127225, "epoch": 1429, "lr": 5.50634208809109e-05}
{"train_loss": 0.15200267732143402, "global_step": 127226, "epoch": 1429, "lr": 5.506284412431215e-05}
{"train_loss": 0.12332174181938171, "global_step": 127227, "epoch": 1429, "lr": 5.5062267367032786e-05}
{"train_loss": 0.11194849014282227, "global_step": 127228, "epoch": 1429, "lr": 5.506169060907286e-05}
{"train_loss": 0.129257470369339, "global_step": 127229, "epoch": 1429, "lr": 5.506111385043243e-05}
{"train_loss": 0.11061255633831024, "global_step": 127230, "epoch": 1429, "lr": 5.506053709111161e-05}
{"train_loss": 0.06881614029407501, "global_step": 127231, "epoch": 1429, "lr": 5.505996033111046e-05}
{"train_loss": 0.08881764858961105, "global_step": 127232, "epoch": 1429, "lr": 5.5059383570429056e-05}
{"train_loss": 0.06738214939832687, "global_step": 127233, "epoch": 1429, "lr": 5.505880680906748e-05}
{"train_loss": 0.10235422104597092, "global_step": 127234, "epoch": 1429, "lr": 5.50582300470258e-05}
{"train_loss": 0.08234985172748566, "global_step": 127235, "epoch": 1429, "lr": 5.505765328430412e-05}
{"train_loss": 0.10501255095005035, "global_step": 127236, "epoch": 1429, "lr": 5.5057076520902494e-05}
{"train_loss": 0.18212315440177917, "global_step": 127237, "epoch": 1429, "lr": 5.505649975682101e-05}
{"train_loss": 0.12380378693342209, "global_step": 127238, "epoch": 1429, "lr": 5.505592299205973e-05}
{"train_loss": 0.15190894901752472, "global_step": 127239, "epoch": 1429, "lr": 5.5055346226618746e-05}
{"train_loss": 0.1763026863336563, "global_step": 127240, "epoch": 1429, "lr": 5.5054769460498124e-05}
{"train_loss": 0.09111667424440384, "global_step": 127241, "epoch": 1429, "lr": 5.505419269369797e-05}
{"train_loss": 0.13866785168647766, "global_step": 127242, "epoch": 1429, "lr": 5.505361592621833e-05}
{"train_loss": 0.13823120296001434, "global_step": 127243, "epoch": 1429, "lr": 5.505303915805929e-05}
{"train_loss": 0.07477067410945892, "global_step": 127244, "epoch": 1429, "lr": 5.505246238922094e-05}
{"train_loss": 0.1235937550663948, "global_step": 127245, "epoch": 1429, "lr": 5.5051885619703344e-05}
{"train_loss": 0.11360388994216919, "global_step": 127246, "epoch": 1429, "lr": 5.5051308849506576e-05}
{"train_loss": 0.07439795136451721, "global_step": 127247, "epoch": 1429, "lr": 5.505073207863074e-05}
{"train_loss": 0.21256296336650848, "global_step": 127248, "epoch": 1429, "lr": 5.505015530707589e-05}
{"train_loss": 0.06537414342164993, "global_step": 127249, "epoch": 1429, "lr": 5.50495785348421e-05}
{"train_loss": 0.13543002307415009, "global_step": 127250, "epoch": 1429, "lr": 5.504900176192945e-05}
{"train_loss": 0.12383504211902618, "global_step": 127251, "epoch": 1429, "lr": 5.504842498833803e-05}
{"train_loss": 0.08441637456417084, "global_step": 127252, "epoch": 1429, "lr": 5.504784821406792e-05}
{"train_loss": 0.17340783774852753, "global_step": 127253, "epoch": 1429, "lr": 5.504727143911918e-05}
{"train_loss": 0.10583581030368805, "global_step": 127254, "epoch": 1429, "lr": 5.504669466349189e-05}
{"train_loss": 0.07904421538114548, "global_step": 127255, "epoch": 1429, "lr": 5.5046117887186146e-05}
{"train_loss": 0.159351646900177, "global_step": 127256, "epoch": 1429, "lr": 5.5045541110202e-05}
{"train_loss": 0.2079779952764511, "global_step": 127257, "epoch": 1429, "lr": 5.504496433253954e-05}
{"train_loss": 0.12190783768892288, "global_step": 127258, "epoch": 1429, "lr": 5.5044387554198853e-05}
{"train_loss": 0.11016805469989777, "global_step": 127259, "epoch": 1429, "lr": 5.504381077518001e-05}
{"train_loss": 0.08464744687080383, "global_step": 127260, "epoch": 1429, "lr": 5.504323399548309e-05}
{"train_loss": 0.12201962620019913, "global_step": 127261, "epoch": 1429, "lr": 5.504265721510816e-05}
{"train_loss": 0.12602803111076355, "global_step": 127262, "epoch": 1429, "lr": 5.504208043405532e-05}
{"train_loss": 0.16251784563064575, "global_step": 127263, "epoch": 1429, "lr": 5.504150365232462e-05}
{"train_loss": 0.15825431048870087, "global_step": 127264, "epoch": 1429, "lr": 5.504092686991617e-05}
{"train_loss": 0.08316195011138916, "global_step": 127265, "epoch": 1429, "lr": 5.5040350086830004e-05}
{"train_loss": 0.16513511538505554, "global_step": 127266, "epoch": 1429, "lr": 5.503977330306624e-05}
{"train_loss": 0.1659984588623047, "global_step": 127267, "epoch": 1429, "lr": 5.503919651862495e-05}
{"train_loss": 0.15530255436897278, "global_step": 127268, "epoch": 1429, "lr": 5.503861973350619e-05}
{"train_loss": 0.12147599642866114, "global_step": 127269, "epoch": 1429, "lr": 5.503804294771004e-05, "val_loss": 5.727848529815674}
{"train_loss": 0.16505484282970428, "global_step": 127270, "epoch": 1430, "lr": 5.503746616123661e-05}
{"train_loss": 0.15030792355537415, "global_step": 127271, "epoch": 1430, "lr": 5.503688937408593e-05}
{"train_loss": 0.10645615309476852, "global_step": 127272, "epoch": 1430, "lr": 5.503631258625811e-05}
{"train_loss": 0.14164535701274872, "global_step": 127273, "epoch": 1430, "lr": 5.5035735797753215e-05}
{"train_loss": 0.07186630368232727, "global_step": 127274, "epoch": 1430, "lr": 5.5035159008571335e-05}
{"train_loss": 0.21998077630996704, "global_step": 127275, "epoch": 1430, "lr": 5.503458221871253e-05}
{"train_loss": 0.08493056148290634, "global_step": 127276, "epoch": 1430, "lr": 5.503400542817689e-05}
{"train_loss": 0.11329089105129242, "global_step": 127277, "epoch": 1430, "lr": 5.503342863696449e-05}
{"train_loss": 0.07566613703966141, "global_step": 127278, "epoch": 1430, "lr": 5.5032851845075406e-05}
{"train_loss": 0.10661066323518753, "global_step": 127279, "epoch": 1430, "lr": 5.503227505250972e-05}
{"train_loss": 0.155108243227005, "global_step": 127280, "epoch": 1430, "lr": 5.5031698259267507e-05}
{"train_loss": 0.16323745250701904, "global_step": 127281, "epoch": 1430, "lr": 5.503112146534884e-05}
{"train_loss": 0.16683818399906158, "global_step": 127282, "epoch": 1430, "lr": 5.5030544670753805e-05}
{"train_loss": 0.0886753723025322, "global_step": 127283, "epoch": 1430, "lr": 5.502996787548247e-05}
{"train_loss": 0.2212115079164505, "global_step": 127284, "epoch": 1430, "lr": 5.502939107953492e-05}
{"train_loss": 0.07621417194604874, "global_step": 127285, "epoch": 1430, "lr": 5.502881428291122e-05}
{"train_loss": 0.08818481117486954, "global_step": 127286, "epoch": 1430, "lr": 5.502823748561148e-05}
{"train_loss": 0.11857800930738449, "global_step": 127287, "epoch": 1430, "lr": 5.502766068763574e-05}
{"train_loss": 0.13527990877628326, "global_step": 127288, "epoch": 1430, "lr": 5.502708388898409e-05}
{"train_loss": 0.21820662915706635, "global_step": 127289, "epoch": 1430, "lr": 5.5026507089656606e-05}
{"train_loss": 0.18234416842460632, "global_step": 127290, "epoch": 1430, "lr": 5.5025930289653386e-05}
{"train_loss": 0.13981464505195618, "global_step": 127291, "epoch": 1430, "lr": 5.5025353488974477e-05}
{"train_loss": 0.09790554642677307, "global_step": 127292, "epoch": 1430, "lr": 5.5024776687619974e-05}
{"train_loss": 0.14443980157375336, "global_step": 127293, "epoch": 1430, "lr": 5.5024199885589946e-05}
{"train_loss": 0.10227086395025253, "global_step": 127294, "epoch": 1430, "lr": 5.502362308288447e-05}
{"train_loss": 0.11527609825134277, "global_step": 127295, "epoch": 1430, "lr": 5.5023046279503645e-05}
{"train_loss": 0.09081585705280304, "global_step": 127296, "epoch": 1430, "lr": 5.502246947544753e-05}
{"train_loss": 0.1036338210105896, "global_step": 127297, "epoch": 1430, "lr": 5.50218926707162e-05}
{"train_loss": 0.10382100939750671, "global_step": 127298, "epoch": 1430, "lr": 5.502131586530974e-05}
{"train_loss": 0.13007506728172302, "global_step": 127299, "epoch": 1430, "lr": 5.502073905922822e-05}
{"train_loss": 0.16080518066883087, "global_step": 127300, "epoch": 1430, "lr": 5.502016225247173e-05}
{"train_loss": 0.09783875197172165, "global_step": 127301, "epoch": 1430, "lr": 5.5019585445040336e-05}
{"train_loss": 0.1034766286611557, "global_step": 127302, "epoch": 1430, "lr": 5.501900863693412e-05}
{"train_loss": 0.1478804349899292, "global_step": 127303, "epoch": 1430, "lr": 5.501843182815316e-05}
{"train_loss": 0.11441747844219208, "global_step": 127304, "epoch": 1430, "lr": 5.501785501869754e-05}
{"train_loss": 0.07942827045917511, "global_step": 127305, "epoch": 1430, "lr": 5.5017278208567334e-05}
{"train_loss": 0.11337891966104507, "global_step": 127306, "epoch": 1430, "lr": 5.5016701397762614e-05}
{"train_loss": 0.09908126294612885, "global_step": 127307, "epoch": 1430, "lr": 5.5016124586283445e-05}
{"train_loss": 0.08108913153409958, "global_step": 127308, "epoch": 1430, "lr": 5.501554777412993e-05}
{"train_loss": 0.16490612924098969, "global_step": 127309, "epoch": 1430, "lr": 5.5014970961302136e-05}
{"train_loss": 0.1254756897687912, "global_step": 127310, "epoch": 1430, "lr": 5.5014394147800144e-05}
{"train_loss": 0.14137157797813416, "global_step": 127311, "epoch": 1430, "lr": 5.501381733362401e-05}
{"train_loss": 0.07333654910326004, "global_step": 127312, "epoch": 1430, "lr": 5.501324051877383e-05}
{"train_loss": 0.17763639986515045, "global_step": 127313, "epoch": 1430, "lr": 5.501266370324969e-05}
{"train_loss": 0.13837295770645142, "global_step": 127314, "epoch": 1430, "lr": 5.501208688705166e-05}
{"train_loss": 0.1651802510023117, "global_step": 127315, "epoch": 1430, "lr": 5.501151007017982e-05}
{"train_loss": 0.12936630845069885, "global_step": 127316, "epoch": 1430, "lr": 5.501093325263423e-05}
{"train_loss": 0.09015066176652908, "global_step": 127317, "epoch": 1430, "lr": 5.5010356434415e-05}
{"train_loss": 0.17263780534267426, "global_step": 127318, "epoch": 1430, "lr": 5.5009779615522186e-05}
{"train_loss": 0.19537979364395142, "global_step": 127319, "epoch": 1430, "lr": 5.500920279595585e-05}
{"train_loss": 0.1326276808977127, "global_step": 127320, "epoch": 1430, "lr": 5.50086259757161e-05}
{"train_loss": 0.12216083705425262, "global_step": 127321, "epoch": 1430, "lr": 5.5008049154802996e-05}
{"train_loss": 0.10979082435369492, "global_step": 127322, "epoch": 1430, "lr": 5.500747233321664e-05}
{"train_loss": 0.09821927547454834, "global_step": 127323, "epoch": 1430, "lr": 5.500689551095707e-05}
{"train_loss": 0.10744038969278336, "global_step": 127324, "epoch": 1430, "lr": 5.500631868802438e-05}
{"train_loss": 0.16185995936393738, "global_step": 127325, "epoch": 1430, "lr": 5.500574186441868e-05}
{"train_loss": 0.13652412593364716, "global_step": 127326, "epoch": 1430, "lr": 5.5005165040139994e-05}
{"train_loss": 0.13577160239219666, "global_step": 127327, "epoch": 1430, "lr": 5.5004588215188435e-05}
{"train_loss": 0.1490413397550583, "global_step": 127328, "epoch": 1430, "lr": 5.500401138956407e-05}
{"train_loss": 0.17933091521263123, "global_step": 127329, "epoch": 1430, "lr": 5.500343456326695e-05}
{"train_loss": 0.1894419640302658, "global_step": 127330, "epoch": 1430, "lr": 5.500285773629721e-05}
{"train_loss": 0.09068761020898819, "global_step": 127331, "epoch": 1430, "lr": 5.500228090865489e-05}
{"train_loss": 0.1762935370206833, "global_step": 127332, "epoch": 1430, "lr": 5.500170408034007e-05}
{"train_loss": 0.1631719172000885, "global_step": 127333, "epoch": 1430, "lr": 5.500112725135284e-05}
{"train_loss": 0.12847775220870972, "global_step": 127334, "epoch": 1430, "lr": 5.500055042169325e-05}
{"train_loss": 0.15584996342658997, "global_step": 127335, "epoch": 1430, "lr": 5.499997359136142e-05}
{"train_loss": 0.10080499202013016, "global_step": 127336, "epoch": 1430, "lr": 5.499939676035739e-05}
{"train_loss": 0.12759855389595032, "global_step": 127337, "epoch": 1430, "lr": 5.4998819928681256e-05}
{"train_loss": 0.09219268709421158, "global_step": 127338, "epoch": 1430, "lr": 5.499824309633309e-05}
{"train_loss": 0.1297011375427246, "global_step": 127339, "epoch": 1430, "lr": 5.4997666263312976e-05}
{"train_loss": 0.1348993480205536, "global_step": 127340, "epoch": 1430, "lr": 5.4997089429620985e-05}
{"train_loss": 0.13041262328624725, "global_step": 127341, "epoch": 1430, "lr": 5.49965125952572e-05}
{"train_loss": 0.12518584728240967, "global_step": 127342, "epoch": 1430, "lr": 5.4995935760221695e-05}
{"train_loss": 0.10366956144571304, "global_step": 127343, "epoch": 1430, "lr": 5.499535892451454e-05}
{"train_loss": 0.10450506210327148, "global_step": 127344, "epoch": 1430, "lr": 5.499478208813583e-05}
{"train_loss": 0.14671653509140015, "global_step": 127345, "epoch": 1430, "lr": 5.499420525108563e-05}
{"train_loss": 0.12354332953691483, "global_step": 127346, "epoch": 1430, "lr": 5.4993628413364016e-05}
{"train_loss": 0.1360761672258377, "global_step": 127347, "epoch": 1430, "lr": 5.499305157497105e-05}
{"train_loss": 0.1126621663570404, "global_step": 127348, "epoch": 1430, "lr": 5.4992474735906854e-05}
{"train_loss": 0.1186724454164505, "global_step": 127349, "epoch": 1430, "lr": 5.499189789617147e-05}
{"train_loss": 0.10064753890037537, "global_step": 127350, "epoch": 1430, "lr": 5.499132105576499e-05}
{"train_loss": 0.11688423156738281, "global_step": 127351, "epoch": 1430, "lr": 5.4990744214687496e-05}
{"train_loss": 0.14729882776737213, "global_step": 127352, "epoch": 1430, "lr": 5.499016737293905e-05}
{"train_loss": 0.10828268527984619, "global_step": 127353, "epoch": 1430, "lr": 5.498959053051973e-05}
{"train_loss": 0.17459295690059662, "global_step": 127354, "epoch": 1430, "lr": 5.4989013687429626e-05}
{"train_loss": 0.14369681477546692, "global_step": 127355, "epoch": 1430, "lr": 5.498843684366881e-05}
{"train_loss": 0.11321517825126648, "global_step": 127356, "epoch": 1430, "lr": 5.4987859999237365e-05}
{"train_loss": 0.11699694395065308, "global_step": 127357, "epoch": 1430, "lr": 5.498728315413536e-05}
{"train_loss": 0.12999460946642952, "global_step": 127358, "epoch": 1430, "lr": 5.498670630836287e-05, "val_loss": 5.840169429779053, "train_action_mse_error": 15.48106575012207}
{"train_loss": 0.08970703929662704, "global_step": 127359, "epoch": 1431, "lr": 5.4986129461919976e-05}
{"train_loss": 0.1255112737417221, "global_step": 127360, "epoch": 1431, "lr": 5.498555261480677e-05}
{"train_loss": 0.13460218906402588, "global_step": 127361, "epoch": 1431, "lr": 5.498497576702331e-05}
{"train_loss": 0.08225499093532562, "global_step": 127362, "epoch": 1431, "lr": 5.498439891856968e-05}
{"train_loss": 0.11538627743721008, "global_step": 127363, "epoch": 1431, "lr": 5.498382206944597e-05}
{"train_loss": 0.16687950491905212, "global_step": 127364, "epoch": 1431, "lr": 5.4983245219652234e-05}
{"train_loss": 0.10428740084171295, "global_step": 127365, "epoch": 1431, "lr": 5.498266836918857e-05}
{"train_loss": 0.07994271069765091, "global_step": 127366, "epoch": 1431, "lr": 5.4982091518055026e-05}
{"train_loss": 0.08330347388982773, "global_step": 127367, "epoch": 1431, "lr": 5.498151466625171e-05}
{"train_loss": 0.1008983701467514, "global_step": 127368, "epoch": 1431, "lr": 5.49809378137787e-05}
{"train_loss": 0.10948494076728821, "global_step": 127369, "epoch": 1431, "lr": 5.4980360960636054e-05}
{"train_loss": 0.18732218444347382, "global_step": 127370, "epoch": 1431, "lr": 5.497978410682387e-05}
{"train_loss": 0.12320594489574432, "global_step": 127371, "epoch": 1431, "lr": 5.4979207252342204e-05}
{"train_loss": 0.18775032460689545, "global_step": 127372, "epoch": 1431, "lr": 5.4978630397191155e-05}
{"train_loss": 0.08396610617637634, "global_step": 127373, "epoch": 1431, "lr": 5.497805354137078e-05}
{"train_loss": 0.19273561239242554, "global_step": 127374, "epoch": 1431, "lr": 5.497747668488117e-05}
{"train_loss": 0.1267542839050293, "global_step": 127375, "epoch": 1431, "lr": 5.4976899827722395e-05}
{"train_loss": 0.16973894834518433, "global_step": 127376, "epoch": 1431, "lr": 5.4976322969894545e-05}
{"train_loss": 0.11821836978197098, "global_step": 127377, "epoch": 1431, "lr": 5.497574611139768e-05}
{"train_loss": 0.08847109228372574, "global_step": 127378, "epoch": 1431, "lr": 5.4975169252231894e-05}
{"train_loss": 0.07962299883365631, "global_step": 127379, "epoch": 1431, "lr": 5.497459239239725e-05}
{"train_loss": 0.13386933505535126, "global_step": 127380, "epoch": 1431, "lr": 5.497401553189384e-05}
{"train_loss": 0.05591947212815285, "global_step": 127381, "epoch": 1431, "lr": 5.497343867072173e-05}
{"train_loss": 0.12965938448905945, "global_step": 127382, "epoch": 1431, "lr": 5.4972861808881e-05}
{"train_loss": 0.1208709105849266, "global_step": 127383, "epoch": 1431, "lr": 5.497228494637174e-05}
{"train_loss": 0.10783431679010391, "global_step": 127384, "epoch": 1431, "lr": 5.497170808319401e-05}
{"train_loss": 0.13572534918785095, "global_step": 127385, "epoch": 1431, "lr": 5.497113121934788e-05}
{"train_loss": 0.14061497151851654, "global_step": 127386, "epoch": 1431, "lr": 5.497055435483346e-05}
{"train_loss": 0.11552444845438004, "global_step": 127387, "epoch": 1431, "lr": 5.49699774896508e-05}
{"train_loss": 0.13849887251853943, "global_step": 127388, "epoch": 1431, "lr": 5.496940062379999e-05}
{"train_loss": 0.21720370650291443, "global_step": 127389, "epoch": 1431, "lr": 5.496882375728112e-05}
{"train_loss": 0.14633923768997192, "global_step": 127390, "epoch": 1431, "lr": 5.496824689009423e-05}
{"train_loss": 0.081821009516716, "global_step": 127391, "epoch": 1431, "lr": 5.496767002223942e-05}
{"train_loss": 0.1354532241821289, "global_step": 127392, "epoch": 1431, "lr": 5.4967093153716787e-05}
{"train_loss": 0.15117253363132477, "global_step": 127393, "epoch": 1431, "lr": 5.4966516284526383e-05}
{"train_loss": 0.09780482202768326, "global_step": 127394, "epoch": 1431, "lr": 5.4965939414668274e-05}
{"train_loss": 0.08175565302371979, "global_step": 127395, "epoch": 1431, "lr": 5.4965362544142575e-05}
{"train_loss": 0.17606189846992493, "global_step": 127396, "epoch": 1431, "lr": 5.496478567294934e-05}
{"train_loss": 0.09943031519651413, "global_step": 127397, "epoch": 1431, "lr": 5.496420880108863e-05}
{"train_loss": 0.21055002510547638, "global_step": 127398, "epoch": 1431, "lr": 5.496363192856057e-05}
{"train_loss": 0.15589556097984314, "global_step": 127399, "epoch": 1431, "lr": 5.4963055055365197e-05}
{"train_loss": 0.14543162286281586, "global_step": 127400, "epoch": 1431, "lr": 5.4962478181502593e-05}
{"train_loss": 0.10381191223859787, "global_step": 127401, "epoch": 1431, "lr": 5.496190130697286e-05}
{"train_loss": 0.11614501476287842, "global_step": 127402, "epoch": 1431, "lr": 5.4961324431776065e-05}
{"train_loss": 0.09945478290319443, "global_step": 127403, "epoch": 1431, "lr": 5.496074755591226e-05}
{"train_loss": 0.1445612609386444, "global_step": 127404, "epoch": 1431, "lr": 5.496017067938155e-05}
{"train_loss": 0.08696193993091583, "global_step": 127405, "epoch": 1431, "lr": 5.495959380218402e-05}
{"train_loss": 0.12383784353733063, "global_step": 127406, "epoch": 1431, "lr": 5.495901692431972e-05}
{"train_loss": 0.17163802683353424, "global_step": 127407, "epoch": 1431, "lr": 5.495844004578874e-05}
{"train_loss": 0.14240773022174835, "global_step": 127408, "epoch": 1431, "lr": 5.4957863166591164e-05}
{"train_loss": 0.16048283874988556, "global_step": 127409, "epoch": 1431, "lr": 5.495728628672705e-05}
{"train_loss": 0.11422865092754364, "global_step": 127410, "epoch": 1431, "lr": 5.4956709406196506e-05}
{"train_loss": 0.13704772293567657, "global_step": 127411, "epoch": 1431, "lr": 5.495613252499958e-05}
{"train_loss": 0.17250195145606995, "global_step": 127412, "epoch": 1431, "lr": 5.495555564313637e-05}
{"train_loss": 0.08844716101884842, "global_step": 127413, "epoch": 1431, "lr": 5.495497876060695e-05}
{"train_loss": 0.1837366372346878, "global_step": 127414, "epoch": 1431, "lr": 5.495440187741139e-05}
{"train_loss": 0.1085742935538292, "global_step": 127415, "epoch": 1431, "lr": 5.495382499354976e-05}
{"train_loss": 0.09861323982477188, "global_step": 127416, "epoch": 1431, "lr": 5.495324810902216e-05}
{"train_loss": 0.11472739279270172, "global_step": 127417, "epoch": 1431, "lr": 5.4952671223828653e-05}
{"train_loss": 0.15649022161960602, "global_step": 127418, "epoch": 1431, "lr": 5.495209433796932e-05}
{"train_loss": 0.11156397312879562, "global_step": 127419, "epoch": 1431, "lr": 5.495151745144423e-05}
{"train_loss": 0.07377456873655319, "global_step": 127420, "epoch": 1431, "lr": 5.4950940564253483e-05}
{"train_loss": 0.06599551439285278, "global_step": 127421, "epoch": 1431, "lr": 5.4950363676397144e-05}
{"train_loss": 0.14233656227588654, "global_step": 127422, "epoch": 1431, "lr": 5.494978678787528e-05}
{"train_loss": 0.13410308957099915, "global_step": 127423, "epoch": 1431, "lr": 5.494920989868797e-05}
{"train_loss": 0.09477202594280243, "global_step": 127424, "epoch": 1431, "lr": 5.4948633008835314e-05}
{"train_loss": 0.12175492197275162, "global_step": 127425, "epoch": 1431, "lr": 5.494805611831737e-05}
{"train_loss": 0.11774604022502899, "global_step": 127426, "epoch": 1431, "lr": 5.4947479227134214e-05}
{"train_loss": 0.1418818235397339, "global_step": 127427, "epoch": 1431, "lr": 5.4946902335285943e-05}
{"train_loss": 0.0897328183054924, "global_step": 127428, "epoch": 1431, "lr": 5.49463254427726e-05}
{"train_loss": 0.107914999127388, "global_step": 127429, "epoch": 1431, "lr": 5.49457485495943e-05}
{"train_loss": 0.100688137114048, "global_step": 127430, "epoch": 1431, "lr": 5.4945171655751106e-05}
{"train_loss": 0.12922528386116028, "global_step": 127431, "epoch": 1431, "lr": 5.4944594761243085e-05}
{"train_loss": 0.07066436111927032, "global_step": 127432, "epoch": 1431, "lr": 5.494401786607033e-05}
{"train_loss": 0.16386841237545013, "global_step": 127433, "epoch": 1431, "lr": 5.4943440970232916e-05}
{"train_loss": 0.10132236778736115, "global_step": 127434, "epoch": 1431, "lr": 5.4942864073730905e-05}
{"train_loss": 0.10727616399526596, "global_step": 127435, "epoch": 1431, "lr": 5.4942287176564386e-05}
{"train_loss": 0.1024056151509285, "global_step": 127436, "epoch": 1431, "lr": 5.494171027873345e-05}
{"train_loss": 0.12619397044181824, "global_step": 127437, "epoch": 1431, "lr": 5.494113338023815e-05}
{"train_loss": 0.07660127431154251, "global_step": 127438, "epoch": 1431, "lr": 5.494055648107859e-05}
{"train_loss": 0.13329000771045685, "global_step": 127439, "epoch": 1431, "lr": 5.4939979581254806e-05}
{"train_loss": 0.09822677075862885, "global_step": 127440, "epoch": 1431, "lr": 5.493940268076693e-05}
{"train_loss": 0.09066662192344666, "global_step": 127441, "epoch": 1431, "lr": 5.493882577961499e-05}
{"train_loss": 0.1565331220626831, "global_step": 127442, "epoch": 1431, "lr": 5.4938248877799095e-05}
{"train_loss": 0.12226102501153946, "global_step": 127443, "epoch": 1431, "lr": 5.4937671975319314e-05}
{"train_loss": 0.12425170093774796, "global_step": 127444, "epoch": 1431, "lr": 5.493709507217572e-05}
{"train_loss": 0.09660109132528305, "global_step": 127445, "epoch": 1431, "lr": 5.493651816836839e-05}
{"train_loss": 0.12699450552463531, "global_step": 127446, "epoch": 1431, "lr": 5.4935941263897407e-05}
{"train_loss": 0.1229920412800955, "global_step": 127447, "epoch": 1431, "lr": 5.4935364358762855e-05, "val_loss": 5.6513671875}
{"train_loss": 0.0939197689294815, "global_step": 127448, "epoch": 1432, "lr": 5.49347874529648e-05}
{"train_loss": 0.11891825497150421, "global_step": 127449, "epoch": 1432, "lr": 5.493421054650332e-05}
{"train_loss": 0.15650947391986847, "global_step": 127450, "epoch": 1432, "lr": 5.493363363937849e-05}
{"train_loss": 0.158731147646904, "global_step": 127451, "epoch": 1432, "lr": 5.493305673159039e-05}
{"train_loss": 0.0801922157406807, "global_step": 127452, "epoch": 1432, "lr": 5.4932479823139116e-05}
{"train_loss": 0.10075776278972626, "global_step": 127453, "epoch": 1432, "lr": 5.493190291402472e-05}
{"train_loss": 0.09857260435819626, "global_step": 127454, "epoch": 1432, "lr": 5.49313260042473e-05}
{"train_loss": 0.07765941321849823, "global_step": 127455, "epoch": 1432, "lr": 5.493074909380692e-05}
{"train_loss": 0.16850623488426208, "global_step": 127456, "epoch": 1432, "lr": 5.493017218270365e-05}
{"train_loss": 0.12000077217817307, "global_step": 127457, "epoch": 1432, "lr": 5.492959527093759e-05}
{"train_loss": 0.08895475417375565, "global_step": 127458, "epoch": 1432, "lr": 5.49290183585088e-05}
{"train_loss": 0.08110018074512482, "global_step": 127459, "epoch": 1432, "lr": 5.4928441445417364e-05}
{"train_loss": 0.10278748720884323, "global_step": 127460, "epoch": 1432, "lr": 5.4927864531663365e-05}
{"train_loss": 0.12984812259674072, "global_step": 127461, "epoch": 1432, "lr": 5.492728761724687e-05}
{"train_loss": 0.10084666311740875, "global_step": 127462, "epoch": 1432, "lr": 5.492671070216796e-05}
{"train_loss": 0.13288426399230957, "global_step": 127463, "epoch": 1432, "lr": 5.4926133786426716e-05}
{"train_loss": 0.09017282724380493, "global_step": 127464, "epoch": 1432, "lr": 5.492555687002322e-05}
{"train_loss": 0.07560177892446518, "global_step": 127465, "epoch": 1432, "lr": 5.492497995295754e-05}
{"train_loss": 0.0776185542345047, "global_step": 127466, "epoch": 1432, "lr": 5.492440303522974e-05}
{"train_loss": 0.10529762506484985, "global_step": 127467, "epoch": 1432, "lr": 5.492382611683993e-05}
{"train_loss": 0.09318911284208298, "global_step": 127468, "epoch": 1432, "lr": 5.4923249197788174e-05}
{"train_loss": 0.1400640904903412, "global_step": 127469, "epoch": 1432, "lr": 5.492267227807454e-05}
{"train_loss": 0.11546530574560165, "global_step": 127470, "epoch": 1432, "lr": 5.4922095357699114e-05}
{"train_loss": 0.15565958619117737, "global_step": 127471, "epoch": 1432, "lr": 5.4921518436661966e-05}
{"train_loss": 0.18067625164985657, "global_step": 127472, "epoch": 1432, "lr": 5.4920941514963184e-05}
{"train_loss": 0.1513822227716446, "global_step": 127473, "epoch": 1432, "lr": 5.492036459260285e-05}
{"train_loss": 0.11601455509662628, "global_step": 127474, "epoch": 1432, "lr": 5.491978766958102e-05}
{"train_loss": 0.13390083611011505, "global_step": 127475, "epoch": 1432, "lr": 5.49192107458978e-05}
{"train_loss": 0.1962876170873642, "global_step": 127476, "epoch": 1432, "lr": 5.491863382155323e-05}
{"train_loss": 0.12716320157051086, "global_step": 127477, "epoch": 1432, "lr": 5.491805689654743e-05}
{"train_loss": 0.16621530055999756, "global_step": 127478, "epoch": 1432, "lr": 5.4917479970880446e-05}
{"train_loss": 0.0721488818526268, "global_step": 127479, "epoch": 1432, "lr": 5.491690304455237e-05}
{"train_loss": 0.12013901025056839, "global_step": 127480, "epoch": 1432, "lr": 5.491632611756328e-05}
{"train_loss": 0.08908302336931229, "global_step": 127481, "epoch": 1432, "lr": 5.4915749189913244e-05}
{"train_loss": 0.11207304149866104, "global_step": 127482, "epoch": 1432, "lr": 5.4915172261602354e-05}
{"train_loss": 0.12967172265052795, "global_step": 127483, "epoch": 1432, "lr": 5.491459533263067e-05}
{"train_loss": 0.14972537755966187, "global_step": 127484, "epoch": 1432, "lr": 5.491401840299828e-05}
{"train_loss": 0.09959464520215988, "global_step": 127485, "epoch": 1432, "lr": 5.491344147270526e-05}
{"train_loss": 0.0782604068517685, "global_step": 127486, "epoch": 1432, "lr": 5.49128645417517e-05}
{"train_loss": 0.1164742037653923, "global_step": 127487, "epoch": 1432, "lr": 5.491228761013765e-05}
{"train_loss": 0.15325616300106049, "global_step": 127488, "epoch": 1432, "lr": 5.491171067786321e-05}
{"train_loss": 0.06354500353336334, "global_step": 127489, "epoch": 1432, "lr": 5.4911133744928445e-05}
{"train_loss": 0.14775608479976654, "global_step": 127490, "epoch": 1432, "lr": 5.491055681133344e-05}
{"train_loss": 0.1266518533229828, "global_step": 127491, "epoch": 1432, "lr": 5.490997987707828e-05}
{"train_loss": 0.12037664651870728, "global_step": 127492, "epoch": 1432, "lr": 5.490940294216301e-05}
{"train_loss": 0.10810603946447372, "global_step": 127493, "epoch": 1432, "lr": 5.4908826006587744e-05}
{"train_loss": 0.08004549890756607, "global_step": 127494, "epoch": 1432, "lr": 5.4908249070352555e-05}
{"train_loss": 0.13036426901817322, "global_step": 127495, "epoch": 1432, "lr": 5.49076721334575e-05}
{"train_loss": 0.1125052273273468, "global_step": 127496, "epoch": 1432, "lr": 5.490709519590268e-05}
{"train_loss": 0.23283588886260986, "global_step": 127497, "epoch": 1432, "lr": 5.490651825768815e-05}
{"train_loss": 0.14693617820739746, "global_step": 127498, "epoch": 1432, "lr": 5.4905941318814004e-05}
{"train_loss": 0.07609131932258606, "global_step": 127499, "epoch": 1432, "lr": 5.490536437928031e-05}
{"train_loss": 0.07166783511638641, "global_step": 127500, "epoch": 1432, "lr": 5.490478743908715e-05}
{"train_loss": 0.18137849867343903, "global_step": 127501, "epoch": 1432, "lr": 5.49042104982346e-05}
{"train_loss": 0.10181118547916412, "global_step": 127502, "epoch": 1432, "lr": 5.490363355672274e-05}
{"train_loss": 0.1444951593875885, "global_step": 127503, "epoch": 1432, "lr": 5.490305661455165e-05}
{"train_loss": 0.2080112099647522, "global_step": 127504, "epoch": 1432, "lr": 5.4902479671721394e-05}
{"train_loss": 0.13528482615947723, "global_step": 127505, "epoch": 1432, "lr": 5.4901902728232066e-05}
{"train_loss": 0.08186265826225281, "global_step": 127506, "epoch": 1432, "lr": 5.4901325784083746e-05}
{"train_loss": 0.13138900697231293, "global_step": 127507, "epoch": 1432, "lr": 5.4900748839276496e-05}
{"train_loss": 0.13139478862285614, "global_step": 127508, "epoch": 1432, "lr": 5.490017189381039e-05}
{"train_loss": 0.13513807952404022, "global_step": 127509, "epoch": 1432, "lr": 5.489959494768553e-05}
{"train_loss": 0.10567045956850052, "global_step": 127510, "epoch": 1432, "lr": 5.489901800090197e-05}
{"train_loss": 0.1454186737537384, "global_step": 127511, "epoch": 1432, "lr": 5.489844105345978e-05}
{"train_loss": 0.12432439625263214, "global_step": 127512, "epoch": 1432, "lr": 5.489786410535908e-05}
{"train_loss": 0.1636887639760971, "global_step": 127513, "epoch": 1432, "lr": 5.489728715659993e-05}
{"train_loss": 0.12000754475593567, "global_step": 127514, "epoch": 1432, "lr": 5.4896710207182375e-05}
{"train_loss": 0.1423851102590561, "global_step": 127515, "epoch": 1432, "lr": 5.489613325710653e-05}
{"train_loss": 0.1279471516609192, "global_step": 127516, "epoch": 1432, "lr": 5.489555630637245e-05}
{"train_loss": 0.07061302661895752, "global_step": 127517, "epoch": 1432, "lr": 5.489497935498024e-05}
{"train_loss": 0.09122003614902496, "global_step": 127518, "epoch": 1432, "lr": 5.489440240292995e-05}
{"train_loss": 0.15963666141033173, "global_step": 127519, "epoch": 1432, "lr": 5.489382545022166e-05}
{"train_loss": 0.1659603863954544, "global_step": 127520, "epoch": 1432, "lr": 5.489324849685546e-05}
{"train_loss": 0.13911433517932892, "global_step": 127521, "epoch": 1432, "lr": 5.489267154283142e-05}
{"train_loss": 0.1367531269788742, "global_step": 127522, "epoch": 1432, "lr": 5.4892094588149625e-05}
{"train_loss": 0.1447179764509201, "global_step": 127523, "epoch": 1432, "lr": 5.4891517632810144e-05}
{"train_loss": 0.07667607069015503, "global_step": 127524, "epoch": 1432, "lr": 5.489094067681306e-05}
{"train_loss": 0.1925603747367859, "global_step": 127525, "epoch": 1432, "lr": 5.489036372015844e-05}
{"train_loss": 0.12529021501541138, "global_step": 127526, "epoch": 1432, "lr": 5.488978676284638e-05}
{"train_loss": 0.19860006868839264, "global_step": 127527, "epoch": 1432, "lr": 5.488920980487695e-05}
{"train_loss": 0.2419082224369049, "global_step": 127528, "epoch": 1432, "lr": 5.4888632846250224e-05}
{"train_loss": 0.07099532335996628, "global_step": 127529, "epoch": 1432, "lr": 5.488805588696626e-05}
{"train_loss": 0.1290925294160843, "global_step": 127530, "epoch": 1432, "lr": 5.488747892702518e-05}
{"train_loss": 0.1627807319164276, "global_step": 127531, "epoch": 1432, "lr": 5.4886901966427026e-05}
{"train_loss": 0.1434594690799713, "global_step": 127532, "epoch": 1432, "lr": 5.4886325005171914e-05}
{"train_loss": 0.11749870330095291, "global_step": 127533, "epoch": 1432, "lr": 5.488574804325988e-05}
{"train_loss": 0.06043563783168793, "global_step": 127534, "epoch": 1432, "lr": 5.4885171080691e-05}
{"train_loss": 0.12120751291513443, "global_step": 127535, "epoch": 1432, "lr": 5.488459411746538e-05}
{"train_loss": 0.12580891388855622, "global_step": 127536, "epoch": 1432, "lr": 5.488401715358309e-05, "val_loss": 5.592989921569824}
{"train_loss": 0.136810302734375, "global_step": 127537, "epoch": 1433, "lr": 5.48834401890442e-05}
{"train_loss": 0.07816625386476517, "global_step": 127538, "epoch": 1433, "lr": 5.488286322384879e-05}
{"train_loss": 0.10234542191028595, "global_step": 127539, "epoch": 1433, "lr": 5.488228625799694e-05}
{"train_loss": 0.12467994540929794, "global_step": 127540, "epoch": 1433, "lr": 5.488170929148873e-05}
{"train_loss": 0.07476534694433212, "global_step": 127541, "epoch": 1433, "lr": 5.4881132324324226e-05}
{"train_loss": 0.10232832282781601, "global_step": 127542, "epoch": 1433, "lr": 5.488055535650352e-05}
{"train_loss": 0.1270350217819214, "global_step": 127543, "epoch": 1433, "lr": 5.487997838802668e-05}
{"train_loss": 0.15681126713752747, "global_step": 127544, "epoch": 1433, "lr": 5.4879401418893786e-05}
{"train_loss": 0.07707810401916504, "global_step": 127545, "epoch": 1433, "lr": 5.4878824449104905e-05}
{"train_loss": 0.07713890075683594, "global_step": 127546, "epoch": 1433, "lr": 5.4878247478660136e-05}
{"train_loss": 0.11996327340602875, "global_step": 127547, "epoch": 1433, "lr": 5.487767050755954e-05}
{"train_loss": 0.09064198285341263, "global_step": 127548, "epoch": 1433, "lr": 5.4877093535803204e-05}
{"train_loss": 0.1477225124835968, "global_step": 127549, "epoch": 1433, "lr": 5.48765165633912e-05}
{"train_loss": 0.09768886864185333, "global_step": 127550, "epoch": 1433, "lr": 5.487593959032362e-05}
{"train_loss": 0.14619293808937073, "global_step": 127551, "epoch": 1433, "lr": 5.487536261660052e-05}
{"train_loss": 0.18452288210391998, "global_step": 127552, "epoch": 1433, "lr": 5.487478564222199e-05}
{"train_loss": 0.15153375267982483, "global_step": 127553, "epoch": 1433, "lr": 5.487420866718811e-05}
{"train_loss": 0.1079750657081604, "global_step": 127554, "epoch": 1433, "lr": 5.487363169149895e-05}
{"train_loss": 0.14787383377552032, "global_step": 127555, "epoch": 1433, "lr": 5.487305471515458e-05}
{"train_loss": 0.14724501967430115, "global_step": 127556, "epoch": 1433, "lr": 5.487247773815509e-05}
{"train_loss": 0.14333853125572205, "global_step": 127557, "epoch": 1433, "lr": 5.487190076050056e-05}
{"train_loss": 0.10855955630540848, "global_step": 127558, "epoch": 1433, "lr": 5.4871323782191065e-05}
{"train_loss": 0.09655178338289261, "global_step": 127559, "epoch": 1433, "lr": 5.487074680322667e-05}
{"train_loss": 0.13146640360355377, "global_step": 127560, "epoch": 1433, "lr": 5.487016982360746e-05}
{"train_loss": 0.10905785113573074, "global_step": 127561, "epoch": 1433, "lr": 5.4869592843333526e-05}
{"train_loss": 0.2088898867368698, "global_step": 127562, "epoch": 1433, "lr": 5.486901586240493e-05}
{"train_loss": 0.14071974158287048, "global_step": 127563, "epoch": 1433, "lr": 5.486843888082176e-05}
{"train_loss": 0.09113927185535431, "global_step": 127564, "epoch": 1433, "lr": 5.486786189858408e-05}
{"train_loss": 0.1661202609539032, "global_step": 127565, "epoch": 1433, "lr": 5.4867284915691966e-05}
{"train_loss": 0.1278700828552246, "global_step": 127566, "epoch": 1433, "lr": 5.486670793214551e-05}
{"train_loss": 0.11930236220359802, "global_step": 127567, "epoch": 1433, "lr": 5.4866130947944795e-05}
{"train_loss": 0.15162168443202972, "global_step": 127568, "epoch": 1433, "lr": 5.4865553963089875e-05}
{"train_loss": 0.20094327628612518, "global_step": 127569, "epoch": 1433, "lr": 5.486497697758086e-05}
{"train_loss": 0.0941329076886177, "global_step": 127570, "epoch": 1433, "lr": 5.486439999141779e-05}
{"train_loss": 0.10038641095161438, "global_step": 127571, "epoch": 1433, "lr": 5.4863823004600776e-05}
{"train_loss": 0.1461094617843628, "global_step": 127572, "epoch": 1433, "lr": 5.486324601712988e-05}
{"train_loss": 0.08541873842477798, "global_step": 127573, "epoch": 1433, "lr": 5.4862669029005174e-05}
{"train_loss": 0.1312035322189331, "global_step": 127574, "epoch": 1433, "lr": 5.486209204022674e-05}
{"train_loss": 0.11982716619968414, "global_step": 127575, "epoch": 1433, "lr": 5.486151505079465e-05}
{"train_loss": 0.12358954548835754, "global_step": 127576, "epoch": 1433, "lr": 5.4860938060709e-05}
{"train_loss": 0.11983594298362732, "global_step": 127577, "epoch": 1433, "lr": 5.4860361069969855e-05}
{"train_loss": 0.18603640794754028, "global_step": 127578, "epoch": 1433, "lr": 5.4859784078577284e-05}
{"train_loss": 0.18957705795764923, "global_step": 127579, "epoch": 1433, "lr": 5.4859207086531386e-05}
{"train_loss": 0.17107491195201874, "global_step": 127580, "epoch": 1433, "lr": 5.485863009383222e-05}
{"train_loss": 0.10738636553287506, "global_step": 127581, "epoch": 1433, "lr": 5.485805310047987e-05}
{"train_loss": 0.08996923267841339, "global_step": 127582, "epoch": 1433, "lr": 5.4857476106474415e-05}
{"train_loss": 0.07006228715181351, "global_step": 127583, "epoch": 1433, "lr": 5.4856899111815915e-05}
{"train_loss": 0.16721123456954956, "global_step": 127584, "epoch": 1433, "lr": 5.485632211650449e-05}
{"train_loss": 0.13339881598949432, "global_step": 127585, "epoch": 1433, "lr": 5.485574512054018e-05}
{"train_loss": 0.1975124478340149, "global_step": 127586, "epoch": 1433, "lr": 5.485516812392307e-05}
{"train_loss": 0.14865891635417938, "global_step": 127587, "epoch": 1433, "lr": 5.4854591126653256e-05}
{"train_loss": 0.16158156096935272, "global_step": 127588, "epoch": 1433, "lr": 5.485401412873079e-05}
{"train_loss": 0.12801851332187653, "global_step": 127589, "epoch": 1433, "lr": 5.4853437130155774e-05}
{"train_loss": 0.10137862712144852, "global_step": 127590, "epoch": 1433, "lr": 5.4852860130928277e-05}
{"train_loss": 0.13632230460643768, "global_step": 127591, "epoch": 1433, "lr": 5.4852283131048354e-05}
{"train_loss": 0.22472800314426422, "global_step": 127592, "epoch": 1433, "lr": 5.48517061305161e-05}
{"train_loss": 0.193964421749115, "global_step": 127593, "epoch": 1433, "lr": 5.4851129129331604e-05}
{"train_loss": 0.1209249421954155, "global_step": 127594, "epoch": 1433, "lr": 5.4850552127494927e-05}
{"train_loss": 0.16309601068496704, "global_step": 127595, "epoch": 1433, "lr": 5.484997512500615e-05}
{"train_loss": 0.13535423576831818, "global_step": 127596, "epoch": 1433, "lr": 5.484939812186536e-05}
{"train_loss": 0.16278114914894104, "global_step": 127597, "epoch": 1433, "lr": 5.4848821118072625e-05}
{"train_loss": 0.2137007862329483, "global_step": 127598, "epoch": 1433, "lr": 5.484824411362802e-05}
{"train_loss": 0.13268934190273285, "global_step": 127599, "epoch": 1433, "lr": 5.484766710853163e-05}
{"train_loss": 0.14513123035430908, "global_step": 127600, "epoch": 1433, "lr": 5.484709010278353e-05}
{"train_loss": 0.09104923903942108, "global_step": 127601, "epoch": 1433, "lr": 5.484651309638379e-05}
{"train_loss": 0.08645883202552795, "global_step": 127602, "epoch": 1433, "lr": 5.484593608933251e-05}
{"train_loss": 0.15612323582172394, "global_step": 127603, "epoch": 1433, "lr": 5.484535908162974e-05}
{"train_loss": 0.10972990095615387, "global_step": 127604, "epoch": 1433, "lr": 5.484478207327558e-05}
{"train_loss": 0.09954600781202316, "global_step": 127605, "epoch": 1433, "lr": 5.4844205064270095e-05}
{"train_loss": 0.18636706471443176, "global_step": 127606, "epoch": 1433, "lr": 5.484362805461337e-05}
{"train_loss": 0.13128142058849335, "global_step": 127607, "epoch": 1433, "lr": 5.484305104430547e-05}
{"train_loss": 0.12162463366985321, "global_step": 127608, "epoch": 1433, "lr": 5.484247403334649e-05}
{"train_loss": 0.1511433720588684, "global_step": 127609, "epoch": 1433, "lr": 5.48418970217365e-05}
{"train_loss": 0.09962454438209534, "global_step": 127610, "epoch": 1433, "lr": 5.484132000947557e-05}
{"train_loss": 0.22495749592781067, "global_step": 127611, "epoch": 1433, "lr": 5.4840742996563786e-05}
{"train_loss": 0.1256657838821411, "global_step": 127612, "epoch": 1433, "lr": 5.4840165983001225e-05}
{"train_loss": 0.19853442907333374, "global_step": 127613, "epoch": 1433, "lr": 5.483958896878796e-05}
{"train_loss": 0.08699759840965271, "global_step": 127614, "epoch": 1433, "lr": 5.483901195392407e-05}
{"train_loss": 0.12610971927642822, "global_step": 127615, "epoch": 1433, "lr": 5.4838434938409635e-05}
{"train_loss": 0.13582655787467957, "global_step": 127616, "epoch": 1433, "lr": 5.483785792224473e-05}
{"train_loss": 0.18009692430496216, "global_step": 127617, "epoch": 1433, "lr": 5.4837280905429434e-05}
{"train_loss": 0.09772610664367676, "global_step": 127618, "epoch": 1433, "lr": 5.483670388796383e-05}
{"train_loss": 0.05662691593170166, "global_step": 127619, "epoch": 1433, "lr": 5.483612686984797e-05}
{"train_loss": 0.07279892265796661, "global_step": 127620, "epoch": 1433, "lr": 5.4835549851081956e-05}
{"train_loss": 0.19422726333141327, "global_step": 127621, "epoch": 1433, "lr": 5.4834972831665875e-05}
{"train_loss": 0.09715386480093002, "global_step": 127622, "epoch": 1433, "lr": 5.4834395811599795e-05}
{"train_loss": 0.19798864424228668, "global_step": 127623, "epoch": 1433, "lr": 5.4833818790883776e-05}
{"train_loss": 0.19322173297405243, "global_step": 127624, "epoch": 1433, "lr": 5.483324176951792e-05}
{"train_loss": 0.13378116714485574, "global_step": 127625, "epoch": 1433, "lr": 5.483266474750228e-05, "val_loss": 5.511391639709473}
{"train_loss": 0.16862192749977112, "global_step": 127626, "epoch": 1434, "lr": 5.4832087724836956e-05}
{"train_loss": 0.14066003262996674, "global_step": 127627, "epoch": 1434, "lr": 5.483151070152202e-05}
{"train_loss": 0.08041562139987946, "global_step": 127628, "epoch": 1434, "lr": 5.483093367755754e-05}
{"train_loss": 0.10527903586626053, "global_step": 127629, "epoch": 1434, "lr": 5.483035665294362e-05}
{"train_loss": 0.12149500846862793, "global_step": 127630, "epoch": 1434, "lr": 5.48297796276803e-05}
{"train_loss": 0.1501242071390152, "global_step": 127631, "epoch": 1434, "lr": 5.482920260176767e-05}
{"train_loss": 0.09352880716323853, "global_step": 127632, "epoch": 1434, "lr": 5.482862557520582e-05}
{"train_loss": 0.14148224890232086, "global_step": 127633, "epoch": 1434, "lr": 5.482804854799481e-05}
{"train_loss": 0.16938181221485138, "global_step": 127634, "epoch": 1434, "lr": 5.482747152013474e-05}
{"train_loss": 0.16484509408473969, "global_step": 127635, "epoch": 1434, "lr": 5.4826894491625667e-05}
{"train_loss": 0.10815009474754333, "global_step": 127636, "epoch": 1434, "lr": 5.482631746246768e-05}
{"train_loss": 0.09650710970163345, "global_step": 127637, "epoch": 1434, "lr": 5.4825740432660854e-05}
{"train_loss": 0.13869334757328033, "global_step": 127638, "epoch": 1434, "lr": 5.482516340220526e-05}
{"train_loss": 0.09336031973361969, "global_step": 127639, "epoch": 1434, "lr": 5.482458637110098e-05}
{"train_loss": 0.11564771831035614, "global_step": 127640, "epoch": 1434, "lr": 5.4824009339348104e-05}
{"train_loss": 0.11318376660346985, "global_step": 127641, "epoch": 1434, "lr": 5.48234323069467e-05}
{"train_loss": 0.09446775913238525, "global_step": 127642, "epoch": 1434, "lr": 5.4822855273896835e-05}
{"train_loss": 0.10724291205406189, "global_step": 127643, "epoch": 1434, "lr": 5.48222782401986e-05}
{"train_loss": 0.12981154024600983, "global_step": 127644, "epoch": 1434, "lr": 5.482170120585207e-05}
{"train_loss": 0.14817318320274353, "global_step": 127645, "epoch": 1434, "lr": 5.482112417085732e-05}
{"train_loss": 0.12213465571403503, "global_step": 127646, "epoch": 1434, "lr": 5.482054713521443e-05}
{"train_loss": 0.1345335841178894, "global_step": 127647, "epoch": 1434, "lr": 5.481997009892348e-05}
{"train_loss": 0.20863837003707886, "global_step": 127648, "epoch": 1434, "lr": 5.481939306198455e-05}
{"train_loss": 0.07916317135095596, "global_step": 127649, "epoch": 1434, "lr": 5.481881602439769e-05}
{"train_loss": 0.1182660311460495, "global_step": 127650, "epoch": 1434, "lr": 5.4818238986163015e-05}
{"train_loss": 0.13311389088630676, "global_step": 127651, "epoch": 1434, "lr": 5.481766194728057e-05}
{"train_loss": 0.12096618115901947, "global_step": 127652, "epoch": 1434, "lr": 5.481708490775047e-05}
{"train_loss": 0.15801890194416046, "global_step": 127653, "epoch": 1434, "lr": 5.481650786757276e-05}
{"train_loss": 0.0908568724989891, "global_step": 127654, "epoch": 1434, "lr": 5.481593082674753e-05}
{"train_loss": 0.07260807603597641, "global_step": 127655, "epoch": 1434, "lr": 5.481535378527486e-05}
{"train_loss": 0.1460898518562317, "global_step": 127656, "epoch": 1434, "lr": 5.481477674315481e-05}
{"train_loss": 0.12617063522338867, "global_step": 127657, "epoch": 1434, "lr": 5.481419970038748e-05}
{"train_loss": 0.1253526657819748, "global_step": 127658, "epoch": 1434, "lr": 5.481362265697294e-05}
{"train_loss": 0.13497062027454376, "global_step": 127659, "epoch": 1434, "lr": 5.4813045612911274e-05}
{"train_loss": 0.1010693907737732, "global_step": 127660, "epoch": 1434, "lr": 5.4812468568202545e-05}
{"train_loss": 0.18259020149707794, "global_step": 127661, "epoch": 1434, "lr": 5.4811891522846846e-05}
{"train_loss": 0.12790943682193756, "global_step": 127662, "epoch": 1434, "lr": 5.4811314476844245e-05}
{"train_loss": 0.16502664983272552, "global_step": 127663, "epoch": 1434, "lr": 5.4810737430194815e-05}
{"train_loss": 0.15312254428863525, "global_step": 127664, "epoch": 1434, "lr": 5.4810160382898645e-05}
{"train_loss": 0.08186937868595123, "global_step": 127665, "epoch": 1434, "lr": 5.48095833349558e-05}
{"train_loss": 0.10683481395244598, "global_step": 127666, "epoch": 1434, "lr": 5.480900628636638e-05}
{"train_loss": 0.10265467315912247, "global_step": 127667, "epoch": 1434, "lr": 5.4808429237130446e-05}
{"train_loss": 0.1593761146068573, "global_step": 127668, "epoch": 1434, "lr": 5.4807852187248074e-05}
{"train_loss": 0.11641572415828705, "global_step": 127669, "epoch": 1434, "lr": 5.480727513671935e-05}
{"train_loss": 0.03979254141449928, "global_step": 127670, "epoch": 1434, "lr": 5.480669808554433e-05}
{"train_loss": 0.1638200432062149, "global_step": 127671, "epoch": 1434, "lr": 5.480612103372312e-05}
{"train_loss": 0.05724233761429787, "global_step": 127672, "epoch": 1434, "lr": 5.480554398125578e-05}
{"train_loss": 0.1267128437757492, "global_step": 127673, "epoch": 1434, "lr": 5.48049669281424e-05}
{"train_loss": 0.07258109748363495, "global_step": 127674, "epoch": 1434, "lr": 5.4804389874383035e-05}
{"train_loss": 0.09664137661457062, "global_step": 127675, "epoch": 1434, "lr": 5.4803812819977785e-05}
{"train_loss": 0.11619891226291656, "global_step": 127676, "epoch": 1434, "lr": 5.4803235764926716e-05}
{"train_loss": 0.09883583337068558, "global_step": 127677, "epoch": 1434, "lr": 5.4802658709229924e-05}
{"train_loss": 0.15207330882549286, "global_step": 127678, "epoch": 1434, "lr": 5.480208165288747e-05}
{"train_loss": 0.09435930848121643, "global_step": 127679, "epoch": 1434, "lr": 5.480150459589943e-05}
{"train_loss": 0.07756586372852325, "global_step": 127680, "epoch": 1434, "lr": 5.4800927538265904e-05}
{"train_loss": 0.08790308982133865, "global_step": 127681, "epoch": 1434, "lr": 5.480035047998693e-05}
{"train_loss": 0.1340527981519699, "global_step": 127682, "epoch": 1434, "lr": 5.4799773421062615e-05}
{"train_loss": 0.16958315670490265, "global_step": 127683, "epoch": 1434, "lr": 5.479919636149303e-05}
{"train_loss": 0.14076493680477142, "global_step": 127684, "epoch": 1434, "lr": 5.4798619301278255e-05}
{"train_loss": 0.2099621742963791, "global_step": 127685, "epoch": 1434, "lr": 5.4798042240418357e-05}
{"train_loss": 0.17742405831813812, "global_step": 127686, "epoch": 1434, "lr": 5.479746517891342e-05}
{"train_loss": 0.12669247388839722, "global_step": 127687, "epoch": 1434, "lr": 5.479688811676354e-05}
{"train_loss": 0.2251514196395874, "global_step": 127688, "epoch": 1434, "lr": 5.4796311053968764e-05}
{"train_loss": 0.15567703545093536, "global_step": 127689, "epoch": 1434, "lr": 5.4795733990529176e-05}
{"train_loss": 0.16463258862495422, "global_step": 127690, "epoch": 1434, "lr": 5.4795156926444854e-05}
{"train_loss": 0.14106516540050507, "global_step": 127691, "epoch": 1434, "lr": 5.47945798617159e-05}
{"train_loss": 0.12159010767936707, "global_step": 127692, "epoch": 1434, "lr": 5.479400279634236e-05}
{"train_loss": 0.12197379022836685, "global_step": 127693, "epoch": 1434, "lr": 5.479342573032432e-05}
{"train_loss": 0.13729366660118103, "global_step": 127694, "epoch": 1434, "lr": 5.479284866366188e-05}
{"train_loss": 0.10956291854381561, "global_step": 127695, "epoch": 1434, "lr": 5.479227159635509e-05}
{"train_loss": 0.13768431544303894, "global_step": 127696, "epoch": 1434, "lr": 5.479169452840404e-05}
{"train_loss": 0.08968888968229294, "global_step": 127697, "epoch": 1434, "lr": 5.4791117459808804e-05}
{"train_loss": 0.11365987360477448, "global_step": 127698, "epoch": 1434, "lr": 5.4790540390569466e-05}
{"train_loss": 0.12829850614070892, "global_step": 127699, "epoch": 1434, "lr": 5.4789963320686085e-05}
{"train_loss": 0.12615032494068146, "global_step": 127700, "epoch": 1434, "lr": 5.478938625015877e-05}
{"train_loss": 0.11519043892621994, "global_step": 127701, "epoch": 1434, "lr": 5.4788809178987566e-05}
{"train_loss": 0.15766669809818268, "global_step": 127702, "epoch": 1434, "lr": 5.478823210717258e-05}
{"train_loss": 0.10136888176202774, "global_step": 127703, "epoch": 1434, "lr": 5.478765503471386e-05}
{"train_loss": 0.1425575613975525, "global_step": 127704, "epoch": 1434, "lr": 5.47870779616115e-05}
{"train_loss": 0.09930086135864258, "global_step": 127705, "epoch": 1434, "lr": 5.4786500887865575e-05}
{"train_loss": 0.2086944431066513, "global_step": 127706, "epoch": 1434, "lr": 5.478592381347618e-05}
{"train_loss": 0.11119779199361801, "global_step": 127707, "epoch": 1434, "lr": 5.478534673844337e-05}
{"train_loss": 0.17684368789196014, "global_step": 127708, "epoch": 1434, "lr": 5.478476966276721e-05}
{"train_loss": 0.13185466825962067, "global_step": 127709, "epoch": 1434, "lr": 5.4784192586447816e-05}
{"train_loss": 0.11792142689228058, "global_step": 127710, "epoch": 1434, "lr": 5.478361550948522e-05}
{"train_loss": 0.06439508497714996, "global_step": 127711, "epoch": 1434, "lr": 5.478303843187954e-05}
{"train_loss": 0.10717257857322693, "global_step": 127712, "epoch": 1434, "lr": 5.478246135363084e-05}
{"train_loss": 0.1492130309343338, "global_step": 127713, "epoch": 1434, "lr": 5.47818842747392e-05}
{"train_loss": 0.12691193115845156, "global_step": 127714, "epoch": 1434, "lr": 5.478130719520469e-05, "val_loss": 5.778502941131592}
{"train_loss": 0.14422062039375305, "global_step": 127715, "epoch": 1435, "lr": 5.4780730115027404e-05}
{"train_loss": 0.11596415191888809, "global_step": 127716, "epoch": 1435, "lr": 5.4780153034207384e-05}
{"train_loss": 0.11514104157686234, "global_step": 127717, "epoch": 1435, "lr": 5.4779575952744744e-05}
{"train_loss": 0.18660451471805573, "global_step": 127718, "epoch": 1435, "lr": 5.477899887063955e-05}
{"train_loss": 0.14507804811000824, "global_step": 127719, "epoch": 1435, "lr": 5.4778421787891875e-05}
{"train_loss": 0.1333676427602768, "global_step": 127720, "epoch": 1435, "lr": 5.47778447045018e-05}
{"train_loss": 0.09147606790065765, "global_step": 127721, "epoch": 1435, "lr": 5.47772676204694e-05}
{"train_loss": 0.1422749012708664, "global_step": 127722, "epoch": 1435, "lr": 5.477669053579475e-05}
{"train_loss": 0.10437879711389542, "global_step": 127723, "epoch": 1435, "lr": 5.477611345047794e-05}
{"train_loss": 0.09057439863681793, "global_step": 127724, "epoch": 1435, "lr": 5.477553636451904e-05}
{"train_loss": 0.1148313581943512, "global_step": 127725, "epoch": 1435, "lr": 5.477495927791813e-05}
{"train_loss": 0.08062632381916046, "global_step": 127726, "epoch": 1435, "lr": 5.477438219067529e-05}
{"train_loss": 0.11681105196475983, "global_step": 127727, "epoch": 1435, "lr": 5.477380510279057e-05}
{"train_loss": 0.18425074219703674, "global_step": 127728, "epoch": 1435, "lr": 5.477322801426408e-05}
{"train_loss": 0.08823270350694656, "global_step": 127729, "epoch": 1435, "lr": 5.477265092509589e-05}
{"train_loss": 0.10803627222776413, "global_step": 127730, "epoch": 1435, "lr": 5.4772073835286074e-05}
{"train_loss": 0.12282034754753113, "global_step": 127731, "epoch": 1435, "lr": 5.4771496744834714e-05}
{"train_loss": 0.12425271421670914, "global_step": 127732, "epoch": 1435, "lr": 5.477091965374188e-05}
{"train_loss": 0.08160942047834396, "global_step": 127733, "epoch": 1435, "lr": 5.4770342562007656e-05}
{"train_loss": 0.0863703265786171, "global_step": 127734, "epoch": 1435, "lr": 5.4769765469632115e-05}
{"train_loss": 0.10391269624233246, "global_step": 127735, "epoch": 1435, "lr": 5.476918837661534e-05}
{"train_loss": 0.10132647305727005, "global_step": 127736, "epoch": 1435, "lr": 5.476861128295741e-05}
{"train_loss": 0.08751613646745682, "global_step": 127737, "epoch": 1435, "lr": 5.476803418865839e-05}
{"train_loss": 0.0924135372042656, "global_step": 127738, "epoch": 1435, "lr": 5.476745709371838e-05}
{"train_loss": 0.11338108777999878, "global_step": 127739, "epoch": 1435, "lr": 5.476687999813743e-05}
{"train_loss": 0.1379302442073822, "global_step": 127740, "epoch": 1435, "lr": 5.476630290191563e-05}
{"train_loss": 0.11139122396707535, "global_step": 127741, "epoch": 1435, "lr": 5.476572580505307e-05}
{"train_loss": 0.13005048036575317, "global_step": 127742, "epoch": 1435, "lr": 5.476514870754981e-05}
{"train_loss": 0.10853138566017151, "global_step": 127743, "epoch": 1435, "lr": 5.476457160940593e-05}
{"train_loss": 0.10916091501712799, "global_step": 127744, "epoch": 1435, "lr": 5.4763994510621516e-05}
{"train_loss": 0.06325478851795197, "global_step": 127745, "epoch": 1435, "lr": 5.476341741119665e-05}
{"train_loss": 0.1755891740322113, "global_step": 127746, "epoch": 1435, "lr": 5.476284031113139e-05}
{"train_loss": 0.13666902482509613, "global_step": 127747, "epoch": 1435, "lr": 5.4762263210425815e-05}
{"train_loss": 0.13800953328609467, "global_step": 127748, "epoch": 1435, "lr": 5.4761686109080025e-05}
{"train_loss": 0.09518330544233322, "global_step": 127749, "epoch": 1435, "lr": 5.476110900709408e-05}
{"train_loss": 0.09608729928731918, "global_step": 127750, "epoch": 1435, "lr": 5.476053190446806e-05}
{"train_loss": 0.09481611847877502, "global_step": 127751, "epoch": 1435, "lr": 5.475995480120205e-05}
{"train_loss": 0.12473827600479126, "global_step": 127752, "epoch": 1435, "lr": 5.475937769729611e-05}
{"train_loss": 0.13008283078670502, "global_step": 127753, "epoch": 1435, "lr": 5.4758800592750345e-05}
{"train_loss": 0.09833016246557236, "global_step": 127754, "epoch": 1435, "lr": 5.4758223487564805e-05}
{"train_loss": 0.1908702254295349, "global_step": 127755, "epoch": 1435, "lr": 5.4757646381739594e-05}
{"train_loss": 0.1274251937866211, "global_step": 127756, "epoch": 1435, "lr": 5.4757069275274764e-05}
{"train_loss": 0.13755899667739868, "global_step": 127757, "epoch": 1435, "lr": 5.475649216817042e-05}
{"train_loss": 0.1641872525215149, "global_step": 127758, "epoch": 1435, "lr": 5.47559150604266e-05}
{"train_loss": 0.19804182648658752, "global_step": 127759, "epoch": 1435, "lr": 5.475533795204342e-05}
{"train_loss": 0.15618962049484253, "global_step": 127760, "epoch": 1435, "lr": 5.4754760843020935e-05}
{"train_loss": 0.22742249071598053, "global_step": 127761, "epoch": 1435, "lr": 5.4754183733359234e-05}
{"train_loss": 0.058146752417087555, "global_step": 127762, "epoch": 1435, "lr": 5.4753606623058395e-05}
{"train_loss": 0.14332346618175507, "global_step": 127763, "epoch": 1435, "lr": 5.475302951211849e-05}
{"train_loss": 0.13500045239925385, "global_step": 127764, "epoch": 1435, "lr": 5.47524524005396e-05}
{"train_loss": 0.21747411787509918, "global_step": 127765, "epoch": 1435, "lr": 5.475187528832179e-05}
{"train_loss": 0.09110138565301895, "global_step": 127766, "epoch": 1435, "lr": 5.475129817546516e-05}
{"train_loss": 0.0838012844324112, "global_step": 127767, "epoch": 1435, "lr": 5.4750721061969765e-05}
{"train_loss": 0.1181580051779747, "global_step": 127768, "epoch": 1435, "lr": 5.4750143947835696e-05}
{"train_loss": 0.100580133497715, "global_step": 127769, "epoch": 1435, "lr": 5.474956683306304e-05}
{"train_loss": 0.05355112627148628, "global_step": 127770, "epoch": 1435, "lr": 5.474898971765185e-05}
{"train_loss": 0.16530415415763855, "global_step": 127771, "epoch": 1435, "lr": 5.4748412601602215e-05}
{"train_loss": 0.1574697196483612, "global_step": 127772, "epoch": 1435, "lr": 5.474783548491422e-05}
{"train_loss": 0.061916522681713104, "global_step": 127773, "epoch": 1435, "lr": 5.474725836758794e-05}
{"train_loss": 0.11556173861026764, "global_step": 127774, "epoch": 1435, "lr": 5.4746681249623446e-05}
{"train_loss": 0.10225199908018112, "global_step": 127775, "epoch": 1435, "lr": 5.4746104131020817e-05}
{"train_loss": 0.1019841656088829, "global_step": 127776, "epoch": 1435, "lr": 5.474552701178013e-05}
{"train_loss": 0.16622787714004517, "global_step": 127777, "epoch": 1435, "lr": 5.474494989190147e-05}
{"train_loss": 0.1575106829404831, "global_step": 127778, "epoch": 1435, "lr": 5.47443727713849e-05}
{"train_loss": 0.149812713265419, "global_step": 127779, "epoch": 1435, "lr": 5.474379565023051e-05}
{"train_loss": 0.12379453331232071, "global_step": 127780, "epoch": 1435, "lr": 5.474321852843839e-05}
{"train_loss": 0.13913869857788086, "global_step": 127781, "epoch": 1435, "lr": 5.4742641406008586e-05}
{"train_loss": 0.13694748282432556, "global_step": 127782, "epoch": 1435, "lr": 5.47420642829412e-05}
{"train_loss": 0.19928976893424988, "global_step": 127783, "epoch": 1435, "lr": 5.4741487159236294e-05}
{"train_loss": 0.14045259356498718, "global_step": 127784, "epoch": 1435, "lr": 5.474091003489395e-05}
{"train_loss": 0.1418202519416809, "global_step": 127785, "epoch": 1435, "lr": 5.4740332909914264e-05}
{"train_loss": 0.10273748636245728, "global_step": 127786, "epoch": 1435, "lr": 5.473975578429727e-05}
{"train_loss": 0.19529840350151062, "global_step": 127787, "epoch": 1435, "lr": 5.47391786580431e-05}
{"train_loss": 0.11669377982616425, "global_step": 127788, "epoch": 1435, "lr": 5.4738601531151804e-05}
{"train_loss": 0.12257646024227142, "global_step": 127789, "epoch": 1435, "lr": 5.4738024403623446e-05}
{"train_loss": 0.12462706118822098, "global_step": 127790, "epoch": 1435, "lr": 5.4737447275458134e-05}
{"train_loss": 0.1192067340016365, "global_step": 127791, "epoch": 1435, "lr": 5.4736870146655914e-05}
{"train_loss": 0.11528069525957108, "global_step": 127792, "epoch": 1435, "lr": 5.473629301721689e-05}
{"train_loss": 0.15727443993091583, "global_step": 127793, "epoch": 1435, "lr": 5.473571588714113e-05}
{"train_loss": 0.10899555683135986, "global_step": 127794, "epoch": 1435, "lr": 5.473513875642871e-05}
{"train_loss": 0.15879960358142853, "global_step": 127795, "epoch": 1435, "lr": 5.4734561625079696e-05}
{"train_loss": 0.1773984581232071, "global_step": 127796, "epoch": 1435, "lr": 5.473398449309419e-05}
{"train_loss": 0.15928959846496582, "global_step": 127797, "epoch": 1435, "lr": 5.473340736047226e-05}
{"train_loss": 0.14148014783859253, "global_step": 127798, "epoch": 1435, "lr": 5.473283022721397e-05}
{"train_loss": 0.17937830090522766, "global_step": 127799, "epoch": 1435, "lr": 5.473225309331942e-05}
{"train_loss": 0.09968236833810806, "global_step": 127800, "epoch": 1435, "lr": 5.4731675958788666e-05}
{"train_loss": 0.1145162358880043, "global_step": 127801, "epoch": 1435, "lr": 5.473109882362179e-05}
{"train_loss": 0.1347273290157318, "global_step": 127802, "epoch": 1435, "lr": 5.473052168781889e-05}
{"train_loss": 0.1267794928607646, "global_step": 127803, "epoch": 1435, "lr": 5.472994455138003e-05, "val_loss": 5.332674503326416, "train_action_mse_error": 25.194610595703125}
{"train_loss": 0.15789730846881866, "global_step": 127804, "epoch": 1436, "lr": 5.472936741430529e-05}
{"train_loss": 0.13321270048618317, "global_step": 127805, "epoch": 1436, "lr": 5.4728790276594736e-05}
{"train_loss": 0.11000905185937881, "global_step": 127806, "epoch": 1436, "lr": 5.472821313824845e-05}
{"train_loss": 0.12193607538938522, "global_step": 127807, "epoch": 1436, "lr": 5.4727635999266516e-05}
{"train_loss": 0.1505212038755417, "global_step": 127808, "epoch": 1436, "lr": 5.472705885964901e-05}
{"train_loss": 0.13174857199192047, "global_step": 127809, "epoch": 1436, "lr": 5.4726481719396005e-05}
{"train_loss": 0.14963175356388092, "global_step": 127810, "epoch": 1436, "lr": 5.4725904578507594e-05}
{"train_loss": 0.07939735054969788, "global_step": 127811, "epoch": 1436, "lr": 5.472532743698383e-05}
{"train_loss": 0.14677050709724426, "global_step": 127812, "epoch": 1436, "lr": 5.472475029482481e-05}
{"train_loss": 0.19616764783859253, "global_step": 127813, "epoch": 1436, "lr": 5.47241731520306e-05}
{"train_loss": 0.13775625824928284, "global_step": 127814, "epoch": 1436, "lr": 5.472359600860128e-05}
{"train_loss": 0.1692771464586258, "global_step": 127815, "epoch": 1436, "lr": 5.4723018864536944e-05}
{"train_loss": 0.1776677966117859, "global_step": 127816, "epoch": 1436, "lr": 5.472244171983765e-05}
{"train_loss": 0.12525352835655212, "global_step": 127817, "epoch": 1436, "lr": 5.472186457450347e-05}
{"train_loss": 0.12119975686073303, "global_step": 127818, "epoch": 1436, "lr": 5.4721287428534505e-05}
{"train_loss": 0.09706131368875504, "global_step": 127819, "epoch": 1436, "lr": 5.4720710281930824e-05}
{"train_loss": 0.11088389903306961, "global_step": 127820, "epoch": 1436, "lr": 5.4720133134692484e-05}
{"train_loss": 0.1260467767715454, "global_step": 127821, "epoch": 1436, "lr": 5.4719555986819595e-05}
{"train_loss": 0.12854968011379242, "global_step": 127822, "epoch": 1436, "lr": 5.471897883831221e-05}
{"train_loss": 0.1690395474433899, "global_step": 127823, "epoch": 1436, "lr": 5.471840168917042e-05}
{"train_loss": 0.08681435883045197, "global_step": 127824, "epoch": 1436, "lr": 5.471782453939429e-05}
{"train_loss": 0.17723102867603302, "global_step": 127825, "epoch": 1436, "lr": 5.4717247388983917e-05}
{"train_loss": 0.10963209718465805, "global_step": 127826, "epoch": 1436, "lr": 5.471667023793937e-05}
{"train_loss": 0.1172601506114006, "global_step": 127827, "epoch": 1436, "lr": 5.4716093086260714e-05}
{"train_loss": 0.07071578502655029, "global_step": 127828, "epoch": 1436, "lr": 5.4715515933948044e-05}
{"train_loss": 0.1405954360961914, "global_step": 127829, "epoch": 1436, "lr": 5.471493878100142e-05}
{"train_loss": 0.09029782563447952, "global_step": 127830, "epoch": 1436, "lr": 5.471436162742094e-05}
{"train_loss": 0.1260181963443756, "global_step": 127831, "epoch": 1436, "lr": 5.471378447320667e-05}
{"train_loss": 0.10177496075630188, "global_step": 127832, "epoch": 1436, "lr": 5.47132073183587e-05}
{"train_loss": 0.156087264418602, "global_step": 127833, "epoch": 1436, "lr": 5.4712630162877074e-05}
{"train_loss": 0.10894525051116943, "global_step": 127834, "epoch": 1436, "lr": 5.471205300676191e-05}
{"train_loss": 0.11559352278709412, "global_step": 127835, "epoch": 1436, "lr": 5.471147585001326e-05}
{"train_loss": 0.1140236034989357, "global_step": 127836, "epoch": 1436, "lr": 5.471089869263121e-05}
{"train_loss": 0.15241169929504395, "global_step": 127837, "epoch": 1436, "lr": 5.471032153461584e-05}
{"train_loss": 0.1258794665336609, "global_step": 127838, "epoch": 1436, "lr": 5.470974437596722e-05}
{"train_loss": 0.13227027654647827, "global_step": 127839, "epoch": 1436, "lr": 5.470916721668543e-05}
{"train_loss": 0.1768067479133606, "global_step": 127840, "epoch": 1436, "lr": 5.470859005677056e-05}
{"train_loss": 0.06022360175848007, "global_step": 127841, "epoch": 1436, "lr": 5.470801289622267e-05}
{"train_loss": 0.11906702816486359, "global_step": 127842, "epoch": 1436, "lr": 5.470743573504184e-05}
{"train_loss": 0.09138555824756622, "global_step": 127843, "epoch": 1436, "lr": 5.470685857322817e-05}
{"train_loss": 0.14446806907653809, "global_step": 127844, "epoch": 1436, "lr": 5.47062814107817e-05}
{"train_loss": 0.13125616312026978, "global_step": 127845, "epoch": 1436, "lr": 5.470570424770254e-05}
{"train_loss": 0.11688905209302902, "global_step": 127846, "epoch": 1436, "lr": 5.470512708399076e-05}
{"train_loss": 0.1492961198091507, "global_step": 127847, "epoch": 1436, "lr": 5.470454991964642e-05}
{"train_loss": 0.11101086437702179, "global_step": 127848, "epoch": 1436, "lr": 5.470397275466962e-05}
{"train_loss": 0.11805108189582825, "global_step": 127849, "epoch": 1436, "lr": 5.4703395589060425e-05}
{"train_loss": 0.15110458433628082, "global_step": 127850, "epoch": 1436, "lr": 5.4702818422818915e-05}
{"train_loss": 0.09966112673282623, "global_step": 127851, "epoch": 1436, "lr": 5.4702241255945166e-05}
{"train_loss": 0.12049276381731033, "global_step": 127852, "epoch": 1436, "lr": 5.4701664088439266e-05}
{"train_loss": 0.1531999409198761, "global_step": 127853, "epoch": 1436, "lr": 5.470108692030128e-05}
{"train_loss": 0.12668973207473755, "global_step": 127854, "epoch": 1436, "lr": 5.470050975153129e-05}
{"train_loss": 0.0823988988995552, "global_step": 127855, "epoch": 1436, "lr": 5.4699932582129365e-05}
{"train_loss": 0.14085699617862701, "global_step": 127856, "epoch": 1436, "lr": 5.469935541209561e-05}
{"train_loss": 0.1508973091840744, "global_step": 127857, "epoch": 1436, "lr": 5.469877824143007e-05}
{"train_loss": 0.14977233111858368, "global_step": 127858, "epoch": 1436, "lr": 5.469820107013284e-05}
{"train_loss": 0.09051927924156189, "global_step": 127859, "epoch": 1436, "lr": 5.4697623898203996e-05}
{"train_loss": 0.09152650088071823, "global_step": 127860, "epoch": 1436, "lr": 5.4697046725643606e-05}
{"train_loss": 0.11730965226888657, "global_step": 127861, "epoch": 1436, "lr": 5.469646955245177e-05}
{"train_loss": 0.12089171260595322, "global_step": 127862, "epoch": 1436, "lr": 5.469589237862854e-05}
{"train_loss": 0.11893776059150696, "global_step": 127863, "epoch": 1436, "lr": 5.4695315204174004e-05}
{"train_loss": 0.09364205598831177, "global_step": 127864, "epoch": 1436, "lr": 5.4694738029088244e-05}
{"train_loss": 0.13561215996742249, "global_step": 127865, "epoch": 1436, "lr": 5.4694160853371335e-05}
{"train_loss": 0.1933988332748413, "global_step": 127866, "epoch": 1436, "lr": 5.469358367702335e-05}
{"train_loss": 0.1544303148984909, "global_step": 127867, "epoch": 1436, "lr": 5.469300650004437e-05}
{"train_loss": 0.07664184272289276, "global_step": 127868, "epoch": 1436, "lr": 5.469242932243448e-05}
{"train_loss": 0.13970443606376648, "global_step": 127869, "epoch": 1436, "lr": 5.469185214419373e-05}
{"train_loss": 0.1466555893421173, "global_step": 127870, "epoch": 1436, "lr": 5.4691274965322235e-05}
{"train_loss": 0.14428678154945374, "global_step": 127871, "epoch": 1436, "lr": 5.469069778582006e-05}
{"train_loss": 0.11729927361011505, "global_step": 127872, "epoch": 1436, "lr": 5.469012060568727e-05}
{"train_loss": 0.11111048609018326, "global_step": 127873, "epoch": 1436, "lr": 5.468954342492395e-05}
{"train_loss": 0.08843604475259781, "global_step": 127874, "epoch": 1436, "lr": 5.468896624353016e-05}
{"train_loss": 0.12146096676588058, "global_step": 127875, "epoch": 1436, "lr": 5.468838906150603e-05}
{"train_loss": 0.1400916874408722, "global_step": 127876, "epoch": 1436, "lr": 5.468781187885158e-05}
{"train_loss": 0.07568857073783875, "global_step": 127877, "epoch": 1436, "lr": 5.468723469556692e-05}
{"train_loss": 0.1474166363477707, "global_step": 127878, "epoch": 1436, "lr": 5.468665751165212e-05}
{"train_loss": 0.11032691597938538, "global_step": 127879, "epoch": 1436, "lr": 5.468608032710725e-05}
{"train_loss": 0.20180967450141907, "global_step": 127880, "epoch": 1436, "lr": 5.468550314193239e-05}
{"train_loss": 0.15038231015205383, "global_step": 127881, "epoch": 1436, "lr": 5.468492595612763e-05}
{"train_loss": 0.11789826303720474, "global_step": 127882, "epoch": 1436, "lr": 5.468434876969304e-05}
{"train_loss": 0.13451632857322693, "global_step": 127883, "epoch": 1436, "lr": 5.4683771582628686e-05}
{"train_loss": 0.091428242623806, "global_step": 127884, "epoch": 1436, "lr": 5.468319439493467e-05}
{"train_loss": 0.1258811354637146, "global_step": 127885, "epoch": 1436, "lr": 5.4682617206611045e-05}
{"train_loss": 0.08155867457389832, "global_step": 127886, "epoch": 1436, "lr": 5.46820400176579e-05}
{"train_loss": 0.12320275604724884, "global_step": 127887, "epoch": 1436, "lr": 5.468146282807531e-05}
{"train_loss": 0.0663226768374443, "global_step": 127888, "epoch": 1436, "lr": 5.468088563786336e-05}
{"train_loss": 0.15136195719242096, "global_step": 127889, "epoch": 1436, "lr": 5.468030844702212e-05}
{"train_loss": 0.07787741720676422, "global_step": 127890, "epoch": 1436, "lr": 5.467973125555167e-05}
{"train_loss": 0.18029822409152985, "global_step": 127891, "epoch": 1436, "lr": 5.467915406345209e-05}
{"train_loss": 0.1260285634505615, "global_step": 127892, "epoch": 1436, "lr": 5.467857687072344e-05, "val_loss": 5.7920684814453125}
{"train_loss": 0.1168961375951767, "global_step": 127893, "epoch": 1437, "lr": 5.467799967736582e-05}
{"train_loss": 0.08763676881790161, "global_step": 127894, "epoch": 1437, "lr": 5.467742248337931e-05}
{"train_loss": 0.1311333328485489, "global_step": 127895, "epoch": 1437, "lr": 5.4676845288763975e-05}
{"train_loss": 0.10648910701274872, "global_step": 127896, "epoch": 1437, "lr": 5.467626809351989e-05}
{"train_loss": 0.15326723456382751, "global_step": 127897, "epoch": 1437, "lr": 5.4675690897647145e-05}
{"train_loss": 0.12337355315685272, "global_step": 127898, "epoch": 1437, "lr": 5.4675113701145796e-05}
{"train_loss": 0.13489305973052979, "global_step": 127899, "epoch": 1437, "lr": 5.467453650401595e-05}
{"train_loss": 0.20543433725833893, "global_step": 127900, "epoch": 1437, "lr": 5.467395930625766e-05}
{"train_loss": 0.08448635786771774, "global_step": 127901, "epoch": 1437, "lr": 5.4673382107871016e-05}
{"train_loss": 0.11123476922512054, "global_step": 127902, "epoch": 1437, "lr": 5.467280490885609e-05}
{"train_loss": 0.10792790353298187, "global_step": 127903, "epoch": 1437, "lr": 5.467222770921298e-05}
{"train_loss": 0.12195312231779099, "global_step": 127904, "epoch": 1437, "lr": 5.467165050894172e-05}
{"train_loss": 0.1009051576256752, "global_step": 127905, "epoch": 1437, "lr": 5.467107330804243e-05}
{"train_loss": 0.08790288865566254, "global_step": 127906, "epoch": 1437, "lr": 5.4670496106515166e-05}
{"train_loss": 0.09881056845188141, "global_step": 127907, "epoch": 1437, "lr": 5.466991890436001e-05}
{"train_loss": 0.08887340128421783, "global_step": 127908, "epoch": 1437, "lr": 5.4669341701577035e-05}
{"train_loss": 0.1018475592136383, "global_step": 127909, "epoch": 1437, "lr": 5.466876449816634e-05}
{"train_loss": 0.12099210172891617, "global_step": 127910, "epoch": 1437, "lr": 5.466818729412796e-05}
{"train_loss": 0.052636947482824326, "global_step": 127911, "epoch": 1437, "lr": 5.466761008946202e-05}
{"train_loss": 0.13378584384918213, "global_step": 127912, "epoch": 1437, "lr": 5.466703288416857e-05}
{"train_loss": 0.0842561349272728, "global_step": 127913, "epoch": 1437, "lr": 5.46664556782477e-05}
{"train_loss": 0.17059503495693207, "global_step": 127914, "epoch": 1437, "lr": 5.466587847169948e-05}
{"train_loss": 0.10560210049152374, "global_step": 127915, "epoch": 1437, "lr": 5.4665301264523994e-05}
{"train_loss": 0.13563129305839539, "global_step": 127916, "epoch": 1437, "lr": 5.466472405672131e-05}
{"train_loss": 0.1419874131679535, "global_step": 127917, "epoch": 1437, "lr": 5.4664146848291506e-05}
{"train_loss": 0.1417844444513321, "global_step": 127918, "epoch": 1437, "lr": 5.466356963923467e-05}
{"train_loss": 0.07064660638570786, "global_step": 127919, "epoch": 1437, "lr": 5.4662992429550865e-05}
{"train_loss": 0.15142755210399628, "global_step": 127920, "epoch": 1437, "lr": 5.4662415219240195e-05}
{"train_loss": 0.07899203896522522, "global_step": 127921, "epoch": 1437, "lr": 5.46618380083027e-05}
{"train_loss": 0.19706673920154572, "global_step": 127922, "epoch": 1437, "lr": 5.466126079673849e-05}
{"train_loss": 0.18176612257957458, "global_step": 127923, "epoch": 1437, "lr": 5.4660683584547625e-05}
{"train_loss": 0.15345127880573273, "global_step": 127924, "epoch": 1437, "lr": 5.466010637173019e-05}
{"train_loss": 0.09385722130537033, "global_step": 127925, "epoch": 1437, "lr": 5.465952915828626e-05}
{"train_loss": 0.18408307433128357, "global_step": 127926, "epoch": 1437, "lr": 5.4658951944215916e-05}
{"train_loss": 0.11762885004281998, "global_step": 127927, "epoch": 1437, "lr": 5.4658374729519236e-05}
{"train_loss": 0.07038794457912445, "global_step": 127928, "epoch": 1437, "lr": 5.465779751419627e-05}
{"train_loss": 0.1210215613245964, "global_step": 127929, "epoch": 1437, "lr": 5.465722029824715e-05}
{"train_loss": 0.1521095484495163, "global_step": 127930, "epoch": 1437, "lr": 5.465664308167191e-05}
{"train_loss": 0.1335005760192871, "global_step": 127931, "epoch": 1437, "lr": 5.465606586447064e-05}
{"train_loss": 0.10301890224218369, "global_step": 127932, "epoch": 1437, "lr": 5.4655488646643417e-05}
{"train_loss": 0.16185468435287476, "global_step": 127933, "epoch": 1437, "lr": 5.4654911428190335e-05}
{"train_loss": 0.10777799040079117, "global_step": 127934, "epoch": 1437, "lr": 5.465433420911146e-05}
{"train_loss": 0.12058304995298386, "global_step": 127935, "epoch": 1437, "lr": 5.465375698940685e-05}
{"train_loss": 0.08340362459421158, "global_step": 127936, "epoch": 1437, "lr": 5.46531797690766e-05}
{"train_loss": 0.1649077832698822, "global_step": 127937, "epoch": 1437, "lr": 5.465260254812079e-05}
{"train_loss": 0.11759072542190552, "global_step": 127938, "epoch": 1437, "lr": 5.465202532653949e-05}
{"train_loss": 0.11503568291664124, "global_step": 127939, "epoch": 1437, "lr": 5.465144810433279e-05}
{"train_loss": 0.2096230834722519, "global_step": 127940, "epoch": 1437, "lr": 5.465087088150076e-05}
{"train_loss": 0.05919585004448891, "global_step": 127941, "epoch": 1437, "lr": 5.465029365804346e-05}
{"train_loss": 0.0849764421582222, "global_step": 127942, "epoch": 1437, "lr": 5.464971643396101e-05}
{"train_loss": 0.1365538090467453, "global_step": 127943, "epoch": 1437, "lr": 5.4649139209253443e-05}
{"train_loss": 0.17899121344089508, "global_step": 127944, "epoch": 1437, "lr": 5.464856198392087e-05}
{"train_loss": 0.13833335041999817, "global_step": 127945, "epoch": 1437, "lr": 5.4647984757963346e-05}
{"train_loss": 0.1038740798830986, "global_step": 127946, "epoch": 1437, "lr": 5.464740753138094e-05}
{"train_loss": 0.1007738709449768, "global_step": 127947, "epoch": 1437, "lr": 5.464683030417377e-05}
{"train_loss": 0.09336879104375839, "global_step": 127948, "epoch": 1437, "lr": 5.464625307634188e-05}
{"train_loss": 0.16586874425411224, "global_step": 127949, "epoch": 1437, "lr": 5.4645675847885356e-05}
{"train_loss": 0.17729489505290985, "global_step": 127950, "epoch": 1437, "lr": 5.4645098618804295e-05}
{"train_loss": 0.06060253828763962, "global_step": 127951, "epoch": 1437, "lr": 5.4644521389098746e-05}
{"train_loss": 0.17626240849494934, "global_step": 127952, "epoch": 1437, "lr": 5.464394415876879e-05}
{"train_loss": 0.1686103194952011, "global_step": 127953, "epoch": 1437, "lr": 5.464336692781453e-05}
{"train_loss": 0.11234109848737717, "global_step": 127954, "epoch": 1437, "lr": 5.464278969623602e-05}
{"train_loss": 0.12244878709316254, "global_step": 127955, "epoch": 1437, "lr": 5.4642212464033335e-05}
{"train_loss": 0.1525341272354126, "global_step": 127956, "epoch": 1437, "lr": 5.464163523120658e-05}
{"train_loss": 0.08929481357336044, "global_step": 127957, "epoch": 1437, "lr": 5.4641057997755796e-05}
{"train_loss": 0.15571950376033783, "global_step": 127958, "epoch": 1437, "lr": 5.464048076368108e-05}
{"train_loss": 0.16411352157592773, "global_step": 127959, "epoch": 1437, "lr": 5.4639903528982504e-05}
{"train_loss": 0.1299600750207901, "global_step": 127960, "epoch": 1437, "lr": 5.4639326293660165e-05}
{"train_loss": 0.0916619747877121, "global_step": 127961, "epoch": 1437, "lr": 5.463874905771411e-05}
{"train_loss": 0.19609104096889496, "global_step": 127962, "epoch": 1437, "lr": 5.463817182114444e-05}
{"train_loss": 0.12441767007112503, "global_step": 127963, "epoch": 1437, "lr": 5.4637594583951226e-05}
{"train_loss": 0.1394861489534378, "global_step": 127964, "epoch": 1437, "lr": 5.4637017346134524e-05}
{"train_loss": 0.100017249584198, "global_step": 127965, "epoch": 1437, "lr": 5.463644010769444e-05}
{"train_loss": 0.164910688996315, "global_step": 127966, "epoch": 1437, "lr": 5.463586286863106e-05}
{"train_loss": 0.1685589998960495, "global_step": 127967, "epoch": 1437, "lr": 5.4635285628944434e-05}
{"train_loss": 0.11141393333673477, "global_step": 127968, "epoch": 1437, "lr": 5.463470838863465e-05}
{"train_loss": 0.1693941354751587, "global_step": 127969, "epoch": 1437, "lr": 5.463413114770178e-05}
{"train_loss": 0.16231496632099152, "global_step": 127970, "epoch": 1437, "lr": 5.4633553906145916e-05}
{"train_loss": 0.15100951492786407, "global_step": 127971, "epoch": 1437, "lr": 5.463297666396713e-05}
{"train_loss": 0.12576919794082642, "global_step": 127972, "epoch": 1437, "lr": 5.46323994211655e-05}
{"train_loss": 0.23448942601680756, "global_step": 127973, "epoch": 1437, "lr": 5.46318221777411e-05}
{"train_loss": 0.13318301737308502, "global_step": 127974, "epoch": 1437, "lr": 5.4631244933694006e-05}
{"train_loss": 0.0793457105755806, "global_step": 127975, "epoch": 1437, "lr": 5.4630667689024286e-05}
{"train_loss": 0.1062823161482811, "global_step": 127976, "epoch": 1437, "lr": 5.463009044373204e-05}
{"train_loss": 0.13801522552967072, "global_step": 127977, "epoch": 1437, "lr": 5.4629513197817326e-05}
{"train_loss": 0.13901160657405853, "global_step": 127978, "epoch": 1437, "lr": 5.462893595128024e-05}
{"train_loss": 0.1293427050113678, "global_step": 127979, "epoch": 1437, "lr": 5.462835870412084e-05}
{"train_loss": 0.09371110796928406, "global_step": 127980, "epoch": 1437, "lr": 5.4627781456339224e-05}
{"train_loss": 0.1287987068965194, "global_step": 127981, "epoch": 1437, "lr": 5.4627204207935444e-05, "val_loss": 5.579408168792725}
{"train_loss": 0.12507344782352448, "global_step": 127982, "epoch": 1438, "lr": 5.46266269589096e-05}
{"train_loss": 0.1498529613018036, "global_step": 127983, "epoch": 1438, "lr": 5.462604970926176e-05}
{"train_loss": 0.1083647757768631, "global_step": 127984, "epoch": 1438, "lr": 5.462547245899201e-05}
{"train_loss": 0.11226898431777954, "global_step": 127985, "epoch": 1438, "lr": 5.462489520810042e-05}
{"train_loss": 0.1328231692314148, "global_step": 127986, "epoch": 1438, "lr": 5.462431795658707e-05}
{"train_loss": 0.16387267410755157, "global_step": 127987, "epoch": 1438, "lr": 5.462374070445203e-05}
{"train_loss": 0.14829418063163757, "global_step": 127988, "epoch": 1438, "lr": 5.462316345169539e-05}
{"train_loss": 0.13151833415031433, "global_step": 127989, "epoch": 1438, "lr": 5.462258619831723e-05}
{"train_loss": 0.06421619653701782, "global_step": 127990, "epoch": 1438, "lr": 5.462200894431762e-05}
{"train_loss": 0.12171553075313568, "global_step": 127991, "epoch": 1438, "lr": 5.4621431689696635e-05}
{"train_loss": 0.10081259161233902, "global_step": 127992, "epoch": 1438, "lr": 5.462085443445435e-05}
{"train_loss": 0.11508698761463165, "global_step": 127993, "epoch": 1438, "lr": 5.462027717859084e-05}
{"train_loss": 0.1050146147608757, "global_step": 127994, "epoch": 1438, "lr": 5.46196999221062e-05}
{"train_loss": 0.12099538743495941, "global_step": 127995, "epoch": 1438, "lr": 5.46191226650005e-05}
{"train_loss": 0.08303800970315933, "global_step": 127996, "epoch": 1438, "lr": 5.46185454072738e-05}
{"train_loss": 0.1129545271396637, "global_step": 127997, "epoch": 1438, "lr": 5.461796814892621e-05}
{"train_loss": 0.1848549097776413, "global_step": 127998, "epoch": 1438, "lr": 5.461739088995779e-05}
{"train_loss": 0.17410439252853394, "global_step": 127999, "epoch": 1438, "lr": 5.4616813630368605e-05}
{"train_loss": 0.11893140524625778, "global_step": 128000, "epoch": 1438, "lr": 5.4616236370158744e-05}
{"train_loss": 0.1912400722503662, "global_step": 128001, "epoch": 1438, "lr": 5.4615659109328285e-05}
{"train_loss": 0.07444077730178833, "global_step": 128002, "epoch": 1438, "lr": 5.4615081847877324e-05}
{"train_loss": 0.10577870905399323, "global_step": 128003, "epoch": 1438, "lr": 5.4614504585805914e-05}
{"train_loss": 0.18324074149131775, "global_step": 128004, "epoch": 1438, "lr": 5.461392732311413e-05}
{"train_loss": 0.09085626900196075, "global_step": 128005, "epoch": 1438, "lr": 5.461335005980208e-05}
{"train_loss": 0.10462114214897156, "global_step": 128006, "epoch": 1438, "lr": 5.46127727958698e-05}
{"train_loss": 0.14329837262630463, "global_step": 128007, "epoch": 1438, "lr": 5.461219553131741e-05}
{"train_loss": 0.13507482409477234, "global_step": 128008, "epoch": 1438, "lr": 5.461161826614496e-05}
{"train_loss": 0.11821617186069489, "global_step": 128009, "epoch": 1438, "lr": 5.461104100035252e-05}
{"train_loss": 0.05750572681427002, "global_step": 128010, "epoch": 1438, "lr": 5.46104637339402e-05}
{"train_loss": 0.08393562585115433, "global_step": 128011, "epoch": 1438, "lr": 5.4609886466908056e-05}
{"train_loss": 0.1266520917415619, "global_step": 128012, "epoch": 1438, "lr": 5.460930919925616e-05}
{"train_loss": 0.11585214734077454, "global_step": 128013, "epoch": 1438, "lr": 5.4608731930984615e-05}
{"train_loss": 0.11635155230760574, "global_step": 128014, "epoch": 1438, "lr": 5.460815466209347e-05}
{"train_loss": 0.11033999174833298, "global_step": 128015, "epoch": 1438, "lr": 5.460757739258281e-05}
{"train_loss": 0.09120230376720428, "global_step": 128016, "epoch": 1438, "lr": 5.460700012245272e-05}
{"train_loss": 0.16126134991645813, "global_step": 128017, "epoch": 1438, "lr": 5.4606422851703277e-05}
{"train_loss": 0.045965999364852905, "global_step": 128018, "epoch": 1438, "lr": 5.4605845580334556e-05}
{"train_loss": 0.14130522310733795, "global_step": 128019, "epoch": 1438, "lr": 5.460526830834663e-05}
{"train_loss": 0.08348490297794342, "global_step": 128020, "epoch": 1438, "lr": 5.4604691035739584e-05}
{"train_loss": 0.10950952023267746, "global_step": 128021, "epoch": 1438, "lr": 5.4604113762513496e-05}
{"train_loss": 0.16595521569252014, "global_step": 128022, "epoch": 1438, "lr": 5.460353648866845e-05}
{"train_loss": 0.13237658143043518, "global_step": 128023, "epoch": 1438, "lr": 5.460295921420451e-05}
{"train_loss": 0.09704212844371796, "global_step": 128024, "epoch": 1438, "lr": 5.460238193912175e-05}
{"train_loss": 0.12687203288078308, "global_step": 128025, "epoch": 1438, "lr": 5.460180466342027e-05}
{"train_loss": 0.15534113347530365, "global_step": 128026, "epoch": 1438, "lr": 5.460122738710012e-05}
{"train_loss": 0.11998613178730011, "global_step": 128027, "epoch": 1438, "lr": 5.460065011016139e-05}
{"train_loss": 0.1712028831243515, "global_step": 128028, "epoch": 1438, "lr": 5.4600072832604164e-05}
{"train_loss": 0.19974486529827118, "global_step": 128029, "epoch": 1438, "lr": 5.459949555442852e-05}
{"train_loss": 0.15566734969615936, "global_step": 128030, "epoch": 1438, "lr": 5.4598918275634536e-05}
{"train_loss": 0.19496119022369385, "global_step": 128031, "epoch": 1438, "lr": 5.459834099622226e-05}
{"train_loss": 0.14874505996704102, "global_step": 128032, "epoch": 1438, "lr": 5.459776371619181e-05}
{"train_loss": 0.20485210418701172, "global_step": 128033, "epoch": 1438, "lr": 5.4597186435543245e-05}
{"train_loss": 0.10865393280982971, "global_step": 128034, "epoch": 1438, "lr": 5.4596609154276646e-05}
{"train_loss": 0.1160946637392044, "global_step": 128035, "epoch": 1438, "lr": 5.4596031872392086e-05}
{"train_loss": 0.12368680536746979, "global_step": 128036, "epoch": 1438, "lr": 5.459545458988964e-05}
{"train_loss": 0.1569678634405136, "global_step": 128037, "epoch": 1438, "lr": 5.459487730676938e-05}
{"train_loss": 0.19687597453594208, "global_step": 128038, "epoch": 1438, "lr": 5.459430002303141e-05}
{"train_loss": 0.08869227021932602, "global_step": 128039, "epoch": 1438, "lr": 5.459372273867579e-05}
{"train_loss": 0.11064447462558746, "global_step": 128040, "epoch": 1438, "lr": 5.45931454537026e-05}
{"train_loss": 0.07544400542974472, "global_step": 128041, "epoch": 1438, "lr": 5.4592568168111926e-05}
{"train_loss": 0.13672344386577606, "global_step": 128042, "epoch": 1438, "lr": 5.459199088190383e-05}
{"train_loss": 0.14590179920196533, "global_step": 128043, "epoch": 1438, "lr": 5.45914135950784e-05}
{"train_loss": 0.08828872442245483, "global_step": 128044, "epoch": 1438, "lr": 5.459083630763571e-05}
{"train_loss": 0.128083735704422, "global_step": 128045, "epoch": 1438, "lr": 5.459025901957583e-05}
{"train_loss": 0.11826703697443008, "global_step": 128046, "epoch": 1438, "lr": 5.4589681730898847e-05}
{"train_loss": 0.06423931568861008, "global_step": 128047, "epoch": 1438, "lr": 5.458910444160484e-05}
{"train_loss": 0.06590177863836288, "global_step": 128048, "epoch": 1438, "lr": 5.458852715169389e-05}
{"train_loss": 0.12475287914276123, "global_step": 128049, "epoch": 1438, "lr": 5.458794986116608e-05}
{"train_loss": 0.11173451691865921, "global_step": 128050, "epoch": 1438, "lr": 5.4587372570021456e-05}
{"train_loss": 0.1486654132604599, "global_step": 128051, "epoch": 1438, "lr": 5.458679527826012e-05}
{"train_loss": 0.13042482733726501, "global_step": 128052, "epoch": 1438, "lr": 5.4586217985882134e-05}
{"train_loss": 0.15360884368419647, "global_step": 128053, "epoch": 1438, "lr": 5.4585640692887605e-05}
{"train_loss": 0.06548100709915161, "global_step": 128054, "epoch": 1438, "lr": 5.458506339927658e-05}
{"train_loss": 0.09450123459100723, "global_step": 128055, "epoch": 1438, "lr": 5.458448610504915e-05}
{"train_loss": 0.1379590481519699, "global_step": 128056, "epoch": 1438, "lr": 5.45839088102054e-05}
{"train_loss": 0.22566573321819305, "global_step": 128057, "epoch": 1438, "lr": 5.458333151474539e-05}
{"train_loss": 0.1868780553340912, "global_step": 128058, "epoch": 1438, "lr": 5.4582754218669215e-05}
{"train_loss": 0.106960728764534, "global_step": 128059, "epoch": 1438, "lr": 5.458217692197693e-05}
{"train_loss": 0.09191031754016876, "global_step": 128060, "epoch": 1438, "lr": 5.458159962466864e-05}
{"train_loss": 0.07250045239925385, "global_step": 128061, "epoch": 1438, "lr": 5.458102232674442e-05}
{"train_loss": 0.16755712032318115, "global_step": 128062, "epoch": 1438, "lr": 5.458044502820433e-05}
{"train_loss": 0.1293741911649704, "global_step": 128063, "epoch": 1438, "lr": 5.457986772904845e-05}
{"train_loss": 0.13391850888729095, "global_step": 128064, "epoch": 1438, "lr": 5.457929042927686e-05}
{"train_loss": 0.09604333341121674, "global_step": 128065, "epoch": 1438, "lr": 5.4578713128889645e-05}
{"train_loss": 0.15902076661586761, "global_step": 128066, "epoch": 1438, "lr": 5.457813582788688e-05}
{"train_loss": 0.16140049695968628, "global_step": 128067, "epoch": 1438, "lr": 5.4577558526268634e-05}
{"train_loss": 0.18255534768104553, "global_step": 128068, "epoch": 1438, "lr": 5.4576981224034994e-05}
{"train_loss": 0.14320522546768188, "global_step": 128069, "epoch": 1438, "lr": 5.457640392118605e-05}
{"train_loss": 0.12796550009692653, "global_step": 128070, "epoch": 1438, "lr": 5.457582661772185e-05, "val_loss": 5.4317851066589355}
{"train_loss": 0.14599278569221497, "global_step": 128071, "epoch": 1439, "lr": 5.457524931364248e-05}
{"train_loss": 0.18581156432628632, "global_step": 128072, "epoch": 1439, "lr": 5.457467200894802e-05}
{"train_loss": 0.08446630835533142, "global_step": 128073, "epoch": 1439, "lr": 5.457409470363856e-05}
{"train_loss": 0.1105566993355751, "global_step": 128074, "epoch": 1439, "lr": 5.457351739771417e-05}
{"train_loss": 0.09339886158704758, "global_step": 128075, "epoch": 1439, "lr": 5.4572940091174914e-05}
{"train_loss": 0.10841835290193558, "global_step": 128076, "epoch": 1439, "lr": 5.45723627840209e-05}
{"train_loss": 0.12102896720170975, "global_step": 128077, "epoch": 1439, "lr": 5.457178547625218e-05}
{"train_loss": 0.14969035983085632, "global_step": 128078, "epoch": 1439, "lr": 5.457120816786884e-05}
{"train_loss": 0.22311252355575562, "global_step": 128079, "epoch": 1439, "lr": 5.4570630858870955e-05}
{"train_loss": 0.1238383799791336, "global_step": 128080, "epoch": 1439, "lr": 5.45700535492586e-05}
{"train_loss": 0.13013650476932526, "global_step": 128081, "epoch": 1439, "lr": 5.456947623903187e-05}
{"train_loss": 0.16534042358398438, "global_step": 128082, "epoch": 1439, "lr": 5.4568898928190826e-05}
{"train_loss": 0.08289631456136703, "global_step": 128083, "epoch": 1439, "lr": 5.456832161673554e-05}
{"train_loss": 0.17494946718215942, "global_step": 128084, "epoch": 1439, "lr": 5.45677443046661e-05}
{"train_loss": 0.11136943846940994, "global_step": 128085, "epoch": 1439, "lr": 5.456716699198259e-05}
{"train_loss": 0.13507062196731567, "global_step": 128086, "epoch": 1439, "lr": 5.456658967868508e-05}
{"train_loss": 0.13859833776950836, "global_step": 128087, "epoch": 1439, "lr": 5.4566012364773646e-05}
{"train_loss": 0.12487480044364929, "global_step": 128088, "epoch": 1439, "lr": 5.456543505024837e-05}
{"train_loss": 0.11981526762247086, "global_step": 128089, "epoch": 1439, "lr": 5.456485773510933e-05}
{"train_loss": 0.12939687073230743, "global_step": 128090, "epoch": 1439, "lr": 5.4564280419356594e-05}
{"train_loss": 0.14703382551670074, "global_step": 128091, "epoch": 1439, "lr": 5.456370310299023e-05}
{"train_loss": 0.14758366346359253, "global_step": 128092, "epoch": 1439, "lr": 5.456312578601035e-05}
{"train_loss": 0.11545919626951218, "global_step": 128093, "epoch": 1439, "lr": 5.456254846841701e-05}
{"train_loss": 0.13894912600517273, "global_step": 128094, "epoch": 1439, "lr": 5.4561971150210286e-05}
{"train_loss": 0.07757093012332916, "global_step": 128095, "epoch": 1439, "lr": 5.456139383139027e-05}
{"train_loss": 0.1081177145242691, "global_step": 128096, "epoch": 1439, "lr": 5.4560816511957025e-05}
{"train_loss": 0.11705338209867477, "global_step": 128097, "epoch": 1439, "lr": 5.4560239191910635e-05}
{"train_loss": 0.11870571970939636, "global_step": 128098, "epoch": 1439, "lr": 5.4559661871251176e-05}
{"train_loss": 0.1562793254852295, "global_step": 128099, "epoch": 1439, "lr": 5.4559084549978723e-05}
{"train_loss": 0.12758488953113556, "global_step": 128100, "epoch": 1439, "lr": 5.455850722809337e-05}
{"train_loss": 0.130435511469841, "global_step": 128101, "epoch": 1439, "lr": 5.455792990559516e-05}
{"train_loss": 0.1058574989438057, "global_step": 128102, "epoch": 1439, "lr": 5.455735258248421e-05}
{"train_loss": 0.15574084222316742, "global_step": 128103, "epoch": 1439, "lr": 5.4556775258760574e-05}
{"train_loss": 0.16266095638275146, "global_step": 128104, "epoch": 1439, "lr": 5.4556197934424335e-05}
{"train_loss": 0.07845500856637955, "global_step": 128105, "epoch": 1439, "lr": 5.455562060947557e-05}
{"train_loss": 0.1342809796333313, "global_step": 128106, "epoch": 1439, "lr": 5.455504328391435e-05}
{"train_loss": 0.1328549087047577, "global_step": 128107, "epoch": 1439, "lr": 5.455446595774079e-05}
{"train_loss": 0.1463136225938797, "global_step": 128108, "epoch": 1439, "lr": 5.455388863095491e-05}
{"train_loss": 0.1312674582004547, "global_step": 128109, "epoch": 1439, "lr": 5.455331130355681e-05}
{"train_loss": 0.06274832040071487, "global_step": 128110, "epoch": 1439, "lr": 5.4552733975546576e-05}
{"train_loss": 0.07761716842651367, "global_step": 128111, "epoch": 1439, "lr": 5.455215664692429e-05}
{"train_loss": 0.2393539398908615, "global_step": 128112, "epoch": 1439, "lr": 5.455157931769002e-05}
{"train_loss": 0.1743258684873581, "global_step": 128113, "epoch": 1439, "lr": 5.455100198784384e-05}
{"train_loss": 0.11145344376564026, "global_step": 128114, "epoch": 1439, "lr": 5.4550424657385846e-05}
{"train_loss": 0.15030083060264587, "global_step": 128115, "epoch": 1439, "lr": 5.4549847326316095e-05}
{"train_loss": 0.09896847605705261, "global_step": 128116, "epoch": 1439, "lr": 5.454926999463468e-05}
{"train_loss": 0.17304614186286926, "global_step": 128117, "epoch": 1439, "lr": 5.454869266234166e-05}
{"train_loss": 0.12407678365707397, "global_step": 128118, "epoch": 1439, "lr": 5.4548115329437135e-05}
{"train_loss": 0.09403491020202637, "global_step": 128119, "epoch": 1439, "lr": 5.454753799592116e-05}
{"train_loss": 0.18379636108875275, "global_step": 128120, "epoch": 1439, "lr": 5.454696066179383e-05}
{"train_loss": 0.11474510282278061, "global_step": 128121, "epoch": 1439, "lr": 5.454638332705522e-05}
{"train_loss": 0.17684462666511536, "global_step": 128122, "epoch": 1439, "lr": 5.454580599170539e-05}
{"train_loss": 0.1495431661605835, "global_step": 128123, "epoch": 1439, "lr": 5.454522865574445e-05}
{"train_loss": 0.20101171731948853, "global_step": 128124, "epoch": 1439, "lr": 5.454465131917245e-05}
{"train_loss": 0.11063594371080399, "global_step": 128125, "epoch": 1439, "lr": 5.454407398198947e-05}
{"train_loss": 0.07911108434200287, "global_step": 128126, "epoch": 1439, "lr": 5.4543496644195614e-05}
{"train_loss": 0.13062497973442078, "global_step": 128127, "epoch": 1439, "lr": 5.454291930579093e-05}
{"train_loss": 0.09163887798786163, "global_step": 128128, "epoch": 1439, "lr": 5.45423419667755e-05}
{"train_loss": 0.13019710779190063, "global_step": 128129, "epoch": 1439, "lr": 5.45417646271494e-05}
{"train_loss": 0.15979497134685516, "global_step": 128130, "epoch": 1439, "lr": 5.454118728691274e-05}
{"train_loss": 0.10161613672971725, "global_step": 128131, "epoch": 1439, "lr": 5.4540609946065555e-05}
{"train_loss": 0.14196835458278656, "global_step": 128132, "epoch": 1439, "lr": 5.454003260460795e-05}
{"train_loss": 0.09738720953464508, "global_step": 128133, "epoch": 1439, "lr": 5.453945526253998e-05}
{"train_loss": 0.10353609174489975, "global_step": 128134, "epoch": 1439, "lr": 5.453887791986175e-05}
{"train_loss": 0.13938340544700623, "global_step": 128135, "epoch": 1439, "lr": 5.453830057657332e-05}
{"train_loss": 0.11912887543439865, "global_step": 128136, "epoch": 1439, "lr": 5.453772323267477e-05}
{"train_loss": 0.0681232288479805, "global_step": 128137, "epoch": 1439, "lr": 5.4537145888166175e-05}
{"train_loss": 0.1030619814991951, "global_step": 128138, "epoch": 1439, "lr": 5.4536568543047616e-05}
{"train_loss": 0.1504218578338623, "global_step": 128139, "epoch": 1439, "lr": 5.453599119731917e-05}
{"train_loss": 0.09908801317214966, "global_step": 128140, "epoch": 1439, "lr": 5.453541385098092e-05}
{"train_loss": 0.2470448762178421, "global_step": 128141, "epoch": 1439, "lr": 5.4534836504032936e-05}
{"train_loss": 0.14060348272323608, "global_step": 128142, "epoch": 1439, "lr": 5.4534259156475295e-05}
{"train_loss": 0.14660243690013885, "global_step": 128143, "epoch": 1439, "lr": 5.453368180830809e-05}
{"train_loss": 0.2559305429458618, "global_step": 128144, "epoch": 1439, "lr": 5.453310445953137e-05}
{"train_loss": 0.07134854793548584, "global_step": 128145, "epoch": 1439, "lr": 5.453252711014524e-05}
{"train_loss": 0.10212744027376175, "global_step": 128146, "epoch": 1439, "lr": 5.453194976014977e-05}
{"train_loss": 0.13522374629974365, "global_step": 128147, "epoch": 1439, "lr": 5.4531372409545025e-05}
{"train_loss": 0.1408390998840332, "global_step": 128148, "epoch": 1439, "lr": 5.4530795058331096e-05}
{"train_loss": 0.10059715062379837, "global_step": 128149, "epoch": 1439, "lr": 5.4530217706508055e-05}
{"train_loss": 0.12804313004016876, "global_step": 128150, "epoch": 1439, "lr": 5.4529640354075984e-05}
{"train_loss": 0.13408790528774261, "global_step": 128151, "epoch": 1439, "lr": 5.4529063001034964e-05}
{"train_loss": 0.07505087554454803, "global_step": 128152, "epoch": 1439, "lr": 5.452848564738506e-05}
{"train_loss": 0.1356518417596817, "global_step": 128153, "epoch": 1439, "lr": 5.452790829312635e-05}
{"train_loss": 0.05722199007868767, "global_step": 128154, "epoch": 1439, "lr": 5.4527330938258935e-05}
{"train_loss": 0.11367511004209518, "global_step": 128155, "epoch": 1439, "lr": 5.4526753582782865e-05}
{"train_loss": 0.10525212436914444, "global_step": 128156, "epoch": 1439, "lr": 5.4526176226698225e-05}
{"train_loss": 0.1404675394296646, "global_step": 128157, "epoch": 1439, "lr": 5.4525598870005104e-05}
{"train_loss": 0.09267105162143707, "global_step": 128158, "epoch": 1439, "lr": 5.452502151270357e-05}
{"train_loss": 0.1288201514673367, "global_step": 128159, "epoch": 1439, "lr": 5.4524444154793695e-05, "val_loss": 5.624472618103027}
{"train_loss": 0.16574014723300934, "global_step": 128160, "epoch": 1440, "lr": 5.452386679627557e-05}
{"train_loss": 0.0777105987071991, "global_step": 128161, "epoch": 1440, "lr": 5.452328943714926e-05}
{"train_loss": 0.18348193168640137, "global_step": 128162, "epoch": 1440, "lr": 5.452271207741485e-05}
{"train_loss": 0.14783687889575958, "global_step": 128163, "epoch": 1440, "lr": 5.4522134717072415e-05}
{"train_loss": 0.0839448869228363, "global_step": 128164, "epoch": 1440, "lr": 5.4521557356122045e-05}
{"train_loss": 0.1100359633564949, "global_step": 128165, "epoch": 1440, "lr": 5.452097999456379e-05}
{"train_loss": 0.07903663069009781, "global_step": 128166, "epoch": 1440, "lr": 5.452040263239776e-05}
{"train_loss": 0.10474224388599396, "global_step": 128167, "epoch": 1440, "lr": 5.451982526962401e-05}
{"train_loss": 0.1140180230140686, "global_step": 128168, "epoch": 1440, "lr": 5.451924790624263e-05}
{"train_loss": 0.07811307907104492, "global_step": 128169, "epoch": 1440, "lr": 5.451867054225369e-05}
{"train_loss": 0.07861573994159698, "global_step": 128170, "epoch": 1440, "lr": 5.4518093177657256e-05}
{"train_loss": 0.03534076362848282, "global_step": 128171, "epoch": 1440, "lr": 5.451751581245343e-05}
{"train_loss": 0.10895561426877975, "global_step": 128172, "epoch": 1440, "lr": 5.4516938446642286e-05}
{"train_loss": 0.10025084763765335, "global_step": 128173, "epoch": 1440, "lr": 5.4516361080223885e-05}
{"train_loss": 0.10407412052154541, "global_step": 128174, "epoch": 1440, "lr": 5.4515783713198324e-05}
{"train_loss": 0.1211470365524292, "global_step": 128175, "epoch": 1440, "lr": 5.4515206345565664e-05}
{"train_loss": 0.15023639798164368, "global_step": 128176, "epoch": 1440, "lr": 5.4514628977326e-05}
{"train_loss": 0.12302328646183014, "global_step": 128177, "epoch": 1440, "lr": 5.4514051608479385e-05}
{"train_loss": 0.10131508857011795, "global_step": 128178, "epoch": 1440, "lr": 5.4513474239025916e-05}
{"train_loss": 0.1151963397860527, "global_step": 128179, "epoch": 1440, "lr": 5.4512896868965666e-05}
{"train_loss": 0.166655033826828, "global_step": 128180, "epoch": 1440, "lr": 5.451231949829872e-05}
{"train_loss": 0.20877186954021454, "global_step": 128181, "epoch": 1440, "lr": 5.451174212702512e-05}
{"train_loss": 0.14202113449573517, "global_step": 128182, "epoch": 1440, "lr": 5.451116475514499e-05}
{"train_loss": 0.1113177090883255, "global_step": 128183, "epoch": 1440, "lr": 5.451058738265839e-05}
{"train_loss": 0.15442006289958954, "global_step": 128184, "epoch": 1440, "lr": 5.4510010009565406e-05}
{"train_loss": 0.13667532801628113, "global_step": 128185, "epoch": 1440, "lr": 5.4509432635866086e-05}
{"train_loss": 0.10826688259840012, "global_step": 128186, "epoch": 1440, "lr": 5.450885526156054e-05}
{"train_loss": 0.16587257385253906, "global_step": 128187, "epoch": 1440, "lr": 5.4508277886648826e-05}
{"train_loss": 0.08347830921411514, "global_step": 128188, "epoch": 1440, "lr": 5.450770051113104e-05}
{"train_loss": 0.10465089231729507, "global_step": 128189, "epoch": 1440, "lr": 5.450712313500724e-05}
{"train_loss": 0.1103215217590332, "global_step": 128190, "epoch": 1440, "lr": 5.4506545758277515e-05}
{"train_loss": 0.1247781291604042, "global_step": 128191, "epoch": 1440, "lr": 5.4505968380941943e-05}
{"train_loss": 0.12436699122190475, "global_step": 128192, "epoch": 1440, "lr": 5.4505391003000594e-05}
{"train_loss": 0.1117665097117424, "global_step": 128193, "epoch": 1440, "lr": 5.4504813624453555e-05}
{"train_loss": 0.14356602728366852, "global_step": 128194, "epoch": 1440, "lr": 5.450423624530089e-05}
{"train_loss": 0.1016414687037468, "global_step": 128195, "epoch": 1440, "lr": 5.4503658865542694e-05}
{"train_loss": 0.1676289439201355, "global_step": 128196, "epoch": 1440, "lr": 5.4503081485179034e-05}
{"train_loss": 0.14436458051204681, "global_step": 128197, "epoch": 1440, "lr": 5.450250410420998e-05}
{"train_loss": 0.1630064696073532, "global_step": 128198, "epoch": 1440, "lr": 5.4501926722635634e-05}
{"train_loss": 0.11238302290439606, "global_step": 128199, "epoch": 1440, "lr": 5.4501349340456043e-05}
{"train_loss": 0.11145137995481491, "global_step": 128200, "epoch": 1440, "lr": 5.450077195767131e-05}
{"train_loss": 0.1900976598262787, "global_step": 128201, "epoch": 1440, "lr": 5.45001945742815e-05}
{"train_loss": 0.14386223256587982, "global_step": 128202, "epoch": 1440, "lr": 5.44996171902867e-05}
{"train_loss": 0.12770387530326843, "global_step": 128203, "epoch": 1440, "lr": 5.449903980568698e-05}
{"train_loss": 0.1311183124780655, "global_step": 128204, "epoch": 1440, "lr": 5.449846242048241e-05}
{"train_loss": 0.18439742922782898, "global_step": 128205, "epoch": 1440, "lr": 5.449788503467308e-05}
{"train_loss": 0.12887674570083618, "global_step": 128206, "epoch": 1440, "lr": 5.4497307648259065e-05}
{"train_loss": 0.12697672843933105, "global_step": 128207, "epoch": 1440, "lr": 5.449673026124045e-05}
{"train_loss": 0.17809166014194489, "global_step": 128208, "epoch": 1440, "lr": 5.4496152873617304e-05}
{"train_loss": 0.12834323942661285, "global_step": 128209, "epoch": 1440, "lr": 5.4495575485389695e-05}
{"train_loss": 0.1067446768283844, "global_step": 128210, "epoch": 1440, "lr": 5.449499809655772e-05}
{"train_loss": 0.13554055988788605, "global_step": 128211, "epoch": 1440, "lr": 5.449442070712143e-05}
{"train_loss": 0.0870525911450386, "global_step": 128212, "epoch": 1440, "lr": 5.449384331708094e-05}
{"train_loss": 0.19647905230522156, "global_step": 128213, "epoch": 1440, "lr": 5.4493265926436296e-05}
{"train_loss": 0.13484999537467957, "global_step": 128214, "epoch": 1440, "lr": 5.4492688535187584e-05}
{"train_loss": 0.18852710723876953, "global_step": 128215, "epoch": 1440, "lr": 5.44921111433349e-05}
{"train_loss": 0.16578072309494019, "global_step": 128216, "epoch": 1440, "lr": 5.44915337508783e-05}
{"train_loss": 0.10164424777030945, "global_step": 128217, "epoch": 1440, "lr": 5.449095635781786e-05}
{"train_loss": 0.18133360147476196, "global_step": 128218, "epoch": 1440, "lr": 5.449037896415368e-05}
{"train_loss": 0.13607072830200195, "global_step": 128219, "epoch": 1440, "lr": 5.448980156988581e-05}
{"train_loss": 0.12663383781909943, "global_step": 128220, "epoch": 1440, "lr": 5.448922417501434e-05}
{"train_loss": 0.12111977487802505, "global_step": 128221, "epoch": 1440, "lr": 5.4488646779539354e-05}
{"train_loss": 0.17330797016620636, "global_step": 128222, "epoch": 1440, "lr": 5.448806938346093e-05}
{"train_loss": 0.09983943402767181, "global_step": 128223, "epoch": 1440, "lr": 5.4487491986779125e-05}
{"train_loss": 0.14897818863391876, "global_step": 128224, "epoch": 1440, "lr": 5.448691458949404e-05}
{"train_loss": 0.17820820212364197, "global_step": 128225, "epoch": 1440, "lr": 5.448633719160575e-05}
{"train_loss": 0.10569833964109421, "global_step": 128226, "epoch": 1440, "lr": 5.448575979311432e-05}
{"train_loss": 0.1567174345254898, "global_step": 128227, "epoch": 1440, "lr": 5.448518239401984e-05}
{"train_loss": 0.11733148992061615, "global_step": 128228, "epoch": 1440, "lr": 5.4484604994322376e-05}
{"train_loss": 0.14625650644302368, "global_step": 128229, "epoch": 1440, "lr": 5.448402759402201e-05}
{"train_loss": 0.11426955461502075, "global_step": 128230, "epoch": 1440, "lr": 5.448345019311882e-05}
{"train_loss": 0.12060689181089401, "global_step": 128231, "epoch": 1440, "lr": 5.448287279161289e-05}
{"train_loss": 0.12689071893692017, "global_step": 128232, "epoch": 1440, "lr": 5.448229538950429e-05}
{"train_loss": 0.14015257358551025, "global_step": 128233, "epoch": 1440, "lr": 5.44817179867931e-05}
{"train_loss": 0.12676694989204407, "global_step": 128234, "epoch": 1440, "lr": 5.44811405834794e-05}
{"train_loss": 0.17728286981582642, "global_step": 128235, "epoch": 1440, "lr": 5.4480563179563274e-05}
{"train_loss": 0.14196808636188507, "global_step": 128236, "epoch": 1440, "lr": 5.447998577504476e-05}
{"train_loss": 0.0586044006049633, "global_step": 128237, "epoch": 1440, "lr": 5.447940836992399e-05}
{"train_loss": 0.2146815061569214, "global_step": 128238, "epoch": 1440, "lr": 5.4478830964201036e-05}
{"train_loss": 0.05694989114999771, "global_step": 128239, "epoch": 1440, "lr": 5.4478253557875924e-05}
{"train_loss": 0.19269052147865295, "global_step": 128240, "epoch": 1440, "lr": 5.447767615094879e-05}
{"train_loss": 0.23772194981575012, "global_step": 128241, "epoch": 1440, "lr": 5.447709874341968e-05}
{"train_loss": 0.14808377623558044, "global_step": 128242, "epoch": 1440, "lr": 5.4476521335288665e-05}
{"train_loss": 0.17637772858142853, "global_step": 128243, "epoch": 1440, "lr": 5.447594392655584e-05}
{"train_loss": 0.13832327723503113, "global_step": 128244, "epoch": 1440, "lr": 5.447536651722129e-05}
{"train_loss": 0.09401309490203857, "global_step": 128245, "epoch": 1440, "lr": 5.447478910728508e-05}
{"train_loss": 0.12857669591903687, "global_step": 128246, "epoch": 1440, "lr": 5.447421169674728e-05}
{"train_loss": 0.16091303527355194, "global_step": 128247, "epoch": 1440, "lr": 5.4473634285607976e-05}
{"train_loss": 0.13259848203049618, "global_step": 128248, "epoch": 1440, "lr": 5.4473056873867254e-05, "val_loss": 5.5362958908081055, "train_action_mse_error": 15.95507526397705}
{"train_loss": 0.19579368829727173, "global_step": 128249, "epoch": 1441, "lr": 5.4472479461525186e-05}
{"train_loss": 0.08612272888422012, "global_step": 128250, "epoch": 1441, "lr": 5.447190204858185e-05}
{"train_loss": 0.15794512629508972, "global_step": 128251, "epoch": 1441, "lr": 5.447132463503731e-05}
{"train_loss": 0.1436426192522049, "global_step": 128252, "epoch": 1441, "lr": 5.447074722089166e-05}
{"train_loss": 0.11982268840074539, "global_step": 128253, "epoch": 1441, "lr": 5.447016980614498e-05}
{"train_loss": 0.18801267445087433, "global_step": 128254, "epoch": 1441, "lr": 5.446959239079733e-05}
{"train_loss": 0.043625637888908386, "global_step": 128255, "epoch": 1441, "lr": 5.4469014974848784e-05}
{"train_loss": 0.14675256609916687, "global_step": 128256, "epoch": 1441, "lr": 5.4468437558299456e-05}
{"train_loss": 0.1137629896402359, "global_step": 128257, "epoch": 1441, "lr": 5.44678601411494e-05}
{"train_loss": 0.08764974772930145, "global_step": 128258, "epoch": 1441, "lr": 5.4467282723398694e-05}
{"train_loss": 0.16059546172618866, "global_step": 128259, "epoch": 1441, "lr": 5.446670530504741e-05}
{"train_loss": 0.1599844992160797, "global_step": 128260, "epoch": 1441, "lr": 5.446612788609563e-05}
{"train_loss": 0.1488444209098816, "global_step": 128261, "epoch": 1441, "lr": 5.4465550466543436e-05}
{"train_loss": 0.10579776018857956, "global_step": 128262, "epoch": 1441, "lr": 5.446497304639091e-05}
{"train_loss": 0.06652464717626572, "global_step": 128263, "epoch": 1441, "lr": 5.446439562563812e-05}
{"train_loss": 0.13093791902065277, "global_step": 128264, "epoch": 1441, "lr": 5.4463818204285144e-05}
{"train_loss": 0.11075355112552643, "global_step": 128265, "epoch": 1441, "lr": 5.446324078233206e-05}
{"train_loss": 0.10255830734968185, "global_step": 128266, "epoch": 1441, "lr": 5.446266335977894e-05}
{"train_loss": 0.1878470778465271, "global_step": 128267, "epoch": 1441, "lr": 5.4462085936625886e-05}
{"train_loss": 0.1024395078420639, "global_step": 128268, "epoch": 1441, "lr": 5.446150851287295e-05}
{"train_loss": 0.13766811788082123, "global_step": 128269, "epoch": 1441, "lr": 5.4460931088520226e-05}
{"train_loss": 0.20292136073112488, "global_step": 128270, "epoch": 1441, "lr": 5.4460353663567775e-05}
{"train_loss": 0.0673847645521164, "global_step": 128271, "epoch": 1441, "lr": 5.445977623801569e-05}
{"train_loss": 0.10834275931119919, "global_step": 128272, "epoch": 1441, "lr": 5.445919881186404e-05}
{"train_loss": 0.09030719846487045, "global_step": 128273, "epoch": 1441, "lr": 5.445862138511289e-05}
{"train_loss": 0.08876939862966537, "global_step": 128274, "epoch": 1441, "lr": 5.445804395776235e-05}
{"train_loss": 0.13595962524414062, "global_step": 128275, "epoch": 1441, "lr": 5.445746652981247e-05}
{"train_loss": 0.08154497295618057, "global_step": 128276, "epoch": 1441, "lr": 5.4456889101263365e-05}
{"train_loss": 0.20190343260765076, "global_step": 128277, "epoch": 1441, "lr": 5.4456311672115066e-05}
{"train_loss": 0.16172631084918976, "global_step": 128278, "epoch": 1441, "lr": 5.445573424236767e-05}
{"train_loss": 0.1738409548997879, "global_step": 128279, "epoch": 1441, "lr": 5.445515681202126e-05}
{"train_loss": 0.11851103603839874, "global_step": 128280, "epoch": 1441, "lr": 5.4454579381075896e-05}
{"train_loss": 0.1378384232521057, "global_step": 128281, "epoch": 1441, "lr": 5.445400194953167e-05}
{"train_loss": 0.15151691436767578, "global_step": 128282, "epoch": 1441, "lr": 5.445342451738867e-05}
{"train_loss": 0.12755857408046722, "global_step": 128283, "epoch": 1441, "lr": 5.445284708464696e-05}
{"train_loss": 0.1238807737827301, "global_step": 128284, "epoch": 1441, "lr": 5.445226965130661e-05}
{"train_loss": 0.13541075587272644, "global_step": 128285, "epoch": 1441, "lr": 5.445169221736771e-05}
{"train_loss": 0.12097392976284027, "global_step": 128286, "epoch": 1441, "lr": 5.445111478283034e-05}
{"train_loss": 0.12477173656225204, "global_step": 128287, "epoch": 1441, "lr": 5.445053734769456e-05}
{"train_loss": 0.17222069203853607, "global_step": 128288, "epoch": 1441, "lr": 5.4449959911960465e-05}
{"train_loss": 0.11920246481895447, "global_step": 128289, "epoch": 1441, "lr": 5.444938247562814e-05}
{"train_loss": 0.10364781320095062, "global_step": 128290, "epoch": 1441, "lr": 5.4448805038697635e-05}
{"train_loss": 0.08290165662765503, "global_step": 128291, "epoch": 1441, "lr": 5.4448227601169036e-05}
{"train_loss": 0.08932748436927795, "global_step": 128292, "epoch": 1441, "lr": 5.444765016304243e-05}
{"train_loss": 0.12267616391181946, "global_step": 128293, "epoch": 1441, "lr": 5.44470727243179e-05}
{"train_loss": 0.18247447907924652, "global_step": 128294, "epoch": 1441, "lr": 5.444649528499551e-05}
{"train_loss": 0.08740444481372833, "global_step": 128295, "epoch": 1441, "lr": 5.444591784507536e-05}
{"train_loss": 0.1621212661266327, "global_step": 128296, "epoch": 1441, "lr": 5.44453404045575e-05}
{"train_loss": 0.1883009672164917, "global_step": 128297, "epoch": 1441, "lr": 5.444476296344202e-05}
{"train_loss": 0.07037873566150665, "global_step": 128298, "epoch": 1441, "lr": 5.4444185521728984e-05}
{"train_loss": 0.16132086515426636, "global_step": 128299, "epoch": 1441, "lr": 5.44436080794185e-05}
{"train_loss": 0.11844415217638016, "global_step": 128300, "epoch": 1441, "lr": 5.444303063651061e-05}
{"train_loss": 0.13581746816635132, "global_step": 128301, "epoch": 1441, "lr": 5.4442453193005406e-05}
{"train_loss": 0.159467875957489, "global_step": 128302, "epoch": 1441, "lr": 5.4441875748902984e-05}
{"train_loss": 0.07932163029909134, "global_step": 128303, "epoch": 1441, "lr": 5.444129830420339e-05}
{"train_loss": 0.16972041130065918, "global_step": 128304, "epoch": 1441, "lr": 5.444072085890673e-05}
{"train_loss": 0.07823759317398071, "global_step": 128305, "epoch": 1441, "lr": 5.444014341301307e-05}
{"train_loss": 0.10377570241689682, "global_step": 128306, "epoch": 1441, "lr": 5.443956596652248e-05}
{"train_loss": 0.1425001621246338, "global_step": 128307, "epoch": 1441, "lr": 5.443898851943504e-05}
{"train_loss": 0.211313396692276, "global_step": 128308, "epoch": 1441, "lr": 5.443841107175084e-05}
{"train_loss": 0.11969242244958878, "global_step": 128309, "epoch": 1441, "lr": 5.443783362346994e-05}
{"train_loss": 0.1174817681312561, "global_step": 128310, "epoch": 1441, "lr": 5.443725617459243e-05}
{"train_loss": 0.21566142141819, "global_step": 128311, "epoch": 1441, "lr": 5.44366787251184e-05}
{"train_loss": 0.08889169991016388, "global_step": 128312, "epoch": 1441, "lr": 5.44361012750479e-05}
{"train_loss": 0.14230464398860931, "global_step": 128313, "epoch": 1441, "lr": 5.443552382438102e-05}
{"train_loss": 0.10762178152799606, "global_step": 128314, "epoch": 1441, "lr": 5.4434946373117846e-05}
{"train_loss": 0.1510135382413864, "global_step": 128315, "epoch": 1441, "lr": 5.443436892125845e-05}
{"train_loss": 0.11230853199958801, "global_step": 128316, "epoch": 1441, "lr": 5.4433791468802896e-05}
{"train_loss": 0.11147784441709518, "global_step": 128317, "epoch": 1441, "lr": 5.443321401575129e-05}
{"train_loss": 0.12348980456590652, "global_step": 128318, "epoch": 1441, "lr": 5.4432636562103664e-05}
{"train_loss": 0.08815529942512512, "global_step": 128319, "epoch": 1441, "lr": 5.4432059107860143e-05}
{"train_loss": 0.1385553628206253, "global_step": 128320, "epoch": 1441, "lr": 5.4431481653020786e-05}
{"train_loss": 0.1358809918165207, "global_step": 128321, "epoch": 1441, "lr": 5.4430904197585666e-05}
{"train_loss": 0.0930352658033371, "global_step": 128322, "epoch": 1441, "lr": 5.443032674155487e-05}
{"train_loss": 0.07759164273738861, "global_step": 128323, "epoch": 1441, "lr": 5.4429749284928465e-05}
{"train_loss": 0.1099838986992836, "global_step": 128324, "epoch": 1441, "lr": 5.4429171827706526e-05}
{"train_loss": 0.10626894980669022, "global_step": 128325, "epoch": 1441, "lr": 5.442859436988915e-05}
{"train_loss": 0.15144383907318115, "global_step": 128326, "epoch": 1441, "lr": 5.442801691147641e-05}
{"train_loss": 0.11447758972644806, "global_step": 128327, "epoch": 1441, "lr": 5.442743945246835e-05}
{"train_loss": 0.0798201858997345, "global_step": 128328, "epoch": 1441, "lr": 5.44268619928651e-05}
{"train_loss": 0.14091970026493073, "global_step": 128329, "epoch": 1441, "lr": 5.4426284532666705e-05}
{"train_loss": 0.1302085518836975, "global_step": 128330, "epoch": 1441, "lr": 5.442570707187324e-05}
{"train_loss": 0.16504335403442383, "global_step": 128331, "epoch": 1441, "lr": 5.44251296104848e-05}
{"train_loss": 0.11422847211360931, "global_step": 128332, "epoch": 1441, "lr": 5.442455214850146e-05}
{"train_loss": 0.09456592053174973, "global_step": 128333, "epoch": 1441, "lr": 5.44239746859233e-05}
{"train_loss": 0.08663918077945709, "global_step": 128334, "epoch": 1441, "lr": 5.44233972227504e-05}
{"train_loss": 0.1400027722120285, "global_step": 128335, "epoch": 1441, "lr": 5.4422819758982804e-05}
{"train_loss": 0.120364248752594, "global_step": 128336, "epoch": 1441, "lr": 5.442224229462062e-05}
{"train_loss": 0.12683991501840314, "global_step": 128337, "epoch": 1441, "lr": 5.442166482966392e-05, "val_loss": 5.577706336975098}
{"train_loss": 0.07660128176212311, "global_step": 128338, "epoch": 1442, "lr": 5.4421087364112777e-05}
{"train_loss": 0.09839729964733124, "global_step": 128339, "epoch": 1442, "lr": 5.442050989796728e-05}
{"train_loss": 0.12799666821956635, "global_step": 128340, "epoch": 1442, "lr": 5.4419932431227494e-05}
{"train_loss": 0.1244535893201828, "global_step": 128341, "epoch": 1442, "lr": 5.44193549638935e-05}
{"train_loss": 0.05556611344218254, "global_step": 128342, "epoch": 1442, "lr": 5.4418777495965375e-05}
{"train_loss": 0.1041380763053894, "global_step": 128343, "epoch": 1442, "lr": 5.441820002744321e-05}
{"train_loss": 0.1545439213514328, "global_step": 128344, "epoch": 1442, "lr": 5.4417622558327065e-05}
{"train_loss": 0.15222619473934174, "global_step": 128345, "epoch": 1442, "lr": 5.4417045088617015e-05}
{"train_loss": 0.07625100761651993, "global_step": 128346, "epoch": 1442, "lr": 5.441646761831316e-05}
{"train_loss": 0.1963176727294922, "global_step": 128347, "epoch": 1442, "lr": 5.441589014741556e-05}
{"train_loss": 0.154932901263237, "global_step": 128348, "epoch": 1442, "lr": 5.44153126759243e-05}
{"train_loss": 0.07292267680168152, "global_step": 128349, "epoch": 1442, "lr": 5.441473520383946e-05}
{"train_loss": 0.13320252299308777, "global_step": 128350, "epoch": 1442, "lr": 5.441415773116111e-05}
{"train_loss": 0.09036920964717865, "global_step": 128351, "epoch": 1442, "lr": 5.441358025788933e-05}
{"train_loss": 0.2209169864654541, "global_step": 128352, "epoch": 1442, "lr": 5.4413002784024194e-05}
{"train_loss": 0.07628387957811356, "global_step": 128353, "epoch": 1442, "lr": 5.4412425309565785e-05}
{"train_loss": 0.13885971903800964, "global_step": 128354, "epoch": 1442, "lr": 5.4411847834514195e-05}
{"train_loss": 0.0841960459947586, "global_step": 128355, "epoch": 1442, "lr": 5.441127035886947e-05}
{"train_loss": 0.12044809013605118, "global_step": 128356, "epoch": 1442, "lr": 5.44106928826317e-05}
{"train_loss": 0.10763541609048843, "global_step": 128357, "epoch": 1442, "lr": 5.441011540580098e-05}
{"train_loss": 0.15783050656318665, "global_step": 128358, "epoch": 1442, "lr": 5.440953792837735e-05}
{"train_loss": 0.16562409698963165, "global_step": 128359, "epoch": 1442, "lr": 5.440896045036094e-05}
{"train_loss": 0.21410658955574036, "global_step": 128360, "epoch": 1442, "lr": 5.440838297175178e-05}
{"train_loss": 0.1591443419456482, "global_step": 128361, "epoch": 1442, "lr": 5.440780549254997e-05}
{"train_loss": 0.13761651515960693, "global_step": 128362, "epoch": 1442, "lr": 5.4407228012755584e-05}
{"train_loss": 0.10063602030277252, "global_step": 128363, "epoch": 1442, "lr": 5.440665053236868e-05}
{"train_loss": 0.0941280648112297, "global_step": 128364, "epoch": 1442, "lr": 5.440607305138938e-05}
{"train_loss": 0.0764371007680893, "global_step": 128365, "epoch": 1442, "lr": 5.4405495569817745e-05}
{"train_loss": 0.13337723910808563, "global_step": 128366, "epoch": 1442, "lr": 5.4404918087653836e-05}
{"train_loss": 0.19192367792129517, "global_step": 128367, "epoch": 1442, "lr": 5.440434060489773e-05}
{"train_loss": 0.10261539369821548, "global_step": 128368, "epoch": 1442, "lr": 5.440376312154952e-05}
{"train_loss": 0.1542472392320633, "global_step": 128369, "epoch": 1442, "lr": 5.440318563760929e-05}
{"train_loss": 0.13928455114364624, "global_step": 128370, "epoch": 1442, "lr": 5.4402608153077083e-05}
{"train_loss": 0.07332507520914078, "global_step": 128371, "epoch": 1442, "lr": 5.440203066795302e-05}
{"train_loss": 0.1018613800406456, "global_step": 128372, "epoch": 1442, "lr": 5.440145318223715e-05}
{"train_loss": 0.15321935713291168, "global_step": 128373, "epoch": 1442, "lr": 5.440087569592956e-05}
{"train_loss": 0.15225808322429657, "global_step": 128374, "epoch": 1442, "lr": 5.440029820903032e-05}
{"train_loss": 0.13553376495838165, "global_step": 128375, "epoch": 1442, "lr": 5.4399720721539515e-05}
{"train_loss": 0.13502053916454315, "global_step": 128376, "epoch": 1442, "lr": 5.4399143233457225e-05}
{"train_loss": 0.16247205436229706, "global_step": 128377, "epoch": 1442, "lr": 5.439856574478352e-05}
{"train_loss": 0.16838830709457397, "global_step": 128378, "epoch": 1442, "lr": 5.4397988255518483e-05}
{"train_loss": 0.1723400205373764, "global_step": 128379, "epoch": 1442, "lr": 5.4397410765662195e-05}
{"train_loss": 0.0836838111281395, "global_step": 128380, "epoch": 1442, "lr": 5.439683327521472e-05}
{"train_loss": 0.16426889598369598, "global_step": 128381, "epoch": 1442, "lr": 5.439625578417615e-05}
{"train_loss": 0.1036587581038475, "global_step": 128382, "epoch": 1442, "lr": 5.439567829254655e-05}
{"train_loss": 0.13315527141094208, "global_step": 128383, "epoch": 1442, "lr": 5.4395100800326e-05}
{"train_loss": 0.11839096248149872, "global_step": 128384, "epoch": 1442, "lr": 5.439452330751459e-05}
{"train_loss": 0.10936857759952545, "global_step": 128385, "epoch": 1442, "lr": 5.439394581411239e-05}
{"train_loss": 0.12223946303129196, "global_step": 128386, "epoch": 1442, "lr": 5.4393368320119486e-05}
{"train_loss": 0.1691511571407318, "global_step": 128387, "epoch": 1442, "lr": 5.4392790825535933e-05}
{"train_loss": 0.13276879489421844, "global_step": 128388, "epoch": 1442, "lr": 5.439221333036184e-05}
{"train_loss": 0.1398225873708725, "global_step": 128389, "epoch": 1442, "lr": 5.4391635834597255e-05}
{"train_loss": 0.1211291179060936, "global_step": 128390, "epoch": 1442, "lr": 5.439105833824227e-05}
{"train_loss": 0.1197228953242302, "global_step": 128391, "epoch": 1442, "lr": 5.439048084129697e-05}
{"train_loss": 0.10978154093027115, "global_step": 128392, "epoch": 1442, "lr": 5.438990334376142e-05}
{"train_loss": 0.15483438968658447, "global_step": 128393, "epoch": 1442, "lr": 5.438932584563569e-05}
{"train_loss": 0.1549549400806427, "global_step": 128394, "epoch": 1442, "lr": 5.438874834691987e-05}
{"train_loss": 0.13875159621238708, "global_step": 128395, "epoch": 1442, "lr": 5.438817084761405e-05}
{"train_loss": 0.11895338445901871, "global_step": 128396, "epoch": 1442, "lr": 5.438759334771828e-05}
{"train_loss": 0.13887441158294678, "global_step": 128397, "epoch": 1442, "lr": 5.438701584723266e-05}
{"train_loss": 0.13317668437957764, "global_step": 128398, "epoch": 1442, "lr": 5.4386438346157255e-05}
{"train_loss": 0.12414605170488358, "global_step": 128399, "epoch": 1442, "lr": 5.438586084449214e-05}
{"train_loss": 0.12132832407951355, "global_step": 128400, "epoch": 1442, "lr": 5.4385283342237404e-05}
{"train_loss": 0.10650458186864853, "global_step": 128401, "epoch": 1442, "lr": 5.438470583939311e-05}
{"train_loss": 0.14354649186134338, "global_step": 128402, "epoch": 1442, "lr": 5.438412833595936e-05}
{"train_loss": 0.13147269189357758, "global_step": 128403, "epoch": 1442, "lr": 5.438355083193623e-05}
{"train_loss": 0.09098432958126068, "global_step": 128404, "epoch": 1442, "lr": 5.4382973327323764e-05}
{"train_loss": 0.08816078305244446, "global_step": 128405, "epoch": 1442, "lr": 5.438239582212207e-05}
{"train_loss": 0.08649351447820663, "global_step": 128406, "epoch": 1442, "lr": 5.4381818316331214e-05}
{"train_loss": 0.10595092922449112, "global_step": 128407, "epoch": 1442, "lr": 5.438124080995127e-05}
{"train_loss": 0.13036246597766876, "global_step": 128408, "epoch": 1442, "lr": 5.438066330298233e-05}
{"train_loss": 0.11192967742681503, "global_step": 128409, "epoch": 1442, "lr": 5.438008579542446e-05}
{"train_loss": 0.11859837919473648, "global_step": 128410, "epoch": 1442, "lr": 5.4379508287277735e-05}
{"train_loss": 0.13253237307071686, "global_step": 128411, "epoch": 1442, "lr": 5.437893077854225e-05}
{"train_loss": 0.0952821671962738, "global_step": 128412, "epoch": 1442, "lr": 5.437835326921807e-05}
{"train_loss": 0.08100955933332443, "global_step": 128413, "epoch": 1442, "lr": 5.437777575930527e-05}
{"train_loss": 0.1503942459821701, "global_step": 128414, "epoch": 1442, "lr": 5.437719824880394e-05}
{"train_loss": 0.10554154217243195, "global_step": 128415, "epoch": 1442, "lr": 5.4376620737714135e-05}
{"train_loss": 0.130443274974823, "global_step": 128416, "epoch": 1442, "lr": 5.437604322603595e-05}
{"train_loss": 0.1932905912399292, "global_step": 128417, "epoch": 1442, "lr": 5.4375465713769456e-05}
{"train_loss": 0.07870913296937943, "global_step": 128418, "epoch": 1442, "lr": 5.437488820091473e-05}
{"train_loss": 0.09859982877969742, "global_step": 128419, "epoch": 1442, "lr": 5.437431068747186e-05}
{"train_loss": 0.16445720195770264, "global_step": 128420, "epoch": 1442, "lr": 5.4373733173440934e-05}
{"train_loss": 0.13313685357570648, "global_step": 128421, "epoch": 1442, "lr": 5.4373155658821996e-05}
{"train_loss": 0.16023986041545868, "global_step": 128422, "epoch": 1442, "lr": 5.4372578143615136e-05}
{"train_loss": 0.08602987974882126, "global_step": 128423, "epoch": 1442, "lr": 5.4372000627820455e-05}
{"train_loss": 0.07668053358793259, "global_step": 128424, "epoch": 1442, "lr": 5.4371423111437995e-05}
{"train_loss": 0.17536576092243195, "global_step": 128425, "epoch": 1442, "lr": 5.4370845594467856e-05}
{"train_loss": 0.1266160069473004, "global_step": 128426, "epoch": 1442, "lr": 5.43702680769101e-05, "val_loss": 5.560896873474121}
{"train_loss": 0.2511691451072693, "global_step": 128427, "epoch": 1443, "lr": 5.436969055876483e-05}
{"train_loss": 0.15724216401576996, "global_step": 128428, "epoch": 1443, "lr": 5.43691130400321e-05}
{"train_loss": 0.09678108990192413, "global_step": 128429, "epoch": 1443, "lr": 5.4368535520712e-05}
{"train_loss": 0.06999105215072632, "global_step": 128430, "epoch": 1443, "lr": 5.436795800080461e-05}
{"train_loss": 0.16316582262516022, "global_step": 128431, "epoch": 1443, "lr": 5.436738048031e-05}
{"train_loss": 0.09974412620067596, "global_step": 128432, "epoch": 1443, "lr": 5.436680295922825e-05}
{"train_loss": 0.14388228952884674, "global_step": 128433, "epoch": 1443, "lr": 5.436622543755943e-05}
{"train_loss": 0.10461849719285965, "global_step": 128434, "epoch": 1443, "lr": 5.436564791530362e-05}
{"train_loss": 0.12348644435405731, "global_step": 128435, "epoch": 1443, "lr": 5.43650703924609e-05}
{"train_loss": 0.1601598858833313, "global_step": 128436, "epoch": 1443, "lr": 5.4364492869031344e-05}
{"train_loss": 0.16368664801120758, "global_step": 128437, "epoch": 1443, "lr": 5.436391534501505e-05}
{"train_loss": 0.11490457504987717, "global_step": 128438, "epoch": 1443, "lr": 5.4363337820412074e-05}
{"train_loss": 0.16512632369995117, "global_step": 128439, "epoch": 1443, "lr": 5.43627602952225e-05}
{"train_loss": 0.08135412633419037, "global_step": 128440, "epoch": 1443, "lr": 5.436218276944641e-05}
{"train_loss": 0.16290345788002014, "global_step": 128441, "epoch": 1443, "lr": 5.436160524308388e-05}
{"train_loss": 0.1218414455652237, "global_step": 128442, "epoch": 1443, "lr": 5.436102771613498e-05}
{"train_loss": 0.09917589277029037, "global_step": 128443, "epoch": 1443, "lr": 5.43604501885998e-05}
{"train_loss": 0.12817060947418213, "global_step": 128444, "epoch": 1443, "lr": 5.4359872660478404e-05}
{"train_loss": 0.1532921940088272, "global_step": 128445, "epoch": 1443, "lr": 5.435929513177087e-05}
{"train_loss": 0.09077899903059006, "global_step": 128446, "epoch": 1443, "lr": 5.4358717602477295e-05}
{"train_loss": 0.11904190480709076, "global_step": 128447, "epoch": 1443, "lr": 5.4358140072597744e-05}
{"train_loss": 0.1573004573583603, "global_step": 128448, "epoch": 1443, "lr": 5.435756254213229e-05}
{"train_loss": 0.11670778691768646, "global_step": 128449, "epoch": 1443, "lr": 5.435698501108101e-05}
{"train_loss": 0.13793668150901794, "global_step": 128450, "epoch": 1443, "lr": 5.435640747944399e-05}
{"train_loss": 0.14943476021289825, "global_step": 128451, "epoch": 1443, "lr": 5.435582994722132e-05}
{"train_loss": 0.1268983632326126, "global_step": 128452, "epoch": 1443, "lr": 5.435525241441304e-05}
{"train_loss": 0.12763291597366333, "global_step": 128453, "epoch": 1443, "lr": 5.435467488101925e-05}
{"train_loss": 0.15748821198940277, "global_step": 128454, "epoch": 1443, "lr": 5.4354097347040024e-05}
{"train_loss": 0.14889322221279144, "global_step": 128455, "epoch": 1443, "lr": 5.435351981247545e-05}
{"train_loss": 0.1570625901222229, "global_step": 128456, "epoch": 1443, "lr": 5.43529422773256e-05}
{"train_loss": 0.06531237065792084, "global_step": 128457, "epoch": 1443, "lr": 5.435236474159056e-05}
{"train_loss": 0.091932512819767, "global_step": 128458, "epoch": 1443, "lr": 5.435178720527038e-05}
{"train_loss": 0.17660348117351532, "global_step": 128459, "epoch": 1443, "lr": 5.435120966836516e-05}
{"train_loss": 0.1870591938495636, "global_step": 128460, "epoch": 1443, "lr": 5.435063213087498e-05}
{"train_loss": 0.11097728461027145, "global_step": 128461, "epoch": 1443, "lr": 5.4350054592799905e-05}
{"train_loss": 0.12844319641590118, "global_step": 128462, "epoch": 1443, "lr": 5.434947705414002e-05}
{"train_loss": 0.1274627298116684, "global_step": 128463, "epoch": 1443, "lr": 5.434889951489539e-05}
{"train_loss": 0.17469851672649384, "global_step": 128464, "epoch": 1443, "lr": 5.434832197506612e-05}
{"train_loss": 0.1457882970571518, "global_step": 128465, "epoch": 1443, "lr": 5.434774443465226e-05}
{"train_loss": 0.09555044025182724, "global_step": 128466, "epoch": 1443, "lr": 5.43471668936539e-05}
{"train_loss": 0.09885337948799133, "global_step": 128467, "epoch": 1443, "lr": 5.434658935207112e-05}
{"train_loss": 0.15810570120811462, "global_step": 128468, "epoch": 1443, "lr": 5.4346011809904e-05}
{"train_loss": 0.12052886933088303, "global_step": 128469, "epoch": 1443, "lr": 5.434543426715261e-05}
{"train_loss": 0.11176487803459167, "global_step": 128470, "epoch": 1443, "lr": 5.434485672381703e-05}
{"train_loss": 0.16875028610229492, "global_step": 128471, "epoch": 1443, "lr": 5.4344279179897326e-05}
{"train_loss": 0.14626029133796692, "global_step": 128472, "epoch": 1443, "lr": 5.434370163539358e-05}
{"train_loss": 0.09220544993877411, "global_step": 128473, "epoch": 1443, "lr": 5.43431240903059e-05}
{"train_loss": 0.10497783869504929, "global_step": 128474, "epoch": 1443, "lr": 5.434254654463432e-05}
{"train_loss": 0.1757327914237976, "global_step": 128475, "epoch": 1443, "lr": 5.4341968998378946e-05}
{"train_loss": 0.06925082951784134, "global_step": 128476, "epoch": 1443, "lr": 5.4341391451539844e-05}
{"train_loss": 0.11867368221282959, "global_step": 128477, "epoch": 1443, "lr": 5.43408139041171e-05}
{"train_loss": 0.13361157476902008, "global_step": 128478, "epoch": 1443, "lr": 5.434023635611079e-05}
{"train_loss": 0.14237752556800842, "global_step": 128479, "epoch": 1443, "lr": 5.433965880752098e-05}
{"train_loss": 0.08841870725154877, "global_step": 128480, "epoch": 1443, "lr": 5.433908125834776e-05}
{"train_loss": 0.09669285267591476, "global_step": 128481, "epoch": 1443, "lr": 5.4338503708591195e-05}
{"train_loss": 0.14124247431755066, "global_step": 128482, "epoch": 1443, "lr": 5.4337926158251374e-05}
{"train_loss": 0.12318162620067596, "global_step": 128483, "epoch": 1443, "lr": 5.433734860732837e-05}
{"train_loss": 0.18745096027851105, "global_step": 128484, "epoch": 1443, "lr": 5.4336771055822275e-05}
{"train_loss": 0.10431142896413803, "global_step": 128485, "epoch": 1443, "lr": 5.433619350373315e-05}
{"train_loss": 0.08562414348125458, "global_step": 128486, "epoch": 1443, "lr": 5.433561595106107e-05}
{"train_loss": 0.13241255283355713, "global_step": 128487, "epoch": 1443, "lr": 5.433503839780613e-05}
{"train_loss": 0.07295332849025726, "global_step": 128488, "epoch": 1443, "lr": 5.433446084396838e-05}
{"train_loss": 0.17274843156337738, "global_step": 128489, "epoch": 1443, "lr": 5.433388328954794e-05}
{"train_loss": 0.1431010216474533, "global_step": 128490, "epoch": 1443, "lr": 5.4333305734544835e-05}
{"train_loss": 0.13492482900619507, "global_step": 128491, "epoch": 1443, "lr": 5.433272817895918e-05}
{"train_loss": 0.07046065479516983, "global_step": 128492, "epoch": 1443, "lr": 5.433215062279104e-05}
{"train_loss": 0.15494143962860107, "global_step": 128493, "epoch": 1443, "lr": 5.4331573066040495e-05}
{"train_loss": 0.13392801582813263, "global_step": 128494, "epoch": 1443, "lr": 5.433099550870764e-05}
{"train_loss": 0.12374237924814224, "global_step": 128495, "epoch": 1443, "lr": 5.433041795079251e-05}
{"train_loss": 0.08025133609771729, "global_step": 128496, "epoch": 1443, "lr": 5.432984039229523e-05}
{"train_loss": 0.11001414805650711, "global_step": 128497, "epoch": 1443, "lr": 5.4329262833215845e-05}
{"train_loss": 0.15720131993293762, "global_step": 128498, "epoch": 1443, "lr": 5.432868527355443e-05}
{"train_loss": 0.1201111301779747, "global_step": 128499, "epoch": 1443, "lr": 5.43281077133111e-05}
{"train_loss": 0.11847054213285446, "global_step": 128500, "epoch": 1443, "lr": 5.432753015248589e-05}
{"train_loss": 0.09904880821704865, "global_step": 128501, "epoch": 1443, "lr": 5.4326952591078915e-05}
{"train_loss": 0.09758026897907257, "global_step": 128502, "epoch": 1443, "lr": 5.432637502909023e-05}
{"train_loss": 0.07785766571760178, "global_step": 128503, "epoch": 1443, "lr": 5.43257974665199e-05}
{"train_loss": 0.12462272495031357, "global_step": 128504, "epoch": 1443, "lr": 5.432521990336803e-05}
{"train_loss": 0.17468464374542236, "global_step": 128505, "epoch": 1443, "lr": 5.432464233963469e-05}
{"train_loss": 0.10102779418230057, "global_step": 128506, "epoch": 1443, "lr": 5.432406477531995e-05}
{"train_loss": 0.14021585881710052, "global_step": 128507, "epoch": 1443, "lr": 5.4323487210423896e-05}
{"train_loss": 0.1550632268190384, "global_step": 128508, "epoch": 1443, "lr": 5.4322909644946606e-05}
{"train_loss": 0.10306030511856079, "global_step": 128509, "epoch": 1443, "lr": 5.4322332078888137e-05}
{"train_loss": 0.1178465187549591, "global_step": 128510, "epoch": 1443, "lr": 5.432175451224859e-05}
{"train_loss": 0.1379779726266861, "global_step": 128511, "epoch": 1443, "lr": 5.432117694502804e-05}
{"train_loss": 0.11434625089168549, "global_step": 128512, "epoch": 1443, "lr": 5.4320599377226564e-05}
{"train_loss": 0.16182644665241241, "global_step": 128513, "epoch": 1443, "lr": 5.4320021808844236e-05}
{"train_loss": 0.12909992039203644, "global_step": 128514, "epoch": 1443, "lr": 5.431944423988112e-05}
{"train_loss": 0.12756316054068254, "global_step": 128515, "epoch": 1443, "lr": 5.431886667033732e-05, "val_loss": 5.6619391441345215}
{"train_loss": 0.20869624614715576, "global_step": 128516, "epoch": 1444, "lr": 5.4318289100212896e-05}
{"train_loss": 0.16510194540023804, "global_step": 128517, "epoch": 1444, "lr": 5.431771152950793e-05}
{"train_loss": 0.11023620516061783, "global_step": 128518, "epoch": 1444, "lr": 5.431713395822251e-05}
{"train_loss": 0.12848973274230957, "global_step": 128519, "epoch": 1444, "lr": 5.4316556386356685e-05}
{"train_loss": 0.14300604164600372, "global_step": 128520, "epoch": 1444, "lr": 5.431597881391057e-05}
{"train_loss": 0.18343572318553925, "global_step": 128521, "epoch": 1444, "lr": 5.4315401240884225e-05}
{"train_loss": 0.15463505685329437, "global_step": 128522, "epoch": 1444, "lr": 5.4314823667277716e-05}
{"train_loss": 0.16239994764328003, "global_step": 128523, "epoch": 1444, "lr": 5.4314246093091125e-05}
{"train_loss": 0.12302820384502411, "global_step": 128524, "epoch": 1444, "lr": 5.4313668518324553e-05}
{"train_loss": 0.10105951875448227, "global_step": 128525, "epoch": 1444, "lr": 5.431309094297805e-05}
{"train_loss": 0.10097774863243103, "global_step": 128526, "epoch": 1444, "lr": 5.431251336705171e-05}
{"train_loss": 0.049791961908340454, "global_step": 128527, "epoch": 1444, "lr": 5.431193579054561e-05}
{"train_loss": 0.10596577823162079, "global_step": 128528, "epoch": 1444, "lr": 5.4311358213459806e-05}
{"train_loss": 0.11329090595245361, "global_step": 128529, "epoch": 1444, "lr": 5.43107806357944e-05}
{"train_loss": 0.131240576505661, "global_step": 128530, "epoch": 1444, "lr": 5.431020305754947e-05}
{"train_loss": 0.12376375496387482, "global_step": 128531, "epoch": 1444, "lr": 5.430962547872508e-05}
{"train_loss": 0.16280323266983032, "global_step": 128532, "epoch": 1444, "lr": 5.430904789932132e-05}
{"train_loss": 0.09935011714696884, "global_step": 128533, "epoch": 1444, "lr": 5.4308470319338246e-05}
{"train_loss": 0.1514863520860672, "global_step": 128534, "epoch": 1444, "lr": 5.430789273877597e-05}
{"train_loss": 0.1305573433637619, "global_step": 128535, "epoch": 1444, "lr": 5.430731515763453e-05}
{"train_loss": 0.09973078966140747, "global_step": 128536, "epoch": 1444, "lr": 5.430673757591403e-05}
{"train_loss": 0.11949562281370163, "global_step": 128537, "epoch": 1444, "lr": 5.4306159993614546e-05}
{"train_loss": 0.09698202461004257, "global_step": 128538, "epoch": 1444, "lr": 5.430558241073616e-05}
{"train_loss": 0.13364340364933014, "global_step": 128539, "epoch": 1444, "lr": 5.4305004827278926e-05}
{"train_loss": 0.12119711190462112, "global_step": 128540, "epoch": 1444, "lr": 5.430442724324294e-05}
{"train_loss": 0.17371883988380432, "global_step": 128541, "epoch": 1444, "lr": 5.4303849658628284e-05}
{"train_loss": 0.15584565699100494, "global_step": 128542, "epoch": 1444, "lr": 5.430327207343502e-05}
{"train_loss": 0.1413595825433731, "global_step": 128543, "epoch": 1444, "lr": 5.430269448766324e-05}
{"train_loss": 0.14086180925369263, "global_step": 128544, "epoch": 1444, "lr": 5.430211690131301e-05}
{"train_loss": 0.13038215041160583, "global_step": 128545, "epoch": 1444, "lr": 5.430153931438441e-05}
{"train_loss": 0.17895011603832245, "global_step": 128546, "epoch": 1444, "lr": 5.430096172687753e-05}
{"train_loss": 0.08894368261098862, "global_step": 128547, "epoch": 1444, "lr": 5.430038413879243e-05}
{"train_loss": 0.11669730395078659, "global_step": 128548, "epoch": 1444, "lr": 5.42998065501292e-05}
{"train_loss": 0.12125740200281143, "global_step": 128549, "epoch": 1444, "lr": 5.4299228960887904e-05}
{"train_loss": 0.22561419010162354, "global_step": 128550, "epoch": 1444, "lr": 5.429865137106864e-05}
{"train_loss": 0.10208091884851456, "global_step": 128551, "epoch": 1444, "lr": 5.4298073780671474e-05}
{"train_loss": 0.091423898935318, "global_step": 128552, "epoch": 1444, "lr": 5.429749618969649e-05}
{"train_loss": 0.13644282519817352, "global_step": 128553, "epoch": 1444, "lr": 5.429691859814374e-05}
{"train_loss": 0.15423175692558289, "global_step": 128554, "epoch": 1444, "lr": 5.429634100601333e-05}
{"train_loss": 0.1555313616991043, "global_step": 128555, "epoch": 1444, "lr": 5.429576341330534e-05}
{"train_loss": 0.16124573349952698, "global_step": 128556, "epoch": 1444, "lr": 5.429518582001982e-05}
{"train_loss": 0.08106695115566254, "global_step": 128557, "epoch": 1444, "lr": 5.4294608226156884e-05}
{"train_loss": 0.11130911856889725, "global_step": 128558, "epoch": 1444, "lr": 5.429403063171658e-05}
{"train_loss": 0.12757731974124908, "global_step": 128559, "epoch": 1444, "lr": 5.429345303669899e-05}
{"train_loss": 0.08523060381412506, "global_step": 128560, "epoch": 1444, "lr": 5.4292875441104206e-05}
{"train_loss": 0.08839404582977295, "global_step": 128561, "epoch": 1444, "lr": 5.429229784493229e-05}
{"train_loss": 0.1712755560874939, "global_step": 128562, "epoch": 1444, "lr": 5.429172024818334e-05}
{"train_loss": 0.11553952842950821, "global_step": 128563, "epoch": 1444, "lr": 5.429114265085741e-05}
{"train_loss": 0.15175549685955048, "global_step": 128564, "epoch": 1444, "lr": 5.429056505295459e-05}
{"train_loss": 0.06442965567111969, "global_step": 128565, "epoch": 1444, "lr": 5.428998745447495e-05}
{"train_loss": 0.07795072346925735, "global_step": 128566, "epoch": 1444, "lr": 5.4289409855418594e-05}
{"train_loss": 0.13421614468097687, "global_step": 128567, "epoch": 1444, "lr": 5.428883225578556e-05}
{"train_loss": 0.149069145321846, "global_step": 128568, "epoch": 1444, "lr": 5.4288254655575956e-05}
{"train_loss": 0.13981667160987854, "global_step": 128569, "epoch": 1444, "lr": 5.4287677054789834e-05}
{"train_loss": 0.16202381253242493, "global_step": 128570, "epoch": 1444, "lr": 5.4287099453427295e-05}
{"train_loss": 0.10672549903392792, "global_step": 128571, "epoch": 1444, "lr": 5.4286521851488415e-05}
{"train_loss": 0.13410569727420807, "global_step": 128572, "epoch": 1444, "lr": 5.4285944248973255e-05}
{"train_loss": 0.17971134185791016, "global_step": 128573, "epoch": 1444, "lr": 5.4285366645881895e-05}
{"train_loss": 0.13227052986621857, "global_step": 128574, "epoch": 1444, "lr": 5.428478904221444e-05}
{"train_loss": 0.07134952396154404, "global_step": 128575, "epoch": 1444, "lr": 5.4284211437970936e-05}
{"train_loss": 0.05922364443540573, "global_step": 128576, "epoch": 1444, "lr": 5.428363383315147e-05}
{"train_loss": 0.18803834915161133, "global_step": 128577, "epoch": 1444, "lr": 5.4283056227756116e-05}
{"train_loss": 0.11443912237882614, "global_step": 128578, "epoch": 1444, "lr": 5.428247862178497e-05}
{"train_loss": 0.11472765356302261, "global_step": 128579, "epoch": 1444, "lr": 5.428190101523809e-05}
{"train_loss": 0.08242484927177429, "global_step": 128580, "epoch": 1444, "lr": 5.428132340811556e-05}
{"train_loss": 0.13181492686271667, "global_step": 128581, "epoch": 1444, "lr": 5.4280745800417465e-05}
{"train_loss": 0.16382397711277008, "global_step": 128582, "epoch": 1444, "lr": 5.428016819214386e-05}
{"train_loss": 0.07229673117399216, "global_step": 128583, "epoch": 1444, "lr": 5.427959058329486e-05}
{"train_loss": 0.1057673767209053, "global_step": 128584, "epoch": 1444, "lr": 5.42790129738705e-05}
{"train_loss": 0.15539787709712982, "global_step": 128585, "epoch": 1444, "lr": 5.4278435363870895e-05}
{"train_loss": 0.09332561492919922, "global_step": 128586, "epoch": 1444, "lr": 5.42778577532961e-05}
{"train_loss": 0.152041956782341, "global_step": 128587, "epoch": 1444, "lr": 5.4277280142146204e-05}
{"train_loss": 0.15736424922943115, "global_step": 128588, "epoch": 1444, "lr": 5.4276702530421275e-05}
{"train_loss": 0.11139766126871109, "global_step": 128589, "epoch": 1444, "lr": 5.427612491812139e-05}
{"train_loss": 0.08037381619215012, "global_step": 128590, "epoch": 1444, "lr": 5.4275547305246646e-05}
{"train_loss": 0.17404596507549286, "global_step": 128591, "epoch": 1444, "lr": 5.4274969691797104e-05}
{"train_loss": 0.1332647055387497, "global_step": 128592, "epoch": 1444, "lr": 5.427439207777284e-05}
{"train_loss": 0.11526688933372498, "global_step": 128593, "epoch": 1444, "lr": 5.427381446317393e-05}
{"train_loss": 0.2234804481267929, "global_step": 128594, "epoch": 1444, "lr": 5.427323684800046e-05}
{"train_loss": 0.11879726499319077, "global_step": 128595, "epoch": 1444, "lr": 5.427265923225252e-05}
{"train_loss": 0.08100488781929016, "global_step": 128596, "epoch": 1444, "lr": 5.427208161593016e-05}
{"train_loss": 0.0913870558142662, "global_step": 128597, "epoch": 1444, "lr": 5.427150399903347e-05}
{"train_loss": 0.09788549691438675, "global_step": 128598, "epoch": 1444, "lr": 5.427092638156254e-05}
{"train_loss": 0.06731666624546051, "global_step": 128599, "epoch": 1444, "lr": 5.427034876351741e-05}
{"train_loss": 0.18175026774406433, "global_step": 128600, "epoch": 1444, "lr": 5.426977114489822e-05}
{"train_loss": 0.10124117136001587, "global_step": 128601, "epoch": 1444, "lr": 5.426919352570499e-05}
{"train_loss": 0.11965432018041611, "global_step": 128602, "epoch": 1444, "lr": 5.426861590593781e-05}
{"train_loss": 0.1287553757429123, "global_step": 128603, "epoch": 1444, "lr": 5.426803828559679e-05}
{"train_loss": 0.1275427672133017, "global_step": 128604, "epoch": 1444, "lr": 5.4267460664681965e-05, "val_loss": 5.6488800048828125}
{"train_loss": 0.10745541006326675, "global_step": 128605, "epoch": 1445, "lr": 5.4266883043193436e-05}
{"train_loss": 0.09032244235277176, "global_step": 128606, "epoch": 1445, "lr": 5.426630542113129e-05}
{"train_loss": 0.10630292445421219, "global_step": 128607, "epoch": 1445, "lr": 5.426572779849558e-05}
{"train_loss": 0.1858070194721222, "global_step": 128608, "epoch": 1445, "lr": 5.4265150175286394e-05}
{"train_loss": 0.15123248100280762, "global_step": 128609, "epoch": 1445, "lr": 5.4264572551503815e-05}
{"train_loss": 0.12839949131011963, "global_step": 128610, "epoch": 1445, "lr": 5.426399492714791e-05}
{"train_loss": 0.10465136170387268, "global_step": 128611, "epoch": 1445, "lr": 5.4263417302218766e-05}
{"train_loss": 0.12183577567338943, "global_step": 128612, "epoch": 1445, "lr": 5.426283967671646e-05}
{"train_loss": 0.10348179936408997, "global_step": 128613, "epoch": 1445, "lr": 5.4262262050641064e-05}
{"train_loss": 0.08886505663394928, "global_step": 128614, "epoch": 1445, "lr": 5.426168442399267e-05}
{"train_loss": 0.13212177157402039, "global_step": 128615, "epoch": 1445, "lr": 5.4261106796771324e-05}
{"train_loss": 0.1529410481452942, "global_step": 128616, "epoch": 1445, "lr": 5.4260529168977136e-05}
{"train_loss": 0.0955166146159172, "global_step": 128617, "epoch": 1445, "lr": 5.425995154061016e-05}
{"train_loss": 0.17510008811950684, "global_step": 128618, "epoch": 1445, "lr": 5.4259373911670496e-05}
{"train_loss": 0.14534135162830353, "global_step": 128619, "epoch": 1445, "lr": 5.425879628215822e-05}
{"train_loss": 0.13834044337272644, "global_step": 128620, "epoch": 1445, "lr": 5.425821865207339e-05}
{"train_loss": 0.11013881117105484, "global_step": 128621, "epoch": 1445, "lr": 5.4257641021416096e-05}
{"train_loss": 0.07333746552467346, "global_step": 128622, "epoch": 1445, "lr": 5.4257063390186416e-05}
{"train_loss": 0.06545570492744446, "global_step": 128623, "epoch": 1445, "lr": 5.425648575838442e-05}
{"train_loss": 0.08515749126672745, "global_step": 128624, "epoch": 1445, "lr": 5.4255908126010203e-05}
{"train_loss": 0.1773708164691925, "global_step": 128625, "epoch": 1445, "lr": 5.425533049306382e-05}
{"train_loss": 0.11531713604927063, "global_step": 128626, "epoch": 1445, "lr": 5.425475285954537e-05}
{"train_loss": 0.07372581213712692, "global_step": 128627, "epoch": 1445, "lr": 5.425417522545491e-05}
{"train_loss": 0.15387193858623505, "global_step": 128628, "epoch": 1445, "lr": 5.425359759079253e-05}
{"train_loss": 0.05766858905553818, "global_step": 128629, "epoch": 1445, "lr": 5.42530199555583e-05}
{"train_loss": 0.15186095237731934, "global_step": 128630, "epoch": 1445, "lr": 5.425244231975232e-05}
{"train_loss": 0.11147881299257278, "global_step": 128631, "epoch": 1445, "lr": 5.425186468337463e-05}
{"train_loss": 0.12075778096914291, "global_step": 128632, "epoch": 1445, "lr": 5.4251287046425345e-05}
{"train_loss": 0.12870483100414276, "global_step": 128633, "epoch": 1445, "lr": 5.425070940890452e-05}
{"train_loss": 0.1469513177871704, "global_step": 128634, "epoch": 1445, "lr": 5.425013177081224e-05}
{"train_loss": 0.11517798155546188, "global_step": 128635, "epoch": 1445, "lr": 5.424955413214858e-05}
{"train_loss": 0.11001397669315338, "global_step": 128636, "epoch": 1445, "lr": 5.4248976492913615e-05}
{"train_loss": 0.13003969192504883, "global_step": 128637, "epoch": 1445, "lr": 5.424839885310743e-05}
{"train_loss": 0.17692942917346954, "global_step": 128638, "epoch": 1445, "lr": 5.42478212127301e-05}
{"train_loss": 0.07826566696166992, "global_step": 128639, "epoch": 1445, "lr": 5.424724357178171e-05}
{"train_loss": 0.098392553627491, "global_step": 128640, "epoch": 1445, "lr": 5.424666593026233e-05}
{"train_loss": 0.2305806428194046, "global_step": 128641, "epoch": 1445, "lr": 5.4246088288172036e-05}
{"train_loss": 0.1313033103942871, "global_step": 128642, "epoch": 1445, "lr": 5.42455106455109e-05}
{"train_loss": 0.14351601898670197, "global_step": 128643, "epoch": 1445, "lr": 5.424493300227901e-05}
{"train_loss": 0.0828833356499672, "global_step": 128644, "epoch": 1445, "lr": 5.424435535847644e-05}
{"train_loss": 0.15885771811008453, "global_step": 128645, "epoch": 1445, "lr": 5.424377771410327e-05}
{"train_loss": 0.061513204127550125, "global_step": 128646, "epoch": 1445, "lr": 5.424320006915957e-05}
{"train_loss": 0.10588081181049347, "global_step": 128647, "epoch": 1445, "lr": 5.424262242364543e-05}
{"train_loss": 0.061553098261356354, "global_step": 128648, "epoch": 1445, "lr": 5.424204477756092e-05}
{"train_loss": 0.17689229547977448, "global_step": 128649, "epoch": 1445, "lr": 5.424146713090612e-05}
{"train_loss": 0.13976114988327026, "global_step": 128650, "epoch": 1445, "lr": 5.42408894836811e-05}
{"train_loss": 0.1121472418308258, "global_step": 128651, "epoch": 1445, "lr": 5.4240311835885946e-05}
{"train_loss": 0.12089158594608307, "global_step": 128652, "epoch": 1445, "lr": 5.423973418752074e-05}
{"train_loss": 0.09882628172636032, "global_step": 128653, "epoch": 1445, "lr": 5.4239156538585554e-05}
{"train_loss": 0.1524980217218399, "global_step": 128654, "epoch": 1445, "lr": 5.423857888908045e-05}
{"train_loss": 0.12122150510549545, "global_step": 128655, "epoch": 1445, "lr": 5.423800123900553e-05}
{"train_loss": 0.13820168375968933, "global_step": 128656, "epoch": 1445, "lr": 5.4237423588360855e-05}
{"train_loss": 0.07032142579555511, "global_step": 128657, "epoch": 1445, "lr": 5.423684593714653e-05}
{"train_loss": 0.12320997565984726, "global_step": 128658, "epoch": 1445, "lr": 5.4236268285362615e-05}
{"train_loss": 0.07807949930429459, "global_step": 128659, "epoch": 1445, "lr": 5.423569063300916e-05}
{"train_loss": 0.08637192100286484, "global_step": 128660, "epoch": 1445, "lr": 5.423511298008629e-05}
{"train_loss": 0.15896686911582947, "global_step": 128661, "epoch": 1445, "lr": 5.423453532659405e-05}
{"train_loss": 0.24193021655082703, "global_step": 128662, "epoch": 1445, "lr": 5.4233957672532524e-05}
{"train_loss": 0.1236557886004448, "global_step": 128663, "epoch": 1445, "lr": 5.4233380017901804e-05}
{"train_loss": 0.11928465217351913, "global_step": 128664, "epoch": 1445, "lr": 5.423280236270195e-05}
{"train_loss": 0.09018239378929138, "global_step": 128665, "epoch": 1445, "lr": 5.4232224706933045e-05}
{"train_loss": 0.12725822627544403, "global_step": 128666, "epoch": 1445, "lr": 5.423164705059518e-05}
{"train_loss": 0.07310114055871964, "global_step": 128667, "epoch": 1445, "lr": 5.42310693936884e-05}
{"train_loss": 0.12837694585323334, "global_step": 128668, "epoch": 1445, "lr": 5.423049173621283e-05}
{"train_loss": 0.08752995729446411, "global_step": 128669, "epoch": 1445, "lr": 5.422991407816851e-05}
{"train_loss": 0.11689476668834686, "global_step": 128670, "epoch": 1445, "lr": 5.4229336419555535e-05}
{"train_loss": 0.1333475410938263, "global_step": 128671, "epoch": 1445, "lr": 5.422875876037398e-05}
{"train_loss": 0.05897287279367447, "global_step": 128672, "epoch": 1445, "lr": 5.42281811006239e-05}
{"train_loss": 0.06390555202960968, "global_step": 128673, "epoch": 1445, "lr": 5.4227603440305396e-05}
{"train_loss": 0.13994134962558746, "global_step": 128674, "epoch": 1445, "lr": 5.422702577941856e-05}
{"train_loss": 0.07680702209472656, "global_step": 128675, "epoch": 1445, "lr": 5.422644811796344e-05}
{"train_loss": 0.15527206659317017, "global_step": 128676, "epoch": 1445, "lr": 5.422587045594013e-05}
{"train_loss": 0.18122433125972748, "global_step": 128677, "epoch": 1445, "lr": 5.422529279334871e-05}
{"train_loss": 0.17313113808631897, "global_step": 128678, "epoch": 1445, "lr": 5.422471513018925e-05}
{"train_loss": 0.11733730882406235, "global_step": 128679, "epoch": 1445, "lr": 5.422413746646182e-05}
{"train_loss": 0.13483044505119324, "global_step": 128680, "epoch": 1445, "lr": 5.4223559802166515e-05}
{"train_loss": 0.10706320405006409, "global_step": 128681, "epoch": 1445, "lr": 5.4222982137303393e-05}
{"train_loss": 0.18285582959651947, "global_step": 128682, "epoch": 1445, "lr": 5.422240447187254e-05}
{"train_loss": 0.05254993215203285, "global_step": 128683, "epoch": 1445, "lr": 5.4221826805874045e-05}
{"train_loss": 0.11017115414142609, "global_step": 128684, "epoch": 1445, "lr": 5.4221249139307974e-05}
{"train_loss": 0.07350918650627136, "global_step": 128685, "epoch": 1445, "lr": 5.422067147217441e-05}
{"train_loss": 0.0925513207912445, "global_step": 128686, "epoch": 1445, "lr": 5.4220093804473425e-05}
{"train_loss": 0.11777723580598831, "global_step": 128687, "epoch": 1445, "lr": 5.421951613620511e-05}
{"train_loss": 0.1127622053027153, "global_step": 128688, "epoch": 1445, "lr": 5.421893846736952e-05}
{"train_loss": 0.1790824681520462, "global_step": 128689, "epoch": 1445, "lr": 5.421836079796675e-05}
{"train_loss": 0.19284333288669586, "global_step": 128690, "epoch": 1445, "lr": 5.421778312799686e-05}
{"train_loss": 0.14838159084320068, "global_step": 128691, "epoch": 1445, "lr": 5.421720545745995e-05}
{"train_loss": 0.11145620048046112, "global_step": 128692, "epoch": 1445, "lr": 5.421662778635609e-05}
{"train_loss": 0.12124932423401415, "global_step": 128693, "epoch": 1445, "lr": 5.4216050114685355e-05, "val_loss": 5.78670597076416, "train_action_mse_error": 27.65015983581543}
{"train_loss": 0.1478009968996048, "global_step": 128694, "epoch": 1446, "lr": 5.421547244244782e-05}
{"train_loss": 0.1836329698562622, "global_step": 128695, "epoch": 1446, "lr": 5.421489476964358e-05}
{"train_loss": 0.16198596358299255, "global_step": 128696, "epoch": 1446, "lr": 5.4214317096272694e-05}
{"train_loss": 0.09549378603696823, "global_step": 128697, "epoch": 1446, "lr": 5.421373942233524e-05}
{"train_loss": 0.0662584975361824, "global_step": 128698, "epoch": 1446, "lr": 5.4213161747831306e-05}
{"train_loss": 0.14521203935146332, "global_step": 128699, "epoch": 1446, "lr": 5.421258407276095e-05}
{"train_loss": 0.21487006545066833, "global_step": 128700, "epoch": 1446, "lr": 5.421200639712428e-05}
{"train_loss": 0.09209819883108139, "global_step": 128701, "epoch": 1446, "lr": 5.421142872092134e-05}
{"train_loss": 0.11127860099077225, "global_step": 128702, "epoch": 1446, "lr": 5.421085104415224e-05}
{"train_loss": 0.09269087761640549, "global_step": 128703, "epoch": 1446, "lr": 5.4210273366817035e-05}
{"train_loss": 0.12324637919664383, "global_step": 128704, "epoch": 1446, "lr": 5.4209695688915806e-05}
{"train_loss": 0.18597444891929626, "global_step": 128705, "epoch": 1446, "lr": 5.4209118010448644e-05}
{"train_loss": 0.13410744071006775, "global_step": 128706, "epoch": 1446, "lr": 5.420854033141561e-05}
{"train_loss": 0.19952361285686493, "global_step": 128707, "epoch": 1446, "lr": 5.420796265181679e-05}
{"train_loss": 0.11647211760282516, "global_step": 128708, "epoch": 1446, "lr": 5.420738497165225e-05}
{"train_loss": 0.13229110836982727, "global_step": 128709, "epoch": 1446, "lr": 5.42068072909221e-05}
{"train_loss": 0.07853258401155472, "global_step": 128710, "epoch": 1446, "lr": 5.420622960962638e-05}
{"train_loss": 0.12365397810935974, "global_step": 128711, "epoch": 1446, "lr": 5.420565192776519e-05}
{"train_loss": 0.11531849950551987, "global_step": 128712, "epoch": 1446, "lr": 5.42050742453386e-05}
{"train_loss": 0.11164659261703491, "global_step": 128713, "epoch": 1446, "lr": 5.420449656234669e-05}
{"train_loss": 0.15164704620838165, "global_step": 128714, "epoch": 1446, "lr": 5.420391887878955e-05}
{"train_loss": 0.1604272425174713, "global_step": 128715, "epoch": 1446, "lr": 5.420334119466722e-05}
{"train_loss": 0.1945313811302185, "global_step": 128716, "epoch": 1446, "lr": 5.4202763509979836e-05}
{"train_loss": 0.11749676614999771, "global_step": 128717, "epoch": 1446, "lr": 5.420218582472741e-05}
{"train_loss": 0.11545871943235397, "global_step": 128718, "epoch": 1446, "lr": 5.4201608138910064e-05}
{"train_loss": 0.12015927582979202, "global_step": 128719, "epoch": 1446, "lr": 5.420103045252787e-05}
{"train_loss": 0.17684170603752136, "global_step": 128720, "epoch": 1446, "lr": 5.4200452765580876e-05}
{"train_loss": 0.18180622160434723, "global_step": 128721, "epoch": 1446, "lr": 5.41998750780692e-05}
{"train_loss": 0.19853265583515167, "global_step": 128722, "epoch": 1446, "lr": 5.4199297389992895e-05}
{"train_loss": 0.08366690576076508, "global_step": 128723, "epoch": 1446, "lr": 5.4198719701352054e-05}
{"train_loss": 0.07179243117570877, "global_step": 128724, "epoch": 1446, "lr": 5.419814201214674e-05}
{"train_loss": 0.14940719306468964, "global_step": 128725, "epoch": 1446, "lr": 5.419756432237704e-05}
{"train_loss": 0.13052114844322205, "global_step": 128726, "epoch": 1446, "lr": 5.4196986632043e-05}
{"train_loss": 0.2247665375471115, "global_step": 128727, "epoch": 1446, "lr": 5.419640894114476e-05}
{"train_loss": 0.0824756771326065, "global_step": 128728, "epoch": 1446, "lr": 5.4195831249682347e-05}
{"train_loss": 0.14302293956279755, "global_step": 128729, "epoch": 1446, "lr": 5.419525355765587e-05}
{"train_loss": 0.1182941123843193, "global_step": 128730, "epoch": 1446, "lr": 5.4194675865065394e-05}
{"train_loss": 0.11374186724424362, "global_step": 128731, "epoch": 1446, "lr": 5.419409817191098e-05}
{"train_loss": 0.10712725669145584, "global_step": 128732, "epoch": 1446, "lr": 5.4193520478192725e-05}
{"train_loss": 0.1755206137895584, "global_step": 128733, "epoch": 1446, "lr": 5.419294278391072e-05}
{"train_loss": 0.14514946937561035, "global_step": 128734, "epoch": 1446, "lr": 5.4192365089065003e-05}
{"train_loss": 0.10997910797595978, "global_step": 128735, "epoch": 1446, "lr": 5.419178739365569e-05}
{"train_loss": 0.09011396765708923, "global_step": 128736, "epoch": 1446, "lr": 5.419120969768283e-05}
{"train_loss": 0.16193567216396332, "global_step": 128737, "epoch": 1446, "lr": 5.419063200114653e-05}
{"train_loss": 0.09288875013589859, "global_step": 128738, "epoch": 1446, "lr": 5.419005430404682e-05}
{"train_loss": 0.15759524703025818, "global_step": 128739, "epoch": 1446, "lr": 5.418947660638383e-05}
{"train_loss": 0.12980002164840698, "global_step": 128740, "epoch": 1446, "lr": 5.4188898908157616e-05}
{"train_loss": 0.09943074733018875, "global_step": 128741, "epoch": 1446, "lr": 5.4188321209368244e-05}
{"train_loss": 0.10911509394645691, "global_step": 128742, "epoch": 1446, "lr": 5.418774351001581e-05}
{"train_loss": 0.11252003163099289, "global_step": 128743, "epoch": 1446, "lr": 5.418716581010039e-05}
{"train_loss": 0.1070023626089096, "global_step": 128744, "epoch": 1446, "lr": 5.4186588109622036e-05}
{"train_loss": 0.13005225360393524, "global_step": 128745, "epoch": 1446, "lr": 5.418601040858087e-05}
{"train_loss": 0.1481393724679947, "global_step": 128746, "epoch": 1446, "lr": 5.4185432706976936e-05}
{"train_loss": 0.06633760780096054, "global_step": 128747, "epoch": 1446, "lr": 5.418485500481032e-05}
{"train_loss": 0.13980437815189362, "global_step": 128748, "epoch": 1446, "lr": 5.41842773020811e-05}
{"train_loss": 0.16502119600772858, "global_step": 128749, "epoch": 1446, "lr": 5.418369959878936e-05}
{"train_loss": 0.22044454514980316, "global_step": 128750, "epoch": 1446, "lr": 5.418312189493517e-05}
{"train_loss": 0.14310607314109802, "global_step": 128751, "epoch": 1446, "lr": 5.418254419051861e-05}
{"train_loss": 0.11855902522802353, "global_step": 128752, "epoch": 1446, "lr": 5.418196648553977e-05}
{"train_loss": 0.17808504402637482, "global_step": 128753, "epoch": 1446, "lr": 5.4181388779998695e-05}
{"train_loss": 0.14760982990264893, "global_step": 128754, "epoch": 1446, "lr": 5.41808110738955e-05}
{"train_loss": 0.0762796550989151, "global_step": 128755, "epoch": 1446, "lr": 5.418023336723025e-05}
{"train_loss": 0.11483465135097504, "global_step": 128756, "epoch": 1446, "lr": 5.417965566000301e-05}
{"train_loss": 0.16279974579811096, "global_step": 128757, "epoch": 1446, "lr": 5.417907795221386e-05}
{"train_loss": 0.16268287599086761, "global_step": 128758, "epoch": 1446, "lr": 5.4178500243862896e-05}
{"train_loss": 0.14599759876728058, "global_step": 128759, "epoch": 1446, "lr": 5.417792253495018e-05}
{"train_loss": 0.12982282042503357, "global_step": 128760, "epoch": 1446, "lr": 5.4177344825475783e-05}
{"train_loss": 0.11077322065830231, "global_step": 128761, "epoch": 1446, "lr": 5.41767671154398e-05}
{"train_loss": 0.12230640649795532, "global_step": 128762, "epoch": 1446, "lr": 5.4176189404842306e-05}
{"train_loss": 0.08654624223709106, "global_step": 128763, "epoch": 1446, "lr": 5.417561169368336e-05}
{"train_loss": 0.08749759942293167, "global_step": 128764, "epoch": 1446, "lr": 5.417503398196305e-05}
{"train_loss": 0.17324891686439514, "global_step": 128765, "epoch": 1446, "lr": 5.4174456269681474e-05}
{"train_loss": 0.0765613242983818, "global_step": 128766, "epoch": 1446, "lr": 5.41738785568387e-05}
{"train_loss": 0.08502741158008575, "global_step": 128767, "epoch": 1446, "lr": 5.4173300843434784e-05}
{"train_loss": 0.14429718255996704, "global_step": 128768, "epoch": 1446, "lr": 5.417272312946981e-05}
{"train_loss": 0.12065646052360535, "global_step": 128769, "epoch": 1446, "lr": 5.417214541494389e-05}
{"train_loss": 0.14553862810134888, "global_step": 128770, "epoch": 1446, "lr": 5.417156769985706e-05}
{"train_loss": 0.13591815531253815, "global_step": 128771, "epoch": 1446, "lr": 5.417098998420942e-05}
{"train_loss": 0.11234677582979202, "global_step": 128772, "epoch": 1446, "lr": 5.417041226800103e-05}
{"train_loss": 0.09643615037202835, "global_step": 128773, "epoch": 1446, "lr": 5.416983455123199e-05}
{"train_loss": 0.16508309543132782, "global_step": 128774, "epoch": 1446, "lr": 5.416925683390238e-05}
{"train_loss": 0.09312431514263153, "global_step": 128775, "epoch": 1446, "lr": 5.4168679116012246e-05}
{"train_loss": 0.15232615172863007, "global_step": 128776, "epoch": 1446, "lr": 5.416810139756169e-05}
{"train_loss": 0.16455817222595215, "global_step": 128777, "epoch": 1446, "lr": 5.416752367855078e-05}
{"train_loss": 0.06680571287870407, "global_step": 128778, "epoch": 1446, "lr": 5.41669459589796e-05}
{"train_loss": 0.11405850201845169, "global_step": 128779, "epoch": 1446, "lr": 5.416636823884822e-05}
{"train_loss": 0.13803018629550934, "global_step": 128780, "epoch": 1446, "lr": 5.416579051815672e-05}
{"train_loss": 0.08367379009723663, "global_step": 128781, "epoch": 1446, "lr": 5.416521279690517e-05}
{"train_loss": 0.13023706374878294, "global_step": 128782, "epoch": 1446, "lr": 5.4164635075093674e-05, "val_loss": 5.657522678375244}
{"train_loss": 0.15093477070331573, "global_step": 128783, "epoch": 1447, "lr": 5.4164057352722294e-05}
{"train_loss": 0.11311796307563782, "global_step": 128784, "epoch": 1447, "lr": 5.41634796297911e-05}
{"train_loss": 0.13300110399723053, "global_step": 128785, "epoch": 1447, "lr": 5.416290190630019e-05}
{"train_loss": 0.09716679900884628, "global_step": 128786, "epoch": 1447, "lr": 5.4162324182249615e-05}
{"train_loss": 0.12511222064495087, "global_step": 128787, "epoch": 1447, "lr": 5.416174645763947e-05}
{"train_loss": 0.04497241973876953, "global_step": 128788, "epoch": 1447, "lr": 5.4161168732469835e-05}
{"train_loss": 0.08842675387859344, "global_step": 128789, "epoch": 1447, "lr": 5.416059100674078e-05}
{"train_loss": 0.15946343541145325, "global_step": 128790, "epoch": 1447, "lr": 5.4160013280452384e-05}
{"train_loss": 0.12191686779260635, "global_step": 128791, "epoch": 1447, "lr": 5.415943555360472e-05}
{"train_loss": 0.13682009279727936, "global_step": 128792, "epoch": 1447, "lr": 5.415885782619787e-05}
{"train_loss": 0.15408053994178772, "global_step": 128793, "epoch": 1447, "lr": 5.415828009823193e-05}
{"train_loss": 0.06982911378145218, "global_step": 128794, "epoch": 1447, "lr": 5.415770236970694e-05}
{"train_loss": 0.11813434213399887, "global_step": 128795, "epoch": 1447, "lr": 5.4157124640623e-05}
{"train_loss": 0.14131014049053192, "global_step": 128796, "epoch": 1447, "lr": 5.415654691098019e-05}
{"train_loss": 0.11046438664197922, "global_step": 128797, "epoch": 1447, "lr": 5.4155969180778586e-05}
{"train_loss": 0.08798842132091522, "global_step": 128798, "epoch": 1447, "lr": 5.415539145001825e-05}
{"train_loss": 0.141182079911232, "global_step": 128799, "epoch": 1447, "lr": 5.415481371869927e-05}
{"train_loss": 0.10323919355869293, "global_step": 128800, "epoch": 1447, "lr": 5.4154235986821735e-05}
{"train_loss": 0.09917963296175003, "global_step": 128801, "epoch": 1447, "lr": 5.415365825438572e-05}
{"train_loss": 0.06917431205511093, "global_step": 128802, "epoch": 1447, "lr": 5.415308052139129e-05}
{"train_loss": 0.232668936252594, "global_step": 128803, "epoch": 1447, "lr": 5.415250278783853e-05}
{"train_loss": 0.19026066362857819, "global_step": 128804, "epoch": 1447, "lr": 5.415192505372752e-05}
{"train_loss": 0.07336107641458511, "global_step": 128805, "epoch": 1447, "lr": 5.4151347319058324e-05}
{"train_loss": 0.09005558490753174, "global_step": 128806, "epoch": 1447, "lr": 5.415076958383104e-05}
{"train_loss": 0.08119212090969086, "global_step": 128807, "epoch": 1447, "lr": 5.4150191848045725e-05}
{"train_loss": 0.05534786731004715, "global_step": 128808, "epoch": 1447, "lr": 5.4149614111702476e-05}
{"train_loss": 0.1326768696308136, "global_step": 128809, "epoch": 1447, "lr": 5.4149036374801364e-05}
{"train_loss": 0.08845466375350952, "global_step": 128810, "epoch": 1447, "lr": 5.414845863734246e-05}
{"train_loss": 0.0700063481926918, "global_step": 128811, "epoch": 1447, "lr": 5.414788089932584e-05}
{"train_loss": 0.11474347859621048, "global_step": 128812, "epoch": 1447, "lr": 5.4147303160751604e-05}
{"train_loss": 0.09641721844673157, "global_step": 128813, "epoch": 1447, "lr": 5.4146725421619814e-05}
{"train_loss": 0.07464714348316193, "global_step": 128814, "epoch": 1447, "lr": 5.4146147681930534e-05}
{"train_loss": 0.0923653095960617, "global_step": 128815, "epoch": 1447, "lr": 5.414556994168386e-05}
{"train_loss": 0.15952399373054504, "global_step": 128816, "epoch": 1447, "lr": 5.414499220087986e-05}
{"train_loss": 0.1357244998216629, "global_step": 128817, "epoch": 1447, "lr": 5.4144414459518614e-05}
{"train_loss": 0.21755702793598175, "global_step": 128818, "epoch": 1447, "lr": 5.4143836717600214e-05}
{"train_loss": 0.18472391366958618, "global_step": 128819, "epoch": 1447, "lr": 5.414325897512471e-05}
{"train_loss": 0.10831178724765778, "global_step": 128820, "epoch": 1447, "lr": 5.4142681232092205e-05}
{"train_loss": 0.09195663034915924, "global_step": 128821, "epoch": 1447, "lr": 5.414210348850277e-05}
{"train_loss": 0.11468103528022766, "global_step": 128822, "epoch": 1447, "lr": 5.414152574435648e-05}
{"train_loss": 0.1390787810087204, "global_step": 128823, "epoch": 1447, "lr": 5.4140947999653414e-05}
{"train_loss": 0.10564889013767242, "global_step": 128824, "epoch": 1447, "lr": 5.414037025439363e-05}
{"train_loss": 0.10137061029672623, "global_step": 128825, "epoch": 1447, "lr": 5.413979250857725e-05}
{"train_loss": 0.09770630300045013, "global_step": 128826, "epoch": 1447, "lr": 5.41392147622043e-05}
{"train_loss": 0.11251131445169449, "global_step": 128827, "epoch": 1447, "lr": 5.413863701527491e-05}
{"train_loss": 0.08207789808511734, "global_step": 128828, "epoch": 1447, "lr": 5.413805926778911e-05}
{"train_loss": 0.13113386929035187, "global_step": 128829, "epoch": 1447, "lr": 5.4137481519747e-05}
{"train_loss": 0.13176898658275604, "global_step": 128830, "epoch": 1447, "lr": 5.413690377114866e-05}
{"train_loss": 0.11044038832187653, "global_step": 128831, "epoch": 1447, "lr": 5.413632602199417e-05}
{"train_loss": 0.15552058815956116, "global_step": 128832, "epoch": 1447, "lr": 5.41357482722836e-05}
{"train_loss": 0.10491929948329926, "global_step": 128833, "epoch": 1447, "lr": 5.413517052201703e-05}
{"train_loss": 0.29848477244377136, "global_step": 128834, "epoch": 1447, "lr": 5.413459277119452e-05}
{"train_loss": 0.1965501308441162, "global_step": 128835, "epoch": 1447, "lr": 5.4134015019816175e-05}
{"train_loss": 0.10059770941734314, "global_step": 128836, "epoch": 1447, "lr": 5.413343726788206e-05}
{"train_loss": 0.143620565533638, "global_step": 128837, "epoch": 1447, "lr": 5.4132859515392266e-05}
{"train_loss": 0.04240976646542549, "global_step": 128838, "epoch": 1447, "lr": 5.413228176234685e-05}
{"train_loss": 0.12462542951107025, "global_step": 128839, "epoch": 1447, "lr": 5.41317040087459e-05}
{"train_loss": 0.133163183927536, "global_step": 128840, "epoch": 1447, "lr": 5.413112625458948e-05}
{"train_loss": 0.09173797816038132, "global_step": 128841, "epoch": 1447, "lr": 5.41305484998777e-05}
{"train_loss": 0.10688386112451553, "global_step": 128842, "epoch": 1447, "lr": 5.4129970744610614e-05}
{"train_loss": 0.1045617014169693, "global_step": 128843, "epoch": 1447, "lr": 5.41293929887883e-05}
{"train_loss": 0.11728791147470474, "global_step": 128844, "epoch": 1447, "lr": 5.4128815232410835e-05}
{"train_loss": 0.10556555539369583, "global_step": 128845, "epoch": 1447, "lr": 5.41282374754783e-05}
{"train_loss": 0.10733918100595474, "global_step": 128846, "epoch": 1447, "lr": 5.4127659717990786e-05}
{"train_loss": 0.10627643764019012, "global_step": 128847, "epoch": 1447, "lr": 5.4127081959948355e-05}
{"train_loss": 0.11642226576805115, "global_step": 128848, "epoch": 1447, "lr": 5.412650420135109e-05}
{"train_loss": 0.10915737599134445, "global_step": 128849, "epoch": 1447, "lr": 5.412592644219906e-05}
{"train_loss": 0.08733879774808884, "global_step": 128850, "epoch": 1447, "lr": 5.412534868249235e-05}
{"train_loss": 0.10332927107810974, "global_step": 128851, "epoch": 1447, "lr": 5.412477092223105e-05}
{"train_loss": 0.18310515582561493, "global_step": 128852, "epoch": 1447, "lr": 5.412419316141521e-05}
{"train_loss": 0.14813530445098877, "global_step": 128853, "epoch": 1447, "lr": 5.4123615400044925e-05}
{"train_loss": 0.20281431078910828, "global_step": 128854, "epoch": 1447, "lr": 5.4123037638120265e-05}
{"train_loss": 0.21999265253543854, "global_step": 128855, "epoch": 1447, "lr": 5.412245987564132e-05}
{"train_loss": 0.17280465364456177, "global_step": 128856, "epoch": 1447, "lr": 5.4121882112608157e-05}
{"train_loss": 0.09418191760778427, "global_step": 128857, "epoch": 1447, "lr": 5.4121304349020865e-05}
{"train_loss": 0.0909080058336258, "global_step": 128858, "epoch": 1447, "lr": 5.412072658487951e-05}
{"train_loss": 0.1712995320558548, "global_step": 128859, "epoch": 1447, "lr": 5.4120148820184166e-05}
{"train_loss": 0.14394550025463104, "global_step": 128860, "epoch": 1447, "lr": 5.4119571054934935e-05}
{"train_loss": 0.1490994691848755, "global_step": 128861, "epoch": 1447, "lr": 5.4118993289131866e-05}
{"train_loss": 0.21878176927566528, "global_step": 128862, "epoch": 1447, "lr": 5.411841552277505e-05}
{"train_loss": 0.09560111165046692, "global_step": 128863, "epoch": 1447, "lr": 5.411783775586456e-05}
{"train_loss": 0.12826868891716003, "global_step": 128864, "epoch": 1447, "lr": 5.411725998840048e-05}
{"train_loss": 0.06598696857690811, "global_step": 128865, "epoch": 1447, "lr": 5.411668222038288e-05}
{"train_loss": 0.1494995653629303, "global_step": 128866, "epoch": 1447, "lr": 5.411610445181185e-05}
{"train_loss": 0.18581920862197876, "global_step": 128867, "epoch": 1447, "lr": 5.4115526682687466e-05}
{"train_loss": 0.07645120471715927, "global_step": 128868, "epoch": 1447, "lr": 5.411494891300979e-05}
{"train_loss": 0.08307510614395142, "global_step": 128869, "epoch": 1447, "lr": 5.411437114277891e-05}
{"train_loss": 0.0889129713177681, "global_step": 128870, "epoch": 1447, "lr": 5.4113793371994906e-05}
{"train_loss": 0.12266510017634777, "global_step": 128871, "epoch": 1447, "lr": 5.411321560065784e-05, "val_loss": 5.749698162078857}
{"train_loss": 0.09582360833883286, "global_step": 128872, "epoch": 1448, "lr": 5.411263782876782e-05}
{"train_loss": 0.1314024180173874, "global_step": 128873, "epoch": 1448, "lr": 5.4112060056324896e-05}
{"train_loss": 0.1255498081445694, "global_step": 128874, "epoch": 1448, "lr": 5.411148228332916e-05}
{"train_loss": 0.12420806288719177, "global_step": 128875, "epoch": 1448, "lr": 5.411090450978068e-05}
{"train_loss": 0.06750628352165222, "global_step": 128876, "epoch": 1448, "lr": 5.411032673567954e-05}
{"train_loss": 0.07814212143421173, "global_step": 128877, "epoch": 1448, "lr": 5.410974896102582e-05}
{"train_loss": 0.13294737040996552, "global_step": 128878, "epoch": 1448, "lr": 5.4109171185819595e-05}
{"train_loss": 0.10414081811904907, "global_step": 128879, "epoch": 1448, "lr": 5.4108593410060935e-05}
{"train_loss": 0.16055908799171448, "global_step": 128880, "epoch": 1448, "lr": 5.410801563374993e-05}
{"train_loss": 0.15230078995227814, "global_step": 128881, "epoch": 1448, "lr": 5.4107437856886655e-05}
{"train_loss": 0.14834217727184296, "global_step": 128882, "epoch": 1448, "lr": 5.410686007947118e-05}
{"train_loss": 0.10070588439702988, "global_step": 128883, "epoch": 1448, "lr": 5.4106282301503585e-05}
{"train_loss": 0.1873844563961029, "global_step": 128884, "epoch": 1448, "lr": 5.4105704522983956e-05}
{"train_loss": 0.11144953221082687, "global_step": 128885, "epoch": 1448, "lr": 5.4105126743912374e-05}
{"train_loss": 0.10303711146116257, "global_step": 128886, "epoch": 1448, "lr": 5.410454896428889e-05}
{"train_loss": 0.1495579481124878, "global_step": 128887, "epoch": 1448, "lr": 5.410397118411361e-05}
{"train_loss": 0.15557177364826202, "global_step": 128888, "epoch": 1448, "lr": 5.41033934033866e-05}
{"train_loss": 0.07800241559743881, "global_step": 128889, "epoch": 1448, "lr": 5.4102815622107936e-05}
{"train_loss": 0.13810554146766663, "global_step": 128890, "epoch": 1448, "lr": 5.41022378402777e-05}
{"train_loss": 0.0753672644495964, "global_step": 128891, "epoch": 1448, "lr": 5.4101660057895964e-05}
{"train_loss": 0.1453842669725418, "global_step": 128892, "epoch": 1448, "lr": 5.410108227496282e-05}
{"train_loss": 0.13206925988197327, "global_step": 128893, "epoch": 1448, "lr": 5.4100504491478324e-05}
{"train_loss": 0.1956566572189331, "global_step": 128894, "epoch": 1448, "lr": 5.409992670744257e-05}
{"train_loss": 0.14688485860824585, "global_step": 128895, "epoch": 1448, "lr": 5.4099348922855634e-05}
{"train_loss": 0.15800444781780243, "global_step": 128896, "epoch": 1448, "lr": 5.409877113771759e-05}
{"train_loss": 0.12118995934724808, "global_step": 128897, "epoch": 1448, "lr": 5.4098193352028517e-05}
{"train_loss": 0.14141163229942322, "global_step": 128898, "epoch": 1448, "lr": 5.409761556578849e-05}
{"train_loss": 0.14414334297180176, "global_step": 128899, "epoch": 1448, "lr": 5.409703777899758e-05}
{"train_loss": 0.1025734543800354, "global_step": 128900, "epoch": 1448, "lr": 5.409645999165588e-05}
{"train_loss": 0.14577914774417877, "global_step": 128901, "epoch": 1448, "lr": 5.409588220376346e-05}
{"train_loss": 0.14940905570983887, "global_step": 128902, "epoch": 1448, "lr": 5.40953044153204e-05}
{"train_loss": 0.08633586019277573, "global_step": 128903, "epoch": 1448, "lr": 5.4094726626326776e-05}
{"train_loss": 0.10774976760149002, "global_step": 128904, "epoch": 1448, "lr": 5.409414883678267e-05}
{"train_loss": 0.12186209857463837, "global_step": 128905, "epoch": 1448, "lr": 5.409357104668815e-05}
{"train_loss": 0.1469220072031021, "global_step": 128906, "epoch": 1448, "lr": 5.40929932560433e-05}
{"train_loss": 0.1187116950750351, "global_step": 128907, "epoch": 1448, "lr": 5.4092415464848204e-05}
{"train_loss": 0.11589063704013824, "global_step": 128908, "epoch": 1448, "lr": 5.409183767310292e-05}
{"train_loss": 0.12504054605960846, "global_step": 128909, "epoch": 1448, "lr": 5.409125988080754e-05}
{"train_loss": 0.12660305202007294, "global_step": 128910, "epoch": 1448, "lr": 5.4090682087962153e-05}
{"train_loss": 0.24773675203323364, "global_step": 128911, "epoch": 1448, "lr": 5.409010429456681e-05}
{"train_loss": 0.14254377782344818, "global_step": 128912, "epoch": 1448, "lr": 5.4089526500621614e-05}
{"train_loss": 0.13936202228069305, "global_step": 128913, "epoch": 1448, "lr": 5.408894870612662e-05}
{"train_loss": 0.12691114842891693, "global_step": 128914, "epoch": 1448, "lr": 5.408837091108192e-05}
{"train_loss": 0.1255679577589035, "global_step": 128915, "epoch": 1448, "lr": 5.40877931154876e-05}
{"train_loss": 0.12542636692523956, "global_step": 128916, "epoch": 1448, "lr": 5.408721531934371e-05}
{"train_loss": 0.12102856487035751, "global_step": 128917, "epoch": 1448, "lr": 5.408663752265034e-05}
{"train_loss": 0.11400812864303589, "global_step": 128918, "epoch": 1448, "lr": 5.4086059725407586e-05}
{"train_loss": 0.15787051618099213, "global_step": 128919, "epoch": 1448, "lr": 5.408548192761551e-05}
{"train_loss": 0.17672398686408997, "global_step": 128920, "epoch": 1448, "lr": 5.408490412927418e-05}
{"train_loss": 0.1359757035970688, "global_step": 128921, "epoch": 1448, "lr": 5.4084326330383694e-05}
{"train_loss": 0.1267046183347702, "global_step": 128922, "epoch": 1448, "lr": 5.4083748530944114e-05}
{"train_loss": 0.11931570619344711, "global_step": 128923, "epoch": 1448, "lr": 5.4083170730955534e-05}
{"train_loss": 0.09366299957036972, "global_step": 128924, "epoch": 1448, "lr": 5.408259293041801e-05}
{"train_loss": 0.08707228302955627, "global_step": 128925, "epoch": 1448, "lr": 5.4082015129331634e-05}
{"train_loss": 0.11892605572938919, "global_step": 128926, "epoch": 1448, "lr": 5.408143732769648e-05}
{"train_loss": 0.14876125752925873, "global_step": 128927, "epoch": 1448, "lr": 5.408085952551263e-05}
{"train_loss": 0.09236635267734528, "global_step": 128928, "epoch": 1448, "lr": 5.408028172278016e-05}
{"train_loss": 0.1662132740020752, "global_step": 128929, "epoch": 1448, "lr": 5.407970391949915e-05}
{"train_loss": 0.11641422659158707, "global_step": 128930, "epoch": 1448, "lr": 5.4079126115669654e-05}
{"train_loss": 0.10445789992809296, "global_step": 128931, "epoch": 1448, "lr": 5.407854831129179e-05}
{"train_loss": 0.14242671430110931, "global_step": 128932, "epoch": 1448, "lr": 5.4077970506365604e-05}
{"train_loss": 0.16626538336277008, "global_step": 128933, "epoch": 1448, "lr": 5.4077392700891185e-05}
{"train_loss": 0.10938812047243118, "global_step": 128934, "epoch": 1448, "lr": 5.407681489486861e-05}
{"train_loss": 0.18837220966815948, "global_step": 128935, "epoch": 1448, "lr": 5.4076237088297964e-05}
{"train_loss": 0.10477197170257568, "global_step": 128936, "epoch": 1448, "lr": 5.407565928117931e-05}
{"train_loss": 0.17307330667972565, "global_step": 128937, "epoch": 1448, "lr": 5.407508147351274e-05}
{"train_loss": 0.173972025513649, "global_step": 128938, "epoch": 1448, "lr": 5.4074503665298315e-05}
{"train_loss": 0.1183379516005516, "global_step": 128939, "epoch": 1448, "lr": 5.4073925856536134e-05}
{"train_loss": 0.16321875154972076, "global_step": 128940, "epoch": 1448, "lr": 5.4073348047226255e-05}
{"train_loss": 0.14247798919677734, "global_step": 128941, "epoch": 1448, "lr": 5.4072770237368766e-05}
{"train_loss": 0.10991458594799042, "global_step": 128942, "epoch": 1448, "lr": 5.4072192426963754e-05}
{"train_loss": 0.13450947403907776, "global_step": 128943, "epoch": 1448, "lr": 5.407161461601127e-05}
{"train_loss": 0.14944486320018768, "global_step": 128944, "epoch": 1448, "lr": 5.407103680451141e-05}
{"train_loss": 0.1213841363787651, "global_step": 128945, "epoch": 1448, "lr": 5.407045899246426e-05}
{"train_loss": 0.10225555300712585, "global_step": 128946, "epoch": 1448, "lr": 5.406988117986986e-05}
{"train_loss": 0.20924971997737885, "global_step": 128947, "epoch": 1448, "lr": 5.406930336672834e-05}
{"train_loss": 0.13676804304122925, "global_step": 128948, "epoch": 1448, "lr": 5.4068725553039746e-05}
{"train_loss": 0.15297147631645203, "global_step": 128949, "epoch": 1448, "lr": 5.406814773880415e-05}
{"train_loss": 0.11818165332078934, "global_step": 128950, "epoch": 1448, "lr": 5.4067569924021655e-05}
{"train_loss": 0.12077641487121582, "global_step": 128951, "epoch": 1448, "lr": 5.406699210869232e-05}
{"train_loss": 0.13038018345832825, "global_step": 128952, "epoch": 1448, "lr": 5.4066414292816227e-05}
{"train_loss": 0.09741489589214325, "global_step": 128953, "epoch": 1448, "lr": 5.406583647639345e-05}
{"train_loss": 0.12210342288017273, "global_step": 128954, "epoch": 1448, "lr": 5.406525865942408e-05}
{"train_loss": 0.11954104155302048, "global_step": 128955, "epoch": 1448, "lr": 5.406468084190818e-05}
{"train_loss": 0.09958990663290024, "global_step": 128956, "epoch": 1448, "lr": 5.406410302384584e-05}
{"train_loss": 0.06217595562338829, "global_step": 128957, "epoch": 1448, "lr": 5.406352520523712e-05}
{"train_loss": 0.07343270629644394, "global_step": 128958, "epoch": 1448, "lr": 5.4062947386082106e-05}
{"train_loss": 0.17154310643672943, "global_step": 128959, "epoch": 1448, "lr": 5.406236956638089e-05}
{"train_loss": 0.12956425618757023, "global_step": 128960, "epoch": 1448, "lr": 5.406179174613353e-05, "val_loss": 5.716286659240723}
{"train_loss": 0.09286260604858398, "global_step": 128961, "epoch": 1449, "lr": 5.406121392534011e-05}
{"train_loss": 0.08313564956188202, "global_step": 128962, "epoch": 1449, "lr": 5.406063610400072e-05}
{"train_loss": 0.14522884786128998, "global_step": 128963, "epoch": 1449, "lr": 5.406005828211542e-05}
{"train_loss": 0.15119490027427673, "global_step": 128964, "epoch": 1449, "lr": 5.40594804596843e-05}
{"train_loss": 0.17814351618289948, "global_step": 128965, "epoch": 1449, "lr": 5.405890263670743e-05}
{"train_loss": 0.156044602394104, "global_step": 128966, "epoch": 1449, "lr": 5.405832481318489e-05}
{"train_loss": 0.12207543104887009, "global_step": 128967, "epoch": 1449, "lr": 5.4057746989116756e-05}
{"train_loss": 0.12651373445987701, "global_step": 128968, "epoch": 1449, "lr": 5.4057169164503105e-05}
{"train_loss": 0.16590873897075653, "global_step": 128969, "epoch": 1449, "lr": 5.405659133934402e-05}
{"train_loss": 0.17234858870506287, "global_step": 128970, "epoch": 1449, "lr": 5.405601351363957e-05}
{"train_loss": 0.12191103398799896, "global_step": 128971, "epoch": 1449, "lr": 5.405543568738984e-05}
{"train_loss": 0.12088846415281296, "global_step": 128972, "epoch": 1449, "lr": 5.405485786059492e-05}
{"train_loss": 0.1065124124288559, "global_step": 128973, "epoch": 1449, "lr": 5.405428003325485e-05}
{"train_loss": 0.12693853676319122, "global_step": 128974, "epoch": 1449, "lr": 5.405370220536975e-05}
{"train_loss": 0.1482723504304886, "global_step": 128975, "epoch": 1449, "lr": 5.405312437693968e-05}
{"train_loss": 0.20116934180259705, "global_step": 128976, "epoch": 1449, "lr": 5.405254654796471e-05}
{"train_loss": 0.14860986173152924, "global_step": 128977, "epoch": 1449, "lr": 5.4051968718444926e-05}
{"train_loss": 0.10232046991586685, "global_step": 128978, "epoch": 1449, "lr": 5.40513908883804e-05}
{"train_loss": 0.1208687350153923, "global_step": 128979, "epoch": 1449, "lr": 5.405081305777122e-05}
{"train_loss": 0.11748644709587097, "global_step": 128980, "epoch": 1449, "lr": 5.405023522661744e-05}
{"train_loss": 0.13284362852573395, "global_step": 128981, "epoch": 1449, "lr": 5.404965739491917e-05}
{"train_loss": 0.15625175833702087, "global_step": 128982, "epoch": 1449, "lr": 5.4049079562676485e-05}
{"train_loss": 0.09098822623491287, "global_step": 128983, "epoch": 1449, "lr": 5.404850172988945e-05}
{"train_loss": 0.1156424731016159, "global_step": 128984, "epoch": 1449, "lr": 5.4047923896558114e-05}
{"train_loss": 0.15260036289691925, "global_step": 128985, "epoch": 1449, "lr": 5.404734606268261e-05}
{"train_loss": 0.13672560453414917, "global_step": 128986, "epoch": 1449, "lr": 5.404676822826298e-05}
{"train_loss": 0.11408504098653793, "global_step": 128987, "epoch": 1449, "lr": 5.404619039329931e-05}
{"train_loss": 0.1028253361582756, "global_step": 128988, "epoch": 1449, "lr": 5.404561255779168e-05}
{"train_loss": 0.23248343169689178, "global_step": 128989, "epoch": 1449, "lr": 5.4045034721740174e-05}
{"train_loss": 0.0987003967165947, "global_step": 128990, "epoch": 1449, "lr": 5.4044456885144856e-05}
{"train_loss": 0.14066000282764435, "global_step": 128991, "epoch": 1449, "lr": 5.4043879048005804e-05}
{"train_loss": 0.10563968122005463, "global_step": 128992, "epoch": 1449, "lr": 5.404330121032311e-05}
{"train_loss": 0.0852903500199318, "global_step": 128993, "epoch": 1449, "lr": 5.404272337209685e-05}
{"train_loss": 0.11424564570188522, "global_step": 128994, "epoch": 1449, "lr": 5.404214553332709e-05}
{"train_loss": 0.15485848486423492, "global_step": 128995, "epoch": 1449, "lr": 5.404156769401391e-05}
{"train_loss": 0.1465502381324768, "global_step": 128996, "epoch": 1449, "lr": 5.4040989854157385e-05}
{"train_loss": 0.09537408500909805, "global_step": 128997, "epoch": 1449, "lr": 5.40404120137576e-05}
{"train_loss": 0.14667782187461853, "global_step": 128998, "epoch": 1449, "lr": 5.403983417281463e-05}
{"train_loss": 0.13347084820270538, "global_step": 128999, "epoch": 1449, "lr": 5.403925633132856e-05}
{"train_loss": 0.0692560225725174, "global_step": 129000, "epoch": 1449, "lr": 5.403867848929945e-05}
{"train_loss": 0.06382744759321213, "global_step": 129001, "epoch": 1449, "lr": 5.4038100646727416e-05}
{"train_loss": 0.10358963161706924, "global_step": 129002, "epoch": 1449, "lr": 5.403752280361249e-05}
{"train_loss": 0.16827872395515442, "global_step": 129003, "epoch": 1449, "lr": 5.403694495995477e-05}
{"train_loss": 0.1521734744310379, "global_step": 129004, "epoch": 1449, "lr": 5.403636711575433e-05}
{"train_loss": 0.20310541987419128, "global_step": 129005, "epoch": 1449, "lr": 5.403578927101125e-05}
{"train_loss": 0.16898272931575775, "global_step": 129006, "epoch": 1449, "lr": 5.403521142572561e-05}
{"train_loss": 0.10805384814739227, "global_step": 129007, "epoch": 1449, "lr": 5.403463357989749e-05}
{"train_loss": 0.19941002130508423, "global_step": 129008, "epoch": 1449, "lr": 5.4034055733526956e-05}
{"train_loss": 0.08402802795171738, "global_step": 129009, "epoch": 1449, "lr": 5.403347788661408e-05}
{"train_loss": 0.10822324454784393, "global_step": 129010, "epoch": 1449, "lr": 5.403290003915897e-05}
{"train_loss": 0.1389341652393341, "global_step": 129011, "epoch": 1449, "lr": 5.4032322191161676e-05}
{"train_loss": 0.11500812321901321, "global_step": 129012, "epoch": 1449, "lr": 5.403174434262229e-05}
{"train_loss": 0.18076173961162567, "global_step": 129013, "epoch": 1449, "lr": 5.403116649354088e-05}
{"train_loss": 0.1538587361574173, "global_step": 129014, "epoch": 1449, "lr": 5.403058864391755e-05}
{"train_loss": 0.11050692200660706, "global_step": 129015, "epoch": 1449, "lr": 5.403001079375233e-05}
{"train_loss": 0.18524327874183655, "global_step": 129016, "epoch": 1449, "lr": 5.4029432943045334e-05}
{"train_loss": 0.23525039851665497, "global_step": 129017, "epoch": 1449, "lr": 5.4028855091796624e-05}
{"train_loss": 0.1539575606584549, "global_step": 129018, "epoch": 1449, "lr": 5.402827724000629e-05}
{"train_loss": 0.10617918521165848, "global_step": 129019, "epoch": 1449, "lr": 5.4027699387674405e-05}
{"train_loss": 0.13662634789943695, "global_step": 129020, "epoch": 1449, "lr": 5.402712153480105e-05}
{"train_loss": 0.10707014799118042, "global_step": 129021, "epoch": 1449, "lr": 5.402654368138629e-05}
{"train_loss": 0.12936855852603912, "global_step": 129022, "epoch": 1449, "lr": 5.4025965827430204e-05}
{"train_loss": 0.07899872213602066, "global_step": 129023, "epoch": 1449, "lr": 5.4025387972932886e-05}
{"train_loss": 0.08214928954839706, "global_step": 129024, "epoch": 1449, "lr": 5.4024810117894395e-05}
{"train_loss": 0.11539097875356674, "global_step": 129025, "epoch": 1449, "lr": 5.402423226231482e-05}
{"train_loss": 0.11117718368768692, "global_step": 129026, "epoch": 1449, "lr": 5.402365440619425e-05}
{"train_loss": 0.17422305047512054, "global_step": 129027, "epoch": 1449, "lr": 5.402307654953274e-05}
{"train_loss": 0.13652841746807098, "global_step": 129028, "epoch": 1449, "lr": 5.402249869233037e-05}
{"train_loss": 0.08682766556739807, "global_step": 129029, "epoch": 1449, "lr": 5.402192083458723e-05}
{"train_loss": 0.10260225832462311, "global_step": 129030, "epoch": 1449, "lr": 5.402134297630338e-05}
{"train_loss": 0.09140828996896744, "global_step": 129031, "epoch": 1449, "lr": 5.4020765117478923e-05}
{"train_loss": 0.16226212680339813, "global_step": 129032, "epoch": 1449, "lr": 5.402018725811392e-05}
{"train_loss": 0.15327614545822144, "global_step": 129033, "epoch": 1449, "lr": 5.401960939820845e-05}
{"train_loss": 0.10994071513414383, "global_step": 129034, "epoch": 1449, "lr": 5.401903153776259e-05}
{"train_loss": 0.08892229199409485, "global_step": 129035, "epoch": 1449, "lr": 5.4018453676776414e-05}
{"train_loss": 0.15101031959056854, "global_step": 129036, "epoch": 1449, "lr": 5.4017875815250016e-05}
{"train_loss": 0.1019289642572403, "global_step": 129037, "epoch": 1449, "lr": 5.401729795318346e-05}
{"train_loss": 0.12076587229967117, "global_step": 129038, "epoch": 1449, "lr": 5.401672009057683e-05}
{"train_loss": 0.10655013471841812, "global_step": 129039, "epoch": 1449, "lr": 5.40161422274302e-05}
{"train_loss": 0.13759934902191162, "global_step": 129040, "epoch": 1449, "lr": 5.401556436374366e-05}
{"train_loss": 0.12442690879106522, "global_step": 129041, "epoch": 1449, "lr": 5.4014986499517276e-05}
{"train_loss": 0.08271566033363342, "global_step": 129042, "epoch": 1449, "lr": 5.401440863475111e-05}
{"train_loss": 0.1501649171113968, "global_step": 129043, "epoch": 1449, "lr": 5.401383076944526e-05}
{"train_loss": 0.16830962896347046, "global_step": 129044, "epoch": 1449, "lr": 5.4013252903599795e-05}
{"train_loss": 0.12820373475551605, "global_step": 129045, "epoch": 1449, "lr": 5.401267503721481e-05}
{"train_loss": 0.11216168105602264, "global_step": 129046, "epoch": 1449, "lr": 5.4012097170290366e-05}
{"train_loss": 0.20135071873664856, "global_step": 129047, "epoch": 1449, "lr": 5.401151930282654e-05}
{"train_loss": 0.16755405068397522, "global_step": 129048, "epoch": 1449, "lr": 5.4010941434823415e-05}
{"train_loss": 0.13129524953579635, "global_step": 129049, "epoch": 1449, "lr": 5.4010363566281064e-05, "val_loss": 5.5888991355896}
{"train_loss": 0.12800052762031555, "global_step": 129050, "epoch": 1450, "lr": 5.400978569719958e-05}
{"train_loss": 0.16535575687885284, "global_step": 129051, "epoch": 1450, "lr": 5.400920782757902e-05}
{"train_loss": 0.1903413087129593, "global_step": 129052, "epoch": 1450, "lr": 5.400862995741946e-05}
{"train_loss": 0.10513178259134293, "global_step": 129053, "epoch": 1450, "lr": 5.400805208672099e-05}
{"train_loss": 0.1439855545759201, "global_step": 129054, "epoch": 1450, "lr": 5.400747421548371e-05}
{"train_loss": 0.13735049962997437, "global_step": 129055, "epoch": 1450, "lr": 5.4006896343707656e-05}
{"train_loss": 0.07681083679199219, "global_step": 129056, "epoch": 1450, "lr": 5.400631847139294e-05}
{"train_loss": 0.1088540256023407, "global_step": 129057, "epoch": 1450, "lr": 5.40057405985396e-05}
{"train_loss": 0.14828167855739594, "global_step": 129058, "epoch": 1450, "lr": 5.400516272514775e-05}
{"train_loss": 0.09303680807352066, "global_step": 129059, "epoch": 1450, "lr": 5.4004584851217464e-05}
{"train_loss": 0.20640449225902557, "global_step": 129060, "epoch": 1450, "lr": 5.400400697674881e-05}
{"train_loss": 0.12410125136375427, "global_step": 129061, "epoch": 1450, "lr": 5.4003429101741855e-05}
{"train_loss": 0.1391502171754837, "global_step": 129062, "epoch": 1450, "lr": 5.400285122619668e-05}
{"train_loss": 0.095376156270504, "global_step": 129063, "epoch": 1450, "lr": 5.400227335011339e-05}
{"train_loss": 0.12375294417142868, "global_step": 129064, "epoch": 1450, "lr": 5.400169547349203e-05}
{"train_loss": 0.13693085312843323, "global_step": 129065, "epoch": 1450, "lr": 5.4001117596332696e-05}
{"train_loss": 0.12086735665798187, "global_step": 129066, "epoch": 1450, "lr": 5.400053971863546e-05}
{"train_loss": 0.11748148500919342, "global_step": 129067, "epoch": 1450, "lr": 5.399996184040039e-05}
{"train_loss": 0.13695085048675537, "global_step": 129068, "epoch": 1450, "lr": 5.399938396162759e-05}
{"train_loss": 0.1249622106552124, "global_step": 129069, "epoch": 1450, "lr": 5.399880608231711e-05}
{"train_loss": 0.15009871125221252, "global_step": 129070, "epoch": 1450, "lr": 5.399822820246905e-05}
{"train_loss": 0.2109415978193283, "global_step": 129071, "epoch": 1450, "lr": 5.399765032208345e-05}
{"train_loss": 0.15782633423805237, "global_step": 129072, "epoch": 1450, "lr": 5.3997072441160436e-05}
{"train_loss": 0.11048891395330429, "global_step": 129073, "epoch": 1450, "lr": 5.399649455970006e-05}
{"train_loss": 0.1631045937538147, "global_step": 129074, "epoch": 1450, "lr": 5.399591667770241e-05}
{"train_loss": 0.1198677346110344, "global_step": 129075, "epoch": 1450, "lr": 5.399533879516756e-05}
{"train_loss": 0.11918380111455917, "global_step": 129076, "epoch": 1450, "lr": 5.399476091209558e-05}
{"train_loss": 0.07631760090589523, "global_step": 129077, "epoch": 1450, "lr": 5.399418302848656e-05}
{"train_loss": 0.0909314751625061, "global_step": 129078, "epoch": 1450, "lr": 5.3993605144340565e-05}
{"train_loss": 0.10290119051933289, "global_step": 129079, "epoch": 1450, "lr": 5.399302725965768e-05}
{"train_loss": 0.10087984055280685, "global_step": 129080, "epoch": 1450, "lr": 5.3992449374437984e-05}
{"train_loss": 0.11645505577325821, "global_step": 129081, "epoch": 1450, "lr": 5.399187148868153e-05}
{"train_loss": 0.12575283646583557, "global_step": 129082, "epoch": 1450, "lr": 5.399129360238844e-05}
{"train_loss": 0.13869009912014008, "global_step": 129083, "epoch": 1450, "lr": 5.399071571555877e-05}
{"train_loss": 0.1669497787952423, "global_step": 129084, "epoch": 1450, "lr": 5.399013782819259e-05}
{"train_loss": 0.11874829232692719, "global_step": 129085, "epoch": 1450, "lr": 5.398955994028998e-05}
{"train_loss": 0.11995283514261246, "global_step": 129086, "epoch": 1450, "lr": 5.398898205185103e-05}
{"train_loss": 0.1267005205154419, "global_step": 129087, "epoch": 1450, "lr": 5.398840416287581e-05}
{"train_loss": 0.1970125436782837, "global_step": 129088, "epoch": 1450, "lr": 5.3987826273364396e-05}
{"train_loss": 0.07452153414487839, "global_step": 129089, "epoch": 1450, "lr": 5.398724838331685e-05}
{"train_loss": 0.08615699410438538, "global_step": 129090, "epoch": 1450, "lr": 5.398667049273328e-05}
{"train_loss": 0.1466521918773651, "global_step": 129091, "epoch": 1450, "lr": 5.3986092601613756e-05}
{"train_loss": 0.11731144040822983, "global_step": 129092, "epoch": 1450, "lr": 5.398551470995835e-05}
{"train_loss": 0.12817329168319702, "global_step": 129093, "epoch": 1450, "lr": 5.398493681776714e-05}
{"train_loss": 0.10560736060142517, "global_step": 129094, "epoch": 1450, "lr": 5.39843589250402e-05}
{"train_loss": 0.14510488510131836, "global_step": 129095, "epoch": 1450, "lr": 5.398378103177761e-05}
{"train_loss": 0.11978261917829514, "global_step": 129096, "epoch": 1450, "lr": 5.3983203137979455e-05}
{"train_loss": 0.20770417153835297, "global_step": 129097, "epoch": 1450, "lr": 5.3982625243645804e-05}
{"train_loss": 0.16254428029060364, "global_step": 129098, "epoch": 1450, "lr": 5.3982047348776754e-05}
{"train_loss": 0.1538459211587906, "global_step": 129099, "epoch": 1450, "lr": 5.3981469453372345e-05}
{"train_loss": 0.1698838770389557, "global_step": 129100, "epoch": 1450, "lr": 5.3980891557432686e-05}
{"train_loss": 0.07850717008113861, "global_step": 129101, "epoch": 1450, "lr": 5.398031366095784e-05}
{"train_loss": 0.13091759383678436, "global_step": 129102, "epoch": 1450, "lr": 5.397973576394789e-05}
{"train_loss": 0.13722147047519684, "global_step": 129103, "epoch": 1450, "lr": 5.3979157866402906e-05}
{"train_loss": 0.1076917052268982, "global_step": 129104, "epoch": 1450, "lr": 5.3978579968322984e-05}
{"train_loss": 0.1429852545261383, "global_step": 129105, "epoch": 1450, "lr": 5.397800206970819e-05}
{"train_loss": 0.08819236606359482, "global_step": 129106, "epoch": 1450, "lr": 5.397742417055859e-05}
{"train_loss": 0.11494060605764389, "global_step": 129107, "epoch": 1450, "lr": 5.397684627087427e-05}
{"train_loss": 0.16408586502075195, "global_step": 129108, "epoch": 1450, "lr": 5.397626837065532e-05}
{"train_loss": 0.13780850172042847, "global_step": 129109, "epoch": 1450, "lr": 5.397569046990182e-05}
{"train_loss": 0.16604040563106537, "global_step": 129110, "epoch": 1450, "lr": 5.397511256861383e-05}
{"train_loss": 0.12123336642980576, "global_step": 129111, "epoch": 1450, "lr": 5.3974534666791434e-05}
{"train_loss": 0.11988602578639984, "global_step": 129112, "epoch": 1450, "lr": 5.397395676443471e-05}
{"train_loss": 0.14160162210464478, "global_step": 129113, "epoch": 1450, "lr": 5.3973378861543733e-05}
{"train_loss": 0.07772617787122726, "global_step": 129114, "epoch": 1450, "lr": 5.397280095811859e-05}
{"train_loss": 0.11980626732110977, "global_step": 129115, "epoch": 1450, "lr": 5.397222305415934e-05}
{"train_loss": 0.10932060331106186, "global_step": 129116, "epoch": 1450, "lr": 5.397164514966609e-05}
{"train_loss": 0.10962240397930145, "global_step": 129117, "epoch": 1450, "lr": 5.397106724463891e-05}
{"train_loss": 0.1413416564464569, "global_step": 129118, "epoch": 1450, "lr": 5.3970489339077844e-05}
{"train_loss": 0.11274220794439316, "global_step": 129119, "epoch": 1450, "lr": 5.3969911432983e-05}
{"train_loss": 0.07340674847364426, "global_step": 129120, "epoch": 1450, "lr": 5.396933352635445e-05}
{"train_loss": 0.10813938826322556, "global_step": 129121, "epoch": 1450, "lr": 5.396875561919228e-05}
{"train_loss": 0.0790012925863266, "global_step": 129122, "epoch": 1450, "lr": 5.396817771149655e-05}
{"train_loss": 0.11398462951183319, "global_step": 129123, "epoch": 1450, "lr": 5.396759980326734e-05}
{"train_loss": 0.1556500941514969, "global_step": 129124, "epoch": 1450, "lr": 5.3967021894504744e-05}
{"train_loss": 0.138076514005661, "global_step": 129125, "epoch": 1450, "lr": 5.3966443985208825e-05}
{"train_loss": 0.18190854787826538, "global_step": 129126, "epoch": 1450, "lr": 5.396586607537967e-05}
{"train_loss": 0.10552351176738739, "global_step": 129127, "epoch": 1450, "lr": 5.396528816501735e-05}
{"train_loss": 0.1315123736858368, "global_step": 129128, "epoch": 1450, "lr": 5.3964710254121955e-05}
{"train_loss": 0.18582642078399658, "global_step": 129129, "epoch": 1450, "lr": 5.396413234269354e-05}
{"train_loss": 0.10970471054315567, "global_step": 129130, "epoch": 1450, "lr": 5.39635544307322e-05}
{"train_loss": 0.076143279671669, "global_step": 129131, "epoch": 1450, "lr": 5.396297651823802e-05}
{"train_loss": 0.053865719586610794, "global_step": 129132, "epoch": 1450, "lr": 5.396239860521105e-05}
{"train_loss": 0.07660189270973206, "global_step": 129133, "epoch": 1450, "lr": 5.396182069165139e-05}
{"train_loss": 0.08879609405994415, "global_step": 129134, "epoch": 1450, "lr": 5.396124277755912e-05}
{"train_loss": 0.11565735936164856, "global_step": 129135, "epoch": 1450, "lr": 5.39606648629343e-05}
{"train_loss": 0.1272008866071701, "global_step": 129136, "epoch": 1450, "lr": 5.396008694777702e-05}
{"train_loss": 0.1135641559958458, "global_step": 129137, "epoch": 1450, "lr": 5.395950903208735e-05}
{"train_loss": 0.12623791961690015, "global_step": 129138, "epoch": 1450, "lr": 5.3958931115865373e-05, "train/sim_max_reward_0": 0.7140702602735799, "train/sim_max_reward_1": 0.938375853326559, "train/sim_max_reward_2": 0.1802055565236731, "train/sim_max_reward_3": 0.43494633939697463, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.44278095056078876, "test/sim_max_reward_4500000": 1.0, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9879838456982045, "test/sim_max_reward_4500003": 0.47521313909156054, "test/sim_max_reward_4500004": 0.715718632675148, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.9684291300109975, "test/sim_max_reward_4500007": 0.9597872222132262, "test/sim_max_reward_4500008": 0.16047111846287299, "test/sim_max_reward_4500009": 0.34350592295061577, "test/sim_max_reward_4500010": 1.0, "test/sim_max_reward_4500011": 0.9853589808502299, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.19327356399474108, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.4785396836604719, "test/sim_max_reward_4500017": 0.9480373125712308, "test/sim_max_reward_4500018": 0.9612652671755935, "test/sim_max_reward_4500019": 0.48383996753691716, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9992025776291067, "test/sim_max_reward_4500023": 1.0, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.0, "test/sim_max_reward_4500027": 0.9631686982891933, "test/sim_max_reward_4500028": 0.4937663568673415, "test/sim_max_reward_4500029": 0.9998737474785306, "test/sim_max_reward_4500030": 0.712381058261623, "test/sim_max_reward_4500031": 0.5461474204767514, "test/sim_max_reward_4500032": 0.7058660987768436, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.946644170456099, "test/sim_max_reward_4500035": 0.9961160597427983, "test/sim_max_reward_4500036": 0.9808456005965723, "test/sim_max_reward_4500037": 0.9958627396518329, "test/sim_max_reward_4500038": 0.17994198080438648, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9782763765407444, "test/sim_max_reward_4500041": 0.5777701542612657, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.5911182754904384, "test/sim_max_reward_4500045": 0.6586885374516083, "test/sim_max_reward_4500046": 0.3951476546842378, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.766261868208062, "train/mean_score": 0.6183964933469293, "test/mean_score": 0.6829700632511849, "val_loss": 5.508387565612793, "train_action_mse_error": 36.845497131347656}
{"train_loss": 0.15122337639331818, "global_step": 129139, "epoch": 1451, "lr": 5.395835319911118e-05}
{"train_loss": 0.12763535976409912, "global_step": 129140, "epoch": 1451, "lr": 5.395777528182482e-05}
{"train_loss": 0.11427830904722214, "global_step": 129141, "epoch": 1451, "lr": 5.395719736400638e-05}
{"train_loss": 0.1248198077082634, "global_step": 129142, "epoch": 1451, "lr": 5.395661944565595e-05}
{"train_loss": 0.16074615716934204, "global_step": 129143, "epoch": 1451, "lr": 5.395604152677361e-05}
{"train_loss": 0.10576049983501434, "global_step": 129144, "epoch": 1451, "lr": 5.3955463607359406e-05}
{"train_loss": 0.1390240490436554, "global_step": 129145, "epoch": 1451, "lr": 5.395488568741345e-05}
{"train_loss": 0.09840293973684311, "global_step": 129146, "epoch": 1451, "lr": 5.395430776693581e-05}
{"train_loss": 0.1657358556985855, "global_step": 129147, "epoch": 1451, "lr": 5.395372984592656e-05}
{"train_loss": 0.14360877871513367, "global_step": 129148, "epoch": 1451, "lr": 5.395315192438578e-05}
{"train_loss": 0.13143393397331238, "global_step": 129149, "epoch": 1451, "lr": 5.395257400231355e-05}
{"train_loss": 0.1534494012594223, "global_step": 129150, "epoch": 1451, "lr": 5.3951996079709944e-05}
{"train_loss": 0.08005128055810928, "global_step": 129151, "epoch": 1451, "lr": 5.395141815657504e-05}
{"train_loss": 0.09947439283132553, "global_step": 129152, "epoch": 1451, "lr": 5.3950840232908915e-05}
{"train_loss": 0.10634114593267441, "global_step": 129153, "epoch": 1451, "lr": 5.395026230871164e-05}
{"train_loss": 0.14198622107505798, "global_step": 129154, "epoch": 1451, "lr": 5.394968438398331e-05}
{"train_loss": 0.07677165418863297, "global_step": 129155, "epoch": 1451, "lr": 5.3949106458724e-05}
{"train_loss": 0.17477478086948395, "global_step": 129156, "epoch": 1451, "lr": 5.3948528532933774e-05}
{"train_loss": 0.15066957473754883, "global_step": 129157, "epoch": 1451, "lr": 5.394795060661272e-05}
{"train_loss": 0.10300508141517639, "global_step": 129158, "epoch": 1451, "lr": 5.39473726797609e-05}
{"train_loss": 0.0888349860906601, "global_step": 129159, "epoch": 1451, "lr": 5.394679475237842e-05}
{"train_loss": 0.11491961777210236, "global_step": 129160, "epoch": 1451, "lr": 5.3946216824465326e-05}
{"train_loss": 0.11368509382009506, "global_step": 129161, "epoch": 1451, "lr": 5.394563889602171e-05}
{"train_loss": 0.09325104206800461, "global_step": 129162, "epoch": 1451, "lr": 5.3945060967047646e-05}
{"train_loss": 0.07944983243942261, "global_step": 129163, "epoch": 1451, "lr": 5.3944483037543236e-05}
{"train_loss": 0.053975753486156464, "global_step": 129164, "epoch": 1451, "lr": 5.394390510750853e-05}
{"train_loss": 0.0908212885260582, "global_step": 129165, "epoch": 1451, "lr": 5.3943327176943614e-05}
{"train_loss": 0.12784188985824585, "global_step": 129166, "epoch": 1451, "lr": 5.394274924584857e-05}
{"train_loss": 0.1607319414615631, "global_step": 129167, "epoch": 1451, "lr": 5.394217131422347e-05}
{"train_loss": 0.11207695305347443, "global_step": 129168, "epoch": 1451, "lr": 5.394159338206838e-05}
{"train_loss": 0.15117822587490082, "global_step": 129169, "epoch": 1451, "lr": 5.39410154493834e-05}
{"train_loss": 0.15288497507572174, "global_step": 129170, "epoch": 1451, "lr": 5.394043751616861e-05}
{"train_loss": 0.04600605368614197, "global_step": 129171, "epoch": 1451, "lr": 5.393985958242406e-05}
{"train_loss": 0.07240207493305206, "global_step": 129172, "epoch": 1451, "lr": 5.393928164814985e-05}
{"train_loss": 0.09909205138683319, "global_step": 129173, "epoch": 1451, "lr": 5.393870371334605e-05}
{"train_loss": 0.2215508222579956, "global_step": 129174, "epoch": 1451, "lr": 5.393812577801274e-05}
{"train_loss": 0.087342768907547, "global_step": 129175, "epoch": 1451, "lr": 5.393754784215002e-05}
{"train_loss": 0.12831078469753265, "global_step": 129176, "epoch": 1451, "lr": 5.3936969905757915e-05}
{"train_loss": 0.10301515460014343, "global_step": 129177, "epoch": 1451, "lr": 5.393639196883654e-05}
{"train_loss": 0.11954967677593231, "global_step": 129178, "epoch": 1451, "lr": 5.393581403138597e-05}
{"train_loss": 0.10948184877634048, "global_step": 129179, "epoch": 1451, "lr": 5.393523609340627e-05}
{"train_loss": 0.11236266046762466, "global_step": 129180, "epoch": 1451, "lr": 5.393465815489752e-05}
{"train_loss": 0.1537468433380127, "global_step": 129181, "epoch": 1451, "lr": 5.393408021585981e-05}
{"train_loss": 0.09339571744203568, "global_step": 129182, "epoch": 1451, "lr": 5.3933502276293216e-05}
{"train_loss": 0.0398113876581192, "global_step": 129183, "epoch": 1451, "lr": 5.39329243361978e-05}
{"train_loss": 0.06195744872093201, "global_step": 129184, "epoch": 1451, "lr": 5.393234639557366e-05}
{"train_loss": 0.09724626690149307, "global_step": 129185, "epoch": 1451, "lr": 5.393176845442086e-05}
{"train_loss": 0.1758740246295929, "global_step": 129186, "epoch": 1451, "lr": 5.3931190512739484e-05}
{"train_loss": 0.12701094150543213, "global_step": 129187, "epoch": 1451, "lr": 5.393061257052961e-05}
{"train_loss": 0.11988730728626251, "global_step": 129188, "epoch": 1451, "lr": 5.39300346277913e-05}
{"train_loss": 0.14471305906772614, "global_step": 129189, "epoch": 1451, "lr": 5.392945668452466e-05}
{"train_loss": 0.11877396702766418, "global_step": 129190, "epoch": 1451, "lr": 5.392887874072975e-05}
{"train_loss": 0.15564976632595062, "global_step": 129191, "epoch": 1451, "lr": 5.392830079640664e-05}
{"train_loss": 0.14081108570098877, "global_step": 129192, "epoch": 1451, "lr": 5.392772285155543e-05}
{"train_loss": 0.1645776778459549, "global_step": 129193, "epoch": 1451, "lr": 5.392714490617618e-05}
{"train_loss": 0.17013126611709595, "global_step": 129194, "epoch": 1451, "lr": 5.392656696026899e-05}
{"train_loss": 0.18812890350818634, "global_step": 129195, "epoch": 1451, "lr": 5.39259890138339e-05}
{"train_loss": 0.0955345556139946, "global_step": 129196, "epoch": 1451, "lr": 5.3925411066871016e-05}
{"train_loss": 0.14322130382061005, "global_step": 129197, "epoch": 1451, "lr": 5.39248331193804e-05}
{"train_loss": 0.16743089258670807, "global_step": 129198, "epoch": 1451, "lr": 5.392425517136212e-05}
{"train_loss": 0.1984708309173584, "global_step": 129199, "epoch": 1451, "lr": 5.39236772228163e-05}
{"train_loss": 0.2222515046596527, "global_step": 129200, "epoch": 1451, "lr": 5.392309927374299e-05}
{"train_loss": 0.10094237327575684, "global_step": 129201, "epoch": 1451, "lr": 5.392252132414226e-05}
{"train_loss": 0.18491721153259277, "global_step": 129202, "epoch": 1451, "lr": 5.3921943374014195e-05}
{"train_loss": 0.12702572345733643, "global_step": 129203, "epoch": 1451, "lr": 5.392136542335887e-05}
{"train_loss": 0.12937691807746887, "global_step": 129204, "epoch": 1451, "lr": 5.392078747217637e-05}
{"train_loss": 0.1193641722202301, "global_step": 129205, "epoch": 1451, "lr": 5.392020952046677e-05}
{"train_loss": 0.18642215430736542, "global_step": 129206, "epoch": 1451, "lr": 5.3919631568230145e-05}
{"train_loss": 0.13221849501132965, "global_step": 129207, "epoch": 1451, "lr": 5.3919053615466576e-05}
{"train_loss": 0.09982214868068695, "global_step": 129208, "epoch": 1451, "lr": 5.3918475662176125e-05}
{"train_loss": 0.06863121688365936, "global_step": 129209, "epoch": 1451, "lr": 5.391789770835889e-05}
{"train_loss": 0.1522919088602066, "global_step": 129210, "epoch": 1451, "lr": 5.391731975401495e-05}
{"train_loss": 0.21198731660842896, "global_step": 129211, "epoch": 1451, "lr": 5.3916741799144365e-05}
{"train_loss": 0.19053643941879272, "global_step": 129212, "epoch": 1451, "lr": 5.391616384374723e-05}
{"train_loss": 0.09003699570894241, "global_step": 129213, "epoch": 1451, "lr": 5.3915585887823596e-05}
{"train_loss": 0.16684122383594513, "global_step": 129214, "epoch": 1451, "lr": 5.3915007931373584e-05}
{"train_loss": 0.11659231781959534, "global_step": 129215, "epoch": 1451, "lr": 5.391442997439723e-05}
{"train_loss": 0.10739152878522873, "global_step": 129216, "epoch": 1451, "lr": 5.391385201689463e-05}
{"train_loss": 0.1569823920726776, "global_step": 129217, "epoch": 1451, "lr": 5.391327405886586e-05}
{"train_loss": 0.12356674671173096, "global_step": 129218, "epoch": 1451, "lr": 5.3912696100311e-05}
{"train_loss": 0.11542253196239471, "global_step": 129219, "epoch": 1451, "lr": 5.391211814123012e-05}
{"train_loss": 0.20094315707683563, "global_step": 129220, "epoch": 1451, "lr": 5.391154018162331e-05}
{"train_loss": 0.15494143962860107, "global_step": 129221, "epoch": 1451, "lr": 5.391096222149065e-05}
{"train_loss": 0.12950187921524048, "global_step": 129222, "epoch": 1451, "lr": 5.391038426083219e-05}
{"train_loss": 0.16765457391738892, "global_step": 129223, "epoch": 1451, "lr": 5.3909806299648035e-05}
{"train_loss": 0.14687058329582214, "global_step": 129224, "epoch": 1451, "lr": 5.3909228337938254e-05}
{"train_loss": 0.15925778448581696, "global_step": 129225, "epoch": 1451, "lr": 5.390865037570293e-05}
{"train_loss": 0.09213690459728241, "global_step": 129226, "epoch": 1451, "lr": 5.390807241294213e-05}
{"train_loss": 0.1293462159928311, "global_step": 129227, "epoch": 1451, "lr": 5.390749444965594e-05, "val_loss": 5.399654865264893}
{"train_loss": 0.13877834379673004, "global_step": 129228, "epoch": 1452, "lr": 5.390691648584444e-05}
{"train_loss": 0.08247506618499756, "global_step": 129229, "epoch": 1452, "lr": 5.3906338521507696e-05}
{"train_loss": 0.10099758207798004, "global_step": 129230, "epoch": 1452, "lr": 5.390576055664579e-05}
{"train_loss": 0.15149396657943726, "global_step": 129231, "epoch": 1452, "lr": 5.390518259125881e-05}
{"train_loss": 0.15588714182376862, "global_step": 129232, "epoch": 1452, "lr": 5.390460462534681e-05}
{"train_loss": 0.11624792218208313, "global_step": 129233, "epoch": 1452, "lr": 5.390402665890991e-05}
{"train_loss": 0.09124571830034256, "global_step": 129234, "epoch": 1452, "lr": 5.3903448691948134e-05}
{"train_loss": 0.17774759232997894, "global_step": 129235, "epoch": 1452, "lr": 5.3902870724461594e-05}
{"train_loss": 0.15960700809955597, "global_step": 129236, "epoch": 1452, "lr": 5.390229275645037e-05}
{"train_loss": 0.12182047218084335, "global_step": 129237, "epoch": 1452, "lr": 5.3901714787914524e-05}
{"train_loss": 0.13013891875743866, "global_step": 129238, "epoch": 1452, "lr": 5.3901136818854136e-05}
{"train_loss": 0.07105734944343567, "global_step": 129239, "epoch": 1452, "lr": 5.39005588492693e-05}
{"train_loss": 0.07981055229902267, "global_step": 129240, "epoch": 1452, "lr": 5.3899980879160074e-05}
{"train_loss": 0.1646706610918045, "global_step": 129241, "epoch": 1452, "lr": 5.389940290852654e-05}
{"train_loss": 0.0975906252861023, "global_step": 129242, "epoch": 1452, "lr": 5.3898824937368786e-05}
{"train_loss": 0.2189088612794876, "global_step": 129243, "epoch": 1452, "lr": 5.3898246965686885e-05}
{"train_loss": 0.05712960660457611, "global_step": 129244, "epoch": 1452, "lr": 5.38976689934809e-05}
{"train_loss": 0.10324974358081818, "global_step": 129245, "epoch": 1452, "lr": 5.389709102075092e-05}
{"train_loss": 0.12190275639295578, "global_step": 129246, "epoch": 1452, "lr": 5.389651304749703e-05}
{"train_loss": 0.1828509271144867, "global_step": 129247, "epoch": 1452, "lr": 5.3895935073719306e-05}
{"train_loss": 0.17541491985321045, "global_step": 129248, "epoch": 1452, "lr": 5.389535709941782e-05}
{"train_loss": 0.18571554124355316, "global_step": 129249, "epoch": 1452, "lr": 5.3894779124592645e-05}
{"train_loss": 0.09762392193078995, "global_step": 129250, "epoch": 1452, "lr": 5.389420114924387e-05}
{"train_loss": 0.14877912402153015, "global_step": 129251, "epoch": 1452, "lr": 5.389362317337157e-05}
{"train_loss": 0.15682467818260193, "global_step": 129252, "epoch": 1452, "lr": 5.389304519697581e-05}
{"train_loss": 0.08799397945404053, "global_step": 129253, "epoch": 1452, "lr": 5.3892467220056675e-05}
{"train_loss": 0.10450156778097153, "global_step": 129254, "epoch": 1452, "lr": 5.389188924261426e-05}
{"train_loss": 0.08642251044511795, "global_step": 129255, "epoch": 1452, "lr": 5.3891311264648614e-05}
{"train_loss": 0.08821409940719604, "global_step": 129256, "epoch": 1452, "lr": 5.3890733286159835e-05}
{"train_loss": 0.09964746981859207, "global_step": 129257, "epoch": 1452, "lr": 5.3890155307147995e-05}
{"train_loss": 0.10566253215074539, "global_step": 129258, "epoch": 1452, "lr": 5.388957732761317e-05}
{"train_loss": 0.13024012744426727, "global_step": 129259, "epoch": 1452, "lr": 5.3888999347555436e-05}
{"train_loss": 0.14708061516284943, "global_step": 129260, "epoch": 1452, "lr": 5.388842136697487e-05}
{"train_loss": 0.10506204515695572, "global_step": 129261, "epoch": 1452, "lr": 5.388784338587156e-05}
{"train_loss": 0.2069845348596573, "global_step": 129262, "epoch": 1452, "lr": 5.388726540424558e-05}
{"train_loss": 0.11764548718929291, "global_step": 129263, "epoch": 1452, "lr": 5.388668742209701e-05}
{"train_loss": 0.09836842864751816, "global_step": 129264, "epoch": 1452, "lr": 5.3886109439425904e-05}
{"train_loss": 0.09228701889514923, "global_step": 129265, "epoch": 1452, "lr": 5.3885531456232365e-05}
{"train_loss": 0.11958213150501251, "global_step": 129266, "epoch": 1452, "lr": 5.3884953472516464e-05}
{"train_loss": 0.12634724378585815, "global_step": 129267, "epoch": 1452, "lr": 5.388437548827828e-05}
{"train_loss": 0.18359512090682983, "global_step": 129268, "epoch": 1452, "lr": 5.3883797503517894e-05}
{"train_loss": 0.18440935015678406, "global_step": 129269, "epoch": 1452, "lr": 5.3883219518235375e-05}
{"train_loss": 0.11199602484703064, "global_step": 129270, "epoch": 1452, "lr": 5.38826415324308e-05}
{"train_loss": 0.18938742578029633, "global_step": 129271, "epoch": 1452, "lr": 5.3882063546104255e-05}
{"train_loss": 0.10005968809127808, "global_step": 129272, "epoch": 1452, "lr": 5.388148555925581e-05}
{"train_loss": 0.09918830543756485, "global_step": 129273, "epoch": 1452, "lr": 5.388090757188555e-05}
{"train_loss": 0.14351564645767212, "global_step": 129274, "epoch": 1452, "lr": 5.388032958399355e-05}
{"train_loss": 0.15935175120830536, "global_step": 129275, "epoch": 1452, "lr": 5.3879751595579885e-05}
{"train_loss": 0.18788981437683105, "global_step": 129276, "epoch": 1452, "lr": 5.387917360664464e-05}
{"train_loss": 0.15949349105358124, "global_step": 129277, "epoch": 1452, "lr": 5.3878595617187876e-05}
{"train_loss": 0.18285363912582397, "global_step": 129278, "epoch": 1452, "lr": 5.38780176272097e-05}
{"train_loss": 0.1458207219839096, "global_step": 129279, "epoch": 1452, "lr": 5.387743963671015e-05}
{"train_loss": 0.06594458222389221, "global_step": 129280, "epoch": 1452, "lr": 5.387686164568934e-05}
{"train_loss": 0.09698399156332016, "global_step": 129281, "epoch": 1452, "lr": 5.387628365414733e-05}
{"train_loss": 0.1075860932469368, "global_step": 129282, "epoch": 1452, "lr": 5.3875705662084195e-05}
{"train_loss": 0.1120641753077507, "global_step": 129283, "epoch": 1452, "lr": 5.387512766950003e-05}
{"train_loss": 0.13807784020900726, "global_step": 129284, "epoch": 1452, "lr": 5.3874549676394895e-05}
{"train_loss": 0.11808668822050095, "global_step": 129285, "epoch": 1452, "lr": 5.387397168276888e-05}
{"train_loss": 0.1860031932592392, "global_step": 129286, "epoch": 1452, "lr": 5.3873393688622044e-05}
{"train_loss": 0.08129459619522095, "global_step": 129287, "epoch": 1452, "lr": 5.387281569395448e-05}
{"train_loss": 0.1966828554868698, "global_step": 129288, "epoch": 1452, "lr": 5.387223769876627e-05}
{"train_loss": 0.1272396445274353, "global_step": 129289, "epoch": 1452, "lr": 5.387165970305749e-05}
{"train_loss": 0.10457506030797958, "global_step": 129290, "epoch": 1452, "lr": 5.38710817068282e-05}
{"train_loss": 0.10870248079299927, "global_step": 129291, "epoch": 1452, "lr": 5.387050371007849e-05}
{"train_loss": 0.11241332441568375, "global_step": 129292, "epoch": 1452, "lr": 5.3869925712808446e-05}
{"train_loss": 0.1403159201145172, "global_step": 129293, "epoch": 1452, "lr": 5.386934771501814e-05}
{"train_loss": 0.14950263500213623, "global_step": 129294, "epoch": 1452, "lr": 5.3868769716707635e-05}
{"train_loss": 0.13207675516605377, "global_step": 129295, "epoch": 1452, "lr": 5.3868191717877034e-05}
{"train_loss": 0.09826929122209549, "global_step": 129296, "epoch": 1452, "lr": 5.38676137185264e-05}
{"train_loss": 0.05860830470919609, "global_step": 129297, "epoch": 1452, "lr": 5.38670357186558e-05}
{"train_loss": 0.151865154504776, "global_step": 129298, "epoch": 1452, "lr": 5.386645771826534e-05}
{"train_loss": 0.13972415030002594, "global_step": 129299, "epoch": 1452, "lr": 5.3865879717355074e-05}
{"train_loss": 0.09268810600042343, "global_step": 129300, "epoch": 1452, "lr": 5.3865301715925085e-05}
{"train_loss": 0.10959426313638687, "global_step": 129301, "epoch": 1452, "lr": 5.386472371397545e-05}
{"train_loss": 0.10477294027805328, "global_step": 129302, "epoch": 1452, "lr": 5.386414571150627e-05}
{"train_loss": 0.1719997078180313, "global_step": 129303, "epoch": 1452, "lr": 5.386356770851758e-05}
{"train_loss": 0.20784702897071838, "global_step": 129304, "epoch": 1452, "lr": 5.3862989705009495e-05}
{"train_loss": 0.10196349024772644, "global_step": 129305, "epoch": 1452, "lr": 5.3862411700982076e-05}
{"train_loss": 0.15115578472614288, "global_step": 129306, "epoch": 1452, "lr": 5.386183369643539e-05}
{"train_loss": 0.17227749526500702, "global_step": 129307, "epoch": 1452, "lr": 5.386125569136954e-05}
{"train_loss": 0.07927249372005463, "global_step": 129308, "epoch": 1452, "lr": 5.386067768578459e-05}
{"train_loss": 0.08748889714479446, "global_step": 129309, "epoch": 1452, "lr": 5.386009967968061e-05}
{"train_loss": 0.1101120114326477, "global_step": 129310, "epoch": 1452, "lr": 5.38595216730577e-05}
{"train_loss": 0.17511287331581116, "global_step": 129311, "epoch": 1452, "lr": 5.3858943665915915e-05}
{"train_loss": 0.12232886999845505, "global_step": 129312, "epoch": 1452, "lr": 5.385836565825535e-05}
{"train_loss": 0.17489448189735413, "global_step": 129313, "epoch": 1452, "lr": 5.3857787650076065e-05}
{"train_loss": 0.1044553816318512, "global_step": 129314, "epoch": 1452, "lr": 5.3857209641378147e-05}
{"train_loss": 0.17961227893829346, "global_step": 129315, "epoch": 1452, "lr": 5.385663163216169e-05}
{"train_loss": 0.12974052863676896, "global_step": 129316, "epoch": 1452, "lr": 5.385605362242674e-05, "val_loss": 5.767844200134277}
{"train_loss": 0.07810581475496292, "global_step": 129317, "epoch": 1453, "lr": 5.38554756121734e-05}
{"train_loss": 0.11386163532733917, "global_step": 129318, "epoch": 1453, "lr": 5.385489760140173e-05}
{"train_loss": 0.12740568816661835, "global_step": 129319, "epoch": 1453, "lr": 5.3854319590111824e-05}
{"train_loss": 0.11709851771593094, "global_step": 129320, "epoch": 1453, "lr": 5.385374157830375e-05}
{"train_loss": 0.10202329605817795, "global_step": 129321, "epoch": 1453, "lr": 5.385316356597758e-05}
{"train_loss": 0.1297607421875, "global_step": 129322, "epoch": 1453, "lr": 5.3852585553133405e-05}
{"train_loss": 0.09412389248609543, "global_step": 129323, "epoch": 1453, "lr": 5.3852007539771296e-05}
{"train_loss": 0.12205091118812561, "global_step": 129324, "epoch": 1453, "lr": 5.385142952589134e-05}
{"train_loss": 0.15413731336593628, "global_step": 129325, "epoch": 1453, "lr": 5.3850851511493594e-05}
{"train_loss": 0.12907922267913818, "global_step": 129326, "epoch": 1453, "lr": 5.385027349657815e-05}
{"train_loss": 0.17589235305786133, "global_step": 129327, "epoch": 1453, "lr": 5.384969548114509e-05}
{"train_loss": 0.12083569169044495, "global_step": 129328, "epoch": 1453, "lr": 5.384911746519448e-05}
{"train_loss": 0.1385699063539505, "global_step": 129329, "epoch": 1453, "lr": 5.38485394487264e-05}
{"train_loss": 0.09110593050718307, "global_step": 129330, "epoch": 1453, "lr": 5.384796143174094e-05}
{"train_loss": 0.13935264945030212, "global_step": 129331, "epoch": 1453, "lr": 5.384738341423816e-05}
{"train_loss": 0.1552925854921341, "global_step": 129332, "epoch": 1453, "lr": 5.384680539621815e-05}
{"train_loss": 0.1698281615972519, "global_step": 129333, "epoch": 1453, "lr": 5.384622737768098e-05}
{"train_loss": 0.08518905192613602, "global_step": 129334, "epoch": 1453, "lr": 5.3845649358626735e-05}
{"train_loss": 0.11845406144857407, "global_step": 129335, "epoch": 1453, "lr": 5.3845071339055496e-05}
{"train_loss": 0.10284475982189178, "global_step": 129336, "epoch": 1453, "lr": 5.384449331896733e-05}
{"train_loss": 0.07014258950948715, "global_step": 129337, "epoch": 1453, "lr": 5.384391529836231e-05}
{"train_loss": 0.1027454063296318, "global_step": 129338, "epoch": 1453, "lr": 5.384333727724053e-05}
{"train_loss": 0.14219066500663757, "global_step": 129339, "epoch": 1453, "lr": 5.384275925560206e-05}
{"train_loss": 0.10513610392808914, "global_step": 129340, "epoch": 1453, "lr": 5.3842181233446975e-05}
{"train_loss": 0.12796719372272491, "global_step": 129341, "epoch": 1453, "lr": 5.384160321077536e-05}
{"train_loss": 0.14557668566703796, "global_step": 129342, "epoch": 1453, "lr": 5.384102518758729e-05}
{"train_loss": 0.17286698520183563, "global_step": 129343, "epoch": 1453, "lr": 5.3840447163882826e-05}
{"train_loss": 0.12005793303251266, "global_step": 129344, "epoch": 1453, "lr": 5.383986913966208e-05}
{"train_loss": 0.08216319233179092, "global_step": 129345, "epoch": 1453, "lr": 5.383929111492509e-05}
{"train_loss": 0.16699481010437012, "global_step": 129346, "epoch": 1453, "lr": 5.3838713089671975e-05}
{"train_loss": 0.15916649997234344, "global_step": 129347, "epoch": 1453, "lr": 5.3838135063902786e-05}
{"train_loss": 0.1714668571949005, "global_step": 129348, "epoch": 1453, "lr": 5.3837557037617604e-05}
{"train_loss": 0.10748022049665451, "global_step": 129349, "epoch": 1453, "lr": 5.3836979010816505e-05}
{"train_loss": 0.10301078110933304, "global_step": 129350, "epoch": 1453, "lr": 5.383640098349957e-05}
{"train_loss": 0.0627443715929985, "global_step": 129351, "epoch": 1453, "lr": 5.383582295566688e-05}
{"train_loss": 0.1330495923757553, "global_step": 129352, "epoch": 1453, "lr": 5.383524492731852e-05}
{"train_loss": 0.09126365929841995, "global_step": 129353, "epoch": 1453, "lr": 5.3834666898454544e-05}
{"train_loss": 0.1642608493566513, "global_step": 129354, "epoch": 1453, "lr": 5.383408886907505e-05}
{"train_loss": 0.11822552978992462, "global_step": 129355, "epoch": 1453, "lr": 5.383351083918011e-05}
{"train_loss": 0.10748405009508133, "global_step": 129356, "epoch": 1453, "lr": 5.383293280876981e-05}
{"train_loss": 0.12105868756771088, "global_step": 129357, "epoch": 1453, "lr": 5.383235477784421e-05}
{"train_loss": 0.14029625058174133, "global_step": 129358, "epoch": 1453, "lr": 5.3831776746403396e-05}
{"train_loss": 0.11931552737951279, "global_step": 129359, "epoch": 1453, "lr": 5.383119871444744e-05}
{"train_loss": 0.1630055457353592, "global_step": 129360, "epoch": 1453, "lr": 5.383062068197643e-05}
{"train_loss": 0.10476826131343842, "global_step": 129361, "epoch": 1453, "lr": 5.383004264899043e-05}
{"train_loss": 0.13388881087303162, "global_step": 129362, "epoch": 1453, "lr": 5.3829464615489544e-05}
{"train_loss": 0.08841396123170853, "global_step": 129363, "epoch": 1453, "lr": 5.382888658147384e-05}
{"train_loss": 0.11515950411558151, "global_step": 129364, "epoch": 1453, "lr": 5.382830854694339e-05}
{"train_loss": 0.16776150465011597, "global_step": 129365, "epoch": 1453, "lr": 5.3827730511898244e-05}
{"train_loss": 0.09279895573854446, "global_step": 129366, "epoch": 1453, "lr": 5.3827152476338526e-05}
{"train_loss": 0.16249702870845795, "global_step": 129367, "epoch": 1453, "lr": 5.3826574440264286e-05}
{"train_loss": 0.09409715235233307, "global_step": 129368, "epoch": 1453, "lr": 5.382599640367561e-05}
{"train_loss": 0.1780548095703125, "global_step": 129369, "epoch": 1453, "lr": 5.382541836657258e-05}
{"train_loss": 0.14576680958271027, "global_step": 129370, "epoch": 1453, "lr": 5.382484032895527e-05}
{"train_loss": 0.07877067476511002, "global_step": 129371, "epoch": 1453, "lr": 5.3824262290823757e-05}
{"train_loss": 0.15253527462482452, "global_step": 129372, "epoch": 1453, "lr": 5.382368425217811e-05}
{"train_loss": 0.11162342131137848, "global_step": 129373, "epoch": 1453, "lr": 5.3823106213018425e-05}
{"train_loss": 0.16682371497154236, "global_step": 129374, "epoch": 1453, "lr": 5.382252817334477e-05}
{"train_loss": 0.14365112781524658, "global_step": 129375, "epoch": 1453, "lr": 5.382195013315721e-05}
{"train_loss": 0.1425047516822815, "global_step": 129376, "epoch": 1453, "lr": 5.3821372092455845e-05}
{"train_loss": 0.1313929706811905, "global_step": 129377, "epoch": 1453, "lr": 5.3820794051240745e-05}
{"train_loss": 0.1205005869269371, "global_step": 129378, "epoch": 1453, "lr": 5.3820216009511984e-05}
{"train_loss": 0.07452534139156342, "global_step": 129379, "epoch": 1453, "lr": 5.381963796726962e-05}
{"train_loss": 0.0714249312877655, "global_step": 129380, "epoch": 1453, "lr": 5.381905992451377e-05}
{"train_loss": 0.11379187554121017, "global_step": 129381, "epoch": 1453, "lr": 5.38184818812445e-05}
{"train_loss": 0.10124030709266663, "global_step": 129382, "epoch": 1453, "lr": 5.381790383746188e-05}
{"train_loss": 0.12190873175859451, "global_step": 129383, "epoch": 1453, "lr": 5.3817325793165994e-05}
{"train_loss": 0.11793918907642365, "global_step": 129384, "epoch": 1453, "lr": 5.3816747748356914e-05}
{"train_loss": 0.14096464216709137, "global_step": 129385, "epoch": 1453, "lr": 5.3816169703034714e-05}
{"train_loss": 0.12033089995384216, "global_step": 129386, "epoch": 1453, "lr": 5.381559165719947e-05}
{"train_loss": 0.09391619265079498, "global_step": 129387, "epoch": 1453, "lr": 5.3815013610851276e-05}
{"train_loss": 0.2053910791873932, "global_step": 129388, "epoch": 1453, "lr": 5.381443556399019e-05}
{"train_loss": 0.0716276690363884, "global_step": 129389, "epoch": 1453, "lr": 5.38138575166163e-05}
{"train_loss": 0.11289863288402557, "global_step": 129390, "epoch": 1453, "lr": 5.3813279468729694e-05}
{"train_loss": 0.09678344428539276, "global_step": 129391, "epoch": 1453, "lr": 5.3812701420330434e-05}
{"train_loss": 0.14929477870464325, "global_step": 129392, "epoch": 1453, "lr": 5.381212337141861e-05}
{"train_loss": 0.08965403586626053, "global_step": 129393, "epoch": 1453, "lr": 5.3811545321994284e-05}
{"train_loss": 0.06618311256170273, "global_step": 129394, "epoch": 1453, "lr": 5.381096727205754e-05}
{"train_loss": 0.10895460098981857, "global_step": 129395, "epoch": 1453, "lr": 5.3810389221608456e-05}
{"train_loss": 0.12740841507911682, "global_step": 129396, "epoch": 1453, "lr": 5.380981117064713e-05}
{"train_loss": 0.10467736423015594, "global_step": 129397, "epoch": 1453, "lr": 5.3809233119173595e-05}
{"train_loss": 0.056011948734521866, "global_step": 129398, "epoch": 1453, "lr": 5.3808655067187963e-05}
{"train_loss": 0.12887164950370789, "global_step": 129399, "epoch": 1453, "lr": 5.3808077014690304e-05}
{"train_loss": 0.15772095322608948, "global_step": 129400, "epoch": 1453, "lr": 5.3807498961680704e-05}
{"train_loss": 0.15838079154491425, "global_step": 129401, "epoch": 1453, "lr": 5.380692090815923e-05}
{"train_loss": 0.14524133503437042, "global_step": 129402, "epoch": 1453, "lr": 5.380634285412597e-05}
{"train_loss": 0.07223062962293625, "global_step": 129403, "epoch": 1453, "lr": 5.380576479958098e-05}
{"train_loss": 0.14160272479057312, "global_step": 129404, "epoch": 1453, "lr": 5.380518674452436e-05}
{"train_loss": 0.12177175187160459, "global_step": 129405, "epoch": 1453, "lr": 5.3804608688956174e-05, "val_loss": 5.673387050628662}
{"train_loss": 0.13203544914722443, "global_step": 129406, "epoch": 1454, "lr": 5.38040306328765e-05}
{"train_loss": 0.1403534710407257, "global_step": 129407, "epoch": 1454, "lr": 5.380345257628543e-05}
{"train_loss": 0.08906836062669754, "global_step": 129408, "epoch": 1454, "lr": 5.380287451918303e-05}
{"train_loss": 0.11589934676885605, "global_step": 129409, "epoch": 1454, "lr": 5.380229646156938e-05}
{"train_loss": 0.11825689673423767, "global_step": 129410, "epoch": 1454, "lr": 5.380171840344456e-05}
{"train_loss": 0.0886550173163414, "global_step": 129411, "epoch": 1454, "lr": 5.380114034480863e-05}
{"train_loss": 0.09669622778892517, "global_step": 129412, "epoch": 1454, "lr": 5.3800562285661696e-05}
{"train_loss": 0.09382845461368561, "global_step": 129413, "epoch": 1454, "lr": 5.3799984226003816e-05}
{"train_loss": 0.10940363258123398, "global_step": 129414, "epoch": 1454, "lr": 5.379940616583509e-05}
{"train_loss": 0.10434545576572418, "global_step": 129415, "epoch": 1454, "lr": 5.379882810515556e-05}
{"train_loss": 0.13095790147781372, "global_step": 129416, "epoch": 1454, "lr": 5.3798250043965326e-05}
{"train_loss": 0.13746367394924164, "global_step": 129417, "epoch": 1454, "lr": 5.379767198226446e-05}
{"train_loss": 0.1403064876794815, "global_step": 129418, "epoch": 1454, "lr": 5.379709392005306e-05}
{"train_loss": 0.1786530762910843, "global_step": 129419, "epoch": 1454, "lr": 5.3796515857331174e-05}
{"train_loss": 0.11005931347608566, "global_step": 129420, "epoch": 1454, "lr": 5.37959377940989e-05}
{"train_loss": 0.1719849705696106, "global_step": 129421, "epoch": 1454, "lr": 5.379535973035632e-05}
{"train_loss": 0.10699615627527237, "global_step": 129422, "epoch": 1454, "lr": 5.379478166610349e-05}
{"train_loss": 0.11185326427221298, "global_step": 129423, "epoch": 1454, "lr": 5.37942036013405e-05}
{"train_loss": 0.11922027170658112, "global_step": 129424, "epoch": 1454, "lr": 5.3793625536067414e-05}
{"train_loss": 0.08645089715719223, "global_step": 129425, "epoch": 1454, "lr": 5.379304747028433e-05}
{"train_loss": 0.1650378257036209, "global_step": 129426, "epoch": 1454, "lr": 5.379246940399132e-05}
{"train_loss": 0.05760003253817558, "global_step": 129427, "epoch": 1454, "lr": 5.379189133718845e-05}
{"train_loss": 0.11397220194339752, "global_step": 129428, "epoch": 1454, "lr": 5.379131326987581e-05}
{"train_loss": 0.12753888964653015, "global_step": 129429, "epoch": 1454, "lr": 5.379073520205348e-05}
{"train_loss": 0.08908383548259735, "global_step": 129430, "epoch": 1454, "lr": 5.379015713372152e-05}
{"train_loss": 0.08182236552238464, "global_step": 129431, "epoch": 1454, "lr": 5.378957906488003e-05}
{"train_loss": 0.09172087162733078, "global_step": 129432, "epoch": 1454, "lr": 5.378900099552907e-05}
{"train_loss": 0.10752260684967041, "global_step": 129433, "epoch": 1454, "lr": 5.378842292566874e-05}
{"train_loss": 0.12056872248649597, "global_step": 129434, "epoch": 1454, "lr": 5.3787844855299075e-05}
{"train_loss": 0.15390342473983765, "global_step": 129435, "epoch": 1454, "lr": 5.37872667844202e-05}
{"train_loss": 0.12035539001226425, "global_step": 129436, "epoch": 1454, "lr": 5.3786688713032164e-05}
{"train_loss": 0.11719861626625061, "global_step": 129437, "epoch": 1454, "lr": 5.378611064113506e-05}
{"train_loss": 0.19361481070518494, "global_step": 129438, "epoch": 1454, "lr": 5.378553256872896e-05}
{"train_loss": 0.14127302169799805, "global_step": 129439, "epoch": 1454, "lr": 5.3784954495813936e-05}
{"train_loss": 0.17309188842773438, "global_step": 129440, "epoch": 1454, "lr": 5.378437642239008e-05}
{"train_loss": 0.17775529623031616, "global_step": 129441, "epoch": 1454, "lr": 5.378379834845746e-05}
{"train_loss": 0.23216506838798523, "global_step": 129442, "epoch": 1454, "lr": 5.378322027401615e-05}
{"train_loss": 0.13075880706310272, "global_step": 129443, "epoch": 1454, "lr": 5.378264219906624e-05}
{"train_loss": 0.13667482137680054, "global_step": 129444, "epoch": 1454, "lr": 5.378206412360779e-05}
{"train_loss": 0.05960458889603615, "global_step": 129445, "epoch": 1454, "lr": 5.3781486047640895e-05}
{"train_loss": 0.1298341304063797, "global_step": 129446, "epoch": 1454, "lr": 5.378090797116562e-05}
{"train_loss": 0.12687045335769653, "global_step": 129447, "epoch": 1454, "lr": 5.3780329894182044e-05}
{"train_loss": 0.056549206376075745, "global_step": 129448, "epoch": 1454, "lr": 5.377975181669025e-05}
{"train_loss": 0.07910836488008499, "global_step": 129449, "epoch": 1454, "lr": 5.3779173738690326e-05}
{"train_loss": 0.06954087316989899, "global_step": 129450, "epoch": 1454, "lr": 5.377859566018233e-05}
{"train_loss": 0.13053496181964874, "global_step": 129451, "epoch": 1454, "lr": 5.377801758116634e-05}
{"train_loss": 0.10814619064331055, "global_step": 129452, "epoch": 1454, "lr": 5.377743950164245e-05}
{"train_loss": 0.17286789417266846, "global_step": 129453, "epoch": 1454, "lr": 5.3776861421610714e-05}
{"train_loss": 0.10123980045318604, "global_step": 129454, "epoch": 1454, "lr": 5.377628334107124e-05}
{"train_loss": 0.14229966700077057, "global_step": 129455, "epoch": 1454, "lr": 5.3775705260024086e-05}
{"train_loss": 0.08513365685939789, "global_step": 129456, "epoch": 1454, "lr": 5.377512717846934e-05}
{"train_loss": 0.1539333015680313, "global_step": 129457, "epoch": 1454, "lr": 5.377454909640708e-05}
{"train_loss": 0.21420490741729736, "global_step": 129458, "epoch": 1454, "lr": 5.3773971013837365e-05}
{"train_loss": 0.15609562397003174, "global_step": 129459, "epoch": 1454, "lr": 5.377339293076029e-05}
{"train_loss": 0.06439196318387985, "global_step": 129460, "epoch": 1454, "lr": 5.3772814847175946e-05}
{"train_loss": 0.09887070208787918, "global_step": 129461, "epoch": 1454, "lr": 5.3772236763084384e-05}
{"train_loss": 0.14396047592163086, "global_step": 129462, "epoch": 1454, "lr": 5.377165867848568e-05}
{"train_loss": 0.19569452106952667, "global_step": 129463, "epoch": 1454, "lr": 5.377108059337992e-05}
{"train_loss": 0.1759871393442154, "global_step": 129464, "epoch": 1454, "lr": 5.37705025077672e-05}
{"train_loss": 0.13152509927749634, "global_step": 129465, "epoch": 1454, "lr": 5.376992442164757e-05}
{"train_loss": 0.11089204251766205, "global_step": 129466, "epoch": 1454, "lr": 5.376934633502112e-05}
{"train_loss": 0.123397596180439, "global_step": 129467, "epoch": 1454, "lr": 5.3768768247887924e-05}
{"train_loss": 0.10723896324634552, "global_step": 129468, "epoch": 1454, "lr": 5.376819016024807e-05}
{"train_loss": 0.14925742149353027, "global_step": 129469, "epoch": 1454, "lr": 5.376761207210162e-05}
{"train_loss": 0.14348578453063965, "global_step": 129470, "epoch": 1454, "lr": 5.3767033983448666e-05}
{"train_loss": 0.1183244064450264, "global_step": 129471, "epoch": 1454, "lr": 5.376645589428928e-05}
{"train_loss": 0.11557751893997192, "global_step": 129472, "epoch": 1454, "lr": 5.376587780462354e-05}
{"train_loss": 0.0699085220694542, "global_step": 129473, "epoch": 1454, "lr": 5.3765299714451524e-05}
{"train_loss": 0.11082129180431366, "global_step": 129474, "epoch": 1454, "lr": 5.3764721623773305e-05}
{"train_loss": 0.12086348235607147, "global_step": 129475, "epoch": 1454, "lr": 5.3764143532588974e-05}
{"train_loss": 0.12059801816940308, "global_step": 129476, "epoch": 1454, "lr": 5.37635654408986e-05}
{"train_loss": 0.12974615395069122, "global_step": 129477, "epoch": 1454, "lr": 5.376298734870225e-05}
{"train_loss": 0.1273643672466278, "global_step": 129478, "epoch": 1454, "lr": 5.3762409256000025e-05}
{"train_loss": 0.21883782744407654, "global_step": 129479, "epoch": 1454, "lr": 5.376183116279199e-05}
{"train_loss": 0.16484561562538147, "global_step": 129480, "epoch": 1454, "lr": 5.376125306907822e-05}
{"train_loss": 0.13289786875247955, "global_step": 129481, "epoch": 1454, "lr": 5.3760674974858795e-05}
{"train_loss": 0.15703003108501434, "global_step": 129482, "epoch": 1454, "lr": 5.376009688013378e-05}
{"train_loss": 0.12855690717697144, "global_step": 129483, "epoch": 1454, "lr": 5.3759518784903286e-05}
{"train_loss": 0.11974829435348511, "global_step": 129484, "epoch": 1454, "lr": 5.375894068916736e-05}
{"train_loss": 0.12927564978599548, "global_step": 129485, "epoch": 1454, "lr": 5.3758362592926085e-05}
{"train_loss": 0.1845368593931198, "global_step": 129486, "epoch": 1454, "lr": 5.375778449617955e-05}
{"train_loss": 0.11712851375341415, "global_step": 129487, "epoch": 1454, "lr": 5.375720639892783e-05}
{"train_loss": 0.11079463362693787, "global_step": 129488, "epoch": 1454, "lr": 5.3756628301170983e-05}
{"train_loss": 0.08461581915616989, "global_step": 129489, "epoch": 1454, "lr": 5.3756050202909116e-05}
{"train_loss": 0.22880403697490692, "global_step": 129490, "epoch": 1454, "lr": 5.375547210414229e-05}
{"train_loss": 0.13782189786434174, "global_step": 129491, "epoch": 1454, "lr": 5.3754894004870595e-05}
{"train_loss": 0.22751961648464203, "global_step": 129492, "epoch": 1454, "lr": 5.375431590509409e-05}
{"train_loss": 0.128419890999794, "global_step": 129493, "epoch": 1454, "lr": 5.375373780481286e-05}
{"train_loss": 0.12872105267610442, "global_step": 129494, "epoch": 1454, "lr": 5.3753159704027e-05, "val_loss": 5.573108673095703}
{"train_loss": 0.16181384027004242, "global_step": 129495, "epoch": 1455, "lr": 5.3752581602736576e-05}
{"train_loss": 0.10398620367050171, "global_step": 129496, "epoch": 1455, "lr": 5.375200350094165e-05}
{"train_loss": 0.1122557669878006, "global_step": 129497, "epoch": 1455, "lr": 5.375142539864232e-05}
{"train_loss": 0.10158118605613708, "global_step": 129498, "epoch": 1455, "lr": 5.3750847295838646e-05}
{"train_loss": 0.11080322414636612, "global_step": 129499, "epoch": 1455, "lr": 5.375026919253073e-05}
{"train_loss": 0.16553553938865662, "global_step": 129500, "epoch": 1455, "lr": 5.374969108871864e-05}
{"train_loss": 0.1234639585018158, "global_step": 129501, "epoch": 1455, "lr": 5.3749112984402436e-05}
{"train_loss": 0.1032332181930542, "global_step": 129502, "epoch": 1455, "lr": 5.3748534879582216e-05}
{"train_loss": 0.14659510552883148, "global_step": 129503, "epoch": 1455, "lr": 5.3747956774258046e-05}
{"train_loss": 0.08672598749399185, "global_step": 129504, "epoch": 1455, "lr": 5.374737866843001e-05}
{"train_loss": 0.09484010189771652, "global_step": 129505, "epoch": 1455, "lr": 5.3746800562098184e-05}
{"train_loss": 0.2325473129749298, "global_step": 129506, "epoch": 1455, "lr": 5.374622245526264e-05}
{"train_loss": 0.2512933909893036, "global_step": 129507, "epoch": 1455, "lr": 5.3745644347923475e-05}
{"train_loss": 0.11205483973026276, "global_step": 129508, "epoch": 1455, "lr": 5.3745066240080744e-05}
{"train_loss": 0.10002288967370987, "global_step": 129509, "epoch": 1455, "lr": 5.3744488131734536e-05}
{"train_loss": 0.10144499689340591, "global_step": 129510, "epoch": 1455, "lr": 5.374391002288493e-05}
{"train_loss": 0.14003704488277435, "global_step": 129511, "epoch": 1455, "lr": 5.3743331913532016e-05}
{"train_loss": 0.10989461839199066, "global_step": 129512, "epoch": 1455, "lr": 5.374275380367584e-05}
{"train_loss": 0.15718697011470795, "global_step": 129513, "epoch": 1455, "lr": 5.37421756933165e-05}
{"train_loss": 0.09917764365673065, "global_step": 129514, "epoch": 1455, "lr": 5.374159758245406e-05}
{"train_loss": 0.05819912254810333, "global_step": 129515, "epoch": 1455, "lr": 5.374101947108861e-05}
{"train_loss": 0.14167018234729767, "global_step": 129516, "epoch": 1455, "lr": 5.374044135922024e-05}
{"train_loss": 0.17096315324306488, "global_step": 129517, "epoch": 1455, "lr": 5.373986324684901e-05}
{"train_loss": 0.205448716878891, "global_step": 129518, "epoch": 1455, "lr": 5.3739285133975006e-05}
{"train_loss": 0.08767679333686829, "global_step": 129519, "epoch": 1455, "lr": 5.373870702059829e-05}
{"train_loss": 0.0912136435508728, "global_step": 129520, "epoch": 1455, "lr": 5.373812890671896e-05}
{"train_loss": 0.10204987227916718, "global_step": 129521, "epoch": 1455, "lr": 5.3737550792337064e-05}
{"train_loss": 0.1078183650970459, "global_step": 129522, "epoch": 1455, "lr": 5.373697267745271e-05}
{"train_loss": 0.07537975162267685, "global_step": 129523, "epoch": 1455, "lr": 5.373639456206597e-05}
{"train_loss": 0.06624016910791397, "global_step": 129524, "epoch": 1455, "lr": 5.3735816446176914e-05}
{"train_loss": 0.14418122172355652, "global_step": 129525, "epoch": 1455, "lr": 5.373523832978562e-05}
{"train_loss": 0.1382162868976593, "global_step": 129526, "epoch": 1455, "lr": 5.373466021289217e-05}
{"train_loss": 0.11670289933681488, "global_step": 129527, "epoch": 1455, "lr": 5.373408209549664e-05}
{"train_loss": 0.11312390118837357, "global_step": 129528, "epoch": 1455, "lr": 5.373350397759911e-05}
{"train_loss": 0.1357554942369461, "global_step": 129529, "epoch": 1455, "lr": 5.373292585919966e-05}
{"train_loss": 0.05469531565904617, "global_step": 129530, "epoch": 1455, "lr": 5.373234774029836e-05}
{"train_loss": 0.10624154657125473, "global_step": 129531, "epoch": 1455, "lr": 5.373176962089529e-05}
{"train_loss": 0.23217643797397614, "global_step": 129532, "epoch": 1455, "lr": 5.373119150099053e-05}
{"train_loss": 0.08150212466716766, "global_step": 129533, "epoch": 1455, "lr": 5.373061338058416e-05}
{"train_loss": 0.16412901878356934, "global_step": 129534, "epoch": 1455, "lr": 5.373003525967625e-05}
{"train_loss": 0.1165996864438057, "global_step": 129535, "epoch": 1455, "lr": 5.3729457138266884e-05}
{"train_loss": 0.20503902435302734, "global_step": 129536, "epoch": 1455, "lr": 5.3728879016356134e-05}
{"train_loss": 0.15722917020320892, "global_step": 129537, "epoch": 1455, "lr": 5.3728300893944104e-05}
{"train_loss": 0.09271009266376495, "global_step": 129538, "epoch": 1455, "lr": 5.372772277103083e-05}
{"train_loss": 0.09322638809680939, "global_step": 129539, "epoch": 1455, "lr": 5.372714464761641e-05}
{"train_loss": 0.15263250470161438, "global_step": 129540, "epoch": 1455, "lr": 5.372656652370092e-05}
{"train_loss": 0.11492665112018585, "global_step": 129541, "epoch": 1455, "lr": 5.372598839928444e-05}
{"train_loss": 0.12760980427265167, "global_step": 129542, "epoch": 1455, "lr": 5.372541027436705e-05}
{"train_loss": 0.138083815574646, "global_step": 129543, "epoch": 1455, "lr": 5.37248321489488e-05}
{"train_loss": 0.07498495280742645, "global_step": 129544, "epoch": 1455, "lr": 5.372425402302982e-05}
{"train_loss": 0.11336590349674225, "global_step": 129545, "epoch": 1455, "lr": 5.3723675896610145e-05}
{"train_loss": 0.11507592350244522, "global_step": 129546, "epoch": 1455, "lr": 5.372309776968988e-05}
{"train_loss": 0.10726270079612732, "global_step": 129547, "epoch": 1455, "lr": 5.372251964226907e-05}
{"train_loss": 0.0824018120765686, "global_step": 129548, "epoch": 1455, "lr": 5.372194151434784e-05}
{"train_loss": 0.15301279723644257, "global_step": 129549, "epoch": 1455, "lr": 5.372136338592622e-05}
{"train_loss": 0.10986526310443878, "global_step": 129550, "epoch": 1455, "lr": 5.372078525700431e-05}
{"train_loss": 0.105647012591362, "global_step": 129551, "epoch": 1455, "lr": 5.372020712758219e-05}
{"train_loss": 0.10818909108638763, "global_step": 129552, "epoch": 1455, "lr": 5.3719628997659945e-05}
{"train_loss": 0.07932067662477493, "global_step": 129553, "epoch": 1455, "lr": 5.371905086723762e-05}
{"train_loss": 0.10888068377971649, "global_step": 129554, "epoch": 1455, "lr": 5.371847273631533e-05}
{"train_loss": 0.09229041635990143, "global_step": 129555, "epoch": 1455, "lr": 5.371789460489313e-05}
{"train_loss": 0.10136467218399048, "global_step": 129556, "epoch": 1455, "lr": 5.37173164729711e-05}
{"train_loss": 0.20798274874687195, "global_step": 129557, "epoch": 1455, "lr": 5.371673834054934e-05}
{"train_loss": 0.12211252003908157, "global_step": 129558, "epoch": 1455, "lr": 5.37161602076279e-05}
{"train_loss": 0.11984871327877045, "global_step": 129559, "epoch": 1455, "lr": 5.371558207420686e-05}
{"train_loss": 0.1563435196876526, "global_step": 129560, "epoch": 1455, "lr": 5.371500394028631e-05}
{"train_loss": 0.07435549050569534, "global_step": 129561, "epoch": 1455, "lr": 5.371442580586631e-05}
{"train_loss": 0.1651994287967682, "global_step": 129562, "epoch": 1455, "lr": 5.371384767094696e-05}
{"train_loss": 0.15338855981826782, "global_step": 129563, "epoch": 1455, "lr": 5.371326953552833e-05}
{"train_loss": 0.10914162546396255, "global_step": 129564, "epoch": 1455, "lr": 5.371269139961051e-05}
{"train_loss": 0.1229945719242096, "global_step": 129565, "epoch": 1455, "lr": 5.371211326319354e-05}
{"train_loss": 0.07600691169500351, "global_step": 129566, "epoch": 1455, "lr": 5.371153512627753e-05}
{"train_loss": 0.10150741040706635, "global_step": 129567, "epoch": 1455, "lr": 5.371095698886256e-05}
{"train_loss": 0.15639153122901917, "global_step": 129568, "epoch": 1455, "lr": 5.371037885094869e-05}
{"train_loss": 0.11099844425916672, "global_step": 129569, "epoch": 1455, "lr": 5.3709800712536006e-05}
{"train_loss": 0.12288761138916016, "global_step": 129570, "epoch": 1455, "lr": 5.370922257362459e-05}
{"train_loss": 0.14106504619121552, "global_step": 129571, "epoch": 1455, "lr": 5.3708644434214496e-05}
{"train_loss": 0.14919348061084747, "global_step": 129572, "epoch": 1455, "lr": 5.3708066294305835e-05}
{"train_loss": 0.12174336612224579, "global_step": 129573, "epoch": 1455, "lr": 5.3707488153898664e-05}
{"train_loss": 0.13428086042404175, "global_step": 129574, "epoch": 1455, "lr": 5.370691001299307e-05}
{"train_loss": 0.11873325705528259, "global_step": 129575, "epoch": 1455, "lr": 5.370633187158912e-05}
{"train_loss": 0.12328256666660309, "global_step": 129576, "epoch": 1455, "lr": 5.3705753729686915e-05}
{"train_loss": 0.18393193185329437, "global_step": 129577, "epoch": 1455, "lr": 5.3705175587286506e-05}
{"train_loss": 0.14396369457244873, "global_step": 129578, "epoch": 1455, "lr": 5.370459744438798e-05}
{"train_loss": 0.08668839186429977, "global_step": 129579, "epoch": 1455, "lr": 5.3704019300991414e-05}
{"train_loss": 0.14727231860160828, "global_step": 129580, "epoch": 1455, "lr": 5.370344115709689e-05}
{"train_loss": 0.1509617418050766, "global_step": 129581, "epoch": 1455, "lr": 5.370286301270449e-05}
{"train_loss": 0.13379648327827454, "global_step": 129582, "epoch": 1455, "lr": 5.370228486781427e-05}
{"train_loss": 0.1253612084167727, "global_step": 129583, "epoch": 1455, "lr": 5.3701706722426335e-05, "val_loss": 5.851455211639404, "train_action_mse_error": 23.158437728881836}
{"train_loss": 0.1447838991880417, "global_step": 129584, "epoch": 1456, "lr": 5.370112857654075e-05}
{"train_loss": 0.10295131802558899, "global_step": 129585, "epoch": 1456, "lr": 5.3700550430157593e-05}
{"train_loss": 0.07068657130002975, "global_step": 129586, "epoch": 1456, "lr": 5.3699972283276936e-05}
{"train_loss": 0.154368057847023, "global_step": 129587, "epoch": 1456, "lr": 5.369939413589886e-05}
{"train_loss": 0.13709308207035065, "global_step": 129588, "epoch": 1456, "lr": 5.369881598802347e-05}
{"train_loss": 0.1057014986872673, "global_step": 129589, "epoch": 1456, "lr": 5.369823783965079e-05}
{"train_loss": 0.120289146900177, "global_step": 129590, "epoch": 1456, "lr": 5.369765969078094e-05}
{"train_loss": 0.08435072004795074, "global_step": 129591, "epoch": 1456, "lr": 5.3697081541413986e-05}
{"train_loss": 0.09826412051916122, "global_step": 129592, "epoch": 1456, "lr": 5.3696503391550004e-05}
{"train_loss": 0.24146045744419098, "global_step": 129593, "epoch": 1456, "lr": 5.369592524118907e-05}
{"train_loss": 0.11313747614622116, "global_step": 129594, "epoch": 1456, "lr": 5.369534709033126e-05}
{"train_loss": 0.14921501278877258, "global_step": 129595, "epoch": 1456, "lr": 5.369476893897667e-05}
{"train_loss": 0.09320691972970963, "global_step": 129596, "epoch": 1456, "lr": 5.3694190787125354e-05}
{"train_loss": 0.07210726290941238, "global_step": 129597, "epoch": 1456, "lr": 5.369361263477739e-05}
{"train_loss": 0.09069503843784332, "global_step": 129598, "epoch": 1456, "lr": 5.369303448193287e-05}
{"train_loss": 0.07424106448888779, "global_step": 129599, "epoch": 1456, "lr": 5.369245632859188e-05}
{"train_loss": 0.09166360646486282, "global_step": 129600, "epoch": 1456, "lr": 5.369187817475447e-05}
{"train_loss": 0.10798488557338715, "global_step": 129601, "epoch": 1456, "lr": 5.369130002042073e-05}
{"train_loss": 0.1680082231760025, "global_step": 129602, "epoch": 1456, "lr": 5.3690721865590756e-05}
{"train_loss": 0.16093914210796356, "global_step": 129603, "epoch": 1456, "lr": 5.36901437102646e-05}
{"train_loss": 0.0749494805932045, "global_step": 129604, "epoch": 1456, "lr": 5.368956555444234e-05}
{"train_loss": 0.16881148517131805, "global_step": 129605, "epoch": 1456, "lr": 5.368898739812407e-05}
{"train_loss": 0.11935646086931229, "global_step": 129606, "epoch": 1456, "lr": 5.3688409241309865e-05}
{"train_loss": 0.0955367311835289, "global_step": 129607, "epoch": 1456, "lr": 5.368783108399979e-05}
{"train_loss": 0.10000687837600708, "global_step": 129608, "epoch": 1456, "lr": 5.368725292619394e-05}
{"train_loss": 0.10674306750297546, "global_step": 129609, "epoch": 1456, "lr": 5.368667476789237e-05}
{"train_loss": 0.08748769760131836, "global_step": 129610, "epoch": 1456, "lr": 5.3686096609095195e-05}
{"train_loss": 0.12717382609844208, "global_step": 129611, "epoch": 1456, "lr": 5.368551844980245e-05}
{"train_loss": 0.07655130326747894, "global_step": 129612, "epoch": 1456, "lr": 5.368494029001424e-05}
{"train_loss": 0.14674288034439087, "global_step": 129613, "epoch": 1456, "lr": 5.368436212973064e-05}
{"train_loss": 0.19381454586982727, "global_step": 129614, "epoch": 1456, "lr": 5.368378396895171e-05}
{"train_loss": 0.16420042514801025, "global_step": 129615, "epoch": 1456, "lr": 5.368320580767754e-05}
{"train_loss": 0.11047690361738205, "global_step": 129616, "epoch": 1456, "lr": 5.3682627645908213e-05}
{"train_loss": 0.1283620446920395, "global_step": 129617, "epoch": 1456, "lr": 5.3682049483643805e-05}
{"train_loss": 0.13930799067020416, "global_step": 129618, "epoch": 1456, "lr": 5.368147132088439e-05}
{"train_loss": 0.1229095458984375, "global_step": 129619, "epoch": 1456, "lr": 5.3680893157630044e-05}
{"train_loss": 0.0989404022693634, "global_step": 129620, "epoch": 1456, "lr": 5.368031499388084e-05}
{"train_loss": 0.10287714004516602, "global_step": 129621, "epoch": 1456, "lr": 5.3679736829636875e-05}
{"train_loss": 0.09946168214082718, "global_step": 129622, "epoch": 1456, "lr": 5.36791586648982e-05}
{"train_loss": 0.13013465702533722, "global_step": 129623, "epoch": 1456, "lr": 5.367858049966491e-05}
{"train_loss": 0.1314040720462799, "global_step": 129624, "epoch": 1456, "lr": 5.36780023339371e-05}
{"train_loss": 0.10075032711029053, "global_step": 129625, "epoch": 1456, "lr": 5.3677424167714795e-05}
{"train_loss": 0.09234440326690674, "global_step": 129626, "epoch": 1456, "lr": 5.367684600099813e-05}
{"train_loss": 0.14220327138900757, "global_step": 129627, "epoch": 1456, "lr": 5.3676267833787144e-05}
{"train_loss": 0.17033997178077698, "global_step": 129628, "epoch": 1456, "lr": 5.3675689666081944e-05}
{"train_loss": 0.21557725965976715, "global_step": 129629, "epoch": 1456, "lr": 5.367511149788258e-05}
{"train_loss": 0.15359215438365936, "global_step": 129630, "epoch": 1456, "lr": 5.3674533329189136e-05}
{"train_loss": 0.1620195358991623, "global_step": 129631, "epoch": 1456, "lr": 5.367395516000171e-05}
{"train_loss": 0.07547004520893097, "global_step": 129632, "epoch": 1456, "lr": 5.3673376990320355e-05}
{"train_loss": 0.07280301302671432, "global_step": 129633, "epoch": 1456, "lr": 5.367279882014517e-05}
{"train_loss": 0.125613272190094, "global_step": 129634, "epoch": 1456, "lr": 5.367222064947621e-05}
{"train_loss": 0.13428492844104767, "global_step": 129635, "epoch": 1456, "lr": 5.367164247831358e-05}
{"train_loss": 0.12648886442184448, "global_step": 129636, "epoch": 1456, "lr": 5.3671064306657325e-05}
{"train_loss": 0.09372428059577942, "global_step": 129637, "epoch": 1456, "lr": 5.367048613450755e-05}
{"train_loss": 0.15849807858467102, "global_step": 129638, "epoch": 1456, "lr": 5.366990796186432e-05}
{"train_loss": 0.08822908252477646, "global_step": 129639, "epoch": 1456, "lr": 5.3669329788727716e-05}
{"train_loss": 0.11336083710193634, "global_step": 129640, "epoch": 1456, "lr": 5.366875161509781e-05}
{"train_loss": 0.17086417973041534, "global_step": 129641, "epoch": 1456, "lr": 5.36681734409747e-05}
{"train_loss": 0.13475292921066284, "global_step": 129642, "epoch": 1456, "lr": 5.366759526635844e-05}
{"train_loss": 0.13546311855316162, "global_step": 129643, "epoch": 1456, "lr": 5.366701709124912e-05}
{"train_loss": 0.16477489471435547, "global_step": 129644, "epoch": 1456, "lr": 5.36664389156468e-05}
{"train_loss": 0.12660077214241028, "global_step": 129645, "epoch": 1456, "lr": 5.366586073955159e-05}
{"train_loss": 0.11448969691991806, "global_step": 129646, "epoch": 1456, "lr": 5.3665282562963535e-05}
{"train_loss": 0.14182524383068085, "global_step": 129647, "epoch": 1456, "lr": 5.366470438588274e-05}
{"train_loss": 0.08060125261545181, "global_step": 129648, "epoch": 1456, "lr": 5.3664126208309266e-05}
{"train_loss": 0.12731261551380157, "global_step": 129649, "epoch": 1456, "lr": 5.366354803024319e-05}
{"train_loss": 0.1865909844636917, "global_step": 129650, "epoch": 1456, "lr": 5.3662969851684594e-05}
{"train_loss": 0.07033652067184448, "global_step": 129651, "epoch": 1456, "lr": 5.366239167263356e-05}
{"train_loss": 0.15207409858703613, "global_step": 129652, "epoch": 1456, "lr": 5.3661813493090165e-05}
{"train_loss": 0.19300633668899536, "global_step": 129653, "epoch": 1456, "lr": 5.366123531305448e-05}
{"train_loss": 0.1163853332400322, "global_step": 129654, "epoch": 1456, "lr": 5.3660657132526594e-05}
{"train_loss": 0.24343560636043549, "global_step": 129655, "epoch": 1456, "lr": 5.3660078951506565e-05}
{"train_loss": 0.11903499811887741, "global_step": 129656, "epoch": 1456, "lr": 5.36595007699945e-05}
{"train_loss": 0.15967683494091034, "global_step": 129657, "epoch": 1456, "lr": 5.365892258799045e-05}
{"train_loss": 0.14860735833644867, "global_step": 129658, "epoch": 1456, "lr": 5.36583444054945e-05}
{"train_loss": 0.07436882704496384, "global_step": 129659, "epoch": 1456, "lr": 5.365776622250672e-05}
{"train_loss": 0.1455812305212021, "global_step": 129660, "epoch": 1456, "lr": 5.365718803902722e-05}
{"train_loss": 0.15758869051933289, "global_step": 129661, "epoch": 1456, "lr": 5.365660985505604e-05}
{"train_loss": 0.11553587764501572, "global_step": 129662, "epoch": 1456, "lr": 5.3656031670593275e-05}
{"train_loss": 0.1918337494134903, "global_step": 129663, "epoch": 1456, "lr": 5.365545348563901e-05}
{"train_loss": 0.16474932432174683, "global_step": 129664, "epoch": 1456, "lr": 5.3654875300193306e-05}
{"train_loss": 0.13340765237808228, "global_step": 129665, "epoch": 1456, "lr": 5.365429711425626e-05}
{"train_loss": 0.08496911078691483, "global_step": 129666, "epoch": 1456, "lr": 5.3653718927827924e-05}
{"train_loss": 0.1553405225276947, "global_step": 129667, "epoch": 1456, "lr": 5.365314074090839e-05}
{"train_loss": 0.15619152784347534, "global_step": 129668, "epoch": 1456, "lr": 5.365256255349774e-05}
{"train_loss": 0.13700230419635773, "global_step": 129669, "epoch": 1456, "lr": 5.365198436559605e-05}
{"train_loss": 0.16664595901966095, "global_step": 129670, "epoch": 1456, "lr": 5.36514061772034e-05}
{"train_loss": 0.10071498155593872, "global_step": 129671, "epoch": 1456, "lr": 5.365082798831985e-05}
{"train_loss": 0.12864706338791365, "global_step": 129672, "epoch": 1456, "lr": 5.365024979894551e-05, "val_loss": 5.672587871551514}
{"train_loss": 0.141727477312088, "global_step": 129673, "epoch": 1457, "lr": 5.364967160908042e-05}
{"train_loss": 0.09755877405405045, "global_step": 129674, "epoch": 1457, "lr": 5.364909341872468e-05}
{"train_loss": 0.12525056302547455, "global_step": 129675, "epoch": 1457, "lr": 5.364851522787837e-05}
{"train_loss": 0.14953523874282837, "global_step": 129676, "epoch": 1457, "lr": 5.364793703654155e-05}
{"train_loss": 0.17990167438983917, "global_step": 129677, "epoch": 1457, "lr": 5.364735884471432e-05}
{"train_loss": 0.11010459810495377, "global_step": 129678, "epoch": 1457, "lr": 5.3646780652396744e-05}
{"train_loss": 0.15806563198566437, "global_step": 129679, "epoch": 1457, "lr": 5.36462024595889e-05}
{"train_loss": 0.1018153578042984, "global_step": 129680, "epoch": 1457, "lr": 5.3645624266290874e-05}
{"train_loss": 0.11661212891340256, "global_step": 129681, "epoch": 1457, "lr": 5.364504607250274e-05}
{"train_loss": 0.10620288550853729, "global_step": 129682, "epoch": 1457, "lr": 5.364446787822457e-05}
{"train_loss": 0.1086004227399826, "global_step": 129683, "epoch": 1457, "lr": 5.364388968345644e-05}
{"train_loss": 0.14252105355262756, "global_step": 129684, "epoch": 1457, "lr": 5.364331148819844e-05}
{"train_loss": 0.1576656848192215, "global_step": 129685, "epoch": 1457, "lr": 5.3642733292450645e-05}
{"train_loss": 0.08997891843318939, "global_step": 129686, "epoch": 1457, "lr": 5.364215509621312e-05}
{"train_loss": 0.10669615864753723, "global_step": 129687, "epoch": 1457, "lr": 5.3641576899485956e-05}
{"train_loss": 0.15983375906944275, "global_step": 129688, "epoch": 1457, "lr": 5.364099870226922e-05}
{"train_loss": 0.13851383328437805, "global_step": 129689, "epoch": 1457, "lr": 5.3640420504563e-05}
{"train_loss": 0.16150929033756256, "global_step": 129690, "epoch": 1457, "lr": 5.363984230636738e-05}
{"train_loss": 0.12308063358068466, "global_step": 129691, "epoch": 1457, "lr": 5.363926410768242e-05}
{"train_loss": 0.12820348143577576, "global_step": 129692, "epoch": 1457, "lr": 5.3638685908508204e-05}
{"train_loss": 0.13986486196517944, "global_step": 129693, "epoch": 1457, "lr": 5.363810770884481e-05}
{"train_loss": 0.0778711810708046, "global_step": 129694, "epoch": 1457, "lr": 5.363752950869232e-05}
{"train_loss": 0.08800684660673141, "global_step": 129695, "epoch": 1457, "lr": 5.36369513080508e-05}
{"train_loss": 0.10201158374547958, "global_step": 129696, "epoch": 1457, "lr": 5.3636373106920345e-05}
{"train_loss": 0.10992719978094101, "global_step": 129697, "epoch": 1457, "lr": 5.363579490530102e-05}
{"train_loss": 0.15507933497428894, "global_step": 129698, "epoch": 1457, "lr": 5.363521670319292e-05}
{"train_loss": 0.09545864909887314, "global_step": 129699, "epoch": 1457, "lr": 5.363463850059609e-05}
{"train_loss": 0.13421812653541565, "global_step": 129700, "epoch": 1457, "lr": 5.363406029751064e-05}
{"train_loss": 0.08853434026241302, "global_step": 129701, "epoch": 1457, "lr": 5.363348209393663e-05}
{"train_loss": 0.08576381206512451, "global_step": 129702, "epoch": 1457, "lr": 5.363290388987414e-05}
{"train_loss": 0.09788373112678528, "global_step": 129703, "epoch": 1457, "lr": 5.363232568532325e-05}
{"train_loss": 0.14255307614803314, "global_step": 129704, "epoch": 1457, "lr": 5.3631747480284044e-05}
{"train_loss": 0.11364179849624634, "global_step": 129705, "epoch": 1457, "lr": 5.363116927475659e-05}
{"train_loss": 0.07567393779754639, "global_step": 129706, "epoch": 1457, "lr": 5.3630591068740976e-05}
{"train_loss": 0.12470521777868271, "global_step": 129707, "epoch": 1457, "lr": 5.363001286223727e-05}
{"train_loss": 0.1340855062007904, "global_step": 129708, "epoch": 1457, "lr": 5.362943465524555e-05}
{"train_loss": 0.17822369933128357, "global_step": 129709, "epoch": 1457, "lr": 5.36288564477659e-05}
{"train_loss": 0.11486679315567017, "global_step": 129710, "epoch": 1457, "lr": 5.362827823979838e-05}
{"train_loss": 0.13472743332386017, "global_step": 129711, "epoch": 1457, "lr": 5.36277000313431e-05}
{"train_loss": 0.1068800613284111, "global_step": 129712, "epoch": 1457, "lr": 5.362712182240011e-05}
{"train_loss": 0.08717929571866989, "global_step": 129713, "epoch": 1457, "lr": 5.3626543612969504e-05}
{"train_loss": 0.12172691524028778, "global_step": 129714, "epoch": 1457, "lr": 5.362596540305135e-05}
{"train_loss": 0.13714222609996796, "global_step": 129715, "epoch": 1457, "lr": 5.3625387192645725e-05}
{"train_loss": 0.12898100912570953, "global_step": 129716, "epoch": 1457, "lr": 5.362480898175273e-05}
{"train_loss": 0.11232930421829224, "global_step": 129717, "epoch": 1457, "lr": 5.3624230770372396e-05}
{"train_loss": 0.13549911975860596, "global_step": 129718, "epoch": 1457, "lr": 5.362365255850484e-05}
{"train_loss": 0.12104243040084839, "global_step": 129719, "epoch": 1457, "lr": 5.362307434615014e-05}
{"train_loss": 0.17083846032619476, "global_step": 129720, "epoch": 1457, "lr": 5.3622496133308355e-05}
{"train_loss": 0.09571612626314163, "global_step": 129721, "epoch": 1457, "lr": 5.3621917919979556e-05}
{"train_loss": 0.1269298493862152, "global_step": 129722, "epoch": 1457, "lr": 5.3621339706163854e-05}
{"train_loss": 0.1772129088640213, "global_step": 129723, "epoch": 1457, "lr": 5.362076149186129e-05}
{"train_loss": 0.09196116030216217, "global_step": 129724, "epoch": 1457, "lr": 5.3620183277071964e-05}
{"train_loss": 0.18130894005298615, "global_step": 129725, "epoch": 1457, "lr": 5.3619605061795964e-05}
{"train_loss": 0.18324671685695648, "global_step": 129726, "epoch": 1457, "lr": 5.361902684603334e-05}
{"train_loss": 0.10610153526067734, "global_step": 129727, "epoch": 1457, "lr": 5.361844862978418e-05}
{"train_loss": 0.1083717793226242, "global_step": 129728, "epoch": 1457, "lr": 5.361787041304857e-05}
{"train_loss": 0.12509655952453613, "global_step": 129729, "epoch": 1457, "lr": 5.3617292195826574e-05}
{"train_loss": 0.14482936263084412, "global_step": 129730, "epoch": 1457, "lr": 5.361671397811827e-05}
{"train_loss": 0.13673661649227142, "global_step": 129731, "epoch": 1457, "lr": 5.361613575992376e-05}
{"train_loss": 0.1264311671257019, "global_step": 129732, "epoch": 1457, "lr": 5.36155575412431e-05}
{"train_loss": 0.11962977796792984, "global_step": 129733, "epoch": 1457, "lr": 5.361497932207637e-05}
{"train_loss": 0.11428932100534439, "global_step": 129734, "epoch": 1457, "lr": 5.361440110242364e-05}
{"train_loss": 0.15627215802669525, "global_step": 129735, "epoch": 1457, "lr": 5.3613822882285014e-05}
{"train_loss": 0.11174295842647552, "global_step": 129736, "epoch": 1457, "lr": 5.361324466166054e-05}
{"train_loss": 0.09586174041032791, "global_step": 129737, "epoch": 1457, "lr": 5.3612666440550316e-05}
{"train_loss": 0.1317475587129593, "global_step": 129738, "epoch": 1457, "lr": 5.3612088218954414e-05}
{"train_loss": 0.12408526241779327, "global_step": 129739, "epoch": 1457, "lr": 5.361150999687291e-05}
{"train_loss": 0.11418239027261734, "global_step": 129740, "epoch": 1457, "lr": 5.361093177430588e-05}
{"train_loss": 0.22500884532928467, "global_step": 129741, "epoch": 1457, "lr": 5.361035355125341e-05}
{"train_loss": 0.16047988831996918, "global_step": 129742, "epoch": 1457, "lr": 5.3609775327715565e-05}
{"train_loss": 0.12954117357730865, "global_step": 129743, "epoch": 1457, "lr": 5.360919710369243e-05}
{"train_loss": 0.11977909505367279, "global_step": 129744, "epoch": 1457, "lr": 5.360861887918408e-05}
{"train_loss": 0.1393759548664093, "global_step": 129745, "epoch": 1457, "lr": 5.360804065419062e-05}
{"train_loss": 0.08536536991596222, "global_step": 129746, "epoch": 1457, "lr": 5.3607462428712086e-05}
{"train_loss": 0.09433063864707947, "global_step": 129747, "epoch": 1457, "lr": 5.3606884202748564e-05}
{"train_loss": 0.16542375087738037, "global_step": 129748, "epoch": 1457, "lr": 5.360630597630015e-05}
{"train_loss": 0.14130660891532898, "global_step": 129749, "epoch": 1457, "lr": 5.3605727749366906e-05}
{"train_loss": 0.10923536121845245, "global_step": 129750, "epoch": 1457, "lr": 5.3605149521948926e-05}
{"train_loss": 0.14028310775756836, "global_step": 129751, "epoch": 1457, "lr": 5.360457129404627e-05}
{"train_loss": 0.11778702586889267, "global_step": 129752, "epoch": 1457, "lr": 5.360399306565902e-05}
{"train_loss": 0.10413037240505219, "global_step": 129753, "epoch": 1457, "lr": 5.360341483678726e-05}
{"train_loss": 0.09470576792955399, "global_step": 129754, "epoch": 1457, "lr": 5.360283660743106e-05}
{"train_loss": 0.15078812837600708, "global_step": 129755, "epoch": 1457, "lr": 5.360225837759052e-05}
{"train_loss": 0.11774292588233948, "global_step": 129756, "epoch": 1457, "lr": 5.360168014726569e-05}
{"train_loss": 0.16305577754974365, "global_step": 129757, "epoch": 1457, "lr": 5.3601101916456656e-05}
{"train_loss": 0.14687177538871765, "global_step": 129758, "epoch": 1457, "lr": 5.360052368516349e-05}
{"train_loss": 0.18929871916770935, "global_step": 129759, "epoch": 1457, "lr": 5.359994545338629e-05}
{"train_loss": 0.11210912466049194, "global_step": 129760, "epoch": 1457, "lr": 5.359936722112511e-05}
{"train_loss": 0.12760441775402326, "global_step": 129761, "epoch": 1457, "lr": 5.3598788988380045e-05, "val_loss": 5.7491135597229}
{"train_loss": 0.18631774187088013, "global_step": 129762, "epoch": 1458, "lr": 5.359821075515117e-05}
{"train_loss": 0.13498090207576752, "global_step": 129763, "epoch": 1458, "lr": 5.359763252143855e-05}
{"train_loss": 0.10669852048158646, "global_step": 129764, "epoch": 1458, "lr": 5.3597054287242286e-05}
{"train_loss": 0.09606284648180008, "global_step": 129765, "epoch": 1458, "lr": 5.359647605256244e-05}
{"train_loss": 0.2195906937122345, "global_step": 129766, "epoch": 1458, "lr": 5.359589781739909e-05}
{"train_loss": 0.1151188388466835, "global_step": 129767, "epoch": 1458, "lr": 5.3595319581752315e-05}
{"train_loss": 0.15261797606945038, "global_step": 129768, "epoch": 1458, "lr": 5.359474134562219e-05}
{"train_loss": 0.17264439165592194, "global_step": 129769, "epoch": 1458, "lr": 5.359416310900879e-05}
{"train_loss": 0.12697206437587738, "global_step": 129770, "epoch": 1458, "lr": 5.3593584871912207e-05}
{"train_loss": 0.10943318158388138, "global_step": 129771, "epoch": 1458, "lr": 5.359300663433251e-05}
{"train_loss": 0.1294373720884323, "global_step": 129772, "epoch": 1458, "lr": 5.359242839626977e-05}
{"train_loss": 0.12556037306785583, "global_step": 129773, "epoch": 1458, "lr": 5.3591850157724075e-05}
{"train_loss": 0.18541483581066132, "global_step": 129774, "epoch": 1458, "lr": 5.359127191869551e-05}
{"train_loss": 0.12262296676635742, "global_step": 129775, "epoch": 1458, "lr": 5.359069367918412e-05}
{"train_loss": 0.0755961537361145, "global_step": 129776, "epoch": 1458, "lr": 5.359011543919003e-05}
{"train_loss": 0.13331085443496704, "global_step": 129777, "epoch": 1458, "lr": 5.358953719871328e-05}
{"train_loss": 0.09937440603971481, "global_step": 129778, "epoch": 1458, "lr": 5.358895895775395e-05}
{"train_loss": 0.10733198374509811, "global_step": 129779, "epoch": 1458, "lr": 5.358838071631214e-05}
{"train_loss": 0.09472370147705078, "global_step": 129780, "epoch": 1458, "lr": 5.358780247438792e-05}
{"train_loss": 0.13640649616718292, "global_step": 129781, "epoch": 1458, "lr": 5.358722423198136e-05}
{"train_loss": 0.09654851257801056, "global_step": 129782, "epoch": 1458, "lr": 5.358664598909253e-05}
{"train_loss": 0.09708868712186813, "global_step": 129783, "epoch": 1458, "lr": 5.3586067745721533e-05}
{"train_loss": 0.11881004273891449, "global_step": 129784, "epoch": 1458, "lr": 5.3585489501868445e-05}
{"train_loss": 0.11627286672592163, "global_step": 129785, "epoch": 1458, "lr": 5.358491125753331e-05}
{"train_loss": 0.08842455595731735, "global_step": 129786, "epoch": 1458, "lr": 5.358433301271624e-05}
{"train_loss": 0.21762898564338684, "global_step": 129787, "epoch": 1458, "lr": 5.358375476741729e-05}
{"train_loss": 0.0928938239812851, "global_step": 129788, "epoch": 1458, "lr": 5.358317652163656e-05}
{"train_loss": 0.10885999351739883, "global_step": 129789, "epoch": 1458, "lr": 5.358259827537411e-05}
{"train_loss": 0.09247569739818573, "global_step": 129790, "epoch": 1458, "lr": 5.358202002863002e-05}
{"train_loss": 0.057169109582901, "global_step": 129791, "epoch": 1458, "lr": 5.358144178140436e-05}
{"train_loss": 0.1515045166015625, "global_step": 129792, "epoch": 1458, "lr": 5.358086353369723e-05}
{"train_loss": 0.13866086304187775, "global_step": 129793, "epoch": 1458, "lr": 5.35802852855087e-05}
{"train_loss": 0.1595810055732727, "global_step": 129794, "epoch": 1458, "lr": 5.357970703683884e-05}
{"train_loss": 0.09501660615205765, "global_step": 129795, "epoch": 1458, "lr": 5.357912878768774e-05}
{"train_loss": 0.12086834758520126, "global_step": 129796, "epoch": 1458, "lr": 5.357855053805546e-05}
{"train_loss": 0.15331247448921204, "global_step": 129797, "epoch": 1458, "lr": 5.357797228794208e-05}
{"train_loss": 0.11839812248945236, "global_step": 129798, "epoch": 1458, "lr": 5.357739403734769e-05}
{"train_loss": 0.08867944777011871, "global_step": 129799, "epoch": 1458, "lr": 5.357681578627238e-05}
{"train_loss": 0.10491084307432175, "global_step": 129800, "epoch": 1458, "lr": 5.35762375347162e-05}
{"train_loss": 0.15649601817131042, "global_step": 129801, "epoch": 1458, "lr": 5.357565928267924e-05}
{"train_loss": 0.11430687457323074, "global_step": 129802, "epoch": 1458, "lr": 5.357508103016158e-05}
{"train_loss": 0.13936595618724823, "global_step": 129803, "epoch": 1458, "lr": 5.357450277716329e-05}
{"train_loss": 0.08696229755878448, "global_step": 129804, "epoch": 1458, "lr": 5.3573924523684446e-05}
{"train_loss": 0.08374447375535965, "global_step": 129805, "epoch": 1458, "lr": 5.357334626972514e-05}
{"train_loss": 0.14433106780052185, "global_step": 129806, "epoch": 1458, "lr": 5.357276801528544e-05}
{"train_loss": 0.10157294571399689, "global_step": 129807, "epoch": 1458, "lr": 5.357218976036542e-05}
{"train_loss": 0.12752673029899597, "global_step": 129808, "epoch": 1458, "lr": 5.357161150496517e-05}
{"train_loss": 0.09326519072055817, "global_step": 129809, "epoch": 1458, "lr": 5.3571033249084746e-05}
{"train_loss": 0.09666328132152557, "global_step": 129810, "epoch": 1458, "lr": 5.357045499272425e-05}
{"train_loss": 0.11949117481708527, "global_step": 129811, "epoch": 1458, "lr": 5.356987673588375e-05}
{"train_loss": 0.1565709114074707, "global_step": 129812, "epoch": 1458, "lr": 5.3569298478563326e-05}
{"train_loss": 0.10488235205411911, "global_step": 129813, "epoch": 1458, "lr": 5.3568720220763046e-05}
{"train_loss": 0.1777658462524414, "global_step": 129814, "epoch": 1458, "lr": 5.3568141962483e-05}
{"train_loss": 0.08795598894357681, "global_step": 129815, "epoch": 1458, "lr": 5.3567563703723244e-05}
{"train_loss": 0.13201361894607544, "global_step": 129816, "epoch": 1458, "lr": 5.356698544448388e-05}
{"train_loss": 0.07580594718456268, "global_step": 129817, "epoch": 1458, "lr": 5.3566407184765e-05}
{"train_loss": 0.1171572282910347, "global_step": 129818, "epoch": 1458, "lr": 5.356582892456664e-05}
{"train_loss": 0.14035563170909882, "global_step": 129819, "epoch": 1458, "lr": 5.356525066388891e-05}
{"train_loss": 0.14765159785747528, "global_step": 129820, "epoch": 1458, "lr": 5.3564672402731866e-05}
{"train_loss": 0.22989439964294434, "global_step": 129821, "epoch": 1458, "lr": 5.35640941410956e-05}
{"train_loss": 0.11899764090776443, "global_step": 129822, "epoch": 1458, "lr": 5.3563515878980195e-05}
{"train_loss": 0.1359390914440155, "global_step": 129823, "epoch": 1458, "lr": 5.356293761638571e-05}
{"train_loss": 0.04672452434897423, "global_step": 129824, "epoch": 1458, "lr": 5.3562359353312216e-05}
{"train_loss": 0.11411452293395996, "global_step": 129825, "epoch": 1458, "lr": 5.356178108975983e-05}
{"train_loss": 0.09999994188547134, "global_step": 129826, "epoch": 1458, "lr": 5.3561202825728595e-05}
{"train_loss": 0.11400771141052246, "global_step": 129827, "epoch": 1458, "lr": 5.3560624561218595e-05}
{"train_loss": 0.08940475434064865, "global_step": 129828, "epoch": 1458, "lr": 5.3560046296229913e-05}
{"train_loss": 0.1277172565460205, "global_step": 129829, "epoch": 1458, "lr": 5.355946803076264e-05}
{"train_loss": 0.16101473569869995, "global_step": 129830, "epoch": 1458, "lr": 5.355888976481682e-05}
{"train_loss": 0.11532961577177048, "global_step": 129831, "epoch": 1458, "lr": 5.355831149839257e-05}
{"train_loss": 0.12575304508209229, "global_step": 129832, "epoch": 1458, "lr": 5.355773323148994e-05}
{"train_loss": 0.13777238130569458, "global_step": 129833, "epoch": 1458, "lr": 5.3557154964109005e-05}
{"train_loss": 0.09391475468873978, "global_step": 129834, "epoch": 1458, "lr": 5.3556576696249864e-05}
{"train_loss": 0.08661464601755142, "global_step": 129835, "epoch": 1458, "lr": 5.3555998427912576e-05}
{"train_loss": 0.13179677724838257, "global_step": 129836, "epoch": 1458, "lr": 5.355542015909725e-05}
{"train_loss": 0.09498598426580429, "global_step": 129837, "epoch": 1458, "lr": 5.355484188980392e-05}
{"train_loss": 0.19549626111984253, "global_step": 129838, "epoch": 1458, "lr": 5.3554263620032687e-05}
{"train_loss": 0.07914933562278748, "global_step": 129839, "epoch": 1458, "lr": 5.355368534978363e-05}
{"train_loss": 0.17477719485759735, "global_step": 129840, "epoch": 1458, "lr": 5.355310707905683e-05}
{"train_loss": 0.11372177302837372, "global_step": 129841, "epoch": 1458, "lr": 5.355252880785236e-05}
{"train_loss": 0.17746111750602722, "global_step": 129842, "epoch": 1458, "lr": 5.355195053617029e-05}
{"train_loss": 0.15680520236492157, "global_step": 129843, "epoch": 1458, "lr": 5.3551372264010704e-05}
{"train_loss": 0.12865103781223297, "global_step": 129844, "epoch": 1458, "lr": 5.355079399137368e-05}
{"train_loss": 0.1156957596540451, "global_step": 129845, "epoch": 1458, "lr": 5.355021571825929e-05}
{"train_loss": 0.08646681159734726, "global_step": 129846, "epoch": 1458, "lr": 5.354963744466762e-05}
{"train_loss": 0.11731632798910141, "global_step": 129847, "epoch": 1458, "lr": 5.354905917059875e-05}
{"train_loss": 0.10158171504735947, "global_step": 129848, "epoch": 1458, "lr": 5.3548480896052746e-05}
{"train_loss": 0.1252470314502716, "global_step": 129849, "epoch": 1458, "lr": 5.354790262102969e-05}
{"train_loss": 0.12298771877134784, "global_step": 129850, "epoch": 1458, "lr": 5.3547324345529656e-05, "val_loss": 5.716617107391357}
{"train_loss": 0.09982317686080933, "global_step": 129851, "epoch": 1459, "lr": 5.3546746069552734e-05}
{"train_loss": 0.11507270485162735, "global_step": 129852, "epoch": 1459, "lr": 5.354616779309899e-05}
{"train_loss": 0.1021176129579544, "global_step": 129853, "epoch": 1459, "lr": 5.3545589516168516e-05}
{"train_loss": 0.13936489820480347, "global_step": 129854, "epoch": 1459, "lr": 5.3545011238761376e-05}
{"train_loss": 0.17344392836093903, "global_step": 129855, "epoch": 1459, "lr": 5.3544432960877654e-05}
{"train_loss": 0.12638096511363983, "global_step": 129856, "epoch": 1459, "lr": 5.354385468251743e-05}
{"train_loss": 0.08827383071184158, "global_step": 129857, "epoch": 1459, "lr": 5.354327640368078e-05}
{"train_loss": 0.13933922350406647, "global_step": 129858, "epoch": 1459, "lr": 5.3542698124367773e-05}
{"train_loss": 0.1143040582537651, "global_step": 129859, "epoch": 1459, "lr": 5.3542119844578485e-05}
{"train_loss": 0.10030469298362732, "global_step": 129860, "epoch": 1459, "lr": 5.3541541564313015e-05}
{"train_loss": 0.1706046611070633, "global_step": 129861, "epoch": 1459, "lr": 5.3540963283571445e-05}
{"train_loss": 0.14514203369617462, "global_step": 129862, "epoch": 1459, "lr": 5.354038500235381e-05}
{"train_loss": 0.16180221736431122, "global_step": 129863, "epoch": 1459, "lr": 5.353980672066022e-05}
{"train_loss": 0.09689997881650925, "global_step": 129864, "epoch": 1459, "lr": 5.353922843849074e-05}
{"train_loss": 0.0599014088511467, "global_step": 129865, "epoch": 1459, "lr": 5.353865015584546e-05}
{"train_loss": 0.07828997075557709, "global_step": 129866, "epoch": 1459, "lr": 5.353807187272446e-05}
{"train_loss": 0.12285120040178299, "global_step": 129867, "epoch": 1459, "lr": 5.3537493589127794e-05}
{"train_loss": 0.1052614226937294, "global_step": 129868, "epoch": 1459, "lr": 5.3536915305055555e-05}
{"train_loss": 0.10844133049249649, "global_step": 129869, "epoch": 1459, "lr": 5.353633702050782e-05}
{"train_loss": 0.10194205492734909, "global_step": 129870, "epoch": 1459, "lr": 5.353575873548467e-05}
{"train_loss": 0.09366174042224884, "global_step": 129871, "epoch": 1459, "lr": 5.3535180449986186e-05}
{"train_loss": 0.14010103046894073, "global_step": 129872, "epoch": 1459, "lr": 5.3534602164012446e-05}
{"train_loss": 0.2209269106388092, "global_step": 129873, "epoch": 1459, "lr": 5.3534023877563514e-05}
{"train_loss": 0.10078354924917221, "global_step": 129874, "epoch": 1459, "lr": 5.353344559063948e-05}
{"train_loss": 0.19194923341274261, "global_step": 129875, "epoch": 1459, "lr": 5.353286730324041e-05}
{"train_loss": 0.093453049659729, "global_step": 129876, "epoch": 1459, "lr": 5.3532289015366396e-05}
{"train_loss": 0.16960929334163666, "global_step": 129877, "epoch": 1459, "lr": 5.35317107270175e-05}
{"train_loss": 0.0969342365860939, "global_step": 129878, "epoch": 1459, "lr": 5.353113243819382e-05}
{"train_loss": 0.09605275094509125, "global_step": 129879, "epoch": 1459, "lr": 5.3530554148895415e-05}
{"train_loss": 0.1317070722579956, "global_step": 129880, "epoch": 1459, "lr": 5.352997585912238e-05}
{"train_loss": 0.08752759546041489, "global_step": 129881, "epoch": 1459, "lr": 5.3529397568874776e-05}
{"train_loss": 0.12317071110010147, "global_step": 129882, "epoch": 1459, "lr": 5.3528819278152684e-05}
{"train_loss": 0.12181837111711502, "global_step": 129883, "epoch": 1459, "lr": 5.352824098695619e-05}
{"train_loss": 0.09608664363622665, "global_step": 129884, "epoch": 1459, "lr": 5.3527662695285364e-05}
{"train_loss": 0.06441617012023926, "global_step": 129885, "epoch": 1459, "lr": 5.3527084403140284e-05}
{"train_loss": 0.06335864961147308, "global_step": 129886, "epoch": 1459, "lr": 5.3526506110521035e-05}
{"train_loss": 0.12413689494132996, "global_step": 129887, "epoch": 1459, "lr": 5.352592781742768e-05}
{"train_loss": 0.12382214516401291, "global_step": 129888, "epoch": 1459, "lr": 5.3525349523860314e-05}
{"train_loss": 0.1610540747642517, "global_step": 129889, "epoch": 1459, "lr": 5.3524771229819006e-05}
{"train_loss": 0.12057631462812424, "global_step": 129890, "epoch": 1459, "lr": 5.352419293530384e-05}
{"train_loss": 0.08854620903730392, "global_step": 129891, "epoch": 1459, "lr": 5.352361464031489e-05}
{"train_loss": 0.17807844281196594, "global_step": 129892, "epoch": 1459, "lr": 5.3523036344852226e-05}
{"train_loss": 0.14742083847522736, "global_step": 129893, "epoch": 1459, "lr": 5.3522458048915936e-05}
{"train_loss": 0.07493782788515091, "global_step": 129894, "epoch": 1459, "lr": 5.35218797525061e-05}
{"train_loss": 0.06654266268014908, "global_step": 129895, "epoch": 1459, "lr": 5.3521301455622785e-05}
{"train_loss": 0.11533991992473602, "global_step": 129896, "epoch": 1459, "lr": 5.3520723158266074e-05}
{"train_loss": 0.07965097576379776, "global_step": 129897, "epoch": 1459, "lr": 5.352014486043605e-05}
{"train_loss": 0.10594460368156433, "global_step": 129898, "epoch": 1459, "lr": 5.3519566562132774e-05}
{"train_loss": 0.07682324200868607, "global_step": 129899, "epoch": 1459, "lr": 5.3518988263356354e-05}
{"train_loss": 0.13680851459503174, "global_step": 129900, "epoch": 1459, "lr": 5.351840996410684e-05}
{"train_loss": 0.16683551669120789, "global_step": 129901, "epoch": 1459, "lr": 5.351783166438432e-05}
{"train_loss": 0.11914097517728806, "global_step": 129902, "epoch": 1459, "lr": 5.351725336418887e-05}
{"train_loss": 0.16837936639785767, "global_step": 129903, "epoch": 1459, "lr": 5.351667506352056e-05}
{"train_loss": 0.12887580692768097, "global_step": 129904, "epoch": 1459, "lr": 5.3516096762379475e-05}
{"train_loss": 0.09533185511827469, "global_step": 129905, "epoch": 1459, "lr": 5.35155184607657e-05}
{"train_loss": 0.12908436357975006, "global_step": 129906, "epoch": 1459, "lr": 5.35149401586793e-05}
{"train_loss": 0.12197621911764145, "global_step": 129907, "epoch": 1459, "lr": 5.351436185612037e-05}
{"train_loss": 0.09785602986812592, "global_step": 129908, "epoch": 1459, "lr": 5.3513783553088956e-05}
{"train_loss": 0.08204644173383713, "global_step": 129909, "epoch": 1459, "lr": 5.351320524958517e-05}
{"train_loss": 0.14593110978603363, "global_step": 129910, "epoch": 1459, "lr": 5.351262694560909e-05}
{"train_loss": 0.16254732012748718, "global_step": 129911, "epoch": 1459, "lr": 5.351204864116076e-05}
{"train_loss": 0.1461830735206604, "global_step": 129912, "epoch": 1459, "lr": 5.351147033624029e-05}
{"train_loss": 0.14198416471481323, "global_step": 129913, "epoch": 1459, "lr": 5.351089203084775e-05}
{"train_loss": 0.12547077238559723, "global_step": 129914, "epoch": 1459, "lr": 5.3510313724983195e-05}
{"train_loss": 0.11339942365884781, "global_step": 129915, "epoch": 1459, "lr": 5.3509735418646735e-05}
{"train_loss": 0.07937205582857132, "global_step": 129916, "epoch": 1459, "lr": 5.3509157111838424e-05}
{"train_loss": 0.11613833904266357, "global_step": 129917, "epoch": 1459, "lr": 5.350857880455836e-05}
{"train_loss": 0.14846651256084442, "global_step": 129918, "epoch": 1459, "lr": 5.350800049680661e-05}
{"train_loss": 0.14911918342113495, "global_step": 129919, "epoch": 1459, "lr": 5.3507422188583255e-05}
{"train_loss": 0.09814997017383575, "global_step": 129920, "epoch": 1459, "lr": 5.350684387988837e-05}
{"train_loss": 0.09757519513368607, "global_step": 129921, "epoch": 1459, "lr": 5.350626557072202e-05}
{"train_loss": 0.11100636422634125, "global_step": 129922, "epoch": 1459, "lr": 5.35056872610843e-05}
{"train_loss": 0.06815474480390549, "global_step": 129923, "epoch": 1459, "lr": 5.350510895097529e-05}
{"train_loss": 0.15886898338794708, "global_step": 129924, "epoch": 1459, "lr": 5.350453064039504e-05}
{"train_loss": 0.07409149408340454, "global_step": 129925, "epoch": 1459, "lr": 5.3503952329343666e-05}
{"train_loss": 0.12966549396514893, "global_step": 129926, "epoch": 1459, "lr": 5.350337401782123e-05}
{"train_loss": 0.13646508753299713, "global_step": 129927, "epoch": 1459, "lr": 5.35027957058278e-05}
{"train_loss": 0.07620539516210556, "global_step": 129928, "epoch": 1459, "lr": 5.350221739336346e-05}
{"train_loss": 0.1406400501728058, "global_step": 129929, "epoch": 1459, "lr": 5.35016390804283e-05}
{"train_loss": 0.13270311057567596, "global_step": 129930, "epoch": 1459, "lr": 5.3501060767022383e-05}
{"train_loss": 0.14273367822170258, "global_step": 129931, "epoch": 1459, "lr": 5.350048245314579e-05}
{"train_loss": 0.10253233462572098, "global_step": 129932, "epoch": 1459, "lr": 5.34999041387986e-05}
{"train_loss": 0.06351438164710999, "global_step": 129933, "epoch": 1459, "lr": 5.34993258239809e-05}
{"train_loss": 0.09331127256155014, "global_step": 129934, "epoch": 1459, "lr": 5.3498747508692736e-05}
{"train_loss": 0.12478744238615036, "global_step": 129935, "epoch": 1459, "lr": 5.349816919293422e-05}
{"train_loss": 0.1344590187072754, "global_step": 129936, "epoch": 1459, "lr": 5.3497590876705426e-05}
{"train_loss": 0.0655851662158966, "global_step": 129937, "epoch": 1459, "lr": 5.349701256000642e-05}
{"train_loss": 0.12893371284008026, "global_step": 129938, "epoch": 1459, "lr": 5.349643424283729e-05}
{"train_loss": 0.11773521453142166, "global_step": 129939, "epoch": 1459, "lr": 5.349585592519809e-05, "val_loss": 5.71494722366333}
{"train_loss": 0.11417927592992783, "global_step": 129940, "epoch": 1460, "lr": 5.349527760708892e-05}
{"train_loss": 0.09423103928565979, "global_step": 129941, "epoch": 1460, "lr": 5.349469928850986e-05}
{"train_loss": 0.21685615181922913, "global_step": 129942, "epoch": 1460, "lr": 5.3494120969460956e-05}
{"train_loss": 0.1446295827627182, "global_step": 129943, "epoch": 1460, "lr": 5.349354264994234e-05}
{"train_loss": 0.14871324598789215, "global_step": 129944, "epoch": 1460, "lr": 5.349296432995404e-05}
{"train_loss": 0.1323915719985962, "global_step": 129945, "epoch": 1460, "lr": 5.3492386009496166e-05}
{"train_loss": 0.14244325459003448, "global_step": 129946, "epoch": 1460, "lr": 5.3491807688568784e-05}
{"train_loss": 0.11650475114583969, "global_step": 129947, "epoch": 1460, "lr": 5.349122936717197e-05}
{"train_loss": 0.09694793820381165, "global_step": 129948, "epoch": 1460, "lr": 5.349065104530579e-05}
{"train_loss": 0.10360616445541382, "global_step": 129949, "epoch": 1460, "lr": 5.3490072722970344e-05}
{"train_loss": 0.11666668951511383, "global_step": 129950, "epoch": 1460, "lr": 5.348949440016569e-05}
{"train_loss": 0.11197284609079361, "global_step": 129951, "epoch": 1460, "lr": 5.3488916076891936e-05}
{"train_loss": 0.12517845630645752, "global_step": 129952, "epoch": 1460, "lr": 5.348833775314913e-05}
{"train_loss": 0.1377912014722824, "global_step": 129953, "epoch": 1460, "lr": 5.348775942893736e-05}
{"train_loss": 0.14940953254699707, "global_step": 129954, "epoch": 1460, "lr": 5.34871811042567e-05}
{"train_loss": 0.11580489575862885, "global_step": 129955, "epoch": 1460, "lr": 5.348660277910723e-05}
{"train_loss": 0.1414172649383545, "global_step": 129956, "epoch": 1460, "lr": 5.3486024453489034e-05}
{"train_loss": 0.22989492118358612, "global_step": 129957, "epoch": 1460, "lr": 5.3485446127402185e-05}
{"train_loss": 0.16086111962795258, "global_step": 129958, "epoch": 1460, "lr": 5.348486780084677e-05}
{"train_loss": 0.10596413165330887, "global_step": 129959, "epoch": 1460, "lr": 5.348428947382285e-05}
{"train_loss": 0.12122060358524323, "global_step": 129960, "epoch": 1460, "lr": 5.3483711146330486e-05}
{"train_loss": 0.11804980784654617, "global_step": 129961, "epoch": 1460, "lr": 5.34831328183698e-05}
{"train_loss": 0.15701723098754883, "global_step": 129962, "epoch": 1460, "lr": 5.348255448994085e-05}
{"train_loss": 0.08146605640649796, "global_step": 129963, "epoch": 1460, "lr": 5.348197616104371e-05}
{"train_loss": 0.13977238535881042, "global_step": 129964, "epoch": 1460, "lr": 5.348139783167846e-05}
{"train_loss": 0.154188334941864, "global_step": 129965, "epoch": 1460, "lr": 5.348081950184518e-05}
{"train_loss": 0.10708338767290115, "global_step": 129966, "epoch": 1460, "lr": 5.348024117154394e-05}
{"train_loss": 0.1283063143491745, "global_step": 129967, "epoch": 1460, "lr": 5.3479662840774836e-05}
{"train_loss": 0.10535451024770737, "global_step": 129968, "epoch": 1460, "lr": 5.3479084509537925e-05}
{"train_loss": 0.08723540604114532, "global_step": 129969, "epoch": 1460, "lr": 5.3478506177833296e-05}
{"train_loss": 0.1473892480134964, "global_step": 129970, "epoch": 1460, "lr": 5.3477927845661024e-05}
{"train_loss": 0.10263566672801971, "global_step": 129971, "epoch": 1460, "lr": 5.347734951302119e-05}
{"train_loss": 0.1232716515660286, "global_step": 129972, "epoch": 1460, "lr": 5.347677117991387e-05}
{"train_loss": 0.09976094961166382, "global_step": 129973, "epoch": 1460, "lr": 5.347619284633913e-05}
{"train_loss": 0.14173023402690887, "global_step": 129974, "epoch": 1460, "lr": 5.3475614512297065e-05}
{"train_loss": 0.06301334500312805, "global_step": 129975, "epoch": 1460, "lr": 5.347503617778774e-05}
{"train_loss": 0.135496124625206, "global_step": 129976, "epoch": 1460, "lr": 5.347445784281124e-05}
{"train_loss": 0.10267581045627594, "global_step": 129977, "epoch": 1460, "lr": 5.3473879507367666e-05}
{"train_loss": 0.12057342380285263, "global_step": 129978, "epoch": 1460, "lr": 5.347330117145704e-05}
{"train_loss": 0.11964072287082672, "global_step": 129979, "epoch": 1460, "lr": 5.3472722835079483e-05}
{"train_loss": 0.10181921720504761, "global_step": 129980, "epoch": 1460, "lr": 5.347214449823505e-05}
{"train_loss": 0.09343487024307251, "global_step": 129981, "epoch": 1460, "lr": 5.3471566160923844e-05}
{"train_loss": 0.09599315375089645, "global_step": 129982, "epoch": 1460, "lr": 5.347098782314592e-05}
{"train_loss": 0.07697765529155731, "global_step": 129983, "epoch": 1460, "lr": 5.347040948490137e-05}
{"train_loss": 0.10727512836456299, "global_step": 129984, "epoch": 1460, "lr": 5.346983114619026e-05}
{"train_loss": 0.10330476611852646, "global_step": 129985, "epoch": 1460, "lr": 5.346925280701268e-05}
{"train_loss": 0.11340127885341644, "global_step": 129986, "epoch": 1460, "lr": 5.3468674467368705e-05}
{"train_loss": 0.13593262434005737, "global_step": 129987, "epoch": 1460, "lr": 5.3468096127258406e-05}
{"train_loss": 0.1244845762848854, "global_step": 129988, "epoch": 1460, "lr": 5.346751778668185e-05}
{"train_loss": 0.09643224626779556, "global_step": 129989, "epoch": 1460, "lr": 5.346693944563914e-05}
{"train_loss": 0.21050232648849487, "global_step": 129990, "epoch": 1460, "lr": 5.346636110413035e-05}
{"train_loss": 0.14326642453670502, "global_step": 129991, "epoch": 1460, "lr": 5.346578276215554e-05}
{"train_loss": 0.12288336455821991, "global_step": 129992, "epoch": 1460, "lr": 5.346520441971479e-05}
{"train_loss": 0.11074300110340118, "global_step": 129993, "epoch": 1460, "lr": 5.3464626076808196e-05}
{"train_loss": 0.08349428325891495, "global_step": 129994, "epoch": 1460, "lr": 5.346404773343582e-05}
{"train_loss": 0.09920832514762878, "global_step": 129995, "epoch": 1460, "lr": 5.346346938959775e-05}
{"train_loss": 0.11211317777633667, "global_step": 129996, "epoch": 1460, "lr": 5.346289104529406e-05}
{"train_loss": 0.09269727021455765, "global_step": 129997, "epoch": 1460, "lr": 5.346231270052482e-05}
{"train_loss": 0.11287134140729904, "global_step": 129998, "epoch": 1460, "lr": 5.3461734355290114e-05}
{"train_loss": 0.08378204703330994, "global_step": 129999, "epoch": 1460, "lr": 5.346115600959003e-05}
{"train_loss": 0.11036095768213272, "global_step": 130000, "epoch": 1460, "lr": 5.346057766342463e-05}
{"train_loss": 0.13393042981624603, "global_step": 130001, "epoch": 1460, "lr": 5.3459999316794e-05}
{"train_loss": 0.13873475790023804, "global_step": 130002, "epoch": 1460, "lr": 5.3459420969698205e-05}
{"train_loss": 0.12222882360219955, "global_step": 130003, "epoch": 1460, "lr": 5.3458842622137337e-05}
{"train_loss": 0.1109919399023056, "global_step": 130004, "epoch": 1460, "lr": 5.345826427411148e-05}
{"train_loss": 0.1175922229886055, "global_step": 130005, "epoch": 1460, "lr": 5.345768592562069e-05}
{"train_loss": 0.13536040484905243, "global_step": 130006, "epoch": 1460, "lr": 5.3457107576665065e-05}
{"train_loss": 0.19708144664764404, "global_step": 130007, "epoch": 1460, "lr": 5.345652922724467e-05}
{"train_loss": 0.1550782173871994, "global_step": 130008, "epoch": 1460, "lr": 5.345595087735959e-05}
{"train_loss": 0.17791932821273804, "global_step": 130009, "epoch": 1460, "lr": 5.345537252700989e-05}
{"train_loss": 0.17893043160438538, "global_step": 130010, "epoch": 1460, "lr": 5.3454794176195666e-05}
{"train_loss": 0.10380435734987259, "global_step": 130011, "epoch": 1460, "lr": 5.3454215824916986e-05}
{"train_loss": 0.1257910132408142, "global_step": 130012, "epoch": 1460, "lr": 5.345363747317392e-05}
{"train_loss": 0.14740358293056488, "global_step": 130013, "epoch": 1460, "lr": 5.3453059120966564e-05}
{"train_loss": 0.11825656145811081, "global_step": 130014, "epoch": 1460, "lr": 5.345248076829499e-05}
{"train_loss": 0.095158152282238, "global_step": 130015, "epoch": 1460, "lr": 5.345190241515926e-05}
{"train_loss": 0.11579874157905579, "global_step": 130016, "epoch": 1460, "lr": 5.345132406155947e-05}
{"train_loss": 0.1515980064868927, "global_step": 130017, "epoch": 1460, "lr": 5.345074570749568e-05}
{"train_loss": 0.1330825537443161, "global_step": 130018, "epoch": 1460, "lr": 5.345016735296801e-05}
{"train_loss": 0.11026225984096527, "global_step": 130019, "epoch": 1460, "lr": 5.344958899797647e-05}
{"train_loss": 0.1261829137802124, "global_step": 130020, "epoch": 1460, "lr": 5.34490106425212e-05}
{"train_loss": 0.09398232400417328, "global_step": 130021, "epoch": 1460, "lr": 5.3448432286602247e-05}
{"train_loss": 0.14446976780891418, "global_step": 130022, "epoch": 1460, "lr": 5.344785393021969e-05}
{"train_loss": 0.15794093906879425, "global_step": 130023, "epoch": 1460, "lr": 5.3447275573373614e-05}
{"train_loss": 0.18646174669265747, "global_step": 130024, "epoch": 1460, "lr": 5.344669721606409e-05}
{"train_loss": 0.10875508189201355, "global_step": 130025, "epoch": 1460, "lr": 5.344611885829121e-05}
{"train_loss": 0.14515089988708496, "global_step": 130026, "epoch": 1460, "lr": 5.344554050005503e-05}
{"train_loss": 0.16013406217098236, "global_step": 130027, "epoch": 1460, "lr": 5.344496214135565e-05}
{"train_loss": 0.125877424572291, "global_step": 130028, "epoch": 1460, "lr": 5.344438378219312e-05, "val_loss": 5.8738017082214355, "train_action_mse_error": 18.541908264160156}
{"train_loss": 0.10153356194496155, "global_step": 130029, "epoch": 1461, "lr": 5.344380542256755e-05}
{"train_loss": 0.08075489103794098, "global_step": 130030, "epoch": 1461, "lr": 5.3443227062479e-05}
{"train_loss": 0.09867662936449051, "global_step": 130031, "epoch": 1461, "lr": 5.3442648701927544e-05}
{"train_loss": 0.14128459990024567, "global_step": 130032, "epoch": 1461, "lr": 5.3442070340913264e-05}
{"train_loss": 0.1583661139011383, "global_step": 130033, "epoch": 1461, "lr": 5.3441491979436254e-05}
{"train_loss": 0.15573185682296753, "global_step": 130034, "epoch": 1461, "lr": 5.344091361749656e-05}
{"train_loss": 0.1325678676366806, "global_step": 130035, "epoch": 1461, "lr": 5.3440335255094285e-05}
{"train_loss": 0.08525358140468597, "global_step": 130036, "epoch": 1461, "lr": 5.3439756892229496e-05}
{"train_loss": 0.17117443680763245, "global_step": 130037, "epoch": 1461, "lr": 5.343917852890228e-05}
{"train_loss": 0.206997349858284, "global_step": 130038, "epoch": 1461, "lr": 5.343860016511271e-05}
{"train_loss": 0.14117331802845, "global_step": 130039, "epoch": 1461, "lr": 5.343802180086085e-05}
{"train_loss": 0.11117809265851974, "global_step": 130040, "epoch": 1461, "lr": 5.3437443436146794e-05}
{"train_loss": 0.09932436048984528, "global_step": 130041, "epoch": 1461, "lr": 5.343686507097062e-05}
{"train_loss": 0.08185125142335892, "global_step": 130042, "epoch": 1461, "lr": 5.34362867053324e-05}
{"train_loss": 0.10069544613361359, "global_step": 130043, "epoch": 1461, "lr": 5.343570833923222e-05}
{"train_loss": 0.08538252860307693, "global_step": 130044, "epoch": 1461, "lr": 5.343512997267014e-05}
{"train_loss": 0.2712283134460449, "global_step": 130045, "epoch": 1461, "lr": 5.3434551605646254e-05}
{"train_loss": 0.0969141349196434, "global_step": 130046, "epoch": 1461, "lr": 5.343397323816063e-05}
{"train_loss": 0.14357216656208038, "global_step": 130047, "epoch": 1461, "lr": 5.343339487021335e-05}
{"train_loss": 0.1141515001654625, "global_step": 130048, "epoch": 1461, "lr": 5.3432816501804495e-05}
{"train_loss": 0.1026386171579361, "global_step": 130049, "epoch": 1461, "lr": 5.343223813293414e-05}
{"train_loss": 0.16026602685451508, "global_step": 130050, "epoch": 1461, "lr": 5.343165976360236e-05}
{"train_loss": 0.13295194506645203, "global_step": 130051, "epoch": 1461, "lr": 5.3431081393809245e-05}
{"train_loss": 0.16829507052898407, "global_step": 130052, "epoch": 1461, "lr": 5.343050302355486e-05}
{"train_loss": 0.1663348376750946, "global_step": 130053, "epoch": 1461, "lr": 5.342992465283927e-05}
{"train_loss": 0.15350037813186646, "global_step": 130054, "epoch": 1461, "lr": 5.3429346281662585e-05}
{"train_loss": 0.0792330875992775, "global_step": 130055, "epoch": 1461, "lr": 5.342876791002486e-05}
{"train_loss": 0.07927902787923813, "global_step": 130056, "epoch": 1461, "lr": 5.342818953792618e-05}
{"train_loss": 0.09317460656166077, "global_step": 130057, "epoch": 1461, "lr": 5.342761116536662e-05}
{"train_loss": 0.09451364725828171, "global_step": 130058, "epoch": 1461, "lr": 5.342703279234625e-05}
{"train_loss": 0.12276442348957062, "global_step": 130059, "epoch": 1461, "lr": 5.342645441886518e-05}
{"train_loss": 0.07831322401762009, "global_step": 130060, "epoch": 1461, "lr": 5.342587604492345e-05}
{"train_loss": 0.12270423769950867, "global_step": 130061, "epoch": 1461, "lr": 5.342529767052116e-05}
{"train_loss": 0.12889334559440613, "global_step": 130062, "epoch": 1461, "lr": 5.3424719295658364e-05}
{"train_loss": 0.07646821439266205, "global_step": 130063, "epoch": 1461, "lr": 5.342414092033518e-05}
{"train_loss": 0.08852101117372513, "global_step": 130064, "epoch": 1461, "lr": 5.3423562544551645e-05}
{"train_loss": 0.07602113485336304, "global_step": 130065, "epoch": 1461, "lr": 5.3422984168307855e-05}
{"train_loss": 0.14222878217697144, "global_step": 130066, "epoch": 1461, "lr": 5.342240579160389e-05}
{"train_loss": 0.16333548724651337, "global_step": 130067, "epoch": 1461, "lr": 5.342182741443983e-05}
{"train_loss": 0.1710706502199173, "global_step": 130068, "epoch": 1461, "lr": 5.3421249036815735e-05}
{"train_loss": 0.09473028779029846, "global_step": 130069, "epoch": 1461, "lr": 5.342067065873171e-05}
{"train_loss": 0.17246024310588837, "global_step": 130070, "epoch": 1461, "lr": 5.3420092280187804e-05}
{"train_loss": 0.08219226449728012, "global_step": 130071, "epoch": 1461, "lr": 5.341951390118412e-05}
{"train_loss": 0.1355789601802826, "global_step": 130072, "epoch": 1461, "lr": 5.3418935521720715e-05}
{"train_loss": 0.15424449741840363, "global_step": 130073, "epoch": 1461, "lr": 5.341835714179767e-05}
{"train_loss": 0.07366172969341278, "global_step": 130074, "epoch": 1461, "lr": 5.341777876141508e-05}
{"train_loss": 0.1630869358778, "global_step": 130075, "epoch": 1461, "lr": 5.341720038057301e-05}
{"train_loss": 0.14410355687141418, "global_step": 130076, "epoch": 1461, "lr": 5.341662199927153e-05}
{"train_loss": 0.05596011132001877, "global_step": 130077, "epoch": 1461, "lr": 5.341604361751074e-05}
{"train_loss": 0.1335791051387787, "global_step": 130078, "epoch": 1461, "lr": 5.34154652352907e-05}
{"train_loss": 0.16464641690254211, "global_step": 130079, "epoch": 1461, "lr": 5.341488685261149e-05}
{"train_loss": 0.1378839910030365, "global_step": 130080, "epoch": 1461, "lr": 5.3414308469473184e-05}
{"train_loss": 0.1891787052154541, "global_step": 130081, "epoch": 1461, "lr": 5.341373008587588e-05}
{"train_loss": 0.12542955577373505, "global_step": 130082, "epoch": 1461, "lr": 5.341315170181963e-05}
{"train_loss": 0.13225215673446655, "global_step": 130083, "epoch": 1461, "lr": 5.341257331730453e-05}
{"train_loss": 0.14052213728427887, "global_step": 130084, "epoch": 1461, "lr": 5.341199493233064e-05}
{"train_loss": 0.10934237390756607, "global_step": 130085, "epoch": 1461, "lr": 5.341141654689806e-05}
{"train_loss": 0.1874120533466339, "global_step": 130086, "epoch": 1461, "lr": 5.341083816100686e-05}
{"train_loss": 0.16697373986244202, "global_step": 130087, "epoch": 1461, "lr": 5.341025977465709e-05}
{"train_loss": 0.09929262101650238, "global_step": 130088, "epoch": 1461, "lr": 5.3409681387848884e-05}
{"train_loss": 0.17060567438602448, "global_step": 130089, "epoch": 1461, "lr": 5.3409103000582276e-05}
{"train_loss": 0.14017412066459656, "global_step": 130090, "epoch": 1461, "lr": 5.340852461285736e-05}
{"train_loss": 0.16979600489139557, "global_step": 130091, "epoch": 1461, "lr": 5.34079462246742e-05}
{"train_loss": 0.10333121567964554, "global_step": 130092, "epoch": 1461, "lr": 5.3407367836032885e-05}
{"train_loss": 0.16866502165794373, "global_step": 130093, "epoch": 1461, "lr": 5.3406789446933494e-05}
{"train_loss": 0.1157115250825882, "global_step": 130094, "epoch": 1461, "lr": 5.34062110573761e-05}
{"train_loss": 0.16441188752651215, "global_step": 130095, "epoch": 1461, "lr": 5.340563266736078e-05}
{"train_loss": 0.12325825542211533, "global_step": 130096, "epoch": 1461, "lr": 5.3405054276887626e-05}
{"train_loss": 0.19890247285366058, "global_step": 130097, "epoch": 1461, "lr": 5.3404475885956684e-05}
{"train_loss": 0.11305709928274155, "global_step": 130098, "epoch": 1461, "lr": 5.340389749456807e-05}
{"train_loss": 0.13303253054618835, "global_step": 130099, "epoch": 1461, "lr": 5.340331910272183e-05}
{"train_loss": 0.11660254746675491, "global_step": 130100, "epoch": 1461, "lr": 5.3402740710418065e-05}
{"train_loss": 0.18635188043117523, "global_step": 130101, "epoch": 1461, "lr": 5.3402162317656835e-05}
{"train_loss": 0.0840529054403305, "global_step": 130102, "epoch": 1461, "lr": 5.3401583924438234e-05}
{"train_loss": 0.155917689204216, "global_step": 130103, "epoch": 1461, "lr": 5.340100553076233e-05}
{"train_loss": 0.13635535538196564, "global_step": 130104, "epoch": 1461, "lr": 5.340042713662919e-05}
{"train_loss": 0.19612360000610352, "global_step": 130105, "epoch": 1461, "lr": 5.3399848742038905e-05}
{"train_loss": 0.16445277631282806, "global_step": 130106, "epoch": 1461, "lr": 5.339927034699156e-05}
{"train_loss": 0.1856103390455246, "global_step": 130107, "epoch": 1461, "lr": 5.339869195148724e-05}
{"train_loss": 0.11823493987321854, "global_step": 130108, "epoch": 1461, "lr": 5.3398113555526e-05}
{"train_loss": 0.09167105704545975, "global_step": 130109, "epoch": 1461, "lr": 5.3397535159107916e-05}
{"train_loss": 0.12178965657949448, "global_step": 130110, "epoch": 1461, "lr": 5.3396956762233074e-05}
{"train_loss": 0.2269236445426941, "global_step": 130111, "epoch": 1461, "lr": 5.339637836490156e-05}
{"train_loss": 0.15638384222984314, "global_step": 130112, "epoch": 1461, "lr": 5.339579996711344e-05}
{"train_loss": 0.12970909476280212, "global_step": 130113, "epoch": 1461, "lr": 5.339522156886878e-05}
{"train_loss": 0.11669906973838806, "global_step": 130114, "epoch": 1461, "lr": 5.339464317016769e-05}
{"train_loss": 0.18626601994037628, "global_step": 130115, "epoch": 1461, "lr": 5.3394064771010235e-05}
{"train_loss": 0.10730206966400146, "global_step": 130116, "epoch": 1461, "lr": 5.339348637139648e-05}
{"train_loss": 0.13274939709834838, "global_step": 130117, "epoch": 1461, "lr": 5.339290797132652e-05, "val_loss": 5.637734413146973}
{"train_loss": 0.13116656243801117, "global_step": 130118, "epoch": 1462, "lr": 5.3392329570800426e-05}
{"train_loss": 0.11259451508522034, "global_step": 130119, "epoch": 1462, "lr": 5.339175116981827e-05}
{"train_loss": 0.1550295650959015, "global_step": 130120, "epoch": 1462, "lr": 5.339117276838014e-05}
{"train_loss": 0.14764021337032318, "global_step": 130121, "epoch": 1462, "lr": 5.3390594366486105e-05}
{"train_loss": 0.09723073989152908, "global_step": 130122, "epoch": 1462, "lr": 5.3390015964136245e-05}
{"train_loss": 0.09315058588981628, "global_step": 130123, "epoch": 1462, "lr": 5.338943756133061e-05}
{"train_loss": 0.12621824443340302, "global_step": 130124, "epoch": 1462, "lr": 5.338885915806934e-05}
{"train_loss": 0.13745521008968353, "global_step": 130125, "epoch": 1462, "lr": 5.338828075435247e-05}
{"train_loss": 0.153899684548378, "global_step": 130126, "epoch": 1462, "lr": 5.33877023501801e-05}
{"train_loss": 0.09628494083881378, "global_step": 130127, "epoch": 1462, "lr": 5.33871239455523e-05}
{"train_loss": 0.12559907138347626, "global_step": 130128, "epoch": 1462, "lr": 5.338654554046912e-05}
{"train_loss": 0.1123398020863533, "global_step": 130129, "epoch": 1462, "lr": 5.338596713493067e-05}
{"train_loss": 0.1738373041152954, "global_step": 130130, "epoch": 1462, "lr": 5.338538872893701e-05}
{"train_loss": 0.11091689765453339, "global_step": 130131, "epoch": 1462, "lr": 5.338481032248823e-05}
{"train_loss": 0.08607299625873566, "global_step": 130132, "epoch": 1462, "lr": 5.3384231915584416e-05}
{"train_loss": 0.1430748850107193, "global_step": 130133, "epoch": 1462, "lr": 5.3383653508225616e-05}
{"train_loss": 0.15074169635772705, "global_step": 130134, "epoch": 1462, "lr": 5.3383075100411926e-05}
{"train_loss": 0.11875174194574356, "global_step": 130135, "epoch": 1462, "lr": 5.3382496692143434e-05}
{"train_loss": 0.11837347596883774, "global_step": 130136, "epoch": 1462, "lr": 5.338191828342019e-05}
{"train_loss": 0.0959760844707489, "global_step": 130137, "epoch": 1462, "lr": 5.338133987424231e-05}
{"train_loss": 0.10198713839054108, "global_step": 130138, "epoch": 1462, "lr": 5.338076146460984e-05}
{"train_loss": 0.10558046400547028, "global_step": 130139, "epoch": 1462, "lr": 5.338018305452286e-05}
{"train_loss": 0.15390931069850922, "global_step": 130140, "epoch": 1462, "lr": 5.337960464398146e-05}
{"train_loss": 0.0979202538728714, "global_step": 130141, "epoch": 1462, "lr": 5.3379026232985705e-05}
{"train_loss": 0.12764407694339752, "global_step": 130142, "epoch": 1462, "lr": 5.3378447821535696e-05}
{"train_loss": 0.16383947432041168, "global_step": 130143, "epoch": 1462, "lr": 5.33778694096315e-05}
{"train_loss": 0.15101489424705505, "global_step": 130144, "epoch": 1462, "lr": 5.3377290997273174e-05}
{"train_loss": 0.10325051844120026, "global_step": 130145, "epoch": 1462, "lr": 5.337671258446082e-05}
{"train_loss": 0.13715510070323944, "global_step": 130146, "epoch": 1462, "lr": 5.3376134171194525e-05}
{"train_loss": 0.12232273072004318, "global_step": 130147, "epoch": 1462, "lr": 5.3375555757474325e-05}
{"train_loss": 0.2393619865179062, "global_step": 130148, "epoch": 1462, "lr": 5.337497734330034e-05}
{"train_loss": 0.15180325508117676, "global_step": 130149, "epoch": 1462, "lr": 5.337439892867262e-05}
{"train_loss": 0.12847624719142914, "global_step": 130150, "epoch": 1462, "lr": 5.3373820513591253e-05}
{"train_loss": 0.15555807948112488, "global_step": 130151, "epoch": 1462, "lr": 5.337324209805632e-05}
{"train_loss": 0.12707571685314178, "global_step": 130152, "epoch": 1462, "lr": 5.337266368206789e-05}
{"train_loss": 0.12186051905155182, "global_step": 130153, "epoch": 1462, "lr": 5.337208526562606e-05}
{"train_loss": 0.08736573904752731, "global_step": 130154, "epoch": 1462, "lr": 5.337150684873088e-05}
{"train_loss": 0.11336615681648254, "global_step": 130155, "epoch": 1462, "lr": 5.337092843138245e-05}
{"train_loss": 0.09880879521369934, "global_step": 130156, "epoch": 1462, "lr": 5.337035001358084e-05}
{"train_loss": 0.08951710909605026, "global_step": 130157, "epoch": 1462, "lr": 5.336977159532612e-05}
{"train_loss": 0.0830143392086029, "global_step": 130158, "epoch": 1462, "lr": 5.336919317661838e-05}
{"train_loss": 0.11076575517654419, "global_step": 130159, "epoch": 1462, "lr": 5.336861475745769e-05}
{"train_loss": 0.0415557362139225, "global_step": 130160, "epoch": 1462, "lr": 5.336803633784413e-05}
{"train_loss": 0.08631180226802826, "global_step": 130161, "epoch": 1462, "lr": 5.3367457917777785e-05}
{"train_loss": 0.12793752551078796, "global_step": 130162, "epoch": 1462, "lr": 5.3366879497258716e-05}
{"train_loss": 0.15363571047782898, "global_step": 130163, "epoch": 1462, "lr": 5.336630107628702e-05}
{"train_loss": 0.12957067787647247, "global_step": 130164, "epoch": 1462, "lr": 5.336572265486277e-05}
{"train_loss": 0.15224683284759521, "global_step": 130165, "epoch": 1462, "lr": 5.336514423298604e-05}
{"train_loss": 0.09505759924650192, "global_step": 130166, "epoch": 1462, "lr": 5.3364565810656905e-05}
{"train_loss": 0.1195225715637207, "global_step": 130167, "epoch": 1462, "lr": 5.336398738787544e-05}
{"train_loss": 0.09037785232067108, "global_step": 130168, "epoch": 1462, "lr": 5.3363408964641724e-05}
{"train_loss": 0.13572412729263306, "global_step": 130169, "epoch": 1462, "lr": 5.3362830540955856e-05}
{"train_loss": 0.0761701911687851, "global_step": 130170, "epoch": 1462, "lr": 5.3362252116817876e-05}
{"train_loss": 0.08490902185440063, "global_step": 130171, "epoch": 1462, "lr": 5.33616736922279e-05}
{"train_loss": 0.14539924263954163, "global_step": 130172, "epoch": 1462, "lr": 5.336109526718598e-05}
{"train_loss": 0.08159218728542328, "global_step": 130173, "epoch": 1462, "lr": 5.33605168416922e-05}
{"train_loss": 0.09291213005781174, "global_step": 130174, "epoch": 1462, "lr": 5.3359938415746635e-05}
{"train_loss": 0.25916340947151184, "global_step": 130175, "epoch": 1462, "lr": 5.335935998934938e-05}
{"train_loss": 0.18283075094223022, "global_step": 130176, "epoch": 1462, "lr": 5.335878156250048e-05}
{"train_loss": 0.11394578963518143, "global_step": 130177, "epoch": 1462, "lr": 5.3358203135200047e-05}
{"train_loss": 0.17072272300720215, "global_step": 130178, "epoch": 1462, "lr": 5.335762470744814e-05}
{"train_loss": 0.11902262270450592, "global_step": 130179, "epoch": 1462, "lr": 5.335704627924484e-05}
{"train_loss": 0.07324406504631042, "global_step": 130180, "epoch": 1462, "lr": 5.335646785059023e-05}
{"train_loss": 0.14023438096046448, "global_step": 130181, "epoch": 1462, "lr": 5.3355889421484386e-05}
{"train_loss": 0.05350352078676224, "global_step": 130182, "epoch": 1462, "lr": 5.3355310991927373e-05}
{"train_loss": 0.08643127977848053, "global_step": 130183, "epoch": 1462, "lr": 5.33547325619193e-05}
{"train_loss": 0.10573005676269531, "global_step": 130184, "epoch": 1462, "lr": 5.33541541314602e-05}
{"train_loss": 0.05510326847434044, "global_step": 130185, "epoch": 1462, "lr": 5.3353575700550195e-05}
{"train_loss": 0.10147295892238617, "global_step": 130186, "epoch": 1462, "lr": 5.335299726918934e-05}
{"train_loss": 0.14395910501480103, "global_step": 130187, "epoch": 1462, "lr": 5.335241883737772e-05}
{"train_loss": 0.08793801069259644, "global_step": 130188, "epoch": 1462, "lr": 5.335184040511538e-05}
{"train_loss": 0.11747957766056061, "global_step": 130189, "epoch": 1462, "lr": 5.3351261972402455e-05}
{"train_loss": 0.16983343660831451, "global_step": 130190, "epoch": 1462, "lr": 5.335068353923898e-05}
{"train_loss": 0.07137876003980637, "global_step": 130191, "epoch": 1462, "lr": 5.3350105105625045e-05}
{"train_loss": 0.12409308552742004, "global_step": 130192, "epoch": 1462, "lr": 5.3349526671560734e-05}
{"train_loss": 0.11511089652776718, "global_step": 130193, "epoch": 1462, "lr": 5.3348948237046126e-05}
{"train_loss": 0.17236866056919098, "global_step": 130194, "epoch": 1462, "lr": 5.3348369802081276e-05}
{"train_loss": 0.05390368402004242, "global_step": 130195, "epoch": 1462, "lr": 5.334779136666629e-05}
{"train_loss": 0.16609513759613037, "global_step": 130196, "epoch": 1462, "lr": 5.334721293080123e-05}
{"train_loss": 0.1199149414896965, "global_step": 130197, "epoch": 1462, "lr": 5.334663449448617e-05}
{"train_loss": 0.08470623195171356, "global_step": 130198, "epoch": 1462, "lr": 5.33460560577212e-05}
{"train_loss": 0.13747085630893707, "global_step": 130199, "epoch": 1462, "lr": 5.334547762050641e-05}
{"train_loss": 0.144566610455513, "global_step": 130200, "epoch": 1462, "lr": 5.334489918284185e-05}
{"train_loss": 0.07882831990718842, "global_step": 130201, "epoch": 1462, "lr": 5.3344320744727615e-05}
{"train_loss": 0.08850867301225662, "global_step": 130202, "epoch": 1462, "lr": 5.334374230616377e-05}
{"train_loss": 0.15059784054756165, "global_step": 130203, "epoch": 1462, "lr": 5.33431638671504e-05}
{"train_loss": 0.07412288337945938, "global_step": 130204, "epoch": 1462, "lr": 5.3342585427687595e-05}
{"train_loss": 0.228801891207695, "global_step": 130205, "epoch": 1462, "lr": 5.334200698777541e-05}
{"train_loss": 0.12084829899367322, "global_step": 130206, "epoch": 1462, "lr": 5.334142854741393e-05, "val_loss": 5.761267185211182}
{"train_loss": 0.1270102858543396, "global_step": 130207, "epoch": 1463, "lr": 5.3340850106603236e-05}
{"train_loss": 0.0868682712316513, "global_step": 130208, "epoch": 1463, "lr": 5.334027166534341e-05}
{"train_loss": 0.10802952945232391, "global_step": 130209, "epoch": 1463, "lr": 5.333969322363452e-05}
{"train_loss": 0.2351411134004593, "global_step": 130210, "epoch": 1463, "lr": 5.333911478147665e-05}
{"train_loss": 0.11180815100669861, "global_step": 130211, "epoch": 1463, "lr": 5.3338536338869886e-05}
{"train_loss": 0.1104486957192421, "global_step": 130212, "epoch": 1463, "lr": 5.3337957895814285e-05}
{"train_loss": 0.08716809004545212, "global_step": 130213, "epoch": 1463, "lr": 5.333737945230994e-05}
{"train_loss": 0.20684337615966797, "global_step": 130214, "epoch": 1463, "lr": 5.3336801008356905e-05}
{"train_loss": 0.11529477685689926, "global_step": 130215, "epoch": 1463, "lr": 5.333622256395531e-05}
{"train_loss": 0.07333649694919586, "global_step": 130216, "epoch": 1463, "lr": 5.333564411910518e-05}
{"train_loss": 0.1422613561153412, "global_step": 130217, "epoch": 1463, "lr": 5.333506567380663e-05}
{"train_loss": 0.11387065052986145, "global_step": 130218, "epoch": 1463, "lr": 5.3334487228059705e-05}
{"train_loss": 0.14220081269741058, "global_step": 130219, "epoch": 1463, "lr": 5.3333908781864505e-05}
{"train_loss": 0.10227514803409576, "global_step": 130220, "epoch": 1463, "lr": 5.33333303352211e-05}
{"train_loss": 0.16676311194896698, "global_step": 130221, "epoch": 1463, "lr": 5.3332751888129574e-05}
{"train_loss": 0.11250057816505432, "global_step": 130222, "epoch": 1463, "lr": 5.3332173440590005e-05}
{"train_loss": 0.08889402449131012, "global_step": 130223, "epoch": 1463, "lr": 5.3331594992602464e-05}
{"train_loss": 0.09232448786497116, "global_step": 130224, "epoch": 1463, "lr": 5.3331016544167025e-05}
{"train_loss": 0.10279949009418488, "global_step": 130225, "epoch": 1463, "lr": 5.333043809528378e-05}
{"train_loss": 0.10913599282503128, "global_step": 130226, "epoch": 1463, "lr": 5.332985964595279e-05}
{"train_loss": 0.12304532527923584, "global_step": 130227, "epoch": 1463, "lr": 5.3329281196174144e-05}
{"train_loss": 0.1057831346988678, "global_step": 130228, "epoch": 1463, "lr": 5.332870274594791e-05}
{"train_loss": 0.08682786673307419, "global_step": 130229, "epoch": 1463, "lr": 5.3328124295274175e-05}
{"train_loss": 0.11260596662759781, "global_step": 130230, "epoch": 1463, "lr": 5.332754584415301e-05}
{"train_loss": 0.10843814164400101, "global_step": 130231, "epoch": 1463, "lr": 5.332696739258451e-05}
{"train_loss": 0.1935662478208542, "global_step": 130232, "epoch": 1463, "lr": 5.3326388940568714e-05}
{"train_loss": 0.10897956043481827, "global_step": 130233, "epoch": 1463, "lr": 5.3325810488105756e-05}
{"train_loss": 0.06943706423044205, "global_step": 130234, "epoch": 1463, "lr": 5.332523203519566e-05}
{"train_loss": 0.09703686088323593, "global_step": 130235, "epoch": 1463, "lr": 5.3324653581838545e-05}
{"train_loss": 0.13567203283309937, "global_step": 130236, "epoch": 1463, "lr": 5.332407512803447e-05}
{"train_loss": 0.09930672496557236, "global_step": 130237, "epoch": 1463, "lr": 5.33234966737835e-05}
{"train_loss": 0.11618608981370926, "global_step": 130238, "epoch": 1463, "lr": 5.3322918219085736e-05}
{"train_loss": 0.18097923696041107, "global_step": 130239, "epoch": 1463, "lr": 5.332233976394124e-05}
{"train_loss": 0.1022595539689064, "global_step": 130240, "epoch": 1463, "lr": 5.33217613083501e-05}
{"train_loss": 0.1377858817577362, "global_step": 130241, "epoch": 1463, "lr": 5.332118285231239e-05}
{"train_loss": 0.09351775795221329, "global_step": 130242, "epoch": 1463, "lr": 5.332060439582819e-05}
{"train_loss": 0.10328852385282516, "global_step": 130243, "epoch": 1463, "lr": 5.3320025938897587e-05}
{"train_loss": 0.16068853437900543, "global_step": 130244, "epoch": 1463, "lr": 5.331944748152064e-05}
{"train_loss": 0.14532630145549774, "global_step": 130245, "epoch": 1463, "lr": 5.331886902369742e-05}
{"train_loss": 0.10512354969978333, "global_step": 130246, "epoch": 1463, "lr": 5.331829056542802e-05}
{"train_loss": 0.0879671722650528, "global_step": 130247, "epoch": 1463, "lr": 5.331771210671252e-05}
{"train_loss": 0.10121095180511475, "global_step": 130248, "epoch": 1463, "lr": 5.3317133647551e-05}
{"train_loss": 0.14480826258659363, "global_step": 130249, "epoch": 1463, "lr": 5.331655518794353e-05}
{"train_loss": 0.18279124796390533, "global_step": 130250, "epoch": 1463, "lr": 5.331597672789017e-05}
{"train_loss": 0.14528068900108337, "global_step": 130251, "epoch": 1463, "lr": 5.331539826739104e-05}
{"train_loss": 0.16768646240234375, "global_step": 130252, "epoch": 1463, "lr": 5.331481980644618e-05}
{"train_loss": 0.1896255910396576, "global_step": 130253, "epoch": 1463, "lr": 5.3314241345055694e-05}
{"train_loss": 0.10026320815086365, "global_step": 130254, "epoch": 1463, "lr": 5.331366288321965e-05}
{"train_loss": 0.14613662660121918, "global_step": 130255, "epoch": 1463, "lr": 5.331308442093812e-05}
{"train_loss": 0.17897622287273407, "global_step": 130256, "epoch": 1463, "lr": 5.331250595821119e-05}
{"train_loss": 0.12631480395793915, "global_step": 130257, "epoch": 1463, "lr": 5.331192749503893e-05}
{"train_loss": 0.12324794381856918, "global_step": 130258, "epoch": 1463, "lr": 5.331134903142142e-05}
{"train_loss": 0.10582146793603897, "global_step": 130259, "epoch": 1463, "lr": 5.331077056735875e-05}
{"train_loss": 0.13501568138599396, "global_step": 130260, "epoch": 1463, "lr": 5.331019210285098e-05}
{"train_loss": 0.12101822346448898, "global_step": 130261, "epoch": 1463, "lr": 5.330961363789819e-05}
{"train_loss": 0.12012388557195663, "global_step": 130262, "epoch": 1463, "lr": 5.330903517250048e-05}
{"train_loss": 0.10820689052343369, "global_step": 130263, "epoch": 1463, "lr": 5.3308456706657896e-05}
{"train_loss": 0.06937622278928757, "global_step": 130264, "epoch": 1463, "lr": 5.3307878240370534e-05}
{"train_loss": 0.14119523763656616, "global_step": 130265, "epoch": 1463, "lr": 5.3307299773638465e-05}
{"train_loss": 0.1989602893590927, "global_step": 130266, "epoch": 1463, "lr": 5.330672130646177e-05}
{"train_loss": 0.13497237861156464, "global_step": 130267, "epoch": 1463, "lr": 5.3306142838840535e-05}
{"train_loss": 0.19663716852664948, "global_step": 130268, "epoch": 1463, "lr": 5.330556437077481e-05}
{"train_loss": 0.14751072227954865, "global_step": 130269, "epoch": 1463, "lr": 5.3304985902264704e-05}
{"train_loss": 0.06329500675201416, "global_step": 130270, "epoch": 1463, "lr": 5.330440743331028e-05}
{"train_loss": 0.11200573295354843, "global_step": 130271, "epoch": 1463, "lr": 5.330382896391163e-05}
{"train_loss": 0.09006404876708984, "global_step": 130272, "epoch": 1463, "lr": 5.33032504940688e-05}
{"train_loss": 0.10092321038246155, "global_step": 130273, "epoch": 1463, "lr": 5.330267202378191e-05}
{"train_loss": 0.16928933560848236, "global_step": 130274, "epoch": 1463, "lr": 5.3302093553051e-05}
{"train_loss": 0.15069912374019623, "global_step": 130275, "epoch": 1463, "lr": 5.3301515081876175e-05}
{"train_loss": 0.19094975292682648, "global_step": 130276, "epoch": 1463, "lr": 5.33009366102575e-05}
{"train_loss": 0.07414761185646057, "global_step": 130277, "epoch": 1463, "lr": 5.330035813819505e-05}
{"train_loss": 0.08294102549552917, "global_step": 130278, "epoch": 1463, "lr": 5.329977966568891e-05}
{"train_loss": 0.12017544358968735, "global_step": 130279, "epoch": 1463, "lr": 5.3299201192739144e-05}
{"train_loss": 0.18081578612327576, "global_step": 130280, "epoch": 1463, "lr": 5.329862271934586e-05}
{"train_loss": 0.11391696333885193, "global_step": 130281, "epoch": 1463, "lr": 5.3298044245509116e-05}
{"train_loss": 0.0841546580195427, "global_step": 130282, "epoch": 1463, "lr": 5.329746577122898e-05}
{"train_loss": 0.1316031962633133, "global_step": 130283, "epoch": 1463, "lr": 5.3296887296505536e-05}
{"train_loss": 0.16628795862197876, "global_step": 130284, "epoch": 1463, "lr": 5.329630882133887e-05}
{"train_loss": 0.1322260946035385, "global_step": 130285, "epoch": 1463, "lr": 5.3295730345729055e-05}
{"train_loss": 0.07049437612295151, "global_step": 130286, "epoch": 1463, "lr": 5.3295151869676174e-05}
{"train_loss": 0.12573237717151642, "global_step": 130287, "epoch": 1463, "lr": 5.329457339318028e-05}
{"train_loss": 0.16076195240020752, "global_step": 130288, "epoch": 1463, "lr": 5.329399491624149e-05}
{"train_loss": 0.19536620378494263, "global_step": 130289, "epoch": 1463, "lr": 5.3293416438859856e-05}
{"train_loss": 0.14026887714862823, "global_step": 130290, "epoch": 1463, "lr": 5.329283796103547e-05}
{"train_loss": 0.0848107784986496, "global_step": 130291, "epoch": 1463, "lr": 5.329225948276839e-05}
{"train_loss": 0.11061766743659973, "global_step": 130292, "epoch": 1463, "lr": 5.329168100405872e-05}
{"train_loss": 0.11238834261894226, "global_step": 130293, "epoch": 1463, "lr": 5.329110252490652e-05}
{"train_loss": 0.10490551590919495, "global_step": 130294, "epoch": 1463, "lr": 5.3290524045311865e-05}
{"train_loss": 0.1257407759012801, "global_step": 130295, "epoch": 1463, "lr": 5.3289945565274846e-05, "val_loss": 5.853541851043701}
{"train_loss": 0.12178322672843933, "global_step": 130296, "epoch": 1464, "lr": 5.328936708479553e-05}
{"train_loss": 0.12058443576097488, "global_step": 130297, "epoch": 1464, "lr": 5.3288788603874e-05}
{"train_loss": 0.13346394896507263, "global_step": 130298, "epoch": 1464, "lr": 5.328821012251033e-05}
{"train_loss": 0.1453019678592682, "global_step": 130299, "epoch": 1464, "lr": 5.328763164070459e-05}
{"train_loss": 0.20740476250648499, "global_step": 130300, "epoch": 1464, "lr": 5.328705315845689e-05}
{"train_loss": 0.07997004687786102, "global_step": 130301, "epoch": 1464, "lr": 5.3286474675767284e-05}
{"train_loss": 0.20554494857788086, "global_step": 130302, "epoch": 1464, "lr": 5.3285896192635845e-05}
{"train_loss": 0.13648536801338196, "global_step": 130303, "epoch": 1464, "lr": 5.3285317709062653e-05}
{"train_loss": 0.18654556572437286, "global_step": 130304, "epoch": 1464, "lr": 5.32847392250478e-05}
{"train_loss": 0.08847028017044067, "global_step": 130305, "epoch": 1464, "lr": 5.328416074059133e-05}
{"train_loss": 0.11912216991186142, "global_step": 130306, "epoch": 1464, "lr": 5.3283582255693365e-05}
{"train_loss": 0.1492466777563095, "global_step": 130307, "epoch": 1464, "lr": 5.3283003770353954e-05}
{"train_loss": 0.129203662276268, "global_step": 130308, "epoch": 1464, "lr": 5.328242528457319e-05}
{"train_loss": 0.09639334678649902, "global_step": 130309, "epoch": 1464, "lr": 5.3281846798351146e-05}
{"train_loss": 0.054739486426115036, "global_step": 130310, "epoch": 1464, "lr": 5.3281268311687884e-05}
{"train_loss": 0.07706457376480103, "global_step": 130311, "epoch": 1464, "lr": 5.328068982458351e-05}
{"train_loss": 0.08603034168481827, "global_step": 130312, "epoch": 1464, "lr": 5.328011133703808e-05}
{"train_loss": 0.15029320120811462, "global_step": 130313, "epoch": 1464, "lr": 5.3279532849051675e-05}
{"train_loss": 0.11691345274448395, "global_step": 130314, "epoch": 1464, "lr": 5.327895436062439e-05}
{"train_loss": 0.17270727455615997, "global_step": 130315, "epoch": 1464, "lr": 5.327837587175628e-05}
{"train_loss": 0.1369209587574005, "global_step": 130316, "epoch": 1464, "lr": 5.3277797382447436e-05}
{"train_loss": 0.09054401516914368, "global_step": 130317, "epoch": 1464, "lr": 5.3277218892697925e-05}
{"train_loss": 0.1650109440088272, "global_step": 130318, "epoch": 1464, "lr": 5.327664040250784e-05}
{"train_loss": 0.07404161244630814, "global_step": 130319, "epoch": 1464, "lr": 5.327606191187725e-05}
{"train_loss": 0.12136293947696686, "global_step": 130320, "epoch": 1464, "lr": 5.327548342080624e-05}
{"train_loss": 0.14282578229904175, "global_step": 130321, "epoch": 1464, "lr": 5.327490492929488e-05}
{"train_loss": 0.1319783329963684, "global_step": 130322, "epoch": 1464, "lr": 5.327432643734323e-05}
{"train_loss": 0.14750848710536957, "global_step": 130323, "epoch": 1464, "lr": 5.327374794495139e-05}
{"train_loss": 0.11065865308046341, "global_step": 130324, "epoch": 1464, "lr": 5.327316945211944e-05}
{"train_loss": 0.1616354137659073, "global_step": 130325, "epoch": 1464, "lr": 5.327259095884746e-05}
{"train_loss": 0.1376609057188034, "global_step": 130326, "epoch": 1464, "lr": 5.327201246513551e-05}
{"train_loss": 0.12641368806362152, "global_step": 130327, "epoch": 1464, "lr": 5.3271433970983686e-05}
{"train_loss": 0.10932326316833496, "global_step": 130328, "epoch": 1464, "lr": 5.3270855476392053e-05}
{"train_loss": 0.13272127509117126, "global_step": 130329, "epoch": 1464, "lr": 5.32702769813607e-05}
{"train_loss": 0.18061770498752594, "global_step": 130330, "epoch": 1464, "lr": 5.3269698485889687e-05}
{"train_loss": 0.16302654147148132, "global_step": 130331, "epoch": 1464, "lr": 5.326911998997911e-05}
{"train_loss": 0.10387920588254929, "global_step": 130332, "epoch": 1464, "lr": 5.326854149362904e-05}
{"train_loss": 0.15273220837116241, "global_step": 130333, "epoch": 1464, "lr": 5.326796299683955e-05}
{"train_loss": 0.10961664468050003, "global_step": 130334, "epoch": 1464, "lr": 5.326738449961073e-05}
{"train_loss": 0.11027562618255615, "global_step": 130335, "epoch": 1464, "lr": 5.3266806001942646e-05}
{"train_loss": 0.2419661432504654, "global_step": 130336, "epoch": 1464, "lr": 5.326622750383537e-05}
{"train_loss": 0.13373088836669922, "global_step": 130337, "epoch": 1464, "lr": 5.326564900528901e-05}
{"train_loss": 0.13268066942691803, "global_step": 130338, "epoch": 1464, "lr": 5.326507050630361e-05}
{"train_loss": 0.1775907725095749, "global_step": 130339, "epoch": 1464, "lr": 5.326449200687927e-05}
{"train_loss": 0.08331063389778137, "global_step": 130340, "epoch": 1464, "lr": 5.326391350701605e-05}
{"train_loss": 0.09152872115373611, "global_step": 130341, "epoch": 1464, "lr": 5.3263335006714035e-05}
{"train_loss": 0.122273750603199, "global_step": 130342, "epoch": 1464, "lr": 5.32627565059733e-05}
{"train_loss": 0.11316821724176407, "global_step": 130343, "epoch": 1464, "lr": 5.326217800479394e-05}
{"train_loss": 0.1623181849718094, "global_step": 130344, "epoch": 1464, "lr": 5.326159950317601e-05}
{"train_loss": 0.14295634627342224, "global_step": 130345, "epoch": 1464, "lr": 5.32610210011196e-05}
{"train_loss": 0.15475495159626007, "global_step": 130346, "epoch": 1464, "lr": 5.3260442498624787e-05}
{"train_loss": 0.15527983009815216, "global_step": 130347, "epoch": 1464, "lr": 5.325986399569165e-05}
{"train_loss": 0.1740666925907135, "global_step": 130348, "epoch": 1464, "lr": 5.325928549232027e-05}
{"train_loss": 0.18825699388980865, "global_step": 130349, "epoch": 1464, "lr": 5.32587069885107e-05}
{"train_loss": 0.11993665248155594, "global_step": 130350, "epoch": 1464, "lr": 5.325812848426305e-05}
{"train_loss": 0.08383545279502869, "global_step": 130351, "epoch": 1464, "lr": 5.325754997957738e-05}
{"train_loss": 0.07652298361063004, "global_step": 130352, "epoch": 1464, "lr": 5.325697147445378e-05}
{"train_loss": 0.09542999416589737, "global_step": 130353, "epoch": 1464, "lr": 5.325639296889231e-05}
{"train_loss": 0.15955698490142822, "global_step": 130354, "epoch": 1464, "lr": 5.325581446289306e-05}
{"train_loss": 0.1561281532049179, "global_step": 130355, "epoch": 1464, "lr": 5.3255235956456106e-05}
{"train_loss": 0.07781288772821426, "global_step": 130356, "epoch": 1464, "lr": 5.3254657449581527e-05}
{"train_loss": 0.11248661577701569, "global_step": 130357, "epoch": 1464, "lr": 5.3254078942269404e-05}
{"train_loss": 0.11761009693145752, "global_step": 130358, "epoch": 1464, "lr": 5.32535004345198e-05}
{"train_loss": 0.20592144131660461, "global_step": 130359, "epoch": 1464, "lr": 5.325292192633281e-05}
{"train_loss": 0.15747249126434326, "global_step": 130360, "epoch": 1464, "lr": 5.3252343417708493e-05}
{"train_loss": 0.13712453842163086, "global_step": 130361, "epoch": 1464, "lr": 5.325176490864695e-05}
{"train_loss": 0.1465606540441513, "global_step": 130362, "epoch": 1464, "lr": 5.325118639914825e-05}
{"train_loss": 0.09959142655134201, "global_step": 130363, "epoch": 1464, "lr": 5.325060788921246e-05}
{"train_loss": 0.15444163978099823, "global_step": 130364, "epoch": 1464, "lr": 5.325002937883966e-05}
{"train_loss": 0.15149028599262238, "global_step": 130365, "epoch": 1464, "lr": 5.324945086802994e-05}
{"train_loss": 0.11229310929775238, "global_step": 130366, "epoch": 1464, "lr": 5.3248872356783363e-05}
{"train_loss": 0.14719130098819733, "global_step": 130367, "epoch": 1464, "lr": 5.324829384510003e-05}
{"train_loss": 0.16376399993896484, "global_step": 130368, "epoch": 1464, "lr": 5.3247715332979984e-05}
{"train_loss": 0.07581715285778046, "global_step": 130369, "epoch": 1464, "lr": 5.324713682042334e-05}
{"train_loss": 0.13752569258213043, "global_step": 130370, "epoch": 1464, "lr": 5.324655830743015e-05}
{"train_loss": 0.12456761300563812, "global_step": 130371, "epoch": 1464, "lr": 5.32459797940005e-05}
{"train_loss": 0.08354490995407104, "global_step": 130372, "epoch": 1464, "lr": 5.324540128013446e-05}
{"train_loss": 0.06719540059566498, "global_step": 130373, "epoch": 1464, "lr": 5.3244822765832136e-05}
{"train_loss": 0.10483670979738235, "global_step": 130374, "epoch": 1464, "lr": 5.324424425109357e-05}
{"train_loss": 0.17095687985420227, "global_step": 130375, "epoch": 1464, "lr": 5.324366573591886e-05}
{"train_loss": 0.09972695261240005, "global_step": 130376, "epoch": 1464, "lr": 5.3243087220308075e-05}
{"train_loss": 0.1215662807226181, "global_step": 130377, "epoch": 1464, "lr": 5.3242508704261294e-05}
{"train_loss": 0.16232194006443024, "global_step": 130378, "epoch": 1464, "lr": 5.3241930187778597e-05}
{"train_loss": 0.12480414658784866, "global_step": 130379, "epoch": 1464, "lr": 5.3241351670860076e-05}
{"train_loss": 0.2122936099767685, "global_step": 130380, "epoch": 1464, "lr": 5.324077315350579e-05}
{"train_loss": 0.18247012794017792, "global_step": 130381, "epoch": 1464, "lr": 5.3240194635715814e-05}
{"train_loss": 0.23440946638584137, "global_step": 130382, "epoch": 1464, "lr": 5.3239616117490234e-05}
{"train_loss": 0.10902851819992065, "global_step": 130383, "epoch": 1464, "lr": 5.323903759882912e-05}
{"train_loss": 0.1333654833894767, "global_step": 130384, "epoch": 1464, "lr": 5.323845907973257e-05, "val_loss": 5.569900989532471}
{"train_loss": 0.07744856923818588, "global_step": 130385, "epoch": 1465, "lr": 5.323788056020065e-05}
{"train_loss": 0.10495982319116592, "global_step": 130386, "epoch": 1465, "lr": 5.3237302040233426e-05}
{"train_loss": 0.12786968052387238, "global_step": 130387, "epoch": 1465, "lr": 5.323672351983099e-05}
{"train_loss": 0.10701387375593185, "global_step": 130388, "epoch": 1465, "lr": 5.323614499899342e-05}
{"train_loss": 0.13864965736865997, "global_step": 130389, "epoch": 1465, "lr": 5.323556647772079e-05}
{"train_loss": 0.1632823497056961, "global_step": 130390, "epoch": 1465, "lr": 5.3234987956013174e-05}
{"train_loss": 0.1336497664451599, "global_step": 130391, "epoch": 1465, "lr": 5.323440943387065e-05}
{"train_loss": 0.12337109446525574, "global_step": 130392, "epoch": 1465, "lr": 5.3233830911293306e-05}
{"train_loss": 0.1586771309375763, "global_step": 130393, "epoch": 1465, "lr": 5.323325238828121e-05}
{"train_loss": 0.09598487615585327, "global_step": 130394, "epoch": 1465, "lr": 5.323267386483444e-05}
{"train_loss": 0.1546163409948349, "global_step": 130395, "epoch": 1465, "lr": 5.323209534095307e-05}
{"train_loss": 0.12011580914258957, "global_step": 130396, "epoch": 1465, "lr": 5.3231516816637194e-05}
{"train_loss": 0.11921367049217224, "global_step": 130397, "epoch": 1465, "lr": 5.323093829188688e-05}
{"train_loss": 0.1446511149406433, "global_step": 130398, "epoch": 1465, "lr": 5.323035976670221e-05}
{"train_loss": 0.1484428197145462, "global_step": 130399, "epoch": 1465, "lr": 5.322978124108324e-05}
{"train_loss": 0.10740330815315247, "global_step": 130400, "epoch": 1465, "lr": 5.322920271503008e-05}
{"train_loss": 0.15670734643936157, "global_step": 130401, "epoch": 1465, "lr": 5.32286241885428e-05}
{"train_loss": 0.19184449315071106, "global_step": 130402, "epoch": 1465, "lr": 5.322804566162145e-05}
{"train_loss": 0.14929260313510895, "global_step": 130403, "epoch": 1465, "lr": 5.322746713426614e-05}
{"train_loss": 0.15816371142864227, "global_step": 130404, "epoch": 1465, "lr": 5.322688860647694e-05}
{"train_loss": 0.18607397377490997, "global_step": 130405, "epoch": 1465, "lr": 5.322631007825393e-05}
{"train_loss": 0.11648624390363693, "global_step": 130406, "epoch": 1465, "lr": 5.322573154959717e-05}
{"train_loss": 0.1814889758825302, "global_step": 130407, "epoch": 1465, "lr": 5.322515302050675e-05}
{"train_loss": 0.1428011953830719, "global_step": 130408, "epoch": 1465, "lr": 5.322457449098275e-05}
{"train_loss": 0.13211002945899963, "global_step": 130409, "epoch": 1465, "lr": 5.322399596102524e-05}
{"train_loss": 0.14397235214710236, "global_step": 130410, "epoch": 1465, "lr": 5.3223417430634304e-05}
{"train_loss": 0.12865878641605377, "global_step": 130411, "epoch": 1465, "lr": 5.3222838899810024e-05}
{"train_loss": 0.16723932325839996, "global_step": 130412, "epoch": 1465, "lr": 5.322226036855247e-05}
{"train_loss": 0.14733563363552094, "global_step": 130413, "epoch": 1465, "lr": 5.322168183686173e-05}
{"train_loss": 0.23413845896720886, "global_step": 130414, "epoch": 1465, "lr": 5.322110330473786e-05}
{"train_loss": 0.22982148826122284, "global_step": 130415, "epoch": 1465, "lr": 5.3220524772180966e-05}
{"train_loss": 0.10892007499933243, "global_step": 130416, "epoch": 1465, "lr": 5.32199462391911e-05}
{"train_loss": 0.1642475575208664, "global_step": 130417, "epoch": 1465, "lr": 5.321936770576836e-05}
{"train_loss": 0.07059382647275925, "global_step": 130418, "epoch": 1465, "lr": 5.321878917191281e-05}
{"train_loss": 0.06139117106795311, "global_step": 130419, "epoch": 1465, "lr": 5.3218210637624535e-05}
{"train_loss": 0.13128407299518585, "global_step": 130420, "epoch": 1465, "lr": 5.321763210290361e-05}
{"train_loss": 0.11193706095218658, "global_step": 130421, "epoch": 1465, "lr": 5.321705356775012e-05}
{"train_loss": 0.11865280568599701, "global_step": 130422, "epoch": 1465, "lr": 5.3216475032164124e-05}
{"train_loss": 0.16233648359775543, "global_step": 130423, "epoch": 1465, "lr": 5.321589649614572e-05}
{"train_loss": 0.10592926293611526, "global_step": 130424, "epoch": 1465, "lr": 5.3215317959694975e-05}
{"train_loss": 0.09744156897068024, "global_step": 130425, "epoch": 1465, "lr": 5.321473942281198e-05}
{"train_loss": 0.1544009894132614, "global_step": 130426, "epoch": 1465, "lr": 5.321416088549679e-05}
{"train_loss": 0.14855340123176575, "global_step": 130427, "epoch": 1465, "lr": 5.3213582347749504e-05}
{"train_loss": 0.09434983879327774, "global_step": 130428, "epoch": 1465, "lr": 5.3213003809570186e-05}
{"train_loss": 0.19380897283554077, "global_step": 130429, "epoch": 1465, "lr": 5.321242527095891e-05}
{"train_loss": 0.13580620288848877, "global_step": 130430, "epoch": 1465, "lr": 5.321184673191577e-05}
{"train_loss": 0.19107437133789062, "global_step": 130431, "epoch": 1465, "lr": 5.3211268192440847e-05}
{"train_loss": 0.15414033830165863, "global_step": 130432, "epoch": 1465, "lr": 5.32106896525342e-05}
{"train_loss": 0.16577696800231934, "global_step": 130433, "epoch": 1465, "lr": 5.32101111121959e-05}
{"train_loss": 0.06321166455745697, "global_step": 130434, "epoch": 1465, "lr": 5.320953257142606e-05}
{"train_loss": 0.07623451203107834, "global_step": 130435, "epoch": 1465, "lr": 5.320895403022474e-05}
{"train_loss": 0.1244458556175232, "global_step": 130436, "epoch": 1465, "lr": 5.3208375488592e-05}
{"train_loss": 0.13007670640945435, "global_step": 130437, "epoch": 1465, "lr": 5.3207796946527936e-05}
{"train_loss": 0.10877789556980133, "global_step": 130438, "epoch": 1465, "lr": 5.320721840403263e-05}
{"train_loss": 0.08363890647888184, "global_step": 130439, "epoch": 1465, "lr": 5.320663986110614e-05}
{"train_loss": 0.20044131577014923, "global_step": 130440, "epoch": 1465, "lr": 5.3206061317748566e-05}
{"train_loss": 0.11768467724323273, "global_step": 130441, "epoch": 1465, "lr": 5.320548277395998e-05}
{"train_loss": 0.11471674591302872, "global_step": 130442, "epoch": 1465, "lr": 5.320490422974045e-05}
{"train_loss": 0.16247445344924927, "global_step": 130443, "epoch": 1465, "lr": 5.320432568509006e-05}
{"train_loss": 0.11803274601697922, "global_step": 130444, "epoch": 1465, "lr": 5.320374714000888e-05}
{"train_loss": 0.1273447722196579, "global_step": 130445, "epoch": 1465, "lr": 5.320316859449701e-05}
{"train_loss": 0.1320270299911499, "global_step": 130446, "epoch": 1465, "lr": 5.3202590048554514e-05}
{"train_loss": 0.12693147361278534, "global_step": 130447, "epoch": 1465, "lr": 5.320201150218146e-05}
{"train_loss": 0.07675144821405411, "global_step": 130448, "epoch": 1465, "lr": 5.320143295537794e-05}
{"train_loss": 0.1000082939863205, "global_step": 130449, "epoch": 1465, "lr": 5.320085440814402e-05}
{"train_loss": 0.1401800811290741, "global_step": 130450, "epoch": 1465, "lr": 5.320027586047979e-05}
{"train_loss": 0.09597189724445343, "global_step": 130451, "epoch": 1465, "lr": 5.319969731238531e-05}
{"train_loss": 0.17290055751800537, "global_step": 130452, "epoch": 1465, "lr": 5.319911876386068e-05}
{"train_loss": 0.14750178158283234, "global_step": 130453, "epoch": 1465, "lr": 5.3198540214905965e-05}
{"train_loss": 0.1374261975288391, "global_step": 130454, "epoch": 1465, "lr": 5.319796166552125e-05}
{"train_loss": 0.17113438248634338, "global_step": 130455, "epoch": 1465, "lr": 5.31973831157066e-05}
{"train_loss": 0.18096095323562622, "global_step": 130456, "epoch": 1465, "lr": 5.31968045654621e-05}
{"train_loss": 0.11452115327119827, "global_step": 130457, "epoch": 1465, "lr": 5.319622601478784e-05}
{"train_loss": 0.17656183242797852, "global_step": 130458, "epoch": 1465, "lr": 5.319564746368387e-05}
{"train_loss": 0.10656197369098663, "global_step": 130459, "epoch": 1465, "lr": 5.31950689121503e-05}
{"train_loss": 0.1089252382516861, "global_step": 130460, "epoch": 1465, "lr": 5.319449036018719e-05}
{"train_loss": 0.0901169553399086, "global_step": 130461, "epoch": 1465, "lr": 5.31939118077946e-05}
{"train_loss": 0.1630243957042694, "global_step": 130462, "epoch": 1465, "lr": 5.319333325497265e-05}
{"train_loss": 0.13969741761684418, "global_step": 130463, "epoch": 1465, "lr": 5.319275470172139e-05}
{"train_loss": 0.08000563830137253, "global_step": 130464, "epoch": 1465, "lr": 5.3192176148040905e-05}
{"train_loss": 0.14125864207744598, "global_step": 130465, "epoch": 1465, "lr": 5.319159759393126e-05}
{"train_loss": 0.17261086404323578, "global_step": 130466, "epoch": 1465, "lr": 5.319101903939255e-05}
{"train_loss": 0.13871578872203827, "global_step": 130467, "epoch": 1465, "lr": 5.319044048442485e-05}
{"train_loss": 0.13316409289836884, "global_step": 130468, "epoch": 1465, "lr": 5.318986192902822e-05}
{"train_loss": 0.058510616421699524, "global_step": 130469, "epoch": 1465, "lr": 5.318928337320278e-05}
{"train_loss": 0.1108119785785675, "global_step": 130470, "epoch": 1465, "lr": 5.318870481694855e-05}
{"train_loss": 0.1521463543176651, "global_step": 130471, "epoch": 1465, "lr": 5.3188126260265656e-05}
{"train_loss": 0.12100718170404434, "global_step": 130472, "epoch": 1465, "lr": 5.318754770315415e-05}
{"train_loss": 0.1338454545632507, "global_step": 130473, "epoch": 1465, "lr": 5.318696914561412e-05, "val_loss": 5.5509514808654785, "train_action_mse_error": 32.11616134643555}
{"train_loss": 0.1892460733652115, "global_step": 130474, "epoch": 1466, "lr": 5.3186390587645643e-05}
{"train_loss": 0.0776728168129921, "global_step": 130475, "epoch": 1466, "lr": 5.3185812029248794e-05}
{"train_loss": 0.09476098418235779, "global_step": 130476, "epoch": 1466, "lr": 5.3185233470423644e-05}
{"train_loss": 0.10939662158489227, "global_step": 130477, "epoch": 1466, "lr": 5.318465491117028e-05}
{"train_loss": 0.12720023095607758, "global_step": 130478, "epoch": 1466, "lr": 5.318407635148879e-05}
{"train_loss": 0.09565757215023041, "global_step": 130479, "epoch": 1466, "lr": 5.3183497791379225e-05}
{"train_loss": 0.1618337780237198, "global_step": 130480, "epoch": 1466, "lr": 5.3182919230841686e-05}
{"train_loss": 0.17693325877189636, "global_step": 130481, "epoch": 1466, "lr": 5.318234066987624e-05}
{"train_loss": 0.19633321464061737, "global_step": 130482, "epoch": 1466, "lr": 5.3181762108482954e-05}
{"train_loss": 0.11498058587312698, "global_step": 130483, "epoch": 1466, "lr": 5.318118354666194e-05}
{"train_loss": 0.08550717681646347, "global_step": 130484, "epoch": 1466, "lr": 5.318060498441324e-05}
{"train_loss": 0.091984324157238, "global_step": 130485, "epoch": 1466, "lr": 5.318002642173696e-05}
{"train_loss": 0.11742746829986572, "global_step": 130486, "epoch": 1466, "lr": 5.317944785863315e-05}
{"train_loss": 0.08767116814851761, "global_step": 130487, "epoch": 1466, "lr": 5.3178869295101897e-05}
{"train_loss": 0.07310064136981964, "global_step": 130488, "epoch": 1466, "lr": 5.31782907311433e-05}
{"train_loss": 0.15726685523986816, "global_step": 130489, "epoch": 1466, "lr": 5.317771216675742e-05}
{"train_loss": 0.08324155956506729, "global_step": 130490, "epoch": 1466, "lr": 5.317713360194434e-05}
{"train_loss": 0.17114832997322083, "global_step": 130491, "epoch": 1466, "lr": 5.3176555036704125e-05}
{"train_loss": 0.1523219347000122, "global_step": 130492, "epoch": 1466, "lr": 5.317597647103686e-05}
{"train_loss": 0.17580312490463257, "global_step": 130493, "epoch": 1466, "lr": 5.3175397904942616e-05}
{"train_loss": 0.09110385179519653, "global_step": 130494, "epoch": 1466, "lr": 5.317481933842149e-05}
{"train_loss": 0.13356438279151917, "global_step": 130495, "epoch": 1466, "lr": 5.317424077147355e-05}
{"train_loss": 0.13025273382663727, "global_step": 130496, "epoch": 1466, "lr": 5.317366220409886e-05}
{"train_loss": 0.12568745017051697, "global_step": 130497, "epoch": 1466, "lr": 5.3173083636297514e-05}
{"train_loss": 0.10317236185073853, "global_step": 130498, "epoch": 1466, "lr": 5.317250506806959e-05}
{"train_loss": 0.058797821402549744, "global_step": 130499, "epoch": 1466, "lr": 5.317192649941516e-05}
{"train_loss": 0.06669459491968155, "global_step": 130500, "epoch": 1466, "lr": 5.3171347930334306e-05}
{"train_loss": 0.15093883872032166, "global_step": 130501, "epoch": 1466, "lr": 5.31707693608271e-05}
{"train_loss": 0.16055120527744293, "global_step": 130502, "epoch": 1466, "lr": 5.317019079089362e-05}
{"train_loss": 0.1417534053325653, "global_step": 130503, "epoch": 1466, "lr": 5.316961222053395e-05}
{"train_loss": 0.18308241665363312, "global_step": 130504, "epoch": 1466, "lr": 5.316903364974815e-05}
{"train_loss": 0.14271536469459534, "global_step": 130505, "epoch": 1466, "lr": 5.316845507853632e-05}
{"train_loss": 0.17150773108005524, "global_step": 130506, "epoch": 1466, "lr": 5.3167876506898537e-05}
{"train_loss": 0.11784113198518753, "global_step": 130507, "epoch": 1466, "lr": 5.316729793483487e-05}
{"train_loss": 0.07243108749389648, "global_step": 130508, "epoch": 1466, "lr": 5.3166719362345407e-05}
{"train_loss": 0.13157786428928375, "global_step": 130509, "epoch": 1466, "lr": 5.316614078943021e-05}
{"train_loss": 0.17571331560611725, "global_step": 130510, "epoch": 1466, "lr": 5.316556221608936e-05}
{"train_loss": 0.10704295337200165, "global_step": 130511, "epoch": 1466, "lr": 5.316498364232294e-05}
{"train_loss": 0.14525340497493744, "global_step": 130512, "epoch": 1466, "lr": 5.316440506813103e-05}
{"train_loss": 0.16957718133926392, "global_step": 130513, "epoch": 1466, "lr": 5.31638264935137e-05}
{"train_loss": 0.16688960790634155, "global_step": 130514, "epoch": 1466, "lr": 5.316324791847102e-05}
{"train_loss": 0.1322205811738968, "global_step": 130515, "epoch": 1466, "lr": 5.31626693430031e-05}
{"train_loss": 0.0976996049284935, "global_step": 130516, "epoch": 1466, "lr": 5.316209076711e-05}
{"train_loss": 0.1085352972149849, "global_step": 130517, "epoch": 1466, "lr": 5.316151219079178e-05}
{"train_loss": 0.09388735145330429, "global_step": 130518, "epoch": 1466, "lr": 5.3160933614048544e-05}
{"train_loss": 0.08825813233852386, "global_step": 130519, "epoch": 1466, "lr": 5.316035503688035e-05}
{"train_loss": 0.15626782178878784, "global_step": 130520, "epoch": 1466, "lr": 5.315977645928729e-05}
{"train_loss": 0.11964134126901627, "global_step": 130521, "epoch": 1466, "lr": 5.315919788126944e-05}
{"train_loss": 0.07063429802656174, "global_step": 130522, "epoch": 1466, "lr": 5.3158619302826864e-05}
{"train_loss": 0.09143796563148499, "global_step": 130523, "epoch": 1466, "lr": 5.315804072395965e-05}
{"train_loss": 0.12768805027008057, "global_step": 130524, "epoch": 1466, "lr": 5.315746214466788e-05}
{"train_loss": 0.11293397843837738, "global_step": 130525, "epoch": 1466, "lr": 5.315688356495163e-05}
{"train_loss": 0.1152089387178421, "global_step": 130526, "epoch": 1466, "lr": 5.3156304984810976e-05}
{"train_loss": 0.13397978246212006, "global_step": 130527, "epoch": 1466, "lr": 5.315572640424601e-05}
{"train_loss": 0.12595844268798828, "global_step": 130528, "epoch": 1466, "lr": 5.315514782325678e-05}
{"train_loss": 0.11984050273895264, "global_step": 130529, "epoch": 1466, "lr": 5.315456924184338e-05}
{"train_loss": 0.10092388093471527, "global_step": 130530, "epoch": 1466, "lr": 5.315399066000588e-05}
{"train_loss": 0.1599702537059784, "global_step": 130531, "epoch": 1466, "lr": 5.315341207774437e-05}
{"train_loss": 0.1138414740562439, "global_step": 130532, "epoch": 1466, "lr": 5.315283349505892e-05}
{"train_loss": 0.2296973019838333, "global_step": 130533, "epoch": 1466, "lr": 5.315225491194962e-05}
{"train_loss": 0.09010365605354309, "global_step": 130534, "epoch": 1466, "lr": 5.315167632841652e-05}
{"train_loss": 0.12097637355327606, "global_step": 130535, "epoch": 1466, "lr": 5.315109774445973e-05}
{"train_loss": 0.10334967821836472, "global_step": 130536, "epoch": 1466, "lr": 5.3150519160079304e-05}
{"train_loss": 0.12428782135248184, "global_step": 130537, "epoch": 1466, "lr": 5.3149940575275335e-05}
{"train_loss": 0.17348530888557434, "global_step": 130538, "epoch": 1466, "lr": 5.31493619900479e-05}
{"train_loss": 0.11909565329551697, "global_step": 130539, "epoch": 1466, "lr": 5.3148783404397065e-05}
{"train_loss": 0.1604248583316803, "global_step": 130540, "epoch": 1466, "lr": 5.314820481832291e-05}
{"train_loss": 0.07555317133665085, "global_step": 130541, "epoch": 1466, "lr": 5.314762623182551e-05}
{"train_loss": 0.07568689435720444, "global_step": 130542, "epoch": 1466, "lr": 5.314704764490496e-05}
{"train_loss": 0.08910063654184341, "global_step": 130543, "epoch": 1466, "lr": 5.3146469057561336e-05}
{"train_loss": 0.09785395860671997, "global_step": 130544, "epoch": 1466, "lr": 5.31458904697947e-05}
{"train_loss": 0.19685061275959015, "global_step": 130545, "epoch": 1466, "lr": 5.314531188160513e-05}
{"train_loss": 0.10698042809963226, "global_step": 130546, "epoch": 1466, "lr": 5.314473329299272e-05}
{"train_loss": 0.13970808684825897, "global_step": 130547, "epoch": 1466, "lr": 5.314415470395755e-05}
{"train_loss": 0.17581132054328918, "global_step": 130548, "epoch": 1466, "lr": 5.314357611449968e-05}
{"train_loss": 0.14402952790260315, "global_step": 130549, "epoch": 1466, "lr": 5.314299752461919e-05}
{"train_loss": 0.17751702666282654, "global_step": 130550, "epoch": 1466, "lr": 5.3142418934316166e-05}
{"train_loss": 0.08323238790035248, "global_step": 130551, "epoch": 1466, "lr": 5.314184034359068e-05}
{"train_loss": 0.1364862620830536, "global_step": 130552, "epoch": 1466, "lr": 5.3141261752442805e-05}
{"train_loss": 0.04710868000984192, "global_step": 130553, "epoch": 1466, "lr": 5.314068316087263e-05}
{"train_loss": 0.0788927972316742, "global_step": 130554, "epoch": 1466, "lr": 5.314010456888024e-05}
{"train_loss": 0.08462900668382645, "global_step": 130555, "epoch": 1466, "lr": 5.313952597646568e-05}
{"train_loss": 0.15993724763393402, "global_step": 130556, "epoch": 1466, "lr": 5.313894738362906e-05}
{"train_loss": 0.196636363863945, "global_step": 130557, "epoch": 1466, "lr": 5.313836879037044e-05}
{"train_loss": 0.13034352660179138, "global_step": 130558, "epoch": 1466, "lr": 5.313779019668992e-05}
{"train_loss": 0.0850212424993515, "global_step": 130559, "epoch": 1466, "lr": 5.313721160258753e-05}
{"train_loss": 0.09153099358081818, "global_step": 130560, "epoch": 1466, "lr": 5.313663300806341e-05}
{"train_loss": 0.09377431124448776, "global_step": 130561, "epoch": 1466, "lr": 5.3136054413117594e-05}
{"train_loss": 0.1250991789477595, "global_step": 130562, "epoch": 1466, "lr": 5.313547581775019e-05, "val_loss": 5.747905254364014}
{"train_loss": 0.1839653104543686, "global_step": 130563, "epoch": 1467, "lr": 5.313489722196125e-05}
{"train_loss": 0.106031134724617, "global_step": 130564, "epoch": 1467, "lr": 5.3134318625750855e-05}
{"train_loss": 0.09950372576713562, "global_step": 130565, "epoch": 1467, "lr": 5.313374002911911e-05}
{"train_loss": 0.15785442292690277, "global_step": 130566, "epoch": 1467, "lr": 5.313316143206606e-05}
{"train_loss": 0.08892542868852615, "global_step": 130567, "epoch": 1467, "lr": 5.313258283459179e-05}
{"train_loss": 0.1322976052761078, "global_step": 130568, "epoch": 1467, "lr": 5.313200423669639e-05}
{"train_loss": 0.1379787027835846, "global_step": 130569, "epoch": 1467, "lr": 5.313142563837992e-05}
{"train_loss": 0.12500064074993134, "global_step": 130570, "epoch": 1467, "lr": 5.3130847039642475e-05}
{"train_loss": 0.1159239262342453, "global_step": 130571, "epoch": 1467, "lr": 5.3130268440484123e-05}
{"train_loss": 0.091914601624012, "global_step": 130572, "epoch": 1467, "lr": 5.312968984090494e-05}
{"train_loss": 0.08327215909957886, "global_step": 130573, "epoch": 1467, "lr": 5.3129111240905016e-05}
{"train_loss": 0.11659982055425644, "global_step": 130574, "epoch": 1467, "lr": 5.3128532640484416e-05}
{"train_loss": 0.11617215722799301, "global_step": 130575, "epoch": 1467, "lr": 5.312795403964323e-05}
{"train_loss": 0.09124982357025146, "global_step": 130576, "epoch": 1467, "lr": 5.312737543838152e-05}
{"train_loss": 0.04397134110331535, "global_step": 130577, "epoch": 1467, "lr": 5.3126796836699366e-05}
{"train_loss": 0.16997593641281128, "global_step": 130578, "epoch": 1467, "lr": 5.312621823459687e-05}
{"train_loss": 0.18926940858364105, "global_step": 130579, "epoch": 1467, "lr": 5.3125639632074076e-05}
{"train_loss": 0.13986393809318542, "global_step": 130580, "epoch": 1467, "lr": 5.312506102913109e-05}
{"train_loss": 0.1396177113056183, "global_step": 130581, "epoch": 1467, "lr": 5.3124482425767975e-05}
{"train_loss": 0.13111978769302368, "global_step": 130582, "epoch": 1467, "lr": 5.312390382198481e-05}
{"train_loss": 0.15566222369670868, "global_step": 130583, "epoch": 1467, "lr": 5.312332521778167e-05}
{"train_loss": 0.08953604847192764, "global_step": 130584, "epoch": 1467, "lr": 5.3122746613158644e-05}
{"train_loss": 0.19703468680381775, "global_step": 130585, "epoch": 1467, "lr": 5.3122168008115806e-05}
{"train_loss": 0.12165655940771103, "global_step": 130586, "epoch": 1467, "lr": 5.312158940265324e-05}
{"train_loss": 0.09333523362874985, "global_step": 130587, "epoch": 1467, "lr": 5.3121010796771e-05}
{"train_loss": 0.13322432339191437, "global_step": 130588, "epoch": 1467, "lr": 5.312043219046917e-05}
{"train_loss": 0.11229626089334488, "global_step": 130589, "epoch": 1467, "lr": 5.311985358374785e-05}
{"train_loss": 0.15248635411262512, "global_step": 130590, "epoch": 1467, "lr": 5.3119274976607094e-05}
{"train_loss": 0.13681145012378693, "global_step": 130591, "epoch": 1467, "lr": 5.3118696369046996e-05}
{"train_loss": 0.15827056765556335, "global_step": 130592, "epoch": 1467, "lr": 5.311811776106762e-05}
{"train_loss": 0.09382779896259308, "global_step": 130593, "epoch": 1467, "lr": 5.3117539152669064e-05}
{"train_loss": 0.06673657894134521, "global_step": 130594, "epoch": 1467, "lr": 5.311696054385138e-05}
{"train_loss": 0.1523902863264084, "global_step": 130595, "epoch": 1467, "lr": 5.311638193461464e-05}
{"train_loss": 0.13950851559638977, "global_step": 130596, "epoch": 1467, "lr": 5.311580332495897e-05}
{"train_loss": 0.11984209716320038, "global_step": 130597, "epoch": 1467, "lr": 5.311522471488442e-05}
{"train_loss": 0.1566934436559677, "global_step": 130598, "epoch": 1467, "lr": 5.311464610439105e-05}
{"train_loss": 0.09903370589017868, "global_step": 130599, "epoch": 1467, "lr": 5.311406749347896e-05}
{"train_loss": 0.13749746978282928, "global_step": 130600, "epoch": 1467, "lr": 5.311348888214822e-05}
{"train_loss": 0.1103329062461853, "global_step": 130601, "epoch": 1467, "lr": 5.3112910270398916e-05}
{"train_loss": 0.16674919426441193, "global_step": 130602, "epoch": 1467, "lr": 5.311233165823112e-05}
{"train_loss": 0.11559302359819412, "global_step": 130603, "epoch": 1467, "lr": 5.3111753045644905e-05}
{"train_loss": 0.1374468356370926, "global_step": 130604, "epoch": 1467, "lr": 5.3111174432640356e-05}
{"train_loss": 0.09542086720466614, "global_step": 130605, "epoch": 1467, "lr": 5.311059581921754e-05}
{"train_loss": 0.12659262120723724, "global_step": 130606, "epoch": 1467, "lr": 5.311001720537655e-05}
{"train_loss": 0.16669517755508423, "global_step": 130607, "epoch": 1467, "lr": 5.3109438591117454e-05}
{"train_loss": 0.1514245867729187, "global_step": 130608, "epoch": 1467, "lr": 5.310885997644033e-05}
{"train_loss": 0.09808535128831863, "global_step": 130609, "epoch": 1467, "lr": 5.310828136134526e-05}
{"train_loss": 0.11685330420732498, "global_step": 130610, "epoch": 1467, "lr": 5.310770274583232e-05}
{"train_loss": 0.11812808364629745, "global_step": 130611, "epoch": 1467, "lr": 5.310712412990159e-05}
{"train_loss": 0.13380588591098785, "global_step": 130612, "epoch": 1467, "lr": 5.310654551355313e-05}
{"train_loss": 0.10242853313684464, "global_step": 130613, "epoch": 1467, "lr": 5.310596689678703e-05}
{"train_loss": 0.10716311633586884, "global_step": 130614, "epoch": 1467, "lr": 5.310538827960338e-05}
{"train_loss": 0.10344256460666656, "global_step": 130615, "epoch": 1467, "lr": 5.310480966200225e-05}
{"train_loss": 0.08331327140331268, "global_step": 130616, "epoch": 1467, "lr": 5.310423104398372e-05}
{"train_loss": 0.15726922452449799, "global_step": 130617, "epoch": 1467, "lr": 5.310365242554787e-05}
{"train_loss": 0.09565048664808273, "global_step": 130618, "epoch": 1467, "lr": 5.310307380669476e-05}
{"train_loss": 0.12998515367507935, "global_step": 130619, "epoch": 1467, "lr": 5.3102495187424486e-05}
{"train_loss": 0.11291816085577011, "global_step": 130620, "epoch": 1467, "lr": 5.310191656773711e-05}
{"train_loss": 0.05200977995991707, "global_step": 130621, "epoch": 1467, "lr": 5.310133794763273e-05}
{"train_loss": 0.08999708294868469, "global_step": 130622, "epoch": 1467, "lr": 5.3100759327111405e-05}
{"train_loss": 0.16552940011024475, "global_step": 130623, "epoch": 1467, "lr": 5.310018070617322e-05}
{"train_loss": 0.13538634777069092, "global_step": 130624, "epoch": 1467, "lr": 5.309960208481827e-05}
{"train_loss": 0.15601591765880585, "global_step": 130625, "epoch": 1467, "lr": 5.3099023463046604e-05}
{"train_loss": 0.10077764093875885, "global_step": 130626, "epoch": 1467, "lr": 5.309844484085831e-05}
{"train_loss": 0.11003559082746506, "global_step": 130627, "epoch": 1467, "lr": 5.309786621825347e-05}
{"train_loss": 0.1460280418395996, "global_step": 130628, "epoch": 1467, "lr": 5.309728759523216e-05}
{"train_loss": 0.22192682325839996, "global_step": 130629, "epoch": 1467, "lr": 5.3096708971794454e-05}
{"train_loss": 0.1360630989074707, "global_step": 130630, "epoch": 1467, "lr": 5.3096130347940444e-05}
{"train_loss": 0.1412602812051773, "global_step": 130631, "epoch": 1467, "lr": 5.309555172367017e-05}
{"train_loss": 0.10491754114627838, "global_step": 130632, "epoch": 1467, "lr": 5.309497309898376e-05}
{"train_loss": 0.10426243394613266, "global_step": 130633, "epoch": 1467, "lr": 5.309439447388126e-05}
{"train_loss": 0.13312922418117523, "global_step": 130634, "epoch": 1467, "lr": 5.309381584836276e-05}
{"train_loss": 0.15935386717319489, "global_step": 130635, "epoch": 1467, "lr": 5.309323722242834e-05}
{"train_loss": 0.20409256219863892, "global_step": 130636, "epoch": 1467, "lr": 5.309265859607806e-05}
{"train_loss": 0.10898754745721817, "global_step": 130637, "epoch": 1467, "lr": 5.3092079969312015e-05}
{"train_loss": 0.1367110013961792, "global_step": 130638, "epoch": 1467, "lr": 5.3091501342130276e-05}
{"train_loss": 0.14450427889823914, "global_step": 130639, "epoch": 1467, "lr": 5.309092271453293e-05}
{"train_loss": 0.16004903614521027, "global_step": 130640, "epoch": 1467, "lr": 5.309034408652005e-05}
{"train_loss": 0.1177535131573677, "global_step": 130641, "epoch": 1467, "lr": 5.30897654580917e-05}
{"train_loss": 0.09453967958688736, "global_step": 130642, "epoch": 1467, "lr": 5.3089186829247974e-05}
{"train_loss": 0.13179060816764832, "global_step": 130643, "epoch": 1467, "lr": 5.3088608199988945e-05}
{"train_loss": 0.13474123179912567, "global_step": 130644, "epoch": 1467, "lr": 5.308802957031469e-05}
{"train_loss": 0.1273236721754074, "global_step": 130645, "epoch": 1467, "lr": 5.3087450940225294e-05}
{"train_loss": 0.14439241588115692, "global_step": 130646, "epoch": 1467, "lr": 5.3086872309720814e-05}
{"train_loss": 0.1428365558385849, "global_step": 130647, "epoch": 1467, "lr": 5.308629367880135e-05}
{"train_loss": 0.20736293494701385, "global_step": 130648, "epoch": 1467, "lr": 5.3085715047466954e-05}
{"train_loss": 0.18903891742229462, "global_step": 130649, "epoch": 1467, "lr": 5.308513641571773e-05}
{"train_loss": 0.19278796017169952, "global_step": 130650, "epoch": 1467, "lr": 5.3084557783553756e-05}
{"train_loss": 0.1292743013983362, "global_step": 130651, "epoch": 1467, "lr": 5.3083979150975096e-05, "val_loss": 5.582157611846924}
{"train_loss": 0.1390918791294098, "global_step": 130652, "epoch": 1468, "lr": 5.308340051798183e-05}
{"train_loss": 0.21753264963626862, "global_step": 130653, "epoch": 1468, "lr": 5.308282188457404e-05}
{"train_loss": 0.07853874564170837, "global_step": 130654, "epoch": 1468, "lr": 5.308224325075181e-05}
{"train_loss": 0.17969290912151337, "global_step": 130655, "epoch": 1468, "lr": 5.3081664616515206e-05}
{"train_loss": 0.08568704128265381, "global_step": 130656, "epoch": 1468, "lr": 5.30810859818643e-05}
{"train_loss": 0.08580175787210464, "global_step": 130657, "epoch": 1468, "lr": 5.308050734679919e-05}
{"train_loss": 0.11150021106004715, "global_step": 130658, "epoch": 1468, "lr": 5.3079928711319936e-05}
{"train_loss": 0.17970579862594604, "global_step": 130659, "epoch": 1468, "lr": 5.3079350075426626e-05}
{"train_loss": 0.14158906042575836, "global_step": 130660, "epoch": 1468, "lr": 5.307877143911935e-05}
{"train_loss": 0.16934867203235626, "global_step": 130661, "epoch": 1468, "lr": 5.3078192802398154e-05}
{"train_loss": 0.1755676567554474, "global_step": 130662, "epoch": 1468, "lr": 5.3077614165263134e-05}
{"train_loss": 0.09640486538410187, "global_step": 130663, "epoch": 1468, "lr": 5.307703552771437e-05}
{"train_loss": 0.1091436967253685, "global_step": 130664, "epoch": 1468, "lr": 5.3076456889751935e-05}
{"train_loss": 0.14732995629310608, "global_step": 130665, "epoch": 1468, "lr": 5.307587825137591e-05}
{"train_loss": 0.14652429521083832, "global_step": 130666, "epoch": 1468, "lr": 5.307529961258637e-05}
{"train_loss": 0.1420259028673172, "global_step": 130667, "epoch": 1468, "lr": 5.3074720973383385e-05}
{"train_loss": 0.09399900585412979, "global_step": 130668, "epoch": 1468, "lr": 5.307414233376704e-05}
{"train_loss": 0.07044489681720734, "global_step": 130669, "epoch": 1468, "lr": 5.307356369373742e-05}
{"train_loss": 0.0863557755947113, "global_step": 130670, "epoch": 1468, "lr": 5.30729850532946e-05}
{"train_loss": 0.09021283686161041, "global_step": 130671, "epoch": 1468, "lr": 5.3072406412438645e-05}
{"train_loss": 0.11183101683855057, "global_step": 130672, "epoch": 1468, "lr": 5.307182777116966e-05}
{"train_loss": 0.09280425310134888, "global_step": 130673, "epoch": 1468, "lr": 5.307124912948769e-05}
{"train_loss": 0.12572568655014038, "global_step": 130674, "epoch": 1468, "lr": 5.307067048739284e-05}
{"train_loss": 0.09260614216327667, "global_step": 130675, "epoch": 1468, "lr": 5.3070091844885164e-05}
{"train_loss": 0.11889688670635223, "global_step": 130676, "epoch": 1468, "lr": 5.306951320196476e-05}
{"train_loss": 0.07043302804231644, "global_step": 130677, "epoch": 1468, "lr": 5.306893455863169e-05}
{"train_loss": 0.1484416127204895, "global_step": 130678, "epoch": 1468, "lr": 5.3068355914886045e-05}
{"train_loss": 0.06272726505994797, "global_step": 130679, "epoch": 1468, "lr": 5.306777727072789e-05}
{"train_loss": 0.1748920977115631, "global_step": 130680, "epoch": 1468, "lr": 5.306719862615732e-05}
{"train_loss": 0.12826338410377502, "global_step": 130681, "epoch": 1468, "lr": 5.3066619981174394e-05}
{"train_loss": 0.16003316640853882, "global_step": 130682, "epoch": 1468, "lr": 5.306604133577922e-05}
{"train_loss": 0.10157731920480728, "global_step": 130683, "epoch": 1468, "lr": 5.3065462689971834e-05}
{"train_loss": 0.09540040045976639, "global_step": 130684, "epoch": 1468, "lr": 5.3064884043752325e-05}
{"train_loss": 0.13080118596553802, "global_step": 130685, "epoch": 1468, "lr": 5.306430539712079e-05}
{"train_loss": 0.11671224236488342, "global_step": 130686, "epoch": 1468, "lr": 5.306372675007729e-05}
{"train_loss": 0.1602647602558136, "global_step": 130687, "epoch": 1468, "lr": 5.306314810262192e-05}
{"train_loss": 0.10261185467243195, "global_step": 130688, "epoch": 1468, "lr": 5.306256945475474e-05}
{"train_loss": 0.1564379781484604, "global_step": 130689, "epoch": 1468, "lr": 5.306199080647584e-05}
{"train_loss": 0.10210579633712769, "global_step": 130690, "epoch": 1468, "lr": 5.30614121577853e-05}
{"train_loss": 0.11379637569189072, "global_step": 130691, "epoch": 1468, "lr": 5.306083350868317e-05}
{"train_loss": 0.09157626330852509, "global_step": 130692, "epoch": 1468, "lr": 5.3060254859169566e-05}
{"train_loss": 0.11365492641925812, "global_step": 130693, "epoch": 1468, "lr": 5.305967620924454e-05}
{"train_loss": 0.09130019694566727, "global_step": 130694, "epoch": 1468, "lr": 5.305909755890818e-05}
{"train_loss": 0.16388224065303802, "global_step": 130695, "epoch": 1468, "lr": 5.305851890816056e-05}
{"train_loss": 0.14699915051460266, "global_step": 130696, "epoch": 1468, "lr": 5.305794025700176e-05}
{"train_loss": 0.18871743977069855, "global_step": 130697, "epoch": 1468, "lr": 5.3057361605431854e-05}
{"train_loss": 0.09308072924613953, "global_step": 130698, "epoch": 1468, "lr": 5.3056782953450935e-05}
{"train_loss": 0.12808531522750854, "global_step": 130699, "epoch": 1468, "lr": 5.3056204301059057e-05}
{"train_loss": 0.13719892501831055, "global_step": 130700, "epoch": 1468, "lr": 5.305562564825631e-05}
{"train_loss": 0.10380590707063675, "global_step": 130701, "epoch": 1468, "lr": 5.305504699504278e-05}
{"train_loss": 0.1273484230041504, "global_step": 130702, "epoch": 1468, "lr": 5.305446834141854e-05}
{"train_loss": 0.16588030755519867, "global_step": 130703, "epoch": 1468, "lr": 5.305388968738365e-05}
{"train_loss": 0.09817194193601608, "global_step": 130704, "epoch": 1468, "lr": 5.305331103293819e-05}
{"train_loss": 0.08840471506118774, "global_step": 130705, "epoch": 1468, "lr": 5.305273237808227e-05}
{"train_loss": 0.13957388699054718, "global_step": 130706, "epoch": 1468, "lr": 5.305215372281594e-05}
{"train_loss": 0.09151735156774521, "global_step": 130707, "epoch": 1468, "lr": 5.305157506713929e-05}
{"train_loss": 0.11921902000904083, "global_step": 130708, "epoch": 1468, "lr": 5.305099641105239e-05}
{"train_loss": 0.17058366537094116, "global_step": 130709, "epoch": 1468, "lr": 5.305041775455533e-05}
{"train_loss": 0.13417726755142212, "global_step": 130710, "epoch": 1468, "lr": 5.304983909764817e-05}
{"train_loss": 0.1195240318775177, "global_step": 130711, "epoch": 1468, "lr": 5.3049260440331e-05}
{"train_loss": 0.12335967272520065, "global_step": 130712, "epoch": 1468, "lr": 5.304868178260388e-05}
{"train_loss": 0.1011054664850235, "global_step": 130713, "epoch": 1468, "lr": 5.304810312446692e-05}
{"train_loss": 0.09911297261714935, "global_step": 130714, "epoch": 1468, "lr": 5.304752446592017e-05}
{"train_loss": 0.14095591008663177, "global_step": 130715, "epoch": 1468, "lr": 5.304694580696372e-05}
{"train_loss": 0.0953621193766594, "global_step": 130716, "epoch": 1468, "lr": 5.304636714759764e-05}
{"train_loss": 0.13224802911281586, "global_step": 130717, "epoch": 1468, "lr": 5.304578848782202e-05}
{"train_loss": 0.1520945429801941, "global_step": 130718, "epoch": 1468, "lr": 5.304520982763693e-05}
{"train_loss": 0.1273462474346161, "global_step": 130719, "epoch": 1468, "lr": 5.304463116704245e-05}
{"train_loss": 0.17023760080337524, "global_step": 130720, "epoch": 1468, "lr": 5.3044052506038656e-05}
{"train_loss": 0.24339458346366882, "global_step": 130721, "epoch": 1468, "lr": 5.304347384462563e-05}
{"train_loss": 0.13680718839168549, "global_step": 130722, "epoch": 1468, "lr": 5.304289518280343e-05}
{"train_loss": 0.10182006657123566, "global_step": 130723, "epoch": 1468, "lr": 5.304231652057217e-05}
{"train_loss": 0.11010958254337311, "global_step": 130724, "epoch": 1468, "lr": 5.304173785793188e-05}
{"train_loss": 0.0993661880493164, "global_step": 130725, "epoch": 1468, "lr": 5.3041159194882695e-05}
{"train_loss": 0.1844930797815323, "global_step": 130726, "epoch": 1468, "lr": 5.304058053142464e-05}
{"train_loss": 0.17593784630298615, "global_step": 130727, "epoch": 1468, "lr": 5.304000186755783e-05}
{"train_loss": 0.19093136489391327, "global_step": 130728, "epoch": 1468, "lr": 5.303942320328233e-05}
{"train_loss": 0.15243732929229736, "global_step": 130729, "epoch": 1468, "lr": 5.303884453859822e-05}
{"train_loss": 0.1520368903875351, "global_step": 130730, "epoch": 1468, "lr": 5.303826587350557e-05}
{"train_loss": 0.07855373620986938, "global_step": 130731, "epoch": 1468, "lr": 5.303768720800445e-05}
{"train_loss": 0.14912638068199158, "global_step": 130732, "epoch": 1468, "lr": 5.3037108542094956e-05}
{"train_loss": 0.12297260016202927, "global_step": 130733, "epoch": 1468, "lr": 5.3036529875777165e-05}
{"train_loss": 0.18200688064098358, "global_step": 130734, "epoch": 1468, "lr": 5.303595120905115e-05}
{"train_loss": 0.10353237390518188, "global_step": 130735, "epoch": 1468, "lr": 5.303537254191698e-05}
{"train_loss": 0.1318168342113495, "global_step": 130736, "epoch": 1468, "lr": 5.3034793874374754e-05}
{"train_loss": 0.1289876401424408, "global_step": 130737, "epoch": 1468, "lr": 5.3034215206424534e-05}
{"train_loss": 0.1060057133436203, "global_step": 130738, "epoch": 1468, "lr": 5.303363653806639e-05}
{"train_loss": 0.17965328693389893, "global_step": 130739, "epoch": 1468, "lr": 5.3033057869300415e-05}
{"train_loss": 0.1288793802763639, "global_step": 130740, "epoch": 1468, "lr": 5.303247920012669e-05, "val_loss": 5.655400276184082}
{"train_loss": 0.06213526800274849, "global_step": 130741, "epoch": 1469, "lr": 5.303190053054529e-05}
{"train_loss": 0.11137650907039642, "global_step": 130742, "epoch": 1469, "lr": 5.3031321860556275e-05}
{"train_loss": 0.08387177437543869, "global_step": 130743, "epoch": 1469, "lr": 5.3030743190159735e-05}
{"train_loss": 0.11135539412498474, "global_step": 130744, "epoch": 1469, "lr": 5.303016451935575e-05}
{"train_loss": 0.18486353754997253, "global_step": 130745, "epoch": 1469, "lr": 5.30295858481444e-05}
{"train_loss": 0.12741945683956146, "global_step": 130746, "epoch": 1469, "lr": 5.3029007176525755e-05}
{"train_loss": 0.07839953899383545, "global_step": 130747, "epoch": 1469, "lr": 5.30284285044999e-05}
{"train_loss": 0.15488511323928833, "global_step": 130748, "epoch": 1469, "lr": 5.302784983206691e-05}
{"train_loss": 0.0821344181895256, "global_step": 130749, "epoch": 1469, "lr": 5.3027271159226867e-05}
{"train_loss": 0.10175594687461853, "global_step": 130750, "epoch": 1469, "lr": 5.302669248597984e-05}
{"train_loss": 0.14025169610977173, "global_step": 130751, "epoch": 1469, "lr": 5.302611381232591e-05}
{"train_loss": 0.1231711134314537, "global_step": 130752, "epoch": 1469, "lr": 5.302553513826515e-05}
{"train_loss": 0.05344514548778534, "global_step": 130753, "epoch": 1469, "lr": 5.302495646379765e-05}
{"train_loss": 0.18123191595077515, "global_step": 130754, "epoch": 1469, "lr": 5.302437778892348e-05}
{"train_loss": 0.21151256561279297, "global_step": 130755, "epoch": 1469, "lr": 5.302379911364273e-05}
{"train_loss": 0.1695699542760849, "global_step": 130756, "epoch": 1469, "lr": 5.302322043795544e-05}
{"train_loss": 0.14423349499702454, "global_step": 130757, "epoch": 1469, "lr": 5.302264176186173e-05}
{"train_loss": 0.09653905779123306, "global_step": 130758, "epoch": 1469, "lr": 5.302206308536167e-05}
{"train_loss": 0.12729738652706146, "global_step": 130759, "epoch": 1469, "lr": 5.302148440845534e-05}
{"train_loss": 0.07890854775905609, "global_step": 130760, "epoch": 1469, "lr": 5.302090573114279e-05}
{"train_loss": 0.08823888748884201, "global_step": 130761, "epoch": 1469, "lr": 5.3020327053424115e-05}
{"train_loss": 0.08131437003612518, "global_step": 130762, "epoch": 1469, "lr": 5.3019748375299395e-05}
{"train_loss": 0.0706821158528328, "global_step": 130763, "epoch": 1469, "lr": 5.301916969676871e-05}
{"train_loss": 0.15158414840698242, "global_step": 130764, "epoch": 1469, "lr": 5.301859101783214e-05}
{"train_loss": 0.13679204881191254, "global_step": 130765, "epoch": 1469, "lr": 5.3018012338489754e-05}
{"train_loss": 0.10383173078298569, "global_step": 130766, "epoch": 1469, "lr": 5.3017433658741625e-05}
{"train_loss": 0.1079479455947876, "global_step": 130767, "epoch": 1469, "lr": 5.301685497858785e-05}
{"train_loss": 0.18747185170650482, "global_step": 130768, "epoch": 1469, "lr": 5.301627629802849e-05}
{"train_loss": 0.1131996139883995, "global_step": 130769, "epoch": 1469, "lr": 5.301569761706362e-05}
{"train_loss": 0.09789971262216568, "global_step": 130770, "epoch": 1469, "lr": 5.3015118935693345e-05}
{"train_loss": 0.06578990817070007, "global_step": 130771, "epoch": 1469, "lr": 5.301454025391771e-05}
{"train_loss": 0.10372324287891388, "global_step": 130772, "epoch": 1469, "lr": 5.3013961571736815e-05}
{"train_loss": 0.1459663212299347, "global_step": 130773, "epoch": 1469, "lr": 5.301338288915072e-05}
{"train_loss": 0.21149079501628876, "global_step": 130774, "epoch": 1469, "lr": 5.301280420615952e-05}
{"train_loss": 0.09134700149297714, "global_step": 130775, "epoch": 1469, "lr": 5.301222552276328e-05}
{"train_loss": 0.14900211989879608, "global_step": 130776, "epoch": 1469, "lr": 5.3011646838962084e-05}
{"train_loss": 0.08699461072683334, "global_step": 130777, "epoch": 1469, "lr": 5.301106815475601e-05}
{"train_loss": 0.12064029276371002, "global_step": 130778, "epoch": 1469, "lr": 5.301048947014514e-05}
{"train_loss": 0.1377033293247223, "global_step": 130779, "epoch": 1469, "lr": 5.300991078512953e-05}
{"train_loss": 0.11342041939496994, "global_step": 130780, "epoch": 1469, "lr": 5.30093320997093e-05}
{"train_loss": 0.11804655939340591, "global_step": 130781, "epoch": 1469, "lr": 5.300875341388448e-05}
{"train_loss": 0.10866202414035797, "global_step": 130782, "epoch": 1469, "lr": 5.300817472765518e-05}
{"train_loss": 0.15864837169647217, "global_step": 130783, "epoch": 1469, "lr": 5.3007596041021466e-05}
{"train_loss": 0.09800992906093597, "global_step": 130784, "epoch": 1469, "lr": 5.300701735398341e-05}
{"train_loss": 0.13939127326011658, "global_step": 130785, "epoch": 1469, "lr": 5.3006438666541114e-05}
{"train_loss": 0.10823646187782288, "global_step": 130786, "epoch": 1469, "lr": 5.3005859978694625e-05}
{"train_loss": 0.18883830308914185, "global_step": 130787, "epoch": 1469, "lr": 5.300528129044404e-05}
{"train_loss": 0.11028807610273361, "global_step": 130788, "epoch": 1469, "lr": 5.300470260178943e-05}
{"train_loss": 0.11143948137760162, "global_step": 130789, "epoch": 1469, "lr": 5.3004123912730875e-05}
{"train_loss": 0.10684514045715332, "global_step": 130790, "epoch": 1469, "lr": 5.300354522326845e-05}
{"train_loss": 0.10839909315109253, "global_step": 130791, "epoch": 1469, "lr": 5.3002966533402235e-05}
{"train_loss": 0.11997003853321075, "global_step": 130792, "epoch": 1469, "lr": 5.30023878431323e-05}
{"train_loss": 0.15604491531848907, "global_step": 130793, "epoch": 1469, "lr": 5.300180915245875e-05}
{"train_loss": 0.20228098332881927, "global_step": 130794, "epoch": 1469, "lr": 5.300123046138164e-05}
{"train_loss": 0.1413576453924179, "global_step": 130795, "epoch": 1469, "lr": 5.300065176990103e-05}
{"train_loss": 0.10174273699522018, "global_step": 130796, "epoch": 1469, "lr": 5.300007307801704e-05}
{"train_loss": 0.12911300361156464, "global_step": 130797, "epoch": 1469, "lr": 5.299949438572972e-05}
{"train_loss": 0.15341372787952423, "global_step": 130798, "epoch": 1469, "lr": 5.299891569303915e-05}
{"train_loss": 0.14943568408489227, "global_step": 130799, "epoch": 1469, "lr": 5.299833699994542e-05}
{"train_loss": 0.16943342983722687, "global_step": 130800, "epoch": 1469, "lr": 5.29977583064486e-05}
{"train_loss": 0.10764099657535553, "global_step": 130801, "epoch": 1469, "lr": 5.2997179612548766e-05}
{"train_loss": 0.10257461667060852, "global_step": 130802, "epoch": 1469, "lr": 5.2996600918245996e-05}
{"train_loss": 0.17761632800102234, "global_step": 130803, "epoch": 1469, "lr": 5.299602222354036e-05}
{"train_loss": 0.12644237279891968, "global_step": 130804, "epoch": 1469, "lr": 5.299544352843196e-05}
{"train_loss": 0.17289645969867706, "global_step": 130805, "epoch": 1469, "lr": 5.2994864832920854e-05}
{"train_loss": 0.09736550599336624, "global_step": 130806, "epoch": 1469, "lr": 5.299428613700713e-05}
{"train_loss": 0.07656684517860413, "global_step": 130807, "epoch": 1469, "lr": 5.299370744069085e-05}
{"train_loss": 0.06542062759399414, "global_step": 130808, "epoch": 1469, "lr": 5.299312874397212e-05}
{"train_loss": 0.12461628019809723, "global_step": 130809, "epoch": 1469, "lr": 5.2992550046850986e-05}
{"train_loss": 0.17545488476753235, "global_step": 130810, "epoch": 1469, "lr": 5.299197134932754e-05}
{"train_loss": 0.11459536850452423, "global_step": 130811, "epoch": 1469, "lr": 5.299139265140186e-05}
{"train_loss": 0.13800545036792755, "global_step": 130812, "epoch": 1469, "lr": 5.2990813953074035e-05}
{"train_loss": 0.14412958920001984, "global_step": 130813, "epoch": 1469, "lr": 5.299023525434412e-05}
{"train_loss": 0.09408537298440933, "global_step": 130814, "epoch": 1469, "lr": 5.2989656555212206e-05}
{"train_loss": 0.14803270995616913, "global_step": 130815, "epoch": 1469, "lr": 5.298907785567837e-05}
{"train_loss": 0.11034800857305527, "global_step": 130816, "epoch": 1469, "lr": 5.298849915574269e-05}
{"train_loss": 0.166714146733284, "global_step": 130817, "epoch": 1469, "lr": 5.298792045540525e-05}
{"train_loss": 0.15131761133670807, "global_step": 130818, "epoch": 1469, "lr": 5.298734175466611e-05}
{"train_loss": 0.13685378432273865, "global_step": 130819, "epoch": 1469, "lr": 5.298676305352536e-05}
{"train_loss": 0.137863427400589, "global_step": 130820, "epoch": 1469, "lr": 5.298618435198308e-05}
{"train_loss": 0.08224482834339142, "global_step": 130821, "epoch": 1469, "lr": 5.298560565003934e-05}
{"train_loss": 0.06960000097751617, "global_step": 130822, "epoch": 1469, "lr": 5.2985026947694213e-05}
{"train_loss": 0.09392500668764114, "global_step": 130823, "epoch": 1469, "lr": 5.2984448244947805e-05}
{"train_loss": 0.18019093573093414, "global_step": 130824, "epoch": 1469, "lr": 5.298386954180016e-05}
{"train_loss": 0.14337047934532166, "global_step": 130825, "epoch": 1469, "lr": 5.2983290838251375e-05}
{"train_loss": 0.1023993194103241, "global_step": 130826, "epoch": 1469, "lr": 5.298271213430153e-05}
{"train_loss": 0.11089671403169632, "global_step": 130827, "epoch": 1469, "lr": 5.2982133429950684e-05}
{"train_loss": 0.1401810199022293, "global_step": 130828, "epoch": 1469, "lr": 5.2981554725198935e-05}
{"train_loss": 0.1244594944075922, "global_step": 130829, "epoch": 1469, "lr": 5.298097602004635e-05, "val_loss": 5.718768119812012}
{"train_loss": 0.20297808945178986, "global_step": 130830, "epoch": 1470, "lr": 5.2980397314493e-05}
{"train_loss": 0.09970089793205261, "global_step": 130831, "epoch": 1470, "lr": 5.297981860853897e-05}
{"train_loss": 0.10078683495521545, "global_step": 130832, "epoch": 1470, "lr": 5.297923990218436e-05}
{"train_loss": 0.09090371429920197, "global_step": 130833, "epoch": 1470, "lr": 5.297866119542921e-05}
{"train_loss": 0.16776441037654877, "global_step": 130834, "epoch": 1470, "lr": 5.297808248827363e-05}
{"train_loss": 0.12196032702922821, "global_step": 130835, "epoch": 1470, "lr": 5.2977503780717674e-05}
{"train_loss": 0.07782827317714691, "global_step": 130836, "epoch": 1470, "lr": 5.297692507276143e-05}
{"train_loss": 0.11232593655586243, "global_step": 130837, "epoch": 1470, "lr": 5.2976346364404974e-05}
{"train_loss": 0.0858626589179039, "global_step": 130838, "epoch": 1470, "lr": 5.297576765564839e-05}
{"train_loss": 0.09779596328735352, "global_step": 130839, "epoch": 1470, "lr": 5.297518894649174e-05}
{"train_loss": 0.0934867188334465, "global_step": 130840, "epoch": 1470, "lr": 5.297461023693512e-05}
{"train_loss": 0.12284848093986511, "global_step": 130841, "epoch": 1470, "lr": 5.2974031526978597e-05}
{"train_loss": 0.13585807383060455, "global_step": 130842, "epoch": 1470, "lr": 5.297345281662225e-05}
{"train_loss": 0.13389723002910614, "global_step": 130843, "epoch": 1470, "lr": 5.297287410586617e-05}
{"train_loss": 0.09067201614379883, "global_step": 130844, "epoch": 1470, "lr": 5.29722953947104e-05}
{"train_loss": 0.10230310261249542, "global_step": 130845, "epoch": 1470, "lr": 5.297171668315506e-05}
{"train_loss": 0.10507617145776749, "global_step": 130846, "epoch": 1470, "lr": 5.2971137971200205e-05}
{"train_loss": 0.09022805839776993, "global_step": 130847, "epoch": 1470, "lr": 5.297055925884592e-05}
{"train_loss": 0.17265769839286804, "global_step": 130848, "epoch": 1470, "lr": 5.2969980546092266e-05}
{"train_loss": 0.11362458765506744, "global_step": 130849, "epoch": 1470, "lr": 5.296940183293933e-05}
{"train_loss": 0.1066625714302063, "global_step": 130850, "epoch": 1470, "lr": 5.296882311938721e-05}
{"train_loss": 0.16191883385181427, "global_step": 130851, "epoch": 1470, "lr": 5.2968244405435985e-05}
{"train_loss": 0.11547581106424332, "global_step": 130852, "epoch": 1470, "lr": 5.2967665691085686e-05}
{"train_loss": 0.0760575458407402, "global_step": 130853, "epoch": 1470, "lr": 5.2967086976336435e-05}
{"train_loss": 0.08396018296480179, "global_step": 130854, "epoch": 1470, "lr": 5.29665082611883e-05}
{"train_loss": 0.14863760769367218, "global_step": 130855, "epoch": 1470, "lr": 5.296592954564134e-05}
{"train_loss": 0.12332762777805328, "global_step": 130856, "epoch": 1470, "lr": 5.296535082969565e-05}
{"train_loss": 0.14162024855613708, "global_step": 130857, "epoch": 1470, "lr": 5.29647721133513e-05}
{"train_loss": 0.15030062198638916, "global_step": 130858, "epoch": 1470, "lr": 5.296419339660838e-05}
{"train_loss": 0.1112559586763382, "global_step": 130859, "epoch": 1470, "lr": 5.296361467946697e-05}
{"train_loss": 0.15573354065418243, "global_step": 130860, "epoch": 1470, "lr": 5.296303596192712e-05}
{"train_loss": 0.14935468137264252, "global_step": 130861, "epoch": 1470, "lr": 5.296245724398894e-05}
{"train_loss": 0.11055458337068558, "global_step": 130862, "epoch": 1470, "lr": 5.2961878525652484e-05}
{"train_loss": 0.182341530919075, "global_step": 130863, "epoch": 1470, "lr": 5.296129980691784e-05}
{"train_loss": 0.17029538750648499, "global_step": 130864, "epoch": 1470, "lr": 5.296072108778509e-05}
{"train_loss": 0.12633320689201355, "global_step": 130865, "epoch": 1470, "lr": 5.2960142368254306e-05}
{"train_loss": 0.05684637650847435, "global_step": 130866, "epoch": 1470, "lr": 5.295956364832557e-05}
{"train_loss": 0.1602322906255722, "global_step": 130867, "epoch": 1470, "lr": 5.2958984927998945e-05}
{"train_loss": 0.10246628522872925, "global_step": 130868, "epoch": 1470, "lr": 5.2958406207274526e-05}
{"train_loss": 0.10626614838838577, "global_step": 130869, "epoch": 1470, "lr": 5.295782748615239e-05}
{"train_loss": 0.13381269574165344, "global_step": 130870, "epoch": 1470, "lr": 5.295724876463262e-05}
{"train_loss": 0.2474122941493988, "global_step": 130871, "epoch": 1470, "lr": 5.2956670042715274e-05}
{"train_loss": 0.12115353345870972, "global_step": 130872, "epoch": 1470, "lr": 5.295609132040044e-05}
{"train_loss": 0.10566096752882004, "global_step": 130873, "epoch": 1470, "lr": 5.2955512597688183e-05}
{"train_loss": 0.15130740404129028, "global_step": 130874, "epoch": 1470, "lr": 5.295493387457861e-05}
{"train_loss": 0.10775889456272125, "global_step": 130875, "epoch": 1470, "lr": 5.2954355151071775e-05}
{"train_loss": 0.10517678409814835, "global_step": 130876, "epoch": 1470, "lr": 5.295377642716777e-05}
{"train_loss": 0.19174572825431824, "global_step": 130877, "epoch": 1470, "lr": 5.295319770286665e-05}
{"train_loss": 0.20491336286067963, "global_step": 130878, "epoch": 1470, "lr": 5.295261897816852e-05}
{"train_loss": 0.22718454897403717, "global_step": 130879, "epoch": 1470, "lr": 5.295204025307344e-05}
{"train_loss": 0.1407841444015503, "global_step": 130880, "epoch": 1470, "lr": 5.29514615275815e-05}
{"train_loss": 0.15801399946212769, "global_step": 130881, "epoch": 1470, "lr": 5.295088280169277e-05}
{"train_loss": 0.11260838061571121, "global_step": 130882, "epoch": 1470, "lr": 5.295030407540733e-05}
{"train_loss": 0.12925218045711517, "global_step": 130883, "epoch": 1470, "lr": 5.294972534872527e-05}
{"train_loss": 0.1063258945941925, "global_step": 130884, "epoch": 1470, "lr": 5.294914662164664e-05}
{"train_loss": 0.11711262166500092, "global_step": 130885, "epoch": 1470, "lr": 5.2948567894171516e-05}
{"train_loss": 0.10510899871587753, "global_step": 130886, "epoch": 1470, "lr": 5.294798916630002e-05}
{"train_loss": 0.07478147000074387, "global_step": 130887, "epoch": 1470, "lr": 5.2947410438032194e-05}
{"train_loss": 0.1415124535560608, "global_step": 130888, "epoch": 1470, "lr": 5.294683170936813e-05}
{"train_loss": 0.13569222390651703, "global_step": 130889, "epoch": 1470, "lr": 5.2946252980307897e-05}
{"train_loss": 0.11048988997936249, "global_step": 130890, "epoch": 1470, "lr": 5.2945674250851585e-05}
{"train_loss": 0.10331394523382187, "global_step": 130891, "epoch": 1470, "lr": 5.2945095520999255e-05}
{"train_loss": 0.1365967094898224, "global_step": 130892, "epoch": 1470, "lr": 5.2944516790751e-05}
{"train_loss": 0.12646004557609558, "global_step": 130893, "epoch": 1470, "lr": 5.294393806010688e-05}
{"train_loss": 0.1828523874282837, "global_step": 130894, "epoch": 1470, "lr": 5.294335932906699e-05}
{"train_loss": 0.12418454140424728, "global_step": 130895, "epoch": 1470, "lr": 5.294278059763139e-05}
{"train_loss": 0.1062915250658989, "global_step": 130896, "epoch": 1470, "lr": 5.294220186580018e-05}
{"train_loss": 0.11934217810630798, "global_step": 130897, "epoch": 1470, "lr": 5.294162313357343e-05}
{"train_loss": 0.1012377217411995, "global_step": 130898, "epoch": 1470, "lr": 5.29410444009512e-05}
{"train_loss": 0.08081167936325073, "global_step": 130899, "epoch": 1470, "lr": 5.2940465667933594e-05}
{"train_loss": 0.19240912795066833, "global_step": 130900, "epoch": 1470, "lr": 5.293988693452067e-05}
{"train_loss": 0.09905858337879181, "global_step": 130901, "epoch": 1470, "lr": 5.2939308200712526e-05}
{"train_loss": 0.14299176633358002, "global_step": 130902, "epoch": 1470, "lr": 5.293872946650922e-05}
{"train_loss": 0.12681610882282257, "global_step": 130903, "epoch": 1470, "lr": 5.293815073191082e-05}
{"train_loss": 0.12434353679418564, "global_step": 130904, "epoch": 1470, "lr": 5.293757199691745e-05}
{"train_loss": 0.12833498418331146, "global_step": 130905, "epoch": 1470, "lr": 5.2936993261529146e-05}
{"train_loss": 0.12274019420146942, "global_step": 130906, "epoch": 1470, "lr": 5.293641452574599e-05}
{"train_loss": 0.11926453560590744, "global_step": 130907, "epoch": 1470, "lr": 5.293583578956809e-05}
{"train_loss": 0.08658351004123688, "global_step": 130908, "epoch": 1470, "lr": 5.293525705299549e-05}
{"train_loss": 0.08669941127300262, "global_step": 130909, "epoch": 1470, "lr": 5.29346783160283e-05}
{"train_loss": 0.1005530059337616, "global_step": 130910, "epoch": 1470, "lr": 5.2934099578666566e-05}
{"train_loss": 0.08522811532020569, "global_step": 130911, "epoch": 1470, "lr": 5.293352084091037e-05}
{"train_loss": 0.14847415685653687, "global_step": 130912, "epoch": 1470, "lr": 5.29329421027598e-05}
{"train_loss": 0.1935386210680008, "global_step": 130913, "epoch": 1470, "lr": 5.293236336421493e-05}
{"train_loss": 0.11901126056909561, "global_step": 130914, "epoch": 1470, "lr": 5.293178462527585e-05}
{"train_loss": 0.11052698642015457, "global_step": 130915, "epoch": 1470, "lr": 5.293120588594261e-05}
{"train_loss": 0.17507877945899963, "global_step": 130916, "epoch": 1470, "lr": 5.293062714621533e-05}
{"train_loss": 0.14020313322544098, "global_step": 130917, "epoch": 1470, "lr": 5.2930048406094044e-05}
{"train_loss": 0.126624649871936, "global_step": 130918, "epoch": 1470, "lr": 5.292946966557884e-05, "val_loss": 5.674862384796143, "train_action_mse_error": 38.640018463134766}
{"train_loss": 0.16454795002937317, "global_step": 130919, "epoch": 1471, "lr": 5.2928890924669826e-05}
{"train_loss": 0.1354859322309494, "global_step": 130920, "epoch": 1471, "lr": 5.292831218336706e-05}
{"train_loss": 0.157115176320076, "global_step": 130921, "epoch": 1471, "lr": 5.29277334416706e-05}
{"train_loss": 0.08080868422985077, "global_step": 130922, "epoch": 1471, "lr": 5.292715469958054e-05}
{"train_loss": 0.11291152238845825, "global_step": 130923, "epoch": 1471, "lr": 5.292657595709697e-05}
{"train_loss": 0.17372694611549377, "global_step": 130924, "epoch": 1471, "lr": 5.292599721421996e-05}
{"train_loss": 0.09439508616924286, "global_step": 130925, "epoch": 1471, "lr": 5.292541847094958e-05}
{"train_loss": 0.13981060683727264, "global_step": 130926, "epoch": 1471, "lr": 5.292483972728591e-05}
{"train_loss": 0.16286154091358185, "global_step": 130927, "epoch": 1471, "lr": 5.292426098322904e-05}
{"train_loss": 0.0757194459438324, "global_step": 130928, "epoch": 1471, "lr": 5.2923682238779035e-05}
{"train_loss": 0.15514473617076874, "global_step": 130929, "epoch": 1471, "lr": 5.292310349393599e-05}
{"train_loss": 0.09285761415958405, "global_step": 130930, "epoch": 1471, "lr": 5.292252474869996e-05}
{"train_loss": 0.06746095418930054, "global_step": 130931, "epoch": 1471, "lr": 5.2921946003071035e-05}
{"train_loss": 0.07786402851343155, "global_step": 130932, "epoch": 1471, "lr": 5.292136725704928e-05}
{"train_loss": 0.13766327500343323, "global_step": 130933, "epoch": 1471, "lr": 5.292078851063479e-05}
{"train_loss": 0.1511441022157669, "global_step": 130934, "epoch": 1471, "lr": 5.292020976382762e-05}
{"train_loss": 0.08613669872283936, "global_step": 130935, "epoch": 1471, "lr": 5.291963101662788e-05}
{"train_loss": 0.12035008519887924, "global_step": 130936, "epoch": 1471, "lr": 5.291905226903563e-05}
{"train_loss": 0.09639135748147964, "global_step": 130937, "epoch": 1471, "lr": 5.2918473521050946e-05}
{"train_loss": 0.09086687862873077, "global_step": 130938, "epoch": 1471, "lr": 5.2917894772673904e-05}
{"train_loss": 0.15901325643062592, "global_step": 130939, "epoch": 1471, "lr": 5.29173160239046e-05}
{"train_loss": 0.15514901280403137, "global_step": 130940, "epoch": 1471, "lr": 5.291673727474307e-05}
{"train_loss": 0.10259127616882324, "global_step": 130941, "epoch": 1471, "lr": 5.291615852518944e-05}
{"train_loss": 0.0873894989490509, "global_step": 130942, "epoch": 1471, "lr": 5.2915579775243776e-05}
{"train_loss": 0.16132807731628418, "global_step": 130943, "epoch": 1471, "lr": 5.291500102490613e-05}
{"train_loss": 0.13887649774551392, "global_step": 130944, "epoch": 1471, "lr": 5.291442227417661e-05}
{"train_loss": 0.0976235568523407, "global_step": 130945, "epoch": 1471, "lr": 5.2913843523055285e-05}
{"train_loss": 0.056627947837114334, "global_step": 130946, "epoch": 1471, "lr": 5.2913264771542215e-05}
{"train_loss": 0.05635377764701843, "global_step": 130947, "epoch": 1471, "lr": 5.2912686019637504e-05}
{"train_loss": 0.23229092359542847, "global_step": 130948, "epoch": 1471, "lr": 5.291210726734123e-05}
{"train_loss": 0.153901606798172, "global_step": 130949, "epoch": 1471, "lr": 5.2911528514653444e-05}
{"train_loss": 0.15163815021514893, "global_step": 130950, "epoch": 1471, "lr": 5.2910949761574224e-05}
{"train_loss": 0.12751151621341705, "global_step": 130951, "epoch": 1471, "lr": 5.2910371008103674e-05}
{"train_loss": 0.10454729199409485, "global_step": 130952, "epoch": 1471, "lr": 5.290979225424186e-05}
{"train_loss": 0.057584505528211594, "global_step": 130953, "epoch": 1471, "lr": 5.290921349998887e-05}
{"train_loss": 0.11437708884477615, "global_step": 130954, "epoch": 1471, "lr": 5.290863474534477e-05}
{"train_loss": 0.10577844083309174, "global_step": 130955, "epoch": 1471, "lr": 5.290805599030963e-05}
{"train_loss": 0.10028880834579468, "global_step": 130956, "epoch": 1471, "lr": 5.290747723488354e-05}
{"train_loss": 0.09811579436063766, "global_step": 130957, "epoch": 1471, "lr": 5.290689847906657e-05}
{"train_loss": 0.07176550477743149, "global_step": 130958, "epoch": 1471, "lr": 5.29063197228588e-05}
{"train_loss": 0.1133296936750412, "global_step": 130959, "epoch": 1471, "lr": 5.290574096626032e-05}
{"train_loss": 0.08202773332595825, "global_step": 130960, "epoch": 1471, "lr": 5.29051622092712e-05}
{"train_loss": 0.1372147649526596, "global_step": 130961, "epoch": 1471, "lr": 5.290458345189151e-05}
{"train_loss": 0.18547020852565765, "global_step": 130962, "epoch": 1471, "lr": 5.290400469412133e-05}
{"train_loss": 0.10051700472831726, "global_step": 130963, "epoch": 1471, "lr": 5.290342593596075e-05}
{"train_loss": 0.13384957611560822, "global_step": 130964, "epoch": 1471, "lr": 5.2902847177409844e-05}
{"train_loss": 0.1064361110329628, "global_step": 130965, "epoch": 1471, "lr": 5.290226841846868e-05}
{"train_loss": 0.14701828360557556, "global_step": 130966, "epoch": 1471, "lr": 5.290168965913734e-05}
{"train_loss": 0.1243753731250763, "global_step": 130967, "epoch": 1471, "lr": 5.290111089941592e-05}
{"train_loss": 0.0781579464673996, "global_step": 130968, "epoch": 1471, "lr": 5.290053213930446e-05}
{"train_loss": 0.15640059113502502, "global_step": 130969, "epoch": 1471, "lr": 5.289995337880307e-05}
{"train_loss": 0.10211393982172012, "global_step": 130970, "epoch": 1471, "lr": 5.2899374617911815e-05}
{"train_loss": 0.15408138930797577, "global_step": 130971, "epoch": 1471, "lr": 5.289879585663077e-05}
{"train_loss": 0.131214439868927, "global_step": 130972, "epoch": 1471, "lr": 5.289821709496001e-05}
{"train_loss": 0.13132280111312866, "global_step": 130973, "epoch": 1471, "lr": 5.289763833289963e-05}
{"train_loss": 0.15295878052711487, "global_step": 130974, "epoch": 1471, "lr": 5.28970595704497e-05}
{"train_loss": 0.06525316834449768, "global_step": 130975, "epoch": 1471, "lr": 5.2896480807610295e-05}
{"train_loss": 0.09006613492965698, "global_step": 130976, "epoch": 1471, "lr": 5.2895902044381486e-05}
{"train_loss": 0.20341449975967407, "global_step": 130977, "epoch": 1471, "lr": 5.289532328076335e-05}
{"train_loss": 0.1576811820268631, "global_step": 130978, "epoch": 1471, "lr": 5.289474451675599e-05}
{"train_loss": 0.10054029524326324, "global_step": 130979, "epoch": 1471, "lr": 5.289416575235947e-05}
{"train_loss": 0.2191452831029892, "global_step": 130980, "epoch": 1471, "lr": 5.289358698757385e-05}
{"train_loss": 0.07799168676137924, "global_step": 130981, "epoch": 1471, "lr": 5.289300822239923e-05}
{"train_loss": 0.12838831543922424, "global_step": 130982, "epoch": 1471, "lr": 5.289242945683568e-05}
{"train_loss": 0.15146590769290924, "global_step": 130983, "epoch": 1471, "lr": 5.289185069088328e-05}
{"train_loss": 0.12414415925741196, "global_step": 130984, "epoch": 1471, "lr": 5.289127192454211e-05}
{"train_loss": 0.1929139345884323, "global_step": 130985, "epoch": 1471, "lr": 5.289069315781223e-05}
{"train_loss": 0.14807561039924622, "global_step": 130986, "epoch": 1471, "lr": 5.2890114390693746e-05}
{"train_loss": 0.14619934558868408, "global_step": 130987, "epoch": 1471, "lr": 5.288953562318673e-05}
{"train_loss": 0.1306402087211609, "global_step": 130988, "epoch": 1471, "lr": 5.288895685529123e-05}
{"train_loss": 0.12009825557470322, "global_step": 130989, "epoch": 1471, "lr": 5.288837808700735e-05}
{"train_loss": 0.1066330224275589, "global_step": 130990, "epoch": 1471, "lr": 5.2887799318335164e-05}
{"train_loss": 0.10125065594911575, "global_step": 130991, "epoch": 1471, "lr": 5.288722054927475e-05}
{"train_loss": 0.06872647255659103, "global_step": 130992, "epoch": 1471, "lr": 5.2886641779826184e-05}
{"train_loss": 0.12526901066303253, "global_step": 130993, "epoch": 1471, "lr": 5.288606300998954e-05}
{"train_loss": 0.06054075434803963, "global_step": 130994, "epoch": 1471, "lr": 5.288548423976489e-05}
{"train_loss": 0.18575526773929596, "global_step": 130995, "epoch": 1471, "lr": 5.288490546915234e-05}
{"train_loss": 0.09452029317617416, "global_step": 130996, "epoch": 1471, "lr": 5.288432669815194e-05}
{"train_loss": 0.14805397391319275, "global_step": 130997, "epoch": 1471, "lr": 5.2883747926763784e-05}
{"train_loss": 0.16506341099739075, "global_step": 130998, "epoch": 1471, "lr": 5.2883169154987945e-05}
{"train_loss": 0.11061152815818787, "global_step": 130999, "epoch": 1471, "lr": 5.2882590382824496e-05}
{"train_loss": 0.1472191959619522, "global_step": 131000, "epoch": 1471, "lr": 5.288201161027352e-05}
{"train_loss": 0.11361312866210938, "global_step": 131001, "epoch": 1471, "lr": 5.288143283733509e-05}
{"train_loss": 0.14538991451263428, "global_step": 131002, "epoch": 1471, "lr": 5.2880854064009275e-05}
{"train_loss": 0.09410326182842255, "global_step": 131003, "epoch": 1471, "lr": 5.2880275290296176e-05}
{"train_loss": 0.13449710607528687, "global_step": 131004, "epoch": 1471, "lr": 5.287969651619587e-05}
{"train_loss": 0.10914180427789688, "global_step": 131005, "epoch": 1471, "lr": 5.2879117741708405e-05}
{"train_loss": 0.1512959599494934, "global_step": 131006, "epoch": 1471, "lr": 5.28785389668339e-05}
{"train_loss": 0.12290619736474552, "global_step": 131007, "epoch": 1471, "lr": 5.287796019157239e-05, "val_loss": 5.660372734069824}
{"train_loss": 0.11406414955854416, "global_step": 131008, "epoch": 1472, "lr": 5.287738141592398e-05}
{"train_loss": 0.19884027540683746, "global_step": 131009, "epoch": 1472, "lr": 5.287680263988874e-05}
{"train_loss": 0.1311759352684021, "global_step": 131010, "epoch": 1472, "lr": 5.2876223863466745e-05}
{"train_loss": 0.10282180458307266, "global_step": 131011, "epoch": 1472, "lr": 5.2875645086658086e-05}
{"train_loss": 0.12850415706634521, "global_step": 131012, "epoch": 1472, "lr": 5.2875066309462815e-05}
{"train_loss": 0.08420310914516449, "global_step": 131013, "epoch": 1472, "lr": 5.287448753188103e-05}
{"train_loss": 0.127811998128891, "global_step": 131014, "epoch": 1472, "lr": 5.2873908753912826e-05}
{"train_loss": 0.1544124335050583, "global_step": 131015, "epoch": 1472, "lr": 5.287332997555824e-05}
{"train_loss": 0.09099801629781723, "global_step": 131016, "epoch": 1472, "lr": 5.287275119681738e-05}
{"train_loss": 0.20282672345638275, "global_step": 131017, "epoch": 1472, "lr": 5.287217241769031e-05}
{"train_loss": 0.14055074751377106, "global_step": 131018, "epoch": 1472, "lr": 5.287159363817711e-05}
{"train_loss": 0.11405082792043686, "global_step": 131019, "epoch": 1472, "lr": 5.2871014858277866e-05}
{"train_loss": 0.13639956712722778, "global_step": 131020, "epoch": 1472, "lr": 5.2870436077992645e-05}
{"train_loss": 0.11056140810251236, "global_step": 131021, "epoch": 1472, "lr": 5.286985729732152e-05}
{"train_loss": 0.15801237523555756, "global_step": 131022, "epoch": 1472, "lr": 5.286927851626459e-05}
{"train_loss": 0.11773091554641724, "global_step": 131023, "epoch": 1472, "lr": 5.2868699734821913e-05}
{"train_loss": 0.1493580937385559, "global_step": 131024, "epoch": 1472, "lr": 5.286812095299358e-05}
{"train_loss": 0.06796429306268692, "global_step": 131025, "epoch": 1472, "lr": 5.2867542170779674e-05}
{"train_loss": 0.1676415354013443, "global_step": 131026, "epoch": 1472, "lr": 5.2866963388180244e-05}
{"train_loss": 0.14295540750026703, "global_step": 131027, "epoch": 1472, "lr": 5.2866384605195394e-05}
{"train_loss": 0.12696221470832825, "global_step": 131028, "epoch": 1472, "lr": 5.2865805821825184e-05}
{"train_loss": 0.1070350632071495, "global_step": 131029, "epoch": 1472, "lr": 5.286522703806971e-05}
{"train_loss": 0.11654866486787796, "global_step": 131030, "epoch": 1472, "lr": 5.2864648253929026e-05}
{"train_loss": 0.1188889592885971, "global_step": 131031, "epoch": 1472, "lr": 5.2864069469403246e-05}
{"train_loss": 0.18237708508968353, "global_step": 131032, "epoch": 1472, "lr": 5.286349068449241e-05}
{"train_loss": 0.12275625765323639, "global_step": 131033, "epoch": 1472, "lr": 5.286291189919662e-05}
{"train_loss": 0.11258739233016968, "global_step": 131034, "epoch": 1472, "lr": 5.286233311351595e-05}
{"train_loss": 0.14316248893737793, "global_step": 131035, "epoch": 1472, "lr": 5.286175432745047e-05}
{"train_loss": 0.09451175481081009, "global_step": 131036, "epoch": 1472, "lr": 5.286117554100026e-05}
{"train_loss": 0.10059365630149841, "global_step": 131037, "epoch": 1472, "lr": 5.28605967541654e-05}
{"train_loss": 0.08144474774599075, "global_step": 131038, "epoch": 1472, "lr": 5.286001796694598e-05}
{"train_loss": 0.12628385424613953, "global_step": 131039, "epoch": 1472, "lr": 5.285943917934205e-05}
{"train_loss": 0.0828661397099495, "global_step": 131040, "epoch": 1472, "lr": 5.285886039135372e-05}
{"train_loss": 0.07636753469705582, "global_step": 131041, "epoch": 1472, "lr": 5.285828160298103e-05}
{"train_loss": 0.11596541106700897, "global_step": 131042, "epoch": 1472, "lr": 5.285770281422409e-05}
{"train_loss": 0.14788833260536194, "global_step": 131043, "epoch": 1472, "lr": 5.2857124025082975e-05}
{"train_loss": 0.19101883471012115, "global_step": 131044, "epoch": 1472, "lr": 5.285654523555774e-05}
{"train_loss": 0.154544860124588, "global_step": 131045, "epoch": 1472, "lr": 5.2855966445648486e-05}
{"train_loss": 0.18863952159881592, "global_step": 131046, "epoch": 1472, "lr": 5.285538765535528e-05}
{"train_loss": 0.1096392497420311, "global_step": 131047, "epoch": 1472, "lr": 5.2854808864678205e-05}
{"train_loss": 0.14396950602531433, "global_step": 131048, "epoch": 1472, "lr": 5.285423007361733e-05}
{"train_loss": 0.10268639773130417, "global_step": 131049, "epoch": 1472, "lr": 5.2853651282172734e-05}
{"train_loss": 0.10478329658508301, "global_step": 131050, "epoch": 1472, "lr": 5.28530724903445e-05}
{"train_loss": 0.17907480895519257, "global_step": 131051, "epoch": 1472, "lr": 5.285249369813271e-05}
{"train_loss": 0.06456895172595978, "global_step": 131052, "epoch": 1472, "lr": 5.285191490553744e-05}
{"train_loss": 0.11489906162023544, "global_step": 131053, "epoch": 1472, "lr": 5.285133611255876e-05}
{"train_loss": 0.13544264435768127, "global_step": 131054, "epoch": 1472, "lr": 5.285075731919675e-05}
{"train_loss": 0.09651429206132889, "global_step": 131055, "epoch": 1472, "lr": 5.28501785254515e-05}
{"train_loss": 0.14616578817367554, "global_step": 131056, "epoch": 1472, "lr": 5.284959973132308e-05}
{"train_loss": 0.10253261774778366, "global_step": 131057, "epoch": 1472, "lr": 5.284902093681156e-05}
{"train_loss": 0.1236419603228569, "global_step": 131058, "epoch": 1472, "lr": 5.284844214191702e-05}
{"train_loss": 0.11991021037101746, "global_step": 131059, "epoch": 1472, "lr": 5.2847863346639545e-05}
{"train_loss": 0.10803394764661789, "global_step": 131060, "epoch": 1472, "lr": 5.2847284550979205e-05}
{"train_loss": 0.13220350444316864, "global_step": 131061, "epoch": 1472, "lr": 5.284670575493609e-05}
{"train_loss": 0.11791370064020157, "global_step": 131062, "epoch": 1472, "lr": 5.284612695851027e-05}
{"train_loss": 0.052601221948862076, "global_step": 131063, "epoch": 1472, "lr": 5.284554816170182e-05}
{"train_loss": 0.12184038758277893, "global_step": 131064, "epoch": 1472, "lr": 5.284496936451083e-05}
{"train_loss": 0.0998145267367363, "global_step": 131065, "epoch": 1472, "lr": 5.2844390566937364e-05}
{"train_loss": 0.16789254546165466, "global_step": 131066, "epoch": 1472, "lr": 5.28438117689815e-05}
{"train_loss": 0.1377047896385193, "global_step": 131067, "epoch": 1472, "lr": 5.2843232970643294e-05}
{"train_loss": 0.11644303053617477, "global_step": 131068, "epoch": 1472, "lr": 5.284265417192289e-05}
{"train_loss": 0.07430868595838547, "global_step": 131069, "epoch": 1472, "lr": 5.284207537282031e-05}
{"train_loss": 0.06747083365917206, "global_step": 131070, "epoch": 1472, "lr": 5.2841496573335644e-05}
{"train_loss": 0.07414118200540543, "global_step": 131071, "epoch": 1472, "lr": 5.284091777346898e-05}
{"train_loss": 0.10043392330408096, "global_step": 131072, "epoch": 1472, "lr": 5.2840338973220386e-05}
{"train_loss": 0.14628858864307404, "global_step": 131073, "epoch": 1472, "lr": 5.283976017258995e-05}
{"train_loss": 0.1210561990737915, "global_step": 131074, "epoch": 1472, "lr": 5.283918137157773e-05}
{"train_loss": 0.1223856583237648, "global_step": 131075, "epoch": 1472, "lr": 5.2838602570183825e-05}
{"train_loss": 0.1373385190963745, "global_step": 131076, "epoch": 1472, "lr": 5.283802376840831e-05}
{"train_loss": 0.08704528957605362, "global_step": 131077, "epoch": 1472, "lr": 5.2837444966251246e-05}
{"train_loss": 0.14651301503181458, "global_step": 131078, "epoch": 1472, "lr": 5.283686616371273e-05}
{"train_loss": 0.08849575370550156, "global_step": 131079, "epoch": 1472, "lr": 5.283628736079282e-05}
{"train_loss": 0.08028973639011383, "global_step": 131080, "epoch": 1472, "lr": 5.283570855749162e-05}
{"train_loss": 0.1345519870519638, "global_step": 131081, "epoch": 1472, "lr": 5.283512975380919e-05}
{"train_loss": 0.09464874863624573, "global_step": 131082, "epoch": 1472, "lr": 5.2834550949745606e-05}
{"train_loss": 0.12753106653690338, "global_step": 131083, "epoch": 1472, "lr": 5.2833972145300966e-05}
{"train_loss": 0.1460786908864975, "global_step": 131084, "epoch": 1472, "lr": 5.2833393340475324e-05}
{"train_loss": 0.07729670405387878, "global_step": 131085, "epoch": 1472, "lr": 5.283281453526876e-05}
{"train_loss": 0.13671453297138214, "global_step": 131086, "epoch": 1472, "lr": 5.283223572968136e-05}
{"train_loss": 0.07551337778568268, "global_step": 131087, "epoch": 1472, "lr": 5.2831656923713204e-05}
{"train_loss": 0.10031400620937347, "global_step": 131088, "epoch": 1472, "lr": 5.2831078117364366e-05}
{"train_loss": 0.08947328478097916, "global_step": 131089, "epoch": 1472, "lr": 5.2830499310634926e-05}
{"train_loss": 0.12879787385463715, "global_step": 131090, "epoch": 1472, "lr": 5.282992050352495e-05}
{"train_loss": 0.18572750687599182, "global_step": 131091, "epoch": 1472, "lr": 5.282934169603454e-05}
{"train_loss": 0.14065824449062347, "global_step": 131092, "epoch": 1472, "lr": 5.282876288816375e-05}
{"train_loss": 0.12414418905973434, "global_step": 131093, "epoch": 1472, "lr": 5.282818407991268e-05}
{"train_loss": 0.11217466741800308, "global_step": 131094, "epoch": 1472, "lr": 5.282760527128138e-05}
{"train_loss": 0.11002472788095474, "global_step": 131095, "epoch": 1472, "lr": 5.2827026462269944e-05}
{"train_loss": 0.12226560649074865, "global_step": 131096, "epoch": 1472, "lr": 5.282644765287846e-05, "val_loss": 5.744833469390869}
{"train_loss": 0.15857915580272675, "global_step": 131097, "epoch": 1473, "lr": 5.282586884310699e-05}
{"train_loss": 0.11379105597734451, "global_step": 131098, "epoch": 1473, "lr": 5.282529003295561e-05}
{"train_loss": 0.08882956206798553, "global_step": 131099, "epoch": 1473, "lr": 5.282471122242441e-05}
{"train_loss": 0.14318223297595978, "global_step": 131100, "epoch": 1473, "lr": 5.282413241151346e-05}
{"train_loss": 0.10700145363807678, "global_step": 131101, "epoch": 1473, "lr": 5.282355360022284e-05}
{"train_loss": 0.12697693705558777, "global_step": 131102, "epoch": 1473, "lr": 5.282297478855263e-05}
{"train_loss": 0.09613122791051865, "global_step": 131103, "epoch": 1473, "lr": 5.2822395976502906e-05}
{"train_loss": 0.07731226086616516, "global_step": 131104, "epoch": 1473, "lr": 5.282181716407374e-05}
{"train_loss": 0.17280900478363037, "global_step": 131105, "epoch": 1473, "lr": 5.282123835126521e-05}
{"train_loss": 0.10794248431921005, "global_step": 131106, "epoch": 1473, "lr": 5.282065953807741e-05}
{"train_loss": 0.16056126356124878, "global_step": 131107, "epoch": 1473, "lr": 5.28200807245104e-05}
{"train_loss": 0.09010545164346695, "global_step": 131108, "epoch": 1473, "lr": 5.2819501910564275e-05}
{"train_loss": 0.1756901741027832, "global_step": 131109, "epoch": 1473, "lr": 5.281892309623909e-05}
{"train_loss": 0.08930635452270508, "global_step": 131110, "epoch": 1473, "lr": 5.2818344281534937e-05}
{"train_loss": 0.10732683539390564, "global_step": 131111, "epoch": 1473, "lr": 5.2817765466451895e-05}
{"train_loss": 0.0935068130493164, "global_step": 131112, "epoch": 1473, "lr": 5.2817186650990046e-05}
{"train_loss": 0.09955677390098572, "global_step": 131113, "epoch": 1473, "lr": 5.281660783514946e-05}
{"train_loss": 0.17881274223327637, "global_step": 131114, "epoch": 1473, "lr": 5.28160290189302e-05}
{"train_loss": 0.12198474258184433, "global_step": 131115, "epoch": 1473, "lr": 5.281545020233237e-05}
{"train_loss": 0.14805148541927338, "global_step": 131116, "epoch": 1473, "lr": 5.2814871385356036e-05}
{"train_loss": 0.13110655546188354, "global_step": 131117, "epoch": 1473, "lr": 5.281429256800128e-05}
{"train_loss": 0.10412301868200302, "global_step": 131118, "epoch": 1473, "lr": 5.281371375026817e-05}
{"train_loss": 0.13270467519760132, "global_step": 131119, "epoch": 1473, "lr": 5.2813134932156794e-05}
{"train_loss": 0.12598375976085663, "global_step": 131120, "epoch": 1473, "lr": 5.281255611366722e-05}
{"train_loss": 0.14265894889831543, "global_step": 131121, "epoch": 1473, "lr": 5.2811977294799543e-05}
{"train_loss": 0.11220379918813705, "global_step": 131122, "epoch": 1473, "lr": 5.281139847555382e-05}
{"train_loss": 0.10341418534517288, "global_step": 131123, "epoch": 1473, "lr": 5.281081965593014e-05}
{"train_loss": 0.15869425237178802, "global_step": 131124, "epoch": 1473, "lr": 5.281024083592859e-05}
{"train_loss": 0.08462575823068619, "global_step": 131125, "epoch": 1473, "lr": 5.2809662015549234e-05}
{"train_loss": 0.2024693638086319, "global_step": 131126, "epoch": 1473, "lr": 5.280908319479214e-05}
{"train_loss": 0.1277047097682953, "global_step": 131127, "epoch": 1473, "lr": 5.280850437365741e-05}
{"train_loss": 0.1546725183725357, "global_step": 131128, "epoch": 1473, "lr": 5.280792555214511e-05}
{"train_loss": 0.13404513895511627, "global_step": 131129, "epoch": 1473, "lr": 5.280734673025531e-05}
{"train_loss": 0.09885760396718979, "global_step": 131130, "epoch": 1473, "lr": 5.280676790798811e-05}
{"train_loss": 0.13507841527462006, "global_step": 131131, "epoch": 1473, "lr": 5.280618908534357e-05}
{"train_loss": 0.01996186189353466, "global_step": 131132, "epoch": 1473, "lr": 5.280561026232177e-05}
{"train_loss": 0.1010311171412468, "global_step": 131133, "epoch": 1473, "lr": 5.2805031438922795e-05}
{"train_loss": 0.1849297732114792, "global_step": 131134, "epoch": 1473, "lr": 5.2804452615146715e-05}
{"train_loss": 0.09143102169036865, "global_step": 131135, "epoch": 1473, "lr": 5.280387379099361e-05}
{"train_loss": 0.14785757660865784, "global_step": 131136, "epoch": 1473, "lr": 5.2803294966463556e-05}
{"train_loss": 0.131523996591568, "global_step": 131137, "epoch": 1473, "lr": 5.280271614155663e-05}
{"train_loss": 0.1621953547000885, "global_step": 131138, "epoch": 1473, "lr": 5.2802137316272916e-05}
{"train_loss": 0.11200869828462601, "global_step": 131139, "epoch": 1473, "lr": 5.280155849061249e-05}
{"train_loss": 0.10298186540603638, "global_step": 131140, "epoch": 1473, "lr": 5.280097966457543e-05}
{"train_loss": 0.10014636069536209, "global_step": 131141, "epoch": 1473, "lr": 5.280040083816181e-05}
{"train_loss": 0.12348021566867828, "global_step": 131142, "epoch": 1473, "lr": 5.279982201137171e-05}
{"train_loss": 0.10524909198284149, "global_step": 131143, "epoch": 1473, "lr": 5.2799243184205216e-05}
{"train_loss": 0.09461195766925812, "global_step": 131144, "epoch": 1473, "lr": 5.279866435666239e-05}
{"train_loss": 0.10430030524730682, "global_step": 131145, "epoch": 1473, "lr": 5.279808552874331e-05}
{"train_loss": 0.13003750145435333, "global_step": 131146, "epoch": 1473, "lr": 5.2797506700448075e-05}
{"train_loss": 0.13837429881095886, "global_step": 131147, "epoch": 1473, "lr": 5.279692787177675e-05}
{"train_loss": 0.1428881585597992, "global_step": 131148, "epoch": 1473, "lr": 5.279634904272941e-05}
{"train_loss": 0.1284288614988327, "global_step": 131149, "epoch": 1473, "lr": 5.279577021330613e-05}
{"train_loss": 0.12943659722805023, "global_step": 131150, "epoch": 1473, "lr": 5.2795191383506995e-05}
{"train_loss": 0.11070531606674194, "global_step": 131151, "epoch": 1473, "lr": 5.2794612553332076e-05}
{"train_loss": 0.11732293665409088, "global_step": 131152, "epoch": 1473, "lr": 5.2794033722781454e-05}
{"train_loss": 0.18337690830230713, "global_step": 131153, "epoch": 1473, "lr": 5.279345489185522e-05}
{"train_loss": 0.14869830012321472, "global_step": 131154, "epoch": 1473, "lr": 5.279287606055343e-05}
{"train_loss": 0.08778185397386551, "global_step": 131155, "epoch": 1473, "lr": 5.2792297228876174e-05}
{"train_loss": 0.12449176609516144, "global_step": 131156, "epoch": 1473, "lr": 5.279171839682353e-05}
{"train_loss": 0.07459486275911331, "global_step": 131157, "epoch": 1473, "lr": 5.279113956439557e-05}
{"train_loss": 0.14241361618041992, "global_step": 131158, "epoch": 1473, "lr": 5.2790560731592385e-05}
{"train_loss": 0.1469264030456543, "global_step": 131159, "epoch": 1473, "lr": 5.278998189841404e-05}
{"train_loss": 0.17439697682857513, "global_step": 131160, "epoch": 1473, "lr": 5.278940306486061e-05}
{"train_loss": 0.14859186112880707, "global_step": 131161, "epoch": 1473, "lr": 5.2788824230932173e-05}
{"train_loss": 0.1499355137348175, "global_step": 131162, "epoch": 1473, "lr": 5.278824539662883e-05}
{"train_loss": 0.20248478651046753, "global_step": 131163, "epoch": 1473, "lr": 5.2787666561950636e-05}
{"train_loss": 0.1190829873085022, "global_step": 131164, "epoch": 1473, "lr": 5.278708772689767e-05}
{"train_loss": 0.12243469059467316, "global_step": 131165, "epoch": 1473, "lr": 5.278650889147001e-05}
{"train_loss": 0.14101479947566986, "global_step": 131166, "epoch": 1473, "lr": 5.2785930055667745e-05}
{"train_loss": 0.09366577863693237, "global_step": 131167, "epoch": 1473, "lr": 5.278535121949094e-05}
{"train_loss": 0.1527296006679535, "global_step": 131168, "epoch": 1473, "lr": 5.2784772382939685e-05}
{"train_loss": 0.14251117408275604, "global_step": 131169, "epoch": 1473, "lr": 5.2784193546014036e-05}
{"train_loss": 0.11647366732358932, "global_step": 131170, "epoch": 1473, "lr": 5.27836147087141e-05}
{"train_loss": 0.11099369078874588, "global_step": 131171, "epoch": 1473, "lr": 5.278303587103994e-05}
{"train_loss": 0.14549745619297028, "global_step": 131172, "epoch": 1473, "lr": 5.2782457032991636e-05}
{"train_loss": 0.11076898127794266, "global_step": 131173, "epoch": 1473, "lr": 5.278187819456926e-05}
{"train_loss": 0.1394912749528885, "global_step": 131174, "epoch": 1473, "lr": 5.278129935577289e-05}
{"train_loss": 0.14190906286239624, "global_step": 131175, "epoch": 1473, "lr": 5.278072051660262e-05}
{"train_loss": 0.1013471931219101, "global_step": 131176, "epoch": 1473, "lr": 5.278014167705851e-05}
{"train_loss": 0.12568756937980652, "global_step": 131177, "epoch": 1473, "lr": 5.277956283714065e-05}
{"train_loss": 0.15450257062911987, "global_step": 131178, "epoch": 1473, "lr": 5.2778983996849105e-05}
{"train_loss": 0.0645902007818222, "global_step": 131179, "epoch": 1473, "lr": 5.277840515618396e-05}
{"train_loss": 0.08912916481494904, "global_step": 131180, "epoch": 1473, "lr": 5.2777826315145295e-05}
{"train_loss": 0.11393523961305618, "global_step": 131181, "epoch": 1473, "lr": 5.277724747373318e-05}
{"train_loss": 0.1416962891817093, "global_step": 131182, "epoch": 1473, "lr": 5.27766686319477e-05}
{"train_loss": 0.1380845457315445, "global_step": 131183, "epoch": 1473, "lr": 5.277608978978893e-05}
{"train_loss": 0.09866440296173096, "global_step": 131184, "epoch": 1473, "lr": 5.2775510947256946e-05}
{"train_loss": 0.1250271135627219, "global_step": 131185, "epoch": 1473, "lr": 5.277493210435184e-05, "val_loss": 5.622585773468018}
{"train_loss": 0.1372317522764206, "global_step": 131186, "epoch": 1474, "lr": 5.277435326107367e-05}
{"train_loss": 0.099085234105587, "global_step": 131187, "epoch": 1474, "lr": 5.2773774417422514e-05}
{"train_loss": 0.046320196241140366, "global_step": 131188, "epoch": 1474, "lr": 5.277319557339847e-05}
{"train_loss": 0.12195462733507156, "global_step": 131189, "epoch": 1474, "lr": 5.2772616729001604e-05}
{"train_loss": 0.116172194480896, "global_step": 131190, "epoch": 1474, "lr": 5.277203788423198e-05}
{"train_loss": 0.1486855149269104, "global_step": 131191, "epoch": 1474, "lr": 5.277145903908971e-05}
{"train_loss": 0.12152158468961716, "global_step": 131192, "epoch": 1474, "lr": 5.277088019357484e-05}
{"train_loss": 0.14868195354938507, "global_step": 131193, "epoch": 1474, "lr": 5.2770301347687456e-05}
{"train_loss": 0.08282355964183807, "global_step": 131194, "epoch": 1474, "lr": 5.2769722501427646e-05}
{"train_loss": 0.12059438228607178, "global_step": 131195, "epoch": 1474, "lr": 5.276914365479547e-05}
{"train_loss": 0.1293085217475891, "global_step": 131196, "epoch": 1474, "lr": 5.276856480779102e-05}
{"train_loss": 0.09969047456979752, "global_step": 131197, "epoch": 1474, "lr": 5.276798596041438e-05}
{"train_loss": 0.12967953085899353, "global_step": 131198, "epoch": 1474, "lr": 5.276740711266561e-05}
{"train_loss": 0.20587332546710968, "global_step": 131199, "epoch": 1474, "lr": 5.2766828264544796e-05}
{"train_loss": 0.08177870512008667, "global_step": 131200, "epoch": 1474, "lr": 5.2766249416052017e-05}
{"train_loss": 0.10755996406078339, "global_step": 131201, "epoch": 1474, "lr": 5.2765670567187356e-05}
{"train_loss": 0.09087174385786057, "global_step": 131202, "epoch": 1474, "lr": 5.276509171795087e-05}
{"train_loss": 0.08878026902675629, "global_step": 131203, "epoch": 1474, "lr": 5.2764512868342655e-05}
{"train_loss": 0.10870449244976044, "global_step": 131204, "epoch": 1474, "lr": 5.276393401836279e-05}
{"train_loss": 0.17945721745491028, "global_step": 131205, "epoch": 1474, "lr": 5.2763355168011344e-05}
{"train_loss": 0.16843532025814056, "global_step": 131206, "epoch": 1474, "lr": 5.27627763172884e-05}
{"train_loss": 0.10490944236516953, "global_step": 131207, "epoch": 1474, "lr": 5.276219746619404e-05}
{"train_loss": 0.1379779726266861, "global_step": 131208, "epoch": 1474, "lr": 5.276161861472832e-05}
{"train_loss": 0.10309860855340958, "global_step": 131209, "epoch": 1474, "lr": 5.276103976289135e-05}
{"train_loss": 0.12602710723876953, "global_step": 131210, "epoch": 1474, "lr": 5.27604609106832e-05}
{"train_loss": 0.08550646156072617, "global_step": 131211, "epoch": 1474, "lr": 5.275988205810391e-05}
{"train_loss": 0.10904352366924286, "global_step": 131212, "epoch": 1474, "lr": 5.27593032051536e-05}
{"train_loss": 0.09049678593873978, "global_step": 131213, "epoch": 1474, "lr": 5.275872435183236e-05}
{"train_loss": 0.09237357974052429, "global_step": 131214, "epoch": 1474, "lr": 5.275814549814021e-05}
{"train_loss": 0.1285264492034912, "global_step": 131215, "epoch": 1474, "lr": 5.2757566644077285e-05}
{"train_loss": 0.12265443056821823, "global_step": 131216, "epoch": 1474, "lr": 5.275698778964362e-05}
{"train_loss": 0.12580490112304688, "global_step": 131217, "epoch": 1474, "lr": 5.275640893483932e-05}
{"train_loss": 0.10532553493976593, "global_step": 131218, "epoch": 1474, "lr": 5.275583007966446e-05}
{"train_loss": 0.13701866567134857, "global_step": 131219, "epoch": 1474, "lr": 5.2755251224119096e-05}
{"train_loss": 0.09194846451282501, "global_step": 131220, "epoch": 1474, "lr": 5.275467236820334e-05}
{"train_loss": 0.07633232325315475, "global_step": 131221, "epoch": 1474, "lr": 5.2754093511917234e-05}
{"train_loss": 0.16764773428440094, "global_step": 131222, "epoch": 1474, "lr": 5.275351465526088e-05}
{"train_loss": 0.18388350307941437, "global_step": 131223, "epoch": 1474, "lr": 5.275293579823436e-05}
{"train_loss": 0.15102975070476532, "global_step": 131224, "epoch": 1474, "lr": 5.275235694083773e-05}
{"train_loss": 0.12271580845117569, "global_step": 131225, "epoch": 1474, "lr": 5.275177808307109e-05}
{"train_loss": 0.08897244930267334, "global_step": 131226, "epoch": 1474, "lr": 5.27511992249345e-05}
{"train_loss": 0.1420736014842987, "global_step": 131227, "epoch": 1474, "lr": 5.2750620366428036e-05}
{"train_loss": 0.1641087830066681, "global_step": 131228, "epoch": 1474, "lr": 5.2750041507551805e-05}
{"train_loss": 0.12619830667972565, "global_step": 131229, "epoch": 1474, "lr": 5.274946264830585e-05}
{"train_loss": 0.10510313510894775, "global_step": 131230, "epoch": 1474, "lr": 5.274888378869025e-05}
{"train_loss": 0.11485625803470612, "global_step": 131231, "epoch": 1474, "lr": 5.274830492870512e-05}
{"train_loss": 0.11626310646533966, "global_step": 131232, "epoch": 1474, "lr": 5.27477260683505e-05}
{"train_loss": 0.08292564004659653, "global_step": 131233, "epoch": 1474, "lr": 5.27471472076265e-05}
{"train_loss": 0.09833851456642151, "global_step": 131234, "epoch": 1474, "lr": 5.274656834653318e-05}
{"train_loss": 0.08294938504695892, "global_step": 131235, "epoch": 1474, "lr": 5.274598948507059e-05}
{"train_loss": 0.13737080991268158, "global_step": 131236, "epoch": 1474, "lr": 5.274541062323885e-05}
{"train_loss": 0.08335456252098083, "global_step": 131237, "epoch": 1474, "lr": 5.2744831761038024e-05}
{"train_loss": 0.04031623527407646, "global_step": 131238, "epoch": 1474, "lr": 5.2744252898468194e-05}
{"train_loss": 0.1238432377576828, "global_step": 131239, "epoch": 1474, "lr": 5.2743674035529424e-05}
{"train_loss": 0.09282826632261276, "global_step": 131240, "epoch": 1474, "lr": 5.27430951722218e-05}
{"train_loss": 0.1280945986509323, "global_step": 131241, "epoch": 1474, "lr": 5.274251630854541e-05}
{"train_loss": 0.11173515766859055, "global_step": 131242, "epoch": 1474, "lr": 5.274193744450031e-05}
{"train_loss": 0.11730658262968063, "global_step": 131243, "epoch": 1474, "lr": 5.27413585800866e-05}
{"train_loss": 0.12858359515666962, "global_step": 131244, "epoch": 1474, "lr": 5.2740779715304346e-05}
{"train_loss": 0.10250800102949142, "global_step": 131245, "epoch": 1474, "lr": 5.274020085015362e-05}
{"train_loss": 0.17490126192569733, "global_step": 131246, "epoch": 1474, "lr": 5.273962198463451e-05}
{"train_loss": 0.12464676797389984, "global_step": 131247, "epoch": 1474, "lr": 5.27390431187471e-05}
{"train_loss": 0.11497458070516586, "global_step": 131248, "epoch": 1474, "lr": 5.273846425249144e-05}
{"train_loss": 0.13732902705669403, "global_step": 131249, "epoch": 1474, "lr": 5.2737885385867636e-05}
{"train_loss": 0.12495569884777069, "global_step": 131250, "epoch": 1474, "lr": 5.2737306518875765e-05}
{"train_loss": 0.16637249290943146, "global_step": 131251, "epoch": 1474, "lr": 5.2736727651515896e-05}
{"train_loss": 0.1368250846862793, "global_step": 131252, "epoch": 1474, "lr": 5.273614878378811e-05}
{"train_loss": 0.1479244977235794, "global_step": 131253, "epoch": 1474, "lr": 5.2735569915692475e-05}
{"train_loss": 0.12594212591648102, "global_step": 131254, "epoch": 1474, "lr": 5.273499104722908e-05}
{"train_loss": 0.12609148025512695, "global_step": 131255, "epoch": 1474, "lr": 5.273441217839799e-05}
{"train_loss": 0.17351968586444855, "global_step": 131256, "epoch": 1474, "lr": 5.2733833309199284e-05}
{"train_loss": 0.10936053842306137, "global_step": 131257, "epoch": 1474, "lr": 5.273325443963307e-05}
{"train_loss": 0.08533494174480438, "global_step": 131258, "epoch": 1474, "lr": 5.2732675569699394e-05}
{"train_loss": 0.09281391650438309, "global_step": 131259, "epoch": 1474, "lr": 5.2732096699398335e-05}
{"train_loss": 0.1005573645234108, "global_step": 131260, "epoch": 1474, "lr": 5.273151782872999e-05}
{"train_loss": 0.18221749365329742, "global_step": 131261, "epoch": 1474, "lr": 5.273093895769442e-05}
{"train_loss": 0.10575621575117111, "global_step": 131262, "epoch": 1474, "lr": 5.273036008629171e-05}
{"train_loss": 0.11043409258127213, "global_step": 131263, "epoch": 1474, "lr": 5.2729781214521935e-05}
{"train_loss": 0.11359953880310059, "global_step": 131264, "epoch": 1474, "lr": 5.272920234238517e-05}
{"train_loss": 0.11809831112623215, "global_step": 131265, "epoch": 1474, "lr": 5.272862346988151e-05}
{"train_loss": 0.10229663550853729, "global_step": 131266, "epoch": 1474, "lr": 5.272804459701101e-05}
{"train_loss": 0.16652311384677887, "global_step": 131267, "epoch": 1474, "lr": 5.2727465723773753e-05}
{"train_loss": 0.05903804674744606, "global_step": 131268, "epoch": 1474, "lr": 5.272688685016983e-05}
{"train_loss": 0.1156201884150505, "global_step": 131269, "epoch": 1474, "lr": 5.272630797619931e-05}
{"train_loss": 0.11850103735923767, "global_step": 131270, "epoch": 1474, "lr": 5.2725729101862266e-05}
{"train_loss": 0.15314124524593353, "global_step": 131271, "epoch": 1474, "lr": 5.2725150227158805e-05}
{"train_loss": 0.12177733331918716, "global_step": 131272, "epoch": 1474, "lr": 5.272457135208896e-05}
{"train_loss": 0.14847439527511597, "global_step": 131273, "epoch": 1474, "lr": 5.2723992476652826e-05}
{"train_loss": 0.1194495220114006, "global_step": 131274, "epoch": 1474, "lr": 5.272341360085049e-05, "val_loss": 5.857463359832764}
{"train_loss": 0.15494920313358307, "global_step": 131275, "epoch": 1475, "lr": 5.272283472468202e-05}
{"train_loss": 0.12287425994873047, "global_step": 131276, "epoch": 1475, "lr": 5.2722255848147506e-05}
{"train_loss": 0.13157576322555542, "global_step": 131277, "epoch": 1475, "lr": 5.272167697124702e-05}
{"train_loss": 0.14752355217933655, "global_step": 131278, "epoch": 1475, "lr": 5.272109809398062e-05}
{"train_loss": 0.08151902258396149, "global_step": 131279, "epoch": 1475, "lr": 5.272051921634842e-05}
{"train_loss": 0.17830577492713928, "global_step": 131280, "epoch": 1475, "lr": 5.271994033835046e-05}
{"train_loss": 0.14538705348968506, "global_step": 131281, "epoch": 1475, "lr": 5.271936145998685e-05}
{"train_loss": 0.17080886662006378, "global_step": 131282, "epoch": 1475, "lr": 5.2718782581257656e-05}
{"train_loss": 0.174392968416214, "global_step": 131283, "epoch": 1475, "lr": 5.271820370216295e-05}
{"train_loss": 0.12199283391237259, "global_step": 131284, "epoch": 1475, "lr": 5.271762482270282e-05}
{"train_loss": 0.13086682558059692, "global_step": 131285, "epoch": 1475, "lr": 5.271704594287733e-05}
{"train_loss": 0.08736904710531235, "global_step": 131286, "epoch": 1475, "lr": 5.271646706268657e-05}
{"train_loss": 0.10501976311206818, "global_step": 131287, "epoch": 1475, "lr": 5.271588818213061e-05}
{"train_loss": 0.1411614865064621, "global_step": 131288, "epoch": 1475, "lr": 5.2715309301209535e-05}
{"train_loss": 0.11997251957654953, "global_step": 131289, "epoch": 1475, "lr": 5.271473041992343e-05}
{"train_loss": 0.15576469898223877, "global_step": 131290, "epoch": 1475, "lr": 5.271415153827236e-05}
{"train_loss": 0.14833927154541016, "global_step": 131291, "epoch": 1475, "lr": 5.271357265625641e-05}
{"train_loss": 0.09762485325336456, "global_step": 131292, "epoch": 1475, "lr": 5.271299377387564e-05}
{"train_loss": 0.11635191738605499, "global_step": 131293, "epoch": 1475, "lr": 5.2712414891130145e-05}
{"train_loss": 0.10667097568511963, "global_step": 131294, "epoch": 1475, "lr": 5.271183600802e-05}
{"train_loss": 0.09966260939836502, "global_step": 131295, "epoch": 1475, "lr": 5.271125712454528e-05}
{"train_loss": 0.09939530491828918, "global_step": 131296, "epoch": 1475, "lr": 5.2710678240706065e-05}
{"train_loss": 0.17511507868766785, "global_step": 131297, "epoch": 1475, "lr": 5.271009935650243e-05}
{"train_loss": 0.11018393188714981, "global_step": 131298, "epoch": 1475, "lr": 5.270952047193446e-05}
{"train_loss": 0.17204789817333221, "global_step": 131299, "epoch": 1475, "lr": 5.2708941587002226e-05}
{"train_loss": 0.10902712494134903, "global_step": 131300, "epoch": 1475, "lr": 5.2708362701705805e-05}
{"train_loss": 0.20082977414131165, "global_step": 131301, "epoch": 1475, "lr": 5.270778381604529e-05}
{"train_loss": 0.1132773831486702, "global_step": 131302, "epoch": 1475, "lr": 5.270720493002073e-05}
{"train_loss": 0.16600437462329865, "global_step": 131303, "epoch": 1475, "lr": 5.270662604363221e-05}
{"train_loss": 0.0962163433432579, "global_step": 131304, "epoch": 1475, "lr": 5.2706047156879835e-05}
{"train_loss": 0.165465846657753, "global_step": 131305, "epoch": 1475, "lr": 5.270546826976367e-05}
{"train_loss": 0.20445957779884338, "global_step": 131306, "epoch": 1475, "lr": 5.270488938228377e-05}
{"train_loss": 0.15037801861763, "global_step": 131307, "epoch": 1475, "lr": 5.270431049444025e-05}
{"train_loss": 0.08565559983253479, "global_step": 131308, "epoch": 1475, "lr": 5.2703731606233156e-05}
{"train_loss": 0.15275302529335022, "global_step": 131309, "epoch": 1475, "lr": 5.2703152717662586e-05}
{"train_loss": 0.08690522611141205, "global_step": 131310, "epoch": 1475, "lr": 5.270257382872862e-05}
{"train_loss": 0.1995048224925995, "global_step": 131311, "epoch": 1475, "lr": 5.270199493943131e-05}
{"train_loss": 0.17630130052566528, "global_step": 131312, "epoch": 1475, "lr": 5.270141604977075e-05}
{"train_loss": 0.17589987814426422, "global_step": 131313, "epoch": 1475, "lr": 5.2700837159747015e-05}
{"train_loss": 0.1088358610868454, "global_step": 131314, "epoch": 1475, "lr": 5.270025826936019e-05}
{"train_loss": 0.190602645277977, "global_step": 131315, "epoch": 1475, "lr": 5.269967937861035e-05}
{"train_loss": 0.1374264657497406, "global_step": 131316, "epoch": 1475, "lr": 5.2699100487497566e-05}
{"train_loss": 0.17351582646369934, "global_step": 131317, "epoch": 1475, "lr": 5.269852159602193e-05}
{"train_loss": 0.155253067612648, "global_step": 131318, "epoch": 1475, "lr": 5.26979427041835e-05}
{"train_loss": 0.13146507740020752, "global_step": 131319, "epoch": 1475, "lr": 5.269736381198237e-05}
{"train_loss": 0.12899938225746155, "global_step": 131320, "epoch": 1475, "lr": 5.2696784919418604e-05}
{"train_loss": 0.10540233552455902, "global_step": 131321, "epoch": 1475, "lr": 5.269620602649229e-05}
{"train_loss": 0.11205736547708511, "global_step": 131322, "epoch": 1475, "lr": 5.2695627133203505e-05}
{"train_loss": 0.09911743551492691, "global_step": 131323, "epoch": 1475, "lr": 5.2695048239552336e-05}
{"train_loss": 0.12881137430667877, "global_step": 131324, "epoch": 1475, "lr": 5.269446934553884e-05}
{"train_loss": 0.14543993771076202, "global_step": 131325, "epoch": 1475, "lr": 5.269389045116312e-05}
{"train_loss": 0.1431846022605896, "global_step": 131326, "epoch": 1475, "lr": 5.269331155642523e-05}
{"train_loss": 0.10336516797542572, "global_step": 131327, "epoch": 1475, "lr": 5.269273266132525e-05}
{"train_loss": 0.11418795585632324, "global_step": 131328, "epoch": 1475, "lr": 5.269215376586327e-05}
{"train_loss": 0.08905181288719177, "global_step": 131329, "epoch": 1475, "lr": 5.269157487003936e-05}
{"train_loss": 0.12240178883075714, "global_step": 131330, "epoch": 1475, "lr": 5.269099597385362e-05}
{"train_loss": 0.10319843888282776, "global_step": 131331, "epoch": 1475, "lr": 5.26904170773061e-05}
{"train_loss": 0.11922004073858261, "global_step": 131332, "epoch": 1475, "lr": 5.268983818039688e-05}
{"train_loss": 0.1446358859539032, "global_step": 131333, "epoch": 1475, "lr": 5.268925928312603e-05}
{"train_loss": 0.15127000212669373, "global_step": 131334, "epoch": 1475, "lr": 5.268868038549366e-05}
{"train_loss": 0.12722256779670715, "global_step": 131335, "epoch": 1475, "lr": 5.268810148749983e-05}
{"train_loss": 0.08328812569379807, "global_step": 131336, "epoch": 1475, "lr": 5.268752258914461e-05}
{"train_loss": 0.14192402362823486, "global_step": 131337, "epoch": 1475, "lr": 5.268694369042808e-05}
{"train_loss": 0.14951810240745544, "global_step": 131338, "epoch": 1475, "lr": 5.268636479135034e-05}
{"train_loss": 0.11511983722448349, "global_step": 131339, "epoch": 1475, "lr": 5.268578589191143e-05}
{"train_loss": 0.08316178619861603, "global_step": 131340, "epoch": 1475, "lr": 5.268520699211146e-05}
{"train_loss": 0.12006974220275879, "global_step": 131341, "epoch": 1475, "lr": 5.268462809195049e-05}
{"train_loss": 0.16120855510234833, "global_step": 131342, "epoch": 1475, "lr": 5.268404919142862e-05}
{"train_loss": 0.14617787301540375, "global_step": 131343, "epoch": 1475, "lr": 5.26834702905459e-05}
{"train_loss": 0.10732617974281311, "global_step": 131344, "epoch": 1475, "lr": 5.2682891389302426e-05}
{"train_loss": 0.17032024264335632, "global_step": 131345, "epoch": 1475, "lr": 5.268231248769827e-05}
{"train_loss": 0.1526165008544922, "global_step": 131346, "epoch": 1475, "lr": 5.2681733585733506e-05}
{"train_loss": 0.10734347254037857, "global_step": 131347, "epoch": 1475, "lr": 5.268115468340822e-05}
{"train_loss": 0.15705884993076324, "global_step": 131348, "epoch": 1475, "lr": 5.268057578072249e-05}
{"train_loss": 0.10343261063098907, "global_step": 131349, "epoch": 1475, "lr": 5.267999687767638e-05}
{"train_loss": 0.1110302284359932, "global_step": 131350, "epoch": 1475, "lr": 5.2679417974269983e-05}
{"train_loss": 0.1281300038099289, "global_step": 131351, "epoch": 1475, "lr": 5.267883907050337e-05}
{"train_loss": 0.08460512012243271, "global_step": 131352, "epoch": 1475, "lr": 5.2678260166376614e-05}
{"train_loss": 0.16035865247249603, "global_step": 131353, "epoch": 1475, "lr": 5.26776812618898e-05}
{"train_loss": 0.11285622417926788, "global_step": 131354, "epoch": 1475, "lr": 5.2677102357043005e-05}
{"train_loss": 0.09901975095272064, "global_step": 131355, "epoch": 1475, "lr": 5.2676523451836315e-05}
{"train_loss": 0.0778714120388031, "global_step": 131356, "epoch": 1475, "lr": 5.2675944546269785e-05}
{"train_loss": 0.14846192300319672, "global_step": 131357, "epoch": 1475, "lr": 5.2675365640343496e-05}
{"train_loss": 0.11457108706235886, "global_step": 131358, "epoch": 1475, "lr": 5.267478673405756e-05}
{"train_loss": 0.10993966460227966, "global_step": 131359, "epoch": 1475, "lr": 5.267420782741202e-05}
{"train_loss": 0.14194512367248535, "global_step": 131360, "epoch": 1475, "lr": 5.267362892040698e-05}
{"train_loss": 0.10247392952442169, "global_step": 131361, "epoch": 1475, "lr": 5.267305001304249e-05}
{"train_loss": 0.1467198133468628, "global_step": 131362, "epoch": 1475, "lr": 5.267247110531864e-05}
{"train_loss": 0.13103565707635345, "global_step": 131363, "epoch": 1475, "lr": 5.2671892197235516e-05, "val_loss": 5.677273273468018, "train_action_mse_error": 7.313180923461914}
{"train_loss": 0.1367173045873642, "global_step": 131364, "epoch": 1476, "lr": 5.267131328879319e-05}
{"train_loss": 0.1694309115409851, "global_step": 131365, "epoch": 1476, "lr": 5.2670734379991737e-05}
{"train_loss": 0.08986613899469376, "global_step": 131366, "epoch": 1476, "lr": 5.2670155470831225e-05}
{"train_loss": 0.11775366961956024, "global_step": 131367, "epoch": 1476, "lr": 5.266957656131176e-05}
{"train_loss": 0.16093331575393677, "global_step": 131368, "epoch": 1476, "lr": 5.266899765143341e-05}
{"train_loss": 0.20212794840335846, "global_step": 131369, "epoch": 1476, "lr": 5.266841874119622e-05}
{"train_loss": 0.0803980752825737, "global_step": 131370, "epoch": 1476, "lr": 5.266783983060031e-05}
{"train_loss": 0.1500042974948883, "global_step": 131371, "epoch": 1476, "lr": 5.2667260919645735e-05}
{"train_loss": 0.19287700951099396, "global_step": 131372, "epoch": 1476, "lr": 5.266668200833258e-05}
{"train_loss": 0.11449585109949112, "global_step": 131373, "epoch": 1476, "lr": 5.266610309666092e-05}
{"train_loss": 0.18956568837165833, "global_step": 131374, "epoch": 1476, "lr": 5.2665524184630846e-05}
{"train_loss": 0.13751670718193054, "global_step": 131375, "epoch": 1476, "lr": 5.266494527224241e-05}
{"train_loss": 0.09650663286447525, "global_step": 131376, "epoch": 1476, "lr": 5.26643663594957e-05}
{"train_loss": 0.12350806593894958, "global_step": 131377, "epoch": 1476, "lr": 5.266378744639081e-05}
{"train_loss": 0.09738214313983917, "global_step": 131378, "epoch": 1476, "lr": 5.266320853292781e-05}
{"train_loss": 0.10777506977319717, "global_step": 131379, "epoch": 1476, "lr": 5.266262961910676e-05}
{"train_loss": 0.08401964604854584, "global_step": 131380, "epoch": 1476, "lr": 5.266205070492777e-05}
{"train_loss": 0.15386201441287994, "global_step": 131381, "epoch": 1476, "lr": 5.266147179039089e-05}
{"train_loss": 0.10375867038965225, "global_step": 131382, "epoch": 1476, "lr": 5.26608928754962e-05}
{"train_loss": 0.10561444610357285, "global_step": 131383, "epoch": 1476, "lr": 5.2660313960243803e-05}
{"train_loss": 0.08417047560214996, "global_step": 131384, "epoch": 1476, "lr": 5.265973504463375e-05}
{"train_loss": 0.1289416402578354, "global_step": 131385, "epoch": 1476, "lr": 5.265915612866612e-05}
{"train_loss": 0.09616757184267044, "global_step": 131386, "epoch": 1476, "lr": 5.265857721234101e-05}
{"train_loss": 0.19663932919502258, "global_step": 131387, "epoch": 1476, "lr": 5.265799829565848e-05}
{"train_loss": 0.12708255648612976, "global_step": 131388, "epoch": 1476, "lr": 5.265741937861863e-05}
{"train_loss": 0.13411447405815125, "global_step": 131389, "epoch": 1476, "lr": 5.265684046122151e-05}
{"train_loss": 0.06481404602527618, "global_step": 131390, "epoch": 1476, "lr": 5.265626154346722e-05}
{"train_loss": 0.16795818507671356, "global_step": 131391, "epoch": 1476, "lr": 5.2655682625355814e-05}
{"train_loss": 0.14830747246742249, "global_step": 131392, "epoch": 1476, "lr": 5.265510370688739e-05}
{"train_loss": 0.10520200431346893, "global_step": 131393, "epoch": 1476, "lr": 5.2654524788062006e-05}
{"train_loss": 0.14184030890464783, "global_step": 131394, "epoch": 1476, "lr": 5.2653945868879774e-05}
{"train_loss": 0.11355868726968765, "global_step": 131395, "epoch": 1476, "lr": 5.2653366949340744e-05}
{"train_loss": 0.17835286259651184, "global_step": 131396, "epoch": 1476, "lr": 5.2652788029445e-05}
{"train_loss": 0.19396385550498962, "global_step": 131397, "epoch": 1476, "lr": 5.265220910919262e-05}
{"train_loss": 0.08442020416259766, "global_step": 131398, "epoch": 1476, "lr": 5.265163018858368e-05}
{"train_loss": 0.15623322129249573, "global_step": 131399, "epoch": 1476, "lr": 5.2651051267618265e-05}
{"train_loss": 0.08963577449321747, "global_step": 131400, "epoch": 1476, "lr": 5.2650472346296454e-05}
{"train_loss": 0.14320459961891174, "global_step": 131401, "epoch": 1476, "lr": 5.2649893424618314e-05}
{"train_loss": 0.10970439016819, "global_step": 131402, "epoch": 1476, "lr": 5.2649314502583925e-05}
{"train_loss": 0.15757137537002563, "global_step": 131403, "epoch": 1476, "lr": 5.264873558019336e-05}
{"train_loss": 0.12796874344348907, "global_step": 131404, "epoch": 1476, "lr": 5.264815665744672e-05}
{"train_loss": 0.17244073748588562, "global_step": 131405, "epoch": 1476, "lr": 5.264757773434407e-05}
{"train_loss": 0.10283161699771881, "global_step": 131406, "epoch": 1476, "lr": 5.264699881088547e-05}
{"train_loss": 0.1321277916431427, "global_step": 131407, "epoch": 1476, "lr": 5.264641988707104e-05}
{"train_loss": 0.11353928595781326, "global_step": 131408, "epoch": 1476, "lr": 5.264584096290082e-05}
{"train_loss": 0.16840016841888428, "global_step": 131409, "epoch": 1476, "lr": 5.264526203837489e-05}
{"train_loss": 0.11577729135751724, "global_step": 131410, "epoch": 1476, "lr": 5.264468311349333e-05}
{"train_loss": 0.14110660552978516, "global_step": 131411, "epoch": 1476, "lr": 5.264410418825623e-05}
{"train_loss": 0.15589794516563416, "global_step": 131412, "epoch": 1476, "lr": 5.2643525262663663e-05}
{"train_loss": 0.13178972899913788, "global_step": 131413, "epoch": 1476, "lr": 5.2642946336715705e-05}
{"train_loss": 0.13257654011249542, "global_step": 131414, "epoch": 1476, "lr": 5.264236741041244e-05}
{"train_loss": 0.1522182822227478, "global_step": 131415, "epoch": 1476, "lr": 5.264178848375394e-05}
{"train_loss": 0.2175203114748001, "global_step": 131416, "epoch": 1476, "lr": 5.264120955674028e-05}
{"train_loss": 0.13634496927261353, "global_step": 131417, "epoch": 1476, "lr": 5.264063062937155e-05}
{"train_loss": 0.12259379029273987, "global_step": 131418, "epoch": 1476, "lr": 5.264005170164781e-05}
{"train_loss": 0.13310480117797852, "global_step": 131419, "epoch": 1476, "lr": 5.2639472773569156e-05}
{"train_loss": 0.08105986565351486, "global_step": 131420, "epoch": 1476, "lr": 5.263889384513565e-05}
{"train_loss": 0.11917971819639206, "global_step": 131421, "epoch": 1476, "lr": 5.263831491634737e-05}
{"train_loss": 0.15366554260253906, "global_step": 131422, "epoch": 1476, "lr": 5.263773598720442e-05}
{"train_loss": 0.10610508918762207, "global_step": 131423, "epoch": 1476, "lr": 5.263715705770685e-05}
{"train_loss": 0.07505878061056137, "global_step": 131424, "epoch": 1476, "lr": 5.263657812785474e-05}
{"train_loss": 0.09898253530263901, "global_step": 131425, "epoch": 1476, "lr": 5.263599919764819e-05}
{"train_loss": 0.15560179948806763, "global_step": 131426, "epoch": 1476, "lr": 5.263542026708725e-05}
{"train_loss": 0.1547108292579651, "global_step": 131427, "epoch": 1476, "lr": 5.263484133617201e-05}
{"train_loss": 0.15554332733154297, "global_step": 131428, "epoch": 1476, "lr": 5.263426240490256e-05}
{"train_loss": 0.10822435468435287, "global_step": 131429, "epoch": 1476, "lr": 5.263368347327895e-05}
{"train_loss": 0.05713026225566864, "global_step": 131430, "epoch": 1476, "lr": 5.263310454130127e-05}
{"train_loss": 0.13685745000839233, "global_step": 131431, "epoch": 1476, "lr": 5.2632525608969604e-05}
{"train_loss": 0.10427972674369812, "global_step": 131432, "epoch": 1476, "lr": 5.263194667628404e-05}
{"train_loss": 0.1329471915960312, "global_step": 131433, "epoch": 1476, "lr": 5.2631367743244633e-05}
{"train_loss": 0.134560227394104, "global_step": 131434, "epoch": 1476, "lr": 5.263078880985147e-05}
{"train_loss": 0.11821231991052628, "global_step": 131435, "epoch": 1476, "lr": 5.263020987610463e-05}
{"train_loss": 0.12532202899456024, "global_step": 131436, "epoch": 1476, "lr": 5.2629630942004196e-05}
{"train_loss": 0.06908837705850601, "global_step": 131437, "epoch": 1476, "lr": 5.262905200755024e-05}
{"train_loss": 0.11278193444013596, "global_step": 131438, "epoch": 1476, "lr": 5.2628473072742836e-05}
{"train_loss": 0.09827068448066711, "global_step": 131439, "epoch": 1476, "lr": 5.2627894137582066e-05}
{"train_loss": 0.17756818234920502, "global_step": 131440, "epoch": 1476, "lr": 5.262731520206801e-05}
{"train_loss": 0.10259919613599777, "global_step": 131441, "epoch": 1476, "lr": 5.262673626620074e-05}
{"train_loss": 0.10486403107643127, "global_step": 131442, "epoch": 1476, "lr": 5.262615732998034e-05}
{"train_loss": 0.09220439195632935, "global_step": 131443, "epoch": 1476, "lr": 5.2625578393406885e-05}
{"train_loss": 0.14795753359794617, "global_step": 131444, "epoch": 1476, "lr": 5.262499945648045e-05}
{"train_loss": 0.08199477195739746, "global_step": 131445, "epoch": 1476, "lr": 5.262442051920112e-05}
{"train_loss": 0.10493920743465424, "global_step": 131446, "epoch": 1476, "lr": 5.262384158156898e-05}
{"train_loss": 0.1057940348982811, "global_step": 131447, "epoch": 1476, "lr": 5.262326264358408e-05}
{"train_loss": 0.08132427930831909, "global_step": 131448, "epoch": 1476, "lr": 5.26226837052465e-05}
{"train_loss": 0.146638423204422, "global_step": 131449, "epoch": 1476, "lr": 5.262210476655636e-05}
{"train_loss": 0.13290968537330627, "global_step": 131450, "epoch": 1476, "lr": 5.2621525827513696e-05}
{"train_loss": 0.11993024498224258, "global_step": 131451, "epoch": 1476, "lr": 5.262094688811861e-05}
{"train_loss": 0.126722439035271, "global_step": 131452, "epoch": 1476, "lr": 5.2620367948371154e-05, "val_loss": 5.708615303039551}
{"train_loss": 0.10602012276649475, "global_step": 131453, "epoch": 1477, "lr": 5.2619789008271436e-05}
{"train_loss": 0.1320696920156479, "global_step": 131454, "epoch": 1477, "lr": 5.2619210067819516e-05}
{"train_loss": 0.09255266934633255, "global_step": 131455, "epoch": 1477, "lr": 5.261863112701547e-05}
{"train_loss": 0.12482457607984543, "global_step": 131456, "epoch": 1477, "lr": 5.261805218585938e-05}
{"train_loss": 0.09068674594163895, "global_step": 131457, "epoch": 1477, "lr": 5.2617473244351335e-05}
{"train_loss": 0.11748021095991135, "global_step": 131458, "epoch": 1477, "lr": 5.2616894302491394e-05}
{"train_loss": 0.1583370417356491, "global_step": 131459, "epoch": 1477, "lr": 5.2616315360279644e-05}
{"train_loss": 0.1865529716014862, "global_step": 131460, "epoch": 1477, "lr": 5.2615736417716165e-05}
{"train_loss": 0.08233051002025604, "global_step": 131461, "epoch": 1477, "lr": 5.2615157474801025e-05}
{"train_loss": 0.14008978009223938, "global_step": 131462, "epoch": 1477, "lr": 5.261457853153432e-05}
{"train_loss": 0.143784761428833, "global_step": 131463, "epoch": 1477, "lr": 5.261399958791611e-05}
{"train_loss": 0.16502662003040314, "global_step": 131464, "epoch": 1477, "lr": 5.2613420643946486e-05}
{"train_loss": 0.08803936839103699, "global_step": 131465, "epoch": 1477, "lr": 5.261284169962553e-05}
{"train_loss": 0.11320260912179947, "global_step": 131466, "epoch": 1477, "lr": 5.261226275495328e-05}
{"train_loss": 0.12248082458972931, "global_step": 131467, "epoch": 1477, "lr": 5.261168380992986e-05}
{"train_loss": 0.10574740171432495, "global_step": 131468, "epoch": 1477, "lr": 5.2611104864555314e-05}
{"train_loss": 0.10070544481277466, "global_step": 131469, "epoch": 1477, "lr": 5.261052591882976e-05}
{"train_loss": 0.09896350651979446, "global_step": 131470, "epoch": 1477, "lr": 5.260994697275324e-05}
{"train_loss": 0.13285920023918152, "global_step": 131471, "epoch": 1477, "lr": 5.2609368026325856e-05}
{"train_loss": 0.08601539582014084, "global_step": 131472, "epoch": 1477, "lr": 5.260878907954766e-05}
{"train_loss": 0.13785529136657715, "global_step": 131473, "epoch": 1477, "lr": 5.260821013241876e-05}
{"train_loss": 0.12051749974489212, "global_step": 131474, "epoch": 1477, "lr": 5.26076311849392e-05}
{"train_loss": 0.14286120235919952, "global_step": 131475, "epoch": 1477, "lr": 5.260705223710909e-05}
{"train_loss": 0.14107675850391388, "global_step": 131476, "epoch": 1477, "lr": 5.260647328892848e-05}
{"train_loss": 0.15903937816619873, "global_step": 131477, "epoch": 1477, "lr": 5.260589434039748e-05}
{"train_loss": 0.20852112770080566, "global_step": 131478, "epoch": 1477, "lr": 5.260531539151613e-05}
{"train_loss": 0.1461486667394638, "global_step": 131479, "epoch": 1477, "lr": 5.260473644228454e-05}
{"train_loss": 0.1316739022731781, "global_step": 131480, "epoch": 1477, "lr": 5.2604157492702764e-05}
{"train_loss": 0.11295685917139053, "global_step": 131481, "epoch": 1477, "lr": 5.2603578542770905e-05}
{"train_loss": 0.1418876200914383, "global_step": 131482, "epoch": 1477, "lr": 5.260299959248901e-05}
{"train_loss": 0.1074075922369957, "global_step": 131483, "epoch": 1477, "lr": 5.2602420641857184e-05}
{"train_loss": 0.20838984847068787, "global_step": 131484, "epoch": 1477, "lr": 5.26018416908755e-05}
{"train_loss": 0.07743393629789352, "global_step": 131485, "epoch": 1477, "lr": 5.260126273954402e-05}
{"train_loss": 0.10733912885189056, "global_step": 131486, "epoch": 1477, "lr": 5.260068378786284e-05}
{"train_loss": 0.12102240324020386, "global_step": 131487, "epoch": 1477, "lr": 5.260010483583202e-05}
{"train_loss": 0.10242090374231339, "global_step": 131488, "epoch": 1477, "lr": 5.259952588345165e-05}
{"train_loss": 0.1331302672624588, "global_step": 131489, "epoch": 1477, "lr": 5.25989469307218e-05}
{"train_loss": 0.2418980598449707, "global_step": 131490, "epoch": 1477, "lr": 5.259836797764257e-05}
{"train_loss": 0.09238508343696594, "global_step": 131491, "epoch": 1477, "lr": 5.2597789024214005e-05}
{"train_loss": 0.10035529732704163, "global_step": 131492, "epoch": 1477, "lr": 5.259721007043621e-05}
{"train_loss": 0.13690023124217987, "global_step": 131493, "epoch": 1477, "lr": 5.259663111630925e-05}
{"train_loss": 0.12078359723091125, "global_step": 131494, "epoch": 1477, "lr": 5.25960521618332e-05}
{"train_loss": 0.137314110994339, "global_step": 131495, "epoch": 1477, "lr": 5.259547320700814e-05}
{"train_loss": 0.11233661323785782, "global_step": 131496, "epoch": 1477, "lr": 5.259489425183416e-05}
{"train_loss": 0.14119543135166168, "global_step": 131497, "epoch": 1477, "lr": 5.259431529631132e-05}
{"train_loss": 0.10813997685909271, "global_step": 131498, "epoch": 1477, "lr": 5.2593736340439715e-05}
{"train_loss": 0.0943116545677185, "global_step": 131499, "epoch": 1477, "lr": 5.25931573842194e-05}
{"train_loss": 0.12638674676418304, "global_step": 131500, "epoch": 1477, "lr": 5.259257842765048e-05}
{"train_loss": 0.1070622056722641, "global_step": 131501, "epoch": 1477, "lr": 5.259199947073301e-05}
{"train_loss": 0.11010045558214188, "global_step": 131502, "epoch": 1477, "lr": 5.2591420513467084e-05}
{"train_loss": 0.10061589628458023, "global_step": 131503, "epoch": 1477, "lr": 5.2590841555852765e-05}
{"train_loss": 0.13119570910930634, "global_step": 131504, "epoch": 1477, "lr": 5.259026259789015e-05}
{"train_loss": 0.1868801862001419, "global_step": 131505, "epoch": 1477, "lr": 5.25896836395793e-05}
{"train_loss": 0.19144362211227417, "global_step": 131506, "epoch": 1477, "lr": 5.25891046809203e-05}
{"train_loss": 0.10035940259695053, "global_step": 131507, "epoch": 1477, "lr": 5.258852572191322e-05}
{"train_loss": 0.1340116709470749, "global_step": 131508, "epoch": 1477, "lr": 5.2587946762558146e-05}
{"train_loss": 0.10889887809753418, "global_step": 131509, "epoch": 1477, "lr": 5.258736780285516e-05}
{"train_loss": 0.1251211166381836, "global_step": 131510, "epoch": 1477, "lr": 5.2586788842804325e-05}
{"train_loss": 0.08941090106964111, "global_step": 131511, "epoch": 1477, "lr": 5.2586209882405734e-05}
{"train_loss": 0.1506984680891037, "global_step": 131512, "epoch": 1477, "lr": 5.2585630921659466e-05}
{"train_loss": 0.10400410741567612, "global_step": 131513, "epoch": 1477, "lr": 5.2585051960565576e-05}
{"train_loss": 0.14860275387763977, "global_step": 131514, "epoch": 1477, "lr": 5.2584472999124165e-05}
{"train_loss": 0.04573279991745949, "global_step": 131515, "epoch": 1477, "lr": 5.2583894037335315e-05}
{"train_loss": 0.17759276926517487, "global_step": 131516, "epoch": 1477, "lr": 5.2583315075199066e-05}
{"train_loss": 0.06591612100601196, "global_step": 131517, "epoch": 1477, "lr": 5.258273611271555e-05}
{"train_loss": 0.12199621647596359, "global_step": 131518, "epoch": 1477, "lr": 5.258215714988479e-05}
{"train_loss": 0.1270667016506195, "global_step": 131519, "epoch": 1477, "lr": 5.25815781867069e-05}
{"train_loss": 0.10419877618551254, "global_step": 131520, "epoch": 1477, "lr": 5.258099922318196e-05}
{"train_loss": 0.08473856002092361, "global_step": 131521, "epoch": 1477, "lr": 5.2580420259310026e-05}
{"train_loss": 0.10684265941381454, "global_step": 131522, "epoch": 1477, "lr": 5.2579841295091195e-05}
{"train_loss": 0.0872756689786911, "global_step": 131523, "epoch": 1477, "lr": 5.2579262330525527e-05}
{"train_loss": 0.07455980032682419, "global_step": 131524, "epoch": 1477, "lr": 5.2578683365613104e-05}
{"train_loss": 0.12621010839939117, "global_step": 131525, "epoch": 1477, "lr": 5.2578104400354014e-05}
{"train_loss": 0.21265096962451935, "global_step": 131526, "epoch": 1477, "lr": 5.257752543474833e-05}
{"train_loss": 0.11596160382032394, "global_step": 131527, "epoch": 1477, "lr": 5.257694646879613e-05}
{"train_loss": 0.18659205734729767, "global_step": 131528, "epoch": 1477, "lr": 5.2576367502497495e-05}
{"train_loss": 0.09384763985872269, "global_step": 131529, "epoch": 1477, "lr": 5.257578853585249e-05}
{"train_loss": 0.11749465763568878, "global_step": 131530, "epoch": 1477, "lr": 5.2575209568861204e-05}
{"train_loss": 0.10781710594892502, "global_step": 131531, "epoch": 1477, "lr": 5.257463060152371e-05}
{"train_loss": 0.11882536113262177, "global_step": 131532, "epoch": 1477, "lr": 5.2574051633840095e-05}
{"train_loss": 0.19967688620090485, "global_step": 131533, "epoch": 1477, "lr": 5.257347266581042e-05}
{"train_loss": 0.09561579674482346, "global_step": 131534, "epoch": 1477, "lr": 5.2572893697434786e-05}
{"train_loss": 0.08454273641109467, "global_step": 131535, "epoch": 1477, "lr": 5.257231472871324e-05}
{"train_loss": 0.07488049566745758, "global_step": 131536, "epoch": 1477, "lr": 5.25717357596459e-05}
{"train_loss": 0.13139159977436066, "global_step": 131537, "epoch": 1477, "lr": 5.257115679023281e-05}
{"train_loss": 0.1331770122051239, "global_step": 131538, "epoch": 1477, "lr": 5.257057782047405e-05}
{"train_loss": 0.1158682107925415, "global_step": 131539, "epoch": 1477, "lr": 5.256999885036973e-05}
{"train_loss": 0.09171783179044724, "global_step": 131540, "epoch": 1477, "lr": 5.256941987991989e-05}
{"train_loss": 0.12318147245920107, "global_step": 131541, "epoch": 1477, "lr": 5.256884090912463e-05, "val_loss": 5.796719074249268}
{"train_loss": 0.14743418991565704, "global_step": 131542, "epoch": 1478, "lr": 5.256826193798401e-05}
{"train_loss": 0.18153223395347595, "global_step": 131543, "epoch": 1478, "lr": 5.256768296649812e-05}
{"train_loss": 0.17571792006492615, "global_step": 131544, "epoch": 1478, "lr": 5.2567103994667044e-05}
{"train_loss": 0.08723042905330658, "global_step": 131545, "epoch": 1478, "lr": 5.256652502249084e-05}
{"train_loss": 0.12607303261756897, "global_step": 131546, "epoch": 1478, "lr": 5.2565946049969614e-05}
{"train_loss": 0.11484246701002121, "global_step": 131547, "epoch": 1478, "lr": 5.256536707710342e-05}
{"train_loss": 0.14367468655109406, "global_step": 131548, "epoch": 1478, "lr": 5.256478810389235e-05}
{"train_loss": 0.10856804251670837, "global_step": 131549, "epoch": 1478, "lr": 5.256420913033646e-05}
{"train_loss": 0.13031479716300964, "global_step": 131550, "epoch": 1478, "lr": 5.256363015643586e-05}
{"train_loss": 0.08306813985109329, "global_step": 131551, "epoch": 1478, "lr": 5.2563051182190605e-05}
{"train_loss": 0.14184221625328064, "global_step": 131552, "epoch": 1478, "lr": 5.2562472207600776e-05}
{"train_loss": 0.06729550659656525, "global_step": 131553, "epoch": 1478, "lr": 5.256189323266645e-05}
{"train_loss": 0.09289517998695374, "global_step": 131554, "epoch": 1478, "lr": 5.256131425738772e-05}
{"train_loss": 0.15604518353939056, "global_step": 131555, "epoch": 1478, "lr": 5.256073528176465e-05}
{"train_loss": 0.12920160591602325, "global_step": 131556, "epoch": 1478, "lr": 5.256015630579732e-05}
{"train_loss": 0.08622467517852783, "global_step": 131557, "epoch": 1478, "lr": 5.2559577329485796e-05}
{"train_loss": 0.11798162013292313, "global_step": 131558, "epoch": 1478, "lr": 5.255899835283018e-05}
{"train_loss": 0.11480151861906052, "global_step": 131559, "epoch": 1478, "lr": 5.255841937583054e-05}
{"train_loss": 0.16303756833076477, "global_step": 131560, "epoch": 1478, "lr": 5.2557840398486944e-05}
{"train_loss": 0.11811905354261398, "global_step": 131561, "epoch": 1478, "lr": 5.2557261420799476e-05}
{"train_loss": 0.18548798561096191, "global_step": 131562, "epoch": 1478, "lr": 5.255668244276823e-05}
{"train_loss": 0.09374480694532394, "global_step": 131563, "epoch": 1478, "lr": 5.255610346439326e-05}
{"train_loss": 0.1714388132095337, "global_step": 131564, "epoch": 1478, "lr": 5.2555524485674646e-05}
{"train_loss": 0.1284784972667694, "global_step": 131565, "epoch": 1478, "lr": 5.2554945506612476e-05}
{"train_loss": 0.12321311980485916, "global_step": 131566, "epoch": 1478, "lr": 5.255436652720683e-05}
{"train_loss": 0.19194166362285614, "global_step": 131567, "epoch": 1478, "lr": 5.255378754745778e-05}
{"train_loss": 0.15966491401195526, "global_step": 131568, "epoch": 1478, "lr": 5.2553208567365394e-05}
{"train_loss": 0.0939001590013504, "global_step": 131569, "epoch": 1478, "lr": 5.255262958692978e-05}
{"train_loss": 0.1893809735774994, "global_step": 131570, "epoch": 1478, "lr": 5.2552050606150985e-05}
{"train_loss": 0.07963486015796661, "global_step": 131571, "epoch": 1478, "lr": 5.255147162502909e-05}
{"train_loss": 0.08498825877904892, "global_step": 131572, "epoch": 1478, "lr": 5.25508926435642e-05}
{"train_loss": 0.128719761967659, "global_step": 131573, "epoch": 1478, "lr": 5.2550313661756355e-05}
{"train_loss": 0.1653836965560913, "global_step": 131574, "epoch": 1478, "lr": 5.254973467960565e-05}
{"train_loss": 0.16942614316940308, "global_step": 131575, "epoch": 1478, "lr": 5.254915569711217e-05}
{"train_loss": 0.182211235165596, "global_step": 131576, "epoch": 1478, "lr": 5.2548576714276e-05}
{"train_loss": 0.12638041377067566, "global_step": 131577, "epoch": 1478, "lr": 5.25479977310972e-05}
{"train_loss": 0.11951766908168793, "global_step": 131578, "epoch": 1478, "lr": 5.254741874757585e-05}
{"train_loss": 0.21968956291675568, "global_step": 131579, "epoch": 1478, "lr": 5.254683976371203e-05}
{"train_loss": 0.1299307644367218, "global_step": 131580, "epoch": 1478, "lr": 5.2546260779505806e-05}
{"train_loss": 0.12808403372764587, "global_step": 131581, "epoch": 1478, "lr": 5.254568179495728e-05}
{"train_loss": 0.14595995843410492, "global_step": 131582, "epoch": 1478, "lr": 5.2545102810066524e-05}
{"train_loss": 0.22302241623401642, "global_step": 131583, "epoch": 1478, "lr": 5.2544523824833604e-05}
{"train_loss": 0.09491205960512161, "global_step": 131584, "epoch": 1478, "lr": 5.25439448392586e-05}
{"train_loss": 0.09801515936851501, "global_step": 131585, "epoch": 1478, "lr": 5.2543365853341594e-05}
{"train_loss": 0.11537794768810272, "global_step": 131586, "epoch": 1478, "lr": 5.2542786867082674e-05}
{"train_loss": 0.08247222751379013, "global_step": 131587, "epoch": 1478, "lr": 5.25422078804819e-05}
{"train_loss": 0.09642460197210312, "global_step": 131588, "epoch": 1478, "lr": 5.2541628893539354e-05}
{"train_loss": 0.10306672751903534, "global_step": 131589, "epoch": 1478, "lr": 5.254104990625511e-05}
{"train_loss": 0.15571749210357666, "global_step": 131590, "epoch": 1478, "lr": 5.254047091862927e-05}
{"train_loss": 0.1019492968916893, "global_step": 131591, "epoch": 1478, "lr": 5.253989193066189e-05}
{"train_loss": 0.1950961947441101, "global_step": 131592, "epoch": 1478, "lr": 5.253931294235305e-05}
{"train_loss": 0.17101746797561646, "global_step": 131593, "epoch": 1478, "lr": 5.253873395370281e-05}
{"train_loss": 0.1425251066684723, "global_step": 131594, "epoch": 1478, "lr": 5.2538154964711294e-05}
{"train_loss": 0.10333127528429031, "global_step": 131595, "epoch": 1478, "lr": 5.253757597537856e-05}
{"train_loss": 0.07711596041917801, "global_step": 131596, "epoch": 1478, "lr": 5.2536996985704666e-05}
{"train_loss": 0.12341156601905823, "global_step": 131597, "epoch": 1478, "lr": 5.25364179956897e-05}
{"train_loss": 0.08649769425392151, "global_step": 131598, "epoch": 1478, "lr": 5.253583900533375e-05}
{"train_loss": 0.08876435458660126, "global_step": 131599, "epoch": 1478, "lr": 5.253526001463689e-05}
{"train_loss": 0.10311692208051682, "global_step": 131600, "epoch": 1478, "lr": 5.253468102359918e-05}
{"train_loss": 0.1797339767217636, "global_step": 131601, "epoch": 1478, "lr": 5.2534102032220735e-05}
{"train_loss": 0.0903659537434578, "global_step": 131602, "epoch": 1478, "lr": 5.2533523040501596e-05}
{"train_loss": 0.05340227112174034, "global_step": 131603, "epoch": 1478, "lr": 5.253294404844187e-05}
{"train_loss": 0.12232273817062378, "global_step": 131604, "epoch": 1478, "lr": 5.253236505604161e-05}
{"train_loss": 0.08900883048772812, "global_step": 131605, "epoch": 1478, "lr": 5.2531786063300906e-05}
{"train_loss": 0.14179232716560364, "global_step": 131606, "epoch": 1478, "lr": 5.2531207070219826e-05}
{"train_loss": 0.18723398447036743, "global_step": 131607, "epoch": 1478, "lr": 5.253062807679846e-05}
{"train_loss": 0.13894598186016083, "global_step": 131608, "epoch": 1478, "lr": 5.253004908303689e-05}
{"train_loss": 0.21177850663661957, "global_step": 131609, "epoch": 1478, "lr": 5.2529470088935184e-05}
{"train_loss": 0.14601801335811615, "global_step": 131610, "epoch": 1478, "lr": 5.2528891094493417e-05}
{"train_loss": 0.13548821210861206, "global_step": 131611, "epoch": 1478, "lr": 5.2528312099711664e-05}
{"train_loss": 0.14261981844902039, "global_step": 131612, "epoch": 1478, "lr": 5.2527733104590014e-05}
{"train_loss": 0.08932947367429733, "global_step": 131613, "epoch": 1478, "lr": 5.252715410912855e-05}
{"train_loss": 0.10820448398590088, "global_step": 131614, "epoch": 1478, "lr": 5.252657511332735e-05}
{"train_loss": 0.1555282324552536, "global_step": 131615, "epoch": 1478, "lr": 5.2525996117186474e-05}
{"train_loss": 0.07619810849428177, "global_step": 131616, "epoch": 1478, "lr": 5.2525417120706e-05}
{"train_loss": 0.1895277053117752, "global_step": 131617, "epoch": 1478, "lr": 5.252483812388601e-05}
{"train_loss": 0.15961895883083344, "global_step": 131618, "epoch": 1478, "lr": 5.252425912672661e-05}
{"train_loss": 0.15381191670894623, "global_step": 131619, "epoch": 1478, "lr": 5.2523680129227824e-05}
{"train_loss": 0.10385337471961975, "global_step": 131620, "epoch": 1478, "lr": 5.252310113138979e-05}
{"train_loss": 0.09618569910526276, "global_step": 131621, "epoch": 1478, "lr": 5.252252213321254e-05}
{"train_loss": 0.11569758504629135, "global_step": 131622, "epoch": 1478, "lr": 5.252194313469616e-05}
{"train_loss": 0.0720798596739769, "global_step": 131623, "epoch": 1478, "lr": 5.2521364135840744e-05}
{"train_loss": 0.15143658220767975, "global_step": 131624, "epoch": 1478, "lr": 5.252078513664637e-05}
{"train_loss": 0.0858248695731163, "global_step": 131625, "epoch": 1478, "lr": 5.2520206137113095e-05}
{"train_loss": 0.19005312025547028, "global_step": 131626, "epoch": 1478, "lr": 5.251962713724101e-05}
{"train_loss": 0.12657204270362854, "global_step": 131627, "epoch": 1478, "lr": 5.251904813703019e-05}
{"train_loss": 0.10094183683395386, "global_step": 131628, "epoch": 1478, "lr": 5.251846913648072e-05}
{"train_loss": 0.11283385753631592, "global_step": 131629, "epoch": 1478, "lr": 5.251789013559266e-05}
{"train_loss": 0.1292863412435805, "global_step": 131630, "epoch": 1478, "lr": 5.251731113436612e-05, "val_loss": 5.709042549133301}
{"train_loss": 0.14713451266288757, "global_step": 131631, "epoch": 1479, "lr": 5.251673213280115e-05}
{"train_loss": 0.16487649083137512, "global_step": 131632, "epoch": 1479, "lr": 5.251615313089784e-05}
{"train_loss": 0.13561676442623138, "global_step": 131633, "epoch": 1479, "lr": 5.251557412865625e-05}
{"train_loss": 0.16007918119430542, "global_step": 131634, "epoch": 1479, "lr": 5.251499512607649e-05}
{"train_loss": 0.11549926549196243, "global_step": 131635, "epoch": 1479, "lr": 5.2514416123158615e-05}
{"train_loss": 0.14181356132030487, "global_step": 131636, "epoch": 1479, "lr": 5.251383711990271e-05}
{"train_loss": 0.12787531316280365, "global_step": 131637, "epoch": 1479, "lr": 5.251325811630884e-05}
{"train_loss": 0.10516668856143951, "global_step": 131638, "epoch": 1479, "lr": 5.251267911237709e-05}
{"train_loss": 0.06622085720300674, "global_step": 131639, "epoch": 1479, "lr": 5.2512100108107555e-05}
{"train_loss": 0.08953746408224106, "global_step": 131640, "epoch": 1479, "lr": 5.2511521103500295e-05}
{"train_loss": 0.2063010334968567, "global_step": 131641, "epoch": 1479, "lr": 5.2510942098555394e-05}
{"train_loss": 0.16222260892391205, "global_step": 131642, "epoch": 1479, "lr": 5.251036309327292e-05}
{"train_loss": 0.1553431749343872, "global_step": 131643, "epoch": 1479, "lr": 5.250978408765296e-05}
{"train_loss": 0.12820552289485931, "global_step": 131644, "epoch": 1479, "lr": 5.250920508169559e-05}
{"train_loss": 0.13269618153572083, "global_step": 131645, "epoch": 1479, "lr": 5.250862607540089e-05}
{"train_loss": 0.15587273240089417, "global_step": 131646, "epoch": 1479, "lr": 5.2508047068768926e-05}
{"train_loss": 0.17998428642749786, "global_step": 131647, "epoch": 1479, "lr": 5.250746806179979e-05}
{"train_loss": 0.11877588927745819, "global_step": 131648, "epoch": 1479, "lr": 5.250688905449356e-05}
{"train_loss": 0.16803567111492157, "global_step": 131649, "epoch": 1479, "lr": 5.2506310046850303e-05}
{"train_loss": 0.13007166981697083, "global_step": 131650, "epoch": 1479, "lr": 5.250573103887011e-05}
{"train_loss": 0.11307098716497421, "global_step": 131651, "epoch": 1479, "lr": 5.250515203055305e-05}
{"train_loss": 0.15665221214294434, "global_step": 131652, "epoch": 1479, "lr": 5.250457302189921e-05}
{"train_loss": 0.10649652779102325, "global_step": 131653, "epoch": 1479, "lr": 5.250399401290866e-05}
{"train_loss": 0.10494652390480042, "global_step": 131654, "epoch": 1479, "lr": 5.250341500358147e-05}
{"train_loss": 0.11466911435127258, "global_step": 131655, "epoch": 1479, "lr": 5.2502835993917734e-05}
{"train_loss": 0.1652986854314804, "global_step": 131656, "epoch": 1479, "lr": 5.250225698391752e-05}
{"train_loss": 0.13755953311920166, "global_step": 131657, "epoch": 1479, "lr": 5.25016779735809e-05}
{"train_loss": 0.16352443397045135, "global_step": 131658, "epoch": 1479, "lr": 5.250109896290797e-05}
{"train_loss": 0.2015857845544815, "global_step": 131659, "epoch": 1479, "lr": 5.2500519951898786e-05}
{"train_loss": 0.11296388506889343, "global_step": 131660, "epoch": 1479, "lr": 5.2499940940553446e-05}
{"train_loss": 0.12503036856651306, "global_step": 131661, "epoch": 1479, "lr": 5.2499361928872025e-05}
{"train_loss": 0.11184218525886536, "global_step": 131662, "epoch": 1479, "lr": 5.249878291685458e-05}
{"train_loss": 0.15034949779510498, "global_step": 131663, "epoch": 1479, "lr": 5.249820390450122e-05}
{"train_loss": 0.13056452572345734, "global_step": 131664, "epoch": 1479, "lr": 5.249762489181199e-05}
{"train_loss": 0.15518051385879517, "global_step": 131665, "epoch": 1479, "lr": 5.2497045878786985e-05}
{"train_loss": 0.09535005688667297, "global_step": 131666, "epoch": 1479, "lr": 5.2496466865426296e-05}
{"train_loss": 0.10065769404172897, "global_step": 131667, "epoch": 1479, "lr": 5.2495887851729987e-05}
{"train_loss": 0.15445227921009064, "global_step": 131668, "epoch": 1479, "lr": 5.249530883769813e-05}
{"train_loss": 0.12363628298044205, "global_step": 131669, "epoch": 1479, "lr": 5.2494729823330814e-05}
{"train_loss": 0.11841392517089844, "global_step": 131670, "epoch": 1479, "lr": 5.249415080862811e-05}
{"train_loss": 0.12864843010902405, "global_step": 131671, "epoch": 1479, "lr": 5.249357179359009e-05}
{"train_loss": 0.14774872362613678, "global_step": 131672, "epoch": 1479, "lr": 5.2492992778216865e-05}
{"train_loss": 0.12966546416282654, "global_step": 131673, "epoch": 1479, "lr": 5.2492413762508466e-05}
{"train_loss": 0.15358677506446838, "global_step": 131674, "epoch": 1479, "lr": 5.2491834746464994e-05}
{"train_loss": 0.17178113758563995, "global_step": 131675, "epoch": 1479, "lr": 5.249125573008653e-05}
{"train_loss": 0.08435440063476562, "global_step": 131676, "epoch": 1479, "lr": 5.2490676713373144e-05}
{"train_loss": 0.08029486984014511, "global_step": 131677, "epoch": 1479, "lr": 5.2490097696324914e-05}
{"train_loss": 0.16503176093101501, "global_step": 131678, "epoch": 1479, "lr": 5.248951867894193e-05}
{"train_loss": 0.12499120831489563, "global_step": 131679, "epoch": 1479, "lr": 5.2488939661224254e-05}
{"train_loss": 0.10662295669317245, "global_step": 131680, "epoch": 1479, "lr": 5.2488360643171966e-05}
{"train_loss": 0.0878758653998375, "global_step": 131681, "epoch": 1479, "lr": 5.248778162478516e-05}
{"train_loss": 0.09644551575183868, "global_step": 131682, "epoch": 1479, "lr": 5.248720260606389e-05}
{"train_loss": 0.12197873741388321, "global_step": 131683, "epoch": 1479, "lr": 5.2486623587008255e-05}
{"train_loss": 0.09743732213973999, "global_step": 131684, "epoch": 1479, "lr": 5.248604456761831e-05}
{"train_loss": 0.10284630209207535, "global_step": 131685, "epoch": 1479, "lr": 5.248546554789415e-05}
{"train_loss": 0.17020156979560852, "global_step": 131686, "epoch": 1479, "lr": 5.248488652783585e-05}
{"train_loss": 0.12602223455905914, "global_step": 131687, "epoch": 1479, "lr": 5.24843075074435e-05}
{"train_loss": 0.1372305005788803, "global_step": 131688, "epoch": 1479, "lr": 5.2483728486717156e-05}
{"train_loss": 0.18521402776241302, "global_step": 131689, "epoch": 1479, "lr": 5.248314946565691e-05}
{"train_loss": 0.08543344587087631, "global_step": 131690, "epoch": 1479, "lr": 5.2482570444262834e-05}
{"train_loss": 0.11805219948291779, "global_step": 131691, "epoch": 1479, "lr": 5.2481991422535005e-05}
{"train_loss": 0.10828660428524017, "global_step": 131692, "epoch": 1479, "lr": 5.2481412400473506e-05}
{"train_loss": 0.1702995002269745, "global_step": 131693, "epoch": 1479, "lr": 5.248083337807841e-05}
{"train_loss": 0.10455393046140671, "global_step": 131694, "epoch": 1479, "lr": 5.2480254355349804e-05}
{"train_loss": 0.16187354922294617, "global_step": 131695, "epoch": 1479, "lr": 5.2479675332287735e-05}
{"train_loss": 0.15760909020900726, "global_step": 131696, "epoch": 1479, "lr": 5.2479096308892315e-05}
{"train_loss": 0.1749037653207779, "global_step": 131697, "epoch": 1479, "lr": 5.247851728516362e-05}
{"train_loss": 0.17950192093849182, "global_step": 131698, "epoch": 1479, "lr": 5.247793826110171e-05}
{"train_loss": 0.159375861287117, "global_step": 131699, "epoch": 1479, "lr": 5.247735923670667e-05}
{"train_loss": 0.1494954228401184, "global_step": 131700, "epoch": 1479, "lr": 5.2476780211978585e-05}
{"train_loss": 0.07048126310110092, "global_step": 131701, "epoch": 1479, "lr": 5.247620118691752e-05}
{"train_loss": 0.08677957206964493, "global_step": 131702, "epoch": 1479, "lr": 5.2475622161523555e-05}
{"train_loss": 0.08773165196180344, "global_step": 131703, "epoch": 1479, "lr": 5.247504313579679e-05}
{"train_loss": 0.14656533300876617, "global_step": 131704, "epoch": 1479, "lr": 5.247446410973727e-05}
{"train_loss": 0.14201970398426056, "global_step": 131705, "epoch": 1479, "lr": 5.24738850833451e-05}
{"train_loss": 0.07458914816379547, "global_step": 131706, "epoch": 1479, "lr": 5.247330605662033e-05}
{"train_loss": 0.0754181295633316, "global_step": 131707, "epoch": 1479, "lr": 5.247272702956307e-05}
{"train_loss": 0.16564011573791504, "global_step": 131708, "epoch": 1479, "lr": 5.2472148002173385e-05}
{"train_loss": 0.1258826106786728, "global_step": 131709, "epoch": 1479, "lr": 5.2471568974451334e-05}
{"train_loss": 0.2102149873971939, "global_step": 131710, "epoch": 1479, "lr": 5.247098994639702e-05}
{"train_loss": 0.1599900722503662, "global_step": 131711, "epoch": 1479, "lr": 5.2470410918010526e-05}
{"train_loss": 0.20418787002563477, "global_step": 131712, "epoch": 1479, "lr": 5.24698318892919e-05}
{"train_loss": 0.09256042540073395, "global_step": 131713, "epoch": 1479, "lr": 5.246925286024124e-05}
{"train_loss": 0.14619551599025726, "global_step": 131714, "epoch": 1479, "lr": 5.24686738308586e-05}
{"train_loss": 0.12485570460557938, "global_step": 131715, "epoch": 1479, "lr": 5.2468094801144094e-05}
{"train_loss": 0.1787077635526657, "global_step": 131716, "epoch": 1479, "lr": 5.2467515771097785e-05}
{"train_loss": 0.17520266771316528, "global_step": 131717, "epoch": 1479, "lr": 5.2466936740719744e-05}
{"train_loss": 0.16219784319400787, "global_step": 131718, "epoch": 1479, "lr": 5.246635771001005e-05}
{"train_loss": 0.13531562811538075, "global_step": 131719, "epoch": 1479, "lr": 5.246577867896879e-05, "val_loss": 5.628498554229736}
{"train_loss": 0.13713470101356506, "global_step": 131720, "epoch": 1480, "lr": 5.246519964759602e-05}
{"train_loss": 0.06541496515274048, "global_step": 131721, "epoch": 1480, "lr": 5.2464620615891846e-05}
{"train_loss": 0.11810844391584396, "global_step": 131722, "epoch": 1480, "lr": 5.246404158385634e-05}
{"train_loss": 0.0711921975016594, "global_step": 131723, "epoch": 1480, "lr": 5.246346255148957e-05}
{"train_loss": 0.07489687204360962, "global_step": 131724, "epoch": 1480, "lr": 5.2462883518791616e-05}
{"train_loss": 0.11901503801345825, "global_step": 131725, "epoch": 1480, "lr": 5.246230448576256e-05}
{"train_loss": 0.164414182305336, "global_step": 131726, "epoch": 1480, "lr": 5.246172545240248e-05}
{"train_loss": 0.14961963891983032, "global_step": 131727, "epoch": 1480, "lr": 5.2461146418711436e-05}
{"train_loss": 0.1448744833469391, "global_step": 131728, "epoch": 1480, "lr": 5.246056738468953e-05}
{"train_loss": 0.17806671559810638, "global_step": 131729, "epoch": 1480, "lr": 5.245998835033683e-05}
{"train_loss": 0.12932176887989044, "global_step": 131730, "epoch": 1480, "lr": 5.245940931565342e-05}
{"train_loss": 0.0839681327342987, "global_step": 131731, "epoch": 1480, "lr": 5.2458830280639385e-05}
{"train_loss": 0.11239749193191528, "global_step": 131732, "epoch": 1480, "lr": 5.245825124529476e-05}
{"train_loss": 0.10878224670886993, "global_step": 131733, "epoch": 1480, "lr": 5.2457672209619676e-05}
{"train_loss": 0.09413263946771622, "global_step": 131734, "epoch": 1480, "lr": 5.245709317361418e-05}
{"train_loss": 0.09125544875860214, "global_step": 131735, "epoch": 1480, "lr": 5.245651413727836e-05}
{"train_loss": 0.14568911492824554, "global_step": 131736, "epoch": 1480, "lr": 5.2455935100612284e-05}
{"train_loss": 0.10690262168645859, "global_step": 131737, "epoch": 1480, "lr": 5.245535606361604e-05}
{"train_loss": 0.07895855605602264, "global_step": 131738, "epoch": 1480, "lr": 5.2454777026289695e-05}
{"train_loss": 0.12729638814926147, "global_step": 131739, "epoch": 1480, "lr": 5.2454197988633344e-05}
{"train_loss": 0.13403157889842987, "global_step": 131740, "epoch": 1480, "lr": 5.245361895064706e-05}
{"train_loss": 0.09398425370454788, "global_step": 131741, "epoch": 1480, "lr": 5.2453039912330914e-05}
{"train_loss": 0.11569797992706299, "global_step": 131742, "epoch": 1480, "lr": 5.2452460873684985e-05}
{"train_loss": 0.10465872287750244, "global_step": 131743, "epoch": 1480, "lr": 5.245188183470935e-05}
{"train_loss": 0.1544988453388214, "global_step": 131744, "epoch": 1480, "lr": 5.24513027954041e-05}
{"train_loss": 0.10947366058826447, "global_step": 131745, "epoch": 1480, "lr": 5.245072375576929e-05}
{"train_loss": 0.11401237547397614, "global_step": 131746, "epoch": 1480, "lr": 5.2450144715805014e-05}
{"train_loss": 0.09112034738063812, "global_step": 131747, "epoch": 1480, "lr": 5.244956567551135e-05}
{"train_loss": 0.10965316742658615, "global_step": 131748, "epoch": 1480, "lr": 5.2448986634888376e-05}
{"train_loss": 0.0999191626906395, "global_step": 131749, "epoch": 1480, "lr": 5.244840759393616e-05}
{"train_loss": 0.0870458111166954, "global_step": 131750, "epoch": 1480, "lr": 5.2447828552654784e-05}
{"train_loss": 0.16258518397808075, "global_step": 131751, "epoch": 1480, "lr": 5.244724951104433e-05}
{"train_loss": 0.07463768124580383, "global_step": 131752, "epoch": 1480, "lr": 5.2446670469104876e-05}
{"train_loss": 0.1772136688232422, "global_step": 131753, "epoch": 1480, "lr": 5.2446091426836485e-05}
{"train_loss": 0.12476062029600143, "global_step": 131754, "epoch": 1480, "lr": 5.244551238423926e-05}
{"train_loss": 0.1499422937631607, "global_step": 131755, "epoch": 1480, "lr": 5.244493334131326e-05}
{"train_loss": 0.10042699426412582, "global_step": 131756, "epoch": 1480, "lr": 5.244435429805855e-05}
{"train_loss": 0.08661512285470963, "global_step": 131757, "epoch": 1480, "lr": 5.2443775254475246e-05}
{"train_loss": 0.1033068522810936, "global_step": 131758, "epoch": 1480, "lr": 5.244319621056341e-05}
{"train_loss": 0.06783845275640488, "global_step": 131759, "epoch": 1480, "lr": 5.24426171663231e-05}
{"train_loss": 0.07012786716222763, "global_step": 131760, "epoch": 1480, "lr": 5.244203812175442e-05}
{"train_loss": 0.08086585998535156, "global_step": 131761, "epoch": 1480, "lr": 5.244145907685745e-05}
{"train_loss": 0.11811164766550064, "global_step": 131762, "epoch": 1480, "lr": 5.2440880031632235e-05}
{"train_loss": 0.1447351723909378, "global_step": 131763, "epoch": 1480, "lr": 5.244030098607888e-05}
{"train_loss": 0.1103820875287056, "global_step": 131764, "epoch": 1480, "lr": 5.243972194019746e-05}
{"train_loss": 0.14348165690898895, "global_step": 131765, "epoch": 1480, "lr": 5.243914289398805e-05}
{"train_loss": 0.10940855741500854, "global_step": 131766, "epoch": 1480, "lr": 5.2438563847450726e-05}
{"train_loss": 0.1201634556055069, "global_step": 131767, "epoch": 1480, "lr": 5.243798480058556e-05}
{"train_loss": 0.13020585477352142, "global_step": 131768, "epoch": 1480, "lr": 5.243740575339264e-05}
{"train_loss": 0.09666798263788223, "global_step": 131769, "epoch": 1480, "lr": 5.243682670587206e-05}
{"train_loss": 0.12922893464565277, "global_step": 131770, "epoch": 1480, "lr": 5.243624765802386e-05}
{"train_loss": 0.08840517699718475, "global_step": 131771, "epoch": 1480, "lr": 5.2435668609848146e-05}
{"train_loss": 0.1828509271144867, "global_step": 131772, "epoch": 1480, "lr": 5.243508956134496e-05}
{"train_loss": 0.154189795255661, "global_step": 131773, "epoch": 1480, "lr": 5.243451051251444e-05}
{"train_loss": 0.14633609354496002, "global_step": 131774, "epoch": 1480, "lr": 5.243393146335659e-05}
{"train_loss": 0.15357805788516998, "global_step": 131775, "epoch": 1480, "lr": 5.243335241387155e-05}
{"train_loss": 0.13626471161842346, "global_step": 131776, "epoch": 1480, "lr": 5.2432773364059386e-05}
{"train_loss": 0.09478035569190979, "global_step": 131777, "epoch": 1480, "lr": 5.243219431392015e-05}
{"train_loss": 0.10640335083007812, "global_step": 131778, "epoch": 1480, "lr": 5.2431615263453934e-05}
{"train_loss": 0.1700602024793625, "global_step": 131779, "epoch": 1480, "lr": 5.243103621266082e-05}
{"train_loss": 0.14672517776489258, "global_step": 131780, "epoch": 1480, "lr": 5.2430457161540894e-05}
{"train_loss": 0.13098646700382233, "global_step": 131781, "epoch": 1480, "lr": 5.242987811009421e-05}
{"train_loss": 0.10841531306505203, "global_step": 131782, "epoch": 1480, "lr": 5.242929905832087e-05}
{"train_loss": 0.12131693214178085, "global_step": 131783, "epoch": 1480, "lr": 5.242872000622093e-05}
{"train_loss": 0.1217866837978363, "global_step": 131784, "epoch": 1480, "lr": 5.242814095379448e-05}
{"train_loss": 0.14230144023895264, "global_step": 131785, "epoch": 1480, "lr": 5.24275619010416e-05}
{"train_loss": 0.09961395710706711, "global_step": 131786, "epoch": 1480, "lr": 5.242698284796236e-05}
{"train_loss": 0.1876375526189804, "global_step": 131787, "epoch": 1480, "lr": 5.242640379455684e-05}
{"train_loss": 0.07007021456956863, "global_step": 131788, "epoch": 1480, "lr": 5.242582474082513e-05}
{"train_loss": 0.12136151641607285, "global_step": 131789, "epoch": 1480, "lr": 5.24252456867673e-05}
{"train_loss": 0.11194465309381485, "global_step": 131790, "epoch": 1480, "lr": 5.2424666632383414e-05}
{"train_loss": 0.10617832839488983, "global_step": 131791, "epoch": 1480, "lr": 5.2424087577673566e-05}
{"train_loss": 0.1013064906001091, "global_step": 131792, "epoch": 1480, "lr": 5.242350852263782e-05}
{"train_loss": 0.12367240339517593, "global_step": 131793, "epoch": 1480, "lr": 5.242292946727626e-05}
{"train_loss": 0.08019848167896271, "global_step": 131794, "epoch": 1480, "lr": 5.2422350411588974e-05}
{"train_loss": 0.17934058606624603, "global_step": 131795, "epoch": 1480, "lr": 5.242177135557603e-05}
{"train_loss": 0.17028085887432098, "global_step": 131796, "epoch": 1480, "lr": 5.2421192299237506e-05}
{"train_loss": 0.1611572951078415, "global_step": 131797, "epoch": 1480, "lr": 5.242061324257349e-05}
{"train_loss": 0.07576172798871994, "global_step": 131798, "epoch": 1480, "lr": 5.2420034185584054e-05}
{"train_loss": 0.12672288715839386, "global_step": 131799, "epoch": 1480, "lr": 5.241945512826927e-05}
{"train_loss": 0.1394532173871994, "global_step": 131800, "epoch": 1480, "lr": 5.241887607062922e-05}
{"train_loss": 0.10622471570968628, "global_step": 131801, "epoch": 1480, "lr": 5.241829701266399e-05}
{"train_loss": 0.1168033555150032, "global_step": 131802, "epoch": 1480, "lr": 5.2417717954373636e-05}
{"train_loss": 0.09490089118480682, "global_step": 131803, "epoch": 1480, "lr": 5.2417138895758254e-05}
{"train_loss": 0.11660149693489075, "global_step": 131804, "epoch": 1480, "lr": 5.2416559836817916e-05}
{"train_loss": 0.12249685078859329, "global_step": 131805, "epoch": 1480, "lr": 5.24159807775527e-05}
{"train_loss": 0.10786519199609756, "global_step": 131806, "epoch": 1480, "lr": 5.2415401717962686e-05}
{"train_loss": 0.12666799128055573, "global_step": 131807, "epoch": 1480, "lr": 5.2414822658047955e-05}
{"train_loss": 0.11919064468212343, "global_step": 131808, "epoch": 1480, "lr": 5.241424359780859e-05, "val_loss": 5.974944114685059, "train_action_mse_error": 19.20280647277832}
{"train_loss": 0.10999944806098938, "global_step": 131809, "epoch": 1481, "lr": 5.2413664537244644e-05}
{"train_loss": 0.15118785202503204, "global_step": 131810, "epoch": 1481, "lr": 5.241308547635622e-05}
{"train_loss": 0.10807256400585175, "global_step": 131811, "epoch": 1481, "lr": 5.241250641514336e-05}
{"train_loss": 0.17870736122131348, "global_step": 131812, "epoch": 1481, "lr": 5.241192735360618e-05}
{"train_loss": 0.17389260232448578, "global_step": 131813, "epoch": 1481, "lr": 5.241134829174475e-05}
{"train_loss": 0.10247767716646194, "global_step": 131814, "epoch": 1481, "lr": 5.2410769229559154e-05}
{"train_loss": 0.09743046760559082, "global_step": 131815, "epoch": 1481, "lr": 5.2410190167049454e-05}
{"train_loss": 0.1657588928937912, "global_step": 131816, "epoch": 1481, "lr": 5.2409611104215726e-05}
{"train_loss": 0.1463189721107483, "global_step": 131817, "epoch": 1481, "lr": 5.240903204105806e-05}
{"train_loss": 0.07332982867956161, "global_step": 131818, "epoch": 1481, "lr": 5.240845297757654e-05}
{"train_loss": 0.11367852240800858, "global_step": 131819, "epoch": 1481, "lr": 5.240787391377121e-05}
{"train_loss": 0.1364191621541977, "global_step": 131820, "epoch": 1481, "lr": 5.240729484964219e-05}
{"train_loss": 0.13328175246715546, "global_step": 131821, "epoch": 1481, "lr": 5.240671578518953e-05}
{"train_loss": 0.16819913685321808, "global_step": 131822, "epoch": 1481, "lr": 5.240613672041331e-05}
{"train_loss": 0.1831335425376892, "global_step": 131823, "epoch": 1481, "lr": 5.2405557655313634e-05}
{"train_loss": 0.10140209645032883, "global_step": 131824, "epoch": 1481, "lr": 5.240497858989054e-05}
{"train_loss": 0.20235325396060944, "global_step": 131825, "epoch": 1481, "lr": 5.240439952414413e-05}
{"train_loss": 0.16048963367938995, "global_step": 131826, "epoch": 1481, "lr": 5.240382045807449e-05}
{"train_loss": 0.14354796707630157, "global_step": 131827, "epoch": 1481, "lr": 5.240324139168169e-05}
{"train_loss": 0.13115659356117249, "global_step": 131828, "epoch": 1481, "lr": 5.240266232496579e-05}
{"train_loss": 0.16677069664001465, "global_step": 131829, "epoch": 1481, "lr": 5.240208325792687e-05}
{"train_loss": 0.15521174669265747, "global_step": 131830, "epoch": 1481, "lr": 5.240150419056503e-05}
{"train_loss": 0.0938468798995018, "global_step": 131831, "epoch": 1481, "lr": 5.240092512288034e-05}
{"train_loss": 0.09366313368082047, "global_step": 131832, "epoch": 1481, "lr": 5.2400346054872874e-05}
{"train_loss": 0.1284409910440445, "global_step": 131833, "epoch": 1481, "lr": 5.239976698654271e-05}
{"train_loss": 0.12843658030033112, "global_step": 131834, "epoch": 1481, "lr": 5.2399187917889926e-05}
{"train_loss": 0.11510766297578812, "global_step": 131835, "epoch": 1481, "lr": 5.23986088489146e-05}
{"train_loss": 0.09497355669736862, "global_step": 131836, "epoch": 1481, "lr": 5.239802977961681e-05}
{"train_loss": 0.17330853641033173, "global_step": 131837, "epoch": 1481, "lr": 5.2397450709996634e-05}
{"train_loss": 0.10827812552452087, "global_step": 131838, "epoch": 1481, "lr": 5.239687164005416e-05}
{"train_loss": 0.11846388876438141, "global_step": 131839, "epoch": 1481, "lr": 5.239629256978944e-05}
{"train_loss": 0.09141960740089417, "global_step": 131840, "epoch": 1481, "lr": 5.239571349920258e-05}
{"train_loss": 0.08793379366397858, "global_step": 131841, "epoch": 1481, "lr": 5.239513442829363e-05}
{"train_loss": 0.13509084284305573, "global_step": 131842, "epoch": 1481, "lr": 5.2394555357062705e-05}
{"train_loss": 0.1869354248046875, "global_step": 131843, "epoch": 1481, "lr": 5.239397628550985e-05}
{"train_loss": 0.12357181310653687, "global_step": 131844, "epoch": 1481, "lr": 5.239339721363515e-05}
{"train_loss": 0.17561203241348267, "global_step": 131845, "epoch": 1481, "lr": 5.239281814143869e-05}
{"train_loss": 0.24018804728984833, "global_step": 131846, "epoch": 1481, "lr": 5.2392239068920546e-05}
{"train_loss": 0.12085799127817154, "global_step": 131847, "epoch": 1481, "lr": 5.239165999608079e-05}
{"train_loss": 0.12108609825372696, "global_step": 131848, "epoch": 1481, "lr": 5.239108092291951e-05}
{"train_loss": 0.11199960112571716, "global_step": 131849, "epoch": 1481, "lr": 5.239050184943678e-05}
{"train_loss": 0.10095760971307755, "global_step": 131850, "epoch": 1481, "lr": 5.238992277563267e-05}
{"train_loss": 0.16184654831886292, "global_step": 131851, "epoch": 1481, "lr": 5.238934370150727e-05}
{"train_loss": 0.1837777942419052, "global_step": 131852, "epoch": 1481, "lr": 5.238876462706065e-05}
{"train_loss": 0.09341592341661453, "global_step": 131853, "epoch": 1481, "lr": 5.238818555229289e-05}
{"train_loss": 0.13676370680332184, "global_step": 131854, "epoch": 1481, "lr": 5.238760647720407e-05}
{"train_loss": 0.12642177939414978, "global_step": 131855, "epoch": 1481, "lr": 5.238702740179425e-05}
{"train_loss": 0.14478588104248047, "global_step": 131856, "epoch": 1481, "lr": 5.238644832606354e-05}
{"train_loss": 0.10564591735601425, "global_step": 131857, "epoch": 1481, "lr": 5.2385869250012e-05}
{"train_loss": 0.14843495190143585, "global_step": 131858, "epoch": 1481, "lr": 5.23852901736397e-05}
{"train_loss": 0.16161710023880005, "global_step": 131859, "epoch": 1481, "lr": 5.238471109694674e-05}
{"train_loss": 0.13751815259456635, "global_step": 131860, "epoch": 1481, "lr": 5.238413201993317e-05}
{"train_loss": 0.1297292411327362, "global_step": 131861, "epoch": 1481, "lr": 5.238355294259909e-05}
{"train_loss": 0.11883760988712311, "global_step": 131862, "epoch": 1481, "lr": 5.238297386494457e-05}
{"train_loss": 0.1903400719165802, "global_step": 131863, "epoch": 1481, "lr": 5.238239478696969e-05}
{"train_loss": 0.13370390236377716, "global_step": 131864, "epoch": 1481, "lr": 5.238181570867452e-05}
{"train_loss": 0.10314784944057465, "global_step": 131865, "epoch": 1481, "lr": 5.238123663005915e-05}
{"train_loss": 0.20956827700138092, "global_step": 131866, "epoch": 1481, "lr": 5.238065755112365e-05}
{"train_loss": 0.13376007974147797, "global_step": 131867, "epoch": 1481, "lr": 5.23800784718681e-05}
{"train_loss": 0.13901206851005554, "global_step": 131868, "epoch": 1481, "lr": 5.2379499392292575e-05}
{"train_loss": 0.13660432398319244, "global_step": 131869, "epoch": 1481, "lr": 5.237892031239715e-05}
{"train_loss": 0.16535435616970062, "global_step": 131870, "epoch": 1481, "lr": 5.2378341232181914e-05}
{"train_loss": 0.12288740277290344, "global_step": 131871, "epoch": 1481, "lr": 5.2377762151646946e-05}
{"train_loss": 0.05662751942873001, "global_step": 131872, "epoch": 1481, "lr": 5.2377183070792314e-05}
{"train_loss": 0.1690153181552887, "global_step": 131873, "epoch": 1481, "lr": 5.237660398961809e-05}
{"train_loss": 0.11199243366718292, "global_step": 131874, "epoch": 1481, "lr": 5.2376024908124365e-05}
{"train_loss": 0.09030749648809433, "global_step": 131875, "epoch": 1481, "lr": 5.237544582631122e-05}
{"train_loss": 0.20730294287204742, "global_step": 131876, "epoch": 1481, "lr": 5.237486674417872e-05}
{"train_loss": 0.1253190040588379, "global_step": 131877, "epoch": 1481, "lr": 5.237428766172694e-05}
{"train_loss": 0.2091614156961441, "global_step": 131878, "epoch": 1481, "lr": 5.237370857895597e-05}
{"train_loss": 0.1627764105796814, "global_step": 131879, "epoch": 1481, "lr": 5.237312949586589e-05}
{"train_loss": 0.23064376413822174, "global_step": 131880, "epoch": 1481, "lr": 5.237255041245677e-05}
{"train_loss": 0.09450479596853256, "global_step": 131881, "epoch": 1481, "lr": 5.237197132872869e-05}
{"train_loss": 0.07640396803617477, "global_step": 131882, "epoch": 1481, "lr": 5.2371392244681726e-05}
{"train_loss": 0.13243527710437775, "global_step": 131883, "epoch": 1481, "lr": 5.237081316031596e-05}
{"train_loss": 0.07343585044145584, "global_step": 131884, "epoch": 1481, "lr": 5.237023407563145e-05}
{"train_loss": 0.14177830517292023, "global_step": 131885, "epoch": 1481, "lr": 5.236965499062832e-05}
{"train_loss": 0.25064578652381897, "global_step": 131886, "epoch": 1481, "lr": 5.2369075905306596e-05}
{"train_loss": 0.1419386863708496, "global_step": 131887, "epoch": 1481, "lr": 5.236849681966638e-05}
{"train_loss": 0.1664687544107437, "global_step": 131888, "epoch": 1481, "lr": 5.236791773370776e-05}
{"train_loss": 0.16109773516654968, "global_step": 131889, "epoch": 1481, "lr": 5.23673386474308e-05}
{"train_loss": 0.1292283833026886, "global_step": 131890, "epoch": 1481, "lr": 5.2366759560835574e-05}
{"train_loss": 0.2150859385728836, "global_step": 131891, "epoch": 1481, "lr": 5.236618047392216e-05}
{"train_loss": 0.10326717048883438, "global_step": 131892, "epoch": 1481, "lr": 5.2365601386690654e-05}
{"train_loss": 0.10978419333696365, "global_step": 131893, "epoch": 1481, "lr": 5.236502229914112e-05}
{"train_loss": 0.09853147715330124, "global_step": 131894, "epoch": 1481, "lr": 5.236444321127364e-05}
{"train_loss": 0.13160328567028046, "global_step": 131895, "epoch": 1481, "lr": 5.236386412308828e-05}
{"train_loss": 0.13497279584407806, "global_step": 131896, "epoch": 1481, "lr": 5.236328503458514e-05}
{"train_loss": 0.13813492446468117, "global_step": 131897, "epoch": 1481, "lr": 5.236270594576428e-05, "val_loss": 5.610205173492432}
{"train_loss": 0.0788080170750618, "global_step": 131898, "epoch": 1482, "lr": 5.236212685662577e-05}
{"train_loss": 0.13836683332920074, "global_step": 131899, "epoch": 1482, "lr": 5.236154776716972e-05}
{"train_loss": 0.09957259148359299, "global_step": 131900, "epoch": 1482, "lr": 5.236096867739617e-05}
{"train_loss": 0.17859850823879242, "global_step": 131901, "epoch": 1482, "lr": 5.236038958730523e-05}
{"train_loss": 0.16454948484897614, "global_step": 131902, "epoch": 1482, "lr": 5.2359810496896964e-05}
{"train_loss": 0.08021119982004166, "global_step": 131903, "epoch": 1482, "lr": 5.235923140617145e-05}
{"train_loss": 0.12671616673469543, "global_step": 131904, "epoch": 1482, "lr": 5.235865231512876e-05}
{"train_loss": 0.13239702582359314, "global_step": 131905, "epoch": 1482, "lr": 5.235807322376899e-05}
{"train_loss": 0.11714130640029907, "global_step": 131906, "epoch": 1482, "lr": 5.235749413209219e-05}
{"train_loss": 0.1341063231229782, "global_step": 131907, "epoch": 1482, "lr": 5.2356915040098466e-05}
{"train_loss": 0.08568891137838364, "global_step": 131908, "epoch": 1482, "lr": 5.235633594778787e-05}
{"train_loss": 0.12450163811445236, "global_step": 131909, "epoch": 1482, "lr": 5.235575685516051e-05}
{"train_loss": 0.15937452018260956, "global_step": 131910, "epoch": 1482, "lr": 5.2355177762216435e-05}
{"train_loss": 0.1429285705089569, "global_step": 131911, "epoch": 1482, "lr": 5.235459866895573e-05}
{"train_loss": 0.10837575048208237, "global_step": 131912, "epoch": 1482, "lr": 5.2354019575378496e-05}
{"train_loss": 0.15463703870773315, "global_step": 131913, "epoch": 1482, "lr": 5.235344048148478e-05}
{"train_loss": 0.1124374270439148, "global_step": 131914, "epoch": 1482, "lr": 5.2352861387274676e-05}
{"train_loss": 0.14827857911586761, "global_step": 131915, "epoch": 1482, "lr": 5.235228229274827e-05}
{"train_loss": 0.11976674199104309, "global_step": 131916, "epoch": 1482, "lr": 5.235170319790561e-05}
{"train_loss": 0.14528809487819672, "global_step": 131917, "epoch": 1482, "lr": 5.23511241027468e-05}
{"train_loss": 0.13723790645599365, "global_step": 131918, "epoch": 1482, "lr": 5.235054500727191e-05}
{"train_loss": 0.11958840489387512, "global_step": 131919, "epoch": 1482, "lr": 5.234996591148101e-05}
{"train_loss": 0.1076500192284584, "global_step": 131920, "epoch": 1482, "lr": 5.23493868153742e-05}
{"train_loss": 0.14600485563278198, "global_step": 131921, "epoch": 1482, "lr": 5.234880771895153e-05}
{"train_loss": 0.11537463217973709, "global_step": 131922, "epoch": 1482, "lr": 5.23482286222131e-05}
{"train_loss": 0.1944885551929474, "global_step": 131923, "epoch": 1482, "lr": 5.2347649525158985e-05}
{"train_loss": 0.08292443305253983, "global_step": 131924, "epoch": 1482, "lr": 5.2347070427789243e-05}
{"train_loss": 0.1425100862979889, "global_step": 131925, "epoch": 1482, "lr": 5.234649133010397e-05}
{"train_loss": 0.08540134131908417, "global_step": 131926, "epoch": 1482, "lr": 5.2345912232103246e-05}
{"train_loss": 0.13415218889713287, "global_step": 131927, "epoch": 1482, "lr": 5.234533313378713e-05}
{"train_loss": 0.15086974203586578, "global_step": 131928, "epoch": 1482, "lr": 5.234475403515573e-05}
{"train_loss": 0.12118957936763763, "global_step": 131929, "epoch": 1482, "lr": 5.23441749362091e-05}
{"train_loss": 0.15583758056163788, "global_step": 131930, "epoch": 1482, "lr": 5.2343595836947324e-05}
{"train_loss": 0.12946538627147675, "global_step": 131931, "epoch": 1482, "lr": 5.234301673737048e-05}
{"train_loss": 0.06070072203874588, "global_step": 131932, "epoch": 1482, "lr": 5.234243763747864e-05}
{"train_loss": 0.09906522184610367, "global_step": 131933, "epoch": 1482, "lr": 5.2341858537271893e-05}
{"train_loss": 0.14702627062797546, "global_step": 131934, "epoch": 1482, "lr": 5.2341279436750315e-05}
{"train_loss": 0.17779265344142914, "global_step": 131935, "epoch": 1482, "lr": 5.234070033591397e-05}
{"train_loss": 0.1339348405599594, "global_step": 131936, "epoch": 1482, "lr": 5.234012123476295e-05}
{"train_loss": 0.14093820750713348, "global_step": 131937, "epoch": 1482, "lr": 5.2339542133297335e-05}
{"train_loss": 0.10027947276830673, "global_step": 131938, "epoch": 1482, "lr": 5.2338963031517195e-05}
{"train_loss": 0.10139712691307068, "global_step": 131939, "epoch": 1482, "lr": 5.233838392942261e-05}
{"train_loss": 0.09417931735515594, "global_step": 131940, "epoch": 1482, "lr": 5.2337804827013645e-05}
{"train_loss": 0.12592238187789917, "global_step": 131941, "epoch": 1482, "lr": 5.2337225724290405e-05}
{"train_loss": 0.07217169553041458, "global_step": 131942, "epoch": 1482, "lr": 5.233664662125295e-05}
{"train_loss": 0.1367955207824707, "global_step": 131943, "epoch": 1482, "lr": 5.233606751790137e-05}
{"train_loss": 0.19868549704551697, "global_step": 131944, "epoch": 1482, "lr": 5.2335488414235724e-05}
{"train_loss": 0.14910998940467834, "global_step": 131945, "epoch": 1482, "lr": 5.23349093102561e-05}
{"train_loss": 0.15338577330112457, "global_step": 131946, "epoch": 1482, "lr": 5.233433020596258e-05}
{"train_loss": 0.19261930882930756, "global_step": 131947, "epoch": 1482, "lr": 5.2333751101355234e-05}
{"train_loss": 0.11641217768192291, "global_step": 131948, "epoch": 1482, "lr": 5.2333171996434146e-05}
{"train_loss": 0.17700186371803284, "global_step": 131949, "epoch": 1482, "lr": 5.233259289119939e-05}
{"train_loss": 0.18961423635482788, "global_step": 131950, "epoch": 1482, "lr": 5.2332013785651047e-05}
{"train_loss": 0.2017318308353424, "global_step": 131951, "epoch": 1482, "lr": 5.233143467978919e-05}
{"train_loss": 0.12669143080711365, "global_step": 131952, "epoch": 1482, "lr": 5.2330855573613904e-05}
{"train_loss": 0.10147707909345627, "global_step": 131953, "epoch": 1482, "lr": 5.2330276467125264e-05}
{"train_loss": 0.20166181027889252, "global_step": 131954, "epoch": 1482, "lr": 5.2329697360323337e-05}
{"train_loss": 0.0647999569773674, "global_step": 131955, "epoch": 1482, "lr": 5.232911825320822e-05}
{"train_loss": 0.1360195279121399, "global_step": 131956, "epoch": 1482, "lr": 5.232853914577997e-05}
{"train_loss": 0.06558503955602646, "global_step": 131957, "epoch": 1482, "lr": 5.232796003803869e-05}
{"train_loss": 0.14286711812019348, "global_step": 131958, "epoch": 1482, "lr": 5.232738092998445e-05}
{"train_loss": 0.09834029525518417, "global_step": 131959, "epoch": 1482, "lr": 5.232680182161731e-05}
{"train_loss": 0.08020880818367004, "global_step": 131960, "epoch": 1482, "lr": 5.232622271293735e-05}
{"train_loss": 0.11622975021600723, "global_step": 131961, "epoch": 1482, "lr": 5.232564360394467e-05}
{"train_loss": 0.275034099817276, "global_step": 131962, "epoch": 1482, "lr": 5.232506449463934e-05}
{"train_loss": 0.12626703083515167, "global_step": 131963, "epoch": 1482, "lr": 5.2324485385021424e-05}
{"train_loss": 0.0960121899843216, "global_step": 131964, "epoch": 1482, "lr": 5.232390627509102e-05}
{"train_loss": 0.2180677056312561, "global_step": 131965, "epoch": 1482, "lr": 5.232332716484819e-05}
{"train_loss": 0.1567726880311966, "global_step": 131966, "epoch": 1482, "lr": 5.2322748054293014e-05}
{"train_loss": 0.14212024211883545, "global_step": 131967, "epoch": 1482, "lr": 5.232216894342558e-05}
{"train_loss": 0.08692877739667892, "global_step": 131968, "epoch": 1482, "lr": 5.232158983224594e-05}
{"train_loss": 0.1225048154592514, "global_step": 131969, "epoch": 1482, "lr": 5.2321010720754216e-05}
{"train_loss": 0.11083929985761642, "global_step": 131970, "epoch": 1482, "lr": 5.232043160895044e-05}
{"train_loss": 0.16765832901000977, "global_step": 131971, "epoch": 1482, "lr": 5.231985249683472e-05}
{"train_loss": 0.15128864347934723, "global_step": 131972, "epoch": 1482, "lr": 5.231927338440712e-05}
{"train_loss": 0.08982741087675095, "global_step": 131973, "epoch": 1482, "lr": 5.231869427166772e-05}
{"train_loss": 0.11517097055912018, "global_step": 131974, "epoch": 1482, "lr": 5.23181151586166e-05}
{"train_loss": 0.08450544625520706, "global_step": 131975, "epoch": 1482, "lr": 5.231753604525385e-05}
{"train_loss": 0.10113832354545593, "global_step": 131976, "epoch": 1482, "lr": 5.231695693157952e-05}
{"train_loss": 0.08134271949529648, "global_step": 131977, "epoch": 1482, "lr": 5.231637781759372e-05}
{"train_loss": 0.1792335957288742, "global_step": 131978, "epoch": 1482, "lr": 5.2315798703296504e-05}
{"train_loss": 0.10822410881519318, "global_step": 131979, "epoch": 1482, "lr": 5.231521958868796e-05}
{"train_loss": 0.09349055588245392, "global_step": 131980, "epoch": 1482, "lr": 5.231464047376815e-05}
{"train_loss": 0.2085360437631607, "global_step": 131981, "epoch": 1482, "lr": 5.231406135853717e-05}
{"train_loss": 0.12888643145561218, "global_step": 131982, "epoch": 1482, "lr": 5.2313482242995095e-05}
{"train_loss": 0.15876984596252441, "global_step": 131983, "epoch": 1482, "lr": 5.231290312714201e-05}
{"train_loss": 0.15727293491363525, "global_step": 131984, "epoch": 1482, "lr": 5.231232401097796e-05}
{"train_loss": 0.11462855339050293, "global_step": 131985, "epoch": 1482, "lr": 5.231174489450307e-05}
{"train_loss": 0.13081504973802674, "global_step": 131986, "epoch": 1482, "lr": 5.231116577771739e-05, "val_loss": 5.682348728179932}
{"train_loss": 0.14989818632602692, "global_step": 131987, "epoch": 1483, "lr": 5.231058666062099e-05}
{"train_loss": 0.09153129905462265, "global_step": 131988, "epoch": 1483, "lr": 5.231000754321397e-05}
{"train_loss": 0.12376614660024643, "global_step": 131989, "epoch": 1483, "lr": 5.230942842549639e-05}
{"train_loss": 0.1627449244260788, "global_step": 131990, "epoch": 1483, "lr": 5.230884930746834e-05}
{"train_loss": 0.15292216837406158, "global_step": 131991, "epoch": 1483, "lr": 5.2308270189129894e-05}
{"train_loss": 0.11091776937246323, "global_step": 131992, "epoch": 1483, "lr": 5.230769107048111e-05}
{"train_loss": 0.11684799939393997, "global_step": 131993, "epoch": 1483, "lr": 5.23071119515221e-05}
{"train_loss": 0.1556713879108429, "global_step": 131994, "epoch": 1483, "lr": 5.230653283225294e-05}
{"train_loss": 0.124456025660038, "global_step": 131995, "epoch": 1483, "lr": 5.230595371267368e-05}
{"train_loss": 0.14671775698661804, "global_step": 131996, "epoch": 1483, "lr": 5.2305374592784425e-05}
{"train_loss": 0.08700158447027206, "global_step": 131997, "epoch": 1483, "lr": 5.2304795472585235e-05}
{"train_loss": 0.1148885190486908, "global_step": 131998, "epoch": 1483, "lr": 5.2304216352076194e-05}
{"train_loss": 0.1577533632516861, "global_step": 131999, "epoch": 1483, "lr": 5.2303637231257375e-05}
{"train_loss": 0.09350751340389252, "global_step": 132000, "epoch": 1483, "lr": 5.230305811012887e-05}
{"train_loss": 0.1405324786901474, "global_step": 132001, "epoch": 1483, "lr": 5.2302478988690726e-05}
{"train_loss": 0.17578518390655518, "global_step": 132002, "epoch": 1483, "lr": 5.230189986694306e-05}
{"train_loss": 0.12629473209381104, "global_step": 132003, "epoch": 1483, "lr": 5.230132074488593e-05}
{"train_loss": 0.11945129185914993, "global_step": 132004, "epoch": 1483, "lr": 5.230074162251941e-05}
{"train_loss": 0.10465683788061142, "global_step": 132005, "epoch": 1483, "lr": 5.2300162499843575e-05}
{"train_loss": 0.14544813334941864, "global_step": 132006, "epoch": 1483, "lr": 5.229958337685853e-05}
{"train_loss": 0.09953264147043228, "global_step": 132007, "epoch": 1483, "lr": 5.229900425356432e-05}
{"train_loss": 0.15623807907104492, "global_step": 132008, "epoch": 1483, "lr": 5.229842512996104e-05}
{"train_loss": 0.12214429676532745, "global_step": 132009, "epoch": 1483, "lr": 5.229784600604877e-05}
{"train_loss": 0.12693127989768982, "global_step": 132010, "epoch": 1483, "lr": 5.229726688182757e-05}
{"train_loss": 0.15446443855762482, "global_step": 132011, "epoch": 1483, "lr": 5.229668775729753e-05}
{"train_loss": 0.16082477569580078, "global_step": 132012, "epoch": 1483, "lr": 5.2296108632458743e-05}
{"train_loss": 0.12350399047136307, "global_step": 132013, "epoch": 1483, "lr": 5.229552950731127e-05}
{"train_loss": 0.1064106673002243, "global_step": 132014, "epoch": 1483, "lr": 5.229495038185519e-05}
{"train_loss": 0.07385576516389847, "global_step": 132015, "epoch": 1483, "lr": 5.229437125609059e-05}
{"train_loss": 0.12697634100914001, "global_step": 132016, "epoch": 1483, "lr": 5.229379213001753e-05}
{"train_loss": 0.1338108628988266, "global_step": 132017, "epoch": 1483, "lr": 5.229321300363609e-05}
{"train_loss": 0.07757241278886795, "global_step": 132018, "epoch": 1483, "lr": 5.229263387694636e-05}
{"train_loss": 0.17655880749225616, "global_step": 132019, "epoch": 1483, "lr": 5.2292054749948424e-05}
{"train_loss": 0.1381891369819641, "global_step": 132020, "epoch": 1483, "lr": 5.2291475622642336e-05}
{"train_loss": 0.126634880900383, "global_step": 132021, "epoch": 1483, "lr": 5.22908964950282e-05}
{"train_loss": 0.09834037721157074, "global_step": 132022, "epoch": 1483, "lr": 5.229031736710607e-05}
{"train_loss": 0.09936175495386124, "global_step": 132023, "epoch": 1483, "lr": 5.228973823887604e-05}
{"train_loss": 0.10986825078725815, "global_step": 132024, "epoch": 1483, "lr": 5.228915911033818e-05}
{"train_loss": 0.1377447098493576, "global_step": 132025, "epoch": 1483, "lr": 5.228857998149257e-05}
{"train_loss": 0.12265831232070923, "global_step": 132026, "epoch": 1483, "lr": 5.2288000852339295e-05}
{"train_loss": 0.11722169816493988, "global_step": 132027, "epoch": 1483, "lr": 5.2287421722878416e-05}
{"train_loss": 0.13384296000003815, "global_step": 132028, "epoch": 1483, "lr": 5.228684259311001e-05}
{"train_loss": 0.15262070298194885, "global_step": 132029, "epoch": 1483, "lr": 5.228626346303418e-05}
{"train_loss": 0.08866582810878754, "global_step": 132030, "epoch": 1483, "lr": 5.2285684332651e-05}
{"train_loss": 0.1005856841802597, "global_step": 132031, "epoch": 1483, "lr": 5.228510520196052e-05}
{"train_loss": 0.09343738108873367, "global_step": 132032, "epoch": 1483, "lr": 5.228452607096285e-05}
{"train_loss": 0.0879889726638794, "global_step": 132033, "epoch": 1483, "lr": 5.228394693965805e-05}
{"train_loss": 0.1256987601518631, "global_step": 132034, "epoch": 1483, "lr": 5.2283367808046194e-05}
{"train_loss": 0.13948579132556915, "global_step": 132035, "epoch": 1483, "lr": 5.228278867612738e-05}
{"train_loss": 0.07553312927484512, "global_step": 132036, "epoch": 1483, "lr": 5.228220954390166e-05}
{"train_loss": 0.10487236082553864, "global_step": 132037, "epoch": 1483, "lr": 5.228163041136913e-05}
{"train_loss": 0.08158750087022781, "global_step": 132038, "epoch": 1483, "lr": 5.228105127852987e-05}
{"train_loss": 0.15469594299793243, "global_step": 132039, "epoch": 1483, "lr": 5.228047214538394e-05}
{"train_loss": 0.15268129110336304, "global_step": 132040, "epoch": 1483, "lr": 5.227989301193142e-05}
{"train_loss": 0.1369391232728958, "global_step": 132041, "epoch": 1483, "lr": 5.227931387817242e-05}
{"train_loss": 0.0884818285703659, "global_step": 132042, "epoch": 1483, "lr": 5.227873474410697e-05}
{"train_loss": 0.1340663731098175, "global_step": 132043, "epoch": 1483, "lr": 5.227815560973519e-05}
{"train_loss": 0.08746597915887833, "global_step": 132044, "epoch": 1483, "lr": 5.227757647505713e-05}
{"train_loss": 0.11365022510290146, "global_step": 132045, "epoch": 1483, "lr": 5.227699734007287e-05}
{"train_loss": 0.12664103507995605, "global_step": 132046, "epoch": 1483, "lr": 5.2276418204782516e-05}
{"train_loss": 0.2363540679216385, "global_step": 132047, "epoch": 1483, "lr": 5.2275839069186104e-05}
{"train_loss": 0.10523153841495514, "global_step": 132048, "epoch": 1483, "lr": 5.227525993328375e-05}
{"train_loss": 0.12265168130397797, "global_step": 132049, "epoch": 1483, "lr": 5.22746807970755e-05}
{"train_loss": 0.11043711751699448, "global_step": 132050, "epoch": 1483, "lr": 5.227410166056146e-05}
{"train_loss": 0.10867121070623398, "global_step": 132051, "epoch": 1483, "lr": 5.227352252374169e-05}
{"train_loss": 0.08621414750814438, "global_step": 132052, "epoch": 1483, "lr": 5.2272943386616266e-05}
{"train_loss": 0.14289751648902893, "global_step": 132053, "epoch": 1483, "lr": 5.227236424918529e-05}
{"train_loss": 0.15345588326454163, "global_step": 132054, "epoch": 1483, "lr": 5.227178511144882e-05}
{"train_loss": 0.1279224306344986, "global_step": 132055, "epoch": 1483, "lr": 5.2271205973406925e-05}
{"train_loss": 0.12537840008735657, "global_step": 132056, "epoch": 1483, "lr": 5.22706268350597e-05}
{"train_loss": 0.10858139395713806, "global_step": 132057, "epoch": 1483, "lr": 5.227004769640721e-05}
{"train_loss": 0.13389042019844055, "global_step": 132058, "epoch": 1483, "lr": 5.226946855744955e-05}
{"train_loss": 0.10766627639532089, "global_step": 132059, "epoch": 1483, "lr": 5.226888941818679e-05}
{"train_loss": 0.08739279210567474, "global_step": 132060, "epoch": 1483, "lr": 5.226831027861898e-05}
{"train_loss": 0.10236088186502457, "global_step": 132061, "epoch": 1483, "lr": 5.2267731138746246e-05}
{"train_loss": 0.13866350054740906, "global_step": 132062, "epoch": 1483, "lr": 5.2267151998568645e-05}
{"train_loss": 0.1706337034702301, "global_step": 132063, "epoch": 1483, "lr": 5.2266572858086235e-05}
{"train_loss": 0.1971241533756256, "global_step": 132064, "epoch": 1483, "lr": 5.226599371729912e-05}
{"train_loss": 0.12003160268068314, "global_step": 132065, "epoch": 1483, "lr": 5.226541457620737e-05}
{"train_loss": 0.13076315820217133, "global_step": 132066, "epoch": 1483, "lr": 5.2264835434811066e-05}
{"train_loss": 0.1137448400259018, "global_step": 132067, "epoch": 1483, "lr": 5.226425629311028e-05}
{"train_loss": 0.125854954123497, "global_step": 132068, "epoch": 1483, "lr": 5.22636771511051e-05}
{"train_loss": 0.17372368276119232, "global_step": 132069, "epoch": 1483, "lr": 5.226309800879559e-05}
{"train_loss": 0.07764046639204025, "global_step": 132070, "epoch": 1483, "lr": 5.226251886618182e-05}
{"train_loss": 0.22276510298252106, "global_step": 132071, "epoch": 1483, "lr": 5.22619397232639e-05}
{"train_loss": 0.1096494272351265, "global_step": 132072, "epoch": 1483, "lr": 5.2261360580041885e-05}
{"train_loss": 0.11168656498193741, "global_step": 132073, "epoch": 1483, "lr": 5.226078143651587e-05}
{"train_loss": 0.21239270269870758, "global_step": 132074, "epoch": 1483, "lr": 5.226020229268591e-05}
{"train_loss": 0.12670672375164674, "global_step": 132075, "epoch": 1483, "lr": 5.2259623148552096e-05, "val_loss": 5.679856300354004}
{"train_loss": 0.12445482611656189, "global_step": 132076, "epoch": 1484, "lr": 5.22590440041145e-05}
{"train_loss": 0.17943435907363892, "global_step": 132077, "epoch": 1484, "lr": 5.2258464859373204e-05}
{"train_loss": 0.15266375243663788, "global_step": 132078, "epoch": 1484, "lr": 5.225788571432828e-05}
{"train_loss": 0.09742732346057892, "global_step": 132079, "epoch": 1484, "lr": 5.2257306568979815e-05}
{"train_loss": 0.12262115627527237, "global_step": 132080, "epoch": 1484, "lr": 5.225672742332788e-05}
{"train_loss": 0.12708112597465515, "global_step": 132081, "epoch": 1484, "lr": 5.2256148277372554e-05}
{"train_loss": 0.1962289661169052, "global_step": 132082, "epoch": 1484, "lr": 5.225556913111392e-05}
{"train_loss": 0.08453510701656342, "global_step": 132083, "epoch": 1484, "lr": 5.2254989984552036e-05}
{"train_loss": 0.09281627833843231, "global_step": 132084, "epoch": 1484, "lr": 5.2254410837687016e-05}
{"train_loss": 0.1282375156879425, "global_step": 132085, "epoch": 1484, "lr": 5.225383169051892e-05}
{"train_loss": 0.18048690259456635, "global_step": 132086, "epoch": 1484, "lr": 5.225325254304782e-05}
{"train_loss": 0.14558666944503784, "global_step": 132087, "epoch": 1484, "lr": 5.225267339527379e-05}
{"train_loss": 0.10569732636213303, "global_step": 132088, "epoch": 1484, "lr": 5.2252094247196914e-05}
{"train_loss": 0.08972861617803574, "global_step": 132089, "epoch": 1484, "lr": 5.225151509881728e-05}
{"train_loss": 0.08754788339138031, "global_step": 132090, "epoch": 1484, "lr": 5.2250935950134964e-05}
{"train_loss": 0.14563721418380737, "global_step": 132091, "epoch": 1484, "lr": 5.225035680115002e-05}
{"train_loss": 0.13454726338386536, "global_step": 132092, "epoch": 1484, "lr": 5.2249777651862556e-05}
{"train_loss": 0.08281020820140839, "global_step": 132093, "epoch": 1484, "lr": 5.2249198502272644e-05}
{"train_loss": 0.1446724683046341, "global_step": 132094, "epoch": 1484, "lr": 5.224861935238035e-05}
{"train_loss": 0.11468570679426193, "global_step": 132095, "epoch": 1484, "lr": 5.224804020218574e-05}
{"train_loss": 0.12636834383010864, "global_step": 132096, "epoch": 1484, "lr": 5.2247461051688916e-05}
{"train_loss": 0.08324838429689407, "global_step": 132097, "epoch": 1484, "lr": 5.2246881900889956e-05}
{"train_loss": 0.10870573669672012, "global_step": 132098, "epoch": 1484, "lr": 5.2246302749788923e-05}
{"train_loss": 0.09660831093788147, "global_step": 132099, "epoch": 1484, "lr": 5.22457235983859e-05}
{"train_loss": 0.1744910031557083, "global_step": 132100, "epoch": 1484, "lr": 5.2245144446680974e-05}
{"train_loss": 0.11111180484294891, "global_step": 132101, "epoch": 1484, "lr": 5.22445652946742e-05}
{"train_loss": 0.08051377534866333, "global_step": 132102, "epoch": 1484, "lr": 5.224398614236568e-05}
{"train_loss": 0.1152682900428772, "global_step": 132103, "epoch": 1484, "lr": 5.224340698975548e-05}
{"train_loss": 0.16120100021362305, "global_step": 132104, "epoch": 1484, "lr": 5.22428278368437e-05}
{"train_loss": 0.11155383288860321, "global_step": 132105, "epoch": 1484, "lr": 5.224224868363038e-05}
{"train_loss": 0.12976861000061035, "global_step": 132106, "epoch": 1484, "lr": 5.2241669530115624e-05}
{"train_loss": 0.09989498555660248, "global_step": 132107, "epoch": 1484, "lr": 5.2241090376299506e-05}
{"train_loss": 0.13045793771743774, "global_step": 132108, "epoch": 1484, "lr": 5.22405112221821e-05}
{"train_loss": 0.1502968817949295, "global_step": 132109, "epoch": 1484, "lr": 5.223993206776348e-05}
{"train_loss": 0.11509986221790314, "global_step": 132110, "epoch": 1484, "lr": 5.2239352913043724e-05}
{"train_loss": 0.11212073266506195, "global_step": 132111, "epoch": 1484, "lr": 5.223877375802293e-05}
{"train_loss": 0.061777323484420776, "global_step": 132112, "epoch": 1484, "lr": 5.2238194602701165e-05}
{"train_loss": 0.14684492349624634, "global_step": 132113, "epoch": 1484, "lr": 5.223761544707848e-05}
{"train_loss": 0.20703169703483582, "global_step": 132114, "epoch": 1484, "lr": 5.2237036291154985e-05}
{"train_loss": 0.13875557482242584, "global_step": 132115, "epoch": 1484, "lr": 5.223645713493075e-05}
{"train_loss": 0.1469520777463913, "global_step": 132116, "epoch": 1484, "lr": 5.223587797840584e-05}
{"train_loss": 0.12298470735549927, "global_step": 132117, "epoch": 1484, "lr": 5.2235298821580345e-05}
{"train_loss": 0.091428741812706, "global_step": 132118, "epoch": 1484, "lr": 5.2234719664454354e-05}
{"train_loss": 0.11235553026199341, "global_step": 132119, "epoch": 1484, "lr": 5.2234140507027916e-05}
{"train_loss": 0.10543288290500641, "global_step": 132120, "epoch": 1484, "lr": 5.223356134930113e-05}
{"train_loss": 0.2050318568944931, "global_step": 132121, "epoch": 1484, "lr": 5.223298219127407e-05}
{"train_loss": 0.050608161836862564, "global_step": 132122, "epoch": 1484, "lr": 5.223240303294682e-05}
{"train_loss": 0.10261139273643494, "global_step": 132123, "epoch": 1484, "lr": 5.223182387431944e-05}
{"train_loss": 0.12309495359659195, "global_step": 132124, "epoch": 1484, "lr": 5.223124471539202e-05}
{"train_loss": 0.09032651036977768, "global_step": 132125, "epoch": 1484, "lr": 5.223066555616465e-05}
{"train_loss": 0.17194512486457825, "global_step": 132126, "epoch": 1484, "lr": 5.2230086396637375e-05}
{"train_loss": 0.10909892618656158, "global_step": 132127, "epoch": 1484, "lr": 5.2229507236810305e-05}
{"train_loss": 0.16979031264781952, "global_step": 132128, "epoch": 1484, "lr": 5.2228928076683506e-05}
{"train_loss": 0.16198058426380157, "global_step": 132129, "epoch": 1484, "lr": 5.2228348916257054e-05}
{"train_loss": 0.13690203428268433, "global_step": 132130, "epoch": 1484, "lr": 5.222776975553102e-05}
{"train_loss": 0.17100663483142853, "global_step": 132131, "epoch": 1484, "lr": 5.222719059450549e-05}
{"train_loss": 0.10947282612323761, "global_step": 132132, "epoch": 1484, "lr": 5.222661143318056e-05}
{"train_loss": 0.1059112474322319, "global_step": 132133, "epoch": 1484, "lr": 5.222603227155627e-05}
{"train_loss": 0.13711921870708466, "global_step": 132134, "epoch": 1484, "lr": 5.222545310963273e-05}
{"train_loss": 0.08534573018550873, "global_step": 132135, "epoch": 1484, "lr": 5.222487394740999e-05}
{"train_loss": 0.10043063759803772, "global_step": 132136, "epoch": 1484, "lr": 5.222429478488815e-05}
{"train_loss": 0.1716536283493042, "global_step": 132137, "epoch": 1484, "lr": 5.222371562206727e-05}
{"train_loss": 0.10122222453355789, "global_step": 132138, "epoch": 1484, "lr": 5.222313645894745e-05}
{"train_loss": 0.1626102179288864, "global_step": 132139, "epoch": 1484, "lr": 5.222255729552875e-05}
{"train_loss": 0.09150904417037964, "global_step": 132140, "epoch": 1484, "lr": 5.2221978131811265e-05}
{"train_loss": 0.10683219879865646, "global_step": 132141, "epoch": 1484, "lr": 5.2221398967795054e-05}
{"train_loss": 0.09602968394756317, "global_step": 132142, "epoch": 1484, "lr": 5.2220819803480206e-05}
{"train_loss": 0.1890862137079239, "global_step": 132143, "epoch": 1484, "lr": 5.222024063886679e-05}
{"train_loss": 0.16895397007465363, "global_step": 132144, "epoch": 1484, "lr": 5.22196614739549e-05}
{"train_loss": 0.15816892683506012, "global_step": 132145, "epoch": 1484, "lr": 5.2219082308744594e-05}
{"train_loss": 0.1424999237060547, "global_step": 132146, "epoch": 1484, "lr": 5.2218503143235965e-05}
{"train_loss": 0.12448756396770477, "global_step": 132147, "epoch": 1484, "lr": 5.2217923977429085e-05}
{"train_loss": 0.13874401152133942, "global_step": 132148, "epoch": 1484, "lr": 5.221734481132403e-05}
{"train_loss": 0.15524962544441223, "global_step": 132149, "epoch": 1484, "lr": 5.2216765644920886e-05}
{"train_loss": 0.129084974527359, "global_step": 132150, "epoch": 1484, "lr": 5.2216186478219716e-05}
{"train_loss": 0.10228802263736725, "global_step": 132151, "epoch": 1484, "lr": 5.2215607311220614e-05}
{"train_loss": 0.14845159649848938, "global_step": 132152, "epoch": 1484, "lr": 5.221502814392365e-05}
{"train_loss": 0.09804223477840424, "global_step": 132153, "epoch": 1484, "lr": 5.22144489763289e-05}
{"train_loss": 0.14170069992542267, "global_step": 132154, "epoch": 1484, "lr": 5.221386980843644e-05}
{"train_loss": 0.1302807629108429, "global_step": 132155, "epoch": 1484, "lr": 5.221329064024635e-05}
{"train_loss": 0.09204177558422089, "global_step": 132156, "epoch": 1484, "lr": 5.2212711471758726e-05}
{"train_loss": 0.10486158728599548, "global_step": 132157, "epoch": 1484, "lr": 5.221213230297362e-05}
{"train_loss": 0.12226254492998123, "global_step": 132158, "epoch": 1484, "lr": 5.221155313389111e-05}
{"train_loss": 0.11835784465074539, "global_step": 132159, "epoch": 1484, "lr": 5.22109739645113e-05}
{"train_loss": 0.12242918461561203, "global_step": 132160, "epoch": 1484, "lr": 5.221039479483424e-05}
{"train_loss": 0.07256972789764404, "global_step": 132161, "epoch": 1484, "lr": 5.2209815624860024e-05}
{"train_loss": 0.13920290768146515, "global_step": 132162, "epoch": 1484, "lr": 5.2209236454588725e-05}
{"train_loss": 0.13099581003189087, "global_step": 132163, "epoch": 1484, "lr": 5.220865728402042e-05}
{"train_loss": 0.1258518389855208, "global_step": 132164, "epoch": 1484, "lr": 5.22080781131552e-05, "val_loss": 5.705056667327881}
{"train_loss": 0.16202697157859802, "global_step": 132165, "epoch": 1485, "lr": 5.220749894199312e-05}
{"train_loss": 0.1368645280599594, "global_step": 132166, "epoch": 1485, "lr": 5.2206919770534266e-05}
{"train_loss": 0.06523341685533524, "global_step": 132167, "epoch": 1485, "lr": 5.220634059877872e-05}
{"train_loss": 0.12278147041797638, "global_step": 132168, "epoch": 1485, "lr": 5.220576142672656e-05}
{"train_loss": 0.08533865958452225, "global_step": 132169, "epoch": 1485, "lr": 5.220518225437786e-05}
{"train_loss": 0.1020079180598259, "global_step": 132170, "epoch": 1485, "lr": 5.2204603081732716e-05}
{"train_loss": 0.1096852570772171, "global_step": 132171, "epoch": 1485, "lr": 5.2204023908791176e-05}
{"train_loss": 0.09893565624952316, "global_step": 132172, "epoch": 1485, "lr": 5.220344473555333e-05}
{"train_loss": 0.14809289574623108, "global_step": 132173, "epoch": 1485, "lr": 5.220286556201927e-05}
{"train_loss": 0.13423024117946625, "global_step": 132174, "epoch": 1485, "lr": 5.2202286388189035e-05}
{"train_loss": 0.08738885074853897, "global_step": 132175, "epoch": 1485, "lr": 5.220170721406275e-05}
{"train_loss": 0.11082948744297028, "global_step": 132176, "epoch": 1485, "lr": 5.2201128039640465e-05}
{"train_loss": 0.29283004999160767, "global_step": 132177, "epoch": 1485, "lr": 5.220054886492227e-05}
{"train_loss": 0.2144748866558075, "global_step": 132178, "epoch": 1485, "lr": 5.219996968990824e-05}
{"train_loss": 0.22055187821388245, "global_step": 132179, "epoch": 1485, "lr": 5.219939051459845e-05}
{"train_loss": 0.11126194149255753, "global_step": 132180, "epoch": 1485, "lr": 5.219881133899297e-05}
{"train_loss": 0.1470990926027298, "global_step": 132181, "epoch": 1485, "lr": 5.219823216309189e-05}
{"train_loss": 0.1698535531759262, "global_step": 132182, "epoch": 1485, "lr": 5.2197652986895284e-05}
{"train_loss": 0.10301899164915085, "global_step": 132183, "epoch": 1485, "lr": 5.219707381040324e-05}
{"train_loss": 0.19600924849510193, "global_step": 132184, "epoch": 1485, "lr": 5.2196494633615825e-05}
{"train_loss": 0.09874381870031357, "global_step": 132185, "epoch": 1485, "lr": 5.219591545653311e-05}
{"train_loss": 0.14209601283073425, "global_step": 132186, "epoch": 1485, "lr": 5.219533627915518e-05}
{"train_loss": 0.13500693440437317, "global_step": 132187, "epoch": 1485, "lr": 5.2194757101482116e-05}
{"train_loss": 0.1339961737394333, "global_step": 132188, "epoch": 1485, "lr": 5.2194177923514e-05}
{"train_loss": 0.18485938012599945, "global_step": 132189, "epoch": 1485, "lr": 5.2193598745250896e-05}
{"train_loss": 0.11024123430252075, "global_step": 132190, "epoch": 1485, "lr": 5.2193019566692904e-05}
{"train_loss": 0.19529090821743011, "global_step": 132191, "epoch": 1485, "lr": 5.219244038784008e-05}
{"train_loss": 0.1014692634344101, "global_step": 132192, "epoch": 1485, "lr": 5.219186120869248e-05}
{"train_loss": 0.12607374787330627, "global_step": 132193, "epoch": 1485, "lr": 5.2191282029250243e-05}
{"train_loss": 0.12528741359710693, "global_step": 132194, "epoch": 1485, "lr": 5.2190702849513406e-05}
{"train_loss": 0.18886728584766388, "global_step": 132195, "epoch": 1485, "lr": 5.219012366948206e-05}
{"train_loss": 0.1629168838262558, "global_step": 132196, "epoch": 1485, "lr": 5.2189544489156275e-05}
{"train_loss": 0.13830047845840454, "global_step": 132197, "epoch": 1485, "lr": 5.218896530853613e-05}
{"train_loss": 0.11362098902463913, "global_step": 132198, "epoch": 1485, "lr": 5.2188386127621714e-05}
{"train_loss": 0.13625042140483856, "global_step": 132199, "epoch": 1485, "lr": 5.2187806946413085e-05}
{"train_loss": 0.16657032072544098, "global_step": 132200, "epoch": 1485, "lr": 5.218722776491034e-05}
{"train_loss": 0.13055288791656494, "global_step": 132201, "epoch": 1485, "lr": 5.218664858311355e-05}
{"train_loss": 0.14324851334095, "global_step": 132202, "epoch": 1485, "lr": 5.218606940102279e-05}
{"train_loss": 0.09830416738986969, "global_step": 132203, "epoch": 1485, "lr": 5.218549021863813e-05}
{"train_loss": 0.19393473863601685, "global_step": 132204, "epoch": 1485, "lr": 5.218491103595967e-05}
{"train_loss": 0.09165444225072861, "global_step": 132205, "epoch": 1485, "lr": 5.2184331852987466e-05}
{"train_loss": 0.12157569825649261, "global_step": 132206, "epoch": 1485, "lr": 5.218375266972161e-05}
{"train_loss": 0.07571887224912643, "global_step": 132207, "epoch": 1485, "lr": 5.2183173486162175e-05}
{"train_loss": 0.14044465124607086, "global_step": 132208, "epoch": 1485, "lr": 5.218259430230924e-05}
{"train_loss": 0.08501239866018295, "global_step": 132209, "epoch": 1485, "lr": 5.2182015118162896e-05}
{"train_loss": 0.10019015520811081, "global_step": 132210, "epoch": 1485, "lr": 5.2181435933723175e-05}
{"train_loss": 0.051847536116838455, "global_step": 132211, "epoch": 1485, "lr": 5.2180856748990205e-05}
{"train_loss": 0.17981339991092682, "global_step": 132212, "epoch": 1485, "lr": 5.218027756396404e-05}
{"train_loss": 0.12678392231464386, "global_step": 132213, "epoch": 1485, "lr": 5.2179698378644774e-05}
{"train_loss": 0.13329198956489563, "global_step": 132214, "epoch": 1485, "lr": 5.217911919303247e-05}
{"train_loss": 0.1433752328157425, "global_step": 132215, "epoch": 1485, "lr": 5.21785400071272e-05}
{"train_loss": 0.09500329196453094, "global_step": 132216, "epoch": 1485, "lr": 5.2177960820929064e-05}
{"train_loss": 0.12969492375850677, "global_step": 132217, "epoch": 1485, "lr": 5.2177381634438114e-05}
{"train_loss": 0.1374213546514511, "global_step": 132218, "epoch": 1485, "lr": 5.217680244765445e-05}
{"train_loss": 0.13354751467704773, "global_step": 132219, "epoch": 1485, "lr": 5.217622326057814e-05}
{"train_loss": 0.11112627387046814, "global_step": 132220, "epoch": 1485, "lr": 5.217564407320927e-05}
{"train_loss": 0.18334342539310455, "global_step": 132221, "epoch": 1485, "lr": 5.217506488554791e-05}
{"train_loss": 0.1467548906803131, "global_step": 132222, "epoch": 1485, "lr": 5.2174485697594135e-05}
{"train_loss": 0.11246631294488907, "global_step": 132223, "epoch": 1485, "lr": 5.2173906509348025e-05}
{"train_loss": 0.10523466020822525, "global_step": 132224, "epoch": 1485, "lr": 5.217332732080966e-05}
{"train_loss": 0.07111340016126633, "global_step": 132225, "epoch": 1485, "lr": 5.217274813197912e-05}
{"train_loss": 0.09681805968284607, "global_step": 132226, "epoch": 1485, "lr": 5.217216894285647e-05}
{"train_loss": 0.15739041566848755, "global_step": 132227, "epoch": 1485, "lr": 5.217158975344181e-05}
{"train_loss": 0.11802971363067627, "global_step": 132228, "epoch": 1485, "lr": 5.217101056373521e-05}
{"train_loss": 0.10831097513437271, "global_step": 132229, "epoch": 1485, "lr": 5.2170431373736736e-05}
{"train_loss": 0.07549817860126495, "global_step": 132230, "epoch": 1485, "lr": 5.216985218344648e-05}
{"train_loss": 0.13548918068408966, "global_step": 132231, "epoch": 1485, "lr": 5.21692729928645e-05}
{"train_loss": 0.16098113358020782, "global_step": 132232, "epoch": 1485, "lr": 5.2168693801990895e-05}
{"train_loss": 0.11279448121786118, "global_step": 132233, "epoch": 1485, "lr": 5.2168114610825746e-05}
{"train_loss": 0.15902380645275116, "global_step": 132234, "epoch": 1485, "lr": 5.2167535419369116e-05}
{"train_loss": 0.12488381564617157, "global_step": 132235, "epoch": 1485, "lr": 5.216695622762108e-05}
{"train_loss": 0.11323815584182739, "global_step": 132236, "epoch": 1485, "lr": 5.2166377035581726e-05}
{"train_loss": 0.11527887731790543, "global_step": 132237, "epoch": 1485, "lr": 5.216579784325113e-05}
{"train_loss": 0.1699342131614685, "global_step": 132238, "epoch": 1485, "lr": 5.216521865062937e-05}
{"train_loss": 0.11788462847471237, "global_step": 132239, "epoch": 1485, "lr": 5.2164639457716525e-05}
{"train_loss": 0.11692637950181961, "global_step": 132240, "epoch": 1485, "lr": 5.2164060264512656e-05}
{"train_loss": 0.22891664505004883, "global_step": 132241, "epoch": 1485, "lr": 5.216348107101787e-05}
{"train_loss": 0.1921578198671341, "global_step": 132242, "epoch": 1485, "lr": 5.216290187723223e-05}
{"train_loss": 0.12211771309375763, "global_step": 132243, "epoch": 1485, "lr": 5.2162322683155816e-05}
{"train_loss": 0.0928497239947319, "global_step": 132244, "epoch": 1485, "lr": 5.2161743488788694e-05}
{"train_loss": 0.0880894809961319, "global_step": 132245, "epoch": 1485, "lr": 5.216116429413096e-05}
{"train_loss": 0.1322081983089447, "global_step": 132246, "epoch": 1485, "lr": 5.216058509918268e-05}
{"train_loss": 0.09203682839870453, "global_step": 132247, "epoch": 1485, "lr": 5.216000590394394e-05}
{"train_loss": 0.09416157007217407, "global_step": 132248, "epoch": 1485, "lr": 5.2159426708414805e-05}
{"train_loss": 0.14273284375667572, "global_step": 132249, "epoch": 1485, "lr": 5.2158847512595366e-05}
{"train_loss": 0.2304668426513672, "global_step": 132250, "epoch": 1485, "lr": 5.2158268316485706e-05}
{"train_loss": 0.15624023973941803, "global_step": 132251, "epoch": 1485, "lr": 5.215768912008589e-05}
{"train_loss": 0.09813187271356583, "global_step": 132252, "epoch": 1485, "lr": 5.215710992339598e-05}
{"train_loss": 0.13256602805484546, "global_step": 132253, "epoch": 1485, "lr": 5.215653072641609e-05, "val_loss": 5.855353355407715, "train_action_mse_error": 35.77189636230469}
{"train_loss": 0.19977062940597534, "global_step": 132254, "epoch": 1486, "lr": 5.215595152914627e-05}
{"train_loss": 0.11615696549415588, "global_step": 132255, "epoch": 1486, "lr": 5.2155372331586614e-05}
{"train_loss": 0.18109484016895294, "global_step": 132256, "epoch": 1486, "lr": 5.2154793133737204e-05}
{"train_loss": 0.11252286285161972, "global_step": 132257, "epoch": 1486, "lr": 5.2154213935598094e-05}
{"train_loss": 0.1306561976671219, "global_step": 132258, "epoch": 1486, "lr": 5.215363473716939e-05}
{"train_loss": 0.1228482648730278, "global_step": 132259, "epoch": 1486, "lr": 5.2153055538451144e-05}
{"train_loss": 0.14135532081127167, "global_step": 132260, "epoch": 1486, "lr": 5.2152476339443446e-05}
{"train_loss": 0.13222770392894745, "global_step": 132261, "epoch": 1486, "lr": 5.215189714014639e-05}
{"train_loss": 0.21093423664569855, "global_step": 132262, "epoch": 1486, "lr": 5.215131794056002e-05}
{"train_loss": 0.19705702364444733, "global_step": 132263, "epoch": 1486, "lr": 5.2150738740684436e-05}
{"train_loss": 0.09997595101594925, "global_step": 132264, "epoch": 1486, "lr": 5.215015954051972e-05}
{"train_loss": 0.09528151154518127, "global_step": 132265, "epoch": 1486, "lr": 5.214958034006593e-05}
{"train_loss": 0.1634245365858078, "global_step": 132266, "epoch": 1486, "lr": 5.214900113932315e-05}
{"train_loss": 0.06592535972595215, "global_step": 132267, "epoch": 1486, "lr": 5.214842193829148e-05}
{"train_loss": 0.10262270271778107, "global_step": 132268, "epoch": 1486, "lr": 5.2147842736970974e-05}
{"train_loss": 0.11165878921747208, "global_step": 132269, "epoch": 1486, "lr": 5.214726353536171e-05}
{"train_loss": 0.1480831503868103, "global_step": 132270, "epoch": 1486, "lr": 5.214668433346378e-05}
{"train_loss": 0.11083205789327621, "global_step": 132271, "epoch": 1486, "lr": 5.214610513127726e-05}
{"train_loss": 0.1176154837012291, "global_step": 132272, "epoch": 1486, "lr": 5.214552592880221e-05}
{"train_loss": 0.13322044909000397, "global_step": 132273, "epoch": 1486, "lr": 5.2144946726038724e-05}
{"train_loss": 0.10367876291275024, "global_step": 132274, "epoch": 1486, "lr": 5.214436752298688e-05}
{"train_loss": 0.1694939285516739, "global_step": 132275, "epoch": 1486, "lr": 5.214378831964675e-05}
{"train_loss": 0.10323014855384827, "global_step": 132276, "epoch": 1486, "lr": 5.214320911601841e-05}
{"train_loss": 0.07153906673192978, "global_step": 132277, "epoch": 1486, "lr": 5.2142629912101934e-05}
{"train_loss": 0.11408517509698868, "global_step": 132278, "epoch": 1486, "lr": 5.2142050707897425e-05}
{"train_loss": 0.10148198157548904, "global_step": 132279, "epoch": 1486, "lr": 5.214147150340494e-05}
{"train_loss": 0.12914320826530457, "global_step": 132280, "epoch": 1486, "lr": 5.2140892298624554e-05}
{"train_loss": 0.10765185952186584, "global_step": 132281, "epoch": 1486, "lr": 5.2140313093556356e-05}
{"train_loss": 0.15650740265846252, "global_step": 132282, "epoch": 1486, "lr": 5.213973388820042e-05}
{"train_loss": 0.0723438486456871, "global_step": 132283, "epoch": 1486, "lr": 5.213915468255681e-05}
{"train_loss": 0.050482042133808136, "global_step": 132284, "epoch": 1486, "lr": 5.213857547662563e-05}
{"train_loss": 0.04755126312375069, "global_step": 132285, "epoch": 1486, "lr": 5.2137996270406944e-05}
{"train_loss": 0.11748208105564117, "global_step": 132286, "epoch": 1486, "lr": 5.213741706390083e-05}
{"train_loss": 0.15905092656612396, "global_step": 132287, "epoch": 1486, "lr": 5.2136837857107354e-05}
{"train_loss": 0.13529393076896667, "global_step": 132288, "epoch": 1486, "lr": 5.2136258650026625e-05}
{"train_loss": 0.1230732798576355, "global_step": 132289, "epoch": 1486, "lr": 5.213567944265869e-05}
{"train_loss": 0.13768255710601807, "global_step": 132290, "epoch": 1486, "lr": 5.213510023500364e-05}
{"train_loss": 0.2270297259092331, "global_step": 132291, "epoch": 1486, "lr": 5.2134521027061555e-05}
{"train_loss": 0.17267851531505585, "global_step": 132292, "epoch": 1486, "lr": 5.213394181883251e-05}
{"train_loss": 0.149314746260643, "global_step": 132293, "epoch": 1486, "lr": 5.213336261031658e-05}
{"train_loss": 0.07856211066246033, "global_step": 132294, "epoch": 1486, "lr": 5.213278340151384e-05}
{"train_loss": 0.0840669795870781, "global_step": 132295, "epoch": 1486, "lr": 5.2132204192424384e-05}
{"train_loss": 0.1989295333623886, "global_step": 132296, "epoch": 1486, "lr": 5.213162498304828e-05}
{"train_loss": 0.06644616276025772, "global_step": 132297, "epoch": 1486, "lr": 5.2131045773385603e-05}
{"train_loss": 0.13683605194091797, "global_step": 132298, "epoch": 1486, "lr": 5.213046656343642e-05}
{"train_loss": 0.10743463784456253, "global_step": 132299, "epoch": 1486, "lr": 5.2129887353200835e-05}
{"train_loss": 0.15215444564819336, "global_step": 132300, "epoch": 1486, "lr": 5.212930814267891e-05}
{"train_loss": 0.08312007784843445, "global_step": 132301, "epoch": 1486, "lr": 5.212872893187072e-05}
{"train_loss": 0.09703399240970612, "global_step": 132302, "epoch": 1486, "lr": 5.212814972077635e-05}
{"train_loss": 0.07730615139007568, "global_step": 132303, "epoch": 1486, "lr": 5.212757050939589e-05}
{"train_loss": 0.1331372857093811, "global_step": 132304, "epoch": 1486, "lr": 5.212699129772939e-05}
{"train_loss": 0.17450165748596191, "global_step": 132305, "epoch": 1486, "lr": 5.212641208577694e-05}
{"train_loss": 0.15988729894161224, "global_step": 132306, "epoch": 1486, "lr": 5.212583287353863e-05}
{"train_loss": 0.10272616147994995, "global_step": 132307, "epoch": 1486, "lr": 5.2125253661014516e-05}
{"train_loss": 0.08428288251161575, "global_step": 132308, "epoch": 1486, "lr": 5.21246744482047e-05}
{"train_loss": 0.10394245386123657, "global_step": 132309, "epoch": 1486, "lr": 5.212409523510924e-05}
{"train_loss": 0.16600003838539124, "global_step": 132310, "epoch": 1486, "lr": 5.2123516021728225e-05}
{"train_loss": 0.20143693685531616, "global_step": 132311, "epoch": 1486, "lr": 5.212293680806173e-05}
{"train_loss": 0.10160306096076965, "global_step": 132312, "epoch": 1486, "lr": 5.212235759410983e-05}
{"train_loss": 0.11752671003341675, "global_step": 132313, "epoch": 1486, "lr": 5.21217783798726e-05}
{"train_loss": 0.14358414709568024, "global_step": 132314, "epoch": 1486, "lr": 5.2121199165350124e-05}
{"train_loss": 0.12159746140241623, "global_step": 132315, "epoch": 1486, "lr": 5.212061995054248e-05}
{"train_loss": 0.08389919251203537, "global_step": 132316, "epoch": 1486, "lr": 5.212004073544975e-05}
{"train_loss": 0.12630434334278107, "global_step": 132317, "epoch": 1486, "lr": 5.2119461520072e-05}
{"train_loss": 0.12357007712125778, "global_step": 132318, "epoch": 1486, "lr": 5.2118882304409314e-05}
{"train_loss": 0.09181015193462372, "global_step": 132319, "epoch": 1486, "lr": 5.211830308846176e-05}
{"train_loss": 0.1338992416858673, "global_step": 132320, "epoch": 1486, "lr": 5.2117723872229454e-05}
{"train_loss": 0.1364346295595169, "global_step": 132321, "epoch": 1486, "lr": 5.2117144655712426e-05}
{"train_loss": 0.059502676129341125, "global_step": 132322, "epoch": 1486, "lr": 5.211656543891078e-05}
{"train_loss": 0.08366716653108597, "global_step": 132323, "epoch": 1486, "lr": 5.211598622182459e-05}
{"train_loss": 0.12849131226539612, "global_step": 132324, "epoch": 1486, "lr": 5.211540700445392e-05}
{"train_loss": 0.10672944039106369, "global_step": 132325, "epoch": 1486, "lr": 5.211482778679887e-05}
{"train_loss": 0.10315659642219543, "global_step": 132326, "epoch": 1486, "lr": 5.21142485688595e-05}
{"train_loss": 0.14426951110363007, "global_step": 132327, "epoch": 1486, "lr": 5.21136693506359e-05}
{"train_loss": 0.1045074537396431, "global_step": 132328, "epoch": 1486, "lr": 5.211309013212814e-05}
{"train_loss": 0.10936792939901352, "global_step": 132329, "epoch": 1486, "lr": 5.21125109133363e-05}
{"train_loss": 0.14855080842971802, "global_step": 132330, "epoch": 1486, "lr": 5.211193169426046e-05}
{"train_loss": 0.16874851286411285, "global_step": 132331, "epoch": 1486, "lr": 5.2111352474900697e-05}
{"train_loss": 0.11378797888755798, "global_step": 132332, "epoch": 1486, "lr": 5.2110773255257094e-05}
{"train_loss": 0.09346117824316025, "global_step": 132333, "epoch": 1486, "lr": 5.2110194035329715e-05}
{"train_loss": 0.11384543776512146, "global_step": 132334, "epoch": 1486, "lr": 5.210961481511865e-05}
{"train_loss": 0.16160747408866882, "global_step": 132335, "epoch": 1486, "lr": 5.210903559462398e-05}
{"train_loss": 0.13117988407611847, "global_step": 132336, "epoch": 1486, "lr": 5.210845637384576e-05}
{"train_loss": 0.11633459478616714, "global_step": 132337, "epoch": 1486, "lr": 5.2107877152784076e-05}
{"train_loss": 0.11599243432283401, "global_step": 132338, "epoch": 1486, "lr": 5.2107297931439035e-05}
{"train_loss": 0.09845706820487976, "global_step": 132339, "epoch": 1486, "lr": 5.21067187098107e-05}
{"train_loss": 0.1418365091085434, "global_step": 132340, "epoch": 1486, "lr": 5.210613948789913e-05}
{"train_loss": 0.12851868569850922, "global_step": 132341, "epoch": 1486, "lr": 5.210556026570441e-05}
{"train_loss": 0.12382220548022999, "global_step": 132342, "epoch": 1486, "lr": 5.210498104322663e-05, "val_loss": 5.536613941192627}
{"train_loss": 0.1460481435060501, "global_step": 132343, "epoch": 1487, "lr": 5.2104401820465865e-05}
{"train_loss": 0.18961729109287262, "global_step": 132344, "epoch": 1487, "lr": 5.210382259742218e-05}
{"train_loss": 0.12556099891662598, "global_step": 132345, "epoch": 1487, "lr": 5.210324337409567e-05}
{"train_loss": 0.06977245211601257, "global_step": 132346, "epoch": 1487, "lr": 5.2102664150486394e-05}
{"train_loss": 0.14447352290153503, "global_step": 132347, "epoch": 1487, "lr": 5.2102084926594454e-05}
{"train_loss": 0.12121470272541046, "global_step": 132348, "epoch": 1487, "lr": 5.2101505702419896e-05}
{"train_loss": 0.10864508897066116, "global_step": 132349, "epoch": 1487, "lr": 5.210092647796283e-05}
{"train_loss": 0.06190227344632149, "global_step": 132350, "epoch": 1487, "lr": 5.2100347253223314e-05}
{"train_loss": 0.1127198189496994, "global_step": 132351, "epoch": 1487, "lr": 5.209976802820143e-05}
{"train_loss": 0.07599544525146484, "global_step": 132352, "epoch": 1487, "lr": 5.209918880289727e-05}
{"train_loss": 0.18136458098888397, "global_step": 132353, "epoch": 1487, "lr": 5.2098609577310885e-05}
{"train_loss": 0.08435938507318497, "global_step": 132354, "epoch": 1487, "lr": 5.209803035144238e-05}
{"train_loss": 0.08719572424888611, "global_step": 132355, "epoch": 1487, "lr": 5.2097451125291806e-05}
{"train_loss": 0.12594756484031677, "global_step": 132356, "epoch": 1487, "lr": 5.2096871898859254e-05}
{"train_loss": 0.0691513642668724, "global_step": 132357, "epoch": 1487, "lr": 5.209629267214482e-05}
{"train_loss": 0.1106109544634819, "global_step": 132358, "epoch": 1487, "lr": 5.2095713445148576e-05}
{"train_loss": 0.12940333783626556, "global_step": 132359, "epoch": 1487, "lr": 5.2095134217870564e-05}
{"train_loss": 0.13109806180000305, "global_step": 132360, "epoch": 1487, "lr": 5.2094554990310896e-05}
{"train_loss": 0.08674626797437668, "global_step": 132361, "epoch": 1487, "lr": 5.2093975762469634e-05}
{"train_loss": 0.12562081217765808, "global_step": 132362, "epoch": 1487, "lr": 5.209339653434687e-05}
{"train_loss": 0.13476398587226868, "global_step": 132363, "epoch": 1487, "lr": 5.209281730594267e-05}
{"train_loss": 0.048762544989585876, "global_step": 132364, "epoch": 1487, "lr": 5.2092238077257116e-05}
{"train_loss": 0.10056571662425995, "global_step": 132365, "epoch": 1487, "lr": 5.209165884829029e-05}
{"train_loss": 0.0869804173707962, "global_step": 132366, "epoch": 1487, "lr": 5.209107961904226e-05}
{"train_loss": 0.07995185256004333, "global_step": 132367, "epoch": 1487, "lr": 5.209050038951312e-05}
{"train_loss": 0.07101651281118393, "global_step": 132368, "epoch": 1487, "lr": 5.208992115970293e-05}
{"train_loss": 0.1560819000005722, "global_step": 132369, "epoch": 1487, "lr": 5.2089341929611777e-05}
{"train_loss": 0.12633071839809418, "global_step": 132370, "epoch": 1487, "lr": 5.2088762699239736e-05}
{"train_loss": 0.14717721939086914, "global_step": 132371, "epoch": 1487, "lr": 5.208818346858688e-05}
{"train_loss": 0.12043110281229019, "global_step": 132372, "epoch": 1487, "lr": 5.208760423765331e-05}
{"train_loss": 0.11039000749588013, "global_step": 132373, "epoch": 1487, "lr": 5.208702500643906e-05}
{"train_loss": 0.12181176245212555, "global_step": 132374, "epoch": 1487, "lr": 5.2086445774944256e-05}
{"train_loss": 0.10160289704799652, "global_step": 132375, "epoch": 1487, "lr": 5.2085866543168947e-05}
{"train_loss": 0.1147339940071106, "global_step": 132376, "epoch": 1487, "lr": 5.208528731111323e-05}
{"train_loss": 0.13998258113861084, "global_step": 132377, "epoch": 1487, "lr": 5.2084708078777156e-05}
{"train_loss": 0.1524149626493454, "global_step": 132378, "epoch": 1487, "lr": 5.208412884616084e-05}
{"train_loss": 0.091055728495121, "global_step": 132379, "epoch": 1487, "lr": 5.2083549613264327e-05}
{"train_loss": 0.09292055666446686, "global_step": 132380, "epoch": 1487, "lr": 5.2082970380087706e-05}
{"train_loss": 0.08421213179826736, "global_step": 132381, "epoch": 1487, "lr": 5.208239114663105e-05}
{"train_loss": 0.12158790230751038, "global_step": 132382, "epoch": 1487, "lr": 5.208181191289445e-05}
{"train_loss": 0.09596845507621765, "global_step": 132383, "epoch": 1487, "lr": 5.208123267887797e-05}
{"train_loss": 0.0932045727968216, "global_step": 132384, "epoch": 1487, "lr": 5.20806534445817e-05}
{"train_loss": 0.1348404735326767, "global_step": 132385, "epoch": 1487, "lr": 5.2080074210005704e-05}
{"train_loss": 0.08933018147945404, "global_step": 132386, "epoch": 1487, "lr": 5.207949497515007e-05}
{"train_loss": 0.10392344743013382, "global_step": 132387, "epoch": 1487, "lr": 5.207891574001487e-05}
{"train_loss": 0.09458683431148529, "global_step": 132388, "epoch": 1487, "lr": 5.207833650460019e-05}
{"train_loss": 0.10384085029363632, "global_step": 132389, "epoch": 1487, "lr": 5.207775726890609e-05}
{"train_loss": 0.16539466381072998, "global_step": 132390, "epoch": 1487, "lr": 5.2077178032932685e-05}
{"train_loss": 0.10952509939670563, "global_step": 132391, "epoch": 1487, "lr": 5.207659879668e-05}
{"train_loss": 0.14404389262199402, "global_step": 132392, "epoch": 1487, "lr": 5.207601956014816e-05}
{"train_loss": 0.12016956508159637, "global_step": 132393, "epoch": 1487, "lr": 5.207544032333722e-05}
{"train_loss": 0.12713950872421265, "global_step": 132394, "epoch": 1487, "lr": 5.2074861086247265e-05}
{"train_loss": 0.09526614844799042, "global_step": 132395, "epoch": 1487, "lr": 5.207428184887836e-05}
{"train_loss": 0.11598249524831772, "global_step": 132396, "epoch": 1487, "lr": 5.2073702611230604e-05}
{"train_loss": 0.08384154736995697, "global_step": 132397, "epoch": 1487, "lr": 5.207312337330408e-05}
{"train_loss": 0.19399785995483398, "global_step": 132398, "epoch": 1487, "lr": 5.207254413509882e-05}
{"train_loss": 0.20905135571956635, "global_step": 132399, "epoch": 1487, "lr": 5.2071964896614956e-05}
{"train_loss": 0.15011200308799744, "global_step": 132400, "epoch": 1487, "lr": 5.2071385657852526e-05}
{"train_loss": 0.09098244458436966, "global_step": 132401, "epoch": 1487, "lr": 5.2070806418811625e-05}
{"train_loss": 0.13109678030014038, "global_step": 132402, "epoch": 1487, "lr": 5.207022717949233e-05}
{"train_loss": 0.1782674938440323, "global_step": 132403, "epoch": 1487, "lr": 5.2069647939894717e-05}
{"train_loss": 0.12776078283786774, "global_step": 132404, "epoch": 1487, "lr": 5.206906870001887e-05}
{"train_loss": 0.13962729275226593, "global_step": 132405, "epoch": 1487, "lr": 5.2068489459864854e-05}
{"train_loss": 0.19415682554244995, "global_step": 132406, "epoch": 1487, "lr": 5.206791021943277e-05}
{"train_loss": 0.08119247108697891, "global_step": 132407, "epoch": 1487, "lr": 5.206733097872266e-05}
{"train_loss": 0.16105125844478607, "global_step": 132408, "epoch": 1487, "lr": 5.206675173773464e-05}
{"train_loss": 0.09081143885850906, "global_step": 132409, "epoch": 1487, "lr": 5.206617249646875e-05}
{"train_loss": 0.17033305764198303, "global_step": 132410, "epoch": 1487, "lr": 5.20655932549251e-05}
{"train_loss": 0.08952460438013077, "global_step": 132411, "epoch": 1487, "lr": 5.2065014013103764e-05}
{"train_loss": 0.13414283096790314, "global_step": 132412, "epoch": 1487, "lr": 5.20644347710048e-05}
{"train_loss": 0.10232584178447723, "global_step": 132413, "epoch": 1487, "lr": 5.20638555286283e-05}
{"train_loss": 0.11072047054767609, "global_step": 132414, "epoch": 1487, "lr": 5.206327628597435e-05}
{"train_loss": 0.13418111205101013, "global_step": 132415, "epoch": 1487, "lr": 5.2062697043043006e-05}
{"train_loss": 0.10128908604383469, "global_step": 132416, "epoch": 1487, "lr": 5.206211779983438e-05}
{"train_loss": 0.09651994705200195, "global_step": 132417, "epoch": 1487, "lr": 5.206153855634851e-05}
{"train_loss": 0.1024533361196518, "global_step": 132418, "epoch": 1487, "lr": 5.206095931258549e-05}
{"train_loss": 0.16052955389022827, "global_step": 132419, "epoch": 1487, "lr": 5.20603800685454e-05}
{"train_loss": 0.1567491739988327, "global_step": 132420, "epoch": 1487, "lr": 5.205980082422832e-05}
{"train_loss": 0.0999564453959465, "global_step": 132421, "epoch": 1487, "lr": 5.205922157963432e-05}
{"train_loss": 0.17638161778450012, "global_step": 132422, "epoch": 1487, "lr": 5.2058642334763485e-05}
{"train_loss": 0.07027047127485275, "global_step": 132423, "epoch": 1487, "lr": 5.20580630896159e-05}
{"train_loss": 0.16228288412094116, "global_step": 132424, "epoch": 1487, "lr": 5.205748384419162e-05}
{"train_loss": 0.12445376813411713, "global_step": 132425, "epoch": 1487, "lr": 5.205690459849074e-05}
{"train_loss": 0.07941734790802002, "global_step": 132426, "epoch": 1487, "lr": 5.2056325352513334e-05}
{"train_loss": 0.11204053461551666, "global_step": 132427, "epoch": 1487, "lr": 5.2055746106259487e-05}
{"train_loss": 0.12965618073940277, "global_step": 132428, "epoch": 1487, "lr": 5.2055166859729256e-05}
{"train_loss": 0.11017292737960815, "global_step": 132429, "epoch": 1487, "lr": 5.2054587612922737e-05}
{"train_loss": 0.09861469268798828, "global_step": 132430, "epoch": 1487, "lr": 5.205400836584e-05}
{"train_loss": 0.11869038754467214, "global_step": 132431, "epoch": 1487, "lr": 5.2053429118481144e-05, "val_loss": 5.5892109870910645}
{"train_loss": 0.15467356145381927, "global_step": 132432, "epoch": 1488, "lr": 5.205284987084622e-05}
{"train_loss": 0.06337971240282059, "global_step": 132433, "epoch": 1488, "lr": 5.2052270622935305e-05}
{"train_loss": 0.09303944557905197, "global_step": 132434, "epoch": 1488, "lr": 5.20516913747485e-05}
{"train_loss": 0.12756319344043732, "global_step": 132435, "epoch": 1488, "lr": 5.205111212628587e-05}
{"train_loss": 0.14492785930633545, "global_step": 132436, "epoch": 1488, "lr": 5.2050532877547506e-05}
{"train_loss": 0.19556927680969238, "global_step": 132437, "epoch": 1488, "lr": 5.2049953628533456e-05}
{"train_loss": 0.08307217061519623, "global_step": 132438, "epoch": 1488, "lr": 5.204937437924382e-05}
{"train_loss": 0.11972346901893616, "global_step": 132439, "epoch": 1488, "lr": 5.2048795129678666e-05}
{"train_loss": 0.13000360131263733, "global_step": 132440, "epoch": 1488, "lr": 5.204821587983807e-05}
{"train_loss": 0.19753798842430115, "global_step": 132441, "epoch": 1488, "lr": 5.204763662972213e-05}
{"train_loss": 0.12220221012830734, "global_step": 132442, "epoch": 1488, "lr": 5.20470573793309e-05}
{"train_loss": 0.06426288932561874, "global_step": 132443, "epoch": 1488, "lr": 5.204647812866448e-05}
{"train_loss": 0.1282244175672531, "global_step": 132444, "epoch": 1488, "lr": 5.204589887772292e-05}
{"train_loss": 0.14966215193271637, "global_step": 132445, "epoch": 1488, "lr": 5.204531962650632e-05}
{"train_loss": 0.06994755566120148, "global_step": 132446, "epoch": 1488, "lr": 5.2044740375014746e-05}
{"train_loss": 0.1178024411201477, "global_step": 132447, "epoch": 1488, "lr": 5.2044161123248294e-05}
{"train_loss": 0.1258353590965271, "global_step": 132448, "epoch": 1488, "lr": 5.2043581871207016e-05}
{"train_loss": 0.10454532504081726, "global_step": 132449, "epoch": 1488, "lr": 5.2043002618891015e-05}
{"train_loss": 0.07563512027263641, "global_step": 132450, "epoch": 1488, "lr": 5.204242336630035e-05}
{"train_loss": 0.18289634585380554, "global_step": 132451, "epoch": 1488, "lr": 5.2041844113435114e-05}
{"train_loss": 0.1412474662065506, "global_step": 132452, "epoch": 1488, "lr": 5.204126486029537e-05}
{"train_loss": 0.13758601248264313, "global_step": 132453, "epoch": 1488, "lr": 5.20406856068812e-05}
{"train_loss": 0.07648688554763794, "global_step": 132454, "epoch": 1488, "lr": 5.2040106353192695e-05}
{"train_loss": 0.13679565489292145, "global_step": 132455, "epoch": 1488, "lr": 5.203952709922993e-05}
{"train_loss": 0.0734802857041359, "global_step": 132456, "epoch": 1488, "lr": 5.203894784499297e-05}
{"train_loss": 0.12354561686515808, "global_step": 132457, "epoch": 1488, "lr": 5.203836859048189e-05}
{"train_loss": 0.14655433595180511, "global_step": 132458, "epoch": 1488, "lr": 5.203778933569678e-05}
{"train_loss": 0.20925725996494293, "global_step": 132459, "epoch": 1488, "lr": 5.203721008063771e-05}
{"train_loss": 0.08325523883104324, "global_step": 132460, "epoch": 1488, "lr": 5.2036630825304754e-05}
{"train_loss": 0.19318316876888275, "global_step": 132461, "epoch": 1488, "lr": 5.203605156969802e-05}
{"train_loss": 0.127971351146698, "global_step": 132462, "epoch": 1488, "lr": 5.2035472313817545e-05}
{"train_loss": 0.14362011849880219, "global_step": 132463, "epoch": 1488, "lr": 5.2034893057663434e-05}
{"train_loss": 0.12684012949466705, "global_step": 132464, "epoch": 1488, "lr": 5.2034313801235746e-05}
{"train_loss": 0.09455903619527817, "global_step": 132465, "epoch": 1488, "lr": 5.2033734544534574e-05}
{"train_loss": 0.1541927009820938, "global_step": 132466, "epoch": 1488, "lr": 5.203315528756e-05}
{"train_loss": 0.07894433289766312, "global_step": 132467, "epoch": 1488, "lr": 5.2032576030312085e-05}
{"train_loss": 0.07888557016849518, "global_step": 132468, "epoch": 1488, "lr": 5.203199677279091e-05}
{"train_loss": 0.12992770969867706, "global_step": 132469, "epoch": 1488, "lr": 5.203141751499657e-05}
{"train_loss": 0.15107619762420654, "global_step": 132470, "epoch": 1488, "lr": 5.203083825692913e-05}
{"train_loss": 0.166842982172966, "global_step": 132471, "epoch": 1488, "lr": 5.2030258998588664e-05}
{"train_loss": 0.1224995031952858, "global_step": 132472, "epoch": 1488, "lr": 5.202967973997526e-05}
{"train_loss": 0.27959567308425903, "global_step": 132473, "epoch": 1488, "lr": 5.202910048108898e-05}
{"train_loss": 0.08227939903736115, "global_step": 132474, "epoch": 1488, "lr": 5.2028521221929926e-05}
{"train_loss": 0.12321785092353821, "global_step": 132475, "epoch": 1488, "lr": 5.2027941962498164e-05}
{"train_loss": 0.06185990944504738, "global_step": 132476, "epoch": 1488, "lr": 5.2027362702793745e-05}
{"train_loss": 0.176752507686615, "global_step": 132477, "epoch": 1488, "lr": 5.20267834428168e-05}
{"train_loss": 0.12084250152111053, "global_step": 132478, "epoch": 1488, "lr": 5.2026204182567365e-05}
{"train_loss": 0.09439411014318466, "global_step": 132479, "epoch": 1488, "lr": 5.2025624922045525e-05}
{"train_loss": 0.16072267293930054, "global_step": 132480, "epoch": 1488, "lr": 5.202504566125137e-05}
{"train_loss": 0.1789039969444275, "global_step": 132481, "epoch": 1488, "lr": 5.202446640018498e-05}
{"train_loss": 0.19844771921634674, "global_step": 132482, "epoch": 1488, "lr": 5.2023887138846406e-05}
{"train_loss": 0.13298805058002472, "global_step": 132483, "epoch": 1488, "lr": 5.202330787723575e-05}
{"train_loss": 0.17123158276081085, "global_step": 132484, "epoch": 1488, "lr": 5.20227286153531e-05}
{"train_loss": 0.14637665450572968, "global_step": 132485, "epoch": 1488, "lr": 5.2022149353198514e-05}
{"train_loss": 0.12704741954803467, "global_step": 132486, "epoch": 1488, "lr": 5.2021570090772064e-05}
{"train_loss": 0.15747205913066864, "global_step": 132487, "epoch": 1488, "lr": 5.2020990828073844e-05}
{"train_loss": 0.056541893631219864, "global_step": 132488, "epoch": 1488, "lr": 5.202041156510393e-05}
{"train_loss": 0.12378346920013428, "global_step": 132489, "epoch": 1488, "lr": 5.2019832301862404e-05}
{"train_loss": 0.15862733125686646, "global_step": 132490, "epoch": 1488, "lr": 5.2019253038349325e-05}
{"train_loss": 0.08774703741073608, "global_step": 132491, "epoch": 1488, "lr": 5.201867377456478e-05}
{"train_loss": 0.11779653280973434, "global_step": 132492, "epoch": 1488, "lr": 5.201809451050885e-05}
{"train_loss": 0.11813605576753616, "global_step": 132493, "epoch": 1488, "lr": 5.2017515246181624e-05}
{"train_loss": 0.17833584547042847, "global_step": 132494, "epoch": 1488, "lr": 5.201693598158317e-05}
{"train_loss": 0.0923604965209961, "global_step": 132495, "epoch": 1488, "lr": 5.201635671671355e-05}
{"train_loss": 0.16824810206890106, "global_step": 132496, "epoch": 1488, "lr": 5.2015777451572856e-05}
{"train_loss": 0.10307114571332932, "global_step": 132497, "epoch": 1488, "lr": 5.2015198186161176e-05}
{"train_loss": 0.08556802570819855, "global_step": 132498, "epoch": 1488, "lr": 5.201461892047856e-05}
{"train_loss": 0.17268887162208557, "global_step": 132499, "epoch": 1488, "lr": 5.2014039654525113e-05}
{"train_loss": 0.11771450936794281, "global_step": 132500, "epoch": 1488, "lr": 5.201346038830089e-05}
{"train_loss": 0.10208538919687271, "global_step": 132501, "epoch": 1488, "lr": 5.201288112180599e-05}
{"train_loss": 0.11858820915222168, "global_step": 132502, "epoch": 1488, "lr": 5.201230185504049e-05}
{"train_loss": 0.1069209948182106, "global_step": 132503, "epoch": 1488, "lr": 5.2011722588004453e-05}
{"train_loss": 0.07658375799655914, "global_step": 132504, "epoch": 1488, "lr": 5.2011143320697975e-05}
{"train_loss": 0.07410728186368942, "global_step": 132505, "epoch": 1488, "lr": 5.201056405312112e-05}
{"train_loss": 0.06409180164337158, "global_step": 132506, "epoch": 1488, "lr": 5.200998478527396e-05}
{"train_loss": 0.12431923300027847, "global_step": 132507, "epoch": 1488, "lr": 5.200940551715659e-05}
{"train_loss": 0.1400417983531952, "global_step": 132508, "epoch": 1488, "lr": 5.200882624876907e-05}
{"train_loss": 0.1322125345468521, "global_step": 132509, "epoch": 1488, "lr": 5.20082469801115e-05}
{"train_loss": 0.10746870189905167, "global_step": 132510, "epoch": 1488, "lr": 5.2007667711183936e-05}
{"train_loss": 0.12000242620706558, "global_step": 132511, "epoch": 1488, "lr": 5.2007088441986474e-05}
{"train_loss": 0.1290488839149475, "global_step": 132512, "epoch": 1488, "lr": 5.200650917251918e-05}
{"train_loss": 0.12955501675605774, "global_step": 132513, "epoch": 1488, "lr": 5.2005929902782144e-05}
{"train_loss": 0.10175062716007233, "global_step": 132514, "epoch": 1488, "lr": 5.2005350632775426e-05}
{"train_loss": 0.10294326394796371, "global_step": 132515, "epoch": 1488, "lr": 5.2004771362499115e-05}
{"train_loss": 0.07643961906433105, "global_step": 132516, "epoch": 1488, "lr": 5.2004192091953284e-05}
{"train_loss": 0.13569392263889313, "global_step": 132517, "epoch": 1488, "lr": 5.2003612821138016e-05}
{"train_loss": 0.18243271112442017, "global_step": 132518, "epoch": 1488, "lr": 5.200303355005337e-05}
{"train_loss": 0.11944487690925598, "global_step": 132519, "epoch": 1488, "lr": 5.200245427869945e-05}
{"train_loss": 0.12649473455849658, "global_step": 132520, "epoch": 1488, "lr": 5.2001875007076336e-05, "val_loss": 5.747655391693115}
{"train_loss": 0.14994218945503235, "global_step": 132521, "epoch": 1489, "lr": 5.200129573518409e-05}
{"train_loss": 0.07063164561986923, "global_step": 132522, "epoch": 1489, "lr": 5.200071646302279e-05}
{"train_loss": 0.10200973600149155, "global_step": 132523, "epoch": 1489, "lr": 5.200013719059252e-05}
{"train_loss": 0.15747696161270142, "global_step": 132524, "epoch": 1489, "lr": 5.199955791789336e-05}
{"train_loss": 0.09997335076332092, "global_step": 132525, "epoch": 1489, "lr": 5.199897864492538e-05}
{"train_loss": 0.12992791831493378, "global_step": 132526, "epoch": 1489, "lr": 5.1998399371688666e-05}
{"train_loss": 0.06714185327291489, "global_step": 132527, "epoch": 1489, "lr": 5.199782009818328e-05}
{"train_loss": 0.15417471528053284, "global_step": 132528, "epoch": 1489, "lr": 5.199724082440932e-05}
{"train_loss": 0.1206774115562439, "global_step": 132529, "epoch": 1489, "lr": 5.1996661550366864e-05}
{"train_loss": 0.10189152508974075, "global_step": 132530, "epoch": 1489, "lr": 5.199608227605597e-05}
{"train_loss": 0.10451275110244751, "global_step": 132531, "epoch": 1489, "lr": 5.199550300147673e-05}
{"train_loss": 0.15096326172351837, "global_step": 132532, "epoch": 1489, "lr": 5.199492372662922e-05}
{"train_loss": 0.1581834852695465, "global_step": 132533, "epoch": 1489, "lr": 5.199434445151352e-05}
{"train_loss": 0.08711390197277069, "global_step": 132534, "epoch": 1489, "lr": 5.199376517612971e-05}
{"train_loss": 0.19260695576667786, "global_step": 132535, "epoch": 1489, "lr": 5.1993185900477834e-05}
{"train_loss": 0.16396069526672363, "global_step": 132536, "epoch": 1489, "lr": 5.199260662455801e-05}
{"train_loss": 0.1423395276069641, "global_step": 132537, "epoch": 1489, "lr": 5.1992027348370306e-05}
{"train_loss": 0.07990670949220657, "global_step": 132538, "epoch": 1489, "lr": 5.199144807191481e-05}
{"train_loss": 0.09670421481132507, "global_step": 132539, "epoch": 1489, "lr": 5.1990868795191586e-05}
{"train_loss": 0.10768862813711166, "global_step": 132540, "epoch": 1489, "lr": 5.199028951820071e-05}
{"train_loss": 0.08513340353965759, "global_step": 132541, "epoch": 1489, "lr": 5.198971024094226e-05}
{"train_loss": 0.13028870522975922, "global_step": 132542, "epoch": 1489, "lr": 5.198913096341632e-05}
{"train_loss": 0.08415002375841141, "global_step": 132543, "epoch": 1489, "lr": 5.1988551685622964e-05}
{"train_loss": 0.15712276101112366, "global_step": 132544, "epoch": 1489, "lr": 5.198797240756228e-05}
{"train_loss": 0.11231760680675507, "global_step": 132545, "epoch": 1489, "lr": 5.198739312923433e-05}
{"train_loss": 0.11006604880094528, "global_step": 132546, "epoch": 1489, "lr": 5.198681385063919e-05}
{"train_loss": 0.10059191286563873, "global_step": 132547, "epoch": 1489, "lr": 5.198623457177696e-05}
{"train_loss": 0.1552492380142212, "global_step": 132548, "epoch": 1489, "lr": 5.1985655292647704e-05}
{"train_loss": 0.12217897176742554, "global_step": 132549, "epoch": 1489, "lr": 5.19850760132515e-05}
{"train_loss": 0.100697360932827, "global_step": 132550, "epoch": 1489, "lr": 5.198449673358843e-05}
{"train_loss": 0.1551181823015213, "global_step": 132551, "epoch": 1489, "lr": 5.1983917453658556e-05}
{"train_loss": 0.10240098834037781, "global_step": 132552, "epoch": 1489, "lr": 5.198333817346199e-05}
{"train_loss": 0.13489320874214172, "global_step": 132553, "epoch": 1489, "lr": 5.1982758892998774e-05}
{"train_loss": 0.08528807759284973, "global_step": 132554, "epoch": 1489, "lr": 5.198217961226901e-05}
{"train_loss": 0.0742480531334877, "global_step": 132555, "epoch": 1489, "lr": 5.198160033127274e-05}
{"train_loss": 0.10695574432611465, "global_step": 132556, "epoch": 1489, "lr": 5.198102105001009e-05}
{"train_loss": 0.14005506038665771, "global_step": 132557, "epoch": 1489, "lr": 5.198044176848111e-05}
{"train_loss": 0.08050550520420074, "global_step": 132558, "epoch": 1489, "lr": 5.197986248668588e-05}
{"train_loss": 0.15792590379714966, "global_step": 132559, "epoch": 1489, "lr": 5.19792832046245e-05}
{"train_loss": 0.08154278993606567, "global_step": 132560, "epoch": 1489, "lr": 5.1978703922297014e-05}
{"train_loss": 0.1815035343170166, "global_step": 132561, "epoch": 1489, "lr": 5.197812463970351e-05}
{"train_loss": 0.09602313488721848, "global_step": 132562, "epoch": 1489, "lr": 5.197754535684408e-05}
{"train_loss": 0.19520629942417145, "global_step": 132563, "epoch": 1489, "lr": 5.197696607371879e-05}
{"train_loss": 0.13233298063278198, "global_step": 132564, "epoch": 1489, "lr": 5.1976386790327725e-05}
{"train_loss": 0.1338752806186676, "global_step": 132565, "epoch": 1489, "lr": 5.197580750667096e-05}
{"train_loss": 0.19354645907878876, "global_step": 132566, "epoch": 1489, "lr": 5.1975228222748564e-05}
{"train_loss": 0.1118466705083847, "global_step": 132567, "epoch": 1489, "lr": 5.197464893856063e-05}
{"train_loss": 0.14195428788661957, "global_step": 132568, "epoch": 1489, "lr": 5.1974069654107224e-05}
{"train_loss": 0.13577604293823242, "global_step": 132569, "epoch": 1489, "lr": 5.197349036938843e-05}
{"train_loss": 0.08994489908218384, "global_step": 132570, "epoch": 1489, "lr": 5.197291108440433e-05}
{"train_loss": 0.10224942117929459, "global_step": 132571, "epoch": 1489, "lr": 5.1972331799154995e-05}
{"train_loss": 0.18032006919384003, "global_step": 132572, "epoch": 1489, "lr": 5.1971752513640506e-05}
{"train_loss": 0.1380368173122406, "global_step": 132573, "epoch": 1489, "lr": 5.197117322786092e-05}
{"train_loss": 0.11175184696912766, "global_step": 132574, "epoch": 1489, "lr": 5.197059394181635e-05}
{"train_loss": 0.10183413326740265, "global_step": 132575, "epoch": 1489, "lr": 5.197001465550685e-05}
{"train_loss": 0.1602356731891632, "global_step": 132576, "epoch": 1489, "lr": 5.1969435368932505e-05}
{"train_loss": 0.15626659989356995, "global_step": 132577, "epoch": 1489, "lr": 5.19688560820934e-05}
{"train_loss": 0.04687520116567612, "global_step": 132578, "epoch": 1489, "lr": 5.19682767949896e-05}
{"train_loss": 0.11527270823717117, "global_step": 132579, "epoch": 1489, "lr": 5.1967697507621206e-05}
{"train_loss": 0.12115078419446945, "global_step": 132580, "epoch": 1489, "lr": 5.196711821998826e-05}
{"train_loss": 0.08484024554491043, "global_step": 132581, "epoch": 1489, "lr": 5.1966538932090866e-05}
{"train_loss": 0.15752829611301422, "global_step": 132582, "epoch": 1489, "lr": 5.1965959643929095e-05}
{"train_loss": 0.11743807047605515, "global_step": 132583, "epoch": 1489, "lr": 5.196538035550302e-05}
{"train_loss": 0.1368498057126999, "global_step": 132584, "epoch": 1489, "lr": 5.196480106681273e-05}
{"train_loss": 0.08967939019203186, "global_step": 132585, "epoch": 1489, "lr": 5.196422177785829e-05}
{"train_loss": 0.11821606755256653, "global_step": 132586, "epoch": 1489, "lr": 5.1963642488639795e-05}
{"train_loss": 0.1877196729183197, "global_step": 132587, "epoch": 1489, "lr": 5.19630631991573e-05}
{"train_loss": 0.09289072453975677, "global_step": 132588, "epoch": 1489, "lr": 5.196248390941091e-05}
{"train_loss": 0.14903472363948822, "global_step": 132589, "epoch": 1489, "lr": 5.196190461940067e-05}
{"train_loss": 0.10984353721141815, "global_step": 132590, "epoch": 1489, "lr": 5.196132532912669e-05}
{"train_loss": 0.10426448285579681, "global_step": 132591, "epoch": 1489, "lr": 5.196074603858902e-05}
{"train_loss": 0.10432540625333786, "global_step": 132592, "epoch": 1489, "lr": 5.196016674778776e-05}
{"train_loss": 0.15764650702476501, "global_step": 132593, "epoch": 1489, "lr": 5.195958745672298e-05}
{"train_loss": 0.1703418642282486, "global_step": 132594, "epoch": 1489, "lr": 5.195900816539476e-05}
{"train_loss": 0.09518623352050781, "global_step": 132595, "epoch": 1489, "lr": 5.1958428873803177e-05}
{"train_loss": 0.09900198876857758, "global_step": 132596, "epoch": 1489, "lr": 5.1957849581948294e-05}
{"train_loss": 0.14352449774742126, "global_step": 132597, "epoch": 1489, "lr": 5.1957270289830215e-05}
{"train_loss": 0.1411459594964981, "global_step": 132598, "epoch": 1489, "lr": 5.1956690997449e-05}
{"train_loss": 0.1652509868144989, "global_step": 132599, "epoch": 1489, "lr": 5.195611170480473e-05}
{"train_loss": 0.21022742986679077, "global_step": 132600, "epoch": 1489, "lr": 5.195553241189749e-05}
{"train_loss": 0.10709239542484283, "global_step": 132601, "epoch": 1489, "lr": 5.1954953118727346e-05}
{"train_loss": 0.11286018788814545, "global_step": 132602, "epoch": 1489, "lr": 5.195437382529438e-05}
{"train_loss": 0.07499229907989502, "global_step": 132603, "epoch": 1489, "lr": 5.195379453159868e-05}
{"train_loss": 0.13421201705932617, "global_step": 132604, "epoch": 1489, "lr": 5.1953215237640315e-05}
{"train_loss": 0.1562589704990387, "global_step": 132605, "epoch": 1489, "lr": 5.195263594341936e-05}
{"train_loss": 0.11756082624197006, "global_step": 132606, "epoch": 1489, "lr": 5.19520566489359e-05}
{"train_loss": 0.17061308026313782, "global_step": 132607, "epoch": 1489, "lr": 5.1951477354190014e-05}
{"train_loss": 0.20857053995132446, "global_step": 132608, "epoch": 1489, "lr": 5.1950898059181764e-05}
{"train_loss": 0.1252478366989768, "global_step": 132609, "epoch": 1489, "lr": 5.195031876391125e-05, "val_loss": 5.787009239196777}
{"train_loss": 0.10313481837511063, "global_step": 132610, "epoch": 1490, "lr": 5.194973946837853e-05}
{"train_loss": 0.196864515542984, "global_step": 132611, "epoch": 1490, "lr": 5.19491601725837e-05}
{"train_loss": 0.14230743050575256, "global_step": 132612, "epoch": 1490, "lr": 5.194858087652683e-05}
{"train_loss": 0.1074058786034584, "global_step": 132613, "epoch": 1490, "lr": 5.1948001580207995e-05}
{"train_loss": 0.16442614793777466, "global_step": 132614, "epoch": 1490, "lr": 5.1947422283627265e-05}
{"train_loss": 0.12533456087112427, "global_step": 132615, "epoch": 1490, "lr": 5.194684298678474e-05}
{"train_loss": 0.1172589510679245, "global_step": 132616, "epoch": 1490, "lr": 5.1946263689680483e-05}
{"train_loss": 0.09975641220808029, "global_step": 132617, "epoch": 1490, "lr": 5.1945684392314574e-05}
{"train_loss": 0.110173299908638, "global_step": 132618, "epoch": 1490, "lr": 5.194510509468709e-05}
{"train_loss": 0.10424087196588516, "global_step": 132619, "epoch": 1490, "lr": 5.1944525796798105e-05}
{"train_loss": 0.1164388507604599, "global_step": 132620, "epoch": 1490, "lr": 5.194394649864771e-05}
{"train_loss": 0.13965755701065063, "global_step": 132621, "epoch": 1490, "lr": 5.194336720023597e-05}
{"train_loss": 0.09780692309141159, "global_step": 132622, "epoch": 1490, "lr": 5.1942787901562975e-05}
{"train_loss": 0.14735271036624908, "global_step": 132623, "epoch": 1490, "lr": 5.1942208602628794e-05}
{"train_loss": 0.1330816149711609, "global_step": 132624, "epoch": 1490, "lr": 5.1941629303433504e-05}
{"train_loss": 0.14097851514816284, "global_step": 132625, "epoch": 1490, "lr": 5.1941050003977185e-05}
{"train_loss": 0.14957907795906067, "global_step": 132626, "epoch": 1490, "lr": 5.194047070425991e-05}
{"train_loss": 0.1152956634759903, "global_step": 132627, "epoch": 1490, "lr": 5.193989140428178e-05}
{"train_loss": 0.06053783744573593, "global_step": 132628, "epoch": 1490, "lr": 5.1939312104042845e-05}
{"train_loss": 0.10451900959014893, "global_step": 132629, "epoch": 1490, "lr": 5.193873280354319e-05}
{"train_loss": 0.10607913881540298, "global_step": 132630, "epoch": 1490, "lr": 5.1938153502782896e-05}
{"train_loss": 0.10443928092718124, "global_step": 132631, "epoch": 1490, "lr": 5.193757420176205e-05}
{"train_loss": 0.06818728893995285, "global_step": 132632, "epoch": 1490, "lr": 5.1936994900480704e-05}
{"train_loss": 0.12560105323791504, "global_step": 132633, "epoch": 1490, "lr": 5.193641559893897e-05}
{"train_loss": 0.09910467267036438, "global_step": 132634, "epoch": 1490, "lr": 5.19358362971369e-05}
{"train_loss": 0.11093758046627045, "global_step": 132635, "epoch": 1490, "lr": 5.193525699507459e-05}
{"train_loss": 0.060146935284137726, "global_step": 132636, "epoch": 1490, "lr": 5.1934677692752096e-05}
{"train_loss": 0.13398940861225128, "global_step": 132637, "epoch": 1490, "lr": 5.193409839016951e-05}
{"train_loss": 0.2035897821187973, "global_step": 132638, "epoch": 1490, "lr": 5.193351908732691e-05}
{"train_loss": 0.1601807028055191, "global_step": 132639, "epoch": 1490, "lr": 5.1932939784224376e-05}
{"train_loss": 0.1145854964852333, "global_step": 132640, "epoch": 1490, "lr": 5.1932360480861975e-05}
{"train_loss": 0.11359109729528427, "global_step": 132641, "epoch": 1490, "lr": 5.193178117723979e-05}
{"train_loss": 0.09639854729175568, "global_step": 132642, "epoch": 1490, "lr": 5.19312018733579e-05}
{"train_loss": 0.13910014927387238, "global_step": 132643, "epoch": 1490, "lr": 5.19306225692164e-05}
{"train_loss": 0.20818763971328735, "global_step": 132644, "epoch": 1490, "lr": 5.193004326481534e-05}
{"train_loss": 0.10065961629152298, "global_step": 132645, "epoch": 1490, "lr": 5.192946396015481e-05}
{"train_loss": 0.1272478550672531, "global_step": 132646, "epoch": 1490, "lr": 5.1928884655234886e-05}
{"train_loss": 0.15630373358726501, "global_step": 132647, "epoch": 1490, "lr": 5.192830535005564e-05}
{"train_loss": 0.14384254813194275, "global_step": 132648, "epoch": 1490, "lr": 5.192772604461716e-05}
{"train_loss": 0.09130069613456726, "global_step": 132649, "epoch": 1490, "lr": 5.192714673891953e-05}
{"train_loss": 0.11348630487918854, "global_step": 132650, "epoch": 1490, "lr": 5.192656743296282e-05}
{"train_loss": 0.22722935676574707, "global_step": 132651, "epoch": 1490, "lr": 5.1925988126747094e-05}
{"train_loss": 0.1591241955757141, "global_step": 132652, "epoch": 1490, "lr": 5.192540882027245e-05}
{"train_loss": 0.06748395413160324, "global_step": 132653, "epoch": 1490, "lr": 5.1924829513538966e-05}
{"train_loss": 0.10168755799531937, "global_step": 132654, "epoch": 1490, "lr": 5.19242502065467e-05}
{"train_loss": 0.13200035691261292, "global_step": 132655, "epoch": 1490, "lr": 5.192367089929574e-05}
{"train_loss": 0.06693563610315323, "global_step": 132656, "epoch": 1490, "lr": 5.192309159178617e-05}
{"train_loss": 0.11987794935703278, "global_step": 132657, "epoch": 1490, "lr": 5.1922512284018064e-05}
{"train_loss": 0.10003155469894409, "global_step": 132658, "epoch": 1490, "lr": 5.192193297599151e-05}
{"train_loss": 0.09394758939743042, "global_step": 132659, "epoch": 1490, "lr": 5.1921353667706564e-05}
{"train_loss": 0.1505090594291687, "global_step": 132660, "epoch": 1490, "lr": 5.1920774359163316e-05}
{"train_loss": 0.1421186327934265, "global_step": 132661, "epoch": 1490, "lr": 5.192019505036184e-05}
{"train_loss": 0.10011551529169083, "global_step": 132662, "epoch": 1490, "lr": 5.191961574130222e-05}
{"train_loss": 0.14463452994823456, "global_step": 132663, "epoch": 1490, "lr": 5.1919036431984534e-05}
{"train_loss": 0.07945634424686432, "global_step": 132664, "epoch": 1490, "lr": 5.191845712240886e-05}
{"train_loss": 0.10081226378679276, "global_step": 132665, "epoch": 1490, "lr": 5.1917877812575264e-05}
{"train_loss": 0.07633800059556961, "global_step": 132666, "epoch": 1490, "lr": 5.191729850248384e-05}
{"train_loss": 0.09052610397338867, "global_step": 132667, "epoch": 1490, "lr": 5.1916719192134654e-05}
{"train_loss": 0.12492509931325912, "global_step": 132668, "epoch": 1490, "lr": 5.191613988152779e-05}
{"train_loss": 0.0872829332947731, "global_step": 132669, "epoch": 1490, "lr": 5.191556057066333e-05}
{"train_loss": 0.11502408981323242, "global_step": 132670, "epoch": 1490, "lr": 5.191498125954133e-05}
{"train_loss": 0.163680300116539, "global_step": 132671, "epoch": 1490, "lr": 5.191440194816189e-05}
{"train_loss": 0.14301274716854095, "global_step": 132672, "epoch": 1490, "lr": 5.191382263652509e-05}
{"train_loss": 0.2230062484741211, "global_step": 132673, "epoch": 1490, "lr": 5.1913243324631e-05}
{"train_loss": 0.14738702774047852, "global_step": 132674, "epoch": 1490, "lr": 5.191266401247969e-05}
{"train_loss": 0.1126837357878685, "global_step": 132675, "epoch": 1490, "lr": 5.191208470007125e-05}
{"train_loss": 0.05793316289782524, "global_step": 132676, "epoch": 1490, "lr": 5.191150538740576e-05}
{"train_loss": 0.15136834979057312, "global_step": 132677, "epoch": 1490, "lr": 5.191092607448328e-05}
{"train_loss": 0.13931474089622498, "global_step": 132678, "epoch": 1490, "lr": 5.1910346761303905e-05}
{"train_loss": 0.0977935716509819, "global_step": 132679, "epoch": 1490, "lr": 5.190976744786771e-05}
{"train_loss": 0.16185951232910156, "global_step": 132680, "epoch": 1490, "lr": 5.190918813417477e-05}
{"train_loss": 0.10239573568105698, "global_step": 132681, "epoch": 1490, "lr": 5.1908608820225156e-05}
{"train_loss": 0.13203386962413788, "global_step": 132682, "epoch": 1490, "lr": 5.190802950601895e-05}
{"train_loss": 0.14313897490501404, "global_step": 132683, "epoch": 1490, "lr": 5.190745019155625e-05}
{"train_loss": 0.12074784934520721, "global_step": 132684, "epoch": 1490, "lr": 5.1906870876837096e-05}
{"train_loss": 0.1687784045934677, "global_step": 132685, "epoch": 1490, "lr": 5.1906291561861596e-05}
{"train_loss": 0.1759137064218521, "global_step": 132686, "epoch": 1490, "lr": 5.1905712246629824e-05}
{"train_loss": 0.10399474203586578, "global_step": 132687, "epoch": 1490, "lr": 5.190513293114184e-05}
{"train_loss": 0.18505756556987762, "global_step": 132688, "epoch": 1490, "lr": 5.190455361539774e-05}
{"train_loss": 0.0974106639623642, "global_step": 132689, "epoch": 1490, "lr": 5.1903974299397604e-05}
{"train_loss": 0.0636044442653656, "global_step": 132690, "epoch": 1490, "lr": 5.19033949831415e-05}
{"train_loss": 0.20153333246707916, "global_step": 132691, "epoch": 1490, "lr": 5.1902815666629486e-05}
{"train_loss": 0.09347276389598846, "global_step": 132692, "epoch": 1490, "lr": 5.190223634986169e-05}
{"train_loss": 0.09746838361024857, "global_step": 132693, "epoch": 1490, "lr": 5.1901657032838147e-05}
{"train_loss": 0.1100538894534111, "global_step": 132694, "epoch": 1490, "lr": 5.1901077715558954e-05}
{"train_loss": 0.1292024701833725, "global_step": 132695, "epoch": 1490, "lr": 5.190049839802418e-05}
{"train_loss": 0.11734869331121445, "global_step": 132696, "epoch": 1490, "lr": 5.189991908023391e-05}
{"train_loss": 0.17364755272865295, "global_step": 132697, "epoch": 1490, "lr": 5.189933976218822e-05}
{"train_loss": 0.1238446957693341, "global_step": 132698, "epoch": 1490, "lr": 5.1898760443887185e-05, "val_loss": 5.57467794418335, "train_action_mse_error": 25.498607635498047}
{"train_loss": 0.12814950942993164, "global_step": 132699, "epoch": 1491, "lr": 5.189818112533088e-05}
{"train_loss": 0.11515414714813232, "global_step": 132700, "epoch": 1491, "lr": 5.189760180651938e-05}
{"train_loss": 0.07801894843578339, "global_step": 132701, "epoch": 1491, "lr": 5.18970224874528e-05}
{"train_loss": 0.10485683381557465, "global_step": 132702, "epoch": 1491, "lr": 5.189644316813117e-05}
{"train_loss": 0.11430100351572037, "global_step": 132703, "epoch": 1491, "lr": 5.189586384855458e-05}
{"train_loss": 0.10217908024787903, "global_step": 132704, "epoch": 1491, "lr": 5.189528452872313e-05}
{"train_loss": 0.11540976911783218, "global_step": 132705, "epoch": 1491, "lr": 5.1894705208636876e-05}
{"train_loss": 0.12462008744478226, "global_step": 132706, "epoch": 1491, "lr": 5.1894125888295897e-05}
{"train_loss": 0.10787166655063629, "global_step": 132707, "epoch": 1491, "lr": 5.189354656770028e-05}
{"train_loss": 0.13622310757637024, "global_step": 132708, "epoch": 1491, "lr": 5.1892967246850096e-05}
{"train_loss": 0.08555003255605698, "global_step": 132709, "epoch": 1491, "lr": 5.1892387925745424e-05}
{"train_loss": 0.10470401495695114, "global_step": 132710, "epoch": 1491, "lr": 5.189180860438635e-05}
{"train_loss": 0.22095754742622375, "global_step": 132711, "epoch": 1491, "lr": 5.1891229282772945e-05}
{"train_loss": 0.0931326225399971, "global_step": 132712, "epoch": 1491, "lr": 5.189064996090528e-05}
{"train_loss": 0.14812572300434113, "global_step": 132713, "epoch": 1491, "lr": 5.189007063878345e-05}
{"train_loss": 0.1800256222486496, "global_step": 132714, "epoch": 1491, "lr": 5.188949131640751e-05}
{"train_loss": 0.13016566634178162, "global_step": 132715, "epoch": 1491, "lr": 5.1888911993777564e-05}
{"train_loss": 0.12007095664739609, "global_step": 132716, "epoch": 1491, "lr": 5.188833267089367e-05}
{"train_loss": 0.08983578532934189, "global_step": 132717, "epoch": 1491, "lr": 5.188775334775592e-05}
{"train_loss": 0.0760701596736908, "global_step": 132718, "epoch": 1491, "lr": 5.188717402436436e-05}
{"train_loss": 0.07467655837535858, "global_step": 132719, "epoch": 1491, "lr": 5.1886594700719114e-05}
{"train_loss": 0.13296441733837128, "global_step": 132720, "epoch": 1491, "lr": 5.1886015376820244e-05}
{"train_loss": 0.05781678855419159, "global_step": 132721, "epoch": 1491, "lr": 5.188543605266781e-05}
{"train_loss": 0.13780950009822845, "global_step": 132722, "epoch": 1491, "lr": 5.188485672826191e-05}
{"train_loss": 0.12737150490283966, "global_step": 132723, "epoch": 1491, "lr": 5.188427740360261e-05}
{"train_loss": 0.1146855354309082, "global_step": 132724, "epoch": 1491, "lr": 5.188369807868999e-05}
{"train_loss": 0.08182992041110992, "global_step": 132725, "epoch": 1491, "lr": 5.188311875352414e-05}
{"train_loss": 0.12200268357992172, "global_step": 132726, "epoch": 1491, "lr": 5.188253942810511e-05}
{"train_loss": 0.18085640668869019, "global_step": 132727, "epoch": 1491, "lr": 5.1881960102433015e-05}
{"train_loss": 0.07802826166152954, "global_step": 132728, "epoch": 1491, "lr": 5.188138077650789e-05}
{"train_loss": 0.08582080155611038, "global_step": 132729, "epoch": 1491, "lr": 5.188080145032985e-05}
{"train_loss": 0.152869313955307, "global_step": 132730, "epoch": 1491, "lr": 5.188022212389897e-05}
{"train_loss": 0.1366879642009735, "global_step": 132731, "epoch": 1491, "lr": 5.18796427972153e-05}
{"train_loss": 0.17499297857284546, "global_step": 132732, "epoch": 1491, "lr": 5.187906347027894e-05}
{"train_loss": 0.1213151142001152, "global_step": 132733, "epoch": 1491, "lr": 5.1878484143089965e-05}
{"train_loss": 0.09712665528059006, "global_step": 132734, "epoch": 1491, "lr": 5.187790481564845e-05}
{"train_loss": 0.09794056415557861, "global_step": 132735, "epoch": 1491, "lr": 5.187732548795448e-05}
{"train_loss": 0.129219189286232, "global_step": 132736, "epoch": 1491, "lr": 5.187674616000809e-05}
{"train_loss": 0.11191734671592712, "global_step": 132737, "epoch": 1491, "lr": 5.187616683180944e-05}
{"train_loss": 0.06857936829328537, "global_step": 132738, "epoch": 1491, "lr": 5.187558750335854e-05}
{"train_loss": 0.07133480161428452, "global_step": 132739, "epoch": 1491, "lr": 5.187500817465549e-05}
{"train_loss": 0.15121206641197205, "global_step": 132740, "epoch": 1491, "lr": 5.1874428845700384e-05}
{"train_loss": 0.12090844660997391, "global_step": 132741, "epoch": 1491, "lr": 5.187384951649328e-05}
{"train_loss": 0.1222931295633316, "global_step": 132742, "epoch": 1491, "lr": 5.187327018703425e-05}
{"train_loss": 0.13592106103897095, "global_step": 132743, "epoch": 1491, "lr": 5.187269085732339e-05}
{"train_loss": 0.1684928983449936, "global_step": 132744, "epoch": 1491, "lr": 5.187211152736076e-05}
{"train_loss": 0.1518935263156891, "global_step": 132745, "epoch": 1491, "lr": 5.187153219714646e-05}
{"train_loss": 0.08006437122821808, "global_step": 132746, "epoch": 1491, "lr": 5.187095286668054e-05}
{"train_loss": 0.06866201013326645, "global_step": 132747, "epoch": 1491, "lr": 5.18703735359631e-05}
{"train_loss": 0.15764406323432922, "global_step": 132748, "epoch": 1491, "lr": 5.186979420499422e-05}
{"train_loss": 0.1482822448015213, "global_step": 132749, "epoch": 1491, "lr": 5.186921487377395e-05}
{"train_loss": 0.11095323413610458, "global_step": 132750, "epoch": 1491, "lr": 5.1868635542302405e-05}
{"train_loss": 0.09688273072242737, "global_step": 132751, "epoch": 1491, "lr": 5.1868056210579626e-05}
{"train_loss": 0.14682485163211823, "global_step": 132752, "epoch": 1491, "lr": 5.1867476878605726e-05}
{"train_loss": 0.1474343240261078, "global_step": 132753, "epoch": 1491, "lr": 5.1866897546380766e-05}
{"train_loss": 0.09621185809373856, "global_step": 132754, "epoch": 1491, "lr": 5.1866318213904806e-05}
{"train_loss": 0.11838424205780029, "global_step": 132755, "epoch": 1491, "lr": 5.186573888117796e-05}
{"train_loss": 0.07088344544172287, "global_step": 132756, "epoch": 1491, "lr": 5.186515954820028e-05}
{"train_loss": 0.09263267368078232, "global_step": 132757, "epoch": 1491, "lr": 5.1864580214971856e-05}
{"train_loss": 0.17376986145973206, "global_step": 132758, "epoch": 1491, "lr": 5.186400088149276e-05}
{"train_loss": 0.14192114770412445, "global_step": 132759, "epoch": 1491, "lr": 5.1863421547763095e-05}
{"train_loss": 0.14703458547592163, "global_step": 132760, "epoch": 1491, "lr": 5.186284221378288e-05}
{"train_loss": 0.10056252032518387, "global_step": 132761, "epoch": 1491, "lr": 5.1862262879552245e-05}
{"train_loss": 0.11263255774974823, "global_step": 132762, "epoch": 1491, "lr": 5.186168354507125e-05}
{"train_loss": 0.18644081056118011, "global_step": 132763, "epoch": 1491, "lr": 5.186110421033997e-05}
{"train_loss": 0.0835980623960495, "global_step": 132764, "epoch": 1491, "lr": 5.1860524875358496e-05}
{"train_loss": 0.1621730625629425, "global_step": 132765, "epoch": 1491, "lr": 5.1859945540126894e-05}
{"train_loss": 0.1310805380344391, "global_step": 132766, "epoch": 1491, "lr": 5.185936620464524e-05}
{"train_loss": 0.11040721088647842, "global_step": 132767, "epoch": 1491, "lr": 5.1858786868913625e-05}
{"train_loss": 0.18446862697601318, "global_step": 132768, "epoch": 1491, "lr": 5.1858207532932114e-05}
{"train_loss": 0.12121136486530304, "global_step": 132769, "epoch": 1491, "lr": 5.185762819670079e-05}
{"train_loss": 0.14857782423496246, "global_step": 132770, "epoch": 1491, "lr": 5.1857048860219725e-05}
{"train_loss": 0.1047498807311058, "global_step": 132771, "epoch": 1491, "lr": 5.185646952348902e-05}
{"train_loss": 0.16513720154762268, "global_step": 132772, "epoch": 1491, "lr": 5.1855890186508705e-05}
{"train_loss": 0.1362890750169754, "global_step": 132773, "epoch": 1491, "lr": 5.1855310849278905e-05}
{"train_loss": 0.05054619535803795, "global_step": 132774, "epoch": 1491, "lr": 5.1854731511799684e-05}
{"train_loss": 0.12605233490467072, "global_step": 132775, "epoch": 1491, "lr": 5.1854152174071116e-05}
{"train_loss": 0.08427084982395172, "global_step": 132776, "epoch": 1491, "lr": 5.185357283609328e-05}
{"train_loss": 0.13689300417900085, "global_step": 132777, "epoch": 1491, "lr": 5.185299349786624e-05}
{"train_loss": 0.159040629863739, "global_step": 132778, "epoch": 1491, "lr": 5.18524141593901e-05}
{"train_loss": 0.14372053742408752, "global_step": 132779, "epoch": 1491, "lr": 5.1851834820664944e-05}
{"train_loss": 0.19446465373039246, "global_step": 132780, "epoch": 1491, "lr": 5.185125548169081e-05}
{"train_loss": 0.13299691677093506, "global_step": 132781, "epoch": 1491, "lr": 5.18506761424678e-05}
{"train_loss": 0.19917051494121552, "global_step": 132782, "epoch": 1491, "lr": 5.1850096802996e-05}
{"train_loss": 0.18921715021133423, "global_step": 132783, "epoch": 1491, "lr": 5.1849517463275453e-05}
{"train_loss": 0.1061677560210228, "global_step": 132784, "epoch": 1491, "lr": 5.184893812330628e-05}
{"train_loss": 0.0885760486125946, "global_step": 132785, "epoch": 1491, "lr": 5.1848358783088534e-05}
{"train_loss": 0.14015910029411316, "global_step": 132786, "epoch": 1491, "lr": 5.18477794426223e-05}
{"train_loss": 0.12336374011434866, "global_step": 132787, "epoch": 1491, "lr": 5.184720010190765e-05, "val_loss": 5.568780422210693}
{"train_loss": 0.07689125835895538, "global_step": 132788, "epoch": 1492, "lr": 5.184662076094467e-05}
{"train_loss": 0.0973518043756485, "global_step": 132789, "epoch": 1492, "lr": 5.184604141973344e-05}
{"train_loss": 0.14089469611644745, "global_step": 132790, "epoch": 1492, "lr": 5.184546207827401e-05}
{"train_loss": 0.12234150618314743, "global_step": 132791, "epoch": 1492, "lr": 5.18448827365665e-05}
{"train_loss": 0.1034458577632904, "global_step": 132792, "epoch": 1492, "lr": 5.184430339461096e-05}
{"train_loss": 0.143588587641716, "global_step": 132793, "epoch": 1492, "lr": 5.1843724052407486e-05}
{"train_loss": 0.12909071147441864, "global_step": 132794, "epoch": 1492, "lr": 5.184314470995614e-05}
{"train_loss": 0.08243363350629807, "global_step": 132795, "epoch": 1492, "lr": 5.1842565367257e-05}
{"train_loss": 0.1522814780473709, "global_step": 132796, "epoch": 1492, "lr": 5.184198602431016e-05}
{"train_loss": 0.16176746785640717, "global_step": 132797, "epoch": 1492, "lr": 5.184140668111568e-05}
{"train_loss": 0.19029666483402252, "global_step": 132798, "epoch": 1492, "lr": 5.1840827337673656e-05}
{"train_loss": 0.2034120261669159, "global_step": 132799, "epoch": 1492, "lr": 5.1840247993984145e-05}
{"train_loss": 0.09446302056312561, "global_step": 132800, "epoch": 1492, "lr": 5.1839668650047246e-05}
{"train_loss": 0.10136516392230988, "global_step": 132801, "epoch": 1492, "lr": 5.183908930586302e-05}
{"train_loss": 0.14247411489486694, "global_step": 132802, "epoch": 1492, "lr": 5.183850996143155e-05}
{"train_loss": 0.08179488033056259, "global_step": 132803, "epoch": 1492, "lr": 5.183793061675291e-05}
{"train_loss": 0.11970876157283783, "global_step": 132804, "epoch": 1492, "lr": 5.183735127182718e-05}
{"train_loss": 0.0837637186050415, "global_step": 132805, "epoch": 1492, "lr": 5.183677192665445e-05}
{"train_loss": 0.12006084620952606, "global_step": 132806, "epoch": 1492, "lr": 5.1836192581234776e-05}
{"train_loss": 0.11508282274007797, "global_step": 132807, "epoch": 1492, "lr": 5.1835613235568256e-05}
{"train_loss": 0.12909157574176788, "global_step": 132808, "epoch": 1492, "lr": 5.183503388965497e-05}
{"train_loss": 0.1418333500623703, "global_step": 132809, "epoch": 1492, "lr": 5.183445454349496e-05}
{"train_loss": 0.08958058059215546, "global_step": 132810, "epoch": 1492, "lr": 5.1833875197088355e-05}
{"train_loss": 0.08304629474878311, "global_step": 132811, "epoch": 1492, "lr": 5.183329585043519e-05}
{"train_loss": 0.14607535302639008, "global_step": 132812, "epoch": 1492, "lr": 5.1832716503535574e-05}
{"train_loss": 0.09362603724002838, "global_step": 132813, "epoch": 1492, "lr": 5.183213715638957e-05}
{"train_loss": 0.13002842664718628, "global_step": 132814, "epoch": 1492, "lr": 5.1831557808997254e-05}
{"train_loss": 0.11682382225990295, "global_step": 132815, "epoch": 1492, "lr": 5.183097846135871e-05}
{"train_loss": 0.07489991933107376, "global_step": 132816, "epoch": 1492, "lr": 5.183039911347402e-05}
{"train_loss": 0.09089162200689316, "global_step": 132817, "epoch": 1492, "lr": 5.1829819765343255e-05}
{"train_loss": 0.12304647266864777, "global_step": 132818, "epoch": 1492, "lr": 5.182924041696649e-05}
{"train_loss": 0.11142869293689728, "global_step": 132819, "epoch": 1492, "lr": 5.1828661068343796e-05}
{"train_loss": 0.10180184245109558, "global_step": 132820, "epoch": 1492, "lr": 5.182808171947526e-05}
{"train_loss": 0.11618369072675705, "global_step": 132821, "epoch": 1492, "lr": 5.182750237036097e-05}
{"train_loss": 0.11701168864965439, "global_step": 132822, "epoch": 1492, "lr": 5.1826923021000996e-05}
{"train_loss": 0.14720472693443298, "global_step": 132823, "epoch": 1492, "lr": 5.182634367139541e-05}
{"train_loss": 0.15735524892807007, "global_step": 132824, "epoch": 1492, "lr": 5.182576432154429e-05}
{"train_loss": 0.11851202696561813, "global_step": 132825, "epoch": 1492, "lr": 5.1825184971447724e-05}
{"train_loss": 0.10375520586967468, "global_step": 132826, "epoch": 1492, "lr": 5.182460562110578e-05}
{"train_loss": 0.16107259690761566, "global_step": 132827, "epoch": 1492, "lr": 5.1824026270518534e-05}
{"train_loss": 0.13600775599479675, "global_step": 132828, "epoch": 1492, "lr": 5.182344691968608e-05}
{"train_loss": 0.0667332336306572, "global_step": 132829, "epoch": 1492, "lr": 5.182286756860848e-05}
{"train_loss": 0.11957497149705887, "global_step": 132830, "epoch": 1492, "lr": 5.182228821728583e-05}
{"train_loss": 0.11386808007955551, "global_step": 132831, "epoch": 1492, "lr": 5.1821708865718176e-05}
{"train_loss": 0.12052734196186066, "global_step": 132832, "epoch": 1492, "lr": 5.182112951390563e-05}
{"train_loss": 0.132314994931221, "global_step": 132833, "epoch": 1492, "lr": 5.182055016184826e-05}
{"train_loss": 0.0872919037938118, "global_step": 132834, "epoch": 1492, "lr": 5.181997080954613e-05}
{"train_loss": 0.18069720268249512, "global_step": 132835, "epoch": 1492, "lr": 5.181939145699932e-05}
{"train_loss": 0.07106473296880722, "global_step": 132836, "epoch": 1492, "lr": 5.181881210420793e-05}
{"train_loss": 0.08629030734300613, "global_step": 132837, "epoch": 1492, "lr": 5.1818232751172025e-05}
{"train_loss": 0.11439690738916397, "global_step": 132838, "epoch": 1492, "lr": 5.1817653397891675e-05}
{"train_loss": 0.14822810888290405, "global_step": 132839, "epoch": 1492, "lr": 5.181707404436697e-05}
{"train_loss": 0.06880596280097961, "global_step": 132840, "epoch": 1492, "lr": 5.181649469059796e-05}
{"train_loss": 0.09144972264766693, "global_step": 132841, "epoch": 1492, "lr": 5.1815915336584765e-05}
{"train_loss": 0.11160381138324738, "global_step": 132842, "epoch": 1492, "lr": 5.181533598232743e-05}
{"train_loss": 0.13116151094436646, "global_step": 132843, "epoch": 1492, "lr": 5.181475662782604e-05}
{"train_loss": 0.1640031933784485, "global_step": 132844, "epoch": 1492, "lr": 5.18141772730807e-05}
{"train_loss": 0.1378178596496582, "global_step": 132845, "epoch": 1492, "lr": 5.1813597918091436e-05}
{"train_loss": 0.12479665875434875, "global_step": 132846, "epoch": 1492, "lr": 5.181301856285837e-05}
{"train_loss": 0.11047035455703735, "global_step": 132847, "epoch": 1492, "lr": 5.181243920738158e-05}
{"train_loss": 0.10675840824842453, "global_step": 132848, "epoch": 1492, "lr": 5.1811859851661114e-05}
{"train_loss": 0.06700201332569122, "global_step": 132849, "epoch": 1492, "lr": 5.1811280495697066e-05}
{"train_loss": 0.16309738159179688, "global_step": 132850, "epoch": 1492, "lr": 5.1810701139489526e-05}
{"train_loss": 0.1464107632637024, "global_step": 132851, "epoch": 1492, "lr": 5.181012178303854e-05}
{"train_loss": 0.16445985436439514, "global_step": 132852, "epoch": 1492, "lr": 5.180954242634423e-05}
{"train_loss": 0.10857290774583817, "global_step": 132853, "epoch": 1492, "lr": 5.1808963069406636e-05}
{"train_loss": 0.12555934488773346, "global_step": 132854, "epoch": 1492, "lr": 5.180838371222584e-05}
{"train_loss": 0.11211156100034714, "global_step": 132855, "epoch": 1492, "lr": 5.1807804354801935e-05}
{"train_loss": 0.16700728237628937, "global_step": 132856, "epoch": 1492, "lr": 5.1807224997135005e-05}
{"train_loss": 0.1409655511379242, "global_step": 132857, "epoch": 1492, "lr": 5.180664563922511e-05}
{"train_loss": 0.12486672401428223, "global_step": 132858, "epoch": 1492, "lr": 5.180606628107233e-05}
{"train_loss": 0.12141401320695877, "global_step": 132859, "epoch": 1492, "lr": 5.180548692267675e-05}
{"train_loss": 0.07450079172849655, "global_step": 132860, "epoch": 1492, "lr": 5.180490756403843e-05}
{"train_loss": 0.08569951355457306, "global_step": 132861, "epoch": 1492, "lr": 5.180432820515748e-05}
{"train_loss": 0.14187857508659363, "global_step": 132862, "epoch": 1492, "lr": 5.180374884603395e-05}
{"train_loss": 0.1018858477473259, "global_step": 132863, "epoch": 1492, "lr": 5.180316948666791e-05}
{"train_loss": 0.09921970218420029, "global_step": 132864, "epoch": 1492, "lr": 5.180259012705948e-05}
{"train_loss": 0.18868708610534668, "global_step": 132865, "epoch": 1492, "lr": 5.180201076720871e-05}
{"train_loss": 0.12423668801784515, "global_step": 132866, "epoch": 1492, "lr": 5.1801431407115686e-05}
{"train_loss": 0.1104009747505188, "global_step": 132867, "epoch": 1492, "lr": 5.180085204678048e-05}
{"train_loss": 0.1377245932817459, "global_step": 132868, "epoch": 1492, "lr": 5.1800272686203166e-05}
{"train_loss": 0.10589314997196198, "global_step": 132869, "epoch": 1492, "lr": 5.179969332538382e-05}
{"train_loss": 0.17549444735050201, "global_step": 132870, "epoch": 1492, "lr": 5.179911396432254e-05}
{"train_loss": 0.17297610640525818, "global_step": 132871, "epoch": 1492, "lr": 5.1798534603019386e-05}
{"train_loss": 0.15902671217918396, "global_step": 132872, "epoch": 1492, "lr": 5.179795524147445e-05}
{"train_loss": 0.10525068640708923, "global_step": 132873, "epoch": 1492, "lr": 5.179737587968779e-05}
{"train_loss": 0.12886148691177368, "global_step": 132874, "epoch": 1492, "lr": 5.17967965176595e-05}
{"train_loss": 0.11565083265304565, "global_step": 132875, "epoch": 1492, "lr": 5.179621715538966e-05}
{"train_loss": 0.1213821651560537, "global_step": 132876, "epoch": 1492, "lr": 5.1795637792878327e-05, "val_loss": 5.618681907653809}
{"train_loss": 0.08821521699428558, "global_step": 132877, "epoch": 1493, "lr": 5.17950584301256e-05}
{"train_loss": 0.15404990315437317, "global_step": 132878, "epoch": 1493, "lr": 5.179447906713154e-05}
{"train_loss": 0.11384079605340958, "global_step": 132879, "epoch": 1493, "lr": 5.179389970389624e-05}
{"train_loss": 0.1425793617963791, "global_step": 132880, "epoch": 1493, "lr": 5.179332034041978e-05}
{"train_loss": 0.11682390421628952, "global_step": 132881, "epoch": 1493, "lr": 5.1792740976702205e-05}
{"train_loss": 0.12929420173168182, "global_step": 132882, "epoch": 1493, "lr": 5.1792161612743636e-05}
{"train_loss": 0.1447443813085556, "global_step": 132883, "epoch": 1493, "lr": 5.179158224854414e-05}
{"train_loss": 0.14743778109550476, "global_step": 132884, "epoch": 1493, "lr": 5.179100288410377e-05}
{"train_loss": 0.11159273982048035, "global_step": 132885, "epoch": 1493, "lr": 5.179042351942264e-05}
{"train_loss": 0.20119725167751312, "global_step": 132886, "epoch": 1493, "lr": 5.1789844154500785e-05}
{"train_loss": 0.19235636293888092, "global_step": 132887, "epoch": 1493, "lr": 5.178926478933833e-05}
{"train_loss": 0.10696972161531448, "global_step": 132888, "epoch": 1493, "lr": 5.178868542393532e-05}
{"train_loss": 0.15711964666843414, "global_step": 132889, "epoch": 1493, "lr": 5.1788106058291855e-05}
{"train_loss": 0.07036817073822021, "global_step": 132890, "epoch": 1493, "lr": 5.178752669240799e-05}
{"train_loss": 0.17924141883850098, "global_step": 132891, "epoch": 1493, "lr": 5.1786947326283806e-05}
{"train_loss": 0.22607921063899994, "global_step": 132892, "epoch": 1493, "lr": 5.17863679599194e-05}
{"train_loss": 0.185153067111969, "global_step": 132893, "epoch": 1493, "lr": 5.1785788593314835e-05}
{"train_loss": 0.07213949412107468, "global_step": 132894, "epoch": 1493, "lr": 5.17852092264702e-05}
{"train_loss": 0.11185918748378754, "global_step": 132895, "epoch": 1493, "lr": 5.178462985938557e-05}
{"train_loss": 0.255911260843277, "global_step": 132896, "epoch": 1493, "lr": 5.178405049206101e-05}
{"train_loss": 0.17404532432556152, "global_step": 132897, "epoch": 1493, "lr": 5.1783471124496595e-05}
{"train_loss": 0.07971570640802383, "global_step": 132898, "epoch": 1493, "lr": 5.178289175669243e-05}
{"train_loss": 0.15282070636749268, "global_step": 132899, "epoch": 1493, "lr": 5.178231238864856e-05}
{"train_loss": 0.11504510790109634, "global_step": 132900, "epoch": 1493, "lr": 5.178173302036509e-05}
{"train_loss": 0.14700406789779663, "global_step": 132901, "epoch": 1493, "lr": 5.178115365184208e-05}
{"train_loss": 0.12965483963489532, "global_step": 132902, "epoch": 1493, "lr": 5.1780574283079616e-05}
{"train_loss": 0.12879543006420135, "global_step": 132903, "epoch": 1493, "lr": 5.177999491407779e-05}
{"train_loss": 0.12379416823387146, "global_step": 132904, "epoch": 1493, "lr": 5.177941554483666e-05}
{"train_loss": 0.1545523703098297, "global_step": 132905, "epoch": 1493, "lr": 5.17788361753563e-05}
{"train_loss": 0.12148180603981018, "global_step": 132906, "epoch": 1493, "lr": 5.1778256805636805e-05}
{"train_loss": 0.1580539494752884, "global_step": 132907, "epoch": 1493, "lr": 5.177767743567825e-05}
{"train_loss": 0.072927325963974, "global_step": 132908, "epoch": 1493, "lr": 5.17770980654807e-05}
{"train_loss": 0.13388483226299286, "global_step": 132909, "epoch": 1493, "lr": 5.177651869504424e-05}
{"train_loss": 0.15963730216026306, "global_step": 132910, "epoch": 1493, "lr": 5.177593932436895e-05}
{"train_loss": 0.11777833849191666, "global_step": 132911, "epoch": 1493, "lr": 5.177535995345492e-05}
{"train_loss": 0.08827637135982513, "global_step": 132912, "epoch": 1493, "lr": 5.1774780582302195e-05}
{"train_loss": 0.22855649888515472, "global_step": 132913, "epoch": 1493, "lr": 5.177420121091088e-05}
{"train_loss": 0.07418408989906311, "global_step": 132914, "epoch": 1493, "lr": 5.1773621839281064e-05}
{"train_loss": 0.07382720708847046, "global_step": 132915, "epoch": 1493, "lr": 5.177304246741278e-05}
{"train_loss": 0.13449302315711975, "global_step": 132916, "epoch": 1493, "lr": 5.1772463095306146e-05}
{"train_loss": 0.1596083790063858, "global_step": 132917, "epoch": 1493, "lr": 5.17718837229612e-05}
{"train_loss": 0.09501496702432632, "global_step": 132918, "epoch": 1493, "lr": 5.1771304350378066e-05}
{"train_loss": 0.0967174768447876, "global_step": 132919, "epoch": 1493, "lr": 5.177072497755681e-05}
{"train_loss": 0.09615803509950638, "global_step": 132920, "epoch": 1493, "lr": 5.177014560449749e-05}
{"train_loss": 0.15953008830547333, "global_step": 132921, "epoch": 1493, "lr": 5.17695662312002e-05}
{"train_loss": 0.18234962224960327, "global_step": 132922, "epoch": 1493, "lr": 5.1768986857665e-05}
{"train_loss": 0.12430936843156815, "global_step": 132923, "epoch": 1493, "lr": 5.1768407483892e-05}
{"train_loss": 0.18153108656406403, "global_step": 132924, "epoch": 1493, "lr": 5.1767828109881253e-05}
{"train_loss": 0.13469016551971436, "global_step": 132925, "epoch": 1493, "lr": 5.176724873563285e-05}
{"train_loss": 0.14747682213783264, "global_step": 132926, "epoch": 1493, "lr": 5.1766669361146845e-05}
{"train_loss": 0.09027377516031265, "global_step": 132927, "epoch": 1493, "lr": 5.176608998642334e-05}
{"train_loss": 0.10341991484165192, "global_step": 132928, "epoch": 1493, "lr": 5.176551061146241e-05}
{"train_loss": 0.12043613940477371, "global_step": 132929, "epoch": 1493, "lr": 5.176493123626413e-05}
{"train_loss": 0.1370953768491745, "global_step": 132930, "epoch": 1493, "lr": 5.176435186082856e-05}
{"train_loss": 0.16205251216888428, "global_step": 132931, "epoch": 1493, "lr": 5.176377248515582e-05}
{"train_loss": 0.09300244599580765, "global_step": 132932, "epoch": 1493, "lr": 5.1763193109245946e-05}
{"train_loss": 0.1576828956604004, "global_step": 132933, "epoch": 1493, "lr": 5.176261373309903e-05}
{"train_loss": 0.17924544215202332, "global_step": 132934, "epoch": 1493, "lr": 5.176203435671517e-05}
{"train_loss": 0.16020481288433075, "global_step": 132935, "epoch": 1493, "lr": 5.176145498009441e-05}
{"train_loss": 0.12930665910243988, "global_step": 132936, "epoch": 1493, "lr": 5.176087560323685e-05}
{"train_loss": 0.05588062107563019, "global_step": 132937, "epoch": 1493, "lr": 5.176029622614256e-05}
{"train_loss": 0.15772829949855804, "global_step": 132938, "epoch": 1493, "lr": 5.175971684881161e-05}
{"train_loss": 0.10218449681997299, "global_step": 132939, "epoch": 1493, "lr": 5.17591374712441e-05}
{"train_loss": 0.15231837332248688, "global_step": 132940, "epoch": 1493, "lr": 5.175855809344009e-05}
{"train_loss": 0.1465674489736557, "global_step": 132941, "epoch": 1493, "lr": 5.175797871539967e-05}
{"train_loss": 0.11579352617263794, "global_step": 132942, "epoch": 1493, "lr": 5.1757399337122905e-05}
{"train_loss": 0.07062989473342896, "global_step": 132943, "epoch": 1493, "lr": 5.1756819958609884e-05}
{"train_loss": 0.11856649070978165, "global_step": 132944, "epoch": 1493, "lr": 5.175624057986067e-05}
{"train_loss": 0.153058260679245, "global_step": 132945, "epoch": 1493, "lr": 5.175566120087536e-05}
{"train_loss": 0.07101885974407196, "global_step": 132946, "epoch": 1493, "lr": 5.1755081821654014e-05}
{"train_loss": 0.11890954524278641, "global_step": 132947, "epoch": 1493, "lr": 5.175450244219673e-05}
{"train_loss": 0.10155990719795227, "global_step": 132948, "epoch": 1493, "lr": 5.175392306250356e-05}
{"train_loss": 0.12308739125728607, "global_step": 132949, "epoch": 1493, "lr": 5.1753343682574605e-05}
{"train_loss": 0.0655718520283699, "global_step": 132950, "epoch": 1493, "lr": 5.175276430240994e-05}
{"train_loss": 0.11163754016160965, "global_step": 132951, "epoch": 1493, "lr": 5.1752184922009626e-05}
{"train_loss": 0.10386615246534348, "global_step": 132952, "epoch": 1493, "lr": 5.175160554137376e-05}
{"train_loss": 0.08097733557224274, "global_step": 132953, "epoch": 1493, "lr": 5.175102616050243e-05}
{"train_loss": 0.1314135640859604, "global_step": 132954, "epoch": 1493, "lr": 5.1750446779395656e-05}
{"train_loss": 0.09825730323791504, "global_step": 132955, "epoch": 1493, "lr": 5.174986739805357e-05}
{"train_loss": 0.10350368171930313, "global_step": 132956, "epoch": 1493, "lr": 5.174928801647625e-05}
{"train_loss": 0.08958461880683899, "global_step": 132957, "epoch": 1493, "lr": 5.174870863466376e-05}
{"train_loss": 0.16352258622646332, "global_step": 132958, "epoch": 1493, "lr": 5.1748129252616165e-05}
{"train_loss": 0.06617952138185501, "global_step": 132959, "epoch": 1493, "lr": 5.1747549870333556e-05}
{"train_loss": 0.13483868539333344, "global_step": 132960, "epoch": 1493, "lr": 5.1746970487816014e-05}
{"train_loss": 0.11403056234121323, "global_step": 132961, "epoch": 1493, "lr": 5.174639110506361e-05}
{"train_loss": 0.13639295101165771, "global_step": 132962, "epoch": 1493, "lr": 5.174581172207643e-05}
{"train_loss": 0.09707624465227127, "global_step": 132963, "epoch": 1493, "lr": 5.174523233885454e-05}
{"train_loss": 0.09449098259210587, "global_step": 132964, "epoch": 1493, "lr": 5.1744652955398034e-05}
{"train_loss": 0.12803593194216825, "global_step": 132965, "epoch": 1493, "lr": 5.174407357170697e-05, "val_loss": 5.455691814422607}
{"train_loss": 0.09719427675008774, "global_step": 132966, "epoch": 1494, "lr": 5.174349418778144e-05}
{"train_loss": 0.09619765728712082, "global_step": 132967, "epoch": 1494, "lr": 5.174291480362152e-05}
{"train_loss": 0.11555027961730957, "global_step": 132968, "epoch": 1494, "lr": 5.174233541922729e-05}
{"train_loss": 0.111019067466259, "global_step": 132969, "epoch": 1494, "lr": 5.174175603459882e-05}
{"train_loss": 0.13208018243312836, "global_step": 132970, "epoch": 1494, "lr": 5.174117664973619e-05}
{"train_loss": 0.207134410738945, "global_step": 132971, "epoch": 1494, "lr": 5.174059726463949e-05}
{"train_loss": 0.24567799270153046, "global_step": 132972, "epoch": 1494, "lr": 5.1740017879308774e-05}
{"train_loss": 0.13477838039398193, "global_step": 132973, "epoch": 1494, "lr": 5.173943849374414e-05}
{"train_loss": 0.092152900993824, "global_step": 132974, "epoch": 1494, "lr": 5.173885910794567e-05}
{"train_loss": 0.1051090657711029, "global_step": 132975, "epoch": 1494, "lr": 5.1738279721913416e-05}
{"train_loss": 0.0879216194152832, "global_step": 132976, "epoch": 1494, "lr": 5.1737700335647476e-05}
{"train_loss": 0.1920507550239563, "global_step": 132977, "epoch": 1494, "lr": 5.173712094914793e-05}
{"train_loss": 0.13576340675354004, "global_step": 132978, "epoch": 1494, "lr": 5.173654156241484e-05}
{"train_loss": 0.11254102736711502, "global_step": 132979, "epoch": 1494, "lr": 5.17359621754483e-05}
{"train_loss": 0.07890518754720688, "global_step": 132980, "epoch": 1494, "lr": 5.1735382788248376e-05}
{"train_loss": 0.082071453332901, "global_step": 132981, "epoch": 1494, "lr": 5.173480340081516e-05}
{"train_loss": 0.08875241875648499, "global_step": 132982, "epoch": 1494, "lr": 5.173422401314871e-05}
{"train_loss": 0.10798875987529755, "global_step": 132983, "epoch": 1494, "lr": 5.173364462524912e-05}
{"train_loss": 0.11434832960367203, "global_step": 132984, "epoch": 1494, "lr": 5.1733065237116464e-05}
{"train_loss": 0.11183580011129379, "global_step": 132985, "epoch": 1494, "lr": 5.173248584875081e-05}
{"train_loss": 0.09611225873231888, "global_step": 132986, "epoch": 1494, "lr": 5.1731906460152256e-05}
{"train_loss": 0.13866503536701202, "global_step": 132987, "epoch": 1494, "lr": 5.1731327071320856e-05}
{"train_loss": 0.12496032565832138, "global_step": 132988, "epoch": 1494, "lr": 5.1730747682256705e-05}
{"train_loss": 0.160296231508255, "global_step": 132989, "epoch": 1494, "lr": 5.173016829295988e-05}
{"train_loss": 0.14516201615333557, "global_step": 132990, "epoch": 1494, "lr": 5.172958890343046e-05}
{"train_loss": 0.07708339393138885, "global_step": 132991, "epoch": 1494, "lr": 5.1729009513668504e-05}
{"train_loss": 0.13989154994487762, "global_step": 132992, "epoch": 1494, "lr": 5.172843012367411e-05}
{"train_loss": 0.0949714332818985, "global_step": 132993, "epoch": 1494, "lr": 5.172785073344736e-05}
{"train_loss": 0.0816449448466301, "global_step": 132994, "epoch": 1494, "lr": 5.17272713429883e-05}
{"train_loss": 0.12913748621940613, "global_step": 132995, "epoch": 1494, "lr": 5.172669195229705e-05}
{"train_loss": 0.09283291548490524, "global_step": 132996, "epoch": 1494, "lr": 5.172611256137365e-05}
{"train_loss": 0.11252463608980179, "global_step": 132997, "epoch": 1494, "lr": 5.172553317021821e-05}
{"train_loss": 0.12407540529966354, "global_step": 132998, "epoch": 1494, "lr": 5.1724953778830784e-05}
{"train_loss": 0.11767459660768509, "global_step": 132999, "epoch": 1494, "lr": 5.172437438721146e-05}
{"train_loss": 0.1064436212182045, "global_step": 133000, "epoch": 1494, "lr": 5.172379499536032e-05}
{"train_loss": 0.11200185865163803, "global_step": 133001, "epoch": 1494, "lr": 5.172321560327743e-05}
{"train_loss": 0.09238634258508682, "global_step": 133002, "epoch": 1494, "lr": 5.172263621096288e-05}
{"train_loss": 0.13998225331306458, "global_step": 133003, "epoch": 1494, "lr": 5.172205681841674e-05}
{"train_loss": 0.19217734038829803, "global_step": 133004, "epoch": 1494, "lr": 5.17214774256391e-05}
{"train_loss": 0.1224948987364769, "global_step": 133005, "epoch": 1494, "lr": 5.1720898032630016e-05}
{"train_loss": 0.06106487289071083, "global_step": 133006, "epoch": 1494, "lr": 5.172031863938958e-05}
{"train_loss": 0.13368238508701324, "global_step": 133007, "epoch": 1494, "lr": 5.171973924591786e-05}
{"train_loss": 0.10038427263498306, "global_step": 133008, "epoch": 1494, "lr": 5.171915985221496e-05}
{"train_loss": 0.10570794343948364, "global_step": 133009, "epoch": 1494, "lr": 5.171858045828093e-05}
{"train_loss": 0.1085224598646164, "global_step": 133010, "epoch": 1494, "lr": 5.171800106411586e-05}
{"train_loss": 0.08202852308750153, "global_step": 133011, "epoch": 1494, "lr": 5.1717421669719825e-05}
{"train_loss": 0.07158894091844559, "global_step": 133012, "epoch": 1494, "lr": 5.1716842275092904e-05}
{"train_loss": 0.141092911362648, "global_step": 133013, "epoch": 1494, "lr": 5.1716262880235176e-05}
{"train_loss": 0.11698613315820694, "global_step": 133014, "epoch": 1494, "lr": 5.171568348514672e-05}
{"train_loss": 0.12018933147192001, "global_step": 133015, "epoch": 1494, "lr": 5.171510408982761e-05}
{"train_loss": 0.16565176844596863, "global_step": 133016, "epoch": 1494, "lr": 5.171452469427791e-05}
{"train_loss": 0.10835923254489899, "global_step": 133017, "epoch": 1494, "lr": 5.1713945298497734e-05}
{"train_loss": 0.1449781060218811, "global_step": 133018, "epoch": 1494, "lr": 5.171336590248712e-05}
{"train_loss": 0.1372767984867096, "global_step": 133019, "epoch": 1494, "lr": 5.171278650624618e-05}
{"train_loss": 0.18094755709171295, "global_step": 133020, "epoch": 1494, "lr": 5.171220710977497e-05}
{"train_loss": 0.19897565245628357, "global_step": 133021, "epoch": 1494, "lr": 5.171162771307357e-05}
{"train_loss": 0.0712282806634903, "global_step": 133022, "epoch": 1494, "lr": 5.1711048316142076e-05}
{"train_loss": 0.10853871703147888, "global_step": 133023, "epoch": 1494, "lr": 5.171046891898055e-05}
{"train_loss": 0.06704266369342804, "global_step": 133024, "epoch": 1494, "lr": 5.170988952158906e-05}
{"train_loss": 0.06696714460849762, "global_step": 133025, "epoch": 1494, "lr": 5.17093101239677e-05}
{"train_loss": 0.10355828702449799, "global_step": 133026, "epoch": 1494, "lr": 5.170873072611655e-05}
{"train_loss": 0.12559182941913605, "global_step": 133027, "epoch": 1494, "lr": 5.170815132803568e-05}
{"train_loss": 0.12708353996276855, "global_step": 133028, "epoch": 1494, "lr": 5.1707571929725176e-05}
{"train_loss": 0.10927418619394302, "global_step": 133029, "epoch": 1494, "lr": 5.1706992531185105e-05}
{"train_loss": 0.08427900820970535, "global_step": 133030, "epoch": 1494, "lr": 5.170641313241554e-05}
{"train_loss": 0.08967612683773041, "global_step": 133031, "epoch": 1494, "lr": 5.1705833733416575e-05}
{"train_loss": 0.13497427105903625, "global_step": 133032, "epoch": 1494, "lr": 5.170525433418829e-05}
{"train_loss": 0.19621211290359497, "global_step": 133033, "epoch": 1494, "lr": 5.170467493473074e-05}
{"train_loss": 0.1283043771982193, "global_step": 133034, "epoch": 1494, "lr": 5.1704095535044026e-05}
{"train_loss": 0.10853146761655807, "global_step": 133035, "epoch": 1494, "lr": 5.170351613512822e-05}
{"train_loss": 0.12077002972364426, "global_step": 133036, "epoch": 1494, "lr": 5.17029367349834e-05}
{"train_loss": 0.10745829343795776, "global_step": 133037, "epoch": 1494, "lr": 5.170235733460963e-05}
{"train_loss": 0.13770490884780884, "global_step": 133038, "epoch": 1494, "lr": 5.1701777934007e-05}
{"train_loss": 0.070691779255867, "global_step": 133039, "epoch": 1494, "lr": 5.170119853317559e-05}
{"train_loss": 0.0730644166469574, "global_step": 133040, "epoch": 1494, "lr": 5.1700619132115465e-05}
{"train_loss": 0.1370329111814499, "global_step": 133041, "epoch": 1494, "lr": 5.170003973082673e-05}
{"train_loss": 0.1178085207939148, "global_step": 133042, "epoch": 1494, "lr": 5.1699460329309435e-05}
{"train_loss": 0.056125689297914505, "global_step": 133043, "epoch": 1494, "lr": 5.1698880927563676e-05}
{"train_loss": 0.11754921823740005, "global_step": 133044, "epoch": 1494, "lr": 5.169830152558951e-05}
{"train_loss": 0.11913899332284927, "global_step": 133045, "epoch": 1494, "lr": 5.169772212338704e-05}
{"train_loss": 0.13604652881622314, "global_step": 133046, "epoch": 1494, "lr": 5.1697142720956327e-05}
{"train_loss": 0.09350769221782684, "global_step": 133047, "epoch": 1494, "lr": 5.169656331829745e-05}
{"train_loss": 0.1270299255847931, "global_step": 133048, "epoch": 1494, "lr": 5.1695983915410496e-05}
{"train_loss": 0.22053126990795135, "global_step": 133049, "epoch": 1494, "lr": 5.169540451229553e-05}
{"train_loss": 0.07747509330511093, "global_step": 133050, "epoch": 1494, "lr": 5.169482510895265e-05}
{"train_loss": 0.09720724076032639, "global_step": 133051, "epoch": 1494, "lr": 5.169424570538191e-05}
{"train_loss": 0.09666240215301514, "global_step": 133052, "epoch": 1494, "lr": 5.169366630158341e-05}
{"train_loss": 0.1326238214969635, "global_step": 133053, "epoch": 1494, "lr": 5.1693086897557205e-05}
{"train_loss": 0.1183200856942809, "global_step": 133054, "epoch": 1494, "lr": 5.16925074933034e-05, "val_loss": 5.63891077041626}
{"train_loss": 0.07761678844690323, "global_step": 133055, "epoch": 1495, "lr": 5.169192808882205e-05}
{"train_loss": 0.1274062544107437, "global_step": 133056, "epoch": 1495, "lr": 5.1691348684113237e-05}
{"train_loss": 0.2015850841999054, "global_step": 133057, "epoch": 1495, "lr": 5.169076927917704e-05}
{"train_loss": 0.13123773038387299, "global_step": 133058, "epoch": 1495, "lr": 5.169018987401355e-05}
{"train_loss": 0.10921169817447662, "global_step": 133059, "epoch": 1495, "lr": 5.168961046862284e-05}
{"train_loss": 0.14558309316635132, "global_step": 133060, "epoch": 1495, "lr": 5.1689031063004964e-05}
{"train_loss": 0.1639930009841919, "global_step": 133061, "epoch": 1495, "lr": 5.1688451657160033e-05}
{"train_loss": 0.14721277356147766, "global_step": 133062, "epoch": 1495, "lr": 5.16878722510881e-05}
{"train_loss": 0.11578746140003204, "global_step": 133063, "epoch": 1495, "lr": 5.1687292844789245e-05}
{"train_loss": 0.18963731825351715, "global_step": 133064, "epoch": 1495, "lr": 5.168671343826358e-05}
{"train_loss": 0.07038326561450958, "global_step": 133065, "epoch": 1495, "lr": 5.168613403151115e-05}
{"train_loss": 0.059631772339344025, "global_step": 133066, "epoch": 1495, "lr": 5.168555462453203e-05}
{"train_loss": 0.08604277670383453, "global_step": 133067, "epoch": 1495, "lr": 5.1684975217326315e-05}
{"train_loss": 0.09634525328874588, "global_step": 133068, "epoch": 1495, "lr": 5.168439580989407e-05}
{"train_loss": 0.10508909076452255, "global_step": 133069, "epoch": 1495, "lr": 5.1683816402235386e-05}
{"train_loss": 0.12324471771717072, "global_step": 133070, "epoch": 1495, "lr": 5.1683236994350325e-05}
{"train_loss": 0.17881916463375092, "global_step": 133071, "epoch": 1495, "lr": 5.168265758623897e-05}
{"train_loss": 0.09425649046897888, "global_step": 133072, "epoch": 1495, "lr": 5.168207817790142e-05}
{"train_loss": 0.22948098182678223, "global_step": 133073, "epoch": 1495, "lr": 5.168149876933772e-05}
{"train_loss": 0.09854515641927719, "global_step": 133074, "epoch": 1495, "lr": 5.1680919360547963e-05}
{"train_loss": 0.15793219208717346, "global_step": 133075, "epoch": 1495, "lr": 5.168033995153223e-05}
{"train_loss": 0.10252378135919571, "global_step": 133076, "epoch": 1495, "lr": 5.16797605422906e-05}
{"train_loss": 0.16555219888687134, "global_step": 133077, "epoch": 1495, "lr": 5.167918113282314e-05}
{"train_loss": 0.10537515580654144, "global_step": 133078, "epoch": 1495, "lr": 5.1678601723129936e-05}
{"train_loss": 0.19246506690979004, "global_step": 133079, "epoch": 1495, "lr": 5.1678022313211075e-05}
{"train_loss": 0.10703300684690475, "global_step": 133080, "epoch": 1495, "lr": 5.1677442903066605e-05}
{"train_loss": 0.12970532476902008, "global_step": 133081, "epoch": 1495, "lr": 5.167686349269663e-05}
{"train_loss": 0.10189124941825867, "global_step": 133082, "epoch": 1495, "lr": 5.1676284082101225e-05}
{"train_loss": 0.18751180171966553, "global_step": 133083, "epoch": 1495, "lr": 5.167570467128047e-05}
{"train_loss": 0.09558816254138947, "global_step": 133084, "epoch": 1495, "lr": 5.167512526023444e-05}
{"train_loss": 0.14055423438549042, "global_step": 133085, "epoch": 1495, "lr": 5.16745458489632e-05}
{"train_loss": 0.0787714421749115, "global_step": 133086, "epoch": 1495, "lr": 5.167396643746682e-05}
{"train_loss": 0.07556860893964767, "global_step": 133087, "epoch": 1495, "lr": 5.167338702574542e-05}
{"train_loss": 0.10918483883142471, "global_step": 133088, "epoch": 1495, "lr": 5.167280761379905e-05}
{"train_loss": 0.1003861278295517, "global_step": 133089, "epoch": 1495, "lr": 5.167222820162779e-05}
{"train_loss": 0.0934918001294136, "global_step": 133090, "epoch": 1495, "lr": 5.167164878923172e-05}
{"train_loss": 0.10050942748785019, "global_step": 133091, "epoch": 1495, "lr": 5.1671069376610915e-05}
{"train_loss": 0.10429307818412781, "global_step": 133092, "epoch": 1495, "lr": 5.1670489963765454e-05}
{"train_loss": 0.09490467607975006, "global_step": 133093, "epoch": 1495, "lr": 5.166991055069542e-05}
{"train_loss": 0.1338900774717331, "global_step": 133094, "epoch": 1495, "lr": 5.1669331137400875e-05}
{"train_loss": 0.21556459367275238, "global_step": 133095, "epoch": 1495, "lr": 5.166875172388193e-05}
{"train_loss": 0.10278505086898804, "global_step": 133096, "epoch": 1495, "lr": 5.166817231013862e-05}
{"train_loss": 0.21301744878292084, "global_step": 133097, "epoch": 1495, "lr": 5.1667592896171054e-05}
{"train_loss": 0.1050555482506752, "global_step": 133098, "epoch": 1495, "lr": 5.166701348197931e-05}
{"train_loss": 0.1610996127128601, "global_step": 133099, "epoch": 1495, "lr": 5.166643406756343e-05}
{"train_loss": 0.10480807721614838, "global_step": 133100, "epoch": 1495, "lr": 5.166585465292354e-05}
{"train_loss": 0.1418401300907135, "global_step": 133101, "epoch": 1495, "lr": 5.166527523805969e-05}
{"train_loss": 0.13856253027915955, "global_step": 133102, "epoch": 1495, "lr": 5.166469582297197e-05}
{"train_loss": 0.13425475358963013, "global_step": 133103, "epoch": 1495, "lr": 5.1664116407660446e-05}
{"train_loss": 0.07514834403991699, "global_step": 133104, "epoch": 1495, "lr": 5.166353699212521e-05}
{"train_loss": 0.12871798872947693, "global_step": 133105, "epoch": 1495, "lr": 5.166295757636632e-05}
{"train_loss": 0.08365320414304733, "global_step": 133106, "epoch": 1495, "lr": 5.1662378160383874e-05}
{"train_loss": 0.08410914242267609, "global_step": 133107, "epoch": 1495, "lr": 5.1661798744177935e-05}
{"train_loss": 0.12426478415727615, "global_step": 133108, "epoch": 1495, "lr": 5.166121932774859e-05}
{"train_loss": 0.14743973314762115, "global_step": 133109, "epoch": 1495, "lr": 5.166063991109592e-05}
{"train_loss": 0.08430109918117523, "global_step": 133110, "epoch": 1495, "lr": 5.166006049421999e-05}
{"train_loss": 0.037198979407548904, "global_step": 133111, "epoch": 1495, "lr": 5.165948107712088e-05}
{"train_loss": 0.14374510943889618, "global_step": 133112, "epoch": 1495, "lr": 5.165890165979868e-05}
{"train_loss": 0.10401390492916107, "global_step": 133113, "epoch": 1495, "lr": 5.165832224225346e-05}
{"train_loss": 0.16024966537952423, "global_step": 133114, "epoch": 1495, "lr": 5.165774282448531e-05}
{"train_loss": 0.07829111069440842, "global_step": 133115, "epoch": 1495, "lr": 5.165716340649428e-05}
{"train_loss": 0.11993304640054703, "global_step": 133116, "epoch": 1495, "lr": 5.165658398828047e-05}
{"train_loss": 0.15478329360485077, "global_step": 133117, "epoch": 1495, "lr": 5.1656004569843955e-05}
{"train_loss": 0.16524972021579742, "global_step": 133118, "epoch": 1495, "lr": 5.16554251511848e-05}
{"train_loss": 0.0831865668296814, "global_step": 133119, "epoch": 1495, "lr": 5.16548457323031e-05}
{"train_loss": 0.12030573934316635, "global_step": 133120, "epoch": 1495, "lr": 5.165426631319893e-05}
{"train_loss": 0.1663440316915512, "global_step": 133121, "epoch": 1495, "lr": 5.165368689387236e-05}
{"train_loss": 0.12185411155223846, "global_step": 133122, "epoch": 1495, "lr": 5.165310747432349e-05}
{"train_loss": 0.10397855192422867, "global_step": 133123, "epoch": 1495, "lr": 5.165252805455236e-05}
{"train_loss": 0.10909634828567505, "global_step": 133124, "epoch": 1495, "lr": 5.165194863455908e-05}
{"train_loss": 0.14227160811424255, "global_step": 133125, "epoch": 1495, "lr": 5.165136921434371e-05}
{"train_loss": 0.1072123795747757, "global_step": 133126, "epoch": 1495, "lr": 5.165078979390633e-05}
{"train_loss": 0.05015943944454193, "global_step": 133127, "epoch": 1495, "lr": 5.165021037324702e-05}
{"train_loss": 0.12808442115783691, "global_step": 133128, "epoch": 1495, "lr": 5.1649630952365854e-05}
{"train_loss": 0.1023443415760994, "global_step": 133129, "epoch": 1495, "lr": 5.1649051531262926e-05}
{"train_loss": 0.17056412994861603, "global_step": 133130, "epoch": 1495, "lr": 5.1648472109938306e-05}
{"train_loss": 0.08729195594787598, "global_step": 133131, "epoch": 1495, "lr": 5.164789268839206e-05}
{"train_loss": 0.07743526250123978, "global_step": 133132, "epoch": 1495, "lr": 5.164731326662427e-05}
{"train_loss": 0.15568499267101288, "global_step": 133133, "epoch": 1495, "lr": 5.164673384463503e-05}
{"train_loss": 0.13140593469142914, "global_step": 133134, "epoch": 1495, "lr": 5.1646154422424405e-05}
{"train_loss": 0.0893276184797287, "global_step": 133135, "epoch": 1495, "lr": 5.164557499999246e-05}
{"train_loss": 0.11595526337623596, "global_step": 133136, "epoch": 1495, "lr": 5.164499557733931e-05}
{"train_loss": 0.13195951282978058, "global_step": 133137, "epoch": 1495, "lr": 5.1644416154465e-05}
{"train_loss": 0.10194481909275055, "global_step": 133138, "epoch": 1495, "lr": 5.164383673136962e-05}
{"train_loss": 0.10253562778234482, "global_step": 133139, "epoch": 1495, "lr": 5.164325730805324e-05}
{"train_loss": 0.14620840549468994, "global_step": 133140, "epoch": 1495, "lr": 5.1642677884515954e-05}
{"train_loss": 0.12694747745990753, "global_step": 133141, "epoch": 1495, "lr": 5.1642098460757835e-05}
{"train_loss": 0.13671496510505676, "global_step": 133142, "epoch": 1495, "lr": 5.164151903677895e-05}
{"train_loss": 0.1218183163260476, "global_step": 133143, "epoch": 1495, "lr": 5.164093961257937e-05, "val_loss": 5.730227947235107, "train_action_mse_error": 14.61825942993164}
{"train_loss": 0.08173289895057678, "global_step": 133144, "epoch": 1496, "lr": 5.16403601881592e-05}
{"train_loss": 0.1100839376449585, "global_step": 133145, "epoch": 1496, "lr": 5.16397807635185e-05}
{"train_loss": 0.14737513661384583, "global_step": 133146, "epoch": 1496, "lr": 5.163920133865735e-05}
{"train_loss": 0.1177765354514122, "global_step": 133147, "epoch": 1496, "lr": 5.163862191357584e-05}
{"train_loss": 0.144467294216156, "global_step": 133148, "epoch": 1496, "lr": 5.163804248827402e-05}
{"train_loss": 0.06601278483867645, "global_step": 133149, "epoch": 1496, "lr": 5.163746306275199e-05}
{"train_loss": 0.12554818391799927, "global_step": 133150, "epoch": 1496, "lr": 5.163688363700983e-05}
{"train_loss": 0.1309068650007248, "global_step": 133151, "epoch": 1496, "lr": 5.1636304211047607e-05}
{"train_loss": 0.14832626283168793, "global_step": 133152, "epoch": 1496, "lr": 5.1635724784865394e-05}
{"train_loss": 0.17213353514671326, "global_step": 133153, "epoch": 1496, "lr": 5.163514535846328e-05}
{"train_loss": 0.08911655843257904, "global_step": 133154, "epoch": 1496, "lr": 5.163456593184134e-05}
{"train_loss": 0.1537531018257141, "global_step": 133155, "epoch": 1496, "lr": 5.163398650499966e-05}
{"train_loss": 0.1205986812710762, "global_step": 133156, "epoch": 1496, "lr": 5.163340707793831e-05}
{"train_loss": 0.06358479708433151, "global_step": 133157, "epoch": 1496, "lr": 5.163282765065737e-05}
{"train_loss": 0.2286759912967682, "global_step": 133158, "epoch": 1496, "lr": 5.163224822315691e-05}
{"train_loss": 0.07560514658689499, "global_step": 133159, "epoch": 1496, "lr": 5.1631668795437014e-05}
{"train_loss": 0.10735410451889038, "global_step": 133160, "epoch": 1496, "lr": 5.163108936749778e-05}
{"train_loss": 0.11839998513460159, "global_step": 133161, "epoch": 1496, "lr": 5.163050993933924e-05}
{"train_loss": 0.13860087096691132, "global_step": 133162, "epoch": 1496, "lr": 5.162993051096151e-05}
{"train_loss": 0.13259704411029816, "global_step": 133163, "epoch": 1496, "lr": 5.162935108236466e-05}
{"train_loss": 0.10627562552690506, "global_step": 133164, "epoch": 1496, "lr": 5.162877165354876e-05}
{"train_loss": 0.23787689208984375, "global_step": 133165, "epoch": 1496, "lr": 5.162819222451388e-05}
{"train_loss": 0.17097726464271545, "global_step": 133166, "epoch": 1496, "lr": 5.162761279526012e-05}
{"train_loss": 0.0918421521782875, "global_step": 133167, "epoch": 1496, "lr": 5.162703336578755e-05}
{"train_loss": 0.09445235878229141, "global_step": 133168, "epoch": 1496, "lr": 5.162645393609624e-05}
{"train_loss": 0.14596126973628998, "global_step": 133169, "epoch": 1496, "lr": 5.1625874506186265e-05}
{"train_loss": 0.12915368378162384, "global_step": 133170, "epoch": 1496, "lr": 5.1625295076057726e-05}
{"train_loss": 0.05571877211332321, "global_step": 133171, "epoch": 1496, "lr": 5.1624715645710666e-05}
{"train_loss": 0.13061431050300598, "global_step": 133172, "epoch": 1496, "lr": 5.1624136215145194e-05}
{"train_loss": 0.12720997631549835, "global_step": 133173, "epoch": 1496, "lr": 5.1623556784361384e-05}
{"train_loss": 0.13406899571418762, "global_step": 133174, "epoch": 1496, "lr": 5.162297735335929e-05}
{"train_loss": 0.1169985979795456, "global_step": 133175, "epoch": 1496, "lr": 5.162239792213902e-05}
{"train_loss": 0.19514036178588867, "global_step": 133176, "epoch": 1496, "lr": 5.1621818490700644e-05}
{"train_loss": 0.10120891034603119, "global_step": 133177, "epoch": 1496, "lr": 5.162123905904423e-05}
{"train_loss": 0.12166771292686462, "global_step": 133178, "epoch": 1496, "lr": 5.162065962716985e-05}
{"train_loss": 0.1698736995458603, "global_step": 133179, "epoch": 1496, "lr": 5.16200801950776e-05}
{"train_loss": 0.05200418829917908, "global_step": 133180, "epoch": 1496, "lr": 5.161950076276756e-05}
{"train_loss": 0.16777831315994263, "global_step": 133181, "epoch": 1496, "lr": 5.161892133023979e-05}
{"train_loss": 0.07502126693725586, "global_step": 133182, "epoch": 1496, "lr": 5.161834189749437e-05}
{"train_loss": 0.11552073061466217, "global_step": 133183, "epoch": 1496, "lr": 5.161776246453138e-05}
{"train_loss": 0.14683836698532104, "global_step": 133184, "epoch": 1496, "lr": 5.161718303135091e-05}
{"train_loss": 0.05508987233042717, "global_step": 133185, "epoch": 1496, "lr": 5.161660359795303e-05}
{"train_loss": 0.1250467151403427, "global_step": 133186, "epoch": 1496, "lr": 5.161602416433781e-05}
{"train_loss": 0.12824207544326782, "global_step": 133187, "epoch": 1496, "lr": 5.161544473050534e-05}
{"train_loss": 0.12134942412376404, "global_step": 133188, "epoch": 1496, "lr": 5.1614865296455694e-05}
{"train_loss": 0.14302928745746613, "global_step": 133189, "epoch": 1496, "lr": 5.161428586218895e-05}
{"train_loss": 0.19999633729457855, "global_step": 133190, "epoch": 1496, "lr": 5.161370642770517e-05}
{"train_loss": 0.13086163997650146, "global_step": 133191, "epoch": 1496, "lr": 5.161312699300446e-05}
{"train_loss": 0.2269897162914276, "global_step": 133192, "epoch": 1496, "lr": 5.161254755808688e-05}
{"train_loss": 0.08462320268154144, "global_step": 133193, "epoch": 1496, "lr": 5.1611968122952517e-05}
{"train_loss": 0.1290220320224762, "global_step": 133194, "epoch": 1496, "lr": 5.161138868760145e-05}
{"train_loss": 0.13949736952781677, "global_step": 133195, "epoch": 1496, "lr": 5.161080925203374e-05}
{"train_loss": 0.1598469763994217, "global_step": 133196, "epoch": 1496, "lr": 5.1610229816249475e-05}
{"train_loss": 0.18101699650287628, "global_step": 133197, "epoch": 1496, "lr": 5.1609650380248755e-05}
{"train_loss": 0.135182723402977, "global_step": 133198, "epoch": 1496, "lr": 5.1609070944031615e-05}
{"train_loss": 0.11289752274751663, "global_step": 133199, "epoch": 1496, "lr": 5.1608491507598176e-05}
{"train_loss": 0.13631978631019592, "global_step": 133200, "epoch": 1496, "lr": 5.1607912070948486e-05}
{"train_loss": 0.145058736205101, "global_step": 133201, "epoch": 1496, "lr": 5.160733263408263e-05}
{"train_loss": 0.17742781341075897, "global_step": 133202, "epoch": 1496, "lr": 5.160675319700068e-05}
{"train_loss": 0.09807387739419937, "global_step": 133203, "epoch": 1496, "lr": 5.1606173759702726e-05}
{"train_loss": 0.12113261222839355, "global_step": 133204, "epoch": 1496, "lr": 5.1605594322188846e-05}
{"train_loss": 0.11643816530704498, "global_step": 133205, "epoch": 1496, "lr": 5.1605014884459115e-05}
{"train_loss": 0.10104218870401382, "global_step": 133206, "epoch": 1496, "lr": 5.16044354465136e-05}
{"train_loss": 0.14682772755622864, "global_step": 133207, "epoch": 1496, "lr": 5.160385600835239e-05}
{"train_loss": 0.12954141199588776, "global_step": 133208, "epoch": 1496, "lr": 5.1603276569975554e-05}
{"train_loss": 0.11929892003536224, "global_step": 133209, "epoch": 1496, "lr": 5.1602697131383184e-05}
{"train_loss": 0.16905617713928223, "global_step": 133210, "epoch": 1496, "lr": 5.1602117692575356e-05}
{"train_loss": 0.10330993682146072, "global_step": 133211, "epoch": 1496, "lr": 5.160153825355214e-05}
{"train_loss": 0.13958492875099182, "global_step": 133212, "epoch": 1496, "lr": 5.160095881431362e-05}
{"train_loss": 0.14572644233703613, "global_step": 133213, "epoch": 1496, "lr": 5.160037937485986e-05}
{"train_loss": 0.16842201352119446, "global_step": 133214, "epoch": 1496, "lr": 5.159979993519096e-05}
{"train_loss": 0.12061800062656403, "global_step": 133215, "epoch": 1496, "lr": 5.159922049530699e-05}
{"train_loss": 0.1550619751214981, "global_step": 133216, "epoch": 1496, "lr": 5.159864105520802e-05}
{"train_loss": 0.1272161900997162, "global_step": 133217, "epoch": 1496, "lr": 5.1598061614894125e-05}
{"train_loss": 0.1793980449438095, "global_step": 133218, "epoch": 1496, "lr": 5.159748217436541e-05}
{"train_loss": 0.14904773235321045, "global_step": 133219, "epoch": 1496, "lr": 5.159690273362191e-05}
{"train_loss": 0.15770235657691956, "global_step": 133220, "epoch": 1496, "lr": 5.159632329266374e-05}
{"train_loss": 0.17203298211097717, "global_step": 133221, "epoch": 1496, "lr": 5.159574385149095e-05}
{"train_loss": 0.09657563269138336, "global_step": 133222, "epoch": 1496, "lr": 5.159516441010364e-05}
{"train_loss": 0.14189907908439636, "global_step": 133223, "epoch": 1496, "lr": 5.159458496850188e-05}
{"train_loss": 0.15081475675106049, "global_step": 133224, "epoch": 1496, "lr": 5.159400552668574e-05}
{"train_loss": 0.1219940334558487, "global_step": 133225, "epoch": 1496, "lr": 5.1593426084655316e-05}
{"train_loss": 0.16282446682453156, "global_step": 133226, "epoch": 1496, "lr": 5.159284664241066e-05}
{"train_loss": 0.16437920928001404, "global_step": 133227, "epoch": 1496, "lr": 5.159226719995187e-05}
{"train_loss": 0.11131705343723297, "global_step": 133228, "epoch": 1496, "lr": 5.1591687757279026e-05}
{"train_loss": 0.10579561442136765, "global_step": 133229, "epoch": 1496, "lr": 5.15911083143922e-05}
{"train_loss": 0.11882918328046799, "global_step": 133230, "epoch": 1496, "lr": 5.1590528871291466e-05}
{"train_loss": 0.06196864694356918, "global_step": 133231, "epoch": 1496, "lr": 5.1589949427976915e-05}
{"train_loss": 0.13075490040558108, "global_step": 133232, "epoch": 1496, "lr": 5.15893699844486e-05, "val_loss": 5.544342994689941}
{"train_loss": 0.09752821922302246, "global_step": 133233, "epoch": 1497, "lr": 5.158879054070662e-05}
{"train_loss": 0.08063504099845886, "global_step": 133234, "epoch": 1497, "lr": 5.1588211096751036e-05}
{"train_loss": 0.10639025270938873, "global_step": 133235, "epoch": 1497, "lr": 5.158763165258196e-05}
{"train_loss": 0.08167268335819244, "global_step": 133236, "epoch": 1497, "lr": 5.1587052208199426e-05}
{"train_loss": 0.13501997292041779, "global_step": 133237, "epoch": 1497, "lr": 5.158647276360353e-05}
{"train_loss": 0.06947049498558044, "global_step": 133238, "epoch": 1497, "lr": 5.158589331879437e-05}
{"train_loss": 0.19059084355831146, "global_step": 133239, "epoch": 1497, "lr": 5.1585313873772e-05}
{"train_loss": 0.15493012964725494, "global_step": 133240, "epoch": 1497, "lr": 5.1584734428536506e-05}
{"train_loss": 0.1386137455701828, "global_step": 133241, "epoch": 1497, "lr": 5.1584154983087954e-05}
{"train_loss": 0.1321692019701004, "global_step": 133242, "epoch": 1497, "lr": 5.158357553742643e-05}
{"train_loss": 0.149088054895401, "global_step": 133243, "epoch": 1497, "lr": 5.158299609155203e-05}
{"train_loss": 0.06780093163251877, "global_step": 133244, "epoch": 1497, "lr": 5.1582416645464794e-05}
{"train_loss": 0.09225299954414368, "global_step": 133245, "epoch": 1497, "lr": 5.158183719916483e-05}
{"train_loss": 0.08337065577507019, "global_step": 133246, "epoch": 1497, "lr": 5.15812577526522e-05}
{"train_loss": 0.1266951709985733, "global_step": 133247, "epoch": 1497, "lr": 5.1580678305927e-05}
{"train_loss": 0.19115442037582397, "global_step": 133248, "epoch": 1497, "lr": 5.1580098858989304e-05}
{"train_loss": 0.12184727191925049, "global_step": 133249, "epoch": 1497, "lr": 5.157951941183917e-05}
{"train_loss": 0.0940861776471138, "global_step": 133250, "epoch": 1497, "lr": 5.157893996447669e-05}
{"train_loss": 0.16212445497512817, "global_step": 133251, "epoch": 1497, "lr": 5.157836051690195e-05}
{"train_loss": 0.15387064218521118, "global_step": 133252, "epoch": 1497, "lr": 5.1577781069115e-05}
{"train_loss": 0.11424701660871506, "global_step": 133253, "epoch": 1497, "lr": 5.157720162111596e-05}
{"train_loss": 0.09020457416772842, "global_step": 133254, "epoch": 1497, "lr": 5.157662217290487e-05}
{"train_loss": 0.11484260857105255, "global_step": 133255, "epoch": 1497, "lr": 5.157604272448183e-05}
{"train_loss": 0.08101694285869598, "global_step": 133256, "epoch": 1497, "lr": 5.1575463275846904e-05}
{"train_loss": 0.15070386230945587, "global_step": 133257, "epoch": 1497, "lr": 5.1574883827000184e-05}
{"train_loss": 0.0871477723121643, "global_step": 133258, "epoch": 1497, "lr": 5.157430437794174e-05}
{"train_loss": 0.10453539341688156, "global_step": 133259, "epoch": 1497, "lr": 5.157372492867164e-05}
{"train_loss": 0.11385030299425125, "global_step": 133260, "epoch": 1497, "lr": 5.1573145479189976e-05}
{"train_loss": 0.08305775374174118, "global_step": 133261, "epoch": 1497, "lr": 5.1572566029496827e-05}
{"train_loss": 0.09994363784790039, "global_step": 133262, "epoch": 1497, "lr": 5.1571986579592244e-05}
{"train_loss": 0.1431725174188614, "global_step": 133263, "epoch": 1497, "lr": 5.157140712947634e-05}
{"train_loss": 0.0995565727353096, "global_step": 133264, "epoch": 1497, "lr": 5.157082767914919e-05}
{"train_loss": 0.11218729615211487, "global_step": 133265, "epoch": 1497, "lr": 5.157024822861085e-05}
{"train_loss": 0.09709403663873672, "global_step": 133266, "epoch": 1497, "lr": 5.1569668777861426e-05}
{"train_loss": 0.1534448266029358, "global_step": 133267, "epoch": 1497, "lr": 5.1569089326900965e-05}
{"train_loss": 0.0742943286895752, "global_step": 133268, "epoch": 1497, "lr": 5.156850987572955e-05}
{"train_loss": 0.08946485072374344, "global_step": 133269, "epoch": 1497, "lr": 5.156793042434729e-05}
{"train_loss": 0.12184792011976242, "global_step": 133270, "epoch": 1497, "lr": 5.1567350972754226e-05}
{"train_loss": 0.12076573073863983, "global_step": 133271, "epoch": 1497, "lr": 5.156677152095045e-05}
{"train_loss": 0.08225855231285095, "global_step": 133272, "epoch": 1497, "lr": 5.1566192068936047e-05}
{"train_loss": 0.16037696599960327, "global_step": 133273, "epoch": 1497, "lr": 5.156561261671109e-05}
{"train_loss": 0.11129622906446457, "global_step": 133274, "epoch": 1497, "lr": 5.1565033164275645e-05}
{"train_loss": 0.13822652399539948, "global_step": 133275, "epoch": 1497, "lr": 5.156445371162981e-05}
{"train_loss": 0.11347818374633789, "global_step": 133276, "epoch": 1497, "lr": 5.1563874258773646e-05}
{"train_loss": 0.11404971778392792, "global_step": 133277, "epoch": 1497, "lr": 5.1563294805707265e-05}
{"train_loss": 0.10426588356494904, "global_step": 133278, "epoch": 1497, "lr": 5.156271535243069e-05}
{"train_loss": 0.10830492526292801, "global_step": 133279, "epoch": 1497, "lr": 5.1562135898944034e-05}
{"train_loss": 0.1298544853925705, "global_step": 133280, "epoch": 1497, "lr": 5.156155644524735e-05}
{"train_loss": 0.0922934040427208, "global_step": 133281, "epoch": 1497, "lr": 5.156097699134076e-05}
{"train_loss": 0.11588267982006073, "global_step": 133282, "epoch": 1497, "lr": 5.1560397537224305e-05}
{"train_loss": 0.1584276407957077, "global_step": 133283, "epoch": 1497, "lr": 5.155981808289807e-05}
{"train_loss": 0.14878757297992706, "global_step": 133284, "epoch": 1497, "lr": 5.155923862836214e-05}
{"train_loss": 0.08607055991888046, "global_step": 133285, "epoch": 1497, "lr": 5.15586591736166e-05}
{"train_loss": 0.15673059225082397, "global_step": 133286, "epoch": 1497, "lr": 5.15580797186615e-05}
{"train_loss": 0.24355880916118622, "global_step": 133287, "epoch": 1497, "lr": 5.1557500263496936e-05}
{"train_loss": 0.15454785525798798, "global_step": 133288, "epoch": 1497, "lr": 5.1556920808123e-05}
{"train_loss": 0.15707452595233917, "global_step": 133289, "epoch": 1497, "lr": 5.155634135253975e-05}
{"train_loss": 0.1433667689561844, "global_step": 133290, "epoch": 1497, "lr": 5.1555761896747254e-05}
{"train_loss": 0.11605629324913025, "global_step": 133291, "epoch": 1497, "lr": 5.155518244074562e-05}
{"train_loss": 0.11181242018938065, "global_step": 133292, "epoch": 1497, "lr": 5.155460298453492e-05}
{"train_loss": 0.09290982037782669, "global_step": 133293, "epoch": 1497, "lr": 5.15540235281152e-05}
{"train_loss": 0.18645347654819489, "global_step": 133294, "epoch": 1497, "lr": 5.1553444071486566e-05}
{"train_loss": 0.138387992978096, "global_step": 133295, "epoch": 1497, "lr": 5.15528646146491e-05}
{"train_loss": 0.11553286761045456, "global_step": 133296, "epoch": 1497, "lr": 5.155228515760288e-05}
{"train_loss": 0.09547200798988342, "global_step": 133297, "epoch": 1497, "lr": 5.155170570034795e-05}
{"train_loss": 0.12751013040542603, "global_step": 133298, "epoch": 1497, "lr": 5.155112624288442e-05}
{"train_loss": 0.11011642217636108, "global_step": 133299, "epoch": 1497, "lr": 5.155054678521236e-05}
{"train_loss": 0.16510452330112457, "global_step": 133300, "epoch": 1497, "lr": 5.1549967327331846e-05}
{"train_loss": 0.1289832890033722, "global_step": 133301, "epoch": 1497, "lr": 5.154938786924296e-05}
{"train_loss": 0.1370047628879547, "global_step": 133302, "epoch": 1497, "lr": 5.154880841094578e-05}
{"train_loss": 0.1393548846244812, "global_step": 133303, "epoch": 1497, "lr": 5.154822895244038e-05}
{"train_loss": 0.12991376221179962, "global_step": 133304, "epoch": 1497, "lr": 5.154764949372683e-05}
{"train_loss": 0.054740093648433685, "global_step": 133305, "epoch": 1497, "lr": 5.154707003480523e-05}
{"train_loss": 0.13727432489395142, "global_step": 133306, "epoch": 1497, "lr": 5.154649057567564e-05}
{"train_loss": 0.13685721158981323, "global_step": 133307, "epoch": 1497, "lr": 5.154591111633815e-05}
{"train_loss": 0.11768776178359985, "global_step": 133308, "epoch": 1497, "lr": 5.154533165679283e-05}
{"train_loss": 0.13960273563861847, "global_step": 133309, "epoch": 1497, "lr": 5.1544752197039755e-05}
{"train_loss": 0.09465943276882172, "global_step": 133310, "epoch": 1497, "lr": 5.154417273707901e-05}
{"train_loss": 0.11581536382436752, "global_step": 133311, "epoch": 1497, "lr": 5.154359327691066e-05}
{"train_loss": 0.19418692588806152, "global_step": 133312, "epoch": 1497, "lr": 5.15430138165348e-05}
{"train_loss": 0.12412966042757034, "global_step": 133313, "epoch": 1497, "lr": 5.154243435595151e-05}
{"train_loss": 0.12428651750087738, "global_step": 133314, "epoch": 1497, "lr": 5.154185489516085e-05}
{"train_loss": 0.11721634119749069, "global_step": 133315, "epoch": 1497, "lr": 5.154127543416292e-05}
{"train_loss": 0.06979946792125702, "global_step": 133316, "epoch": 1497, "lr": 5.1540695972957763e-05}
{"train_loss": 0.07636167109012604, "global_step": 133317, "epoch": 1497, "lr": 5.154011651154548e-05}
{"train_loss": 0.1445247232913971, "global_step": 133318, "epoch": 1497, "lr": 5.1539537049926146e-05}
{"train_loss": 0.15556637942790985, "global_step": 133319, "epoch": 1497, "lr": 5.1538957588099845e-05}
{"train_loss": 0.0799306333065033, "global_step": 133320, "epoch": 1497, "lr": 5.153837812606665e-05}
{"train_loss": 0.12114045685261823, "global_step": 133321, "epoch": 1497, "lr": 5.153779866382664e-05, "val_loss": 5.6149139404296875}
{"train_loss": 0.18387047946453094, "global_step": 133322, "epoch": 1498, "lr": 5.153721920137989e-05}
{"train_loss": 0.13418035209178925, "global_step": 133323, "epoch": 1498, "lr": 5.153663973872649e-05}
{"train_loss": 0.17473554611206055, "global_step": 133324, "epoch": 1498, "lr": 5.1536060275866495e-05}
{"train_loss": 0.071220763027668, "global_step": 133325, "epoch": 1498, "lr": 5.1535480812799996e-05}
{"train_loss": 0.10541829466819763, "global_step": 133326, "epoch": 1498, "lr": 5.153490134952708e-05}
{"train_loss": 0.09592750668525696, "global_step": 133327, "epoch": 1498, "lr": 5.1534321886047796e-05}
{"train_loss": 0.10949639230966568, "global_step": 133328, "epoch": 1498, "lr": 5.1533742422362265e-05}
{"train_loss": 0.19575750827789307, "global_step": 133329, "epoch": 1498, "lr": 5.1533162958470525e-05}
{"train_loss": 0.10824117064476013, "global_step": 133330, "epoch": 1498, "lr": 5.153258349437268e-05}
{"train_loss": 0.07491114735603333, "global_step": 133331, "epoch": 1498, "lr": 5.1532004030068794e-05}
{"train_loss": 0.09044840931892395, "global_step": 133332, "epoch": 1498, "lr": 5.153142456555894e-05}
{"train_loss": 0.11723650991916656, "global_step": 133333, "epoch": 1498, "lr": 5.153084510084322e-05}
{"train_loss": 0.08950142562389374, "global_step": 133334, "epoch": 1498, "lr": 5.153026563592168e-05}
{"train_loss": 0.07574433833360672, "global_step": 133335, "epoch": 1498, "lr": 5.152968617079442e-05}
{"train_loss": 0.12618649005889893, "global_step": 133336, "epoch": 1498, "lr": 5.152910670546152e-05}
{"train_loss": 0.10854217410087585, "global_step": 133337, "epoch": 1498, "lr": 5.152852723992304e-05}
{"train_loss": 0.10101021826267242, "global_step": 133338, "epoch": 1498, "lr": 5.1527947774179075e-05}
{"train_loss": 0.08721336722373962, "global_step": 133339, "epoch": 1498, "lr": 5.1527368308229695e-05}
{"train_loss": 0.08247383683919907, "global_step": 133340, "epoch": 1498, "lr": 5.152678884207498e-05}
{"train_loss": 0.08761072158813477, "global_step": 133341, "epoch": 1498, "lr": 5.152620937571501e-05}
{"train_loss": 0.15774402022361755, "global_step": 133342, "epoch": 1498, "lr": 5.152562990914985e-05}
{"train_loss": 0.09785950928926468, "global_step": 133343, "epoch": 1498, "lr": 5.1525050442379595e-05}
{"train_loss": 0.09524930268526077, "global_step": 133344, "epoch": 1498, "lr": 5.1524470975404316e-05}
{"train_loss": 0.15219348669052124, "global_step": 133345, "epoch": 1498, "lr": 5.152389150822409e-05}
{"train_loss": 0.11107730120420456, "global_step": 133346, "epoch": 1498, "lr": 5.152331204083899e-05}
{"train_loss": 0.078109972178936, "global_step": 133347, "epoch": 1498, "lr": 5.15227325732491e-05}
{"train_loss": 0.09036703407764435, "global_step": 133348, "epoch": 1498, "lr": 5.152215310545451e-05}
{"train_loss": 0.14377711713314056, "global_step": 133349, "epoch": 1498, "lr": 5.152157363745527e-05}
{"train_loss": 0.09728258103132248, "global_step": 133350, "epoch": 1498, "lr": 5.152099416925148e-05}
{"train_loss": 0.14795833826065063, "global_step": 133351, "epoch": 1498, "lr": 5.15204147008432e-05}
{"train_loss": 0.13932469487190247, "global_step": 133352, "epoch": 1498, "lr": 5.151983523223053e-05}
{"train_loss": 0.08738619834184647, "global_step": 133353, "epoch": 1498, "lr": 5.151925576341353e-05}
{"train_loss": 0.18598252534866333, "global_step": 133354, "epoch": 1498, "lr": 5.151867629439229e-05}
{"train_loss": 0.11470764130353928, "global_step": 133355, "epoch": 1498, "lr": 5.1518096825166886e-05}
{"train_loss": 0.15327274799346924, "global_step": 133356, "epoch": 1498, "lr": 5.151751735573739e-05}
{"train_loss": 0.12757430970668793, "global_step": 133357, "epoch": 1498, "lr": 5.1516937886103876e-05}
{"train_loss": 0.08163157850503922, "global_step": 133358, "epoch": 1498, "lr": 5.151635841626643e-05}
{"train_loss": 0.07760624587535858, "global_step": 133359, "epoch": 1498, "lr": 5.151577894622513e-05}
{"train_loss": 0.12822696566581726, "global_step": 133360, "epoch": 1498, "lr": 5.1515199475980056e-05}
{"train_loss": 0.12769807875156403, "global_step": 133361, "epoch": 1498, "lr": 5.151462000553128e-05}
{"train_loss": 0.06537427008152008, "global_step": 133362, "epoch": 1498, "lr": 5.1514040534878875e-05}
{"train_loss": 0.0922277420759201, "global_step": 133363, "epoch": 1498, "lr": 5.1513461064022925e-05}
{"train_loss": 0.06944799423217773, "global_step": 133364, "epoch": 1498, "lr": 5.1512881592963525e-05}
{"train_loss": 0.04909293353557587, "global_step": 133365, "epoch": 1498, "lr": 5.1512302121700715e-05}
{"train_loss": 0.11690430343151093, "global_step": 133366, "epoch": 1498, "lr": 5.1511722650234603e-05}
{"train_loss": 0.1437821239233017, "global_step": 133367, "epoch": 1498, "lr": 5.151114317856527e-05}
{"train_loss": 0.10164075344800949, "global_step": 133368, "epoch": 1498, "lr": 5.1510563706692774e-05}
{"train_loss": 0.11210985481739044, "global_step": 133369, "epoch": 1498, "lr": 5.150998423461719e-05}
{"train_loss": 0.10788922756910324, "global_step": 133370, "epoch": 1498, "lr": 5.150940476233862e-05}
{"train_loss": 0.0760323703289032, "global_step": 133371, "epoch": 1498, "lr": 5.150882528985713e-05}
{"train_loss": 0.127423495054245, "global_step": 133372, "epoch": 1498, "lr": 5.15082458171728e-05}
{"train_loss": 0.10616420954465866, "global_step": 133373, "epoch": 1498, "lr": 5.1507666344285684e-05}
{"train_loss": 0.07075923681259155, "global_step": 133374, "epoch": 1498, "lr": 5.15070868711959e-05}
{"train_loss": 0.06320028007030487, "global_step": 133375, "epoch": 1498, "lr": 5.15065073979035e-05}
{"train_loss": 0.11306759715080261, "global_step": 133376, "epoch": 1498, "lr": 5.150592792440857e-05}
{"train_loss": 0.19419950246810913, "global_step": 133377, "epoch": 1498, "lr": 5.150534845071118e-05}
{"train_loss": 0.15275414288043976, "global_step": 133378, "epoch": 1498, "lr": 5.1504768976811427e-05}
{"train_loss": 0.12791316211223602, "global_step": 133379, "epoch": 1498, "lr": 5.150418950270937e-05}
{"train_loss": 0.095553457736969, "global_step": 133380, "epoch": 1498, "lr": 5.150361002840508e-05}
{"train_loss": 0.11288522183895111, "global_step": 133381, "epoch": 1498, "lr": 5.150303055389867e-05}
{"train_loss": 0.14134851098060608, "global_step": 133382, "epoch": 1498, "lr": 5.1502451079190184e-05}
{"train_loss": 0.07580830156803131, "global_step": 133383, "epoch": 1498, "lr": 5.1501871604279704e-05}
{"train_loss": 0.149795800447464, "global_step": 133384, "epoch": 1498, "lr": 5.150129212916733e-05}
{"train_loss": 0.09542489796876907, "global_step": 133385, "epoch": 1498, "lr": 5.1500712653853124e-05}
{"train_loss": 0.16167958080768585, "global_step": 133386, "epoch": 1498, "lr": 5.1500133178337165e-05}
{"train_loss": 0.12313871830701828, "global_step": 133387, "epoch": 1498, "lr": 5.149955370261952e-05}
{"train_loss": 0.1375257819890976, "global_step": 133388, "epoch": 1498, "lr": 5.149897422670029e-05}
{"train_loss": 0.10316728800535202, "global_step": 133389, "epoch": 1498, "lr": 5.1498394750579535e-05}
{"train_loss": 0.13215692341327667, "global_step": 133390, "epoch": 1498, "lr": 5.149781527425734e-05}
{"train_loss": 0.14301562309265137, "global_step": 133391, "epoch": 1498, "lr": 5.149723579773379e-05}
{"train_loss": 0.1301371157169342, "global_step": 133392, "epoch": 1498, "lr": 5.1496656321008944e-05}
{"train_loss": 0.09756336361169815, "global_step": 133393, "epoch": 1498, "lr": 5.14960768440829e-05}
{"train_loss": 0.13306432962417603, "global_step": 133394, "epoch": 1498, "lr": 5.149549736695571e-05}
{"train_loss": 0.14904548227787018, "global_step": 133395, "epoch": 1498, "lr": 5.149491788962749e-05}
{"train_loss": 0.16076192259788513, "global_step": 133396, "epoch": 1498, "lr": 5.149433841209828e-05}
{"train_loss": 0.08914400637149811, "global_step": 133397, "epoch": 1498, "lr": 5.149375893436819e-05}
{"train_loss": 0.16459277272224426, "global_step": 133398, "epoch": 1498, "lr": 5.149317945643727e-05}
{"train_loss": 0.16909845173358917, "global_step": 133399, "epoch": 1498, "lr": 5.149259997830561e-05}
{"train_loss": 0.08764687925577164, "global_step": 133400, "epoch": 1498, "lr": 5.1492020499973295e-05}
{"train_loss": 0.094960056245327, "global_step": 133401, "epoch": 1498, "lr": 5.14914410214404e-05}
{"train_loss": 0.13937842845916748, "global_step": 133402, "epoch": 1498, "lr": 5.149086154270699e-05}
{"train_loss": 0.13822710514068604, "global_step": 133403, "epoch": 1498, "lr": 5.149028206377315e-05}
{"train_loss": 0.0894218385219574, "global_step": 133404, "epoch": 1498, "lr": 5.148970258463897e-05}
{"train_loss": 0.13975143432617188, "global_step": 133405, "epoch": 1498, "lr": 5.14891231053045e-05}
{"train_loss": 0.07793448865413666, "global_step": 133406, "epoch": 1498, "lr": 5.148854362576986e-05}
{"train_loss": 0.10780467838048935, "global_step": 133407, "epoch": 1498, "lr": 5.148796414603508e-05}
{"train_loss": 0.09359318763017654, "global_step": 133408, "epoch": 1498, "lr": 5.148738466610028e-05}
{"train_loss": 0.10141395032405853, "global_step": 133409, "epoch": 1498, "lr": 5.148680518596551e-05}
{"train_loss": 0.11506287895896461, "global_step": 133410, "epoch": 1498, "lr": 5.148622570563085e-05, "val_loss": 5.721409320831299}
{"train_loss": 0.10474860668182373, "global_step": 133411, "epoch": 1499, "lr": 5.1485646225096397e-05}
{"train_loss": 0.06229473277926445, "global_step": 133412, "epoch": 1499, "lr": 5.148506674436222e-05}
{"train_loss": 0.10362366586923599, "global_step": 133413, "epoch": 1499, "lr": 5.148448726342838e-05}
{"train_loss": 0.09761270135641098, "global_step": 133414, "epoch": 1499, "lr": 5.148390778229498e-05}
{"train_loss": 0.15133386850357056, "global_step": 133415, "epoch": 1499, "lr": 5.148332830096209e-05}
{"train_loss": 0.11203449964523315, "global_step": 133416, "epoch": 1499, "lr": 5.148274881942977e-05}
{"train_loss": 0.10095169395208359, "global_step": 133417, "epoch": 1499, "lr": 5.148216933769812e-05}
{"train_loss": 0.11238245666027069, "global_step": 133418, "epoch": 1499, "lr": 5.148158985576721e-05}
{"train_loss": 0.16746416687965393, "global_step": 133419, "epoch": 1499, "lr": 5.148101037363712e-05}
{"train_loss": 0.12162528187036514, "global_step": 133420, "epoch": 1499, "lr": 5.148043089130792e-05}
{"train_loss": 0.17127466201782227, "global_step": 133421, "epoch": 1499, "lr": 5.14798514087797e-05}
{"train_loss": 0.07320519536733627, "global_step": 133422, "epoch": 1499, "lr": 5.147927192605253e-05}
{"train_loss": 0.0787200927734375, "global_step": 133423, "epoch": 1499, "lr": 5.1478692443126496e-05}
{"train_loss": 0.1017015352845192, "global_step": 133424, "epoch": 1499, "lr": 5.147811296000167e-05}
{"train_loss": 0.14954592287540436, "global_step": 133425, "epoch": 1499, "lr": 5.147753347667811e-05}
{"train_loss": 0.1735813170671463, "global_step": 133426, "epoch": 1499, "lr": 5.147695399315593e-05}
{"train_loss": 0.095690056681633, "global_step": 133427, "epoch": 1499, "lr": 5.1476374509435196e-05}
{"train_loss": 0.17783695459365845, "global_step": 133428, "epoch": 1499, "lr": 5.147579502551597e-05}
{"train_loss": 0.1214211955666542, "global_step": 133429, "epoch": 1499, "lr": 5.1475215541398346e-05}
{"train_loss": 0.1235702782869339, "global_step": 133430, "epoch": 1499, "lr": 5.1474636057082406e-05}
{"train_loss": 0.09818942844867706, "global_step": 133431, "epoch": 1499, "lr": 5.1474056572568206e-05}
{"train_loss": 0.08995882421731949, "global_step": 133432, "epoch": 1499, "lr": 5.1473477087855834e-05}
{"train_loss": 0.15274283289909363, "global_step": 133433, "epoch": 1499, "lr": 5.147289760294539e-05}
{"train_loss": 0.08946441859006882, "global_step": 133434, "epoch": 1499, "lr": 5.1472318117836914e-05}
{"train_loss": 0.1512458175420761, "global_step": 133435, "epoch": 1499, "lr": 5.147173863253052e-05}
{"train_loss": 0.22548575699329376, "global_step": 133436, "epoch": 1499, "lr": 5.147115914702625e-05}
{"train_loss": 0.11975090205669403, "global_step": 133437, "epoch": 1499, "lr": 5.147057966132422e-05}
{"train_loss": 0.18928800523281097, "global_step": 133438, "epoch": 1499, "lr": 5.1470000175424474e-05}
{"train_loss": 0.15696001052856445, "global_step": 133439, "epoch": 1499, "lr": 5.146942068932711e-05}
{"train_loss": 0.0856565609574318, "global_step": 133440, "epoch": 1499, "lr": 5.1468841203032195e-05}
{"train_loss": 0.13024526834487915, "global_step": 133441, "epoch": 1499, "lr": 5.1468261716539814e-05}
{"train_loss": 0.16277015209197998, "global_step": 133442, "epoch": 1499, "lr": 5.146768222985005e-05}
{"train_loss": 0.11078500747680664, "global_step": 133443, "epoch": 1499, "lr": 5.146710274296297e-05}
{"train_loss": 0.10635495185852051, "global_step": 133444, "epoch": 1499, "lr": 5.146652325587864e-05}
{"train_loss": 0.16494786739349365, "global_step": 133445, "epoch": 1499, "lr": 5.146594376859718e-05}
{"train_loss": 0.12124259024858475, "global_step": 133446, "epoch": 1499, "lr": 5.146536428111863e-05}
{"train_loss": 0.09881528466939926, "global_step": 133447, "epoch": 1499, "lr": 5.1464784793443075e-05}
{"train_loss": 0.13258539140224457, "global_step": 133448, "epoch": 1499, "lr": 5.146420530557061e-05}
{"train_loss": 0.13705237209796906, "global_step": 133449, "epoch": 1499, "lr": 5.1463625817501294e-05}
{"train_loss": 0.18223510682582855, "global_step": 133450, "epoch": 1499, "lr": 5.146304632923521e-05}
{"train_loss": 0.14124499261379242, "global_step": 133451, "epoch": 1499, "lr": 5.1462466840772436e-05}
{"train_loss": 0.06976941227912903, "global_step": 133452, "epoch": 1499, "lr": 5.146188735211305e-05}
{"train_loss": 0.09190265834331512, "global_step": 133453, "epoch": 1499, "lr": 5.146130786325714e-05}
{"train_loss": 0.06868283450603485, "global_step": 133454, "epoch": 1499, "lr": 5.146072837420477e-05}
{"train_loss": 0.05191812291741371, "global_step": 133455, "epoch": 1499, "lr": 5.146014888495602e-05}
{"train_loss": 0.12806402146816254, "global_step": 133456, "epoch": 1499, "lr": 5.145956939551096e-05}
{"train_loss": 0.18790195882320404, "global_step": 133457, "epoch": 1499, "lr": 5.1458989905869705e-05}
{"train_loss": 0.09941798448562622, "global_step": 133458, "epoch": 1499, "lr": 5.145841041603229e-05}
{"train_loss": 0.13086514174938202, "global_step": 133459, "epoch": 1499, "lr": 5.145783092599881e-05}
{"train_loss": 0.11747990548610687, "global_step": 133460, "epoch": 1499, "lr": 5.1457251435769346e-05}
{"train_loss": 0.18444760143756866, "global_step": 133461, "epoch": 1499, "lr": 5.145667194534397e-05}
{"train_loss": 0.0946650505065918, "global_step": 133462, "epoch": 1499, "lr": 5.145609245472275e-05}
{"train_loss": 0.12203802913427353, "global_step": 133463, "epoch": 1499, "lr": 5.1455512963905796e-05}
{"train_loss": 0.0622512586414814, "global_step": 133464, "epoch": 1499, "lr": 5.145493347289315e-05}
{"train_loss": 0.05845841392874718, "global_step": 133465, "epoch": 1499, "lr": 5.145435398168493e-05}
{"train_loss": 0.10124213248491287, "global_step": 133466, "epoch": 1499, "lr": 5.1453774490281183e-05}
{"train_loss": 0.0938701331615448, "global_step": 133467, "epoch": 1499, "lr": 5.1453194998681987e-05}
{"train_loss": 0.12031794339418411, "global_step": 133468, "epoch": 1499, "lr": 5.145261550688742e-05}
{"train_loss": 0.2039860039949417, "global_step": 133469, "epoch": 1499, "lr": 5.1452036014897575e-05}
{"train_loss": 0.1349729597568512, "global_step": 133470, "epoch": 1499, "lr": 5.1451456522712516e-05}
{"train_loss": 0.08824753761291504, "global_step": 133471, "epoch": 1499, "lr": 5.1450877030332334e-05}
{"train_loss": 0.17309719324111938, "global_step": 133472, "epoch": 1499, "lr": 5.1450297537757084e-05}
{"train_loss": 0.12586364150047302, "global_step": 133473, "epoch": 1499, "lr": 5.1449718044986874e-05}
{"train_loss": 0.13193409144878387, "global_step": 133474, "epoch": 1499, "lr": 5.144913855202176e-05}
{"train_loss": 0.10743115097284317, "global_step": 133475, "epoch": 1499, "lr": 5.144855905886183e-05}
{"train_loss": 0.1269223839044571, "global_step": 133476, "epoch": 1499, "lr": 5.144797956550716e-05}
{"train_loss": 0.1851288080215454, "global_step": 133477, "epoch": 1499, "lr": 5.144740007195783e-05}
{"train_loss": 0.05925591662526131, "global_step": 133478, "epoch": 1499, "lr": 5.144682057821391e-05}
{"train_loss": 0.1600932776927948, "global_step": 133479, "epoch": 1499, "lr": 5.144624108427548e-05}
{"train_loss": 0.1914946585893631, "global_step": 133480, "epoch": 1499, "lr": 5.1445661590142614e-05}
{"train_loss": 0.1374317854642868, "global_step": 133481, "epoch": 1499, "lr": 5.14450820958154e-05}
{"train_loss": 0.10190930962562561, "global_step": 133482, "epoch": 1499, "lr": 5.144450260129392e-05}
{"train_loss": 0.16721060872077942, "global_step": 133483, "epoch": 1499, "lr": 5.144392310657824e-05}
{"train_loss": 0.1030566394329071, "global_step": 133484, "epoch": 1499, "lr": 5.144334361166846e-05}
{"train_loss": 0.08589299023151398, "global_step": 133485, "epoch": 1499, "lr": 5.1442764116564625e-05}
{"train_loss": 0.19174465537071228, "global_step": 133486, "epoch": 1499, "lr": 5.144218462126682e-05}
{"train_loss": 0.07734236121177673, "global_step": 133487, "epoch": 1499, "lr": 5.1441605125775136e-05}
{"train_loss": 0.1018359363079071, "global_step": 133488, "epoch": 1499, "lr": 5.1441025630089655e-05}
{"train_loss": 0.15618713200092316, "global_step": 133489, "epoch": 1499, "lr": 5.144044613421044e-05}
{"train_loss": 0.10206956416368484, "global_step": 133490, "epoch": 1499, "lr": 5.143986663813757e-05}
{"train_loss": 0.11088021844625473, "global_step": 133491, "epoch": 1499, "lr": 5.1439287141871126e-05}
{"train_loss": 0.11426103860139847, "global_step": 133492, "epoch": 1499, "lr": 5.14387076454112e-05}
{"train_loss": 0.10111144185066223, "global_step": 133493, "epoch": 1499, "lr": 5.1438128148757845e-05}
{"train_loss": 0.11121338605880737, "global_step": 133494, "epoch": 1499, "lr": 5.143754865191115e-05}
{"train_loss": 0.17073096334934235, "global_step": 133495, "epoch": 1499, "lr": 5.1436969154871203e-05}
{"train_loss": 0.045655280351638794, "global_step": 133496, "epoch": 1499, "lr": 5.143638965763807e-05}
{"train_loss": 0.16291220486164093, "global_step": 133497, "epoch": 1499, "lr": 5.143581016021183e-05}
{"train_loss": 0.11304119229316711, "global_step": 133498, "epoch": 1499, "lr": 5.143523066259256e-05}
{"train_loss": 0.12299963199857915, "global_step": 133499, "epoch": 1499, "lr": 5.1434651164780336e-05, "val_loss": 5.7272162437438965}
{"train_loss": 0.08328834176063538, "global_step": 133500, "epoch": 1500, "lr": 5.143407166677525e-05}
{"train_loss": 0.08243013173341751, "global_step": 133501, "epoch": 1500, "lr": 5.143349216857737e-05}
{"train_loss": 0.10413352400064468, "global_step": 133502, "epoch": 1500, "lr": 5.143291267018677e-05}
{"train_loss": 0.14619313180446625, "global_step": 133503, "epoch": 1500, "lr": 5.143233317160355e-05}
{"train_loss": 0.08367826044559479, "global_step": 133504, "epoch": 1500, "lr": 5.143175367282776e-05}
{"train_loss": 0.1451738029718399, "global_step": 133505, "epoch": 1500, "lr": 5.143117417385949e-05}
{"train_loss": 0.17511360347270966, "global_step": 133506, "epoch": 1500, "lr": 5.143059467469881e-05}
{"train_loss": 0.11029861867427826, "global_step": 133507, "epoch": 1500, "lr": 5.14300151753458e-05}
{"train_loss": 0.0966685339808464, "global_step": 133508, "epoch": 1500, "lr": 5.142943567580055e-05}
{"train_loss": 0.1319890022277832, "global_step": 133509, "epoch": 1500, "lr": 5.1428856176063135e-05}
{"train_loss": 0.10089558362960815, "global_step": 133510, "epoch": 1500, "lr": 5.142827667613361e-05}
{"train_loss": 0.11935476213693619, "global_step": 133511, "epoch": 1500, "lr": 5.142769717601208e-05}
{"train_loss": 0.06895239651203156, "global_step": 133512, "epoch": 1500, "lr": 5.142711767569862e-05}
{"train_loss": 0.15398502349853516, "global_step": 133513, "epoch": 1500, "lr": 5.142653817519329e-05}
{"train_loss": 0.13861657679080963, "global_step": 133514, "epoch": 1500, "lr": 5.142595867449619e-05}
{"train_loss": 0.09351050853729248, "global_step": 133515, "epoch": 1500, "lr": 5.142537917360738e-05}
{"train_loss": 0.13245467841625214, "global_step": 133516, "epoch": 1500, "lr": 5.1424799672526935e-05}
{"train_loss": 0.1433534026145935, "global_step": 133517, "epoch": 1500, "lr": 5.1424220171254954e-05}
{"train_loss": 0.1316535621881485, "global_step": 133518, "epoch": 1500, "lr": 5.142364066979151e-05}
{"train_loss": 0.07735464721918106, "global_step": 133519, "epoch": 1500, "lr": 5.142306116813667e-05}
{"train_loss": 0.10029731690883636, "global_step": 133520, "epoch": 1500, "lr": 5.1422481666290515e-05}
{"train_loss": 0.08653311431407928, "global_step": 133521, "epoch": 1500, "lr": 5.142190216425312e-05}
{"train_loss": 0.11519058793783188, "global_step": 133522, "epoch": 1500, "lr": 5.1421322662024575e-05}
{"train_loss": 0.16644129157066345, "global_step": 133523, "epoch": 1500, "lr": 5.1420743159604966e-05}
{"train_loss": 0.086420938372612, "global_step": 133524, "epoch": 1500, "lr": 5.1420163656994334e-05}
{"train_loss": 0.21857626736164093, "global_step": 133525, "epoch": 1500, "lr": 5.141958415419279e-05}
{"train_loss": 0.07169584929943085, "global_step": 133526, "epoch": 1500, "lr": 5.141900465120039e-05}
{"train_loss": 0.12952329218387604, "global_step": 133527, "epoch": 1500, "lr": 5.141842514801722e-05}
{"train_loss": 0.1020214706659317, "global_step": 133528, "epoch": 1500, "lr": 5.1417845644643366e-05}
{"train_loss": 0.10698403418064117, "global_step": 133529, "epoch": 1500, "lr": 5.14172661410789e-05}
{"train_loss": 0.1777445524930954, "global_step": 133530, "epoch": 1500, "lr": 5.141668663732391e-05}
{"train_loss": 0.12968139350414276, "global_step": 133531, "epoch": 1500, "lr": 5.1416107133378446e-05}
{"train_loss": 0.09726426750421524, "global_step": 133532, "epoch": 1500, "lr": 5.141552762924261e-05}
{"train_loss": 0.12205254286527634, "global_step": 133533, "epoch": 1500, "lr": 5.1414948124916475e-05}
{"train_loss": 0.12341932952404022, "global_step": 133534, "epoch": 1500, "lr": 5.14143686204001e-05}
{"train_loss": 0.13034218549728394, "global_step": 133535, "epoch": 1500, "lr": 5.14137891156936e-05}
{"train_loss": 0.11476673930883408, "global_step": 133536, "epoch": 1500, "lr": 5.141320961079703e-05}
{"train_loss": 0.11484494060277939, "global_step": 133537, "epoch": 1500, "lr": 5.141263010571047e-05}
{"train_loss": 0.14772477746009827, "global_step": 133538, "epoch": 1500, "lr": 5.1412050600434e-05}
{"train_loss": 0.10793651640415192, "global_step": 133539, "epoch": 1500, "lr": 5.14114710949677e-05}
{"train_loss": 0.12123484164476395, "global_step": 133540, "epoch": 1500, "lr": 5.141089158931165e-05}
{"train_loss": 0.07940530776977539, "global_step": 133541, "epoch": 1500, "lr": 5.1410312083465906e-05}
{"train_loss": 0.13733050227165222, "global_step": 133542, "epoch": 1500, "lr": 5.140973257743059e-05}
{"train_loss": 0.1123332530260086, "global_step": 133543, "epoch": 1500, "lr": 5.140915307120573e-05}
{"train_loss": 0.14989927411079407, "global_step": 133544, "epoch": 1500, "lr": 5.1408573564791426e-05}
{"train_loss": 0.10365287214517593, "global_step": 133545, "epoch": 1500, "lr": 5.140799405818777e-05}
{"train_loss": 0.09734506905078888, "global_step": 133546, "epoch": 1500, "lr": 5.1407414551394817e-05}
{"train_loss": 0.08162061870098114, "global_step": 133547, "epoch": 1500, "lr": 5.140683504441266e-05}
{"train_loss": 0.10859585553407669, "global_step": 133548, "epoch": 1500, "lr": 5.1406255537241367e-05}
{"train_loss": 0.1752515286207199, "global_step": 133549, "epoch": 1500, "lr": 5.1405676029881014e-05}
{"train_loss": 0.10048477351665497, "global_step": 133550, "epoch": 1500, "lr": 5.1405096522331685e-05}
{"train_loss": 0.135443314909935, "global_step": 133551, "epoch": 1500, "lr": 5.140451701459347e-05}
{"train_loss": 0.07217928767204285, "global_step": 133552, "epoch": 1500, "lr": 5.1403937506666424e-05}
{"train_loss": 0.11742080003023148, "global_step": 133553, "epoch": 1500, "lr": 5.140335799855064e-05}
{"train_loss": 0.11087220907211304, "global_step": 133554, "epoch": 1500, "lr": 5.1402778490246185e-05}
{"train_loss": 0.06108623370528221, "global_step": 133555, "epoch": 1500, "lr": 5.1402198981753146e-05}
{"train_loss": 0.06805428117513657, "global_step": 133556, "epoch": 1500, "lr": 5.1401619473071606e-05}
{"train_loss": 0.17210376262664795, "global_step": 133557, "epoch": 1500, "lr": 5.140103996420164e-05}
{"train_loss": 0.1607365906238556, "global_step": 133558, "epoch": 1500, "lr": 5.140046045514331e-05}
{"train_loss": 0.12944455444812775, "global_step": 133559, "epoch": 1500, "lr": 5.13998809458967e-05}
{"train_loss": 0.1312997043132782, "global_step": 133560, "epoch": 1500, "lr": 5.139930143646191e-05}
{"train_loss": 0.12261436879634857, "global_step": 133561, "epoch": 1500, "lr": 5.1398721926839e-05}
{"train_loss": 0.08745444566011429, "global_step": 133562, "epoch": 1500, "lr": 5.139814241702804e-05}
{"train_loss": 0.12128409743309021, "global_step": 133563, "epoch": 1500, "lr": 5.139756290702912e-05}
{"train_loss": 0.11824368685483932, "global_step": 133564, "epoch": 1500, "lr": 5.139698339684231e-05}
{"train_loss": 0.20921896398067474, "global_step": 133565, "epoch": 1500, "lr": 5.1396403886467694e-05}
{"train_loss": 0.18873964250087738, "global_step": 133566, "epoch": 1500, "lr": 5.1395824375905354e-05}
{"train_loss": 0.12761767208576202, "global_step": 133567, "epoch": 1500, "lr": 5.139524486515536e-05}
{"train_loss": 0.08599495887756348, "global_step": 133568, "epoch": 1500, "lr": 5.139466535421778e-05}
{"train_loss": 0.10495692491531372, "global_step": 133569, "epoch": 1500, "lr": 5.139408584309272e-05}
{"train_loss": 0.07306072860956192, "global_step": 133570, "epoch": 1500, "lr": 5.139350633178024e-05}
{"train_loss": 0.14452019333839417, "global_step": 133571, "epoch": 1500, "lr": 5.1392926820280406e-05}
{"train_loss": 0.0959743782877922, "global_step": 133572, "epoch": 1500, "lr": 5.139234730859333e-05}
{"train_loss": 0.11120542138814926, "global_step": 133573, "epoch": 1500, "lr": 5.139176779671905e-05}
{"train_loss": 0.13623985648155212, "global_step": 133574, "epoch": 1500, "lr": 5.139118828465768e-05}
{"train_loss": 0.08714679628610611, "global_step": 133575, "epoch": 1500, "lr": 5.1390608772409274e-05}
{"train_loss": 0.09156838804483414, "global_step": 133576, "epoch": 1500, "lr": 5.1390029259973914e-05}
{"train_loss": 0.11581216007471085, "global_step": 133577, "epoch": 1500, "lr": 5.138944974735169e-05}
{"train_loss": 0.1460656374692917, "global_step": 133578, "epoch": 1500, "lr": 5.1388870234542674e-05}
{"train_loss": 0.191664457321167, "global_step": 133579, "epoch": 1500, "lr": 5.138829072154694e-05}
{"train_loss": 0.1497253179550171, "global_step": 133580, "epoch": 1500, "lr": 5.1387711208364555e-05}
{"train_loss": 0.20455282926559448, "global_step": 133581, "epoch": 1500, "lr": 5.138713169499563e-05}
{"train_loss": 0.1414223313331604, "global_step": 133582, "epoch": 1500, "lr": 5.1386552181440206e-05}
{"train_loss": 0.14101777970790863, "global_step": 133583, "epoch": 1500, "lr": 5.138597266769839e-05}
{"train_loss": 0.0798710435628891, "global_step": 133584, "epoch": 1500, "lr": 5.138539315377023e-05}
{"train_loss": 0.0726284310221672, "global_step": 133585, "epoch": 1500, "lr": 5.1384813639655834e-05}
{"train_loss": 0.20333746075630188, "global_step": 133586, "epoch": 1500, "lr": 5.138423412535526e-05}
{"train_loss": 0.08772256225347519, "global_step": 133587, "epoch": 1500, "lr": 5.138365461086859e-05}
{"train_loss": 0.12111718538269568, "global_step": 133588, "epoch": 1500, "lr": 5.138307509619591e-05, "train/sim_max_reward_0": 0.41878296855964436, "train/sim_max_reward_1": 0.18553172373985946, "train/sim_max_reward_2": 0.36133762153294957, "train/sim_max_reward_3": 0.3535008414609347, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9855625259507437, "test/sim_max_reward_4500000": 0.9636084287892138, "test/sim_max_reward_4500001": 0.9681708576221627, "test/sim_max_reward_4500002": 0.9600652014548745, "test/sim_max_reward_4500003": 0.18115513148320392, "test/sim_max_reward_4500004": 0.5822861919036513, "test/sim_max_reward_4500005": 0.2651592790510816, "test/sim_max_reward_4500006": 0.9595377520555847, "test/sim_max_reward_4500007": 0.877880594733799, "test/sim_max_reward_4500008": 0.07521390434036344, "test/sim_max_reward_4500009": 0.04971296352052077, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9762759458395749, "test/sim_max_reward_4500012": 0.9400581447754593, "test/sim_max_reward_4500013": 0.43411594152706323, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.5319918298325601, "test/sim_max_reward_4500016": 0.5707882112071179, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 0.4199660494201476, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9827789890708771, "test/sim_max_reward_4500022": 0.9959140516298626, "test/sim_max_reward_4500023": 0.8852687728776031, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.40635517284526546, "test/sim_max_reward_4500026": 0.9532913206100202, "test/sim_max_reward_4500027": 0.9545247114798883, "test/sim_max_reward_4500028": 0.5523966468087884, "test/sim_max_reward_4500029": 0.4372636007515444, "test/sim_max_reward_4500030": 0.24587532428892037, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6454746379148736, "test/sim_max_reward_4500033": 0.9967343602172727, "test/sim_max_reward_4500034": 0.5628492780681487, "test/sim_max_reward_4500035": 0.421973445000001, "test/sim_max_reward_4500036": 0.9590649448657114, "test/sim_max_reward_4500037": 0.7259955778717214, "test/sim_max_reward_4500038": 0.49487525152772666, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.5486839151628591, "test/sim_max_reward_4500041": 0.520349733235387, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.7852552493292749, "test/sim_max_reward_4500044": 0.955684918543368, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.5897302954777527, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.8254949023838045, "train/mean_score": 0.5507859468740219, "test/mean_score": 0.6040364305503411, "val_loss": 5.883962154388428, "train_action_mse_error": 20.048959732055664}
{"train_loss": 0.10970675945281982, "global_step": 133589, "epoch": 1501, "lr": 5.1382495581337276e-05}
{"train_loss": 0.1900486797094345, "global_step": 133590, "epoch": 1501, "lr": 5.1381916066292804e-05}
{"train_loss": 0.11103493720293045, "global_step": 133591, "epoch": 1501, "lr": 5.138133655106254e-05}
{"train_loss": 0.13882941007614136, "global_step": 133592, "epoch": 1501, "lr": 5.138075703564658e-05}
{"train_loss": 0.09282892197370529, "global_step": 133593, "epoch": 1501, "lr": 5.138017752004498e-05}
{"train_loss": 0.15216326713562012, "global_step": 133594, "epoch": 1501, "lr": 5.1379598004257845e-05}
{"train_loss": 0.1817084401845932, "global_step": 133595, "epoch": 1501, "lr": 5.1379018488285234e-05}
{"train_loss": 0.10174787789583206, "global_step": 133596, "epoch": 1501, "lr": 5.137843897212724e-05}
{"train_loss": 0.14816640317440033, "global_step": 133597, "epoch": 1501, "lr": 5.137785945578392e-05}
{"train_loss": 0.09406714886426926, "global_step": 133598, "epoch": 1501, "lr": 5.1377279939255377e-05}
{"train_loss": 0.13740287721157074, "global_step": 133599, "epoch": 1501, "lr": 5.137670042254166e-05}
{"train_loss": 0.17101992666721344, "global_step": 133600, "epoch": 1501, "lr": 5.1376120905642876e-05}
{"train_loss": 0.092356376349926, "global_step": 133601, "epoch": 1501, "lr": 5.137554138855909e-05}
{"train_loss": 0.1346786916255951, "global_step": 133602, "epoch": 1501, "lr": 5.137496187129036e-05}
{"train_loss": 0.19161583483219147, "global_step": 133603, "epoch": 1501, "lr": 5.1374382353836804e-05}
{"train_loss": 0.16152113676071167, "global_step": 133604, "epoch": 1501, "lr": 5.137380283619847e-05}
{"train_loss": 0.11137401312589645, "global_step": 133605, "epoch": 1501, "lr": 5.137322331837544e-05}
{"train_loss": 0.11720245331525803, "global_step": 133606, "epoch": 1501, "lr": 5.13726438003678e-05}
{"train_loss": 0.1734667867422104, "global_step": 133607, "epoch": 1501, "lr": 5.137206428217561e-05}
{"train_loss": 0.12674090266227722, "global_step": 133608, "epoch": 1501, "lr": 5.137148476379898e-05}
{"train_loss": 0.12374196946620941, "global_step": 133609, "epoch": 1501, "lr": 5.137090524523798e-05}
{"train_loss": 0.06197666376829147, "global_step": 133610, "epoch": 1501, "lr": 5.1370325726492664e-05}
{"train_loss": 0.09125351905822754, "global_step": 133611, "epoch": 1501, "lr": 5.136974620756313e-05}
{"train_loss": 0.06160050630569458, "global_step": 133612, "epoch": 1501, "lr": 5.136916668844944e-05}
{"train_loss": 0.13208019733428955, "global_step": 133613, "epoch": 1501, "lr": 5.136858716915169e-05}
{"train_loss": 0.07745744287967682, "global_step": 133614, "epoch": 1501, "lr": 5.136800764966996e-05}
{"train_loss": 0.20111404359340668, "global_step": 133615, "epoch": 1501, "lr": 5.136742813000431e-05}
{"train_loss": 0.09271567314863205, "global_step": 133616, "epoch": 1501, "lr": 5.1366848610154826e-05}
{"train_loss": 0.14007891714572906, "global_step": 133617, "epoch": 1501, "lr": 5.136626909012159e-05}
{"train_loss": 0.08666781336069107, "global_step": 133618, "epoch": 1501, "lr": 5.136568956990466e-05}
{"train_loss": 0.11565639078617096, "global_step": 133619, "epoch": 1501, "lr": 5.1365110049504164e-05}
{"train_loss": 0.14091327786445618, "global_step": 133620, "epoch": 1501, "lr": 5.1364530528920117e-05}
{"train_loss": 0.16260020434856415, "global_step": 133621, "epoch": 1501, "lr": 5.1363951008152635e-05}
{"train_loss": 0.10369885712862015, "global_step": 133622, "epoch": 1501, "lr": 5.1363371487201775e-05}
{"train_loss": 0.14205050468444824, "global_step": 133623, "epoch": 1501, "lr": 5.136279196606764e-05}
{"train_loss": 0.1331612765789032, "global_step": 133624, "epoch": 1501, "lr": 5.136221244475028e-05}
{"train_loss": 0.0869118794798851, "global_step": 133625, "epoch": 1501, "lr": 5.136163292324978e-05}
{"train_loss": 0.10872159898281097, "global_step": 133626, "epoch": 1501, "lr": 5.136105340156624e-05}
{"train_loss": 0.11621887236833572, "global_step": 133627, "epoch": 1501, "lr": 5.1360473879699725e-05}
{"train_loss": 0.12346041947603226, "global_step": 133628, "epoch": 1501, "lr": 5.1359894357650304e-05}
{"train_loss": 0.09232571721076965, "global_step": 133629, "epoch": 1501, "lr": 5.135931483541807e-05}
{"train_loss": 0.09137595444917679, "global_step": 133630, "epoch": 1501, "lr": 5.135873531300309e-05}
{"train_loss": 0.14817479252815247, "global_step": 133631, "epoch": 1501, "lr": 5.135815579040545e-05}
{"train_loss": 0.1312636286020279, "global_step": 133632, "epoch": 1501, "lr": 5.135757626762521e-05}
{"train_loss": 0.2000901699066162, "global_step": 133633, "epoch": 1501, "lr": 5.135699674466247e-05}
{"train_loss": 0.10154008120298386, "global_step": 133634, "epoch": 1501, "lr": 5.135641722151729e-05}
{"train_loss": 0.15742747485637665, "global_step": 133635, "epoch": 1501, "lr": 5.1355837698189765e-05}
{"train_loss": 0.21604886651039124, "global_step": 133636, "epoch": 1501, "lr": 5.1355258174679964e-05}
{"train_loss": 0.15855905413627625, "global_step": 133637, "epoch": 1501, "lr": 5.1354678650987956e-05}
{"train_loss": 0.13236945867538452, "global_step": 133638, "epoch": 1501, "lr": 5.135409912711383e-05}
{"train_loss": 0.1373937427997589, "global_step": 133639, "epoch": 1501, "lr": 5.135351960305768e-05}
{"train_loss": 0.08808869868516922, "global_step": 133640, "epoch": 1501, "lr": 5.135294007881956e-05}
{"train_loss": 0.09070815145969391, "global_step": 133641, "epoch": 1501, "lr": 5.1352360554399534e-05}
{"train_loss": 0.09408590942621231, "global_step": 133642, "epoch": 1501, "lr": 5.1351781029797704e-05}
{"train_loss": 0.19113609194755554, "global_step": 133643, "epoch": 1501, "lr": 5.135120150501415e-05}
{"train_loss": 0.2218158096075058, "global_step": 133644, "epoch": 1501, "lr": 5.135062198004894e-05}
{"train_loss": 0.10090572386980057, "global_step": 133645, "epoch": 1501, "lr": 5.135004245490216e-05}
{"train_loss": 0.12306570261716843, "global_step": 133646, "epoch": 1501, "lr": 5.134946292957388e-05}
{"train_loss": 0.15079091489315033, "global_step": 133647, "epoch": 1501, "lr": 5.1348883404064185e-05}
{"train_loss": 0.10900449007749557, "global_step": 133648, "epoch": 1501, "lr": 5.134830387837315e-05}
{"train_loss": 0.10842873156070709, "global_step": 133649, "epoch": 1501, "lr": 5.134772435250085e-05}
{"train_loss": 0.09216229617595673, "global_step": 133650, "epoch": 1501, "lr": 5.134714482644736e-05}
{"train_loss": 0.1667039543390274, "global_step": 133651, "epoch": 1501, "lr": 5.134656530021277e-05}
{"train_loss": 0.13259252905845642, "global_step": 133652, "epoch": 1501, "lr": 5.134598577379714e-05}
{"train_loss": 0.16640226542949677, "global_step": 133653, "epoch": 1501, "lr": 5.134540624720057e-05}
{"train_loss": 0.05774941295385361, "global_step": 133654, "epoch": 1501, "lr": 5.1344826720423126e-05}
{"train_loss": 0.11657306551933289, "global_step": 133655, "epoch": 1501, "lr": 5.134424719346488e-05}
{"train_loss": 0.21900199353694916, "global_step": 133656, "epoch": 1501, "lr": 5.1343667666325926e-05}
{"train_loss": 0.08280474692583084, "global_step": 133657, "epoch": 1501, "lr": 5.134308813900631e-05}
{"train_loss": 0.20404395461082458, "global_step": 133658, "epoch": 1501, "lr": 5.134250861150617e-05}
{"train_loss": 0.13073278963565826, "global_step": 133659, "epoch": 1501, "lr": 5.134192908382552e-05}
{"train_loss": 0.12062741816043854, "global_step": 133660, "epoch": 1501, "lr": 5.134134955596447e-05}
{"train_loss": 0.1581706553697586, "global_step": 133661, "epoch": 1501, "lr": 5.134077002792308e-05}
{"train_loss": 0.11502711474895477, "global_step": 133662, "epoch": 1501, "lr": 5.134019049970145e-05}
{"train_loss": 0.19390857219696045, "global_step": 133663, "epoch": 1501, "lr": 5.1339610971299646e-05}
{"train_loss": 0.0974874421954155, "global_step": 133664, "epoch": 1501, "lr": 5.1339031442717744e-05}
{"train_loss": 0.11497820168733597, "global_step": 133665, "epoch": 1501, "lr": 5.133845191395583e-05}
{"train_loss": 0.18080922961235046, "global_step": 133666, "epoch": 1501, "lr": 5.1337872385013987e-05}
{"train_loss": 0.1284799724817276, "global_step": 133667, "epoch": 1501, "lr": 5.1337292855892274e-05}
{"train_loss": 0.16775470972061157, "global_step": 133668, "epoch": 1501, "lr": 5.1336713326590766e-05}
{"train_loss": 0.14043265581130981, "global_step": 133669, "epoch": 1501, "lr": 5.133613379710957e-05}
{"train_loss": 0.11393076926469803, "global_step": 133670, "epoch": 1501, "lr": 5.133555426744875e-05}
{"train_loss": 0.17651619017124176, "global_step": 133671, "epoch": 1501, "lr": 5.1334974737608364e-05}
{"train_loss": 0.10741538554430008, "global_step": 133672, "epoch": 1501, "lr": 5.1334395207588525e-05}
{"train_loss": 0.1948479562997818, "global_step": 133673, "epoch": 1501, "lr": 5.1333815677389284e-05}
{"train_loss": 0.08826043456792831, "global_step": 133674, "epoch": 1501, "lr": 5.1333236147010723e-05}
{"train_loss": 0.12745164334774017, "global_step": 133675, "epoch": 1501, "lr": 5.133265661645294e-05}
{"train_loss": 0.14878517389297485, "global_step": 133676, "epoch": 1501, "lr": 5.133207708571598e-05}
{"train_loss": 0.1327115824718154, "global_step": 133677, "epoch": 1501, "lr": 5.1331497554799946e-05, "val_loss": 5.659011363983154}
{"train_loss": 0.09717796742916107, "global_step": 133678, "epoch": 1502, "lr": 5.1330918023704924e-05}
{"train_loss": 0.15582694113254547, "global_step": 133679, "epoch": 1502, "lr": 5.133033849243094e-05}
{"train_loss": 0.11647573113441467, "global_step": 133680, "epoch": 1502, "lr": 5.132975896097813e-05}
{"train_loss": 0.09098870307207108, "global_step": 133681, "epoch": 1502, "lr": 5.132917942934655e-05}
{"train_loss": 0.1123378723859787, "global_step": 133682, "epoch": 1502, "lr": 5.1328599897536276e-05}
{"train_loss": 0.14697612822055817, "global_step": 133683, "epoch": 1502, "lr": 5.132802036554739e-05}
{"train_loss": 0.12969200313091278, "global_step": 133684, "epoch": 1502, "lr": 5.132744083337997e-05}
{"train_loss": 0.13220848143100739, "global_step": 133685, "epoch": 1502, "lr": 5.132686130103409e-05}
{"train_loss": 0.11079362034797668, "global_step": 133686, "epoch": 1502, "lr": 5.1326281768509833e-05}
{"train_loss": 0.10609186440706253, "global_step": 133687, "epoch": 1502, "lr": 5.132570223580726e-05}
{"train_loss": 0.21316294372081757, "global_step": 133688, "epoch": 1502, "lr": 5.1325122702926474e-05}
{"train_loss": 0.153494730591774, "global_step": 133689, "epoch": 1502, "lr": 5.1324543169867535e-05}
{"train_loss": 0.10068371891975403, "global_step": 133690, "epoch": 1502, "lr": 5.1323963636630536e-05}
{"train_loss": 0.15695753693580627, "global_step": 133691, "epoch": 1502, "lr": 5.1323384103215546e-05}
{"train_loss": 0.11593183875083923, "global_step": 133692, "epoch": 1502, "lr": 5.132280456962263e-05}
{"train_loss": 0.10922721773386002, "global_step": 133693, "epoch": 1502, "lr": 5.132222503585189e-05}
{"train_loss": 0.1486504226922989, "global_step": 133694, "epoch": 1502, "lr": 5.132164550190339e-05}
{"train_loss": 0.19610102474689484, "global_step": 133695, "epoch": 1502, "lr": 5.132106596777721e-05}
{"train_loss": 0.1405816376209259, "global_step": 133696, "epoch": 1502, "lr": 5.132048643347343e-05}
{"train_loss": 0.15286274254322052, "global_step": 133697, "epoch": 1502, "lr": 5.131990689899214e-05}
{"train_loss": 0.16634689271450043, "global_step": 133698, "epoch": 1502, "lr": 5.131932736433338e-05}
{"train_loss": 0.09532611072063446, "global_step": 133699, "epoch": 1502, "lr": 5.131874782949726e-05}
{"train_loss": 0.13783137500286102, "global_step": 133700, "epoch": 1502, "lr": 5.131816829448386e-05}
{"train_loss": 0.09421935677528381, "global_step": 133701, "epoch": 1502, "lr": 5.131758875929325e-05}
{"train_loss": 0.10182235389947891, "global_step": 133702, "epoch": 1502, "lr": 5.1317009223925494e-05}
{"train_loss": 0.17379677295684814, "global_step": 133703, "epoch": 1502, "lr": 5.131642968838068e-05}
{"train_loss": 0.07566339522600174, "global_step": 133704, "epoch": 1502, "lr": 5.1315850152658896e-05}
{"train_loss": 0.12905578315258026, "global_step": 133705, "epoch": 1502, "lr": 5.1315270616760215e-05}
{"train_loss": 0.10517145693302155, "global_step": 133706, "epoch": 1502, "lr": 5.131469108068471e-05}
{"train_loss": 0.20658010244369507, "global_step": 133707, "epoch": 1502, "lr": 5.1314111544432454e-05}
{"train_loss": 0.17386339604854584, "global_step": 133708, "epoch": 1502, "lr": 5.1313532008003536e-05}
{"train_loss": 0.09698053449392319, "global_step": 133709, "epoch": 1502, "lr": 5.131295247139804e-05}
{"train_loss": 0.07607150822877884, "global_step": 133710, "epoch": 1502, "lr": 5.1312372934616015e-05}
{"train_loss": 0.10057464987039566, "global_step": 133711, "epoch": 1502, "lr": 5.131179339765757e-05}
{"train_loss": 0.10724092274904251, "global_step": 133712, "epoch": 1502, "lr": 5.131121386052277e-05}
{"train_loss": 0.09626984596252441, "global_step": 133713, "epoch": 1502, "lr": 5.131063432321168e-05}
{"train_loss": 0.10476119071245193, "global_step": 133714, "epoch": 1502, "lr": 5.131005478572441e-05}
{"train_loss": 0.10105086863040924, "global_step": 133715, "epoch": 1502, "lr": 5.1309475248061004e-05}
{"train_loss": 0.15528064966201782, "global_step": 133716, "epoch": 1502, "lr": 5.130889571022156e-05}
{"train_loss": 0.117888905107975, "global_step": 133717, "epoch": 1502, "lr": 5.130831617220615e-05}
{"train_loss": 0.16875553131103516, "global_step": 133718, "epoch": 1502, "lr": 5.1307736634014856e-05}
{"train_loss": 0.09607598185539246, "global_step": 133719, "epoch": 1502, "lr": 5.130715709564775e-05}
{"train_loss": 0.20893937349319458, "global_step": 133720, "epoch": 1502, "lr": 5.130657755710492e-05}
{"train_loss": 0.07944006472826004, "global_step": 133721, "epoch": 1502, "lr": 5.130599801838642e-05}
{"train_loss": 0.11204662173986435, "global_step": 133722, "epoch": 1502, "lr": 5.130541847949236e-05}
{"train_loss": 0.10500257462263107, "global_step": 133723, "epoch": 1502, "lr": 5.1304838940422796e-05}
{"train_loss": 0.10648878663778305, "global_step": 133724, "epoch": 1502, "lr": 5.130425940117782e-05}
{"train_loss": 0.1977093368768692, "global_step": 133725, "epoch": 1502, "lr": 5.1303679861757494e-05}
{"train_loss": 0.13500428199768066, "global_step": 133726, "epoch": 1502, "lr": 5.1303100322161914e-05}
{"train_loss": 0.08614733815193176, "global_step": 133727, "epoch": 1502, "lr": 5.130252078239113e-05}
{"train_loss": 0.11653865873813629, "global_step": 133728, "epoch": 1502, "lr": 5.130194124244525e-05}
{"train_loss": 0.13982374966144562, "global_step": 133729, "epoch": 1502, "lr": 5.130136170232434e-05}
{"train_loss": 0.13797055184841156, "global_step": 133730, "epoch": 1502, "lr": 5.130078216202847e-05}
{"train_loss": 0.09329783171415329, "global_step": 133731, "epoch": 1502, "lr": 5.1300202621557736e-05}
{"train_loss": 0.08618645370006561, "global_step": 133732, "epoch": 1502, "lr": 5.129962308091221e-05}
{"train_loss": 0.17015422880649567, "global_step": 133733, "epoch": 1502, "lr": 5.1299043540091953e-05}
{"train_loss": 0.15834049880504608, "global_step": 133734, "epoch": 1502, "lr": 5.1298463999097054e-05}
{"train_loss": 0.15120363235473633, "global_step": 133735, "epoch": 1502, "lr": 5.1297884457927606e-05}
{"train_loss": 0.14418953657150269, "global_step": 133736, "epoch": 1502, "lr": 5.1297304916583655e-05}
{"train_loss": 0.18152424693107605, "global_step": 133737, "epoch": 1502, "lr": 5.1296725375065316e-05}
{"train_loss": 0.12179108709096909, "global_step": 133738, "epoch": 1502, "lr": 5.129614583337263e-05}
{"train_loss": 0.11049577593803406, "global_step": 133739, "epoch": 1502, "lr": 5.12955662915057e-05}
{"train_loss": 0.10963859409093857, "global_step": 133740, "epoch": 1502, "lr": 5.12949867494646e-05}
{"train_loss": 0.14025911688804626, "global_step": 133741, "epoch": 1502, "lr": 5.1294407207249406e-05}
{"train_loss": 0.10280655324459076, "global_step": 133742, "epoch": 1502, "lr": 5.1293827664860196e-05}
{"train_loss": 0.11495566368103027, "global_step": 133743, "epoch": 1502, "lr": 5.1293248122297046e-05}
{"train_loss": 0.07553982734680176, "global_step": 133744, "epoch": 1502, "lr": 5.129266857956003e-05}
{"train_loss": 0.14123281836509705, "global_step": 133745, "epoch": 1502, "lr": 5.1292089036649225e-05}
{"train_loss": 0.12932150065898895, "global_step": 133746, "epoch": 1502, "lr": 5.129150949356473e-05}
{"train_loss": 0.13401876389980316, "global_step": 133747, "epoch": 1502, "lr": 5.12909299503066e-05}
{"train_loss": 0.12361932545900345, "global_step": 133748, "epoch": 1502, "lr": 5.129035040687491e-05}
{"train_loss": 0.14993199706077576, "global_step": 133749, "epoch": 1502, "lr": 5.128977086326976e-05}
{"train_loss": 0.10015830397605896, "global_step": 133750, "epoch": 1502, "lr": 5.128919131949121e-05}
{"train_loss": 0.20014379918575287, "global_step": 133751, "epoch": 1502, "lr": 5.1288611775539354e-05}
{"train_loss": 0.16230319440364838, "global_step": 133752, "epoch": 1502, "lr": 5.1288032231414255e-05}
{"train_loss": 0.1095595583319664, "global_step": 133753, "epoch": 1502, "lr": 5.128745268711599e-05}
{"train_loss": 0.10178640484809875, "global_step": 133754, "epoch": 1502, "lr": 5.128687314264464e-05}
{"train_loss": 0.08862707763910294, "global_step": 133755, "epoch": 1502, "lr": 5.12862935980003e-05}
{"train_loss": 0.10108795017004013, "global_step": 133756, "epoch": 1502, "lr": 5.128571405318302e-05}
{"train_loss": 0.12773798406124115, "global_step": 133757, "epoch": 1502, "lr": 5.12851345081929e-05}
{"train_loss": 0.1049904078245163, "global_step": 133758, "epoch": 1502, "lr": 5.1284554963030004e-05}
{"train_loss": 0.21835346519947052, "global_step": 133759, "epoch": 1502, "lr": 5.128397541769442e-05}
{"train_loss": 0.053893640637397766, "global_step": 133760, "epoch": 1502, "lr": 5.1283395872186225e-05}
{"train_loss": 0.07136882841587067, "global_step": 133761, "epoch": 1502, "lr": 5.128281632650548e-05}
{"train_loss": 0.05509498342871666, "global_step": 133762, "epoch": 1502, "lr": 5.128223678065228e-05}
{"train_loss": 0.1308727115392685, "global_step": 133763, "epoch": 1502, "lr": 5.12816572346267e-05}
{"train_loss": 0.09273385256528854, "global_step": 133764, "epoch": 1502, "lr": 5.128107768842882e-05}
{"train_loss": 0.10986076295375824, "global_step": 133765, "epoch": 1502, "lr": 5.128049814205872e-05}
{"train_loss": 0.12573214214337006, "global_step": 133766, "epoch": 1502, "lr": 5.127991859551646e-05, "val_loss": 5.674472808837891}
{"train_loss": 0.11540171504020691, "global_step": 133767, "epoch": 1503, "lr": 5.1279339048802144e-05}
{"train_loss": 0.07231041043996811, "global_step": 133768, "epoch": 1503, "lr": 5.127875950191583e-05}
{"train_loss": 0.11474435031414032, "global_step": 133769, "epoch": 1503, "lr": 5.1278179954857594e-05}
{"train_loss": 0.0856235921382904, "global_step": 133770, "epoch": 1503, "lr": 5.127760040762754e-05}
{"train_loss": 0.060976747423410416, "global_step": 133771, "epoch": 1503, "lr": 5.127702086022571e-05}
{"train_loss": 0.12741327285766602, "global_step": 133772, "epoch": 1503, "lr": 5.12764413126522e-05}
{"train_loss": 0.09937310963869095, "global_step": 133773, "epoch": 1503, "lr": 5.127586176490711e-05}
{"train_loss": 0.10655194520950317, "global_step": 133774, "epoch": 1503, "lr": 5.127528221699047e-05}
{"train_loss": 0.12823784351348877, "global_step": 133775, "epoch": 1503, "lr": 5.12747026689024e-05}
{"train_loss": 0.09311830252408981, "global_step": 133776, "epoch": 1503, "lr": 5.127412312064296e-05}
{"train_loss": 0.12656129896640778, "global_step": 133777, "epoch": 1503, "lr": 5.127354357221222e-05}
{"train_loss": 0.10545510053634644, "global_step": 133778, "epoch": 1503, "lr": 5.127296402361028e-05}
{"train_loss": 0.0758618488907814, "global_step": 133779, "epoch": 1503, "lr": 5.12723844748372e-05}
{"train_loss": 0.11836493760347366, "global_step": 133780, "epoch": 1503, "lr": 5.127180492589306e-05}
{"train_loss": 0.08672133088111877, "global_step": 133781, "epoch": 1503, "lr": 5.127122537677794e-05}
{"train_loss": 0.06575310975313187, "global_step": 133782, "epoch": 1503, "lr": 5.1270645827491924e-05}
{"train_loss": 0.08503318578004837, "global_step": 133783, "epoch": 1503, "lr": 5.127006627803509e-05}
{"train_loss": 0.1112276166677475, "global_step": 133784, "epoch": 1503, "lr": 5.126948672840751e-05}
{"train_loss": 0.10698599368333817, "global_step": 133785, "epoch": 1503, "lr": 5.126890717860926e-05}
{"train_loss": 0.13471628725528717, "global_step": 133786, "epoch": 1503, "lr": 5.126832762864042e-05}
{"train_loss": 0.12835200130939484, "global_step": 133787, "epoch": 1503, "lr": 5.1267748078501064e-05}
{"train_loss": 0.06157964468002319, "global_step": 133788, "epoch": 1503, "lr": 5.1267168528191276e-05}
{"train_loss": 0.10812856256961823, "global_step": 133789, "epoch": 1503, "lr": 5.126658897771114e-05}
{"train_loss": 0.113188736140728, "global_step": 133790, "epoch": 1503, "lr": 5.126600942706072e-05}
{"train_loss": 0.0785609558224678, "global_step": 133791, "epoch": 1503, "lr": 5.126542987624011e-05}
{"train_loss": 0.09710917621850967, "global_step": 133792, "epoch": 1503, "lr": 5.126485032524937e-05}
{"train_loss": 0.1372380554676056, "global_step": 133793, "epoch": 1503, "lr": 5.1264270774088585e-05}
{"train_loss": 0.14246532320976257, "global_step": 133794, "epoch": 1503, "lr": 5.126369122275784e-05}
{"train_loss": 0.07519254833459854, "global_step": 133795, "epoch": 1503, "lr": 5.12631116712572e-05}
{"train_loss": 0.11960715055465698, "global_step": 133796, "epoch": 1503, "lr": 5.126253211958676e-05}
{"train_loss": 0.12720553576946259, "global_step": 133797, "epoch": 1503, "lr": 5.126195256774659e-05}
{"train_loss": 0.13821297883987427, "global_step": 133798, "epoch": 1503, "lr": 5.1261373015736744e-05}
{"train_loss": 0.14685046672821045, "global_step": 133799, "epoch": 1503, "lr": 5.1260793463557344e-05}
{"train_loss": 0.08843456208705902, "global_step": 133800, "epoch": 1503, "lr": 5.1260213911208436e-05}
{"train_loss": 0.1056891456246376, "global_step": 133801, "epoch": 1503, "lr": 5.1259634358690114e-05}
{"train_loss": 0.18747931718826294, "global_step": 133802, "epoch": 1503, "lr": 5.125905480600244e-05}
{"train_loss": 0.10976344347000122, "global_step": 133803, "epoch": 1503, "lr": 5.125847525314551e-05}
{"train_loss": 0.0894191712141037, "global_step": 133804, "epoch": 1503, "lr": 5.125789570011939e-05}
{"train_loss": 0.06243838369846344, "global_step": 133805, "epoch": 1503, "lr": 5.125731614692416e-05}
{"train_loss": 0.07376884669065475, "global_step": 133806, "epoch": 1503, "lr": 5.125673659355988e-05}
{"train_loss": 0.07628992199897766, "global_step": 133807, "epoch": 1503, "lr": 5.125615704002668e-05}
{"train_loss": 0.14210298657417297, "global_step": 133808, "epoch": 1503, "lr": 5.125557748632459e-05}
{"train_loss": 0.10494199395179749, "global_step": 133809, "epoch": 1503, "lr": 5.125499793245371e-05}
{"train_loss": 0.13727660477161407, "global_step": 133810, "epoch": 1503, "lr": 5.12544183784141e-05}
{"train_loss": 0.11737139523029327, "global_step": 133811, "epoch": 1503, "lr": 5.1253838824205854e-05}
{"train_loss": 0.15204952657222748, "global_step": 133812, "epoch": 1503, "lr": 5.1253259269829054e-05}
{"train_loss": 0.11278814822435379, "global_step": 133813, "epoch": 1503, "lr": 5.1252679715283756e-05}
{"train_loss": 0.18081748485565186, "global_step": 133814, "epoch": 1503, "lr": 5.125210016057006e-05}
{"train_loss": 0.12863202393054962, "global_step": 133815, "epoch": 1503, "lr": 5.125152060568802e-05}
{"train_loss": 0.18949876725673676, "global_step": 133816, "epoch": 1503, "lr": 5.1250941050637743e-05}
{"train_loss": 0.12050867825746536, "global_step": 133817, "epoch": 1503, "lr": 5.1250361495419284e-05}
{"train_loss": 0.06613924354314804, "global_step": 133818, "epoch": 1503, "lr": 5.1249781940032726e-05}
{"train_loss": 0.08047465234994888, "global_step": 133819, "epoch": 1503, "lr": 5.124920238447816e-05}
{"train_loss": 0.12010002136230469, "global_step": 133820, "epoch": 1503, "lr": 5.124862282875565e-05}
{"train_loss": 0.135599285364151, "global_step": 133821, "epoch": 1503, "lr": 5.124804327286527e-05}
{"train_loss": 0.10634095221757889, "global_step": 133822, "epoch": 1503, "lr": 5.124746371680712e-05}
{"train_loss": 0.1037847101688385, "global_step": 133823, "epoch": 1503, "lr": 5.124688416058125e-05}
{"train_loss": 0.1355455368757248, "global_step": 133824, "epoch": 1503, "lr": 5.124630460418777e-05}
{"train_loss": 0.21993865072727203, "global_step": 133825, "epoch": 1503, "lr": 5.1245725047626716e-05}
{"train_loss": 0.09193646162748337, "global_step": 133826, "epoch": 1503, "lr": 5.124514549089819e-05}
{"train_loss": 0.0946800634264946, "global_step": 133827, "epoch": 1503, "lr": 5.12445659340023e-05}
{"train_loss": 0.10033579915761948, "global_step": 133828, "epoch": 1503, "lr": 5.124398637693907e-05}
{"train_loss": 0.0823330283164978, "global_step": 133829, "epoch": 1503, "lr": 5.124340681970861e-05}
{"train_loss": 0.10693071782588959, "global_step": 133830, "epoch": 1503, "lr": 5.1242827262310976e-05}
{"train_loss": 0.1184978112578392, "global_step": 133831, "epoch": 1503, "lr": 5.124224770474627e-05}
{"train_loss": 0.06162692978978157, "global_step": 133832, "epoch": 1503, "lr": 5.1241668147014555e-05}
{"train_loss": 0.18977275490760803, "global_step": 133833, "epoch": 1503, "lr": 5.1241088589115914e-05}
{"train_loss": 0.048416633158922195, "global_step": 133834, "epoch": 1503, "lr": 5.1240509031050435e-05}
{"train_loss": 0.10271797329187393, "global_step": 133835, "epoch": 1503, "lr": 5.123992947281817e-05}
{"train_loss": 0.12434466183185577, "global_step": 133836, "epoch": 1503, "lr": 5.123934991441921e-05}
{"train_loss": 0.1157512366771698, "global_step": 133837, "epoch": 1503, "lr": 5.123877035585364e-05}
{"train_loss": 0.1643214225769043, "global_step": 133838, "epoch": 1503, "lr": 5.123819079712153e-05}
{"train_loss": 0.11354637891054153, "global_step": 133839, "epoch": 1503, "lr": 5.123761123822297e-05}
{"train_loss": 0.11066412925720215, "global_step": 133840, "epoch": 1503, "lr": 5.123703167915802e-05}
{"train_loss": 0.16741223633289337, "global_step": 133841, "epoch": 1503, "lr": 5.123645211992676e-05}
{"train_loss": 0.1136903464794159, "global_step": 133842, "epoch": 1503, "lr": 5.123587256052928e-05}
{"train_loss": 0.14430713653564453, "global_step": 133843, "epoch": 1503, "lr": 5.1235293000965646e-05}
{"train_loss": 0.14432799816131592, "global_step": 133844, "epoch": 1503, "lr": 5.123471344123595e-05}
{"train_loss": 0.08394574373960495, "global_step": 133845, "epoch": 1503, "lr": 5.123413388134026e-05}
{"train_loss": 0.1936103254556656, "global_step": 133846, "epoch": 1503, "lr": 5.1233554321278666e-05}
{"train_loss": 0.13682979345321655, "global_step": 133847, "epoch": 1503, "lr": 5.123297476105122e-05}
{"train_loss": 0.06903871893882751, "global_step": 133848, "epoch": 1503, "lr": 5.1232395200658027e-05}
{"train_loss": 0.12843675911426544, "global_step": 133849, "epoch": 1503, "lr": 5.1231815640099147e-05}
{"train_loss": 0.1796768307685852, "global_step": 133850, "epoch": 1503, "lr": 5.123123607937467e-05}
{"train_loss": 0.19904492795467377, "global_step": 133851, "epoch": 1503, "lr": 5.1230656518484666e-05}
{"train_loss": 0.07177116721868515, "global_step": 133852, "epoch": 1503, "lr": 5.123007695742921e-05}
{"train_loss": 0.0923105999827385, "global_step": 133853, "epoch": 1503, "lr": 5.122949739620838e-05}
{"train_loss": 0.14260153472423553, "global_step": 133854, "epoch": 1503, "lr": 5.1228917834822276e-05}
{"train_loss": 0.1148964450851585, "global_step": 133855, "epoch": 1503, "lr": 5.122833827327095e-05, "val_loss": 5.815269947052002}
{"train_loss": 0.11433519423007965, "global_step": 133856, "epoch": 1504, "lr": 5.122775871155448e-05}
{"train_loss": 0.12640312314033508, "global_step": 133857, "epoch": 1504, "lr": 5.1227179149672964e-05}
{"train_loss": 0.11429767310619354, "global_step": 133858, "epoch": 1504, "lr": 5.122659958762647e-05}
{"train_loss": 0.023722289130091667, "global_step": 133859, "epoch": 1504, "lr": 5.122602002541507e-05}
{"train_loss": 0.1341298520565033, "global_step": 133860, "epoch": 1504, "lr": 5.122544046303886e-05}
{"train_loss": 0.09641942381858826, "global_step": 133861, "epoch": 1504, "lr": 5.122486090049787e-05}
{"train_loss": 0.1432570219039917, "global_step": 133862, "epoch": 1504, "lr": 5.122428133779224e-05}
{"train_loss": 0.0835881233215332, "global_step": 133863, "epoch": 1504, "lr": 5.122370177492201e-05}
{"train_loss": 0.13698916137218475, "global_step": 133864, "epoch": 1504, "lr": 5.122312221188728e-05}
{"train_loss": 0.12978044152259827, "global_step": 133865, "epoch": 1504, "lr": 5.12225426486881e-05}
{"train_loss": 0.1531948298215866, "global_step": 133866, "epoch": 1504, "lr": 5.122196308532459e-05}
{"train_loss": 0.115882009267807, "global_step": 133867, "epoch": 1504, "lr": 5.122138352179678e-05}
{"train_loss": 0.1069844514131546, "global_step": 133868, "epoch": 1504, "lr": 5.122080395810478e-05}
{"train_loss": 0.14650022983551025, "global_step": 133869, "epoch": 1504, "lr": 5.122022439424865e-05}
{"train_loss": 0.14542365074157715, "global_step": 133870, "epoch": 1504, "lr": 5.121964483022848e-05}
{"train_loss": 0.12137113511562347, "global_step": 133871, "epoch": 1504, "lr": 5.1219065266044344e-05}
{"train_loss": 0.1234554871916771, "global_step": 133872, "epoch": 1504, "lr": 5.121848570169632e-05}
{"train_loss": 0.08650295436382294, "global_step": 133873, "epoch": 1504, "lr": 5.121790613718448e-05}
{"train_loss": 0.09741732478141785, "global_step": 133874, "epoch": 1504, "lr": 5.1217326572508904e-05}
{"train_loss": 0.1518305391073227, "global_step": 133875, "epoch": 1504, "lr": 5.121674700766968e-05}
{"train_loss": 0.1132269874215126, "global_step": 133876, "epoch": 1504, "lr": 5.121616744266688e-05}
{"train_loss": 0.08606379479169846, "global_step": 133877, "epoch": 1504, "lr": 5.121558787750058e-05}
{"train_loss": 0.09532778710126877, "global_step": 133878, "epoch": 1504, "lr": 5.121500831217087e-05}
{"train_loss": 0.046902790665626526, "global_step": 133879, "epoch": 1504, "lr": 5.121442874667779e-05}
{"train_loss": 0.17653074860572815, "global_step": 133880, "epoch": 1504, "lr": 5.121384918102146e-05}
{"train_loss": 0.12390697747468948, "global_step": 133881, "epoch": 1504, "lr": 5.121326961520194e-05}
{"train_loss": 0.07070160657167435, "global_step": 133882, "epoch": 1504, "lr": 5.1212690049219314e-05}
{"train_loss": 0.11244030296802521, "global_step": 133883, "epoch": 1504, "lr": 5.121211048307366e-05}
{"train_loss": 0.1016928032040596, "global_step": 133884, "epoch": 1504, "lr": 5.121153091676505e-05}
{"train_loss": 0.12346173077821732, "global_step": 133885, "epoch": 1504, "lr": 5.121095135029358e-05}
{"train_loss": 0.1409558355808258, "global_step": 133886, "epoch": 1504, "lr": 5.121037178365929e-05}
{"train_loss": 0.17136290669441223, "global_step": 133887, "epoch": 1504, "lr": 5.1209792216862285e-05}
{"train_loss": 0.0846487507224083, "global_step": 133888, "epoch": 1504, "lr": 5.120921264990264e-05}
{"train_loss": 0.15768787264823914, "global_step": 133889, "epoch": 1504, "lr": 5.1208633082780435e-05}
{"train_loss": 0.0999581515789032, "global_step": 133890, "epoch": 1504, "lr": 5.1208053515495734e-05}
{"train_loss": 0.21799829602241516, "global_step": 133891, "epoch": 1504, "lr": 5.120747394804863e-05}
{"train_loss": 0.08120740205049515, "global_step": 133892, "epoch": 1504, "lr": 5.12068943804392e-05}
{"train_loss": 0.1039726734161377, "global_step": 133893, "epoch": 1504, "lr": 5.120631481266751e-05}
{"train_loss": 0.11081269383430481, "global_step": 133894, "epoch": 1504, "lr": 5.1205735244733655e-05}
{"train_loss": 0.09341233223676682, "global_step": 133895, "epoch": 1504, "lr": 5.120515567663769e-05}
{"train_loss": 0.10806445777416229, "global_step": 133896, "epoch": 1504, "lr": 5.120457610837972e-05}
{"train_loss": 0.14007462561130524, "global_step": 133897, "epoch": 1504, "lr": 5.1203996539959796e-05}
{"train_loss": 0.09454499185085297, "global_step": 133898, "epoch": 1504, "lr": 5.1203416971378016e-05}
{"train_loss": 0.10406003147363663, "global_step": 133899, "epoch": 1504, "lr": 5.120283740263446e-05}
{"train_loss": 0.09732196480035782, "global_step": 133900, "epoch": 1504, "lr": 5.120225783372918e-05}
{"train_loss": 0.1014968752861023, "global_step": 133901, "epoch": 1504, "lr": 5.120167826466229e-05}
{"train_loss": 0.17196621000766754, "global_step": 133902, "epoch": 1504, "lr": 5.120109869543384e-05}
{"train_loss": 0.08788052946329117, "global_step": 133903, "epoch": 1504, "lr": 5.120051912604392e-05}
{"train_loss": 0.11526890099048615, "global_step": 133904, "epoch": 1504, "lr": 5.1199939556492614e-05}
{"train_loss": 0.09807337075471878, "global_step": 133905, "epoch": 1504, "lr": 5.119935998677997e-05}
{"train_loss": 0.08302710205316544, "global_step": 133906, "epoch": 1504, "lr": 5.11987804169061e-05}
{"train_loss": 0.15340952575206757, "global_step": 133907, "epoch": 1504, "lr": 5.119820084687107e-05}
{"train_loss": 0.1563379019498825, "global_step": 133908, "epoch": 1504, "lr": 5.119762127667495e-05}
{"train_loss": 0.11875617504119873, "global_step": 133909, "epoch": 1504, "lr": 5.119704170631783e-05}
{"train_loss": 0.1395418792963028, "global_step": 133910, "epoch": 1504, "lr": 5.119646213579977e-05}
{"train_loss": 0.10784321278333664, "global_step": 133911, "epoch": 1504, "lr": 5.1195882565120866e-05}
{"train_loss": 0.1447371393442154, "global_step": 133912, "epoch": 1504, "lr": 5.119530299428119e-05}
{"train_loss": 0.11608666181564331, "global_step": 133913, "epoch": 1504, "lr": 5.1194723423280825e-05}
{"train_loss": 0.10556462407112122, "global_step": 133914, "epoch": 1504, "lr": 5.119414385211984e-05}
{"train_loss": 0.13991756737232208, "global_step": 133915, "epoch": 1504, "lr": 5.119356428079831e-05}
{"train_loss": 0.08790180087089539, "global_step": 133916, "epoch": 1504, "lr": 5.1192984709316326e-05}
{"train_loss": 0.11356546729803085, "global_step": 133917, "epoch": 1504, "lr": 5.119240513767396e-05}
{"train_loss": 0.1298505812883377, "global_step": 133918, "epoch": 1504, "lr": 5.1191825565871286e-05}
{"train_loss": 0.11121766269207001, "global_step": 133919, "epoch": 1504, "lr": 5.11912459939084e-05}
{"train_loss": 0.07936133444309235, "global_step": 133920, "epoch": 1504, "lr": 5.1190666421785346e-05}
{"train_loss": 0.08388525247573853, "global_step": 133921, "epoch": 1504, "lr": 5.1190086849502236e-05}
{"train_loss": 0.1294230818748474, "global_step": 133922, "epoch": 1504, "lr": 5.118950727705913e-05}
{"train_loss": 0.0884089320898056, "global_step": 133923, "epoch": 1504, "lr": 5.1188927704456114e-05}
{"train_loss": 0.14961916208267212, "global_step": 133924, "epoch": 1504, "lr": 5.1188348131693264e-05}
{"train_loss": 0.08681754022836685, "global_step": 133925, "epoch": 1504, "lr": 5.1187768558770644e-05}
{"train_loss": 0.10223618149757385, "global_step": 133926, "epoch": 1504, "lr": 5.118718898568835e-05}
{"train_loss": 0.10517691820859909, "global_step": 133927, "epoch": 1504, "lr": 5.118660941244644e-05}
{"train_loss": 0.19553923606872559, "global_step": 133928, "epoch": 1504, "lr": 5.1186029839045013e-05}
{"train_loss": 0.11072792112827301, "global_step": 133929, "epoch": 1504, "lr": 5.118545026548415e-05}
{"train_loss": 0.15319743752479553, "global_step": 133930, "epoch": 1504, "lr": 5.1184870691763896e-05}
{"train_loss": 0.12192636728286743, "global_step": 133931, "epoch": 1504, "lr": 5.1184291117884366e-05}
{"train_loss": 0.18129926919937134, "global_step": 133932, "epoch": 1504, "lr": 5.1183711543845614e-05}
{"train_loss": 0.13143591582775116, "global_step": 133933, "epoch": 1504, "lr": 5.118313196964772e-05}
{"train_loss": 0.09957899153232574, "global_step": 133934, "epoch": 1504, "lr": 5.1182552395290783e-05}
{"train_loss": 0.15884451568126678, "global_step": 133935, "epoch": 1504, "lr": 5.118197282077486e-05}
{"train_loss": 0.17143301665782928, "global_step": 133936, "epoch": 1504, "lr": 5.118139324610004e-05}
{"train_loss": 0.10188080370426178, "global_step": 133937, "epoch": 1504, "lr": 5.118081367126639e-05}
{"train_loss": 0.11037859320640564, "global_step": 133938, "epoch": 1504, "lr": 5.1180234096273997e-05}
{"train_loss": 0.08177896589040756, "global_step": 133939, "epoch": 1504, "lr": 5.1179654521122934e-05}
{"train_loss": 0.12459997832775116, "global_step": 133940, "epoch": 1504, "lr": 5.117907494581329e-05}
{"train_loss": 0.07317700982093811, "global_step": 133941, "epoch": 1504, "lr": 5.1178495370345116e-05}
{"train_loss": 0.10115289688110352, "global_step": 133942, "epoch": 1504, "lr": 5.117791579471852e-05}
{"train_loss": 0.14010003209114075, "global_step": 133943, "epoch": 1504, "lr": 5.117733621893358e-05}
{"train_loss": 0.11774181329718467, "global_step": 133944, "epoch": 1504, "lr": 5.117675664299034e-05, "val_loss": 5.81926965713501}
{"train_loss": 0.16761183738708496, "global_step": 133945, "epoch": 1505, "lr": 5.117617706688891e-05}
{"train_loss": 0.1788356900215149, "global_step": 133946, "epoch": 1505, "lr": 5.117559749062936e-05}
{"train_loss": 0.16752110421657562, "global_step": 133947, "epoch": 1505, "lr": 5.117501791421175e-05}
{"train_loss": 0.20936980843544006, "global_step": 133948, "epoch": 1505, "lr": 5.117443833763619e-05}
{"train_loss": 0.14828990399837494, "global_step": 133949, "epoch": 1505, "lr": 5.117385876090273e-05}
{"train_loss": 0.15012520551681519, "global_step": 133950, "epoch": 1505, "lr": 5.117327918401146e-05}
{"train_loss": 0.17284128069877625, "global_step": 133951, "epoch": 1505, "lr": 5.117269960696246e-05}
{"train_loss": 0.07486651837825775, "global_step": 133952, "epoch": 1505, "lr": 5.11721200297558e-05}
{"train_loss": 0.1038060411810875, "global_step": 133953, "epoch": 1505, "lr": 5.117154045239156e-05}
{"train_loss": 0.11454588919878006, "global_step": 133954, "epoch": 1505, "lr": 5.117096087486983e-05}
{"train_loss": 0.1362256407737732, "global_step": 133955, "epoch": 1505, "lr": 5.117038129719067e-05}
{"train_loss": 0.1501491665840149, "global_step": 133956, "epoch": 1505, "lr": 5.1169801719354184e-05}
{"train_loss": 0.07125331461429596, "global_step": 133957, "epoch": 1505, "lr": 5.116922214136042e-05}
{"train_loss": 0.11001907289028168, "global_step": 133958, "epoch": 1505, "lr": 5.116864256320947e-05}
{"train_loss": 0.1622966229915619, "global_step": 133959, "epoch": 1505, "lr": 5.11680629849014e-05}
{"train_loss": 0.16777141392230988, "global_step": 133960, "epoch": 1505, "lr": 5.116748340643631e-05}
{"train_loss": 0.09763166308403015, "global_step": 133961, "epoch": 1505, "lr": 5.116690382781426e-05}
{"train_loss": 0.11556151509284973, "global_step": 133962, "epoch": 1505, "lr": 5.1166324249035355e-05}
{"train_loss": 0.12750278413295746, "global_step": 133963, "epoch": 1505, "lr": 5.1165744670099634e-05}
{"train_loss": 0.1436159759759903, "global_step": 133964, "epoch": 1505, "lr": 5.116516509100719e-05}
{"train_loss": 0.1640430986881256, "global_step": 133965, "epoch": 1505, "lr": 5.116458551175811e-05}
{"train_loss": 0.1480100303888321, "global_step": 133966, "epoch": 1505, "lr": 5.116400593235247e-05}
{"train_loss": 0.08433622121810913, "global_step": 133967, "epoch": 1505, "lr": 5.116342635279033e-05}
{"train_loss": 0.10094004124403, "global_step": 133968, "epoch": 1505, "lr": 5.1162846773071795e-05}
{"train_loss": 0.2130117267370224, "global_step": 133969, "epoch": 1505, "lr": 5.116226719319691e-05}
{"train_loss": 0.10755237191915512, "global_step": 133970, "epoch": 1505, "lr": 5.1161687613165774e-05}
{"train_loss": 0.1404769867658615, "global_step": 133971, "epoch": 1505, "lr": 5.116110803297848e-05}
{"train_loss": 0.17506057024002075, "global_step": 133972, "epoch": 1505, "lr": 5.1160528452635084e-05}
{"train_loss": 0.20009371638298035, "global_step": 133973, "epoch": 1505, "lr": 5.1159948872135666e-05}
{"train_loss": 0.09122289717197418, "global_step": 133974, "epoch": 1505, "lr": 5.1159369291480306e-05}
{"train_loss": 0.09908919781446457, "global_step": 133975, "epoch": 1505, "lr": 5.1158789710669085e-05}
{"train_loss": 0.12767109274864197, "global_step": 133976, "epoch": 1505, "lr": 5.115821012970209e-05}
{"train_loss": 0.10835517942905426, "global_step": 133977, "epoch": 1505, "lr": 5.115763054857937e-05}
{"train_loss": 0.10431250929832458, "global_step": 133978, "epoch": 1505, "lr": 5.1157050967301036e-05}
{"train_loss": 0.13961397111415863, "global_step": 133979, "epoch": 1505, "lr": 5.1156471385867144e-05}
{"train_loss": 0.09161670506000519, "global_step": 133980, "epoch": 1505, "lr": 5.115589180427778e-05}
{"train_loss": 0.07299094647169113, "global_step": 133981, "epoch": 1505, "lr": 5.115531222253301e-05}
{"train_loss": 0.07763370871543884, "global_step": 133982, "epoch": 1505, "lr": 5.115473264063294e-05}
{"train_loss": 0.22619451582431793, "global_step": 133983, "epoch": 1505, "lr": 5.1154153058577624e-05}
{"train_loss": 0.1508234441280365, "global_step": 133984, "epoch": 1505, "lr": 5.1153573476367146e-05}
{"train_loss": 0.1138422042131424, "global_step": 133985, "epoch": 1505, "lr": 5.115299389400158e-05}
{"train_loss": 0.13773414492607117, "global_step": 133986, "epoch": 1505, "lr": 5.115241431148101e-05}
{"train_loss": 0.10747542977333069, "global_step": 133987, "epoch": 1505, "lr": 5.1151834728805505e-05}
{"train_loss": 0.08152761310338974, "global_step": 133988, "epoch": 1505, "lr": 5.1151255145975154e-05}
{"train_loss": 0.11588052660226822, "global_step": 133989, "epoch": 1505, "lr": 5.1150675562990026e-05}
{"train_loss": 0.10198965668678284, "global_step": 133990, "epoch": 1505, "lr": 5.115009597985021e-05}
{"train_loss": 0.10781426727771759, "global_step": 133991, "epoch": 1505, "lr": 5.114951639655577e-05}
{"train_loss": 0.10378118604421616, "global_step": 133992, "epoch": 1505, "lr": 5.1148936813106804e-05}
{"train_loss": 0.054561104625463486, "global_step": 133993, "epoch": 1505, "lr": 5.1148357229503376e-05}
{"train_loss": 0.12419683486223221, "global_step": 133994, "epoch": 1505, "lr": 5.1147777645745565e-05}
{"train_loss": 0.14372582733631134, "global_step": 133995, "epoch": 1505, "lr": 5.1147198061833444e-05}
{"train_loss": 0.13384604454040527, "global_step": 133996, "epoch": 1505, "lr": 5.11466184777671e-05}
{"train_loss": 0.13588836789131165, "global_step": 133997, "epoch": 1505, "lr": 5.114603889354661e-05}
{"train_loss": 0.09143713116645813, "global_step": 133998, "epoch": 1505, "lr": 5.114545930917204e-05}
{"train_loss": 0.10849808901548386, "global_step": 133999, "epoch": 1505, "lr": 5.114487972464348e-05}
{"train_loss": 0.09889069944620132, "global_step": 134000, "epoch": 1505, "lr": 5.114430013996101e-05}
{"train_loss": 0.08404684066772461, "global_step": 134001, "epoch": 1505, "lr": 5.114372055512471e-05}
{"train_loss": 0.12386152148246765, "global_step": 134002, "epoch": 1505, "lr": 5.114314097013464e-05}
{"train_loss": 0.08673038333654404, "global_step": 134003, "epoch": 1505, "lr": 5.1142561384990895e-05}
{"train_loss": 0.09092220664024353, "global_step": 134004, "epoch": 1505, "lr": 5.114198179969354e-05}
{"train_loss": 0.09020698815584183, "global_step": 134005, "epoch": 1505, "lr": 5.114140221424265e-05}
{"train_loss": 0.16428698599338531, "global_step": 134006, "epoch": 1505, "lr": 5.114082262863832e-05}
{"train_loss": 0.13912518322467804, "global_step": 134007, "epoch": 1505, "lr": 5.114024304288062e-05}
{"train_loss": 0.19435690343379974, "global_step": 134008, "epoch": 1505, "lr": 5.113966345696963e-05}
{"train_loss": 0.10344208776950836, "global_step": 134009, "epoch": 1505, "lr": 5.1139083870905426e-05}
{"train_loss": 0.0709996446967125, "global_step": 134010, "epoch": 1505, "lr": 5.1138504284688095e-05}
{"train_loss": 0.09425559639930725, "global_step": 134011, "epoch": 1505, "lr": 5.1137924698317695e-05}
{"train_loss": 0.09319593012332916, "global_step": 134012, "epoch": 1505, "lr": 5.113734511179431e-05}
{"train_loss": 0.11548459529876709, "global_step": 134013, "epoch": 1505, "lr": 5.1136765525118034e-05}
{"train_loss": 0.07971550524234772, "global_step": 134014, "epoch": 1505, "lr": 5.113618593828893e-05}
{"train_loss": 0.1921006143093109, "global_step": 134015, "epoch": 1505, "lr": 5.1135606351307086e-05}
{"train_loss": 0.08312766999006271, "global_step": 134016, "epoch": 1505, "lr": 5.113502676417257e-05}
{"train_loss": 0.12047398090362549, "global_step": 134017, "epoch": 1505, "lr": 5.113444717688546e-05}
{"train_loss": 0.1487131416797638, "global_step": 134018, "epoch": 1505, "lr": 5.113386758944584e-05}
{"train_loss": 0.16074888408184052, "global_step": 134019, "epoch": 1505, "lr": 5.113328800185378e-05}
{"train_loss": 0.10097769647836685, "global_step": 134020, "epoch": 1505, "lr": 5.1132708414109374e-05}
{"train_loss": 0.1070009097456932, "global_step": 134021, "epoch": 1505, "lr": 5.113212882621269e-05}
{"train_loss": 0.10683075338602066, "global_step": 134022, "epoch": 1505, "lr": 5.113154923816379e-05}
{"train_loss": 0.11085353791713715, "global_step": 134023, "epoch": 1505, "lr": 5.113096964996278e-05}
{"train_loss": 0.071498803794384, "global_step": 134024, "epoch": 1505, "lr": 5.11303900616097e-05}
{"train_loss": 0.09494763612747192, "global_step": 134025, "epoch": 1505, "lr": 5.1129810473104676e-05}
{"train_loss": 0.07391063868999481, "global_step": 134026, "epoch": 1505, "lr": 5.1129230884447755e-05}
{"train_loss": 0.1411639004945755, "global_step": 134027, "epoch": 1505, "lr": 5.112865129563903e-05}
{"train_loss": 0.08778085559606552, "global_step": 134028, "epoch": 1505, "lr": 5.112807170667857e-05}
{"train_loss": 0.08994337171316147, "global_step": 134029, "epoch": 1505, "lr": 5.112749211756646e-05}
{"train_loss": 0.1592835634946823, "global_step": 134030, "epoch": 1505, "lr": 5.112691252830276e-05}
{"train_loss": 0.1436142921447754, "global_step": 134031, "epoch": 1505, "lr": 5.1126332938887565e-05}
{"train_loss": 0.15027976036071777, "global_step": 134032, "epoch": 1505, "lr": 5.112575334932095e-05}
{"train_loss": 0.12352728588360079, "global_step": 134033, "epoch": 1505, "lr": 5.112517375960298e-05, "val_loss": 5.776339054107666, "train_action_mse_error": 17.141145706176758}
{"train_loss": 0.12705789506435394, "global_step": 134034, "epoch": 1506, "lr": 5.1124594169733766e-05}
{"train_loss": 0.16474655270576477, "global_step": 134035, "epoch": 1506, "lr": 5.112401457971334e-05}
{"train_loss": 0.16665105521678925, "global_step": 134036, "epoch": 1506, "lr": 5.112343498954182e-05}
{"train_loss": 0.08533510565757751, "global_step": 134037, "epoch": 1506, "lr": 5.112285539921926e-05}
{"train_loss": 0.07550748437643051, "global_step": 134038, "epoch": 1506, "lr": 5.112227580874576e-05}
{"train_loss": 0.1250806599855423, "global_step": 134039, "epoch": 1506, "lr": 5.1121696218121377e-05}
{"train_loss": 0.12133552134037018, "global_step": 134040, "epoch": 1506, "lr": 5.11211166273462e-05}
{"train_loss": 0.1154896691441536, "global_step": 134041, "epoch": 1506, "lr": 5.112053703642029e-05}
{"train_loss": 0.1400754153728485, "global_step": 134042, "epoch": 1506, "lr": 5.111995744534373e-05}
{"train_loss": 0.1352381855249405, "global_step": 134043, "epoch": 1506, "lr": 5.111937785411661e-05}
{"train_loss": 0.10024689137935638, "global_step": 134044, "epoch": 1506, "lr": 5.111879826273902e-05}
{"train_loss": 0.09319622069597244, "global_step": 134045, "epoch": 1506, "lr": 5.1118218671211e-05}
{"train_loss": 0.14158228039741516, "global_step": 134046, "epoch": 1506, "lr": 5.111763907953267e-05}
{"train_loss": 0.1090688556432724, "global_step": 134047, "epoch": 1506, "lr": 5.1117059487704075e-05}
{"train_loss": 0.15120476484298706, "global_step": 134048, "epoch": 1506, "lr": 5.1116479895725314e-05}
{"train_loss": 0.19237542152404785, "global_step": 134049, "epoch": 1506, "lr": 5.1115900303596445e-05}
{"train_loss": 0.13010984659194946, "global_step": 134050, "epoch": 1506, "lr": 5.111532071131756e-05}
{"train_loss": 0.10155365616083145, "global_step": 134051, "epoch": 1506, "lr": 5.1114741118888744e-05}
{"train_loss": 0.17579160630702972, "global_step": 134052, "epoch": 1506, "lr": 5.111416152631005e-05}
{"train_loss": 0.09680810570716858, "global_step": 134053, "epoch": 1506, "lr": 5.111358193358158e-05}
{"train_loss": 0.11504494398832321, "global_step": 134054, "epoch": 1506, "lr": 5.11130023407034e-05}
{"train_loss": 0.09650852531194687, "global_step": 134055, "epoch": 1506, "lr": 5.111242274767558e-05}
{"train_loss": 0.06912875175476074, "global_step": 134056, "epoch": 1506, "lr": 5.111184315449822e-05}
{"train_loss": 0.08236081153154373, "global_step": 134057, "epoch": 1506, "lr": 5.111126356117139e-05}
{"train_loss": 0.11364126950502396, "global_step": 134058, "epoch": 1506, "lr": 5.111068396769515e-05}
{"train_loss": 0.17779237031936646, "global_step": 134059, "epoch": 1506, "lr": 5.111010437406961e-05}
{"train_loss": 0.08050400018692017, "global_step": 134060, "epoch": 1506, "lr": 5.1109524780294816e-05}
{"train_loss": 0.1243029311299324, "global_step": 134061, "epoch": 1506, "lr": 5.110894518637086e-05}
{"train_loss": 0.12219465523958206, "global_step": 134062, "epoch": 1506, "lr": 5.110836559229782e-05}
{"train_loss": 0.1252584606409073, "global_step": 134063, "epoch": 1506, "lr": 5.1107785998075786e-05}
{"train_loss": 0.09250607341527939, "global_step": 134064, "epoch": 1506, "lr": 5.110720640370481e-05}
{"train_loss": 0.031440213322639465, "global_step": 134065, "epoch": 1506, "lr": 5.1106626809185e-05}
{"train_loss": 0.10768242180347443, "global_step": 134066, "epoch": 1506, "lr": 5.1106047214516396e-05}
{"train_loss": 0.11955301463603973, "global_step": 134067, "epoch": 1506, "lr": 5.110546761969911e-05}
{"train_loss": 0.13523735105991364, "global_step": 134068, "epoch": 1506, "lr": 5.110488802473321e-05}
{"train_loss": 0.1398555338382721, "global_step": 134069, "epoch": 1506, "lr": 5.110430842961876e-05}
{"train_loss": 0.06975255161523819, "global_step": 134070, "epoch": 1506, "lr": 5.110372883435586e-05}
{"train_loss": 0.17458367347717285, "global_step": 134071, "epoch": 1506, "lr": 5.110314923894457e-05}
{"train_loss": 0.1043350026011467, "global_step": 134072, "epoch": 1506, "lr": 5.110256964338498e-05}
{"train_loss": 0.10867176204919815, "global_step": 134073, "epoch": 1506, "lr": 5.110199004767716e-05}
{"train_loss": 0.10885602980852127, "global_step": 134074, "epoch": 1506, "lr": 5.11014104518212e-05}
{"train_loss": 0.12871825695037842, "global_step": 134075, "epoch": 1506, "lr": 5.110083085581715e-05}
{"train_loss": 0.05312476307153702, "global_step": 134076, "epoch": 1506, "lr": 5.110025125966512e-05}
{"train_loss": 0.07797995954751968, "global_step": 134077, "epoch": 1506, "lr": 5.1099671663365164e-05}
{"train_loss": 0.13458135724067688, "global_step": 134078, "epoch": 1506, "lr": 5.109909206691739e-05}
{"train_loss": 0.08546929806470871, "global_step": 134079, "epoch": 1506, "lr": 5.1098512470321846e-05}
{"train_loss": 0.07623948156833649, "global_step": 134080, "epoch": 1506, "lr": 5.109793287357861e-05}
{"train_loss": 0.0837603360414505, "global_step": 134081, "epoch": 1506, "lr": 5.1097353276687786e-05}
{"train_loss": 0.10400176048278809, "global_step": 134082, "epoch": 1506, "lr": 5.1096773679649426e-05}
{"train_loss": 0.15585605800151825, "global_step": 134083, "epoch": 1506, "lr": 5.109619408246362e-05}
{"train_loss": 0.08998946845531464, "global_step": 134084, "epoch": 1506, "lr": 5.109561448513045e-05}
{"train_loss": 0.09365132451057434, "global_step": 134085, "epoch": 1506, "lr": 5.1095034887649984e-05}
{"train_loss": 0.12090843170881271, "global_step": 134086, "epoch": 1506, "lr": 5.1094455290022306e-05}
{"train_loss": 0.12724436819553375, "global_step": 134087, "epoch": 1506, "lr": 5.109387569224748e-05}
{"train_loss": 0.11764751374721527, "global_step": 134088, "epoch": 1506, "lr": 5.109329609432561e-05}
{"train_loss": 0.06715147197246552, "global_step": 134089, "epoch": 1506, "lr": 5.109271649625675e-05}
{"train_loss": 0.11252598464488983, "global_step": 134090, "epoch": 1506, "lr": 5.1092136898041e-05}
{"train_loss": 0.06731396168470383, "global_step": 134091, "epoch": 1506, "lr": 5.109155729967841e-05}
{"train_loss": 0.17151883244514465, "global_step": 134092, "epoch": 1506, "lr": 5.109097770116908e-05}
{"train_loss": 0.12008389085531235, "global_step": 134093, "epoch": 1506, "lr": 5.109039810251309e-05}
{"train_loss": 0.08780056983232498, "global_step": 134094, "epoch": 1506, "lr": 5.1089818503710506e-05}
{"train_loss": 0.12079881131649017, "global_step": 134095, "epoch": 1506, "lr": 5.10892389047614e-05}
{"train_loss": 0.08543965220451355, "global_step": 134096, "epoch": 1506, "lr": 5.1088659305665865e-05}
{"train_loss": 0.08162233233451843, "global_step": 134097, "epoch": 1506, "lr": 5.108807970642397e-05}
{"train_loss": 0.1487838476896286, "global_step": 134098, "epoch": 1506, "lr": 5.10875001070358e-05}
{"train_loss": 0.2487584799528122, "global_step": 134099, "epoch": 1506, "lr": 5.108692050750143e-05}
{"train_loss": 0.1367373764514923, "global_step": 134100, "epoch": 1506, "lr": 5.108634090782094e-05}
{"train_loss": 0.10302923619747162, "global_step": 134101, "epoch": 1506, "lr": 5.10857613079944e-05}
{"train_loss": 0.1283668428659439, "global_step": 134102, "epoch": 1506, "lr": 5.108518170802189e-05}
{"train_loss": 0.0821104645729065, "global_step": 134103, "epoch": 1506, "lr": 5.108460210790349e-05}
{"train_loss": 0.11967839300632477, "global_step": 134104, "epoch": 1506, "lr": 5.108402250763929e-05}
{"train_loss": 0.125532865524292, "global_step": 134105, "epoch": 1506, "lr": 5.108344290722934e-05}
{"train_loss": 0.08149656653404236, "global_step": 134106, "epoch": 1506, "lr": 5.108286330667373e-05}
{"train_loss": 0.09721554070711136, "global_step": 134107, "epoch": 1506, "lr": 5.1082283705972557e-05}
{"train_loss": 0.10961879044771194, "global_step": 134108, "epoch": 1506, "lr": 5.108170410512588e-05}
{"train_loss": 0.11510248482227325, "global_step": 134109, "epoch": 1506, "lr": 5.108112450413378e-05}
{"train_loss": 0.12169384956359863, "global_step": 134110, "epoch": 1506, "lr": 5.108054490299634e-05}
{"train_loss": 0.09602401405572891, "global_step": 134111, "epoch": 1506, "lr": 5.107996530171364e-05}
{"train_loss": 0.17719203233718872, "global_step": 134112, "epoch": 1506, "lr": 5.107938570028574e-05}
{"train_loss": 0.15705525875091553, "global_step": 134113, "epoch": 1506, "lr": 5.107880609871273e-05}
{"train_loss": 0.21046984195709229, "global_step": 134114, "epoch": 1506, "lr": 5.10782264969947e-05}
{"train_loss": 0.10262014716863632, "global_step": 134115, "epoch": 1506, "lr": 5.1077646895131706e-05}
{"train_loss": 0.15047797560691833, "global_step": 134116, "epoch": 1506, "lr": 5.1077067293123824e-05}
{"train_loss": 0.06517040729522705, "global_step": 134117, "epoch": 1506, "lr": 5.107648769097116e-05}
{"train_loss": 0.1449040174484253, "global_step": 134118, "epoch": 1506, "lr": 5.107590808867378e-05}
{"train_loss": 0.07144039869308472, "global_step": 134119, "epoch": 1506, "lr": 5.107532848623175e-05}
{"train_loss": 0.10756159573793411, "global_step": 134120, "epoch": 1506, "lr": 5.107474888364514e-05}
{"train_loss": 0.14561676979064941, "global_step": 134121, "epoch": 1506, "lr": 5.1074169280914054e-05}
{"train_loss": 0.11623788039001186, "global_step": 134122, "epoch": 1506, "lr": 5.1073589678038566e-05, "val_loss": 5.892007350921631}
{"train_loss": 0.09180422127246857, "global_step": 134123, "epoch": 1507, "lr": 5.1073010075018745e-05}
{"train_loss": 0.15224029123783112, "global_step": 134124, "epoch": 1507, "lr": 5.107243047185466e-05}
{"train_loss": 0.1565852165222168, "global_step": 134125, "epoch": 1507, "lr": 5.107185086854641e-05}
{"train_loss": 0.15746192634105682, "global_step": 134126, "epoch": 1507, "lr": 5.1071271265094066e-05}
{"train_loss": 0.07789290696382523, "global_step": 134127, "epoch": 1507, "lr": 5.107069166149769e-05}
{"train_loss": 0.13658228516578674, "global_step": 134128, "epoch": 1507, "lr": 5.107011205775738e-05}
{"train_loss": 0.1177535206079483, "global_step": 134129, "epoch": 1507, "lr": 5.106953245387322e-05}
{"train_loss": 0.11850545555353165, "global_step": 134130, "epoch": 1507, "lr": 5.106895284984525e-05}
{"train_loss": 0.07627389580011368, "global_step": 134131, "epoch": 1507, "lr": 5.1068373245673585e-05}
{"train_loss": 0.15344588458538055, "global_step": 134132, "epoch": 1507, "lr": 5.1067793641358284e-05}
{"train_loss": 0.169822558760643, "global_step": 134133, "epoch": 1507, "lr": 5.1067214036899435e-05}
{"train_loss": 0.10099752247333527, "global_step": 134134, "epoch": 1507, "lr": 5.106663443229711e-05}
{"train_loss": 0.1079375296831131, "global_step": 134135, "epoch": 1507, "lr": 5.10660548275514e-05}
{"train_loss": 0.12429007887840271, "global_step": 134136, "epoch": 1507, "lr": 5.106547522266236e-05}
{"train_loss": 0.12385229766368866, "global_step": 134137, "epoch": 1507, "lr": 5.1064895617630085e-05}
{"train_loss": 0.15053193271160126, "global_step": 134138, "epoch": 1507, "lr": 5.106431601245465e-05}
{"train_loss": 0.08701781928539276, "global_step": 134139, "epoch": 1507, "lr": 5.106373640713612e-05}
{"train_loss": 0.10538697242736816, "global_step": 134140, "epoch": 1507, "lr": 5.106315680167459e-05}
{"train_loss": 0.18287838995456696, "global_step": 134141, "epoch": 1507, "lr": 5.106257719607014e-05}
{"train_loss": 0.11008619517087936, "global_step": 134142, "epoch": 1507, "lr": 5.106199759032283e-05}
{"train_loss": 0.12250040471553802, "global_step": 134143, "epoch": 1507, "lr": 5.106141798443275e-05}
{"train_loss": 0.1024012342095375, "global_step": 134144, "epoch": 1507, "lr": 5.106083837839997e-05}
{"train_loss": 0.13815546035766602, "global_step": 134145, "epoch": 1507, "lr": 5.106025877222458e-05}
{"train_loss": 0.11311055719852448, "global_step": 134146, "epoch": 1507, "lr": 5.105967916590664e-05}
{"train_loss": 0.14771787822246552, "global_step": 134147, "epoch": 1507, "lr": 5.105909955944625e-05}
{"train_loss": 0.1280677765607834, "global_step": 134148, "epoch": 1507, "lr": 5.1058519952843474e-05}
{"train_loss": 0.10918910056352615, "global_step": 134149, "epoch": 1507, "lr": 5.105794034609839e-05}
{"train_loss": 0.08029469847679138, "global_step": 134150, "epoch": 1507, "lr": 5.105736073921108e-05}
{"train_loss": 0.090059295296669, "global_step": 134151, "epoch": 1507, "lr": 5.105678113218163e-05}
{"train_loss": 0.15678443014621735, "global_step": 134152, "epoch": 1507, "lr": 5.105620152501009e-05}
{"train_loss": 0.13231511414051056, "global_step": 134153, "epoch": 1507, "lr": 5.105562191769657e-05}
{"train_loss": 0.1705511510372162, "global_step": 134154, "epoch": 1507, "lr": 5.1055042310241145e-05}
{"train_loss": 0.06455682963132858, "global_step": 134155, "epoch": 1507, "lr": 5.105446270264387e-05}
{"train_loss": 0.12321025133132935, "global_step": 134156, "epoch": 1507, "lr": 5.1053883094904834e-05}
{"train_loss": 0.11693837493658066, "global_step": 134157, "epoch": 1507, "lr": 5.105330348702412e-05}
{"train_loss": 0.08291498571634293, "global_step": 134158, "epoch": 1507, "lr": 5.1052723879001796e-05}
{"train_loss": 0.17107932269573212, "global_step": 134159, "epoch": 1507, "lr": 5.1052144270837944e-05}
{"train_loss": 0.07042068243026733, "global_step": 134160, "epoch": 1507, "lr": 5.105156466253265e-05}
{"train_loss": 0.1213696151971817, "global_step": 134161, "epoch": 1507, "lr": 5.105098505408599e-05}
{"train_loss": 0.12984150648117065, "global_step": 134162, "epoch": 1507, "lr": 5.105040544549803e-05}
{"train_loss": 0.1014716625213623, "global_step": 134163, "epoch": 1507, "lr": 5.104982583676886e-05}
{"train_loss": 0.10057735443115234, "global_step": 134164, "epoch": 1507, "lr": 5.104924622789855e-05}
{"train_loss": 0.0932379961013794, "global_step": 134165, "epoch": 1507, "lr": 5.104866661888719e-05}
{"train_loss": 0.14719252288341522, "global_step": 134166, "epoch": 1507, "lr": 5.104808700973485e-05}
{"train_loss": 0.10247434675693512, "global_step": 134167, "epoch": 1507, "lr": 5.1047507400441597e-05}
{"train_loss": 0.12057799100875854, "global_step": 134168, "epoch": 1507, "lr": 5.1046927791007526e-05}
{"train_loss": 0.15664292871952057, "global_step": 134169, "epoch": 1507, "lr": 5.104634818143269e-05}
{"train_loss": 0.12294982373714447, "global_step": 134170, "epoch": 1507, "lr": 5.104576857171721e-05}
{"train_loss": 0.14949004352092743, "global_step": 134171, "epoch": 1507, "lr": 5.1045188961861136e-05}
{"train_loss": 0.1032753735780716, "global_step": 134172, "epoch": 1507, "lr": 5.104460935186455e-05}
{"train_loss": 0.10123763978481293, "global_step": 134173, "epoch": 1507, "lr": 5.1044029741727515e-05}
{"train_loss": 0.11399673670530319, "global_step": 134174, "epoch": 1507, "lr": 5.104345013145013e-05}
{"train_loss": 0.10492359101772308, "global_step": 134175, "epoch": 1507, "lr": 5.104287052103247e-05}
{"train_loss": 0.14670871198177338, "global_step": 134176, "epoch": 1507, "lr": 5.10422909104746e-05}
{"train_loss": 0.08202867954969406, "global_step": 134177, "epoch": 1507, "lr": 5.1041711299776615e-05}
{"train_loss": 0.0830378457903862, "global_step": 134178, "epoch": 1507, "lr": 5.1041131688938584e-05}
{"train_loss": 0.13239842653274536, "global_step": 134179, "epoch": 1507, "lr": 5.1040552077960576e-05}
{"train_loss": 0.10987923294305801, "global_step": 134180, "epoch": 1507, "lr": 5.103997246684269e-05}
{"train_loss": 0.11113295704126358, "global_step": 134181, "epoch": 1507, "lr": 5.103939285558499e-05}
{"train_loss": 0.16446788609027863, "global_step": 134182, "epoch": 1507, "lr": 5.1038813244187556e-05}
{"train_loss": 0.05502105876803398, "global_step": 134183, "epoch": 1507, "lr": 5.1038233632650466e-05}
{"train_loss": 0.1469757854938507, "global_step": 134184, "epoch": 1507, "lr": 5.103765402097379e-05}
{"train_loss": 0.10462352633476257, "global_step": 134185, "epoch": 1507, "lr": 5.103707440915763e-05}
{"train_loss": 0.21443626284599304, "global_step": 134186, "epoch": 1507, "lr": 5.103649479720204e-05}
{"train_loss": 0.1656413972377777, "global_step": 134187, "epoch": 1507, "lr": 5.103591518510709e-05}
{"train_loss": 0.13397423923015594, "global_step": 134188, "epoch": 1507, "lr": 5.103533557287289e-05}
{"train_loss": 0.11115779727697372, "global_step": 134189, "epoch": 1507, "lr": 5.10347559604995e-05}
{"train_loss": 0.12022587656974792, "global_step": 134190, "epoch": 1507, "lr": 5.103417634798701e-05}
{"train_loss": 0.09544733911752701, "global_step": 134191, "epoch": 1507, "lr": 5.103359673533548e-05}
{"train_loss": 0.09098906815052032, "global_step": 134192, "epoch": 1507, "lr": 5.103301712254499e-05}
{"train_loss": 0.06114467233419418, "global_step": 134193, "epoch": 1507, "lr": 5.103243750961563e-05}
{"train_loss": 0.10630655288696289, "global_step": 134194, "epoch": 1507, "lr": 5.103185789654746e-05}
{"train_loss": 0.20361846685409546, "global_step": 134195, "epoch": 1507, "lr": 5.103127828334058e-05}
{"train_loss": 0.1866377592086792, "global_step": 134196, "epoch": 1507, "lr": 5.103069866999506e-05}
{"train_loss": 0.07329809665679932, "global_step": 134197, "epoch": 1507, "lr": 5.103011905651096e-05}
{"train_loss": 0.092926986515522, "global_step": 134198, "epoch": 1507, "lr": 5.1029539442888385e-05}
{"train_loss": 0.08594631403684616, "global_step": 134199, "epoch": 1507, "lr": 5.10289598291274e-05}
{"train_loss": 0.09493707865476608, "global_step": 134200, "epoch": 1507, "lr": 5.1028380215228076e-05}
{"train_loss": 0.12078554183244705, "global_step": 134201, "epoch": 1507, "lr": 5.102780060119051e-05}
{"train_loss": 0.11398905515670776, "global_step": 134202, "epoch": 1507, "lr": 5.102722098701477e-05}
{"train_loss": 0.08732763677835464, "global_step": 134203, "epoch": 1507, "lr": 5.102664137270092e-05}
{"train_loss": 0.14565618336200714, "global_step": 134204, "epoch": 1507, "lr": 5.102606175824906e-05}
{"train_loss": 0.09701154381036758, "global_step": 134205, "epoch": 1507, "lr": 5.102548214365925e-05}
{"train_loss": 0.1479886770248413, "global_step": 134206, "epoch": 1507, "lr": 5.102490252893157e-05}
{"train_loss": 0.14599177241325378, "global_step": 134207, "epoch": 1507, "lr": 5.1024322914066124e-05}
{"train_loss": 0.1715228259563446, "global_step": 134208, "epoch": 1507, "lr": 5.102374329906297e-05}
{"train_loss": 0.15182456374168396, "global_step": 134209, "epoch": 1507, "lr": 5.102316368392218e-05}
{"train_loss": 0.13257111608982086, "global_step": 134210, "epoch": 1507, "lr": 5.102258406864384e-05}
{"train_loss": 0.12088249016846164, "global_step": 134211, "epoch": 1507, "lr": 5.102200445322802e-05, "val_loss": 5.7284979820251465}
{"train_loss": 0.07074812799692154, "global_step": 134212, "epoch": 1508, "lr": 5.102142483767482e-05}
{"train_loss": 0.16456195712089539, "global_step": 134213, "epoch": 1508, "lr": 5.1020845221984284e-05}
{"train_loss": 0.09435582160949707, "global_step": 134214, "epoch": 1508, "lr": 5.1020265606156504e-05}
{"train_loss": 0.1565956473350525, "global_step": 134215, "epoch": 1508, "lr": 5.101968599019158e-05}
{"train_loss": 0.0777798667550087, "global_step": 134216, "epoch": 1508, "lr": 5.101910637408955e-05}
{"train_loss": 0.17784078419208527, "global_step": 134217, "epoch": 1508, "lr": 5.1018526757850526e-05}
{"train_loss": 0.12196707725524902, "global_step": 134218, "epoch": 1508, "lr": 5.1017947141474574e-05}
{"train_loss": 0.20104198157787323, "global_step": 134219, "epoch": 1508, "lr": 5.101736752496178e-05}
{"train_loss": 0.11580298095941544, "global_step": 134220, "epoch": 1508, "lr": 5.1016787908312203e-05}
{"train_loss": 0.14454351365566254, "global_step": 134221, "epoch": 1508, "lr": 5.101620829152592e-05}
{"train_loss": 0.13628333806991577, "global_step": 134222, "epoch": 1508, "lr": 5.101562867460304e-05}
{"train_loss": 0.1225360631942749, "global_step": 134223, "epoch": 1508, "lr": 5.1015049057543616e-05}
{"train_loss": 0.09924347698688507, "global_step": 134224, "epoch": 1508, "lr": 5.101446944034771e-05}
{"train_loss": 0.1155286580324173, "global_step": 134225, "epoch": 1508, "lr": 5.101388982301545e-05}
{"train_loss": 0.1929185688495636, "global_step": 134226, "epoch": 1508, "lr": 5.101331020554687e-05}
{"train_loss": 0.1185193732380867, "global_step": 134227, "epoch": 1508, "lr": 5.101273058794207e-05}
{"train_loss": 0.09480325877666473, "global_step": 134228, "epoch": 1508, "lr": 5.101215097020112e-05}
{"train_loss": 0.12180153280496597, "global_step": 134229, "epoch": 1508, "lr": 5.1011571352324105e-05}
{"train_loss": 0.23101069033145905, "global_step": 134230, "epoch": 1508, "lr": 5.101099173431109e-05}
{"train_loss": 0.18115434050559998, "global_step": 134231, "epoch": 1508, "lr": 5.1010412116162154e-05}
{"train_loss": 0.11706680059432983, "global_step": 134232, "epoch": 1508, "lr": 5.100983249787738e-05}
{"train_loss": 0.15380904078483582, "global_step": 134233, "epoch": 1508, "lr": 5.100925287945686e-05}
{"train_loss": 0.23229506611824036, "global_step": 134234, "epoch": 1508, "lr": 5.100867326090064e-05}
{"train_loss": 0.08359599113464355, "global_step": 134235, "epoch": 1508, "lr": 5.100809364220883e-05}
{"train_loss": 0.12234284728765488, "global_step": 134236, "epoch": 1508, "lr": 5.1007514023381485e-05}
{"train_loss": 0.047805044800043106, "global_step": 134237, "epoch": 1508, "lr": 5.1006934404418694e-05}
{"train_loss": 0.20121634006500244, "global_step": 134238, "epoch": 1508, "lr": 5.100635478532054e-05}
{"train_loss": 0.11164110153913498, "global_step": 134239, "epoch": 1508, "lr": 5.100577516608709e-05}
{"train_loss": 0.21342968940734863, "global_step": 134240, "epoch": 1508, "lr": 5.100519554671842e-05}
{"train_loss": 0.08095841109752655, "global_step": 134241, "epoch": 1508, "lr": 5.1004615927214615e-05}
{"train_loss": 0.23222529888153076, "global_step": 134242, "epoch": 1508, "lr": 5.100403630757574e-05}
{"train_loss": 0.13826073706150055, "global_step": 134243, "epoch": 1508, "lr": 5.100345668780191e-05}
{"train_loss": 0.05592489242553711, "global_step": 134244, "epoch": 1508, "lr": 5.100287706789316e-05}
{"train_loss": 0.13219967484474182, "global_step": 134245, "epoch": 1508, "lr": 5.10022974478496e-05}
{"train_loss": 0.09792278707027435, "global_step": 134246, "epoch": 1508, "lr": 5.100171782767128e-05}
{"train_loss": 0.08970321714878082, "global_step": 134247, "epoch": 1508, "lr": 5.100113820735831e-05}
{"train_loss": 0.09191443026065826, "global_step": 134248, "epoch": 1508, "lr": 5.100055858691074e-05}
{"train_loss": 0.1041335016489029, "global_step": 134249, "epoch": 1508, "lr": 5.099997896632864e-05}
{"train_loss": 0.16018491983413696, "global_step": 134250, "epoch": 1508, "lr": 5.099939934561213e-05}
{"train_loss": 0.16399608552455902, "global_step": 134251, "epoch": 1508, "lr": 5.099881972476125e-05}
{"train_loss": 0.12028352171182632, "global_step": 134252, "epoch": 1508, "lr": 5.099824010377608e-05}
{"train_loss": 0.10362780094146729, "global_step": 134253, "epoch": 1508, "lr": 5.0997660482656726e-05}
{"train_loss": 0.08514134585857391, "global_step": 134254, "epoch": 1508, "lr": 5.0997080861403246e-05}
{"train_loss": 0.10309222340583801, "global_step": 134255, "epoch": 1508, "lr": 5.099650124001572e-05}
{"train_loss": 0.16074474155902863, "global_step": 134256, "epoch": 1508, "lr": 5.0995921618494224e-05}
{"train_loss": 0.14179658889770508, "global_step": 134257, "epoch": 1508, "lr": 5.0995341996838844e-05}
{"train_loss": 0.08196771889925003, "global_step": 134258, "epoch": 1508, "lr": 5.0994762375049646e-05}
{"train_loss": 0.11810044944286346, "global_step": 134259, "epoch": 1508, "lr": 5.0994182753126705e-05}
{"train_loss": 0.15765459835529327, "global_step": 134260, "epoch": 1508, "lr": 5.099360313107011e-05}
{"train_loss": 0.14303050935268402, "global_step": 134261, "epoch": 1508, "lr": 5.099302350887995e-05}
{"train_loss": 0.11615538597106934, "global_step": 134262, "epoch": 1508, "lr": 5.0992443886556285e-05}
{"train_loss": 0.13848665356636047, "global_step": 134263, "epoch": 1508, "lr": 5.09918642640992e-05}
{"train_loss": 0.10878732055425644, "global_step": 134264, "epoch": 1508, "lr": 5.0991284641508775e-05}
{"train_loss": 0.10099884867668152, "global_step": 134265, "epoch": 1508, "lr": 5.0990705018785076e-05}
{"train_loss": 0.10262642800807953, "global_step": 134266, "epoch": 1508, "lr": 5.099012539592819e-05}
{"train_loss": 0.1736883670091629, "global_step": 134267, "epoch": 1508, "lr": 5.0989545772938205e-05}
{"train_loss": 0.12077741324901581, "global_step": 134268, "epoch": 1508, "lr": 5.0988966149815174e-05}
{"train_loss": 0.12850426137447357, "global_step": 134269, "epoch": 1508, "lr": 5.0988386526559193e-05}
{"train_loss": 0.14091618359088898, "global_step": 134270, "epoch": 1508, "lr": 5.0987806903170344e-05}
{"train_loss": 0.1345849335193634, "global_step": 134271, "epoch": 1508, "lr": 5.0987227279648686e-05}
{"train_loss": 0.15386022627353668, "global_step": 134272, "epoch": 1508, "lr": 5.0986647655994315e-05}
{"train_loss": 0.10113547742366791, "global_step": 134273, "epoch": 1508, "lr": 5.098606803220729e-05}
{"train_loss": 0.1547575145959854, "global_step": 134274, "epoch": 1508, "lr": 5.0985488408287704e-05}
{"train_loss": 0.11184481531381607, "global_step": 134275, "epoch": 1508, "lr": 5.098490878423563e-05}
{"train_loss": 0.1031312420964241, "global_step": 134276, "epoch": 1508, "lr": 5.098432916005115e-05}
{"train_loss": 0.06652364134788513, "global_step": 134277, "epoch": 1508, "lr": 5.098374953573434e-05}
{"train_loss": 0.13104650378227234, "global_step": 134278, "epoch": 1508, "lr": 5.098316991128527e-05}
{"train_loss": 0.11594130098819733, "global_step": 134279, "epoch": 1508, "lr": 5.0982590286704025e-05}
{"train_loss": 0.15498726069927216, "global_step": 134280, "epoch": 1508, "lr": 5.098201066199069e-05}
{"train_loss": 0.06678301095962524, "global_step": 134281, "epoch": 1508, "lr": 5.098143103714533e-05}
{"train_loss": 0.11429715156555176, "global_step": 134282, "epoch": 1508, "lr": 5.098085141216803e-05}
{"train_loss": 0.08665039390325546, "global_step": 134283, "epoch": 1508, "lr": 5.098027178705887e-05}
{"train_loss": 0.12590397894382477, "global_step": 134284, "epoch": 1508, "lr": 5.097969216181793e-05}
{"train_loss": 0.1336764395236969, "global_step": 134285, "epoch": 1508, "lr": 5.097911253644527e-05}
{"train_loss": 0.09892867505550385, "global_step": 134286, "epoch": 1508, "lr": 5.0978532910941e-05}
{"train_loss": 0.1339612603187561, "global_step": 134287, "epoch": 1508, "lr": 5.0977953285305156e-05}
{"train_loss": 0.08715472370386124, "global_step": 134288, "epoch": 1508, "lr": 5.097737365953785e-05}
{"train_loss": 0.10014892369508743, "global_step": 134289, "epoch": 1508, "lr": 5.097679403363914e-05}
{"train_loss": 0.13515746593475342, "global_step": 134290, "epoch": 1508, "lr": 5.097621440760913e-05}
{"train_loss": 0.11672193557024002, "global_step": 134291, "epoch": 1508, "lr": 5.097563478144786e-05}
{"train_loss": 0.20294074714183807, "global_step": 134292, "epoch": 1508, "lr": 5.0975055155155426e-05}
{"train_loss": 0.1328747570514679, "global_step": 134293, "epoch": 1508, "lr": 5.0974475528731914e-05}
{"train_loss": 0.17836648225784302, "global_step": 134294, "epoch": 1508, "lr": 5.0973895902177404e-05}
{"train_loss": 0.13394558429718018, "global_step": 134295, "epoch": 1508, "lr": 5.0973316275491956e-05}
{"train_loss": 0.16181375086307526, "global_step": 134296, "epoch": 1508, "lr": 5.0972736648675645e-05}
{"train_loss": 0.06999171525239944, "global_step": 134297, "epoch": 1508, "lr": 5.097215702172857e-05}
{"train_loss": 0.08776125311851501, "global_step": 134298, "epoch": 1508, "lr": 5.097157739465081e-05}
{"train_loss": 0.058770421892404556, "global_step": 134299, "epoch": 1508, "lr": 5.097099776744243e-05}
{"train_loss": 0.12635010129280305, "global_step": 134300, "epoch": 1508, "lr": 5.097041814010352e-05, "val_loss": 5.708576679229736}
{"train_loss": 0.19395749270915985, "global_step": 134301, "epoch": 1509, "lr": 5.0969838512634137e-05}
{"train_loss": 0.0751015692949295, "global_step": 134302, "epoch": 1509, "lr": 5.0969258885034374e-05}
{"train_loss": 0.08277297019958496, "global_step": 134303, "epoch": 1509, "lr": 5.096867925730431e-05}
{"train_loss": 0.15845747292041779, "global_step": 134304, "epoch": 1509, "lr": 5.096809962944401e-05}
{"train_loss": 0.06764581054449081, "global_step": 134305, "epoch": 1509, "lr": 5.0967520001453574e-05}
{"train_loss": 0.09074357151985168, "global_step": 134306, "epoch": 1509, "lr": 5.096694037333306e-05}
{"train_loss": 0.16409601271152496, "global_step": 134307, "epoch": 1509, "lr": 5.0966360745082555e-05}
{"train_loss": 0.11500803381204605, "global_step": 134308, "epoch": 1509, "lr": 5.0965781116702126e-05}
{"train_loss": 0.11921168118715286, "global_step": 134309, "epoch": 1509, "lr": 5.096520148819186e-05}
{"train_loss": 0.07980258762836456, "global_step": 134310, "epoch": 1509, "lr": 5.0964621859551844e-05}
{"train_loss": 0.1317330002784729, "global_step": 134311, "epoch": 1509, "lr": 5.096404223078214e-05}
{"train_loss": 0.12777681648731232, "global_step": 134312, "epoch": 1509, "lr": 5.096346260188284e-05}
{"train_loss": 0.08227995783090591, "global_step": 134313, "epoch": 1509, "lr": 5.0962882972853996e-05}
{"train_loss": 0.1871633380651474, "global_step": 134314, "epoch": 1509, "lr": 5.0962303343695714e-05}
{"train_loss": 0.1299542635679245, "global_step": 134315, "epoch": 1509, "lr": 5.096172371440806e-05}
{"train_loss": 0.12400413304567337, "global_step": 134316, "epoch": 1509, "lr": 5.096114408499112e-05}
{"train_loss": 0.14545005559921265, "global_step": 134317, "epoch": 1509, "lr": 5.0960564455444957e-05}
{"train_loss": 0.0916697084903717, "global_step": 134318, "epoch": 1509, "lr": 5.095998482576967e-05}
{"train_loss": 0.10338500887155533, "global_step": 134319, "epoch": 1509, "lr": 5.095940519596531e-05}
{"train_loss": 0.14044123888015747, "global_step": 134320, "epoch": 1509, "lr": 5.095882556603199e-05}
{"train_loss": 0.187017023563385, "global_step": 134321, "epoch": 1509, "lr": 5.095824593596975e-05}
{"train_loss": 0.10605250298976898, "global_step": 134322, "epoch": 1509, "lr": 5.095766630577869e-05}
{"train_loss": 0.08466280996799469, "global_step": 134323, "epoch": 1509, "lr": 5.0957086675458884e-05}
{"train_loss": 0.1349405199289322, "global_step": 134324, "epoch": 1509, "lr": 5.0956507045010414e-05}
{"train_loss": 0.14730016887187958, "global_step": 134325, "epoch": 1509, "lr": 5.095592741443337e-05}
{"train_loss": 0.12596121430397034, "global_step": 134326, "epoch": 1509, "lr": 5.095534778372779e-05}
{"train_loss": 0.13484437763690948, "global_step": 134327, "epoch": 1509, "lr": 5.095476815289377e-05}
{"train_loss": 0.08584655076265335, "global_step": 134328, "epoch": 1509, "lr": 5.095418852193141e-05}
{"train_loss": 0.10382474958896637, "global_step": 134329, "epoch": 1509, "lr": 5.0953608890840754e-05}
{"train_loss": 0.07703585177659988, "global_step": 134330, "epoch": 1509, "lr": 5.095302925962191e-05}
{"train_loss": 0.11086133867502213, "global_step": 134331, "epoch": 1509, "lr": 5.0952449628274936e-05}
{"train_loss": 0.14376458525657654, "global_step": 134332, "epoch": 1509, "lr": 5.095186999679992e-05}
{"train_loss": 0.09693845361471176, "global_step": 134333, "epoch": 1509, "lr": 5.095129036519693e-05}
{"train_loss": 0.09257853776216507, "global_step": 134334, "epoch": 1509, "lr": 5.095071073346606e-05}
{"train_loss": 0.10823366791009903, "global_step": 134335, "epoch": 1509, "lr": 5.0950131101607365e-05}
{"train_loss": 0.1602974385023117, "global_step": 134336, "epoch": 1509, "lr": 5.094955146962095e-05}
{"train_loss": 0.0921189934015274, "global_step": 134337, "epoch": 1509, "lr": 5.094897183750687e-05}
{"train_loss": 0.14802035689353943, "global_step": 134338, "epoch": 1509, "lr": 5.0948392205265214e-05}
{"train_loss": 0.10025522112846375, "global_step": 134339, "epoch": 1509, "lr": 5.094781257289607e-05}
{"train_loss": 0.1213059276342392, "global_step": 134340, "epoch": 1509, "lr": 5.094723294039949e-05}
{"train_loss": 0.10712531954050064, "global_step": 134341, "epoch": 1509, "lr": 5.094665330777557e-05}
{"train_loss": 0.19970044493675232, "global_step": 134342, "epoch": 1509, "lr": 5.09460736750244e-05}
{"train_loss": 0.12116548418998718, "global_step": 134343, "epoch": 1509, "lr": 5.094549404214603e-05}
{"train_loss": 0.0796482264995575, "global_step": 134344, "epoch": 1509, "lr": 5.0944914409140555e-05}
{"train_loss": 0.14840330183506012, "global_step": 134345, "epoch": 1509, "lr": 5.094433477600804e-05}
{"train_loss": 0.10431265085935593, "global_step": 134346, "epoch": 1509, "lr": 5.094375514274856e-05}
{"train_loss": 0.1165299117565155, "global_step": 134347, "epoch": 1509, "lr": 5.0943175509362215e-05}
{"train_loss": 0.16293838620185852, "global_step": 134348, "epoch": 1509, "lr": 5.094259587584908e-05}
{"train_loss": 0.128618061542511, "global_step": 134349, "epoch": 1509, "lr": 5.094201624220921e-05}
{"train_loss": 0.19643491506576538, "global_step": 134350, "epoch": 1509, "lr": 5.0941436608442704e-05}
{"train_loss": 0.1138739213347435, "global_step": 134351, "epoch": 1509, "lr": 5.094085697454962e-05}
{"train_loss": 0.13340477645397186, "global_step": 134352, "epoch": 1509, "lr": 5.094027734053006e-05}
{"train_loss": 0.14140412211418152, "global_step": 134353, "epoch": 1509, "lr": 5.0939697706384103e-05}
{"train_loss": 0.12190023809671402, "global_step": 134354, "epoch": 1509, "lr": 5.09391180721118e-05}
{"train_loss": 0.11781613528728485, "global_step": 134355, "epoch": 1509, "lr": 5.093853843771326e-05}
{"train_loss": 0.11247961223125458, "global_step": 134356, "epoch": 1509, "lr": 5.093795880318852e-05}
{"train_loss": 0.15858624875545502, "global_step": 134357, "epoch": 1509, "lr": 5.0937379168537704e-05}
{"train_loss": 0.20419812202453613, "global_step": 134358, "epoch": 1509, "lr": 5.093679953376086e-05}
{"train_loss": 0.1905217170715332, "global_step": 134359, "epoch": 1509, "lr": 5.093621989885808e-05}
{"train_loss": 0.0785866230726242, "global_step": 134360, "epoch": 1509, "lr": 5.0935640263829435e-05}
{"train_loss": 0.1472444236278534, "global_step": 134361, "epoch": 1509, "lr": 5.0935060628675e-05}
{"train_loss": 0.15108518302440643, "global_step": 134362, "epoch": 1509, "lr": 5.0934480993394864e-05}
{"train_loss": 0.16026152670383453, "global_step": 134363, "epoch": 1509, "lr": 5.09339013579891e-05}
{"train_loss": 0.21990032494068146, "global_step": 134364, "epoch": 1509, "lr": 5.0933321722457785e-05}
{"train_loss": 0.11121922731399536, "global_step": 134365, "epoch": 1509, "lr": 5.093274208680099e-05}
{"train_loss": 0.07212696224451065, "global_step": 134366, "epoch": 1509, "lr": 5.0932162451018796e-05}
{"train_loss": 0.11882327497005463, "global_step": 134367, "epoch": 1509, "lr": 5.093158281511129e-05}
{"train_loss": 0.08958688378334045, "global_step": 134368, "epoch": 1509, "lr": 5.093100317907854e-05}
{"train_loss": 0.11633852124214172, "global_step": 134369, "epoch": 1509, "lr": 5.093042354292061e-05}
{"train_loss": 0.11043566465377808, "global_step": 134370, "epoch": 1509, "lr": 5.092984390663762e-05}
{"train_loss": 0.14784541726112366, "global_step": 134371, "epoch": 1509, "lr": 5.092926427022963e-05}
{"train_loss": 0.11740212142467499, "global_step": 134372, "epoch": 1509, "lr": 5.09286846336967e-05}
{"train_loss": 0.1061471477150917, "global_step": 134373, "epoch": 1509, "lr": 5.092810499703892e-05}
{"train_loss": 0.07942870259284973, "global_step": 134374, "epoch": 1509, "lr": 5.0927525360256365e-05}
{"train_loss": 0.1275288313627243, "global_step": 134375, "epoch": 1509, "lr": 5.092694572334912e-05}
{"train_loss": 0.1492006629705429, "global_step": 134376, "epoch": 1509, "lr": 5.092636608631726e-05}
{"train_loss": 0.1734066754579544, "global_step": 134377, "epoch": 1509, "lr": 5.092578644916085e-05}
{"train_loss": 0.15520741045475006, "global_step": 134378, "epoch": 1509, "lr": 5.092520681187999e-05}
{"train_loss": 0.14573653042316437, "global_step": 134379, "epoch": 1509, "lr": 5.092462717447475e-05}
{"train_loss": 0.1299147754907608, "global_step": 134380, "epoch": 1509, "lr": 5.0924047536945194e-05}
{"train_loss": 0.19211681187152863, "global_step": 134381, "epoch": 1509, "lr": 5.092346789929142e-05}
{"train_loss": 0.06918985396623611, "global_step": 134382, "epoch": 1509, "lr": 5.09228882615135e-05}
{"train_loss": 0.08174009621143341, "global_step": 134383, "epoch": 1509, "lr": 5.0922308623611515e-05}
{"train_loss": 0.1011696308851242, "global_step": 134384, "epoch": 1509, "lr": 5.092172898558553e-05}
{"train_loss": 0.10918788611888885, "global_step": 134385, "epoch": 1509, "lr": 5.092114934743561e-05}
{"train_loss": 0.09573711454868317, "global_step": 134386, "epoch": 1509, "lr": 5.092056970916187e-05}
{"train_loss": 0.15491299331188202, "global_step": 134387, "epoch": 1509, "lr": 5.091999007076436e-05}
{"train_loss": 0.0771840289235115, "global_step": 134388, "epoch": 1509, "lr": 5.0919410432243176e-05}
{"train_loss": 0.1254286826326606, "global_step": 134389, "epoch": 1509, "lr": 5.0918830793598395e-05, "val_loss": 5.671451568603516}
{"train_loss": 0.16063731908798218, "global_step": 134390, "epoch": 1510, "lr": 5.091825115483008e-05}
{"train_loss": 0.13631796836853027, "global_step": 134391, "epoch": 1510, "lr": 5.0917671515938315e-05}
{"train_loss": 0.09168735891580582, "global_step": 134392, "epoch": 1510, "lr": 5.09170918769232e-05}
{"train_loss": 0.1632072478532791, "global_step": 134393, "epoch": 1510, "lr": 5.0916512237784766e-05}
{"train_loss": 0.1261865794658661, "global_step": 134394, "epoch": 1510, "lr": 5.091593259852313e-05}
{"train_loss": 0.13664717972278595, "global_step": 134395, "epoch": 1510, "lr": 5.091535295913836e-05}
{"train_loss": 0.1647786796092987, "global_step": 134396, "epoch": 1510, "lr": 5.091477331963054e-05}
{"train_loss": 0.0877712294459343, "global_step": 134397, "epoch": 1510, "lr": 5.091419367999972e-05}
{"train_loss": 0.08467871695756912, "global_step": 134398, "epoch": 1510, "lr": 5.0913614040246016e-05}
{"train_loss": 0.1205931082367897, "global_step": 134399, "epoch": 1510, "lr": 5.0913034400369474e-05}
{"train_loss": 0.10764792561531067, "global_step": 134400, "epoch": 1510, "lr": 5.09124547603702e-05}
{"train_loss": 0.14244627952575684, "global_step": 134401, "epoch": 1510, "lr": 5.091187512024824e-05}
{"train_loss": 0.09940172731876373, "global_step": 134402, "epoch": 1510, "lr": 5.0911295480003706e-05}
{"train_loss": 0.11140042543411255, "global_step": 134403, "epoch": 1510, "lr": 5.091071583963666e-05}
{"train_loss": 0.17347314953804016, "global_step": 134404, "epoch": 1510, "lr": 5.0910136199147174e-05}
{"train_loss": 0.17911900579929352, "global_step": 134405, "epoch": 1510, "lr": 5.090955655853531e-05}
{"train_loss": 0.1736077219247818, "global_step": 134406, "epoch": 1510, "lr": 5.0908976917801186e-05}
{"train_loss": 0.1586538404226303, "global_step": 134407, "epoch": 1510, "lr": 5.090839727694486e-05}
{"train_loss": 0.12034212797880173, "global_step": 134408, "epoch": 1510, "lr": 5.090781763596641e-05}
{"train_loss": 0.10230861604213715, "global_step": 134409, "epoch": 1510, "lr": 5.090723799486592e-05}
{"train_loss": 0.2187630832195282, "global_step": 134410, "epoch": 1510, "lr": 5.090665835364347e-05}
{"train_loss": 0.17718598246574402, "global_step": 134411, "epoch": 1510, "lr": 5.090607871229911e-05}
{"train_loss": 0.10803630948066711, "global_step": 134412, "epoch": 1510, "lr": 5.0905499070832954e-05}
{"train_loss": 0.04853847622871399, "global_step": 134413, "epoch": 1510, "lr": 5.090491942924506e-05}
{"train_loss": 0.11459704488515854, "global_step": 134414, "epoch": 1510, "lr": 5.0904339787535517e-05}
{"train_loss": 0.06534407287836075, "global_step": 134415, "epoch": 1510, "lr": 5.090376014570438e-05}
{"train_loss": 0.16776208579540253, "global_step": 134416, "epoch": 1510, "lr": 5.090318050375176e-05}
{"train_loss": 0.1394561231136322, "global_step": 134417, "epoch": 1510, "lr": 5.0902600861677716e-05}
{"train_loss": 0.09897816926240921, "global_step": 134418, "epoch": 1510, "lr": 5.090202121948232e-05}
{"train_loss": 0.04820169508457184, "global_step": 134419, "epoch": 1510, "lr": 5.090144157716567e-05}
{"train_loss": 0.07243655622005463, "global_step": 134420, "epoch": 1510, "lr": 5.090086193472782e-05}
{"train_loss": 0.15301507711410522, "global_step": 134421, "epoch": 1510, "lr": 5.090028229216888e-05}
{"train_loss": 0.17906661331653595, "global_step": 134422, "epoch": 1510, "lr": 5.089970264948889e-05}
{"train_loss": 0.10613247752189636, "global_step": 134423, "epoch": 1510, "lr": 5.089912300668793e-05}
{"train_loss": 0.14185212552547455, "global_step": 134424, "epoch": 1510, "lr": 5.089854336376612e-05}
{"train_loss": 0.11409691721200943, "global_step": 134425, "epoch": 1510, "lr": 5.089796372072351e-05}
{"train_loss": 0.11358396708965302, "global_step": 134426, "epoch": 1510, "lr": 5.089738407756017e-05}
{"train_loss": 0.15211310982704163, "global_step": 134427, "epoch": 1510, "lr": 5.089680443427619e-05}
{"train_loss": 0.16008210182189941, "global_step": 134428, "epoch": 1510, "lr": 5.089622479087165e-05}
{"train_loss": 0.13225875794887543, "global_step": 134429, "epoch": 1510, "lr": 5.0895645147346623e-05}
{"train_loss": 0.10347110778093338, "global_step": 134430, "epoch": 1510, "lr": 5.089506550370119e-05}
{"train_loss": 0.09276548027992249, "global_step": 134431, "epoch": 1510, "lr": 5.089448585993541e-05}
{"train_loss": 0.13604889810085297, "global_step": 134432, "epoch": 1510, "lr": 5.0893906216049394e-05}
{"train_loss": 0.09916780143976212, "global_step": 134433, "epoch": 1510, "lr": 5.08933265720432e-05}
{"train_loss": 0.0657007172703743, "global_step": 134434, "epoch": 1510, "lr": 5.089274692791691e-05}
{"train_loss": 0.09632451087236404, "global_step": 134435, "epoch": 1510, "lr": 5.08921672836706e-05}
{"train_loss": 0.13047093152999878, "global_step": 134436, "epoch": 1510, "lr": 5.089158763930435e-05}
{"train_loss": 0.12614765763282776, "global_step": 134437, "epoch": 1510, "lr": 5.089100799481823e-05}
{"train_loss": 0.113436259329319, "global_step": 134438, "epoch": 1510, "lr": 5.0890428350212324e-05}
{"train_loss": 0.1550472527742386, "global_step": 134439, "epoch": 1510, "lr": 5.088984870548672e-05}
{"train_loss": 0.158846914768219, "global_step": 134440, "epoch": 1510, "lr": 5.0889269060641483e-05}
{"train_loss": 0.1678805649280548, "global_step": 134441, "epoch": 1510, "lr": 5.088868941567669e-05}
{"train_loss": 0.07852673530578613, "global_step": 134442, "epoch": 1510, "lr": 5.088810977059243e-05}
{"train_loss": 0.15459203720092773, "global_step": 134443, "epoch": 1510, "lr": 5.088753012538877e-05}
{"train_loss": 0.06257876753807068, "global_step": 134444, "epoch": 1510, "lr": 5.0886950480065795e-05}
{"train_loss": 0.14118626713752747, "global_step": 134445, "epoch": 1510, "lr": 5.088637083462359e-05}
{"train_loss": 0.1904340386390686, "global_step": 134446, "epoch": 1510, "lr": 5.0885791189062215e-05}
{"train_loss": 0.10142301768064499, "global_step": 134447, "epoch": 1510, "lr": 5.088521154338175e-05}
{"train_loss": 0.1049446165561676, "global_step": 134448, "epoch": 1510, "lr": 5.088463189758228e-05}
{"train_loss": 0.11900988221168518, "global_step": 134449, "epoch": 1510, "lr": 5.0884052251663895e-05}
{"train_loss": 0.13556475937366486, "global_step": 134450, "epoch": 1510, "lr": 5.0883472605626645e-05}
{"train_loss": 0.14054220914840698, "global_step": 134451, "epoch": 1510, "lr": 5.088289295947063e-05}
{"train_loss": 0.1496037393808365, "global_step": 134452, "epoch": 1510, "lr": 5.0882313313195916e-05}
{"train_loss": 0.09986337274312973, "global_step": 134453, "epoch": 1510, "lr": 5.0881733666802586e-05}
{"train_loss": 0.1415354609489441, "global_step": 134454, "epoch": 1510, "lr": 5.088115402029072e-05}
{"train_loss": 0.08730848878622055, "global_step": 134455, "epoch": 1510, "lr": 5.08805743736604e-05}
{"train_loss": 0.11287864297628403, "global_step": 134456, "epoch": 1510, "lr": 5.087999472691169e-05}
{"train_loss": 0.08219391852617264, "global_step": 134457, "epoch": 1510, "lr": 5.087941508004468e-05}
{"train_loss": 0.060099128633737564, "global_step": 134458, "epoch": 1510, "lr": 5.087883543305943e-05}
{"train_loss": 0.05818610265851021, "global_step": 134459, "epoch": 1510, "lr": 5.087825578595605e-05}
{"train_loss": 0.11894349753856659, "global_step": 134460, "epoch": 1510, "lr": 5.0877676138734595e-05}
{"train_loss": 0.12835809588432312, "global_step": 134461, "epoch": 1510, "lr": 5.0877096491395136e-05}
{"train_loss": 0.16952775418758392, "global_step": 134462, "epoch": 1510, "lr": 5.087651684393777e-05}
{"train_loss": 0.11234144121408463, "global_step": 134463, "epoch": 1510, "lr": 5.0875937196362566e-05}
{"train_loss": 0.09547483921051025, "global_step": 134464, "epoch": 1510, "lr": 5.087535754866961e-05}
{"train_loss": 0.09406115114688873, "global_step": 134465, "epoch": 1510, "lr": 5.0874777900858964e-05}
{"train_loss": 0.13450168073177338, "global_step": 134466, "epoch": 1510, "lr": 5.0874198252930715e-05}
{"train_loss": 0.14071767032146454, "global_step": 134467, "epoch": 1510, "lr": 5.087361860488494e-05}
{"train_loss": 0.16548652946949005, "global_step": 134468, "epoch": 1510, "lr": 5.087303895672173e-05}
{"train_loss": 0.12819233536720276, "global_step": 134469, "epoch": 1510, "lr": 5.087245930844113e-05}
{"train_loss": 0.07064062356948853, "global_step": 134470, "epoch": 1510, "lr": 5.0871879660043254e-05}
{"train_loss": 0.1229526549577713, "global_step": 134471, "epoch": 1510, "lr": 5.087130001152817e-05}
{"train_loss": 0.11967366933822632, "global_step": 134472, "epoch": 1510, "lr": 5.087072036289593e-05}
{"train_loss": 0.10853210836648941, "global_step": 134473, "epoch": 1510, "lr": 5.087014071414664e-05}
{"train_loss": 0.09792336076498032, "global_step": 134474, "epoch": 1510, "lr": 5.086956106528038e-05}
{"train_loss": 0.0863189697265625, "global_step": 134475, "epoch": 1510, "lr": 5.086898141629721e-05}
{"train_loss": 0.10409694910049438, "global_step": 134476, "epoch": 1510, "lr": 5.086840176719722e-05}
{"train_loss": 0.1077447161078453, "global_step": 134477, "epoch": 1510, "lr": 5.0867822117980476e-05}
{"train_loss": 0.12227252443854729, "global_step": 134478, "epoch": 1510, "lr": 5.086724246864707e-05, "val_loss": 5.468319892883301, "train_action_mse_error": 4.395900249481201}
{"train_loss": 0.06457839906215668, "global_step": 134479, "epoch": 1511, "lr": 5.0866662819197074e-05}
{"train_loss": 0.1361577808856964, "global_step": 134480, "epoch": 1511, "lr": 5.086608316963056e-05}
{"train_loss": 0.09739542007446289, "global_step": 134481, "epoch": 1511, "lr": 5.086550351994761e-05}
{"train_loss": 0.13210909068584442, "global_step": 134482, "epoch": 1511, "lr": 5.086492387014832e-05}
{"train_loss": 0.10037128627300262, "global_step": 134483, "epoch": 1511, "lr": 5.086434422023274e-05}
{"train_loss": 0.1128353551030159, "global_step": 134484, "epoch": 1511, "lr": 5.086376457020096e-05}
{"train_loss": 0.1352434605360031, "global_step": 134485, "epoch": 1511, "lr": 5.0863184920053056e-05}
{"train_loss": 0.10184290260076523, "global_step": 134486, "epoch": 1511, "lr": 5.086260526978911e-05}
{"train_loss": 0.1410471498966217, "global_step": 134487, "epoch": 1511, "lr": 5.086202561940919e-05}
{"train_loss": 0.09940951317548752, "global_step": 134488, "epoch": 1511, "lr": 5.086144596891339e-05}
{"train_loss": 0.11582963913679123, "global_step": 134489, "epoch": 1511, "lr": 5.086086631830178e-05}
{"train_loss": 0.11973173171281815, "global_step": 134490, "epoch": 1511, "lr": 5.086028666757443e-05}
{"train_loss": 0.1391407996416092, "global_step": 134491, "epoch": 1511, "lr": 5.085970701673143e-05}
{"train_loss": 0.13037484884262085, "global_step": 134492, "epoch": 1511, "lr": 5.085912736577285e-05}
{"train_loss": 0.0753757581114769, "global_step": 134493, "epoch": 1511, "lr": 5.085854771469877e-05}
{"train_loss": 0.1658472716808319, "global_step": 134494, "epoch": 1511, "lr": 5.0857968063509276e-05}
{"train_loss": 0.08148591220378876, "global_step": 134495, "epoch": 1511, "lr": 5.085738841220443e-05}
{"train_loss": 0.09016851335763931, "global_step": 134496, "epoch": 1511, "lr": 5.0856808760784325e-05}
{"train_loss": 0.12806211411952972, "global_step": 134497, "epoch": 1511, "lr": 5.085622910924903e-05}
{"train_loss": 0.16796354949474335, "global_step": 134498, "epoch": 1511, "lr": 5.085564945759862e-05}
{"train_loss": 0.09518548101186752, "global_step": 134499, "epoch": 1511, "lr": 5.0855069805833186e-05}
{"train_loss": 0.15381096303462982, "global_step": 134500, "epoch": 1511, "lr": 5.08544901539528e-05}
{"train_loss": 0.1681840419769287, "global_step": 134501, "epoch": 1511, "lr": 5.0853910501957523e-05}
{"train_loss": 0.03029472567141056, "global_step": 134502, "epoch": 1511, "lr": 5.085333084984747e-05}
{"train_loss": 0.18451951444149017, "global_step": 134503, "epoch": 1511, "lr": 5.085275119762268e-05}
{"train_loss": 0.11603991687297821, "global_step": 134504, "epoch": 1511, "lr": 5.0852171545283255e-05}
{"train_loss": 0.08039167523384094, "global_step": 134505, "epoch": 1511, "lr": 5.085159189282926e-05}
{"train_loss": 0.12042948603630066, "global_step": 134506, "epoch": 1511, "lr": 5.0851012240260785e-05}
{"train_loss": 0.1294146329164505, "global_step": 134507, "epoch": 1511, "lr": 5.0850432587577903e-05}
{"train_loss": 0.08924365043640137, "global_step": 134508, "epoch": 1511, "lr": 5.084985293478069e-05}
{"train_loss": 0.11131777614355087, "global_step": 134509, "epoch": 1511, "lr": 5.084927328186923e-05}
{"train_loss": 0.16146628558635712, "global_step": 134510, "epoch": 1511, "lr": 5.084869362884358e-05}
{"train_loss": 0.1256224364042282, "global_step": 134511, "epoch": 1511, "lr": 5.0848113975703846e-05}
{"train_loss": 0.13492092490196228, "global_step": 134512, "epoch": 1511, "lr": 5.0847534322450085e-05}
{"train_loss": 0.07995674759149551, "global_step": 134513, "epoch": 1511, "lr": 5.0846954669082394e-05}
{"train_loss": 0.10415785759687424, "global_step": 134514, "epoch": 1511, "lr": 5.0846375015600836e-05}
{"train_loss": 0.06597717106342316, "global_step": 134515, "epoch": 1511, "lr": 5.084579536200549e-05}
{"train_loss": 0.13212986290454865, "global_step": 134516, "epoch": 1511, "lr": 5.084521570829643e-05}
{"train_loss": 0.10148608684539795, "global_step": 134517, "epoch": 1511, "lr": 5.084463605447375e-05}
{"train_loss": 0.08138535916805267, "global_step": 134518, "epoch": 1511, "lr": 5.084405640053752e-05}
{"train_loss": 0.11991336196660995, "global_step": 134519, "epoch": 1511, "lr": 5.0843476746487815e-05}
{"train_loss": 0.11229023337364197, "global_step": 134520, "epoch": 1511, "lr": 5.084289709232471e-05}
{"train_loss": 0.11892411857843399, "global_step": 134521, "epoch": 1511, "lr": 5.0842317438048306e-05}
{"train_loss": 0.13510139286518097, "global_step": 134522, "epoch": 1511, "lr": 5.084173778365864e-05}
{"train_loss": 0.1087755560874939, "global_step": 134523, "epoch": 1511, "lr": 5.084115812915583e-05}
{"train_loss": 0.12019456923007965, "global_step": 134524, "epoch": 1511, "lr": 5.0840578474539925e-05}
{"train_loss": 0.14523790776729584, "global_step": 134525, "epoch": 1511, "lr": 5.083999881981102e-05}
{"train_loss": 0.19048361480236053, "global_step": 134526, "epoch": 1511, "lr": 5.083941916496918e-05}
{"train_loss": 0.07495774328708649, "global_step": 134527, "epoch": 1511, "lr": 5.08388395100145e-05}
{"train_loss": 0.13012287020683289, "global_step": 134528, "epoch": 1511, "lr": 5.0838259854947046e-05}
{"train_loss": 0.05575529485940933, "global_step": 134529, "epoch": 1511, "lr": 5.083768019976689e-05}
{"train_loss": 0.12666374444961548, "global_step": 134530, "epoch": 1511, "lr": 5.083710054447413e-05}
{"train_loss": 0.10016407072544098, "global_step": 134531, "epoch": 1511, "lr": 5.083652088906883e-05}
{"train_loss": 0.1027921587228775, "global_step": 134532, "epoch": 1511, "lr": 5.0835941233551064e-05}
{"train_loss": 0.03769930824637413, "global_step": 134533, "epoch": 1511, "lr": 5.0835361577920915e-05}
{"train_loss": 0.12849976122379303, "global_step": 134534, "epoch": 1511, "lr": 5.083478192217846e-05}
{"train_loss": 0.12998785078525543, "global_step": 134535, "epoch": 1511, "lr": 5.0834202266323795e-05}
{"train_loss": 0.14495015144348145, "global_step": 134536, "epoch": 1511, "lr": 5.083362261035697e-05}
{"train_loss": 0.10688408464193344, "global_step": 134537, "epoch": 1511, "lr": 5.0833042954278066e-05}
{"train_loss": 0.19340015947818756, "global_step": 134538, "epoch": 1511, "lr": 5.083246329808719e-05}
{"train_loss": 0.13336333632469177, "global_step": 134539, "epoch": 1511, "lr": 5.0831883641784386e-05}
{"train_loss": 0.12531743943691254, "global_step": 134540, "epoch": 1511, "lr": 5.0831303985369735e-05}
{"train_loss": 0.11249876022338867, "global_step": 134541, "epoch": 1511, "lr": 5.083072432884335e-05}
{"train_loss": 0.14563871920108795, "global_step": 134542, "epoch": 1511, "lr": 5.083014467220527e-05}
{"train_loss": 0.1442108452320099, "global_step": 134543, "epoch": 1511, "lr": 5.082956501545558e-05}
{"train_loss": 0.16497193276882172, "global_step": 134544, "epoch": 1511, "lr": 5.082898535859438e-05}
{"train_loss": 0.12367890030145645, "global_step": 134545, "epoch": 1511, "lr": 5.082840570162173e-05}
{"train_loss": 0.09597846120595932, "global_step": 134546, "epoch": 1511, "lr": 5.082782604453771e-05}
{"train_loss": 0.12121886759996414, "global_step": 134547, "epoch": 1511, "lr": 5.08272463873424e-05}
{"train_loss": 0.13174015283584595, "global_step": 134548, "epoch": 1511, "lr": 5.0826666730035864e-05}
{"train_loss": 0.1057337075471878, "global_step": 134549, "epoch": 1511, "lr": 5.082608707261821e-05}
{"train_loss": 0.17714793980121613, "global_step": 134550, "epoch": 1511, "lr": 5.082550741508948e-05}
{"train_loss": 0.16885462403297424, "global_step": 134551, "epoch": 1511, "lr": 5.082492775744978e-05}
{"train_loss": 0.10199283808469772, "global_step": 134552, "epoch": 1511, "lr": 5.082434809969919e-05}
{"train_loss": 0.10638570785522461, "global_step": 134553, "epoch": 1511, "lr": 5.0823768441837775e-05}
{"train_loss": 0.08956047147512436, "global_step": 134554, "epoch": 1511, "lr": 5.082318878386559e-05}
{"train_loss": 0.09411369264125824, "global_step": 134555, "epoch": 1511, "lr": 5.082260912578276e-05}
{"train_loss": 0.10283811390399933, "global_step": 134556, "epoch": 1511, "lr": 5.082202946758934e-05}
{"train_loss": 0.1510983109474182, "global_step": 134557, "epoch": 1511, "lr": 5.082144980928541e-05}
{"train_loss": 0.13687705993652344, "global_step": 134558, "epoch": 1511, "lr": 5.0820870150871026e-05}
{"train_loss": 0.09881658107042313, "global_step": 134559, "epoch": 1511, "lr": 5.08202904923463e-05}
{"train_loss": 0.12471765279769897, "global_step": 134560, "epoch": 1511, "lr": 5.08197108337113e-05}
{"train_loss": 0.14659357070922852, "global_step": 134561, "epoch": 1511, "lr": 5.081913117496608e-05}
{"train_loss": 0.1303657591342926, "global_step": 134562, "epoch": 1511, "lr": 5.0818551516110755e-05}
{"train_loss": 0.1726236641407013, "global_step": 134563, "epoch": 1511, "lr": 5.081797185714538e-05}
{"train_loss": 0.14231252670288086, "global_step": 134564, "epoch": 1511, "lr": 5.0817392198070046e-05}
{"train_loss": 0.14737829566001892, "global_step": 134565, "epoch": 1511, "lr": 5.081681253888482e-05}
{"train_loss": 0.12568901479244232, "global_step": 134566, "epoch": 1511, "lr": 5.081623287958979e-05}
{"train_loss": 0.12068673162564134, "global_step": 134567, "epoch": 1511, "lr": 5.081565322018501e-05, "val_loss": 5.671334743499756}
{"train_loss": 0.14564134180545807, "global_step": 134568, "epoch": 1512, "lr": 5.0815073560670576e-05}
{"train_loss": 0.15649491548538208, "global_step": 134569, "epoch": 1512, "lr": 5.081449390104658e-05}
{"train_loss": 0.11952678114175797, "global_step": 134570, "epoch": 1512, "lr": 5.0813914241313076e-05}
{"train_loss": 0.08315352350473404, "global_step": 134571, "epoch": 1512, "lr": 5.0813334581470164e-05}
{"train_loss": 0.12689155340194702, "global_step": 134572, "epoch": 1512, "lr": 5.0812754921517904e-05}
{"train_loss": 0.1039116159081459, "global_step": 134573, "epoch": 1512, "lr": 5.0812175261456377e-05}
{"train_loss": 0.09369739145040512, "global_step": 134574, "epoch": 1512, "lr": 5.0811595601285656e-05}
{"train_loss": 0.1283208727836609, "global_step": 134575, "epoch": 1512, "lr": 5.0811015941005825e-05}
{"train_loss": 0.06103815138339996, "global_step": 134576, "epoch": 1512, "lr": 5.081043628061697e-05}
{"train_loss": 0.19831600785255432, "global_step": 134577, "epoch": 1512, "lr": 5.080985662011916e-05}
{"train_loss": 0.11597199738025665, "global_step": 134578, "epoch": 1512, "lr": 5.0809276959512474e-05}
{"train_loss": 0.10530006885528564, "global_step": 134579, "epoch": 1512, "lr": 5.080869729879699e-05}
{"train_loss": 0.1339864581823349, "global_step": 134580, "epoch": 1512, "lr": 5.0808117637972794e-05}
{"train_loss": 0.13113096356391907, "global_step": 134581, "epoch": 1512, "lr": 5.080753797703994e-05}
{"train_loss": 0.07890570908784866, "global_step": 134582, "epoch": 1512, "lr": 5.0806958315998545e-05}
{"train_loss": 0.10288836061954498, "global_step": 134583, "epoch": 1512, "lr": 5.0806378654848654e-05}
{"train_loss": 0.1677972376346588, "global_step": 134584, "epoch": 1512, "lr": 5.080579899359035e-05}
{"train_loss": 0.164052352309227, "global_step": 134585, "epoch": 1512, "lr": 5.080521933222372e-05}
{"train_loss": 0.13324248790740967, "global_step": 134586, "epoch": 1512, "lr": 5.0804639670748835e-05}
{"train_loss": 0.15764844417572021, "global_step": 134587, "epoch": 1512, "lr": 5.0804060009165765e-05}
{"train_loss": 0.08274891972541809, "global_step": 134588, "epoch": 1512, "lr": 5.080348034747461e-05}
{"train_loss": 0.12140685319900513, "global_step": 134589, "epoch": 1512, "lr": 5.0802900685675435e-05}
{"train_loss": 0.14180317521095276, "global_step": 134590, "epoch": 1512, "lr": 5.080232102376834e-05}
{"train_loss": 0.1390334814786911, "global_step": 134591, "epoch": 1512, "lr": 5.080174136175336e-05}
{"train_loss": 0.13012367486953735, "global_step": 134592, "epoch": 1512, "lr": 5.0801161699630604e-05}
{"train_loss": 0.14355793595314026, "global_step": 134593, "epoch": 1512, "lr": 5.0800582037400144e-05}
{"train_loss": 0.09088737517595291, "global_step": 134594, "epoch": 1512, "lr": 5.080000237506205e-05}
{"train_loss": 0.10444426536560059, "global_step": 134595, "epoch": 1512, "lr": 5.079942271261641e-05}
{"train_loss": 0.09615218639373779, "global_step": 134596, "epoch": 1512, "lr": 5.0798843050063296e-05}
{"train_loss": 0.14211799204349518, "global_step": 134597, "epoch": 1512, "lr": 5.079826338740278e-05}
{"train_loss": 0.09745926409959793, "global_step": 134598, "epoch": 1512, "lr": 5.079768372463496e-05}
{"train_loss": 0.10320188850164413, "global_step": 134599, "epoch": 1512, "lr": 5.0797104061759896e-05}
{"train_loss": 0.16500645875930786, "global_step": 134600, "epoch": 1512, "lr": 5.079652439877767e-05}
{"train_loss": 0.13663595914840698, "global_step": 134601, "epoch": 1512, "lr": 5.079594473568835e-05}
{"train_loss": 0.13041354715824127, "global_step": 134602, "epoch": 1512, "lr": 5.079536507249204e-05}
{"train_loss": 0.14719824492931366, "global_step": 134603, "epoch": 1512, "lr": 5.079478540918879e-05}
{"train_loss": 0.1320982575416565, "global_step": 134604, "epoch": 1512, "lr": 5.07942057457787e-05}
{"train_loss": 0.1174938753247261, "global_step": 134605, "epoch": 1512, "lr": 5.0793626082261826e-05}
{"train_loss": 0.08377096056938171, "global_step": 134606, "epoch": 1512, "lr": 5.079304641863827e-05}
{"train_loss": 0.10103365778923035, "global_step": 134607, "epoch": 1512, "lr": 5.07924667549081e-05}
{"train_loss": 0.21067260205745697, "global_step": 134608, "epoch": 1512, "lr": 5.0791887091071385e-05}
{"train_loss": 0.06294605880975723, "global_step": 134609, "epoch": 1512, "lr": 5.0791307427128224e-05}
{"train_loss": 0.11970160901546478, "global_step": 134610, "epoch": 1512, "lr": 5.079072776307868e-05}
{"train_loss": 0.08460404723882675, "global_step": 134611, "epoch": 1512, "lr": 5.079014809892283e-05}
{"train_loss": 0.10816080868244171, "global_step": 134612, "epoch": 1512, "lr": 5.078956843466075e-05}
{"train_loss": 0.0856260135769844, "global_step": 134613, "epoch": 1512, "lr": 5.078898877029252e-05}
{"train_loss": 0.11695490777492523, "global_step": 134614, "epoch": 1512, "lr": 5.078840910581822e-05}
{"train_loss": 0.14038163423538208, "global_step": 134615, "epoch": 1512, "lr": 5.0787829441237936e-05}
{"train_loss": 0.07725081592798233, "global_step": 134616, "epoch": 1512, "lr": 5.078724977655174e-05}
{"train_loss": 0.09413900971412659, "global_step": 134617, "epoch": 1512, "lr": 5.078667011175969e-05}
{"train_loss": 0.07092618197202682, "global_step": 134618, "epoch": 1512, "lr": 5.078609044686189e-05}
{"train_loss": 0.10077288746833801, "global_step": 134619, "epoch": 1512, "lr": 5.0785510781858415e-05}
{"train_loss": 0.08839874714612961, "global_step": 134620, "epoch": 1512, "lr": 5.0784931116749337e-05}
{"train_loss": 0.17226974666118622, "global_step": 134621, "epoch": 1512, "lr": 5.078435145153473e-05}
{"train_loss": 0.10512874275445938, "global_step": 134622, "epoch": 1512, "lr": 5.0783771786214674e-05}
{"train_loss": 0.119888074696064, "global_step": 134623, "epoch": 1512, "lr": 5.0783192120789246e-05}
{"train_loss": 0.07790280133485794, "global_step": 134624, "epoch": 1512, "lr": 5.0782612455258535e-05}
{"train_loss": 0.06857378780841827, "global_step": 134625, "epoch": 1512, "lr": 5.0782032789622616e-05}
{"train_loss": 0.10708626359701157, "global_step": 134626, "epoch": 1512, "lr": 5.078145312388155e-05}
{"train_loss": 0.07337336987257004, "global_step": 134627, "epoch": 1512, "lr": 5.078087345803544e-05}
{"train_loss": 0.184470996260643, "global_step": 134628, "epoch": 1512, "lr": 5.078029379208434e-05}
{"train_loss": 0.15671272575855255, "global_step": 134629, "epoch": 1512, "lr": 5.077971412602837e-05}
{"train_loss": 0.09910286217927933, "global_step": 134630, "epoch": 1512, "lr": 5.077913445986754e-05}
{"train_loss": 0.09303023666143417, "global_step": 134631, "epoch": 1512, "lr": 5.077855479360198e-05}
{"train_loss": 0.054813649505376816, "global_step": 134632, "epoch": 1512, "lr": 5.077797512723175e-05}
{"train_loss": 0.1094101145863533, "global_step": 134633, "epoch": 1512, "lr": 5.077739546075693e-05}
{"train_loss": 0.12354272603988647, "global_step": 134634, "epoch": 1512, "lr": 5.07768157941776e-05}
{"train_loss": 0.10845598578453064, "global_step": 134635, "epoch": 1512, "lr": 5.0776236127493836e-05}
{"train_loss": 0.18399886786937714, "global_step": 134636, "epoch": 1512, "lr": 5.0775656460705715e-05}
{"train_loss": 0.08811759203672409, "global_step": 134637, "epoch": 1512, "lr": 5.0775076793813314e-05}
{"train_loss": 0.13432638347148895, "global_step": 134638, "epoch": 1512, "lr": 5.077449712681672e-05}
{"train_loss": 0.1199801117181778, "global_step": 134639, "epoch": 1512, "lr": 5.0773917459716004e-05}
{"train_loss": 0.16769737005233765, "global_step": 134640, "epoch": 1512, "lr": 5.0773337792511244e-05}
{"train_loss": 0.11468661576509476, "global_step": 134641, "epoch": 1512, "lr": 5.07727581252025e-05}
{"train_loss": 0.06425607949495316, "global_step": 134642, "epoch": 1512, "lr": 5.077217845778989e-05}
{"train_loss": 0.1322248876094818, "global_step": 134643, "epoch": 1512, "lr": 5.0771598790273454e-05}
{"train_loss": 0.09774881601333618, "global_step": 134644, "epoch": 1512, "lr": 5.0771019122653296e-05}
{"train_loss": 0.06336285918951035, "global_step": 134645, "epoch": 1512, "lr": 5.077043945492949e-05}
{"train_loss": 0.09862487018108368, "global_step": 134646, "epoch": 1512, "lr": 5.0769859787102105e-05}
{"train_loss": 0.10015087574720383, "global_step": 134647, "epoch": 1512, "lr": 5.0769280119171214e-05}
{"train_loss": 0.15511900186538696, "global_step": 134648, "epoch": 1512, "lr": 5.076870045113692e-05}
{"train_loss": 0.13047489523887634, "global_step": 134649, "epoch": 1512, "lr": 5.076812078299926e-05}
{"train_loss": 0.10508901625871658, "global_step": 134650, "epoch": 1512, "lr": 5.076754111475835e-05}
{"train_loss": 0.10671806335449219, "global_step": 134651, "epoch": 1512, "lr": 5.076696144641425e-05}
{"train_loss": 0.12244553864002228, "global_step": 134652, "epoch": 1512, "lr": 5.076638177796704e-05}
{"train_loss": 0.08187040686607361, "global_step": 134653, "epoch": 1512, "lr": 5.07658021094168e-05}
{"train_loss": 0.10705842077732086, "global_step": 134654, "epoch": 1512, "lr": 5.07652224407636e-05}
{"train_loss": 0.10770233720541, "global_step": 134655, "epoch": 1512, "lr": 5.076464277200754e-05}
{"train_loss": 0.11700181156564295, "global_step": 134656, "epoch": 1512, "lr": 5.076406310314867e-05, "val_loss": 5.724551200866699}
{"train_loss": 0.17695027589797974, "global_step": 134657, "epoch": 1513, "lr": 5.07634834341871e-05}
{"train_loss": 0.12158599495887756, "global_step": 134658, "epoch": 1513, "lr": 5.076290376512287e-05}
{"train_loss": 0.10346433520317078, "global_step": 134659, "epoch": 1513, "lr": 5.076232409595607e-05}
{"train_loss": 0.08124606311321259, "global_step": 134660, "epoch": 1513, "lr": 5.0761744426686795e-05}
{"train_loss": 0.12300031632184982, "global_step": 134661, "epoch": 1513, "lr": 5.0761164757315115e-05}
{"train_loss": 0.09780852496623993, "global_step": 134662, "epoch": 1513, "lr": 5.076058508784111e-05}
{"train_loss": 0.14254504442214966, "global_step": 134663, "epoch": 1513, "lr": 5.0760005418264844e-05}
{"train_loss": 0.08263081312179565, "global_step": 134664, "epoch": 1513, "lr": 5.075942574858641e-05}
{"train_loss": 0.14450617134571075, "global_step": 134665, "epoch": 1513, "lr": 5.075884607880589e-05}
{"train_loss": 0.11227503418922424, "global_step": 134666, "epoch": 1513, "lr": 5.075826640892334e-05}
{"train_loss": 0.16019262373447418, "global_step": 134667, "epoch": 1513, "lr": 5.075768673893885e-05}
{"train_loss": 0.1033000573515892, "global_step": 134668, "epoch": 1513, "lr": 5.075710706885252e-05}
{"train_loss": 0.10845193266868591, "global_step": 134669, "epoch": 1513, "lr": 5.075652739866438e-05}
{"train_loss": 0.10336597263813019, "global_step": 134670, "epoch": 1513, "lr": 5.075594772837454e-05}
{"train_loss": 0.13316260278224945, "global_step": 134671, "epoch": 1513, "lr": 5.0755368057983076e-05}
{"train_loss": 0.10239803791046143, "global_step": 134672, "epoch": 1513, "lr": 5.075478838749006e-05}
{"train_loss": 0.10793653875589371, "global_step": 134673, "epoch": 1513, "lr": 5.075420871689557e-05}
{"train_loss": 0.16823796927928925, "global_step": 134674, "epoch": 1513, "lr": 5.075362904619969e-05}
{"train_loss": 0.09710397571325302, "global_step": 134675, "epoch": 1513, "lr": 5.075304937540248e-05}
{"train_loss": 0.13958896696567535, "global_step": 134676, "epoch": 1513, "lr": 5.075246970450405e-05}
{"train_loss": 0.10684043914079666, "global_step": 134677, "epoch": 1513, "lr": 5.0751890033504446e-05}
{"train_loss": 0.10539114475250244, "global_step": 134678, "epoch": 1513, "lr": 5.075131036240376e-05}
{"train_loss": 0.11882650852203369, "global_step": 134679, "epoch": 1513, "lr": 5.075073069120208e-05}
{"train_loss": 0.1197836622595787, "global_step": 134680, "epoch": 1513, "lr": 5.075015101989946e-05}
{"train_loss": 0.1352226734161377, "global_step": 134681, "epoch": 1513, "lr": 5.0749571348496006e-05}
{"train_loss": 0.1018599271774292, "global_step": 134682, "epoch": 1513, "lr": 5.0748991676991774e-05}
{"train_loss": 0.14190006256103516, "global_step": 134683, "epoch": 1513, "lr": 5.074841200538685e-05}
{"train_loss": 0.08744526654481888, "global_step": 134684, "epoch": 1513, "lr": 5.0747832333681314e-05}
{"train_loss": 0.06746593117713928, "global_step": 134685, "epoch": 1513, "lr": 5.0747252661875234e-05}
{"train_loss": 0.1439807415008545, "global_step": 134686, "epoch": 1513, "lr": 5.07466729899687e-05}
{"train_loss": 0.18329814076423645, "global_step": 134687, "epoch": 1513, "lr": 5.0746093317961795e-05}
{"train_loss": 0.06829172372817993, "global_step": 134688, "epoch": 1513, "lr": 5.074551364585458e-05}
{"train_loss": 0.12253144383430481, "global_step": 134689, "epoch": 1513, "lr": 5.0744933973647134e-05}
{"train_loss": 0.049964115023612976, "global_step": 134690, "epoch": 1513, "lr": 5.074435430133955e-05}
{"train_loss": 0.09972195327281952, "global_step": 134691, "epoch": 1513, "lr": 5.0743774628931886e-05}
{"train_loss": 0.11505801230669022, "global_step": 134692, "epoch": 1513, "lr": 5.074319495642423e-05}
{"train_loss": 0.0793936625123024, "global_step": 134693, "epoch": 1513, "lr": 5.074261528381666e-05}
{"train_loss": 0.09621801972389221, "global_step": 134694, "epoch": 1513, "lr": 5.074203561110926e-05}
{"train_loss": 0.06361227482557297, "global_step": 134695, "epoch": 1513, "lr": 5.0741455938302095e-05}
{"train_loss": 0.13415253162384033, "global_step": 134696, "epoch": 1513, "lr": 5.074087626539525e-05}
{"train_loss": 0.1308629810810089, "global_step": 134697, "epoch": 1513, "lr": 5.074029659238881e-05}
{"train_loss": 0.08608736842870712, "global_step": 134698, "epoch": 1513, "lr": 5.0739716919282854e-05}
{"train_loss": 0.12541209161281586, "global_step": 134699, "epoch": 1513, "lr": 5.073913724607744e-05}
{"train_loss": 0.1122673824429512, "global_step": 134700, "epoch": 1513, "lr": 5.073855757277267e-05}
{"train_loss": 0.15232473611831665, "global_step": 134701, "epoch": 1513, "lr": 5.07379778993686e-05}
{"train_loss": 0.16731220483779907, "global_step": 134702, "epoch": 1513, "lr": 5.073739822586532e-05}
{"train_loss": 0.127860426902771, "global_step": 134703, "epoch": 1513, "lr": 5.073681855226291e-05}
{"train_loss": 0.09081660211086273, "global_step": 134704, "epoch": 1513, "lr": 5.073623887856144e-05}
{"train_loss": 0.11698515713214874, "global_step": 134705, "epoch": 1513, "lr": 5.073565920476099e-05}
{"train_loss": 0.111722432076931, "global_step": 134706, "epoch": 1513, "lr": 5.0735079530861654e-05}
{"train_loss": 0.05617193132638931, "global_step": 134707, "epoch": 1513, "lr": 5.0734499856863474e-05}
{"train_loss": 0.09445735067129135, "global_step": 134708, "epoch": 1513, "lr": 5.0733920182766556e-05}
{"train_loss": 0.152382954955101, "global_step": 134709, "epoch": 1513, "lr": 5.073334050857098e-05}
{"train_loss": 0.11119619756937027, "global_step": 134710, "epoch": 1513, "lr": 5.073276083427681e-05}
{"train_loss": 0.11168139427900314, "global_step": 134711, "epoch": 1513, "lr": 5.073218115988412e-05}
{"train_loss": 0.11440669745206833, "global_step": 134712, "epoch": 1513, "lr": 5.073160148539301e-05}
{"train_loss": 0.09304184466600418, "global_step": 134713, "epoch": 1513, "lr": 5.0731021810803536e-05}
{"train_loss": 0.10027111321687698, "global_step": 134714, "epoch": 1513, "lr": 5.0730442136115775e-05}
{"train_loss": 0.10766109079122543, "global_step": 134715, "epoch": 1513, "lr": 5.072986246132983e-05}
{"train_loss": 0.0804438591003418, "global_step": 134716, "epoch": 1513, "lr": 5.0729282786445764e-05}
{"train_loss": 0.13933877646923065, "global_step": 134717, "epoch": 1513, "lr": 5.0728703111463663e-05}
{"train_loss": 0.1054106280207634, "global_step": 134718, "epoch": 1513, "lr": 5.072812343638358e-05}
{"train_loss": 0.11980316042900085, "global_step": 134719, "epoch": 1513, "lr": 5.072754376120562e-05}
{"train_loss": 0.10030369460582733, "global_step": 134720, "epoch": 1513, "lr": 5.072696408592984e-05}
{"train_loss": 0.14177948236465454, "global_step": 134721, "epoch": 1513, "lr": 5.072638441055634e-05}
{"train_loss": 0.1617095172405243, "global_step": 134722, "epoch": 1513, "lr": 5.072580473508518e-05}
{"train_loss": 0.13897499442100525, "global_step": 134723, "epoch": 1513, "lr": 5.072522505951645e-05}
{"train_loss": 0.12128814309835434, "global_step": 134724, "epoch": 1513, "lr": 5.072464538385022e-05}
{"train_loss": 0.17652305960655212, "global_step": 134725, "epoch": 1513, "lr": 5.0724065708086566e-05}
{"train_loss": 0.17703449726104736, "global_step": 134726, "epoch": 1513, "lr": 5.0723486032225583e-05}
{"train_loss": 0.12827831506729126, "global_step": 134727, "epoch": 1513, "lr": 5.072290635626733e-05}
{"train_loss": 0.0914464071393013, "global_step": 134728, "epoch": 1513, "lr": 5.072232668021189e-05}
{"train_loss": 0.10746069252490997, "global_step": 134729, "epoch": 1513, "lr": 5.072174700405934e-05}
{"train_loss": 0.13025923073291779, "global_step": 134730, "epoch": 1513, "lr": 5.072116732780976e-05}
{"train_loss": 0.10898908227682114, "global_step": 134731, "epoch": 1513, "lr": 5.0720587651463226e-05}
{"train_loss": 0.147223562002182, "global_step": 134732, "epoch": 1513, "lr": 5.072000797501981e-05}
{"train_loss": 0.1353323757648468, "global_step": 134733, "epoch": 1513, "lr": 5.071942829847961e-05}
{"train_loss": 0.11879494786262512, "global_step": 134734, "epoch": 1513, "lr": 5.071884862184268e-05}
{"train_loss": 0.13989445567131042, "global_step": 134735, "epoch": 1513, "lr": 5.071826894510913e-05}
{"train_loss": 0.11415460705757141, "global_step": 134736, "epoch": 1513, "lr": 5.0717689268279e-05}
{"train_loss": 0.16477751731872559, "global_step": 134737, "epoch": 1513, "lr": 5.0717109591352394e-05}
{"train_loss": 0.14342564344406128, "global_step": 134738, "epoch": 1513, "lr": 5.0716529914329384e-05}
{"train_loss": 0.08844492584466934, "global_step": 134739, "epoch": 1513, "lr": 5.071595023721003e-05}
{"train_loss": 0.10033218562602997, "global_step": 134740, "epoch": 1513, "lr": 5.071537055999444e-05}
{"train_loss": 0.23225875198841095, "global_step": 134741, "epoch": 1513, "lr": 5.071479088268267e-05}
{"train_loss": 0.16632601618766785, "global_step": 134742, "epoch": 1513, "lr": 5.071421120527481e-05}
{"train_loss": 0.1270085871219635, "global_step": 134743, "epoch": 1513, "lr": 5.0713631527770935e-05}
{"train_loss": 0.1178116500377655, "global_step": 134744, "epoch": 1513, "lr": 5.071305185017111e-05}
{"train_loss": 0.12055898700537307, "global_step": 134745, "epoch": 1513, "lr": 5.071247217247545e-05, "val_loss": 5.741894245147705}
{"train_loss": 0.25861606001853943, "global_step": 134746, "epoch": 1514, "lr": 5.0711892494683975e-05}
{"train_loss": 0.09220930933952332, "global_step": 134747, "epoch": 1514, "lr": 5.071131281679681e-05}
{"train_loss": 0.1385057419538498, "global_step": 134748, "epoch": 1514, "lr": 5.0710733138814015e-05}
{"train_loss": 0.09544887393712997, "global_step": 134749, "epoch": 1514, "lr": 5.071015346073566e-05}
{"train_loss": 0.11533892154693604, "global_step": 134750, "epoch": 1514, "lr": 5.070957378256183e-05}
{"train_loss": 0.17645639181137085, "global_step": 134751, "epoch": 1514, "lr": 5.070899410429263e-05}
{"train_loss": 0.11893359571695328, "global_step": 134752, "epoch": 1514, "lr": 5.070841442592811e-05}
{"train_loss": 0.21036329865455627, "global_step": 134753, "epoch": 1514, "lr": 5.070783474746834e-05}
{"train_loss": 0.0986851155757904, "global_step": 134754, "epoch": 1514, "lr": 5.070725506891342e-05}
{"train_loss": 0.10425201803445816, "global_step": 134755, "epoch": 1514, "lr": 5.0706675390263415e-05}
{"train_loss": 0.13734988868236542, "global_step": 134756, "epoch": 1514, "lr": 5.070609571151841e-05}
{"train_loss": 0.11509176343679428, "global_step": 134757, "epoch": 1514, "lr": 5.070551603267847e-05}
{"train_loss": 0.09894777089357376, "global_step": 134758, "epoch": 1514, "lr": 5.0704936353743695e-05}
{"train_loss": 0.14065077900886536, "global_step": 134759, "epoch": 1514, "lr": 5.070435667471414e-05}
{"train_loss": 0.11627493798732758, "global_step": 134760, "epoch": 1514, "lr": 5.070377699558989e-05}
{"train_loss": 0.1186252161860466, "global_step": 134761, "epoch": 1514, "lr": 5.070319731637103e-05}
{"train_loss": 0.09616301953792572, "global_step": 134762, "epoch": 1514, "lr": 5.070261763705764e-05}
{"train_loss": 0.11884140968322754, "global_step": 134763, "epoch": 1514, "lr": 5.070203795764978e-05}
{"train_loss": 0.1139235720038414, "global_step": 134764, "epoch": 1514, "lr": 5.070145827814756e-05}
{"train_loss": 0.11086247861385345, "global_step": 134765, "epoch": 1514, "lr": 5.0700878598551015e-05}
{"train_loss": 0.11273574084043503, "global_step": 134766, "epoch": 1514, "lr": 5.070029891886026e-05}
{"train_loss": 0.09112338721752167, "global_step": 134767, "epoch": 1514, "lr": 5.069971923907535e-05}
{"train_loss": 0.11628761142492294, "global_step": 134768, "epoch": 1514, "lr": 5.069913955919636e-05}
{"train_loss": 0.1085314154624939, "global_step": 134769, "epoch": 1514, "lr": 5.0698559879223395e-05}
{"train_loss": 0.15091572701931, "global_step": 134770, "epoch": 1514, "lr": 5.0697980199156514e-05}
{"train_loss": 0.09263915568590164, "global_step": 134771, "epoch": 1514, "lr": 5.0697400518995796e-05}
{"train_loss": 0.14224299788475037, "global_step": 134772, "epoch": 1514, "lr": 5.069682083874132e-05}
{"train_loss": 0.06249408423900604, "global_step": 134773, "epoch": 1514, "lr": 5.069624115839318e-05}
{"train_loss": 0.08582065254449844, "global_step": 134774, "epoch": 1514, "lr": 5.0695661477951415e-05}
{"train_loss": 0.12035122513771057, "global_step": 134775, "epoch": 1514, "lr": 5.0695081797416146e-05}
{"train_loss": 0.11782616376876831, "global_step": 134776, "epoch": 1514, "lr": 5.0694502116787425e-05}
{"train_loss": 0.06925037503242493, "global_step": 134777, "epoch": 1514, "lr": 5.0693922436065334e-05}
{"train_loss": 0.11961128562688828, "global_step": 134778, "epoch": 1514, "lr": 5.069334275524996e-05}
{"train_loss": 0.1409008502960205, "global_step": 134779, "epoch": 1514, "lr": 5.069276307434137e-05}
{"train_loss": 0.08669325709342957, "global_step": 134780, "epoch": 1514, "lr": 5.069218339333964e-05}
{"train_loss": 0.1547427475452423, "global_step": 134781, "epoch": 1514, "lr": 5.0691603712244875e-05}
{"train_loss": 0.16296802461147308, "global_step": 134782, "epoch": 1514, "lr": 5.069102403105711e-05}
{"train_loss": 0.09210092574357986, "global_step": 134783, "epoch": 1514, "lr": 5.0690444349776454e-05}
{"train_loss": 0.09440023452043533, "global_step": 134784, "epoch": 1514, "lr": 5.068986466840299e-05}
{"train_loss": 0.13368280231952667, "global_step": 134785, "epoch": 1514, "lr": 5.068928498693677e-05}
{"train_loss": 0.09661062806844711, "global_step": 134786, "epoch": 1514, "lr": 5.068870530537787e-05}
{"train_loss": 0.11578211188316345, "global_step": 134787, "epoch": 1514, "lr": 5.068812562372639e-05}
{"train_loss": 0.13582822680473328, "global_step": 134788, "epoch": 1514, "lr": 5.068754594198241e-05}
{"train_loss": 0.08602578192949295, "global_step": 134789, "epoch": 1514, "lr": 5.068696626014599e-05}
{"train_loss": 0.06173905357718468, "global_step": 134790, "epoch": 1514, "lr": 5.0686386578217214e-05}
{"train_loss": 0.06780381500720978, "global_step": 134791, "epoch": 1514, "lr": 5.068580689619616e-05}
{"train_loss": 0.12314876168966293, "global_step": 134792, "epoch": 1514, "lr": 5.068522721408292e-05}
{"train_loss": 0.07640842348337173, "global_step": 134793, "epoch": 1514, "lr": 5.068464753187755e-05}
{"train_loss": 0.1396416574716568, "global_step": 134794, "epoch": 1514, "lr": 5.068406784958014e-05}
{"train_loss": 0.17533932626247406, "global_step": 134795, "epoch": 1514, "lr": 5.0683488167190776e-05}
{"train_loss": 0.12326264381408691, "global_step": 134796, "epoch": 1514, "lr": 5.068290848470951e-05}
{"train_loss": 0.1381235420703888, "global_step": 134797, "epoch": 1514, "lr": 5.068232880213644e-05}
{"train_loss": 0.19308708608150482, "global_step": 134798, "epoch": 1514, "lr": 5.068174911947163e-05}
{"train_loss": 0.11049383133649826, "global_step": 134799, "epoch": 1514, "lr": 5.068116943671518e-05}
{"train_loss": 0.08454418182373047, "global_step": 134800, "epoch": 1514, "lr": 5.068058975386716e-05}
{"train_loss": 0.12508688867092133, "global_step": 134801, "epoch": 1514, "lr": 5.068001007092763e-05}
{"train_loss": 0.11792999505996704, "global_step": 134802, "epoch": 1514, "lr": 5.067943038789669e-05}
{"train_loss": 0.11364655196666718, "global_step": 134803, "epoch": 1514, "lr": 5.0678850704774404e-05}
{"train_loss": 0.12733522057533264, "global_step": 134804, "epoch": 1514, "lr": 5.067827102156085e-05}
{"train_loss": 0.1655251532793045, "global_step": 134805, "epoch": 1514, "lr": 5.0677691338256116e-05}
{"train_loss": 0.1289830505847931, "global_step": 134806, "epoch": 1514, "lr": 5.0677111654860274e-05}
{"train_loss": 0.0795319527387619, "global_step": 134807, "epoch": 1514, "lr": 5.0676531971373406e-05}
{"train_loss": 0.14884637296199799, "global_step": 134808, "epoch": 1514, "lr": 5.067595228779558e-05}
{"train_loss": 0.09623397886753082, "global_step": 134809, "epoch": 1514, "lr": 5.067537260412688e-05}
{"train_loss": 0.17666882276535034, "global_step": 134810, "epoch": 1514, "lr": 5.0674792920367396e-05}
{"train_loss": 0.10428786277770996, "global_step": 134811, "epoch": 1514, "lr": 5.0674213236517185e-05}
{"train_loss": 0.13006821274757385, "global_step": 134812, "epoch": 1514, "lr": 5.0673633552576335e-05}
{"train_loss": 0.06551387906074524, "global_step": 134813, "epoch": 1514, "lr": 5.067305386854493e-05}
{"train_loss": 0.14233529567718506, "global_step": 134814, "epoch": 1514, "lr": 5.0672474184423036e-05}
{"train_loss": 0.1581675410270691, "global_step": 134815, "epoch": 1514, "lr": 5.067189450021073e-05}
{"train_loss": 0.12860989570617676, "global_step": 134816, "epoch": 1514, "lr": 5.06713148159081e-05}
{"train_loss": 0.17443393170833588, "global_step": 134817, "epoch": 1514, "lr": 5.0670735131515226e-05}
{"train_loss": 0.19243334233760834, "global_step": 134818, "epoch": 1514, "lr": 5.067015544703217e-05}
{"train_loss": 0.17186245322227478, "global_step": 134819, "epoch": 1514, "lr": 5.0669575762459023e-05}
{"train_loss": 0.12399802356958389, "global_step": 134820, "epoch": 1514, "lr": 5.066899607779586e-05}
{"train_loss": 0.17512351274490356, "global_step": 134821, "epoch": 1514, "lr": 5.0668416393042764e-05}
{"train_loss": 0.10509848594665527, "global_step": 134822, "epoch": 1514, "lr": 5.0667836708199803e-05}
{"train_loss": 0.13600462675094604, "global_step": 134823, "epoch": 1514, "lr": 5.066725702326706e-05}
{"train_loss": 0.10181219130754471, "global_step": 134824, "epoch": 1514, "lr": 5.0666677338244614e-05}
{"train_loss": 0.17767708003520966, "global_step": 134825, "epoch": 1514, "lr": 5.066609765313254e-05}
{"train_loss": 0.09981435537338257, "global_step": 134826, "epoch": 1514, "lr": 5.066551796793092e-05}
{"train_loss": 0.07510081678628922, "global_step": 134827, "epoch": 1514, "lr": 5.066493828263983e-05}
{"train_loss": 0.13170000910758972, "global_step": 134828, "epoch": 1514, "lr": 5.0664358597259344e-05}
{"train_loss": 0.1102643832564354, "global_step": 134829, "epoch": 1514, "lr": 5.066377891178954e-05}
{"train_loss": 0.14750324189662933, "global_step": 134830, "epoch": 1514, "lr": 5.0663199226230496e-05}
{"train_loss": 0.10064768046140671, "global_step": 134831, "epoch": 1514, "lr": 5.06626195405823e-05}
{"train_loss": 0.10327280312776566, "global_step": 134832, "epoch": 1514, "lr": 5.066203985484502e-05}
{"train_loss": 0.17620684206485748, "global_step": 134833, "epoch": 1514, "lr": 5.0661460169018736e-05}
{"train_loss": 0.12345401369286386, "global_step": 134834, "epoch": 1514, "lr": 5.0660880483103534e-05, "val_loss": 5.70078182220459}
{"train_loss": 0.0836682841181755, "global_step": 134835, "epoch": 1515, "lr": 5.066030079709948e-05}
{"train_loss": 0.1806730479001999, "global_step": 134836, "epoch": 1515, "lr": 5.065972111100665e-05}
{"train_loss": 0.17066128551959991, "global_step": 134837, "epoch": 1515, "lr": 5.065914142482514e-05}
{"train_loss": 0.12222688645124435, "global_step": 134838, "epoch": 1515, "lr": 5.065856173855501e-05}
{"train_loss": 0.1459064930677414, "global_step": 134839, "epoch": 1515, "lr": 5.065798205219635e-05}
{"train_loss": 0.14836496114730835, "global_step": 134840, "epoch": 1515, "lr": 5.065740236574923e-05}
{"train_loss": 0.12278339266777039, "global_step": 134841, "epoch": 1515, "lr": 5.0656822679213724e-05}
{"train_loss": 0.11691561341285706, "global_step": 134842, "epoch": 1515, "lr": 5.065624299258992e-05}
{"train_loss": 0.09145209938287735, "global_step": 134843, "epoch": 1515, "lr": 5.06556633058779e-05}
{"train_loss": 0.08126808702945709, "global_step": 134844, "epoch": 1515, "lr": 5.065508361907773e-05}
{"train_loss": 0.12241965532302856, "global_step": 134845, "epoch": 1515, "lr": 5.065450393218948e-05}
{"train_loss": 0.12586982548236847, "global_step": 134846, "epoch": 1515, "lr": 5.0653924245213245e-05}
{"train_loss": 0.13239745795726776, "global_step": 134847, "epoch": 1515, "lr": 5.065334455814911e-05}
{"train_loss": 0.16938719153404236, "global_step": 134848, "epoch": 1515, "lr": 5.065276487099713e-05}
{"train_loss": 0.13838285207748413, "global_step": 134849, "epoch": 1515, "lr": 5.0652185183757396e-05}
{"train_loss": 0.1047879233956337, "global_step": 134850, "epoch": 1515, "lr": 5.065160549642998e-05}
{"train_loss": 0.11110903322696686, "global_step": 134851, "epoch": 1515, "lr": 5.065102580901497e-05}
{"train_loss": 0.10719557106494904, "global_step": 134852, "epoch": 1515, "lr": 5.065044612151243e-05}
{"train_loss": 0.16873542964458466, "global_step": 134853, "epoch": 1515, "lr": 5.0649866433922454e-05}
{"train_loss": 0.13241375982761383, "global_step": 134854, "epoch": 1515, "lr": 5.0649286746245105e-05}
{"train_loss": 0.0978279858827591, "global_step": 134855, "epoch": 1515, "lr": 5.064870705848047e-05}
{"train_loss": 0.11953835189342499, "global_step": 134856, "epoch": 1515, "lr": 5.064812737062863e-05}
{"train_loss": 0.09774331003427505, "global_step": 134857, "epoch": 1515, "lr": 5.0647547682689644e-05}
{"train_loss": 0.16503271460533142, "global_step": 134858, "epoch": 1515, "lr": 5.064696799466362e-05}
{"train_loss": 0.1252942681312561, "global_step": 134859, "epoch": 1515, "lr": 5.0646388306550606e-05}
{"train_loss": 0.10527058690786362, "global_step": 134860, "epoch": 1515, "lr": 5.06458086183507e-05}
{"train_loss": 0.12160055339336395, "global_step": 134861, "epoch": 1515, "lr": 5.064522893006396e-05}
{"train_loss": 0.08968798071146011, "global_step": 134862, "epoch": 1515, "lr": 5.064464924169049e-05}
{"train_loss": 0.09005232155323029, "global_step": 134863, "epoch": 1515, "lr": 5.0644069553230354e-05}
{"train_loss": 0.0913274884223938, "global_step": 134864, "epoch": 1515, "lr": 5.0643489864683624e-05}
{"train_loss": 0.11250030994415283, "global_step": 134865, "epoch": 1515, "lr": 5.0642910176050394e-05}
{"train_loss": 0.1290677785873413, "global_step": 134866, "epoch": 1515, "lr": 5.0642330487330726e-05}
{"train_loss": 0.12545496225357056, "global_step": 134867, "epoch": 1515, "lr": 5.06417507985247e-05}
{"train_loss": 0.12319307774305344, "global_step": 134868, "epoch": 1515, "lr": 5.06411711096324e-05}
{"train_loss": 0.07831072062253952, "global_step": 134869, "epoch": 1515, "lr": 5.0640591420653915e-05}
{"train_loss": 0.1769411414861679, "global_step": 134870, "epoch": 1515, "lr": 5.0640011731589295e-05}
{"train_loss": 0.11393702775239944, "global_step": 134871, "epoch": 1515, "lr": 5.063943204243864e-05}
{"train_loss": 0.12417251616716385, "global_step": 134872, "epoch": 1515, "lr": 5.063885235320203e-05}
{"train_loss": 0.12996245920658112, "global_step": 134873, "epoch": 1515, "lr": 5.063827266387952e-05}
{"train_loss": 0.1003047525882721, "global_step": 134874, "epoch": 1515, "lr": 5.0637692974471205e-05}
{"train_loss": 0.13744868338108063, "global_step": 134875, "epoch": 1515, "lr": 5.0637113284977166e-05}
{"train_loss": 0.13106857240200043, "global_step": 134876, "epoch": 1515, "lr": 5.0636533595397476e-05}
{"train_loss": 0.14486318826675415, "global_step": 134877, "epoch": 1515, "lr": 5.06359539057322e-05}
{"train_loss": 0.12517718970775604, "global_step": 134878, "epoch": 1515, "lr": 5.063537421598144e-05}
{"train_loss": 0.15154069662094116, "global_step": 134879, "epoch": 1515, "lr": 5.0634794526145244e-05}
{"train_loss": 0.07058219611644745, "global_step": 134880, "epoch": 1515, "lr": 5.0634214836223725e-05}
{"train_loss": 0.08347517251968384, "global_step": 134881, "epoch": 1515, "lr": 5.063363514621694e-05}
{"train_loss": 0.12580446898937225, "global_step": 134882, "epoch": 1515, "lr": 5.063305545612497e-05}
{"train_loss": 0.14422135055065155, "global_step": 134883, "epoch": 1515, "lr": 5.0632475765947884e-05}
{"train_loss": 0.10135452449321747, "global_step": 134884, "epoch": 1515, "lr": 5.063189607568578e-05}
{"train_loss": 0.15878865122795105, "global_step": 134885, "epoch": 1515, "lr": 5.063131638533872e-05}
{"train_loss": 0.22830535471439362, "global_step": 134886, "epoch": 1515, "lr": 5.06307366949068e-05}
{"train_loss": 0.1137053519487381, "global_step": 134887, "epoch": 1515, "lr": 5.063015700439007e-05}
{"train_loss": 0.16137360036373138, "global_step": 134888, "epoch": 1515, "lr": 5.0629577313788625e-05}
{"train_loss": 0.16958697140216827, "global_step": 134889, "epoch": 1515, "lr": 5.0628997623102547e-05}
{"train_loss": 0.14011551439762115, "global_step": 134890, "epoch": 1515, "lr": 5.062841793233191e-05}
{"train_loss": 0.15897628664970398, "global_step": 134891, "epoch": 1515, "lr": 5.0627838241476775e-05}
{"train_loss": 0.21357806026935577, "global_step": 134892, "epoch": 1515, "lr": 5.062725855053725e-05}
{"train_loss": 0.1209985613822937, "global_step": 134893, "epoch": 1515, "lr": 5.062667885951339e-05}
{"train_loss": 0.1851947009563446, "global_step": 134894, "epoch": 1515, "lr": 5.0626099168405285e-05}
{"train_loss": 0.11535549908876419, "global_step": 134895, "epoch": 1515, "lr": 5.062551947721301e-05}
{"train_loss": 0.13611841201782227, "global_step": 134896, "epoch": 1515, "lr": 5.0624939785936634e-05}
{"train_loss": 0.1222790852189064, "global_step": 134897, "epoch": 1515, "lr": 5.062436009457625e-05}
{"train_loss": 0.14391939342021942, "global_step": 134898, "epoch": 1515, "lr": 5.062378040313193e-05}
{"train_loss": 0.1759544163942337, "global_step": 134899, "epoch": 1515, "lr": 5.0623200711603736e-05}
{"train_loss": 0.1706613004207611, "global_step": 134900, "epoch": 1515, "lr": 5.0622621019991776e-05}
{"train_loss": 0.1385490447282791, "global_step": 134901, "epoch": 1515, "lr": 5.06220413282961e-05}
{"train_loss": 0.07865671068429947, "global_step": 134902, "epoch": 1515, "lr": 5.062146163651681e-05}
{"train_loss": 0.14976836740970612, "global_step": 134903, "epoch": 1515, "lr": 5.0620881944653965e-05}
{"train_loss": 0.14377248287200928, "global_step": 134904, "epoch": 1515, "lr": 5.062030225270765e-05}
{"train_loss": 0.08725668489933014, "global_step": 134905, "epoch": 1515, "lr": 5.061972256067794e-05}
{"train_loss": 0.10841663926839828, "global_step": 134906, "epoch": 1515, "lr": 5.061914286856493e-05}
{"train_loss": 0.13877221941947937, "global_step": 134907, "epoch": 1515, "lr": 5.0618563176368674e-05}
{"train_loss": 0.164028137922287, "global_step": 134908, "epoch": 1515, "lr": 5.0617983484089245e-05}
{"train_loss": 0.10069610178470612, "global_step": 134909, "epoch": 1515, "lr": 5.061740379172676e-05}
{"train_loss": 0.11723043024539948, "global_step": 134910, "epoch": 1515, "lr": 5.061682409928126e-05}
{"train_loss": 0.07279261946678162, "global_step": 134911, "epoch": 1515, "lr": 5.061624440675283e-05}
{"train_loss": 0.10592126101255417, "global_step": 134912, "epoch": 1515, "lr": 5.061566471414157e-05}
{"train_loss": 0.1014108955860138, "global_step": 134913, "epoch": 1515, "lr": 5.061508502144752e-05}
{"train_loss": 0.14251859486103058, "global_step": 134914, "epoch": 1515, "lr": 5.06145053286708e-05}
{"train_loss": 0.14610402286052704, "global_step": 134915, "epoch": 1515, "lr": 5.061392563581147e-05}
{"train_loss": 0.26561078429222107, "global_step": 134916, "epoch": 1515, "lr": 5.0613345942869584e-05}
{"train_loss": 0.11172335594892502, "global_step": 134917, "epoch": 1515, "lr": 5.0612766249845254e-05}
{"train_loss": 0.1537591814994812, "global_step": 134918, "epoch": 1515, "lr": 5.061218655673855e-05}
{"train_loss": 0.12200500816106796, "global_step": 134919, "epoch": 1515, "lr": 5.061160686354953e-05}
{"train_loss": 0.08955077826976776, "global_step": 134920, "epoch": 1515, "lr": 5.06110271702783e-05}
{"train_loss": 0.12039405852556229, "global_step": 134921, "epoch": 1515, "lr": 5.061044747692492e-05}
{"train_loss": 0.14294284582138062, "global_step": 134922, "epoch": 1515, "lr": 5.060986778348947e-05}
{"train_loss": 0.12966453242167997, "global_step": 134923, "epoch": 1515, "lr": 5.060928808997203e-05, "val_loss": 5.585501670837402, "train_action_mse_error": 19.768936157226562}
{"train_loss": 0.059972841292619705, "global_step": 134924, "epoch": 1516, "lr": 5.0608708396372684e-05}
{"train_loss": 0.14617781341075897, "global_step": 134925, "epoch": 1516, "lr": 5.0608128702691503e-05}
{"train_loss": 0.12450480461120605, "global_step": 134926, "epoch": 1516, "lr": 5.060754900892857e-05}
{"train_loss": 0.15425939857959747, "global_step": 134927, "epoch": 1516, "lr": 5.060696931508395e-05}
{"train_loss": 0.07745736837387085, "global_step": 134928, "epoch": 1516, "lr": 5.0606389621157735e-05}
{"train_loss": 0.10314767062664032, "global_step": 134929, "epoch": 1516, "lr": 5.0605809927150006e-05}
{"train_loss": 0.11171762645244598, "global_step": 134930, "epoch": 1516, "lr": 5.060523023306082e-05}
{"train_loss": 0.10163021087646484, "global_step": 134931, "epoch": 1516, "lr": 5.0604650538890266e-05}
{"train_loss": 0.1720995455980301, "global_step": 134932, "epoch": 1516, "lr": 5.060407084463843e-05}
{"train_loss": 0.11733242869377136, "global_step": 134933, "epoch": 1516, "lr": 5.060349115030539e-05}
{"train_loss": 0.12614335119724274, "global_step": 134934, "epoch": 1516, "lr": 5.0602911455891225e-05}
{"train_loss": 0.1181902065873146, "global_step": 134935, "epoch": 1516, "lr": 5.060233176139599e-05}
{"train_loss": 0.11615981161594391, "global_step": 134936, "epoch": 1516, "lr": 5.060175206681978e-05}
{"train_loss": 0.13457727432250977, "global_step": 134937, "epoch": 1516, "lr": 5.0601172372162685e-05}
{"train_loss": 0.17536108195781708, "global_step": 134938, "epoch": 1516, "lr": 5.060059267742476e-05}
{"train_loss": 0.16978271305561066, "global_step": 134939, "epoch": 1516, "lr": 5.0600012982606096e-05}
{"train_loss": 0.07272840291261673, "global_step": 134940, "epoch": 1516, "lr": 5.059943328770677e-05}
{"train_loss": 0.10085386782884598, "global_step": 134941, "epoch": 1516, "lr": 5.059885359272685e-05}
{"train_loss": 0.07497912645339966, "global_step": 134942, "epoch": 1516, "lr": 5.0598273897666414e-05}
{"train_loss": 0.10164782404899597, "global_step": 134943, "epoch": 1516, "lr": 5.059769420252556e-05}
{"train_loss": 0.07898717373609543, "global_step": 134944, "epoch": 1516, "lr": 5.059711450730436e-05}
{"train_loss": 0.1554071456193924, "global_step": 134945, "epoch": 1516, "lr": 5.059653481200287e-05}
{"train_loss": 0.15608105063438416, "global_step": 134946, "epoch": 1516, "lr": 5.059595511662119e-05}
{"train_loss": 0.1259496957063675, "global_step": 134947, "epoch": 1516, "lr": 5.059537542115938e-05}
{"train_loss": 0.12665769457817078, "global_step": 134948, "epoch": 1516, "lr": 5.059479572561755e-05}
{"train_loss": 0.1249404177069664, "global_step": 134949, "epoch": 1516, "lr": 5.0594216029995735e-05}
{"train_loss": 0.10125542432069778, "global_step": 134950, "epoch": 1516, "lr": 5.0593636334294036e-05}
{"train_loss": 0.10961666703224182, "global_step": 134951, "epoch": 1516, "lr": 5.0593056638512546e-05}
{"train_loss": 0.07990977168083191, "global_step": 134952, "epoch": 1516, "lr": 5.059247694265132e-05}
{"train_loss": 0.14153729379177094, "global_step": 134953, "epoch": 1516, "lr": 5.059189724671045e-05}
{"train_loss": 0.0979500338435173, "global_step": 134954, "epoch": 1516, "lr": 5.059131755069e-05}
{"train_loss": 0.08535141497850418, "global_step": 134955, "epoch": 1516, "lr": 5.059073785459005e-05}
{"train_loss": 0.18891772627830505, "global_step": 134956, "epoch": 1516, "lr": 5.059015815841069e-05}
{"train_loss": 0.09718353301286697, "global_step": 134957, "epoch": 1516, "lr": 5.0589578462151985e-05}
{"train_loss": 0.21323443949222565, "global_step": 134958, "epoch": 1516, "lr": 5.058899876581402e-05}
{"train_loss": 0.10890055447816849, "global_step": 134959, "epoch": 1516, "lr": 5.058841906939688e-05}
{"train_loss": 0.0939079225063324, "global_step": 134960, "epoch": 1516, "lr": 5.058783937290063e-05}
{"train_loss": 0.13609719276428223, "global_step": 134961, "epoch": 1516, "lr": 5.058725967632534e-05}
{"train_loss": 0.07807581126689911, "global_step": 134962, "epoch": 1516, "lr": 5.0586679979671114e-05}
{"train_loss": 0.1494305431842804, "global_step": 134963, "epoch": 1516, "lr": 5.058610028293801e-05}
{"train_loss": 0.09057988971471786, "global_step": 134964, "epoch": 1516, "lr": 5.058552058612611e-05}
{"train_loss": 0.11115257441997528, "global_step": 134965, "epoch": 1516, "lr": 5.05849408892355e-05}
{"train_loss": 0.11210794746875763, "global_step": 134966, "epoch": 1516, "lr": 5.058436119226625e-05}
{"train_loss": 0.13464882969856262, "global_step": 134967, "epoch": 1516, "lr": 5.058378149521843e-05}
{"train_loss": 0.0613589771091938, "global_step": 134968, "epoch": 1516, "lr": 5.0583201798092125e-05}
{"train_loss": 0.12976236641407013, "global_step": 134969, "epoch": 1516, "lr": 5.0582622100887424e-05}
{"train_loss": 0.17817293107509613, "global_step": 134970, "epoch": 1516, "lr": 5.05820424036044e-05}
{"train_loss": 0.10955848544836044, "global_step": 134971, "epoch": 1516, "lr": 5.0581462706243135e-05}
{"train_loss": 0.12926799058914185, "global_step": 134972, "epoch": 1516, "lr": 5.058088300880369e-05}
{"train_loss": 0.10939077287912369, "global_step": 134973, "epoch": 1516, "lr": 5.058030331128616e-05}
{"train_loss": 0.14055941998958588, "global_step": 134974, "epoch": 1516, "lr": 5.057972361369061e-05}
{"train_loss": 0.13589385151863098, "global_step": 134975, "epoch": 1516, "lr": 5.0579143916017116e-05}
{"train_loss": 0.13427886366844177, "global_step": 134976, "epoch": 1516, "lr": 5.0578564218265766e-05}
{"train_loss": 0.1265508383512497, "global_step": 134977, "epoch": 1516, "lr": 5.057798452043664e-05}
{"train_loss": 0.0865178108215332, "global_step": 134978, "epoch": 1516, "lr": 5.0577404822529814e-05}
{"train_loss": 0.10122082382440567, "global_step": 134979, "epoch": 1516, "lr": 5.0576825124545355e-05}
{"train_loss": 0.12683595716953278, "global_step": 134980, "epoch": 1516, "lr": 5.057624542648335e-05}
{"train_loss": 0.11474085599184036, "global_step": 134981, "epoch": 1516, "lr": 5.057566572834388e-05}
{"train_loss": 0.14057736098766327, "global_step": 134982, "epoch": 1516, "lr": 5.0575086030127015e-05}
{"train_loss": 0.14761114120483398, "global_step": 134983, "epoch": 1516, "lr": 5.057450633183284e-05}
{"train_loss": 0.14466026425361633, "global_step": 134984, "epoch": 1516, "lr": 5.057392663346142e-05}
{"train_loss": 0.08415487408638, "global_step": 134985, "epoch": 1516, "lr": 5.0573346935012855e-05}
{"train_loss": 0.0983567014336586, "global_step": 134986, "epoch": 1516, "lr": 5.0572767236487195e-05}
{"train_loss": 0.10723381489515305, "global_step": 134987, "epoch": 1516, "lr": 5.0572187537884544e-05}
{"train_loss": 0.17380988597869873, "global_step": 134988, "epoch": 1516, "lr": 5.057160783920497e-05}
{"train_loss": 0.1468971222639084, "global_step": 134989, "epoch": 1516, "lr": 5.057102814044855e-05}
{"train_loss": 0.11326342821121216, "global_step": 134990, "epoch": 1516, "lr": 5.0570448441615356e-05}
{"train_loss": 0.08718064427375793, "global_step": 134991, "epoch": 1516, "lr": 5.0569868742705496e-05}
{"train_loss": 0.04454159736633301, "global_step": 134992, "epoch": 1516, "lr": 5.0569289043719e-05}
{"train_loss": 0.1761663258075714, "global_step": 134993, "epoch": 1516, "lr": 5.0568709344655986e-05}
{"train_loss": 0.09433260560035706, "global_step": 134994, "epoch": 1516, "lr": 5.0568129645516505e-05}
{"train_loss": 0.12129681557416916, "global_step": 134995, "epoch": 1516, "lr": 5.056754994630064e-05}
{"train_loss": 0.08044329285621643, "global_step": 134996, "epoch": 1516, "lr": 5.0566970247008485e-05}
{"train_loss": 0.11372838169336319, "global_step": 134997, "epoch": 1516, "lr": 5.056639054764011e-05}
{"train_loss": 0.10564450174570084, "global_step": 134998, "epoch": 1516, "lr": 5.056581084819558e-05}
{"train_loss": 0.11605461686849594, "global_step": 134999, "epoch": 1516, "lr": 5.0565231148674983e-05}
{"train_loss": 0.11177215725183487, "global_step": 135000, "epoch": 1516, "lr": 5.056465144907842e-05}
{"train_loss": 0.15452462434768677, "global_step": 135001, "epoch": 1516, "lr": 5.0564071749405915e-05}
{"train_loss": 0.1428810954093933, "global_step": 135002, "epoch": 1516, "lr": 5.05634920496576e-05}
{"train_loss": 0.1028391420841217, "global_step": 135003, "epoch": 1516, "lr": 5.056291234983353e-05}
{"train_loss": 0.10866990685462952, "global_step": 135004, "epoch": 1516, "lr": 5.056233264993375e-05}
{"train_loss": 0.125135600566864, "global_step": 135005, "epoch": 1516, "lr": 5.056175294995841e-05}
{"train_loss": 0.08998428285121918, "global_step": 135006, "epoch": 1516, "lr": 5.0561173249907536e-05}
{"train_loss": 0.07730092108249664, "global_step": 135007, "epoch": 1516, "lr": 5.056059354978121e-05}
{"train_loss": 0.09709402173757553, "global_step": 135008, "epoch": 1516, "lr": 5.0560013849579534e-05}
{"train_loss": 0.1243453174829483, "global_step": 135009, "epoch": 1516, "lr": 5.055943414930256e-05}
{"train_loss": 0.0989103838801384, "global_step": 135010, "epoch": 1516, "lr": 5.055885444895039e-05}
{"train_loss": 0.13705407083034515, "global_step": 135011, "epoch": 1516, "lr": 5.055827474852309e-05}
{"train_loss": 0.11835576485047179, "global_step": 135012, "epoch": 1516, "lr": 5.055769504802073e-05, "val_loss": 5.479670524597168}
{"train_loss": 0.1345723420381546, "global_step": 135013, "epoch": 1517, "lr": 5.0557115347443386e-05}
{"train_loss": 0.12110666185617447, "global_step": 135014, "epoch": 1517, "lr": 5.055653564679116e-05}
{"train_loss": 0.087786003947258, "global_step": 135015, "epoch": 1517, "lr": 5.05559559460641e-05}
{"train_loss": 0.1465223729610443, "global_step": 135016, "epoch": 1517, "lr": 5.055537624526231e-05}
{"train_loss": 0.12226731330156326, "global_step": 135017, "epoch": 1517, "lr": 5.0554796544385854e-05}
{"train_loss": 0.08528763055801392, "global_step": 135018, "epoch": 1517, "lr": 5.05542168434348e-05}
{"train_loss": 0.13543473184108734, "global_step": 135019, "epoch": 1517, "lr": 5.055363714240925e-05}
{"train_loss": 0.1307673305273056, "global_step": 135020, "epoch": 1517, "lr": 5.055305744130927e-05}
{"train_loss": 0.12299440801143646, "global_step": 135021, "epoch": 1517, "lr": 5.055247774013493e-05}
{"train_loss": 0.113906629383564, "global_step": 135022, "epoch": 1517, "lr": 5.0551898038886314e-05}
{"train_loss": 0.1042669489979744, "global_step": 135023, "epoch": 1517, "lr": 5.0551318337563515e-05}
{"train_loss": 0.15284225344657898, "global_step": 135024, "epoch": 1517, "lr": 5.055073863616658e-05}
{"train_loss": 0.06300665438175201, "global_step": 135025, "epoch": 1517, "lr": 5.055015893469563e-05}
{"train_loss": 0.07843330502510071, "global_step": 135026, "epoch": 1517, "lr": 5.05495792331507e-05}
{"train_loss": 0.12867027521133423, "global_step": 135027, "epoch": 1517, "lr": 5.0548999531531894e-05}
{"train_loss": 0.12493737787008286, "global_step": 135028, "epoch": 1517, "lr": 5.054841982983928e-05}
{"train_loss": 0.07263653725385666, "global_step": 135029, "epoch": 1517, "lr": 5.0547840128072934e-05}
{"train_loss": 0.12032138556241989, "global_step": 135030, "epoch": 1517, "lr": 5.0547260426232956e-05}
{"train_loss": 0.16771164536476135, "global_step": 135031, "epoch": 1517, "lr": 5.054668072431938e-05}
{"train_loss": 0.15336254239082336, "global_step": 135032, "epoch": 1517, "lr": 5.0546101022332325e-05}
{"train_loss": 0.10614828020334244, "global_step": 135033, "epoch": 1517, "lr": 5.054552132027185e-05}
{"train_loss": 0.1361340433359146, "global_step": 135034, "epoch": 1517, "lr": 5.054494161813803e-05}
{"train_loss": 0.11988945305347443, "global_step": 135035, "epoch": 1517, "lr": 5.054436191593096e-05}
{"train_loss": 0.08816167712211609, "global_step": 135036, "epoch": 1517, "lr": 5.0543782213650704e-05}
{"train_loss": 0.14119558036327362, "global_step": 135037, "epoch": 1517, "lr": 5.054320251129734e-05}
{"train_loss": 0.13748100399971008, "global_step": 135038, "epoch": 1517, "lr": 5.054262280887094e-05}
{"train_loss": 0.17791400849819183, "global_step": 135039, "epoch": 1517, "lr": 5.0542043106371606e-05}
{"train_loss": 0.09599263221025467, "global_step": 135040, "epoch": 1517, "lr": 5.0541463403799386e-05}
{"train_loss": 0.11155951768159866, "global_step": 135041, "epoch": 1517, "lr": 5.0540883701154374e-05}
{"train_loss": 0.13866494596004486, "global_step": 135042, "epoch": 1517, "lr": 5.0540303998436654e-05}
{"train_loss": 0.15541110932826996, "global_step": 135043, "epoch": 1517, "lr": 5.05397242956463e-05}
{"train_loss": 0.13121087849140167, "global_step": 135044, "epoch": 1517, "lr": 5.053914459278338e-05}
{"train_loss": 0.13273322582244873, "global_step": 135045, "epoch": 1517, "lr": 5.053856488984798e-05}
{"train_loss": 0.09947872906923294, "global_step": 135046, "epoch": 1517, "lr": 5.053798518684019e-05}
{"train_loss": 0.14710350334644318, "global_step": 135047, "epoch": 1517, "lr": 5.053740548376006e-05}
{"train_loss": 0.12319200485944748, "global_step": 135048, "epoch": 1517, "lr": 5.053682578060768e-05}
{"train_loss": 0.0944952517747879, "global_step": 135049, "epoch": 1517, "lr": 5.053624607738315e-05}
{"train_loss": 0.09734206646680832, "global_step": 135050, "epoch": 1517, "lr": 5.053566637408651e-05}
{"train_loss": 0.10097270458936691, "global_step": 135051, "epoch": 1517, "lr": 5.0535086670717854e-05}
{"train_loss": 0.09691298007965088, "global_step": 135052, "epoch": 1517, "lr": 5.0534506967277275e-05}
{"train_loss": 0.13066333532333374, "global_step": 135053, "epoch": 1517, "lr": 5.0533927263764835e-05}
{"train_loss": 0.12612470984458923, "global_step": 135054, "epoch": 1517, "lr": 5.053334756018061e-05}
{"train_loss": 0.10461102426052094, "global_step": 135055, "epoch": 1517, "lr": 5.0532767856524685e-05}
{"train_loss": 0.1341352015733719, "global_step": 135056, "epoch": 1517, "lr": 5.053218815279713e-05}
{"train_loss": 0.14159537851810455, "global_step": 135057, "epoch": 1517, "lr": 5.0531608448998034e-05}
{"train_loss": 0.11514151841402054, "global_step": 135058, "epoch": 1517, "lr": 5.053102874512746e-05}
{"train_loss": 0.07352958619594574, "global_step": 135059, "epoch": 1517, "lr": 5.0530449041185514e-05}
{"train_loss": 0.19883257150650024, "global_step": 135060, "epoch": 1517, "lr": 5.052986933717224e-05}
{"train_loss": 0.07661371678113937, "global_step": 135061, "epoch": 1517, "lr": 5.052928963308774e-05}
{"train_loss": 0.15374459326267242, "global_step": 135062, "epoch": 1517, "lr": 5.052870992893208e-05}
{"train_loss": 0.13745231926441193, "global_step": 135063, "epoch": 1517, "lr": 5.052813022470535e-05}
{"train_loss": 0.1314413994550705, "global_step": 135064, "epoch": 1517, "lr": 5.052755052040761e-05}
{"train_loss": 0.141866996884346, "global_step": 135065, "epoch": 1517, "lr": 5.0526970816038945e-05}
{"train_loss": 0.09897202253341675, "global_step": 135066, "epoch": 1517, "lr": 5.052639111159945e-05}
{"train_loss": 0.1203526183962822, "global_step": 135067, "epoch": 1517, "lr": 5.052581140708917e-05}
{"train_loss": 0.10340224206447601, "global_step": 135068, "epoch": 1517, "lr": 5.0525231702508215e-05}
{"train_loss": 0.09408332407474518, "global_step": 135069, "epoch": 1517, "lr": 5.052465199785665e-05}
{"train_loss": 0.09441777318716049, "global_step": 135070, "epoch": 1517, "lr": 5.052407229313455e-05}
{"train_loss": 0.1081731989979744, "global_step": 135071, "epoch": 1517, "lr": 5.052349258834198e-05}
{"train_loss": 0.10687591135501862, "global_step": 135072, "epoch": 1517, "lr": 5.0522912883479044e-05}
{"train_loss": 0.10516368597745895, "global_step": 135073, "epoch": 1517, "lr": 5.052233317854581e-05}
{"train_loss": 0.1337527185678482, "global_step": 135074, "epoch": 1517, "lr": 5.0521753473542354e-05}
{"train_loss": 0.0925811305642128, "global_step": 135075, "epoch": 1517, "lr": 5.052117376846876e-05}
{"train_loss": 0.08806774765253067, "global_step": 135076, "epoch": 1517, "lr": 5.052059406332508e-05}
{"train_loss": 0.12034826725721359, "global_step": 135077, "epoch": 1517, "lr": 5.0520014358111424e-05}
{"train_loss": 0.11396592110395432, "global_step": 135078, "epoch": 1517, "lr": 5.051943465282786e-05}
{"train_loss": 0.12081839889287949, "global_step": 135079, "epoch": 1517, "lr": 5.0518854947474456e-05}
{"train_loss": 0.1298145055770874, "global_step": 135080, "epoch": 1517, "lr": 5.051827524205132e-05}
{"train_loss": 0.1413717120885849, "global_step": 135081, "epoch": 1517, "lr": 5.051769553655849e-05}
{"train_loss": 0.06047596409916878, "global_step": 135082, "epoch": 1517, "lr": 5.0517115830996065e-05}
{"train_loss": 0.0958457663655281, "global_step": 135083, "epoch": 1517, "lr": 5.0516536125364124e-05}
{"train_loss": 0.08239757269620895, "global_step": 135084, "epoch": 1517, "lr": 5.051595641966273e-05}
{"train_loss": 0.13332664966583252, "global_step": 135085, "epoch": 1517, "lr": 5.0515376713891994e-05}
{"train_loss": 0.14285282790660858, "global_step": 135086, "epoch": 1517, "lr": 5.051479700805195e-05}
{"train_loss": 0.10320878773927689, "global_step": 135087, "epoch": 1517, "lr": 5.051421730214271e-05}
{"train_loss": 0.10951582342386246, "global_step": 135088, "epoch": 1517, "lr": 5.051363759616435e-05}
{"train_loss": 0.09001024812459946, "global_step": 135089, "epoch": 1517, "lr": 5.0513057890116924e-05}
{"train_loss": 0.10398274660110474, "global_step": 135090, "epoch": 1517, "lr": 5.051247818400052e-05}
{"train_loss": 0.09563294053077698, "global_step": 135091, "epoch": 1517, "lr": 5.051189847781523e-05}
{"train_loss": 0.15221774578094482, "global_step": 135092, "epoch": 1517, "lr": 5.0511318771561114e-05}
{"train_loss": 0.1594986915588379, "global_step": 135093, "epoch": 1517, "lr": 5.051073906523826e-05}
{"train_loss": 0.1546253263950348, "global_step": 135094, "epoch": 1517, "lr": 5.0510159358846746e-05}
{"train_loss": 0.1614684760570526, "global_step": 135095, "epoch": 1517, "lr": 5.050957965238663e-05}
{"train_loss": 0.0737898200750351, "global_step": 135096, "epoch": 1517, "lr": 5.0508999945858014e-05}
{"train_loss": 0.09888091683387756, "global_step": 135097, "epoch": 1517, "lr": 5.050842023926098e-05}
{"train_loss": 0.16534891724586487, "global_step": 135098, "epoch": 1517, "lr": 5.050784053259559e-05}
{"train_loss": 0.12977935373783112, "global_step": 135099, "epoch": 1517, "lr": 5.050726082586192e-05}
{"train_loss": 0.1290641874074936, "global_step": 135100, "epoch": 1517, "lr": 5.050668111906007e-05}
{"train_loss": 0.11852987859858556, "global_step": 135101, "epoch": 1517, "lr": 5.05061014121901e-05, "val_loss": 5.694600582122803}
{"train_loss": 0.15371274948120117, "global_step": 135102, "epoch": 1518, "lr": 5.050552170525208e-05}
{"train_loss": 0.10395533591508865, "global_step": 135103, "epoch": 1518, "lr": 5.05049419982461e-05}
{"train_loss": 0.09933202713727951, "global_step": 135104, "epoch": 1518, "lr": 5.0504362291172244e-05}
{"train_loss": 0.11551522463560104, "global_step": 135105, "epoch": 1518, "lr": 5.050378258403059e-05}
{"train_loss": 0.15151475369930267, "global_step": 135106, "epoch": 1518, "lr": 5.05032028768212e-05}
{"train_loss": 0.10878399014472961, "global_step": 135107, "epoch": 1518, "lr": 5.0502623169544164e-05}
{"train_loss": 0.09458968043327332, "global_step": 135108, "epoch": 1518, "lr": 5.0502043462199557e-05}
{"train_loss": 0.07692297548055649, "global_step": 135109, "epoch": 1518, "lr": 5.050146375478745e-05}
{"train_loss": 0.1905401051044464, "global_step": 135110, "epoch": 1518, "lr": 5.050088404730793e-05}
{"train_loss": 0.18541757762432098, "global_step": 135111, "epoch": 1518, "lr": 5.0500304339761076e-05}
{"train_loss": 0.10686159133911133, "global_step": 135112, "epoch": 1518, "lr": 5.049972463214695e-05}
{"train_loss": 0.062301650643348694, "global_step": 135113, "epoch": 1518, "lr": 5.0499144924465636e-05}
{"train_loss": 0.13844551146030426, "global_step": 135114, "epoch": 1518, "lr": 5.049856521671723e-05}
{"train_loss": 0.14623381197452545, "global_step": 135115, "epoch": 1518, "lr": 5.04979855089018e-05}
{"train_loss": 0.182169571518898, "global_step": 135116, "epoch": 1518, "lr": 5.049740580101942e-05}
{"train_loss": 0.14188380539417267, "global_step": 135117, "epoch": 1518, "lr": 5.049682609307017e-05}
{"train_loss": 0.16544502973556519, "global_step": 135118, "epoch": 1518, "lr": 5.0496246385054136e-05}
{"train_loss": 0.10243254899978638, "global_step": 135119, "epoch": 1518, "lr": 5.0495666676971376e-05}
{"train_loss": 0.12478222697973251, "global_step": 135120, "epoch": 1518, "lr": 5.049508696882199e-05}
{"train_loss": 0.14599409699440002, "global_step": 135121, "epoch": 1518, "lr": 5.049450726060604e-05}
{"train_loss": 0.13443124294281006, "global_step": 135122, "epoch": 1518, "lr": 5.04939275523236e-05}
{"train_loss": 0.08605873584747314, "global_step": 135123, "epoch": 1518, "lr": 5.049334784397477e-05}
{"train_loss": 0.10118886828422546, "global_step": 135124, "epoch": 1518, "lr": 5.049276813555962e-05}
{"train_loss": 0.09956423938274384, "global_step": 135125, "epoch": 1518, "lr": 5.049218842707821e-05}
{"train_loss": 0.13757120072841644, "global_step": 135126, "epoch": 1518, "lr": 5.0491608718530634e-05}
{"train_loss": 0.12332674115896225, "global_step": 135127, "epoch": 1518, "lr": 5.049102900991698e-05}
{"train_loss": 0.1093197762966156, "global_step": 135128, "epoch": 1518, "lr": 5.049044930123731e-05}
{"train_loss": 0.14351168274879456, "global_step": 135129, "epoch": 1518, "lr": 5.04898695924917e-05}
{"train_loss": 0.13274399936199188, "global_step": 135130, "epoch": 1518, "lr": 5.048928988368022e-05}
{"train_loss": 0.17803898453712463, "global_step": 135131, "epoch": 1518, "lr": 5.0488710174802966e-05}
{"train_loss": 0.11279390752315521, "global_step": 135132, "epoch": 1518, "lr": 5.048813046586002e-05}
{"train_loss": 0.14632944762706757, "global_step": 135133, "epoch": 1518, "lr": 5.048755075685144e-05}
{"train_loss": 0.18259765207767487, "global_step": 135134, "epoch": 1518, "lr": 5.0486971047777334e-05}
{"train_loss": 0.060609787702560425, "global_step": 135135, "epoch": 1518, "lr": 5.048639133863775e-05}
{"train_loss": 0.11560402065515518, "global_step": 135136, "epoch": 1518, "lr": 5.0485811629432765e-05}
{"train_loss": 0.1650213599205017, "global_step": 135137, "epoch": 1518, "lr": 5.048523192016249e-05}
{"train_loss": 0.12349815666675568, "global_step": 135138, "epoch": 1518, "lr": 5.0484652210826965e-05}
{"train_loss": 0.12499313801527023, "global_step": 135139, "epoch": 1518, "lr": 5.0484072501426297e-05}
{"train_loss": 0.1513632833957672, "global_step": 135140, "epoch": 1518, "lr": 5.048349279196054e-05}
{"train_loss": 0.15017704665660858, "global_step": 135141, "epoch": 1518, "lr": 5.0482913082429783e-05}
{"train_loss": 0.10438399016857147, "global_step": 135142, "epoch": 1518, "lr": 5.0482333372834115e-05}
{"train_loss": 0.1424441933631897, "global_step": 135143, "epoch": 1518, "lr": 5.048175366317359e-05}
{"train_loss": 0.16914822161197662, "global_step": 135144, "epoch": 1518, "lr": 5.048117395344831e-05}
{"train_loss": 0.13658151030540466, "global_step": 135145, "epoch": 1518, "lr": 5.048059424365834e-05}
{"train_loss": 0.1678558886051178, "global_step": 135146, "epoch": 1518, "lr": 5.0480014533803764e-05}
{"train_loss": 0.07482588291168213, "global_step": 135147, "epoch": 1518, "lr": 5.047943482388465e-05}
{"train_loss": 0.11683535575866699, "global_step": 135148, "epoch": 1518, "lr": 5.047885511390109e-05}
{"train_loss": 0.12871026992797852, "global_step": 135149, "epoch": 1518, "lr": 5.047827540385314e-05}
{"train_loss": 0.11453427374362946, "global_step": 135150, "epoch": 1518, "lr": 5.047769569374089e-05}
{"train_loss": 0.13387508690357208, "global_step": 135151, "epoch": 1518, "lr": 5.0477115983564425e-05}
{"train_loss": 0.11843641847372055, "global_step": 135152, "epoch": 1518, "lr": 5.0476536273323826e-05}
{"train_loss": 0.13672669231891632, "global_step": 135153, "epoch": 1518, "lr": 5.047595656301916e-05}
{"train_loss": 0.06496191024780273, "global_step": 135154, "epoch": 1518, "lr": 5.04753768526505e-05}
{"train_loss": 0.12570315599441528, "global_step": 135155, "epoch": 1518, "lr": 5.0474797142217935e-05}
{"train_loss": 0.09738660603761673, "global_step": 135156, "epoch": 1518, "lr": 5.047421743172154e-05}
{"train_loss": 0.1147812232375145, "global_step": 135157, "epoch": 1518, "lr": 5.047363772116139e-05}
{"train_loss": 0.12563465535640717, "global_step": 135158, "epoch": 1518, "lr": 5.0473058010537566e-05}
{"train_loss": 0.09795929491519928, "global_step": 135159, "epoch": 1518, "lr": 5.0472478299850146e-05}
{"train_loss": 0.108512744307518, "global_step": 135160, "epoch": 1518, "lr": 5.047189858909921e-05}
{"train_loss": 0.14941947162151337, "global_step": 135161, "epoch": 1518, "lr": 5.0471318878284826e-05}
{"train_loss": 0.21417635679244995, "global_step": 135162, "epoch": 1518, "lr": 5.047073916740709e-05}
{"train_loss": 0.10574512183666229, "global_step": 135163, "epoch": 1518, "lr": 5.047015945646606e-05}
{"train_loss": 0.11618407070636749, "global_step": 135164, "epoch": 1518, "lr": 5.046957974546182e-05}
{"train_loss": 0.11292745172977448, "global_step": 135165, "epoch": 1518, "lr": 5.0469000034394465e-05}
{"train_loss": 0.19164860248565674, "global_step": 135166, "epoch": 1518, "lr": 5.0468420323264054e-05}
{"train_loss": 0.12932884693145752, "global_step": 135167, "epoch": 1518, "lr": 5.046784061207065e-05}
{"train_loss": 0.0913655236363411, "global_step": 135168, "epoch": 1518, "lr": 5.046726090081436e-05}
{"train_loss": 0.08849700540304184, "global_step": 135169, "epoch": 1518, "lr": 5.046668118949527e-05}
{"train_loss": 0.12833663821220398, "global_step": 135170, "epoch": 1518, "lr": 5.046610147811342e-05}
{"train_loss": 0.1375301331281662, "global_step": 135171, "epoch": 1518, "lr": 5.0465521766668925e-05}
{"train_loss": 0.11938899755477905, "global_step": 135172, "epoch": 1518, "lr": 5.0464942055161835e-05}
{"train_loss": 0.1602650284767151, "global_step": 135173, "epoch": 1518, "lr": 5.046436234359224e-05}
{"train_loss": 0.12039878219366074, "global_step": 135174, "epoch": 1518, "lr": 5.0463782631960223e-05}
{"train_loss": 0.1257365494966507, "global_step": 135175, "epoch": 1518, "lr": 5.046320292026585e-05}
{"train_loss": 0.1268492341041565, "global_step": 135176, "epoch": 1518, "lr": 5.0462623208509205e-05}
{"train_loss": 0.09111696481704712, "global_step": 135177, "epoch": 1518, "lr": 5.046204349669037e-05}
{"train_loss": 0.09021066129207611, "global_step": 135178, "epoch": 1518, "lr": 5.0461463784809416e-05}
{"train_loss": 0.12831631302833557, "global_step": 135179, "epoch": 1518, "lr": 5.046088407286642e-05}
{"train_loss": 0.14988461136817932, "global_step": 135180, "epoch": 1518, "lr": 5.0460304360861466e-05}
{"train_loss": 0.13022702932357788, "global_step": 135181, "epoch": 1518, "lr": 5.045972464879464e-05}
{"train_loss": 0.12527430057525635, "global_step": 135182, "epoch": 1518, "lr": 5.0459144936666e-05}
{"train_loss": 0.13554763793945312, "global_step": 135183, "epoch": 1518, "lr": 5.0458565224475626e-05}
{"train_loss": 0.1820526272058487, "global_step": 135184, "epoch": 1518, "lr": 5.0457985512223615e-05}
{"train_loss": 0.04596962034702301, "global_step": 135185, "epoch": 1518, "lr": 5.045740579991004e-05}
{"train_loss": 0.07733327895402908, "global_step": 135186, "epoch": 1518, "lr": 5.045682608753496e-05}
{"train_loss": 0.12454865127801895, "global_step": 135187, "epoch": 1518, "lr": 5.045624637509847e-05}
{"train_loss": 0.1317235231399536, "global_step": 135188, "epoch": 1518, "lr": 5.045566666260064e-05}
{"train_loss": 0.1458800733089447, "global_step": 135189, "epoch": 1518, "lr": 5.045508695004155e-05}
{"train_loss": 0.12575548045923202, "global_step": 135190, "epoch": 1518, "lr": 5.045450723742129e-05, "val_loss": 5.377065181732178}
{"train_loss": 0.09995568543672562, "global_step": 135191, "epoch": 1519, "lr": 5.045392752473992e-05}
{"train_loss": 0.0713546946644783, "global_step": 135192, "epoch": 1519, "lr": 5.045334781199752e-05}
{"train_loss": 0.11724919825792313, "global_step": 135193, "epoch": 1519, "lr": 5.0452768099194184e-05}
{"train_loss": 0.09071537107229233, "global_step": 135194, "epoch": 1519, "lr": 5.045218838632997e-05}
{"train_loss": 0.21195417642593384, "global_step": 135195, "epoch": 1519, "lr": 5.045160867340496e-05}
{"train_loss": 0.11570937931537628, "global_step": 135196, "epoch": 1519, "lr": 5.045102896041924e-05}
{"train_loss": 0.13808131217956543, "global_step": 135197, "epoch": 1519, "lr": 5.04504492473729e-05}
{"train_loss": 0.08057425916194916, "global_step": 135198, "epoch": 1519, "lr": 5.044986953426599e-05}
{"train_loss": 0.0854005515575409, "global_step": 135199, "epoch": 1519, "lr": 5.04492898210986e-05}
{"train_loss": 0.08611167222261429, "global_step": 135200, "epoch": 1519, "lr": 5.044871010787081e-05}
{"train_loss": 0.07153508812189102, "global_step": 135201, "epoch": 1519, "lr": 5.044813039458269e-05}
{"train_loss": 0.15820398926734924, "global_step": 135202, "epoch": 1519, "lr": 5.0447550681234336e-05}
{"train_loss": 0.10213413834571838, "global_step": 135203, "epoch": 1519, "lr": 5.044697096782581e-05}
{"train_loss": 0.15530909597873688, "global_step": 135204, "epoch": 1519, "lr": 5.044639125435719e-05}
{"train_loss": 0.12906110286712646, "global_step": 135205, "epoch": 1519, "lr": 5.044581154082856e-05}
{"train_loss": 0.08685776591300964, "global_step": 135206, "epoch": 1519, "lr": 5.0445231827240004e-05}
{"train_loss": 0.12595625221729279, "global_step": 135207, "epoch": 1519, "lr": 5.044465211359158e-05}
{"train_loss": 0.07616013288497925, "global_step": 135208, "epoch": 1519, "lr": 5.044407239988338e-05}
{"train_loss": 0.09231463819742203, "global_step": 135209, "epoch": 1519, "lr": 5.044349268611549e-05}
{"train_loss": 0.15387266874313354, "global_step": 135210, "epoch": 1519, "lr": 5.044291297228797e-05}
{"train_loss": 0.13729731738567352, "global_step": 135211, "epoch": 1519, "lr": 5.0442333258400906e-05}
{"train_loss": 0.09590096026659012, "global_step": 135212, "epoch": 1519, "lr": 5.0441753544454374e-05}
{"train_loss": 0.09684265404939651, "global_step": 135213, "epoch": 1519, "lr": 5.044117383044846e-05}
{"train_loss": 0.09731321036815643, "global_step": 135214, "epoch": 1519, "lr": 5.044059411638323e-05}
{"train_loss": 0.08833945542573929, "global_step": 135215, "epoch": 1519, "lr": 5.0440014402258776e-05}
{"train_loss": 0.19248029589653015, "global_step": 135216, "epoch": 1519, "lr": 5.043943468807515e-05}
{"train_loss": 0.12968534231185913, "global_step": 135217, "epoch": 1519, "lr": 5.043885497383246e-05}
{"train_loss": 0.10968317836523056, "global_step": 135218, "epoch": 1519, "lr": 5.043827525953078e-05}
{"train_loss": 0.08885788917541504, "global_step": 135219, "epoch": 1519, "lr": 5.043769554517016e-05}
{"train_loss": 0.08095022290945053, "global_step": 135220, "epoch": 1519, "lr": 5.0437115830750704e-05}
{"train_loss": 0.11753475666046143, "global_step": 135221, "epoch": 1519, "lr": 5.043653611627248e-05}
{"train_loss": 0.053189944475889206, "global_step": 135222, "epoch": 1519, "lr": 5.0435956401735586e-05}
{"train_loss": 0.10877634584903717, "global_step": 135223, "epoch": 1519, "lr": 5.0435376687140065e-05}
{"train_loss": 0.10468673706054688, "global_step": 135224, "epoch": 1519, "lr": 5.043479697248602e-05}
{"train_loss": 0.20508381724357605, "global_step": 135225, "epoch": 1519, "lr": 5.043421725777352e-05}
{"train_loss": 0.09211085736751556, "global_step": 135226, "epoch": 1519, "lr": 5.043363754300264e-05}
{"train_loss": 0.16294890642166138, "global_step": 135227, "epoch": 1519, "lr": 5.0433057828173466e-05}
{"train_loss": 0.0852249413728714, "global_step": 135228, "epoch": 1519, "lr": 5.043247811328608e-05}
{"train_loss": 0.0854773297905922, "global_step": 135229, "epoch": 1519, "lr": 5.043189839834055e-05}
{"train_loss": 0.1886746883392334, "global_step": 135230, "epoch": 1519, "lr": 5.043131868333695e-05}
{"train_loss": 0.12596844136714935, "global_step": 135231, "epoch": 1519, "lr": 5.043073896827537e-05}
{"train_loss": 0.11857742816209793, "global_step": 135232, "epoch": 1519, "lr": 5.0430159253155886e-05}
{"train_loss": 0.20856906473636627, "global_step": 135233, "epoch": 1519, "lr": 5.042957953797857e-05}
{"train_loss": 0.09577034413814545, "global_step": 135234, "epoch": 1519, "lr": 5.0428999822743497e-05}
{"train_loss": 0.0888490155339241, "global_step": 135235, "epoch": 1519, "lr": 5.0428420107450745e-05}
{"train_loss": 0.1974770724773407, "global_step": 135236, "epoch": 1519, "lr": 5.04278403921004e-05}
{"train_loss": 0.06659120321273804, "global_step": 135237, "epoch": 1519, "lr": 5.042726067669254e-05}
{"train_loss": 0.11312933266162872, "global_step": 135238, "epoch": 1519, "lr": 5.042668096122725e-05}
{"train_loss": 0.15415184199810028, "global_step": 135239, "epoch": 1519, "lr": 5.0426101245704584e-05}
{"train_loss": 0.1205642893910408, "global_step": 135240, "epoch": 1519, "lr": 5.042552153012464e-05}
{"train_loss": 0.12026909738779068, "global_step": 135241, "epoch": 1519, "lr": 5.042494181448749e-05}
{"train_loss": 0.12256798893213272, "global_step": 135242, "epoch": 1519, "lr": 5.042436209879321e-05}
{"train_loss": 0.09120167046785355, "global_step": 135243, "epoch": 1519, "lr": 5.0423782383041886e-05}
{"train_loss": 0.09456315636634827, "global_step": 135244, "epoch": 1519, "lr": 5.042320266723358e-05}
{"train_loss": 0.09106162190437317, "global_step": 135245, "epoch": 1519, "lr": 5.042262295136838e-05}
{"train_loss": 0.14923541247844696, "global_step": 135246, "epoch": 1519, "lr": 5.042204323544637e-05}
{"train_loss": 0.14565415680408478, "global_step": 135247, "epoch": 1519, "lr": 5.0421463519467624e-05}
{"train_loss": 0.12251489609479904, "global_step": 135248, "epoch": 1519, "lr": 5.042088380343221e-05}
{"train_loss": 0.1230652928352356, "global_step": 135249, "epoch": 1519, "lr": 5.042030408734022e-05}
{"train_loss": 0.10880054533481598, "global_step": 135250, "epoch": 1519, "lr": 5.041972437119171e-05}
{"train_loss": 0.1274072527885437, "global_step": 135251, "epoch": 1519, "lr": 5.0419144654986794e-05}
{"train_loss": 0.12538538873195648, "global_step": 135252, "epoch": 1519, "lr": 5.041856493872552e-05}
{"train_loss": 0.10994388908147812, "global_step": 135253, "epoch": 1519, "lr": 5.041798522240797e-05}
{"train_loss": 0.07354295253753662, "global_step": 135254, "epoch": 1519, "lr": 5.041740550603423e-05}
{"train_loss": 0.16049125790596008, "global_step": 135255, "epoch": 1519, "lr": 5.0416825789604374e-05}
{"train_loss": 0.1266290545463562, "global_step": 135256, "epoch": 1519, "lr": 5.041624607311849e-05}
{"train_loss": 0.0741044208407402, "global_step": 135257, "epoch": 1519, "lr": 5.041566635657664e-05}
{"train_loss": 0.11915713548660278, "global_step": 135258, "epoch": 1519, "lr": 5.04150866399789e-05}
{"train_loss": 0.11096369475126266, "global_step": 135259, "epoch": 1519, "lr": 5.041450692332537e-05}
{"train_loss": 0.14146561920642853, "global_step": 135260, "epoch": 1519, "lr": 5.0413927206616106e-05}
{"train_loss": 0.18413977324962616, "global_step": 135261, "epoch": 1519, "lr": 5.0413347489851204e-05}
{"train_loss": 0.0902082771062851, "global_step": 135262, "epoch": 1519, "lr": 5.0412767773030724e-05}
{"train_loss": 0.15913386642932892, "global_step": 135263, "epoch": 1519, "lr": 5.041218805615475e-05}
{"train_loss": 0.15116435289382935, "global_step": 135264, "epoch": 1519, "lr": 5.041160833922337e-05}
{"train_loss": 0.17561495304107666, "global_step": 135265, "epoch": 1519, "lr": 5.0411028622236656e-05}
{"train_loss": 0.12675224244594574, "global_step": 135266, "epoch": 1519, "lr": 5.0410448905194674e-05}
{"train_loss": 0.09477830678224564, "global_step": 135267, "epoch": 1519, "lr": 5.0409869188097525e-05}
{"train_loss": 0.10775987058877945, "global_step": 135268, "epoch": 1519, "lr": 5.0409289470945266e-05}
{"train_loss": 0.1561061590909958, "global_step": 135269, "epoch": 1519, "lr": 5.040870975373798e-05}
{"train_loss": 0.130426287651062, "global_step": 135270, "epoch": 1519, "lr": 5.040813003647575e-05}
{"train_loss": 0.11625812202692032, "global_step": 135271, "epoch": 1519, "lr": 5.040755031915866e-05}
{"train_loss": 0.11653871089220047, "global_step": 135272, "epoch": 1519, "lr": 5.040697060178677e-05}
{"train_loss": 0.13108035922050476, "global_step": 135273, "epoch": 1519, "lr": 5.040639088436016e-05}
{"train_loss": 0.09602854400873184, "global_step": 135274, "epoch": 1519, "lr": 5.0405811166878936e-05}
{"train_loss": 0.1208709254860878, "global_step": 135275, "epoch": 1519, "lr": 5.0405231449343146e-05}
{"train_loss": 0.1467537134885788, "global_step": 135276, "epoch": 1519, "lr": 5.040465173175288e-05}
{"train_loss": 0.12079620361328125, "global_step": 135277, "epoch": 1519, "lr": 5.040407201410822e-05}
{"train_loss": 0.08398161828517914, "global_step": 135278, "epoch": 1519, "lr": 5.040349229640923e-05}
{"train_loss": 0.11898848761835795, "global_step": 135279, "epoch": 1519, "lr": 5.040291257865599e-05, "val_loss": 5.679883003234863}
{"train_loss": 0.12988537549972534, "global_step": 135280, "epoch": 1520, "lr": 5.040233286084859e-05}
{"train_loss": 0.10855279117822647, "global_step": 135281, "epoch": 1520, "lr": 5.04017531429871e-05}
{"train_loss": 0.10185855627059937, "global_step": 135282, "epoch": 1520, "lr": 5.040117342507159e-05}
{"train_loss": 0.12136244773864746, "global_step": 135283, "epoch": 1520, "lr": 5.040059370710216e-05}
{"train_loss": 0.09354795515537262, "global_step": 135284, "epoch": 1520, "lr": 5.040001398907886e-05}
{"train_loss": 0.13854902982711792, "global_step": 135285, "epoch": 1520, "lr": 5.03994342710018e-05}
{"train_loss": 0.08742457628250122, "global_step": 135286, "epoch": 1520, "lr": 5.039885455287103e-05}
{"train_loss": 0.1337118297815323, "global_step": 135287, "epoch": 1520, "lr": 5.039827483468665e-05}
{"train_loss": 0.08465855568647385, "global_step": 135288, "epoch": 1520, "lr": 5.039769511644872e-05}
{"train_loss": 0.08490324765443802, "global_step": 135289, "epoch": 1520, "lr": 5.0397115398157324e-05}
{"train_loss": 0.13570420444011688, "global_step": 135290, "epoch": 1520, "lr": 5.039653567981254e-05}
{"train_loss": 0.11057061702013016, "global_step": 135291, "epoch": 1520, "lr": 5.0395955961414454e-05}
{"train_loss": 0.1304132491350174, "global_step": 135292, "epoch": 1520, "lr": 5.039537624296313e-05}
{"train_loss": 0.13452471792697906, "global_step": 135293, "epoch": 1520, "lr": 5.0394796524458653e-05}
{"train_loss": 0.09439820796251297, "global_step": 135294, "epoch": 1520, "lr": 5.0394216805901085e-05}
{"train_loss": 0.10729681700468063, "global_step": 135295, "epoch": 1520, "lr": 5.039363708729053e-05}
{"train_loss": 0.11804791539907455, "global_step": 135296, "epoch": 1520, "lr": 5.0393057368627085e-05}
{"train_loss": 0.150962233543396, "global_step": 135297, "epoch": 1520, "lr": 5.0392477649910775e-05}
{"train_loss": 0.09362876415252686, "global_step": 135298, "epoch": 1520, "lr": 5.03918979311417e-05}
{"train_loss": 0.09017840027809143, "global_step": 135299, "epoch": 1520, "lr": 5.039131821231994e-05}
{"train_loss": 0.06681334227323532, "global_step": 135300, "epoch": 1520, "lr": 5.0390738493445565e-05}
{"train_loss": 0.1069134920835495, "global_step": 135301, "epoch": 1520, "lr": 5.039015877451867e-05}
{"train_loss": 0.10053811967372894, "global_step": 135302, "epoch": 1520, "lr": 5.038957905553932e-05}
{"train_loss": 0.12007574737071991, "global_step": 135303, "epoch": 1520, "lr": 5.0388999336507595e-05}
{"train_loss": 0.14876039326190948, "global_step": 135304, "epoch": 1520, "lr": 5.0388419617423564e-05}
{"train_loss": 0.10993779450654984, "global_step": 135305, "epoch": 1520, "lr": 5.038783989828732e-05}
{"train_loss": 0.14680379629135132, "global_step": 135306, "epoch": 1520, "lr": 5.038726017909895e-05}
{"train_loss": 0.1159396767616272, "global_step": 135307, "epoch": 1520, "lr": 5.0386680459858505e-05}
{"train_loss": 0.13539743423461914, "global_step": 135308, "epoch": 1520, "lr": 5.038610074056608e-05}
{"train_loss": 0.08895005285739899, "global_step": 135309, "epoch": 1520, "lr": 5.038552102122175e-05}
{"train_loss": 0.07549850642681122, "global_step": 135310, "epoch": 1520, "lr": 5.0384941301825586e-05}
{"train_loss": 0.10850867629051208, "global_step": 135311, "epoch": 1520, "lr": 5.038436158237767e-05}
{"train_loss": 0.14176727831363678, "global_step": 135312, "epoch": 1520, "lr": 5.038378186287808e-05}
{"train_loss": 0.11007348448038101, "global_step": 135313, "epoch": 1520, "lr": 5.0383202143326893e-05}
{"train_loss": 0.15403763949871063, "global_step": 135314, "epoch": 1520, "lr": 5.038262242372419e-05}
{"train_loss": 0.08035649359226227, "global_step": 135315, "epoch": 1520, "lr": 5.0382042704070064e-05}
{"train_loss": 0.09527800232172012, "global_step": 135316, "epoch": 1520, "lr": 5.038146298436457e-05}
{"train_loss": 0.13817954063415527, "global_step": 135317, "epoch": 1520, "lr": 5.038088326460779e-05}
{"train_loss": 0.11881398409605026, "global_step": 135318, "epoch": 1520, "lr": 5.03803035447998e-05}
{"train_loss": 0.10745055973529816, "global_step": 135319, "epoch": 1520, "lr": 5.037972382494069e-05}
{"train_loss": 0.08558531105518341, "global_step": 135320, "epoch": 1520, "lr": 5.037914410503053e-05}
{"train_loss": 0.06966851651668549, "global_step": 135321, "epoch": 1520, "lr": 5.037856438506939e-05}
{"train_loss": 0.10099221765995026, "global_step": 135322, "epoch": 1520, "lr": 5.037798466505738e-05}
{"train_loss": 0.10257288813591003, "global_step": 135323, "epoch": 1520, "lr": 5.037740494499453e-05}
{"train_loss": 0.10080451518297195, "global_step": 135324, "epoch": 1520, "lr": 5.037682522488095e-05}
{"train_loss": 0.06670669466257095, "global_step": 135325, "epoch": 1520, "lr": 5.0376245504716715e-05}
{"train_loss": 0.11625196039676666, "global_step": 135326, "epoch": 1520, "lr": 5.037566578450189e-05}
{"train_loss": 0.13916125893592834, "global_step": 135327, "epoch": 1520, "lr": 5.0375086064236566e-05}
{"train_loss": 0.06339569389820099, "global_step": 135328, "epoch": 1520, "lr": 5.0374506343920815e-05}
{"train_loss": 0.13106794655323029, "global_step": 135329, "epoch": 1520, "lr": 5.037392662355472e-05}
{"train_loss": 0.1642756462097168, "global_step": 135330, "epoch": 1520, "lr": 5.037334690313834e-05}
{"train_loss": 0.12802156805992126, "global_step": 135331, "epoch": 1520, "lr": 5.037276718267178e-05}
{"train_loss": 0.09051875025033951, "global_step": 135332, "epoch": 1520, "lr": 5.0372187462155104e-05}
{"train_loss": 0.09093771129846573, "global_step": 135333, "epoch": 1520, "lr": 5.03716077415884e-05}
{"train_loss": 0.19299866259098053, "global_step": 135334, "epoch": 1520, "lr": 5.037102802097174e-05}
{"train_loss": 0.09579794108867645, "global_step": 135335, "epoch": 1520, "lr": 5.03704483003052e-05}
{"train_loss": 0.08161196857690811, "global_step": 135336, "epoch": 1520, "lr": 5.036986857958884e-05}
{"train_loss": 0.13280080258846283, "global_step": 135337, "epoch": 1520, "lr": 5.036928885882277e-05}
{"train_loss": 0.13844306766986847, "global_step": 135338, "epoch": 1520, "lr": 5.036870913800704e-05}
{"train_loss": 0.08504827320575714, "global_step": 135339, "epoch": 1520, "lr": 5.036812941714176e-05}
{"train_loss": 0.12231593579053879, "global_step": 135340, "epoch": 1520, "lr": 5.036754969622698e-05}
{"train_loss": 0.07694318890571594, "global_step": 135341, "epoch": 1520, "lr": 5.036696997526279e-05}
{"train_loss": 0.08992331475019455, "global_step": 135342, "epoch": 1520, "lr": 5.0366390254249264e-05}
{"train_loss": 0.09069148451089859, "global_step": 135343, "epoch": 1520, "lr": 5.036581053318648e-05}
{"train_loss": 0.122980035841465, "global_step": 135344, "epoch": 1520, "lr": 5.0365230812074514e-05}
{"train_loss": 0.09286820143461227, "global_step": 135345, "epoch": 1520, "lr": 5.036465109091345e-05}
{"train_loss": 0.11901772767305374, "global_step": 135346, "epoch": 1520, "lr": 5.036407136970336e-05}
{"train_loss": 0.15061163902282715, "global_step": 135347, "epoch": 1520, "lr": 5.0363491648444326e-05}
{"train_loss": 0.0950838252902031, "global_step": 135348, "epoch": 1520, "lr": 5.0362911927136435e-05}
{"train_loss": 0.09631615877151489, "global_step": 135349, "epoch": 1520, "lr": 5.036233220577973e-05}
{"train_loss": 0.16241462528705597, "global_step": 135350, "epoch": 1520, "lr": 5.036175248437434e-05}
{"train_loss": 0.11927127838134766, "global_step": 135351, "epoch": 1520, "lr": 5.0361172762920304e-05}
{"train_loss": 0.13567917048931122, "global_step": 135352, "epoch": 1520, "lr": 5.036059304141772e-05}
{"train_loss": 0.15824465453624725, "global_step": 135353, "epoch": 1520, "lr": 5.036001331986665e-05}
{"train_loss": 0.10619273036718369, "global_step": 135354, "epoch": 1520, "lr": 5.0359433598267205e-05}
{"train_loss": 0.1651855707168579, "global_step": 135355, "epoch": 1520, "lr": 5.0358853876619416e-05}
{"train_loss": 0.07262121140956879, "global_step": 135356, "epoch": 1520, "lr": 5.035827415492338e-05}
{"train_loss": 0.11844446510076523, "global_step": 135357, "epoch": 1520, "lr": 5.035769443317919e-05}
{"train_loss": 0.10611788928508759, "global_step": 135358, "epoch": 1520, "lr": 5.0357114711386914e-05}
{"train_loss": 0.0944560095667839, "global_step": 135359, "epoch": 1520, "lr": 5.035653498954661e-05}
{"train_loss": 0.13464871048927307, "global_step": 135360, "epoch": 1520, "lr": 5.0355955267658394e-05}
{"train_loss": 0.10912855714559555, "global_step": 135361, "epoch": 1520, "lr": 5.0355375545722315e-05}
{"train_loss": 0.09029313176870346, "global_step": 135362, "epoch": 1520, "lr": 5.0354795823738465e-05}
{"train_loss": 0.11411010473966599, "global_step": 135363, "epoch": 1520, "lr": 5.035421610170691e-05}
{"train_loss": 0.10227315872907639, "global_step": 135364, "epoch": 1520, "lr": 5.035363637962774e-05}
{"train_loss": 0.13223746418952942, "global_step": 135365, "epoch": 1520, "lr": 5.035305665750103e-05}
{"train_loss": 0.1658971756696701, "global_step": 135366, "epoch": 1520, "lr": 5.035247693532685e-05}
{"train_loss": 0.09646323323249817, "global_step": 135367, "epoch": 1520, "lr": 5.035189721310527e-05}
{"train_loss": 0.11364841419324447, "global_step": 135368, "epoch": 1520, "lr": 5.03513174908364e-05, "val_loss": 5.770643711090088, "train_action_mse_error": 24.266300201416016}
{"train_loss": 0.08921106159687042, "global_step": 135369, "epoch": 1521, "lr": 5.03507377685203e-05}
{"train_loss": 0.1328965425491333, "global_step": 135370, "epoch": 1521, "lr": 5.035015804615705e-05}
{"train_loss": 0.09435563534498215, "global_step": 135371, "epoch": 1521, "lr": 5.0349578323746716e-05}
{"train_loss": 0.17111362516880035, "global_step": 135372, "epoch": 1521, "lr": 5.034899860128939e-05}
{"train_loss": 0.10542431473731995, "global_step": 135373, "epoch": 1521, "lr": 5.034841887878515e-05}
{"train_loss": 0.11398531496524811, "global_step": 135374, "epoch": 1521, "lr": 5.034783915623407e-05}
{"train_loss": 0.1005871593952179, "global_step": 135375, "epoch": 1521, "lr": 5.034725943363622e-05}
{"train_loss": 0.1352173388004303, "global_step": 135376, "epoch": 1521, "lr": 5.034667971099168e-05}
{"train_loss": 0.11520073562860489, "global_step": 135377, "epoch": 1521, "lr": 5.034609998830055e-05}
{"train_loss": 0.07610625773668289, "global_step": 135378, "epoch": 1521, "lr": 5.034552026556288e-05}
{"train_loss": 0.06806757301092148, "global_step": 135379, "epoch": 1521, "lr": 5.034494054277876e-05}
{"train_loss": 0.11057019233703613, "global_step": 135380, "epoch": 1521, "lr": 5.0344360819948265e-05}
{"train_loss": 0.06866972893476486, "global_step": 135381, "epoch": 1521, "lr": 5.0343781097071486e-05}
{"train_loss": 0.1541036069393158, "global_step": 135382, "epoch": 1521, "lr": 5.034320137414847e-05}
{"train_loss": 0.11106529086828232, "global_step": 135383, "epoch": 1521, "lr": 5.034262165117933e-05}
{"train_loss": 0.1413908451795578, "global_step": 135384, "epoch": 1521, "lr": 5.034204192816413e-05}
{"train_loss": 0.12457238882780075, "global_step": 135385, "epoch": 1521, "lr": 5.034146220510293e-05}
{"train_loss": 0.12035051733255386, "global_step": 135386, "epoch": 1521, "lr": 5.0340882481995833e-05}
{"train_loss": 0.12392593920230865, "global_step": 135387, "epoch": 1521, "lr": 5.034030275884291e-05}
{"train_loss": 0.14451760053634644, "global_step": 135388, "epoch": 1521, "lr": 5.0339723035644235e-05}
{"train_loss": 0.1130659207701683, "global_step": 135389, "epoch": 1521, "lr": 5.03391433123999e-05}
{"train_loss": 0.07623953372240067, "global_step": 135390, "epoch": 1521, "lr": 5.0338563589109956e-05}
{"train_loss": 0.03888528794050217, "global_step": 135391, "epoch": 1521, "lr": 5.0337983865774505e-05}
{"train_loss": 0.12459688633680344, "global_step": 135392, "epoch": 1521, "lr": 5.033740414239363e-05}
{"train_loss": 0.12674807012081146, "global_step": 135393, "epoch": 1521, "lr": 5.033682441896738e-05}
{"train_loss": 0.08683747053146362, "global_step": 135394, "epoch": 1521, "lr": 5.033624469549585e-05}
{"train_loss": 0.10168328881263733, "global_step": 135395, "epoch": 1521, "lr": 5.033566497197911e-05}
{"train_loss": 0.17459984123706818, "global_step": 135396, "epoch": 1521, "lr": 5.0335085248417244e-05}
{"train_loss": 0.10483148694038391, "global_step": 135397, "epoch": 1521, "lr": 5.0334505524810336e-05}
{"train_loss": 0.12736837565898895, "global_step": 135398, "epoch": 1521, "lr": 5.0333925801158454e-05}
{"train_loss": 0.09467917680740356, "global_step": 135399, "epoch": 1521, "lr": 5.0333346077461685e-05}
{"train_loss": 0.06731163710355759, "global_step": 135400, "epoch": 1521, "lr": 5.03327663537201e-05}
{"train_loss": 0.14069940149784088, "global_step": 135401, "epoch": 1521, "lr": 5.0332186629933775e-05}
{"train_loss": 0.1332745999097824, "global_step": 135402, "epoch": 1521, "lr": 5.033160690610279e-05}
{"train_loss": 0.14323115348815918, "global_step": 135403, "epoch": 1521, "lr": 5.033102718222722e-05}
{"train_loss": 0.11734495311975479, "global_step": 135404, "epoch": 1521, "lr": 5.033044745830715e-05}
{"train_loss": 0.1296217143535614, "global_step": 135405, "epoch": 1521, "lr": 5.0329867734342665e-05}
{"train_loss": 0.10108807682991028, "global_step": 135406, "epoch": 1521, "lr": 5.032928801033383e-05}
{"train_loss": 0.1285555511713028, "global_step": 135407, "epoch": 1521, "lr": 5.0328708286280724e-05}
{"train_loss": 0.08166570961475372, "global_step": 135408, "epoch": 1521, "lr": 5.032812856218343e-05}
{"train_loss": 0.07380905747413635, "global_step": 135409, "epoch": 1521, "lr": 5.032754883804202e-05}
{"train_loss": 0.07539162039756775, "global_step": 135410, "epoch": 1521, "lr": 5.032696911385658e-05}
{"train_loss": 0.12321410328149796, "global_step": 135411, "epoch": 1521, "lr": 5.032638938962718e-05}
{"train_loss": 0.0947251170873642, "global_step": 135412, "epoch": 1521, "lr": 5.0325809665353904e-05}
{"train_loss": 0.12024339288473129, "global_step": 135413, "epoch": 1521, "lr": 5.0325229941036836e-05}
{"train_loss": 0.08626531809568405, "global_step": 135414, "epoch": 1521, "lr": 5.032465021667603e-05}
{"train_loss": 0.061348333954811096, "global_step": 135415, "epoch": 1521, "lr": 5.032407049227158e-05}
{"train_loss": 0.1393224447965622, "global_step": 135416, "epoch": 1521, "lr": 5.0323490767823555e-05}
{"train_loss": 0.1003163605928421, "global_step": 135417, "epoch": 1521, "lr": 5.032291104333205e-05}
{"train_loss": 0.17078249156475067, "global_step": 135418, "epoch": 1521, "lr": 5.0322331318797136e-05}
{"train_loss": 0.13643422722816467, "global_step": 135419, "epoch": 1521, "lr": 5.032175159421888e-05}
{"train_loss": 0.09036291390657425, "global_step": 135420, "epoch": 1521, "lr": 5.032117186959737e-05}
{"train_loss": 0.2070290595293045, "global_step": 135421, "epoch": 1521, "lr": 5.0320592144932676e-05}
{"train_loss": 0.10762018710374832, "global_step": 135422, "epoch": 1521, "lr": 5.0320012420224896e-05}
{"train_loss": 0.12411820888519287, "global_step": 135423, "epoch": 1521, "lr": 5.0319432695474086e-05}
{"train_loss": 0.1330578774213791, "global_step": 135424, "epoch": 1521, "lr": 5.031885297068034e-05}
{"train_loss": 0.10229139029979706, "global_step": 135425, "epoch": 1521, "lr": 5.0318273245843725e-05}
{"train_loss": 0.10456962883472443, "global_step": 135426, "epoch": 1521, "lr": 5.031769352096433e-05}
{"train_loss": 0.14060474932193756, "global_step": 135427, "epoch": 1521, "lr": 5.0317113796042216e-05}
{"train_loss": 0.11588392406702042, "global_step": 135428, "epoch": 1521, "lr": 5.031653407107747e-05}
{"train_loss": 0.08063878864049911, "global_step": 135429, "epoch": 1521, "lr": 5.031595434607017e-05}
{"train_loss": 0.08779651671648026, "global_step": 135430, "epoch": 1521, "lr": 5.03153746210204e-05}
{"train_loss": 0.0871867835521698, "global_step": 135431, "epoch": 1521, "lr": 5.031479489592823e-05}
{"train_loss": 0.1538763791322708, "global_step": 135432, "epoch": 1521, "lr": 5.031421517079373e-05}
{"train_loss": 0.12047037482261658, "global_step": 135433, "epoch": 1521, "lr": 5.0313635445616994e-05}
{"train_loss": 0.10335540771484375, "global_step": 135434, "epoch": 1521, "lr": 5.0313055720398094e-05}
{"train_loss": 0.09291206300258636, "global_step": 135435, "epoch": 1521, "lr": 5.031247599513711e-05}
{"train_loss": 0.1018807590007782, "global_step": 135436, "epoch": 1521, "lr": 5.031189626983411e-05}
{"train_loss": 0.11279533803462982, "global_step": 135437, "epoch": 1521, "lr": 5.031131654448918e-05}
{"train_loss": 0.1188134104013443, "global_step": 135438, "epoch": 1521, "lr": 5.03107368191024e-05}
{"train_loss": 0.12883281707763672, "global_step": 135439, "epoch": 1521, "lr": 5.0310157093673835e-05}
{"train_loss": 0.09166641533374786, "global_step": 135440, "epoch": 1521, "lr": 5.030957736820359e-05}
{"train_loss": 0.12860903143882751, "global_step": 135441, "epoch": 1521, "lr": 5.030899764269171e-05}
{"train_loss": 0.11113230884075165, "global_step": 135442, "epoch": 1521, "lr": 5.030841791713831e-05}
{"train_loss": 0.12580648064613342, "global_step": 135443, "epoch": 1521, "lr": 5.030783819154343e-05}
{"train_loss": 0.09549439698457718, "global_step": 135444, "epoch": 1521, "lr": 5.030725846590717e-05}
{"train_loss": 0.1483680158853531, "global_step": 135445, "epoch": 1521, "lr": 5.0306678740229605e-05}
{"train_loss": 0.13136355578899384, "global_step": 135446, "epoch": 1521, "lr": 5.030609901451081e-05}
{"train_loss": 0.12586578726768494, "global_step": 135447, "epoch": 1521, "lr": 5.0305519288750855e-05}
{"train_loss": 0.12598732113838196, "global_step": 135448, "epoch": 1521, "lr": 5.030493956294984e-05}
{"train_loss": 0.12777495384216309, "global_step": 135449, "epoch": 1521, "lr": 5.030435983710783e-05}
{"train_loss": 0.11854998022317886, "global_step": 135450, "epoch": 1521, "lr": 5.030378011122491e-05}
{"train_loss": 0.12221367657184601, "global_step": 135451, "epoch": 1521, "lr": 5.030320038530113e-05}
{"train_loss": 0.07017632573843002, "global_step": 135452, "epoch": 1521, "lr": 5.0302620659336595e-05}
{"train_loss": 0.1605101078748703, "global_step": 135453, "epoch": 1521, "lr": 5.0302040933331385e-05}
{"train_loss": 0.15511329472064972, "global_step": 135454, "epoch": 1521, "lr": 5.0301461207285563e-05}
{"train_loss": 0.11824411898851395, "global_step": 135455, "epoch": 1521, "lr": 5.03008814811992e-05}
{"train_loss": 0.05731918662786484, "global_step": 135456, "epoch": 1521, "lr": 5.0300301755072396e-05}
{"train_loss": 0.11380935594272078, "global_step": 135457, "epoch": 1521, "lr": 5.029972202890522e-05, "val_loss": 5.7785820960998535}
{"train_loss": 0.09179361164569855, "global_step": 135458, "epoch": 1522, "lr": 5.0299142302697757e-05}
{"train_loss": 0.1413576900959015, "global_step": 135459, "epoch": 1522, "lr": 5.029856257645007e-05}
{"train_loss": 0.12153446674346924, "global_step": 135460, "epoch": 1522, "lr": 5.029798285016225e-05}
{"train_loss": 0.1097642108798027, "global_step": 135461, "epoch": 1522, "lr": 5.0297403123834366e-05}
{"train_loss": 0.13755527138710022, "global_step": 135462, "epoch": 1522, "lr": 5.0296823397466496e-05}
{"train_loss": 0.11807511746883392, "global_step": 135463, "epoch": 1522, "lr": 5.029624367105873e-05}
{"train_loss": 0.08001869916915894, "global_step": 135464, "epoch": 1522, "lr": 5.0295663944611135e-05}
{"train_loss": 0.1533653289079666, "global_step": 135465, "epoch": 1522, "lr": 5.02950842181238e-05}
{"train_loss": 0.1171896904706955, "global_step": 135466, "epoch": 1522, "lr": 5.029450449159678e-05}
{"train_loss": 0.15860459208488464, "global_step": 135467, "epoch": 1522, "lr": 5.029392476503018e-05}
{"train_loss": 0.13676953315734863, "global_step": 135468, "epoch": 1522, "lr": 5.0293345038424055e-05}
{"train_loss": 0.07326039671897888, "global_step": 135469, "epoch": 1522, "lr": 5.02927653117785e-05}
{"train_loss": 0.128424733877182, "global_step": 135470, "epoch": 1522, "lr": 5.02921855850936e-05}
{"train_loss": 0.13950146734714508, "global_step": 135471, "epoch": 1522, "lr": 5.02916058583694e-05}
{"train_loss": 0.1349141001701355, "global_step": 135472, "epoch": 1522, "lr": 5.0291026131606e-05}
{"train_loss": 0.126746267080307, "global_step": 135473, "epoch": 1522, "lr": 5.029044640480347e-05}
{"train_loss": 0.057675451040267944, "global_step": 135474, "epoch": 1522, "lr": 5.0289866677961906e-05}
{"train_loss": 0.17302179336547852, "global_step": 135475, "epoch": 1522, "lr": 5.028928695108136e-05}
{"train_loss": 0.11374896764755249, "global_step": 135476, "epoch": 1522, "lr": 5.028870722416192e-05}
{"train_loss": 0.11656177043914795, "global_step": 135477, "epoch": 1522, "lr": 5.0288127497203676e-05}
{"train_loss": 0.09515190869569778, "global_step": 135478, "epoch": 1522, "lr": 5.02875477702067e-05}
{"train_loss": 0.11949703097343445, "global_step": 135479, "epoch": 1522, "lr": 5.0286968043171054e-05}
{"train_loss": 0.08993227034807205, "global_step": 135480, "epoch": 1522, "lr": 5.028638831609685e-05}
{"train_loss": 0.1204635500907898, "global_step": 135481, "epoch": 1522, "lr": 5.028580858898412e-05}
{"train_loss": 0.1303294450044632, "global_step": 135482, "epoch": 1522, "lr": 5.028522886183298e-05}
{"train_loss": 0.12068481743335724, "global_step": 135483, "epoch": 1522, "lr": 5.0284649134643494e-05}
{"train_loss": 0.16149330139160156, "global_step": 135484, "epoch": 1522, "lr": 5.0284069407415736e-05}
{"train_loss": 0.1294122338294983, "global_step": 135485, "epoch": 1522, "lr": 5.0283489680149795e-05}
{"train_loss": 0.05464748293161392, "global_step": 135486, "epoch": 1522, "lr": 5.028290995284574e-05}
{"train_loss": 0.17550373077392578, "global_step": 135487, "epoch": 1522, "lr": 5.028233022550365e-05}
{"train_loss": 0.13403718173503876, "global_step": 135488, "epoch": 1522, "lr": 5.02817504981236e-05}
{"train_loss": 0.16687357425689697, "global_step": 135489, "epoch": 1522, "lr": 5.028117077070569e-05}
{"train_loss": 0.1146593764424324, "global_step": 135490, "epoch": 1522, "lr": 5.0280591043249966e-05}
{"train_loss": 0.1250203400850296, "global_step": 135491, "epoch": 1522, "lr": 5.0280011315756525e-05}
{"train_loss": 0.1551646888256073, "global_step": 135492, "epoch": 1522, "lr": 5.0279431588225435e-05}
{"train_loss": 0.06316877901554108, "global_step": 135493, "epoch": 1522, "lr": 5.027885186065677e-05}
{"train_loss": 0.09718939661979675, "global_step": 135494, "epoch": 1522, "lr": 5.027827213305062e-05}
{"train_loss": 0.08359087258577347, "global_step": 135495, "epoch": 1522, "lr": 5.027769240540707e-05}
{"train_loss": 0.08205723762512207, "global_step": 135496, "epoch": 1522, "lr": 5.027711267772618e-05}
{"train_loss": 0.10568154603242874, "global_step": 135497, "epoch": 1522, "lr": 5.027653295000804e-05}
{"train_loss": 0.080844447016716, "global_step": 135498, "epoch": 1522, "lr": 5.027595322225273e-05}
{"train_loss": 0.0887063667178154, "global_step": 135499, "epoch": 1522, "lr": 5.027537349446031e-05}
{"train_loss": 0.1102677509188652, "global_step": 135500, "epoch": 1522, "lr": 5.027479376663087e-05}
{"train_loss": 0.15229599177837372, "global_step": 135501, "epoch": 1522, "lr": 5.027421403876449e-05}
{"train_loss": 0.08275631070137024, "global_step": 135502, "epoch": 1522, "lr": 5.027363431086124e-05}
{"train_loss": 0.17828994989395142, "global_step": 135503, "epoch": 1522, "lr": 5.027305458292121e-05}
{"train_loss": 0.09736394137144089, "global_step": 135504, "epoch": 1522, "lr": 5.027247485494447e-05}
{"train_loss": 0.15458925068378448, "global_step": 135505, "epoch": 1522, "lr": 5.02718951269311e-05}
{"train_loss": 0.12925371527671814, "global_step": 135506, "epoch": 1522, "lr": 5.027131539888118e-05}
{"train_loss": 0.2031281441450119, "global_step": 135507, "epoch": 1522, "lr": 5.0270735670794775e-05}
{"train_loss": 0.13136205077171326, "global_step": 135508, "epoch": 1522, "lr": 5.0270155942671995e-05}
{"train_loss": 0.14353978633880615, "global_step": 135509, "epoch": 1522, "lr": 5.026957621451288e-05}
{"train_loss": 0.16443634033203125, "global_step": 135510, "epoch": 1522, "lr": 5.026899648631752e-05}
{"train_loss": 0.11724091321229935, "global_step": 135511, "epoch": 1522, "lr": 5.0268416758086e-05}
{"train_loss": 0.06555481255054474, "global_step": 135512, "epoch": 1522, "lr": 5.026783702981839e-05}
{"train_loss": 0.07036899775266647, "global_step": 135513, "epoch": 1522, "lr": 5.026725730151478e-05}
{"train_loss": 0.11987311393022537, "global_step": 135514, "epoch": 1522, "lr": 5.026667757317524e-05}
{"train_loss": 0.1772812306880951, "global_step": 135515, "epoch": 1522, "lr": 5.0266097844799845e-05}
{"train_loss": 0.11164513230323792, "global_step": 135516, "epoch": 1522, "lr": 5.0265518116388674e-05}
{"train_loss": 0.16848152875900269, "global_step": 135517, "epoch": 1522, "lr": 5.0264938387941806e-05}
{"train_loss": 0.08671194314956665, "global_step": 135518, "epoch": 1522, "lr": 5.026435865945933e-05}
{"train_loss": 0.1479853242635727, "global_step": 135519, "epoch": 1522, "lr": 5.026377893094131e-05}
{"train_loss": 0.11094215512275696, "global_step": 135520, "epoch": 1522, "lr": 5.0263199202387824e-05}
{"train_loss": 0.10650075972080231, "global_step": 135521, "epoch": 1522, "lr": 5.026261947379895e-05}
{"train_loss": 0.19353565573692322, "global_step": 135522, "epoch": 1522, "lr": 5.02620397451748e-05}
{"train_loss": 0.12711107730865479, "global_step": 135523, "epoch": 1522, "lr": 5.026146001651539e-05}
{"train_loss": 0.07246040552854538, "global_step": 135524, "epoch": 1522, "lr": 5.026088028782083e-05}
{"train_loss": 0.08015269786119461, "global_step": 135525, "epoch": 1522, "lr": 5.026030055909121e-05}
{"train_loss": 0.07449723035097122, "global_step": 135526, "epoch": 1522, "lr": 5.0259720830326595e-05}
{"train_loss": 0.08848415315151215, "global_step": 135527, "epoch": 1522, "lr": 5.025914110152706e-05}
{"train_loss": 0.09124498069286346, "global_step": 135528, "epoch": 1522, "lr": 5.02585613726927e-05}
{"train_loss": 0.11680080741643906, "global_step": 135529, "epoch": 1522, "lr": 5.025798164382357e-05}
{"train_loss": 0.10234896838665009, "global_step": 135530, "epoch": 1522, "lr": 5.0257401914919745e-05}
{"train_loss": 0.10312161594629288, "global_step": 135531, "epoch": 1522, "lr": 5.0256822185981333e-05}
{"train_loss": 0.09317251294851303, "global_step": 135532, "epoch": 1522, "lr": 5.025624245700838e-05}
{"train_loss": 0.11765820533037186, "global_step": 135533, "epoch": 1522, "lr": 5.025566272800098e-05}
{"train_loss": 0.09688130766153336, "global_step": 135534, "epoch": 1522, "lr": 5.0255082998959226e-05}
{"train_loss": 0.08729462325572968, "global_step": 135535, "epoch": 1522, "lr": 5.025450326988317e-05}
{"train_loss": 0.12034765630960464, "global_step": 135536, "epoch": 1522, "lr": 5.02539235407729e-05}
{"train_loss": 0.2382378876209259, "global_step": 135537, "epoch": 1522, "lr": 5.025334381162848e-05}
{"train_loss": 0.09715879708528519, "global_step": 135538, "epoch": 1522, "lr": 5.025276408245002e-05}
{"train_loss": 0.11816999316215515, "global_step": 135539, "epoch": 1522, "lr": 5.025218435323756e-05}
{"train_loss": 0.10259220004081726, "global_step": 135540, "epoch": 1522, "lr": 5.025160462399121e-05}
{"train_loss": 0.1207110583782196, "global_step": 135541, "epoch": 1522, "lr": 5.025102489471104e-05}
{"train_loss": 0.16475556790828705, "global_step": 135542, "epoch": 1522, "lr": 5.025044516539711e-05}
{"train_loss": 0.11984951794147491, "global_step": 135543, "epoch": 1522, "lr": 5.024986543604952e-05}
{"train_loss": 0.11580722033977509, "global_step": 135544, "epoch": 1522, "lr": 5.024928570666834e-05}
{"train_loss": 0.13173596560955048, "global_step": 135545, "epoch": 1522, "lr": 5.024870597725363e-05}
{"train_loss": 0.11980817785088936, "global_step": 135546, "epoch": 1522, "lr": 5.0248126247805495e-05, "val_loss": 5.656352519989014}
{"train_loss": 0.11486563831567764, "global_step": 135547, "epoch": 1523, "lr": 5.024754651832402e-05}
{"train_loss": 0.06314562261104584, "global_step": 135548, "epoch": 1523, "lr": 5.024696678880923e-05}
{"train_loss": 0.1678265780210495, "global_step": 135549, "epoch": 1523, "lr": 5.0246387059261256e-05}
{"train_loss": 0.140917107462883, "global_step": 135550, "epoch": 1523, "lr": 5.024580732968016e-05}
{"train_loss": 0.09923531860113144, "global_step": 135551, "epoch": 1523, "lr": 5.024522760006602e-05}
{"train_loss": 0.09771328419446945, "global_step": 135552, "epoch": 1523, "lr": 5.0244647870418916e-05}
{"train_loss": 0.12105175107717514, "global_step": 135553, "epoch": 1523, "lr": 5.024406814073891e-05}
{"train_loss": 0.08452869206666946, "global_step": 135554, "epoch": 1523, "lr": 5.0243488411026093e-05}
{"train_loss": 0.11074569821357727, "global_step": 135555, "epoch": 1523, "lr": 5.0242908681280554e-05}
{"train_loss": 0.1332128942012787, "global_step": 135556, "epoch": 1523, "lr": 5.024232895150235e-05}
{"train_loss": 0.13863061368465424, "global_step": 135557, "epoch": 1523, "lr": 5.024174922169157e-05}
{"train_loss": 0.1192062646150589, "global_step": 135558, "epoch": 1523, "lr": 5.0241169491848286e-05}
{"train_loss": 0.1397590935230255, "global_step": 135559, "epoch": 1523, "lr": 5.0240589761972575e-05}
{"train_loss": 0.10151522606611252, "global_step": 135560, "epoch": 1523, "lr": 5.0240010032064535e-05}
{"train_loss": 0.19710657000541687, "global_step": 135561, "epoch": 1523, "lr": 5.0239430302124225e-05}
{"train_loss": 0.09603269398212433, "global_step": 135562, "epoch": 1523, "lr": 5.0238850572151716e-05}
{"train_loss": 0.07382114976644516, "global_step": 135563, "epoch": 1523, "lr": 5.0238270842147106e-05}
{"train_loss": 0.15449951589107513, "global_step": 135564, "epoch": 1523, "lr": 5.023769111211045e-05}
{"train_loss": 0.09868617355823517, "global_step": 135565, "epoch": 1523, "lr": 5.0237111382041855e-05}
{"train_loss": 0.14300468564033508, "global_step": 135566, "epoch": 1523, "lr": 5.0236531651941375e-05}
{"train_loss": 0.15435999631881714, "global_step": 135567, "epoch": 1523, "lr": 5.02359519218091e-05}
{"train_loss": 0.07882118970155716, "global_step": 135568, "epoch": 1523, "lr": 5.02353721916451e-05}
{"train_loss": 0.21293862164020538, "global_step": 135569, "epoch": 1523, "lr": 5.023479246144946e-05}
{"train_loss": 0.09288270771503448, "global_step": 135570, "epoch": 1523, "lr": 5.0234212731222266e-05}
{"train_loss": 0.12285681068897247, "global_step": 135571, "epoch": 1523, "lr": 5.023363300096357e-05}
{"train_loss": 0.15101706981658936, "global_step": 135572, "epoch": 1523, "lr": 5.023305327067347e-05}
{"train_loss": 0.12150745838880539, "global_step": 135573, "epoch": 1523, "lr": 5.0232473540352034e-05}
{"train_loss": 0.09152474254369736, "global_step": 135574, "epoch": 1523, "lr": 5.0231893809999356e-05}
{"train_loss": 0.13544824719429016, "global_step": 135575, "epoch": 1523, "lr": 5.0231314079615486e-05}
{"train_loss": 0.0922439694404602, "global_step": 135576, "epoch": 1523, "lr": 5.0230734349200527e-05}
{"train_loss": 0.1665906459093094, "global_step": 135577, "epoch": 1523, "lr": 5.023015461875455e-05}
{"train_loss": 0.10696843266487122, "global_step": 135578, "epoch": 1523, "lr": 5.022957488827763e-05}
{"train_loss": 0.1119508370757103, "global_step": 135579, "epoch": 1523, "lr": 5.022899515776985e-05}
{"train_loss": 0.094793900847435, "global_step": 135580, "epoch": 1523, "lr": 5.022841542723128e-05}
{"train_loss": 0.10782045125961304, "global_step": 135581, "epoch": 1523, "lr": 5.022783569666201e-05}
{"train_loss": 0.11874599009752274, "global_step": 135582, "epoch": 1523, "lr": 5.0227255966062106e-05}
{"train_loss": 0.11606034636497498, "global_step": 135583, "epoch": 1523, "lr": 5.022667623543165e-05}
{"train_loss": 0.15031209588050842, "global_step": 135584, "epoch": 1523, "lr": 5.022609650477072e-05}
{"train_loss": 0.1530476212501526, "global_step": 135585, "epoch": 1523, "lr": 5.022551677407938e-05}
{"train_loss": 0.13376018404960632, "global_step": 135586, "epoch": 1523, "lr": 5.022493704335774e-05}
{"train_loss": 0.10613417625427246, "global_step": 135587, "epoch": 1523, "lr": 5.022435731260585e-05}
{"train_loss": 0.11406233161687851, "global_step": 135588, "epoch": 1523, "lr": 5.02237775818238e-05}
{"train_loss": 0.1452123075723648, "global_step": 135589, "epoch": 1523, "lr": 5.0223197851011673e-05}
{"train_loss": 0.10005268454551697, "global_step": 135590, "epoch": 1523, "lr": 5.022261812016954e-05}
{"train_loss": 0.1225368082523346, "global_step": 135591, "epoch": 1523, "lr": 5.022203838929748e-05}
{"train_loss": 0.08379087597131729, "global_step": 135592, "epoch": 1523, "lr": 5.022145865839556e-05}
{"train_loss": 0.07897377759218216, "global_step": 135593, "epoch": 1523, "lr": 5.022087892746387e-05}
{"train_loss": 0.19163814187049866, "global_step": 135594, "epoch": 1523, "lr": 5.0220299196502484e-05}
{"train_loss": 0.166429340839386, "global_step": 135595, "epoch": 1523, "lr": 5.021971946551148e-05}
{"train_loss": 0.1568768322467804, "global_step": 135596, "epoch": 1523, "lr": 5.021913973449095e-05}
{"train_loss": 0.12975627183914185, "global_step": 135597, "epoch": 1523, "lr": 5.021856000344095e-05}
{"train_loss": 0.11153536289930344, "global_step": 135598, "epoch": 1523, "lr": 5.0217980272361565e-05}
{"train_loss": 0.14348328113555908, "global_step": 135599, "epoch": 1523, "lr": 5.021740054125288e-05}
{"train_loss": 0.1497177630662918, "global_step": 135600, "epoch": 1523, "lr": 5.021682081011496e-05}
{"train_loss": 0.17472557723522186, "global_step": 135601, "epoch": 1523, "lr": 5.02162410789479e-05}
{"train_loss": 0.14302915334701538, "global_step": 135602, "epoch": 1523, "lr": 5.021566134775176e-05}
{"train_loss": 0.06766599416732788, "global_step": 135603, "epoch": 1523, "lr": 5.021508161652664e-05}
{"train_loss": 0.11588863283395767, "global_step": 135604, "epoch": 1523, "lr": 5.02145018852726e-05}
{"train_loss": 0.10050787031650543, "global_step": 135605, "epoch": 1523, "lr": 5.0213922153989726e-05}
{"train_loss": 0.14468376338481903, "global_step": 135606, "epoch": 1523, "lr": 5.021334242267809e-05}
{"train_loss": 0.12182910740375519, "global_step": 135607, "epoch": 1523, "lr": 5.0212762691337765e-05}
{"train_loss": 0.1394263505935669, "global_step": 135608, "epoch": 1523, "lr": 5.0212182959968844e-05}
{"train_loss": 0.141938716173172, "global_step": 135609, "epoch": 1523, "lr": 5.0211603228571394e-05}
{"train_loss": 0.14217786490917206, "global_step": 135610, "epoch": 1523, "lr": 5.0211023497145504e-05}
{"train_loss": 0.14668799936771393, "global_step": 135611, "epoch": 1523, "lr": 5.021044376569123e-05}
{"train_loss": 0.12306148558855057, "global_step": 135612, "epoch": 1523, "lr": 5.0209864034208676e-05}
{"train_loss": 0.08699586987495422, "global_step": 135613, "epoch": 1523, "lr": 5.020928430269791e-05}
{"train_loss": 0.059414852410554886, "global_step": 135614, "epoch": 1523, "lr": 5.0208704571159e-05}
{"train_loss": 0.1391167789697647, "global_step": 135615, "epoch": 1523, "lr": 5.0208124839592044e-05}
{"train_loss": 0.12913358211517334, "global_step": 135616, "epoch": 1523, "lr": 5.020754510799709e-05}
{"train_loss": 0.10634005814790726, "global_step": 135617, "epoch": 1523, "lr": 5.020696537637425e-05}
{"train_loss": 0.08101490139961243, "global_step": 135618, "epoch": 1523, "lr": 5.020638564472357e-05}
{"train_loss": 0.08753293007612228, "global_step": 135619, "epoch": 1523, "lr": 5.0205805913045166e-05}
{"train_loss": 0.23211786150932312, "global_step": 135620, "epoch": 1523, "lr": 5.0205226181339084e-05}
{"train_loss": 0.16939149796962738, "global_step": 135621, "epoch": 1523, "lr": 5.020464644960541e-05}
{"train_loss": 0.10809709131717682, "global_step": 135622, "epoch": 1523, "lr": 5.0204066717844224e-05}
{"train_loss": 0.14275750517845154, "global_step": 135623, "epoch": 1523, "lr": 5.02034869860556e-05}
{"train_loss": 0.08750824630260468, "global_step": 135624, "epoch": 1523, "lr": 5.0202907254239615e-05}
{"train_loss": 0.11817793548107147, "global_step": 135625, "epoch": 1523, "lr": 5.020232752239636e-05}
{"train_loss": 0.1304224282503128, "global_step": 135626, "epoch": 1523, "lr": 5.020174779052591e-05}
{"train_loss": 0.18941627442836761, "global_step": 135627, "epoch": 1523, "lr": 5.0201168058628336e-05}
{"train_loss": 0.1822461485862732, "global_step": 135628, "epoch": 1523, "lr": 5.0200588326703714e-05}
{"train_loss": 0.1379213035106659, "global_step": 135629, "epoch": 1523, "lr": 5.020000859475212e-05}
{"train_loss": 0.13180223107337952, "global_step": 135630, "epoch": 1523, "lr": 5.019942886277365e-05}
{"train_loss": 0.12117049098014832, "global_step": 135631, "epoch": 1523, "lr": 5.0198849130768353e-05}
{"train_loss": 0.0653790682554245, "global_step": 135632, "epoch": 1523, "lr": 5.0198269398736334e-05}
{"train_loss": 0.18527314066886902, "global_step": 135633, "epoch": 1523, "lr": 5.019768966667766e-05}
{"train_loss": 0.12771475315093994, "global_step": 135634, "epoch": 1523, "lr": 5.01971099345924e-05}
{"train_loss": 0.12582263622558518, "global_step": 135635, "epoch": 1523, "lr": 5.019653020248065e-05, "val_loss": 5.700037956237793}
{"train_loss": 0.14260132610797882, "global_step": 135636, "epoch": 1524, "lr": 5.0195950470342476e-05}
{"train_loss": 0.1415606141090393, "global_step": 135637, "epoch": 1524, "lr": 5.019537073817796e-05}
{"train_loss": 0.12390544265508652, "global_step": 135638, "epoch": 1524, "lr": 5.019479100598717e-05}
{"train_loss": 0.1582721322774887, "global_step": 135639, "epoch": 1524, "lr": 5.019421127377021e-05}
{"train_loss": 0.1151372492313385, "global_step": 135640, "epoch": 1524, "lr": 5.019363154152713e-05}
{"train_loss": 0.06827916949987411, "global_step": 135641, "epoch": 1524, "lr": 5.019305180925802e-05}
{"train_loss": 0.1263246238231659, "global_step": 135642, "epoch": 1524, "lr": 5.019247207696296e-05}
{"train_loss": 0.11045101284980774, "global_step": 135643, "epoch": 1524, "lr": 5.019189234464202e-05}
{"train_loss": 0.13837845623493195, "global_step": 135644, "epoch": 1524, "lr": 5.0191312612295284e-05}
{"train_loss": 0.15412886440753937, "global_step": 135645, "epoch": 1524, "lr": 5.019073287992283e-05}
{"train_loss": 0.14943824708461761, "global_step": 135646, "epoch": 1524, "lr": 5.019015314752473e-05}
{"train_loss": 0.131611168384552, "global_step": 135647, "epoch": 1524, "lr": 5.018957341510106e-05}
{"train_loss": 0.08354102075099945, "global_step": 135648, "epoch": 1524, "lr": 5.018899368265192e-05}
{"train_loss": 0.08855219185352325, "global_step": 135649, "epoch": 1524, "lr": 5.018841395017736e-05}
{"train_loss": 0.049109507352113724, "global_step": 135650, "epoch": 1524, "lr": 5.018783421767747e-05}
{"train_loss": 0.14374221861362457, "global_step": 135651, "epoch": 1524, "lr": 5.018725448515234e-05}
{"train_loss": 0.10512638092041016, "global_step": 135652, "epoch": 1524, "lr": 5.018667475260203e-05}
{"train_loss": 0.11681704223155975, "global_step": 135653, "epoch": 1524, "lr": 5.018609502002663e-05}
{"train_loss": 0.08873758465051651, "global_step": 135654, "epoch": 1524, "lr": 5.01855152874262e-05}
{"train_loss": 0.15229086577892303, "global_step": 135655, "epoch": 1524, "lr": 5.0184935554800824e-05}
{"train_loss": 0.1466130018234253, "global_step": 135656, "epoch": 1524, "lr": 5.01843558221506e-05}
{"train_loss": 0.12283878028392792, "global_step": 135657, "epoch": 1524, "lr": 5.018377608947557e-05}
{"train_loss": 0.0906498059630394, "global_step": 135658, "epoch": 1524, "lr": 5.018319635677586e-05}
{"train_loss": 0.13190041482448578, "global_step": 135659, "epoch": 1524, "lr": 5.0182616624051524e-05}
{"train_loss": 0.10367947816848755, "global_step": 135660, "epoch": 1524, "lr": 5.018203689130262e-05}
{"train_loss": 0.10488514602184296, "global_step": 135661, "epoch": 1524, "lr": 5.018145715852925e-05}
{"train_loss": 0.057783935219049454, "global_step": 135662, "epoch": 1524, "lr": 5.018087742573149e-05}
{"train_loss": 0.09905743598937988, "global_step": 135663, "epoch": 1524, "lr": 5.0180297692909404e-05}
{"train_loss": 0.22800293564796448, "global_step": 135664, "epoch": 1524, "lr": 5.017971796006308e-05}
{"train_loss": 0.09888722002506256, "global_step": 135665, "epoch": 1524, "lr": 5.01791382271926e-05}
{"train_loss": 0.11751070618629456, "global_step": 135666, "epoch": 1524, "lr": 5.0178558494298036e-05}
{"train_loss": 0.13354791700839996, "global_step": 135667, "epoch": 1524, "lr": 5.0177978761379464e-05}
{"train_loss": 0.08808407932519913, "global_step": 135668, "epoch": 1524, "lr": 5.0177399028436965e-05}
{"train_loss": 0.11764434725046158, "global_step": 135669, "epoch": 1524, "lr": 5.017681929547062e-05}
{"train_loss": 0.1587117314338684, "global_step": 135670, "epoch": 1524, "lr": 5.01762395624805e-05}
{"train_loss": 0.0846513882279396, "global_step": 135671, "epoch": 1524, "lr": 5.017565982946669e-05}
{"train_loss": 0.12517495453357697, "global_step": 135672, "epoch": 1524, "lr": 5.017508009642926e-05}
{"train_loss": 0.16449598968029022, "global_step": 135673, "epoch": 1524, "lr": 5.0174500363368304e-05}
{"train_loss": 0.09204351902008057, "global_step": 135674, "epoch": 1524, "lr": 5.0173920630283876e-05}
{"train_loss": 0.16020698845386505, "global_step": 135675, "epoch": 1524, "lr": 5.017334089717607e-05}
{"train_loss": 0.12814263999462128, "global_step": 135676, "epoch": 1524, "lr": 5.0172761164044954e-05}
{"train_loss": 0.12469235807657242, "global_step": 135677, "epoch": 1524, "lr": 5.017218143089063e-05}
{"train_loss": 0.13316084444522858, "global_step": 135678, "epoch": 1524, "lr": 5.0171601697713157e-05}
{"train_loss": 0.12956921756267548, "global_step": 135679, "epoch": 1524, "lr": 5.01710219645126e-05}
{"train_loss": 0.12199266254901886, "global_step": 135680, "epoch": 1524, "lr": 5.017044223128906e-05}
{"train_loss": 0.07238944619894028, "global_step": 135681, "epoch": 1524, "lr": 5.01698624980426e-05}
{"train_loss": 0.06276867538690567, "global_step": 135682, "epoch": 1524, "lr": 5.016928276477331e-05}
{"train_loss": 0.1003299430012703, "global_step": 135683, "epoch": 1524, "lr": 5.0168703031481255e-05}
{"train_loss": 0.1419849991798401, "global_step": 135684, "epoch": 1524, "lr": 5.016812329816652e-05}
{"train_loss": 0.05634751543402672, "global_step": 135685, "epoch": 1524, "lr": 5.016754356482919e-05}
{"train_loss": 0.12375020235776901, "global_step": 135686, "epoch": 1524, "lr": 5.0166963831469335e-05}
{"train_loss": 0.1337236613035202, "global_step": 135687, "epoch": 1524, "lr": 5.016638409808703e-05}
{"train_loss": 0.12216217070817947, "global_step": 135688, "epoch": 1524, "lr": 5.0165804364682356e-05}
{"train_loss": 0.1866932511329651, "global_step": 135689, "epoch": 1524, "lr": 5.01652246312554e-05}
{"train_loss": 0.0828479751944542, "global_step": 135690, "epoch": 1524, "lr": 5.016464489780622e-05}
{"train_loss": 0.13229452073574066, "global_step": 135691, "epoch": 1524, "lr": 5.0164065164334916e-05}
{"train_loss": 0.09781691431999207, "global_step": 135692, "epoch": 1524, "lr": 5.016348543084155e-05}
{"train_loss": 0.08238435536623001, "global_step": 135693, "epoch": 1524, "lr": 5.016290569732619e-05}
{"train_loss": 0.09295491129159927, "global_step": 135694, "epoch": 1524, "lr": 5.0162325963788945e-05}
{"train_loss": 0.15503869950771332, "global_step": 135695, "epoch": 1524, "lr": 5.016174623022988e-05}
{"train_loss": 0.10006074607372284, "global_step": 135696, "epoch": 1524, "lr": 5.016116649664907e-05}
{"train_loss": 0.11303138732910156, "global_step": 135697, "epoch": 1524, "lr": 5.01605867630466e-05}
{"train_loss": 0.08351839333772659, "global_step": 135698, "epoch": 1524, "lr": 5.016000702942253e-05}
{"train_loss": 0.1837393343448639, "global_step": 135699, "epoch": 1524, "lr": 5.015942729577695e-05}
{"train_loss": 0.06872352212667465, "global_step": 135700, "epoch": 1524, "lr": 5.015884756210994e-05}
{"train_loss": 0.14339132606983185, "global_step": 135701, "epoch": 1524, "lr": 5.015826782842157e-05}
{"train_loss": 0.12772639095783234, "global_step": 135702, "epoch": 1524, "lr": 5.015768809471193e-05}
{"train_loss": 0.12450724840164185, "global_step": 135703, "epoch": 1524, "lr": 5.0157108360981085e-05}
{"train_loss": 0.13397692143917084, "global_step": 135704, "epoch": 1524, "lr": 5.0156528627229114e-05}
{"train_loss": 0.043093323707580566, "global_step": 135705, "epoch": 1524, "lr": 5.015594889345611e-05}
{"train_loss": 0.11351616680622101, "global_step": 135706, "epoch": 1524, "lr": 5.015536915966214e-05}
{"train_loss": 0.1551569700241089, "global_step": 135707, "epoch": 1524, "lr": 5.015478942584727e-05}
{"train_loss": 0.13980714976787567, "global_step": 135708, "epoch": 1524, "lr": 5.01542096920116e-05}
{"train_loss": 0.12911248207092285, "global_step": 135709, "epoch": 1524, "lr": 5.01536299581552e-05}
{"train_loss": 0.1278475522994995, "global_step": 135710, "epoch": 1524, "lr": 5.0153050224278145e-05}
{"train_loss": 0.08378201723098755, "global_step": 135711, "epoch": 1524, "lr": 5.01524704903805e-05}
{"train_loss": 0.08537128567695618, "global_step": 135712, "epoch": 1524, "lr": 5.015189075646237e-05}
{"train_loss": 0.0954902246594429, "global_step": 135713, "epoch": 1524, "lr": 5.015131102252382e-05}
{"train_loss": 0.14682839810848236, "global_step": 135714, "epoch": 1524, "lr": 5.0150731288564934e-05}
{"train_loss": 0.1380946934223175, "global_step": 135715, "epoch": 1524, "lr": 5.0150151554585775e-05}
{"train_loss": 0.05131502076983452, "global_step": 135716, "epoch": 1524, "lr": 5.0149571820586436e-05}
{"train_loss": 0.08554337173700333, "global_step": 135717, "epoch": 1524, "lr": 5.014899208656699e-05}
{"train_loss": 0.12996113300323486, "global_step": 135718, "epoch": 1524, "lr": 5.014841235252752e-05}
{"train_loss": 0.1168801560997963, "global_step": 135719, "epoch": 1524, "lr": 5.014783261846808e-05}
{"train_loss": 0.14005735516548157, "global_step": 135720, "epoch": 1524, "lr": 5.0147252884388776e-05}
{"train_loss": 0.08128020167350769, "global_step": 135721, "epoch": 1524, "lr": 5.014667315028967e-05}
{"train_loss": 0.12669269740581512, "global_step": 135722, "epoch": 1524, "lr": 5.0146093416170846e-05}
{"train_loss": 0.11342283338308334, "global_step": 135723, "epoch": 1524, "lr": 5.0145513682032384e-05}
{"train_loss": 0.1171477512362298, "global_step": 135724, "epoch": 1524, "lr": 5.014493394787436e-05, "val_loss": 5.765751361846924}
{"train_loss": 0.08263593912124634, "global_step": 135725, "epoch": 1525, "lr": 5.014435421369685e-05}
{"train_loss": 0.0994013100862503, "global_step": 135726, "epoch": 1525, "lr": 5.014377447949994e-05}
{"train_loss": 0.12572309374809265, "global_step": 135727, "epoch": 1525, "lr": 5.014319474528369e-05}
{"train_loss": 0.1233898401260376, "global_step": 135728, "epoch": 1525, "lr": 5.01426150110482e-05}
{"train_loss": 0.09326008707284927, "global_step": 135729, "epoch": 1525, "lr": 5.014203527679353e-05}
{"train_loss": 0.10919079184532166, "global_step": 135730, "epoch": 1525, "lr": 5.0141455542519755e-05}
{"train_loss": 0.12717759609222412, "global_step": 135731, "epoch": 1525, "lr": 5.0140875808226984e-05}
{"train_loss": 0.15790463984012604, "global_step": 135732, "epoch": 1525, "lr": 5.014029607391526e-05}
{"train_loss": 0.08937805145978928, "global_step": 135733, "epoch": 1525, "lr": 5.0139716339584684e-05}
{"train_loss": 0.11373975872993469, "global_step": 135734, "epoch": 1525, "lr": 5.013913660523532e-05}
{"train_loss": 0.14605613052845, "global_step": 135735, "epoch": 1525, "lr": 5.0138556870867256e-05}
{"train_loss": 0.1196979284286499, "global_step": 135736, "epoch": 1525, "lr": 5.0137977136480566e-05}
{"train_loss": 0.13019460439682007, "global_step": 135737, "epoch": 1525, "lr": 5.0137397402075316e-05}
{"train_loss": 0.11425816267728806, "global_step": 135738, "epoch": 1525, "lr": 5.01368176676516e-05}
{"train_loss": 0.079107865691185, "global_step": 135739, "epoch": 1525, "lr": 5.0136237933209494e-05}
{"train_loss": 0.10064190626144409, "global_step": 135740, "epoch": 1525, "lr": 5.013565819874907e-05}
{"train_loss": 0.08106464892625809, "global_step": 135741, "epoch": 1525, "lr": 5.01350784642704e-05}
{"train_loss": 0.12533411383628845, "global_step": 135742, "epoch": 1525, "lr": 5.013449872977358e-05}
{"train_loss": 0.12239590287208557, "global_step": 135743, "epoch": 1525, "lr": 5.013391899525868e-05}
{"train_loss": 0.08159834891557693, "global_step": 135744, "epoch": 1525, "lr": 5.013333926072578e-05}
{"train_loss": 0.21013788878917694, "global_step": 135745, "epoch": 1525, "lr": 5.013275952617493e-05}
{"train_loss": 0.13423572480678558, "global_step": 135746, "epoch": 1525, "lr": 5.013217979160626e-05}
{"train_loss": 0.11147429794073105, "global_step": 135747, "epoch": 1525, "lr": 5.01316000570198e-05}
{"train_loss": 0.14161254465579987, "global_step": 135748, "epoch": 1525, "lr": 5.0131020322415644e-05}
{"train_loss": 0.1953422874212265, "global_step": 135749, "epoch": 1525, "lr": 5.013044058779389e-05}
{"train_loss": 0.12582501769065857, "global_step": 135750, "epoch": 1525, "lr": 5.0129860853154596e-05}
{"train_loss": 0.13084283471107483, "global_step": 135751, "epoch": 1525, "lr": 5.012928111849784e-05}
{"train_loss": 0.1005229577422142, "global_step": 135752, "epoch": 1525, "lr": 5.0128701383823706e-05}
{"train_loss": 0.1317020207643509, "global_step": 135753, "epoch": 1525, "lr": 5.0128121649132274e-05}
{"train_loss": 0.13544000685214996, "global_step": 135754, "epoch": 1525, "lr": 5.0127541914423617e-05}
{"train_loss": 0.08809351921081543, "global_step": 135755, "epoch": 1525, "lr": 5.012696217969782e-05}
{"train_loss": 0.13563084602355957, "global_step": 135756, "epoch": 1525, "lr": 5.0126382444954944e-05}
{"train_loss": 0.08954521268606186, "global_step": 135757, "epoch": 1525, "lr": 5.012580271019508e-05}
{"train_loss": 0.10125256329774857, "global_step": 135758, "epoch": 1525, "lr": 5.0125222975418306e-05}
{"train_loss": 0.05555380508303642, "global_step": 135759, "epoch": 1525, "lr": 5.012464324062469e-05}
{"train_loss": 0.05365977808833122, "global_step": 135760, "epoch": 1525, "lr": 5.0124063505814326e-05}
{"train_loss": 0.10856661200523376, "global_step": 135761, "epoch": 1525, "lr": 5.012348377098728e-05}
{"train_loss": 0.08627650886774063, "global_step": 135762, "epoch": 1525, "lr": 5.0122904036143626e-05}
{"train_loss": 0.11695726215839386, "global_step": 135763, "epoch": 1525, "lr": 5.012232430128345e-05}
{"train_loss": 0.13667313754558563, "global_step": 135764, "epoch": 1525, "lr": 5.012174456640684e-05}
{"train_loss": 0.07573175430297852, "global_step": 135765, "epoch": 1525, "lr": 5.012116483151386e-05}
{"train_loss": 0.1586141139268875, "global_step": 135766, "epoch": 1525, "lr": 5.012058509660458e-05}
{"train_loss": 0.09182805567979813, "global_step": 135767, "epoch": 1525, "lr": 5.012000536167909e-05}
{"train_loss": 0.13294176757335663, "global_step": 135768, "epoch": 1525, "lr": 5.0119425626737474e-05}
{"train_loss": 0.11619232594966888, "global_step": 135769, "epoch": 1525, "lr": 5.0118845891779795e-05}
{"train_loss": 0.17419354617595673, "global_step": 135770, "epoch": 1525, "lr": 5.0118266156806146e-05}
{"train_loss": 0.10006069391965866, "global_step": 135771, "epoch": 1525, "lr": 5.01176864218166e-05}
{"train_loss": 0.07094302028417587, "global_step": 135772, "epoch": 1525, "lr": 5.0117106686811236e-05}
{"train_loss": 0.09605469554662704, "global_step": 135773, "epoch": 1525, "lr": 5.011652695179011e-05}
{"train_loss": 0.07821240276098251, "global_step": 135774, "epoch": 1525, "lr": 5.011594721675335e-05}
{"train_loss": 0.1208137720823288, "global_step": 135775, "epoch": 1525, "lr": 5.0115367481700984e-05}
{"train_loss": 0.11087041348218918, "global_step": 135776, "epoch": 1525, "lr": 5.0114787746633096e-05}
{"train_loss": 0.09920153021812439, "global_step": 135777, "epoch": 1525, "lr": 5.011420801154979e-05}
{"train_loss": 0.1421271413564682, "global_step": 135778, "epoch": 1525, "lr": 5.011362827645113e-05}
{"train_loss": 0.12919333577156067, "global_step": 135779, "epoch": 1525, "lr": 5.0113048541337195e-05}
{"train_loss": 0.17160648107528687, "global_step": 135780, "epoch": 1525, "lr": 5.0112468806208044e-05}
{"train_loss": 0.21325792372226715, "global_step": 135781, "epoch": 1525, "lr": 5.011188907106379e-05}
{"train_loss": 0.14711454510688782, "global_step": 135782, "epoch": 1525, "lr": 5.01113093359045e-05}
{"train_loss": 0.09989102929830551, "global_step": 135783, "epoch": 1525, "lr": 5.0110729600730234e-05}
{"train_loss": 0.12125340104103088, "global_step": 135784, "epoch": 1525, "lr": 5.0110149865541065e-05}
{"train_loss": 0.14253877103328705, "global_step": 135785, "epoch": 1525, "lr": 5.010957013033711e-05}
{"train_loss": 0.13674691319465637, "global_step": 135786, "epoch": 1525, "lr": 5.0108990395118424e-05}
{"train_loss": 0.14933539927005768, "global_step": 135787, "epoch": 1525, "lr": 5.010841065988509e-05}
{"train_loss": 0.08778040111064911, "global_step": 135788, "epoch": 1525, "lr": 5.010783092463717e-05}
{"train_loss": 0.08608601242303848, "global_step": 135789, "epoch": 1525, "lr": 5.010725118937476e-05}
{"train_loss": 0.07522799074649811, "global_step": 135790, "epoch": 1525, "lr": 5.010667145409792e-05}
{"train_loss": 0.15376920998096466, "global_step": 135791, "epoch": 1525, "lr": 5.0106091718806755e-05}
{"train_loss": 0.1073303073644638, "global_step": 135792, "epoch": 1525, "lr": 5.010551198350132e-05}
{"train_loss": 0.14506253600120544, "global_step": 135793, "epoch": 1525, "lr": 5.0104932248181714e-05}
{"train_loss": 0.0946575179696083, "global_step": 135794, "epoch": 1525, "lr": 5.0104352512847986e-05}
{"train_loss": 0.11841844767332077, "global_step": 135795, "epoch": 1525, "lr": 5.010377277750023e-05}
{"train_loss": 0.09481660276651382, "global_step": 135796, "epoch": 1525, "lr": 5.010319304213853e-05}
{"train_loss": 0.16521158814430237, "global_step": 135797, "epoch": 1525, "lr": 5.010261330676295e-05}
{"train_loss": 0.14026211202144623, "global_step": 135798, "epoch": 1525, "lr": 5.0102033571373575e-05}
{"train_loss": 0.15225626528263092, "global_step": 135799, "epoch": 1525, "lr": 5.010145383597048e-05}
{"train_loss": 0.11103329807519913, "global_step": 135800, "epoch": 1525, "lr": 5.010087410055375e-05}
{"train_loss": 0.11679589003324509, "global_step": 135801, "epoch": 1525, "lr": 5.0100294365123454e-05}
{"train_loss": 0.090780109167099, "global_step": 135802, "epoch": 1525, "lr": 5.0099714629679674e-05}
{"train_loss": 0.13621535897254944, "global_step": 135803, "epoch": 1525, "lr": 5.00991348942225e-05}
{"train_loss": 0.11943470686674118, "global_step": 135804, "epoch": 1525, "lr": 5.009855515875199e-05}
{"train_loss": 0.10512470453977585, "global_step": 135805, "epoch": 1525, "lr": 5.009797542326823e-05}
{"train_loss": 0.09119799733161926, "global_step": 135806, "epoch": 1525, "lr": 5.00973956877713e-05}
{"train_loss": 0.118568055331707, "global_step": 135807, "epoch": 1525, "lr": 5.009681595226128e-05}
{"train_loss": 0.10411397367715836, "global_step": 135808, "epoch": 1525, "lr": 5.0096236216738243e-05}
{"train_loss": 0.12135818600654602, "global_step": 135809, "epoch": 1525, "lr": 5.009565648120227e-05}
{"train_loss": 0.07838074117898941, "global_step": 135810, "epoch": 1525, "lr": 5.009507674565343e-05}
{"train_loss": 0.0941048264503479, "global_step": 135811, "epoch": 1525, "lr": 5.009449701009181e-05}
{"train_loss": 0.1012924313545227, "global_step": 135812, "epoch": 1525, "lr": 5.009391727451749e-05}
{"train_loss": 0.11667584694838257, "global_step": 135813, "epoch": 1525, "lr": 5.009333753893055e-05, "val_loss": 5.826213836669922, "train_action_mse_error": 20.329299926757812}
{"train_loss": 0.08287303149700165, "global_step": 135814, "epoch": 1526, "lr": 5.009275780333106e-05}
{"train_loss": 0.12012455612421036, "global_step": 135815, "epoch": 1526, "lr": 5.00921780677191e-05}
{"train_loss": 0.11470475047826767, "global_step": 135816, "epoch": 1526, "lr": 5.009159833209474e-05}
{"train_loss": 0.09628718346357346, "global_step": 135817, "epoch": 1526, "lr": 5.009101859645806e-05}
{"train_loss": 0.20150582492351532, "global_step": 135818, "epoch": 1526, "lr": 5.0090438860809153e-05}
{"train_loss": 0.17535409331321716, "global_step": 135819, "epoch": 1526, "lr": 5.0089859125148085e-05}
{"train_loss": 0.18130148947238922, "global_step": 135820, "epoch": 1526, "lr": 5.008927938947493e-05}
{"train_loss": 0.18498975038528442, "global_step": 135821, "epoch": 1526, "lr": 5.008869965378977e-05}
{"train_loss": 0.08238983154296875, "global_step": 135822, "epoch": 1526, "lr": 5.00881199180927e-05}
{"train_loss": 0.16172832250595093, "global_step": 135823, "epoch": 1526, "lr": 5.0087540182383785e-05}
{"train_loss": 0.15360844135284424, "global_step": 135824, "epoch": 1526, "lr": 5.0086960446663086e-05}
{"train_loss": 0.16230836510658264, "global_step": 135825, "epoch": 1526, "lr": 5.008638071093071e-05}
{"train_loss": 0.0670211911201477, "global_step": 135826, "epoch": 1526, "lr": 5.0085800975186715e-05}
{"train_loss": 0.15746192634105682, "global_step": 135827, "epoch": 1526, "lr": 5.008522123943119e-05}
{"train_loss": 0.11506661772727966, "global_step": 135828, "epoch": 1526, "lr": 5.00846415036642e-05}
{"train_loss": 0.15861448645591736, "global_step": 135829, "epoch": 1526, "lr": 5.008406176788584e-05}
{"train_loss": 0.12496000528335571, "global_step": 135830, "epoch": 1526, "lr": 5.008348203209617e-05}
{"train_loss": 0.13792668282985687, "global_step": 135831, "epoch": 1526, "lr": 5.008290229629529e-05}
{"train_loss": 0.16064755618572235, "global_step": 135832, "epoch": 1526, "lr": 5.0082322560483267e-05}
{"train_loss": 0.09416462481021881, "global_step": 135833, "epoch": 1526, "lr": 5.008174282466016e-05}
{"train_loss": 0.11036764830350876, "global_step": 135834, "epoch": 1526, "lr": 5.0081163088826065e-05}
{"train_loss": 0.11444088071584702, "global_step": 135835, "epoch": 1526, "lr": 5.008058335298107e-05}
{"train_loss": 0.10954741388559341, "global_step": 135836, "epoch": 1526, "lr": 5.0080003617125223e-05}
{"train_loss": 0.0979573130607605, "global_step": 135837, "epoch": 1526, "lr": 5.007942388125864e-05}
{"train_loss": 0.0864131972193718, "global_step": 135838, "epoch": 1526, "lr": 5.007884414538135e-05}
{"train_loss": 0.10895784199237823, "global_step": 135839, "epoch": 1526, "lr": 5.007826440949349e-05}
{"train_loss": 0.1417030543088913, "global_step": 135840, "epoch": 1526, "lr": 5.0077684673595104e-05}
{"train_loss": 0.07583659887313843, "global_step": 135841, "epoch": 1526, "lr": 5.007710493768627e-05}
{"train_loss": 0.10229358822107315, "global_step": 135842, "epoch": 1526, "lr": 5.0076525201767066e-05}
{"train_loss": 0.11330779641866684, "global_step": 135843, "epoch": 1526, "lr": 5.007594546583758e-05}
{"train_loss": 0.1453532576560974, "global_step": 135844, "epoch": 1526, "lr": 5.007536572989788e-05}
{"train_loss": 0.04305360093712807, "global_step": 135845, "epoch": 1526, "lr": 5.007478599394806e-05}
{"train_loss": 0.0961083248257637, "global_step": 135846, "epoch": 1526, "lr": 5.0074206257988176e-05}
{"train_loss": 0.14114663004875183, "global_step": 135847, "epoch": 1526, "lr": 5.0073626522018313e-05}
{"train_loss": 0.04322449862957001, "global_step": 135848, "epoch": 1526, "lr": 5.007304678603856e-05}
{"train_loss": 0.07273965328931808, "global_step": 135849, "epoch": 1526, "lr": 5.0072467050048976e-05}
{"train_loss": 0.16616296768188477, "global_step": 135850, "epoch": 1526, "lr": 5.007188731404966e-05}
{"train_loss": 0.04796295613050461, "global_step": 135851, "epoch": 1526, "lr": 5.007130757804068e-05}
{"train_loss": 0.13713575899600983, "global_step": 135852, "epoch": 1526, "lr": 5.007072784202211e-05}
{"train_loss": 0.06855789572000504, "global_step": 135853, "epoch": 1526, "lr": 5.007014810599403e-05}
{"train_loss": 0.05310971289873123, "global_step": 135854, "epoch": 1526, "lr": 5.006956836995652e-05}
{"train_loss": 0.1424572914838791, "global_step": 135855, "epoch": 1526, "lr": 5.0068988633909664e-05}
{"train_loss": 0.10328223556280136, "global_step": 135856, "epoch": 1526, "lr": 5.006840889785353e-05}
{"train_loss": 0.13816547393798828, "global_step": 135857, "epoch": 1526, "lr": 5.006782916178817e-05}
{"train_loss": 0.07905027270317078, "global_step": 135858, "epoch": 1526, "lr": 5.0067249425713726e-05}
{"train_loss": 0.08767133206129074, "global_step": 135859, "epoch": 1526, "lr": 5.006666968963023e-05}
{"train_loss": 0.11430945247411728, "global_step": 135860, "epoch": 1526, "lr": 5.0066089953537774e-05}
{"train_loss": 0.1319429725408554, "global_step": 135861, "epoch": 1526, "lr": 5.006551021743643e-05}
{"train_loss": 0.13310818374156952, "global_step": 135862, "epoch": 1526, "lr": 5.006493048132628e-05}
{"train_loss": 0.06398472189903259, "global_step": 135863, "epoch": 1526, "lr": 5.00643507452074e-05}
{"train_loss": 0.08791455626487732, "global_step": 135864, "epoch": 1526, "lr": 5.0063771009079875e-05}
{"train_loss": 0.17036190629005432, "global_step": 135865, "epoch": 1526, "lr": 5.006319127294378e-05}
{"train_loss": 0.06211988255381584, "global_step": 135866, "epoch": 1526, "lr": 5.0062611536799176e-05}
{"train_loss": 0.10955530405044556, "global_step": 135867, "epoch": 1526, "lr": 5.006203180064616e-05}
{"train_loss": 0.10570985823869705, "global_step": 135868, "epoch": 1526, "lr": 5.0061452064484805e-05}
{"train_loss": 0.11268536001443863, "global_step": 135869, "epoch": 1526, "lr": 5.006087232831519e-05}
{"train_loss": 0.10509189963340759, "global_step": 135870, "epoch": 1526, "lr": 5.006029259213739e-05}
{"train_loss": 0.11769893765449524, "global_step": 135871, "epoch": 1526, "lr": 5.0059712855951503e-05}
{"train_loss": 0.11871934682130814, "global_step": 135872, "epoch": 1526, "lr": 5.0059133119757564e-05}
{"train_loss": 0.21486076712608337, "global_step": 135873, "epoch": 1526, "lr": 5.005855338355568e-05}
{"train_loss": 0.06836085021495819, "global_step": 135874, "epoch": 1526, "lr": 5.0057973647345926e-05}
{"train_loss": 0.12728382647037506, "global_step": 135875, "epoch": 1526, "lr": 5.005739391112837e-05}
{"train_loss": 0.09716346114873886, "global_step": 135876, "epoch": 1526, "lr": 5.00568141749031e-05}
{"train_loss": 0.11310503631830215, "global_step": 135877, "epoch": 1526, "lr": 5.00562344386702e-05}
{"train_loss": 0.11890584975481033, "global_step": 135878, "epoch": 1526, "lr": 5.005565470242974e-05}
{"train_loss": 0.12377747893333435, "global_step": 135879, "epoch": 1526, "lr": 5.005507496618179e-05}
{"train_loss": 0.17426574230194092, "global_step": 135880, "epoch": 1526, "lr": 5.005449522992644e-05}
{"train_loss": 0.06759458035230637, "global_step": 135881, "epoch": 1526, "lr": 5.005391549366377e-05}
{"train_loss": 0.08895019441843033, "global_step": 135882, "epoch": 1526, "lr": 5.005333575739385e-05}
{"train_loss": 0.12843577563762665, "global_step": 135883, "epoch": 1526, "lr": 5.005275602111675e-05}
{"train_loss": 0.1522260159254074, "global_step": 135884, "epoch": 1526, "lr": 5.005217628483257e-05}
{"train_loss": 0.09818992018699646, "global_step": 135885, "epoch": 1526, "lr": 5.0051596548541365e-05}
{"train_loss": 0.08511433750391006, "global_step": 135886, "epoch": 1526, "lr": 5.0051016812243233e-05}
{"train_loss": 0.1120045930147171, "global_step": 135887, "epoch": 1526, "lr": 5.005043707593824e-05}
{"train_loss": 0.1666751652956009, "global_step": 135888, "epoch": 1526, "lr": 5.004985733962647e-05}
{"train_loss": 0.10220107436180115, "global_step": 135889, "epoch": 1526, "lr": 5.0049277603307984e-05}
{"train_loss": 0.2051650583744049, "global_step": 135890, "epoch": 1526, "lr": 5.00486978669829e-05}
{"train_loss": 0.14893898367881775, "global_step": 135891, "epoch": 1526, "lr": 5.004811813065124e-05}
{"train_loss": 0.11715622246265411, "global_step": 135892, "epoch": 1526, "lr": 5.004753839431313e-05}
{"train_loss": 0.056270208209753036, "global_step": 135893, "epoch": 1526, "lr": 5.00469586579686e-05}
{"train_loss": 0.10156046599149704, "global_step": 135894, "epoch": 1526, "lr": 5.0046378921617785e-05}
{"train_loss": 0.088591068983078, "global_step": 135895, "epoch": 1526, "lr": 5.004579918526072e-05}
{"train_loss": 0.126275435090065, "global_step": 135896, "epoch": 1526, "lr": 5.004521944889751e-05}
{"train_loss": 0.1672210395336151, "global_step": 135897, "epoch": 1526, "lr": 5.004463971252822e-05}
{"train_loss": 0.14843450486660004, "global_step": 135898, "epoch": 1526, "lr": 5.004405997615293e-05}
{"train_loss": 0.08707041293382645, "global_step": 135899, "epoch": 1526, "lr": 5.00434802397717e-05}
{"train_loss": 0.11442530900239944, "global_step": 135900, "epoch": 1526, "lr": 5.0042900503384635e-05}
{"train_loss": 0.12179668247699738, "global_step": 135901, "epoch": 1526, "lr": 5.0042320766991805e-05}
{"train_loss": 0.1173359670079826, "global_step": 135902, "epoch": 1526, "lr": 5.004174103059328e-05, "val_loss": 5.929417133331299}
{"train_loss": 0.12720225751399994, "global_step": 135903, "epoch": 1527, "lr": 5.004116129418916e-05}
{"train_loss": 0.16612277925014496, "global_step": 135904, "epoch": 1527, "lr": 5.004058155777949e-05}
{"train_loss": 0.05809497833251953, "global_step": 135905, "epoch": 1527, "lr": 5.004000182136436e-05}
{"train_loss": 0.11869736760854721, "global_step": 135906, "epoch": 1527, "lr": 5.0039422084943867e-05}
{"train_loss": 0.09422498196363449, "global_step": 135907, "epoch": 1527, "lr": 5.003884234851807e-05}
{"train_loss": 0.13956576585769653, "global_step": 135908, "epoch": 1527, "lr": 5.003826261208704e-05}
{"train_loss": 0.13652855157852173, "global_step": 135909, "epoch": 1527, "lr": 5.0037682875650885e-05}
{"train_loss": 0.09262363612651825, "global_step": 135910, "epoch": 1527, "lr": 5.003710313920965e-05}
{"train_loss": 0.14583203196525574, "global_step": 135911, "epoch": 1527, "lr": 5.0036523402763425e-05}
{"train_loss": 0.10055924206972122, "global_step": 135912, "epoch": 1527, "lr": 5.003594366631229e-05}
{"train_loss": 0.09046857804059982, "global_step": 135913, "epoch": 1527, "lr": 5.003536392985634e-05}
{"train_loss": 0.15408313274383545, "global_step": 135914, "epoch": 1527, "lr": 5.003478419339561e-05}
{"train_loss": 0.09821753203868866, "global_step": 135915, "epoch": 1527, "lr": 5.003420445693022e-05}
{"train_loss": 0.17078013718128204, "global_step": 135916, "epoch": 1527, "lr": 5.003362472046024e-05}
{"train_loss": 0.07038743793964386, "global_step": 135917, "epoch": 1527, "lr": 5.003304498398572e-05}
{"train_loss": 0.13445670902729034, "global_step": 135918, "epoch": 1527, "lr": 5.0032465247506764e-05}
{"train_loss": 0.13665857911109924, "global_step": 135919, "epoch": 1527, "lr": 5.0031885511023444e-05}
{"train_loss": 0.18710631132125854, "global_step": 135920, "epoch": 1527, "lr": 5.003130577453584e-05}
{"train_loss": 0.05750486999750137, "global_step": 135921, "epoch": 1527, "lr": 5.003072603804403e-05}
{"train_loss": 0.12856550514698029, "global_step": 135922, "epoch": 1527, "lr": 5.0030146301548086e-05}
{"train_loss": 0.10466843843460083, "global_step": 135923, "epoch": 1527, "lr": 5.002956656504809e-05}
{"train_loss": 0.10789970308542252, "global_step": 135924, "epoch": 1527, "lr": 5.002898682854412e-05}
{"train_loss": 0.13009381294250488, "global_step": 135925, "epoch": 1527, "lr": 5.0028407092036245e-05}
{"train_loss": 0.13740895688533783, "global_step": 135926, "epoch": 1527, "lr": 5.002782735552456e-05}
{"train_loss": 0.11553703248500824, "global_step": 135927, "epoch": 1527, "lr": 5.002724761900913e-05}
{"train_loss": 0.1450483500957489, "global_step": 135928, "epoch": 1527, "lr": 5.002666788249004e-05}
{"train_loss": 0.14743466675281525, "global_step": 135929, "epoch": 1527, "lr": 5.0026088145967364e-05}
{"train_loss": 0.07416950166225433, "global_step": 135930, "epoch": 1527, "lr": 5.002550840944118e-05}
{"train_loss": 0.09584363549947739, "global_step": 135931, "epoch": 1527, "lr": 5.002492867291156e-05}
{"train_loss": 0.13686271011829376, "global_step": 135932, "epoch": 1527, "lr": 5.00243489363786e-05}
{"train_loss": 0.09163211286067963, "global_step": 135933, "epoch": 1527, "lr": 5.002376919984236e-05}
{"train_loss": 0.07665600627660751, "global_step": 135934, "epoch": 1527, "lr": 5.002318946330292e-05}
{"train_loss": 0.09892522543668747, "global_step": 135935, "epoch": 1527, "lr": 5.002260972676037e-05}
{"train_loss": 0.12998032569885254, "global_step": 135936, "epoch": 1527, "lr": 5.002202999021478e-05}
{"train_loss": 0.15235652029514313, "global_step": 135937, "epoch": 1527, "lr": 5.0021450253666234e-05}
{"train_loss": 0.16679750382900238, "global_step": 135938, "epoch": 1527, "lr": 5.002087051711479e-05}
{"train_loss": 0.17700673639774323, "global_step": 135939, "epoch": 1527, "lr": 5.002029078056055e-05}
{"train_loss": 0.12581564486026764, "global_step": 135940, "epoch": 1527, "lr": 5.001971104400358e-05}
{"train_loss": 0.15003816783428192, "global_step": 135941, "epoch": 1527, "lr": 5.0019131307443966e-05}
{"train_loss": 0.13400417566299438, "global_step": 135942, "epoch": 1527, "lr": 5.001855157088176e-05}
{"train_loss": 0.10988495498895645, "global_step": 135943, "epoch": 1527, "lr": 5.001797183431708e-05}
{"train_loss": 0.10457221418619156, "global_step": 135944, "epoch": 1527, "lr": 5.0017392097749985e-05}
{"train_loss": 0.12736281752586365, "global_step": 135945, "epoch": 1527, "lr": 5.0016812361180546e-05}
{"train_loss": 0.08295060694217682, "global_step": 135946, "epoch": 1527, "lr": 5.0016232624608836e-05}
{"train_loss": 0.0775180533528328, "global_step": 135947, "epoch": 1527, "lr": 5.001565288803496e-05}
{"train_loss": 0.10574491322040558, "global_step": 135948, "epoch": 1527, "lr": 5.001507315145897e-05}
{"train_loss": 0.12609288096427917, "global_step": 135949, "epoch": 1527, "lr": 5.001449341488096e-05}
{"train_loss": 0.09766680002212524, "global_step": 135950, "epoch": 1527, "lr": 5.0013913678301e-05}
{"train_loss": 0.14604230225086212, "global_step": 135951, "epoch": 1527, "lr": 5.0013333941719164e-05}
{"train_loss": 0.1335531622171402, "global_step": 135952, "epoch": 1527, "lr": 5.0012754205135534e-05}
{"train_loss": 0.12049520760774612, "global_step": 135953, "epoch": 1527, "lr": 5.00121744685502e-05}
{"train_loss": 0.08355120569467545, "global_step": 135954, "epoch": 1527, "lr": 5.001159473196322e-05}
{"train_loss": 0.09380709379911423, "global_step": 135955, "epoch": 1527, "lr": 5.0011014995374686e-05}
{"train_loss": 0.08365370333194733, "global_step": 135956, "epoch": 1527, "lr": 5.001043525878467e-05}
{"train_loss": 0.1412261575460434, "global_step": 135957, "epoch": 1527, "lr": 5.000985552219325e-05}
{"train_loss": 0.18721243739128113, "global_step": 135958, "epoch": 1527, "lr": 5.000927578560051e-05}
{"train_loss": 0.09422892332077026, "global_step": 135959, "epoch": 1527, "lr": 5.000869604900651e-05}
{"train_loss": 0.10863952338695526, "global_step": 135960, "epoch": 1527, "lr": 5.0008116312411355e-05}
{"train_loss": 0.1330309510231018, "global_step": 135961, "epoch": 1527, "lr": 5.000753657581511e-05}
{"train_loss": 0.12369338423013687, "global_step": 135962, "epoch": 1527, "lr": 5.000695683921783e-05}
{"train_loss": 0.1204860657453537, "global_step": 135963, "epoch": 1527, "lr": 5.000637710261964e-05}
{"train_loss": 0.07181350886821747, "global_step": 135964, "epoch": 1527, "lr": 5.000579736602058e-05}
{"train_loss": 0.09645968675613403, "global_step": 135965, "epoch": 1527, "lr": 5.0005217629420744e-05}
{"train_loss": 0.15659181773662567, "global_step": 135966, "epoch": 1527, "lr": 5.000463789282021e-05}
{"train_loss": 0.07507001608610153, "global_step": 135967, "epoch": 1527, "lr": 5.000405815621905e-05}
{"train_loss": 0.13661997020244598, "global_step": 135968, "epoch": 1527, "lr": 5.000347841961735e-05}
{"train_loss": 0.1630864292383194, "global_step": 135969, "epoch": 1527, "lr": 5.0002898683015164e-05}
{"train_loss": 0.11906947195529938, "global_step": 135970, "epoch": 1527, "lr": 5.0002318946412595e-05}
{"train_loss": 0.07770200073719025, "global_step": 135971, "epoch": 1527, "lr": 5.000173920980973e-05}
{"train_loss": 0.10765910148620605, "global_step": 135972, "epoch": 1527, "lr": 5.000115947320662e-05}
{"train_loss": 0.0810774490237236, "global_step": 135973, "epoch": 1527, "lr": 5.000057973660335e-05}
{"train_loss": 0.15871335566043854, "global_step": 135974, "epoch": 1527, "lr": 5e-05}
{"train_loss": 0.059490740299224854, "global_step": 135975, "epoch": 1527, "lr": 4.9999420263396654e-05}
{"train_loss": 0.17254053056240082, "global_step": 135976, "epoch": 1527, "lr": 4.9998840526793394e-05}
{"train_loss": 0.10385909676551819, "global_step": 135977, "epoch": 1527, "lr": 4.9998260790190276e-05}
{"train_loss": 0.1580759733915329, "global_step": 135978, "epoch": 1527, "lr": 4.999768105358741e-05}
{"train_loss": 0.11811947822570801, "global_step": 135979, "epoch": 1527, "lr": 4.9997101316984834e-05}
{"train_loss": 0.16439352929592133, "global_step": 135980, "epoch": 1527, "lr": 4.999652158038267e-05}
{"train_loss": 0.1157560721039772, "global_step": 135981, "epoch": 1527, "lr": 4.999594184378095e-05}
{"train_loss": 0.14815981686115265, "global_step": 135982, "epoch": 1527, "lr": 4.9995362107179804e-05}
{"train_loss": 0.12707450985908508, "global_step": 135983, "epoch": 1527, "lr": 4.999478237057926e-05}
{"train_loss": 0.14569684863090515, "global_step": 135984, "epoch": 1527, "lr": 4.999420263397943e-05}
{"train_loss": 0.10342667251825333, "global_step": 135985, "epoch": 1527, "lr": 4.999362289738037e-05}
{"train_loss": 0.12129534780979156, "global_step": 135986, "epoch": 1527, "lr": 4.999304316078217e-05}
{"train_loss": 0.0899115577340126, "global_step": 135987, "epoch": 1527, "lr": 4.999246342418491e-05}
{"train_loss": 0.07242771983146667, "global_step": 135988, "epoch": 1527, "lr": 4.999188368758867e-05}
{"train_loss": 0.09623987227678299, "global_step": 135989, "epoch": 1527, "lr": 4.9991303950993494e-05}
{"train_loss": 0.16253462433815002, "global_step": 135990, "epoch": 1527, "lr": 4.999072421439952e-05}
{"train_loss": 0.11932460637239928, "global_step": 135991, "epoch": 1527, "lr": 4.999014447780676e-05, "val_loss": 5.835041046142578}
{"train_loss": 0.10734739899635315, "global_step": 135992, "epoch": 1528, "lr": 4.998956474121536e-05}
{"train_loss": 0.04307558387517929, "global_step": 135993, "epoch": 1528, "lr": 4.9988985004625326e-05}
{"train_loss": 0.10706428438425064, "global_step": 135994, "epoch": 1528, "lr": 4.9988405268036784e-05}
{"train_loss": 0.1761832982301712, "global_step": 135995, "epoch": 1528, "lr": 4.9987825531449814e-05}
{"train_loss": 0.17018714547157288, "global_step": 135996, "epoch": 1528, "lr": 4.9987245794864464e-05}
{"train_loss": 0.14743520319461823, "global_step": 135997, "epoch": 1528, "lr": 4.998666605828085e-05}
{"train_loss": 0.12129737436771393, "global_step": 135998, "epoch": 1528, "lr": 4.998608632169901e-05}
{"train_loss": 0.13928715884685516, "global_step": 135999, "epoch": 1528, "lr": 4.998550658511905e-05}
{"train_loss": 0.07765529304742813, "global_step": 136000, "epoch": 1528, "lr": 4.998492684854103e-05}
{"train_loss": 0.12033604830503464, "global_step": 136001, "epoch": 1528, "lr": 4.9984347111965054e-05}
{"train_loss": 0.09710614383220673, "global_step": 136002, "epoch": 1528, "lr": 4.998376737539117e-05}
{"train_loss": 0.09337744861841202, "global_step": 136003, "epoch": 1528, "lr": 4.9983187638819466e-05}
{"train_loss": 0.06952236592769623, "global_step": 136004, "epoch": 1528, "lr": 4.9982607902250026e-05}
{"train_loss": 0.17217548191547394, "global_step": 136005, "epoch": 1528, "lr": 4.9982028165682924e-05}
{"train_loss": 0.16008034348487854, "global_step": 136006, "epoch": 1528, "lr": 4.998144842911824e-05}
{"train_loss": 0.10917419195175171, "global_step": 136007, "epoch": 1528, "lr": 4.998086869255606e-05}
{"train_loss": 0.21433836221694946, "global_step": 136008, "epoch": 1528, "lr": 4.998028895599643e-05}
{"train_loss": 0.13355927169322968, "global_step": 136009, "epoch": 1528, "lr": 4.997970921943947e-05}
{"train_loss": 0.1460915207862854, "global_step": 136010, "epoch": 1528, "lr": 4.997912948288522e-05}
{"train_loss": 0.08323963731527328, "global_step": 136011, "epoch": 1528, "lr": 4.997854974633377e-05}
{"train_loss": 0.18013764917850494, "global_step": 136012, "epoch": 1528, "lr": 4.997797000978522e-05}
{"train_loss": 0.1261451244354248, "global_step": 136013, "epoch": 1528, "lr": 4.997739027323963e-05}
{"train_loss": 0.10858669131994247, "global_step": 136014, "epoch": 1528, "lr": 4.9976810536697083e-05}
{"train_loss": 0.08506039530038834, "global_step": 136015, "epoch": 1528, "lr": 4.9976230800157635e-05}
{"train_loss": 0.08342928439378738, "global_step": 136016, "epoch": 1528, "lr": 4.9975651063621406e-05}
{"train_loss": 0.12516504526138306, "global_step": 136017, "epoch": 1528, "lr": 4.9975071327088436e-05}
{"train_loss": 0.0851375162601471, "global_step": 136018, "epoch": 1528, "lr": 4.997449159055883e-05}
{"train_loss": 0.14930854737758636, "global_step": 136019, "epoch": 1528, "lr": 4.9973911854032634e-05}
{"train_loss": 0.09728184342384338, "global_step": 136020, "epoch": 1528, "lr": 4.997333211750997e-05}
{"train_loss": 0.10300910472869873, "global_step": 136021, "epoch": 1528, "lr": 4.997275238099088e-05}
{"train_loss": 0.0755477175116539, "global_step": 136022, "epoch": 1528, "lr": 4.997217264447545e-05}
{"train_loss": 0.1464242786169052, "global_step": 136023, "epoch": 1528, "lr": 4.997159290796376e-05}
{"train_loss": 0.15693311393260956, "global_step": 136024, "epoch": 1528, "lr": 4.9971013171455896e-05}
{"train_loss": 0.1103043183684349, "global_step": 136025, "epoch": 1528, "lr": 4.997043343495192e-05}
{"train_loss": 0.1452709138393402, "global_step": 136026, "epoch": 1528, "lr": 4.996985369845193e-05}
{"train_loss": 0.15464241802692413, "global_step": 136027, "epoch": 1528, "lr": 4.996927396195598e-05}
{"train_loss": 0.15833371877670288, "global_step": 136028, "epoch": 1528, "lr": 4.996869422546418e-05}
{"train_loss": 0.041576869785785675, "global_step": 136029, "epoch": 1528, "lr": 4.996811448897656e-05}
{"train_loss": 0.1585114300251007, "global_step": 136030, "epoch": 1528, "lr": 4.9967534752493234e-05}
{"train_loss": 0.08537408709526062, "global_step": 136031, "epoch": 1528, "lr": 4.996695501601429e-05}
{"train_loss": 0.08305060863494873, "global_step": 136032, "epoch": 1528, "lr": 4.9966375279539766e-05}
{"train_loss": 0.08848299086093903, "global_step": 136033, "epoch": 1528, "lr": 4.996579554306979e-05}
{"train_loss": 0.1221579983830452, "global_step": 136034, "epoch": 1528, "lr": 4.9965215806604385e-05}
{"train_loss": 0.11688835173845291, "global_step": 136035, "epoch": 1528, "lr": 4.996463607014368e-05}
{"train_loss": 0.18553029000759125, "global_step": 136036, "epoch": 1528, "lr": 4.9964056333687706e-05}
{"train_loss": 0.14837764203548431, "global_step": 136037, "epoch": 1528, "lr": 4.9963476597236586e-05}
{"train_loss": 0.11809957027435303, "global_step": 136038, "epoch": 1528, "lr": 4.996289686079035e-05}
{"train_loss": 0.10749255865812302, "global_step": 136039, "epoch": 1528, "lr": 4.9962317124349133e-05}
{"train_loss": 0.12352165579795837, "global_step": 136040, "epoch": 1528, "lr": 4.996173738791297e-05}
{"train_loss": 0.1466667205095291, "global_step": 136041, "epoch": 1528, "lr": 4.996115765148195e-05}
{"train_loss": 0.13689079880714417, "global_step": 136042, "epoch": 1528, "lr": 4.996057791505615e-05}
{"train_loss": 0.09073516726493835, "global_step": 136043, "epoch": 1528, "lr": 4.995999817863565e-05}
{"train_loss": 0.1282244324684143, "global_step": 136044, "epoch": 1528, "lr": 4.9959418442220524e-05}
{"train_loss": 0.09552266448736191, "global_step": 136045, "epoch": 1528, "lr": 4.9958838705810875e-05}
{"train_loss": 0.11084671318531036, "global_step": 136046, "epoch": 1528, "lr": 4.995825896940673e-05}
{"train_loss": 0.0884580984711647, "global_step": 136047, "epoch": 1528, "lr": 4.995767923300819e-05}
{"train_loss": 0.10171584784984589, "global_step": 136048, "epoch": 1528, "lr": 4.995709949661537e-05}
{"train_loss": 0.11806859076023102, "global_step": 136049, "epoch": 1528, "lr": 4.99565197602283e-05}
{"train_loss": 0.06516768038272858, "global_step": 136050, "epoch": 1528, "lr": 4.995594002384709e-05}
{"train_loss": 0.13951513171195984, "global_step": 136051, "epoch": 1528, "lr": 4.9955360287471785e-05}
{"train_loss": 0.12212860584259033, "global_step": 136052, "epoch": 1528, "lr": 4.9954780551102495e-05}
{"train_loss": 0.11184042692184448, "global_step": 136053, "epoch": 1528, "lr": 4.995420081473928e-05}
{"train_loss": 0.09457755088806152, "global_step": 136054, "epoch": 1528, "lr": 4.9953621078382226e-05}
{"train_loss": 0.08147991448640823, "global_step": 136055, "epoch": 1528, "lr": 4.9953041342031396e-05}
{"train_loss": 0.06817957758903503, "global_step": 136056, "epoch": 1528, "lr": 4.9952461605686894e-05}
{"train_loss": 0.11173093318939209, "global_step": 136057, "epoch": 1528, "lr": 4.995188186934877e-05}
{"train_loss": 0.16995373368263245, "global_step": 136058, "epoch": 1528, "lr": 4.995130213301712e-05}
{"train_loss": 0.1552116870880127, "global_step": 136059, "epoch": 1528, "lr": 4.995072239669202e-05}
{"train_loss": 0.15026496350765228, "global_step": 136060, "epoch": 1528, "lr": 4.9950142660373546e-05}
{"train_loss": 0.095958411693573, "global_step": 136061, "epoch": 1528, "lr": 4.9949562924061774e-05}
{"train_loss": 0.12443438172340393, "global_step": 136062, "epoch": 1528, "lr": 4.994898318775678e-05}
{"train_loss": 0.12518122792243958, "global_step": 136063, "epoch": 1528, "lr": 4.994840345145864e-05}
{"train_loss": 0.08681298047304153, "global_step": 136064, "epoch": 1528, "lr": 4.994782371516745e-05}
{"train_loss": 0.11317440122365952, "global_step": 136065, "epoch": 1528, "lr": 4.9947243978883254e-05}
{"train_loss": 0.1370813250541687, "global_step": 136066, "epoch": 1528, "lr": 4.994666424260615e-05}
{"train_loss": 0.12787073850631714, "global_step": 136067, "epoch": 1528, "lr": 4.994608450633624e-05}
{"train_loss": 0.11799563467502594, "global_step": 136068, "epoch": 1528, "lr": 4.9945504770073555e-05}
{"train_loss": 0.15915684401988983, "global_step": 136069, "epoch": 1528, "lr": 4.9944925033818216e-05}
{"train_loss": 0.11913684755563736, "global_step": 136070, "epoch": 1528, "lr": 4.994434529757026e-05}
{"train_loss": 0.15326716005802155, "global_step": 136071, "epoch": 1528, "lr": 4.994376556132981e-05}
{"train_loss": 0.09526603668928146, "global_step": 136072, "epoch": 1528, "lr": 4.994318582509689e-05}
{"train_loss": 0.11465609073638916, "global_step": 136073, "epoch": 1528, "lr": 4.994260608887164e-05}
{"train_loss": 0.09747998416423798, "global_step": 136074, "epoch": 1528, "lr": 4.994202635265408e-05}
{"train_loss": 0.1563195288181305, "global_step": 136075, "epoch": 1528, "lr": 4.9941446616444336e-05}
{"train_loss": 0.07349973917007446, "global_step": 136076, "epoch": 1528, "lr": 4.994086688024245e-05}
{"train_loss": 0.15575124323368073, "global_step": 136077, "epoch": 1528, "lr": 4.994028714404853e-05}
{"train_loss": 0.08197982609272003, "global_step": 136078, "epoch": 1528, "lr": 4.993970740786262e-05}
{"train_loss": 0.07731055468320847, "global_step": 136079, "epoch": 1528, "lr": 4.993912767168482e-05}
{"train_loss": 0.11912253239516461, "global_step": 136080, "epoch": 1528, "lr": 4.9938547935515207e-05, "val_loss": 5.8807501792907715}
{"train_loss": 0.10217849165201187, "global_step": 136081, "epoch": 1529, "lr": 4.9937968199353854e-05}
{"train_loss": 0.07968518882989883, "global_step": 136082, "epoch": 1529, "lr": 4.9937388463200836e-05}
{"train_loss": 0.1367412507534027, "global_step": 136083, "epoch": 1529, "lr": 4.993680872705623e-05}
{"train_loss": 0.06623794883489609, "global_step": 136084, "epoch": 1529, "lr": 4.9936228990920136e-05}
{"train_loss": 0.176502525806427, "global_step": 136085, "epoch": 1529, "lr": 4.99356492547926e-05}
{"train_loss": 0.16260313987731934, "global_step": 136086, "epoch": 1529, "lr": 4.9935069518673725e-05}
{"train_loss": 0.09610039740800858, "global_step": 136087, "epoch": 1529, "lr": 4.993448978256357e-05}
{"train_loss": 0.13625772297382355, "global_step": 136088, "epoch": 1529, "lr": 4.993391004646224e-05}
{"train_loss": 0.06349164247512817, "global_step": 136089, "epoch": 1529, "lr": 4.993333031036977e-05}
{"train_loss": 0.06712023913860321, "global_step": 136090, "epoch": 1529, "lr": 4.9932750574286286e-05}
{"train_loss": 0.12368687987327576, "global_step": 136091, "epoch": 1529, "lr": 4.9932170838211825e-05}
{"train_loss": 0.060284193605184555, "global_step": 136092, "epoch": 1529, "lr": 4.9931591102146505e-05}
{"train_loss": 0.13797006011009216, "global_step": 136093, "epoch": 1529, "lr": 4.9931011366090354e-05}
{"train_loss": 0.08921507745981216, "global_step": 136094, "epoch": 1529, "lr": 4.99304316300435e-05}
{"train_loss": 0.10893774032592773, "global_step": 136095, "epoch": 1529, "lr": 4.992985189400598e-05}
{"train_loss": 0.19281500577926636, "global_step": 136096, "epoch": 1529, "lr": 4.992927215797791e-05}
{"train_loss": 0.10466650873422623, "global_step": 136097, "epoch": 1529, "lr": 4.992869242195933e-05}
{"train_loss": 0.09440948069095612, "global_step": 136098, "epoch": 1529, "lr": 4.9928112685950354e-05}
{"train_loss": 0.16142600774765015, "global_step": 136099, "epoch": 1529, "lr": 4.9927532949951035e-05}
{"train_loss": 0.07143588364124298, "global_step": 136100, "epoch": 1529, "lr": 4.992695321396146e-05}
{"train_loss": 0.18380065262317657, "global_step": 136101, "epoch": 1529, "lr": 4.99263734779817e-05}
{"train_loss": 0.05818013474345207, "global_step": 136102, "epoch": 1529, "lr": 4.992579374201183e-05}
{"train_loss": 0.0981559157371521, "global_step": 136103, "epoch": 1529, "lr": 4.9925214006051955e-05}
{"train_loss": 0.09030543267726898, "global_step": 136104, "epoch": 1529, "lr": 4.9924634270102116e-05}
{"train_loss": 0.1283317655324936, "global_step": 136105, "epoch": 1529, "lr": 4.992405453416243e-05}
{"train_loss": 0.09874691069126129, "global_step": 136106, "epoch": 1529, "lr": 4.992347479823293e-05}
{"train_loss": 0.13997918367385864, "global_step": 136107, "epoch": 1529, "lr": 4.992289506231374e-05}
{"train_loss": 0.12547965347766876, "global_step": 136108, "epoch": 1529, "lr": 4.99223153264049e-05}
{"train_loss": 0.07723750174045563, "global_step": 136109, "epoch": 1529, "lr": 4.992173559050652e-05}
{"train_loss": 0.15198951959609985, "global_step": 136110, "epoch": 1529, "lr": 4.992115585461864e-05}
{"train_loss": 0.08666414767503738, "global_step": 136111, "epoch": 1529, "lr": 4.992057611874139e-05}
{"train_loss": 0.14797072112560272, "global_step": 136112, "epoch": 1529, "lr": 4.991999638287478e-05}
{"train_loss": 0.08118610829114914, "global_step": 136113, "epoch": 1529, "lr": 4.991941664701896e-05}
{"train_loss": 0.09834980219602585, "global_step": 136114, "epoch": 1529, "lr": 4.991883691117395e-05}
{"train_loss": 0.08755098283290863, "global_step": 136115, "epoch": 1529, "lr": 4.991825717533987e-05}
{"train_loss": 0.114710733294487, "global_step": 136116, "epoch": 1529, "lr": 4.991767743951675e-05}
{"train_loss": 0.11180602014064789, "global_step": 136117, "epoch": 1529, "lr": 4.991709770370473e-05}
{"train_loss": 0.10747448354959488, "global_step": 136118, "epoch": 1529, "lr": 4.991651796790384e-05}
{"train_loss": 0.07967893779277802, "global_step": 136119, "epoch": 1529, "lr": 4.991593823211417e-05}
{"train_loss": 0.11950752139091492, "global_step": 136120, "epoch": 1529, "lr": 4.9915358496335806e-05}
{"train_loss": 0.06934723258018494, "global_step": 136121, "epoch": 1529, "lr": 4.991477876056881e-05}
{"train_loss": 0.11089940369129181, "global_step": 136122, "epoch": 1529, "lr": 4.9914199024813296e-05}
{"train_loss": 0.16120174527168274, "global_step": 136123, "epoch": 1529, "lr": 4.991361928906929e-05}
{"train_loss": 0.10035667568445206, "global_step": 136124, "epoch": 1529, "lr": 4.991303955333692e-05}
{"train_loss": 0.1355590969324112, "global_step": 136125, "epoch": 1529, "lr": 4.991245981761622e-05}
{"train_loss": 0.17174626886844635, "global_step": 136126, "epoch": 1529, "lr": 4.991188008190731e-05}
{"train_loss": 0.10330459475517273, "global_step": 136127, "epoch": 1529, "lr": 4.991130034621022e-05}
{"train_loss": 0.14653509855270386, "global_step": 136128, "epoch": 1529, "lr": 4.991072061052508e-05}
{"train_loss": 0.07759109884500504, "global_step": 136129, "epoch": 1529, "lr": 4.9910140874851926e-05}
{"train_loss": 0.17548665404319763, "global_step": 136130, "epoch": 1529, "lr": 4.990956113919087e-05}
{"train_loss": 0.1312628537416458, "global_step": 136131, "epoch": 1529, "lr": 4.990898140354195e-05}
{"train_loss": 0.07232620567083359, "global_step": 136132, "epoch": 1529, "lr": 4.990840166790529e-05}
{"train_loss": 0.16803070902824402, "global_step": 136133, "epoch": 1529, "lr": 4.990782193228092e-05}
{"train_loss": 0.1176886036992073, "global_step": 136134, "epoch": 1529, "lr": 4.9907242196668965e-05}
{"train_loss": 0.14112918078899384, "global_step": 136135, "epoch": 1529, "lr": 4.990666246106946e-05}
{"train_loss": 0.07786408066749573, "global_step": 136136, "epoch": 1529, "lr": 4.9906082725482524e-05}
{"train_loss": 0.07469657808542252, "global_step": 136137, "epoch": 1529, "lr": 4.99055029899082e-05}
{"train_loss": 0.07590072602033615, "global_step": 136138, "epoch": 1529, "lr": 4.990492325434657e-05}
{"train_loss": 0.07871945947408676, "global_step": 136139, "epoch": 1529, "lr": 4.9904343518797746e-05}
{"train_loss": 0.10672542452812195, "global_step": 136140, "epoch": 1529, "lr": 4.990376378326176e-05}
{"train_loss": 0.12235457450151443, "global_step": 136141, "epoch": 1529, "lr": 4.990318404773873e-05}
{"train_loss": 0.12883543968200684, "global_step": 136142, "epoch": 1529, "lr": 4.9902604312228694e-05}
{"train_loss": 0.10427936911582947, "global_step": 136143, "epoch": 1529, "lr": 4.990202457673178e-05}
{"train_loss": 0.11908842623233795, "global_step": 136144, "epoch": 1529, "lr": 4.9901444841248014e-05}
{"train_loss": 0.13408227264881134, "global_step": 136145, "epoch": 1529, "lr": 4.9900865105777515e-05}
{"train_loss": 0.1797824651002884, "global_step": 136146, "epoch": 1529, "lr": 4.9900285370320324e-05}
{"train_loss": 0.15115711092948914, "global_step": 136147, "epoch": 1529, "lr": 4.989970563487656e-05}
{"train_loss": 0.14270377159118652, "global_step": 136148, "epoch": 1529, "lr": 4.989912589944626e-05}
{"train_loss": 0.11918109655380249, "global_step": 136149, "epoch": 1529, "lr": 4.989854616402954e-05}
{"train_loss": 0.11650346219539642, "global_step": 136150, "epoch": 1529, "lr": 4.9897966428626436e-05}
{"train_loss": 0.1801844835281372, "global_step": 136151, "epoch": 1529, "lr": 4.9897386693237076e-05}
{"train_loss": 0.09031985700130463, "global_step": 136152, "epoch": 1529, "lr": 4.9896806957861485e-05}
{"train_loss": 0.18725644052028656, "global_step": 136153, "epoch": 1529, "lr": 4.989622722249979e-05}
{"train_loss": 0.0947486162185669, "global_step": 136154, "epoch": 1529, "lr": 4.9895647487152025e-05}
{"train_loss": 0.07637593150138855, "global_step": 136155, "epoch": 1529, "lr": 4.989506775181831e-05}
{"train_loss": 0.1209130585193634, "global_step": 136156, "epoch": 1529, "lr": 4.989448801649869e-05}
{"train_loss": 0.11577669531106949, "global_step": 136157, "epoch": 1529, "lr": 4.989390828119325e-05}
{"train_loss": 0.14139601588249207, "global_step": 136158, "epoch": 1529, "lr": 4.9893328545902084e-05}
{"train_loss": 0.14661942422389984, "global_step": 136159, "epoch": 1529, "lr": 4.9892748810625244e-05}
{"train_loss": 0.12593770027160645, "global_step": 136160, "epoch": 1529, "lr": 4.9892169075362844e-05}
{"train_loss": 0.08962921798229218, "global_step": 136161, "epoch": 1529, "lr": 4.989158934011492e-05}
{"train_loss": 0.12367919832468033, "global_step": 136162, "epoch": 1529, "lr": 4.989100960488159e-05}
{"train_loss": 0.09088359773159027, "global_step": 136163, "epoch": 1529, "lr": 4.989042986966289e-05}
{"train_loss": 0.1407770961523056, "global_step": 136164, "epoch": 1529, "lr": 4.988985013445894e-05}
{"train_loss": 0.1594191938638687, "global_step": 136165, "epoch": 1529, "lr": 4.988927039926978e-05}
{"train_loss": 0.1343088448047638, "global_step": 136166, "epoch": 1529, "lr": 4.988869066409553e-05}
{"train_loss": 0.15348035097122192, "global_step": 136167, "epoch": 1529, "lr": 4.988811092893622e-05}
{"train_loss": 0.1798681914806366, "global_step": 136168, "epoch": 1529, "lr": 4.988753119379197e-05}
{"train_loss": 0.11797719867376799, "global_step": 136169, "epoch": 1529, "lr": 4.988695145866282e-05, "val_loss": 5.7593302726745605}
{"train_loss": 0.155848428606987, "global_step": 136170, "epoch": 1530, "lr": 4.98863717235489e-05}
{"train_loss": 0.1544037014245987, "global_step": 136171, "epoch": 1530, "lr": 4.988579198845022e-05}
{"train_loss": 0.06655224412679672, "global_step": 136172, "epoch": 1530, "lr": 4.988521225336692e-05}
{"train_loss": 0.1368987262248993, "global_step": 136173, "epoch": 1530, "lr": 4.9884632518299034e-05}
{"train_loss": 0.08492571115493774, "global_step": 136174, "epoch": 1530, "lr": 4.9884052783246665e-05}
{"train_loss": 0.17117339372634888, "global_step": 136175, "epoch": 1530, "lr": 4.9883473048209894e-05}
{"train_loss": 0.0719049796462059, "global_step": 136176, "epoch": 1530, "lr": 4.988289331318877e-05}
{"train_loss": 0.11906605213880539, "global_step": 136177, "epoch": 1530, "lr": 4.9882313578183406e-05}
{"train_loss": 0.1488512009382248, "global_step": 136178, "epoch": 1530, "lr": 4.988173384319385e-05}
{"train_loss": 0.11165869235992432, "global_step": 136179, "epoch": 1530, "lr": 4.9881154108220216e-05}
{"train_loss": 0.07821090519428253, "global_step": 136180, "epoch": 1530, "lr": 4.988057437326253e-05}
{"train_loss": 0.06337537616491318, "global_step": 136181, "epoch": 1530, "lr": 4.987999463832092e-05}
{"train_loss": 0.10393045097589493, "global_step": 136182, "epoch": 1530, "lr": 4.987941490339543e-05}
{"train_loss": 0.11650566756725311, "global_step": 136183, "epoch": 1530, "lr": 4.9878835168486166e-05}
{"train_loss": 0.17317402362823486, "global_step": 136184, "epoch": 1530, "lr": 4.987825543359317e-05}
{"train_loss": 0.1020503044128418, "global_step": 136185, "epoch": 1530, "lr": 4.987767569871657e-05}
{"train_loss": 0.14391347765922546, "global_step": 136186, "epoch": 1530, "lr": 4.987709596385638e-05}
{"train_loss": 0.09424940496683121, "global_step": 136187, "epoch": 1530, "lr": 4.9876516229012745e-05}
{"train_loss": 0.15687090158462524, "global_step": 136188, "epoch": 1530, "lr": 4.9875936494185686e-05}
{"train_loss": 0.07734094560146332, "global_step": 136189, "epoch": 1530, "lr": 4.987535675937533e-05}
{"train_loss": 0.07741303741931915, "global_step": 136190, "epoch": 1530, "lr": 4.9874777024581705e-05}
{"train_loss": 0.13190588355064392, "global_step": 136191, "epoch": 1530, "lr": 4.987419728980493e-05}
{"train_loss": 0.1592857986688614, "global_step": 136192, "epoch": 1530, "lr": 4.987361755504506e-05}
{"train_loss": 0.08621320873498917, "global_step": 136193, "epoch": 1530, "lr": 4.987303782030219e-05}
{"train_loss": 0.06939759850502014, "global_step": 136194, "epoch": 1530, "lr": 4.9872458085576395e-05}
{"train_loss": 0.0830858051776886, "global_step": 136195, "epoch": 1530, "lr": 4.9871878350867724e-05}
{"train_loss": 0.12105464935302734, "global_step": 136196, "epoch": 1530, "lr": 4.98712986161763e-05}
{"train_loss": 0.17300643026828766, "global_step": 136197, "epoch": 1530, "lr": 4.987071888150215e-05}
{"train_loss": 0.10696737468242645, "global_step": 136198, "epoch": 1530, "lr": 4.9870139146845415e-05}
{"train_loss": 0.10940000414848328, "global_step": 136199, "epoch": 1530, "lr": 4.9869559412206106e-05}
{"train_loss": 0.13837221264839172, "global_step": 136200, "epoch": 1530, "lr": 4.986897967758437e-05}
{"train_loss": 0.11286372691392899, "global_step": 136201, "epoch": 1530, "lr": 4.986839994298021e-05}
{"train_loss": 0.17276570200920105, "global_step": 136202, "epoch": 1530, "lr": 4.9867820208393766e-05}
{"train_loss": 0.10533109307289124, "global_step": 136203, "epoch": 1530, "lr": 4.986724047382507e-05}
{"train_loss": 0.1187000721693039, "global_step": 136204, "epoch": 1530, "lr": 4.9866660739274255e-05}
{"train_loss": 0.09728481620550156, "global_step": 136205, "epoch": 1530, "lr": 4.986608100474133e-05}
{"train_loss": 0.10950936377048492, "global_step": 136206, "epoch": 1530, "lr": 4.9865501270226436e-05}
{"train_loss": 0.07252103835344315, "global_step": 136207, "epoch": 1530, "lr": 4.9864921535729604e-05}
{"train_loss": 0.11439534276723862, "global_step": 136208, "epoch": 1530, "lr": 4.9864341801250954e-05}
{"train_loss": 0.2075779139995575, "global_step": 136209, "epoch": 1530, "lr": 4.986376206679052e-05}
{"train_loss": 0.13961249589920044, "global_step": 136210, "epoch": 1530, "lr": 4.9863182332348405e-05}
{"train_loss": 0.2132137268781662, "global_step": 136211, "epoch": 1530, "lr": 4.986260259792469e-05}
{"train_loss": 0.10770624876022339, "global_step": 136212, "epoch": 1530, "lr": 4.9862022863519446e-05}
{"train_loss": 0.10263926535844803, "global_step": 136213, "epoch": 1530, "lr": 4.986144312913275e-05}
{"train_loss": 0.10485439747571945, "global_step": 136214, "epoch": 1530, "lr": 4.986086339476469e-05}
{"train_loss": 0.11961489915847778, "global_step": 136215, "epoch": 1530, "lr": 4.986028366041533e-05}
{"train_loss": 0.15852998197078705, "global_step": 136216, "epoch": 1530, "lr": 4.985970392608474e-05}
{"train_loss": 0.07814832776784897, "global_step": 136217, "epoch": 1530, "lr": 4.9859124191773034e-05}
{"train_loss": 0.10949172079563141, "global_step": 136218, "epoch": 1530, "lr": 4.9858544457480236e-05}
{"train_loss": 0.08290892839431763, "global_step": 136219, "epoch": 1530, "lr": 4.985796472320649e-05}
{"train_loss": 0.23413141071796417, "global_step": 136220, "epoch": 1530, "lr": 4.985738498895181e-05}
{"train_loss": 0.095590740442276, "global_step": 136221, "epoch": 1530, "lr": 4.9856805254716326e-05}
{"train_loss": 0.14501452445983887, "global_step": 136222, "epoch": 1530, "lr": 4.985622552050007e-05}
{"train_loss": 0.15568238496780396, "global_step": 136223, "epoch": 1530, "lr": 4.9855645786303166e-05}
{"train_loss": 0.1751897633075714, "global_step": 136224, "epoch": 1530, "lr": 4.9855066052125645e-05}
{"train_loss": 0.07576944679021835, "global_step": 136225, "epoch": 1530, "lr": 4.985448631796763e-05}
{"train_loss": 0.07715272903442383, "global_step": 136226, "epoch": 1530, "lr": 4.985390658382916e-05}
{"train_loss": 0.08551284670829773, "global_step": 136227, "epoch": 1530, "lr": 4.9853326849710354e-05}
{"train_loss": 0.1354699283838272, "global_step": 136228, "epoch": 1530, "lr": 4.9852747115611235e-05}
{"train_loss": 0.10843583196401596, "global_step": 136229, "epoch": 1530, "lr": 4.985216738153193e-05}
{"train_loss": 0.12745094299316406, "global_step": 136230, "epoch": 1530, "lr": 4.98515876474725e-05}
{"train_loss": 0.15276215970516205, "global_step": 136231, "epoch": 1530, "lr": 4.985100791343302e-05}
{"train_loss": 0.09491025656461716, "global_step": 136232, "epoch": 1530, "lr": 4.985042817941357e-05}
{"train_loss": 0.17551949620246887, "global_step": 136233, "epoch": 1530, "lr": 4.984984844541423e-05}
{"train_loss": 0.15440593659877777, "global_step": 136234, "epoch": 1530, "lr": 4.9849268711435085e-05}
{"train_loss": 0.16686412692070007, "global_step": 136235, "epoch": 1530, "lr": 4.9848688977476173e-05}
{"train_loss": 0.13404163718223572, "global_step": 136236, "epoch": 1530, "lr": 4.984810924353764e-05}
{"train_loss": 0.1660807728767395, "global_step": 136237, "epoch": 1530, "lr": 4.98475295096195e-05}
{"train_loss": 0.10500114411115646, "global_step": 136238, "epoch": 1530, "lr": 4.984694977572188e-05}
{"train_loss": 0.10809005796909332, "global_step": 136239, "epoch": 1530, "lr": 4.9846370041844805e-05}
{"train_loss": 0.10886424034833908, "global_step": 136240, "epoch": 1530, "lr": 4.984579030798842e-05}
{"train_loss": 0.18228791654109955, "global_step": 136241, "epoch": 1530, "lr": 4.9845210574152734e-05}
{"train_loss": 0.11816607415676117, "global_step": 136242, "epoch": 1530, "lr": 4.9844630840337886e-05}
{"train_loss": 0.15093843638896942, "global_step": 136243, "epoch": 1530, "lr": 4.9844051106543896e-05}
{"train_loss": 0.14949309825897217, "global_step": 136244, "epoch": 1530, "lr": 4.9843471372770905e-05}
{"train_loss": 0.10671819746494293, "global_step": 136245, "epoch": 1530, "lr": 4.984289163901893e-05}
{"train_loss": 0.11342402547597885, "global_step": 136246, "epoch": 1530, "lr": 4.984231190528809e-05}
{"train_loss": 0.11073290556669235, "global_step": 136247, "epoch": 1530, "lr": 4.984173217157844e-05}
{"train_loss": 0.08778493851423264, "global_step": 136248, "epoch": 1530, "lr": 4.984115243789007e-05}
{"train_loss": 0.10083436220884323, "global_step": 136249, "epoch": 1530, "lr": 4.984057270422306e-05}
{"train_loss": 0.12182827293872833, "global_step": 136250, "epoch": 1530, "lr": 4.983999297057748e-05}
{"train_loss": 0.10375026613473892, "global_step": 136251, "epoch": 1530, "lr": 4.9839413236953406e-05}
{"train_loss": 0.11006667464971542, "global_step": 136252, "epoch": 1530, "lr": 4.9838833503350926e-05}
{"train_loss": 0.09936536848545074, "global_step": 136253, "epoch": 1530, "lr": 4.983825376977013e-05}
{"train_loss": 0.15531982481479645, "global_step": 136254, "epoch": 1530, "lr": 4.983767403621105e-05}
{"train_loss": 0.13187776505947113, "global_step": 136255, "epoch": 1530, "lr": 4.983709430267382e-05}
{"train_loss": 0.05418995022773743, "global_step": 136256, "epoch": 1530, "lr": 4.983651456915846e-05}
{"train_loss": 0.13780450820922852, "global_step": 136257, "epoch": 1530, "lr": 4.983593483566511e-05}
{"train_loss": 0.1219413723336177, "global_step": 136258, "epoch": 1530, "lr": 4.9835355102193784e-05, "val_loss": 5.861598491668701, "train_action_mse_error": 15.615167617797852}
{"train_loss": 0.12425815314054489, "global_step": 136259, "epoch": 1531, "lr": 4.9834775368744624e-05}
{"train_loss": 0.14222964644432068, "global_step": 136260, "epoch": 1531, "lr": 4.983419563531765e-05}
{"train_loss": 0.14818446338176727, "global_step": 136261, "epoch": 1531, "lr": 4.983361590191299e-05}
{"train_loss": 0.1016681119799614, "global_step": 136262, "epoch": 1531, "lr": 4.983303616853068e-05}
{"train_loss": 0.10607261955738068, "global_step": 136263, "epoch": 1531, "lr": 4.983245643517083e-05}
{"train_loss": 0.15838684141635895, "global_step": 136264, "epoch": 1531, "lr": 4.983187670183348e-05}
{"train_loss": 0.1127447783946991, "global_step": 136265, "epoch": 1531, "lr": 4.9831296968518743e-05}
{"train_loss": 0.11782626807689667, "global_step": 136266, "epoch": 1531, "lr": 4.9830717235226705e-05}
{"train_loss": 0.12738679349422455, "global_step": 136267, "epoch": 1531, "lr": 4.983013750195741e-05}
{"train_loss": 0.13575591146945953, "global_step": 136268, "epoch": 1531, "lr": 4.982955776871095e-05}
{"train_loss": 0.09838870167732239, "global_step": 136269, "epoch": 1531, "lr": 4.9828978035487414e-05}
{"train_loss": 0.09908962994813919, "global_step": 136270, "epoch": 1531, "lr": 4.982839830228686e-05}
{"train_loss": 0.1456003040075302, "global_step": 136271, "epoch": 1531, "lr": 4.982781856910938e-05}
{"train_loss": 0.14302006363868713, "global_step": 136272, "epoch": 1531, "lr": 4.982723883595505e-05}
{"train_loss": 0.15913401544094086, "global_step": 136273, "epoch": 1531, "lr": 4.9826659102823935e-05}
{"train_loss": 0.07283978909254074, "global_step": 136274, "epoch": 1531, "lr": 4.982607936971614e-05}
{"train_loss": 0.09898228198289871, "global_step": 136275, "epoch": 1531, "lr": 4.982549963663171e-05}
{"train_loss": 0.08857128769159317, "global_step": 136276, "epoch": 1531, "lr": 4.982491990357075e-05}
{"train_loss": 0.10690057277679443, "global_step": 136277, "epoch": 1531, "lr": 4.982434017053332e-05}
{"train_loss": 0.14260905981063843, "global_step": 136278, "epoch": 1531, "lr": 4.982376043751952e-05}
{"train_loss": 0.0836537703871727, "global_step": 136279, "epoch": 1531, "lr": 4.982318070452939e-05}
{"train_loss": 0.22899895906448364, "global_step": 136280, "epoch": 1531, "lr": 4.982260097156305e-05}
{"train_loss": 0.2035207599401474, "global_step": 136281, "epoch": 1531, "lr": 4.982202123862055e-05}
{"train_loss": 0.12331312894821167, "global_step": 136282, "epoch": 1531, "lr": 4.982144150570199e-05}
{"train_loss": 0.10258134454488754, "global_step": 136283, "epoch": 1531, "lr": 4.982086177280741e-05}
{"train_loss": 0.11360710859298706, "global_step": 136284, "epoch": 1531, "lr": 4.982028203993692e-05}
{"train_loss": 0.13325421512126923, "global_step": 136285, "epoch": 1531, "lr": 4.981970230709061e-05}
{"train_loss": 0.07785546779632568, "global_step": 136286, "epoch": 1531, "lr": 4.981912257426852e-05}
{"train_loss": 0.117017462849617, "global_step": 136287, "epoch": 1531, "lr": 4.981854284147076e-05}
{"train_loss": 0.11327798664569855, "global_step": 136288, "epoch": 1531, "lr": 4.981796310869739e-05}
{"train_loss": 0.131258025765419, "global_step": 136289, "epoch": 1531, "lr": 4.981738337594849e-05}
{"train_loss": 0.11247434467077255, "global_step": 136290, "epoch": 1531, "lr": 4.9816803643224144e-05}
{"train_loss": 0.10133335739374161, "global_step": 136291, "epoch": 1531, "lr": 4.981622391052443e-05}
{"train_loss": 0.06982526928186417, "global_step": 136292, "epoch": 1531, "lr": 4.9815644177849406e-05}
{"train_loss": 0.12033034861087799, "global_step": 136293, "epoch": 1531, "lr": 4.9815064445199194e-05}
{"train_loss": 0.09380771219730377, "global_step": 136294, "epoch": 1531, "lr": 4.981448471257382e-05}
{"train_loss": 0.1134488433599472, "global_step": 136295, "epoch": 1531, "lr": 4.9813904979973405e-05}
{"train_loss": 0.08789169788360596, "global_step": 136296, "epoch": 1531, "lr": 4.9813325247397976e-05}
{"train_loss": 0.09444969147443771, "global_step": 136297, "epoch": 1531, "lr": 4.981274551484768e-05}
{"train_loss": 0.08134140074253082, "global_step": 136298, "epoch": 1531, "lr": 4.981216578232253e-05}
{"train_loss": 0.1016344279050827, "global_step": 136299, "epoch": 1531, "lr": 4.981158604982266e-05}
{"train_loss": 0.132059708237648, "global_step": 136300, "epoch": 1531, "lr": 4.9811006317348095e-05}
{"train_loss": 0.10137606412172318, "global_step": 136301, "epoch": 1531, "lr": 4.9810426584898935e-05}
{"train_loss": 0.09544113278388977, "global_step": 136302, "epoch": 1531, "lr": 4.980984685247528e-05}
{"train_loss": 0.10668916255235672, "global_step": 136303, "epoch": 1531, "lr": 4.9809267120077173e-05}
{"train_loss": 0.03970826789736748, "global_step": 136304, "epoch": 1531, "lr": 4.980868738770473e-05}
{"train_loss": 0.08595718443393707, "global_step": 136305, "epoch": 1531, "lr": 4.980810765535799e-05}
{"train_loss": 0.13977810740470886, "global_step": 136306, "epoch": 1531, "lr": 4.980752792303706e-05}
{"train_loss": 0.09758663177490234, "global_step": 136307, "epoch": 1531, "lr": 4.980694819074199e-05}
{"train_loss": 0.08645186573266983, "global_step": 136308, "epoch": 1531, "lr": 4.980636845847288e-05}
{"train_loss": 0.15301884710788727, "global_step": 136309, "epoch": 1531, "lr": 4.9805788726229803e-05}
{"train_loss": 0.1481940597295761, "global_step": 136310, "epoch": 1531, "lr": 4.980520899401284e-05}
{"train_loss": 0.11534135788679123, "global_step": 136311, "epoch": 1531, "lr": 4.980462926182205e-05}
{"train_loss": 0.13785187900066376, "global_step": 136312, "epoch": 1531, "lr": 4.980404952965754e-05}
{"train_loss": 0.17240458726882935, "global_step": 136313, "epoch": 1531, "lr": 4.980346979751936e-05}
{"train_loss": 0.13895146548748016, "global_step": 136314, "epoch": 1531, "lr": 4.980289006540762e-05}
{"train_loss": 0.07738877832889557, "global_step": 136315, "epoch": 1531, "lr": 4.9802310333322355e-05}
{"train_loss": 0.0966845378279686, "global_step": 136316, "epoch": 1531, "lr": 4.980173060126369e-05}
{"train_loss": 0.07129450142383575, "global_step": 136317, "epoch": 1531, "lr": 4.980115086923166e-05}
{"train_loss": 0.07716183364391327, "global_step": 136318, "epoch": 1531, "lr": 4.980057113722638e-05}
{"train_loss": 0.06396897882223129, "global_step": 136319, "epoch": 1531, "lr": 4.9799991405247883e-05}
{"train_loss": 0.0788053348660469, "global_step": 136320, "epoch": 1531, "lr": 4.9799411673296284e-05}
{"train_loss": 0.11390352994203568, "global_step": 136321, "epoch": 1531, "lr": 4.9798831941371675e-05}
{"train_loss": 0.07978788763284683, "global_step": 136322, "epoch": 1531, "lr": 4.979825220947409e-05}
{"train_loss": 0.05321415513753891, "global_step": 136323, "epoch": 1531, "lr": 4.979767247760364e-05}
{"train_loss": 0.20162290334701538, "global_step": 136324, "epoch": 1531, "lr": 4.979709274576039e-05}
{"train_loss": 0.13835741579532623, "global_step": 136325, "epoch": 1531, "lr": 4.979651301394441e-05}
{"train_loss": 0.11852477490901947, "global_step": 136326, "epoch": 1531, "lr": 4.9795933282155795e-05}
{"train_loss": 0.10733133554458618, "global_step": 136327, "epoch": 1531, "lr": 4.979535355039461e-05}
{"train_loss": 0.08959399908781052, "global_step": 136328, "epoch": 1531, "lr": 4.9794773818660934e-05}
{"train_loss": 0.11092323809862137, "global_step": 136329, "epoch": 1531, "lr": 4.9794194086954846e-05}
{"train_loss": 0.11676079034805298, "global_step": 136330, "epoch": 1531, "lr": 4.979361435527643e-05}
{"train_loss": 0.1128244549036026, "global_step": 136331, "epoch": 1531, "lr": 4.9793034623625773e-05}
{"train_loss": 0.07869989424943924, "global_step": 136332, "epoch": 1531, "lr": 4.979245489200292e-05}
{"train_loss": 0.13958488404750824, "global_step": 136333, "epoch": 1531, "lr": 4.979187516040799e-05}
{"train_loss": 0.11263415217399597, "global_step": 136334, "epoch": 1531, "lr": 4.979129542884101e-05}
{"train_loss": 0.1513177901506424, "global_step": 136335, "epoch": 1531, "lr": 4.979071569730211e-05}
{"train_loss": 0.08600444346666336, "global_step": 136336, "epoch": 1531, "lr": 4.9790135965791336e-05}
{"train_loss": 0.1435561329126358, "global_step": 136337, "epoch": 1531, "lr": 4.978955623430876e-05}
{"train_loss": 0.15202510356903076, "global_step": 136338, "epoch": 1531, "lr": 4.9788976502854514e-05}
{"train_loss": 0.11287138611078262, "global_step": 136339, "epoch": 1531, "lr": 4.978839677142861e-05}
{"train_loss": 0.07721920311450958, "global_step": 136340, "epoch": 1531, "lr": 4.978781704003117e-05}
{"train_loss": 0.1132434830069542, "global_step": 136341, "epoch": 1531, "lr": 4.978723730866224e-05}
{"train_loss": 0.09221623837947845, "global_step": 136342, "epoch": 1531, "lr": 4.978665757732193e-05}
{"train_loss": 0.09766070544719696, "global_step": 136343, "epoch": 1531, "lr": 4.978607784601028e-05}
{"train_loss": 0.06639697402715683, "global_step": 136344, "epoch": 1531, "lr": 4.978549811472741e-05}
{"train_loss": 0.16891337931156158, "global_step": 136345, "epoch": 1531, "lr": 4.9784918383473365e-05}
{"train_loss": 0.11623339354991913, "global_step": 136346, "epoch": 1531, "lr": 4.978433865224824e-05}
{"train_loss": 0.11339577015363768, "global_step": 136347, "epoch": 1531, "lr": 4.978375892105211e-05, "val_loss": 5.733619213104248}
{"train_loss": 0.06627274304628372, "global_step": 136348, "epoch": 1532, "lr": 4.978317918988505e-05}
{"train_loss": 0.10122764110565186, "global_step": 136349, "epoch": 1532, "lr": 4.978259945874713e-05}
{"train_loss": 0.09398817270994186, "global_step": 136350, "epoch": 1532, "lr": 4.978201972763846e-05}
{"train_loss": 0.12469618022441864, "global_step": 136351, "epoch": 1532, "lr": 4.978143999655907e-05}
{"train_loss": 0.08937360346317291, "global_step": 136352, "epoch": 1532, "lr": 4.978086026550908e-05}
{"train_loss": 0.10048618912696838, "global_step": 136353, "epoch": 1532, "lr": 4.9780280534488524e-05}
{"train_loss": 0.16173779964447021, "global_step": 136354, "epoch": 1532, "lr": 4.9779700803497535e-05}
{"train_loss": 0.07122395187616348, "global_step": 136355, "epoch": 1532, "lr": 4.9779121072536144e-05}
{"train_loss": 0.13139373064041138, "global_step": 136356, "epoch": 1532, "lr": 4.977854134160445e-05}
{"train_loss": 0.18148484826087952, "global_step": 136357, "epoch": 1532, "lr": 4.977796161070254e-05}
{"train_loss": 0.11952408403158188, "global_step": 136358, "epoch": 1532, "lr": 4.977738187983046e-05}
{"train_loss": 0.10323753952980042, "global_step": 136359, "epoch": 1532, "lr": 4.977680214898833e-05}
{"train_loss": 0.11803946644067764, "global_step": 136360, "epoch": 1532, "lr": 4.9776222418176195e-05}
{"train_loss": 0.09541212022304535, "global_step": 136361, "epoch": 1532, "lr": 4.977564268739416e-05}
{"train_loss": 0.17468096315860748, "global_step": 136362, "epoch": 1532, "lr": 4.977506295664226e-05}
{"train_loss": 0.1271427720785141, "global_step": 136363, "epoch": 1532, "lr": 4.9774483225920624e-05}
{"train_loss": 0.10234384983778, "global_step": 136364, "epoch": 1532, "lr": 4.9773903495229294e-05}
{"train_loss": 0.1060011237859726, "global_step": 136365, "epoch": 1532, "lr": 4.977332376456837e-05}
{"train_loss": 0.13008949160575867, "global_step": 136366, "epoch": 1532, "lr": 4.9772744033937905e-05}
{"train_loss": 0.06545588374137878, "global_step": 136367, "epoch": 1532, "lr": 4.9772164303338e-05}
{"train_loss": 0.13277889788150787, "global_step": 136368, "epoch": 1532, "lr": 4.9771584572768726e-05}
{"train_loss": 0.10345488041639328, "global_step": 136369, "epoch": 1532, "lr": 4.9771004842230175e-05}
{"train_loss": 0.1576964557170868, "global_step": 136370, "epoch": 1532, "lr": 4.9770425111722374e-05}
{"train_loss": 0.08215688169002533, "global_step": 136371, "epoch": 1532, "lr": 4.9769845381245466e-05}
{"train_loss": 0.20389573276042938, "global_step": 136372, "epoch": 1532, "lr": 4.976926565079948e-05}
{"train_loss": 0.15237760543823242, "global_step": 136373, "epoch": 1532, "lr": 4.976868592038453e-05}
{"train_loss": 0.11425101011991501, "global_step": 136374, "epoch": 1532, "lr": 4.976810619000066e-05}
{"train_loss": 0.0994722917675972, "global_step": 136375, "epoch": 1532, "lr": 4.9767526459647964e-05}
{"train_loss": 0.1585606038570404, "global_step": 136376, "epoch": 1532, "lr": 4.9766946729326544e-05}
{"train_loss": 0.08386585116386414, "global_step": 136377, "epoch": 1532, "lr": 4.976636699903643e-05}
{"train_loss": 0.11982797831296921, "global_step": 136378, "epoch": 1532, "lr": 4.976578726877775e-05}
{"train_loss": 0.15027838945388794, "global_step": 136379, "epoch": 1532, "lr": 4.9765207538550536e-05}
{"train_loss": 0.06252111494541168, "global_step": 136380, "epoch": 1532, "lr": 4.976462780835491e-05}
{"train_loss": 0.14020442962646484, "global_step": 136381, "epoch": 1532, "lr": 4.97640480781909e-05}
{"train_loss": 0.09719860553741455, "global_step": 136382, "epoch": 1532, "lr": 4.9763468348058636e-05}
{"train_loss": 0.10561586171388626, "global_step": 136383, "epoch": 1532, "lr": 4.976288861795816e-05}
{"train_loss": 0.08965978026390076, "global_step": 136384, "epoch": 1532, "lr": 4.976230888788955e-05}
{"train_loss": 0.11001837253570557, "global_step": 136385, "epoch": 1532, "lr": 4.976172915785291e-05}
{"train_loss": 0.06533656269311905, "global_step": 136386, "epoch": 1532, "lr": 4.976114942784829e-05}
{"train_loss": 0.182585671544075, "global_step": 136387, "epoch": 1532, "lr": 4.976056969787579e-05}
{"train_loss": 0.138590469956398, "global_step": 136388, "epoch": 1532, "lr": 4.975998996793549e-05}
{"train_loss": 0.10690455138683319, "global_step": 136389, "epoch": 1532, "lr": 4.975941023802743e-05}
{"train_loss": 0.0596630834043026, "global_step": 136390, "epoch": 1532, "lr": 4.975883050815174e-05}
{"train_loss": 0.14624299108982086, "global_step": 136391, "epoch": 1532, "lr": 4.9758250778308445e-05}
{"train_loss": 0.1468687355518341, "global_step": 136392, "epoch": 1532, "lr": 4.975767104849765e-05}
{"train_loss": 0.15546293556690216, "global_step": 136393, "epoch": 1532, "lr": 4.9757091318719464e-05}
{"train_loss": 0.20566906034946442, "global_step": 136394, "epoch": 1532, "lr": 4.9756511588973905e-05}
{"train_loss": 0.1600206196308136, "global_step": 136395, "epoch": 1532, "lr": 4.97559318592611e-05}
{"train_loss": 0.06595267355442047, "global_step": 136396, "epoch": 1532, "lr": 4.975535212958109e-05}
{"train_loss": 0.08483462780714035, "global_step": 136397, "epoch": 1532, "lr": 4.975477239993399e-05}
{"train_loss": 0.12256712466478348, "global_step": 136398, "epoch": 1532, "lr": 4.9754192670319835e-05}
{"train_loss": 0.19945593178272247, "global_step": 136399, "epoch": 1532, "lr": 4.975361294073875e-05}
{"train_loss": 0.07716305553913116, "global_step": 136400, "epoch": 1532, "lr": 4.9753033211190765e-05}
{"train_loss": 0.09209516644477844, "global_step": 136401, "epoch": 1532, "lr": 4.9752453481676005e-05}
{"train_loss": 0.16370071470737457, "global_step": 136402, "epoch": 1532, "lr": 4.975187375219451e-05}
{"train_loss": 0.11455146968364716, "global_step": 136403, "epoch": 1532, "lr": 4.9751294022746375e-05}
{"train_loss": 0.15505775809288025, "global_step": 136404, "epoch": 1532, "lr": 4.975071429333168e-05}
{"train_loss": 0.1367892026901245, "global_step": 136405, "epoch": 1532, "lr": 4.975013456395049e-05}
{"train_loss": 0.03375779837369919, "global_step": 136406, "epoch": 1532, "lr": 4.97495548346029e-05}
{"train_loss": 0.08092879503965378, "global_step": 136407, "epoch": 1532, "lr": 4.9748975105288986e-05}
{"train_loss": 0.1183714047074318, "global_step": 136408, "epoch": 1532, "lr": 4.9748395376008794e-05}
{"train_loss": 0.10553067177534103, "global_step": 136409, "epoch": 1532, "lr": 4.9747815646762455e-05}
{"train_loss": 0.17025995254516602, "global_step": 136410, "epoch": 1532, "lr": 4.974723591755e-05}
{"train_loss": 0.09143953770399094, "global_step": 136411, "epoch": 1532, "lr": 4.974665618837152e-05}
{"train_loss": 0.1234065443277359, "global_step": 136412, "epoch": 1532, "lr": 4.9746076459227114e-05}
{"train_loss": 0.10344456881284714, "global_step": 136413, "epoch": 1532, "lr": 4.974549673011683e-05}
{"train_loss": 0.1619560718536377, "global_step": 136414, "epoch": 1532, "lr": 4.9744917001040785e-05}
{"train_loss": 0.07666561007499695, "global_step": 136415, "epoch": 1532, "lr": 4.974433727199901e-05}
{"train_loss": 0.09146212041378021, "global_step": 136416, "epoch": 1532, "lr": 4.974375754299163e-05}
{"train_loss": 0.14828133583068848, "global_step": 136417, "epoch": 1532, "lr": 4.974317781401867e-05}
{"train_loss": 0.1351843923330307, "global_step": 136418, "epoch": 1532, "lr": 4.974259808508026e-05}
{"train_loss": 0.09208391606807709, "global_step": 136419, "epoch": 1532, "lr": 4.9742018356176434e-05}
{"train_loss": 0.11163657903671265, "global_step": 136420, "epoch": 1532, "lr": 4.974143862730732e-05}
{"train_loss": 0.12132244557142258, "global_step": 136421, "epoch": 1532, "lr": 4.974085889847295e-05}
{"train_loss": 0.17482610046863556, "global_step": 136422, "epoch": 1532, "lr": 4.974027916967341e-05}
{"train_loss": 0.14412489533424377, "global_step": 136423, "epoch": 1532, "lr": 4.9739699440908796e-05}
{"train_loss": 0.11656686663627625, "global_step": 136424, "epoch": 1532, "lr": 4.9739119712179174e-05}
{"train_loss": 0.10645163804292679, "global_step": 136425, "epoch": 1532, "lr": 4.973853998348462e-05}
{"train_loss": 0.13521061837673187, "global_step": 136426, "epoch": 1532, "lr": 4.973796025482522e-05}
{"train_loss": 0.11139677464962006, "global_step": 136427, "epoch": 1532, "lr": 4.973738052620105e-05}
{"train_loss": 0.12230277806520462, "global_step": 136428, "epoch": 1532, "lr": 4.973680079761217e-05}
{"train_loss": 0.09586389362812042, "global_step": 136429, "epoch": 1532, "lr": 4.97362210690587e-05}
{"train_loss": 0.08672884106636047, "global_step": 136430, "epoch": 1532, "lr": 4.9735641340540676e-05}
{"train_loss": 0.07620566338300705, "global_step": 136431, "epoch": 1532, "lr": 4.97350616120582e-05}
{"train_loss": 0.1415189802646637, "global_step": 136432, "epoch": 1532, "lr": 4.973448188361133e-05}
{"train_loss": 0.10366802662611008, "global_step": 136433, "epoch": 1532, "lr": 4.973390215520017e-05}
{"train_loss": 0.09852902591228485, "global_step": 136434, "epoch": 1532, "lr": 4.973332242682477e-05}
{"train_loss": 0.13007616996765137, "global_step": 136435, "epoch": 1532, "lr": 4.9732742698485236e-05}
{"train_loss": 0.11830818690778164, "global_step": 136436, "epoch": 1532, "lr": 4.973216297018161e-05, "val_loss": 5.580807209014893}
{"train_loss": 0.1264171600341797, "global_step": 136437, "epoch": 1533, "lr": 4.973158324191402e-05}
{"train_loss": 0.10298396646976471, "global_step": 136438, "epoch": 1533, "lr": 4.973100351368249e-05}
{"train_loss": 0.071892149746418, "global_step": 136439, "epoch": 1533, "lr": 4.973042378548714e-05}
{"train_loss": 0.14242790639400482, "global_step": 136440, "epoch": 1533, "lr": 4.9729844057328016e-05}
{"train_loss": 0.12337420880794525, "global_step": 136441, "epoch": 1533, "lr": 4.972926432920523e-05}
{"train_loss": 0.1124408021569252, "global_step": 136442, "epoch": 1533, "lr": 4.972868460111883e-05}
{"train_loss": 0.11786072701215744, "global_step": 136443, "epoch": 1533, "lr": 4.972810487306891e-05}
{"train_loss": 0.04929596930742264, "global_step": 136444, "epoch": 1533, "lr": 4.9727525145055534e-05}
{"train_loss": 0.09310301393270493, "global_step": 136445, "epoch": 1533, "lr": 4.9726945417078796e-05}
{"train_loss": 0.13241440057754517, "global_step": 136446, "epoch": 1533, "lr": 4.9726365689138765e-05}
{"train_loss": 0.10074770450592041, "global_step": 136447, "epoch": 1533, "lr": 4.972578596123551e-05}
{"train_loss": 0.1697031557559967, "global_step": 136448, "epoch": 1533, "lr": 4.972520623336914e-05}
{"train_loss": 0.10379230976104736, "global_step": 136449, "epoch": 1533, "lr": 4.9724626505539694e-05}
{"train_loss": 0.10461515188217163, "global_step": 136450, "epoch": 1533, "lr": 4.972404677774729e-05}
{"train_loss": 0.0817076712846756, "global_step": 136451, "epoch": 1533, "lr": 4.972346704999196e-05}
{"train_loss": 0.11436575651168823, "global_step": 136452, "epoch": 1533, "lr": 4.972288732227383e-05}
{"train_loss": 0.11386539787054062, "global_step": 136453, "epoch": 1533, "lr": 4.9722307594592933e-05}
{"train_loss": 0.16272981464862823, "global_step": 136454, "epoch": 1533, "lr": 4.972172786694939e-05}
{"train_loss": 0.14486365020275116, "global_step": 136455, "epoch": 1533, "lr": 4.9721148139343234e-05}
{"train_loss": 0.13467910885810852, "global_step": 136456, "epoch": 1533, "lr": 4.972056841177459e-05}
{"train_loss": 0.07085128128528595, "global_step": 136457, "epoch": 1533, "lr": 4.971998868424349e-05}
{"train_loss": 0.08147697895765305, "global_step": 136458, "epoch": 1533, "lr": 4.971940895675005e-05}
{"train_loss": 0.09257712960243225, "global_step": 136459, "epoch": 1533, "lr": 4.9718829229294314e-05}
{"train_loss": 0.10516207665205002, "global_step": 136460, "epoch": 1533, "lr": 4.971824950187641e-05}
{"train_loss": 0.18499617278575897, "global_step": 136461, "epoch": 1533, "lr": 4.971766977449636e-05}
{"train_loss": 0.17578092217445374, "global_step": 136462, "epoch": 1533, "lr": 4.971709004715427e-05}
{"train_loss": 0.07378635555505753, "global_step": 136463, "epoch": 1533, "lr": 4.971651031985021e-05}
{"train_loss": 0.10839543491601944, "global_step": 136464, "epoch": 1533, "lr": 4.971593059258426e-05}
{"train_loss": 0.12992218136787415, "global_step": 136465, "epoch": 1533, "lr": 4.971535086535651e-05}
{"train_loss": 0.11537081003189087, "global_step": 136466, "epoch": 1533, "lr": 4.971477113816701e-05}
{"train_loss": 0.09489405900239944, "global_step": 136467, "epoch": 1533, "lr": 4.971419141101589e-05}
{"train_loss": 0.09970231354236603, "global_step": 136468, "epoch": 1533, "lr": 4.971361168390316e-05}
{"train_loss": 0.06826868653297424, "global_step": 136469, "epoch": 1533, "lr": 4.971303195682895e-05}
{"train_loss": 0.11175122112035751, "global_step": 136470, "epoch": 1533, "lr": 4.971245222979331e-05}
{"train_loss": 0.09422547370195389, "global_step": 136471, "epoch": 1533, "lr": 4.9711872502796336e-05}
{"train_loss": 0.12947776913642883, "global_step": 136472, "epoch": 1533, "lr": 4.9711292775838084e-05}
{"train_loss": 0.1313185691833496, "global_step": 136473, "epoch": 1533, "lr": 4.971071304891866e-05}
{"train_loss": 0.15459245443344116, "global_step": 136474, "epoch": 1533, "lr": 4.971013332203811e-05}
{"train_loss": 0.12507155537605286, "global_step": 136475, "epoch": 1533, "lr": 4.970955359519655e-05}
{"train_loss": 0.11218828707933426, "global_step": 136476, "epoch": 1533, "lr": 4.970897386839401e-05}
{"train_loss": 0.14543206989765167, "global_step": 136477, "epoch": 1533, "lr": 4.970839414163063e-05}
{"train_loss": 0.17531883716583252, "global_step": 136478, "epoch": 1533, "lr": 4.9707814414906415e-05}
{"train_loss": 0.07796253263950348, "global_step": 136479, "epoch": 1533, "lr": 4.970723468822151e-05}
{"train_loss": 0.10641800612211227, "global_step": 136480, "epoch": 1533, "lr": 4.9706654961575956e-05}
{"train_loss": 0.14197354018688202, "global_step": 136481, "epoch": 1533, "lr": 4.970607523496983e-05}
{"train_loss": 0.14094646275043488, "global_step": 136482, "epoch": 1533, "lr": 4.970549550840323e-05}
{"train_loss": 0.12139604985713959, "global_step": 136483, "epoch": 1533, "lr": 4.9704915781876205e-05}
{"train_loss": 0.14653344452381134, "global_step": 136484, "epoch": 1533, "lr": 4.970433605538887e-05}
{"train_loss": 0.0443245954811573, "global_step": 136485, "epoch": 1533, "lr": 4.9703756328941264e-05}
{"train_loss": 0.05920900031924248, "global_step": 136486, "epoch": 1533, "lr": 4.970317660253351e-05}
{"train_loss": 0.12185947597026825, "global_step": 136487, "epoch": 1533, "lr": 4.970259687616564e-05}
{"train_loss": 0.10324816405773163, "global_step": 136488, "epoch": 1533, "lr": 4.970201714983776e-05}
{"train_loss": 0.07648438215255737, "global_step": 136489, "epoch": 1533, "lr": 4.970143742354993e-05}
{"train_loss": 0.06873945146799088, "global_step": 136490, "epoch": 1533, "lr": 4.970085769730226e-05}
{"train_loss": 0.13802100718021393, "global_step": 136491, "epoch": 1533, "lr": 4.9700277971094786e-05}
{"train_loss": 0.06427192687988281, "global_step": 136492, "epoch": 1533, "lr": 4.9699698244927616e-05}
{"train_loss": 0.08355482667684555, "global_step": 136493, "epoch": 1533, "lr": 4.9699118518800805e-05}
{"train_loss": 0.12387154996395111, "global_step": 136494, "epoch": 1533, "lr": 4.969853879271447e-05}
{"train_loss": 0.05743466317653656, "global_step": 136495, "epoch": 1533, "lr": 4.9697959066668634e-05}
{"train_loss": 0.11871299892663956, "global_step": 136496, "epoch": 1533, "lr": 4.969737934066342e-05}
{"train_loss": 0.07942749559879303, "global_step": 136497, "epoch": 1533, "lr": 4.969679961469888e-05}
{"train_loss": 0.12000974267721176, "global_step": 136498, "epoch": 1533, "lr": 4.969621988877512e-05}
{"train_loss": 0.10807324945926666, "global_step": 136499, "epoch": 1533, "lr": 4.969564016289218e-05}
{"train_loss": 0.13110987842082977, "global_step": 136500, "epoch": 1533, "lr": 4.9695060437050165e-05}
{"train_loss": 0.1488640457391739, "global_step": 136501, "epoch": 1533, "lr": 4.969448071124915e-05}
{"train_loss": 0.06257320195436478, "global_step": 136502, "epoch": 1533, "lr": 4.969390098548919e-05}
{"train_loss": 0.10388447344303131, "global_step": 136503, "epoch": 1533, "lr": 4.96933212597704e-05}
{"train_loss": 0.1455363631248474, "global_step": 136504, "epoch": 1533, "lr": 4.969274153409283e-05}
{"train_loss": 0.14087486267089844, "global_step": 136505, "epoch": 1533, "lr": 4.969216180845658e-05}
{"train_loss": 0.07822288572788239, "global_step": 136506, "epoch": 1533, "lr": 4.969158208286169e-05}
{"train_loss": 0.09391364455223083, "global_step": 136507, "epoch": 1533, "lr": 4.9691002357308294e-05}
{"train_loss": 0.07637777179479599, "global_step": 136508, "epoch": 1533, "lr": 4.9690422631796415e-05}
{"train_loss": 0.07654806971549988, "global_step": 136509, "epoch": 1533, "lr": 4.968984290632617e-05}
{"train_loss": 0.08092420548200607, "global_step": 136510, "epoch": 1533, "lr": 4.9689263180897605e-05}
{"train_loss": 0.07773163914680481, "global_step": 136511, "epoch": 1533, "lr": 4.968868345551084e-05}
{"train_loss": 0.06162594258785248, "global_step": 136512, "epoch": 1533, "lr": 4.9688103730165905e-05}
{"train_loss": 0.1201736330986023, "global_step": 136513, "epoch": 1533, "lr": 4.968752400486291e-05}
{"train_loss": 0.11181014776229858, "global_step": 136514, "epoch": 1533, "lr": 4.968694427960192e-05}
{"train_loss": 0.0917867124080658, "global_step": 136515, "epoch": 1533, "lr": 4.9686364554383024e-05}
{"train_loss": 0.12252253293991089, "global_step": 136516, "epoch": 1533, "lr": 4.968578482920628e-05}
{"train_loss": 0.08624375611543655, "global_step": 136517, "epoch": 1533, "lr": 4.96852051040718e-05}
{"train_loss": 0.15798057615756989, "global_step": 136518, "epoch": 1533, "lr": 4.968462537897962e-05}
{"train_loss": 0.07758637517690659, "global_step": 136519, "epoch": 1533, "lr": 4.9684045653929836e-05}
{"train_loss": 0.11611997336149216, "global_step": 136520, "epoch": 1533, "lr": 4.9683465928922545e-05}
{"train_loss": 0.11721523851156235, "global_step": 136521, "epoch": 1533, "lr": 4.968288620395779e-05}
{"train_loss": 0.13773980736732483, "global_step": 136522, "epoch": 1533, "lr": 4.968230647903568e-05}
{"train_loss": 0.08886788785457611, "global_step": 136523, "epoch": 1533, "lr": 4.968172675415627e-05}
{"train_loss": 0.11552898585796356, "global_step": 136524, "epoch": 1533, "lr": 4.968114702931967e-05}
{"train_loss": 0.10915514784917403, "global_step": 136525, "epoch": 1533, "lr": 4.9680567304525905e-05, "val_loss": 5.865535736083984}
{"train_loss": 0.08617939800024033, "global_step": 136526, "epoch": 1534, "lr": 4.967998757977511e-05}
{"train_loss": 0.12839724123477936, "global_step": 136527, "epoch": 1534, "lr": 4.9679407855067316e-05}
{"train_loss": 0.09105291962623596, "global_step": 136528, "epoch": 1534, "lr": 4.9678828130402646e-05}
{"train_loss": 0.0687016099691391, "global_step": 136529, "epoch": 1534, "lr": 4.967824840578113e-05}
{"train_loss": 0.09477293491363525, "global_step": 136530, "epoch": 1534, "lr": 4.967766868120289e-05}
{"train_loss": 0.199183851480484, "global_step": 136531, "epoch": 1534, "lr": 4.967708895666796e-05}
{"train_loss": 0.06498274952173233, "global_step": 136532, "epoch": 1534, "lr": 4.967650923217646e-05}
{"train_loss": 0.10399437695741653, "global_step": 136533, "epoch": 1534, "lr": 4.967592950772844e-05}
{"train_loss": 0.14407405257225037, "global_step": 136534, "epoch": 1534, "lr": 4.9675349783324e-05}
{"train_loss": 0.1012088879942894, "global_step": 136535, "epoch": 1534, "lr": 4.967477005896318e-05}
{"train_loss": 0.0778636559844017, "global_step": 136536, "epoch": 1534, "lr": 4.9674190334646114e-05}
{"train_loss": 0.06506326049566269, "global_step": 136537, "epoch": 1534, "lr": 4.9673610610372835e-05}
{"train_loss": 0.11072853952646255, "global_step": 136538, "epoch": 1534, "lr": 4.9673030886143425e-05}
{"train_loss": 0.13725405931472778, "global_step": 136539, "epoch": 1534, "lr": 4.967245116195799e-05}
{"train_loss": 0.06320083141326904, "global_step": 136540, "epoch": 1534, "lr": 4.967187143781657e-05}
{"train_loss": 0.11462969332933426, "global_step": 136541, "epoch": 1534, "lr": 4.967129171371929e-05}
{"train_loss": 0.14149364829063416, "global_step": 136542, "epoch": 1534, "lr": 4.967071198966617e-05}
{"train_loss": 0.12150522321462631, "global_step": 136543, "epoch": 1534, "lr": 4.967013226565734e-05}
{"train_loss": 0.12390562891960144, "global_step": 136544, "epoch": 1534, "lr": 4.966955254169284e-05}
{"train_loss": 0.04946774244308472, "global_step": 136545, "epoch": 1534, "lr": 4.966897281777279e-05}
{"train_loss": 0.12323901057243347, "global_step": 136546, "epoch": 1534, "lr": 4.9668393093897215e-05}
{"train_loss": 0.1836266666650772, "global_step": 136547, "epoch": 1534, "lr": 4.966781337006624e-05}
{"train_loss": 0.1502688229084015, "global_step": 136548, "epoch": 1534, "lr": 4.9667233646279906e-05}
{"train_loss": 0.14908313751220703, "global_step": 136549, "epoch": 1534, "lr": 4.966665392253834e-05}
{"train_loss": 0.10554613918066025, "global_step": 136550, "epoch": 1534, "lr": 4.966607419884155e-05}
{"train_loss": 0.06714553385972977, "global_step": 136551, "epoch": 1534, "lr": 4.966549447518968e-05}
{"train_loss": 0.12538762390613556, "global_step": 136552, "epoch": 1534, "lr": 4.966491475158277e-05}
{"train_loss": 0.10032451897859573, "global_step": 136553, "epoch": 1534, "lr": 4.966433502802091e-05}
{"train_loss": 0.11219733953475952, "global_step": 136554, "epoch": 1534, "lr": 4.9663755304504164e-05}
{"train_loss": 0.14852716028690338, "global_step": 136555, "epoch": 1534, "lr": 4.966317558103263e-05}
{"train_loss": 0.07314442843198776, "global_step": 136556, "epoch": 1534, "lr": 4.966259585760638e-05}
{"train_loss": 0.07982202619314194, "global_step": 136557, "epoch": 1534, "lr": 4.966201613422549e-05}
{"train_loss": 0.1462964415550232, "global_step": 136558, "epoch": 1534, "lr": 4.966143641089005e-05}
{"train_loss": 0.09101536124944687, "global_step": 136559, "epoch": 1534, "lr": 4.966085668760011e-05}
{"train_loss": 0.16991513967514038, "global_step": 136560, "epoch": 1534, "lr": 4.9660276964355776e-05}
{"train_loss": 0.1216820776462555, "global_step": 136561, "epoch": 1534, "lr": 4.965969724115709e-05}
{"train_loss": 0.11802012473344803, "global_step": 136562, "epoch": 1534, "lr": 4.965911751800418e-05}
{"train_loss": 0.10830702632665634, "global_step": 136563, "epoch": 1534, "lr": 4.9658537794897073e-05}
{"train_loss": 0.1030040979385376, "global_step": 136564, "epoch": 1534, "lr": 4.96579580718359e-05}
{"train_loss": 0.1551588922739029, "global_step": 136565, "epoch": 1534, "lr": 4.965737834882068e-05}
{"train_loss": 0.10583794862031937, "global_step": 136566, "epoch": 1534, "lr": 4.965679862585155e-05}
{"train_loss": 0.07816164940595627, "global_step": 136567, "epoch": 1534, "lr": 4.965621890292853e-05}
{"train_loss": 0.12992441654205322, "global_step": 136568, "epoch": 1534, "lr": 4.965563918005175e-05}
{"train_loss": 0.12030234187841415, "global_step": 136569, "epoch": 1534, "lr": 4.965505945722125e-05}
{"train_loss": 0.10281948745250702, "global_step": 136570, "epoch": 1534, "lr": 4.965447973443714e-05}
{"train_loss": 0.11306550353765488, "global_step": 136571, "epoch": 1534, "lr": 4.965390001169946e-05}
{"train_loss": 0.06541532278060913, "global_step": 136572, "epoch": 1534, "lr": 4.965332028900833e-05}
{"train_loss": 0.0942763164639473, "global_step": 136573, "epoch": 1534, "lr": 4.9652740566363785e-05}
{"train_loss": 0.0869341492652893, "global_step": 136574, "epoch": 1534, "lr": 4.9652160843765944e-05}
{"train_loss": 0.06924821436405182, "global_step": 136575, "epoch": 1534, "lr": 4.965158112121486e-05}
{"train_loss": 0.11743515729904175, "global_step": 136576, "epoch": 1534, "lr": 4.965100139871061e-05}
{"train_loss": 0.13357113301753998, "global_step": 136577, "epoch": 1534, "lr": 4.9650421676253296e-05}
{"train_loss": 0.14118005335330963, "global_step": 136578, "epoch": 1534, "lr": 4.9649841953842954e-05}
{"train_loss": 0.10912615060806274, "global_step": 136579, "epoch": 1534, "lr": 4.9649262231479714e-05}
{"train_loss": 0.0888010635972023, "global_step": 136580, "epoch": 1534, "lr": 4.9648682509163595e-05}
{"train_loss": 0.1221844032406807, "global_step": 136581, "epoch": 1534, "lr": 4.964810278689474e-05}
{"train_loss": 0.15767022967338562, "global_step": 136582, "epoch": 1534, "lr": 4.9647523064673166e-05}
{"train_loss": 0.1257399469614029, "global_step": 136583, "epoch": 1534, "lr": 4.9646943342499e-05}
{"train_loss": 0.07505668699741364, "global_step": 136584, "epoch": 1534, "lr": 4.964636362037227e-05}
{"train_loss": 0.14280425012111664, "global_step": 136585, "epoch": 1534, "lr": 4.9645783898293116e-05}
{"train_loss": 0.17679263651371002, "global_step": 136586, "epoch": 1534, "lr": 4.9645204176261554e-05}
{"train_loss": 0.12652350962162018, "global_step": 136587, "epoch": 1534, "lr": 4.96446244542777e-05}
{"train_loss": 0.09867751598358154, "global_step": 136588, "epoch": 1534, "lr": 4.964404473234162e-05}
{"train_loss": 0.08333025872707367, "global_step": 136589, "epoch": 1534, "lr": 4.964346501045341e-05}
{"train_loss": 0.10608580708503723, "global_step": 136590, "epoch": 1534, "lr": 4.9642885288613105e-05}
{"train_loss": 0.11210346966981888, "global_step": 136591, "epoch": 1534, "lr": 4.9642305566820826e-05}
{"train_loss": 0.07847987115383148, "global_step": 136592, "epoch": 1534, "lr": 4.9641725845076624e-05}
{"train_loss": 0.08869519084692001, "global_step": 136593, "epoch": 1534, "lr": 4.96411461233806e-05}
{"train_loss": 0.08427823334932327, "global_step": 136594, "epoch": 1534, "lr": 4.964056640173281e-05}
{"train_loss": 0.07913781702518463, "global_step": 136595, "epoch": 1534, "lr": 4.9639986680133346e-05}
{"train_loss": 0.0732777863740921, "global_step": 136596, "epoch": 1534, "lr": 4.9639406958582295e-05}
{"train_loss": 0.06775577366352081, "global_step": 136597, "epoch": 1534, "lr": 4.96388272370797e-05}
{"train_loss": 0.11849618703126907, "global_step": 136598, "epoch": 1534, "lr": 4.963824751562567e-05}
{"train_loss": 0.15411120653152466, "global_step": 136599, "epoch": 1534, "lr": 4.9637667794220264e-05}
{"train_loss": 0.1391853243112564, "global_step": 136600, "epoch": 1534, "lr": 4.963708807286359e-05}
{"train_loss": 0.12817826867103577, "global_step": 136601, "epoch": 1534, "lr": 4.9636508351555685e-05}
{"train_loss": 0.15640196204185486, "global_step": 136602, "epoch": 1534, "lr": 4.9635928630296664e-05}
{"train_loss": 0.09203856438398361, "global_step": 136603, "epoch": 1534, "lr": 4.963534890908657e-05}
{"train_loss": 0.10578227043151855, "global_step": 136604, "epoch": 1534, "lr": 4.963476918792551e-05}
{"train_loss": 0.0819869339466095, "global_step": 136605, "epoch": 1534, "lr": 4.9634189466813535e-05}
{"train_loss": 0.049643680453300476, "global_step": 136606, "epoch": 1534, "lr": 4.963360974575076e-05}
{"train_loss": 0.14462783932685852, "global_step": 136607, "epoch": 1534, "lr": 4.9633030024737224e-05}
{"train_loss": 0.1028800681233406, "global_step": 136608, "epoch": 1534, "lr": 4.9632450303773045e-05}
{"train_loss": 0.1101447194814682, "global_step": 136609, "epoch": 1534, "lr": 4.963187058285825e-05}
{"train_loss": 0.11550354957580566, "global_step": 136610, "epoch": 1534, "lr": 4.9631290861992965e-05}
{"train_loss": 0.13272370398044586, "global_step": 136611, "epoch": 1534, "lr": 4.963071114117725e-05}
{"train_loss": 0.13393260538578033, "global_step": 136612, "epoch": 1534, "lr": 4.963013142041117e-05}
{"train_loss": 0.08637277781963348, "global_step": 136613, "epoch": 1534, "lr": 4.962955169969482e-05}
{"train_loss": 0.11076489258348272, "global_step": 136614, "epoch": 1534, "lr": 4.962897197902827e-05, "val_loss": 5.980443477630615}
{"train_loss": 0.1368236094713211, "global_step": 136615, "epoch": 1535, "lr": 4.962839225841162e-05}
{"train_loss": 0.138371080160141, "global_step": 136616, "epoch": 1535, "lr": 4.9627812537844894e-05}
{"train_loss": 0.09843479096889496, "global_step": 136617, "epoch": 1535, "lr": 4.9627232817328226e-05}
{"train_loss": 0.12798205018043518, "global_step": 136618, "epoch": 1535, "lr": 4.9626653096861655e-05}
{"train_loss": 0.14487959444522858, "global_step": 136619, "epoch": 1535, "lr": 4.96260733764453e-05}
{"train_loss": 0.1484367847442627, "global_step": 136620, "epoch": 1535, "lr": 4.962549365607919e-05}
{"train_loss": 0.09994228929281235, "global_step": 136621, "epoch": 1535, "lr": 4.962491393576346e-05}
{"train_loss": 0.19394469261169434, "global_step": 136622, "epoch": 1535, "lr": 4.962433421549812e-05}
{"train_loss": 0.11662224680185318, "global_step": 136623, "epoch": 1535, "lr": 4.962375449528331e-05}
{"train_loss": 0.10132381319999695, "global_step": 136624, "epoch": 1535, "lr": 4.9623174775119056e-05}
{"train_loss": 0.08966226875782013, "global_step": 136625, "epoch": 1535, "lr": 4.9622595055005494e-05}
{"train_loss": 0.17200058698654175, "global_step": 136626, "epoch": 1535, "lr": 4.962201533494264e-05}
{"train_loss": 0.1577804684638977, "global_step": 136627, "epoch": 1535, "lr": 4.962143561493062e-05}
{"train_loss": 0.1026189997792244, "global_step": 136628, "epoch": 1535, "lr": 4.9620855894969484e-05}
{"train_loss": 0.11657184362411499, "global_step": 136629, "epoch": 1535, "lr": 4.962027617505932e-05}
{"train_loss": 0.11433790624141693, "global_step": 136630, "epoch": 1535, "lr": 4.9619696455200206e-05}
{"train_loss": 0.11440224200487137, "global_step": 136631, "epoch": 1535, "lr": 4.961911673539222e-05}
{"train_loss": 0.13491368293762207, "global_step": 136632, "epoch": 1535, "lr": 4.961853701563544e-05}
{"train_loss": 0.12400740385055542, "global_step": 136633, "epoch": 1535, "lr": 4.961795729592994e-05}
{"train_loss": 0.07261718809604645, "global_step": 136634, "epoch": 1535, "lr": 4.961737757627582e-05}
{"train_loss": 0.11790307611227036, "global_step": 136635, "epoch": 1535, "lr": 4.961679785667311e-05}
{"train_loss": 0.12635357677936554, "global_step": 136636, "epoch": 1535, "lr": 4.961621813712194e-05}
{"train_loss": 0.11855467408895493, "global_step": 136637, "epoch": 1535, "lr": 4.961563841762234e-05}
{"train_loss": 0.1228770911693573, "global_step": 136638, "epoch": 1535, "lr": 4.961505869817444e-05}
{"train_loss": 0.15712285041809082, "global_step": 136639, "epoch": 1535, "lr": 4.961447897877827e-05}
{"train_loss": 0.14409008622169495, "global_step": 136640, "epoch": 1535, "lr": 4.961389925943394e-05}
{"train_loss": 0.08553647249937057, "global_step": 136641, "epoch": 1535, "lr": 4.9613319540141506e-05}
{"train_loss": 0.08666041493415833, "global_step": 136642, "epoch": 1535, "lr": 4.961273982090108e-05}
{"train_loss": 0.08665534108877182, "global_step": 136643, "epoch": 1535, "lr": 4.961216010171268e-05}
{"train_loss": 0.11194204539060593, "global_step": 136644, "epoch": 1535, "lr": 4.9611580382576454e-05}
{"train_loss": 0.11088940501213074, "global_step": 136645, "epoch": 1535, "lr": 4.9611000663492424e-05}
{"train_loss": 0.1340271681547165, "global_step": 136646, "epoch": 1535, "lr": 4.9610420944460686e-05}
{"train_loss": 0.10865514725446701, "global_step": 136647, "epoch": 1535, "lr": 4.960984122548135e-05}
{"train_loss": 0.11642362177371979, "global_step": 136648, "epoch": 1535, "lr": 4.960926150655445e-05}
{"train_loss": 0.1572028398513794, "global_step": 136649, "epoch": 1535, "lr": 4.960868178768008e-05}
{"train_loss": 0.2090737372636795, "global_step": 136650, "epoch": 1535, "lr": 4.960810206885831e-05}
{"train_loss": 0.12192923575639725, "global_step": 136651, "epoch": 1535, "lr": 4.960752235008924e-05}
{"train_loss": 0.11699067056179047, "global_step": 136652, "epoch": 1535, "lr": 4.9606942631372934e-05}
{"train_loss": 0.10519073903560638, "global_step": 136653, "epoch": 1535, "lr": 4.960636291270947e-05}
{"train_loss": 0.1405995488166809, "global_step": 136654, "epoch": 1535, "lr": 4.9605783194098906e-05}
{"train_loss": 0.11108756810426712, "global_step": 136655, "epoch": 1535, "lr": 4.9605203475541365e-05}
{"train_loss": 0.22381176054477692, "global_step": 136656, "epoch": 1535, "lr": 4.960462375703688e-05}
{"train_loss": 0.12181985378265381, "global_step": 136657, "epoch": 1535, "lr": 4.9604044038585565e-05}
{"train_loss": 0.056230805814266205, "global_step": 136658, "epoch": 1535, "lr": 4.960346432018746e-05}
{"train_loss": 0.10468892753124237, "global_step": 136659, "epoch": 1535, "lr": 4.9602884601842695e-05}
{"train_loss": 0.1306852400302887, "global_step": 136660, "epoch": 1535, "lr": 4.9602304883551284e-05}
{"train_loss": 0.17054417729377747, "global_step": 136661, "epoch": 1535, "lr": 4.960172516531337e-05}
{"train_loss": 0.13658268749713898, "global_step": 136662, "epoch": 1535, "lr": 4.960114544712897e-05}
{"train_loss": 0.15434226393699646, "global_step": 136663, "epoch": 1535, "lr": 4.960056572899821e-05}
{"train_loss": 0.16255567967891693, "global_step": 136664, "epoch": 1535, "lr": 4.959998601092114e-05}
{"train_loss": 0.09898015111684799, "global_step": 136665, "epoch": 1535, "lr": 4.959940629289784e-05}
{"train_loss": 0.14471642673015594, "global_step": 136666, "epoch": 1535, "lr": 4.959882657492842e-05}
{"train_loss": 0.12517982721328735, "global_step": 136667, "epoch": 1535, "lr": 4.9598246857012914e-05}
{"train_loss": 0.12228964269161224, "global_step": 136668, "epoch": 1535, "lr": 4.9597667139151424e-05}
{"train_loss": 0.16470925509929657, "global_step": 136669, "epoch": 1535, "lr": 4.959708742134402e-05}
{"train_loss": 0.13541719317436218, "global_step": 136670, "epoch": 1535, "lr": 4.959650770359079e-05}
{"train_loss": 0.0891125276684761, "global_step": 136671, "epoch": 1535, "lr": 4.95959279858918e-05}
{"train_loss": 0.14245359599590302, "global_step": 136672, "epoch": 1535, "lr": 4.959534826824713e-05}
{"train_loss": 0.07787565886974335, "global_step": 136673, "epoch": 1535, "lr": 4.959476855065686e-05}
{"train_loss": 0.08682174980640411, "global_step": 136674, "epoch": 1535, "lr": 4.959418883312108e-05}
{"train_loss": 0.12550245225429535, "global_step": 136675, "epoch": 1535, "lr": 4.959360911563984e-05}
{"train_loss": 0.1629779189825058, "global_step": 136676, "epoch": 1535, "lr": 4.959302939821325e-05}
{"train_loss": 0.12386532872915268, "global_step": 136677, "epoch": 1535, "lr": 4.959244968084135e-05}
{"train_loss": 0.13635863363742828, "global_step": 136678, "epoch": 1535, "lr": 4.959186996352427e-05}
{"train_loss": 0.0878012403845787, "global_step": 136679, "epoch": 1535, "lr": 4.959129024626203e-05}
{"train_loss": 0.09948507696390152, "global_step": 136680, "epoch": 1535, "lr": 4.959071052905476e-05}
{"train_loss": 0.12465790659189224, "global_step": 136681, "epoch": 1535, "lr": 4.959013081190249e-05}
{"train_loss": 0.17264865338802338, "global_step": 136682, "epoch": 1535, "lr": 4.9589551094805324e-05}
{"train_loss": 0.1525733321905136, "global_step": 136683, "epoch": 1535, "lr": 4.958897137776336e-05}
{"train_loss": 0.12510938942432404, "global_step": 136684, "epoch": 1535, "lr": 4.9588391660776626e-05}
{"train_loss": 0.15571056306362152, "global_step": 136685, "epoch": 1535, "lr": 4.958781194384526e-05}
{"train_loss": 0.15413326025009155, "global_step": 136686, "epoch": 1535, "lr": 4.9587232226969274e-05}
{"train_loss": 0.1318470686674118, "global_step": 136687, "epoch": 1535, "lr": 4.9586652510148814e-05}
{"train_loss": 0.14514754712581635, "global_step": 136688, "epoch": 1535, "lr": 4.9586072793383906e-05}
{"train_loss": 0.13188135623931885, "global_step": 136689, "epoch": 1535, "lr": 4.9585493076674645e-05}
{"train_loss": 0.05541636049747467, "global_step": 136690, "epoch": 1535, "lr": 4.9584913360021104e-05}
{"train_loss": 0.08410076797008514, "global_step": 136691, "epoch": 1535, "lr": 4.958433364342337e-05}
{"train_loss": 0.1595958173274994, "global_step": 136692, "epoch": 1535, "lr": 4.9583753926881526e-05}
{"train_loss": 0.11594347655773163, "global_step": 136693, "epoch": 1535, "lr": 4.9583174210395644e-05}
{"train_loss": 0.2014772891998291, "global_step": 136694, "epoch": 1535, "lr": 4.958259449396578e-05}
{"train_loss": 0.12410067021846771, "global_step": 136695, "epoch": 1535, "lr": 4.9582014777592054e-05}
{"train_loss": 0.06851520389318466, "global_step": 136696, "epoch": 1535, "lr": 4.9581435061274494e-05}
{"train_loss": 0.09742460399866104, "global_step": 136697, "epoch": 1535, "lr": 4.958085534501323e-05}
{"train_loss": 0.09479457139968872, "global_step": 136698, "epoch": 1535, "lr": 4.9580275628808293e-05}
{"train_loss": 0.1283351480960846, "global_step": 136699, "epoch": 1535, "lr": 4.957969591265981e-05}
{"train_loss": 0.1570630967617035, "global_step": 136700, "epoch": 1535, "lr": 4.95791161965678e-05}
{"train_loss": 0.1503458023071289, "global_step": 136701, "epoch": 1535, "lr": 4.957853648053238e-05}
{"train_loss": 0.18414407968521118, "global_step": 136702, "epoch": 1535, "lr": 4.9577956764553634e-05}
{"train_loss": 0.12721348671096094, "global_step": 136703, "epoch": 1535, "lr": 4.957737704863161e-05, "val_loss": 5.693562030792236, "train_action_mse_error": 9.473909378051758}
{"train_loss": 0.16636548936367035, "global_step": 136704, "epoch": 1536, "lr": 4.957679733276643e-05}
{"train_loss": 0.1474931389093399, "global_step": 136705, "epoch": 1536, "lr": 4.957621761695811e-05}
{"train_loss": 0.14369438588619232, "global_step": 136706, "epoch": 1536, "lr": 4.95756379012068e-05}
{"train_loss": 0.048241447657346725, "global_step": 136707, "epoch": 1536, "lr": 4.957505818551252e-05}
{"train_loss": 0.21489188075065613, "global_step": 136708, "epoch": 1536, "lr": 4.957447846987537e-05}
{"train_loss": 0.16983070969581604, "global_step": 136709, "epoch": 1536, "lr": 4.957389875429542e-05}
{"train_loss": 0.0904267281293869, "global_step": 136710, "epoch": 1536, "lr": 4.9573319038772764e-05}
{"train_loss": 0.1583186239004135, "global_step": 136711, "epoch": 1536, "lr": 4.957273932330746e-05}
{"train_loss": 0.15820032358169556, "global_step": 136712, "epoch": 1536, "lr": 4.9572159607899616e-05}
{"train_loss": 0.15927532315254211, "global_step": 136713, "epoch": 1536, "lr": 4.9571579892549266e-05}
{"train_loss": 0.17279145121574402, "global_step": 136714, "epoch": 1536, "lr": 4.9571000177256535e-05}
{"train_loss": 0.10183559358119965, "global_step": 136715, "epoch": 1536, "lr": 4.957042046202145e-05}
{"train_loss": 0.11029413342475891, "global_step": 136716, "epoch": 1536, "lr": 4.9569840746844146e-05}
{"train_loss": 0.10880210995674133, "global_step": 136717, "epoch": 1536, "lr": 4.956926103172464e-05}
{"train_loss": 0.08678798377513885, "global_step": 136718, "epoch": 1536, "lr": 4.956868131666305e-05}
{"train_loss": 0.13194096088409424, "global_step": 136719, "epoch": 1536, "lr": 4.956810160165946e-05}
{"train_loss": 0.12426836788654327, "global_step": 136720, "epoch": 1536, "lr": 4.956752188671392e-05}
{"train_loss": 0.14025621116161346, "global_step": 136721, "epoch": 1536, "lr": 4.956694217182654e-05}
{"train_loss": 0.17418010532855988, "global_step": 136722, "epoch": 1536, "lr": 4.956636245699736e-05}
{"train_loss": 0.1742825210094452, "global_step": 136723, "epoch": 1536, "lr": 4.9565782742226494e-05}
{"train_loss": 0.1358707994222641, "global_step": 136724, "epoch": 1536, "lr": 4.956520302751399e-05}
{"train_loss": 0.17279785871505737, "global_step": 136725, "epoch": 1536, "lr": 4.9564623312859946e-05}
{"train_loss": 0.09592115879058838, "global_step": 136726, "epoch": 1536, "lr": 4.956404359826443e-05}
{"train_loss": 0.14003849029541016, "global_step": 136727, "epoch": 1536, "lr": 4.9563463883727525e-05}
{"train_loss": 0.14670391380786896, "global_step": 136728, "epoch": 1536, "lr": 4.956288416924931e-05}
{"train_loss": 0.08559093624353409, "global_step": 136729, "epoch": 1536, "lr": 4.956230445482985e-05}
{"train_loss": 0.09058897197246552, "global_step": 136730, "epoch": 1536, "lr": 4.956172474046924e-05}
{"train_loss": 0.08041924983263016, "global_step": 136731, "epoch": 1536, "lr": 4.9561145026167564e-05}
{"train_loss": 0.07375628501176834, "global_step": 136732, "epoch": 1536, "lr": 4.9560565311924855e-05}
{"train_loss": 0.10307130962610245, "global_step": 136733, "epoch": 1536, "lr": 4.9559985597741256e-05}
{"train_loss": 0.09228324145078659, "global_step": 136734, "epoch": 1536, "lr": 4.955940588361678e-05}
{"train_loss": 0.09315658360719681, "global_step": 136735, "epoch": 1536, "lr": 4.955882616955156e-05}
{"train_loss": 0.1583249568939209, "global_step": 136736, "epoch": 1536, "lr": 4.955824645554564e-05}
{"train_loss": 0.07783567160367966, "global_step": 136737, "epoch": 1536, "lr": 4.95576667415991e-05}
{"train_loss": 0.10068313777446747, "global_step": 136738, "epoch": 1536, "lr": 4.955708702771204e-05}
{"train_loss": 0.09402082115411758, "global_step": 136739, "epoch": 1536, "lr": 4.955650731388451e-05}
{"train_loss": 0.08397451788187027, "global_step": 136740, "epoch": 1536, "lr": 4.955592760011663e-05}
{"train_loss": 0.161703422665596, "global_step": 136741, "epoch": 1536, "lr": 4.9555347886408424e-05}
{"train_loss": 0.1026226207613945, "global_step": 136742, "epoch": 1536, "lr": 4.9554768172760014e-05}
{"train_loss": 0.12751561403274536, "global_step": 136743, "epoch": 1536, "lr": 4.955418845917143e-05}
{"train_loss": 0.11861871182918549, "global_step": 136744, "epoch": 1536, "lr": 4.955360874564282e-05}
{"train_loss": 0.08752620965242386, "global_step": 136745, "epoch": 1536, "lr": 4.9553029032174205e-05}
{"train_loss": 0.14324922859668732, "global_step": 136746, "epoch": 1536, "lr": 4.9552449318765675e-05}
{"train_loss": 0.15194645524024963, "global_step": 136747, "epoch": 1536, "lr": 4.955186960541732e-05}
{"train_loss": 0.07646259665489197, "global_step": 136748, "epoch": 1536, "lr": 4.95512898921292e-05}
{"train_loss": 0.08182752877473831, "global_step": 136749, "epoch": 1536, "lr": 4.955071017890141e-05}
{"train_loss": 0.13216710090637207, "global_step": 136750, "epoch": 1536, "lr": 4.955013046573403e-05}
{"train_loss": 0.08457813411951065, "global_step": 136751, "epoch": 1536, "lr": 4.9549550752627114e-05}
{"train_loss": 0.1848946213722229, "global_step": 136752, "epoch": 1536, "lr": 4.954897103958077e-05}
{"train_loss": 0.12218614667654037, "global_step": 136753, "epoch": 1536, "lr": 4.9548391326595044e-05}
{"train_loss": 0.10405711829662323, "global_step": 136754, "epoch": 1536, "lr": 4.9547811613670055e-05}
{"train_loss": 0.17293794453144073, "global_step": 136755, "epoch": 1536, "lr": 4.9547231900805835e-05}
{"train_loss": 0.09825277328491211, "global_step": 136756, "epoch": 1536, "lr": 4.954665218800248e-05}
{"train_loss": 0.1317892074584961, "global_step": 136757, "epoch": 1536, "lr": 4.95460724752601e-05}
{"train_loss": 0.1464942842721939, "global_step": 136758, "epoch": 1536, "lr": 4.954549276257871e-05}
{"train_loss": 0.11088428646326065, "global_step": 136759, "epoch": 1536, "lr": 4.954491304995846e-05}
{"train_loss": 0.12645815312862396, "global_step": 136760, "epoch": 1536, "lr": 4.9544333337399355e-05}
{"train_loss": 0.17019128799438477, "global_step": 136761, "epoch": 1536, "lr": 4.954375362490154e-05}
{"train_loss": 0.09125696867704391, "global_step": 136762, "epoch": 1536, "lr": 4.954317391246504e-05}
{"train_loss": 0.146601602435112, "global_step": 136763, "epoch": 1536, "lr": 4.954259420008998e-05}
{"train_loss": 0.11496561020612717, "global_step": 136764, "epoch": 1536, "lr": 4.954201448777639e-05}
{"train_loss": 0.13318683207035065, "global_step": 136765, "epoch": 1536, "lr": 4.954143477552438e-05}
{"train_loss": 0.06811163574457169, "global_step": 136766, "epoch": 1536, "lr": 4.954085506333401e-05}
{"train_loss": 0.12464579939842224, "global_step": 136767, "epoch": 1536, "lr": 4.9540275351205375e-05}
{"train_loss": 0.14755557477474213, "global_step": 136768, "epoch": 1536, "lr": 4.953969563913854e-05}
{"train_loss": 0.08604995161294937, "global_step": 136769, "epoch": 1536, "lr": 4.95391159271336e-05}
{"train_loss": 0.131058007478714, "global_step": 136770, "epoch": 1536, "lr": 4.95385362151906e-05}
{"train_loss": 0.12000039964914322, "global_step": 136771, "epoch": 1536, "lr": 4.953795650330966e-05}
{"train_loss": 0.1341419517993927, "global_step": 136772, "epoch": 1536, "lr": 4.953737679149081e-05}
{"train_loss": 0.0818798765540123, "global_step": 136773, "epoch": 1536, "lr": 4.9536797079734154e-05}
{"train_loss": 0.09125049412250519, "global_step": 136774, "epoch": 1536, "lr": 4.9536217368039795e-05}
{"train_loss": 0.1045432910323143, "global_step": 136775, "epoch": 1536, "lr": 4.9535637656407764e-05}
{"train_loss": 0.10441683232784271, "global_step": 136776, "epoch": 1536, "lr": 4.953505794483818e-05}
{"train_loss": 0.07795828580856323, "global_step": 136777, "epoch": 1536, "lr": 4.953447823333108e-05}
{"train_loss": 0.12426190078258514, "global_step": 136778, "epoch": 1536, "lr": 4.953389852188659e-05}
{"train_loss": 0.1467442810535431, "global_step": 136779, "epoch": 1536, "lr": 4.953331881050474e-05}
{"train_loss": 0.09573528915643692, "global_step": 136780, "epoch": 1536, "lr": 4.953273909918564e-05}
{"train_loss": 0.06840019673109055, "global_step": 136781, "epoch": 1536, "lr": 4.953215938792935e-05}
{"train_loss": 0.07735399901866913, "global_step": 136782, "epoch": 1536, "lr": 4.953157967673597e-05}
{"train_loss": 0.08194954693317413, "global_step": 136783, "epoch": 1536, "lr": 4.953099996560555e-05}
{"train_loss": 0.07804229855537415, "global_step": 136784, "epoch": 1536, "lr": 4.953042025453819e-05}
{"train_loss": 0.14927461743354797, "global_step": 136785, "epoch": 1536, "lr": 4.952984054353396e-05}
{"train_loss": 0.14762452244758606, "global_step": 136786, "epoch": 1536, "lr": 4.952926083259293e-05}
{"train_loss": 0.1551949679851532, "global_step": 136787, "epoch": 1536, "lr": 4.952868112171518e-05}
{"train_loss": 0.11167692393064499, "global_step": 136788, "epoch": 1536, "lr": 4.95281014109008e-05}
{"train_loss": 0.11598663032054901, "global_step": 136789, "epoch": 1536, "lr": 4.9527521700149866e-05}
{"train_loss": 0.17638550698757172, "global_step": 136790, "epoch": 1536, "lr": 4.952694198946245e-05}
{"train_loss": 0.08370055258274078, "global_step": 136791, "epoch": 1536, "lr": 4.952636227883862e-05}
{"train_loss": 0.12060594981473484, "global_step": 136792, "epoch": 1536, "lr": 4.952578256827846e-05, "val_loss": 5.843952655792236}
{"train_loss": 0.16050273180007935, "global_step": 136793, "epoch": 1537, "lr": 4.952520285778208e-05}
{"train_loss": 0.13225166499614716, "global_step": 136794, "epoch": 1537, "lr": 4.9524623147349506e-05}
{"train_loss": 0.17578282952308655, "global_step": 136795, "epoch": 1537, "lr": 4.9524043436980855e-05}
{"train_loss": 0.12825803458690643, "global_step": 136796, "epoch": 1537, "lr": 4.952346372667618e-05}
{"train_loss": 0.12355288118124008, "global_step": 136797, "epoch": 1537, "lr": 4.9522884016435586e-05}
{"train_loss": 0.13273179531097412, "global_step": 136798, "epoch": 1537, "lr": 4.952230430625911e-05}
{"train_loss": 0.0754476860165596, "global_step": 136799, "epoch": 1537, "lr": 4.952172459614688e-05}
{"train_loss": 0.12253089994192123, "global_step": 136800, "epoch": 1537, "lr": 4.9521144886098916e-05}
{"train_loss": 0.09444776177406311, "global_step": 136801, "epoch": 1537, "lr": 4.952056517611536e-05}
{"train_loss": 0.0826169028878212, "global_step": 136802, "epoch": 1537, "lr": 4.951998546619624e-05}
{"train_loss": 0.12514406442642212, "global_step": 136803, "epoch": 1537, "lr": 4.951940575634167e-05}
{"train_loss": 0.08705878257751465, "global_step": 136804, "epoch": 1537, "lr": 4.95188260465517e-05}
{"train_loss": 0.11371644586324692, "global_step": 136805, "epoch": 1537, "lr": 4.9518246336826414e-05}
{"train_loss": 0.11959123611450195, "global_step": 136806, "epoch": 1537, "lr": 4.95176666271659e-05}
{"train_loss": 0.12893223762512207, "global_step": 136807, "epoch": 1537, "lr": 4.951708691757022e-05}
{"train_loss": 0.1346040815114975, "global_step": 136808, "epoch": 1537, "lr": 4.951650720803947e-05}
{"train_loss": 0.07132891565561295, "global_step": 136809, "epoch": 1537, "lr": 4.951592749857371e-05}
{"train_loss": 0.06600102037191391, "global_step": 136810, "epoch": 1537, "lr": 4.951534778917305e-05}
{"train_loss": 0.14404596388339996, "global_step": 136811, "epoch": 1537, "lr": 4.951476807983751e-05}
{"train_loss": 0.07811538130044937, "global_step": 136812, "epoch": 1537, "lr": 4.951418837056724e-05}
{"train_loss": 0.13484862446784973, "global_step": 136813, "epoch": 1537, "lr": 4.9513608661362256e-05}
{"train_loss": 0.11959864944219589, "global_step": 136814, "epoch": 1537, "lr": 4.951302895222268e-05}
{"train_loss": 0.12221148610115051, "global_step": 136815, "epoch": 1537, "lr": 4.951244924314855e-05}
{"train_loss": 0.1086248978972435, "global_step": 136816, "epoch": 1537, "lr": 4.951186953413999e-05}
{"train_loss": 0.09168103337287903, "global_step": 136817, "epoch": 1537, "lr": 4.951128982519703e-05}
{"train_loss": 0.132085382938385, "global_step": 136818, "epoch": 1537, "lr": 4.9510710116319796e-05}
{"train_loss": 0.1546778827905655, "global_step": 136819, "epoch": 1537, "lr": 4.951013040750832e-05}
{"train_loss": 0.16922679543495178, "global_step": 136820, "epoch": 1537, "lr": 4.950955069876271e-05}
{"train_loss": 0.13220448791980743, "global_step": 136821, "epoch": 1537, "lr": 4.9508970990083023e-05}
{"train_loss": 0.1256253570318222, "global_step": 136822, "epoch": 1537, "lr": 4.950839128146937e-05}
{"train_loss": 0.1891380399465561, "global_step": 136823, "epoch": 1537, "lr": 4.95078115729218e-05}
{"train_loss": 0.07561337202787399, "global_step": 136824, "epoch": 1537, "lr": 4.950723186444039e-05}
{"train_loss": 0.07792093604803085, "global_step": 136825, "epoch": 1537, "lr": 4.950665215602524e-05}
{"train_loss": 0.15424573421478271, "global_step": 136826, "epoch": 1537, "lr": 4.95060724476764e-05}
{"train_loss": 0.13665816187858582, "global_step": 136827, "epoch": 1537, "lr": 4.950549273939398e-05}
{"train_loss": 0.11075916141271591, "global_step": 136828, "epoch": 1537, "lr": 4.950491303117801e-05}
{"train_loss": 0.11112087219953537, "global_step": 136829, "epoch": 1537, "lr": 4.9504333323028635e-05}
{"train_loss": 0.12759792804718018, "global_step": 136830, "epoch": 1537, "lr": 4.950375361494586e-05}
{"train_loss": 0.07942338287830353, "global_step": 136831, "epoch": 1537, "lr": 4.9503173906929835e-05}
{"train_loss": 0.16014035046100616, "global_step": 136832, "epoch": 1537, "lr": 4.950259419898058e-05}
{"train_loss": 0.06014684960246086, "global_step": 136833, "epoch": 1537, "lr": 4.950201449109821e-05}
{"train_loss": 0.11860410869121552, "global_step": 136834, "epoch": 1537, "lr": 4.950143478328277e-05}
{"train_loss": 0.12276961654424667, "global_step": 136835, "epoch": 1537, "lr": 4.9500855075534376e-05}
{"train_loss": 0.08035211265087128, "global_step": 136836, "epoch": 1537, "lr": 4.950027536785306e-05}
{"train_loss": 0.11452777683734894, "global_step": 136837, "epoch": 1537, "lr": 4.9499695660238956e-05}
{"train_loss": 0.13483130931854248, "global_step": 136838, "epoch": 1537, "lr": 4.9499115952692084e-05}
{"train_loss": 0.0977495089173317, "global_step": 136839, "epoch": 1537, "lr": 4.9498536245212575e-05}
{"train_loss": 0.08041603118181229, "global_step": 136840, "epoch": 1537, "lr": 4.9497956537800455e-05}
{"train_loss": 0.09982406347990036, "global_step": 136841, "epoch": 1537, "lr": 4.9497376830455854e-05}
{"train_loss": 0.15539713203907013, "global_step": 136842, "epoch": 1537, "lr": 4.949679712317882e-05}
{"train_loss": 0.15636849403381348, "global_step": 136843, "epoch": 1537, "lr": 4.9496217415969426e-05}
{"train_loss": 0.11580321937799454, "global_step": 136844, "epoch": 1537, "lr": 4.949563770882776e-05}
{"train_loss": 0.15357495844364166, "global_step": 136845, "epoch": 1537, "lr": 4.949505800175389e-05}
{"train_loss": 0.10980618745088577, "global_step": 136846, "epoch": 1537, "lr": 4.949447829474793e-05}
{"train_loss": 0.09200501441955566, "global_step": 136847, "epoch": 1537, "lr": 4.94938985878099e-05}
{"train_loss": 0.1489308923482895, "global_step": 136848, "epoch": 1537, "lr": 4.9493318880939946e-05}
{"train_loss": 0.08630762249231339, "global_step": 136849, "epoch": 1537, "lr": 4.949273917413807e-05}
{"train_loss": 0.10425007343292236, "global_step": 136850, "epoch": 1537, "lr": 4.9492159467404417e-05}
{"train_loss": 0.07813908904790878, "global_step": 136851, "epoch": 1537, "lr": 4.949157976073902e-05}
{"train_loss": 0.1249098926782608, "global_step": 136852, "epoch": 1537, "lr": 4.9491000054142e-05}
{"train_loss": 0.11215392500162125, "global_step": 136853, "epoch": 1537, "lr": 4.9490420347613375e-05}
{"train_loss": 0.11841607093811035, "global_step": 136854, "epoch": 1537, "lr": 4.948984064115328e-05}
{"train_loss": 0.08214499801397324, "global_step": 136855, "epoch": 1537, "lr": 4.948926093476175e-05}
{"train_loss": 0.0854424387216568, "global_step": 136856, "epoch": 1537, "lr": 4.948868122843891e-05}
{"train_loss": 0.11544344574213028, "global_step": 136857, "epoch": 1537, "lr": 4.948810152218478e-05}
{"train_loss": 0.1296474039554596, "global_step": 136858, "epoch": 1537, "lr": 4.94875218159995e-05}
{"train_loss": 0.13108699023723602, "global_step": 136859, "epoch": 1537, "lr": 4.948694210988308e-05}
{"train_loss": 0.11121649295091629, "global_step": 136860, "epoch": 1537, "lr": 4.9486362403835674e-05}
{"train_loss": 0.11895401030778885, "global_step": 136861, "epoch": 1537, "lr": 4.9485782697857296e-05}
{"train_loss": 0.1292266547679901, "global_step": 136862, "epoch": 1537, "lr": 4.948520299194806e-05}
{"train_loss": 0.137368842959404, "global_step": 136863, "epoch": 1537, "lr": 4.9484623286108024e-05}
{"train_loss": 0.15770111978054047, "global_step": 136864, "epoch": 1537, "lr": 4.948404358033726e-05}
{"train_loss": 0.1292542666196823, "global_step": 136865, "epoch": 1537, "lr": 4.948346387463589e-05}
{"train_loss": 0.17479299008846283, "global_step": 136866, "epoch": 1537, "lr": 4.9482884169003933e-05}
{"train_loss": 0.08552790433168411, "global_step": 136867, "epoch": 1537, "lr": 4.948230446344152e-05}
{"train_loss": 0.10258432477712631, "global_step": 136868, "epoch": 1537, "lr": 4.9481724757948686e-05}
{"train_loss": 0.14802439510822296, "global_step": 136869, "epoch": 1537, "lr": 4.948114505252555e-05}
{"train_loss": 0.13050521910190582, "global_step": 136870, "epoch": 1537, "lr": 4.948056534717214e-05}
{"train_loss": 0.17676854133605957, "global_step": 136871, "epoch": 1537, "lr": 4.947998564188859e-05}
{"train_loss": 0.08689422160387039, "global_step": 136872, "epoch": 1537, "lr": 4.9479405936674925e-05}
{"train_loss": 0.10717080533504486, "global_step": 136873, "epoch": 1537, "lr": 4.947882623153127e-05}
{"train_loss": 0.10120946913957596, "global_step": 136874, "epoch": 1537, "lr": 4.947824652645766e-05}
{"train_loss": 0.12225671112537384, "global_step": 136875, "epoch": 1537, "lr": 4.947766682145421e-05}
{"train_loss": 0.1931988000869751, "global_step": 136876, "epoch": 1537, "lr": 4.947708711652097e-05}
{"train_loss": 0.12310560792684555, "global_step": 136877, "epoch": 1537, "lr": 4.947650741165804e-05}
{"train_loss": 0.17381241917610168, "global_step": 136878, "epoch": 1537, "lr": 4.9475927706865466e-05}
{"train_loss": 0.10353301465511322, "global_step": 136879, "epoch": 1537, "lr": 4.9475348002143375e-05}
{"train_loss": 0.13803443312644958, "global_step": 136880, "epoch": 1537, "lr": 4.94747682974918e-05}
{"train_loss": 0.12015263991576902, "global_step": 136881, "epoch": 1537, "lr": 4.947418859291084e-05, "val_loss": 5.820777893066406}
{"train_loss": 0.09993766248226166, "global_step": 136882, "epoch": 1538, "lr": 4.947360888840057e-05}
{"train_loss": 0.09451255947351456, "global_step": 136883, "epoch": 1538, "lr": 4.947302918396105e-05}
{"train_loss": 0.1022610068321228, "global_step": 136884, "epoch": 1538, "lr": 4.9472449479592404e-05}
{"train_loss": 0.11850875616073608, "global_step": 136885, "epoch": 1538, "lr": 4.9471869775294656e-05}
{"train_loss": 0.13634133338928223, "global_step": 136886, "epoch": 1538, "lr": 4.947129007106793e-05}
{"train_loss": 0.13809140026569366, "global_step": 136887, "epoch": 1538, "lr": 4.947071036691226e-05}
{"train_loss": 0.14434024691581726, "global_step": 136888, "epoch": 1538, "lr": 4.9470130662827765e-05}
{"train_loss": 0.06783939898014069, "global_step": 136889, "epoch": 1538, "lr": 4.946955095881449e-05}
{"train_loss": 0.08095297962427139, "global_step": 136890, "epoch": 1538, "lr": 4.946897125487255e-05}
{"train_loss": 0.12031052261590958, "global_step": 136891, "epoch": 1538, "lr": 4.946839155100198e-05}
{"train_loss": 0.12720473110675812, "global_step": 136892, "epoch": 1538, "lr": 4.946781184720289e-05}
{"train_loss": 0.17395295202732086, "global_step": 136893, "epoch": 1538, "lr": 4.946723214347533e-05}
{"train_loss": 0.09159603714942932, "global_step": 136894, "epoch": 1538, "lr": 4.946665243981941e-05}
{"train_loss": 0.10754349827766418, "global_step": 136895, "epoch": 1538, "lr": 4.946607273623518e-05}
{"train_loss": 0.21422795951366425, "global_step": 136896, "epoch": 1538, "lr": 4.946549303272274e-05}
{"train_loss": 0.13588115572929382, "global_step": 136897, "epoch": 1538, "lr": 4.946491332928215e-05}
{"train_loss": 0.10333165526390076, "global_step": 136898, "epoch": 1538, "lr": 4.946433362591351e-05}
{"train_loss": 0.16005024313926697, "global_step": 136899, "epoch": 1538, "lr": 4.9463753922616865e-05}
{"train_loss": 0.13936659693717957, "global_step": 136900, "epoch": 1538, "lr": 4.9463174219392314e-05}
{"train_loss": 0.12129873037338257, "global_step": 136901, "epoch": 1538, "lr": 4.946259451623996e-05}
{"train_loss": 0.07801064848899841, "global_step": 136902, "epoch": 1538, "lr": 4.9462014813159815e-05}
{"train_loss": 0.14044977724552155, "global_step": 136903, "epoch": 1538, "lr": 4.946143511015203e-05}
{"train_loss": 0.17132245004177094, "global_step": 136904, "epoch": 1538, "lr": 4.9460855407216615e-05}
{"train_loss": 0.10167399793863297, "global_step": 136905, "epoch": 1538, "lr": 4.946027570435371e-05}
{"train_loss": 0.14806197583675385, "global_step": 136906, "epoch": 1538, "lr": 4.9459696001563344e-05}
{"train_loss": 0.16610732674598694, "global_step": 136907, "epoch": 1538, "lr": 4.945911629884564e-05}
{"train_loss": 0.06819207221269608, "global_step": 136908, "epoch": 1538, "lr": 4.945853659620062e-05}
{"train_loss": 0.13276582956314087, "global_step": 136909, "epoch": 1538, "lr": 4.945795689362842e-05}
{"train_loss": 0.12268954515457153, "global_step": 136910, "epoch": 1538, "lr": 4.945737719112907e-05}
{"train_loss": 0.10821876674890518, "global_step": 136911, "epoch": 1538, "lr": 4.945679748870269e-05}
{"train_loss": 0.09967684000730515, "global_step": 136912, "epoch": 1538, "lr": 4.9456217786349314e-05}
{"train_loss": 0.10667306184768677, "global_step": 136913, "epoch": 1538, "lr": 4.945563808406907e-05}
{"train_loss": 0.1702299863100052, "global_step": 136914, "epoch": 1538, "lr": 4.945505838186198e-05}
{"train_loss": 0.10463014245033264, "global_step": 136915, "epoch": 1538, "lr": 4.945447867972817e-05}
{"train_loss": 0.13061678409576416, "global_step": 136916, "epoch": 1538, "lr": 4.9453898977667687e-05}
{"train_loss": 0.054885365068912506, "global_step": 136917, "epoch": 1538, "lr": 4.945331927568063e-05}
{"train_loss": 0.11524313688278198, "global_step": 136918, "epoch": 1538, "lr": 4.945273957376706e-05}
{"train_loss": 0.08027321845293045, "global_step": 136919, "epoch": 1538, "lr": 4.9452159871927064e-05}
{"train_loss": 0.15274354815483093, "global_step": 136920, "epoch": 1538, "lr": 4.945158017016073e-05}
{"train_loss": 0.08319872617721558, "global_step": 136921, "epoch": 1538, "lr": 4.9451000468468104e-05}
{"train_loss": 0.12967924773693085, "global_step": 136922, "epoch": 1538, "lr": 4.945042076684931e-05}
{"train_loss": 0.10386308282613754, "global_step": 136923, "epoch": 1538, "lr": 4.944984106530438e-05}
{"train_loss": 0.09406790137290955, "global_step": 136924, "epoch": 1538, "lr": 4.944926136383342e-05}
{"train_loss": 0.10647907853126526, "global_step": 136925, "epoch": 1538, "lr": 4.944868166243649e-05}
{"train_loss": 0.09190794825553894, "global_step": 136926, "epoch": 1538, "lr": 4.94481019611137e-05}
{"train_loss": 0.09110566973686218, "global_step": 136927, "epoch": 1538, "lr": 4.944752225986508e-05}
{"train_loss": 0.0842607170343399, "global_step": 136928, "epoch": 1538, "lr": 4.944694255869076e-05}
{"train_loss": 0.08124155551195145, "global_step": 136929, "epoch": 1538, "lr": 4.944636285759076e-05}
{"train_loss": 0.0973062664270401, "global_step": 136930, "epoch": 1538, "lr": 4.9445783156565225e-05}
{"train_loss": 0.15516357123851776, "global_step": 136931, "epoch": 1538, "lr": 4.9445203455614164e-05}
{"train_loss": 0.1525551676750183, "global_step": 136932, "epoch": 1538, "lr": 4.944462375473772e-05}
{"train_loss": 0.08383093029260635, "global_step": 136933, "epoch": 1538, "lr": 4.944404405393591e-05}
{"train_loss": 0.15136781334877014, "global_step": 136934, "epoch": 1538, "lr": 4.9443464353208866e-05}
{"train_loss": 0.09391471743583679, "global_step": 136935, "epoch": 1538, "lr": 4.944288465255662e-05}
{"train_loss": 0.1978129744529724, "global_step": 136936, "epoch": 1538, "lr": 4.944230495197928e-05}
{"train_loss": 0.1696518510580063, "global_step": 136937, "epoch": 1538, "lr": 4.944172525147692e-05}
{"train_loss": 0.08555461466312408, "global_step": 136938, "epoch": 1538, "lr": 4.944114555104961e-05}
{"train_loss": 0.13280968368053436, "global_step": 136939, "epoch": 1538, "lr": 4.944056585069745e-05}
{"train_loss": 0.10224311798810959, "global_step": 136940, "epoch": 1538, "lr": 4.943998615042047e-05}
{"train_loss": 0.08255864679813385, "global_step": 136941, "epoch": 1538, "lr": 4.94394064502188e-05}
{"train_loss": 0.1538122147321701, "global_step": 136942, "epoch": 1538, "lr": 4.943882675009247e-05}
{"train_loss": 0.10222993046045303, "global_step": 136943, "epoch": 1538, "lr": 4.943824705004161e-05}
{"train_loss": 0.14369794726371765, "global_step": 136944, "epoch": 1538, "lr": 4.943766735006624e-05}
{"train_loss": 0.11695665121078491, "global_step": 136945, "epoch": 1538, "lr": 4.9437087650166504e-05}
{"train_loss": 0.16863715648651123, "global_step": 136946, "epoch": 1538, "lr": 4.943650795034242e-05}
{"train_loss": 0.1012614369392395, "global_step": 136947, "epoch": 1538, "lr": 4.9435928250594096e-05}
{"train_loss": 0.09505102783441544, "global_step": 136948, "epoch": 1538, "lr": 4.94353485509216e-05}
{"train_loss": 0.10657563805580139, "global_step": 136949, "epoch": 1538, "lr": 4.943476885132503e-05}
{"train_loss": 0.17733338475227356, "global_step": 136950, "epoch": 1538, "lr": 4.943418915180443e-05}
{"train_loss": 0.12360227108001709, "global_step": 136951, "epoch": 1538, "lr": 4.9433609452359916e-05}
{"train_loss": 0.1424497812986374, "global_step": 136952, "epoch": 1538, "lr": 4.943302975299152e-05}
{"train_loss": 0.17840507626533508, "global_step": 136953, "epoch": 1538, "lr": 4.9432450053699376e-05}
{"train_loss": 0.1039019376039505, "global_step": 136954, "epoch": 1538, "lr": 4.9431870354483513e-05}
{"train_loss": 0.10283136367797852, "global_step": 136955, "epoch": 1538, "lr": 4.943129065534403e-05}
{"train_loss": 0.11152597516775131, "global_step": 136956, "epoch": 1538, "lr": 4.943071095628101e-05}
{"train_loss": 0.11715685576200485, "global_step": 136957, "epoch": 1538, "lr": 4.9430131257294516e-05}
{"train_loss": 0.12206500768661499, "global_step": 136958, "epoch": 1538, "lr": 4.942955155838465e-05}
{"train_loss": 0.15223059058189392, "global_step": 136959, "epoch": 1538, "lr": 4.942897185955145e-05}
{"train_loss": 0.13709548115730286, "global_step": 136960, "epoch": 1538, "lr": 4.9428392160795037e-05}
{"train_loss": 0.09128978848457336, "global_step": 136961, "epoch": 1538, "lr": 4.9427812462115454e-05}
{"train_loss": 0.19735246896743774, "global_step": 136962, "epoch": 1538, "lr": 4.9427232763512816e-05}
{"train_loss": 0.1015247255563736, "global_step": 136963, "epoch": 1538, "lr": 4.9426653064987157e-05}
{"train_loss": 0.12860007584095, "global_step": 136964, "epoch": 1538, "lr": 4.94260733665386e-05}
{"train_loss": 0.15680596232414246, "global_step": 136965, "epoch": 1538, "lr": 4.942549366816717e-05}
{"train_loss": 0.13771618902683258, "global_step": 136966, "epoch": 1538, "lr": 4.9424913969873e-05}
{"train_loss": 0.15386024117469788, "global_step": 136967, "epoch": 1538, "lr": 4.942433427165613e-05}
{"train_loss": 0.10637878626585007, "global_step": 136968, "epoch": 1538, "lr": 4.9423754573516664e-05}
{"train_loss": 0.13794784247875214, "global_step": 136969, "epoch": 1538, "lr": 4.942317487545465e-05}
{"train_loss": 0.12252547098009774, "global_step": 136970, "epoch": 1538, "lr": 4.942259517747021e-05, "val_loss": 5.75746488571167}
{"train_loss": 0.1720319539308548, "global_step": 136971, "epoch": 1539, "lr": 4.9422015479563366e-05}
{"train_loss": 0.06669057905673981, "global_step": 136972, "epoch": 1539, "lr": 4.942143578173424e-05}
{"train_loss": 0.15133558213710785, "global_step": 136973, "epoch": 1539, "lr": 4.942085608398289e-05}
{"train_loss": 0.08001585304737091, "global_step": 136974, "epoch": 1539, "lr": 4.9420276386309406e-05}
{"train_loss": 0.1282625049352646, "global_step": 136975, "epoch": 1539, "lr": 4.941969668871385e-05}
{"train_loss": 0.09570938348770142, "global_step": 136976, "epoch": 1539, "lr": 4.9419116991196315e-05}
{"train_loss": 0.11388392746448517, "global_step": 136977, "epoch": 1539, "lr": 4.941853729375688e-05}
{"train_loss": 0.08869919925928116, "global_step": 136978, "epoch": 1539, "lr": 4.9417957596395596e-05}
{"train_loss": 0.13609784841537476, "global_step": 136979, "epoch": 1539, "lr": 4.941737789911258e-05}
{"train_loss": 0.09789734333753586, "global_step": 136980, "epoch": 1539, "lr": 4.941679820190787e-05}
{"train_loss": 0.08840443938970566, "global_step": 136981, "epoch": 1539, "lr": 4.941621850478159e-05}
{"train_loss": 0.11344119161367416, "global_step": 136982, "epoch": 1539, "lr": 4.941563880773377e-05}
{"train_loss": 0.12010091543197632, "global_step": 136983, "epoch": 1539, "lr": 4.941505911076453e-05}
{"train_loss": 0.12826628983020782, "global_step": 136984, "epoch": 1539, "lr": 4.94144794138739e-05}
{"train_loss": 0.13352783024311066, "global_step": 136985, "epoch": 1539, "lr": 4.941389971706201e-05}
{"train_loss": 0.08008962869644165, "global_step": 136986, "epoch": 1539, "lr": 4.9413320020328904e-05}
{"train_loss": 0.07227862626314163, "global_step": 136987, "epoch": 1539, "lr": 4.941274032367468e-05}
{"train_loss": 0.12050080299377441, "global_step": 136988, "epoch": 1539, "lr": 4.9412160627099384e-05}
{"train_loss": 0.12677083909511566, "global_step": 136989, "epoch": 1539, "lr": 4.941158093060315e-05}
{"train_loss": 0.09571778029203415, "global_step": 136990, "epoch": 1539, "lr": 4.941100123418598e-05}
{"train_loss": 0.07226433604955673, "global_step": 136991, "epoch": 1539, "lr": 4.941042153784802e-05}
{"train_loss": 0.15649449825286865, "global_step": 136992, "epoch": 1539, "lr": 4.940984184158932e-05}
{"train_loss": 0.10304349660873413, "global_step": 136993, "epoch": 1539, "lr": 4.9409262145409954e-05}
{"train_loss": 0.08382571488618851, "global_step": 136994, "epoch": 1539, "lr": 4.940868244931001e-05}
{"train_loss": 0.09953352808952332, "global_step": 136995, "epoch": 1539, "lr": 4.940810275328956e-05}
{"train_loss": 0.10576830059289932, "global_step": 136996, "epoch": 1539, "lr": 4.940752305734869e-05}
{"train_loss": 0.07919074594974518, "global_step": 136997, "epoch": 1539, "lr": 4.940694336148746e-05}
{"train_loss": 0.16624124348163605, "global_step": 136998, "epoch": 1539, "lr": 4.940636366570597e-05}
{"train_loss": 0.13184261322021484, "global_step": 136999, "epoch": 1539, "lr": 4.940578397000427e-05}
{"train_loss": 0.08463313430547714, "global_step": 137000, "epoch": 1539, "lr": 4.9405204274382476e-05}
{"train_loss": 0.11549931764602661, "global_step": 137001, "epoch": 1539, "lr": 4.940462457884062e-05}
{"train_loss": 0.06527871638536453, "global_step": 137002, "epoch": 1539, "lr": 4.9404044883378837e-05}
{"train_loss": 0.11282302439212799, "global_step": 137003, "epoch": 1539, "lr": 4.9403465187997146e-05}
{"train_loss": 0.09032199531793594, "global_step": 137004, "epoch": 1539, "lr": 4.9402885492695674e-05}
{"train_loss": 0.13326014578342438, "global_step": 137005, "epoch": 1539, "lr": 4.940230579747445e-05}
{"train_loss": 0.10475528240203857, "global_step": 137006, "epoch": 1539, "lr": 4.9401726102333604e-05}
{"train_loss": 0.12741568684577942, "global_step": 137007, "epoch": 1539, "lr": 4.940114640727317e-05}
{"train_loss": 0.1307155191898346, "global_step": 137008, "epoch": 1539, "lr": 4.9400566712293264e-05}
{"train_loss": 0.10972683876752853, "global_step": 137009, "epoch": 1539, "lr": 4.9399987017393916e-05}
{"train_loss": 0.17040656507015228, "global_step": 137010, "epoch": 1539, "lr": 4.939940732257525e-05}
{"train_loss": 0.1415501981973648, "global_step": 137011, "epoch": 1539, "lr": 4.939882762783733e-05}
{"train_loss": 0.07825972884893417, "global_step": 137012, "epoch": 1539, "lr": 4.939824793318022e-05}
{"train_loss": 0.15209881961345673, "global_step": 137013, "epoch": 1539, "lr": 4.939766823860402e-05}
{"train_loss": 0.09871617704629898, "global_step": 137014, "epoch": 1539, "lr": 4.939708854410879e-05}
{"train_loss": 0.1411711722612381, "global_step": 137015, "epoch": 1539, "lr": 4.939650884969461e-05}
{"train_loss": 0.13409589231014252, "global_step": 137016, "epoch": 1539, "lr": 4.939592915536157e-05}
{"train_loss": 0.07363543659448624, "global_step": 137017, "epoch": 1539, "lr": 4.9395349461109746e-05}
{"train_loss": 0.11660510301589966, "global_step": 137018, "epoch": 1539, "lr": 4.939476976693919e-05}
{"train_loss": 0.14496776461601257, "global_step": 137019, "epoch": 1539, "lr": 4.939419007285002e-05}
{"train_loss": 0.050037868320941925, "global_step": 137020, "epoch": 1539, "lr": 4.939361037884228e-05}
{"train_loss": 0.07763006538152695, "global_step": 137021, "epoch": 1539, "lr": 4.939303068491607e-05}
{"train_loss": 0.11176148056983948, "global_step": 137022, "epoch": 1539, "lr": 4.9392450991071444e-05}
{"train_loss": 0.13295002281665802, "global_step": 137023, "epoch": 1539, "lr": 4.9391871297308515e-05}
{"train_loss": 0.11186975985765457, "global_step": 137024, "epoch": 1539, "lr": 4.939129160362733e-05}
{"train_loss": 0.13157737255096436, "global_step": 137025, "epoch": 1539, "lr": 4.9390711910027985e-05}
{"train_loss": 0.09107677638530731, "global_step": 137026, "epoch": 1539, "lr": 4.939013221651054e-05}
{"train_loss": 0.061393264681100845, "global_step": 137027, "epoch": 1539, "lr": 4.938955252307508e-05}
{"train_loss": 0.12419092655181885, "global_step": 137028, "epoch": 1539, "lr": 4.9388972829721716e-05}
{"train_loss": 0.09749463945627213, "global_step": 137029, "epoch": 1539, "lr": 4.9388393136450466e-05}
{"train_loss": 0.1054232195019722, "global_step": 137030, "epoch": 1539, "lr": 4.9387813443261466e-05}
{"train_loss": 0.12945617735385895, "global_step": 137031, "epoch": 1539, "lr": 4.938723375015476e-05}
{"train_loss": 0.11072611808776855, "global_step": 137032, "epoch": 1539, "lr": 4.938665405713042e-05}
{"train_loss": 0.10890571027994156, "global_step": 137033, "epoch": 1539, "lr": 4.9386074364188545e-05}
{"train_loss": 0.120278000831604, "global_step": 137034, "epoch": 1539, "lr": 4.938549467132921e-05}
{"train_loss": 0.08002275228500366, "global_step": 137035, "epoch": 1539, "lr": 4.938491497855248e-05}
{"train_loss": 0.07020581513643265, "global_step": 137036, "epoch": 1539, "lr": 4.938433528585845e-05}
{"train_loss": 0.09902931749820709, "global_step": 137037, "epoch": 1539, "lr": 4.9383755593247175e-05}
{"train_loss": 0.13302601873874664, "global_step": 137038, "epoch": 1539, "lr": 4.9383175900718766e-05}
{"train_loss": 0.10672324895858765, "global_step": 137039, "epoch": 1539, "lr": 4.938259620827325e-05}
{"train_loss": 0.1596195101737976, "global_step": 137040, "epoch": 1539, "lr": 4.9382016515910766e-05}
{"train_loss": 0.12115217745304108, "global_step": 137041, "epoch": 1539, "lr": 4.9381436823631344e-05}
{"train_loss": 0.19830667972564697, "global_step": 137042, "epoch": 1539, "lr": 4.93808571314351e-05}
{"train_loss": 0.11033740639686584, "global_step": 137043, "epoch": 1539, "lr": 4.938027743932207e-05}
{"train_loss": 0.09566674381494522, "global_step": 137044, "epoch": 1539, "lr": 4.9379697747292374e-05}
{"train_loss": 0.2058688998222351, "global_step": 137045, "epoch": 1539, "lr": 4.9379118055346046e-05}
{"train_loss": 0.090181365609169, "global_step": 137046, "epoch": 1539, "lr": 4.9378538363483195e-05}
{"train_loss": 0.07140932232141495, "global_step": 137047, "epoch": 1539, "lr": 4.93779586717039e-05}
{"train_loss": 0.12727980315685272, "global_step": 137048, "epoch": 1539, "lr": 4.937737898000823e-05}
{"train_loss": 0.18434882164001465, "global_step": 137049, "epoch": 1539, "lr": 4.937679928839627e-05}
{"train_loss": 0.09323287755250931, "global_step": 137050, "epoch": 1539, "lr": 4.937621959686809e-05}
{"train_loss": 0.13888999819755554, "global_step": 137051, "epoch": 1539, "lr": 4.937563990542376e-05}
{"train_loss": 0.08401526510715485, "global_step": 137052, "epoch": 1539, "lr": 4.937506021406337e-05}
{"train_loss": 0.12503911554813385, "global_step": 137053, "epoch": 1539, "lr": 4.9374480522787005e-05}
{"train_loss": 0.0987849161028862, "global_step": 137054, "epoch": 1539, "lr": 4.9373900831594727e-05}
{"train_loss": 0.1419188678264618, "global_step": 137055, "epoch": 1539, "lr": 4.937332114048663e-05}
{"train_loss": 0.11143551766872406, "global_step": 137056, "epoch": 1539, "lr": 4.937274144946276e-05}
{"train_loss": 0.14380811154842377, "global_step": 137057, "epoch": 1539, "lr": 4.937216175852324e-05}
{"train_loss": 0.06340701878070831, "global_step": 137058, "epoch": 1539, "lr": 4.9371582067668104e-05}
{"train_loss": 0.11370541452524367, "global_step": 137059, "epoch": 1539, "lr": 4.937100237689747e-05, "val_loss": 5.8841872215271}
{"train_loss": 0.10550297796726227, "global_step": 137060, "epoch": 1540, "lr": 4.937042268621138e-05}
{"train_loss": 0.15200503170490265, "global_step": 137061, "epoch": 1540, "lr": 4.936984299560995e-05}
{"train_loss": 0.09449002146720886, "global_step": 137062, "epoch": 1540, "lr": 4.936926330509322e-05}
{"train_loss": 0.11804014444351196, "global_step": 137063, "epoch": 1540, "lr": 4.936868361466128e-05}
{"train_loss": 0.08062997460365295, "global_step": 137064, "epoch": 1540, "lr": 4.936810392431423e-05}
{"train_loss": 0.1575617790222168, "global_step": 137065, "epoch": 1540, "lr": 4.936752423405211e-05}
{"train_loss": 0.09057820588350296, "global_step": 137066, "epoch": 1540, "lr": 4.936694454387504e-05}
{"train_loss": 0.14881515502929688, "global_step": 137067, "epoch": 1540, "lr": 4.936636485378306e-05}
{"train_loss": 0.11300740391016006, "global_step": 137068, "epoch": 1540, "lr": 4.936578516377629e-05}
{"train_loss": 0.11416460573673248, "global_step": 137069, "epoch": 1540, "lr": 4.936520547385477e-05}
{"train_loss": 0.09364532679319382, "global_step": 137070, "epoch": 1540, "lr": 4.9364625784018575e-05}
{"train_loss": 0.10475282371044159, "global_step": 137071, "epoch": 1540, "lr": 4.936404609426781e-05}
{"train_loss": 0.07051768153905869, "global_step": 137072, "epoch": 1540, "lr": 4.936346640460254e-05}
{"train_loss": 0.135600283741951, "global_step": 137073, "epoch": 1540, "lr": 4.936288671502285e-05}
{"train_loss": 0.12740401923656464, "global_step": 137074, "epoch": 1540, "lr": 4.936230702552881e-05}
{"train_loss": 0.1462874859571457, "global_step": 137075, "epoch": 1540, "lr": 4.936172733612049e-05}
{"train_loss": 0.07078303396701813, "global_step": 137076, "epoch": 1540, "lr": 4.9361147646798e-05}
{"train_loss": 0.12030595541000366, "global_step": 137077, "epoch": 1540, "lr": 4.936056795756137e-05}
{"train_loss": 0.14764276146888733, "global_step": 137078, "epoch": 1540, "lr": 4.935998826841072e-05}
{"train_loss": 0.14434273540973663, "global_step": 137079, "epoch": 1540, "lr": 4.93594085793461e-05}
{"train_loss": 0.108856700360775, "global_step": 137080, "epoch": 1540, "lr": 4.9358828890367616e-05}
{"train_loss": 0.23499245941638947, "global_step": 137081, "epoch": 1540, "lr": 4.935824920147531e-05}
{"train_loss": 0.11553584784269333, "global_step": 137082, "epoch": 1540, "lr": 4.935766951266927e-05}
{"train_loss": 0.16683904826641083, "global_step": 137083, "epoch": 1540, "lr": 4.935708982394962e-05}
{"train_loss": 0.11124479025602341, "global_step": 137084, "epoch": 1540, "lr": 4.9356510135316374e-05}
{"train_loss": 0.09477096796035767, "global_step": 137085, "epoch": 1540, "lr": 4.935593044676966e-05}
{"train_loss": 0.12003908306360245, "global_step": 137086, "epoch": 1540, "lr": 4.935535075830951e-05}
{"train_loss": 0.08269285410642624, "global_step": 137087, "epoch": 1540, "lr": 4.9354771069936046e-05}
{"train_loss": 0.13155491650104523, "global_step": 137088, "epoch": 1540, "lr": 4.935419138164932e-05}
{"train_loss": 0.14002098143100739, "global_step": 137089, "epoch": 1540, "lr": 4.9353611693449406e-05}
{"train_loss": 0.09213463962078094, "global_step": 137090, "epoch": 1540, "lr": 4.9353032005336396e-05}
{"train_loss": 0.11158574372529984, "global_step": 137091, "epoch": 1540, "lr": 4.935245231731036e-05}
{"train_loss": 0.0967467650771141, "global_step": 137092, "epoch": 1540, "lr": 4.935187262937138e-05}
{"train_loss": 0.0997556522488594, "global_step": 137093, "epoch": 1540, "lr": 4.935129294151955e-05}
{"train_loss": 0.07611433416604996, "global_step": 137094, "epoch": 1540, "lr": 4.935071325375491e-05}
{"train_loss": 0.1263231337070465, "global_step": 137095, "epoch": 1540, "lr": 4.9350133566077564e-05}
{"train_loss": 0.1121172234416008, "global_step": 137096, "epoch": 1540, "lr": 4.9349553878487574e-05}
{"train_loss": 0.14953765273094177, "global_step": 137097, "epoch": 1540, "lr": 4.934897419098505e-05}
{"train_loss": 0.12341612577438354, "global_step": 137098, "epoch": 1540, "lr": 4.9348394503570025e-05}
{"train_loss": 0.09463238716125488, "global_step": 137099, "epoch": 1540, "lr": 4.93478148162426e-05}
{"train_loss": 0.08456992357969284, "global_step": 137100, "epoch": 1540, "lr": 4.9347235129002884e-05}
{"train_loss": 0.08883176743984222, "global_step": 137101, "epoch": 1540, "lr": 4.934665544185089e-05}
{"train_loss": 0.12615588307380676, "global_step": 137102, "epoch": 1540, "lr": 4.9346075754786767e-05}
{"train_loss": 0.10309255123138428, "global_step": 137103, "epoch": 1540, "lr": 4.934549606781052e-05}
{"train_loss": 0.07956051081418991, "global_step": 137104, "epoch": 1540, "lr": 4.934491638092229e-05}
{"train_loss": 0.09117034077644348, "global_step": 137105, "epoch": 1540, "lr": 4.934433669412211e-05}
{"train_loss": 0.11573893576860428, "global_step": 137106, "epoch": 1540, "lr": 4.934375700741009e-05}
{"train_loss": 0.1584538221359253, "global_step": 137107, "epoch": 1540, "lr": 4.934317732078629e-05}
{"train_loss": 0.07336372137069702, "global_step": 137108, "epoch": 1540, "lr": 4.9342597634250786e-05}
{"train_loss": 0.12068117409944534, "global_step": 137109, "epoch": 1540, "lr": 4.934201794780367e-05}
{"train_loss": 0.11717455834150314, "global_step": 137110, "epoch": 1540, "lr": 4.9341438261445e-05}
{"train_loss": 0.08324114233255386, "global_step": 137111, "epoch": 1540, "lr": 4.934085857517487e-05}
{"train_loss": 0.11913201957941055, "global_step": 137112, "epoch": 1540, "lr": 4.9340278888993366e-05}
{"train_loss": 0.08311644196510315, "global_step": 137113, "epoch": 1540, "lr": 4.933969920290053e-05}
{"train_loss": 0.10150039196014404, "global_step": 137114, "epoch": 1540, "lr": 4.9339119516896484e-05}
{"train_loss": 0.10018525272607803, "global_step": 137115, "epoch": 1540, "lr": 4.933853983098127e-05}
{"train_loss": 0.18237148225307465, "global_step": 137116, "epoch": 1540, "lr": 4.9337960145155e-05}
{"train_loss": 0.07773860543966293, "global_step": 137117, "epoch": 1540, "lr": 4.9337380459417706e-05}
{"train_loss": 0.1458228975534439, "global_step": 137118, "epoch": 1540, "lr": 4.93368007737695e-05}
{"train_loss": 0.09928740561008453, "global_step": 137119, "epoch": 1540, "lr": 4.933622108821048e-05}
{"train_loss": 0.07758523523807526, "global_step": 137120, "epoch": 1540, "lr": 4.933564140274066e-05}
{"train_loss": 0.09650491178035736, "global_step": 137121, "epoch": 1540, "lr": 4.933506171736019e-05}
{"train_loss": 0.08795883506536484, "global_step": 137122, "epoch": 1540, "lr": 4.9334482032069085e-05}
{"train_loss": 0.12784811854362488, "global_step": 137123, "epoch": 1540, "lr": 4.933390234686747e-05}
{"train_loss": 0.09444977343082428, "global_step": 137124, "epoch": 1540, "lr": 4.933332266175539e-05}
{"train_loss": 0.09538164734840393, "global_step": 137125, "epoch": 1540, "lr": 4.933274297673295e-05}
{"train_loss": 0.08365703374147415, "global_step": 137126, "epoch": 1540, "lr": 4.9332163291800195e-05}
{"train_loss": 0.09123532474040985, "global_step": 137127, "epoch": 1540, "lr": 4.933158360695725e-05}
{"train_loss": 0.14032231271266937, "global_step": 137128, "epoch": 1540, "lr": 4.933100392220415e-05}
{"train_loss": 0.1633186638355255, "global_step": 137129, "epoch": 1540, "lr": 4.933042423754098e-05}
{"train_loss": 0.1697411835193634, "global_step": 137130, "epoch": 1540, "lr": 4.932984455296784e-05}
{"train_loss": 0.1534986048936844, "global_step": 137131, "epoch": 1540, "lr": 4.9329264868484785e-05}
{"train_loss": 0.1762239933013916, "global_step": 137132, "epoch": 1540, "lr": 4.932868518409191e-05}
{"train_loss": 0.1527494490146637, "global_step": 137133, "epoch": 1540, "lr": 4.932810549978929e-05}
{"train_loss": 0.14857040345668793, "global_step": 137134, "epoch": 1540, "lr": 4.932752581557698e-05}
{"train_loss": 0.09558847546577454, "global_step": 137135, "epoch": 1540, "lr": 4.93269461314551e-05}
{"train_loss": 0.09467823803424835, "global_step": 137136, "epoch": 1540, "lr": 4.9326366447423677e-05}
{"train_loss": 0.08610011637210846, "global_step": 137137, "epoch": 1540, "lr": 4.932578676348282e-05}
{"train_loss": 0.1242290809750557, "global_step": 137138, "epoch": 1540, "lr": 4.9325207079632616e-05}
{"train_loss": 0.11851036548614502, "global_step": 137139, "epoch": 1540, "lr": 4.932462739587311e-05}
{"train_loss": 0.11827560514211655, "global_step": 137140, "epoch": 1540, "lr": 4.932404771220443e-05}
{"train_loss": 0.17161937057971954, "global_step": 137141, "epoch": 1540, "lr": 4.93234680286266e-05}
{"train_loss": 0.1327739655971527, "global_step": 137142, "epoch": 1540, "lr": 4.932288834513973e-05}
{"train_loss": 0.09728045761585236, "global_step": 137143, "epoch": 1540, "lr": 4.932230866174388e-05}
{"train_loss": 0.13766196370124817, "global_step": 137144, "epoch": 1540, "lr": 4.932172897843916e-05}
{"train_loss": 0.14063206315040588, "global_step": 137145, "epoch": 1540, "lr": 4.93211492952256e-05}
{"train_loss": 0.09561881422996521, "global_step": 137146, "epoch": 1540, "lr": 4.932056961210333e-05}
{"train_loss": 0.0803895890712738, "global_step": 137147, "epoch": 1540, "lr": 4.931998992907238e-05}
{"train_loss": 0.11624476464276903, "global_step": 137148, "epoch": 1540, "lr": 4.9319410246132855e-05, "val_loss": 5.789829254150391, "train_action_mse_error": 35.48281478881836}
{"train_loss": 0.20357728004455566, "global_step": 137149, "epoch": 1541, "lr": 4.9318830563284826e-05}
{"train_loss": 0.06746693700551987, "global_step": 137150, "epoch": 1541, "lr": 4.931825088052837e-05}
{"train_loss": 0.19287315011024475, "global_step": 137151, "epoch": 1541, "lr": 4.9317671197863574e-05}
{"train_loss": 0.17187318205833435, "global_step": 137152, "epoch": 1541, "lr": 4.931709151529051e-05}
{"train_loss": 0.14413847029209137, "global_step": 137153, "epoch": 1541, "lr": 4.931651183280924e-05}
{"train_loss": 0.11261904984712601, "global_step": 137154, "epoch": 1541, "lr": 4.931593215041986e-05}
{"train_loss": 0.08258748054504395, "global_step": 137155, "epoch": 1541, "lr": 4.9315352468122456e-05}
{"train_loss": 0.14711850881576538, "global_step": 137156, "epoch": 1541, "lr": 4.931477278591708e-05}
{"train_loss": 0.14660291373729706, "global_step": 137157, "epoch": 1541, "lr": 4.9314193103803844e-05}
{"train_loss": 0.13959792256355286, "global_step": 137158, "epoch": 1541, "lr": 4.9313613421782784e-05}
{"train_loss": 0.1313733011484146, "global_step": 137159, "epoch": 1541, "lr": 4.9313033739854024e-05}
{"train_loss": 0.16078412532806396, "global_step": 137160, "epoch": 1541, "lr": 4.9312454058017596e-05}
{"train_loss": 0.11445339024066925, "global_step": 137161, "epoch": 1541, "lr": 4.931187437627362e-05}
{"train_loss": 0.16796648502349854, "global_step": 137162, "epoch": 1541, "lr": 4.9311294694622135e-05}
{"train_loss": 0.19039055705070496, "global_step": 137163, "epoch": 1541, "lr": 4.931071501306326e-05}
{"train_loss": 0.09467397630214691, "global_step": 137164, "epoch": 1541, "lr": 4.9310135331597024e-05}
{"train_loss": 0.06430884450674057, "global_step": 137165, "epoch": 1541, "lr": 4.930955565022356e-05}
{"train_loss": 0.1416970044374466, "global_step": 137166, "epoch": 1541, "lr": 4.93089759689429e-05}
{"train_loss": 0.1394048035144806, "global_step": 137167, "epoch": 1541, "lr": 4.9308396287755144e-05}
{"train_loss": 0.2055196613073349, "global_step": 137168, "epoch": 1541, "lr": 4.9307816606660365e-05}
{"train_loss": 0.13118116557598114, "global_step": 137169, "epoch": 1541, "lr": 4.930723692565864e-05}
{"train_loss": 0.12189740687608719, "global_step": 137170, "epoch": 1541, "lr": 4.930665724475006e-05}
{"train_loss": 0.13470458984375, "global_step": 137171, "epoch": 1541, "lr": 4.9306077563934684e-05}
{"train_loss": 0.16233785450458527, "global_step": 137172, "epoch": 1541, "lr": 4.930549788321259e-05}
{"train_loss": 0.12076670676469803, "global_step": 137173, "epoch": 1541, "lr": 4.930491820258385e-05}
{"train_loss": 0.06220826506614685, "global_step": 137174, "epoch": 1541, "lr": 4.930433852204859e-05}
{"train_loss": 0.0703970193862915, "global_step": 137175, "epoch": 1541, "lr": 4.9303758841606826e-05}
{"train_loss": 0.09240635484457016, "global_step": 137176, "epoch": 1541, "lr": 4.930317916125868e-05}
{"train_loss": 0.09269321709871292, "global_step": 137177, "epoch": 1541, "lr": 4.93025994810042e-05}
{"train_loss": 0.07587484270334244, "global_step": 137178, "epoch": 1541, "lr": 4.93020198008435e-05}
{"train_loss": 0.10073297470808029, "global_step": 137179, "epoch": 1541, "lr": 4.93014401207766e-05}
{"train_loss": 0.10978009551763535, "global_step": 137180, "epoch": 1541, "lr": 4.930086044080365e-05}
{"train_loss": 0.10438372194766998, "global_step": 137181, "epoch": 1541, "lr": 4.930028076092466e-05}
{"train_loss": 0.1576467901468277, "global_step": 137182, "epoch": 1541, "lr": 4.929970108113977e-05}
{"train_loss": 0.12703192234039307, "global_step": 137183, "epoch": 1541, "lr": 4.929912140144899e-05}
{"train_loss": 0.10996189713478088, "global_step": 137184, "epoch": 1541, "lr": 4.929854172185246e-05}
{"train_loss": 0.0867294892668724, "global_step": 137185, "epoch": 1541, "lr": 4.9297962042350235e-05}
{"train_loss": 0.038063473999500275, "global_step": 137186, "epoch": 1541, "lr": 4.929738236294238e-05}
{"train_loss": 0.10831671953201294, "global_step": 137187, "epoch": 1541, "lr": 4.929680268362898e-05}
{"train_loss": 0.11819969117641449, "global_step": 137188, "epoch": 1541, "lr": 4.929622300441012e-05}
{"train_loss": 0.09337395429611206, "global_step": 137189, "epoch": 1541, "lr": 4.929564332528588e-05}
{"train_loss": 0.10357799381017685, "global_step": 137190, "epoch": 1541, "lr": 4.929506364625631e-05}
{"train_loss": 0.12425478547811508, "global_step": 137191, "epoch": 1541, "lr": 4.929448396732154e-05}
{"train_loss": 0.07224401086568832, "global_step": 137192, "epoch": 1541, "lr": 4.92939042884816e-05}
{"train_loss": 0.09119464457035065, "global_step": 137193, "epoch": 1541, "lr": 4.9293324609736596e-05}
{"train_loss": 0.12589535117149353, "global_step": 137194, "epoch": 1541, "lr": 4.9292744931086585e-05}
{"train_loss": 0.10544534027576447, "global_step": 137195, "epoch": 1541, "lr": 4.929216525253167e-05}
{"train_loss": 0.13728883862495422, "global_step": 137196, "epoch": 1541, "lr": 4.92915855740719e-05}
{"train_loss": 0.11166897416114807, "global_step": 137197, "epoch": 1541, "lr": 4.9291005895707384e-05}
{"train_loss": 0.1577242612838745, "global_step": 137198, "epoch": 1541, "lr": 4.929042621743816e-05}
{"train_loss": 0.14888252317905426, "global_step": 137199, "epoch": 1541, "lr": 4.9289846539264355e-05}
{"train_loss": 0.10845472663640976, "global_step": 137200, "epoch": 1541, "lr": 4.9289266861185997e-05}
{"train_loss": 0.07399095594882965, "global_step": 137201, "epoch": 1541, "lr": 4.928868718320322e-05}
{"train_loss": 0.11853405833244324, "global_step": 137202, "epoch": 1541, "lr": 4.928810750531604e-05}
{"train_loss": 0.12153343111276627, "global_step": 137203, "epoch": 1541, "lr": 4.9287527827524585e-05}
{"train_loss": 0.18226732313632965, "global_step": 137204, "epoch": 1541, "lr": 4.92869481498289e-05}
{"train_loss": 0.19093844294548035, "global_step": 137205, "epoch": 1541, "lr": 4.928636847222908e-05}
{"train_loss": 0.06933391094207764, "global_step": 137206, "epoch": 1541, "lr": 4.92857887947252e-05}
{"train_loss": 0.1540134698152542, "global_step": 137207, "epoch": 1541, "lr": 4.928520911731734e-05}
{"train_loss": 0.10141336172819138, "global_step": 137208, "epoch": 1541, "lr": 4.9284629440005574e-05}
{"train_loss": 0.13787296414375305, "global_step": 137209, "epoch": 1541, "lr": 4.928404976278997e-05}
{"train_loss": 0.08981103450059891, "global_step": 137210, "epoch": 1541, "lr": 4.9283470085670635e-05}
{"train_loss": 0.09179610013961792, "global_step": 137211, "epoch": 1541, "lr": 4.928289040864761e-05}
{"train_loss": 0.1252308040857315, "global_step": 137212, "epoch": 1541, "lr": 4.928231073172101e-05}
{"train_loss": 0.1169159859418869, "global_step": 137213, "epoch": 1541, "lr": 4.928173105489088e-05}
{"train_loss": 0.12620241940021515, "global_step": 137214, "epoch": 1541, "lr": 4.928115137815733e-05}
{"train_loss": 0.18383461236953735, "global_step": 137215, "epoch": 1541, "lr": 4.9280571701520396e-05}
{"train_loss": 0.1172322928905487, "global_step": 137216, "epoch": 1541, "lr": 4.92799920249802e-05}
{"train_loss": 0.13018937408924103, "global_step": 137217, "epoch": 1541, "lr": 4.9279412348536786e-05}
{"train_loss": 0.11494937539100647, "global_step": 137218, "epoch": 1541, "lr": 4.9278832672190264e-05}
{"train_loss": 0.11522456258535385, "global_step": 137219, "epoch": 1541, "lr": 4.927825299594067e-05}
{"train_loss": 0.12222813069820404, "global_step": 137220, "epoch": 1541, "lr": 4.927767331978813e-05}
{"train_loss": 0.11019373685121536, "global_step": 137221, "epoch": 1541, "lr": 4.927709364373268e-05}
{"train_loss": 0.10313424468040466, "global_step": 137222, "epoch": 1541, "lr": 4.9276513967774435e-05}
{"train_loss": 0.10089937597513199, "global_step": 137223, "epoch": 1541, "lr": 4.927593429191344e-05}
{"train_loss": 0.05374005436897278, "global_step": 137224, "epoch": 1541, "lr": 4.927535461614979e-05}
{"train_loss": 0.1205969899892807, "global_step": 137225, "epoch": 1541, "lr": 4.927477494048356e-05}
{"train_loss": 0.19776909053325653, "global_step": 137226, "epoch": 1541, "lr": 4.9274195264914816e-05}
{"train_loss": 0.14975640177726746, "global_step": 137227, "epoch": 1541, "lr": 4.927361558944367e-05}
{"train_loss": 0.07009003311395645, "global_step": 137228, "epoch": 1541, "lr": 4.927303591407015e-05}
{"train_loss": 0.1326991319656372, "global_step": 137229, "epoch": 1541, "lr": 4.9272456238794396e-05}
{"train_loss": 0.08084356039762497, "global_step": 137230, "epoch": 1541, "lr": 4.927187656361642e-05}
{"train_loss": 0.09850184619426727, "global_step": 137231, "epoch": 1541, "lr": 4.9271296888536355e-05}
{"train_loss": 0.1608458012342453, "global_step": 137232, "epoch": 1541, "lr": 4.9270717213554234e-05}
{"train_loss": 0.12570776045322418, "global_step": 137233, "epoch": 1541, "lr": 4.927013753867018e-05}
{"train_loss": 0.09071781486272812, "global_step": 137234, "epoch": 1541, "lr": 4.9269557863884216e-05}
{"train_loss": 0.09912201762199402, "global_step": 137235, "epoch": 1541, "lr": 4.926897818919648e-05}
{"train_loss": 0.11489667743444443, "global_step": 137236, "epoch": 1541, "lr": 4.9268398514607005e-05}
{"train_loss": 0.1204697017589312, "global_step": 137237, "epoch": 1541, "lr": 4.92678188401159e-05, "val_loss": 5.803440093994141}
{"train_loss": 0.08741674572229385, "global_step": 137238, "epoch": 1542, "lr": 4.9267239165723204e-05}
{"train_loss": 0.11015869677066803, "global_step": 137239, "epoch": 1542, "lr": 4.926665949142904e-05}
{"train_loss": 0.10852910578250885, "global_step": 137240, "epoch": 1542, "lr": 4.926607981723345e-05}
{"train_loss": 0.07467710226774216, "global_step": 137241, "epoch": 1542, "lr": 4.9265500143136545e-05}
{"train_loss": 0.13958176970481873, "global_step": 137242, "epoch": 1542, "lr": 4.9264920469138364e-05}
{"train_loss": 0.18372316658496857, "global_step": 137243, "epoch": 1542, "lr": 4.926434079523903e-05}
{"train_loss": 0.06954384595155716, "global_step": 137244, "epoch": 1542, "lr": 4.926376112143858e-05}
{"train_loss": 0.0830787718296051, "global_step": 137245, "epoch": 1542, "lr": 4.92631814477371e-05}
{"train_loss": 0.13991378247737885, "global_step": 137246, "epoch": 1542, "lr": 4.926260177413469e-05}
{"train_loss": 0.10133683681488037, "global_step": 137247, "epoch": 1542, "lr": 4.92620221006314e-05}
{"train_loss": 0.15379901230335236, "global_step": 137248, "epoch": 1542, "lr": 4.926144242722734e-05}
{"train_loss": 0.126156747341156, "global_step": 137249, "epoch": 1542, "lr": 4.9260862753922556e-05}
{"train_loss": 0.10814471542835236, "global_step": 137250, "epoch": 1542, "lr": 4.926028308071716e-05}
{"train_loss": 0.1361803114414215, "global_step": 137251, "epoch": 1542, "lr": 4.925970340761118e-05}
{"train_loss": 0.07145776599645615, "global_step": 137252, "epoch": 1542, "lr": 4.925912373460475e-05}
{"train_loss": 0.10031214356422424, "global_step": 137253, "epoch": 1542, "lr": 4.92585440616979e-05}
{"train_loss": 0.06513895094394684, "global_step": 137254, "epoch": 1542, "lr": 4.925796438889075e-05}
{"train_loss": 0.12632331252098083, "global_step": 137255, "epoch": 1542, "lr": 4.925738471618334e-05}
{"train_loss": 0.1366044282913208, "global_step": 137256, "epoch": 1542, "lr": 4.9256805043575786e-05}
{"train_loss": 0.1043807789683342, "global_step": 137257, "epoch": 1542, "lr": 4.9256225371068125e-05}
{"train_loss": 0.11710648238658905, "global_step": 137258, "epoch": 1542, "lr": 4.925564569866048e-05}
{"train_loss": 0.1591687649488449, "global_step": 137259, "epoch": 1542, "lr": 4.925506602635288e-05}
{"train_loss": 0.0892655998468399, "global_step": 137260, "epoch": 1542, "lr": 4.925448635414544e-05}
{"train_loss": 0.1292521208524704, "global_step": 137261, "epoch": 1542, "lr": 4.9253906682038216e-05}
{"train_loss": 0.11956629902124405, "global_step": 137262, "epoch": 1542, "lr": 4.925332701003131e-05}
{"train_loss": 0.12868402898311615, "global_step": 137263, "epoch": 1542, "lr": 4.925274733812477e-05}
{"train_loss": 0.14106611907482147, "global_step": 137264, "epoch": 1542, "lr": 4.925216766631869e-05}
{"train_loss": 0.09878765046596527, "global_step": 137265, "epoch": 1542, "lr": 4.9251587994613164e-05}
{"train_loss": 0.11050301045179367, "global_step": 137266, "epoch": 1542, "lr": 4.9251008323008224e-05}
{"train_loss": 0.09688875824213028, "global_step": 137267, "epoch": 1542, "lr": 4.9250428651504006e-05}
{"train_loss": 0.1200638934969902, "global_step": 137268, "epoch": 1542, "lr": 4.924984898010054e-05}
{"train_loss": 0.10282736271619797, "global_step": 137269, "epoch": 1542, "lr": 4.9249269308797934e-05}
{"train_loss": 0.1431528776884079, "global_step": 137270, "epoch": 1542, "lr": 4.924868963759624e-05}
{"train_loss": 0.11906803399324417, "global_step": 137271, "epoch": 1542, "lr": 4.924810996649557e-05}
{"train_loss": 0.07778063416481018, "global_step": 137272, "epoch": 1542, "lr": 4.924753029549596e-05}
{"train_loss": 0.11217176914215088, "global_step": 137273, "epoch": 1542, "lr": 4.9246950624597536e-05}
{"train_loss": 0.07849626988172531, "global_step": 137274, "epoch": 1542, "lr": 4.924637095380033e-05}
{"train_loss": 0.19994935393333435, "global_step": 137275, "epoch": 1542, "lr": 4.9245791283104456e-05}
{"train_loss": 0.16985422372817993, "global_step": 137276, "epoch": 1542, "lr": 4.924521161250996e-05}
{"train_loss": 0.12576772272586823, "global_step": 137277, "epoch": 1542, "lr": 4.924463194201695e-05}
{"train_loss": 0.127170130610466, "global_step": 137278, "epoch": 1542, "lr": 4.9244052271625476e-05}
{"train_loss": 0.08050723373889923, "global_step": 137279, "epoch": 1542, "lr": 4.9243472601335646e-05}
{"train_loss": 0.12415342032909393, "global_step": 137280, "epoch": 1542, "lr": 4.92428929311475e-05}
{"train_loss": 0.0846492275595665, "global_step": 137281, "epoch": 1542, "lr": 4.924231326106115e-05}
{"train_loss": 0.15410107374191284, "global_step": 137282, "epoch": 1542, "lr": 4.924173359107668e-05}
{"train_loss": 0.11091804504394531, "global_step": 137283, "epoch": 1542, "lr": 4.9241153921194114e-05}
{"train_loss": 0.12076050788164139, "global_step": 137284, "epoch": 1542, "lr": 4.9240574251413595e-05}
{"train_loss": 0.12154269963502884, "global_step": 137285, "epoch": 1542, "lr": 4.9239994581735154e-05}
{"train_loss": 0.18130545318126678, "global_step": 137286, "epoch": 1542, "lr": 4.9239414912158905e-05}
{"train_loss": 0.12481594830751419, "global_step": 137287, "epoch": 1542, "lr": 4.923883524268488e-05}
{"train_loss": 0.1315675526857376, "global_step": 137288, "epoch": 1542, "lr": 4.9238255573313216e-05}
{"train_loss": 0.09265170991420746, "global_step": 137289, "epoch": 1542, "lr": 4.9237675904043926e-05}
{"train_loss": 0.19902530312538147, "global_step": 137290, "epoch": 1542, "lr": 4.923709623487715e-05}
{"train_loss": 0.08833715319633484, "global_step": 137291, "epoch": 1542, "lr": 4.923651656581292e-05}
{"train_loss": 0.10719907283782959, "global_step": 137292, "epoch": 1542, "lr": 4.9235936896851345e-05}
{"train_loss": 0.11294932663440704, "global_step": 137293, "epoch": 1542, "lr": 4.9235357227992466e-05}
{"train_loss": 0.07434368878602982, "global_step": 137294, "epoch": 1542, "lr": 4.923477755923641e-05}
{"train_loss": 0.12492325156927109, "global_step": 137295, "epoch": 1542, "lr": 4.923419789058321e-05}
{"train_loss": 0.15557189285755157, "global_step": 137296, "epoch": 1542, "lr": 4.923361822203298e-05}
{"train_loss": 0.15174268186092377, "global_step": 137297, "epoch": 1542, "lr": 4.923303855358576e-05}
{"train_loss": 0.15653091669082642, "global_step": 137298, "epoch": 1542, "lr": 4.9232458885241674e-05}
{"train_loss": 0.13082093000411987, "global_step": 137299, "epoch": 1542, "lr": 4.923187921700074e-05}
{"train_loss": 0.09152796119451523, "global_step": 137300, "epoch": 1542, "lr": 4.9231299548863094e-05}
{"train_loss": 0.1281193196773529, "global_step": 137301, "epoch": 1542, "lr": 4.9230719880828804e-05}
{"train_loss": 0.1236814558506012, "global_step": 137302, "epoch": 1542, "lr": 4.92301402128979e-05}
{"train_loss": 0.07812303304672241, "global_step": 137303, "epoch": 1542, "lr": 4.9229560545070524e-05}
{"train_loss": 0.07870757579803467, "global_step": 137304, "epoch": 1542, "lr": 4.9228980877346696e-05}
{"train_loss": 0.13310407102108002, "global_step": 137305, "epoch": 1542, "lr": 4.922840120972656e-05}
{"train_loss": 0.13968417048454285, "global_step": 137306, "epoch": 1542, "lr": 4.922782154221012e-05}
{"train_loss": 0.10160240530967712, "global_step": 137307, "epoch": 1542, "lr": 4.922724187479751e-05}
{"train_loss": 0.15003708004951477, "global_step": 137308, "epoch": 1542, "lr": 4.9226662207488774e-05}
{"train_loss": 0.08626510202884674, "global_step": 137309, "epoch": 1542, "lr": 4.922608254028402e-05}
{"train_loss": 0.09027327597141266, "global_step": 137310, "epoch": 1542, "lr": 4.922550287318329e-05}
{"train_loss": 0.09804675728082657, "global_step": 137311, "epoch": 1542, "lr": 4.9224923206186705e-05}
{"train_loss": 0.15388530492782593, "global_step": 137312, "epoch": 1542, "lr": 4.922434353929429e-05}
{"train_loss": 0.11000590771436691, "global_step": 137313, "epoch": 1542, "lr": 4.922376387250618e-05}
{"train_loss": 0.16494207084178925, "global_step": 137314, "epoch": 1542, "lr": 4.922318420582241e-05}
{"train_loss": 0.08541973680257797, "global_step": 137315, "epoch": 1542, "lr": 4.922260453924309e-05}
{"train_loss": 0.10835929960012436, "global_step": 137316, "epoch": 1542, "lr": 4.922202487276826e-05}
{"train_loss": 0.10138597339391708, "global_step": 137317, "epoch": 1542, "lr": 4.922144520639803e-05}
{"train_loss": 0.11500126868486404, "global_step": 137318, "epoch": 1542, "lr": 4.9220865540132466e-05}
{"train_loss": 0.1184070035815239, "global_step": 137319, "epoch": 1542, "lr": 4.922028587397165e-05}
{"train_loss": 0.10888679325580597, "global_step": 137320, "epoch": 1542, "lr": 4.9219706207915664e-05}
{"train_loss": 0.1112215667963028, "global_step": 137321, "epoch": 1542, "lr": 4.921912654196456e-05}
{"train_loss": 0.09459884464740753, "global_step": 137322, "epoch": 1542, "lr": 4.9218546876118455e-05}
{"train_loss": 0.09234718233346939, "global_step": 137323, "epoch": 1542, "lr": 4.921796721037739e-05}
{"train_loss": 0.09490548819303513, "global_step": 137324, "epoch": 1542, "lr": 4.921738754474147e-05}
{"train_loss": 0.14663037657737732, "global_step": 137325, "epoch": 1542, "lr": 4.921680787921075e-05}
{"train_loss": 0.11785176379627056, "global_step": 137326, "epoch": 1542, "lr": 4.9216228213785344e-05, "val_loss": 6.022990703582764}
{"train_loss": 0.06674640625715256, "global_step": 137327, "epoch": 1543, "lr": 4.921564854846528e-05}
{"train_loss": 0.09407884627580643, "global_step": 137328, "epoch": 1543, "lr": 4.921506888325069e-05}
{"train_loss": 0.0716872438788414, "global_step": 137329, "epoch": 1543, "lr": 4.92144892181416e-05}
{"train_loss": 0.21938340365886688, "global_step": 137330, "epoch": 1543, "lr": 4.921390955313813e-05}
{"train_loss": 0.1380985826253891, "global_step": 137331, "epoch": 1543, "lr": 4.921332988824032e-05}
{"train_loss": 0.10130321234464645, "global_step": 137332, "epoch": 1543, "lr": 4.921275022344829e-05}
{"train_loss": 0.07592325657606125, "global_step": 137333, "epoch": 1543, "lr": 4.9212170558762076e-05}
{"train_loss": 0.12529775500297546, "global_step": 137334, "epoch": 1543, "lr": 4.92115908941818e-05}
{"train_loss": 0.14881065487861633, "global_step": 137335, "epoch": 1543, "lr": 4.9211011229707485e-05}
{"train_loss": 0.09611360728740692, "global_step": 137336, "epoch": 1543, "lr": 4.921043156533926e-05}
{"train_loss": 0.10332904011011124, "global_step": 137337, "epoch": 1543, "lr": 4.9209851901077184e-05}
{"train_loss": 0.11201278120279312, "global_step": 137338, "epoch": 1543, "lr": 4.920927223692133e-05}
{"train_loss": 0.11659608036279678, "global_step": 137339, "epoch": 1543, "lr": 4.920869257287179e-05}
{"train_loss": 0.14552992582321167, "global_step": 137340, "epoch": 1543, "lr": 4.9208112908928606e-05}
{"train_loss": 0.10081975907087326, "global_step": 137341, "epoch": 1543, "lr": 4.920753324509191e-05}
{"train_loss": 0.11391008645296097, "global_step": 137342, "epoch": 1543, "lr": 4.9206953581361724e-05}
{"train_loss": 0.12984289228916168, "global_step": 137343, "epoch": 1543, "lr": 4.920637391773818e-05}
{"train_loss": 0.15798573195934296, "global_step": 137344, "epoch": 1543, "lr": 4.920579425422131e-05}
{"train_loss": 0.13572685420513153, "global_step": 137345, "epoch": 1543, "lr": 4.9205214590811226e-05}
{"train_loss": 0.09786161035299301, "global_step": 137346, "epoch": 1543, "lr": 4.9204634927507975e-05}
{"train_loss": 0.08997108042240143, "global_step": 137347, "epoch": 1543, "lr": 4.920405526431167e-05}
{"train_loss": 0.15745596587657928, "global_step": 137348, "epoch": 1543, "lr": 4.920347560122235e-05}
{"train_loss": 0.09391676634550095, "global_step": 137349, "epoch": 1543, "lr": 4.9202895938240136e-05}
{"train_loss": 0.14768889546394348, "global_step": 137350, "epoch": 1543, "lr": 4.9202316275365057e-05}
{"train_loss": 0.08117570728063583, "global_step": 137351, "epoch": 1543, "lr": 4.920173661259724e-05}
{"train_loss": 0.09202136099338531, "global_step": 137352, "epoch": 1543, "lr": 4.9201156949936716e-05}
{"train_loss": 0.12089599668979645, "global_step": 137353, "epoch": 1543, "lr": 4.92005772873836e-05}
{"train_loss": 0.14496323466300964, "global_step": 137354, "epoch": 1543, "lr": 4.919999762493796e-05}
{"train_loss": 0.07099571079015732, "global_step": 137355, "epoch": 1543, "lr": 4.919941796259986e-05}
{"train_loss": 0.12207147479057312, "global_step": 137356, "epoch": 1543, "lr": 4.91988383003694e-05}
{"train_loss": 0.10887610167264938, "global_step": 137357, "epoch": 1543, "lr": 4.919825863824664e-05}
{"train_loss": 0.11097245663404465, "global_step": 137358, "epoch": 1543, "lr": 4.9197678976231674e-05}
{"train_loss": 0.13654972612857819, "global_step": 137359, "epoch": 1543, "lr": 4.9197099314324556e-05}
{"train_loss": 0.11833338439464569, "global_step": 137360, "epoch": 1543, "lr": 4.91965196525254e-05}
{"train_loss": 0.15111082792282104, "global_step": 137361, "epoch": 1543, "lr": 4.919593999083423e-05}
{"train_loss": 0.09004838019609451, "global_step": 137362, "epoch": 1543, "lr": 4.919536032925119e-05}
{"train_loss": 0.10248187184333801, "global_step": 137363, "epoch": 1543, "lr": 4.91947806677763e-05}
{"train_loss": 0.16108755767345428, "global_step": 137364, "epoch": 1543, "lr": 4.9194201006409675e-05}
{"train_loss": 0.13323254883289337, "global_step": 137365, "epoch": 1543, "lr": 4.9193621345151364e-05}
{"train_loss": 0.09363055229187012, "global_step": 137366, "epoch": 1543, "lr": 4.919304168400149e-05}
{"train_loss": 0.08687134832143784, "global_step": 137367, "epoch": 1543, "lr": 4.9192462022960064e-05}
{"train_loss": 0.14457586407661438, "global_step": 137368, "epoch": 1543, "lr": 4.919188236202723e-05}
{"train_loss": 0.15320917963981628, "global_step": 137369, "epoch": 1543, "lr": 4.9191302701203015e-05}
{"train_loss": 0.1091025248169899, "global_step": 137370, "epoch": 1543, "lr": 4.9190723040487544e-05}
{"train_loss": 0.10350318998098373, "global_step": 137371, "epoch": 1543, "lr": 4.9190143379880846e-05}
{"train_loss": 0.14136873185634613, "global_step": 137372, "epoch": 1543, "lr": 4.918956371938303e-05}
{"train_loss": 0.08847113698720932, "global_step": 137373, "epoch": 1543, "lr": 4.918898405899418e-05}
{"train_loss": 0.09046090394258499, "global_step": 137374, "epoch": 1543, "lr": 4.9188404398714355e-05}
{"train_loss": 0.1738237738609314, "global_step": 137375, "epoch": 1543, "lr": 4.918782473854364e-05}
{"train_loss": 0.08838716149330139, "global_step": 137376, "epoch": 1543, "lr": 4.9187245078482114e-05}
{"train_loss": 0.10400902479887009, "global_step": 137377, "epoch": 1543, "lr": 4.918666541852985e-05}
{"train_loss": 0.11912642419338226, "global_step": 137378, "epoch": 1543, "lr": 4.918608575868693e-05}
{"train_loss": 0.14325305819511414, "global_step": 137379, "epoch": 1543, "lr": 4.918550609895344e-05}
{"train_loss": 0.09671638906002045, "global_step": 137380, "epoch": 1543, "lr": 4.918492643932943e-05}
{"train_loss": 0.06445950269699097, "global_step": 137381, "epoch": 1543, "lr": 4.918434677981501e-05}
{"train_loss": 0.18081051111221313, "global_step": 137382, "epoch": 1543, "lr": 4.918376712041023e-05}
{"train_loss": 0.08634571731090546, "global_step": 137383, "epoch": 1543, "lr": 4.918318746111521e-05}
{"train_loss": 0.04766266420483589, "global_step": 137384, "epoch": 1543, "lr": 4.9182607801929966e-05}
{"train_loss": 0.10125084966421127, "global_step": 137385, "epoch": 1543, "lr": 4.9182028142854635e-05}
{"train_loss": 0.2351130247116089, "global_step": 137386, "epoch": 1543, "lr": 4.9181448483889256e-05}
{"train_loss": 0.0793987438082695, "global_step": 137387, "epoch": 1543, "lr": 4.918086882503394e-05}
{"train_loss": 0.10434851050376892, "global_step": 137388, "epoch": 1543, "lr": 4.918028916628872e-05}
{"train_loss": 0.1829855591058731, "global_step": 137389, "epoch": 1543, "lr": 4.917970950765372e-05}
{"train_loss": 0.07540665566921234, "global_step": 137390, "epoch": 1543, "lr": 4.9179129849128985e-05}
{"train_loss": 0.1169290542602539, "global_step": 137391, "epoch": 1543, "lr": 4.91785501907146e-05}
{"train_loss": 0.11929377168416977, "global_step": 137392, "epoch": 1543, "lr": 4.917797053241067e-05}
{"train_loss": 0.1050223633646965, "global_step": 137393, "epoch": 1543, "lr": 4.917739087421725e-05}
{"train_loss": 0.2158380150794983, "global_step": 137394, "epoch": 1543, "lr": 4.917681121613441e-05}
{"train_loss": 0.07097186893224716, "global_step": 137395, "epoch": 1543, "lr": 4.9176231558162244e-05}
{"train_loss": 0.12147585302591324, "global_step": 137396, "epoch": 1543, "lr": 4.917565190030082e-05}
{"train_loss": 0.06214212253689766, "global_step": 137397, "epoch": 1543, "lr": 4.917507224255022e-05}
{"train_loss": 0.17396573722362518, "global_step": 137398, "epoch": 1543, "lr": 4.917449258491054e-05}
{"train_loss": 0.18366149067878723, "global_step": 137399, "epoch": 1543, "lr": 4.91739129273818e-05}
{"train_loss": 0.19715356826782227, "global_step": 137400, "epoch": 1543, "lr": 4.9173333269964155e-05}
{"train_loss": 0.08235261589288712, "global_step": 137401, "epoch": 1543, "lr": 4.917275361265762e-05}
{"train_loss": 0.12240896373987198, "global_step": 137402, "epoch": 1543, "lr": 4.9172173955462315e-05}
{"train_loss": 0.08940008282661438, "global_step": 137403, "epoch": 1543, "lr": 4.9171594298378285e-05}
{"train_loss": 0.14082598686218262, "global_step": 137404, "epoch": 1543, "lr": 4.9171014641405636e-05}
{"train_loss": 0.12475840002298355, "global_step": 137405, "epoch": 1543, "lr": 4.917043498454442e-05}
{"train_loss": 0.08885108679533005, "global_step": 137406, "epoch": 1543, "lr": 4.916985532779476e-05}
{"train_loss": 0.07299881428480148, "global_step": 137407, "epoch": 1543, "lr": 4.916927567115667e-05}
{"train_loss": 0.07777527719736099, "global_step": 137408, "epoch": 1543, "lr": 4.9168696014630256e-05}
{"train_loss": 0.11047302931547165, "global_step": 137409, "epoch": 1543, "lr": 4.916811635821563e-05}
{"train_loss": 0.12497631460428238, "global_step": 137410, "epoch": 1543, "lr": 4.916753670191282e-05}
{"train_loss": 0.1328759789466858, "global_step": 137411, "epoch": 1543, "lr": 4.916695704572194e-05}
{"train_loss": 0.08516039699316025, "global_step": 137412, "epoch": 1543, "lr": 4.916637738964305e-05}
{"train_loss": 0.16208116710186005, "global_step": 137413, "epoch": 1543, "lr": 4.916579773367622e-05}
{"train_loss": 0.1251007467508316, "global_step": 137414, "epoch": 1543, "lr": 4.916521807782154e-05}
{"train_loss": 0.11779586267605256, "global_step": 137415, "epoch": 1543, "lr": 4.91646384220791e-05, "val_loss": 5.86928653717041}
{"train_loss": 0.10362264513969421, "global_step": 137416, "epoch": 1544, "lr": 4.916405876644895e-05}
{"train_loss": 0.1267976462841034, "global_step": 137417, "epoch": 1544, "lr": 4.9163479110931196e-05}
{"train_loss": 0.10143186151981354, "global_step": 137418, "epoch": 1544, "lr": 4.916289945552588e-05}
{"train_loss": 0.1710660606622696, "global_step": 137419, "epoch": 1544, "lr": 4.916231980023313e-05}
{"train_loss": 0.13607603311538696, "global_step": 137420, "epoch": 1544, "lr": 4.916174014505297e-05}
{"train_loss": 0.11171536892652512, "global_step": 137421, "epoch": 1544, "lr": 4.9161160489985524e-05}
{"train_loss": 0.11639462411403656, "global_step": 137422, "epoch": 1544, "lr": 4.9160580835030826e-05}
{"train_loss": 0.09274004399776459, "global_step": 137423, "epoch": 1544, "lr": 4.9160001180189e-05}
{"train_loss": 0.1108212098479271, "global_step": 137424, "epoch": 1544, "lr": 4.915942152546009e-05}
{"train_loss": 0.08991385251283646, "global_step": 137425, "epoch": 1544, "lr": 4.9158841870844195e-05}
{"train_loss": 0.13353538513183594, "global_step": 137426, "epoch": 1544, "lr": 4.9158262216341365e-05}
{"train_loss": 0.13721296191215515, "global_step": 137427, "epoch": 1544, "lr": 4.91576825619517e-05}
{"train_loss": 0.03654009848833084, "global_step": 137428, "epoch": 1544, "lr": 4.91571029076753e-05}
{"train_loss": 0.1460019052028656, "global_step": 137429, "epoch": 1544, "lr": 4.915652325351218e-05}
{"train_loss": 0.1145283505320549, "global_step": 137430, "epoch": 1544, "lr": 4.915594359946249e-05}
{"train_loss": 0.13273268938064575, "global_step": 137431, "epoch": 1544, "lr": 4.9155363945526256e-05}
{"train_loss": 0.16136625409126282, "global_step": 137432, "epoch": 1544, "lr": 4.915478429170357e-05}
{"train_loss": 0.10264790058135986, "global_step": 137433, "epoch": 1544, "lr": 4.915420463799453e-05}
{"train_loss": 0.17142555117607117, "global_step": 137434, "epoch": 1544, "lr": 4.915362498439918e-05}
{"train_loss": 0.16062909364700317, "global_step": 137435, "epoch": 1544, "lr": 4.915304533091762e-05}
{"train_loss": 0.18050585687160492, "global_step": 137436, "epoch": 1544, "lr": 4.9152465677549933e-05}
{"train_loss": 0.2097412645816803, "global_step": 137437, "epoch": 1544, "lr": 4.9151886024296165e-05}
{"train_loss": 0.09935251623392105, "global_step": 137438, "epoch": 1544, "lr": 4.915130637115644e-05}
{"train_loss": 0.10312680155038834, "global_step": 137439, "epoch": 1544, "lr": 4.9150726718130784e-05}
{"train_loss": 0.10000969469547272, "global_step": 137440, "epoch": 1544, "lr": 4.915014706521933e-05}
{"train_loss": 0.10184887051582336, "global_step": 137441, "epoch": 1544, "lr": 4.91495674124221e-05}
{"train_loss": 0.11764967441558838, "global_step": 137442, "epoch": 1544, "lr": 4.9148987759739234e-05}
{"train_loss": 0.2117716372013092, "global_step": 137443, "epoch": 1544, "lr": 4.914840810717074e-05}
{"train_loss": 0.1227928102016449, "global_step": 137444, "epoch": 1544, "lr": 4.914782845471674e-05}
{"train_loss": 0.12305985391139984, "global_step": 137445, "epoch": 1544, "lr": 4.914724880237733e-05}
{"train_loss": 0.047769855707883835, "global_step": 137446, "epoch": 1544, "lr": 4.914666915015254e-05}
{"train_loss": 0.1410233974456787, "global_step": 137447, "epoch": 1544, "lr": 4.914608949804248e-05}
{"train_loss": 0.12349928170442581, "global_step": 137448, "epoch": 1544, "lr": 4.914550984604721e-05}
{"train_loss": 0.0804724395275116, "global_step": 137449, "epoch": 1544, "lr": 4.9144930194166826e-05}
{"train_loss": 0.09803854674100876, "global_step": 137450, "epoch": 1544, "lr": 4.914435054240139e-05}
{"train_loss": 0.08343911170959473, "global_step": 137451, "epoch": 1544, "lr": 4.914377089075099e-05}
{"train_loss": 0.08783459663391113, "global_step": 137452, "epoch": 1544, "lr": 4.914319123921569e-05}
{"train_loss": 0.18883438408374786, "global_step": 137453, "epoch": 1544, "lr": 4.9142611587795576e-05}
{"train_loss": 0.13916349411010742, "global_step": 137454, "epoch": 1544, "lr": 4.914203193649074e-05}
{"train_loss": 0.11741706728935242, "global_step": 137455, "epoch": 1544, "lr": 4.914145228530125e-05}
{"train_loss": 0.10471943020820618, "global_step": 137456, "epoch": 1544, "lr": 4.914087263422716e-05}
{"train_loss": 0.07129741460084915, "global_step": 137457, "epoch": 1544, "lr": 4.9140292983268594e-05}
{"train_loss": 0.12179102748632431, "global_step": 137458, "epoch": 1544, "lr": 4.913971333242558e-05}
{"train_loss": 0.12624625861644745, "global_step": 137459, "epoch": 1544, "lr": 4.9139133681698244e-05}
{"train_loss": 0.06954890489578247, "global_step": 137460, "epoch": 1544, "lr": 4.913855403108662e-05}
{"train_loss": 0.11149945110082626, "global_step": 137461, "epoch": 1544, "lr": 4.913797438059083e-05}
{"train_loss": 0.13908806443214417, "global_step": 137462, "epoch": 1544, "lr": 4.913739473021091e-05}
{"train_loss": 0.136480450630188, "global_step": 137463, "epoch": 1544, "lr": 4.913681507994695e-05}
{"train_loss": 0.10173650830984116, "global_step": 137464, "epoch": 1544, "lr": 4.9136235429799054e-05}
{"train_loss": 0.08623822033405304, "global_step": 137465, "epoch": 1544, "lr": 4.9135655779767264e-05}
{"train_loss": 0.1152167096734047, "global_step": 137466, "epoch": 1544, "lr": 4.91350761298517e-05}
{"train_loss": 0.13717108964920044, "global_step": 137467, "epoch": 1544, "lr": 4.913449648005238e-05}
{"train_loss": 0.06444545090198517, "global_step": 137468, "epoch": 1544, "lr": 4.9133916830369456e-05}
{"train_loss": 0.1153450533747673, "global_step": 137469, "epoch": 1544, "lr": 4.9133337180802944e-05}
{"train_loss": 0.1304534673690796, "global_step": 137470, "epoch": 1544, "lr": 4.9132757531352944e-05}
{"train_loss": 0.15722452104091644, "global_step": 137471, "epoch": 1544, "lr": 4.9132177882019535e-05}
{"train_loss": 0.0814947783946991, "global_step": 137472, "epoch": 1544, "lr": 4.9131598232802794e-05}
{"train_loss": 0.10144741088151932, "global_step": 137473, "epoch": 1544, "lr": 4.91310185837028e-05}
{"train_loss": 0.12904511392116547, "global_step": 137474, "epoch": 1544, "lr": 4.913043893471964e-05}
{"train_loss": 0.1320783793926239, "global_step": 137475, "epoch": 1544, "lr": 4.912985928585337e-05}
{"train_loss": 0.09689076244831085, "global_step": 137476, "epoch": 1544, "lr": 4.912927963710409e-05}
{"train_loss": 0.11519229412078857, "global_step": 137477, "epoch": 1544, "lr": 4.912869998847185e-05}
{"train_loss": 0.1437854766845703, "global_step": 137478, "epoch": 1544, "lr": 4.912812033995677e-05}
{"train_loss": 0.10391884297132492, "global_step": 137479, "epoch": 1544, "lr": 4.9127540691558874e-05}
{"train_loss": 0.13154077529907227, "global_step": 137480, "epoch": 1544, "lr": 4.912696104327827e-05}
{"train_loss": 0.07201150804758072, "global_step": 137481, "epoch": 1544, "lr": 4.9126381395115065e-05}
{"train_loss": 0.09642352908849716, "global_step": 137482, "epoch": 1544, "lr": 4.912580174706928e-05}
{"train_loss": 0.07980626821517944, "global_step": 137483, "epoch": 1544, "lr": 4.912522209914104e-05}
{"train_loss": 0.0929255411028862, "global_step": 137484, "epoch": 1544, "lr": 4.912464245133039e-05}
{"train_loss": 0.14503200352191925, "global_step": 137485, "epoch": 1544, "lr": 4.912406280363744e-05}
{"train_loss": 0.17527985572814941, "global_step": 137486, "epoch": 1544, "lr": 4.912348315606222e-05}
{"train_loss": 0.14127901196479797, "global_step": 137487, "epoch": 1544, "lr": 4.912290350860487e-05}
{"train_loss": 0.08496738970279694, "global_step": 137488, "epoch": 1544, "lr": 4.912232386126541e-05}
{"train_loss": 0.15056966245174408, "global_step": 137489, "epoch": 1544, "lr": 4.912174421404396e-05}
{"train_loss": 0.06144427880644798, "global_step": 137490, "epoch": 1544, "lr": 4.912116456694057e-05}
{"train_loss": 0.07641945779323578, "global_step": 137491, "epoch": 1544, "lr": 4.912058491995533e-05}
{"train_loss": 0.0845942497253418, "global_step": 137492, "epoch": 1544, "lr": 4.912000527308832e-05}
{"train_loss": 0.13201525807380676, "global_step": 137493, "epoch": 1544, "lr": 4.911942562633961e-05}
{"train_loss": 0.1580582559108734, "global_step": 137494, "epoch": 1544, "lr": 4.9118845979709286e-05}
{"train_loss": 0.06336959451436996, "global_step": 137495, "epoch": 1544, "lr": 4.911826633319743e-05}
{"train_loss": 0.07021333277225494, "global_step": 137496, "epoch": 1544, "lr": 4.9117686686804096e-05}
{"train_loss": 0.12975645065307617, "global_step": 137497, "epoch": 1544, "lr": 4.911710704052939e-05}
{"train_loss": 0.09543521702289581, "global_step": 137498, "epoch": 1544, "lr": 4.911652739437337e-05}
{"train_loss": 0.09133455902338028, "global_step": 137499, "epoch": 1544, "lr": 4.911594774833611e-05}
{"train_loss": 0.12220081686973572, "global_step": 137500, "epoch": 1544, "lr": 4.911536810241772e-05}
{"train_loss": 0.18724778294563293, "global_step": 137501, "epoch": 1544, "lr": 4.911478845661825e-05}
{"train_loss": 0.08355449140071869, "global_step": 137502, "epoch": 1544, "lr": 4.9114208810937804e-05}
{"train_loss": 0.04921131953597069, "global_step": 137503, "epoch": 1544, "lr": 4.911362916537641e-05}
{"train_loss": 0.11625777055205924, "global_step": 137504, "epoch": 1544, "lr": 4.911304951993421e-05, "val_loss": 5.767259120941162}
{"train_loss": 0.14427302777767181, "global_step": 137505, "epoch": 1545, "lr": 4.9112469874611225e-05}
{"train_loss": 0.13147105276584625, "global_step": 137506, "epoch": 1545, "lr": 4.911189022940758e-05}
{"train_loss": 0.07900942862033844, "global_step": 137507, "epoch": 1545, "lr": 4.9111310584323306e-05}
{"train_loss": 0.062175992876291275, "global_step": 137508, "epoch": 1545, "lr": 4.911073093935853e-05}
{"train_loss": 0.07910317927598953, "global_step": 137509, "epoch": 1545, "lr": 4.911015129451329e-05}
{"train_loss": 0.09312465786933899, "global_step": 137510, "epoch": 1545, "lr": 4.910957164978768e-05}
{"train_loss": 0.06175669655203819, "global_step": 137511, "epoch": 1545, "lr": 4.910899200518178e-05}
{"train_loss": 0.07395651936531067, "global_step": 137512, "epoch": 1545, "lr": 4.9108412360695664e-05}
{"train_loss": 0.1265542358160019, "global_step": 137513, "epoch": 1545, "lr": 4.910783271632942e-05}
{"train_loss": 0.20811349153518677, "global_step": 137514, "epoch": 1545, "lr": 4.9107253072083115e-05}
{"train_loss": 0.10650847107172012, "global_step": 137515, "epoch": 1545, "lr": 4.910667342795681e-05}
{"train_loss": 0.1220192015171051, "global_step": 137516, "epoch": 1545, "lr": 4.9106093783950624e-05}
{"train_loss": 0.09060050547122955, "global_step": 137517, "epoch": 1545, "lr": 4.910551414006459e-05}
{"train_loss": 0.09915006905794144, "global_step": 137518, "epoch": 1545, "lr": 4.9104934496298814e-05}
{"train_loss": 0.09215356409549713, "global_step": 137519, "epoch": 1545, "lr": 4.910435485265339e-05}
{"train_loss": 0.11604560166597366, "global_step": 137520, "epoch": 1545, "lr": 4.910377520912835e-05}
{"train_loss": 0.12560799717903137, "global_step": 137521, "epoch": 1545, "lr": 4.910319556572382e-05}
{"train_loss": 0.11019764840602875, "global_step": 137522, "epoch": 1545, "lr": 4.910261592243983e-05}
{"train_loss": 0.17163456976413727, "global_step": 137523, "epoch": 1545, "lr": 4.9102036279276504e-05}
{"train_loss": 0.10306818038225174, "global_step": 137524, "epoch": 1545, "lr": 4.910145663623388e-05}
{"train_loss": 0.08709076792001724, "global_step": 137525, "epoch": 1545, "lr": 4.910087699331207e-05}
{"train_loss": 0.09975025057792664, "global_step": 137526, "epoch": 1545, "lr": 4.9100297350511124e-05}
{"train_loss": 0.08559418469667435, "global_step": 137527, "epoch": 1545, "lr": 4.9099717707831146e-05}
{"train_loss": 0.18504609167575836, "global_step": 137528, "epoch": 1545, "lr": 4.909913806527219e-05}
{"train_loss": 0.17412829399108887, "global_step": 137529, "epoch": 1545, "lr": 4.9098558422834344e-05}
{"train_loss": 0.10112271457910538, "global_step": 137530, "epoch": 1545, "lr": 4.909797878051769e-05}
{"train_loss": 0.14583510160446167, "global_step": 137531, "epoch": 1545, "lr": 4.9097399138322295e-05}
{"train_loss": 0.14829480648040771, "global_step": 137532, "epoch": 1545, "lr": 4.909681949624825e-05}
{"train_loss": 0.13188597559928894, "global_step": 137533, "epoch": 1545, "lr": 4.909623985429563e-05}
{"train_loss": 0.09815594553947449, "global_step": 137534, "epoch": 1545, "lr": 4.90956602124645e-05}
{"train_loss": 0.15105654299259186, "global_step": 137535, "epoch": 1545, "lr": 4.909508057075494e-05}
{"train_loss": 0.1296217143535614, "global_step": 137536, "epoch": 1545, "lr": 4.909450092916705e-05}
{"train_loss": 0.12854598462581635, "global_step": 137537, "epoch": 1545, "lr": 4.9093921287700885e-05}
{"train_loss": 0.14059244096279144, "global_step": 137538, "epoch": 1545, "lr": 4.909334164635654e-05}
{"train_loss": 0.07609657198190689, "global_step": 137539, "epoch": 1545, "lr": 4.909276200513407e-05}
{"train_loss": 0.10231238603591919, "global_step": 137540, "epoch": 1545, "lr": 4.909218236403359e-05}
{"train_loss": 0.09638085216283798, "global_step": 137541, "epoch": 1545, "lr": 4.9091602723055135e-05}
{"train_loss": 0.15437230467796326, "global_step": 137542, "epoch": 1545, "lr": 4.9091023082198826e-05}
{"train_loss": 0.14246317744255066, "global_step": 137543, "epoch": 1545, "lr": 4.9090443441464686e-05}
{"train_loss": 0.18297982215881348, "global_step": 137544, "epoch": 1545, "lr": 4.908986380085285e-05}
{"train_loss": 0.13110578060150146, "global_step": 137545, "epoch": 1545, "lr": 4.9089284160363355e-05}
{"train_loss": 0.07805096358060837, "global_step": 137546, "epoch": 1545, "lr": 4.908870451999631e-05}
{"train_loss": 0.1558663547039032, "global_step": 137547, "epoch": 1545, "lr": 4.908812487975177e-05}
{"train_loss": 0.1503150910139084, "global_step": 137548, "epoch": 1545, "lr": 4.908754523962982e-05}
{"train_loss": 0.1288629174232483, "global_step": 137549, "epoch": 1545, "lr": 4.908696559963054e-05}
{"train_loss": 0.1080952063202858, "global_step": 137550, "epoch": 1545, "lr": 4.9086385959754e-05}
{"train_loss": 0.08533339202404022, "global_step": 137551, "epoch": 1545, "lr": 4.9085806320000286e-05}
{"train_loss": 0.1896350383758545, "global_step": 137552, "epoch": 1545, "lr": 4.908522668036949e-05}
{"train_loss": 0.18032515048980713, "global_step": 137553, "epoch": 1545, "lr": 4.9084647040861645e-05}
{"train_loss": 0.09707620739936829, "global_step": 137554, "epoch": 1545, "lr": 4.908406740147687e-05}
{"train_loss": 0.08260975778102875, "global_step": 137555, "epoch": 1545, "lr": 4.908348776221524e-05}
{"train_loss": 0.11368487775325775, "global_step": 137556, "epoch": 1545, "lr": 4.908290812307681e-05}
{"train_loss": 0.11596664786338806, "global_step": 137557, "epoch": 1545, "lr": 4.908232848406169e-05}
{"train_loss": 0.11859921365976334, "global_step": 137558, "epoch": 1545, "lr": 4.908174884516992e-05}
{"train_loss": 0.11276236921548843, "global_step": 137559, "epoch": 1545, "lr": 4.9081169206401616e-05}
{"train_loss": 0.2125302106142044, "global_step": 137560, "epoch": 1545, "lr": 4.908058956775682e-05}
{"train_loss": 0.0792471393942833, "global_step": 137561, "epoch": 1545, "lr": 4.908000992923565e-05}
{"train_loss": 0.12325987219810486, "global_step": 137562, "epoch": 1545, "lr": 4.907943029083813e-05}
{"train_loss": 0.11037320643663406, "global_step": 137563, "epoch": 1545, "lr": 4.9078850652564406e-05}
{"train_loss": 0.06318081170320511, "global_step": 137564, "epoch": 1545, "lr": 4.907827101441449e-05}
{"train_loss": 0.10289046168327332, "global_step": 137565, "epoch": 1545, "lr": 4.907769137638851e-05}
{"train_loss": 0.19846737384796143, "global_step": 137566, "epoch": 1545, "lr": 4.907711173848652e-05}
{"train_loss": 0.17528308928012848, "global_step": 137567, "epoch": 1545, "lr": 4.907653210070859e-05}
{"train_loss": 0.12158270180225372, "global_step": 137568, "epoch": 1545, "lr": 4.907595246305482e-05}
{"train_loss": 0.11542628705501556, "global_step": 137569, "epoch": 1545, "lr": 4.907537282552526e-05}
{"train_loss": 0.11745480448007584, "global_step": 137570, "epoch": 1545, "lr": 4.9074793188120014e-05}
{"train_loss": 0.09174119681119919, "global_step": 137571, "epoch": 1545, "lr": 4.907421355083914e-05}
{"train_loss": 0.12920846045017242, "global_step": 137572, "epoch": 1545, "lr": 4.907363391368275e-05}
{"train_loss": 0.13331912457942963, "global_step": 137573, "epoch": 1545, "lr": 4.9073054276650885e-05}
{"train_loss": 0.1601899415254593, "global_step": 137574, "epoch": 1545, "lr": 4.9072474639743647e-05}
{"train_loss": 0.08068805187940598, "global_step": 137575, "epoch": 1545, "lr": 4.9071895002961085e-05}
{"train_loss": 0.1537754088640213, "global_step": 137576, "epoch": 1545, "lr": 4.9071315366303315e-05}
{"train_loss": 0.147583469748497, "global_step": 137577, "epoch": 1545, "lr": 4.907073572977038e-05}
{"train_loss": 0.07458551973104477, "global_step": 137578, "epoch": 1545, "lr": 4.907015609336239e-05}
{"train_loss": 0.12262723594903946, "global_step": 137579, "epoch": 1545, "lr": 4.906957645707938e-05}
{"train_loss": 0.10473259538412094, "global_step": 137580, "epoch": 1545, "lr": 4.906899682092148e-05}
{"train_loss": 0.07239691913127899, "global_step": 137581, "epoch": 1545, "lr": 4.906841718488872e-05}
{"train_loss": 0.19393177330493927, "global_step": 137582, "epoch": 1545, "lr": 4.906783754898123e-05}
{"train_loss": 0.1398485004901886, "global_step": 137583, "epoch": 1545, "lr": 4.906725791319902e-05}
{"train_loss": 0.1360916793346405, "global_step": 137584, "epoch": 1545, "lr": 4.906667827754224e-05}
{"train_loss": 0.12468405812978745, "global_step": 137585, "epoch": 1545, "lr": 4.906609864201091e-05}
{"train_loss": 0.08154827356338501, "global_step": 137586, "epoch": 1545, "lr": 4.906551900660515e-05}
{"train_loss": 0.1503874659538269, "global_step": 137587, "epoch": 1545, "lr": 4.906493937132501e-05}
{"train_loss": 0.2309325635433197, "global_step": 137588, "epoch": 1545, "lr": 4.906435973617058e-05}
{"train_loss": 0.15863874554634094, "global_step": 137589, "epoch": 1545, "lr": 4.906378010114193e-05}
{"train_loss": 0.09022969752550125, "global_step": 137590, "epoch": 1545, "lr": 4.9063200466239136e-05}
{"train_loss": 0.16351856291294098, "global_step": 137591, "epoch": 1545, "lr": 4.906262083146231e-05}
{"train_loss": 0.07229389995336533, "global_step": 137592, "epoch": 1545, "lr": 4.906204119681148e-05}
{"train_loss": 0.12349574519007393, "global_step": 137593, "epoch": 1545, "lr": 4.906146156228676e-05, "val_loss": 5.586440563201904, "train_action_mse_error": 29.126583099365234}
{"train_loss": 0.15329793095588684, "global_step": 137594, "epoch": 1546, "lr": 4.9060881927888196e-05}
{"train_loss": 0.16568736732006073, "global_step": 137595, "epoch": 1546, "lr": 4.9060302293615915e-05}
{"train_loss": 0.13529658317565918, "global_step": 137596, "epoch": 1546, "lr": 4.905972265946993e-05}
{"train_loss": 0.0625862330198288, "global_step": 137597, "epoch": 1546, "lr": 4.905914302545039e-05}
{"train_loss": 0.14972682297229767, "global_step": 137598, "epoch": 1546, "lr": 4.905856339155731e-05}
{"train_loss": 0.13276545703411102, "global_step": 137599, "epoch": 1546, "lr": 4.9057983757790807e-05}
{"train_loss": 0.12508940696716309, "global_step": 137600, "epoch": 1546, "lr": 4.9057404124150934e-05}
{"train_loss": 0.10704977810382843, "global_step": 137601, "epoch": 1546, "lr": 4.90568244906378e-05}
{"train_loss": 0.18819428980350494, "global_step": 137602, "epoch": 1546, "lr": 4.9056244857251444e-05}
{"train_loss": 0.10894086956977844, "global_step": 137603, "epoch": 1546, "lr": 4.9055665223991984e-05}
{"train_loss": 0.1644463986158371, "global_step": 137604, "epoch": 1546, "lr": 4.9055085590859463e-05}
{"train_loss": 0.10040782392024994, "global_step": 137605, "epoch": 1546, "lr": 4.905450595785399e-05}
{"train_loss": 0.180805042386055, "global_step": 137606, "epoch": 1546, "lr": 4.905392632497562e-05}
{"train_loss": 0.06955303251743317, "global_step": 137607, "epoch": 1546, "lr": 4.905334669222442e-05}
{"train_loss": 0.1377306431531906, "global_step": 137608, "epoch": 1546, "lr": 4.905276705960052e-05}
{"train_loss": 0.09284092485904694, "global_step": 137609, "epoch": 1546, "lr": 4.905218742710393e-05}
{"train_loss": 0.11121393740177155, "global_step": 137610, "epoch": 1546, "lr": 4.905160779473479e-05}
{"train_loss": 0.08742576837539673, "global_step": 137611, "epoch": 1546, "lr": 4.905102816249313e-05}
{"train_loss": 0.14920300245285034, "global_step": 137612, "epoch": 1546, "lr": 4.9050448530379063e-05}
{"train_loss": 0.09587883949279785, "global_step": 137613, "epoch": 1546, "lr": 4.904986889839264e-05}
{"train_loss": 0.15882977843284607, "global_step": 137614, "epoch": 1546, "lr": 4.904928926653396e-05}
{"train_loss": 0.14566874504089355, "global_step": 137615, "epoch": 1546, "lr": 4.904870963480308e-05}
{"train_loss": 0.11441405117511749, "global_step": 137616, "epoch": 1546, "lr": 4.90481300032001e-05}
{"train_loss": 0.09191488474607468, "global_step": 137617, "epoch": 1546, "lr": 4.904755037172507e-05}
{"train_loss": 0.14424662292003632, "global_step": 137618, "epoch": 1546, "lr": 4.9046970740378114e-05}
{"train_loss": 0.12606585025787354, "global_step": 137619, "epoch": 1546, "lr": 4.904639110915925e-05}
{"train_loss": 0.11822306364774704, "global_step": 137620, "epoch": 1546, "lr": 4.9045811478068615e-05}
{"train_loss": 0.11059533804655075, "global_step": 137621, "epoch": 1546, "lr": 4.904523184710624e-05}
{"train_loss": 0.165470689535141, "global_step": 137622, "epoch": 1546, "lr": 4.9044652216272235e-05}
{"train_loss": 0.09132778644561768, "global_step": 137623, "epoch": 1546, "lr": 4.904407258556665e-05}
{"train_loss": 0.14144426584243774, "global_step": 137624, "epoch": 1546, "lr": 4.90434929549896e-05}
{"train_loss": 0.055595096200704575, "global_step": 137625, "epoch": 1546, "lr": 4.904291332454112e-05}
{"train_loss": 0.15608210861682892, "global_step": 137626, "epoch": 1546, "lr": 4.9042333694221305e-05}
{"train_loss": 0.12216559797525406, "global_step": 137627, "epoch": 1546, "lr": 4.904175406403026e-05}
{"train_loss": 0.131597638130188, "global_step": 137628, "epoch": 1546, "lr": 4.904117443396801e-05}
{"train_loss": 0.1634255349636078, "global_step": 137629, "epoch": 1546, "lr": 4.90405948040347e-05}
{"train_loss": 0.151980459690094, "global_step": 137630, "epoch": 1546, "lr": 4.904001517423033e-05}
{"train_loss": 0.0816335454583168, "global_step": 137631, "epoch": 1546, "lr": 4.903943554455505e-05}
{"train_loss": 0.04943310096859932, "global_step": 137632, "epoch": 1546, "lr": 4.903885591500888e-05}
{"train_loss": 0.12011340260505676, "global_step": 137633, "epoch": 1546, "lr": 4.903827628559195e-05}
{"train_loss": 0.10044270008802414, "global_step": 137634, "epoch": 1546, "lr": 4.903769665630429e-05}
{"train_loss": 0.08892843127250671, "global_step": 137635, "epoch": 1546, "lr": 4.9037117027146015e-05}
{"train_loss": 0.044666800647974014, "global_step": 137636, "epoch": 1546, "lr": 4.903653739811718e-05}
{"train_loss": 0.09323716163635254, "global_step": 137637, "epoch": 1546, "lr": 4.9035957769217886e-05}
{"train_loss": 0.07632118463516235, "global_step": 137638, "epoch": 1546, "lr": 4.903537814044817e-05}
{"train_loss": 0.12625634670257568, "global_step": 137639, "epoch": 1546, "lr": 4.9034798511808154e-05}
{"train_loss": 0.11497070640325546, "global_step": 137640, "epoch": 1546, "lr": 4.9034218883297885e-05}
{"train_loss": 0.14539173245429993, "global_step": 137641, "epoch": 1546, "lr": 4.903363925491747e-05}
{"train_loss": 0.15250244736671448, "global_step": 137642, "epoch": 1546, "lr": 4.903305962666695e-05}
{"train_loss": 0.0939880833029747, "global_step": 137643, "epoch": 1546, "lr": 4.9032479998546445e-05}
{"train_loss": 0.10099201649427414, "global_step": 137644, "epoch": 1546, "lr": 4.9031900370556e-05}
{"train_loss": 0.09340598434209824, "global_step": 137645, "epoch": 1546, "lr": 4.9031320742695694e-05}
{"train_loss": 0.13153433799743652, "global_step": 137646, "epoch": 1546, "lr": 4.903074111496564e-05}
{"train_loss": 0.16891752183437347, "global_step": 137647, "epoch": 1546, "lr": 4.903016148736587e-05}
{"train_loss": 0.1374589204788208, "global_step": 137648, "epoch": 1546, "lr": 4.9029581859896495e-05}
{"train_loss": 0.10976782441139221, "global_step": 137649, "epoch": 1546, "lr": 4.9029002232557564e-05}
{"train_loss": 0.11518349498510361, "global_step": 137650, "epoch": 1546, "lr": 4.902842260534919e-05}
{"train_loss": 0.1284501999616623, "global_step": 137651, "epoch": 1546, "lr": 4.902784297827142e-05}
{"train_loss": 0.11804207414388657, "global_step": 137652, "epoch": 1546, "lr": 4.9027263351324367e-05}
{"train_loss": 0.1263131946325302, "global_step": 137653, "epoch": 1546, "lr": 4.9026683724508055e-05}
{"train_loss": 0.10574711859226227, "global_step": 137654, "epoch": 1546, "lr": 4.902610409782262e-05}
{"train_loss": 0.0828084647655487, "global_step": 137655, "epoch": 1546, "lr": 4.902552447126809e-05}
{"train_loss": 0.09016525745391846, "global_step": 137656, "epoch": 1546, "lr": 4.902494484484459e-05}
{"train_loss": 0.1707002818584442, "global_step": 137657, "epoch": 1546, "lr": 4.9024365218552155e-05}
{"train_loss": 0.13375407457351685, "global_step": 137658, "epoch": 1546, "lr": 4.9023785592390905e-05}
{"train_loss": 0.11834658682346344, "global_step": 137659, "epoch": 1546, "lr": 4.902320596636087e-05}
{"train_loss": 0.1335700899362564, "global_step": 137660, "epoch": 1546, "lr": 4.9022626340462174e-05}
{"train_loss": 0.11418076604604721, "global_step": 137661, "epoch": 1546, "lr": 4.9022046714694856e-05}
{"train_loss": 0.04219568520784378, "global_step": 137662, "epoch": 1546, "lr": 4.902146708905902e-05}
{"train_loss": 0.11702848970890045, "global_step": 137663, "epoch": 1546, "lr": 4.9020887463554747e-05}
{"train_loss": 0.11545334756374359, "global_step": 137664, "epoch": 1546, "lr": 4.902030783818208e-05}
{"train_loss": 0.05576616898179054, "global_step": 137665, "epoch": 1546, "lr": 4.901972821294114e-05}
{"train_loss": 0.08759883046150208, "global_step": 137666, "epoch": 1546, "lr": 4.9019148587831966e-05}
{"train_loss": 0.1606927067041397, "global_step": 137667, "epoch": 1546, "lr": 4.901856896285468e-05}
{"train_loss": 0.12493009120225906, "global_step": 137668, "epoch": 1546, "lr": 4.901798933800931e-05}
{"train_loss": 0.14280033111572266, "global_step": 137669, "epoch": 1546, "lr": 4.9017409713295986e-05}
{"train_loss": 0.15174804627895355, "global_step": 137670, "epoch": 1546, "lr": 4.901683008871473e-05}
{"train_loss": 0.0963316261768341, "global_step": 137671, "epoch": 1546, "lr": 4.901625046426569e-05}
{"train_loss": 0.08210411667823792, "global_step": 137672, "epoch": 1546, "lr": 4.9015670839948855e-05}
{"train_loss": 0.08707664161920547, "global_step": 137673, "epoch": 1546, "lr": 4.901509121576439e-05}
{"train_loss": 0.16324888169765472, "global_step": 137674, "epoch": 1546, "lr": 4.901451159171231e-05}
{"train_loss": 0.10920775681734085, "global_step": 137675, "epoch": 1546, "lr": 4.9013931967792734e-05}
{"train_loss": 0.13766542077064514, "global_step": 137676, "epoch": 1546, "lr": 4.9013352344005704e-05}
{"train_loss": 0.13470375537872314, "global_step": 137677, "epoch": 1546, "lr": 4.901277272035133e-05}
{"train_loss": 0.0868121087551117, "global_step": 137678, "epoch": 1546, "lr": 4.901219309682967e-05}
{"train_loss": 0.10717221349477768, "global_step": 137679, "epoch": 1546, "lr": 4.901161347344082e-05}
{"train_loss": 0.09085385501384735, "global_step": 137680, "epoch": 1546, "lr": 4.901103385018483e-05}
{"train_loss": 0.13190536201000214, "global_step": 137681, "epoch": 1546, "lr": 4.90104542270618e-05}
{"train_loss": 0.11834566671861692, "global_step": 137682, "epoch": 1546, "lr": 4.900987460407182e-05, "val_loss": 5.94494104385376}
{"train_loss": 0.1741102784872055, "global_step": 137683, "epoch": 1547, "lr": 4.900929498121493e-05}
{"train_loss": 0.09296251088380814, "global_step": 137684, "epoch": 1547, "lr": 4.900871535849124e-05}
{"train_loss": 0.14254963397979736, "global_step": 137685, "epoch": 1547, "lr": 4.90081357359008e-05}
{"train_loss": 0.10485906898975372, "global_step": 137686, "epoch": 1547, "lr": 4.900755611344373e-05}
{"train_loss": 0.13546520471572876, "global_step": 137687, "epoch": 1547, "lr": 4.900697649112005e-05}
{"train_loss": 0.0704742819070816, "global_step": 137688, "epoch": 1547, "lr": 4.9006396868929895e-05}
{"train_loss": 0.18180188536643982, "global_step": 137689, "epoch": 1547, "lr": 4.90058172468733e-05}
{"train_loss": 0.17023050785064697, "global_step": 137690, "epoch": 1547, "lr": 4.900523762495038e-05}
{"train_loss": 0.11093740165233612, "global_step": 137691, "epoch": 1547, "lr": 4.9004658003161175e-05}
{"train_loss": 0.14807698130607605, "global_step": 137692, "epoch": 1547, "lr": 4.9004078381505795e-05}
{"train_loss": 0.061323534697294235, "global_step": 137693, "epoch": 1547, "lr": 4.9003498759984287e-05}
{"train_loss": 0.07682086527347565, "global_step": 137694, "epoch": 1547, "lr": 4.900291913859677e-05}
{"train_loss": 0.12250880897045135, "global_step": 137695, "epoch": 1547, "lr": 4.900233951734328e-05}
{"train_loss": 0.1455952674150467, "global_step": 137696, "epoch": 1547, "lr": 4.900175989622393e-05}
{"train_loss": 0.14335106313228607, "global_step": 137697, "epoch": 1547, "lr": 4.900118027523876e-05}
{"train_loss": 0.1301448494195938, "global_step": 137698, "epoch": 1547, "lr": 4.900060065438789e-05}
{"train_loss": 0.11312784254550934, "global_step": 137699, "epoch": 1547, "lr": 4.9000021033671365e-05}
{"train_loss": 0.11824416369199753, "global_step": 137700, "epoch": 1547, "lr": 4.899944141308928e-05}
{"train_loss": 0.11482151597738266, "global_step": 137701, "epoch": 1547, "lr": 4.8998861792641705e-05}
{"train_loss": 0.09937911480665207, "global_step": 137702, "epoch": 1547, "lr": 4.8998282172328716e-05}
{"train_loss": 0.10819007456302643, "global_step": 137703, "epoch": 1547, "lr": 4.899770255215041e-05}
{"train_loss": 0.07779712975025177, "global_step": 137704, "epoch": 1547, "lr": 4.8997122932106835e-05}
{"train_loss": 0.07178574055433273, "global_step": 137705, "epoch": 1547, "lr": 4.8996543312198105e-05}
{"train_loss": 0.09539914131164551, "global_step": 137706, "epoch": 1547, "lr": 4.899596369242425e-05}
{"train_loss": 0.12697933614253998, "global_step": 137707, "epoch": 1547, "lr": 4.8995384072785404e-05}
{"train_loss": 0.15281708538532257, "global_step": 137708, "epoch": 1547, "lr": 4.899480445328159e-05}
{"train_loss": 0.10528677701950073, "global_step": 137709, "epoch": 1547, "lr": 4.899422483391294e-05}
{"train_loss": 0.08154547214508057, "global_step": 137710, "epoch": 1547, "lr": 4.899364521467947e-05}
{"train_loss": 0.17067161202430725, "global_step": 137711, "epoch": 1547, "lr": 4.8993065595581325e-05}
{"train_loss": 0.19015656411647797, "global_step": 137712, "epoch": 1547, "lr": 4.899248597661853e-05}
{"train_loss": 0.09794475883245468, "global_step": 137713, "epoch": 1547, "lr": 4.899190635779119e-05}
{"train_loss": 0.14037340879440308, "global_step": 137714, "epoch": 1547, "lr": 4.899132673909937e-05}
{"train_loss": 0.11685917526483536, "global_step": 137715, "epoch": 1547, "lr": 4.899074712054317e-05}
{"train_loss": 0.06183282658457756, "global_step": 137716, "epoch": 1547, "lr": 4.899016750212262e-05}
{"train_loss": 0.09525833278894424, "global_step": 137717, "epoch": 1547, "lr": 4.898958788383786e-05}
{"train_loss": 0.12756651639938354, "global_step": 137718, "epoch": 1547, "lr": 4.898900826568892e-05}
{"train_loss": 0.09027820825576782, "global_step": 137719, "epoch": 1547, "lr": 4.898842864767591e-05}
{"train_loss": 0.14357344806194305, "global_step": 137720, "epoch": 1547, "lr": 4.898784902979888e-05}
{"train_loss": 0.12513409554958344, "global_step": 137721, "epoch": 1547, "lr": 4.8987269412057933e-05}
{"train_loss": 0.06359865516424179, "global_step": 137722, "epoch": 1547, "lr": 4.898668979445314e-05}
{"train_loss": 0.12938129901885986, "global_step": 137723, "epoch": 1547, "lr": 4.8986110176984555e-05}
{"train_loss": 0.0819549486041069, "global_step": 137724, "epoch": 1547, "lr": 4.8985530559652296e-05}
{"train_loss": 0.09210849553346634, "global_step": 137725, "epoch": 1547, "lr": 4.8984950942456395e-05}
{"train_loss": 0.058254387229681015, "global_step": 137726, "epoch": 1547, "lr": 4.898437132539698e-05}
{"train_loss": 0.15884321928024292, "global_step": 137727, "epoch": 1547, "lr": 4.8983791708474084e-05}
{"train_loss": 0.10630065947771072, "global_step": 137728, "epoch": 1547, "lr": 4.898321209168782e-05}
{"train_loss": 0.13764363527297974, "global_step": 137729, "epoch": 1547, "lr": 4.898263247503824e-05}
{"train_loss": 0.14496822655200958, "global_step": 137730, "epoch": 1547, "lr": 4.898205285852544e-05}
{"train_loss": 0.0899115800857544, "global_step": 137731, "epoch": 1547, "lr": 4.898147324214948e-05}
{"train_loss": 0.12755580246448517, "global_step": 137732, "epoch": 1547, "lr": 4.898089362591047e-05}
{"train_loss": 0.07299663871526718, "global_step": 137733, "epoch": 1547, "lr": 4.898031400980844e-05}
{"train_loss": 0.10322029143571854, "global_step": 137734, "epoch": 1547, "lr": 4.897973439384349e-05}
{"train_loss": 0.14652006328105927, "global_step": 137735, "epoch": 1547, "lr": 4.8979154778015735e-05}
{"train_loss": 0.08481836318969727, "global_step": 137736, "epoch": 1547, "lr": 4.89785751623252e-05}
{"train_loss": 0.1863081157207489, "global_step": 137737, "epoch": 1547, "lr": 4.897799554677199e-05}
{"train_loss": 0.07745267450809479, "global_step": 137738, "epoch": 1547, "lr": 4.8977415931356165e-05}
{"train_loss": 0.06572119891643524, "global_step": 137739, "epoch": 1547, "lr": 4.8976836316077825e-05}
{"train_loss": 0.13188643753528595, "global_step": 137740, "epoch": 1547, "lr": 4.897625670093704e-05}
{"train_loss": 0.10784585773944855, "global_step": 137741, "epoch": 1547, "lr": 4.897567708593389e-05}
{"train_loss": 0.12914547324180603, "global_step": 137742, "epoch": 1547, "lr": 4.897509747106842e-05}
{"train_loss": 0.07732291519641876, "global_step": 137743, "epoch": 1547, "lr": 4.8974517856340765e-05}
{"train_loss": 0.14377504587173462, "global_step": 137744, "epoch": 1547, "lr": 4.897393824175095e-05}
{"train_loss": 0.08243368566036224, "global_step": 137745, "epoch": 1547, "lr": 4.8973358627299095e-05}
{"train_loss": 0.07563511282205582, "global_step": 137746, "epoch": 1547, "lr": 4.897277901298524e-05}
{"train_loss": 0.05767901986837387, "global_step": 137747, "epoch": 1547, "lr": 4.897219939880951e-05}
{"train_loss": 0.10015831142663956, "global_step": 137748, "epoch": 1547, "lr": 4.897161978477193e-05}
{"train_loss": 0.08945222198963165, "global_step": 137749, "epoch": 1547, "lr": 4.8971040170872626e-05}
{"train_loss": 0.17066052556037903, "global_step": 137750, "epoch": 1547, "lr": 4.897046055711163e-05}
{"train_loss": 0.15934407711029053, "global_step": 137751, "epoch": 1547, "lr": 4.896988094348906e-05}
{"train_loss": 0.13300774991512299, "global_step": 137752, "epoch": 1547, "lr": 4.896930133000496e-05}
{"train_loss": 0.09830476343631744, "global_step": 137753, "epoch": 1547, "lr": 4.896872171665942e-05}
{"train_loss": 0.1267620027065277, "global_step": 137754, "epoch": 1547, "lr": 4.896814210345255e-05}
{"train_loss": 0.10483640432357788, "global_step": 137755, "epoch": 1547, "lr": 4.896756249038438e-05}
{"train_loss": 0.12451063096523285, "global_step": 137756, "epoch": 1547, "lr": 4.896698287745502e-05}
{"train_loss": 0.1240142434835434, "global_step": 137757, "epoch": 1547, "lr": 4.896640326466453e-05}
{"train_loss": 0.05562290549278259, "global_step": 137758, "epoch": 1547, "lr": 4.8965823652013004e-05}
{"train_loss": 0.07543404400348663, "global_step": 137759, "epoch": 1547, "lr": 4.8965244039500504e-05}
{"train_loss": 0.10214149206876755, "global_step": 137760, "epoch": 1547, "lr": 4.896466442712713e-05}
{"train_loss": 0.13386014103889465, "global_step": 137761, "epoch": 1547, "lr": 4.8964084814892916e-05}
{"train_loss": 0.06382585316896439, "global_step": 137762, "epoch": 1547, "lr": 4.896350520279799e-05}
{"train_loss": 0.08449340611696243, "global_step": 137763, "epoch": 1547, "lr": 4.8962925590842384e-05}
{"train_loss": 0.13294366002082825, "global_step": 137764, "epoch": 1547, "lr": 4.8962345979026234e-05}
{"train_loss": 0.16498976945877075, "global_step": 137765, "epoch": 1547, "lr": 4.8961766367349546e-05}
{"train_loss": 0.1100107878446579, "global_step": 137766, "epoch": 1547, "lr": 4.896118675581247e-05}
{"train_loss": 0.11078154295682907, "global_step": 137767, "epoch": 1547, "lr": 4.8960607144415024e-05}
{"train_loss": 0.11962074041366577, "global_step": 137768, "epoch": 1547, "lr": 4.896002753315733e-05}
{"train_loss": 0.08526482433080673, "global_step": 137769, "epoch": 1547, "lr": 4.895944792203943e-05}
{"train_loss": 0.10175899416208267, "global_step": 137770, "epoch": 1547, "lr": 4.895886831106144e-05}
{"train_loss": 0.11324096810114517, "global_step": 137771, "epoch": 1547, "lr": 4.895828870022339e-05, "val_loss": 6.120667934417725}
{"train_loss": 0.07701045274734497, "global_step": 137772, "epoch": 1548, "lr": 4.89577090895254e-05}
{"train_loss": 0.09293042123317719, "global_step": 137773, "epoch": 1548, "lr": 4.8957129478967545e-05}
{"train_loss": 0.09855177253484726, "global_step": 137774, "epoch": 1548, "lr": 4.895654986854988e-05}
{"train_loss": 0.13110798597335815, "global_step": 137775, "epoch": 1548, "lr": 4.895597025827249e-05}
{"train_loss": 0.15898148715496063, "global_step": 137776, "epoch": 1548, "lr": 4.895539064813547e-05}
{"train_loss": 0.06503550708293915, "global_step": 137777, "epoch": 1548, "lr": 4.8954811038138876e-05}
{"train_loss": 0.13935735821723938, "global_step": 137778, "epoch": 1548, "lr": 4.89542314282828e-05}
{"train_loss": 0.08358030021190643, "global_step": 137779, "epoch": 1548, "lr": 4.895365181856732e-05}
{"train_loss": 0.11332022398710251, "global_step": 137780, "epoch": 1548, "lr": 4.895307220899249e-05}
{"train_loss": 0.07605840265750885, "global_step": 137781, "epoch": 1548, "lr": 4.895249259955842e-05}
{"train_loss": 0.07919178903102875, "global_step": 137782, "epoch": 1548, "lr": 4.895191299026517e-05}
{"train_loss": 0.18989969789981842, "global_step": 137783, "epoch": 1548, "lr": 4.8951333381112836e-05}
{"train_loss": 0.1433708667755127, "global_step": 137784, "epoch": 1548, "lr": 4.895075377210146e-05}
{"train_loss": 0.13972891867160797, "global_step": 137785, "epoch": 1548, "lr": 4.8950174163231163e-05}
{"train_loss": 0.11526927351951599, "global_step": 137786, "epoch": 1548, "lr": 4.894959455450198e-05}
{"train_loss": 0.1380213052034378, "global_step": 137787, "epoch": 1548, "lr": 4.8949014945914035e-05}
{"train_loss": 0.13821756839752197, "global_step": 137788, "epoch": 1548, "lr": 4.8948435337467355e-05}
{"train_loss": 0.13960106670856476, "global_step": 137789, "epoch": 1548, "lr": 4.894785572916205e-05}
{"train_loss": 0.10489784926176071, "global_step": 137790, "epoch": 1548, "lr": 4.894727612099822e-05}
{"train_loss": 0.11876090615987778, "global_step": 137791, "epoch": 1548, "lr": 4.8946696512975885e-05}
{"train_loss": 0.042002011090517044, "global_step": 137792, "epoch": 1548, "lr": 4.894611690509518e-05}
{"train_loss": 0.10738083720207214, "global_step": 137793, "epoch": 1548, "lr": 4.894553729735615e-05}
{"train_loss": 0.11976048350334167, "global_step": 137794, "epoch": 1548, "lr": 4.894495768975887e-05}
{"train_loss": 0.09258460998535156, "global_step": 137795, "epoch": 1548, "lr": 4.894437808230343e-05}
{"train_loss": 0.10269935429096222, "global_step": 137796, "epoch": 1548, "lr": 4.894379847498991e-05}
{"train_loss": 0.07004915922880173, "global_step": 137797, "epoch": 1548, "lr": 4.894321886781839e-05}
{"train_loss": 0.09799449890851974, "global_step": 137798, "epoch": 1548, "lr": 4.894263926078894e-05}
{"train_loss": 0.08345028012990952, "global_step": 137799, "epoch": 1548, "lr": 4.894205965390162e-05}
{"train_loss": 0.10042396932840347, "global_step": 137800, "epoch": 1548, "lr": 4.8941480047156544e-05}
{"train_loss": 0.12052445113658905, "global_step": 137801, "epoch": 1548, "lr": 4.894090044055376e-05}
{"train_loss": 0.10487601161003113, "global_step": 137802, "epoch": 1548, "lr": 4.894032083409338e-05}
{"train_loss": 0.13646788895130157, "global_step": 137803, "epoch": 1548, "lr": 4.893974122777544e-05}
{"train_loss": 0.09206598252058029, "global_step": 137804, "epoch": 1548, "lr": 4.8939161621600055e-05}
{"train_loss": 0.06663040071725845, "global_step": 137805, "epoch": 1548, "lr": 4.893858201556727e-05}
{"train_loss": 0.07928077131509781, "global_step": 137806, "epoch": 1548, "lr": 4.89380024096772e-05}
{"train_loss": 0.09034226834774017, "global_step": 137807, "epoch": 1548, "lr": 4.893742280392988e-05}
{"train_loss": 0.11012744903564453, "global_step": 137808, "epoch": 1548, "lr": 4.893684319832541e-05}
{"train_loss": 0.13969206809997559, "global_step": 137809, "epoch": 1548, "lr": 4.893626359286389e-05}
{"train_loss": 0.1794097125530243, "global_step": 137810, "epoch": 1548, "lr": 4.893568398754535e-05}
{"train_loss": 0.07538893073797226, "global_step": 137811, "epoch": 1548, "lr": 4.8935104382369927e-05}
{"train_loss": 0.08021952956914902, "global_step": 137812, "epoch": 1548, "lr": 4.893452477733765e-05}
{"train_loss": 0.06413031369447708, "global_step": 137813, "epoch": 1548, "lr": 4.8933945172448616e-05}
{"train_loss": 0.06641987711191177, "global_step": 137814, "epoch": 1548, "lr": 4.893336556770289e-05}
{"train_loss": 0.09185033291578293, "global_step": 137815, "epoch": 1548, "lr": 4.893278596310058e-05}
{"train_loss": 0.07905439287424088, "global_step": 137816, "epoch": 1548, "lr": 4.893220635864173e-05}
{"train_loss": 0.16061098873615265, "global_step": 137817, "epoch": 1548, "lr": 4.8931626754326427e-05}
{"train_loss": 0.12689296901226044, "global_step": 137818, "epoch": 1548, "lr": 4.893104715015476e-05}
{"train_loss": 0.0817747637629509, "global_step": 137819, "epoch": 1548, "lr": 4.8930467546126815e-05}
{"train_loss": 0.10872932523488998, "global_step": 137820, "epoch": 1548, "lr": 4.892988794224263e-05}
{"train_loss": 0.13456694781780243, "global_step": 137821, "epoch": 1548, "lr": 4.8929308338502326e-05}
{"train_loss": 0.16605527698993683, "global_step": 137822, "epoch": 1548, "lr": 4.8928728734905946e-05}
{"train_loss": 0.0729171633720398, "global_step": 137823, "epoch": 1548, "lr": 4.892814913145361e-05}
{"train_loss": 0.18004962801933289, "global_step": 137824, "epoch": 1548, "lr": 4.892756952814534e-05}
{"train_loss": 0.104888416826725, "global_step": 137825, "epoch": 1548, "lr": 4.892698992498126e-05}
{"train_loss": 0.15208014845848083, "global_step": 137826, "epoch": 1548, "lr": 4.8926410321961445e-05}
{"train_loss": 0.12498262524604797, "global_step": 137827, "epoch": 1548, "lr": 4.8925830719085944e-05}
{"train_loss": 0.09243019670248032, "global_step": 137828, "epoch": 1548, "lr": 4.892525111635487e-05}
{"train_loss": 0.14290709793567657, "global_step": 137829, "epoch": 1548, "lr": 4.892467151376826e-05}
{"train_loss": 0.12879331409931183, "global_step": 137830, "epoch": 1548, "lr": 4.892409191132624e-05}
{"train_loss": 0.09919489920139313, "global_step": 137831, "epoch": 1548, "lr": 4.8923512309028844e-05}
{"train_loss": 0.17257972061634064, "global_step": 137832, "epoch": 1548, "lr": 4.892293270687619e-05}
{"train_loss": 0.146784245967865, "global_step": 137833, "epoch": 1548, "lr": 4.892235310486831e-05}
{"train_loss": 0.1343412548303604, "global_step": 137834, "epoch": 1548, "lr": 4.892177350300532e-05}
{"train_loss": 0.08723843842744827, "global_step": 137835, "epoch": 1548, "lr": 4.8921193901287274e-05}
{"train_loss": 0.14492957293987274, "global_step": 137836, "epoch": 1548, "lr": 4.892061429971427e-05}
{"train_loss": 0.08533000946044922, "global_step": 137837, "epoch": 1548, "lr": 4.892003469828638e-05}
{"train_loss": 0.07689890265464783, "global_step": 137838, "epoch": 1548, "lr": 4.891945509700368e-05}
{"train_loss": 0.15047767758369446, "global_step": 137839, "epoch": 1548, "lr": 4.891887549586622e-05}
{"train_loss": 0.10459087789058685, "global_step": 137840, "epoch": 1548, "lr": 4.8918295894874136e-05}
{"train_loss": 0.17748790979385376, "global_step": 137841, "epoch": 1548, "lr": 4.891771629402745e-05}
{"train_loss": 0.08967986702919006, "global_step": 137842, "epoch": 1548, "lr": 4.891713669332628e-05}
{"train_loss": 0.08487122505903244, "global_step": 137843, "epoch": 1548, "lr": 4.8916557092770675e-05}
{"train_loss": 0.17226381599903107, "global_step": 137844, "epoch": 1548, "lr": 4.891597749236072e-05}
{"train_loss": 0.09990847110748291, "global_step": 137845, "epoch": 1548, "lr": 4.891539789209652e-05}
{"train_loss": 0.11502867937088013, "global_step": 137846, "epoch": 1548, "lr": 4.8914818291978116e-05}
{"train_loss": 0.09588533639907837, "global_step": 137847, "epoch": 1548, "lr": 4.8914238692005613e-05}
{"train_loss": 0.12349535524845123, "global_step": 137848, "epoch": 1548, "lr": 4.8913659092179065e-05}
{"train_loss": 0.11167149990797043, "global_step": 137849, "epoch": 1548, "lr": 4.891307949249857e-05}
{"train_loss": 0.11075016856193542, "global_step": 137850, "epoch": 1548, "lr": 4.891249989296419e-05}
{"train_loss": 0.14326268434524536, "global_step": 137851, "epoch": 1548, "lr": 4.891192029357604e-05}
{"train_loss": 0.0825711265206337, "global_step": 137852, "epoch": 1548, "lr": 4.891134069433414e-05}
{"train_loss": 0.16036801040172577, "global_step": 137853, "epoch": 1548, "lr": 4.891076109523861e-05}
{"train_loss": 0.08969906717538834, "global_step": 137854, "epoch": 1548, "lr": 4.8910181496289505e-05}
{"train_loss": 0.17441219091415405, "global_step": 137855, "epoch": 1548, "lr": 4.890960189748692e-05}
{"train_loss": 0.06841305643320084, "global_step": 137856, "epoch": 1548, "lr": 4.890902229883092e-05}
{"train_loss": 0.11571173369884491, "global_step": 137857, "epoch": 1548, "lr": 4.8908442700321605e-05}
{"train_loss": 0.2360047847032547, "global_step": 137858, "epoch": 1548, "lr": 4.8907863101959015e-05}
{"train_loss": 0.11039983481168747, "global_step": 137859, "epoch": 1548, "lr": 4.890728350374326e-05}
{"train_loss": 0.11439272991559479, "global_step": 137860, "epoch": 1548, "lr": 4.89067039056744e-05, "val_loss": 5.928466320037842}
{"train_loss": 0.1635458618402481, "global_step": 137861, "epoch": 1549, "lr": 4.890612430775251e-05}
{"train_loss": 0.10566457360982895, "global_step": 137862, "epoch": 1549, "lr": 4.8905544709977705e-05}
{"train_loss": 0.1686072200536728, "global_step": 137863, "epoch": 1549, "lr": 4.8904965112350014e-05}
{"train_loss": 0.08670914173126221, "global_step": 137864, "epoch": 1549, "lr": 4.890438551486956e-05}
{"train_loss": 0.09969820827245712, "global_step": 137865, "epoch": 1549, "lr": 4.8903805917536376e-05}
{"train_loss": 0.05183183401823044, "global_step": 137866, "epoch": 1549, "lr": 4.890322632035058e-05}
{"train_loss": 0.18897980451583862, "global_step": 137867, "epoch": 1549, "lr": 4.890264672331222e-05}
{"train_loss": 0.10450878739356995, "global_step": 137868, "epoch": 1549, "lr": 4.89020671264214e-05}
{"train_loss": 0.1494661569595337, "global_step": 137869, "epoch": 1549, "lr": 4.890148752967816e-05}
{"train_loss": 0.16009338200092316, "global_step": 137870, "epoch": 1549, "lr": 4.890090793308262e-05}
{"train_loss": 0.09600220620632172, "global_step": 137871, "epoch": 1549, "lr": 4.890032833663484e-05}
{"train_loss": 0.07299266010522842, "global_step": 137872, "epoch": 1549, "lr": 4.889974874033489e-05}
{"train_loss": 0.11963990330696106, "global_step": 137873, "epoch": 1549, "lr": 4.889916914418286e-05}
{"train_loss": 0.13338617980480194, "global_step": 137874, "epoch": 1549, "lr": 4.889858954817882e-05}
{"train_loss": 0.14102217555046082, "global_step": 137875, "epoch": 1549, "lr": 4.8898009952322846e-05}
{"train_loss": 0.1279902458190918, "global_step": 137876, "epoch": 1549, "lr": 4.8897430356615046e-05}
{"train_loss": 0.07543329149484634, "global_step": 137877, "epoch": 1549, "lr": 4.889685076105544e-05}
{"train_loss": 0.14827494323253632, "global_step": 137878, "epoch": 1549, "lr": 4.8896271165644166e-05}
{"train_loss": 0.12797218561172485, "global_step": 137879, "epoch": 1549, "lr": 4.889569157038125e-05}
{"train_loss": 0.13307420909404755, "global_step": 137880, "epoch": 1549, "lr": 4.889511197526679e-05}
{"train_loss": 0.098695307970047, "global_step": 137881, "epoch": 1549, "lr": 4.88945323803009e-05}
{"train_loss": 0.08236435800790787, "global_step": 137882, "epoch": 1549, "lr": 4.88939527854836e-05}
{"train_loss": 0.08268111199140549, "global_step": 137883, "epoch": 1549, "lr": 4.8893373190815016e-05}
{"train_loss": 0.14117313921451569, "global_step": 137884, "epoch": 1549, "lr": 4.8892793596295184e-05}
{"train_loss": 0.10775281488895416, "global_step": 137885, "epoch": 1549, "lr": 4.8892214001924226e-05}
{"train_loss": 0.11651753634214401, "global_step": 137886, "epoch": 1549, "lr": 4.8891634407702176e-05}
{"train_loss": 0.13900721073150635, "global_step": 137887, "epoch": 1549, "lr": 4.8891054813629144e-05}
{"train_loss": 0.115746408700943, "global_step": 137888, "epoch": 1549, "lr": 4.889047521970518e-05}
{"train_loss": 0.1283080130815506, "global_step": 137889, "epoch": 1549, "lr": 4.8889895625930406e-05}
{"train_loss": 0.14821815490722656, "global_step": 137890, "epoch": 1549, "lr": 4.888931603230486e-05}
{"train_loss": 0.18945389986038208, "global_step": 137891, "epoch": 1549, "lr": 4.8888736438828624e-05}
{"train_loss": 0.10648556053638458, "global_step": 137892, "epoch": 1549, "lr": 4.8888156845501786e-05}
{"train_loss": 0.1327202022075653, "global_step": 137893, "epoch": 1549, "lr": 4.8887577252324427e-05}
{"train_loss": 0.10569106787443161, "global_step": 137894, "epoch": 1549, "lr": 4.888699765929661e-05}
{"train_loss": 0.17436297237873077, "global_step": 137895, "epoch": 1549, "lr": 4.8886418066418445e-05}
{"train_loss": 0.11672208458185196, "global_step": 137896, "epoch": 1549, "lr": 4.888583847368996e-05}
{"train_loss": 0.15943904221057892, "global_step": 137897, "epoch": 1549, "lr": 4.888525888111128e-05}
{"train_loss": 0.18091100454330444, "global_step": 137898, "epoch": 1549, "lr": 4.888467928868245e-05}
{"train_loss": 0.12835492193698883, "global_step": 137899, "epoch": 1549, "lr": 4.888409969640355e-05}
{"train_loss": 0.11958061158657074, "global_step": 137900, "epoch": 1549, "lr": 4.88835201042747e-05}
{"train_loss": 0.11070471256971359, "global_step": 137901, "epoch": 1549, "lr": 4.888294051229592e-05}
{"train_loss": 0.16656213998794556, "global_step": 137902, "epoch": 1549, "lr": 4.888236092046734e-05}
{"train_loss": 0.1372554451227188, "global_step": 137903, "epoch": 1549, "lr": 4.888178132878899e-05}
{"train_loss": 0.14831943809986115, "global_step": 137904, "epoch": 1549, "lr": 4.888120173726099e-05}
{"train_loss": 0.11245277523994446, "global_step": 137905, "epoch": 1549, "lr": 4.888062214588338e-05}
{"train_loss": 0.05826302990317345, "global_step": 137906, "epoch": 1549, "lr": 4.888004255465628e-05}
{"train_loss": 0.2383904755115509, "global_step": 137907, "epoch": 1549, "lr": 4.887946296357972e-05}
{"train_loss": 0.07347629964351654, "global_step": 137908, "epoch": 1549, "lr": 4.887888337265383e-05}
{"train_loss": 0.09670910984277725, "global_step": 137909, "epoch": 1549, "lr": 4.887830378187864e-05}
{"train_loss": 0.12408103793859482, "global_step": 137910, "epoch": 1549, "lr": 4.887772419125425e-05}
{"train_loss": 0.19233030080795288, "global_step": 137911, "epoch": 1549, "lr": 4.8877144600780744e-05}
{"train_loss": 0.1479090303182602, "global_step": 137912, "epoch": 1549, "lr": 4.887656501045819e-05}
{"train_loss": 0.14147412776947021, "global_step": 137913, "epoch": 1549, "lr": 4.8875985420286666e-05}
{"train_loss": 0.17612074315547943, "global_step": 137914, "epoch": 1549, "lr": 4.8875405830266265e-05}
{"train_loss": 0.1797924041748047, "global_step": 137915, "epoch": 1549, "lr": 4.887482624039702e-05}
{"train_loss": 0.07442481815814972, "global_step": 137916, "epoch": 1549, "lr": 4.8874246650679054e-05}
{"train_loss": 0.10161648690700531, "global_step": 137917, "epoch": 1549, "lr": 4.887366706111245e-05}
{"train_loss": 0.17338785529136658, "global_step": 137918, "epoch": 1549, "lr": 4.887308747169724e-05}
{"train_loss": 0.10221915692090988, "global_step": 137919, "epoch": 1549, "lr": 4.887250788243356e-05}
{"train_loss": 0.12199420481920242, "global_step": 137920, "epoch": 1549, "lr": 4.887192829332143e-05}
{"train_loss": 0.1396818310022354, "global_step": 137921, "epoch": 1549, "lr": 4.887134870436098e-05}
{"train_loss": 0.14741577208042145, "global_step": 137922, "epoch": 1549, "lr": 4.8870769115552236e-05}
{"train_loss": 0.09848103672266006, "global_step": 137923, "epoch": 1549, "lr": 4.8870189526895335e-05}
{"train_loss": 0.0930308923125267, "global_step": 137924, "epoch": 1549, "lr": 4.88696099383903e-05}
{"train_loss": 0.06639616191387177, "global_step": 137925, "epoch": 1549, "lr": 4.886903035003725e-05}
{"train_loss": 0.1341116726398468, "global_step": 137926, "epoch": 1549, "lr": 4.886845076183622e-05}
{"train_loss": 0.17374348640441895, "global_step": 137927, "epoch": 1549, "lr": 4.8867871173787336e-05}
{"train_loss": 0.13465338945388794, "global_step": 137928, "epoch": 1549, "lr": 4.886729158589064e-05}
{"train_loss": 0.1098221093416214, "global_step": 137929, "epoch": 1549, "lr": 4.886671199814623e-05}
{"train_loss": 0.17915242910385132, "global_step": 137930, "epoch": 1549, "lr": 4.8866132410554176e-05}
{"train_loss": 0.09453637152910233, "global_step": 137931, "epoch": 1549, "lr": 4.886555282311456e-05}
{"train_loss": 0.12770670652389526, "global_step": 137932, "epoch": 1549, "lr": 4.886497323582745e-05}
{"train_loss": 0.11985190957784653, "global_step": 137933, "epoch": 1549, "lr": 4.886439364869293e-05}
{"train_loss": 0.13133832812309265, "global_step": 137934, "epoch": 1549, "lr": 4.8863814061711076e-05}
{"train_loss": 0.08996772766113281, "global_step": 137935, "epoch": 1549, "lr": 4.8863234474881964e-05}
{"train_loss": 0.09268930554389954, "global_step": 137936, "epoch": 1549, "lr": 4.88626548882057e-05}
{"train_loss": 0.13913056254386902, "global_step": 137937, "epoch": 1549, "lr": 4.886207530168231e-05}
{"train_loss": 0.16314968466758728, "global_step": 137938, "epoch": 1549, "lr": 4.8861495715311924e-05}
{"train_loss": 0.12178713083267212, "global_step": 137939, "epoch": 1549, "lr": 4.886091612909457e-05}
{"train_loss": 0.07184416800737381, "global_step": 137940, "epoch": 1549, "lr": 4.8860336543030385e-05}
{"train_loss": 0.09287040680646896, "global_step": 137941, "epoch": 1549, "lr": 4.885975695711938e-05}
{"train_loss": 0.13722100853919983, "global_step": 137942, "epoch": 1549, "lr": 4.885917737136169e-05}
{"train_loss": 0.07368457317352295, "global_step": 137943, "epoch": 1549, "lr": 4.8858597785757356e-05}
{"train_loss": 0.06961255520582199, "global_step": 137944, "epoch": 1549, "lr": 4.885801820030649e-05}
{"train_loss": 0.16564854979515076, "global_step": 137945, "epoch": 1549, "lr": 4.885743861500912e-05}
{"train_loss": 0.10494956374168396, "global_step": 137946, "epoch": 1549, "lr": 4.885685902986538e-05}
{"train_loss": 0.12739430367946625, "global_step": 137947, "epoch": 1549, "lr": 4.88562794448753e-05}
{"train_loss": 0.1815759390592575, "global_step": 137948, "epoch": 1549, "lr": 4.8855699860039e-05}
{"train_loss": 0.12573696367358894, "global_step": 137949, "epoch": 1549, "lr": 4.885512027535653e-05, "val_loss": 5.741920471191406}
{"train_loss": 0.14595243334770203, "global_step": 137950, "epoch": 1550, "lr": 4.885454069082797e-05}
{"train_loss": 0.14974181354045868, "global_step": 137951, "epoch": 1550, "lr": 4.8853961106453403e-05}
{"train_loss": 0.11253450810909271, "global_step": 137952, "epoch": 1550, "lr": 4.88533815222329e-05}
{"train_loss": 0.1263696402311325, "global_step": 137953, "epoch": 1550, "lr": 4.885280193816657e-05}
{"train_loss": 0.0740509182214737, "global_step": 137954, "epoch": 1550, "lr": 4.8852222354254433e-05}
{"train_loss": 0.08936983346939087, "global_step": 137955, "epoch": 1550, "lr": 4.885164277049663e-05}
{"train_loss": 0.10039160400629044, "global_step": 137956, "epoch": 1550, "lr": 4.8851063186893194e-05}
{"train_loss": 0.17990393936634064, "global_step": 137957, "epoch": 1550, "lr": 4.885048360344423e-05}
{"train_loss": 0.16481956839561462, "global_step": 137958, "epoch": 1550, "lr": 4.884990402014979e-05}
{"train_loss": 0.1089002788066864, "global_step": 137959, "epoch": 1550, "lr": 4.8849324437009985e-05}
{"train_loss": 0.138384148478508, "global_step": 137960, "epoch": 1550, "lr": 4.884874485402485e-05}
{"train_loss": 0.1275549829006195, "global_step": 137961, "epoch": 1550, "lr": 4.8848165271194514e-05}
{"train_loss": 0.13161113858222961, "global_step": 137962, "epoch": 1550, "lr": 4.884758568851901e-05}
{"train_loss": 0.07459480315446854, "global_step": 137963, "epoch": 1550, "lr": 4.8847006105998447e-05}
{"train_loss": 0.07982693612575531, "global_step": 137964, "epoch": 1550, "lr": 4.8846426523632866e-05}
{"train_loss": 0.07043250650167465, "global_step": 137965, "epoch": 1550, "lr": 4.8845846941422394e-05}
{"train_loss": 0.1700892150402069, "global_step": 137966, "epoch": 1550, "lr": 4.884526735936707e-05}
{"train_loss": 0.08488745242357254, "global_step": 137967, "epoch": 1550, "lr": 4.8844687777467e-05}
{"train_loss": 0.09308808296918869, "global_step": 137968, "epoch": 1550, "lr": 4.884410819572224e-05}
{"train_loss": 0.13799215853214264, "global_step": 137969, "epoch": 1550, "lr": 4.8843528614132874e-05}
{"train_loss": 0.1594884693622589, "global_step": 137970, "epoch": 1550, "lr": 4.8842949032698975e-05}
{"train_loss": 0.07059777528047562, "global_step": 137971, "epoch": 1550, "lr": 4.8842369451420626e-05}
{"train_loss": 0.1440802365541458, "global_step": 137972, "epoch": 1550, "lr": 4.884178987029793e-05}
{"train_loss": 0.1381368339061737, "global_step": 137973, "epoch": 1550, "lr": 4.8841210289330907e-05}
{"train_loss": 0.1933453232049942, "global_step": 137974, "epoch": 1550, "lr": 4.88406307085197e-05}
{"train_loss": 0.10173629969358444, "global_step": 137975, "epoch": 1550, "lr": 4.884005112786433e-05}
{"train_loss": 0.13010312616825104, "global_step": 137976, "epoch": 1550, "lr": 4.883947154736493e-05}
{"train_loss": 0.09633053094148636, "global_step": 137977, "epoch": 1550, "lr": 4.883889196702152e-05}
{"train_loss": 0.08216879516839981, "global_step": 137978, "epoch": 1550, "lr": 4.883831238683424e-05}
{"train_loss": 0.09864569455385208, "global_step": 137979, "epoch": 1550, "lr": 4.883773280680309e-05}
{"train_loss": 0.12063208222389221, "global_step": 137980, "epoch": 1550, "lr": 4.8837153226928236e-05}
{"train_loss": 0.1115100085735321, "global_step": 137981, "epoch": 1550, "lr": 4.883657364720968e-05}
{"train_loss": 0.13891737163066864, "global_step": 137982, "epoch": 1550, "lr": 4.883599406764756e-05}
{"train_loss": 0.13229604065418243, "global_step": 137983, "epoch": 1550, "lr": 4.88354144882419e-05}
{"train_loss": 0.07758354395627975, "global_step": 137984, "epoch": 1550, "lr": 4.8834834908992826e-05}
{"train_loss": 0.1605517864227295, "global_step": 137985, "epoch": 1550, "lr": 4.883425532990038e-05}
{"train_loss": 0.10847816616296768, "global_step": 137986, "epoch": 1550, "lr": 4.883367575096467e-05}
{"train_loss": 0.1278683841228485, "global_step": 137987, "epoch": 1550, "lr": 4.8833096172185746e-05}
{"train_loss": 0.1199985221028328, "global_step": 137988, "epoch": 1550, "lr": 4.8832516593563685e-05}
{"train_loss": 0.04413393512368202, "global_step": 137989, "epoch": 1550, "lr": 4.88319370150986e-05}
{"train_loss": 0.04891781881451607, "global_step": 137990, "epoch": 1550, "lr": 4.8831357436790534e-05}
{"train_loss": 0.06877170503139496, "global_step": 137991, "epoch": 1550, "lr": 4.88307778586396e-05}
{"train_loss": 0.14691922068595886, "global_step": 137992, "epoch": 1550, "lr": 4.883019828064582e-05}
{"train_loss": 0.07874640077352524, "global_step": 137993, "epoch": 1550, "lr": 4.882961870280934e-05}
{"train_loss": 0.16010509431362152, "global_step": 137994, "epoch": 1550, "lr": 4.8829039125130174e-05}
{"train_loss": 0.1326500028371811, "global_step": 137995, "epoch": 1550, "lr": 4.882845954760845e-05}
{"train_loss": 0.09720483422279358, "global_step": 137996, "epoch": 1550, "lr": 4.8827879970244206e-05}
{"train_loss": 0.08624809980392456, "global_step": 137997, "epoch": 1550, "lr": 4.8827300393037564e-05}
{"train_loss": 0.19263838231563568, "global_step": 137998, "epoch": 1550, "lr": 4.8826720815988545e-05}
{"train_loss": 0.051712073385715485, "global_step": 137999, "epoch": 1550, "lr": 4.882614123909729e-05}
{"train_loss": 0.12781848013401031, "global_step": 138000, "epoch": 1550, "lr": 4.882556166236382e-05}
{"train_loss": 0.10866294801235199, "global_step": 138001, "epoch": 1550, "lr": 4.882498208578827e-05}
{"train_loss": 0.14453235268592834, "global_step": 138002, "epoch": 1550, "lr": 4.8824402509370654e-05}
{"train_loss": 0.09919067472219467, "global_step": 138003, "epoch": 1550, "lr": 4.882382293311111e-05}
{"train_loss": 0.05209684744477272, "global_step": 138004, "epoch": 1550, "lr": 4.882324335700967e-05}
{"train_loss": 0.15729577839374542, "global_step": 138005, "epoch": 1550, "lr": 4.882266378106645e-05}
{"train_loss": 0.07831447571516037, "global_step": 138006, "epoch": 1550, "lr": 4.882208420528149e-05}
{"train_loss": 0.06755167245864868, "global_step": 138007, "epoch": 1550, "lr": 4.882150462965488e-05}
{"train_loss": 0.0917726680636406, "global_step": 138008, "epoch": 1550, "lr": 4.882092505418673e-05}
{"train_loss": 0.13906006515026093, "global_step": 138009, "epoch": 1550, "lr": 4.8820345478877064e-05}
{"train_loss": 0.13309694826602936, "global_step": 138010, "epoch": 1550, "lr": 4.8819765903726015e-05}
{"train_loss": 0.20833203196525574, "global_step": 138011, "epoch": 1550, "lr": 4.881918632873361e-05}
{"train_loss": 0.12634195387363434, "global_step": 138012, "epoch": 1550, "lr": 4.881860675389998e-05}
{"train_loss": 0.20572388172149658, "global_step": 138013, "epoch": 1550, "lr": 4.881802717922514e-05}
{"train_loss": 0.10064481198787689, "global_step": 138014, "epoch": 1550, "lr": 4.8817447604709235e-05}
{"train_loss": 0.10838019102811813, "global_step": 138015, "epoch": 1550, "lr": 4.8816868030352276e-05}
{"train_loss": 0.07736753672361374, "global_step": 138016, "epoch": 1550, "lr": 4.8816288456154404e-05}
{"train_loss": 0.1304951012134552, "global_step": 138017, "epoch": 1550, "lr": 4.8815708882115645e-05}
{"train_loss": 0.0791134387254715, "global_step": 138018, "epoch": 1550, "lr": 4.881512930823612e-05}
{"train_loss": 0.1329488456249237, "global_step": 138019, "epoch": 1550, "lr": 4.881454973451587e-05}
{"train_loss": 0.12912677228450775, "global_step": 138020, "epoch": 1550, "lr": 4.8813970160955005e-05}
{"train_loss": 0.17926032841205597, "global_step": 138021, "epoch": 1550, "lr": 4.881339058755357e-05}
{"train_loss": 0.14890290796756744, "global_step": 138022, "epoch": 1550, "lr": 4.881281101431168e-05}
{"train_loss": 0.10595922917127609, "global_step": 138023, "epoch": 1550, "lr": 4.881223144122937e-05}
{"train_loss": 0.07610511034727097, "global_step": 138024, "epoch": 1550, "lr": 4.8811651868306754e-05}
{"train_loss": 0.13804306089878082, "global_step": 138025, "epoch": 1550, "lr": 4.8811072295543905e-05}
{"train_loss": 0.11088667809963226, "global_step": 138026, "epoch": 1550, "lr": 4.881049272294087e-05}
{"train_loss": 0.1316942572593689, "global_step": 138027, "epoch": 1550, "lr": 4.880991315049777e-05}
{"train_loss": 0.17590942978858948, "global_step": 138028, "epoch": 1550, "lr": 4.8809333578214645e-05}
{"train_loss": 0.2174738645553589, "global_step": 138029, "epoch": 1550, "lr": 4.880875400609161e-05}
{"train_loss": 0.10402253270149231, "global_step": 138030, "epoch": 1550, "lr": 4.8808174434128706e-05}
{"train_loss": 0.1847800463438034, "global_step": 138031, "epoch": 1550, "lr": 4.880759486232605e-05}
{"train_loss": 0.12915106117725372, "global_step": 138032, "epoch": 1550, "lr": 4.880701529068367e-05}
{"train_loss": 0.10181180387735367, "global_step": 138033, "epoch": 1550, "lr": 4.8806435719201706e-05}
{"train_loss": 0.1251661330461502, "global_step": 138034, "epoch": 1550, "lr": 4.880585614788017e-05}
{"train_loss": 0.11094629764556885, "global_step": 138035, "epoch": 1550, "lr": 4.8805276576719193e-05}
{"train_loss": 0.149931862950325, "global_step": 138036, "epoch": 1550, "lr": 4.8804697005718816e-05}
{"train_loss": 0.07258450239896774, "global_step": 138037, "epoch": 1550, "lr": 4.880411743487915e-05}
{"train_loss": 0.11935218804505433, "global_step": 138038, "epoch": 1550, "lr": 4.880353786420024e-05, "train/sim_max_reward_0": 0.5125209002349327, "train/sim_max_reward_1": 0.4969065035368364, "train/sim_max_reward_2": 0.6690301517798593, "train/sim_max_reward_3": 0.5969877181183441, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.7219858448263732, "test/sim_max_reward_4500000": 0.8954356387477502, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9697278081489974, "test/sim_max_reward_4500003": 0.44516317743204903, "test/sim_max_reward_4500004": 0.5934698445146418, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9629595377515596, "test/sim_max_reward_4500008": 0.15515447962338658, "test/sim_max_reward_4500009": 0.9395513365117031, "test/sim_max_reward_4500010": 0.999050940136008, "test/sim_max_reward_4500011": 0.8879811373816842, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.6213437041336942, "test/sim_max_reward_4500014": 0.98434384863099, "test/sim_max_reward_4500015": 0.99456914837894, "test/sim_max_reward_4500016": 0.9972118485167746, "test/sim_max_reward_4500017": 0.4223402771728485, "test/sim_max_reward_4500018": 0.8467187599980941, "test/sim_max_reward_4500019": 0.933566336277069, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9646512481954621, "test/sim_max_reward_4500022": 0.9888645653921521, "test/sim_max_reward_4500023": 0.5780638927647764, "test/sim_max_reward_4500024": 0.6491480581230538, "test/sim_max_reward_4500025": 0.425463727947344, "test/sim_max_reward_4500026": 0.0, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6886642054354931, "test/sim_max_reward_4500029": 0.9841943014059713, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.43594415116172247, "test/sim_max_reward_4500032": 0.9770635540087097, "test/sim_max_reward_4500033": 0.9829940718003958, "test/sim_max_reward_4500034": 0.08972452873716112, "test/sim_max_reward_4500035": 0.6735953784068665, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.9393560801194797, "test/sim_max_reward_4500038": 0.5706851294749262, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9814197154462754, "test/sim_max_reward_4500041": 0.37255068069734004, "test/sim_max_reward_4500042": 0.0024226635144497643, "test/sim_max_reward_4500043": 0.5601864085588584, "test/sim_max_reward_4500044": 0.5426168877215367, "test/sim_max_reward_4500045": 0.6958615871803373, "test/sim_max_reward_4500046": 0.6926940617814025, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.5762906818504969, "train/mean_score": 0.666238519749391, "test/mean_score": 0.700420868061608, "val_loss": 5.970267295837402, "train_action_mse_error": 20.481082916259766}
{"train_loss": 0.1231219694018364, "global_step": 138039, "epoch": 1551, "lr": 4.88029582936822e-05}
{"train_loss": 0.08587369322776794, "global_step": 138040, "epoch": 1551, "lr": 4.8802378723325065e-05}
{"train_loss": 0.1328536719083786, "global_step": 138041, "epoch": 1551, "lr": 4.8801799153128956e-05}
{"train_loss": 0.09557283669710159, "global_step": 138042, "epoch": 1551, "lr": 4.880121958309391e-05}
{"train_loss": 0.10278313606977463, "global_step": 138043, "epoch": 1551, "lr": 4.8800640013220034e-05}
{"train_loss": 0.09077966958284378, "global_step": 138044, "epoch": 1551, "lr": 4.88000604435074e-05}
{"train_loss": 0.07769717276096344, "global_step": 138045, "epoch": 1551, "lr": 4.8799480873956084e-05}
{"train_loss": 0.09861341863870621, "global_step": 138046, "epoch": 1551, "lr": 4.879890130456617e-05}
{"train_loss": 0.08201658725738525, "global_step": 138047, "epoch": 1551, "lr": 4.879832173533771e-05}
{"train_loss": 0.06565837562084198, "global_step": 138048, "epoch": 1551, "lr": 4.879774216627082e-05}
{"train_loss": 0.07123814523220062, "global_step": 138049, "epoch": 1551, "lr": 4.8797162597365545e-05}
{"train_loss": 0.09404587745666504, "global_step": 138050, "epoch": 1551, "lr": 4.879658302862199e-05}
{"train_loss": 0.06053661182522774, "global_step": 138051, "epoch": 1551, "lr": 4.87960034600402e-05}
{"train_loss": 0.07853669673204422, "global_step": 138052, "epoch": 1551, "lr": 4.8795423891620294e-05}
{"train_loss": 0.1696704775094986, "global_step": 138053, "epoch": 1551, "lr": 4.879484432336231e-05}
{"train_loss": 0.13995487987995148, "global_step": 138054, "epoch": 1551, "lr": 4.879426475526637e-05}
{"train_loss": 0.11998672038316727, "global_step": 138055, "epoch": 1551, "lr": 4.8793685187332496e-05}
{"train_loss": 0.12188791483640671, "global_step": 138056, "epoch": 1551, "lr": 4.879310561956082e-05}
{"train_loss": 0.08230556547641754, "global_step": 138057, "epoch": 1551, "lr": 4.879252605195138e-05}
{"train_loss": 0.14334969222545624, "global_step": 138058, "epoch": 1551, "lr": 4.8791946484504284e-05}
{"train_loss": 0.10211910307407379, "global_step": 138059, "epoch": 1551, "lr": 4.879136691721958e-05}
{"train_loss": 0.08503357321023941, "global_step": 138060, "epoch": 1551, "lr": 4.879078735009738e-05}
{"train_loss": 0.14365637302398682, "global_step": 138061, "epoch": 1551, "lr": 4.879020778313773e-05}
{"train_loss": 0.11507430672645569, "global_step": 138062, "epoch": 1551, "lr": 4.878962821634072e-05}
{"train_loss": 0.1194961741566658, "global_step": 138063, "epoch": 1551, "lr": 4.878904864970644e-05}
{"train_loss": 0.11529861390590668, "global_step": 138064, "epoch": 1551, "lr": 4.878846908323495e-05}
{"train_loss": 0.09552363306283951, "global_step": 138065, "epoch": 1551, "lr": 4.878788951692635e-05}
{"train_loss": 0.13675107061862946, "global_step": 138066, "epoch": 1551, "lr": 4.8787309950780684e-05}
{"train_loss": 0.08852138370275497, "global_step": 138067, "epoch": 1551, "lr": 4.878673038479807e-05}
{"train_loss": 0.06079600006341934, "global_step": 138068, "epoch": 1551, "lr": 4.8786150818978546e-05}
{"train_loss": 0.12167078256607056, "global_step": 138069, "epoch": 1551, "lr": 4.878557125332222e-05}
{"train_loss": 0.13731196522712708, "global_step": 138070, "epoch": 1551, "lr": 4.8784991687829144e-05}
{"train_loss": 0.12896636128425598, "global_step": 138071, "epoch": 1551, "lr": 4.878441212249944e-05}
{"train_loss": 0.11236559599637985, "global_step": 138072, "epoch": 1551, "lr": 4.8783832557333123e-05}
{"train_loss": 0.16947907209396362, "global_step": 138073, "epoch": 1551, "lr": 4.8783252992330336e-05}
{"train_loss": 0.12224215269088745, "global_step": 138074, "epoch": 1551, "lr": 4.878267342749111e-05}
{"train_loss": 0.12163402885198593, "global_step": 138075, "epoch": 1551, "lr": 4.8782093862815545e-05}
{"train_loss": 0.1217082291841507, "global_step": 138076, "epoch": 1551, "lr": 4.878151429830369e-05}
{"train_loss": 0.10520022362470627, "global_step": 138077, "epoch": 1551, "lr": 4.878093473395568e-05}
{"train_loss": 0.09844247251749039, "global_step": 138078, "epoch": 1551, "lr": 4.878035516977153e-05}
{"train_loss": 0.1855476051568985, "global_step": 138079, "epoch": 1551, "lr": 4.877977560575136e-05}
{"train_loss": 0.136017844080925, "global_step": 138080, "epoch": 1551, "lr": 4.877919604189523e-05}
{"train_loss": 0.0481564886868, "global_step": 138081, "epoch": 1551, "lr": 4.877861647820323e-05}
{"train_loss": 0.1063159927725792, "global_step": 138082, "epoch": 1551, "lr": 4.8778036914675424e-05}
{"train_loss": 0.11653662472963333, "global_step": 138083, "epoch": 1551, "lr": 4.8777457351311896e-05}
{"train_loss": 0.1695721298456192, "global_step": 138084, "epoch": 1551, "lr": 4.877687778811273e-05}
{"train_loss": 0.10785888135433197, "global_step": 138085, "epoch": 1551, "lr": 4.877629822507799e-05}
{"train_loss": 0.07562650740146637, "global_step": 138086, "epoch": 1551, "lr": 4.8775718662207774e-05}
{"train_loss": 0.18621183931827545, "global_step": 138087, "epoch": 1551, "lr": 4.8775139099502126e-05}
{"train_loss": 0.08886135369539261, "global_step": 138088, "epoch": 1551, "lr": 4.8774559536961175e-05}
{"train_loss": 0.11222100257873535, "global_step": 138089, "epoch": 1551, "lr": 4.8773979974584935e-05}
{"train_loss": 0.11982617527246475, "global_step": 138090, "epoch": 1551, "lr": 4.877340041237355e-05}
{"train_loss": 0.04819927364587784, "global_step": 138091, "epoch": 1551, "lr": 4.877282085032704e-05}
{"train_loss": 0.12415167689323425, "global_step": 138092, "epoch": 1551, "lr": 4.8772241288445536e-05}
{"train_loss": 0.07965531945228577, "global_step": 138093, "epoch": 1551, "lr": 4.877166172672906e-05}
{"train_loss": 0.13669930398464203, "global_step": 138094, "epoch": 1551, "lr": 4.877108216517775e-05}
{"train_loss": 0.09959609806537628, "global_step": 138095, "epoch": 1551, "lr": 4.877050260379162e-05}
{"train_loss": 0.0932120680809021, "global_step": 138096, "epoch": 1551, "lr": 4.876992304257082e-05}
{"train_loss": 0.08695439994335175, "global_step": 138097, "epoch": 1551, "lr": 4.876934348151535e-05}
{"train_loss": 0.14553087949752808, "global_step": 138098, "epoch": 1551, "lr": 4.876876392062534e-05}
{"train_loss": 0.17100100219249725, "global_step": 138099, "epoch": 1551, "lr": 4.8768184359900865e-05}
{"train_loss": 0.12698130309581757, "global_step": 138100, "epoch": 1551, "lr": 4.876760479934198e-05}
{"train_loss": 0.07665885239839554, "global_step": 138101, "epoch": 1551, "lr": 4.8767025238948785e-05}
{"train_loss": 0.08697183430194855, "global_step": 138102, "epoch": 1551, "lr": 4.8766445678721345e-05}
{"train_loss": 0.14137305319309235, "global_step": 138103, "epoch": 1551, "lr": 4.8765866118659755e-05}
{"train_loss": 0.12354397773742676, "global_step": 138104, "epoch": 1551, "lr": 4.876528655876405e-05}
{"train_loss": 0.11695823818445206, "global_step": 138105, "epoch": 1551, "lr": 4.8764706999034365e-05}
{"train_loss": 0.0823952704668045, "global_step": 138106, "epoch": 1551, "lr": 4.876412743947072e-05}
{"train_loss": 0.10477489978075027, "global_step": 138107, "epoch": 1551, "lr": 4.876354788007326e-05}
{"train_loss": 0.08625634759664536, "global_step": 138108, "epoch": 1551, "lr": 4.876296832084199e-05}
{"train_loss": 0.10411480814218521, "global_step": 138109, "epoch": 1551, "lr": 4.8762388761777055e-05}
{"train_loss": 0.10695262253284454, "global_step": 138110, "epoch": 1551, "lr": 4.8761809202878475e-05}
{"train_loss": 0.16235974431037903, "global_step": 138111, "epoch": 1551, "lr": 4.876122964414638e-05}
{"train_loss": 0.12627582252025604, "global_step": 138112, "epoch": 1551, "lr": 4.87606500855808e-05}
{"train_loss": 0.12606030702590942, "global_step": 138113, "epoch": 1551, "lr": 4.876007052718185e-05}
{"train_loss": 0.1110759750008583, "global_step": 138114, "epoch": 1551, "lr": 4.875949096894958e-05}
{"train_loss": 0.13029944896697998, "global_step": 138115, "epoch": 1551, "lr": 4.8758911410884084e-05}
{"train_loss": 0.1181008443236351, "global_step": 138116, "epoch": 1551, "lr": 4.875833185298545e-05}
{"train_loss": 0.15047234296798706, "global_step": 138117, "epoch": 1551, "lr": 4.875775229525374e-05}
{"train_loss": 0.16395936906337738, "global_step": 138118, "epoch": 1551, "lr": 4.875717273768903e-05}
{"train_loss": 0.13623633980751038, "global_step": 138119, "epoch": 1551, "lr": 4.8756593180291404e-05}
{"train_loss": 0.10844177007675171, "global_step": 138120, "epoch": 1551, "lr": 4.875601362306094e-05}
{"train_loss": 0.08397883176803589, "global_step": 138121, "epoch": 1551, "lr": 4.8755434065997714e-05}
{"train_loss": 0.10400258749723434, "global_step": 138122, "epoch": 1551, "lr": 4.875485450910182e-05}
{"train_loss": 0.1133652999997139, "global_step": 138123, "epoch": 1551, "lr": 4.875427495237328e-05}
{"train_loss": 0.12432628124952316, "global_step": 138124, "epoch": 1551, "lr": 4.8753695395812255e-05}
{"train_loss": 0.15401768684387207, "global_step": 138125, "epoch": 1551, "lr": 4.875311583941876e-05}
{"train_loss": 0.14024391770362854, "global_step": 138126, "epoch": 1551, "lr": 4.87525362831929e-05}
{"train_loss": 0.11364144714695684, "global_step": 138127, "epoch": 1551, "lr": 4.875195672713473e-05, "val_loss": 6.002225399017334}
{"train_loss": 0.1564839482307434, "global_step": 138128, "epoch": 1552, "lr": 4.875137717124437e-05}
{"train_loss": 0.1693059206008911, "global_step": 138129, "epoch": 1552, "lr": 4.875079761552185e-05}
{"train_loss": 0.21835553646087646, "global_step": 138130, "epoch": 1552, "lr": 4.875021805996729e-05}
{"train_loss": 0.09833795577287674, "global_step": 138131, "epoch": 1552, "lr": 4.874963850458073e-05}
{"train_loss": 0.11593810468912125, "global_step": 138132, "epoch": 1552, "lr": 4.874905894936228e-05}
{"train_loss": 0.11351314932107925, "global_step": 138133, "epoch": 1552, "lr": 4.874847939431198e-05}
{"train_loss": 0.12121850997209549, "global_step": 138134, "epoch": 1552, "lr": 4.8747899839429944e-05}
{"train_loss": 0.13167770206928253, "global_step": 138135, "epoch": 1552, "lr": 4.8747320284716256e-05}
{"train_loss": 0.1340724378824234, "global_step": 138136, "epoch": 1552, "lr": 4.874674073017096e-05}
{"train_loss": 0.08423382043838501, "global_step": 138137, "epoch": 1552, "lr": 4.874616117579415e-05}
{"train_loss": 0.1741011142730713, "global_step": 138138, "epoch": 1552, "lr": 4.8745581621585903e-05}
{"train_loss": 0.1673406958580017, "global_step": 138139, "epoch": 1552, "lr": 4.8745002067546304e-05}
{"train_loss": 0.09772439301013947, "global_step": 138140, "epoch": 1552, "lr": 4.874442251367541e-05}
{"train_loss": 0.12258046120405197, "global_step": 138141, "epoch": 1552, "lr": 4.874384295997334e-05}
{"train_loss": 0.10137590765953064, "global_step": 138142, "epoch": 1552, "lr": 4.874326340644012e-05}
{"train_loss": 0.08934415131807327, "global_step": 138143, "epoch": 1552, "lr": 4.874268385307587e-05}
{"train_loss": 0.09295488893985748, "global_step": 138144, "epoch": 1552, "lr": 4.8742104299880624e-05}
{"train_loss": 0.09713217616081238, "global_step": 138145, "epoch": 1552, "lr": 4.8741524746854516e-05}
{"train_loss": 0.0822255089879036, "global_step": 138146, "epoch": 1552, "lr": 4.874094519399757e-05}
{"train_loss": 0.11805371195077896, "global_step": 138147, "epoch": 1552, "lr": 4.874036564130992e-05}
{"train_loss": 0.13240209221839905, "global_step": 138148, "epoch": 1552, "lr": 4.873978608879157e-05}
{"train_loss": 0.0823838859796524, "global_step": 138149, "epoch": 1552, "lr": 4.8739206536442675e-05}
{"train_loss": 0.10855409502983093, "global_step": 138150, "epoch": 1552, "lr": 4.873862698426326e-05}
{"train_loss": 0.059090785682201385, "global_step": 138151, "epoch": 1552, "lr": 4.8738047432253444e-05}
{"train_loss": 0.07074365764856339, "global_step": 138152, "epoch": 1552, "lr": 4.873746788041325e-05}
{"train_loss": 0.12857311964035034, "global_step": 138153, "epoch": 1552, "lr": 4.8736888328742795e-05}
{"train_loss": 0.14176207780838013, "global_step": 138154, "epoch": 1552, "lr": 4.873630877724217e-05}
{"train_loss": 0.08574162423610687, "global_step": 138155, "epoch": 1552, "lr": 4.873572922591142e-05}
{"train_loss": 0.13146747648715973, "global_step": 138156, "epoch": 1552, "lr": 4.873514967475064e-05}
{"train_loss": 0.14589717984199524, "global_step": 138157, "epoch": 1552, "lr": 4.8734570123759907e-05}
{"train_loss": 0.08395922929048538, "global_step": 138158, "epoch": 1552, "lr": 4.873399057293929e-05}
{"train_loss": 0.13699853420257568, "global_step": 138159, "epoch": 1552, "lr": 4.8733411022288866e-05}
{"train_loss": 0.10212108492851257, "global_step": 138160, "epoch": 1552, "lr": 4.873283147180873e-05}
{"train_loss": 0.11131671816110611, "global_step": 138161, "epoch": 1552, "lr": 4.873225192149895e-05}
{"train_loss": 0.22236615419387817, "global_step": 138162, "epoch": 1552, "lr": 4.87316723713596e-05}
{"train_loss": 0.1210261732339859, "global_step": 138163, "epoch": 1552, "lr": 4.8731092821390756e-05}
{"train_loss": 0.1522293984889984, "global_step": 138164, "epoch": 1552, "lr": 4.8730513271592515e-05}
{"train_loss": 0.13388308882713318, "global_step": 138165, "epoch": 1552, "lr": 4.872993372196492e-05}
{"train_loss": 0.12887677550315857, "global_step": 138166, "epoch": 1552, "lr": 4.872935417250809e-05}
{"train_loss": 0.11505769938230515, "global_step": 138167, "epoch": 1552, "lr": 4.8728774623222064e-05}
{"train_loss": 0.07896634191274643, "global_step": 138168, "epoch": 1552, "lr": 4.872819507410696e-05}
{"train_loss": 0.17844067513942719, "global_step": 138169, "epoch": 1552, "lr": 4.872761552516281e-05}
{"train_loss": 0.11038585752248764, "global_step": 138170, "epoch": 1552, "lr": 4.8727035976389724e-05}
{"train_loss": 0.0962248370051384, "global_step": 138171, "epoch": 1552, "lr": 4.8726456427787784e-05}
{"train_loss": 0.109958715736866, "global_step": 138172, "epoch": 1552, "lr": 4.872587687935704e-05}
{"train_loss": 0.10165195912122726, "global_step": 138173, "epoch": 1552, "lr": 4.8725297331097614e-05}
{"train_loss": 0.0844087079167366, "global_step": 138174, "epoch": 1552, "lr": 4.8724717783009526e-05}
{"train_loss": 0.08816568553447723, "global_step": 138175, "epoch": 1552, "lr": 4.872413823509291e-05}
{"train_loss": 0.07833697646856308, "global_step": 138176, "epoch": 1552, "lr": 4.87235586873478e-05}
{"train_loss": 0.135529562830925, "global_step": 138177, "epoch": 1552, "lr": 4.87229791397743e-05}
{"train_loss": 0.14201076328754425, "global_step": 138178, "epoch": 1552, "lr": 4.872239959237248e-05}
{"train_loss": 0.09812624752521515, "global_step": 138179, "epoch": 1552, "lr": 4.872182004514241e-05}
{"train_loss": 0.11150684952735901, "global_step": 138180, "epoch": 1552, "lr": 4.8721240498084185e-05}
{"train_loss": 0.18549789488315582, "global_step": 138181, "epoch": 1552, "lr": 4.872066095119788e-05}
{"train_loss": 0.18345770239830017, "global_step": 138182, "epoch": 1552, "lr": 4.8720081404483544e-05}
{"train_loss": 0.19320355355739594, "global_step": 138183, "epoch": 1552, "lr": 4.8719501857941305e-05}
{"train_loss": 0.16484688222408295, "global_step": 138184, "epoch": 1552, "lr": 4.871892231157119e-05}
{"train_loss": 0.1398514360189438, "global_step": 138185, "epoch": 1552, "lr": 4.871834276537331e-05}
{"train_loss": 0.08635435998439789, "global_step": 138186, "epoch": 1552, "lr": 4.871776321934773e-05}
{"train_loss": 0.1189739927649498, "global_step": 138187, "epoch": 1552, "lr": 4.871718367349454e-05}
{"train_loss": 0.13444384932518005, "global_step": 138188, "epoch": 1552, "lr": 4.8716604127813794e-05}
{"train_loss": 0.12047337740659714, "global_step": 138189, "epoch": 1552, "lr": 4.871602458230558e-05}
{"train_loss": 0.10802491754293442, "global_step": 138190, "epoch": 1552, "lr": 4.871544503697e-05}
{"train_loss": 0.12368164211511612, "global_step": 138191, "epoch": 1552, "lr": 4.8714865491807097e-05}
{"train_loss": 0.10972649604082108, "global_step": 138192, "epoch": 1552, "lr": 4.871428594681699e-05}
{"train_loss": 0.16283366084098816, "global_step": 138193, "epoch": 1552, "lr": 4.8713706401999706e-05}
{"train_loss": 0.0609179325401783, "global_step": 138194, "epoch": 1552, "lr": 4.871312685735536e-05}
{"train_loss": 0.11383531987667084, "global_step": 138195, "epoch": 1552, "lr": 4.871254731288402e-05}
{"train_loss": 0.14498116075992584, "global_step": 138196, "epoch": 1552, "lr": 4.871196776858576e-05}
{"train_loss": 0.15863670408725739, "global_step": 138197, "epoch": 1552, "lr": 4.871138822446066e-05}
{"train_loss": 0.09899146854877472, "global_step": 138198, "epoch": 1552, "lr": 4.8710808680508794e-05}
{"train_loss": 0.153339684009552, "global_step": 138199, "epoch": 1552, "lr": 4.871022913673025e-05}
{"train_loss": 0.10832305252552032, "global_step": 138200, "epoch": 1552, "lr": 4.8709649593125103e-05}
{"train_loss": 0.1020461842417717, "global_step": 138201, "epoch": 1552, "lr": 4.870907004969342e-05}
{"train_loss": 0.14875008165836334, "global_step": 138202, "epoch": 1552, "lr": 4.8708490506435296e-05}
{"train_loss": 0.15950416028499603, "global_step": 138203, "epoch": 1552, "lr": 4.870791096335078e-05}
{"train_loss": 0.08317555487155914, "global_step": 138204, "epoch": 1552, "lr": 4.8707331420439994e-05}
{"train_loss": 0.1473403424024582, "global_step": 138205, "epoch": 1552, "lr": 4.8706751877702965e-05}
{"train_loss": 0.14838072657585144, "global_step": 138206, "epoch": 1552, "lr": 4.87061723351398e-05}
{"train_loss": 0.12902186810970306, "global_step": 138207, "epoch": 1552, "lr": 4.8705592792750605e-05}
{"train_loss": 0.1151464432477951, "global_step": 138208, "epoch": 1552, "lr": 4.8705013250535396e-05}
{"train_loss": 0.08745342493057251, "global_step": 138209, "epoch": 1552, "lr": 4.870443370849431e-05}
{"train_loss": 0.09201700240373611, "global_step": 138210, "epoch": 1552, "lr": 4.8703854166627366e-05}
{"train_loss": 0.09820031374692917, "global_step": 138211, "epoch": 1552, "lr": 4.87032746249347e-05}
{"train_loss": 0.09534578025341034, "global_step": 138212, "epoch": 1552, "lr": 4.870269508341634e-05}
{"train_loss": 0.13139967620372772, "global_step": 138213, "epoch": 1552, "lr": 4.870211554207241e-05}
{"train_loss": 0.13636554777622223, "global_step": 138214, "epoch": 1552, "lr": 4.870153600090295e-05}
{"train_loss": 0.1622103452682495, "global_step": 138215, "epoch": 1552, "lr": 4.870095645990806e-05}
{"train_loss": 0.12283016109232152, "global_step": 138216, "epoch": 1552, "lr": 4.870037691908781e-05, "val_loss": 5.862557888031006}
{"train_loss": 0.05192204564809799, "global_step": 138217, "epoch": 1553, "lr": 4.8699797378442275e-05}
{"train_loss": 0.09099983423948288, "global_step": 138218, "epoch": 1553, "lr": 4.8699217837971534e-05}
{"train_loss": 0.14281967282295227, "global_step": 138219, "epoch": 1553, "lr": 4.869863829767568e-05}
{"train_loss": 0.12225797772407532, "global_step": 138220, "epoch": 1553, "lr": 4.869805875755476e-05}
{"train_loss": 0.15484236180782318, "global_step": 138221, "epoch": 1553, "lr": 4.8697479217608885e-05}
{"train_loss": 0.15324313938617706, "global_step": 138222, "epoch": 1553, "lr": 4.8696899677838104e-05}
{"train_loss": 0.09752737730741501, "global_step": 138223, "epoch": 1553, "lr": 4.869632013824253e-05}
{"train_loss": 0.17081108689308167, "global_step": 138224, "epoch": 1553, "lr": 4.869574059882219e-05}
{"train_loss": 0.08605366200208664, "global_step": 138225, "epoch": 1553, "lr": 4.86951610595772e-05}
{"train_loss": 0.18453674018383026, "global_step": 138226, "epoch": 1553, "lr": 4.869458152050765e-05}
{"train_loss": 0.13612928986549377, "global_step": 138227, "epoch": 1553, "lr": 4.8694001981613575e-05}
{"train_loss": 0.08402850478887558, "global_step": 138228, "epoch": 1553, "lr": 4.8693422442895094e-05}
{"train_loss": 0.07726827263832092, "global_step": 138229, "epoch": 1553, "lr": 4.869284290435225e-05}
{"train_loss": 0.1071590781211853, "global_step": 138230, "epoch": 1553, "lr": 4.8692263365985156e-05}
{"train_loss": 0.11767832189798355, "global_step": 138231, "epoch": 1553, "lr": 4.869168382779385e-05}
{"train_loss": 0.10300661623477936, "global_step": 138232, "epoch": 1553, "lr": 4.869110428977845e-05}
{"train_loss": 0.13025140762329102, "global_step": 138233, "epoch": 1553, "lr": 4.869052475193901e-05}
{"train_loss": 0.12298446148633957, "global_step": 138234, "epoch": 1553, "lr": 4.8689945214275604e-05}
{"train_loss": 0.0823364406824112, "global_step": 138235, "epoch": 1553, "lr": 4.868936567678833e-05}
{"train_loss": 0.09079471230506897, "global_step": 138236, "epoch": 1553, "lr": 4.868878613947725e-05}
{"train_loss": 0.10081572085618973, "global_step": 138237, "epoch": 1553, "lr": 4.8688206602342444e-05}
{"train_loss": 0.12027525156736374, "global_step": 138238, "epoch": 1553, "lr": 4.8687627065384e-05}
{"train_loss": 0.08212795853614807, "global_step": 138239, "epoch": 1553, "lr": 4.868704752860198e-05}
{"train_loss": 0.09905785322189331, "global_step": 138240, "epoch": 1553, "lr": 4.868646799199649e-05}
{"train_loss": 0.16524472832679749, "global_step": 138241, "epoch": 1553, "lr": 4.868588845556756e-05}
{"train_loss": 0.1138535812497139, "global_step": 138242, "epoch": 1553, "lr": 4.868530891931529e-05}
{"train_loss": 0.11308427155017853, "global_step": 138243, "epoch": 1553, "lr": 4.86847293832398e-05}
{"train_loss": 0.1512785702943802, "global_step": 138244, "epoch": 1553, "lr": 4.86841498473411e-05}
{"train_loss": 0.11431490629911423, "global_step": 138245, "epoch": 1553, "lr": 4.868357031161933e-05}
{"train_loss": 0.048523012548685074, "global_step": 138246, "epoch": 1553, "lr": 4.868299077607451e-05}
{"train_loss": 0.13216011226177216, "global_step": 138247, "epoch": 1553, "lr": 4.868241124070677e-05}
{"train_loss": 0.07438648492097855, "global_step": 138248, "epoch": 1553, "lr": 4.868183170551614e-05}
{"train_loss": 0.07919484376907349, "global_step": 138249, "epoch": 1553, "lr": 4.868125217050275e-05}
{"train_loss": 0.099981889128685, "global_step": 138250, "epoch": 1553, "lr": 4.8680672635666615e-05}
{"train_loss": 0.09413748234510422, "global_step": 138251, "epoch": 1553, "lr": 4.868009310100788e-05}
{"train_loss": 0.18779531121253967, "global_step": 138252, "epoch": 1553, "lr": 4.867951356652658e-05}
{"train_loss": 0.1263727843761444, "global_step": 138253, "epoch": 1553, "lr": 4.86789340322228e-05}
{"train_loss": 0.13039495050907135, "global_step": 138254, "epoch": 1553, "lr": 4.8678354498096626e-05}
{"train_loss": 0.09587498754262924, "global_step": 138255, "epoch": 1553, "lr": 4.867777496414812e-05}
{"train_loss": 0.0936691164970398, "global_step": 138256, "epoch": 1553, "lr": 4.867719543037738e-05}
{"train_loss": 0.07724083960056305, "global_step": 138257, "epoch": 1553, "lr": 4.8676615896784486e-05}
{"train_loss": 0.20414717495441437, "global_step": 138258, "epoch": 1553, "lr": 4.8676036363369475e-05}
{"train_loss": 0.11759524047374725, "global_step": 138259, "epoch": 1553, "lr": 4.867545683013248e-05}
{"train_loss": 0.10775101184844971, "global_step": 138260, "epoch": 1553, "lr": 4.867487729707354e-05}
{"train_loss": 0.15818536281585693, "global_step": 138261, "epoch": 1553, "lr": 4.867429776419274e-05}
{"train_loss": 0.12498710304498672, "global_step": 138262, "epoch": 1553, "lr": 4.8673718231490185e-05}
{"train_loss": 0.042697757482528687, "global_step": 138263, "epoch": 1553, "lr": 4.8673138698965913e-05}
{"train_loss": 0.09944133460521698, "global_step": 138264, "epoch": 1553, "lr": 4.867255916662004e-05}
{"train_loss": 0.07142478972673416, "global_step": 138265, "epoch": 1553, "lr": 4.867197963445261e-05}
{"train_loss": 0.1847982108592987, "global_step": 138266, "epoch": 1553, "lr": 4.8671400102463735e-05}
{"train_loss": 0.084202341735363, "global_step": 138267, "epoch": 1553, "lr": 4.8670820570653445e-05}
{"train_loss": 0.1431972235441208, "global_step": 138268, "epoch": 1553, "lr": 4.867024103902187e-05}
{"train_loss": 0.0995911955833435, "global_step": 138269, "epoch": 1553, "lr": 4.8669661507569057e-05}
{"train_loss": 0.08824536949396133, "global_step": 138270, "epoch": 1553, "lr": 4.86690819762951e-05}
{"train_loss": 0.12025467306375504, "global_step": 138271, "epoch": 1553, "lr": 4.8668502445200065e-05}
{"train_loss": 0.16717371344566345, "global_step": 138272, "epoch": 1553, "lr": 4.866792291428403e-05}
{"train_loss": 0.15404389798641205, "global_step": 138273, "epoch": 1553, "lr": 4.8667343383547074e-05}
{"train_loss": 0.10259704291820526, "global_step": 138274, "epoch": 1553, "lr": 4.866676385298928e-05}
{"train_loss": 0.07218854874372482, "global_step": 138275, "epoch": 1553, "lr": 4.866618432261073e-05}
{"train_loss": 0.044956058263778687, "global_step": 138276, "epoch": 1553, "lr": 4.866560479241149e-05}
{"train_loss": 0.13400021195411682, "global_step": 138277, "epoch": 1553, "lr": 4.866502526239164e-05}
{"train_loss": 0.16848593950271606, "global_step": 138278, "epoch": 1553, "lr": 4.866444573255128e-05}
{"train_loss": 0.12981891632080078, "global_step": 138279, "epoch": 1553, "lr": 4.866386620289044e-05}
{"train_loss": 0.0806632936000824, "global_step": 138280, "epoch": 1553, "lr": 4.8663286673409225e-05}
{"train_loss": 0.10709080100059509, "global_step": 138281, "epoch": 1553, "lr": 4.8662707144107745e-05}
{"train_loss": 0.08864882588386536, "global_step": 138282, "epoch": 1553, "lr": 4.866212761498602e-05}
{"train_loss": 0.04842573404312134, "global_step": 138283, "epoch": 1553, "lr": 4.866154808604418e-05}
{"train_loss": 0.05061197280883789, "global_step": 138284, "epoch": 1553, "lr": 4.8660968557282254e-05}
{"train_loss": 0.04803706705570221, "global_step": 138285, "epoch": 1553, "lr": 4.8660389028700365e-05}
{"train_loss": 0.11188101023435593, "global_step": 138286, "epoch": 1553, "lr": 4.8659809500298556e-05}
{"train_loss": 0.17433956265449524, "global_step": 138287, "epoch": 1553, "lr": 4.8659229972076934e-05}
{"train_loss": 0.13419920206069946, "global_step": 138288, "epoch": 1553, "lr": 4.865865044403554e-05}
{"train_loss": 0.0699361264705658, "global_step": 138289, "epoch": 1553, "lr": 4.8658070916174504e-05}
{"train_loss": 0.14291633665561676, "global_step": 138290, "epoch": 1553, "lr": 4.8657491388493845e-05}
{"train_loss": 0.1346023976802826, "global_step": 138291, "epoch": 1553, "lr": 4.865691186099369e-05}
{"train_loss": 0.07102233916521072, "global_step": 138292, "epoch": 1553, "lr": 4.865633233367409e-05}
{"train_loss": 0.17125137150287628, "global_step": 138293, "epoch": 1553, "lr": 4.8655752806535134e-05}
{"train_loss": 0.1483670324087143, "global_step": 138294, "epoch": 1553, "lr": 4.8655173279576886e-05}
{"train_loss": 0.09745330363512039, "global_step": 138295, "epoch": 1553, "lr": 4.865459375279944e-05}
{"train_loss": 0.0657069981098175, "global_step": 138296, "epoch": 1553, "lr": 4.865401422620286e-05}
{"train_loss": 0.07324326783418655, "global_step": 138297, "epoch": 1553, "lr": 4.865343469978723e-05}
{"train_loss": 0.10263550281524658, "global_step": 138298, "epoch": 1553, "lr": 4.865285517355265e-05}
{"train_loss": 0.13063332438468933, "global_step": 138299, "epoch": 1553, "lr": 4.865227564749916e-05}
{"train_loss": 0.10398334264755249, "global_step": 138300, "epoch": 1553, "lr": 4.8651696121626866e-05}
{"train_loss": 0.1630987673997879, "global_step": 138301, "epoch": 1553, "lr": 4.865111659593581e-05}
{"train_loss": 0.14209090173244476, "global_step": 138302, "epoch": 1553, "lr": 4.8650537070426127e-05}
{"train_loss": 0.14163663983345032, "global_step": 138303, "epoch": 1553, "lr": 4.864995754509784e-05}
{"train_loss": 0.10144902765750885, "global_step": 138304, "epoch": 1553, "lr": 4.864937801995107e-05}
{"train_loss": 0.11342641815878032, "global_step": 138305, "epoch": 1553, "lr": 4.8648798494985856e-05, "val_loss": 5.949792385101318}
{"train_loss": 0.10423998534679413, "global_step": 138306, "epoch": 1554, "lr": 4.864821897020231e-05}
{"train_loss": 0.13337910175323486, "global_step": 138307, "epoch": 1554, "lr": 4.864763944560047e-05}
{"train_loss": 0.08035421371459961, "global_step": 138308, "epoch": 1554, "lr": 4.864705992118047e-05}
{"train_loss": 0.14593133330345154, "global_step": 138309, "epoch": 1554, "lr": 4.864648039694233e-05}
{"train_loss": 0.10703319311141968, "global_step": 138310, "epoch": 1554, "lr": 4.8645900872886176e-05}
{"train_loss": 0.08337611705064774, "global_step": 138311, "epoch": 1554, "lr": 4.864532134901205e-05}
{"train_loss": 0.09895289689302444, "global_step": 138312, "epoch": 1554, "lr": 4.864474182532005e-05}
{"train_loss": 0.11955425143241882, "global_step": 138313, "epoch": 1554, "lr": 4.864416230181025e-05}
{"train_loss": 0.09430818259716034, "global_step": 138314, "epoch": 1554, "lr": 4.8643582778482714e-05}
{"train_loss": 0.0996297299861908, "global_step": 138315, "epoch": 1554, "lr": 4.864300325533754e-05}
{"train_loss": 0.11825907975435257, "global_step": 138316, "epoch": 1554, "lr": 4.864242373237479e-05}
{"train_loss": 0.10391326248645782, "global_step": 138317, "epoch": 1554, "lr": 4.864184420959457e-05}
{"train_loss": 0.10866018384695053, "global_step": 138318, "epoch": 1554, "lr": 4.8641264686996907e-05}
{"train_loss": 0.11757051199674606, "global_step": 138319, "epoch": 1554, "lr": 4.864068516458194e-05}
{"train_loss": 0.1850229650735855, "global_step": 138320, "epoch": 1554, "lr": 4.864010564234969e-05}
{"train_loss": 0.06735871732234955, "global_step": 138321, "epoch": 1554, "lr": 4.863952612030028e-05}
{"train_loss": 0.1228780522942543, "global_step": 138322, "epoch": 1554, "lr": 4.8638946598433756e-05}
{"train_loss": 0.15001490712165833, "global_step": 138323, "epoch": 1554, "lr": 4.8638367076750226e-05}
{"train_loss": 0.10006904602050781, "global_step": 138324, "epoch": 1554, "lr": 4.863778755524973e-05}
{"train_loss": 0.10287530720233917, "global_step": 138325, "epoch": 1554, "lr": 4.863720803393239e-05}
{"train_loss": 0.1471843123435974, "global_step": 138326, "epoch": 1554, "lr": 4.863662851279824e-05}
{"train_loss": 0.06764405220746994, "global_step": 138327, "epoch": 1554, "lr": 4.8636048991847397e-05}
{"train_loss": 0.10134647786617279, "global_step": 138328, "epoch": 1554, "lr": 4.8635469471079895e-05}
{"train_loss": 0.09033435583114624, "global_step": 138329, "epoch": 1554, "lr": 4.863488995049587e-05}
{"train_loss": 0.1104273721575737, "global_step": 138330, "epoch": 1554, "lr": 4.863431043009535e-05}
{"train_loss": 0.10372302681207657, "global_step": 138331, "epoch": 1554, "lr": 4.8633730909878425e-05}
{"train_loss": 0.16757869720458984, "global_step": 138332, "epoch": 1554, "lr": 4.8633151389845186e-05}
{"train_loss": 0.08738010376691818, "global_step": 138333, "epoch": 1554, "lr": 4.863257186999569e-05}
{"train_loss": 0.0587286502122879, "global_step": 138334, "epoch": 1554, "lr": 4.863199235033005e-05}
{"train_loss": 0.1363987773656845, "global_step": 138335, "epoch": 1554, "lr": 4.86314128308483e-05}
{"train_loss": 0.08300300687551498, "global_step": 138336, "epoch": 1554, "lr": 4.863083331155056e-05}
{"train_loss": 0.07574666291475296, "global_step": 138337, "epoch": 1554, "lr": 4.863025379243687e-05}
{"train_loss": 0.0982341319322586, "global_step": 138338, "epoch": 1554, "lr": 4.862967427350735e-05}
{"train_loss": 0.10907676070928574, "global_step": 138339, "epoch": 1554, "lr": 4.8629094754762025e-05}
{"train_loss": 0.10460947453975677, "global_step": 138340, "epoch": 1554, "lr": 4.8628515236201024e-05}
{"train_loss": 0.1174086481332779, "global_step": 138341, "epoch": 1554, "lr": 4.862793571782438e-05}
{"train_loss": 0.10894791036844254, "global_step": 138342, "epoch": 1554, "lr": 4.8627356199632216e-05}
{"train_loss": 0.13843217492103577, "global_step": 138343, "epoch": 1554, "lr": 4.862677668162457e-05}
{"train_loss": 0.08481620252132416, "global_step": 138344, "epoch": 1554, "lr": 4.862619716380156e-05}
{"train_loss": 0.09601176530122757, "global_step": 138345, "epoch": 1554, "lr": 4.862561764616321e-05}
{"train_loss": 0.16996589303016663, "global_step": 138346, "epoch": 1554, "lr": 4.862503812870965e-05}
{"train_loss": 0.08635777235031128, "global_step": 138347, "epoch": 1554, "lr": 4.8624458611440925e-05}
{"train_loss": 0.19971951842308044, "global_step": 138348, "epoch": 1554, "lr": 4.862387909435714e-05}
{"train_loss": 0.14691342413425446, "global_step": 138349, "epoch": 1554, "lr": 4.862329957745835e-05}
{"train_loss": 0.13801082968711853, "global_step": 138350, "epoch": 1554, "lr": 4.862272006074464e-05}
{"train_loss": 0.0850655660033226, "global_step": 138351, "epoch": 1554, "lr": 4.8622140544216086e-05}
{"train_loss": 0.13677112758159637, "global_step": 138352, "epoch": 1554, "lr": 4.8621561027872766e-05}
{"train_loss": 0.15012012422084808, "global_step": 138353, "epoch": 1554, "lr": 4.862098151171478e-05}
{"train_loss": 0.16569100320339203, "global_step": 138354, "epoch": 1554, "lr": 4.8620401995742153e-05}
{"train_loss": 0.09906112402677536, "global_step": 138355, "epoch": 1554, "lr": 4.861982247995502e-05}
{"train_loss": 0.0717427134513855, "global_step": 138356, "epoch": 1554, "lr": 4.8619242964353426e-05}
{"train_loss": 0.12684683501720428, "global_step": 138357, "epoch": 1554, "lr": 4.8618663448937465e-05}
{"train_loss": 0.08935341984033585, "global_step": 138358, "epoch": 1554, "lr": 4.86180839337072e-05}
{"train_loss": 0.0840429961681366, "global_step": 138359, "epoch": 1554, "lr": 4.861750441866273e-05}
{"train_loss": 0.08265822380781174, "global_step": 138360, "epoch": 1554, "lr": 4.86169249038041e-05}
{"train_loss": 0.13475917279720306, "global_step": 138361, "epoch": 1554, "lr": 4.861634538913143e-05}
{"train_loss": 0.14955070614814758, "global_step": 138362, "epoch": 1554, "lr": 4.861576587464475e-05}
{"train_loss": 0.1352866142988205, "global_step": 138363, "epoch": 1554, "lr": 4.861518636034419e-05}
{"train_loss": 0.1067834123969078, "global_step": 138364, "epoch": 1554, "lr": 4.8614606846229774e-05}
{"train_loss": 0.08219730108976364, "global_step": 138365, "epoch": 1554, "lr": 4.8614027332301636e-05}
{"train_loss": 0.14601825177669525, "global_step": 138366, "epoch": 1554, "lr": 4.86134478185598e-05}
{"train_loss": 0.13966530561447144, "global_step": 138367, "epoch": 1554, "lr": 4.861286830500439e-05}
{"train_loss": 0.07386928051710129, "global_step": 138368, "epoch": 1554, "lr": 4.861228879163545e-05}
{"train_loss": 0.0764869824051857, "global_step": 138369, "epoch": 1554, "lr": 4.861170927845307e-05}
{"train_loss": 0.0984085276722908, "global_step": 138370, "epoch": 1554, "lr": 4.8611129765457344e-05}
{"train_loss": 0.09001800417900085, "global_step": 138371, "epoch": 1554, "lr": 4.861055025264831e-05}
{"train_loss": 0.08356761932373047, "global_step": 138372, "epoch": 1554, "lr": 4.860997074002609e-05}
{"train_loss": 0.1732054203748703, "global_step": 138373, "epoch": 1554, "lr": 4.860939122759073e-05}
{"train_loss": 0.09727794677019119, "global_step": 138374, "epoch": 1554, "lr": 4.8608811715342334e-05}
{"train_loss": 0.13938304781913757, "global_step": 138375, "epoch": 1554, "lr": 4.860823220328095e-05}
{"train_loss": 0.2138516902923584, "global_step": 138376, "epoch": 1554, "lr": 4.860765269140669e-05}
{"train_loss": 0.2492104470729828, "global_step": 138377, "epoch": 1554, "lr": 4.860707317971959e-05}
{"train_loss": 0.08119853585958481, "global_step": 138378, "epoch": 1554, "lr": 4.860649366821978e-05}
{"train_loss": 0.06980583071708679, "global_step": 138379, "epoch": 1554, "lr": 4.860591415690729e-05}
{"train_loss": 0.11370650678873062, "global_step": 138380, "epoch": 1554, "lr": 4.8605334645782235e-05}
{"train_loss": 0.08751785010099411, "global_step": 138381, "epoch": 1554, "lr": 4.860475513484465e-05}
{"train_loss": 0.06406717747449875, "global_step": 138382, "epoch": 1554, "lr": 4.8604175624094665e-05}
{"train_loss": 0.10174593329429626, "global_step": 138383, "epoch": 1554, "lr": 4.860359611353231e-05}
{"train_loss": 0.13652384281158447, "global_step": 138384, "epoch": 1554, "lr": 4.860301660315771e-05}
{"train_loss": 0.07814832776784897, "global_step": 138385, "epoch": 1554, "lr": 4.8602437092970894e-05}
{"train_loss": 0.11497370153665543, "global_step": 138386, "epoch": 1554, "lr": 4.860185758297198e-05}
{"train_loss": 0.05808082967996597, "global_step": 138387, "epoch": 1554, "lr": 4.860127807316101e-05}
{"train_loss": 0.10409192740917206, "global_step": 138388, "epoch": 1554, "lr": 4.860069856353809e-05}
{"train_loss": 0.085159532725811, "global_step": 138389, "epoch": 1554, "lr": 4.86001190541033e-05}
{"train_loss": 0.13698327541351318, "global_step": 138390, "epoch": 1554, "lr": 4.8599539544856696e-05}
{"train_loss": 0.14746500551700592, "global_step": 138391, "epoch": 1554, "lr": 4.859896003579838e-05}
{"train_loss": 0.11135877668857574, "global_step": 138392, "epoch": 1554, "lr": 4.859838052692839e-05}
{"train_loss": 0.12872765958309174, "global_step": 138393, "epoch": 1554, "lr": 4.859780101824686e-05}
{"train_loss": 0.1134577581889174, "global_step": 138394, "epoch": 1554, "lr": 4.859722150975381e-05, "val_loss": 6.113479137420654}
{"train_loss": 0.08870873600244522, "global_step": 138395, "epoch": 1555, "lr": 4.859664200144938e-05}
{"train_loss": 0.11350797116756439, "global_step": 138396, "epoch": 1555, "lr": 4.859606249333358e-05}
{"train_loss": 0.21633943915367126, "global_step": 138397, "epoch": 1555, "lr": 4.859548298540656e-05}
{"train_loss": 0.10420183837413788, "global_step": 138398, "epoch": 1555, "lr": 4.859490347766832e-05}
{"train_loss": 0.13428683578968048, "global_step": 138399, "epoch": 1555, "lr": 4.8594323970119004e-05}
{"train_loss": 0.16592107713222504, "global_step": 138400, "epoch": 1555, "lr": 4.859374446275865e-05}
{"train_loss": 0.11729560792446136, "global_step": 138401, "epoch": 1555, "lr": 4.8593164955587364e-05}
{"train_loss": 0.08058742433786392, "global_step": 138402, "epoch": 1555, "lr": 4.8592585448605195e-05}
{"train_loss": 0.13928574323654175, "global_step": 138403, "epoch": 1555, "lr": 4.859200594181225e-05}
{"train_loss": 0.13622315227985382, "global_step": 138404, "epoch": 1555, "lr": 4.859142643520858e-05}
{"train_loss": 0.13785840570926666, "global_step": 138405, "epoch": 1555, "lr": 4.8590846928794295e-05}
{"train_loss": 0.13184207677841187, "global_step": 138406, "epoch": 1555, "lr": 4.859026742256943e-05}
{"train_loss": 0.16555072367191315, "global_step": 138407, "epoch": 1555, "lr": 4.8589687916534085e-05}
{"train_loss": 0.16067171096801758, "global_step": 138408, "epoch": 1555, "lr": 4.8589108410688364e-05}
{"train_loss": 0.12736545503139496, "global_step": 138409, "epoch": 1555, "lr": 4.8588528905032296e-05}
{"train_loss": 0.11350088566541672, "global_step": 138410, "epoch": 1555, "lr": 4.858794939956601e-05}
{"train_loss": 0.08845510333776474, "global_step": 138411, "epoch": 1555, "lr": 4.858736989428952e-05}
{"train_loss": 0.13957467675209045, "global_step": 138412, "epoch": 1555, "lr": 4.858679038920298e-05}
{"train_loss": 0.07618913799524307, "global_step": 138413, "epoch": 1555, "lr": 4.85862108843064e-05}
{"train_loss": 0.11064410954713821, "global_step": 138414, "epoch": 1555, "lr": 4.8585631379599905e-05}
{"train_loss": 0.15379852056503296, "global_step": 138415, "epoch": 1555, "lr": 4.858505187508354e-05}
{"train_loss": 0.10680875927209854, "global_step": 138416, "epoch": 1555, "lr": 4.858447237075741e-05}
{"train_loss": 0.08398792147636414, "global_step": 138417, "epoch": 1555, "lr": 4.8583892866621566e-05}
{"train_loss": 0.06203924119472504, "global_step": 138418, "epoch": 1555, "lr": 4.858331336267612e-05}
{"train_loss": 0.15212418138980865, "global_step": 138419, "epoch": 1555, "lr": 4.858273385892111e-05}
{"train_loss": 0.17996205389499664, "global_step": 138420, "epoch": 1555, "lr": 4.858215435535665e-05}
{"train_loss": 0.09043920040130615, "global_step": 138421, "epoch": 1555, "lr": 4.858157485198279e-05}
{"train_loss": 0.12221606820821762, "global_step": 138422, "epoch": 1555, "lr": 4.858099534879964e-05}
{"train_loss": 0.12181666493415833, "global_step": 138423, "epoch": 1555, "lr": 4.8580415845807224e-05}
{"train_loss": 0.13125954568386078, "global_step": 138424, "epoch": 1555, "lr": 4.857983634300568e-05}
{"train_loss": 0.146437406539917, "global_step": 138425, "epoch": 1555, "lr": 4.857925684039505e-05}
{"train_loss": 0.1277661919593811, "global_step": 138426, "epoch": 1555, "lr": 4.857867733797542e-05}
{"train_loss": 0.1541742980480194, "global_step": 138427, "epoch": 1555, "lr": 4.857809783574689e-05}
{"train_loss": 0.08610308915376663, "global_step": 138428, "epoch": 1555, "lr": 4.857751833370949e-05}
{"train_loss": 0.1050589308142662, "global_step": 138429, "epoch": 1555, "lr": 4.857693883186334e-05}
{"train_loss": 0.19837415218353271, "global_step": 138430, "epoch": 1555, "lr": 4.8576359330208496e-05}
{"train_loss": 0.08431542664766312, "global_step": 138431, "epoch": 1555, "lr": 4.857577982874506e-05}
{"train_loss": 0.14265546202659607, "global_step": 138432, "epoch": 1555, "lr": 4.857520032747306e-05}
{"train_loss": 0.15611042082309723, "global_step": 138433, "epoch": 1555, "lr": 4.857462082639264e-05}
{"train_loss": 0.05063405632972717, "global_step": 138434, "epoch": 1555, "lr": 4.857404132550382e-05}
{"train_loss": 0.0802433043718338, "global_step": 138435, "epoch": 1555, "lr": 4.857346182480673e-05}
{"train_loss": 0.1348498910665512, "global_step": 138436, "epoch": 1555, "lr": 4.857288232430139e-05}
{"train_loss": 0.08615130186080933, "global_step": 138437, "epoch": 1555, "lr": 4.857230282398793e-05}
{"train_loss": 0.1714775413274765, "global_step": 138438, "epoch": 1555, "lr": 4.85717233238664e-05}
{"train_loss": 0.1118202954530716, "global_step": 138439, "epoch": 1555, "lr": 4.857114382393689e-05}
{"train_loss": 0.17983174324035645, "global_step": 138440, "epoch": 1555, "lr": 4.8570564324199455e-05}
{"train_loss": 0.1568533182144165, "global_step": 138441, "epoch": 1555, "lr": 4.856998482465422e-05}
{"train_loss": 0.08491775393486023, "global_step": 138442, "epoch": 1555, "lr": 4.8569405325301204e-05}
{"train_loss": 0.1698947697877884, "global_step": 138443, "epoch": 1555, "lr": 4.8568825826140524e-05}
{"train_loss": 0.11094603687524796, "global_step": 138444, "epoch": 1555, "lr": 4.856824632717225e-05}
{"train_loss": 0.08978088200092316, "global_step": 138445, "epoch": 1555, "lr": 4.856766682839646e-05}
{"train_loss": 0.14986862242221832, "global_step": 138446, "epoch": 1555, "lr": 4.856708732981323e-05}
{"train_loss": 0.11387165635824203, "global_step": 138447, "epoch": 1555, "lr": 4.856650783142263e-05}
{"train_loss": 0.07474477589130402, "global_step": 138448, "epoch": 1555, "lr": 4.856592833322476e-05}
{"train_loss": 0.129993736743927, "global_step": 138449, "epoch": 1555, "lr": 4.856534883521966e-05}
{"train_loss": 0.08382757008075714, "global_step": 138450, "epoch": 1555, "lr": 4.856476933740746e-05}
{"train_loss": 0.07207988202571869, "global_step": 138451, "epoch": 1555, "lr": 4.856418983978818e-05}
{"train_loss": 0.13525594770908356, "global_step": 138452, "epoch": 1555, "lr": 4.856361034236195e-05}
{"train_loss": 0.12187279760837555, "global_step": 138453, "epoch": 1555, "lr": 4.856303084512881e-05}
{"train_loss": 0.13677050173282623, "global_step": 138454, "epoch": 1555, "lr": 4.856245134808887e-05}
{"train_loss": 0.10338002443313599, "global_step": 138455, "epoch": 1555, "lr": 4.8561871851242166e-05}
{"train_loss": 0.10848142951726913, "global_step": 138456, "epoch": 1555, "lr": 4.8561292354588826e-05}
{"train_loss": 0.1399613618850708, "global_step": 138457, "epoch": 1555, "lr": 4.856071285812888e-05}
{"train_loss": 0.09289097785949707, "global_step": 138458, "epoch": 1555, "lr": 4.8560133361862456e-05}
{"train_loss": 0.13154593110084534, "global_step": 138459, "epoch": 1555, "lr": 4.855955386578958e-05}
{"train_loss": 0.12652839720249176, "global_step": 138460, "epoch": 1555, "lr": 4.8558974369910356e-05}
{"train_loss": 0.1409928947687149, "global_step": 138461, "epoch": 1555, "lr": 4.855839487422487e-05}
{"train_loss": 0.11239834874868393, "global_step": 138462, "epoch": 1555, "lr": 4.855781537873319e-05}
{"train_loss": 0.1446017324924469, "global_step": 138463, "epoch": 1555, "lr": 4.855723588343539e-05}
{"train_loss": 0.10872285068035126, "global_step": 138464, "epoch": 1555, "lr": 4.8556656388331554e-05}
{"train_loss": 0.11836729198694229, "global_step": 138465, "epoch": 1555, "lr": 4.855607689342177e-05}
{"train_loss": 0.1496078372001648, "global_step": 138466, "epoch": 1555, "lr": 4.855549739870608e-05}
{"train_loss": 0.06722262501716614, "global_step": 138467, "epoch": 1555, "lr": 4.855491790418461e-05}
{"train_loss": 0.0883505642414093, "global_step": 138468, "epoch": 1555, "lr": 4.855433840985739e-05}
{"train_loss": 0.0801030844449997, "global_step": 138469, "epoch": 1555, "lr": 4.855375891572454e-05}
{"train_loss": 0.10770748555660248, "global_step": 138470, "epoch": 1555, "lr": 4.8553179421786104e-05}
{"train_loss": 0.059723351150751114, "global_step": 138471, "epoch": 1555, "lr": 4.85525999280422e-05}
{"train_loss": 0.205256387591362, "global_step": 138472, "epoch": 1555, "lr": 4.855202043449285e-05}
{"train_loss": 0.15745779871940613, "global_step": 138473, "epoch": 1555, "lr": 4.8551440941138186e-05}
{"train_loss": 0.0782976821064949, "global_step": 138474, "epoch": 1555, "lr": 4.8550861447978246e-05}
{"train_loss": 0.11294969171285629, "global_step": 138475, "epoch": 1555, "lr": 4.855028195501315e-05}
{"train_loss": 0.11874277889728546, "global_step": 138476, "epoch": 1555, "lr": 4.854970246224293e-05}
{"train_loss": 0.1205563023686409, "global_step": 138477, "epoch": 1555, "lr": 4.854912296966769e-05}
{"train_loss": 0.19188375771045685, "global_step": 138478, "epoch": 1555, "lr": 4.8548543477287496e-05}
{"train_loss": 0.13379497826099396, "global_step": 138479, "epoch": 1555, "lr": 4.8547963985102444e-05}
{"train_loss": 0.080651193857193, "global_step": 138480, "epoch": 1555, "lr": 4.854738449311259e-05}
{"train_loss": 0.14571227133274078, "global_step": 138481, "epoch": 1555, "lr": 4.854680500131803e-05}
{"train_loss": 0.12109953165054321, "global_step": 138482, "epoch": 1555, "lr": 4.8546225509718835e-05}
{"train_loss": 0.12162828884935112, "global_step": 138483, "epoch": 1555, "lr": 4.854564601831508e-05, "val_loss": 5.862825870513916, "train_action_mse_error": 38.43241882324219}
{"train_loss": 0.14607523381710052, "global_step": 138484, "epoch": 1556, "lr": 4.854506652710685e-05}
{"train_loss": 0.13302914798259735, "global_step": 138485, "epoch": 1556, "lr": 4.85444870360942e-05}
{"train_loss": 0.09359508752822876, "global_step": 138486, "epoch": 1556, "lr": 4.854390754527726e-05}
{"train_loss": 0.14170514047145844, "global_step": 138487, "epoch": 1556, "lr": 4.854332805465604e-05}
{"train_loss": 0.08867788314819336, "global_step": 138488, "epoch": 1556, "lr": 4.854274856423068e-05}
{"train_loss": 0.0623435415327549, "global_step": 138489, "epoch": 1556, "lr": 4.8542169074001195e-05}
{"train_loss": 0.12198309600353241, "global_step": 138490, "epoch": 1556, "lr": 4.8541589583967736e-05}
{"train_loss": 0.08910305052995682, "global_step": 138491, "epoch": 1556, "lr": 4.854101009413031e-05}
{"train_loss": 0.1771547943353653, "global_step": 138492, "epoch": 1556, "lr": 4.854043060448905e-05}
{"train_loss": 0.1117263212800026, "global_step": 138493, "epoch": 1556, "lr": 4.853985111504399e-05}
{"train_loss": 0.15772941708564758, "global_step": 138494, "epoch": 1556, "lr": 4.853927162579526e-05}
{"train_loss": 0.08555953949689865, "global_step": 138495, "epoch": 1556, "lr": 4.853869213674287e-05}
{"train_loss": 0.11114098876714706, "global_step": 138496, "epoch": 1556, "lr": 4.8538112647886944e-05}
{"train_loss": 0.08806006610393524, "global_step": 138497, "epoch": 1556, "lr": 4.8537533159227575e-05}
{"train_loss": 0.1420735865831375, "global_step": 138498, "epoch": 1556, "lr": 4.85369536707648e-05}
{"train_loss": 0.14894899725914001, "global_step": 138499, "epoch": 1556, "lr": 4.8536374182498724e-05}
{"train_loss": 0.10150018334388733, "global_step": 138500, "epoch": 1556, "lr": 4.8535794694429405e-05}
{"train_loss": 0.11446315795183182, "global_step": 138501, "epoch": 1556, "lr": 4.8535215206556936e-05}
{"train_loss": 0.08262890577316284, "global_step": 138502, "epoch": 1556, "lr": 4.8534635718881386e-05}
{"train_loss": 0.0950421690940857, "global_step": 138503, "epoch": 1556, "lr": 4.853405623140283e-05}
{"train_loss": 0.07635223865509033, "global_step": 138504, "epoch": 1556, "lr": 4.853347674412136e-05}
{"train_loss": 0.13310544192790985, "global_step": 138505, "epoch": 1556, "lr": 4.8532897257037056e-05}
{"train_loss": 0.0948726162314415, "global_step": 138506, "epoch": 1556, "lr": 4.8532317770149965e-05}
{"train_loss": 0.152555450797081, "global_step": 138507, "epoch": 1556, "lr": 4.8531738283460205e-05}
{"train_loss": 0.11706078052520752, "global_step": 138508, "epoch": 1556, "lr": 4.853115879696782e-05}
{"train_loss": 0.08819303661584854, "global_step": 138509, "epoch": 1556, "lr": 4.8530579310672917e-05}
{"train_loss": 0.08336795121431351, "global_step": 138510, "epoch": 1556, "lr": 4.852999982457554e-05}
{"train_loss": 0.1169866994023323, "global_step": 138511, "epoch": 1556, "lr": 4.852942033867581e-05}
{"train_loss": 0.09035927057266235, "global_step": 138512, "epoch": 1556, "lr": 4.8528840852973757e-05}
{"train_loss": 0.09209353476762772, "global_step": 138513, "epoch": 1556, "lr": 4.852826136746951e-05}
{"train_loss": 0.1524643450975418, "global_step": 138514, "epoch": 1556, "lr": 4.85276818821631e-05}
{"train_loss": 0.0835149884223938, "global_step": 138515, "epoch": 1556, "lr": 4.852710239705461e-05}
{"train_loss": 0.11413991451263428, "global_step": 138516, "epoch": 1556, "lr": 4.852652291214417e-05}
{"train_loss": 0.06976908445358276, "global_step": 138517, "epoch": 1556, "lr": 4.852594342743179e-05}
{"train_loss": 0.07488125562667847, "global_step": 138518, "epoch": 1556, "lr": 4.852536394291761e-05}
{"train_loss": 0.11686094850301743, "global_step": 138519, "epoch": 1556, "lr": 4.8524784458601665e-05}
{"train_loss": 0.10953068733215332, "global_step": 138520, "epoch": 1556, "lr": 4.852420497448404e-05}
{"train_loss": 0.142863467335701, "global_step": 138521, "epoch": 1556, "lr": 4.8523625490564816e-05}
{"train_loss": 0.12389475852251053, "global_step": 138522, "epoch": 1556, "lr": 4.8523046006844076e-05}
{"train_loss": 0.09650212526321411, "global_step": 138523, "epoch": 1556, "lr": 4.8522466523321894e-05}
{"train_loss": 0.1300419420003891, "global_step": 138524, "epoch": 1556, "lr": 4.852188703999836e-05}
{"train_loss": 0.12071546167135239, "global_step": 138525, "epoch": 1556, "lr": 4.852130755687352e-05}
{"train_loss": 0.15989966690540314, "global_step": 138526, "epoch": 1556, "lr": 4.852072807394749e-05}
{"train_loss": 0.14797770977020264, "global_step": 138527, "epoch": 1556, "lr": 4.852014859122031e-05}
{"train_loss": 0.092434361577034, "global_step": 138528, "epoch": 1556, "lr": 4.85195691086921e-05}
{"train_loss": 0.16898523271083832, "global_step": 138529, "epoch": 1556, "lr": 4.851898962636289e-05}
{"train_loss": 0.08901023864746094, "global_step": 138530, "epoch": 1556, "lr": 4.851841014423281e-05}
{"train_loss": 0.06403566896915436, "global_step": 138531, "epoch": 1556, "lr": 4.8517830662301885e-05}
{"train_loss": 0.13715583086013794, "global_step": 138532, "epoch": 1556, "lr": 4.851725118057025e-05}
{"train_loss": 0.10243004560470581, "global_step": 138533, "epoch": 1556, "lr": 4.851667169903793e-05}
{"train_loss": 0.1390800029039383, "global_step": 138534, "epoch": 1556, "lr": 4.851609221770502e-05}
{"train_loss": 0.12133467942476273, "global_step": 138535, "epoch": 1556, "lr": 4.851551273657163e-05}
{"train_loss": 0.12691079080104828, "global_step": 138536, "epoch": 1556, "lr": 4.8514933255637785e-05}
{"train_loss": 0.10272667557001114, "global_step": 138537, "epoch": 1556, "lr": 4.851435377490361e-05}
{"train_loss": 0.1262212097644806, "global_step": 138538, "epoch": 1556, "lr": 4.8513774294369155e-05}
{"train_loss": 0.17441587150096893, "global_step": 138539, "epoch": 1556, "lr": 4.85131948140345e-05}
{"train_loss": 0.05813421681523323, "global_step": 138540, "epoch": 1556, "lr": 4.851261533389974e-05}
{"train_loss": 0.16857849061489105, "global_step": 138541, "epoch": 1556, "lr": 4.8512035853964924e-05}
{"train_loss": 0.15539821982383728, "global_step": 138542, "epoch": 1556, "lr": 4.851145637423016e-05}
{"train_loss": 0.1806892305612564, "global_step": 138543, "epoch": 1556, "lr": 4.851087689469551e-05}
{"train_loss": 0.0982118770480156, "global_step": 138544, "epoch": 1556, "lr": 4.8510297415361044e-05}
{"train_loss": 0.14465191960334778, "global_step": 138545, "epoch": 1556, "lr": 4.8509717936226866e-05}
{"train_loss": 0.09624605625867844, "global_step": 138546, "epoch": 1556, "lr": 4.850913845729303e-05}
{"train_loss": 0.14335474371910095, "global_step": 138547, "epoch": 1556, "lr": 4.8508558978559624e-05}
{"train_loss": 0.11938203871250153, "global_step": 138548, "epoch": 1556, "lr": 4.850797950002671e-05}
{"train_loss": 0.0796537697315216, "global_step": 138549, "epoch": 1556, "lr": 4.850740002169441e-05}
{"train_loss": 0.18837106227874756, "global_step": 138550, "epoch": 1556, "lr": 4.850682054356274e-05}
{"train_loss": 0.09567883610725403, "global_step": 138551, "epoch": 1556, "lr": 4.850624106563181e-05}
{"train_loss": 0.08716081082820892, "global_step": 138552, "epoch": 1556, "lr": 4.850566158790172e-05}
{"train_loss": 0.14525239169597626, "global_step": 138553, "epoch": 1556, "lr": 4.8505082110372515e-05}
{"train_loss": 0.11769431829452515, "global_step": 138554, "epoch": 1556, "lr": 4.850450263304429e-05}
{"train_loss": 0.04301857948303223, "global_step": 138555, "epoch": 1556, "lr": 4.85039231559171e-05}
{"train_loss": 0.13343383371829987, "global_step": 138556, "epoch": 1556, "lr": 4.850334367899107e-05}
{"train_loss": 0.12641167640686035, "global_step": 138557, "epoch": 1556, "lr": 4.8502764202266224e-05}
{"train_loss": 0.1674506664276123, "global_step": 138558, "epoch": 1556, "lr": 4.8502184725742675e-05}
{"train_loss": 0.13414834439754486, "global_step": 138559, "epoch": 1556, "lr": 4.8501605249420476e-05}
{"train_loss": 0.14368368685245514, "global_step": 138560, "epoch": 1556, "lr": 4.850102577329972e-05}
{"train_loss": 0.07419086247682571, "global_step": 138561, "epoch": 1556, "lr": 4.8500446297380494e-05}
{"train_loss": 0.05715407803654671, "global_step": 138562, "epoch": 1556, "lr": 4.849986682166287e-05}
{"train_loss": 0.1873025894165039, "global_step": 138563, "epoch": 1556, "lr": 4.8499287346146895e-05}
{"train_loss": 0.15945599973201752, "global_step": 138564, "epoch": 1556, "lr": 4.849870787083269e-05}
{"train_loss": 0.1135939434170723, "global_step": 138565, "epoch": 1556, "lr": 4.849812839572031e-05}
{"train_loss": 0.11008608341217041, "global_step": 138566, "epoch": 1556, "lr": 4.849754892080984e-05}
{"train_loss": 0.12731562554836273, "global_step": 138567, "epoch": 1556, "lr": 4.849696944610135e-05}
{"train_loss": 0.1472475826740265, "global_step": 138568, "epoch": 1556, "lr": 4.849638997159494e-05}
{"train_loss": 0.16964934766292572, "global_step": 138569, "epoch": 1556, "lr": 4.849581049729065e-05}
{"train_loss": 0.09067167341709137, "global_step": 138570, "epoch": 1556, "lr": 4.849523102318858e-05}
{"train_loss": 0.08712027966976166, "global_step": 138571, "epoch": 1556, "lr": 4.8494651549288826e-05}
{"train_loss": 0.11827798091461149, "global_step": 138572, "epoch": 1556, "lr": 4.849407207559143e-05, "val_loss": 6.218445777893066}
{"train_loss": 0.1056152731180191, "global_step": 138573, "epoch": 1557, "lr": 4.8493492602096515e-05}
{"train_loss": 0.14020831882953644, "global_step": 138574, "epoch": 1557, "lr": 4.84929131288041e-05}
{"train_loss": 0.058435529470443726, "global_step": 138575, "epoch": 1557, "lr": 4.849233365571432e-05}
{"train_loss": 0.11456602811813354, "global_step": 138576, "epoch": 1557, "lr": 4.849175418282722e-05}
{"train_loss": 0.17123356461524963, "global_step": 138577, "epoch": 1557, "lr": 4.8491174710142885e-05}
{"train_loss": 0.07538311183452606, "global_step": 138578, "epoch": 1557, "lr": 4.849059523766139e-05}
{"train_loss": 0.13750147819519043, "global_step": 138579, "epoch": 1557, "lr": 4.849001576538282e-05}
{"train_loss": 0.06325159966945648, "global_step": 138580, "epoch": 1557, "lr": 4.848943629330724e-05}
{"train_loss": 0.10174515098333359, "global_step": 138581, "epoch": 1557, "lr": 4.848885682143475e-05}
{"train_loss": 0.08788532018661499, "global_step": 138582, "epoch": 1557, "lr": 4.84882773497654e-05}
{"train_loss": 0.15650774538516998, "global_step": 138583, "epoch": 1557, "lr": 4.8487697878299303e-05}
{"train_loss": 0.11723136156797409, "global_step": 138584, "epoch": 1557, "lr": 4.8487118407036493e-05}
{"train_loss": 0.10087604075670242, "global_step": 138585, "epoch": 1557, "lr": 4.848653893597709e-05}
{"train_loss": 0.18007425963878632, "global_step": 138586, "epoch": 1557, "lr": 4.848595946512113e-05}
{"train_loss": 0.12139547616243362, "global_step": 138587, "epoch": 1557, "lr": 4.8485379994468726e-05}
{"train_loss": 0.10766370594501495, "global_step": 138588, "epoch": 1557, "lr": 4.8484800524019955e-05}
{"train_loss": 0.10447756946086884, "global_step": 138589, "epoch": 1557, "lr": 4.848422105377487e-05}
{"train_loss": 0.08369208127260208, "global_step": 138590, "epoch": 1557, "lr": 4.848364158373358e-05}
{"train_loss": 0.08025709539651871, "global_step": 138591, "epoch": 1557, "lr": 4.848306211389613e-05}
{"train_loss": 0.09657236188650131, "global_step": 138592, "epoch": 1557, "lr": 4.848248264426263e-05}
{"train_loss": 0.1518314629793167, "global_step": 138593, "epoch": 1557, "lr": 4.848190317483312e-05}
{"train_loss": 0.12525574862957, "global_step": 138594, "epoch": 1557, "lr": 4.8481323705607714e-05}
{"train_loss": 0.11184009909629822, "global_step": 138595, "epoch": 1557, "lr": 4.848074423658648e-05}
{"train_loss": 0.10734745860099792, "global_step": 138596, "epoch": 1557, "lr": 4.848016476776948e-05}
{"train_loss": 0.06808370351791382, "global_step": 138597, "epoch": 1557, "lr": 4.847958529915681e-05}
{"train_loss": 0.07031209766864777, "global_step": 138598, "epoch": 1557, "lr": 4.847900583074854e-05}
{"train_loss": 0.12705466151237488, "global_step": 138599, "epoch": 1557, "lr": 4.847842636254474e-05}
{"train_loss": 0.1564127653837204, "global_step": 138600, "epoch": 1557, "lr": 4.847784689454552e-05}
{"train_loss": 0.09823370724916458, "global_step": 138601, "epoch": 1557, "lr": 4.8477267426750914e-05}
{"train_loss": 0.09001854807138443, "global_step": 138602, "epoch": 1557, "lr": 4.8476687959161035e-05}
{"train_loss": 0.07701359689235687, "global_step": 138603, "epoch": 1557, "lr": 4.847610849177592e-05}
{"train_loss": 0.08661570399999619, "global_step": 138604, "epoch": 1557, "lr": 4.847552902459571e-05}
{"train_loss": 0.1254286915063858, "global_step": 138605, "epoch": 1557, "lr": 4.847494955762042e-05}
{"train_loss": 0.13814866542816162, "global_step": 138606, "epoch": 1557, "lr": 4.847437009085015e-05}
{"train_loss": 0.11746145039796829, "global_step": 138607, "epoch": 1557, "lr": 4.8473790624285e-05}
{"train_loss": 0.12109848856925964, "global_step": 138608, "epoch": 1557, "lr": 4.8473211157925016e-05}
{"train_loss": 0.17705008387565613, "global_step": 138609, "epoch": 1557, "lr": 4.847263169177031e-05}
{"train_loss": 0.08478384464979172, "global_step": 138610, "epoch": 1557, "lr": 4.847205222582092e-05}
{"train_loss": 0.10904061049222946, "global_step": 138611, "epoch": 1557, "lr": 4.847147276007697e-05}
{"train_loss": 0.12236440926790237, "global_step": 138612, "epoch": 1557, "lr": 4.847089329453848e-05}
{"train_loss": 0.10912558436393738, "global_step": 138613, "epoch": 1557, "lr": 4.8470313829205585e-05}
{"train_loss": 0.07991693913936615, "global_step": 138614, "epoch": 1557, "lr": 4.846973436407833e-05}
{"train_loss": 0.07989563792943954, "global_step": 138615, "epoch": 1557, "lr": 4.84691548991568e-05}
{"train_loss": 0.1117362529039383, "global_step": 138616, "epoch": 1557, "lr": 4.8468575434441075e-05}
{"train_loss": 0.10781443119049072, "global_step": 138617, "epoch": 1557, "lr": 4.8467995969931225e-05}
{"train_loss": 0.09558728337287903, "global_step": 138618, "epoch": 1557, "lr": 4.846741650562733e-05}
{"train_loss": 0.1165754646062851, "global_step": 138619, "epoch": 1557, "lr": 4.8466837041529486e-05}
{"train_loss": 0.10179299861192703, "global_step": 138620, "epoch": 1557, "lr": 4.8466257577637747e-05}
{"train_loss": 0.15356378257274628, "global_step": 138621, "epoch": 1557, "lr": 4.8465678113952215e-05}
{"train_loss": 0.0819772481918335, "global_step": 138622, "epoch": 1557, "lr": 4.8465098650472934e-05}
{"train_loss": 0.07718722522258759, "global_step": 138623, "epoch": 1557, "lr": 4.84645191872e-05}
{"train_loss": 0.07091952860355377, "global_step": 138624, "epoch": 1557, "lr": 4.846393972413352e-05}
{"train_loss": 0.14365693926811218, "global_step": 138625, "epoch": 1557, "lr": 4.846336026127352e-05}
{"train_loss": 0.07791146636009216, "global_step": 138626, "epoch": 1557, "lr": 4.846278079862011e-05}
{"train_loss": 0.1968531757593155, "global_step": 138627, "epoch": 1557, "lr": 4.846220133617336e-05}
{"train_loss": 0.168686643242836, "global_step": 138628, "epoch": 1557, "lr": 4.846162187393336e-05}
{"train_loss": 0.0818178728222847, "global_step": 138629, "epoch": 1557, "lr": 4.846104241190015e-05}
{"train_loss": 0.06682462245225906, "global_step": 138630, "epoch": 1557, "lr": 4.8460462950073866e-05}
{"train_loss": 0.12527941167354584, "global_step": 138631, "epoch": 1557, "lr": 4.8459883488454526e-05}
{"train_loss": 0.11975596845149994, "global_step": 138632, "epoch": 1557, "lr": 4.8459304027042255e-05}
{"train_loss": 0.08005796372890472, "global_step": 138633, "epoch": 1557, "lr": 4.84587245658371e-05}
{"train_loss": 0.18324518203735352, "global_step": 138634, "epoch": 1557, "lr": 4.8458145104839164e-05}
{"train_loss": 0.0921047255396843, "global_step": 138635, "epoch": 1557, "lr": 4.84575656440485e-05}
{"train_loss": 0.08083343505859375, "global_step": 138636, "epoch": 1557, "lr": 4.8456986183465204e-05}
{"train_loss": 0.19361993670463562, "global_step": 138637, "epoch": 1557, "lr": 4.845640672308934e-05}
{"train_loss": 0.10710281878709793, "global_step": 138638, "epoch": 1557, "lr": 4.8455827262921e-05}
{"train_loss": 0.10994886606931686, "global_step": 138639, "epoch": 1557, "lr": 4.845524780296025e-05}
{"train_loss": 0.10136178880929947, "global_step": 138640, "epoch": 1557, "lr": 4.8454668343207194e-05}
{"train_loss": 0.12391362339258194, "global_step": 138641, "epoch": 1557, "lr": 4.845408888366186e-05}
{"train_loss": 0.11358801275491714, "global_step": 138642, "epoch": 1557, "lr": 4.845350942432435e-05}
{"train_loss": 0.1160627231001854, "global_step": 138643, "epoch": 1557, "lr": 4.845292996519478e-05}
{"train_loss": 0.09943924099206924, "global_step": 138644, "epoch": 1557, "lr": 4.845235050627316e-05}
{"train_loss": 0.1365603655576706, "global_step": 138645, "epoch": 1557, "lr": 4.845177104755963e-05}
{"train_loss": 0.0762084499001503, "global_step": 138646, "epoch": 1557, "lr": 4.845119158905422e-05}
{"train_loss": 0.19264043867588043, "global_step": 138647, "epoch": 1557, "lr": 4.845061213075705e-05}
{"train_loss": 0.08865261822938919, "global_step": 138648, "epoch": 1557, "lr": 4.845003267266816e-05}
{"train_loss": 0.11038972437381744, "global_step": 138649, "epoch": 1557, "lr": 4.8449453214787656e-05}
{"train_loss": 0.0916985347867012, "global_step": 138650, "epoch": 1557, "lr": 4.8448873757115584e-05}
{"train_loss": 0.1698547750711441, "global_step": 138651, "epoch": 1557, "lr": 4.8448294299652066e-05}
{"train_loss": 0.1352289915084839, "global_step": 138652, "epoch": 1557, "lr": 4.8447714842397135e-05}
{"train_loss": 0.12674568593502045, "global_step": 138653, "epoch": 1557, "lr": 4.844713538535091e-05}
{"train_loss": 0.1449984610080719, "global_step": 138654, "epoch": 1557, "lr": 4.844655592851344e-05}
{"train_loss": 0.05875881761312485, "global_step": 138655, "epoch": 1557, "lr": 4.844597647188481e-05}
{"train_loss": 0.1141832172870636, "global_step": 138656, "epoch": 1557, "lr": 4.84453970154651e-05}
{"train_loss": 0.08890197426080704, "global_step": 138657, "epoch": 1557, "lr": 4.8444817559254385e-05}
{"train_loss": 0.14361700415611267, "global_step": 138658, "epoch": 1557, "lr": 4.844423810325275e-05}
{"train_loss": 0.12546327710151672, "global_step": 138659, "epoch": 1557, "lr": 4.844365864746028e-05}
{"train_loss": 0.12018982321023941, "global_step": 138660, "epoch": 1557, "lr": 4.844307919187701e-05}
{"train_loss": 0.11285381320487248, "global_step": 138661, "epoch": 1557, "lr": 4.8442499736503055e-05, "val_loss": 5.8597283363342285}
{"train_loss": 0.14446592330932617, "global_step": 138662, "epoch": 1558, "lr": 4.844192028133851e-05}
{"train_loss": 0.1095159500837326, "global_step": 138663, "epoch": 1558, "lr": 4.8441340826383405e-05}
{"train_loss": 0.1554984599351883, "global_step": 138664, "epoch": 1558, "lr": 4.844076137163787e-05}
{"train_loss": 0.12740004062652588, "global_step": 138665, "epoch": 1558, "lr": 4.844018191710193e-05}
{"train_loss": 0.13527841866016388, "global_step": 138666, "epoch": 1558, "lr": 4.8439602462775714e-05}
{"train_loss": 0.049688056111335754, "global_step": 138667, "epoch": 1558, "lr": 4.843902300865924e-05}
{"train_loss": 0.08912338316440582, "global_step": 138668, "epoch": 1558, "lr": 4.843844355475265e-05}
{"train_loss": 0.08892388641834259, "global_step": 138669, "epoch": 1558, "lr": 4.843786410105598e-05}
{"train_loss": 0.12335889041423798, "global_step": 138670, "epoch": 1558, "lr": 4.8437284647569333e-05}
{"train_loss": 0.1656995713710785, "global_step": 138671, "epoch": 1558, "lr": 4.843670519429275e-05}
{"train_loss": 0.09476344287395477, "global_step": 138672, "epoch": 1558, "lr": 4.843612574122636e-05}
{"train_loss": 0.1278003603219986, "global_step": 138673, "epoch": 1558, "lr": 4.84355462883702e-05}
{"train_loss": 0.08984286338090897, "global_step": 138674, "epoch": 1558, "lr": 4.843496683572436e-05}
{"train_loss": 0.1252698004245758, "global_step": 138675, "epoch": 1558, "lr": 4.8434387383288923e-05}
{"train_loss": 0.08452179282903671, "global_step": 138676, "epoch": 1558, "lr": 4.8433807931063965e-05}
{"train_loss": 0.05946839973330498, "global_step": 138677, "epoch": 1558, "lr": 4.843322847904956e-05}
{"train_loss": 0.17107152938842773, "global_step": 138678, "epoch": 1558, "lr": 4.843264902724577e-05}
{"train_loss": 0.1217958927154541, "global_step": 138679, "epoch": 1558, "lr": 4.843206957565273e-05}
{"train_loss": 0.1212620735168457, "global_step": 138680, "epoch": 1558, "lr": 4.8431490124270445e-05}
{"train_loss": 0.060681987553834915, "global_step": 138681, "epoch": 1558, "lr": 4.843091067309905e-05}
{"train_loss": 0.12172829359769821, "global_step": 138682, "epoch": 1558, "lr": 4.843033122213858e-05}
{"train_loss": 0.10439871996641159, "global_step": 138683, "epoch": 1558, "lr": 4.842975177138915e-05}
{"train_loss": 0.06574960798025131, "global_step": 138684, "epoch": 1558, "lr": 4.8429172320850806e-05}
{"train_loss": 0.15352478623390198, "global_step": 138685, "epoch": 1558, "lr": 4.842859287052366e-05}
{"train_loss": 0.13018231093883514, "global_step": 138686, "epoch": 1558, "lr": 4.842801342040775e-05}
{"train_loss": 0.11888866126537323, "global_step": 138687, "epoch": 1558, "lr": 4.842743397050319e-05}
{"train_loss": 0.08075196295976639, "global_step": 138688, "epoch": 1558, "lr": 4.842685452081003e-05}
{"train_loss": 0.11296553164720535, "global_step": 138689, "epoch": 1558, "lr": 4.842627507132837e-05}
{"train_loss": 0.06476224958896637, "global_step": 138690, "epoch": 1558, "lr": 4.842569562205827e-05}
{"train_loss": 0.13609561324119568, "global_step": 138691, "epoch": 1558, "lr": 4.842511617299983e-05}
{"train_loss": 0.15083461999893188, "global_step": 138692, "epoch": 1558, "lr": 4.842453672415311e-05}
{"train_loss": 0.11848125606775284, "global_step": 138693, "epoch": 1558, "lr": 4.8423957275518186e-05}
{"train_loss": 0.1130778044462204, "global_step": 138694, "epoch": 1558, "lr": 4.842337782709514e-05}
{"train_loss": 0.22155620157718658, "global_step": 138695, "epoch": 1558, "lr": 4.842279837888405e-05}
{"train_loss": 0.0843404233455658, "global_step": 138696, "epoch": 1558, "lr": 4.8422218930885e-05}
{"train_loss": 0.0862332433462143, "global_step": 138697, "epoch": 1558, "lr": 4.842163948309805e-05}
{"train_loss": 0.1346374750137329, "global_step": 138698, "epoch": 1558, "lr": 4.842106003552332e-05}
{"train_loss": 0.18535245954990387, "global_step": 138699, "epoch": 1558, "lr": 4.8420480588160834e-05}
{"train_loss": 0.08943670243024826, "global_step": 138700, "epoch": 1558, "lr": 4.841990114101071e-05}
{"train_loss": 0.1386851817369461, "global_step": 138701, "epoch": 1558, "lr": 4.8419321694072996e-05}
{"train_loss": 0.06418255716562271, "global_step": 138702, "epoch": 1558, "lr": 4.84187422473478e-05}
{"train_loss": 0.10522373765707016, "global_step": 138703, "epoch": 1558, "lr": 4.8418162800835173e-05}
{"train_loss": 0.0398118682205677, "global_step": 138704, "epoch": 1558, "lr": 4.8417583354535224e-05}
{"train_loss": 0.0926196277141571, "global_step": 138705, "epoch": 1558, "lr": 4.841700390844798e-05}
{"train_loss": 0.09698765724897385, "global_step": 138706, "epoch": 1558, "lr": 4.841642446257358e-05}
{"train_loss": 0.13772691786289215, "global_step": 138707, "epoch": 1558, "lr": 4.841584501691205e-05}
{"train_loss": 0.06936806440353394, "global_step": 138708, "epoch": 1558, "lr": 4.841526557146352e-05}
{"train_loss": 0.18770159780979156, "global_step": 138709, "epoch": 1558, "lr": 4.841468612622801e-05}
{"train_loss": 0.07774919271469116, "global_step": 138710, "epoch": 1558, "lr": 4.8414106681205646e-05}
{"train_loss": 0.18213117122650146, "global_step": 138711, "epoch": 1558, "lr": 4.8413527236396474e-05}
{"train_loss": 0.0689050704240799, "global_step": 138712, "epoch": 1558, "lr": 4.841294779180059e-05}
{"train_loss": 0.11823413521051407, "global_step": 138713, "epoch": 1558, "lr": 4.8412368347418054e-05}
{"train_loss": 0.14727231860160828, "global_step": 138714, "epoch": 1558, "lr": 4.8411788903248955e-05}
{"train_loss": 0.0865020826458931, "global_step": 138715, "epoch": 1558, "lr": 4.8411209459293397e-05}
{"train_loss": 0.13743658363819122, "global_step": 138716, "epoch": 1558, "lr": 4.84106300155514e-05}
{"train_loss": 0.09123176336288452, "global_step": 138717, "epoch": 1558, "lr": 4.8410050572023104e-05}
{"train_loss": 0.14102023839950562, "global_step": 138718, "epoch": 1558, "lr": 4.8409471128708525e-05}
{"train_loss": 0.1712653785943985, "global_step": 138719, "epoch": 1558, "lr": 4.840889168560781e-05}
{"train_loss": 0.14539189636707306, "global_step": 138720, "epoch": 1558, "lr": 4.840831224272097e-05}
{"train_loss": 0.08105725049972534, "global_step": 138721, "epoch": 1558, "lr": 4.840773280004813e-05}
{"train_loss": 0.07789016515016556, "global_step": 138722, "epoch": 1558, "lr": 4.840715335758934e-05}
{"train_loss": 0.1322147399187088, "global_step": 138723, "epoch": 1558, "lr": 4.84065739153447e-05}
{"train_loss": 0.11498712748289108, "global_step": 138724, "epoch": 1558, "lr": 4.8405994473314266e-05}
{"train_loss": 0.06505870819091797, "global_step": 138725, "epoch": 1558, "lr": 4.840541503149814e-05}
{"train_loss": 0.17437143623828888, "global_step": 138726, "epoch": 1558, "lr": 4.840483558989637e-05}
{"train_loss": 0.14458215236663818, "global_step": 138727, "epoch": 1558, "lr": 4.8404256148509067e-05}
{"train_loss": 0.1410355418920517, "global_step": 138728, "epoch": 1558, "lr": 4.8403676707336275e-05}
{"train_loss": 0.12477479130029678, "global_step": 138729, "epoch": 1558, "lr": 4.840309726637811e-05}
{"train_loss": 0.07466010004281998, "global_step": 138730, "epoch": 1558, "lr": 4.840251782563461e-05}
{"train_loss": 0.1294133961200714, "global_step": 138731, "epoch": 1558, "lr": 4.8401938385105886e-05}
{"train_loss": 0.13766343891620636, "global_step": 138732, "epoch": 1558, "lr": 4.8401358944792e-05}
{"train_loss": 0.12329971045255661, "global_step": 138733, "epoch": 1558, "lr": 4.840077950469301e-05}
{"train_loss": 0.11796564608812332, "global_step": 138734, "epoch": 1558, "lr": 4.840020006480905e-05}
{"train_loss": 0.14632514119148254, "global_step": 138735, "epoch": 1558, "lr": 4.8399620625140135e-05}
{"train_loss": 0.1497935801744461, "global_step": 138736, "epoch": 1558, "lr": 4.839904118568639e-05}
{"train_loss": 0.11573421955108643, "global_step": 138737, "epoch": 1558, "lr": 4.839846174644786e-05}
{"train_loss": 0.07269366085529327, "global_step": 138738, "epoch": 1558, "lr": 4.8397882307424656e-05}
{"train_loss": 0.10302494466304779, "global_step": 138739, "epoch": 1558, "lr": 4.8397302868616814e-05}
{"train_loss": 0.10726853460073471, "global_step": 138740, "epoch": 1558, "lr": 4.839672343002445e-05}
{"train_loss": 0.07741362601518631, "global_step": 138741, "epoch": 1558, "lr": 4.839614399164762e-05}
{"train_loss": 0.09586015343666077, "global_step": 138742, "epoch": 1558, "lr": 4.8395564553486425e-05}
{"train_loss": 0.12619155645370483, "global_step": 138743, "epoch": 1558, "lr": 4.8394985115540904e-05}
{"train_loss": 0.06639325618743896, "global_step": 138744, "epoch": 1558, "lr": 4.839440567781118e-05}
{"train_loss": 0.040501825511455536, "global_step": 138745, "epoch": 1558, "lr": 4.8393826240297286e-05}
{"train_loss": 0.22068656980991364, "global_step": 138746, "epoch": 1558, "lr": 4.8393246802999345e-05}
{"train_loss": 0.09424147009849548, "global_step": 138747, "epoch": 1558, "lr": 4.839266736591739e-05}
{"train_loss": 0.13839437067508698, "global_step": 138748, "epoch": 1558, "lr": 4.839208792905154e-05}
{"train_loss": 0.09710909426212311, "global_step": 138749, "epoch": 1558, "lr": 4.8391508492401836e-05}
{"train_loss": 0.11507328302505311, "global_step": 138750, "epoch": 1558, "lr": 4.8390929055968383e-05, "val_loss": 5.849431037902832}
{"train_loss": 0.09517986327409744, "global_step": 138751, "epoch": 1559, "lr": 4.8390349619751264e-05}
{"train_loss": 0.07720629870891571, "global_step": 138752, "epoch": 1559, "lr": 4.8389770183750516e-05}
{"train_loss": 0.1571376472711563, "global_step": 138753, "epoch": 1559, "lr": 4.838919074796627e-05}
{"train_loss": 0.09871229529380798, "global_step": 138754, "epoch": 1559, "lr": 4.8388611312398554e-05}
{"train_loss": 0.13276724517345428, "global_step": 138755, "epoch": 1559, "lr": 4.8388031877047495e-05}
{"train_loss": 0.11367104947566986, "global_step": 138756, "epoch": 1559, "lr": 4.838745244191312e-05}
{"train_loss": 0.08211065828800201, "global_step": 138757, "epoch": 1559, "lr": 4.838687300699555e-05}
{"train_loss": 0.093578040599823, "global_step": 138758, "epoch": 1559, "lr": 4.838629357229483e-05}
{"train_loss": 0.10195543617010117, "global_step": 138759, "epoch": 1559, "lr": 4.838571413781108e-05}
{"train_loss": 0.24661758542060852, "global_step": 138760, "epoch": 1559, "lr": 4.8385134703544324e-05}
{"train_loss": 0.0707259401679039, "global_step": 138761, "epoch": 1559, "lr": 4.838455526949468e-05}
{"train_loss": 0.1438906490802765, "global_step": 138762, "epoch": 1559, "lr": 4.8383975835662195e-05}
{"train_loss": 0.10673671215772629, "global_step": 138763, "epoch": 1559, "lr": 4.8383396402047e-05}
{"train_loss": 0.10041234642267227, "global_step": 138764, "epoch": 1559, "lr": 4.83828169686491e-05}
{"train_loss": 0.10353679955005646, "global_step": 138765, "epoch": 1559, "lr": 4.8382237535468636e-05}
{"train_loss": 0.14749301970005035, "global_step": 138766, "epoch": 1559, "lr": 4.838165810250565e-05}
{"train_loss": 0.06113508716225624, "global_step": 138767, "epoch": 1559, "lr": 4.838107866976024e-05}
{"train_loss": 0.1293582022190094, "global_step": 138768, "epoch": 1559, "lr": 4.8380499237232454e-05}
{"train_loss": 0.09957493096590042, "global_step": 138769, "epoch": 1559, "lr": 4.8379919804922405e-05}
{"train_loss": 0.12518371641635895, "global_step": 138770, "epoch": 1559, "lr": 4.837934037283016e-05}
{"train_loss": 0.1262064427137375, "global_step": 138771, "epoch": 1559, "lr": 4.837876094095578e-05}
{"train_loss": 0.0634433925151825, "global_step": 138772, "epoch": 1559, "lr": 4.837818150929937e-05}
{"train_loss": 0.13430942595005035, "global_step": 138773, "epoch": 1559, "lr": 4.837760207786097e-05}
{"train_loss": 0.17310109734535217, "global_step": 138774, "epoch": 1559, "lr": 4.8377022646640715e-05}
{"train_loss": 0.16994325816631317, "global_step": 138775, "epoch": 1559, "lr": 4.837644321563862e-05}
{"train_loss": 0.04884614050388336, "global_step": 138776, "epoch": 1559, "lr": 4.837586378485481e-05}
{"train_loss": 0.09893962740898132, "global_step": 138777, "epoch": 1559, "lr": 4.837528435428934e-05}
{"train_loss": 0.10025057941675186, "global_step": 138778, "epoch": 1559, "lr": 4.83747049239423e-05}
{"train_loss": 0.09912750124931335, "global_step": 138779, "epoch": 1559, "lr": 4.837412549381374e-05}
{"train_loss": 0.10711859166622162, "global_step": 138780, "epoch": 1559, "lr": 4.8373546063903783e-05}
{"train_loss": 0.1099701002240181, "global_step": 138781, "epoch": 1559, "lr": 4.837296663421246e-05}
{"train_loss": 0.09338171035051346, "global_step": 138782, "epoch": 1559, "lr": 4.83723872047399e-05}
{"train_loss": 0.11140572279691696, "global_step": 138783, "epoch": 1559, "lr": 4.837180777548613e-05}
{"train_loss": 0.07796253263950348, "global_step": 138784, "epoch": 1559, "lr": 4.8371228346451267e-05}
{"train_loss": 0.12545686960220337, "global_step": 138785, "epoch": 1559, "lr": 4.837064891763535e-05}
{"train_loss": 0.08562435209751129, "global_step": 138786, "epoch": 1559, "lr": 4.8370069489038505e-05}
{"train_loss": 0.10752256214618683, "global_step": 138787, "epoch": 1559, "lr": 4.8369490060660764e-05}
{"train_loss": 0.08544130623340607, "global_step": 138788, "epoch": 1559, "lr": 4.8368910632502234e-05}
{"train_loss": 0.11919274926185608, "global_step": 138789, "epoch": 1559, "lr": 4.8368331204563e-05}
{"train_loss": 0.07986239343881607, "global_step": 138790, "epoch": 1559, "lr": 4.836775177684309e-05}
{"train_loss": 0.1204356700181961, "global_step": 138791, "epoch": 1559, "lr": 4.836717234934264e-05}
{"train_loss": 0.14943376183509827, "global_step": 138792, "epoch": 1559, "lr": 4.836659292206169e-05}
{"train_loss": 0.12108052521944046, "global_step": 138793, "epoch": 1559, "lr": 4.836601349500035e-05}
{"train_loss": 0.13928645849227905, "global_step": 138794, "epoch": 1559, "lr": 4.8365434068158663e-05}
{"train_loss": 0.07747649401426315, "global_step": 138795, "epoch": 1559, "lr": 4.836485464153673e-05}
{"train_loss": 0.12849858403205872, "global_step": 138796, "epoch": 1559, "lr": 4.836427521513461e-05}
{"train_loss": 0.11370624601840973, "global_step": 138797, "epoch": 1559, "lr": 4.8363695788952425e-05}
{"train_loss": 0.10478003323078156, "global_step": 138798, "epoch": 1559, "lr": 4.8363116362990184e-05}
{"train_loss": 0.16811597347259521, "global_step": 138799, "epoch": 1559, "lr": 4.836253693724803e-05}
{"train_loss": 0.1293724924325943, "global_step": 138800, "epoch": 1559, "lr": 4.8361957511725995e-05}
{"train_loss": 0.1490761637687683, "global_step": 138801, "epoch": 1559, "lr": 4.836137808642419e-05}
{"train_loss": 0.063643679022789, "global_step": 138802, "epoch": 1559, "lr": 4.836079866134266e-05}
{"train_loss": 0.09614691138267517, "global_step": 138803, "epoch": 1559, "lr": 4.836021923648152e-05}
{"train_loss": 0.10700353235006332, "global_step": 138804, "epoch": 1559, "lr": 4.8359639811840806e-05}
{"train_loss": 0.1113748773932457, "global_step": 138805, "epoch": 1559, "lr": 4.835906038742063e-05}
{"train_loss": 0.17211222648620605, "global_step": 138806, "epoch": 1559, "lr": 4.8358480963221066e-05}
{"train_loss": 0.12102341651916504, "global_step": 138807, "epoch": 1559, "lr": 4.835790153924217e-05}
{"train_loss": 0.09113157540559769, "global_step": 138808, "epoch": 1559, "lr": 4.835732211548406e-05}
{"train_loss": 0.1100352481007576, "global_step": 138809, "epoch": 1559, "lr": 4.8356742691946756e-05}
{"train_loss": 0.09055481106042862, "global_step": 138810, "epoch": 1559, "lr": 4.8356163268630396e-05}
{"train_loss": 0.09285563975572586, "global_step": 138811, "epoch": 1559, "lr": 4.8355583845535e-05}
{"train_loss": 0.1350703239440918, "global_step": 138812, "epoch": 1559, "lr": 4.8355004422660704e-05}
{"train_loss": 0.11531221866607666, "global_step": 138813, "epoch": 1559, "lr": 4.835442500000753e-05}
{"train_loss": 0.15247732400894165, "global_step": 138814, "epoch": 1559, "lr": 4.835384557757561e-05}
{"train_loss": 0.12986217439174652, "global_step": 138815, "epoch": 1559, "lr": 4.8353266155364976e-05}
{"train_loss": 0.1295325607061386, "global_step": 138816, "epoch": 1559, "lr": 4.835268673337574e-05}
{"train_loss": 0.12963910400867462, "global_step": 138817, "epoch": 1559, "lr": 4.835210731160795e-05}
{"train_loss": 0.054211340844631195, "global_step": 138818, "epoch": 1559, "lr": 4.835152789006172e-05}
{"train_loss": 0.09932310879230499, "global_step": 138819, "epoch": 1559, "lr": 4.8350948468737085e-05}
{"train_loss": 0.13463009893894196, "global_step": 138820, "epoch": 1559, "lr": 4.835036904763416e-05}
{"train_loss": 0.15477630496025085, "global_step": 138821, "epoch": 1559, "lr": 4.8349789626752996e-05}
{"train_loss": 0.1843588650226593, "global_step": 138822, "epoch": 1559, "lr": 4.8349210206093696e-05}
{"train_loss": 0.14826780557632446, "global_step": 138823, "epoch": 1559, "lr": 4.8348630785656305e-05}
{"train_loss": 0.13395170867443085, "global_step": 138824, "epoch": 1559, "lr": 4.834805136544093e-05}
{"train_loss": 0.08528605103492737, "global_step": 138825, "epoch": 1559, "lr": 4.834747194544765e-05}
{"train_loss": 0.13642698526382446, "global_step": 138826, "epoch": 1559, "lr": 4.8346892525676515e-05}
{"train_loss": 0.11358954757452011, "global_step": 138827, "epoch": 1559, "lr": 4.834631310612764e-05}
{"train_loss": 0.12303738296031952, "global_step": 138828, "epoch": 1559, "lr": 4.8345733686801064e-05}
{"train_loss": 0.09227140247821808, "global_step": 138829, "epoch": 1559, "lr": 4.834515426769691e-05}
{"train_loss": 0.0936180055141449, "global_step": 138830, "epoch": 1559, "lr": 4.83445748488152e-05}
{"train_loss": 0.17263327538967133, "global_step": 138831, "epoch": 1559, "lr": 4.834399543015606e-05}
{"train_loss": 0.1762229949235916, "global_step": 138832, "epoch": 1559, "lr": 4.834341601171953e-05}
{"train_loss": 0.12701255083084106, "global_step": 138833, "epoch": 1559, "lr": 4.834283659350573e-05}
{"train_loss": 0.08177046477794647, "global_step": 138834, "epoch": 1559, "lr": 4.83422571755147e-05}
{"train_loss": 0.12865398824214935, "global_step": 138835, "epoch": 1559, "lr": 4.834167775774655e-05}
{"train_loss": 0.1067875549197197, "global_step": 138836, "epoch": 1559, "lr": 4.834109834020132e-05}
{"train_loss": 0.11954484134912491, "global_step": 138837, "epoch": 1559, "lr": 4.8340518922879135e-05}
{"train_loss": 0.15160636603832245, "global_step": 138838, "epoch": 1559, "lr": 4.8339939505780016e-05}
{"train_loss": 0.11659726613525594, "global_step": 138839, "epoch": 1559, "lr": 4.833936008890411e-05, "val_loss": 5.9306464195251465}
{"train_loss": 0.10808169096708298, "global_step": 138840, "epoch": 1560, "lr": 4.8338780672251416e-05}
{"train_loss": 0.0807921439409256, "global_step": 138841, "epoch": 1560, "lr": 4.833820125582208e-05}
{"train_loss": 0.1500914841890335, "global_step": 138842, "epoch": 1560, "lr": 4.833762183961614e-05}
{"train_loss": 0.08013796806335449, "global_step": 138843, "epoch": 1560, "lr": 4.8337042423633686e-05}
{"train_loss": 0.06382429599761963, "global_step": 138844, "epoch": 1560, "lr": 4.8336463007874804e-05}
{"train_loss": 0.14587141573429108, "global_step": 138845, "epoch": 1560, "lr": 4.833588359233956e-05}
{"train_loss": 0.12074438482522964, "global_step": 138846, "epoch": 1560, "lr": 4.833530417702804e-05}
{"train_loss": 0.18776173889636993, "global_step": 138847, "epoch": 1560, "lr": 4.833472476194031e-05}
{"train_loss": 0.0905415490269661, "global_step": 138848, "epoch": 1560, "lr": 4.833414534707648e-05}
{"train_loss": 0.16949686408042908, "global_step": 138849, "epoch": 1560, "lr": 4.833356593243657e-05}
{"train_loss": 0.1483813226222992, "global_step": 138850, "epoch": 1560, "lr": 4.833298651802072e-05}
{"train_loss": 0.1282193660736084, "global_step": 138851, "epoch": 1560, "lr": 4.833240710382895e-05}
{"train_loss": 0.1507941037416458, "global_step": 138852, "epoch": 1560, "lr": 4.83318276898614e-05}
{"train_loss": 0.14617207646369934, "global_step": 138853, "epoch": 1560, "lr": 4.8331248276118084e-05}
{"train_loss": 0.13366639614105225, "global_step": 138854, "epoch": 1560, "lr": 4.833066886259914e-05}
{"train_loss": 0.09286951273679733, "global_step": 138855, "epoch": 1560, "lr": 4.833008944930459e-05}
{"train_loss": 0.09245976060628891, "global_step": 138856, "epoch": 1560, "lr": 4.832951003623457e-05}
{"train_loss": 0.1531272977590561, "global_step": 138857, "epoch": 1560, "lr": 4.8328930623389096e-05}
{"train_loss": 0.09319733083248138, "global_step": 138858, "epoch": 1560, "lr": 4.83283512107683e-05}
{"train_loss": 0.08005419373512268, "global_step": 138859, "epoch": 1560, "lr": 4.832777179837222e-05}
{"train_loss": 0.09284329414367676, "global_step": 138860, "epoch": 1560, "lr": 4.832719238620095e-05}
{"train_loss": 0.12714459002017975, "global_step": 138861, "epoch": 1560, "lr": 4.8326612974254584e-05}
{"train_loss": 0.13142314553260803, "global_step": 138862, "epoch": 1560, "lr": 4.832603356253318e-05}
{"train_loss": 0.1534169763326645, "global_step": 138863, "epoch": 1560, "lr": 4.832545415103682e-05}
{"train_loss": 0.11077098548412323, "global_step": 138864, "epoch": 1560, "lr": 4.832487473976558e-05}
{"train_loss": 0.11986467242240906, "global_step": 138865, "epoch": 1560, "lr": 4.832429532871954e-05}
{"train_loss": 0.10147702693939209, "global_step": 138866, "epoch": 1560, "lr": 4.832371591789878e-05}
{"train_loss": 0.11469724774360657, "global_step": 138867, "epoch": 1560, "lr": 4.8323136507303384e-05}
{"train_loss": 0.055281609296798706, "global_step": 138868, "epoch": 1560, "lr": 4.83225570969334e-05}
{"train_loss": 0.11061082780361176, "global_step": 138869, "epoch": 1560, "lr": 4.832197768678896e-05}
{"train_loss": 0.1390983909368515, "global_step": 138870, "epoch": 1560, "lr": 4.8321398276870075e-05}
{"train_loss": 0.05386759340763092, "global_step": 138871, "epoch": 1560, "lr": 4.8320818867176884e-05}
{"train_loss": 0.16370946168899536, "global_step": 138872, "epoch": 1560, "lr": 4.832023945770942e-05}
{"train_loss": 0.08745671063661575, "global_step": 138873, "epoch": 1560, "lr": 4.831966004846779e-05}
{"train_loss": 0.15093758702278137, "global_step": 138874, "epoch": 1560, "lr": 4.831908063945205e-05}
{"train_loss": 0.09363363683223724, "global_step": 138875, "epoch": 1560, "lr": 4.83185012306623e-05}
{"train_loss": 0.12528091669082642, "global_step": 138876, "epoch": 1560, "lr": 4.83179218220986e-05}
{"train_loss": 0.09763263165950775, "global_step": 138877, "epoch": 1560, "lr": 4.8317342413761026e-05}
{"train_loss": 0.12795349955558777, "global_step": 138878, "epoch": 1560, "lr": 4.831676300564969e-05}
{"train_loss": 0.13856397569179535, "global_step": 138879, "epoch": 1560, "lr": 4.831618359776462e-05}
{"train_loss": 0.09996891021728516, "global_step": 138880, "epoch": 1560, "lr": 4.831560419010594e-05}
{"train_loss": 0.10945602506399155, "global_step": 138881, "epoch": 1560, "lr": 4.8315024782673696e-05}
{"train_loss": 0.07076754420995712, "global_step": 138882, "epoch": 1560, "lr": 4.8314445375467983e-05}
{"train_loss": 0.11100539565086365, "global_step": 138883, "epoch": 1560, "lr": 4.831386596848886e-05}
{"train_loss": 0.11421561241149902, "global_step": 138884, "epoch": 1560, "lr": 4.831328656173643e-05}
{"train_loss": 0.10506454855203629, "global_step": 138885, "epoch": 1560, "lr": 4.831270715521075e-05}
{"train_loss": 0.14868247509002686, "global_step": 138886, "epoch": 1560, "lr": 4.831212774891192e-05}
{"train_loss": 0.15479877591133118, "global_step": 138887, "epoch": 1560, "lr": 4.831154834283998e-05}
{"train_loss": 0.1086745485663414, "global_step": 138888, "epoch": 1560, "lr": 4.8310968936995054e-05}
{"train_loss": 0.08204513043165207, "global_step": 138889, "epoch": 1560, "lr": 4.8310389531377176e-05}
{"train_loss": 0.08389860391616821, "global_step": 138890, "epoch": 1560, "lr": 4.8309810125986464e-05}
{"train_loss": 0.12723475694656372, "global_step": 138891, "epoch": 1560, "lr": 4.830923072082296e-05}
{"train_loss": 0.1361006498336792, "global_step": 138892, "epoch": 1560, "lr": 4.830865131588678e-05}
{"train_loss": 0.09677594155073166, "global_step": 138893, "epoch": 1560, "lr": 4.8308071911177966e-05}
{"train_loss": 0.08073543757200241, "global_step": 138894, "epoch": 1560, "lr": 4.830749250669662e-05}
{"train_loss": 0.12250230461359024, "global_step": 138895, "epoch": 1560, "lr": 4.83069131024428e-05}
{"train_loss": 0.06612709909677505, "global_step": 138896, "epoch": 1560, "lr": 4.830633369841659e-05}
{"train_loss": 0.09645013511180878, "global_step": 138897, "epoch": 1560, "lr": 4.830575429461809e-05}
{"train_loss": 0.15289616584777832, "global_step": 138898, "epoch": 1560, "lr": 4.830517489104735e-05}
{"train_loss": 0.13231469690799713, "global_step": 138899, "epoch": 1560, "lr": 4.830459548770447e-05}
{"train_loss": 0.09288344532251358, "global_step": 138900, "epoch": 1560, "lr": 4.8304016084589515e-05}
{"train_loss": 0.09177746623754501, "global_step": 138901, "epoch": 1560, "lr": 4.8303436681702566e-05}
{"train_loss": 0.10497282445430756, "global_step": 138902, "epoch": 1560, "lr": 4.830285727904369e-05}
{"train_loss": 0.07389867305755615, "global_step": 138903, "epoch": 1560, "lr": 4.8302277876612975e-05}
{"train_loss": 0.07968071103096008, "global_step": 138904, "epoch": 1560, "lr": 4.83016984744105e-05}
{"train_loss": 0.07943800091743469, "global_step": 138905, "epoch": 1560, "lr": 4.830111907243635e-05}
{"train_loss": 0.12091892212629318, "global_step": 138906, "epoch": 1560, "lr": 4.830053967069058e-05}
{"train_loss": 0.11204944550991058, "global_step": 138907, "epoch": 1560, "lr": 4.8299960269173293e-05}
{"train_loss": 0.11326701194047928, "global_step": 138908, "epoch": 1560, "lr": 4.829938086788454e-05}
{"train_loss": 0.18616020679473877, "global_step": 138909, "epoch": 1560, "lr": 4.829880146682444e-05}
{"train_loss": 0.16201941668987274, "global_step": 138910, "epoch": 1560, "lr": 4.829822206599301e-05}
{"train_loss": 0.13254521787166595, "global_step": 138911, "epoch": 1560, "lr": 4.8297642665390396e-05}
{"train_loss": 0.12770624458789825, "global_step": 138912, "epoch": 1560, "lr": 4.829706326501662e-05}
{"train_loss": 0.205045685172081, "global_step": 138913, "epoch": 1560, "lr": 4.82964838648718e-05}
{"train_loss": 0.10703539103269577, "global_step": 138914, "epoch": 1560, "lr": 4.8295904464955986e-05}
{"train_loss": 0.06763935834169388, "global_step": 138915, "epoch": 1560, "lr": 4.8295325065269255e-05}
{"train_loss": 0.16472798585891724, "global_step": 138916, "epoch": 1560, "lr": 4.8294745665811725e-05}
{"train_loss": 0.1370399445295334, "global_step": 138917, "epoch": 1560, "lr": 4.8294166266583416e-05}
{"train_loss": 0.08594221621751785, "global_step": 138918, "epoch": 1560, "lr": 4.829358686758447e-05}
{"train_loss": 0.130015566945076, "global_step": 138919, "epoch": 1560, "lr": 4.829300746881491e-05}
{"train_loss": 0.10770520567893982, "global_step": 138920, "epoch": 1560, "lr": 4.8292428070274836e-05}
{"train_loss": 0.138373002409935, "global_step": 138921, "epoch": 1560, "lr": 4.8291848671964324e-05}
{"train_loss": 0.11982938647270203, "global_step": 138922, "epoch": 1560, "lr": 4.829126927388345e-05}
{"train_loss": 0.12306874245405197, "global_step": 138923, "epoch": 1560, "lr": 4.8290689876032304e-05}
{"train_loss": 0.11511234194040298, "global_step": 138924, "epoch": 1560, "lr": 4.829011047841096e-05}
{"train_loss": 0.05833742022514343, "global_step": 138925, "epoch": 1560, "lr": 4.828953108101947e-05}
{"train_loss": 0.11074590682983398, "global_step": 138926, "epoch": 1560, "lr": 4.828895168385795e-05}
{"train_loss": 0.04865681007504463, "global_step": 138927, "epoch": 1560, "lr": 4.8288372286926434e-05}
{"train_loss": 0.11518238740188352, "global_step": 138928, "epoch": 1560, "lr": 4.828779289022505e-05, "val_loss": 5.889358043670654, "train_action_mse_error": 4.180087089538574}
{"train_loss": 0.06450333446264267, "global_step": 138929, "epoch": 1561, "lr": 4.828721349375383e-05}
{"train_loss": 0.08110132813453674, "global_step": 138930, "epoch": 1561, "lr": 4.828663409751289e-05}
{"train_loss": 0.15580448508262634, "global_step": 138931, "epoch": 1561, "lr": 4.8286054701502285e-05}
{"train_loss": 0.12725618481636047, "global_step": 138932, "epoch": 1561, "lr": 4.8285475305722084e-05}
{"train_loss": 0.17932303249835968, "global_step": 138933, "epoch": 1561, "lr": 4.828489591017241e-05}
{"train_loss": 0.13422441482543945, "global_step": 138934, "epoch": 1561, "lr": 4.828431651485328e-05}
{"train_loss": 0.11219702661037445, "global_step": 138935, "epoch": 1561, "lr": 4.8283737119764836e-05}
{"train_loss": 0.11048799753189087, "global_step": 138936, "epoch": 1561, "lr": 4.8283157724907094e-05}
{"train_loss": 0.06593059748411179, "global_step": 138937, "epoch": 1561, "lr": 4.8282578330280187e-05}
{"train_loss": 0.14399878680706024, "global_step": 138938, "epoch": 1561, "lr": 4.8281998935884146e-05}
{"train_loss": 0.12820705771446228, "global_step": 138939, "epoch": 1561, "lr": 4.828141954171908e-05}
{"train_loss": 0.05814334377646446, "global_step": 138940, "epoch": 1561, "lr": 4.828084014778505e-05}
{"train_loss": 0.10444021970033646, "global_step": 138941, "epoch": 1561, "lr": 4.828026075408214e-05}
{"train_loss": 0.13805152475833893, "global_step": 138942, "epoch": 1561, "lr": 4.827968136061043e-05}
{"train_loss": 0.16907595098018646, "global_step": 138943, "epoch": 1561, "lr": 4.827910196737001e-05}
{"train_loss": 0.07755700498819351, "global_step": 138944, "epoch": 1561, "lr": 4.827852257436092e-05}
{"train_loss": 0.15679241716861725, "global_step": 138945, "epoch": 1561, "lr": 4.827794318158328e-05}
{"train_loss": 0.08643104135990143, "global_step": 138946, "epoch": 1561, "lr": 4.827736378903713e-05}
{"train_loss": 0.11367984116077423, "global_step": 138947, "epoch": 1561, "lr": 4.8276784396722587e-05}
{"train_loss": 0.06299589574337006, "global_step": 138948, "epoch": 1561, "lr": 4.827620500463969e-05}
{"train_loss": 0.07380576431751251, "global_step": 138949, "epoch": 1561, "lr": 4.827562561278856e-05}
{"train_loss": 0.07754585891962051, "global_step": 138950, "epoch": 1561, "lr": 4.827504622116922e-05}
{"train_loss": 0.10349997878074646, "global_step": 138951, "epoch": 1561, "lr": 4.827446682978179e-05}
{"train_loss": 0.12000393122434616, "global_step": 138952, "epoch": 1561, "lr": 4.827388743862635e-05}
{"train_loss": 0.09029673039913177, "global_step": 138953, "epoch": 1561, "lr": 4.827330804770295e-05}
{"train_loss": 0.08038435876369476, "global_step": 138954, "epoch": 1561, "lr": 4.827272865701171e-05}
{"train_loss": 0.11612724512815475, "global_step": 138955, "epoch": 1561, "lr": 4.8272149266552645e-05}
{"train_loss": 0.1346246302127838, "global_step": 138956, "epoch": 1561, "lr": 4.82715698763259e-05}
{"train_loss": 0.14804579317569733, "global_step": 138957, "epoch": 1561, "lr": 4.82709904863315e-05}
{"train_loss": 0.09506017714738846, "global_step": 138958, "epoch": 1561, "lr": 4.827041109656956e-05}
{"train_loss": 0.13251248002052307, "global_step": 138959, "epoch": 1561, "lr": 4.826983170704013e-05}
{"train_loss": 0.10929001122713089, "global_step": 138960, "epoch": 1561, "lr": 4.82692523177433e-05}
{"train_loss": 0.0931873694062233, "global_step": 138961, "epoch": 1561, "lr": 4.826867292867915e-05}
{"train_loss": 0.11320976912975311, "global_step": 138962, "epoch": 1561, "lr": 4.8268093539847755e-05}
{"train_loss": 0.08197376877069473, "global_step": 138963, "epoch": 1561, "lr": 4.82675141512492e-05}
{"train_loss": 0.07087188214063644, "global_step": 138964, "epoch": 1561, "lr": 4.826693476288356e-05}
{"train_loss": 0.06189587712287903, "global_step": 138965, "epoch": 1561, "lr": 4.8266355374750896e-05}
{"train_loss": 0.11035220324993134, "global_step": 138966, "epoch": 1561, "lr": 4.826577598685131e-05}
{"train_loss": 0.1778106689453125, "global_step": 138967, "epoch": 1561, "lr": 4.8265196599184855e-05}
{"train_loss": 0.18737471103668213, "global_step": 138968, "epoch": 1561, "lr": 4.826461721175162e-05}
{"train_loss": 0.11416880786418915, "global_step": 138969, "epoch": 1561, "lr": 4.826403782455171e-05}
{"train_loss": 0.1498022824525833, "global_step": 138970, "epoch": 1561, "lr": 4.826345843758516e-05}
{"train_loss": 0.1016787514090538, "global_step": 138971, "epoch": 1561, "lr": 4.826287905085208e-05}
{"train_loss": 0.09011206030845642, "global_step": 138972, "epoch": 1561, "lr": 4.826229966435252e-05}
{"train_loss": 0.22137010097503662, "global_step": 138973, "epoch": 1561, "lr": 4.826172027808659e-05}
{"train_loss": 0.07352505624294281, "global_step": 138974, "epoch": 1561, "lr": 4.8261140892054335e-05}
{"train_loss": 0.14333190023899078, "global_step": 138975, "epoch": 1561, "lr": 4.826056150625586e-05}
{"train_loss": 0.09881925582885742, "global_step": 138976, "epoch": 1561, "lr": 4.8259982120691224e-05}
{"train_loss": 0.09825225174427032, "global_step": 138977, "epoch": 1561, "lr": 4.825940273536052e-05}
{"train_loss": 0.11203587055206299, "global_step": 138978, "epoch": 1561, "lr": 4.825882335026381e-05}
{"train_loss": 0.09025796502828598, "global_step": 138979, "epoch": 1561, "lr": 4.825824396540119e-05}
{"train_loss": 0.0748007744550705, "global_step": 138980, "epoch": 1561, "lr": 4.825766458077272e-05}
{"train_loss": 0.12986062467098236, "global_step": 138981, "epoch": 1561, "lr": 4.8257085196378485e-05}
{"train_loss": 0.13307762145996094, "global_step": 138982, "epoch": 1561, "lr": 4.8256505812218567e-05}
{"train_loss": 0.0895005539059639, "global_step": 138983, "epoch": 1561, "lr": 4.825592642829305e-05}
{"train_loss": 0.14350832998752594, "global_step": 138984, "epoch": 1561, "lr": 4.8255347044601984e-05}
{"train_loss": 0.20997503399848938, "global_step": 138985, "epoch": 1561, "lr": 4.825476766114548e-05}
{"train_loss": 0.1844833344221115, "global_step": 138986, "epoch": 1561, "lr": 4.825418827792358e-05}
{"train_loss": 0.1140158548951149, "global_step": 138987, "epoch": 1561, "lr": 4.825360889493639e-05}
{"train_loss": 0.07007394731044769, "global_step": 138988, "epoch": 1561, "lr": 4.8253029512184e-05}
{"train_loss": 0.11194790154695511, "global_step": 138989, "epoch": 1561, "lr": 4.825245012966644e-05}
{"train_loss": 0.09367832541465759, "global_step": 138990, "epoch": 1561, "lr": 4.825187074738385e-05}
{"train_loss": 0.16312752664089203, "global_step": 138991, "epoch": 1561, "lr": 4.8251291365336246e-05}
{"train_loss": 0.11261281371116638, "global_step": 138992, "epoch": 1561, "lr": 4.825071198352376e-05}
{"train_loss": 0.1622495800256729, "global_step": 138993, "epoch": 1561, "lr": 4.825013260194642e-05}
{"train_loss": 0.05323372781276703, "global_step": 138994, "epoch": 1561, "lr": 4.824955322060435e-05}
{"train_loss": 0.13011819124221802, "global_step": 138995, "epoch": 1561, "lr": 4.8248973839497584e-05}
{"train_loss": 0.08400947600603104, "global_step": 138996, "epoch": 1561, "lr": 4.824839445862625e-05}
{"train_loss": 0.12197810411453247, "global_step": 138997, "epoch": 1561, "lr": 4.824781507799038e-05}
{"train_loss": 0.12050101161003113, "global_step": 138998, "epoch": 1561, "lr": 4.8247235697590074e-05}
{"train_loss": 0.11446648091077805, "global_step": 138999, "epoch": 1561, "lr": 4.82466563174254e-05}
{"train_loss": 0.057006482034921646, "global_step": 139000, "epoch": 1561, "lr": 4.8246076937496445e-05}
{"train_loss": 0.11899103969335556, "global_step": 139001, "epoch": 1561, "lr": 4.824549755780329e-05}
{"train_loss": 0.0710873231291771, "global_step": 139002, "epoch": 1561, "lr": 4.8244918178346004e-05}
{"train_loss": 0.13759241998195648, "global_step": 139003, "epoch": 1561, "lr": 4.8244338799124654e-05}
{"train_loss": 0.1243690550327301, "global_step": 139004, "epoch": 1561, "lr": 4.8243759420139335e-05}
{"train_loss": 0.1221662163734436, "global_step": 139005, "epoch": 1561, "lr": 4.8243180041390134e-05}
{"train_loss": 0.09468276798725128, "global_step": 139006, "epoch": 1561, "lr": 4.82426006628771e-05}
{"train_loss": 0.09179369360208511, "global_step": 139007, "epoch": 1561, "lr": 4.824202128460035e-05}
{"train_loss": 0.1500108540058136, "global_step": 139008, "epoch": 1561, "lr": 4.824144190655991e-05}
{"train_loss": 0.1075219213962555, "global_step": 139009, "epoch": 1561, "lr": 4.8240862528755916e-05}
{"train_loss": 0.0784769207239151, "global_step": 139010, "epoch": 1561, "lr": 4.824028315118839e-05}
{"train_loss": 0.15252627432346344, "global_step": 139011, "epoch": 1561, "lr": 4.823970377385746e-05}
{"train_loss": 0.03672377020120621, "global_step": 139012, "epoch": 1561, "lr": 4.823912439676316e-05}
{"train_loss": 0.1139356940984726, "global_step": 139013, "epoch": 1561, "lr": 4.8238545019905605e-05}
{"train_loss": 0.1321316659450531, "global_step": 139014, "epoch": 1561, "lr": 4.8237965643284836e-05}
{"train_loss": 0.06763012707233429, "global_step": 139015, "epoch": 1561, "lr": 4.823738626690098e-05}
{"train_loss": 0.05941367149353027, "global_step": 139016, "epoch": 1561, "lr": 4.8236806890754066e-05}
{"train_loss": 0.11164415920718332, "global_step": 139017, "epoch": 1561, "lr": 4.8236227514844194e-05, "val_loss": 5.956677436828613}
{"train_loss": 0.08717001974582672, "global_step": 139018, "epoch": 1562, "lr": 4.8235648139171444e-05}
{"train_loss": 0.15690681338310242, "global_step": 139019, "epoch": 1562, "lr": 4.8235068763735884e-05}
{"train_loss": 0.09845787286758423, "global_step": 139020, "epoch": 1562, "lr": 4.82344893885376e-05}
{"train_loss": 0.08110901713371277, "global_step": 139021, "epoch": 1562, "lr": 4.823391001357668e-05}
{"train_loss": 0.13550734519958496, "global_step": 139022, "epoch": 1562, "lr": 4.823333063885317e-05}
{"train_loss": 0.1855596899986267, "global_step": 139023, "epoch": 1562, "lr": 4.823275126436716e-05}
{"train_loss": 0.055577926337718964, "global_step": 139024, "epoch": 1562, "lr": 4.823217189011876e-05}
{"train_loss": 0.07244434207677841, "global_step": 139025, "epoch": 1562, "lr": 4.8231592516107995e-05}
{"train_loss": 0.14869336783885956, "global_step": 139026, "epoch": 1562, "lr": 4.8231013142335005e-05}
{"train_loss": 0.07680566608905792, "global_step": 139027, "epoch": 1562, "lr": 4.823043376879981e-05}
{"train_loss": 0.09034454822540283, "global_step": 139028, "epoch": 1562, "lr": 4.8229854395502524e-05}
{"train_loss": 0.15635129809379578, "global_step": 139029, "epoch": 1562, "lr": 4.8229275022443196e-05}
{"train_loss": 0.061084307730197906, "global_step": 139030, "epoch": 1562, "lr": 4.822869564962194e-05}
{"train_loss": 0.1230321004986763, "global_step": 139031, "epoch": 1562, "lr": 4.82281162770388e-05}
{"train_loss": 0.10538359731435776, "global_step": 139032, "epoch": 1562, "lr": 4.822753690469388e-05}
{"train_loss": 0.07773631811141968, "global_step": 139033, "epoch": 1562, "lr": 4.822695753258723e-05}
{"train_loss": 0.1073513850569725, "global_step": 139034, "epoch": 1562, "lr": 4.822637816071897e-05}
{"train_loss": 0.1034947857260704, "global_step": 139035, "epoch": 1562, "lr": 4.822579878908913e-05}
{"train_loss": 0.1535429209470749, "global_step": 139036, "epoch": 1562, "lr": 4.822521941769782e-05}
{"train_loss": 0.12247000634670258, "global_step": 139037, "epoch": 1562, "lr": 4.82246400465451e-05}
{"train_loss": 0.0961032286286354, "global_step": 139038, "epoch": 1562, "lr": 4.8224060675631055e-05}
{"train_loss": 0.1116703674197197, "global_step": 139039, "epoch": 1562, "lr": 4.822348130495577e-05}
{"train_loss": 0.0732271671295166, "global_step": 139040, "epoch": 1562, "lr": 4.822290193451932e-05}
{"train_loss": 0.17719797790050507, "global_step": 139041, "epoch": 1562, "lr": 4.822232256432176e-05}
{"train_loss": 0.11393803358078003, "global_step": 139042, "epoch": 1562, "lr": 4.822174319436319e-05}
{"train_loss": 0.07724137604236603, "global_step": 139043, "epoch": 1562, "lr": 4.822116382464371e-05}
{"train_loss": 0.08388926088809967, "global_step": 139044, "epoch": 1562, "lr": 4.8220584455163346e-05}
{"train_loss": 0.12014015763998032, "global_step": 139045, "epoch": 1562, "lr": 4.822000508592222e-05}
{"train_loss": 0.1030401736497879, "global_step": 139046, "epoch": 1562, "lr": 4.8219425716920375e-05}
{"train_loss": 0.13312117755413055, "global_step": 139047, "epoch": 1562, "lr": 4.821884634815793e-05}
{"train_loss": 0.13490794599056244, "global_step": 139048, "epoch": 1562, "lr": 4.821826697963492e-05}
{"train_loss": 0.1029861718416214, "global_step": 139049, "epoch": 1562, "lr": 4.821768761135146e-05}
{"train_loss": 0.14538554847240448, "global_step": 139050, "epoch": 1562, "lr": 4.821710824330758e-05}
{"train_loss": 0.14982450008392334, "global_step": 139051, "epoch": 1562, "lr": 4.8216528875503424e-05}
{"train_loss": 0.1324807107448578, "global_step": 139052, "epoch": 1562, "lr": 4.821594950793901e-05}
{"train_loss": 0.047049034386873245, "global_step": 139053, "epoch": 1562, "lr": 4.821537014061446e-05}
{"train_loss": 0.191499724984169, "global_step": 139054, "epoch": 1562, "lr": 4.821479077352982e-05}
{"train_loss": 0.14316239953041077, "global_step": 139055, "epoch": 1562, "lr": 4.821421140668517e-05}
{"train_loss": 0.14688372611999512, "global_step": 139056, "epoch": 1562, "lr": 4.8213632040080606e-05}
{"train_loss": 0.2217763066291809, "global_step": 139057, "epoch": 1562, "lr": 4.82130526737162e-05}
{"train_loss": 0.08013616502285004, "global_step": 139058, "epoch": 1562, "lr": 4.8212473307592025e-05}
{"train_loss": 0.11705008149147034, "global_step": 139059, "epoch": 1562, "lr": 4.821189394170815e-05}
{"train_loss": 0.09110478311777115, "global_step": 139060, "epoch": 1562, "lr": 4.8211314576064684e-05}
{"train_loss": 0.1402449607849121, "global_step": 139061, "epoch": 1562, "lr": 4.821073521066167e-05}
{"train_loss": 0.13190296292304993, "global_step": 139062, "epoch": 1562, "lr": 4.821015584549922e-05}
{"train_loss": 0.14184214174747467, "global_step": 139063, "epoch": 1562, "lr": 4.8209576480577364e-05}
{"train_loss": 0.09514069557189941, "global_step": 139064, "epoch": 1562, "lr": 4.8208997115896236e-05}
{"train_loss": 0.1495947390794754, "global_step": 139065, "epoch": 1562, "lr": 4.820841775145587e-05}
{"train_loss": 0.13224419951438904, "global_step": 139066, "epoch": 1562, "lr": 4.820783838725637e-05}
{"train_loss": 0.041706714779138565, "global_step": 139067, "epoch": 1562, "lr": 4.820725902329779e-05}
{"train_loss": 0.1150587797164917, "global_step": 139068, "epoch": 1562, "lr": 4.820667965958024e-05}
{"train_loss": 0.19274553656578064, "global_step": 139069, "epoch": 1562, "lr": 4.820610029610377e-05}
{"train_loss": 0.13830305635929108, "global_step": 139070, "epoch": 1562, "lr": 4.8205520932868474e-05}
{"train_loss": 0.12863858044147491, "global_step": 139071, "epoch": 1562, "lr": 4.820494156987441e-05}
{"train_loss": 0.1490020453929901, "global_step": 139072, "epoch": 1562, "lr": 4.820436220712169e-05}
{"train_loss": 0.12387445569038391, "global_step": 139073, "epoch": 1562, "lr": 4.820378284461035e-05}
{"train_loss": 0.09758733212947845, "global_step": 139074, "epoch": 1562, "lr": 4.8203203482340514e-05}
{"train_loss": 0.10754081606864929, "global_step": 139075, "epoch": 1562, "lr": 4.8202624120312225e-05}
{"train_loss": 0.08225581049919128, "global_step": 139076, "epoch": 1562, "lr": 4.8202044758525564e-05}
{"train_loss": 0.16367006301879883, "global_step": 139077, "epoch": 1562, "lr": 4.820146539698062e-05}
{"train_loss": 0.09509749710559845, "global_step": 139078, "epoch": 1562, "lr": 4.820088603567746e-05}
{"train_loss": 0.1258680671453476, "global_step": 139079, "epoch": 1562, "lr": 4.820030667461619e-05}
{"train_loss": 0.14231902360916138, "global_step": 139080, "epoch": 1562, "lr": 4.819972731379684e-05}
{"train_loss": 0.09026212990283966, "global_step": 139081, "epoch": 1562, "lr": 4.819914795321954e-05}
{"train_loss": 0.12132205814123154, "global_step": 139082, "epoch": 1562, "lr": 4.819856859288431e-05}
{"train_loss": 0.12740367650985718, "global_step": 139083, "epoch": 1562, "lr": 4.8197989232791296e-05}
{"train_loss": 0.07946310937404633, "global_step": 139084, "epoch": 1562, "lr": 4.819740987294052e-05}
{"train_loss": 0.10139351338148117, "global_step": 139085, "epoch": 1562, "lr": 4.8196830513332094e-05}
{"train_loss": 0.09541735053062439, "global_step": 139086, "epoch": 1562, "lr": 4.819625115396606e-05}
{"train_loss": 0.1836203783750534, "global_step": 139087, "epoch": 1562, "lr": 4.819567179484254e-05}
{"train_loss": 0.1885683238506317, "global_step": 139088, "epoch": 1562, "lr": 4.8195092435961577e-05}
{"train_loss": 0.10109300166368484, "global_step": 139089, "epoch": 1562, "lr": 4.8194513077323276e-05}
{"train_loss": 0.09965524822473526, "global_step": 139090, "epoch": 1562, "lr": 4.819393371892768e-05}
{"train_loss": 0.08885214477777481, "global_step": 139091, "epoch": 1562, "lr": 4.819335436077491e-05}
{"train_loss": 0.09087468683719635, "global_step": 139092, "epoch": 1562, "lr": 4.8192775002865e-05}
{"train_loss": 0.11773783713579178, "global_step": 139093, "epoch": 1562, "lr": 4.8192195645198076e-05}
{"train_loss": 0.1293611079454422, "global_step": 139094, "epoch": 1562, "lr": 4.819161628777416e-05}
{"train_loss": 0.12242317199707031, "global_step": 139095, "epoch": 1562, "lr": 4.819103693059337e-05}
{"train_loss": 0.11662668734788895, "global_step": 139096, "epoch": 1562, "lr": 4.819045757365578e-05}
{"train_loss": 0.05517379939556122, "global_step": 139097, "epoch": 1562, "lr": 4.8189878216961456e-05}
{"train_loss": 0.08750570565462112, "global_step": 139098, "epoch": 1562, "lr": 4.818929886051049e-05}
{"train_loss": 0.2371634542942047, "global_step": 139099, "epoch": 1562, "lr": 4.8188719504302926e-05}
{"train_loss": 0.13479921221733093, "global_step": 139100, "epoch": 1562, "lr": 4.81881401483389e-05}
{"train_loss": 0.12750424444675446, "global_step": 139101, "epoch": 1562, "lr": 4.818756079261842e-05}
{"train_loss": 0.09160850942134857, "global_step": 139102, "epoch": 1562, "lr": 4.8186981437141635e-05}
{"train_loss": 0.09841527789831161, "global_step": 139103, "epoch": 1562, "lr": 4.818640208190856e-05}
{"train_loss": 0.09567496180534363, "global_step": 139104, "epoch": 1562, "lr": 4.818582272691933e-05}
{"train_loss": 0.09308461844921112, "global_step": 139105, "epoch": 1562, "lr": 4.818524337217396e-05}
{"train_loss": 0.11788505424609345, "global_step": 139106, "epoch": 1562, "lr": 4.818466401767259e-05, "val_loss": 5.898762226104736}
{"train_loss": 0.09156384319067001, "global_step": 139107, "epoch": 1563, "lr": 4.8184084663415254e-05}
{"train_loss": 0.10715090483427048, "global_step": 139108, "epoch": 1563, "lr": 4.818350530940206e-05}
{"train_loss": 0.1373724639415741, "global_step": 139109, "epoch": 1563, "lr": 4.818292595563305e-05}
{"train_loss": 0.13439683616161346, "global_step": 139110, "epoch": 1563, "lr": 4.818234660210835e-05}
{"train_loss": 0.11063659191131592, "global_step": 139111, "epoch": 1563, "lr": 4.818176724882798e-05}
{"train_loss": 0.15194475650787354, "global_step": 139112, "epoch": 1563, "lr": 4.818118789579208e-05}
{"train_loss": 0.10730472952127457, "global_step": 139113, "epoch": 1563, "lr": 4.818060854300068e-05}
{"train_loss": 0.06212765723466873, "global_step": 139114, "epoch": 1563, "lr": 4.8180029190453876e-05}
{"train_loss": 0.13243186473846436, "global_step": 139115, "epoch": 1563, "lr": 4.8179449838151755e-05}
{"train_loss": 0.12483056634664536, "global_step": 139116, "epoch": 1563, "lr": 4.817887048609437e-05}
{"train_loss": 0.058705031871795654, "global_step": 139117, "epoch": 1563, "lr": 4.817829113428183e-05}
{"train_loss": 0.1032264456152916, "global_step": 139118, "epoch": 1563, "lr": 4.817771178271417e-05}
{"train_loss": 0.09882143884897232, "global_step": 139119, "epoch": 1563, "lr": 4.8177132431391526e-05}
{"train_loss": 0.13814064860343933, "global_step": 139120, "epoch": 1563, "lr": 4.8176553080313916e-05}
{"train_loss": 0.09671155363321304, "global_step": 139121, "epoch": 1563, "lr": 4.817597372948148e-05}
{"train_loss": 0.05622297525405884, "global_step": 139122, "epoch": 1563, "lr": 4.8175394378894226e-05}
{"train_loss": 0.08685703575611115, "global_step": 139123, "epoch": 1563, "lr": 4.81748150285523e-05}
{"train_loss": 0.1423051953315735, "global_step": 139124, "epoch": 1563, "lr": 4.817423567845572e-05}
{"train_loss": 0.11536197364330292, "global_step": 139125, "epoch": 1563, "lr": 4.817365632860462e-05}
{"train_loss": 0.0842403769493103, "global_step": 139126, "epoch": 1563, "lr": 4.8173076978999016e-05}
{"train_loss": 0.10359784960746765, "global_step": 139127, "epoch": 1563, "lr": 4.817249762963905e-05}
{"train_loss": 0.12687020003795624, "global_step": 139128, "epoch": 1563, "lr": 4.817191828052475e-05}
{"train_loss": 0.09335219860076904, "global_step": 139129, "epoch": 1563, "lr": 4.817133893165623e-05}
{"train_loss": 0.15861517190933228, "global_step": 139130, "epoch": 1563, "lr": 4.8170759583033524e-05}
{"train_loss": 0.08331175148487091, "global_step": 139131, "epoch": 1563, "lr": 4.8170180234656756e-05}
{"train_loss": 0.1610826551914215, "global_step": 139132, "epoch": 1563, "lr": 4.8169600886525994e-05}
{"train_loss": 0.11380331963300705, "global_step": 139133, "epoch": 1563, "lr": 4.816902153864129e-05}
{"train_loss": 0.1337233930826187, "global_step": 139134, "epoch": 1563, "lr": 4.816844219100276e-05}
{"train_loss": 0.12931124866008759, "global_step": 139135, "epoch": 1563, "lr": 4.8167862843610425e-05}
{"train_loss": 0.14882546663284302, "global_step": 139136, "epoch": 1563, "lr": 4.816728349646443e-05}
{"train_loss": 0.1681951880455017, "global_step": 139137, "epoch": 1563, "lr": 4.81667041495648e-05}
{"train_loss": 0.17442196607589722, "global_step": 139138, "epoch": 1563, "lr": 4.8166124802911664e-05}
{"train_loss": 0.11824992299079895, "global_step": 139139, "epoch": 1563, "lr": 4.8165545456505034e-05}
{"train_loss": 0.18181228637695312, "global_step": 139140, "epoch": 1563, "lr": 4.816496611034505e-05}
{"train_loss": 0.14740116894245148, "global_step": 139141, "epoch": 1563, "lr": 4.816438676443175e-05}
{"train_loss": 0.13737890124320984, "global_step": 139142, "epoch": 1563, "lr": 4.8163807418765236e-05}
{"train_loss": 0.12306895852088928, "global_step": 139143, "epoch": 1563, "lr": 4.816322807334556e-05}
{"train_loss": 0.13465294241905212, "global_step": 139144, "epoch": 1563, "lr": 4.8162648728172835e-05}
{"train_loss": 0.1741059124469757, "global_step": 139145, "epoch": 1563, "lr": 4.8162069383247104e-05}
{"train_loss": 0.13433194160461426, "global_step": 139146, "epoch": 1563, "lr": 4.816149003856848e-05}
{"train_loss": 0.12767082452774048, "global_step": 139147, "epoch": 1563, "lr": 4.8160910694137e-05}
{"train_loss": 0.10214845836162567, "global_step": 139148, "epoch": 1563, "lr": 4.816033134995278e-05}
{"train_loss": 0.13935989141464233, "global_step": 139149, "epoch": 1563, "lr": 4.815975200601586e-05}
{"train_loss": 0.10539663583040237, "global_step": 139150, "epoch": 1563, "lr": 4.815917266232635e-05}
{"train_loss": 0.09685955196619034, "global_step": 139151, "epoch": 1563, "lr": 4.8158593318884334e-05}
{"train_loss": 0.14784075319766998, "global_step": 139152, "epoch": 1563, "lr": 4.815801397568984e-05}
{"train_loss": 0.13246500492095947, "global_step": 139153, "epoch": 1563, "lr": 4.815743463274301e-05}
{"train_loss": 0.16576655209064484, "global_step": 139154, "epoch": 1563, "lr": 4.8156855290043864e-05}
{"train_loss": 0.09017059952020645, "global_step": 139155, "epoch": 1563, "lr": 4.8156275947592526e-05}
{"train_loss": 0.07679685950279236, "global_step": 139156, "epoch": 1563, "lr": 4.8155696605389036e-05}
{"train_loss": 0.0762360617518425, "global_step": 139157, "epoch": 1563, "lr": 4.815511726343351e-05}
{"train_loss": 0.08926182985305786, "global_step": 139158, "epoch": 1563, "lr": 4.815453792172599e-05}
{"train_loss": 0.12578986585140228, "global_step": 139159, "epoch": 1563, "lr": 4.815395858026659e-05}
{"train_loss": 0.10823851078748703, "global_step": 139160, "epoch": 1563, "lr": 4.815337923905534e-05}
{"train_loss": 0.1714664101600647, "global_step": 139161, "epoch": 1563, "lr": 4.815279989809237e-05}
{"train_loss": 0.08054641634225845, "global_step": 139162, "epoch": 1563, "lr": 4.8152220557377716e-05}
{"train_loss": 0.1056680679321289, "global_step": 139163, "epoch": 1563, "lr": 4.815164121691149e-05}
{"train_loss": 0.13298489153385162, "global_step": 139164, "epoch": 1563, "lr": 4.815106187669373e-05}
{"train_loss": 0.10685187578201294, "global_step": 139165, "epoch": 1563, "lr": 4.8150482536724565e-05}
{"train_loss": 0.09469115734100342, "global_step": 139166, "epoch": 1563, "lr": 4.814990319700402e-05}
{"train_loss": 0.14737385511398315, "global_step": 139167, "epoch": 1563, "lr": 4.814932385753222e-05}
{"train_loss": 0.1284002959728241, "global_step": 139168, "epoch": 1563, "lr": 4.8148744518309194e-05}
{"train_loss": 0.16861867904663086, "global_step": 139169, "epoch": 1563, "lr": 4.814816517933507e-05}
{"train_loss": 0.1695459485054016, "global_step": 139170, "epoch": 1563, "lr": 4.81475858406099e-05}
{"train_loss": 0.0975676104426384, "global_step": 139171, "epoch": 1563, "lr": 4.814700650213375e-05}
{"train_loss": 0.13627979159355164, "global_step": 139172, "epoch": 1563, "lr": 4.814642716390674e-05}
{"train_loss": 0.11562565714120865, "global_step": 139173, "epoch": 1563, "lr": 4.814584782592889e-05}
{"train_loss": 0.07821228355169296, "global_step": 139174, "epoch": 1563, "lr": 4.814526848820033e-05}
{"train_loss": 0.16008056700229645, "global_step": 139175, "epoch": 1563, "lr": 4.814468915072109e-05}
{"train_loss": 0.1768682301044464, "global_step": 139176, "epoch": 1563, "lr": 4.8144109813491307e-05}
{"train_loss": 0.1378793865442276, "global_step": 139177, "epoch": 1563, "lr": 4.8143530476511e-05}
{"train_loss": 0.11857107281684875, "global_step": 139178, "epoch": 1563, "lr": 4.814295113978029e-05}
{"train_loss": 0.13640961050987244, "global_step": 139179, "epoch": 1563, "lr": 4.814237180329922e-05}
{"train_loss": 0.15541096031665802, "global_step": 139180, "epoch": 1563, "lr": 4.814179246706791e-05}
{"train_loss": 0.08967537432909012, "global_step": 139181, "epoch": 1563, "lr": 4.814121313108639e-05}
{"train_loss": 0.12589150667190552, "global_step": 139182, "epoch": 1563, "lr": 4.814063379535478e-05}
{"train_loss": 0.12875458598136902, "global_step": 139183, "epoch": 1563, "lr": 4.814005445987312e-05}
{"train_loss": 0.08838765323162079, "global_step": 139184, "epoch": 1563, "lr": 4.813947512464153e-05}
{"train_loss": 0.19797788560390472, "global_step": 139185, "epoch": 1563, "lr": 4.813889578966003e-05}
{"train_loss": 0.12484098225831985, "global_step": 139186, "epoch": 1563, "lr": 4.8138316454928755e-05}
{"train_loss": 0.15725550055503845, "global_step": 139187, "epoch": 1563, "lr": 4.813773712044776e-05}
{"train_loss": 0.1266966611146927, "global_step": 139188, "epoch": 1563, "lr": 4.8137157786217126e-05}
{"train_loss": 0.11834879964590073, "global_step": 139189, "epoch": 1563, "lr": 4.8136578452236924e-05}
{"train_loss": 0.13166184723377228, "global_step": 139190, "epoch": 1563, "lr": 4.813599911850723e-05}
{"train_loss": 0.17659622430801392, "global_step": 139191, "epoch": 1563, "lr": 4.8135419785028155e-05}
{"train_loss": 0.12263712286949158, "global_step": 139192, "epoch": 1563, "lr": 4.813484045179972e-05}
{"train_loss": 0.121268630027771, "global_step": 139193, "epoch": 1563, "lr": 4.813426111882206e-05}
{"train_loss": 0.07711050659418106, "global_step": 139194, "epoch": 1563, "lr": 4.813368178609519e-05}
{"train_loss": 0.12337465642878179, "global_step": 139195, "epoch": 1563, "lr": 4.813310245361926e-05, "val_loss": 5.886610984802246}
{"train_loss": 0.09474923461675644, "global_step": 139196, "epoch": 1564, "lr": 4.8132523121394285e-05}
{"train_loss": 0.13358575105667114, "global_step": 139197, "epoch": 1564, "lr": 4.8131943789420385e-05}
{"train_loss": 0.06639857590198517, "global_step": 139198, "epoch": 1564, "lr": 4.813136445769761e-05}
{"train_loss": 0.09487848728895187, "global_step": 139199, "epoch": 1564, "lr": 4.813078512622607e-05}
{"train_loss": 0.15687380731105804, "global_step": 139200, "epoch": 1564, "lr": 4.81302057950058e-05}
{"train_loss": 0.18756262958049774, "global_step": 139201, "epoch": 1564, "lr": 4.812962646403692e-05}
{"train_loss": 0.12591056525707245, "global_step": 139202, "epoch": 1564, "lr": 4.812904713331947e-05}
{"train_loss": 0.13282309472560883, "global_step": 139203, "epoch": 1564, "lr": 4.812846780285357e-05}
{"train_loss": 0.12756773829460144, "global_step": 139204, "epoch": 1564, "lr": 4.8127888472639246e-05}
{"train_loss": 0.09590445458889008, "global_step": 139205, "epoch": 1564, "lr": 4.812730914267662e-05}
{"train_loss": 0.1096983551979065, "global_step": 139206, "epoch": 1564, "lr": 4.812672981296576e-05}
{"train_loss": 0.10180948674678802, "global_step": 139207, "epoch": 1564, "lr": 4.812615048350673e-05}
{"train_loss": 0.15480394661426544, "global_step": 139208, "epoch": 1564, "lr": 4.812557115429962e-05}
{"train_loss": 0.08204390108585358, "global_step": 139209, "epoch": 1564, "lr": 4.8124991825344506e-05}
{"train_loss": 0.1002594381570816, "global_step": 139210, "epoch": 1564, "lr": 4.8124412496641474e-05}
{"train_loss": 0.10787689685821533, "global_step": 139211, "epoch": 1564, "lr": 4.812383316819057e-05}
{"train_loss": 0.1260443478822708, "global_step": 139212, "epoch": 1564, "lr": 4.812325383999191e-05}
{"train_loss": 0.11988886445760727, "global_step": 139213, "epoch": 1564, "lr": 4.812267451204554e-05}
{"train_loss": 0.1401451826095581, "global_step": 139214, "epoch": 1564, "lr": 4.8122095184351575e-05}
{"train_loss": 0.15185551345348358, "global_step": 139215, "epoch": 1564, "lr": 4.812151585691005e-05}
{"train_loss": 0.11970622837543488, "global_step": 139216, "epoch": 1564, "lr": 4.8120936529721084e-05}
{"train_loss": 0.09789109975099564, "global_step": 139217, "epoch": 1564, "lr": 4.812035720278471e-05}
{"train_loss": 0.05854488164186478, "global_step": 139218, "epoch": 1564, "lr": 4.8119777876101055e-05}
{"train_loss": 0.10266920924186707, "global_step": 139219, "epoch": 1564, "lr": 4.811919854967015e-05}
{"train_loss": 0.07547258585691452, "global_step": 139220, "epoch": 1564, "lr": 4.8118619223492126e-05}
{"train_loss": 0.11546630412340164, "global_step": 139221, "epoch": 1564, "lr": 4.8118039897567e-05}
{"train_loss": 0.11807531863451004, "global_step": 139222, "epoch": 1564, "lr": 4.8117460571894886e-05}
{"train_loss": 0.14840508997440338, "global_step": 139223, "epoch": 1564, "lr": 4.8116881246475876e-05}
{"train_loss": 0.1085437759757042, "global_step": 139224, "epoch": 1564, "lr": 4.811630192131001e-05}
{"train_loss": 0.10202806442975998, "global_step": 139225, "epoch": 1564, "lr": 4.81157225963974e-05}
{"train_loss": 0.10329582542181015, "global_step": 139226, "epoch": 1564, "lr": 4.81151432717381e-05}
{"train_loss": 0.09080547094345093, "global_step": 139227, "epoch": 1564, "lr": 4.811456394733219e-05}
{"train_loss": 0.0917191356420517, "global_step": 139228, "epoch": 1564, "lr": 4.811398462317976e-05}
{"train_loss": 0.07211395353078842, "global_step": 139229, "epoch": 1564, "lr": 4.81134052992809e-05}
{"train_loss": 0.1580105870962143, "global_step": 139230, "epoch": 1564, "lr": 4.811282597563564e-05}
{"train_loss": 0.08946797996759415, "global_step": 139231, "epoch": 1564, "lr": 4.811224665224411e-05}
{"train_loss": 0.142139732837677, "global_step": 139232, "epoch": 1564, "lr": 4.811166732910634e-05}
{"train_loss": 0.09956508129835129, "global_step": 139233, "epoch": 1564, "lr": 4.811108800622246e-05}
{"train_loss": 0.15692850947380066, "global_step": 139234, "epoch": 1564, "lr": 4.81105086835925e-05}
{"train_loss": 0.1592431366443634, "global_step": 139235, "epoch": 1564, "lr": 4.8109929361216575e-05}
{"train_loss": 0.13697940111160278, "global_step": 139236, "epoch": 1564, "lr": 4.810935003909473e-05}
{"train_loss": 0.10090143978595734, "global_step": 139237, "epoch": 1564, "lr": 4.810877071722708e-05}
{"train_loss": 0.0872887447476387, "global_step": 139238, "epoch": 1564, "lr": 4.810819139561366e-05}
{"train_loss": 0.08918886631727219, "global_step": 139239, "epoch": 1564, "lr": 4.8107612074254594e-05}
{"train_loss": 0.10674162209033966, "global_step": 139240, "epoch": 1564, "lr": 4.8107032753149916e-05}
{"train_loss": 0.11610868573188782, "global_step": 139241, "epoch": 1564, "lr": 4.810645343229972e-05}
{"train_loss": 0.13038791716098785, "global_step": 139242, "epoch": 1564, "lr": 4.8105874111704115e-05}
{"train_loss": 0.05271335691213608, "global_step": 139243, "epoch": 1564, "lr": 4.8105294791363135e-05}
{"train_loss": 0.07629720121622086, "global_step": 139244, "epoch": 1564, "lr": 4.810471547127688e-05}
{"train_loss": 0.12805552780628204, "global_step": 139245, "epoch": 1564, "lr": 4.810413615144542e-05}
{"train_loss": 0.14577877521514893, "global_step": 139246, "epoch": 1564, "lr": 4.8103556831868846e-05}
{"train_loss": 0.1378471851348877, "global_step": 139247, "epoch": 1564, "lr": 4.810297751254721e-05}
{"train_loss": 0.09815610945224762, "global_step": 139248, "epoch": 1564, "lr": 4.810239819348063e-05}
{"train_loss": 0.20340900123119354, "global_step": 139249, "epoch": 1564, "lr": 4.810181887466913e-05}
{"train_loss": 0.11327441036701202, "global_step": 139250, "epoch": 1564, "lr": 4.810123955611284e-05}
{"train_loss": 0.15216179192066193, "global_step": 139251, "epoch": 1564, "lr": 4.8100660237811794e-05}
{"train_loss": 0.10888051241636276, "global_step": 139252, "epoch": 1564, "lr": 4.810008091976612e-05}
{"train_loss": 0.13452355563640594, "global_step": 139253, "epoch": 1564, "lr": 4.809950160197583e-05}
{"train_loss": 0.15447349846363068, "global_step": 139254, "epoch": 1564, "lr": 4.809892228444107e-05}
{"train_loss": 0.12294049561023712, "global_step": 139255, "epoch": 1564, "lr": 4.8098342967161865e-05}
{"train_loss": 0.09369587153196335, "global_step": 139256, "epoch": 1564, "lr": 4.8097763650138334e-05}
{"train_loss": 0.11638981848955154, "global_step": 139257, "epoch": 1564, "lr": 4.809718433337052e-05}
{"train_loss": 0.08794662356376648, "global_step": 139258, "epoch": 1564, "lr": 4.809660501685851e-05}
{"train_loss": 0.12802809476852417, "global_step": 139259, "epoch": 1564, "lr": 4.809602570060241e-05}
{"train_loss": 0.1862940937280655, "global_step": 139260, "epoch": 1564, "lr": 4.8095446384602255e-05}
{"train_loss": 0.12050861120223999, "global_step": 139261, "epoch": 1564, "lr": 4.8094867068858163e-05}
{"train_loss": 0.13001827895641327, "global_step": 139262, "epoch": 1564, "lr": 4.809428775337019e-05}
{"train_loss": 0.11117333918809891, "global_step": 139263, "epoch": 1564, "lr": 4.809370843813841e-05}
{"train_loss": 0.16434353590011597, "global_step": 139264, "epoch": 1564, "lr": 4.809312912316291e-05}
{"train_loss": 0.24002322554588318, "global_step": 139265, "epoch": 1564, "lr": 4.809254980844377e-05}
{"train_loss": 0.13551802933216095, "global_step": 139266, "epoch": 1564, "lr": 4.8091970493981054e-05}
{"train_loss": 0.12426642328500748, "global_step": 139267, "epoch": 1564, "lr": 4.809139117977486e-05}
{"train_loss": 0.11124399304389954, "global_step": 139268, "epoch": 1564, "lr": 4.809081186582525e-05}
{"train_loss": 0.11471882462501526, "global_step": 139269, "epoch": 1564, "lr": 4.809023255213231e-05}
{"train_loss": 0.09650002419948578, "global_step": 139270, "epoch": 1564, "lr": 4.80896532386961e-05}
{"train_loss": 0.15405908226966858, "global_step": 139271, "epoch": 1564, "lr": 4.8089073925516736e-05}
{"train_loss": 0.12078747153282166, "global_step": 139272, "epoch": 1564, "lr": 4.8088494612594254e-05}
{"train_loss": 0.10181521624326706, "global_step": 139273, "epoch": 1564, "lr": 4.808791529992877e-05}
{"train_loss": 0.1259177327156067, "global_step": 139274, "epoch": 1564, "lr": 4.808733598752031e-05}
{"train_loss": 0.10453777015209198, "global_step": 139275, "epoch": 1564, "lr": 4.808675667536902e-05}
{"train_loss": 0.12834975123405457, "global_step": 139276, "epoch": 1564, "lr": 4.808617736347491e-05}
{"train_loss": 0.08472734689712524, "global_step": 139277, "epoch": 1564, "lr": 4.8085598051838105e-05}
{"train_loss": 0.11451893299818039, "global_step": 139278, "epoch": 1564, "lr": 4.808501874045867e-05}
{"train_loss": 0.1657128632068634, "global_step": 139279, "epoch": 1564, "lr": 4.808443942933668e-05}
{"train_loss": 0.11359098553657532, "global_step": 139280, "epoch": 1564, "lr": 4.808386011847222e-05}
{"train_loss": 0.12915274500846863, "global_step": 139281, "epoch": 1564, "lr": 4.8083280807865364e-05}
{"train_loss": 0.12896981835365295, "global_step": 139282, "epoch": 1564, "lr": 4.8082701497516175e-05}
{"train_loss": 0.09701356291770935, "global_step": 139283, "epoch": 1564, "lr": 4.808212218742475e-05}
{"train_loss": 0.12010326136029169, "global_step": 139284, "epoch": 1564, "lr": 4.808154287759116e-05, "val_loss": 5.877323150634766}
{"train_loss": 0.06327180564403534, "global_step": 139285, "epoch": 1565, "lr": 4.808096356801548e-05}
{"train_loss": 0.10996825993061066, "global_step": 139286, "epoch": 1565, "lr": 4.8080384258697805e-05}
{"train_loss": 0.12485868483781815, "global_step": 139287, "epoch": 1565, "lr": 4.807980494963817e-05}
{"train_loss": 0.09337334334850311, "global_step": 139288, "epoch": 1565, "lr": 4.807922564083671e-05}
{"train_loss": 0.10058030486106873, "global_step": 139289, "epoch": 1565, "lr": 4.8078646332293454e-05}
{"train_loss": 0.10850603878498077, "global_step": 139290, "epoch": 1565, "lr": 4.8078067024008525e-05}
{"train_loss": 0.0877823606133461, "global_step": 139291, "epoch": 1565, "lr": 4.807748771598195e-05}
{"train_loss": 0.08074488490819931, "global_step": 139292, "epoch": 1565, "lr": 4.807690840821385e-05}
{"train_loss": 0.1250707060098648, "global_step": 139293, "epoch": 1565, "lr": 4.807632910070427e-05}
{"train_loss": 0.10223469138145447, "global_step": 139294, "epoch": 1565, "lr": 4.807574979345333e-05}
{"train_loss": 0.13643503189086914, "global_step": 139295, "epoch": 1565, "lr": 4.807517048646105e-05}
{"train_loss": 0.07050052285194397, "global_step": 139296, "epoch": 1565, "lr": 4.807459117972755e-05}
{"train_loss": 0.14688673615455627, "global_step": 139297, "epoch": 1565, "lr": 4.807401187325291e-05}
{"train_loss": 0.11923398077487946, "global_step": 139298, "epoch": 1565, "lr": 4.807343256703718e-05}
{"train_loss": 0.128554567694664, "global_step": 139299, "epoch": 1565, "lr": 4.8072853261080474e-05}
{"train_loss": 0.10030379891395569, "global_step": 139300, "epoch": 1565, "lr": 4.8072273955382844e-05}
{"train_loss": 0.1154840812087059, "global_step": 139301, "epoch": 1565, "lr": 4.8071694649944365e-05}
{"train_loss": 0.11615178734064102, "global_step": 139302, "epoch": 1565, "lr": 4.8071115344765126e-05}
{"train_loss": 0.10519729554653168, "global_step": 139303, "epoch": 1565, "lr": 4.807053603984521e-05}
{"train_loss": 0.12229970842599869, "global_step": 139304, "epoch": 1565, "lr": 4.806995673518467e-05}
{"train_loss": 0.12690913677215576, "global_step": 139305, "epoch": 1565, "lr": 4.806937743078361e-05}
{"train_loss": 0.07318072021007538, "global_step": 139306, "epoch": 1565, "lr": 4.80687981266421e-05}
{"train_loss": 0.10616491734981537, "global_step": 139307, "epoch": 1565, "lr": 4.806821882276023e-05}
{"train_loss": 0.14347627758979797, "global_step": 139308, "epoch": 1565, "lr": 4.806763951913804e-05}
{"train_loss": 0.1631287932395935, "global_step": 139309, "epoch": 1565, "lr": 4.806706021577565e-05}
{"train_loss": 0.20936168730258942, "global_step": 139310, "epoch": 1565, "lr": 4.80664809126731e-05}
{"train_loss": 0.07829133421182632, "global_step": 139311, "epoch": 1565, "lr": 4.806590160983051e-05}
{"train_loss": 0.11363767087459564, "global_step": 139312, "epoch": 1565, "lr": 4.8065322307247916e-05}
{"train_loss": 0.09216627478599548, "global_step": 139313, "epoch": 1565, "lr": 4.8064743004925414e-05}
{"train_loss": 0.08474955707788467, "global_step": 139314, "epoch": 1565, "lr": 4.8064163702863107e-05}
{"train_loss": 0.12508393824100494, "global_step": 139315, "epoch": 1565, "lr": 4.806358440106103e-05}
{"train_loss": 0.10241259634494781, "global_step": 139316, "epoch": 1565, "lr": 4.80630050995193e-05}
{"train_loss": 0.14462482929229736, "global_step": 139317, "epoch": 1565, "lr": 4.806242579823795e-05}
{"train_loss": 0.15920168161392212, "global_step": 139318, "epoch": 1565, "lr": 4.806184649721711e-05}
{"train_loss": 0.07957226037979126, "global_step": 139319, "epoch": 1565, "lr": 4.8061267196456813e-05}
{"train_loss": 0.12046417593955994, "global_step": 139320, "epoch": 1565, "lr": 4.806068789595717e-05}
{"train_loss": 0.1267589032649994, "global_step": 139321, "epoch": 1565, "lr": 4.806010859571824e-05}
{"train_loss": 0.08449118584394455, "global_step": 139322, "epoch": 1565, "lr": 4.805952929574009e-05}
{"train_loss": 0.1619339883327484, "global_step": 139323, "epoch": 1565, "lr": 4.805894999602282e-05}
{"train_loss": 0.11214800179004669, "global_step": 139324, "epoch": 1565, "lr": 4.805837069656651e-05}
{"train_loss": 0.16355612874031067, "global_step": 139325, "epoch": 1565, "lr": 4.805779139737122e-05}
{"train_loss": 0.05993030592799187, "global_step": 139326, "epoch": 1565, "lr": 4.8057212098437044e-05}
{"train_loss": 0.19372674822807312, "global_step": 139327, "epoch": 1565, "lr": 4.8056632799764034e-05}
{"train_loss": 0.09353042393922806, "global_step": 139328, "epoch": 1565, "lr": 4.805605350135231e-05}
{"train_loss": 0.10324345529079437, "global_step": 139329, "epoch": 1565, "lr": 4.80554742032019e-05}
{"train_loss": 0.11050176620483398, "global_step": 139330, "epoch": 1565, "lr": 4.805489490531294e-05}
{"train_loss": 0.12802106142044067, "global_step": 139331, "epoch": 1565, "lr": 4.805431560768544e-05}
{"train_loss": 0.12133318930864334, "global_step": 139332, "epoch": 1565, "lr": 4.805373631031952e-05}
{"train_loss": 0.08279656618833542, "global_step": 139333, "epoch": 1565, "lr": 4.805315701321527e-05}
{"train_loss": 0.14644455909729004, "global_step": 139334, "epoch": 1565, "lr": 4.805257771637273e-05}
{"train_loss": 0.10925713926553726, "global_step": 139335, "epoch": 1565, "lr": 4.8051998419792024e-05}
{"train_loss": 0.11354539543390274, "global_step": 139336, "epoch": 1565, "lr": 4.8051419123473176e-05}
{"train_loss": 0.13221126794815063, "global_step": 139337, "epoch": 1565, "lr": 4.805083982741631e-05}
{"train_loss": 0.13667255640029907, "global_step": 139338, "epoch": 1565, "lr": 4.8050260531621466e-05}
{"train_loss": 0.1758180558681488, "global_step": 139339, "epoch": 1565, "lr": 4.804968123608876e-05}
{"train_loss": 0.08003699779510498, "global_step": 139340, "epoch": 1565, "lr": 4.804910194081824e-05}
{"train_loss": 0.20924285054206848, "global_step": 139341, "epoch": 1565, "lr": 4.8048522645810004e-05}
{"train_loss": 0.15281878411769867, "global_step": 139342, "epoch": 1565, "lr": 4.804794335106411e-05}
{"train_loss": 0.12415265291929245, "global_step": 139343, "epoch": 1565, "lr": 4.8047364056580654e-05}
{"train_loss": 0.13851483166217804, "global_step": 139344, "epoch": 1565, "lr": 4.80467847623597e-05}
{"train_loss": 0.12051437050104141, "global_step": 139345, "epoch": 1565, "lr": 4.804620546840134e-05}
{"train_loss": 0.14287307858467102, "global_step": 139346, "epoch": 1565, "lr": 4.8045626174705625e-05}
{"train_loss": 0.14155718684196472, "global_step": 139347, "epoch": 1565, "lr": 4.804504688127268e-05}
{"train_loss": 0.0992865115404129, "global_step": 139348, "epoch": 1565, "lr": 4.8044467588102524e-05}
{"train_loss": 0.08592148125171661, "global_step": 139349, "epoch": 1565, "lr": 4.8043888295195275e-05}
{"train_loss": 0.10772766172885895, "global_step": 139350, "epoch": 1565, "lr": 4.804330900255101e-05}
{"train_loss": 0.1223573237657547, "global_step": 139351, "epoch": 1565, "lr": 4.804272971016979e-05}
{"train_loss": 0.1714886575937271, "global_step": 139352, "epoch": 1565, "lr": 4.804215041805171e-05}
{"train_loss": 0.1705242544412613, "global_step": 139353, "epoch": 1565, "lr": 4.804157112619683e-05}
{"train_loss": 0.13220196962356567, "global_step": 139354, "epoch": 1565, "lr": 4.804099183460525e-05}
{"train_loss": 0.08614599704742432, "global_step": 139355, "epoch": 1565, "lr": 4.804041254327701e-05}
{"train_loss": 0.15999111533164978, "global_step": 139356, "epoch": 1565, "lr": 4.803983325221224e-05}
{"train_loss": 0.0684138685464859, "global_step": 139357, "epoch": 1565, "lr": 4.8039253961410975e-05}
{"train_loss": 0.10000373423099518, "global_step": 139358, "epoch": 1565, "lr": 4.8038674670873323e-05}
{"train_loss": 0.09625037759542465, "global_step": 139359, "epoch": 1565, "lr": 4.803809538059934e-05}
{"train_loss": 0.10988195985555649, "global_step": 139360, "epoch": 1565, "lr": 4.803751609058911e-05}
{"train_loss": 0.09800007939338684, "global_step": 139361, "epoch": 1565, "lr": 4.803693680084271e-05}
{"train_loss": 0.13193170726299286, "global_step": 139362, "epoch": 1565, "lr": 4.8036357511360216e-05}
{"train_loss": 0.06166255474090576, "global_step": 139363, "epoch": 1565, "lr": 4.8035778222141714e-05}
{"train_loss": 0.1546429544687271, "global_step": 139364, "epoch": 1565, "lr": 4.803519893318729e-05}
{"train_loss": 0.0884818285703659, "global_step": 139365, "epoch": 1565, "lr": 4.803461964449699e-05}
{"train_loss": 0.11134856939315796, "global_step": 139366, "epoch": 1565, "lr": 4.803404035607093e-05}
{"train_loss": 0.06943762302398682, "global_step": 139367, "epoch": 1565, "lr": 4.8033461067909145e-05}
{"train_loss": 0.12077803909778595, "global_step": 139368, "epoch": 1565, "lr": 4.803288178001174e-05}
{"train_loss": 0.1339292824268341, "global_step": 139369, "epoch": 1565, "lr": 4.803230249237881e-05}
{"train_loss": 0.10444235801696777, "global_step": 139370, "epoch": 1565, "lr": 4.803172320501039e-05}
{"train_loss": 0.09566717594861984, "global_step": 139371, "epoch": 1565, "lr": 4.8031143917906606e-05}
{"train_loss": 0.08187132328748703, "global_step": 139372, "epoch": 1565, "lr": 4.803056463106749e-05}
{"train_loss": 0.11837348304270359, "global_step": 139373, "epoch": 1565, "lr": 4.802998534449316e-05, "val_loss": 5.920333385467529, "train_action_mse_error": 9.945781707763672}
{"train_loss": 0.06273884326219559, "global_step": 139374, "epoch": 1566, "lr": 4.8029406058183656e-05}
{"train_loss": 0.14189349114894867, "global_step": 139375, "epoch": 1566, "lr": 4.802882677213909e-05}
{"train_loss": 0.1418728530406952, "global_step": 139376, "epoch": 1566, "lr": 4.8028247486359505e-05}
{"train_loss": 0.14038249850273132, "global_step": 139377, "epoch": 1566, "lr": 4.8027668200845024e-05}
{"train_loss": 0.09049680083990097, "global_step": 139378, "epoch": 1566, "lr": 4.8027088915595684e-05}
{"train_loss": 0.0904933512210846, "global_step": 139379, "epoch": 1566, "lr": 4.802650963061158e-05}
{"train_loss": 0.15583592653274536, "global_step": 139380, "epoch": 1566, "lr": 4.802593034589278e-05}
{"train_loss": 0.09072165191173553, "global_step": 139381, "epoch": 1566, "lr": 4.802535106143938e-05}
{"train_loss": 0.18697413802146912, "global_step": 139382, "epoch": 1566, "lr": 4.802477177725144e-05}
{"train_loss": 0.12766379117965698, "global_step": 139383, "epoch": 1566, "lr": 4.802419249332906e-05}
{"train_loss": 0.08800707012414932, "global_step": 139384, "epoch": 1566, "lr": 4.802361320967229e-05}
{"train_loss": 0.1501849740743637, "global_step": 139385, "epoch": 1566, "lr": 4.802303392628121e-05}
{"train_loss": 0.07796766608953476, "global_step": 139386, "epoch": 1566, "lr": 4.802245464315593e-05}
{"train_loss": 0.06781376898288727, "global_step": 139387, "epoch": 1566, "lr": 4.802187536029649e-05}
{"train_loss": 0.11249854415655136, "global_step": 139388, "epoch": 1566, "lr": 4.8021296077703004e-05}
{"train_loss": 0.1329776495695114, "global_step": 139389, "epoch": 1566, "lr": 4.802071679537551e-05}
{"train_loss": 0.11558448523283005, "global_step": 139390, "epoch": 1566, "lr": 4.802013751331413e-05}
{"train_loss": 0.08829894661903381, "global_step": 139391, "epoch": 1566, "lr": 4.8019558231518894e-05}
{"train_loss": 0.10270457714796066, "global_step": 139392, "epoch": 1566, "lr": 4.8018978949989926e-05}
{"train_loss": 0.1580614298582077, "global_step": 139393, "epoch": 1566, "lr": 4.801839966872726e-05}
{"train_loss": 0.1432049721479416, "global_step": 139394, "epoch": 1566, "lr": 4.801782038773101e-05}
{"train_loss": 0.11575527489185333, "global_step": 139395, "epoch": 1566, "lr": 4.801724110700123e-05}
{"train_loss": 0.1137208417057991, "global_step": 139396, "epoch": 1566, "lr": 4.801666182653803e-05}
{"train_loss": 0.1648184061050415, "global_step": 139397, "epoch": 1566, "lr": 4.801608254634145e-05}
{"train_loss": 0.14475560188293457, "global_step": 139398, "epoch": 1566, "lr": 4.8015503266411584e-05}
{"train_loss": 0.13241606950759888, "global_step": 139399, "epoch": 1566, "lr": 4.8014923986748514e-05}
{"train_loss": 0.08474025130271912, "global_step": 139400, "epoch": 1566, "lr": 4.801434470735231e-05}
{"train_loss": 0.1195988655090332, "global_step": 139401, "epoch": 1566, "lr": 4.8013765428223045e-05}
{"train_loss": 0.1316232532262802, "global_step": 139402, "epoch": 1566, "lr": 4.801318614936083e-05}
{"train_loss": 0.11642147600650787, "global_step": 139403, "epoch": 1566, "lr": 4.8012606870765684e-05}
{"train_loss": 0.14404985308647156, "global_step": 139404, "epoch": 1566, "lr": 4.801202759243773e-05}
{"train_loss": 0.15492640435695648, "global_step": 139405, "epoch": 1566, "lr": 4.801144831437704e-05}
{"train_loss": 0.15310075879096985, "global_step": 139406, "epoch": 1566, "lr": 4.8010869036583685e-05}
{"train_loss": 0.07591096311807632, "global_step": 139407, "epoch": 1566, "lr": 4.8010289759057754e-05}
{"train_loss": 0.1358175277709961, "global_step": 139408, "epoch": 1566, "lr": 4.8009710481799296e-05}
{"train_loss": 0.09799293428659439, "global_step": 139409, "epoch": 1566, "lr": 4.8009131204808425e-05}
{"train_loss": 0.05552161857485771, "global_step": 139410, "epoch": 1566, "lr": 4.800855192808519e-05}
{"train_loss": 0.10686296224594116, "global_step": 139411, "epoch": 1566, "lr": 4.80079726516297e-05}
{"train_loss": 0.15962852537631989, "global_step": 139412, "epoch": 1566, "lr": 4.800739337544199e-05}
{"train_loss": 0.1274711936712265, "global_step": 139413, "epoch": 1566, "lr": 4.800681409952218e-05}
{"train_loss": 0.1261064112186432, "global_step": 139414, "epoch": 1566, "lr": 4.800623482387031e-05}
{"train_loss": 0.1654939204454422, "global_step": 139415, "epoch": 1566, "lr": 4.80056555484865e-05}
{"train_loss": 0.10374131053686142, "global_step": 139416, "epoch": 1566, "lr": 4.80050762733708e-05}
{"train_loss": 0.11776906251907349, "global_step": 139417, "epoch": 1566, "lr": 4.8004496998523286e-05}
{"train_loss": 0.1582067310810089, "global_step": 139418, "epoch": 1566, "lr": 4.800391772394404e-05}
{"train_loss": 0.12155865877866745, "global_step": 139419, "epoch": 1566, "lr": 4.800333844963315e-05}
{"train_loss": 0.11520709842443466, "global_step": 139420, "epoch": 1566, "lr": 4.8002759175590684e-05}
{"train_loss": 0.1292121410369873, "global_step": 139421, "epoch": 1566, "lr": 4.800217990181672e-05}
{"train_loss": 0.09021513164043427, "global_step": 139422, "epoch": 1566, "lr": 4.800160062831135e-05}
{"train_loss": 0.1020665168762207, "global_step": 139423, "epoch": 1566, "lr": 4.800102135507461e-05}
{"train_loss": 0.09768155217170715, "global_step": 139424, "epoch": 1566, "lr": 4.800044208210665e-05}
{"train_loss": 0.22301284968852997, "global_step": 139425, "epoch": 1566, "lr": 4.7999862809407476e-05}
{"train_loss": 0.10513473302125931, "global_step": 139426, "epoch": 1566, "lr": 4.799928353697722e-05}
{"train_loss": 0.12307526171207428, "global_step": 139427, "epoch": 1566, "lr": 4.799870426481591e-05}
{"train_loss": 0.05467146635055542, "global_step": 139428, "epoch": 1566, "lr": 4.7998124992923675e-05}
{"train_loss": 0.12083715200424194, "global_step": 139429, "epoch": 1566, "lr": 4.799754572130054e-05}
{"train_loss": 0.1715582311153412, "global_step": 139430, "epoch": 1566, "lr": 4.799696644994664e-05}
{"train_loss": 0.15988482534885406, "global_step": 139431, "epoch": 1566, "lr": 4.7996387178862e-05}
{"train_loss": 0.1064152866601944, "global_step": 139432, "epoch": 1566, "lr": 4.799580790804674e-05}
{"train_loss": 0.0725742056965828, "global_step": 139433, "epoch": 1566, "lr": 4.79952286375009e-05}
{"train_loss": 0.10119659453630447, "global_step": 139434, "epoch": 1566, "lr": 4.79946493672246e-05}
{"train_loss": 0.08389240503311157, "global_step": 139435, "epoch": 1566, "lr": 4.799407009721787e-05}
{"train_loss": 0.12043313682079315, "global_step": 139436, "epoch": 1566, "lr": 4.799349082748084e-05}
{"train_loss": 0.11303259432315826, "global_step": 139437, "epoch": 1566, "lr": 4.7992911558013545e-05}
{"train_loss": 0.10998893529176712, "global_step": 139438, "epoch": 1566, "lr": 4.7992332288816075e-05}
{"train_loss": 0.07829941809177399, "global_step": 139439, "epoch": 1566, "lr": 4.799175301988851e-05}
{"train_loss": 0.11253078281879425, "global_step": 139440, "epoch": 1566, "lr": 4.799117375123093e-05}
{"train_loss": 0.07163810729980469, "global_step": 139441, "epoch": 1566, "lr": 4.7990594482843425e-05}
{"train_loss": 0.14614295959472656, "global_step": 139442, "epoch": 1566, "lr": 4.7990015214726044e-05}
{"train_loss": 0.12301348894834518, "global_step": 139443, "epoch": 1566, "lr": 4.79894359468789e-05}
{"train_loss": 0.10634759068489075, "global_step": 139444, "epoch": 1566, "lr": 4.798885667930203e-05}
{"train_loss": 0.07479491084814072, "global_step": 139445, "epoch": 1566, "lr": 4.798827741199555e-05}
{"train_loss": 0.1024971529841423, "global_step": 139446, "epoch": 1566, "lr": 4.798769814495951e-05}
{"train_loss": 0.2072945237159729, "global_step": 139447, "epoch": 1566, "lr": 4.798711887819402e-05}
{"train_loss": 0.06794165074825287, "global_step": 139448, "epoch": 1566, "lr": 4.798653961169911e-05}
{"train_loss": 0.07506220042705536, "global_step": 139449, "epoch": 1566, "lr": 4.7985960345474905e-05}
{"train_loss": 0.13032692670822144, "global_step": 139450, "epoch": 1566, "lr": 4.798538107952145e-05}
{"train_loss": 0.07531769573688507, "global_step": 139451, "epoch": 1566, "lr": 4.7984801813838855e-05}
{"train_loss": 0.16200381517410278, "global_step": 139452, "epoch": 1566, "lr": 4.7984222548427156e-05}
{"train_loss": 0.07706960290670395, "global_step": 139453, "epoch": 1566, "lr": 4.798364328328647e-05}
{"train_loss": 0.12620432674884796, "global_step": 139454, "epoch": 1566, "lr": 4.798306401841685e-05}
{"train_loss": 0.10624801367521286, "global_step": 139455, "epoch": 1566, "lr": 4.7982484753818394e-05}
{"train_loss": 0.09204092621803284, "global_step": 139456, "epoch": 1566, "lr": 4.7981905489491154e-05}
{"train_loss": 0.08743122220039368, "global_step": 139457, "epoch": 1566, "lr": 4.798132622543523e-05}
{"train_loss": 0.1063319519162178, "global_step": 139458, "epoch": 1566, "lr": 4.7980746961650686e-05}
{"train_loss": 0.1264529675245285, "global_step": 139459, "epoch": 1566, "lr": 4.79801676981376e-05}
{"train_loss": 0.11119541525840759, "global_step": 139460, "epoch": 1566, "lr": 4.797958843489608e-05}
{"train_loss": 0.1267150193452835, "global_step": 139461, "epoch": 1566, "lr": 4.797900917192615e-05}
{"train_loss": 0.11607598923565296, "global_step": 139462, "epoch": 1566, "lr": 4.797842990922795e-05, "val_loss": 6.025111198425293}
{"train_loss": 0.14017237722873688, "global_step": 139463, "epoch": 1567, "lr": 4.797785064680149e-05}
{"train_loss": 0.10445468872785568, "global_step": 139464, "epoch": 1567, "lr": 4.797727138464692e-05}
{"train_loss": 0.1505110263824463, "global_step": 139465, "epoch": 1567, "lr": 4.797669212276424e-05}
{"train_loss": 0.08201989531517029, "global_step": 139466, "epoch": 1567, "lr": 4.797611286115361e-05}
{"train_loss": 0.09477852284908295, "global_step": 139467, "epoch": 1567, "lr": 4.797553359981504e-05}
{"train_loss": 0.14071901142597198, "global_step": 139468, "epoch": 1567, "lr": 4.797495433874865e-05}
{"train_loss": 0.10195709019899368, "global_step": 139469, "epoch": 1567, "lr": 4.7974375077954486e-05}
{"train_loss": 0.13676950335502625, "global_step": 139470, "epoch": 1567, "lr": 4.797379581743267e-05}
{"train_loss": 0.1657603532075882, "global_step": 139471, "epoch": 1567, "lr": 4.797321655718321e-05}
{"train_loss": 0.057795584201812744, "global_step": 139472, "epoch": 1567, "lr": 4.7972637297206267e-05}
{"train_loss": 0.06791854649782181, "global_step": 139473, "epoch": 1567, "lr": 4.7972058037501855e-05}
{"train_loss": 0.082863949239254, "global_step": 139474, "epoch": 1567, "lr": 4.797147877807009e-05}
{"train_loss": 0.07830392569303513, "global_step": 139475, "epoch": 1567, "lr": 4.7970899518911034e-05}
{"train_loss": 0.07772468030452728, "global_step": 139476, "epoch": 1567, "lr": 4.797032026002475e-05}
{"train_loss": 0.08763589709997177, "global_step": 139477, "epoch": 1567, "lr": 4.796974100141135e-05}
{"train_loss": 0.07230537384748459, "global_step": 139478, "epoch": 1567, "lr": 4.7969161743070876e-05}
{"train_loss": 0.12995263934135437, "global_step": 139479, "epoch": 1567, "lr": 4.796858248500344e-05}
{"train_loss": 0.12085247784852982, "global_step": 139480, "epoch": 1567, "lr": 4.796800322720909e-05}
{"train_loss": 0.13379956781864166, "global_step": 139481, "epoch": 1567, "lr": 4.796742396968793e-05}
{"train_loss": 0.13182495534420013, "global_step": 139482, "epoch": 1567, "lr": 4.7966844712440005e-05}
{"train_loss": 0.1759633868932724, "global_step": 139483, "epoch": 1567, "lr": 4.796626545546544e-05}
{"train_loss": 0.10528784990310669, "global_step": 139484, "epoch": 1567, "lr": 4.796568619876426e-05}
{"train_loss": 0.15704026818275452, "global_step": 139485, "epoch": 1567, "lr": 4.796510694233659e-05}
{"train_loss": 0.10628383606672287, "global_step": 139486, "epoch": 1567, "lr": 4.7964527686182466e-05}
{"train_loss": 0.18303975462913513, "global_step": 139487, "epoch": 1567, "lr": 4.796394843030201e-05}
{"train_loss": 0.11803027242422104, "global_step": 139488, "epoch": 1567, "lr": 4.796336917469525e-05}
{"train_loss": 0.10648217052221298, "global_step": 139489, "epoch": 1567, "lr": 4.7962789919362315e-05}
{"train_loss": 0.13340404629707336, "global_step": 139490, "epoch": 1567, "lr": 4.7962210664303235e-05}
{"train_loss": 0.1366373747587204, "global_step": 139491, "epoch": 1567, "lr": 4.796163140951814e-05}
{"train_loss": 0.13177506625652313, "global_step": 139492, "epoch": 1567, "lr": 4.796105215500705e-05}
{"train_loss": 0.12446919083595276, "global_step": 139493, "epoch": 1567, "lr": 4.796047290077009e-05}
{"train_loss": 0.10007773339748383, "global_step": 139494, "epoch": 1567, "lr": 4.795989364680731e-05}
{"train_loss": 0.14810574054718018, "global_step": 139495, "epoch": 1567, "lr": 4.795931439311879e-05}
{"train_loss": 0.13186189532279968, "global_step": 139496, "epoch": 1567, "lr": 4.795873513970464e-05}
{"train_loss": 0.08688350021839142, "global_step": 139497, "epoch": 1567, "lr": 4.7958155886564884e-05}
{"train_loss": 0.09376587718725204, "global_step": 139498, "epoch": 1567, "lr": 4.795757663369966e-05}
{"train_loss": 0.1021941751241684, "global_step": 139499, "epoch": 1567, "lr": 4.795699738110899e-05}
{"train_loss": 0.07663574814796448, "global_step": 139500, "epoch": 1567, "lr": 4.7956418128792996e-05}
{"train_loss": 0.07127582281827927, "global_step": 139501, "epoch": 1567, "lr": 4.795583887675171e-05}
{"train_loss": 0.14997869729995728, "global_step": 139502, "epoch": 1567, "lr": 4.7955259624985265e-05}
{"train_loss": 0.08902255445718765, "global_step": 139503, "epoch": 1567, "lr": 4.7954680373493685e-05}
{"train_loss": 0.13604047894477844, "global_step": 139504, "epoch": 1567, "lr": 4.79541011222771e-05}
{"train_loss": 0.09847955405712128, "global_step": 139505, "epoch": 1567, "lr": 4.795352187133554e-05}
{"train_loss": 0.1048663780093193, "global_step": 139506, "epoch": 1567, "lr": 4.795294262066912e-05}
{"train_loss": 0.14270517230033875, "global_step": 139507, "epoch": 1567, "lr": 4.795236337027789e-05}
{"train_loss": 0.09898895025253296, "global_step": 139508, "epoch": 1567, "lr": 4.7951784120161944e-05}
{"train_loss": 0.10354408621788025, "global_step": 139509, "epoch": 1567, "lr": 4.7951204870321346e-05}
{"train_loss": 0.10400786995887756, "global_step": 139510, "epoch": 1567, "lr": 4.795062562075621e-05}
{"train_loss": 0.0953686535358429, "global_step": 139511, "epoch": 1567, "lr": 4.7950046371466556e-05}
{"train_loss": 0.15424764156341553, "global_step": 139512, "epoch": 1567, "lr": 4.7949467122452506e-05}
{"train_loss": 0.10458695888519287, "global_step": 139513, "epoch": 1567, "lr": 4.7948887873714135e-05}
{"train_loss": 0.10155637562274933, "global_step": 139514, "epoch": 1567, "lr": 4.79483086252515e-05}
{"train_loss": 0.0830627903342247, "global_step": 139515, "epoch": 1567, "lr": 4.79477293770647e-05}
{"train_loss": 0.15264801681041718, "global_step": 139516, "epoch": 1567, "lr": 4.7947150129153785e-05}
{"train_loss": 0.11853344738483429, "global_step": 139517, "epoch": 1567, "lr": 4.794657088151887e-05}
{"train_loss": 0.09402140229940414, "global_step": 139518, "epoch": 1567, "lr": 4.7945991634159995e-05}
{"train_loss": 0.10641604661941528, "global_step": 139519, "epoch": 1567, "lr": 4.7945412387077275e-05}
{"train_loss": 0.09451714158058167, "global_step": 139520, "epoch": 1567, "lr": 4.794483314027075e-05}
{"train_loss": 0.15436124801635742, "global_step": 139521, "epoch": 1567, "lr": 4.794425389374054e-05}
{"train_loss": 0.09945499897003174, "global_step": 139522, "epoch": 1567, "lr": 4.794367464748668e-05}
{"train_loss": 0.08738468587398529, "global_step": 139523, "epoch": 1567, "lr": 4.794309540150928e-05}
{"train_loss": 0.1567930281162262, "global_step": 139524, "epoch": 1567, "lr": 4.794251615580839e-05}
{"train_loss": 0.10784310102462769, "global_step": 139525, "epoch": 1567, "lr": 4.794193691038413e-05}
{"train_loss": 0.1119002178311348, "global_step": 139526, "epoch": 1567, "lr": 4.794135766523652e-05}
{"train_loss": 0.14147616922855377, "global_step": 139527, "epoch": 1567, "lr": 4.79407784203657e-05}
{"train_loss": 0.18775607645511627, "global_step": 139528, "epoch": 1567, "lr": 4.7940199175771686e-05}
{"train_loss": 0.06675397604703903, "global_step": 139529, "epoch": 1567, "lr": 4.7939619931454617e-05}
{"train_loss": 0.08847905695438385, "global_step": 139530, "epoch": 1567, "lr": 4.7939040687414514e-05}
{"train_loss": 0.10672735422849655, "global_step": 139531, "epoch": 1567, "lr": 4.79384614436515e-05}
{"train_loss": 0.10690515488386154, "global_step": 139532, "epoch": 1567, "lr": 4.793788220016563e-05}
{"train_loss": 0.1551193743944168, "global_step": 139533, "epoch": 1567, "lr": 4.7937302956956985e-05}
{"train_loss": 0.1599408984184265, "global_step": 139534, "epoch": 1567, "lr": 4.793672371402566e-05}
{"train_loss": 0.11063487082719803, "global_step": 139535, "epoch": 1567, "lr": 4.793614447137169e-05}
{"train_loss": 0.13574661314487457, "global_step": 139536, "epoch": 1567, "lr": 4.793556522899521e-05}
{"train_loss": 0.1277206987142563, "global_step": 139537, "epoch": 1567, "lr": 4.793498598689624e-05}
{"train_loss": 0.10126762092113495, "global_step": 139538, "epoch": 1567, "lr": 4.79344067450749e-05}
{"train_loss": 0.12004831433296204, "global_step": 139539, "epoch": 1567, "lr": 4.793382750353125e-05}
{"train_loss": 0.11377023160457611, "global_step": 139540, "epoch": 1567, "lr": 4.7933248262265383e-05}
{"train_loss": 0.09539595246315002, "global_step": 139541, "epoch": 1567, "lr": 4.7932669021277344e-05}
{"train_loss": 0.1450737714767456, "global_step": 139542, "epoch": 1567, "lr": 4.793208978056726e-05}
{"train_loss": 0.09559661149978638, "global_step": 139543, "epoch": 1567, "lr": 4.793151054013515e-05}
{"train_loss": 0.16943132877349854, "global_step": 139544, "epoch": 1567, "lr": 4.7930931299981154e-05}
{"train_loss": 0.10944470018148422, "global_step": 139545, "epoch": 1567, "lr": 4.7930352060105295e-05}
{"train_loss": 0.07907045632600784, "global_step": 139546, "epoch": 1567, "lr": 4.792977282050769e-05}
{"train_loss": 0.0633118748664856, "global_step": 139547, "epoch": 1567, "lr": 4.7929193581188387e-05}
{"train_loss": 0.08226234465837479, "global_step": 139548, "epoch": 1567, "lr": 4.79286143421475e-05}
{"train_loss": 0.14327898621559143, "global_step": 139549, "epoch": 1567, "lr": 4.792803510338506e-05}
{"train_loss": 0.11892558634281158, "global_step": 139550, "epoch": 1567, "lr": 4.7927455864901184e-05}
{"train_loss": 0.11473435297440947, "global_step": 139551, "epoch": 1567, "lr": 4.792687662669594e-05, "val_loss": 5.972692012786865}
{"train_loss": 0.16959650814533234, "global_step": 139552, "epoch": 1568, "lr": 4.7926297388769394e-05}
{"train_loss": 0.07319369912147522, "global_step": 139553, "epoch": 1568, "lr": 4.7925718151121645e-05}
{"train_loss": 0.11721061170101166, "global_step": 139554, "epoch": 1568, "lr": 4.792513891375274e-05}
{"train_loss": 0.12123105674982071, "global_step": 139555, "epoch": 1568, "lr": 4.792455967666279e-05}
{"train_loss": 0.07130029052495956, "global_step": 139556, "epoch": 1568, "lr": 4.7923980439851845e-05}
{"train_loss": 0.09666343778371811, "global_step": 139557, "epoch": 1568, "lr": 4.792340120332001e-05}
{"train_loss": 0.14033803343772888, "global_step": 139558, "epoch": 1568, "lr": 4.7922821967067333e-05}
{"train_loss": 0.11441498249769211, "global_step": 139559, "epoch": 1568, "lr": 4.7922242731093925e-05}
{"train_loss": 0.13453391194343567, "global_step": 139560, "epoch": 1568, "lr": 4.792166349539983e-05}
{"train_loss": 0.16901662945747375, "global_step": 139561, "epoch": 1568, "lr": 4.7921084259985156e-05}
{"train_loss": 0.17377500236034393, "global_step": 139562, "epoch": 1568, "lr": 4.792050502484995e-05}
{"train_loss": 0.08297182619571686, "global_step": 139563, "epoch": 1568, "lr": 4.791992578999432e-05}
{"train_loss": 0.12353651970624924, "global_step": 139564, "epoch": 1568, "lr": 4.7919346555418314e-05}
{"train_loss": 0.05734923109412193, "global_step": 139565, "epoch": 1568, "lr": 4.791876732112205e-05}
{"train_loss": 0.0942346379160881, "global_step": 139566, "epoch": 1568, "lr": 4.791818808710557e-05}
{"train_loss": 0.1313650757074356, "global_step": 139567, "epoch": 1568, "lr": 4.7917608853368956e-05}
{"train_loss": 0.09317067265510559, "global_step": 139568, "epoch": 1568, "lr": 4.7917029619912306e-05}
{"train_loss": 0.13349740207195282, "global_step": 139569, "epoch": 1568, "lr": 4.7916450386735685e-05}
{"train_loss": 0.11506387591362, "global_step": 139570, "epoch": 1568, "lr": 4.791587115383917e-05}
{"train_loss": 0.11923974752426147, "global_step": 139571, "epoch": 1568, "lr": 4.7915291921222836e-05}
{"train_loss": 0.09692145138978958, "global_step": 139572, "epoch": 1568, "lr": 4.791471268888679e-05}
{"train_loss": 0.12444839626550674, "global_step": 139573, "epoch": 1568, "lr": 4.791413345683105e-05}
{"train_loss": 0.08638134598731995, "global_step": 139574, "epoch": 1568, "lr": 4.7913554225055755e-05}
{"train_loss": 0.12494668364524841, "global_step": 139575, "epoch": 1568, "lr": 4.7912974993560935e-05}
{"train_loss": 0.13805226981639862, "global_step": 139576, "epoch": 1568, "lr": 4.791239576234672e-05}
{"train_loss": 0.10400110483169556, "global_step": 139577, "epoch": 1568, "lr": 4.7911816531413124e-05}
{"train_loss": 0.12404067069292068, "global_step": 139578, "epoch": 1568, "lr": 4.791123730076029e-05}
{"train_loss": 0.11266433447599411, "global_step": 139579, "epoch": 1568, "lr": 4.7910658070388235e-05}
{"train_loss": 0.09308023750782013, "global_step": 139580, "epoch": 1568, "lr": 4.79100788402971e-05}
{"train_loss": 0.10894052684307098, "global_step": 139581, "epoch": 1568, "lr": 4.79094996104869e-05}
{"train_loss": 0.15705427527427673, "global_step": 139582, "epoch": 1568, "lr": 4.7908920380957757e-05}
{"train_loss": 0.10380381345748901, "global_step": 139583, "epoch": 1568, "lr": 4.790834115170972e-05}
{"train_loss": 0.09180198609828949, "global_step": 139584, "epoch": 1568, "lr": 4.79077619227429e-05}
{"train_loss": 0.10074998438358307, "global_step": 139585, "epoch": 1568, "lr": 4.790718269405734e-05}
{"train_loss": 0.0835951566696167, "global_step": 139586, "epoch": 1568, "lr": 4.7906603465653145e-05}
{"train_loss": 0.0817703902721405, "global_step": 139587, "epoch": 1568, "lr": 4.790602423753037e-05}
{"train_loss": 0.15934526920318604, "global_step": 139588, "epoch": 1568, "lr": 4.7905445009689115e-05}
{"train_loss": 0.15447793900966644, "global_step": 139589, "epoch": 1568, "lr": 4.790486578212944e-05}
{"train_loss": 0.135236918926239, "global_step": 139590, "epoch": 1568, "lr": 4.790428655485144e-05}
{"train_loss": 0.07112745195627213, "global_step": 139591, "epoch": 1568, "lr": 4.790370732785519e-05}
{"train_loss": 0.10520551353693008, "global_step": 139592, "epoch": 1568, "lr": 4.790312810114074e-05}
{"train_loss": 0.11823916435241699, "global_step": 139593, "epoch": 1568, "lr": 4.7902548874708206e-05}
{"train_loss": 0.08319807052612305, "global_step": 139594, "epoch": 1568, "lr": 4.790196964855763e-05}
{"train_loss": 0.10532437264919281, "global_step": 139595, "epoch": 1568, "lr": 4.790139042268913e-05}
{"train_loss": 0.0970255434513092, "global_step": 139596, "epoch": 1568, "lr": 4.790081119710274e-05}
{"train_loss": 0.1435634195804596, "global_step": 139597, "epoch": 1568, "lr": 4.790023197179859e-05}
{"train_loss": 0.10096031427383423, "global_step": 139598, "epoch": 1568, "lr": 4.78996527467767e-05}
{"train_loss": 0.1211550310254097, "global_step": 139599, "epoch": 1568, "lr": 4.789907352203719e-05}
{"train_loss": 0.10469402372837067, "global_step": 139600, "epoch": 1568, "lr": 4.789849429758011e-05}
{"train_loss": 0.14647790789604187, "global_step": 139601, "epoch": 1568, "lr": 4.789791507340558e-05}
{"train_loss": 0.07264329493045807, "global_step": 139602, "epoch": 1568, "lr": 4.789733584951361e-05}
{"train_loss": 0.14942209422588348, "global_step": 139603, "epoch": 1568, "lr": 4.7896756625904335e-05}
{"train_loss": 0.07721391320228577, "global_step": 139604, "epoch": 1568, "lr": 4.789617740257783e-05}
{"train_loss": 0.11525817215442657, "global_step": 139605, "epoch": 1568, "lr": 4.7895598179534146e-05}
{"train_loss": 0.0944666862487793, "global_step": 139606, "epoch": 1568, "lr": 4.789501895677338e-05}
{"train_loss": 0.14507299661636353, "global_step": 139607, "epoch": 1568, "lr": 4.789443973429559e-05}
{"train_loss": 0.09456778317689896, "global_step": 139608, "epoch": 1568, "lr": 4.789386051210088e-05}
{"train_loss": 0.10163634270429611, "global_step": 139609, "epoch": 1568, "lr": 4.789328129018931e-05}
{"train_loss": 0.09256856888532639, "global_step": 139610, "epoch": 1568, "lr": 4.7892702068560976e-05}
{"train_loss": 0.06854037195444107, "global_step": 139611, "epoch": 1568, "lr": 4.789212284721592e-05}
{"train_loss": 0.11414285749197006, "global_step": 139612, "epoch": 1568, "lr": 4.7891543626154264e-05}
{"train_loss": 0.0928037166595459, "global_step": 139613, "epoch": 1568, "lr": 4.789096440537604e-05}
{"train_loss": 0.18457648158073425, "global_step": 139614, "epoch": 1568, "lr": 4.789038518488137e-05}
{"train_loss": 0.08954915404319763, "global_step": 139615, "epoch": 1568, "lr": 4.788980596467029e-05}
{"train_loss": 0.12145229429006577, "global_step": 139616, "epoch": 1568, "lr": 4.7889226744742924e-05}
{"train_loss": 0.18057338893413544, "global_step": 139617, "epoch": 1568, "lr": 4.7888647525099315e-05}
{"train_loss": 0.12024085223674774, "global_step": 139618, "epoch": 1568, "lr": 4.788806830573956e-05}
{"train_loss": 0.20022830367088318, "global_step": 139619, "epoch": 1568, "lr": 4.7887489086663703e-05}
{"train_loss": 0.15007397532463074, "global_step": 139620, "epoch": 1568, "lr": 4.7886909867871884e-05}
{"train_loss": 0.13836725056171417, "global_step": 139621, "epoch": 1568, "lr": 4.7886330649364105e-05}
{"train_loss": 0.10609883069992065, "global_step": 139622, "epoch": 1568, "lr": 4.7885751431140496e-05}
{"train_loss": 0.09079824388027191, "global_step": 139623, "epoch": 1568, "lr": 4.7885172213201144e-05}
{"train_loss": 0.11102661490440369, "global_step": 139624, "epoch": 1568, "lr": 4.788459299554609e-05}
{"train_loss": 0.10922235250473022, "global_step": 139625, "epoch": 1568, "lr": 4.788401377817542e-05}
{"train_loss": 0.1020982414484024, "global_step": 139626, "epoch": 1568, "lr": 4.788343456108923e-05}
{"train_loss": 0.07092264294624329, "global_step": 139627, "epoch": 1568, "lr": 4.7882855344287586e-05}
{"train_loss": 0.17200753092765808, "global_step": 139628, "epoch": 1568, "lr": 4.788227612777056e-05}
{"train_loss": 0.07842482626438141, "global_step": 139629, "epoch": 1568, "lr": 4.788169691153825e-05}
{"train_loss": 0.11583205312490463, "global_step": 139630, "epoch": 1568, "lr": 4.788111769559069e-05}
{"train_loss": 0.16500183939933777, "global_step": 139631, "epoch": 1568, "lr": 4.788053847992802e-05}
{"train_loss": 0.07567835599184036, "global_step": 139632, "epoch": 1568, "lr": 4.787995926455027e-05}
{"train_loss": 0.08876847475767136, "global_step": 139633, "epoch": 1568, "lr": 4.787938004945754e-05}
{"train_loss": 0.11868118494749069, "global_step": 139634, "epoch": 1568, "lr": 4.787880083464989e-05}
{"train_loss": 0.17697575688362122, "global_step": 139635, "epoch": 1568, "lr": 4.7878221620127426e-05}
{"train_loss": 0.2549191415309906, "global_step": 139636, "epoch": 1568, "lr": 4.787764240589019e-05}
{"train_loss": 0.19446051120758057, "global_step": 139637, "epoch": 1568, "lr": 4.78770631919383e-05}
{"train_loss": 0.14033159613609314, "global_step": 139638, "epoch": 1568, "lr": 4.787648397827179e-05}
{"train_loss": 0.08526427298784256, "global_step": 139639, "epoch": 1568, "lr": 4.787590476489076e-05}
{"train_loss": 0.11795896639147502, "global_step": 139640, "epoch": 1568, "lr": 4.787532555179531e-05, "val_loss": 5.950619697570801}
{"train_loss": 0.11487534642219543, "global_step": 139641, "epoch": 1569, "lr": 4.787474633898548e-05}
{"train_loss": 0.11543085426092148, "global_step": 139642, "epoch": 1569, "lr": 4.787416712646138e-05}
{"train_loss": 0.08437425643205643, "global_step": 139643, "epoch": 1569, "lr": 4.787358791422307e-05}
{"train_loss": 0.09859878569841385, "global_step": 139644, "epoch": 1569, "lr": 4.787300870227062e-05}
{"train_loss": 0.13955368101596832, "global_step": 139645, "epoch": 1569, "lr": 4.7872429490604124e-05}
{"train_loss": 0.09147128462791443, "global_step": 139646, "epoch": 1569, "lr": 4.7871850279223654e-05}
{"train_loss": 0.10775459557771683, "global_step": 139647, "epoch": 1569, "lr": 4.787127106812928e-05}
{"train_loss": 0.1334221512079239, "global_step": 139648, "epoch": 1569, "lr": 4.78706918573211e-05}
{"train_loss": 0.08846642822027206, "global_step": 139649, "epoch": 1569, "lr": 4.787011264679918e-05}
{"train_loss": 0.10345856845378876, "global_step": 139650, "epoch": 1569, "lr": 4.78695334365636e-05}
{"train_loss": 0.06428288668394089, "global_step": 139651, "epoch": 1569, "lr": 4.7868954226614415e-05}
{"train_loss": 0.13494908809661865, "global_step": 139652, "epoch": 1569, "lr": 4.786837501695175e-05}
{"train_loss": 0.057613857090473175, "global_step": 139653, "epoch": 1569, "lr": 4.786779580757563e-05}
{"train_loss": 0.12569227814674377, "global_step": 139654, "epoch": 1569, "lr": 4.786721659848618e-05}
{"train_loss": 0.09713484346866608, "global_step": 139655, "epoch": 1569, "lr": 4.786663738968343e-05}
{"train_loss": 0.1596253216266632, "global_step": 139656, "epoch": 1569, "lr": 4.786605818116751e-05}
{"train_loss": 0.10640982538461685, "global_step": 139657, "epoch": 1569, "lr": 4.7865478972938456e-05}
{"train_loss": 0.08091508597135544, "global_step": 139658, "epoch": 1569, "lr": 4.7864899764996357e-05}
{"train_loss": 0.08960508555173874, "global_step": 139659, "epoch": 1569, "lr": 4.7864320557341326e-05}
{"train_loss": 0.1258038878440857, "global_step": 139660, "epoch": 1569, "lr": 4.786374134997338e-05}
{"train_loss": 0.10673674196004868, "global_step": 139661, "epoch": 1569, "lr": 4.786316214289265e-05}
{"train_loss": 0.09212954342365265, "global_step": 139662, "epoch": 1569, "lr": 4.7862582936099175e-05}
{"train_loss": 0.14875003695487976, "global_step": 139663, "epoch": 1569, "lr": 4.786200372959307e-05}
{"train_loss": 0.13789746165275574, "global_step": 139664, "epoch": 1569, "lr": 4.786142452337438e-05}
{"train_loss": 0.13272561132907867, "global_step": 139665, "epoch": 1569, "lr": 4.78608453174432e-05}
{"train_loss": 0.08189269155263901, "global_step": 139666, "epoch": 1569, "lr": 4.7860266111799594e-05}
{"train_loss": 0.07125823944807053, "global_step": 139667, "epoch": 1569, "lr": 4.7859686906443655e-05}
{"train_loss": 0.1095387190580368, "global_step": 139668, "epoch": 1569, "lr": 4.785910770137546e-05}
{"train_loss": 0.0706910714507103, "global_step": 139669, "epoch": 1569, "lr": 4.785852849659509e-05}
{"train_loss": 0.11400458961725235, "global_step": 139670, "epoch": 1569, "lr": 4.785794929210258e-05}
{"train_loss": 0.11130170524120331, "global_step": 139671, "epoch": 1569, "lr": 4.785737008789808e-05}
{"train_loss": 0.15494205057621002, "global_step": 139672, "epoch": 1569, "lr": 4.785679088398161e-05}
{"train_loss": 0.13415522873401642, "global_step": 139673, "epoch": 1569, "lr": 4.785621168035328e-05}
{"train_loss": 0.12071043252944946, "global_step": 139674, "epoch": 1569, "lr": 4.7855632477013134e-05}
{"train_loss": 0.09818273037672043, "global_step": 139675, "epoch": 1569, "lr": 4.78550532739613e-05}
{"train_loss": 0.12455347925424576, "global_step": 139676, "epoch": 1569, "lr": 4.7854474071197805e-05}
{"train_loss": 0.10560799390077591, "global_step": 139677, "epoch": 1569, "lr": 4.785389486872275e-05}
{"train_loss": 0.1227291077375412, "global_step": 139678, "epoch": 1569, "lr": 4.785331566653623e-05}
{"train_loss": 0.18053729832172394, "global_step": 139679, "epoch": 1569, "lr": 4.7852736464638285e-05}
{"train_loss": 0.06167198345065117, "global_step": 139680, "epoch": 1569, "lr": 4.785215726302904e-05}
{"train_loss": 0.06687833368778229, "global_step": 139681, "epoch": 1569, "lr": 4.7851578061708526e-05}
{"train_loss": 0.12027864903211594, "global_step": 139682, "epoch": 1569, "lr": 4.785099886067685e-05}
{"train_loss": 0.15613047778606415, "global_step": 139683, "epoch": 1569, "lr": 4.7850419659934086e-05}
{"train_loss": 0.10079300403594971, "global_step": 139684, "epoch": 1569, "lr": 4.78498404594803e-05}
{"train_loss": 0.13328313827514648, "global_step": 139685, "epoch": 1569, "lr": 4.7849261259315575e-05}
{"train_loss": 0.09038051962852478, "global_step": 139686, "epoch": 1569, "lr": 4.7848682059439994e-05}
{"train_loss": 0.0828782320022583, "global_step": 139687, "epoch": 1569, "lr": 4.784810285985363e-05}
{"train_loss": 0.11308526992797852, "global_step": 139688, "epoch": 1569, "lr": 4.784752366055657e-05}
{"train_loss": 0.08859273791313171, "global_step": 139689, "epoch": 1569, "lr": 4.784694446154887e-05}
{"train_loss": 0.09522904455661774, "global_step": 139690, "epoch": 1569, "lr": 4.784636526283063e-05}
{"train_loss": 0.08838839828968048, "global_step": 139691, "epoch": 1569, "lr": 4.784578606440192e-05}
{"train_loss": 0.1184665784239769, "global_step": 139692, "epoch": 1569, "lr": 4.784520686626282e-05}
{"train_loss": 0.14635801315307617, "global_step": 139693, "epoch": 1569, "lr": 4.784462766841339e-05}
{"train_loss": 0.08092252910137177, "global_step": 139694, "epoch": 1569, "lr": 4.7844048470853726e-05}
{"train_loss": 0.16311274468898773, "global_step": 139695, "epoch": 1569, "lr": 4.784346927358392e-05}
{"train_loss": 0.15564271807670593, "global_step": 139696, "epoch": 1569, "lr": 4.784289007660402e-05}
{"train_loss": 0.09899678081274033, "global_step": 139697, "epoch": 1569, "lr": 4.784231087991413e-05}
{"train_loss": 0.10136549919843674, "global_step": 139698, "epoch": 1569, "lr": 4.78417316835143e-05}
{"train_loss": 0.15708737075328827, "global_step": 139699, "epoch": 1569, "lr": 4.784115248740464e-05}
{"train_loss": 0.10298983752727509, "global_step": 139700, "epoch": 1569, "lr": 4.784057329158519e-05}
{"train_loss": 0.1505919098854065, "global_step": 139701, "epoch": 1569, "lr": 4.783999409605607e-05}
{"train_loss": 0.1271866112947464, "global_step": 139702, "epoch": 1569, "lr": 4.783941490081733e-05}
{"train_loss": 0.10164422541856766, "global_step": 139703, "epoch": 1569, "lr": 4.783883570586905e-05}
{"train_loss": 0.08153653144836426, "global_step": 139704, "epoch": 1569, "lr": 4.783825651121132e-05}
{"train_loss": 0.09161150455474854, "global_step": 139705, "epoch": 1569, "lr": 4.7837677316844195e-05}
{"train_loss": 0.11638463288545609, "global_step": 139706, "epoch": 1569, "lr": 4.7837098122767774e-05}
{"train_loss": 0.06503699719905853, "global_step": 139707, "epoch": 1569, "lr": 4.783651892898215e-05}
{"train_loss": 0.1453404277563095, "global_step": 139708, "epoch": 1569, "lr": 4.783593973548735e-05}
{"train_loss": 0.13016840815544128, "global_step": 139709, "epoch": 1569, "lr": 4.78353605422835e-05}
{"train_loss": 0.14011643826961517, "global_step": 139710, "epoch": 1569, "lr": 4.783478134937064e-05}
{"train_loss": 0.2103273719549179, "global_step": 139711, "epoch": 1569, "lr": 4.783420215674889e-05}
{"train_loss": 0.09345164149999619, "global_step": 139712, "epoch": 1569, "lr": 4.7833622964418286e-05}
{"train_loss": 0.09906011819839478, "global_step": 139713, "epoch": 1569, "lr": 4.7833043772378924e-05}
{"train_loss": 0.11300692707300186, "global_step": 139714, "epoch": 1569, "lr": 4.7832464580630895e-05}
{"train_loss": 0.10181068629026413, "global_step": 139715, "epoch": 1569, "lr": 4.783188538917425e-05}
{"train_loss": 0.07581467181444168, "global_step": 139716, "epoch": 1569, "lr": 4.783130619800911e-05}
{"train_loss": 0.05935578793287277, "global_step": 139717, "epoch": 1569, "lr": 4.78307270071355e-05}
{"train_loss": 0.1598624736070633, "global_step": 139718, "epoch": 1569, "lr": 4.7830147816553534e-05}
{"train_loss": 0.16604287922382355, "global_step": 139719, "epoch": 1569, "lr": 4.782956862626327e-05}
{"train_loss": 0.09186337143182755, "global_step": 139720, "epoch": 1569, "lr": 4.78289894362648e-05}
{"train_loss": 0.0929739698767662, "global_step": 139721, "epoch": 1569, "lr": 4.78284102465582e-05}
{"train_loss": 0.13421325385570526, "global_step": 139722, "epoch": 1569, "lr": 4.7827831057143534e-05}
{"train_loss": 0.06677944958209991, "global_step": 139723, "epoch": 1569, "lr": 4.7827251868020894e-05}
{"train_loss": 0.1263800859451294, "global_step": 139724, "epoch": 1569, "lr": 4.782667267919036e-05}
{"train_loss": 0.1559087485074997, "global_step": 139725, "epoch": 1569, "lr": 4.7826093490651994e-05}
{"train_loss": 0.14070208370685577, "global_step": 139726, "epoch": 1569, "lr": 4.78255143024059e-05}
{"train_loss": 0.08385109156370163, "global_step": 139727, "epoch": 1569, "lr": 4.782493511445211e-05}
{"train_loss": 0.09349173307418823, "global_step": 139728, "epoch": 1569, "lr": 4.7824355926790755e-05}
{"train_loss": 0.114228590276469, "global_step": 139729, "epoch": 1569, "lr": 4.7823776739421865e-05, "val_loss": 6.144947528839111}
{"train_loss": 0.07498858869075775, "global_step": 139730, "epoch": 1570, "lr": 4.782319755234555e-05}
{"train_loss": 0.055389951914548874, "global_step": 139731, "epoch": 1570, "lr": 4.782261836556189e-05}
{"train_loss": 0.0898933932185173, "global_step": 139732, "epoch": 1570, "lr": 4.782203917907094e-05}
{"train_loss": 0.1327124983072281, "global_step": 139733, "epoch": 1570, "lr": 4.782145999287281e-05}
{"train_loss": 0.07101801037788391, "global_step": 139734, "epoch": 1570, "lr": 4.782088080696754e-05}
{"train_loss": 0.0942852646112442, "global_step": 139735, "epoch": 1570, "lr": 4.782030162135524e-05}
{"train_loss": 0.08774562180042267, "global_step": 139736, "epoch": 1570, "lr": 4.781972243603596e-05}
{"train_loss": 0.08208946883678436, "global_step": 139737, "epoch": 1570, "lr": 4.7819143251009807e-05}
{"train_loss": 0.09886812418699265, "global_step": 139738, "epoch": 1570, "lr": 4.781856406627682e-05}
{"train_loss": 0.08675068616867065, "global_step": 139739, "epoch": 1570, "lr": 4.781798488183713e-05}
{"train_loss": 0.15124890208244324, "global_step": 139740, "epoch": 1570, "lr": 4.7817405697690774e-05}
{"train_loss": 0.16214421391487122, "global_step": 139741, "epoch": 1570, "lr": 4.781682651383783e-05}
{"train_loss": 0.11169720441102982, "global_step": 139742, "epoch": 1570, "lr": 4.7816247330278394e-05}
{"train_loss": 0.1055373027920723, "global_step": 139743, "epoch": 1570, "lr": 4.7815668147012545e-05}
{"train_loss": 0.1729585975408554, "global_step": 139744, "epoch": 1570, "lr": 4.781508896404035e-05}
{"train_loss": 0.11441536992788315, "global_step": 139745, "epoch": 1570, "lr": 4.7814509781361886e-05}
{"train_loss": 0.1673038750886917, "global_step": 139746, "epoch": 1570, "lr": 4.781393059897722e-05}
{"train_loss": 0.08548878133296967, "global_step": 139747, "epoch": 1570, "lr": 4.7813351416886474e-05}
{"train_loss": 0.09996700286865234, "global_step": 139748, "epoch": 1570, "lr": 4.781277223508967e-05}
{"train_loss": 0.06962891668081284, "global_step": 139749, "epoch": 1570, "lr": 4.781219305358691e-05}
{"train_loss": 0.11209956556558609, "global_step": 139750, "epoch": 1570, "lr": 4.78116138723783e-05}
{"train_loss": 0.1579740047454834, "global_step": 139751, "epoch": 1570, "lr": 4.781103469146387e-05}
{"train_loss": 0.18292781710624695, "global_step": 139752, "epoch": 1570, "lr": 4.781045551084374e-05}
{"train_loss": 0.05170605331659317, "global_step": 139753, "epoch": 1570, "lr": 4.780987633051794e-05}
{"train_loss": 0.15335635840892792, "global_step": 139754, "epoch": 1570, "lr": 4.7809297150486606e-05}
{"train_loss": 0.11026697605848312, "global_step": 139755, "epoch": 1570, "lr": 4.7808717970749755e-05}
{"train_loss": 0.2409464567899704, "global_step": 139756, "epoch": 1570, "lr": 4.780813879130752e-05}
{"train_loss": 0.14903989434242249, "global_step": 139757, "epoch": 1570, "lr": 4.7807559612159934e-05}
{"train_loss": 0.04591897130012512, "global_step": 139758, "epoch": 1570, "lr": 4.780698043330712e-05}
{"train_loss": 0.10382245481014252, "global_step": 139759, "epoch": 1570, "lr": 4.7806401254749116e-05}
{"train_loss": 0.07947501540184021, "global_step": 139760, "epoch": 1570, "lr": 4.7805822076486014e-05}
{"train_loss": 0.06592579185962677, "global_step": 139761, "epoch": 1570, "lr": 4.7805242898517896e-05}
{"train_loss": 0.06701233983039856, "global_step": 139762, "epoch": 1570, "lr": 4.780466372084483e-05}
{"train_loss": 0.13409624993801117, "global_step": 139763, "epoch": 1570, "lr": 4.7804084543466905e-05}
{"train_loss": 0.09395643323659897, "global_step": 139764, "epoch": 1570, "lr": 4.7803505366384194e-05}
{"train_loss": 0.10809376835823059, "global_step": 139765, "epoch": 1570, "lr": 4.780292618959677e-05}
{"train_loss": 0.18270204961299896, "global_step": 139766, "epoch": 1570, "lr": 4.780234701310471e-05}
{"train_loss": 0.17041388154029846, "global_step": 139767, "epoch": 1570, "lr": 4.7801767836908115e-05}
{"train_loss": 0.07135865092277527, "global_step": 139768, "epoch": 1570, "lr": 4.780118866100703e-05}
{"train_loss": 0.06777957081794739, "global_step": 139769, "epoch": 1570, "lr": 4.780060948540157e-05}
{"train_loss": 0.09160404652357101, "global_step": 139770, "epoch": 1570, "lr": 4.780003031009177e-05}
{"train_loss": 0.17323270440101624, "global_step": 139771, "epoch": 1570, "lr": 4.779945113507774e-05}
{"train_loss": 0.0836993008852005, "global_step": 139772, "epoch": 1570, "lr": 4.779887196035954e-05}
{"train_loss": 0.1394466906785965, "global_step": 139773, "epoch": 1570, "lr": 4.7798292785937265e-05}
{"train_loss": 0.11747994273900986, "global_step": 139774, "epoch": 1570, "lr": 4.779771361181096e-05}
{"train_loss": 0.14253424108028412, "global_step": 139775, "epoch": 1570, "lr": 4.779713443798076e-05}
{"train_loss": 0.12774409353733063, "global_step": 139776, "epoch": 1570, "lr": 4.7796555264446674e-05}
{"train_loss": 0.10520248860120773, "global_step": 139777, "epoch": 1570, "lr": 4.779597609120884e-05}
{"train_loss": 0.08854019641876221, "global_step": 139778, "epoch": 1570, "lr": 4.7795396918267296e-05}
{"train_loss": 0.07777304202318192, "global_step": 139779, "epoch": 1570, "lr": 4.779481774562214e-05}
{"train_loss": 0.15231400728225708, "global_step": 139780, "epoch": 1570, "lr": 4.7794238573273445e-05}
{"train_loss": 0.09981005638837814, "global_step": 139781, "epoch": 1570, "lr": 4.779365940122129e-05}
{"train_loss": 0.10133161395788193, "global_step": 139782, "epoch": 1570, "lr": 4.7793080229465745e-05}
{"train_loss": 0.08752712607383728, "global_step": 139783, "epoch": 1570, "lr": 4.779250105800689e-05}
{"train_loss": 0.10128331929445267, "global_step": 139784, "epoch": 1570, "lr": 4.7791921886844813e-05}
{"train_loss": 0.11249420791864395, "global_step": 139785, "epoch": 1570, "lr": 4.7791342715979575e-05}
{"train_loss": 0.10828865319490433, "global_step": 139786, "epoch": 1570, "lr": 4.779076354541128e-05}
{"train_loss": 0.1291719377040863, "global_step": 139787, "epoch": 1570, "lr": 4.7790184375139974e-05}
{"train_loss": 0.17846161127090454, "global_step": 139788, "epoch": 1570, "lr": 4.778960520516577e-05}
{"train_loss": 0.12765201926231384, "global_step": 139789, "epoch": 1570, "lr": 4.7789026035488706e-05}
{"train_loss": 0.1466602385044098, "global_step": 139790, "epoch": 1570, "lr": 4.77884468661089e-05}
{"train_loss": 0.12464290112257004, "global_step": 139791, "epoch": 1570, "lr": 4.778786769702639e-05}
{"train_loss": 0.12549066543579102, "global_step": 139792, "epoch": 1570, "lr": 4.778728852824129e-05}
{"train_loss": 0.10141933709383011, "global_step": 139793, "epoch": 1570, "lr": 4.778670935975365e-05}
{"train_loss": 0.09900906682014465, "global_step": 139794, "epoch": 1570, "lr": 4.778613019156357e-05}
{"train_loss": 0.11744532734155655, "global_step": 139795, "epoch": 1570, "lr": 4.7785551023671106e-05}
{"train_loss": 0.07356973737478256, "global_step": 139796, "epoch": 1570, "lr": 4.7784971856076364e-05}
{"train_loss": 0.08617039769887924, "global_step": 139797, "epoch": 1570, "lr": 4.778439268877939e-05}
{"train_loss": 0.08444619923830032, "global_step": 139798, "epoch": 1570, "lr": 4.7783813521780295e-05}
{"train_loss": 0.09283646196126938, "global_step": 139799, "epoch": 1570, "lr": 4.778323435507913e-05}
{"train_loss": 0.13225021958351135, "global_step": 139800, "epoch": 1570, "lr": 4.778265518867598e-05}
{"train_loss": 0.2269919067621231, "global_step": 139801, "epoch": 1570, "lr": 4.7782076022570926e-05}
{"train_loss": 0.10914798825979233, "global_step": 139802, "epoch": 1570, "lr": 4.778149685676405e-05}
{"train_loss": 0.06835593283176422, "global_step": 139803, "epoch": 1570, "lr": 4.778091769125542e-05}
{"train_loss": 0.07398686558008194, "global_step": 139804, "epoch": 1570, "lr": 4.778033852604511e-05}
{"train_loss": 0.1146668791770935, "global_step": 139805, "epoch": 1570, "lr": 4.777975936113322e-05}
{"train_loss": 0.12695622444152832, "global_step": 139806, "epoch": 1570, "lr": 4.77791801965198e-05}
{"train_loss": 0.1120627149939537, "global_step": 139807, "epoch": 1570, "lr": 4.777860103220496e-05}
{"train_loss": 0.0841628760099411, "global_step": 139808, "epoch": 1570, "lr": 4.777802186818874e-05}
{"train_loss": 0.09878745675086975, "global_step": 139809, "epoch": 1570, "lr": 4.777744270447126e-05}
{"train_loss": 0.10820813477039337, "global_step": 139810, "epoch": 1570, "lr": 4.777686354105255e-05}
{"train_loss": 0.19284692406654358, "global_step": 139811, "epoch": 1570, "lr": 4.777628437793274e-05}
{"train_loss": 0.18119916319847107, "global_step": 139812, "epoch": 1570, "lr": 4.777570521511186e-05}
{"train_loss": 0.1316928267478943, "global_step": 139813, "epoch": 1570, "lr": 4.777512605259003e-05}
{"train_loss": 0.06018127128481865, "global_step": 139814, "epoch": 1570, "lr": 4.777454689036728e-05}
{"train_loss": 0.08643162995576859, "global_step": 139815, "epoch": 1570, "lr": 4.7773967728443744e-05}
{"train_loss": 0.1160658523440361, "global_step": 139816, "epoch": 1570, "lr": 4.777338856681945e-05}
{"train_loss": 0.09031390398740768, "global_step": 139817, "epoch": 1570, "lr": 4.7772809405494515e-05}
{"train_loss": 0.11272941826936904, "global_step": 139818, "epoch": 1570, "lr": 4.777223024446899e-05, "val_loss": 6.013211727142334, "train_action_mse_error": 22.432636260986328}
{"train_loss": 0.12384117394685745, "global_step": 139819, "epoch": 1571, "lr": 4.7771651083742964e-05}
{"train_loss": 0.13953690230846405, "global_step": 139820, "epoch": 1571, "lr": 4.7771071923316506e-05}
{"train_loss": 0.19188207387924194, "global_step": 139821, "epoch": 1571, "lr": 4.777049276318969e-05}
{"train_loss": 0.08322753757238388, "global_step": 139822, "epoch": 1571, "lr": 4.776991360336263e-05}
{"train_loss": 0.1004309356212616, "global_step": 139823, "epoch": 1571, "lr": 4.7769334443835354e-05}
{"train_loss": 0.13060416281223297, "global_step": 139824, "epoch": 1571, "lr": 4.776875528460798e-05}
{"train_loss": 0.08513524383306503, "global_step": 139825, "epoch": 1571, "lr": 4.7768176125680555e-05}
{"train_loss": 0.10708677023649216, "global_step": 139826, "epoch": 1571, "lr": 4.776759696705319e-05}
{"train_loss": 0.17850349843502045, "global_step": 139827, "epoch": 1571, "lr": 4.776701780872593e-05}
{"train_loss": 0.10218362510204315, "global_step": 139828, "epoch": 1571, "lr": 4.7766438650698883e-05}
{"train_loss": 0.07333948463201523, "global_step": 139829, "epoch": 1571, "lr": 4.776585949297209e-05}
{"train_loss": 0.13305391371250153, "global_step": 139830, "epoch": 1571, "lr": 4.776528033554567e-05}
{"train_loss": 0.046434562653303146, "global_step": 139831, "epoch": 1571, "lr": 4.776470117841966e-05}
{"train_loss": 0.1433623880147934, "global_step": 139832, "epoch": 1571, "lr": 4.776412202159418e-05}
{"train_loss": 0.0859338715672493, "global_step": 139833, "epoch": 1571, "lr": 4.7763542865069266e-05}
{"train_loss": 0.1602504402399063, "global_step": 139834, "epoch": 1571, "lr": 4.776296370884504e-05}
{"train_loss": 0.14695745706558228, "global_step": 139835, "epoch": 1571, "lr": 4.776238455292153e-05}
{"train_loss": 0.1200699433684349, "global_step": 139836, "epoch": 1571, "lr": 4.776180539729887e-05}
{"train_loss": 0.13423004746437073, "global_step": 139837, "epoch": 1571, "lr": 4.776122624197708e-05}
{"train_loss": 0.09912682324647903, "global_step": 139838, "epoch": 1571, "lr": 4.776064708695628e-05}
{"train_loss": 0.06995332986116409, "global_step": 139839, "epoch": 1571, "lr": 4.776006793223653e-05}
{"train_loss": 0.16843530535697937, "global_step": 139840, "epoch": 1571, "lr": 4.7759488777817904e-05}
{"train_loss": 0.1279941201210022, "global_step": 139841, "epoch": 1571, "lr": 4.7758909623700506e-05}
{"train_loss": 0.1269780993461609, "global_step": 139842, "epoch": 1571, "lr": 4.7758330469884374e-05}
{"train_loss": 0.09175495058298111, "global_step": 139843, "epoch": 1571, "lr": 4.775775131636963e-05}
{"train_loss": 0.054469771683216095, "global_step": 139844, "epoch": 1571, "lr": 4.775717216315631e-05}
{"train_loss": 0.15257611870765686, "global_step": 139845, "epoch": 1571, "lr": 4.775659301024452e-05}
{"train_loss": 0.13585606217384338, "global_step": 139846, "epoch": 1571, "lr": 4.775601385763432e-05}
{"train_loss": 0.09314926713705063, "global_step": 139847, "epoch": 1571, "lr": 4.775543470532581e-05}
{"train_loss": 0.1658836454153061, "global_step": 139848, "epoch": 1571, "lr": 4.775485555331904e-05}
{"train_loss": 0.14377819001674652, "global_step": 139849, "epoch": 1571, "lr": 4.7754276401614125e-05}
{"train_loss": 0.08969438076019287, "global_step": 139850, "epoch": 1571, "lr": 4.7753697250211095e-05}
{"train_loss": 0.12438279390335083, "global_step": 139851, "epoch": 1571, "lr": 4.775311809911007e-05}
{"train_loss": 0.14447219669818878, "global_step": 139852, "epoch": 1571, "lr": 4.7752538948311096e-05}
{"train_loss": 0.07647490501403809, "global_step": 139853, "epoch": 1571, "lr": 4.775195979781428e-05}
{"train_loss": 0.06973880529403687, "global_step": 139854, "epoch": 1571, "lr": 4.775138064761967e-05}
{"train_loss": 0.11196593195199966, "global_step": 139855, "epoch": 1571, "lr": 4.775080149772738e-05}
{"train_loss": 0.1635219156742096, "global_step": 139856, "epoch": 1571, "lr": 4.7750222348137456e-05}
{"train_loss": 0.14265161752700806, "global_step": 139857, "epoch": 1571, "lr": 4.774964319884998e-05}
{"train_loss": 0.1996261179447174, "global_step": 139858, "epoch": 1571, "lr": 4.7749064049865055e-05}
{"train_loss": 0.16778333485126495, "global_step": 139859, "epoch": 1571, "lr": 4.7748484901182715e-05}
{"train_loss": 0.0768340677022934, "global_step": 139860, "epoch": 1571, "lr": 4.774790575280309e-05}
{"train_loss": 0.11630354076623917, "global_step": 139861, "epoch": 1571, "lr": 4.774732660472621e-05}
{"train_loss": 0.10437887907028198, "global_step": 139862, "epoch": 1571, "lr": 4.77467474569522e-05}
{"train_loss": 0.14942650496959686, "global_step": 139863, "epoch": 1571, "lr": 4.7746168309481085e-05}
{"train_loss": 0.18369074165821075, "global_step": 139864, "epoch": 1571, "lr": 4.7745589162312996e-05}
{"train_loss": 0.13139565289020538, "global_step": 139865, "epoch": 1571, "lr": 4.774501001544796e-05}
{"train_loss": 0.11431670188903809, "global_step": 139866, "epoch": 1571, "lr": 4.7744430868886104e-05}
{"train_loss": 0.06472957134246826, "global_step": 139867, "epoch": 1571, "lr": 4.774385172262746e-05}
{"train_loss": 0.11839993298053741, "global_step": 139868, "epoch": 1571, "lr": 4.7743272576672144e-05}
{"train_loss": 0.08845342695713043, "global_step": 139869, "epoch": 1571, "lr": 4.77426934310202e-05}
{"train_loss": 0.10576135665178299, "global_step": 139870, "epoch": 1571, "lr": 4.7742114285671744e-05}
{"train_loss": 0.11099701374769211, "global_step": 139871, "epoch": 1571, "lr": 4.7741535140626815e-05}
{"train_loss": 0.04536882042884827, "global_step": 139872, "epoch": 1571, "lr": 4.774095599588553e-05}
{"train_loss": 0.12949660420417786, "global_step": 139873, "epoch": 1571, "lr": 4.774037685144792e-05}
{"train_loss": 0.17164775729179382, "global_step": 139874, "epoch": 1571, "lr": 4.773979770731411e-05}
{"train_loss": 0.0634312704205513, "global_step": 139875, "epoch": 1571, "lr": 4.7739218563484136e-05}
{"train_loss": 0.15026096999645233, "global_step": 139876, "epoch": 1571, "lr": 4.773863941995811e-05}
{"train_loss": 0.10868632793426514, "global_step": 139877, "epoch": 1571, "lr": 4.7738060276736106e-05}
{"train_loss": 0.09884658455848694, "global_step": 139878, "epoch": 1571, "lr": 4.773748113381817e-05}
{"train_loss": 0.17964133620262146, "global_step": 139879, "epoch": 1571, "lr": 4.773690199120443e-05}
{"train_loss": 0.1026868149638176, "global_step": 139880, "epoch": 1571, "lr": 4.773632284889491e-05}
{"train_loss": 0.12034552544355392, "global_step": 139881, "epoch": 1571, "lr": 4.7735743706889735e-05}
{"train_loss": 0.085237517952919, "global_step": 139882, "epoch": 1571, "lr": 4.773516456518894e-05}
{"train_loss": 0.08301635831594467, "global_step": 139883, "epoch": 1571, "lr": 4.773458542379264e-05}
{"train_loss": 0.15140339732170105, "global_step": 139884, "epoch": 1571, "lr": 4.773400628270088e-05}
{"train_loss": 0.08611690998077393, "global_step": 139885, "epoch": 1571, "lr": 4.7733427141913784e-05}
{"train_loss": 0.12341519445180893, "global_step": 139886, "epoch": 1571, "lr": 4.773284800143137e-05}
{"train_loss": 0.10353145003318787, "global_step": 139887, "epoch": 1571, "lr": 4.773226886125377e-05}
{"train_loss": 0.11288946866989136, "global_step": 139888, "epoch": 1571, "lr": 4.773168972138102e-05}
{"train_loss": 0.14137983322143555, "global_step": 139889, "epoch": 1571, "lr": 4.7731110581813245e-05}
{"train_loss": 0.0840243324637413, "global_step": 139890, "epoch": 1571, "lr": 4.773053144255046e-05}
{"train_loss": 0.12669505178928375, "global_step": 139891, "epoch": 1571, "lr": 4.7729952303592804e-05}
{"train_loss": 0.1447940468788147, "global_step": 139892, "epoch": 1571, "lr": 4.772937316494031e-05}
{"train_loss": 0.09836862981319427, "global_step": 139893, "epoch": 1571, "lr": 4.772879402659308e-05}
{"train_loss": 0.09172726422548294, "global_step": 139894, "epoch": 1571, "lr": 4.7728214888551195e-05}
{"train_loss": 0.09426461160182953, "global_step": 139895, "epoch": 1571, "lr": 4.772763575081472e-05}
{"train_loss": 0.0730496495962143, "global_step": 139896, "epoch": 1571, "lr": 4.772705661338374e-05}
{"train_loss": 0.08260886371135712, "global_step": 139897, "epoch": 1571, "lr": 4.772647747625832e-05}
{"train_loss": 0.10293451696634293, "global_step": 139898, "epoch": 1571, "lr": 4.772589833943855e-05}
{"train_loss": 0.11974813044071198, "global_step": 139899, "epoch": 1571, "lr": 4.77253192029245e-05}
{"train_loss": 0.19048811495304108, "global_step": 139900, "epoch": 1571, "lr": 4.772474006671627e-05}
{"train_loss": 0.11452220380306244, "global_step": 139901, "epoch": 1571, "lr": 4.77241609308139e-05}
{"train_loss": 0.07063641399145126, "global_step": 139902, "epoch": 1571, "lr": 4.772358179521751e-05}
{"train_loss": 0.07263462245464325, "global_step": 139903, "epoch": 1571, "lr": 4.772300265992713e-05}
{"train_loss": 0.07928963750600815, "global_step": 139904, "epoch": 1571, "lr": 4.7722423524942894e-05}
{"train_loss": 0.10936128348112106, "global_step": 139905, "epoch": 1571, "lr": 4.772184439026483e-05}
{"train_loss": 0.10749579221010208, "global_step": 139906, "epoch": 1571, "lr": 4.772126525589304e-05}
{"train_loss": 0.11604402268703064, "global_step": 139907, "epoch": 1571, "lr": 4.77206861218276e-05, "val_loss": 6.143138408660889}
{"train_loss": 0.1942407339811325, "global_step": 139908, "epoch": 1572, "lr": 4.7720106988068596e-05}
{"train_loss": 0.12743841111660004, "global_step": 139909, "epoch": 1572, "lr": 4.771952785461608e-05}
{"train_loss": 0.19771304726600647, "global_step": 139910, "epoch": 1572, "lr": 4.771894872147016e-05}
{"train_loss": 0.09631222486495972, "global_step": 139911, "epoch": 1572, "lr": 4.771836958863088e-05}
{"train_loss": 0.0980088859796524, "global_step": 139912, "epoch": 1572, "lr": 4.7717790456098346e-05}
{"train_loss": 0.13786299526691437, "global_step": 139913, "epoch": 1572, "lr": 4.771721132387263e-05}
{"train_loss": 0.11112236231565475, "global_step": 139914, "epoch": 1572, "lr": 4.771663219195381e-05}
{"train_loss": 0.15234968066215515, "global_step": 139915, "epoch": 1572, "lr": 4.771605306034197e-05}
{"train_loss": 0.07349562644958496, "global_step": 139916, "epoch": 1572, "lr": 4.7715473929037155e-05}
{"train_loss": 0.12878024578094482, "global_step": 139917, "epoch": 1572, "lr": 4.771489479803949e-05}
{"train_loss": 0.18787269294261932, "global_step": 139918, "epoch": 1572, "lr": 4.771431566734901e-05}
{"train_loss": 0.041684575378894806, "global_step": 139919, "epoch": 1572, "lr": 4.771373653696583e-05}
{"train_loss": 0.10633984208106995, "global_step": 139920, "epoch": 1572, "lr": 4.7713157406889994e-05}
{"train_loss": 0.0531575046479702, "global_step": 139921, "epoch": 1572, "lr": 4.771257827712161e-05}
{"train_loss": 0.11363337934017181, "global_step": 139922, "epoch": 1572, "lr": 4.771199914766072e-05}
{"train_loss": 0.06216965615749359, "global_step": 139923, "epoch": 1572, "lr": 4.771142001850746e-05}
{"train_loss": 0.057618748396635056, "global_step": 139924, "epoch": 1572, "lr": 4.771084088966183e-05}
{"train_loss": 0.09214386343955994, "global_step": 139925, "epoch": 1572, "lr": 4.771026176112399e-05}
{"train_loss": 0.11388153582811356, "global_step": 139926, "epoch": 1572, "lr": 4.7709682632893944e-05}
{"train_loss": 0.10585054755210876, "global_step": 139927, "epoch": 1572, "lr": 4.7709103504971826e-05}
{"train_loss": 0.1262006014585495, "global_step": 139928, "epoch": 1572, "lr": 4.770852437735767e-05}
{"train_loss": 0.1761852502822876, "global_step": 139929, "epoch": 1572, "lr": 4.77079452500516e-05}
{"train_loss": 0.1297144889831543, "global_step": 139930, "epoch": 1572, "lr": 4.7707366123053643e-05}
{"train_loss": 0.11359985917806625, "global_step": 139931, "epoch": 1572, "lr": 4.770678699636392e-05}
{"train_loss": 0.1097835823893547, "global_step": 139932, "epoch": 1572, "lr": 4.7706207869982485e-05}
{"train_loss": 0.0972767323255539, "global_step": 139933, "epoch": 1572, "lr": 4.7705628743909426e-05}
{"train_loss": 0.11374548077583313, "global_step": 139934, "epoch": 1572, "lr": 4.770504961814482e-05}
{"train_loss": 0.06329990178346634, "global_step": 139935, "epoch": 1572, "lr": 4.7704470492688737e-05}
{"train_loss": 0.10179238766431808, "global_step": 139936, "epoch": 1572, "lr": 4.770389136754126e-05}
{"train_loss": 0.0934155061841011, "global_step": 139937, "epoch": 1572, "lr": 4.7703312242702465e-05}
{"train_loss": 0.07179777324199677, "global_step": 139938, "epoch": 1572, "lr": 4.770273311817244e-05}
{"train_loss": 0.0867341086268425, "global_step": 139939, "epoch": 1572, "lr": 4.770215399395124e-05}
{"train_loss": 0.13740229606628418, "global_step": 139940, "epoch": 1572, "lr": 4.770157487003898e-05}
{"train_loss": 0.12444094568490982, "global_step": 139941, "epoch": 1572, "lr": 4.770099574643569e-05}
{"train_loss": 0.0852566584944725, "global_step": 139942, "epoch": 1572, "lr": 4.77004166231415e-05}
{"train_loss": 0.06416135281324387, "global_step": 139943, "epoch": 1572, "lr": 4.769983750015643e-05}
{"train_loss": 0.11848339438438416, "global_step": 139944, "epoch": 1572, "lr": 4.769925837748062e-05}
{"train_loss": 0.11164361238479614, "global_step": 139945, "epoch": 1572, "lr": 4.7698679255114086e-05}
{"train_loss": 0.1102890893816948, "global_step": 139946, "epoch": 1572, "lr": 4.769810013305696e-05}
{"train_loss": 0.08015935122966766, "global_step": 139947, "epoch": 1572, "lr": 4.7697521011309285e-05}
{"train_loss": 0.10903920233249664, "global_step": 139948, "epoch": 1572, "lr": 4.769694188987115e-05}
{"train_loss": 0.13966725766658783, "global_step": 139949, "epoch": 1572, "lr": 4.7696362768742636e-05}
{"train_loss": 0.10635685175657272, "global_step": 139950, "epoch": 1572, "lr": 4.769578364792381e-05}
{"train_loss": 0.13416649401187897, "global_step": 139951, "epoch": 1572, "lr": 4.769520452741477e-05}
{"train_loss": 0.08905012160539627, "global_step": 139952, "epoch": 1572, "lr": 4.769462540721558e-05}
{"train_loss": 0.12575504183769226, "global_step": 139953, "epoch": 1572, "lr": 4.769404628732633e-05}
{"train_loss": 0.13380330801010132, "global_step": 139954, "epoch": 1572, "lr": 4.7693467167747065e-05}
{"train_loss": 0.04933485388755798, "global_step": 139955, "epoch": 1572, "lr": 4.7692888048477904e-05}
{"train_loss": 0.08588279783725739, "global_step": 139956, "epoch": 1572, "lr": 4.7692308929518885e-05}
{"train_loss": 0.15283076465129852, "global_step": 139957, "epoch": 1572, "lr": 4.769172981087013e-05}
{"train_loss": 0.09030574560165405, "global_step": 139958, "epoch": 1572, "lr": 4.769115069253167e-05}
{"train_loss": 0.13827396929264069, "global_step": 139959, "epoch": 1572, "lr": 4.7690571574503635e-05}
{"train_loss": 0.14484015107154846, "global_step": 139960, "epoch": 1572, "lr": 4.768999245678604e-05}
{"train_loss": 0.08433151245117188, "global_step": 139961, "epoch": 1572, "lr": 4.768941333937903e-05}
{"train_loss": 0.13565252721309662, "global_step": 139962, "epoch": 1572, "lr": 4.768883422228263e-05}
{"train_loss": 0.1269354224205017, "global_step": 139963, "epoch": 1572, "lr": 4.7688255105496956e-05}
{"train_loss": 0.12210533022880554, "global_step": 139964, "epoch": 1572, "lr": 4.768767598902204e-05}
{"train_loss": 0.11292542517185211, "global_step": 139965, "epoch": 1572, "lr": 4.7687096872858015e-05}
{"train_loss": 0.20950287580490112, "global_step": 139966, "epoch": 1572, "lr": 4.768651775700491e-05}
{"train_loss": 0.18875393271446228, "global_step": 139967, "epoch": 1572, "lr": 4.7685938641462836e-05}
{"train_loss": 0.12302564084529877, "global_step": 139968, "epoch": 1572, "lr": 4.768535952623186e-05}
{"train_loss": 0.09933771193027496, "global_step": 139969, "epoch": 1572, "lr": 4.768478041131206e-05}
{"train_loss": 0.05150633677840233, "global_step": 139970, "epoch": 1572, "lr": 4.768420129670351e-05}
{"train_loss": 0.11229680478572845, "global_step": 139971, "epoch": 1572, "lr": 4.768362218240629e-05}
{"train_loss": 0.11022362858057022, "global_step": 139972, "epoch": 1572, "lr": 4.768304306842049e-05}
{"train_loss": 0.09001480787992477, "global_step": 139973, "epoch": 1572, "lr": 4.768246395474616e-05}
{"train_loss": 0.09901461005210876, "global_step": 139974, "epoch": 1572, "lr": 4.768188484138341e-05}
{"train_loss": 0.10844745486974716, "global_step": 139975, "epoch": 1572, "lr": 4.7681305728332285e-05}
{"train_loss": 0.08966731280088425, "global_step": 139976, "epoch": 1572, "lr": 4.7680726615592894e-05}
{"train_loss": 0.10602694004774094, "global_step": 139977, "epoch": 1572, "lr": 4.768014750316529e-05}
{"train_loss": 0.10148579627275467, "global_step": 139978, "epoch": 1572, "lr": 4.767956839104958e-05}
{"train_loss": 0.11564402282238007, "global_step": 139979, "epoch": 1572, "lr": 4.76789892792458e-05}
{"train_loss": 0.15685118734836578, "global_step": 139980, "epoch": 1572, "lr": 4.767841016775407e-05}
{"train_loss": 0.09369993209838867, "global_step": 139981, "epoch": 1572, "lr": 4.767783105657443e-05}
{"train_loss": 0.11954841017723083, "global_step": 139982, "epoch": 1572, "lr": 4.7677251945707005e-05}
{"train_loss": 0.11231756210327148, "global_step": 139983, "epoch": 1572, "lr": 4.767667283515183e-05}
{"train_loss": 0.09075237810611725, "global_step": 139984, "epoch": 1572, "lr": 4.767609372490898e-05}
{"train_loss": 0.1348022222518921, "global_step": 139985, "epoch": 1572, "lr": 4.767551461497858e-05}
{"train_loss": 0.12055416405200958, "global_step": 139986, "epoch": 1572, "lr": 4.767493550536067e-05}
{"train_loss": 0.1149180606007576, "global_step": 139987, "epoch": 1572, "lr": 4.767435639605533e-05}
{"train_loss": 0.11048421263694763, "global_step": 139988, "epoch": 1572, "lr": 4.767377728706265e-05}
{"train_loss": 0.09562992304563522, "global_step": 139989, "epoch": 1572, "lr": 4.76731981783827e-05}
{"train_loss": 0.10383865237236023, "global_step": 139990, "epoch": 1572, "lr": 4.767261907001557e-05}
{"train_loss": 0.08020301163196564, "global_step": 139991, "epoch": 1572, "lr": 4.767203996196131e-05}
{"train_loss": 0.14195901155471802, "global_step": 139992, "epoch": 1572, "lr": 4.767146085422003e-05}
{"train_loss": 0.11656539142131805, "global_step": 139993, "epoch": 1572, "lr": 4.76708817467918e-05}
{"train_loss": 0.07658923417329788, "global_step": 139994, "epoch": 1572, "lr": 4.767030263967667e-05}
{"train_loss": 0.16461117565631866, "global_step": 139995, "epoch": 1572, "lr": 4.766972353287476e-05}
{"train_loss": 0.11311550325389659, "global_step": 139996, "epoch": 1572, "lr": 4.766914442638611e-05, "val_loss": 6.046497821807861}
{"train_loss": 0.12426739931106567, "global_step": 139997, "epoch": 1573, "lr": 4.766856532021083e-05}
{"train_loss": 0.10925529152154922, "global_step": 139998, "epoch": 1573, "lr": 4.7667986214348965e-05}
{"train_loss": 0.07453396171331406, "global_step": 139999, "epoch": 1573, "lr": 4.766740710880063e-05}
{"train_loss": 0.10055434703826904, "global_step": 140000, "epoch": 1573, "lr": 4.7666828003565865e-05}
{"train_loss": 0.14594992995262146, "global_step": 140001, "epoch": 1573, "lr": 4.766624889864479e-05}
{"train_loss": 0.12175285816192627, "global_step": 140002, "epoch": 1573, "lr": 4.766566979403743e-05}
{"train_loss": 0.07334443926811218, "global_step": 140003, "epoch": 1573, "lr": 4.76650906897439e-05}
{"train_loss": 0.08800966292619705, "global_step": 140004, "epoch": 1573, "lr": 4.766451158576429e-05}
{"train_loss": 0.15295366942882538, "global_step": 140005, "epoch": 1573, "lr": 4.766393248209863e-05}
{"train_loss": 0.08926121890544891, "global_step": 140006, "epoch": 1573, "lr": 4.7663353378747055e-05}
{"train_loss": 0.10042067617177963, "global_step": 140007, "epoch": 1573, "lr": 4.7662774275709606e-05}
{"train_loss": 0.10818753391504288, "global_step": 140008, "epoch": 1573, "lr": 4.766219517298636e-05}
{"train_loss": 0.11383191496133804, "global_step": 140009, "epoch": 1573, "lr": 4.766161607057741e-05}
{"train_loss": 0.08229173719882965, "global_step": 140010, "epoch": 1573, "lr": 4.7661036968482817e-05}
{"train_loss": 0.10736922919750214, "global_step": 140011, "epoch": 1573, "lr": 4.7660457866702676e-05}
{"train_loss": 0.11059471219778061, "global_step": 140012, "epoch": 1573, "lr": 4.765987876523707e-05}
{"train_loss": 0.079767607152462, "global_step": 140013, "epoch": 1573, "lr": 4.7659299664086045e-05}
{"train_loss": 0.1469411849975586, "global_step": 140014, "epoch": 1573, "lr": 4.7658720563249716e-05}
{"train_loss": 0.07084052264690399, "global_step": 140015, "epoch": 1573, "lr": 4.7658141462728125e-05}
{"train_loss": 0.1357443779706955, "global_step": 140016, "epoch": 1573, "lr": 4.7657562362521386e-05}
{"train_loss": 0.14899033308029175, "global_step": 140017, "epoch": 1573, "lr": 4.765698326262954e-05}
{"train_loss": 0.11310728639364243, "global_step": 140018, "epoch": 1573, "lr": 4.76564041630527e-05}
{"train_loss": 0.13004818558692932, "global_step": 140019, "epoch": 1573, "lr": 4.765582506379091e-05}
{"train_loss": 0.08844996243715286, "global_step": 140020, "epoch": 1573, "lr": 4.765524596484427e-05}
{"train_loss": 0.0901598408818245, "global_step": 140021, "epoch": 1573, "lr": 4.765466686621287e-05}
{"train_loss": 0.1348615437746048, "global_step": 140022, "epoch": 1573, "lr": 4.765408776789675e-05}
{"train_loss": 0.10711443424224854, "global_step": 140023, "epoch": 1573, "lr": 4.765350866989603e-05}
{"train_loss": 0.18309319019317627, "global_step": 140024, "epoch": 1573, "lr": 4.765292957221076e-05}
{"train_loss": 0.1528322696685791, "global_step": 140025, "epoch": 1573, "lr": 4.765235047484103e-05}
{"train_loss": 0.1070258617401123, "global_step": 140026, "epoch": 1573, "lr": 4.765177137778691e-05}
{"train_loss": 0.1576070785522461, "global_step": 140027, "epoch": 1573, "lr": 4.765119228104847e-05}
{"train_loss": 0.15072035789489746, "global_step": 140028, "epoch": 1573, "lr": 4.7650613184625816e-05}
{"train_loss": 0.05790876969695091, "global_step": 140029, "epoch": 1573, "lr": 4.7650034088518994e-05}
{"train_loss": 0.18650604784488678, "global_step": 140030, "epoch": 1573, "lr": 4.76494549927281e-05}
{"train_loss": 0.12141964584589005, "global_step": 140031, "epoch": 1573, "lr": 4.764887589725323e-05}
{"train_loss": 0.1733730435371399, "global_step": 140032, "epoch": 1573, "lr": 4.76482968020944e-05}
{"train_loss": 0.059998996555805206, "global_step": 140033, "epoch": 1573, "lr": 4.764771770725176e-05}
{"train_loss": 0.05881557613611221, "global_step": 140034, "epoch": 1573, "lr": 4.764713861272533e-05}
{"train_loss": 0.10766008496284485, "global_step": 140035, "epoch": 1573, "lr": 4.764655951851524e-05}
{"train_loss": 0.08561257272958755, "global_step": 140036, "epoch": 1573, "lr": 4.7645980424621516e-05}
{"train_loss": 0.1202482208609581, "global_step": 140037, "epoch": 1573, "lr": 4.764540133104428e-05}
{"train_loss": 0.09350160509347916, "global_step": 140038, "epoch": 1573, "lr": 4.764482223778358e-05}
{"train_loss": 0.11175118386745453, "global_step": 140039, "epoch": 1573, "lr": 4.7644243144839494e-05}
{"train_loss": 0.0998486801981926, "global_step": 140040, "epoch": 1573, "lr": 4.764366405221214e-05}
{"train_loss": 0.1580507457256317, "global_step": 140041, "epoch": 1573, "lr": 4.764308495990154e-05}
{"train_loss": 0.10668623447418213, "global_step": 140042, "epoch": 1573, "lr": 4.764250586790782e-05}
{"train_loss": 0.11706677079200745, "global_step": 140043, "epoch": 1573, "lr": 4.7641926776231015e-05}
{"train_loss": 0.12792570888996124, "global_step": 140044, "epoch": 1573, "lr": 4.764134768487125e-05}
{"train_loss": 0.09614851325750351, "global_step": 140045, "epoch": 1573, "lr": 4.764076859382856e-05}
{"train_loss": 0.21639129519462585, "global_step": 140046, "epoch": 1573, "lr": 4.764018950310305e-05}
{"train_loss": 0.09216917306184769, "global_step": 140047, "epoch": 1573, "lr": 4.763961041269478e-05}
{"train_loss": 0.0891125500202179, "global_step": 140048, "epoch": 1573, "lr": 4.7639031322603835e-05}
{"train_loss": 0.13373076915740967, "global_step": 140049, "epoch": 1573, "lr": 4.763845223283029e-05}
{"train_loss": 0.13509616255760193, "global_step": 140050, "epoch": 1573, "lr": 4.7637873143374244e-05}
{"train_loss": 0.05759146064519882, "global_step": 140051, "epoch": 1573, "lr": 4.763729405423574e-05}
{"train_loss": 0.09445695579051971, "global_step": 140052, "epoch": 1573, "lr": 4.7636714965414884e-05}
{"train_loss": 0.11503979563713074, "global_step": 140053, "epoch": 1573, "lr": 4.763613587691173e-05}
{"train_loss": 0.056682735681533813, "global_step": 140054, "epoch": 1573, "lr": 4.7635556788726385e-05}
{"train_loss": 0.09828644245862961, "global_step": 140055, "epoch": 1573, "lr": 4.763497770085889e-05}
{"train_loss": 0.1082170158624649, "global_step": 140056, "epoch": 1573, "lr": 4.7634398613309364e-05}
{"train_loss": 0.06542346626520157, "global_step": 140057, "epoch": 1573, "lr": 4.763381952607784e-05}
{"train_loss": 0.15627475082874298, "global_step": 140058, "epoch": 1573, "lr": 4.763324043916443e-05}
{"train_loss": 0.08134141564369202, "global_step": 140059, "epoch": 1573, "lr": 4.763266135256921e-05}
{"train_loss": 0.1625015139579773, "global_step": 140060, "epoch": 1573, "lr": 4.7632082266292236e-05}
{"train_loss": 0.10102711617946625, "global_step": 140061, "epoch": 1573, "lr": 4.763150318033363e-05}
{"train_loss": 0.13955500721931458, "global_step": 140062, "epoch": 1573, "lr": 4.76309240946934e-05}
{"train_loss": 0.06199847161769867, "global_step": 140063, "epoch": 1573, "lr": 4.7630345009371695e-05}
{"train_loss": 0.1344919055700302, "global_step": 140064, "epoch": 1573, "lr": 4.762976592436855e-05}
{"train_loss": 0.07122455537319183, "global_step": 140065, "epoch": 1573, "lr": 4.762918683968406e-05}
{"train_loss": 0.20887048542499542, "global_step": 140066, "epoch": 1573, "lr": 4.7628607755318286e-05}
{"train_loss": 0.11405452340841293, "global_step": 140067, "epoch": 1573, "lr": 4.762802867127132e-05}
{"train_loss": 0.13063974678516388, "global_step": 140068, "epoch": 1573, "lr": 4.7627449587543236e-05}
{"train_loss": 0.1562892496585846, "global_step": 140069, "epoch": 1573, "lr": 4.762687050413413e-05}
{"train_loss": 0.11506431549787521, "global_step": 140070, "epoch": 1573, "lr": 4.7626291421044036e-05}
{"train_loss": 0.1364368349313736, "global_step": 140071, "epoch": 1573, "lr": 4.762571233827308e-05}
{"train_loss": 0.14147327840328217, "global_step": 140072, "epoch": 1573, "lr": 4.76251332558213e-05}
{"train_loss": 0.10026620328426361, "global_step": 140073, "epoch": 1573, "lr": 4.762455417368881e-05}
{"train_loss": 0.09644846618175507, "global_step": 140074, "epoch": 1573, "lr": 4.7623975091875646e-05}
{"train_loss": 0.15094931423664093, "global_step": 140075, "epoch": 1573, "lr": 4.7623396010381904e-05}
{"train_loss": 0.10265110433101654, "global_step": 140076, "epoch": 1573, "lr": 4.76228169292077e-05}
{"train_loss": 0.11382805556058884, "global_step": 140077, "epoch": 1573, "lr": 4.762223784835305e-05}
{"train_loss": 0.17803877592086792, "global_step": 140078, "epoch": 1573, "lr": 4.762165876781809e-05}
{"train_loss": 0.19526086747646332, "global_step": 140079, "epoch": 1573, "lr": 4.762107968760284e-05}
{"train_loss": 0.12064767628908157, "global_step": 140080, "epoch": 1573, "lr": 4.7620500607707436e-05}
{"train_loss": 0.15976545214653015, "global_step": 140081, "epoch": 1573, "lr": 4.76199215281319e-05}
{"train_loss": 0.11807174235582352, "global_step": 140082, "epoch": 1573, "lr": 4.7619342448876364e-05}
{"train_loss": 0.1332665979862213, "global_step": 140083, "epoch": 1573, "lr": 4.761876336994086e-05}
{"train_loss": 0.11080735921859741, "global_step": 140084, "epoch": 1573, "lr": 4.761818429132549e-05}
{"train_loss": 0.11807573122087489, "global_step": 140085, "epoch": 1573, "lr": 4.761760521303033e-05, "val_loss": 6.162757873535156}
{"train_loss": 0.10107097774744034, "global_step": 140086, "epoch": 1574, "lr": 4.7617026135055446e-05}
{"train_loss": 0.12515999376773834, "global_step": 140087, "epoch": 1574, "lr": 4.761644705740092e-05}
{"train_loss": 0.07539881765842438, "global_step": 140088, "epoch": 1574, "lr": 4.761586798006685e-05}
{"train_loss": 0.14854754507541656, "global_step": 140089, "epoch": 1574, "lr": 4.761528890305328e-05}
{"train_loss": 0.11084989458322525, "global_step": 140090, "epoch": 1574, "lr": 4.7614709826360323e-05}
{"train_loss": 0.10927869379520416, "global_step": 140091, "epoch": 1574, "lr": 4.7614130749988014e-05}
{"train_loss": 0.09452839195728302, "global_step": 140092, "epoch": 1574, "lr": 4.761355167393648e-05}
{"train_loss": 0.13165971636772156, "global_step": 140093, "epoch": 1574, "lr": 4.761297259820575e-05}
{"train_loss": 0.1322983354330063, "global_step": 140094, "epoch": 1574, "lr": 4.761239352279593e-05}
{"train_loss": 0.10787755250930786, "global_step": 140095, "epoch": 1574, "lr": 4.761181444770712e-05}
{"train_loss": 0.1025572195649147, "global_step": 140096, "epoch": 1574, "lr": 4.761123537293935e-05}
{"train_loss": 0.07920863479375839, "global_step": 140097, "epoch": 1574, "lr": 4.761065629849274e-05}
{"train_loss": 0.09225428849458694, "global_step": 140098, "epoch": 1574, "lr": 4.761007722436733e-05}
{"train_loss": 0.12393739074468613, "global_step": 140099, "epoch": 1574, "lr": 4.760949815056323e-05}
{"train_loss": 0.14845417439937592, "global_step": 140100, "epoch": 1574, "lr": 4.760891907708049e-05}
{"train_loss": 0.08902686089277267, "global_step": 140101, "epoch": 1574, "lr": 4.7608340003919214e-05}
{"train_loss": 0.10775233805179596, "global_step": 140102, "epoch": 1574, "lr": 4.7607760931079465e-05}
{"train_loss": 0.16816531121730804, "global_step": 140103, "epoch": 1574, "lr": 4.7607181858561325e-05}
{"train_loss": 0.08346683531999588, "global_step": 140104, "epoch": 1574, "lr": 4.760660278636486e-05}
{"train_loss": 0.12823310494422913, "global_step": 140105, "epoch": 1574, "lr": 4.760602371449017e-05}
{"train_loss": 0.10940393060445786, "global_step": 140106, "epoch": 1574, "lr": 4.760544464293731e-05}
{"train_loss": 0.13949955999851227, "global_step": 140107, "epoch": 1574, "lr": 4.760486557170638e-05}
{"train_loss": 0.10881111025810242, "global_step": 140108, "epoch": 1574, "lr": 4.760428650079744e-05}
{"train_loss": 0.16670700907707214, "global_step": 140109, "epoch": 1574, "lr": 4.760370743021058e-05}
{"train_loss": 0.14411990344524384, "global_step": 140110, "epoch": 1574, "lr": 4.760312835994586e-05}
{"train_loss": 0.1658141314983368, "global_step": 140111, "epoch": 1574, "lr": 4.7602549290003364e-05}
{"train_loss": 0.07125233113765717, "global_step": 140112, "epoch": 1574, "lr": 4.76019702203832e-05}
{"train_loss": 0.10889208316802979, "global_step": 140113, "epoch": 1574, "lr": 4.76013911510854e-05}
{"train_loss": 0.08373720943927765, "global_step": 140114, "epoch": 1574, "lr": 4.7600812082110086e-05}
{"train_loss": 0.11332619190216064, "global_step": 140115, "epoch": 1574, "lr": 4.760023301345729e-05}
{"train_loss": 0.12771832942962646, "global_step": 140116, "epoch": 1574, "lr": 4.759965394512714e-05}
{"train_loss": 0.09813138097524643, "global_step": 140117, "epoch": 1574, "lr": 4.759907487711966e-05}
{"train_loss": 0.15196813642978668, "global_step": 140118, "epoch": 1574, "lr": 4.759849580943498e-05}
{"train_loss": 0.09071746468544006, "global_step": 140119, "epoch": 1574, "lr": 4.759791674207313e-05}
{"train_loss": 0.10355952382087708, "global_step": 140120, "epoch": 1574, "lr": 4.7597337675034234e-05}
{"train_loss": 0.11303538084030151, "global_step": 140121, "epoch": 1574, "lr": 4.7596758608318324e-05}
{"train_loss": 0.09272801131010056, "global_step": 140122, "epoch": 1574, "lr": 4.7596179541925526e-05}
{"train_loss": 0.10075723379850388, "global_step": 140123, "epoch": 1574, "lr": 4.7595600475855875e-05}
{"train_loss": 0.09205802530050278, "global_step": 140124, "epoch": 1574, "lr": 4.759502141010947e-05}
{"train_loss": 0.12157286703586578, "global_step": 140125, "epoch": 1574, "lr": 4.759444234468638e-05}
{"train_loss": 0.12517020106315613, "global_step": 140126, "epoch": 1574, "lr": 4.7593863279586694e-05}
{"train_loss": 0.09732027351856232, "global_step": 140127, "epoch": 1574, "lr": 4.759328421481048e-05}
{"train_loss": 0.13116009533405304, "global_step": 140128, "epoch": 1574, "lr": 4.759270515035784e-05}
{"train_loss": 0.1379142701625824, "global_step": 140129, "epoch": 1574, "lr": 4.759212608622879e-05}
{"train_loss": 0.10233084857463837, "global_step": 140130, "epoch": 1574, "lr": 4.7591547022423464e-05}
{"train_loss": 0.07814054936170578, "global_step": 140131, "epoch": 1574, "lr": 4.7590967958941946e-05}
{"train_loss": 0.07366711646318436, "global_step": 140132, "epoch": 1574, "lr": 4.759038889578427e-05}
{"train_loss": 0.12009720504283905, "global_step": 140133, "epoch": 1574, "lr": 4.758980983295056e-05}
{"train_loss": 0.1052846685051918, "global_step": 140134, "epoch": 1574, "lr": 4.7589230770440844e-05}
{"train_loss": 0.07174189388751984, "global_step": 140135, "epoch": 1574, "lr": 4.758865170825525e-05}
{"train_loss": 0.11711373180150986, "global_step": 140136, "epoch": 1574, "lr": 4.758807264639381e-05}
{"train_loss": 0.18558178842067719, "global_step": 140137, "epoch": 1574, "lr": 4.758749358485665e-05}
{"train_loss": 0.14455392956733704, "global_step": 140138, "epoch": 1574, "lr": 4.75869145236438e-05}
{"train_loss": 0.13982897996902466, "global_step": 140139, "epoch": 1574, "lr": 4.7586335462755374e-05}
{"train_loss": 0.12187453359365463, "global_step": 140140, "epoch": 1574, "lr": 4.7585756402191424e-05}
{"train_loss": 0.1076667308807373, "global_step": 140141, "epoch": 1574, "lr": 4.7585177341952056e-05}
{"train_loss": 0.1338738203048706, "global_step": 140142, "epoch": 1574, "lr": 4.7584598282037326e-05}
{"train_loss": 0.09534350782632828, "global_step": 140143, "epoch": 1574, "lr": 4.7584019222447314e-05}
{"train_loss": 0.06778666377067566, "global_step": 140144, "epoch": 1574, "lr": 4.7583440163182095e-05}
{"train_loss": 0.11217647045850754, "global_step": 140145, "epoch": 1574, "lr": 4.758286110424176e-05}
{"train_loss": 0.13087479770183563, "global_step": 140146, "epoch": 1574, "lr": 4.7582282045626376e-05}
{"train_loss": 0.1360710710287094, "global_step": 140147, "epoch": 1574, "lr": 4.758170298733602e-05}
{"train_loss": 0.08225230127573013, "global_step": 140148, "epoch": 1574, "lr": 4.758112392937079e-05}
{"train_loss": 0.10725667327642441, "global_step": 140149, "epoch": 1574, "lr": 4.7580544871730724e-05}
{"train_loss": 0.11201214045286179, "global_step": 140150, "epoch": 1574, "lr": 4.757996581441596e-05}
{"train_loss": 0.0838974118232727, "global_step": 140151, "epoch": 1574, "lr": 4.75793867574265e-05}
{"train_loss": 0.12039108574390411, "global_step": 140152, "epoch": 1574, "lr": 4.75788077007625e-05}
{"train_loss": 0.11942244321107864, "global_step": 140153, "epoch": 1574, "lr": 4.757822864442397e-05}
{"train_loss": 0.056409358978271484, "global_step": 140154, "epoch": 1574, "lr": 4.757764958841104e-05}
{"train_loss": 0.13219116628170013, "global_step": 140155, "epoch": 1574, "lr": 4.757707053272375e-05}
{"train_loss": 0.07811742275953293, "global_step": 140156, "epoch": 1574, "lr": 4.75764914773622e-05}
{"train_loss": 0.09798337519168854, "global_step": 140157, "epoch": 1574, "lr": 4.757591242232645e-05}
{"train_loss": 0.10585952550172806, "global_step": 140158, "epoch": 1574, "lr": 4.757533336761661e-05}
{"train_loss": 0.1266038715839386, "global_step": 140159, "epoch": 1574, "lr": 4.7574754313232714e-05}
{"train_loss": 0.12153566628694534, "global_step": 140160, "epoch": 1574, "lr": 4.7574175259174885e-05}
{"train_loss": 0.13962988555431366, "global_step": 140161, "epoch": 1574, "lr": 4.757359620544317e-05}
{"train_loss": 0.11247825622558594, "global_step": 140162, "epoch": 1574, "lr": 4.757301715203765e-05}
{"train_loss": 0.12980332970619202, "global_step": 140163, "epoch": 1574, "lr": 4.757243809895841e-05}
{"train_loss": 0.10424725711345673, "global_step": 140164, "epoch": 1574, "lr": 4.757185904620553e-05}
{"train_loss": 0.09707902371883392, "global_step": 140165, "epoch": 1574, "lr": 4.757127999377908e-05}
{"train_loss": 0.08124065399169922, "global_step": 140166, "epoch": 1574, "lr": 4.757070094167914e-05}
{"train_loss": 0.14542953670024872, "global_step": 140167, "epoch": 1574, "lr": 4.75701218899058e-05}
{"train_loss": 0.18482016026973724, "global_step": 140168, "epoch": 1574, "lr": 4.756954283845911e-05}
{"train_loss": 0.07820992916822433, "global_step": 140169, "epoch": 1574, "lr": 4.7568963787339185e-05}
{"train_loss": 0.10511265695095062, "global_step": 140170, "epoch": 1574, "lr": 4.7568384736546064e-05}
{"train_loss": 0.07254589349031448, "global_step": 140171, "epoch": 1574, "lr": 4.7567805686079864e-05}
{"train_loss": 0.10434029251337051, "global_step": 140172, "epoch": 1574, "lr": 4.756722663594062e-05}
{"train_loss": 0.14596989750862122, "global_step": 140173, "epoch": 1574, "lr": 4.756664758612846e-05}
{"train_loss": 0.11286181881186667, "global_step": 140174, "epoch": 1574, "lr": 4.7566068536643407e-05, "val_loss": 5.955848217010498}
{"train_loss": 0.0928306058049202, "global_step": 140175, "epoch": 1575, "lr": 4.7565489487485595e-05}
{"train_loss": 0.07114674896001816, "global_step": 140176, "epoch": 1575, "lr": 4.756491043865504e-05}
{"train_loss": 0.15002432465553284, "global_step": 140177, "epoch": 1575, "lr": 4.7564331390151886e-05}
{"train_loss": 0.15609882771968842, "global_step": 140178, "epoch": 1575, "lr": 4.7563752341976144e-05}
{"train_loss": 0.09980063140392303, "global_step": 140179, "epoch": 1575, "lr": 4.756317329412796e-05}
{"train_loss": 0.165403351187706, "global_step": 140180, "epoch": 1575, "lr": 4.756259424660736e-05}
{"train_loss": 0.11335940659046173, "global_step": 140181, "epoch": 1575, "lr": 4.756201519941444e-05}
{"train_loss": 0.10138001292943954, "global_step": 140182, "epoch": 1575, "lr": 4.7561436152549285e-05}
{"train_loss": 0.07135677337646484, "global_step": 140183, "epoch": 1575, "lr": 4.7560857106011957e-05}
{"train_loss": 0.09484036266803741, "global_step": 140184, "epoch": 1575, "lr": 4.756027805980255e-05}
{"train_loss": 0.13260258734226227, "global_step": 140185, "epoch": 1575, "lr": 4.755969901392112e-05}
{"train_loss": 0.12239426374435425, "global_step": 140186, "epoch": 1575, "lr": 4.755911996836777e-05}
{"train_loss": 0.0678199827671051, "global_step": 140187, "epoch": 1575, "lr": 4.7558540923142556e-05}
{"train_loss": 0.18994468450546265, "global_step": 140188, "epoch": 1575, "lr": 4.7557961878245584e-05}
{"train_loss": 0.10646931082010269, "global_step": 140189, "epoch": 1575, "lr": 4.7557382833676895e-05}
{"train_loss": 0.06142314523458481, "global_step": 140190, "epoch": 1575, "lr": 4.7556803789436604e-05}
{"train_loss": 0.09578247368335724, "global_step": 140191, "epoch": 1575, "lr": 4.755622474552475e-05}
{"train_loss": 0.12262716889381409, "global_step": 140192, "epoch": 1575, "lr": 4.755564570194146e-05}
{"train_loss": 0.11417824774980545, "global_step": 140193, "epoch": 1575, "lr": 4.755506665868675e-05}
{"train_loss": 0.10993093997240067, "global_step": 140194, "epoch": 1575, "lr": 4.755448761576076e-05}
{"train_loss": 0.07965154945850372, "global_step": 140195, "epoch": 1575, "lr": 4.755390857316352e-05}
{"train_loss": 0.1676521599292755, "global_step": 140196, "epoch": 1575, "lr": 4.755332953089515e-05}
{"train_loss": 0.058702949434518814, "global_step": 140197, "epoch": 1575, "lr": 4.755275048895568e-05}
{"train_loss": 0.130442813038826, "global_step": 140198, "epoch": 1575, "lr": 4.7552171447345234e-05}
{"train_loss": 0.08751557767391205, "global_step": 140199, "epoch": 1575, "lr": 4.755159240606385e-05}
{"train_loss": 0.10547246038913727, "global_step": 140200, "epoch": 1575, "lr": 4.755101336511164e-05}
{"train_loss": 0.07765254378318787, "global_step": 140201, "epoch": 1575, "lr": 4.755043432448866e-05}
{"train_loss": 0.12778714299201965, "global_step": 140202, "epoch": 1575, "lr": 4.7549855284194984e-05}
{"train_loss": 0.059816230088472366, "global_step": 140203, "epoch": 1575, "lr": 4.754927624423072e-05}
{"train_loss": 0.11149647831916809, "global_step": 140204, "epoch": 1575, "lr": 4.75486972045959e-05}
{"train_loss": 0.10455607622861862, "global_step": 140205, "epoch": 1575, "lr": 4.7548118165290655e-05}
{"train_loss": 0.1393129825592041, "global_step": 140206, "epoch": 1575, "lr": 4.754753912631501e-05}
{"train_loss": 0.08076499402523041, "global_step": 140207, "epoch": 1575, "lr": 4.7546960087669104e-05}
{"train_loss": 0.09017053246498108, "global_step": 140208, "epoch": 1575, "lr": 4.754638104935294e-05}
{"train_loss": 0.14509771764278412, "global_step": 140209, "epoch": 1575, "lr": 4.754580201136667e-05}
{"train_loss": 0.11591651290655136, "global_step": 140210, "epoch": 1575, "lr": 4.754522297371031e-05}
{"train_loss": 0.07046684622764587, "global_step": 140211, "epoch": 1575, "lr": 4.754464393638398e-05}
{"train_loss": 0.09091091901063919, "global_step": 140212, "epoch": 1575, "lr": 4.754406489938773e-05}
{"train_loss": 0.11238500475883484, "global_step": 140213, "epoch": 1575, "lr": 4.7543485862721666e-05}
{"train_loss": 0.10722425580024719, "global_step": 140214, "epoch": 1575, "lr": 4.754290682638583e-05}
{"train_loss": 0.10889290273189545, "global_step": 140215, "epoch": 1575, "lr": 4.754232779038035e-05}
{"train_loss": 0.09991885721683502, "global_step": 140216, "epoch": 1575, "lr": 4.754174875470524e-05}
{"train_loss": 0.07683636248111725, "global_step": 140217, "epoch": 1575, "lr": 4.754116971936064e-05}
{"train_loss": 0.08195453137159348, "global_step": 140218, "epoch": 1575, "lr": 4.754059068434659e-05}
{"train_loss": 0.12012923508882523, "global_step": 140219, "epoch": 1575, "lr": 4.754001164966318e-05}
{"train_loss": 0.12370619177818298, "global_step": 140220, "epoch": 1575, "lr": 4.753943261531048e-05}
{"train_loss": 0.08899696916341782, "global_step": 140221, "epoch": 1575, "lr": 4.753885358128856e-05}
{"train_loss": 0.13404420018196106, "global_step": 140222, "epoch": 1575, "lr": 4.753827454759754e-05}
{"train_loss": 0.09510420262813568, "global_step": 140223, "epoch": 1575, "lr": 4.753769551423744e-05}
{"train_loss": 0.08758759498596191, "global_step": 140224, "epoch": 1575, "lr": 4.7537116481208396e-05}
{"train_loss": 0.09261316061019897, "global_step": 140225, "epoch": 1575, "lr": 4.753653744851043e-05}
{"train_loss": 0.1555662304162979, "global_step": 140226, "epoch": 1575, "lr": 4.7535958416143674e-05}
{"train_loss": 0.061125896871089935, "global_step": 140227, "epoch": 1575, "lr": 4.7535379384108145e-05}
{"train_loss": 0.11724378168582916, "global_step": 140228, "epoch": 1575, "lr": 4.753480035240398e-05}
{"train_loss": 0.0653039813041687, "global_step": 140229, "epoch": 1575, "lr": 4.753422132103122e-05}
{"train_loss": 0.08514275401830673, "global_step": 140230, "epoch": 1575, "lr": 4.7533642289989966e-05}
{"train_loss": 0.12377955764532089, "global_step": 140231, "epoch": 1575, "lr": 4.753306325928026e-05}
{"train_loss": 0.17695896327495575, "global_step": 140232, "epoch": 1575, "lr": 4.753248422890223e-05}
{"train_loss": 0.11402463167905807, "global_step": 140233, "epoch": 1575, "lr": 4.753190519885591e-05}
{"train_loss": 0.12594236433506012, "global_step": 140234, "epoch": 1575, "lr": 4.7531326169141415e-05}
{"train_loss": 0.12536659836769104, "global_step": 140235, "epoch": 1575, "lr": 4.7530747139758773e-05}
{"train_loss": 0.17137369513511658, "global_step": 140236, "epoch": 1575, "lr": 4.753016811070812e-05}
{"train_loss": 0.15282829105854034, "global_step": 140237, "epoch": 1575, "lr": 4.7529589081989486e-05}
{"train_loss": 0.20984117686748505, "global_step": 140238, "epoch": 1575, "lr": 4.752901005360298e-05}
{"train_loss": 0.0660860687494278, "global_step": 140239, "epoch": 1575, "lr": 4.752843102554868e-05}
{"train_loss": 0.10338953882455826, "global_step": 140240, "epoch": 1575, "lr": 4.752785199782662e-05}
{"train_loss": 0.1351911425590515, "global_step": 140241, "epoch": 1575, "lr": 4.7527272970436945e-05}
{"train_loss": 0.18411682546138763, "global_step": 140242, "epoch": 1575, "lr": 4.7526693943379666e-05}
{"train_loss": 0.05411122739315033, "global_step": 140243, "epoch": 1575, "lr": 4.752611491665492e-05}
{"train_loss": 0.13568411767482758, "global_step": 140244, "epoch": 1575, "lr": 4.752553589026273e-05}
{"train_loss": 0.0949241891503334, "global_step": 140245, "epoch": 1575, "lr": 4.752495686420323e-05}
{"train_loss": 0.14027321338653564, "global_step": 140246, "epoch": 1575, "lr": 4.752437783847645e-05}
{"train_loss": 0.09572935104370117, "global_step": 140247, "epoch": 1575, "lr": 4.75237988130825e-05}
{"train_loss": 0.12797144055366516, "global_step": 140248, "epoch": 1575, "lr": 4.752321978802143e-05}
{"train_loss": 0.19404786825180054, "global_step": 140249, "epoch": 1575, "lr": 4.752264076329335e-05}
{"train_loss": 0.10495775938034058, "global_step": 140250, "epoch": 1575, "lr": 4.75220617388983e-05}
{"train_loss": 0.0993906632065773, "global_step": 140251, "epoch": 1575, "lr": 4.75214827148364e-05}
{"train_loss": 0.13070610165596008, "global_step": 140252, "epoch": 1575, "lr": 4.752090369110769e-05}
{"train_loss": 0.1047845408320427, "global_step": 140253, "epoch": 1575, "lr": 4.7520324667712276e-05}
{"train_loss": 0.12705278396606445, "global_step": 140254, "epoch": 1575, "lr": 4.7519745644650215e-05}
{"train_loss": 0.11411982029676437, "global_step": 140255, "epoch": 1575, "lr": 4.7519166621921606e-05}
{"train_loss": 0.08663346618413925, "global_step": 140256, "epoch": 1575, "lr": 4.75185875995265e-05}
{"train_loss": 0.11227496713399887, "global_step": 140257, "epoch": 1575, "lr": 4.751800857746499e-05}
{"train_loss": 0.12181371450424194, "global_step": 140258, "epoch": 1575, "lr": 4.751742955573718e-05}
{"train_loss": 0.09996647387742996, "global_step": 140259, "epoch": 1575, "lr": 4.751685053434309e-05}
{"train_loss": 0.12211061269044876, "global_step": 140260, "epoch": 1575, "lr": 4.751627151328285e-05}
{"train_loss": 0.14202027022838593, "global_step": 140261, "epoch": 1575, "lr": 4.7515692492556504e-05}
{"train_loss": 0.14079846441745758, "global_step": 140262, "epoch": 1575, "lr": 4.7515113472164155e-05}
{"train_loss": 0.1124791648615612, "global_step": 140263, "epoch": 1575, "lr": 4.751453445210585e-05, "val_loss": 5.8544230461120605, "train_action_mse_error": 15.796369552612305}
{"train_loss": 0.1193951740860939, "global_step": 140264, "epoch": 1576, "lr": 4.7513955432381706e-05}
{"train_loss": 0.05247234180569649, "global_step": 140265, "epoch": 1576, "lr": 4.751337641299176e-05}
{"train_loss": 0.06933636218309402, "global_step": 140266, "epoch": 1576, "lr": 4.7512797393936134e-05}
{"train_loss": 0.052343711256980896, "global_step": 140267, "epoch": 1576, "lr": 4.751221837521486e-05}
{"train_loss": 0.09950809180736542, "global_step": 140268, "epoch": 1576, "lr": 4.751163935682805e-05}
{"train_loss": 0.13562075793743134, "global_step": 140269, "epoch": 1576, "lr": 4.751106033877576e-05}
{"train_loss": 0.08670030534267426, "global_step": 140270, "epoch": 1576, "lr": 4.7510481321058094e-05}
{"train_loss": 0.09248396009206772, "global_step": 140271, "epoch": 1576, "lr": 4.750990230367509e-05}
{"train_loss": 0.10896588861942291, "global_step": 140272, "epoch": 1576, "lr": 4.7509323286626875e-05}
{"train_loss": 0.1167992651462555, "global_step": 140273, "epoch": 1576, "lr": 4.750874426991348e-05}
{"train_loss": 0.1476890742778778, "global_step": 140274, "epoch": 1576, "lr": 4.750816525353501e-05}
{"train_loss": 0.14411284029483795, "global_step": 140275, "epoch": 1576, "lr": 4.750758623749154e-05}
{"train_loss": 0.14487354457378387, "global_step": 140276, "epoch": 1576, "lr": 4.750700722178315e-05}
{"train_loss": 0.11820042133331299, "global_step": 140277, "epoch": 1576, "lr": 4.7506428206409915e-05}
{"train_loss": 0.12755130231380463, "global_step": 140278, "epoch": 1576, "lr": 4.750584919137189e-05}
{"train_loss": 0.12011949717998505, "global_step": 140279, "epoch": 1576, "lr": 4.75052701766692e-05}
{"train_loss": 0.1512020379304886, "global_step": 140280, "epoch": 1576, "lr": 4.7504691162301875e-05}
{"train_loss": 0.12107211351394653, "global_step": 140281, "epoch": 1576, "lr": 4.750411214827003e-05}
{"train_loss": 0.09309770166873932, "global_step": 140282, "epoch": 1576, "lr": 4.75035331345737e-05}
{"train_loss": 0.12702950835227966, "global_step": 140283, "epoch": 1576, "lr": 4.750295412121302e-05}
{"train_loss": 0.05006593465805054, "global_step": 140284, "epoch": 1576, "lr": 4.7502375108188014e-05}
{"train_loss": 0.08429735898971558, "global_step": 140285, "epoch": 1576, "lr": 4.7501796095498806e-05}
{"train_loss": 0.11132193356752396, "global_step": 140286, "epoch": 1576, "lr": 4.750121708314543e-05}
{"train_loss": 0.16631558537483215, "global_step": 140287, "epoch": 1576, "lr": 4.7500638071128e-05}
{"train_loss": 0.14125338196754456, "global_step": 140288, "epoch": 1576, "lr": 4.750005905944656e-05}
{"train_loss": 0.10985095053911209, "global_step": 140289, "epoch": 1576, "lr": 4.749948004810123e-05}
{"train_loss": 0.11376149952411652, "global_step": 140290, "epoch": 1576, "lr": 4.749890103709204e-05}
{"train_loss": 0.09288053959608078, "global_step": 140291, "epoch": 1576, "lr": 4.749832202641912e-05}
{"train_loss": 0.05066315084695816, "global_step": 140292, "epoch": 1576, "lr": 4.7497743016082495e-05}
{"train_loss": 0.11859464645385742, "global_step": 140293, "epoch": 1576, "lr": 4.749716400608228e-05}
{"train_loss": 0.09560640156269073, "global_step": 140294, "epoch": 1576, "lr": 4.7496584996418536e-05}
{"train_loss": 0.12180734425783157, "global_step": 140295, "epoch": 1576, "lr": 4.749600598709135e-05}
{"train_loss": 0.0800950676202774, "global_step": 140296, "epoch": 1576, "lr": 4.74954269781008e-05}
{"train_loss": 0.14659984409809113, "global_step": 140297, "epoch": 1576, "lr": 4.749484796944695e-05}
{"train_loss": 0.12764199078083038, "global_step": 140298, "epoch": 1576, "lr": 4.74942689611299e-05}
{"train_loss": 0.06136362999677658, "global_step": 140299, "epoch": 1576, "lr": 4.7493689953149695e-05}
{"train_loss": 0.08472268283367157, "global_step": 140300, "epoch": 1576, "lr": 4.749311094550645e-05}
{"train_loss": 0.1086258590221405, "global_step": 140301, "epoch": 1576, "lr": 4.749253193820021e-05}
{"train_loss": 0.0801314190030098, "global_step": 140302, "epoch": 1576, "lr": 4.749195293123109e-05}
{"train_loss": 0.10677915066480637, "global_step": 140303, "epoch": 1576, "lr": 4.7491373924599126e-05}
{"train_loss": 0.13030588626861572, "global_step": 140304, "epoch": 1576, "lr": 4.7490794918304434e-05}
{"train_loss": 0.08770304918289185, "global_step": 140305, "epoch": 1576, "lr": 4.749021591234705e-05}
{"train_loss": 0.1587444543838501, "global_step": 140306, "epoch": 1576, "lr": 4.7489636906727105e-05}
{"train_loss": 0.0943983793258667, "global_step": 140307, "epoch": 1576, "lr": 4.7489057901444624e-05}
{"train_loss": 0.1083553284406662, "global_step": 140308, "epoch": 1576, "lr": 4.748847889649973e-05}
{"train_loss": 0.06128040328621864, "global_step": 140309, "epoch": 1576, "lr": 4.748789989189245e-05}
{"train_loss": 0.16029852628707886, "global_step": 140310, "epoch": 1576, "lr": 4.748732088762292e-05}
{"train_loss": 0.14908021688461304, "global_step": 140311, "epoch": 1576, "lr": 4.7486741883691164e-05}
{"train_loss": 0.046925634145736694, "global_step": 140312, "epoch": 1576, "lr": 4.74861628800973e-05}
{"train_loss": 0.12271814048290253, "global_step": 140313, "epoch": 1576, "lr": 4.748558387684139e-05}
{"train_loss": 0.10655305534601212, "global_step": 140314, "epoch": 1576, "lr": 4.7485004873923514e-05}
{"train_loss": 0.1708371490240097, "global_step": 140315, "epoch": 1576, "lr": 4.748442587134376e-05}
{"train_loss": 0.0600961297750473, "global_step": 140316, "epoch": 1576, "lr": 4.748384686910217e-05}
{"train_loss": 0.08797216415405273, "global_step": 140317, "epoch": 1576, "lr": 4.748326786719886e-05}
{"train_loss": 0.1427600234746933, "global_step": 140318, "epoch": 1576, "lr": 4.7482688865633885e-05}
{"train_loss": 0.18500763177871704, "global_step": 140319, "epoch": 1576, "lr": 4.748210986440735e-05}
{"train_loss": 0.08861728012561798, "global_step": 140320, "epoch": 1576, "lr": 4.748153086351928e-05}
{"train_loss": 0.11482354253530502, "global_step": 140321, "epoch": 1576, "lr": 4.7480951862969826e-05}
{"train_loss": 0.11098680645227432, "global_step": 140322, "epoch": 1576, "lr": 4.7480372862759e-05}
{"train_loss": 0.14181947708129883, "global_step": 140323, "epoch": 1576, "lr": 4.747979386288693e-05}
{"train_loss": 0.07649869471788406, "global_step": 140324, "epoch": 1576, "lr": 4.747921486335365e-05}
{"train_loss": 0.15856124460697174, "global_step": 140325, "epoch": 1576, "lr": 4.7478635864159275e-05}
{"train_loss": 0.1174044981598854, "global_step": 140326, "epoch": 1576, "lr": 4.747805686530385e-05}
{"train_loss": 0.16420488059520721, "global_step": 140327, "epoch": 1576, "lr": 4.747747786678749e-05}
{"train_loss": 0.12839250266551971, "global_step": 140328, "epoch": 1576, "lr": 4.747689886861023e-05}
{"train_loss": 0.09064755588769913, "global_step": 140329, "epoch": 1576, "lr": 4.747631987077218e-05}
{"train_loss": 0.10882821679115295, "global_step": 140330, "epoch": 1576, "lr": 4.747574087327341e-05}
{"train_loss": 0.05981462821364403, "global_step": 140331, "epoch": 1576, "lr": 4.7475161876113996e-05}
{"train_loss": 0.1360839605331421, "global_step": 140332, "epoch": 1576, "lr": 4.747458287929402e-05}
{"train_loss": 0.12180232256650925, "global_step": 140333, "epoch": 1576, "lr": 4.7474003882813545e-05}
{"train_loss": 0.19875070452690125, "global_step": 140334, "epoch": 1576, "lr": 4.7473424886672664e-05}
{"train_loss": 0.14057175815105438, "global_step": 140335, "epoch": 1576, "lr": 4.7472845890871456e-05}
{"train_loss": 0.13136638700962067, "global_step": 140336, "epoch": 1576, "lr": 4.747226689541e-05}
{"train_loss": 0.12461274862289429, "global_step": 140337, "epoch": 1576, "lr": 4.747168790028834e-05}
{"train_loss": 0.09533889591693878, "global_step": 140338, "epoch": 1576, "lr": 4.747110890550661e-05}
{"train_loss": 0.14690013229846954, "global_step": 140339, "epoch": 1576, "lr": 4.7470529911064835e-05}
{"train_loss": 0.11955927312374115, "global_step": 140340, "epoch": 1576, "lr": 4.7469950916963135e-05}
{"train_loss": 0.093027263879776, "global_step": 140341, "epoch": 1576, "lr": 4.746937192320154e-05}
{"train_loss": 0.09146548062562943, "global_step": 140342, "epoch": 1576, "lr": 4.746879292978019e-05}
{"train_loss": 0.2184888869524002, "global_step": 140343, "epoch": 1576, "lr": 4.746821393669911e-05}
{"train_loss": 0.05969633907079697, "global_step": 140344, "epoch": 1576, "lr": 4.746763494395842e-05}
{"train_loss": 0.07197925448417664, "global_step": 140345, "epoch": 1576, "lr": 4.746705595155815e-05}
{"train_loss": 0.08123066276311874, "global_step": 140346, "epoch": 1576, "lr": 4.746647695949842e-05}
{"train_loss": 0.07650226354598999, "global_step": 140347, "epoch": 1576, "lr": 4.746589796777928e-05}
{"train_loss": 0.13976818323135376, "global_step": 140348, "epoch": 1576, "lr": 4.746531897640081e-05}
{"train_loss": 0.12136846035718918, "global_step": 140349, "epoch": 1576, "lr": 4.7464739985363124e-05}
{"train_loss": 0.1319444328546524, "global_step": 140350, "epoch": 1576, "lr": 4.746416099466626e-05}
{"train_loss": 0.051715146750211716, "global_step": 140351, "epoch": 1576, "lr": 4.746358200431031e-05}
{"train_loss": 0.11239590872539564, "global_step": 140352, "epoch": 1576, "lr": 4.7463003014295346e-05, "val_loss": 5.911663055419922}
{"train_loss": 0.09491594135761261, "global_step": 140353, "epoch": 1577, "lr": 4.746242402462145e-05}
{"train_loss": 0.1851259171962738, "global_step": 140354, "epoch": 1577, "lr": 4.746184503528871e-05}
{"train_loss": 0.13646171987056732, "global_step": 140355, "epoch": 1577, "lr": 4.74612660462972e-05}
{"train_loss": 0.15797275304794312, "global_step": 140356, "epoch": 1577, "lr": 4.746068705764697e-05}
{"train_loss": 0.15897858142852783, "global_step": 140357, "epoch": 1577, "lr": 4.7460108069338136e-05}
{"train_loss": 0.10149436444044113, "global_step": 140358, "epoch": 1577, "lr": 4.745952908137074e-05}
{"train_loss": 0.16431450843811035, "global_step": 140359, "epoch": 1577, "lr": 4.745895009374491e-05}
{"train_loss": 0.08119277656078339, "global_step": 140360, "epoch": 1577, "lr": 4.745837110646066e-05}
{"train_loss": 0.0847051739692688, "global_step": 140361, "epoch": 1577, "lr": 4.745779211951813e-05}
{"train_loss": 0.16732484102249146, "global_step": 140362, "epoch": 1577, "lr": 4.7457213132917344e-05}
{"train_loss": 0.0970834344625473, "global_step": 140363, "epoch": 1577, "lr": 4.7456634146658424e-05}
{"train_loss": 0.10094482451677322, "global_step": 140364, "epoch": 1577, "lr": 4.7456055160741407e-05}
{"train_loss": 0.10046274214982986, "global_step": 140365, "epoch": 1577, "lr": 4.7455476175166394e-05}
{"train_loss": 0.1137794628739357, "global_step": 140366, "epoch": 1577, "lr": 4.745489718993349e-05}
{"train_loss": 0.18299974501132965, "global_step": 140367, "epoch": 1577, "lr": 4.7454318205042716e-05}
{"train_loss": 0.08752325177192688, "global_step": 140368, "epoch": 1577, "lr": 4.74537392204942e-05}
{"train_loss": 0.08886095881462097, "global_step": 140369, "epoch": 1577, "lr": 4.7453160236287985e-05}
{"train_loss": 0.14465822279453278, "global_step": 140370, "epoch": 1577, "lr": 4.745258125242417e-05}
{"train_loss": 0.16694897413253784, "global_step": 140371, "epoch": 1577, "lr": 4.745200226890281e-05}
{"train_loss": 0.10111699998378754, "global_step": 140372, "epoch": 1577, "lr": 4.745142328572401e-05}
{"train_loss": 0.16666725277900696, "global_step": 140373, "epoch": 1577, "lr": 4.7450844302887825e-05}
{"train_loss": 0.09371133893728256, "global_step": 140374, "epoch": 1577, "lr": 4.745026532039436e-05}
{"train_loss": 0.13726705312728882, "global_step": 140375, "epoch": 1577, "lr": 4.744968633824366e-05}
{"train_loss": 0.10269811749458313, "global_step": 140376, "epoch": 1577, "lr": 4.744910735643583e-05}
{"train_loss": 0.09670592844486237, "global_step": 140377, "epoch": 1577, "lr": 4.7448528374970916e-05}
{"train_loss": 0.11559092998504639, "global_step": 140378, "epoch": 1577, "lr": 4.744794939384904e-05}
{"train_loss": 0.15026162564754486, "global_step": 140379, "epoch": 1577, "lr": 4.7447370413070234e-05}
{"train_loss": 0.03614245727658272, "global_step": 140380, "epoch": 1577, "lr": 4.744679143263462e-05}
{"train_loss": 0.08626066893339157, "global_step": 140381, "epoch": 1577, "lr": 4.744621245254223e-05}
{"train_loss": 0.09559904038906097, "global_step": 140382, "epoch": 1577, "lr": 4.7445633472793194e-05}
{"train_loss": 0.11499694734811783, "global_step": 140383, "epoch": 1577, "lr": 4.744505449338753e-05}
{"train_loss": 0.09656000137329102, "global_step": 140384, "epoch": 1577, "lr": 4.744447551432535e-05}
{"train_loss": 0.09444879740476608, "global_step": 140385, "epoch": 1577, "lr": 4.744389653560676e-05}
{"train_loss": 0.09414102882146835, "global_step": 140386, "epoch": 1577, "lr": 4.744331755723178e-05}
{"train_loss": 0.10273934155702591, "global_step": 140387, "epoch": 1577, "lr": 4.744273857920053e-05}
{"train_loss": 0.14545921981334686, "global_step": 140388, "epoch": 1577, "lr": 4.744215960151307e-05}
{"train_loss": 0.09720040112733841, "global_step": 140389, "epoch": 1577, "lr": 4.7441580624169473e-05}
{"train_loss": 0.13377301394939423, "global_step": 140390, "epoch": 1577, "lr": 4.744100164716983e-05}
{"train_loss": 0.09207901358604431, "global_step": 140391, "epoch": 1577, "lr": 4.744042267051421e-05}
{"train_loss": 0.13873131573200226, "global_step": 140392, "epoch": 1577, "lr": 4.7439843694202695e-05}
{"train_loss": 0.11046720296144485, "global_step": 140393, "epoch": 1577, "lr": 4.7439264718235374e-05}
{"train_loss": 0.08224129676818848, "global_step": 140394, "epoch": 1577, "lr": 4.7438685742612294e-05}
{"train_loss": 0.08830126374959946, "global_step": 140395, "epoch": 1577, "lr": 4.743810676733357e-05}
{"train_loss": 0.12990806996822357, "global_step": 140396, "epoch": 1577, "lr": 4.7437527792399236e-05}
{"train_loss": 0.03723234683275223, "global_step": 140397, "epoch": 1577, "lr": 4.743694881780942e-05}
{"train_loss": 0.06061278283596039, "global_step": 140398, "epoch": 1577, "lr": 4.743636984356415e-05}
{"train_loss": 0.109360471367836, "global_step": 140399, "epoch": 1577, "lr": 4.743579086966356e-05}
{"train_loss": 0.13307122886180878, "global_step": 140400, "epoch": 1577, "lr": 4.743521189610767e-05}
{"train_loss": 0.15205782651901245, "global_step": 140401, "epoch": 1577, "lr": 4.7434632922896575e-05}
{"train_loss": 0.10256362706422806, "global_step": 140402, "epoch": 1577, "lr": 4.74340539500304e-05}
{"train_loss": 0.12531819939613342, "global_step": 140403, "epoch": 1577, "lr": 4.743347497750915e-05}
{"train_loss": 0.1553269475698471, "global_step": 140404, "epoch": 1577, "lr": 4.743289600533297e-05}
{"train_loss": 0.08237053453922272, "global_step": 140405, "epoch": 1577, "lr": 4.743231703350188e-05}
{"train_loss": 0.1253909170627594, "global_step": 140406, "epoch": 1577, "lr": 4.7431738062016004e-05}
{"train_loss": 0.06252211332321167, "global_step": 140407, "epoch": 1577, "lr": 4.743115909087539e-05}
{"train_loss": 0.09349178522825241, "global_step": 140408, "epoch": 1577, "lr": 4.743058012008012e-05}
{"train_loss": 0.11367757618427277, "global_step": 140409, "epoch": 1577, "lr": 4.743000114963029e-05}
{"train_loss": 0.11204315721988678, "global_step": 140410, "epoch": 1577, "lr": 4.7429422179525954e-05}
{"train_loss": 0.15417945384979248, "global_step": 140411, "epoch": 1577, "lr": 4.74288432097672e-05}
{"train_loss": 0.1939554512500763, "global_step": 140412, "epoch": 1577, "lr": 4.7428264240354126e-05}
{"train_loss": 0.09890249371528625, "global_step": 140413, "epoch": 1577, "lr": 4.742768527128676e-05}
{"train_loss": 0.182176873087883, "global_step": 140414, "epoch": 1577, "lr": 4.742710630256524e-05}
{"train_loss": 0.0850299522280693, "global_step": 140415, "epoch": 1577, "lr": 4.742652733418958e-05}
{"train_loss": 0.06186620518565178, "global_step": 140416, "epoch": 1577, "lr": 4.742594836615993e-05}
{"train_loss": 0.1494210660457611, "global_step": 140417, "epoch": 1577, "lr": 4.74253693984763e-05}
{"train_loss": 0.14849895238876343, "global_step": 140418, "epoch": 1577, "lr": 4.742479043113882e-05}
{"train_loss": 0.1037769541144371, "global_step": 140419, "epoch": 1577, "lr": 4.742421146414753e-05}
{"train_loss": 0.05385914817452431, "global_step": 140420, "epoch": 1577, "lr": 4.742363249750251e-05}
{"train_loss": 0.14245498180389404, "global_step": 140421, "epoch": 1577, "lr": 4.742305353120388e-05}
{"train_loss": 0.12809011340141296, "global_step": 140422, "epoch": 1577, "lr": 4.7422474565251667e-05}
{"train_loss": 0.1136772558093071, "global_step": 140423, "epoch": 1577, "lr": 4.7421895599646e-05}
{"train_loss": 0.12197154760360718, "global_step": 140424, "epoch": 1577, "lr": 4.7421316634386894e-05}
{"train_loss": 0.0712248831987381, "global_step": 140425, "epoch": 1577, "lr": 4.742073766947449e-05}
{"train_loss": 0.12467891722917557, "global_step": 140426, "epoch": 1577, "lr": 4.7420158704908824e-05}
{"train_loss": 0.10976650565862656, "global_step": 140427, "epoch": 1577, "lr": 4.7419579740689985e-05}
{"train_loss": 0.12275131791830063, "global_step": 140428, "epoch": 1577, "lr": 4.741900077681805e-05}
{"train_loss": 0.09827422350645065, "global_step": 140429, "epoch": 1577, "lr": 4.74184218132931e-05}
{"train_loss": 0.09884811192750931, "global_step": 140430, "epoch": 1577, "lr": 4.741784285011521e-05}
{"train_loss": 0.11149635910987854, "global_step": 140431, "epoch": 1577, "lr": 4.741726388728448e-05}
{"train_loss": 0.11594520509243011, "global_step": 140432, "epoch": 1577, "lr": 4.741668492480094e-05}
{"train_loss": 0.12741784751415253, "global_step": 140433, "epoch": 1577, "lr": 4.741610596266472e-05}
{"train_loss": 0.06370244175195694, "global_step": 140434, "epoch": 1577, "lr": 4.741552700087584e-05}
{"train_loss": 0.2028592824935913, "global_step": 140435, "epoch": 1577, "lr": 4.741494803943444e-05}
{"train_loss": 0.18602614104747772, "global_step": 140436, "epoch": 1577, "lr": 4.7414369078340546e-05}
{"train_loss": 0.12593533098697662, "global_step": 140437, "epoch": 1577, "lr": 4.741379011759428e-05}
{"train_loss": 0.14960311353206635, "global_step": 140438, "epoch": 1577, "lr": 4.741321115719568e-05}
{"train_loss": 0.11303359270095825, "global_step": 140439, "epoch": 1577, "lr": 4.741263219714484e-05}
{"train_loss": 0.16821852326393127, "global_step": 140440, "epoch": 1577, "lr": 4.741205323744186e-05}
{"train_loss": 0.11673511970746384, "global_step": 140441, "epoch": 1577, "lr": 4.741147427808678e-05, "val_loss": 5.958446025848389}
{"train_loss": 0.1322975903749466, "global_step": 140442, "epoch": 1578, "lr": 4.741089531907972e-05}
{"train_loss": 0.13878656923770905, "global_step": 140443, "epoch": 1578, "lr": 4.7410316360420704e-05}
{"train_loss": 0.08361218124628067, "global_step": 140444, "epoch": 1578, "lr": 4.7409737402109866e-05}
{"train_loss": 0.1313128024339676, "global_step": 140445, "epoch": 1578, "lr": 4.740915844414725e-05}
{"train_loss": 0.09834522008895874, "global_step": 140446, "epoch": 1578, "lr": 4.740857948653293e-05}
{"train_loss": 0.1143227145075798, "global_step": 140447, "epoch": 1578, "lr": 4.7408000529266996e-05}
{"train_loss": 0.09588145464658737, "global_step": 140448, "epoch": 1578, "lr": 4.7407421572349534e-05}
{"train_loss": 0.11759429425001144, "global_step": 140449, "epoch": 1578, "lr": 4.740684261578061e-05}
{"train_loss": 0.09135201573371887, "global_step": 140450, "epoch": 1578, "lr": 4.74062636595603e-05}
{"train_loss": 0.14154495298862457, "global_step": 140451, "epoch": 1578, "lr": 4.740568470368869e-05}
{"train_loss": 0.08636051416397095, "global_step": 140452, "epoch": 1578, "lr": 4.7405105748165864e-05}
{"train_loss": 0.1479460597038269, "global_step": 140453, "epoch": 1578, "lr": 4.740452679299187e-05}
{"train_loss": 0.06437129527330399, "global_step": 140454, "epoch": 1578, "lr": 4.740394783816682e-05}
{"train_loss": 0.07806396484375, "global_step": 140455, "epoch": 1578, "lr": 4.7403368883690765e-05}
{"train_loss": 0.09183201193809509, "global_step": 140456, "epoch": 1578, "lr": 4.740278992956379e-05}
{"train_loss": 0.11173227429389954, "global_step": 140457, "epoch": 1578, "lr": 4.7402210975786006e-05}
{"train_loss": 0.10665721446275711, "global_step": 140458, "epoch": 1578, "lr": 4.740163202235743e-05}
{"train_loss": 0.13460014760494232, "global_step": 140459, "epoch": 1578, "lr": 4.740105306927821e-05}
{"train_loss": 0.08196178078651428, "global_step": 140460, "epoch": 1578, "lr": 4.740047411654835e-05}
{"train_loss": 0.10218188911676407, "global_step": 140461, "epoch": 1578, "lr": 4.7399895164167995e-05}
{"train_loss": 0.09066331386566162, "global_step": 140462, "epoch": 1578, "lr": 4.739931621213717e-05}
{"train_loss": 0.12207720428705215, "global_step": 140463, "epoch": 1578, "lr": 4.7398737260455996e-05}
{"train_loss": 0.13758225739002228, "global_step": 140464, "epoch": 1578, "lr": 4.7398158309124506e-05}
{"train_loss": 0.22722631692886353, "global_step": 140465, "epoch": 1578, "lr": 4.739757935814282e-05}
{"train_loss": 0.07447467744350433, "global_step": 140466, "epoch": 1578, "lr": 4.7397000407511e-05}
{"train_loss": 0.10996465384960175, "global_step": 140467, "epoch": 1578, "lr": 4.739642145722911e-05}
{"train_loss": 0.08964220434427261, "global_step": 140468, "epoch": 1578, "lr": 4.739584250729725e-05}
{"train_loss": 0.08955474197864532, "global_step": 140469, "epoch": 1578, "lr": 4.739526355771548e-05}
{"train_loss": 0.07687842845916748, "global_step": 140470, "epoch": 1578, "lr": 4.739468460848388e-05}
{"train_loss": 0.15333762764930725, "global_step": 140471, "epoch": 1578, "lr": 4.739410565960255e-05}
{"train_loss": 0.08133523166179657, "global_step": 140472, "epoch": 1578, "lr": 4.7393526711071526e-05}
{"train_loss": 0.11057578772306442, "global_step": 140473, "epoch": 1578, "lr": 4.739294776289094e-05}
{"train_loss": 0.06253794580698013, "global_step": 140474, "epoch": 1578, "lr": 4.739236881506081e-05}
{"train_loss": 0.08907899260520935, "global_step": 140475, "epoch": 1578, "lr": 4.739178986758125e-05}
{"train_loss": 0.15701277554035187, "global_step": 140476, "epoch": 1578, "lr": 4.739121092045235e-05}
{"train_loss": 0.08358000963926315, "global_step": 140477, "epoch": 1578, "lr": 4.739063197367415e-05}
{"train_loss": 0.11709269881248474, "global_step": 140478, "epoch": 1578, "lr": 4.739005302724677e-05}
{"train_loss": 0.12655840814113617, "global_step": 140479, "epoch": 1578, "lr": 4.7389474081170246e-05}
{"train_loss": 0.09993894398212433, "global_step": 140480, "epoch": 1578, "lr": 4.738889513544469e-05}
{"train_loss": 0.16596822440624237, "global_step": 140481, "epoch": 1578, "lr": 4.738831619007015e-05}
{"train_loss": 0.15668021142482758, "global_step": 140482, "epoch": 1578, "lr": 4.7387737245046735e-05}
{"train_loss": 0.10516010969877243, "global_step": 140483, "epoch": 1578, "lr": 4.738715830037449e-05}
{"train_loss": 0.08030420541763306, "global_step": 140484, "epoch": 1578, "lr": 4.7386579356053525e-05}
{"train_loss": 0.15118902921676636, "global_step": 140485, "epoch": 1578, "lr": 4.73860004120839e-05}
{"train_loss": 0.11082815378904343, "global_step": 140486, "epoch": 1578, "lr": 4.738542146846569e-05}
{"train_loss": 0.10987155139446259, "global_step": 140487, "epoch": 1578, "lr": 4.738484252519898e-05}
{"train_loss": 0.09641485661268234, "global_step": 140488, "epoch": 1578, "lr": 4.7384263582283847e-05}
{"train_loss": 0.08057896047830582, "global_step": 140489, "epoch": 1578, "lr": 4.738368463972037e-05}
{"train_loss": 0.0904427021741867, "global_step": 140490, "epoch": 1578, "lr": 4.738310569750863e-05}
{"train_loss": 0.07360781729221344, "global_step": 140491, "epoch": 1578, "lr": 4.7382526755648684e-05}
{"train_loss": 0.14039264619350433, "global_step": 140492, "epoch": 1578, "lr": 4.738194781414062e-05}
{"train_loss": 0.11056119948625565, "global_step": 140493, "epoch": 1578, "lr": 4.7381368872984545e-05}
{"train_loss": 0.08978059142827988, "global_step": 140494, "epoch": 1578, "lr": 4.738078993218049e-05}
{"train_loss": 0.1068677008152008, "global_step": 140495, "epoch": 1578, "lr": 4.738021099172858e-05}
{"train_loss": 0.09324559569358826, "global_step": 140496, "epoch": 1578, "lr": 4.7379632051628844e-05}
{"train_loss": 0.07616730034351349, "global_step": 140497, "epoch": 1578, "lr": 4.737905311188141e-05}
{"train_loss": 0.1354091912508011, "global_step": 140498, "epoch": 1578, "lr": 4.73784741724863e-05}
{"train_loss": 0.17329145967960358, "global_step": 140499, "epoch": 1578, "lr": 4.7377895233443655e-05}
{"train_loss": 0.16071408987045288, "global_step": 140500, "epoch": 1578, "lr": 4.7377316294753495e-05}
{"train_loss": 0.1204233467578888, "global_step": 140501, "epoch": 1578, "lr": 4.7376737356415944e-05}
{"train_loss": 0.06968020647764206, "global_step": 140502, "epoch": 1578, "lr": 4.737615841843103e-05}
{"train_loss": 0.09742005169391632, "global_step": 140503, "epoch": 1578, "lr": 4.7375579480798896e-05}
{"train_loss": 0.11960127204656601, "global_step": 140504, "epoch": 1578, "lr": 4.737500054351956e-05}
{"train_loss": 0.12309952080249786, "global_step": 140505, "epoch": 1578, "lr": 4.737442160659313e-05}
{"train_loss": 0.09124310314655304, "global_step": 140506, "epoch": 1578, "lr": 4.7373842670019675e-05}
{"train_loss": 0.14793148636817932, "global_step": 140507, "epoch": 1578, "lr": 4.7373263733799274e-05}
{"train_loss": 0.1634943187236786, "global_step": 140508, "epoch": 1578, "lr": 4.7372684797932e-05}
{"train_loss": 0.15458272397518158, "global_step": 140509, "epoch": 1578, "lr": 4.737210586241796e-05}
{"train_loss": 0.11929307132959366, "global_step": 140510, "epoch": 1578, "lr": 4.7371526927257176e-05}
{"train_loss": 0.08748134225606918, "global_step": 140511, "epoch": 1578, "lr": 4.737094799244976e-05}
{"train_loss": 0.10563234984874725, "global_step": 140512, "epoch": 1578, "lr": 4.737036905799581e-05}
{"train_loss": 0.11176392436027527, "global_step": 140513, "epoch": 1578, "lr": 4.736979012389536e-05}
{"train_loss": 0.08125422149896622, "global_step": 140514, "epoch": 1578, "lr": 4.736921119014854e-05}
{"train_loss": 0.09051636606454849, "global_step": 140515, "epoch": 1578, "lr": 4.736863225675537e-05}
{"train_loss": 0.10479805618524551, "global_step": 140516, "epoch": 1578, "lr": 4.736805332371597e-05}
{"train_loss": 0.06835197657346725, "global_step": 140517, "epoch": 1578, "lr": 4.7367474391030394e-05}
{"train_loss": 0.13728760182857513, "global_step": 140518, "epoch": 1578, "lr": 4.736689545869874e-05}
{"train_loss": 0.10768464207649231, "global_step": 140519, "epoch": 1578, "lr": 4.736631652672105e-05}
{"train_loss": 0.08004534989595413, "global_step": 140520, "epoch": 1578, "lr": 4.7365737595097465e-05}
{"train_loss": 0.1403496116399765, "global_step": 140521, "epoch": 1578, "lr": 4.736515866382799e-05}
{"train_loss": 0.13070541620254517, "global_step": 140522, "epoch": 1578, "lr": 4.736457973291276e-05}
{"train_loss": 0.18835972249507904, "global_step": 140523, "epoch": 1578, "lr": 4.7364000802351826e-05}
{"train_loss": 0.04599350690841675, "global_step": 140524, "epoch": 1578, "lr": 4.7363421872145264e-05}
{"train_loss": 0.1227412149310112, "global_step": 140525, "epoch": 1578, "lr": 4.736284294229316e-05}
{"train_loss": 0.083421990275383, "global_step": 140526, "epoch": 1578, "lr": 4.7362264012795585e-05}
{"train_loss": 0.11853209882974625, "global_step": 140527, "epoch": 1578, "lr": 4.736168508365263e-05}
{"train_loss": 0.07786707580089569, "global_step": 140528, "epoch": 1578, "lr": 4.736110615486435e-05}
{"train_loss": 0.0442555733025074, "global_step": 140529, "epoch": 1578, "lr": 4.7360527226430855e-05}
{"train_loss": 0.11043692233689716, "global_step": 140530, "epoch": 1578, "lr": 4.735994829835219e-05, "val_loss": 6.061934471130371}
{"train_loss": 0.1660294532775879, "global_step": 140531, "epoch": 1579, "lr": 4.735936937062846e-05}
{"train_loss": 0.12466439604759216, "global_step": 140532, "epoch": 1579, "lr": 4.735879044325972e-05}
{"train_loss": 0.15313708782196045, "global_step": 140533, "epoch": 1579, "lr": 4.735821151624607e-05}
{"train_loss": 0.12277257442474365, "global_step": 140534, "epoch": 1579, "lr": 4.735763258958756e-05}
{"train_loss": 0.13556000590324402, "global_step": 140535, "epoch": 1579, "lr": 4.735705366328431e-05}
{"train_loss": 0.1722290813922882, "global_step": 140536, "epoch": 1579, "lr": 4.735647473733634e-05}
{"train_loss": 0.11192534863948822, "global_step": 140537, "epoch": 1579, "lr": 4.735589581174379e-05}
{"train_loss": 0.15657995641231537, "global_step": 140538, "epoch": 1579, "lr": 4.7355316886506676e-05}
{"train_loss": 0.0642523467540741, "global_step": 140539, "epoch": 1579, "lr": 4.735473796162514e-05}
{"train_loss": 0.12496520578861237, "global_step": 140540, "epoch": 1579, "lr": 4.73541590370992e-05}
{"train_loss": 0.1693175733089447, "global_step": 140541, "epoch": 1579, "lr": 4.7353580112928986e-05}
{"train_loss": 0.08820158988237381, "global_step": 140542, "epoch": 1579, "lr": 4.735300118911454e-05}
{"train_loss": 0.16266950964927673, "global_step": 140543, "epoch": 1579, "lr": 4.7352422265655944e-05}
{"train_loss": 0.10001114010810852, "global_step": 140544, "epoch": 1579, "lr": 4.7351843342553284e-05}
{"train_loss": 0.13204877078533173, "global_step": 140545, "epoch": 1579, "lr": 4.735126441980664e-05}
{"train_loss": 0.06996511667966843, "global_step": 140546, "epoch": 1579, "lr": 4.7350685497416086e-05}
{"train_loss": 0.11435196548700333, "global_step": 140547, "epoch": 1579, "lr": 4.735010657538169e-05}
{"train_loss": 0.08472658693790436, "global_step": 140548, "epoch": 1579, "lr": 4.7349527653703564e-05}
{"train_loss": 0.15958070755004883, "global_step": 140549, "epoch": 1579, "lr": 4.734894873238174e-05}
{"train_loss": 0.11412402242422104, "global_step": 140550, "epoch": 1579, "lr": 4.7348369811416334e-05}
{"train_loss": 0.09231240302324295, "global_step": 140551, "epoch": 1579, "lr": 4.734779089080738e-05}
{"train_loss": 0.10174430906772614, "global_step": 140552, "epoch": 1579, "lr": 4.734721197055502e-05}
{"train_loss": 0.10490851104259491, "global_step": 140553, "epoch": 1579, "lr": 4.734663305065926e-05}
{"train_loss": 0.08378278464078903, "global_step": 140554, "epoch": 1579, "lr": 4.7346054131120244e-05}
{"train_loss": 0.232064351439476, "global_step": 140555, "epoch": 1579, "lr": 4.734547521193799e-05}
{"train_loss": 0.2179609090089798, "global_step": 140556, "epoch": 1579, "lr": 4.734489629311263e-05}
{"train_loss": 0.1274632066488266, "global_step": 140557, "epoch": 1579, "lr": 4.734431737464419e-05}
{"train_loss": 0.11655844002962112, "global_step": 140558, "epoch": 1579, "lr": 4.7343738456532804e-05}
{"train_loss": 0.14080248773097992, "global_step": 140559, "epoch": 1579, "lr": 4.734315953877849e-05}
{"train_loss": 0.11409670114517212, "global_step": 140560, "epoch": 1579, "lr": 4.734258062138138e-05}
{"train_loss": 0.08379281312227249, "global_step": 140561, "epoch": 1579, "lr": 4.7342001704341524e-05}
{"train_loss": 0.16054217517375946, "global_step": 140562, "epoch": 1579, "lr": 4.7341422787658994e-05}
{"train_loss": 0.08828606456518173, "global_step": 140563, "epoch": 1579, "lr": 4.7340843871333886e-05}
{"train_loss": 0.13016100227832794, "global_step": 140564, "epoch": 1579, "lr": 4.7340264955366265e-05}
{"train_loss": 0.136070117354393, "global_step": 140565, "epoch": 1579, "lr": 4.733968603975621e-05}
{"train_loss": 0.17794981598854065, "global_step": 140566, "epoch": 1579, "lr": 4.7339107124503795e-05}
{"train_loss": 0.13653364777565002, "global_step": 140567, "epoch": 1579, "lr": 4.733852820960912e-05}
{"train_loss": 0.1402336210012436, "global_step": 140568, "epoch": 1579, "lr": 4.733794929507223e-05}
{"train_loss": 0.2230238914489746, "global_step": 140569, "epoch": 1579, "lr": 4.733737038089325e-05}
{"train_loss": 0.15574313700199127, "global_step": 140570, "epoch": 1579, "lr": 4.7336791467072194e-05}
{"train_loss": 0.07833432406187057, "global_step": 140571, "epoch": 1579, "lr": 4.7336212553609196e-05}
{"train_loss": 0.09881292283535004, "global_step": 140572, "epoch": 1579, "lr": 4.7335633640504304e-05}
{"train_loss": 0.09324654191732407, "global_step": 140573, "epoch": 1579, "lr": 4.733505472775761e-05}
{"train_loss": 0.09780923277139664, "global_step": 140574, "epoch": 1579, "lr": 4.7334475815369166e-05}
{"train_loss": 0.09352371841669083, "global_step": 140575, "epoch": 1579, "lr": 4.7333896903339096e-05}
{"train_loss": 0.14284713566303253, "global_step": 140576, "epoch": 1579, "lr": 4.733331799166743e-05}
{"train_loss": 0.06730545312166214, "global_step": 140577, "epoch": 1579, "lr": 4.733273908035429e-05}
{"train_loss": 0.1216629222035408, "global_step": 140578, "epoch": 1579, "lr": 4.73321601693997e-05}
{"train_loss": 0.16081970930099487, "global_step": 140579, "epoch": 1579, "lr": 4.7331581258803796e-05}
{"train_loss": 0.11562783271074295, "global_step": 140580, "epoch": 1579, "lr": 4.7331002348566604e-05}
{"train_loss": 0.11865943670272827, "global_step": 140581, "epoch": 1579, "lr": 4.7330423438688256e-05}
{"train_loss": 0.1055118665099144, "global_step": 140582, "epoch": 1579, "lr": 4.732984452916878e-05}
{"train_loss": 0.15717054903507233, "global_step": 140583, "epoch": 1579, "lr": 4.732926562000827e-05}
{"train_loss": 0.08850843459367752, "global_step": 140584, "epoch": 1579, "lr": 4.732868671120682e-05}
{"train_loss": 0.10273538529872894, "global_step": 140585, "epoch": 1579, "lr": 4.732810780276448e-05}
{"train_loss": 0.14977754652500153, "global_step": 140586, "epoch": 1579, "lr": 4.732752889468137e-05}
{"train_loss": 0.12181004881858826, "global_step": 140587, "epoch": 1579, "lr": 4.732694998695751e-05}
{"train_loss": 0.12913642823696136, "global_step": 140588, "epoch": 1579, "lr": 4.732637107959303e-05}
{"train_loss": 0.09195861965417862, "global_step": 140589, "epoch": 1579, "lr": 4.7325792172587976e-05}
{"train_loss": 0.14683890342712402, "global_step": 140590, "epoch": 1579, "lr": 4.732521326594245e-05}
{"train_loss": 0.1470184028148651, "global_step": 140591, "epoch": 1579, "lr": 4.7324634359656495e-05}
{"train_loss": 0.11795265227556229, "global_step": 140592, "epoch": 1579, "lr": 4.732405545373024e-05}
{"train_loss": 0.09324834495782852, "global_step": 140593, "epoch": 1579, "lr": 4.7323476548163704e-05}
{"train_loss": 0.0700480118393898, "global_step": 140594, "epoch": 1579, "lr": 4.7322897642957014e-05}
{"train_loss": 0.11980506032705307, "global_step": 140595, "epoch": 1579, "lr": 4.732231873811021e-05}
{"train_loss": 0.0859411284327507, "global_step": 140596, "epoch": 1579, "lr": 4.732173983362341e-05}
{"train_loss": 0.13784904778003693, "global_step": 140597, "epoch": 1579, "lr": 4.732116092949665e-05}
{"train_loss": 0.13116216659545898, "global_step": 140598, "epoch": 1579, "lr": 4.7320582025730035e-05}
{"train_loss": 0.11508127301931381, "global_step": 140599, "epoch": 1579, "lr": 4.732000312232362e-05}
{"train_loss": 0.13466079533100128, "global_step": 140600, "epoch": 1579, "lr": 4.731942421927753e-05}
{"train_loss": 0.1034424901008606, "global_step": 140601, "epoch": 1579, "lr": 4.731884531659179e-05}
{"train_loss": 0.09248381108045578, "global_step": 140602, "epoch": 1579, "lr": 4.731826641426649e-05}
{"train_loss": 0.11613349616527557, "global_step": 140603, "epoch": 1579, "lr": 4.731768751230174e-05}
{"train_loss": 0.2167837917804718, "global_step": 140604, "epoch": 1579, "lr": 4.731710861069757e-05}
{"train_loss": 0.10232985764741898, "global_step": 140605, "epoch": 1579, "lr": 4.731652970945411e-05}
{"train_loss": 0.12876681983470917, "global_step": 140606, "epoch": 1579, "lr": 4.731595080857138e-05}
{"train_loss": 0.10040254890918732, "global_step": 140607, "epoch": 1579, "lr": 4.731537190804951e-05}
{"train_loss": 0.14477549493312836, "global_step": 140608, "epoch": 1579, "lr": 4.7314793007888544e-05}
{"train_loss": 0.13095718622207642, "global_step": 140609, "epoch": 1579, "lr": 4.731421410808859e-05}
{"train_loss": 0.08767250925302505, "global_step": 140610, "epoch": 1579, "lr": 4.7313635208649676e-05}
{"train_loss": 0.09531192481517792, "global_step": 140611, "epoch": 1579, "lr": 4.7313056309571934e-05}
{"train_loss": 0.14767582714557648, "global_step": 140612, "epoch": 1579, "lr": 4.7312477410855406e-05}
{"train_loss": 0.1019243597984314, "global_step": 140613, "epoch": 1579, "lr": 4.7311898512500196e-05}
{"train_loss": 0.11747635900974274, "global_step": 140614, "epoch": 1579, "lr": 4.731131961450635e-05}
{"train_loss": 0.11493169516324997, "global_step": 140615, "epoch": 1579, "lr": 4.7310740716873985e-05}
{"train_loss": 0.15330253541469574, "global_step": 140616, "epoch": 1579, "lr": 4.731016181960314e-05}
{"train_loss": 0.11514950543642044, "global_step": 140617, "epoch": 1579, "lr": 4.730958292269393e-05}
{"train_loss": 0.0669456422328949, "global_step": 140618, "epoch": 1579, "lr": 4.730900402614639e-05}
{"train_loss": 0.12343877068396365, "global_step": 140619, "epoch": 1579, "lr": 4.730842512996063e-05, "val_loss": 6.014057159423828}
{"train_loss": 0.05826655775308609, "global_step": 140620, "epoch": 1580, "lr": 4.730784623413674e-05}
{"train_loss": 0.136736199259758, "global_step": 140621, "epoch": 1580, "lr": 4.730726733867475e-05}
{"train_loss": 0.12212818115949631, "global_step": 140622, "epoch": 1580, "lr": 4.730668844357479e-05}
{"train_loss": 0.15377739071846008, "global_step": 140623, "epoch": 1580, "lr": 4.7306109548836884e-05}
{"train_loss": 0.09534195065498352, "global_step": 140624, "epoch": 1580, "lr": 4.7305530654461164e-05}
{"train_loss": 0.05827093869447708, "global_step": 140625, "epoch": 1580, "lr": 4.730495176044766e-05}
{"train_loss": 0.11517952382564545, "global_step": 140626, "epoch": 1580, "lr": 4.73043728667965e-05}
{"train_loss": 0.09965325146913528, "global_step": 140627, "epoch": 1580, "lr": 4.730379397350771e-05}
{"train_loss": 0.09816864132881165, "global_step": 140628, "epoch": 1580, "lr": 4.730321508058141e-05}
{"train_loss": 0.11703556776046753, "global_step": 140629, "epoch": 1580, "lr": 4.730263618801764e-05}
{"train_loss": 0.1313241571187973, "global_step": 140630, "epoch": 1580, "lr": 4.7302057295816524e-05}
{"train_loss": 0.08650641143321991, "global_step": 140631, "epoch": 1580, "lr": 4.730147840397808e-05}
{"train_loss": 0.0542851984500885, "global_step": 140632, "epoch": 1580, "lr": 4.730089951250245e-05}
{"train_loss": 0.11301326006650925, "global_step": 140633, "epoch": 1580, "lr": 4.730032062138966e-05}
{"train_loss": 0.10316132754087448, "global_step": 140634, "epoch": 1580, "lr": 4.729974173063983e-05}
{"train_loss": 0.1449793428182602, "global_step": 140635, "epoch": 1580, "lr": 4.729916284025299e-05}
{"train_loss": 0.10540500283241272, "global_step": 140636, "epoch": 1580, "lr": 4.729858395022928e-05}
{"train_loss": 0.13704125583171844, "global_step": 140637, "epoch": 1580, "lr": 4.7298005060568704e-05}
{"train_loss": 0.06941821426153183, "global_step": 140638, "epoch": 1580, "lr": 4.72974261712714e-05}
{"train_loss": 0.08449037373065948, "global_step": 140639, "epoch": 1580, "lr": 4.7296847282337426e-05}
{"train_loss": 0.07314478605985641, "global_step": 140640, "epoch": 1580, "lr": 4.729626839376684e-05}
{"train_loss": 0.10631226003170013, "global_step": 140641, "epoch": 1580, "lr": 4.729568950555976e-05}
{"train_loss": 0.0940745547413826, "global_step": 140642, "epoch": 1580, "lr": 4.729511061771622e-05}
{"train_loss": 0.1626577377319336, "global_step": 140643, "epoch": 1580, "lr": 4.7294531730236344e-05}
{"train_loss": 0.11195649206638336, "global_step": 140644, "epoch": 1580, "lr": 4.729395284312016e-05}
{"train_loss": 0.14680331945419312, "global_step": 140645, "epoch": 1580, "lr": 4.7293373956367794e-05}
{"train_loss": 0.06797464936971664, "global_step": 140646, "epoch": 1580, "lr": 4.7292795069979276e-05}
{"train_loss": 0.14074179530143738, "global_step": 140647, "epoch": 1580, "lr": 4.729221618395474e-05}
{"train_loss": 0.1297377198934555, "global_step": 140648, "epoch": 1580, "lr": 4.72916372982942e-05}
{"train_loss": 0.09541041404008865, "global_step": 140649, "epoch": 1580, "lr": 4.72910584129978e-05}
{"train_loss": 0.1415652334690094, "global_step": 140650, "epoch": 1580, "lr": 4.7290479528065554e-05}
{"train_loss": 0.1337641328573227, "global_step": 140651, "epoch": 1580, "lr": 4.728990064349759e-05}
{"train_loss": 0.10908982902765274, "global_step": 140652, "epoch": 1580, "lr": 4.7289321759293946e-05}
{"train_loss": 0.1417023241519928, "global_step": 140653, "epoch": 1580, "lr": 4.728874287545474e-05}
{"train_loss": 0.12146519869565964, "global_step": 140654, "epoch": 1580, "lr": 4.728816399198001e-05}
{"train_loss": 0.10857650637626648, "global_step": 140655, "epoch": 1580, "lr": 4.7287585108869866e-05}
{"train_loss": 0.15695177018642426, "global_step": 140656, "epoch": 1580, "lr": 4.728700622612437e-05}
{"train_loss": 0.09940725564956665, "global_step": 140657, "epoch": 1580, "lr": 4.7286427343743604e-05}
{"train_loss": 0.12005136907100677, "global_step": 140658, "epoch": 1580, "lr": 4.728584846172765e-05}
{"train_loss": 0.12161203473806381, "global_step": 140659, "epoch": 1580, "lr": 4.728526958007657e-05}
{"train_loss": 0.15088458359241486, "global_step": 140660, "epoch": 1580, "lr": 4.728469069879047e-05}
{"train_loss": 0.08696338534355164, "global_step": 140661, "epoch": 1580, "lr": 4.7284111817869386e-05}
{"train_loss": 0.15536373853683472, "global_step": 140662, "epoch": 1580, "lr": 4.728353293731344e-05}
{"train_loss": 0.12192118167877197, "global_step": 140663, "epoch": 1580, "lr": 4.728295405712268e-05}
{"train_loss": 0.12875670194625854, "global_step": 140664, "epoch": 1580, "lr": 4.72823751772972e-05}
{"train_loss": 0.08977212756872177, "global_step": 140665, "epoch": 1580, "lr": 4.728179629783705e-05}
{"train_loss": 0.10531483590602875, "global_step": 140666, "epoch": 1580, "lr": 4.728121741874236e-05}
{"train_loss": 0.09671339392662048, "global_step": 140667, "epoch": 1580, "lr": 4.728063854001316e-05}
{"train_loss": 0.1423666775226593, "global_step": 140668, "epoch": 1580, "lr": 4.728005966164956e-05}
{"train_loss": 0.11206776648759842, "global_step": 140669, "epoch": 1580, "lr": 4.72794807836516e-05}
{"train_loss": 0.08931273221969604, "global_step": 140670, "epoch": 1580, "lr": 4.72789019060194e-05}
{"train_loss": 0.11952557414770126, "global_step": 140671, "epoch": 1580, "lr": 4.7278323028753e-05}
{"train_loss": 0.10619121789932251, "global_step": 140672, "epoch": 1580, "lr": 4.727774415185252e-05}
{"train_loss": 0.11786545068025589, "global_step": 140673, "epoch": 1580, "lr": 4.727716527531799e-05}
{"train_loss": 0.0688604936003685, "global_step": 140674, "epoch": 1580, "lr": 4.727658639914953e-05}
{"train_loss": 0.15555760264396667, "global_step": 140675, "epoch": 1580, "lr": 4.7276007523347186e-05}
{"train_loss": 0.19289612770080566, "global_step": 140676, "epoch": 1580, "lr": 4.7275428647911054e-05}
{"train_loss": 0.11652787029743195, "global_step": 140677, "epoch": 1580, "lr": 4.7274849772841214e-05}
{"train_loss": 0.05869339779019356, "global_step": 140678, "epoch": 1580, "lr": 4.727427089813773e-05}
{"train_loss": 0.10639715194702148, "global_step": 140679, "epoch": 1580, "lr": 4.7273692023800703e-05}
{"train_loss": 0.05557909980416298, "global_step": 140680, "epoch": 1580, "lr": 4.7273113149830176e-05}
{"train_loss": 0.08923130482435226, "global_step": 140681, "epoch": 1580, "lr": 4.727253427622625e-05}
{"train_loss": 0.14292339980602264, "global_step": 140682, "epoch": 1580, "lr": 4.727195540298899e-05}
{"train_loss": 0.10385285317897797, "global_step": 140683, "epoch": 1580, "lr": 4.727137653011851e-05}
{"train_loss": 0.08240258693695068, "global_step": 140684, "epoch": 1580, "lr": 4.727079765761483e-05}
{"train_loss": 0.046981025487184525, "global_step": 140685, "epoch": 1580, "lr": 4.727021878547808e-05}
{"train_loss": 0.1399308443069458, "global_step": 140686, "epoch": 1580, "lr": 4.7269639913708297e-05}
{"train_loss": 0.13568925857543945, "global_step": 140687, "epoch": 1580, "lr": 4.72690610423056e-05}
{"train_loss": 0.11013039946556091, "global_step": 140688, "epoch": 1580, "lr": 4.726848217127002e-05}
{"train_loss": 0.13673727214336395, "global_step": 140689, "epoch": 1580, "lr": 4.726790330060168e-05}
{"train_loss": 0.1168525218963623, "global_step": 140690, "epoch": 1580, "lr": 4.726732443030062e-05}
{"train_loss": 0.17317070066928864, "global_step": 140691, "epoch": 1580, "lr": 4.7266745560366954e-05}
{"train_loss": 0.08459366112947464, "global_step": 140692, "epoch": 1580, "lr": 4.726616669080072e-05}
{"train_loss": 0.1075989231467247, "global_step": 140693, "epoch": 1580, "lr": 4.7265587821602024e-05}
{"train_loss": 0.0809805765748024, "global_step": 140694, "epoch": 1580, "lr": 4.726500895277094e-05}
{"train_loss": 0.16805589199066162, "global_step": 140695, "epoch": 1580, "lr": 4.7264430084307544e-05}
{"train_loss": 0.10025663673877716, "global_step": 140696, "epoch": 1580, "lr": 4.726385121621191e-05}
{"train_loss": 0.09556670486927032, "global_step": 140697, "epoch": 1580, "lr": 4.726327234848411e-05}
{"train_loss": 0.08165847510099411, "global_step": 140698, "epoch": 1580, "lr": 4.7262693481124246e-05}
{"train_loss": 0.15569384396076202, "global_step": 140699, "epoch": 1580, "lr": 4.726211461413236e-05}
{"train_loss": 0.13133911788463593, "global_step": 140700, "epoch": 1580, "lr": 4.726153574750857e-05}
{"train_loss": 0.12635478377342224, "global_step": 140701, "epoch": 1580, "lr": 4.7260956881252916e-05}
{"train_loss": 0.12075386196374893, "global_step": 140702, "epoch": 1580, "lr": 4.726037801536551e-05}
{"train_loss": 0.07948239147663116, "global_step": 140703, "epoch": 1580, "lr": 4.725979914984639e-05}
{"train_loss": 0.10444716364145279, "global_step": 140704, "epoch": 1580, "lr": 4.725922028469568e-05}
{"train_loss": 0.09226896613836288, "global_step": 140705, "epoch": 1580, "lr": 4.725864141991341e-05}
{"train_loss": 0.12129373848438263, "global_step": 140706, "epoch": 1580, "lr": 4.725806255549971e-05}
{"train_loss": 0.12660199403762817, "global_step": 140707, "epoch": 1580, "lr": 4.7257483691454604e-05}
{"train_loss": 0.11307260297824827, "global_step": 140708, "epoch": 1580, "lr": 4.725690482777822e-05, "val_loss": 5.989227294921875, "train_action_mse_error": 11.15035629272461}
{"train_loss": 0.1409519910812378, "global_step": 140709, "epoch": 1581, "lr": 4.725632596447058e-05}
{"train_loss": 0.08084962517023087, "global_step": 140710, "epoch": 1581, "lr": 4.7255747101531804e-05}
{"train_loss": 0.08808092027902603, "global_step": 140711, "epoch": 1581, "lr": 4.725516823896198e-05}
{"train_loss": 0.09329890459775925, "global_step": 140712, "epoch": 1581, "lr": 4.725458937676115e-05}
{"train_loss": 0.09285901486873627, "global_step": 140713, "epoch": 1581, "lr": 4.7254010514929414e-05}
{"train_loss": 0.10407988727092743, "global_step": 140714, "epoch": 1581, "lr": 4.725343165346684e-05}
{"train_loss": 0.1188352033495903, "global_step": 140715, "epoch": 1581, "lr": 4.725285279237351e-05}
{"train_loss": 0.13358163833618164, "global_step": 140716, "epoch": 1581, "lr": 4.725227393164949e-05}
{"train_loss": 0.1313607394695282, "global_step": 140717, "epoch": 1581, "lr": 4.725169507129489e-05}
{"train_loss": 0.06768399477005005, "global_step": 140718, "epoch": 1581, "lr": 4.7251116211309746e-05}
{"train_loss": 0.11590185016393661, "global_step": 140719, "epoch": 1581, "lr": 4.725053735169418e-05}
{"train_loss": 0.13407249748706818, "global_step": 140720, "epoch": 1581, "lr": 4.7249958492448206e-05}
{"train_loss": 0.08596359938383102, "global_step": 140721, "epoch": 1581, "lr": 4.7249379633571976e-05}
{"train_loss": 0.09418069571256638, "global_step": 140722, "epoch": 1581, "lr": 4.724880077506551e-05}
{"train_loss": 0.13203385472297668, "global_step": 140723, "epoch": 1581, "lr": 4.724822191692893e-05}
{"train_loss": 0.12761083245277405, "global_step": 140724, "epoch": 1581, "lr": 4.724764305916228e-05}
{"train_loss": 0.10930988192558289, "global_step": 140725, "epoch": 1581, "lr": 4.724706420176566e-05}
{"train_loss": 0.12127500027418137, "global_step": 140726, "epoch": 1581, "lr": 4.7246485344739124e-05}
{"train_loss": 0.10663129389286041, "global_step": 140727, "epoch": 1581, "lr": 4.7245906488082784e-05}
{"train_loss": 0.0694645345211029, "global_step": 140728, "epoch": 1581, "lr": 4.724532763179668e-05}
{"train_loss": 0.10708653926849365, "global_step": 140729, "epoch": 1581, "lr": 4.7244748775880895e-05}
{"train_loss": 0.13717176020145416, "global_step": 140730, "epoch": 1581, "lr": 4.724416992033556e-05}
{"train_loss": 0.13577449321746826, "global_step": 140731, "epoch": 1581, "lr": 4.724359106516069e-05}
{"train_loss": 0.1412203460931778, "global_step": 140732, "epoch": 1581, "lr": 4.724301221035639e-05}
{"train_loss": 0.10188335925340652, "global_step": 140733, "epoch": 1581, "lr": 4.724243335592273e-05}
{"train_loss": 0.07790102809667587, "global_step": 140734, "epoch": 1581, "lr": 4.72418545018598e-05}
{"train_loss": 0.13223862648010254, "global_step": 140735, "epoch": 1581, "lr": 4.724127564816766e-05}
{"train_loss": 0.0843399167060852, "global_step": 140736, "epoch": 1581, "lr": 4.7240696794846406e-05}
{"train_loss": 0.08303751796483994, "global_step": 140737, "epoch": 1581, "lr": 4.724011794189609e-05}
{"train_loss": 0.0926240086555481, "global_step": 140738, "epoch": 1581, "lr": 4.723953908931683e-05}
{"train_loss": 0.08938132226467133, "global_step": 140739, "epoch": 1581, "lr": 4.7238960237108656e-05}
{"train_loss": 0.1310407519340515, "global_step": 140740, "epoch": 1581, "lr": 4.723838138527169e-05}
{"train_loss": 0.11616968363523483, "global_step": 140741, "epoch": 1581, "lr": 4.723780253380597e-05}
{"train_loss": 0.10419655591249466, "global_step": 140742, "epoch": 1581, "lr": 4.723722368271162e-05}
{"train_loss": 0.09739645570516586, "global_step": 140743, "epoch": 1581, "lr": 4.723664483198866e-05}
{"train_loss": 0.08002793043851852, "global_step": 140744, "epoch": 1581, "lr": 4.723606598163723e-05}
{"train_loss": 0.10753456503152847, "global_step": 140745, "epoch": 1581, "lr": 4.723548713165735e-05}
{"train_loss": 0.13914038240909576, "global_step": 140746, "epoch": 1581, "lr": 4.723490828204913e-05}
{"train_loss": 0.11404526978731155, "global_step": 140747, "epoch": 1581, "lr": 4.723432943281266e-05}
{"train_loss": 0.07685142010450363, "global_step": 140748, "epoch": 1581, "lr": 4.723375058394798e-05}
{"train_loss": 0.10815558582544327, "global_step": 140749, "epoch": 1581, "lr": 4.7233171735455216e-05}
{"train_loss": 0.12348763644695282, "global_step": 140750, "epoch": 1581, "lr": 4.7232592887334406e-05}
{"train_loss": 0.09983282536268234, "global_step": 140751, "epoch": 1581, "lr": 4.7232014039585634e-05}
{"train_loss": 0.11025018244981766, "global_step": 140752, "epoch": 1581, "lr": 4.7231435192208986e-05}
{"train_loss": 0.10059604048728943, "global_step": 140753, "epoch": 1581, "lr": 4.723085634520454e-05}
{"train_loss": 0.12335135787725449, "global_step": 140754, "epoch": 1581, "lr": 4.7230277498572366e-05}
{"train_loss": 0.09716697037220001, "global_step": 140755, "epoch": 1581, "lr": 4.722969865231256e-05}
{"train_loss": 0.17804814875125885, "global_step": 140756, "epoch": 1581, "lr": 4.722911980642517e-05}
{"train_loss": 0.09708188474178314, "global_step": 140757, "epoch": 1581, "lr": 4.722854096091032e-05}
{"train_loss": 0.2019108235836029, "global_step": 140758, "epoch": 1581, "lr": 4.722796211576802e-05}
{"train_loss": 0.2077254056930542, "global_step": 140759, "epoch": 1581, "lr": 4.722738327099842e-05}
{"train_loss": 0.126564159989357, "global_step": 140760, "epoch": 1581, "lr": 4.722680442660154e-05}
{"train_loss": 0.14937365055084229, "global_step": 140761, "epoch": 1581, "lr": 4.72262255825775e-05}
{"train_loss": 0.10145147889852524, "global_step": 140762, "epoch": 1581, "lr": 4.7225646738926345e-05}
{"train_loss": 0.14501547813415527, "global_step": 140763, "epoch": 1581, "lr": 4.7225067895648186e-05}
{"train_loss": 0.07663346827030182, "global_step": 140764, "epoch": 1581, "lr": 4.722448905274306e-05}
{"train_loss": 0.09217605739831924, "global_step": 140765, "epoch": 1581, "lr": 4.722391021021107e-05}
{"train_loss": 0.1087438091635704, "global_step": 140766, "epoch": 1581, "lr": 4.7223331368052306e-05}
{"train_loss": 0.10145683586597443, "global_step": 140767, "epoch": 1581, "lr": 4.722275252626682e-05}
{"train_loss": 0.15699855983257294, "global_step": 140768, "epoch": 1581, "lr": 4.722217368485472e-05}
{"train_loss": 0.11177004873752594, "global_step": 140769, "epoch": 1581, "lr": 4.722159484381605e-05}
{"train_loss": 0.11607193946838379, "global_step": 140770, "epoch": 1581, "lr": 4.7221016003150906e-05}
{"train_loss": 0.09090153127908707, "global_step": 140771, "epoch": 1581, "lr": 4.7220437162859364e-05}
{"train_loss": 0.07431478798389435, "global_step": 140772, "epoch": 1581, "lr": 4.72198583229415e-05}
{"train_loss": 0.1010606437921524, "global_step": 140773, "epoch": 1581, "lr": 4.7219279483397386e-05}
{"train_loss": 0.1474882960319519, "global_step": 140774, "epoch": 1581, "lr": 4.7218700644227126e-05}
{"train_loss": 0.13605304062366486, "global_step": 140775, "epoch": 1581, "lr": 4.7218121805430746e-05}
{"train_loss": 0.1412782520055771, "global_step": 140776, "epoch": 1581, "lr": 4.721754296700839e-05}
{"train_loss": 0.06384517252445221, "global_step": 140777, "epoch": 1581, "lr": 4.721696412896007e-05}
{"train_loss": 0.17430947721004486, "global_step": 140778, "epoch": 1581, "lr": 4.721638529128592e-05}
{"train_loss": 0.08526332676410675, "global_step": 140779, "epoch": 1581, "lr": 4.721580645398597e-05}
{"train_loss": 0.10514135658740997, "global_step": 140780, "epoch": 1581, "lr": 4.721522761706034e-05}
{"train_loss": 0.05230499058961868, "global_step": 140781, "epoch": 1581, "lr": 4.721464878050907e-05}
{"train_loss": 0.1421368420124054, "global_step": 140782, "epoch": 1581, "lr": 4.721406994433226e-05}
{"train_loss": 0.15774083137512207, "global_step": 140783, "epoch": 1581, "lr": 4.721349110853e-05}
{"train_loss": 0.11614672094583511, "global_step": 140784, "epoch": 1581, "lr": 4.721291227310233e-05}
{"train_loss": 0.12424816936254501, "global_step": 140785, "epoch": 1581, "lr": 4.721233343804938e-05}
{"train_loss": 0.05558871105313301, "global_step": 140786, "epoch": 1581, "lr": 4.7211754603371176e-05}
{"train_loss": 0.09385279566049576, "global_step": 140787, "epoch": 1581, "lr": 4.721117576906783e-05}
{"train_loss": 0.10674098134040833, "global_step": 140788, "epoch": 1581, "lr": 4.72105969351394e-05}
{"train_loss": 0.07869940251111984, "global_step": 140789, "epoch": 1581, "lr": 4.7210018101585975e-05}
{"train_loss": 0.09557986259460449, "global_step": 140790, "epoch": 1581, "lr": 4.7209439268407627e-05}
{"train_loss": 0.12077391147613525, "global_step": 140791, "epoch": 1581, "lr": 4.720886043560443e-05}
{"train_loss": 0.039868950843811035, "global_step": 140792, "epoch": 1581, "lr": 4.720828160317647e-05}
{"train_loss": 0.1389273703098297, "global_step": 140793, "epoch": 1581, "lr": 4.720770277112383e-05}
{"train_loss": 0.11012978851795197, "global_step": 140794, "epoch": 1581, "lr": 4.720712393944658e-05}
{"train_loss": 0.17688128352165222, "global_step": 140795, "epoch": 1581, "lr": 4.72065451081448e-05}
{"train_loss": 0.13627606630325317, "global_step": 140796, "epoch": 1581, "lr": 4.720596627721855e-05}
{"train_loss": 0.11233005978250771, "global_step": 140797, "epoch": 1581, "lr": 4.720538744666795e-05, "val_loss": 5.919121742248535}
{"train_loss": 0.13307897746562958, "global_step": 140798, "epoch": 1582, "lr": 4.7204808616493016e-05}
{"train_loss": 0.1444326788187027, "global_step": 140799, "epoch": 1582, "lr": 4.72042297866939e-05}
{"train_loss": 0.07221530377864838, "global_step": 140800, "epoch": 1582, "lr": 4.7203650957270606e-05}
{"train_loss": 0.12019944190979004, "global_step": 140801, "epoch": 1582, "lr": 4.720307212822325e-05}
{"train_loss": 0.12458357214927673, "global_step": 140802, "epoch": 1582, "lr": 4.720249329955193e-05}
{"train_loss": 0.15205742418766022, "global_step": 140803, "epoch": 1582, "lr": 4.720191447125668e-05}
{"train_loss": 0.07999308407306671, "global_step": 140804, "epoch": 1582, "lr": 4.7201335643337624e-05}
{"train_loss": 0.11927208304405212, "global_step": 140805, "epoch": 1582, "lr": 4.720075681579479e-05}
{"train_loss": 0.06550772488117218, "global_step": 140806, "epoch": 1582, "lr": 4.7200177988628294e-05}
{"train_loss": 0.07823386043310165, "global_step": 140807, "epoch": 1582, "lr": 4.7199599161838186e-05}
{"train_loss": 0.16058140993118286, "global_step": 140808, "epoch": 1582, "lr": 4.719902033542458e-05}
{"train_loss": 0.1390296071767807, "global_step": 140809, "epoch": 1582, "lr": 4.719844150938752e-05}
{"train_loss": 0.10032933950424194, "global_step": 140810, "epoch": 1582, "lr": 4.719786268372709e-05}
{"train_loss": 0.10465079545974731, "global_step": 140811, "epoch": 1582, "lr": 4.719728385844338e-05}
{"train_loss": 0.12271903455257416, "global_step": 140812, "epoch": 1582, "lr": 4.719670503353646e-05}
{"train_loss": 0.07722978293895721, "global_step": 140813, "epoch": 1582, "lr": 4.719612620900641e-05}
{"train_loss": 0.10891346633434296, "global_step": 140814, "epoch": 1582, "lr": 4.719554738485331e-05}
{"train_loss": 0.11277244240045547, "global_step": 140815, "epoch": 1582, "lr": 4.7194968561077217e-05}
{"train_loss": 0.10919264703989029, "global_step": 140816, "epoch": 1582, "lr": 4.719438973767825e-05}
{"train_loss": 0.15592068433761597, "global_step": 140817, "epoch": 1582, "lr": 4.719381091465644e-05}
{"train_loss": 0.08835246413946152, "global_step": 140818, "epoch": 1582, "lr": 4.719323209201191e-05}
{"train_loss": 0.1524793803691864, "global_step": 140819, "epoch": 1582, "lr": 4.7192653269744694e-05}
{"train_loss": 0.060043517500162125, "global_step": 140820, "epoch": 1582, "lr": 4.719207444785489e-05}
{"train_loss": 0.0997239276766777, "global_step": 140821, "epoch": 1582, "lr": 4.7191495626342596e-05}
{"train_loss": 0.08452855050563812, "global_step": 140822, "epoch": 1582, "lr": 4.7190916805207856e-05}
{"train_loss": 0.10379151999950409, "global_step": 140823, "epoch": 1582, "lr": 4.7190337984450784e-05}
{"train_loss": 0.07692567259073257, "global_step": 140824, "epoch": 1582, "lr": 4.7189759164071415e-05}
{"train_loss": 0.13798053562641144, "global_step": 140825, "epoch": 1582, "lr": 4.718918034406987e-05}
{"train_loss": 0.12782111763954163, "global_step": 140826, "epoch": 1582, "lr": 4.718860152444618e-05}
{"train_loss": 0.10115097463130951, "global_step": 140827, "epoch": 1582, "lr": 4.718802270520047e-05}
{"train_loss": 0.14629177749156952, "global_step": 140828, "epoch": 1582, "lr": 4.718744388633279e-05}
{"train_loss": 0.07913282513618469, "global_step": 140829, "epoch": 1582, "lr": 4.7186865067843224e-05}
{"train_loss": 0.1007719337940216, "global_step": 140830, "epoch": 1582, "lr": 4.718628624973184e-05}
{"train_loss": 0.104779452085495, "global_step": 140831, "epoch": 1582, "lr": 4.718570743199874e-05}
{"train_loss": 0.08527170121669769, "global_step": 140832, "epoch": 1582, "lr": 4.7185128614643976e-05}
{"train_loss": 0.12036269158124924, "global_step": 140833, "epoch": 1582, "lr": 4.718454979766765e-05}
{"train_loss": 0.08605672419071198, "global_step": 140834, "epoch": 1582, "lr": 4.718397098106981e-05}
{"train_loss": 0.18101032078266144, "global_step": 140835, "epoch": 1582, "lr": 4.718339216485057e-05}
{"train_loss": 0.08580205589532852, "global_step": 140836, "epoch": 1582, "lr": 4.718281334900997e-05}
{"train_loss": 0.15974225103855133, "global_step": 140837, "epoch": 1582, "lr": 4.7182234533548097e-05}
{"train_loss": 0.12445715814828873, "global_step": 140838, "epoch": 1582, "lr": 4.718165571846507e-05}
{"train_loss": 0.10324928909540176, "global_step": 140839, "epoch": 1582, "lr": 4.718107690376091e-05}
{"train_loss": 0.13029618561267853, "global_step": 140840, "epoch": 1582, "lr": 4.718049808943574e-05}
{"train_loss": 0.12872739136219025, "global_step": 140841, "epoch": 1582, "lr": 4.717991927548959e-05}
{"train_loss": 0.136739581823349, "global_step": 140842, "epoch": 1582, "lr": 4.71793404619226e-05}
{"train_loss": 0.12946118414402008, "global_step": 140843, "epoch": 1582, "lr": 4.7178761648734785e-05}
{"train_loss": 0.1449941247701645, "global_step": 140844, "epoch": 1582, "lr": 4.717818283592627e-05}
{"train_loss": 0.16353170573711395, "global_step": 140845, "epoch": 1582, "lr": 4.71776040234971e-05}
{"train_loss": 0.13266383111476898, "global_step": 140846, "epoch": 1582, "lr": 4.717702521144738e-05}
{"train_loss": 0.11017134040594101, "global_step": 140847, "epoch": 1582, "lr": 4.717644639977717e-05}
{"train_loss": 0.1044057086110115, "global_step": 140848, "epoch": 1582, "lr": 4.717586758848655e-05}
{"train_loss": 0.17041943967342377, "global_step": 140849, "epoch": 1582, "lr": 4.7175288777575604e-05}
{"train_loss": 0.10936356335878372, "global_step": 140850, "epoch": 1582, "lr": 4.71747099670444e-05}
{"train_loss": 0.0500725582242012, "global_step": 140851, "epoch": 1582, "lr": 4.7174131156893024e-05}
{"train_loss": 0.14674128592014313, "global_step": 140852, "epoch": 1582, "lr": 4.717355234712156e-05}
{"train_loss": 0.1105089783668518, "global_step": 140853, "epoch": 1582, "lr": 4.717297353773006e-05}
{"train_loss": 0.04765749350190163, "global_step": 140854, "epoch": 1582, "lr": 4.7172394728718646e-05}
{"train_loss": 0.0790998637676239, "global_step": 140855, "epoch": 1582, "lr": 4.717181592008734e-05}
{"train_loss": 0.08732424676418304, "global_step": 140856, "epoch": 1582, "lr": 4.717123711183625e-05}
{"train_loss": 0.06756344437599182, "global_step": 140857, "epoch": 1582, "lr": 4.717065830396547e-05}
{"train_loss": 0.1257920116186142, "global_step": 140858, "epoch": 1582, "lr": 4.717007949647505e-05}
{"train_loss": 0.10951751470565796, "global_step": 140859, "epoch": 1582, "lr": 4.7169500689365085e-05}
{"train_loss": 0.09021873772144318, "global_step": 140860, "epoch": 1582, "lr": 4.716892188263563e-05}
{"train_loss": 0.09243835508823395, "global_step": 140861, "epoch": 1582, "lr": 4.71683430762868e-05}
{"train_loss": 0.13434387743473053, "global_step": 140862, "epoch": 1582, "lr": 4.716776427031864e-05}
{"train_loss": 0.1202007308602333, "global_step": 140863, "epoch": 1582, "lr": 4.716718546473125e-05}
{"train_loss": 0.10425975173711777, "global_step": 140864, "epoch": 1582, "lr": 4.716660665952468e-05}
{"train_loss": 0.08540266752243042, "global_step": 140865, "epoch": 1582, "lr": 4.716602785469905e-05}
{"train_loss": 0.08764706552028656, "global_step": 140866, "epoch": 1582, "lr": 4.71654490502544e-05}
{"train_loss": 0.07837264239788055, "global_step": 140867, "epoch": 1582, "lr": 4.7164870246190815e-05}
{"train_loss": 0.08547823876142502, "global_step": 140868, "epoch": 1582, "lr": 4.716429144250838e-05}
{"train_loss": 0.1542796492576599, "global_step": 140869, "epoch": 1582, "lr": 4.716371263920718e-05}
{"train_loss": 0.14992640912532806, "global_step": 140870, "epoch": 1582, "lr": 4.7163133836287283e-05}
{"train_loss": 0.12945902347564697, "global_step": 140871, "epoch": 1582, "lr": 4.716255503374877e-05}
{"train_loss": 0.0812147781252861, "global_step": 140872, "epoch": 1582, "lr": 4.7161976231591704e-05}
{"train_loss": 0.14092308282852173, "global_step": 140873, "epoch": 1582, "lr": 4.716139742981617e-05}
{"train_loss": 0.11259018629789352, "global_step": 140874, "epoch": 1582, "lr": 4.7160818628422275e-05}
{"train_loss": 0.1304692029953003, "global_step": 140875, "epoch": 1582, "lr": 4.716023982741005e-05}
{"train_loss": 0.09464094787836075, "global_step": 140876, "epoch": 1582, "lr": 4.715966102677962e-05}
{"train_loss": 0.11995844542980194, "global_step": 140877, "epoch": 1582, "lr": 4.715908222653102e-05}
{"train_loss": 0.17232348024845123, "global_step": 140878, "epoch": 1582, "lr": 4.715850342666437e-05}
{"train_loss": 0.1542098969221115, "global_step": 140879, "epoch": 1582, "lr": 4.7157924627179695e-05}
{"train_loss": 0.09346968680620193, "global_step": 140880, "epoch": 1582, "lr": 4.715734582807713e-05}
{"train_loss": 0.08001986145973206, "global_step": 140881, "epoch": 1582, "lr": 4.71567670293567e-05}
{"train_loss": 0.11704592406749725, "global_step": 140882, "epoch": 1582, "lr": 4.715618823101853e-05}
{"train_loss": 0.10139588266611099, "global_step": 140883, "epoch": 1582, "lr": 4.7155609433062654e-05}
{"train_loss": 0.12081338465213776, "global_step": 140884, "epoch": 1582, "lr": 4.715503063548919e-05}
{"train_loss": 0.07604046165943146, "global_step": 140885, "epoch": 1582, "lr": 4.715445183829819e-05}
{"train_loss": 0.11240457818749246, "global_step": 140886, "epoch": 1582, "lr": 4.715387304148974e-05, "val_loss": 6.0044965744018555}
{"train_loss": 0.12577563524246216, "global_step": 140887, "epoch": 1583, "lr": 4.715329424506392e-05}
{"train_loss": 0.08319330215454102, "global_step": 140888, "epoch": 1583, "lr": 4.71527154490208e-05}
{"train_loss": 0.0782446637749672, "global_step": 140889, "epoch": 1583, "lr": 4.715213665336047e-05}
{"train_loss": 0.1521550565958023, "global_step": 140890, "epoch": 1583, "lr": 4.715155785808301e-05}
{"train_loss": 0.11205704510211945, "global_step": 140891, "epoch": 1583, "lr": 4.7150979063188453e-05}
{"train_loss": 0.060568880289793015, "global_step": 140892, "epoch": 1583, "lr": 4.715040026867693e-05}
{"train_loss": 0.09143541008234024, "global_step": 140893, "epoch": 1583, "lr": 4.714982147454851e-05}
{"train_loss": 0.13210497796535492, "global_step": 140894, "epoch": 1583, "lr": 4.714924268080325e-05}
{"train_loss": 0.08193767070770264, "global_step": 140895, "epoch": 1583, "lr": 4.7148663887441254e-05}
{"train_loss": 0.08915892988443375, "global_step": 140896, "epoch": 1583, "lr": 4.714808509446256e-05}
{"train_loss": 0.13635125756263733, "global_step": 140897, "epoch": 1583, "lr": 4.71475063018673e-05}
{"train_loss": 0.15805773437023163, "global_step": 140898, "epoch": 1583, "lr": 4.71469275096555e-05}
{"train_loss": 0.1404370814561844, "global_step": 140899, "epoch": 1583, "lr": 4.7146348717827284e-05}
{"train_loss": 0.10307136178016663, "global_step": 140900, "epoch": 1583, "lr": 4.714576992638268e-05}
{"train_loss": 0.08143315464258194, "global_step": 140901, "epoch": 1583, "lr": 4.714519113532182e-05}
{"train_loss": 0.16043031215667725, "global_step": 140902, "epoch": 1583, "lr": 4.714461234464473e-05}
{"train_loss": 0.10778288543224335, "global_step": 140903, "epoch": 1583, "lr": 4.714403355435153e-05}
{"train_loss": 0.08267314732074738, "global_step": 140904, "epoch": 1583, "lr": 4.7143454764442276e-05}
{"train_loss": 0.0706757977604866, "global_step": 140905, "epoch": 1583, "lr": 4.7142875974917044e-05}
{"train_loss": 0.1124505028128624, "global_step": 140906, "epoch": 1583, "lr": 4.714229718577592e-05}
{"train_loss": 0.07318016141653061, "global_step": 140907, "epoch": 1583, "lr": 4.714171839701898e-05}
{"train_loss": 0.11389701813459396, "global_step": 140908, "epoch": 1583, "lr": 4.71411396086463e-05}
{"train_loss": 0.17588944733142853, "global_step": 140909, "epoch": 1583, "lr": 4.7140560820657945e-05}
{"train_loss": 0.08506348729133606, "global_step": 140910, "epoch": 1583, "lr": 4.7139982033054034e-05}
{"train_loss": 0.1271427571773529, "global_step": 140911, "epoch": 1583, "lr": 4.7139403245834595e-05}
{"train_loss": 0.10065615177154541, "global_step": 140912, "epoch": 1583, "lr": 4.713882445899975e-05}
{"train_loss": 0.0821804404258728, "global_step": 140913, "epoch": 1583, "lr": 4.713824567254953e-05}
{"train_loss": 0.10056187957525253, "global_step": 140914, "epoch": 1583, "lr": 4.7137666886484064e-05}
{"train_loss": 0.11290764808654785, "global_step": 140915, "epoch": 1583, "lr": 4.713708810080338e-05}
{"train_loss": 0.13360756635665894, "global_step": 140916, "epoch": 1583, "lr": 4.71365093155076e-05}
{"train_loss": 0.1435166895389557, "global_step": 140917, "epoch": 1583, "lr": 4.713593053059676e-05}
{"train_loss": 0.07503998279571533, "global_step": 140918, "epoch": 1583, "lr": 4.7135351746070986e-05}
{"train_loss": 0.1088501587510109, "global_step": 140919, "epoch": 1583, "lr": 4.71347729619303e-05}
{"train_loss": 0.07424824684858322, "global_step": 140920, "epoch": 1583, "lr": 4.7134194178174834e-05}
{"train_loss": 0.11416478455066681, "global_step": 140921, "epoch": 1583, "lr": 4.713361539480462e-05}
{"train_loss": 0.13592210412025452, "global_step": 140922, "epoch": 1583, "lr": 4.7133036611819775e-05}
{"train_loss": 0.19371561706066132, "global_step": 140923, "epoch": 1583, "lr": 4.713245782922034e-05}
{"train_loss": 0.15549075603485107, "global_step": 140924, "epoch": 1583, "lr": 4.713187904700643e-05}
{"train_loss": 0.12402766197919846, "global_step": 140925, "epoch": 1583, "lr": 4.71313002651781e-05}
{"train_loss": 0.2342151403427124, "global_step": 140926, "epoch": 1583, "lr": 4.7130721483735424e-05}
{"train_loss": 0.08010159432888031, "global_step": 140927, "epoch": 1583, "lr": 4.713014270267849e-05}
{"train_loss": 0.07639914005994797, "global_step": 140928, "epoch": 1583, "lr": 4.712956392200736e-05}
{"train_loss": 0.19204096496105194, "global_step": 140929, "epoch": 1583, "lr": 4.7128985141722145e-05}
{"train_loss": 0.1234099268913269, "global_step": 140930, "epoch": 1583, "lr": 4.712840636182289e-05}
{"train_loss": 0.1377861201763153, "global_step": 140931, "epoch": 1583, "lr": 4.71278275823097e-05}
{"train_loss": 0.0990670770406723, "global_step": 140932, "epoch": 1583, "lr": 4.7127248803182624e-05}
{"train_loss": 0.14108479022979736, "global_step": 140933, "epoch": 1583, "lr": 4.712667002444177e-05}
{"train_loss": 0.12034157663583755, "global_step": 140934, "epoch": 1583, "lr": 4.712609124608718e-05}
{"train_loss": 0.07073185592889786, "global_step": 140935, "epoch": 1583, "lr": 4.712551246811897e-05}
{"train_loss": 0.11485779285430908, "global_step": 140936, "epoch": 1583, "lr": 4.712493369053718e-05}
{"train_loss": 0.10078510642051697, "global_step": 140937, "epoch": 1583, "lr": 4.712435491334193e-05}
{"train_loss": 0.08972673118114471, "global_step": 140938, "epoch": 1583, "lr": 4.712377613653326e-05}
{"train_loss": 0.13076815009117126, "global_step": 140939, "epoch": 1583, "lr": 4.712319736011128e-05}
{"train_loss": 0.11851153522729874, "global_step": 140940, "epoch": 1583, "lr": 4.7122618584076025e-05}
{"train_loss": 0.10360806435346603, "global_step": 140941, "epoch": 1583, "lr": 4.712203980842763e-05}
{"train_loss": 0.059341464191675186, "global_step": 140942, "epoch": 1583, "lr": 4.7121461033166117e-05}
{"train_loss": 0.06407469511032104, "global_step": 140943, "epoch": 1583, "lr": 4.712088225829161e-05}
{"train_loss": 0.13789555430412292, "global_step": 140944, "epoch": 1583, "lr": 4.712030348380415e-05}
{"train_loss": 0.10675540566444397, "global_step": 140945, "epoch": 1583, "lr": 4.7119724709703836e-05}
{"train_loss": 0.08697759360074997, "global_step": 140946, "epoch": 1583, "lr": 4.711914593599073e-05}
{"train_loss": 0.11284437030553818, "global_step": 140947, "epoch": 1583, "lr": 4.7118567162664915e-05}
{"train_loss": 0.08824682235717773, "global_step": 140948, "epoch": 1583, "lr": 4.71179883897265e-05}
{"train_loss": 0.12248221039772034, "global_step": 140949, "epoch": 1583, "lr": 4.711740961717551e-05}
{"train_loss": 0.06306800246238708, "global_step": 140950, "epoch": 1583, "lr": 4.711683084501207e-05}
{"train_loss": 0.14049032330513, "global_step": 140951, "epoch": 1583, "lr": 4.7116252073236214e-05}
{"train_loss": 0.1373376101255417, "global_step": 140952, "epoch": 1583, "lr": 4.7115673301848064e-05}
{"train_loss": 0.14695006608963013, "global_step": 140953, "epoch": 1583, "lr": 4.711509453084766e-05}
{"train_loss": 0.0863458588719368, "global_step": 140954, "epoch": 1583, "lr": 4.7114515760235115e-05}
{"train_loss": 0.12554232776165009, "global_step": 140955, "epoch": 1583, "lr": 4.711393699001047e-05}
{"train_loss": 0.14070989191532135, "global_step": 140956, "epoch": 1583, "lr": 4.711335822017384e-05}
{"train_loss": 0.1624297797679901, "global_step": 140957, "epoch": 1583, "lr": 4.711277945072526e-05}
{"train_loss": 0.1248827800154686, "global_step": 140958, "epoch": 1583, "lr": 4.711220068166486e-05}
{"train_loss": 0.09454569965600967, "global_step": 140959, "epoch": 1583, "lr": 4.711162191299266e-05}
{"train_loss": 0.17897632718086243, "global_step": 140960, "epoch": 1583, "lr": 4.71110431447088e-05}
{"train_loss": 0.12656879425048828, "global_step": 140961, "epoch": 1583, "lr": 4.711046437681328e-05}
{"train_loss": 0.14819513261318207, "global_step": 140962, "epoch": 1583, "lr": 4.7109885609306266e-05}
{"train_loss": 0.14587712287902832, "global_step": 140963, "epoch": 1583, "lr": 4.710930684218778e-05}
{"train_loss": 0.11014585942029953, "global_step": 140964, "epoch": 1583, "lr": 4.710872807545789e-05}
{"train_loss": 0.07047057151794434, "global_step": 140965, "epoch": 1583, "lr": 4.710814930911673e-05}
{"train_loss": 0.11087700724601746, "global_step": 140966, "epoch": 1583, "lr": 4.7107570543164317e-05}
{"train_loss": 0.15146683156490326, "global_step": 140967, "epoch": 1583, "lr": 4.710699177760078e-05}
{"train_loss": 0.1292964071035385, "global_step": 140968, "epoch": 1583, "lr": 4.7106413012426145e-05}
{"train_loss": 0.16735005378723145, "global_step": 140969, "epoch": 1583, "lr": 4.710583424764054e-05}
{"train_loss": 0.12734736502170563, "global_step": 140970, "epoch": 1583, "lr": 4.710525548324401e-05}
{"train_loss": 0.1406281590461731, "global_step": 140971, "epoch": 1583, "lr": 4.710467671923666e-05}
{"train_loss": 0.11747965216636658, "global_step": 140972, "epoch": 1583, "lr": 4.710409795561852e-05}
{"train_loss": 0.1651834398508072, "global_step": 140973, "epoch": 1583, "lr": 4.710351919238973e-05}
{"train_loss": 0.08841514587402344, "global_step": 140974, "epoch": 1583, "lr": 4.7102940429550305e-05}
{"train_loss": 0.11654298216774223, "global_step": 140975, "epoch": 1583, "lr": 4.7102361667100386e-05, "val_loss": 5.834299564361572}
{"train_loss": 0.10121139883995056, "global_step": 140976, "epoch": 1584, "lr": 4.7101782905039994e-05}
{"train_loss": 0.13833314180374146, "global_step": 140977, "epoch": 1584, "lr": 4.710120414336926e-05}
{"train_loss": 0.07652716338634491, "global_step": 140978, "epoch": 1584, "lr": 4.7100625382088197e-05}
{"train_loss": 0.08693580329418182, "global_step": 140979, "epoch": 1584, "lr": 4.7100046621196954e-05}
{"train_loss": 0.07412026077508926, "global_step": 140980, "epoch": 1584, "lr": 4.709946786069554e-05}
{"train_loss": 0.1495116502046585, "global_step": 140981, "epoch": 1584, "lr": 4.7098889100584106e-05}
{"train_loss": 0.09035499393939972, "global_step": 140982, "epoch": 1584, "lr": 4.709831034086267e-05}
{"train_loss": 0.08499594032764435, "global_step": 140983, "epoch": 1584, "lr": 4.7097731581531316e-05}
{"train_loss": 0.10377819836139679, "global_step": 140984, "epoch": 1584, "lr": 4.709715282259017e-05}
{"train_loss": 0.11804300546646118, "global_step": 140985, "epoch": 1584, "lr": 4.709657406403924e-05}
{"train_loss": 0.15777310729026794, "global_step": 140986, "epoch": 1584, "lr": 4.709599530587868e-05}
{"train_loss": 0.12109754979610443, "global_step": 140987, "epoch": 1584, "lr": 4.7095416548108495e-05}
{"train_loss": 0.10325229167938232, "global_step": 140988, "epoch": 1584, "lr": 4.709483779072882e-05}
{"train_loss": 0.08971519768238068, "global_step": 140989, "epoch": 1584, "lr": 4.709425903373968e-05}
{"train_loss": 0.08785087615251541, "global_step": 140990, "epoch": 1584, "lr": 4.7093680277141216e-05}
{"train_loss": 0.12654273211956024, "global_step": 140991, "epoch": 1584, "lr": 4.709310152093344e-05}
{"train_loss": 0.11455347388982773, "global_step": 140992, "epoch": 1584, "lr": 4.709252276511649e-05}
{"train_loss": 0.1274937391281128, "global_step": 140993, "epoch": 1584, "lr": 4.7091944009690385e-05}
{"train_loss": 0.10064451396465302, "global_step": 140994, "epoch": 1584, "lr": 4.7091365254655264e-05}
{"train_loss": 0.09441918879747391, "global_step": 140995, "epoch": 1584, "lr": 4.709078650001114e-05}
{"train_loss": 0.08930268883705139, "global_step": 140996, "epoch": 1584, "lr": 4.7090207745758155e-05}
{"train_loss": 0.11782261729240417, "global_step": 140997, "epoch": 1584, "lr": 4.708962899189633e-05}
{"train_loss": 0.08054570853710175, "global_step": 140998, "epoch": 1584, "lr": 4.7089050238425795e-05}
{"train_loss": 0.10028078407049179, "global_step": 140999, "epoch": 1584, "lr": 4.7088471485346574e-05}
{"train_loss": 0.09856259077787399, "global_step": 141000, "epoch": 1584, "lr": 4.708789273265879e-05}
{"train_loss": 0.18420813977718353, "global_step": 141001, "epoch": 1584, "lr": 4.708731398036251e-05}
{"train_loss": 0.08066224306821823, "global_step": 141002, "epoch": 1584, "lr": 4.708673522845778e-05}
{"train_loss": 0.17130866646766663, "global_step": 141003, "epoch": 1584, "lr": 4.708615647694473e-05}
{"train_loss": 0.09050253033638, "global_step": 141004, "epoch": 1584, "lr": 4.7085577725823385e-05}
{"train_loss": 0.08063347637653351, "global_step": 141005, "epoch": 1584, "lr": 4.7084998975093875e-05}
{"train_loss": 0.13526202738285065, "global_step": 141006, "epoch": 1584, "lr": 4.708442022475623e-05}
{"train_loss": 0.14379258453845978, "global_step": 141007, "epoch": 1584, "lr": 4.708384147481057e-05}
{"train_loss": 0.10863039642572403, "global_step": 141008, "epoch": 1584, "lr": 4.7083262725256924e-05}
{"train_loss": 0.09930696338415146, "global_step": 141009, "epoch": 1584, "lr": 4.708268397609543e-05}
{"train_loss": 0.11693298071622849, "global_step": 141010, "epoch": 1584, "lr": 4.708210522732611e-05}
{"train_loss": 0.08066415041685104, "global_step": 141011, "epoch": 1584, "lr": 4.708152647894908e-05}
{"train_loss": 0.07461748272180557, "global_step": 141012, "epoch": 1584, "lr": 4.708094773096438e-05}
{"train_loss": 0.10000017285346985, "global_step": 141013, "epoch": 1584, "lr": 4.708036898337214e-05}
{"train_loss": 0.08896486461162567, "global_step": 141014, "epoch": 1584, "lr": 4.7079790236172385e-05}
{"train_loss": 0.07626061141490936, "global_step": 141015, "epoch": 1584, "lr": 4.707921148936524e-05}
{"train_loss": 0.15199756622314453, "global_step": 141016, "epoch": 1584, "lr": 4.707863274295073e-05}
{"train_loss": 0.12425178289413452, "global_step": 141017, "epoch": 1584, "lr": 4.7078053996929e-05}
{"train_loss": 0.1555650383234024, "global_step": 141018, "epoch": 1584, "lr": 4.707747525130005e-05}
{"train_loss": 0.10028740763664246, "global_step": 141019, "epoch": 1584, "lr": 4.7076896506064014e-05}
{"train_loss": 0.08449891954660416, "global_step": 141020, "epoch": 1584, "lr": 4.707631776122097e-05}
{"train_loss": 0.06984095275402069, "global_step": 141021, "epoch": 1584, "lr": 4.7075739016770955e-05}
{"train_loss": 0.08506405353546143, "global_step": 141022, "epoch": 1584, "lr": 4.707516027271409e-05}
{"train_loss": 0.09649710357189178, "global_step": 141023, "epoch": 1584, "lr": 4.707458152905042e-05}
{"train_loss": 0.06388933211565018, "global_step": 141024, "epoch": 1584, "lr": 4.7074002785780053e-05}
{"train_loss": 0.10268528759479523, "global_step": 141025, "epoch": 1584, "lr": 4.707342404290303e-05}
{"train_loss": 0.10566134750843048, "global_step": 141026, "epoch": 1584, "lr": 4.707284530041947e-05}
{"train_loss": 0.09254749864339828, "global_step": 141027, "epoch": 1584, "lr": 4.707226655832941e-05}
{"train_loss": 0.10783218592405319, "global_step": 141028, "epoch": 1584, "lr": 4.707168781663297e-05}
{"train_loss": 0.11197320371866226, "global_step": 141029, "epoch": 1584, "lr": 4.7071109075330186e-05}
{"train_loss": 0.0838247612118721, "global_step": 141030, "epoch": 1584, "lr": 4.707053033442117e-05}
{"train_loss": 0.06395987421274185, "global_step": 141031, "epoch": 1584, "lr": 4.706995159390597e-05}
{"train_loss": 0.0807022750377655, "global_step": 141032, "epoch": 1584, "lr": 4.70693728537847e-05}
{"train_loss": 0.13033485412597656, "global_step": 141033, "epoch": 1584, "lr": 4.7068794114057394e-05}
{"train_loss": 0.08767500519752502, "global_step": 141034, "epoch": 1584, "lr": 4.7068215374724176e-05}
{"train_loss": 0.16556721925735474, "global_step": 141035, "epoch": 1584, "lr": 4.706763663578508e-05}
{"train_loss": 0.05959181487560272, "global_step": 141036, "epoch": 1584, "lr": 4.706705789724021e-05}
{"train_loss": 0.08428743481636047, "global_step": 141037, "epoch": 1584, "lr": 4.7066479159089643e-05}
{"train_loss": 0.08408248424530029, "global_step": 141038, "epoch": 1584, "lr": 4.706590042133345e-05}
{"train_loss": 0.1405673325061798, "global_step": 141039, "epoch": 1584, "lr": 4.7065321683971714e-05}
{"train_loss": 0.08334336429834366, "global_step": 141040, "epoch": 1584, "lr": 4.706474294700451e-05}
{"train_loss": 0.13298307359218597, "global_step": 141041, "epoch": 1584, "lr": 4.706416421043192e-05}
{"train_loss": 0.054935991764068604, "global_step": 141042, "epoch": 1584, "lr": 4.7063585474254005e-05}
{"train_loss": 0.09054818749427795, "global_step": 141043, "epoch": 1584, "lr": 4.706300673847087e-05}
{"train_loss": 0.18679209053516388, "global_step": 141044, "epoch": 1584, "lr": 4.706242800308256e-05}
{"train_loss": 0.11747775971889496, "global_step": 141045, "epoch": 1584, "lr": 4.7061849268089185e-05}
{"train_loss": 0.10436980426311493, "global_step": 141046, "epoch": 1584, "lr": 4.70612705334908e-05}
{"train_loss": 0.07628779113292694, "global_step": 141047, "epoch": 1584, "lr": 4.7060691799287506e-05}
{"train_loss": 0.1098649725317955, "global_step": 141048, "epoch": 1584, "lr": 4.7060113065479336e-05}
{"train_loss": 0.11708219349384308, "global_step": 141049, "epoch": 1584, "lr": 4.705953433206643e-05}
{"train_loss": 0.09029681980609894, "global_step": 141050, "epoch": 1584, "lr": 4.705895559904881e-05}
{"train_loss": 0.09559657424688339, "global_step": 141051, "epoch": 1584, "lr": 4.7058376866426604e-05}
{"train_loss": 0.1345595270395279, "global_step": 141052, "epoch": 1584, "lr": 4.705779813419983e-05}
{"train_loss": 0.11088990420103073, "global_step": 141053, "epoch": 1584, "lr": 4.705721940236863e-05}
{"train_loss": 0.11053571105003357, "global_step": 141054, "epoch": 1584, "lr": 4.7056640670933024e-05}
{"train_loss": 0.07988578826189041, "global_step": 141055, "epoch": 1584, "lr": 4.7056061939893134e-05}
{"train_loss": 0.12077193707227707, "global_step": 141056, "epoch": 1584, "lr": 4.705548320924902e-05}
{"train_loss": 0.08575668931007385, "global_step": 141057, "epoch": 1584, "lr": 4.705490447900076e-05}
{"train_loss": 0.1966615617275238, "global_step": 141058, "epoch": 1584, "lr": 4.705432574914843e-05}
{"train_loss": 0.11797498911619186, "global_step": 141059, "epoch": 1584, "lr": 4.70537470196921e-05}
{"train_loss": 0.08771125972270966, "global_step": 141060, "epoch": 1584, "lr": 4.705316829063188e-05}
{"train_loss": 0.11047933250665665, "global_step": 141061, "epoch": 1584, "lr": 4.7052589561967804e-05}
{"train_loss": 0.08246508985757828, "global_step": 141062, "epoch": 1584, "lr": 4.705201083369999e-05}
{"train_loss": 0.07068558037281036, "global_step": 141063, "epoch": 1584, "lr": 4.705143210582848e-05}
{"train_loss": 0.10530130640509423, "global_step": 141064, "epoch": 1584, "lr": 4.705085337835339e-05, "val_loss": 6.127336025238037}
{"train_loss": 0.12646378576755524, "global_step": 141065, "epoch": 1585, "lr": 4.705027465127475e-05}
{"train_loss": 0.09261894226074219, "global_step": 141066, "epoch": 1585, "lr": 4.704969592459268e-05}
{"train_loss": 0.07280325144529343, "global_step": 141067, "epoch": 1585, "lr": 4.7049117198307235e-05}
{"train_loss": 0.09021150320768356, "global_step": 141068, "epoch": 1585, "lr": 4.704853847241852e-05}
{"train_loss": 0.1168166771531105, "global_step": 141069, "epoch": 1585, "lr": 4.704795974692657e-05}
{"train_loss": 0.04399772733449936, "global_step": 141070, "epoch": 1585, "lr": 4.70473810218315e-05}
{"train_loss": 0.13440755009651184, "global_step": 141071, "epoch": 1585, "lr": 4.7046802297133354e-05}
{"train_loss": 0.05542677268385887, "global_step": 141072, "epoch": 1585, "lr": 4.704622357283226e-05}
{"train_loss": 0.1514982134103775, "global_step": 141073, "epoch": 1585, "lr": 4.704564484892824e-05}
{"train_loss": 0.04237111657857895, "global_step": 141074, "epoch": 1585, "lr": 4.70450661254214e-05}
{"train_loss": 0.1729043573141098, "global_step": 141075, "epoch": 1585, "lr": 4.704448740231183e-05}
{"train_loss": 0.14037641882896423, "global_step": 141076, "epoch": 1585, "lr": 4.704390867959958e-05}
{"train_loss": 0.09716290980577469, "global_step": 141077, "epoch": 1585, "lr": 4.7043329957284745e-05}
{"train_loss": 0.14958825707435608, "global_step": 141078, "epoch": 1585, "lr": 4.704275123536739e-05}
{"train_loss": 0.09100250899791718, "global_step": 141079, "epoch": 1585, "lr": 4.704217251384763e-05}
{"train_loss": 0.1704711765050888, "global_step": 141080, "epoch": 1585, "lr": 4.704159379272547e-05}
{"train_loss": 0.10965649038553238, "global_step": 141081, "epoch": 1585, "lr": 4.7041015072001074e-05}
{"train_loss": 0.11796916276216507, "global_step": 141082, "epoch": 1585, "lr": 4.704043635167444e-05}
{"train_loss": 0.07353022694587708, "global_step": 141083, "epoch": 1585, "lr": 4.703985763174571e-05}
{"train_loss": 0.11754651367664337, "global_step": 141084, "epoch": 1585, "lr": 4.703927891221492e-05}
{"train_loss": 0.08061939477920532, "global_step": 141085, "epoch": 1585, "lr": 4.703870019308218e-05}
{"train_loss": 0.13546495139598846, "global_step": 141086, "epoch": 1585, "lr": 4.703812147434752e-05}
{"train_loss": 0.11966082453727722, "global_step": 141087, "epoch": 1585, "lr": 4.7037542756011085e-05}
{"train_loss": 0.0771726667881012, "global_step": 141088, "epoch": 1585, "lr": 4.703696403807288e-05}
{"train_loss": 0.10379861295223236, "global_step": 141089, "epoch": 1585, "lr": 4.703638532053306e-05}
{"train_loss": 0.1422259509563446, "global_step": 141090, "epoch": 1585, "lr": 4.703580660339162e-05}
{"train_loss": 0.06003263220191002, "global_step": 141091, "epoch": 1585, "lr": 4.703522788664869e-05}
{"train_loss": 0.05059066787362099, "global_step": 141092, "epoch": 1585, "lr": 4.703464917030436e-05}
{"train_loss": 0.10098621994256973, "global_step": 141093, "epoch": 1585, "lr": 4.703407045435867e-05}
{"train_loss": 0.10572104901075363, "global_step": 141094, "epoch": 1585, "lr": 4.703349173881172e-05}
{"train_loss": 0.10460197925567627, "global_step": 141095, "epoch": 1585, "lr": 4.703291302366357e-05}
{"train_loss": 0.1282699555158615, "global_step": 141096, "epoch": 1585, "lr": 4.703233430891432e-05}
{"train_loss": 0.1213257685303688, "global_step": 141097, "epoch": 1585, "lr": 4.703175559456403e-05}
{"train_loss": 0.10321410745382309, "global_step": 141098, "epoch": 1585, "lr": 4.70311768806128e-05}
{"train_loss": 0.09683168679475784, "global_step": 141099, "epoch": 1585, "lr": 4.7030598167060665e-05}
{"train_loss": 0.09151428937911987, "global_step": 141100, "epoch": 1585, "lr": 4.703001945390775e-05}
{"train_loss": 0.08813649415969849, "global_step": 141101, "epoch": 1585, "lr": 4.70294407411541e-05}
{"train_loss": 0.13682343065738678, "global_step": 141102, "epoch": 1585, "lr": 4.702886202879982e-05}
{"train_loss": 0.13105015456676483, "global_step": 141103, "epoch": 1585, "lr": 4.702828331684495e-05}
{"train_loss": 0.11636150628328323, "global_step": 141104, "epoch": 1585, "lr": 4.702770460528962e-05}
{"train_loss": 0.1458907425403595, "global_step": 141105, "epoch": 1585, "lr": 4.702712589413385e-05}
{"train_loss": 0.11493372917175293, "global_step": 141106, "epoch": 1585, "lr": 4.702654718337777e-05}
{"train_loss": 0.12155916541814804, "global_step": 141107, "epoch": 1585, "lr": 4.7025968473021415e-05}
{"train_loss": 0.11248447746038437, "global_step": 141108, "epoch": 1585, "lr": 4.7025389763064906e-05}
{"train_loss": 0.20259477198123932, "global_step": 141109, "epoch": 1585, "lr": 4.702481105350827e-05}
{"train_loss": 0.09833571314811707, "global_step": 141110, "epoch": 1585, "lr": 4.702423234435161e-05}
{"train_loss": 0.1501242071390152, "global_step": 141111, "epoch": 1585, "lr": 4.702365363559504e-05}
{"train_loss": 0.13943679630756378, "global_step": 141112, "epoch": 1585, "lr": 4.702307492723858e-05}
{"train_loss": 0.1646689474582672, "global_step": 141113, "epoch": 1585, "lr": 4.702249621928234e-05}
{"train_loss": 0.09011638909578323, "global_step": 141114, "epoch": 1585, "lr": 4.702191751172638e-05}
{"train_loss": 0.090056873857975, "global_step": 141115, "epoch": 1585, "lr": 4.70213388045708e-05}
{"train_loss": 0.10414600372314453, "global_step": 141116, "epoch": 1585, "lr": 4.7020760097815656e-05}
{"train_loss": 0.11642015725374222, "global_step": 141117, "epoch": 1585, "lr": 4.702018139146104e-05}
{"train_loss": 0.09416786581277847, "global_step": 141118, "epoch": 1585, "lr": 4.701960268550701e-05}
{"train_loss": 0.09954391419887543, "global_step": 141119, "epoch": 1585, "lr": 4.701902397995368e-05}
{"train_loss": 0.14694744348526, "global_step": 141120, "epoch": 1585, "lr": 4.701844527480108e-05}
{"train_loss": 0.13507886230945587, "global_step": 141121, "epoch": 1585, "lr": 4.701786657004934e-05}
{"train_loss": 0.12660489976406097, "global_step": 141122, "epoch": 1585, "lr": 4.701728786569848e-05}
{"train_loss": 0.08003953099250793, "global_step": 141123, "epoch": 1585, "lr": 4.7016709161748644e-05}
{"train_loss": 0.08663321286439896, "global_step": 141124, "epoch": 1585, "lr": 4.7016130458199844e-05}
{"train_loss": 0.17887409031391144, "global_step": 141125, "epoch": 1585, "lr": 4.701555175505222e-05}
{"train_loss": 0.06599392741918564, "global_step": 141126, "epoch": 1585, "lr": 4.701497305230579e-05}
{"train_loss": 0.18496298789978027, "global_step": 141127, "epoch": 1585, "lr": 4.701439434996066e-05}
{"train_loss": 0.17798927426338196, "global_step": 141128, "epoch": 1585, "lr": 4.7013815648016933e-05}
{"train_loss": 0.1607734113931656, "global_step": 141129, "epoch": 1585, "lr": 4.701323694647464e-05}
{"train_loss": 0.10999217629432678, "global_step": 141130, "epoch": 1585, "lr": 4.70126582453339e-05}
{"train_loss": 0.08959818631410599, "global_step": 141131, "epoch": 1585, "lr": 4.7012079544594764e-05}
{"train_loss": 0.1391892433166504, "global_step": 141132, "epoch": 1585, "lr": 4.701150084425732e-05}
{"train_loss": 0.07499762624502182, "global_step": 141133, "epoch": 1585, "lr": 4.701092214432164e-05}
{"train_loss": 0.11386742442846298, "global_step": 141134, "epoch": 1585, "lr": 4.70103434447878e-05}
{"train_loss": 0.08077479153871536, "global_step": 141135, "epoch": 1585, "lr": 4.700976474565589e-05}
{"train_loss": 0.0691002905368805, "global_step": 141136, "epoch": 1585, "lr": 4.7009186046925976e-05}
{"train_loss": 0.12883597612380981, "global_step": 141137, "epoch": 1585, "lr": 4.7008607348598146e-05}
{"train_loss": 0.13031065464019775, "global_step": 141138, "epoch": 1585, "lr": 4.700802865067248e-05}
{"train_loss": 0.038448527455329895, "global_step": 141139, "epoch": 1585, "lr": 4.700744995314903e-05}
{"train_loss": 0.09581630676984787, "global_step": 141140, "epoch": 1585, "lr": 4.7006871256027905e-05}
{"train_loss": 0.09920449554920197, "global_step": 141141, "epoch": 1585, "lr": 4.700629255930916e-05}
{"train_loss": 0.09830201417207718, "global_step": 141142, "epoch": 1585, "lr": 4.7005713862992896e-05}
{"train_loss": 0.11245325952768326, "global_step": 141143, "epoch": 1585, "lr": 4.700513516707915e-05}
{"train_loss": 0.12039557844400406, "global_step": 141144, "epoch": 1585, "lr": 4.7004556471568057e-05}
{"train_loss": 0.13880085945129395, "global_step": 141145, "epoch": 1585, "lr": 4.700397777645964e-05}
{"train_loss": 0.07345905900001526, "global_step": 141146, "epoch": 1585, "lr": 4.700339908175401e-05}
{"train_loss": 0.12636353075504303, "global_step": 141147, "epoch": 1585, "lr": 4.7002820387451245e-05}
{"train_loss": 0.1070936992764473, "global_step": 141148, "epoch": 1585, "lr": 4.7002241693551405e-05}
{"train_loss": 0.1380401849746704, "global_step": 141149, "epoch": 1585, "lr": 4.700166300005459e-05}
{"train_loss": 0.12260177731513977, "global_step": 141150, "epoch": 1585, "lr": 4.7001084306960846e-05}
{"train_loss": 0.0627567395567894, "global_step": 141151, "epoch": 1585, "lr": 4.700050561427029e-05}
{"train_loss": 0.07709112763404846, "global_step": 141152, "epoch": 1585, "lr": 4.699992692198297e-05}
{"train_loss": 0.11112389140082209, "global_step": 141153, "epoch": 1585, "lr": 4.6999348230098974e-05, "val_loss": 5.9678544998168945, "train_action_mse_error": 16.62075424194336}
{"train_loss": 0.10993801057338715, "global_step": 141154, "epoch": 1586, "lr": 4.6998769538618375e-05}
{"train_loss": 0.09516695141792297, "global_step": 141155, "epoch": 1586, "lr": 4.699819084754126e-05}
{"train_loss": 0.12510870397090912, "global_step": 141156, "epoch": 1586, "lr": 4.69976121568677e-05}
{"train_loss": 0.21056534349918365, "global_step": 141157, "epoch": 1586, "lr": 4.699703346659778e-05}
{"train_loss": 0.09422403573989868, "global_step": 141158, "epoch": 1586, "lr": 4.699645477673156e-05}
{"train_loss": 0.1441839039325714, "global_step": 141159, "epoch": 1586, "lr": 4.699587608726915e-05}
{"train_loss": 0.11159292608499527, "global_step": 141160, "epoch": 1586, "lr": 4.699529739821058e-05}
{"train_loss": 0.056280191987752914, "global_step": 141161, "epoch": 1586, "lr": 4.6994718709555985e-05}
{"train_loss": 0.1126963421702385, "global_step": 141162, "epoch": 1586, "lr": 4.6994140021305386e-05}
{"train_loss": 0.1196814477443695, "global_step": 141163, "epoch": 1586, "lr": 4.699356133345889e-05}
{"train_loss": 0.07551468908786774, "global_step": 141164, "epoch": 1586, "lr": 4.699298264601659e-05}
{"train_loss": 0.16764099895954132, "global_step": 141165, "epoch": 1586, "lr": 4.699240395897853e-05}
{"train_loss": 0.09185823798179626, "global_step": 141166, "epoch": 1586, "lr": 4.699182527234483e-05}
{"train_loss": 0.14605849981307983, "global_step": 141167, "epoch": 1586, "lr": 4.699124658611552e-05}
{"train_loss": 0.13877762854099274, "global_step": 141168, "epoch": 1586, "lr": 4.699066790029071e-05}
{"train_loss": 0.13901033997535706, "global_step": 141169, "epoch": 1586, "lr": 4.699008921487046e-05}
{"train_loss": 0.12313501536846161, "global_step": 141170, "epoch": 1586, "lr": 4.6989510529854876e-05}
{"train_loss": 0.1216939240694046, "global_step": 141171, "epoch": 1586, "lr": 4.6988931845244e-05}
{"train_loss": 0.10847914218902588, "global_step": 141172, "epoch": 1586, "lr": 4.698835316103792e-05}
{"train_loss": 0.06281431764364243, "global_step": 141173, "epoch": 1586, "lr": 4.6987774477236726e-05}
{"train_loss": 0.13866497576236725, "global_step": 141174, "epoch": 1586, "lr": 4.6987195793840494e-05}
{"train_loss": 0.08602102845907211, "global_step": 141175, "epoch": 1586, "lr": 4.698661711084929e-05}
{"train_loss": 0.09339609742164612, "global_step": 141176, "epoch": 1586, "lr": 4.698603842826321e-05}
{"train_loss": 0.14209851622581482, "global_step": 141177, "epoch": 1586, "lr": 4.698545974608231e-05}
{"train_loss": 0.08771868050098419, "global_step": 141178, "epoch": 1586, "lr": 4.6984881064306687e-05}
{"train_loss": 0.0837363600730896, "global_step": 141179, "epoch": 1586, "lr": 4.698430238293639e-05}
{"train_loss": 0.13796338438987732, "global_step": 141180, "epoch": 1586, "lr": 4.698372370197154e-05}
{"train_loss": 0.16572673618793488, "global_step": 141181, "epoch": 1586, "lr": 4.698314502141217e-05}
{"train_loss": 0.07745590060949326, "global_step": 141182, "epoch": 1586, "lr": 4.698256634125837e-05}
{"train_loss": 0.16585665941238403, "global_step": 141183, "epoch": 1586, "lr": 4.6981987661510265e-05}
{"train_loss": 0.10577241331338882, "global_step": 141184, "epoch": 1586, "lr": 4.6981408982167864e-05}
{"train_loss": 0.09765937179327011, "global_step": 141185, "epoch": 1586, "lr": 4.69808303032313e-05}
{"train_loss": 0.13480941951274872, "global_step": 141186, "epoch": 1586, "lr": 4.69802516247006e-05}
{"train_loss": 0.09280315786600113, "global_step": 141187, "epoch": 1586, "lr": 4.69796729465759e-05}
{"train_loss": 0.10041745752096176, "global_step": 141188, "epoch": 1586, "lr": 4.697909426885722e-05}
{"train_loss": 0.064452163875103, "global_step": 141189, "epoch": 1586, "lr": 4.697851559154468e-05}
{"train_loss": 0.1199253499507904, "global_step": 141190, "epoch": 1586, "lr": 4.6977936914638335e-05}
{"train_loss": 0.12670721113681793, "global_step": 141191, "epoch": 1586, "lr": 4.697735823813827e-05}
{"train_loss": 0.06374413520097733, "global_step": 141192, "epoch": 1586, "lr": 4.697677956204457e-05}
{"train_loss": 0.13200822472572327, "global_step": 141193, "epoch": 1586, "lr": 4.697620088635729e-05}
{"train_loss": 0.10765256732702255, "global_step": 141194, "epoch": 1586, "lr": 4.697562221107653e-05}
{"train_loss": 0.12871253490447998, "global_step": 141195, "epoch": 1586, "lr": 4.697504353620237e-05}
{"train_loss": 0.14117538928985596, "global_step": 141196, "epoch": 1586, "lr": 4.697446486173486e-05}
{"train_loss": 0.09374196827411652, "global_step": 141197, "epoch": 1586, "lr": 4.697388618767412e-05}
{"train_loss": 0.07316919416189194, "global_step": 141198, "epoch": 1586, "lr": 4.6973307514020174e-05}
{"train_loss": 0.12179280072450638, "global_step": 141199, "epoch": 1586, "lr": 4.697272884077316e-05}
{"train_loss": 0.12110597640275955, "global_step": 141200, "epoch": 1586, "lr": 4.69721501679331e-05}
{"train_loss": 0.06913149356842041, "global_step": 141201, "epoch": 1586, "lr": 4.69715714955001e-05}
{"train_loss": 0.06658884882926941, "global_step": 141202, "epoch": 1586, "lr": 4.697099282347425e-05}
{"train_loss": 0.10886164009571075, "global_step": 141203, "epoch": 1586, "lr": 4.6970414151855604e-05}
{"train_loss": 0.08972303569316864, "global_step": 141204, "epoch": 1586, "lr": 4.696983548064426e-05}
{"train_loss": 0.14002510905265808, "global_step": 141205, "epoch": 1586, "lr": 4.696925680984027e-05}
{"train_loss": 0.128745436668396, "global_step": 141206, "epoch": 1586, "lr": 4.696867813944374e-05}
{"train_loss": 0.07633163034915924, "global_step": 141207, "epoch": 1586, "lr": 4.6968099469454715e-05}
{"train_loss": 0.12843187153339386, "global_step": 141208, "epoch": 1586, "lr": 4.696752079987332e-05}
{"train_loss": 0.11506565660238266, "global_step": 141209, "epoch": 1586, "lr": 4.696694213069959e-05}
{"train_loss": 0.07945968955755234, "global_step": 141210, "epoch": 1586, "lr": 4.6966363461933616e-05}
{"train_loss": 0.13530389964580536, "global_step": 141211, "epoch": 1586, "lr": 4.696578479357548e-05}
{"train_loss": 0.1840198040008545, "global_step": 141212, "epoch": 1586, "lr": 4.696520612562526e-05}
{"train_loss": 0.09095306694507599, "global_step": 141213, "epoch": 1586, "lr": 4.696462745808302e-05}
{"train_loss": 0.07236044853925705, "global_step": 141214, "epoch": 1586, "lr": 4.6964048790948874e-05}
{"train_loss": 0.1239369586110115, "global_step": 141215, "epoch": 1586, "lr": 4.696347012422284e-05}
{"train_loss": 0.1576891839504242, "global_step": 141216, "epoch": 1586, "lr": 4.696289145790506e-05}
{"train_loss": 0.10607918351888657, "global_step": 141217, "epoch": 1586, "lr": 4.696231279199556e-05}
{"train_loss": 0.12547138333320618, "global_step": 141218, "epoch": 1586, "lr": 4.696173412649444e-05}
{"train_loss": 0.08366663008928299, "global_step": 141219, "epoch": 1586, "lr": 4.6961155461401795e-05}
{"train_loss": 0.12006794661283493, "global_step": 141220, "epoch": 1586, "lr": 4.696057679671767e-05}
{"train_loss": 0.12694872915744781, "global_step": 141221, "epoch": 1586, "lr": 4.695999813244217e-05}
{"train_loss": 0.08127904683351517, "global_step": 141222, "epoch": 1586, "lr": 4.695941946857535e-05}
{"train_loss": 0.1974153369665146, "global_step": 141223, "epoch": 1586, "lr": 4.695884080511732e-05}
{"train_loss": 0.08842872083187103, "global_step": 141224, "epoch": 1586, "lr": 4.695826214206811e-05}
{"train_loss": 0.11100853234529495, "global_step": 141225, "epoch": 1586, "lr": 4.695768347942785e-05}
{"train_loss": 0.12208936363458633, "global_step": 141226, "epoch": 1586, "lr": 4.6957104817196566e-05}
{"train_loss": 0.09194080531597137, "global_step": 141227, "epoch": 1586, "lr": 4.695652615537439e-05}
{"train_loss": 0.14498743414878845, "global_step": 141228, "epoch": 1586, "lr": 4.695594749396136e-05}
{"train_loss": 0.12613944709300995, "global_step": 141229, "epoch": 1586, "lr": 4.6955368832957564e-05}
{"train_loss": 0.10593871772289276, "global_step": 141230, "epoch": 1586, "lr": 4.6954790172363086e-05}
{"train_loss": 0.1225019320845604, "global_step": 141231, "epoch": 1586, "lr": 4.695421151217799e-05}
{"train_loss": 0.12139986455440521, "global_step": 141232, "epoch": 1586, "lr": 4.695363285240237e-05}
{"train_loss": 0.05402982607483864, "global_step": 141233, "epoch": 1586, "lr": 4.6953054193036305e-05}
{"train_loss": 0.20765255391597748, "global_step": 141234, "epoch": 1586, "lr": 4.695247553407984e-05}
{"train_loss": 0.11073283851146698, "global_step": 141235, "epoch": 1586, "lr": 4.695189687553311e-05}
{"train_loss": 0.1059742271900177, "global_step": 141236, "epoch": 1586, "lr": 4.695131821739613e-05}
{"train_loss": 0.11130737513303757, "global_step": 141237, "epoch": 1586, "lr": 4.695073955966901e-05}
{"train_loss": 0.13349895179271698, "global_step": 141238, "epoch": 1586, "lr": 4.695016090235185e-05}
{"train_loss": 0.0913611501455307, "global_step": 141239, "epoch": 1586, "lr": 4.6949582245444675e-05}
{"train_loss": 0.09547736495733261, "global_step": 141240, "epoch": 1586, "lr": 4.6949003588947614e-05}
{"train_loss": 0.05698799714446068, "global_step": 141241, "epoch": 1586, "lr": 4.694842493286071e-05}
{"train_loss": 0.11267708710740122, "global_step": 141242, "epoch": 1586, "lr": 4.694784627718407e-05, "val_loss": 6.048781394958496}
{"train_loss": 0.06219121813774109, "global_step": 141243, "epoch": 1587, "lr": 4.6947267621917726e-05}
{"train_loss": 0.09556622803211212, "global_step": 141244, "epoch": 1587, "lr": 4.6946688967061815e-05}
{"train_loss": 0.0801357626914978, "global_step": 141245, "epoch": 1587, "lr": 4.694611031261636e-05}
{"train_loss": 0.12040948122739792, "global_step": 141246, "epoch": 1587, "lr": 4.694553165858149e-05}
{"train_loss": 0.0646025761961937, "global_step": 141247, "epoch": 1587, "lr": 4.694495300495724e-05}
{"train_loss": 0.06658408790826797, "global_step": 141248, "epoch": 1587, "lr": 4.6944374351743706e-05}
{"train_loss": 0.1210467591881752, "global_step": 141249, "epoch": 1587, "lr": 4.6943795698940955e-05}
{"train_loss": 0.10884387046098709, "global_step": 141250, "epoch": 1587, "lr": 4.694321704654908e-05}
{"train_loss": 0.1163235604763031, "global_step": 141251, "epoch": 1587, "lr": 4.694263839456816e-05}
{"train_loss": 0.09442782402038574, "global_step": 141252, "epoch": 1587, "lr": 4.6942059742998253e-05}
{"train_loss": 0.11492113023996353, "global_step": 141253, "epoch": 1587, "lr": 4.694148109183945e-05}
{"train_loss": 0.0862560048699379, "global_step": 141254, "epoch": 1587, "lr": 4.694090244109182e-05}
{"train_loss": 0.13731229305267334, "global_step": 141255, "epoch": 1587, "lr": 4.694032379075547e-05}
{"train_loss": 0.08852047473192215, "global_step": 141256, "epoch": 1587, "lr": 4.693974514083044e-05}
{"train_loss": 0.10532169044017792, "global_step": 141257, "epoch": 1587, "lr": 4.693916649131684e-05}
{"train_loss": 0.10968145728111267, "global_step": 141258, "epoch": 1587, "lr": 4.693858784221471e-05}
{"train_loss": 0.174719899892807, "global_step": 141259, "epoch": 1587, "lr": 4.693800919352417e-05}
{"train_loss": 0.0683857649564743, "global_step": 141260, "epoch": 1587, "lr": 4.6937430545245257e-05}
{"train_loss": 0.10013169050216675, "global_step": 141261, "epoch": 1587, "lr": 4.693685189737809e-05}
{"train_loss": 0.10918434709310532, "global_step": 141262, "epoch": 1587, "lr": 4.693627324992271e-05}
{"train_loss": 0.14134065806865692, "global_step": 141263, "epoch": 1587, "lr": 4.693569460287922e-05}
{"train_loss": 0.13939756155014038, "global_step": 141264, "epoch": 1587, "lr": 4.693511595624768e-05}
{"train_loss": 0.07888960838317871, "global_step": 141265, "epoch": 1587, "lr": 4.693453731002819e-05}
{"train_loss": 0.0736219510436058, "global_step": 141266, "epoch": 1587, "lr": 4.6933958664220794e-05}
{"train_loss": 0.18221984803676605, "global_step": 141267, "epoch": 1587, "lr": 4.693338001882561e-05}
{"train_loss": 0.1220201849937439, "global_step": 141268, "epoch": 1587, "lr": 4.6932801373842696e-05}
{"train_loss": 0.08808288723230362, "global_step": 141269, "epoch": 1587, "lr": 4.6932222729272116e-05}
{"train_loss": 0.10044389218091965, "global_step": 141270, "epoch": 1587, "lr": 4.693164408511397e-05}
{"train_loss": 0.07829355448484421, "global_step": 141271, "epoch": 1587, "lr": 4.693106544136832e-05}
{"train_loss": 0.11734934151172638, "global_step": 141272, "epoch": 1587, "lr": 4.693048679803526e-05}
{"train_loss": 0.18493814766407013, "global_step": 141273, "epoch": 1587, "lr": 4.692990815511484e-05}
{"train_loss": 0.0693892389535904, "global_step": 141274, "epoch": 1587, "lr": 4.692932951260717e-05}
{"train_loss": 0.1177210733294487, "global_step": 141275, "epoch": 1587, "lr": 4.692875087051231e-05}
{"train_loss": 0.18140587210655212, "global_step": 141276, "epoch": 1587, "lr": 4.6928172228830354e-05}
{"train_loss": 0.1406383514404297, "global_step": 141277, "epoch": 1587, "lr": 4.6927593587561347e-05}
{"train_loss": 0.1497214287519455, "global_step": 141278, "epoch": 1587, "lr": 4.692701494670541e-05}
{"train_loss": 0.11805462837219238, "global_step": 141279, "epoch": 1587, "lr": 4.692643630626258e-05}
{"train_loss": 0.09242426604032516, "global_step": 141280, "epoch": 1587, "lr": 4.692585766623297e-05}
{"train_loss": 0.10065765678882599, "global_step": 141281, "epoch": 1587, "lr": 4.692527902661662e-05}
{"train_loss": 0.10194262117147446, "global_step": 141282, "epoch": 1587, "lr": 4.6924700387413654e-05}
{"train_loss": 0.16960792243480682, "global_step": 141283, "epoch": 1587, "lr": 4.69241217486241e-05}
{"train_loss": 0.1061471551656723, "global_step": 141284, "epoch": 1587, "lr": 4.6923543110248084e-05}
{"train_loss": 0.1365666538476944, "global_step": 141285, "epoch": 1587, "lr": 4.6922964472285635e-05}
{"train_loss": 0.10370636731386185, "global_step": 141286, "epoch": 1587, "lr": 4.6922385834736884e-05}
{"train_loss": 0.11036074906587601, "global_step": 141287, "epoch": 1587, "lr": 4.6921807197601864e-05}
{"train_loss": 0.0845949575304985, "global_step": 141288, "epoch": 1587, "lr": 4.692122856088067e-05}
{"train_loss": 0.12444241344928741, "global_step": 141289, "epoch": 1587, "lr": 4.692064992457338e-05}
{"train_loss": 0.05895902216434479, "global_step": 141290, "epoch": 1587, "lr": 4.692007128868006e-05}
{"train_loss": 0.0901816263794899, "global_step": 141291, "epoch": 1587, "lr": 4.691949265320082e-05}
{"train_loss": 0.1109965369105339, "global_step": 141292, "epoch": 1587, "lr": 4.69189140181357e-05}
{"train_loss": 0.08751214295625687, "global_step": 141293, "epoch": 1587, "lr": 4.691833538348481e-05}
{"train_loss": 0.08198317140340805, "global_step": 141294, "epoch": 1587, "lr": 4.691775674924819e-05}
{"train_loss": 0.07116993516683578, "global_step": 141295, "epoch": 1587, "lr": 4.691717811542597e-05}
{"train_loss": 0.14186905324459076, "global_step": 141296, "epoch": 1587, "lr": 4.691659948201817e-05}
{"train_loss": 0.17018654942512512, "global_step": 141297, "epoch": 1587, "lr": 4.6916020849024915e-05}
{"train_loss": 0.1199616938829422, "global_step": 141298, "epoch": 1587, "lr": 4.691544221644625e-05}
{"train_loss": 0.06593673676252365, "global_step": 141299, "epoch": 1587, "lr": 4.691486358428228e-05}
{"train_loss": 0.1201377660036087, "global_step": 141300, "epoch": 1587, "lr": 4.691428495253305e-05}
{"train_loss": 0.0659831315279007, "global_step": 141301, "epoch": 1587, "lr": 4.691370632119868e-05}
{"train_loss": 0.14530950784683228, "global_step": 141302, "epoch": 1587, "lr": 4.6913127690279204e-05}
{"train_loss": 0.031921498477458954, "global_step": 141303, "epoch": 1587, "lr": 4.691254905977474e-05}
{"train_loss": 0.07424129545688629, "global_step": 141304, "epoch": 1587, "lr": 4.691197042968532e-05}
{"train_loss": 0.09712124615907669, "global_step": 141305, "epoch": 1587, "lr": 4.6911391800011073e-05}
{"train_loss": 0.07167771458625793, "global_step": 141306, "epoch": 1587, "lr": 4.6910813170752044e-05}
{"train_loss": 0.14461275935173035, "global_step": 141307, "epoch": 1587, "lr": 4.691023454190832e-05}
{"train_loss": 0.05831307917833328, "global_step": 141308, "epoch": 1587, "lr": 4.6909655913479964e-05}
{"train_loss": 0.11963563412427902, "global_step": 141309, "epoch": 1587, "lr": 4.690907728546707e-05}
{"train_loss": 0.12865005433559418, "global_step": 141310, "epoch": 1587, "lr": 4.690849865786973e-05}
{"train_loss": 0.13427820801734924, "global_step": 141311, "epoch": 1587, "lr": 4.6907920030687977e-05}
{"train_loss": 0.07721070945262909, "global_step": 141312, "epoch": 1587, "lr": 4.690734140392195e-05}
{"train_loss": 0.07745437324047089, "global_step": 141313, "epoch": 1587, "lr": 4.6906762777571664e-05}
{"train_loss": 0.15669503808021545, "global_step": 141314, "epoch": 1587, "lr": 4.690618415163725e-05}
{"train_loss": 0.10059568285942078, "global_step": 141315, "epoch": 1587, "lr": 4.690560552611874e-05}
{"train_loss": 0.07225269079208374, "global_step": 141316, "epoch": 1587, "lr": 4.690502690101625e-05}
{"train_loss": 0.09564962238073349, "global_step": 141317, "epoch": 1587, "lr": 4.6904448276329824e-05}
{"train_loss": 0.12405863404273987, "global_step": 141318, "epoch": 1587, "lr": 4.690386965205958e-05}
{"train_loss": 0.18195860087871552, "global_step": 141319, "epoch": 1587, "lr": 4.690329102820555e-05}
{"train_loss": 0.14379607141017914, "global_step": 141320, "epoch": 1587, "lr": 4.690271240476786e-05}
{"train_loss": 0.19978737831115723, "global_step": 141321, "epoch": 1587, "lr": 4.690213378174654e-05}
{"train_loss": 0.05711996927857399, "global_step": 141322, "epoch": 1587, "lr": 4.690155515914171e-05}
{"train_loss": 0.15137964487075806, "global_step": 141323, "epoch": 1587, "lr": 4.690097653695341e-05}
{"train_loss": 0.10446661710739136, "global_step": 141324, "epoch": 1587, "lr": 4.6900397915181756e-05}
{"train_loss": 0.1052173376083374, "global_step": 141325, "epoch": 1587, "lr": 4.689981929382679e-05}
{"train_loss": 0.09960625320672989, "global_step": 141326, "epoch": 1587, "lr": 4.689924067288861e-05}
{"train_loss": 0.14020393788814545, "global_step": 141327, "epoch": 1587, "lr": 4.6898662052367285e-05}
{"train_loss": 0.11017487198114395, "global_step": 141328, "epoch": 1587, "lr": 4.689808343226289e-05}
{"train_loss": 0.11674898117780685, "global_step": 141329, "epoch": 1587, "lr": 4.689750481257553e-05}
{"train_loss": 0.08419404178857803, "global_step": 141330, "epoch": 1587, "lr": 4.6896926193305244e-05}
{"train_loss": 0.10913255948866352, "global_step": 141331, "epoch": 1587, "lr": 4.689634757445215e-05, "val_loss": 5.994290351867676}
{"train_loss": 0.16398559510707855, "global_step": 141332, "epoch": 1588, "lr": 4.689576895601628e-05}
{"train_loss": 0.11814744770526886, "global_step": 141333, "epoch": 1588, "lr": 4.6895190337997754e-05}
{"train_loss": 0.14766117930412292, "global_step": 141334, "epoch": 1588, "lr": 4.6894611720396616e-05}
{"train_loss": 0.1372230052947998, "global_step": 141335, "epoch": 1588, "lr": 4.689403310321298e-05}
{"train_loss": 0.08118312060832977, "global_step": 141336, "epoch": 1588, "lr": 4.689345448644688e-05}
{"train_loss": 0.11822839826345444, "global_step": 141337, "epoch": 1588, "lr": 4.689287587009844e-05}
{"train_loss": 0.11212701350450516, "global_step": 141338, "epoch": 1588, "lr": 4.6892297254167696e-05}
{"train_loss": 0.13197530806064606, "global_step": 141339, "epoch": 1588, "lr": 4.6891718638654764e-05}
{"train_loss": 0.14766426384449005, "global_step": 141340, "epoch": 1588, "lr": 4.689114002355968e-05}
{"train_loss": 0.11739892512559891, "global_step": 141341, "epoch": 1588, "lr": 4.6890561408882565e-05}
{"train_loss": 0.1845158040523529, "global_step": 141342, "epoch": 1588, "lr": 4.688998279462345e-05}
{"train_loss": 0.07700085639953613, "global_step": 141343, "epoch": 1588, "lr": 4.6889404180782475e-05}
{"train_loss": 0.15951624512672424, "global_step": 141344, "epoch": 1588, "lr": 4.688882556735966e-05}
{"train_loss": 0.16930221021175385, "global_step": 141345, "epoch": 1588, "lr": 4.68882469543551e-05}
{"train_loss": 0.12116598337888718, "global_step": 141346, "epoch": 1588, "lr": 4.688766834176889e-05}
{"train_loss": 0.05213332548737526, "global_step": 141347, "epoch": 1588, "lr": 4.688708972960108e-05}
{"train_loss": 0.094273142516613, "global_step": 141348, "epoch": 1588, "lr": 4.688651111785178e-05}
{"train_loss": 0.1466602087020874, "global_step": 141349, "epoch": 1588, "lr": 4.688593250652103e-05}
{"train_loss": 0.0987112820148468, "global_step": 141350, "epoch": 1588, "lr": 4.688535389560896e-05}
{"train_loss": 0.14190207421779633, "global_step": 141351, "epoch": 1588, "lr": 4.6884775285115587e-05}
{"train_loss": 0.12412793934345245, "global_step": 141352, "epoch": 1588, "lr": 4.6884196675041034e-05}
{"train_loss": 0.0667530819773674, "global_step": 141353, "epoch": 1588, "lr": 4.688361806538535e-05}
{"train_loss": 0.10916414856910706, "global_step": 141354, "epoch": 1588, "lr": 4.688303945614864e-05}
{"train_loss": 0.1387677937746048, "global_step": 141355, "epoch": 1588, "lr": 4.6882460847330955e-05}
{"train_loss": 0.13659925758838654, "global_step": 141356, "epoch": 1588, "lr": 4.68818822389324e-05}
{"train_loss": 0.14768628776073456, "global_step": 141357, "epoch": 1588, "lr": 4.6881303630953016e-05}
{"train_loss": 0.06270880252122879, "global_step": 141358, "epoch": 1588, "lr": 4.6880725023392924e-05}
{"train_loss": 0.09860430657863617, "global_step": 141359, "epoch": 1588, "lr": 4.688014641625216e-05}
{"train_loss": 0.09418482333421707, "global_step": 141360, "epoch": 1588, "lr": 4.687956780953085e-05}
{"train_loss": 0.13267943263053894, "global_step": 141361, "epoch": 1588, "lr": 4.687898920322902e-05}
{"train_loss": 0.12134845554828644, "global_step": 141362, "epoch": 1588, "lr": 4.6878410597346786e-05}
{"train_loss": 0.1220180094242096, "global_step": 141363, "epoch": 1588, "lr": 4.6877831991884206e-05}
{"train_loss": 0.11619757115840912, "global_step": 141364, "epoch": 1588, "lr": 4.6877253386841354e-05}
{"train_loss": 0.14645910263061523, "global_step": 141365, "epoch": 1588, "lr": 4.687667478221833e-05}
{"train_loss": 0.11707447469234467, "global_step": 141366, "epoch": 1588, "lr": 4.687609617801519e-05}
{"train_loss": 0.11967051774263382, "global_step": 141367, "epoch": 1588, "lr": 4.6875517574232036e-05}
{"train_loss": 0.09745723754167557, "global_step": 141368, "epoch": 1588, "lr": 4.687493897086891e-05}
{"train_loss": 0.10912783443927765, "global_step": 141369, "epoch": 1588, "lr": 4.687436036792593e-05}
{"train_loss": 0.12849873304367065, "global_step": 141370, "epoch": 1588, "lr": 4.6873781765403135e-05}
{"train_loss": 0.10623679310083389, "global_step": 141371, "epoch": 1588, "lr": 4.6873203163300646e-05}
{"train_loss": 0.14879634976387024, "global_step": 141372, "epoch": 1588, "lr": 4.687262456161849e-05}
{"train_loss": 0.043344032019376755, "global_step": 141373, "epoch": 1588, "lr": 4.687204596035679e-05}
{"train_loss": 0.12180979549884796, "global_step": 141374, "epoch": 1588, "lr": 4.687146735951559e-05}
{"train_loss": 0.1230345219373703, "global_step": 141375, "epoch": 1588, "lr": 4.6870888759095e-05}
{"train_loss": 0.1252480000257492, "global_step": 141376, "epoch": 1588, "lr": 4.6870310159095066e-05}
{"train_loss": 0.0849526897072792, "global_step": 141377, "epoch": 1588, "lr": 4.68697315595159e-05}
{"train_loss": 0.12753388285636902, "global_step": 141378, "epoch": 1588, "lr": 4.6869152960357537e-05}
{"train_loss": 0.10064882040023804, "global_step": 141379, "epoch": 1588, "lr": 4.68685743616201e-05}
{"train_loss": 0.08913134783506393, "global_step": 141380, "epoch": 1588, "lr": 4.6867995763303624e-05}
{"train_loss": 0.15334245562553406, "global_step": 141381, "epoch": 1588, "lr": 4.6867417165408226e-05}
{"train_loss": 0.1388060599565506, "global_step": 141382, "epoch": 1588, "lr": 4.686683856793395e-05}
{"train_loss": 0.10882603377103806, "global_step": 141383, "epoch": 1588, "lr": 4.68662599708809e-05}
{"train_loss": 0.11615293473005295, "global_step": 141384, "epoch": 1588, "lr": 4.686568137424915e-05}
{"train_loss": 0.14111778140068054, "global_step": 141385, "epoch": 1588, "lr": 4.686510277803875e-05}
{"train_loss": 0.14033903181552887, "global_step": 141386, "epoch": 1588, "lr": 4.686452418224982e-05}
{"train_loss": 0.10662326216697693, "global_step": 141387, "epoch": 1588, "lr": 4.68639455868824e-05}
{"train_loss": 0.08073157072067261, "global_step": 141388, "epoch": 1588, "lr": 4.6863366991936604e-05}
{"train_loss": 0.09439825266599655, "global_step": 141389, "epoch": 1588, "lr": 4.686278839741246e-05}
{"train_loss": 0.0992625430226326, "global_step": 141390, "epoch": 1588, "lr": 4.686220980331011e-05}
{"train_loss": 0.1460246443748474, "global_step": 141391, "epoch": 1588, "lr": 4.6861631209629565e-05}
{"train_loss": 0.07362361997365952, "global_step": 141392, "epoch": 1588, "lr": 4.686105261637097e-05}
{"train_loss": 0.14806997776031494, "global_step": 141393, "epoch": 1588, "lr": 4.6860474023534335e-05}
{"train_loss": 0.1583229899406433, "global_step": 141394, "epoch": 1588, "lr": 4.6859895431119795e-05}
{"train_loss": 0.10013425350189209, "global_step": 141395, "epoch": 1588, "lr": 4.685931683912738e-05}
{"train_loss": 0.10314129292964935, "global_step": 141396, "epoch": 1588, "lr": 4.685873824755722e-05}
{"train_loss": 0.07124829292297363, "global_step": 141397, "epoch": 1588, "lr": 4.685815965640934e-05}
{"train_loss": 0.09059849381446838, "global_step": 141398, "epoch": 1588, "lr": 4.685758106568386e-05}
{"train_loss": 0.08900892734527588, "global_step": 141399, "epoch": 1588, "lr": 4.6857002475380824e-05}
{"train_loss": 0.10752295702695847, "global_step": 141400, "epoch": 1588, "lr": 4.6856423885500336e-05}
{"train_loss": 0.08388534933328629, "global_step": 141401, "epoch": 1588, "lr": 4.6855845296042456e-05}
{"train_loss": 0.10102251172065735, "global_step": 141402, "epoch": 1588, "lr": 4.685526670700727e-05}
{"train_loss": 0.09299256652593613, "global_step": 141403, "epoch": 1588, "lr": 4.6854688118394875e-05}
{"train_loss": 0.10895147919654846, "global_step": 141404, "epoch": 1588, "lr": 4.68541095302053e-05}
{"train_loss": 0.1497187614440918, "global_step": 141405, "epoch": 1588, "lr": 4.6853530942438676e-05}
{"train_loss": 0.1247173547744751, "global_step": 141406, "epoch": 1588, "lr": 4.685295235509504e-05}
{"train_loss": 0.11635411530733109, "global_step": 141407, "epoch": 1588, "lr": 4.6852373768174503e-05}
{"train_loss": 0.10687287151813507, "global_step": 141408, "epoch": 1588, "lr": 4.68517951816771e-05}
{"train_loss": 0.08650311082601547, "global_step": 141409, "epoch": 1588, "lr": 4.685121659560297e-05}
{"train_loss": 0.062230803072452545, "global_step": 141410, "epoch": 1588, "lr": 4.685063800995212e-05}
{"train_loss": 0.0592099167406559, "global_step": 141411, "epoch": 1588, "lr": 4.685005942472468e-05}
{"train_loss": 0.06148622930049896, "global_step": 141412, "epoch": 1588, "lr": 4.68494808399207e-05}
{"train_loss": 0.0886632651090622, "global_step": 141413, "epoch": 1588, "lr": 4.6848902255540296e-05}
{"train_loss": 0.12715691328048706, "global_step": 141414, "epoch": 1588, "lr": 4.684832367158349e-05}
{"train_loss": 0.1188303753733635, "global_step": 141415, "epoch": 1588, "lr": 4.684774508805041e-05}
{"train_loss": 0.1408236026763916, "global_step": 141416, "epoch": 1588, "lr": 4.684716650494109e-05}
{"train_loss": 0.09916206449270248, "global_step": 141417, "epoch": 1588, "lr": 4.6846587922255644e-05}
{"train_loss": 0.14634649455547333, "global_step": 141418, "epoch": 1588, "lr": 4.684600933999413e-05}
{"train_loss": 0.1392085701227188, "global_step": 141419, "epoch": 1588, "lr": 4.684543075815663e-05}
{"train_loss": 0.11484396260943305, "global_step": 141420, "epoch": 1588, "lr": 4.6844852176743234e-05, "val_loss": 6.114807605743408}
{"train_loss": 0.10684513300657272, "global_step": 141421, "epoch": 1589, "lr": 4.6844273595754004e-05}
{"train_loss": 0.11256829649209976, "global_step": 141422, "epoch": 1589, "lr": 4.6843695015189035e-05}
{"train_loss": 0.09827610105276108, "global_step": 141423, "epoch": 1589, "lr": 4.684311643504837e-05}
{"train_loss": 0.08866028487682343, "global_step": 141424, "epoch": 1589, "lr": 4.684253785533213e-05}
{"train_loss": 0.10927935689687729, "global_step": 141425, "epoch": 1589, "lr": 4.684195927604035e-05}
{"train_loss": 0.1270187944173813, "global_step": 141426, "epoch": 1589, "lr": 4.6841380697173155e-05}
{"train_loss": 0.10714995861053467, "global_step": 141427, "epoch": 1589, "lr": 4.684080211873057e-05}
{"train_loss": 0.10475354641675949, "global_step": 141428, "epoch": 1589, "lr": 4.684022354071273e-05}
{"train_loss": 0.20313005149364471, "global_step": 141429, "epoch": 1589, "lr": 4.683964496311966e-05}
{"train_loss": 0.11539247632026672, "global_step": 141430, "epoch": 1589, "lr": 4.683906638595148e-05}
{"train_loss": 0.1325698345899582, "global_step": 141431, "epoch": 1589, "lr": 4.683848780920823e-05}
{"train_loss": 0.14260628819465637, "global_step": 141432, "epoch": 1589, "lr": 4.683790923289003e-05}
{"train_loss": 0.061279553920030594, "global_step": 141433, "epoch": 1589, "lr": 4.6837330656996906e-05}
{"train_loss": 0.056197188794612885, "global_step": 141434, "epoch": 1589, "lr": 4.683675208152899e-05}
{"train_loss": 0.08761794865131378, "global_step": 141435, "epoch": 1589, "lr": 4.6836173506486315e-05}
{"train_loss": 0.13168878853321075, "global_step": 141436, "epoch": 1589, "lr": 4.683559493186898e-05}
{"train_loss": 0.1139446273446083, "global_step": 141437, "epoch": 1589, "lr": 4.683501635767707e-05}
{"train_loss": 0.13252057135105133, "global_step": 141438, "epoch": 1589, "lr": 4.683443778391065e-05}
{"train_loss": 0.12665876746177673, "global_step": 141439, "epoch": 1589, "lr": 4.68338592105698e-05}
{"train_loss": 0.0852450579404831, "global_step": 141440, "epoch": 1589, "lr": 4.6833280637654605e-05}
{"train_loss": 0.08760353177785873, "global_step": 141441, "epoch": 1589, "lr": 4.683270206516514e-05}
{"train_loss": 0.1328183114528656, "global_step": 141442, "epoch": 1589, "lr": 4.683212349310147e-05}
{"train_loss": 0.06644364446401596, "global_step": 141443, "epoch": 1589, "lr": 4.6831544921463685e-05}
{"train_loss": 0.11423158645629883, "global_step": 141444, "epoch": 1589, "lr": 4.683096635025185e-05}
{"train_loss": 0.1318848431110382, "global_step": 141445, "epoch": 1589, "lr": 4.683038777946608e-05}
{"train_loss": 0.09412927180528641, "global_step": 141446, "epoch": 1589, "lr": 4.6829809209106395e-05}
{"train_loss": 0.11374252289533615, "global_step": 141447, "epoch": 1589, "lr": 4.682923063917293e-05}
{"train_loss": 0.16179051995277405, "global_step": 141448, "epoch": 1589, "lr": 4.6828652069665705e-05}
{"train_loss": 0.09770011156797409, "global_step": 141449, "epoch": 1589, "lr": 4.682807350058486e-05}
{"train_loss": 0.11064186692237854, "global_step": 141450, "epoch": 1589, "lr": 4.682749493193042e-05}
{"train_loss": 0.14471030235290527, "global_step": 141451, "epoch": 1589, "lr": 4.6826916363702504e-05}
{"train_loss": 0.16549621522426605, "global_step": 141452, "epoch": 1589, "lr": 4.682633779590115e-05}
{"train_loss": 0.09816579520702362, "global_step": 141453, "epoch": 1589, "lr": 4.6825759228526476e-05}
{"train_loss": 0.06585416197776794, "global_step": 141454, "epoch": 1589, "lr": 4.6825180661578525e-05}
{"train_loss": 0.13204170763492584, "global_step": 141455, "epoch": 1589, "lr": 4.682460209505739e-05}
{"train_loss": 0.09488411247730255, "global_step": 141456, "epoch": 1589, "lr": 4.682402352896316e-05}
{"train_loss": 0.13084585964679718, "global_step": 141457, "epoch": 1589, "lr": 4.6823444963295893e-05}
{"train_loss": 0.13307440280914307, "global_step": 141458, "epoch": 1589, "lr": 4.6822866398055675e-05}
{"train_loss": 0.15126700699329376, "global_step": 141459, "epoch": 1589, "lr": 4.6822287833242585e-05}
{"train_loss": 0.11296670138835907, "global_step": 141460, "epoch": 1589, "lr": 4.682170926885671e-05}
{"train_loss": 0.11426381021738052, "global_step": 141461, "epoch": 1589, "lr": 4.6821130704898095e-05}
{"train_loss": 0.0822179764509201, "global_step": 141462, "epoch": 1589, "lr": 4.682055214136687e-05}
{"train_loss": 0.14616930484771729, "global_step": 141463, "epoch": 1589, "lr": 4.681997357826305e-05}
{"train_loss": 0.1119241863489151, "global_step": 141464, "epoch": 1589, "lr": 4.6819395015586776e-05}
{"train_loss": 0.08518476039171219, "global_step": 141465, "epoch": 1589, "lr": 4.6818816453338075e-05}
{"train_loss": 0.10281749069690704, "global_step": 141466, "epoch": 1589, "lr": 4.681823789151706e-05}
{"train_loss": 0.09768268465995789, "global_step": 141467, "epoch": 1589, "lr": 4.681765933012378e-05}
{"train_loss": 0.07014402747154236, "global_step": 141468, "epoch": 1589, "lr": 4.681708076915834e-05}
{"train_loss": 0.13763105869293213, "global_step": 141469, "epoch": 1589, "lr": 4.6816502208620786e-05}
{"train_loss": 0.09063328802585602, "global_step": 141470, "epoch": 1589, "lr": 4.6815923648511236e-05}
{"train_loss": 0.07901489734649658, "global_step": 141471, "epoch": 1589, "lr": 4.681534508882973e-05}
{"train_loss": 0.09483562409877777, "global_step": 141472, "epoch": 1589, "lr": 4.6814766529576354e-05}
{"train_loss": 0.11001574993133545, "global_step": 141473, "epoch": 1589, "lr": 4.6814187970751224e-05}
{"train_loss": 0.08793725073337555, "global_step": 141474, "epoch": 1589, "lr": 4.681360941235437e-05}
{"train_loss": 0.09438496828079224, "global_step": 141475, "epoch": 1589, "lr": 4.681303085438589e-05}
{"train_loss": 0.15588785707950592, "global_step": 141476, "epoch": 1589, "lr": 4.6812452296845856e-05}
{"train_loss": 0.07323838770389557, "global_step": 141477, "epoch": 1589, "lr": 4.6811873739734355e-05}
{"train_loss": 0.08442773669958115, "global_step": 141478, "epoch": 1589, "lr": 4.681129518305145e-05}
{"train_loss": 0.12259580940008163, "global_step": 141479, "epoch": 1589, "lr": 4.681071662679724e-05}
{"train_loss": 0.10420706868171692, "global_step": 141480, "epoch": 1589, "lr": 4.681013807097178e-05}
{"train_loss": 0.07084286212921143, "global_step": 141481, "epoch": 1589, "lr": 4.6809559515575166e-05}
{"train_loss": 0.09495341777801514, "global_step": 141482, "epoch": 1589, "lr": 4.680898096060746e-05}
{"train_loss": 0.1765376478433609, "global_step": 141483, "epoch": 1589, "lr": 4.6808402406068756e-05}
{"train_loss": 0.1133534237742424, "global_step": 141484, "epoch": 1589, "lr": 4.680782385195911e-05}
{"train_loss": 0.08105317503213882, "global_step": 141485, "epoch": 1589, "lr": 4.680724529827863e-05}
{"train_loss": 0.12493837624788284, "global_step": 141486, "epoch": 1589, "lr": 4.680666674502736e-05}
{"train_loss": 0.09703141450881958, "global_step": 141487, "epoch": 1589, "lr": 4.680608819220541e-05}
{"train_loss": 0.09564337134361267, "global_step": 141488, "epoch": 1589, "lr": 4.680550963981283e-05}
{"train_loss": 0.14689819514751434, "global_step": 141489, "epoch": 1589, "lr": 4.6804931087849725e-05}
{"train_loss": 0.14018826186656952, "global_step": 141490, "epoch": 1589, "lr": 4.680435253631613e-05}
{"train_loss": 0.08845020085573196, "global_step": 141491, "epoch": 1589, "lr": 4.680377398521216e-05}
{"train_loss": 0.13380718231201172, "global_step": 141492, "epoch": 1589, "lr": 4.68031954345379e-05}
{"train_loss": 0.17390945553779602, "global_step": 141493, "epoch": 1589, "lr": 4.68026168842934e-05}
{"train_loss": 0.12219151854515076, "global_step": 141494, "epoch": 1589, "lr": 4.680203833447876e-05}
{"train_loss": 0.10707680135965347, "global_step": 141495, "epoch": 1589, "lr": 4.6801459785094046e-05}
{"train_loss": 0.10396627336740494, "global_step": 141496, "epoch": 1589, "lr": 4.680088123613933e-05}
{"train_loss": 0.1693994402885437, "global_step": 141497, "epoch": 1589, "lr": 4.68003026876147e-05}
{"train_loss": 0.097507543861866, "global_step": 141498, "epoch": 1589, "lr": 4.679972413952023e-05}
{"train_loss": 0.12707993388175964, "global_step": 141499, "epoch": 1589, "lr": 4.6799145591856e-05}
{"train_loss": 0.10057534277439117, "global_step": 141500, "epoch": 1589, "lr": 4.679856704462209e-05}
{"train_loss": 0.14816933870315552, "global_step": 141501, "epoch": 1589, "lr": 4.679798849781855e-05}
{"train_loss": 0.12820273637771606, "global_step": 141502, "epoch": 1589, "lr": 4.679740995144551e-05}
{"train_loss": 0.11688552796840668, "global_step": 141503, "epoch": 1589, "lr": 4.6796831405502994e-05}
{"train_loss": 0.13665828108787537, "global_step": 141504, "epoch": 1589, "lr": 4.679625285999113e-05}
{"train_loss": 0.16002540290355682, "global_step": 141505, "epoch": 1589, "lr": 4.679567431490995e-05}
{"train_loss": 0.08055010437965393, "global_step": 141506, "epoch": 1589, "lr": 4.679509577025957e-05}
{"train_loss": 0.21239344775676727, "global_step": 141507, "epoch": 1589, "lr": 4.679451722604003e-05}
{"train_loss": 0.07206905633211136, "global_step": 141508, "epoch": 1589, "lr": 4.6793938682251425e-05}
{"train_loss": 0.11454924811305625, "global_step": 141509, "epoch": 1589, "lr": 4.679336013889386e-05, "val_loss": 6.331664085388184}
{"train_loss": 0.08074773848056793, "global_step": 141510, "epoch": 1590, "lr": 4.679278159596737e-05}
{"train_loss": 0.1454187035560608, "global_step": 141511, "epoch": 1590, "lr": 4.679220305347207e-05}
{"train_loss": 0.06523016095161438, "global_step": 141512, "epoch": 1590, "lr": 4.6791624511408e-05}
{"train_loss": 0.12283863127231598, "global_step": 141513, "epoch": 1590, "lr": 4.6791045969775275e-05}
{"train_loss": 0.13523761928081512, "global_step": 141514, "epoch": 1590, "lr": 4.6790467428573946e-05}
{"train_loss": 0.07097049802541733, "global_step": 141515, "epoch": 1590, "lr": 4.6789888887804104e-05}
{"train_loss": 0.16380801796913147, "global_step": 141516, "epoch": 1590, "lr": 4.678931034746582e-05}
{"train_loss": 0.07493320852518082, "global_step": 141517, "epoch": 1590, "lr": 4.678873180755917e-05}
{"train_loss": 0.13180094957351685, "global_step": 141518, "epoch": 1590, "lr": 4.678815326808423e-05}
{"train_loss": 0.11721561849117279, "global_step": 141519, "epoch": 1590, "lr": 4.678757472904111e-05}
{"train_loss": 0.1316332370042801, "global_step": 141520, "epoch": 1590, "lr": 4.678699619042983e-05}
{"train_loss": 0.12471942603588104, "global_step": 141521, "epoch": 1590, "lr": 4.678641765225053e-05}
{"train_loss": 0.1056215688586235, "global_step": 141522, "epoch": 1590, "lr": 4.6785839114503224e-05}
{"train_loss": 0.1061370000243187, "global_step": 141523, "epoch": 1590, "lr": 4.678526057718805e-05}
{"train_loss": 0.1362801194190979, "global_step": 141524, "epoch": 1590, "lr": 4.6784682040305036e-05}
{"train_loss": 0.11223170161247253, "global_step": 141525, "epoch": 1590, "lr": 4.6784103503854296e-05}
{"train_loss": 0.09233850240707397, "global_step": 141526, "epoch": 1590, "lr": 4.678352496783589e-05}
{"train_loss": 0.1408129632472992, "global_step": 141527, "epoch": 1590, "lr": 4.6782946432249886e-05}
{"train_loss": 0.08750072866678238, "global_step": 141528, "epoch": 1590, "lr": 4.67823678970964e-05}
{"train_loss": 0.08740661293268204, "global_step": 141529, "epoch": 1590, "lr": 4.6781789362375464e-05}
{"train_loss": 0.12812386453151703, "global_step": 141530, "epoch": 1590, "lr": 4.6781210828087205e-05}
{"train_loss": 0.14314353466033936, "global_step": 141531, "epoch": 1590, "lr": 4.6780632294231644e-05}
{"train_loss": 0.13280561566352844, "global_step": 141532, "epoch": 1590, "lr": 4.6780053760808904e-05}
{"train_loss": 0.06148901581764221, "global_step": 141533, "epoch": 1590, "lr": 4.6779475227819045e-05}
{"train_loss": 0.07375036925077438, "global_step": 141534, "epoch": 1590, "lr": 4.677889669526215e-05}
{"train_loss": 0.10732394456863403, "global_step": 141535, "epoch": 1590, "lr": 4.677831816313828e-05}
{"train_loss": 0.10919291526079178, "global_step": 141536, "epoch": 1590, "lr": 4.6777739631447535e-05}
{"train_loss": 0.06943529844284058, "global_step": 141537, "epoch": 1590, "lr": 4.677716110018999e-05}
{"train_loss": 0.1327819675207138, "global_step": 141538, "epoch": 1590, "lr": 4.6776582569365714e-05}
{"train_loss": 0.13441292941570282, "global_step": 141539, "epoch": 1590, "lr": 4.677600403897477e-05}
{"train_loss": 0.10716105252504349, "global_step": 141540, "epoch": 1590, "lr": 4.677542550901728e-05}
{"train_loss": 0.11109773814678192, "global_step": 141541, "epoch": 1590, "lr": 4.6774846979493265e-05}
{"train_loss": 0.17648246884346008, "global_step": 141542, "epoch": 1590, "lr": 4.677426845040286e-05}
{"train_loss": 0.1279616802930832, "global_step": 141543, "epoch": 1590, "lr": 4.677368992174609e-05}
{"train_loss": 0.13799349963665009, "global_step": 141544, "epoch": 1590, "lr": 4.6773111393523055e-05}
{"train_loss": 0.08433915674686432, "global_step": 141545, "epoch": 1590, "lr": 4.677253286573386e-05}
{"train_loss": 0.09554782509803772, "global_step": 141546, "epoch": 1590, "lr": 4.677195433837854e-05}
{"train_loss": 0.13043737411499023, "global_step": 141547, "epoch": 1590, "lr": 4.677137581145722e-05}
{"train_loss": 0.07204391807317734, "global_step": 141548, "epoch": 1590, "lr": 4.677079728496991e-05}
{"train_loss": 0.15913711488246918, "global_step": 141549, "epoch": 1590, "lr": 4.6770218758916764e-05}
{"train_loss": 0.0650140568614006, "global_step": 141550, "epoch": 1590, "lr": 4.676964023329779e-05}
{"train_loss": 0.12316829711198807, "global_step": 141551, "epoch": 1590, "lr": 4.6769061708113126e-05}
{"train_loss": 0.11133527755737305, "global_step": 141552, "epoch": 1590, "lr": 4.676848318336282e-05}
{"train_loss": 0.054885994642972946, "global_step": 141553, "epoch": 1590, "lr": 4.6767904659046935e-05}
{"train_loss": 0.1472705751657486, "global_step": 141554, "epoch": 1590, "lr": 4.6767326135165574e-05}
{"train_loss": 0.11014808714389801, "global_step": 141555, "epoch": 1590, "lr": 4.676674761171881e-05}
{"train_loss": 0.07695712149143219, "global_step": 141556, "epoch": 1590, "lr": 4.676616908870671e-05}
{"train_loss": 0.10468999296426773, "global_step": 141557, "epoch": 1590, "lr": 4.6765590566129376e-05}
{"train_loss": 0.14101730287075043, "global_step": 141558, "epoch": 1590, "lr": 4.6765012043986844e-05}
{"train_loss": 0.1060718223452568, "global_step": 141559, "epoch": 1590, "lr": 4.676443352227924e-05}
{"train_loss": 0.07441670447587967, "global_step": 141560, "epoch": 1590, "lr": 4.67638550010066e-05}
{"train_loss": 0.17216241359710693, "global_step": 141561, "epoch": 1590, "lr": 4.676327648016903e-05}
{"train_loss": 0.09124878793954849, "global_step": 141562, "epoch": 1590, "lr": 4.6762697959766585e-05}
{"train_loss": 0.10688459873199463, "global_step": 141563, "epoch": 1590, "lr": 4.676211943979936e-05}
{"train_loss": 0.10018343478441238, "global_step": 141564, "epoch": 1590, "lr": 4.676154092026744e-05}
{"train_loss": 0.06988495588302612, "global_step": 141565, "epoch": 1590, "lr": 4.6760962401170875e-05}
{"train_loss": 0.09432923793792725, "global_step": 141566, "epoch": 1590, "lr": 4.6760383882509784e-05}
{"train_loss": 0.11106398701667786, "global_step": 141567, "epoch": 1590, "lr": 4.675980536428419e-05}
{"train_loss": 0.12988048791885376, "global_step": 141568, "epoch": 1590, "lr": 4.675922684649423e-05}
{"train_loss": 0.07232275605201721, "global_step": 141569, "epoch": 1590, "lr": 4.675864832913993e-05}
{"train_loss": 0.12808184325695038, "global_step": 141570, "epoch": 1590, "lr": 4.675806981222141e-05}
{"train_loss": 0.0651455819606781, "global_step": 141571, "epoch": 1590, "lr": 4.675749129573872e-05}
{"train_loss": 0.08054160326719284, "global_step": 141572, "epoch": 1590, "lr": 4.675691277969194e-05}
{"train_loss": 0.06209004670381546, "global_step": 141573, "epoch": 1590, "lr": 4.6756334264081154e-05}
{"train_loss": 0.10810615867376328, "global_step": 141574, "epoch": 1590, "lr": 4.6755755748906444e-05}
{"train_loss": 0.1643732637166977, "global_step": 141575, "epoch": 1590, "lr": 4.675517723416788e-05}
{"train_loss": 0.11655425280332565, "global_step": 141576, "epoch": 1590, "lr": 4.6754598719865556e-05}
{"train_loss": 0.12790560722351074, "global_step": 141577, "epoch": 1590, "lr": 4.675402020599951e-05}
{"train_loss": 0.12183555215597153, "global_step": 141578, "epoch": 1590, "lr": 4.675344169256988e-05}
{"train_loss": 0.17654424905776978, "global_step": 141579, "epoch": 1590, "lr": 4.675286317957667e-05}
{"train_loss": 0.14189079403877258, "global_step": 141580, "epoch": 1590, "lr": 4.6752284667020034e-05}
{"train_loss": 0.10530881583690643, "global_step": 141581, "epoch": 1590, "lr": 4.675170615489999e-05}
{"train_loss": 0.17515143752098083, "global_step": 141582, "epoch": 1590, "lr": 4.6751127643216635e-05}
{"train_loss": 0.1393379271030426, "global_step": 141583, "epoch": 1590, "lr": 4.675054913197008e-05}
{"train_loss": 0.07867792993783951, "global_step": 141584, "epoch": 1590, "lr": 4.6749970621160346e-05}
{"train_loss": 0.12598808109760284, "global_step": 141585, "epoch": 1590, "lr": 4.674939211078756e-05}
{"train_loss": 0.10634973645210266, "global_step": 141586, "epoch": 1590, "lr": 4.6748813600851757e-05}
{"train_loss": 0.1226641982793808, "global_step": 141587, "epoch": 1590, "lr": 4.674823509135306e-05}
{"train_loss": 0.12452132254838943, "global_step": 141588, "epoch": 1590, "lr": 4.6747656582291505e-05}
{"train_loss": 0.11014916747808456, "global_step": 141589, "epoch": 1590, "lr": 4.6747078073667206e-05}
{"train_loss": 0.07214538007974625, "global_step": 141590, "epoch": 1590, "lr": 4.674649956548021e-05}
{"train_loss": 0.09169375151395798, "global_step": 141591, "epoch": 1590, "lr": 4.6745921057730615e-05}
{"train_loss": 0.0808686912059784, "global_step": 141592, "epoch": 1590, "lr": 4.674534255041848e-05}
{"train_loss": 0.09658592939376831, "global_step": 141593, "epoch": 1590, "lr": 4.6744764043543905e-05}
{"train_loss": 0.14949633181095123, "global_step": 141594, "epoch": 1590, "lr": 4.674418553710695e-05}
{"train_loss": 0.1380872279405594, "global_step": 141595, "epoch": 1590, "lr": 4.67436070311077e-05}
{"train_loss": 0.062220677733421326, "global_step": 141596, "epoch": 1590, "lr": 4.674302852554623e-05}
{"train_loss": 0.1421252191066742, "global_step": 141597, "epoch": 1590, "lr": 4.674245002042264e-05}
{"train_loss": 0.11120520929774541, "global_step": 141598, "epoch": 1590, "lr": 4.6741871515736957e-05, "val_loss": 6.123825550079346, "train_action_mse_error": 27.213787078857422}
{"train_loss": 0.10307545214891434, "global_step": 141599, "epoch": 1591, "lr": 4.67412930114893e-05}
{"train_loss": 0.09514431655406952, "global_step": 141600, "epoch": 1591, "lr": 4.674071450767975e-05}
{"train_loss": 0.18240836262702942, "global_step": 141601, "epoch": 1591, "lr": 4.674013600430835e-05}
{"train_loss": 0.08982263505458832, "global_step": 141602, "epoch": 1591, "lr": 4.673955750137522e-05}
{"train_loss": 0.16396726667881012, "global_step": 141603, "epoch": 1591, "lr": 4.67389789988804e-05}
{"train_loss": 0.08763236552476883, "global_step": 141604, "epoch": 1591, "lr": 4.6738400496824004e-05}
{"train_loss": 0.08544953167438507, "global_step": 141605, "epoch": 1591, "lr": 4.673782199520606e-05}
{"train_loss": 0.14281746745109558, "global_step": 141606, "epoch": 1591, "lr": 4.673724349402671e-05}
{"train_loss": 0.09348562359809875, "global_step": 141607, "epoch": 1591, "lr": 4.673666499328597e-05}
{"train_loss": 0.0978628471493721, "global_step": 141608, "epoch": 1591, "lr": 4.673608649298397e-05}
{"train_loss": 0.1047031506896019, "global_step": 141609, "epoch": 1591, "lr": 4.673550799312074e-05}
{"train_loss": 0.11915667355060577, "global_step": 141610, "epoch": 1591, "lr": 4.6734929493696404e-05}
{"train_loss": 0.0784948319196701, "global_step": 141611, "epoch": 1591, "lr": 4.673435099471101e-05}
{"train_loss": 0.10312134027481079, "global_step": 141612, "epoch": 1591, "lr": 4.673377249616463e-05}
{"train_loss": 0.10943639278411865, "global_step": 141613, "epoch": 1591, "lr": 4.6733193998057366e-05}
{"train_loss": 0.08682674914598465, "global_step": 141614, "epoch": 1591, "lr": 4.673261550038928e-05}
{"train_loss": 0.1387862265110016, "global_step": 141615, "epoch": 1591, "lr": 4.6732037003160456e-05}
{"train_loss": 0.10395935922861099, "global_step": 141616, "epoch": 1591, "lr": 4.673145850637098e-05}
{"train_loss": 0.08859699964523315, "global_step": 141617, "epoch": 1591, "lr": 4.6730880010020897e-05}
{"train_loss": 0.09192082285881042, "global_step": 141618, "epoch": 1591, "lr": 4.673030151411031e-05}
{"train_loss": 0.12039396911859512, "global_step": 141619, "epoch": 1591, "lr": 4.6729723018639316e-05}
{"train_loss": 0.10607293993234634, "global_step": 141620, "epoch": 1591, "lr": 4.6729144523607945e-05}
{"train_loss": 0.10241103917360306, "global_step": 141621, "epoch": 1591, "lr": 4.6728566029016325e-05}
{"train_loss": 0.0771254152059555, "global_step": 141622, "epoch": 1591, "lr": 4.6727987534864486e-05}
{"train_loss": 0.11080751568078995, "global_step": 141623, "epoch": 1591, "lr": 4.6727409041152554e-05}
{"train_loss": 0.11893279105424881, "global_step": 141624, "epoch": 1591, "lr": 4.672683054788056e-05}
{"train_loss": 0.1004946306347847, "global_step": 141625, "epoch": 1591, "lr": 4.672625205504862e-05}
{"train_loss": 0.09892306476831436, "global_step": 141626, "epoch": 1591, "lr": 4.672567356265678e-05}
{"train_loss": 0.1215144544839859, "global_step": 141627, "epoch": 1591, "lr": 4.672509507070515e-05}
{"train_loss": 0.07465749979019165, "global_step": 141628, "epoch": 1591, "lr": 4.672451657919377e-05}
{"train_loss": 0.1211690679192543, "global_step": 141629, "epoch": 1591, "lr": 4.6723938088122766e-05}
{"train_loss": 0.10389280319213867, "global_step": 141630, "epoch": 1591, "lr": 4.672335959749217e-05}
{"train_loss": 0.08493399620056152, "global_step": 141631, "epoch": 1591, "lr": 4.672278110730208e-05}
{"train_loss": 0.09977101534605026, "global_step": 141632, "epoch": 1591, "lr": 4.672220261755258e-05}
{"train_loss": 0.04115630313754082, "global_step": 141633, "epoch": 1591, "lr": 4.672162412824374e-05}
{"train_loss": 0.14624446630477905, "global_step": 141634, "epoch": 1591, "lr": 4.672104563937563e-05}
{"train_loss": 0.07524305582046509, "global_step": 141635, "epoch": 1591, "lr": 4.672046715094832e-05}
{"train_loss": 0.1474982649087906, "global_step": 141636, "epoch": 1591, "lr": 4.6719888662961934e-05}
{"train_loss": 0.0718061774969101, "global_step": 141637, "epoch": 1591, "lr": 4.67193101754165e-05}
{"train_loss": 0.1128130778670311, "global_step": 141638, "epoch": 1591, "lr": 4.671873168831213e-05}
{"train_loss": 0.12251883000135422, "global_step": 141639, "epoch": 1591, "lr": 4.671815320164886e-05}
{"train_loss": 0.11614400148391724, "global_step": 141640, "epoch": 1591, "lr": 4.6717574715426826e-05}
{"train_loss": 0.13142932951450348, "global_step": 141641, "epoch": 1591, "lr": 4.671699622964605e-05}
{"train_loss": 0.055478088557720184, "global_step": 141642, "epoch": 1591, "lr": 4.6716417744306654e-05}
{"train_loss": 0.1102914810180664, "global_step": 141643, "epoch": 1591, "lr": 4.671583925940867e-05}
{"train_loss": 0.15766113996505737, "global_step": 141644, "epoch": 1591, "lr": 4.671526077495223e-05}
{"train_loss": 0.13140469789505005, "global_step": 141645, "epoch": 1591, "lr": 4.671468229093736e-05}
{"train_loss": 0.15094618499279022, "global_step": 141646, "epoch": 1591, "lr": 4.671410380736418e-05}
{"train_loss": 0.10533903539180756, "global_step": 141647, "epoch": 1591, "lr": 4.671352532423273e-05}
{"train_loss": 0.148602694272995, "global_step": 141648, "epoch": 1591, "lr": 4.671294684154312e-05}
{"train_loss": 0.09041395038366318, "global_step": 141649, "epoch": 1591, "lr": 4.6712368359295414e-05}
{"train_loss": 0.08022958040237427, "global_step": 141650, "epoch": 1591, "lr": 4.671178987748969e-05}
{"train_loss": 0.12154518067836761, "global_step": 141651, "epoch": 1591, "lr": 4.6711211396126014e-05}
{"train_loss": 0.10650010406970978, "global_step": 141652, "epoch": 1591, "lr": 4.6710632915204484e-05}
{"train_loss": 0.078304223716259, "global_step": 141653, "epoch": 1591, "lr": 4.671005443472517e-05}
{"train_loss": 0.1327577829360962, "global_step": 141654, "epoch": 1591, "lr": 4.670947595468813e-05}
{"train_loss": 0.12145491689443588, "global_step": 141655, "epoch": 1591, "lr": 4.67088974750935e-05}
{"train_loss": 0.08342285454273224, "global_step": 141656, "epoch": 1591, "lr": 4.670831899594128e-05}
{"train_loss": 0.1133762076497078, "global_step": 141657, "epoch": 1591, "lr": 4.670774051723161e-05}
{"train_loss": 0.10245170444250107, "global_step": 141658, "epoch": 1591, "lr": 4.6707162038964534e-05}
{"train_loss": 0.0648583173751831, "global_step": 141659, "epoch": 1591, "lr": 4.670658356114015e-05}
{"train_loss": 0.1330692619085312, "global_step": 141660, "epoch": 1591, "lr": 4.670600508375851e-05}
{"train_loss": 0.19132311642169952, "global_step": 141661, "epoch": 1591, "lr": 4.670542660681972e-05}
{"train_loss": 0.09803034365177155, "global_step": 141662, "epoch": 1591, "lr": 4.670484813032384e-05}
{"train_loss": 0.11611280590295792, "global_step": 141663, "epoch": 1591, "lr": 4.6704269654270964e-05}
{"train_loss": 0.09312435239553452, "global_step": 141664, "epoch": 1591, "lr": 4.670369117866114e-05}
{"train_loss": 0.06607019156217575, "global_step": 141665, "epoch": 1591, "lr": 4.670311270349448e-05}
{"train_loss": 0.12349146604537964, "global_step": 141666, "epoch": 1591, "lr": 4.670253422877103e-05}
{"train_loss": 0.09929326176643372, "global_step": 141667, "epoch": 1591, "lr": 4.670195575449091e-05}
{"train_loss": 0.08862681686878204, "global_step": 141668, "epoch": 1591, "lr": 4.670137728065416e-05}
{"train_loss": 0.09769617021083832, "global_step": 141669, "epoch": 1591, "lr": 4.670079880726086e-05}
{"train_loss": 0.15906573832035065, "global_step": 141670, "epoch": 1591, "lr": 4.670022033431111e-05}
{"train_loss": 0.11024011671543121, "global_step": 141671, "epoch": 1591, "lr": 4.669964186180495e-05}
{"train_loss": 0.10623560100793839, "global_step": 141672, "epoch": 1591, "lr": 4.669906338974251e-05}
{"train_loss": 0.15167848765850067, "global_step": 141673, "epoch": 1591, "lr": 4.669848491812382e-05}
{"train_loss": 0.09764500707387924, "global_step": 141674, "epoch": 1591, "lr": 4.6697906446949005e-05}
{"train_loss": 0.13148152828216553, "global_step": 141675, "epoch": 1591, "lr": 4.6697327976218094e-05}
{"train_loss": 0.18621720373630524, "global_step": 141676, "epoch": 1591, "lr": 4.6696749505931204e-05}
{"train_loss": 0.14456841349601746, "global_step": 141677, "epoch": 1591, "lr": 4.669617103608838e-05}
{"train_loss": 0.12426205724477768, "global_step": 141678, "epoch": 1591, "lr": 4.6695592566689726e-05}
{"train_loss": 0.07243045419454575, "global_step": 141679, "epoch": 1591, "lr": 4.66950140977353e-05}
{"train_loss": 0.13812729716300964, "global_step": 141680, "epoch": 1591, "lr": 4.66944356292252e-05}
{"train_loss": 0.156717449426651, "global_step": 141681, "epoch": 1591, "lr": 4.669385716115948e-05}
{"train_loss": 0.16285011172294617, "global_step": 141682, "epoch": 1591, "lr": 4.669327869353825e-05}
{"train_loss": 0.09316658973693848, "global_step": 141683, "epoch": 1591, "lr": 4.669270022636155e-05}
{"train_loss": 0.1891726702451706, "global_step": 141684, "epoch": 1591, "lr": 4.6692121759629485e-05}
{"train_loss": 0.13831502199172974, "global_step": 141685, "epoch": 1591, "lr": 4.669154329334211e-05}
{"train_loss": 0.08151665329933167, "global_step": 141686, "epoch": 1591, "lr": 4.6690964827499546e-05}
{"train_loss": 0.11163298848472285, "global_step": 141687, "epoch": 1591, "lr": 4.669038636210182e-05, "val_loss": 5.993581295013428}
{"train_loss": 0.06798495352268219, "global_step": 141688, "epoch": 1592, "lr": 4.6689807897149035e-05}
{"train_loss": 0.08534468710422516, "global_step": 141689, "epoch": 1592, "lr": 4.668922943264126e-05}
{"train_loss": 0.09864148497581482, "global_step": 141690, "epoch": 1592, "lr": 4.668865096857857e-05}
{"train_loss": 0.13750429451465607, "global_step": 141691, "epoch": 1592, "lr": 4.668807250496108e-05}
{"train_loss": 0.08879933506250381, "global_step": 141692, "epoch": 1592, "lr": 4.668749404178881e-05}
{"train_loss": 0.14222244918346405, "global_step": 141693, "epoch": 1592, "lr": 4.6686915579061894e-05}
{"train_loss": 0.11911933124065399, "global_step": 141694, "epoch": 1592, "lr": 4.6686337116780346e-05}
{"train_loss": 0.1093311458826065, "global_step": 141695, "epoch": 1592, "lr": 4.668575865494431e-05}
{"train_loss": 0.22209475934505463, "global_step": 141696, "epoch": 1592, "lr": 4.6685180193553815e-05}
{"train_loss": 0.0917353481054306, "global_step": 141697, "epoch": 1592, "lr": 4.6684601732608974e-05}
{"train_loss": 0.09286995232105255, "global_step": 141698, "epoch": 1592, "lr": 4.668402327210983e-05}
{"train_loss": 0.0870388075709343, "global_step": 141699, "epoch": 1592, "lr": 4.668344481205649e-05}
{"train_loss": 0.09235949069261551, "global_step": 141700, "epoch": 1592, "lr": 4.6682866352449014e-05}
{"train_loss": 0.0959552451968193, "global_step": 141701, "epoch": 1592, "lr": 4.6682287893287495e-05}
{"train_loss": 0.09003252536058426, "global_step": 141702, "epoch": 1592, "lr": 4.6681709434571983e-05}
{"train_loss": 0.1818258911371231, "global_step": 141703, "epoch": 1592, "lr": 4.668113097630261e-05}
{"train_loss": 0.1798473298549652, "global_step": 141704, "epoch": 1592, "lr": 4.668055251847938e-05}
{"train_loss": 0.16148877143859863, "global_step": 141705, "epoch": 1592, "lr": 4.667997406110244e-05}
{"train_loss": 0.09354697167873383, "global_step": 141706, "epoch": 1592, "lr": 4.6679395604171815e-05}
{"train_loss": 0.10558146983385086, "global_step": 141707, "epoch": 1592, "lr": 4.667881714768761e-05}
{"train_loss": 0.1369040608406067, "global_step": 141708, "epoch": 1592, "lr": 4.6678238691649903e-05}
{"train_loss": 0.06470607966184616, "global_step": 141709, "epoch": 1592, "lr": 4.667766023605876e-05}
{"train_loss": 0.08105674386024475, "global_step": 141710, "epoch": 1592, "lr": 4.6677081780914276e-05}
{"train_loss": 0.111202672123909, "global_step": 141711, "epoch": 1592, "lr": 4.66765033262165e-05}
{"train_loss": 0.14813008904457092, "global_step": 141712, "epoch": 1592, "lr": 4.667592487196555e-05}
{"train_loss": 0.10945624858140945, "global_step": 141713, "epoch": 1592, "lr": 4.667534641816146e-05}
{"train_loss": 0.09111785143613815, "global_step": 141714, "epoch": 1592, "lr": 4.6674767964804346e-05}
{"train_loss": 0.15628746151924133, "global_step": 141715, "epoch": 1592, "lr": 4.667418951189425e-05}
{"train_loss": 0.13328859210014343, "global_step": 141716, "epoch": 1592, "lr": 4.667361105943129e-05}
{"train_loss": 0.2578166127204895, "global_step": 141717, "epoch": 1592, "lr": 4.66730326074155e-05}
{"train_loss": 0.07073880732059479, "global_step": 141718, "epoch": 1592, "lr": 4.6672454155847006e-05}
{"train_loss": 0.08190344274044037, "global_step": 141719, "epoch": 1592, "lr": 4.6671875704725836e-05}
{"train_loss": 0.13807111978530884, "global_step": 141720, "epoch": 1592, "lr": 4.6671297254052117e-05}
{"train_loss": 0.12826870381832123, "global_step": 141721, "epoch": 1592, "lr": 4.6670718803825874e-05}
{"train_loss": 0.0871749296784401, "global_step": 141722, "epoch": 1592, "lr": 4.667014035404724e-05}
{"train_loss": 0.11416812241077423, "global_step": 141723, "epoch": 1592, "lr": 4.6669561904716234e-05}
{"train_loss": 0.10446581244468689, "global_step": 141724, "epoch": 1592, "lr": 4.666898345583299e-05}
{"train_loss": 0.07390772551298141, "global_step": 141725, "epoch": 1592, "lr": 4.666840500739754e-05}
{"train_loss": 0.09542533755302429, "global_step": 141726, "epoch": 1592, "lr": 4.666782655941e-05}
{"train_loss": 0.0553869791328907, "global_step": 141727, "epoch": 1592, "lr": 4.666724811187043e-05}
{"train_loss": 0.08516199141740799, "global_step": 141728, "epoch": 1592, "lr": 4.6666669664778894e-05}
{"train_loss": 0.08844293653964996, "global_step": 141729, "epoch": 1592, "lr": 4.6666091218135506e-05}
{"train_loss": 0.10687164962291718, "global_step": 141730, "epoch": 1592, "lr": 4.6665512771940294e-05}
{"train_loss": 0.10873029381036758, "global_step": 141731, "epoch": 1592, "lr": 4.6664934326193385e-05}
{"train_loss": 0.156112939119339, "global_step": 141732, "epoch": 1592, "lr": 4.6664355880894814e-05}
{"train_loss": 0.07709653675556183, "global_step": 141733, "epoch": 1592, "lr": 4.666377743604471e-05}
{"train_loss": 0.13250799477100372, "global_step": 141734, "epoch": 1592, "lr": 4.6663198991643086e-05}
{"train_loss": 0.07874918729066849, "global_step": 141735, "epoch": 1592, "lr": 4.6662620547690085e-05}
{"train_loss": 0.0872107595205307, "global_step": 141736, "epoch": 1592, "lr": 4.6662042104185726e-05}
{"train_loss": 0.13875225186347961, "global_step": 141737, "epoch": 1592, "lr": 4.666146366113014e-05}
{"train_loss": 0.09885220974683762, "global_step": 141738, "epoch": 1592, "lr": 4.666088521852336e-05}
{"train_loss": 0.15970753133296967, "global_step": 141739, "epoch": 1592, "lr": 4.6660306776365504e-05}
{"train_loss": 0.09289946407079697, "global_step": 141740, "epoch": 1592, "lr": 4.66597283346566e-05}
{"train_loss": 0.12420576810836792, "global_step": 141741, "epoch": 1592, "lr": 4.665914989339678e-05}
{"train_loss": 0.20485959947109222, "global_step": 141742, "epoch": 1592, "lr": 4.6658571452586083e-05}
{"train_loss": 0.10143503546714783, "global_step": 141743, "epoch": 1592, "lr": 4.665799301222462e-05}
{"train_loss": 0.07423135638237, "global_step": 141744, "epoch": 1592, "lr": 4.665741457231242e-05}
{"train_loss": 0.14148850739002228, "global_step": 141745, "epoch": 1592, "lr": 4.6656836132849605e-05}
{"train_loss": 0.1127321720123291, "global_step": 141746, "epoch": 1592, "lr": 4.665625769383624e-05}
{"train_loss": 0.14822426438331604, "global_step": 141747, "epoch": 1592, "lr": 4.665567925527239e-05}
{"train_loss": 0.08490157127380371, "global_step": 141748, "epoch": 1592, "lr": 4.6655100817158156e-05}
{"train_loss": 0.08263970166444778, "global_step": 141749, "epoch": 1592, "lr": 4.6654522379493595e-05}
{"train_loss": 0.10114294290542603, "global_step": 141750, "epoch": 1592, "lr": 4.66539439422788e-05}
{"train_loss": 0.04567865654826164, "global_step": 141751, "epoch": 1592, "lr": 4.6653365505513824e-05}
{"train_loss": 0.0739343985915184, "global_step": 141752, "epoch": 1592, "lr": 4.665278706919879e-05}
{"train_loss": 0.12186061590909958, "global_step": 141753, "epoch": 1592, "lr": 4.665220863333371e-05}
{"train_loss": 0.09608189016580582, "global_step": 141754, "epoch": 1592, "lr": 4.665163019791874e-05}
{"train_loss": 0.10297010838985443, "global_step": 141755, "epoch": 1592, "lr": 4.665105176295389e-05}
{"train_loss": 0.14258551597595215, "global_step": 141756, "epoch": 1592, "lr": 4.6650473328439285e-05}
{"train_loss": 0.07817769050598145, "global_step": 141757, "epoch": 1592, "lr": 4.664989489437496e-05}
{"train_loss": 0.15782618522644043, "global_step": 141758, "epoch": 1592, "lr": 4.6649316460761045e-05}
{"train_loss": 0.14995238184928894, "global_step": 141759, "epoch": 1592, "lr": 4.664873802759756e-05}
{"train_loss": 0.12117496877908707, "global_step": 141760, "epoch": 1592, "lr": 4.6648159594884635e-05}
{"train_loss": 0.15266162157058716, "global_step": 141761, "epoch": 1592, "lr": 4.66475811626223e-05}
{"train_loss": 0.1580372452735901, "global_step": 141762, "epoch": 1592, "lr": 4.664700273081068e-05}
{"train_loss": 0.11504753679037094, "global_step": 141763, "epoch": 1592, "lr": 4.664642429944981e-05}
{"train_loss": 0.12406472861766815, "global_step": 141764, "epoch": 1592, "lr": 4.6645845868539794e-05}
{"train_loss": 0.08993770182132721, "global_step": 141765, "epoch": 1592, "lr": 4.664526743808072e-05}
{"train_loss": 0.11645182967185974, "global_step": 141766, "epoch": 1592, "lr": 4.6644689008072625e-05}
{"train_loss": 0.11469057947397232, "global_step": 141767, "epoch": 1592, "lr": 4.664411057851563e-05}
{"train_loss": 0.10288561135530472, "global_step": 141768, "epoch": 1592, "lr": 4.664353214940978e-05}
{"train_loss": 0.1038331612944603, "global_step": 141769, "epoch": 1592, "lr": 4.6642953720755173e-05}
{"train_loss": 0.03931396082043648, "global_step": 141770, "epoch": 1592, "lr": 4.664237529255186e-05}
{"train_loss": 0.14019879698753357, "global_step": 141771, "epoch": 1592, "lr": 4.664179686479997e-05}
{"train_loss": 0.09162073582410812, "global_step": 141772, "epoch": 1592, "lr": 4.664121843749953e-05}
{"train_loss": 0.1190769299864769, "global_step": 141773, "epoch": 1592, "lr": 4.664064001065065e-05}
{"train_loss": 0.12085610628128052, "global_step": 141774, "epoch": 1592, "lr": 4.664006158425338e-05}
{"train_loss": 0.12326043844223022, "global_step": 141775, "epoch": 1592, "lr": 4.663948315830783e-05}
{"train_loss": 0.11296313553211394, "global_step": 141776, "epoch": 1592, "lr": 4.663890473281404e-05, "val_loss": 5.9760003089904785}
{"train_loss": 0.09779167175292969, "global_step": 141777, "epoch": 1593, "lr": 4.6638326307772127e-05}
{"train_loss": 0.0653131902217865, "global_step": 141778, "epoch": 1593, "lr": 4.663774788318213e-05}
{"train_loss": 0.08645463734865189, "global_step": 141779, "epoch": 1593, "lr": 4.663716945904417e-05}
{"train_loss": 0.15935637056827545, "global_step": 141780, "epoch": 1593, "lr": 4.663659103535828e-05}
{"train_loss": 0.12084400653839111, "global_step": 141781, "epoch": 1593, "lr": 4.6636012612124566e-05}
{"train_loss": 0.1742219626903534, "global_step": 141782, "epoch": 1593, "lr": 4.6635434189343106e-05}
{"train_loss": 0.1000465601682663, "global_step": 141783, "epoch": 1593, "lr": 4.663485576701397e-05}
{"train_loss": 0.12104400247335434, "global_step": 141784, "epoch": 1593, "lr": 4.663427734513725e-05}
{"train_loss": 0.05705184116959572, "global_step": 141785, "epoch": 1593, "lr": 4.663369892371298e-05}
{"train_loss": 0.1117827445268631, "global_step": 141786, "epoch": 1593, "lr": 4.663312050274129e-05}
{"train_loss": 0.10662125051021576, "global_step": 141787, "epoch": 1593, "lr": 4.663254208222222e-05}
{"train_loss": 0.08941426128149033, "global_step": 141788, "epoch": 1593, "lr": 4.663196366215588e-05}
{"train_loss": 0.13953617215156555, "global_step": 141789, "epoch": 1593, "lr": 4.6631385242542315e-05}
{"train_loss": 0.08425840735435486, "global_step": 141790, "epoch": 1593, "lr": 4.6630806823381636e-05}
{"train_loss": 0.12341063469648361, "global_step": 141791, "epoch": 1593, "lr": 4.663022840467389e-05}
{"train_loss": 0.11055291444063187, "global_step": 141792, "epoch": 1593, "lr": 4.6629649986419184e-05}
{"train_loss": 0.12179771065711975, "global_step": 141793, "epoch": 1593, "lr": 4.662907156861756e-05}
{"train_loss": 0.10461080074310303, "global_step": 141794, "epoch": 1593, "lr": 4.662849315126914e-05}
{"train_loss": 0.09477560222148895, "global_step": 141795, "epoch": 1593, "lr": 4.662791473437395e-05}
{"train_loss": 0.10314438492059708, "global_step": 141796, "epoch": 1593, "lr": 4.662733631793212e-05}
{"train_loss": 0.08591952174901962, "global_step": 141797, "epoch": 1593, "lr": 4.662675790194369e-05}
{"train_loss": 0.11437246203422546, "global_step": 141798, "epoch": 1593, "lr": 4.662617948640876e-05}
{"train_loss": 0.10815386474132538, "global_step": 141799, "epoch": 1593, "lr": 4.662560107132739e-05}
{"train_loss": 0.14223627746105194, "global_step": 141800, "epoch": 1593, "lr": 4.662502265669967e-05}
{"train_loss": 0.1902923882007599, "global_step": 141801, "epoch": 1593, "lr": 4.662444424252568e-05}
{"train_loss": 0.13916611671447754, "global_step": 141802, "epoch": 1593, "lr": 4.662386582880549e-05}
{"train_loss": 0.08422788977622986, "global_step": 141803, "epoch": 1593, "lr": 4.662328741553919e-05}
{"train_loss": 0.09046979993581772, "global_step": 141804, "epoch": 1593, "lr": 4.6622709002726824e-05}
{"train_loss": 0.08336184918880463, "global_step": 141805, "epoch": 1593, "lr": 4.662213059036852e-05}
{"train_loss": 0.14784003794193268, "global_step": 141806, "epoch": 1593, "lr": 4.66215521784643e-05}
{"train_loss": 0.13047729432582855, "global_step": 141807, "epoch": 1593, "lr": 4.66209737670143e-05}
{"train_loss": 0.18499188125133514, "global_step": 141808, "epoch": 1593, "lr": 4.6620395356018544e-05}
{"train_loss": 0.2016802579164505, "global_step": 141809, "epoch": 1593, "lr": 4.661981694547716e-05}
{"train_loss": 0.14920945465564728, "global_step": 141810, "epoch": 1593, "lr": 4.661923853539017e-05}
{"train_loss": 0.13581876456737518, "global_step": 141811, "epoch": 1593, "lr": 4.661866012575771e-05}
{"train_loss": 0.08729983121156693, "global_step": 141812, "epoch": 1593, "lr": 4.661808171657981e-05}
{"train_loss": 0.118800088763237, "global_step": 141813, "epoch": 1593, "lr": 4.661750330785659e-05}
{"train_loss": 0.11185317486524582, "global_step": 141814, "epoch": 1593, "lr": 4.661692489958808e-05}
{"train_loss": 0.11853300780057907, "global_step": 141815, "epoch": 1593, "lr": 4.66163464917744e-05}
{"train_loss": 0.11284711956977844, "global_step": 141816, "epoch": 1593, "lr": 4.66157680844156e-05}
{"train_loss": 0.11732370406389236, "global_step": 141817, "epoch": 1593, "lr": 4.6615189677511775e-05}
{"train_loss": 0.0778593048453331, "global_step": 141818, "epoch": 1593, "lr": 4.6614611271062994e-05}
{"train_loss": 0.08275706321001053, "global_step": 141819, "epoch": 1593, "lr": 4.661403286506935e-05}
{"train_loss": 0.11411131918430328, "global_step": 141820, "epoch": 1593, "lr": 4.661345445953089e-05}
{"train_loss": 0.10606073588132858, "global_step": 141821, "epoch": 1593, "lr": 4.661287605444772e-05}
{"train_loss": 0.09478431940078735, "global_step": 141822, "epoch": 1593, "lr": 4.6612297649819905e-05}
{"train_loss": 0.09464691579341888, "global_step": 141823, "epoch": 1593, "lr": 4.661171924564752e-05}
{"train_loss": 0.10000189393758774, "global_step": 141824, "epoch": 1593, "lr": 4.6611140841930666e-05}
{"train_loss": 0.14561165869235992, "global_step": 141825, "epoch": 1593, "lr": 4.661056243866938e-05}
{"train_loss": 0.09752461314201355, "global_step": 141826, "epoch": 1593, "lr": 4.660998403586379e-05}
{"train_loss": 0.1127464547753334, "global_step": 141827, "epoch": 1593, "lr": 4.6609405633513914e-05}
{"train_loss": 0.13087226450443268, "global_step": 141828, "epoch": 1593, "lr": 4.6608827231619886e-05}
{"train_loss": 0.1984691321849823, "global_step": 141829, "epoch": 1593, "lr": 4.6608248830181745e-05}
{"train_loss": 0.08669818192720413, "global_step": 141830, "epoch": 1593, "lr": 4.660767042919959e-05}
{"train_loss": 0.09592963010072708, "global_step": 141831, "epoch": 1593, "lr": 4.660709202867349e-05}
{"train_loss": 0.13494190573692322, "global_step": 141832, "epoch": 1593, "lr": 4.660651362860354e-05}
{"train_loss": 0.05669640377163887, "global_step": 141833, "epoch": 1593, "lr": 4.6605935228989777e-05}
{"train_loss": 0.12857586145401, "global_step": 141834, "epoch": 1593, "lr": 4.660535682983232e-05}
{"train_loss": 0.19350355863571167, "global_step": 141835, "epoch": 1593, "lr": 4.660477843113122e-05}
{"train_loss": 0.11281437426805496, "global_step": 141836, "epoch": 1593, "lr": 4.6604200032886566e-05}
{"train_loss": 0.12205665558576584, "global_step": 141837, "epoch": 1593, "lr": 4.660362163509846e-05}
{"train_loss": 0.1492965668439865, "global_step": 141838, "epoch": 1593, "lr": 4.6603043237766944e-05}
{"train_loss": 0.18324396014213562, "global_step": 141839, "epoch": 1593, "lr": 4.66024648408921e-05}
{"train_loss": 0.13794109225273132, "global_step": 141840, "epoch": 1593, "lr": 4.660188644447402e-05}
{"train_loss": 0.09657629579305649, "global_step": 141841, "epoch": 1593, "lr": 4.660130804851277e-05}
{"train_loss": 0.1682390421628952, "global_step": 141842, "epoch": 1593, "lr": 4.660072965300844e-05}
{"train_loss": 0.08694922924041748, "global_step": 141843, "epoch": 1593, "lr": 4.66001512579611e-05}
{"train_loss": 0.07241992652416229, "global_step": 141844, "epoch": 1593, "lr": 4.6599572863370814e-05}
{"train_loss": 0.05980528146028519, "global_step": 141845, "epoch": 1593, "lr": 4.6598994469237695e-05}
{"train_loss": 0.1385955810546875, "global_step": 141846, "epoch": 1593, "lr": 4.659841607556178e-05}
{"train_loss": 0.06540445238351822, "global_step": 141847, "epoch": 1593, "lr": 4.6597837682343184e-05}
{"train_loss": 0.1305229365825653, "global_step": 141848, "epoch": 1593, "lr": 4.659725928958195e-05}
{"train_loss": 0.11424236744642258, "global_step": 141849, "epoch": 1593, "lr": 4.659668089727819e-05}
{"train_loss": 0.12542259693145752, "global_step": 141850, "epoch": 1593, "lr": 4.6596102505431944e-05}
{"train_loss": 0.06776437908411026, "global_step": 141851, "epoch": 1593, "lr": 4.6595524114043334e-05}
{"train_loss": 0.14460040628910065, "global_step": 141852, "epoch": 1593, "lr": 4.659494572311239e-05}
{"train_loss": 0.12797851860523224, "global_step": 141853, "epoch": 1593, "lr": 4.6594367332639215e-05}
{"train_loss": 0.09389185905456543, "global_step": 141854, "epoch": 1593, "lr": 4.659378894262391e-05}
{"train_loss": 0.05936595797538757, "global_step": 141855, "epoch": 1593, "lr": 4.6593210553066504e-05}
{"train_loss": 0.05199677124619484, "global_step": 141856, "epoch": 1593, "lr": 4.6592632163967126e-05}
{"train_loss": 0.16073772311210632, "global_step": 141857, "epoch": 1593, "lr": 4.659205377532581e-05}
{"train_loss": 0.14048472046852112, "global_step": 141858, "epoch": 1593, "lr": 4.659147538714266e-05}
{"train_loss": 0.0988074466586113, "global_step": 141859, "epoch": 1593, "lr": 4.659089699941773e-05}
{"train_loss": 0.11582136899232864, "global_step": 141860, "epoch": 1593, "lr": 4.659031861215113e-05}
{"train_loss": 0.0832163393497467, "global_step": 141861, "epoch": 1593, "lr": 4.6589740225342905e-05}
{"train_loss": 0.09255782514810562, "global_step": 141862, "epoch": 1593, "lr": 4.6589161838993166e-05}
{"train_loss": 0.19043433666229248, "global_step": 141863, "epoch": 1593, "lr": 4.658858345310195e-05}
{"train_loss": 0.09654193371534348, "global_step": 141864, "epoch": 1593, "lr": 4.6588005067669376e-05}
{"train_loss": 0.11544084553135915, "global_step": 141865, "epoch": 1593, "lr": 4.658742668269548e-05, "val_loss": 6.0017876625061035}
{"train_loss": 0.11667769402265549, "global_step": 141866, "epoch": 1594, "lr": 4.658684829818039e-05}
{"train_loss": 0.0871925950050354, "global_step": 141867, "epoch": 1594, "lr": 4.658626991412414e-05}
{"train_loss": 0.1558896154165268, "global_step": 141868, "epoch": 1594, "lr": 4.6585691530526834e-05}
{"train_loss": 0.04616909474134445, "global_step": 141869, "epoch": 1594, "lr": 4.658511314738853e-05}
{"train_loss": 0.11179034411907196, "global_step": 141870, "epoch": 1594, "lr": 4.658453476470933e-05}
{"train_loss": 0.12274856865406036, "global_step": 141871, "epoch": 1594, "lr": 4.6583956382489274e-05}
{"train_loss": 0.050556764006614685, "global_step": 141872, "epoch": 1594, "lr": 4.658337800072847e-05}
{"train_loss": 0.10240139067173004, "global_step": 141873, "epoch": 1594, "lr": 4.6582799619427e-05}
{"train_loss": 0.09989101439714432, "global_step": 141874, "epoch": 1594, "lr": 4.658222123858492e-05}
{"train_loss": 0.11372929066419601, "global_step": 141875, "epoch": 1594, "lr": 4.658164285820233e-05}
{"train_loss": 0.10259070992469788, "global_step": 141876, "epoch": 1594, "lr": 4.6581064478279304e-05}
{"train_loss": 0.13212141394615173, "global_step": 141877, "epoch": 1594, "lr": 4.6580486098815894e-05}
{"train_loss": 0.06406662613153458, "global_step": 141878, "epoch": 1594, "lr": 4.657990771981221e-05}
{"train_loss": 0.08999742567539215, "global_step": 141879, "epoch": 1594, "lr": 4.657932934126831e-05}
{"train_loss": 0.1742066740989685, "global_step": 141880, "epoch": 1594, "lr": 4.657875096318427e-05}
{"train_loss": 0.07006315141916275, "global_step": 141881, "epoch": 1594, "lr": 4.657817258556019e-05}
{"train_loss": 0.13465096056461334, "global_step": 141882, "epoch": 1594, "lr": 4.657759420839612e-05}
{"train_loss": 0.06913378089666367, "global_step": 141883, "epoch": 1594, "lr": 4.657701583169216e-05}
{"train_loss": 0.10102680325508118, "global_step": 141884, "epoch": 1594, "lr": 4.657643745544837e-05}
{"train_loss": 0.13846378028392792, "global_step": 141885, "epoch": 1594, "lr": 4.657585907966485e-05}
{"train_loss": 0.11174531280994415, "global_step": 141886, "epoch": 1594, "lr": 4.657528070434164e-05}
{"train_loss": 0.10207868367433548, "global_step": 141887, "epoch": 1594, "lr": 4.657470232947887e-05}
{"train_loss": 0.0706682950258255, "global_step": 141888, "epoch": 1594, "lr": 4.657412395507657e-05}
{"train_loss": 0.13841606676578522, "global_step": 141889, "epoch": 1594, "lr": 4.6573545581134825e-05}
{"train_loss": 0.11060190945863724, "global_step": 141890, "epoch": 1594, "lr": 4.657296720765375e-05}
{"train_loss": 0.15465864539146423, "global_step": 141891, "epoch": 1594, "lr": 4.657238883463338e-05}
{"train_loss": 0.10721558332443237, "global_step": 141892, "epoch": 1594, "lr": 4.6571810462073834e-05}
{"train_loss": 0.20818156003952026, "global_step": 141893, "epoch": 1594, "lr": 4.657123208997514e-05}
{"train_loss": 0.05404727905988693, "global_step": 141894, "epoch": 1594, "lr": 4.6570653718337434e-05}
{"train_loss": 0.10018523037433624, "global_step": 141895, "epoch": 1594, "lr": 4.657007534716074e-05}
{"train_loss": 0.10380496084690094, "global_step": 141896, "epoch": 1594, "lr": 4.656949697644516e-05}
{"train_loss": 0.15692287683486938, "global_step": 141897, "epoch": 1594, "lr": 4.656891860619077e-05}
{"train_loss": 0.10256047546863556, "global_step": 141898, "epoch": 1594, "lr": 4.656834023639765e-05}
{"train_loss": 0.08622798323631287, "global_step": 141899, "epoch": 1594, "lr": 4.656776186706587e-05}
{"train_loss": 0.0704994946718216, "global_step": 141900, "epoch": 1594, "lr": 4.6567183498195524e-05}
{"train_loss": 0.19807825982570648, "global_step": 141901, "epoch": 1594, "lr": 4.656660512978665e-05}
{"train_loss": 0.10884976387023926, "global_step": 141902, "epoch": 1594, "lr": 4.656602676183939e-05}
{"train_loss": 0.08266375958919525, "global_step": 141903, "epoch": 1594, "lr": 4.6565448394353764e-05}
{"train_loss": 0.10714147984981537, "global_step": 141904, "epoch": 1594, "lr": 4.656487002732988e-05}
{"train_loss": 0.10986292362213135, "global_step": 141905, "epoch": 1594, "lr": 4.656429166076779e-05}
{"train_loss": 0.11089928448200226, "global_step": 141906, "epoch": 1594, "lr": 4.656371329466762e-05}
{"train_loss": 0.07493513077497482, "global_step": 141907, "epoch": 1594, "lr": 4.656313492902939e-05}
{"train_loss": 0.09745825827121735, "global_step": 141908, "epoch": 1594, "lr": 4.65625565638532e-05}
{"train_loss": 0.10495394468307495, "global_step": 141909, "epoch": 1594, "lr": 4.656197819913915e-05}
{"train_loss": 0.12675262987613678, "global_step": 141910, "epoch": 1594, "lr": 4.656139983488729e-05}
{"train_loss": 0.06810005754232407, "global_step": 141911, "epoch": 1594, "lr": 4.656082147109773e-05}
{"train_loss": 0.12729044258594513, "global_step": 141912, "epoch": 1594, "lr": 4.6560243107770495e-05}
{"train_loss": 0.12026414275169373, "global_step": 141913, "epoch": 1594, "lr": 4.6559664744905726e-05}
{"train_loss": 0.06952454149723053, "global_step": 141914, "epoch": 1594, "lr": 4.6559086382503444e-05}
{"train_loss": 0.1116747260093689, "global_step": 141915, "epoch": 1594, "lr": 4.6558508020563765e-05}
{"train_loss": 0.10964394360780716, "global_step": 141916, "epoch": 1594, "lr": 4.655792965908674e-05}
{"train_loss": 0.13329753279685974, "global_step": 141917, "epoch": 1594, "lr": 4.655735129807247e-05}
{"train_loss": 0.0984930545091629, "global_step": 141918, "epoch": 1594, "lr": 4.6556772937521014e-05}
{"train_loss": 0.11204960942268372, "global_step": 141919, "epoch": 1594, "lr": 4.6556194577432474e-05}
{"train_loss": 0.09587204456329346, "global_step": 141920, "epoch": 1594, "lr": 4.655561621780689e-05}
{"train_loss": 0.17094257473945618, "global_step": 141921, "epoch": 1594, "lr": 4.655503785864438e-05}
{"train_loss": 0.07212267816066742, "global_step": 141922, "epoch": 1594, "lr": 4.655445949994498e-05}
{"train_loss": 0.09192337840795517, "global_step": 141923, "epoch": 1594, "lr": 4.655388114170882e-05}
{"train_loss": 0.13155783712863922, "global_step": 141924, "epoch": 1594, "lr": 4.655330278393592e-05}
{"train_loss": 0.08507689833641052, "global_step": 141925, "epoch": 1594, "lr": 4.6552724426626384e-05}
{"train_loss": 0.117015041410923, "global_step": 141926, "epoch": 1594, "lr": 4.655214606978032e-05}
{"train_loss": 0.08257453143596649, "global_step": 141927, "epoch": 1594, "lr": 4.655156771339775e-05}
{"train_loss": 0.15062172710895538, "global_step": 141928, "epoch": 1594, "lr": 4.655098935747881e-05}
{"train_loss": 0.19016477465629578, "global_step": 141929, "epoch": 1594, "lr": 4.655041100202352e-05}
{"train_loss": 0.13728511333465576, "global_step": 141930, "epoch": 1594, "lr": 4.654983264703201e-05}
{"train_loss": 0.08692062646150589, "global_step": 141931, "epoch": 1594, "lr": 4.654925429250431e-05}
{"train_loss": 0.07771465927362442, "global_step": 141932, "epoch": 1594, "lr": 4.654867593844054e-05}
{"train_loss": 0.13633684813976288, "global_step": 141933, "epoch": 1594, "lr": 4.654809758484075e-05}
{"train_loss": 0.1150287389755249, "global_step": 141934, "epoch": 1594, "lr": 4.654751923170503e-05}
{"train_loss": 0.1823192536830902, "global_step": 141935, "epoch": 1594, "lr": 4.654694087903345e-05}
{"train_loss": 0.07771705090999603, "global_step": 141936, "epoch": 1594, "lr": 4.654636252682609e-05}
{"train_loss": 0.1391080766916275, "global_step": 141937, "epoch": 1594, "lr": 4.654578417508303e-05}
{"train_loss": 0.13754241168498993, "global_step": 141938, "epoch": 1594, "lr": 4.654520582380435e-05}
{"train_loss": 0.11102667450904846, "global_step": 141939, "epoch": 1594, "lr": 4.654462747299012e-05}
{"train_loss": 0.1087501272559166, "global_step": 141940, "epoch": 1594, "lr": 4.654404912264044e-05}
{"train_loss": 0.1409275382757187, "global_step": 141941, "epoch": 1594, "lr": 4.6543470772755344e-05}
{"train_loss": 0.1454041302204132, "global_step": 141942, "epoch": 1594, "lr": 4.654289242333496e-05}
{"train_loss": 0.08103016018867493, "global_step": 141943, "epoch": 1594, "lr": 4.654231407437932e-05}
{"train_loss": 0.05053474381566048, "global_step": 141944, "epoch": 1594, "lr": 4.6541735725888524e-05}
{"train_loss": 0.10509885102510452, "global_step": 141945, "epoch": 1594, "lr": 4.654115737786267e-05}
{"train_loss": 0.11664828658103943, "global_step": 141946, "epoch": 1594, "lr": 4.654057903030179e-05}
{"train_loss": 0.08311101049184799, "global_step": 141947, "epoch": 1594, "lr": 4.654000068320601e-05}
{"train_loss": 0.13880258798599243, "global_step": 141948, "epoch": 1594, "lr": 4.653942233657537e-05}
{"train_loss": 0.09768824279308319, "global_step": 141949, "epoch": 1594, "lr": 4.6538843990409984e-05}
{"train_loss": 0.21490073204040527, "global_step": 141950, "epoch": 1594, "lr": 4.6538265644709884e-05}
{"train_loss": 0.04607908055186272, "global_step": 141951, "epoch": 1594, "lr": 4.653768729947519e-05}
{"train_loss": 0.17683438956737518, "global_step": 141952, "epoch": 1594, "lr": 4.6537108954705947e-05}
{"train_loss": 0.0899052768945694, "global_step": 141953, "epoch": 1594, "lr": 4.6536530610402265e-05}
{"train_loss": 0.1123922634660528, "global_step": 141954, "epoch": 1594, "lr": 4.653595226656419e-05, "val_loss": 6.096213340759277}
{"train_loss": 0.13391412794589996, "global_step": 141955, "epoch": 1595, "lr": 4.6535373923191815e-05}
{"train_loss": 0.06929440796375275, "global_step": 141956, "epoch": 1595, "lr": 4.6534795580285216e-05}
{"train_loss": 0.0797279104590416, "global_step": 141957, "epoch": 1595, "lr": 4.653421723784448e-05}
{"train_loss": 0.10103728622198105, "global_step": 141958, "epoch": 1595, "lr": 4.653363889586967e-05}
{"train_loss": 0.09288133680820465, "global_step": 141959, "epoch": 1595, "lr": 4.653306055436088e-05}
{"train_loss": 0.10925412178039551, "global_step": 141960, "epoch": 1595, "lr": 4.653248221331816e-05}
{"train_loss": 0.12799042463302612, "global_step": 141961, "epoch": 1595, "lr": 4.653190387274162e-05}
{"train_loss": 0.08889703452587128, "global_step": 141962, "epoch": 1595, "lr": 4.6531325532631307e-05}
{"train_loss": 0.12802456319332123, "global_step": 141963, "epoch": 1595, "lr": 4.653074719298731e-05}
{"train_loss": 0.0998525619506836, "global_step": 141964, "epoch": 1595, "lr": 4.653016885380974e-05}
{"train_loss": 0.07519969344139099, "global_step": 141965, "epoch": 1595, "lr": 4.6529590515098626e-05}
{"train_loss": 0.0948958545923233, "global_step": 141966, "epoch": 1595, "lr": 4.652901217685408e-05}
{"train_loss": 0.10148104280233383, "global_step": 141967, "epoch": 1595, "lr": 4.6528433839076154e-05}
{"train_loss": 0.09393973648548126, "global_step": 141968, "epoch": 1595, "lr": 4.652785550176495e-05}
{"train_loss": 0.12980078160762787, "global_step": 141969, "epoch": 1595, "lr": 4.652727716492052e-05}
{"train_loss": 0.12934105098247528, "global_step": 141970, "epoch": 1595, "lr": 4.652669882854297e-05}
{"train_loss": 0.05301670357584953, "global_step": 141971, "epoch": 1595, "lr": 4.6526120492632346e-05}
{"train_loss": 0.12577320635318756, "global_step": 141972, "epoch": 1595, "lr": 4.6525542157188765e-05}
{"train_loss": 0.1042838990688324, "global_step": 141973, "epoch": 1595, "lr": 4.6524963822212264e-05}
{"train_loss": 0.1338062584400177, "global_step": 141974, "epoch": 1595, "lr": 4.652438548770295e-05}
{"train_loss": 0.1390850991010666, "global_step": 141975, "epoch": 1595, "lr": 4.652380715366088e-05}
{"train_loss": 0.05832408741116524, "global_step": 141976, "epoch": 1595, "lr": 4.652322882008615e-05}
{"train_loss": 0.12116196006536484, "global_step": 141977, "epoch": 1595, "lr": 4.652265048697882e-05}
{"train_loss": 0.10637395828962326, "global_step": 141978, "epoch": 1595, "lr": 4.6522072154338995e-05}
{"train_loss": 0.06659547239542007, "global_step": 141979, "epoch": 1595, "lr": 4.6521493822166715e-05}
{"train_loss": 0.06416060030460358, "global_step": 141980, "epoch": 1595, "lr": 4.652091549046207e-05}
{"train_loss": 0.11224987357854843, "global_step": 141981, "epoch": 1595, "lr": 4.652033715922518e-05}
{"train_loss": 0.07680384814739227, "global_step": 141982, "epoch": 1595, "lr": 4.651975882845606e-05}
{"train_loss": 0.08868439495563507, "global_step": 141983, "epoch": 1595, "lr": 4.6519180498154834e-05}
{"train_loss": 0.13584312796592712, "global_step": 141984, "epoch": 1595, "lr": 4.651860216832154e-05}
{"train_loss": 0.07594635337591171, "global_step": 141985, "epoch": 1595, "lr": 4.6518023838956306e-05}
{"train_loss": 0.13877250254154205, "global_step": 141986, "epoch": 1595, "lr": 4.651744551005916e-05}
{"train_loss": 0.15903572738170624, "global_step": 141987, "epoch": 1595, "lr": 4.6516867181630215e-05}
{"train_loss": 0.07871507108211517, "global_step": 141988, "epoch": 1595, "lr": 4.651628885366951e-05}
{"train_loss": 0.07301110774278641, "global_step": 141989, "epoch": 1595, "lr": 4.651571052617718e-05}
{"train_loss": 0.1177254170179367, "global_step": 141990, "epoch": 1595, "lr": 4.651513219915325e-05}
{"train_loss": 0.12022954970598221, "global_step": 141991, "epoch": 1595, "lr": 4.6514553872597826e-05}
{"train_loss": 0.08815932273864746, "global_step": 141992, "epoch": 1595, "lr": 4.651397554651097e-05}
{"train_loss": 0.12244710326194763, "global_step": 141993, "epoch": 1595, "lr": 4.6513397220892775e-05}
{"train_loss": 0.183865487575531, "global_step": 141994, "epoch": 1595, "lr": 4.651281889574331e-05}
{"train_loss": 0.17014890909194946, "global_step": 141995, "epoch": 1595, "lr": 4.651224057106265e-05}
{"train_loss": 0.06832896918058395, "global_step": 141996, "epoch": 1595, "lr": 4.651166224685088e-05}
{"train_loss": 0.13222819566726685, "global_step": 141997, "epoch": 1595, "lr": 4.651108392310808e-05}
{"train_loss": 0.15051573514938354, "global_step": 141998, "epoch": 1595, "lr": 4.651050559983431e-05}
{"train_loss": 0.09425538778305054, "global_step": 141999, "epoch": 1595, "lr": 4.6509927277029654e-05}
{"train_loss": 0.127491295337677, "global_step": 142000, "epoch": 1595, "lr": 4.650934895469422e-05}
{"train_loss": 0.14571869373321533, "global_step": 142001, "epoch": 1595, "lr": 4.6508770632828034e-05}
{"train_loss": 0.07216648012399673, "global_step": 142002, "epoch": 1595, "lr": 4.6508192311431234e-05}
{"train_loss": 0.10762488842010498, "global_step": 142003, "epoch": 1595, "lr": 4.650761399050383e-05}
{"train_loss": 0.08506694436073303, "global_step": 142004, "epoch": 1595, "lr": 4.6507035670045965e-05}
{"train_loss": 0.14052225649356842, "global_step": 142005, "epoch": 1595, "lr": 4.6506457350057665e-05}
{"train_loss": 0.15887194871902466, "global_step": 142006, "epoch": 1595, "lr": 4.650587903053905e-05}
{"train_loss": 0.07376793771982193, "global_step": 142007, "epoch": 1595, "lr": 4.650530071149015e-05}
{"train_loss": 0.07189943641424179, "global_step": 142008, "epoch": 1595, "lr": 4.6504722392911095e-05}
{"train_loss": 0.13576434552669525, "global_step": 142009, "epoch": 1595, "lr": 4.6504144074801914e-05}
{"train_loss": 0.10536445677280426, "global_step": 142010, "epoch": 1595, "lr": 4.6503565757162735e-05}
{"train_loss": 0.07235965877771378, "global_step": 142011, "epoch": 1595, "lr": 4.650298743999359e-05}
{"train_loss": 0.12309721112251282, "global_step": 142012, "epoch": 1595, "lr": 4.6502409123294585e-05}
{"train_loss": 0.13184629380702972, "global_step": 142013, "epoch": 1595, "lr": 4.6501830807065784e-05}
{"train_loss": 0.04800647497177124, "global_step": 142014, "epoch": 1595, "lr": 4.650125249130727e-05}
{"train_loss": 0.16667337715625763, "global_step": 142015, "epoch": 1595, "lr": 4.650067417601912e-05}
{"train_loss": 0.07670290768146515, "global_step": 142016, "epoch": 1595, "lr": 4.65000958612014e-05}
{"train_loss": 0.12452824413776398, "global_step": 142017, "epoch": 1595, "lr": 4.649951754685422e-05}
{"train_loss": 0.13341768085956573, "global_step": 142018, "epoch": 1595, "lr": 4.6498939232977615e-05}
{"train_loss": 0.09109456837177277, "global_step": 142019, "epoch": 1595, "lr": 4.649836091957171e-05}
{"train_loss": 0.11306893080472946, "global_step": 142020, "epoch": 1595, "lr": 4.649778260663654e-05}
{"train_loss": 0.06342069804668427, "global_step": 142021, "epoch": 1595, "lr": 4.6497204294172215e-05}
{"train_loss": 0.18465621769428253, "global_step": 142022, "epoch": 1595, "lr": 4.6496625982178775e-05}
{"train_loss": 0.07261352241039276, "global_step": 142023, "epoch": 1595, "lr": 4.6496047670656346e-05}
{"train_loss": 0.08297225087881088, "global_step": 142024, "epoch": 1595, "lr": 4.649546935960496e-05}
{"train_loss": 0.0402771532535553, "global_step": 142025, "epoch": 1595, "lr": 4.649489104902474e-05}
{"train_loss": 0.14403070509433746, "global_step": 142026, "epoch": 1595, "lr": 4.649431273891571e-05}
{"train_loss": 0.06523316353559494, "global_step": 142027, "epoch": 1595, "lr": 4.6493734429278e-05}
{"train_loss": 0.09281051158905029, "global_step": 142028, "epoch": 1595, "lr": 4.649315612011164e-05}
{"train_loss": 0.05592954158782959, "global_step": 142029, "epoch": 1595, "lr": 4.649257781141676e-05}
{"train_loss": 0.09342383593320847, "global_step": 142030, "epoch": 1595, "lr": 4.64919995031934e-05}
{"train_loss": 0.13571861386299133, "global_step": 142031, "epoch": 1595, "lr": 4.649142119544165e-05}
{"train_loss": 0.06484223157167435, "global_step": 142032, "epoch": 1595, "lr": 4.649084288816158e-05}
{"train_loss": 0.1187634989619255, "global_step": 142033, "epoch": 1595, "lr": 4.649026458135328e-05}
{"train_loss": 0.10669577866792679, "global_step": 142034, "epoch": 1595, "lr": 4.648968627501682e-05}
{"train_loss": 0.10957492887973785, "global_step": 142035, "epoch": 1595, "lr": 4.6489107969152256e-05}
{"train_loss": 0.10195343941450119, "global_step": 142036, "epoch": 1595, "lr": 4.648852966375972e-05}
{"train_loss": 0.10246381908655167, "global_step": 142037, "epoch": 1595, "lr": 4.648795135883923e-05}
{"train_loss": 0.1060803011059761, "global_step": 142038, "epoch": 1595, "lr": 4.6487373054390924e-05}
{"train_loss": 0.12164080888032913, "global_step": 142039, "epoch": 1595, "lr": 4.648679475041482e-05}
{"train_loss": 0.25798025727272034, "global_step": 142040, "epoch": 1595, "lr": 4.648621644691105e-05}
{"train_loss": 0.1758880913257599, "global_step": 142041, "epoch": 1595, "lr": 4.648563814387964e-05}
{"train_loss": 0.14061671495437622, "global_step": 142042, "epoch": 1595, "lr": 4.6485059841320714e-05}
{"train_loss": 0.10863273649403218, "global_step": 142043, "epoch": 1595, "lr": 4.6484481539234307e-05, "val_loss": 6.12042760848999, "train_action_mse_error": 4.897022247314453}
{"train_loss": 0.10884714871644974, "global_step": 142044, "epoch": 1596, "lr": 4.6483903237620543e-05}
{"train_loss": 0.1282808780670166, "global_step": 142045, "epoch": 1596, "lr": 4.648332493647946e-05}
{"train_loss": 0.08499216288328171, "global_step": 142046, "epoch": 1596, "lr": 4.648274663581116e-05}
{"train_loss": 0.14580240845680237, "global_step": 142047, "epoch": 1596, "lr": 4.64821683356157e-05}
{"train_loss": 0.08672397583723068, "global_step": 142048, "epoch": 1596, "lr": 4.648159003589318e-05}
{"train_loss": 0.12229545414447784, "global_step": 142049, "epoch": 1596, "lr": 4.6481011736643664e-05}
{"train_loss": 0.1237238198518753, "global_step": 142050, "epoch": 1596, "lr": 4.648043343786723e-05}
{"train_loss": 0.12591639161109924, "global_step": 142051, "epoch": 1596, "lr": 4.6479855139563964e-05}
{"train_loss": 0.08468262851238251, "global_step": 142052, "epoch": 1596, "lr": 4.6479276841733924e-05}
{"train_loss": 0.11630002409219742, "global_step": 142053, "epoch": 1596, "lr": 4.647869854437723e-05}
{"train_loss": 0.15753541886806488, "global_step": 142054, "epoch": 1596, "lr": 4.64781202474939e-05}
{"train_loss": 0.1026720255613327, "global_step": 142055, "epoch": 1596, "lr": 4.647754195108407e-05}
{"train_loss": 0.12676963210105896, "global_step": 142056, "epoch": 1596, "lr": 4.647696365514777e-05}
{"train_loss": 0.15379080176353455, "global_step": 142057, "epoch": 1596, "lr": 4.647638535968513e-05}
{"train_loss": 0.09238005429506302, "global_step": 142058, "epoch": 1596, "lr": 4.647580706469616e-05}
{"train_loss": 0.14989112317562103, "global_step": 142059, "epoch": 1596, "lr": 4.6475228770181006e-05}
{"train_loss": 0.12149952352046967, "global_step": 142060, "epoch": 1596, "lr": 4.6474650476139684e-05}
{"train_loss": 0.10440241545438766, "global_step": 142061, "epoch": 1596, "lr": 4.6474072182572334e-05}
{"train_loss": 0.12562058866024017, "global_step": 142062, "epoch": 1596, "lr": 4.647349388947898e-05}
{"train_loss": 0.20702825486660004, "global_step": 142063, "epoch": 1596, "lr": 4.647291559685974e-05}
{"train_loss": 0.07190349698066711, "global_step": 142064, "epoch": 1596, "lr": 4.647233730471465e-05}
{"train_loss": 0.08089844137430191, "global_step": 142065, "epoch": 1596, "lr": 4.647175901304384e-05}
{"train_loss": 0.1183127835392952, "global_step": 142066, "epoch": 1596, "lr": 4.647118072184732e-05}
{"train_loss": 0.09439413249492645, "global_step": 142067, "epoch": 1596, "lr": 4.647060243112525e-05}
{"train_loss": 0.1338062733411789, "global_step": 142068, "epoch": 1596, "lr": 4.647002414087763e-05}
{"train_loss": 0.07793684303760529, "global_step": 142069, "epoch": 1596, "lr": 4.6469445851104596e-05}
{"train_loss": 0.06649255752563477, "global_step": 142070, "epoch": 1596, "lr": 4.64688675618062e-05}
{"train_loss": 0.12289439886808395, "global_step": 142071, "epoch": 1596, "lr": 4.64682892729825e-05}
{"train_loss": 0.07223278284072876, "global_step": 142072, "epoch": 1596, "lr": 4.6467710984633615e-05}
{"train_loss": 0.06606150418519974, "global_step": 142073, "epoch": 1596, "lr": 4.6467132696759594e-05}
{"train_loss": 0.12377317249774933, "global_step": 142074, "epoch": 1596, "lr": 4.6466554409360535e-05}
{"train_loss": 0.0990503802895546, "global_step": 142075, "epoch": 1596, "lr": 4.6465976122436484e-05}
{"train_loss": 0.13364277780056, "global_step": 142076, "epoch": 1596, "lr": 4.6465397835987566e-05}
{"train_loss": 0.11539434641599655, "global_step": 142077, "epoch": 1596, "lr": 4.6464819550013805e-05}
{"train_loss": 0.17050473392009735, "global_step": 142078, "epoch": 1596, "lr": 4.646424126451533e-05}
{"train_loss": 0.10155774652957916, "global_step": 142079, "epoch": 1596, "lr": 4.646366297949218e-05}
{"train_loss": 0.11673492193222046, "global_step": 142080, "epoch": 1596, "lr": 4.646308469494446e-05}
{"train_loss": 0.17966686189174652, "global_step": 142081, "epoch": 1596, "lr": 4.646250641087222e-05}
{"train_loss": 0.05783959850668907, "global_step": 142082, "epoch": 1596, "lr": 4.646192812727557e-05}
{"train_loss": 0.07888085395097733, "global_step": 142083, "epoch": 1596, "lr": 4.646134984415454e-05}
{"train_loss": 0.12698997557163239, "global_step": 142084, "epoch": 1596, "lr": 4.6460771561509276e-05}
{"train_loss": 0.0680447369813919, "global_step": 142085, "epoch": 1596, "lr": 4.646019327933979e-05}
{"train_loss": 0.1653933972120285, "global_step": 142086, "epoch": 1596, "lr": 4.645961499764621e-05}
{"train_loss": 0.15292523801326752, "global_step": 142087, "epoch": 1596, "lr": 4.645903671642857e-05}
{"train_loss": 0.07087160646915436, "global_step": 142088, "epoch": 1596, "lr": 4.645845843568699e-05}
{"train_loss": 0.11530144512653351, "global_step": 142089, "epoch": 1596, "lr": 4.645788015542152e-05}
{"train_loss": 0.14886116981506348, "global_step": 142090, "epoch": 1596, "lr": 4.645730187563223e-05}
{"train_loss": 0.08880838006734848, "global_step": 142091, "epoch": 1596, "lr": 4.645672359631923e-05}
{"train_loss": 0.1250508576631546, "global_step": 142092, "epoch": 1596, "lr": 4.645614531748257e-05}
{"train_loss": 0.11421326547861099, "global_step": 142093, "epoch": 1596, "lr": 4.645556703912236e-05}
{"train_loss": 0.12367710471153259, "global_step": 142094, "epoch": 1596, "lr": 4.645498876123862e-05}
{"train_loss": 0.09837298840284348, "global_step": 142095, "epoch": 1596, "lr": 4.6454410483831495e-05}
{"train_loss": 0.09939944744110107, "global_step": 142096, "epoch": 1596, "lr": 4.6453832206901004e-05}
{"train_loss": 0.08949960023164749, "global_step": 142097, "epoch": 1596, "lr": 4.6453253930447284e-05}
{"train_loss": 0.18133318424224854, "global_step": 142098, "epoch": 1596, "lr": 4.645267565447035e-05}
{"train_loss": 0.11254189163446426, "global_step": 142099, "epoch": 1596, "lr": 4.6452097378970334e-05}
{"train_loss": 0.15286317467689514, "global_step": 142100, "epoch": 1596, "lr": 4.645151910394727e-05}
{"train_loss": 0.1372450888156891, "global_step": 142101, "epoch": 1596, "lr": 4.645094082940128e-05}
{"train_loss": 0.10914342105388641, "global_step": 142102, "epoch": 1596, "lr": 4.6450362555332386e-05}
{"train_loss": 0.10936792194843292, "global_step": 142103, "epoch": 1596, "lr": 4.644978428174073e-05}
{"train_loss": 0.11207104474306107, "global_step": 142104, "epoch": 1596, "lr": 4.644920600862633e-05}
{"train_loss": 0.16343480348587036, "global_step": 142105, "epoch": 1596, "lr": 4.6448627735989315e-05}
{"train_loss": 0.07453498244285583, "global_step": 142106, "epoch": 1596, "lr": 4.644804946382972e-05}
{"train_loss": 0.10426674038171768, "global_step": 142107, "epoch": 1596, "lr": 4.644747119214765e-05}
{"train_loss": 0.09981309622526169, "global_step": 142108, "epoch": 1596, "lr": 4.644689292094318e-05}
{"train_loss": 0.15247930586338043, "global_step": 142109, "epoch": 1596, "lr": 4.6446314650216366e-05}
{"train_loss": 0.05891665816307068, "global_step": 142110, "epoch": 1596, "lr": 4.644573637996732e-05}
{"train_loss": 0.10910801589488983, "global_step": 142111, "epoch": 1596, "lr": 4.644515811019608e-05}
{"train_loss": 0.11619932949542999, "global_step": 142112, "epoch": 1596, "lr": 4.644457984090277e-05}
{"train_loss": 0.07587099075317383, "global_step": 142113, "epoch": 1596, "lr": 4.644400157208742e-05}
{"train_loss": 0.09499453753232956, "global_step": 142114, "epoch": 1596, "lr": 4.644342330375015e-05}
{"train_loss": 0.1178193986415863, "global_step": 142115, "epoch": 1596, "lr": 4.6442845035891e-05}
{"train_loss": 0.14932279288768768, "global_step": 142116, "epoch": 1596, "lr": 4.644226676851009e-05}
{"train_loss": 0.11495205014944077, "global_step": 142117, "epoch": 1596, "lr": 4.644168850160744e-05}
{"train_loss": 0.0967683419585228, "global_step": 142118, "epoch": 1596, "lr": 4.6441110235183196e-05}
{"train_loss": 0.0830257385969162, "global_step": 142119, "epoch": 1596, "lr": 4.6440531969237374e-05}
{"train_loss": 0.15694934129714966, "global_step": 142120, "epoch": 1596, "lr": 4.64399537037701e-05}
{"train_loss": 0.11171688884496689, "global_step": 142121, "epoch": 1596, "lr": 4.643937543878141e-05}
{"train_loss": 0.09191206097602844, "global_step": 142122, "epoch": 1596, "lr": 4.643879717427143e-05}
{"train_loss": 0.09767518192529678, "global_step": 142123, "epoch": 1596, "lr": 4.643821891024018e-05}
{"train_loss": 0.11874111741781235, "global_step": 142124, "epoch": 1596, "lr": 4.6437640646687795e-05}
{"train_loss": 0.18862709403038025, "global_step": 142125, "epoch": 1596, "lr": 4.64370623836143e-05}
{"train_loss": 0.13879212737083435, "global_step": 142126, "epoch": 1596, "lr": 4.643648412101982e-05}
{"train_loss": 0.18624623119831085, "global_step": 142127, "epoch": 1596, "lr": 4.643590585890441e-05}
{"train_loss": 0.10103112459182739, "global_step": 142128, "epoch": 1596, "lr": 4.643532759726814e-05}
{"train_loss": 0.08310569822788239, "global_step": 142129, "epoch": 1596, "lr": 4.643474933611111e-05}
{"train_loss": 0.19751261174678802, "global_step": 142130, "epoch": 1596, "lr": 4.643417107543336e-05}
{"train_loss": 0.09986178576946259, "global_step": 142131, "epoch": 1596, "lr": 4.643359281523502e-05}
{"train_loss": 0.11567063231983882, "global_step": 142132, "epoch": 1596, "lr": 4.643301455551611e-05, "val_loss": 5.972206115722656}
{"train_loss": 0.04516366496682167, "global_step": 142133, "epoch": 1597, "lr": 4.643243629627677e-05}
{"train_loss": 0.1400270164012909, "global_step": 142134, "epoch": 1597, "lr": 4.643185803751701e-05}
{"train_loss": 0.12212423980236053, "global_step": 142135, "epoch": 1597, "lr": 4.643127977923697e-05}
{"train_loss": 0.12628847360610962, "global_step": 142136, "epoch": 1597, "lr": 4.6430701521436686e-05}
{"train_loss": 0.11999090015888214, "global_step": 142137, "epoch": 1597, "lr": 4.643012326411627e-05}
{"train_loss": 0.10298193991184235, "global_step": 142138, "epoch": 1597, "lr": 4.642954500727576e-05}
{"train_loss": 0.06959806382656097, "global_step": 142139, "epoch": 1597, "lr": 4.642896675091527e-05}
{"train_loss": 0.04934779182076454, "global_step": 142140, "epoch": 1597, "lr": 4.6428388495034845e-05}
{"train_loss": 0.1325807273387909, "global_step": 142141, "epoch": 1597, "lr": 4.64278102396346e-05}
{"train_loss": 0.12560877203941345, "global_step": 142142, "epoch": 1597, "lr": 4.642723198471457e-05}
{"train_loss": 0.12374778836965561, "global_step": 142143, "epoch": 1597, "lr": 4.642665373027487e-05}
{"train_loss": 0.19434498250484467, "global_step": 142144, "epoch": 1597, "lr": 4.642607547631556e-05}
{"train_loss": 0.12821336090564728, "global_step": 142145, "epoch": 1597, "lr": 4.642549722283672e-05}
{"train_loss": 0.0708433985710144, "global_step": 142146, "epoch": 1597, "lr": 4.642491896983844e-05}
{"train_loss": 0.08171346038579941, "global_step": 142147, "epoch": 1597, "lr": 4.6424340717320766e-05}
{"train_loss": 0.1052100658416748, "global_step": 142148, "epoch": 1597, "lr": 4.642376246528381e-05}
{"train_loss": 0.13407406210899353, "global_step": 142149, "epoch": 1597, "lr": 4.6423184213727625e-05}
{"train_loss": 0.1031961515545845, "global_step": 142150, "epoch": 1597, "lr": 4.642260596265231e-05}
{"train_loss": 0.1277681291103363, "global_step": 142151, "epoch": 1597, "lr": 4.642202771205791e-05}
{"train_loss": 0.13137121498584747, "global_step": 142152, "epoch": 1597, "lr": 4.642144946194455e-05}
{"train_loss": 0.10034476220607758, "global_step": 142153, "epoch": 1597, "lr": 4.642087121231227e-05}
{"train_loss": 0.10477601736783981, "global_step": 142154, "epoch": 1597, "lr": 4.6420292963161176e-05}
{"train_loss": 0.17310388386249542, "global_step": 142155, "epoch": 1597, "lr": 4.641971471449131e-05}
{"train_loss": 0.14278924465179443, "global_step": 142156, "epoch": 1597, "lr": 4.641913646630278e-05}
{"train_loss": 0.06751727312803268, "global_step": 142157, "epoch": 1597, "lr": 4.6418558218595644e-05}
{"train_loss": 0.16411949694156647, "global_step": 142158, "epoch": 1597, "lr": 4.641797997137001e-05}
{"train_loss": 0.093241386115551, "global_step": 142159, "epoch": 1597, "lr": 4.641740172462591e-05}
{"train_loss": 0.14604589343070984, "global_step": 142160, "epoch": 1597, "lr": 4.6416823478363465e-05}
{"train_loss": 0.09501370787620544, "global_step": 142161, "epoch": 1597, "lr": 4.641624523258272e-05}
{"train_loss": 0.12318345904350281, "global_step": 142162, "epoch": 1597, "lr": 4.641566698728377e-05}
{"train_loss": 0.12680311501026154, "global_step": 142163, "epoch": 1597, "lr": 4.641508874246669e-05}
{"train_loss": 0.13888093829154968, "global_step": 142164, "epoch": 1597, "lr": 4.641451049813157e-05}
{"train_loss": 0.0933324322104454, "global_step": 142165, "epoch": 1597, "lr": 4.641393225427848e-05}
{"train_loss": 0.13177765905857086, "global_step": 142166, "epoch": 1597, "lr": 4.641335401090746e-05}
{"train_loss": 0.11378709226846695, "global_step": 142167, "epoch": 1597, "lr": 4.641277576801866e-05}
{"train_loss": 0.09465373307466507, "global_step": 142168, "epoch": 1597, "lr": 4.641219752561209e-05}
{"train_loss": 0.2184736430644989, "global_step": 142169, "epoch": 1597, "lr": 4.6411619283687875e-05}
{"train_loss": 0.15488934516906738, "global_step": 142170, "epoch": 1597, "lr": 4.641104104224605e-05}
{"train_loss": 0.10686898231506348, "global_step": 142171, "epoch": 1597, "lr": 4.6410462801286746e-05}
{"train_loss": 0.0948047786951065, "global_step": 142172, "epoch": 1597, "lr": 4.6409884560809985e-05}
{"train_loss": 0.06225812807679176, "global_step": 142173, "epoch": 1597, "lr": 4.640930632081589e-05}
{"train_loss": 0.08769792318344116, "global_step": 142174, "epoch": 1597, "lr": 4.640872808130451e-05}
{"train_loss": 0.11090087890625, "global_step": 142175, "epoch": 1597, "lr": 4.6408149842275944e-05}
{"train_loss": 0.1348484307527542, "global_step": 142176, "epoch": 1597, "lr": 4.6407571603730234e-05}
{"train_loss": 0.09533374011516571, "global_step": 142177, "epoch": 1597, "lr": 4.640699336566752e-05}
{"train_loss": 0.103153757750988, "global_step": 142178, "epoch": 1597, "lr": 4.6406415128087805e-05}
{"train_loss": 0.12766797840595245, "global_step": 142179, "epoch": 1597, "lr": 4.6405836890991213e-05}
{"train_loss": 0.11837223917245865, "global_step": 142180, "epoch": 1597, "lr": 4.6405258654377824e-05}
{"train_loss": 0.1547275185585022, "global_step": 142181, "epoch": 1597, "lr": 4.6404680418247703e-05}
{"train_loss": 0.10245130956172943, "global_step": 142182, "epoch": 1597, "lr": 4.640410218260092e-05}
{"train_loss": 0.1296858936548233, "global_step": 142183, "epoch": 1597, "lr": 4.640352394743757e-05}
{"train_loss": 0.0628889799118042, "global_step": 142184, "epoch": 1597, "lr": 4.640294571275772e-05}
{"train_loss": 0.0557221993803978, "global_step": 142185, "epoch": 1597, "lr": 4.640236747856145e-05}
{"train_loss": 0.0962517186999321, "global_step": 142186, "epoch": 1597, "lr": 4.640178924484885e-05}
{"train_loss": 0.08501898497343063, "global_step": 142187, "epoch": 1597, "lr": 4.640121101161996e-05}
{"train_loss": 0.14892368018627167, "global_step": 142188, "epoch": 1597, "lr": 4.640063277887491e-05}
{"train_loss": 0.07259342819452286, "global_step": 142189, "epoch": 1597, "lr": 4.640005454661372e-05}
{"train_loss": 0.18725860118865967, "global_step": 142190, "epoch": 1597, "lr": 4.6399476314836525e-05}
{"train_loss": 0.13963690400123596, "global_step": 142191, "epoch": 1597, "lr": 4.639889808354336e-05}
{"train_loss": 0.1619424968957901, "global_step": 142192, "epoch": 1597, "lr": 4.639831985273434e-05}
{"train_loss": 0.13109740614891052, "global_step": 142193, "epoch": 1597, "lr": 4.6397741622409496e-05}
{"train_loss": 0.10726293921470642, "global_step": 142194, "epoch": 1597, "lr": 4.6397163392568955e-05}
{"train_loss": 0.13787730038166046, "global_step": 142195, "epoch": 1597, "lr": 4.639658516321275e-05}
{"train_loss": 0.10156611353158951, "global_step": 142196, "epoch": 1597, "lr": 4.6396006934341e-05}
{"train_loss": 0.1229075938463211, "global_step": 142197, "epoch": 1597, "lr": 4.6395428705953745e-05}
{"train_loss": 0.105984628200531, "global_step": 142198, "epoch": 1597, "lr": 4.639485047805108e-05}
{"train_loss": 0.11953172832727432, "global_step": 142199, "epoch": 1597, "lr": 4.6394272250633106e-05}
{"train_loss": 0.07534252852201462, "global_step": 142200, "epoch": 1597, "lr": 4.6393694023699866e-05}
{"train_loss": 0.1660252809524536, "global_step": 142201, "epoch": 1597, "lr": 4.639311579725145e-05}
{"train_loss": 0.11558111011981964, "global_step": 142202, "epoch": 1597, "lr": 4.639253757128793e-05}
{"train_loss": 0.11485854536294937, "global_step": 142203, "epoch": 1597, "lr": 4.6391959345809393e-05}
{"train_loss": 0.16266189515590668, "global_step": 142204, "epoch": 1597, "lr": 4.639138112081591e-05}
{"train_loss": 0.10311312228441238, "global_step": 142205, "epoch": 1597, "lr": 4.6390802896307587e-05}
{"train_loss": 0.12235724925994873, "global_step": 142206, "epoch": 1597, "lr": 4.639022467228444e-05}
{"train_loss": 0.09733884781599045, "global_step": 142207, "epoch": 1597, "lr": 4.638964644874661e-05}
{"train_loss": 0.11388511210680008, "global_step": 142208, "epoch": 1597, "lr": 4.638906822569413e-05}
{"train_loss": 0.10877084732055664, "global_step": 142209, "epoch": 1597, "lr": 4.638849000312711e-05}
{"train_loss": 0.11271636188030243, "global_step": 142210, "epoch": 1597, "lr": 4.63879117810456e-05}
{"train_loss": 0.06325510144233704, "global_step": 142211, "epoch": 1597, "lr": 4.63873335594497e-05}
{"train_loss": 0.09260211884975433, "global_step": 142212, "epoch": 1597, "lr": 4.638675533833947e-05}
{"train_loss": 0.09346535801887512, "global_step": 142213, "epoch": 1597, "lr": 4.638617711771501e-05}
{"train_loss": 0.15270943939685822, "global_step": 142214, "epoch": 1597, "lr": 4.6385598897576365e-05}
{"train_loss": 0.07297857105731964, "global_step": 142215, "epoch": 1597, "lr": 4.638502067792366e-05}
{"train_loss": 0.07787131518125534, "global_step": 142216, "epoch": 1597, "lr": 4.638444245875692e-05}
{"train_loss": 0.05801375210285187, "global_step": 142217, "epoch": 1597, "lr": 4.638386424007624e-05}
{"train_loss": 0.11754830181598663, "global_step": 142218, "epoch": 1597, "lr": 4.6383286021881736e-05}
{"train_loss": 0.12564806640148163, "global_step": 142219, "epoch": 1597, "lr": 4.6382707804173444e-05}
{"train_loss": 0.09290695190429688, "global_step": 142220, "epoch": 1597, "lr": 4.638212958695145e-05}
{"train_loss": 0.11429398067379265, "global_step": 142221, "epoch": 1597, "lr": 4.638155137021584e-05, "val_loss": 6.083418846130371}
{"train_loss": 0.07203428447246552, "global_step": 142222, "epoch": 1598, "lr": 4.638097315396668e-05}
{"train_loss": 0.11832442134618759, "global_step": 142223, "epoch": 1598, "lr": 4.6380394938204054e-05}
{"train_loss": 0.22990424931049347, "global_step": 142224, "epoch": 1598, "lr": 4.637981672292805e-05}
{"train_loss": 0.1269279271364212, "global_step": 142225, "epoch": 1598, "lr": 4.637923850813872e-05}
{"train_loss": 0.08041457086801529, "global_step": 142226, "epoch": 1598, "lr": 4.637866029383617e-05}
{"train_loss": 0.09884689003229141, "global_step": 142227, "epoch": 1598, "lr": 4.637808208002045e-05}
{"train_loss": 0.12105818092823029, "global_step": 142228, "epoch": 1598, "lr": 4.637750386669167e-05}
{"train_loss": 0.12697051465511322, "global_step": 142229, "epoch": 1598, "lr": 4.6376925653849873e-05}
{"train_loss": 0.04717562347650528, "global_step": 142230, "epoch": 1598, "lr": 4.637634744149517e-05}
{"train_loss": 0.11876271665096283, "global_step": 142231, "epoch": 1598, "lr": 4.637576922962761e-05}
{"train_loss": 0.13179811835289001, "global_step": 142232, "epoch": 1598, "lr": 4.63751910182473e-05}
{"train_loss": 0.08051455020904541, "global_step": 142233, "epoch": 1598, "lr": 4.637461280735428e-05}
{"train_loss": 0.05798549950122833, "global_step": 142234, "epoch": 1598, "lr": 4.637403459694865e-05}
{"train_loss": 0.19301609694957733, "global_step": 142235, "epoch": 1598, "lr": 4.637345638703051e-05}
{"train_loss": 0.1064516007900238, "global_step": 142236, "epoch": 1598, "lr": 4.637287817759989e-05}
{"train_loss": 0.10875425487756729, "global_step": 142237, "epoch": 1598, "lr": 4.637229996865691e-05}
{"train_loss": 0.10054737329483032, "global_step": 142238, "epoch": 1598, "lr": 4.637172176020163e-05}
{"train_loss": 0.07370410859584808, "global_step": 142239, "epoch": 1598, "lr": 4.637114355223412e-05}
{"train_loss": 0.09921098500490189, "global_step": 142240, "epoch": 1598, "lr": 4.637056534475447e-05}
{"train_loss": 0.19417212903499603, "global_step": 142241, "epoch": 1598, "lr": 4.636998713776275e-05}
{"train_loss": 0.08224263787269592, "global_step": 142242, "epoch": 1598, "lr": 4.636940893125904e-05}
{"train_loss": 0.15021784603595734, "global_step": 142243, "epoch": 1598, "lr": 4.636883072524343e-05}
{"train_loss": 0.11466692388057709, "global_step": 142244, "epoch": 1598, "lr": 4.636825251971597e-05}
{"train_loss": 0.12387917935848236, "global_step": 142245, "epoch": 1598, "lr": 4.636767431467677e-05}
{"train_loss": 0.1382683962583542, "global_step": 142246, "epoch": 1598, "lr": 4.636709611012587e-05}
{"train_loss": 0.11131218820810318, "global_step": 142247, "epoch": 1598, "lr": 4.6366517906063396e-05}
{"train_loss": 0.09457501024007797, "global_step": 142248, "epoch": 1598, "lr": 4.636593970248937e-05}
{"train_loss": 0.0816379189491272, "global_step": 142249, "epoch": 1598, "lr": 4.636536149940393e-05}
{"train_loss": 0.14780738949775696, "global_step": 142250, "epoch": 1598, "lr": 4.63647832968071e-05}
{"train_loss": 0.18813075125217438, "global_step": 142251, "epoch": 1598, "lr": 4.6364205094699e-05}
{"train_loss": 0.07050628215074539, "global_step": 142252, "epoch": 1598, "lr": 4.6363626893079666e-05}
{"train_loss": 0.09463249146938324, "global_step": 142253, "epoch": 1598, "lr": 4.63630486919492e-05}
{"train_loss": 0.05519506707787514, "global_step": 142254, "epoch": 1598, "lr": 4.63624704913077e-05}
{"train_loss": 0.08501429110765457, "global_step": 142255, "epoch": 1598, "lr": 4.636189229115519e-05}
{"train_loss": 0.12217346578836441, "global_step": 142256, "epoch": 1598, "lr": 4.636131409149181e-05}
{"train_loss": 0.12388885021209717, "global_step": 142257, "epoch": 1598, "lr": 4.63607358923176e-05}
{"train_loss": 0.09488082677125931, "global_step": 142258, "epoch": 1598, "lr": 4.636015769363263e-05}
{"train_loss": 0.15446598827838898, "global_step": 142259, "epoch": 1598, "lr": 4.6359579495437e-05}
{"train_loss": 0.05628317967057228, "global_step": 142260, "epoch": 1598, "lr": 4.6359001297730784e-05}
{"train_loss": 0.08586270362138748, "global_step": 142261, "epoch": 1598, "lr": 4.635842310051406e-05}
{"train_loss": 0.20916777849197388, "global_step": 142262, "epoch": 1598, "lr": 4.63578449037869e-05}
{"train_loss": 0.05325639620423317, "global_step": 142263, "epoch": 1598, "lr": 4.635726670754937e-05}
{"train_loss": 0.15111561119556427, "global_step": 142264, "epoch": 1598, "lr": 4.635668851180158e-05}
{"train_loss": 0.12495558708906174, "global_step": 142265, "epoch": 1598, "lr": 4.6356110316543574e-05}
{"train_loss": 0.106698177754879, "global_step": 142266, "epoch": 1598, "lr": 4.6355532121775455e-05}
{"train_loss": 0.08667302131652832, "global_step": 142267, "epoch": 1598, "lr": 4.635495392749727e-05}
{"train_loss": 0.12050539255142212, "global_step": 142268, "epoch": 1598, "lr": 4.635437573370915e-05}
{"train_loss": 0.11046304553747177, "global_step": 142269, "epoch": 1598, "lr": 4.635379754041111e-05}
{"train_loss": 0.13045255839824677, "global_step": 142270, "epoch": 1598, "lr": 4.6353219347603254e-05}
{"train_loss": 0.05069824308156967, "global_step": 142271, "epoch": 1598, "lr": 4.635264115528569e-05}
{"train_loss": 0.13831989467144012, "global_step": 142272, "epoch": 1598, "lr": 4.635206296345844e-05}
{"train_loss": 0.09242618829011917, "global_step": 142273, "epoch": 1598, "lr": 4.6351484772121646e-05}
{"train_loss": 0.10322760790586472, "global_step": 142274, "epoch": 1598, "lr": 4.6350906581275317e-05}
{"train_loss": 0.17651820182800293, "global_step": 142275, "epoch": 1598, "lr": 4.635032839091959e-05}
{"train_loss": 0.11465315520763397, "global_step": 142276, "epoch": 1598, "lr": 4.634975020105451e-05}
{"train_loss": 0.14357362687587738, "global_step": 142277, "epoch": 1598, "lr": 4.6349172011680154e-05}
{"train_loss": 0.07281021773815155, "global_step": 142278, "epoch": 1598, "lr": 4.634859382279661e-05}
{"train_loss": 0.16193507611751556, "global_step": 142279, "epoch": 1598, "lr": 4.634801563440396e-05}
{"train_loss": 0.058578792959451675, "global_step": 142280, "epoch": 1598, "lr": 4.634743744650226e-05}
{"train_loss": 0.05108148604631424, "global_step": 142281, "epoch": 1598, "lr": 4.634685925909162e-05}
{"train_loss": 0.06927105039358139, "global_step": 142282, "epoch": 1598, "lr": 4.634628107217209e-05}
{"train_loss": 0.12128240615129471, "global_step": 142283, "epoch": 1598, "lr": 4.634570288574377e-05}
{"train_loss": 0.08019514381885529, "global_step": 142284, "epoch": 1598, "lr": 4.6345124699806705e-05}
{"train_loss": 0.14509381353855133, "global_step": 142285, "epoch": 1598, "lr": 4.634454651436101e-05}
{"train_loss": 0.0957014337182045, "global_step": 142286, "epoch": 1598, "lr": 4.634396832940673e-05}
{"train_loss": 0.09137354791164398, "global_step": 142287, "epoch": 1598, "lr": 4.634339014494398e-05}
{"train_loss": 0.06890146434307098, "global_step": 142288, "epoch": 1598, "lr": 4.634281196097279e-05}
{"train_loss": 0.1432742476463318, "global_step": 142289, "epoch": 1598, "lr": 4.634223377749327e-05}
{"train_loss": 0.10192063450813293, "global_step": 142290, "epoch": 1598, "lr": 4.634165559450552e-05}
{"train_loss": 0.05899222940206528, "global_step": 142291, "epoch": 1598, "lr": 4.634107741200956e-05}
{"train_loss": 0.1787218451499939, "global_step": 142292, "epoch": 1598, "lr": 4.634049923000552e-05}
{"train_loss": 0.07148058712482452, "global_step": 142293, "epoch": 1598, "lr": 4.6339921048493426e-05}
{"train_loss": 0.0676330029964447, "global_step": 142294, "epoch": 1598, "lr": 4.633934286747342e-05}
{"train_loss": 0.13718822598457336, "global_step": 142295, "epoch": 1598, "lr": 4.6338764686945515e-05}
{"train_loss": 0.1332135796546936, "global_step": 142296, "epoch": 1598, "lr": 4.633818650690984e-05}
{"train_loss": 0.061434537172317505, "global_step": 142297, "epoch": 1598, "lr": 4.633760832736645e-05}
{"train_loss": 0.0880562886595726, "global_step": 142298, "epoch": 1598, "lr": 4.633703014831541e-05}
{"train_loss": 0.156964510679245, "global_step": 142299, "epoch": 1598, "lr": 4.6336451969756825e-05}
{"train_loss": 0.10563254356384277, "global_step": 142300, "epoch": 1598, "lr": 4.6335873791690746e-05}
{"train_loss": 0.062496379017829895, "global_step": 142301, "epoch": 1598, "lr": 4.6335295614117274e-05}
{"train_loss": 0.08990075439214706, "global_step": 142302, "epoch": 1598, "lr": 4.6334717437036483e-05}
{"train_loss": 0.07081414014101028, "global_step": 142303, "epoch": 1598, "lr": 4.633413926044843e-05}
{"train_loss": 0.0983453094959259, "global_step": 142304, "epoch": 1598, "lr": 4.633356108435322e-05}
{"train_loss": 0.12280521541833878, "global_step": 142305, "epoch": 1598, "lr": 4.63329829087509e-05}
{"train_loss": 0.08581480383872986, "global_step": 142306, "epoch": 1598, "lr": 4.633240473364156e-05}
{"train_loss": 0.0941857248544693, "global_step": 142307, "epoch": 1598, "lr": 4.633182655902531e-05}
{"train_loss": 0.11365365236997604, "global_step": 142308, "epoch": 1598, "lr": 4.633124838490218e-05}
{"train_loss": 0.10652340203523636, "global_step": 142309, "epoch": 1598, "lr": 4.6330670211272295e-05}
{"train_loss": 0.10843042920479613, "global_step": 142310, "epoch": 1598, "lr": 4.6330092038135683e-05, "val_loss": 6.005672931671143}
{"train_loss": 0.1006094217300415, "global_step": 142311, "epoch": 1599, "lr": 4.6329513865492465e-05}
{"train_loss": 0.07942858338356018, "global_step": 142312, "epoch": 1599, "lr": 4.632893569334267e-05}
{"train_loss": 0.08737704902887344, "global_step": 142313, "epoch": 1599, "lr": 4.632835752168644e-05}
{"train_loss": 0.13157989084720612, "global_step": 142314, "epoch": 1599, "lr": 4.6327779350523785e-05}
{"train_loss": 0.07387151569128036, "global_step": 142315, "epoch": 1599, "lr": 4.6327201179854844e-05}
{"train_loss": 0.07074586302042007, "global_step": 142316, "epoch": 1599, "lr": 4.6326623009679656e-05}
{"train_loss": 0.12122020125389099, "global_step": 142317, "epoch": 1599, "lr": 4.632604483999831e-05}
{"train_loss": 0.1522287279367447, "global_step": 142318, "epoch": 1599, "lr": 4.6325466670810876e-05}
{"train_loss": 0.08745396882295609, "global_step": 142319, "epoch": 1599, "lr": 4.632488850211744e-05}
{"train_loss": 0.08074350655078888, "global_step": 142320, "epoch": 1599, "lr": 4.6324310333918074e-05}
{"train_loss": 0.06300558149814606, "global_step": 142321, "epoch": 1599, "lr": 4.6323732166212875e-05}
{"train_loss": 0.07708432525396347, "global_step": 142322, "epoch": 1599, "lr": 4.632315399900188e-05}
{"train_loss": 0.0806991383433342, "global_step": 142323, "epoch": 1599, "lr": 4.632257583228522e-05}
{"train_loss": 0.08081020414829254, "global_step": 142324, "epoch": 1599, "lr": 4.632199766606292e-05}
{"train_loss": 0.05644766613841057, "global_step": 142325, "epoch": 1599, "lr": 4.632141950033508e-05}
{"train_loss": 0.08914797008037567, "global_step": 142326, "epoch": 1599, "lr": 4.632084133510181e-05}
{"train_loss": 0.08959272503852844, "global_step": 142327, "epoch": 1599, "lr": 4.632026317036313e-05}
{"train_loss": 0.14964386820793152, "global_step": 142328, "epoch": 1599, "lr": 4.631968500611917e-05}
{"train_loss": 0.09919607639312744, "global_step": 142329, "epoch": 1599, "lr": 4.631910684236996e-05}
{"train_loss": 0.11205561459064484, "global_step": 142330, "epoch": 1599, "lr": 4.631852867911562e-05}
{"train_loss": 0.05831891670823097, "global_step": 142331, "epoch": 1599, "lr": 4.631795051635619e-05}
{"train_loss": 0.1295953392982483, "global_step": 142332, "epoch": 1599, "lr": 4.631737235409179e-05}
{"train_loss": 0.09141858667135239, "global_step": 142333, "epoch": 1599, "lr": 4.6316794192322456e-05}
{"train_loss": 0.08863694965839386, "global_step": 142334, "epoch": 1599, "lr": 4.6316216031048304e-05}
{"train_loss": 0.06294237077236176, "global_step": 142335, "epoch": 1599, "lr": 4.631563787026938e-05}
{"train_loss": 0.06498254090547562, "global_step": 142336, "epoch": 1599, "lr": 4.631505970998577e-05}
{"train_loss": 0.09046841412782669, "global_step": 142337, "epoch": 1599, "lr": 4.6314481550197556e-05}
{"train_loss": 0.13883575797080994, "global_step": 142338, "epoch": 1599, "lr": 4.6313903390904816e-05}
{"train_loss": 0.11647345125675201, "global_step": 142339, "epoch": 1599, "lr": 4.631332523210763e-05}
{"train_loss": 0.10423633456230164, "global_step": 142340, "epoch": 1599, "lr": 4.6312747073806084e-05}
{"train_loss": 0.08563582599163055, "global_step": 142341, "epoch": 1599, "lr": 4.631216891600021e-05}
{"train_loss": 0.1451243758201599, "global_step": 142342, "epoch": 1599, "lr": 4.631159075869016e-05}
{"train_loss": 0.1468134969472885, "global_step": 142343, "epoch": 1599, "lr": 4.6311012601875934e-05}
{"train_loss": 0.08953983336687088, "global_step": 142344, "epoch": 1599, "lr": 4.631043444555766e-05}
{"train_loss": 0.0770106166601181, "global_step": 142345, "epoch": 1599, "lr": 4.630985628973542e-05}
{"train_loss": 0.09839188307523727, "global_step": 142346, "epoch": 1599, "lr": 4.630927813440925e-05}
{"train_loss": 0.17857490479946136, "global_step": 142347, "epoch": 1599, "lr": 4.630869997957928e-05}
{"train_loss": 0.0971052497625351, "global_step": 142348, "epoch": 1599, "lr": 4.6308121825245534e-05}
{"train_loss": 0.14513561129570007, "global_step": 142349, "epoch": 1599, "lr": 4.630754367140814e-05}
{"train_loss": 0.113815538585186, "global_step": 142350, "epoch": 1599, "lr": 4.630696551806712e-05}
{"train_loss": 0.11593255400657654, "global_step": 142351, "epoch": 1599, "lr": 4.630638736522262e-05}
{"train_loss": 0.13221105933189392, "global_step": 142352, "epoch": 1599, "lr": 4.630580921287465e-05}
{"train_loss": 0.058259446173906326, "global_step": 142353, "epoch": 1599, "lr": 4.630523106102335e-05}
{"train_loss": 0.07420126348733902, "global_step": 142354, "epoch": 1599, "lr": 4.630465290966875e-05}
{"train_loss": 0.09562551230192184, "global_step": 142355, "epoch": 1599, "lr": 4.630407475881095e-05}
{"train_loss": 0.13517139852046967, "global_step": 142356, "epoch": 1599, "lr": 4.630349660845001e-05}
{"train_loss": 0.10579276829957962, "global_step": 142357, "epoch": 1599, "lr": 4.6302918458586026e-05}
{"train_loss": 0.12452694028615952, "global_step": 142358, "epoch": 1599, "lr": 4.630234030921907e-05}
{"train_loss": 0.11998011916875839, "global_step": 142359, "epoch": 1599, "lr": 4.6301762160349226e-05}
{"train_loss": 0.10024016350507736, "global_step": 142360, "epoch": 1599, "lr": 4.630118401197655e-05}
{"train_loss": 0.17663797736167908, "global_step": 142361, "epoch": 1599, "lr": 4.630060586410113e-05}
{"train_loss": 0.09084902703762054, "global_step": 142362, "epoch": 1599, "lr": 4.6300027716723075e-05}
{"train_loss": 0.08768118917942047, "global_step": 142363, "epoch": 1599, "lr": 4.629944956984241e-05}
{"train_loss": 0.11778397858142853, "global_step": 142364, "epoch": 1599, "lr": 4.629887142345927e-05}
{"train_loss": 0.07470421493053436, "global_step": 142365, "epoch": 1599, "lr": 4.629829327757367e-05}
{"train_loss": 0.11601509898900986, "global_step": 142366, "epoch": 1599, "lr": 4.6297715132185736e-05}
{"train_loss": 0.09330319613218307, "global_step": 142367, "epoch": 1599, "lr": 4.629713698729552e-05}
{"train_loss": 0.09609027206897736, "global_step": 142368, "epoch": 1599, "lr": 4.6296558842903125e-05}
{"train_loss": 0.1336331069469452, "global_step": 142369, "epoch": 1599, "lr": 4.629598069900859e-05}
{"train_loss": 0.12802638113498688, "global_step": 142370, "epoch": 1599, "lr": 4.629540255561204e-05}
{"train_loss": 0.08513001352548599, "global_step": 142371, "epoch": 1599, "lr": 4.62948244127135e-05}
{"train_loss": 0.14726339280605316, "global_step": 142372, "epoch": 1599, "lr": 4.62942462703131e-05}
{"train_loss": 0.13809026777744293, "global_step": 142373, "epoch": 1599, "lr": 4.6293668128410885e-05}
{"train_loss": 0.16953228414058685, "global_step": 142374, "epoch": 1599, "lr": 4.6293089987006947e-05}
{"train_loss": 0.14662624895572662, "global_step": 142375, "epoch": 1599, "lr": 4.629251184610135e-05}
{"train_loss": 0.14816327393054962, "global_step": 142376, "epoch": 1599, "lr": 4.629193370569418e-05}
{"train_loss": 0.07858141511678696, "global_step": 142377, "epoch": 1599, "lr": 4.629135556578551e-05}
{"train_loss": 0.07576783746480942, "global_step": 142378, "epoch": 1599, "lr": 4.6290777426375445e-05}
{"train_loss": 0.051153577864170074, "global_step": 142379, "epoch": 1599, "lr": 4.629019928746401e-05}
{"train_loss": 0.1151244193315506, "global_step": 142380, "epoch": 1599, "lr": 4.628962114905131e-05}
{"train_loss": 0.12443491816520691, "global_step": 142381, "epoch": 1599, "lr": 4.628904301113745e-05}
{"train_loss": 0.07229305803775787, "global_step": 142382, "epoch": 1599, "lr": 4.6288464873722465e-05}
{"train_loss": 0.1546519547700882, "global_step": 142383, "epoch": 1599, "lr": 4.6287886736806466e-05}
{"train_loss": 0.09019967913627625, "global_step": 142384, "epoch": 1599, "lr": 4.62873086003895e-05}
{"train_loss": 0.07328400760889053, "global_step": 142385, "epoch": 1599, "lr": 4.6286730464471676e-05}
{"train_loss": 0.1420614868402481, "global_step": 142386, "epoch": 1599, "lr": 4.6286152329053035e-05}
{"train_loss": 0.1629003882408142, "global_step": 142387, "epoch": 1599, "lr": 4.62855741941337e-05}
{"train_loss": 0.09227227419614792, "global_step": 142388, "epoch": 1599, "lr": 4.6284996059713706e-05}
{"train_loss": 0.1249341070652008, "global_step": 142389, "epoch": 1599, "lr": 4.628441792579316e-05}
{"train_loss": 0.11456330120563507, "global_step": 142390, "epoch": 1599, "lr": 4.6283839792372114e-05}
{"train_loss": 0.08843321353197098, "global_step": 142391, "epoch": 1599, "lr": 4.628326165945068e-05}
{"train_loss": 0.21677108108997345, "global_step": 142392, "epoch": 1599, "lr": 4.628268352702891e-05}
{"train_loss": 0.0996396392583847, "global_step": 142393, "epoch": 1599, "lr": 4.628210539510688e-05}
{"train_loss": 0.11559877544641495, "global_step": 142394, "epoch": 1599, "lr": 4.6281527263684685e-05}
{"train_loss": 0.07017107307910919, "global_step": 142395, "epoch": 1599, "lr": 4.6280949132762386e-05}
{"train_loss": 0.15228451788425446, "global_step": 142396, "epoch": 1599, "lr": 4.628037100234007e-05}
{"train_loss": 0.06111834943294525, "global_step": 142397, "epoch": 1599, "lr": 4.62797928724178e-05}
{"train_loss": 0.10602252185344696, "global_step": 142398, "epoch": 1599, "lr": 4.62792147429957e-05}
{"train_loss": 0.10647478611784035, "global_step": 142399, "epoch": 1599, "lr": 4.627863661407378e-05, "val_loss": 6.143744945526123}
{"train_loss": 0.14906053245067596, "global_step": 142400, "epoch": 1600, "lr": 4.627805848565217e-05}
{"train_loss": 0.07539156824350357, "global_step": 142401, "epoch": 1600, "lr": 4.627748035773092e-05}
{"train_loss": 0.09314046055078506, "global_step": 142402, "epoch": 1600, "lr": 4.627690223031014e-05}
{"train_loss": 0.06886414438486099, "global_step": 142403, "epoch": 1600, "lr": 4.627632410338985e-05}
{"train_loss": 0.0933556854724884, "global_step": 142404, "epoch": 1600, "lr": 4.62757459769702e-05}
{"train_loss": 0.09205936640501022, "global_step": 142405, "epoch": 1600, "lr": 4.6275167851051194e-05}
{"train_loss": 0.07892433553934097, "global_step": 142406, "epoch": 1600, "lr": 4.6274589725632976e-05}
{"train_loss": 0.07320453971624374, "global_step": 142407, "epoch": 1600, "lr": 4.627401160071557e-05}
{"train_loss": 0.0699552372097969, "global_step": 142408, "epoch": 1600, "lr": 4.62734334762991e-05}
{"train_loss": 0.052277572453022, "global_step": 142409, "epoch": 1600, "lr": 4.62728553523836e-05}
{"train_loss": 0.048923879861831665, "global_step": 142410, "epoch": 1600, "lr": 4.62722772289692e-05}
{"train_loss": 0.14060378074645996, "global_step": 142411, "epoch": 1600, "lr": 4.627169910605591e-05}
{"train_loss": 0.1273641139268875, "global_step": 142412, "epoch": 1600, "lr": 4.627112098364387e-05}
{"train_loss": 0.15341810882091522, "global_step": 142413, "epoch": 1600, "lr": 4.627054286173312e-05}
{"train_loss": 0.17724308371543884, "global_step": 142414, "epoch": 1600, "lr": 4.626996474032376e-05}
{"train_loss": 0.08478720486164093, "global_step": 142415, "epoch": 1600, "lr": 4.626938661941586e-05}
{"train_loss": 0.0825105756521225, "global_step": 142416, "epoch": 1600, "lr": 4.626880849900947e-05}
{"train_loss": 0.10506998747587204, "global_step": 142417, "epoch": 1600, "lr": 4.626823037910472e-05}
{"train_loss": 0.09327458590269089, "global_step": 142418, "epoch": 1600, "lr": 4.626765225970164e-05}
{"train_loss": 0.11053940653800964, "global_step": 142419, "epoch": 1600, "lr": 4.6267074140800354e-05}
{"train_loss": 0.11781571805477142, "global_step": 142420, "epoch": 1600, "lr": 4.626649602240089e-05}
{"train_loss": 0.09438006579875946, "global_step": 142421, "epoch": 1600, "lr": 4.626591790450337e-05}
{"train_loss": 0.0792403295636177, "global_step": 142422, "epoch": 1600, "lr": 4.626533978710783e-05}
{"train_loss": 0.1472056359052658, "global_step": 142423, "epoch": 1600, "lr": 4.62647616702144e-05}
{"train_loss": 0.1351967453956604, "global_step": 142424, "epoch": 1600, "lr": 4.626418355382309e-05}
{"train_loss": 0.16823497414588928, "global_step": 142425, "epoch": 1600, "lr": 4.626360543793405e-05}
{"train_loss": 0.12840883433818817, "global_step": 142426, "epoch": 1600, "lr": 4.6263027322547295e-05}
{"train_loss": 0.12083534151315689, "global_step": 142427, "epoch": 1600, "lr": 4.6262449207662955e-05}
{"train_loss": 0.059765178710222244, "global_step": 142428, "epoch": 1600, "lr": 4.6261871093281054e-05}
{"train_loss": 0.08837644010782242, "global_step": 142429, "epoch": 1600, "lr": 4.6261292979401735e-05}
{"train_loss": 0.09435213357210159, "global_step": 142430, "epoch": 1600, "lr": 4.6260714866025006e-05}
{"train_loss": 0.08473271131515503, "global_step": 142431, "epoch": 1600, "lr": 4.6260136753151e-05}
{"train_loss": 0.10096637904644012, "global_step": 142432, "epoch": 1600, "lr": 4.6259558640779766e-05}
{"train_loss": 0.13997428119182587, "global_step": 142433, "epoch": 1600, "lr": 4.625898052891138e-05}
{"train_loss": 0.06685265898704529, "global_step": 142434, "epoch": 1600, "lr": 4.625840241754595e-05}
{"train_loss": 0.14474134147167206, "global_step": 142435, "epoch": 1600, "lr": 4.625782430668351e-05}
{"train_loss": 0.12559260427951813, "global_step": 142436, "epoch": 1600, "lr": 4.625724619632418e-05}
{"train_loss": 0.0696631595492363, "global_step": 142437, "epoch": 1600, "lr": 4.625666808646799e-05}
{"train_loss": 0.11926361173391342, "global_step": 142438, "epoch": 1600, "lr": 4.625608997711507e-05}
{"train_loss": 0.08894862979650497, "global_step": 142439, "epoch": 1600, "lr": 4.6255511868265456e-05}
{"train_loss": 0.05746208876371384, "global_step": 142440, "epoch": 1600, "lr": 4.625493375991926e-05}
{"train_loss": 0.07102897018194199, "global_step": 142441, "epoch": 1600, "lr": 4.625435565207652e-05}
{"train_loss": 0.09362123906612396, "global_step": 142442, "epoch": 1600, "lr": 4.625377754473737e-05}
{"train_loss": 0.12843789160251617, "global_step": 142443, "epoch": 1600, "lr": 4.625319943790182e-05}
{"train_loss": 0.09900059551000595, "global_step": 142444, "epoch": 1600, "lr": 4.6252621331570006e-05}
{"train_loss": 0.0895179957151413, "global_step": 142445, "epoch": 1600, "lr": 4.6252043225741966e-05}
{"train_loss": 0.10864411294460297, "global_step": 142446, "epoch": 1600, "lr": 4.625146512041781e-05}
{"train_loss": 0.17304003238677979, "global_step": 142447, "epoch": 1600, "lr": 4.625088701559757e-05}
{"train_loss": 0.1679016500711441, "global_step": 142448, "epoch": 1600, "lr": 4.625030891128138e-05}
{"train_loss": 0.051810476928949356, "global_step": 142449, "epoch": 1600, "lr": 4.624973080746927e-05}
{"train_loss": 0.10114489495754242, "global_step": 142450, "epoch": 1600, "lr": 4.6249152704161365e-05}
{"train_loss": 0.09828310459852219, "global_step": 142451, "epoch": 1600, "lr": 4.6248574601357694e-05}
{"train_loss": 0.0932055190205574, "global_step": 142452, "epoch": 1600, "lr": 4.624799649905835e-05}
{"train_loss": 0.11046276986598969, "global_step": 142453, "epoch": 1600, "lr": 4.624741839726344e-05}
{"train_loss": 0.10197581350803375, "global_step": 142454, "epoch": 1600, "lr": 4.6246840295973e-05}
{"train_loss": 0.12335450947284698, "global_step": 142455, "epoch": 1600, "lr": 4.624626219518714e-05}
{"train_loss": 0.08477890491485596, "global_step": 142456, "epoch": 1600, "lr": 4.624568409490591e-05}
{"train_loss": 0.16548964381217957, "global_step": 142457, "epoch": 1600, "lr": 4.6245105995129424e-05}
{"train_loss": 0.08324908465147018, "global_step": 142458, "epoch": 1600, "lr": 4.624452789585771e-05}
{"train_loss": 0.09475256502628326, "global_step": 142459, "epoch": 1600, "lr": 4.6243949797090896e-05}
{"train_loss": 0.12519492208957672, "global_step": 142460, "epoch": 1600, "lr": 4.624337169882902e-05}
{"train_loss": 0.14710532128810883, "global_step": 142461, "epoch": 1600, "lr": 4.6242793601072196e-05}
{"train_loss": 0.11928506940603256, "global_step": 142462, "epoch": 1600, "lr": 4.6242215503820466e-05}
{"train_loss": 0.04569948464632034, "global_step": 142463, "epoch": 1600, "lr": 4.6241637407073934e-05}
{"train_loss": 0.07405378669500351, "global_step": 142464, "epoch": 1600, "lr": 4.624105931083266e-05}
{"train_loss": 0.14122115075588226, "global_step": 142465, "epoch": 1600, "lr": 4.624048121509674e-05}
{"train_loss": 0.1487443745136261, "global_step": 142466, "epoch": 1600, "lr": 4.623990311986623e-05}
{"train_loss": 0.06831545382738113, "global_step": 142467, "epoch": 1600, "lr": 4.623932502514124e-05}
{"train_loss": 0.13151612877845764, "global_step": 142468, "epoch": 1600, "lr": 4.62387469309218e-05}
{"train_loss": 0.09253381937742233, "global_step": 142469, "epoch": 1600, "lr": 4.623816883720803e-05}
{"train_loss": 0.14860674738883972, "global_step": 142470, "epoch": 1600, "lr": 4.623759074399999e-05}
{"train_loss": 0.12433072179555893, "global_step": 142471, "epoch": 1600, "lr": 4.6237012651297746e-05}
{"train_loss": 0.16787362098693848, "global_step": 142472, "epoch": 1600, "lr": 4.623643455910141e-05}
{"train_loss": 0.11114002019166946, "global_step": 142473, "epoch": 1600, "lr": 4.6235856467411024e-05}
{"train_loss": 0.10621035844087601, "global_step": 142474, "epoch": 1600, "lr": 4.62352783762267e-05}
{"train_loss": 0.10374021530151367, "global_step": 142475, "epoch": 1600, "lr": 4.6234700285548474e-05}
{"train_loss": 0.13789620995521545, "global_step": 142476, "epoch": 1600, "lr": 4.6234122195376476e-05}
{"train_loss": 0.10009043663740158, "global_step": 142477, "epoch": 1600, "lr": 4.6233544105710726e-05}
{"train_loss": 0.09928009659051895, "global_step": 142478, "epoch": 1600, "lr": 4.623296601655135e-05}
{"train_loss": 0.07112211734056473, "global_step": 142479, "epoch": 1600, "lr": 4.623238792789838e-05}
{"train_loss": 0.07642827928066254, "global_step": 142480, "epoch": 1600, "lr": 4.623180983975195e-05}
{"train_loss": 0.08005931973457336, "global_step": 142481, "epoch": 1600, "lr": 4.623123175211209e-05}
{"train_loss": 0.14758458733558655, "global_step": 142482, "epoch": 1600, "lr": 4.6230653664978905e-05}
{"train_loss": 0.09306845813989639, "global_step": 142483, "epoch": 1600, "lr": 4.6230075578352444e-05}
{"train_loss": 0.1564483493566513, "global_step": 142484, "epoch": 1600, "lr": 4.622949749223283e-05}
{"train_loss": 0.1613231748342514, "global_step": 142485, "epoch": 1600, "lr": 4.622891940662008e-05}
{"train_loss": 0.16036029160022736, "global_step": 142486, "epoch": 1600, "lr": 4.6228341321514345e-05}
{"train_loss": 0.12228452414274216, "global_step": 142487, "epoch": 1600, "lr": 4.6227763236915634e-05}
{"train_loss": 0.1082548453064447, "global_step": 142488, "epoch": 1600, "lr": 4.6227185152824065e-05, "train/sim_max_reward_0": 0.6416645960930053, "train/sim_max_reward_1": 0.6852234951964223, "train/sim_max_reward_2": 0.6094392254873223, "train/sim_max_reward_3": 0.7200129005285153, "train/sim_max_reward_4": 0.19944521875356488, "train/sim_max_reward_5": 0.5750402516162949, "test/sim_max_reward_4500000": 0.4589861821946362, "test/sim_max_reward_4500001": 0.9836648550533708, "test/sim_max_reward_4500002": 0.9480662702121955, "test/sim_max_reward_4500003": 0.6179631561368021, "test/sim_max_reward_4500004": 0.9452737924841788, "test/sim_max_reward_4500005": 0.9667115496524755, "test/sim_max_reward_4500006": 0.8226209127778309, "test/sim_max_reward_4500007": 0.974935078106003, "test/sim_max_reward_4500008": 0.21588588396054442, "test/sim_max_reward_4500009": 0.04988548088654186, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 1.0, "test/sim_max_reward_4500013": 0.5986845100390735, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9111290883849007, "test/sim_max_reward_4500016": 0.986526783368986, "test/sim_max_reward_4500017": 0.8826389143033979, "test/sim_max_reward_4500018": 0.15085675195028703, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9999864375341095, "test/sim_max_reward_4500023": 0.6921665433791871, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9094887518935058, "test/sim_max_reward_4500026": 0.9574936395638899, "test/sim_max_reward_4500027": 0.9627905042333214, "test/sim_max_reward_4500028": 0.5296941259052506, "test/sim_max_reward_4500029": 0.40784137240087764, "test/sim_max_reward_4500030": 0.28801082309427195, "test/sim_max_reward_4500031": 0.4731189055445301, "test/sim_max_reward_4500032": 0.6461299956804547, "test/sim_max_reward_4500033": 0.9821056110913678, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.5687843371440258, "test/sim_max_reward_4500036": 0.9956758352041859, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.2181658417782123, "test/sim_max_reward_4500039": 0.9907543669203215, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.66006971661858, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.2826987333264295, "test/sim_max_reward_4500044": 0.6830346514368244, "test/sim_max_reward_4500045": 0.9636319804628439, "test/sim_max_reward_4500046": 0.4260263241651776, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.9955672799313124, "train/mean_score": 0.5718042812791876, "test/mean_score": 0.6629412997363981, "val_loss": 6.184322357177734, "train_action_mse_error": 14.015870094299316}
{"train_loss": 0.08716470748186111, "global_step": 142489, "epoch": 1601, "lr": 4.6226607069239714e-05}
{"train_loss": 0.09530167281627655, "global_step": 142490, "epoch": 1601, "lr": 4.6226028986162626e-05}
{"train_loss": 0.07836945354938507, "global_step": 142491, "epoch": 1601, "lr": 4.622545090359293e-05}
{"train_loss": 0.11665540933609009, "global_step": 142492, "epoch": 1601, "lr": 4.622487282153065e-05}
{"train_loss": 0.11932934075593948, "global_step": 142493, "epoch": 1601, "lr": 4.622429473997592e-05}
{"train_loss": 0.103688545525074, "global_step": 142494, "epoch": 1601, "lr": 4.6223716658928756e-05}
{"train_loss": 0.1212167963385582, "global_step": 142495, "epoch": 1601, "lr": 4.622313857838929e-05}
{"train_loss": 0.21848788857460022, "global_step": 142496, "epoch": 1601, "lr": 4.622256049835756e-05}
{"train_loss": 0.10309308022260666, "global_step": 142497, "epoch": 1601, "lr": 4.622198241883368e-05}
{"train_loss": 0.07661298662424088, "global_step": 142498, "epoch": 1601, "lr": 4.622140433981769e-05}
{"train_loss": 0.060745563358068466, "global_step": 142499, "epoch": 1601, "lr": 4.62208262613097e-05}
{"train_loss": 0.08243251591920853, "global_step": 142500, "epoch": 1601, "lr": 4.6220248183309754e-05}
{"train_loss": 0.1497146040201187, "global_step": 142501, "epoch": 1601, "lr": 4.621967010581798e-05}
{"train_loss": 0.1173451691865921, "global_step": 142502, "epoch": 1601, "lr": 4.6219092028834395e-05}
{"train_loss": 0.14120565354824066, "global_step": 142503, "epoch": 1601, "lr": 4.621851395235913e-05}
{"train_loss": 0.1101001650094986, "global_step": 142504, "epoch": 1601, "lr": 4.621793587639222e-05}
{"train_loss": 0.11082237213850021, "global_step": 142505, "epoch": 1601, "lr": 4.621735780093379e-05}
{"train_loss": 0.09732489287853241, "global_step": 142506, "epoch": 1601, "lr": 4.621677972598386e-05}
{"train_loss": 0.11566892266273499, "global_step": 142507, "epoch": 1601, "lr": 4.621620165154254e-05}
{"train_loss": 0.09020635485649109, "global_step": 142508, "epoch": 1601, "lr": 4.6215623577609935e-05}
{"train_loss": 0.09821922332048416, "global_step": 142509, "epoch": 1601, "lr": 4.621504550418606e-05}
{"train_loss": 0.15649908781051636, "global_step": 142510, "epoch": 1601, "lr": 4.621446743127105e-05}
{"train_loss": 0.18320021033287048, "global_step": 142511, "epoch": 1601, "lr": 4.621388935886494e-05}
{"train_loss": 0.06294140219688416, "global_step": 142512, "epoch": 1601, "lr": 4.621331128696785e-05}
{"train_loss": 0.12448203563690186, "global_step": 142513, "epoch": 1601, "lr": 4.621273321557981e-05}
{"train_loss": 0.08956077694892883, "global_step": 142514, "epoch": 1601, "lr": 4.6212155144700944e-05}
{"train_loss": 0.1315649002790451, "global_step": 142515, "epoch": 1601, "lr": 4.621157707433128e-05}
{"train_loss": 0.08993684500455856, "global_step": 142516, "epoch": 1601, "lr": 4.621099900447095e-05}
{"train_loss": 0.134517639875412, "global_step": 142517, "epoch": 1601, "lr": 4.621042093511998e-05}
{"train_loss": 0.062016282230615616, "global_step": 142518, "epoch": 1601, "lr": 4.62098428662785e-05}
{"train_loss": 0.10738620162010193, "global_step": 142519, "epoch": 1601, "lr": 4.6209264797946534e-05}
{"train_loss": 0.17051033675670624, "global_step": 142520, "epoch": 1601, "lr": 4.6208686730124205e-05}
{"train_loss": 0.10653355717658997, "global_step": 142521, "epoch": 1601, "lr": 4.620810866281156e-05}
{"train_loss": 0.12143731862306595, "global_step": 142522, "epoch": 1601, "lr": 4.6207530596008706e-05}
{"train_loss": 0.06240832060575485, "global_step": 142523, "epoch": 1601, "lr": 4.620695252971568e-05}
{"train_loss": 0.09119722992181778, "global_step": 142524, "epoch": 1601, "lr": 4.62063744639326e-05}
{"train_loss": 0.07338500767946243, "global_step": 142525, "epoch": 1601, "lr": 4.620579639865951e-05}
{"train_loss": 0.09728343784809113, "global_step": 142526, "epoch": 1601, "lr": 4.6205218333896524e-05}
{"train_loss": 0.08385492861270905, "global_step": 142527, "epoch": 1601, "lr": 4.620464026964369e-05}
{"train_loss": 0.10758662968873978, "global_step": 142528, "epoch": 1601, "lr": 4.6204062205901095e-05}
{"train_loss": 0.09682992100715637, "global_step": 142529, "epoch": 1601, "lr": 4.620348414266883e-05}
{"train_loss": 0.14733193814754486, "global_step": 142530, "epoch": 1601, "lr": 4.620290607994694e-05}
{"train_loss": 0.15504848957061768, "global_step": 142531, "epoch": 1601, "lr": 4.620232801773555e-05}
{"train_loss": 0.08645249903202057, "global_step": 142532, "epoch": 1601, "lr": 4.620174995603468e-05}
{"train_loss": 0.1209721714258194, "global_step": 142533, "epoch": 1601, "lr": 4.620117189484446e-05}
{"train_loss": 0.10528367757797241, "global_step": 142534, "epoch": 1601, "lr": 4.620059383416493e-05}
{"train_loss": 0.12009991705417633, "global_step": 142535, "epoch": 1601, "lr": 4.62000157739962e-05}
{"train_loss": 0.16602835059165955, "global_step": 142536, "epoch": 1601, "lr": 4.6199437714338316e-05}
{"train_loss": 0.13857224583625793, "global_step": 142537, "epoch": 1601, "lr": 4.619885965519139e-05}
{"train_loss": 0.18330098688602448, "global_step": 142538, "epoch": 1601, "lr": 4.619828159655546e-05}
{"train_loss": 0.084465391933918, "global_step": 142539, "epoch": 1601, "lr": 4.619770353843065e-05}
{"train_loss": 0.170610249042511, "global_step": 142540, "epoch": 1601, "lr": 4.6197125480816986e-05}
{"train_loss": 0.11896829307079315, "global_step": 142541, "epoch": 1601, "lr": 4.619654742371459e-05}
{"train_loss": 0.11048524081707001, "global_step": 142542, "epoch": 1601, "lr": 4.6195969367123505e-05}
{"train_loss": 0.12797746062278748, "global_step": 142543, "epoch": 1601, "lr": 4.619539131104383e-05}
{"train_loss": 0.11145385354757309, "global_step": 142544, "epoch": 1601, "lr": 4.619481325547565e-05}
{"train_loss": 0.08748223632574081, "global_step": 142545, "epoch": 1601, "lr": 4.619423520041902e-05}
{"train_loss": 0.06635749340057373, "global_step": 142546, "epoch": 1601, "lr": 4.6193657145874036e-05}
{"train_loss": 0.09128612279891968, "global_step": 142547, "epoch": 1601, "lr": 4.6193079091840766e-05}
{"train_loss": 0.07406992465257645, "global_step": 142548, "epoch": 1601, "lr": 4.61925010383193e-05}
{"train_loss": 0.14973321557044983, "global_step": 142549, "epoch": 1601, "lr": 4.619192298530969e-05}
{"train_loss": 0.10931511223316193, "global_step": 142550, "epoch": 1601, "lr": 4.619134493281205e-05}
{"train_loss": 0.09363026171922684, "global_step": 142551, "epoch": 1601, "lr": 4.619076688082641e-05}
{"train_loss": 0.09944627434015274, "global_step": 142552, "epoch": 1601, "lr": 4.61901888293529e-05}
{"train_loss": 0.09655493497848511, "global_step": 142553, "epoch": 1601, "lr": 4.618961077839155e-05}
{"train_loss": 0.09884664416313171, "global_step": 142554, "epoch": 1601, "lr": 4.618903272794248e-05}
{"train_loss": 0.06840578466653824, "global_step": 142555, "epoch": 1601, "lr": 4.618845467800573e-05}
{"train_loss": 0.12539778649806976, "global_step": 142556, "epoch": 1601, "lr": 4.6187876628581415e-05}
{"train_loss": 0.07788776606321335, "global_step": 142557, "epoch": 1601, "lr": 4.618729857966957e-05}
{"train_loss": 0.08497582376003265, "global_step": 142558, "epoch": 1601, "lr": 4.6186720531270324e-05}
{"train_loss": 0.08076277375221252, "global_step": 142559, "epoch": 1601, "lr": 4.61861424833837e-05}
{"train_loss": 0.09102249890565872, "global_step": 142560, "epoch": 1601, "lr": 4.6185564436009805e-05}
{"train_loss": 0.10817378014326096, "global_step": 142561, "epoch": 1601, "lr": 4.6184986389148736e-05}
{"train_loss": 0.13362866640090942, "global_step": 142562, "epoch": 1601, "lr": 4.618440834280054e-05}
{"train_loss": 0.13282345235347748, "global_step": 142563, "epoch": 1601, "lr": 4.61838302969653e-05}
{"train_loss": 0.10455973446369171, "global_step": 142564, "epoch": 1601, "lr": 4.61832522516431e-05}
{"train_loss": 0.0980195552110672, "global_step": 142565, "epoch": 1601, "lr": 4.618267420683402e-05}
{"train_loss": 0.07479214668273926, "global_step": 142566, "epoch": 1601, "lr": 4.6182096162538116e-05}
{"train_loss": 0.06077631935477257, "global_step": 142567, "epoch": 1601, "lr": 4.6181518118755504e-05}
{"train_loss": 0.13807706534862518, "global_step": 142568, "epoch": 1601, "lr": 4.618094007548622e-05}
{"train_loss": 0.10133282095193863, "global_step": 142569, "epoch": 1601, "lr": 4.618036203273039e-05}
{"train_loss": 0.14336533844470978, "global_step": 142570, "epoch": 1601, "lr": 4.6179783990488034e-05}
{"train_loss": 0.05874309688806534, "global_step": 142571, "epoch": 1601, "lr": 4.617920594875928e-05}
{"train_loss": 0.12328317016363144, "global_step": 142572, "epoch": 1601, "lr": 4.617862790754417e-05}
{"train_loss": 0.10221049934625626, "global_step": 142573, "epoch": 1601, "lr": 4.617804986684281e-05}
{"train_loss": 0.1744152009487152, "global_step": 142574, "epoch": 1601, "lr": 4.617747182665525e-05}
{"train_loss": 0.11552757769823074, "global_step": 142575, "epoch": 1601, "lr": 4.61768937869816e-05}
{"train_loss": 0.12823402881622314, "global_step": 142576, "epoch": 1601, "lr": 4.6176315747821904e-05}
{"train_loss": 0.10975140236904112, "global_step": 142577, "epoch": 1601, "lr": 4.617573770917627e-05, "val_loss": 6.013136386871338}
{"train_loss": 0.14602206647396088, "global_step": 142578, "epoch": 1602, "lr": 4.617515967104474e-05}
{"train_loss": 0.11924678087234497, "global_step": 142579, "epoch": 1602, "lr": 4.617458163342742e-05}
{"train_loss": 0.10124624520540237, "global_step": 142580, "epoch": 1602, "lr": 4.61740035963244e-05}
{"train_loss": 0.13424274325370789, "global_step": 142581, "epoch": 1602, "lr": 4.6173425559735725e-05}
{"train_loss": 0.11184534430503845, "global_step": 142582, "epoch": 1602, "lr": 4.617284752366149e-05}
{"train_loss": 0.11664849519729614, "global_step": 142583, "epoch": 1602, "lr": 4.617226948810176e-05}
{"train_loss": 0.1186431422829628, "global_step": 142584, "epoch": 1602, "lr": 4.617169145305663e-05}
{"train_loss": 0.12810690701007843, "global_step": 142585, "epoch": 1602, "lr": 4.617111341852617e-05}
{"train_loss": 0.12971603870391846, "global_step": 142586, "epoch": 1602, "lr": 4.617053538451047e-05}
{"train_loss": 0.06254076212644577, "global_step": 142587, "epoch": 1602, "lr": 4.616995735100956e-05}
{"train_loss": 0.12171625345945358, "global_step": 142588, "epoch": 1602, "lr": 4.616937931802359e-05}
{"train_loss": 0.09117960929870605, "global_step": 142589, "epoch": 1602, "lr": 4.6168801285552574e-05}
{"train_loss": 0.08489767462015152, "global_step": 142590, "epoch": 1602, "lr": 4.616822325359663e-05}
{"train_loss": 0.07502349466085434, "global_step": 142591, "epoch": 1602, "lr": 4.616764522215581e-05}
{"train_loss": 0.06576430797576904, "global_step": 142592, "epoch": 1602, "lr": 4.616706719123022e-05}
{"train_loss": 0.09944575279951096, "global_step": 142593, "epoch": 1602, "lr": 4.61664891608199e-05}
{"train_loss": 0.12172818183898926, "global_step": 142594, "epoch": 1602, "lr": 4.616591113092497e-05}
{"train_loss": 0.12473271042108536, "global_step": 142595, "epoch": 1602, "lr": 4.616533310154546e-05}
{"train_loss": 0.07111669331789017, "global_step": 142596, "epoch": 1602, "lr": 4.6164755072681504e-05}
{"train_loss": 0.10419727861881256, "global_step": 142597, "epoch": 1602, "lr": 4.6164177044333126e-05}
{"train_loss": 0.1112745925784111, "global_step": 142598, "epoch": 1602, "lr": 4.616359901650042e-05}
{"train_loss": 0.07463338971138, "global_step": 142599, "epoch": 1602, "lr": 4.616302098918351e-05}
{"train_loss": 0.05694388598203659, "global_step": 142600, "epoch": 1602, "lr": 4.6162442962382414e-05}
{"train_loss": 0.09043988585472107, "global_step": 142601, "epoch": 1602, "lr": 4.6161864936097226e-05}
{"train_loss": 0.09911563247442245, "global_step": 142602, "epoch": 1602, "lr": 4.6161286910328036e-05}
{"train_loss": 0.05280282348394394, "global_step": 142603, "epoch": 1602, "lr": 4.616070888507491e-05}
{"train_loss": 0.07062909752130508, "global_step": 142604, "epoch": 1602, "lr": 4.616013086033794e-05}
{"train_loss": 0.062026940286159515, "global_step": 142605, "epoch": 1602, "lr": 4.615955283611719e-05}
{"train_loss": 0.07023796439170837, "global_step": 142606, "epoch": 1602, "lr": 4.6158974812412724e-05}
{"train_loss": 0.08785117417573929, "global_step": 142607, "epoch": 1602, "lr": 4.615839678922466e-05}
{"train_loss": 0.1777641326189041, "global_step": 142608, "epoch": 1602, "lr": 4.6157818766553036e-05}
{"train_loss": 0.08405281603336334, "global_step": 142609, "epoch": 1602, "lr": 4.615724074439796e-05}
{"train_loss": 0.16096962988376617, "global_step": 142610, "epoch": 1602, "lr": 4.615666272275948e-05}
{"train_loss": 0.0972323790192604, "global_step": 142611, "epoch": 1602, "lr": 4.615608470163771e-05}
{"train_loss": 0.0955796167254448, "global_step": 142612, "epoch": 1602, "lr": 4.615550668103268e-05}
{"train_loss": 0.09378733485937119, "global_step": 142613, "epoch": 1602, "lr": 4.615492866094453e-05}
{"train_loss": 0.10173829644918442, "global_step": 142614, "epoch": 1602, "lr": 4.615435064137327e-05}
{"train_loss": 0.08491839468479156, "global_step": 142615, "epoch": 1602, "lr": 4.615377262231901e-05}
{"train_loss": 0.0795145034790039, "global_step": 142616, "epoch": 1602, "lr": 4.615319460378186e-05}
{"train_loss": 0.10030508041381836, "global_step": 142617, "epoch": 1602, "lr": 4.615261658576184e-05}
{"train_loss": 0.14886543154716492, "global_step": 142618, "epoch": 1602, "lr": 4.615203856825907e-05}
{"train_loss": 0.14958710968494415, "global_step": 142619, "epoch": 1602, "lr": 4.6151460551273603e-05}
{"train_loss": 0.1047920361161232, "global_step": 142620, "epoch": 1602, "lr": 4.6150882534805534e-05}
{"train_loss": 0.11723916232585907, "global_step": 142621, "epoch": 1602, "lr": 4.6150304518854926e-05}
{"train_loss": 0.11652606725692749, "global_step": 142622, "epoch": 1602, "lr": 4.6149726503421856e-05}
{"train_loss": 0.12008687108755112, "global_step": 142623, "epoch": 1602, "lr": 4.614914848850642e-05}
{"train_loss": 0.08356308937072754, "global_step": 142624, "epoch": 1602, "lr": 4.614857047410868e-05}
{"train_loss": 0.08594594150781631, "global_step": 142625, "epoch": 1602, "lr": 4.614799246022871e-05}
{"train_loss": 0.10109703987836838, "global_step": 142626, "epoch": 1602, "lr": 4.6147414446866614e-05}
{"train_loss": 0.04283595085144043, "global_step": 142627, "epoch": 1602, "lr": 4.614683643402243e-05}
{"train_loss": 0.14472757279872894, "global_step": 142628, "epoch": 1602, "lr": 4.6146258421696275e-05}
{"train_loss": 0.13858428597450256, "global_step": 142629, "epoch": 1602, "lr": 4.614568040988819e-05}
{"train_loss": 0.13341748714447021, "global_step": 142630, "epoch": 1602, "lr": 4.614510239859829e-05}
{"train_loss": 0.18455007672309875, "global_step": 142631, "epoch": 1602, "lr": 4.6144524387826616e-05}
{"train_loss": 0.08222848176956177, "global_step": 142632, "epoch": 1602, "lr": 4.6143946377573274e-05}
{"train_loss": 0.13164815306663513, "global_step": 142633, "epoch": 1602, "lr": 4.614336836783832e-05}
{"train_loss": 0.11116226762533188, "global_step": 142634, "epoch": 1602, "lr": 4.6142790358621845e-05}
{"train_loss": 0.0951961800456047, "global_step": 142635, "epoch": 1602, "lr": 4.614221234992395e-05}
{"train_loss": 0.12825606763362885, "global_step": 142636, "epoch": 1602, "lr": 4.614163434174466e-05}
{"train_loss": 0.09494538605213165, "global_step": 142637, "epoch": 1602, "lr": 4.614105633408409e-05}
{"train_loss": 0.1634603589773178, "global_step": 142638, "epoch": 1602, "lr": 4.6140478326942307e-05}
{"train_loss": 0.1768348664045334, "global_step": 142639, "epoch": 1602, "lr": 4.6139900320319395e-05}
{"train_loss": 0.03848125785589218, "global_step": 142640, "epoch": 1602, "lr": 4.6139322314215424e-05}
{"train_loss": 0.11260790377855301, "global_step": 142641, "epoch": 1602, "lr": 4.6138744308630474e-05}
{"train_loss": 0.11189018189907074, "global_step": 142642, "epoch": 1602, "lr": 4.613816630356462e-05}
{"train_loss": 0.12190193682909012, "global_step": 142643, "epoch": 1602, "lr": 4.613758829901794e-05}
{"train_loss": 0.06512672454118729, "global_step": 142644, "epoch": 1602, "lr": 4.613701029499052e-05}
{"train_loss": 0.14411485195159912, "global_step": 142645, "epoch": 1602, "lr": 4.6136432291482437e-05}
{"train_loss": 0.08994562923908234, "global_step": 142646, "epoch": 1602, "lr": 4.613585428849375e-05}
{"train_loss": 0.09557172656059265, "global_step": 142647, "epoch": 1602, "lr": 4.613527628602457e-05}
{"train_loss": 0.09252549707889557, "global_step": 142648, "epoch": 1602, "lr": 4.613469828407493e-05}
{"train_loss": 0.08398385345935822, "global_step": 142649, "epoch": 1602, "lr": 4.613412028264496e-05}
{"train_loss": 0.09379904717206955, "global_step": 142650, "epoch": 1602, "lr": 4.6133542281734674e-05}
{"train_loss": 0.09959588944911957, "global_step": 142651, "epoch": 1602, "lr": 4.61329642813442e-05}
{"train_loss": 0.11812617629766464, "global_step": 142652, "epoch": 1602, "lr": 4.6132386281473616e-05}
{"train_loss": 0.10528132319450378, "global_step": 142653, "epoch": 1602, "lr": 4.613180828212297e-05}
{"train_loss": 0.09209324419498444, "global_step": 142654, "epoch": 1602, "lr": 4.613123028329237e-05}
{"train_loss": 0.20267441868782043, "global_step": 142655, "epoch": 1602, "lr": 4.613065228498186e-05}
{"train_loss": 0.07669098675251007, "global_step": 142656, "epoch": 1602, "lr": 4.613007428719156e-05}
{"train_loss": 0.13377416133880615, "global_step": 142657, "epoch": 1602, "lr": 4.6129496289921504e-05}
{"train_loss": 0.13133953511714935, "global_step": 142658, "epoch": 1602, "lr": 4.612891829317181e-05}
{"train_loss": 0.15189529955387115, "global_step": 142659, "epoch": 1602, "lr": 4.612834029694253e-05}
{"train_loss": 0.055642105638980865, "global_step": 142660, "epoch": 1602, "lr": 4.612776230123374e-05}
{"train_loss": 0.1020960882306099, "global_step": 142661, "epoch": 1602, "lr": 4.612718430604553e-05}
{"train_loss": 0.11977928131818771, "global_step": 142662, "epoch": 1602, "lr": 4.612660631137797e-05}
{"train_loss": 0.11453214287757874, "global_step": 142663, "epoch": 1602, "lr": 4.612602831723114e-05}
{"train_loss": 0.10995922237634659, "global_step": 142664, "epoch": 1602, "lr": 4.612545032360513e-05}
{"train_loss": 0.10674121230840683, "global_step": 142665, "epoch": 1602, "lr": 4.612487233049998e-05}
{"train_loss": 0.10690125019362803, "global_step": 142666, "epoch": 1602, "lr": 4.6124294337915824e-05, "val_loss": 6.114889621734619}
{"train_loss": 0.13747675716876984, "global_step": 142667, "epoch": 1603, "lr": 4.612371634585268e-05}
{"train_loss": 0.09801163524389267, "global_step": 142668, "epoch": 1603, "lr": 4.612313835431068e-05}
{"train_loss": 0.20027872920036316, "global_step": 142669, "epoch": 1603, "lr": 4.612256036328986e-05}
{"train_loss": 0.14869913458824158, "global_step": 142670, "epoch": 1603, "lr": 4.612198237279031e-05}
{"train_loss": 0.07849828153848648, "global_step": 142671, "epoch": 1603, "lr": 4.6121404382812135e-05}
{"train_loss": 0.0632043108344078, "global_step": 142672, "epoch": 1603, "lr": 4.6120826393355363e-05}
{"train_loss": 0.07431362569332123, "global_step": 142673, "epoch": 1603, "lr": 4.612024840442013e-05}
{"train_loss": 0.13035611808300018, "global_step": 142674, "epoch": 1603, "lr": 4.611967041600645e-05}
{"train_loss": 0.06414219737052917, "global_step": 142675, "epoch": 1603, "lr": 4.6119092428114455e-05}
{"train_loss": 0.1171550527215004, "global_step": 142676, "epoch": 1603, "lr": 4.611851444074419e-05}
{"train_loss": 0.1235528513789177, "global_step": 142677, "epoch": 1603, "lr": 4.611793645389576e-05}
{"train_loss": 0.09883671998977661, "global_step": 142678, "epoch": 1603, "lr": 4.611735846756921e-05}
{"train_loss": 0.06593646854162216, "global_step": 142679, "epoch": 1603, "lr": 4.6116780481764636e-05}
{"train_loss": 0.13107968866825104, "global_step": 142680, "epoch": 1603, "lr": 4.611620249648212e-05}
{"train_loss": 0.12403889000415802, "global_step": 142681, "epoch": 1603, "lr": 4.611562451172173e-05}
{"train_loss": 0.11903117597103119, "global_step": 142682, "epoch": 1603, "lr": 4.611504652748354e-05}
{"train_loss": 0.09899989515542984, "global_step": 142683, "epoch": 1603, "lr": 4.6114468543767654e-05}
{"train_loss": 0.1279391348361969, "global_step": 142684, "epoch": 1603, "lr": 4.61138905605741e-05}
{"train_loss": 0.08908967673778534, "global_step": 142685, "epoch": 1603, "lr": 4.611331257790302e-05}
{"train_loss": 0.1340237855911255, "global_step": 142686, "epoch": 1603, "lr": 4.6112734595754426e-05}
{"train_loss": 0.10894975066184998, "global_step": 142687, "epoch": 1603, "lr": 4.611215661412843e-05}
{"train_loss": 0.13807986676692963, "global_step": 142688, "epoch": 1603, "lr": 4.6111578633025134e-05}
{"train_loss": 0.12275684624910355, "global_step": 142689, "epoch": 1603, "lr": 4.611100065244456e-05}
{"train_loss": 0.20698711276054382, "global_step": 142690, "epoch": 1603, "lr": 4.611042267238684e-05}
{"train_loss": 0.11414483189582825, "global_step": 142691, "epoch": 1603, "lr": 4.6109844692852e-05}
{"train_loss": 0.19032372534275055, "global_step": 142692, "epoch": 1603, "lr": 4.610926671384018e-05}
{"train_loss": 0.07450245320796967, "global_step": 142693, "epoch": 1603, "lr": 4.6108688735351384e-05}
{"train_loss": 0.12661412358283997, "global_step": 142694, "epoch": 1603, "lr": 4.610811075738575e-05}
{"train_loss": 0.12373355776071548, "global_step": 142695, "epoch": 1603, "lr": 4.6107532779943316e-05}
{"train_loss": 0.13979294896125793, "global_step": 142696, "epoch": 1603, "lr": 4.6106954803024204e-05}
{"train_loss": 0.1276823878288269, "global_step": 142697, "epoch": 1603, "lr": 4.610637682662844e-05}
{"train_loss": 0.08216239511966705, "global_step": 142698, "epoch": 1603, "lr": 4.610579885075614e-05}
{"train_loss": 0.13268990814685822, "global_step": 142699, "epoch": 1603, "lr": 4.610522087540736e-05}
{"train_loss": 0.0879458487033844, "global_step": 142700, "epoch": 1603, "lr": 4.610464290058219e-05}
{"train_loss": 0.05978496000170708, "global_step": 142701, "epoch": 1603, "lr": 4.61040649262807e-05}
{"train_loss": 0.18288390338420868, "global_step": 142702, "epoch": 1603, "lr": 4.610348695250298e-05}
{"train_loss": 0.11498227715492249, "global_step": 142703, "epoch": 1603, "lr": 4.610290897924908e-05}
{"train_loss": 0.1061054989695549, "global_step": 142704, "epoch": 1603, "lr": 4.6102331006519124e-05}
{"train_loss": 0.15270040929317474, "global_step": 142705, "epoch": 1603, "lr": 4.610175303431313e-05}
{"train_loss": 0.12243463099002838, "global_step": 142706, "epoch": 1603, "lr": 4.610117506263121e-05}
{"train_loss": 0.1759379357099533, "global_step": 142707, "epoch": 1603, "lr": 4.610059709147347e-05}
{"train_loss": 0.07842417806386948, "global_step": 142708, "epoch": 1603, "lr": 4.6100019120839924e-05}
{"train_loss": 0.08301687240600586, "global_step": 142709, "epoch": 1603, "lr": 4.609944115073071e-05}
{"train_loss": 0.14292897284030914, "global_step": 142710, "epoch": 1603, "lr": 4.6098863181145855e-05}
{"train_loss": 0.13966654241085052, "global_step": 142711, "epoch": 1603, "lr": 4.609828521208549e-05}
{"train_loss": 0.11600712686777115, "global_step": 142712, "epoch": 1603, "lr": 4.609770724354963e-05}
{"train_loss": 0.17564566433429718, "global_step": 142713, "epoch": 1603, "lr": 4.609712927553842e-05}
{"train_loss": 0.10269350558519363, "global_step": 142714, "epoch": 1603, "lr": 4.6096551308051864e-05}
{"train_loss": 0.11006944626569748, "global_step": 142715, "epoch": 1603, "lr": 4.609597334109011e-05}
{"train_loss": 0.11530892550945282, "global_step": 142716, "epoch": 1603, "lr": 4.6095395374653195e-05}
{"train_loss": 0.12227565050125122, "global_step": 142717, "epoch": 1603, "lr": 4.609481740874121e-05}
{"train_loss": 0.052155427634716034, "global_step": 142718, "epoch": 1603, "lr": 4.6094239443354226e-05}
{"train_loss": 0.09109188616275787, "global_step": 142719, "epoch": 1603, "lr": 4.609366147849232e-05}
{"train_loss": 0.06847630441188812, "global_step": 142720, "epoch": 1603, "lr": 4.609308351415558e-05}
{"train_loss": 0.11883494257926941, "global_step": 142721, "epoch": 1603, "lr": 4.6092505550344086e-05}
{"train_loss": 0.13096393644809723, "global_step": 142722, "epoch": 1603, "lr": 4.609192758705788e-05}
{"train_loss": 0.112423375248909, "global_step": 142723, "epoch": 1603, "lr": 4.609134962429709e-05}
{"train_loss": 0.1387176513671875, "global_step": 142724, "epoch": 1603, "lr": 4.609077166206176e-05}
{"train_loss": 0.15815167129039764, "global_step": 142725, "epoch": 1603, "lr": 4.609019370035196e-05}
{"train_loss": 0.07447268068790436, "global_step": 142726, "epoch": 1603, "lr": 4.6089615739167815e-05}
{"train_loss": 0.11412089318037033, "global_step": 142727, "epoch": 1603, "lr": 4.608903777850936e-05}
{"train_loss": 0.13117364048957825, "global_step": 142728, "epoch": 1603, "lr": 4.60884598183767e-05}
{"train_loss": 0.12271139025688171, "global_step": 142729, "epoch": 1603, "lr": 4.6087881858769874e-05}
{"train_loss": 0.10472287237644196, "global_step": 142730, "epoch": 1603, "lr": 4.608730389968902e-05}
{"train_loss": 0.09526687860488892, "global_step": 142731, "epoch": 1603, "lr": 4.608672594113414e-05}
{"train_loss": 0.12076219171285629, "global_step": 142732, "epoch": 1603, "lr": 4.608614798310539e-05}
{"train_loss": 0.09986206144094467, "global_step": 142733, "epoch": 1603, "lr": 4.608557002560277e-05}
{"train_loss": 0.11758458614349365, "global_step": 142734, "epoch": 1603, "lr": 4.6084992068626435e-05}
{"train_loss": 0.1339719295501709, "global_step": 142735, "epoch": 1603, "lr": 4.6084414112176416e-05}
{"train_loss": 0.13591422140598297, "global_step": 142736, "epoch": 1603, "lr": 4.608383615625279e-05}
{"train_loss": 0.1239553689956665, "global_step": 142737, "epoch": 1603, "lr": 4.608325820085565e-05}
{"train_loss": 0.1389404535293579, "global_step": 142738, "epoch": 1603, "lr": 4.608268024598507e-05}
{"train_loss": 0.0859125480055809, "global_step": 142739, "epoch": 1603, "lr": 4.608210229164112e-05}
{"train_loss": 0.11510562151670456, "global_step": 142740, "epoch": 1603, "lr": 4.6081524337823887e-05}
{"train_loss": 0.07913081347942352, "global_step": 142741, "epoch": 1603, "lr": 4.608094638453344e-05}
{"train_loss": 0.144998237490654, "global_step": 142742, "epoch": 1603, "lr": 4.6080368431769854e-05}
{"train_loss": 0.1702166348695755, "global_step": 142743, "epoch": 1603, "lr": 4.6079790479533234e-05}
{"train_loss": 0.1392957866191864, "global_step": 142744, "epoch": 1603, "lr": 4.6079212527823625e-05}
{"train_loss": 0.16415449976921082, "global_step": 142745, "epoch": 1603, "lr": 4.6078634576641134e-05}
{"train_loss": 0.11791612207889557, "global_step": 142746, "epoch": 1603, "lr": 4.6078056625985796e-05}
{"train_loss": 0.1486218273639679, "global_step": 142747, "epoch": 1603, "lr": 4.607747867585775e-05}
{"train_loss": 0.13428747653961182, "global_step": 142748, "epoch": 1603, "lr": 4.607690072625701e-05}
{"train_loss": 0.13582409918308258, "global_step": 142749, "epoch": 1603, "lr": 4.607632277718371e-05}
{"train_loss": 0.18100996315479279, "global_step": 142750, "epoch": 1603, "lr": 4.607574482863787e-05}
{"train_loss": 0.20191825926303864, "global_step": 142751, "epoch": 1603, "lr": 4.607516688061962e-05}
{"train_loss": 0.08955732733011246, "global_step": 142752, "epoch": 1603, "lr": 4.6074588933129e-05}
{"train_loss": 0.13140055537223816, "global_step": 142753, "epoch": 1603, "lr": 4.6074010986166126e-05}
{"train_loss": 0.13787640631198883, "global_step": 142754, "epoch": 1603, "lr": 4.607343303973103e-05}
{"train_loss": 0.12189738707763426, "global_step": 142755, "epoch": 1603, "lr": 4.607285509382383e-05, "val_loss": 6.127155303955078}
{"train_loss": 0.08950304239988327, "global_step": 142756, "epoch": 1604, "lr": 4.6072277148444584e-05}
{"train_loss": 0.08812947571277618, "global_step": 142757, "epoch": 1604, "lr": 4.6071699203593366e-05}
{"train_loss": 0.12271512299776077, "global_step": 142758, "epoch": 1604, "lr": 4.6071121259270264e-05}
{"train_loss": 0.12297248840332031, "global_step": 142759, "epoch": 1604, "lr": 4.607054331547535e-05}
{"train_loss": 0.0703435018658638, "global_step": 142760, "epoch": 1604, "lr": 4.606996537220871e-05}
{"train_loss": 0.14202778041362762, "global_step": 142761, "epoch": 1604, "lr": 4.60693874294704e-05}
{"train_loss": 0.12519517540931702, "global_step": 142762, "epoch": 1604, "lr": 4.606880948726053e-05}
{"train_loss": 0.1757306307554245, "global_step": 142763, "epoch": 1604, "lr": 4.606823154557914e-05}
{"train_loss": 0.12008290737867355, "global_step": 142764, "epoch": 1604, "lr": 4.606765360442635e-05}
{"train_loss": 0.12148987501859665, "global_step": 142765, "epoch": 1604, "lr": 4.6067075663802194e-05}
{"train_loss": 0.07201751321554184, "global_step": 142766, "epoch": 1604, "lr": 4.6066497723706795e-05}
{"train_loss": 0.14692974090576172, "global_step": 142767, "epoch": 1604, "lr": 4.6065919784140196e-05}
{"train_loss": 0.119960717856884, "global_step": 142768, "epoch": 1604, "lr": 4.606534184510249e-05}
{"train_loss": 0.07932214438915253, "global_step": 142769, "epoch": 1604, "lr": 4.606476390659374e-05}
{"train_loss": 0.11516159772872925, "global_step": 142770, "epoch": 1604, "lr": 4.6064185968614055e-05}
{"train_loss": 0.15982623398303986, "global_step": 142771, "epoch": 1604, "lr": 4.6063608031163466e-05}
{"train_loss": 0.1564931720495224, "global_step": 142772, "epoch": 1604, "lr": 4.6063030094242103e-05}
{"train_loss": 0.14765416085720062, "global_step": 142773, "epoch": 1604, "lr": 4.606245215785e-05}
{"train_loss": 0.09125564992427826, "global_step": 142774, "epoch": 1604, "lr": 4.606187422198727e-05}
{"train_loss": 0.06893796473741531, "global_step": 142775, "epoch": 1604, "lr": 4.606129628665396e-05}
{"train_loss": 0.15141576528549194, "global_step": 142776, "epoch": 1604, "lr": 4.606071835185016e-05}
{"train_loss": 0.07933054864406586, "global_step": 142777, "epoch": 1604, "lr": 4.6060140417575946e-05}
{"train_loss": 0.19750072062015533, "global_step": 142778, "epoch": 1604, "lr": 4.60595624838314e-05}
{"train_loss": 0.07852516323328018, "global_step": 142779, "epoch": 1604, "lr": 4.605898455061661e-05}
{"train_loss": 0.10459744185209274, "global_step": 142780, "epoch": 1604, "lr": 4.605840661793162e-05}
{"train_loss": 0.0893234834074974, "global_step": 142781, "epoch": 1604, "lr": 4.605782868577655e-05}
{"train_loss": 0.11602392047643661, "global_step": 142782, "epoch": 1604, "lr": 4.6057250754151434e-05}
{"train_loss": 0.13102397322654724, "global_step": 142783, "epoch": 1604, "lr": 4.6056672823056405e-05}
{"train_loss": 0.18104961514472961, "global_step": 142784, "epoch": 1604, "lr": 4.6056094892491476e-05}
{"train_loss": 0.1433214694261551, "global_step": 142785, "epoch": 1604, "lr": 4.605551696245678e-05}
{"train_loss": 0.17300018668174744, "global_step": 142786, "epoch": 1604, "lr": 4.605493903295235e-05}
{"train_loss": 0.1148713007569313, "global_step": 142787, "epoch": 1604, "lr": 4.605436110397831e-05}
{"train_loss": 0.09538377076387405, "global_step": 142788, "epoch": 1604, "lr": 4.6053783175534686e-05}
{"train_loss": 0.1284782737493515, "global_step": 142789, "epoch": 1604, "lr": 4.6053205247621614e-05}
{"train_loss": 0.17880462110042572, "global_step": 142790, "epoch": 1604, "lr": 4.605262732023911e-05}
{"train_loss": 0.09522826969623566, "global_step": 142791, "epoch": 1604, "lr": 4.605204939338731e-05}
{"train_loss": 0.07549580931663513, "global_step": 142792, "epoch": 1604, "lr": 4.605147146706624e-05}
{"train_loss": 0.08832888305187225, "global_step": 142793, "epoch": 1604, "lr": 4.605089354127602e-05}
{"train_loss": 0.13631229102611542, "global_step": 142794, "epoch": 1604, "lr": 4.60503156160167e-05}
{"train_loss": 0.055627550929784775, "global_step": 142795, "epoch": 1604, "lr": 4.6049737691288364e-05}
{"train_loss": 0.1452893614768982, "global_step": 142796, "epoch": 1604, "lr": 4.6049159767091096e-05}
{"train_loss": 0.12377244979143143, "global_step": 142797, "epoch": 1604, "lr": 4.604858184342496e-05}
{"train_loss": 0.15347403287887573, "global_step": 142798, "epoch": 1604, "lr": 4.604800392029007e-05}
{"train_loss": 0.11583483964204788, "global_step": 142799, "epoch": 1604, "lr": 4.604742599768645e-05}
{"train_loss": 0.11241354793310165, "global_step": 142800, "epoch": 1604, "lr": 4.6046848075614224e-05}
{"train_loss": 0.13342700898647308, "global_step": 142801, "epoch": 1604, "lr": 4.6046270154073434e-05}
{"train_loss": 0.079500213265419, "global_step": 142802, "epoch": 1604, "lr": 4.6045692233064195e-05}
{"train_loss": 0.13334524631500244, "global_step": 142803, "epoch": 1604, "lr": 4.604511431258655e-05}
{"train_loss": 0.14093875885009766, "global_step": 142804, "epoch": 1604, "lr": 4.60445363926406e-05}
{"train_loss": 0.14090964198112488, "global_step": 142805, "epoch": 1604, "lr": 4.6043958473226405e-05}
{"train_loss": 0.11221451312303543, "global_step": 142806, "epoch": 1604, "lr": 4.6043380554344066e-05}
{"train_loss": 0.1331397145986557, "global_step": 142807, "epoch": 1604, "lr": 4.604280263599363e-05}
{"train_loss": 0.09185539931058884, "global_step": 142808, "epoch": 1604, "lr": 4.6042224718175204e-05}
{"train_loss": 0.06443598866462708, "global_step": 142809, "epoch": 1604, "lr": 4.604164680088884e-05}
{"train_loss": 0.10845957696437836, "global_step": 142810, "epoch": 1604, "lr": 4.6041068884134645e-05}
{"train_loss": 0.08236178755760193, "global_step": 142811, "epoch": 1604, "lr": 4.604049096791266e-05}
{"train_loss": 0.18205608427524567, "global_step": 142812, "epoch": 1604, "lr": 4.6039913052223e-05}
{"train_loss": 0.2199767380952835, "global_step": 142813, "epoch": 1604, "lr": 4.603933513706572e-05}
{"train_loss": 0.06718607246875763, "global_step": 142814, "epoch": 1604, "lr": 4.603875722244089e-05}
{"train_loss": 0.08733046799898148, "global_step": 142815, "epoch": 1604, "lr": 4.603817930834862e-05}
{"train_loss": 0.08608878403902054, "global_step": 142816, "epoch": 1604, "lr": 4.603760139478895e-05}
{"train_loss": 0.11979416012763977, "global_step": 142817, "epoch": 1604, "lr": 4.6037023481762e-05}
{"train_loss": 0.1218840628862381, "global_step": 142818, "epoch": 1604, "lr": 4.603644556926779e-05}
{"train_loss": 0.14201918244361877, "global_step": 142819, "epoch": 1604, "lr": 4.603586765730647e-05}
{"train_loss": 0.1355951577425003, "global_step": 142820, "epoch": 1604, "lr": 4.603528974587805e-05}
{"train_loss": 0.14888665080070496, "global_step": 142821, "epoch": 1604, "lr": 4.6034711834982655e-05}
{"train_loss": 0.07896362990140915, "global_step": 142822, "epoch": 1604, "lr": 4.603413392462033e-05}
{"train_loss": 0.07380740344524384, "global_step": 142823, "epoch": 1604, "lr": 4.603355601479119e-05}
{"train_loss": 0.11243047565221786, "global_step": 142824, "epoch": 1604, "lr": 4.603297810549526e-05}
{"train_loss": 0.13930287957191467, "global_step": 142825, "epoch": 1604, "lr": 4.6032400196732675e-05}
{"train_loss": 0.10560125857591629, "global_step": 142826, "epoch": 1604, "lr": 4.603182228850346e-05}
{"train_loss": 0.1079549565911293, "global_step": 142827, "epoch": 1604, "lr": 4.6031244380807745e-05}
{"train_loss": 0.09515250474214554, "global_step": 142828, "epoch": 1604, "lr": 4.603066647364556e-05}
{"train_loss": 0.09759722650051117, "global_step": 142829, "epoch": 1604, "lr": 4.603008856701702e-05}
{"train_loss": 0.1541450321674347, "global_step": 142830, "epoch": 1604, "lr": 4.602951066092217e-05}
{"train_loss": 0.030814750120043755, "global_step": 142831, "epoch": 1604, "lr": 4.602893275536112e-05}
{"train_loss": 0.11597388237714767, "global_step": 142832, "epoch": 1604, "lr": 4.602835485033392e-05}
{"train_loss": 0.07329985499382019, "global_step": 142833, "epoch": 1604, "lr": 4.602777694584065e-05}
{"train_loss": 0.12485067546367645, "global_step": 142834, "epoch": 1604, "lr": 4.6027199041881425e-05}
{"train_loss": 0.10993649810552597, "global_step": 142835, "epoch": 1604, "lr": 4.6026621138456265e-05}
{"train_loss": 0.12858808040618896, "global_step": 142836, "epoch": 1604, "lr": 4.6026043235565304e-05}
{"train_loss": 0.10957764089107513, "global_step": 142837, "epoch": 1604, "lr": 4.6025465333208564e-05}
{"train_loss": 0.08186667412519455, "global_step": 142838, "epoch": 1604, "lr": 4.602488743138618e-05}
{"train_loss": 0.06343550235033035, "global_step": 142839, "epoch": 1604, "lr": 4.602430953009818e-05}
{"train_loss": 0.08686163276433945, "global_step": 142840, "epoch": 1604, "lr": 4.602373162934468e-05}
{"train_loss": 0.1099267452955246, "global_step": 142841, "epoch": 1604, "lr": 4.6023153729125725e-05}
{"train_loss": 0.10533292591571808, "global_step": 142842, "epoch": 1604, "lr": 4.602257582944143e-05}
{"train_loss": 0.09965449571609497, "global_step": 142843, "epoch": 1604, "lr": 4.6021997930291826e-05}
{"train_loss": 0.11494610703477029, "global_step": 142844, "epoch": 1604, "lr": 4.6021420031677034e-05, "val_loss": 6.0770978927612305}
{"train_loss": 0.057917188853025436, "global_step": 142845, "epoch": 1605, "lr": 4.6020842133597106e-05}
{"train_loss": 0.10845071822404861, "global_step": 142846, "epoch": 1605, "lr": 4.6020264236052137e-05}
{"train_loss": 0.12283521145582199, "global_step": 142847, "epoch": 1605, "lr": 4.6019686339042173e-05}
{"train_loss": 0.0934801697731018, "global_step": 142848, "epoch": 1605, "lr": 4.601910844256734e-05}
{"train_loss": 0.12593001127243042, "global_step": 142849, "epoch": 1605, "lr": 4.6018530546627666e-05}
{"train_loss": 0.08648008108139038, "global_step": 142850, "epoch": 1605, "lr": 4.601795265122327e-05}
{"train_loss": 0.06215747073292732, "global_step": 142851, "epoch": 1605, "lr": 4.60173747563542e-05}
{"train_loss": 0.06341151893138885, "global_step": 142852, "epoch": 1605, "lr": 4.601679686202054e-05}
{"train_loss": 0.12406224012374878, "global_step": 142853, "epoch": 1605, "lr": 4.6016218968222394e-05}
{"train_loss": 0.07110634446144104, "global_step": 142854, "epoch": 1605, "lr": 4.6015641074959806e-05}
{"train_loss": 0.07102419435977936, "global_step": 142855, "epoch": 1605, "lr": 4.6015063182232875e-05}
{"train_loss": 0.05916723608970642, "global_step": 142856, "epoch": 1605, "lr": 4.6014485290041656e-05}
{"train_loss": 0.1132173240184784, "global_step": 142857, "epoch": 1605, "lr": 4.6013907398386255e-05}
{"train_loss": 0.0950106754899025, "global_step": 142858, "epoch": 1605, "lr": 4.6013329507266715e-05}
{"train_loss": 0.1196567714214325, "global_step": 142859, "epoch": 1605, "lr": 4.6012751616683164e-05}
{"train_loss": 0.11667924374341965, "global_step": 142860, "epoch": 1605, "lr": 4.601217372663562e-05}
{"train_loss": 0.14156049489974976, "global_step": 142861, "epoch": 1605, "lr": 4.601159583712422e-05}
{"train_loss": 0.1443316489458084, "global_step": 142862, "epoch": 1605, "lr": 4.601101794814898e-05}
{"train_loss": 0.14211909472942352, "global_step": 142863, "epoch": 1605, "lr": 4.601044005971004e-05}
{"train_loss": 0.09999670833349228, "global_step": 142864, "epoch": 1605, "lr": 4.600986217180743e-05}
{"train_loss": 0.14593617618083954, "global_step": 142865, "epoch": 1605, "lr": 4.6009284284441256e-05}
{"train_loss": 0.05848619341850281, "global_step": 142866, "epoch": 1605, "lr": 4.6008706397611565e-05}
{"train_loss": 0.1480664312839508, "global_step": 142867, "epoch": 1605, "lr": 4.600812851131848e-05}
{"train_loss": 0.13299468159675598, "global_step": 142868, "epoch": 1605, "lr": 4.6007550625562034e-05}
{"train_loss": 0.10392361879348755, "global_step": 142869, "epoch": 1605, "lr": 4.6006972740342336e-05}
{"train_loss": 0.1219903975725174, "global_step": 142870, "epoch": 1605, "lr": 4.600639485565945e-05}
{"train_loss": 0.110806904733181, "global_step": 142871, "epoch": 1605, "lr": 4.600581697151345e-05}
{"train_loss": 0.08402050286531448, "global_step": 142872, "epoch": 1605, "lr": 4.6005239087904433e-05}
{"train_loss": 0.09523282945156097, "global_step": 142873, "epoch": 1605, "lr": 4.6004661204832445e-05}
{"train_loss": 0.1584661900997162, "global_step": 142874, "epoch": 1605, "lr": 4.60040833222976e-05}
{"train_loss": 0.09428255259990692, "global_step": 142875, "epoch": 1605, "lr": 4.600350544029993e-05}
{"train_loss": 0.1397150158882141, "global_step": 142876, "epoch": 1605, "lr": 4.600292755883957e-05}
{"train_loss": 0.09724224358797073, "global_step": 142877, "epoch": 1605, "lr": 4.6002349677916544e-05}
{"train_loss": 0.12453556060791016, "global_step": 142878, "epoch": 1605, "lr": 4.600177179753098e-05}
{"train_loss": 0.16531434655189514, "global_step": 142879, "epoch": 1605, "lr": 4.60011939176829e-05}
{"train_loss": 0.09363257139921188, "global_step": 142880, "epoch": 1605, "lr": 4.6000616038372434e-05}
{"train_loss": 0.08608752489089966, "global_step": 142881, "epoch": 1605, "lr": 4.600003815959962e-05}
{"train_loss": 0.0645233541727066, "global_step": 142882, "epoch": 1605, "lr": 4.599946028136457e-05}
{"train_loss": 0.096043661236763, "global_step": 142883, "epoch": 1605, "lr": 4.5998882403667315e-05}
{"train_loss": 0.11367509514093399, "global_step": 142884, "epoch": 1605, "lr": 4.599830452650799e-05}
{"train_loss": 0.09707050770521164, "global_step": 142885, "epoch": 1605, "lr": 4.5997726649886625e-05}
{"train_loss": 0.16547322273254395, "global_step": 142886, "epoch": 1605, "lr": 4.5997148773803336e-05}
{"train_loss": 0.046068836003541946, "global_step": 142887, "epoch": 1605, "lr": 4.5996570898258156e-05}
{"train_loss": 0.054007258266210556, "global_step": 142888, "epoch": 1605, "lr": 4.59959930232512e-05}
{"train_loss": 0.09278887510299683, "global_step": 142889, "epoch": 1605, "lr": 4.599541514878254e-05}
{"train_loss": 0.130374476313591, "global_step": 142890, "epoch": 1605, "lr": 4.599483727485224e-05}
{"train_loss": 0.131947323679924, "global_step": 142891, "epoch": 1605, "lr": 4.59942594014604e-05}
{"train_loss": 0.14075905084609985, "global_step": 142892, "epoch": 1605, "lr": 4.5993681528607067e-05}
{"train_loss": 0.07868040353059769, "global_step": 142893, "epoch": 1605, "lr": 4.599310365629235e-05}
{"train_loss": 0.08013647049665451, "global_step": 142894, "epoch": 1605, "lr": 4.59925257845163e-05}
{"train_loss": 0.12034585326910019, "global_step": 142895, "epoch": 1605, "lr": 4.599194791327901e-05}
{"train_loss": 0.09119433909654617, "global_step": 142896, "epoch": 1605, "lr": 4.599137004258054e-05}
{"train_loss": 0.1531422734260559, "global_step": 142897, "epoch": 1605, "lr": 4.5990792172421e-05}
{"train_loss": 0.1469583362340927, "global_step": 142898, "epoch": 1605, "lr": 4.599021430280043e-05}
{"train_loss": 0.155803382396698, "global_step": 142899, "epoch": 1605, "lr": 4.5989636433718954e-05}
{"train_loss": 0.07962911576032639, "global_step": 142900, "epoch": 1605, "lr": 4.5989058565176596e-05}
{"train_loss": 0.0932115688920021, "global_step": 142901, "epoch": 1605, "lr": 4.598848069717349e-05}
{"train_loss": 0.1783319115638733, "global_step": 142902, "epoch": 1605, "lr": 4.598790282970965e-05}
{"train_loss": 0.08750366419553757, "global_step": 142903, "epoch": 1605, "lr": 4.598732496278521e-05}
{"train_loss": 0.0914309024810791, "global_step": 142904, "epoch": 1605, "lr": 4.598674709640021e-05}
{"train_loss": 0.08257085829973221, "global_step": 142905, "epoch": 1605, "lr": 4.598616923055475e-05}
{"train_loss": 0.10566139966249466, "global_step": 142906, "epoch": 1605, "lr": 4.5985591365248905e-05}
{"train_loss": 0.0734555572271347, "global_step": 142907, "epoch": 1605, "lr": 4.598501350048274e-05}
{"train_loss": 0.1813160926103592, "global_step": 142908, "epoch": 1605, "lr": 4.598443563625634e-05}
{"train_loss": 0.1396474391222, "global_step": 142909, "epoch": 1605, "lr": 4.5983857772569794e-05}
{"train_loss": 0.07462402433156967, "global_step": 142910, "epoch": 1605, "lr": 4.598327990942318e-05}
{"train_loss": 0.19514183700084686, "global_step": 142911, "epoch": 1605, "lr": 4.598270204681654e-05}
{"train_loss": 0.08623184263706207, "global_step": 142912, "epoch": 1605, "lr": 4.5982124184749995e-05}
{"train_loss": 0.10068630427122116, "global_step": 142913, "epoch": 1605, "lr": 4.5981546323223584e-05}
{"train_loss": 0.12801304459571838, "global_step": 142914, "epoch": 1605, "lr": 4.598096846223743e-05}
{"train_loss": 0.09048210084438324, "global_step": 142915, "epoch": 1605, "lr": 4.598039060179156e-05}
{"train_loss": 0.03306747227907181, "global_step": 142916, "epoch": 1605, "lr": 4.59798127418861e-05}
{"train_loss": 0.12699952721595764, "global_step": 142917, "epoch": 1605, "lr": 4.597923488252109e-05}
{"train_loss": 0.13591566681861877, "global_step": 142918, "epoch": 1605, "lr": 4.597865702369664e-05}
{"train_loss": 0.1367296725511551, "global_step": 142919, "epoch": 1605, "lr": 4.5978079165412784e-05}
{"train_loss": 0.07143552601337433, "global_step": 142920, "epoch": 1605, "lr": 4.5977501307669655e-05}
{"train_loss": 0.15386803448200226, "global_step": 142921, "epoch": 1605, "lr": 4.597692345046728e-05}
{"train_loss": 0.1222289651632309, "global_step": 142922, "epoch": 1605, "lr": 4.597634559380578e-05}
{"train_loss": 0.11253444850444794, "global_step": 142923, "epoch": 1605, "lr": 4.5975767737685186e-05}
{"train_loss": 0.08215998113155365, "global_step": 142924, "epoch": 1605, "lr": 4.597518988210561e-05}
{"train_loss": 0.12130572646856308, "global_step": 142925, "epoch": 1605, "lr": 4.5974612027067126e-05}
{"train_loss": 0.044248759746551514, "global_step": 142926, "epoch": 1605, "lr": 4.59740341725698e-05}
{"train_loss": 0.1117417961359024, "global_step": 142927, "epoch": 1605, "lr": 4.597345631861372e-05}
{"train_loss": 0.1258533000946045, "global_step": 142928, "epoch": 1605, "lr": 4.5972878465198966e-05}
{"train_loss": 0.10391265153884888, "global_step": 142929, "epoch": 1605, "lr": 4.5972300612325613e-05}
{"train_loss": 0.17354492843151093, "global_step": 142930, "epoch": 1605, "lr": 4.597172275999371e-05}
{"train_loss": 0.07912836968898773, "global_step": 142931, "epoch": 1605, "lr": 4.597114490820339e-05}
{"train_loss": 0.07054007798433304, "global_step": 142932, "epoch": 1605, "lr": 4.597056705695467e-05}
{"train_loss": 0.10967204937439286, "global_step": 142933, "epoch": 1605, "lr": 4.596998920624769e-05, "val_loss": 6.032112121582031, "train_action_mse_error": 16.777042388916016}
{"train_loss": 0.07228196412324905, "global_step": 142934, "epoch": 1606, "lr": 4.596941135608247e-05}
{"train_loss": 0.11852133274078369, "global_step": 142935, "epoch": 1606, "lr": 4.596883350645913e-05}
{"train_loss": 0.07833259552717209, "global_step": 142936, "epoch": 1606, "lr": 4.596825565737772e-05}
{"train_loss": 0.2045920193195343, "global_step": 142937, "epoch": 1606, "lr": 4.596767780883834e-05}
{"train_loss": 0.09975387901067734, "global_step": 142938, "epoch": 1606, "lr": 4.596709996084104e-05}
{"train_loss": 0.15103982388973236, "global_step": 142939, "epoch": 1606, "lr": 4.596652211338593e-05}
{"train_loss": 0.06548140197992325, "global_step": 142940, "epoch": 1606, "lr": 4.596594426647306e-05}
{"train_loss": 0.17780616879463196, "global_step": 142941, "epoch": 1606, "lr": 4.596536642010251e-05}
{"train_loss": 0.09972307085990906, "global_step": 142942, "epoch": 1606, "lr": 4.59647885742744e-05}
{"train_loss": 0.1342090517282486, "global_step": 142943, "epoch": 1606, "lr": 4.596421072898875e-05}
{"train_loss": 0.06999853253364563, "global_step": 142944, "epoch": 1606, "lr": 4.5963632884245675e-05}
{"train_loss": 0.08080758154392242, "global_step": 142945, "epoch": 1606, "lr": 4.596305504004524e-05}
{"train_loss": 0.1386852264404297, "global_step": 142946, "epoch": 1606, "lr": 4.596247719638752e-05}
{"train_loss": 0.05806358903646469, "global_step": 142947, "epoch": 1606, "lr": 4.596189935327259e-05}
{"train_loss": 0.057789236307144165, "global_step": 142948, "epoch": 1606, "lr": 4.596132151070055e-05}
{"train_loss": 0.12665174901485443, "global_step": 142949, "epoch": 1606, "lr": 4.5960743668671444e-05}
{"train_loss": 0.11857528984546661, "global_step": 142950, "epoch": 1606, "lr": 4.596016582718538e-05}
{"train_loss": 0.1503923088312149, "global_step": 142951, "epoch": 1606, "lr": 4.59595879862424e-05}
{"train_loss": 0.102809377014637, "global_step": 142952, "epoch": 1606, "lr": 4.595901014584263e-05}
{"train_loss": 0.12518513202667236, "global_step": 142953, "epoch": 1606, "lr": 4.59584323059861e-05}
{"train_loss": 0.1115148663520813, "global_step": 142954, "epoch": 1606, "lr": 4.5957854466672934e-05}
{"train_loss": 0.0904984176158905, "global_step": 142955, "epoch": 1606, "lr": 4.595727662790316e-05}
{"train_loss": 0.07792338728904724, "global_step": 142956, "epoch": 1606, "lr": 4.5956698789676905e-05}
{"train_loss": 0.10495901852846146, "global_step": 142957, "epoch": 1606, "lr": 4.59561209519942e-05}
{"train_loss": 0.09158481657505035, "global_step": 142958, "epoch": 1606, "lr": 4.595554311485516e-05}
{"train_loss": 0.12427450716495514, "global_step": 142959, "epoch": 1606, "lr": 4.595496527825984e-05}
{"train_loss": 0.03864035755395889, "global_step": 142960, "epoch": 1606, "lr": 4.5954387442208315e-05}
{"train_loss": 0.07650038599967957, "global_step": 142961, "epoch": 1606, "lr": 4.5953809606700696e-05}
{"train_loss": 0.1368151307106018, "global_step": 142962, "epoch": 1606, "lr": 4.595323177173703e-05}
{"train_loss": 0.08358636498451233, "global_step": 142963, "epoch": 1606, "lr": 4.5952653937317406e-05}
{"train_loss": 0.13252340257167816, "global_step": 142964, "epoch": 1606, "lr": 4.595207610344189e-05}
{"train_loss": 0.10056057572364807, "global_step": 142965, "epoch": 1606, "lr": 4.595149827011057e-05}
{"train_loss": 0.1531749665737152, "global_step": 142966, "epoch": 1606, "lr": 4.5950920437323526e-05}
{"train_loss": 0.12091255933046341, "global_step": 142967, "epoch": 1606, "lr": 4.595034260508084e-05}
{"train_loss": 0.11591977626085281, "global_step": 142968, "epoch": 1606, "lr": 4.594976477338256e-05}
{"train_loss": 0.12197433412075043, "global_step": 142969, "epoch": 1606, "lr": 4.5949186942228804e-05}
{"train_loss": 0.1035940945148468, "global_step": 142970, "epoch": 1606, "lr": 4.594860911161961e-05}
{"train_loss": 0.10910037904977798, "global_step": 142971, "epoch": 1606, "lr": 4.59480312815551e-05}
{"train_loss": 0.07284997403621674, "global_step": 142972, "epoch": 1606, "lr": 4.5947453452035305e-05}
{"train_loss": 0.09930798411369324, "global_step": 142973, "epoch": 1606, "lr": 4.594687562306034e-05}
{"train_loss": 0.15224526822566986, "global_step": 142974, "epoch": 1606, "lr": 4.5946297794630254e-05}
{"train_loss": 0.146571546792984, "global_step": 142975, "epoch": 1606, "lr": 4.5945719966745166e-05}
{"train_loss": 0.11932075023651123, "global_step": 142976, "epoch": 1606, "lr": 4.594514213940509e-05}
{"train_loss": 0.13324302434921265, "global_step": 142977, "epoch": 1606, "lr": 4.5944564312610176e-05}
{"train_loss": 0.15164008736610413, "global_step": 142978, "epoch": 1606, "lr": 4.594398648636044e-05}
{"train_loss": 0.09129404276609421, "global_step": 142979, "epoch": 1606, "lr": 4.5943408660655985e-05}
{"train_loss": 0.05971846356987953, "global_step": 142980, "epoch": 1606, "lr": 4.5942830835496906e-05}
{"train_loss": 0.06689654290676117, "global_step": 142981, "epoch": 1606, "lr": 4.594225301088325e-05}
{"train_loss": 0.12856970727443695, "global_step": 142982, "epoch": 1606, "lr": 4.594167518681513e-05}
{"train_loss": 0.12826190888881683, "global_step": 142983, "epoch": 1606, "lr": 4.5941097363292585e-05}
{"train_loss": 0.05682956799864769, "global_step": 142984, "epoch": 1606, "lr": 4.5940519540315714e-05}
{"train_loss": 0.16356873512268066, "global_step": 142985, "epoch": 1606, "lr": 4.593994171788459e-05}
{"train_loss": 0.17967376112937927, "global_step": 142986, "epoch": 1606, "lr": 4.5939363895999286e-05}
{"train_loss": 0.18198776245117188, "global_step": 142987, "epoch": 1606, "lr": 4.593878607465989e-05}
{"train_loss": 0.12512770295143127, "global_step": 142988, "epoch": 1606, "lr": 4.593820825386649e-05}
{"train_loss": 0.16960550844669342, "global_step": 142989, "epoch": 1606, "lr": 4.5937630433619125e-05}
{"train_loss": 0.14162376523017883, "global_step": 142990, "epoch": 1606, "lr": 4.593705261391791e-05}
{"train_loss": 0.1716552972793579, "global_step": 142991, "epoch": 1606, "lr": 4.5936474794762895e-05}
{"train_loss": 0.10940144956111908, "global_step": 142992, "epoch": 1606, "lr": 4.593589697615419e-05}
{"train_loss": 0.11383956670761108, "global_step": 142993, "epoch": 1606, "lr": 4.5935319158091834e-05}
{"train_loss": 0.1681777387857437, "global_step": 142994, "epoch": 1606, "lr": 4.593474134057594e-05}
{"train_loss": 0.09098304808139801, "global_step": 142995, "epoch": 1606, "lr": 4.593416352360655e-05}
{"train_loss": 0.13374702632427216, "global_step": 142996, "epoch": 1606, "lr": 4.593358570718377e-05}
{"train_loss": 0.11304552108049393, "global_step": 142997, "epoch": 1606, "lr": 4.5933007891307686e-05}
{"train_loss": 0.11393938958644867, "global_step": 142998, "epoch": 1606, "lr": 4.593243007597834e-05}
{"train_loss": 0.19568587839603424, "global_step": 142999, "epoch": 1606, "lr": 4.593185226119586e-05}
{"train_loss": 0.15857695043087006, "global_step": 143000, "epoch": 1606, "lr": 4.593127444696026e-05}
{"train_loss": 0.15061980485916138, "global_step": 143001, "epoch": 1606, "lr": 4.593069663327167e-05}
{"train_loss": 0.20633821189403534, "global_step": 143002, "epoch": 1606, "lr": 4.593011882013014e-05}
{"train_loss": 0.10609179735183716, "global_step": 143003, "epoch": 1606, "lr": 4.592954100753576e-05}
{"train_loss": 0.11231854557991028, "global_step": 143004, "epoch": 1606, "lr": 4.59289631954886e-05}
{"train_loss": 0.07215499132871628, "global_step": 143005, "epoch": 1606, "lr": 4.5928385383988744e-05}
{"train_loss": 0.120524100959301, "global_step": 143006, "epoch": 1606, "lr": 4.5927807573036264e-05}
{"train_loss": 0.1886976808309555, "global_step": 143007, "epoch": 1606, "lr": 4.592722976263125e-05}
{"train_loss": 0.12954692542552948, "global_step": 143008, "epoch": 1606, "lr": 4.592665195277375e-05}
{"train_loss": 0.1455533355474472, "global_step": 143009, "epoch": 1606, "lr": 4.592607414346389e-05}
{"train_loss": 0.18968521058559418, "global_step": 143010, "epoch": 1606, "lr": 4.592549633470169e-05}
{"train_loss": 0.08595061302185059, "global_step": 143011, "epoch": 1606, "lr": 4.592491852648728e-05}
{"train_loss": 0.07240027189254761, "global_step": 143012, "epoch": 1606, "lr": 4.592434071882069e-05}
{"train_loss": 0.07323459535837173, "global_step": 143013, "epoch": 1606, "lr": 4.592376291170206e-05}
{"train_loss": 0.12387176603078842, "global_step": 143014, "epoch": 1606, "lr": 4.592318510513139e-05}
{"train_loss": 0.08967970311641693, "global_step": 143015, "epoch": 1606, "lr": 4.592260729910881e-05}
{"train_loss": 0.10419398546218872, "global_step": 143016, "epoch": 1606, "lr": 4.592202949363441e-05}
{"train_loss": 0.10023871064186096, "global_step": 143017, "epoch": 1606, "lr": 4.592145168870821e-05}
{"train_loss": 0.10578188300132751, "global_step": 143018, "epoch": 1606, "lr": 4.592087388433035e-05}
{"train_loss": 0.05715662240982056, "global_step": 143019, "epoch": 1606, "lr": 4.5920296080500856e-05}
{"train_loss": 0.12546366453170776, "global_step": 143020, "epoch": 1606, "lr": 4.591971827721985e-05}
{"train_loss": 0.13790559768676758, "global_step": 143021, "epoch": 1606, "lr": 4.5919140474487374e-05}
{"train_loss": 0.11765352653318577, "global_step": 143022, "epoch": 1606, "lr": 4.591856267230353e-05, "val_loss": 5.935215473175049}
{"train_loss": 0.0741531103849411, "global_step": 143023, "epoch": 1607, "lr": 4.591798487066837e-05}
{"train_loss": 0.14294211566448212, "global_step": 143024, "epoch": 1607, "lr": 4.5917407069582e-05}
{"train_loss": 0.2274058759212494, "global_step": 143025, "epoch": 1607, "lr": 4.5916829269044484e-05}
{"train_loss": 0.12414319068193436, "global_step": 143026, "epoch": 1607, "lr": 4.5916251469055905e-05}
{"train_loss": 0.0761563628911972, "global_step": 143027, "epoch": 1607, "lr": 4.591567366961632e-05}
{"train_loss": 0.1067781150341034, "global_step": 143028, "epoch": 1607, "lr": 4.591509587072584e-05}
{"train_loss": 0.08427469432353973, "global_step": 143029, "epoch": 1607, "lr": 4.5914518072384506e-05}
{"train_loss": 0.10539428889751434, "global_step": 143030, "epoch": 1607, "lr": 4.591394027459244e-05}
{"train_loss": 0.07150396704673767, "global_step": 143031, "epoch": 1607, "lr": 4.591336247734966e-05}
{"train_loss": 0.057525865733623505, "global_step": 143032, "epoch": 1607, "lr": 4.59127846806563e-05}
{"train_loss": 0.1476515829563141, "global_step": 143033, "epoch": 1607, "lr": 4.591220688451242e-05}
{"train_loss": 0.15243171155452728, "global_step": 143034, "epoch": 1607, "lr": 4.591162908891808e-05}
{"train_loss": 0.12269877642393112, "global_step": 143035, "epoch": 1607, "lr": 4.591105129387339e-05}
{"train_loss": 0.0793895497918129, "global_step": 143036, "epoch": 1607, "lr": 4.591047349937839e-05}
{"train_loss": 0.14609409868717194, "global_step": 143037, "epoch": 1607, "lr": 4.59098957054332e-05}
{"train_loss": 0.17867721617221832, "global_step": 143038, "epoch": 1607, "lr": 4.590931791203785e-05}
{"train_loss": 0.14364969730377197, "global_step": 143039, "epoch": 1607, "lr": 4.590874011919247e-05}
{"train_loss": 0.07829777896404266, "global_step": 143040, "epoch": 1607, "lr": 4.590816232689709e-05}
{"train_loss": 0.09246690571308136, "global_step": 143041, "epoch": 1607, "lr": 4.5907584535151815e-05}
{"train_loss": 0.10372215509414673, "global_step": 143042, "epoch": 1607, "lr": 4.590700674395671e-05}
{"train_loss": 0.07641218602657318, "global_step": 143043, "epoch": 1607, "lr": 4.590642895331187e-05}
{"train_loss": 0.0746942088007927, "global_step": 143044, "epoch": 1607, "lr": 4.590585116321734e-05}
{"train_loss": 0.12549249827861786, "global_step": 143045, "epoch": 1607, "lr": 4.590527337367325e-05}
{"train_loss": 0.1784161627292633, "global_step": 143046, "epoch": 1607, "lr": 4.590469558467961e-05}
{"train_loss": 0.14630502462387085, "global_step": 143047, "epoch": 1607, "lr": 4.5904117796236564e-05}
{"train_loss": 0.07803790271282196, "global_step": 143048, "epoch": 1607, "lr": 4.590354000834413e-05}
{"train_loss": 0.1038016676902771, "global_step": 143049, "epoch": 1607, "lr": 4.590296222100244e-05}
{"train_loss": 0.11311476677656174, "global_step": 143050, "epoch": 1607, "lr": 4.5902384434211524e-05}
{"train_loss": 0.12991110980510712, "global_step": 143051, "epoch": 1607, "lr": 4.590180664797149e-05}
{"train_loss": 0.11491227895021439, "global_step": 143052, "epoch": 1607, "lr": 4.5901228862282424e-05}
{"train_loss": 0.12026949971914291, "global_step": 143053, "epoch": 1607, "lr": 4.5900651077144364e-05}
{"train_loss": 0.10385759174823761, "global_step": 143054, "epoch": 1607, "lr": 4.590007329255744e-05}
{"train_loss": 0.052218034863471985, "global_step": 143055, "epoch": 1607, "lr": 4.5899495508521674e-05}
{"train_loss": 0.06584914028644562, "global_step": 143056, "epoch": 1607, "lr": 4.589891772503719e-05}
{"train_loss": 0.120318703353405, "global_step": 143057, "epoch": 1607, "lr": 4.589833994210403e-05}
{"train_loss": 0.05723176524043083, "global_step": 143058, "epoch": 1607, "lr": 4.5897762159722315e-05}
{"train_loss": 0.06571943312883377, "global_step": 143059, "epoch": 1607, "lr": 4.5897184377892075e-05}
{"train_loss": 0.11529689282178879, "global_step": 143060, "epoch": 1607, "lr": 4.589660659661341e-05}
{"train_loss": 0.08768422901630402, "global_step": 143061, "epoch": 1607, "lr": 4.58960288158864e-05}
{"train_loss": 0.13739870488643646, "global_step": 143062, "epoch": 1607, "lr": 4.589545103571112e-05}
{"train_loss": 0.11405717581510544, "global_step": 143063, "epoch": 1607, "lr": 4.589487325608764e-05}
{"train_loss": 0.11942345649003983, "global_step": 143064, "epoch": 1607, "lr": 4.5894295477016055e-05}
{"train_loss": 0.07246334850788116, "global_step": 143065, "epoch": 1607, "lr": 4.589371769849642e-05}
{"train_loss": 0.17825043201446533, "global_step": 143066, "epoch": 1607, "lr": 4.589313992052884e-05}
{"train_loss": 0.1334480196237564, "global_step": 143067, "epoch": 1607, "lr": 4.5892562143113356e-05}
{"train_loss": 0.12831489741802216, "global_step": 143068, "epoch": 1607, "lr": 4.589198436625007e-05}
{"train_loss": 0.1354481279850006, "global_step": 143069, "epoch": 1607, "lr": 4.589140658993908e-05}
{"train_loss": 0.0795462355017662, "global_step": 143070, "epoch": 1607, "lr": 4.589082881418041e-05}
{"train_loss": 0.0986250713467598, "global_step": 143071, "epoch": 1607, "lr": 4.589025103897419e-05}
{"train_loss": 0.12018194049596786, "global_step": 143072, "epoch": 1607, "lr": 4.588967326432046e-05}
{"train_loss": 0.137160524725914, "global_step": 143073, "epoch": 1607, "lr": 4.588909549021933e-05}
{"train_loss": 0.08745518326759338, "global_step": 143074, "epoch": 1607, "lr": 4.588851771667084e-05}
{"train_loss": 0.0735427513718605, "global_step": 143075, "epoch": 1607, "lr": 4.5887939943675115e-05}
{"train_loss": 0.15687240660190582, "global_step": 143076, "epoch": 1607, "lr": 4.588736217123218e-05}
{"train_loss": 0.12857134640216827, "global_step": 143077, "epoch": 1607, "lr": 4.5886784399342166e-05}
{"train_loss": 0.10616720467805862, "global_step": 143078, "epoch": 1607, "lr": 4.5886206628005105e-05}
{"train_loss": 0.10958599299192429, "global_step": 143079, "epoch": 1607, "lr": 4.5885628857221103e-05}
{"train_loss": 0.1245778426527977, "global_step": 143080, "epoch": 1607, "lr": 4.588505108699022e-05}
{"train_loss": 0.10522118210792542, "global_step": 143081, "epoch": 1607, "lr": 4.5884473317312546e-05}
{"train_loss": 0.07479958981275558, "global_step": 143082, "epoch": 1607, "lr": 4.588389554818815e-05}
{"train_loss": 0.07821725308895111, "global_step": 143083, "epoch": 1607, "lr": 4.588331777961712e-05}
{"train_loss": 0.1613171547651291, "global_step": 143084, "epoch": 1607, "lr": 4.588274001159953e-05}
{"train_loss": 0.06309453397989273, "global_step": 143085, "epoch": 1607, "lr": 4.5882162244135457e-05}
{"train_loss": 0.06705769896507263, "global_step": 143086, "epoch": 1607, "lr": 4.588158447722496e-05}
{"train_loss": 0.07722780853509903, "global_step": 143087, "epoch": 1607, "lr": 4.588100671086814e-05}
{"train_loss": 0.12291814386844635, "global_step": 143088, "epoch": 1607, "lr": 4.5880428945065076e-05}
{"train_loss": 0.11390820145606995, "global_step": 143089, "epoch": 1607, "lr": 4.5879851179815825e-05}
{"train_loss": 0.09660318493843079, "global_step": 143090, "epoch": 1607, "lr": 4.58792734151205e-05}
{"train_loss": 0.09448452293872833, "global_step": 143091, "epoch": 1607, "lr": 4.587869565097913e-05}
{"train_loss": 0.11949941515922546, "global_step": 143092, "epoch": 1607, "lr": 4.587811788739185e-05}
{"train_loss": 0.12447793781757355, "global_step": 143093, "epoch": 1607, "lr": 4.587754012435868e-05}
{"train_loss": 0.07722398638725281, "global_step": 143094, "epoch": 1607, "lr": 4.587696236187975e-05}
{"train_loss": 0.12062130868434906, "global_step": 143095, "epoch": 1607, "lr": 4.587638459995508e-05}
{"train_loss": 0.07474645972251892, "global_step": 143096, "epoch": 1607, "lr": 4.5875806838584805e-05}
{"train_loss": 0.12798276543617249, "global_step": 143097, "epoch": 1607, "lr": 4.587522907776896e-05}
{"train_loss": 0.13608577847480774, "global_step": 143098, "epoch": 1607, "lr": 4.587465131750766e-05}
{"train_loss": 0.10849720239639282, "global_step": 143099, "epoch": 1607, "lr": 4.587407355780095e-05}
{"train_loss": 0.1637120246887207, "global_step": 143100, "epoch": 1607, "lr": 4.587349579864892e-05}
{"train_loss": 0.08279488235712051, "global_step": 143101, "epoch": 1607, "lr": 4.5872918040051657e-05}
{"train_loss": 0.10574692487716675, "global_step": 143102, "epoch": 1607, "lr": 4.587234028200922e-05}
{"train_loss": 0.09597224742174149, "global_step": 143103, "epoch": 1607, "lr": 4.58717625245217e-05}
{"train_loss": 0.04611789807677269, "global_step": 143104, "epoch": 1607, "lr": 4.5871184767589183e-05}
{"train_loss": 0.1253863275051117, "global_step": 143105, "epoch": 1607, "lr": 4.5870607011211714e-05}
{"train_loss": 0.0765315592288971, "global_step": 143106, "epoch": 1607, "lr": 4.587002925538939e-05}
{"train_loss": 0.15814511477947235, "global_step": 143107, "epoch": 1607, "lr": 4.5869451500122314e-05}
{"train_loss": 0.07178668677806854, "global_step": 143108, "epoch": 1607, "lr": 4.586887374541051e-05}
{"train_loss": 0.12115661054849625, "global_step": 143109, "epoch": 1607, "lr": 4.586829599125412e-05}
{"train_loss": 0.07146463543176651, "global_step": 143110, "epoch": 1607, "lr": 4.5867718237653156e-05}
{"train_loss": 0.10888104442130314, "global_step": 143111, "epoch": 1607, "lr": 4.586714048460775e-05, "val_loss": 5.998705863952637}
{"train_loss": 0.15381135046482086, "global_step": 143112, "epoch": 1608, "lr": 4.586656273211794e-05}
{"train_loss": 0.10620401054620743, "global_step": 143113, "epoch": 1608, "lr": 4.586598498018384e-05}
{"train_loss": 0.08931110054254532, "global_step": 143114, "epoch": 1608, "lr": 4.586540722880548e-05}
{"train_loss": 0.1088646799325943, "global_step": 143115, "epoch": 1608, "lr": 4.5864829477983e-05}
{"train_loss": 0.10439959168434143, "global_step": 143116, "epoch": 1608, "lr": 4.58642517277164e-05}
{"train_loss": 0.10974077880382538, "global_step": 143117, "epoch": 1608, "lr": 4.586367397800584e-05}
{"train_loss": 0.08689771592617035, "global_step": 143118, "epoch": 1608, "lr": 4.586309622885135e-05}
{"train_loss": 0.06013188138604164, "global_step": 143119, "epoch": 1608, "lr": 4.586251848025301e-05}
{"train_loss": 0.08273304253816605, "global_step": 143120, "epoch": 1608, "lr": 4.58619407322109e-05}
{"train_loss": 0.05686189606785774, "global_step": 143121, "epoch": 1608, "lr": 4.586136298472511e-05}
{"train_loss": 0.10286029428243637, "global_step": 143122, "epoch": 1608, "lr": 4.58607852377957e-05}
{"train_loss": 0.10268278419971466, "global_step": 143123, "epoch": 1608, "lr": 4.586020749142275e-05}
{"train_loss": 0.09180095791816711, "global_step": 143124, "epoch": 1608, "lr": 4.585962974560637e-05}
{"train_loss": 0.13559561967849731, "global_step": 143125, "epoch": 1608, "lr": 4.585905200034659e-05}
{"train_loss": 0.11729878932237625, "global_step": 143126, "epoch": 1608, "lr": 4.585847425564353e-05}
{"train_loss": 0.11485152691602707, "global_step": 143127, "epoch": 1608, "lr": 4.5857896511497226e-05}
{"train_loss": 0.1325889378786087, "global_step": 143128, "epoch": 1608, "lr": 4.58573187679078e-05}
{"train_loss": 0.09478458017110825, "global_step": 143129, "epoch": 1608, "lr": 4.585674102487528e-05}
{"train_loss": 0.08327994495630264, "global_step": 143130, "epoch": 1608, "lr": 4.58561632823998e-05}
{"train_loss": 0.08606665581464767, "global_step": 143131, "epoch": 1608, "lr": 4.5855585540481384e-05}
{"train_loss": 0.11493944376707077, "global_step": 143132, "epoch": 1608, "lr": 4.5855007799120156e-05}
{"train_loss": 0.09728304296731949, "global_step": 143133, "epoch": 1608, "lr": 4.5854430058316146e-05}
{"train_loss": 0.08933752030134201, "global_step": 143134, "epoch": 1608, "lr": 4.5853852318069485e-05}
{"train_loss": 0.1552809327840805, "global_step": 143135, "epoch": 1608, "lr": 4.58532745783802e-05}
{"train_loss": 0.11587083339691162, "global_step": 143136, "epoch": 1608, "lr": 4.585269683924841e-05}
{"train_loss": 0.09933224320411682, "global_step": 143137, "epoch": 1608, "lr": 4.585211910067416e-05}
{"train_loss": 0.10991086810827255, "global_step": 143138, "epoch": 1608, "lr": 4.5851541362657555e-05}
{"train_loss": 0.13123025000095367, "global_step": 143139, "epoch": 1608, "lr": 4.585096362519865e-05}
{"train_loss": 0.0710287094116211, "global_step": 143140, "epoch": 1608, "lr": 4.585038588829753e-05}
{"train_loss": 0.16012787818908691, "global_step": 143141, "epoch": 1608, "lr": 4.584980815195428e-05}
{"train_loss": 0.12520574033260345, "global_step": 143142, "epoch": 1608, "lr": 4.584923041616897e-05}
{"train_loss": 0.13412176072597504, "global_step": 143143, "epoch": 1608, "lr": 4.584865268094168e-05}
{"train_loss": 0.057381924241781235, "global_step": 143144, "epoch": 1608, "lr": 4.584807494627249e-05}
{"train_loss": 0.10362498462200165, "global_step": 143145, "epoch": 1608, "lr": 4.584749721216148e-05}
{"train_loss": 0.06688749045133591, "global_step": 143146, "epoch": 1608, "lr": 4.584691947860871e-05}
{"train_loss": 0.10066086053848267, "global_step": 143147, "epoch": 1608, "lr": 4.58463417456143e-05}
{"train_loss": 0.12025802582502365, "global_step": 143148, "epoch": 1608, "lr": 4.584576401317826e-05}
{"train_loss": 0.15571044385433197, "global_step": 143149, "epoch": 1608, "lr": 4.584518628130074e-05}
{"train_loss": 0.11215411871671677, "global_step": 143150, "epoch": 1608, "lr": 4.584460854998176e-05}
{"train_loss": 0.1486809104681015, "global_step": 143151, "epoch": 1608, "lr": 4.584403081922144e-05}
{"train_loss": 0.05390438437461853, "global_step": 143152, "epoch": 1608, "lr": 4.584345308901982e-05}
{"train_loss": 0.07879315316677094, "global_step": 143153, "epoch": 1608, "lr": 4.584287535937702e-05}
{"train_loss": 0.10719218105077744, "global_step": 143154, "epoch": 1608, "lr": 4.584229763029307e-05}
{"train_loss": 0.08254282921552658, "global_step": 143155, "epoch": 1608, "lr": 4.58417199017681e-05}
{"train_loss": 0.1892361044883728, "global_step": 143156, "epoch": 1608, "lr": 4.584114217380214e-05}
{"train_loss": 0.09006039053201675, "global_step": 143157, "epoch": 1608, "lr": 4.58405644463953e-05}
{"train_loss": 0.08531826734542847, "global_step": 143158, "epoch": 1608, "lr": 4.5839986719547634e-05}
{"train_loss": 0.08350595831871033, "global_step": 143159, "epoch": 1608, "lr": 4.583940899325922e-05}
{"train_loss": 0.06540428102016449, "global_step": 143160, "epoch": 1608, "lr": 4.583883126753018e-05}
{"train_loss": 0.09425088763237, "global_step": 143161, "epoch": 1608, "lr": 4.5838253542360525e-05}
{"train_loss": 0.1745484471321106, "global_step": 143162, "epoch": 1608, "lr": 4.5837675817750396e-05}
{"train_loss": 0.16004891693592072, "global_step": 143163, "epoch": 1608, "lr": 4.583709809369982e-05}
{"train_loss": 0.14964622259140015, "global_step": 143164, "epoch": 1608, "lr": 4.5836520370208904e-05}
{"train_loss": 0.10364487022161484, "global_step": 143165, "epoch": 1608, "lr": 4.583594264727771e-05}
{"train_loss": 0.03821929544210434, "global_step": 143166, "epoch": 1608, "lr": 4.583536492490634e-05}
{"train_loss": 0.12593607604503632, "global_step": 143167, "epoch": 1608, "lr": 4.5834787203094826e-05}
{"train_loss": 0.1048608049750328, "global_step": 143168, "epoch": 1608, "lr": 4.5834209481843305e-05}
{"train_loss": 0.12364703416824341, "global_step": 143169, "epoch": 1608, "lr": 4.5833631761151795e-05}
{"train_loss": 0.13491636514663696, "global_step": 143170, "epoch": 1608, "lr": 4.583305404102043e-05}
{"train_loss": 0.09836117178201675, "global_step": 143171, "epoch": 1608, "lr": 4.5832476321449234e-05}
{"train_loss": 0.10078535974025726, "global_step": 143172, "epoch": 1608, "lr": 4.583189860243833e-05}
{"train_loss": 0.06491898000240326, "global_step": 143173, "epoch": 1608, "lr": 4.583132088398776e-05}
{"train_loss": 0.147028386592865, "global_step": 143174, "epoch": 1608, "lr": 4.5830743166097644e-05}
{"train_loss": 0.12729105353355408, "global_step": 143175, "epoch": 1608, "lr": 4.5830165448768015e-05}
{"train_loss": 0.09714137017726898, "global_step": 143176, "epoch": 1608, "lr": 4.582958773199897e-05}
{"train_loss": 0.11283335089683533, "global_step": 143177, "epoch": 1608, "lr": 4.582901001579059e-05}
{"train_loss": 0.17225830256938934, "global_step": 143178, "epoch": 1608, "lr": 4.582843230014294e-05}
{"train_loss": 0.10019152611494064, "global_step": 143179, "epoch": 1608, "lr": 4.582785458505612e-05}
{"train_loss": 0.06535680592060089, "global_step": 143180, "epoch": 1608, "lr": 4.582727687053018e-05}
{"train_loss": 0.10164476186037064, "global_step": 143181, "epoch": 1608, "lr": 4.582669915656523e-05}
{"train_loss": 0.1494145542383194, "global_step": 143182, "epoch": 1608, "lr": 4.5826121443161305e-05}
{"train_loss": 0.16131503880023956, "global_step": 143183, "epoch": 1608, "lr": 4.582554373031853e-05}
{"train_loss": 0.09739253669977188, "global_step": 143184, "epoch": 1608, "lr": 4.582496601803694e-05}
{"train_loss": 0.05094751715660095, "global_step": 143185, "epoch": 1608, "lr": 4.582438830631666e-05}
{"train_loss": 0.07230456918478012, "global_step": 143186, "epoch": 1608, "lr": 4.5823810595157706e-05}
{"train_loss": 0.0865839347243309, "global_step": 143187, "epoch": 1608, "lr": 4.582323288456022e-05}
{"train_loss": 0.07945196330547333, "global_step": 143188, "epoch": 1608, "lr": 4.582265517452423e-05}
{"train_loss": 0.08683133125305176, "global_step": 143189, "epoch": 1608, "lr": 4.5822077465049847e-05}
{"train_loss": 0.11276011168956757, "global_step": 143190, "epoch": 1608, "lr": 4.5821499756137116e-05}
{"train_loss": 0.04994075372815132, "global_step": 143191, "epoch": 1608, "lr": 4.582092204778616e-05}
{"train_loss": 0.07081988453865051, "global_step": 143192, "epoch": 1608, "lr": 4.5820344339997e-05}
{"train_loss": 0.13477103412151337, "global_step": 143193, "epoch": 1608, "lr": 4.581976663276977e-05}
{"train_loss": 0.12070588767528534, "global_step": 143194, "epoch": 1608, "lr": 4.581918892610451e-05}
{"train_loss": 0.09849472343921661, "global_step": 143195, "epoch": 1608, "lr": 4.5818611220001296e-05}
{"train_loss": 0.14024975895881653, "global_step": 143196, "epoch": 1608, "lr": 4.5818033514460244e-05}
{"train_loss": 0.10726571828126907, "global_step": 143197, "epoch": 1608, "lr": 4.581745580948138e-05}
{"train_loss": 0.08034924417734146, "global_step": 143198, "epoch": 1608, "lr": 4.581687810506483e-05}
{"train_loss": 0.10702770203351974, "global_step": 143199, "epoch": 1608, "lr": 4.581630040121064e-05}
{"train_loss": 0.1062135649531075, "global_step": 143200, "epoch": 1608, "lr": 4.5815722697918905e-05, "val_loss": 6.039279460906982}
{"train_loss": 0.08308259397745132, "global_step": 143201, "epoch": 1609, "lr": 4.581514499518968e-05}
{"train_loss": 0.1783539354801178, "global_step": 143202, "epoch": 1609, "lr": 4.581456729302308e-05}
{"train_loss": 0.06523812562227249, "global_step": 143203, "epoch": 1609, "lr": 4.5813989591419134e-05}
{"train_loss": 0.16067840158939362, "global_step": 143204, "epoch": 1609, "lr": 4.581341189037797e-05}
{"train_loss": 0.10565493255853653, "global_step": 143205, "epoch": 1609, "lr": 4.5812834189899626e-05}
{"train_loss": 0.17153112590312958, "global_step": 143206, "epoch": 1609, "lr": 4.581225648998421e-05}
{"train_loss": 0.1442277878522873, "global_step": 143207, "epoch": 1609, "lr": 4.581167879063176e-05}
{"train_loss": 0.08891613781452179, "global_step": 143208, "epoch": 1609, "lr": 4.581110109184241e-05}
{"train_loss": 0.14960332214832306, "global_step": 143209, "epoch": 1609, "lr": 4.581052339361618e-05}
{"train_loss": 0.09598203003406525, "global_step": 143210, "epoch": 1609, "lr": 4.5809945695953196e-05}
{"train_loss": 0.07163146883249283, "global_step": 143211, "epoch": 1609, "lr": 4.580936799885349e-05}
{"train_loss": 0.0742102563381195, "global_step": 143212, "epoch": 1609, "lr": 4.580879030231719e-05}
{"train_loss": 0.14955319464206696, "global_step": 143213, "epoch": 1609, "lr": 4.580821260634432e-05}
{"train_loss": 0.15633054077625275, "global_step": 143214, "epoch": 1609, "lr": 4.5807634910935e-05}
{"train_loss": 0.11154843121767044, "global_step": 143215, "epoch": 1609, "lr": 4.58070572160893e-05}
{"train_loss": 0.11906065046787262, "global_step": 143216, "epoch": 1609, "lr": 4.580647952180727e-05}
{"train_loss": 0.13892094790935516, "global_step": 143217, "epoch": 1609, "lr": 4.5805901828089026e-05}
{"train_loss": 0.11165069788694382, "global_step": 143218, "epoch": 1609, "lr": 4.580532413493461e-05}
{"train_loss": 0.07321082800626755, "global_step": 143219, "epoch": 1609, "lr": 4.580474644234414e-05}
{"train_loss": 0.15339763462543488, "global_step": 143220, "epoch": 1609, "lr": 4.5804168750317645e-05}
{"train_loss": 0.1680261492729187, "global_step": 143221, "epoch": 1609, "lr": 4.5803591058855256e-05}
{"train_loss": 0.09833960980176926, "global_step": 143222, "epoch": 1609, "lr": 4.580301336795699e-05}
{"train_loss": 0.1059039831161499, "global_step": 143223, "epoch": 1609, "lr": 4.580243567762299e-05}
{"train_loss": 0.1658479869365692, "global_step": 143224, "epoch": 1609, "lr": 4.5801857987853274e-05}
{"train_loss": 0.22545868158340454, "global_step": 143225, "epoch": 1609, "lr": 4.580128029864798e-05}
{"train_loss": 0.11430497467517853, "global_step": 143226, "epoch": 1609, "lr": 4.580070261000711e-05}
{"train_loss": 0.10959772765636444, "global_step": 143227, "epoch": 1609, "lr": 4.580012492193082e-05}
{"train_loss": 0.07165572047233582, "global_step": 143228, "epoch": 1609, "lr": 4.5799547234419136e-05}
{"train_loss": 0.145933136343956, "global_step": 143229, "epoch": 1609, "lr": 4.579896954747216e-05}
{"train_loss": 0.09161917120218277, "global_step": 143230, "epoch": 1609, "lr": 4.579839186108995e-05}
{"train_loss": 0.1560906022787094, "global_step": 143231, "epoch": 1609, "lr": 4.5797814175272606e-05}
{"train_loss": 0.11297594010829926, "global_step": 143232, "epoch": 1609, "lr": 4.5797236490020176e-05}
{"train_loss": 0.0826890841126442, "global_step": 143233, "epoch": 1609, "lr": 4.579665880533277e-05}
{"train_loss": 0.06889377534389496, "global_step": 143234, "epoch": 1609, "lr": 4.5796081121210464e-05}
{"train_loss": 0.08510863035917282, "global_step": 143235, "epoch": 1609, "lr": 4.5795503437653305e-05}
{"train_loss": 0.12248130142688751, "global_step": 143236, "epoch": 1609, "lr": 4.579492575466141e-05}
{"train_loss": 0.08472316712141037, "global_step": 143237, "epoch": 1609, "lr": 4.579434807223481e-05}
{"train_loss": 0.09781454503536224, "global_step": 143238, "epoch": 1609, "lr": 4.5793770390373626e-05}
{"train_loss": 0.11155477911233902, "global_step": 143239, "epoch": 1609, "lr": 4.57931927090779e-05}
{"train_loss": 0.09455141425132751, "global_step": 143240, "epoch": 1609, "lr": 4.5792615028347754e-05}
{"train_loss": 0.13773943483829498, "global_step": 143241, "epoch": 1609, "lr": 4.579203734818321e-05}
{"train_loss": 0.10907148569822311, "global_step": 143242, "epoch": 1609, "lr": 4.579145966858441e-05}
{"train_loss": 0.11238645017147064, "global_step": 143243, "epoch": 1609, "lr": 4.579088198955137e-05}
{"train_loss": 0.06338603794574738, "global_step": 143244, "epoch": 1609, "lr": 4.579030431108421e-05}
{"train_loss": 0.10682876408100128, "global_step": 143245, "epoch": 1609, "lr": 4.578972663318297e-05}
{"train_loss": 0.09534796327352524, "global_step": 143246, "epoch": 1609, "lr": 4.578914895584778e-05}
{"train_loss": 0.08601564913988113, "global_step": 143247, "epoch": 1609, "lr": 4.5788571279078665e-05}
{"train_loss": 0.1297638863325119, "global_step": 143248, "epoch": 1609, "lr": 4.5787993602875744e-05}
{"train_loss": 0.06666155159473419, "global_step": 143249, "epoch": 1609, "lr": 4.5787415927239055e-05}
{"train_loss": 0.19750259816646576, "global_step": 143250, "epoch": 1609, "lr": 4.5786838252168706e-05}
{"train_loss": 0.1468796581029892, "global_step": 143251, "epoch": 1609, "lr": 4.5786260577664764e-05}
{"train_loss": 0.08482453972101212, "global_step": 143252, "epoch": 1609, "lr": 4.578568290372731e-05}
{"train_loss": 0.10151468962430954, "global_step": 143253, "epoch": 1609, "lr": 4.5785105230356434e-05}
{"train_loss": 0.11908140033483505, "global_step": 143254, "epoch": 1609, "lr": 4.578452755755217e-05}
{"train_loss": 0.13927966356277466, "global_step": 143255, "epoch": 1609, "lr": 4.578394988531465e-05}
{"train_loss": 0.09332186728715897, "global_step": 143256, "epoch": 1609, "lr": 4.5783372213643904e-05}
{"train_loss": 0.11228015273809433, "global_step": 143257, "epoch": 1609, "lr": 4.578279454254006e-05}
{"train_loss": 0.1015428677201271, "global_step": 143258, "epoch": 1609, "lr": 4.578221687200314e-05}
{"train_loss": 0.11140763014554977, "global_step": 143259, "epoch": 1609, "lr": 4.578163920203327e-05}
{"train_loss": 0.16532479226589203, "global_step": 143260, "epoch": 1609, "lr": 4.578106153263049e-05}
{"train_loss": 0.07523419708013535, "global_step": 143261, "epoch": 1609, "lr": 4.5780483863794916e-05}
{"train_loss": 0.10500804334878922, "global_step": 143262, "epoch": 1609, "lr": 4.577990619552659e-05}
{"train_loss": 0.14245425164699554, "global_step": 143263, "epoch": 1609, "lr": 4.577932852782561e-05}
{"train_loss": 0.11170899122953415, "global_step": 143264, "epoch": 1609, "lr": 4.577875086069203e-05}
{"train_loss": 0.13359998166561127, "global_step": 143265, "epoch": 1609, "lr": 4.577817319412597e-05}
{"train_loss": 0.061301860958337784, "global_step": 143266, "epoch": 1609, "lr": 4.577759552812746e-05}
{"train_loss": 0.12517762184143066, "global_step": 143267, "epoch": 1609, "lr": 4.577701786269663e-05}
{"train_loss": 0.11195981502532959, "global_step": 143268, "epoch": 1609, "lr": 4.5776440197833503e-05}
{"train_loss": 0.06091950833797455, "global_step": 143269, "epoch": 1609, "lr": 4.5775862533538186e-05}
{"train_loss": 0.11921148747205734, "global_step": 143270, "epoch": 1609, "lr": 4.5775284869810766e-05}
{"train_loss": 0.09255538880825043, "global_step": 143271, "epoch": 1609, "lr": 4.57747072066513e-05}
{"train_loss": 0.11378349363803864, "global_step": 143272, "epoch": 1609, "lr": 4.577412954405988e-05}
{"train_loss": 0.1348620504140854, "global_step": 143273, "epoch": 1609, "lr": 4.577355188203656e-05}
{"train_loss": 0.15905578434467316, "global_step": 143274, "epoch": 1609, "lr": 4.577297422058145e-05}
{"train_loss": 0.06219840049743652, "global_step": 143275, "epoch": 1609, "lr": 4.5772396559694596e-05}
{"train_loss": 0.11346612870693207, "global_step": 143276, "epoch": 1609, "lr": 4.577181889937612e-05}
{"train_loss": 0.05105511099100113, "global_step": 143277, "epoch": 1609, "lr": 4.577124123962604e-05}
{"train_loss": 0.10543902218341827, "global_step": 143278, "epoch": 1609, "lr": 4.577066358044449e-05}
{"train_loss": 0.09262190759181976, "global_step": 143279, "epoch": 1609, "lr": 4.57700859218315e-05}
{"train_loss": 0.1055419072508812, "global_step": 143280, "epoch": 1609, "lr": 4.576950826378719e-05}
{"train_loss": 0.08318348973989487, "global_step": 143281, "epoch": 1609, "lr": 4.57689306063116e-05}
{"train_loss": 0.14703580737113953, "global_step": 143282, "epoch": 1609, "lr": 4.5768352949404844e-05}
{"train_loss": 0.07366741448640823, "global_step": 143283, "epoch": 1609, "lr": 4.5767775293066966e-05}
{"train_loss": 0.07679037749767303, "global_step": 143284, "epoch": 1609, "lr": 4.576719763729808e-05}
{"train_loss": 0.23311541974544525, "global_step": 143285, "epoch": 1609, "lr": 4.5766619982098214e-05}
{"train_loss": 0.059147272258996964, "global_step": 143286, "epoch": 1609, "lr": 4.576604232746748e-05}
{"train_loss": 0.1616324484348297, "global_step": 143287, "epoch": 1609, "lr": 4.576546467340597e-05}
{"train_loss": 0.1204320415854454, "global_step": 143288, "epoch": 1609, "lr": 4.576488701991373e-05}
{"train_loss": 0.1137157663535536, "global_step": 143289, "epoch": 1609, "lr": 4.576430936699085e-05, "val_loss": 6.1604533195495605}
{"train_loss": 0.1314990073442459, "global_step": 143290, "epoch": 1610, "lr": 4.57637317146374e-05}
{"train_loss": 0.12896931171417236, "global_step": 143291, "epoch": 1610, "lr": 4.576315406285349e-05}
{"train_loss": 0.08248873054981232, "global_step": 143292, "epoch": 1610, "lr": 4.5762576411639136e-05}
{"train_loss": 0.1568785309791565, "global_step": 143293, "epoch": 1610, "lr": 4.5761998760994486e-05}
{"train_loss": 0.10360950976610184, "global_step": 143294, "epoch": 1610, "lr": 4.576142111091956e-05}
{"train_loss": 0.0555395781993866, "global_step": 143295, "epoch": 1610, "lr": 4.576084346141447e-05}
{"train_loss": 0.12601591646671295, "global_step": 143296, "epoch": 1610, "lr": 4.576026581247927e-05}
{"train_loss": 0.14932438731193542, "global_step": 143297, "epoch": 1610, "lr": 4.5759688164114066e-05}
{"train_loss": 0.13387006521224976, "global_step": 143298, "epoch": 1610, "lr": 4.575911051631891e-05}
{"train_loss": 0.15541604161262512, "global_step": 143299, "epoch": 1610, "lr": 4.57585328690939e-05}
{"train_loss": 0.12290909141302109, "global_step": 143300, "epoch": 1610, "lr": 4.5757955222439093e-05}
{"train_loss": 0.10328996181488037, "global_step": 143301, "epoch": 1610, "lr": 4.575737757635459e-05}
{"train_loss": 0.10100610554218292, "global_step": 143302, "epoch": 1610, "lr": 4.575679993084044e-05}
{"train_loss": 0.12173771858215332, "global_step": 143303, "epoch": 1610, "lr": 4.575622228589675e-05}
{"train_loss": 0.13018646836280823, "global_step": 143304, "epoch": 1610, "lr": 4.575564464152357e-05}
{"train_loss": 0.059375837445259094, "global_step": 143305, "epoch": 1610, "lr": 4.5755066997720995e-05}
{"train_loss": 0.06696075946092606, "global_step": 143306, "epoch": 1610, "lr": 4.575448935448911e-05}
{"train_loss": 0.11554175615310669, "global_step": 143307, "epoch": 1610, "lr": 4.575391171182798e-05}
{"train_loss": 0.08642850071191788, "global_step": 143308, "epoch": 1610, "lr": 4.575333406973768e-05}
{"train_loss": 0.11700954288244247, "global_step": 143309, "epoch": 1610, "lr": 4.575275642821829e-05}
{"train_loss": 0.12486781179904938, "global_step": 143310, "epoch": 1610, "lr": 4.5752178787269906e-05}
{"train_loss": 0.16162805259227753, "global_step": 143311, "epoch": 1610, "lr": 4.575160114689257e-05}
{"train_loss": 0.18093010783195496, "global_step": 143312, "epoch": 1610, "lr": 4.57510235070864e-05}
{"train_loss": 0.12866120040416718, "global_step": 143313, "epoch": 1610, "lr": 4.575044586785143e-05}
{"train_loss": 0.12026356905698776, "global_step": 143314, "epoch": 1610, "lr": 4.574986822918778e-05}
{"train_loss": 0.13093027472496033, "global_step": 143315, "epoch": 1610, "lr": 4.574929059109549e-05}
{"train_loss": 0.13178907334804535, "global_step": 143316, "epoch": 1610, "lr": 4.574871295357468e-05}
{"train_loss": 0.12321986258029938, "global_step": 143317, "epoch": 1610, "lr": 4.5748135316625373e-05}
{"train_loss": 0.09740572422742844, "global_step": 143318, "epoch": 1610, "lr": 4.574755768024771e-05}
{"train_loss": 0.06214441731572151, "global_step": 143319, "epoch": 1610, "lr": 4.5746980044441704e-05}
{"train_loss": 0.1594521850347519, "global_step": 143320, "epoch": 1610, "lr": 4.57464024092075e-05}
{"train_loss": 0.13693886995315552, "global_step": 143321, "epoch": 1610, "lr": 4.5745824774545105e-05}
{"train_loss": 0.08610621094703674, "global_step": 143322, "epoch": 1610, "lr": 4.5745247140454636e-05}
{"train_loss": 0.1155073270201683, "global_step": 143323, "epoch": 1610, "lr": 4.5744669506936186e-05}
{"train_loss": 0.12282389402389526, "global_step": 143324, "epoch": 1610, "lr": 4.57440918739898e-05}
{"train_loss": 0.1261547952890396, "global_step": 143325, "epoch": 1610, "lr": 4.5743514241615584e-05}
{"train_loss": 0.1479707956314087, "global_step": 143326, "epoch": 1610, "lr": 4.5742936609813595e-05}
{"train_loss": 0.1352088451385498, "global_step": 143327, "epoch": 1610, "lr": 4.574235897858391e-05}
{"train_loss": 0.09050699323415756, "global_step": 143328, "epoch": 1610, "lr": 4.574178134792662e-05}
{"train_loss": 0.047379277646541595, "global_step": 143329, "epoch": 1610, "lr": 4.574120371784179e-05}
{"train_loss": 0.12620730698108673, "global_step": 143330, "epoch": 1610, "lr": 4.574062608832951e-05}
{"train_loss": 0.12773923575878143, "global_step": 143331, "epoch": 1610, "lr": 4.574004845938985e-05}
{"train_loss": 0.11172982305288315, "global_step": 143332, "epoch": 1610, "lr": 4.5739470831022875e-05}
{"train_loss": 0.10100962966680527, "global_step": 143333, "epoch": 1610, "lr": 4.57388932032287e-05}
{"train_loss": 0.08506765961647034, "global_step": 143334, "epoch": 1610, "lr": 4.573831557600735e-05}
{"train_loss": 0.08750833570957184, "global_step": 143335, "epoch": 1610, "lr": 4.573773794935896e-05}
{"train_loss": 0.1206931471824646, "global_step": 143336, "epoch": 1610, "lr": 4.573716032328355e-05}
{"train_loss": 0.08833714574575424, "global_step": 143337, "epoch": 1610, "lr": 4.573658269778125e-05}
{"train_loss": 0.1811046004295349, "global_step": 143338, "epoch": 1610, "lr": 4.57360050728521e-05}
{"train_loss": 0.11279433965682983, "global_step": 143339, "epoch": 1610, "lr": 4.573542744849621e-05}
{"train_loss": 0.1040615513920784, "global_step": 143340, "epoch": 1610, "lr": 4.573484982471362e-05}
{"train_loss": 0.08858223259449005, "global_step": 143341, "epoch": 1610, "lr": 4.5734272201504426e-05}
{"train_loss": 0.13054843246936798, "global_step": 143342, "epoch": 1610, "lr": 4.573369457886872e-05}
{"train_loss": 0.08601654320955276, "global_step": 143343, "epoch": 1610, "lr": 4.573311695680656e-05}
{"train_loss": 0.10787784308195114, "global_step": 143344, "epoch": 1610, "lr": 4.573253933531804e-05}
{"train_loss": 0.11458371579647064, "global_step": 143345, "epoch": 1610, "lr": 4.573196171440323e-05}
{"train_loss": 0.10189972817897797, "global_step": 143346, "epoch": 1610, "lr": 4.573138409406219e-05}
{"train_loss": 0.0712774470448494, "global_step": 143347, "epoch": 1610, "lr": 4.5730806474295024e-05}
{"train_loss": 0.1417493373155594, "global_step": 143348, "epoch": 1610, "lr": 4.57302288551018e-05}
{"train_loss": 0.041550323367118835, "global_step": 143349, "epoch": 1610, "lr": 4.572965123648259e-05}
{"train_loss": 0.13633659482002258, "global_step": 143350, "epoch": 1610, "lr": 4.572907361843749e-05}
{"train_loss": 0.09440626949071884, "global_step": 143351, "epoch": 1610, "lr": 4.5728496000966536e-05}
{"train_loss": 0.1071450412273407, "global_step": 143352, "epoch": 1610, "lr": 4.572791838406986e-05}
{"train_loss": 0.10856804251670837, "global_step": 143353, "epoch": 1610, "lr": 4.572734076774749e-05}
{"train_loss": 0.13574795424938202, "global_step": 143354, "epoch": 1610, "lr": 4.572676315199955e-05}
{"train_loss": 0.15072894096374512, "global_step": 143355, "epoch": 1610, "lr": 4.572618553682608e-05}
{"train_loss": 0.10599105060100555, "global_step": 143356, "epoch": 1610, "lr": 4.5725607922227186e-05}
{"train_loss": 0.09035109728574753, "global_step": 143357, "epoch": 1610, "lr": 4.572503030820291e-05}
{"train_loss": 0.2311227172613144, "global_step": 143358, "epoch": 1610, "lr": 4.572445269475337e-05}
{"train_loss": 0.06716743856668472, "global_step": 143359, "epoch": 1610, "lr": 4.572387508187861e-05}
{"train_loss": 0.1083047166466713, "global_step": 143360, "epoch": 1610, "lr": 4.572329746957872e-05}
{"train_loss": 0.09167751669883728, "global_step": 143361, "epoch": 1610, "lr": 4.572271985785381e-05}
{"train_loss": 0.08981315791606903, "global_step": 143362, "epoch": 1610, "lr": 4.57221422467039e-05}
{"train_loss": 0.08596371859312057, "global_step": 143363, "epoch": 1610, "lr": 4.572156463612911e-05}
{"train_loss": 0.09500811994075775, "global_step": 143364, "epoch": 1610, "lr": 4.5720987026129505e-05}
{"train_loss": 0.10296423733234406, "global_step": 143365, "epoch": 1610, "lr": 4.572040941670516e-05}
{"train_loss": 0.2058965563774109, "global_step": 143366, "epoch": 1610, "lr": 4.571983180785614e-05}
{"train_loss": 0.10351940989494324, "global_step": 143367, "epoch": 1610, "lr": 4.5719254199582554e-05}
{"train_loss": 0.15767772495746613, "global_step": 143368, "epoch": 1610, "lr": 4.571867659188445e-05}
{"train_loss": 0.13308151066303253, "global_step": 143369, "epoch": 1610, "lr": 4.5718098984761934e-05}
{"train_loss": 0.0751262679696083, "global_step": 143370, "epoch": 1610, "lr": 4.5717521378215046e-05}
{"train_loss": 0.12619058787822723, "global_step": 143371, "epoch": 1610, "lr": 4.57169437722439e-05}
{"train_loss": 0.03988143801689148, "global_step": 143372, "epoch": 1610, "lr": 4.5716366166848545e-05}
{"train_loss": 0.07693097740411758, "global_step": 143373, "epoch": 1610, "lr": 4.5715788562029096e-05}
{"train_loss": 0.11468616127967834, "global_step": 143374, "epoch": 1610, "lr": 4.5715210957785574e-05}
{"train_loss": 0.12294957041740417, "global_step": 143375, "epoch": 1610, "lr": 4.571463335411812e-05}
{"train_loss": 0.09445483982563019, "global_step": 143376, "epoch": 1610, "lr": 4.571405575102676e-05}
{"train_loss": 0.1174883171916008, "global_step": 143377, "epoch": 1610, "lr": 4.571347814851159e-05}
{"train_loss": 0.11334112223781896, "global_step": 143378, "epoch": 1610, "lr": 4.571290054657271e-05, "val_loss": 5.988471984863281, "train_action_mse_error": 18.508464813232422}
{"train_loss": 0.07369348406791687, "global_step": 143379, "epoch": 1611, "lr": 4.5712322945210164e-05}
{"train_loss": 0.1011231318116188, "global_step": 143380, "epoch": 1611, "lr": 4.571174534442406e-05}
{"train_loss": 0.12846925854682922, "global_step": 143381, "epoch": 1611, "lr": 4.571116774421444e-05}
{"train_loss": 0.16441264748573303, "global_step": 143382, "epoch": 1611, "lr": 4.5710590144581424e-05}
{"train_loss": 0.11023857444524765, "global_step": 143383, "epoch": 1611, "lr": 4.571001254552505e-05}
{"train_loss": 0.052494633942842484, "global_step": 143384, "epoch": 1611, "lr": 4.570943494704541e-05}
{"train_loss": 0.07640920579433441, "global_step": 143385, "epoch": 1611, "lr": 4.57088573491426e-05}
{"train_loss": 0.10021710395812988, "global_step": 143386, "epoch": 1611, "lr": 4.570827975181667e-05}
{"train_loss": 0.15361884236335754, "global_step": 143387, "epoch": 1611, "lr": 4.570770215506771e-05}
{"train_loss": 0.17392778396606445, "global_step": 143388, "epoch": 1611, "lr": 4.570712455889581e-05}
{"train_loss": 0.11225935071706772, "global_step": 143389, "epoch": 1611, "lr": 4.570654696330102e-05}
{"train_loss": 0.11209702491760254, "global_step": 143390, "epoch": 1611, "lr": 4.570596936828345e-05}
{"train_loss": 0.06467841565608978, "global_step": 143391, "epoch": 1611, "lr": 4.5705391773843134e-05}
{"train_loss": 0.08715080469846725, "global_step": 143392, "epoch": 1611, "lr": 4.570481417998019e-05}
{"train_loss": 0.07304374873638153, "global_step": 143393, "epoch": 1611, "lr": 4.570423658669467e-05}
{"train_loss": 0.07955722510814667, "global_step": 143394, "epoch": 1611, "lr": 4.5703658993986685e-05}
{"train_loss": 0.06723353266716003, "global_step": 143395, "epoch": 1611, "lr": 4.570308140185627e-05}
{"train_loss": 0.14745549857616425, "global_step": 143396, "epoch": 1611, "lr": 4.570250381030352e-05}
{"train_loss": 0.09723129123449326, "global_step": 143397, "epoch": 1611, "lr": 4.570192621932854e-05}
{"train_loss": 0.09700124710798264, "global_step": 143398, "epoch": 1611, "lr": 4.570134862893136e-05}
{"train_loss": 0.08235762268304825, "global_step": 143399, "epoch": 1611, "lr": 4.57007710391121e-05}
{"train_loss": 0.07151353359222412, "global_step": 143400, "epoch": 1611, "lr": 4.570019344987081e-05}
{"train_loss": 0.08899975568056107, "global_step": 143401, "epoch": 1611, "lr": 4.569961586120758e-05}
{"train_loss": 0.05966360867023468, "global_step": 143402, "epoch": 1611, "lr": 4.569903827312249e-05}
{"train_loss": 0.20779506862163544, "global_step": 143403, "epoch": 1611, "lr": 4.56984606856156e-05}
{"train_loss": 0.11608324944972992, "global_step": 143404, "epoch": 1611, "lr": 4.5697883098687e-05}
{"train_loss": 0.125258207321167, "global_step": 143405, "epoch": 1611, "lr": 4.5697305512336776e-05}
{"train_loss": 0.12457701563835144, "global_step": 143406, "epoch": 1611, "lr": 4.569672792656499e-05}
{"train_loss": 0.17397597432136536, "global_step": 143407, "epoch": 1611, "lr": 4.569615034137174e-05}
{"train_loss": 0.13027507066726685, "global_step": 143408, "epoch": 1611, "lr": 4.5695572756757064e-05}
{"train_loss": 0.11433659493923187, "global_step": 143409, "epoch": 1611, "lr": 4.569499517272109e-05}
{"train_loss": 0.08683091402053833, "global_step": 143410, "epoch": 1611, "lr": 4.5694417589263854e-05}
{"train_loss": 0.13303396105766296, "global_step": 143411, "epoch": 1611, "lr": 4.569384000638547e-05}
{"train_loss": 0.09884819388389587, "global_step": 143412, "epoch": 1611, "lr": 4.569326242408597e-05}
{"train_loss": 0.16085697710514069, "global_step": 143413, "epoch": 1611, "lr": 4.569268484236547e-05}
{"train_loss": 0.1470271199941635, "global_step": 143414, "epoch": 1611, "lr": 4.569210726122405e-05}
{"train_loss": 0.06341168284416199, "global_step": 143415, "epoch": 1611, "lr": 4.569152968066175e-05}
{"train_loss": 0.0736517608165741, "global_step": 143416, "epoch": 1611, "lr": 4.56909521006787e-05}
{"train_loss": 0.08484956622123718, "global_step": 143417, "epoch": 1611, "lr": 4.569037452127492e-05}
{"train_loss": 0.16746781766414642, "global_step": 143418, "epoch": 1611, "lr": 4.568979694245054e-05}
{"train_loss": 0.0571724995970726, "global_step": 143419, "epoch": 1611, "lr": 4.5689219364205595e-05}
{"train_loss": 0.10983613133430481, "global_step": 143420, "epoch": 1611, "lr": 4.56886417865402e-05}
{"train_loss": 0.09902910888195038, "global_step": 143421, "epoch": 1611, "lr": 4.5688064209454404e-05}
{"train_loss": 0.04371751472353935, "global_step": 143422, "epoch": 1611, "lr": 4.56874866329483e-05}
{"train_loss": 0.11319558322429657, "global_step": 143423, "epoch": 1611, "lr": 4.568690905702196e-05}
{"train_loss": 0.08721446990966797, "global_step": 143424, "epoch": 1611, "lr": 4.568633148167546e-05}
{"train_loss": 0.0839306116104126, "global_step": 143425, "epoch": 1611, "lr": 4.568575390690888e-05}
{"train_loss": 0.14007991552352905, "global_step": 143426, "epoch": 1611, "lr": 4.56851763327223e-05}
{"train_loss": 0.10062458366155624, "global_step": 143427, "epoch": 1611, "lr": 4.5684598759115794e-05}
{"train_loss": 0.08790307492017746, "global_step": 143428, "epoch": 1611, "lr": 4.5684021186089455e-05}
{"train_loss": 0.0463164784014225, "global_step": 143429, "epoch": 1611, "lr": 4.568344361364332e-05}
{"train_loss": 0.10262962430715561, "global_step": 143430, "epoch": 1611, "lr": 4.568286604177752e-05}
{"train_loss": 0.08090972900390625, "global_step": 143431, "epoch": 1611, "lr": 4.5682288470492076e-05}
{"train_loss": 0.1133367046713829, "global_step": 143432, "epoch": 1611, "lr": 4.56817108997871e-05}
{"train_loss": 0.06123063340783119, "global_step": 143433, "epoch": 1611, "lr": 4.568113332966269e-05}
{"train_loss": 0.07849026471376419, "global_step": 143434, "epoch": 1611, "lr": 4.568055576011888e-05}
{"train_loss": 0.10228084772825241, "global_step": 143435, "epoch": 1611, "lr": 4.567997819115578e-05}
{"train_loss": 0.14195454120635986, "global_step": 143436, "epoch": 1611, "lr": 4.567940062277344e-05}
{"train_loss": 0.11260277777910233, "global_step": 143437, "epoch": 1611, "lr": 4.567882305497197e-05}
{"train_loss": 0.07767914980649948, "global_step": 143438, "epoch": 1611, "lr": 4.56782454877514e-05}
{"train_loss": 0.1056542620062828, "global_step": 143439, "epoch": 1611, "lr": 4.5677667921111875e-05}
{"train_loss": 0.13307908177375793, "global_step": 143440, "epoch": 1611, "lr": 4.5677090355053405e-05}
{"train_loss": 0.10622260719537735, "global_step": 143441, "epoch": 1611, "lr": 4.5676512789576116e-05}
{"train_loss": 0.09023682028055191, "global_step": 143442, "epoch": 1611, "lr": 4.567593522468006e-05}
{"train_loss": 0.16639412939548492, "global_step": 143443, "epoch": 1611, "lr": 4.567535766036532e-05}
{"train_loss": 0.09520044177770615, "global_step": 143444, "epoch": 1611, "lr": 4.5674780096631975e-05}
{"train_loss": 0.08602286130189896, "global_step": 143445, "epoch": 1611, "lr": 4.567420253348011e-05}
{"train_loss": 0.1384984403848648, "global_step": 143446, "epoch": 1611, "lr": 4.567362497090979e-05}
{"train_loss": 0.0645260289311409, "global_step": 143447, "epoch": 1611, "lr": 4.567304740892111e-05}
{"train_loss": 0.09617382287979126, "global_step": 143448, "epoch": 1611, "lr": 4.567246984751411e-05}
{"train_loss": 0.1428012251853943, "global_step": 143449, "epoch": 1611, "lr": 4.56718922866889e-05}
{"train_loss": 0.11285856366157532, "global_step": 143450, "epoch": 1611, "lr": 4.567131472644557e-05}
{"train_loss": 0.1415613889694214, "global_step": 143451, "epoch": 1611, "lr": 4.567073716678416e-05}
{"train_loss": 0.11784560233354568, "global_step": 143452, "epoch": 1611, "lr": 4.567015960770479e-05}
{"train_loss": 0.14803577959537506, "global_step": 143453, "epoch": 1611, "lr": 4.566958204920748e-05}
{"train_loss": 0.135574072599411, "global_step": 143454, "epoch": 1611, "lr": 4.5669004491292374e-05}
{"train_loss": 0.1094241589307785, "global_step": 143455, "epoch": 1611, "lr": 4.5668426933959496e-05}
{"train_loss": 0.11267243325710297, "global_step": 143456, "epoch": 1611, "lr": 4.566784937720896e-05}
{"train_loss": 0.10380041599273682, "global_step": 143457, "epoch": 1611, "lr": 4.5667271821040816e-05}
{"train_loss": 0.06789182126522064, "global_step": 143458, "epoch": 1611, "lr": 4.566669426545518e-05}
{"train_loss": 0.12962903082370758, "global_step": 143459, "epoch": 1611, "lr": 4.566611671045207e-05}
{"train_loss": 0.18970923125743866, "global_step": 143460, "epoch": 1611, "lr": 4.5665539156031626e-05}
{"train_loss": 0.14877904951572418, "global_step": 143461, "epoch": 1611, "lr": 4.5664961602193883e-05}
{"train_loss": 0.14427116513252258, "global_step": 143462, "epoch": 1611, "lr": 4.566438404893894e-05}
{"train_loss": 0.13758060336112976, "global_step": 143463, "epoch": 1611, "lr": 4.5663806496266864e-05}
{"train_loss": 0.08180838823318481, "global_step": 143464, "epoch": 1611, "lr": 4.566322894417773e-05}
{"train_loss": 0.08419223874807358, "global_step": 143465, "epoch": 1611, "lr": 4.566265139267163e-05}
{"train_loss": 0.1207490935921669, "global_step": 143466, "epoch": 1611, "lr": 4.5662073841748645e-05}
{"train_loss": 0.10846477177705657, "global_step": 143467, "epoch": 1611, "lr": 4.566149629140881e-05, "val_loss": 6.097923755645752}
{"train_loss": 0.14644449949264526, "global_step": 143468, "epoch": 1612, "lr": 4.566091874165225e-05}
{"train_loss": 0.08137422800064087, "global_step": 143469, "epoch": 1612, "lr": 4.5660341192479034e-05}
{"train_loss": 0.16174690425395966, "global_step": 143470, "epoch": 1612, "lr": 4.565976364388922e-05}
{"train_loss": 0.08238698542118073, "global_step": 143471, "epoch": 1612, "lr": 4.565918609588291e-05}
{"train_loss": 0.0792311355471611, "global_step": 143472, "epoch": 1612, "lr": 4.5658608548460154e-05}
{"train_loss": 0.20278534293174744, "global_step": 143473, "epoch": 1612, "lr": 4.5658031001621066e-05}
{"train_loss": 0.05891900137066841, "global_step": 143474, "epoch": 1612, "lr": 4.565745345536568e-05}
{"train_loss": 0.08302280306816101, "global_step": 143475, "epoch": 1612, "lr": 4.565687590969412e-05}
{"train_loss": 0.09135610610246658, "global_step": 143476, "epoch": 1612, "lr": 4.5656298364606416e-05}
{"train_loss": 0.08223909884691238, "global_step": 143477, "epoch": 1612, "lr": 4.565572082010269e-05}
{"train_loss": 0.10582521557807922, "global_step": 143478, "epoch": 1612, "lr": 4.565514327618298e-05}
{"train_loss": 0.16504605114459991, "global_step": 143479, "epoch": 1612, "lr": 4.565456573284741e-05}
{"train_loss": 0.20835426449775696, "global_step": 143480, "epoch": 1612, "lr": 4.5653988190096014e-05}
{"train_loss": 0.07615502923727036, "global_step": 143481, "epoch": 1612, "lr": 4.565341064792889e-05}
{"train_loss": 0.12095778435468674, "global_step": 143482, "epoch": 1612, "lr": 4.5652833106346106e-05}
{"train_loss": 0.10016946494579315, "global_step": 143483, "epoch": 1612, "lr": 4.565225556534775e-05}
{"train_loss": 0.15694639086723328, "global_step": 143484, "epoch": 1612, "lr": 4.5651678024933894e-05}
{"train_loss": 0.11353505402803421, "global_step": 143485, "epoch": 1612, "lr": 4.5651100485104626e-05}
{"train_loss": 0.09052809327840805, "global_step": 143486, "epoch": 1612, "lr": 4.5650522945859994e-05}
{"train_loss": 0.1236938089132309, "global_step": 143487, "epoch": 1612, "lr": 4.56499454072001e-05}
{"train_loss": 0.20757631957530975, "global_step": 143488, "epoch": 1612, "lr": 4.564936786912504e-05}
{"train_loss": 0.08227555453777313, "global_step": 143489, "epoch": 1612, "lr": 4.564879033163484e-05}
{"train_loss": 0.06657467782497406, "global_step": 143490, "epoch": 1612, "lr": 4.564821279472963e-05}
{"train_loss": 0.10511159151792526, "global_step": 143491, "epoch": 1612, "lr": 4.5647635258409446e-05}
{"train_loss": 0.04234892874956131, "global_step": 143492, "epoch": 1612, "lr": 4.564705772267441e-05}
{"train_loss": 0.09610134363174438, "global_step": 143493, "epoch": 1612, "lr": 4.5646480187524546e-05}
{"train_loss": 0.07487247884273529, "global_step": 143494, "epoch": 1612, "lr": 4.564590265295998e-05}
{"train_loss": 0.06012805178761482, "global_step": 143495, "epoch": 1612, "lr": 4.5645325118980756e-05}
{"train_loss": 0.13041086494922638, "global_step": 143496, "epoch": 1612, "lr": 4.564474758558698e-05}
{"train_loss": 0.10675802826881409, "global_step": 143497, "epoch": 1612, "lr": 4.5644170052778695e-05}
{"train_loss": 0.1477956473827362, "global_step": 143498, "epoch": 1612, "lr": 4.564359252055602e-05}
{"train_loss": 0.0900115892291069, "global_step": 143499, "epoch": 1612, "lr": 4.5643014988919004e-05}
{"train_loss": 0.09960537403821945, "global_step": 143500, "epoch": 1612, "lr": 4.564243745786773e-05}
{"train_loss": 0.17104263603687286, "global_step": 143501, "epoch": 1612, "lr": 4.564185992740227e-05}
{"train_loss": 0.10589436441659927, "global_step": 143502, "epoch": 1612, "lr": 4.564128239752271e-05}
{"train_loss": 0.05865122377872467, "global_step": 143503, "epoch": 1612, "lr": 4.5640704868229136e-05}
{"train_loss": 0.14595022797584534, "global_step": 143504, "epoch": 1612, "lr": 4.56401273395216e-05}
{"train_loss": 0.0889497846364975, "global_step": 143505, "epoch": 1612, "lr": 4.563954981140021e-05}
{"train_loss": 0.10665058344602585, "global_step": 143506, "epoch": 1612, "lr": 4.563897228386502e-05}
{"train_loss": 0.05745123326778412, "global_step": 143507, "epoch": 1612, "lr": 4.563839475691613e-05}
{"train_loss": 0.09385651350021362, "global_step": 143508, "epoch": 1612, "lr": 4.5637817230553585e-05}
{"train_loss": 0.11505455523729324, "global_step": 143509, "epoch": 1612, "lr": 4.563723970477751e-05}
{"train_loss": 0.12899720668792725, "global_step": 143510, "epoch": 1612, "lr": 4.5636662179587924e-05}
{"train_loss": 0.08169308304786682, "global_step": 143511, "epoch": 1612, "lr": 4.563608465498496e-05}
{"train_loss": 0.08652351796627045, "global_step": 143512, "epoch": 1612, "lr": 4.5635507130968654e-05}
{"train_loss": 0.1493569016456604, "global_step": 143513, "epoch": 1612, "lr": 4.563492960753912e-05}
{"train_loss": 0.08097492158412933, "global_step": 143514, "epoch": 1612, "lr": 4.5634352084696397e-05}
{"train_loss": 0.10938472300767899, "global_step": 143515, "epoch": 1612, "lr": 4.56337745624406e-05}
{"train_loss": 0.12184837460517883, "global_step": 143516, "epoch": 1612, "lr": 4.563319704077177e-05}
{"train_loss": 0.1131538450717926, "global_step": 143517, "epoch": 1612, "lr": 4.5632619519690024e-05}
{"train_loss": 0.21857896447181702, "global_step": 143518, "epoch": 1612, "lr": 4.563204199919541e-05}
{"train_loss": 0.10686344653367996, "global_step": 143519, "epoch": 1612, "lr": 4.5631464479288e-05}
{"train_loss": 0.09714881330728531, "global_step": 143520, "epoch": 1612, "lr": 4.56308869599679e-05}
{"train_loss": 0.16634809970855713, "global_step": 143521, "epoch": 1612, "lr": 4.5630309441235184e-05}
{"train_loss": 0.122947558760643, "global_step": 143522, "epoch": 1612, "lr": 4.5629731923089905e-05}
{"train_loss": 0.12129320204257965, "global_step": 143523, "epoch": 1612, "lr": 4.562915440553215e-05}
{"train_loss": 0.07251056283712387, "global_step": 143524, "epoch": 1612, "lr": 4.562857688856202e-05}
{"train_loss": 0.0635785162448883, "global_step": 143525, "epoch": 1612, "lr": 4.562799937217955e-05}
{"train_loss": 0.10449031740427017, "global_step": 143526, "epoch": 1612, "lr": 4.562742185638487e-05}
{"train_loss": 0.14284294843673706, "global_step": 143527, "epoch": 1612, "lr": 4.562684434117801e-05}
{"train_loss": 0.08494038134813309, "global_step": 143528, "epoch": 1612, "lr": 4.5626266826559084e-05}
{"train_loss": 0.05912318453192711, "global_step": 143529, "epoch": 1612, "lr": 4.562568931252813e-05}
{"train_loss": 0.059970010071992874, "global_step": 143530, "epoch": 1612, "lr": 4.562511179908527e-05}
{"train_loss": 0.13727416098117828, "global_step": 143531, "epoch": 1612, "lr": 4.562453428623055e-05}
{"train_loss": 0.14171645045280457, "global_step": 143532, "epoch": 1612, "lr": 4.562395677396407e-05}
{"train_loss": 0.12798574566841125, "global_step": 143533, "epoch": 1612, "lr": 4.562337926228588e-05}
{"train_loss": 0.10835156589746475, "global_step": 143534, "epoch": 1612, "lr": 4.5622801751196084e-05}
{"train_loss": 0.14524373412132263, "global_step": 143535, "epoch": 1612, "lr": 4.5622224240694736e-05}
{"train_loss": 0.09182355552911758, "global_step": 143536, "epoch": 1612, "lr": 4.562164673078195e-05}
{"train_loss": 0.10693880170583725, "global_step": 143537, "epoch": 1612, "lr": 4.5621069221457765e-05}
{"train_loss": 0.11678864806890488, "global_step": 143538, "epoch": 1612, "lr": 4.562049171272228e-05}
{"train_loss": 0.07328662276268005, "global_step": 143539, "epoch": 1612, "lr": 4.561991420457555e-05}
{"train_loss": 0.11372258514165878, "global_step": 143540, "epoch": 1612, "lr": 4.561933669701767e-05}
{"train_loss": 0.16286395490169525, "global_step": 143541, "epoch": 1612, "lr": 4.561875919004874e-05}
{"train_loss": 0.09659041464328766, "global_step": 143542, "epoch": 1612, "lr": 4.561818168366879e-05}
{"train_loss": 0.11363556236028671, "global_step": 143543, "epoch": 1612, "lr": 4.561760417787795e-05}
{"train_loss": 0.19972510635852814, "global_step": 143544, "epoch": 1612, "lr": 4.561702667267624e-05}
{"train_loss": 0.07686448842287064, "global_step": 143545, "epoch": 1612, "lr": 4.561644916806379e-05}
{"train_loss": 0.11313705146312714, "global_step": 143546, "epoch": 1612, "lr": 4.561587166404063e-05}
{"train_loss": 0.09727198630571365, "global_step": 143547, "epoch": 1612, "lr": 4.561529416060689e-05}
{"train_loss": 0.08645416051149368, "global_step": 143548, "epoch": 1612, "lr": 4.56147166577626e-05}
{"train_loss": 0.08683385699987411, "global_step": 143549, "epoch": 1612, "lr": 4.561413915550788e-05}
{"train_loss": 0.12347929924726486, "global_step": 143550, "epoch": 1612, "lr": 4.5613561653842763e-05}
{"train_loss": 0.17298579216003418, "global_step": 143551, "epoch": 1612, "lr": 4.561298415276737e-05}
{"train_loss": 0.09740111976861954, "global_step": 143552, "epoch": 1612, "lr": 4.561240665228173e-05}
{"train_loss": 0.12018975615501404, "global_step": 143553, "epoch": 1612, "lr": 4.561182915238598e-05}
{"train_loss": 0.17800363898277283, "global_step": 143554, "epoch": 1612, "lr": 4.561125165308014e-05}
{"train_loss": 0.1418842375278473, "global_step": 143555, "epoch": 1612, "lr": 4.561067415436433e-05}
{"train_loss": 0.11267641220199928, "global_step": 143556, "epoch": 1612, "lr": 4.56100966562386e-05, "val_loss": 6.178694725036621}
{"train_loss": 0.1370171159505844, "global_step": 143557, "epoch": 1613, "lr": 4.560951915870305e-05}
{"train_loss": 0.10117124766111374, "global_step": 143558, "epoch": 1613, "lr": 4.560894166175774e-05}
{"train_loss": 0.061242423951625824, "global_step": 143559, "epoch": 1613, "lr": 4.560836416540274e-05}
{"train_loss": 0.14747215807437897, "global_step": 143560, "epoch": 1613, "lr": 4.5607786669638175e-05}
{"train_loss": 0.13147348165512085, "global_step": 143561, "epoch": 1613, "lr": 4.560720917446406e-05}
{"train_loss": 0.0688694640994072, "global_step": 143562, "epoch": 1613, "lr": 4.560663167988053e-05}
{"train_loss": 0.10088077932596207, "global_step": 143563, "epoch": 1613, "lr": 4.5606054185887605e-05}
{"train_loss": 0.08555233478546143, "global_step": 143564, "epoch": 1613, "lr": 4.560547669248541e-05}
{"train_loss": 0.07356467843055725, "global_step": 143565, "epoch": 1613, "lr": 4.5604899199674e-05}
{"train_loss": 0.062244709581136703, "global_step": 143566, "epoch": 1613, "lr": 4.560432170745347e-05}
{"train_loss": 0.13931065797805786, "global_step": 143567, "epoch": 1613, "lr": 4.5603744215823865e-05}
{"train_loss": 0.09715935587882996, "global_step": 143568, "epoch": 1613, "lr": 4.56031667247853e-05}
{"train_loss": 0.11409834027290344, "global_step": 143569, "epoch": 1613, "lr": 4.560258923433782e-05}
{"train_loss": 0.1840905100107193, "global_step": 143570, "epoch": 1613, "lr": 4.5602011744481535e-05}
{"train_loss": 0.08132407814264297, "global_step": 143571, "epoch": 1613, "lr": 4.560143425521649e-05}
{"train_loss": 0.10782042890787125, "global_step": 143572, "epoch": 1613, "lr": 4.560085676654279e-05}
{"train_loss": 0.07589223235845566, "global_step": 143573, "epoch": 1613, "lr": 4.560027927846049e-05}
{"train_loss": 0.07872750610113144, "global_step": 143574, "epoch": 1613, "lr": 4.559970179096969e-05}
{"train_loss": 0.11354753375053406, "global_step": 143575, "epoch": 1613, "lr": 4.559912430407045e-05}
{"train_loss": 0.0811515524983406, "global_step": 143576, "epoch": 1613, "lr": 4.5598546817762856e-05}
{"train_loss": 0.11664086580276489, "global_step": 143577, "epoch": 1613, "lr": 4.5597969332047e-05}
{"train_loss": 0.12091659009456635, "global_step": 143578, "epoch": 1613, "lr": 4.559739184692291e-05}
{"train_loss": 0.09354883432388306, "global_step": 143579, "epoch": 1613, "lr": 4.559681436239073e-05}
{"train_loss": 0.15598395466804504, "global_step": 143580, "epoch": 1613, "lr": 4.559623687845048e-05}
{"train_loss": 0.11908992379903793, "global_step": 143581, "epoch": 1613, "lr": 4.5595659395102275e-05}
{"train_loss": 0.057635482400655746, "global_step": 143582, "epoch": 1613, "lr": 4.559508191234617e-05}
{"train_loss": 0.08609996736049652, "global_step": 143583, "epoch": 1613, "lr": 4.559450443018227e-05}
{"train_loss": 0.09369757026433945, "global_step": 143584, "epoch": 1613, "lr": 4.559392694861061e-05}
{"train_loss": 0.15094970166683197, "global_step": 143585, "epoch": 1613, "lr": 4.559334946763132e-05}
{"train_loss": 0.13897667825222015, "global_step": 143586, "epoch": 1613, "lr": 4.559277198724443e-05}
{"train_loss": 0.07642799615859985, "global_step": 143587, "epoch": 1613, "lr": 4.5592194507450057e-05}
{"train_loss": 0.10013489425182343, "global_step": 143588, "epoch": 1613, "lr": 4.5591617028248236e-05}
{"train_loss": 0.12523826956748962, "global_step": 143589, "epoch": 1613, "lr": 4.5591039549639094e-05}
{"train_loss": 0.10155081003904343, "global_step": 143590, "epoch": 1613, "lr": 4.559046207162265e-05}
{"train_loss": 0.07862891256809235, "global_step": 143591, "epoch": 1613, "lr": 4.558988459419905e-05}
{"train_loss": 0.13452769815921783, "global_step": 143592, "epoch": 1613, "lr": 4.558930711736831e-05}
{"train_loss": 0.15883812308311462, "global_step": 143593, "epoch": 1613, "lr": 4.558872964113056e-05}
{"train_loss": 0.10055434703826904, "global_step": 143594, "epoch": 1613, "lr": 4.5588152165485823e-05}
{"train_loss": 0.12181700021028519, "global_step": 143595, "epoch": 1613, "lr": 4.5587574690434214e-05}
{"train_loss": 0.12869928777217865, "global_step": 143596, "epoch": 1613, "lr": 4.558699721597582e-05}
{"train_loss": 0.12268514186143875, "global_step": 143597, "epoch": 1613, "lr": 4.558641974211068e-05}
{"train_loss": 0.07397350668907166, "global_step": 143598, "epoch": 1613, "lr": 4.55858422688389e-05}
{"train_loss": 0.12108288705348969, "global_step": 143599, "epoch": 1613, "lr": 4.558526479616054e-05}
{"train_loss": 0.06943162530660629, "global_step": 143600, "epoch": 1613, "lr": 4.55846873240757e-05}
{"train_loss": 0.1312919408082962, "global_step": 143601, "epoch": 1613, "lr": 4.5584109852584436e-05}
{"train_loss": 0.14325979351997375, "global_step": 143602, "epoch": 1613, "lr": 4.558353238168685e-05}
{"train_loss": 0.07870902866125107, "global_step": 143603, "epoch": 1613, "lr": 4.5582954911382983e-05}
{"train_loss": 0.07295215874910355, "global_step": 143604, "epoch": 1613, "lr": 4.558237744167295e-05}
{"train_loss": 0.23250234127044678, "global_step": 143605, "epoch": 1613, "lr": 4.55817999725568e-05}
{"train_loss": 0.05177893117070198, "global_step": 143606, "epoch": 1613, "lr": 4.558122250403464e-05}
{"train_loss": 0.06075073406100273, "global_step": 143607, "epoch": 1613, "lr": 4.558064503610651e-05}
{"train_loss": 0.1683807224035263, "global_step": 143608, "epoch": 1613, "lr": 4.558006756877253e-05}
{"train_loss": 0.14231589436531067, "global_step": 143609, "epoch": 1613, "lr": 4.5579490102032734e-05}
{"train_loss": 0.08624453842639923, "global_step": 143610, "epoch": 1613, "lr": 4.557891263588724e-05}
{"train_loss": 0.11046553403139114, "global_step": 143611, "epoch": 1613, "lr": 4.5578335170336096e-05}
{"train_loss": 0.11107248067855835, "global_step": 143612, "epoch": 1613, "lr": 4.5577757705379404e-05}
{"train_loss": 0.13903485238552094, "global_step": 143613, "epoch": 1613, "lr": 4.557718024101721e-05}
{"train_loss": 0.07728739082813263, "global_step": 143614, "epoch": 1613, "lr": 4.557660277724962e-05}
{"train_loss": 0.10771588236093521, "global_step": 143615, "epoch": 1613, "lr": 4.557602531407671e-05}
{"train_loss": 0.0973607748746872, "global_step": 143616, "epoch": 1613, "lr": 4.557544785149853e-05}
{"train_loss": 0.10938850045204163, "global_step": 143617, "epoch": 1613, "lr": 4.55748703895152e-05}
{"train_loss": 0.10975267738103867, "global_step": 143618, "epoch": 1613, "lr": 4.5574292928126757e-05}
{"train_loss": 0.165633425116539, "global_step": 143619, "epoch": 1613, "lr": 4.557371546733331e-05}
{"train_loss": 0.07579176127910614, "global_step": 143620, "epoch": 1613, "lr": 4.55731380071349e-05}
{"train_loss": 0.08803939074277878, "global_step": 143621, "epoch": 1613, "lr": 4.5572560547531654e-05}
{"train_loss": 0.08526470512151718, "global_step": 143622, "epoch": 1613, "lr": 4.557198308852361e-05}
{"train_loss": 0.09194671362638474, "global_step": 143623, "epoch": 1613, "lr": 4.557140563011087e-05}
{"train_loss": 0.1801171749830246, "global_step": 143624, "epoch": 1613, "lr": 4.557082817229348e-05}
{"train_loss": 0.17013326287269592, "global_step": 143625, "epoch": 1613, "lr": 4.557025071507156e-05}
{"train_loss": 0.12853091955184937, "global_step": 143626, "epoch": 1613, "lr": 4.5569673258445146e-05}
{"train_loss": 0.10335563868284225, "global_step": 143627, "epoch": 1613, "lr": 4.556909580241435e-05}
{"train_loss": 0.08133889734745026, "global_step": 143628, "epoch": 1613, "lr": 4.5568518346979225e-05}
{"train_loss": 0.1332152932882309, "global_step": 143629, "epoch": 1613, "lr": 4.556794089213987e-05}
{"train_loss": 0.1367746740579605, "global_step": 143630, "epoch": 1613, "lr": 4.556736343789634e-05}
{"train_loss": 0.13553224503993988, "global_step": 143631, "epoch": 1613, "lr": 4.5566785984248725e-05}
{"train_loss": 0.06851586699485779, "global_step": 143632, "epoch": 1613, "lr": 4.556620853119711e-05}
{"train_loss": 0.15802901983261108, "global_step": 143633, "epoch": 1613, "lr": 4.556563107874155e-05}
{"train_loss": 0.06045231223106384, "global_step": 143634, "epoch": 1613, "lr": 4.5565053626882166e-05}
{"train_loss": 0.14849112927913666, "global_step": 143635, "epoch": 1613, "lr": 4.5564476175618975e-05}
{"train_loss": 0.11691787838935852, "global_step": 143636, "epoch": 1613, "lr": 4.556389872495211e-05}
{"train_loss": 0.19166484475135803, "global_step": 143637, "epoch": 1613, "lr": 4.55633212748816e-05}
{"train_loss": 0.10380067676305771, "global_step": 143638, "epoch": 1613, "lr": 4.5562743825407574e-05}
{"train_loss": 0.18043214082717896, "global_step": 143639, "epoch": 1613, "lr": 4.5562166376530054e-05}
{"train_loss": 0.10809358209371567, "global_step": 143640, "epoch": 1613, "lr": 4.556158892824918e-05}
{"train_loss": 0.11247609555721283, "global_step": 143641, "epoch": 1613, "lr": 4.5561011480564966e-05}
{"train_loss": 0.13699612021446228, "global_step": 143642, "epoch": 1613, "lr": 4.5560434033477545e-05}
{"train_loss": 0.13652962446212769, "global_step": 143643, "epoch": 1613, "lr": 4.5559856586986945e-05}
{"train_loss": 0.10301601886749268, "global_step": 143644, "epoch": 1613, "lr": 4.555927914109329e-05}
{"train_loss": 0.11418161373794748, "global_step": 143645, "epoch": 1613, "lr": 4.5558701695796615e-05, "val_loss": 6.009692668914795}
{"train_loss": 0.08160997182130814, "global_step": 143646, "epoch": 1614, "lr": 4.555812425109704e-05}
{"train_loss": 0.09592173993587494, "global_step": 143647, "epoch": 1614, "lr": 4.55575468069946e-05}
{"train_loss": 0.11863122135400772, "global_step": 143648, "epoch": 1614, "lr": 4.555696936348942e-05}
{"train_loss": 0.07504275441169739, "global_step": 143649, "epoch": 1614, "lr": 4.555639192058152e-05}
{"train_loss": 0.1335940957069397, "global_step": 143650, "epoch": 1614, "lr": 4.555581447827103e-05}
{"train_loss": 0.14841534197330475, "global_step": 143651, "epoch": 1614, "lr": 4.5555237036558e-05}
{"train_loss": 0.14087696373462677, "global_step": 143652, "epoch": 1614, "lr": 4.555465959544251e-05}
{"train_loss": 0.10079383105039597, "global_step": 143653, "epoch": 1614, "lr": 4.555408215492466e-05}
{"train_loss": 0.09068941324949265, "global_step": 143654, "epoch": 1614, "lr": 4.5553504715004486e-05}
{"train_loss": 0.11709989607334137, "global_step": 143655, "epoch": 1614, "lr": 4.555292727568211e-05}
{"train_loss": 0.11358422040939331, "global_step": 143656, "epoch": 1614, "lr": 4.5552349836957567e-05}
{"train_loss": 0.08657814562320709, "global_step": 143657, "epoch": 1614, "lr": 4.5551772398830976e-05}
{"train_loss": 0.0920410007238388, "global_step": 143658, "epoch": 1614, "lr": 4.5551194961302376e-05}
{"train_loss": 0.1513289511203766, "global_step": 143659, "epoch": 1614, "lr": 4.5550617524371884e-05}
{"train_loss": 0.09739954769611359, "global_step": 143660, "epoch": 1614, "lr": 4.555004008803954e-05}
{"train_loss": 0.07609879970550537, "global_step": 143661, "epoch": 1614, "lr": 4.554946265230546e-05}
{"train_loss": 0.11398626863956451, "global_step": 143662, "epoch": 1614, "lr": 4.554888521716967e-05}
{"train_loss": 0.17971061170101166, "global_step": 143663, "epoch": 1614, "lr": 4.554830778263231e-05}
{"train_loss": 0.13283579051494598, "global_step": 143664, "epoch": 1614, "lr": 4.55477303486934e-05}
{"train_loss": 0.14051932096481323, "global_step": 143665, "epoch": 1614, "lr": 4.5547152915353065e-05}
{"train_loss": 0.12134655565023422, "global_step": 143666, "epoch": 1614, "lr": 4.5546575482611333e-05}
{"train_loss": 0.09043960273265839, "global_step": 143667, "epoch": 1614, "lr": 4.554599805046833e-05}
{"train_loss": 0.07191289216279984, "global_step": 143668, "epoch": 1614, "lr": 4.554542061892411e-05}
{"train_loss": 0.0489777997136116, "global_step": 143669, "epoch": 1614, "lr": 4.554484318797875e-05}
{"train_loss": 0.12425386905670166, "global_step": 143670, "epoch": 1614, "lr": 4.554426575763234e-05}
{"train_loss": 0.11402545869350433, "global_step": 143671, "epoch": 1614, "lr": 4.5543688327884946e-05}
{"train_loss": 0.05637545883655548, "global_step": 143672, "epoch": 1614, "lr": 4.5543110898736647e-05}
{"train_loss": 0.20967598259449005, "global_step": 143673, "epoch": 1614, "lr": 4.554253347018752e-05}
{"train_loss": 0.14330007135868073, "global_step": 143674, "epoch": 1614, "lr": 4.554195604223766e-05}
{"train_loss": 0.1057562306523323, "global_step": 143675, "epoch": 1614, "lr": 4.554137861488711e-05}
{"train_loss": 0.065322145819664, "global_step": 143676, "epoch": 1614, "lr": 4.5540801188135984e-05}
{"train_loss": 0.07886940240859985, "global_step": 143677, "epoch": 1614, "lr": 4.5540223761984325e-05}
{"train_loss": 0.10435362160205841, "global_step": 143678, "epoch": 1614, "lr": 4.553964633643224e-05}
{"train_loss": 0.0753144770860672, "global_step": 143679, "epoch": 1614, "lr": 4.5539068911479786e-05}
{"train_loss": 0.0984993726015091, "global_step": 143680, "epoch": 1614, "lr": 4.553849148712707e-05}
{"train_loss": 0.09094393253326416, "global_step": 143681, "epoch": 1614, "lr": 4.553791406337412e-05}
{"train_loss": 0.0803649052977562, "global_step": 143682, "epoch": 1614, "lr": 4.553733664022107e-05}
{"train_loss": 0.08962850272655487, "global_step": 143683, "epoch": 1614, "lr": 4.5536759217667954e-05}
{"train_loss": 0.09253834187984467, "global_step": 143684, "epoch": 1614, "lr": 4.553618179571488e-05}
{"train_loss": 0.067327581346035, "global_step": 143685, "epoch": 1614, "lr": 4.5535604374361894e-05}
{"train_loss": 0.11749010533094406, "global_step": 143686, "epoch": 1614, "lr": 4.553502695360909e-05}
{"train_loss": 0.10510796308517456, "global_step": 143687, "epoch": 1614, "lr": 4.553444953345657e-05}
{"train_loss": 0.059614937752485275, "global_step": 143688, "epoch": 1614, "lr": 4.5533872113904374e-05}
{"train_loss": 0.18799589574337006, "global_step": 143689, "epoch": 1614, "lr": 4.5533294694952596e-05}
{"train_loss": 0.11542639136314392, "global_step": 143690, "epoch": 1614, "lr": 4.5532717276601325e-05}
{"train_loss": 0.1252271831035614, "global_step": 143691, "epoch": 1614, "lr": 4.5532139858850606e-05}
{"train_loss": 0.14004570245742798, "global_step": 143692, "epoch": 1614, "lr": 4.553156244170055e-05}
{"train_loss": 0.07369223982095718, "global_step": 143693, "epoch": 1614, "lr": 4.553098502515122e-05}
{"train_loss": 0.08640803396701813, "global_step": 143694, "epoch": 1614, "lr": 4.5530407609202683e-05}
{"train_loss": 0.12017800658941269, "global_step": 143695, "epoch": 1614, "lr": 4.552983019385505e-05}
{"train_loss": 0.11148831248283386, "global_step": 143696, "epoch": 1614, "lr": 4.552925277910835e-05}
{"train_loss": 0.07461851090192795, "global_step": 143697, "epoch": 1614, "lr": 4.5528675364962714e-05}
{"train_loss": 0.0923304557800293, "global_step": 143698, "epoch": 1614, "lr": 4.5528097951418164e-05}
{"train_loss": 0.13821880519390106, "global_step": 143699, "epoch": 1614, "lr": 4.552752053847483e-05}
{"train_loss": 0.11280176788568497, "global_step": 143700, "epoch": 1614, "lr": 4.552694312613275e-05}
{"train_loss": 0.10564704239368439, "global_step": 143701, "epoch": 1614, "lr": 4.5526365714392036e-05}
{"train_loss": 0.10487450659275055, "global_step": 143702, "epoch": 1614, "lr": 4.5525788303252734e-05}
{"train_loss": 0.13705480098724365, "global_step": 143703, "epoch": 1614, "lr": 4.552521089271493e-05}
{"train_loss": 0.12108899652957916, "global_step": 143704, "epoch": 1614, "lr": 4.5524633482778725e-05}
{"train_loss": 0.08912984281778336, "global_step": 143705, "epoch": 1614, "lr": 4.552405607344415e-05}
{"train_loss": 0.04547083005309105, "global_step": 143706, "epoch": 1614, "lr": 4.552347866471135e-05}
{"train_loss": 0.11372889578342438, "global_step": 143707, "epoch": 1614, "lr": 4.552290125658034e-05}
{"train_loss": 0.11307529360055923, "global_step": 143708, "epoch": 1614, "lr": 4.552232384905123e-05}
{"train_loss": 0.06836868077516556, "global_step": 143709, "epoch": 1614, "lr": 4.552174644212408e-05}
{"train_loss": 0.12415198236703873, "global_step": 143710, "epoch": 1614, "lr": 4.552116903579898e-05}
{"train_loss": 0.16665013134479523, "global_step": 143711, "epoch": 1614, "lr": 4.552059163007601e-05}
{"train_loss": 0.07601844519376755, "global_step": 143712, "epoch": 1614, "lr": 4.552001422495525e-05}
{"train_loss": 0.08676256239414215, "global_step": 143713, "epoch": 1614, "lr": 4.5519436820436745e-05}
{"train_loss": 0.12743736803531647, "global_step": 143714, "epoch": 1614, "lr": 4.551885941652062e-05}
{"train_loss": 0.08266561478376389, "global_step": 143715, "epoch": 1614, "lr": 4.55182820132069e-05}
{"train_loss": 0.11032024025917053, "global_step": 143716, "epoch": 1614, "lr": 4.551770461049573e-05}
{"train_loss": 0.0904124453663826, "global_step": 143717, "epoch": 1614, "lr": 4.551712720838712e-05}
{"train_loss": 0.13435986638069153, "global_step": 143718, "epoch": 1614, "lr": 4.55165498068812e-05}
{"train_loss": 0.1406445950269699, "global_step": 143719, "epoch": 1614, "lr": 4.5515972405978005e-05}
{"train_loss": 0.08261508494615555, "global_step": 143720, "epoch": 1614, "lr": 4.551539500567765e-05}
{"train_loss": 0.09603011608123779, "global_step": 143721, "epoch": 1614, "lr": 4.551481760598018e-05}
{"train_loss": 0.13510073721408844, "global_step": 143722, "epoch": 1614, "lr": 4.551424020688568e-05}
{"train_loss": 0.13611161708831787, "global_step": 143723, "epoch": 1614, "lr": 4.551366280839426e-05}
{"train_loss": 0.11597386002540588, "global_step": 143724, "epoch": 1614, "lr": 4.551308541050595e-05}
{"train_loss": 0.09104814380407333, "global_step": 143725, "epoch": 1614, "lr": 4.551250801322088e-05}
{"train_loss": 0.07786989212036133, "global_step": 143726, "epoch": 1614, "lr": 4.551193061653908e-05}
{"train_loss": 0.0804978758096695, "global_step": 143727, "epoch": 1614, "lr": 4.551135322046066e-05}
{"train_loss": 0.09865134209394455, "global_step": 143728, "epoch": 1614, "lr": 4.5510775824985665e-05}
{"train_loss": 0.114597849547863, "global_step": 143729, "epoch": 1614, "lr": 4.55101984301142e-05}
{"train_loss": 0.13249969482421875, "global_step": 143730, "epoch": 1614, "lr": 4.550962103584634e-05}
{"train_loss": 0.08182008564472198, "global_step": 143731, "epoch": 1614, "lr": 4.550904364218216e-05}
{"train_loss": 0.08006077259778976, "global_step": 143732, "epoch": 1614, "lr": 4.5508466249121714e-05}
{"train_loss": 0.11868078261613846, "global_step": 143733, "epoch": 1614, "lr": 4.550788885666513e-05}
{"train_loss": 0.10626079435094019, "global_step": 143734, "epoch": 1614, "lr": 4.550731146481242e-05, "val_loss": 6.16564416885376}
{"train_loss": 0.07558484375476837, "global_step": 143735, "epoch": 1615, "lr": 4.550673407356373e-05}
{"train_loss": 0.09387555718421936, "global_step": 143736, "epoch": 1615, "lr": 4.5506156682919075e-05}
{"train_loss": 0.18076041340827942, "global_step": 143737, "epoch": 1615, "lr": 4.5505579292878585e-05}
{"train_loss": 0.12417413294315338, "global_step": 143738, "epoch": 1615, "lr": 4.55050019034423e-05}
{"train_loss": 0.08087591826915741, "global_step": 143739, "epoch": 1615, "lr": 4.550442451461033e-05}
{"train_loss": 0.07722850888967514, "global_step": 143740, "epoch": 1615, "lr": 4.5503847126382715e-05}
{"train_loss": 0.13993088901042938, "global_step": 143741, "epoch": 1615, "lr": 4.5503269738759555e-05}
{"train_loss": 0.12227670848369598, "global_step": 143742, "epoch": 1615, "lr": 4.550269235174094e-05}
{"train_loss": 0.12713968753814697, "global_step": 143743, "epoch": 1615, "lr": 4.550211496532692e-05}
{"train_loss": 0.0823746919631958, "global_step": 143744, "epoch": 1615, "lr": 4.55015375795176e-05}
{"train_loss": 0.07967514544725418, "global_step": 143745, "epoch": 1615, "lr": 4.550096019431304e-05}
{"train_loss": 0.1700730174779892, "global_step": 143746, "epoch": 1615, "lr": 4.550038280971331e-05}
{"train_loss": 0.11393677443265915, "global_step": 143747, "epoch": 1615, "lr": 4.54998054257185e-05}
{"train_loss": 0.08780238777399063, "global_step": 143748, "epoch": 1615, "lr": 4.54992280423287e-05}
{"train_loss": 0.08710572123527527, "global_step": 143749, "epoch": 1615, "lr": 4.549865065954397e-05}
{"train_loss": 0.10951223224401474, "global_step": 143750, "epoch": 1615, "lr": 4.54980732773644e-05}
{"train_loss": 0.10270743072032928, "global_step": 143751, "epoch": 1615, "lr": 4.549749589579003e-05}
{"train_loss": 0.1149628534913063, "global_step": 143752, "epoch": 1615, "lr": 4.5496918514821e-05}
{"train_loss": 0.06092194467782974, "global_step": 143753, "epoch": 1615, "lr": 4.549634113445732e-05}
{"train_loss": 0.1140219122171402, "global_step": 143754, "epoch": 1615, "lr": 4.549576375469913e-05}
{"train_loss": 0.11175170540809631, "global_step": 143755, "epoch": 1615, "lr": 4.549518637554646e-05}
{"train_loss": 0.07266352325677872, "global_step": 143756, "epoch": 1615, "lr": 4.5494608996999424e-05}
{"train_loss": 0.12422943115234375, "global_step": 143757, "epoch": 1615, "lr": 4.549403161905807e-05}
{"train_loss": 0.08227236568927765, "global_step": 143758, "epoch": 1615, "lr": 4.549345424172248e-05}
{"train_loss": 0.16112704575061798, "global_step": 143759, "epoch": 1615, "lr": 4.549287686499277e-05}
{"train_loss": 0.11250081658363342, "global_step": 143760, "epoch": 1615, "lr": 4.5492299488868964e-05}
{"train_loss": 0.10265610367059708, "global_step": 143761, "epoch": 1615, "lr": 4.549172211335118e-05}
{"train_loss": 0.09994592517614365, "global_step": 143762, "epoch": 1615, "lr": 4.549114473843946e-05}
{"train_loss": 0.1713252067565918, "global_step": 143763, "epoch": 1615, "lr": 4.549056736413392e-05}
{"train_loss": 0.1092941015958786, "global_step": 143764, "epoch": 1615, "lr": 4.548998999043461e-05}
{"train_loss": 0.06507542729377747, "global_step": 143765, "epoch": 1615, "lr": 4.548941261734161e-05}
{"train_loss": 0.176410511136055, "global_step": 143766, "epoch": 1615, "lr": 4.548883524485501e-05}
{"train_loss": 0.10301246494054794, "global_step": 143767, "epoch": 1615, "lr": 4.548825787297488e-05}
{"train_loss": 0.16482505202293396, "global_step": 143768, "epoch": 1615, "lr": 4.54876805017013e-05}
{"train_loss": 0.08833987265825272, "global_step": 143769, "epoch": 1615, "lr": 4.5487103131034345e-05}
{"train_loss": 0.09544234722852707, "global_step": 143770, "epoch": 1615, "lr": 4.5486525760974096e-05}
{"train_loss": 0.10309944301843643, "global_step": 143771, "epoch": 1615, "lr": 4.548594839152064e-05}
{"train_loss": 0.06972595304250717, "global_step": 143772, "epoch": 1615, "lr": 4.548537102267402e-05}
{"train_loss": 0.15362346172332764, "global_step": 143773, "epoch": 1615, "lr": 4.5484793654434355e-05}
{"train_loss": 0.06908021122217178, "global_step": 143774, "epoch": 1615, "lr": 4.548421628680168e-05}
{"train_loss": 0.0823216661810875, "global_step": 143775, "epoch": 1615, "lr": 4.5483638919776127e-05}
{"train_loss": 0.09543845057487488, "global_step": 143776, "epoch": 1615, "lr": 4.548306155335772e-05}
{"train_loss": 0.11273529380559921, "global_step": 143777, "epoch": 1615, "lr": 4.548248418754656e-05}
{"train_loss": 0.06056087091565132, "global_step": 143778, "epoch": 1615, "lr": 4.548190682234275e-05}
{"train_loss": 0.10047940164804459, "global_step": 143779, "epoch": 1615, "lr": 4.5481329457746315e-05}
{"train_loss": 0.1638564169406891, "global_step": 143780, "epoch": 1615, "lr": 4.548075209375738e-05}
{"train_loss": 0.09790690243244171, "global_step": 143781, "epoch": 1615, "lr": 4.5480174730375986e-05}
{"train_loss": 0.1390768587589264, "global_step": 143782, "epoch": 1615, "lr": 4.547959736760225e-05}
{"train_loss": 0.029805950820446014, "global_step": 143783, "epoch": 1615, "lr": 4.54790200054362e-05}
{"train_loss": 0.10835523158311844, "global_step": 143784, "epoch": 1615, "lr": 4.5478442643877966e-05}
{"train_loss": 0.08860017359256744, "global_step": 143785, "epoch": 1615, "lr": 4.547786528292759e-05}
{"train_loss": 0.08023538440465927, "global_step": 143786, "epoch": 1615, "lr": 4.547728792258516e-05}
{"train_loss": 0.13378871977329254, "global_step": 143787, "epoch": 1615, "lr": 4.5476710562850756e-05}
{"train_loss": 0.09672358632087708, "global_step": 143788, "epoch": 1615, "lr": 4.547613320372445e-05}
{"train_loss": 0.1540694534778595, "global_step": 143789, "epoch": 1615, "lr": 4.5475555845206316e-05}
{"train_loss": 0.1050049364566803, "global_step": 143790, "epoch": 1615, "lr": 4.547497848729646e-05}
{"train_loss": 0.09387572109699249, "global_step": 143791, "epoch": 1615, "lr": 4.547440112999491e-05}
{"train_loss": 0.1563110500574112, "global_step": 143792, "epoch": 1615, "lr": 4.54738237733018e-05}
{"train_loss": 0.1001989021897316, "global_step": 143793, "epoch": 1615, "lr": 4.5473246417217153e-05}
{"train_loss": 0.05116725713014603, "global_step": 143794, "epoch": 1615, "lr": 4.547266906174107e-05}
{"train_loss": 0.05581459775567055, "global_step": 143795, "epoch": 1615, "lr": 4.547209170687366e-05}
{"train_loss": 0.15699049830436707, "global_step": 143796, "epoch": 1615, "lr": 4.547151435261494e-05}
{"train_loss": 0.18754857778549194, "global_step": 143797, "epoch": 1615, "lr": 4.547093699896505e-05}
{"train_loss": 0.0948096439242363, "global_step": 143798, "epoch": 1615, "lr": 4.547035964592401e-05}
{"train_loss": 0.1545238047838211, "global_step": 143799, "epoch": 1615, "lr": 4.546978229349195e-05}
{"train_loss": 0.11021658033132553, "global_step": 143800, "epoch": 1615, "lr": 4.54692049416689e-05}
{"train_loss": 0.06957809627056122, "global_step": 143801, "epoch": 1615, "lr": 4.5468627590454986e-05}
{"train_loss": 0.08122476935386658, "global_step": 143802, "epoch": 1615, "lr": 4.5468050239850236e-05}
{"train_loss": 0.14524555206298828, "global_step": 143803, "epoch": 1615, "lr": 4.546747288985477e-05}
{"train_loss": 0.10635954141616821, "global_step": 143804, "epoch": 1615, "lr": 4.546689554046863e-05}
{"train_loss": 0.1452280879020691, "global_step": 143805, "epoch": 1615, "lr": 4.546631819169192e-05}
{"train_loss": 0.09107697755098343, "global_step": 143806, "epoch": 1615, "lr": 4.546574084352471e-05}
{"train_loss": 0.12009744346141815, "global_step": 143807, "epoch": 1615, "lr": 4.5465163495967076e-05}
{"train_loss": 0.09349768608808517, "global_step": 143808, "epoch": 1615, "lr": 4.5464586149019096e-05}
{"train_loss": 0.08178935199975967, "global_step": 143809, "epoch": 1615, "lr": 4.546400880268085e-05}
{"train_loss": 0.09111838787794113, "global_step": 143810, "epoch": 1615, "lr": 4.54634314569524e-05}
{"train_loss": 0.11949032545089722, "global_step": 143811, "epoch": 1615, "lr": 4.5462854111833844e-05}
{"train_loss": 0.04879693686962128, "global_step": 143812, "epoch": 1615, "lr": 4.5462276767325244e-05}
{"train_loss": 0.12523554265499115, "global_step": 143813, "epoch": 1615, "lr": 4.5461699423426684e-05}
{"train_loss": 0.08542892336845398, "global_step": 143814, "epoch": 1615, "lr": 4.546112208013826e-05}
{"train_loss": 0.11718439310789108, "global_step": 143815, "epoch": 1615, "lr": 4.5460544737460016e-05}
{"train_loss": 0.132367342710495, "global_step": 143816, "epoch": 1615, "lr": 4.5459967395392064e-05}
{"train_loss": 0.14667542278766632, "global_step": 143817, "epoch": 1615, "lr": 4.545939005393444e-05}
{"train_loss": 0.13299040496349335, "global_step": 143818, "epoch": 1615, "lr": 4.5458812713087276e-05}
{"train_loss": 0.11327800899744034, "global_step": 143819, "epoch": 1615, "lr": 4.545823537285059e-05}
{"train_loss": 0.054910022765398026, "global_step": 143820, "epoch": 1615, "lr": 4.545765803322452e-05}
{"train_loss": 0.09495725482702255, "global_step": 143821, "epoch": 1615, "lr": 4.5457080694209075e-05}
{"train_loss": 0.16393578052520752, "global_step": 143822, "epoch": 1615, "lr": 4.5456503355804405e-05}
{"train_loss": 0.10889664832293318, "global_step": 143823, "epoch": 1615, "lr": 4.5455926018010533e-05, "val_loss": 5.934206962585449, "train_action_mse_error": 9.147405624389648}
{"train_loss": 0.15627789497375488, "global_step": 143824, "epoch": 1616, "lr": 4.545534868082756e-05}
{"train_loss": 0.10165809839963913, "global_step": 143825, "epoch": 1616, "lr": 4.545477134425557e-05}
{"train_loss": 0.14448611438274384, "global_step": 143826, "epoch": 1616, "lr": 4.5454194008294614e-05}
{"train_loss": 0.18388792872428894, "global_step": 143827, "epoch": 1616, "lr": 4.54536166729448e-05}
{"train_loss": 0.21719799935817719, "global_step": 143828, "epoch": 1616, "lr": 4.5453039338206195e-05}
{"train_loss": 0.07540830969810486, "global_step": 143829, "epoch": 1616, "lr": 4.545246200407885e-05}
{"train_loss": 0.18058013916015625, "global_step": 143830, "epoch": 1616, "lr": 4.545188467056287e-05}
{"train_loss": 0.14706477522850037, "global_step": 143831, "epoch": 1616, "lr": 4.545130733765835e-05}
{"train_loss": 0.14583848416805267, "global_step": 143832, "epoch": 1616, "lr": 4.545073000536532e-05}
{"train_loss": 0.13272076845169067, "global_step": 143833, "epoch": 1616, "lr": 4.545015267368392e-05}
{"train_loss": 0.0906146764755249, "global_step": 143834, "epoch": 1616, "lr": 4.544957534261415e-05}
{"train_loss": 0.1001814678311348, "global_step": 143835, "epoch": 1616, "lr": 4.544899801215616e-05}
{"train_loss": 0.1328951120376587, "global_step": 143836, "epoch": 1616, "lr": 4.544842068230998e-05}
{"train_loss": 0.08577178418636322, "global_step": 143837, "epoch": 1616, "lr": 4.544784335307572e-05}
{"train_loss": 0.10640449821949005, "global_step": 143838, "epoch": 1616, "lr": 4.544726602445342e-05}
{"train_loss": 0.09453578293323517, "global_step": 143839, "epoch": 1616, "lr": 4.544668869644321e-05}
{"train_loss": 0.06125727668404579, "global_step": 143840, "epoch": 1616, "lr": 4.544611136904511e-05}
{"train_loss": 0.059219710528850555, "global_step": 143841, "epoch": 1616, "lr": 4.544553404225924e-05}
{"train_loss": 0.061170272529125214, "global_step": 143842, "epoch": 1616, "lr": 4.5444956716085654e-05}
{"train_loss": 0.11643190681934357, "global_step": 143843, "epoch": 1616, "lr": 4.544437939052444e-05}
{"train_loss": 0.08342906832695007, "global_step": 143844, "epoch": 1616, "lr": 4.544380206557568e-05}
{"train_loss": 0.11406921595335007, "global_step": 143845, "epoch": 1616, "lr": 4.544322474123944e-05}
{"train_loss": 0.13357512652873993, "global_step": 143846, "epoch": 1616, "lr": 4.544264741751581e-05}
{"train_loss": 0.09736070036888123, "global_step": 143847, "epoch": 1616, "lr": 4.544207009440485e-05}
{"train_loss": 0.12203546613454819, "global_step": 143848, "epoch": 1616, "lr": 4.544149277190665e-05}
{"train_loss": 0.16711601614952087, "global_step": 143849, "epoch": 1616, "lr": 4.544091545002127e-05}
{"train_loss": 0.08936681598424911, "global_step": 143850, "epoch": 1616, "lr": 4.5440338128748835e-05}
{"train_loss": 0.1638399213552475, "global_step": 143851, "epoch": 1616, "lr": 4.543976080808936e-05}
{"train_loss": 0.1527813971042633, "global_step": 143852, "epoch": 1616, "lr": 4.5439183488042987e-05}
{"train_loss": 0.07675264775753021, "global_step": 143853, "epoch": 1616, "lr": 4.5438606168609733e-05}
{"train_loss": 0.15466934442520142, "global_step": 143854, "epoch": 1616, "lr": 4.543802884978972e-05}
{"train_loss": 0.13857482373714447, "global_step": 143855, "epoch": 1616, "lr": 4.543745153158299e-05}
{"train_loss": 0.13741974532604218, "global_step": 143856, "epoch": 1616, "lr": 4.543687421398967e-05}
{"train_loss": 0.12256339937448502, "global_step": 143857, "epoch": 1616, "lr": 4.543629689700977e-05}
{"train_loss": 0.10027703642845154, "global_step": 143858, "epoch": 1616, "lr": 4.543571958064343e-05}
{"train_loss": 0.07490053027868271, "global_step": 143859, "epoch": 1616, "lr": 4.5435142264890685e-05}
{"train_loss": 0.06257322430610657, "global_step": 143860, "epoch": 1616, "lr": 4.5434564949751645e-05}
{"train_loss": 0.14764787256717682, "global_step": 143861, "epoch": 1616, "lr": 4.5433987635226365e-05}
{"train_loss": 0.08615448325872421, "global_step": 143862, "epoch": 1616, "lr": 4.5433410321314934e-05}
{"train_loss": 0.14927725493907928, "global_step": 143863, "epoch": 1616, "lr": 4.543283300801742e-05}
{"train_loss": 0.09431964159011841, "global_step": 143864, "epoch": 1616, "lr": 4.5432255695333906e-05}
{"train_loss": 0.06926776468753815, "global_step": 143865, "epoch": 1616, "lr": 4.543167838326447e-05}
{"train_loss": 0.12663738429546356, "global_step": 143866, "epoch": 1616, "lr": 4.54311010718092e-05}
{"train_loss": 0.13794049620628357, "global_step": 143867, "epoch": 1616, "lr": 4.543052376096814e-05}
{"train_loss": 0.12523731589317322, "global_step": 143868, "epoch": 1616, "lr": 4.5429946450741396e-05}
{"train_loss": 0.07744024693965912, "global_step": 143869, "epoch": 1616, "lr": 4.542936914112905e-05}
{"train_loss": 0.12832897901535034, "global_step": 143870, "epoch": 1616, "lr": 4.542879183213116e-05}
{"train_loss": 0.04906217381358147, "global_step": 143871, "epoch": 1616, "lr": 4.542821452374783e-05}
{"train_loss": 0.09116861969232559, "global_step": 143872, "epoch": 1616, "lr": 4.5427637215979103e-05}
{"train_loss": 0.1122068464756012, "global_step": 143873, "epoch": 1616, "lr": 4.542705990882509e-05}
{"train_loss": 0.1507207453250885, "global_step": 143874, "epoch": 1616, "lr": 4.5426482602285834e-05}
{"train_loss": 0.16045084595680237, "global_step": 143875, "epoch": 1616, "lr": 4.5425905296361455e-05}
{"train_loss": 0.131463885307312, "global_step": 143876, "epoch": 1616, "lr": 4.542532799105198e-05}
{"train_loss": 0.0788305327296257, "global_step": 143877, "epoch": 1616, "lr": 4.542475068635754e-05}
{"train_loss": 0.1499941051006317, "global_step": 143878, "epoch": 1616, "lr": 4.542417338227817e-05}
{"train_loss": 0.10909761488437653, "global_step": 143879, "epoch": 1616, "lr": 4.542359607881398e-05}
{"train_loss": 0.11850178986787796, "global_step": 143880, "epoch": 1616, "lr": 4.542301877596502e-05}
{"train_loss": 0.05635109916329384, "global_step": 143881, "epoch": 1616, "lr": 4.542244147373138e-05}
{"train_loss": 0.13298299908638, "global_step": 143882, "epoch": 1616, "lr": 4.542186417211314e-05}
{"train_loss": 0.121165931224823, "global_step": 143883, "epoch": 1616, "lr": 4.542128687111037e-05}
{"train_loss": 0.15568964183330536, "global_step": 143884, "epoch": 1616, "lr": 4.542070957072315e-05}
{"train_loss": 0.08232958614826202, "global_step": 143885, "epoch": 1616, "lr": 4.542013227095155e-05}
{"train_loss": 0.08246622234582901, "global_step": 143886, "epoch": 1616, "lr": 4.5419554971795685e-05}
{"train_loss": 0.1016906276345253, "global_step": 143887, "epoch": 1616, "lr": 4.5418977673255584e-05}
{"train_loss": 0.1293611377477646, "global_step": 143888, "epoch": 1616, "lr": 4.541840037533136e-05}
{"train_loss": 0.10000433027744293, "global_step": 143889, "epoch": 1616, "lr": 4.541782307802306e-05}
{"train_loss": 0.10065878927707672, "global_step": 143890, "epoch": 1616, "lr": 4.5417245781330796e-05}
{"train_loss": 0.16135409474372864, "global_step": 143891, "epoch": 1616, "lr": 4.541666848525461e-05}
{"train_loss": 0.12476089596748352, "global_step": 143892, "epoch": 1616, "lr": 4.5416091189794614e-05}
{"train_loss": 0.11334651708602905, "global_step": 143893, "epoch": 1616, "lr": 4.541551389495085e-05}
{"train_loss": 0.10133612155914307, "global_step": 143894, "epoch": 1616, "lr": 4.541493660072343e-05}
{"train_loss": 0.04565604031085968, "global_step": 143895, "epoch": 1616, "lr": 4.5414359307112406e-05}
{"train_loss": 0.11153163760900497, "global_step": 143896, "epoch": 1616, "lr": 4.541378201411788e-05}
{"train_loss": 0.15381789207458496, "global_step": 143897, "epoch": 1616, "lr": 4.541320472173989e-05}
{"train_loss": 0.12208128720521927, "global_step": 143898, "epoch": 1616, "lr": 4.541262742997856e-05}
{"train_loss": 0.10046982765197754, "global_step": 143899, "epoch": 1616, "lr": 4.541205013883394e-05}
{"train_loss": 0.15764601528644562, "global_step": 143900, "epoch": 1616, "lr": 4.5411472848306126e-05}
{"train_loss": 0.10336625576019287, "global_step": 143901, "epoch": 1616, "lr": 4.541089555839517e-05}
{"train_loss": 0.07630332559347153, "global_step": 143902, "epoch": 1616, "lr": 4.541031826910116e-05}
{"train_loss": 0.11478354781866074, "global_step": 143903, "epoch": 1616, "lr": 4.540974098042418e-05}
{"train_loss": 0.13886958360671997, "global_step": 143904, "epoch": 1616, "lr": 4.5409163692364295e-05}
{"train_loss": 0.18020501732826233, "global_step": 143905, "epoch": 1616, "lr": 4.540858640492161e-05}
{"train_loss": 0.16101135313510895, "global_step": 143906, "epoch": 1616, "lr": 4.5408009118096174e-05}
{"train_loss": 0.1288488805294037, "global_step": 143907, "epoch": 1616, "lr": 4.5407431831888086e-05}
{"train_loss": 0.10832387208938599, "global_step": 143908, "epoch": 1616, "lr": 4.54068545462974e-05}
{"train_loss": 0.08416981250047684, "global_step": 143909, "epoch": 1616, "lr": 4.540627726132422e-05}
{"train_loss": 0.12970256805419922, "global_step": 143910, "epoch": 1616, "lr": 4.540569997696859e-05}
{"train_loss": 0.11865150183439255, "global_step": 143911, "epoch": 1616, "lr": 4.5405122693230626e-05}
{"train_loss": 0.11587787988815415, "global_step": 143912, "epoch": 1616, "lr": 4.540454541011037e-05, "val_loss": 5.923130035400391}
{"train_loss": 0.14234116673469543, "global_step": 143913, "epoch": 1617, "lr": 4.5403968127607946e-05}
{"train_loss": 0.14735126495361328, "global_step": 143914, "epoch": 1617, "lr": 4.540339084572337e-05}
{"train_loss": 0.17715105414390564, "global_step": 143915, "epoch": 1617, "lr": 4.540281356445677e-05}
{"train_loss": 0.0948759987950325, "global_step": 143916, "epoch": 1617, "lr": 4.5402236283808197e-05}
{"train_loss": 0.12614597380161285, "global_step": 143917, "epoch": 1617, "lr": 4.540165900377775e-05}
{"train_loss": 0.11554387211799622, "global_step": 143918, "epoch": 1617, "lr": 4.540108172436548e-05}
{"train_loss": 0.11007096618413925, "global_step": 143919, "epoch": 1617, "lr": 4.5400504445571495e-05}
{"train_loss": 0.12537185847759247, "global_step": 143920, "epoch": 1617, "lr": 4.539992716739585e-05}
{"train_loss": 0.2430659383535385, "global_step": 143921, "epoch": 1617, "lr": 4.53993498898386e-05}
{"train_loss": 0.14792968332767487, "global_step": 143922, "epoch": 1617, "lr": 4.5398772612899895e-05}
{"train_loss": 0.09089615941047668, "global_step": 143923, "epoch": 1617, "lr": 4.539819533657974e-05}
{"train_loss": 0.0926036536693573, "global_step": 143924, "epoch": 1617, "lr": 4.539761806087826e-05}
{"train_loss": 0.07745176553726196, "global_step": 143925, "epoch": 1617, "lr": 4.5397040785795496e-05}
{"train_loss": 0.08149315416812897, "global_step": 143926, "epoch": 1617, "lr": 4.539646351133156e-05}
{"train_loss": 0.08212408423423767, "global_step": 143927, "epoch": 1617, "lr": 4.5395886237486495e-05}
{"train_loss": 0.09478937089443207, "global_step": 143928, "epoch": 1617, "lr": 4.539530896426043e-05}
{"train_loss": 0.08287937939167023, "global_step": 143929, "epoch": 1617, "lr": 4.539473169165337e-05}
{"train_loss": 0.10997051745653152, "global_step": 143930, "epoch": 1617, "lr": 4.5394154419665456e-05}
{"train_loss": 0.06739295274019241, "global_step": 143931, "epoch": 1617, "lr": 4.539357714829673e-05}
{"train_loss": 0.1777774542570114, "global_step": 143932, "epoch": 1617, "lr": 4.53929998775473e-05}
{"train_loss": 0.09504441171884537, "global_step": 143933, "epoch": 1617, "lr": 4.53924226074172e-05}
{"train_loss": 0.09688567370176315, "global_step": 143934, "epoch": 1617, "lr": 4.539184533790656e-05}
{"train_loss": 0.10335767269134521, "global_step": 143935, "epoch": 1617, "lr": 4.5391268069015404e-05}
{"train_loss": 0.1079992949962616, "global_step": 143936, "epoch": 1617, "lr": 4.539069080074385e-05}
{"train_loss": 0.19978170096874237, "global_step": 143937, "epoch": 1617, "lr": 4.539011353309195e-05}
{"train_loss": 0.1262388974428177, "global_step": 143938, "epoch": 1617, "lr": 4.538953626605981e-05}
{"train_loss": 0.10926438122987747, "global_step": 143939, "epoch": 1617, "lr": 4.538895899964749e-05}
{"train_loss": 0.06868717074394226, "global_step": 143940, "epoch": 1617, "lr": 4.538838173385504e-05}
{"train_loss": 0.07123272120952606, "global_step": 143941, "epoch": 1617, "lr": 4.5387804468682605e-05}
{"train_loss": 0.09806335717439651, "global_step": 143942, "epoch": 1617, "lr": 4.538722720413019e-05}
{"train_loss": 0.08765675872564316, "global_step": 143943, "epoch": 1617, "lr": 4.5386649940197936e-05}
{"train_loss": 0.11344245076179504, "global_step": 143944, "epoch": 1617, "lr": 4.538607267688586e-05}
{"train_loss": 0.10888027399778366, "global_step": 143945, "epoch": 1617, "lr": 4.5385495414194104e-05}
{"train_loss": 0.04990820959210396, "global_step": 143946, "epoch": 1617, "lr": 4.538491815212268e-05}
{"train_loss": 0.10079557448625565, "global_step": 143947, "epoch": 1617, "lr": 4.538434089067172e-05}
{"train_loss": 0.06496232748031616, "global_step": 143948, "epoch": 1617, "lr": 4.538376362984126e-05}
{"train_loss": 0.11434793472290039, "global_step": 143949, "epoch": 1617, "lr": 4.538318636963142e-05}
{"train_loss": 0.07498401403427124, "global_step": 143950, "epoch": 1617, "lr": 4.538260911004223e-05}
{"train_loss": 0.0926906168460846, "global_step": 143951, "epoch": 1617, "lr": 4.5382031851073814e-05}
{"train_loss": 0.11675269156694412, "global_step": 143952, "epoch": 1617, "lr": 4.5381454592726206e-05}
{"train_loss": 0.10203668475151062, "global_step": 143953, "epoch": 1617, "lr": 4.5380877334999526e-05}
{"train_loss": 0.17265330255031586, "global_step": 143954, "epoch": 1617, "lr": 4.538030007789381e-05}
{"train_loss": 0.14238592982292175, "global_step": 143955, "epoch": 1617, "lr": 4.537972282140918e-05}
{"train_loss": 0.09870246797800064, "global_step": 143956, "epoch": 1617, "lr": 4.537914556554567e-05}
{"train_loss": 0.10551077127456665, "global_step": 143957, "epoch": 1617, "lr": 4.537856831030338e-05}
{"train_loss": 0.1394912600517273, "global_step": 143958, "epoch": 1617, "lr": 4.53779910556824e-05}
{"train_loss": 0.15457457304000854, "global_step": 143959, "epoch": 1617, "lr": 4.5377413801682775e-05}
{"train_loss": 0.14445874094963074, "global_step": 143960, "epoch": 1617, "lr": 4.5376836548304614e-05}
{"train_loss": 0.11705859005451202, "global_step": 143961, "epoch": 1617, "lr": 4.5376259295547965e-05}
{"train_loss": 0.09949489682912827, "global_step": 143962, "epoch": 1617, "lr": 4.537568204341294e-05}
{"train_loss": 0.1499091386795044, "global_step": 143963, "epoch": 1617, "lr": 4.537510479189958e-05}
{"train_loss": 0.09824711084365845, "global_step": 143964, "epoch": 1617, "lr": 4.5374527541008005e-05}
{"train_loss": 0.10054517537355423, "global_step": 143965, "epoch": 1617, "lr": 4.5373950290738245e-05}
{"train_loss": 0.12438388913869858, "global_step": 143966, "epoch": 1617, "lr": 4.5373373041090416e-05}
{"train_loss": 0.16877597570419312, "global_step": 143967, "epoch": 1617, "lr": 4.537279579206457e-05}
{"train_loss": 0.12698081135749817, "global_step": 143968, "epoch": 1617, "lr": 4.537221854366081e-05}
{"train_loss": 0.11280592530965805, "global_step": 143969, "epoch": 1617, "lr": 4.5371641295879176e-05}
{"train_loss": 0.08683852106332779, "global_step": 143970, "epoch": 1617, "lr": 4.537106404871979e-05}
{"train_loss": 0.11927642673254013, "global_step": 143971, "epoch": 1617, "lr": 4.5370486802182685e-05}
{"train_loss": 0.0814070850610733, "global_step": 143972, "epoch": 1617, "lr": 4.536990955626799e-05}
{"train_loss": 0.19597652554512024, "global_step": 143973, "epoch": 1617, "lr": 4.5369332310975726e-05}
{"train_loss": 0.14901120960712433, "global_step": 143974, "epoch": 1617, "lr": 4.5368755066306026e-05}
{"train_loss": 0.09022895246744156, "global_step": 143975, "epoch": 1617, "lr": 4.5368177822258915e-05}
{"train_loss": 0.09467814117670059, "global_step": 143976, "epoch": 1617, "lr": 4.536760057883451e-05}
{"train_loss": 0.06714066863059998, "global_step": 143977, "epoch": 1617, "lr": 4.536702333603288e-05}
{"train_loss": 0.08236715197563171, "global_step": 143978, "epoch": 1617, "lr": 4.536644609385408e-05}
{"train_loss": 0.12760433554649353, "global_step": 143979, "epoch": 1617, "lr": 4.5365868852298224e-05}
{"train_loss": 0.20013649761676788, "global_step": 143980, "epoch": 1617, "lr": 4.536529161136535e-05}
{"train_loss": 0.07662765681743622, "global_step": 143981, "epoch": 1617, "lr": 4.536471437105558e-05}
{"train_loss": 0.15898312628269196, "global_step": 143982, "epoch": 1617, "lr": 4.5364137131368945e-05}
{"train_loss": 0.14225012063980103, "global_step": 143983, "epoch": 1617, "lr": 4.5363559892305566e-05}
{"train_loss": 0.08455076068639755, "global_step": 143984, "epoch": 1617, "lr": 4.5362982653865474e-05}
{"train_loss": 0.04149802774190903, "global_step": 143985, "epoch": 1617, "lr": 4.5362405416048806e-05}
{"train_loss": 0.08276854455471039, "global_step": 143986, "epoch": 1617, "lr": 4.536182817885557e-05}
{"train_loss": 0.1411455273628235, "global_step": 143987, "epoch": 1617, "lr": 4.536125094228591e-05}
{"train_loss": 0.1098238080739975, "global_step": 143988, "epoch": 1617, "lr": 4.5360673706339854e-05}
{"train_loss": 0.14336510002613068, "global_step": 143989, "epoch": 1617, "lr": 4.536009647101751e-05}
{"train_loss": 0.1894252896308899, "global_step": 143990, "epoch": 1617, "lr": 4.535951923631893e-05}
{"train_loss": 0.13678288459777832, "global_step": 143991, "epoch": 1617, "lr": 4.535894200224423e-05}
{"train_loss": 0.12237519770860672, "global_step": 143992, "epoch": 1617, "lr": 4.535836476879344e-05}
{"train_loss": 0.14644433557987213, "global_step": 143993, "epoch": 1617, "lr": 4.5357787535966676e-05}
{"train_loss": 0.1508316546678543, "global_step": 143994, "epoch": 1617, "lr": 4.5357210303763986e-05}
{"train_loss": 0.051971979439258575, "global_step": 143995, "epoch": 1617, "lr": 4.535663307218547e-05}
{"train_loss": 0.126698836684227, "global_step": 143996, "epoch": 1617, "lr": 4.535605584123121e-05}
{"train_loss": 0.06854946911334991, "global_step": 143997, "epoch": 1617, "lr": 4.535547861090126e-05}
{"train_loss": 0.07603904604911804, "global_step": 143998, "epoch": 1617, "lr": 4.535490138119572e-05}
{"train_loss": 0.10408589988946915, "global_step": 143999, "epoch": 1617, "lr": 4.5354324152114635e-05}
{"train_loss": 0.1391395628452301, "global_step": 144000, "epoch": 1617, "lr": 4.535374692365813e-05}
{"train_loss": 0.11509778045034141, "global_step": 144001, "epoch": 1617, "lr": 4.5353169695826236e-05, "val_loss": 5.997143745422363}
{"train_loss": 0.11430437117815018, "global_step": 144002, "epoch": 1618, "lr": 4.535259246861907e-05}
{"train_loss": 0.07000648975372314, "global_step": 144003, "epoch": 1618, "lr": 4.535201524203667e-05}
{"train_loss": 0.2004338651895523, "global_step": 144004, "epoch": 1618, "lr": 4.5351438016079164e-05}
{"train_loss": 0.087479367852211, "global_step": 144005, "epoch": 1618, "lr": 4.535086079074657e-05}
{"train_loss": 0.1841396689414978, "global_step": 144006, "epoch": 1618, "lr": 4.535028356603902e-05}
{"train_loss": 0.08233656734228134, "global_step": 144007, "epoch": 1618, "lr": 4.5349706341956546e-05}
{"train_loss": 0.20559638738632202, "global_step": 144008, "epoch": 1618, "lr": 4.534912911849927e-05}
{"train_loss": 0.06978939473628998, "global_step": 144009, "epoch": 1618, "lr": 4.534855189566722e-05}
{"train_loss": 0.1906767189502716, "global_step": 144010, "epoch": 1618, "lr": 4.534797467346053e-05}
{"train_loss": 0.11378684639930725, "global_step": 144011, "epoch": 1618, "lr": 4.534739745187922e-05}
{"train_loss": 0.08663186430931091, "global_step": 144012, "epoch": 1618, "lr": 4.5346820230923407e-05}
{"train_loss": 0.1717294305562973, "global_step": 144013, "epoch": 1618, "lr": 4.5346243010593164e-05}
{"train_loss": 0.14322222769260406, "global_step": 144014, "epoch": 1618, "lr": 4.534566579088856e-05}
{"train_loss": 0.10255195945501328, "global_step": 144015, "epoch": 1618, "lr": 4.534508857180967e-05}
{"train_loss": 0.07369889318943024, "global_step": 144016, "epoch": 1618, "lr": 4.5344511353356575e-05}
{"train_loss": 0.059080783277750015, "global_step": 144017, "epoch": 1618, "lr": 4.534393413552937e-05}
{"train_loss": 0.05425581336021423, "global_step": 144018, "epoch": 1618, "lr": 4.534335691832809e-05}
{"train_loss": 0.07606413215398788, "global_step": 144019, "epoch": 1618, "lr": 4.534277970175287e-05}
{"train_loss": 0.07719332724809647, "global_step": 144020, "epoch": 1618, "lr": 4.5342202485803726e-05}
{"train_loss": 0.12044251710176468, "global_step": 144021, "epoch": 1618, "lr": 4.534162527048079e-05}
{"train_loss": 0.109103262424469, "global_step": 144022, "epoch": 1618, "lr": 4.5341048055784095e-05}
{"train_loss": 0.1043870821595192, "global_step": 144023, "epoch": 1618, "lr": 4.534047084171376e-05}
{"train_loss": 0.1001584380865097, "global_step": 144024, "epoch": 1618, "lr": 4.533989362826982e-05}
{"train_loss": 0.10824403166770935, "global_step": 144025, "epoch": 1618, "lr": 4.533931641545239e-05}
{"train_loss": 0.14610601961612701, "global_step": 144026, "epoch": 1618, "lr": 4.533873920326152e-05}
{"train_loss": 0.13776305317878723, "global_step": 144027, "epoch": 1618, "lr": 4.533816199169732e-05}
{"train_loss": 0.05989979952573776, "global_step": 144028, "epoch": 1618, "lr": 4.533758478075982e-05}
{"train_loss": 0.13356846570968628, "global_step": 144029, "epoch": 1618, "lr": 4.5337007570449154e-05}
{"train_loss": 0.09512881934642792, "global_step": 144030, "epoch": 1618, "lr": 4.533643036076535e-05}
{"train_loss": 0.10694737732410431, "global_step": 144031, "epoch": 1618, "lr": 4.5335853151708506e-05}
{"train_loss": 0.062152646481990814, "global_step": 144032, "epoch": 1618, "lr": 4.5335275943278703e-05}
{"train_loss": 0.15157191455364227, "global_step": 144033, "epoch": 1618, "lr": 4.5334698735476024e-05}
{"train_loss": 0.1746302992105484, "global_step": 144034, "epoch": 1618, "lr": 4.533412152830053e-05}
{"train_loss": 0.17753803730010986, "global_step": 144035, "epoch": 1618, "lr": 4.5333544321752305e-05}
{"train_loss": 0.0804767832159996, "global_step": 144036, "epoch": 1618, "lr": 4.533296711583144e-05}
{"train_loss": 0.19762852787971497, "global_step": 144037, "epoch": 1618, "lr": 4.5332389910537985e-05}
{"train_loss": 0.105021633207798, "global_step": 144038, "epoch": 1618, "lr": 4.533181270587205e-05}
{"train_loss": 0.10586259514093399, "global_step": 144039, "epoch": 1618, "lr": 4.533123550183368e-05}
{"train_loss": 0.1200290396809578, "global_step": 144040, "epoch": 1618, "lr": 4.533065829842298e-05}
{"train_loss": 0.18042439222335815, "global_step": 144041, "epoch": 1618, "lr": 4.5330081095640004e-05}
{"train_loss": 0.09016173332929611, "global_step": 144042, "epoch": 1618, "lr": 4.532950389348486e-05}
{"train_loss": 0.08124833554029465, "global_step": 144043, "epoch": 1618, "lr": 4.532892669195758e-05}
{"train_loss": 0.14088699221611023, "global_step": 144044, "epoch": 1618, "lr": 4.5328349491058296e-05}
{"train_loss": 0.06824953109025955, "global_step": 144045, "epoch": 1618, "lr": 4.532777229078704e-05}
{"train_loss": 0.11226846277713776, "global_step": 144046, "epoch": 1618, "lr": 4.5327195091143925e-05}
{"train_loss": 0.13137246668338776, "global_step": 144047, "epoch": 1618, "lr": 4.532661789212899e-05}
{"train_loss": 0.10494551062583923, "global_step": 144048, "epoch": 1618, "lr": 4.532604069374234e-05}
{"train_loss": 0.15881872177124023, "global_step": 144049, "epoch": 1618, "lr": 4.532546349598406e-05}
{"train_loss": 0.0747215524315834, "global_step": 144050, "epoch": 1618, "lr": 4.532488629885421e-05}
{"train_loss": 0.08521591871976852, "global_step": 144051, "epoch": 1618, "lr": 4.5324309102352866e-05}
{"train_loss": 0.151382714509964, "global_step": 144052, "epoch": 1618, "lr": 4.532373190648012e-05}
{"train_loss": 0.0716681182384491, "global_step": 144053, "epoch": 1618, "lr": 4.532315471123603e-05}
{"train_loss": 0.09877178072929382, "global_step": 144054, "epoch": 1618, "lr": 4.53225775166207e-05}
{"train_loss": 0.10567121207714081, "global_step": 144055, "epoch": 1618, "lr": 4.532200032263419e-05}
{"train_loss": 0.0670330747961998, "global_step": 144056, "epoch": 1618, "lr": 4.5321423129276566e-05}
{"train_loss": 0.08306080847978592, "global_step": 144057, "epoch": 1618, "lr": 4.5320845936547936e-05}
{"train_loss": 0.08645062893629074, "global_step": 144058, "epoch": 1618, "lr": 4.532026874444834e-05}
{"train_loss": 0.10228822380304337, "global_step": 144059, "epoch": 1618, "lr": 4.53196915529779e-05}
{"train_loss": 0.10500681400299072, "global_step": 144060, "epoch": 1618, "lr": 4.531911436213665e-05}
{"train_loss": 0.08986796438694, "global_step": 144061, "epoch": 1618, "lr": 4.531853717192471e-05}
{"train_loss": 0.11756637692451477, "global_step": 144062, "epoch": 1618, "lr": 4.5317959982342114e-05}
{"train_loss": 0.15287761390209198, "global_step": 144063, "epoch": 1618, "lr": 4.531738279338898e-05}
{"train_loss": 0.12052001804113388, "global_step": 144064, "epoch": 1618, "lr": 4.531680560506535e-05}
{"train_loss": 0.06581958383321762, "global_step": 144065, "epoch": 1618, "lr": 4.531622841737133e-05}
{"train_loss": 0.07433106005191803, "global_step": 144066, "epoch": 1618, "lr": 4.5315651230306974e-05}
{"train_loss": 0.17169341444969177, "global_step": 144067, "epoch": 1618, "lr": 4.531507404387237e-05}
{"train_loss": 0.13195165991783142, "global_step": 144068, "epoch": 1618, "lr": 4.5314496858067615e-05}
{"train_loss": 0.08753690123558044, "global_step": 144069, "epoch": 1618, "lr": 4.5313919672892764e-05}
{"train_loss": 0.12277808040380478, "global_step": 144070, "epoch": 1618, "lr": 4.5313342488347895e-05}
{"train_loss": 0.09957986325025558, "global_step": 144071, "epoch": 1618, "lr": 4.531276530443309e-05}
{"train_loss": 0.12592802941799164, "global_step": 144072, "epoch": 1618, "lr": 4.531218812114842e-05}
{"train_loss": 0.07205141335725784, "global_step": 144073, "epoch": 1618, "lr": 4.531161093849399e-05}
{"train_loss": 0.11575876176357269, "global_step": 144074, "epoch": 1618, "lr": 4.531103375646985e-05}
{"train_loss": 0.1266327053308487, "global_step": 144075, "epoch": 1618, "lr": 4.5310456575076064e-05}
{"train_loss": 0.1357993483543396, "global_step": 144076, "epoch": 1618, "lr": 4.530987939431276e-05}
{"train_loss": 0.07666876912117004, "global_step": 144077, "epoch": 1618, "lr": 4.5309302214179954e-05}
{"train_loss": 0.1093481183052063, "global_step": 144078, "epoch": 1618, "lr": 4.530872503467778e-05}
{"train_loss": 0.14582976698875427, "global_step": 144079, "epoch": 1618, "lr": 4.530814785580627e-05}
{"train_loss": 0.10598815232515335, "global_step": 144080, "epoch": 1618, "lr": 4.5307570677565545e-05}
{"train_loss": 0.156239852309227, "global_step": 144081, "epoch": 1618, "lr": 4.530699349995564e-05}
{"train_loss": 0.1268157660961151, "global_step": 144082, "epoch": 1618, "lr": 4.530641632297667e-05}
{"train_loss": 0.09858772903680801, "global_step": 144083, "epoch": 1618, "lr": 4.530583914662868e-05}
{"train_loss": 0.15512728691101074, "global_step": 144084, "epoch": 1618, "lr": 4.5305261970911754e-05}
{"train_loss": 0.062107086181640625, "global_step": 144085, "epoch": 1618, "lr": 4.5304684795826e-05}
{"train_loss": 0.08908914774656296, "global_step": 144086, "epoch": 1618, "lr": 4.530410762137146e-05}
{"train_loss": 0.10609336197376251, "global_step": 144087, "epoch": 1618, "lr": 4.530353044754824e-05}
{"train_loss": 0.12152495980262756, "global_step": 144088, "epoch": 1618, "lr": 4.53029532743564e-05}
{"train_loss": 0.11251258105039597, "global_step": 144089, "epoch": 1618, "lr": 4.5302376101796016e-05}
{"train_loss": 0.11248674632960491, "global_step": 144090, "epoch": 1618, "lr": 4.530179892986717e-05, "val_loss": 6.040406703948975}
{"train_loss": 0.12337812781333923, "global_step": 144091, "epoch": 1619, "lr": 4.530122175856994e-05}
{"train_loss": 0.11274327337741852, "global_step": 144092, "epoch": 1619, "lr": 4.5300644587904405e-05}
{"train_loss": 0.12789450585842133, "global_step": 144093, "epoch": 1619, "lr": 4.530006741787065e-05}
{"train_loss": 0.10132288932800293, "global_step": 144094, "epoch": 1619, "lr": 4.529949024846872e-05}
{"train_loss": 0.1234278753399849, "global_step": 144095, "epoch": 1619, "lr": 4.529891307969874e-05}
{"train_loss": 0.062414027750492096, "global_step": 144096, "epoch": 1619, "lr": 4.5298335911560746e-05}
{"train_loss": 0.09132876992225647, "global_step": 144097, "epoch": 1619, "lr": 4.529775874405485e-05}
{"train_loss": 0.13255412876605988, "global_step": 144098, "epoch": 1619, "lr": 4.52971815771811e-05}
{"train_loss": 0.10309962183237076, "global_step": 144099, "epoch": 1619, "lr": 4.52966044109396e-05}
{"train_loss": 0.096592977643013, "global_step": 144100, "epoch": 1619, "lr": 4.529602724533039e-05}
{"train_loss": 0.09952320903539658, "global_step": 144101, "epoch": 1619, "lr": 4.52954500803536e-05}
{"train_loss": 0.10635609924793243, "global_step": 144102, "epoch": 1619, "lr": 4.5294872916009255e-05}
{"train_loss": 0.0795745924115181, "global_step": 144103, "epoch": 1619, "lr": 4.529429575229745e-05}
{"train_loss": 0.09552763402462006, "global_step": 144104, "epoch": 1619, "lr": 4.52937185892183e-05}
{"train_loss": 0.09048232436180115, "global_step": 144105, "epoch": 1619, "lr": 4.529314142677183e-05}
{"train_loss": 0.10660786926746368, "global_step": 144106, "epoch": 1619, "lr": 4.529256426495816e-05}
{"train_loss": 0.1259351521730423, "global_step": 144107, "epoch": 1619, "lr": 4.529198710377734e-05}
{"train_loss": 0.09859014302492142, "global_step": 144108, "epoch": 1619, "lr": 4.529140994322946e-05}
{"train_loss": 0.07002738118171692, "global_step": 144109, "epoch": 1619, "lr": 4.529083278331458e-05}
{"train_loss": 0.18274465203285217, "global_step": 144110, "epoch": 1619, "lr": 4.5290255624032794e-05}
{"train_loss": 0.12042398005723953, "global_step": 144111, "epoch": 1619, "lr": 4.5289678465384175e-05}
{"train_loss": 0.07283825427293777, "global_step": 144112, "epoch": 1619, "lr": 4.528910130736881e-05}
{"train_loss": 0.09284093976020813, "global_step": 144113, "epoch": 1619, "lr": 4.528852414998676e-05}
{"train_loss": 0.20459435880184174, "global_step": 144114, "epoch": 1619, "lr": 4.5287946993238115e-05}
{"train_loss": 0.12336818128824234, "global_step": 144115, "epoch": 1619, "lr": 4.528736983712293e-05}
{"train_loss": 0.14827954769134521, "global_step": 144116, "epoch": 1619, "lr": 4.5286792681641334e-05}
{"train_loss": 0.07563044130802155, "global_step": 144117, "epoch": 1619, "lr": 4.528621552679334e-05}
{"train_loss": 0.10321847349405289, "global_step": 144118, "epoch": 1619, "lr": 4.528563837257907e-05}
{"train_loss": 0.1246805489063263, "global_step": 144119, "epoch": 1619, "lr": 4.5285061218998584e-05}
{"train_loss": 0.10965514928102493, "global_step": 144120, "epoch": 1619, "lr": 4.528448406605198e-05}
{"train_loss": 0.13872352242469788, "global_step": 144121, "epoch": 1619, "lr": 4.528390691373929e-05}
{"train_loss": 0.11226367205381393, "global_step": 144122, "epoch": 1619, "lr": 4.5283329762060635e-05}
{"train_loss": 0.13547120988368988, "global_step": 144123, "epoch": 1619, "lr": 4.5282752611016095e-05}
{"train_loss": 0.0697193592786789, "global_step": 144124, "epoch": 1619, "lr": 4.5282175460605705e-05}
{"train_loss": 0.09354619681835175, "global_step": 144125, "epoch": 1619, "lr": 4.5281598310829594e-05}
{"train_loss": 0.17086105048656464, "global_step": 144126, "epoch": 1619, "lr": 4.5281021161687795e-05}
{"train_loss": 0.06780123710632324, "global_step": 144127, "epoch": 1619, "lr": 4.5280444013180424e-05}
{"train_loss": 0.10955478996038437, "global_step": 144128, "epoch": 1619, "lr": 4.527986686530753e-05}
{"train_loss": 0.06393716484308243, "global_step": 144129, "epoch": 1619, "lr": 4.5279289718069195e-05}
{"train_loss": 0.11893366277217865, "global_step": 144130, "epoch": 1619, "lr": 4.5278712571465506e-05}
{"train_loss": 0.16064049303531647, "global_step": 144131, "epoch": 1619, "lr": 4.527813542549654e-05}
{"train_loss": 0.10143834352493286, "global_step": 144132, "epoch": 1619, "lr": 4.527755828016237e-05}
{"train_loss": 0.051947690546512604, "global_step": 144133, "epoch": 1619, "lr": 4.527698113546308e-05}
{"train_loss": 0.04975094273686409, "global_step": 144134, "epoch": 1619, "lr": 4.527640399139872e-05}
{"train_loss": 0.1250431090593338, "global_step": 144135, "epoch": 1619, "lr": 4.527582684796942e-05}
{"train_loss": 0.1477595865726471, "global_step": 144136, "epoch": 1619, "lr": 4.527524970517521e-05}
{"train_loss": 0.08208977431058884, "global_step": 144137, "epoch": 1619, "lr": 4.527467256301619e-05}
{"train_loss": 0.06672501564025879, "global_step": 144138, "epoch": 1619, "lr": 4.5274095421492424e-05}
{"train_loss": 0.07935333997011185, "global_step": 144139, "epoch": 1619, "lr": 4.527351828060399e-05}
{"train_loss": 0.09770401567220688, "global_step": 144140, "epoch": 1619, "lr": 4.5272941140351e-05}
{"train_loss": 0.09001894295215607, "global_step": 144141, "epoch": 1619, "lr": 4.527236400073349e-05}
{"train_loss": 0.07305677980184555, "global_step": 144142, "epoch": 1619, "lr": 4.5271786861751565e-05}
{"train_loss": 0.10434744507074356, "global_step": 144143, "epoch": 1619, "lr": 4.527120972340527e-05}
{"train_loss": 0.05853421241044998, "global_step": 144144, "epoch": 1619, "lr": 4.527063258569473e-05}
{"train_loss": 0.1160133108496666, "global_step": 144145, "epoch": 1619, "lr": 4.527005544861997e-05}
{"train_loss": 0.11765258759260178, "global_step": 144146, "epoch": 1619, "lr": 4.526947831218111e-05}
{"train_loss": 0.07501563429832458, "global_step": 144147, "epoch": 1619, "lr": 4.526890117637821e-05}
{"train_loss": 0.12491948157548904, "global_step": 144148, "epoch": 1619, "lr": 4.5268324041211345e-05}
{"train_loss": 0.14758406579494476, "global_step": 144149, "epoch": 1619, "lr": 4.52677469066806e-05}
{"train_loss": 0.10223277658224106, "global_step": 144150, "epoch": 1619, "lr": 4.526716977278604e-05}
{"train_loss": 0.08616697043180466, "global_step": 144151, "epoch": 1619, "lr": 4.5266592639527754e-05}
{"train_loss": 0.060134850442409515, "global_step": 144152, "epoch": 1619, "lr": 4.5266015506905835e-05}
{"train_loss": 0.09197632223367691, "global_step": 144153, "epoch": 1619, "lr": 4.526543837492031e-05}
{"train_loss": 0.1625901609659195, "global_step": 144154, "epoch": 1619, "lr": 4.5264861243571325e-05}
{"train_loss": 0.056832216680049896, "global_step": 144155, "epoch": 1619, "lr": 4.526428411285888e-05}
{"train_loss": 0.13490763306617737, "global_step": 144156, "epoch": 1619, "lr": 4.526370698278313e-05}
{"train_loss": 0.08884864300489426, "global_step": 144157, "epoch": 1619, "lr": 4.526312985334409e-05}
{"train_loss": 0.13868315517902374, "global_step": 144158, "epoch": 1619, "lr": 4.526255272454187e-05}
{"train_loss": 0.10511302947998047, "global_step": 144159, "epoch": 1619, "lr": 4.526197559637656e-05}
{"train_loss": 0.05232786387205124, "global_step": 144160, "epoch": 1619, "lr": 4.52613984688482e-05}
{"train_loss": 0.1396474689245224, "global_step": 144161, "epoch": 1619, "lr": 4.526082134195691e-05}
{"train_loss": 0.14155711233615875, "global_step": 144162, "epoch": 1619, "lr": 4.526024421570272e-05}
{"train_loss": 0.05577067658305168, "global_step": 144163, "epoch": 1619, "lr": 4.5259667090085755e-05}
{"train_loss": 0.12456433475017548, "global_step": 144164, "epoch": 1619, "lr": 4.525908996510605e-05}
{"train_loss": 0.07883622497320175, "global_step": 144165, "epoch": 1619, "lr": 4.5258512840763724e-05}
{"train_loss": 0.12138956040143967, "global_step": 144166, "epoch": 1619, "lr": 4.525793571705882e-05}
{"train_loss": 0.15809841454029083, "global_step": 144167, "epoch": 1619, "lr": 4.5257358593991426e-05}
{"train_loss": 0.1400129646062851, "global_step": 144168, "epoch": 1619, "lr": 4.5256781471561624e-05}
{"train_loss": 0.12864813208580017, "global_step": 144169, "epoch": 1619, "lr": 4.5256204349769496e-05}
{"train_loss": 0.09745201468467712, "global_step": 144170, "epoch": 1619, "lr": 4.52556272286151e-05}
{"train_loss": 0.14185024797916412, "global_step": 144171, "epoch": 1619, "lr": 4.525505010809855e-05}
{"train_loss": 0.10403013974428177, "global_step": 144172, "epoch": 1619, "lr": 4.5254472988219884e-05}
{"train_loss": 0.13520990312099457, "global_step": 144173, "epoch": 1619, "lr": 4.525389586897921e-05}
{"train_loss": 0.07824838161468506, "global_step": 144174, "epoch": 1619, "lr": 4.5253318750376566e-05}
{"train_loss": 0.08400439471006393, "global_step": 144175, "epoch": 1619, "lr": 4.525274163241206e-05}
{"train_loss": 0.099347785115242, "global_step": 144176, "epoch": 1619, "lr": 4.5252164515085785e-05}
{"train_loss": 0.12032472342252731, "global_step": 144177, "epoch": 1619, "lr": 4.5251587398397776e-05}
{"train_loss": 0.06569550186395645, "global_step": 144178, "epoch": 1619, "lr": 4.525101028234816e-05}
{"train_loss": 0.10627182087536609, "global_step": 144179, "epoch": 1619, "lr": 4.525043316693696e-05, "val_loss": 6.156589031219482}
{"train_loss": 0.0748019814491272, "global_step": 144180, "epoch": 1620, "lr": 4.524985605216431e-05}
{"train_loss": 0.0932430550456047, "global_step": 144181, "epoch": 1620, "lr": 4.524927893803023e-05}
{"train_loss": 0.10605738312005997, "global_step": 144182, "epoch": 1620, "lr": 4.524870182453485e-05}
{"train_loss": 0.05955316871404648, "global_step": 144183, "epoch": 1620, "lr": 4.524812471167821e-05}
{"train_loss": 0.08094003051519394, "global_step": 144184, "epoch": 1620, "lr": 4.5247547599460415e-05}
{"train_loss": 0.059700630605220795, "global_step": 144185, "epoch": 1620, "lr": 4.5246970487881525e-05}
{"train_loss": 0.09717396646738052, "global_step": 144186, "epoch": 1620, "lr": 4.524639337694162e-05}
{"train_loss": 0.10800907015800476, "global_step": 144187, "epoch": 1620, "lr": 4.524581626664077e-05}
{"train_loss": 0.1007314920425415, "global_step": 144188, "epoch": 1620, "lr": 4.5245239156979076e-05}
{"train_loss": 0.12212271988391876, "global_step": 144189, "epoch": 1620, "lr": 4.524466204795659e-05}
{"train_loss": 0.05041905492544174, "global_step": 144190, "epoch": 1620, "lr": 4.5244084939573423e-05}
{"train_loss": 0.10330189764499664, "global_step": 144191, "epoch": 1620, "lr": 4.52435078318296e-05}
{"train_loss": 0.08293690532445908, "global_step": 144192, "epoch": 1620, "lr": 4.5242930724725254e-05}
{"train_loss": 0.1981976330280304, "global_step": 144193, "epoch": 1620, "lr": 4.524235361826042e-05}
{"train_loss": 0.1349889189004898, "global_step": 144194, "epoch": 1620, "lr": 4.5241776512435186e-05}
{"train_loss": 0.10214205086231232, "global_step": 144195, "epoch": 1620, "lr": 4.524119940724967e-05}
{"train_loss": 0.06183694675564766, "global_step": 144196, "epoch": 1620, "lr": 4.524062230270388e-05}
{"train_loss": 0.09780386090278625, "global_step": 144197, "epoch": 1620, "lr": 4.5240045198797966e-05}
{"train_loss": 0.08359755575656891, "global_step": 144198, "epoch": 1620, "lr": 4.523946809553194e-05}
{"train_loss": 0.16586382687091827, "global_step": 144199, "epoch": 1620, "lr": 4.523889099290594e-05}
{"train_loss": 0.07001922279596329, "global_step": 144200, "epoch": 1620, "lr": 4.523831389091998e-05}
{"train_loss": 0.1565488874912262, "global_step": 144201, "epoch": 1620, "lr": 4.52377367895742e-05}
{"train_loss": 0.09904474020004272, "global_step": 144202, "epoch": 1620, "lr": 4.523715968886862e-05}
{"train_loss": 0.06605552136898041, "global_step": 144203, "epoch": 1620, "lr": 4.523658258880337e-05}
{"train_loss": 0.12590628862380981, "global_step": 144204, "epoch": 1620, "lr": 4.5236005489378495e-05}
{"train_loss": 0.14202725887298584, "global_step": 144205, "epoch": 1620, "lr": 4.523542839059408e-05}
{"train_loss": 0.09504538774490356, "global_step": 144206, "epoch": 1620, "lr": 4.52348512924502e-05}
{"train_loss": 0.13042153418064117, "global_step": 144207, "epoch": 1620, "lr": 4.523427419494695e-05}
{"train_loss": 0.12638182938098907, "global_step": 144208, "epoch": 1620, "lr": 4.5233697098084374e-05}
{"train_loss": 0.09160545468330383, "global_step": 144209, "epoch": 1620, "lr": 4.523312000186259e-05}
{"train_loss": 0.131103977560997, "global_step": 144210, "epoch": 1620, "lr": 4.523254290628164e-05}
{"train_loss": 0.18036818504333496, "global_step": 144211, "epoch": 1620, "lr": 4.52319658113416e-05}
{"train_loss": 0.10476823151111603, "global_step": 144212, "epoch": 1620, "lr": 4.52313887170426e-05}
{"train_loss": 0.10495971888303757, "global_step": 144213, "epoch": 1620, "lr": 4.5230811623384656e-05}
{"train_loss": 0.1268724799156189, "global_step": 144214, "epoch": 1620, "lr": 4.523023453036789e-05}
{"train_loss": 0.059553273022174835, "global_step": 144215, "epoch": 1620, "lr": 4.522965743799235e-05}
{"train_loss": 0.07972753793001175, "global_step": 144216, "epoch": 1620, "lr": 4.5229080346258127e-05}
{"train_loss": 0.09031105041503906, "global_step": 144217, "epoch": 1620, "lr": 4.522850325516529e-05}
{"train_loss": 0.0733410194516182, "global_step": 144218, "epoch": 1620, "lr": 4.522792616471394e-05}
{"train_loss": 0.1344434916973114, "global_step": 144219, "epoch": 1620, "lr": 4.522734907490411e-05}
{"train_loss": 0.09817154705524445, "global_step": 144220, "epoch": 1620, "lr": 4.5226771985735933e-05}
{"train_loss": 0.14658954739570618, "global_step": 144221, "epoch": 1620, "lr": 4.522619489720943e-05}
{"train_loss": 0.11868172138929367, "global_step": 144222, "epoch": 1620, "lr": 4.522561780932474e-05}
{"train_loss": 0.08918360620737076, "global_step": 144223, "epoch": 1620, "lr": 4.522504072208189e-05}
{"train_loss": 0.10932481288909912, "global_step": 144224, "epoch": 1620, "lr": 4.522446363548097e-05}
{"train_loss": 0.07860665023326874, "global_step": 144225, "epoch": 1620, "lr": 4.522388654952207e-05}
{"train_loss": 0.12102621048688889, "global_step": 144226, "epoch": 1620, "lr": 4.5223309464205256e-05}
{"train_loss": 0.09474670886993408, "global_step": 144227, "epoch": 1620, "lr": 4.522273237953061e-05}
{"train_loss": 0.06972578912973404, "global_step": 144228, "epoch": 1620, "lr": 4.522215529549822e-05}
{"train_loss": 0.10525330156087875, "global_step": 144229, "epoch": 1620, "lr": 4.522157821210814e-05}
{"train_loss": 0.11862898617982864, "global_step": 144230, "epoch": 1620, "lr": 4.522100112936045e-05}
{"train_loss": 0.10701555013656616, "global_step": 144231, "epoch": 1620, "lr": 4.522042404725527e-05}
{"train_loss": 0.13992804288864136, "global_step": 144232, "epoch": 1620, "lr": 4.5219846965792614e-05}
{"train_loss": 0.11277543753385544, "global_step": 144233, "epoch": 1620, "lr": 4.5219269884972615e-05}
{"train_loss": 0.11659059673547745, "global_step": 144234, "epoch": 1620, "lr": 4.52186928047953e-05}
{"train_loss": 0.0856245681643486, "global_step": 144235, "epoch": 1620, "lr": 4.521811572526081e-05}
{"train_loss": 0.1063188910484314, "global_step": 144236, "epoch": 1620, "lr": 4.521753864636916e-05}
{"train_loss": 0.11025022715330124, "global_step": 144237, "epoch": 1620, "lr": 4.5216961568120465e-05}
{"train_loss": 0.13902606070041656, "global_step": 144238, "epoch": 1620, "lr": 4.521638449051478e-05}
{"train_loss": 0.1477888971567154, "global_step": 144239, "epoch": 1620, "lr": 4.521580741355221e-05}
{"train_loss": 0.11283846944570541, "global_step": 144240, "epoch": 1620, "lr": 4.5215230337232794e-05}
{"train_loss": 0.1059751883149147, "global_step": 144241, "epoch": 1620, "lr": 4.521465326155666e-05}
{"train_loss": 0.11206251382827759, "global_step": 144242, "epoch": 1620, "lr": 4.521407618652383e-05}
{"train_loss": 0.14501696825027466, "global_step": 144243, "epoch": 1620, "lr": 4.521349911213443e-05}
{"train_loss": 0.1360880732536316, "global_step": 144244, "epoch": 1620, "lr": 4.521292203838851e-05}
{"train_loss": 0.0876806229352951, "global_step": 144245, "epoch": 1620, "lr": 4.5212344965286154e-05}
{"train_loss": 0.10939603298902512, "global_step": 144246, "epoch": 1620, "lr": 4.521176789282744e-05}
{"train_loss": 0.08854611217975616, "global_step": 144247, "epoch": 1620, "lr": 4.5211190821012445e-05}
{"train_loss": 0.08228401839733124, "global_step": 144248, "epoch": 1620, "lr": 4.521061374984125e-05}
{"train_loss": 0.19058938324451447, "global_step": 144249, "epoch": 1620, "lr": 4.5210036679313906e-05}
{"train_loss": 0.07969622313976288, "global_step": 144250, "epoch": 1620, "lr": 4.520945960943055e-05}
{"train_loss": 0.16433943808078766, "global_step": 144251, "epoch": 1620, "lr": 4.5208882540191194e-05}
{"train_loss": 0.06576027721166611, "global_step": 144252, "epoch": 1620, "lr": 4.520830547159597e-05}
{"train_loss": 0.13612587749958038, "global_step": 144253, "epoch": 1620, "lr": 4.520772840364491e-05}
{"train_loss": 0.1334230899810791, "global_step": 144254, "epoch": 1620, "lr": 4.520715133633813e-05}
{"train_loss": 0.10313741862773895, "global_step": 144255, "epoch": 1620, "lr": 4.520657426967567e-05}
{"train_loss": 0.1774037480354309, "global_step": 144256, "epoch": 1620, "lr": 4.520599720365766e-05}
{"train_loss": 0.13788136839866638, "global_step": 144257, "epoch": 1620, "lr": 4.520542013828411e-05}
{"train_loss": 0.12263301759958267, "global_step": 144258, "epoch": 1620, "lr": 4.520484307355516e-05}
{"train_loss": 0.08691640198230743, "global_step": 144259, "epoch": 1620, "lr": 4.5204266009470835e-05}
{"train_loss": 0.08392934501171112, "global_step": 144260, "epoch": 1620, "lr": 4.520368894603126e-05}
{"train_loss": 0.05652415007352829, "global_step": 144261, "epoch": 1620, "lr": 4.5203111883236475e-05}
{"train_loss": 0.11304263025522232, "global_step": 144262, "epoch": 1620, "lr": 4.520253482108659e-05}
{"train_loss": 0.16001009941101074, "global_step": 144263, "epoch": 1620, "lr": 4.520195775958166e-05}
{"train_loss": 0.11239151656627655, "global_step": 144264, "epoch": 1620, "lr": 4.520138069872176e-05}
{"train_loss": 0.1255597621202469, "global_step": 144265, "epoch": 1620, "lr": 4.5200803638506986e-05}
{"train_loss": 0.06767205148935318, "global_step": 144266, "epoch": 1620, "lr": 4.520022657893739e-05}
{"train_loss": 0.12663660943508148, "global_step": 144267, "epoch": 1620, "lr": 4.519964952001308e-05}
{"train_loss": 0.109086732725414, "global_step": 144268, "epoch": 1620, "lr": 4.51990724617341e-05, "val_loss": 6.3046064376831055, "train_action_mse_error": 5.757967948913574}
{"train_loss": 0.07618775963783264, "global_step": 144269, "epoch": 1621, "lr": 4.519849540410057e-05}
{"train_loss": 0.1244083046913147, "global_step": 144270, "epoch": 1621, "lr": 4.519791834711253e-05}
{"train_loss": 0.11487569659948349, "global_step": 144271, "epoch": 1621, "lr": 4.519734129077009e-05}
{"train_loss": 0.08242547512054443, "global_step": 144272, "epoch": 1621, "lr": 4.5196764235073275e-05}
{"train_loss": 0.08304066956043243, "global_step": 144273, "epoch": 1621, "lr": 4.519618718002223e-05}
{"train_loss": 0.08651944994926453, "global_step": 144274, "epoch": 1621, "lr": 4.519561012561697e-05}
{"train_loss": 0.12091449648141861, "global_step": 144275, "epoch": 1621, "lr": 4.519503307185763e-05}
{"train_loss": 0.1024337187409401, "global_step": 144276, "epoch": 1621, "lr": 4.519445601874423e-05}
{"train_loss": 0.12662316858768463, "global_step": 144277, "epoch": 1621, "lr": 4.5193878966276905e-05}
{"train_loss": 0.0930492952466011, "global_step": 144278, "epoch": 1621, "lr": 4.5193301914455675e-05}
{"train_loss": 0.0925719365477562, "global_step": 144279, "epoch": 1621, "lr": 4.5192724863280676e-05}
{"train_loss": 0.10457222163677216, "global_step": 144280, "epoch": 1621, "lr": 4.519214781275193e-05}
{"train_loss": 0.11270521581172943, "global_step": 144281, "epoch": 1621, "lr": 4.519157076286957e-05}
{"train_loss": 0.14705178141593933, "global_step": 144282, "epoch": 1621, "lr": 4.519099371363363e-05}
{"train_loss": 0.14706523716449738, "global_step": 144283, "epoch": 1621, "lr": 4.51904166650442e-05}
{"train_loss": 0.12808865308761597, "global_step": 144284, "epoch": 1621, "lr": 4.518983961710137e-05}
{"train_loss": 0.13768385350704193, "global_step": 144285, "epoch": 1621, "lr": 4.5189262569805176e-05}
{"train_loss": 0.1204201877117157, "global_step": 144286, "epoch": 1621, "lr": 4.518868552315577e-05}
{"train_loss": 0.15978606045246124, "global_step": 144287, "epoch": 1621, "lr": 4.518810847715315e-05}
{"train_loss": 0.06619743257761002, "global_step": 144288, "epoch": 1621, "lr": 4.518753143179746e-05}
{"train_loss": 0.08075936138629913, "global_step": 144289, "epoch": 1621, "lr": 4.5186954387088724e-05}
{"train_loss": 0.06648870557546616, "global_step": 144290, "epoch": 1621, "lr": 4.518637734302706e-05}
{"train_loss": 0.05476547032594681, "global_step": 144291, "epoch": 1621, "lr": 4.518580029961252e-05}
{"train_loss": 0.16482658684253693, "global_step": 144292, "epoch": 1621, "lr": 4.51852232568452e-05}
{"train_loss": 0.07903705537319183, "global_step": 144293, "epoch": 1621, "lr": 4.518464621472515e-05}
{"train_loss": 0.08372080326080322, "global_step": 144294, "epoch": 1621, "lr": 4.5184069173252494e-05}
{"train_loss": 0.08853910863399506, "global_step": 144295, "epoch": 1621, "lr": 4.518349213242725e-05}
{"train_loss": 0.11762546002864838, "global_step": 144296, "epoch": 1621, "lr": 4.518291509224956e-05}
{"train_loss": 0.1093793585896492, "global_step": 144297, "epoch": 1621, "lr": 4.518233805271943e-05}
{"train_loss": 0.10387594252824783, "global_step": 144298, "epoch": 1621, "lr": 4.518176101383701e-05}
{"train_loss": 0.17331546545028687, "global_step": 144299, "epoch": 1621, "lr": 4.518118397560232e-05}
{"train_loss": 0.09501699358224869, "global_step": 144300, "epoch": 1621, "lr": 4.518060693801548e-05}
{"train_loss": 0.10765356570482254, "global_step": 144301, "epoch": 1621, "lr": 4.518002990107654e-05}
{"train_loss": 0.1513681858778, "global_step": 144302, "epoch": 1621, "lr": 4.517945286478557e-05}
{"train_loss": 0.08084148168563843, "global_step": 144303, "epoch": 1621, "lr": 4.5178875829142683e-05}
{"train_loss": 0.1209709644317627, "global_step": 144304, "epoch": 1621, "lr": 4.5178298794147925e-05}
{"train_loss": 0.16249752044677734, "global_step": 144305, "epoch": 1621, "lr": 4.5177721759801405e-05}
{"train_loss": 0.08477979153394699, "global_step": 144306, "epoch": 1621, "lr": 4.517714472610316e-05}
{"train_loss": 0.06774481385946274, "global_step": 144307, "epoch": 1621, "lr": 4.5176567693053315e-05}
{"train_loss": 0.06189894303679466, "global_step": 144308, "epoch": 1621, "lr": 4.5175990660651894e-05}
{"train_loss": 0.18217355012893677, "global_step": 144309, "epoch": 1621, "lr": 4.517541362889903e-05}
{"train_loss": 0.14421431720256805, "global_step": 144310, "epoch": 1621, "lr": 4.517483659779474e-05}
{"train_loss": 0.08915677666664124, "global_step": 144311, "epoch": 1621, "lr": 4.5174259567339165e-05}
{"train_loss": 0.11548823863267899, "global_step": 144312, "epoch": 1621, "lr": 4.517368253753233e-05}
{"train_loss": 0.12333567440509796, "global_step": 144313, "epoch": 1621, "lr": 4.517310550837435e-05}
{"train_loss": 0.1385660320520401, "global_step": 144314, "epoch": 1621, "lr": 4.517252847986527e-05}
{"train_loss": 0.06440593302249908, "global_step": 144315, "epoch": 1621, "lr": 4.517195145200521e-05}
{"train_loss": 0.09380359947681427, "global_step": 144316, "epoch": 1621, "lr": 4.5171374424794194e-05}
{"train_loss": 0.056884776800870895, "global_step": 144317, "epoch": 1621, "lr": 4.5170797398232354e-05}
{"train_loss": 0.12253870815038681, "global_step": 144318, "epoch": 1621, "lr": 4.517022037231971e-05}
{"train_loss": 0.11852313578128815, "global_step": 144319, "epoch": 1621, "lr": 4.516964334705641e-05}
{"train_loss": 0.08338271081447601, "global_step": 144320, "epoch": 1621, "lr": 4.516906632244247e-05}
{"train_loss": 0.15820635855197906, "global_step": 144321, "epoch": 1621, "lr": 4.516848929847798e-05}
{"train_loss": 0.07534979283809662, "global_step": 144322, "epoch": 1621, "lr": 4.5167912275163055e-05}
{"train_loss": 0.1389012187719345, "global_step": 144323, "epoch": 1621, "lr": 4.5167335252497716e-05}
{"train_loss": 0.06944981962442398, "global_step": 144324, "epoch": 1621, "lr": 4.51667582304821e-05}
{"train_loss": 0.1391952782869339, "global_step": 144325, "epoch": 1621, "lr": 4.516618120911623e-05}
{"train_loss": 0.0840567946434021, "global_step": 144326, "epoch": 1621, "lr": 4.5165604188400223e-05}
{"train_loss": 0.12451091408729553, "global_step": 144327, "epoch": 1621, "lr": 4.516502716833412e-05}
{"train_loss": 0.13607683777809143, "global_step": 144328, "epoch": 1621, "lr": 4.516445014891805e-05}
{"train_loss": 0.12299942970275879, "global_step": 144329, "epoch": 1621, "lr": 4.5163873130152036e-05}
{"train_loss": 0.09784718602895737, "global_step": 144330, "epoch": 1621, "lr": 4.5163296112036205e-05}
{"train_loss": 0.1019207313656807, "global_step": 144331, "epoch": 1621, "lr": 4.5162719094570584e-05}
{"train_loss": 0.18001531064510345, "global_step": 144332, "epoch": 1621, "lr": 4.51621420777553e-05}
{"train_loss": 0.09268578141927719, "global_step": 144333, "epoch": 1621, "lr": 4.5161565061590384e-05}
{"train_loss": 0.05213863030076027, "global_step": 144334, "epoch": 1621, "lr": 4.5160988046075954e-05}
{"train_loss": 0.08170686662197113, "global_step": 144335, "epoch": 1621, "lr": 4.516041103121205e-05}
{"train_loss": 0.136753112077713, "global_step": 144336, "epoch": 1621, "lr": 4.51598340169988e-05}
{"train_loss": 0.09086862951517105, "global_step": 144337, "epoch": 1621, "lr": 4.5159257003436225e-05}
{"train_loss": 0.07817428559064865, "global_step": 144338, "epoch": 1621, "lr": 4.5158679990524436e-05}
{"train_loss": 0.10829654335975647, "global_step": 144339, "epoch": 1621, "lr": 4.5158102978263514e-05}
{"train_loss": 0.10540170222520828, "global_step": 144340, "epoch": 1621, "lr": 4.515752596665351e-05}
{"train_loss": 0.117208331823349, "global_step": 144341, "epoch": 1621, "lr": 4.515694895569454e-05}
{"train_loss": 0.10096694529056549, "global_step": 144342, "epoch": 1621, "lr": 4.515637194538663e-05}
{"train_loss": 0.0753367692232132, "global_step": 144343, "epoch": 1621, "lr": 4.515579493572992e-05}
{"train_loss": 0.10362992435693741, "global_step": 144344, "epoch": 1621, "lr": 4.515521792672442e-05}
{"train_loss": 0.11683884263038635, "global_step": 144345, "epoch": 1621, "lr": 4.5154640918370264e-05}
{"train_loss": 0.07384684681892395, "global_step": 144346, "epoch": 1621, "lr": 4.515406391066749e-05}
{"train_loss": 0.09974067658185959, "global_step": 144347, "epoch": 1621, "lr": 4.515348690361622e-05}
{"train_loss": 0.084994375705719, "global_step": 144348, "epoch": 1621, "lr": 4.5152909897216475e-05}
{"train_loss": 0.127757266163826, "global_step": 144349, "epoch": 1621, "lr": 4.515233289146839e-05}
{"train_loss": 0.15310132503509521, "global_step": 144350, "epoch": 1621, "lr": 4.515175588637199e-05}
{"train_loss": 0.07431317865848541, "global_step": 144351, "epoch": 1621, "lr": 4.51511788819274e-05}
{"train_loss": 0.06736878305673599, "global_step": 144352, "epoch": 1621, "lr": 4.515060187813466e-05}
{"train_loss": 0.09513108432292938, "global_step": 144353, "epoch": 1621, "lr": 4.5150024874993876e-05}
{"train_loss": 0.1643335074186325, "global_step": 144354, "epoch": 1621, "lr": 4.5149447872505085e-05}
{"train_loss": 0.10160569101572037, "global_step": 144355, "epoch": 1621, "lr": 4.514887087066842e-05}
{"train_loss": 0.17079225182533264, "global_step": 144356, "epoch": 1621, "lr": 4.514829386948391e-05}
{"train_loss": 0.10749167999189892, "global_step": 144357, "epoch": 1621, "lr": 4.5147716868951665e-05, "val_loss": 5.997386932373047}
{"train_loss": 0.1367848962545395, "global_step": 144358, "epoch": 1622, "lr": 4.514713986907174e-05}
{"train_loss": 0.07570643723011017, "global_step": 144359, "epoch": 1622, "lr": 4.514656286984423e-05}
{"train_loss": 0.08814530819654465, "global_step": 144360, "epoch": 1622, "lr": 4.514598587126922e-05}
{"train_loss": 0.1332746297121048, "global_step": 144361, "epoch": 1622, "lr": 4.514540887334674e-05}
{"train_loss": 0.07433261722326279, "global_step": 144362, "epoch": 1622, "lr": 4.5144831876076934e-05}
{"train_loss": 0.0953345075249672, "global_step": 144363, "epoch": 1622, "lr": 4.5144254879459816e-05}
{"train_loss": 0.07649121433496475, "global_step": 144364, "epoch": 1622, "lr": 4.514367788349553e-05}
{"train_loss": 0.10740037262439728, "global_step": 144365, "epoch": 1622, "lr": 4.514310088818408e-05}
{"train_loss": 0.20503053069114685, "global_step": 144366, "epoch": 1622, "lr": 4.514252389352561e-05}
{"train_loss": 0.08382178843021393, "global_step": 144367, "epoch": 1622, "lr": 4.5141946899520145e-05}
{"train_loss": 0.06696178019046783, "global_step": 144368, "epoch": 1622, "lr": 4.514136990616781e-05}
{"train_loss": 0.12804168462753296, "global_step": 144369, "epoch": 1622, "lr": 4.5140792913468626e-05}
{"train_loss": 0.11579650640487671, "global_step": 144370, "epoch": 1622, "lr": 4.5140215921422735e-05}
{"train_loss": 0.06586867570877075, "global_step": 144371, "epoch": 1622, "lr": 4.5139638930030156e-05}
{"train_loss": 0.13853508234024048, "global_step": 144372, "epoch": 1622, "lr": 4.513906193929102e-05}
{"train_loss": 0.16365841031074524, "global_step": 144373, "epoch": 1622, "lr": 4.513848494920536e-05}
{"train_loss": 0.09273619204759598, "global_step": 144374, "epoch": 1622, "lr": 4.513790795977328e-05}
{"train_loss": 0.11676506698131561, "global_step": 144375, "epoch": 1622, "lr": 4.513733097099484e-05}
{"train_loss": 0.1209496259689331, "global_step": 144376, "epoch": 1622, "lr": 4.5136753982870126e-05}
{"train_loss": 0.10270148515701294, "global_step": 144377, "epoch": 1622, "lr": 4.513617699539923e-05}
{"train_loss": 0.11514485627412796, "global_step": 144378, "epoch": 1622, "lr": 4.513560000858221e-05}
{"train_loss": 0.06438310444355011, "global_step": 144379, "epoch": 1622, "lr": 4.5135023022419157e-05}
{"train_loss": 0.1396481692790985, "global_step": 144380, "epoch": 1622, "lr": 4.5134446036910117e-05}
{"train_loss": 0.0892840027809143, "global_step": 144381, "epoch": 1622, "lr": 4.5133869052055216e-05}
{"train_loss": 0.10273951292037964, "global_step": 144382, "epoch": 1622, "lr": 4.513329206785449e-05}
{"train_loss": 0.10045081377029419, "global_step": 144383, "epoch": 1622, "lr": 4.5132715084308046e-05}
{"train_loss": 0.09332993626594543, "global_step": 144384, "epoch": 1622, "lr": 4.513213810141593e-05}
{"train_loss": 0.06232938915491104, "global_step": 144385, "epoch": 1622, "lr": 4.513156111917827e-05}
{"train_loss": 0.11254921555519104, "global_step": 144386, "epoch": 1622, "lr": 4.5130984137595076e-05}
{"train_loss": 0.10819344967603683, "global_step": 144387, "epoch": 1622, "lr": 4.513040715666649e-05}
{"train_loss": 0.10071305930614471, "global_step": 144388, "epoch": 1622, "lr": 4.512983017639254e-05}
{"train_loss": 0.1436622440814972, "global_step": 144389, "epoch": 1622, "lr": 4.512925319677335e-05}
{"train_loss": 0.15080061554908752, "global_step": 144390, "epoch": 1622, "lr": 4.5128676217808954e-05}
{"train_loss": 0.14092111587524414, "global_step": 144391, "epoch": 1622, "lr": 4.5128099239499463e-05}
{"train_loss": 0.10551808029413223, "global_step": 144392, "epoch": 1622, "lr": 4.512752226184492e-05}
{"train_loss": 0.16605809330940247, "global_step": 144393, "epoch": 1622, "lr": 4.512694528484543e-05}
{"train_loss": 0.11071433126926422, "global_step": 144394, "epoch": 1622, "lr": 4.512636830850106e-05}
{"train_loss": 0.14209584891796112, "global_step": 144395, "epoch": 1622, "lr": 4.51257913328119e-05}
{"train_loss": 0.09789159148931503, "global_step": 144396, "epoch": 1622, "lr": 4.5125214357778014e-05}
{"train_loss": 0.12193980067968369, "global_step": 144397, "epoch": 1622, "lr": 4.512463738339948e-05}
{"train_loss": 0.13491865992546082, "global_step": 144398, "epoch": 1622, "lr": 4.51240604096764e-05}
{"train_loss": 0.05121101438999176, "global_step": 144399, "epoch": 1622, "lr": 4.5123483436608794e-05}
{"train_loss": 0.11990456283092499, "global_step": 144400, "epoch": 1622, "lr": 4.512290646419681e-05}
{"train_loss": 0.17969106137752533, "global_step": 144401, "epoch": 1622, "lr": 4.5122329492440465e-05}
{"train_loss": 0.13506947457790375, "global_step": 144402, "epoch": 1622, "lr": 4.512175252133988e-05}
{"train_loss": 0.07824836671352386, "global_step": 144403, "epoch": 1622, "lr": 4.51211755508951e-05}
{"train_loss": 0.10670267045497894, "global_step": 144404, "epoch": 1622, "lr": 4.5120598581106246e-05}
{"train_loss": 0.1324210911989212, "global_step": 144405, "epoch": 1622, "lr": 4.512002161197334e-05}
{"train_loss": 0.08577825874090195, "global_step": 144406, "epoch": 1622, "lr": 4.511944464349651e-05}
{"train_loss": 0.14246779680252075, "global_step": 144407, "epoch": 1622, "lr": 4.5118867675675786e-05}
{"train_loss": 0.17319071292877197, "global_step": 144408, "epoch": 1622, "lr": 4.51182907085113e-05}
{"train_loss": 0.1520131379365921, "global_step": 144409, "epoch": 1622, "lr": 4.511771374200307e-05}
{"train_loss": 0.1515839844942093, "global_step": 144410, "epoch": 1622, "lr": 4.511713677615123e-05}
{"train_loss": 0.09144788235425949, "global_step": 144411, "epoch": 1622, "lr": 4.511655981095581e-05}
{"train_loss": 0.12133271247148514, "global_step": 144412, "epoch": 1622, "lr": 4.511598284641692e-05}
{"train_loss": 0.06844914704561234, "global_step": 144413, "epoch": 1622, "lr": 4.511540588253463e-05}
{"train_loss": 0.1100255697965622, "global_step": 144414, "epoch": 1622, "lr": 4.5114828919309e-05}
{"train_loss": 0.11158469319343567, "global_step": 144415, "epoch": 1622, "lr": 4.511425195674014e-05}
{"train_loss": 0.11238346993923187, "global_step": 144416, "epoch": 1622, "lr": 4.51136749948281e-05}
{"train_loss": 0.18540655076503754, "global_step": 144417, "epoch": 1622, "lr": 4.511309803357298e-05}
{"train_loss": 0.07339887320995331, "global_step": 144418, "epoch": 1622, "lr": 4.5112521072974826e-05}
{"train_loss": 0.08915382623672485, "global_step": 144419, "epoch": 1622, "lr": 4.5111944113033746e-05}
{"train_loss": 0.06901837140321732, "global_step": 144420, "epoch": 1622, "lr": 4.511136715374979e-05}
{"train_loss": 0.09317698329687119, "global_step": 144421, "epoch": 1622, "lr": 4.511079019512307e-05}
{"train_loss": 0.13942886888980865, "global_step": 144422, "epoch": 1622, "lr": 4.511021323715363e-05}
{"train_loss": 0.14058291912078857, "global_step": 144423, "epoch": 1622, "lr": 4.510963627984157e-05}
{"train_loss": 0.1423502266407013, "global_step": 144424, "epoch": 1622, "lr": 4.510905932318695e-05}
{"train_loss": 0.19235479831695557, "global_step": 144425, "epoch": 1622, "lr": 4.5108482367189875e-05}
{"train_loss": 0.11015772074460983, "global_step": 144426, "epoch": 1622, "lr": 4.5107905411850386e-05}
{"train_loss": 0.11364572495222092, "global_step": 144427, "epoch": 1622, "lr": 4.5107328457168595e-05}
{"train_loss": 0.08919990062713623, "global_step": 144428, "epoch": 1622, "lr": 4.510675150314455e-05}
{"train_loss": 0.10850982367992401, "global_step": 144429, "epoch": 1622, "lr": 4.510617454977834e-05}
{"train_loss": 0.15341916680335999, "global_step": 144430, "epoch": 1622, "lr": 4.5105597597070064e-05}
{"train_loss": 0.09299416840076447, "global_step": 144431, "epoch": 1622, "lr": 4.5105020645019774e-05}
{"train_loss": 0.12443655729293823, "global_step": 144432, "epoch": 1622, "lr": 4.5104443693627555e-05}
{"train_loss": 0.11024878174066544, "global_step": 144433, "epoch": 1622, "lr": 4.510386674289348e-05}
{"train_loss": 0.1279725432395935, "global_step": 144434, "epoch": 1622, "lr": 4.510328979281763e-05}
{"train_loss": 0.07417191565036774, "global_step": 144435, "epoch": 1622, "lr": 4.510271284340009e-05}
{"train_loss": 0.11627605557441711, "global_step": 144436, "epoch": 1622, "lr": 4.5102135894640934e-05}
{"train_loss": 0.12148410826921463, "global_step": 144437, "epoch": 1622, "lr": 4.510155894654021e-05}
{"train_loss": 0.07075687497854233, "global_step": 144438, "epoch": 1622, "lr": 4.510098199909806e-05}
{"train_loss": 0.10128552466630936, "global_step": 144439, "epoch": 1622, "lr": 4.510040505231449e-05}
{"train_loss": 0.1400635689496994, "global_step": 144440, "epoch": 1622, "lr": 4.509982810618963e-05}
{"train_loss": 0.11067553609609604, "global_step": 144441, "epoch": 1622, "lr": 4.509925116072352e-05}
{"train_loss": 0.11906048655509949, "global_step": 144442, "epoch": 1622, "lr": 4.509867421591628e-05}
{"train_loss": 0.1093059703707695, "global_step": 144443, "epoch": 1622, "lr": 4.509809727176794e-05}
{"train_loss": 0.14190427958965302, "global_step": 144444, "epoch": 1622, "lr": 4.5097520328278625e-05}
{"train_loss": 0.1609775274991989, "global_step": 144445, "epoch": 1622, "lr": 4.509694338544836e-05}
{"train_loss": 0.11567095349009117, "global_step": 144446, "epoch": 1622, "lr": 4.509636644327728e-05, "val_loss": 6.231956481933594}
{"train_loss": 0.15331411361694336, "global_step": 144447, "epoch": 1623, "lr": 4.509578950176541e-05}
{"train_loss": 0.11786045879125595, "global_step": 144448, "epoch": 1623, "lr": 4.5095212560912856e-05}
{"train_loss": 0.11848095059394836, "global_step": 144449, "epoch": 1623, "lr": 4.5094635620719705e-05}
{"train_loss": 0.09642922878265381, "global_step": 144450, "epoch": 1623, "lr": 4.5094058681186015e-05}
{"train_loss": 0.16138629615306854, "global_step": 144451, "epoch": 1623, "lr": 4.509348174231186e-05}
{"train_loss": 0.09502608329057693, "global_step": 144452, "epoch": 1623, "lr": 4.5092904804097336e-05}
{"train_loss": 0.0875958576798439, "global_step": 144453, "epoch": 1623, "lr": 4.5092327866542504e-05}
{"train_loss": 0.2555530071258545, "global_step": 144454, "epoch": 1623, "lr": 4.5091750929647456e-05}
{"train_loss": 0.17040792107582092, "global_step": 144455, "epoch": 1623, "lr": 4.5091173993412274e-05}
{"train_loss": 0.11416461318731308, "global_step": 144456, "epoch": 1623, "lr": 4.509059705783699e-05}
{"train_loss": 0.19684219360351562, "global_step": 144457, "epoch": 1623, "lr": 4.509002012292175e-05}
{"train_loss": 0.13378673791885376, "global_step": 144458, "epoch": 1623, "lr": 4.508944318866657e-05}
{"train_loss": 0.09890079498291016, "global_step": 144459, "epoch": 1623, "lr": 4.508886625507158e-05}
{"train_loss": 0.07394973933696747, "global_step": 144460, "epoch": 1623, "lr": 4.50882893221368e-05}
{"train_loss": 0.1369336098432541, "global_step": 144461, "epoch": 1623, "lr": 4.5087712389862376e-05}
{"train_loss": 0.10722679644823074, "global_step": 144462, "epoch": 1623, "lr": 4.5087135458248314e-05}
{"train_loss": 0.09778886288404465, "global_step": 144463, "epoch": 1623, "lr": 4.508655852729476e-05}
{"train_loss": 0.17634578049182892, "global_step": 144464, "epoch": 1623, "lr": 4.5085981597001726e-05}
{"train_loss": 0.17117923498153687, "global_step": 144465, "epoch": 1623, "lr": 4.508540466736935e-05}
{"train_loss": 0.10917620360851288, "global_step": 144466, "epoch": 1623, "lr": 4.508482773839766e-05}
{"train_loss": 0.11520840227603912, "global_step": 144467, "epoch": 1623, "lr": 4.5084250810086754e-05}
{"train_loss": 0.1930101066827774, "global_step": 144468, "epoch": 1623, "lr": 4.5083673882436725e-05}
{"train_loss": 0.12997251749038696, "global_step": 144469, "epoch": 1623, "lr": 4.508309695544763e-05}
{"train_loss": 0.10625984519720078, "global_step": 144470, "epoch": 1623, "lr": 4.5082520029119565e-05}
{"train_loss": 0.0698300376534462, "global_step": 144471, "epoch": 1623, "lr": 4.5081943103452576e-05}
{"train_loss": 0.13275843858718872, "global_step": 144472, "epoch": 1623, "lr": 4.5081366178446774e-05}
{"train_loss": 0.08421789109706879, "global_step": 144473, "epoch": 1623, "lr": 4.508078925410222e-05}
{"train_loss": 0.0841236487030983, "global_step": 144474, "epoch": 1623, "lr": 4.508021233041898e-05}
{"train_loss": 0.1464558243751526, "global_step": 144475, "epoch": 1623, "lr": 4.507963540739716e-05}
{"train_loss": 0.12797924876213074, "global_step": 144476, "epoch": 1623, "lr": 4.5079058485036834e-05}
{"train_loss": 0.12127003818750381, "global_step": 144477, "epoch": 1623, "lr": 4.5078481563338046e-05}
{"train_loss": 0.07323965430259705, "global_step": 144478, "epoch": 1623, "lr": 4.507790464230091e-05}
{"train_loss": 0.0638914480805397, "global_step": 144479, "epoch": 1623, "lr": 4.507732772192547e-05}
{"train_loss": 0.08298172056674957, "global_step": 144480, "epoch": 1623, "lr": 4.507675080221185e-05}
{"train_loss": 0.08115525543689728, "global_step": 144481, "epoch": 1623, "lr": 4.5076173883160075e-05}
{"train_loss": 0.09607940912246704, "global_step": 144482, "epoch": 1623, "lr": 4.507559696477028e-05}
{"train_loss": 0.09835575520992279, "global_step": 144483, "epoch": 1623, "lr": 4.5075020047042474e-05}
{"train_loss": 0.10257833451032639, "global_step": 144484, "epoch": 1623, "lr": 4.5074443129976787e-05}
{"train_loss": 0.14725527167320251, "global_step": 144485, "epoch": 1623, "lr": 4.507386621357329e-05}
{"train_loss": 0.077679842710495, "global_step": 144486, "epoch": 1623, "lr": 4.5073289297832035e-05}
{"train_loss": 0.07884658873081207, "global_step": 144487, "epoch": 1623, "lr": 4.507271238275314e-05}
{"train_loss": 0.12845227122306824, "global_step": 144488, "epoch": 1623, "lr": 4.507213546833663e-05}
{"train_loss": 0.08749659359455109, "global_step": 144489, "epoch": 1623, "lr": 4.507155855458264e-05}
{"train_loss": 0.11137079447507858, "global_step": 144490, "epoch": 1623, "lr": 4.5070981641491204e-05}
{"train_loss": 0.08389682322740555, "global_step": 144491, "epoch": 1623, "lr": 4.507040472906242e-05}
{"train_loss": 0.10041696578264236, "global_step": 144492, "epoch": 1623, "lr": 4.506982781729636e-05}
{"train_loss": 0.10790493339300156, "global_step": 144493, "epoch": 1623, "lr": 4.506925090619309e-05}
{"train_loss": 0.15130172669887543, "global_step": 144494, "epoch": 1623, "lr": 4.5068673995752706e-05}
{"train_loss": 0.13429999351501465, "global_step": 144495, "epoch": 1623, "lr": 4.5068097085975296e-05}
{"train_loss": 0.1277453750371933, "global_step": 144496, "epoch": 1623, "lr": 4.5067520176860896e-05}
{"train_loss": 0.12291982024908066, "global_step": 144497, "epoch": 1623, "lr": 4.506694326840962e-05}
{"train_loss": 0.07113911211490631, "global_step": 144498, "epoch": 1623, "lr": 4.506636636062152e-05}
{"train_loss": 0.07352983206510544, "global_step": 144499, "epoch": 1623, "lr": 4.506578945349671e-05}
{"train_loss": 0.09919509291648865, "global_step": 144500, "epoch": 1623, "lr": 4.506521254703522e-05}
{"train_loss": 0.12388379871845245, "global_step": 144501, "epoch": 1623, "lr": 4.506463564123717e-05}
{"train_loss": 0.1317681521177292, "global_step": 144502, "epoch": 1623, "lr": 4.50640587361026e-05}
{"train_loss": 0.0717155858874321, "global_step": 144503, "epoch": 1623, "lr": 4.506348183163161e-05}
{"train_loss": 0.1649879515171051, "global_step": 144504, "epoch": 1623, "lr": 4.5062904927824294e-05}
{"train_loss": 0.056323472410440445, "global_step": 144505, "epoch": 1623, "lr": 4.506232802468069e-05}
{"train_loss": 0.09580143541097641, "global_step": 144506, "epoch": 1623, "lr": 4.506175112220092e-05}
{"train_loss": 0.06834061443805695, "global_step": 144507, "epoch": 1623, "lr": 4.5061174220385014e-05}
{"train_loss": 0.20659217238426208, "global_step": 144508, "epoch": 1623, "lr": 4.506059731923309e-05}
{"train_loss": 0.13927416503429413, "global_step": 144509, "epoch": 1623, "lr": 4.50600204187452e-05}
{"train_loss": 0.11839302629232407, "global_step": 144510, "epoch": 1623, "lr": 4.505944351892143e-05}
{"train_loss": 0.05298661068081856, "global_step": 144511, "epoch": 1623, "lr": 4.505886661976186e-05}
{"train_loss": 0.14684125781059265, "global_step": 144512, "epoch": 1623, "lr": 4.5058289721266564e-05}
{"train_loss": 0.09087000042200089, "global_step": 144513, "epoch": 1623, "lr": 4.505771282343562e-05}
{"train_loss": 0.07116075605154037, "global_step": 144514, "epoch": 1623, "lr": 4.505713592626912e-05}
{"train_loss": 0.10030060261487961, "global_step": 144515, "epoch": 1623, "lr": 4.5056559029767095e-05}
{"train_loss": 0.15372249484062195, "global_step": 144516, "epoch": 1623, "lr": 4.505598213392969e-05}
{"train_loss": 0.09354911744594574, "global_step": 144517, "epoch": 1623, "lr": 4.505540523875692e-05}
{"train_loss": 0.08811047673225403, "global_step": 144518, "epoch": 1623, "lr": 4.505482834424892e-05}
{"train_loss": 0.0993606373667717, "global_step": 144519, "epoch": 1623, "lr": 4.50542514504057e-05}
{"train_loss": 0.08666133135557175, "global_step": 144520, "epoch": 1623, "lr": 4.5053674557227395e-05}
{"train_loss": 0.11476186662912369, "global_step": 144521, "epoch": 1623, "lr": 4.5053097664714075e-05}
{"train_loss": 0.13679061830043793, "global_step": 144522, "epoch": 1623, "lr": 4.505252077286578e-05}
{"train_loss": 0.1582537442445755, "global_step": 144523, "epoch": 1623, "lr": 4.505194388168264e-05}
{"train_loss": 0.11371877044439316, "global_step": 144524, "epoch": 1623, "lr": 4.5051366991164684e-05}
{"train_loss": 0.10483699291944504, "global_step": 144525, "epoch": 1623, "lr": 4.505079010131204e-05}
{"train_loss": 0.1112065315246582, "global_step": 144526, "epoch": 1623, "lr": 4.5050213212124724e-05}
{"train_loss": 0.1006876528263092, "global_step": 144527, "epoch": 1623, "lr": 4.5049636323602874e-05}
{"train_loss": 0.13640987873077393, "global_step": 144528, "epoch": 1623, "lr": 4.504905943574653e-05}
{"train_loss": 0.0937848687171936, "global_step": 144529, "epoch": 1623, "lr": 4.5048482548555774e-05}
{"train_loss": 0.11923158168792725, "global_step": 144530, "epoch": 1623, "lr": 4.504790566203069e-05}
{"train_loss": 0.08452606201171875, "global_step": 144531, "epoch": 1623, "lr": 4.504732877617136e-05}
{"train_loss": 0.077362559735775, "global_step": 144532, "epoch": 1623, "lr": 4.504675189097785e-05}
{"train_loss": 0.09107548743486404, "global_step": 144533, "epoch": 1623, "lr": 4.5046175006450265e-05}
{"train_loss": 0.05840412899851799, "global_step": 144534, "epoch": 1623, "lr": 4.504559812258863e-05}
{"train_loss": 0.11273923101887275, "global_step": 144535, "epoch": 1623, "lr": 4.504502123939308e-05, "val_loss": 6.24290132522583}
{"train_loss": 0.14340604841709137, "global_step": 144536, "epoch": 1624, "lr": 4.504444435686364e-05}
{"train_loss": 0.09931641072034836, "global_step": 144537, "epoch": 1624, "lr": 4.504386747500044e-05}
{"train_loss": 0.11044257879257202, "global_step": 144538, "epoch": 1624, "lr": 4.504329059380351e-05}
{"train_loss": 0.13244108855724335, "global_step": 144539, "epoch": 1624, "lr": 4.504271371327295e-05}
{"train_loss": 0.09993986785411835, "global_step": 144540, "epoch": 1624, "lr": 4.5042136833408855e-05}
{"train_loss": 0.046813689172267914, "global_step": 144541, "epoch": 1624, "lr": 4.5041559954211266e-05}
{"train_loss": 0.09450248628854752, "global_step": 144542, "epoch": 1624, "lr": 4.50409830756803e-05}
{"train_loss": 0.1298350840806961, "global_step": 144543, "epoch": 1624, "lr": 4.5040406197815986e-05}
{"train_loss": 0.09946325421333313, "global_step": 144544, "epoch": 1624, "lr": 4.503982932061845e-05}
{"train_loss": 0.09855576604604721, "global_step": 144545, "epoch": 1624, "lr": 4.5039252444087736e-05}
{"train_loss": 0.13694071769714355, "global_step": 144546, "epoch": 1624, "lr": 4.503867556822395e-05}
{"train_loss": 0.09911458939313889, "global_step": 144547, "epoch": 1624, "lr": 4.5038098693027144e-05}
{"train_loss": 0.14826679229736328, "global_step": 144548, "epoch": 1624, "lr": 4.503752181849741e-05}
{"train_loss": 0.0910734236240387, "global_step": 144549, "epoch": 1624, "lr": 4.5036944944634815e-05}
{"train_loss": 0.07729441672563553, "global_step": 144550, "epoch": 1624, "lr": 4.5036368071439444e-05}
{"train_loss": 0.09687411040067673, "global_step": 144551, "epoch": 1624, "lr": 4.503579119891137e-05}
{"train_loss": 0.08441943675279617, "global_step": 144552, "epoch": 1624, "lr": 4.503521432705069e-05}
{"train_loss": 0.16578130424022675, "global_step": 144553, "epoch": 1624, "lr": 4.503463745585744e-05}
{"train_loss": 0.13089117407798767, "global_step": 144554, "epoch": 1624, "lr": 4.503406058533174e-05}
{"train_loss": 0.14650556445121765, "global_step": 144555, "epoch": 1624, "lr": 4.5033483715473635e-05}
{"train_loss": 0.07921557873487473, "global_step": 144556, "epoch": 1624, "lr": 4.503290684628321e-05}
{"train_loss": 0.15453994274139404, "global_step": 144557, "epoch": 1624, "lr": 4.503232997776058e-05}
{"train_loss": 0.10790210962295532, "global_step": 144558, "epoch": 1624, "lr": 4.503175310990577e-05}
{"train_loss": 0.09046315401792526, "global_step": 144559, "epoch": 1624, "lr": 4.5031176242718895e-05}
{"train_loss": 0.1832636594772339, "global_step": 144560, "epoch": 1624, "lr": 4.50305993762e-05}
{"train_loss": 0.08106570690870285, "global_step": 144561, "epoch": 1624, "lr": 4.5030022510349203e-05}
{"train_loss": 0.12586168944835663, "global_step": 144562, "epoch": 1624, "lr": 4.502944564516654e-05}
{"train_loss": 0.14635781943798065, "global_step": 144563, "epoch": 1624, "lr": 4.502886878065213e-05}
{"train_loss": 0.09594579041004181, "global_step": 144564, "epoch": 1624, "lr": 4.5028291916806e-05}
{"train_loss": 0.08882800489664078, "global_step": 144565, "epoch": 1624, "lr": 4.5027715053628275e-05}
{"train_loss": 0.07594606280326843, "global_step": 144566, "epoch": 1624, "lr": 4.502713819111901e-05}
{"train_loss": 0.13313239812850952, "global_step": 144567, "epoch": 1624, "lr": 4.5026561329278285e-05}
{"train_loss": 0.07042402029037476, "global_step": 144568, "epoch": 1624, "lr": 4.5025984468106174e-05}
{"train_loss": 0.1291590929031372, "global_step": 144569, "epoch": 1624, "lr": 4.502540760760276e-05}
{"train_loss": 0.091699980199337, "global_step": 144570, "epoch": 1624, "lr": 4.502483074776812e-05}
{"train_loss": 0.16812537610530853, "global_step": 144571, "epoch": 1624, "lr": 4.502425388860233e-05}
{"train_loss": 0.10820360481739044, "global_step": 144572, "epoch": 1624, "lr": 4.502367703010547e-05}
{"train_loss": 0.13343700766563416, "global_step": 144573, "epoch": 1624, "lr": 4.5023100172277624e-05}
{"train_loss": 0.0729990005493164, "global_step": 144574, "epoch": 1624, "lr": 4.502252331511884e-05}
{"train_loss": 0.1378766894340515, "global_step": 144575, "epoch": 1624, "lr": 4.5021946458629224e-05}
{"train_loss": 0.084690622985363, "global_step": 144576, "epoch": 1624, "lr": 4.502136960280886e-05}
{"train_loss": 0.13596166670322418, "global_step": 144577, "epoch": 1624, "lr": 4.5020792747657794e-05}
{"train_loss": 0.12386442720890045, "global_step": 144578, "epoch": 1624, "lr": 4.5020215893176145e-05}
{"train_loss": 0.1345142424106598, "global_step": 144579, "epoch": 1624, "lr": 4.5019639039363944e-05}
{"train_loss": 0.08760219812393188, "global_step": 144580, "epoch": 1624, "lr": 4.501906218622131e-05}
{"train_loss": 0.1318899542093277, "global_step": 144581, "epoch": 1624, "lr": 4.501848533374828e-05}
{"train_loss": 0.08187668025493622, "global_step": 144582, "epoch": 1624, "lr": 4.501790848194498e-05}
{"train_loss": 0.15915685892105103, "global_step": 144583, "epoch": 1624, "lr": 4.501733163081144e-05}
{"train_loss": 0.10384291410446167, "global_step": 144584, "epoch": 1624, "lr": 4.501675478034778e-05}
{"train_loss": 0.16846464574337006, "global_step": 144585, "epoch": 1624, "lr": 4.5016177930554036e-05}
{"train_loss": 0.08516908437013626, "global_step": 144586, "epoch": 1624, "lr": 4.501560108143033e-05}
{"train_loss": 0.06968023627996445, "global_step": 144587, "epoch": 1624, "lr": 4.50150242329767e-05}
{"train_loss": 0.0757477656006813, "global_step": 144588, "epoch": 1624, "lr": 4.501444738519324e-05}
{"train_loss": 0.09825899451971054, "global_step": 144589, "epoch": 1624, "lr": 4.501387053808003e-05}
{"train_loss": 0.12337534874677658, "global_step": 144590, "epoch": 1624, "lr": 4.501329369163714e-05}
{"train_loss": 0.13548240065574646, "global_step": 144591, "epoch": 1624, "lr": 4.501271684586465e-05}
{"train_loss": 0.06223258003592491, "global_step": 144592, "epoch": 1624, "lr": 4.501214000076266e-05}
{"train_loss": 0.0909796953201294, "global_step": 144593, "epoch": 1624, "lr": 4.50115631563312e-05}
{"train_loss": 0.1488237977027893, "global_step": 144594, "epoch": 1624, "lr": 4.501098631257037e-05}
{"train_loss": 0.07588490843772888, "global_step": 144595, "epoch": 1624, "lr": 4.501040946948028e-05}
{"train_loss": 0.16431832313537598, "global_step": 144596, "epoch": 1624, "lr": 4.5009832627060956e-05}
{"train_loss": 0.12929238379001617, "global_step": 144597, "epoch": 1624, "lr": 4.500925578531252e-05}
{"train_loss": 0.10129416733980179, "global_step": 144598, "epoch": 1624, "lr": 4.500867894423501e-05}
{"train_loss": 0.1251286417245865, "global_step": 144599, "epoch": 1624, "lr": 4.500810210382854e-05}
{"train_loss": 0.08001069724559784, "global_step": 144600, "epoch": 1624, "lr": 4.500752526409315e-05}
{"train_loss": 0.18445266783237457, "global_step": 144601, "epoch": 1624, "lr": 4.500694842502896e-05}
{"train_loss": 0.1783628761768341, "global_step": 144602, "epoch": 1624, "lr": 4.5006371586635996e-05}
{"train_loss": 0.09412533044815063, "global_step": 144603, "epoch": 1624, "lr": 4.500579474891439e-05}
{"train_loss": 0.10735171288251877, "global_step": 144604, "epoch": 1624, "lr": 4.500521791186418e-05}
{"train_loss": 0.08543667942285538, "global_step": 144605, "epoch": 1624, "lr": 4.5004641075485474e-05}
{"train_loss": 0.11011733114719391, "global_step": 144606, "epoch": 1624, "lr": 4.5004064239778324e-05}
{"train_loss": 0.16786924004554749, "global_step": 144607, "epoch": 1624, "lr": 4.500348740474281e-05}
{"train_loss": 0.1283065676689148, "global_step": 144608, "epoch": 1624, "lr": 4.500291057037902e-05}
{"train_loss": 0.17513759434223175, "global_step": 144609, "epoch": 1624, "lr": 4.500233373668703e-05}
{"train_loss": 0.10840786248445511, "global_step": 144610, "epoch": 1624, "lr": 4.500175690366691e-05}
{"train_loss": 0.10441303253173828, "global_step": 144611, "epoch": 1624, "lr": 4.5001180071318735e-05}
{"train_loss": 0.10072540491819382, "global_step": 144612, "epoch": 1624, "lr": 4.500060323964262e-05}
{"train_loss": 0.1345648616552353, "global_step": 144613, "epoch": 1624, "lr": 4.5000026408638584e-05}
{"train_loss": 0.18508274853229523, "global_step": 144614, "epoch": 1624, "lr": 4.499944957830676e-05}
{"train_loss": 0.10829752683639526, "global_step": 144615, "epoch": 1624, "lr": 4.499887274864717e-05}
{"train_loss": 0.19266512989997864, "global_step": 144616, "epoch": 1624, "lr": 4.499829591965994e-05}
{"train_loss": 0.08746124804019928, "global_step": 144617, "epoch": 1624, "lr": 4.4997719091345114e-05}
{"train_loss": 0.10553202033042908, "global_step": 144618, "epoch": 1624, "lr": 4.49971422637028e-05}
{"train_loss": 0.14206428825855255, "global_step": 144619, "epoch": 1624, "lr": 4.499656543673305e-05}
{"train_loss": 0.07689899951219559, "global_step": 144620, "epoch": 1624, "lr": 4.499598861043596e-05}
{"train_loss": 0.19352629780769348, "global_step": 144621, "epoch": 1624, "lr": 4.499541178481158e-05}
{"train_loss": 0.10797697305679321, "global_step": 144622, "epoch": 1624, "lr": 4.4994834959860024e-05}
{"train_loss": 0.15050745010375977, "global_step": 144623, "epoch": 1624, "lr": 4.4994258135581334e-05}
{"train_loss": 0.11818352195151736, "global_step": 144624, "epoch": 1624, "lr": 4.499368131197563e-05, "val_loss": 6.054816246032715}
{"train_loss": 0.11369922757148743, "global_step": 144625, "epoch": 1625, "lr": 4.499310448904294e-05}
{"train_loss": 0.12694577872753143, "global_step": 144626, "epoch": 1625, "lr": 4.499252766678338e-05}
{"train_loss": 0.13202261924743652, "global_step": 144627, "epoch": 1625, "lr": 4.499195084519701e-05}
{"train_loss": 0.1285778433084488, "global_step": 144628, "epoch": 1625, "lr": 4.499137402428391e-05}
{"train_loss": 0.164181187748909, "global_step": 144629, "epoch": 1625, "lr": 4.499079720404416e-05}
{"train_loss": 0.06749705225229263, "global_step": 144630, "epoch": 1625, "lr": 4.499022038447782e-05}
{"train_loss": 0.15153998136520386, "global_step": 144631, "epoch": 1625, "lr": 4.498964356558501e-05}
{"train_loss": 0.12371936440467834, "global_step": 144632, "epoch": 1625, "lr": 4.498906674736576e-05}
{"train_loss": 0.03327213600277901, "global_step": 144633, "epoch": 1625, "lr": 4.4988489929820194e-05}
{"train_loss": 0.09120027720928192, "global_step": 144634, "epoch": 1625, "lr": 4.498791311294834e-05}
{"train_loss": 0.07054726779460907, "global_step": 144635, "epoch": 1625, "lr": 4.4987336296750316e-05}
{"train_loss": 0.11981333792209625, "global_step": 144636, "epoch": 1625, "lr": 4.498675948122617e-05}
{"train_loss": 0.1327361911535263, "global_step": 144637, "epoch": 1625, "lr": 4.498618266637601e-05}
{"train_loss": 0.07357092201709747, "global_step": 144638, "epoch": 1625, "lr": 4.4985605852199874e-05}
{"train_loss": 0.13663944602012634, "global_step": 144639, "epoch": 1625, "lr": 4.498502903869789e-05}
{"train_loss": 0.10281572490930557, "global_step": 144640, "epoch": 1625, "lr": 4.4984452225870075e-05}
{"train_loss": 0.07550087571144104, "global_step": 144641, "epoch": 1625, "lr": 4.4983875413716566e-05}
{"train_loss": 0.10523822158575058, "global_step": 144642, "epoch": 1625, "lr": 4.49832986022374e-05}
{"train_loss": 0.08183926343917847, "global_step": 144643, "epoch": 1625, "lr": 4.4982721791432684e-05}
{"train_loss": 0.0856085792183876, "global_step": 144644, "epoch": 1625, "lr": 4.498214498130247e-05}
{"train_loss": 0.11890624463558197, "global_step": 144645, "epoch": 1625, "lr": 4.4981568171846846e-05}
{"train_loss": 0.12937568128108978, "global_step": 144646, "epoch": 1625, "lr": 4.498099136306589e-05}
{"train_loss": 0.09904280304908752, "global_step": 144647, "epoch": 1625, "lr": 4.498041455495966e-05}
{"train_loss": 0.17668476700782776, "global_step": 144648, "epoch": 1625, "lr": 4.497983774752828e-05}
{"train_loss": 0.13824035227298737, "global_step": 144649, "epoch": 1625, "lr": 4.4979260940771775e-05}
{"train_loss": 0.06824721395969391, "global_step": 144650, "epoch": 1625, "lr": 4.497868413469027e-05}
{"train_loss": 0.09625223278999329, "global_step": 144651, "epoch": 1625, "lr": 4.49781073292838e-05}
{"train_loss": 0.08758725225925446, "global_step": 144652, "epoch": 1625, "lr": 4.4977530524552484e-05}
{"train_loss": 0.10334709286689758, "global_step": 144653, "epoch": 1625, "lr": 4.497695372049635e-05}
{"train_loss": 0.12608379125595093, "global_step": 144654, "epoch": 1625, "lr": 4.497637691711554e-05}
{"train_loss": 0.13389705121517181, "global_step": 144655, "epoch": 1625, "lr": 4.497580011441006e-05}
{"train_loss": 0.10315209627151489, "global_step": 144656, "epoch": 1625, "lr": 4.497522331238005e-05}
{"train_loss": 0.10275536775588989, "global_step": 144657, "epoch": 1625, "lr": 4.497464651102553e-05}
{"train_loss": 0.1364954710006714, "global_step": 144658, "epoch": 1625, "lr": 4.497406971034664e-05}
{"train_loss": 0.21307367086410522, "global_step": 144659, "epoch": 1625, "lr": 4.49734929103434e-05}
{"train_loss": 0.09011375159025192, "global_step": 144660, "epoch": 1625, "lr": 4.497291611101593e-05}
{"train_loss": 0.13763993978500366, "global_step": 144661, "epoch": 1625, "lr": 4.497233931236427e-05}
{"train_loss": 0.0687907412648201, "global_step": 144662, "epoch": 1625, "lr": 4.497176251438854e-05}
{"train_loss": 0.06509260833263397, "global_step": 144663, "epoch": 1625, "lr": 4.497118571708879e-05}
{"train_loss": 0.06254487484693527, "global_step": 144664, "epoch": 1625, "lr": 4.4970608920465094e-05}
{"train_loss": 0.09805704653263092, "global_step": 144665, "epoch": 1625, "lr": 4.497003212451754e-05}
{"train_loss": 0.0678752064704895, "global_step": 144666, "epoch": 1625, "lr": 4.496945532924619e-05}
{"train_loss": 0.10559437423944473, "global_step": 144667, "epoch": 1625, "lr": 4.496887853465117e-05}
{"train_loss": 0.11511003971099854, "global_step": 144668, "epoch": 1625, "lr": 4.496830174073249e-05}
{"train_loss": 0.1295393705368042, "global_step": 144669, "epoch": 1625, "lr": 4.496772494749029e-05}
{"train_loss": 0.09388532489538193, "global_step": 144670, "epoch": 1625, "lr": 4.4967148154924586e-05}
{"train_loss": 0.11698858439922333, "global_step": 144671, "epoch": 1625, "lr": 4.496657136303552e-05}
{"train_loss": 0.07828985154628754, "global_step": 144672, "epoch": 1625, "lr": 4.496599457182311e-05}
{"train_loss": 0.1466970294713974, "global_step": 144673, "epoch": 1625, "lr": 4.4965417781287485e-05}
{"train_loss": 0.08555033802986145, "global_step": 144674, "epoch": 1625, "lr": 4.496484099142867e-05}
{"train_loss": 0.08821099251508713, "global_step": 144675, "epoch": 1625, "lr": 4.4964264202246796e-05}
{"train_loss": 0.18648496270179749, "global_step": 144676, "epoch": 1625, "lr": 4.496368741374189e-05}
{"train_loss": 0.09378356486558914, "global_step": 144677, "epoch": 1625, "lr": 4.496311062591409e-05}
{"train_loss": 0.12620024383068085, "global_step": 144678, "epoch": 1625, "lr": 4.4962533838763404e-05}
{"train_loss": 0.1906539350748062, "global_step": 144679, "epoch": 1625, "lr": 4.496195705228997e-05}
{"train_loss": 0.07972802966833115, "global_step": 144680, "epoch": 1625, "lr": 4.4961380266493816e-05}
{"train_loss": 0.16385401785373688, "global_step": 144681, "epoch": 1625, "lr": 4.496080348137507e-05}
{"train_loss": 0.15173465013504028, "global_step": 144682, "epoch": 1625, "lr": 4.4960226696933764e-05}
{"train_loss": 0.1549503058195114, "global_step": 144683, "epoch": 1625, "lr": 4.495964991316999e-05}
{"train_loss": 0.13476940989494324, "global_step": 144684, "epoch": 1625, "lr": 4.495907313008385e-05}
{"train_loss": 0.11611970514059067, "global_step": 144685, "epoch": 1625, "lr": 4.4958496347675375e-05}
{"train_loss": 0.0657113641500473, "global_step": 144686, "epoch": 1625, "lr": 4.4957919565944695e-05}
{"train_loss": 0.09794019162654877, "global_step": 144687, "epoch": 1625, "lr": 4.495734278489184e-05}
{"train_loss": 0.1003379374742508, "global_step": 144688, "epoch": 1625, "lr": 4.4956766004516925e-05}
{"train_loss": 0.09051906317472458, "global_step": 144689, "epoch": 1625, "lr": 4.495618922481999e-05}
{"train_loss": 0.09893407672643661, "global_step": 144690, "epoch": 1625, "lr": 4.495561244580116e-05}
{"train_loss": 0.07996121793985367, "global_step": 144691, "epoch": 1625, "lr": 4.4955035667460466e-05}
{"train_loss": 0.09178375452756882, "global_step": 144692, "epoch": 1625, "lr": 4.4954458889798024e-05}
{"train_loss": 0.09342744201421738, "global_step": 144693, "epoch": 1625, "lr": 4.495388211281387e-05}
{"train_loss": 0.12408146262168884, "global_step": 144694, "epoch": 1625, "lr": 4.4953305336508134e-05}
{"train_loss": 0.11777031421661377, "global_step": 144695, "epoch": 1625, "lr": 4.4952728560880834e-05}
{"train_loss": 0.10387670248746872, "global_step": 144696, "epoch": 1625, "lr": 4.495215178593211e-05}
{"train_loss": 0.15193241834640503, "global_step": 144697, "epoch": 1625, "lr": 4.495157501166197e-05}
{"train_loss": 0.06662840396165848, "global_step": 144698, "epoch": 1625, "lr": 4.495099823807056e-05}
{"train_loss": 0.10467865318059921, "global_step": 144699, "epoch": 1625, "lr": 4.495042146515791e-05}
{"train_loss": 0.09206700325012207, "global_step": 144700, "epoch": 1625, "lr": 4.4949844692924134e-05}
{"train_loss": 0.05874893069267273, "global_step": 144701, "epoch": 1625, "lr": 4.494926792136927e-05}
{"train_loss": 0.11829189211130142, "global_step": 144702, "epoch": 1625, "lr": 4.4948691150493415e-05}
{"train_loss": 0.13895776867866516, "global_step": 144703, "epoch": 1625, "lr": 4.494811438029667e-05}
{"train_loss": 0.061877887696027756, "global_step": 144704, "epoch": 1625, "lr": 4.4947537610779064e-05}
{"train_loss": 0.13264869153499603, "global_step": 144705, "epoch": 1625, "lr": 4.494696084194071e-05}
{"train_loss": 0.07547459006309509, "global_step": 144706, "epoch": 1625, "lr": 4.494638407378167e-05}
{"train_loss": 0.09072209149599075, "global_step": 144707, "epoch": 1625, "lr": 4.494580730630205e-05}
{"train_loss": 0.05772385746240616, "global_step": 144708, "epoch": 1625, "lr": 4.494523053950187e-05}
{"train_loss": 0.13041973114013672, "global_step": 144709, "epoch": 1625, "lr": 4.494465377338127e-05}
{"train_loss": 0.10163421928882599, "global_step": 144710, "epoch": 1625, "lr": 4.4944077007940276e-05}
{"train_loss": 0.09918934106826782, "global_step": 144711, "epoch": 1625, "lr": 4.494350024317901e-05}
{"train_loss": 0.10736542195081711, "global_step": 144712, "epoch": 1625, "lr": 4.494292347909752e-05}
{"train_loss": 0.10808263744196195, "global_step": 144713, "epoch": 1625, "lr": 4.49423467156959e-05, "val_loss": 6.293009281158447, "train_action_mse_error": 16.56369972229004}
{"train_loss": 0.09511350095272064, "global_step": 144714, "epoch": 1626, "lr": 4.49417699529742e-05}
{"train_loss": 0.10849539935588837, "global_step": 144715, "epoch": 1626, "lr": 4.494119319093254e-05}
{"train_loss": 0.06465530395507812, "global_step": 144716, "epoch": 1626, "lr": 4.494061642957096e-05}
{"train_loss": 0.07585427910089493, "global_step": 144717, "epoch": 1626, "lr": 4.4940039668889565e-05}
{"train_loss": 0.11587013304233551, "global_step": 144718, "epoch": 1626, "lr": 4.4939462908888404e-05}
{"train_loss": 0.1474333256483078, "global_step": 144719, "epoch": 1626, "lr": 4.4938886149567586e-05}
{"train_loss": 0.1214742660522461, "global_step": 144720, "epoch": 1626, "lr": 4.493830939092716e-05}
{"train_loss": 0.11793682724237442, "global_step": 144721, "epoch": 1626, "lr": 4.493773263296722e-05}
{"train_loss": 0.12901203334331512, "global_step": 144722, "epoch": 1626, "lr": 4.493715587568785e-05}
{"train_loss": 0.10231351107358932, "global_step": 144723, "epoch": 1626, "lr": 4.493657911908911e-05}
{"train_loss": 0.08835224062204361, "global_step": 144724, "epoch": 1626, "lr": 4.493600236317109e-05}
{"train_loss": 0.13628339767456055, "global_step": 144725, "epoch": 1626, "lr": 4.493542560793385e-05}
{"train_loss": 0.06338836997747421, "global_step": 144726, "epoch": 1626, "lr": 4.4934848853377495e-05}
{"train_loss": 0.16145259141921997, "global_step": 144727, "epoch": 1626, "lr": 4.493427209950207e-05}
{"train_loss": 0.08130347728729248, "global_step": 144728, "epoch": 1626, "lr": 4.49336953463077e-05}
{"train_loss": 0.12143867462873459, "global_step": 144729, "epoch": 1626, "lr": 4.49331185937944e-05}
{"train_loss": 0.08134874701499939, "global_step": 144730, "epoch": 1626, "lr": 4.4932541841962306e-05}
{"train_loss": 0.17381562292575836, "global_step": 144731, "epoch": 1626, "lr": 4.493196509081145e-05}
{"train_loss": 0.09641776233911514, "global_step": 144732, "epoch": 1626, "lr": 4.4931388340341953e-05}
{"train_loss": 0.14723782241344452, "global_step": 144733, "epoch": 1626, "lr": 4.493081159055384e-05}
{"train_loss": 0.1372021734714508, "global_step": 144734, "epoch": 1626, "lr": 4.4930234841447245e-05}
{"train_loss": 0.08127789199352264, "global_step": 144735, "epoch": 1626, "lr": 4.492965809302219e-05}
{"train_loss": 0.05131426826119423, "global_step": 144736, "epoch": 1626, "lr": 4.492908134527881e-05}
{"train_loss": 0.06572313606739044, "global_step": 144737, "epoch": 1626, "lr": 4.4928504598217134e-05}
{"train_loss": 0.13051719963550568, "global_step": 144738, "epoch": 1626, "lr": 4.492792785183726e-05}
{"train_loss": 0.12061423808336258, "global_step": 144739, "epoch": 1626, "lr": 4.492735110613927e-05}
{"train_loss": 0.15623649954795837, "global_step": 144740, "epoch": 1626, "lr": 4.4926774361123234e-05}
{"train_loss": 0.07425635308027267, "global_step": 144741, "epoch": 1626, "lr": 4.492619761678925e-05}
{"train_loss": 0.09969709068536758, "global_step": 144742, "epoch": 1626, "lr": 4.492562087313734e-05}
{"train_loss": 0.09286931902170181, "global_step": 144743, "epoch": 1626, "lr": 4.492504413016765e-05}
{"train_loss": 0.06997716426849365, "global_step": 144744, "epoch": 1626, "lr": 4.49244673878802e-05}
{"train_loss": 0.13019435107707977, "global_step": 144745, "epoch": 1626, "lr": 4.492389064627512e-05}
{"train_loss": 0.09117387235164642, "global_step": 144746, "epoch": 1626, "lr": 4.492331390535243e-05}
{"train_loss": 0.06783725321292877, "global_step": 144747, "epoch": 1626, "lr": 4.492273716511227e-05}
{"train_loss": 0.1227659285068512, "global_step": 144748, "epoch": 1626, "lr": 4.492216042555466e-05}
{"train_loss": 0.09662992507219315, "global_step": 144749, "epoch": 1626, "lr": 4.492158368667973e-05}
{"train_loss": 0.17184333503246307, "global_step": 144750, "epoch": 1626, "lr": 4.4921006948487507e-05}
{"train_loss": 0.08228324353694916, "global_step": 144751, "epoch": 1626, "lr": 4.492043021097811e-05}
{"train_loss": 0.11914398521184921, "global_step": 144752, "epoch": 1626, "lr": 4.491985347415158e-05}
{"train_loss": 0.09263163805007935, "global_step": 144753, "epoch": 1626, "lr": 4.491927673800804e-05}
{"train_loss": 0.06469964981079102, "global_step": 144754, "epoch": 1626, "lr": 4.491870000254752e-05}
{"train_loss": 0.09324005246162415, "global_step": 144755, "epoch": 1626, "lr": 4.4918123267770136e-05}
{"train_loss": 0.0468427874147892, "global_step": 144756, "epoch": 1626, "lr": 4.4917546533675925e-05}
{"train_loss": 0.05835622921586037, "global_step": 144757, "epoch": 1626, "lr": 4.4916969800265005e-05}
{"train_loss": 0.14349600672721863, "global_step": 144758, "epoch": 1626, "lr": 4.491639306753743e-05}
{"train_loss": 0.09696850925683975, "global_step": 144759, "epoch": 1626, "lr": 4.491581633549329e-05}
{"train_loss": 0.07430117577314377, "global_step": 144760, "epoch": 1626, "lr": 4.4915239604132656e-05}
{"train_loss": 0.08268842101097107, "global_step": 144761, "epoch": 1626, "lr": 4.49146628734556e-05}
{"train_loss": 0.12869031727313995, "global_step": 144762, "epoch": 1626, "lr": 4.491408614346222e-05}
{"train_loss": 0.10422484576702118, "global_step": 144763, "epoch": 1626, "lr": 4.491350941415255e-05}
{"train_loss": 0.06730881333351135, "global_step": 144764, "epoch": 1626, "lr": 4.491293268552673e-05}
{"train_loss": 0.09493005275726318, "global_step": 144765, "epoch": 1626, "lr": 4.491235595758478e-05}
{"train_loss": 0.11091060936450958, "global_step": 144766, "epoch": 1626, "lr": 4.491177923032682e-05}
{"train_loss": 0.11387734860181808, "global_step": 144767, "epoch": 1626, "lr": 4.491120250375289e-05}
{"train_loss": 0.1025007888674736, "global_step": 144768, "epoch": 1626, "lr": 4.491062577786311e-05}
{"train_loss": 0.07488076388835907, "global_step": 144769, "epoch": 1626, "lr": 4.491004905265751e-05}
{"train_loss": 0.07021209597587585, "global_step": 144770, "epoch": 1626, "lr": 4.490947232813621e-05}
{"train_loss": 0.0877046212553978, "global_step": 144771, "epoch": 1626, "lr": 4.490889560429925e-05}
{"train_loss": 0.06521203368902206, "global_step": 144772, "epoch": 1626, "lr": 4.4908318881146746e-05}
{"train_loss": 0.12403973191976547, "global_step": 144773, "epoch": 1626, "lr": 4.4907742158678746e-05}
{"train_loss": 0.036838460713624954, "global_step": 144774, "epoch": 1626, "lr": 4.490716543689534e-05}
{"train_loss": 0.0654224380850792, "global_step": 144775, "epoch": 1626, "lr": 4.4906588715796605e-05}
{"train_loss": 0.10128691047430038, "global_step": 144776, "epoch": 1626, "lr": 4.490601199538262e-05}
{"train_loss": 0.09874894469976425, "global_step": 144777, "epoch": 1626, "lr": 4.490543527565346e-05}
{"train_loss": 0.11119142174720764, "global_step": 144778, "epoch": 1626, "lr": 4.49048585566092e-05}
{"train_loss": 0.13332265615463257, "global_step": 144779, "epoch": 1626, "lr": 4.4904281838249925e-05}
{"train_loss": 0.0866883248090744, "global_step": 144780, "epoch": 1626, "lr": 4.490370512057569e-05}
{"train_loss": 0.06028091162443161, "global_step": 144781, "epoch": 1626, "lr": 4.4903128403586605e-05}
{"train_loss": 0.05360260233283043, "global_step": 144782, "epoch": 1626, "lr": 4.490255168728272e-05}
{"train_loss": 0.14313437044620514, "global_step": 144783, "epoch": 1626, "lr": 4.490197497166415e-05}
{"train_loss": 0.07076260447502136, "global_step": 144784, "epoch": 1626, "lr": 4.4901398256730914e-05}
{"train_loss": 0.08660464733839035, "global_step": 144785, "epoch": 1626, "lr": 4.490082154248315e-05}
{"train_loss": 0.11984811723232269, "global_step": 144786, "epoch": 1626, "lr": 4.490024482892089e-05}
{"train_loss": 0.09698306024074554, "global_step": 144787, "epoch": 1626, "lr": 4.4899668116044244e-05}
{"train_loss": 0.09661368280649185, "global_step": 144788, "epoch": 1626, "lr": 4.489909140385326e-05}
{"train_loss": 0.12801994383335114, "global_step": 144789, "epoch": 1626, "lr": 4.489851469234805e-05}
{"train_loss": 0.0769544169306755, "global_step": 144790, "epoch": 1626, "lr": 4.489793798152865e-05}
{"train_loss": 0.07610917091369629, "global_step": 144791, "epoch": 1626, "lr": 4.4897361271395185e-05}
{"train_loss": 0.09684096276760101, "global_step": 144792, "epoch": 1626, "lr": 4.489678456194769e-05}
{"train_loss": 0.08747423440217972, "global_step": 144793, "epoch": 1626, "lr": 4.489620785318625e-05}
{"train_loss": 0.15187832713127136, "global_step": 144794, "epoch": 1626, "lr": 4.4895631145110976e-05}
{"train_loss": 0.1376573145389557, "global_step": 144795, "epoch": 1626, "lr": 4.489505443772192e-05}
{"train_loss": 0.15402641892433167, "global_step": 144796, "epoch": 1626, "lr": 4.489447773101915e-05}
{"train_loss": 0.11576946824789047, "global_step": 144797, "epoch": 1626, "lr": 4.489390102500276e-05}
{"train_loss": 0.15004320442676544, "global_step": 144798, "epoch": 1626, "lr": 4.4893324319672835e-05}
{"train_loss": 0.12025459855794907, "global_step": 144799, "epoch": 1626, "lr": 4.489274761502942e-05}
{"train_loss": 0.08159185200929642, "global_step": 144800, "epoch": 1626, "lr": 4.489217091107263e-05}
{"train_loss": 0.08611632883548737, "global_step": 144801, "epoch": 1626, "lr": 4.4891594207802504e-05}
{"train_loss": 0.10088280483745457, "global_step": 144802, "epoch": 1626, "lr": 4.4891017505219165e-05, "val_loss": 6.13787317276001}
{"train_loss": 0.06101538985967636, "global_step": 144803, "epoch": 1627, "lr": 4.489044080332265e-05}
{"train_loss": 0.11313105374574661, "global_step": 144804, "epoch": 1627, "lr": 4.488986410211307e-05}
{"train_loss": 0.08105888217687607, "global_step": 144805, "epoch": 1627, "lr": 4.488928740159046e-05}
{"train_loss": 0.08984256535768509, "global_step": 144806, "epoch": 1627, "lr": 4.488871070175495e-05}
{"train_loss": 0.0843600407242775, "global_step": 144807, "epoch": 1627, "lr": 4.4888134002606565e-05}
{"train_loss": 0.10111753642559052, "global_step": 144808, "epoch": 1627, "lr": 4.488755730414543e-05}
{"train_loss": 0.10827140510082245, "global_step": 144809, "epoch": 1627, "lr": 4.488698060637158e-05}
{"train_loss": 0.14875759184360504, "global_step": 144810, "epoch": 1627, "lr": 4.488640390928511e-05}
{"train_loss": 0.07187262922525406, "global_step": 144811, "epoch": 1627, "lr": 4.488582721288613e-05}
{"train_loss": 0.067855104804039, "global_step": 144812, "epoch": 1627, "lr": 4.4885250517174674e-05}
{"train_loss": 0.06518889218568802, "global_step": 144813, "epoch": 1627, "lr": 4.488467382215083e-05}
{"train_loss": 0.11904989182949066, "global_step": 144814, "epoch": 1627, "lr": 4.488409712781468e-05}
{"train_loss": 0.10229276120662689, "global_step": 144815, "epoch": 1627, "lr": 4.4883520434166306e-05}
{"train_loss": 0.06618408113718033, "global_step": 144816, "epoch": 1627, "lr": 4.488294374120578e-05}
{"train_loss": 0.18022987246513367, "global_step": 144817, "epoch": 1627, "lr": 4.4882367048933173e-05}
{"train_loss": 0.056166645139455795, "global_step": 144818, "epoch": 1627, "lr": 4.488179035734857e-05}
{"train_loss": 0.06466202437877655, "global_step": 144819, "epoch": 1627, "lr": 4.4881213666452064e-05}
{"train_loss": 0.08932258188724518, "global_step": 144820, "epoch": 1627, "lr": 4.4880636976243694e-05}
{"train_loss": 0.14686788618564606, "global_step": 144821, "epoch": 1627, "lr": 4.488006028672358e-05}
{"train_loss": 0.06185077130794525, "global_step": 144822, "epoch": 1627, "lr": 4.487948359789176e-05}
{"train_loss": 0.09714961796998978, "global_step": 144823, "epoch": 1627, "lr": 4.4878906909748356e-05}
{"train_loss": 0.09539161622524261, "global_step": 144824, "epoch": 1627, "lr": 4.487833022229339e-05}
{"train_loss": 0.12688419222831726, "global_step": 144825, "epoch": 1627, "lr": 4.4877753535526996e-05}
{"train_loss": 0.11360732465982437, "global_step": 144826, "epoch": 1627, "lr": 4.4877176849449206e-05}
{"train_loss": 0.10058525949716568, "global_step": 144827, "epoch": 1627, "lr": 4.4876600164060134e-05}
{"train_loss": 0.06250321865081787, "global_step": 144828, "epoch": 1627, "lr": 4.4876023479359816e-05}
{"train_loss": 0.10310735553503036, "global_step": 144829, "epoch": 1627, "lr": 4.4875446795348365e-05}
{"train_loss": 0.11387462913990021, "global_step": 144830, "epoch": 1627, "lr": 4.4874870112025865e-05}
{"train_loss": 0.10256791114807129, "global_step": 144831, "epoch": 1627, "lr": 4.487429342939235e-05}
{"train_loss": 0.11850748211145401, "global_step": 144832, "epoch": 1627, "lr": 4.4873716747447955e-05}
{"train_loss": 0.09174715727567673, "global_step": 144833, "epoch": 1627, "lr": 4.48731400661927e-05}
{"train_loss": 0.08416051417589188, "global_step": 144834, "epoch": 1627, "lr": 4.4872563385626705e-05}
{"train_loss": 0.11389264464378357, "global_step": 144835, "epoch": 1627, "lr": 4.487198670575002e-05}
{"train_loss": 0.11802658438682556, "global_step": 144836, "epoch": 1627, "lr": 4.487141002656274e-05}
{"train_loss": 0.07952359318733215, "global_step": 144837, "epoch": 1627, "lr": 4.4870833348064935e-05}
{"train_loss": 0.14268141984939575, "global_step": 144838, "epoch": 1627, "lr": 4.487025667025669e-05}
{"train_loss": 0.10123492777347565, "global_step": 144839, "epoch": 1627, "lr": 4.4869679993138064e-05}
{"train_loss": 0.13541600108146667, "global_step": 144840, "epoch": 1627, "lr": 4.4869103316709165e-05}
{"train_loss": 0.049441561102867126, "global_step": 144841, "epoch": 1627, "lr": 4.4868526640970035e-05}
{"train_loss": 0.061850860714912415, "global_step": 144842, "epoch": 1627, "lr": 4.486794996592078e-05}
{"train_loss": 0.11162784695625305, "global_step": 144843, "epoch": 1627, "lr": 4.4867373291561456e-05}
{"train_loss": 0.09041614830493927, "global_step": 144844, "epoch": 1627, "lr": 4.486679661789216e-05}
{"train_loss": 0.04212887957692146, "global_step": 144845, "epoch": 1627, "lr": 4.486621994491295e-05}
{"train_loss": 0.06821718066930771, "global_step": 144846, "epoch": 1627, "lr": 4.486564327262393e-05}
{"train_loss": 0.12007797509431839, "global_step": 144847, "epoch": 1627, "lr": 4.486506660102515e-05}
{"train_loss": 0.13306856155395508, "global_step": 144848, "epoch": 1627, "lr": 4.486448993011669e-05}
{"train_loss": 0.12611635029315948, "global_step": 144849, "epoch": 1627, "lr": 4.4863913259898654e-05}
{"train_loss": 0.14804717898368835, "global_step": 144850, "epoch": 1627, "lr": 4.486333659037109e-05}
{"train_loss": 0.12668071687221527, "global_step": 144851, "epoch": 1627, "lr": 4.4862759921534104e-05}
{"train_loss": 0.06354229152202606, "global_step": 144852, "epoch": 1627, "lr": 4.4862183253387744e-05}
{"train_loss": 0.09952761232852936, "global_step": 144853, "epoch": 1627, "lr": 4.486160658593211e-05}
{"train_loss": 0.1238655149936676, "global_step": 144854, "epoch": 1627, "lr": 4.4861029919167265e-05}
{"train_loss": 0.10980338603258133, "global_step": 144855, "epoch": 1627, "lr": 4.4860453253093296e-05}
{"train_loss": 0.18710175156593323, "global_step": 144856, "epoch": 1627, "lr": 4.485987658771027e-05}
{"train_loss": 0.11124924570322037, "global_step": 144857, "epoch": 1627, "lr": 4.4859299923018284e-05}
{"train_loss": 0.08813945949077606, "global_step": 144858, "epoch": 1627, "lr": 4.4858723259017384e-05}
{"train_loss": 0.08688106387853622, "global_step": 144859, "epoch": 1627, "lr": 4.485814659570768e-05}
{"train_loss": 0.0538683645427227, "global_step": 144860, "epoch": 1627, "lr": 4.485756993308922e-05}
{"train_loss": 0.07389283180236816, "global_step": 144861, "epoch": 1627, "lr": 4.4856993271162116e-05}
{"train_loss": 0.26247769594192505, "global_step": 144862, "epoch": 1627, "lr": 4.485641660992641e-05}
{"train_loss": 0.12960286438465118, "global_step": 144863, "epoch": 1627, "lr": 4.4855839949382213e-05}
{"train_loss": 0.07642339915037155, "global_step": 144864, "epoch": 1627, "lr": 4.485526328952957e-05}
{"train_loss": 0.06321054697036743, "global_step": 144865, "epoch": 1627, "lr": 4.4854686630368565e-05}
{"train_loss": 0.06814121454954147, "global_step": 144866, "epoch": 1627, "lr": 4.485410997189931e-05}
{"train_loss": 0.10499610006809235, "global_step": 144867, "epoch": 1627, "lr": 4.485353331412184e-05}
{"train_loss": 0.04758699983358383, "global_step": 144868, "epoch": 1627, "lr": 4.4852956657036266e-05}
{"train_loss": 0.22279620170593262, "global_step": 144869, "epoch": 1627, "lr": 4.485238000064263e-05}
{"train_loss": 0.11456424742937088, "global_step": 144870, "epoch": 1627, "lr": 4.4851803344941054e-05}
{"train_loss": 0.1483665108680725, "global_step": 144871, "epoch": 1627, "lr": 4.485122668993157e-05}
{"train_loss": 0.10421948879957199, "global_step": 144872, "epoch": 1627, "lr": 4.4850650035614284e-05}
{"train_loss": 0.12437906861305237, "global_step": 144873, "epoch": 1627, "lr": 4.485007338198926e-05}
{"train_loss": 0.12563595175743103, "global_step": 144874, "epoch": 1627, "lr": 4.4849496729056586e-05}
{"train_loss": 0.15281184017658234, "global_step": 144875, "epoch": 1627, "lr": 4.484892007681633e-05}
{"train_loss": 0.07280486077070236, "global_step": 144876, "epoch": 1627, "lr": 4.4848343425268585e-05}
{"train_loss": 0.08957232534885406, "global_step": 144877, "epoch": 1627, "lr": 4.48477667744134e-05}
{"train_loss": 0.06711713969707489, "global_step": 144878, "epoch": 1627, "lr": 4.484719012425089e-05}
{"train_loss": 0.10824885964393616, "global_step": 144879, "epoch": 1627, "lr": 4.4846613474781084e-05}
{"train_loss": 0.09176710247993469, "global_step": 144880, "epoch": 1627, "lr": 4.484603682600412e-05}
{"train_loss": 0.04233311489224434, "global_step": 144881, "epoch": 1627, "lr": 4.484546017792001e-05}
{"train_loss": 0.07379917800426483, "global_step": 144882, "epoch": 1627, "lr": 4.48448835305289e-05}
{"train_loss": 0.0674348697066307, "global_step": 144883, "epoch": 1627, "lr": 4.48443068838308e-05}
{"train_loss": 0.09091086685657501, "global_step": 144884, "epoch": 1627, "lr": 4.4843730237825826e-05}
{"train_loss": 0.10443899035453796, "global_step": 144885, "epoch": 1627, "lr": 4.484315359251407e-05}
{"train_loss": 0.10317998379468918, "global_step": 144886, "epoch": 1627, "lr": 4.484257694789557e-05}
{"train_loss": 0.09465579688549042, "global_step": 144887, "epoch": 1627, "lr": 4.484200030397044e-05}
{"train_loss": 0.16652442514896393, "global_step": 144888, "epoch": 1627, "lr": 4.484142366073873e-05}
{"train_loss": 0.12644565105438232, "global_step": 144889, "epoch": 1627, "lr": 4.484084701820053e-05}
{"train_loss": 0.16709105670452118, "global_step": 144890, "epoch": 1627, "lr": 4.484027037635591e-05}
{"train_loss": 0.10336369627647185, "global_step": 144891, "epoch": 1627, "lr": 4.4839693735204965e-05, "val_loss": 6.220203876495361}
{"train_loss": 0.06933432072401047, "global_step": 144892, "epoch": 1628, "lr": 4.483911709474775e-05}
{"train_loss": 0.14241552352905273, "global_step": 144893, "epoch": 1628, "lr": 4.4838540454984355e-05}
{"train_loss": 0.11254338920116425, "global_step": 144894, "epoch": 1628, "lr": 4.483796381591486e-05}
{"train_loss": 0.061248283833265305, "global_step": 144895, "epoch": 1628, "lr": 4.483738717753934e-05}
{"train_loss": 0.1537935882806778, "global_step": 144896, "epoch": 1628, "lr": 4.4836810539857856e-05}
{"train_loss": 0.060624510049819946, "global_step": 144897, "epoch": 1628, "lr": 4.483623390287052e-05}
{"train_loss": 0.10454311966896057, "global_step": 144898, "epoch": 1628, "lr": 4.483565726657736e-05}
{"train_loss": 0.09130457043647766, "global_step": 144899, "epoch": 1628, "lr": 4.483508063097851e-05}
{"train_loss": 0.16228702664375305, "global_step": 144900, "epoch": 1628, "lr": 4.4834503996074e-05}
{"train_loss": 0.08496490120887756, "global_step": 144901, "epoch": 1628, "lr": 4.4833927361863925e-05}
{"train_loss": 0.1151723638176918, "global_step": 144902, "epoch": 1628, "lr": 4.483335072834839e-05}
{"train_loss": 0.08517375588417053, "global_step": 144903, "epoch": 1628, "lr": 4.483277409552743e-05}
{"train_loss": 0.08840849250555038, "global_step": 144904, "epoch": 1628, "lr": 4.483219746340116e-05}
{"train_loss": 0.08857164531946182, "global_step": 144905, "epoch": 1628, "lr": 4.483162083196961e-05}
{"train_loss": 0.13328377902507782, "global_step": 144906, "epoch": 1628, "lr": 4.4831044201232916e-05}
{"train_loss": 0.11139516532421112, "global_step": 144907, "epoch": 1628, "lr": 4.48304675711911e-05}
{"train_loss": 0.16964253783226013, "global_step": 144908, "epoch": 1628, "lr": 4.482989094184428e-05}
{"train_loss": 0.07873459905385971, "global_step": 144909, "epoch": 1628, "lr": 4.482931431319252e-05}
{"train_loss": 0.07312324643135071, "global_step": 144910, "epoch": 1628, "lr": 4.4828737685235886e-05}
{"train_loss": 0.10696966201066971, "global_step": 144911, "epoch": 1628, "lr": 4.4828161057974464e-05}
{"train_loss": 0.08927837014198303, "global_step": 144912, "epoch": 1628, "lr": 4.4827584431408334e-05}
{"train_loss": 0.14853593707084656, "global_step": 144913, "epoch": 1628, "lr": 4.482700780553758e-05}
{"train_loss": 0.0771930143237114, "global_step": 144914, "epoch": 1628, "lr": 4.4826431180362276e-05}
{"train_loss": 0.07002653926610947, "global_step": 144915, "epoch": 1628, "lr": 4.4825854555882476e-05}
{"train_loss": 0.10673339664936066, "global_step": 144916, "epoch": 1628, "lr": 4.4825277932098294e-05}
{"train_loss": 0.18564030528068542, "global_step": 144917, "epoch": 1628, "lr": 4.4824701309009776e-05}
{"train_loss": 0.07854923605918884, "global_step": 144918, "epoch": 1628, "lr": 4.482412468661703e-05}
{"train_loss": 0.1178300529718399, "global_step": 144919, "epoch": 1628, "lr": 4.4823548064920095e-05}
{"train_loss": 0.1168980747461319, "global_step": 144920, "epoch": 1628, "lr": 4.482297144391907e-05}
{"train_loss": 0.0902852937579155, "global_step": 144921, "epoch": 1628, "lr": 4.482239482361406e-05}
{"train_loss": 0.08828576654195786, "global_step": 144922, "epoch": 1628, "lr": 4.4821818204005094e-05}
{"train_loss": 0.09615439921617508, "global_step": 144923, "epoch": 1628, "lr": 4.4821241585092294e-05}
{"train_loss": 0.10392840206623077, "global_step": 144924, "epoch": 1628, "lr": 4.482066496687569e-05}
{"train_loss": 0.08889762312173843, "global_step": 144925, "epoch": 1628, "lr": 4.482008834935541e-05}
{"train_loss": 0.12229751795530319, "global_step": 144926, "epoch": 1628, "lr": 4.481951173253148e-05}
{"train_loss": 0.054913125932216644, "global_step": 144927, "epoch": 1628, "lr": 4.4818935116404026e-05}
{"train_loss": 0.07362048327922821, "global_step": 144928, "epoch": 1628, "lr": 4.481835850097308e-05}
{"train_loss": 0.10231122374534607, "global_step": 144929, "epoch": 1628, "lr": 4.4817781886238766e-05}
{"train_loss": 0.11755306273698807, "global_step": 144930, "epoch": 1628, "lr": 4.481720527220112e-05}
{"train_loss": 0.09780848771333694, "global_step": 144931, "epoch": 1628, "lr": 4.481662865886025e-05}
{"train_loss": 0.16606201231479645, "global_step": 144932, "epoch": 1628, "lr": 4.481605204621622e-05}
{"train_loss": 0.14489078521728516, "global_step": 144933, "epoch": 1628, "lr": 4.48154754342691e-05}
{"train_loss": 0.12686140835285187, "global_step": 144934, "epoch": 1628, "lr": 4.481489882301898e-05}
{"train_loss": 0.11062850803136826, "global_step": 144935, "epoch": 1628, "lr": 4.4814322212465945e-05}
{"train_loss": 0.16309726238250732, "global_step": 144936, "epoch": 1628, "lr": 4.481374560261004e-05}
{"train_loss": 0.08429718017578125, "global_step": 144937, "epoch": 1628, "lr": 4.4813168993451366e-05}
{"train_loss": 0.12370999902486801, "global_step": 144938, "epoch": 1628, "lr": 4.481259238499002e-05}
{"train_loss": 0.13273662328720093, "global_step": 144939, "epoch": 1628, "lr": 4.481201577722603e-05}
{"train_loss": 0.06700261682271957, "global_step": 144940, "epoch": 1628, "lr": 4.481143917015952e-05}
{"train_loss": 0.12085457891225815, "global_step": 144941, "epoch": 1628, "lr": 4.481086256379054e-05}
{"train_loss": 0.04468652233481407, "global_step": 144942, "epoch": 1628, "lr": 4.481028595811919e-05}
{"train_loss": 0.11799506843090057, "global_step": 144943, "epoch": 1628, "lr": 4.4809709353145515e-05}
{"train_loss": 0.10528174787759781, "global_step": 144944, "epoch": 1628, "lr": 4.4809132748869633e-05}
{"train_loss": 0.12012845277786255, "global_step": 144945, "epoch": 1628, "lr": 4.480855614529158e-05}
{"train_loss": 0.12707509100437164, "global_step": 144946, "epoch": 1628, "lr": 4.4807979542411464e-05}
{"train_loss": 0.0970737561583519, "global_step": 144947, "epoch": 1628, "lr": 4.480740294022934e-05}
{"train_loss": 0.06618574261665344, "global_step": 144948, "epoch": 1628, "lr": 4.480682633874532e-05}
{"train_loss": 0.0743250623345375, "global_step": 144949, "epoch": 1628, "lr": 4.4806249737959436e-05}
{"train_loss": 0.09232550114393234, "global_step": 144950, "epoch": 1628, "lr": 4.48056731378718e-05}
{"train_loss": 0.2576618194580078, "global_step": 144951, "epoch": 1628, "lr": 4.480509653848248e-05}
{"train_loss": 0.13073907792568207, "global_step": 144952, "epoch": 1628, "lr": 4.4804519939791545e-05}
{"train_loss": 0.09676260501146317, "global_step": 144953, "epoch": 1628, "lr": 4.480394334179908e-05}
{"train_loss": 0.10752220451831818, "global_step": 144954, "epoch": 1628, "lr": 4.480336674450517e-05}
{"train_loss": 0.1289338767528534, "global_step": 144955, "epoch": 1628, "lr": 4.4802790147909866e-05}
{"train_loss": 0.11520890146493912, "global_step": 144956, "epoch": 1628, "lr": 4.480221355201326e-05}
{"train_loss": 0.07119454443454742, "global_step": 144957, "epoch": 1628, "lr": 4.4801636956815464e-05}
{"train_loss": 0.08741778135299683, "global_step": 144958, "epoch": 1628, "lr": 4.480106036231649e-05}
{"train_loss": 0.07774714380502701, "global_step": 144959, "epoch": 1628, "lr": 4.480048376851648e-05}
{"train_loss": 0.0812256932258606, "global_step": 144960, "epoch": 1628, "lr": 4.479990717541546e-05}
{"train_loss": 0.12817434966564178, "global_step": 144961, "epoch": 1628, "lr": 4.479933058301355e-05}
{"train_loss": 0.19327028095722198, "global_step": 144962, "epoch": 1628, "lr": 4.479875399131078e-05}
{"train_loss": 0.1092166006565094, "global_step": 144963, "epoch": 1628, "lr": 4.479817740030728e-05}
{"train_loss": 0.11179131269454956, "global_step": 144964, "epoch": 1628, "lr": 4.479760081000308e-05}
{"train_loss": 0.10025037080049515, "global_step": 144965, "epoch": 1628, "lr": 4.47970242203983e-05}
{"train_loss": 0.11769290268421173, "global_step": 144966, "epoch": 1628, "lr": 4.479644763149298e-05}
{"train_loss": 0.08437100797891617, "global_step": 144967, "epoch": 1628, "lr": 4.479587104328723e-05}
{"train_loss": 0.08928012102842331, "global_step": 144968, "epoch": 1628, "lr": 4.47952944557811e-05}
{"train_loss": 0.10481012612581253, "global_step": 144969, "epoch": 1628, "lr": 4.479471786897468e-05}
{"train_loss": 0.14037930965423584, "global_step": 144970, "epoch": 1628, "lr": 4.479414128286805e-05}
{"train_loss": 0.1022750660777092, "global_step": 144971, "epoch": 1628, "lr": 4.479356469746128e-05}
{"train_loss": 0.0883103609085083, "global_step": 144972, "epoch": 1628, "lr": 4.4792988112754454e-05}
{"train_loss": 0.06656316667795181, "global_step": 144973, "epoch": 1628, "lr": 4.479241152874766e-05}
{"train_loss": 0.11262646317481995, "global_step": 144974, "epoch": 1628, "lr": 4.4791834945440944e-05}
{"train_loss": 0.08593039214611053, "global_step": 144975, "epoch": 1628, "lr": 4.4791258362834394e-05}
{"train_loss": 0.09234557300806046, "global_step": 144976, "epoch": 1628, "lr": 4.479068178092812e-05}
{"train_loss": 0.10423041135072708, "global_step": 144977, "epoch": 1628, "lr": 4.479010519972215e-05}
{"train_loss": 0.06771223992109299, "global_step": 144978, "epoch": 1628, "lr": 4.478952861921662e-05}
{"train_loss": 0.08132803440093994, "global_step": 144979, "epoch": 1628, "lr": 4.478895203941154e-05}
{"train_loss": 0.10678217459595606, "global_step": 144980, "epoch": 1628, "lr": 4.478837546030705e-05, "val_loss": 6.176540851593018}
{"train_loss": 0.11644498258829117, "global_step": 144981, "epoch": 1629, "lr": 4.478779888190318e-05}
{"train_loss": 0.09722557663917542, "global_step": 144982, "epoch": 1629, "lr": 4.478722230420005e-05}
{"train_loss": 0.1390911489725113, "global_step": 144983, "epoch": 1629, "lr": 4.4786645727197686e-05}
{"train_loss": 0.13697347044944763, "global_step": 144984, "epoch": 1629, "lr": 4.478606915089621e-05}
{"train_loss": 0.1392187774181366, "global_step": 144985, "epoch": 1629, "lr": 4.478549257529568e-05}
{"train_loss": 0.19018150866031647, "global_step": 144986, "epoch": 1629, "lr": 4.478491600039619e-05}
{"train_loss": 0.09009113162755966, "global_step": 144987, "epoch": 1629, "lr": 4.4784339426197784e-05}
{"train_loss": 0.09183380007743835, "global_step": 144988, "epoch": 1629, "lr": 4.478376285270057e-05}
{"train_loss": 0.06159641221165657, "global_step": 144989, "epoch": 1629, "lr": 4.4783186279904616e-05}
{"train_loss": 0.09526604413986206, "global_step": 144990, "epoch": 1629, "lr": 4.4782609707809996e-05}
{"train_loss": 0.12603957951068878, "global_step": 144991, "epoch": 1629, "lr": 4.478203313641679e-05}
{"train_loss": 0.2076110541820526, "global_step": 144992, "epoch": 1629, "lr": 4.478145656572506e-05}
{"train_loss": 0.10966730117797852, "global_step": 144993, "epoch": 1629, "lr": 4.4780879995734934e-05}
{"train_loss": 0.1713375300168991, "global_step": 144994, "epoch": 1629, "lr": 4.478030342644643e-05}
{"train_loss": 0.08819585293531418, "global_step": 144995, "epoch": 1629, "lr": 4.4779726857859674e-05}
{"train_loss": 0.07065942883491516, "global_step": 144996, "epoch": 1629, "lr": 4.477915028997469e-05}
{"train_loss": 0.07502922415733337, "global_step": 144997, "epoch": 1629, "lr": 4.477857372279161e-05}
{"train_loss": 0.18827538192272186, "global_step": 144998, "epoch": 1629, "lr": 4.4777997156310466e-05}
{"train_loss": 0.06747959554195404, "global_step": 144999, "epoch": 1629, "lr": 4.4777420590531385e-05}
{"train_loss": 0.1542132943868637, "global_step": 145000, "epoch": 1629, "lr": 4.4776844025454385e-05}
{"train_loss": 0.1287773996591568, "global_step": 145001, "epoch": 1629, "lr": 4.47762674610796e-05}
{"train_loss": 0.11022080481052399, "global_step": 145002, "epoch": 1629, "lr": 4.477569089740706e-05}
{"train_loss": 0.14774450659751892, "global_step": 145003, "epoch": 1629, "lr": 4.477511433443689e-05}
{"train_loss": 0.10498474538326263, "global_step": 145004, "epoch": 1629, "lr": 4.477453777216911e-05}
{"train_loss": 0.16035477817058563, "global_step": 145005, "epoch": 1629, "lr": 4.477396121060386e-05}
{"train_loss": 0.1081208735704422, "global_step": 145006, "epoch": 1629, "lr": 4.4773384649741176e-05}
{"train_loss": 0.15045765042304993, "global_step": 145007, "epoch": 1629, "lr": 4.477280808958115e-05}
{"train_loss": 0.12179391086101532, "global_step": 145008, "epoch": 1629, "lr": 4.477223153012385e-05}
{"train_loss": 0.12695996463298798, "global_step": 145009, "epoch": 1629, "lr": 4.477165497136936e-05}
{"train_loss": 0.11348418891429901, "global_step": 145010, "epoch": 1629, "lr": 4.477107841331776e-05}
{"train_loss": 0.14207172393798828, "global_step": 145011, "epoch": 1629, "lr": 4.477050185596911e-05}
{"train_loss": 0.1118423342704773, "global_step": 145012, "epoch": 1629, "lr": 4.476992529932353e-05}
{"train_loss": 0.1358325481414795, "global_step": 145013, "epoch": 1629, "lr": 4.476934874338104e-05}
{"train_loss": 0.1375006139278412, "global_step": 145014, "epoch": 1629, "lr": 4.476877218814178e-05}
{"train_loss": 0.08095297962427139, "global_step": 145015, "epoch": 1629, "lr": 4.4768195633605767e-05}
{"train_loss": 0.10303451865911484, "global_step": 145016, "epoch": 1629, "lr": 4.476761907977313e-05}
{"train_loss": 0.10681721568107605, "global_step": 145017, "epoch": 1629, "lr": 4.4767042526643895e-05}
{"train_loss": 0.14101305603981018, "global_step": 145018, "epoch": 1629, "lr": 4.4766465974218193e-05}
{"train_loss": 0.1586434692144394, "global_step": 145019, "epoch": 1629, "lr": 4.476588942249606e-05}
{"train_loss": 0.08834283798933029, "global_step": 145020, "epoch": 1629, "lr": 4.47653128714776e-05}
{"train_loss": 0.11847606301307678, "global_step": 145021, "epoch": 1629, "lr": 4.476473632116286e-05}
{"train_loss": 0.11671003699302673, "global_step": 145022, "epoch": 1629, "lr": 4.4764159771551966e-05}
{"train_loss": 0.15400554239749908, "global_step": 145023, "epoch": 1629, "lr": 4.476358322264494e-05}
{"train_loss": 0.14903931319713593, "global_step": 145024, "epoch": 1629, "lr": 4.476300667444191e-05}
{"train_loss": 0.07686509937047958, "global_step": 145025, "epoch": 1629, "lr": 4.476243012694292e-05}
{"train_loss": 0.07165713608264923, "global_step": 145026, "epoch": 1629, "lr": 4.476185358014806e-05}
{"train_loss": 0.0867132917046547, "global_step": 145027, "epoch": 1629, "lr": 4.47612770340574e-05}
{"train_loss": 0.09848930686712265, "global_step": 145028, "epoch": 1629, "lr": 4.476070048867101e-05}
{"train_loss": 0.12957143783569336, "global_step": 145029, "epoch": 1629, "lr": 4.476012394398901e-05}
{"train_loss": 0.12002264708280563, "global_step": 145030, "epoch": 1629, "lr": 4.475954740001142e-05}
{"train_loss": 0.16217561066150665, "global_step": 145031, "epoch": 1629, "lr": 4.475897085673837e-05}
{"train_loss": 0.07618514448404312, "global_step": 145032, "epoch": 1629, "lr": 4.4758394314169885e-05}
{"train_loss": 0.17442190647125244, "global_step": 145033, "epoch": 1629, "lr": 4.4757817772306095e-05}
{"train_loss": 0.10804267227649689, "global_step": 145034, "epoch": 1629, "lr": 4.475724123114703e-05}
{"train_loss": 0.1429683417081833, "global_step": 145035, "epoch": 1629, "lr": 4.475666469069281e-05}
{"train_loss": 0.12938708066940308, "global_step": 145036, "epoch": 1629, "lr": 4.475608815094347e-05}
{"train_loss": 0.11093084514141083, "global_step": 145037, "epoch": 1629, "lr": 4.4755511611899134e-05}
{"train_loss": 0.13998809456825256, "global_step": 145038, "epoch": 1629, "lr": 4.475493507355984e-05}
{"train_loss": 0.1155230775475502, "global_step": 145039, "epoch": 1629, "lr": 4.475435853592569e-05}
{"train_loss": 0.09897347539663315, "global_step": 145040, "epoch": 1629, "lr": 4.475378199899674e-05}
{"train_loss": 0.10992268472909927, "global_step": 145041, "epoch": 1629, "lr": 4.47532054627731e-05}
{"train_loss": 0.15437018871307373, "global_step": 145042, "epoch": 1629, "lr": 4.4752628927254806e-05}
{"train_loss": 0.0944419726729393, "global_step": 145043, "epoch": 1629, "lr": 4.475205239244198e-05}
{"train_loss": 0.08719722181558609, "global_step": 145044, "epoch": 1629, "lr": 4.4751475858334655e-05}
{"train_loss": 0.12212736904621124, "global_step": 145045, "epoch": 1629, "lr": 4.475089932493295e-05}
{"train_loss": 0.0892806276679039, "global_step": 145046, "epoch": 1629, "lr": 4.475032279223691e-05}
{"train_loss": 0.09251991659402847, "global_step": 145047, "epoch": 1629, "lr": 4.474974626024662e-05}
{"train_loss": 0.10108364373445511, "global_step": 145048, "epoch": 1629, "lr": 4.474916972896218e-05}
{"train_loss": 0.07141881436109543, "global_step": 145049, "epoch": 1629, "lr": 4.474859319838363e-05}
{"train_loss": 0.13921651244163513, "global_step": 145050, "epoch": 1629, "lr": 4.474801666851109e-05}
{"train_loss": 0.08801886439323425, "global_step": 145051, "epoch": 1629, "lr": 4.4747440139344584e-05}
{"train_loss": 0.11476963013410568, "global_step": 145052, "epoch": 1629, "lr": 4.474686361088426e-05}
{"train_loss": 0.11298533529043198, "global_step": 145053, "epoch": 1629, "lr": 4.4746287083130124e-05}
{"train_loss": 0.11382058262825012, "global_step": 145054, "epoch": 1629, "lr": 4.47457105560823e-05}
{"train_loss": 0.08071984350681305, "global_step": 145055, "epoch": 1629, "lr": 4.4745134029740844e-05}
{"train_loss": 0.10679131746292114, "global_step": 145056, "epoch": 1629, "lr": 4.474455750410585e-05}
{"train_loss": 0.09704292565584183, "global_step": 145057, "epoch": 1629, "lr": 4.4743980979177374e-05}
{"train_loss": 0.09532243758440018, "global_step": 145058, "epoch": 1629, "lr": 4.4743404454955525e-05}
{"train_loss": 0.10216107219457626, "global_step": 145059, "epoch": 1629, "lr": 4.474282793144034e-05}
{"train_loss": 0.12058126926422119, "global_step": 145060, "epoch": 1629, "lr": 4.4742251408631936e-05}
{"train_loss": 0.13261309266090393, "global_step": 145061, "epoch": 1629, "lr": 4.474167488653035e-05}
{"train_loss": 0.13209979236125946, "global_step": 145062, "epoch": 1629, "lr": 4.474109836513571e-05}
{"train_loss": 0.1112324446439743, "global_step": 145063, "epoch": 1629, "lr": 4.4740521844448034e-05}
{"train_loss": 0.10510479658842087, "global_step": 145064, "epoch": 1629, "lr": 4.473994532446745e-05}
{"train_loss": 0.09255479276180267, "global_step": 145065, "epoch": 1629, "lr": 4.4739368805194026e-05}
{"train_loss": 0.1337292641401291, "global_step": 145066, "epoch": 1629, "lr": 4.473879228662781e-05}
{"train_loss": 0.15984536707401276, "global_step": 145067, "epoch": 1629, "lr": 4.473821576876891e-05}
{"train_loss": 0.08010443300008774, "global_step": 145068, "epoch": 1629, "lr": 4.473763925161738e-05}
{"train_loss": 0.11701057579242781, "global_step": 145069, "epoch": 1629, "lr": 4.473706273517333e-05, "val_loss": 6.199300289154053}
{"train_loss": 0.0764475017786026, "global_step": 145070, "epoch": 1630, "lr": 4.4736486219436797e-05}
{"train_loss": 0.11382684111595154, "global_step": 145071, "epoch": 1630, "lr": 4.47359097044079e-05}
{"train_loss": 0.13896584510803223, "global_step": 145072, "epoch": 1630, "lr": 4.473533319008667e-05}
{"train_loss": 0.07804682105779648, "global_step": 145073, "epoch": 1630, "lr": 4.473475667647324e-05}
{"train_loss": 0.13450992107391357, "global_step": 145074, "epoch": 1630, "lr": 4.473418016356763e-05}
{"train_loss": 0.11331664770841599, "global_step": 145075, "epoch": 1630, "lr": 4.473360365136997e-05}
{"train_loss": 0.12814687192440033, "global_step": 145076, "epoch": 1630, "lr": 4.473302713988029e-05}
{"train_loss": 0.10653045773506165, "global_step": 145077, "epoch": 1630, "lr": 4.473245062909871e-05}
{"train_loss": 0.09871038049459457, "global_step": 145078, "epoch": 1630, "lr": 4.4731874119025266e-05}
{"train_loss": 0.11234330385923386, "global_step": 145079, "epoch": 1630, "lr": 4.473129760966008e-05}
{"train_loss": 0.07325214147567749, "global_step": 145080, "epoch": 1630, "lr": 4.473072110100319e-05}
{"train_loss": 0.19147557020187378, "global_step": 145081, "epoch": 1630, "lr": 4.473014459305471e-05}
{"train_loss": 0.0784863606095314, "global_step": 145082, "epoch": 1630, "lr": 4.4729568085814675e-05}
{"train_loss": 0.07387762516736984, "global_step": 145083, "epoch": 1630, "lr": 4.47289915792832e-05}
{"train_loss": 0.11796797066926956, "global_step": 145084, "epoch": 1630, "lr": 4.472841507346035e-05}
{"train_loss": 0.05466318503022194, "global_step": 145085, "epoch": 1630, "lr": 4.472783856834618e-05}
{"train_loss": 0.10602568835020065, "global_step": 145086, "epoch": 1630, "lr": 4.4727262063940814e-05}
{"train_loss": 0.07339825481176376, "global_step": 145087, "epoch": 1630, "lr": 4.472668556024428e-05}
{"train_loss": 0.12944746017456055, "global_step": 145088, "epoch": 1630, "lr": 4.4726109057256706e-05}
{"train_loss": 0.08426529169082642, "global_step": 145089, "epoch": 1630, "lr": 4.472553255497811e-05}
{"train_loss": 0.1297035664319992, "global_step": 145090, "epoch": 1630, "lr": 4.472495605340863e-05}
{"train_loss": 0.11124636977910995, "global_step": 145091, "epoch": 1630, "lr": 4.472437955254829e-05}
{"train_loss": 0.1143273413181305, "global_step": 145092, "epoch": 1630, "lr": 4.472380305239722e-05}
{"train_loss": 0.11763546615839005, "global_step": 145093, "epoch": 1630, "lr": 4.4723226552955444e-05}
{"train_loss": 0.07436349242925644, "global_step": 145094, "epoch": 1630, "lr": 4.472265005422309e-05}
{"train_loss": 0.09671378135681152, "global_step": 145095, "epoch": 1630, "lr": 4.4722073556200196e-05}
{"train_loss": 0.09501087665557861, "global_step": 145096, "epoch": 1630, "lr": 4.472149705888686e-05}
{"train_loss": 0.08303055167198181, "global_step": 145097, "epoch": 1630, "lr": 4.4720920562283146e-05}
{"train_loss": 0.08409970253705978, "global_step": 145098, "epoch": 1630, "lr": 4.472034406638916e-05}
{"train_loss": 0.11802633106708527, "global_step": 145099, "epoch": 1630, "lr": 4.4719767571204936e-05}
{"train_loss": 0.1050441563129425, "global_step": 145100, "epoch": 1630, "lr": 4.47191910767306e-05}
{"train_loss": 0.1389913558959961, "global_step": 145101, "epoch": 1630, "lr": 4.471861458296618e-05}
{"train_loss": 0.14610429108142853, "global_step": 145102, "epoch": 1630, "lr": 4.471803808991179e-05}
{"train_loss": 0.0656273365020752, "global_step": 145103, "epoch": 1630, "lr": 4.47174615975675e-05}
{"train_loss": 0.11313402652740479, "global_step": 145104, "epoch": 1630, "lr": 4.471688510593337e-05}
{"train_loss": 0.10411392897367477, "global_step": 145105, "epoch": 1630, "lr": 4.47163086150095e-05}
{"train_loss": 0.14849522709846497, "global_step": 145106, "epoch": 1630, "lr": 4.471573212479594e-05}
{"train_loss": 0.12823569774627686, "global_step": 145107, "epoch": 1630, "lr": 4.4715155635292815e-05}
{"train_loss": 0.1483580321073532, "global_step": 145108, "epoch": 1630, "lr": 4.4714579146500145e-05}
{"train_loss": 0.12526439130306244, "global_step": 145109, "epoch": 1630, "lr": 4.471400265841806e-05}
{"train_loss": 0.08248045295476913, "global_step": 145110, "epoch": 1630, "lr": 4.471342617104659e-05}
{"train_loss": 0.08681079000234604, "global_step": 145111, "epoch": 1630, "lr": 4.471284968438586e-05}
{"train_loss": 0.06681465357542038, "global_step": 145112, "epoch": 1630, "lr": 4.4712273198435894e-05}
{"train_loss": 0.1329164057970047, "global_step": 145113, "epoch": 1630, "lr": 4.4711696713196825e-05}
{"train_loss": 0.09312256425619125, "global_step": 145114, "epoch": 1630, "lr": 4.471112022866869e-05}
{"train_loss": 0.1102859377861023, "global_step": 145115, "epoch": 1630, "lr": 4.471054374485159e-05}
{"train_loss": 0.09806211292743683, "global_step": 145116, "epoch": 1630, "lr": 4.4709967261745575e-05}
{"train_loss": 0.11505063623189926, "global_step": 145117, "epoch": 1630, "lr": 4.470939077935076e-05}
{"train_loss": 0.1632702797651291, "global_step": 145118, "epoch": 1630, "lr": 4.4708814297667184e-05}
{"train_loss": 0.11702173203229904, "global_step": 145119, "epoch": 1630, "lr": 4.470823781669495e-05}
{"train_loss": 0.11123389005661011, "global_step": 145120, "epoch": 1630, "lr": 4.470766133643414e-05}
{"train_loss": 0.13186600804328918, "global_step": 145121, "epoch": 1630, "lr": 4.4707084856884815e-05}
{"train_loss": 0.09978437423706055, "global_step": 145122, "epoch": 1630, "lr": 4.4706508378047066e-05}
{"train_loss": 0.12533706426620483, "global_step": 145123, "epoch": 1630, "lr": 4.470593189992094e-05}
{"train_loss": 0.1282849907875061, "global_step": 145124, "epoch": 1630, "lr": 4.470535542250656e-05}
{"train_loss": 0.13231468200683594, "global_step": 145125, "epoch": 1630, "lr": 4.470477894580396e-05}
{"train_loss": 0.08722436428070068, "global_step": 145126, "epoch": 1630, "lr": 4.470420246981326e-05}
{"train_loss": 0.13946394622325897, "global_step": 145127, "epoch": 1630, "lr": 4.4703625994534484e-05}
{"train_loss": 0.12468927353620529, "global_step": 145128, "epoch": 1630, "lr": 4.470304951996778e-05}
{"train_loss": 0.07273046672344208, "global_step": 145129, "epoch": 1630, "lr": 4.470247304611315e-05}
{"train_loss": 0.12342964857816696, "global_step": 145130, "epoch": 1630, "lr": 4.4701896572970737e-05}
{"train_loss": 0.17150826752185822, "global_step": 145131, "epoch": 1630, "lr": 4.470132010054057e-05}
{"train_loss": 0.07960063219070435, "global_step": 145132, "epoch": 1630, "lr": 4.470074362882276e-05}
{"train_loss": 0.11658815294504166, "global_step": 145133, "epoch": 1630, "lr": 4.4700167157817355e-05}
{"train_loss": 0.18551406264305115, "global_step": 145134, "epoch": 1630, "lr": 4.469959068752447e-05}
{"train_loss": 0.11330989748239517, "global_step": 145135, "epoch": 1630, "lr": 4.469901421794413e-05}
{"train_loss": 0.0946306362748146, "global_step": 145136, "epoch": 1630, "lr": 4.4698437749076474e-05}
{"train_loss": 0.051618944853544235, "global_step": 145137, "epoch": 1630, "lr": 4.4697861280921524e-05}
{"train_loss": 0.12249454855918884, "global_step": 145138, "epoch": 1630, "lr": 4.469728481347939e-05}
{"train_loss": 0.1055489331483841, "global_step": 145139, "epoch": 1630, "lr": 4.4696708346750146e-05}
{"train_loss": 0.04230625927448273, "global_step": 145140, "epoch": 1630, "lr": 4.469613188073386e-05}
{"train_loss": 0.12967638671398163, "global_step": 145141, "epoch": 1630, "lr": 4.4695555415430626e-05}
{"train_loss": 0.09955436736345291, "global_step": 145142, "epoch": 1630, "lr": 4.469497895084049e-05}
{"train_loss": 0.11210982501506805, "global_step": 145143, "epoch": 1630, "lr": 4.4694402486963574e-05}
{"train_loss": 0.09287868440151215, "global_step": 145144, "epoch": 1630, "lr": 4.4693826023799905e-05}
{"train_loss": 0.1335655152797699, "global_step": 145145, "epoch": 1630, "lr": 4.469324956134961e-05}
{"train_loss": 0.05562904477119446, "global_step": 145146, "epoch": 1630, "lr": 4.469267309961272e-05}
{"train_loss": 0.12971773743629456, "global_step": 145147, "epoch": 1630, "lr": 4.469209663858936e-05}
{"train_loss": 0.15483401715755463, "global_step": 145148, "epoch": 1630, "lr": 4.4691520178279556e-05}
{"train_loss": 0.09376725554466248, "global_step": 145149, "epoch": 1630, "lr": 4.4690943718683434e-05}
{"train_loss": 0.14111797511577606, "global_step": 145150, "epoch": 1630, "lr": 4.4690367259801035e-05}
{"train_loss": 0.10749215632677078, "global_step": 145151, "epoch": 1630, "lr": 4.468979080163247e-05}
{"train_loss": 0.16707487404346466, "global_step": 145152, "epoch": 1630, "lr": 4.4689214344177775e-05}
{"train_loss": 0.12239942699670792, "global_step": 145153, "epoch": 1630, "lr": 4.4688637887437075e-05}
{"train_loss": 0.08015799522399902, "global_step": 145154, "epoch": 1630, "lr": 4.46880614314104e-05}
{"train_loss": 0.09314203262329102, "global_step": 145155, "epoch": 1630, "lr": 4.4687484976097856e-05}
{"train_loss": 0.1458173394203186, "global_step": 145156, "epoch": 1630, "lr": 4.468690852149952e-05}
{"train_loss": 0.10438112914562225, "global_step": 145157, "epoch": 1630, "lr": 4.468633206761546e-05}
{"train_loss": 0.10995956384733821, "global_step": 145158, "epoch": 1630, "lr": 4.4685755614445766e-05, "val_loss": 6.200064182281494, "train_action_mse_error": 16.812049865722656}
{"train_loss": 0.08559798449277878, "global_step": 145159, "epoch": 1631, "lr": 4.46851791619905e-05}
{"train_loss": 0.13805878162384033, "global_step": 145160, "epoch": 1631, "lr": 4.468460271024975e-05}
{"train_loss": 0.04987913370132446, "global_step": 145161, "epoch": 1631, "lr": 4.468402625922359e-05}
{"train_loss": 0.09588254988193512, "global_step": 145162, "epoch": 1631, "lr": 4.468344980891211e-05}
{"train_loss": 0.1566890925168991, "global_step": 145163, "epoch": 1631, "lr": 4.468287335931535e-05}
{"train_loss": 0.13250944018363953, "global_step": 145164, "epoch": 1631, "lr": 4.468229691043344e-05}
{"train_loss": 0.06135762110352516, "global_step": 145165, "epoch": 1631, "lr": 4.46817204622664e-05}
{"train_loss": 0.04988301545381546, "global_step": 145166, "epoch": 1631, "lr": 4.468114401481437e-05}
{"train_loss": 0.053378522396087646, "global_step": 145167, "epoch": 1631, "lr": 4.468056756807737e-05}
{"train_loss": 0.11721068620681763, "global_step": 145168, "epoch": 1631, "lr": 4.467999112205552e-05}
{"train_loss": 0.0902768149971962, "global_step": 145169, "epoch": 1631, "lr": 4.467941467674886e-05}
{"train_loss": 0.09006975591182709, "global_step": 145170, "epoch": 1631, "lr": 4.4678838232157515e-05}
{"train_loss": 0.09996102750301361, "global_step": 145171, "epoch": 1631, "lr": 4.4678261788281504e-05}
{"train_loss": 0.09666694700717926, "global_step": 145172, "epoch": 1631, "lr": 4.4677685345120965e-05}
{"train_loss": 0.10028122365474701, "global_step": 145173, "epoch": 1631, "lr": 4.467710890267593e-05}
{"train_loss": 0.12825040519237518, "global_step": 145174, "epoch": 1631, "lr": 4.4676532460946486e-05}
{"train_loss": 0.06384943425655365, "global_step": 145175, "epoch": 1631, "lr": 4.4675956019932735e-05}
{"train_loss": 0.08706586807966232, "global_step": 145176, "epoch": 1631, "lr": 4.467537957963473e-05}
{"train_loss": 0.11102790385484695, "global_step": 145177, "epoch": 1631, "lr": 4.467480314005256e-05}
{"train_loss": 0.12469830363988876, "global_step": 145178, "epoch": 1631, "lr": 4.467422670118629e-05}
{"train_loss": 0.14699649810791016, "global_step": 145179, "epoch": 1631, "lr": 4.467365026303601e-05}
{"train_loss": 0.15373367071151733, "global_step": 145180, "epoch": 1631, "lr": 4.467307382560179e-05}
{"train_loss": 0.11816344410181046, "global_step": 145181, "epoch": 1631, "lr": 4.4672497388883724e-05}
{"train_loss": 0.1506364941596985, "global_step": 145182, "epoch": 1631, "lr": 4.4671920952881853e-05}
{"train_loss": 0.12647010385990143, "global_step": 145183, "epoch": 1631, "lr": 4.4671344517596295e-05}
{"train_loss": 0.20617851614952087, "global_step": 145184, "epoch": 1631, "lr": 4.467076808302709e-05}
{"train_loss": 0.1408705711364746, "global_step": 145185, "epoch": 1631, "lr": 4.4670191649174365e-05}
{"train_loss": 0.13027070462703705, "global_step": 145186, "epoch": 1631, "lr": 4.4669615216038136e-05}
{"train_loss": 0.09300266206264496, "global_step": 145187, "epoch": 1631, "lr": 4.466903878361854e-05}
{"train_loss": 0.19128218293190002, "global_step": 145188, "epoch": 1631, "lr": 4.46684623519156e-05}
{"train_loss": 0.10206384211778641, "global_step": 145189, "epoch": 1631, "lr": 4.466788592092945e-05}
{"train_loss": 0.06907626986503601, "global_step": 145190, "epoch": 1631, "lr": 4.466730949066011e-05}
{"train_loss": 0.12863683700561523, "global_step": 145191, "epoch": 1631, "lr": 4.466673306110769e-05}
{"train_loss": 0.11968424171209335, "global_step": 145192, "epoch": 1631, "lr": 4.466615663227227e-05}
{"train_loss": 0.1591140776872635, "global_step": 145193, "epoch": 1631, "lr": 4.4665580204153914e-05}
{"train_loss": 0.1560678780078888, "global_step": 145194, "epoch": 1631, "lr": 4.466500377675272e-05}
{"train_loss": 0.16584047675132751, "global_step": 145195, "epoch": 1631, "lr": 4.466442735006874e-05}
{"train_loss": 0.11357022076845169, "global_step": 145196, "epoch": 1631, "lr": 4.466385092410207e-05}
{"train_loss": 0.10225952416658401, "global_step": 145197, "epoch": 1631, "lr": 4.4663274498852775e-05}
{"train_loss": 0.15722982585430145, "global_step": 145198, "epoch": 1631, "lr": 4.4662698074320944e-05}
{"train_loss": 0.11162177473306656, "global_step": 145199, "epoch": 1631, "lr": 4.466212165050664e-05}
{"train_loss": 0.06598935276269913, "global_step": 145200, "epoch": 1631, "lr": 4.4661545227409965e-05}
{"train_loss": 0.10026822984218597, "global_step": 145201, "epoch": 1631, "lr": 4.466096880503095e-05}
{"train_loss": 0.14616738259792328, "global_step": 145202, "epoch": 1631, "lr": 4.466039238336973e-05}
{"train_loss": 0.09281155467033386, "global_step": 145203, "epoch": 1631, "lr": 4.465981596242633e-05}
{"train_loss": 0.0979519784450531, "global_step": 145204, "epoch": 1631, "lr": 4.465923954220088e-05}
{"train_loss": 0.06534550338983536, "global_step": 145205, "epoch": 1631, "lr": 4.4658663122693405e-05}
{"train_loss": 0.09473089873790741, "global_step": 145206, "epoch": 1631, "lr": 4.465808670390404e-05}
{"train_loss": 0.1707802414894104, "global_step": 145207, "epoch": 1631, "lr": 4.465751028583279e-05}
{"train_loss": 0.055267736315727234, "global_step": 145208, "epoch": 1631, "lr": 4.465693386847981e-05}
{"train_loss": 0.1917954385280609, "global_step": 145209, "epoch": 1631, "lr": 4.465635745184511e-05}
{"train_loss": 0.08769325911998749, "global_step": 145210, "epoch": 1631, "lr": 4.4655781035928795e-05}
{"train_loss": 0.08337488770484924, "global_step": 145211, "epoch": 1631, "lr": 4.4655204620730975e-05}
{"train_loss": 0.12273595482110977, "global_step": 145212, "epoch": 1631, "lr": 4.465462820625167e-05}
{"train_loss": 0.0818195566534996, "global_step": 145213, "epoch": 1631, "lr": 4.465405179249101e-05}
{"train_loss": 0.19130167365074158, "global_step": 145214, "epoch": 1631, "lr": 4.4653475379449036e-05}
{"train_loss": 0.13009968400001526, "global_step": 145215, "epoch": 1631, "lr": 4.4652898967125836e-05}
{"train_loss": 0.08109351992607117, "global_step": 145216, "epoch": 1631, "lr": 4.465232255552149e-05}
{"train_loss": 0.07063143700361252, "global_step": 145217, "epoch": 1631, "lr": 4.465174614463608e-05}
{"train_loss": 0.15075263381004333, "global_step": 145218, "epoch": 1631, "lr": 4.465116973446966e-05}
{"train_loss": 0.09553738683462143, "global_step": 145219, "epoch": 1631, "lr": 4.465059332502235e-05}
{"train_loss": 0.1163358986377716, "global_step": 145220, "epoch": 1631, "lr": 4.465001691629418e-05}
{"train_loss": 0.11906488239765167, "global_step": 145221, "epoch": 1631, "lr": 4.464944050828527e-05}
{"train_loss": 0.09742432832717896, "global_step": 145222, "epoch": 1631, "lr": 4.4648864100995655e-05}
{"train_loss": 0.13126903772354126, "global_step": 145223, "epoch": 1631, "lr": 4.4648287694425456e-05}
{"train_loss": 0.13134747743606567, "global_step": 145224, "epoch": 1631, "lr": 4.4647711288574706e-05}
{"train_loss": 0.10615964978933334, "global_step": 145225, "epoch": 1631, "lr": 4.4647134883443534e-05}
{"train_loss": 0.12294626235961914, "global_step": 145226, "epoch": 1631, "lr": 4.464655847903196e-05}
{"train_loss": 0.13239562511444092, "global_step": 145227, "epoch": 1631, "lr": 4.464598207534012e-05}
{"train_loss": 0.14082500338554382, "global_step": 145228, "epoch": 1631, "lr": 4.4645405672368036e-05}
{"train_loss": 0.12184267491102219, "global_step": 145229, "epoch": 1631, "lr": 4.4644829270115814e-05}
{"train_loss": 0.13005267083644867, "global_step": 145230, "epoch": 1631, "lr": 4.464425286858355e-05}
{"train_loss": 0.0602167584002018, "global_step": 145231, "epoch": 1631, "lr": 4.464367646777128e-05}
{"train_loss": 0.09779376536607742, "global_step": 145232, "epoch": 1631, "lr": 4.4643100067679125e-05}
{"train_loss": 0.16197216510772705, "global_step": 145233, "epoch": 1631, "lr": 4.4642523668307126e-05}
{"train_loss": 0.12638653814792633, "global_step": 145234, "epoch": 1631, "lr": 4.464194726965538e-05}
{"train_loss": 0.10756538063287735, "global_step": 145235, "epoch": 1631, "lr": 4.464137087172395e-05}
{"train_loss": 0.1305636465549469, "global_step": 145236, "epoch": 1631, "lr": 4.4640794474512924e-05}
{"train_loss": 0.12384495884180069, "global_step": 145237, "epoch": 1631, "lr": 4.464021807802239e-05}
{"train_loss": 0.1401359885931015, "global_step": 145238, "epoch": 1631, "lr": 4.463964168225241e-05}
{"train_loss": 0.14172719419002533, "global_step": 145239, "epoch": 1631, "lr": 4.463906528720305e-05}
{"train_loss": 0.11940474063158035, "global_step": 145240, "epoch": 1631, "lr": 4.463848889287443e-05}
{"train_loss": 0.18213674426078796, "global_step": 145241, "epoch": 1631, "lr": 4.4637912499266563e-05}
{"train_loss": 0.07740680128335953, "global_step": 145242, "epoch": 1631, "lr": 4.46373361063796e-05}
{"train_loss": 0.11993630230426788, "global_step": 145243, "epoch": 1631, "lr": 4.463675971421355e-05}
{"train_loss": 0.1333136111497879, "global_step": 145244, "epoch": 1631, "lr": 4.4636183322768543e-05}
{"train_loss": 0.07993250340223312, "global_step": 145245, "epoch": 1631, "lr": 4.463560693204462e-05}
{"train_loss": 0.15149220824241638, "global_step": 145246, "epoch": 1631, "lr": 4.463503054204187e-05}
{"train_loss": 0.11661721079537038, "global_step": 145247, "epoch": 1631, "lr": 4.46344541527604e-05, "val_loss": 5.961129188537598}
{"train_loss": 0.12408022582530975, "global_step": 145248, "epoch": 1632, "lr": 4.463387776420024e-05}
{"train_loss": 0.05719083920121193, "global_step": 145249, "epoch": 1632, "lr": 4.463330137636151e-05}
{"train_loss": 0.11518050730228424, "global_step": 145250, "epoch": 1632, "lr": 4.4632724989244246e-05}
{"train_loss": 0.08361544460058212, "global_step": 145251, "epoch": 1632, "lr": 4.463214860284857e-05}
{"train_loss": 0.09141236543655396, "global_step": 145252, "epoch": 1632, "lr": 4.463157221717452e-05}
{"train_loss": 0.08670516312122345, "global_step": 145253, "epoch": 1632, "lr": 4.46309958322222e-05}
{"train_loss": 0.0887984037399292, "global_step": 145254, "epoch": 1632, "lr": 4.4630419447991664e-05}
{"train_loss": 0.09915503114461899, "global_step": 145255, "epoch": 1632, "lr": 4.4629843064483014e-05}
{"train_loss": 0.08357987552881241, "global_step": 145256, "epoch": 1632, "lr": 4.4629266681696314e-05}
{"train_loss": 0.16730907559394836, "global_step": 145257, "epoch": 1632, "lr": 4.4628690299631645e-05}
{"train_loss": 0.12532135844230652, "global_step": 145258, "epoch": 1632, "lr": 4.4628113918289075e-05}
{"train_loss": 0.1365082561969757, "global_step": 145259, "epoch": 1632, "lr": 4.4627537537668706e-05}
{"train_loss": 0.0896664410829544, "global_step": 145260, "epoch": 1632, "lr": 4.462696115777058e-05}
{"train_loss": 0.10068173706531525, "global_step": 145261, "epoch": 1632, "lr": 4.4626384778594806e-05}
{"train_loss": 0.15044274926185608, "global_step": 145262, "epoch": 1632, "lr": 4.462580840014143e-05}
{"train_loss": 0.13022640347480774, "global_step": 145263, "epoch": 1632, "lr": 4.4625232022410576e-05}
{"train_loss": 0.10725683718919754, "global_step": 145264, "epoch": 1632, "lr": 4.462465564540227e-05}
{"train_loss": 0.09921708703041077, "global_step": 145265, "epoch": 1632, "lr": 4.462407926911661e-05}
{"train_loss": 0.07431627810001373, "global_step": 145266, "epoch": 1632, "lr": 4.46235028935537e-05}
{"train_loss": 0.14166781306266785, "global_step": 145267, "epoch": 1632, "lr": 4.462292651871358e-05}
{"train_loss": 0.10458647459745407, "global_step": 145268, "epoch": 1632, "lr": 4.4622350144596356e-05}
{"train_loss": 0.07285314053297043, "global_step": 145269, "epoch": 1632, "lr": 4.462177377120207e-05}
{"train_loss": 0.13541796803474426, "global_step": 145270, "epoch": 1632, "lr": 4.462119739853084e-05}
{"train_loss": 0.10900581628084183, "global_step": 145271, "epoch": 1632, "lr": 4.4620621026582715e-05}
{"train_loss": 0.12164664268493652, "global_step": 145272, "epoch": 1632, "lr": 4.46200446553578e-05}
{"train_loss": 0.14384518563747406, "global_step": 145273, "epoch": 1632, "lr": 4.461946828485614e-05}
{"train_loss": 0.1706802397966385, "global_step": 145274, "epoch": 1632, "lr": 4.4618891915077824e-05}
{"train_loss": 0.13957446813583374, "global_step": 145275, "epoch": 1632, "lr": 4.461831554602294e-05}
{"train_loss": 0.13211606442928314, "global_step": 145276, "epoch": 1632, "lr": 4.4617739177691546e-05}
{"train_loss": 0.19872444868087769, "global_step": 145277, "epoch": 1632, "lr": 4.4617162810083743e-05}
{"train_loss": 0.08677133917808533, "global_step": 145278, "epoch": 1632, "lr": 4.46165864431996e-05}
{"train_loss": 0.17765407264232635, "global_step": 145279, "epoch": 1632, "lr": 4.461601007703917e-05}
{"train_loss": 0.12540790438652039, "global_step": 145280, "epoch": 1632, "lr": 4.461543371160258e-05}
{"train_loss": 0.04304629936814308, "global_step": 145281, "epoch": 1632, "lr": 4.4614857346889856e-05}
{"train_loss": 0.12077005207538605, "global_step": 145282, "epoch": 1632, "lr": 4.4614280982901094e-05}
{"train_loss": 0.16713875532150269, "global_step": 145283, "epoch": 1632, "lr": 4.4613704619636396e-05}
{"train_loss": 0.10790325701236725, "global_step": 145284, "epoch": 1632, "lr": 4.46131282570958e-05}
{"train_loss": 0.10446556657552719, "global_step": 145285, "epoch": 1632, "lr": 4.461255189527943e-05}
{"train_loss": 0.1507250815629959, "global_step": 145286, "epoch": 1632, "lr": 4.4611975534187316e-05}
{"train_loss": 0.1235499382019043, "global_step": 145287, "epoch": 1632, "lr": 4.461139917381957e-05}
{"train_loss": 0.0869063213467598, "global_step": 145288, "epoch": 1632, "lr": 4.4610822814176244e-05}
{"train_loss": 0.10586366057395935, "global_step": 145289, "epoch": 1632, "lr": 4.461024645525744e-05}
{"train_loss": 0.23453345894813538, "global_step": 145290, "epoch": 1632, "lr": 4.4609670097063205e-05}
{"train_loss": 0.13205116987228394, "global_step": 145291, "epoch": 1632, "lr": 4.4609093739593656e-05}
{"train_loss": 0.12917838990688324, "global_step": 145292, "epoch": 1632, "lr": 4.4608517382848834e-05}
{"train_loss": 0.06417689472436905, "global_step": 145293, "epoch": 1632, "lr": 4.4607941026828836e-05}
{"train_loss": 0.13558053970336914, "global_step": 145294, "epoch": 1632, "lr": 4.4607364671533735e-05}
{"train_loss": 0.08188565075397491, "global_step": 145295, "epoch": 1632, "lr": 4.460678831696361e-05}
{"train_loss": 0.16695205867290497, "global_step": 145296, "epoch": 1632, "lr": 4.4606211963118525e-05}
{"train_loss": 0.10672518610954285, "global_step": 145297, "epoch": 1632, "lr": 4.460563560999859e-05}
{"train_loss": 0.08733952045440674, "global_step": 145298, "epoch": 1632, "lr": 4.460505925760384e-05}
{"train_loss": 0.14629854261875153, "global_step": 145299, "epoch": 1632, "lr": 4.460448290593439e-05}
{"train_loss": 0.13331201672554016, "global_step": 145300, "epoch": 1632, "lr": 4.460390655499028e-05}
{"train_loss": 0.08325766026973724, "global_step": 145301, "epoch": 1632, "lr": 4.460333020477161e-05}
{"train_loss": 0.08911174535751343, "global_step": 145302, "epoch": 1632, "lr": 4.4602753855278485e-05}
{"train_loss": 0.1795036494731903, "global_step": 145303, "epoch": 1632, "lr": 4.4602177506510925e-05}
{"train_loss": 0.14654922485351562, "global_step": 145304, "epoch": 1632, "lr": 4.460160115846906e-05}
{"train_loss": 0.0640307366847992, "global_step": 145305, "epoch": 1632, "lr": 4.4601024811152924e-05}
{"train_loss": 0.12918251752853394, "global_step": 145306, "epoch": 1632, "lr": 4.460044846456264e-05}
{"train_loss": 0.10122363269329071, "global_step": 145307, "epoch": 1632, "lr": 4.4599872118698227e-05}
{"train_loss": 0.12081456184387207, "global_step": 145308, "epoch": 1632, "lr": 4.4599295773559823e-05}
{"train_loss": 0.13640525937080383, "global_step": 145309, "epoch": 1632, "lr": 4.4598719429147456e-05}
{"train_loss": 0.11943384259939194, "global_step": 145310, "epoch": 1632, "lr": 4.459814308546125e-05}
{"train_loss": 0.12156901508569717, "global_step": 145311, "epoch": 1632, "lr": 4.459756674250124e-05}
{"train_loss": 0.09704270213842392, "global_step": 145312, "epoch": 1632, "lr": 4.459699040026753e-05}
{"train_loss": 0.16035465896129608, "global_step": 145313, "epoch": 1632, "lr": 4.4596414058760184e-05}
{"train_loss": 0.11812160909175873, "global_step": 145314, "epoch": 1632, "lr": 4.4595837717979284e-05}
{"train_loss": 0.11345367878675461, "global_step": 145315, "epoch": 1632, "lr": 4.459526137792491e-05}
{"train_loss": 0.06906614452600479, "global_step": 145316, "epoch": 1632, "lr": 4.4594685038597144e-05}
{"train_loss": 0.10117889195680618, "global_step": 145317, "epoch": 1632, "lr": 4.459410869999604e-05}
{"train_loss": 0.09162833541631699, "global_step": 145318, "epoch": 1632, "lr": 4.459353236212169e-05}
{"train_loss": 0.14069899916648865, "global_step": 145319, "epoch": 1632, "lr": 4.4592956024974195e-05}
{"train_loss": 0.046621158719062805, "global_step": 145320, "epoch": 1632, "lr": 4.459237968855359e-05}
{"train_loss": 0.07470479607582092, "global_step": 145321, "epoch": 1632, "lr": 4.459180335285999e-05}
{"train_loss": 0.09290257841348648, "global_step": 145322, "epoch": 1632, "lr": 4.459122701789344e-05}
{"train_loss": 0.14913049340248108, "global_step": 145323, "epoch": 1632, "lr": 4.459065068365406e-05}
{"train_loss": 0.1392853856086731, "global_step": 145324, "epoch": 1632, "lr": 4.459007435014187e-05}
{"train_loss": 0.057412102818489075, "global_step": 145325, "epoch": 1632, "lr": 4.4589498017357e-05}
{"train_loss": 0.07591516524553299, "global_step": 145326, "epoch": 1632, "lr": 4.458892168529949e-05}
{"train_loss": 0.14967915415763855, "global_step": 145327, "epoch": 1632, "lr": 4.458834535396945e-05}
{"train_loss": 0.09172862768173218, "global_step": 145328, "epoch": 1632, "lr": 4.4587769023366924e-05}
{"train_loss": 0.12127776443958282, "global_step": 145329, "epoch": 1632, "lr": 4.458719269349203e-05}
{"train_loss": 0.11821316927671432, "global_step": 145330, "epoch": 1632, "lr": 4.4586616364344807e-05}
{"train_loss": 0.1248672679066658, "global_step": 145331, "epoch": 1632, "lr": 4.458604003592535e-05}
{"train_loss": 0.14271435141563416, "global_step": 145332, "epoch": 1632, "lr": 4.458546370823373e-05}
{"train_loss": 0.1200333684682846, "global_step": 145333, "epoch": 1632, "lr": 4.458488738127003e-05}
{"train_loss": 0.07068182528018951, "global_step": 145334, "epoch": 1632, "lr": 4.458431105503433e-05}
{"train_loss": 0.11854095011949539, "global_step": 145335, "epoch": 1632, "lr": 4.4583734729526714e-05}
{"train_loss": 0.11593814169088107, "global_step": 145336, "epoch": 1632, "lr": 4.458315840474722e-05, "val_loss": 6.312122821807861}
{"train_loss": 0.11659500002861023, "global_step": 145337, "epoch": 1633, "lr": 4.458258208069596e-05}
{"train_loss": 0.12800168991088867, "global_step": 145338, "epoch": 1633, "lr": 4.458200575737303e-05}
{"train_loss": 0.10399983078241348, "global_step": 145339, "epoch": 1633, "lr": 4.458142943477845e-05}
{"train_loss": 0.11162111908197403, "global_step": 145340, "epoch": 1633, "lr": 4.458085311291236e-05}
{"train_loss": 0.1321810632944107, "global_step": 145341, "epoch": 1633, "lr": 4.4580276791774786e-05}
{"train_loss": 0.06348980963230133, "global_step": 145342, "epoch": 1633, "lr": 4.4579700471365856e-05}
{"train_loss": 0.10139606893062592, "global_step": 145343, "epoch": 1633, "lr": 4.457912415168559e-05}
{"train_loss": 0.07707464694976807, "global_step": 145344, "epoch": 1633, "lr": 4.457854783273412e-05}
{"train_loss": 0.10309033840894699, "global_step": 145345, "epoch": 1633, "lr": 4.4577971514511475e-05}
{"train_loss": 0.11388054490089417, "global_step": 145346, "epoch": 1633, "lr": 4.4577395197017774e-05}
{"train_loss": 0.10580940544605255, "global_step": 145347, "epoch": 1633, "lr": 4.457681888025306e-05}
{"train_loss": 0.10447002202272415, "global_step": 145348, "epoch": 1633, "lr": 4.4576242564217445e-05}
{"train_loss": 0.11532993614673615, "global_step": 145349, "epoch": 1633, "lr": 4.457566624891097e-05}
{"train_loss": 0.1265280395746231, "global_step": 145350, "epoch": 1633, "lr": 4.457508993433375e-05}
{"train_loss": 0.11697255820035934, "global_step": 145351, "epoch": 1633, "lr": 4.457451362048583e-05}
{"train_loss": 0.11563428491353989, "global_step": 145352, "epoch": 1633, "lr": 4.4573937307367296e-05}
{"train_loss": 0.1328442543745041, "global_step": 145353, "epoch": 1633, "lr": 4.457336099497824e-05}
{"train_loss": 0.11755185574293137, "global_step": 145354, "epoch": 1633, "lr": 4.4572784683318735e-05}
{"train_loss": 0.2047591358423233, "global_step": 145355, "epoch": 1633, "lr": 4.4572208372388835e-05}
{"train_loss": 0.1275085210800171, "global_step": 145356, "epoch": 1633, "lr": 4.457163206218863e-05}
{"train_loss": 0.0798143744468689, "global_step": 145357, "epoch": 1633, "lr": 4.457105575271823e-05}
{"train_loss": 0.072652667760849, "global_step": 145358, "epoch": 1633, "lr": 4.457047944397766e-05}
{"train_loss": 0.15735678374767303, "global_step": 145359, "epoch": 1633, "lr": 4.4569903135967047e-05}
{"train_loss": 0.1925351470708847, "global_step": 145360, "epoch": 1633, "lr": 4.456932682868642e-05}
{"train_loss": 0.07731828838586807, "global_step": 145361, "epoch": 1633, "lr": 4.4568750522135906e-05}
{"train_loss": 0.11801771074533463, "global_step": 145362, "epoch": 1633, "lr": 4.4568174216315525e-05}
{"train_loss": 0.10030049085617065, "global_step": 145363, "epoch": 1633, "lr": 4.456759791122542e-05}
{"train_loss": 0.15543366968631744, "global_step": 145364, "epoch": 1633, "lr": 4.456702160686561e-05}
{"train_loss": 0.06277713179588318, "global_step": 145365, "epoch": 1633, "lr": 4.456644530323623e-05}
{"train_loss": 0.13351739943027496, "global_step": 145366, "epoch": 1633, "lr": 4.456586900033728e-05}
{"train_loss": 0.09639821946620941, "global_step": 145367, "epoch": 1633, "lr": 4.456529269816892e-05}
{"train_loss": 0.10739906132221222, "global_step": 145368, "epoch": 1633, "lr": 4.4564716396731185e-05}
{"train_loss": 0.1037856936454773, "global_step": 145369, "epoch": 1633, "lr": 4.4564140096024146e-05}
{"train_loss": 0.10736291855573654, "global_step": 145370, "epoch": 1633, "lr": 4.45635637960479e-05}
{"train_loss": 0.09572633355855942, "global_step": 145371, "epoch": 1633, "lr": 4.456298749680251e-05}
{"train_loss": 0.15159602463245392, "global_step": 145372, "epoch": 1633, "lr": 4.4562411198288064e-05}
{"train_loss": 0.06799566000699997, "global_step": 145373, "epoch": 1633, "lr": 4.456183490050462e-05}
{"train_loss": 0.13122273981571198, "global_step": 145374, "epoch": 1633, "lr": 4.45612586034523e-05}
{"train_loss": 0.1358511745929718, "global_step": 145375, "epoch": 1633, "lr": 4.4560682307131124e-05}
{"train_loss": 0.07252034544944763, "global_step": 145376, "epoch": 1633, "lr": 4.456010601154123e-05}
{"train_loss": 0.1269516795873642, "global_step": 145377, "epoch": 1633, "lr": 4.4559529716682625e-05}
{"train_loss": 0.10029707103967667, "global_step": 145378, "epoch": 1633, "lr": 4.455895342255546e-05}
{"train_loss": 0.16302502155303955, "global_step": 145379, "epoch": 1633, "lr": 4.455837712915975e-05}
{"train_loss": 0.11632561683654785, "global_step": 145380, "epoch": 1633, "lr": 4.455780083649562e-05}
{"train_loss": 0.07885898649692535, "global_step": 145381, "epoch": 1633, "lr": 4.4557224544563106e-05}
{"train_loss": 0.06518924236297607, "global_step": 145382, "epoch": 1633, "lr": 4.455664825336234e-05}
{"train_loss": 0.13967961072921753, "global_step": 145383, "epoch": 1633, "lr": 4.455607196289333e-05}
{"train_loss": 0.09063905477523804, "global_step": 145384, "epoch": 1633, "lr": 4.455549567315622e-05}
{"train_loss": 0.0904410257935524, "global_step": 145385, "epoch": 1633, "lr": 4.455491938415103e-05}
{"train_loss": 0.09663329273462296, "global_step": 145386, "epoch": 1633, "lr": 4.4554343095877894e-05}
{"train_loss": 0.128734290599823, "global_step": 145387, "epoch": 1633, "lr": 4.455376680833683e-05}
{"train_loss": 0.1059972271323204, "global_step": 145388, "epoch": 1633, "lr": 4.455319052152797e-05}
{"train_loss": 0.11153727769851685, "global_step": 145389, "epoch": 1633, "lr": 4.4552614235451364e-05}
{"train_loss": 0.19081947207450867, "global_step": 145390, "epoch": 1633, "lr": 4.455203795010708e-05}
{"train_loss": 0.09860117733478546, "global_step": 145391, "epoch": 1633, "lr": 4.4551461665495215e-05}
{"train_loss": 0.1727687567472458, "global_step": 145392, "epoch": 1633, "lr": 4.4550885381615834e-05}
{"train_loss": 0.11950916796922684, "global_step": 145393, "epoch": 1633, "lr": 4.455030909846903e-05}
{"train_loss": 0.12210844457149506, "global_step": 145394, "epoch": 1633, "lr": 4.454973281605486e-05}
{"train_loss": 0.07001582533121109, "global_step": 145395, "epoch": 1633, "lr": 4.4549156534373425e-05}
{"train_loss": 0.08493676781654358, "global_step": 145396, "epoch": 1633, "lr": 4.454858025342477e-05}
{"train_loss": 0.11503072828054428, "global_step": 145397, "epoch": 1633, "lr": 4.454800397320901e-05}
{"train_loss": 0.09949113428592682, "global_step": 145398, "epoch": 1633, "lr": 4.454742769372619e-05}
{"train_loss": 0.18897542357444763, "global_step": 145399, "epoch": 1633, "lr": 4.454685141497642e-05}
{"train_loss": 0.2111348658800125, "global_step": 145400, "epoch": 1633, "lr": 4.454627513695974e-05}
{"train_loss": 0.16004639863967896, "global_step": 145401, "epoch": 1633, "lr": 4.4545698859676266e-05}
{"train_loss": 0.08735303580760956, "global_step": 145402, "epoch": 1633, "lr": 4.454512258312604e-05}
{"train_loss": 0.16466814279556274, "global_step": 145403, "epoch": 1633, "lr": 4.454454630730917e-05}
{"train_loss": 0.19069458544254303, "global_step": 145404, "epoch": 1633, "lr": 4.4543970032225704e-05}
{"train_loss": 0.16438616812229156, "global_step": 145405, "epoch": 1633, "lr": 4.454339375787575e-05}
{"train_loss": 0.15672272443771362, "global_step": 145406, "epoch": 1633, "lr": 4.454281748425935e-05}
{"train_loss": 0.18422116339206696, "global_step": 145407, "epoch": 1633, "lr": 4.4542241211376625e-05}
{"train_loss": 0.1199055016040802, "global_step": 145408, "epoch": 1633, "lr": 4.454166493922762e-05}
{"train_loss": 0.12224511057138443, "global_step": 145409, "epoch": 1633, "lr": 4.454108866781241e-05}
{"train_loss": 0.10326812416315079, "global_step": 145410, "epoch": 1633, "lr": 4.4540512397131104e-05}
{"train_loss": 0.1436094492673874, "global_step": 145411, "epoch": 1633, "lr": 4.453993612718374e-05}
{"train_loss": 0.10528470575809479, "global_step": 145412, "epoch": 1633, "lr": 4.453935985797044e-05}
{"train_loss": 0.13916893303394318, "global_step": 145413, "epoch": 1633, "lr": 4.4538783589491226e-05}
{"train_loss": 0.13214585185050964, "global_step": 145414, "epoch": 1633, "lr": 4.453820732174623e-05}
{"train_loss": 0.09404748678207397, "global_step": 145415, "epoch": 1633, "lr": 4.453763105473548e-05}
{"train_loss": 0.11810345202684402, "global_step": 145416, "epoch": 1633, "lr": 4.453705478845911e-05}
{"train_loss": 0.10848987847566605, "global_step": 145417, "epoch": 1633, "lr": 4.453647852291714e-05}
{"train_loss": 0.2330942451953888, "global_step": 145418, "epoch": 1633, "lr": 4.45359022581097e-05}
{"train_loss": 0.14755171537399292, "global_step": 145419, "epoch": 1633, "lr": 4.453532599403681e-05}
{"train_loss": 0.11776524782180786, "global_step": 145420, "epoch": 1633, "lr": 4.453474973069861e-05}
{"train_loss": 0.145041823387146, "global_step": 145421, "epoch": 1633, "lr": 4.4534173468095124e-05}
{"train_loss": 0.10039420425891876, "global_step": 145422, "epoch": 1633, "lr": 4.4533597206226477e-05}
{"train_loss": 0.1766754537820816, "global_step": 145423, "epoch": 1633, "lr": 4.453302094509269e-05}
{"train_loss": 0.12033719569444656, "global_step": 145424, "epoch": 1633, "lr": 4.45324446846939e-05}
{"train_loss": 0.12199509838658772, "global_step": 145425, "epoch": 1633, "lr": 4.453186842503013e-05, "val_loss": 6.173898220062256}
{"train_loss": 0.1701248735189438, "global_step": 145426, "epoch": 1634, "lr": 4.4531292166101504e-05}
{"train_loss": 0.11481545120477676, "global_step": 145427, "epoch": 1634, "lr": 4.453071590790807e-05}
{"train_loss": 0.1109880805015564, "global_step": 145428, "epoch": 1634, "lr": 4.4530139650449904e-05}
{"train_loss": 0.0745728462934494, "global_step": 145429, "epoch": 1634, "lr": 4.452956339372711e-05}
{"train_loss": 0.13541929423809052, "global_step": 145430, "epoch": 1634, "lr": 4.452898713773974e-05}
{"train_loss": 0.11865003407001495, "global_step": 145431, "epoch": 1634, "lr": 4.4528410882487896e-05}
{"train_loss": 0.09124685823917389, "global_step": 145432, "epoch": 1634, "lr": 4.452783462797162e-05}
{"train_loss": 0.1075640320777893, "global_step": 145433, "epoch": 1634, "lr": 4.4527258374191036e-05}
{"train_loss": 0.11375948786735535, "global_step": 145434, "epoch": 1634, "lr": 4.452668212114616e-05}
{"train_loss": 0.16278569400310516, "global_step": 145435, "epoch": 1634, "lr": 4.452610586883714e-05}
{"train_loss": 0.16028252243995667, "global_step": 145436, "epoch": 1634, "lr": 4.4525529617264e-05}
{"train_loss": 0.13667675852775574, "global_step": 145437, "epoch": 1634, "lr": 4.452495336642684e-05}
{"train_loss": 0.10631880164146423, "global_step": 145438, "epoch": 1634, "lr": 4.452437711632572e-05}
{"train_loss": 0.08085574954748154, "global_step": 145439, "epoch": 1634, "lr": 4.4523800866960755e-05}
{"train_loss": 0.10659272968769073, "global_step": 145440, "epoch": 1634, "lr": 4.452322461833198e-05}
{"train_loss": 0.09350742399692535, "global_step": 145441, "epoch": 1634, "lr": 4.452264837043951e-05}
{"train_loss": 0.09042055159807205, "global_step": 145442, "epoch": 1634, "lr": 4.452207212328338e-05}
{"train_loss": 0.09103113412857056, "global_step": 145443, "epoch": 1634, "lr": 4.45214958768637e-05}
{"train_loss": 0.09281858801841736, "global_step": 145444, "epoch": 1634, "lr": 4.4520919631180534e-05}
{"train_loss": 0.17757464945316315, "global_step": 145445, "epoch": 1634, "lr": 4.452034338623397e-05}
{"train_loss": 0.13218456506729126, "global_step": 145446, "epoch": 1634, "lr": 4.451976714202409e-05}
{"train_loss": 0.16069546341896057, "global_step": 145447, "epoch": 1634, "lr": 4.451919089855093e-05}
{"train_loss": 0.050988081842660904, "global_step": 145448, "epoch": 1634, "lr": 4.4518614655814625e-05}
{"train_loss": 0.10217677056789398, "global_step": 145449, "epoch": 1634, "lr": 4.451803841381521e-05}
{"train_loss": 0.11502199620008469, "global_step": 145450, "epoch": 1634, "lr": 4.451746217255279e-05}
{"train_loss": 0.1183491051197052, "global_step": 145451, "epoch": 1634, "lr": 4.451688593202742e-05}
{"train_loss": 0.11749666929244995, "global_step": 145452, "epoch": 1634, "lr": 4.4516309692239205e-05}
{"train_loss": 0.09245814383029938, "global_step": 145453, "epoch": 1634, "lr": 4.451573345318818e-05}
{"train_loss": 0.08171509206295013, "global_step": 145454, "epoch": 1634, "lr": 4.451515721487447e-05}
{"train_loss": 0.0732339546084404, "global_step": 145455, "epoch": 1634, "lr": 4.451458097729811e-05}
{"train_loss": 0.08041151612997055, "global_step": 145456, "epoch": 1634, "lr": 4.451400474045922e-05}
{"train_loss": 0.21253754198551178, "global_step": 145457, "epoch": 1634, "lr": 4.4513428504357836e-05}
{"train_loss": 0.0739048570394516, "global_step": 145458, "epoch": 1634, "lr": 4.451285226899407e-05}
{"train_loss": 0.04765048623085022, "global_step": 145459, "epoch": 1634, "lr": 4.451227603436797e-05}
{"train_loss": 0.06241091340780258, "global_step": 145460, "epoch": 1634, "lr": 4.4511699800479644e-05}
{"train_loss": 0.09264401346445084, "global_step": 145461, "epoch": 1634, "lr": 4.451112356732913e-05}
{"train_loss": 0.05979218706488609, "global_step": 145462, "epoch": 1634, "lr": 4.451054733491655e-05}
{"train_loss": 0.11951044201850891, "global_step": 145463, "epoch": 1634, "lr": 4.4509971103241945e-05}
{"train_loss": 0.08746310323476791, "global_step": 145464, "epoch": 1634, "lr": 4.450939487230541e-05}
{"train_loss": 0.0997830480337143, "global_step": 145465, "epoch": 1634, "lr": 4.450881864210703e-05}
{"train_loss": 0.1105690523982048, "global_step": 145466, "epoch": 1634, "lr": 4.450824241264686e-05}
{"train_loss": 0.11343585699796677, "global_step": 145467, "epoch": 1634, "lr": 4.450766618392501e-05}
{"train_loss": 0.13968601822853088, "global_step": 145468, "epoch": 1634, "lr": 4.450708995594152e-05}
{"train_loss": 0.1702393889427185, "global_step": 145469, "epoch": 1634, "lr": 4.450651372869649e-05}
{"train_loss": 0.1262613832950592, "global_step": 145470, "epoch": 1634, "lr": 4.450593750218998e-05}
{"train_loss": 0.0763968750834465, "global_step": 145471, "epoch": 1634, "lr": 4.450536127642211e-05}
{"train_loss": 0.06871727854013443, "global_step": 145472, "epoch": 1634, "lr": 4.45047850513929e-05}
{"train_loss": 0.12410183995962143, "global_step": 145473, "epoch": 1634, "lr": 4.450420882710247e-05}
{"train_loss": 0.058985959738492966, "global_step": 145474, "epoch": 1634, "lr": 4.450363260355086e-05}
{"train_loss": 0.09996654093265533, "global_step": 145475, "epoch": 1634, "lr": 4.4503056380738197e-05}
{"train_loss": 0.16597668826580048, "global_step": 145476, "epoch": 1634, "lr": 4.4502480158664506e-05}
{"train_loss": 0.08465816080570221, "global_step": 145477, "epoch": 1634, "lr": 4.4501903937329915e-05}
{"train_loss": 0.09860072284936905, "global_step": 145478, "epoch": 1634, "lr": 4.4501327716734446e-05}
{"train_loss": 0.1537535935640335, "global_step": 145479, "epoch": 1634, "lr": 4.450075149687823e-05}
{"train_loss": 0.09021676331758499, "global_step": 145480, "epoch": 1634, "lr": 4.450017527776131e-05}
{"train_loss": 0.09422560036182404, "global_step": 145481, "epoch": 1634, "lr": 4.4499599059383786e-05}
{"train_loss": 0.10435796529054642, "global_step": 145482, "epoch": 1634, "lr": 4.449902284174571e-05}
{"train_loss": 0.13282307982444763, "global_step": 145483, "epoch": 1634, "lr": 4.449844662484718e-05}
{"train_loss": 0.11704827845096588, "global_step": 145484, "epoch": 1634, "lr": 4.449787040868827e-05}
{"train_loss": 0.10512400418519974, "global_step": 145485, "epoch": 1634, "lr": 4.449729419326905e-05}
{"train_loss": 0.14320001006126404, "global_step": 145486, "epoch": 1634, "lr": 4.449671797858961e-05}
{"train_loss": 0.118428073823452, "global_step": 145487, "epoch": 1634, "lr": 4.449614176465e-05}
{"train_loss": 0.14519371092319489, "global_step": 145488, "epoch": 1634, "lr": 4.4495565551450343e-05}
{"train_loss": 0.15291650593280792, "global_step": 145489, "epoch": 1634, "lr": 4.449498933899066e-05}
{"train_loss": 0.10699585825204849, "global_step": 145490, "epoch": 1634, "lr": 4.449441312727109e-05}
{"train_loss": 0.07540233433246613, "global_step": 145491, "epoch": 1634, "lr": 4.449383691629165e-05}
{"train_loss": 0.09810996055603027, "global_step": 145492, "epoch": 1634, "lr": 4.449326070605247e-05}
{"train_loss": 0.07833821326494217, "global_step": 145493, "epoch": 1634, "lr": 4.449268449655358e-05}
{"train_loss": 0.05375722795724869, "global_step": 145494, "epoch": 1634, "lr": 4.4492108287795106e-05}
{"train_loss": 0.09541189670562744, "global_step": 145495, "epoch": 1634, "lr": 4.449153207977708e-05}
{"train_loss": 0.07813145965337753, "global_step": 145496, "epoch": 1634, "lr": 4.4490955872499625e-05}
{"train_loss": 0.12131254374980927, "global_step": 145497, "epoch": 1634, "lr": 4.449037966596277e-05}
{"train_loss": 0.08377082645893097, "global_step": 145498, "epoch": 1634, "lr": 4.448980346016664e-05}
{"train_loss": 0.06786958873271942, "global_step": 145499, "epoch": 1634, "lr": 4.448922725511126e-05}
{"train_loss": 0.1540345549583435, "global_step": 145500, "epoch": 1634, "lr": 4.448865105079675e-05}
{"train_loss": 0.077689528465271, "global_step": 145501, "epoch": 1634, "lr": 4.448807484722317e-05}
{"train_loss": 0.16138792037963867, "global_step": 145502, "epoch": 1634, "lr": 4.4487498644390606e-05}
{"train_loss": 0.09605240821838379, "global_step": 145503, "epoch": 1634, "lr": 4.4486922442299134e-05}
{"train_loss": 0.0946260616183281, "global_step": 145504, "epoch": 1634, "lr": 4.448634624094882e-05}
{"train_loss": 0.08340218663215637, "global_step": 145505, "epoch": 1634, "lr": 4.4485770040339766e-05}
{"train_loss": 0.15301856398582458, "global_step": 145506, "epoch": 1634, "lr": 4.448519384047201e-05}
{"train_loss": 0.1553838551044464, "global_step": 145507, "epoch": 1634, "lr": 4.448461764134568e-05}
{"train_loss": 0.08406461030244827, "global_step": 145508, "epoch": 1634, "lr": 4.4484041442960794e-05}
{"train_loss": 0.13147185742855072, "global_step": 145509, "epoch": 1634, "lr": 4.4483465245317485e-05}
{"train_loss": 0.1009020060300827, "global_step": 145510, "epoch": 1634, "lr": 4.448288904841579e-05}
{"train_loss": 0.15459531545639038, "global_step": 145511, "epoch": 1634, "lr": 4.448231285225582e-05}
{"train_loss": 0.11640104651451111, "global_step": 145512, "epoch": 1634, "lr": 4.448173665683762e-05}
{"train_loss": 0.07204418629407883, "global_step": 145513, "epoch": 1634, "lr": 4.4481160462161296e-05}
{"train_loss": 0.10884913806332631, "global_step": 145514, "epoch": 1634, "lr": 4.44805842682269e-05, "val_loss": 6.205440521240234}
{"train_loss": 0.07110337167978287, "global_step": 145515, "epoch": 1635, "lr": 4.448000807503454e-05}
{"train_loss": 0.13799244165420532, "global_step": 145516, "epoch": 1635, "lr": 4.447943188258425e-05}
{"train_loss": 0.08509356528520584, "global_step": 145517, "epoch": 1635, "lr": 4.4478855690876164e-05}
{"train_loss": 0.08607757091522217, "global_step": 145518, "epoch": 1635, "lr": 4.44782794999103e-05}
{"train_loss": 0.215982586145401, "global_step": 145519, "epoch": 1635, "lr": 4.447770330968678e-05}
{"train_loss": 0.07351595908403397, "global_step": 145520, "epoch": 1635, "lr": 4.4477127120205666e-05}
{"train_loss": 0.15991057455539703, "global_step": 145521, "epoch": 1635, "lr": 4.4476550931467025e-05}
{"train_loss": 0.17011000216007233, "global_step": 145522, "epoch": 1635, "lr": 4.447597474347095e-05}
{"train_loss": 0.10951323807239532, "global_step": 145523, "epoch": 1635, "lr": 4.447539855621752e-05}
{"train_loss": 0.07187182456254959, "global_step": 145524, "epoch": 1635, "lr": 4.447482236970681e-05}
{"train_loss": 0.1418699324131012, "global_step": 145525, "epoch": 1635, "lr": 4.4474246183938864e-05}
{"train_loss": 0.13587547838687897, "global_step": 145526, "epoch": 1635, "lr": 4.447366999891382e-05}
{"train_loss": 0.10497298091650009, "global_step": 145527, "epoch": 1635, "lr": 4.44730938146317e-05}
{"train_loss": 0.18018247187137604, "global_step": 145528, "epoch": 1635, "lr": 4.447251763109262e-05}
{"train_loss": 0.1117757111787796, "global_step": 145529, "epoch": 1635, "lr": 4.4471941448296635e-05}
{"train_loss": 0.16200710833072662, "global_step": 145530, "epoch": 1635, "lr": 4.447136526624384e-05}
{"train_loss": 0.13470980525016785, "global_step": 145531, "epoch": 1635, "lr": 4.447078908493428e-05}
{"train_loss": 0.08339373767375946, "global_step": 145532, "epoch": 1635, "lr": 4.4470212904368085e-05}
{"train_loss": 0.11186110973358154, "global_step": 145533, "epoch": 1635, "lr": 4.446963672454528e-05}
{"train_loss": 0.18434593081474304, "global_step": 145534, "epoch": 1635, "lr": 4.446906054546598e-05}
{"train_loss": 0.11041794717311859, "global_step": 145535, "epoch": 1635, "lr": 4.446848436713023e-05}
{"train_loss": 0.12434549629688263, "global_step": 145536, "epoch": 1635, "lr": 4.446790818953812e-05}
{"train_loss": 0.10795219987630844, "global_step": 145537, "epoch": 1635, "lr": 4.446733201268976e-05}
{"train_loss": 0.2361924797296524, "global_step": 145538, "epoch": 1635, "lr": 4.446675583658519e-05}
{"train_loss": 0.08258071541786194, "global_step": 145539, "epoch": 1635, "lr": 4.4466179661224494e-05}
{"train_loss": 0.1967044174671173, "global_step": 145540, "epoch": 1635, "lr": 4.4465603486607755e-05}
{"train_loss": 0.12691953778266907, "global_step": 145541, "epoch": 1635, "lr": 4.446502731273504e-05}
{"train_loss": 0.12029265612363815, "global_step": 145542, "epoch": 1635, "lr": 4.4464451139606446e-05}
{"train_loss": 0.09049519151449203, "global_step": 145543, "epoch": 1635, "lr": 4.446387496722204e-05}
{"train_loss": 0.10713408142328262, "global_step": 145544, "epoch": 1635, "lr": 4.446329879558189e-05}
{"train_loss": 0.09005065262317657, "global_step": 145545, "epoch": 1635, "lr": 4.4462722624686084e-05}
{"train_loss": 0.05026734620332718, "global_step": 145546, "epoch": 1635, "lr": 4.446214645453469e-05}
{"train_loss": 0.12631826102733612, "global_step": 145547, "epoch": 1635, "lr": 4.446157028512781e-05}
{"train_loss": 0.13870522379875183, "global_step": 145548, "epoch": 1635, "lr": 4.446099411646549e-05}
{"train_loss": 0.15092270076274872, "global_step": 145549, "epoch": 1635, "lr": 4.446041794854783e-05}
{"train_loss": 0.07591219246387482, "global_step": 145550, "epoch": 1635, "lr": 4.4459841781374886e-05}
{"train_loss": 0.16475385427474976, "global_step": 145551, "epoch": 1635, "lr": 4.4459265614946774e-05}
{"train_loss": 0.1342622935771942, "global_step": 145552, "epoch": 1635, "lr": 4.4458689449263516e-05}
{"train_loss": 0.1280273050069809, "global_step": 145553, "epoch": 1635, "lr": 4.445811328432525e-05}
{"train_loss": 0.0886315107345581, "global_step": 145554, "epoch": 1635, "lr": 4.4457537120132e-05}
{"train_loss": 0.13446444272994995, "global_step": 145555, "epoch": 1635, "lr": 4.4456960956683865e-05}
{"train_loss": 0.11692608147859573, "global_step": 145556, "epoch": 1635, "lr": 4.445638479398094e-05}
{"train_loss": 0.12667520344257355, "global_step": 145557, "epoch": 1635, "lr": 4.445580863202328e-05}
{"train_loss": 0.09535371512174606, "global_step": 145558, "epoch": 1635, "lr": 4.445523247081097e-05}
{"train_loss": 0.09555144608020782, "global_step": 145559, "epoch": 1635, "lr": 4.445465631034408e-05}
{"train_loss": 0.13181544840335846, "global_step": 145560, "epoch": 1635, "lr": 4.44540801506227e-05}
{"train_loss": 0.12618358433246613, "global_step": 145561, "epoch": 1635, "lr": 4.445350399164691e-05}
{"train_loss": 0.13639649748802185, "global_step": 145562, "epoch": 1635, "lr": 4.4452927833416774e-05}
{"train_loss": 0.21732565760612488, "global_step": 145563, "epoch": 1635, "lr": 4.445235167593236e-05}
{"train_loss": 0.08695518225431442, "global_step": 145564, "epoch": 1635, "lr": 4.4451775519193777e-05}
{"train_loss": 0.0739646703004837, "global_step": 145565, "epoch": 1635, "lr": 4.445119936320107e-05}
{"train_loss": 0.05931609123945236, "global_step": 145566, "epoch": 1635, "lr": 4.445062320795435e-05}
{"train_loss": 0.13110515475273132, "global_step": 145567, "epoch": 1635, "lr": 4.445004705345365e-05}
{"train_loss": 0.11418106406927109, "global_step": 145568, "epoch": 1635, "lr": 4.4449470899699104e-05}
{"train_loss": 0.11531522125005722, "global_step": 145569, "epoch": 1635, "lr": 4.444889474669073e-05}
{"train_loss": 0.15584678947925568, "global_step": 145570, "epoch": 1635, "lr": 4.444831859442866e-05}
{"train_loss": 0.1335759311914444, "global_step": 145571, "epoch": 1635, "lr": 4.444774244291292e-05}
{"train_loss": 0.1788192242383957, "global_step": 145572, "epoch": 1635, "lr": 4.444716629214362e-05}
{"train_loss": 0.18000352382659912, "global_step": 145573, "epoch": 1635, "lr": 4.444659014212085e-05}
{"train_loss": 0.12513597309589386, "global_step": 145574, "epoch": 1635, "lr": 4.4446013992844644e-05}
{"train_loss": 0.13632456958293915, "global_step": 145575, "epoch": 1635, "lr": 4.444543784431513e-05}
{"train_loss": 0.13320715725421906, "global_step": 145576, "epoch": 1635, "lr": 4.444486169653234e-05}
{"train_loss": 0.07726156711578369, "global_step": 145577, "epoch": 1635, "lr": 4.444428554949638e-05}
{"train_loss": 0.1508452445268631, "global_step": 145578, "epoch": 1635, "lr": 4.4443709403207314e-05}
{"train_loss": 0.10594646632671356, "global_step": 145579, "epoch": 1635, "lr": 4.444313325766522e-05}
{"train_loss": 0.07489065080881119, "global_step": 145580, "epoch": 1635, "lr": 4.4442557112870186e-05}
{"train_loss": 0.12697277963161469, "global_step": 145581, "epoch": 1635, "lr": 4.444198096882229e-05}
{"train_loss": 0.1318562775850296, "global_step": 145582, "epoch": 1635, "lr": 4.4441404825521585e-05}
{"train_loss": 0.10020381212234497, "global_step": 145583, "epoch": 1635, "lr": 4.4440828682968174e-05}
{"train_loss": 0.0839846283197403, "global_step": 145584, "epoch": 1635, "lr": 4.444025254116212e-05}
{"train_loss": 0.1730198711156845, "global_step": 145585, "epoch": 1635, "lr": 4.4439676400103516e-05}
{"train_loss": 0.12326324731111526, "global_step": 145586, "epoch": 1635, "lr": 4.443910025979241e-05}
{"train_loss": 0.16559313237667084, "global_step": 145587, "epoch": 1635, "lr": 4.4438524120228924e-05}
{"train_loss": 0.10665130615234375, "global_step": 145588, "epoch": 1635, "lr": 4.443794798141308e-05}
{"train_loss": 0.17798547446727753, "global_step": 145589, "epoch": 1635, "lr": 4.443737184334503e-05}
{"train_loss": 0.09529919177293777, "global_step": 145590, "epoch": 1635, "lr": 4.443679570602477e-05}
{"train_loss": 0.19935818016529083, "global_step": 145591, "epoch": 1635, "lr": 4.4436219569452416e-05}
{"train_loss": 0.11106454581022263, "global_step": 145592, "epoch": 1635, "lr": 4.4435643433628064e-05}
{"train_loss": 0.08926302939653397, "global_step": 145593, "epoch": 1635, "lr": 4.443506729855175e-05}
{"train_loss": 0.14939242601394653, "global_step": 145594, "epoch": 1635, "lr": 4.44344911642236e-05}
{"train_loss": 0.07126861810684204, "global_step": 145595, "epoch": 1635, "lr": 4.4433915030643656e-05}
{"train_loss": 0.11938659846782684, "global_step": 145596, "epoch": 1635, "lr": 4.4433338897812e-05}
{"train_loss": 0.04255922511219978, "global_step": 145597, "epoch": 1635, "lr": 4.4432762765728716e-05}
{"train_loss": 0.06045025214552879, "global_step": 145598, "epoch": 1635, "lr": 4.443218663439388e-05}
{"train_loss": 0.12935389578342438, "global_step": 145599, "epoch": 1635, "lr": 4.4431610503807564e-05}
{"train_loss": 0.16131922602653503, "global_step": 145600, "epoch": 1635, "lr": 4.443103437396986e-05}
{"train_loss": 0.1786162555217743, "global_step": 145601, "epoch": 1635, "lr": 4.4430458244880816e-05}
{"train_loss": 0.10314695537090302, "global_step": 145602, "epoch": 1635, "lr": 4.4429882116540556e-05}
{"train_loss": 0.125897122986531, "global_step": 145603, "epoch": 1635, "lr": 4.44293059889491e-05, "val_loss": 6.134518146514893, "train_action_mse_error": 30.14325714111328}
{"train_loss": 0.1376357227563858, "global_step": 145604, "epoch": 1636, "lr": 4.4428729862106586e-05}
{"train_loss": 0.13221172988414764, "global_step": 145605, "epoch": 1636, "lr": 4.442815373601303e-05}
{"train_loss": 0.08277000486850739, "global_step": 145606, "epoch": 1636, "lr": 4.442757761066857e-05}
{"train_loss": 0.0964183509349823, "global_step": 145607, "epoch": 1636, "lr": 4.4427001486073234e-05}
{"train_loss": 0.10014931112527847, "global_step": 145608, "epoch": 1636, "lr": 4.4426425362227135e-05}
{"train_loss": 0.11382127553224564, "global_step": 145609, "epoch": 1636, "lr": 4.442584923913032e-05}
{"train_loss": 0.10140476375818253, "global_step": 145610, "epoch": 1636, "lr": 4.442527311678288e-05}
{"train_loss": 0.12655554711818695, "global_step": 145611, "epoch": 1636, "lr": 4.442469699518491e-05}
{"train_loss": 0.06144605949521065, "global_step": 145612, "epoch": 1636, "lr": 4.442412087433645e-05}
{"train_loss": 0.14106760919094086, "global_step": 145613, "epoch": 1636, "lr": 4.442354475423763e-05}
{"train_loss": 0.14293886721134186, "global_step": 145614, "epoch": 1636, "lr": 4.442296863488848e-05}
{"train_loss": 0.09544262290000916, "global_step": 145615, "epoch": 1636, "lr": 4.4422392516289094e-05}
{"train_loss": 0.11709608137607574, "global_step": 145616, "epoch": 1636, "lr": 4.442181639843955e-05}
{"train_loss": 0.1407265067100525, "global_step": 145617, "epoch": 1636, "lr": 4.4421240281339924e-05}
{"train_loss": 0.12257330119609833, "global_step": 145618, "epoch": 1636, "lr": 4.4420664164990296e-05}
{"train_loss": 0.18221336603164673, "global_step": 145619, "epoch": 1636, "lr": 4.442008804939074e-05}
{"train_loss": 0.11669646203517914, "global_step": 145620, "epoch": 1636, "lr": 4.441951193454134e-05}
{"train_loss": 0.11318904161453247, "global_step": 145621, "epoch": 1636, "lr": 4.441893582044217e-05}
{"train_loss": 0.11043767631053925, "global_step": 145622, "epoch": 1636, "lr": 4.441835970709329e-05}
{"train_loss": 0.12024373561143875, "global_step": 145623, "epoch": 1636, "lr": 4.441778359449481e-05}
{"train_loss": 0.14461667835712433, "global_step": 145624, "epoch": 1636, "lr": 4.441720748264678e-05}
{"train_loss": 0.1099960058927536, "global_step": 145625, "epoch": 1636, "lr": 4.44166313715493e-05}
{"train_loss": 0.09906448423862457, "global_step": 145626, "epoch": 1636, "lr": 4.4416055261202416e-05}
{"train_loss": 0.24917933344841003, "global_step": 145627, "epoch": 1636, "lr": 4.4415479151606224e-05}
{"train_loss": 0.11157159507274628, "global_step": 145628, "epoch": 1636, "lr": 4.441490304276083e-05}
{"train_loss": 0.10542070865631104, "global_step": 145629, "epoch": 1636, "lr": 4.441432693466626e-05}
{"train_loss": 0.07870209217071533, "global_step": 145630, "epoch": 1636, "lr": 4.441375082732264e-05}
{"train_loss": 0.09792274981737137, "global_step": 145631, "epoch": 1636, "lr": 4.441317472073e-05}
{"train_loss": 0.10203224420547485, "global_step": 145632, "epoch": 1636, "lr": 4.441259861488846e-05}
{"train_loss": 0.08983476459980011, "global_step": 145633, "epoch": 1636, "lr": 4.4412022509798056e-05}
{"train_loss": 0.08273781090974808, "global_step": 145634, "epoch": 1636, "lr": 4.441144640545891e-05}
{"train_loss": 0.07676994800567627, "global_step": 145635, "epoch": 1636, "lr": 4.4410870301871073e-05}
{"train_loss": 0.12978073954582214, "global_step": 145636, "epoch": 1636, "lr": 4.441029419903462e-05}
{"train_loss": 0.06849473714828491, "global_step": 145637, "epoch": 1636, "lr": 4.4409718096949635e-05}
{"train_loss": 0.10057082772254944, "global_step": 145638, "epoch": 1636, "lr": 4.4409141995616194e-05}
{"train_loss": 0.11941662430763245, "global_step": 145639, "epoch": 1636, "lr": 4.440856589503438e-05}
{"train_loss": 0.11252886801958084, "global_step": 145640, "epoch": 1636, "lr": 4.440798979520427e-05}
{"train_loss": 0.06346718221902847, "global_step": 145641, "epoch": 1636, "lr": 4.440741369612592e-05}
{"train_loss": 0.144249826669693, "global_step": 145642, "epoch": 1636, "lr": 4.440683759779944e-05}
{"train_loss": 0.1202797219157219, "global_step": 145643, "epoch": 1636, "lr": 4.4406261500224884e-05}
{"train_loss": 0.16729004681110382, "global_step": 145644, "epoch": 1636, "lr": 4.4405685403402355e-05}
{"train_loss": 0.07036406546831131, "global_step": 145645, "epoch": 1636, "lr": 4.440510930733189e-05}
{"train_loss": 0.11992087960243225, "global_step": 145646, "epoch": 1636, "lr": 4.440453321201359e-05}
{"train_loss": 0.12446079403162003, "global_step": 145647, "epoch": 1636, "lr": 4.440395711744755e-05}
{"train_loss": 0.11015740036964417, "global_step": 145648, "epoch": 1636, "lr": 4.440338102363381e-05}
{"train_loss": 0.1205187737941742, "global_step": 145649, "epoch": 1636, "lr": 4.440280493057249e-05}
{"train_loss": 0.04019559547305107, "global_step": 145650, "epoch": 1636, "lr": 4.4402228838263625e-05}
{"train_loss": 0.07571601122617722, "global_step": 145651, "epoch": 1636, "lr": 4.440165274670734e-05}
{"train_loss": 0.09383126348257065, "global_step": 145652, "epoch": 1636, "lr": 4.440107665590365e-05}
{"train_loss": 0.1381225883960724, "global_step": 145653, "epoch": 1636, "lr": 4.440050056585269e-05}
{"train_loss": 0.09935140609741211, "global_step": 145654, "epoch": 1636, "lr": 4.439992447655451e-05}
{"train_loss": 0.10490426421165466, "global_step": 145655, "epoch": 1636, "lr": 4.439934838800919e-05}
{"train_loss": 0.13244041800498962, "global_step": 145656, "epoch": 1636, "lr": 4.4398772300216804e-05}
{"train_loss": 0.14816918969154358, "global_step": 145657, "epoch": 1636, "lr": 4.4398196213177444e-05}
{"train_loss": 0.08544463664293289, "global_step": 145658, "epoch": 1636, "lr": 4.439762012689118e-05}
{"train_loss": 0.11619438976049423, "global_step": 145659, "epoch": 1636, "lr": 4.439704404135809e-05}
{"train_loss": 0.14431031048297882, "global_step": 145660, "epoch": 1636, "lr": 4.4396467956578225e-05}
{"train_loss": 0.1221901997923851, "global_step": 145661, "epoch": 1636, "lr": 4.4395891872551714e-05}
{"train_loss": 0.14715704321861267, "global_step": 145662, "epoch": 1636, "lr": 4.439531578927859e-05}
{"train_loss": 0.08131293207406998, "global_step": 145663, "epoch": 1636, "lr": 4.4394739706758935e-05}
{"train_loss": 0.09358200430870056, "global_step": 145664, "epoch": 1636, "lr": 4.439416362499287e-05}
{"train_loss": 0.051808569580316544, "global_step": 145665, "epoch": 1636, "lr": 4.439358754398042e-05}
{"train_loss": 0.10965683311223984, "global_step": 145666, "epoch": 1636, "lr": 4.43930114637217e-05}
{"train_loss": 0.1514861136674881, "global_step": 145667, "epoch": 1636, "lr": 4.4392435384216755e-05}
{"train_loss": 0.1299324929714203, "global_step": 145668, "epoch": 1636, "lr": 4.4391859305465694e-05}
{"train_loss": 0.10970869660377502, "global_step": 145669, "epoch": 1636, "lr": 4.439128322746856e-05}
{"train_loss": 0.11834470182657242, "global_step": 145670, "epoch": 1636, "lr": 4.4390707150225475e-05}
{"train_loss": 0.11370214819908142, "global_step": 145671, "epoch": 1636, "lr": 4.4390131073736466e-05}
{"train_loss": 0.11449220031499863, "global_step": 145672, "epoch": 1636, "lr": 4.4389554998001656e-05}
{"train_loss": 0.07049732655286789, "global_step": 145673, "epoch": 1636, "lr": 4.43889789230211e-05}
{"train_loss": 0.10082738101482391, "global_step": 145674, "epoch": 1636, "lr": 4.4388402848794866e-05}
{"train_loss": 0.0704115629196167, "global_step": 145675, "epoch": 1636, "lr": 4.438782677532305e-05}
{"train_loss": 0.10863382369279861, "global_step": 145676, "epoch": 1636, "lr": 4.4387250702605716e-05}
{"train_loss": 0.1288018822669983, "global_step": 145677, "epoch": 1636, "lr": 4.438667463064295e-05}
{"train_loss": 0.09260202944278717, "global_step": 145678, "epoch": 1636, "lr": 4.438609855943484e-05}
{"train_loss": 0.11843929439783096, "global_step": 145679, "epoch": 1636, "lr": 4.438552248898143e-05}
{"train_loss": 0.09283870458602905, "global_step": 145680, "epoch": 1636, "lr": 4.438494641928284e-05}
{"train_loss": 0.08800812810659409, "global_step": 145681, "epoch": 1636, "lr": 4.43843703503391e-05}
{"train_loss": 0.05349584296345711, "global_step": 145682, "epoch": 1636, "lr": 4.438379428215032e-05}
{"train_loss": 0.18066492676734924, "global_step": 145683, "epoch": 1636, "lr": 4.438321821471658e-05}
{"train_loss": 0.13430173695087433, "global_step": 145684, "epoch": 1636, "lr": 4.438264214803793e-05}
{"train_loss": 0.12322543561458588, "global_step": 145685, "epoch": 1636, "lr": 4.43820660821145e-05}
{"train_loss": 0.14011140167713165, "global_step": 145686, "epoch": 1636, "lr": 4.43814900169463e-05}
{"train_loss": 0.15795695781707764, "global_step": 145687, "epoch": 1636, "lr": 4.438091395253346e-05}
{"train_loss": 0.10960575193166733, "global_step": 145688, "epoch": 1636, "lr": 4.438033788887602e-05}
{"train_loss": 0.1381235122680664, "global_step": 145689, "epoch": 1636, "lr": 4.43797618259741e-05}
{"train_loss": 0.09743928164243698, "global_step": 145690, "epoch": 1636, "lr": 4.4379185763827724e-05}
{"train_loss": 0.10572374612092972, "global_step": 145691, "epoch": 1636, "lr": 4.437860970243702e-05}
{"train_loss": 0.11280337729480829, "global_step": 145692, "epoch": 1636, "lr": 4.437803364180204e-05, "val_loss": 6.033027648925781}
{"train_loss": 0.05855659395456314, "global_step": 145693, "epoch": 1637, "lr": 4.437745758192285e-05}
{"train_loss": 0.13333995640277863, "global_step": 145694, "epoch": 1637, "lr": 4.437688152279956e-05}
{"train_loss": 0.05990443378686905, "global_step": 145695, "epoch": 1637, "lr": 4.437630546443222e-05}
{"train_loss": 0.07632128149271011, "global_step": 145696, "epoch": 1637, "lr": 4.437572940682092e-05}
{"train_loss": 0.0905039981007576, "global_step": 145697, "epoch": 1637, "lr": 4.437515334996574e-05}
{"train_loss": 0.09886571019887924, "global_step": 145698, "epoch": 1637, "lr": 4.4374577293866735e-05}
{"train_loss": 0.11164838075637817, "global_step": 145699, "epoch": 1637, "lr": 4.4374001238524e-05}
{"train_loss": 0.07317257672548294, "global_step": 145700, "epoch": 1637, "lr": 4.4373425183937636e-05}
{"train_loss": 0.09674708545207977, "global_step": 145701, "epoch": 1637, "lr": 4.4372849130107666e-05}
{"train_loss": 0.06742151081562042, "global_step": 145702, "epoch": 1637, "lr": 4.437227307703423e-05}
{"train_loss": 0.14524883031845093, "global_step": 145703, "epoch": 1637, "lr": 4.437169702471735e-05}
{"train_loss": 0.11668287962675095, "global_step": 145704, "epoch": 1637, "lr": 4.437112097315714e-05}
{"train_loss": 0.05787235125899315, "global_step": 145705, "epoch": 1637, "lr": 4.437054492235365e-05}
{"train_loss": 0.1098637580871582, "global_step": 145706, "epoch": 1637, "lr": 4.436996887230699e-05}
{"train_loss": 0.1225779578089714, "global_step": 145707, "epoch": 1637, "lr": 4.43693928230172e-05}
{"train_loss": 0.12083785235881805, "global_step": 145708, "epoch": 1637, "lr": 4.43688167744844e-05}
{"train_loss": 0.14522822201251984, "global_step": 145709, "epoch": 1637, "lr": 4.436824072670861e-05}
{"train_loss": 0.10948808491230011, "global_step": 145710, "epoch": 1637, "lr": 4.436766467968998e-05}
{"train_loss": 0.12532636523246765, "global_step": 145711, "epoch": 1637, "lr": 4.436708863342853e-05}
{"train_loss": 0.13628999888896942, "global_step": 145712, "epoch": 1637, "lr": 4.436651258792435e-05}
{"train_loss": 0.08948030322790146, "global_step": 145713, "epoch": 1637, "lr": 4.4365936543177536e-05}
{"train_loss": 0.13662920892238617, "global_step": 145714, "epoch": 1637, "lr": 4.4365360499188145e-05}
{"train_loss": 0.08842217177152634, "global_step": 145715, "epoch": 1637, "lr": 4.4364784455956266e-05}
{"train_loss": 0.09990862756967545, "global_step": 145716, "epoch": 1637, "lr": 4.436420841348198e-05}
{"train_loss": 0.13077598810195923, "global_step": 145717, "epoch": 1637, "lr": 4.436363237176534e-05}
{"train_loss": 0.12085666507482529, "global_step": 145718, "epoch": 1637, "lr": 4.436305633080644e-05}
{"train_loss": 0.09554214775562286, "global_step": 145719, "epoch": 1637, "lr": 4.436248029060538e-05}
{"train_loss": 0.10146987438201904, "global_step": 145720, "epoch": 1637, "lr": 4.4361904251162186e-05}
{"train_loss": 0.08756453543901443, "global_step": 145721, "epoch": 1637, "lr": 4.4361328212477e-05}
{"train_loss": 0.058068305253982544, "global_step": 145722, "epoch": 1637, "lr": 4.4360752174549834e-05}
{"train_loss": 0.16173717379570007, "global_step": 145723, "epoch": 1637, "lr": 4.4360176137380826e-05}
{"train_loss": 0.15177568793296814, "global_step": 145724, "epoch": 1637, "lr": 4.4359600100969995e-05}
{"train_loss": 0.1279911994934082, "global_step": 145725, "epoch": 1637, "lr": 4.4359024065317474e-05}
{"train_loss": 0.15107131004333496, "global_step": 145726, "epoch": 1637, "lr": 4.4358448030423286e-05}
{"train_loss": 0.1184043437242508, "global_step": 145727, "epoch": 1637, "lr": 4.435787199628756e-05}
{"train_loss": 0.08763124793767929, "global_step": 145728, "epoch": 1637, "lr": 4.435729596291033e-05}
{"train_loss": 0.12538675963878632, "global_step": 145729, "epoch": 1637, "lr": 4.435671993029171e-05}
{"train_loss": 0.12371449172496796, "global_step": 145730, "epoch": 1637, "lr": 4.435614389843174e-05}
{"train_loss": 0.08666902780532837, "global_step": 145731, "epoch": 1637, "lr": 4.435556786733054e-05}
{"train_loss": 0.1321559101343155, "global_step": 145732, "epoch": 1637, "lr": 4.435499183698816e-05}
{"train_loss": 0.08083688467741013, "global_step": 145733, "epoch": 1637, "lr": 4.435441580740468e-05}
{"train_loss": 0.11581454426050186, "global_step": 145734, "epoch": 1637, "lr": 4.435383977858017e-05}
{"train_loss": 0.12452244758605957, "global_step": 145735, "epoch": 1637, "lr": 4.435326375051473e-05}
{"train_loss": 0.11424437165260315, "global_step": 145736, "epoch": 1637, "lr": 4.435268772320842e-05}
{"train_loss": 0.05885901302099228, "global_step": 145737, "epoch": 1637, "lr": 4.435211169666131e-05}
{"train_loss": 0.09473468363285065, "global_step": 145738, "epoch": 1637, "lr": 4.4351535670873506e-05}
{"train_loss": 0.1346975415945053, "global_step": 145739, "epoch": 1637, "lr": 4.4350959645845055e-05}
{"train_loss": 0.056369077414274216, "global_step": 145740, "epoch": 1637, "lr": 4.435038362157607e-05}
{"train_loss": 0.11169960349798203, "global_step": 145741, "epoch": 1637, "lr": 4.4349807598066585e-05}
{"train_loss": 0.09104692190885544, "global_step": 145742, "epoch": 1637, "lr": 4.434923157531672e-05}
{"train_loss": 0.08617480844259262, "global_step": 145743, "epoch": 1637, "lr": 4.434865555332651e-05}
{"train_loss": 0.11118151992559433, "global_step": 145744, "epoch": 1637, "lr": 4.4348079532096074e-05}
{"train_loss": 0.08084988594055176, "global_step": 145745, "epoch": 1637, "lr": 4.4347503511625454e-05}
{"train_loss": 0.09991922974586487, "global_step": 145746, "epoch": 1637, "lr": 4.4346927491914765e-05}
{"train_loss": 0.15242846310138702, "global_step": 145747, "epoch": 1637, "lr": 4.434635147296404e-05}
{"train_loss": 0.1267346441745758, "global_step": 145748, "epoch": 1637, "lr": 4.4345775454773395e-05}
{"train_loss": 0.06765749305486679, "global_step": 145749, "epoch": 1637, "lr": 4.434519943734288e-05}
{"train_loss": 0.08523879945278168, "global_step": 145750, "epoch": 1637, "lr": 4.4344623420672596e-05}
{"train_loss": 0.17571960389614105, "global_step": 145751, "epoch": 1637, "lr": 4.43440474047626e-05}
{"train_loss": 0.171725332736969, "global_step": 145752, "epoch": 1637, "lr": 4.434347138961298e-05}
{"train_loss": 0.09779971092939377, "global_step": 145753, "epoch": 1637, "lr": 4.434289537522381e-05}
{"train_loss": 0.07866093516349792, "global_step": 145754, "epoch": 1637, "lr": 4.4342319361595156e-05}
{"train_loss": 0.10484406352043152, "global_step": 145755, "epoch": 1637, "lr": 4.434174334872713e-05}
{"train_loss": 0.1337289810180664, "global_step": 145756, "epoch": 1637, "lr": 4.434116733661977e-05}
{"train_loss": 0.08971016108989716, "global_step": 145757, "epoch": 1637, "lr": 4.434059132527319e-05}
{"train_loss": 0.08005047589540482, "global_step": 145758, "epoch": 1637, "lr": 4.434001531468743e-05}
{"train_loss": 0.0778750330209732, "global_step": 145759, "epoch": 1637, "lr": 4.4339439304862605e-05}
{"train_loss": 0.1225530281662941, "global_step": 145760, "epoch": 1637, "lr": 4.433886329579875e-05}
{"train_loss": 0.11027845740318298, "global_step": 145761, "epoch": 1637, "lr": 4.4338287287495986e-05}
{"train_loss": 0.1219269260764122, "global_step": 145762, "epoch": 1637, "lr": 4.433771127995436e-05}
{"train_loss": 0.09337697923183441, "global_step": 145763, "epoch": 1637, "lr": 4.433713527317397e-05}
{"train_loss": 0.14407901465892792, "global_step": 145764, "epoch": 1637, "lr": 4.433655926715486e-05}
{"train_loss": 0.13178546726703644, "global_step": 145765, "epoch": 1637, "lr": 4.4335983261897163e-05}
{"train_loss": 0.06915432959794998, "global_step": 145766, "epoch": 1637, "lr": 4.433540725740089e-05}
{"train_loss": 0.08920945972204208, "global_step": 145767, "epoch": 1637, "lr": 4.433483125366619e-05}
{"train_loss": 0.11125282198190689, "global_step": 145768, "epoch": 1637, "lr": 4.4334255250693065e-05}
{"train_loss": 0.07532300055027008, "global_step": 145769, "epoch": 1637, "lr": 4.433367924848166e-05}
{"train_loss": 0.12402724474668503, "global_step": 145770, "epoch": 1637, "lr": 4.4333103247032015e-05}
{"train_loss": 0.05803769826889038, "global_step": 145771, "epoch": 1637, "lr": 4.433252724634421e-05}
{"train_loss": 0.14481908082962036, "global_step": 145772, "epoch": 1637, "lr": 4.433195124641833e-05}
{"train_loss": 0.07314765453338623, "global_step": 145773, "epoch": 1637, "lr": 4.433137524725444e-05}
{"train_loss": 0.09700515866279602, "global_step": 145774, "epoch": 1637, "lr": 4.4330799248852645e-05}
{"train_loss": 0.11730498820543289, "global_step": 145775, "epoch": 1637, "lr": 4.4330223251212985e-05}
{"train_loss": 0.08133556693792343, "global_step": 145776, "epoch": 1637, "lr": 4.432964725433558e-05}
{"train_loss": 0.07573102414608002, "global_step": 145777, "epoch": 1637, "lr": 4.4329071258220463e-05}
{"train_loss": 0.059151340276002884, "global_step": 145778, "epoch": 1637, "lr": 4.432849526286775e-05}
{"train_loss": 0.08724644035100937, "global_step": 145779, "epoch": 1637, "lr": 4.4327919268277484e-05}
{"train_loss": 0.11842790991067886, "global_step": 145780, "epoch": 1637, "lr": 4.4327343274449784e-05}
{"train_loss": 0.10547251844506585, "global_step": 145781, "epoch": 1637, "lr": 4.432676728138468e-05, "val_loss": 6.11961030960083}
{"train_loss": 0.10552412271499634, "global_step": 145782, "epoch": 1638, "lr": 4.43261912890823e-05}
{"train_loss": 0.09680365025997162, "global_step": 145783, "epoch": 1638, "lr": 4.432561529754266e-05}
{"train_loss": 0.12580358982086182, "global_step": 145784, "epoch": 1638, "lr": 4.4325039306765904e-05}
{"train_loss": 0.12517467141151428, "global_step": 145785, "epoch": 1638, "lr": 4.4324463316752054e-05}
{"train_loss": 0.06916412711143494, "global_step": 145786, "epoch": 1638, "lr": 4.432388732750123e-05}
{"train_loss": 0.1145521029829979, "global_step": 145787, "epoch": 1638, "lr": 4.4323311339013465e-05}
{"train_loss": 0.10793435573577881, "global_step": 145788, "epoch": 1638, "lr": 4.4322735351288894e-05}
{"train_loss": 0.09286696463823318, "global_step": 145789, "epoch": 1638, "lr": 4.432215936432754e-05}
{"train_loss": 0.11395365744829178, "global_step": 145790, "epoch": 1638, "lr": 4.43215833781295e-05}
{"train_loss": 0.07629396766424179, "global_step": 145791, "epoch": 1638, "lr": 4.432100739269487e-05}
{"train_loss": 0.09051711112260818, "global_step": 145792, "epoch": 1638, "lr": 4.4320431408023694e-05}
{"train_loss": 0.10406187921762466, "global_step": 145793, "epoch": 1638, "lr": 4.4319855424116084e-05}
{"train_loss": 0.12204066663980484, "global_step": 145794, "epoch": 1638, "lr": 4.4319279440972085e-05}
{"train_loss": 0.1296357363462448, "global_step": 145795, "epoch": 1638, "lr": 4.431870345859181e-05}
{"train_loss": 0.1016777902841568, "global_step": 145796, "epoch": 1638, "lr": 4.431812747697529e-05}
{"train_loss": 0.13252083957195282, "global_step": 145797, "epoch": 1638, "lr": 4.431755149612265e-05}
{"train_loss": 0.09205503761768341, "global_step": 145798, "epoch": 1638, "lr": 4.4316975516033926e-05}
{"train_loss": 0.12317431718111038, "global_step": 145799, "epoch": 1638, "lr": 4.431639953670923e-05}
{"train_loss": 0.1521531343460083, "global_step": 145800, "epoch": 1638, "lr": 4.4315823558148616e-05}
{"train_loss": 0.14024566113948822, "global_step": 145801, "epoch": 1638, "lr": 4.431524758035219e-05}
{"train_loss": 0.1063685417175293, "global_step": 145802, "epoch": 1638, "lr": 4.431467160331998e-05}
{"train_loss": 0.07172540575265884, "global_step": 145803, "epoch": 1638, "lr": 4.4314095627052124e-05}
{"train_loss": 0.10170275717973709, "global_step": 145804, "epoch": 1638, "lr": 4.431351965154864e-05}
{"train_loss": 0.12087134271860123, "global_step": 145805, "epoch": 1638, "lr": 4.431294367680966e-05}
{"train_loss": 0.04687410965561867, "global_step": 145806, "epoch": 1638, "lr": 4.4312367702835206e-05}
{"train_loss": 0.08367222547531128, "global_step": 145807, "epoch": 1638, "lr": 4.431179172962541e-05}
{"train_loss": 0.08138798177242279, "global_step": 145808, "epoch": 1638, "lr": 4.431121575718032e-05}
{"train_loss": 0.07960247993469238, "global_step": 145809, "epoch": 1638, "lr": 4.4310639785500004e-05}
{"train_loss": 0.10317901521921158, "global_step": 145810, "epoch": 1638, "lr": 4.4310063814584565e-05}
{"train_loss": 0.08176239579916, "global_step": 145811, "epoch": 1638, "lr": 4.430948784443406e-05}
{"train_loss": 0.07860822230577469, "global_step": 145812, "epoch": 1638, "lr": 4.430891187504859e-05}
{"train_loss": 0.06293603032827377, "global_step": 145813, "epoch": 1638, "lr": 4.43083359064282e-05}
{"train_loss": 0.13144254684448242, "global_step": 145814, "epoch": 1638, "lr": 4.4307759938573e-05}
{"train_loss": 0.1239565908908844, "global_step": 145815, "epoch": 1638, "lr": 4.4307183971483034e-05}
{"train_loss": 0.07328576594591141, "global_step": 145816, "epoch": 1638, "lr": 4.4306608005158426e-05}
{"train_loss": 0.08534003049135208, "global_step": 145817, "epoch": 1638, "lr": 4.43060320395992e-05}
{"train_loss": 0.09547918289899826, "global_step": 145818, "epoch": 1638, "lr": 4.4305456074805475e-05}
{"train_loss": 0.14525054395198822, "global_step": 145819, "epoch": 1638, "lr": 4.430488011077729e-05}
{"train_loss": 0.12745629251003265, "global_step": 145820, "epoch": 1638, "lr": 4.430430414751477e-05}
{"train_loss": 0.1294688731431961, "global_step": 145821, "epoch": 1638, "lr": 4.4303728185017946e-05}
{"train_loss": 0.0993436649441719, "global_step": 145822, "epoch": 1638, "lr": 4.430315222328693e-05}
{"train_loss": 0.05774357169866562, "global_step": 145823, "epoch": 1638, "lr": 4.430257626232177e-05}
{"train_loss": 0.1178397610783577, "global_step": 145824, "epoch": 1638, "lr": 4.430200030212258e-05}
{"train_loss": 0.12121985852718353, "global_step": 145825, "epoch": 1638, "lr": 4.43014243426894e-05}
{"train_loss": 0.11208786070346832, "global_step": 145826, "epoch": 1638, "lr": 4.430084838402233e-05}
{"train_loss": 0.10990177094936371, "global_step": 145827, "epoch": 1638, "lr": 4.4300272426121456e-05}
{"train_loss": 0.18922480940818787, "global_step": 145828, "epoch": 1638, "lr": 4.4299696468986816e-05}
{"train_loss": 0.14262248575687408, "global_step": 145829, "epoch": 1638, "lr": 4.429912051261854e-05}
{"train_loss": 0.13259467482566833, "global_step": 145830, "epoch": 1638, "lr": 4.429854455701664e-05}
{"train_loss": 0.13814619183540344, "global_step": 145831, "epoch": 1638, "lr": 4.429796860218127e-05}
{"train_loss": 0.1343018263578415, "global_step": 145832, "epoch": 1638, "lr": 4.429739264811244e-05}
{"train_loss": 0.1627807468175888, "global_step": 145833, "epoch": 1638, "lr": 4.429681669481028e-05}
{"train_loss": 0.06799201667308807, "global_step": 145834, "epoch": 1638, "lr": 4.429624074227481e-05}
{"train_loss": 0.09102802723646164, "global_step": 145835, "epoch": 1638, "lr": 4.429566479050618e-05}
{"train_loss": 0.10397104173898697, "global_step": 145836, "epoch": 1638, "lr": 4.429508883950439e-05}
{"train_loss": 0.07211307436227798, "global_step": 145837, "epoch": 1638, "lr": 4.4294512889269595e-05}
{"train_loss": 0.050334930419921875, "global_step": 145838, "epoch": 1638, "lr": 4.42939369398018e-05}
{"train_loss": 0.0678003653883934, "global_step": 145839, "epoch": 1638, "lr": 4.429336099110114e-05}
{"train_loss": 0.14481572806835175, "global_step": 145840, "epoch": 1638, "lr": 4.429278504316765e-05}
{"train_loss": 0.16419833898544312, "global_step": 145841, "epoch": 1638, "lr": 4.429220909600145e-05}
{"train_loss": 0.1589231789112091, "global_step": 145842, "epoch": 1638, "lr": 4.429163314960256e-05}
{"train_loss": 0.14014562964439392, "global_step": 145843, "epoch": 1638, "lr": 4.429105720397111e-05}
{"train_loss": 0.12687557935714722, "global_step": 145844, "epoch": 1638, "lr": 4.4290481259107155e-05}
{"train_loss": 0.10720442235469818, "global_step": 145845, "epoch": 1638, "lr": 4.428990531501077e-05}
{"train_loss": 0.13535703718662262, "global_step": 145846, "epoch": 1638, "lr": 4.4289329371682044e-05}
{"train_loss": 0.10568884015083313, "global_step": 145847, "epoch": 1638, "lr": 4.428875342912105e-05}
{"train_loss": 0.11936113983392715, "global_step": 145848, "epoch": 1638, "lr": 4.428817748732786e-05}
{"train_loss": 0.16720952093601227, "global_step": 145849, "epoch": 1638, "lr": 4.428760154630255e-05}
{"train_loss": 0.1267792135477066, "global_step": 145850, "epoch": 1638, "lr": 4.4287025606045216e-05}
{"train_loss": 0.12490510940551758, "global_step": 145851, "epoch": 1638, "lr": 4.42864496665559e-05}
{"train_loss": 0.1250123679637909, "global_step": 145852, "epoch": 1638, "lr": 4.4285873727834725e-05}
{"train_loss": 0.11608947813510895, "global_step": 145853, "epoch": 1638, "lr": 4.428529778988172e-05}
{"train_loss": 0.12267865985631943, "global_step": 145854, "epoch": 1638, "lr": 4.428472185269701e-05}
{"train_loss": 0.14052768051624298, "global_step": 145855, "epoch": 1638, "lr": 4.428414591628064e-05}
{"train_loss": 0.13036075234413147, "global_step": 145856, "epoch": 1638, "lr": 4.42835699806327e-05}
{"train_loss": 0.14273352921009064, "global_step": 145857, "epoch": 1638, "lr": 4.428299404575325e-05}
{"train_loss": 0.11376724392175674, "global_step": 145858, "epoch": 1638, "lr": 4.428241811164241e-05}
{"train_loss": 0.10969507694244385, "global_step": 145859, "epoch": 1638, "lr": 4.42818421783002e-05}
{"train_loss": 0.12480982393026352, "global_step": 145860, "epoch": 1638, "lr": 4.428126624572675e-05}
{"train_loss": 0.09568487852811813, "global_step": 145861, "epoch": 1638, "lr": 4.428069031392209e-05}
{"train_loss": 0.14068831503391266, "global_step": 145862, "epoch": 1638, "lr": 4.4280114382886344e-05}
{"train_loss": 0.0817948505282402, "global_step": 145863, "epoch": 1638, "lr": 4.427953845261955e-05}
{"train_loss": 0.16939157247543335, "global_step": 145864, "epoch": 1638, "lr": 4.4278962523121805e-05}
{"train_loss": 0.12796884775161743, "global_step": 145865, "epoch": 1638, "lr": 4.427838659439319e-05}
{"train_loss": 0.05520758405327797, "global_step": 145866, "epoch": 1638, "lr": 4.427781066643378e-05}
{"train_loss": 0.11677447706460953, "global_step": 145867, "epoch": 1638, "lr": 4.427723473924365e-05}
{"train_loss": 0.07335108518600464, "global_step": 145868, "epoch": 1638, "lr": 4.427665881282286e-05}
{"train_loss": 0.16187834739685059, "global_step": 145869, "epoch": 1638, "lr": 4.427608288717152e-05}
{"train_loss": 0.11208712050084317, "global_step": 145870, "epoch": 1638, "lr": 4.427550696228968e-05, "val_loss": 6.327461242675781}
{"train_loss": 0.08486037701368332, "global_step": 145871, "epoch": 1639, "lr": 4.427493103817744e-05}
{"train_loss": 0.0673939436674118, "global_step": 145872, "epoch": 1639, "lr": 4.427435511483484e-05}
{"train_loss": 0.15457040071487427, "global_step": 145873, "epoch": 1639, "lr": 4.427377919226202e-05}
{"train_loss": 0.12816664576530457, "global_step": 145874, "epoch": 1639, "lr": 4.4273203270458985e-05}
{"train_loss": 0.11261747777462006, "global_step": 145875, "epoch": 1639, "lr": 4.427262734942588e-05}
{"train_loss": 0.11705148965120316, "global_step": 145876, "epoch": 1639, "lr": 4.427205142916272e-05}
{"train_loss": 0.16431617736816406, "global_step": 145877, "epoch": 1639, "lr": 4.427147550966964e-05}
{"train_loss": 0.08034933358430862, "global_step": 145878, "epoch": 1639, "lr": 4.427089959094667e-05}
{"train_loss": 0.09189106523990631, "global_step": 145879, "epoch": 1639, "lr": 4.4270323672993926e-05}
{"train_loss": 0.08053983002901077, "global_step": 145880, "epoch": 1639, "lr": 4.426974775581144e-05}
{"train_loss": 0.12194596230983734, "global_step": 145881, "epoch": 1639, "lr": 4.426917183939934e-05}
{"train_loss": 0.10555107891559601, "global_step": 145882, "epoch": 1639, "lr": 4.426859592375767e-05}
{"train_loss": 0.09554322063922882, "global_step": 145883, "epoch": 1639, "lr": 4.4268020008886526e-05}
{"train_loss": 0.17516374588012695, "global_step": 145884, "epoch": 1639, "lr": 4.426744409478597e-05}
{"train_loss": 0.12887145578861237, "global_step": 145885, "epoch": 1639, "lr": 4.426686818145608e-05}
{"train_loss": 0.11426479369401932, "global_step": 145886, "epoch": 1639, "lr": 4.426629226889696e-05}
{"train_loss": 0.097300224006176, "global_step": 145887, "epoch": 1639, "lr": 4.426571635710865e-05}
{"train_loss": 0.07918120175600052, "global_step": 145888, "epoch": 1639, "lr": 4.426514044609126e-05}
{"train_loss": 0.09609478712081909, "global_step": 145889, "epoch": 1639, "lr": 4.426456453584483e-05}
{"train_loss": 0.05837429314851761, "global_step": 145890, "epoch": 1639, "lr": 4.4263988626369476e-05}
{"train_loss": 0.061605799943208694, "global_step": 145891, "epoch": 1639, "lr": 4.4263412717665246e-05}
{"train_loss": 0.17445619404315948, "global_step": 145892, "epoch": 1639, "lr": 4.426283680973225e-05}
{"train_loss": 0.11638005822896957, "global_step": 145893, "epoch": 1639, "lr": 4.426226090257052e-05}
{"train_loss": 0.11152700334787369, "global_step": 145894, "epoch": 1639, "lr": 4.426168499618019e-05}
{"train_loss": 0.12642821669578552, "global_step": 145895, "epoch": 1639, "lr": 4.426110909056127e-05}
{"train_loss": 0.09509631991386414, "global_step": 145896, "epoch": 1639, "lr": 4.426053318571391e-05}
{"train_loss": 0.14429648220539093, "global_step": 145897, "epoch": 1639, "lr": 4.425995728163812e-05}
{"train_loss": 0.13822583854198456, "global_step": 145898, "epoch": 1639, "lr": 4.4259381378334035e-05}
{"train_loss": 0.13636493682861328, "global_step": 145899, "epoch": 1639, "lr": 4.425880547580169e-05}
{"train_loss": 0.11805472522974014, "global_step": 145900, "epoch": 1639, "lr": 4.4258229574041176e-05}
{"train_loss": 0.16179020702838898, "global_step": 145901, "epoch": 1639, "lr": 4.4257653673052576e-05}
{"train_loss": 0.10705317556858063, "global_step": 145902, "epoch": 1639, "lr": 4.4257077772835975e-05}
{"train_loss": 0.09912779927253723, "global_step": 145903, "epoch": 1639, "lr": 4.425650187339143e-05}
{"train_loss": 0.14364992082118988, "global_step": 145904, "epoch": 1639, "lr": 4.4255925974719037e-05}
{"train_loss": 0.0641392171382904, "global_step": 145905, "epoch": 1639, "lr": 4.425535007681887e-05}
{"train_loss": 0.07103965431451797, "global_step": 145906, "epoch": 1639, "lr": 4.4254774179690976e-05}
{"train_loss": 0.07252533733844757, "global_step": 145907, "epoch": 1639, "lr": 4.4254198283335484e-05}
{"train_loss": 0.10940233618021011, "global_step": 145908, "epoch": 1639, "lr": 4.4253622387752425e-05}
{"train_loss": 0.10777867585420609, "global_step": 145909, "epoch": 1639, "lr": 4.4253046492941914e-05}
{"train_loss": 0.06048087403178215, "global_step": 145910, "epoch": 1639, "lr": 4.425247059890399e-05}
{"train_loss": 0.13745009899139404, "global_step": 145911, "epoch": 1639, "lr": 4.425189470563877e-05}
{"train_loss": 0.11924166232347488, "global_step": 145912, "epoch": 1639, "lr": 4.42513188131463e-05}
{"train_loss": 0.14110638201236725, "global_step": 145913, "epoch": 1639, "lr": 4.425074292142669e-05}
{"train_loss": 0.12284958362579346, "global_step": 145914, "epoch": 1639, "lr": 4.425016703047997e-05}
{"train_loss": 0.09377346932888031, "global_step": 145915, "epoch": 1639, "lr": 4.424959114030627e-05}
{"train_loss": 0.10047335922718048, "global_step": 145916, "epoch": 1639, "lr": 4.4249015250905617e-05}
{"train_loss": 0.13573095202445984, "global_step": 145917, "epoch": 1639, "lr": 4.4248439362278114e-05}
{"train_loss": 0.09178723394870758, "global_step": 145918, "epoch": 1639, "lr": 4.4247863474423864e-05}
{"train_loss": 0.15168748795986176, "global_step": 145919, "epoch": 1639, "lr": 4.424728758734291e-05}
{"train_loss": 0.1815139800310135, "global_step": 145920, "epoch": 1639, "lr": 4.4246711701035334e-05}
{"train_loss": 0.10040570050477982, "global_step": 145921, "epoch": 1639, "lr": 4.4246135815501216e-05}
{"train_loss": 0.14759257435798645, "global_step": 145922, "epoch": 1639, "lr": 4.4245559930740636e-05}
{"train_loss": 0.09694433212280273, "global_step": 145923, "epoch": 1639, "lr": 4.424498404675367e-05}
{"train_loss": 0.09302666038274765, "global_step": 145924, "epoch": 1639, "lr": 4.4244408163540407e-05}
{"train_loss": 0.1307254582643509, "global_step": 145925, "epoch": 1639, "lr": 4.424383228110089e-05}
{"train_loss": 0.08342255651950836, "global_step": 145926, "epoch": 1639, "lr": 4.424325639943524e-05}
{"train_loss": 0.10474736988544464, "global_step": 145927, "epoch": 1639, "lr": 4.42426805185435e-05}
{"train_loss": 0.07383013516664505, "global_step": 145928, "epoch": 1639, "lr": 4.424210463842578e-05}
{"train_loss": 0.07386823743581772, "global_step": 145929, "epoch": 1639, "lr": 4.424152875908212e-05}
{"train_loss": 0.1407710611820221, "global_step": 145930, "epoch": 1639, "lr": 4.424095288051263e-05}
{"train_loss": 0.08928435295820236, "global_step": 145931, "epoch": 1639, "lr": 4.424037700271736e-05}
{"train_loss": 0.09098272770643234, "global_step": 145932, "epoch": 1639, "lr": 4.4239801125696415e-05}
{"train_loss": 0.11204010248184204, "global_step": 145933, "epoch": 1639, "lr": 4.423922524944984e-05}
{"train_loss": 0.140994131565094, "global_step": 145934, "epoch": 1639, "lr": 4.423864937397775e-05}
{"train_loss": 0.07795698940753937, "global_step": 145935, "epoch": 1639, "lr": 4.4238073499280195e-05}
{"train_loss": 0.10070394724607468, "global_step": 145936, "epoch": 1639, "lr": 4.4237497625357245e-05}
{"train_loss": 0.056588683277368546, "global_step": 145937, "epoch": 1639, "lr": 4.423692175220902e-05}
{"train_loss": 0.1153305172920227, "global_step": 145938, "epoch": 1639, "lr": 4.4236345879835564e-05}
{"train_loss": 0.11172042042016983, "global_step": 145939, "epoch": 1639, "lr": 4.4235770008236954e-05}
{"train_loss": 0.14075882732868195, "global_step": 145940, "epoch": 1639, "lr": 4.423519413741328e-05}
{"train_loss": 0.12520262598991394, "global_step": 145941, "epoch": 1639, "lr": 4.4234618267364616e-05}
{"train_loss": 0.15727850794792175, "global_step": 145942, "epoch": 1639, "lr": 4.423404239809103e-05}
{"train_loss": 0.12277651578187943, "global_step": 145943, "epoch": 1639, "lr": 4.423346652959262e-05}
{"train_loss": 0.12718823552131653, "global_step": 145944, "epoch": 1639, "lr": 4.423289066186943e-05}
{"train_loss": 0.07874249666929245, "global_step": 145945, "epoch": 1639, "lr": 4.4232314794921574e-05}
{"train_loss": 0.1903354823589325, "global_step": 145946, "epoch": 1639, "lr": 4.42317389287491e-05}
{"train_loss": 0.14483679831027985, "global_step": 145947, "epoch": 1639, "lr": 4.423116306335212e-05}
{"train_loss": 0.1453613042831421, "global_step": 145948, "epoch": 1639, "lr": 4.4230587198730654e-05}
{"train_loss": 0.08262693136930466, "global_step": 145949, "epoch": 1639, "lr": 4.4230011334884855e-05}
{"train_loss": 0.06021134927868843, "global_step": 145950, "epoch": 1639, "lr": 4.422943547181473e-05}
{"train_loss": 0.08125875145196915, "global_step": 145951, "epoch": 1639, "lr": 4.422885960952041e-05}
{"train_loss": 0.11401332914829254, "global_step": 145952, "epoch": 1639, "lr": 4.422828374800192e-05}
{"train_loss": 0.06969963759183884, "global_step": 145953, "epoch": 1639, "lr": 4.422770788725938e-05}
{"train_loss": 0.0596502460539341, "global_step": 145954, "epoch": 1639, "lr": 4.422713202729287e-05}
{"train_loss": 0.13612963259220123, "global_step": 145955, "epoch": 1639, "lr": 4.422655616810243e-05}
{"train_loss": 0.07830257713794708, "global_step": 145956, "epoch": 1639, "lr": 4.422598030968817e-05}
{"train_loss": 0.08603422343730927, "global_step": 145957, "epoch": 1639, "lr": 4.422540445205017e-05}
{"train_loss": 0.09245742112398148, "global_step": 145958, "epoch": 1639, "lr": 4.422482859518848e-05}
{"train_loss": 0.11048143905367744, "global_step": 145959, "epoch": 1639, "lr": 4.422425273910319e-05, "val_loss": 6.350788593292236}
{"train_loss": 0.11771880090236664, "global_step": 145960, "epoch": 1640, "lr": 4.4223676883794376e-05}
{"train_loss": 0.1723490059375763, "global_step": 145961, "epoch": 1640, "lr": 4.4223101029262125e-05}
{"train_loss": 0.10215013474225998, "global_step": 145962, "epoch": 1640, "lr": 4.422252517550651e-05}
{"train_loss": 0.1552024483680725, "global_step": 145963, "epoch": 1640, "lr": 4.42219493225276e-05}
{"train_loss": 0.12946817278862, "global_step": 145964, "epoch": 1640, "lr": 4.422137347032549e-05}
{"train_loss": 0.08768797665834427, "global_step": 145965, "epoch": 1640, "lr": 4.422079761890023e-05}
{"train_loss": 0.09787045419216156, "global_step": 145966, "epoch": 1640, "lr": 4.422022176825193e-05}
{"train_loss": 0.09765520691871643, "global_step": 145967, "epoch": 1640, "lr": 4.421964591838063e-05}
{"train_loss": 0.09981047362089157, "global_step": 145968, "epoch": 1640, "lr": 4.4219070069286444e-05}
{"train_loss": 0.09535190463066101, "global_step": 145969, "epoch": 1640, "lr": 4.421849422096942e-05}
{"train_loss": 0.08631208539009094, "global_step": 145970, "epoch": 1640, "lr": 4.4217918373429665e-05}
{"train_loss": 0.18066349625587463, "global_step": 145971, "epoch": 1640, "lr": 4.421734252666723e-05}
{"train_loss": 0.1189715638756752, "global_step": 145972, "epoch": 1640, "lr": 4.421676668068219e-05}
{"train_loss": 0.09565382450819016, "global_step": 145973, "epoch": 1640, "lr": 4.421619083547466e-05}
{"train_loss": 0.2005700320005417, "global_step": 145974, "epoch": 1640, "lr": 4.421561499104467e-05}
{"train_loss": 0.18723921477794647, "global_step": 145975, "epoch": 1640, "lr": 4.421503914739235e-05}
{"train_loss": 0.11244723200798035, "global_step": 145976, "epoch": 1640, "lr": 4.4214463304517717e-05}
{"train_loss": 0.15649212896823883, "global_step": 145977, "epoch": 1640, "lr": 4.4213887462420905e-05}
{"train_loss": 0.12223673611879349, "global_step": 145978, "epoch": 1640, "lr": 4.421331162110195e-05}
{"train_loss": 0.09199924021959305, "global_step": 145979, "epoch": 1640, "lr": 4.4212735780560946e-05}
{"train_loss": 0.13285069167613983, "global_step": 145980, "epoch": 1640, "lr": 4.4212159940797974e-05}
{"train_loss": 0.18958841264247894, "global_step": 145981, "epoch": 1640, "lr": 4.42115841018131e-05}
{"train_loss": 0.10195459425449371, "global_step": 145982, "epoch": 1640, "lr": 4.4211008263606414e-05}
{"train_loss": 0.1142590120434761, "global_step": 145983, "epoch": 1640, "lr": 4.4210432426177994e-05}
{"train_loss": 0.058048028498888016, "global_step": 145984, "epoch": 1640, "lr": 4.4209856589527895e-05}
{"train_loss": 0.11122094094753265, "global_step": 145985, "epoch": 1640, "lr": 4.420928075365622e-05}
{"train_loss": 0.07822102308273315, "global_step": 145986, "epoch": 1640, "lr": 4.420870491856303e-05}
{"train_loss": 0.07884997874498367, "global_step": 145987, "epoch": 1640, "lr": 4.4208129084248425e-05}
{"train_loss": 0.11245612800121307, "global_step": 145988, "epoch": 1640, "lr": 4.420755325071244e-05}
{"train_loss": 0.07673366367816925, "global_step": 145989, "epoch": 1640, "lr": 4.4206977417955203e-05}
{"train_loss": 0.05752227082848549, "global_step": 145990, "epoch": 1640, "lr": 4.420640158597675e-05}
{"train_loss": 0.15014605224132538, "global_step": 145991, "epoch": 1640, "lr": 4.420582575477717e-05}
{"train_loss": 0.09442158043384552, "global_step": 145992, "epoch": 1640, "lr": 4.420524992435657e-05}
{"train_loss": 0.1603955328464508, "global_step": 145993, "epoch": 1640, "lr": 4.420467409471498e-05}
{"train_loss": 0.11409773677587509, "global_step": 145994, "epoch": 1640, "lr": 4.4204098265852526e-05}
{"train_loss": 0.11555968225002289, "global_step": 145995, "epoch": 1640, "lr": 4.420352243776924e-05}
{"train_loss": 0.1070985496044159, "global_step": 145996, "epoch": 1640, "lr": 4.420294661046523e-05}
{"train_loss": 0.1231272965669632, "global_step": 145997, "epoch": 1640, "lr": 4.420237078394056e-05}
{"train_loss": 0.1538098156452179, "global_step": 145998, "epoch": 1640, "lr": 4.420179495819531e-05}
{"train_loss": 0.12766481935977936, "global_step": 145999, "epoch": 1640, "lr": 4.4201219133229567e-05}
{"train_loss": 0.07945843040943146, "global_step": 146000, "epoch": 1640, "lr": 4.420064330904339e-05}
{"train_loss": 0.0764276459813118, "global_step": 146001, "epoch": 1640, "lr": 4.420006748563687e-05}
{"train_loss": 0.1496247947216034, "global_step": 146002, "epoch": 1640, "lr": 4.4199491663010084e-05}
{"train_loss": 0.07608842104673386, "global_step": 146003, "epoch": 1640, "lr": 4.4198915841163086e-05}
{"train_loss": 0.10947015136480331, "global_step": 146004, "epoch": 1640, "lr": 4.4198340020095995e-05}
{"train_loss": 0.08328734338283539, "global_step": 146005, "epoch": 1640, "lr": 4.419776419980885e-05}
{"train_loss": 0.09898696094751358, "global_step": 146006, "epoch": 1640, "lr": 4.4197188380301766e-05}
{"train_loss": 0.11595775932073593, "global_step": 146007, "epoch": 1640, "lr": 4.4196612561574776e-05}
{"train_loss": 0.14214284718036652, "global_step": 146008, "epoch": 1640, "lr": 4.419603674362798e-05}
{"train_loss": 0.1362452507019043, "global_step": 146009, "epoch": 1640, "lr": 4.4195460926461474e-05}
{"train_loss": 0.10469301789999008, "global_step": 146010, "epoch": 1640, "lr": 4.4194885110075304e-05}
{"train_loss": 0.07843577861785889, "global_step": 146011, "epoch": 1640, "lr": 4.419430929446958e-05}
{"train_loss": 0.1261233240365982, "global_step": 146012, "epoch": 1640, "lr": 4.419373347964433e-05}
{"train_loss": 0.10830570757389069, "global_step": 146013, "epoch": 1640, "lr": 4.4193157665599694e-05}
{"train_loss": 0.1106579452753067, "global_step": 146014, "epoch": 1640, "lr": 4.419258185233569e-05}
{"train_loss": 0.10094137489795685, "global_step": 146015, "epoch": 1640, "lr": 4.419200603985245e-05}
{"train_loss": 0.06231575459241867, "global_step": 146016, "epoch": 1640, "lr": 4.4191430228150005e-05}
{"train_loss": 0.06525686383247375, "global_step": 146017, "epoch": 1640, "lr": 4.419085441722846e-05}
{"train_loss": 0.10483104735612869, "global_step": 146018, "epoch": 1640, "lr": 4.4190278607087876e-05}
{"train_loss": 0.14454615116119385, "global_step": 146019, "epoch": 1640, "lr": 4.418970279772834e-05}
{"train_loss": 0.0766529068350792, "global_step": 146020, "epoch": 1640, "lr": 4.418912698914993e-05}
{"train_loss": 0.09370461106300354, "global_step": 146021, "epoch": 1640, "lr": 4.4188551181352734e-05}
{"train_loss": 0.08984001725912094, "global_step": 146022, "epoch": 1640, "lr": 4.418797537433679e-05}
{"train_loss": 0.06823531538248062, "global_step": 146023, "epoch": 1640, "lr": 4.418739956810222e-05}
{"train_loss": 0.0697997584939003, "global_step": 146024, "epoch": 1640, "lr": 4.418682376264907e-05}
{"train_loss": 0.15496045351028442, "global_step": 146025, "epoch": 1640, "lr": 4.418624795797745e-05}
{"train_loss": 0.1513102650642395, "global_step": 146026, "epoch": 1640, "lr": 4.41856721540874e-05}
{"train_loss": 0.0646611824631691, "global_step": 146027, "epoch": 1640, "lr": 4.4185096350979005e-05}
{"train_loss": 0.08988252282142639, "global_step": 146028, "epoch": 1640, "lr": 4.418452054865238e-05}
{"train_loss": 0.13303843140602112, "global_step": 146029, "epoch": 1640, "lr": 4.418394474710756e-05}
{"train_loss": 0.10631254315376282, "global_step": 146030, "epoch": 1640, "lr": 4.418336894634465e-05}
{"train_loss": 0.11246057599782944, "global_step": 146031, "epoch": 1640, "lr": 4.4182793146363696e-05}
{"train_loss": 0.06914202123880386, "global_step": 146032, "epoch": 1640, "lr": 4.418221734716482e-05}
{"train_loss": 0.13570575416088104, "global_step": 146033, "epoch": 1640, "lr": 4.4181641548748055e-05}
{"train_loss": 0.10912542790174484, "global_step": 146034, "epoch": 1640, "lr": 4.418106575111351e-05}
{"train_loss": 0.09514663368463516, "global_step": 146035, "epoch": 1640, "lr": 4.4180489954261245e-05}
{"train_loss": 0.0846053808927536, "global_step": 146036, "epoch": 1640, "lr": 4.4179914158191336e-05}
{"train_loss": 0.1532956063747406, "global_step": 146037, "epoch": 1640, "lr": 4.4179338362903876e-05}
{"train_loss": 0.16329240798950195, "global_step": 146038, "epoch": 1640, "lr": 4.4178762568398926e-05}
{"train_loss": 0.0896514430642128, "global_step": 146039, "epoch": 1640, "lr": 4.417818677467657e-05}
{"train_loss": 0.11632943153381348, "global_step": 146040, "epoch": 1640, "lr": 4.41776109817369e-05}
{"train_loss": 0.13152466714382172, "global_step": 146041, "epoch": 1640, "lr": 4.4177035189579956e-05}
{"train_loss": 0.104475237429142, "global_step": 146042, "epoch": 1640, "lr": 4.417645939820586e-05}
{"train_loss": 0.1068078875541687, "global_step": 146043, "epoch": 1640, "lr": 4.417588360761465e-05}
{"train_loss": 0.09254566580057144, "global_step": 146044, "epoch": 1640, "lr": 4.417530781780642e-05}
{"train_loss": 0.07199858129024506, "global_step": 146045, "epoch": 1640, "lr": 4.417473202878127e-05}
{"train_loss": 0.1413566619157791, "global_step": 146046, "epoch": 1640, "lr": 4.417415624053923e-05}
{"train_loss": 0.07171820104122162, "global_step": 146047, "epoch": 1640, "lr": 4.417358045308043e-05}
{"train_loss": 0.1107375757412964, "global_step": 146048, "epoch": 1640, "lr": 4.41730046664049e-05, "val_loss": 6.340421199798584, "train_action_mse_error": 17.874778747558594}
{"train_loss": 0.12231896072626114, "global_step": 146049, "epoch": 1641, "lr": 4.417242888051276e-05}
{"train_loss": 0.11773932725191116, "global_step": 146050, "epoch": 1641, "lr": 4.417185309540404e-05}
{"train_loss": 0.13344617187976837, "global_step": 146051, "epoch": 1641, "lr": 4.4171277311078866e-05}
{"train_loss": 0.10836847126483917, "global_step": 146052, "epoch": 1641, "lr": 4.417070152753728e-05}
{"train_loss": 0.09301171451807022, "global_step": 146053, "epoch": 1641, "lr": 4.417012574477938e-05}
{"train_loss": 0.17150454223155975, "global_step": 146054, "epoch": 1641, "lr": 4.416954996280523e-05}
{"train_loss": 0.09589125216007233, "global_step": 146055, "epoch": 1641, "lr": 4.416897418161492e-05}
{"train_loss": 0.09074930101633072, "global_step": 146056, "epoch": 1641, "lr": 4.4168398401208516e-05}
{"train_loss": 0.08148117363452911, "global_step": 146057, "epoch": 1641, "lr": 4.41678226215861e-05}
{"train_loss": 0.10915042459964752, "global_step": 146058, "epoch": 1641, "lr": 4.4167246842747745e-05}
{"train_loss": 0.10666456818580627, "global_step": 146059, "epoch": 1641, "lr": 4.416667106469354e-05}
{"train_loss": 0.09736304730176926, "global_step": 146060, "epoch": 1641, "lr": 4.4166095287423544e-05}
{"train_loss": 0.05809735879302025, "global_step": 146061, "epoch": 1641, "lr": 4.4165519510937864e-05}
{"train_loss": 0.10315194725990295, "global_step": 146062, "epoch": 1641, "lr": 4.416494373523654e-05}
{"train_loss": 0.13999629020690918, "global_step": 146063, "epoch": 1641, "lr": 4.416436796031966e-05}
{"train_loss": 0.08105962723493576, "global_step": 146064, "epoch": 1641, "lr": 4.416379218618734e-05}
{"train_loss": 0.1269984245300293, "global_step": 146065, "epoch": 1641, "lr": 4.416321641283959e-05}
{"train_loss": 0.10284227132797241, "global_step": 146066, "epoch": 1641, "lr": 4.416264064027657e-05}
{"train_loss": 0.0867476537823677, "global_step": 146067, "epoch": 1641, "lr": 4.416206486849826e-05}
{"train_loss": 0.14182734489440918, "global_step": 146068, "epoch": 1641, "lr": 4.416148909750484e-05}
{"train_loss": 0.10951033234596252, "global_step": 146069, "epoch": 1641, "lr": 4.41609133272963e-05}
{"train_loss": 0.06463101506233215, "global_step": 146070, "epoch": 1641, "lr": 4.416033755787278e-05}
{"train_loss": 0.09483350813388824, "global_step": 146071, "epoch": 1641, "lr": 4.41597617892343e-05}
{"train_loss": 0.09790245443582535, "global_step": 146072, "epoch": 1641, "lr": 4.4159186021381006e-05}
{"train_loss": 0.12015558034181595, "global_step": 146073, "epoch": 1641, "lr": 4.41586102543129e-05}
{"train_loss": 0.14825379848480225, "global_step": 146074, "epoch": 1641, "lr": 4.4158034488030134e-05}
{"train_loss": 0.12091625481843948, "global_step": 146075, "epoch": 1641, "lr": 4.4157458722532735e-05}
{"train_loss": 0.06779156625270844, "global_step": 146076, "epoch": 1641, "lr": 4.4156882957820786e-05}
{"train_loss": 0.06968126446008682, "global_step": 146077, "epoch": 1641, "lr": 4.415630719389439e-05}
{"train_loss": 0.08213535696268082, "global_step": 146078, "epoch": 1641, "lr": 4.415573143075359e-05}
{"train_loss": 0.11296680569648743, "global_step": 146079, "epoch": 1641, "lr": 4.415515566839849e-05}
{"train_loss": 0.07255378365516663, "global_step": 146080, "epoch": 1641, "lr": 4.415457990682914e-05}
{"train_loss": 0.09367872774600983, "global_step": 146081, "epoch": 1641, "lr": 4.415400414604566e-05}
{"train_loss": 0.1166246235370636, "global_step": 146082, "epoch": 1641, "lr": 4.4153428386048086e-05}
{"train_loss": 0.05840035900473595, "global_step": 146083, "epoch": 1641, "lr": 4.4152852626836525e-05}
{"train_loss": 0.08530616015195847, "global_step": 146084, "epoch": 1641, "lr": 4.415227686841103e-05}
{"train_loss": 0.07373171299695969, "global_step": 146085, "epoch": 1641, "lr": 4.415170111077171e-05}
{"train_loss": 0.10036306828260422, "global_step": 146086, "epoch": 1641, "lr": 4.41511253539186e-05}
{"train_loss": 0.05513424426317215, "global_step": 146087, "epoch": 1641, "lr": 4.4150549597851817e-05}
{"train_loss": 0.11217378824949265, "global_step": 146088, "epoch": 1641, "lr": 4.41499738425714e-05}
{"train_loss": 0.10755913704633713, "global_step": 146089, "epoch": 1641, "lr": 4.4149398088077477e-05}
{"train_loss": 0.15600095689296722, "global_step": 146090, "epoch": 1641, "lr": 4.4148822334370064e-05}
{"train_loss": 0.08688794076442719, "global_step": 146091, "epoch": 1641, "lr": 4.41482465814493e-05}
{"train_loss": 0.0988142341375351, "global_step": 146092, "epoch": 1641, "lr": 4.414767082931521e-05}
{"train_loss": 0.17643986642360687, "global_step": 146093, "epoch": 1641, "lr": 4.4147095077967915e-05}
{"train_loss": 0.1531788557767868, "global_step": 146094, "epoch": 1641, "lr": 4.414651932740746e-05}
{"train_loss": 0.08981896936893463, "global_step": 146095, "epoch": 1641, "lr": 4.414594357763393e-05}
{"train_loss": 0.14005382359027863, "global_step": 146096, "epoch": 1641, "lr": 4.4145367828647415e-05}
{"train_loss": 0.10124196112155914, "global_step": 146097, "epoch": 1641, "lr": 4.414479208044798e-05}
{"train_loss": 0.09702199697494507, "global_step": 146098, "epoch": 1641, "lr": 4.41442163330357e-05}
{"train_loss": 0.12203474342823029, "global_step": 146099, "epoch": 1641, "lr": 4.414364058641066e-05}
{"train_loss": 0.06653094291687012, "global_step": 146100, "epoch": 1641, "lr": 4.414306484057295e-05}
{"train_loss": 0.07278496026992798, "global_step": 146101, "epoch": 1641, "lr": 4.414248909552262e-05}
{"train_loss": 0.12863311171531677, "global_step": 146102, "epoch": 1641, "lr": 4.414191335125977e-05}
{"train_loss": 0.10963545739650726, "global_step": 146103, "epoch": 1641, "lr": 4.414133760778446e-05}
{"train_loss": 0.12693339586257935, "global_step": 146104, "epoch": 1641, "lr": 4.414076186509679e-05}
{"train_loss": 0.08601517975330353, "global_step": 146105, "epoch": 1641, "lr": 4.41401861231968e-05}
{"train_loss": 0.11363973468542099, "global_step": 146106, "epoch": 1641, "lr": 4.413961038208461e-05}
{"train_loss": 0.08768205344676971, "global_step": 146107, "epoch": 1641, "lr": 4.4139034641760264e-05}
{"train_loss": 0.11030622571706772, "global_step": 146108, "epoch": 1641, "lr": 4.4138458902223875e-05}
{"train_loss": 0.06807120889425278, "global_step": 146109, "epoch": 1641, "lr": 4.413788316347547e-05}
{"train_loss": 0.1434832066297531, "global_step": 146110, "epoch": 1641, "lr": 4.413730742551518e-05}
{"train_loss": 0.10144813358783722, "global_step": 146111, "epoch": 1641, "lr": 4.4136731688343044e-05}
{"train_loss": 0.102715402841568, "global_step": 146112, "epoch": 1641, "lr": 4.413615595195916e-05}
{"train_loss": 0.0434858500957489, "global_step": 146113, "epoch": 1641, "lr": 4.41355802163636e-05}
{"train_loss": 0.08005309104919434, "global_step": 146114, "epoch": 1641, "lr": 4.413500448155643e-05}
{"train_loss": 0.11780434846878052, "global_step": 146115, "epoch": 1641, "lr": 4.4134428747537745e-05}
{"train_loss": 0.19907620549201965, "global_step": 146116, "epoch": 1641, "lr": 4.4133853014307605e-05}
{"train_loss": 0.05483749881386757, "global_step": 146117, "epoch": 1641, "lr": 4.413327728186611e-05}
{"train_loss": 0.08339697122573853, "global_step": 146118, "epoch": 1641, "lr": 4.4132701550213305e-05}
{"train_loss": 0.09429877996444702, "global_step": 146119, "epoch": 1641, "lr": 4.4132125819349314e-05}
{"train_loss": 0.12213844060897827, "global_step": 146120, "epoch": 1641, "lr": 4.413155008927416e-05}
{"train_loss": 0.0939992368221283, "global_step": 146121, "epoch": 1641, "lr": 4.413097435998797e-05}
{"train_loss": 0.07337291538715363, "global_step": 146122, "epoch": 1641, "lr": 4.413039863149078e-05}
{"train_loss": 0.05850279703736305, "global_step": 146123, "epoch": 1641, "lr": 4.4129822903782704e-05}
{"train_loss": 0.05754540115594864, "global_step": 146124, "epoch": 1641, "lr": 4.412924717686378e-05}
{"train_loss": 0.09279915690422058, "global_step": 146125, "epoch": 1641, "lr": 4.412867145073413e-05}
{"train_loss": 0.11031437665224075, "global_step": 146126, "epoch": 1641, "lr": 4.412809572539379e-05}
{"train_loss": 0.11521652340888977, "global_step": 146127, "epoch": 1641, "lr": 4.4127520000842874e-05}
{"train_loss": 0.1388363093137741, "global_step": 146128, "epoch": 1641, "lr": 4.4126944277081424e-05}
{"train_loss": 0.18890069425106049, "global_step": 146129, "epoch": 1641, "lr": 4.412636855410955e-05}
{"train_loss": 0.14502230286598206, "global_step": 146130, "epoch": 1641, "lr": 4.4125792831927295e-05}
{"train_loss": 0.07790371030569077, "global_step": 146131, "epoch": 1641, "lr": 4.4125217110534786e-05}
{"train_loss": 0.11022508889436722, "global_step": 146132, "epoch": 1641, "lr": 4.412464138993205e-05}
{"train_loss": 0.11027491837739944, "global_step": 146133, "epoch": 1641, "lr": 4.4124065670119184e-05}
{"train_loss": 0.1266634315252304, "global_step": 146134, "epoch": 1641, "lr": 4.412348995109627e-05}
{"train_loss": 0.05647451430559158, "global_step": 146135, "epoch": 1641, "lr": 4.4122914232863366e-05}
{"train_loss": 0.09143926948308945, "global_step": 146136, "epoch": 1641, "lr": 4.412233851542059e-05}
{"train_loss": 0.10331712244601732, "global_step": 146137, "epoch": 1641, "lr": 4.412176279876797e-05, "val_loss": 6.237856864929199}
{"train_loss": 0.07852409034967422, "global_step": 146138, "epoch": 1642, "lr": 4.412118708290563e-05}
{"train_loss": 0.10318908840417862, "global_step": 146139, "epoch": 1642, "lr": 4.41206113678336e-05}
{"train_loss": 0.105515256524086, "global_step": 146140, "epoch": 1642, "lr": 4.412003565355201e-05}
{"train_loss": 0.15154719352722168, "global_step": 146141, "epoch": 1642, "lr": 4.4119459940060886e-05}
{"train_loss": 0.15595680475234985, "global_step": 146142, "epoch": 1642, "lr": 4.4118884227360345e-05}
{"train_loss": 0.13716092705726624, "global_step": 146143, "epoch": 1642, "lr": 4.411830851545044e-05}
{"train_loss": 0.1092020571231842, "global_step": 146144, "epoch": 1642, "lr": 4.4117732804331266e-05}
{"train_loss": 0.11413832753896713, "global_step": 146145, "epoch": 1642, "lr": 4.411715709400287e-05}
{"train_loss": 0.12308885902166367, "global_step": 146146, "epoch": 1642, "lr": 4.411658138446537e-05}
{"train_loss": 0.058527447283267975, "global_step": 146147, "epoch": 1642, "lr": 4.411600567571881e-05}
{"train_loss": 0.1272987723350525, "global_step": 146148, "epoch": 1642, "lr": 4.41154299677633e-05}
{"train_loss": 0.10720464587211609, "global_step": 146149, "epoch": 1642, "lr": 4.4114854260598876e-05}
{"train_loss": 0.07845909893512726, "global_step": 146150, "epoch": 1642, "lr": 4.4114278554225664e-05}
{"train_loss": 0.1268310397863388, "global_step": 146151, "epoch": 1642, "lr": 4.41137028486437e-05}
{"train_loss": 0.11222339421510696, "global_step": 146152, "epoch": 1642, "lr": 4.411312714385308e-05}
{"train_loss": 0.10891753435134888, "global_step": 146153, "epoch": 1642, "lr": 4.411255143985388e-05}
{"train_loss": 0.10019781440496445, "global_step": 146154, "epoch": 1642, "lr": 4.4111975736646155e-05}
{"train_loss": 0.1320246160030365, "global_step": 146155, "epoch": 1642, "lr": 4.411140003423003e-05}
{"train_loss": 0.06335224956274033, "global_step": 146156, "epoch": 1642, "lr": 4.4110824332605536e-05}
{"train_loss": 0.1071082055568695, "global_step": 146157, "epoch": 1642, "lr": 4.411024863177279e-05}
{"train_loss": 0.1417652815580368, "global_step": 146158, "epoch": 1642, "lr": 4.410967293173183e-05}
{"train_loss": 0.1609569936990738, "global_step": 146159, "epoch": 1642, "lr": 4.4109097232482774e-05}
{"train_loss": 0.10170658677816391, "global_step": 146160, "epoch": 1642, "lr": 4.4108521534025654e-05}
{"train_loss": 0.09236007183790207, "global_step": 146161, "epoch": 1642, "lr": 4.41079458363606e-05}
{"train_loss": 0.08966588228940964, "global_step": 146162, "epoch": 1642, "lr": 4.410737013948763e-05}
{"train_loss": 0.11772595345973969, "global_step": 146163, "epoch": 1642, "lr": 4.4106794443406874e-05}
{"train_loss": 0.12425780296325684, "global_step": 146164, "epoch": 1642, "lr": 4.4106218748118375e-05}
{"train_loss": 0.06349122524261475, "global_step": 146165, "epoch": 1642, "lr": 4.410564305362224e-05}
{"train_loss": 0.06587646901607513, "global_step": 146166, "epoch": 1642, "lr": 4.41050673599185e-05}
{"train_loss": 0.06086619198322296, "global_step": 146167, "epoch": 1642, "lr": 4.410449166700729e-05}
{"train_loss": 0.10126969963312149, "global_step": 146168, "epoch": 1642, "lr": 4.4103915974888645e-05}
{"train_loss": 0.07204819470643997, "global_step": 146169, "epoch": 1642, "lr": 4.410334028356267e-05}
{"train_loss": 0.1371595710515976, "global_step": 146170, "epoch": 1642, "lr": 4.410276459302942e-05}
{"train_loss": 0.09225758165121078, "global_step": 146171, "epoch": 1642, "lr": 4.4102188903288966e-05}
{"train_loss": 0.06971687078475952, "global_step": 146172, "epoch": 1642, "lr": 4.410161321434142e-05}
{"train_loss": 0.13220442831516266, "global_step": 146173, "epoch": 1642, "lr": 4.410103752618683e-05}
{"train_loss": 0.07145442813634872, "global_step": 146174, "epoch": 1642, "lr": 4.41004618388253e-05}
{"train_loss": 0.12265955656766891, "global_step": 146175, "epoch": 1642, "lr": 4.409988615225687e-05}
{"train_loss": 0.06724105775356293, "global_step": 146176, "epoch": 1642, "lr": 4.4099310466481665e-05}
{"train_loss": 0.09237860143184662, "global_step": 146177, "epoch": 1642, "lr": 4.409873478149971e-05}
{"train_loss": 0.07610413432121277, "global_step": 146178, "epoch": 1642, "lr": 4.409815909731113e-05}
{"train_loss": 0.07454869896173477, "global_step": 146179, "epoch": 1642, "lr": 4.409758341391595e-05}
{"train_loss": 0.06414814293384552, "global_step": 146180, "epoch": 1642, "lr": 4.409700773131431e-05}
{"train_loss": 0.08763065189123154, "global_step": 146181, "epoch": 1642, "lr": 4.409643204950623e-05}
{"train_loss": 0.12690608203411102, "global_step": 146182, "epoch": 1642, "lr": 4.4095856368491836e-05}
{"train_loss": 0.11179161071777344, "global_step": 146183, "epoch": 1642, "lr": 4.409528068827116e-05}
{"train_loss": 0.08896025270223618, "global_step": 146184, "epoch": 1642, "lr": 4.409470500884432e-05}
{"train_loss": 0.11403827369213104, "global_step": 146185, "epoch": 1642, "lr": 4.409412933021136e-05}
{"train_loss": 0.16259348392486572, "global_step": 146186, "epoch": 1642, "lr": 4.409355365237239e-05}
{"train_loss": 0.14695186913013458, "global_step": 146187, "epoch": 1642, "lr": 4.409297797532745e-05}
{"train_loss": 0.11615709960460663, "global_step": 146188, "epoch": 1642, "lr": 4.409240229907665e-05}
{"train_loss": 0.138675257563591, "global_step": 146189, "epoch": 1642, "lr": 4.409182662362005e-05}
{"train_loss": 0.08067889511585236, "global_step": 146190, "epoch": 1642, "lr": 4.409125094895773e-05}
{"train_loss": 0.11855286359786987, "global_step": 146191, "epoch": 1642, "lr": 4.4090675275089785e-05}
{"train_loss": 0.11344390362501144, "global_step": 146192, "epoch": 1642, "lr": 4.409009960201625e-05}
{"train_loss": 0.18970182538032532, "global_step": 146193, "epoch": 1642, "lr": 4.408952392973725e-05}
{"train_loss": 0.14205579459667206, "global_step": 146194, "epoch": 1642, "lr": 4.4088948258252826e-05}
{"train_loss": 0.09125358611345291, "global_step": 146195, "epoch": 1642, "lr": 4.4088372587563086e-05}
{"train_loss": 0.11070766299962997, "global_step": 146196, "epoch": 1642, "lr": 4.408779691766808e-05}
{"train_loss": 0.11452039331197739, "global_step": 146197, "epoch": 1642, "lr": 4.4087221248567915e-05}
{"train_loss": 0.1555817723274231, "global_step": 146198, "epoch": 1642, "lr": 4.4086645580262626e-05}
{"train_loss": 0.13474318385124207, "global_step": 146199, "epoch": 1642, "lr": 4.4086069912752336e-05}
{"train_loss": 0.0963202491402626, "global_step": 146200, "epoch": 1642, "lr": 4.4085494246037084e-05}
{"train_loss": 0.12057847529649734, "global_step": 146201, "epoch": 1642, "lr": 4.408491858011699e-05}
{"train_loss": 0.0491197295486927, "global_step": 146202, "epoch": 1642, "lr": 4.408434291499208e-05}
{"train_loss": 0.08385397493839264, "global_step": 146203, "epoch": 1642, "lr": 4.4083767250662476e-05}
{"train_loss": 0.14782333374023438, "global_step": 146204, "epoch": 1642, "lr": 4.4083191587128226e-05}
{"train_loss": 0.2020539790391922, "global_step": 146205, "epoch": 1642, "lr": 4.4082615924389435e-05}
{"train_loss": 0.10345140099525452, "global_step": 146206, "epoch": 1642, "lr": 4.408204026244614e-05}
{"train_loss": 0.10215412080287933, "global_step": 146207, "epoch": 1642, "lr": 4.408146460129845e-05}
{"train_loss": 0.2014859914779663, "global_step": 146208, "epoch": 1642, "lr": 4.4080888940946445e-05}
{"train_loss": 0.08057542890310287, "global_step": 146209, "epoch": 1642, "lr": 4.408031328139019e-05}
{"train_loss": 0.15041665732860565, "global_step": 146210, "epoch": 1642, "lr": 4.4079737622629774e-05}
{"train_loss": 0.08462820947170258, "global_step": 146211, "epoch": 1642, "lr": 4.407916196466525e-05}
{"train_loss": 0.11202320456504822, "global_step": 146212, "epoch": 1642, "lr": 4.407858630749672e-05}
{"train_loss": 0.18439243733882904, "global_step": 146213, "epoch": 1642, "lr": 4.4078010651124236e-05}
{"train_loss": 0.10990601032972336, "global_step": 146214, "epoch": 1642, "lr": 4.407743499554791e-05}
{"train_loss": 0.13228684663772583, "global_step": 146215, "epoch": 1642, "lr": 4.4076859340767785e-05}
{"train_loss": 0.08813304454088211, "global_step": 146216, "epoch": 1642, "lr": 4.4076283686783974e-05}
{"train_loss": 0.03557601198554039, "global_step": 146217, "epoch": 1642, "lr": 4.407570803359651e-05}
{"train_loss": 0.09345968812704086, "global_step": 146218, "epoch": 1642, "lr": 4.407513238120551e-05}
{"train_loss": 0.10725407302379608, "global_step": 146219, "epoch": 1642, "lr": 4.4074556729611026e-05}
{"train_loss": 0.13744650781154633, "global_step": 146220, "epoch": 1642, "lr": 4.407398107881316e-05}
{"train_loss": 0.1755675971508026, "global_step": 146221, "epoch": 1642, "lr": 4.4073405428811965e-05}
{"train_loss": 0.05573715269565582, "global_step": 146222, "epoch": 1642, "lr": 4.407282977960754e-05}
{"train_loss": 0.10566207766532898, "global_step": 146223, "epoch": 1642, "lr": 4.407225413119993e-05}
{"train_loss": 0.07933822274208069, "global_step": 146224, "epoch": 1642, "lr": 4.4071678483589256e-05}
{"train_loss": 0.12441152334213257, "global_step": 146225, "epoch": 1642, "lr": 4.407110283677555e-05}
{"train_loss": 0.11037054816993434, "global_step": 146226, "epoch": 1642, "lr": 4.4070527190758926e-05, "val_loss": 6.054592132568359}
{"train_loss": 0.14188049733638763, "global_step": 146227, "epoch": 1643, "lr": 4.4069951545539446e-05}
{"train_loss": 0.10721639543771744, "global_step": 146228, "epoch": 1643, "lr": 4.406937590111719e-05}
{"train_loss": 0.1860072910785675, "global_step": 146229, "epoch": 1643, "lr": 4.4068800257492236e-05}
{"train_loss": 0.04146244376897812, "global_step": 146230, "epoch": 1643, "lr": 4.406822461466465e-05}
{"train_loss": 0.11368738114833832, "global_step": 146231, "epoch": 1643, "lr": 4.406764897263453e-05}
{"train_loss": 0.11655743420124054, "global_step": 146232, "epoch": 1643, "lr": 4.406707333140193e-05}
{"train_loss": 0.12482061237096786, "global_step": 146233, "epoch": 1643, "lr": 4.406649769096696e-05}
{"train_loss": 0.1391432136297226, "global_step": 146234, "epoch": 1643, "lr": 4.406592205132965e-05}
{"train_loss": 0.07598527520895004, "global_step": 146235, "epoch": 1643, "lr": 4.406534641249013e-05}
{"train_loss": 0.16798259317874908, "global_step": 146236, "epoch": 1643, "lr": 4.406477077444843e-05}
{"train_loss": 0.16156774759292603, "global_step": 146237, "epoch": 1643, "lr": 4.4064195137204664e-05}
{"train_loss": 0.17209139466285706, "global_step": 146238, "epoch": 1643, "lr": 4.406361950075888e-05}
{"train_loss": 0.06367815285921097, "global_step": 146239, "epoch": 1643, "lr": 4.406304386511119e-05}
{"train_loss": 0.07264850288629532, "global_step": 146240, "epoch": 1643, "lr": 4.406246823026163e-05}
{"train_loss": 0.12256351113319397, "global_step": 146241, "epoch": 1643, "lr": 4.406189259621032e-05}
{"train_loss": 0.1260601282119751, "global_step": 146242, "epoch": 1643, "lr": 4.4061316962957295e-05}
{"train_loss": 0.10965844988822937, "global_step": 146243, "epoch": 1643, "lr": 4.4060741330502676e-05}
{"train_loss": 0.08574677258729935, "global_step": 146244, "epoch": 1643, "lr": 4.40601656988465e-05}
{"train_loss": 0.061110809445381165, "global_step": 146245, "epoch": 1643, "lr": 4.405959006798887e-05}
{"train_loss": 0.12726305425167084, "global_step": 146246, "epoch": 1643, "lr": 4.405901443792985e-05}
{"train_loss": 0.09685768187046051, "global_step": 146247, "epoch": 1643, "lr": 4.4058438808669526e-05}
{"train_loss": 0.12950466573238373, "global_step": 146248, "epoch": 1643, "lr": 4.405786318020798e-05}
{"train_loss": 0.11332277953624725, "global_step": 146249, "epoch": 1643, "lr": 4.405728755254527e-05}
{"train_loss": 0.14658239483833313, "global_step": 146250, "epoch": 1643, "lr": 4.4056711925681506e-05}
{"train_loss": 0.09520772099494934, "global_step": 146251, "epoch": 1643, "lr": 4.4056136299616714e-05}
{"train_loss": 0.08989302814006805, "global_step": 146252, "epoch": 1643, "lr": 4.4055560674351026e-05}
{"train_loss": 0.10568610578775406, "global_step": 146253, "epoch": 1643, "lr": 4.405498504988448e-05}
{"train_loss": 0.11794479191303253, "global_step": 146254, "epoch": 1643, "lr": 4.405440942621719e-05}
{"train_loss": 0.08100450783967972, "global_step": 146255, "epoch": 1643, "lr": 4.405383380334918e-05}
{"train_loss": 0.1104663833975792, "global_step": 146256, "epoch": 1643, "lr": 4.4053258181280594e-05}
{"train_loss": 0.07591371238231659, "global_step": 146257, "epoch": 1643, "lr": 4.4052682560011445e-05}
{"train_loss": 0.11678316444158554, "global_step": 146258, "epoch": 1643, "lr": 4.405210693954187e-05}
{"train_loss": 0.12317033857107162, "global_step": 146259, "epoch": 1643, "lr": 4.4051531319871884e-05}
{"train_loss": 0.0651867613196373, "global_step": 146260, "epoch": 1643, "lr": 4.405095570100163e-05}
{"train_loss": 0.08978679031133652, "global_step": 146261, "epoch": 1643, "lr": 4.405038008293112e-05}
{"train_loss": 0.11217333376407623, "global_step": 146262, "epoch": 1643, "lr": 4.404980446566048e-05}
{"train_loss": 0.11002220213413239, "global_step": 146263, "epoch": 1643, "lr": 4.404922884918977e-05}
{"train_loss": 0.13215841352939606, "global_step": 146264, "epoch": 1643, "lr": 4.404865323351907e-05}
{"train_loss": 0.04768208786845207, "global_step": 146265, "epoch": 1643, "lr": 4.4048077618648456e-05}
{"train_loss": 0.08342421799898148, "global_step": 146266, "epoch": 1643, "lr": 4.404750200457801e-05}
{"train_loss": 0.13048309087753296, "global_step": 146267, "epoch": 1643, "lr": 4.4046926391307806e-05}
{"train_loss": 0.07975219190120697, "global_step": 146268, "epoch": 1643, "lr": 4.404635077883791e-05}
{"train_loss": 0.0658225268125534, "global_step": 146269, "epoch": 1643, "lr": 4.4045775167168426e-05}
{"train_loss": 0.16305509209632874, "global_step": 146270, "epoch": 1643, "lr": 4.4045199556299385e-05}
{"train_loss": 0.16787508130073547, "global_step": 146271, "epoch": 1643, "lr": 4.4044623946230925e-05}
{"train_loss": 0.12201974540948868, "global_step": 146272, "epoch": 1643, "lr": 4.4044048336963074e-05}
{"train_loss": 0.11586196720600128, "global_step": 146273, "epoch": 1643, "lr": 4.404347272849594e-05}
{"train_loss": 0.09642142057418823, "global_step": 146274, "epoch": 1643, "lr": 4.404289712082957e-05}
{"train_loss": 0.11289945989847183, "global_step": 146275, "epoch": 1643, "lr": 4.404232151396409e-05}
{"train_loss": 0.08769392222166061, "global_step": 146276, "epoch": 1643, "lr": 4.404174590789952e-05}
{"train_loss": 0.1123364195227623, "global_step": 146277, "epoch": 1643, "lr": 4.404117030263598e-05}
{"train_loss": 0.09246701747179031, "global_step": 146278, "epoch": 1643, "lr": 4.4040594698173524e-05}
{"train_loss": 0.08241061121225357, "global_step": 146279, "epoch": 1643, "lr": 4.4040019094512245e-05}
{"train_loss": 0.0810045674443245, "global_step": 146280, "epoch": 1643, "lr": 4.40394434916522e-05}
{"train_loss": 0.08698085695505142, "global_step": 146281, "epoch": 1643, "lr": 4.4038867889593486e-05}
{"train_loss": 0.09017344564199448, "global_step": 146282, "epoch": 1643, "lr": 4.4038292288336174e-05}
{"train_loss": 0.10841197520494461, "global_step": 146283, "epoch": 1643, "lr": 4.403771668788034e-05}
{"train_loss": 0.1398506611585617, "global_step": 146284, "epoch": 1643, "lr": 4.403714108822606e-05}
{"train_loss": 0.07496094703674316, "global_step": 146285, "epoch": 1643, "lr": 4.403656548937342e-05}
{"train_loss": 0.1179308295249939, "global_step": 146286, "epoch": 1643, "lr": 4.40359898913225e-05}
{"train_loss": 0.1470356434583664, "global_step": 146287, "epoch": 1643, "lr": 4.4035414294073346e-05}
{"train_loss": 0.11624941229820251, "global_step": 146288, "epoch": 1643, "lr": 4.403483869762608e-05}
{"train_loss": 0.10161547362804413, "global_step": 146289, "epoch": 1643, "lr": 4.403426310198073e-05}
{"train_loss": 0.10757346451282501, "global_step": 146290, "epoch": 1643, "lr": 4.403368750713743e-05}
{"train_loss": 0.0843871682882309, "global_step": 146291, "epoch": 1643, "lr": 4.403311191309621e-05}
{"train_loss": 0.0810852199792862, "global_step": 146292, "epoch": 1643, "lr": 4.403253631985718e-05}
{"train_loss": 0.12705546617507935, "global_step": 146293, "epoch": 1643, "lr": 4.403196072742038e-05}
{"train_loss": 0.11969101428985596, "global_step": 146294, "epoch": 1643, "lr": 4.403138513578593e-05}
{"train_loss": 0.06526617705821991, "global_step": 146295, "epoch": 1643, "lr": 4.403080954495387e-05}
{"train_loss": 0.11153122037649155, "global_step": 146296, "epoch": 1643, "lr": 4.403023395492432e-05}
{"train_loss": 0.12267805635929108, "global_step": 146297, "epoch": 1643, "lr": 4.402965836569731e-05}
{"train_loss": 0.12004152685403824, "global_step": 146298, "epoch": 1643, "lr": 4.402908277727293e-05}
{"train_loss": 0.08897923678159714, "global_step": 146299, "epoch": 1643, "lr": 4.40285071896513e-05}
{"train_loss": 0.1617615520954132, "global_step": 146300, "epoch": 1643, "lr": 4.4027931602832444e-05}
{"train_loss": 0.12530335783958435, "global_step": 146301, "epoch": 1643, "lr": 4.402735601681647e-05}
{"train_loss": 0.08043645322322845, "global_step": 146302, "epoch": 1643, "lr": 4.402678043160344e-05}
{"train_loss": 0.08833654224872589, "global_step": 146303, "epoch": 1643, "lr": 4.4026204847193436e-05}
{"train_loss": 0.09806187450885773, "global_step": 146304, "epoch": 1643, "lr": 4.402562926358653e-05}
{"train_loss": 0.11214198917150497, "global_step": 146305, "epoch": 1643, "lr": 4.402505368078282e-05}
{"train_loss": 0.13194189965724945, "global_step": 146306, "epoch": 1643, "lr": 4.4024478098782356e-05}
{"train_loss": 0.2160416543483734, "global_step": 146307, "epoch": 1643, "lr": 4.4023902517585245e-05}
{"train_loss": 0.15158218145370483, "global_step": 146308, "epoch": 1643, "lr": 4.402332693719152e-05}
{"train_loss": 0.09734024852514267, "global_step": 146309, "epoch": 1643, "lr": 4.402275135760132e-05}
{"train_loss": 0.07646963745355606, "global_step": 146310, "epoch": 1643, "lr": 4.4022175778814654e-05}
{"train_loss": 0.08173360675573349, "global_step": 146311, "epoch": 1643, "lr": 4.402160020083166e-05}
{"train_loss": 0.10379914939403534, "global_step": 146312, "epoch": 1643, "lr": 4.4021024623652365e-05}
{"train_loss": 0.09247469156980515, "global_step": 146313, "epoch": 1643, "lr": 4.40204490472769e-05}
{"train_loss": 0.042846839874982834, "global_step": 146314, "epoch": 1643, "lr": 4.4019873471705284e-05}
{"train_loss": 0.10811746304624537, "global_step": 146315, "epoch": 1643, "lr": 4.401929789693764e-05, "val_loss": 6.1262311935424805}
{"train_loss": 0.14485594630241394, "global_step": 146316, "epoch": 1644, "lr": 4.4018722322974015e-05}
{"train_loss": 0.10674626380205154, "global_step": 146317, "epoch": 1644, "lr": 4.4018146749814494e-05}
{"train_loss": 0.09155916422605515, "global_step": 146318, "epoch": 1644, "lr": 4.401757117745919e-05}
{"train_loss": 0.04166635125875473, "global_step": 146319, "epoch": 1644, "lr": 4.401699560590812e-05}
{"train_loss": 0.1784428358078003, "global_step": 146320, "epoch": 1644, "lr": 4.4016420035161424e-05}
{"train_loss": 0.0723106861114502, "global_step": 146321, "epoch": 1644, "lr": 4.4015844465219124e-05}
{"train_loss": 0.0638367235660553, "global_step": 146322, "epoch": 1644, "lr": 4.401526889608133e-05}
{"train_loss": 0.10455282032489777, "global_step": 146323, "epoch": 1644, "lr": 4.401469332774811e-05}
{"train_loss": 0.13206768035888672, "global_step": 146324, "epoch": 1644, "lr": 4.401411776021953e-05}
{"train_loss": 0.0582939088344574, "global_step": 146325, "epoch": 1644, "lr": 4.4013542193495694e-05}
{"train_loss": 0.10922906547784805, "global_step": 146326, "epoch": 1644, "lr": 4.4012966627576666e-05}
{"train_loss": 0.1867130696773529, "global_step": 146327, "epoch": 1644, "lr": 4.401239106246251e-05}
{"train_loss": 0.11254426091909409, "global_step": 146328, "epoch": 1644, "lr": 4.401181549815332e-05}
{"train_loss": 0.09291881322860718, "global_step": 146329, "epoch": 1644, "lr": 4.4011239934649164e-05}
{"train_loss": 0.11276543140411377, "global_step": 146330, "epoch": 1644, "lr": 4.4010664371950135e-05}
{"train_loss": 0.11425404995679855, "global_step": 146331, "epoch": 1644, "lr": 4.401008881005628e-05}
{"train_loss": 0.09258576482534409, "global_step": 146332, "epoch": 1644, "lr": 4.400951324896772e-05}
{"train_loss": 0.07117760926485062, "global_step": 146333, "epoch": 1644, "lr": 4.400893768868448e-05}
{"train_loss": 0.07767200469970703, "global_step": 146334, "epoch": 1644, "lr": 4.400836212920667e-05}
{"train_loss": 0.053380049765110016, "global_step": 146335, "epoch": 1644, "lr": 4.400778657053439e-05}
{"train_loss": 0.14707084000110626, "global_step": 146336, "epoch": 1644, "lr": 4.400721101266766e-05}
{"train_loss": 0.034329600632190704, "global_step": 146337, "epoch": 1644, "lr": 4.400663545560661e-05}
{"train_loss": 0.1607711762189865, "global_step": 146338, "epoch": 1644, "lr": 4.400605989935127e-05}
{"train_loss": 0.11263015121221542, "global_step": 146339, "epoch": 1644, "lr": 4.4005484343901774e-05}
{"train_loss": 0.11440886557102203, "global_step": 146340, "epoch": 1644, "lr": 4.400490878925815e-05}
{"train_loss": 0.12260406464338303, "global_step": 146341, "epoch": 1644, "lr": 4.4004333235420494e-05}
{"train_loss": 0.11280376464128494, "global_step": 146342, "epoch": 1644, "lr": 4.400375768238888e-05}
{"train_loss": 0.11497730016708374, "global_step": 146343, "epoch": 1644, "lr": 4.400318213016339e-05}
{"train_loss": 0.12890267372131348, "global_step": 146344, "epoch": 1644, "lr": 4.40026065787441e-05}
{"train_loss": 0.08407416939735413, "global_step": 146345, "epoch": 1644, "lr": 4.40020310281311e-05}
{"train_loss": 0.13103042542934418, "global_step": 146346, "epoch": 1644, "lr": 4.4001455478324424e-05}
{"train_loss": 0.02254834771156311, "global_step": 146347, "epoch": 1644, "lr": 4.40008799293242e-05}
{"train_loss": 0.08834244310855865, "global_step": 146348, "epoch": 1644, "lr": 4.400030438113047e-05}
{"train_loss": 0.11230532824993134, "global_step": 146349, "epoch": 1644, "lr": 4.399972883374335e-05}
{"train_loss": 0.17315231263637543, "global_step": 146350, "epoch": 1644, "lr": 4.399915328716287e-05}
{"train_loss": 0.11371325701475143, "global_step": 146351, "epoch": 1644, "lr": 4.399857774138915e-05}
{"train_loss": 0.1311301440000534, "global_step": 146352, "epoch": 1644, "lr": 4.3998002196422224e-05}
{"train_loss": 0.08488701283931732, "global_step": 146353, "epoch": 1644, "lr": 4.399742665226219e-05}
{"train_loss": 0.14913663268089294, "global_step": 146354, "epoch": 1644, "lr": 4.399685110890916e-05}
{"train_loss": 0.10486534982919693, "global_step": 146355, "epoch": 1644, "lr": 4.3996275566363155e-05}
{"train_loss": 0.05887213349342346, "global_step": 146356, "epoch": 1644, "lr": 4.39957000246243e-05}
{"train_loss": 0.08273321390151978, "global_step": 146357, "epoch": 1644, "lr": 4.399512448369262e-05}
{"train_loss": 0.06424576044082642, "global_step": 146358, "epoch": 1644, "lr": 4.3994548943568254e-05}
{"train_loss": 0.10105840861797333, "global_step": 146359, "epoch": 1644, "lr": 4.399397340425123e-05}
{"train_loss": 0.07602931559085846, "global_step": 146360, "epoch": 1644, "lr": 4.399339786574165e-05}
{"train_loss": 0.07613498717546463, "global_step": 146361, "epoch": 1644, "lr": 4.399282232803959e-05}
{"train_loss": 0.08905348181724548, "global_step": 146362, "epoch": 1644, "lr": 4.399224679114511e-05}
{"train_loss": 0.07015050202608109, "global_step": 146363, "epoch": 1644, "lr": 4.39916712550583e-05}
{"train_loss": 0.07534264028072357, "global_step": 146364, "epoch": 1644, "lr": 4.399109571977925e-05}
{"train_loss": 0.17943838238716125, "global_step": 146365, "epoch": 1644, "lr": 4.3990520185308006e-05}
{"train_loss": 0.08753801137208939, "global_step": 146366, "epoch": 1644, "lr": 4.398994465164468e-05}
{"train_loss": 0.1622290164232254, "global_step": 146367, "epoch": 1644, "lr": 4.398936911878931e-05}
{"train_loss": 0.10028575360774994, "global_step": 146368, "epoch": 1644, "lr": 4.3988793586742015e-05}
{"train_loss": 0.07142432034015656, "global_step": 146369, "epoch": 1644, "lr": 4.398821805550284e-05}
{"train_loss": 0.1687958538532257, "global_step": 146370, "epoch": 1644, "lr": 4.3987642525071896e-05}
{"train_loss": 0.0868012011051178, "global_step": 146371, "epoch": 1644, "lr": 4.398706699544922e-05}
{"train_loss": 0.07101114839315414, "global_step": 146372, "epoch": 1644, "lr": 4.39864914666349e-05}
{"train_loss": 0.1409105509519577, "global_step": 146373, "epoch": 1644, "lr": 4.398591593862905e-05}
{"train_loss": 0.06576959043741226, "global_step": 146374, "epoch": 1644, "lr": 4.39853404114317e-05}
{"train_loss": 0.13180887699127197, "global_step": 146375, "epoch": 1644, "lr": 4.3984764885042976e-05}
{"train_loss": 0.1073453426361084, "global_step": 146376, "epoch": 1644, "lr": 4.398418935946289e-05}
{"train_loss": 0.1345883458852768, "global_step": 146377, "epoch": 1644, "lr": 4.398361383469159e-05}
{"train_loss": 0.08381184935569763, "global_step": 146378, "epoch": 1644, "lr": 4.39830383107291e-05}
{"train_loss": 0.13381333649158478, "global_step": 146379, "epoch": 1644, "lr": 4.398246278757553e-05}
{"train_loss": 0.0789218321442604, "global_step": 146380, "epoch": 1644, "lr": 4.3981887265230936e-05}
{"train_loss": 0.18187540769577026, "global_step": 146381, "epoch": 1644, "lr": 4.3981311743695404e-05}
{"train_loss": 0.10959091782569885, "global_step": 146382, "epoch": 1644, "lr": 4.398073622296901e-05}
{"train_loss": 0.09075459837913513, "global_step": 146383, "epoch": 1644, "lr": 4.398016070305185e-05}
{"train_loss": 0.07050982117652893, "global_step": 146384, "epoch": 1644, "lr": 4.397958518394395e-05}
{"train_loss": 0.12103419750928879, "global_step": 146385, "epoch": 1644, "lr": 4.3979009665645454e-05}
{"train_loss": 0.11591228097677231, "global_step": 146386, "epoch": 1644, "lr": 4.397843414815638e-05}
{"train_loss": 0.07522767037153244, "global_step": 146387, "epoch": 1644, "lr": 4.397785863147686e-05}
{"train_loss": 0.15815898776054382, "global_step": 146388, "epoch": 1644, "lr": 4.3977283115606916e-05}
{"train_loss": 0.07929804176092148, "global_step": 146389, "epoch": 1644, "lr": 4.397670760054666e-05}
{"train_loss": 0.0736919417977333, "global_step": 146390, "epoch": 1644, "lr": 4.397613208629617e-05}
{"train_loss": 0.04672042652964592, "global_step": 146391, "epoch": 1644, "lr": 4.397555657285551e-05}
{"train_loss": 0.10276547819375992, "global_step": 146392, "epoch": 1644, "lr": 4.3974981060224774e-05}
{"train_loss": 0.10796155780553818, "global_step": 146393, "epoch": 1644, "lr": 4.3974405548404005e-05}
{"train_loss": 0.09753767400979996, "global_step": 146394, "epoch": 1644, "lr": 4.397383003739334e-05}
{"train_loss": 0.10776381194591522, "global_step": 146395, "epoch": 1644, "lr": 4.397325452719279e-05}
{"train_loss": 0.11468513309955597, "global_step": 146396, "epoch": 1644, "lr": 4.3972679017802474e-05}
{"train_loss": 0.07885947823524475, "global_step": 146397, "epoch": 1644, "lr": 4.397210350922246e-05}
{"train_loss": 0.10930511355400085, "global_step": 146398, "epoch": 1644, "lr": 4.397152800145282e-05}
{"train_loss": 0.07116147875785828, "global_step": 146399, "epoch": 1644, "lr": 4.397095249449363e-05}
{"train_loss": 0.06261365115642548, "global_step": 146400, "epoch": 1644, "lr": 4.3970376988344974e-05}
{"train_loss": 0.18775251507759094, "global_step": 146401, "epoch": 1644, "lr": 4.3969801483006935e-05}
{"train_loss": 0.10099459439516068, "global_step": 146402, "epoch": 1644, "lr": 4.3969225978479587e-05}
{"train_loss": 0.09104164689779282, "global_step": 146403, "epoch": 1644, "lr": 4.396865047476298e-05}
{"train_loss": 0.1034635914426841, "global_step": 146404, "epoch": 1644, "lr": 4.396807497185724e-05, "val_loss": 6.315601348876953}
{"train_loss": 0.07799863070249557, "global_step": 146405, "epoch": 1645, "lr": 4.396749946976239e-05}
{"train_loss": 0.1244577169418335, "global_step": 146406, "epoch": 1645, "lr": 4.396692396847856e-05}
{"train_loss": 0.1509876698255539, "global_step": 146407, "epoch": 1645, "lr": 4.396634846800579e-05}
{"train_loss": 0.06820498406887054, "global_step": 146408, "epoch": 1645, "lr": 4.396577296834417e-05}
{"train_loss": 0.1346466839313507, "global_step": 146409, "epoch": 1645, "lr": 4.396519746949379e-05}
{"train_loss": 0.11772776395082474, "global_step": 146410, "epoch": 1645, "lr": 4.39646219714547e-05}
{"train_loss": 0.05265537649393082, "global_step": 146411, "epoch": 1645, "lr": 4.3964046474227015e-05}
{"train_loss": 0.09979598969221115, "global_step": 146412, "epoch": 1645, "lr": 4.3963470977810767e-05}
{"train_loss": 0.11784317344427109, "global_step": 146413, "epoch": 1645, "lr": 4.396289548220608e-05}
{"train_loss": 0.08014313876628876, "global_step": 146414, "epoch": 1645, "lr": 4.396231998741299e-05}
{"train_loss": 0.06785781681537628, "global_step": 146415, "epoch": 1645, "lr": 4.3961744493431606e-05}
{"train_loss": 0.09054867923259735, "global_step": 146416, "epoch": 1645, "lr": 4.396116900026199e-05}
{"train_loss": 0.1612556427717209, "global_step": 146417, "epoch": 1645, "lr": 4.396059350790421e-05}
{"train_loss": 0.13045057654380798, "global_step": 146418, "epoch": 1645, "lr": 4.396001801635836e-05}
{"train_loss": 0.13163389265537262, "global_step": 146419, "epoch": 1645, "lr": 4.395944252562452e-05}
{"train_loss": 0.08168170601129532, "global_step": 146420, "epoch": 1645, "lr": 4.395886703570275e-05}
{"train_loss": 0.07349599152803421, "global_step": 146421, "epoch": 1645, "lr": 4.395829154659314e-05}
{"train_loss": 0.1576937884092331, "global_step": 146422, "epoch": 1645, "lr": 4.395771605829576e-05}
{"train_loss": 0.08739668875932693, "global_step": 146423, "epoch": 1645, "lr": 4.395714057081071e-05}
{"train_loss": 0.09595315158367157, "global_step": 146424, "epoch": 1645, "lr": 4.395656508413802e-05}
{"train_loss": 0.1493801474571228, "global_step": 146425, "epoch": 1645, "lr": 4.3955989598277796e-05}
{"train_loss": 0.11286377161741257, "global_step": 146426, "epoch": 1645, "lr": 4.395541411323014e-05}
{"train_loss": 0.09304608404636383, "global_step": 146427, "epoch": 1645, "lr": 4.3954838628995085e-05}
{"train_loss": 0.11860617995262146, "global_step": 146428, "epoch": 1645, "lr": 4.395426314557274e-05}
{"train_loss": 0.09828732907772064, "global_step": 146429, "epoch": 1645, "lr": 4.3953687662963164e-05}
{"train_loss": 0.12183418869972229, "global_step": 146430, "epoch": 1645, "lr": 4.395311218116645e-05}
{"train_loss": 0.16348060965538025, "global_step": 146431, "epoch": 1645, "lr": 4.395253670018266e-05}
{"train_loss": 0.08349897712469101, "global_step": 146432, "epoch": 1645, "lr": 4.3951961220011884e-05}
{"train_loss": 0.0691571980714798, "global_step": 146433, "epoch": 1645, "lr": 4.395138574065418e-05}
{"train_loss": 0.15612375736236572, "global_step": 146434, "epoch": 1645, "lr": 4.3950810262109656e-05}
{"train_loss": 0.08446548879146576, "global_step": 146435, "epoch": 1645, "lr": 4.3950234784378345e-05}
{"train_loss": 0.08968111127614975, "global_step": 146436, "epoch": 1645, "lr": 4.394965930746038e-05}
{"train_loss": 0.08738133311271667, "global_step": 146437, "epoch": 1645, "lr": 4.39490838313558e-05}
{"train_loss": 0.14406701922416687, "global_step": 146438, "epoch": 1645, "lr": 4.394850835606469e-05}
{"train_loss": 0.16842475533485413, "global_step": 146439, "epoch": 1645, "lr": 4.394793288158713e-05}
{"train_loss": 0.09439253807067871, "global_step": 146440, "epoch": 1645, "lr": 4.3947357407923195e-05}
{"train_loss": 0.058199796825647354, "global_step": 146441, "epoch": 1645, "lr": 4.394678193507297e-05}
{"train_loss": 0.15062615275382996, "global_step": 146442, "epoch": 1645, "lr": 4.394620646303653e-05}
{"train_loss": 0.09032780677080154, "global_step": 146443, "epoch": 1645, "lr": 4.394563099181392e-05}
{"train_loss": 0.13748399913311005, "global_step": 146444, "epoch": 1645, "lr": 4.3945055521405256e-05}
{"train_loss": 0.07269348204135895, "global_step": 146445, "epoch": 1645, "lr": 4.394448005181063e-05}
{"train_loss": 0.08984845876693726, "global_step": 146446, "epoch": 1645, "lr": 4.394390458303007e-05}
{"train_loss": 0.13048288226127625, "global_step": 146447, "epoch": 1645, "lr": 4.39433291150637e-05}
{"train_loss": 0.11438152194023132, "global_step": 146448, "epoch": 1645, "lr": 4.3942753647911555e-05}
{"train_loss": 0.1367754489183426, "global_step": 146449, "epoch": 1645, "lr": 4.394217818157375e-05}
{"train_loss": 0.14288492500782013, "global_step": 146450, "epoch": 1645, "lr": 4.3941602716050325e-05}
{"train_loss": 0.15455369651317596, "global_step": 146451, "epoch": 1645, "lr": 4.3941027251341407e-05}
{"train_loss": 0.18030928075313568, "global_step": 146452, "epoch": 1645, "lr": 4.3940451787447015e-05}
{"train_loss": 0.10414951294660568, "global_step": 146453, "epoch": 1645, "lr": 4.393987632436728e-05}
{"train_loss": 0.12857551872730255, "global_step": 146454, "epoch": 1645, "lr": 4.393930086210223e-05}
{"train_loss": 0.1542198359966278, "global_step": 146455, "epoch": 1645, "lr": 4.393872540065199e-05}
{"train_loss": 0.1636587232351303, "global_step": 146456, "epoch": 1645, "lr": 4.3938149940016605e-05}
{"train_loss": 0.04568066820502281, "global_step": 146457, "epoch": 1645, "lr": 4.393757448019616e-05}
{"train_loss": 0.11869131028652191, "global_step": 146458, "epoch": 1645, "lr": 4.393699902119073e-05}
{"train_loss": 0.177860826253891, "global_step": 146459, "epoch": 1645, "lr": 4.393642356300041e-05}
{"train_loss": 0.13313576579093933, "global_step": 146460, "epoch": 1645, "lr": 4.393584810562526e-05}
{"train_loss": 0.11792730540037155, "global_step": 146461, "epoch": 1645, "lr": 4.393527264906534e-05}
{"train_loss": 0.1420011818408966, "global_step": 146462, "epoch": 1645, "lr": 4.393469719332077e-05}
{"train_loss": 0.12438268959522247, "global_step": 146463, "epoch": 1645, "lr": 4.3934121738391594e-05}
{"train_loss": 0.07805785536766052, "global_step": 146464, "epoch": 1645, "lr": 4.393354628427792e-05}
{"train_loss": 0.06374701112508774, "global_step": 146465, "epoch": 1645, "lr": 4.393297083097979e-05}
{"train_loss": 0.13336274027824402, "global_step": 146466, "epoch": 1645, "lr": 4.393239537849731e-05}
{"train_loss": 0.19530604779720306, "global_step": 146467, "epoch": 1645, "lr": 4.393181992683053e-05}
{"train_loss": 0.06914461404085159, "global_step": 146468, "epoch": 1645, "lr": 4.393124447597956e-05}
{"train_loss": 0.14250262081623077, "global_step": 146469, "epoch": 1645, "lr": 4.393066902594444e-05}
{"train_loss": 0.07983772456645966, "global_step": 146470, "epoch": 1645, "lr": 4.3930093576725295e-05}
{"train_loss": 0.15363584458827972, "global_step": 146471, "epoch": 1645, "lr": 4.3929518128322154e-05}
{"train_loss": 0.12208468466997147, "global_step": 146472, "epoch": 1645, "lr": 4.3928942680735125e-05}
{"train_loss": 0.129904106259346, "global_step": 146473, "epoch": 1645, "lr": 4.3928367233964263e-05}
{"train_loss": 0.14330002665519714, "global_step": 146474, "epoch": 1645, "lr": 4.392779178800968e-05}
{"train_loss": 0.07477333396673203, "global_step": 146475, "epoch": 1645, "lr": 4.3927216342871414e-05}
{"train_loss": 0.14076358079910278, "global_step": 146476, "epoch": 1645, "lr": 4.392664089854956e-05}
{"train_loss": 0.1427999883890152, "global_step": 146477, "epoch": 1645, "lr": 4.392606545504421e-05}
{"train_loss": 0.11922641098499298, "global_step": 146478, "epoch": 1645, "lr": 4.392549001235541e-05}
{"train_loss": 0.1417580544948578, "global_step": 146479, "epoch": 1645, "lr": 4.392491457048326e-05}
{"train_loss": 0.14653941988945007, "global_step": 146480, "epoch": 1645, "lr": 4.392433912942782e-05}
{"train_loss": 0.1743784248828888, "global_step": 146481, "epoch": 1645, "lr": 4.3923763689189205e-05}
{"train_loss": 0.12978726625442505, "global_step": 146482, "epoch": 1645, "lr": 4.392318824976743e-05}
{"train_loss": 0.11889271438121796, "global_step": 146483, "epoch": 1645, "lr": 4.3922612811162645e-05}
{"train_loss": 0.08289111405611038, "global_step": 146484, "epoch": 1645, "lr": 4.3922037373374866e-05}
{"train_loss": 0.0986512079834938, "global_step": 146485, "epoch": 1645, "lr": 4.392146193640421e-05}
{"train_loss": 0.09990733861923218, "global_step": 146486, "epoch": 1645, "lr": 4.3920886500250716e-05}
{"train_loss": 0.08153983950614929, "global_step": 146487, "epoch": 1645, "lr": 4.3920311064914515e-05}
{"train_loss": 0.17717482149600983, "global_step": 146488, "epoch": 1645, "lr": 4.391973563039563e-05}
{"train_loss": 0.12226703017950058, "global_step": 146489, "epoch": 1645, "lr": 4.3919160196694185e-05}
{"train_loss": 0.10557664185762405, "global_step": 146490, "epoch": 1645, "lr": 4.391858476381021e-05}
{"train_loss": 0.11132726818323135, "global_step": 146491, "epoch": 1645, "lr": 4.391800933174383e-05}
{"train_loss": 0.12471699714660645, "global_step": 146492, "epoch": 1645, "lr": 4.391743390049508e-05}
{"train_loss": 0.11698317963085818, "global_step": 146493, "epoch": 1645, "lr": 4.3916858470064076e-05, "val_loss": 5.9833598136901855, "train_action_mse_error": 12.610434532165527}
{"train_loss": 0.18524014949798584, "global_step": 146494, "epoch": 1646, "lr": 4.391628304045087e-05}
{"train_loss": 0.1323549598455429, "global_step": 146495, "epoch": 1646, "lr": 4.391570761165554e-05}
{"train_loss": 0.1359286904335022, "global_step": 146496, "epoch": 1646, "lr": 4.391513218367817e-05}
{"train_loss": 0.09821540117263794, "global_step": 146497, "epoch": 1646, "lr": 4.391455675651884e-05}
{"train_loss": 0.10113482922315598, "global_step": 146498, "epoch": 1646, "lr": 4.391398133017762e-05}
{"train_loss": 0.1466275155544281, "global_step": 146499, "epoch": 1646, "lr": 4.391340590465458e-05}
{"train_loss": 0.12384804338216782, "global_step": 146500, "epoch": 1646, "lr": 4.3912830479949826e-05}
{"train_loss": 0.09672945737838745, "global_step": 146501, "epoch": 1646, "lr": 4.39122550560634e-05}
{"train_loss": 0.10146886110305786, "global_step": 146502, "epoch": 1646, "lr": 4.3911679632995424e-05}
{"train_loss": 0.14067326486110687, "global_step": 146503, "epoch": 1646, "lr": 4.391110421074592e-05}
{"train_loss": 0.10510548949241638, "global_step": 146504, "epoch": 1646, "lr": 4.391052878931502e-05}
{"train_loss": 0.0808490514755249, "global_step": 146505, "epoch": 1646, "lr": 4.390995336870275e-05}
{"train_loss": 0.10249537974596024, "global_step": 146506, "epoch": 1646, "lr": 4.390937794890924e-05}
{"train_loss": 0.1269335299730301, "global_step": 146507, "epoch": 1646, "lr": 4.390880252993451e-05}
{"train_loss": 0.11504670232534409, "global_step": 146508, "epoch": 1646, "lr": 4.390822711177869e-05}
{"train_loss": 0.12732985615730286, "global_step": 146509, "epoch": 1646, "lr": 4.390765169444182e-05}
{"train_loss": 0.10543369501829147, "global_step": 146510, "epoch": 1646, "lr": 4.390707627792401e-05}
{"train_loss": 0.14632025361061096, "global_step": 146511, "epoch": 1646, "lr": 4.39065008622253e-05}
{"train_loss": 0.1409493386745453, "global_step": 146512, "epoch": 1646, "lr": 4.3905925447345805e-05}
{"train_loss": 0.09815753251314163, "global_step": 146513, "epoch": 1646, "lr": 4.3905350033285574e-05}
{"train_loss": 0.12370375543832779, "global_step": 146514, "epoch": 1646, "lr": 4.3904774620044695e-05}
{"train_loss": 0.11930644512176514, "global_step": 146515, "epoch": 1646, "lr": 4.390419920762324e-05}
{"train_loss": 0.1015750840306282, "global_step": 146516, "epoch": 1646, "lr": 4.390362379602129e-05}
{"train_loss": 0.11026348173618317, "global_step": 146517, "epoch": 1646, "lr": 4.390304838523894e-05}
{"train_loss": 0.10712671279907227, "global_step": 146518, "epoch": 1646, "lr": 4.390247297527623e-05}
{"train_loss": 0.1182628720998764, "global_step": 146519, "epoch": 1646, "lr": 4.3901897566133273e-05}
{"train_loss": 0.1275496780872345, "global_step": 146520, "epoch": 1646, "lr": 4.390132215781012e-05}
{"train_loss": 0.11741109192371368, "global_step": 146521, "epoch": 1646, "lr": 4.390074675030687e-05}
{"train_loss": 0.08796907961368561, "global_step": 146522, "epoch": 1646, "lr": 4.390017134362358e-05}
{"train_loss": 0.1298474669456482, "global_step": 146523, "epoch": 1646, "lr": 4.389959593776035e-05}
{"train_loss": 0.10246666520833969, "global_step": 146524, "epoch": 1646, "lr": 4.389902053271723e-05}
{"train_loss": 0.062142565846443176, "global_step": 146525, "epoch": 1646, "lr": 4.389844512849433e-05}
{"train_loss": 0.07658375799655914, "global_step": 146526, "epoch": 1646, "lr": 4.389786972509169e-05}
{"train_loss": 0.1239505261182785, "global_step": 146527, "epoch": 1646, "lr": 4.389729432250943e-05}
{"train_loss": 0.12312018126249313, "global_step": 146528, "epoch": 1646, "lr": 4.389671892074758e-05}
{"train_loss": 0.12562215328216553, "global_step": 146529, "epoch": 1646, "lr": 4.389614351980626e-05}
{"train_loss": 0.05997619777917862, "global_step": 146530, "epoch": 1646, "lr": 4.389556811968552e-05}
{"train_loss": 0.11753658205270767, "global_step": 146531, "epoch": 1646, "lr": 4.3894992720385466e-05}
{"train_loss": 0.14140623807907104, "global_step": 146532, "epoch": 1646, "lr": 4.389441732190612e-05}
{"train_loss": 0.11417295038700104, "global_step": 146533, "epoch": 1646, "lr": 4.389384192424763e-05}
{"train_loss": 0.12203215062618256, "global_step": 146534, "epoch": 1646, "lr": 4.389326652741002e-05}
{"train_loss": 0.1418685019016266, "global_step": 146535, "epoch": 1646, "lr": 4.389269113139338e-05}
{"train_loss": 0.1126209944486618, "global_step": 146536, "epoch": 1646, "lr": 4.389211573619781e-05}
{"train_loss": 0.07823151350021362, "global_step": 146537, "epoch": 1646, "lr": 4.389154034182336e-05}
{"train_loss": 0.0833083987236023, "global_step": 146538, "epoch": 1646, "lr": 4.389096494827013e-05}
{"train_loss": 0.07896102964878082, "global_step": 146539, "epoch": 1646, "lr": 4.3890389555538165e-05}
{"train_loss": 0.11369980871677399, "global_step": 146540, "epoch": 1646, "lr": 4.388981416362759e-05}
{"train_loss": 0.1122482493519783, "global_step": 146541, "epoch": 1646, "lr": 4.388923877253842e-05}
{"train_loss": 0.13100916147232056, "global_step": 146542, "epoch": 1646, "lr": 4.38886633822708e-05}
{"train_loss": 0.1000637412071228, "global_step": 146543, "epoch": 1646, "lr": 4.388808799282475e-05}
{"train_loss": 0.1375567615032196, "global_step": 146544, "epoch": 1646, "lr": 4.3887512604200394e-05}
{"train_loss": 0.06520716100931168, "global_step": 146545, "epoch": 1646, "lr": 4.3886937216397764e-05}
{"train_loss": 0.12185592949390411, "global_step": 146546, "epoch": 1646, "lr": 4.388636182941698e-05}
{"train_loss": 0.04914874583482742, "global_step": 146547, "epoch": 1646, "lr": 4.388578644325809e-05}
{"train_loss": 0.10433866083621979, "global_step": 146548, "epoch": 1646, "lr": 4.388521105792119e-05}
{"train_loss": 0.14617274701595306, "global_step": 146549, "epoch": 1646, "lr": 4.388463567340634e-05}
{"train_loss": 0.1471390724182129, "global_step": 146550, "epoch": 1646, "lr": 4.388406028971363e-05}
{"train_loss": 0.09658943116664886, "global_step": 146551, "epoch": 1646, "lr": 4.388348490684313e-05}
{"train_loss": 0.23820549249649048, "global_step": 146552, "epoch": 1646, "lr": 4.388290952479493e-05}
{"train_loss": 0.07472289353609085, "global_step": 146553, "epoch": 1646, "lr": 4.38823341435691e-05}
{"train_loss": 0.13606534898281097, "global_step": 146554, "epoch": 1646, "lr": 4.388175876316569e-05}
{"train_loss": 0.10527338832616806, "global_step": 146555, "epoch": 1646, "lr": 4.3881183383584836e-05}
{"train_loss": 0.1358698606491089, "global_step": 146556, "epoch": 1646, "lr": 4.388060800482656e-05}
{"train_loss": 0.1182144433259964, "global_step": 146557, "epoch": 1646, "lr": 4.3880032626890976e-05}
{"train_loss": 0.23040363192558289, "global_step": 146558, "epoch": 1646, "lr": 4.3879457249778125e-05}
{"train_loss": 0.1774868220090866, "global_step": 146559, "epoch": 1646, "lr": 4.3878881873488137e-05}
{"train_loss": 0.04816245660185814, "global_step": 146560, "epoch": 1646, "lr": 4.387830649802103e-05}
{"train_loss": 0.10153688490390778, "global_step": 146561, "epoch": 1646, "lr": 4.3877731123376935e-05}
{"train_loss": 0.07354316115379333, "global_step": 146562, "epoch": 1646, "lr": 4.3877155749555885e-05}
{"train_loss": 0.05426959693431854, "global_step": 146563, "epoch": 1646, "lr": 4.3876580376557995e-05}
{"train_loss": 0.11111614853143692, "global_step": 146564, "epoch": 1646, "lr": 4.3876005004383306e-05}
{"train_loss": 0.15573740005493164, "global_step": 146565, "epoch": 1646, "lr": 4.3875429633031926e-05}
{"train_loss": 0.07891584187746048, "global_step": 146566, "epoch": 1646, "lr": 4.387485426250391e-05}
{"train_loss": 0.0925927460193634, "global_step": 146567, "epoch": 1646, "lr": 4.387427889279936e-05}
{"train_loss": 0.06604574620723724, "global_step": 146568, "epoch": 1646, "lr": 4.387370352391831e-05}
{"train_loss": 0.08088313788175583, "global_step": 146569, "epoch": 1646, "lr": 4.38731281558609e-05}
{"train_loss": 0.06896746158599854, "global_step": 146570, "epoch": 1646, "lr": 4.3872552788627145e-05}
{"train_loss": 0.11015896499156952, "global_step": 146571, "epoch": 1646, "lr": 4.387197742221717e-05}
{"train_loss": 0.11648914963006973, "global_step": 146572, "epoch": 1646, "lr": 4.387140205663103e-05}
{"train_loss": 0.11508956551551819, "global_step": 146573, "epoch": 1646, "lr": 4.3870826691868795e-05}
{"train_loss": 0.12379548698663712, "global_step": 146574, "epoch": 1646, "lr": 4.387025132793057e-05}
{"train_loss": 0.1708882749080658, "global_step": 146575, "epoch": 1646, "lr": 4.386967596481639e-05}
{"train_loss": 0.1204848363995552, "global_step": 146576, "epoch": 1646, "lr": 4.386910060252639e-05}
{"train_loss": 0.09731094539165497, "global_step": 146577, "epoch": 1646, "lr": 4.386852524106058e-05}
{"train_loss": 0.12742947041988373, "global_step": 146578, "epoch": 1646, "lr": 4.3867949880419105e-05}
{"train_loss": 0.06830897927284241, "global_step": 146579, "epoch": 1646, "lr": 4.3867374520601976e-05}
{"train_loss": 0.06909914314746857, "global_step": 146580, "epoch": 1646, "lr": 4.3866799161609335e-05}
{"train_loss": 0.06683224439620972, "global_step": 146581, "epoch": 1646, "lr": 4.38662238034412e-05}
{"train_loss": 0.11339753284380677, "global_step": 146582, "epoch": 1646, "lr": 4.38656484460977e-05, "val_loss": 6.1462202072143555}
{"train_loss": 0.1540151983499527, "global_step": 146583, "epoch": 1647, "lr": 4.386507308957887e-05}
{"train_loss": 0.09911616891622543, "global_step": 146584, "epoch": 1647, "lr": 4.386449773388483e-05}
{"train_loss": 0.11701784282922745, "global_step": 146585, "epoch": 1647, "lr": 4.3863922379015604e-05}
{"train_loss": 0.07719516009092331, "global_step": 146586, "epoch": 1647, "lr": 4.386334702497132e-05}
{"train_loss": 0.1284000277519226, "global_step": 146587, "epoch": 1647, "lr": 4.3862771671752025e-05}
{"train_loss": 0.12094337493181229, "global_step": 146588, "epoch": 1647, "lr": 4.3862196319357805e-05}
{"train_loss": 0.05801871791481972, "global_step": 146589, "epoch": 1647, "lr": 4.3861620967788744e-05}
{"train_loss": 0.14178955554962158, "global_step": 146590, "epoch": 1647, "lr": 4.386104561704491e-05}
{"train_loss": 0.11844887584447861, "global_step": 146591, "epoch": 1647, "lr": 4.386047026712639e-05}
{"train_loss": 0.1527968943119049, "global_step": 146592, "epoch": 1647, "lr": 4.385989491803324e-05}
{"train_loss": 0.09237128496170044, "global_step": 146593, "epoch": 1647, "lr": 4.385931956976557e-05}
{"train_loss": 0.0809035450220108, "global_step": 146594, "epoch": 1647, "lr": 4.385874422232342e-05}
{"train_loss": 0.13159862160682678, "global_step": 146595, "epoch": 1647, "lr": 4.3858168875706915e-05}
{"train_loss": 0.09493829309940338, "global_step": 146596, "epoch": 1647, "lr": 4.3857593529916076e-05}
{"train_loss": 0.07988487929105759, "global_step": 146597, "epoch": 1647, "lr": 4.3857018184951026e-05}
{"train_loss": 0.1668730229139328, "global_step": 146598, "epoch": 1647, "lr": 4.3856442840811806e-05}
{"train_loss": 0.06756817549467087, "global_step": 146599, "epoch": 1647, "lr": 4.385586749749854e-05}
{"train_loss": 0.16437309980392456, "global_step": 146600, "epoch": 1647, "lr": 4.385529215501125e-05}
{"train_loss": 0.15316776931285858, "global_step": 146601, "epoch": 1647, "lr": 4.385471681335007e-05}
{"train_loss": 0.10515594482421875, "global_step": 146602, "epoch": 1647, "lr": 4.385414147251502e-05}
{"train_loss": 0.10603079199790955, "global_step": 146603, "epoch": 1647, "lr": 4.385356613250623e-05}
{"train_loss": 0.10771394520998001, "global_step": 146604, "epoch": 1647, "lr": 4.385299079332373e-05}
{"train_loss": 0.13334426283836365, "global_step": 146605, "epoch": 1647, "lr": 4.385241545496764e-05}
{"train_loss": 0.1423841118812561, "global_step": 146606, "epoch": 1647, "lr": 4.3851840117437996e-05}
{"train_loss": 0.12031452357769012, "global_step": 146607, "epoch": 1647, "lr": 4.385126478073491e-05}
{"train_loss": 0.15873457491397858, "global_step": 146608, "epoch": 1647, "lr": 4.385068944485845e-05}
{"train_loss": 0.13656124472618103, "global_step": 146609, "epoch": 1647, "lr": 4.385011410980869e-05}
{"train_loss": 0.09264688938856125, "global_step": 146610, "epoch": 1647, "lr": 4.3849538775585716e-05}
{"train_loss": 0.14020182192325592, "global_step": 146611, "epoch": 1647, "lr": 4.384896344218957e-05}
{"train_loss": 0.06635671854019165, "global_step": 146612, "epoch": 1647, "lr": 4.384838810962039e-05}
{"train_loss": 0.12372196465730667, "global_step": 146613, "epoch": 1647, "lr": 4.3847812777878186e-05}
{"train_loss": 0.12554340064525604, "global_step": 146614, "epoch": 1647, "lr": 4.384723744696309e-05}
{"train_loss": 0.07834691554307938, "global_step": 146615, "epoch": 1647, "lr": 4.384666211687514e-05}
{"train_loss": 0.0767730250954628, "global_step": 146616, "epoch": 1647, "lr": 4.3846086787614455e-05}
{"train_loss": 0.09506962448358536, "global_step": 146617, "epoch": 1647, "lr": 4.384551145918106e-05}
{"train_loss": 0.17338985204696655, "global_step": 146618, "epoch": 1647, "lr": 4.3844936131575085e-05}
{"train_loss": 0.18796241283416748, "global_step": 146619, "epoch": 1647, "lr": 4.384436080479657e-05}
{"train_loss": 0.17388398945331573, "global_step": 146620, "epoch": 1647, "lr": 4.384378547884562e-05}
{"train_loss": 0.1562882959842682, "global_step": 146621, "epoch": 1647, "lr": 4.384321015372228e-05}
{"train_loss": 0.1466754525899887, "global_step": 146622, "epoch": 1647, "lr": 4.384263482942667e-05}
{"train_loss": 0.17160606384277344, "global_step": 146623, "epoch": 1647, "lr": 4.384205950595881e-05}
{"train_loss": 0.07489906251430511, "global_step": 146624, "epoch": 1647, "lr": 4.3841484183318834e-05}
{"train_loss": 0.13760483264923096, "global_step": 146625, "epoch": 1647, "lr": 4.384090886150678e-05}
{"train_loss": 0.08372223377227783, "global_step": 146626, "epoch": 1647, "lr": 4.384033354052275e-05}
{"train_loss": 0.11500591784715652, "global_step": 146627, "epoch": 1647, "lr": 4.3839758220366814e-05}
{"train_loss": 0.09280835837125778, "global_step": 146628, "epoch": 1647, "lr": 4.383918290103904e-05}
{"train_loss": 0.12030816078186035, "global_step": 146629, "epoch": 1647, "lr": 4.3838607582539527e-05}
{"train_loss": 0.12387952953577042, "global_step": 146630, "epoch": 1647, "lr": 4.383803226486832e-05}
{"train_loss": 0.17994703352451324, "global_step": 146631, "epoch": 1647, "lr": 4.383745694802553e-05}
{"train_loss": 0.08966436237096786, "global_step": 146632, "epoch": 1647, "lr": 4.38368816320112e-05}
{"train_loss": 0.14340846240520477, "global_step": 146633, "epoch": 1647, "lr": 4.3836306316825446e-05}
{"train_loss": 0.06867685168981552, "global_step": 146634, "epoch": 1647, "lr": 4.383573100246831e-05}
{"train_loss": 0.14808890223503113, "global_step": 146635, "epoch": 1647, "lr": 4.38351556889399e-05}
{"train_loss": 0.09235093742609024, "global_step": 146636, "epoch": 1647, "lr": 4.383458037624026e-05}
{"train_loss": 0.0928654596209526, "global_step": 146637, "epoch": 1647, "lr": 4.383400506436951e-05}
{"train_loss": 0.05694124102592468, "global_step": 146638, "epoch": 1647, "lr": 4.3833429753327677e-05}
{"train_loss": 0.10105390101671219, "global_step": 146639, "epoch": 1647, "lr": 4.3832854443114884e-05}
{"train_loss": 0.06836942583322525, "global_step": 146640, "epoch": 1647, "lr": 4.383227913373117e-05}
{"train_loss": 0.10110852122306824, "global_step": 146641, "epoch": 1647, "lr": 4.383170382517665e-05}
{"train_loss": 0.11234311759471893, "global_step": 146642, "epoch": 1647, "lr": 4.383112851745136e-05}
{"train_loss": 0.10143532603979111, "global_step": 146643, "epoch": 1647, "lr": 4.3830553210555414e-05}
{"train_loss": 0.14422491192817688, "global_step": 146644, "epoch": 1647, "lr": 4.3829977904488875e-05}
{"train_loss": 0.06844744086265564, "global_step": 146645, "epoch": 1647, "lr": 4.3829402599251825e-05}
{"train_loss": 0.11331818997859955, "global_step": 146646, "epoch": 1647, "lr": 4.3828827294844325e-05}
{"train_loss": 0.09168732911348343, "global_step": 146647, "epoch": 1647, "lr": 4.382825199126647e-05}
{"train_loss": 0.15407679975032806, "global_step": 146648, "epoch": 1647, "lr": 4.382767668851833e-05}
{"train_loss": 0.10600972175598145, "global_step": 146649, "epoch": 1647, "lr": 4.382710138659999e-05}
{"train_loss": 0.08585072308778763, "global_step": 146650, "epoch": 1647, "lr": 4.382652608551152e-05}
{"train_loss": 0.08257447183132172, "global_step": 146651, "epoch": 1647, "lr": 4.3825950785252987e-05}
{"train_loss": 0.07981475442647934, "global_step": 146652, "epoch": 1647, "lr": 4.3825375485824496e-05}
{"train_loss": 0.059258073568344116, "global_step": 146653, "epoch": 1647, "lr": 4.382480018722609e-05}
{"train_loss": 0.09134399145841599, "global_step": 146654, "epoch": 1647, "lr": 4.3824224889457885e-05}
{"train_loss": 0.1420336216688156, "global_step": 146655, "epoch": 1647, "lr": 4.382364959251991e-05}
{"train_loss": 0.10722551494836807, "global_step": 146656, "epoch": 1647, "lr": 4.3823074296412295e-05}
{"train_loss": 0.08216740936040878, "global_step": 146657, "epoch": 1647, "lr": 4.382249900113508e-05}
{"train_loss": 0.08133595436811447, "global_step": 146658, "epoch": 1647, "lr": 4.382192370668836e-05}
{"train_loss": 0.06197898089885712, "global_step": 146659, "epoch": 1647, "lr": 4.3821348413072196e-05}
{"train_loss": 0.09192655235528946, "global_step": 146660, "epoch": 1647, "lr": 4.382077312028669e-05}
{"train_loss": 0.10558164864778519, "global_step": 146661, "epoch": 1647, "lr": 4.38201978283319e-05}
{"train_loss": 0.11421100795269012, "global_step": 146662, "epoch": 1647, "lr": 4.3819622537207895e-05}
{"train_loss": 0.07381299138069153, "global_step": 146663, "epoch": 1647, "lr": 4.381904724691479e-05}
{"train_loss": 0.07885363698005676, "global_step": 146664, "epoch": 1647, "lr": 4.381847195745263e-05}
{"train_loss": 0.06113483011722565, "global_step": 146665, "epoch": 1647, "lr": 4.3817896668821506e-05}
{"train_loss": 0.11404746025800705, "global_step": 146666, "epoch": 1647, "lr": 4.3817321381021483e-05}
{"train_loss": 0.11821267753839493, "global_step": 146667, "epoch": 1647, "lr": 4.381674609405265e-05}
{"train_loss": 0.11413545906543732, "global_step": 146668, "epoch": 1647, "lr": 4.381617080791508e-05}
{"train_loss": 0.1258327066898346, "global_step": 146669, "epoch": 1647, "lr": 4.381559552260886e-05}
{"train_loss": 0.08776483684778214, "global_step": 146670, "epoch": 1647, "lr": 4.381502023813404e-05}
{"train_loss": 0.1108331959485338, "global_step": 146671, "epoch": 1647, "lr": 4.381444495449073e-05, "val_loss": 6.261208534240723}
{"train_loss": 0.11491301655769348, "global_step": 146672, "epoch": 1648, "lr": 4.381386967167897e-05}
{"train_loss": 0.10309211164712906, "global_step": 146673, "epoch": 1648, "lr": 4.3813294389698886e-05}
{"train_loss": 0.10938991606235504, "global_step": 146674, "epoch": 1648, "lr": 4.3812719108550505e-05}
{"train_loss": 0.10166987031698227, "global_step": 146675, "epoch": 1648, "lr": 4.381214382823396e-05}
{"train_loss": 0.10214291512966156, "global_step": 146676, "epoch": 1648, "lr": 4.381156854874926e-05}
{"train_loss": 0.1575770378112793, "global_step": 146677, "epoch": 1648, "lr": 4.381099327009655e-05}
{"train_loss": 0.11643350124359131, "global_step": 146678, "epoch": 1648, "lr": 4.381041799227585e-05}
{"train_loss": 0.06304444372653961, "global_step": 146679, "epoch": 1648, "lr": 4.380984271528727e-05}
{"train_loss": 0.136033296585083, "global_step": 146680, "epoch": 1648, "lr": 4.38092674391309e-05}
{"train_loss": 0.09296929091215134, "global_step": 146681, "epoch": 1648, "lr": 4.380869216380678e-05}
{"train_loss": 0.15118899941444397, "global_step": 146682, "epoch": 1648, "lr": 4.380811688931502e-05}
{"train_loss": 0.11470192670822144, "global_step": 146683, "epoch": 1648, "lr": 4.3807541615655675e-05}
{"train_loss": 0.08213263750076294, "global_step": 146684, "epoch": 1648, "lr": 4.380696634282883e-05}
{"train_loss": 0.169671893119812, "global_step": 146685, "epoch": 1648, "lr": 4.380639107083456e-05}
{"train_loss": 0.13309556245803833, "global_step": 146686, "epoch": 1648, "lr": 4.380581579967296e-05}
{"train_loss": 0.208939790725708, "global_step": 146687, "epoch": 1648, "lr": 4.380524052934408e-05}
{"train_loss": 0.11453300714492798, "global_step": 146688, "epoch": 1648, "lr": 4.380466525984802e-05}
{"train_loss": 0.15141725540161133, "global_step": 146689, "epoch": 1648, "lr": 4.380408999118483e-05}
{"train_loss": 0.059806205332279205, "global_step": 146690, "epoch": 1648, "lr": 4.3803514723354624e-05}
{"train_loss": 0.11012285947799683, "global_step": 146691, "epoch": 1648, "lr": 4.380293945635744e-05}
{"train_loss": 0.09519488364458084, "global_step": 146692, "epoch": 1648, "lr": 4.3802364190193395e-05}
{"train_loss": 0.0849473848938942, "global_step": 146693, "epoch": 1648, "lr": 4.380178892486253e-05}
{"train_loss": 0.12179964035749435, "global_step": 146694, "epoch": 1648, "lr": 4.380121366036496e-05}
{"train_loss": 0.05620364844799042, "global_step": 146695, "epoch": 1648, "lr": 4.380063839670071e-05}
{"train_loss": 0.08854427188634872, "global_step": 146696, "epoch": 1648, "lr": 4.380006313386992e-05}
{"train_loss": 0.1301146000623703, "global_step": 146697, "epoch": 1648, "lr": 4.3799487871872606e-05}
{"train_loss": 0.10666946321725845, "global_step": 146698, "epoch": 1648, "lr": 4.379891261070888e-05}
{"train_loss": 0.052047617733478546, "global_step": 146699, "epoch": 1648, "lr": 4.3798337350378836e-05}
{"train_loss": 0.06812864542007446, "global_step": 146700, "epoch": 1648, "lr": 4.379776209088251e-05}
{"train_loss": 0.06857006251811981, "global_step": 146701, "epoch": 1648, "lr": 4.3797186832220025e-05}
{"train_loss": 0.0844961628317833, "global_step": 146702, "epoch": 1648, "lr": 4.379661157439141e-05}
{"train_loss": 0.14082492887973785, "global_step": 146703, "epoch": 1648, "lr": 4.3796036317396775e-05}
{"train_loss": 0.11679020524024963, "global_step": 146704, "epoch": 1648, "lr": 4.3795461061236184e-05}
{"train_loss": 0.08128340542316437, "global_step": 146705, "epoch": 1648, "lr": 4.3794885805909717e-05}
{"train_loss": 0.1058724895119667, "global_step": 146706, "epoch": 1648, "lr": 4.379431055141745e-05}
{"train_loss": 0.13834412395954132, "global_step": 146707, "epoch": 1648, "lr": 4.3793735297759474e-05}
{"train_loss": 0.09488406777381897, "global_step": 146708, "epoch": 1648, "lr": 4.379316004493584e-05}
{"train_loss": 0.09562710672616959, "global_step": 146709, "epoch": 1648, "lr": 4.3792584792946656e-05}
{"train_loss": 0.07450801879167557, "global_step": 146710, "epoch": 1648, "lr": 4.379200954179196e-05}
{"train_loss": 0.14527106285095215, "global_step": 146711, "epoch": 1648, "lr": 4.379143429147188e-05}
{"train_loss": 0.1150929406285286, "global_step": 146712, "epoch": 1648, "lr": 4.3790859041986436e-05}
{"train_loss": 0.08344003558158875, "global_step": 146713, "epoch": 1648, "lr": 4.379028379333576e-05}
{"train_loss": 0.15686379373073578, "global_step": 146714, "epoch": 1648, "lr": 4.378970854551989e-05}
{"train_loss": 0.15964721143245697, "global_step": 146715, "epoch": 1648, "lr": 4.378913329853891e-05}
{"train_loss": 0.10719260573387146, "global_step": 146716, "epoch": 1648, "lr": 4.378855805239293e-05}
{"train_loss": 0.09513230621814728, "global_step": 146717, "epoch": 1648, "lr": 4.378798280708198e-05}
{"train_loss": 0.1816982924938202, "global_step": 146718, "epoch": 1648, "lr": 4.378740756260618e-05}
{"train_loss": 0.08074682205915451, "global_step": 146719, "epoch": 1648, "lr": 4.378683231896557e-05}
{"train_loss": 0.08033709228038788, "global_step": 146720, "epoch": 1648, "lr": 4.378625707616026e-05}
{"train_loss": 0.09698350727558136, "global_step": 146721, "epoch": 1648, "lr": 4.3785681834190306e-05}
{"train_loss": 0.0995965451002121, "global_step": 146722, "epoch": 1648, "lr": 4.378510659305579e-05}
{"train_loss": 0.16791117191314697, "global_step": 146723, "epoch": 1648, "lr": 4.37845313527568e-05}
{"train_loss": 0.07822670042514801, "global_step": 146724, "epoch": 1648, "lr": 4.378395611329339e-05}
{"train_loss": 0.13517896831035614, "global_step": 146725, "epoch": 1648, "lr": 4.378338087466566e-05}
{"train_loss": 0.08579900860786438, "global_step": 146726, "epoch": 1648, "lr": 4.378280563687368e-05}
{"train_loss": 0.1292869746685028, "global_step": 146727, "epoch": 1648, "lr": 4.378223039991751e-05}
{"train_loss": 0.10927337408065796, "global_step": 146728, "epoch": 1648, "lr": 4.378165516379727e-05}
{"train_loss": 0.1354476809501648, "global_step": 146729, "epoch": 1648, "lr": 4.378107992851298e-05}
{"train_loss": 0.15505406260490417, "global_step": 146730, "epoch": 1648, "lr": 4.378050469406478e-05}
{"train_loss": 0.09421753138303757, "global_step": 146731, "epoch": 1648, "lr": 4.3779929460452684e-05}
{"train_loss": 0.11244309693574905, "global_step": 146732, "epoch": 1648, "lr": 4.377935422767683e-05}
{"train_loss": 0.12152481824159622, "global_step": 146733, "epoch": 1648, "lr": 4.377877899573723e-05}
{"train_loss": 0.1520356386899948, "global_step": 146734, "epoch": 1648, "lr": 4.377820376463401e-05}
{"train_loss": 0.04961822181940079, "global_step": 146735, "epoch": 1648, "lr": 4.377762853436726e-05}
{"train_loss": 0.0972958654165268, "global_step": 146736, "epoch": 1648, "lr": 4.377705330493701e-05}
{"train_loss": 0.08764855563640594, "global_step": 146737, "epoch": 1648, "lr": 4.377647807634337e-05}
{"train_loss": 0.08768469840288162, "global_step": 146738, "epoch": 1648, "lr": 4.377590284858639e-05}
{"train_loss": 0.07014559209346771, "global_step": 146739, "epoch": 1648, "lr": 4.3775327621666186e-05}
{"train_loss": 0.0785631611943245, "global_step": 146740, "epoch": 1648, "lr": 4.37747523955828e-05}
{"train_loss": 0.14647045731544495, "global_step": 146741, "epoch": 1648, "lr": 4.377417717033633e-05}
{"train_loss": 0.11615163087844849, "global_step": 146742, "epoch": 1648, "lr": 4.377360194592684e-05}
{"train_loss": 0.07511207461357117, "global_step": 146743, "epoch": 1648, "lr": 4.3773026722354414e-05}
{"train_loss": 0.07493764907121658, "global_step": 146744, "epoch": 1648, "lr": 4.3772451499619125e-05}
{"train_loss": 0.10377892851829529, "global_step": 146745, "epoch": 1648, "lr": 4.377187627772108e-05}
{"train_loss": 0.09872856736183167, "global_step": 146746, "epoch": 1648, "lr": 4.377130105666029e-05}
{"train_loss": 0.1281527280807495, "global_step": 146747, "epoch": 1648, "lr": 4.37707258364369e-05}
{"train_loss": 0.08966725319623947, "global_step": 146748, "epoch": 1648, "lr": 4.3770150617050945e-05}
{"train_loss": 0.06962523609399796, "global_step": 146749, "epoch": 1648, "lr": 4.3769575398502534e-05}
{"train_loss": 0.09312312304973602, "global_step": 146750, "epoch": 1648, "lr": 4.376900018079171e-05}
{"train_loss": 0.13812728226184845, "global_step": 146751, "epoch": 1648, "lr": 4.376842496391858e-05}
{"train_loss": 0.13667306303977966, "global_step": 146752, "epoch": 1648, "lr": 4.376784974788319e-05}
{"train_loss": 0.10771934688091278, "global_step": 146753, "epoch": 1648, "lr": 4.376727453268564e-05}
{"train_loss": 0.097873754799366, "global_step": 146754, "epoch": 1648, "lr": 4.376669931832603e-05}
{"train_loss": 0.08549201488494873, "global_step": 146755, "epoch": 1648, "lr": 4.376612410480438e-05}
{"train_loss": 0.07193756848573685, "global_step": 146756, "epoch": 1648, "lr": 4.3765548892120834e-05}
{"train_loss": 0.18339380621910095, "global_step": 146757, "epoch": 1648, "lr": 4.37649736802754e-05}
{"train_loss": 0.1194068193435669, "global_step": 146758, "epoch": 1648, "lr": 4.376439846926822e-05}
{"train_loss": 0.06515350192785263, "global_step": 146759, "epoch": 1648, "lr": 4.376382325909932e-05}
{"train_loss": 0.10880005928907502, "global_step": 146760, "epoch": 1648, "lr": 4.376324804976881e-05, "val_loss": 6.265961170196533}
{"train_loss": 0.12887528538703918, "global_step": 146761, "epoch": 1649, "lr": 4.3762672841276745e-05}
{"train_loss": 0.1179877370595932, "global_step": 146762, "epoch": 1649, "lr": 4.376209763362322e-05}
{"train_loss": 0.10276425629854202, "global_step": 146763, "epoch": 1649, "lr": 4.37615224268083e-05}
{"train_loss": 0.07132761180400848, "global_step": 146764, "epoch": 1649, "lr": 4.3760947220832065e-05}
{"train_loss": 0.1283327043056488, "global_step": 146765, "epoch": 1649, "lr": 4.37603720156946e-05}
{"train_loss": 0.06571415811777115, "global_step": 146766, "epoch": 1649, "lr": 4.3759796811395984e-05}
{"train_loss": 0.06060270965099335, "global_step": 146767, "epoch": 1649, "lr": 4.3759221607936265e-05}
{"train_loss": 0.10549913346767426, "global_step": 146768, "epoch": 1649, "lr": 4.375864640531557e-05}
{"train_loss": 0.1018308773636818, "global_step": 146769, "epoch": 1649, "lr": 4.3758071203533924e-05}
{"train_loss": 0.10119591653347015, "global_step": 146770, "epoch": 1649, "lr": 4.3757496002591424e-05}
{"train_loss": 0.12805168330669403, "global_step": 146771, "epoch": 1649, "lr": 4.375692080248818e-05}
{"train_loss": 0.15815387666225433, "global_step": 146772, "epoch": 1649, "lr": 4.375634560322422e-05}
{"train_loss": 0.16838888823986053, "global_step": 146773, "epoch": 1649, "lr": 4.3755770404799665e-05}
{"train_loss": 0.11375926434993744, "global_step": 146774, "epoch": 1649, "lr": 4.3755195207214546e-05}
{"train_loss": 0.06638617068529129, "global_step": 146775, "epoch": 1649, "lr": 4.3754620010468986e-05}
{"train_loss": 0.09619569778442383, "global_step": 146776, "epoch": 1649, "lr": 4.3754044814563025e-05}
{"train_loss": 0.10518277436494827, "global_step": 146777, "epoch": 1649, "lr": 4.375346961949678e-05}
{"train_loss": 0.08794784545898438, "global_step": 146778, "epoch": 1649, "lr": 4.3752894425270275e-05}
{"train_loss": 0.08626298606395721, "global_step": 146779, "epoch": 1649, "lr": 4.375231923188364e-05}
{"train_loss": 0.12499693781137466, "global_step": 146780, "epoch": 1649, "lr": 4.3751744039336925e-05}
{"train_loss": 0.1401059627532959, "global_step": 146781, "epoch": 1649, "lr": 4.375116884763021e-05}
{"train_loss": 0.10460838675498962, "global_step": 146782, "epoch": 1649, "lr": 4.3750593656763565e-05}
{"train_loss": 0.11756613105535507, "global_step": 146783, "epoch": 1649, "lr": 4.3750018466737084e-05}
{"train_loss": 0.08982156217098236, "global_step": 146784, "epoch": 1649, "lr": 4.374944327755084e-05}
{"train_loss": 0.09015924483537674, "global_step": 146785, "epoch": 1649, "lr": 4.374886808920491e-05}
{"train_loss": 0.13024316728115082, "global_step": 146786, "epoch": 1649, "lr": 4.3748292901699357e-05}
{"train_loss": 0.11177128553390503, "global_step": 146787, "epoch": 1649, "lr": 4.374771771503429e-05}
{"train_loss": 0.1523078829050064, "global_step": 146788, "epoch": 1649, "lr": 4.3747142529209734e-05}
{"train_loss": 0.08582637459039688, "global_step": 146789, "epoch": 1649, "lr": 4.374656734422581e-05}
{"train_loss": 0.08987356722354889, "global_step": 146790, "epoch": 1649, "lr": 4.3745992160082595e-05}
{"train_loss": 0.08747431635856628, "global_step": 146791, "epoch": 1649, "lr": 4.374541697678014e-05}
{"train_loss": 0.08052846044301987, "global_step": 146792, "epoch": 1649, "lr": 4.374484179431856e-05}
{"train_loss": 0.07132774591445923, "global_step": 146793, "epoch": 1649, "lr": 4.374426661269788e-05}
{"train_loss": 0.06207878887653351, "global_step": 146794, "epoch": 1649, "lr": 4.3743691431918235e-05}
{"train_loss": 0.05475383996963501, "global_step": 146795, "epoch": 1649, "lr": 4.374311625197965e-05}
{"train_loss": 0.16122490167617798, "global_step": 146796, "epoch": 1649, "lr": 4.374254107288226e-05}
{"train_loss": 0.15383470058441162, "global_step": 146797, "epoch": 1649, "lr": 4.3741965894626075e-05}
{"train_loss": 0.12486972659826279, "global_step": 146798, "epoch": 1649, "lr": 4.374139071721123e-05}
{"train_loss": 0.10297591239213943, "global_step": 146799, "epoch": 1649, "lr": 4.3740815540637766e-05}
{"train_loss": 0.0731901302933693, "global_step": 146800, "epoch": 1649, "lr": 4.374024036490578e-05}
{"train_loss": 0.10402249544858932, "global_step": 146801, "epoch": 1649, "lr": 4.3739665190015336e-05}
{"train_loss": 0.07960592210292816, "global_step": 146802, "epoch": 1649, "lr": 4.373909001596652e-05}
{"train_loss": 0.08600897341966629, "global_step": 146803, "epoch": 1649, "lr": 4.3738514842759415e-05}
{"train_loss": 0.11015987396240234, "global_step": 146804, "epoch": 1649, "lr": 4.373793967039409e-05}
{"train_loss": 0.16871093213558197, "global_step": 146805, "epoch": 1649, "lr": 4.3737364498870606e-05}
{"train_loss": 0.12828411161899567, "global_step": 146806, "epoch": 1649, "lr": 4.3736789328189056e-05}
{"train_loss": 0.10456202924251556, "global_step": 146807, "epoch": 1649, "lr": 4.3736214158349545e-05}
{"train_loss": 0.07188616693019867, "global_step": 146808, "epoch": 1649, "lr": 4.373563898935209e-05}
{"train_loss": 0.10620299726724625, "global_step": 146809, "epoch": 1649, "lr": 4.373506382119683e-05}
{"train_loss": 0.08895041048526764, "global_step": 146810, "epoch": 1649, "lr": 4.3734488653883794e-05}
{"train_loss": 0.1705484241247177, "global_step": 146811, "epoch": 1649, "lr": 4.37339134874131e-05}
{"train_loss": 0.12457305938005447, "global_step": 146812, "epoch": 1649, "lr": 4.373333832178478e-05}
{"train_loss": 0.08478524535894394, "global_step": 146813, "epoch": 1649, "lr": 4.373276315699897e-05}
{"train_loss": 0.12111332267522812, "global_step": 146814, "epoch": 1649, "lr": 4.373218799305568e-05}
{"train_loss": 0.13781963288784027, "global_step": 146815, "epoch": 1649, "lr": 4.373161282995505e-05}
{"train_loss": 0.07700339704751968, "global_step": 146816, "epoch": 1649, "lr": 4.37310376676971e-05}
{"train_loss": 0.11489861458539963, "global_step": 146817, "epoch": 1649, "lr": 4.373046250628196e-05}
{"train_loss": 0.04962731525301933, "global_step": 146818, "epoch": 1649, "lr": 4.372988734570967e-05}
{"train_loss": 0.11372403800487518, "global_step": 146819, "epoch": 1649, "lr": 4.372931218598032e-05}
{"train_loss": 0.08416149765253067, "global_step": 146820, "epoch": 1649, "lr": 4.372873702709399e-05}
{"train_loss": 0.13792677223682404, "global_step": 146821, "epoch": 1649, "lr": 4.3728161869050755e-05}
{"train_loss": 0.07360812276601791, "global_step": 146822, "epoch": 1649, "lr": 4.3727586711850697e-05}
{"train_loss": 0.15180052816867828, "global_step": 146823, "epoch": 1649, "lr": 4.372701155549389e-05}
{"train_loss": 0.18099036812782288, "global_step": 146824, "epoch": 1649, "lr": 4.3726436399980396e-05}
{"train_loss": 0.1401304006576538, "global_step": 146825, "epoch": 1649, "lr": 4.3725861245310304e-05}
{"train_loss": 0.06411990523338318, "global_step": 146826, "epoch": 1649, "lr": 4.372528609148371e-05}
{"train_loss": 0.0852004736661911, "global_step": 146827, "epoch": 1649, "lr": 4.372471093850066e-05}
{"train_loss": 0.10001062601804733, "global_step": 146828, "epoch": 1649, "lr": 4.3724135786361264e-05}
{"train_loss": 0.06649763882160187, "global_step": 146829, "epoch": 1649, "lr": 4.372356063506556e-05}
{"train_loss": 0.051167313009500504, "global_step": 146830, "epoch": 1649, "lr": 4.3722985484613665e-05}
{"train_loss": 0.07850424200296402, "global_step": 146831, "epoch": 1649, "lr": 4.372241033500562e-05}
{"train_loss": 0.11277151107788086, "global_step": 146832, "epoch": 1649, "lr": 4.372183518624155e-05}
{"train_loss": 0.12708425521850586, "global_step": 146833, "epoch": 1649, "lr": 4.3721260038321466e-05}
{"train_loss": 0.10989425331354141, "global_step": 146834, "epoch": 1649, "lr": 4.372068489124551e-05}
{"train_loss": 0.121274434030056, "global_step": 146835, "epoch": 1649, "lr": 4.372010974501372e-05}
{"train_loss": 0.09205285459756851, "global_step": 146836, "epoch": 1649, "lr": 4.371953459962619e-05}
{"train_loss": 0.08894644677639008, "global_step": 146837, "epoch": 1649, "lr": 4.371895945508299e-05}
{"train_loss": 0.08543511480093002, "global_step": 146838, "epoch": 1649, "lr": 4.37183843113842e-05}
{"train_loss": 0.13341669738292694, "global_step": 146839, "epoch": 1649, "lr": 4.371780916852989e-05}
{"train_loss": 0.18056148290634155, "global_step": 146840, "epoch": 1649, "lr": 4.371723402652015e-05}
{"train_loss": 0.05975709855556488, "global_step": 146841, "epoch": 1649, "lr": 4.371665888535505e-05}
{"train_loss": 0.07479552179574966, "global_step": 146842, "epoch": 1649, "lr": 4.3716083745034656e-05}
{"train_loss": 0.10473557561635971, "global_step": 146843, "epoch": 1649, "lr": 4.371550860555909e-05}
{"train_loss": 0.13339638710021973, "global_step": 146844, "epoch": 1649, "lr": 4.3714933466928356e-05}
{"train_loss": 0.07676693797111511, "global_step": 146845, "epoch": 1649, "lr": 4.371435832914261e-05}
{"train_loss": 0.1363905817270279, "global_step": 146846, "epoch": 1649, "lr": 4.371378319220186e-05}
{"train_loss": 0.09144513309001923, "global_step": 146847, "epoch": 1649, "lr": 4.3713208056106245e-05}
{"train_loss": 0.12605096399784088, "global_step": 146848, "epoch": 1649, "lr": 4.371263292085579e-05}
{"train_loss": 0.1054357664471262, "global_step": 146849, "epoch": 1649, "lr": 4.371205778645062e-05, "val_loss": 6.282163143157959}
{"train_loss": 0.07426666468381882, "global_step": 146850, "epoch": 1650, "lr": 4.3711482652890753e-05}
{"train_loss": 0.07283705472946167, "global_step": 146851, "epoch": 1650, "lr": 4.371090752017633e-05}
{"train_loss": 0.12557241320610046, "global_step": 146852, "epoch": 1650, "lr": 4.371033238830738e-05}
{"train_loss": 0.13355165719985962, "global_step": 146853, "epoch": 1650, "lr": 4.370975725728401e-05}
{"train_loss": 0.1974753439426422, "global_step": 146854, "epoch": 1650, "lr": 4.3709182127106275e-05}
{"train_loss": 0.1625014990568161, "global_step": 146855, "epoch": 1650, "lr": 4.3708606997774285e-05}
{"train_loss": 0.14470802247524261, "global_step": 146856, "epoch": 1650, "lr": 4.370803186928808e-05}
{"train_loss": 0.07871202379465103, "global_step": 146857, "epoch": 1650, "lr": 4.370745674164776e-05}
{"train_loss": 0.10251429677009583, "global_step": 146858, "epoch": 1650, "lr": 4.370688161485339e-05}
{"train_loss": 0.13721607625484467, "global_step": 146859, "epoch": 1650, "lr": 4.370630648890506e-05}
{"train_loss": 0.07787282764911652, "global_step": 146860, "epoch": 1650, "lr": 4.3705731363802835e-05}
{"train_loss": 0.12672998011112213, "global_step": 146861, "epoch": 1650, "lr": 4.370515623954679e-05}
{"train_loss": 0.07873684167861938, "global_step": 146862, "epoch": 1650, "lr": 4.370458111613702e-05}
{"train_loss": 0.13539329171180725, "global_step": 146863, "epoch": 1650, "lr": 4.3704005993573585e-05}
{"train_loss": 0.07713621854782104, "global_step": 146864, "epoch": 1650, "lr": 4.3703430871856584e-05}
{"train_loss": 0.10342704504728317, "global_step": 146865, "epoch": 1650, "lr": 4.370285575098606e-05}
{"train_loss": 0.13601620495319366, "global_step": 146866, "epoch": 1650, "lr": 4.3702280630962136e-05}
{"train_loss": 0.1423605978488922, "global_step": 146867, "epoch": 1650, "lr": 4.370170551178484e-05}
{"train_loss": 0.061328813433647156, "global_step": 146868, "epoch": 1650, "lr": 4.370113039345429e-05}
{"train_loss": 0.05797746405005455, "global_step": 146869, "epoch": 1650, "lr": 4.370055527597052e-05}
{"train_loss": 0.11735640466213226, "global_step": 146870, "epoch": 1650, "lr": 4.3699980159333664e-05}
{"train_loss": 0.1166132241487503, "global_step": 146871, "epoch": 1650, "lr": 4.369940504354375e-05}
{"train_loss": 0.13961370289325714, "global_step": 146872, "epoch": 1650, "lr": 4.3698829928600894e-05}
{"train_loss": 0.09760395437479019, "global_step": 146873, "epoch": 1650, "lr": 4.369825481450512e-05}
{"train_loss": 0.10785558819770813, "global_step": 146874, "epoch": 1650, "lr": 4.369767970125658e-05}
{"train_loss": 0.08461698144674301, "global_step": 146875, "epoch": 1650, "lr": 4.3697104588855275e-05}
{"train_loss": 0.08715241402387619, "global_step": 146876, "epoch": 1650, "lr": 4.369652947730134e-05}
{"train_loss": 0.111823670566082, "global_step": 146877, "epoch": 1650, "lr": 4.369595436659483e-05}
{"train_loss": 0.06315337121486664, "global_step": 146878, "epoch": 1650, "lr": 4.3695379256735806e-05}
{"train_loss": 0.1644015610218048, "global_step": 146879, "epoch": 1650, "lr": 4.3694804147724374e-05}
{"train_loss": 0.11643607914447784, "global_step": 146880, "epoch": 1650, "lr": 4.369422903956058e-05}
{"train_loss": 0.12019173800945282, "global_step": 146881, "epoch": 1650, "lr": 4.3693653932244546e-05}
{"train_loss": 0.08065309375524521, "global_step": 146882, "epoch": 1650, "lr": 4.36930788257763e-05}
{"train_loss": 0.09708885848522186, "global_step": 146883, "epoch": 1650, "lr": 4.369250372015597e-05}
{"train_loss": 0.13421110808849335, "global_step": 146884, "epoch": 1650, "lr": 4.369192861538357e-05}
{"train_loss": 0.10914343595504761, "global_step": 146885, "epoch": 1650, "lr": 4.369135351145925e-05}
{"train_loss": 0.08246751874685287, "global_step": 146886, "epoch": 1650, "lr": 4.369077840838302e-05}
{"train_loss": 0.11986043304204941, "global_step": 146887, "epoch": 1650, "lr": 4.3690203306155016e-05}
{"train_loss": 0.06500079482793808, "global_step": 146888, "epoch": 1650, "lr": 4.3689628204775256e-05}
{"train_loss": 0.14187480509281158, "global_step": 146889, "epoch": 1650, "lr": 4.3689053104243884e-05}
{"train_loss": 0.09336362779140472, "global_step": 146890, "epoch": 1650, "lr": 4.3688478004560907e-05}
{"train_loss": 0.09975128620862961, "global_step": 146891, "epoch": 1650, "lr": 4.368790290572646e-05}
{"train_loss": 0.07942944020032883, "global_step": 146892, "epoch": 1650, "lr": 4.3687327807740585e-05}
{"train_loss": 0.12568166851997375, "global_step": 146893, "epoch": 1650, "lr": 4.368675271060339e-05}
{"train_loss": 0.060837820172309875, "global_step": 146894, "epoch": 1650, "lr": 4.368617761431491e-05}
{"train_loss": 0.09233497828245163, "global_step": 146895, "epoch": 1650, "lr": 4.368560251887527e-05}
{"train_loss": 0.11561035364866257, "global_step": 146896, "epoch": 1650, "lr": 4.3685027424284516e-05}
{"train_loss": 0.08301497995853424, "global_step": 146897, "epoch": 1650, "lr": 4.3684452330542716e-05}
{"train_loss": 0.07951705157756805, "global_step": 146898, "epoch": 1650, "lr": 4.368387723764999e-05}
{"train_loss": 0.07583864033222198, "global_step": 146899, "epoch": 1650, "lr": 4.368330214560636e-05}
{"train_loss": 0.12166450917720795, "global_step": 146900, "epoch": 1650, "lr": 4.368272705441197e-05}
{"train_loss": 0.1403786540031433, "global_step": 146901, "epoch": 1650, "lr": 4.368215196406683e-05}
{"train_loss": 0.17309609055519104, "global_step": 146902, "epoch": 1650, "lr": 4.3681576874571065e-05}
{"train_loss": 0.08268264681100845, "global_step": 146903, "epoch": 1650, "lr": 4.3681001785924724e-05}
{"train_loss": 0.2017229199409485, "global_step": 146904, "epoch": 1650, "lr": 4.368042669812791e-05}
{"train_loss": 0.0742972269654274, "global_step": 146905, "epoch": 1650, "lr": 4.367985161118067e-05}
{"train_loss": 0.11027327179908752, "global_step": 146906, "epoch": 1650, "lr": 4.3679276525083105e-05}
{"train_loss": 0.07440939545631409, "global_step": 146907, "epoch": 1650, "lr": 4.367870143983528e-05}
{"train_loss": 0.14186535775661469, "global_step": 146908, "epoch": 1650, "lr": 4.367812635543728e-05}
{"train_loss": 0.09864091873168945, "global_step": 146909, "epoch": 1650, "lr": 4.367755127188917e-05}
{"train_loss": 0.14701928198337555, "global_step": 146910, "epoch": 1650, "lr": 4.367697618919105e-05}
{"train_loss": 0.1319463551044464, "global_step": 146911, "epoch": 1650, "lr": 4.367640110734297e-05}
{"train_loss": 0.13991403579711914, "global_step": 146912, "epoch": 1650, "lr": 4.367582602634503e-05}
{"train_loss": 0.1259630173444748, "global_step": 146913, "epoch": 1650, "lr": 4.367525094619729e-05}
{"train_loss": 0.11564186960458755, "global_step": 146914, "epoch": 1650, "lr": 4.3674675866899856e-05}
{"train_loss": 0.17741505801677704, "global_step": 146915, "epoch": 1650, "lr": 4.367410078845277e-05}
{"train_loss": 0.10703691095113754, "global_step": 146916, "epoch": 1650, "lr": 4.3673525710856114e-05}
{"train_loss": 0.09706485271453857, "global_step": 146917, "epoch": 1650, "lr": 4.3672950634109996e-05}
{"train_loss": 0.0928548127412796, "global_step": 146918, "epoch": 1650, "lr": 4.367237555821445e-05}
{"train_loss": 0.12434504181146622, "global_step": 146919, "epoch": 1650, "lr": 4.36718004831696e-05}
{"train_loss": 0.10444970428943634, "global_step": 146920, "epoch": 1650, "lr": 4.367122540897548e-05}
{"train_loss": 0.11723075062036514, "global_step": 146921, "epoch": 1650, "lr": 4.36706503356322e-05}
{"train_loss": 0.09178584069013596, "global_step": 146922, "epoch": 1650, "lr": 4.367007526313981e-05}
{"train_loss": 0.16502781212329865, "global_step": 146923, "epoch": 1650, "lr": 4.3669500191498414e-05}
{"train_loss": 0.0950964018702507, "global_step": 146924, "epoch": 1650, "lr": 4.366892512070806e-05}
{"train_loss": 0.10561664402484894, "global_step": 146925, "epoch": 1650, "lr": 4.366835005076887e-05}
{"train_loss": 0.09891244769096375, "global_step": 146926, "epoch": 1650, "lr": 4.3667774981680867e-05}
{"train_loss": 0.10417687892913818, "global_step": 146927, "epoch": 1650, "lr": 4.366719991344417e-05}
{"train_loss": 0.11791053414344788, "global_step": 146928, "epoch": 1650, "lr": 4.3666624846058824e-05}
{"train_loss": 0.05895775556564331, "global_step": 146929, "epoch": 1650, "lr": 4.366604977952494e-05}
{"train_loss": 0.09465669840574265, "global_step": 146930, "epoch": 1650, "lr": 4.366547471384257e-05}
{"train_loss": 0.1480971723794937, "global_step": 146931, "epoch": 1650, "lr": 4.366489964901181e-05}
{"train_loss": 0.08069764077663422, "global_step": 146932, "epoch": 1650, "lr": 4.366432458503271e-05}
{"train_loss": 0.13158099353313446, "global_step": 146933, "epoch": 1650, "lr": 4.3663749521905375e-05}
{"train_loss": 0.10860716551542282, "global_step": 146934, "epoch": 1650, "lr": 4.366317445962988e-05}
{"train_loss": 0.14176127314567566, "global_step": 146935, "epoch": 1650, "lr": 4.366259939820627e-05}
{"train_loss": 0.05472567304968834, "global_step": 146936, "epoch": 1650, "lr": 4.3662024337634674e-05}
{"train_loss": 0.12024357914924622, "global_step": 146937, "epoch": 1650, "lr": 4.366144927791511e-05}
{"train_loss": 0.10947730375474758, "global_step": 146938, "epoch": 1650, "lr": 4.366087421904772e-05, "train/sim_max_reward_0": 0.5528023409838152, "train/sim_max_reward_1": 0.9017127681522926, "train/sim_max_reward_2": 0.9840385735668348, "train/sim_max_reward_3": 0.40919483030379805, "train/sim_max_reward_4": 0.2457020537624324, "train/sim_max_reward_5": 0.6485938322864392, "test/sim_max_reward_4500000": 0.42614211028441523, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9365717402527775, "test/sim_max_reward_4500003": 0.10937968762145041, "test/sim_max_reward_4500004": 0.5263946025014115, "test/sim_max_reward_4500005": 0.9824841749830836, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9785497593996321, "test/sim_max_reward_4500008": 0.8894181728691372, "test/sim_max_reward_4500009": 0.050998891705189286, "test/sim_max_reward_4500010": 1.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5594455936945921, "test/sim_max_reward_4500014": 0.9993026451443555, "test/sim_max_reward_4500015": 0.9789480660191344, "test/sim_max_reward_4500016": 0.9753277862078785, "test/sim_max_reward_4500017": 0.8695305466395785, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.7563874698699824, "test/sim_max_reward_4500021": 0.6868563531843392, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6997713967751701, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.9933856584693154, "test/sim_max_reward_4500027": 0.9697774793266631, "test/sim_max_reward_4500028": 0.6562995708924391, "test/sim_max_reward_4500029": 0.7093633835534313, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.3793886411882573, "test/sim_max_reward_4500032": 0.4531571722508592, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.7208293884534358, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.7628054367116062, "test/sim_max_reward_4500038": 0.16263183781831383, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6269313015861611, "test/sim_max_reward_4500042": 0.423534625665682, "test/sim_max_reward_4500043": 0.21250893692528328, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.6776729175088028, "test/sim_max_reward_4500046": 0.6066135341397182, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9391687820485286, "test/sim_max_reward_4500049": 0.5647634777495953, "train/mean_score": 0.6236740665092687, "test/mean_score": 0.7056868228288045, "val_loss": 6.274550437927246, "train_action_mse_error": 32.12614059448242}
{"train_loss": 0.14086055755615234, "global_step": 146939, "epoch": 1651, "lr": 4.366029916103253e-05}
{"train_loss": 0.1043061912059784, "global_step": 146940, "epoch": 1651, "lr": 4.365972410386964e-05}
{"train_loss": 0.08794243633747101, "global_step": 146941, "epoch": 1651, "lr": 4.365914904755911e-05}
{"train_loss": 0.10011248290538788, "global_step": 146942, "epoch": 1651, "lr": 4.3658573992101057e-05}
{"train_loss": 0.15003491938114166, "global_step": 146943, "epoch": 1651, "lr": 4.3657998937495506e-05}
{"train_loss": 0.08672372251749039, "global_step": 146944, "epoch": 1651, "lr": 4.365742388374258e-05}
{"train_loss": 0.19029037654399872, "global_step": 146945, "epoch": 1651, "lr": 4.365684883084231e-05}
{"train_loss": 0.11026765406131744, "global_step": 146946, "epoch": 1651, "lr": 4.365627377879483e-05}
{"train_loss": 0.10921710729598999, "global_step": 146947, "epoch": 1651, "lr": 4.365569872760016e-05}
{"train_loss": 0.107350654900074, "global_step": 146948, "epoch": 1651, "lr": 4.3655123677258425e-05}
{"train_loss": 0.14811542630195618, "global_step": 146949, "epoch": 1651, "lr": 4.3654548627769665e-05}
{"train_loss": 0.08107703179121017, "global_step": 146950, "epoch": 1651, "lr": 4.365397357913399e-05}
{"train_loss": 0.11863800138235092, "global_step": 146951, "epoch": 1651, "lr": 4.3653398531351444e-05}
{"train_loss": 0.14684608578681946, "global_step": 146952, "epoch": 1651, "lr": 4.365282348442212e-05}
{"train_loss": 0.10437992960214615, "global_step": 146953, "epoch": 1651, "lr": 4.365224843834612e-05}
{"train_loss": 0.11015606671571732, "global_step": 146954, "epoch": 1651, "lr": 4.365167339312347e-05}
{"train_loss": 0.07036054879426956, "global_step": 146955, "epoch": 1651, "lr": 4.3651098348754304e-05}
{"train_loss": 0.215841606259346, "global_step": 146956, "epoch": 1651, "lr": 4.3650523305238644e-05}
{"train_loss": 0.14349116384983063, "global_step": 146957, "epoch": 1651, "lr": 4.364994826257661e-05}
{"train_loss": 0.08560759574174881, "global_step": 146958, "epoch": 1651, "lr": 4.364937322076825e-05}
{"train_loss": 0.11143001168966293, "global_step": 146959, "epoch": 1651, "lr": 4.364879817981367e-05}
{"train_loss": 0.09860783070325851, "global_step": 146960, "epoch": 1651, "lr": 4.364822313971291e-05}
{"train_loss": 0.1095006912946701, "global_step": 146961, "epoch": 1651, "lr": 4.36476481004661e-05}
{"train_loss": 0.06649746745824814, "global_step": 146962, "epoch": 1651, "lr": 4.3647073062073255e-05}
{"train_loss": 0.1305852234363556, "global_step": 146963, "epoch": 1651, "lr": 4.364649802453451e-05}
{"train_loss": 0.15004469454288483, "global_step": 146964, "epoch": 1651, "lr": 4.364592298784989e-05}
{"train_loss": 0.041794005781412125, "global_step": 146965, "epoch": 1651, "lr": 4.364534795201952e-05}
{"train_loss": 0.10447181016206741, "global_step": 146966, "epoch": 1651, "lr": 4.364477291704344e-05}
{"train_loss": 0.11448401212692261, "global_step": 146967, "epoch": 1651, "lr": 4.3644197882921764e-05}
{"train_loss": 0.06860452145338058, "global_step": 146968, "epoch": 1651, "lr": 4.364362284965452e-05}
{"train_loss": 0.10883887112140656, "global_step": 146969, "epoch": 1651, "lr": 4.364304781724183e-05}
{"train_loss": 0.09438467770814896, "global_step": 146970, "epoch": 1651, "lr": 4.364247278568375e-05}
{"train_loss": 0.07812477648258209, "global_step": 146971, "epoch": 1651, "lr": 4.3641897754980366e-05}
{"train_loss": 0.08763393014669418, "global_step": 146972, "epoch": 1651, "lr": 4.364132272513176e-05}
{"train_loss": 0.19739320874214172, "global_step": 146973, "epoch": 1651, "lr": 4.364074769613798e-05}
{"train_loss": 0.08878879249095917, "global_step": 146974, "epoch": 1651, "lr": 4.364017266799914e-05}
{"train_loss": 0.09102068841457367, "global_step": 146975, "epoch": 1651, "lr": 4.3639597640715284e-05}
{"train_loss": 0.044891107827425, "global_step": 146976, "epoch": 1651, "lr": 4.363902261428653e-05}
{"train_loss": 0.11808834224939346, "global_step": 146977, "epoch": 1651, "lr": 4.363844758871291e-05}
{"train_loss": 0.11854521185159683, "global_step": 146978, "epoch": 1651, "lr": 4.3637872563994545e-05}
{"train_loss": 0.07440491765737534, "global_step": 146979, "epoch": 1651, "lr": 4.3637297540131465e-05}
{"train_loss": 0.07674045115709305, "global_step": 146980, "epoch": 1651, "lr": 4.36367225171238e-05}
{"train_loss": 0.04710715636610985, "global_step": 146981, "epoch": 1651, "lr": 4.363614749497157e-05}
{"train_loss": 0.1350284367799759, "global_step": 146982, "epoch": 1651, "lr": 4.3635572473674915e-05}
{"train_loss": 0.10370106250047684, "global_step": 146983, "epoch": 1651, "lr": 4.363499745323385e-05}
{"train_loss": 0.09651698917150497, "global_step": 146984, "epoch": 1651, "lr": 4.363442243364851e-05}
{"train_loss": 0.10862289369106293, "global_step": 146985, "epoch": 1651, "lr": 4.363384741491892e-05}
{"train_loss": 0.07149172574281693, "global_step": 146986, "epoch": 1651, "lr": 4.36332723970452e-05}
{"train_loss": 0.09521980583667755, "global_step": 146987, "epoch": 1651, "lr": 4.363269738002739e-05}
{"train_loss": 0.08410720527172089, "global_step": 146988, "epoch": 1651, "lr": 4.363212236386559e-05}
{"train_loss": 0.09794546663761139, "global_step": 146989, "epoch": 1651, "lr": 4.3631547348559885e-05}
{"train_loss": 0.12445587664842606, "global_step": 146990, "epoch": 1651, "lr": 4.3630972334110336e-05}
{"train_loss": 0.09654537588357925, "global_step": 146991, "epoch": 1651, "lr": 4.363039732051702e-05}
{"train_loss": 0.03130847215652466, "global_step": 146992, "epoch": 1651, "lr": 4.3629822307780033e-05}
{"train_loss": 0.13198637962341309, "global_step": 146993, "epoch": 1651, "lr": 4.362924729589944e-05}
{"train_loss": 0.08315028995275497, "global_step": 146994, "epoch": 1651, "lr": 4.3628672284875294e-05}
{"train_loss": 0.0701858252286911, "global_step": 146995, "epoch": 1651, "lr": 4.362809727470772e-05}
{"train_loss": 0.068424291908741, "global_step": 146996, "epoch": 1651, "lr": 4.362752226539675e-05}
{"train_loss": 0.07828234881162643, "global_step": 146997, "epoch": 1651, "lr": 4.36269472569425e-05}
{"train_loss": 0.10240191221237183, "global_step": 146998, "epoch": 1651, "lr": 4.3626372249345015e-05}
{"train_loss": 0.08729824423789978, "global_step": 146999, "epoch": 1651, "lr": 4.36257972426044e-05}
{"train_loss": 0.18182802200317383, "global_step": 147000, "epoch": 1651, "lr": 4.36252222367207e-05}
{"train_loss": 0.10853366553783417, "global_step": 147001, "epoch": 1651, "lr": 4.3624647231694036e-05}
{"train_loss": 0.13148072361946106, "global_step": 147002, "epoch": 1651, "lr": 4.362407222752444e-05}
{"train_loss": 0.1304975003004074, "global_step": 147003, "epoch": 1651, "lr": 4.3623497224212025e-05}
{"train_loss": 0.09050264954566956, "global_step": 147004, "epoch": 1651, "lr": 4.362292222175683e-05}
{"train_loss": 0.11022080481052399, "global_step": 147005, "epoch": 1651, "lr": 4.362234722015898e-05}
{"train_loss": 0.06848412752151489, "global_step": 147006, "epoch": 1651, "lr": 4.362177221941851e-05}
{"train_loss": 0.0541866309940815, "global_step": 147007, "epoch": 1651, "lr": 4.362119721953552e-05}
{"train_loss": 0.08230996876955032, "global_step": 147008, "epoch": 1651, "lr": 4.362062222051009e-05}
{"train_loss": 0.11990052461624146, "global_step": 147009, "epoch": 1651, "lr": 4.362004722234228e-05}
{"train_loss": 0.07184795290231705, "global_step": 147010, "epoch": 1651, "lr": 4.361947222503218e-05}
{"train_loss": 0.11935527622699738, "global_step": 147011, "epoch": 1651, "lr": 4.3618897228579866e-05}
{"train_loss": 0.0955565869808197, "global_step": 147012, "epoch": 1651, "lr": 4.361832223298543e-05}
{"train_loss": 0.1412387192249298, "global_step": 147013, "epoch": 1651, "lr": 4.3617747238248904e-05}
{"train_loss": 0.1285506784915924, "global_step": 147014, "epoch": 1651, "lr": 4.361717224437041e-05}
{"train_loss": 0.0760747492313385, "global_step": 147015, "epoch": 1651, "lr": 4.361659725135e-05}
{"train_loss": 0.10442063957452774, "global_step": 147016, "epoch": 1651, "lr": 4.361602225918777e-05}
{"train_loss": 0.0862543135881424, "global_step": 147017, "epoch": 1651, "lr": 4.361544726788378e-05}
{"train_loss": 0.07654831558465958, "global_step": 147018, "epoch": 1651, "lr": 4.361487227743813e-05}
{"train_loss": 0.048570454120635986, "global_step": 147019, "epoch": 1651, "lr": 4.361429728785086e-05}
{"train_loss": 0.09841258823871613, "global_step": 147020, "epoch": 1651, "lr": 4.3613722299122095e-05}
{"train_loss": 0.13796642422676086, "global_step": 147021, "epoch": 1651, "lr": 4.3613147311251865e-05}
{"train_loss": 0.09785430133342743, "global_step": 147022, "epoch": 1651, "lr": 4.361257232424029e-05}
{"train_loss": 0.09790273010730743, "global_step": 147023, "epoch": 1651, "lr": 4.36119973380874e-05}
{"train_loss": 0.07357526570558548, "global_step": 147024, "epoch": 1651, "lr": 4.36114223527933e-05}
{"train_loss": 0.10255591571331024, "global_step": 147025, "epoch": 1651, "lr": 4.3610847368358096e-05}
{"train_loss": 0.0716717466711998, "global_step": 147026, "epoch": 1651, "lr": 4.361027238478182e-05}
{"train_loss": 0.10233684202258507, "global_step": 147027, "epoch": 1651, "lr": 4.360969740206457e-05, "val_loss": 6.230957984924316}
{"train_loss": 0.07200444489717484, "global_step": 147028, "epoch": 1652, "lr": 4.3609122420206414e-05}
{"train_loss": 0.1386111080646515, "global_step": 147029, "epoch": 1652, "lr": 4.360854743920742e-05}
{"train_loss": 0.06467960774898529, "global_step": 147030, "epoch": 1652, "lr": 4.3607972459067703e-05}
{"train_loss": 0.12091320008039474, "global_step": 147031, "epoch": 1652, "lr": 4.3607397479787316e-05}
{"train_loss": 0.10021217912435532, "global_step": 147032, "epoch": 1652, "lr": 4.3606822501366314e-05}
{"train_loss": 0.06888358294963837, "global_step": 147033, "epoch": 1652, "lr": 4.360624752380482e-05}
{"train_loss": 0.12049756199121475, "global_step": 147034, "epoch": 1652, "lr": 4.360567254710287e-05}
{"train_loss": 0.12492144107818604, "global_step": 147035, "epoch": 1652, "lr": 4.3605097571260574e-05}
{"train_loss": 0.04362183064222336, "global_step": 147036, "epoch": 1652, "lr": 4.360452259627798e-05}
{"train_loss": 0.1121048629283905, "global_step": 147037, "epoch": 1652, "lr": 4.360394762215519e-05}
{"train_loss": 0.06887654960155487, "global_step": 147038, "epoch": 1652, "lr": 4.360337264889226e-05}
{"train_loss": 0.08039819449186325, "global_step": 147039, "epoch": 1652, "lr": 4.36027976764893e-05}
{"train_loss": 0.09387291222810745, "global_step": 147040, "epoch": 1652, "lr": 4.3602222704946336e-05}
{"train_loss": 0.0931038111448288, "global_step": 147041, "epoch": 1652, "lr": 4.360164773426351e-05}
{"train_loss": 0.09293331205844879, "global_step": 147042, "epoch": 1652, "lr": 4.360107276444083e-05}
{"train_loss": 0.15094490349292755, "global_step": 147043, "epoch": 1652, "lr": 4.360049779547841e-05}
{"train_loss": 0.10994365066289902, "global_step": 147044, "epoch": 1652, "lr": 4.359992282737636e-05}
{"train_loss": 0.13353537023067474, "global_step": 147045, "epoch": 1652, "lr": 4.35993478601347e-05}
{"train_loss": 0.09424419701099396, "global_step": 147046, "epoch": 1652, "lr": 4.3598772893753526e-05}
{"train_loss": 0.07689115405082703, "global_step": 147047, "epoch": 1652, "lr": 4.3598197928232924e-05}
{"train_loss": 0.0822078064084053, "global_step": 147048, "epoch": 1652, "lr": 4.359762296357297e-05}
{"train_loss": 0.08656521141529083, "global_step": 147049, "epoch": 1652, "lr": 4.3597047999773735e-05}
{"train_loss": 0.1144038662314415, "global_step": 147050, "epoch": 1652, "lr": 4.35964730368353e-05}
{"train_loss": 0.08074849098920822, "global_step": 147051, "epoch": 1652, "lr": 4.3595898074757734e-05}
{"train_loss": 0.13305312395095825, "global_step": 147052, "epoch": 1652, "lr": 4.3595323113541143e-05}
{"train_loss": 0.08988559246063232, "global_step": 147053, "epoch": 1652, "lr": 4.359474815318555e-05}
{"train_loss": 0.0835963636636734, "global_step": 147054, "epoch": 1652, "lr": 4.359417319369109e-05}
{"train_loss": 0.11246201395988464, "global_step": 147055, "epoch": 1652, "lr": 4.35935982350578e-05}
{"train_loss": 0.10065534710884094, "global_step": 147056, "epoch": 1652, "lr": 4.359302327728579e-05}
{"train_loss": 0.1543884128332138, "global_step": 147057, "epoch": 1652, "lr": 4.35924483203751e-05}
{"train_loss": 0.08110654354095459, "global_step": 147058, "epoch": 1652, "lr": 4.359187336432585e-05}
{"train_loss": 0.16748033463954926, "global_step": 147059, "epoch": 1652, "lr": 4.359129840913807e-05}
{"train_loss": 0.09772007912397385, "global_step": 147060, "epoch": 1652, "lr": 4.359072345481185e-05}
{"train_loss": 0.06283712387084961, "global_step": 147061, "epoch": 1652, "lr": 4.3590148501347315e-05}
{"train_loss": 0.09088609367609024, "global_step": 147062, "epoch": 1652, "lr": 4.358957354874449e-05}
{"train_loss": 0.1048136055469513, "global_step": 147063, "epoch": 1652, "lr": 4.3588998597003485e-05}
{"train_loss": 0.11946368962526321, "global_step": 147064, "epoch": 1652, "lr": 4.358842364612434e-05}
{"train_loss": 0.1020994484424591, "global_step": 147065, "epoch": 1652, "lr": 4.358784869610716e-05}
{"train_loss": 0.1001477837562561, "global_step": 147066, "epoch": 1652, "lr": 4.358727374695202e-05}
{"train_loss": 0.11065985262393951, "global_step": 147067, "epoch": 1652, "lr": 4.358669879865899e-05}
{"train_loss": 0.11123990267515182, "global_step": 147068, "epoch": 1652, "lr": 4.358612385122815e-05}
{"train_loss": 0.1072930097579956, "global_step": 147069, "epoch": 1652, "lr": 4.358554890465959e-05}
{"train_loss": 0.14760342240333557, "global_step": 147070, "epoch": 1652, "lr": 4.358497395895335e-05}
{"train_loss": 0.09680215269327164, "global_step": 147071, "epoch": 1652, "lr": 4.3584399014109556e-05}
{"train_loss": 0.14995688199996948, "global_step": 147072, "epoch": 1652, "lr": 4.3583824070128236e-05}
{"train_loss": 0.15415823459625244, "global_step": 147073, "epoch": 1652, "lr": 4.3583249127009514e-05}
{"train_loss": 0.1569909006357193, "global_step": 147074, "epoch": 1652, "lr": 4.358267418475344e-05}
{"train_loss": 0.09649405628442764, "global_step": 147075, "epoch": 1652, "lr": 4.3582099243360095e-05}
{"train_loss": 0.07402345538139343, "global_step": 147076, "epoch": 1652, "lr": 4.3581524302829554e-05}
{"train_loss": 0.08580390363931656, "global_step": 147077, "epoch": 1652, "lr": 4.3580949363161916e-05}
{"train_loss": 0.15776285529136658, "global_step": 147078, "epoch": 1652, "lr": 4.358037442435722e-05}
{"train_loss": 0.13585536181926727, "global_step": 147079, "epoch": 1652, "lr": 4.357979948641556e-05}
{"train_loss": 0.14320158958435059, "global_step": 147080, "epoch": 1652, "lr": 4.357922454933704e-05}
{"train_loss": 0.18049325048923492, "global_step": 147081, "epoch": 1652, "lr": 4.35786496131217e-05}
{"train_loss": 0.19070661067962646, "global_step": 147082, "epoch": 1652, "lr": 4.357807467776964e-05}
{"train_loss": 0.15615320205688477, "global_step": 147083, "epoch": 1652, "lr": 4.357749974328093e-05}
{"train_loss": 0.11443530023097992, "global_step": 147084, "epoch": 1652, "lr": 4.357692480965565e-05}
{"train_loss": 0.03366396948695183, "global_step": 147085, "epoch": 1652, "lr": 4.3576349876893866e-05}
{"train_loss": 0.09471431374549866, "global_step": 147086, "epoch": 1652, "lr": 4.357577494499567e-05}
{"train_loss": 0.06854328513145447, "global_step": 147087, "epoch": 1652, "lr": 4.3575200013961125e-05}
{"train_loss": 0.134803906083107, "global_step": 147088, "epoch": 1652, "lr": 4.357462508379033e-05}
{"train_loss": 0.15457773208618164, "global_step": 147089, "epoch": 1652, "lr": 4.357405015448333e-05}
{"train_loss": 0.17434754967689514, "global_step": 147090, "epoch": 1652, "lr": 4.357347522604024e-05}
{"train_loss": 0.1517956703901291, "global_step": 147091, "epoch": 1652, "lr": 4.3572900298461096e-05}
{"train_loss": 0.12682335078716278, "global_step": 147092, "epoch": 1652, "lr": 4.357232537174602e-05}
{"train_loss": 0.13038848340511322, "global_step": 147093, "epoch": 1652, "lr": 4.357175044589504e-05}
{"train_loss": 0.04729318991303444, "global_step": 147094, "epoch": 1652, "lr": 4.3571175520908294e-05}
{"train_loss": 0.10220647603273392, "global_step": 147095, "epoch": 1652, "lr": 4.357060059678579e-05}
{"train_loss": 0.10677945613861084, "global_step": 147096, "epoch": 1652, "lr": 4.3570025673527646e-05}
{"train_loss": 0.15263541042804718, "global_step": 147097, "epoch": 1652, "lr": 4.356945075113396e-05}
{"train_loss": 0.17617063224315643, "global_step": 147098, "epoch": 1652, "lr": 4.356887582960475e-05}
{"train_loss": 0.12122628092765808, "global_step": 147099, "epoch": 1652, "lr": 4.3568300908940154e-05}
{"train_loss": 0.10235219448804855, "global_step": 147100, "epoch": 1652, "lr": 4.35677259891402e-05}
{"train_loss": 0.12190929055213928, "global_step": 147101, "epoch": 1652, "lr": 4.356715107020501e-05}
{"train_loss": 0.057049427181482315, "global_step": 147102, "epoch": 1652, "lr": 4.3566576152134624e-05}
{"train_loss": 0.13399086892604828, "global_step": 147103, "epoch": 1652, "lr": 4.3566001234929133e-05}
{"train_loss": 0.13108006119728088, "global_step": 147104, "epoch": 1652, "lr": 4.356542631858862e-05}
{"train_loss": 0.1159924566745758, "global_step": 147105, "epoch": 1652, "lr": 4.356485140311315e-05}
{"train_loss": 0.14651232957839966, "global_step": 147106, "epoch": 1652, "lr": 4.356427648850281e-05}
{"train_loss": 0.09280066192150116, "global_step": 147107, "epoch": 1652, "lr": 4.356370157475768e-05}
{"train_loss": 0.10463103652000427, "global_step": 147108, "epoch": 1652, "lr": 4.356312666187783e-05}
{"train_loss": 0.1487269550561905, "global_step": 147109, "epoch": 1652, "lr": 4.356255174986335e-05}
{"train_loss": 0.06062250956892967, "global_step": 147110, "epoch": 1652, "lr": 4.356197683871428e-05}
{"train_loss": 0.24790485203266144, "global_step": 147111, "epoch": 1652, "lr": 4.356140192843074e-05}
{"train_loss": 0.1091955155134201, "global_step": 147112, "epoch": 1652, "lr": 4.3560827019012785e-05}
{"train_loss": 0.09977790713310242, "global_step": 147113, "epoch": 1652, "lr": 4.356025211046051e-05}
{"train_loss": 0.1404101550579071, "global_step": 147114, "epoch": 1652, "lr": 4.355967720277396e-05}
{"train_loss": 0.09730403870344162, "global_step": 147115, "epoch": 1652, "lr": 4.355910229595324e-05}
{"train_loss": 0.11339883037497488, "global_step": 147116, "epoch": 1652, "lr": 4.3558527389998425e-05, "val_loss": 6.299259662628174}
{"train_loss": 0.07287133485078812, "global_step": 147117, "epoch": 1653, "lr": 4.355795248490958e-05}
{"train_loss": 0.1842038631439209, "global_step": 147118, "epoch": 1653, "lr": 4.3557377580686806e-05}
{"train_loss": 0.08927592635154724, "global_step": 147119, "epoch": 1653, "lr": 4.355680267733015e-05}
{"train_loss": 0.07390332221984863, "global_step": 147120, "epoch": 1653, "lr": 4.355622777483972e-05}
{"train_loss": 0.09456207603216171, "global_step": 147121, "epoch": 1653, "lr": 4.355565287321555e-05}
{"train_loss": 0.12131910771131516, "global_step": 147122, "epoch": 1653, "lr": 4.355507797245777e-05}
{"train_loss": 0.12163905054330826, "global_step": 147123, "epoch": 1653, "lr": 4.355450307256642e-05}
{"train_loss": 0.13928750157356262, "global_step": 147124, "epoch": 1653, "lr": 4.3553928173541594e-05}
{"train_loss": 0.08898917585611343, "global_step": 147125, "epoch": 1653, "lr": 4.3553353275383354e-05}
{"train_loss": 0.10415977239608765, "global_step": 147126, "epoch": 1653, "lr": 4.35527783780918e-05}
{"train_loss": 0.05867275968194008, "global_step": 147127, "epoch": 1653, "lr": 4.355220348166698e-05}
{"train_loss": 0.08367201685905457, "global_step": 147128, "epoch": 1653, "lr": 4.3551628586109e-05}
{"train_loss": 0.10424791276454926, "global_step": 147129, "epoch": 1653, "lr": 4.355105369141792e-05}
{"train_loss": 0.06635799258947372, "global_step": 147130, "epoch": 1653, "lr": 4.355047879759383e-05}
{"train_loss": 0.09016348421573639, "global_step": 147131, "epoch": 1653, "lr": 4.354990390463679e-05}
{"train_loss": 0.09099473059177399, "global_step": 147132, "epoch": 1653, "lr": 4.354932901254689e-05}
{"train_loss": 0.08493155241012573, "global_step": 147133, "epoch": 1653, "lr": 4.3548754121324195e-05}
{"train_loss": 0.08610118180513382, "global_step": 147134, "epoch": 1653, "lr": 4.354817923096879e-05}
{"train_loss": 0.14653712511062622, "global_step": 147135, "epoch": 1653, "lr": 4.354760434148078e-05}
{"train_loss": 0.06617391109466553, "global_step": 147136, "epoch": 1653, "lr": 4.354702945286018e-05}
{"train_loss": 0.15354610979557037, "global_step": 147137, "epoch": 1653, "lr": 4.354645456510714e-05}
{"train_loss": 0.10526983439922333, "global_step": 147138, "epoch": 1653, "lr": 4.354587967822167e-05}
{"train_loss": 0.09736062586307526, "global_step": 147139, "epoch": 1653, "lr": 4.35453047922039e-05}
{"train_loss": 0.12140394002199173, "global_step": 147140, "epoch": 1653, "lr": 4.354472990705387e-05}
{"train_loss": 0.0891137421131134, "global_step": 147141, "epoch": 1653, "lr": 4.354415502277169e-05}
{"train_loss": 0.12584228813648224, "global_step": 147142, "epoch": 1653, "lr": 4.3543580139357404e-05}
{"train_loss": 0.09022623300552368, "global_step": 147143, "epoch": 1653, "lr": 4.3543005256811106e-05}
{"train_loss": 0.11644592136144638, "global_step": 147144, "epoch": 1653, "lr": 4.354243037513288e-05}
{"train_loss": 0.11939528584480286, "global_step": 147145, "epoch": 1653, "lr": 4.354185549432279e-05}
{"train_loss": 0.11085539311170578, "global_step": 147146, "epoch": 1653, "lr": 4.3541280614380916e-05}
{"train_loss": 0.09751856327056885, "global_step": 147147, "epoch": 1653, "lr": 4.354070573530735e-05}
{"train_loss": 0.08563396334648132, "global_step": 147148, "epoch": 1653, "lr": 4.3540130857102145e-05}
{"train_loss": 0.10276040434837341, "global_step": 147149, "epoch": 1653, "lr": 4.3539555979765405e-05}
{"train_loss": 0.08807583153247833, "global_step": 147150, "epoch": 1653, "lr": 4.353898110329717e-05}
{"train_loss": 0.12927864491939545, "global_step": 147151, "epoch": 1653, "lr": 4.353840622769754e-05}
{"train_loss": 0.120437391102314, "global_step": 147152, "epoch": 1653, "lr": 4.353783135296662e-05}
{"train_loss": 0.07851453125476837, "global_step": 147153, "epoch": 1653, "lr": 4.3537256479104434e-05}
{"train_loss": 0.1655905842781067, "global_step": 147154, "epoch": 1653, "lr": 4.353668160611111e-05}
{"train_loss": 0.15180709958076477, "global_step": 147155, "epoch": 1653, "lr": 4.353610673398667e-05}
{"train_loss": 0.15227298438549042, "global_step": 147156, "epoch": 1653, "lr": 4.3535531862731254e-05}
{"train_loss": 0.09487616270780563, "global_step": 147157, "epoch": 1653, "lr": 4.353495699234488e-05}
{"train_loss": 0.12635163962841034, "global_step": 147158, "epoch": 1653, "lr": 4.353438212282768e-05}
{"train_loss": 0.11804002523422241, "global_step": 147159, "epoch": 1653, "lr": 4.353380725417967e-05}
{"train_loss": 0.04780905693769455, "global_step": 147160, "epoch": 1653, "lr": 4.3533232386400994e-05}
{"train_loss": 0.10873721539974213, "global_step": 147161, "epoch": 1653, "lr": 4.3532657519491676e-05}
{"train_loss": 0.12442158907651901, "global_step": 147162, "epoch": 1653, "lr": 4.3532082653451814e-05}
{"train_loss": 0.16609929502010345, "global_step": 147163, "epoch": 1653, "lr": 4.353150778828149e-05}
{"train_loss": 0.0647696852684021, "global_step": 147164, "epoch": 1653, "lr": 4.3530932923980776e-05}
{"train_loss": 0.13888123631477356, "global_step": 147165, "epoch": 1653, "lr": 4.353035806054975e-05}
{"train_loss": 0.1166715919971466, "global_step": 147166, "epoch": 1653, "lr": 4.3529783197988495e-05}
{"train_loss": 0.07597627490758896, "global_step": 147167, "epoch": 1653, "lr": 4.352920833629706e-05}
{"train_loss": 0.12771427631378174, "global_step": 147168, "epoch": 1653, "lr": 4.352863347547557e-05}
{"train_loss": 0.09748346358537674, "global_step": 147169, "epoch": 1653, "lr": 4.352805861552405e-05}
{"train_loss": 0.10105086117982864, "global_step": 147170, "epoch": 1653, "lr": 4.3527483756442604e-05}
{"train_loss": 0.10740779340267181, "global_step": 147171, "epoch": 1653, "lr": 4.352690889823133e-05}
{"train_loss": 0.10615886747837067, "global_step": 147172, "epoch": 1653, "lr": 4.352633404089027e-05}
{"train_loss": 0.12078965455293655, "global_step": 147173, "epoch": 1653, "lr": 4.352575918441953e-05}
{"train_loss": 0.09644642472267151, "global_step": 147174, "epoch": 1653, "lr": 4.352518432881915e-05}
{"train_loss": 0.07772579044103622, "global_step": 147175, "epoch": 1653, "lr": 4.3524609474089265e-05}
{"train_loss": 0.09319189190864563, "global_step": 147176, "epoch": 1653, "lr": 4.352403462022988e-05}
{"train_loss": 0.05915031209588051, "global_step": 147177, "epoch": 1653, "lr": 4.3523459767241135e-05}
{"train_loss": 0.1063997745513916, "global_step": 147178, "epoch": 1653, "lr": 4.3522884915123066e-05}
{"train_loss": 0.11229386180639267, "global_step": 147179, "epoch": 1653, "lr": 4.352231006387578e-05}
{"train_loss": 0.11372191458940506, "global_step": 147180, "epoch": 1653, "lr": 4.352173521349934e-05}
{"train_loss": 0.16334281861782074, "global_step": 147181, "epoch": 1653, "lr": 4.352116036399381e-05}
{"train_loss": 0.09400729835033417, "global_step": 147182, "epoch": 1653, "lr": 4.352058551535929e-05}
{"train_loss": 0.05542655661702156, "global_step": 147183, "epoch": 1653, "lr": 4.3520010667595845e-05}
{"train_loss": 0.1575494408607483, "global_step": 147184, "epoch": 1653, "lr": 4.3519435820703556e-05}
{"train_loss": 0.1519266813993454, "global_step": 147185, "epoch": 1653, "lr": 4.351886097468252e-05}
{"train_loss": 0.09737300872802734, "global_step": 147186, "epoch": 1653, "lr": 4.351828612953276e-05}
{"train_loss": 0.10168345272541046, "global_step": 147187, "epoch": 1653, "lr": 4.351771128525439e-05}
{"train_loss": 0.12826229631900787, "global_step": 147188, "epoch": 1653, "lr": 4.351713644184751e-05}
{"train_loss": 0.11689338833093643, "global_step": 147189, "epoch": 1653, "lr": 4.351656159931215e-05}
{"train_loss": 0.14338605105876923, "global_step": 147190, "epoch": 1653, "lr": 4.351598675764843e-05}
{"train_loss": 0.07424891740083694, "global_step": 147191, "epoch": 1653, "lr": 4.351541191685638e-05}
{"train_loss": 0.11888176947832108, "global_step": 147192, "epoch": 1653, "lr": 4.351483707693613e-05}
{"train_loss": 0.10742048919200897, "global_step": 147193, "epoch": 1653, "lr": 4.351426223788771e-05}
{"train_loss": 0.08141361176967621, "global_step": 147194, "epoch": 1653, "lr": 4.351368739971124e-05}
{"train_loss": 0.09188229590654373, "global_step": 147195, "epoch": 1653, "lr": 4.351311256240676e-05}
{"train_loss": 0.10298381745815277, "global_step": 147196, "epoch": 1653, "lr": 4.351253772597437e-05}
{"train_loss": 0.1213177740573883, "global_step": 147197, "epoch": 1653, "lr": 4.351196289041413e-05}
{"train_loss": 0.10870455950498581, "global_step": 147198, "epoch": 1653, "lr": 4.351138805572614e-05}
{"train_loss": 0.07485583424568176, "global_step": 147199, "epoch": 1653, "lr": 4.351081322191046e-05}
{"train_loss": 0.12934300303459167, "global_step": 147200, "epoch": 1653, "lr": 4.351023838896717e-05}
{"train_loss": 0.10841093212366104, "global_step": 147201, "epoch": 1653, "lr": 4.350966355689636e-05}
{"train_loss": 0.10270620882511139, "global_step": 147202, "epoch": 1653, "lr": 4.350908872569808e-05}
{"train_loss": 0.06014465168118477, "global_step": 147203, "epoch": 1653, "lr": 4.350851389537244e-05}
{"train_loss": 0.12113408744335175, "global_step": 147204, "epoch": 1653, "lr": 4.35079390659195e-05}
{"train_loss": 0.10679661499315433, "global_step": 147205, "epoch": 1653, "lr": 4.350736423733932e-05, "val_loss": 6.270332336425781}
{"train_loss": 0.10778693109750748, "global_step": 147206, "epoch": 1654, "lr": 4.3506789409631996e-05}
{"train_loss": 0.06558634340763092, "global_step": 147207, "epoch": 1654, "lr": 4.350621458279763e-05}
{"train_loss": 0.08373747766017914, "global_step": 147208, "epoch": 1654, "lr": 4.350563975683625e-05}
{"train_loss": 0.1529255509376526, "global_step": 147209, "epoch": 1654, "lr": 4.350506493174797e-05}
{"train_loss": 0.15695367753505707, "global_step": 147210, "epoch": 1654, "lr": 4.350449010753285e-05}
{"train_loss": 0.14791062474250793, "global_step": 147211, "epoch": 1654, "lr": 4.350391528419098e-05}
{"train_loss": 0.10687375068664551, "global_step": 147212, "epoch": 1654, "lr": 4.350334046172242e-05}
{"train_loss": 0.0765024945139885, "global_step": 147213, "epoch": 1654, "lr": 4.350276564012727e-05}
{"train_loss": 0.08333948254585266, "global_step": 147214, "epoch": 1654, "lr": 4.3502190819405576e-05}
{"train_loss": 0.08728912472724915, "global_step": 147215, "epoch": 1654, "lr": 4.350161599955745e-05}
{"train_loss": 0.10834469646215439, "global_step": 147216, "epoch": 1654, "lr": 4.350104118058293e-05}
{"train_loss": 0.08858717232942581, "global_step": 147217, "epoch": 1654, "lr": 4.350046636248215e-05}
{"train_loss": 0.06294619292020798, "global_step": 147218, "epoch": 1654, "lr": 4.349989154525513e-05}
{"train_loss": 0.11928020417690277, "global_step": 147219, "epoch": 1654, "lr": 4.3499316728901974e-05}
{"train_loss": 0.16385029256343842, "global_step": 147220, "epoch": 1654, "lr": 4.3498741913422756e-05}
{"train_loss": 0.11790208518505096, "global_step": 147221, "epoch": 1654, "lr": 4.3498167098817546e-05}
{"train_loss": 0.1488238275051117, "global_step": 147222, "epoch": 1654, "lr": 4.349759228508644e-05}
{"train_loss": 0.07789672911167145, "global_step": 147223, "epoch": 1654, "lr": 4.349701747222948e-05}
{"train_loss": 0.07331430166959763, "global_step": 147224, "epoch": 1654, "lr": 4.349644266024681e-05}
{"train_loss": 0.09921563416719437, "global_step": 147225, "epoch": 1654, "lr": 4.349586784913842e-05}
{"train_loss": 0.09301542490720749, "global_step": 147226, "epoch": 1654, "lr": 4.349529303890446e-05}
{"train_loss": 0.15861822664737701, "global_step": 147227, "epoch": 1654, "lr": 4.3494718229544965e-05}
{"train_loss": 0.10730436444282532, "global_step": 147228, "epoch": 1654, "lr": 4.3494143421060036e-05}
{"train_loss": 0.07293541729450226, "global_step": 147229, "epoch": 1654, "lr": 4.349356861344972e-05}
{"train_loss": 0.08955201506614685, "global_step": 147230, "epoch": 1654, "lr": 4.3492993806714144e-05}
{"train_loss": 0.08837830275297165, "global_step": 147231, "epoch": 1654, "lr": 4.3492419000853324e-05}
{"train_loss": 0.11870969831943512, "global_step": 147232, "epoch": 1654, "lr": 4.34918441958674e-05}
{"train_loss": 0.10746585577726364, "global_step": 147233, "epoch": 1654, "lr": 4.349126939175639e-05}
{"train_loss": 0.105106420814991, "global_step": 147234, "epoch": 1654, "lr": 4.349069458852043e-05}
{"train_loss": 0.08822301030158997, "global_step": 147235, "epoch": 1654, "lr": 4.349011978615953e-05}
{"train_loss": 0.08280109614133835, "global_step": 147236, "epoch": 1654, "lr": 4.3489544984673834e-05}
{"train_loss": 0.09922125190496445, "global_step": 147237, "epoch": 1654, "lr": 4.348897018406337e-05}
{"train_loss": 0.07319725304841995, "global_step": 147238, "epoch": 1654, "lr": 4.348839538432825e-05}
{"train_loss": 0.08115945011377335, "global_step": 147239, "epoch": 1654, "lr": 4.348782058546853e-05}
{"train_loss": 0.08496588468551636, "global_step": 147240, "epoch": 1654, "lr": 4.34872457874843e-05}
{"train_loss": 0.13309118151664734, "global_step": 147241, "epoch": 1654, "lr": 4.348667099037562e-05}
{"train_loss": 0.09121018648147583, "global_step": 147242, "epoch": 1654, "lr": 4.3486096194142566e-05}
{"train_loss": 0.16405290365219116, "global_step": 147243, "epoch": 1654, "lr": 4.348552139878526e-05}
{"train_loss": 0.0917975902557373, "global_step": 147244, "epoch": 1654, "lr": 4.348494660430371e-05}
{"train_loss": 0.07094062119722366, "global_step": 147245, "epoch": 1654, "lr": 4.348437181069806e-05}
{"train_loss": 0.1341683566570282, "global_step": 147246, "epoch": 1654, "lr": 4.348379701796833e-05}
{"train_loss": 0.11732949316501617, "global_step": 147247, "epoch": 1654, "lr": 4.348322222611465e-05}
{"train_loss": 0.12465853244066238, "global_step": 147248, "epoch": 1654, "lr": 4.3482647435137045e-05}
{"train_loss": 0.09775520116090775, "global_step": 147249, "epoch": 1654, "lr": 4.3482072645035645e-05}
{"train_loss": 0.12685906887054443, "global_step": 147250, "epoch": 1654, "lr": 4.3481497855810484e-05}
{"train_loss": 0.08211541920900345, "global_step": 147251, "epoch": 1654, "lr": 4.348092306746167e-05}
{"train_loss": 0.13903450965881348, "global_step": 147252, "epoch": 1654, "lr": 4.348034827998925e-05}
{"train_loss": 0.09919554740190506, "global_step": 147253, "epoch": 1654, "lr": 4.3479773493393334e-05}
{"train_loss": 0.12795674800872803, "global_step": 147254, "epoch": 1654, "lr": 4.347919870767396e-05}
{"train_loss": 0.1075257807970047, "global_step": 147255, "epoch": 1654, "lr": 4.347862392283125e-05}
{"train_loss": 0.07981760054826736, "global_step": 147256, "epoch": 1654, "lr": 4.3478049138865245e-05}
{"train_loss": 0.12403442710638046, "global_step": 147257, "epoch": 1654, "lr": 4.347747435577606e-05}
{"train_loss": 0.13097505271434784, "global_step": 147258, "epoch": 1654, "lr": 4.3476899573563726e-05}
{"train_loss": 0.07121092826128006, "global_step": 147259, "epoch": 1654, "lr": 4.347632479222835e-05}
{"train_loss": 0.11058995127677917, "global_step": 147260, "epoch": 1654, "lr": 4.347575001177001e-05}
{"train_loss": 0.14607083797454834, "global_step": 147261, "epoch": 1654, "lr": 4.347517523218875e-05}
{"train_loss": 0.1190662831068039, "global_step": 147262, "epoch": 1654, "lr": 4.34746004534847e-05}
{"train_loss": 0.1126575917005539, "global_step": 147263, "epoch": 1654, "lr": 4.34740256756579e-05}
{"train_loss": 0.1147717833518982, "global_step": 147264, "epoch": 1654, "lr": 4.347345089870845e-05}
{"train_loss": 0.07246711850166321, "global_step": 147265, "epoch": 1654, "lr": 4.3472876122636394e-05}
{"train_loss": 0.08795598149299622, "global_step": 147266, "epoch": 1654, "lr": 4.347230134744185e-05}
{"train_loss": 0.06619541347026825, "global_step": 147267, "epoch": 1654, "lr": 4.347172657312486e-05}
{"train_loss": 0.10415084660053253, "global_step": 147268, "epoch": 1654, "lr": 4.347115179968553e-05}
{"train_loss": 0.07079711556434631, "global_step": 147269, "epoch": 1654, "lr": 4.3470577027123906e-05}
{"train_loss": 0.07423309981822968, "global_step": 147270, "epoch": 1654, "lr": 4.3470002255440106e-05}
{"train_loss": 0.10090001672506332, "global_step": 147271, "epoch": 1654, "lr": 4.3469427484634164e-05}
{"train_loss": 0.09693581610918045, "global_step": 147272, "epoch": 1654, "lr": 4.3468852714706196e-05}
{"train_loss": 0.07738788425922394, "global_step": 147273, "epoch": 1654, "lr": 4.346827794565624e-05}
{"train_loss": 0.12050472944974899, "global_step": 147274, "epoch": 1654, "lr": 4.346770317748442e-05}
{"train_loss": 0.06764519214630127, "global_step": 147275, "epoch": 1654, "lr": 4.346712841019077e-05}
{"train_loss": 0.07239387184381485, "global_step": 147276, "epoch": 1654, "lr": 4.34665536437754e-05}
{"train_loss": 0.05764289200305939, "global_step": 147277, "epoch": 1654, "lr": 4.346597887823835e-05}
{"train_loss": 0.10593582689762115, "global_step": 147278, "epoch": 1654, "lr": 4.346540411357973e-05}
{"train_loss": 0.16105259954929352, "global_step": 147279, "epoch": 1654, "lr": 4.346482934979962e-05}
{"train_loss": 0.08189717680215836, "global_step": 147280, "epoch": 1654, "lr": 4.3464254586898054e-05}
{"train_loss": 0.10589927434921265, "global_step": 147281, "epoch": 1654, "lr": 4.346367982487517e-05}
{"train_loss": 0.11331956088542938, "global_step": 147282, "epoch": 1654, "lr": 4.346310506373099e-05}
{"train_loss": 0.08176849782466888, "global_step": 147283, "epoch": 1654, "lr": 4.346253030346564e-05}
{"train_loss": 0.08487515896558762, "global_step": 147284, "epoch": 1654, "lr": 4.3461955544079154e-05}
{"train_loss": 0.13122987747192383, "global_step": 147285, "epoch": 1654, "lr": 4.346138078557165e-05}
{"train_loss": 0.06116883084177971, "global_step": 147286, "epoch": 1654, "lr": 4.3460806027943155e-05}
{"train_loss": 0.08905404061079025, "global_step": 147287, "epoch": 1654, "lr": 4.34602312711938e-05}
{"train_loss": 0.09597137570381165, "global_step": 147288, "epoch": 1654, "lr": 4.3459656515323616e-05}
{"train_loss": 0.12826380133628845, "global_step": 147289, "epoch": 1654, "lr": 4.345908176033272e-05}
{"train_loss": 0.1303728222846985, "global_step": 147290, "epoch": 1654, "lr": 4.345850700622115e-05}
{"train_loss": 0.09010977298021317, "global_step": 147291, "epoch": 1654, "lr": 4.345793225298903e-05}
{"train_loss": 0.09245166927576065, "global_step": 147292, "epoch": 1654, "lr": 4.345735750063639e-05}
{"train_loss": 0.09455470740795135, "global_step": 147293, "epoch": 1654, "lr": 4.345678274916334e-05}
{"train_loss": 0.1023824555234293, "global_step": 147294, "epoch": 1654, "lr": 4.345620799856994e-05, "val_loss": 6.310296058654785}
{"train_loss": 0.10932374745607376, "global_step": 147295, "epoch": 1655, "lr": 4.345563324885628e-05}
{"train_loss": 0.08310266584157944, "global_step": 147296, "epoch": 1655, "lr": 4.345505850002243e-05}
{"train_loss": 0.06288857758045197, "global_step": 147297, "epoch": 1655, "lr": 4.345448375206845e-05}
{"train_loss": 0.08703751116991043, "global_step": 147298, "epoch": 1655, "lr": 4.345390900499446e-05}
{"train_loss": 0.09994079917669296, "global_step": 147299, "epoch": 1655, "lr": 4.345333425880049e-05}
{"train_loss": 0.10761836916208267, "global_step": 147300, "epoch": 1655, "lr": 4.345275951348666e-05}
{"train_loss": 0.09626257419586182, "global_step": 147301, "epoch": 1655, "lr": 4.3452184769053e-05}
{"train_loss": 0.08917295932769775, "global_step": 147302, "epoch": 1655, "lr": 4.3451610025499646e-05}
{"train_loss": 0.1019270196557045, "global_step": 147303, "epoch": 1655, "lr": 4.3451035282826615e-05}
{"train_loss": 0.08432483673095703, "global_step": 147304, "epoch": 1655, "lr": 4.345046054103403e-05}
{"train_loss": 0.0583566352725029, "global_step": 147305, "epoch": 1655, "lr": 4.3449885800121936e-05}
{"train_loss": 0.03904831409454346, "global_step": 147306, "epoch": 1655, "lr": 4.344931106009045e-05}
{"train_loss": 0.09977619349956512, "global_step": 147307, "epoch": 1655, "lr": 4.34487363209396e-05}
{"train_loss": 0.13394904136657715, "global_step": 147308, "epoch": 1655, "lr": 4.344816158266951e-05}
{"train_loss": 0.08787435293197632, "global_step": 147309, "epoch": 1655, "lr": 4.344758684528021e-05}
{"train_loss": 0.09272991865873337, "global_step": 147310, "epoch": 1655, "lr": 4.3447012108771816e-05}
{"train_loss": 0.1284271627664566, "global_step": 147311, "epoch": 1655, "lr": 4.3446437373144386e-05}
{"train_loss": 0.09471357613801956, "global_step": 147312, "epoch": 1655, "lr": 4.344586263839801e-05}
{"train_loss": 0.07071305811405182, "global_step": 147313, "epoch": 1655, "lr": 4.344528790453274e-05}
{"train_loss": 0.1138150617480278, "global_step": 147314, "epoch": 1655, "lr": 4.344471317154869e-05}
{"train_loss": 0.09242892265319824, "global_step": 147315, "epoch": 1655, "lr": 4.344413843944593e-05}
{"train_loss": 0.18202705681324005, "global_step": 147316, "epoch": 1655, "lr": 4.3443563708224485e-05}
{"train_loss": 0.11532048135995865, "global_step": 147317, "epoch": 1655, "lr": 4.344298897788451e-05}
{"train_loss": 0.09140804409980774, "global_step": 147318, "epoch": 1655, "lr": 4.344241424842602e-05}
{"train_loss": 0.0775868371129036, "global_step": 147319, "epoch": 1655, "lr": 4.344183951984914e-05}
{"train_loss": 0.1607617437839508, "global_step": 147320, "epoch": 1655, "lr": 4.344126479215391e-05}
{"train_loss": 0.11084862053394318, "global_step": 147321, "epoch": 1655, "lr": 4.344069006534043e-05}
{"train_loss": 0.16215550899505615, "global_step": 147322, "epoch": 1655, "lr": 4.344011533940876e-05}
{"train_loss": 0.05228733643889427, "global_step": 147323, "epoch": 1655, "lr": 4.3439540614359e-05}
{"train_loss": 0.11324943602085114, "global_step": 147324, "epoch": 1655, "lr": 4.34389658901912e-05}
{"train_loss": 0.05784039944410324, "global_step": 147325, "epoch": 1655, "lr": 4.343839116690546e-05}
{"train_loss": 0.1257513463497162, "global_step": 147326, "epoch": 1655, "lr": 4.343781644450183e-05}
{"train_loss": 0.09500207006931305, "global_step": 147327, "epoch": 1655, "lr": 4.3437241722980436e-05}
{"train_loss": 0.08723825961351395, "global_step": 147328, "epoch": 1655, "lr": 4.34366670023413e-05}
{"train_loss": 0.13940276205539703, "global_step": 147329, "epoch": 1655, "lr": 4.3436092282584544e-05}
{"train_loss": 0.052927277982234955, "global_step": 147330, "epoch": 1655, "lr": 4.3435517563710206e-05}
{"train_loss": 0.1376107931137085, "global_step": 147331, "epoch": 1655, "lr": 4.34349428457184e-05}
{"train_loss": 0.08669660240411758, "global_step": 147332, "epoch": 1655, "lr": 4.3434368128609164e-05}
{"train_loss": 0.15494483709335327, "global_step": 147333, "epoch": 1655, "lr": 4.343379341238261e-05}
{"train_loss": 0.10069675743579865, "global_step": 147334, "epoch": 1655, "lr": 4.343321869703881e-05}
{"train_loss": 0.07666605710983276, "global_step": 147335, "epoch": 1655, "lr": 4.343264398257783e-05}
{"train_loss": 0.12958893179893494, "global_step": 147336, "epoch": 1655, "lr": 4.343206926899975e-05}
{"train_loss": 0.15819118916988373, "global_step": 147337, "epoch": 1655, "lr": 4.343149455630464e-05}
{"train_loss": 0.1380590796470642, "global_step": 147338, "epoch": 1655, "lr": 4.3430919844492604e-05}
{"train_loss": 0.09310512989759445, "global_step": 147339, "epoch": 1655, "lr": 4.343034513356367e-05}
{"train_loss": 0.11585689336061478, "global_step": 147340, "epoch": 1655, "lr": 4.342977042351798e-05}
{"train_loss": 0.1424493044614792, "global_step": 147341, "epoch": 1655, "lr": 4.342919571435555e-05}
{"train_loss": 0.14107085764408112, "global_step": 147342, "epoch": 1655, "lr": 4.342862100607651e-05}
{"train_loss": 0.11955533176660538, "global_step": 147343, "epoch": 1655, "lr": 4.342804629868088e-05}
{"train_loss": 0.050399377942085266, "global_step": 147344, "epoch": 1655, "lr": 4.34274715921688e-05}
{"train_loss": 0.11294776201248169, "global_step": 147345, "epoch": 1655, "lr": 4.342689688654029e-05}
{"train_loss": 0.1008613258600235, "global_step": 147346, "epoch": 1655, "lr": 4.3426322181795485e-05}
{"train_loss": 0.08554970473051071, "global_step": 147347, "epoch": 1655, "lr": 4.34257474779344e-05}
{"train_loss": 0.10731764882802963, "global_step": 147348, "epoch": 1655, "lr": 4.342517277495717e-05}
{"train_loss": 0.09184867143630981, "global_step": 147349, "epoch": 1655, "lr": 4.342459807286381e-05}
{"train_loss": 0.06909473985433578, "global_step": 147350, "epoch": 1655, "lr": 4.342402337165446e-05}
{"train_loss": 0.07824094593524933, "global_step": 147351, "epoch": 1655, "lr": 4.342344867132916e-05}
{"train_loss": 0.18150648474693298, "global_step": 147352, "epoch": 1655, "lr": 4.3422873971888e-05}
{"train_loss": 0.11143288016319275, "global_step": 147353, "epoch": 1655, "lr": 4.342229927333106e-05}
{"train_loss": 0.1598365306854248, "global_step": 147354, "epoch": 1655, "lr": 4.342172457565841e-05}
{"train_loss": 0.09508054703474045, "global_step": 147355, "epoch": 1655, "lr": 4.3421149878870135e-05}
{"train_loss": 0.08298001438379288, "global_step": 147356, "epoch": 1655, "lr": 4.3420575182966286e-05}
{"train_loss": 0.0875154584646225, "global_step": 147357, "epoch": 1655, "lr": 4.3420000487946984e-05}
{"train_loss": 0.17268149554729462, "global_step": 147358, "epoch": 1655, "lr": 4.3419425793812256e-05}
{"train_loss": 0.12114676833152771, "global_step": 147359, "epoch": 1655, "lr": 4.341885110056223e-05}
{"train_loss": 0.12387330830097198, "global_step": 147360, "epoch": 1655, "lr": 4.341827640819694e-05}
{"train_loss": 0.09540014714002609, "global_step": 147361, "epoch": 1655, "lr": 4.341770171671651e-05}
{"train_loss": 0.0924421176314354, "global_step": 147362, "epoch": 1655, "lr": 4.341712702612096e-05}
{"train_loss": 0.06758017838001251, "global_step": 147363, "epoch": 1655, "lr": 4.341655233641041e-05}
{"train_loss": 0.14042812585830688, "global_step": 147364, "epoch": 1655, "lr": 4.3415977647584914e-05}
{"train_loss": 0.12183189392089844, "global_step": 147365, "epoch": 1655, "lr": 4.3415402959644585e-05}
{"train_loss": 0.12189566344022751, "global_step": 147366, "epoch": 1655, "lr": 4.3414828272589446e-05}
{"train_loss": 0.06384506076574326, "global_step": 147367, "epoch": 1655, "lr": 4.3414253586419626e-05}
{"train_loss": 0.1461315155029297, "global_step": 147368, "epoch": 1655, "lr": 4.341367890113516e-05}
{"train_loss": 0.06617659330368042, "global_step": 147369, "epoch": 1655, "lr": 4.341310421673616e-05}
{"train_loss": 0.16277942061424255, "global_step": 147370, "epoch": 1655, "lr": 4.341252953322268e-05}
{"train_loss": 0.19482634961605072, "global_step": 147371, "epoch": 1655, "lr": 4.341195485059481e-05}
{"train_loss": 0.08557422459125519, "global_step": 147372, "epoch": 1655, "lr": 4.341138016885261e-05}
{"train_loss": 0.10650990158319473, "global_step": 147373, "epoch": 1655, "lr": 4.341080548799618e-05}
{"train_loss": 0.08104551583528519, "global_step": 147374, "epoch": 1655, "lr": 4.34102308080256e-05}
{"train_loss": 0.07056375592947006, "global_step": 147375, "epoch": 1655, "lr": 4.3409656128940903e-05}
{"train_loss": 0.06337413191795349, "global_step": 147376, "epoch": 1655, "lr": 4.3409081450742226e-05}
{"train_loss": 0.11137578636407852, "global_step": 147377, "epoch": 1655, "lr": 4.34085067734296e-05}
{"train_loss": 0.11012202501296997, "global_step": 147378, "epoch": 1655, "lr": 4.340793209700314e-05}
{"train_loss": 0.2219686359167099, "global_step": 147379, "epoch": 1655, "lr": 4.340735742146287e-05}
{"train_loss": 0.08485731482505798, "global_step": 147380, "epoch": 1655, "lr": 4.3406782746808936e-05}
{"train_loss": 0.06280232220888138, "global_step": 147381, "epoch": 1655, "lr": 4.340620807304135e-05}
{"train_loss": 0.08407554030418396, "global_step": 147382, "epoch": 1655, "lr": 4.340563340016024e-05}
{"train_loss": 0.10697091935893123, "global_step": 147383, "epoch": 1655, "lr": 4.340505872816564e-05, "val_loss": 6.41335916519165, "train_action_mse_error": 27.315383911132812}
{"train_loss": 0.058728668838739395, "global_step": 147384, "epoch": 1656, "lr": 4.340448405705768e-05}
{"train_loss": 0.19253413379192352, "global_step": 147385, "epoch": 1656, "lr": 4.340390938683638e-05}
{"train_loss": 0.09397564083337784, "global_step": 147386, "epoch": 1656, "lr": 4.340333471750186e-05}
{"train_loss": 0.11684832721948624, "global_step": 147387, "epoch": 1656, "lr": 4.340276004905417e-05}
{"train_loss": 0.09674894064664841, "global_step": 147388, "epoch": 1656, "lr": 4.34021853814934e-05}
{"train_loss": 0.11617649346590042, "global_step": 147389, "epoch": 1656, "lr": 4.340161071481964e-05}
{"train_loss": 0.2208075076341629, "global_step": 147390, "epoch": 1656, "lr": 4.340103604903294e-05}
{"train_loss": 0.09720437973737717, "global_step": 147391, "epoch": 1656, "lr": 4.340046138413339e-05}
{"train_loss": 0.15672501921653748, "global_step": 147392, "epoch": 1656, "lr": 4.339988672012107e-05}
{"train_loss": 0.16588526964187622, "global_step": 147393, "epoch": 1656, "lr": 4.3399312056996064e-05}
{"train_loss": 0.1346169114112854, "global_step": 147394, "epoch": 1656, "lr": 4.339873739475843e-05}
{"train_loss": 0.1887696534395218, "global_step": 147395, "epoch": 1656, "lr": 4.3398162733408265e-05}
{"train_loss": 0.10626651346683502, "global_step": 147396, "epoch": 1656, "lr": 4.339758807294562e-05}
{"train_loss": 0.11329135298728943, "global_step": 147397, "epoch": 1656, "lr": 4.3397013413370605e-05}
{"train_loss": 0.10918489098548889, "global_step": 147398, "epoch": 1656, "lr": 4.339643875468326e-05}
{"train_loss": 0.14264711737632751, "global_step": 147399, "epoch": 1656, "lr": 4.339586409688371e-05}
{"train_loss": 0.13786669075489044, "global_step": 147400, "epoch": 1656, "lr": 4.3395289439971984e-05}
{"train_loss": 0.1314551681280136, "global_step": 147401, "epoch": 1656, "lr": 4.33947147839482e-05}
{"train_loss": 0.156203031539917, "global_step": 147402, "epoch": 1656, "lr": 4.339414012881239e-05}
{"train_loss": 0.09939184784889221, "global_step": 147403, "epoch": 1656, "lr": 4.3393565474564684e-05}
{"train_loss": 0.09895095229148865, "global_step": 147404, "epoch": 1656, "lr": 4.339299082120511e-05}
{"train_loss": 0.11966883391141891, "global_step": 147405, "epoch": 1656, "lr": 4.3392416168733766e-05}
{"train_loss": 0.05554836615920067, "global_step": 147406, "epoch": 1656, "lr": 4.3391841517150754e-05}
{"train_loss": 0.06361495703458786, "global_step": 147407, "epoch": 1656, "lr": 4.3391266866456124e-05}
{"train_loss": 0.11721151322126389, "global_step": 147408, "epoch": 1656, "lr": 4.339069221664995e-05}
{"train_loss": 0.10288278758525848, "global_step": 147409, "epoch": 1656, "lr": 4.339011756773232e-05}
{"train_loss": 0.17172668874263763, "global_step": 147410, "epoch": 1656, "lr": 4.338954291970331e-05}
{"train_loss": 0.1041775792837143, "global_step": 147411, "epoch": 1656, "lr": 4.338896827256299e-05}
{"train_loss": 0.07242037355899811, "global_step": 147412, "epoch": 1656, "lr": 4.338839362631145e-05}
{"train_loss": 0.17724820971488953, "global_step": 147413, "epoch": 1656, "lr": 4.338781898094875e-05}
{"train_loss": 0.15298204123973846, "global_step": 147414, "epoch": 1656, "lr": 4.338724433647499e-05}
{"train_loss": 0.07864547520875931, "global_step": 147415, "epoch": 1656, "lr": 4.3386669692890214e-05}
{"train_loss": 0.1031036376953125, "global_step": 147416, "epoch": 1656, "lr": 4.338609505019455e-05}
{"train_loss": 0.11906782537698746, "global_step": 147417, "epoch": 1656, "lr": 4.338552040838801e-05}
{"train_loss": 0.09346728771924973, "global_step": 147418, "epoch": 1656, "lr": 4.338494576747074e-05}
{"train_loss": 0.1893819123506546, "global_step": 147419, "epoch": 1656, "lr": 4.338437112744275e-05}
{"train_loss": 0.19651666283607483, "global_step": 147420, "epoch": 1656, "lr": 4.338379648830418e-05}
{"train_loss": 0.09222384542226791, "global_step": 147421, "epoch": 1656, "lr": 4.338322185005506e-05}
{"train_loss": 0.10542912036180496, "global_step": 147422, "epoch": 1656, "lr": 4.338264721269549e-05}
{"train_loss": 0.1099194809794426, "global_step": 147423, "epoch": 1656, "lr": 4.338207257622554e-05}
{"train_loss": 0.11720456928014755, "global_step": 147424, "epoch": 1656, "lr": 4.338149794064528e-05}
{"train_loss": 0.11978934705257416, "global_step": 147425, "epoch": 1656, "lr": 4.3380923305954826e-05}
{"train_loss": 0.08520567417144775, "global_step": 147426, "epoch": 1656, "lr": 4.33803486721542e-05}
{"train_loss": 0.07347945123910904, "global_step": 147427, "epoch": 1656, "lr": 4.337977403924352e-05}
{"train_loss": 0.1177213117480278, "global_step": 147428, "epoch": 1656, "lr": 4.337919940722284e-05}
{"train_loss": 0.1299450695514679, "global_step": 147429, "epoch": 1656, "lr": 4.3378624776092255e-05}
{"train_loss": 0.11252205818891525, "global_step": 147430, "epoch": 1656, "lr": 4.3378050145851823e-05}
{"train_loss": 0.16888415813446045, "global_step": 147431, "epoch": 1656, "lr": 4.337747551650165e-05}
{"train_loss": 0.15924029052257538, "global_step": 147432, "epoch": 1656, "lr": 4.337690088804177e-05}
{"train_loss": 0.0774381011724472, "global_step": 147433, "epoch": 1656, "lr": 4.337632626047231e-05}
{"train_loss": 0.16221335530281067, "global_step": 147434, "epoch": 1656, "lr": 4.3375751633793296e-05}
{"train_loss": 0.11216308176517487, "global_step": 147435, "epoch": 1656, "lr": 4.3375177008004856e-05}
{"train_loss": 0.08592120558023453, "global_step": 147436, "epoch": 1656, "lr": 4.3374602383107014e-05}
{"train_loss": 0.16028647124767303, "global_step": 147437, "epoch": 1656, "lr": 4.3374027759099905e-05}
{"train_loss": 0.09216052293777466, "global_step": 147438, "epoch": 1656, "lr": 4.3373453135983556e-05}
{"train_loss": 0.05410684272646904, "global_step": 147439, "epoch": 1656, "lr": 4.337287851375809e-05}
{"train_loss": 0.1604251265525818, "global_step": 147440, "epoch": 1656, "lr": 4.3372303892423525e-05}
{"train_loss": 0.11214990168809891, "global_step": 147441, "epoch": 1656, "lr": 4.3371729271979985e-05}
{"train_loss": 0.09294643998146057, "global_step": 147442, "epoch": 1656, "lr": 4.337115465242755e-05}
{"train_loss": 0.12460768222808838, "global_step": 147443, "epoch": 1656, "lr": 4.3370580033766263e-05}
{"train_loss": 0.16133232414722443, "global_step": 147444, "epoch": 1656, "lr": 4.3370005415996244e-05}
{"train_loss": 0.20800532400608063, "global_step": 147445, "epoch": 1656, "lr": 4.336943079911754e-05}
{"train_loss": 0.10074248909950256, "global_step": 147446, "epoch": 1656, "lr": 4.336885618313022e-05}
{"train_loss": 0.12253634631633759, "global_step": 147447, "epoch": 1656, "lr": 4.336828156803439e-05}
{"train_loss": 0.06836682558059692, "global_step": 147448, "epoch": 1656, "lr": 4.336770695383011e-05}
{"train_loss": 0.16443386673927307, "global_step": 147449, "epoch": 1656, "lr": 4.3367132340517465e-05}
{"train_loss": 0.13117548823356628, "global_step": 147450, "epoch": 1656, "lr": 4.3366557728096526e-05}
{"train_loss": 0.17204634845256805, "global_step": 147451, "epoch": 1656, "lr": 4.3365983116567364e-05}
{"train_loss": 0.0993434265255928, "global_step": 147452, "epoch": 1656, "lr": 4.336540850593009e-05}
{"train_loss": 0.07839282602071762, "global_step": 147453, "epoch": 1656, "lr": 4.336483389618472e-05}
{"train_loss": 0.13311323523521423, "global_step": 147454, "epoch": 1656, "lr": 4.33642592873314e-05}
{"train_loss": 0.0859115943312645, "global_step": 147455, "epoch": 1656, "lr": 4.336368467937015e-05}
{"train_loss": 0.10696104913949966, "global_step": 147456, "epoch": 1656, "lr": 4.336311007230109e-05}
{"train_loss": 0.11441431939601898, "global_step": 147457, "epoch": 1656, "lr": 4.336253546612426e-05}
{"train_loss": 0.13570000231266022, "global_step": 147458, "epoch": 1656, "lr": 4.336196086083978e-05}
{"train_loss": 0.09167389571666718, "global_step": 147459, "epoch": 1656, "lr": 4.336138625644768e-05}
{"train_loss": 0.14417114853858948, "global_step": 147460, "epoch": 1656, "lr": 4.3360811652948056e-05}
{"train_loss": 0.0873325914144516, "global_step": 147461, "epoch": 1656, "lr": 4.336023705034101e-05}
{"train_loss": 0.07328645139932632, "global_step": 147462, "epoch": 1656, "lr": 4.335966244862659e-05}
{"train_loss": 0.15400326251983643, "global_step": 147463, "epoch": 1656, "lr": 4.33590878478049e-05}
{"train_loss": 0.14168721437454224, "global_step": 147464, "epoch": 1656, "lr": 4.335851324787597e-05}
{"train_loss": 0.15890498459339142, "global_step": 147465, "epoch": 1656, "lr": 4.3357938648839934e-05}
{"train_loss": 0.16247603297233582, "global_step": 147466, "epoch": 1656, "lr": 4.3357364050696834e-05}
{"train_loss": 0.12557561695575714, "global_step": 147467, "epoch": 1656, "lr": 4.335678945344675e-05}
{"train_loss": 0.1106153056025505, "global_step": 147468, "epoch": 1656, "lr": 4.3356214857089764e-05}
{"train_loss": 0.1374950408935547, "global_step": 147469, "epoch": 1656, "lr": 4.3355640261625965e-05}
{"train_loss": 0.080185666680336, "global_step": 147470, "epoch": 1656, "lr": 4.3355065667055406e-05}
{"train_loss": 0.13792216777801514, "global_step": 147471, "epoch": 1656, "lr": 4.3354491073378195e-05}
{"train_loss": 0.12155561211906123, "global_step": 147472, "epoch": 1656, "lr": 4.335391648059437e-05, "val_loss": 6.047389030456543}
{"train_loss": 0.12411129474639893, "global_step": 147473, "epoch": 1657, "lr": 4.3353341888704044e-05}
{"train_loss": 0.1376684606075287, "global_step": 147474, "epoch": 1657, "lr": 4.3352767297707265e-05}
{"train_loss": 0.11032238602638245, "global_step": 147475, "epoch": 1657, "lr": 4.335219270760415e-05}
{"train_loss": 0.0871591567993164, "global_step": 147476, "epoch": 1657, "lr": 4.3351618118394724e-05}
{"train_loss": 0.12495088577270508, "global_step": 147477, "epoch": 1657, "lr": 4.3351043530079097e-05}
{"train_loss": 0.11662641167640686, "global_step": 147478, "epoch": 1657, "lr": 4.335046894265736e-05}
{"train_loss": 0.12247761338949203, "global_step": 147479, "epoch": 1657, "lr": 4.334989435612955e-05}
{"train_loss": 0.10876430571079254, "global_step": 147480, "epoch": 1657, "lr": 4.334931977049579e-05}
{"train_loss": 0.05871793255209923, "global_step": 147481, "epoch": 1657, "lr": 4.3348745185756104e-05}
{"train_loss": 0.09718764573335648, "global_step": 147482, "epoch": 1657, "lr": 4.334817060191063e-05}
{"train_loss": 0.15386894345283508, "global_step": 147483, "epoch": 1657, "lr": 4.334759601895938e-05}
{"train_loss": 0.14150555431842804, "global_step": 147484, "epoch": 1657, "lr": 4.3347021436902495e-05}
{"train_loss": 0.1885230988264084, "global_step": 147485, "epoch": 1657, "lr": 4.334644685574001e-05}
{"train_loss": 0.09802208095788956, "global_step": 147486, "epoch": 1657, "lr": 4.334587227547201e-05}
{"train_loss": 0.13370786607265472, "global_step": 147487, "epoch": 1657, "lr": 4.334529769609858e-05}
{"train_loss": 0.05665537714958191, "global_step": 147488, "epoch": 1657, "lr": 4.334472311761979e-05}
{"train_loss": 0.12298949807882309, "global_step": 147489, "epoch": 1657, "lr": 4.334414854003573e-05}
{"train_loss": 0.08901267498731613, "global_step": 147490, "epoch": 1657, "lr": 4.334357396334647e-05}
{"train_loss": 0.1065436378121376, "global_step": 147491, "epoch": 1657, "lr": 4.334299938755206e-05}
{"train_loss": 0.08409484475851059, "global_step": 147492, "epoch": 1657, "lr": 4.334242481265264e-05}
{"train_loss": 0.20929567515850067, "global_step": 147493, "epoch": 1657, "lr": 4.3341850238648215e-05}
{"train_loss": 0.07710573822259903, "global_step": 147494, "epoch": 1657, "lr": 4.334127566553892e-05}
{"train_loss": 0.13707473874092102, "global_step": 147495, "epoch": 1657, "lr": 4.3340701093324795e-05}
{"train_loss": 0.1353898048400879, "global_step": 147496, "epoch": 1657, "lr": 4.334012652200593e-05}
{"train_loss": 0.06449002772569656, "global_step": 147497, "epoch": 1657, "lr": 4.3339551951582414e-05}
{"train_loss": 0.12305107712745667, "global_step": 147498, "epoch": 1657, "lr": 4.33389773820543e-05}
{"train_loss": 0.1002313420176506, "global_step": 147499, "epoch": 1657, "lr": 4.33384028134217e-05}
{"train_loss": 0.10455938428640366, "global_step": 147500, "epoch": 1657, "lr": 4.333782824568465e-05}
{"train_loss": 0.11618867516517639, "global_step": 147501, "epoch": 1657, "lr": 4.333725367884327e-05}
{"train_loss": 0.05888696387410164, "global_step": 147502, "epoch": 1657, "lr": 4.33366791128976e-05}
{"train_loss": 0.15636992454528809, "global_step": 147503, "epoch": 1657, "lr": 4.333610454784774e-05}
{"train_loss": 0.09420772641897202, "global_step": 147504, "epoch": 1657, "lr": 4.3335529983693756e-05}
{"train_loss": 0.13090549409389496, "global_step": 147505, "epoch": 1657, "lr": 4.333495542043573e-05}
{"train_loss": 0.07724370062351227, "global_step": 147506, "epoch": 1657, "lr": 4.333438085807374e-05}
{"train_loss": 0.06075894087553024, "global_step": 147507, "epoch": 1657, "lr": 4.3333806296607856e-05}
{"train_loss": 0.07804454118013382, "global_step": 147508, "epoch": 1657, "lr": 4.333323173603816e-05}
{"train_loss": 0.07764280587434769, "global_step": 147509, "epoch": 1657, "lr": 4.333265717636474e-05}
{"train_loss": 0.12359927594661713, "global_step": 147510, "epoch": 1657, "lr": 4.333208261758765e-05}
{"train_loss": 0.09193681925535202, "global_step": 147511, "epoch": 1657, "lr": 4.333150805970699e-05}
{"train_loss": 0.08189760148525238, "global_step": 147512, "epoch": 1657, "lr": 4.3330933502722815e-05}
{"train_loss": 0.09549518674612045, "global_step": 147513, "epoch": 1657, "lr": 4.333035894663523e-05}
{"train_loss": 0.0858958438038826, "global_step": 147514, "epoch": 1657, "lr": 4.3329784391444285e-05}
{"train_loss": 0.09278733283281326, "global_step": 147515, "epoch": 1657, "lr": 4.332920983715006e-05}
{"train_loss": 0.0936850905418396, "global_step": 147516, "epoch": 1657, "lr": 4.332863528375267e-05}
{"train_loss": 0.0913124680519104, "global_step": 147517, "epoch": 1657, "lr": 4.3328060731252136e-05}
{"train_loss": 0.09628190100193024, "global_step": 147518, "epoch": 1657, "lr": 4.332748617964858e-05}
{"train_loss": 0.12200956791639328, "global_step": 147519, "epoch": 1657, "lr": 4.332691162894205e-05}
{"train_loss": 0.12143836915493011, "global_step": 147520, "epoch": 1657, "lr": 4.332633707913266e-05}
{"train_loss": 0.07686762511730194, "global_step": 147521, "epoch": 1657, "lr": 4.332576253022043e-05}
{"train_loss": 0.15267179906368256, "global_step": 147522, "epoch": 1657, "lr": 4.3325187982205495e-05}
{"train_loss": 0.18523745238780975, "global_step": 147523, "epoch": 1657, "lr": 4.33246134350879e-05}
{"train_loss": 0.09228183329105377, "global_step": 147524, "epoch": 1657, "lr": 4.3324038888867726e-05}
{"train_loss": 0.2788615822792053, "global_step": 147525, "epoch": 1657, "lr": 4.332346434354506e-05}
{"train_loss": 0.06259768456220627, "global_step": 147526, "epoch": 1657, "lr": 4.332288979911997e-05}
{"train_loss": 0.09223660826683044, "global_step": 147527, "epoch": 1657, "lr": 4.332231525559253e-05}
{"train_loss": 0.174479678273201, "global_step": 147528, "epoch": 1657, "lr": 4.3321740712962836e-05}
{"train_loss": 0.12606185674667358, "global_step": 147529, "epoch": 1657, "lr": 4.332116617123093e-05}
{"train_loss": 0.07874076068401337, "global_step": 147530, "epoch": 1657, "lr": 4.332059163039694e-05}
{"train_loss": 0.06935936212539673, "global_step": 147531, "epoch": 1657, "lr": 4.332001709046089e-05}
{"train_loss": 0.05910859629511833, "global_step": 147532, "epoch": 1657, "lr": 4.331944255142288e-05}
{"train_loss": 0.054792098701000214, "global_step": 147533, "epoch": 1657, "lr": 4.331886801328301e-05}
{"train_loss": 0.10515252500772476, "global_step": 147534, "epoch": 1657, "lr": 4.331829347604132e-05}
{"train_loss": 0.10459034144878387, "global_step": 147535, "epoch": 1657, "lr": 4.331771893969793e-05}
{"train_loss": 0.07991965115070343, "global_step": 147536, "epoch": 1657, "lr": 4.331714440425286e-05}
{"train_loss": 0.12778396904468536, "global_step": 147537, "epoch": 1657, "lr": 4.3316569869706244e-05}
{"train_loss": 0.09740203619003296, "global_step": 147538, "epoch": 1657, "lr": 4.331599533605811e-05}
{"train_loss": 0.07619377225637436, "global_step": 147539, "epoch": 1657, "lr": 4.331542080330858e-05}
{"train_loss": 0.09581848978996277, "global_step": 147540, "epoch": 1657, "lr": 4.3314846271457695e-05}
{"train_loss": 0.10968251526355743, "global_step": 147541, "epoch": 1657, "lr": 4.3314271740505564e-05}
{"train_loss": 0.0783112645149231, "global_step": 147542, "epoch": 1657, "lr": 4.331369721045224e-05}
{"train_loss": 0.12321121245622635, "global_step": 147543, "epoch": 1657, "lr": 4.33131226812978e-05}
{"train_loss": 0.11834415048360825, "global_step": 147544, "epoch": 1657, "lr": 4.331254815304234e-05}
{"train_loss": 0.11868759244680405, "global_step": 147545, "epoch": 1657, "lr": 4.331197362568592e-05}
{"train_loss": 0.05882716923952103, "global_step": 147546, "epoch": 1657, "lr": 4.331139909922862e-05}
{"train_loss": 0.10582345724105835, "global_step": 147547, "epoch": 1657, "lr": 4.331082457367054e-05}
{"train_loss": 0.14365196228027344, "global_step": 147548, "epoch": 1657, "lr": 4.331025004901171e-05}
{"train_loss": 0.07832974940538406, "global_step": 147549, "epoch": 1657, "lr": 4.330967552525226e-05}
{"train_loss": 0.11811121553182602, "global_step": 147550, "epoch": 1657, "lr": 4.330910100239222e-05}
{"train_loss": 0.10352928936481476, "global_step": 147551, "epoch": 1657, "lr": 4.330852648043169e-05}
{"train_loss": 0.08930355310440063, "global_step": 147552, "epoch": 1657, "lr": 4.330795195937077e-05}
{"train_loss": 0.08778303116559982, "global_step": 147553, "epoch": 1657, "lr": 4.3307377439209496e-05}
{"train_loss": 0.08141909539699554, "global_step": 147554, "epoch": 1657, "lr": 4.330680291994798e-05}
{"train_loss": 0.13250398635864258, "global_step": 147555, "epoch": 1657, "lr": 4.3306228401586254e-05}
{"train_loss": 0.1235203668475151, "global_step": 147556, "epoch": 1657, "lr": 4.3305653884124465e-05}
{"train_loss": 0.10390321910381317, "global_step": 147557, "epoch": 1657, "lr": 4.330507936756261e-05}
{"train_loss": 0.09589242935180664, "global_step": 147558, "epoch": 1657, "lr": 4.3304504851900835e-05}
{"train_loss": 0.152665913105011, "global_step": 147559, "epoch": 1657, "lr": 4.330393033713916e-05}
{"train_loss": 0.12165935337543488, "global_step": 147560, "epoch": 1657, "lr": 4.3303355823277725e-05}
{"train_loss": 0.10705122678132539, "global_step": 147561, "epoch": 1657, "lr": 4.330278131031655e-05, "val_loss": 6.2735795974731445}
{"train_loss": 0.08145572245121002, "global_step": 147562, "epoch": 1658, "lr": 4.330220679825574e-05}
{"train_loss": 0.10664454847574234, "global_step": 147563, "epoch": 1658, "lr": 4.330163228709536e-05}
{"train_loss": 0.1596975177526474, "global_step": 147564, "epoch": 1658, "lr": 4.33010577768355e-05}
{"train_loss": 0.1021587923169136, "global_step": 147565, "epoch": 1658, "lr": 4.330048326747623e-05}
{"train_loss": 0.10898436605930328, "global_step": 147566, "epoch": 1658, "lr": 4.329990875901763e-05}
{"train_loss": 0.09893397241830826, "global_step": 147567, "epoch": 1658, "lr": 4.329933425145977e-05}
{"train_loss": 0.09711463004350662, "global_step": 147568, "epoch": 1658, "lr": 4.329875974480272e-05}
{"train_loss": 0.10431963205337524, "global_step": 147569, "epoch": 1658, "lr": 4.329818523904659e-05}
{"train_loss": 0.15528135001659393, "global_step": 147570, "epoch": 1658, "lr": 4.329761073419143e-05}
{"train_loss": 0.08142156898975372, "global_step": 147571, "epoch": 1658, "lr": 4.3297036230237346e-05}
{"train_loss": 0.13475622236728668, "global_step": 147572, "epoch": 1658, "lr": 4.329646172718436e-05}
{"train_loss": 0.09351007640361786, "global_step": 147573, "epoch": 1658, "lr": 4.32958872250326e-05}
{"train_loss": 0.08471180498600006, "global_step": 147574, "epoch": 1658, "lr": 4.329531272378212e-05}
{"train_loss": 0.2087569534778595, "global_step": 147575, "epoch": 1658, "lr": 4.3294738223433015e-05}
{"train_loss": 0.10841508954763412, "global_step": 147576, "epoch": 1658, "lr": 4.329416372398534e-05}
{"train_loss": 0.13342253863811493, "global_step": 147577, "epoch": 1658, "lr": 4.32935892254392e-05}
{"train_loss": 0.14523866772651672, "global_step": 147578, "epoch": 1658, "lr": 4.329301472779463e-05}
{"train_loss": 0.12153183668851852, "global_step": 147579, "epoch": 1658, "lr": 4.329244023105175e-05}
{"train_loss": 0.043583549559116364, "global_step": 147580, "epoch": 1658, "lr": 4.32918657352106e-05}
{"train_loss": 0.07472755014896393, "global_step": 147581, "epoch": 1658, "lr": 4.329129124027131e-05}
{"train_loss": 0.09932329505681992, "global_step": 147582, "epoch": 1658, "lr": 4.32907167462339e-05}
{"train_loss": 0.12502622604370117, "global_step": 147583, "epoch": 1658, "lr": 4.3290142253098475e-05}
{"train_loss": 0.1295180320739746, "global_step": 147584, "epoch": 1658, "lr": 4.3289567760865115e-05}
{"train_loss": 0.07455386221408844, "global_step": 147585, "epoch": 1658, "lr": 4.3288993269533886e-05}
{"train_loss": 0.07507777214050293, "global_step": 147586, "epoch": 1658, "lr": 4.328841877910487e-05}
{"train_loss": 0.1066136583685875, "global_step": 147587, "epoch": 1658, "lr": 4.328784428957813e-05}
{"train_loss": 0.0933539941906929, "global_step": 147588, "epoch": 1658, "lr": 4.328726980095378e-05}
{"train_loss": 0.12279006838798523, "global_step": 147589, "epoch": 1658, "lr": 4.328669531323186e-05}
{"train_loss": 0.09383545815944672, "global_step": 147590, "epoch": 1658, "lr": 4.328612082641248e-05}
{"train_loss": 0.09738220274448395, "global_step": 147591, "epoch": 1658, "lr": 4.3285546340495674e-05}
{"train_loss": 0.10639587044715881, "global_step": 147592, "epoch": 1658, "lr": 4.328497185548157e-05}
{"train_loss": 0.1260756105184555, "global_step": 147593, "epoch": 1658, "lr": 4.32843973713702e-05}
{"train_loss": 0.05352155119180679, "global_step": 147594, "epoch": 1658, "lr": 4.3283822888161676e-05}
{"train_loss": 0.10195264220237732, "global_step": 147595, "epoch": 1658, "lr": 4.3283248405856044e-05}
{"train_loss": 0.07873118668794632, "global_step": 147596, "epoch": 1658, "lr": 4.3282673924453416e-05}
{"train_loss": 0.08600469678640366, "global_step": 147597, "epoch": 1658, "lr": 4.328209944395383e-05}
{"train_loss": 0.07050512731075287, "global_step": 147598, "epoch": 1658, "lr": 4.3281524964357406e-05}
{"train_loss": 0.09635590016841888, "global_step": 147599, "epoch": 1658, "lr": 4.328095048566418e-05}
{"train_loss": 0.09708228707313538, "global_step": 147600, "epoch": 1658, "lr": 4.328037600787426e-05}
{"train_loss": 0.05154140293598175, "global_step": 147601, "epoch": 1658, "lr": 4.327980153098771e-05}
{"train_loss": 0.08703406900167465, "global_step": 147602, "epoch": 1658, "lr": 4.327922705500461e-05}
{"train_loss": 0.07239095866680145, "global_step": 147603, "epoch": 1658, "lr": 4.3278652579925036e-05}
{"train_loss": 0.11127300560474396, "global_step": 147604, "epoch": 1658, "lr": 4.327807810574904e-05}
{"train_loss": 0.11116746068000793, "global_step": 147605, "epoch": 1658, "lr": 4.327750363247676e-05}
{"train_loss": 0.0701148584485054, "global_step": 147606, "epoch": 1658, "lr": 4.327692916010821e-05}
{"train_loss": 0.14616645872592926, "global_step": 147607, "epoch": 1658, "lr": 4.327635468864353e-05}
{"train_loss": 0.08078838139772415, "global_step": 147608, "epoch": 1658, "lr": 4.327578021808273e-05}
{"train_loss": 0.13487324118614197, "global_step": 147609, "epoch": 1658, "lr": 4.327520574842594e-05}
{"train_loss": 0.09602683782577515, "global_step": 147610, "epoch": 1658, "lr": 4.3274631279673195e-05}
{"train_loss": 0.14581428468227386, "global_step": 147611, "epoch": 1658, "lr": 4.327405681182461e-05}
{"train_loss": 0.15310919284820557, "global_step": 147612, "epoch": 1658, "lr": 4.3273482344880234e-05}
{"train_loss": 0.0637364462018013, "global_step": 147613, "epoch": 1658, "lr": 4.3272907878840176e-05}
{"train_loss": 0.11721771955490112, "global_step": 147614, "epoch": 1658, "lr": 4.327233341370447e-05}
{"train_loss": 0.05459710210561752, "global_step": 147615, "epoch": 1658, "lr": 4.327175894947324e-05}
{"train_loss": 0.14542177319526672, "global_step": 147616, "epoch": 1658, "lr": 4.327118448614652e-05}
{"train_loss": 0.14144060015678406, "global_step": 147617, "epoch": 1658, "lr": 4.327061002372443e-05}
{"train_loss": 0.12911854684352875, "global_step": 147618, "epoch": 1658, "lr": 4.327003556220699e-05}
{"train_loss": 0.11197421699762344, "global_step": 147619, "epoch": 1658, "lr": 4.326946110159435e-05}
{"train_loss": 0.19155731797218323, "global_step": 147620, "epoch": 1658, "lr": 4.326888664188653e-05}
{"train_loss": 0.10647962242364883, "global_step": 147621, "epoch": 1658, "lr": 4.3268312183083624e-05}
{"train_loss": 0.10152510553598404, "global_step": 147622, "epoch": 1658, "lr": 4.326773772518572e-05}
{"train_loss": 0.13032247126102448, "global_step": 147623, "epoch": 1658, "lr": 4.326716326819287e-05}
{"train_loss": 0.12442303448915482, "global_step": 147624, "epoch": 1658, "lr": 4.326658881210519e-05}
{"train_loss": 0.1426696926355362, "global_step": 147625, "epoch": 1658, "lr": 4.326601435692271e-05}
{"train_loss": 0.09462545067071915, "global_step": 147626, "epoch": 1658, "lr": 4.3265439902645554e-05}
{"train_loss": 0.0950256735086441, "global_step": 147627, "epoch": 1658, "lr": 4.326486544927376e-05}
{"train_loss": 0.15754321217536926, "global_step": 147628, "epoch": 1658, "lr": 4.3264290996807445e-05}
{"train_loss": 0.10659698396921158, "global_step": 147629, "epoch": 1658, "lr": 4.326371654524664e-05}
{"train_loss": 0.18128199875354767, "global_step": 147630, "epoch": 1658, "lr": 4.326314209459147e-05}
{"train_loss": 0.26276707649230957, "global_step": 147631, "epoch": 1658, "lr": 4.3262567644841966e-05}
{"train_loss": 0.14827874302864075, "global_step": 147632, "epoch": 1658, "lr": 4.326199319599825e-05}
{"train_loss": 0.0969989001750946, "global_step": 147633, "epoch": 1658, "lr": 4.326141874806035e-05}
{"train_loss": 0.0926593542098999, "global_step": 147634, "epoch": 1658, "lr": 4.32608443010284e-05}
{"train_loss": 0.09913095831871033, "global_step": 147635, "epoch": 1658, "lr": 4.326026985490242e-05}
{"train_loss": 0.09472880512475967, "global_step": 147636, "epoch": 1658, "lr": 4.325969540968253e-05}
{"train_loss": 0.10237428545951843, "global_step": 147637, "epoch": 1658, "lr": 4.3259120965368785e-05}
{"train_loss": 0.0845608040690422, "global_step": 147638, "epoch": 1658, "lr": 4.3258546521961284e-05}
{"train_loss": 0.0946173369884491, "global_step": 147639, "epoch": 1658, "lr": 4.3257972079460074e-05}
{"train_loss": 0.08405585587024689, "global_step": 147640, "epoch": 1658, "lr": 4.325739763786526e-05}
{"train_loss": 0.12181362509727478, "global_step": 147641, "epoch": 1658, "lr": 4.325682319717689e-05}
{"train_loss": 0.10758183151483536, "global_step": 147642, "epoch": 1658, "lr": 4.3256248757395054e-05}
{"train_loss": 0.09604299068450928, "global_step": 147643, "epoch": 1658, "lr": 4.325567431851986e-05}
{"train_loss": 0.13723550736904144, "global_step": 147644, "epoch": 1658, "lr": 4.325509988055133e-05}
{"train_loss": 0.13253581523895264, "global_step": 147645, "epoch": 1658, "lr": 4.32545254434896e-05}
{"train_loss": 0.09477904438972473, "global_step": 147646, "epoch": 1658, "lr": 4.3253951007334684e-05}
{"train_loss": 0.1415545642375946, "global_step": 147647, "epoch": 1658, "lr": 4.325337657208672e-05}
{"train_loss": 0.185285285115242, "global_step": 147648, "epoch": 1658, "lr": 4.325280213774573e-05}
{"train_loss": 0.0827861800789833, "global_step": 147649, "epoch": 1658, "lr": 4.3252227704311846e-05}
{"train_loss": 0.11077162732234162, "global_step": 147650, "epoch": 1658, "lr": 4.32516532717851e-05, "val_loss": 6.302161693572998}
{"train_loss": 0.13215693831443787, "global_step": 147651, "epoch": 1659, "lr": 4.32510788401656e-05}
{"train_loss": 0.09865005314350128, "global_step": 147652, "epoch": 1659, "lr": 4.325050440945339e-05}
{"train_loss": 0.16256745159626007, "global_step": 147653, "epoch": 1659, "lr": 4.324992997964859e-05}
{"train_loss": 0.06342799216508865, "global_step": 147654, "epoch": 1659, "lr": 4.3249355550751234e-05}
{"train_loss": 0.15620149672031403, "global_step": 147655, "epoch": 1659, "lr": 4.324878112276145e-05}
{"train_loss": 0.13863615691661835, "global_step": 147656, "epoch": 1659, "lr": 4.324820669567925e-05}
{"train_loss": 0.059861838817596436, "global_step": 147657, "epoch": 1659, "lr": 4.324763226950477e-05}
{"train_loss": 0.11210425198078156, "global_step": 147658, "epoch": 1659, "lr": 4.324705784423806e-05}
{"train_loss": 0.08610299974679947, "global_step": 147659, "epoch": 1659, "lr": 4.324648341987918e-05}
{"train_loss": 0.08391476422548294, "global_step": 147660, "epoch": 1659, "lr": 4.3245908996428255e-05}
{"train_loss": 0.08225554972887039, "global_step": 147661, "epoch": 1659, "lr": 4.324533457388532e-05}
{"train_loss": 0.10645806044340134, "global_step": 147662, "epoch": 1659, "lr": 4.324476015225048e-05}
{"train_loss": 0.1878759264945984, "global_step": 147663, "epoch": 1659, "lr": 4.3244185731523776e-05}
{"train_loss": 0.1343306303024292, "global_step": 147664, "epoch": 1659, "lr": 4.324361131170534e-05}
{"train_loss": 0.11736731231212616, "global_step": 147665, "epoch": 1659, "lr": 4.3243036892795194e-05}
{"train_loss": 0.13457687199115753, "global_step": 147666, "epoch": 1659, "lr": 4.324246247479346e-05}
{"train_loss": 0.06964593380689621, "global_step": 147667, "epoch": 1659, "lr": 4.3241888057700166e-05}
{"train_loss": 0.12807607650756836, "global_step": 147668, "epoch": 1659, "lr": 4.3241313641515444e-05}
{"train_loss": 0.07084938883781433, "global_step": 147669, "epoch": 1659, "lr": 4.324073922623932e-05}
{"train_loss": 0.07190029323101044, "global_step": 147670, "epoch": 1659, "lr": 4.324016481187193e-05}
{"train_loss": 0.06778007000684738, "global_step": 147671, "epoch": 1659, "lr": 4.323959039841329e-05}
{"train_loss": 0.12664265930652618, "global_step": 147672, "epoch": 1659, "lr": 4.323901598586352e-05}
{"train_loss": 0.11786054074764252, "global_step": 147673, "epoch": 1659, "lr": 4.3238441574222664e-05}
{"train_loss": 0.12613190710544586, "global_step": 147674, "epoch": 1659, "lr": 4.323786716349084e-05}
{"train_loss": 0.08330006152391434, "global_step": 147675, "epoch": 1659, "lr": 4.323729275366808e-05}
{"train_loss": 0.08710356056690216, "global_step": 147676, "epoch": 1659, "lr": 4.323671834475451e-05}
{"train_loss": 0.08464284986257553, "global_step": 147677, "epoch": 1659, "lr": 4.323614393675015e-05}
{"train_loss": 0.10254478454589844, "global_step": 147678, "epoch": 1659, "lr": 4.323556952965513e-05}
{"train_loss": 0.1318390816450119, "global_step": 147679, "epoch": 1659, "lr": 4.323499512346951e-05}
{"train_loss": 0.12273039668798447, "global_step": 147680, "epoch": 1659, "lr": 4.3234420718193336e-05}
{"train_loss": 0.14878815412521362, "global_step": 147681, "epoch": 1659, "lr": 4.3233846313826735e-05}
{"train_loss": 0.12003947049379349, "global_step": 147682, "epoch": 1659, "lr": 4.323327191036975e-05}
{"train_loss": 0.07608433812856674, "global_step": 147683, "epoch": 1659, "lr": 4.323269750782248e-05}
{"train_loss": 0.08938315510749817, "global_step": 147684, "epoch": 1659, "lr": 4.323212310618497e-05}
{"train_loss": 0.06466715037822723, "global_step": 147685, "epoch": 1659, "lr": 4.323154870545734e-05}
{"train_loss": 0.11572292447090149, "global_step": 147686, "epoch": 1659, "lr": 4.3230974305639624e-05}
{"train_loss": 0.12901969254016876, "global_step": 147687, "epoch": 1659, "lr": 4.323039990673195e-05}
{"train_loss": 0.11243452876806259, "global_step": 147688, "epoch": 1659, "lr": 4.3229825508734337e-05}
{"train_loss": 0.1289127767086029, "global_step": 147689, "epoch": 1659, "lr": 4.322925111164692e-05}
{"train_loss": 0.18999701738357544, "global_step": 147690, "epoch": 1659, "lr": 4.322867671546972e-05}
{"train_loss": 0.09497017413377762, "global_step": 147691, "epoch": 1659, "lr": 4.322810232020287e-05}
{"train_loss": 0.11780794709920883, "global_step": 147692, "epoch": 1659, "lr": 4.322752792584639e-05}
{"train_loss": 0.09005812555551529, "global_step": 147693, "epoch": 1659, "lr": 4.3226953532400414e-05}
{"train_loss": 0.09553138166666031, "global_step": 147694, "epoch": 1659, "lr": 4.322637913986496e-05}
{"train_loss": 0.15324316918849945, "global_step": 147695, "epoch": 1659, "lr": 4.3225804748240154e-05}
{"train_loss": 0.09110618382692337, "global_step": 147696, "epoch": 1659, "lr": 4.322523035752606e-05}
{"train_loss": 0.17098234593868256, "global_step": 147697, "epoch": 1659, "lr": 4.322465596772275e-05}
{"train_loss": 0.1393689066171646, "global_step": 147698, "epoch": 1659, "lr": 4.3224081578830305e-05}
{"train_loss": 0.0800161212682724, "global_step": 147699, "epoch": 1659, "lr": 4.322350719084879e-05}
{"train_loss": 0.06798119843006134, "global_step": 147700, "epoch": 1659, "lr": 4.3222932803778305e-05}
{"train_loss": 0.1331099271774292, "global_step": 147701, "epoch": 1659, "lr": 4.322235841761889e-05}
{"train_loss": 0.11896064132452011, "global_step": 147702, "epoch": 1659, "lr": 4.3221784032370675e-05}
{"train_loss": 0.07581281661987305, "global_step": 147703, "epoch": 1659, "lr": 4.3221209648033686e-05}
{"train_loss": 0.06615972518920898, "global_step": 147704, "epoch": 1659, "lr": 4.3220635264608044e-05}
{"train_loss": 0.07120697945356369, "global_step": 147705, "epoch": 1659, "lr": 4.322006088209378e-05}
{"train_loss": 0.08537483215332031, "global_step": 147706, "epoch": 1659, "lr": 4.3219486500491016e-05}
{"train_loss": 0.0965258777141571, "global_step": 147707, "epoch": 1659, "lr": 4.321891211979979e-05}
{"train_loss": 0.10516064614057541, "global_step": 147708, "epoch": 1659, "lr": 4.3218337740020235e-05}
{"train_loss": 0.06362384557723999, "global_step": 147709, "epoch": 1659, "lr": 4.3217763361152355e-05}
{"train_loss": 0.15655873715877533, "global_step": 147710, "epoch": 1659, "lr": 4.321718898319628e-05}
{"train_loss": 0.13255076110363007, "global_step": 147711, "epoch": 1659, "lr": 4.3216614606152065e-05}
{"train_loss": 0.09796192497015, "global_step": 147712, "epoch": 1659, "lr": 4.32160402300198e-05}
{"train_loss": 0.08993128687143326, "global_step": 147713, "epoch": 1659, "lr": 4.3215465854799546e-05}
{"train_loss": 0.05946512892842293, "global_step": 147714, "epoch": 1659, "lr": 4.321489148049139e-05}
{"train_loss": 0.09486855566501617, "global_step": 147715, "epoch": 1659, "lr": 4.321431710709542e-05}
{"train_loss": 0.13362468779087067, "global_step": 147716, "epoch": 1659, "lr": 4.32137427346117e-05}
{"train_loss": 0.08902847021818161, "global_step": 147717, "epoch": 1659, "lr": 4.321316836304032e-05}
{"train_loss": 0.11253554373979568, "global_step": 147718, "epoch": 1659, "lr": 4.321259399238132e-05}
{"train_loss": 0.10583087801933289, "global_step": 147719, "epoch": 1659, "lr": 4.321201962263483e-05}
{"train_loss": 0.06460904330015182, "global_step": 147720, "epoch": 1659, "lr": 4.321144525380088e-05}
{"train_loss": 0.11120807379484177, "global_step": 147721, "epoch": 1659, "lr": 4.321087088587959e-05}
{"train_loss": 0.15180498361587524, "global_step": 147722, "epoch": 1659, "lr": 4.321029651887099e-05}
{"train_loss": 0.14169718325138092, "global_step": 147723, "epoch": 1659, "lr": 4.3209722152775214e-05}
{"train_loss": 0.059955984354019165, "global_step": 147724, "epoch": 1659, "lr": 4.3209147787592276e-05}
{"train_loss": 0.10686685144901276, "global_step": 147725, "epoch": 1659, "lr": 4.320857342332231e-05}
{"train_loss": 0.08992321789264679, "global_step": 147726, "epoch": 1659, "lr": 4.3207999059965346e-05}
{"train_loss": 0.12779313325881958, "global_step": 147727, "epoch": 1659, "lr": 4.320742469752151e-05}
{"train_loss": 0.0826825350522995, "global_step": 147728, "epoch": 1659, "lr": 4.320685033599083e-05}
{"train_loss": 0.07951605319976807, "global_step": 147729, "epoch": 1659, "lr": 4.320627597537342e-05}
{"train_loss": 0.06791194528341293, "global_step": 147730, "epoch": 1659, "lr": 4.320570161566933e-05}
{"train_loss": 0.11271732300519943, "global_step": 147731, "epoch": 1659, "lr": 4.320512725687866e-05}
{"train_loss": 0.11168921738862991, "global_step": 147732, "epoch": 1659, "lr": 4.320455289900148e-05}
{"train_loss": 0.13562341034412384, "global_step": 147733, "epoch": 1659, "lr": 4.320397854203786e-05}
{"train_loss": 0.06948674470186234, "global_step": 147734, "epoch": 1659, "lr": 4.320340418598788e-05}
{"train_loss": 0.09420547634363174, "global_step": 147735, "epoch": 1659, "lr": 4.3202829830851635e-05}
{"train_loss": 0.08802345395088196, "global_step": 147736, "epoch": 1659, "lr": 4.320225547662918e-05}
{"train_loss": 0.06291059404611588, "global_step": 147737, "epoch": 1659, "lr": 4.320168112332059e-05}
{"train_loss": 0.10411682724952698, "global_step": 147738, "epoch": 1659, "lr": 4.320110677092596e-05}
{"train_loss": 0.10509481687056885, "global_step": 147739, "epoch": 1659, "lr": 4.3200532419445344e-05, "val_loss": 6.508068561553955}
{"train_loss": 0.07887934148311615, "global_step": 147740, "epoch": 1660, "lr": 4.3199958068878854e-05}
{"train_loss": 0.0916256457567215, "global_step": 147741, "epoch": 1660, "lr": 4.319938371922653e-05}
{"train_loss": 0.0726366713643074, "global_step": 147742, "epoch": 1660, "lr": 4.319880937048848e-05}
{"train_loss": 0.1351872831583023, "global_step": 147743, "epoch": 1660, "lr": 4.319823502266474e-05}
{"train_loss": 0.08315638452768326, "global_step": 147744, "epoch": 1660, "lr": 4.319766067575545e-05}
{"train_loss": 0.08024320006370544, "global_step": 147745, "epoch": 1660, "lr": 4.319708632976063e-05}
{"train_loss": 0.11985043436288834, "global_step": 147746, "epoch": 1660, "lr": 4.319651198468039e-05}
{"train_loss": 0.14916253089904785, "global_step": 147747, "epoch": 1660, "lr": 4.319593764051478e-05}
{"train_loss": 0.11390221863985062, "global_step": 147748, "epoch": 1660, "lr": 4.319536329726392e-05}
{"train_loss": 0.10953771322965622, "global_step": 147749, "epoch": 1660, "lr": 4.319478895492784e-05}
{"train_loss": 0.11874671280384064, "global_step": 147750, "epoch": 1660, "lr": 4.3194214613506633e-05}
{"train_loss": 0.14672823250293732, "global_step": 147751, "epoch": 1660, "lr": 4.319364027300039e-05}
{"train_loss": 0.11800823360681534, "global_step": 147752, "epoch": 1660, "lr": 4.3193065933409175e-05}
{"train_loss": 0.19871407747268677, "global_step": 147753, "epoch": 1660, "lr": 4.3192491594733077e-05}
{"train_loss": 0.13778655230998993, "global_step": 147754, "epoch": 1660, "lr": 4.319191725697216e-05}
{"train_loss": 0.0872747153043747, "global_step": 147755, "epoch": 1660, "lr": 4.319134292012652e-05}
{"train_loss": 0.09486986696720123, "global_step": 147756, "epoch": 1660, "lr": 4.3190768584196204e-05}
{"train_loss": 0.09639881551265717, "global_step": 147757, "epoch": 1660, "lr": 4.3190194249181313e-05}
{"train_loss": 0.05261697620153427, "global_step": 147758, "epoch": 1660, "lr": 4.318961991508191e-05}
{"train_loss": 0.06882177293300629, "global_step": 147759, "epoch": 1660, "lr": 4.3189045581898095e-05}
{"train_loss": 0.1527940183877945, "global_step": 147760, "epoch": 1660, "lr": 4.3188471249629916e-05}
{"train_loss": 0.10766331851482391, "global_step": 147761, "epoch": 1660, "lr": 4.3187896918277474e-05}
{"train_loss": 0.07555008679628372, "global_step": 147762, "epoch": 1660, "lr": 4.318732258784082e-05}
{"train_loss": 0.11541779339313507, "global_step": 147763, "epoch": 1660, "lr": 4.318674825832007e-05}
{"train_loss": 0.1104038655757904, "global_step": 147764, "epoch": 1660, "lr": 4.318617392971526e-05}
{"train_loss": 0.0775466188788414, "global_step": 147765, "epoch": 1660, "lr": 4.31855996020265e-05}
{"train_loss": 0.08185549825429916, "global_step": 147766, "epoch": 1660, "lr": 4.3185025275253835e-05}
{"train_loss": 0.19185706973075867, "global_step": 147767, "epoch": 1660, "lr": 4.318445094939738e-05}
{"train_loss": 0.09934331476688385, "global_step": 147768, "epoch": 1660, "lr": 4.318387662445717e-05}
{"train_loss": 0.07113461196422577, "global_step": 147769, "epoch": 1660, "lr": 4.318330230043332e-05}
{"train_loss": 0.06001291424036026, "global_step": 147770, "epoch": 1660, "lr": 4.318272797732589e-05}
{"train_loss": 0.08927909284830093, "global_step": 147771, "epoch": 1660, "lr": 4.318215365513496e-05}
{"train_loss": 0.21963000297546387, "global_step": 147772, "epoch": 1660, "lr": 4.318157933386061e-05}
{"train_loss": 0.10164456814527512, "global_step": 147773, "epoch": 1660, "lr": 4.3181005013502904e-05}
{"train_loss": 0.0741422027349472, "global_step": 147774, "epoch": 1660, "lr": 4.318043069406195e-05}
{"train_loss": 0.08357372879981995, "global_step": 147775, "epoch": 1660, "lr": 4.317985637553777e-05}
{"train_loss": 0.10498464852571487, "global_step": 147776, "epoch": 1660, "lr": 4.317928205793051e-05}
{"train_loss": 0.11100494861602783, "global_step": 147777, "epoch": 1660, "lr": 4.3178707741240185e-05}
{"train_loss": 0.10526813566684723, "global_step": 147778, "epoch": 1660, "lr": 4.3178133425466915e-05}
{"train_loss": 0.13156136870384216, "global_step": 147779, "epoch": 1660, "lr": 4.3177559110610745e-05}
{"train_loss": 0.12058260291814804, "global_step": 147780, "epoch": 1660, "lr": 4.31769847966718e-05}
{"train_loss": 0.11879440397024155, "global_step": 147781, "epoch": 1660, "lr": 4.317641048365009e-05}
{"train_loss": 0.11640073359012604, "global_step": 147782, "epoch": 1660, "lr": 4.317583617154576e-05}
{"train_loss": 0.13983890414237976, "global_step": 147783, "epoch": 1660, "lr": 4.317526186035883e-05}
{"train_loss": 0.09516969323158264, "global_step": 147784, "epoch": 1660, "lr": 4.317468755008942e-05}
{"train_loss": 0.1392611265182495, "global_step": 147785, "epoch": 1660, "lr": 4.317411324073757e-05}
{"train_loss": 0.06002764403820038, "global_step": 147786, "epoch": 1660, "lr": 4.317353893230339e-05}
{"train_loss": 0.12342757731676102, "global_step": 147787, "epoch": 1660, "lr": 4.317296462478695e-05}
{"train_loss": 0.10327400267124176, "global_step": 147788, "epoch": 1660, "lr": 4.317239031818832e-05}
{"train_loss": 0.11843656003475189, "global_step": 147789, "epoch": 1660, "lr": 4.3171816012507576e-05}
{"train_loss": 0.15108992159366608, "global_step": 147790, "epoch": 1660, "lr": 4.31712417077448e-05}
{"train_loss": 0.09997809678316116, "global_step": 147791, "epoch": 1660, "lr": 4.3170667403900064e-05}
{"train_loss": 0.08351575583219528, "global_step": 147792, "epoch": 1660, "lr": 4.317009310097345e-05}
{"train_loss": 0.09003255516290665, "global_step": 147793, "epoch": 1660, "lr": 4.316951879896504e-05}
{"train_loss": 0.1233491450548172, "global_step": 147794, "epoch": 1660, "lr": 4.3168944497874894e-05}
{"train_loss": 0.09718094021081924, "global_step": 147795, "epoch": 1660, "lr": 4.316837019770312e-05}
{"train_loss": 0.07328156381845474, "global_step": 147796, "epoch": 1660, "lr": 4.316779589844975e-05}
{"train_loss": 0.10612399876117706, "global_step": 147797, "epoch": 1660, "lr": 4.31672216001149e-05}
{"train_loss": 0.07879983633756638, "global_step": 147798, "epoch": 1660, "lr": 4.3166647302698624e-05}
{"train_loss": 0.08598969876766205, "global_step": 147799, "epoch": 1660, "lr": 4.3166073006201026e-05}
{"train_loss": 0.09215378761291504, "global_step": 147800, "epoch": 1660, "lr": 4.316549871062214e-05}
{"train_loss": 0.07759420573711395, "global_step": 147801, "epoch": 1660, "lr": 4.31649244159621e-05}
{"train_loss": 0.07908362150192261, "global_step": 147802, "epoch": 1660, "lr": 4.316435012222093e-05}
{"train_loss": 0.10078920423984528, "global_step": 147803, "epoch": 1660, "lr": 4.316377582939875e-05}
{"train_loss": 0.20725515484809875, "global_step": 147804, "epoch": 1660, "lr": 4.316320153749559e-05}
{"train_loss": 0.09473540633916855, "global_step": 147805, "epoch": 1660, "lr": 4.316262724651156e-05}
{"train_loss": 0.05746745318174362, "global_step": 147806, "epoch": 1660, "lr": 4.3162052956446754e-05}
{"train_loss": 0.11455502361059189, "global_step": 147807, "epoch": 1660, "lr": 4.3161478667301206e-05}
{"train_loss": 0.10247070342302322, "global_step": 147808, "epoch": 1660, "lr": 4.3160904379075026e-05}
{"train_loss": 0.15953442454338074, "global_step": 147809, "epoch": 1660, "lr": 4.316033009176828e-05}
{"train_loss": 0.0881752297282219, "global_step": 147810, "epoch": 1660, "lr": 4.315975580538103e-05}
{"train_loss": 0.0846007689833641, "global_step": 147811, "epoch": 1660, "lr": 4.3159181519913376e-05}
{"train_loss": 0.06327394396066666, "global_step": 147812, "epoch": 1660, "lr": 4.315860723536539e-05}
{"train_loss": 0.05881975591182709, "global_step": 147813, "epoch": 1660, "lr": 4.3158032951737146e-05}
{"train_loss": 0.14627155661582947, "global_step": 147814, "epoch": 1660, "lr": 4.315745866902873e-05}
{"train_loss": 0.07330433279275894, "global_step": 147815, "epoch": 1660, "lr": 4.315688438724018e-05}
{"train_loss": 0.05676666274666786, "global_step": 147816, "epoch": 1660, "lr": 4.315631010637163e-05}
{"train_loss": 0.15884020924568176, "global_step": 147817, "epoch": 1660, "lr": 4.315573582642312e-05}
{"train_loss": 0.08044283837080002, "global_step": 147818, "epoch": 1660, "lr": 4.315516154739475e-05}
{"train_loss": 0.12350312620401382, "global_step": 147819, "epoch": 1660, "lr": 4.315458726928656e-05}
{"train_loss": 0.13013426959514618, "global_step": 147820, "epoch": 1660, "lr": 4.315401299209868e-05}
{"train_loss": 0.10401798784732819, "global_step": 147821, "epoch": 1660, "lr": 4.315343871583113e-05}
{"train_loss": 0.1233179047703743, "global_step": 147822, "epoch": 1660, "lr": 4.315286444048403e-05}
{"train_loss": 0.08315557986497879, "global_step": 147823, "epoch": 1660, "lr": 4.315229016605745e-05}
{"train_loss": 0.11958888173103333, "global_step": 147824, "epoch": 1660, "lr": 4.315171589255145e-05}
{"train_loss": 0.1314755231142044, "global_step": 147825, "epoch": 1660, "lr": 4.315114161996614e-05}
{"train_loss": 0.08478937298059464, "global_step": 147826, "epoch": 1660, "lr": 4.315056734830155e-05}
{"train_loss": 0.0864545926451683, "global_step": 147827, "epoch": 1660, "lr": 4.3149993077557805e-05}
{"train_loss": 0.10575345723649089, "global_step": 147828, "epoch": 1660, "lr": 4.3149418807734955e-05, "val_loss": 6.232666492462158, "train_action_mse_error": 25.827198028564453}
{"train_loss": 0.12445099651813507, "global_step": 147829, "epoch": 1661, "lr": 4.314884453883308e-05}
{"train_loss": 0.10572133213281631, "global_step": 147830, "epoch": 1661, "lr": 4.3148270270852256e-05}
{"train_loss": 0.07979106903076172, "global_step": 147831, "epoch": 1661, "lr": 4.314769600379256e-05}
{"train_loss": 0.11960386484861374, "global_step": 147832, "epoch": 1661, "lr": 4.3147121737654075e-05}
{"train_loss": 0.08403631299734116, "global_step": 147833, "epoch": 1661, "lr": 4.3146547472436895e-05}
{"train_loss": 0.15545262396335602, "global_step": 147834, "epoch": 1661, "lr": 4.314597320814104e-05}
{"train_loss": 0.18152685463428497, "global_step": 147835, "epoch": 1661, "lr": 4.314539894476666e-05}
{"train_loss": 0.061568208038806915, "global_step": 147836, "epoch": 1661, "lr": 4.314482468231378e-05}
{"train_loss": 0.1366136074066162, "global_step": 147837, "epoch": 1661, "lr": 4.3144250420782505e-05}
{"train_loss": 0.10877606272697449, "global_step": 147838, "epoch": 1661, "lr": 4.3143676160172886e-05}
{"train_loss": 0.21287968754768372, "global_step": 147839, "epoch": 1661, "lr": 4.3143101900485036e-05}
{"train_loss": 0.14059779047966003, "global_step": 147840, "epoch": 1661, "lr": 4.314252764171899e-05}
{"train_loss": 0.08673552423715591, "global_step": 147841, "epoch": 1661, "lr": 4.314195338387485e-05}
{"train_loss": 0.0933549776673317, "global_step": 147842, "epoch": 1661, "lr": 4.314137912695271e-05}
{"train_loss": 0.15733212232589722, "global_step": 147843, "epoch": 1661, "lr": 4.314080487095261e-05}
{"train_loss": 0.10633165389299393, "global_step": 147844, "epoch": 1661, "lr": 4.314023061587467e-05}
{"train_loss": 0.09199677407741547, "global_step": 147845, "epoch": 1661, "lr": 4.313965636171891e-05}
{"train_loss": 0.11186492443084717, "global_step": 147846, "epoch": 1661, "lr": 4.3139082108485465e-05}
{"train_loss": 0.12239310145378113, "global_step": 147847, "epoch": 1661, "lr": 4.313850785617438e-05}
{"train_loss": 0.12116701900959015, "global_step": 147848, "epoch": 1661, "lr": 4.3137933604785735e-05}
{"train_loss": 0.1624988615512848, "global_step": 147849, "epoch": 1661, "lr": 4.3137359354319614e-05}
{"train_loss": 0.12393441051244736, "global_step": 147850, "epoch": 1661, "lr": 4.313678510477609e-05}
{"train_loss": 0.1288985311985016, "global_step": 147851, "epoch": 1661, "lr": 4.313621085615524e-05}
{"train_loss": 0.0572044812142849, "global_step": 147852, "epoch": 1661, "lr": 4.313563660845715e-05}
{"train_loss": 0.06619182229042053, "global_step": 147853, "epoch": 1661, "lr": 4.313506236168188e-05}
{"train_loss": 0.11738715320825577, "global_step": 147854, "epoch": 1661, "lr": 4.313448811582953e-05}
{"train_loss": 0.0711427628993988, "global_step": 147855, "epoch": 1661, "lr": 4.3133913870900146e-05}
{"train_loss": 0.09441732615232468, "global_step": 147856, "epoch": 1661, "lr": 4.3133339626893844e-05}
{"train_loss": 0.13394805788993835, "global_step": 147857, "epoch": 1661, "lr": 4.313276538381066e-05}
{"train_loss": 0.14150750637054443, "global_step": 147858, "epoch": 1661, "lr": 4.313219114165069e-05}
{"train_loss": 0.06249161809682846, "global_step": 147859, "epoch": 1661, "lr": 4.313161690041404e-05}
{"train_loss": 0.10952925682067871, "global_step": 147860, "epoch": 1661, "lr": 4.313104266010074e-05}
{"train_loss": 0.07341759651899338, "global_step": 147861, "epoch": 1661, "lr": 4.31304684207109e-05}
{"train_loss": 0.22027136385440826, "global_step": 147862, "epoch": 1661, "lr": 4.3129894182244565e-05}
{"train_loss": 0.11400146782398224, "global_step": 147863, "epoch": 1661, "lr": 4.3129319944701856e-05}
{"train_loss": 0.10897582769393921, "global_step": 147864, "epoch": 1661, "lr": 4.3128745708082805e-05}
{"train_loss": 0.1044078916311264, "global_step": 147865, "epoch": 1661, "lr": 4.3128171472387536e-05}
{"train_loss": 0.07762228697538376, "global_step": 147866, "epoch": 1661, "lr": 4.312759723761608e-05}
{"train_loss": 0.15540564060211182, "global_step": 147867, "epoch": 1661, "lr": 4.312702300376854e-05}
{"train_loss": 0.09061232954263687, "global_step": 147868, "epoch": 1661, "lr": 4.312644877084499e-05}
{"train_loss": 0.10646317154169083, "global_step": 147869, "epoch": 1661, "lr": 4.31258745388455e-05}
{"train_loss": 0.10159142315387726, "global_step": 147870, "epoch": 1661, "lr": 4.312530030777016e-05}
{"train_loss": 0.09015297889709473, "global_step": 147871, "epoch": 1661, "lr": 4.312472607761905e-05}
{"train_loss": 0.09906638413667679, "global_step": 147872, "epoch": 1661, "lr": 4.312415184839221e-05}
{"train_loss": 0.13338786363601685, "global_step": 147873, "epoch": 1661, "lr": 4.3123577620089766e-05}
{"train_loss": 0.08834362775087357, "global_step": 147874, "epoch": 1661, "lr": 4.312300339271175e-05}
{"train_loss": 0.14190000295639038, "global_step": 147875, "epoch": 1661, "lr": 4.312242916625829e-05}
{"train_loss": 0.13360516726970673, "global_step": 147876, "epoch": 1661, "lr": 4.312185494072941e-05}
{"train_loss": 0.16197419166564941, "global_step": 147877, "epoch": 1661, "lr": 4.312128071612521e-05}
{"train_loss": 0.07441845536231995, "global_step": 147878, "epoch": 1661, "lr": 4.31207064924458e-05}
{"train_loss": 0.10513918101787567, "global_step": 147879, "epoch": 1661, "lr": 4.31201322696912e-05}
{"train_loss": 0.09447908401489258, "global_step": 147880, "epoch": 1661, "lr": 4.3119558047861534e-05}
{"train_loss": 0.04828881844878197, "global_step": 147881, "epoch": 1661, "lr": 4.3118983826956835e-05}
{"train_loss": 0.16143955290317535, "global_step": 147882, "epoch": 1661, "lr": 4.311840960697724e-05}
{"train_loss": 0.09539610892534256, "global_step": 147883, "epoch": 1661, "lr": 4.311783538792276e-05}
{"train_loss": 0.11678384989500046, "global_step": 147884, "epoch": 1661, "lr": 4.311726116979352e-05}
{"train_loss": 0.09044353663921356, "global_step": 147885, "epoch": 1661, "lr": 4.311668695258958e-05}
{"train_loss": 0.10290185362100601, "global_step": 147886, "epoch": 1661, "lr": 4.311611273631101e-05}
{"train_loss": 0.10761889815330505, "global_step": 147887, "epoch": 1661, "lr": 4.31155385209579e-05}
{"train_loss": 0.08920139819383621, "global_step": 147888, "epoch": 1661, "lr": 4.311496430653032e-05}
{"train_loss": 0.07241667807102203, "global_step": 147889, "epoch": 1661, "lr": 4.311439009302834e-05}
{"train_loss": 0.062268685549497604, "global_step": 147890, "epoch": 1661, "lr": 4.311381588045207e-05}
{"train_loss": 0.08780486136674881, "global_step": 147891, "epoch": 1661, "lr": 4.311324166880154e-05}
{"train_loss": 0.09755107760429382, "global_step": 147892, "epoch": 1661, "lr": 4.311266745807687e-05}
{"train_loss": 0.09911717474460602, "global_step": 147893, "epoch": 1661, "lr": 4.3112093248278094e-05}
{"train_loss": 0.07680186629295349, "global_step": 147894, "epoch": 1661, "lr": 4.311151903940534e-05}
{"train_loss": 0.08962950855493546, "global_step": 147895, "epoch": 1661, "lr": 4.3110944831458635e-05}
{"train_loss": 0.14926040172576904, "global_step": 147896, "epoch": 1661, "lr": 4.311037062443808e-05}
{"train_loss": 0.15226393938064575, "global_step": 147897, "epoch": 1661, "lr": 4.310979641834377e-05}
{"train_loss": 0.09544944763183594, "global_step": 147898, "epoch": 1661, "lr": 4.310922221317574e-05}
{"train_loss": 0.07507730275392532, "global_step": 147899, "epoch": 1661, "lr": 4.3108648008934125e-05}
{"train_loss": 0.11738467961549759, "global_step": 147900, "epoch": 1661, "lr": 4.310807380561893e-05}
{"train_loss": 0.10336339473724365, "global_step": 147901, "epoch": 1661, "lr": 4.3107499603230306e-05}
{"train_loss": 0.10885322839021683, "global_step": 147902, "epoch": 1661, "lr": 4.3106925401768266e-05}
{"train_loss": 0.10026222467422485, "global_step": 147903, "epoch": 1661, "lr": 4.310635120123294e-05}
{"train_loss": 0.07021629065275192, "global_step": 147904, "epoch": 1661, "lr": 4.310577700162437e-05}
{"train_loss": 0.13717685639858246, "global_step": 147905, "epoch": 1661, "lr": 4.310520280294264e-05}
{"train_loss": 0.09205002337694168, "global_step": 147906, "epoch": 1661, "lr": 4.310462860518783e-05}
{"train_loss": 0.20544776320457458, "global_step": 147907, "epoch": 1661, "lr": 4.3104054408360026e-05}
{"train_loss": 0.1080404743552208, "global_step": 147908, "epoch": 1661, "lr": 4.31034802124593e-05}
{"train_loss": 0.12530775368213654, "global_step": 147909, "epoch": 1661, "lr": 4.310290601748572e-05}
{"train_loss": 0.07430675625801086, "global_step": 147910, "epoch": 1661, "lr": 4.3102331823439374e-05}
{"train_loss": 0.08311019092798233, "global_step": 147911, "epoch": 1661, "lr": 4.3101757630320344e-05}
{"train_loss": 0.09023275226354599, "global_step": 147912, "epoch": 1661, "lr": 4.310118343812868e-05}
{"train_loss": 0.10842406004667282, "global_step": 147913, "epoch": 1661, "lr": 4.3100609246864476e-05}
{"train_loss": 0.11160275340080261, "global_step": 147914, "epoch": 1661, "lr": 4.3100035056527834e-05}
{"train_loss": 0.09556370228528976, "global_step": 147915, "epoch": 1661, "lr": 4.309946086711878e-05}
{"train_loss": 0.123455710709095, "global_step": 147916, "epoch": 1661, "lr": 4.3098886678637454e-05}
{"train_loss": 0.10968640459220061, "global_step": 147917, "epoch": 1661, "lr": 4.3098312491083865e-05, "val_loss": 6.012617588043213}
{"train_loss": 0.08313146978616714, "global_step": 147918, "epoch": 1662, "lr": 4.309773830445815e-05}
{"train_loss": 0.10234445333480835, "global_step": 147919, "epoch": 1662, "lr": 4.3097164118760334e-05}
{"train_loss": 0.1132846400141716, "global_step": 147920, "epoch": 1662, "lr": 4.309658993399054e-05}
{"train_loss": 0.1623029261827469, "global_step": 147921, "epoch": 1662, "lr": 4.3096015750148814e-05}
{"train_loss": 0.10870353132486343, "global_step": 147922, "epoch": 1662, "lr": 4.3095441567235264e-05}
{"train_loss": 0.12723420560359955, "global_step": 147923, "epoch": 1662, "lr": 4.309486738524992e-05}
{"train_loss": 0.09420039504766464, "global_step": 147924, "epoch": 1662, "lr": 4.309429320419292e-05}
{"train_loss": 0.13522788882255554, "global_step": 147925, "epoch": 1662, "lr": 4.309371902406428e-05}
{"train_loss": 0.13618510961532593, "global_step": 147926, "epoch": 1662, "lr": 4.309314484486412e-05}
{"train_loss": 0.17327716946601868, "global_step": 147927, "epoch": 1662, "lr": 4.309257066659249e-05}
{"train_loss": 0.13246387243270874, "global_step": 147928, "epoch": 1662, "lr": 4.3091996489249485e-05}
{"train_loss": 0.13907016813755035, "global_step": 147929, "epoch": 1662, "lr": 4.3091422312835186e-05}
{"train_loss": 0.07462713122367859, "global_step": 147930, "epoch": 1662, "lr": 4.309084813734966e-05}
{"train_loss": 0.08508957177400589, "global_step": 147931, "epoch": 1662, "lr": 4.309027396279297e-05}
{"train_loss": 0.11121023446321487, "global_step": 147932, "epoch": 1662, "lr": 4.308969978916521e-05}
{"train_loss": 0.08541252464056015, "global_step": 147933, "epoch": 1662, "lr": 4.308912561646647e-05}
{"train_loss": 0.11960673332214355, "global_step": 147934, "epoch": 1662, "lr": 4.3088551444696795e-05}
{"train_loss": 0.11115202307701111, "global_step": 147935, "epoch": 1662, "lr": 4.30879772738563e-05}
{"train_loss": 0.09174572676420212, "global_step": 147936, "epoch": 1662, "lr": 4.308740310394502e-05}
{"train_loss": 0.05594778433442116, "global_step": 147937, "epoch": 1662, "lr": 4.308682893496308e-05}
{"train_loss": 0.14686992764472961, "global_step": 147938, "epoch": 1662, "lr": 4.308625476691051e-05}
{"train_loss": 0.14271298050880432, "global_step": 147939, "epoch": 1662, "lr": 4.308568059978742e-05}
{"train_loss": 0.10995201021432877, "global_step": 147940, "epoch": 1662, "lr": 4.308510643359387e-05}
{"train_loss": 0.12410665303468704, "global_step": 147941, "epoch": 1662, "lr": 4.308453226832996e-05}
{"train_loss": 0.09533018618822098, "global_step": 147942, "epoch": 1662, "lr": 4.3083958103995725e-05}
{"train_loss": 0.13368137180805206, "global_step": 147943, "epoch": 1662, "lr": 4.3083383940591296e-05}
{"train_loss": 0.104567751288414, "global_step": 147944, "epoch": 1662, "lr": 4.30828097781167e-05}
{"train_loss": 0.1715947389602661, "global_step": 147945, "epoch": 1662, "lr": 4.308223561657204e-05}
{"train_loss": 0.08001091331243515, "global_step": 147946, "epoch": 1662, "lr": 4.3081661455957386e-05}
{"train_loss": 0.09372589737176895, "global_step": 147947, "epoch": 1662, "lr": 4.308108729627283e-05}
{"train_loss": 0.07397734373807907, "global_step": 147948, "epoch": 1662, "lr": 4.308051313751843e-05}
{"train_loss": 0.12791329622268677, "global_step": 147949, "epoch": 1662, "lr": 4.3079938979694256e-05}
{"train_loss": 0.10335180908441544, "global_step": 147950, "epoch": 1662, "lr": 4.307936482280043e-05}
{"train_loss": 0.05842910334467888, "global_step": 147951, "epoch": 1662, "lr": 4.307879066683697e-05}
{"train_loss": 0.07046784460544586, "global_step": 147952, "epoch": 1662, "lr": 4.3078216511804006e-05}
{"train_loss": 0.08543726056814194, "global_step": 147953, "epoch": 1662, "lr": 4.307764235770158e-05}
{"train_loss": 0.116945281624794, "global_step": 147954, "epoch": 1662, "lr": 4.307706820452979e-05}
{"train_loss": 0.062314026057720184, "global_step": 147955, "epoch": 1662, "lr": 4.307649405228868e-05}
{"train_loss": 0.11034560203552246, "global_step": 147956, "epoch": 1662, "lr": 4.307591990097838e-05}
{"train_loss": 0.09324038028717041, "global_step": 147957, "epoch": 1662, "lr": 4.307534575059891e-05}
{"train_loss": 0.09777803719043732, "global_step": 147958, "epoch": 1662, "lr": 4.3074771601150406e-05}
{"train_loss": 0.14829079806804657, "global_step": 147959, "epoch": 1662, "lr": 4.3074197452632886e-05}
{"train_loss": 0.10692029446363449, "global_step": 147960, "epoch": 1662, "lr": 4.307362330504648e-05}
{"train_loss": 0.15835534036159515, "global_step": 147961, "epoch": 1662, "lr": 4.3073049158391224e-05}
{"train_loss": 0.05366327613592148, "global_step": 147962, "epoch": 1662, "lr": 4.3072475012667225e-05}
{"train_loss": 0.08289554715156555, "global_step": 147963, "epoch": 1662, "lr": 4.3071900867874535e-05}
{"train_loss": 0.14736640453338623, "global_step": 147964, "epoch": 1662, "lr": 4.307132672401325e-05}
{"train_loss": 0.12395536154508591, "global_step": 147965, "epoch": 1662, "lr": 4.307075258108344e-05}
{"train_loss": 0.13678644597530365, "global_step": 147966, "epoch": 1662, "lr": 4.307017843908518e-05}
{"train_loss": 0.14736655354499817, "global_step": 147967, "epoch": 1662, "lr": 4.306960429801855e-05}
{"train_loss": 0.12527871131896973, "global_step": 147968, "epoch": 1662, "lr": 4.306903015788362e-05}
{"train_loss": 0.09348490089178085, "global_step": 147969, "epoch": 1662, "lr": 4.3068456018680494e-05}
{"train_loss": 0.08548006415367126, "global_step": 147970, "epoch": 1662, "lr": 4.306788188040921e-05}
{"train_loss": 0.09710931032896042, "global_step": 147971, "epoch": 1662, "lr": 4.3067307743069884e-05}
{"train_loss": 0.08839632570743561, "global_step": 147972, "epoch": 1662, "lr": 4.306673360666256e-05}
{"train_loss": 0.06072608008980751, "global_step": 147973, "epoch": 1662, "lr": 4.306615947118734e-05}
{"train_loss": 0.054950110614299774, "global_step": 147974, "epoch": 1662, "lr": 4.3065585336644275e-05}
{"train_loss": 0.09033004939556122, "global_step": 147975, "epoch": 1662, "lr": 4.306501120303348e-05}
{"train_loss": 0.14191870391368866, "global_step": 147976, "epoch": 1662, "lr": 4.306443707035498e-05}
{"train_loss": 0.07036848366260529, "global_step": 147977, "epoch": 1662, "lr": 4.3063862938608914e-05}
{"train_loss": 0.08535727858543396, "global_step": 147978, "epoch": 1662, "lr": 4.3063288807795297e-05}
{"train_loss": 0.10029132664203644, "global_step": 147979, "epoch": 1662, "lr": 4.306271467791426e-05}
{"train_loss": 0.11786656826734543, "global_step": 147980, "epoch": 1662, "lr": 4.306214054896583e-05}
{"train_loss": 0.14349265396595, "global_step": 147981, "epoch": 1662, "lr": 4.306156642095014e-05}
{"train_loss": 0.0938555896282196, "global_step": 147982, "epoch": 1662, "lr": 4.306099229386723e-05}
{"train_loss": 0.12317977845668793, "global_step": 147983, "epoch": 1662, "lr": 4.306041816771717e-05}
{"train_loss": 0.15352120995521545, "global_step": 147984, "epoch": 1662, "lr": 4.305984404250007e-05}
{"train_loss": 0.08228178322315216, "global_step": 147985, "epoch": 1662, "lr": 4.3059269918215975e-05}
{"train_loss": 0.06717711687088013, "global_step": 147986, "epoch": 1662, "lr": 4.305869579486499e-05}
{"train_loss": 0.13328732550144196, "global_step": 147987, "epoch": 1662, "lr": 4.305812167244716e-05}
{"train_loss": 0.10887496173381805, "global_step": 147988, "epoch": 1662, "lr": 4.305754755096261e-05}
{"train_loss": 0.12723314762115479, "global_step": 147989, "epoch": 1662, "lr": 4.305697343041136e-05}
{"train_loss": 0.07163495570421219, "global_step": 147990, "epoch": 1662, "lr": 4.305639931079353e-05}
{"train_loss": 0.13407737016677856, "global_step": 147991, "epoch": 1662, "lr": 4.305582519210917e-05}
{"train_loss": 0.05519867688417435, "global_step": 147992, "epoch": 1662, "lr": 4.305525107435839e-05}
{"train_loss": 0.0853792130947113, "global_step": 147993, "epoch": 1662, "lr": 4.305467695754122e-05}
{"train_loss": 0.1609964817762375, "global_step": 147994, "epoch": 1662, "lr": 4.3054102841657795e-05}
{"train_loss": 0.14641548693180084, "global_step": 147995, "epoch": 1662, "lr": 4.3053528726708136e-05}
{"train_loss": 0.134520024061203, "global_step": 147996, "epoch": 1662, "lr": 4.305295461269236e-05}
{"train_loss": 0.12230367958545685, "global_step": 147997, "epoch": 1662, "lr": 4.305238049961052e-05}
{"train_loss": 0.1499023735523224, "global_step": 147998, "epoch": 1662, "lr": 4.3051806387462716e-05}
{"train_loss": 0.10325180739164352, "global_step": 147999, "epoch": 1662, "lr": 4.3051232276248994e-05}
{"train_loss": 0.08124342560768127, "global_step": 148000, "epoch": 1662, "lr": 4.305065816596947e-05}
{"train_loss": 0.07810749858617783, "global_step": 148001, "epoch": 1662, "lr": 4.3050084056624186e-05}
{"train_loss": 0.11261843889951706, "global_step": 148002, "epoch": 1662, "lr": 4.3049509948213235e-05}
{"train_loss": 0.12474720925092697, "global_step": 148003, "epoch": 1662, "lr": 4.304893584073669e-05}
{"train_loss": 0.11974513530731201, "global_step": 148004, "epoch": 1662, "lr": 4.304836173419463e-05}
{"train_loss": 0.12274592369794846, "global_step": 148005, "epoch": 1662, "lr": 4.304778762858715e-05}
{"train_loss": 0.10874284406223994, "global_step": 148006, "epoch": 1662, "lr": 4.3047213523914285e-05, "val_loss": 6.1872663497924805}
{"train_loss": 0.14087837934494019, "global_step": 148007, "epoch": 1663, "lr": 4.304663942017616e-05}
{"train_loss": 0.12115872651338577, "global_step": 148008, "epoch": 1663, "lr": 4.304606531737281e-05}
{"train_loss": 0.08679907768964767, "global_step": 148009, "epoch": 1663, "lr": 4.3045491215504354e-05}
{"train_loss": 0.12038283050060272, "global_step": 148010, "epoch": 1663, "lr": 4.304491711457082e-05}
{"train_loss": 0.12713775038719177, "global_step": 148011, "epoch": 1663, "lr": 4.304434301457234e-05}
{"train_loss": 0.1917625367641449, "global_step": 148012, "epoch": 1663, "lr": 4.304376891550894e-05}
{"train_loss": 0.1650519073009491, "global_step": 148013, "epoch": 1663, "lr": 4.304319481738074e-05}
{"train_loss": 0.10572177171707153, "global_step": 148014, "epoch": 1663, "lr": 4.304262072018778e-05}
{"train_loss": 0.12773965299129486, "global_step": 148015, "epoch": 1663, "lr": 4.304204662393017e-05}
{"train_loss": 0.07497432082891464, "global_step": 148016, "epoch": 1663, "lr": 4.3041472528607954e-05}
{"train_loss": 0.08854005485773087, "global_step": 148017, "epoch": 1663, "lr": 4.3040898434221247e-05}
{"train_loss": 0.15968306362628937, "global_step": 148018, "epoch": 1663, "lr": 4.304032434077009e-05}
{"train_loss": 0.11824314296245575, "global_step": 148019, "epoch": 1663, "lr": 4.3039750248254593e-05}
{"train_loss": 0.13511726260185242, "global_step": 148020, "epoch": 1663, "lr": 4.30391761566748e-05}
{"train_loss": 0.08358639478683472, "global_step": 148021, "epoch": 1663, "lr": 4.303860206603082e-05}
{"train_loss": 0.1812409609556198, "global_step": 148022, "epoch": 1663, "lr": 4.303802797632271e-05}
{"train_loss": 0.13936908543109894, "global_step": 148023, "epoch": 1663, "lr": 4.303745388755054e-05}
{"train_loss": 0.09990988671779633, "global_step": 148024, "epoch": 1663, "lr": 4.3036879799714425e-05}
{"train_loss": 0.11882252246141434, "global_step": 148025, "epoch": 1663, "lr": 4.3036305712814384e-05}
{"train_loss": 0.1509363055229187, "global_step": 148026, "epoch": 1663, "lr": 4.3035731626850565e-05}
{"train_loss": 0.13651303946971893, "global_step": 148027, "epoch": 1663, "lr": 4.303515754182297e-05}
{"train_loss": 0.14842593669891357, "global_step": 148028, "epoch": 1663, "lr": 4.3034583457731736e-05}
{"train_loss": 0.13964180648326874, "global_step": 148029, "epoch": 1663, "lr": 4.30340093745769e-05}
{"train_loss": 0.13331511616706848, "global_step": 148030, "epoch": 1663, "lr": 4.303343529235858e-05}
{"train_loss": 0.19730783998966217, "global_step": 148031, "epoch": 1663, "lr": 4.303286121107681e-05}
{"train_loss": 0.1404474377632141, "global_step": 148032, "epoch": 1663, "lr": 4.3032287130731704e-05}
{"train_loss": 0.08211102336645126, "global_step": 148033, "epoch": 1663, "lr": 4.30317130513233e-05}
{"train_loss": 0.08716997504234314, "global_step": 148034, "epoch": 1663, "lr": 4.303113897285172e-05}
{"train_loss": 0.0889301598072052, "global_step": 148035, "epoch": 1663, "lr": 4.303056489531699e-05}
{"train_loss": 0.07579179108142853, "global_step": 148036, "epoch": 1663, "lr": 4.302999081871925e-05}
{"train_loss": 0.09679628908634186, "global_step": 148037, "epoch": 1663, "lr": 4.302941674305852e-05}
{"train_loss": 0.1428011953830719, "global_step": 148038, "epoch": 1663, "lr": 4.302884266833491e-05}
{"train_loss": 0.12385773658752441, "global_step": 148039, "epoch": 1663, "lr": 4.3028268594548474e-05}
{"train_loss": 0.12512394785881042, "global_step": 148040, "epoch": 1663, "lr": 4.3027694521699313e-05}
{"train_loss": 0.12696635723114014, "global_step": 148041, "epoch": 1663, "lr": 4.30271204497875e-05}
{"train_loss": 0.09183508902788162, "global_step": 148042, "epoch": 1663, "lr": 4.302654637881309e-05}
{"train_loss": 0.04649277403950691, "global_step": 148043, "epoch": 1663, "lr": 4.302597230877619e-05}
{"train_loss": 0.07043415307998657, "global_step": 148044, "epoch": 1663, "lr": 4.3025398239676854e-05}
{"train_loss": 0.17031550407409668, "global_step": 148045, "epoch": 1663, "lr": 4.3024824171515174e-05}
{"train_loss": 0.10953490436077118, "global_step": 148046, "epoch": 1663, "lr": 4.302425010429121e-05}
{"train_loss": 0.1286848932504654, "global_step": 148047, "epoch": 1663, "lr": 4.302367603800507e-05}
{"train_loss": 0.09728415310382843, "global_step": 148048, "epoch": 1663, "lr": 4.302310197265679e-05}
{"train_loss": 0.11169282346963882, "global_step": 148049, "epoch": 1663, "lr": 4.302252790824649e-05}
{"train_loss": 0.07533655315637589, "global_step": 148050, "epoch": 1663, "lr": 4.30219538447742e-05}
{"train_loss": 0.11114250868558884, "global_step": 148051, "epoch": 1663, "lr": 4.302137978224005e-05}
{"train_loss": 0.09432639181613922, "global_step": 148052, "epoch": 1663, "lr": 4.302080572064407e-05}
{"train_loss": 0.04870413988828659, "global_step": 148053, "epoch": 1663, "lr": 4.302023165998638e-05}
{"train_loss": 0.1498122364282608, "global_step": 148054, "epoch": 1663, "lr": 4.301965760026701e-05}
{"train_loss": 0.07109975069761276, "global_step": 148055, "epoch": 1663, "lr": 4.301908354148608e-05}
{"train_loss": 0.11833042651414871, "global_step": 148056, "epoch": 1663, "lr": 4.301850948364363e-05}
{"train_loss": 0.10897994786500931, "global_step": 148057, "epoch": 1663, "lr": 4.301793542673979e-05}
{"train_loss": 0.11469340324401855, "global_step": 148058, "epoch": 1663, "lr": 4.301736137077457e-05}
{"train_loss": 0.11481700092554092, "global_step": 148059, "epoch": 1663, "lr": 4.301678731574809e-05}
{"train_loss": 0.0748760774731636, "global_step": 148060, "epoch": 1663, "lr": 4.301621326166043e-05}
{"train_loss": 0.09664243459701538, "global_step": 148061, "epoch": 1663, "lr": 4.3015639208511636e-05}
{"train_loss": 0.09557397663593292, "global_step": 148062, "epoch": 1663, "lr": 4.301506515630182e-05}
{"train_loss": 0.12996505200862885, "global_step": 148063, "epoch": 1663, "lr": 4.301449110503103e-05}
{"train_loss": 0.06397384405136108, "global_step": 148064, "epoch": 1663, "lr": 4.301391705469937e-05}
{"train_loss": 0.09229632467031479, "global_step": 148065, "epoch": 1663, "lr": 4.301334300530689e-05}
{"train_loss": 0.13374486565589905, "global_step": 148066, "epoch": 1663, "lr": 4.30127689568537e-05}
{"train_loss": 0.08737773448228836, "global_step": 148067, "epoch": 1663, "lr": 4.3012194909339835e-05}
{"train_loss": 0.05835871398448944, "global_step": 148068, "epoch": 1663, "lr": 4.301162086276542e-05}
{"train_loss": 0.07817161828279495, "global_step": 148069, "epoch": 1663, "lr": 4.3011046817130487e-05}
{"train_loss": 0.10430718958377838, "global_step": 148070, "epoch": 1663, "lr": 4.301047277243514e-05}
{"train_loss": 0.09829643368721008, "global_step": 148071, "epoch": 1663, "lr": 4.300989872867944e-05}
{"train_loss": 0.1111970990896225, "global_step": 148072, "epoch": 1663, "lr": 4.30093246858635e-05}
{"train_loss": 0.11466610431671143, "global_step": 148073, "epoch": 1663, "lr": 4.300875064398734e-05}
{"train_loss": 0.09327755123376846, "global_step": 148074, "epoch": 1663, "lr": 4.30081766030511e-05}
{"train_loss": 0.09501885622739792, "global_step": 148075, "epoch": 1663, "lr": 4.30076025630548e-05}
{"train_loss": 0.06467108428478241, "global_step": 148076, "epoch": 1663, "lr": 4.300702852399855e-05}
{"train_loss": 0.0743427649140358, "global_step": 148077, "epoch": 1663, "lr": 4.300645448588242e-05}
{"train_loss": 0.08381104469299316, "global_step": 148078, "epoch": 1663, "lr": 4.3005880448706495e-05}
{"train_loss": 0.06826300173997879, "global_step": 148079, "epoch": 1663, "lr": 4.300530641247085e-05}
{"train_loss": 0.09170744568109512, "global_step": 148080, "epoch": 1663, "lr": 4.3004732377175524e-05}
{"train_loss": 0.10305586457252502, "global_step": 148081, "epoch": 1663, "lr": 4.3004158342820664e-05}
{"train_loss": 0.11854010075330734, "global_step": 148082, "epoch": 1663, "lr": 4.3003584309406274e-05}
{"train_loss": 0.1127924770116806, "global_step": 148083, "epoch": 1663, "lr": 4.30030102769325e-05}
{"train_loss": 0.09860239177942276, "global_step": 148084, "epoch": 1663, "lr": 4.3002436245399355e-05}
{"train_loss": 0.11518285423517227, "global_step": 148085, "epoch": 1663, "lr": 4.3001862214806974e-05}
{"train_loss": 0.11064772307872772, "global_step": 148086, "epoch": 1663, "lr": 4.300128818515538e-05}
{"train_loss": 0.09173089265823364, "global_step": 148087, "epoch": 1663, "lr": 4.300071415644471e-05}
{"train_loss": 0.0513945110142231, "global_step": 148088, "epoch": 1663, "lr": 4.300014012867498e-05}
{"train_loss": 0.09590393304824829, "global_step": 148089, "epoch": 1663, "lr": 4.299956610184632e-05}
{"train_loss": 0.13792242109775543, "global_step": 148090, "epoch": 1663, "lr": 4.299899207595877e-05}
{"train_loss": 0.08219480514526367, "global_step": 148091, "epoch": 1663, "lr": 4.299841805101243e-05}
{"train_loss": 0.08367389440536499, "global_step": 148092, "epoch": 1663, "lr": 4.2997844027007345e-05}
{"train_loss": 0.17368698120117188, "global_step": 148093, "epoch": 1663, "lr": 4.2997270003943644e-05}
{"train_loss": 0.07965200394392014, "global_step": 148094, "epoch": 1663, "lr": 4.299669598182134e-05}
{"train_loss": 0.10867932572793425, "global_step": 148095, "epoch": 1663, "lr": 4.299612196064057e-05, "val_loss": 6.261978626251221}
{"train_loss": 0.08340758830308914, "global_step": 148096, "epoch": 1664, "lr": 4.2995547940401376e-05}
{"train_loss": 0.05455987527966499, "global_step": 148097, "epoch": 1664, "lr": 4.2994973921103854e-05}
{"train_loss": 0.11799941211938858, "global_step": 148098, "epoch": 1664, "lr": 4.299439990274807e-05}
{"train_loss": 0.09780599921941757, "global_step": 148099, "epoch": 1664, "lr": 4.299382588533409e-05}
{"train_loss": 0.06661161780357361, "global_step": 148100, "epoch": 1664, "lr": 4.2993251868862036e-05}
{"train_loss": 0.0845668688416481, "global_step": 148101, "epoch": 1664, "lr": 4.299267785333192e-05}
{"train_loss": 0.12122184038162231, "global_step": 148102, "epoch": 1664, "lr": 4.299210383874388e-05}
{"train_loss": 0.10056482255458832, "global_step": 148103, "epoch": 1664, "lr": 4.2991529825097934e-05}
{"train_loss": 0.11793272197246552, "global_step": 148104, "epoch": 1664, "lr": 4.299095581239423e-05}
{"train_loss": 0.10040109604597092, "global_step": 148105, "epoch": 1664, "lr": 4.2990381800632776e-05}
{"train_loss": 0.09031466394662857, "global_step": 148106, "epoch": 1664, "lr": 4.2989807789813695e-05}
{"train_loss": 0.14417259395122528, "global_step": 148107, "epoch": 1664, "lr": 4.298923377993704e-05}
{"train_loss": 0.11367929726839066, "global_step": 148108, "epoch": 1664, "lr": 4.298865977100291e-05}
{"train_loss": 0.10103920102119446, "global_step": 148109, "epoch": 1664, "lr": 4.2988085763011346e-05}
{"train_loss": 0.07485731691122055, "global_step": 148110, "epoch": 1664, "lr": 4.298751175596248e-05}
{"train_loss": 0.0711081400513649, "global_step": 148111, "epoch": 1664, "lr": 4.298693774985633e-05}
{"train_loss": 0.0846320316195488, "global_step": 148112, "epoch": 1664, "lr": 4.298636374469301e-05}
{"train_loss": 0.1717490404844284, "global_step": 148113, "epoch": 1664, "lr": 4.2985789740472585e-05}
{"train_loss": 0.088559590280056, "global_step": 148114, "epoch": 1664, "lr": 4.2985215737195147e-05}
{"train_loss": 0.09974215924739838, "global_step": 148115, "epoch": 1664, "lr": 4.298464173486075e-05}
{"train_loss": 0.06893035769462585, "global_step": 148116, "epoch": 1664, "lr": 4.298406773346948e-05}
{"train_loss": 0.09271182119846344, "global_step": 148117, "epoch": 1664, "lr": 4.298349373302144e-05}
{"train_loss": 0.0590384379029274, "global_step": 148118, "epoch": 1664, "lr": 4.298291973351666e-05}
{"train_loss": 0.07626277953386307, "global_step": 148119, "epoch": 1664, "lr": 4.2982345734955255e-05}
{"train_loss": 0.09532234072685242, "global_step": 148120, "epoch": 1664, "lr": 4.2981771737337264e-05}
{"train_loss": 0.11685428023338318, "global_step": 148121, "epoch": 1664, "lr": 4.298119774066282e-05}
{"train_loss": 0.1007397398352623, "global_step": 148122, "epoch": 1664, "lr": 4.2980623744931944e-05}
{"train_loss": 0.08284848928451538, "global_step": 148123, "epoch": 1664, "lr": 4.298004975014476e-05}
{"train_loss": 0.04356636852025986, "global_step": 148124, "epoch": 1664, "lr": 4.2979475756301294e-05}
{"train_loss": 0.10490500181913376, "global_step": 148125, "epoch": 1664, "lr": 4.2978901763401677e-05}
{"train_loss": 0.0677027627825737, "global_step": 148126, "epoch": 1664, "lr": 4.2978327771445945e-05}
{"train_loss": 0.05459760129451752, "global_step": 148127, "epoch": 1664, "lr": 4.297775378043421e-05}
{"train_loss": 0.06773363053798676, "global_step": 148128, "epoch": 1664, "lr": 4.2977179790366506e-05}
{"train_loss": 0.08738159388303757, "global_step": 148129, "epoch": 1664, "lr": 4.2976605801242954e-05}
{"train_loss": 0.09296348690986633, "global_step": 148130, "epoch": 1664, "lr": 4.2976031813063593e-05}
{"train_loss": 0.13199658691883087, "global_step": 148131, "epoch": 1664, "lr": 4.297545782582853e-05}
{"train_loss": 0.06380175054073334, "global_step": 148132, "epoch": 1664, "lr": 4.297488383953784e-05}
{"train_loss": 0.10837382823228836, "global_step": 148133, "epoch": 1664, "lr": 4.297430985419157e-05}
{"train_loss": 0.07283307611942291, "global_step": 148134, "epoch": 1664, "lr": 4.2973735869789837e-05}
{"train_loss": 0.10153777152299881, "global_step": 148135, "epoch": 1664, "lr": 4.29731618863327e-05}
{"train_loss": 0.08222576230764389, "global_step": 148136, "epoch": 1664, "lr": 4.2972587903820225e-05}
{"train_loss": 0.09645713120698929, "global_step": 148137, "epoch": 1664, "lr": 4.2972013922252505e-05}
{"train_loss": 0.14865060150623322, "global_step": 148138, "epoch": 1664, "lr": 4.297143994162962e-05}
{"train_loss": 0.10277330875396729, "global_step": 148139, "epoch": 1664, "lr": 4.2970865961951625e-05}
{"train_loss": 0.06711422652006149, "global_step": 148140, "epoch": 1664, "lr": 4.2970291983218626e-05}
{"train_loss": 0.056062519550323486, "global_step": 148141, "epoch": 1664, "lr": 4.2969718005430666e-05}
{"train_loss": 0.12300901859998703, "global_step": 148142, "epoch": 1664, "lr": 4.296914402858786e-05}
{"train_loss": 0.10426592081785202, "global_step": 148143, "epoch": 1664, "lr": 4.296857005269025e-05}
{"train_loss": 0.12262776494026184, "global_step": 148144, "epoch": 1664, "lr": 4.296799607773795e-05}
{"train_loss": 0.11540283262729645, "global_step": 148145, "epoch": 1664, "lr": 4.2967422103731e-05}
{"train_loss": 0.10331050306558609, "global_step": 148146, "epoch": 1664, "lr": 4.2966848130669504e-05}
{"train_loss": 0.05816458910703659, "global_step": 148147, "epoch": 1664, "lr": 4.296627415855352e-05}
{"train_loss": 0.08762866258621216, "global_step": 148148, "epoch": 1664, "lr": 4.296570018738315e-05}
{"train_loss": 0.0939280167222023, "global_step": 148149, "epoch": 1664, "lr": 4.2965126217158435e-05}
{"train_loss": 0.029013603925704956, "global_step": 148150, "epoch": 1664, "lr": 4.2964552247879476e-05}
{"train_loss": 0.07605994492769241, "global_step": 148151, "epoch": 1664, "lr": 4.296397827954637e-05}
{"train_loss": 0.10673235356807709, "global_step": 148152, "epoch": 1664, "lr": 4.296340431215915e-05}
{"train_loss": 0.0758286789059639, "global_step": 148153, "epoch": 1664, "lr": 4.296283034571793e-05}
{"train_loss": 0.10382747650146484, "global_step": 148154, "epoch": 1664, "lr": 4.296225638022276e-05}
{"train_loss": 0.09769908338785172, "global_step": 148155, "epoch": 1664, "lr": 4.296168241567374e-05}
{"train_loss": 0.07947461307048798, "global_step": 148156, "epoch": 1664, "lr": 4.296110845207092e-05}
{"train_loss": 0.1015431359410286, "global_step": 148157, "epoch": 1664, "lr": 4.296053448941442e-05}
{"train_loss": 0.1279517412185669, "global_step": 148158, "epoch": 1664, "lr": 4.2959960527704256e-05}
{"train_loss": 0.1292465478181839, "global_step": 148159, "epoch": 1664, "lr": 4.295938656694057e-05}
{"train_loss": 0.06978234648704529, "global_step": 148160, "epoch": 1664, "lr": 4.2958812607123386e-05}
{"train_loss": 0.09443686902523041, "global_step": 148161, "epoch": 1664, "lr": 4.2958238648252816e-05}
{"train_loss": 0.16759927570819855, "global_step": 148162, "epoch": 1664, "lr": 4.2957664690328916e-05}
{"train_loss": 0.15367788076400757, "global_step": 148163, "epoch": 1664, "lr": 4.295709073335179e-05}
{"train_loss": 0.10862903296947479, "global_step": 148164, "epoch": 1664, "lr": 4.295651677732147e-05}
{"train_loss": 0.11509200930595398, "global_step": 148165, "epoch": 1664, "lr": 4.2955942822238085e-05}
{"train_loss": 0.15630392730236053, "global_step": 148166, "epoch": 1664, "lr": 4.295536886810167e-05}
{"train_loss": 0.0922335609793663, "global_step": 148167, "epoch": 1664, "lr": 4.295479491491232e-05}
{"train_loss": 0.10701650381088257, "global_step": 148168, "epoch": 1664, "lr": 4.295422096267013e-05}
{"train_loss": 0.09117573499679565, "global_step": 148169, "epoch": 1664, "lr": 4.2953647011375134e-05}
{"train_loss": 0.06951930373907089, "global_step": 148170, "epoch": 1664, "lr": 4.2953073061027463e-05}
{"train_loss": 0.06736979633569717, "global_step": 148171, "epoch": 1664, "lr": 4.295249911162715e-05}
{"train_loss": 0.11480499804019928, "global_step": 148172, "epoch": 1664, "lr": 4.295192516317429e-05}
{"train_loss": 0.14303617179393768, "global_step": 148173, "epoch": 1664, "lr": 4.295135121566896e-05}
{"train_loss": 0.1377653181552887, "global_step": 148174, "epoch": 1664, "lr": 4.295077726911122e-05}
{"train_loss": 0.09945691376924515, "global_step": 148175, "epoch": 1664, "lr": 4.2950203323501184e-05}
{"train_loss": 0.14163149893283844, "global_step": 148176, "epoch": 1664, "lr": 4.2949629378838907e-05}
{"train_loss": 0.08463674038648605, "global_step": 148177, "epoch": 1664, "lr": 4.294905543512444e-05}
{"train_loss": 0.12252777814865112, "global_step": 148178, "epoch": 1664, "lr": 4.2948481492357916e-05}
{"train_loss": 0.08769460767507553, "global_step": 148179, "epoch": 1664, "lr": 4.294790755053936e-05}
{"train_loss": 0.07532558590173721, "global_step": 148180, "epoch": 1664, "lr": 4.2947333609668896e-05}
{"train_loss": 0.1132422536611557, "global_step": 148181, "epoch": 1664, "lr": 4.294675966974655e-05}
{"train_loss": 0.07877970486879349, "global_step": 148182, "epoch": 1664, "lr": 4.294618573077245e-05}
{"train_loss": 0.12794166803359985, "global_step": 148183, "epoch": 1664, "lr": 4.294561179274662e-05}
{"train_loss": 0.0980101064098685, "global_step": 148184, "epoch": 1664, "lr": 4.2945037855669205e-05, "val_loss": 6.1721696853637695}
{"train_loss": 0.10356198251247406, "global_step": 148185, "epoch": 1665, "lr": 4.294446391954021e-05}
{"train_loss": 0.1158650815486908, "global_step": 148186, "epoch": 1665, "lr": 4.294388998435974e-05}
{"train_loss": 0.15024298429489136, "global_step": 148187, "epoch": 1665, "lr": 4.2943316050127914e-05}
{"train_loss": 0.10460631549358368, "global_step": 148188, "epoch": 1665, "lr": 4.2942742116844746e-05}
{"train_loss": 0.12080331891775131, "global_step": 148189, "epoch": 1665, "lr": 4.294216818451036e-05}
{"train_loss": 0.14257751405239105, "global_step": 148190, "epoch": 1665, "lr": 4.2941594253124804e-05}
{"train_loss": 0.08742110431194305, "global_step": 148191, "epoch": 1665, "lr": 4.294102032268816e-05}
{"train_loss": 0.12562979757785797, "global_step": 148192, "epoch": 1665, "lr": 4.294044639320052e-05}
{"train_loss": 0.08477762341499329, "global_step": 148193, "epoch": 1665, "lr": 4.293987246466195e-05}
{"train_loss": 0.12107446789741516, "global_step": 148194, "epoch": 1665, "lr": 4.293929853707252e-05}
{"train_loss": 0.11575448513031006, "global_step": 148195, "epoch": 1665, "lr": 4.293872461043233e-05}
{"train_loss": 0.07830639183521271, "global_step": 148196, "epoch": 1665, "lr": 4.293815068474142e-05}
{"train_loss": 0.11690973490476608, "global_step": 148197, "epoch": 1665, "lr": 4.293757675999991e-05}
{"train_loss": 0.1561015546321869, "global_step": 148198, "epoch": 1665, "lr": 4.293700283620784e-05}
{"train_loss": 0.11081315577030182, "global_step": 148199, "epoch": 1665, "lr": 4.293642891336532e-05}
{"train_loss": 0.11946921050548553, "global_step": 148200, "epoch": 1665, "lr": 4.293585499147239e-05}
{"train_loss": 0.09955956041812897, "global_step": 148201, "epoch": 1665, "lr": 4.293528107052917e-05}
{"train_loss": 0.11127964407205582, "global_step": 148202, "epoch": 1665, "lr": 4.29347071505357e-05}
{"train_loss": 0.06745818257331848, "global_step": 148203, "epoch": 1665, "lr": 4.293413323149206e-05}
{"train_loss": 0.09245984256267548, "global_step": 148204, "epoch": 1665, "lr": 4.2933559313398375e-05}
{"train_loss": 0.07990410923957825, "global_step": 148205, "epoch": 1665, "lr": 4.293298539625465e-05}
{"train_loss": 0.07819335907697678, "global_step": 148206, "epoch": 1665, "lr": 4.293241148006104e-05}
{"train_loss": 0.10680247843265533, "global_step": 148207, "epoch": 1665, "lr": 4.293183756481754e-05}
{"train_loss": 0.0877104252576828, "global_step": 148208, "epoch": 1665, "lr": 4.2931263650524304e-05}
{"train_loss": 0.14611059427261353, "global_step": 148209, "epoch": 1665, "lr": 4.293068973718136e-05}
{"train_loss": 0.11637058109045029, "global_step": 148210, "epoch": 1665, "lr": 4.29301158247888e-05}
{"train_loss": 0.06334775686264038, "global_step": 148211, "epoch": 1665, "lr": 4.292954191334669e-05}
{"train_loss": 0.10184755176305771, "global_step": 148212, "epoch": 1665, "lr": 4.2928968002855125e-05}
{"train_loss": 0.12212559580802917, "global_step": 148213, "epoch": 1665, "lr": 4.2928394093314175e-05}
{"train_loss": 0.07138731330633163, "global_step": 148214, "epoch": 1665, "lr": 4.292782018472392e-05}
{"train_loss": 0.0891072154045105, "global_step": 148215, "epoch": 1665, "lr": 4.2927246277084423e-05}
{"train_loss": 0.1180747002363205, "global_step": 148216, "epoch": 1665, "lr": 4.2926672370395785e-05}
{"train_loss": 0.13839498162269592, "global_step": 148217, "epoch": 1665, "lr": 4.2926098464658046e-05}
{"train_loss": 0.05810490623116493, "global_step": 148218, "epoch": 1665, "lr": 4.2925524559871336e-05}
{"train_loss": 0.10540607571601868, "global_step": 148219, "epoch": 1665, "lr": 4.2924950656035674e-05}
{"train_loss": 0.0761382207274437, "global_step": 148220, "epoch": 1665, "lr": 4.292437675315118e-05}
{"train_loss": 0.06790484488010406, "global_step": 148221, "epoch": 1665, "lr": 4.2923802851217916e-05}
{"train_loss": 0.10728463530540466, "global_step": 148222, "epoch": 1665, "lr": 4.292322895023595e-05}
{"train_loss": 0.11432567238807678, "global_step": 148223, "epoch": 1665, "lr": 4.292265505020539e-05}
{"train_loss": 0.10262151062488556, "global_step": 148224, "epoch": 1665, "lr": 4.292208115112628e-05}
{"train_loss": 0.07891815900802612, "global_step": 148225, "epoch": 1665, "lr": 4.2921507252998724e-05}
{"train_loss": 0.07970987260341644, "global_step": 148226, "epoch": 1665, "lr": 4.2920933355822754e-05}
{"train_loss": 0.10334814339876175, "global_step": 148227, "epoch": 1665, "lr": 4.2920359459598516e-05}
{"train_loss": 0.0911388248205185, "global_step": 148228, "epoch": 1665, "lr": 4.291978556432603e-05}
{"train_loss": 0.07878465950489044, "global_step": 148229, "epoch": 1665, "lr": 4.2919211670005396e-05}
{"train_loss": 0.09848597645759583, "global_step": 148230, "epoch": 1665, "lr": 4.291863777663669e-05}
{"train_loss": 0.10651364177465439, "global_step": 148231, "epoch": 1665, "lr": 4.291806388421998e-05}
{"train_loss": 0.0850958600640297, "global_step": 148232, "epoch": 1665, "lr": 4.291748999275536e-05}
{"train_loss": 0.10899746417999268, "global_step": 148233, "epoch": 1665, "lr": 4.29169161022429e-05}
{"train_loss": 0.05705545097589493, "global_step": 148234, "epoch": 1665, "lr": 4.291634221268265e-05}
{"train_loss": 0.08563060313463211, "global_step": 148235, "epoch": 1665, "lr": 4.291576832407474e-05}
{"train_loss": 0.0646732747554779, "global_step": 148236, "epoch": 1665, "lr": 4.291519443641919e-05}
{"train_loss": 0.13200004398822784, "global_step": 148237, "epoch": 1665, "lr": 4.2914620549716135e-05}
{"train_loss": 0.13111259043216705, "global_step": 148238, "epoch": 1665, "lr": 4.29140466639656e-05}
{"train_loss": 0.11180464178323746, "global_step": 148239, "epoch": 1665, "lr": 4.291347277916768e-05}
{"train_loss": 0.09486030042171478, "global_step": 148240, "epoch": 1665, "lr": 4.291289889532249e-05}
{"train_loss": 0.13806666433811188, "global_step": 148241, "epoch": 1665, "lr": 4.291232501243004e-05}
{"train_loss": 0.09464215487241745, "global_step": 148242, "epoch": 1665, "lr": 4.291175113049047e-05}
{"train_loss": 0.08535376191139221, "global_step": 148243, "epoch": 1665, "lr": 4.29111772495038e-05}
{"train_loss": 0.08094899356365204, "global_step": 148244, "epoch": 1665, "lr": 4.2910603369470165e-05}
{"train_loss": 0.11366144567728043, "global_step": 148245, "epoch": 1665, "lr": 4.291002949038959e-05}
{"train_loss": 0.13309140503406525, "global_step": 148246, "epoch": 1665, "lr": 4.29094556122622e-05}
{"train_loss": 0.13043929636478424, "global_step": 148247, "epoch": 1665, "lr": 4.290888173508802e-05}
{"train_loss": 0.06884345412254333, "global_step": 148248, "epoch": 1665, "lr": 4.2908307858867166e-05}
{"train_loss": 0.08987922221422195, "global_step": 148249, "epoch": 1665, "lr": 4.290773398359971e-05}
{"train_loss": 0.1006050780415535, "global_step": 148250, "epoch": 1665, "lr": 4.2907160109285714e-05}
{"train_loss": 0.14837294816970825, "global_step": 148251, "epoch": 1665, "lr": 4.2906586235925275e-05}
{"train_loss": 0.10761793702840805, "global_step": 148252, "epoch": 1665, "lr": 4.2906012363518445e-05}
{"train_loss": 0.17511488497257233, "global_step": 148253, "epoch": 1665, "lr": 4.290543849206532e-05}
{"train_loss": 0.12332743406295776, "global_step": 148254, "epoch": 1665, "lr": 4.290486462156599e-05}
{"train_loss": 0.15517909824848175, "global_step": 148255, "epoch": 1665, "lr": 4.2904290752020483e-05}
{"train_loss": 0.11491435021162033, "global_step": 148256, "epoch": 1665, "lr": 4.290371688342893e-05}
{"train_loss": 0.1051056757569313, "global_step": 148257, "epoch": 1665, "lr": 4.290314301579137e-05}
{"train_loss": 0.11032941937446594, "global_step": 148258, "epoch": 1665, "lr": 4.290256914910789e-05}
{"train_loss": 0.08208470791578293, "global_step": 148259, "epoch": 1665, "lr": 4.2901995283378596e-05}
{"train_loss": 0.1308528631925583, "global_step": 148260, "epoch": 1665, "lr": 4.2901421418603526e-05}
{"train_loss": 0.089647576212883, "global_step": 148261, "epoch": 1665, "lr": 4.290084755478279e-05}
{"train_loss": 0.16938906908035278, "global_step": 148262, "epoch": 1665, "lr": 4.2900273691916426e-05}
{"train_loss": 0.1039295494556427, "global_step": 148263, "epoch": 1665, "lr": 4.289969983000456e-05}
{"train_loss": 0.10811102390289307, "global_step": 148264, "epoch": 1665, "lr": 4.2899125969047216e-05}
{"train_loss": 0.10226459801197052, "global_step": 148265, "epoch": 1665, "lr": 4.2898552109044526e-05}
{"train_loss": 0.0904214084148407, "global_step": 148266, "epoch": 1665, "lr": 4.289797824999651e-05}
{"train_loss": 0.09298735111951828, "global_step": 148267, "epoch": 1665, "lr": 4.28974043919033e-05}
{"train_loss": 0.06657230854034424, "global_step": 148268, "epoch": 1665, "lr": 4.289683053476494e-05}
{"train_loss": 0.07773888856172562, "global_step": 148269, "epoch": 1665, "lr": 4.289625667858151e-05}
{"train_loss": 0.1036103367805481, "global_step": 148270, "epoch": 1665, "lr": 4.289568282335309e-05}
{"train_loss": 0.06657977402210236, "global_step": 148271, "epoch": 1665, "lr": 4.289510896907977e-05}
{"train_loss": 0.07570141553878784, "global_step": 148272, "epoch": 1665, "lr": 4.2894535115761604e-05}
{"train_loss": 0.10347336183270711, "global_step": 148273, "epoch": 1665, "lr": 4.289396126339869e-05, "val_loss": 6.434586048126221, "train_action_mse_error": 21.120357513427734}
{"train_loss": 0.10196233540773392, "global_step": 148274, "epoch": 1666, "lr": 4.289338741199108e-05}
{"train_loss": 0.15074311196804047, "global_step": 148275, "epoch": 1666, "lr": 4.289281356153889e-05}
{"train_loss": 0.10021590441465378, "global_step": 148276, "epoch": 1666, "lr": 4.2892239712042156e-05}
{"train_loss": 0.07493077218532562, "global_step": 148277, "epoch": 1666, "lr": 4.289166586350097e-05}
{"train_loss": 0.08687350898981094, "global_step": 148278, "epoch": 1666, "lr": 4.2891092015915435e-05}
{"train_loss": 0.08874057233333588, "global_step": 148279, "epoch": 1666, "lr": 4.289051816928558e-05}
{"train_loss": 0.12235903739929199, "global_step": 148280, "epoch": 1666, "lr": 4.288994432361154e-05}
{"train_loss": 0.1083841547369957, "global_step": 148281, "epoch": 1666, "lr": 4.288937047889333e-05}
{"train_loss": 0.10293316096067429, "global_step": 148282, "epoch": 1666, "lr": 4.288879663513108e-05}
{"train_loss": 0.13313104212284088, "global_step": 148283, "epoch": 1666, "lr": 4.2888222792324825e-05}
{"train_loss": 0.17714254558086395, "global_step": 148284, "epoch": 1666, "lr": 4.288764895047468e-05}
{"train_loss": 0.1618281602859497, "global_step": 148285, "epoch": 1666, "lr": 4.288707510958069e-05}
{"train_loss": 0.06428196281194687, "global_step": 148286, "epoch": 1666, "lr": 4.288650126964296e-05}
{"train_loss": 0.06896175444126129, "global_step": 148287, "epoch": 1666, "lr": 4.2885927430661545e-05}
{"train_loss": 0.10170770436525345, "global_step": 148288, "epoch": 1666, "lr": 4.288535359263653e-05}
{"train_loss": 0.08927636593580246, "global_step": 148289, "epoch": 1666, "lr": 4.288477975556799e-05}
{"train_loss": 0.06567880511283875, "global_step": 148290, "epoch": 1666, "lr": 4.288420591945601e-05}
{"train_loss": 0.08179128170013428, "global_step": 148291, "epoch": 1666, "lr": 4.2883632084300654e-05}
{"train_loss": 0.1865638941526413, "global_step": 148292, "epoch": 1666, "lr": 4.2883058250102024e-05}
{"train_loss": 0.10248355567455292, "global_step": 148293, "epoch": 1666, "lr": 4.288248441686016e-05}
{"train_loss": 0.07638895511627197, "global_step": 148294, "epoch": 1666, "lr": 4.288191058457516e-05}
{"train_loss": 0.1542455404996872, "global_step": 148295, "epoch": 1666, "lr": 4.288133675324711e-05}
{"train_loss": 0.11889608204364777, "global_step": 148296, "epoch": 1666, "lr": 4.2880762922876075e-05}
{"train_loss": 0.09301631152629852, "global_step": 148297, "epoch": 1666, "lr": 4.2880189093462144e-05}
{"train_loss": 0.18231111764907837, "global_step": 148298, "epoch": 1666, "lr": 4.287961526500536e-05}
{"train_loss": 0.09671152383089066, "global_step": 148299, "epoch": 1666, "lr": 4.287904143750585e-05}
{"train_loss": 0.14826808869838715, "global_step": 148300, "epoch": 1666, "lr": 4.2878467610963655e-05}
{"train_loss": 0.07376409322023392, "global_step": 148301, "epoch": 1666, "lr": 4.287789378537887e-05}
{"train_loss": 0.07703863829374313, "global_step": 148302, "epoch": 1666, "lr": 4.287731996075155e-05}
{"train_loss": 0.05399209260940552, "global_step": 148303, "epoch": 1666, "lr": 4.287674613708181e-05}
{"train_loss": 0.07748278230428696, "global_step": 148304, "epoch": 1666, "lr": 4.2876172314369676e-05}
{"train_loss": 0.13121137022972107, "global_step": 148305, "epoch": 1666, "lr": 4.287559849261529e-05}
{"train_loss": 0.0690370723605156, "global_step": 148306, "epoch": 1666, "lr": 4.287502467181867e-05}
{"train_loss": 0.07434001564979553, "global_step": 148307, "epoch": 1666, "lr": 4.287445085197992e-05}
{"train_loss": 0.07787568867206573, "global_step": 148308, "epoch": 1666, "lr": 4.2873877033099113e-05}
{"train_loss": 0.057113394141197205, "global_step": 148309, "epoch": 1666, "lr": 4.287330321517632e-05}
{"train_loss": 0.12494876980781555, "global_step": 148310, "epoch": 1666, "lr": 4.2872729398211635e-05}
{"train_loss": 0.09374973177909851, "global_step": 148311, "epoch": 1666, "lr": 4.287215558220513e-05}
{"train_loss": 0.11288028210401535, "global_step": 148312, "epoch": 1666, "lr": 4.287158176715686e-05}
{"train_loss": 0.10788998007774353, "global_step": 148313, "epoch": 1666, "lr": 4.287100795306691e-05}
{"train_loss": 0.08977555483579636, "global_step": 148314, "epoch": 1666, "lr": 4.28704341399354e-05}
{"train_loss": 0.07191885262727737, "global_step": 148315, "epoch": 1666, "lr": 4.286986032776234e-05}
{"train_loss": 0.06709912419319153, "global_step": 148316, "epoch": 1666, "lr": 4.2869286516547865e-05}
{"train_loss": 0.07075905054807663, "global_step": 148317, "epoch": 1666, "lr": 4.286871270629201e-05}
{"train_loss": 0.16971473395824432, "global_step": 148318, "epoch": 1666, "lr": 4.286813889699489e-05}
{"train_loss": 0.10015974938869476, "global_step": 148319, "epoch": 1666, "lr": 4.286756508865653e-05}
{"train_loss": 0.08783486485481262, "global_step": 148320, "epoch": 1666, "lr": 4.286699128127708e-05}
{"train_loss": 0.10361750423908234, "global_step": 148321, "epoch": 1666, "lr": 4.286641747485654e-05}
{"train_loss": 0.06437835842370987, "global_step": 148322, "epoch": 1666, "lr": 4.286584366939505e-05}
{"train_loss": 0.09344998002052307, "global_step": 148323, "epoch": 1666, "lr": 4.2865269864892637e-05}
{"train_loss": 0.0911417007446289, "global_step": 148324, "epoch": 1666, "lr": 4.286469606134942e-05}
{"train_loss": 0.04951360076665878, "global_step": 148325, "epoch": 1666, "lr": 4.286412225876546e-05}
{"train_loss": 0.08605480194091797, "global_step": 148326, "epoch": 1666, "lr": 4.286354845714082e-05}
{"train_loss": 0.10774602741003036, "global_step": 148327, "epoch": 1666, "lr": 4.286297465647559e-05}
{"train_loss": 0.12087354063987732, "global_step": 148328, "epoch": 1666, "lr": 4.2862400856769847e-05}
{"train_loss": 0.09080673009157181, "global_step": 148329, "epoch": 1666, "lr": 4.2861827058023665e-05}
{"train_loss": 0.11859472095966339, "global_step": 148330, "epoch": 1666, "lr": 4.286125326023712e-05}
{"train_loss": 0.092713363468647, "global_step": 148331, "epoch": 1666, "lr": 4.2860679463410306e-05}
{"train_loss": 0.10344907641410828, "global_step": 148332, "epoch": 1666, "lr": 4.286010566754327e-05}
{"train_loss": 0.12086546421051025, "global_step": 148333, "epoch": 1666, "lr": 4.285953187263613e-05}
{"train_loss": 0.09884776175022125, "global_step": 148334, "epoch": 1666, "lr": 4.28589580786889e-05}
{"train_loss": 0.08217453211545944, "global_step": 148335, "epoch": 1666, "lr": 4.285838428570174e-05}
{"train_loss": 0.14376090466976166, "global_step": 148336, "epoch": 1666, "lr": 4.285781049367465e-05}
{"train_loss": 0.07644368708133698, "global_step": 148337, "epoch": 1666, "lr": 4.2857236702607764e-05}
{"train_loss": 0.07108978182077408, "global_step": 148338, "epoch": 1666, "lr": 4.285666291250111e-05}
{"train_loss": 0.09548837691545486, "global_step": 148339, "epoch": 1666, "lr": 4.285608912335482e-05}
{"train_loss": 0.10095978528261185, "global_step": 148340, "epoch": 1666, "lr": 4.2855515335168915e-05}
{"train_loss": 0.09317398071289062, "global_step": 148341, "epoch": 1666, "lr": 4.285494154794352e-05}
{"train_loss": 0.12333479523658752, "global_step": 148342, "epoch": 1666, "lr": 4.285436776167867e-05}
{"train_loss": 0.1004251167178154, "global_step": 148343, "epoch": 1666, "lr": 4.285379397637449e-05}
{"train_loss": 0.08478569984436035, "global_step": 148344, "epoch": 1666, "lr": 4.2853220192031015e-05}
{"train_loss": 0.10661925375461578, "global_step": 148345, "epoch": 1666, "lr": 4.285264640864834e-05}
{"train_loss": 0.07937111705541611, "global_step": 148346, "epoch": 1666, "lr": 4.2852072626226534e-05}
{"train_loss": 0.06968317925930023, "global_step": 148347, "epoch": 1666, "lr": 4.2851498844765687e-05}
{"train_loss": 0.09177554398775101, "global_step": 148348, "epoch": 1666, "lr": 4.285092506426586e-05}
{"train_loss": 0.11806771904230118, "global_step": 148349, "epoch": 1666, "lr": 4.2850351284727135e-05}
{"train_loss": 0.10124393552541733, "global_step": 148350, "epoch": 1666, "lr": 4.2849777506149614e-05}
{"train_loss": 0.11915100365877151, "global_step": 148351, "epoch": 1666, "lr": 4.284920372853333e-05}
{"train_loss": 0.16790048778057098, "global_step": 148352, "epoch": 1666, "lr": 4.28486299518784e-05}
{"train_loss": 0.12696573138237, "global_step": 148353, "epoch": 1666, "lr": 4.284805617618487e-05}
{"train_loss": 0.08274379372596741, "global_step": 148354, "epoch": 1666, "lr": 4.2847482401452854e-05}
{"train_loss": 0.1159011721611023, "global_step": 148355, "epoch": 1666, "lr": 4.284690862768239e-05}
{"train_loss": 0.055741600692272186, "global_step": 148356, "epoch": 1666, "lr": 4.284633485487358e-05}
{"train_loss": 0.17193089425563812, "global_step": 148357, "epoch": 1666, "lr": 4.2845761083026484e-05}
{"train_loss": 0.12165471911430359, "global_step": 148358, "epoch": 1666, "lr": 4.28451873121412e-05}
{"train_loss": 0.07328531891107559, "global_step": 148359, "epoch": 1666, "lr": 4.2844613542217775e-05}
{"train_loss": 0.09787878394126892, "global_step": 148360, "epoch": 1666, "lr": 4.284403977325633e-05}
{"train_loss": 0.04595349356532097, "global_step": 148361, "epoch": 1666, "lr": 4.284346600525689e-05}
{"train_loss": 0.10038976647545782, "global_step": 148362, "epoch": 1666, "lr": 4.284289223821958e-05, "val_loss": 6.513915061950684}
{"train_loss": 0.10701416432857513, "global_step": 148363, "epoch": 1667, "lr": 4.284231847214445e-05}
{"train_loss": 0.09824244678020477, "global_step": 148364, "epoch": 1667, "lr": 4.284174470703158e-05}
{"train_loss": 0.13285699486732483, "global_step": 148365, "epoch": 1667, "lr": 4.284117094288105e-05}
{"train_loss": 0.06935858726501465, "global_step": 148366, "epoch": 1667, "lr": 4.284059717969293e-05}
{"train_loss": 0.1021178737282753, "global_step": 148367, "epoch": 1667, "lr": 4.2840023417467316e-05}
{"train_loss": 0.1401432901620865, "global_step": 148368, "epoch": 1667, "lr": 4.283944965620426e-05}
{"train_loss": 0.0973634421825409, "global_step": 148369, "epoch": 1667, "lr": 4.2838875895903874e-05}
{"train_loss": 0.11619885265827179, "global_step": 148370, "epoch": 1667, "lr": 4.283830213656619e-05}
{"train_loss": 0.1301458775997162, "global_step": 148371, "epoch": 1667, "lr": 4.2837728378191336e-05}
{"train_loss": 0.107799232006073, "global_step": 148372, "epoch": 1667, "lr": 4.283715462077933e-05}
{"train_loss": 0.08136972039937973, "global_step": 148373, "epoch": 1667, "lr": 4.2836580864330317e-05}
{"train_loss": 0.11553582549095154, "global_step": 148374, "epoch": 1667, "lr": 4.28360071088443e-05}
{"train_loss": 0.11646337062120438, "global_step": 148375, "epoch": 1667, "lr": 4.283543335432143e-05}
{"train_loss": 0.06739248335361481, "global_step": 148376, "epoch": 1667, "lr": 4.283485960076172e-05}
{"train_loss": 0.06301374733448029, "global_step": 148377, "epoch": 1667, "lr": 4.2834285848165305e-05}
{"train_loss": 0.10395965725183487, "global_step": 148378, "epoch": 1667, "lr": 4.283371209653221e-05}
{"train_loss": 0.1266132891178131, "global_step": 148379, "epoch": 1667, "lr": 4.283313834586255e-05}
{"train_loss": 0.16705304384231567, "global_step": 148380, "epoch": 1667, "lr": 4.2832564596156366e-05}
{"train_loss": 0.13816562294960022, "global_step": 148381, "epoch": 1667, "lr": 4.2831990847413785e-05}
{"train_loss": 0.17081202566623688, "global_step": 148382, "epoch": 1667, "lr": 4.283141709963483e-05}
{"train_loss": 0.10649606585502625, "global_step": 148383, "epoch": 1667, "lr": 4.283084335281962e-05}
{"train_loss": 0.09900796413421631, "global_step": 148384, "epoch": 1667, "lr": 4.283026960696821e-05}
{"train_loss": 0.11490900814533234, "global_step": 148385, "epoch": 1667, "lr": 4.282969586208067e-05}
{"train_loss": 0.13991495966911316, "global_step": 148386, "epoch": 1667, "lr": 4.282912211815712e-05}
{"train_loss": 0.0901324450969696, "global_step": 148387, "epoch": 1667, "lr": 4.282854837519758e-05}
{"train_loss": 0.06250151991844177, "global_step": 148388, "epoch": 1667, "lr": 4.282797463320217e-05}
{"train_loss": 0.17399446666240692, "global_step": 148389, "epoch": 1667, "lr": 4.282740089217093e-05}
{"train_loss": 0.120289646089077, "global_step": 148390, "epoch": 1667, "lr": 4.2826827152103985e-05}
{"train_loss": 0.13180401921272278, "global_step": 148391, "epoch": 1667, "lr": 4.2826253413001364e-05}
{"train_loss": 0.1797083169221878, "global_step": 148392, "epoch": 1667, "lr": 4.282567967486319e-05}
{"train_loss": 0.1261083483695984, "global_step": 148393, "epoch": 1667, "lr": 4.282510593768949e-05}
{"train_loss": 0.14257225394248962, "global_step": 148394, "epoch": 1667, "lr": 4.28245322014804e-05}
{"train_loss": 0.1417129933834076, "global_step": 148395, "epoch": 1667, "lr": 4.2823958466235923e-05}
{"train_loss": 0.13835833966732025, "global_step": 148396, "epoch": 1667, "lr": 4.282338473195621e-05}
{"train_loss": 0.07885593175888062, "global_step": 148397, "epoch": 1667, "lr": 4.282281099864128e-05}
{"train_loss": 0.11575628072023392, "global_step": 148398, "epoch": 1667, "lr": 4.282223726629126e-05}
{"train_loss": 0.07355807721614838, "global_step": 148399, "epoch": 1667, "lr": 4.282166353490618e-05}
{"train_loss": 0.11656259000301361, "global_step": 148400, "epoch": 1667, "lr": 4.282108980448616e-05}
{"train_loss": 0.05745904892683029, "global_step": 148401, "epoch": 1667, "lr": 4.282051607503124e-05}
{"train_loss": 0.10856815427541733, "global_step": 148402, "epoch": 1667, "lr": 4.281994234654153e-05}
{"train_loss": 0.12295249849557877, "global_step": 148403, "epoch": 1667, "lr": 4.281936861901708e-05}
{"train_loss": 0.11458194255828857, "global_step": 148404, "epoch": 1667, "lr": 4.2818794892457984e-05}
{"train_loss": 0.12734565138816833, "global_step": 148405, "epoch": 1667, "lr": 4.281822116686431e-05}
{"train_loss": 0.11712184548377991, "global_step": 148406, "epoch": 1667, "lr": 4.2817647442236136e-05}
{"train_loss": 0.057626254856586456, "global_step": 148407, "epoch": 1667, "lr": 4.281707371857355e-05}
{"train_loss": 0.14599263668060303, "global_step": 148408, "epoch": 1667, "lr": 4.2816499995876616e-05}
{"train_loss": 0.05817576125264168, "global_step": 148409, "epoch": 1667, "lr": 4.281592627414542e-05}
{"train_loss": 0.07152026891708374, "global_step": 148410, "epoch": 1667, "lr": 4.281535255338002e-05}
{"train_loss": 0.08506280183792114, "global_step": 148411, "epoch": 1667, "lr": 4.2814778833580535e-05}
{"train_loss": 0.11001919955015182, "global_step": 148412, "epoch": 1667, "lr": 4.281420511474699e-05}
{"train_loss": 0.08881661295890808, "global_step": 148413, "epoch": 1667, "lr": 4.281363139687951e-05}
{"train_loss": 0.12049802392721176, "global_step": 148414, "epoch": 1667, "lr": 4.281305767997812e-05}
{"train_loss": 0.10350191593170166, "global_step": 148415, "epoch": 1667, "lr": 4.281248396404296e-05}
{"train_loss": 0.1154240295290947, "global_step": 148416, "epoch": 1667, "lr": 4.281191024907405e-05}
{"train_loss": 0.05298880487680435, "global_step": 148417, "epoch": 1667, "lr": 4.281133653507151e-05}
{"train_loss": 0.08772186189889908, "global_step": 148418, "epoch": 1667, "lr": 4.2810762822035375e-05}
{"train_loss": 0.05554882064461708, "global_step": 148419, "epoch": 1667, "lr": 4.281018910996577e-05}
{"train_loss": 0.10256090760231018, "global_step": 148420, "epoch": 1667, "lr": 4.2809615398862724e-05}
{"train_loss": 0.12187838554382324, "global_step": 148421, "epoch": 1667, "lr": 4.280904168872635e-05}
{"train_loss": 0.10830467194318771, "global_step": 148422, "epoch": 1667, "lr": 4.280846797955672e-05}
{"train_loss": 0.06240379065275192, "global_step": 148423, "epoch": 1667, "lr": 4.280789427135389e-05}
{"train_loss": 0.10278499871492386, "global_step": 148424, "epoch": 1667, "lr": 4.280732056411797e-05}
{"train_loss": 0.12092079222202301, "global_step": 148425, "epoch": 1667, "lr": 4.280674685784899e-05}
{"train_loss": 0.1053331196308136, "global_step": 148426, "epoch": 1667, "lr": 4.280617315254708e-05}
{"train_loss": 0.06750737130641937, "global_step": 148427, "epoch": 1667, "lr": 4.280559944821227e-05}
{"train_loss": 0.109468474984169, "global_step": 148428, "epoch": 1667, "lr": 4.2805025744844686e-05}
{"train_loss": 0.0893835574388504, "global_step": 148429, "epoch": 1667, "lr": 4.280445204244435e-05}
{"train_loss": 0.07929041981697083, "global_step": 148430, "epoch": 1667, "lr": 4.28038783410114e-05}
{"train_loss": 0.08841785788536072, "global_step": 148431, "epoch": 1667, "lr": 4.2803304640545846e-05}
{"train_loss": 0.07051054388284683, "global_step": 148432, "epoch": 1667, "lr": 4.280273094104782e-05}
{"train_loss": 0.07359591126441956, "global_step": 148433, "epoch": 1667, "lr": 4.280215724251737e-05}
{"train_loss": 0.0740923210978508, "global_step": 148434, "epoch": 1667, "lr": 4.28015835449546e-05}
{"train_loss": 0.06712254881858826, "global_step": 148435, "epoch": 1667, "lr": 4.280100984835955e-05}
{"train_loss": 0.1408206671476364, "global_step": 148436, "epoch": 1667, "lr": 4.280043615273233e-05}
{"train_loss": 0.07347076386213303, "global_step": 148437, "epoch": 1667, "lr": 4.2799862458072985e-05}
{"train_loss": 0.09771517664194107, "global_step": 148438, "epoch": 1667, "lr": 4.2799288764381636e-05}
{"train_loss": 0.13336357474327087, "global_step": 148439, "epoch": 1667, "lr": 4.279871507165831e-05}
{"train_loss": 0.06720034778118134, "global_step": 148440, "epoch": 1667, "lr": 4.2798141379903125e-05}
{"train_loss": 0.11351320892572403, "global_step": 148441, "epoch": 1667, "lr": 4.2797567689116144e-05}
{"train_loss": 0.14546482264995575, "global_step": 148442, "epoch": 1667, "lr": 4.279699399929743e-05}
{"train_loss": 0.13094238936901093, "global_step": 148443, "epoch": 1667, "lr": 4.2796420310447086e-05}
{"train_loss": 0.11650664359331131, "global_step": 148444, "epoch": 1667, "lr": 4.2795846622565164e-05}
{"train_loss": 0.10653457790613174, "global_step": 148445, "epoch": 1667, "lr": 4.279527293565177e-05}
{"train_loss": 0.09615789353847504, "global_step": 148446, "epoch": 1667, "lr": 4.2794699249706935e-05}
{"train_loss": 0.148087739944458, "global_step": 148447, "epoch": 1667, "lr": 4.279412556473079e-05}
{"train_loss": 0.12864957749843597, "global_step": 148448, "epoch": 1667, "lr": 4.279355188072337e-05}
{"train_loss": 0.10203012079000473, "global_step": 148449, "epoch": 1667, "lr": 4.2792978197684786e-05}
{"train_loss": 0.11826273053884506, "global_step": 148450, "epoch": 1667, "lr": 4.2792404515615076e-05}
{"train_loss": 0.10685440194740724, "global_step": 148451, "epoch": 1667, "lr": 4.2791830834514364e-05, "val_loss": 6.223662376403809}
{"train_loss": 0.1346115618944168, "global_step": 148452, "epoch": 1668, "lr": 4.279125715438268e-05}
{"train_loss": 0.07621820271015167, "global_step": 148453, "epoch": 1668, "lr": 4.279068347522015e-05}
{"train_loss": 0.1027592346072197, "global_step": 148454, "epoch": 1668, "lr": 4.279010979702679e-05}
{"train_loss": 0.1578904539346695, "global_step": 148455, "epoch": 1668, "lr": 4.278953611980274e-05}
{"train_loss": 0.12878362834453583, "global_step": 148456, "epoch": 1668, "lr": 4.278896244354803e-05}
{"train_loss": 0.10205238312482834, "global_step": 148457, "epoch": 1668, "lr": 4.2788388768262765e-05}
{"train_loss": 0.102200947701931, "global_step": 148458, "epoch": 1668, "lr": 4.278781509394702e-05}
{"train_loss": 0.06733977049589157, "global_step": 148459, "epoch": 1668, "lr": 4.278724142060086e-05}
{"train_loss": 0.06847874075174332, "global_step": 148460, "epoch": 1668, "lr": 4.278666774822438e-05}
{"train_loss": 0.07114246487617493, "global_step": 148461, "epoch": 1668, "lr": 4.278609407681762e-05}
{"train_loss": 0.04331877455115318, "global_step": 148462, "epoch": 1668, "lr": 4.27855204063807e-05}
{"train_loss": 0.08146266639232635, "global_step": 148463, "epoch": 1668, "lr": 4.278494673691366e-05}
{"train_loss": 0.10503287613391876, "global_step": 148464, "epoch": 1668, "lr": 4.278437306841662e-05}
{"train_loss": 0.10243391990661621, "global_step": 148465, "epoch": 1668, "lr": 4.2783799400889614e-05}
{"train_loss": 0.06447960436344147, "global_step": 148466, "epoch": 1668, "lr": 4.2783225734332756e-05}
{"train_loss": 0.11479300260543823, "global_step": 148467, "epoch": 1668, "lr": 4.278265206874608e-05}
{"train_loss": 0.1429784595966339, "global_step": 148468, "epoch": 1668, "lr": 4.278207840412972e-05}
{"train_loss": 0.07368077337741852, "global_step": 148469, "epoch": 1668, "lr": 4.278150474048369e-05}
{"train_loss": 0.05065879225730896, "global_step": 148470, "epoch": 1668, "lr": 4.278093107780813e-05}
{"train_loss": 0.13704194128513336, "global_step": 148471, "epoch": 1668, "lr": 4.2780357416103055e-05}
{"train_loss": 0.06394340097904205, "global_step": 148472, "epoch": 1668, "lr": 4.2779783755368594e-05}
{"train_loss": 0.09250318259000778, "global_step": 148473, "epoch": 1668, "lr": 4.277921009560479e-05}
{"train_loss": 0.10390662401914597, "global_step": 148474, "epoch": 1668, "lr": 4.277863643681175e-05}
{"train_loss": 0.0826483964920044, "global_step": 148475, "epoch": 1668, "lr": 4.2778062778989516e-05}
{"train_loss": 0.07051859050989151, "global_step": 148476, "epoch": 1668, "lr": 4.2777489122138185e-05}
{"train_loss": 0.09145155549049377, "global_step": 148477, "epoch": 1668, "lr": 4.2776915466257846e-05}
{"train_loss": 0.0831899344921112, "global_step": 148478, "epoch": 1668, "lr": 4.277634181134855e-05}
{"train_loss": 0.11364215612411499, "global_step": 148479, "epoch": 1668, "lr": 4.27757681574104e-05}
{"train_loss": 0.07962996512651443, "global_step": 148480, "epoch": 1668, "lr": 4.277519450444345e-05}
{"train_loss": 0.12606287002563477, "global_step": 148481, "epoch": 1668, "lr": 4.27746208524478e-05}
{"train_loss": 0.1542242467403412, "global_step": 148482, "epoch": 1668, "lr": 4.277404720142349e-05}
{"train_loss": 0.05413602292537689, "global_step": 148483, "epoch": 1668, "lr": 4.277347355137065e-05}
{"train_loss": 0.05757981166243553, "global_step": 148484, "epoch": 1668, "lr": 4.27728999022893e-05}
{"train_loss": 0.04973124340176582, "global_step": 148485, "epoch": 1668, "lr": 4.2772326254179566e-05}
{"train_loss": 0.0609564483165741, "global_step": 148486, "epoch": 1668, "lr": 4.277175260704149e-05}
{"train_loss": 0.102351114153862, "global_step": 148487, "epoch": 1668, "lr": 4.277117896087518e-05}
{"train_loss": 0.10563302040100098, "global_step": 148488, "epoch": 1668, "lr": 4.2770605315680676e-05}
{"train_loss": 0.1499868929386139, "global_step": 148489, "epoch": 1668, "lr": 4.27700316714581e-05}
{"train_loss": 0.07375053316354752, "global_step": 148490, "epoch": 1668, "lr": 4.276945802820748e-05}
{"train_loss": 0.1604437530040741, "global_step": 148491, "epoch": 1668, "lr": 4.2768884385928945e-05}
{"train_loss": 0.09941326826810837, "global_step": 148492, "epoch": 1668, "lr": 4.276831074462252e-05}
{"train_loss": 0.08222444355487823, "global_step": 148493, "epoch": 1668, "lr": 4.276773710428831e-05}
{"train_loss": 0.18507178127765656, "global_step": 148494, "epoch": 1668, "lr": 4.276716346492641e-05}
{"train_loss": 0.0923420786857605, "global_step": 148495, "epoch": 1668, "lr": 4.2766589826536866e-05}
{"train_loss": 0.06244432553648949, "global_step": 148496, "epoch": 1668, "lr": 4.2766016189119764e-05}
{"train_loss": 0.09879432618618011, "global_step": 148497, "epoch": 1668, "lr": 4.2765442552675185e-05}
{"train_loss": 0.13177019357681274, "global_step": 148498, "epoch": 1668, "lr": 4.27648689172032e-05}
{"train_loss": 0.06757806241512299, "global_step": 148499, "epoch": 1668, "lr": 4.27642952827039e-05}
{"train_loss": 0.22281789779663086, "global_step": 148500, "epoch": 1668, "lr": 4.2763721649177356e-05}
{"train_loss": 0.12230224907398224, "global_step": 148501, "epoch": 1668, "lr": 4.276314801662362e-05}
{"train_loss": 0.09590555727481842, "global_step": 148502, "epoch": 1668, "lr": 4.276257438504281e-05}
{"train_loss": 0.13590918481349945, "global_step": 148503, "epoch": 1668, "lr": 4.276200075443497e-05}
{"train_loss": 0.12722979485988617, "global_step": 148504, "epoch": 1668, "lr": 4.27614271248002e-05}
{"train_loss": 0.07460558414459229, "global_step": 148505, "epoch": 1668, "lr": 4.276085349613856e-05}
{"train_loss": 0.03128800541162491, "global_step": 148506, "epoch": 1668, "lr": 4.276027986845015e-05}
{"train_loss": 0.08559777587652206, "global_step": 148507, "epoch": 1668, "lr": 4.275970624173501e-05}
{"train_loss": 0.09590984135866165, "global_step": 148508, "epoch": 1668, "lr": 4.2759132615993255e-05}
{"train_loss": 0.11964727938175201, "global_step": 148509, "epoch": 1668, "lr": 4.275855899122494e-05}
{"train_loss": 0.15462325513362885, "global_step": 148510, "epoch": 1668, "lr": 4.275798536743016e-05}
{"train_loss": 0.09549941122531891, "global_step": 148511, "epoch": 1668, "lr": 4.275741174460895e-05}
{"train_loss": 0.09585698693990707, "global_step": 148512, "epoch": 1668, "lr": 4.2756838122761435e-05}
{"train_loss": 0.07900316268205643, "global_step": 148513, "epoch": 1668, "lr": 4.2756264501887686e-05}
{"train_loss": 0.073260597884655, "global_step": 148514, "epoch": 1668, "lr": 4.275569088198776e-05}
{"train_loss": 0.112945057451725, "global_step": 148515, "epoch": 1668, "lr": 4.275511726306175e-05}
{"train_loss": 0.0653332844376564, "global_step": 148516, "epoch": 1668, "lr": 4.275454364510972e-05}
{"train_loss": 0.1155964732170105, "global_step": 148517, "epoch": 1668, "lr": 4.2753970028131755e-05}
{"train_loss": 0.148050457239151, "global_step": 148518, "epoch": 1668, "lr": 4.275339641212793e-05}
{"train_loss": 0.13012875616550446, "global_step": 148519, "epoch": 1668, "lr": 4.275282279709834e-05}
{"train_loss": 0.121821828186512, "global_step": 148520, "epoch": 1668, "lr": 4.275224918304301e-05}
{"train_loss": 0.0913226529955864, "global_step": 148521, "epoch": 1668, "lr": 4.275167556996208e-05}
{"train_loss": 0.11492893844842911, "global_step": 148522, "epoch": 1668, "lr": 4.2751101957855574e-05}
{"train_loss": 0.10629160702228546, "global_step": 148523, "epoch": 1668, "lr": 4.275052834672362e-05}
{"train_loss": 0.1437416523694992, "global_step": 148524, "epoch": 1668, "lr": 4.274995473656625e-05}
{"train_loss": 0.15140005946159363, "global_step": 148525, "epoch": 1668, "lr": 4.274938112738357e-05}
{"train_loss": 0.0822741761803627, "global_step": 148526, "epoch": 1668, "lr": 4.274880751917564e-05}
{"train_loss": 0.07593617588281631, "global_step": 148527, "epoch": 1668, "lr": 4.2748233911942555e-05}
{"train_loss": 0.08778499066829681, "global_step": 148528, "epoch": 1668, "lr": 4.2747660305684366e-05}
{"train_loss": 0.08856485038995743, "global_step": 148529, "epoch": 1668, "lr": 4.274708670040119e-05}
{"train_loss": 0.15859276056289673, "global_step": 148530, "epoch": 1668, "lr": 4.274651309609305e-05}
{"train_loss": 0.06462268531322479, "global_step": 148531, "epoch": 1668, "lr": 4.2745939492760055e-05}
{"train_loss": 0.11262267082929611, "global_step": 148532, "epoch": 1668, "lr": 4.27453658904023e-05}
{"train_loss": 0.10620775073766708, "global_step": 148533, "epoch": 1668, "lr": 4.2744792289019835e-05}
{"train_loss": 0.07898333668708801, "global_step": 148534, "epoch": 1668, "lr": 4.274421868861275e-05}
{"train_loss": 0.18762613832950592, "global_step": 148535, "epoch": 1668, "lr": 4.274364508918111e-05}
{"train_loss": 0.08113511651754379, "global_step": 148536, "epoch": 1668, "lr": 4.2743071490724995e-05}
{"train_loss": 0.1237030103802681, "global_step": 148537, "epoch": 1668, "lr": 4.274249789324449e-05}
{"train_loss": 0.14644384384155273, "global_step": 148538, "epoch": 1668, "lr": 4.2741924296739674e-05}
{"train_loss": 0.08480054140090942, "global_step": 148539, "epoch": 1668, "lr": 4.27413507012106e-05}
{"train_loss": 0.10288276514980231, "global_step": 148540, "epoch": 1668, "lr": 4.2740777106657385e-05, "val_loss": 6.160365581512451}
{"train_loss": 0.10669394582509995, "global_step": 148541, "epoch": 1669, "lr": 4.274020351308006e-05}
{"train_loss": 0.11353820562362671, "global_step": 148542, "epoch": 1669, "lr": 4.273962992047875e-05}
{"train_loss": 0.14695923030376434, "global_step": 148543, "epoch": 1669, "lr": 4.2739056328853486e-05}
{"train_loss": 0.13567574322223663, "global_step": 148544, "epoch": 1669, "lr": 4.2738482738204394e-05}
{"train_loss": 0.08394463360309601, "global_step": 148545, "epoch": 1669, "lr": 4.27379091485315e-05}
{"train_loss": 0.10544872283935547, "global_step": 148546, "epoch": 1669, "lr": 4.2737335559834924e-05}
{"train_loss": 0.09684430807828903, "global_step": 148547, "epoch": 1669, "lr": 4.27367619721147e-05}
{"train_loss": 0.08312175422906876, "global_step": 148548, "epoch": 1669, "lr": 4.273618838537094e-05}
{"train_loss": 0.12273229658603668, "global_step": 148549, "epoch": 1669, "lr": 4.2735614799603727e-05}
{"train_loss": 0.06994286179542542, "global_step": 148550, "epoch": 1669, "lr": 4.27350412148131e-05}
{"train_loss": 0.12141633778810501, "global_step": 148551, "epoch": 1669, "lr": 4.273446763099918e-05}
{"train_loss": 0.09191413223743439, "global_step": 148552, "epoch": 1669, "lr": 4.2733894048162007e-05}
{"train_loss": 0.1044526994228363, "global_step": 148553, "epoch": 1669, "lr": 4.273332046630168e-05}
{"train_loss": 0.10108250379562378, "global_step": 148554, "epoch": 1669, "lr": 4.273274688541827e-05}
{"train_loss": 0.07680849730968475, "global_step": 148555, "epoch": 1669, "lr": 4.2732173305511845e-05}
{"train_loss": 0.14778338372707367, "global_step": 148556, "epoch": 1669, "lr": 4.2731599726582504e-05}
{"train_loss": 0.1121630072593689, "global_step": 148557, "epoch": 1669, "lr": 4.2731026148630314e-05}
{"train_loss": 0.13116052746772766, "global_step": 148558, "epoch": 1669, "lr": 4.273045257165533e-05}
{"train_loss": 0.08742405474185944, "global_step": 148559, "epoch": 1669, "lr": 4.2729878995657663e-05}
{"train_loss": 0.1086030825972557, "global_step": 148560, "epoch": 1669, "lr": 4.272930542063735e-05}
{"train_loss": 0.16354386508464813, "global_step": 148561, "epoch": 1669, "lr": 4.272873184659453e-05}
{"train_loss": 0.1073533147573471, "global_step": 148562, "epoch": 1669, "lr": 4.272815827352922e-05}
{"train_loss": 0.1662418097257614, "global_step": 148563, "epoch": 1669, "lr": 4.272758470144153e-05}
{"train_loss": 0.12973229587078094, "global_step": 148564, "epoch": 1669, "lr": 4.272701113033152e-05}
{"train_loss": 0.0764499381184578, "global_step": 148565, "epoch": 1669, "lr": 4.2726437560199286e-05}
{"train_loss": 0.1170310229063034, "global_step": 148566, "epoch": 1669, "lr": 4.2725863991044875e-05}
{"train_loss": 0.1221480518579483, "global_step": 148567, "epoch": 1669, "lr": 4.272529042286838e-05}
{"train_loss": 0.09518186002969742, "global_step": 148568, "epoch": 1669, "lr": 4.272471685566991e-05}
{"train_loss": 0.10695285350084305, "global_step": 148569, "epoch": 1669, "lr": 4.2724143289449485e-05}
{"train_loss": 0.05834704637527466, "global_step": 148570, "epoch": 1669, "lr": 4.2723569724207234e-05}
{"train_loss": 0.08769331872463226, "global_step": 148571, "epoch": 1669, "lr": 4.27229961599432e-05}
{"train_loss": 0.10761773586273193, "global_step": 148572, "epoch": 1669, "lr": 4.272242259665747e-05}
{"train_loss": 0.10832031816244125, "global_step": 148573, "epoch": 1669, "lr": 4.272184903435012e-05}
{"train_loss": 0.09587433189153671, "global_step": 148574, "epoch": 1669, "lr": 4.272127547302123e-05}
{"train_loss": 0.13092108070850372, "global_step": 148575, "epoch": 1669, "lr": 4.272070191267088e-05}
{"train_loss": 0.1788790076971054, "global_step": 148576, "epoch": 1669, "lr": 4.272012835329915e-05}
{"train_loss": 0.20229041576385498, "global_step": 148577, "epoch": 1669, "lr": 4.271955479490608e-05}
{"train_loss": 0.10971466451883316, "global_step": 148578, "epoch": 1669, "lr": 4.271898123749182e-05}
{"train_loss": 0.1010308712720871, "global_step": 148579, "epoch": 1669, "lr": 4.271840768105637e-05}
{"train_loss": 0.06434229761362076, "global_step": 148580, "epoch": 1669, "lr": 4.2717834125599854e-05}
{"train_loss": 0.08848710358142853, "global_step": 148581, "epoch": 1669, "lr": 4.271726057112233e-05}
{"train_loss": 0.1527329385280609, "global_step": 148582, "epoch": 1669, "lr": 4.271668701762389e-05}
{"train_loss": 0.07728660106658936, "global_step": 148583, "epoch": 1669, "lr": 4.271611346510459e-05}
{"train_loss": 0.1509680449962616, "global_step": 148584, "epoch": 1669, "lr": 4.271553991356452e-05}
{"train_loss": 0.09384221583604813, "global_step": 148585, "epoch": 1669, "lr": 4.271496636300377e-05}
{"train_loss": 0.08582695573568344, "global_step": 148586, "epoch": 1669, "lr": 4.271439281342239e-05}
{"train_loss": 0.23127110302448273, "global_step": 148587, "epoch": 1669, "lr": 4.271381926482049e-05}
{"train_loss": 0.14462222158908844, "global_step": 148588, "epoch": 1669, "lr": 4.271324571719812e-05}
{"train_loss": 0.06110839545726776, "global_step": 148589, "epoch": 1669, "lr": 4.271267217055537e-05}
{"train_loss": 0.13430741429328918, "global_step": 148590, "epoch": 1669, "lr": 4.2712098624892304e-05}
{"train_loss": 0.09862738102674484, "global_step": 148591, "epoch": 1669, "lr": 4.271152508020901e-05}
{"train_loss": 0.1239859014749527, "global_step": 148592, "epoch": 1669, "lr": 4.2710951536505564e-05}
{"train_loss": 0.04738227277994156, "global_step": 148593, "epoch": 1669, "lr": 4.2710377993782036e-05}
{"train_loss": 0.0732627660036087, "global_step": 148594, "epoch": 1669, "lr": 4.270980445203851e-05}
{"train_loss": 0.09340734034776688, "global_step": 148595, "epoch": 1669, "lr": 4.2709230911275066e-05}
{"train_loss": 0.11570877581834793, "global_step": 148596, "epoch": 1669, "lr": 4.270865737149178e-05}
{"train_loss": 0.11143609881401062, "global_step": 148597, "epoch": 1669, "lr": 4.270808383268873e-05}
{"train_loss": 0.20752021670341492, "global_step": 148598, "epoch": 1669, "lr": 4.270751029486597e-05}
{"train_loss": 0.09473808109760284, "global_step": 148599, "epoch": 1669, "lr": 4.2706936758023616e-05}
{"train_loss": 0.06290211528539658, "global_step": 148600, "epoch": 1669, "lr": 4.270636322216171e-05}
{"train_loss": 0.14064393937587738, "global_step": 148601, "epoch": 1669, "lr": 4.270578968728036e-05}
{"train_loss": 0.1269833743572235, "global_step": 148602, "epoch": 1669, "lr": 4.270521615337961e-05}
{"train_loss": 0.09411998093128204, "global_step": 148603, "epoch": 1669, "lr": 4.2704642620459555e-05}
{"train_loss": 0.12956593930721283, "global_step": 148604, "epoch": 1669, "lr": 4.2704069088520295e-05}
{"train_loss": 0.0794593021273613, "global_step": 148605, "epoch": 1669, "lr": 4.270349555756186e-05}
{"train_loss": 0.11153262108564377, "global_step": 148606, "epoch": 1669, "lr": 4.270292202758437e-05}
{"train_loss": 0.07035008072853088, "global_step": 148607, "epoch": 1669, "lr": 4.270234849858786e-05}
{"train_loss": 0.08703668415546417, "global_step": 148608, "epoch": 1669, "lr": 4.270177497057245e-05}
{"train_loss": 0.12700368463993073, "global_step": 148609, "epoch": 1669, "lr": 4.2701201443538184e-05}
{"train_loss": 0.09266958385705948, "global_step": 148610, "epoch": 1669, "lr": 4.270062791748517e-05}
{"train_loss": 0.05727504938840866, "global_step": 148611, "epoch": 1669, "lr": 4.270005439241347e-05}
{"train_loss": 0.1268901526927948, "global_step": 148612, "epoch": 1669, "lr": 4.269948086832314e-05}
{"train_loss": 0.0744272917509079, "global_step": 148613, "epoch": 1669, "lr": 4.269890734521428e-05}
{"train_loss": 0.08208827674388885, "global_step": 148614, "epoch": 1669, "lr": 4.269833382308697e-05}
{"train_loss": 0.0847817063331604, "global_step": 148615, "epoch": 1669, "lr": 4.269776030194127e-05}
{"train_loss": 0.10548976808786392, "global_step": 148616, "epoch": 1669, "lr": 4.269718678177729e-05}
{"train_loss": 0.16841337084770203, "global_step": 148617, "epoch": 1669, "lr": 4.2696613262595056e-05}
{"train_loss": 0.12097050994634628, "global_step": 148618, "epoch": 1669, "lr": 4.269603974439469e-05}
{"train_loss": 0.07922130823135376, "global_step": 148619, "epoch": 1669, "lr": 4.269546622717624e-05}
{"train_loss": 0.06901871412992477, "global_step": 148620, "epoch": 1669, "lr": 4.2694892710939786e-05}
{"train_loss": 0.08661036193370819, "global_step": 148621, "epoch": 1669, "lr": 4.269431919568544e-05}
{"train_loss": 0.10652858763933182, "global_step": 148622, "epoch": 1669, "lr": 4.2693745681413236e-05}
{"train_loss": 0.0637693926692009, "global_step": 148623, "epoch": 1669, "lr": 4.269317216812329e-05}
{"train_loss": 0.06027806177735329, "global_step": 148624, "epoch": 1669, "lr": 4.269259865581564e-05}
{"train_loss": 0.1274787038564682, "global_step": 148625, "epoch": 1669, "lr": 4.269202514449039e-05}
{"train_loss": 0.10478240251541138, "global_step": 148626, "epoch": 1669, "lr": 4.2691451634147596e-05}
{"train_loss": 0.08576536178588867, "global_step": 148627, "epoch": 1669, "lr": 4.269087812478736e-05}
{"train_loss": 0.18536323308944702, "global_step": 148628, "epoch": 1669, "lr": 4.269030461640974e-05}
{"train_loss": 0.10870429827423578, "global_step": 148629, "epoch": 1669, "lr": 4.268973110901483e-05, "val_loss": 6.142725467681885}
{"train_loss": 0.185069277882576, "global_step": 148630, "epoch": 1670, "lr": 4.2689157602602685e-05}
{"train_loss": 0.10290152579545975, "global_step": 148631, "epoch": 1670, "lr": 4.26885840971734e-05}
{"train_loss": 0.18774260580539703, "global_step": 148632, "epoch": 1670, "lr": 4.268801059272704e-05}
{"train_loss": 0.11694298684597015, "global_step": 148633, "epoch": 1670, "lr": 4.2687437089263695e-05}
{"train_loss": 0.05751464515924454, "global_step": 148634, "epoch": 1670, "lr": 4.268686358678343e-05}
{"train_loss": 0.07176968455314636, "global_step": 148635, "epoch": 1670, "lr": 4.2686290085286335e-05}
{"train_loss": 0.12496236711740494, "global_step": 148636, "epoch": 1670, "lr": 4.268571658477246e-05}
{"train_loss": 0.17669056355953217, "global_step": 148637, "epoch": 1670, "lr": 4.2685143085241926e-05}
{"train_loss": 0.10404307395219803, "global_step": 148638, "epoch": 1670, "lr": 4.268456958669476e-05}
{"train_loss": 0.09632702171802521, "global_step": 148639, "epoch": 1670, "lr": 4.268399608913107e-05}
{"train_loss": 0.16174988448619843, "global_step": 148640, "epoch": 1670, "lr": 4.268342259255094e-05}
{"train_loss": 0.12468079477548599, "global_step": 148641, "epoch": 1670, "lr": 4.268284909695442e-05}
{"train_loss": 0.17403921484947205, "global_step": 148642, "epoch": 1670, "lr": 4.2682275602341623e-05}
{"train_loss": 0.12202552706003189, "global_step": 148643, "epoch": 1670, "lr": 4.268170210871258e-05}
{"train_loss": 0.17720632255077362, "global_step": 148644, "epoch": 1670, "lr": 4.2681128616067415e-05}
{"train_loss": 0.12312057614326477, "global_step": 148645, "epoch": 1670, "lr": 4.2680555124406165e-05}
{"train_loss": 0.1389262080192566, "global_step": 148646, "epoch": 1670, "lr": 4.267998163372894e-05}
{"train_loss": 0.09899608045816422, "global_step": 148647, "epoch": 1670, "lr": 4.267940814403579e-05}
{"train_loss": 0.08775509893894196, "global_step": 148648, "epoch": 1670, "lr": 4.2678834655326815e-05}
{"train_loss": 0.11718215048313141, "global_step": 148649, "epoch": 1670, "lr": 4.267826116760208e-05}
{"train_loss": 0.11190814524888992, "global_step": 148650, "epoch": 1670, "lr": 4.267768768086166e-05}
{"train_loss": 0.10547997802495956, "global_step": 148651, "epoch": 1670, "lr": 4.267711419510564e-05}
{"train_loss": 0.11261279881000519, "global_step": 148652, "epoch": 1670, "lr": 4.267654071033408e-05}
{"train_loss": 0.1329154521226883, "global_step": 148653, "epoch": 1670, "lr": 4.2675967226547086e-05}
{"train_loss": 0.10296910256147385, "global_step": 148654, "epoch": 1670, "lr": 4.2675393743744715e-05}
{"train_loss": 0.15211375057697296, "global_step": 148655, "epoch": 1670, "lr": 4.267482026192704e-05}
{"train_loss": 0.13679568469524384, "global_step": 148656, "epoch": 1670, "lr": 4.267424678109416e-05}
{"train_loss": 0.11300301551818848, "global_step": 148657, "epoch": 1670, "lr": 4.267367330124612e-05}
{"train_loss": 0.09522269666194916, "global_step": 148658, "epoch": 1670, "lr": 4.267309982238301e-05}
{"train_loss": 0.12306380271911621, "global_step": 148659, "epoch": 1670, "lr": 4.2672526344504935e-05}
{"train_loss": 0.10579806566238403, "global_step": 148660, "epoch": 1670, "lr": 4.2671952867611934e-05}
{"train_loss": 0.11124934256076813, "global_step": 148661, "epoch": 1670, "lr": 4.267137939170412e-05}
{"train_loss": 0.1210467517375946, "global_step": 148662, "epoch": 1670, "lr": 4.267080591678152e-05}
{"train_loss": 0.15880626440048218, "global_step": 148663, "epoch": 1670, "lr": 4.267023244284427e-05}
{"train_loss": 0.07353979349136353, "global_step": 148664, "epoch": 1670, "lr": 4.2669658969892394e-05}
{"train_loss": 0.10642549395561218, "global_step": 148665, "epoch": 1670, "lr": 4.2669085497926014e-05}
{"train_loss": 0.12401711195707321, "global_step": 148666, "epoch": 1670, "lr": 4.266851202694517e-05}
{"train_loss": 0.10873957723379135, "global_step": 148667, "epoch": 1670, "lr": 4.266793855694997e-05}
{"train_loss": 0.1356865018606186, "global_step": 148668, "epoch": 1670, "lr": 4.266736508794047e-05}
{"train_loss": 0.16086603701114655, "global_step": 148669, "epoch": 1670, "lr": 4.266679161991675e-05}
{"train_loss": 0.12228719890117645, "global_step": 148670, "epoch": 1670, "lr": 4.266621815287889e-05}
{"train_loss": 0.07370758801698685, "global_step": 148671, "epoch": 1670, "lr": 4.2665644686826966e-05}
{"train_loss": 0.11367083340883255, "global_step": 148672, "epoch": 1670, "lr": 4.2665071221761066e-05}
{"train_loss": 0.1315520852804184, "global_step": 148673, "epoch": 1670, "lr": 4.266449775768126e-05}
{"train_loss": 0.10199318826198578, "global_step": 148674, "epoch": 1670, "lr": 4.2663924294587606e-05}
{"train_loss": 0.08763578534126282, "global_step": 148675, "epoch": 1670, "lr": 4.2663350832480195e-05}
{"train_loss": 0.15707986056804657, "global_step": 148676, "epoch": 1670, "lr": 4.266277737135913e-05}
{"train_loss": 0.08339837938547134, "global_step": 148677, "epoch": 1670, "lr": 4.266220391122444e-05}
{"train_loss": 0.09628260880708694, "global_step": 148678, "epoch": 1670, "lr": 4.266163045207625e-05}
{"train_loss": 0.15971601009368896, "global_step": 148679, "epoch": 1670, "lr": 4.2661056993914596e-05}
{"train_loss": 0.1212470680475235, "global_step": 148680, "epoch": 1670, "lr": 4.2660483536739594e-05}
{"train_loss": 0.13675741851329803, "global_step": 148681, "epoch": 1670, "lr": 4.265991008055128e-05}
{"train_loss": 0.10795798152685165, "global_step": 148682, "epoch": 1670, "lr": 4.265933662534977e-05}
{"train_loss": 0.09526948630809784, "global_step": 148683, "epoch": 1670, "lr": 4.2658763171135104e-05}
{"train_loss": 0.07265740633010864, "global_step": 148684, "epoch": 1670, "lr": 4.26581897179074e-05}
{"train_loss": 0.1278359740972519, "global_step": 148685, "epoch": 1670, "lr": 4.2657616265666687e-05}
{"train_loss": 0.10277022421360016, "global_step": 148686, "epoch": 1670, "lr": 4.2657042814413095e-05}
{"train_loss": 0.09132228791713715, "global_step": 148687, "epoch": 1670, "lr": 4.2656469364146656e-05}
{"train_loss": 0.11622385680675507, "global_step": 148688, "epoch": 1670, "lr": 4.265589591486747e-05}
{"train_loss": 0.07802757620811462, "global_step": 148689, "epoch": 1670, "lr": 4.265532246657561e-05}
{"train_loss": 0.0937076285481453, "global_step": 148690, "epoch": 1670, "lr": 4.265474901927115e-05}
{"train_loss": 0.07269509136676788, "global_step": 148691, "epoch": 1670, "lr": 4.265417557295417e-05}
{"train_loss": 0.09384164959192276, "global_step": 148692, "epoch": 1670, "lr": 4.265360212762476e-05}
{"train_loss": 0.11929965019226074, "global_step": 148693, "epoch": 1670, "lr": 4.2653028683282955e-05}
{"train_loss": 0.1649957299232483, "global_step": 148694, "epoch": 1670, "lr": 4.2652455239928874e-05}
{"train_loss": 0.13021883368492126, "global_step": 148695, "epoch": 1670, "lr": 4.265188179756259e-05}
{"train_loss": 0.09223610162734985, "global_step": 148696, "epoch": 1670, "lr": 4.2651308356184156e-05}
{"train_loss": 0.07705426961183548, "global_step": 148697, "epoch": 1670, "lr": 4.265073491579368e-05}
{"train_loss": 0.13869215548038483, "global_step": 148698, "epoch": 1670, "lr": 4.265016147639121e-05}
{"train_loss": 0.08490286767482758, "global_step": 148699, "epoch": 1670, "lr": 4.264958803797685e-05}
{"train_loss": 0.09701920300722122, "global_step": 148700, "epoch": 1670, "lr": 4.2649014600550645e-05}
{"train_loss": 0.13039225339889526, "global_step": 148701, "epoch": 1670, "lr": 4.264844116411271e-05}
{"train_loss": 0.14649087190628052, "global_step": 148702, "epoch": 1670, "lr": 4.264786772866308e-05}
{"train_loss": 0.12227523326873779, "global_step": 148703, "epoch": 1670, "lr": 4.2647294294201876e-05}
{"train_loss": 0.14643478393554688, "global_step": 148704, "epoch": 1670, "lr": 4.264672086072914e-05}
{"train_loss": 0.13689075410366058, "global_step": 148705, "epoch": 1670, "lr": 4.264614742824498e-05}
{"train_loss": 0.11867456138134003, "global_step": 148706, "epoch": 1670, "lr": 4.2645573996749446e-05}
{"train_loss": 0.061639077961444855, "global_step": 148707, "epoch": 1670, "lr": 4.264500056624262e-05}
{"train_loss": 0.09922610223293304, "global_step": 148708, "epoch": 1670, "lr": 4.2644427136724584e-05}
{"train_loss": 0.12277509272098541, "global_step": 148709, "epoch": 1670, "lr": 4.2643853708195416e-05}
{"train_loss": 0.06038575619459152, "global_step": 148710, "epoch": 1670, "lr": 4.2643280280655196e-05}
{"train_loss": 0.10304045677185059, "global_step": 148711, "epoch": 1670, "lr": 4.2642706854103986e-05}
{"train_loss": 0.05469265207648277, "global_step": 148712, "epoch": 1670, "lr": 4.26421334285419e-05}
{"train_loss": 0.18893124163150787, "global_step": 148713, "epoch": 1670, "lr": 4.264156000396896e-05}
{"train_loss": 0.13629785180091858, "global_step": 148714, "epoch": 1670, "lr": 4.26409865803853e-05}
{"train_loss": 0.07820554822683334, "global_step": 148715, "epoch": 1670, "lr": 4.264041315779094e-05}
{"train_loss": 0.09163428097963333, "global_step": 148716, "epoch": 1670, "lr": 4.263983973618602e-05}
{"train_loss": 0.10052575170993805, "global_step": 148717, "epoch": 1670, "lr": 4.2639266315570545e-05}
{"train_loss": 0.11568993710902299, "global_step": 148718, "epoch": 1670, "lr": 4.263869289594466e-05, "val_loss": 6.118277072906494, "train_action_mse_error": 28.74212646484375}
{"train_loss": 0.10289300978183746, "global_step": 148719, "epoch": 1671, "lr": 4.2638119477308405e-05}
{"train_loss": 0.08449863642454147, "global_step": 148720, "epoch": 1671, "lr": 4.263754605966187e-05}
{"train_loss": 0.07295463234186172, "global_step": 148721, "epoch": 1671, "lr": 4.2636972643005115e-05}
{"train_loss": 0.0812079906463623, "global_step": 148722, "epoch": 1671, "lr": 4.2636399227338244e-05}
{"train_loss": 0.049341559410095215, "global_step": 148723, "epoch": 1671, "lr": 4.26358258126613e-05}
{"train_loss": 0.10005873441696167, "global_step": 148724, "epoch": 1671, "lr": 4.26352523989744e-05}
{"train_loss": 0.11001701653003693, "global_step": 148725, "epoch": 1671, "lr": 4.263467898627759e-05}
{"train_loss": 0.11539699137210846, "global_step": 148726, "epoch": 1671, "lr": 4.263410557457096e-05}
{"train_loss": 0.11484503000974655, "global_step": 148727, "epoch": 1671, "lr": 4.263353216385459e-05}
{"train_loss": 0.10116041451692581, "global_step": 148728, "epoch": 1671, "lr": 4.2632958754128546e-05}
{"train_loss": 0.10272730141878128, "global_step": 148729, "epoch": 1671, "lr": 4.263238534539291e-05}
{"train_loss": 0.11227274686098099, "global_step": 148730, "epoch": 1671, "lr": 4.263181193764775e-05}
{"train_loss": 0.16988106071949005, "global_step": 148731, "epoch": 1671, "lr": 4.2631238530893184e-05}
{"train_loss": 0.10358597338199615, "global_step": 148732, "epoch": 1671, "lr": 4.263066512512922e-05}
{"train_loss": 0.1072382852435112, "global_step": 148733, "epoch": 1671, "lr": 4.2630091720356005e-05}
{"train_loss": 0.11685197800397873, "global_step": 148734, "epoch": 1671, "lr": 4.2629518316573555e-05}
{"train_loss": 0.1165287047624588, "global_step": 148735, "epoch": 1671, "lr": 4.262894491378201e-05}
{"train_loss": 0.0918794572353363, "global_step": 148736, "epoch": 1671, "lr": 4.262837151198138e-05}
{"train_loss": 0.07978407293558121, "global_step": 148737, "epoch": 1671, "lr": 4.26277981111718e-05}
{"train_loss": 0.1698213517665863, "global_step": 148738, "epoch": 1671, "lr": 4.262722471135331e-05}
{"train_loss": 0.08928228914737701, "global_step": 148739, "epoch": 1671, "lr": 4.262665131252601e-05}
{"train_loss": 0.12827035784721375, "global_step": 148740, "epoch": 1671, "lr": 4.2626077914689953e-05}
{"train_loss": 0.1498025357723236, "global_step": 148741, "epoch": 1671, "lr": 4.262550451784525e-05}
{"train_loss": 0.08096686750650406, "global_step": 148742, "epoch": 1671, "lr": 4.262493112199193e-05}
{"train_loss": 0.12712307274341583, "global_step": 148743, "epoch": 1671, "lr": 4.262435772713013e-05}
{"train_loss": 0.18320222198963165, "global_step": 148744, "epoch": 1671, "lr": 4.262378433325986e-05}
{"train_loss": 0.12036849558353424, "global_step": 148745, "epoch": 1671, "lr": 4.2623210940381266e-05}
{"train_loss": 0.08263857662677765, "global_step": 148746, "epoch": 1671, "lr": 4.2622637548494385e-05}
{"train_loss": 0.1391461044549942, "global_step": 148747, "epoch": 1671, "lr": 4.262206415759927e-05}
{"train_loss": 0.1508921980857849, "global_step": 148748, "epoch": 1671, "lr": 4.262149076769607e-05}
{"train_loss": 0.10957520455121994, "global_step": 148749, "epoch": 1671, "lr": 4.262091737878478e-05}
{"train_loss": 0.1436341553926468, "global_step": 148750, "epoch": 1671, "lr": 4.262034399086556e-05}
{"train_loss": 0.11179760098457336, "global_step": 148751, "epoch": 1671, "lr": 4.2619770603938414e-05}
{"train_loss": 0.10050935298204422, "global_step": 148752, "epoch": 1671, "lr": 4.261919721800347e-05}
{"train_loss": 0.047069646418094635, "global_step": 148753, "epoch": 1671, "lr": 4.261862383306076e-05}
{"train_loss": 0.08771010488271713, "global_step": 148754, "epoch": 1671, "lr": 4.2618050449110415e-05}
{"train_loss": 0.14103110134601593, "global_step": 148755, "epoch": 1671, "lr": 4.2617477066152453e-05}
{"train_loss": 0.13807334005832672, "global_step": 148756, "epoch": 1671, "lr": 4.261690368418702e-05}
{"train_loss": 0.15309712290763855, "global_step": 148757, "epoch": 1671, "lr": 4.2616330303214115e-05}
{"train_loss": 0.061281248927116394, "global_step": 148758, "epoch": 1671, "lr": 4.261575692323388e-05}
{"train_loss": 0.12675943970680237, "global_step": 148759, "epoch": 1671, "lr": 4.261518354424635e-05}
{"train_loss": 0.1238013356924057, "global_step": 148760, "epoch": 1671, "lr": 4.261461016625163e-05}
{"train_loss": 0.1505364030599594, "global_step": 148761, "epoch": 1671, "lr": 4.261403678924977e-05}
{"train_loss": 0.07638555765151978, "global_step": 148762, "epoch": 1671, "lr": 4.2613463413240887e-05}
{"train_loss": 0.10626110434532166, "global_step": 148763, "epoch": 1671, "lr": 4.261289003822501e-05}
{"train_loss": 0.10934314876794815, "global_step": 148764, "epoch": 1671, "lr": 4.261231666420227e-05}
{"train_loss": 0.10385684669017792, "global_step": 148765, "epoch": 1671, "lr": 4.2611743291172687e-05}
{"train_loss": 0.13779771327972412, "global_step": 148766, "epoch": 1671, "lr": 4.261116991913636e-05}
{"train_loss": 0.07400805503129959, "global_step": 148767, "epoch": 1671, "lr": 4.26105965480934e-05}
{"train_loss": 0.06009488180279732, "global_step": 148768, "epoch": 1671, "lr": 4.261002317804383e-05}
{"train_loss": 0.10652714967727661, "global_step": 148769, "epoch": 1671, "lr": 4.260944980898777e-05}
{"train_loss": 0.08358784765005112, "global_step": 148770, "epoch": 1671, "lr": 4.260887644092526e-05}
{"train_loss": 0.09524287283420563, "global_step": 148771, "epoch": 1671, "lr": 4.260830307385642e-05}
{"train_loss": 0.07132667303085327, "global_step": 148772, "epoch": 1671, "lr": 4.260772970778128e-05}
{"train_loss": 0.1318388432264328, "global_step": 148773, "epoch": 1671, "lr": 4.260715634269997e-05}
{"train_loss": 0.09829726815223694, "global_step": 148774, "epoch": 1671, "lr": 4.2606582978612506e-05}
{"train_loss": 0.14644379913806915, "global_step": 148775, "epoch": 1671, "lr": 4.260600961551902e-05}
{"train_loss": 0.10410822182893753, "global_step": 148776, "epoch": 1671, "lr": 4.260543625341955e-05}
{"train_loss": 0.057921137660741806, "global_step": 148777, "epoch": 1671, "lr": 4.26048628923142e-05}
{"train_loss": 0.09507104754447937, "global_step": 148778, "epoch": 1671, "lr": 4.260428953220302e-05}
{"train_loss": 0.15254975855350494, "global_step": 148779, "epoch": 1671, "lr": 4.2603716173086125e-05}
{"train_loss": 0.07994179427623749, "global_step": 148780, "epoch": 1671, "lr": 4.260314281496355e-05}
{"train_loss": 0.08486493676900864, "global_step": 148781, "epoch": 1671, "lr": 4.260256945783541e-05}
{"train_loss": 0.09144634008407593, "global_step": 148782, "epoch": 1671, "lr": 4.260199610170175e-05}
{"train_loss": 0.10831030458211899, "global_step": 148783, "epoch": 1671, "lr": 4.260142274656268e-05}
{"train_loss": 0.0698823556303978, "global_step": 148784, "epoch": 1671, "lr": 4.260084939241824e-05}
{"train_loss": 0.10572642087936401, "global_step": 148785, "epoch": 1671, "lr": 4.260027603926853e-05}
{"train_loss": 0.0863446444272995, "global_step": 148786, "epoch": 1671, "lr": 4.259970268711363e-05}
{"train_loss": 0.1262357234954834, "global_step": 148787, "epoch": 1671, "lr": 4.259912933595359e-05}
{"train_loss": 0.21123667061328888, "global_step": 148788, "epoch": 1671, "lr": 4.259855598578854e-05}
{"train_loss": 0.1505911946296692, "global_step": 148789, "epoch": 1671, "lr": 4.259798263661849e-05}
{"train_loss": 0.06772365421056747, "global_step": 148790, "epoch": 1671, "lr": 4.2597409288443576e-05}
{"train_loss": 0.0940721407532692, "global_step": 148791, "epoch": 1671, "lr": 4.259683594126383e-05}
{"train_loss": 0.13298490643501282, "global_step": 148792, "epoch": 1671, "lr": 4.259626259507937e-05}
{"train_loss": 0.16026581823825836, "global_step": 148793, "epoch": 1671, "lr": 4.2595689249890236e-05}
{"train_loss": 0.07928439974784851, "global_step": 148794, "epoch": 1671, "lr": 4.2595115905696534e-05}
{"train_loss": 0.11121857911348343, "global_step": 148795, "epoch": 1671, "lr": 4.259454256249831e-05}
{"train_loss": 0.13421763479709625, "global_step": 148796, "epoch": 1671, "lr": 4.259396922029568e-05}
{"train_loss": 0.10617954283952713, "global_step": 148797, "epoch": 1671, "lr": 4.259339587908868e-05}
{"train_loss": 0.11668461561203003, "global_step": 148798, "epoch": 1671, "lr": 4.259282253887744e-05}
{"train_loss": 0.04494095221161842, "global_step": 148799, "epoch": 1671, "lr": 4.2592249199661965e-05}
{"train_loss": 0.1280914694070816, "global_step": 148800, "epoch": 1671, "lr": 4.2591675861442407e-05}
{"train_loss": 0.1322658509016037, "global_step": 148801, "epoch": 1671, "lr": 4.2591102524218776e-05}
{"train_loss": 0.1076982319355011, "global_step": 148802, "epoch": 1671, "lr": 4.25905291879912e-05}
{"train_loss": 0.1048460379242897, "global_step": 148803, "epoch": 1671, "lr": 4.2589955852759745e-05}
{"train_loss": 0.17264941334724426, "global_step": 148804, "epoch": 1671, "lr": 4.258938251852446e-05}
{"train_loss": 0.11341500282287598, "global_step": 148805, "epoch": 1671, "lr": 4.258880918528546e-05}
{"train_loss": 0.101263627409935, "global_step": 148806, "epoch": 1671, "lr": 4.258823585304278e-05}
{"train_loss": 0.11022601657536592, "global_step": 148807, "epoch": 1671, "lr": 4.258766252179655e-05, "val_loss": 6.294381141662598}
{"train_loss": 0.07266376167535782, "global_step": 148808, "epoch": 1672, "lr": 4.258708919154679e-05}
{"train_loss": 0.07856695353984833, "global_step": 148809, "epoch": 1672, "lr": 4.258651586229363e-05}
{"train_loss": 0.05190729349851608, "global_step": 148810, "epoch": 1672, "lr": 4.258594253403711e-05}
{"train_loss": 0.07439734041690826, "global_step": 148811, "epoch": 1672, "lr": 4.258536920677733e-05}
{"train_loss": 0.08656930923461914, "global_step": 148812, "epoch": 1672, "lr": 4.2584795880514336e-05}
{"train_loss": 0.09585554152727127, "global_step": 148813, "epoch": 1672, "lr": 4.2584222555248255e-05}
{"train_loss": 0.14591895043849945, "global_step": 148814, "epoch": 1672, "lr": 4.2583649230979106e-05}
{"train_loss": 0.102236807346344, "global_step": 148815, "epoch": 1672, "lr": 4.258307590770702e-05}
{"train_loss": 0.09336115419864655, "global_step": 148816, "epoch": 1672, "lr": 4.2582502585432025e-05}
{"train_loss": 0.08104673773050308, "global_step": 148817, "epoch": 1672, "lr": 4.2581929264154244e-05}
{"train_loss": 0.11882168799638748, "global_step": 148818, "epoch": 1672, "lr": 4.258135594387371e-05}
{"train_loss": 0.1679200381040573, "global_step": 148819, "epoch": 1672, "lr": 4.258078262459056e-05}
{"train_loss": 0.07898419350385666, "global_step": 148820, "epoch": 1672, "lr": 4.258020930630479e-05}
{"train_loss": 0.1341264545917511, "global_step": 148821, "epoch": 1672, "lr": 4.257963598901654e-05}
{"train_loss": 0.10375957190990448, "global_step": 148822, "epoch": 1672, "lr": 4.257906267272587e-05}
{"train_loss": 0.08668909221887589, "global_step": 148823, "epoch": 1672, "lr": 4.257848935743284e-05}
{"train_loss": 0.09926890581846237, "global_step": 148824, "epoch": 1672, "lr": 4.257791604313757e-05}
{"train_loss": 0.0763455256819725, "global_step": 148825, "epoch": 1672, "lr": 4.2577342729840075e-05}
{"train_loss": 0.09169038385152817, "global_step": 148826, "epoch": 1672, "lr": 4.25767694175405e-05}
{"train_loss": 0.04207337647676468, "global_step": 148827, "epoch": 1672, "lr": 4.257619610623885e-05}
{"train_loss": 0.10212968289852142, "global_step": 148828, "epoch": 1672, "lr": 4.257562279593528e-05}
{"train_loss": 0.13620342314243317, "global_step": 148829, "epoch": 1672, "lr": 4.257504948662979e-05}
{"train_loss": 0.05193595215678215, "global_step": 148830, "epoch": 1672, "lr": 4.257447617832252e-05}
{"train_loss": 0.12663112580776215, "global_step": 148831, "epoch": 1672, "lr": 4.25739028710135e-05}
{"train_loss": 0.07955914735794067, "global_step": 148832, "epoch": 1672, "lr": 4.257332956470286e-05}
{"train_loss": 0.1397770792245865, "global_step": 148833, "epoch": 1672, "lr": 4.257275625939061e-05}
{"train_loss": 0.1374620944261551, "global_step": 148834, "epoch": 1672, "lr": 4.257218295507689e-05}
{"train_loss": 0.12477236241102219, "global_step": 148835, "epoch": 1672, "lr": 4.257160965176172e-05}
{"train_loss": 0.07812997698783875, "global_step": 148836, "epoch": 1672, "lr": 4.257103634944524e-05}
{"train_loss": 0.13581055402755737, "global_step": 148837, "epoch": 1672, "lr": 4.257046304812747e-05}
{"train_loss": 0.18214167654514313, "global_step": 148838, "epoch": 1672, "lr": 4.2569889747808516e-05}
{"train_loss": 0.08917326480150223, "global_step": 148839, "epoch": 1672, "lr": 4.256931644848845e-05}
{"train_loss": 0.057087212800979614, "global_step": 148840, "epoch": 1672, "lr": 4.256874315016735e-05}
{"train_loss": 0.056278131902217865, "global_step": 148841, "epoch": 1672, "lr": 4.25681698528453e-05}
{"train_loss": 0.09310327470302582, "global_step": 148842, "epoch": 1672, "lr": 4.256759655652236e-05}
{"train_loss": 0.1218554675579071, "global_step": 148843, "epoch": 1672, "lr": 4.2567023261198625e-05}
{"train_loss": 0.034460265189409256, "global_step": 148844, "epoch": 1672, "lr": 4.2566449966874144e-05}
{"train_loss": 0.1186811774969101, "global_step": 148845, "epoch": 1672, "lr": 4.256587667354904e-05}
{"train_loss": 0.04729938134551048, "global_step": 148846, "epoch": 1672, "lr": 4.256530338122334e-05}
{"train_loss": 0.05773378908634186, "global_step": 148847, "epoch": 1672, "lr": 4.2564730089897156e-05}
{"train_loss": 0.07955168932676315, "global_step": 148848, "epoch": 1672, "lr": 4.256415679957054e-05}
{"train_loss": 0.05117375776171684, "global_step": 148849, "epoch": 1672, "lr": 4.25635835102436e-05}
{"train_loss": 0.09036584943532944, "global_step": 148850, "epoch": 1672, "lr": 4.2563010221916384e-05}
{"train_loss": 0.12060021609067917, "global_step": 148851, "epoch": 1672, "lr": 4.2562436934589e-05}
{"train_loss": 0.10566964745521545, "global_step": 148852, "epoch": 1672, "lr": 4.2561863648261476e-05}
{"train_loss": 0.05166846886277199, "global_step": 148853, "epoch": 1672, "lr": 4.256129036293394e-05}
{"train_loss": 0.06843958050012589, "global_step": 148854, "epoch": 1672, "lr": 4.256071707860643e-05}
{"train_loss": 0.09677598625421524, "global_step": 148855, "epoch": 1672, "lr": 4.256014379527907e-05}
{"train_loss": 0.13255465030670166, "global_step": 148856, "epoch": 1672, "lr": 4.255957051295187e-05}
{"train_loss": 0.13936200737953186, "global_step": 148857, "epoch": 1672, "lr": 4.255899723162497e-05}
{"train_loss": 0.13935579359531403, "global_step": 148858, "epoch": 1672, "lr": 4.255842395129842e-05}
{"train_loss": 0.08890193700790405, "global_step": 148859, "epoch": 1672, "lr": 4.2557850671972295e-05}
{"train_loss": 0.08950947225093842, "global_step": 148860, "epoch": 1672, "lr": 4.255727739364668e-05}
{"train_loss": 0.11914447695016861, "global_step": 148861, "epoch": 1672, "lr": 4.255670411632164e-05}
{"train_loss": 0.23605948686599731, "global_step": 148862, "epoch": 1672, "lr": 4.255613083999728e-05}
{"train_loss": 0.07413604110479355, "global_step": 148863, "epoch": 1672, "lr": 4.255555756467363e-05}
{"train_loss": 0.1439167559146881, "global_step": 148864, "epoch": 1672, "lr": 4.2554984290350824e-05}
{"train_loss": 0.09907931089401245, "global_step": 148865, "epoch": 1672, "lr": 4.255441101702888e-05}
{"train_loss": 0.10933429002761841, "global_step": 148866, "epoch": 1672, "lr": 4.255383774470794e-05}
{"train_loss": 0.13256479799747467, "global_step": 148867, "epoch": 1672, "lr": 4.2553264473388004e-05}
{"train_loss": 0.13844606280326843, "global_step": 148868, "epoch": 1672, "lr": 4.255269120306923e-05}
{"train_loss": 0.16693776845932007, "global_step": 148869, "epoch": 1672, "lr": 4.255211793375162e-05}
{"train_loss": 0.10045310109853745, "global_step": 148870, "epoch": 1672, "lr": 4.255154466543532e-05}
{"train_loss": 0.09137064218521118, "global_step": 148871, "epoch": 1672, "lr": 4.2550971398120346e-05}
{"train_loss": 0.10600259155035019, "global_step": 148872, "epoch": 1672, "lr": 4.2550398131806824e-05}
{"train_loss": 0.06155427545309067, "global_step": 148873, "epoch": 1672, "lr": 4.254982486649479e-05}
{"train_loss": 0.13785099983215332, "global_step": 148874, "epoch": 1672, "lr": 4.254925160218434e-05}
{"train_loss": 0.09090888500213623, "global_step": 148875, "epoch": 1672, "lr": 4.254867833887558e-05}
{"train_loss": 0.16388174891471863, "global_step": 148876, "epoch": 1672, "lr": 4.2548105076568547e-05}
{"train_loss": 0.16130903363227844, "global_step": 148877, "epoch": 1672, "lr": 4.254753181526333e-05}
{"train_loss": 0.14101283252239227, "global_step": 148878, "epoch": 1672, "lr": 4.254695855496e-05}
{"train_loss": 0.16122843325138092, "global_step": 148879, "epoch": 1672, "lr": 4.254638529565864e-05}
{"train_loss": 0.11142551153898239, "global_step": 148880, "epoch": 1672, "lr": 4.254581203735933e-05}
{"train_loss": 0.08325697481632233, "global_step": 148881, "epoch": 1672, "lr": 4.254523878006217e-05}
{"train_loss": 0.12813635170459747, "global_step": 148882, "epoch": 1672, "lr": 4.254466552376718e-05}
{"train_loss": 0.06251703947782516, "global_step": 148883, "epoch": 1672, "lr": 4.254409226847448e-05}
{"train_loss": 0.09017924219369888, "global_step": 148884, "epoch": 1672, "lr": 4.2543519014184125e-05}
{"train_loss": 0.14145103096961975, "global_step": 148885, "epoch": 1672, "lr": 4.254294576089623e-05}
{"train_loss": 0.06113088130950928, "global_step": 148886, "epoch": 1672, "lr": 4.254237250861081e-05}
{"train_loss": 0.06177039071917534, "global_step": 148887, "epoch": 1672, "lr": 4.254179925732801e-05}
{"train_loss": 0.11004479974508286, "global_step": 148888, "epoch": 1672, "lr": 4.254122600704784e-05}
{"train_loss": 0.08466638624668121, "global_step": 148889, "epoch": 1672, "lr": 4.2540652757770446e-05}
{"train_loss": 0.18126514554023743, "global_step": 148890, "epoch": 1672, "lr": 4.254007950949584e-05}
{"train_loss": 0.08608375489711761, "global_step": 148891, "epoch": 1672, "lr": 4.253950626222415e-05}
{"train_loss": 0.07819081097841263, "global_step": 148892, "epoch": 1672, "lr": 4.2538933015955414e-05}
{"train_loss": 0.08174304664134979, "global_step": 148893, "epoch": 1672, "lr": 4.253835977068973e-05}
{"train_loss": 0.09356872737407684, "global_step": 148894, "epoch": 1672, "lr": 4.2537786526427195e-05}
{"train_loss": 0.13087333738803864, "global_step": 148895, "epoch": 1672, "lr": 4.2537213283167856e-05}
{"train_loss": 0.10421176924464408, "global_step": 148896, "epoch": 1672, "lr": 4.253664004091179e-05, "val_loss": 6.397773265838623}
{"train_loss": 0.04811520501971245, "global_step": 148897, "epoch": 1673, "lr": 4.253606679965908e-05}
{"train_loss": 0.18147583305835724, "global_step": 148898, "epoch": 1673, "lr": 4.253549355940981e-05}
{"train_loss": 0.1295737475156784, "global_step": 148899, "epoch": 1673, "lr": 4.2534920320164054e-05}
{"train_loss": 0.08944237977266312, "global_step": 148900, "epoch": 1673, "lr": 4.2534347081921895e-05}
{"train_loss": 0.07131293416023254, "global_step": 148901, "epoch": 1673, "lr": 4.2533773844683375e-05}
{"train_loss": 0.15870153903961182, "global_step": 148902, "epoch": 1673, "lr": 4.253320060844863e-05}
{"train_loss": 0.07876169681549072, "global_step": 148903, "epoch": 1673, "lr": 4.253262737321768e-05}
{"train_loss": 0.07040710002183914, "global_step": 148904, "epoch": 1673, "lr": 4.253205413899064e-05}
{"train_loss": 0.06847238540649414, "global_step": 148905, "epoch": 1673, "lr": 4.253148090576756e-05}
{"train_loss": 0.19527855515480042, "global_step": 148906, "epoch": 1673, "lr": 4.253090767354856e-05}
{"train_loss": 0.09642931073904037, "global_step": 148907, "epoch": 1673, "lr": 4.253033444233365e-05}
{"train_loss": 0.1398390829563141, "global_step": 148908, "epoch": 1673, "lr": 4.252976121212298e-05}
{"train_loss": 0.07533823698759079, "global_step": 148909, "epoch": 1673, "lr": 4.252918798291657e-05}
{"train_loss": 0.1086340919137001, "global_step": 148910, "epoch": 1673, "lr": 4.252861475471454e-05}
{"train_loss": 0.07801320403814316, "global_step": 148911, "epoch": 1673, "lr": 4.252804152751692e-05}
{"train_loss": 0.21462908387184143, "global_step": 148912, "epoch": 1673, "lr": 4.252746830132382e-05}
{"train_loss": 0.14738060534000397, "global_step": 148913, "epoch": 1673, "lr": 4.2526895076135334e-05}
{"train_loss": 0.14509938657283783, "global_step": 148914, "epoch": 1673, "lr": 4.25263218519515e-05}
{"train_loss": 0.047345954924821854, "global_step": 148915, "epoch": 1673, "lr": 4.252574862877241e-05}
{"train_loss": 0.07101292908191681, "global_step": 148916, "epoch": 1673, "lr": 4.2525175406598145e-05}
{"train_loss": 0.1340220719575882, "global_step": 148917, "epoch": 1673, "lr": 4.2524602185428775e-05}
{"train_loss": 0.10959315299987793, "global_step": 148918, "epoch": 1673, "lr": 4.2524028965264386e-05}
{"train_loss": 0.09589642286300659, "global_step": 148919, "epoch": 1673, "lr": 4.252345574610506e-05}
{"train_loss": 0.14960111677646637, "global_step": 148920, "epoch": 1673, "lr": 4.252288252795084e-05}
{"train_loss": 0.13895922899246216, "global_step": 148921, "epoch": 1673, "lr": 4.2522309310801856e-05}
{"train_loss": 0.12614372372627258, "global_step": 148922, "epoch": 1673, "lr": 4.252173609465813e-05}
{"train_loss": 0.07015379518270493, "global_step": 148923, "epoch": 1673, "lr": 4.252116287951978e-05}
{"train_loss": 0.1352115124464035, "global_step": 148924, "epoch": 1673, "lr": 4.252058966538686e-05}
{"train_loss": 0.12028492987155914, "global_step": 148925, "epoch": 1673, "lr": 4.252001645225947e-05}
{"train_loss": 0.060755956918001175, "global_step": 148926, "epoch": 1673, "lr": 4.2519443240137645e-05}
{"train_loss": 0.16371333599090576, "global_step": 148927, "epoch": 1673, "lr": 4.2518870029021526e-05}
{"train_loss": 0.10810549557209015, "global_step": 148928, "epoch": 1673, "lr": 4.251829681891112e-05}
{"train_loss": 0.04175296798348427, "global_step": 148929, "epoch": 1673, "lr": 4.251772360980654e-05}
{"train_loss": 0.0845685824751854, "global_step": 148930, "epoch": 1673, "lr": 4.2517150401707886e-05}
{"train_loss": 0.12244956940412521, "global_step": 148931, "epoch": 1673, "lr": 4.251657719461519e-05}
{"train_loss": 0.08059895783662796, "global_step": 148932, "epoch": 1673, "lr": 4.251600398852856e-05}
{"train_loss": 0.06690151244401932, "global_step": 148933, "epoch": 1673, "lr": 4.251543078344806e-05}
{"train_loss": 0.10588249564170837, "global_step": 148934, "epoch": 1673, "lr": 4.2514857579373766e-05}
{"train_loss": 0.14943403005599976, "global_step": 148935, "epoch": 1673, "lr": 4.251428437630576e-05}
{"train_loss": 0.06330262124538422, "global_step": 148936, "epoch": 1673, "lr": 4.251371117424412e-05}
{"train_loss": 0.08049050718545914, "global_step": 148937, "epoch": 1673, "lr": 4.251313797318892e-05}
{"train_loss": 0.09514061361551285, "global_step": 148938, "epoch": 1673, "lr": 4.251256477314024e-05}
{"train_loss": 0.0908103883266449, "global_step": 148939, "epoch": 1673, "lr": 4.251199157409814e-05}
{"train_loss": 0.09417161345481873, "global_step": 148940, "epoch": 1673, "lr": 4.251141837606274e-05}
{"train_loss": 0.11865328997373581, "global_step": 148941, "epoch": 1673, "lr": 4.251084517903405e-05}
{"train_loss": 0.0912872850894928, "global_step": 148942, "epoch": 1673, "lr": 4.251027198301222e-05}
{"train_loss": 0.09485939890146255, "global_step": 148943, "epoch": 1673, "lr": 4.250969878799726e-05}
{"train_loss": 0.11612794548273087, "global_step": 148944, "epoch": 1673, "lr": 4.250912559398931e-05}
{"train_loss": 0.12948520481586456, "global_step": 148945, "epoch": 1673, "lr": 4.25085524009884e-05}
{"train_loss": 0.05982225760817528, "global_step": 148946, "epoch": 1673, "lr": 4.2507979208994635e-05}
{"train_loss": 0.243565633893013, "global_step": 148947, "epoch": 1673, "lr": 4.250740601800806e-05}
{"train_loss": 0.1213727742433548, "global_step": 148948, "epoch": 1673, "lr": 4.250683282802877e-05}
{"train_loss": 0.10879442095756531, "global_step": 148949, "epoch": 1673, "lr": 4.250625963905688e-05}
{"train_loss": 0.07491999119520187, "global_step": 148950, "epoch": 1673, "lr": 4.25056864510924e-05}
{"train_loss": 0.08585801720619202, "global_step": 148951, "epoch": 1673, "lr": 4.2505113264135464e-05}
{"train_loss": 0.1421050727367401, "global_step": 148952, "epoch": 1673, "lr": 4.2504540078186103e-05}
{"train_loss": 0.0924583449959755, "global_step": 148953, "epoch": 1673, "lr": 4.250396689324444e-05}
{"train_loss": 0.08261200040578842, "global_step": 148954, "epoch": 1673, "lr": 4.250339370931051e-05}
{"train_loss": 0.08657293021678925, "global_step": 148955, "epoch": 1673, "lr": 4.2502820526384414e-05}
{"train_loss": 0.05638042837381363, "global_step": 148956, "epoch": 1673, "lr": 4.250224734446621e-05}
{"train_loss": 0.13315336406230927, "global_step": 148957, "epoch": 1673, "lr": 4.2501674163556e-05}
{"train_loss": 0.046414826065301895, "global_step": 148958, "epoch": 1673, "lr": 4.250110098365384e-05}
{"train_loss": 0.08514020591974258, "global_step": 148959, "epoch": 1673, "lr": 4.2500527804759835e-05}
{"train_loss": 0.10939224809408188, "global_step": 148960, "epoch": 1673, "lr": 4.249995462687402e-05}
{"train_loss": 0.07956115156412125, "global_step": 148961, "epoch": 1673, "lr": 4.249938144999651e-05}
{"train_loss": 0.1125141829252243, "global_step": 148962, "epoch": 1673, "lr": 4.249880827412734e-05}
{"train_loss": 0.15796367824077606, "global_step": 148963, "epoch": 1673, "lr": 4.2498235099266645e-05}
{"train_loss": 0.05960473790764809, "global_step": 148964, "epoch": 1673, "lr": 4.2497661925414454e-05}
{"train_loss": 0.07747918367385864, "global_step": 148965, "epoch": 1673, "lr": 4.2497088752570854e-05}
{"train_loss": 0.06433068215847015, "global_step": 148966, "epoch": 1673, "lr": 4.249651558073595e-05}
{"train_loss": 0.0840790867805481, "global_step": 148967, "epoch": 1673, "lr": 4.249594240990978e-05}
{"train_loss": 0.1433475911617279, "global_step": 148968, "epoch": 1673, "lr": 4.249536924009245e-05}
{"train_loss": 0.1436619609594345, "global_step": 148969, "epoch": 1673, "lr": 4.249479607128402e-05}
{"train_loss": 0.10628590732812881, "global_step": 148970, "epoch": 1673, "lr": 4.249422290348458e-05}
{"train_loss": 0.12282541394233704, "global_step": 148971, "epoch": 1673, "lr": 4.249364973669419e-05}
{"train_loss": 0.1720813810825348, "global_step": 148972, "epoch": 1673, "lr": 4.2493076570912955e-05}
{"train_loss": 0.09261904656887054, "global_step": 148973, "epoch": 1673, "lr": 4.249250340614093e-05}
{"train_loss": 0.09468815475702286, "global_step": 148974, "epoch": 1673, "lr": 4.249193024237819e-05}
{"train_loss": 0.11373919993638992, "global_step": 148975, "epoch": 1673, "lr": 4.249135707962482e-05}
{"train_loss": 0.1094982847571373, "global_step": 148976, "epoch": 1673, "lr": 4.249078391788089e-05}
{"train_loss": 0.058676932007074356, "global_step": 148977, "epoch": 1673, "lr": 4.2490210757146484e-05}
{"train_loss": 0.10235849767923355, "global_step": 148978, "epoch": 1673, "lr": 4.248963759742169e-05}
{"train_loss": 0.1399928331375122, "global_step": 148979, "epoch": 1673, "lr": 4.248906443870655e-05}
{"train_loss": 0.10372234880924225, "global_step": 148980, "epoch": 1673, "lr": 4.248849128100119e-05}
{"train_loss": 0.174624964594841, "global_step": 148981, "epoch": 1673, "lr": 4.2487918124305635e-05}
{"train_loss": 0.10848058015108109, "global_step": 148982, "epoch": 1673, "lr": 4.248734496862001e-05}
{"train_loss": 0.09382282197475433, "global_step": 148983, "epoch": 1673, "lr": 4.248677181394435e-05}
{"train_loss": 0.09583096206188202, "global_step": 148984, "epoch": 1673, "lr": 4.248619866027875e-05}
{"train_loss": 0.10677231161781911, "global_step": 148985, "epoch": 1673, "lr": 4.2485625507623315e-05, "val_loss": 6.316447734832764}
{"train_loss": 0.05936211347579956, "global_step": 148986, "epoch": 1674, "lr": 4.248505235597807e-05}
{"train_loss": 0.06781401485204697, "global_step": 148987, "epoch": 1674, "lr": 4.248447920534314e-05}
{"train_loss": 0.11453469842672348, "global_step": 148988, "epoch": 1674, "lr": 4.2483906055718564e-05}
{"train_loss": 0.1330660730600357, "global_step": 148989, "epoch": 1674, "lr": 4.248333290710445e-05}
{"train_loss": 0.07936302572488785, "global_step": 148990, "epoch": 1674, "lr": 4.248275975950085e-05}
{"train_loss": 0.06604894250631332, "global_step": 148991, "epoch": 1674, "lr": 4.2482186612907856e-05}
{"train_loss": 0.09669451415538788, "global_step": 148992, "epoch": 1674, "lr": 4.248161346732554e-05}
{"train_loss": 0.1197267398238182, "global_step": 148993, "epoch": 1674, "lr": 4.2481040322753974e-05}
{"train_loss": 0.10299023985862732, "global_step": 148994, "epoch": 1674, "lr": 4.2480467179193246e-05}
{"train_loss": 0.12696519494056702, "global_step": 148995, "epoch": 1674, "lr": 4.247989403664343e-05}
{"train_loss": 0.07839858531951904, "global_step": 148996, "epoch": 1674, "lr": 4.247932089510459e-05}
{"train_loss": 0.13398781418800354, "global_step": 148997, "epoch": 1674, "lr": 4.2478747754576833e-05}
{"train_loss": 0.12234268337488174, "global_step": 148998, "epoch": 1674, "lr": 4.247817461506019e-05}
{"train_loss": 0.09368514269590378, "global_step": 148999, "epoch": 1674, "lr": 4.247760147655479e-05}
{"train_loss": 0.06811956316232681, "global_step": 149000, "epoch": 1674, "lr": 4.247702833906066e-05}
{"train_loss": 0.07347752898931503, "global_step": 149001, "epoch": 1674, "lr": 4.24764552025779e-05}
{"train_loss": 0.10616602003574371, "global_step": 149002, "epoch": 1674, "lr": 4.247588206710662e-05}
{"train_loss": 0.06731932610273361, "global_step": 149003, "epoch": 1674, "lr": 4.2475308932646835e-05}
{"train_loss": 0.1570151150226593, "global_step": 149004, "epoch": 1674, "lr": 4.247473579919868e-05}
{"train_loss": 0.11844730377197266, "global_step": 149005, "epoch": 1674, "lr": 4.247416266676218e-05}
{"train_loss": 0.03836502507328987, "global_step": 149006, "epoch": 1674, "lr": 4.247358953533746e-05}
{"train_loss": 0.1356888860464096, "global_step": 149007, "epoch": 1674, "lr": 4.2473016404924556e-05}
{"train_loss": 0.0941050723195076, "global_step": 149008, "epoch": 1674, "lr": 4.2472443275523576e-05}
{"train_loss": 0.12268757820129395, "global_step": 149009, "epoch": 1674, "lr": 4.247187014713458e-05}
{"train_loss": 0.06999532878398895, "global_step": 149010, "epoch": 1674, "lr": 4.247129701975765e-05}
{"train_loss": 0.05908537283539772, "global_step": 149011, "epoch": 1674, "lr": 4.247072389339286e-05}
{"train_loss": 0.08249058574438095, "global_step": 149012, "epoch": 1674, "lr": 4.247015076804029e-05}
{"train_loss": 0.09591873735189438, "global_step": 149013, "epoch": 1674, "lr": 4.246957764370002e-05}
{"train_loss": 0.0794185996055603, "global_step": 149014, "epoch": 1674, "lr": 4.2469004520372117e-05}
{"train_loss": 0.11653366684913635, "global_step": 149015, "epoch": 1674, "lr": 4.246843139805667e-05}
{"train_loss": 0.14861392974853516, "global_step": 149016, "epoch": 1674, "lr": 4.246785827675376e-05}
{"train_loss": 0.059338465332984924, "global_step": 149017, "epoch": 1674, "lr": 4.246728515646343e-05}
{"train_loss": 0.1077384427189827, "global_step": 149018, "epoch": 1674, "lr": 4.246671203718581e-05}
{"train_loss": 0.15555088222026825, "global_step": 149019, "epoch": 1674, "lr": 4.246613891892092e-05}
{"train_loss": 0.09926803410053253, "global_step": 149020, "epoch": 1674, "lr": 4.2465565801668864e-05}
{"train_loss": 0.062027983367443085, "global_step": 149021, "epoch": 1674, "lr": 4.246499268542975e-05}
{"train_loss": 0.11450802534818649, "global_step": 149022, "epoch": 1674, "lr": 4.24644195702036e-05}
{"train_loss": 0.09169124066829681, "global_step": 149023, "epoch": 1674, "lr": 4.246384645599054e-05}
{"train_loss": 0.11065208166837692, "global_step": 149024, "epoch": 1674, "lr": 4.24632733427906e-05}
{"train_loss": 0.10136516392230988, "global_step": 149025, "epoch": 1674, "lr": 4.246270023060391e-05}
{"train_loss": 0.07174200564622879, "global_step": 149026, "epoch": 1674, "lr": 4.246212711943049e-05}
{"train_loss": 0.11922962218523026, "global_step": 149027, "epoch": 1674, "lr": 4.2461554009270465e-05}
{"train_loss": 0.09562364965677261, "global_step": 149028, "epoch": 1674, "lr": 4.246098090012388e-05}
{"train_loss": 0.07258521765470505, "global_step": 149029, "epoch": 1674, "lr": 4.246040779199083e-05}
{"train_loss": 0.1401507407426834, "global_step": 149030, "epoch": 1674, "lr": 4.245983468487139e-05}
{"train_loss": 0.09555115550756454, "global_step": 149031, "epoch": 1674, "lr": 4.245926157876563e-05}
{"train_loss": 0.1776924878358841, "global_step": 149032, "epoch": 1674, "lr": 4.2458688473673626e-05}
{"train_loss": 0.12193246185779572, "global_step": 149033, "epoch": 1674, "lr": 4.245811536959546e-05}
{"train_loss": 0.08384228497743607, "global_step": 149034, "epoch": 1674, "lr": 4.2457542266531215e-05}
{"train_loss": 0.12605608999729156, "global_step": 149035, "epoch": 1674, "lr": 4.245696916448096e-05}
{"train_loss": 0.0835244208574295, "global_step": 149036, "epoch": 1674, "lr": 4.2456396063444765e-05}
{"train_loss": 0.09469516575336456, "global_step": 149037, "epoch": 1674, "lr": 4.245582296342273e-05}
{"train_loss": 0.13304175436496735, "global_step": 149038, "epoch": 1674, "lr": 4.24552498644149e-05}
{"train_loss": 0.1120515763759613, "global_step": 149039, "epoch": 1674, "lr": 4.245467676642136e-05}
{"train_loss": 0.12121793627738953, "global_step": 149040, "epoch": 1674, "lr": 4.245410366944223e-05}
{"train_loss": 0.06112457066774368, "global_step": 149041, "epoch": 1674, "lr": 4.245353057347753e-05}
{"train_loss": 0.13674934208393097, "global_step": 149042, "epoch": 1674, "lr": 4.245295747852738e-05}
{"train_loss": 0.10054676979780197, "global_step": 149043, "epoch": 1674, "lr": 4.2452384384591816e-05}
{"train_loss": 0.09538541734218597, "global_step": 149044, "epoch": 1674, "lr": 4.245181129167096e-05}
{"train_loss": 0.05250505730509758, "global_step": 149045, "epoch": 1674, "lr": 4.245123819976484e-05}
{"train_loss": 0.16331568360328674, "global_step": 149046, "epoch": 1674, "lr": 4.245066510887359e-05}
{"train_loss": 0.12594346702098846, "global_step": 149047, "epoch": 1674, "lr": 4.245009201899723e-05}
{"train_loss": 0.08641580492258072, "global_step": 149048, "epoch": 1674, "lr": 4.244951893013588e-05}
{"train_loss": 0.1337953358888626, "global_step": 149049, "epoch": 1674, "lr": 4.244894584228959e-05}
{"train_loss": 0.10051760077476501, "global_step": 149050, "epoch": 1674, "lr": 4.2448372755458455e-05}
{"train_loss": 0.08770207315683365, "global_step": 149051, "epoch": 1674, "lr": 4.244779966964254e-05}
{"train_loss": 0.10156208276748657, "global_step": 149052, "epoch": 1674, "lr": 4.244722658484193e-05}
{"train_loss": 0.15389132499694824, "global_step": 149053, "epoch": 1674, "lr": 4.244665350105669e-05}
{"train_loss": 0.10640978068113327, "global_step": 149054, "epoch": 1674, "lr": 4.2446080418286904e-05}
{"train_loss": 0.16364361345767975, "global_step": 149055, "epoch": 1674, "lr": 4.244550733653266e-05}
{"train_loss": 0.16842308640480042, "global_step": 149056, "epoch": 1674, "lr": 4.244493425579401e-05}
{"train_loss": 0.12475351244211197, "global_step": 149057, "epoch": 1674, "lr": 4.244436117607107e-05}
{"train_loss": 0.04737631976604462, "global_step": 149058, "epoch": 1674, "lr": 4.2443788097363865e-05}
{"train_loss": 0.04886474460363388, "global_step": 149059, "epoch": 1674, "lr": 4.244321501967253e-05}
{"train_loss": 0.11636363714933395, "global_step": 149060, "epoch": 1674, "lr": 4.244264194299709e-05}
{"train_loss": 0.05274796485900879, "global_step": 149061, "epoch": 1674, "lr": 4.2442068867337666e-05}
{"train_loss": 0.14119648933410645, "global_step": 149062, "epoch": 1674, "lr": 4.244149579269428e-05}
{"train_loss": 0.14424774050712585, "global_step": 149063, "epoch": 1674, "lr": 4.244092271906708e-05}
{"train_loss": 0.15726719796657562, "global_step": 149064, "epoch": 1674, "lr": 4.244034964645608e-05}
{"train_loss": 0.08595336973667145, "global_step": 149065, "epoch": 1674, "lr": 4.243977657486141e-05}
{"train_loss": 0.10525169968605042, "global_step": 149066, "epoch": 1674, "lr": 4.2439203504283085e-05}
{"train_loss": 0.11185862123966217, "global_step": 149067, "epoch": 1674, "lr": 4.243863043472125e-05}
{"train_loss": 0.09009812027215958, "global_step": 149068, "epoch": 1674, "lr": 4.2438057366175926e-05}
{"train_loss": 0.11796023696660995, "global_step": 149069, "epoch": 1674, "lr": 4.243748429864723e-05}
{"train_loss": 0.11479346454143524, "global_step": 149070, "epoch": 1674, "lr": 4.243691123213521e-05}
{"train_loss": 0.07594632357358932, "global_step": 149071, "epoch": 1674, "lr": 4.243633816663996e-05}
{"train_loss": 0.07633274793624878, "global_step": 149072, "epoch": 1674, "lr": 4.2435765102161554e-05}
{"train_loss": 0.08691875636577606, "global_step": 149073, "epoch": 1674, "lr": 4.243519203870006e-05}
{"train_loss": 0.10377990500478262, "global_step": 149074, "epoch": 1674, "lr": 4.2434618976255566e-05, "val_loss": 6.350495338439941}
{"train_loss": 0.05862796679139137, "global_step": 149075, "epoch": 1675, "lr": 4.243404591482813e-05}
{"train_loss": 0.05641356110572815, "global_step": 149076, "epoch": 1675, "lr": 4.243347285441788e-05}
{"train_loss": 0.05088188126683235, "global_step": 149077, "epoch": 1675, "lr": 4.243289979502482e-05}
{"train_loss": 0.1655711978673935, "global_step": 149078, "epoch": 1675, "lr": 4.24323267366491e-05}
{"train_loss": 0.12091158330440521, "global_step": 149079, "epoch": 1675, "lr": 4.2431753679290724e-05}
{"train_loss": 0.155509814620018, "global_step": 149080, "epoch": 1675, "lr": 4.243118062294984e-05}
{"train_loss": 0.12711526453495026, "global_step": 149081, "epoch": 1675, "lr": 4.2430607567626464e-05}
{"train_loss": 0.11550291627645493, "global_step": 149082, "epoch": 1675, "lr": 4.243003451332073e-05}
{"train_loss": 0.11593969911336899, "global_step": 149083, "epoch": 1675, "lr": 4.242946146003265e-05}
{"train_loss": 0.08363475650548935, "global_step": 149084, "epoch": 1675, "lr": 4.242888840776237e-05}
{"train_loss": 0.11130402237176895, "global_step": 149085, "epoch": 1675, "lr": 4.242831535650991e-05}
{"train_loss": 0.06551895290613174, "global_step": 149086, "epoch": 1675, "lr": 4.2427742306275386e-05}
{"train_loss": 0.06603290140628815, "global_step": 149087, "epoch": 1675, "lr": 4.242716925705884e-05}
{"train_loss": 0.0787886306643486, "global_step": 149088, "epoch": 1675, "lr": 4.2426596208860395e-05}
{"train_loss": 0.11385427415370941, "global_step": 149089, "epoch": 1675, "lr": 4.242602316168009e-05}
{"train_loss": 0.10142523050308228, "global_step": 149090, "epoch": 1675, "lr": 4.242545011551801e-05}
{"train_loss": 0.13083702325820923, "global_step": 149091, "epoch": 1675, "lr": 4.242487707037424e-05}
{"train_loss": 0.12541820108890533, "global_step": 149092, "epoch": 1675, "lr": 4.242430402624884e-05}
{"train_loss": 0.10327804088592529, "global_step": 149093, "epoch": 1675, "lr": 4.242373098314193e-05}
{"train_loss": 0.04647214338183403, "global_step": 149094, "epoch": 1675, "lr": 4.242315794105353e-05}
{"train_loss": 0.07969119399785995, "global_step": 149095, "epoch": 1675, "lr": 4.242258489998376e-05}
{"train_loss": 0.15879866480827332, "global_step": 149096, "epoch": 1675, "lr": 4.242201185993267e-05}
{"train_loss": 0.10627824068069458, "global_step": 149097, "epoch": 1675, "lr": 4.242143882090036e-05}
{"train_loss": 0.11733917891979218, "global_step": 149098, "epoch": 1675, "lr": 4.242086578288687e-05}
{"train_loss": 0.11228219419717789, "global_step": 149099, "epoch": 1675, "lr": 4.242029274589233e-05}
{"train_loss": 0.13884975016117096, "global_step": 149100, "epoch": 1675, "lr": 4.2419719709916773e-05}
{"train_loss": 0.06938683986663818, "global_step": 149101, "epoch": 1675, "lr": 4.2419146674960306e-05}
{"train_loss": 0.1264144331216812, "global_step": 149102, "epoch": 1675, "lr": 4.2418573641022974e-05}
{"train_loss": 0.13420765101909637, "global_step": 149103, "epoch": 1675, "lr": 4.241800060810489e-05}
{"train_loss": 0.09424685686826706, "global_step": 149104, "epoch": 1675, "lr": 4.24174275762061e-05}
{"train_loss": 0.07491923868656158, "global_step": 149105, "epoch": 1675, "lr": 4.2416854545326714e-05}
{"train_loss": 0.09931599348783493, "global_step": 149106, "epoch": 1675, "lr": 4.241628151546676e-05}
{"train_loss": 0.09087298810482025, "global_step": 149107, "epoch": 1675, "lr": 4.2415708486626374e-05}
{"train_loss": 0.12745748460292816, "global_step": 149108, "epoch": 1675, "lr": 4.241513545880559e-05}
{"train_loss": 0.08455616980791092, "global_step": 149109, "epoch": 1675, "lr": 4.2414562432004504e-05}
{"train_loss": 0.07945745438337326, "global_step": 149110, "epoch": 1675, "lr": 4.241398940622318e-05}
{"train_loss": 0.07480024546384811, "global_step": 149111, "epoch": 1675, "lr": 4.241341638146169e-05}
{"train_loss": 0.06822153180837631, "global_step": 149112, "epoch": 1675, "lr": 4.241284335772015e-05}
{"train_loss": 0.0920976772904396, "global_step": 149113, "epoch": 1675, "lr": 4.241227033499859e-05}
{"train_loss": 0.07661138474941254, "global_step": 149114, "epoch": 1675, "lr": 4.241169731329713e-05}
{"train_loss": 0.1447211354970932, "global_step": 149115, "epoch": 1675, "lr": 4.24111242926158e-05}
{"train_loss": 0.08596441149711609, "global_step": 149116, "epoch": 1675, "lr": 4.2410551272954723e-05}
{"train_loss": 0.08837812393903732, "global_step": 149117, "epoch": 1675, "lr": 4.240997825431394e-05}
{"train_loss": 0.13422991335391998, "global_step": 149118, "epoch": 1675, "lr": 4.240940523669355e-05}
{"train_loss": 0.09752783924341202, "global_step": 149119, "epoch": 1675, "lr": 4.240883222009361e-05}
{"train_loss": 0.1344386637210846, "global_step": 149120, "epoch": 1675, "lr": 4.240825920451424e-05}
{"train_loss": 0.10241732001304626, "global_step": 149121, "epoch": 1675, "lr": 4.240768618995545e-05}
{"train_loss": 0.07377377152442932, "global_step": 149122, "epoch": 1675, "lr": 4.240711317641739e-05}
{"train_loss": 0.144303560256958, "global_step": 149123, "epoch": 1675, "lr": 4.240654016390007e-05}
{"train_loss": 0.13899527490139008, "global_step": 149124, "epoch": 1675, "lr": 4.2405967152403626e-05}
{"train_loss": 0.0825563371181488, "global_step": 149125, "epoch": 1675, "lr": 4.240539414192808e-05}
{"train_loss": 0.17677392065525055, "global_step": 149126, "epoch": 1675, "lr": 4.240482113247356e-05}
{"train_loss": 0.12351773679256439, "global_step": 149127, "epoch": 1675, "lr": 4.2404248124040116e-05}
{"train_loss": 0.08514520525932312, "global_step": 149128, "epoch": 1675, "lr": 4.2403675116627804e-05}
{"train_loss": 0.12453031539916992, "global_step": 149129, "epoch": 1675, "lr": 4.240310211023675e-05}
{"train_loss": 0.10511962324380875, "global_step": 149130, "epoch": 1675, "lr": 4.2402529104866994e-05}
{"train_loss": 0.11344361305236816, "global_step": 149131, "epoch": 1675, "lr": 4.240195610051864e-05}
{"train_loss": 0.15167255699634552, "global_step": 149132, "epoch": 1675, "lr": 4.240138309719173e-05}
{"train_loss": 0.14184577763080597, "global_step": 149133, "epoch": 1675, "lr": 4.2400810094886375e-05}
{"train_loss": 0.11691048741340637, "global_step": 149134, "epoch": 1675, "lr": 4.240023709360263e-05}
{"train_loss": 0.08735629171133041, "global_step": 149135, "epoch": 1675, "lr": 4.23996640933406e-05}
{"train_loss": 0.11746574938297272, "global_step": 149136, "epoch": 1675, "lr": 4.239909109410031e-05}
{"train_loss": 0.1414065957069397, "global_step": 149137, "epoch": 1675, "lr": 4.23985180958819e-05}
{"train_loss": 0.13483066856861115, "global_step": 149138, "epoch": 1675, "lr": 4.2397945098685396e-05}
{"train_loss": 0.12745444476604462, "global_step": 149139, "epoch": 1675, "lr": 4.2397372102510914e-05}
{"train_loss": 0.054293904453516006, "global_step": 149140, "epoch": 1675, "lr": 4.239679910735849e-05}
{"train_loss": 0.15137912333011627, "global_step": 149141, "epoch": 1675, "lr": 4.239622611322825e-05}
{"train_loss": 0.1446147859096527, "global_step": 149142, "epoch": 1675, "lr": 4.239565312012022e-05}
{"train_loss": 0.1128205731511116, "global_step": 149143, "epoch": 1675, "lr": 4.239508012803453e-05}
{"train_loss": 0.14735190570354462, "global_step": 149144, "epoch": 1675, "lr": 4.23945071369712e-05}
{"train_loss": 0.05524854362010956, "global_step": 149145, "epoch": 1675, "lr": 4.2393934146930354e-05}
{"train_loss": 0.13369502127170563, "global_step": 149146, "epoch": 1675, "lr": 4.239336115791204e-05}
{"train_loss": 0.10034160315990448, "global_step": 149147, "epoch": 1675, "lr": 4.239278816991634e-05}
{"train_loss": 0.05990029126405716, "global_step": 149148, "epoch": 1675, "lr": 4.239221518294336e-05}
{"train_loss": 0.09554636478424072, "global_step": 149149, "epoch": 1675, "lr": 4.239164219699313e-05}
{"train_loss": 0.11658868193626404, "global_step": 149150, "epoch": 1675, "lr": 4.239106921206577e-05}
{"train_loss": 0.07681585103273392, "global_step": 149151, "epoch": 1675, "lr": 4.239049622816132e-05}
{"train_loss": 0.15128853917121887, "global_step": 149152, "epoch": 1675, "lr": 4.238992324527989e-05}
{"train_loss": 0.08377896249294281, "global_step": 149153, "epoch": 1675, "lr": 4.2389350263421535e-05}
{"train_loss": 0.08641870319843292, "global_step": 149154, "epoch": 1675, "lr": 4.238877728258635e-05}
{"train_loss": 0.12227138876914978, "global_step": 149155, "epoch": 1675, "lr": 4.238820430277438e-05}
{"train_loss": 0.11989620327949524, "global_step": 149156, "epoch": 1675, "lr": 4.2387631323985744e-05}
{"train_loss": 0.07135064154863358, "global_step": 149157, "epoch": 1675, "lr": 4.238705834622048e-05}
{"train_loss": 0.083926260471344, "global_step": 149158, "epoch": 1675, "lr": 4.238648536947871e-05}
{"train_loss": 0.1468212604522705, "global_step": 149159, "epoch": 1675, "lr": 4.238591239376045e-05}
{"train_loss": 0.04880887642502785, "global_step": 149160, "epoch": 1675, "lr": 4.238533941906584e-05}
{"train_loss": 0.0617477223277092, "global_step": 149161, "epoch": 1675, "lr": 4.238476644539491e-05}
{"train_loss": 0.08291421830654144, "global_step": 149162, "epoch": 1675, "lr": 4.238419347274777e-05}
{"train_loss": 0.1051326176005133, "global_step": 149163, "epoch": 1675, "lr": 4.2383620501124465e-05, "val_loss": 6.276211738586426, "train_action_mse_error": 18.226337432861328}
{"train_loss": 0.07307247072458267, "global_step": 149164, "epoch": 1676, "lr": 4.238304753052511e-05}
{"train_loss": 0.1608690321445465, "global_step": 149165, "epoch": 1676, "lr": 4.238247456094975e-05}
{"train_loss": 0.0948607474565506, "global_step": 149166, "epoch": 1676, "lr": 4.238190159239846e-05}
{"train_loss": 0.10765019804239273, "global_step": 149167, "epoch": 1676, "lr": 4.2381328624871354e-05}
{"train_loss": 0.10406429320573807, "global_step": 149168, "epoch": 1676, "lr": 4.2380755658368463e-05}
{"train_loss": 0.0872800201177597, "global_step": 149169, "epoch": 1676, "lr": 4.2380182692889914e-05}
{"train_loss": 0.09682701528072357, "global_step": 149170, "epoch": 1676, "lr": 4.237960972843573e-05}
{"train_loss": 0.09658980369567871, "global_step": 149171, "epoch": 1676, "lr": 4.237903676500603e-05}
{"train_loss": 0.11514484882354736, "global_step": 149172, "epoch": 1676, "lr": 4.2378463802600865e-05}
{"train_loss": 0.07230038195848465, "global_step": 149173, "epoch": 1676, "lr": 4.237789084122034e-05}
{"train_loss": 0.11976097524166107, "global_step": 149174, "epoch": 1676, "lr": 4.23773178808645e-05}
{"train_loss": 0.10213641822338104, "global_step": 149175, "epoch": 1676, "lr": 4.237674492153345e-05}
{"train_loss": 0.08099596202373505, "global_step": 149176, "epoch": 1676, "lr": 4.237617196322723e-05}
{"train_loss": 0.0979660227894783, "global_step": 149177, "epoch": 1676, "lr": 4.2375599005945967e-05}
{"train_loss": 0.08916550129652023, "global_step": 149178, "epoch": 1676, "lr": 4.2375026049689695e-05}
{"train_loss": 0.1124323308467865, "global_step": 149179, "epoch": 1676, "lr": 4.237445309445852e-05}
{"train_loss": 0.11850174516439438, "global_step": 149180, "epoch": 1676, "lr": 4.2373880140252484e-05}
{"train_loss": 0.1194082722067833, "global_step": 149181, "epoch": 1676, "lr": 4.237330718707172e-05}
{"train_loss": 0.08903908729553223, "global_step": 149182, "epoch": 1676, "lr": 4.237273423491625e-05}
{"train_loss": 0.10382524132728577, "global_step": 149183, "epoch": 1676, "lr": 4.237216128378617e-05}
{"train_loss": 0.09249746054410934, "global_step": 149184, "epoch": 1676, "lr": 4.237158833368157e-05}
{"train_loss": 0.06860243529081345, "global_step": 149185, "epoch": 1676, "lr": 4.2371015384602505e-05}
{"train_loss": 0.060950227081775665, "global_step": 149186, "epoch": 1676, "lr": 4.237044243654909e-05}
{"train_loss": 0.09844633936882019, "global_step": 149187, "epoch": 1676, "lr": 4.236986948952135e-05}
{"train_loss": 0.13402213156223297, "global_step": 149188, "epoch": 1676, "lr": 4.236929654351941e-05}
{"train_loss": 0.0830451250076294, "global_step": 149189, "epoch": 1676, "lr": 4.23687235985433e-05}
{"train_loss": 0.07900292426347733, "global_step": 149190, "epoch": 1676, "lr": 4.236815065459314e-05}
{"train_loss": 0.13326729834079742, "global_step": 149191, "epoch": 1676, "lr": 4.236757771166898e-05}
{"train_loss": 0.12117786705493927, "global_step": 149192, "epoch": 1676, "lr": 4.236700476977092e-05}
{"train_loss": 0.09251775592565536, "global_step": 149193, "epoch": 1676, "lr": 4.2366431828899e-05}
{"train_loss": 0.1470227837562561, "global_step": 149194, "epoch": 1676, "lr": 4.236585888905334e-05}
{"train_loss": 0.1171102374792099, "global_step": 149195, "epoch": 1676, "lr": 4.236528595023398e-05}
{"train_loss": 0.09106719493865967, "global_step": 149196, "epoch": 1676, "lr": 4.236471301244104e-05}
{"train_loss": 0.08886055648326874, "global_step": 149197, "epoch": 1676, "lr": 4.2364140075674546e-05}
{"train_loss": 0.06365785747766495, "global_step": 149198, "epoch": 1676, "lr": 4.2363567139934625e-05}
{"train_loss": 0.10421718657016754, "global_step": 149199, "epoch": 1676, "lr": 4.23629942052213e-05}
{"train_loss": 0.12588007748126984, "global_step": 149200, "epoch": 1676, "lr": 4.2362421271534704e-05}
{"train_loss": 0.19540686905384064, "global_step": 149201, "epoch": 1676, "lr": 4.2361848338874866e-05}
{"train_loss": 0.10097828507423401, "global_step": 149202, "epoch": 1676, "lr": 4.236127540724189e-05}
{"train_loss": 0.12679798901081085, "global_step": 149203, "epoch": 1676, "lr": 4.236070247663585e-05}
{"train_loss": 0.08644485473632812, "global_step": 149204, "epoch": 1676, "lr": 4.2360129547056826e-05}
{"train_loss": 0.07185711711645126, "global_step": 149205, "epoch": 1676, "lr": 4.2359556618504894e-05}
{"train_loss": 0.11706581711769104, "global_step": 149206, "epoch": 1676, "lr": 4.23589836909801e-05}
{"train_loss": 0.10238156467676163, "global_step": 149207, "epoch": 1676, "lr": 4.235841076448258e-05}
{"train_loss": 0.07120347023010254, "global_step": 149208, "epoch": 1676, "lr": 4.2357837839012356e-05}
{"train_loss": 0.13378570973873138, "global_step": 149209, "epoch": 1676, "lr": 4.2357264914569544e-05}
{"train_loss": 0.07189017534255981, "global_step": 149210, "epoch": 1676, "lr": 4.235669199115418e-05}
{"train_loss": 0.12191631644964218, "global_step": 149211, "epoch": 1676, "lr": 4.2356119068766395e-05}
{"train_loss": 0.12942621111869812, "global_step": 149212, "epoch": 1676, "lr": 4.2355546147406214e-05}
{"train_loss": 0.14211392402648926, "global_step": 149213, "epoch": 1676, "lr": 4.2354973227073754e-05}
{"train_loss": 0.20173880457878113, "global_step": 149214, "epoch": 1676, "lr": 4.235440030776906e-05}
{"train_loss": 0.12781241536140442, "global_step": 149215, "epoch": 1676, "lr": 4.235382738949224e-05}
{"train_loss": 0.09370135515928268, "global_step": 149216, "epoch": 1676, "lr": 4.235325447224334e-05}
{"train_loss": 0.08715401589870453, "global_step": 149217, "epoch": 1676, "lr": 4.235268155602246e-05}
{"train_loss": 0.14457303285598755, "global_step": 149218, "epoch": 1676, "lr": 4.235210864082966e-05}
{"train_loss": 0.1116529107093811, "global_step": 149219, "epoch": 1676, "lr": 4.2351535726665024e-05}
{"train_loss": 0.19701547920703888, "global_step": 149220, "epoch": 1676, "lr": 4.235096281352864e-05}
{"train_loss": 0.10733354091644287, "global_step": 149221, "epoch": 1676, "lr": 4.235038990142057e-05}
{"train_loss": 0.06577513366937637, "global_step": 149222, "epoch": 1676, "lr": 4.234981699034091e-05}
{"train_loss": 0.13545210659503937, "global_step": 149223, "epoch": 1676, "lr": 4.234924408028971e-05}
{"train_loss": 0.11668402701616287, "global_step": 149224, "epoch": 1676, "lr": 4.234867117126707e-05}
{"train_loss": 0.1771148294210434, "global_step": 149225, "epoch": 1676, "lr": 4.234809826327305e-05}
{"train_loss": 0.047147370874881744, "global_step": 149226, "epoch": 1676, "lr": 4.2347525356307746e-05}
{"train_loss": 0.0971902385354042, "global_step": 149227, "epoch": 1676, "lr": 4.234695245037121e-05}
{"train_loss": 0.05629042536020279, "global_step": 149228, "epoch": 1676, "lr": 4.2346379545463544e-05}
{"train_loss": 0.11639728397130966, "global_step": 149229, "epoch": 1676, "lr": 4.23458066415848e-05}
{"train_loss": 0.14204035699367523, "global_step": 149230, "epoch": 1676, "lr": 4.234523373873509e-05}
{"train_loss": 0.06817420572042465, "global_step": 149231, "epoch": 1676, "lr": 4.2344660836914446e-05}
{"train_loss": 0.09545756131410599, "global_step": 149232, "epoch": 1676, "lr": 4.234408793612299e-05}
{"train_loss": 0.07880557328462601, "global_step": 149233, "epoch": 1676, "lr": 4.234351503636076e-05}
{"train_loss": 0.1375458687543869, "global_step": 149234, "epoch": 1676, "lr": 4.234294213762787e-05}
{"train_loss": 0.10255807638168335, "global_step": 149235, "epoch": 1676, "lr": 4.2342369239924364e-05}
{"train_loss": 0.1203589215874672, "global_step": 149236, "epoch": 1676, "lr": 4.234179634325035e-05}
{"train_loss": 0.0881035178899765, "global_step": 149237, "epoch": 1676, "lr": 4.234122344760586e-05}
{"train_loss": 0.14234371483325958, "global_step": 149238, "epoch": 1676, "lr": 4.234065055299101e-05}
{"train_loss": 0.08954991400241852, "global_step": 149239, "epoch": 1676, "lr": 4.2340077659405875e-05}
{"train_loss": 0.08695811033248901, "global_step": 149240, "epoch": 1676, "lr": 4.233950476685052e-05}
{"train_loss": 0.09044226258993149, "global_step": 149241, "epoch": 1676, "lr": 4.233893187532502e-05}
{"train_loss": 0.07880769670009613, "global_step": 149242, "epoch": 1676, "lr": 4.233835898482946e-05}
{"train_loss": 0.18562258780002594, "global_step": 149243, "epoch": 1676, "lr": 4.233778609536393e-05}
{"train_loss": 0.08699442446231842, "global_step": 149244, "epoch": 1676, "lr": 4.233721320692847e-05}
{"train_loss": 0.1337234526872635, "global_step": 149245, "epoch": 1676, "lr": 4.233664031952319e-05}
{"train_loss": 0.07775641977787018, "global_step": 149246, "epoch": 1676, "lr": 4.233606743314814e-05}
{"train_loss": 0.11618328094482422, "global_step": 149247, "epoch": 1676, "lr": 4.233549454780343e-05}
{"train_loss": 0.11046988517045975, "global_step": 149248, "epoch": 1676, "lr": 4.2334921663489104e-05}
{"train_loss": 0.05143246054649353, "global_step": 149249, "epoch": 1676, "lr": 4.233434878020528e-05}
{"train_loss": 0.06067885458469391, "global_step": 149250, "epoch": 1676, "lr": 4.2333775897951977e-05}
{"train_loss": 0.09181773662567139, "global_step": 149251, "epoch": 1676, "lr": 4.233320301672933e-05}
{"train_loss": 0.10624388412813122, "global_step": 149252, "epoch": 1676, "lr": 4.233263013653737e-05, "val_loss": 6.32265043258667}
{"train_loss": 0.1219821497797966, "global_step": 149253, "epoch": 1677, "lr": 4.2332057257376216e-05}
{"train_loss": 0.07277945429086685, "global_step": 149254, "epoch": 1677, "lr": 4.23314843792459e-05}
{"train_loss": 0.07858100533485413, "global_step": 149255, "epoch": 1677, "lr": 4.2330911502146523e-05}
{"train_loss": 0.130490243434906, "global_step": 149256, "epoch": 1677, "lr": 4.233033862607818e-05}
{"train_loss": 0.11124800145626068, "global_step": 149257, "epoch": 1677, "lr": 4.2329765751040926e-05}
{"train_loss": 0.14763988554477692, "global_step": 149258, "epoch": 1677, "lr": 4.232919287703484e-05}
{"train_loss": 0.11492577195167542, "global_step": 149259, "epoch": 1677, "lr": 4.232862000405999e-05}
{"train_loss": 0.12437190115451813, "global_step": 149260, "epoch": 1677, "lr": 4.232804713211648e-05}
{"train_loss": 0.03983285278081894, "global_step": 149261, "epoch": 1677, "lr": 4.232747426120436e-05}
{"train_loss": 0.13355837762355804, "global_step": 149262, "epoch": 1677, "lr": 4.232690139132373e-05}
{"train_loss": 0.12141452729701996, "global_step": 149263, "epoch": 1677, "lr": 4.232632852247463e-05}
{"train_loss": 0.10805539041757584, "global_step": 149264, "epoch": 1677, "lr": 4.2325755654657187e-05}
{"train_loss": 0.1496225744485855, "global_step": 149265, "epoch": 1677, "lr": 4.232518278787144e-05}
{"train_loss": 0.08786782622337341, "global_step": 149266, "epoch": 1677, "lr": 4.232460992211749e-05}
{"train_loss": 0.1942444145679474, "global_step": 149267, "epoch": 1677, "lr": 4.232403705739538e-05}
{"train_loss": 0.08780373632907867, "global_step": 149268, "epoch": 1677, "lr": 4.232346419370523e-05}
{"train_loss": 0.08812607824802399, "global_step": 149269, "epoch": 1677, "lr": 4.232289133104708e-05}
{"train_loss": 0.1064375787973404, "global_step": 149270, "epoch": 1677, "lr": 4.232231846942105e-05}
{"train_loss": 0.09231846034526825, "global_step": 149271, "epoch": 1677, "lr": 4.232174560882716e-05}
{"train_loss": 0.1432133913040161, "global_step": 149272, "epoch": 1677, "lr": 4.232117274926554e-05}
{"train_loss": 0.1143546998500824, "global_step": 149273, "epoch": 1677, "lr": 4.232059989073622e-05}
{"train_loss": 0.1004132404923439, "global_step": 149274, "epoch": 1677, "lr": 4.232002703323931e-05}
{"train_loss": 0.12599073350429535, "global_step": 149275, "epoch": 1677, "lr": 4.2319454176774896e-05}
{"train_loss": 0.129696786403656, "global_step": 149276, "epoch": 1677, "lr": 4.231888132134303e-05}
{"train_loss": 0.14575399458408356, "global_step": 149277, "epoch": 1677, "lr": 4.23183084669438e-05}
{"train_loss": 0.0616014264523983, "global_step": 149278, "epoch": 1677, "lr": 4.231773561357727e-05}
{"train_loss": 0.12674908339977264, "global_step": 149279, "epoch": 1677, "lr": 4.231716276124353e-05}
{"train_loss": 0.1106758713722229, "global_step": 149280, "epoch": 1677, "lr": 4.2316589909942664e-05}
{"train_loss": 0.07965981215238571, "global_step": 149281, "epoch": 1677, "lr": 4.231601705967473e-05}
{"train_loss": 0.12183300405740738, "global_step": 149282, "epoch": 1677, "lr": 4.231544421043981e-05}
{"train_loss": 0.10474447160959244, "global_step": 149283, "epoch": 1677, "lr": 4.2314871362238e-05}
{"train_loss": 0.08488587290048599, "global_step": 149284, "epoch": 1677, "lr": 4.231429851506934e-05}
{"train_loss": 0.10517893731594086, "global_step": 149285, "epoch": 1677, "lr": 4.231372566893395e-05}
{"train_loss": 0.10228267312049866, "global_step": 149286, "epoch": 1677, "lr": 4.231315282383186e-05}
{"train_loss": 0.12149626761674881, "global_step": 149287, "epoch": 1677, "lr": 4.231257997976321e-05}
{"train_loss": 0.08829285204410553, "global_step": 149288, "epoch": 1677, "lr": 4.231200713672801e-05}
{"train_loss": 0.22656209766864777, "global_step": 149289, "epoch": 1677, "lr": 4.2311434294726384e-05}
{"train_loss": 0.06483818590641022, "global_step": 149290, "epoch": 1677, "lr": 4.2310861453758364e-05}
{"train_loss": 0.07835419476032257, "global_step": 149291, "epoch": 1677, "lr": 4.231028861382409e-05}
{"train_loss": 0.09323477745056152, "global_step": 149292, "epoch": 1677, "lr": 4.2309715774923576e-05}
{"train_loss": 0.1490594744682312, "global_step": 149293, "epoch": 1677, "lr": 4.230914293705693e-05}
{"train_loss": 0.11320282518863678, "global_step": 149294, "epoch": 1677, "lr": 4.230857010022424e-05}
{"train_loss": 0.13462397456169128, "global_step": 149295, "epoch": 1677, "lr": 4.230799726442556e-05}
{"train_loss": 0.20289823412895203, "global_step": 149296, "epoch": 1677, "lr": 4.230742442966099e-05}
{"train_loss": 0.1996380090713501, "global_step": 149297, "epoch": 1677, "lr": 4.2306851595930585e-05}
{"train_loss": 0.10991387814283371, "global_step": 149298, "epoch": 1677, "lr": 4.230627876323443e-05}
{"train_loss": 0.15869808197021484, "global_step": 149299, "epoch": 1677, "lr": 4.23057059315726e-05}
{"train_loss": 0.12323515862226486, "global_step": 149300, "epoch": 1677, "lr": 4.230513310094517e-05}
{"train_loss": 0.12198156863451004, "global_step": 149301, "epoch": 1677, "lr": 4.230456027135223e-05}
{"train_loss": 0.15797753632068634, "global_step": 149302, "epoch": 1677, "lr": 4.230398744279386e-05}
{"train_loss": 0.09404071420431137, "global_step": 149303, "epoch": 1677, "lr": 4.230341461527009e-05}
{"train_loss": 0.06897947192192078, "global_step": 149304, "epoch": 1677, "lr": 4.2302841788781065e-05}
{"train_loss": 0.10362767428159714, "global_step": 149305, "epoch": 1677, "lr": 4.23022689633268e-05}
{"train_loss": 0.1042037159204483, "global_step": 149306, "epoch": 1677, "lr": 4.230169613890743e-05}
{"train_loss": 0.07280676811933517, "global_step": 149307, "epoch": 1677, "lr": 4.2301123315522975e-05}
{"train_loss": 0.10908686369657516, "global_step": 149308, "epoch": 1677, "lr": 4.2300550493173566e-05}
{"train_loss": 0.1377057284116745, "global_step": 149309, "epoch": 1677, "lr": 4.229997767185924e-05}
{"train_loss": 0.0887570008635521, "global_step": 149310, "epoch": 1677, "lr": 4.229940485158007e-05}
{"train_loss": 0.16573403775691986, "global_step": 149311, "epoch": 1677, "lr": 4.229883203233619e-05}
{"train_loss": 0.12411843985319138, "global_step": 149312, "epoch": 1677, "lr": 4.229825921412761e-05}
{"train_loss": 0.15617790818214417, "global_step": 149313, "epoch": 1677, "lr": 4.229768639695445e-05}
{"train_loss": 0.15848205983638763, "global_step": 149314, "epoch": 1677, "lr": 4.229711358081676e-05}
{"train_loss": 0.096082903444767, "global_step": 149315, "epoch": 1677, "lr": 4.229654076571465e-05}
{"train_loss": 0.08635899424552917, "global_step": 149316, "epoch": 1677, "lr": 4.2295967951648156e-05}
{"train_loss": 0.09819149971008301, "global_step": 149317, "epoch": 1677, "lr": 4.229539513861739e-05}
{"train_loss": 0.09930390119552612, "global_step": 149318, "epoch": 1677, "lr": 4.2294822326622403e-05}
{"train_loss": 0.21740718185901642, "global_step": 149319, "epoch": 1677, "lr": 4.2294249515663284e-05}
{"train_loss": 0.12732236087322235, "global_step": 149320, "epoch": 1677, "lr": 4.229367670574011e-05}
{"train_loss": 0.08115221560001373, "global_step": 149321, "epoch": 1677, "lr": 4.229310389685297e-05}
{"train_loss": 0.120537668466568, "global_step": 149322, "epoch": 1677, "lr": 4.2292531089001906e-05}
{"train_loss": 0.07117748260498047, "global_step": 149323, "epoch": 1677, "lr": 4.229195828218705e-05}
{"train_loss": 0.11307021975517273, "global_step": 149324, "epoch": 1677, "lr": 4.229138547640841e-05}
{"train_loss": 0.08894862979650497, "global_step": 149325, "epoch": 1677, "lr": 4.2290812671666124e-05}
{"train_loss": 0.11511503905057907, "global_step": 149326, "epoch": 1677, "lr": 4.229023986796022e-05}
{"train_loss": 0.08925875276327133, "global_step": 149327, "epoch": 1677, "lr": 4.228966706529083e-05}
{"train_loss": 0.07299699634313583, "global_step": 149328, "epoch": 1677, "lr": 4.228909426365797e-05}
{"train_loss": 0.10858959704637527, "global_step": 149329, "epoch": 1677, "lr": 4.228852146306175e-05}
{"train_loss": 0.11602733284235, "global_step": 149330, "epoch": 1677, "lr": 4.228794866350226e-05}
{"train_loss": 0.10443364083766937, "global_step": 149331, "epoch": 1677, "lr": 4.228737586497955e-05}
{"train_loss": 0.10618393868207932, "global_step": 149332, "epoch": 1677, "lr": 4.228680306749373e-05}
{"train_loss": 0.1008734479546547, "global_step": 149333, "epoch": 1677, "lr": 4.228623027104482e-05}
{"train_loss": 0.10651979595422745, "global_step": 149334, "epoch": 1677, "lr": 4.2285657475632964e-05}
{"train_loss": 0.07274722307920456, "global_step": 149335, "epoch": 1677, "lr": 4.22850846812582e-05}
{"train_loss": 0.13368161022663116, "global_step": 149336, "epoch": 1677, "lr": 4.2284511887920605e-05}
{"train_loss": 0.11707843095064163, "global_step": 149337, "epoch": 1677, "lr": 4.228393909562027e-05}
{"train_loss": 0.1472790241241455, "global_step": 149338, "epoch": 1677, "lr": 4.2283366304357264e-05}
{"train_loss": 0.0864676684141159, "global_step": 149339, "epoch": 1677, "lr": 4.228279351413166e-05}
{"train_loss": 0.1327032446861267, "global_step": 149340, "epoch": 1677, "lr": 4.228222072494356e-05}
{"train_loss": 0.11409610874030028, "global_step": 149341, "epoch": 1677, "lr": 4.228164793679299e-05, "val_loss": 5.99057149887085}
{"train_loss": 0.07041370868682861, "global_step": 149342, "epoch": 1678, "lr": 4.228107514968009e-05}
{"train_loss": 0.07994118332862854, "global_step": 149343, "epoch": 1678, "lr": 4.228050236360487e-05}
{"train_loss": 0.09268948435783386, "global_step": 149344, "epoch": 1678, "lr": 4.227992957856748e-05}
{"train_loss": 0.13111336529254913, "global_step": 149345, "epoch": 1678, "lr": 4.227935679456792e-05}
{"train_loss": 0.1093350350856781, "global_step": 149346, "epoch": 1678, "lr": 4.227878401160632e-05}
{"train_loss": 0.1248370110988617, "global_step": 149347, "epoch": 1678, "lr": 4.227821122968276e-05}
{"train_loss": 0.1254516839981079, "global_step": 149348, "epoch": 1678, "lr": 4.2277638448797284e-05}
{"train_loss": 0.13631339371204376, "global_step": 149349, "epoch": 1678, "lr": 4.227706566895e-05}
{"train_loss": 0.11262889951467514, "global_step": 149350, "epoch": 1678, "lr": 4.227649289014095e-05}
{"train_loss": 0.09869497269392014, "global_step": 149351, "epoch": 1678, "lr": 4.227592011237026e-05}
{"train_loss": 0.0892057865858078, "global_step": 149352, "epoch": 1678, "lr": 4.227534733563795e-05}
{"train_loss": 0.1802489310503006, "global_step": 149353, "epoch": 1678, "lr": 4.2274774559944146e-05}
{"train_loss": 0.08302617818117142, "global_step": 149354, "epoch": 1678, "lr": 4.22742017852889e-05}
{"train_loss": 0.06449256837368011, "global_step": 149355, "epoch": 1678, "lr": 4.227362901167229e-05}
{"train_loss": 0.0790025144815445, "global_step": 149356, "epoch": 1678, "lr": 4.2273056239094395e-05}
{"train_loss": 0.11313749849796295, "global_step": 149357, "epoch": 1678, "lr": 4.227248346755529e-05}
{"train_loss": 0.06070340424776077, "global_step": 149358, "epoch": 1678, "lr": 4.227191069705506e-05}
{"train_loss": 0.1469285935163498, "global_step": 149359, "epoch": 1678, "lr": 4.227133792759379e-05}
{"train_loss": 0.1032051295042038, "global_step": 149360, "epoch": 1678, "lr": 4.2270765159171524e-05}
{"train_loss": 0.08560197055339813, "global_step": 149361, "epoch": 1678, "lr": 4.227019239178838e-05}
{"train_loss": 0.0890200212597847, "global_step": 149362, "epoch": 1678, "lr": 4.226961962544438e-05}
{"train_loss": 0.06437734514474869, "global_step": 149363, "epoch": 1678, "lr": 4.2269046860139675e-05}
{"train_loss": 0.08055335283279419, "global_step": 149364, "epoch": 1678, "lr": 4.226847409587427e-05}
{"train_loss": 0.10169441252946854, "global_step": 149365, "epoch": 1678, "lr": 4.226790133264828e-05}
{"train_loss": 0.12313808500766754, "global_step": 149366, "epoch": 1678, "lr": 4.226732857046179e-05}
{"train_loss": 0.06760488450527191, "global_step": 149367, "epoch": 1678, "lr": 4.226675580931485e-05}
{"train_loss": 0.1387985348701477, "global_step": 149368, "epoch": 1678, "lr": 4.226618304920758e-05}
{"train_loss": 0.13566626608371735, "global_step": 149369, "epoch": 1678, "lr": 4.226561029013999e-05}
{"train_loss": 0.10147419571876526, "global_step": 149370, "epoch": 1678, "lr": 4.226503753211222e-05}
{"train_loss": 0.10172359645366669, "global_step": 149371, "epoch": 1678, "lr": 4.22644647751243e-05}
{"train_loss": 0.10010839253664017, "global_step": 149372, "epoch": 1678, "lr": 4.226389201917636e-05}
{"train_loss": 0.10898621380329132, "global_step": 149373, "epoch": 1678, "lr": 4.226331926426842e-05}
{"train_loss": 0.13984620571136475, "global_step": 149374, "epoch": 1678, "lr": 4.2262746510400584e-05}
{"train_loss": 0.10896671563386917, "global_step": 149375, "epoch": 1678, "lr": 4.2262173757572935e-05}
{"train_loss": 0.09319821745157242, "global_step": 149376, "epoch": 1678, "lr": 4.226160100578554e-05}
{"train_loss": 0.14482933282852173, "global_step": 149377, "epoch": 1678, "lr": 4.2261028255038474e-05}
{"train_loss": 0.06228085607290268, "global_step": 149378, "epoch": 1678, "lr": 4.2260455505331837e-05}
{"train_loss": 0.08161357790231705, "global_step": 149379, "epoch": 1678, "lr": 4.225988275666566e-05}
{"train_loss": 0.043314386159181595, "global_step": 149380, "epoch": 1678, "lr": 4.2259310009040075e-05}
{"train_loss": 0.10846492648124695, "global_step": 149381, "epoch": 1678, "lr": 4.225873726245511e-05}
{"train_loss": 0.13321781158447266, "global_step": 149382, "epoch": 1678, "lr": 4.225816451691085e-05}
{"train_loss": 0.056738436222076416, "global_step": 149383, "epoch": 1678, "lr": 4.2257591772407416e-05}
{"train_loss": 0.10703416168689728, "global_step": 149384, "epoch": 1678, "lr": 4.2257019028944834e-05}
{"train_loss": 0.09452523291110992, "global_step": 149385, "epoch": 1678, "lr": 4.225644628652322e-05}
{"train_loss": 0.06590285152196884, "global_step": 149386, "epoch": 1678, "lr": 4.2255873545142615e-05}
{"train_loss": 0.08496227860450745, "global_step": 149387, "epoch": 1678, "lr": 4.225530080480313e-05}
{"train_loss": 0.14009059965610504, "global_step": 149388, "epoch": 1678, "lr": 4.225472806550481e-05}
{"train_loss": 0.09534832835197449, "global_step": 149389, "epoch": 1678, "lr": 4.225415532724777e-05}
{"train_loss": 0.1056196466088295, "global_step": 149390, "epoch": 1678, "lr": 4.225358259003204e-05}
{"train_loss": 0.09347765147686005, "global_step": 149391, "epoch": 1678, "lr": 4.225300985385774e-05}
{"train_loss": 0.099298395216465, "global_step": 149392, "epoch": 1678, "lr": 4.225243711872491e-05}
{"train_loss": 0.07402080297470093, "global_step": 149393, "epoch": 1678, "lr": 4.225186438463366e-05}
{"train_loss": 0.06738035380840302, "global_step": 149394, "epoch": 1678, "lr": 4.225129165158405e-05}
{"train_loss": 0.04574073851108551, "global_step": 149395, "epoch": 1678, "lr": 4.225071891957616e-05}
{"train_loss": 0.08731960505247116, "global_step": 149396, "epoch": 1678, "lr": 4.2250146188610056e-05}
{"train_loss": 0.09538526833057404, "global_step": 149397, "epoch": 1678, "lr": 4.224957345868583e-05}
{"train_loss": 0.07416877895593643, "global_step": 149398, "epoch": 1678, "lr": 4.2249000729803555e-05}
{"train_loss": 0.12304788082838058, "global_step": 149399, "epoch": 1678, "lr": 4.2248428001963324e-05}
{"train_loss": 0.06677290797233582, "global_step": 149400, "epoch": 1678, "lr": 4.2247855275165174e-05}
{"train_loss": 0.10840494930744171, "global_step": 149401, "epoch": 1678, "lr": 4.2247282549409194e-05}
{"train_loss": 0.20371146500110626, "global_step": 149402, "epoch": 1678, "lr": 4.224670982469551e-05}
{"train_loss": 0.06812653690576553, "global_step": 149403, "epoch": 1678, "lr": 4.224613710102413e-05}
{"train_loss": 0.10383117198944092, "global_step": 149404, "epoch": 1678, "lr": 4.224556437839519e-05}
{"train_loss": 0.08085311204195023, "global_step": 149405, "epoch": 1678, "lr": 4.224499165680871e-05}
{"train_loss": 0.10675202310085297, "global_step": 149406, "epoch": 1678, "lr": 4.2244418936264827e-05}
{"train_loss": 0.09960584342479706, "global_step": 149407, "epoch": 1678, "lr": 4.2243846216763554e-05}
{"train_loss": 0.13530561327934265, "global_step": 149408, "epoch": 1678, "lr": 4.224327349830503e-05}
{"train_loss": 0.15871934592723846, "global_step": 149409, "epoch": 1678, "lr": 4.2242700780889284e-05}
{"train_loss": 0.12078090012073517, "global_step": 149410, "epoch": 1678, "lr": 4.224212806451643e-05}
{"train_loss": 0.11757764965295792, "global_step": 149411, "epoch": 1678, "lr": 4.224155534918651e-05}
{"train_loss": 0.1074909046292305, "global_step": 149412, "epoch": 1678, "lr": 4.224098263489963e-05}
{"train_loss": 0.1140228658914566, "global_step": 149413, "epoch": 1678, "lr": 4.224040992165585e-05}
{"train_loss": 0.08125370740890503, "global_step": 149414, "epoch": 1678, "lr": 4.2239837209455266e-05}
{"train_loss": 0.0871022567152977, "global_step": 149415, "epoch": 1678, "lr": 4.2239264498297925e-05}
{"train_loss": 0.10501463711261749, "global_step": 149416, "epoch": 1678, "lr": 4.2238691788183925e-05}
{"train_loss": 0.09187211841344833, "global_step": 149417, "epoch": 1678, "lr": 4.2238119079113335e-05}
{"train_loss": 0.04987499117851257, "global_step": 149418, "epoch": 1678, "lr": 4.223754637108625e-05}
{"train_loss": 0.07405912131071091, "global_step": 149419, "epoch": 1678, "lr": 4.2236973664102706e-05}
{"train_loss": 0.10931334644556046, "global_step": 149420, "epoch": 1678, "lr": 4.223640095816282e-05}
{"train_loss": 0.09903476387262344, "global_step": 149421, "epoch": 1678, "lr": 4.223582825326666e-05}
{"train_loss": 0.11658471822738647, "global_step": 149422, "epoch": 1678, "lr": 4.223525554941428e-05}
{"train_loss": 0.1113608106970787, "global_step": 149423, "epoch": 1678, "lr": 4.223468284660581e-05}
{"train_loss": 0.08932319283485413, "global_step": 149424, "epoch": 1678, "lr": 4.2234110144841255e-05}
{"train_loss": 0.15425769984722137, "global_step": 149425, "epoch": 1678, "lr": 4.2233537444120756e-05}
{"train_loss": 0.09203411638736725, "global_step": 149426, "epoch": 1678, "lr": 4.2232964744444344e-05}
{"train_loss": 0.09555467218160629, "global_step": 149427, "epoch": 1678, "lr": 4.223239204581213e-05}
{"train_loss": 0.14790970087051392, "global_step": 149428, "epoch": 1678, "lr": 4.2231819348224165e-05}
{"train_loss": 0.10825143754482269, "global_step": 149429, "epoch": 1678, "lr": 4.2231246651680555e-05}
{"train_loss": 0.10159536848744649, "global_step": 149430, "epoch": 1678, "lr": 4.223067395618133e-05, "val_loss": 6.176976680755615}
{"train_loss": 0.11051246523857117, "global_step": 149431, "epoch": 1679, "lr": 4.223010126172663e-05}
{"train_loss": 0.11536063998937607, "global_step": 149432, "epoch": 1679, "lr": 4.222952856831648e-05}
{"train_loss": 0.07180586457252502, "global_step": 149433, "epoch": 1679, "lr": 4.222895587595098e-05}
{"train_loss": 0.07637333124876022, "global_step": 149434, "epoch": 1679, "lr": 4.2228383184630195e-05}
{"train_loss": 0.11120510846376419, "global_step": 149435, "epoch": 1679, "lr": 4.2227810494354215e-05}
{"train_loss": 0.08684977889060974, "global_step": 149436, "epoch": 1679, "lr": 4.222723780512311e-05}
{"train_loss": 0.07788560539484024, "global_step": 149437, "epoch": 1679, "lr": 4.222666511693694e-05}
{"train_loss": 0.07583347707986832, "global_step": 149438, "epoch": 1679, "lr": 4.222609242979583e-05}
{"train_loss": 0.13144777715206146, "global_step": 149439, "epoch": 1679, "lr": 4.22255197436998e-05}
{"train_loss": 0.07615847885608673, "global_step": 149440, "epoch": 1679, "lr": 4.2224947058648986e-05}
{"train_loss": 0.11477755755186081, "global_step": 149441, "epoch": 1679, "lr": 4.22243743746434e-05}
{"train_loss": 0.07869388163089752, "global_step": 149442, "epoch": 1679, "lr": 4.2223801691683185e-05}
{"train_loss": 0.1005568653345108, "global_step": 149443, "epoch": 1679, "lr": 4.222322900976835e-05}
{"train_loss": 0.14157333970069885, "global_step": 149444, "epoch": 1679, "lr": 4.222265632889904e-05}
{"train_loss": 0.21030300855636597, "global_step": 149445, "epoch": 1679, "lr": 4.222208364907527e-05}
{"train_loss": 0.11098514497280121, "global_step": 149446, "epoch": 1679, "lr": 4.2221510970297176e-05}
{"train_loss": 0.07655565440654755, "global_step": 149447, "epoch": 1679, "lr": 4.222093829256478e-05}
{"train_loss": 0.1083594262599945, "global_step": 149448, "epoch": 1679, "lr": 4.22203656158782e-05}
{"train_loss": 0.1404971331357956, "global_step": 149449, "epoch": 1679, "lr": 4.221979294023748e-05}
{"train_loss": 0.09222840517759323, "global_step": 149450, "epoch": 1679, "lr": 4.221922026564273e-05}
{"train_loss": 0.11166650056838989, "global_step": 149451, "epoch": 1679, "lr": 4.2218647592094005e-05}
{"train_loss": 0.07463659346103668, "global_step": 149452, "epoch": 1679, "lr": 4.221807491959139e-05}
{"train_loss": 0.1652049422264099, "global_step": 149453, "epoch": 1679, "lr": 4.221750224813496e-05}
{"train_loss": 0.07845504581928253, "global_step": 149454, "epoch": 1679, "lr": 4.221692957772479e-05}
{"train_loss": 0.10028782486915588, "global_step": 149455, "epoch": 1679, "lr": 4.221635690836096e-05}
{"train_loss": 0.06501974910497665, "global_step": 149456, "epoch": 1679, "lr": 4.221578424004353e-05}
{"train_loss": 0.1358548253774643, "global_step": 149457, "epoch": 1679, "lr": 4.221521157277262e-05}
{"train_loss": 0.07368286699056625, "global_step": 149458, "epoch": 1679, "lr": 4.221463890654826e-05}
{"train_loss": 0.0972793847322464, "global_step": 149459, "epoch": 1679, "lr": 4.221406624137056e-05}
{"train_loss": 0.13826167583465576, "global_step": 149460, "epoch": 1679, "lr": 4.221349357723957e-05}
{"train_loss": 0.09646236896514893, "global_step": 149461, "epoch": 1679, "lr": 4.22129209141554e-05}
{"train_loss": 0.051201511174440384, "global_step": 149462, "epoch": 1679, "lr": 4.221234825211809e-05}
{"train_loss": 0.0686529129743576, "global_step": 149463, "epoch": 1679, "lr": 4.221177559112775e-05}
{"train_loss": 0.1193196251988411, "global_step": 149464, "epoch": 1679, "lr": 4.221120293118442e-05}
{"train_loss": 0.06130177900195122, "global_step": 149465, "epoch": 1679, "lr": 4.2210630272288235e-05}
{"train_loss": 0.14471647143363953, "global_step": 149466, "epoch": 1679, "lr": 4.22100576144392e-05}
{"train_loss": 0.07825005054473877, "global_step": 149467, "epoch": 1679, "lr": 4.220948495763745e-05}
{"train_loss": 0.11602490395307541, "global_step": 149468, "epoch": 1679, "lr": 4.220891230188302e-05}
{"train_loss": 0.1135985478758812, "global_step": 149469, "epoch": 1679, "lr": 4.220833964717603e-05}
{"train_loss": 0.07888903468847275, "global_step": 149470, "epoch": 1679, "lr": 4.220776699351652e-05}
{"train_loss": 0.0832037478685379, "global_step": 149471, "epoch": 1679, "lr": 4.220719434090458e-05}
{"train_loss": 0.11857884377241135, "global_step": 149472, "epoch": 1679, "lr": 4.2206621689340296e-05}
{"train_loss": 0.060477208346128464, "global_step": 149473, "epoch": 1679, "lr": 4.220604903882372e-05}
{"train_loss": 0.08760686218738556, "global_step": 149474, "epoch": 1679, "lr": 4.2205476389354966e-05}
{"train_loss": 0.1468881219625473, "global_step": 149475, "epoch": 1679, "lr": 4.220490374093407e-05}
{"train_loss": 0.07881519943475723, "global_step": 149476, "epoch": 1679, "lr": 4.2204331093561155e-05}
{"train_loss": 0.06170649081468582, "global_step": 149477, "epoch": 1679, "lr": 4.2203758447236245e-05}
{"train_loss": 0.07394969463348389, "global_step": 149478, "epoch": 1679, "lr": 4.220318580195947e-05}
{"train_loss": 0.0731041207909584, "global_step": 149479, "epoch": 1679, "lr": 4.220261315773086e-05}
{"train_loss": 0.15108782052993774, "global_step": 149480, "epoch": 1679, "lr": 4.2202040514550536e-05}
{"train_loss": 0.09028518944978714, "global_step": 149481, "epoch": 1679, "lr": 4.220146787241853e-05}
{"train_loss": 0.17602592706680298, "global_step": 149482, "epoch": 1679, "lr": 4.220089523133496e-05}
{"train_loss": 0.08224135637283325, "global_step": 149483, "epoch": 1679, "lr": 4.220032259129987e-05}
{"train_loss": 0.17374546825885773, "global_step": 149484, "epoch": 1679, "lr": 4.2199749952313365e-05}
{"train_loss": 0.11573787778615952, "global_step": 149485, "epoch": 1679, "lr": 4.219917731437549e-05}
{"train_loss": 0.09912250936031342, "global_step": 149486, "epoch": 1679, "lr": 4.2198604677486364e-05}
{"train_loss": 0.031018255278468132, "global_step": 149487, "epoch": 1679, "lr": 4.219803204164602e-05}
{"train_loss": 0.16815264523029327, "global_step": 149488, "epoch": 1679, "lr": 4.2197459406854575e-05}
{"train_loss": 0.09724467247724533, "global_step": 149489, "epoch": 1679, "lr": 4.2196886773112074e-05}
{"train_loss": 0.14504843950271606, "global_step": 149490, "epoch": 1679, "lr": 4.219631414041861e-05}
{"train_loss": 0.06242518126964569, "global_step": 149491, "epoch": 1679, "lr": 4.2195741508774255e-05}
{"train_loss": 0.12859280407428741, "global_step": 149492, "epoch": 1679, "lr": 4.219516887817908e-05}
{"train_loss": 0.14999271929264069, "global_step": 149493, "epoch": 1679, "lr": 4.219459624863318e-05}
{"train_loss": 0.08265405148267746, "global_step": 149494, "epoch": 1679, "lr": 4.219402362013661e-05}
{"train_loss": 0.09864099323749542, "global_step": 149495, "epoch": 1679, "lr": 4.2193450992689477e-05}
{"train_loss": 0.1402556449174881, "global_step": 149496, "epoch": 1679, "lr": 4.2192878366291816e-05}
{"train_loss": 0.12388616800308228, "global_step": 149497, "epoch": 1679, "lr": 4.219230574094375e-05}
{"train_loss": 0.133682519197464, "global_step": 149498, "epoch": 1679, "lr": 4.2191733116645315e-05}
{"train_loss": 0.06784649193286896, "global_step": 149499, "epoch": 1679, "lr": 4.2191160493396625e-05}
{"train_loss": 0.05648268759250641, "global_step": 149500, "epoch": 1679, "lr": 4.219058787119772e-05}
{"train_loss": 0.08617837727069855, "global_step": 149501, "epoch": 1679, "lr": 4.219001525004871e-05}
{"train_loss": 0.09843725711107254, "global_step": 149502, "epoch": 1679, "lr": 4.218944262994964e-05}
{"train_loss": 0.07164281606674194, "global_step": 149503, "epoch": 1679, "lr": 4.2188870010900626e-05}
{"train_loss": 0.12284021824598312, "global_step": 149504, "epoch": 1679, "lr": 4.21882973929017e-05}
{"train_loss": 0.09572270512580872, "global_step": 149505, "epoch": 1679, "lr": 4.2187724775952987e-05}
{"train_loss": 0.11477947980165482, "global_step": 149506, "epoch": 1679, "lr": 4.218715216005451e-05}
{"train_loss": 0.08418504893779755, "global_step": 149507, "epoch": 1679, "lr": 4.218657954520641e-05}
{"train_loss": 0.13498175144195557, "global_step": 149508, "epoch": 1679, "lr": 4.2186006931408707e-05}
{"train_loss": 0.07276572287082672, "global_step": 149509, "epoch": 1679, "lr": 4.218543431866149e-05}
{"train_loss": 0.1165802925825119, "global_step": 149510, "epoch": 1679, "lr": 4.218486170696487e-05}
{"train_loss": 0.1112862229347229, "global_step": 149511, "epoch": 1679, "lr": 4.2184289096318886e-05}
{"train_loss": 0.05364016070961952, "global_step": 149512, "epoch": 1679, "lr": 4.218371648672364e-05}
{"train_loss": 0.06930700689554214, "global_step": 149513, "epoch": 1679, "lr": 4.218314387817918e-05}
{"train_loss": 0.09038151800632477, "global_step": 149514, "epoch": 1679, "lr": 4.218257127068562e-05}
{"train_loss": 0.10730089247226715, "global_step": 149515, "epoch": 1679, "lr": 4.2181998664243005e-05}
{"train_loss": 0.10702205449342728, "global_step": 149516, "epoch": 1679, "lr": 4.2181426058851436e-05}
{"train_loss": 0.056226786226034164, "global_step": 149517, "epoch": 1679, "lr": 4.218085345451097e-05}
{"train_loss": 0.14790956676006317, "global_step": 149518, "epoch": 1679, "lr": 4.21802808512217e-05}
{"train_loss": 0.1025108794859621, "global_step": 149519, "epoch": 1679, "lr": 4.217970824898368e-05, "val_loss": 6.132906913757324}
{"train_loss": 0.06062512844800949, "global_step": 149520, "epoch": 1680, "lr": 4.217913564779703e-05}
{"train_loss": 0.07118036597967148, "global_step": 149521, "epoch": 1680, "lr": 4.217856304766177e-05}
{"train_loss": 0.07632743567228317, "global_step": 149522, "epoch": 1680, "lr": 4.217799044857804e-05}
{"train_loss": 0.11953886598348618, "global_step": 149523, "epoch": 1680, "lr": 4.217741785054585e-05}
{"train_loss": 0.06746610999107361, "global_step": 149524, "epoch": 1680, "lr": 4.217684525356533e-05}
{"train_loss": 0.10925564169883728, "global_step": 149525, "epoch": 1680, "lr": 4.217627265763653e-05}
{"train_loss": 0.12851329147815704, "global_step": 149526, "epoch": 1680, "lr": 4.217570006275955e-05}
{"train_loss": 0.1871875375509262, "global_step": 149527, "epoch": 1680, "lr": 4.217512746893443e-05}
{"train_loss": 0.12699028849601746, "global_step": 149528, "epoch": 1680, "lr": 4.2174554876161274e-05}
{"train_loss": 0.09072103351354599, "global_step": 149529, "epoch": 1680, "lr": 4.217398228444017e-05}
{"train_loss": 0.06589454412460327, "global_step": 149530, "epoch": 1680, "lr": 4.217340969377116e-05}
{"train_loss": 0.06684630364179611, "global_step": 149531, "epoch": 1680, "lr": 4.217283710415436e-05}
{"train_loss": 0.1442241221666336, "global_step": 149532, "epoch": 1680, "lr": 4.2172264515589806e-05}
{"train_loss": 0.10859829187393188, "global_step": 149533, "epoch": 1680, "lr": 4.217169192807761e-05}
{"train_loss": 0.09070978313684464, "global_step": 149534, "epoch": 1680, "lr": 4.217111934161782e-05}
{"train_loss": 0.08733413368463516, "global_step": 149535, "epoch": 1680, "lr": 4.217054675621055e-05}
{"train_loss": 0.046467289328575134, "global_step": 149536, "epoch": 1680, "lr": 4.216997417185582e-05}
{"train_loss": 0.12296422570943832, "global_step": 149537, "epoch": 1680, "lr": 4.216940158855378e-05}
{"train_loss": 0.11151506751775742, "global_step": 149538, "epoch": 1680, "lr": 4.216882900630444e-05}
{"train_loss": 0.12800800800323486, "global_step": 149539, "epoch": 1680, "lr": 4.216825642510793e-05}
{"train_loss": 0.08374372124671936, "global_step": 149540, "epoch": 1680, "lr": 4.2167683844964276e-05}
{"train_loss": 0.08218368887901306, "global_step": 149541, "epoch": 1680, "lr": 4.21671112658736e-05}
{"train_loss": 0.10170718282461166, "global_step": 149542, "epoch": 1680, "lr": 4.2166538687835945e-05}
{"train_loss": 0.11307745426893234, "global_step": 149543, "epoch": 1680, "lr": 4.216596611085142e-05}
{"train_loss": 0.05506560951471329, "global_step": 149544, "epoch": 1680, "lr": 4.216539353492007e-05}
{"train_loss": 0.09923114627599716, "global_step": 149545, "epoch": 1680, "lr": 4.2164820960042004e-05}
{"train_loss": 0.10757619887590408, "global_step": 149546, "epoch": 1680, "lr": 4.2164248386217266e-05}
{"train_loss": 0.14497338235378265, "global_step": 149547, "epoch": 1680, "lr": 4.216367581344596e-05}
{"train_loss": 0.08826079964637756, "global_step": 149548, "epoch": 1680, "lr": 4.216310324172816e-05}
{"train_loss": 0.10791343450546265, "global_step": 149549, "epoch": 1680, "lr": 4.216253067106391e-05}
{"train_loss": 0.05933167785406113, "global_step": 149550, "epoch": 1680, "lr": 4.216195810145334e-05}
{"train_loss": 0.11541425436735153, "global_step": 149551, "epoch": 1680, "lr": 4.216138553289647e-05}
{"train_loss": 0.13840937614440918, "global_step": 149552, "epoch": 1680, "lr": 4.216081296539344e-05}
{"train_loss": 0.08798731863498688, "global_step": 149553, "epoch": 1680, "lr": 4.2160240398944264e-05}
{"train_loss": 0.2035371959209442, "global_step": 149554, "epoch": 1680, "lr": 4.215966783354907e-05}
{"train_loss": 0.0818367600440979, "global_step": 149555, "epoch": 1680, "lr": 4.2159095269207895e-05}
{"train_loss": 0.1346064954996109, "global_step": 149556, "epoch": 1680, "lr": 4.215852270592085e-05}
{"train_loss": 0.09298107028007507, "global_step": 149557, "epoch": 1680, "lr": 4.2157950143687974e-05}
{"train_loss": 0.09778103232383728, "global_step": 149558, "epoch": 1680, "lr": 4.21573775825094e-05}
{"train_loss": 0.08612670004367828, "global_step": 149559, "epoch": 1680, "lr": 4.215680502238513e-05}
{"train_loss": 0.09881491214036942, "global_step": 149560, "epoch": 1680, "lr": 4.2156232463315315e-05}
{"train_loss": 0.05275854095816612, "global_step": 149561, "epoch": 1680, "lr": 4.215565990529998e-05}
{"train_loss": 0.08522073179483414, "global_step": 149562, "epoch": 1680, "lr": 4.215508734833923e-05}
{"train_loss": 0.08371080458164215, "global_step": 149563, "epoch": 1680, "lr": 4.2154514792433125e-05}
{"train_loss": 0.08194439113140106, "global_step": 149564, "epoch": 1680, "lr": 4.2153942237581755e-05}
{"train_loss": 0.21397775411605835, "global_step": 149565, "epoch": 1680, "lr": 4.215336968378519e-05}
{"train_loss": 0.09599260240793228, "global_step": 149566, "epoch": 1680, "lr": 4.215279713104352e-05}
{"train_loss": 0.09407184273004532, "global_step": 149567, "epoch": 1680, "lr": 4.2152224579356806e-05}
{"train_loss": 0.07461616396903992, "global_step": 149568, "epoch": 1680, "lr": 4.215165202872512e-05}
{"train_loss": 0.05365457385778427, "global_step": 149569, "epoch": 1680, "lr": 4.215107947914856e-05}
{"train_loss": 0.12324775010347366, "global_step": 149570, "epoch": 1680, "lr": 4.2150506930627176e-05}
{"train_loss": 0.10061953216791153, "global_step": 149571, "epoch": 1680, "lr": 4.214993438316108e-05}
{"train_loss": 0.07240547239780426, "global_step": 149572, "epoch": 1680, "lr": 4.214936183675031e-05}
{"train_loss": 0.0611313097178936, "global_step": 149573, "epoch": 1680, "lr": 4.214878929139498e-05}
{"train_loss": 0.08071578294038773, "global_step": 149574, "epoch": 1680, "lr": 4.2148216747095136e-05}
{"train_loss": 0.10058582574129105, "global_step": 149575, "epoch": 1680, "lr": 4.214764420385088e-05}
{"train_loss": 0.15240994095802307, "global_step": 149576, "epoch": 1680, "lr": 4.214707166166226e-05}
{"train_loss": 0.09668198972940445, "global_step": 149577, "epoch": 1680, "lr": 4.2146499120529405e-05}
{"train_loss": 0.1244063675403595, "global_step": 149578, "epoch": 1680, "lr": 4.214592658045232e-05}
{"train_loss": 0.067725770175457, "global_step": 149579, "epoch": 1680, "lr": 4.2145354041431145e-05}
{"train_loss": 0.13109658658504486, "global_step": 149580, "epoch": 1680, "lr": 4.214478150346591e-05}
{"train_loss": 0.12325966358184814, "global_step": 149581, "epoch": 1680, "lr": 4.2144208966556734e-05}
{"train_loss": 0.07445083558559418, "global_step": 149582, "epoch": 1680, "lr": 4.2143636430703656e-05}
{"train_loss": 0.09603814035654068, "global_step": 149583, "epoch": 1680, "lr": 4.2143063895906774e-05}
{"train_loss": 0.08926445990800858, "global_step": 149584, "epoch": 1680, "lr": 4.214249136216617e-05}
{"train_loss": 0.11392821371555328, "global_step": 149585, "epoch": 1680, "lr": 4.214191882948191e-05}
{"train_loss": 0.12309414148330688, "global_step": 149586, "epoch": 1680, "lr": 4.2141346297854076e-05}
{"train_loss": 0.09379561245441437, "global_step": 149587, "epoch": 1680, "lr": 4.214077376728273e-05}
{"train_loss": 0.06877443939447403, "global_step": 149588, "epoch": 1680, "lr": 4.214020123776798e-05}
{"train_loss": 0.07814337313175201, "global_step": 149589, "epoch": 1680, "lr": 4.2139628709309865e-05}
{"train_loss": 0.10649128258228302, "global_step": 149590, "epoch": 1680, "lr": 4.2139056181908495e-05}
{"train_loss": 0.04912417009472847, "global_step": 149591, "epoch": 1680, "lr": 4.213848365556392e-05}
{"train_loss": 0.13180850446224213, "global_step": 149592, "epoch": 1680, "lr": 4.213791113027625e-05}
{"train_loss": 0.08282998949289322, "global_step": 149593, "epoch": 1680, "lr": 4.213733860604552e-05}
{"train_loss": 0.10982462763786316, "global_step": 149594, "epoch": 1680, "lr": 4.213676608287184e-05}
{"train_loss": 0.10992025583982468, "global_step": 149595, "epoch": 1680, "lr": 4.2136193560755265e-05}
{"train_loss": 0.08252125978469849, "global_step": 149596, "epoch": 1680, "lr": 4.21356210396959e-05}
{"train_loss": 0.04450030252337456, "global_step": 149597, "epoch": 1680, "lr": 4.213504851969379e-05}
{"train_loss": 0.21716590225696564, "global_step": 149598, "epoch": 1680, "lr": 4.2134476000749044e-05}
{"train_loss": 0.0879635289311409, "global_step": 149599, "epoch": 1680, "lr": 4.2133903482861704e-05}
{"train_loss": 0.10403595119714737, "global_step": 149600, "epoch": 1680, "lr": 4.213333096603187e-05}
{"train_loss": 0.13833948969841003, "global_step": 149601, "epoch": 1680, "lr": 4.213275845025962e-05}
{"train_loss": 0.15273018181324005, "global_step": 149602, "epoch": 1680, "lr": 4.213218593554501e-05}
{"train_loss": 0.09963641315698624, "global_step": 149603, "epoch": 1680, "lr": 4.213161342188815e-05}
{"train_loss": 0.14592581987380981, "global_step": 149604, "epoch": 1680, "lr": 4.2131040909289086e-05}
{"train_loss": 0.08628835529088974, "global_step": 149605, "epoch": 1680, "lr": 4.213046839774793e-05}
{"train_loss": 0.11697825789451599, "global_step": 149606, "epoch": 1680, "lr": 4.2129895887264706e-05}
{"train_loss": 0.11727709323167801, "global_step": 149607, "epoch": 1680, "lr": 4.2129323377839544e-05}
{"train_loss": 0.10199611207072655, "global_step": 149608, "epoch": 1680, "lr": 4.212875086947248e-05, "val_loss": 6.2939300537109375, "train_action_mse_error": 13.256003379821777}
{"train_loss": 0.12506906688213348, "global_step": 149609, "epoch": 1681, "lr": 4.2128178362163635e-05}
{"train_loss": 0.07498321682214737, "global_step": 149610, "epoch": 1681, "lr": 4.212760585591303e-05}
{"train_loss": 0.10853521525859833, "global_step": 149611, "epoch": 1681, "lr": 4.2127033350720804e-05}
{"train_loss": 0.14417938888072968, "global_step": 149612, "epoch": 1681, "lr": 4.212646084658697e-05}
{"train_loss": 0.1077096089720726, "global_step": 149613, "epoch": 1681, "lr": 4.212588834351167e-05}
{"train_loss": 0.06841173022985458, "global_step": 149614, "epoch": 1681, "lr": 4.212531584149492e-05}
{"train_loss": 0.09047051519155502, "global_step": 149615, "epoch": 1681, "lr": 4.212474334053684e-05}
{"train_loss": 0.14463232457637787, "global_step": 149616, "epoch": 1681, "lr": 4.212417084063748e-05}
{"train_loss": 0.16013476252555847, "global_step": 149617, "epoch": 1681, "lr": 4.212359834179695e-05}
{"train_loss": 0.18140824139118195, "global_step": 149618, "epoch": 1681, "lr": 4.212302584401528e-05}
{"train_loss": 0.1534925103187561, "global_step": 149619, "epoch": 1681, "lr": 4.212245334729259e-05}
{"train_loss": 0.08390112221240997, "global_step": 149620, "epoch": 1681, "lr": 4.2121880851628934e-05}
{"train_loss": 0.09842119365930557, "global_step": 149621, "epoch": 1681, "lr": 4.21213083570244e-05}
{"train_loss": 0.18703050911426544, "global_step": 149622, "epoch": 1681, "lr": 4.212073586347906e-05}
{"train_loss": 0.10672549903392792, "global_step": 149623, "epoch": 1681, "lr": 4.2120163370992986e-05}
{"train_loss": 0.0864429920911789, "global_step": 149624, "epoch": 1681, "lr": 4.2119590879566265e-05}
{"train_loss": 0.1461244374513626, "global_step": 149625, "epoch": 1681, "lr": 4.211901838919897e-05}
{"train_loss": 0.11605626344680786, "global_step": 149626, "epoch": 1681, "lr": 4.211844589989119e-05}
{"train_loss": 0.07602645456790924, "global_step": 149627, "epoch": 1681, "lr": 4.2117873411642954e-05}
{"train_loss": 0.0788242295384407, "global_step": 149628, "epoch": 1681, "lr": 4.2117300924454406e-05}
{"train_loss": 0.07741910964250565, "global_step": 149629, "epoch": 1681, "lr": 4.211672843832557e-05}
{"train_loss": 0.1058279424905777, "global_step": 149630, "epoch": 1681, "lr": 4.211615595325656e-05}
{"train_loss": 0.11607503890991211, "global_step": 149631, "epoch": 1681, "lr": 4.2115583469247424e-05}
{"train_loss": 0.09603903442621231, "global_step": 149632, "epoch": 1681, "lr": 4.211501098629826e-05}
{"train_loss": 0.12171485275030136, "global_step": 149633, "epoch": 1681, "lr": 4.2114438504409125e-05}
{"train_loss": 0.05095995217561722, "global_step": 149634, "epoch": 1681, "lr": 4.211386602358013e-05}
{"train_loss": 0.08710326999425888, "global_step": 149635, "epoch": 1681, "lr": 4.21132935438113e-05}
{"train_loss": 0.1403195858001709, "global_step": 149636, "epoch": 1681, "lr": 4.211272106510275e-05}
{"train_loss": 0.06284209340810776, "global_step": 149637, "epoch": 1681, "lr": 4.211214858745456e-05}
{"train_loss": 0.12905548512935638, "global_step": 149638, "epoch": 1681, "lr": 4.211157611086678e-05}
{"train_loss": 0.06982421875, "global_step": 149639, "epoch": 1681, "lr": 4.2111003635339524e-05}
{"train_loss": 0.08479072898626328, "global_step": 149640, "epoch": 1681, "lr": 4.2110431160872835e-05}
{"train_loss": 0.15660879015922546, "global_step": 149641, "epoch": 1681, "lr": 4.210985868746681e-05}
{"train_loss": 0.1109999269247055, "global_step": 149642, "epoch": 1681, "lr": 4.2109286215121515e-05}
{"train_loss": 0.11816417425870895, "global_step": 149643, "epoch": 1681, "lr": 4.210871374383703e-05}
{"train_loss": 0.15963155031204224, "global_step": 149644, "epoch": 1681, "lr": 4.2108141273613424e-05}
{"train_loss": 0.06241432577371597, "global_step": 149645, "epoch": 1681, "lr": 4.21075688044508e-05}
{"train_loss": 0.04548601433634758, "global_step": 149646, "epoch": 1681, "lr": 4.210699633634919e-05}
{"train_loss": 0.04371820017695427, "global_step": 149647, "epoch": 1681, "lr": 4.2106423869308725e-05}
{"train_loss": 0.08059307932853699, "global_step": 149648, "epoch": 1681, "lr": 4.2105851403329434e-05}
{"train_loss": 0.11721053719520569, "global_step": 149649, "epoch": 1681, "lr": 4.2105278938411436e-05}
{"train_loss": 0.11972310394048691, "global_step": 149650, "epoch": 1681, "lr": 4.210470647455477e-05}
{"train_loss": 0.11957567185163498, "global_step": 149651, "epoch": 1681, "lr": 4.2104134011759537e-05}
{"train_loss": 0.14000925421714783, "global_step": 149652, "epoch": 1681, "lr": 4.21035615500258e-05}
{"train_loss": 0.11005307734012604, "global_step": 149653, "epoch": 1681, "lr": 4.210298908935366e-05}
{"train_loss": 0.1727638989686966, "global_step": 149654, "epoch": 1681, "lr": 4.2102416629743154e-05}
{"train_loss": 0.08455341309309006, "global_step": 149655, "epoch": 1681, "lr": 4.210184417119438e-05}
{"train_loss": 0.11312399059534073, "global_step": 149656, "epoch": 1681, "lr": 4.210127171370744e-05}
{"train_loss": 0.15645526349544525, "global_step": 149657, "epoch": 1681, "lr": 4.210069925728237e-05}
{"train_loss": 0.08463973551988602, "global_step": 149658, "epoch": 1681, "lr": 4.210012680191928e-05}
{"train_loss": 0.07253960520029068, "global_step": 149659, "epoch": 1681, "lr": 4.209955434761822e-05}
{"train_loss": 0.0929270014166832, "global_step": 149660, "epoch": 1681, "lr": 4.209898189437928e-05}
{"train_loss": 0.1098962053656578, "global_step": 149661, "epoch": 1681, "lr": 4.2098409442202544e-05}
{"train_loss": 0.11194968223571777, "global_step": 149662, "epoch": 1681, "lr": 4.209783699108807e-05}
{"train_loss": 0.11997789889574051, "global_step": 149663, "epoch": 1681, "lr": 4.2097264541035955e-05}
{"train_loss": 0.07378457486629486, "global_step": 149664, "epoch": 1681, "lr": 4.2096692092046275e-05}
{"train_loss": 0.08979771286249161, "global_step": 149665, "epoch": 1681, "lr": 4.209611964411907e-05}
{"train_loss": 0.12926556169986725, "global_step": 149666, "epoch": 1681, "lr": 4.209554719725448e-05}
{"train_loss": 0.05689071863889694, "global_step": 149667, "epoch": 1681, "lr": 4.209497475145251e-05}
{"train_loss": 0.12219760566949844, "global_step": 149668, "epoch": 1681, "lr": 4.209440230671331e-05}
{"train_loss": 0.07477350533008575, "global_step": 149669, "epoch": 1681, "lr": 4.209382986303689e-05}
{"train_loss": 0.05436095967888832, "global_step": 149670, "epoch": 1681, "lr": 4.209325742042338e-05}
{"train_loss": 0.1001911535859108, "global_step": 149671, "epoch": 1681, "lr": 4.209268497887281e-05}
{"train_loss": 0.11626075953245163, "global_step": 149672, "epoch": 1681, "lr": 4.209211253838531e-05}
{"train_loss": 0.0770396962761879, "global_step": 149673, "epoch": 1681, "lr": 4.209154009896091e-05}
{"train_loss": 0.0762505829334259, "global_step": 149674, "epoch": 1681, "lr": 4.2090967660599705e-05}
{"train_loss": 0.12836752831935883, "global_step": 149675, "epoch": 1681, "lr": 4.209039522330179e-05}
{"train_loss": 0.08736516535282135, "global_step": 149676, "epoch": 1681, "lr": 4.2089822787067214e-05}
{"train_loss": 0.1126592755317688, "global_step": 149677, "epoch": 1681, "lr": 4.208925035189608e-05}
{"train_loss": 0.13138896226882935, "global_step": 149678, "epoch": 1681, "lr": 4.2088677917788435e-05}
{"train_loss": 0.11621467024087906, "global_step": 149679, "epoch": 1681, "lr": 4.2088105484744374e-05}
{"train_loss": 0.1086859181523323, "global_step": 149680, "epoch": 1681, "lr": 4.208753305276397e-05}
{"train_loss": 0.06428634375333786, "global_step": 149681, "epoch": 1681, "lr": 4.2086960621847296e-05}
{"train_loss": 0.08355653285980225, "global_step": 149682, "epoch": 1681, "lr": 4.208638819199444e-05}
{"train_loss": 0.09957696497440338, "global_step": 149683, "epoch": 1681, "lr": 4.208581576320548e-05}
{"train_loss": 0.09736146032810211, "global_step": 149684, "epoch": 1681, "lr": 4.2085243335480476e-05}
{"train_loss": 0.15275852382183075, "global_step": 149685, "epoch": 1681, "lr": 4.208467090881953e-05}
{"train_loss": 0.10684911906719208, "global_step": 149686, "epoch": 1681, "lr": 4.208409848322267e-05}
{"train_loss": 0.07842035591602325, "global_step": 149687, "epoch": 1681, "lr": 4.2083526058690034e-05}
{"train_loss": 0.1306542158126831, "global_step": 149688, "epoch": 1681, "lr": 4.2082953635221654e-05}
{"train_loss": 0.10965210944414139, "global_step": 149689, "epoch": 1681, "lr": 4.208238121281765e-05}
{"train_loss": 0.09830661863088608, "global_step": 149690, "epoch": 1681, "lr": 4.208180879147804e-05}
{"train_loss": 0.1461220383644104, "global_step": 149691, "epoch": 1681, "lr": 4.2081236371202944e-05}
{"train_loss": 0.15121658146381378, "global_step": 149692, "epoch": 1681, "lr": 4.208066395199244e-05}
{"train_loss": 0.10889143496751785, "global_step": 149693, "epoch": 1681, "lr": 4.208009153384658e-05}
{"train_loss": 0.09094321727752686, "global_step": 149694, "epoch": 1681, "lr": 4.207951911676548e-05}
{"train_loss": 0.09624364227056503, "global_step": 149695, "epoch": 1681, "lr": 4.2078946700749165e-05}
{"train_loss": 0.1210247129201889, "global_step": 149696, "epoch": 1681, "lr": 4.207837428579777e-05}
{"train_loss": 0.10639545794450835, "global_step": 149697, "epoch": 1681, "lr": 4.2077801871911316e-05, "val_loss": 6.0869221687316895}
{"train_loss": 0.07704324275255203, "global_step": 149698, "epoch": 1682, "lr": 4.2077229459089916e-05}
{"train_loss": 0.08218061923980713, "global_step": 149699, "epoch": 1682, "lr": 4.207665704733363e-05}
{"train_loss": 0.09367525577545166, "global_step": 149700, "epoch": 1682, "lr": 4.2076084636642545e-05}
{"train_loss": 0.08453802019357681, "global_step": 149701, "epoch": 1682, "lr": 4.2075512227016736e-05}
{"train_loss": 0.08876408636569977, "global_step": 149702, "epoch": 1682, "lr": 4.207493981845628e-05}
{"train_loss": 0.09984814375638962, "global_step": 149703, "epoch": 1682, "lr": 4.207436741096124e-05}
{"train_loss": 0.13461704552173615, "global_step": 149704, "epoch": 1682, "lr": 4.207379500453171e-05}
{"train_loss": 0.08882027119398117, "global_step": 149705, "epoch": 1682, "lr": 4.207322259916776e-05}
{"train_loss": 0.08267085999250412, "global_step": 149706, "epoch": 1682, "lr": 4.207265019486948e-05}
{"train_loss": 0.10225091129541397, "global_step": 149707, "epoch": 1682, "lr": 4.207207779163691e-05}
{"train_loss": 0.10317155718803406, "global_step": 149708, "epoch": 1682, "lr": 4.207150538947018e-05}
{"train_loss": 0.14418049156665802, "global_step": 149709, "epoch": 1682, "lr": 4.207093298836932e-05}
{"train_loss": 0.0808456763625145, "global_step": 149710, "epoch": 1682, "lr": 4.2070360588334415e-05}
{"train_loss": 0.12354189902544022, "global_step": 149711, "epoch": 1682, "lr": 4.206978818936559e-05}
{"train_loss": 0.06000664457678795, "global_step": 149712, "epoch": 1682, "lr": 4.206921579146286e-05}
{"train_loss": 0.12606178224086761, "global_step": 149713, "epoch": 1682, "lr": 4.206864339462634e-05}
{"train_loss": 0.07823501527309418, "global_step": 149714, "epoch": 1682, "lr": 4.206807099885608e-05}
{"train_loss": 0.08625602722167969, "global_step": 149715, "epoch": 1682, "lr": 4.206749860415219e-05}
{"train_loss": 0.08729912340641022, "global_step": 149716, "epoch": 1682, "lr": 4.206692621051472e-05}
{"train_loss": 0.13402843475341797, "global_step": 149717, "epoch": 1682, "lr": 4.2066353817943755e-05}
{"train_loss": 0.1086869016289711, "global_step": 149718, "epoch": 1682, "lr": 4.206578142643937e-05}
{"train_loss": 0.16760054230690002, "global_step": 149719, "epoch": 1682, "lr": 4.206520903600164e-05}
{"train_loss": 0.08329187333583832, "global_step": 149720, "epoch": 1682, "lr": 4.206463664663065e-05}
{"train_loss": 0.08283663541078568, "global_step": 149721, "epoch": 1682, "lr": 4.2064064258326486e-05}
{"train_loss": 0.1414598524570465, "global_step": 149722, "epoch": 1682, "lr": 4.206349187108919e-05}
{"train_loss": 0.08071315288543701, "global_step": 149723, "epoch": 1682, "lr": 4.206291948491889e-05}
{"train_loss": 0.07554846256971359, "global_step": 149724, "epoch": 1682, "lr": 4.206234709981559e-05}
{"train_loss": 0.0840529203414917, "global_step": 149725, "epoch": 1682, "lr": 4.206177471577945e-05}
{"train_loss": 0.0945025384426117, "global_step": 149726, "epoch": 1682, "lr": 4.206120233281048e-05}
{"train_loss": 0.10590341687202454, "global_step": 149727, "epoch": 1682, "lr": 4.2060629950908786e-05}
{"train_loss": 0.07442190498113632, "global_step": 149728, "epoch": 1682, "lr": 4.206005757007447e-05}
{"train_loss": 0.08983869850635529, "global_step": 149729, "epoch": 1682, "lr": 4.205948519030755e-05}
{"train_loss": 0.08499487489461899, "global_step": 149730, "epoch": 1682, "lr": 4.2058912811608163e-05}
{"train_loss": 0.07360157370567322, "global_step": 149731, "epoch": 1682, "lr": 4.205834043397634e-05}
{"train_loss": 0.06072693318128586, "global_step": 149732, "epoch": 1682, "lr": 4.205776805741219e-05}
{"train_loss": 0.0994981974363327, "global_step": 149733, "epoch": 1682, "lr": 4.205719568191576e-05}
{"train_loss": 0.10163571685552597, "global_step": 149734, "epoch": 1682, "lr": 4.205662330748717e-05}
{"train_loss": 0.06946393102407455, "global_step": 149735, "epoch": 1682, "lr": 4.205605093412645e-05}
{"train_loss": 0.14653553068637848, "global_step": 149736, "epoch": 1682, "lr": 4.2055478561833705e-05}
{"train_loss": 0.04789368063211441, "global_step": 149737, "epoch": 1682, "lr": 4.2054906190609e-05}
{"train_loss": 0.10719545185565948, "global_step": 149738, "epoch": 1682, "lr": 4.2054333820452415e-05}
{"train_loss": 0.1239212229847908, "global_step": 149739, "epoch": 1682, "lr": 4.2053761451364024e-05}
{"train_loss": 0.09551158547401428, "global_step": 149740, "epoch": 1682, "lr": 4.205318908334391e-05}
{"train_loss": 0.08492182940244675, "global_step": 149741, "epoch": 1682, "lr": 4.205261671639216e-05}
{"train_loss": 0.1546958088874817, "global_step": 149742, "epoch": 1682, "lr": 4.205204435050884e-05}
{"train_loss": 0.10632655769586563, "global_step": 149743, "epoch": 1682, "lr": 4.2051471985694005e-05}
{"train_loss": 0.1191115602850914, "global_step": 149744, "epoch": 1682, "lr": 4.205089962194777e-05}
{"train_loss": 0.06293880939483643, "global_step": 149745, "epoch": 1682, "lr": 4.205032725927018e-05}
{"train_loss": 0.06899487227201462, "global_step": 149746, "epoch": 1682, "lr": 4.204975489766132e-05}
{"train_loss": 0.08918025344610214, "global_step": 149747, "epoch": 1682, "lr": 4.20491825371213e-05}
{"train_loss": 0.03838096186518669, "global_step": 149748, "epoch": 1682, "lr": 4.204861017765014e-05}
{"train_loss": 0.09936029464006424, "global_step": 149749, "epoch": 1682, "lr": 4.204803781924798e-05}
{"train_loss": 0.04773768410086632, "global_step": 149750, "epoch": 1682, "lr": 4.204746546191484e-05}
{"train_loss": 0.10686834156513214, "global_step": 149751, "epoch": 1682, "lr": 4.204689310565084e-05}
{"train_loss": 0.06605327874422073, "global_step": 149752, "epoch": 1682, "lr": 4.2046320750456025e-05}
{"train_loss": 0.12373953312635422, "global_step": 149753, "epoch": 1682, "lr": 4.20457483963305e-05}
{"train_loss": 0.15888264775276184, "global_step": 149754, "epoch": 1682, "lr": 4.2045176043274306e-05}
{"train_loss": 0.10948095470666885, "global_step": 149755, "epoch": 1682, "lr": 4.2044603691287565e-05}
{"train_loss": 0.12419111281633377, "global_step": 149756, "epoch": 1682, "lr": 4.2044031340370314e-05}
{"train_loss": 0.0827588140964508, "global_step": 149757, "epoch": 1682, "lr": 4.204345899052265e-05}
{"train_loss": 0.11611194163560867, "global_step": 149758, "epoch": 1682, "lr": 4.2042886641744645e-05}
{"train_loss": 0.08615174144506454, "global_step": 149759, "epoch": 1682, "lr": 4.2042314294036375e-05}
{"train_loss": 0.0901229977607727, "global_step": 149760, "epoch": 1682, "lr": 4.204174194739792e-05}
{"train_loss": 0.1518964022397995, "global_step": 149761, "epoch": 1682, "lr": 4.204116960182937e-05}
{"train_loss": 0.08502262830734253, "global_step": 149762, "epoch": 1682, "lr": 4.204059725733077e-05}
{"train_loss": 0.15792393684387207, "global_step": 149763, "epoch": 1682, "lr": 4.2040024913902203e-05}
{"train_loss": 0.10472093522548676, "global_step": 149764, "epoch": 1682, "lr": 4.203945257154379e-05}
{"train_loss": 0.08264533430337906, "global_step": 149765, "epoch": 1682, "lr": 4.2038880230255546e-05}
{"train_loss": 0.1480342000722885, "global_step": 149766, "epoch": 1682, "lr": 4.2038307890037596e-05}
{"train_loss": 0.12248840928077698, "global_step": 149767, "epoch": 1682, "lr": 4.203773555088999e-05}
{"train_loss": 0.07272998988628387, "global_step": 149768, "epoch": 1682, "lr": 4.2037163212812823e-05}
{"train_loss": 0.06183730810880661, "global_step": 149769, "epoch": 1682, "lr": 4.203659087580615e-05}
{"train_loss": 0.15125657618045807, "global_step": 149770, "epoch": 1682, "lr": 4.2036018539870074e-05}
{"train_loss": 0.12813445925712585, "global_step": 149771, "epoch": 1682, "lr": 4.203544620500464e-05}
{"train_loss": 0.10105772316455841, "global_step": 149772, "epoch": 1682, "lr": 4.203487387120997e-05}
{"train_loss": 0.11588326096534729, "global_step": 149773, "epoch": 1682, "lr": 4.203430153848608e-05}
{"train_loss": 0.10862985998392105, "global_step": 149774, "epoch": 1682, "lr": 4.203372920683311e-05}
{"train_loss": 0.06487524509429932, "global_step": 149775, "epoch": 1682, "lr": 4.20331568762511e-05}
{"train_loss": 0.05055106058716774, "global_step": 149776, "epoch": 1682, "lr": 4.2032584546740134e-05}
{"train_loss": 0.11609943956136703, "global_step": 149777, "epoch": 1682, "lr": 4.2032012218300285e-05}
{"train_loss": 0.07435853034257889, "global_step": 149778, "epoch": 1682, "lr": 4.203143989093164e-05}
{"train_loss": 0.1010945737361908, "global_step": 149779, "epoch": 1682, "lr": 4.203086756463427e-05}
{"train_loss": 0.07445164769887924, "global_step": 149780, "epoch": 1682, "lr": 4.2030295239408255e-05}
{"train_loss": 0.10731581598520279, "global_step": 149781, "epoch": 1682, "lr": 4.202972291525366e-05}
{"train_loss": 0.10786394774913788, "global_step": 149782, "epoch": 1682, "lr": 4.202915059217056e-05}
{"train_loss": 0.11793281883001328, "global_step": 149783, "epoch": 1682, "lr": 4.202857827015908e-05}
{"train_loss": 0.09960558265447617, "global_step": 149784, "epoch": 1682, "lr": 4.202800594921923e-05}
{"train_loss": 0.19105404615402222, "global_step": 149785, "epoch": 1682, "lr": 4.2027433629351145e-05}
{"train_loss": 0.1000835666160905, "global_step": 149786, "epoch": 1682, "lr": 4.202686131055484e-05, "val_loss": 6.3217244148254395}
{"train_loss": 0.06886475533246994, "global_step": 149787, "epoch": 1683, "lr": 4.202628899283046e-05}
{"train_loss": 0.15197214484214783, "global_step": 149788, "epoch": 1683, "lr": 4.2025716676178016e-05}
{"train_loss": 0.10026995092630386, "global_step": 149789, "epoch": 1683, "lr": 4.202514436059765e-05}
{"train_loss": 0.10181830078363419, "global_step": 149790, "epoch": 1683, "lr": 4.202457204608938e-05}
{"train_loss": 0.07213806360960007, "global_step": 149791, "epoch": 1683, "lr": 4.202399973265333e-05}
{"train_loss": 0.1558753103017807, "global_step": 149792, "epoch": 1683, "lr": 4.202342742028954e-05}
{"train_loss": 0.10108093917369843, "global_step": 149793, "epoch": 1683, "lr": 4.2022855108998115e-05}
{"train_loss": 0.1579822152853012, "global_step": 149794, "epoch": 1683, "lr": 4.202228279877911e-05}
{"train_loss": 0.0633329227566719, "global_step": 149795, "epoch": 1683, "lr": 4.202171048963262e-05}
{"train_loss": 0.09120233356952667, "global_step": 149796, "epoch": 1683, "lr": 4.202113818155871e-05}
{"train_loss": 0.07242385298013687, "global_step": 149797, "epoch": 1683, "lr": 4.202056587455745e-05}
{"train_loss": 0.06951982527971268, "global_step": 149798, "epoch": 1683, "lr": 4.201999356862895e-05}
{"train_loss": 0.09457718580961227, "global_step": 149799, "epoch": 1683, "lr": 4.201942126377325e-05}
{"train_loss": 0.1662318855524063, "global_step": 149800, "epoch": 1683, "lr": 4.2018848959990445e-05}
{"train_loss": 0.11040227115154266, "global_step": 149801, "epoch": 1683, "lr": 4.201827665728059e-05}
{"train_loss": 0.07081461697816849, "global_step": 149802, "epoch": 1683, "lr": 4.2017704355643805e-05}
{"train_loss": 0.05167434737086296, "global_step": 149803, "epoch": 1683, "lr": 4.201713205508013e-05}
{"train_loss": 0.13279299437999725, "global_step": 149804, "epoch": 1683, "lr": 4.201655975558967e-05}
{"train_loss": 0.12396742403507233, "global_step": 149805, "epoch": 1683, "lr": 4.2015987457172464e-05}
{"train_loss": 0.11045196652412415, "global_step": 149806, "epoch": 1683, "lr": 4.201541515982863e-05}
{"train_loss": 0.08714214712381363, "global_step": 149807, "epoch": 1683, "lr": 4.201484286355821e-05}
{"train_loss": 0.1358948051929474, "global_step": 149808, "epoch": 1683, "lr": 4.2014270568361316e-05}
{"train_loss": 0.08009649813175201, "global_step": 149809, "epoch": 1683, "lr": 4.2013698274237984e-05}
{"train_loss": 0.10707002878189087, "global_step": 149810, "epoch": 1683, "lr": 4.201312598118834e-05}
{"train_loss": 0.15000993013381958, "global_step": 149811, "epoch": 1683, "lr": 4.20125536892124e-05}
{"train_loss": 0.09745034575462341, "global_step": 149812, "epoch": 1683, "lr": 4.201198139831031e-05}
{"train_loss": 0.11288383603096008, "global_step": 149813, "epoch": 1683, "lr": 4.201140910848209e-05}
{"train_loss": 0.0752723217010498, "global_step": 149814, "epoch": 1683, "lr": 4.201083681972785e-05}
{"train_loss": 0.10553184896707535, "global_step": 149815, "epoch": 1683, "lr": 4.201026453204765e-05}
{"train_loss": 0.09501773118972778, "global_step": 149816, "epoch": 1683, "lr": 4.200969224544157e-05}
{"train_loss": 0.14021480083465576, "global_step": 149817, "epoch": 1683, "lr": 4.2009119959909685e-05}
{"train_loss": 0.13807976245880127, "global_step": 149818, "epoch": 1683, "lr": 4.2008547675452076e-05}
{"train_loss": 0.060183048248291016, "global_step": 149819, "epoch": 1683, "lr": 4.200797539206884e-05}
{"train_loss": 0.07152899354696274, "global_step": 149820, "epoch": 1683, "lr": 4.200740310976001e-05}
{"train_loss": 0.1042177751660347, "global_step": 149821, "epoch": 1683, "lr": 4.200683082852571e-05}
{"train_loss": 0.12944868206977844, "global_step": 149822, "epoch": 1683, "lr": 4.2006258548365975e-05}
{"train_loss": 0.08309020847082138, "global_step": 149823, "epoch": 1683, "lr": 4.2005686269280913e-05}
{"train_loss": 0.09442188590765, "global_step": 149824, "epoch": 1683, "lr": 4.200511399127058e-05}
{"train_loss": 0.0877612978219986, "global_step": 149825, "epoch": 1683, "lr": 4.200454171433508e-05}
{"train_loss": 0.05687127634882927, "global_step": 149826, "epoch": 1683, "lr": 4.200396943847445e-05}
{"train_loss": 0.09366193413734436, "global_step": 149827, "epoch": 1683, "lr": 4.20033971636888e-05}
{"train_loss": 0.11615781486034393, "global_step": 149828, "epoch": 1683, "lr": 4.2002824889978185e-05}
{"train_loss": 0.08265362679958344, "global_step": 149829, "epoch": 1683, "lr": 4.200225261734271e-05}
{"train_loss": 0.07020653039216995, "global_step": 149830, "epoch": 1683, "lr": 4.200168034578241e-05}
{"train_loss": 0.09441984444856644, "global_step": 149831, "epoch": 1683, "lr": 4.2001108075297416e-05}
{"train_loss": 0.12060210108757019, "global_step": 149832, "epoch": 1683, "lr": 4.200053580588777e-05}
{"train_loss": 0.190556138753891, "global_step": 149833, "epoch": 1683, "lr": 4.1999963537553546e-05}
{"train_loss": 0.1264580637216568, "global_step": 149834, "epoch": 1683, "lr": 4.1999391270294826e-05}
{"train_loss": 0.1399531215429306, "global_step": 149835, "epoch": 1683, "lr": 4.1998819004111695e-05}
{"train_loss": 0.20236587524414062, "global_step": 149836, "epoch": 1683, "lr": 4.199824673900422e-05}
{"train_loss": 0.21166926622390747, "global_step": 149837, "epoch": 1683, "lr": 4.199767447497248e-05}
{"train_loss": 0.09943720698356628, "global_step": 149838, "epoch": 1683, "lr": 4.199710221201658e-05}
{"train_loss": 0.06463941186666489, "global_step": 149839, "epoch": 1683, "lr": 4.199652995013654e-05}
{"train_loss": 0.138234943151474, "global_step": 149840, "epoch": 1683, "lr": 4.1995957689332494e-05}
{"train_loss": 0.08970515429973602, "global_step": 149841, "epoch": 1683, "lr": 4.199538542960447e-05}
{"train_loss": 0.11466989666223526, "global_step": 149842, "epoch": 1683, "lr": 4.1994813170952596e-05}
{"train_loss": 0.10432890802621841, "global_step": 149843, "epoch": 1683, "lr": 4.19942409133769e-05}
{"train_loss": 0.09006159007549286, "global_step": 149844, "epoch": 1683, "lr": 4.1993668656877496e-05}
{"train_loss": 0.052547018975019455, "global_step": 149845, "epoch": 1683, "lr": 4.199309640145443e-05}
{"train_loss": 0.09207446873188019, "global_step": 149846, "epoch": 1683, "lr": 4.199252414710782e-05}
{"train_loss": 0.11765847355127335, "global_step": 149847, "epoch": 1683, "lr": 4.1991951893837684e-05}
{"train_loss": 0.09763395041227341, "global_step": 149848, "epoch": 1683, "lr": 4.1991379641644156e-05}
{"train_loss": 0.055852010846138, "global_step": 149849, "epoch": 1683, "lr": 4.1990807390527275e-05}
{"train_loss": 0.05869019776582718, "global_step": 149850, "epoch": 1683, "lr": 4.1990235140487156e-05}
{"train_loss": 0.18539488315582275, "global_step": 149851, "epoch": 1683, "lr": 4.198966289152383e-05}
{"train_loss": 0.10249601304531097, "global_step": 149852, "epoch": 1683, "lr": 4.1989090643637406e-05}
{"train_loss": 0.06462712585926056, "global_step": 149853, "epoch": 1683, "lr": 4.1988518396827945e-05}
{"train_loss": 0.11081179231405258, "global_step": 149854, "epoch": 1683, "lr": 4.1987946151095523e-05}
{"train_loss": 0.08865813910961151, "global_step": 149855, "epoch": 1683, "lr": 4.198737390644025e-05}
{"train_loss": 0.12910769879817963, "global_step": 149856, "epoch": 1683, "lr": 4.1986801662862145e-05}
{"train_loss": 0.09176426380872726, "global_step": 149857, "epoch": 1683, "lr": 4.1986229420361344e-05}
{"train_loss": 0.07503525167703629, "global_step": 149858, "epoch": 1683, "lr": 4.1985657178937874e-05}
{"train_loss": 0.11861934512853622, "global_step": 149859, "epoch": 1683, "lr": 4.198508493859186e-05}
{"train_loss": 0.1267663985490799, "global_step": 149860, "epoch": 1683, "lr": 4.198451269932333e-05}
{"train_loss": 0.1091352254152298, "global_step": 149861, "epoch": 1683, "lr": 4.19839404611324e-05}
{"train_loss": 0.11727044731378555, "global_step": 149862, "epoch": 1683, "lr": 4.1983368224019114e-05}
{"train_loss": 0.12329810112714767, "global_step": 149863, "epoch": 1683, "lr": 4.1982795987983594e-05}
{"train_loss": 0.08058039844036102, "global_step": 149864, "epoch": 1683, "lr": 4.1982223753025865e-05}
{"train_loss": 0.10672261565923691, "global_step": 149865, "epoch": 1683, "lr": 4.1981651519146044e-05}
{"train_loss": 0.05427684262394905, "global_step": 149866, "epoch": 1683, "lr": 4.198107928634417e-05}
{"train_loss": 0.11158057302236557, "global_step": 149867, "epoch": 1683, "lr": 4.198050705462037e-05}
{"train_loss": 0.13006040453910828, "global_step": 149868, "epoch": 1683, "lr": 4.197993482397467e-05}
{"train_loss": 0.12770850956439972, "global_step": 149869, "epoch": 1683, "lr": 4.197936259440719e-05}
{"train_loss": 0.12539337575435638, "global_step": 149870, "epoch": 1683, "lr": 4.197879036591797e-05}
{"train_loss": 0.1329614669084549, "global_step": 149871, "epoch": 1683, "lr": 4.197821813850712e-05}
{"train_loss": 0.1543087661266327, "global_step": 149872, "epoch": 1683, "lr": 4.197764591217469e-05}
{"train_loss": 0.08355586230754852, "global_step": 149873, "epoch": 1683, "lr": 4.197707368692076e-05}
{"train_loss": 0.13563162088394165, "global_step": 149874, "epoch": 1683, "lr": 4.197650146274544e-05}
{"train_loss": 0.10622917492403074, "global_step": 149875, "epoch": 1683, "lr": 4.197592923964875e-05, "val_loss": 6.383666515350342}
{"train_loss": 0.09687279909849167, "global_step": 149876, "epoch": 1684, "lr": 4.197535701763083e-05}
{"train_loss": 0.08443228155374527, "global_step": 149877, "epoch": 1684, "lr": 4.19747847966917e-05}
{"train_loss": 0.13671816885471344, "global_step": 149878, "epoch": 1684, "lr": 4.197421257683148e-05}
{"train_loss": 0.09958644956350327, "global_step": 149879, "epoch": 1684, "lr": 4.197364035805021e-05}
{"train_loss": 0.1913980394601822, "global_step": 149880, "epoch": 1684, "lr": 4.197306814034801e-05}
{"train_loss": 0.09705904126167297, "global_step": 149881, "epoch": 1684, "lr": 4.197249592372491e-05}
{"train_loss": 0.16201451420783997, "global_step": 149882, "epoch": 1684, "lr": 4.197192370818103e-05}
{"train_loss": 0.1548326462507248, "global_step": 149883, "epoch": 1684, "lr": 4.197135149371641e-05}
{"train_loss": 0.2018784135580063, "global_step": 149884, "epoch": 1684, "lr": 4.197077928033116e-05}
{"train_loss": 0.0738540068268776, "global_step": 149885, "epoch": 1684, "lr": 4.1970207068025324e-05}
{"train_loss": 0.11279000341892242, "global_step": 149886, "epoch": 1684, "lr": 4.196963485679901e-05}
{"train_loss": 0.11969136446714401, "global_step": 149887, "epoch": 1684, "lr": 4.196906264665227e-05}
{"train_loss": 0.06940432637929916, "global_step": 149888, "epoch": 1684, "lr": 4.1968490437585205e-05}
{"train_loss": 0.0659630075097084, "global_step": 149889, "epoch": 1684, "lr": 4.1967918229597866e-05}
{"train_loss": 0.09619119763374329, "global_step": 149890, "epoch": 1684, "lr": 4.1967346022690345e-05}
{"train_loss": 0.10483254492282867, "global_step": 149891, "epoch": 1684, "lr": 4.1966773816862724e-05}
{"train_loss": 0.09219104796648026, "global_step": 149892, "epoch": 1684, "lr": 4.1966201612115055e-05}
{"train_loss": 0.0750412717461586, "global_step": 149893, "epoch": 1684, "lr": 4.196562940844746e-05}
{"train_loss": 0.09762079268693924, "global_step": 149894, "epoch": 1684, "lr": 4.196505720585996e-05}
{"train_loss": 0.06576353311538696, "global_step": 149895, "epoch": 1684, "lr": 4.1964485004352686e-05}
{"train_loss": 0.07107067108154297, "global_step": 149896, "epoch": 1684, "lr": 4.1963912803925665e-05}
{"train_loss": 0.15043219923973083, "global_step": 149897, "epoch": 1684, "lr": 4.196334060457902e-05}
{"train_loss": 0.10383494198322296, "global_step": 149898, "epoch": 1684, "lr": 4.196276840631278e-05}
{"train_loss": 0.1317778080701828, "global_step": 149899, "epoch": 1684, "lr": 4.196219620912707e-05}
{"train_loss": 0.14145813882350922, "global_step": 149900, "epoch": 1684, "lr": 4.1961624013021926e-05}
{"train_loss": 0.11411414295434952, "global_step": 149901, "epoch": 1684, "lr": 4.1961051817997464e-05}
{"train_loss": 0.1303526759147644, "global_step": 149902, "epoch": 1684, "lr": 4.196047962405373e-05}
{"train_loss": 0.08783126622438431, "global_step": 149903, "epoch": 1684, "lr": 4.195990743119082e-05}
{"train_loss": 0.11437324434518814, "global_step": 149904, "epoch": 1684, "lr": 4.195933523940878e-05}
{"train_loss": 0.1166580468416214, "global_step": 149905, "epoch": 1684, "lr": 4.1958763048707736e-05}
{"train_loss": 0.106717050075531, "global_step": 149906, "epoch": 1684, "lr": 4.195819085908772e-05}
{"train_loss": 0.07160244137048721, "global_step": 149907, "epoch": 1684, "lr": 4.1957618670548844e-05}
{"train_loss": 0.15371093153953552, "global_step": 149908, "epoch": 1684, "lr": 4.1957046483091144e-05}
{"train_loss": 0.06199921295046806, "global_step": 149909, "epoch": 1684, "lr": 4.195647429671474e-05}
{"train_loss": 0.0960894525051117, "global_step": 149910, "epoch": 1684, "lr": 4.195590211141969e-05}
{"train_loss": 0.1955351084470749, "global_step": 149911, "epoch": 1684, "lr": 4.195532992720606e-05}
{"train_loss": 0.11463776230812073, "global_step": 149912, "epoch": 1684, "lr": 4.195475774407396e-05}
{"train_loss": 0.11121761053800583, "global_step": 149913, "epoch": 1684, "lr": 4.1954185562023414e-05}
{"train_loss": 0.11278321593999863, "global_step": 149914, "epoch": 1684, "lr": 4.1953613381054556e-05}
{"train_loss": 0.09046514332294464, "global_step": 149915, "epoch": 1684, "lr": 4.1953041201167406e-05}
{"train_loss": 0.10304421931505203, "global_step": 149916, "epoch": 1684, "lr": 4.195246902236211e-05}
{"train_loss": 0.0841088742017746, "global_step": 149917, "epoch": 1684, "lr": 4.195189684463867e-05}
{"train_loss": 0.14819686114788055, "global_step": 149918, "epoch": 1684, "lr": 4.195132466799723e-05}
{"train_loss": 0.071816086769104, "global_step": 149919, "epoch": 1684, "lr": 4.195075249243781e-05}
{"train_loss": 0.09385555982589722, "global_step": 149920, "epoch": 1684, "lr": 4.1950180317960545e-05}
{"train_loss": 0.13136087357997894, "global_step": 149921, "epoch": 1684, "lr": 4.194960814456544e-05}
{"train_loss": 0.08777458220720291, "global_step": 149922, "epoch": 1684, "lr": 4.194903597225265e-05}
{"train_loss": 0.052671320736408234, "global_step": 149923, "epoch": 1684, "lr": 4.1948463801022184e-05}
{"train_loss": 0.08541716635227203, "global_step": 149924, "epoch": 1684, "lr": 4.194789163087418e-05}
{"train_loss": 0.07559634000062943, "global_step": 149925, "epoch": 1684, "lr": 4.194731946180865e-05}
{"train_loss": 0.10067887604236603, "global_step": 149926, "epoch": 1684, "lr": 4.194674729382573e-05}
{"train_loss": 0.16634418070316315, "global_step": 149927, "epoch": 1684, "lr": 4.1946175126925454e-05}
{"train_loss": 0.06403911113739014, "global_step": 149928, "epoch": 1684, "lr": 4.194560296110792e-05}
{"train_loss": 0.046881839632987976, "global_step": 149929, "epoch": 1684, "lr": 4.194503079637322e-05}
{"train_loss": 0.08630736917257309, "global_step": 149930, "epoch": 1684, "lr": 4.194445863272139e-05}
{"train_loss": 0.07692921906709671, "global_step": 149931, "epoch": 1684, "lr": 4.194388647015255e-05}
{"train_loss": 0.08300658315420151, "global_step": 149932, "epoch": 1684, "lr": 4.1943314308666735e-05}
{"train_loss": 0.0940563902258873, "global_step": 149933, "epoch": 1684, "lr": 4.194274214826406e-05}
{"train_loss": 0.10085778683423996, "global_step": 149934, "epoch": 1684, "lr": 4.194216998894457e-05}
{"train_loss": 0.11667559295892715, "global_step": 149935, "epoch": 1684, "lr": 4.1941597830708376e-05}
{"train_loss": 0.10155685245990753, "global_step": 149936, "epoch": 1684, "lr": 4.194102567355551e-05}
{"train_loss": 0.10427584499120712, "global_step": 149937, "epoch": 1684, "lr": 4.19404535174861e-05}
{"train_loss": 0.06149551644921303, "global_step": 149938, "epoch": 1684, "lr": 4.193988136250018e-05}
{"train_loss": 0.09420273452997208, "global_step": 149939, "epoch": 1684, "lr": 4.193930920859786e-05}
{"train_loss": 0.07262330502271652, "global_step": 149940, "epoch": 1684, "lr": 4.1938737055779186e-05}
{"train_loss": 0.0870383009314537, "global_step": 149941, "epoch": 1684, "lr": 4.193816490404427e-05}
{"train_loss": 0.11651777476072311, "global_step": 149942, "epoch": 1684, "lr": 4.1937592753393144e-05}
{"train_loss": 0.1435999721288681, "global_step": 149943, "epoch": 1684, "lr": 4.193702060382594e-05}
{"train_loss": 0.1307940036058426, "global_step": 149944, "epoch": 1684, "lr": 4.193644845534268e-05}
{"train_loss": 0.13547813892364502, "global_step": 149945, "epoch": 1684, "lr": 4.193587630794347e-05}
{"train_loss": 0.10891907662153244, "global_step": 149946, "epoch": 1684, "lr": 4.193530416162838e-05}
{"train_loss": 0.11086839437484741, "global_step": 149947, "epoch": 1684, "lr": 4.193473201639751e-05}
{"train_loss": 0.1660962700843811, "global_step": 149948, "epoch": 1684, "lr": 4.193415987225092e-05}
{"train_loss": 0.20035766065120697, "global_step": 149949, "epoch": 1684, "lr": 4.1933587729188654e-05}
{"train_loss": 0.17075605690479279, "global_step": 149950, "epoch": 1684, "lr": 4.193301558721084e-05}
{"train_loss": 0.07799737900495529, "global_step": 149951, "epoch": 1684, "lr": 4.1932443446317526e-05}
{"train_loss": 0.11460113525390625, "global_step": 149952, "epoch": 1684, "lr": 4.193187130650881e-05}
{"train_loss": 0.1234879344701767, "global_step": 149953, "epoch": 1684, "lr": 4.193129916778473e-05}
{"train_loss": 0.1365448236465454, "global_step": 149954, "epoch": 1684, "lr": 4.193072703014542e-05}
{"train_loss": 0.09617042541503906, "global_step": 149955, "epoch": 1684, "lr": 4.193015489359089e-05}
{"train_loss": 0.06572208553552628, "global_step": 149956, "epoch": 1684, "lr": 4.1929582758121284e-05}
{"train_loss": 0.055149760097265244, "global_step": 149957, "epoch": 1684, "lr": 4.192901062373662e-05}
{"train_loss": 0.13358435034751892, "global_step": 149958, "epoch": 1684, "lr": 4.192843849043703e-05}
{"train_loss": 0.08545078337192535, "global_step": 149959, "epoch": 1684, "lr": 4.192786635822254e-05}
{"train_loss": 0.1209694892168045, "global_step": 149960, "epoch": 1684, "lr": 4.192729422709327e-05}
{"train_loss": 0.07654478400945663, "global_step": 149961, "epoch": 1684, "lr": 4.1926722097049254e-05}
{"train_loss": 0.10855449736118317, "global_step": 149962, "epoch": 1684, "lr": 4.192614996809062e-05}
{"train_loss": 0.09246274828910828, "global_step": 149963, "epoch": 1684, "lr": 4.192557784021738e-05}
{"train_loss": 0.10674382716919599, "global_step": 149964, "epoch": 1684, "lr": 4.192500571342967e-05, "val_loss": 6.383766174316406}
{"train_loss": 0.11916625499725342, "global_step": 149965, "epoch": 1685, "lr": 4.1924433587727544e-05}
{"train_loss": 0.06703560799360275, "global_step": 149966, "epoch": 1685, "lr": 4.192386146311108e-05}
{"train_loss": 0.08732061833143234, "global_step": 149967, "epoch": 1685, "lr": 4.192328933958035e-05}
{"train_loss": 0.08496497571468353, "global_step": 149968, "epoch": 1685, "lr": 4.192271721713544e-05}
{"train_loss": 0.11544007807970047, "global_step": 149969, "epoch": 1685, "lr": 4.1922145095776434e-05}
{"train_loss": 0.09664743393659592, "global_step": 149970, "epoch": 1685, "lr": 4.192157297550337e-05}
{"train_loss": 0.0737711638212204, "global_step": 149971, "epoch": 1685, "lr": 4.1921000856316375e-05}
{"train_loss": 0.1330074965953827, "global_step": 149972, "epoch": 1685, "lr": 4.192042873821548e-05}
{"train_loss": 0.06710819154977798, "global_step": 149973, "epoch": 1685, "lr": 4.1919856621200815e-05}
{"train_loss": 0.14842967689037323, "global_step": 149974, "epoch": 1685, "lr": 4.19192845052724e-05}
{"train_loss": 0.11196492612361908, "global_step": 149975, "epoch": 1685, "lr": 4.191871239043036e-05}
{"train_loss": 0.07471060752868652, "global_step": 149976, "epoch": 1685, "lr": 4.191814027667473e-05}
{"train_loss": 0.08821576833724976, "global_step": 149977, "epoch": 1685, "lr": 4.191756816400563e-05}
{"train_loss": 0.09964008629322052, "global_step": 149978, "epoch": 1685, "lr": 4.191699605242309e-05}
{"train_loss": 0.11108804494142532, "global_step": 149979, "epoch": 1685, "lr": 4.191642394192723e-05}
{"train_loss": 0.12008792161941528, "global_step": 149980, "epoch": 1685, "lr": 4.191585183251809e-05}
{"train_loss": 0.10440139472484589, "global_step": 149981, "epoch": 1685, "lr": 4.1915279724195766e-05}
{"train_loss": 0.1645359843969345, "global_step": 149982, "epoch": 1685, "lr": 4.191470761696035e-05}
{"train_loss": 0.10307764261960983, "global_step": 149983, "epoch": 1685, "lr": 4.19141355108119e-05}
{"train_loss": 0.10734041035175323, "global_step": 149984, "epoch": 1685, "lr": 4.1913563405750496e-05}
{"train_loss": 0.14710964262485504, "global_step": 149985, "epoch": 1685, "lr": 4.191299130177621e-05}
{"train_loss": 0.1009698137640953, "global_step": 149986, "epoch": 1685, "lr": 4.1912419198889124e-05}
{"train_loss": 0.10336066037416458, "global_step": 149987, "epoch": 1685, "lr": 4.191184709708932e-05}
{"train_loss": 0.11734585464000702, "global_step": 149988, "epoch": 1685, "lr": 4.191127499637688e-05}
{"train_loss": 0.2067834734916687, "global_step": 149989, "epoch": 1685, "lr": 4.191070289675184e-05}
{"train_loss": 0.17660973966121674, "global_step": 149990, "epoch": 1685, "lr": 4.191013079821434e-05}
{"train_loss": 0.13358540832996368, "global_step": 149991, "epoch": 1685, "lr": 4.190955870076439e-05}
{"train_loss": 0.09026177227497101, "global_step": 149992, "epoch": 1685, "lr": 4.190898660440213e-05}
{"train_loss": 0.1009235680103302, "global_step": 149993, "epoch": 1685, "lr": 4.190841450912759e-05}
{"train_loss": 0.11940246820449829, "global_step": 149994, "epoch": 1685, "lr": 4.190784241494088e-05}
{"train_loss": 0.1270008087158203, "global_step": 149995, "epoch": 1685, "lr": 4.190727032184205e-05}
{"train_loss": 0.1273161768913269, "global_step": 149996, "epoch": 1685, "lr": 4.190669822983121e-05}
{"train_loss": 0.08923006057739258, "global_step": 149997, "epoch": 1685, "lr": 4.190612613890838e-05}
{"train_loss": 0.09188716858625412, "global_step": 149998, "epoch": 1685, "lr": 4.19055540490737e-05}
{"train_loss": 0.1174866110086441, "global_step": 149999, "epoch": 1685, "lr": 4.19049819603272e-05}
{"train_loss": 0.07959060370922089, "global_step": 150000, "epoch": 1685, "lr": 4.190440987266898e-05}
{"train_loss": 0.11623922735452652, "global_step": 150001, "epoch": 1685, "lr": 4.190383778609913e-05}
{"train_loss": 0.059307992458343506, "global_step": 150002, "epoch": 1685, "lr": 4.1903265700617705e-05}
{"train_loss": 0.07750043272972107, "global_step": 150003, "epoch": 1685, "lr": 4.1902693616224786e-05}
{"train_loss": 0.06943102180957794, "global_step": 150004, "epoch": 1685, "lr": 4.1902121532920446e-05}
{"train_loss": 0.19813279807567596, "global_step": 150005, "epoch": 1685, "lr": 4.1901549450704774e-05}
{"train_loss": 0.05608309432864189, "global_step": 150006, "epoch": 1685, "lr": 4.190097736957784e-05}
{"train_loss": 0.16762635111808777, "global_step": 150007, "epoch": 1685, "lr": 4.1900405289539724e-05}
{"train_loss": 0.14825241267681122, "global_step": 150008, "epoch": 1685, "lr": 4.189983321059048e-05}
{"train_loss": 0.06813005357980728, "global_step": 150009, "epoch": 1685, "lr": 4.1899261132730226e-05}
{"train_loss": 0.0679829940199852, "global_step": 150010, "epoch": 1685, "lr": 4.1898689055959004e-05}
{"train_loss": 0.07302196323871613, "global_step": 150011, "epoch": 1685, "lr": 4.189811698027692e-05}
{"train_loss": 0.10303138941526413, "global_step": 150012, "epoch": 1685, "lr": 4.189754490568402e-05}
{"train_loss": 0.10316496342420578, "global_step": 150013, "epoch": 1685, "lr": 4.1896972832180406e-05}
{"train_loss": 0.14386898279190063, "global_step": 150014, "epoch": 1685, "lr": 4.189640075976613e-05}
{"train_loss": 0.04717063903808594, "global_step": 150015, "epoch": 1685, "lr": 4.1895828688441304e-05}
{"train_loss": 0.08961517363786697, "global_step": 150016, "epoch": 1685, "lr": 4.189525661820597e-05}
{"train_loss": 0.12550625205039978, "global_step": 150017, "epoch": 1685, "lr": 4.1894684549060214e-05}
{"train_loss": 0.07279081642627716, "global_step": 150018, "epoch": 1685, "lr": 4.189411248100414e-05}
{"train_loss": 0.09046022593975067, "global_step": 150019, "epoch": 1685, "lr": 4.1893540414037786e-05}
{"train_loss": 0.09198547899723053, "global_step": 150020, "epoch": 1685, "lr": 4.1892968348161264e-05}
{"train_loss": 0.06366757303476334, "global_step": 150021, "epoch": 1685, "lr": 4.189239628337462e-05}
{"train_loss": 0.12097548693418503, "global_step": 150022, "epoch": 1685, "lr": 4.189182421967796e-05}
{"train_loss": 0.13923831284046173, "global_step": 150023, "epoch": 1685, "lr": 4.189125215707133e-05}
{"train_loss": 0.10587211698293686, "global_step": 150024, "epoch": 1685, "lr": 4.1890680095554824e-05}
{"train_loss": 0.12300629913806915, "global_step": 150025, "epoch": 1685, "lr": 4.1890108035128526e-05}
{"train_loss": 0.08520637452602386, "global_step": 150026, "epoch": 1685, "lr": 4.188953597579251e-05}
{"train_loss": 0.17480380833148956, "global_step": 150027, "epoch": 1685, "lr": 4.188896391754682e-05}
{"train_loss": 0.12940898537635803, "global_step": 150028, "epoch": 1685, "lr": 4.1888391860391584e-05}
{"train_loss": 0.1065855622291565, "global_step": 150029, "epoch": 1685, "lr": 4.188781980432684e-05}
{"train_loss": 0.0779375433921814, "global_step": 150030, "epoch": 1685, "lr": 4.18872477493527e-05}
{"train_loss": 0.10353598743677139, "global_step": 150031, "epoch": 1685, "lr": 4.188667569546919e-05}
{"train_loss": 0.06375708431005478, "global_step": 150032, "epoch": 1685, "lr": 4.1886103642676444e-05}
{"train_loss": 0.11453872174024582, "global_step": 150033, "epoch": 1685, "lr": 4.188553159097449e-05}
{"train_loss": 0.06201466545462608, "global_step": 150034, "epoch": 1685, "lr": 4.188495954036345e-05}
{"train_loss": 0.08825881034135818, "global_step": 150035, "epoch": 1685, "lr": 4.188438749084336e-05}
{"train_loss": 0.10556627064943314, "global_step": 150036, "epoch": 1685, "lr": 4.1883815442414315e-05}
{"train_loss": 0.05621720477938652, "global_step": 150037, "epoch": 1685, "lr": 4.188324339507641e-05}
{"train_loss": 0.1575494259595871, "global_step": 150038, "epoch": 1685, "lr": 4.188267134882969e-05}
{"train_loss": 0.11169502884149551, "global_step": 150039, "epoch": 1685, "lr": 4.188209930367426e-05}
{"train_loss": 0.06357737630605698, "global_step": 150040, "epoch": 1685, "lr": 4.188152725961017e-05}
{"train_loss": 0.10578840970993042, "global_step": 150041, "epoch": 1685, "lr": 4.188095521663751e-05}
{"train_loss": 0.09279927611351013, "global_step": 150042, "epoch": 1685, "lr": 4.188038317475637e-05}
{"train_loss": 0.12816767394542694, "global_step": 150043, "epoch": 1685, "lr": 4.1879811133966794e-05}
{"train_loss": 0.08489984273910522, "global_step": 150044, "epoch": 1685, "lr": 4.187923909426889e-05}
{"train_loss": 0.10343664884567261, "global_step": 150045, "epoch": 1685, "lr": 4.1878667055662735e-05}
{"train_loss": 0.07435808330774307, "global_step": 150046, "epoch": 1685, "lr": 4.1878095018148366e-05}
{"train_loss": 0.07199005782604218, "global_step": 150047, "epoch": 1685, "lr": 4.187752298172591e-05}
{"train_loss": 0.11436104029417038, "global_step": 150048, "epoch": 1685, "lr": 4.187695094639541e-05}
{"train_loss": 0.08169795572757721, "global_step": 150049, "epoch": 1685, "lr": 4.1876378912156965e-05}
{"train_loss": 0.11696697771549225, "global_step": 150050, "epoch": 1685, "lr": 4.187580687901063e-05}
{"train_loss": 0.11723325401544571, "global_step": 150051, "epoch": 1685, "lr": 4.187523484695651e-05}
{"train_loss": 0.07696221023797989, "global_step": 150052, "epoch": 1685, "lr": 4.187466281599465e-05}
{"train_loss": 0.10413635057512294, "global_step": 150053, "epoch": 1685, "lr": 4.187409078612515e-05, "val_loss": 6.293452262878418, "train_action_mse_error": 20.325824737548828}
{"train_loss": 0.0726243257522583, "global_step": 150054, "epoch": 1686, "lr": 4.187351875734807e-05}
{"train_loss": 0.09406228363513947, "global_step": 150055, "epoch": 1686, "lr": 4.187294672966349e-05}
{"train_loss": 0.09426715224981308, "global_step": 150056, "epoch": 1686, "lr": 4.1872374703071524e-05}
{"train_loss": 0.08217913657426834, "global_step": 150057, "epoch": 1686, "lr": 4.187180267757219e-05}
{"train_loss": 0.0946316048502922, "global_step": 150058, "epoch": 1686, "lr": 4.1871230653165614e-05}
{"train_loss": 0.08073887228965759, "global_step": 150059, "epoch": 1686, "lr": 4.187065862985185e-05}
{"train_loss": 0.10487605631351471, "global_step": 150060, "epoch": 1686, "lr": 4.187008660763096e-05}
{"train_loss": 0.09943937510251999, "global_step": 150061, "epoch": 1686, "lr": 4.186951458650304e-05}
{"train_loss": 0.08896899968385696, "global_step": 150062, "epoch": 1686, "lr": 4.186894256646818e-05}
{"train_loss": 0.07155543565750122, "global_step": 150063, "epoch": 1686, "lr": 4.186837054752643e-05}
{"train_loss": 0.09264758974313736, "global_step": 150064, "epoch": 1686, "lr": 4.186779852967789e-05}
{"train_loss": 0.07682493329048157, "global_step": 150065, "epoch": 1686, "lr": 4.1867226512922606e-05}
{"train_loss": 0.08823076635599136, "global_step": 150066, "epoch": 1686, "lr": 4.1866654497260705e-05}
{"train_loss": 0.061067741364240646, "global_step": 150067, "epoch": 1686, "lr": 4.18660824826922e-05}
{"train_loss": 0.1076880395412445, "global_step": 150068, "epoch": 1686, "lr": 4.186551046921722e-05}
{"train_loss": 0.0861416682600975, "global_step": 150069, "epoch": 1686, "lr": 4.186493845683581e-05}
{"train_loss": 0.12335202842950821, "global_step": 150070, "epoch": 1686, "lr": 4.1864366445548084e-05}
{"train_loss": 0.12340576201677322, "global_step": 150071, "epoch": 1686, "lr": 4.1863794435354065e-05}
{"train_loss": 0.08651110529899597, "global_step": 150072, "epoch": 1686, "lr": 4.1863222426253857e-05}
{"train_loss": 0.10557208955287933, "global_step": 150073, "epoch": 1686, "lr": 4.186265041824757e-05}
{"train_loss": 0.07079144567251205, "global_step": 150074, "epoch": 1686, "lr": 4.1862078411335225e-05}
{"train_loss": 0.07913589477539062, "global_step": 150075, "epoch": 1686, "lr": 4.186150640551695e-05}
{"train_loss": 0.12094184011220932, "global_step": 150076, "epoch": 1686, "lr": 4.1860934400792766e-05}
{"train_loss": 0.06815320253372192, "global_step": 150077, "epoch": 1686, "lr": 4.1860362397162805e-05}
{"train_loss": 0.09207066893577576, "global_step": 150078, "epoch": 1686, "lr": 4.1859790394627115e-05}
{"train_loss": 0.057714786380529404, "global_step": 150079, "epoch": 1686, "lr": 4.1859218393185765e-05}
{"train_loss": 0.08239783346652985, "global_step": 150080, "epoch": 1686, "lr": 4.185864639283885e-05}
{"train_loss": 0.0641910657286644, "global_step": 150081, "epoch": 1686, "lr": 4.185807439358644e-05}
{"train_loss": 0.10345561057329178, "global_step": 150082, "epoch": 1686, "lr": 4.185750239542861e-05}
{"train_loss": 0.15391358733177185, "global_step": 150083, "epoch": 1686, "lr": 4.1856930398365454e-05}
{"train_loss": 0.1142725870013237, "global_step": 150084, "epoch": 1686, "lr": 4.185635840239702e-05}
{"train_loss": 0.09391086548566818, "global_step": 150085, "epoch": 1686, "lr": 4.1855786407523416e-05}
{"train_loss": 0.11845539510250092, "global_step": 150086, "epoch": 1686, "lr": 4.185521441374468e-05}
{"train_loss": 0.10216999799013138, "global_step": 150087, "epoch": 1686, "lr": 4.185464242106093e-05}
{"train_loss": 0.1327066272497177, "global_step": 150088, "epoch": 1686, "lr": 4.18540704294722e-05}
{"train_loss": 0.07529055327177048, "global_step": 150089, "epoch": 1686, "lr": 4.185349843897862e-05}
{"train_loss": 0.10641143471002579, "global_step": 150090, "epoch": 1686, "lr": 4.185292644958022e-05}
{"train_loss": 0.10231894999742508, "global_step": 150091, "epoch": 1686, "lr": 4.185235446127708e-05}
{"train_loss": 0.11896450817584991, "global_step": 150092, "epoch": 1686, "lr": 4.1851782474069327e-05}
{"train_loss": 0.14425289630889893, "global_step": 150093, "epoch": 1686, "lr": 4.185121048795697e-05}
{"train_loss": 0.14880387485027313, "global_step": 150094, "epoch": 1686, "lr": 4.1850638502940144e-05}
{"train_loss": 0.13796751201152802, "global_step": 150095, "epoch": 1686, "lr": 4.1850066519018884e-05}
{"train_loss": 0.08884657174348831, "global_step": 150096, "epoch": 1686, "lr": 4.18494945361933e-05}
{"train_loss": 0.10256808251142502, "global_step": 150097, "epoch": 1686, "lr": 4.184892255446343e-05}
{"train_loss": 0.10221236944198608, "global_step": 150098, "epoch": 1686, "lr": 4.184835057382939e-05}
{"train_loss": 0.13703882694244385, "global_step": 150099, "epoch": 1686, "lr": 4.1847778594291234e-05}
{"train_loss": 0.06782475858926773, "global_step": 150100, "epoch": 1686, "lr": 4.184720661584905e-05}
{"train_loss": 0.06066732853651047, "global_step": 150101, "epoch": 1686, "lr": 4.184663463850291e-05}
{"train_loss": 0.0831158310174942, "global_step": 150102, "epoch": 1686, "lr": 4.184606266225288e-05}
{"train_loss": 0.05358028784394264, "global_step": 150103, "epoch": 1686, "lr": 4.1845490687099056e-05}
{"train_loss": 0.07582058012485504, "global_step": 150104, "epoch": 1686, "lr": 4.1844918713041516e-05}
{"train_loss": 0.09515254944562912, "global_step": 150105, "epoch": 1686, "lr": 4.18443467400803e-05}
{"train_loss": 0.1228620782494545, "global_step": 150106, "epoch": 1686, "lr": 4.184377476821554e-05}
{"train_loss": 0.12942656874656677, "global_step": 150107, "epoch": 1686, "lr": 4.1843202797447264e-05}
{"train_loss": 0.19225549697875977, "global_step": 150108, "epoch": 1686, "lr": 4.1842630827775565e-05}
{"train_loss": 0.12287489324808121, "global_step": 150109, "epoch": 1686, "lr": 4.184205885920055e-05}
{"train_loss": 0.11547479778528214, "global_step": 150110, "epoch": 1686, "lr": 4.184148689172225e-05}
{"train_loss": 0.15403595566749573, "global_step": 150111, "epoch": 1686, "lr": 4.184091492534078e-05}
{"train_loss": 0.1057056337594986, "global_step": 150112, "epoch": 1686, "lr": 4.184034296005618e-05}
{"train_loss": 0.05909125134348869, "global_step": 150113, "epoch": 1686, "lr": 4.1839770995868566e-05}
{"train_loss": 0.07987242937088013, "global_step": 150114, "epoch": 1686, "lr": 4.1839199032777974e-05}
{"train_loss": 0.08423770219087601, "global_step": 150115, "epoch": 1686, "lr": 4.183862707078453e-05}
{"train_loss": 0.0640876442193985, "global_step": 150116, "epoch": 1686, "lr": 4.183805510988825e-05}
{"train_loss": 0.15725727379322052, "global_step": 150117, "epoch": 1686, "lr": 4.183748315008927e-05}
{"train_loss": 0.13605819642543793, "global_step": 150118, "epoch": 1686, "lr": 4.183691119138764e-05}
{"train_loss": 0.10594959557056427, "global_step": 150119, "epoch": 1686, "lr": 4.183633923378343e-05}
{"train_loss": 0.10340356081724167, "global_step": 150120, "epoch": 1686, "lr": 4.183576727727673e-05}
{"train_loss": 0.1935899704694748, "global_step": 150121, "epoch": 1686, "lr": 4.18351953218676e-05}
{"train_loss": 0.10604355484247208, "global_step": 150122, "epoch": 1686, "lr": 4.1834623367556135e-05}
{"train_loss": 0.06537118554115295, "global_step": 150123, "epoch": 1686, "lr": 4.1834051414342415e-05}
{"train_loss": 0.10744894295930862, "global_step": 150124, "epoch": 1686, "lr": 4.183347946222649e-05}
{"train_loss": 0.1093759760260582, "global_step": 150125, "epoch": 1686, "lr": 4.1832907511208476e-05}
{"train_loss": 0.1072409451007843, "global_step": 150126, "epoch": 1686, "lr": 4.183233556128841e-05}
{"train_loss": 0.07408902794122696, "global_step": 150127, "epoch": 1686, "lr": 4.183176361246638e-05}
{"train_loss": 0.1351086050271988, "global_step": 150128, "epoch": 1686, "lr": 4.183119166474249e-05}
{"train_loss": 0.08933910727500916, "global_step": 150129, "epoch": 1686, "lr": 4.183061971811677e-05}
{"train_loss": 0.12466698884963989, "global_step": 150130, "epoch": 1686, "lr": 4.1830047772589356e-05}
{"train_loss": 0.1080998107790947, "global_step": 150131, "epoch": 1686, "lr": 4.182947582816026e-05}
{"train_loss": 0.07713115215301514, "global_step": 150132, "epoch": 1686, "lr": 4.182890388482962e-05}
{"train_loss": 0.08737494051456451, "global_step": 150133, "epoch": 1686, "lr": 4.1828331942597466e-05}
{"train_loss": 0.08135832101106644, "global_step": 150134, "epoch": 1686, "lr": 4.1827760001463914e-05}
{"train_loss": 0.11907051503658295, "global_step": 150135, "epoch": 1686, "lr": 4.1827188061428997e-05}
{"train_loss": 0.09802456200122833, "global_step": 150136, "epoch": 1686, "lr": 4.182661612249284e-05}
{"train_loss": 0.08950947970151901, "global_step": 150137, "epoch": 1686, "lr": 4.182604418465547e-05}
{"train_loss": 0.1406966745853424, "global_step": 150138, "epoch": 1686, "lr": 4.1825472247917e-05}
{"train_loss": 0.10135875642299652, "global_step": 150139, "epoch": 1686, "lr": 4.182490031227749e-05}
{"train_loss": 0.13742318749427795, "global_step": 150140, "epoch": 1686, "lr": 4.1824328377737035e-05}
{"train_loss": 0.1290130764245987, "global_step": 150141, "epoch": 1686, "lr": 4.1823756444295687e-05}
{"train_loss": 0.10221975296735764, "global_step": 150142, "epoch": 1686, "lr": 4.182318451195355e-05, "val_loss": 6.350804805755615}
{"train_loss": 0.11190121620893478, "global_step": 150143, "epoch": 1687, "lr": 4.1822612580710665e-05}
{"train_loss": 0.09537467360496521, "global_step": 150144, "epoch": 1687, "lr": 4.182204065056713e-05}
{"train_loss": 0.09379784762859344, "global_step": 150145, "epoch": 1687, "lr": 4.1821468721523046e-05}
{"train_loss": 0.17454195022583008, "global_step": 150146, "epoch": 1687, "lr": 4.182089679357843e-05}
{"train_loss": 0.11707393825054169, "global_step": 150147, "epoch": 1687, "lr": 4.182032486673343e-05}
{"train_loss": 0.07692559063434601, "global_step": 150148, "epoch": 1687, "lr": 4.181975294098807e-05}
{"train_loss": 0.09901370108127594, "global_step": 150149, "epoch": 1687, "lr": 4.181918101634246e-05}
{"train_loss": 0.15082509815692902, "global_step": 150150, "epoch": 1687, "lr": 4.181860909279664e-05}
{"train_loss": 0.09005219489336014, "global_step": 150151, "epoch": 1687, "lr": 4.1818037170350726e-05}
{"train_loss": 0.0989169254899025, "global_step": 150152, "epoch": 1687, "lr": 4.1817465249004764e-05}
{"train_loss": 0.10303047299385071, "global_step": 150153, "epoch": 1687, "lr": 4.181689332875886e-05}
{"train_loss": 0.09449344873428345, "global_step": 150154, "epoch": 1687, "lr": 4.181632140961305e-05}
{"train_loss": 0.15585552155971527, "global_step": 150155, "epoch": 1687, "lr": 4.181574949156746e-05}
{"train_loss": 0.11046263575553894, "global_step": 150156, "epoch": 1687, "lr": 4.181517757462214e-05}
{"train_loss": 0.08965404331684113, "global_step": 150157, "epoch": 1687, "lr": 4.181460565877716e-05}
{"train_loss": 0.1444368213415146, "global_step": 150158, "epoch": 1687, "lr": 4.181403374403261e-05}
{"train_loss": 0.07855958491563797, "global_step": 150159, "epoch": 1687, "lr": 4.181346183038856e-05}
{"train_loss": 0.08994363993406296, "global_step": 150160, "epoch": 1687, "lr": 4.18128899178451e-05}
{"train_loss": 0.19128026068210602, "global_step": 150161, "epoch": 1687, "lr": 4.18123180064023e-05}
{"train_loss": 0.10784360021352768, "global_step": 150162, "epoch": 1687, "lr": 4.181174609606021e-05}
{"train_loss": 0.08262399584054947, "global_step": 150163, "epoch": 1687, "lr": 4.181117418681894e-05}
{"train_loss": 0.07429017871618271, "global_step": 150164, "epoch": 1687, "lr": 4.181060227867857e-05}
{"train_loss": 0.05000605061650276, "global_step": 150165, "epoch": 1687, "lr": 4.1810030371639144e-05}
{"train_loss": 0.09953747689723969, "global_step": 150166, "epoch": 1687, "lr": 4.180945846570079e-05}
{"train_loss": 0.12825116515159607, "global_step": 150167, "epoch": 1687, "lr": 4.1808886560863524e-05}
{"train_loss": 0.1135454997420311, "global_step": 150168, "epoch": 1687, "lr": 4.180831465712748e-05}
{"train_loss": 0.12601105868816376, "global_step": 150169, "epoch": 1687, "lr": 4.1807742754492685e-05}
{"train_loss": 0.16685383021831512, "global_step": 150170, "epoch": 1687, "lr": 4.180717085295926e-05}
{"train_loss": 0.17046847939491272, "global_step": 150171, "epoch": 1687, "lr": 4.180659895252724e-05}
{"train_loss": 0.09633301198482513, "global_step": 150172, "epoch": 1687, "lr": 4.1806027053196745e-05}
{"train_loss": 0.09193843603134155, "global_step": 150173, "epoch": 1687, "lr": 4.180545515496781e-05}
{"train_loss": 0.097696453332901, "global_step": 150174, "epoch": 1687, "lr": 4.1804883257840546e-05}
{"train_loss": 0.12414286285638809, "global_step": 150175, "epoch": 1687, "lr": 4.1804311361815015e-05}
{"train_loss": 0.12454694509506226, "global_step": 150176, "epoch": 1687, "lr": 4.180373946689129e-05}
{"train_loss": 0.150864377617836, "global_step": 150177, "epoch": 1687, "lr": 4.180316757306945e-05}
{"train_loss": 0.07172820717096329, "global_step": 150178, "epoch": 1687, "lr": 4.1802595680349574e-05}
{"train_loss": 0.0815725326538086, "global_step": 150179, "epoch": 1687, "lr": 4.1802023788731744e-05}
{"train_loss": 0.09408323466777802, "global_step": 150180, "epoch": 1687, "lr": 4.180145189821604e-05}
{"train_loss": 0.07723574340343475, "global_step": 150181, "epoch": 1687, "lr": 4.180088000880251e-05}
{"train_loss": 0.05429137498140335, "global_step": 150182, "epoch": 1687, "lr": 4.1800308120491245e-05}
{"train_loss": 0.10265184193849564, "global_step": 150183, "epoch": 1687, "lr": 4.179973623328236e-05}
{"train_loss": 0.14031371474266052, "global_step": 150184, "epoch": 1687, "lr": 4.179916434717587e-05}
{"train_loss": 0.15793979167938232, "global_step": 150185, "epoch": 1687, "lr": 4.179859246217191e-05}
{"train_loss": 0.08151119947433472, "global_step": 150186, "epoch": 1687, "lr": 4.179802057827051e-05}
{"train_loss": 0.08829563111066818, "global_step": 150187, "epoch": 1687, "lr": 4.179744869547178e-05}
{"train_loss": 0.07919007539749146, "global_step": 150188, "epoch": 1687, "lr": 4.1796876813775764e-05}
{"train_loss": 0.07465420663356781, "global_step": 150189, "epoch": 1687, "lr": 4.1796304933182584e-05}
{"train_loss": 0.10322621464729309, "global_step": 150190, "epoch": 1687, "lr": 4.179573305369227e-05}
{"train_loss": 0.0754791721701622, "global_step": 150191, "epoch": 1687, "lr": 4.179516117530493e-05}
{"train_loss": 0.1931646466255188, "global_step": 150192, "epoch": 1687, "lr": 4.179458929802062e-05}
{"train_loss": 0.14674311876296997, "global_step": 150193, "epoch": 1687, "lr": 4.179401742183945e-05}
{"train_loss": 0.07969096302986145, "global_step": 150194, "epoch": 1687, "lr": 4.1793445546761457e-05}
{"train_loss": 0.12774218618869781, "global_step": 150195, "epoch": 1687, "lr": 4.1792873672786734e-05}
{"train_loss": 0.09594441205263138, "global_step": 150196, "epoch": 1687, "lr": 4.179230179991537e-05}
{"train_loss": 0.08107787370681763, "global_step": 150197, "epoch": 1687, "lr": 4.179172992814742e-05}
{"train_loss": 0.09349602460861206, "global_step": 150198, "epoch": 1687, "lr": 4.179115805748297e-05}
{"train_loss": 0.052848778665065765, "global_step": 150199, "epoch": 1687, "lr": 4.17905861879221e-05}
{"train_loss": 0.10932666808366776, "global_step": 150200, "epoch": 1687, "lr": 4.1790014319464896e-05}
{"train_loss": 0.11141301691532135, "global_step": 150201, "epoch": 1687, "lr": 4.178944245211141e-05}
{"train_loss": 0.12439647316932678, "global_step": 150202, "epoch": 1687, "lr": 4.178887058586175e-05}
{"train_loss": 0.07388246804475784, "global_step": 150203, "epoch": 1687, "lr": 4.1788298720715956e-05}
{"train_loss": 0.056773897260427475, "global_step": 150204, "epoch": 1687, "lr": 4.178772685667415e-05}
{"train_loss": 0.16930389404296875, "global_step": 150205, "epoch": 1687, "lr": 4.178715499373636e-05}
{"train_loss": 0.051653873175382614, "global_step": 150206, "epoch": 1687, "lr": 4.178658313190271e-05}
{"train_loss": 0.1383553445339203, "global_step": 150207, "epoch": 1687, "lr": 4.1786011271173234e-05}
{"train_loss": 0.0718400701880455, "global_step": 150208, "epoch": 1687, "lr": 4.178543941154804e-05}
{"train_loss": 0.06543499231338501, "global_step": 150209, "epoch": 1687, "lr": 4.1784867553027185e-05}
{"train_loss": 0.04964052513241768, "global_step": 150210, "epoch": 1687, "lr": 4.178429569561077e-05}
{"train_loss": 0.07478123903274536, "global_step": 150211, "epoch": 1687, "lr": 4.178372383929883e-05}
{"train_loss": 0.11985272169113159, "global_step": 150212, "epoch": 1687, "lr": 4.17831519840915e-05}
{"train_loss": 0.0566038116812706, "global_step": 150213, "epoch": 1687, "lr": 4.1782580129988794e-05}
{"train_loss": 0.1574711799621582, "global_step": 150214, "epoch": 1687, "lr": 4.178200827699085e-05}
{"train_loss": 0.07638023048639297, "global_step": 150215, "epoch": 1687, "lr": 4.17814364250977e-05}
{"train_loss": 0.09575318545103073, "global_step": 150216, "epoch": 1687, "lr": 4.178086457430944e-05}
{"train_loss": 0.12861329317092896, "global_step": 150217, "epoch": 1687, "lr": 4.178029272462614e-05}
{"train_loss": 0.08781424164772034, "global_step": 150218, "epoch": 1687, "lr": 4.177972087604786e-05}
{"train_loss": 0.12293476611375809, "global_step": 150219, "epoch": 1687, "lr": 4.1779149028574726e-05}
{"train_loss": 0.07286232709884644, "global_step": 150220, "epoch": 1687, "lr": 4.1778577182206766e-05}
{"train_loss": 0.07274370640516281, "global_step": 150221, "epoch": 1687, "lr": 4.1778005336944095e-05}
{"train_loss": 0.10784000158309937, "global_step": 150222, "epoch": 1687, "lr": 4.177743349278674e-05}
{"train_loss": 0.11354038119316101, "global_step": 150223, "epoch": 1687, "lr": 4.177686164973485e-05}
{"train_loss": 0.09470833837985992, "global_step": 150224, "epoch": 1687, "lr": 4.177628980778842e-05}
{"train_loss": 0.16209819912910461, "global_step": 150225, "epoch": 1687, "lr": 4.17757179669476e-05}
{"train_loss": 0.07718469202518463, "global_step": 150226, "epoch": 1687, "lr": 4.177514612721241e-05}
{"train_loss": 0.07217389345169067, "global_step": 150227, "epoch": 1687, "lr": 4.177457428858297e-05}
{"train_loss": 0.07220674306154251, "global_step": 150228, "epoch": 1687, "lr": 4.177400245105932e-05}
{"train_loss": 0.08382993936538696, "global_step": 150229, "epoch": 1687, "lr": 4.177343061464158e-05}
{"train_loss": 0.11412282288074493, "global_step": 150230, "epoch": 1687, "lr": 4.177285877932977e-05}
{"train_loss": 0.10295638877354311, "global_step": 150231, "epoch": 1687, "lr": 4.1772286945124026e-05, "val_loss": 6.424319267272949}
{"train_loss": 0.1323423832654953, "global_step": 150232, "epoch": 1688, "lr": 4.1771715112024376e-05}
{"train_loss": 0.11942896246910095, "global_step": 150233, "epoch": 1688, "lr": 4.1771143280030944e-05}
{"train_loss": 0.12414655834436417, "global_step": 150234, "epoch": 1688, "lr": 4.177057144914377e-05}
{"train_loss": 0.09399089962244034, "global_step": 150235, "epoch": 1688, "lr": 4.176999961936293e-05}
{"train_loss": 0.1627195179462433, "global_step": 150236, "epoch": 1688, "lr": 4.176942779068854e-05}
{"train_loss": 0.14144425094127655, "global_step": 150237, "epoch": 1688, "lr": 4.176885596312062e-05}
{"train_loss": 0.10161600261926651, "global_step": 150238, "epoch": 1688, "lr": 4.17682841366593e-05}
{"train_loss": 0.12209329754114151, "global_step": 150239, "epoch": 1688, "lr": 4.176771231130462e-05}
{"train_loss": 0.08202450722455978, "global_step": 150240, "epoch": 1688, "lr": 4.176714048705668e-05}
{"train_loss": 0.08707743883132935, "global_step": 150241, "epoch": 1688, "lr": 4.176656866391554e-05}
{"train_loss": 0.1415022909641266, "global_step": 150242, "epoch": 1688, "lr": 4.1765996841881295e-05}
{"train_loss": 0.10393970459699631, "global_step": 150243, "epoch": 1688, "lr": 4.176542502095399e-05}
{"train_loss": 0.09191535413265228, "global_step": 150244, "epoch": 1688, "lr": 4.176485320113375e-05}
{"train_loss": 0.06887306272983551, "global_step": 150245, "epoch": 1688, "lr": 4.17642813824206e-05}
{"train_loss": 0.06307056546211243, "global_step": 150246, "epoch": 1688, "lr": 4.176370956481467e-05}
{"train_loss": 0.05711769312620163, "global_step": 150247, "epoch": 1688, "lr": 4.1763137748315975e-05}
{"train_loss": 0.11969012022018433, "global_step": 150248, "epoch": 1688, "lr": 4.176256593292466e-05}
{"train_loss": 0.07753554731607437, "global_step": 150249, "epoch": 1688, "lr": 4.176199411864074e-05}
{"train_loss": 0.12582212686538696, "global_step": 150250, "epoch": 1688, "lr": 4.176142230546434e-05}
{"train_loss": 0.1828530728816986, "global_step": 150251, "epoch": 1688, "lr": 4.17608504933955e-05}
{"train_loss": 0.10121779888868332, "global_step": 150252, "epoch": 1688, "lr": 4.176027868243433e-05}
{"train_loss": 0.09405391663312912, "global_step": 150253, "epoch": 1688, "lr": 4.1759706872580886e-05}
{"train_loss": 0.08595229685306549, "global_step": 150254, "epoch": 1688, "lr": 4.175913506383523e-05}
{"train_loss": 0.1918194591999054, "global_step": 150255, "epoch": 1688, "lr": 4.175856325619748e-05}
{"train_loss": 0.06972566992044449, "global_step": 150256, "epoch": 1688, "lr": 4.175799144966767e-05}
{"train_loss": 0.08318427950143814, "global_step": 150257, "epoch": 1688, "lr": 4.175741964424592e-05}
{"train_loss": 0.08644574880599976, "global_step": 150258, "epoch": 1688, "lr": 4.1756847839932265e-05}
{"train_loss": 0.15054570138454437, "global_step": 150259, "epoch": 1688, "lr": 4.175627603672682e-05}
{"train_loss": 0.09999562799930573, "global_step": 150260, "epoch": 1688, "lr": 4.1755704234629626e-05}
{"train_loss": 0.08119328320026398, "global_step": 150261, "epoch": 1688, "lr": 4.17551324336408e-05}
{"train_loss": 0.1109696477651596, "global_step": 150262, "epoch": 1688, "lr": 4.1754560633760363e-05}
{"train_loss": 0.10469216108322144, "global_step": 150263, "epoch": 1688, "lr": 4.175398883498846e-05}
{"train_loss": 0.030687376856803894, "global_step": 150264, "epoch": 1688, "lr": 4.17534170373251e-05}
{"train_loss": 0.12021425366401672, "global_step": 150265, "epoch": 1688, "lr": 4.1752845240770424e-05}
{"train_loss": 0.11913023889064789, "global_step": 150266, "epoch": 1688, "lr": 4.1752273445324444e-05}
{"train_loss": 0.07470924407243729, "global_step": 150267, "epoch": 1688, "lr": 4.17517016509873e-05}
{"train_loss": 0.07317192852497101, "global_step": 150268, "epoch": 1688, "lr": 4.175112985775902e-05}
{"train_loss": 0.08136825263500214, "global_step": 150269, "epoch": 1688, "lr": 4.175055806563972e-05}
{"train_loss": 0.10949913412332535, "global_step": 150270, "epoch": 1688, "lr": 4.174998627462943e-05}
{"train_loss": 0.12723448872566223, "global_step": 150271, "epoch": 1688, "lr": 4.174941448472827e-05}
{"train_loss": 0.08840312063694, "global_step": 150272, "epoch": 1688, "lr": 4.174884269593631e-05}
{"train_loss": 0.13172639906406403, "global_step": 150273, "epoch": 1688, "lr": 4.174827090825359e-05}
{"train_loss": 0.08159427344799042, "global_step": 150274, "epoch": 1688, "lr": 4.174769912168024e-05}
{"train_loss": 0.07472024112939835, "global_step": 150275, "epoch": 1688, "lr": 4.17471273362163e-05}
{"train_loss": 0.07908862829208374, "global_step": 150276, "epoch": 1688, "lr": 4.174655555186187e-05}
{"train_loss": 0.07733476907014847, "global_step": 150277, "epoch": 1688, "lr": 4.1745983768616985e-05}
{"train_loss": 0.07021187990903854, "global_step": 150278, "epoch": 1688, "lr": 4.174541198648179e-05}
{"train_loss": 0.05590582638978958, "global_step": 150279, "epoch": 1688, "lr": 4.1744840205456295e-05}
{"train_loss": 0.05931076779961586, "global_step": 150280, "epoch": 1688, "lr": 4.1744268425540623e-05}
{"train_loss": 0.16227011382579803, "global_step": 150281, "epoch": 1688, "lr": 4.1743696646734816e-05}
{"train_loss": 0.08232158422470093, "global_step": 150282, "epoch": 1688, "lr": 4.174312486903899e-05}
{"train_loss": 0.12468317151069641, "global_step": 150283, "epoch": 1688, "lr": 4.174255309245317e-05}
{"train_loss": 0.1005578488111496, "global_step": 150284, "epoch": 1688, "lr": 4.17419813169775e-05}
{"train_loss": 0.07314755022525787, "global_step": 150285, "epoch": 1688, "lr": 4.1741409542611994e-05}
{"train_loss": 0.12324745208024979, "global_step": 150286, "epoch": 1688, "lr": 4.174083776935677e-05}
{"train_loss": 0.07151614129543304, "global_step": 150287, "epoch": 1688, "lr": 4.174026599721187e-05}
{"train_loss": 0.11501538008451462, "global_step": 150288, "epoch": 1688, "lr": 4.1739694226177414e-05}
{"train_loss": 0.09180948138237, "global_step": 150289, "epoch": 1688, "lr": 4.173912245625344e-05}
{"train_loss": 0.10734360665082932, "global_step": 150290, "epoch": 1688, "lr": 4.1738550687440045e-05}
{"train_loss": 0.09214592725038528, "global_step": 150291, "epoch": 1688, "lr": 4.173797891973731e-05}
{"train_loss": 0.1012597382068634, "global_step": 150292, "epoch": 1688, "lr": 4.173740715314528e-05}
{"train_loss": 0.04482061415910721, "global_step": 150293, "epoch": 1688, "lr": 4.173683538766408e-05}
{"train_loss": 0.19755935668945312, "global_step": 150294, "epoch": 1688, "lr": 4.173626362329374e-05}
{"train_loss": 0.15913650393486023, "global_step": 150295, "epoch": 1688, "lr": 4.1735691860034384e-05}
{"train_loss": 0.11708612740039825, "global_step": 150296, "epoch": 1688, "lr": 4.1735120097886035e-05}
{"train_loss": 0.087944895029068, "global_step": 150297, "epoch": 1688, "lr": 4.1734548336848824e-05}
{"train_loss": 0.13748469948768616, "global_step": 150298, "epoch": 1688, "lr": 4.173397657692278e-05}
{"train_loss": 0.1185990497469902, "global_step": 150299, "epoch": 1688, "lr": 4.1733404818108026e-05}
{"train_loss": 0.1039826050400734, "global_step": 150300, "epoch": 1688, "lr": 4.173283306040459e-05}
{"train_loss": 0.12085670232772827, "global_step": 150301, "epoch": 1688, "lr": 4.173226130381259e-05}
{"train_loss": 0.10703500360250473, "global_step": 150302, "epoch": 1688, "lr": 4.173168954833207e-05}
{"train_loss": 0.1380416303873062, "global_step": 150303, "epoch": 1688, "lr": 4.173111779396315e-05}
{"train_loss": 0.10572067648172379, "global_step": 150304, "epoch": 1688, "lr": 4.173054604070586e-05}
{"train_loss": 0.07910171151161194, "global_step": 150305, "epoch": 1688, "lr": 4.172997428856031e-05}
{"train_loss": 0.11164785176515579, "global_step": 150306, "epoch": 1688, "lr": 4.172940253752655e-05}
{"train_loss": 0.06595764309167862, "global_step": 150307, "epoch": 1688, "lr": 4.172883078760469e-05}
{"train_loss": 0.12834466993808746, "global_step": 150308, "epoch": 1688, "lr": 4.1728259038794766e-05}
{"train_loss": 0.17270785570144653, "global_step": 150309, "epoch": 1688, "lr": 4.172768729109689e-05}
{"train_loss": 0.1364586353302002, "global_step": 150310, "epoch": 1688, "lr": 4.172711554451113e-05}
{"train_loss": 0.09380904585123062, "global_step": 150311, "epoch": 1688, "lr": 4.1726543799037545e-05}
{"train_loss": 0.10687685757875443, "global_step": 150312, "epoch": 1688, "lr": 4.172597205467624e-05}
{"train_loss": 0.07173006981611252, "global_step": 150313, "epoch": 1688, "lr": 4.1725400311427254e-05}
{"train_loss": 0.15579259395599365, "global_step": 150314, "epoch": 1688, "lr": 4.1724828569290716e-05}
{"train_loss": 0.09785077720880508, "global_step": 150315, "epoch": 1688, "lr": 4.172425682826665e-05}
{"train_loss": 0.16512997448444366, "global_step": 150316, "epoch": 1688, "lr": 4.172368508835517e-05}
{"train_loss": 0.2019290030002594, "global_step": 150317, "epoch": 1688, "lr": 4.172311334955633e-05}
{"train_loss": 0.06748621165752411, "global_step": 150318, "epoch": 1688, "lr": 4.172254161187024e-05}
{"train_loss": 0.1375918686389923, "global_step": 150319, "epoch": 1688, "lr": 4.172196987529692e-05}
{"train_loss": 0.10693941955988327, "global_step": 150320, "epoch": 1688, "lr": 4.172139813983651e-05, "val_loss": 6.473386287689209}
{"train_loss": 0.16224218904972076, "global_step": 150321, "epoch": 1689, "lr": 4.172082640548903e-05}
{"train_loss": 0.09297182410955429, "global_step": 150322, "epoch": 1689, "lr": 4.172025467225461e-05}
{"train_loss": 0.1560169905424118, "global_step": 150323, "epoch": 1689, "lr": 4.1719682940133276e-05}
{"train_loss": 0.14780046045780182, "global_step": 150324, "epoch": 1689, "lr": 4.171911120912516e-05}
{"train_loss": 0.10997184365987778, "global_step": 150325, "epoch": 1689, "lr": 4.171853947923027e-05}
{"train_loss": 0.09372887760400772, "global_step": 150326, "epoch": 1689, "lr": 4.171796775044875e-05}
{"train_loss": 0.06465709209442139, "global_step": 150327, "epoch": 1689, "lr": 4.1717396022780636e-05}
{"train_loss": 0.0952712669968605, "global_step": 150328, "epoch": 1689, "lr": 4.171682429622602e-05}
{"train_loss": 0.1194826140999794, "global_step": 150329, "epoch": 1689, "lr": 4.171625257078498e-05}
{"train_loss": 0.08878900855779648, "global_step": 150330, "epoch": 1689, "lr": 4.171568084645759e-05}
{"train_loss": 0.09801046550273895, "global_step": 150331, "epoch": 1689, "lr": 4.171510912324393e-05}
{"train_loss": 0.18882258236408234, "global_step": 150332, "epoch": 1689, "lr": 4.171453740114405e-05}
{"train_loss": 0.09307139366865158, "global_step": 150333, "epoch": 1689, "lr": 4.1713965680158077e-05}
{"train_loss": 0.09604838490486145, "global_step": 150334, "epoch": 1689, "lr": 4.1713393960286043e-05}
{"train_loss": 0.06834030896425247, "global_step": 150335, "epoch": 1689, "lr": 4.1712822241528055e-05}
{"train_loss": 0.10820992290973663, "global_step": 150336, "epoch": 1689, "lr": 4.171225052388415e-05}
{"train_loss": 0.041738905012607574, "global_step": 150337, "epoch": 1689, "lr": 4.171167880735447e-05}
{"train_loss": 0.09403139352798462, "global_step": 150338, "epoch": 1689, "lr": 4.171110709193902e-05}
{"train_loss": 0.11656998097896576, "global_step": 150339, "epoch": 1689, "lr": 4.171053537763794e-05}
{"train_loss": 0.1111445426940918, "global_step": 150340, "epoch": 1689, "lr": 4.170996366445125e-05}
{"train_loss": 0.05783718824386597, "global_step": 150341, "epoch": 1689, "lr": 4.170939195237908e-05}
{"train_loss": 0.09492295235395432, "global_step": 150342, "epoch": 1689, "lr": 4.170882024142146e-05}
{"train_loss": 0.12676039338111877, "global_step": 150343, "epoch": 1689, "lr": 4.1708248531578504e-05}
{"train_loss": 0.08785456418991089, "global_step": 150344, "epoch": 1689, "lr": 4.170767682285025e-05}
{"train_loss": 0.13274388015270233, "global_step": 150345, "epoch": 1689, "lr": 4.170710511523682e-05}
{"train_loss": 0.10276313126087189, "global_step": 150346, "epoch": 1689, "lr": 4.170653340873826e-05}
{"train_loss": 0.07623885571956635, "global_step": 150347, "epoch": 1689, "lr": 4.1705961703354655e-05}
{"train_loss": 0.13384677469730377, "global_step": 150348, "epoch": 1689, "lr": 4.170538999908609e-05}
{"train_loss": 0.07815827429294586, "global_step": 150349, "epoch": 1689, "lr": 4.170481829593262e-05}
{"train_loss": 0.07422379404306412, "global_step": 150350, "epoch": 1689, "lr": 4.170424659389436e-05}
{"train_loss": 0.11642798781394958, "global_step": 150351, "epoch": 1689, "lr": 4.170367489297134e-05}
{"train_loss": 0.08997538685798645, "global_step": 150352, "epoch": 1689, "lr": 4.170310319316368e-05}
{"train_loss": 0.09316025674343109, "global_step": 150353, "epoch": 1689, "lr": 4.170253149447141e-05}
{"train_loss": 0.12426762282848358, "global_step": 150354, "epoch": 1689, "lr": 4.170195979689466e-05}
{"train_loss": 0.10365408658981323, "global_step": 150355, "epoch": 1689, "lr": 4.1701388100433456e-05}
{"train_loss": 0.10911400616168976, "global_step": 150356, "epoch": 1689, "lr": 4.170081640508792e-05}
{"train_loss": 0.0930824875831604, "global_step": 150357, "epoch": 1689, "lr": 4.170024471085809e-05}
{"train_loss": 0.08489729464054108, "global_step": 150358, "epoch": 1689, "lr": 4.1699673017744076e-05}
{"train_loss": 0.11487411707639694, "global_step": 150359, "epoch": 1689, "lr": 4.1699101325745925e-05}
{"train_loss": 0.10871957242488861, "global_step": 150360, "epoch": 1689, "lr": 4.169852963486375e-05}
{"train_loss": 0.07520930469036102, "global_step": 150361, "epoch": 1689, "lr": 4.1697957945097586e-05}
{"train_loss": 0.10277014225721359, "global_step": 150362, "epoch": 1689, "lr": 4.169738625644753e-05}
{"train_loss": 0.15907539427280426, "global_step": 150363, "epoch": 1689, "lr": 4.169681456891368e-05}
{"train_loss": 0.2055598795413971, "global_step": 150364, "epoch": 1689, "lr": 4.169624288249608e-05}
{"train_loss": 0.06293296068906784, "global_step": 150365, "epoch": 1689, "lr": 4.169567119719482e-05}
{"train_loss": 0.10245340317487717, "global_step": 150366, "epoch": 1689, "lr": 4.1695099513009974e-05}
{"train_loss": 0.13077063858509064, "global_step": 150367, "epoch": 1689, "lr": 4.1694527829941624e-05}
{"train_loss": 0.10173238068819046, "global_step": 150368, "epoch": 1689, "lr": 4.169395614798983e-05}
{"train_loss": 0.14925722777843475, "global_step": 150369, "epoch": 1689, "lr": 4.169338446715471e-05}
{"train_loss": 0.16194023191928864, "global_step": 150370, "epoch": 1689, "lr": 4.1692812787436286e-05}
{"train_loss": 0.13498280942440033, "global_step": 150371, "epoch": 1689, "lr": 4.1692241108834687e-05}
{"train_loss": 0.15428468585014343, "global_step": 150372, "epoch": 1689, "lr": 4.169166943134993e-05}
{"train_loss": 0.15789614617824554, "global_step": 150373, "epoch": 1689, "lr": 4.169109775498216e-05}
{"train_loss": 0.15152326226234436, "global_step": 150374, "epoch": 1689, "lr": 4.1690526079731395e-05}
{"train_loss": 0.10864967852830887, "global_step": 150375, "epoch": 1689, "lr": 4.1689954405597766e-05}
{"train_loss": 0.1374472975730896, "global_step": 150376, "epoch": 1689, "lr": 4.168938273258129e-05}
{"train_loss": 0.13757845759391785, "global_step": 150377, "epoch": 1689, "lr": 4.168881106068211e-05}
{"train_loss": 0.09440328925848007, "global_step": 150378, "epoch": 1689, "lr": 4.1688239389900235e-05}
{"train_loss": 0.13475283980369568, "global_step": 150379, "epoch": 1689, "lr": 4.16876677202358e-05}
{"train_loss": 0.08350381255149841, "global_step": 150380, "epoch": 1689, "lr": 4.1687096051688836e-05}
{"train_loss": 0.12234175950288773, "global_step": 150381, "epoch": 1689, "lr": 4.168652438425944e-05}
{"train_loss": 0.07915496081113815, "global_step": 150382, "epoch": 1689, "lr": 4.168595271794771e-05}
{"train_loss": 0.10461866110563278, "global_step": 150383, "epoch": 1689, "lr": 4.1685381052753693e-05}
{"train_loss": 0.08097335696220398, "global_step": 150384, "epoch": 1689, "lr": 4.168480938867747e-05}
{"train_loss": 0.10060105472803116, "global_step": 150385, "epoch": 1689, "lr": 4.1684237725719124e-05}
{"train_loss": 0.09904831647872925, "global_step": 150386, "epoch": 1689, "lr": 4.1683666063878735e-05}
{"train_loss": 0.13810214400291443, "global_step": 150387, "epoch": 1689, "lr": 4.1683094403156374e-05}
{"train_loss": 0.0673353299498558, "global_step": 150388, "epoch": 1689, "lr": 4.168252274355213e-05}
{"train_loss": 0.12903977930545807, "global_step": 150389, "epoch": 1689, "lr": 4.168195108506605e-05}
{"train_loss": 0.07463587820529938, "global_step": 150390, "epoch": 1689, "lr": 4.1681379427698255e-05}
{"train_loss": 0.09807393699884415, "global_step": 150391, "epoch": 1689, "lr": 4.168080777144877e-05}
{"train_loss": 0.10607118904590607, "global_step": 150392, "epoch": 1689, "lr": 4.168023611631772e-05}
{"train_loss": 0.08462337404489517, "global_step": 150393, "epoch": 1689, "lr": 4.167966446230514e-05}
{"train_loss": 0.09088534861803055, "global_step": 150394, "epoch": 1689, "lr": 4.1679092809411156e-05}
{"train_loss": 0.09509699791669846, "global_step": 150395, "epoch": 1689, "lr": 4.167852115763579e-05}
{"train_loss": 0.11260638386011124, "global_step": 150396, "epoch": 1689, "lr": 4.167794950697917e-05}
{"train_loss": 0.09462696313858032, "global_step": 150397, "epoch": 1689, "lr": 4.167737785744132e-05}
{"train_loss": 0.15030615031719208, "global_step": 150398, "epoch": 1689, "lr": 4.1676806209022344e-05}
{"train_loss": 0.05294501408934593, "global_step": 150399, "epoch": 1689, "lr": 4.167623456172235e-05}
{"train_loss": 0.09178610891103745, "global_step": 150400, "epoch": 1689, "lr": 4.167566291554136e-05}
{"train_loss": 0.10701671242713928, "global_step": 150401, "epoch": 1689, "lr": 4.167509127047949e-05}
{"train_loss": 0.09617313742637634, "global_step": 150402, "epoch": 1689, "lr": 4.16745196265368e-05}
{"train_loss": 0.053101710975170135, "global_step": 150403, "epoch": 1689, "lr": 4.167394798371337e-05}
{"train_loss": 0.16736772656440735, "global_step": 150404, "epoch": 1689, "lr": 4.167337634200927e-05}
{"train_loss": 0.0989302545785904, "global_step": 150405, "epoch": 1689, "lr": 4.167280470142459e-05}
{"train_loss": 0.07413201779127121, "global_step": 150406, "epoch": 1689, "lr": 4.16722330619594e-05}
{"train_loss": 0.05808784440159798, "global_step": 150407, "epoch": 1689, "lr": 4.1671661423613784e-05}
{"train_loss": 0.08198313415050507, "global_step": 150408, "epoch": 1689, "lr": 4.167108978638779e-05}
{"train_loss": 0.10695936488971282, "global_step": 150409, "epoch": 1689, "lr": 4.167051815028154e-05, "val_loss": 6.455347537994385}
{"train_loss": 0.10422802716493607, "global_step": 150410, "epoch": 1690, "lr": 4.166994651529506e-05}
{"train_loss": 0.14164133369922638, "global_step": 150411, "epoch": 1690, "lr": 4.166937488142848e-05}
{"train_loss": 0.07433516532182693, "global_step": 150412, "epoch": 1690, "lr": 4.166880324868183e-05}
{"train_loss": 0.0721387043595314, "global_step": 150413, "epoch": 1690, "lr": 4.166823161705523e-05}
{"train_loss": 0.06788113713264465, "global_step": 150414, "epoch": 1690, "lr": 4.1667659986548715e-05}
{"train_loss": 0.1005115956068039, "global_step": 150415, "epoch": 1690, "lr": 4.16670883571624e-05}
{"train_loss": 0.07080382853746414, "global_step": 150416, "epoch": 1690, "lr": 4.1666516728896315e-05}
{"train_loss": 0.08184384554624557, "global_step": 150417, "epoch": 1690, "lr": 4.166594510175058e-05}
{"train_loss": 0.1033673956990242, "global_step": 150418, "epoch": 1690, "lr": 4.166537347572527e-05}
{"train_loss": 0.0628238394856453, "global_step": 150419, "epoch": 1690, "lr": 4.166480185082042e-05}
{"train_loss": 0.1054069772362709, "global_step": 150420, "epoch": 1690, "lr": 4.1664230227036165e-05}
{"train_loss": 0.10193321108818054, "global_step": 150421, "epoch": 1690, "lr": 4.166365860437255e-05}
{"train_loss": 0.08996802568435669, "global_step": 150422, "epoch": 1690, "lr": 4.166308698282964e-05}
{"train_loss": 0.09154979139566422, "global_step": 150423, "epoch": 1690, "lr": 4.1662515362407536e-05}
{"train_loss": 0.0879102274775505, "global_step": 150424, "epoch": 1690, "lr": 4.16619437431063e-05}
{"train_loss": 0.10120395570993423, "global_step": 150425, "epoch": 1690, "lr": 4.166137212492602e-05}
{"train_loss": 0.1892552524805069, "global_step": 150426, "epoch": 1690, "lr": 4.166080050786676e-05}
{"train_loss": 0.09885977953672409, "global_step": 150427, "epoch": 1690, "lr": 4.166022889192861e-05}
{"train_loss": 0.11778371036052704, "global_step": 150428, "epoch": 1690, "lr": 4.1659657277111646e-05}
{"train_loss": 0.09330695867538452, "global_step": 150429, "epoch": 1690, "lr": 4.165908566341592e-05}
{"train_loss": 0.11470113694667816, "global_step": 150430, "epoch": 1690, "lr": 4.165851405084155e-05}
{"train_loss": 0.09245471656322479, "global_step": 150431, "epoch": 1690, "lr": 4.165794243938857e-05}
{"train_loss": 0.10731527954339981, "global_step": 150432, "epoch": 1690, "lr": 4.1657370829057095e-05}
{"train_loss": 0.20848222076892853, "global_step": 150433, "epoch": 1690, "lr": 4.165679921984717e-05}
{"train_loss": 0.1210210919380188, "global_step": 150434, "epoch": 1690, "lr": 4.165622761175891e-05}
{"train_loss": 0.06432721018791199, "global_step": 150435, "epoch": 1690, "lr": 4.1655656004792344e-05}
{"train_loss": 0.07796802371740341, "global_step": 150436, "epoch": 1690, "lr": 4.165508439894757e-05}
{"train_loss": 0.09807884693145752, "global_step": 150437, "epoch": 1690, "lr": 4.165451279422469e-05}
{"train_loss": 0.08007930219173431, "global_step": 150438, "epoch": 1690, "lr": 4.165394119062374e-05}
{"train_loss": 0.11158356070518494, "global_step": 150439, "epoch": 1690, "lr": 4.1653369588144834e-05}
{"train_loss": 0.11009426414966583, "global_step": 150440, "epoch": 1690, "lr": 4.165279798678801e-05}
{"train_loss": 0.09526854008436203, "global_step": 150441, "epoch": 1690, "lr": 4.16522263865534e-05}
{"train_loss": 0.0839519277215004, "global_step": 150442, "epoch": 1690, "lr": 4.165165478744102e-05}
{"train_loss": 0.10552314668893814, "global_step": 150443, "epoch": 1690, "lr": 4.165108318945097e-05}
{"train_loss": 0.1352926790714264, "global_step": 150444, "epoch": 1690, "lr": 4.165051159258334e-05}
{"train_loss": 0.06368161737918854, "global_step": 150445, "epoch": 1690, "lr": 4.1649939996838194e-05}
{"train_loss": 0.08307422697544098, "global_step": 150446, "epoch": 1690, "lr": 4.1649368402215615e-05}
{"train_loss": 0.07160990685224533, "global_step": 150447, "epoch": 1690, "lr": 4.1648796808715686e-05}
{"train_loss": 0.10407917946577072, "global_step": 150448, "epoch": 1690, "lr": 4.164822521633845e-05}
{"train_loss": 0.05214136466383934, "global_step": 150449, "epoch": 1690, "lr": 4.164765362508403e-05}
{"train_loss": 0.07933726906776428, "global_step": 150450, "epoch": 1690, "lr": 4.164708203495247e-05}
{"train_loss": 0.08304057270288467, "global_step": 150451, "epoch": 1690, "lr": 4.164651044594386e-05}
{"train_loss": 0.06764426082372665, "global_step": 150452, "epoch": 1690, "lr": 4.164593885805827e-05}
{"train_loss": 0.08945536613464355, "global_step": 150453, "epoch": 1690, "lr": 4.164536727129578e-05}
{"train_loss": 0.1034831702709198, "global_step": 150454, "epoch": 1690, "lr": 4.164479568565648e-05}
{"train_loss": 0.10366839915513992, "global_step": 150455, "epoch": 1690, "lr": 4.164422410114042e-05}
{"train_loss": 0.12017564475536346, "global_step": 150456, "epoch": 1690, "lr": 4.1643652517747714e-05}
{"train_loss": 0.09702219069004059, "global_step": 150457, "epoch": 1690, "lr": 4.164308093547839e-05}
{"train_loss": 0.07204946130514145, "global_step": 150458, "epoch": 1690, "lr": 4.164250935433258e-05}
{"train_loss": 0.11901897937059402, "global_step": 150459, "epoch": 1690, "lr": 4.164193777431032e-05}
{"train_loss": 0.09524638950824738, "global_step": 150460, "epoch": 1690, "lr": 4.1641366195411705e-05}
{"train_loss": 0.10323324054479599, "global_step": 150461, "epoch": 1690, "lr": 4.16407946176368e-05}
{"train_loss": 0.09834036231040955, "global_step": 150462, "epoch": 1690, "lr": 4.164022304098569e-05}
{"train_loss": 0.07440897077322006, "global_step": 150463, "epoch": 1690, "lr": 4.163965146545845e-05}
{"train_loss": 0.09735181927680969, "global_step": 150464, "epoch": 1690, "lr": 4.163907989105516e-05}
{"train_loss": 0.11054433137178421, "global_step": 150465, "epoch": 1690, "lr": 4.1638508317775896e-05}
{"train_loss": 0.07704845815896988, "global_step": 150466, "epoch": 1690, "lr": 4.163793674562073e-05}
{"train_loss": 0.11987592279911041, "global_step": 150467, "epoch": 1690, "lr": 4.163736517458973e-05}
{"train_loss": 0.06516755372285843, "global_step": 150468, "epoch": 1690, "lr": 4.1636793604683006e-05}
{"train_loss": 0.0851307362318039, "global_step": 150469, "epoch": 1690, "lr": 4.163622203590059e-05}
{"train_loss": 0.07773681730031967, "global_step": 150470, "epoch": 1690, "lr": 4.163565046824261e-05}
{"train_loss": 0.11930336803197861, "global_step": 150471, "epoch": 1690, "lr": 4.163507890170908e-05}
{"train_loss": 0.07566557824611664, "global_step": 150472, "epoch": 1690, "lr": 4.1634507336300115e-05}
{"train_loss": 0.09411609172821045, "global_step": 150473, "epoch": 1690, "lr": 4.163393577201582e-05}
{"train_loss": 0.039086826145648956, "global_step": 150474, "epoch": 1690, "lr": 4.163336420885621e-05}
{"train_loss": 0.09480604529380798, "global_step": 150475, "epoch": 1690, "lr": 4.163279264682141e-05}
{"train_loss": 0.05736315995454788, "global_step": 150476, "epoch": 1690, "lr": 4.1632221085911464e-05}
{"train_loss": 0.1642354130744934, "global_step": 150477, "epoch": 1690, "lr": 4.163164952612649e-05}
{"train_loss": 0.11875930428504944, "global_step": 150478, "epoch": 1690, "lr": 4.163107796746651e-05}
{"train_loss": 0.16566802561283112, "global_step": 150479, "epoch": 1690, "lr": 4.1630506409931655e-05}
{"train_loss": 0.06645045429468155, "global_step": 150480, "epoch": 1690, "lr": 4.1629934853521964e-05}
{"train_loss": 0.12003815919160843, "global_step": 150481, "epoch": 1690, "lr": 4.162936329823752e-05}
{"train_loss": 0.09248878806829453, "global_step": 150482, "epoch": 1690, "lr": 4.1628791744078416e-05}
{"train_loss": 0.15627606213092804, "global_step": 150483, "epoch": 1690, "lr": 4.1628220191044714e-05}
{"train_loss": 0.10706564784049988, "global_step": 150484, "epoch": 1690, "lr": 4.1627648639136495e-05}
{"train_loss": 0.13032928109169006, "global_step": 150485, "epoch": 1690, "lr": 4.162707708835385e-05}
{"train_loss": 0.12224403768777847, "global_step": 150486, "epoch": 1690, "lr": 4.162650553869682e-05}
{"train_loss": 0.12166619300842285, "global_step": 150487, "epoch": 1690, "lr": 4.1625933990165524e-05}
{"train_loss": 0.07955681532621384, "global_step": 150488, "epoch": 1690, "lr": 4.1625362442760004e-05}
{"train_loss": 0.07754485309123993, "global_step": 150489, "epoch": 1690, "lr": 4.1624790896480346e-05}
{"train_loss": 0.11745163798332214, "global_step": 150490, "epoch": 1690, "lr": 4.162421935132665e-05}
{"train_loss": 0.10989979654550552, "global_step": 150491, "epoch": 1690, "lr": 4.1623647807298965e-05}
{"train_loss": 0.138467475771904, "global_step": 150492, "epoch": 1690, "lr": 4.1623076264397396e-05}
{"train_loss": 0.10839047282934189, "global_step": 150493, "epoch": 1690, "lr": 4.162250472262198e-05}
{"train_loss": 0.07885510474443436, "global_step": 150494, "epoch": 1690, "lr": 4.162193318197284e-05}
{"train_loss": 0.11439579725265503, "global_step": 150495, "epoch": 1690, "lr": 4.162136164245e-05}
{"train_loss": 0.11028694361448288, "global_step": 150496, "epoch": 1690, "lr": 4.162079010405359e-05}
{"train_loss": 0.08742271363735199, "global_step": 150497, "epoch": 1690, "lr": 4.162021856678364e-05}
{"train_loss": 0.0986791377824344, "global_step": 150498, "epoch": 1690, "lr": 4.161964703064028e-05, "val_loss": 6.439560413360596, "train_action_mse_error": 16.82312774658203}
{"train_loss": 0.10228423774242401, "global_step": 150499, "epoch": 1691, "lr": 4.1619075495623536e-05}
{"train_loss": 0.10673884302377701, "global_step": 150500, "epoch": 1691, "lr": 4.161850396173351e-05}
{"train_loss": 0.04592607170343399, "global_step": 150501, "epoch": 1691, "lr": 4.1617932428970276e-05}
{"train_loss": 0.09980326890945435, "global_step": 150502, "epoch": 1691, "lr": 4.16173608973339e-05}
{"train_loss": 0.09554547816514969, "global_step": 150503, "epoch": 1691, "lr": 4.161678936682448e-05}
{"train_loss": 0.11411365866661072, "global_step": 150504, "epoch": 1691, "lr": 4.1616217837442084e-05}
{"train_loss": 0.09834125638008118, "global_step": 150505, "epoch": 1691, "lr": 4.161564630918676e-05}
{"train_loss": 0.1250859946012497, "global_step": 150506, "epoch": 1691, "lr": 4.161507478205864e-05}
{"train_loss": 0.08371083438396454, "global_step": 150507, "epoch": 1691, "lr": 4.161450325605775e-05}
{"train_loss": 0.11022314429283142, "global_step": 150508, "epoch": 1691, "lr": 4.161393173118417e-05}
{"train_loss": 0.09148436039686203, "global_step": 150509, "epoch": 1691, "lr": 4.161336020743804e-05}
{"train_loss": 0.15285389125347137, "global_step": 150510, "epoch": 1691, "lr": 4.161278868481935e-05}
{"train_loss": 0.10900863260030746, "global_step": 150511, "epoch": 1691, "lr": 4.1612217163328255e-05}
{"train_loss": 0.0828288123011589, "global_step": 150512, "epoch": 1691, "lr": 4.161164564296477e-05}
{"train_loss": 0.07413331419229507, "global_step": 150513, "epoch": 1691, "lr": 4.161107412372902e-05}
{"train_loss": 0.07855421304702759, "global_step": 150514, "epoch": 1691, "lr": 4.1610502605621025e-05}
{"train_loss": 0.10398244857788086, "global_step": 150515, "epoch": 1691, "lr": 4.160993108864093e-05}
{"train_loss": 0.12391830235719681, "global_step": 150516, "epoch": 1691, "lr": 4.1609359572788754e-05}
{"train_loss": 0.07888732850551605, "global_step": 150517, "epoch": 1691, "lr": 4.160878805806461e-05}
{"train_loss": 0.10064981132745743, "global_step": 150518, "epoch": 1691, "lr": 4.160821654446856e-05}
{"train_loss": 0.07871060073375702, "global_step": 150519, "epoch": 1691, "lr": 4.160764503200069e-05}
{"train_loss": 0.09105345606803894, "global_step": 150520, "epoch": 1691, "lr": 4.160707352066105e-05}
{"train_loss": 0.1146392971277237, "global_step": 150521, "epoch": 1691, "lr": 4.160650201044975e-05}
{"train_loss": 0.06781569868326187, "global_step": 150522, "epoch": 1691, "lr": 4.1605930501366855e-05}
{"train_loss": 0.0882478728890419, "global_step": 150523, "epoch": 1691, "lr": 4.160535899341245e-05}
{"train_loss": 0.06218014657497406, "global_step": 150524, "epoch": 1691, "lr": 4.160478748658657e-05}
{"train_loss": 0.07070281356573105, "global_step": 150525, "epoch": 1691, "lr": 4.160421598088934e-05}
{"train_loss": 0.13662344217300415, "global_step": 150526, "epoch": 1691, "lr": 4.160364447632083e-05}
{"train_loss": 0.07260142266750336, "global_step": 150527, "epoch": 1691, "lr": 4.160307297288109e-05}
{"train_loss": 0.09120607376098633, "global_step": 150528, "epoch": 1691, "lr": 4.1602501470570234e-05}
{"train_loss": 0.12731598317623138, "global_step": 150529, "epoch": 1691, "lr": 4.1601929969388305e-05}
{"train_loss": 0.12405909597873688, "global_step": 150530, "epoch": 1691, "lr": 4.160135846933541e-05}
{"train_loss": 0.06380908936262131, "global_step": 150531, "epoch": 1691, "lr": 4.160078697041159e-05}
{"train_loss": 0.07445358484983444, "global_step": 150532, "epoch": 1691, "lr": 4.160021547261696e-05}
{"train_loss": 0.10886632651090622, "global_step": 150533, "epoch": 1691, "lr": 4.159964397595156e-05}
{"train_loss": 0.06993231922388077, "global_step": 150534, "epoch": 1691, "lr": 4.159907248041551e-05}
{"train_loss": 0.07118500769138336, "global_step": 150535, "epoch": 1691, "lr": 4.159850098600884e-05}
{"train_loss": 0.0994538739323616, "global_step": 150536, "epoch": 1691, "lr": 4.1597929492731665e-05}
{"train_loss": 0.10644953697919846, "global_step": 150537, "epoch": 1691, "lr": 4.159735800058404e-05}
{"train_loss": 0.13294661045074463, "global_step": 150538, "epoch": 1691, "lr": 4.1596786509566046e-05}
{"train_loss": 0.07789924740791321, "global_step": 150539, "epoch": 1691, "lr": 4.159621501967776e-05}
{"train_loss": 0.11160027235746384, "global_step": 150540, "epoch": 1691, "lr": 4.1595643530919265e-05}
{"train_loss": 0.13387741148471832, "global_step": 150541, "epoch": 1691, "lr": 4.159507204329063e-05}
{"train_loss": 0.12415331602096558, "global_step": 150542, "epoch": 1691, "lr": 4.159450055679194e-05}
{"train_loss": 0.10477103292942047, "global_step": 150543, "epoch": 1691, "lr": 4.1593929071423256e-05}
{"train_loss": 0.10324018448591232, "global_step": 150544, "epoch": 1691, "lr": 4.159335758718466e-05}
{"train_loss": 0.07186788320541382, "global_step": 150545, "epoch": 1691, "lr": 4.159278610407627e-05}
{"train_loss": 0.08716884255409241, "global_step": 150546, "epoch": 1691, "lr": 4.1592214622098086e-05}
{"train_loss": 0.14146572351455688, "global_step": 150547, "epoch": 1691, "lr": 4.159164314125026e-05}
{"train_loss": 0.09133142232894897, "global_step": 150548, "epoch": 1691, "lr": 4.15910716615328e-05}
{"train_loss": 0.09768526256084442, "global_step": 150549, "epoch": 1691, "lr": 4.159050018294585e-05}
{"train_loss": 0.09425884485244751, "global_step": 150550, "epoch": 1691, "lr": 4.1589928705489435e-05}
{"train_loss": 0.0921144187450409, "global_step": 150551, "epoch": 1691, "lr": 4.1589357229163674e-05}
{"train_loss": 0.14575420320034027, "global_step": 150552, "epoch": 1691, "lr": 4.158878575396859e-05}
{"train_loss": 0.13940384984016418, "global_step": 150553, "epoch": 1691, "lr": 4.158821427990432e-05}
{"train_loss": 0.1416490525007248, "global_step": 150554, "epoch": 1691, "lr": 4.158764280697089e-05}
{"train_loss": 0.07880131155252457, "global_step": 150555, "epoch": 1691, "lr": 4.158707133516842e-05}
{"train_loss": 0.12439164519309998, "global_step": 150556, "epoch": 1691, "lr": 4.158649986449694e-05}
{"train_loss": 0.11158416420221329, "global_step": 150557, "epoch": 1691, "lr": 4.1585928394956584e-05}
{"train_loss": 0.07838281989097595, "global_step": 150558, "epoch": 1691, "lr": 4.158535692654738e-05}
{"train_loss": 0.11092908680438995, "global_step": 150559, "epoch": 1691, "lr": 4.1584785459269424e-05}
{"train_loss": 0.1079082041978836, "global_step": 150560, "epoch": 1691, "lr": 4.158421399312279e-05}
{"train_loss": 0.12788911163806915, "global_step": 150561, "epoch": 1691, "lr": 4.1583642528107554e-05}
{"train_loss": 0.09164872020483017, "global_step": 150562, "epoch": 1691, "lr": 4.158307106422379e-05}
{"train_loss": 0.07403148710727692, "global_step": 150563, "epoch": 1691, "lr": 4.158249960147158e-05}
{"train_loss": 0.1469767838716507, "global_step": 150564, "epoch": 1691, "lr": 4.158192813985102e-05}
{"train_loss": 0.1467134803533554, "global_step": 150565, "epoch": 1691, "lr": 4.158135667936214e-05}
{"train_loss": 0.07196040451526642, "global_step": 150566, "epoch": 1691, "lr": 4.158078522000507e-05}
{"train_loss": 0.11595343798398972, "global_step": 150567, "epoch": 1691, "lr": 4.1580213761779834e-05}
{"train_loss": 0.09945139288902283, "global_step": 150568, "epoch": 1691, "lr": 4.157964230468656e-05}
{"train_loss": 0.10957708209753036, "global_step": 150569, "epoch": 1691, "lr": 4.157907084872528e-05}
{"train_loss": 0.05902773141860962, "global_step": 150570, "epoch": 1691, "lr": 4.157849939389612e-05}
{"train_loss": 0.09207046031951904, "global_step": 150571, "epoch": 1691, "lr": 4.157792794019909e-05}
{"train_loss": 0.07157235592603683, "global_step": 150572, "epoch": 1691, "lr": 4.157735648763434e-05}
{"train_loss": 0.1467578411102295, "global_step": 150573, "epoch": 1691, "lr": 4.157678503620188e-05}
{"train_loss": 0.1358293741941452, "global_step": 150574, "epoch": 1691, "lr": 4.157621358590184e-05}
{"train_loss": 0.08527365326881409, "global_step": 150575, "epoch": 1691, "lr": 4.157564213673427e-05}
{"train_loss": 0.07194770127534866, "global_step": 150576, "epoch": 1691, "lr": 4.157507068869926e-05}
{"train_loss": 0.10523153841495514, "global_step": 150577, "epoch": 1691, "lr": 4.157449924179687e-05}
{"train_loss": 0.07621041685342789, "global_step": 150578, "epoch": 1691, "lr": 4.157392779602719e-05}
{"train_loss": 0.09868041425943375, "global_step": 150579, "epoch": 1691, "lr": 4.1573356351390294e-05}
{"train_loss": 0.18354378640651703, "global_step": 150580, "epoch": 1691, "lr": 4.157278490788624e-05}
{"train_loss": 0.09187005460262299, "global_step": 150581, "epoch": 1691, "lr": 4.157221346551515e-05}
{"train_loss": 0.09719040244817734, "global_step": 150582, "epoch": 1691, "lr": 4.1571642024277043e-05}
{"train_loss": 0.0645352154970169, "global_step": 150583, "epoch": 1691, "lr": 4.157107058417206e-05}
{"train_loss": 0.14817731082439423, "global_step": 150584, "epoch": 1691, "lr": 4.157049914520021e-05}
{"train_loss": 0.12219233065843582, "global_step": 150585, "epoch": 1691, "lr": 4.1569927707361636e-05}
{"train_loss": 0.09594747424125671, "global_step": 150586, "epoch": 1691, "lr": 4.156935627065635e-05}
{"train_loss": 0.10167754675899998, "global_step": 150587, "epoch": 1691, "lr": 4.156878483508449e-05, "val_loss": 6.495944499969482}
{"train_loss": 0.13520357012748718, "global_step": 150588, "epoch": 1692, "lr": 4.156821340064608e-05}
{"train_loss": 0.09448734670877457, "global_step": 150589, "epoch": 1692, "lr": 4.156764196734124e-05}
{"train_loss": 0.1572933942079544, "global_step": 150590, "epoch": 1692, "lr": 4.1567070535170014e-05}
{"train_loss": 0.07869873195886612, "global_step": 150591, "epoch": 1692, "lr": 4.156649910413251e-05}
{"train_loss": 0.17093481123447418, "global_step": 150592, "epoch": 1692, "lr": 4.156592767422877e-05}
{"train_loss": 0.12432758510112762, "global_step": 150593, "epoch": 1692, "lr": 4.15653562454589e-05}
{"train_loss": 0.1287248432636261, "global_step": 150594, "epoch": 1692, "lr": 4.156478481782295e-05}
{"train_loss": 0.16787712275981903, "global_step": 150595, "epoch": 1692, "lr": 4.156421339132103e-05}
{"train_loss": 0.13369420170783997, "global_step": 150596, "epoch": 1692, "lr": 4.156364196595319e-05}
{"train_loss": 0.09719722718000412, "global_step": 150597, "epoch": 1692, "lr": 4.1563070541719515e-05}
{"train_loss": 0.11183402687311172, "global_step": 150598, "epoch": 1692, "lr": 4.1562499118620084e-05}
{"train_loss": 0.12251007556915283, "global_step": 150599, "epoch": 1692, "lr": 4.156192769665496e-05}
{"train_loss": 0.15571917593479156, "global_step": 150600, "epoch": 1692, "lr": 4.1561356275824256e-05}
{"train_loss": 0.1272527128458023, "global_step": 150601, "epoch": 1692, "lr": 4.1560784856128e-05}
{"train_loss": 0.09128348529338837, "global_step": 150602, "epoch": 1692, "lr": 4.1560213437566315e-05}
{"train_loss": 0.07119925320148468, "global_step": 150603, "epoch": 1692, "lr": 4.1559642020139234e-05}
{"train_loss": 0.061878494918346405, "global_step": 150604, "epoch": 1692, "lr": 4.1559070603846886e-05}
{"train_loss": 0.1255304366350174, "global_step": 150605, "epoch": 1692, "lr": 4.155849918868929e-05}
{"train_loss": 0.12277568876743317, "global_step": 150606, "epoch": 1692, "lr": 4.155792777466657e-05}
{"train_loss": 0.08281854540109634, "global_step": 150607, "epoch": 1692, "lr": 4.1557356361778765e-05}
{"train_loss": 0.18449941277503967, "global_step": 150608, "epoch": 1692, "lr": 4.1556784950026e-05}
{"train_loss": 0.20062890648841858, "global_step": 150609, "epoch": 1692, "lr": 4.155621353940828e-05}
{"train_loss": 0.051027677953243256, "global_step": 150610, "epoch": 1692, "lr": 4.155564212992577e-05}
{"train_loss": 0.06294277310371399, "global_step": 150611, "epoch": 1692, "lr": 4.155507072157846e-05}
{"train_loss": 0.136355459690094, "global_step": 150612, "epoch": 1692, "lr": 4.15544993143665e-05}
{"train_loss": 0.10394661128520966, "global_step": 150613, "epoch": 1692, "lr": 4.155392790828991e-05}
{"train_loss": 0.06916721910238266, "global_step": 150614, "epoch": 1692, "lr": 4.1553356503348815e-05}
{"train_loss": 0.0768471360206604, "global_step": 150615, "epoch": 1692, "lr": 4.1552785099543256e-05}
{"train_loss": 0.10845248401165009, "global_step": 150616, "epoch": 1692, "lr": 4.155221369687331e-05}
{"train_loss": 0.11828602105379105, "global_step": 150617, "epoch": 1692, "lr": 4.155164229533909e-05}
{"train_loss": 0.0865917056798935, "global_step": 150618, "epoch": 1692, "lr": 4.155107089494062e-05}
{"train_loss": 0.07753477245569229, "global_step": 150619, "epoch": 1692, "lr": 4.155049949567804e-05}
{"train_loss": 0.1283852458000183, "global_step": 150620, "epoch": 1692, "lr": 4.154992809755136e-05}
{"train_loss": 0.09510467946529388, "global_step": 150621, "epoch": 1692, "lr": 4.1549356700560715e-05}
{"train_loss": 0.08379174023866653, "global_step": 150622, "epoch": 1692, "lr": 4.154878530470614e-05}
{"train_loss": 0.11089707911014557, "global_step": 150623, "epoch": 1692, "lr": 4.154821390998774e-05}
{"train_loss": 0.07411327958106995, "global_step": 150624, "epoch": 1692, "lr": 4.154764251640556e-05}
{"train_loss": 0.1344119906425476, "global_step": 150625, "epoch": 1692, "lr": 4.154707112395972e-05}
{"train_loss": 0.04475622624158859, "global_step": 150626, "epoch": 1692, "lr": 4.154649973265025e-05}
{"train_loss": 0.11189501732587814, "global_step": 150627, "epoch": 1692, "lr": 4.1545928342477275e-05}
{"train_loss": 0.11866501718759537, "global_step": 150628, "epoch": 1692, "lr": 4.154535695344082e-05}
{"train_loss": 0.11041154712438583, "global_step": 150629, "epoch": 1692, "lr": 4.154478556554102e-05}
{"train_loss": 0.11717423051595688, "global_step": 150630, "epoch": 1692, "lr": 4.154421417877789e-05}
{"train_loss": 0.12913648784160614, "global_step": 150631, "epoch": 1692, "lr": 4.154364279315156e-05}
{"train_loss": 0.04306456819176674, "global_step": 150632, "epoch": 1692, "lr": 4.154307140866206e-05}
{"train_loss": 0.13979429006576538, "global_step": 150633, "epoch": 1692, "lr": 4.1542500025309525e-05}
{"train_loss": 0.11417820304632187, "global_step": 150634, "epoch": 1692, "lr": 4.1541928643093976e-05}
{"train_loss": 0.08249296993017197, "global_step": 150635, "epoch": 1692, "lr": 4.1541357262015504e-05}
{"train_loss": 0.07578613609075546, "global_step": 150636, "epoch": 1692, "lr": 4.154078588207421e-05}
{"train_loss": 0.09451895952224731, "global_step": 150637, "epoch": 1692, "lr": 4.154021450327014e-05}
{"train_loss": 0.1000223159790039, "global_step": 150638, "epoch": 1692, "lr": 4.153964312560341e-05}
{"train_loss": 0.09546202421188354, "global_step": 150639, "epoch": 1692, "lr": 4.153907174907404e-05}
{"train_loss": 0.0951588824391365, "global_step": 150640, "epoch": 1692, "lr": 4.1538500373682166e-05}
{"train_loss": 0.05918020009994507, "global_step": 150641, "epoch": 1692, "lr": 4.153792899942781e-05}
{"train_loss": 0.1377074271440506, "global_step": 150642, "epoch": 1692, "lr": 4.1537357626311105e-05}
{"train_loss": 0.1254010945558548, "global_step": 150643, "epoch": 1692, "lr": 4.153678625433207e-05}
{"train_loss": 0.1599958837032318, "global_step": 150644, "epoch": 1692, "lr": 4.153621488349084e-05}
{"train_loss": 0.13125327229499817, "global_step": 150645, "epoch": 1692, "lr": 4.153564351378744e-05}
{"train_loss": 0.07856042683124542, "global_step": 150646, "epoch": 1692, "lr": 4.153507214522199e-05}
{"train_loss": 0.08538880944252014, "global_step": 150647, "epoch": 1692, "lr": 4.153450077779452e-05}
{"train_loss": 0.16237054765224457, "global_step": 150648, "epoch": 1692, "lr": 4.153392941150516e-05}
{"train_loss": 0.1494019776582718, "global_step": 150649, "epoch": 1692, "lr": 4.1533358046353935e-05}
{"train_loss": 0.14655236899852753, "global_step": 150650, "epoch": 1692, "lr": 4.1532786682340975e-05}
{"train_loss": 0.04704749211668968, "global_step": 150651, "epoch": 1692, "lr": 4.15322153194663e-05}
{"train_loss": 0.16180145740509033, "global_step": 150652, "epoch": 1692, "lr": 4.153164395773003e-05}
{"train_loss": 0.06574434041976929, "global_step": 150653, "epoch": 1692, "lr": 4.153107259713224e-05}
{"train_loss": 0.09348461776971817, "global_step": 150654, "epoch": 1692, "lr": 4.153050123767297e-05}
{"train_loss": 0.10337111353874207, "global_step": 150655, "epoch": 1692, "lr": 4.152992987935234e-05}
{"train_loss": 0.09927360713481903, "global_step": 150656, "epoch": 1692, "lr": 4.152935852217039e-05}
{"train_loss": 0.14001242816448212, "global_step": 150657, "epoch": 1692, "lr": 4.152878716612723e-05}
{"train_loss": 0.08785787969827652, "global_step": 150658, "epoch": 1692, "lr": 4.15282158112229e-05}
{"train_loss": 0.11184880137443542, "global_step": 150659, "epoch": 1692, "lr": 4.152764445745752e-05}
{"train_loss": 0.0752430409193039, "global_step": 150660, "epoch": 1692, "lr": 4.152707310483114e-05}
{"train_loss": 0.11083252727985382, "global_step": 150661, "epoch": 1692, "lr": 4.1526501753343845e-05}
{"train_loss": 0.21972551941871643, "global_step": 150662, "epoch": 1692, "lr": 4.1525930402995695e-05}
{"train_loss": 0.06806006282567978, "global_step": 150663, "epoch": 1692, "lr": 4.15253590537868e-05}
{"train_loss": 0.08512952923774719, "global_step": 150664, "epoch": 1692, "lr": 4.152478770571719e-05}
{"train_loss": 0.08272930234670639, "global_step": 150665, "epoch": 1692, "lr": 4.1524216358787e-05}
{"train_loss": 0.0986151248216629, "global_step": 150666, "epoch": 1692, "lr": 4.152364501299625e-05}
{"train_loss": 0.061798933893442154, "global_step": 150667, "epoch": 1692, "lr": 4.152307366834506e-05}
{"train_loss": 0.16724425554275513, "global_step": 150668, "epoch": 1692, "lr": 4.1522502324833476e-05}
{"train_loss": 0.1074535995721817, "global_step": 150669, "epoch": 1692, "lr": 4.1521930982461605e-05}
{"train_loss": 0.07293383777141571, "global_step": 150670, "epoch": 1692, "lr": 4.152135964122949e-05}
{"train_loss": 0.22406171262264252, "global_step": 150671, "epoch": 1692, "lr": 4.1520788301137226e-05}
{"train_loss": 0.14502446353435516, "global_step": 150672, "epoch": 1692, "lr": 4.1520216962184894e-05}
{"train_loss": 0.042814433574676514, "global_step": 150673, "epoch": 1692, "lr": 4.151964562437257e-05}
{"train_loss": 0.11451851576566696, "global_step": 150674, "epoch": 1692, "lr": 4.151907428770033e-05}
{"train_loss": 0.10594446212053299, "global_step": 150675, "epoch": 1692, "lr": 4.1518502952168226e-05}
{"train_loss": 0.11066869709096598, "global_step": 150676, "epoch": 1692, "lr": 4.151793161777638e-05, "val_loss": 6.370994567871094}
{"train_loss": 0.09021761268377304, "global_step": 150677, "epoch": 1693, "lr": 4.151736028452482e-05}
{"train_loss": 0.1267634928226471, "global_step": 150678, "epoch": 1693, "lr": 4.1516788952413674e-05}
{"train_loss": 0.07883673906326294, "global_step": 150679, "epoch": 1693, "lr": 4.151621762144297e-05}
{"train_loss": 0.11904636025428772, "global_step": 150680, "epoch": 1693, "lr": 4.151564629161283e-05}
{"train_loss": 0.10803643614053726, "global_step": 150681, "epoch": 1693, "lr": 4.1515074962923284e-05}
{"train_loss": 0.12698718905448914, "global_step": 150682, "epoch": 1693, "lr": 4.151450363537445e-05}
{"train_loss": 0.06979101151227951, "global_step": 150683, "epoch": 1693, "lr": 4.151393230896638e-05}
{"train_loss": 0.08462858945131302, "global_step": 150684, "epoch": 1693, "lr": 4.1513360983699164e-05}
{"train_loss": 0.08677108585834503, "global_step": 150685, "epoch": 1693, "lr": 4.151278965957286e-05}
{"train_loss": 0.08457350730895996, "global_step": 150686, "epoch": 1693, "lr": 4.151221833658757e-05}
{"train_loss": 0.16790802776813507, "global_step": 150687, "epoch": 1693, "lr": 4.1511647014743346e-05}
{"train_loss": 0.1105794906616211, "global_step": 150688, "epoch": 1693, "lr": 4.15110756940403e-05}
{"train_loss": 0.07701678574085236, "global_step": 150689, "epoch": 1693, "lr": 4.151050437447846e-05}
{"train_loss": 0.050714246928691864, "global_step": 150690, "epoch": 1693, "lr": 4.1509933056057946e-05}
{"train_loss": 0.13053320348262787, "global_step": 150691, "epoch": 1693, "lr": 4.1509361738778816e-05}
{"train_loss": 0.06859008222818375, "global_step": 150692, "epoch": 1693, "lr": 4.150879042264114e-05}
{"train_loss": 0.11425670981407166, "global_step": 150693, "epoch": 1693, "lr": 4.150821910764503e-05}
{"train_loss": 0.1261911541223526, "global_step": 150694, "epoch": 1693, "lr": 4.150764779379051e-05}
{"train_loss": 0.0965966209769249, "global_step": 150695, "epoch": 1693, "lr": 4.15070764810777e-05}
{"train_loss": 0.07412219047546387, "global_step": 150696, "epoch": 1693, "lr": 4.1506505169506634e-05}
{"train_loss": 0.10947530716657639, "global_step": 150697, "epoch": 1693, "lr": 4.1505933859077446e-05}
{"train_loss": 0.06219495087862015, "global_step": 150698, "epoch": 1693, "lr": 4.150536254979015e-05}
{"train_loss": 0.1375967115163803, "global_step": 150699, "epoch": 1693, "lr": 4.1504791241644884e-05}
{"train_loss": 0.06305935978889465, "global_step": 150700, "epoch": 1693, "lr": 4.1504219934641674e-05}
{"train_loss": 0.07697166502475739, "global_step": 150701, "epoch": 1693, "lr": 4.150364862878064e-05}
{"train_loss": 0.08465854078531265, "global_step": 150702, "epoch": 1693, "lr": 4.150307732406181e-05}
{"train_loss": 0.12629054486751556, "global_step": 150703, "epoch": 1693, "lr": 4.1502506020485317e-05}
{"train_loss": 0.09821461886167526, "global_step": 150704, "epoch": 1693, "lr": 4.1501934718051184e-05}
{"train_loss": 0.08555568009614944, "global_step": 150705, "epoch": 1693, "lr": 4.150136341675953e-05}
{"train_loss": 0.05719124525785446, "global_step": 150706, "epoch": 1693, "lr": 4.1500792116610395e-05}
{"train_loss": 0.11287711560726166, "global_step": 150707, "epoch": 1693, "lr": 4.1500220817603886e-05}
{"train_loss": 0.08978378772735596, "global_step": 150708, "epoch": 1693, "lr": 4.1499649519740066e-05}
{"train_loss": 0.11026061326265335, "global_step": 150709, "epoch": 1693, "lr": 4.1499078223019015e-05}
{"train_loss": 0.12901580333709717, "global_step": 150710, "epoch": 1693, "lr": 4.149850692744081e-05}
{"train_loss": 0.10688762366771698, "global_step": 150711, "epoch": 1693, "lr": 4.149793563300553e-05}
{"train_loss": 0.09635327756404877, "global_step": 150712, "epoch": 1693, "lr": 4.1497364339713265e-05}
{"train_loss": 0.12267212569713593, "global_step": 150713, "epoch": 1693, "lr": 4.1496793047564044e-05}
{"train_loss": 0.07455497235059738, "global_step": 150714, "epoch": 1693, "lr": 4.1496221756558e-05}
{"train_loss": 0.10508633404970169, "global_step": 150715, "epoch": 1693, "lr": 4.1495650466695164e-05}
{"train_loss": 0.11062593758106232, "global_step": 150716, "epoch": 1693, "lr": 4.1495079177975654e-05}
{"train_loss": 0.1441648006439209, "global_step": 150717, "epoch": 1693, "lr": 4.14945078903995e-05}
{"train_loss": 0.10946212708950043, "global_step": 150718, "epoch": 1693, "lr": 4.1493936603966846e-05}
{"train_loss": 0.07114379853010178, "global_step": 150719, "epoch": 1693, "lr": 4.149336531867769e-05}
{"train_loss": 0.12750841677188873, "global_step": 150720, "epoch": 1693, "lr": 4.149279403453217e-05}
{"train_loss": 0.12900091707706451, "global_step": 150721, "epoch": 1693, "lr": 4.149222275153033e-05}
{"train_loss": 0.1164446771144867, "global_step": 150722, "epoch": 1693, "lr": 4.149165146967227e-05}
{"train_loss": 0.08764015138149261, "global_step": 150723, "epoch": 1693, "lr": 4.149108018895803e-05}
{"train_loss": 0.08877111226320267, "global_step": 150724, "epoch": 1693, "lr": 4.149050890938774e-05}
{"train_loss": 0.11411456018686295, "global_step": 150725, "epoch": 1693, "lr": 4.148993763096142e-05}
{"train_loss": 0.09534865617752075, "global_step": 150726, "epoch": 1693, "lr": 4.1489366353679186e-05}
{"train_loss": 0.13124150037765503, "global_step": 150727, "epoch": 1693, "lr": 4.1488795077541104e-05}
{"train_loss": 0.08363469690084457, "global_step": 150728, "epoch": 1693, "lr": 4.148822380254725e-05}
{"train_loss": 0.10526799410581589, "global_step": 150729, "epoch": 1693, "lr": 4.14876525286977e-05}
{"train_loss": 0.11600383371114731, "global_step": 150730, "epoch": 1693, "lr": 4.148708125599253e-05}
{"train_loss": 0.06496953219175339, "global_step": 150731, "epoch": 1693, "lr": 4.148650998443183e-05}
{"train_loss": 0.14167426526546478, "global_step": 150732, "epoch": 1693, "lr": 4.1485938714015645e-05}
{"train_loss": 0.09881467372179031, "global_step": 150733, "epoch": 1693, "lr": 4.1485367444744086e-05}
{"train_loss": 0.08516480773687363, "global_step": 150734, "epoch": 1693, "lr": 4.14847961766172e-05}
{"train_loss": 0.09346802532672882, "global_step": 150735, "epoch": 1693, "lr": 4.14842249096351e-05}
{"train_loss": 0.1027563288807869, "global_step": 150736, "epoch": 1693, "lr": 4.148365364379783e-05}
{"train_loss": 0.1341743916273117, "global_step": 150737, "epoch": 1693, "lr": 4.1483082379105486e-05}
{"train_loss": 0.13612475991249084, "global_step": 150738, "epoch": 1693, "lr": 4.148251111555812e-05}
{"train_loss": 0.08456028997898102, "global_step": 150739, "epoch": 1693, "lr": 4.148193985315586e-05}
{"train_loss": 0.08855441212654114, "global_step": 150740, "epoch": 1693, "lr": 4.1481368591898714e-05}
{"train_loss": 0.11843442916870117, "global_step": 150741, "epoch": 1693, "lr": 4.148079733178682e-05}
{"train_loss": 0.12171869724988937, "global_step": 150742, "epoch": 1693, "lr": 4.14802260728202e-05}
{"train_loss": 0.1540033370256424, "global_step": 150743, "epoch": 1693, "lr": 4.147965481499897e-05}
{"train_loss": 0.10413241386413574, "global_step": 150744, "epoch": 1693, "lr": 4.147908355832322e-05}
{"train_loss": 0.15833832323551178, "global_step": 150745, "epoch": 1693, "lr": 4.147851230279298e-05}
{"train_loss": 0.12435741722583771, "global_step": 150746, "epoch": 1693, "lr": 4.1477941048408363e-05}
{"train_loss": 0.08740121126174927, "global_step": 150747, "epoch": 1693, "lr": 4.147736979516943e-05}
{"train_loss": 0.11653497815132141, "global_step": 150748, "epoch": 1693, "lr": 4.1476798543076255e-05}
{"train_loss": 0.15334823727607727, "global_step": 150749, "epoch": 1693, "lr": 4.147622729212892e-05}
{"train_loss": 0.09634630382061005, "global_step": 150750, "epoch": 1693, "lr": 4.1475656042327523e-05}
{"train_loss": 0.09907934814691544, "global_step": 150751, "epoch": 1693, "lr": 4.147508479367209e-05}
{"train_loss": 0.07701359689235687, "global_step": 150752, "epoch": 1693, "lr": 4.147451354616275e-05}
{"train_loss": 0.05465045943856239, "global_step": 150753, "epoch": 1693, "lr": 4.147394229979954e-05}
{"train_loss": 0.13163331151008606, "global_step": 150754, "epoch": 1693, "lr": 4.1473371054582575e-05}
{"train_loss": 0.12162812054157257, "global_step": 150755, "epoch": 1693, "lr": 4.147279981051189e-05}
{"train_loss": 0.08943402767181396, "global_step": 150756, "epoch": 1693, "lr": 4.14722285675876e-05}
{"train_loss": 0.12326621264219284, "global_step": 150757, "epoch": 1693, "lr": 4.147165732580975e-05}
{"train_loss": 0.08750830590724945, "global_step": 150758, "epoch": 1693, "lr": 4.147108608517845e-05}
{"train_loss": 0.06256304681301117, "global_step": 150759, "epoch": 1693, "lr": 4.147051484569374e-05}
{"train_loss": 0.07130488008260727, "global_step": 150760, "epoch": 1693, "lr": 4.146994360735573e-05}
{"train_loss": 0.10315433889627457, "global_step": 150761, "epoch": 1693, "lr": 4.1469372370164474e-05}
{"train_loss": 0.11466804146766663, "global_step": 150762, "epoch": 1693, "lr": 4.1468801134120045e-05}
{"train_loss": 0.10733035206794739, "global_step": 150763, "epoch": 1693, "lr": 4.1468229899222553e-05}
{"train_loss": 0.16151732206344604, "global_step": 150764, "epoch": 1693, "lr": 4.1467658665472046e-05}
{"train_loss": 0.10414380668086952, "global_step": 150765, "epoch": 1693, "lr": 4.146708743286861e-05, "val_loss": 6.337235450744629}
{"train_loss": 0.17869488894939423, "global_step": 150766, "epoch": 1694, "lr": 4.146651620141232e-05}
{"train_loss": 0.11139875650405884, "global_step": 150767, "epoch": 1694, "lr": 4.146594497110325e-05}
{"train_loss": 0.10599085688591003, "global_step": 150768, "epoch": 1694, "lr": 4.1465373741941485e-05}
{"train_loss": 0.1335260570049286, "global_step": 150769, "epoch": 1694, "lr": 4.1464802513927085e-05}
{"train_loss": 0.1472560167312622, "global_step": 150770, "epoch": 1694, "lr": 4.146423128706015e-05}
{"train_loss": 0.1307590752840042, "global_step": 150771, "epoch": 1694, "lr": 4.146366006134075e-05}
{"train_loss": 0.0931071788072586, "global_step": 150772, "epoch": 1694, "lr": 4.1463088836768934e-05}
{"train_loss": 0.12221512198448181, "global_step": 150773, "epoch": 1694, "lr": 4.146251761334483e-05}
{"train_loss": 0.09367747604846954, "global_step": 150774, "epoch": 1694, "lr": 4.146194639106846e-05}
{"train_loss": 0.09113846719264984, "global_step": 150775, "epoch": 1694, "lr": 4.146137516993995e-05}
{"train_loss": 0.1256117969751358, "global_step": 150776, "epoch": 1694, "lr": 4.146080394995933e-05}
{"train_loss": 0.14179806411266327, "global_step": 150777, "epoch": 1694, "lr": 4.146023273112672e-05}
{"train_loss": 0.11893106251955032, "global_step": 150778, "epoch": 1694, "lr": 4.1459661513442166e-05}
{"train_loss": 0.12163344025611877, "global_step": 150779, "epoch": 1694, "lr": 4.145909029690575e-05}
{"train_loss": 0.14640174806118011, "global_step": 150780, "epoch": 1694, "lr": 4.145851908151758e-05}
{"train_loss": 0.12382762879133224, "global_step": 150781, "epoch": 1694, "lr": 4.145794786727769e-05}
{"train_loss": 0.12383746355772018, "global_step": 150782, "epoch": 1694, "lr": 4.1457376654186186e-05}
{"train_loss": 0.12365205585956573, "global_step": 150783, "epoch": 1694, "lr": 4.145680544224312e-05}
{"train_loss": 0.08611718565225601, "global_step": 150784, "epoch": 1694, "lr": 4.14562342314486e-05}
{"train_loss": 0.08098512142896652, "global_step": 150785, "epoch": 1694, "lr": 4.145566302180267e-05}
{"train_loss": 0.1795438677072525, "global_step": 150786, "epoch": 1694, "lr": 4.145509181330543e-05}
{"train_loss": 0.06571776419878006, "global_step": 150787, "epoch": 1694, "lr": 4.1454520605956944e-05}
{"train_loss": 0.08481310307979584, "global_step": 150788, "epoch": 1694, "lr": 4.1453949399757295e-05}
{"train_loss": 0.09085065126419067, "global_step": 150789, "epoch": 1694, "lr": 4.145337819470656e-05}
{"train_loss": 0.1223241463303566, "global_step": 150790, "epoch": 1694, "lr": 4.1452806990804824e-05}
{"train_loss": 0.06534869968891144, "global_step": 150791, "epoch": 1694, "lr": 4.145223578805213e-05}
{"train_loss": 0.044692110270261765, "global_step": 150792, "epoch": 1694, "lr": 4.1451664586448604e-05}
{"train_loss": 0.10568400472402573, "global_step": 150793, "epoch": 1694, "lr": 4.145109338599427e-05}
{"train_loss": 0.09700421243906021, "global_step": 150794, "epoch": 1694, "lr": 4.145052218668926e-05}
{"train_loss": 0.06281401216983795, "global_step": 150795, "epoch": 1694, "lr": 4.14499509885336e-05}
{"train_loss": 0.1264563500881195, "global_step": 150796, "epoch": 1694, "lr": 4.144937979152741e-05}
{"train_loss": 0.08149059116840363, "global_step": 150797, "epoch": 1694, "lr": 4.144880859567073e-05}
{"train_loss": 0.0403425358235836, "global_step": 150798, "epoch": 1694, "lr": 4.144823740096365e-05}
{"train_loss": 0.10958400368690491, "global_step": 150799, "epoch": 1694, "lr": 4.144766620740626e-05}
{"train_loss": 0.10092797875404358, "global_step": 150800, "epoch": 1694, "lr": 4.144709501499863e-05}
{"train_loss": 0.11715315282344818, "global_step": 150801, "epoch": 1694, "lr": 4.1446523823740835e-05}
{"train_loss": 0.09135296195745468, "global_step": 150802, "epoch": 1694, "lr": 4.1445952633632935e-05}
{"train_loss": 0.0669570341706276, "global_step": 150803, "epoch": 1694, "lr": 4.144538144467505e-05}
{"train_loss": 0.07952363044023514, "global_step": 150804, "epoch": 1694, "lr": 4.1444810256867214e-05}
{"train_loss": 0.15013746917247772, "global_step": 150805, "epoch": 1694, "lr": 4.1444239070209516e-05}
{"train_loss": 0.1226830706000328, "global_step": 150806, "epoch": 1694, "lr": 4.1443667884702035e-05}
{"train_loss": 0.13599973917007446, "global_step": 150807, "epoch": 1694, "lr": 4.144309670034485e-05}
{"train_loss": 0.0804208517074585, "global_step": 150808, "epoch": 1694, "lr": 4.144252551713804e-05}
{"train_loss": 0.14689631760120392, "global_step": 150809, "epoch": 1694, "lr": 4.144195433508169e-05}
{"train_loss": 0.11233274638652802, "global_step": 150810, "epoch": 1694, "lr": 4.144138315417584e-05}
{"train_loss": 0.1017260029911995, "global_step": 150811, "epoch": 1694, "lr": 4.144081197442061e-05}
{"train_loss": 0.09410165250301361, "global_step": 150812, "epoch": 1694, "lr": 4.144024079581604e-05}
{"train_loss": 0.09957870841026306, "global_step": 150813, "epoch": 1694, "lr": 4.143966961836225e-05}
{"train_loss": 0.1052912026643753, "global_step": 150814, "epoch": 1694, "lr": 4.1439098442059274e-05}
{"train_loss": 0.08605510741472244, "global_step": 150815, "epoch": 1694, "lr": 4.1438527266907224e-05}
{"train_loss": 0.07022544741630554, "global_step": 150816, "epoch": 1694, "lr": 4.143795609290613e-05}
{"train_loss": 0.08559072762727737, "global_step": 150817, "epoch": 1694, "lr": 4.143738492005611e-05}
{"train_loss": 0.1259545385837555, "global_step": 150818, "epoch": 1694, "lr": 4.1436813748357246e-05}
{"train_loss": 0.07323130965232849, "global_step": 150819, "epoch": 1694, "lr": 4.143624257780957e-05}
{"train_loss": 0.14396685361862183, "global_step": 150820, "epoch": 1694, "lr": 4.143567140841323e-05}
{"train_loss": 0.10127826035022736, "global_step": 150821, "epoch": 1694, "lr": 4.143510024016821e-05}
{"train_loss": 0.0640367716550827, "global_step": 150822, "epoch": 1694, "lr": 4.1434529073074675e-05}
{"train_loss": 0.10465416312217712, "global_step": 150823, "epoch": 1694, "lr": 4.1433957907132644e-05}
{"train_loss": 0.14992831647396088, "global_step": 150824, "epoch": 1694, "lr": 4.143338674234222e-05}
{"train_loss": 0.10105690360069275, "global_step": 150825, "epoch": 1694, "lr": 4.143281557870347e-05}
{"train_loss": 0.149211123585701, "global_step": 150826, "epoch": 1694, "lr": 4.143224441621647e-05}
{"train_loss": 0.20706620812416077, "global_step": 150827, "epoch": 1694, "lr": 4.143167325488131e-05}
{"train_loss": 0.08372195065021515, "global_step": 150828, "epoch": 1694, "lr": 4.143110209469806e-05}
{"train_loss": 0.09983651340007782, "global_step": 150829, "epoch": 1694, "lr": 4.143053093566677e-05}
{"train_loss": 0.19358691573143005, "global_step": 150830, "epoch": 1694, "lr": 4.1429959777787566e-05}
{"train_loss": 0.09647145867347717, "global_step": 150831, "epoch": 1694, "lr": 4.142938862106048e-05}
{"train_loss": 0.12842673063278198, "global_step": 150832, "epoch": 1694, "lr": 4.1428817465485633e-05}
{"train_loss": 0.17372120916843414, "global_step": 150833, "epoch": 1694, "lr": 4.142824631106305e-05}
{"train_loss": 0.12864333391189575, "global_step": 150834, "epoch": 1694, "lr": 4.1427675157792834e-05}
{"train_loss": 0.15963850915431976, "global_step": 150835, "epoch": 1694, "lr": 4.142710400567508e-05}
{"train_loss": 0.15002216398715973, "global_step": 150836, "epoch": 1694, "lr": 4.142653285470983e-05}
{"train_loss": 0.10889212042093277, "global_step": 150837, "epoch": 1694, "lr": 4.1425961704897206e-05}
{"train_loss": 0.17562782764434814, "global_step": 150838, "epoch": 1694, "lr": 4.142539055623723e-05}
{"train_loss": 0.04250464588403702, "global_step": 150839, "epoch": 1694, "lr": 4.1424819408730025e-05}
{"train_loss": 0.09674068540334702, "global_step": 150840, "epoch": 1694, "lr": 4.142424826237563e-05}
{"train_loss": 0.08253682404756546, "global_step": 150841, "epoch": 1694, "lr": 4.1423677117174165e-05}
{"train_loss": 0.17922157049179077, "global_step": 150842, "epoch": 1694, "lr": 4.142310597312567e-05}
{"train_loss": 0.12908440828323364, "global_step": 150843, "epoch": 1694, "lr": 4.1422534830230224e-05}
{"train_loss": 0.07047007977962494, "global_step": 150844, "epoch": 1694, "lr": 4.1421963688487925e-05}
{"train_loss": 0.09999725222587585, "global_step": 150845, "epoch": 1694, "lr": 4.142139254789883e-05}
{"train_loss": 0.09779084473848343, "global_step": 150846, "epoch": 1694, "lr": 4.142082140846303e-05}
{"train_loss": 0.08938372880220413, "global_step": 150847, "epoch": 1694, "lr": 4.1420250270180603e-05}
{"train_loss": 0.08247391879558563, "global_step": 150848, "epoch": 1694, "lr": 4.1419679133051605e-05}
{"train_loss": 0.11487900465726852, "global_step": 150849, "epoch": 1694, "lr": 4.141910799707614e-05}
{"train_loss": 0.14509116113185883, "global_step": 150850, "epoch": 1694, "lr": 4.141853686225425e-05}
{"train_loss": 0.13659559190273285, "global_step": 150851, "epoch": 1694, "lr": 4.141796572858606e-05}
{"train_loss": 0.048743873834609985, "global_step": 150852, "epoch": 1694, "lr": 4.14173945960716e-05}
{"train_loss": 0.12317502498626709, "global_step": 150853, "epoch": 1694, "lr": 4.1416823464710955e-05}
{"train_loss": 0.11170303336020267, "global_step": 150854, "epoch": 1694, "lr": 4.141625233450425e-05, "val_loss": 6.324124813079834}
{"train_loss": 0.11243336647748947, "global_step": 150855, "epoch": 1695, "lr": 4.141568120545149e-05}
{"train_loss": 0.14509828388690948, "global_step": 150856, "epoch": 1695, "lr": 4.1415110077552815e-05}
{"train_loss": 0.09276145696640015, "global_step": 150857, "epoch": 1695, "lr": 4.141453895080826e-05}
{"train_loss": 0.17446550726890564, "global_step": 150858, "epoch": 1695, "lr": 4.141396782521792e-05}
{"train_loss": 0.11652069538831711, "global_step": 150859, "epoch": 1695, "lr": 4.141339670078186e-05}
{"train_loss": 0.15699218213558197, "global_step": 150860, "epoch": 1695, "lr": 4.141282557750018e-05}
{"train_loss": 0.06090865284204483, "global_step": 150861, "epoch": 1695, "lr": 4.1412254455372926e-05}
{"train_loss": 0.09200798720121384, "global_step": 150862, "epoch": 1695, "lr": 4.14116833344002e-05}
{"train_loss": 0.09001322835683823, "global_step": 150863, "epoch": 1695, "lr": 4.141111221458206e-05}
{"train_loss": 0.15823295712471008, "global_step": 150864, "epoch": 1695, "lr": 4.14105410959186e-05}
{"train_loss": 0.14049389958381653, "global_step": 150865, "epoch": 1695, "lr": 4.140996997840988e-05}
{"train_loss": 0.1215476468205452, "global_step": 150866, "epoch": 1695, "lr": 4.1409398862055996e-05}
{"train_loss": 0.13019204139709473, "global_step": 150867, "epoch": 1695, "lr": 4.140882774685699e-05}
{"train_loss": 0.11717987060546875, "global_step": 150868, "epoch": 1695, "lr": 4.140825663281299e-05}
{"train_loss": 0.08568458259105682, "global_step": 150869, "epoch": 1695, "lr": 4.140768551992402e-05}
{"train_loss": 0.12026753276586533, "global_step": 150870, "epoch": 1695, "lr": 4.140711440819018e-05}
{"train_loss": 0.08074050396680832, "global_step": 150871, "epoch": 1695, "lr": 4.140654329761158e-05}
{"train_loss": 0.11780105531215668, "global_step": 150872, "epoch": 1695, "lr": 4.140597218818824e-05}
{"train_loss": 0.18040074408054352, "global_step": 150873, "epoch": 1695, "lr": 4.140540107992028e-05}
{"train_loss": 0.07657977193593979, "global_step": 150874, "epoch": 1695, "lr": 4.140482997280774e-05}
{"train_loss": 0.1042289137840271, "global_step": 150875, "epoch": 1695, "lr": 4.140425886685074e-05}
{"train_loss": 0.05654323473572731, "global_step": 150876, "epoch": 1695, "lr": 4.14036877620493e-05}
{"train_loss": 0.10785095393657684, "global_step": 150877, "epoch": 1695, "lr": 4.140311665840356e-05}
{"train_loss": 0.15698397159576416, "global_step": 150878, "epoch": 1695, "lr": 4.140254555591355e-05}
{"train_loss": 0.13054385781288147, "global_step": 150879, "epoch": 1695, "lr": 4.140197445457937e-05}
{"train_loss": 0.11886599659919739, "global_step": 150880, "epoch": 1695, "lr": 4.1401403354401094e-05}
{"train_loss": 0.15250760316848755, "global_step": 150881, "epoch": 1695, "lr": 4.1400832255378786e-05}
{"train_loss": 0.11964750289916992, "global_step": 150882, "epoch": 1695, "lr": 4.140026115751253e-05}
{"train_loss": 0.06968970596790314, "global_step": 150883, "epoch": 1695, "lr": 4.139969006080241e-05}
{"train_loss": 0.09812307357788086, "global_step": 150884, "epoch": 1695, "lr": 4.139911896524849e-05}
{"train_loss": 0.11304949969053268, "global_step": 150885, "epoch": 1695, "lr": 4.139854787085086e-05}
{"train_loss": 0.12720635533332825, "global_step": 150886, "epoch": 1695, "lr": 4.1397976777609585e-05}
{"train_loss": 0.06050172448158264, "global_step": 150887, "epoch": 1695, "lr": 4.139740568552476e-05}
{"train_loss": 0.0855005607008934, "global_step": 150888, "epoch": 1695, "lr": 4.139683459459643e-05}
{"train_loss": 0.0896637961268425, "global_step": 150889, "epoch": 1695, "lr": 4.1396263504824685e-05}
{"train_loss": 0.14663179218769073, "global_step": 150890, "epoch": 1695, "lr": 4.1395692416209634e-05}
{"train_loss": 0.10418210178613663, "global_step": 150891, "epoch": 1695, "lr": 4.1395121328751305e-05}
{"train_loss": 0.08030012249946594, "global_step": 150892, "epoch": 1695, "lr": 4.139455024244981e-05}
{"train_loss": 0.17503902316093445, "global_step": 150893, "epoch": 1695, "lr": 4.1393979157305206e-05}
{"train_loss": 0.10435500741004944, "global_step": 150894, "epoch": 1695, "lr": 4.139340807331759e-05}
{"train_loss": 0.1348198801279068, "global_step": 150895, "epoch": 1695, "lr": 4.1392836990487e-05}
{"train_loss": 0.1381935179233551, "global_step": 150896, "epoch": 1695, "lr": 4.139226590881357e-05}
{"train_loss": 0.09898088872432709, "global_step": 150897, "epoch": 1695, "lr": 4.139169482829732e-05}
{"train_loss": 0.09961369633674622, "global_step": 150898, "epoch": 1695, "lr": 4.1391123748938374e-05}
{"train_loss": 0.21769702434539795, "global_step": 150899, "epoch": 1695, "lr": 4.139055267073676e-05}
{"train_loss": 0.1052764430642128, "global_step": 150900, "epoch": 1695, "lr": 4.1389981593692606e-05}
{"train_loss": 0.1107424646615982, "global_step": 150901, "epoch": 1695, "lr": 4.138941051780595e-05}
{"train_loss": 0.10071440041065216, "global_step": 150902, "epoch": 1695, "lr": 4.138883944307689e-05}
{"train_loss": 0.05499684065580368, "global_step": 150903, "epoch": 1695, "lr": 4.138826836950549e-05}
{"train_loss": 0.11652062833309174, "global_step": 150904, "epoch": 1695, "lr": 4.1387697297091834e-05}
{"train_loss": 0.11886761337518692, "global_step": 150905, "epoch": 1695, "lr": 4.138712622583599e-05}
{"train_loss": 0.07524023950099945, "global_step": 150906, "epoch": 1695, "lr": 4.1386555155738044e-05}
{"train_loss": 0.1269887089729309, "global_step": 150907, "epoch": 1695, "lr": 4.1385984086798085e-05}
{"train_loss": 0.11115767061710358, "global_step": 150908, "epoch": 1695, "lr": 4.138541301901615e-05}
{"train_loss": 0.0792122408747673, "global_step": 150909, "epoch": 1695, "lr": 4.138484195239237e-05}
{"train_loss": 0.12672317028045654, "global_step": 150910, "epoch": 1695, "lr": 4.138427088692677e-05}
{"train_loss": 0.06180940195918083, "global_step": 150911, "epoch": 1695, "lr": 4.138369982261947e-05}
{"train_loss": 0.133549764752388, "global_step": 150912, "epoch": 1695, "lr": 4.1383128759470507e-05}
{"train_loss": 0.07720673829317093, "global_step": 150913, "epoch": 1695, "lr": 4.138255769748e-05}
{"train_loss": 0.12758880853652954, "global_step": 150914, "epoch": 1695, "lr": 4.1381986636647974e-05}
{"train_loss": 0.10062795877456665, "global_step": 150915, "epoch": 1695, "lr": 4.138141557697456e-05}
{"train_loss": 0.12077479064464569, "global_step": 150916, "epoch": 1695, "lr": 4.1380844518459796e-05}
{"train_loss": 0.1377459019422531, "global_step": 150917, "epoch": 1695, "lr": 4.138027346110378e-05}
{"train_loss": 0.10896475613117218, "global_step": 150918, "epoch": 1695, "lr": 4.137970240490657e-05}
{"train_loss": 0.054821401834487915, "global_step": 150919, "epoch": 1695, "lr": 4.137913134986827e-05}
{"train_loss": 0.10018868744373322, "global_step": 150920, "epoch": 1695, "lr": 4.1378560295988936e-05}
{"train_loss": 0.14016856253147125, "global_step": 150921, "epoch": 1695, "lr": 4.137798924326865e-05}
{"train_loss": 0.04586112126708031, "global_step": 150922, "epoch": 1695, "lr": 4.137741819170747e-05}
{"train_loss": 0.12259030342102051, "global_step": 150923, "epoch": 1695, "lr": 4.137684714130551e-05}
{"train_loss": 0.13225015997886658, "global_step": 150924, "epoch": 1695, "lr": 4.1376276092062825e-05}
{"train_loss": 0.1433895081281662, "global_step": 150925, "epoch": 1695, "lr": 4.137570504397948e-05}
{"train_loss": 0.0789509266614914, "global_step": 150926, "epoch": 1695, "lr": 4.1375133997055594e-05}
{"train_loss": 0.08570714294910431, "global_step": 150927, "epoch": 1695, "lr": 4.137456295129119e-05}
{"train_loss": 0.0665968731045723, "global_step": 150928, "epoch": 1695, "lr": 4.137399190668639e-05}
{"train_loss": 0.08875838667154312, "global_step": 150929, "epoch": 1695, "lr": 4.137342086324123e-05}
{"train_loss": 0.14474964141845703, "global_step": 150930, "epoch": 1695, "lr": 4.137284982095583e-05}
{"train_loss": 0.13487915694713593, "global_step": 150931, "epoch": 1695, "lr": 4.137227877983023e-05}
{"train_loss": 0.07491431385278702, "global_step": 150932, "epoch": 1695, "lr": 4.137170773986454e-05}
{"train_loss": 0.07988092303276062, "global_step": 150933, "epoch": 1695, "lr": 4.13711367010588e-05}
{"train_loss": 0.13237932324409485, "global_step": 150934, "epoch": 1695, "lr": 4.137056566341312e-05}
{"train_loss": 0.09796464443206787, "global_step": 150935, "epoch": 1695, "lr": 4.136999462692754e-05}
{"train_loss": 0.0902719497680664, "global_step": 150936, "epoch": 1695, "lr": 4.1369423591602193e-05}
{"train_loss": 0.14345383644104004, "global_step": 150937, "epoch": 1695, "lr": 4.136885255743709e-05}
{"train_loss": 0.09470503777265549, "global_step": 150938, "epoch": 1695, "lr": 4.136828152443236e-05}
{"train_loss": 0.11479716002941132, "global_step": 150939, "epoch": 1695, "lr": 4.1367710492588054e-05}
{"train_loss": 0.12914340198040009, "global_step": 150940, "epoch": 1695, "lr": 4.136713946190425e-05}
{"train_loss": 0.08046629279851913, "global_step": 150941, "epoch": 1695, "lr": 4.1366568432381036e-05}
{"train_loss": 0.12130092829465866, "global_step": 150942, "epoch": 1695, "lr": 4.1365997404018475e-05}
{"train_loss": 0.11124711019102107, "global_step": 150943, "epoch": 1695, "lr": 4.1365426376816654e-05, "val_loss": 6.280616283416748, "train_action_mse_error": 30.90293312072754}
{"train_loss": 0.11328324675559998, "global_step": 150944, "epoch": 1696, "lr": 4.136485535077563e-05}
{"train_loss": 0.10124928504228592, "global_step": 150945, "epoch": 1696, "lr": 4.136428432589552e-05}
{"train_loss": 0.15608680248260498, "global_step": 150946, "epoch": 1696, "lr": 4.136371330217636e-05}
{"train_loss": 0.034334696829319, "global_step": 150947, "epoch": 1696, "lr": 4.136314227961826e-05}
{"train_loss": 0.14418239891529083, "global_step": 150948, "epoch": 1696, "lr": 4.1362571258221264e-05}
{"train_loss": 0.1133795902132988, "global_step": 150949, "epoch": 1696, "lr": 4.1362000237985475e-05}
{"train_loss": 0.14631575345993042, "global_step": 150950, "epoch": 1696, "lr": 4.1361429218910955e-05}
{"train_loss": 0.10156922042369843, "global_step": 150951, "epoch": 1696, "lr": 4.136085820099779e-05}
{"train_loss": 0.11050531268119812, "global_step": 150952, "epoch": 1696, "lr": 4.136028718424604e-05}
{"train_loss": 0.05466749519109726, "global_step": 150953, "epoch": 1696, "lr": 4.1359716168655814e-05}
{"train_loss": 0.12178928405046463, "global_step": 150954, "epoch": 1696, "lr": 4.135914515422715e-05}
{"train_loss": 0.18822108209133148, "global_step": 150955, "epoch": 1696, "lr": 4.135857414096016e-05}
{"train_loss": 0.06794260442256927, "global_step": 150956, "epoch": 1696, "lr": 4.135800312885489e-05}
{"train_loss": 0.10524462908506393, "global_step": 150957, "epoch": 1696, "lr": 4.1357432117911444e-05}
{"train_loss": 0.0656338632106781, "global_step": 150958, "epoch": 1696, "lr": 4.135686110812988e-05}
{"train_loss": 0.11185469478368759, "global_step": 150959, "epoch": 1696, "lr": 4.135629009951028e-05}
{"train_loss": 0.09323761612176895, "global_step": 150960, "epoch": 1696, "lr": 4.135571909205272e-05}
{"train_loss": 0.14641186594963074, "global_step": 150961, "epoch": 1696, "lr": 4.135514808575727e-05}
{"train_loss": 0.05975142493844032, "global_step": 150962, "epoch": 1696, "lr": 4.1354577080624034e-05}
{"train_loss": 0.08949586004018784, "global_step": 150963, "epoch": 1696, "lr": 4.1354006076653053e-05}
{"train_loss": 0.07126087695360184, "global_step": 150964, "epoch": 1696, "lr": 4.135343507384444e-05}
{"train_loss": 0.12834158539772034, "global_step": 150965, "epoch": 1696, "lr": 4.1352864072198225e-05}
{"train_loss": 0.12349892407655716, "global_step": 150966, "epoch": 1696, "lr": 4.135229307171454e-05}
{"train_loss": 0.1382116824388504, "global_step": 150967, "epoch": 1696, "lr": 4.135172207239341e-05}
{"train_loss": 0.09478452056646347, "global_step": 150968, "epoch": 1696, "lr": 4.135115107423496e-05}
{"train_loss": 0.07484066486358643, "global_step": 150969, "epoch": 1696, "lr": 4.1350580077239215e-05}
{"train_loss": 0.14642111957073212, "global_step": 150970, "epoch": 1696, "lr": 4.1350009081406304e-05}
{"train_loss": 0.131363645195961, "global_step": 150971, "epoch": 1696, "lr": 4.134943808673626e-05}
{"train_loss": 0.13023212552070618, "global_step": 150972, "epoch": 1696, "lr": 4.134886709322919e-05}
{"train_loss": 0.132609561085701, "global_step": 150973, "epoch": 1696, "lr": 4.1348296100885155e-05}
{"train_loss": 0.06197446957230568, "global_step": 150974, "epoch": 1696, "lr": 4.1347725109704244e-05}
{"train_loss": 0.11216211318969727, "global_step": 150975, "epoch": 1696, "lr": 4.134715411968651e-05}
{"train_loss": 0.06021776422858238, "global_step": 150976, "epoch": 1696, "lr": 4.134658313083206e-05}
{"train_loss": 0.06761926412582397, "global_step": 150977, "epoch": 1696, "lr": 4.134601214314096e-05}
{"train_loss": 0.07594204694032669, "global_step": 150978, "epoch": 1696, "lr": 4.1345441156613266e-05}
{"train_loss": 0.08450585603713989, "global_step": 150979, "epoch": 1696, "lr": 4.1344870171249085e-05}
{"train_loss": 0.1306336224079132, "global_step": 150980, "epoch": 1696, "lr": 4.1344299187048463e-05}
{"train_loss": 0.09703534841537476, "global_step": 150981, "epoch": 1696, "lr": 4.1343728204011516e-05}
{"train_loss": 0.12583236396312714, "global_step": 150982, "epoch": 1696, "lr": 4.134315722213828e-05}
{"train_loss": 0.11920341849327087, "global_step": 150983, "epoch": 1696, "lr": 4.1342586241428874e-05}
{"train_loss": 0.09126723557710648, "global_step": 150984, "epoch": 1696, "lr": 4.134201526188333e-05}
{"train_loss": 0.09075301885604858, "global_step": 150985, "epoch": 1696, "lr": 4.134144428350176e-05}
{"train_loss": 0.08375926315784454, "global_step": 150986, "epoch": 1696, "lr": 4.1340873306284216e-05}
{"train_loss": 0.12611708045005798, "global_step": 150987, "epoch": 1696, "lr": 4.1340302330230804e-05}
{"train_loss": 0.12217165529727936, "global_step": 150988, "epoch": 1696, "lr": 4.133973135534156e-05}
{"train_loss": 0.13078108429908752, "global_step": 150989, "epoch": 1696, "lr": 4.133916038161661e-05}
{"train_loss": 0.08455528318881989, "global_step": 150990, "epoch": 1696, "lr": 4.1338589409055975e-05}
{"train_loss": 0.1296941339969635, "global_step": 150991, "epoch": 1696, "lr": 4.1338018437659795e-05}
{"train_loss": 0.08605728298425674, "global_step": 150992, "epoch": 1696, "lr": 4.133744746742808e-05}
{"train_loss": 0.06803351640701294, "global_step": 150993, "epoch": 1696, "lr": 4.1336876498360964e-05}
{"train_loss": 0.088701993227005, "global_step": 150994, "epoch": 1696, "lr": 4.1336305530458476e-05}
{"train_loss": 0.07437904179096222, "global_step": 150995, "epoch": 1696, "lr": 4.133573456372074e-05}
{"train_loss": 0.1052422747015953, "global_step": 150996, "epoch": 1696, "lr": 4.1335163598147795e-05}
{"train_loss": 0.047135937958955765, "global_step": 150997, "epoch": 1696, "lr": 4.1334592633739725e-05}
{"train_loss": 0.06273388117551804, "global_step": 150998, "epoch": 1696, "lr": 4.1334021670496636e-05}
{"train_loss": 0.12864787876605988, "global_step": 150999, "epoch": 1696, "lr": 4.1333450708418556e-05}
{"train_loss": 0.08094228804111481, "global_step": 151000, "epoch": 1696, "lr": 4.1332879747505615e-05}
{"train_loss": 0.15206368267536163, "global_step": 151001, "epoch": 1696, "lr": 4.133230878775783e-05}
{"train_loss": 0.15872621536254883, "global_step": 151002, "epoch": 1696, "lr": 4.133173782917534e-05}
{"train_loss": 0.09585801512002945, "global_step": 151003, "epoch": 1696, "lr": 4.133116687175817e-05}
{"train_loss": 0.06916974484920502, "global_step": 151004, "epoch": 1696, "lr": 4.133059591550645e-05}
{"train_loss": 0.14637860655784607, "global_step": 151005, "epoch": 1696, "lr": 4.133002496042019e-05}
{"train_loss": 0.0817490890622139, "global_step": 151006, "epoch": 1696, "lr": 4.132945400649953e-05}
{"train_loss": 0.10248591750860214, "global_step": 151007, "epoch": 1696, "lr": 4.13288830537445e-05}
{"train_loss": 0.12720030546188354, "global_step": 151008, "epoch": 1696, "lr": 4.132831210215522e-05}
{"train_loss": 0.07652126997709274, "global_step": 151009, "epoch": 1696, "lr": 4.132774115173171e-05}
{"train_loss": 0.07355686277151108, "global_step": 151010, "epoch": 1696, "lr": 4.132717020247411e-05}
{"train_loss": 0.04129403457045555, "global_step": 151011, "epoch": 1696, "lr": 4.1326599254382435e-05}
{"train_loss": 0.08852678537368774, "global_step": 151012, "epoch": 1696, "lr": 4.132602830745682e-05}
{"train_loss": 0.12184854596853256, "global_step": 151013, "epoch": 1696, "lr": 4.13254573616973e-05}
{"train_loss": 0.1321270763874054, "global_step": 151014, "epoch": 1696, "lr": 4.132488641710398e-05}
{"train_loss": 0.05886421352624893, "global_step": 151015, "epoch": 1696, "lr": 4.13243154736769e-05}
{"train_loss": 0.1092320904135704, "global_step": 151016, "epoch": 1696, "lr": 4.132374453141618e-05}
{"train_loss": 0.1647021472454071, "global_step": 151017, "epoch": 1696, "lr": 4.1323173590321876e-05}
{"train_loss": 0.1473657786846161, "global_step": 151018, "epoch": 1696, "lr": 4.1322602650394054e-05}
{"train_loss": 0.12382553517818451, "global_step": 151019, "epoch": 1696, "lr": 4.132203171163282e-05}
{"train_loss": 0.07388670742511749, "global_step": 151020, "epoch": 1696, "lr": 4.1321460774038214e-05}
{"train_loss": 0.0914899930357933, "global_step": 151021, "epoch": 1696, "lr": 4.132088983761035e-05}
{"train_loss": 0.09206228703260422, "global_step": 151022, "epoch": 1696, "lr": 4.132031890234926e-05}
{"train_loss": 0.14430825412273407, "global_step": 151023, "epoch": 1696, "lr": 4.131974796825508e-05}
{"train_loss": 0.07579784840345383, "global_step": 151024, "epoch": 1696, "lr": 4.1319177035327826e-05}
{"train_loss": 0.15687206387519836, "global_step": 151025, "epoch": 1696, "lr": 4.1318606103567624e-05}
{"train_loss": 0.08036018908023834, "global_step": 151026, "epoch": 1696, "lr": 4.131803517297451e-05}
{"train_loss": 0.10365290939807892, "global_step": 151027, "epoch": 1696, "lr": 4.131746424354861e-05}
{"train_loss": 0.09835531562566757, "global_step": 151028, "epoch": 1696, "lr": 4.131689331528994e-05}
{"train_loss": 0.08189854770898819, "global_step": 151029, "epoch": 1696, "lr": 4.1316322388198625e-05}
{"train_loss": 0.09012659639120102, "global_step": 151030, "epoch": 1696, "lr": 4.131575146227471e-05}
{"train_loss": 0.09962627291679382, "global_step": 151031, "epoch": 1696, "lr": 4.131518053751831e-05}
{"train_loss": 0.10340626098299294, "global_step": 151032, "epoch": 1696, "lr": 4.131460961392945e-05, "val_loss": 6.454154014587402}
{"train_loss": 0.11084242165088654, "global_step": 151033, "epoch": 1697, "lr": 4.131403869150825e-05}
{"train_loss": 0.10492897778749466, "global_step": 151034, "epoch": 1697, "lr": 4.131346777025477e-05}
{"train_loss": 0.1029137372970581, "global_step": 151035, "epoch": 1697, "lr": 4.1312896850169095e-05}
{"train_loss": 0.0671277791261673, "global_step": 151036, "epoch": 1697, "lr": 4.1312325931251295e-05}
{"train_loss": 0.09817188233137131, "global_step": 151037, "epoch": 1697, "lr": 4.131175501350143e-05}
{"train_loss": 0.07524894177913666, "global_step": 151038, "epoch": 1697, "lr": 4.131118409691962e-05}
{"train_loss": 0.09227089583873749, "global_step": 151039, "epoch": 1697, "lr": 4.1310613181505885e-05}
{"train_loss": 0.15321548283100128, "global_step": 151040, "epoch": 1697, "lr": 4.131004226726036e-05}
{"train_loss": 0.163474440574646, "global_step": 151041, "epoch": 1697, "lr": 4.130947135418307e-05}
{"train_loss": 0.10243289172649384, "global_step": 151042, "epoch": 1697, "lr": 4.1308900442274134e-05}
{"train_loss": 0.2098144292831421, "global_step": 151043, "epoch": 1697, "lr": 4.1308329531533596e-05}
{"train_loss": 0.1070796400308609, "global_step": 151044, "epoch": 1697, "lr": 4.1307758621961565e-05}
{"train_loss": 0.08394510298967361, "global_step": 151045, "epoch": 1697, "lr": 4.130718771355808e-05}
{"train_loss": 0.13304443657398224, "global_step": 151046, "epoch": 1697, "lr": 4.1306616806323255e-05}
{"train_loss": 0.05563179403543472, "global_step": 151047, "epoch": 1697, "lr": 4.1306045900257134e-05}
{"train_loss": 0.15289269387722015, "global_step": 151048, "epoch": 1697, "lr": 4.130547499535983e-05}
{"train_loss": 0.07307099550962448, "global_step": 151049, "epoch": 1697, "lr": 4.130490409163137e-05}
{"train_loss": 0.14950011670589447, "global_step": 151050, "epoch": 1697, "lr": 4.130433318907189e-05}
{"train_loss": 0.11425343155860901, "global_step": 151051, "epoch": 1697, "lr": 4.130376228768141e-05}
{"train_loss": 0.07559486478567123, "global_step": 151052, "epoch": 1697, "lr": 4.130319138746004e-05}
{"train_loss": 0.08362343162298203, "global_step": 151053, "epoch": 1697, "lr": 4.130262048840786e-05}
{"train_loss": 0.11676815897226334, "global_step": 151054, "epoch": 1697, "lr": 4.130204959052493e-05}
{"train_loss": 0.0718710869550705, "global_step": 151055, "epoch": 1697, "lr": 4.1301478693811345e-05}
{"train_loss": 0.08628808706998825, "global_step": 151056, "epoch": 1697, "lr": 4.130090779826715e-05}
{"train_loss": 0.12019053101539612, "global_step": 151057, "epoch": 1697, "lr": 4.130033690389246e-05}
{"train_loss": 0.10620304942131042, "global_step": 151058, "epoch": 1697, "lr": 4.1299766010687314e-05}
{"train_loss": 0.12636014819145203, "global_step": 151059, "epoch": 1697, "lr": 4.129919511865183e-05}
{"train_loss": 0.09231670945882797, "global_step": 151060, "epoch": 1697, "lr": 4.129862422778604e-05}
{"train_loss": 0.14113923907279968, "global_step": 151061, "epoch": 1697, "lr": 4.129805333809007e-05}
{"train_loss": 0.09548711776733398, "global_step": 151062, "epoch": 1697, "lr": 4.129748244956395e-05}
{"train_loss": 0.12962038815021515, "global_step": 151063, "epoch": 1697, "lr": 4.1296911562207796e-05}
{"train_loss": 0.0959516167640686, "global_step": 151064, "epoch": 1697, "lr": 4.1296340676021646e-05}
{"train_loss": 0.08950146287679672, "global_step": 151065, "epoch": 1697, "lr": 4.1295769791005615e-05}
{"train_loss": 0.15800516307353973, "global_step": 151066, "epoch": 1697, "lr": 4.1295198907159735e-05}
{"train_loss": 0.07074984163045883, "global_step": 151067, "epoch": 1697, "lr": 4.1294628024484144e-05}
{"train_loss": 0.15514686703681946, "global_step": 151068, "epoch": 1697, "lr": 4.129405714297886e-05}
{"train_loss": 0.1532667726278305, "global_step": 151069, "epoch": 1697, "lr": 4.1293486262644e-05}
{"train_loss": 0.11860238015651703, "global_step": 151070, "epoch": 1697, "lr": 4.1292915383479605e-05}
{"train_loss": 0.09668729454278946, "global_step": 151071, "epoch": 1697, "lr": 4.129234450548578e-05}
{"train_loss": 0.09490711987018585, "global_step": 151072, "epoch": 1697, "lr": 4.12917736286626e-05}
{"train_loss": 0.10153605043888092, "global_step": 151073, "epoch": 1697, "lr": 4.129120275301013e-05}
{"train_loss": 0.10243972390890121, "global_step": 151074, "epoch": 1697, "lr": 4.1290631878528464e-05}
{"train_loss": 0.0767902359366417, "global_step": 151075, "epoch": 1697, "lr": 4.129006100521764e-05}
{"train_loss": 0.15491880476474762, "global_step": 151076, "epoch": 1697, "lr": 4.1289490133077784e-05}
{"train_loss": 0.06600494682788849, "global_step": 151077, "epoch": 1697, "lr": 4.1288919262108934e-05}
{"train_loss": 0.17815646529197693, "global_step": 151078, "epoch": 1697, "lr": 4.12883483923112e-05}
{"train_loss": 0.08798374235630035, "global_step": 151079, "epoch": 1697, "lr": 4.128777752368462e-05}
{"train_loss": 0.08504097163677216, "global_step": 151080, "epoch": 1697, "lr": 4.128720665622931e-05}
{"train_loss": 0.12535350024700165, "global_step": 151081, "epoch": 1697, "lr": 4.128663578994531e-05}
{"train_loss": 0.07442843168973923, "global_step": 151082, "epoch": 1697, "lr": 4.1286064924832746e-05}
{"train_loss": 0.10741829872131348, "global_step": 151083, "epoch": 1697, "lr": 4.128549406089163e-05}
{"train_loss": 0.10227834433317184, "global_step": 151084, "epoch": 1697, "lr": 4.12849231981221e-05}
{"train_loss": 0.1016593798995018, "global_step": 151085, "epoch": 1697, "lr": 4.128435233652419e-05}
{"train_loss": 0.14617332816123962, "global_step": 151086, "epoch": 1697, "lr": 4.128378147609801e-05}
{"train_loss": 0.054319482296705246, "global_step": 151087, "epoch": 1697, "lr": 4.128321061684359e-05}
{"train_loss": 0.09213519096374512, "global_step": 151088, "epoch": 1697, "lr": 4.128263975876106e-05}
{"train_loss": 0.10072774440050125, "global_step": 151089, "epoch": 1697, "lr": 4.128206890185047e-05}
{"train_loss": 0.1074863076210022, "global_step": 151090, "epoch": 1697, "lr": 4.1281498046111886e-05}
{"train_loss": 0.0805598720908165, "global_step": 151091, "epoch": 1697, "lr": 4.1280927191545415e-05}
{"train_loss": 0.1496504694223404, "global_step": 151092, "epoch": 1697, "lr": 4.12803563381511e-05}
{"train_loss": 0.05303395539522171, "global_step": 151093, "epoch": 1697, "lr": 4.127978548592906e-05}
{"train_loss": 0.1017807126045227, "global_step": 151094, "epoch": 1697, "lr": 4.127921463487932e-05}
{"train_loss": 0.06488868594169617, "global_step": 151095, "epoch": 1697, "lr": 4.127864378500201e-05}
{"train_loss": 0.08977146446704865, "global_step": 151096, "epoch": 1697, "lr": 4.127807293629715e-05}
{"train_loss": 0.04775438830256462, "global_step": 151097, "epoch": 1697, "lr": 4.1277502088764877e-05}
{"train_loss": 0.10570105910301208, "global_step": 151098, "epoch": 1697, "lr": 4.127693124240521e-05}
{"train_loss": 0.11692115664482117, "global_step": 151099, "epoch": 1697, "lr": 4.1276360397218276e-05}
{"train_loss": 0.12341970950365067, "global_step": 151100, "epoch": 1697, "lr": 4.127578955320411e-05}
{"train_loss": 0.07681813091039658, "global_step": 151101, "epoch": 1697, "lr": 4.127521871036283e-05}
{"train_loss": 0.106268011033535, "global_step": 151102, "epoch": 1697, "lr": 4.127464786869446e-05}
{"train_loss": 0.0942278653383255, "global_step": 151103, "epoch": 1697, "lr": 4.127407702819914e-05}
{"train_loss": 0.14669577777385712, "global_step": 151104, "epoch": 1697, "lr": 4.127350618887689e-05}
{"train_loss": 0.11231224983930588, "global_step": 151105, "epoch": 1697, "lr": 4.127293535072783e-05}
{"train_loss": 0.10185810178518295, "global_step": 151106, "epoch": 1697, "lr": 4.1272364513751994e-05}
{"train_loss": 0.13244271278381348, "global_step": 151107, "epoch": 1697, "lr": 4.127179367794949e-05}
{"train_loss": 0.10444170236587524, "global_step": 151108, "epoch": 1697, "lr": 4.127122284332039e-05}
{"train_loss": 0.13942793011665344, "global_step": 151109, "epoch": 1697, "lr": 4.1270652009864776e-05}
{"train_loss": 0.15165065228939056, "global_step": 151110, "epoch": 1697, "lr": 4.1270081177582705e-05}
{"train_loss": 0.1110464483499527, "global_step": 151111, "epoch": 1697, "lr": 4.126951034647427e-05}
{"train_loss": 0.0906791165471077, "global_step": 151112, "epoch": 1697, "lr": 4.126893951653955e-05}
{"train_loss": 0.08186057209968567, "global_step": 151113, "epoch": 1697, "lr": 4.12683686877786e-05}
{"train_loss": 0.09188593924045563, "global_step": 151114, "epoch": 1697, "lr": 4.126779786019153e-05}
{"train_loss": 0.08943551778793335, "global_step": 151115, "epoch": 1697, "lr": 4.126722703377838e-05}
{"train_loss": 0.07785391807556152, "global_step": 151116, "epoch": 1697, "lr": 4.1266656208539256e-05}
{"train_loss": 0.14618095755577087, "global_step": 151117, "epoch": 1697, "lr": 4.126608538447421e-05}
{"train_loss": 0.13893568515777588, "global_step": 151118, "epoch": 1697, "lr": 4.126551456158335e-05}
{"train_loss": 0.06076500564813614, "global_step": 151119, "epoch": 1697, "lr": 4.126494373986672e-05}
{"train_loss": 0.12886759638786316, "global_step": 151120, "epoch": 1697, "lr": 4.126437291932443e-05}
{"train_loss": 0.10632039920500155, "global_step": 151121, "epoch": 1697, "lr": 4.126380209995652e-05, "val_loss": 6.485751152038574}
{"train_loss": 0.14642837643623352, "global_step": 151122, "epoch": 1698, "lr": 4.12632312817631e-05}
{"train_loss": 0.14079628884792328, "global_step": 151123, "epoch": 1698, "lr": 4.1262660464744214e-05}
{"train_loss": 0.0969250351190567, "global_step": 151124, "epoch": 1698, "lr": 4.126208964889996e-05}
{"train_loss": 0.05596039071679115, "global_step": 151125, "epoch": 1698, "lr": 4.126151883423043e-05}
{"train_loss": 0.04221515357494354, "global_step": 151126, "epoch": 1698, "lr": 4.126094802073566e-05}
{"train_loss": 0.15380631387233734, "global_step": 151127, "epoch": 1698, "lr": 4.126037720841577e-05}
{"train_loss": 0.08743153512477875, "global_step": 151128, "epoch": 1698, "lr": 4.12598063972708e-05}
{"train_loss": 0.08744185417890549, "global_step": 151129, "epoch": 1698, "lr": 4.1259235587300853e-05}
{"train_loss": 0.1108359768986702, "global_step": 151130, "epoch": 1698, "lr": 4.125866477850599e-05}
{"train_loss": 0.13759273290634155, "global_step": 151131, "epoch": 1698, "lr": 4.12580939708863e-05}
{"train_loss": 0.08806706219911575, "global_step": 151132, "epoch": 1698, "lr": 4.1257523164441844e-05}
{"train_loss": 0.05954625830054283, "global_step": 151133, "epoch": 1698, "lr": 4.125695235917272e-05}
{"train_loss": 0.14714881777763367, "global_step": 151134, "epoch": 1698, "lr": 4.1256381555078974e-05}
{"train_loss": 0.13362416625022888, "global_step": 151135, "epoch": 1698, "lr": 4.125581075216072e-05}
{"train_loss": 0.10497339814901352, "global_step": 151136, "epoch": 1698, "lr": 4.125523995041799e-05}
{"train_loss": 0.0797066017985344, "global_step": 151137, "epoch": 1698, "lr": 4.125466914985091e-05}
{"train_loss": 0.08596453070640564, "global_step": 151138, "epoch": 1698, "lr": 4.125409835045951e-05}
{"train_loss": 0.1926414519548416, "global_step": 151139, "epoch": 1698, "lr": 4.125352755224392e-05}
{"train_loss": 0.06132306158542633, "global_step": 151140, "epoch": 1698, "lr": 4.125295675520417e-05}
{"train_loss": 0.06301331520080566, "global_step": 151141, "epoch": 1698, "lr": 4.125238595934036e-05}
{"train_loss": 0.0823332890868187, "global_step": 151142, "epoch": 1698, "lr": 4.1251815164652544e-05}
{"train_loss": 0.08151789754629135, "global_step": 151143, "epoch": 1698, "lr": 4.1251244371140804e-05}
{"train_loss": 0.04500548914074898, "global_step": 151144, "epoch": 1698, "lr": 4.125067357880527e-05}
{"train_loss": 0.08855542540550232, "global_step": 151145, "epoch": 1698, "lr": 4.125010278764594e-05}
{"train_loss": 0.13983364403247833, "global_step": 151146, "epoch": 1698, "lr": 4.1249531997662954e-05}
{"train_loss": 0.11906679719686508, "global_step": 151147, "epoch": 1698, "lr": 4.1248961208856354e-05}
{"train_loss": 0.11092980951070786, "global_step": 151148, "epoch": 1698, "lr": 4.1248390421226214e-05}
{"train_loss": 0.12535762786865234, "global_step": 151149, "epoch": 1698, "lr": 4.124781963477263e-05}
{"train_loss": 0.09287896007299423, "global_step": 151150, "epoch": 1698, "lr": 4.124724884949567e-05}
{"train_loss": 0.09372086077928543, "global_step": 151151, "epoch": 1698, "lr": 4.124667806539541e-05}
{"train_loss": 0.16186118125915527, "global_step": 151152, "epoch": 1698, "lr": 4.124610728247193e-05}
{"train_loss": 0.14536121487617493, "global_step": 151153, "epoch": 1698, "lr": 4.124553650072529e-05}
{"train_loss": 0.12290561944246292, "global_step": 151154, "epoch": 1698, "lr": 4.1244965720155604e-05}
{"train_loss": 0.09222794324159622, "global_step": 151155, "epoch": 1698, "lr": 4.12443949407629e-05}
{"train_loss": 0.029739482328295708, "global_step": 151156, "epoch": 1698, "lr": 4.12438241625473e-05}
{"train_loss": 0.11470954865217209, "global_step": 151157, "epoch": 1698, "lr": 4.124325338550885e-05}
{"train_loss": 0.07629385590553284, "global_step": 151158, "epoch": 1698, "lr": 4.124268260964765e-05}
{"train_loss": 0.10796556621789932, "global_step": 151159, "epoch": 1698, "lr": 4.1242111834963744e-05}
{"train_loss": 0.07499280571937561, "global_step": 151160, "epoch": 1698, "lr": 4.1241541061457235e-05}
{"train_loss": 0.08134328573942184, "global_step": 151161, "epoch": 1698, "lr": 4.124097028912821e-05}
{"train_loss": 0.07388495653867722, "global_step": 151162, "epoch": 1698, "lr": 4.124039951797671e-05}
{"train_loss": 0.09342547506093979, "global_step": 151163, "epoch": 1698, "lr": 4.123982874800285e-05}
{"train_loss": 0.03745465725660324, "global_step": 151164, "epoch": 1698, "lr": 4.123925797920667e-05}
{"train_loss": 0.14265109598636627, "global_step": 151165, "epoch": 1698, "lr": 4.123868721158829e-05}
{"train_loss": 0.0707901120185852, "global_step": 151166, "epoch": 1698, "lr": 4.123811644514775e-05}
{"train_loss": 0.1016436442732811, "global_step": 151167, "epoch": 1698, "lr": 4.1237545679885126e-05}
{"train_loss": 0.08779565244913101, "global_step": 151168, "epoch": 1698, "lr": 4.123697491580052e-05}
{"train_loss": 0.1320657730102539, "global_step": 151169, "epoch": 1698, "lr": 4.1236404152893995e-05}
{"train_loss": 0.12088744342327118, "global_step": 151170, "epoch": 1698, "lr": 4.123583339116562e-05}
{"train_loss": 0.13904865086078644, "global_step": 151171, "epoch": 1698, "lr": 4.12352626306155e-05}
{"train_loss": 0.14674848318099976, "global_step": 151172, "epoch": 1698, "lr": 4.123469187124367e-05}
{"train_loss": 0.06876170635223389, "global_step": 151173, "epoch": 1698, "lr": 4.123412111305025e-05}
{"train_loss": 0.07613104581832886, "global_step": 151174, "epoch": 1698, "lr": 4.1233550356035264e-05}
{"train_loss": 0.0718970000743866, "global_step": 151175, "epoch": 1698, "lr": 4.123297960019886e-05}
{"train_loss": 0.0766524076461792, "global_step": 151176, "epoch": 1698, "lr": 4.1232408845541035e-05}
{"train_loss": 0.14914704859256744, "global_step": 151177, "epoch": 1698, "lr": 4.123183809206194e-05}
{"train_loss": 0.07428453117609024, "global_step": 151178, "epoch": 1698, "lr": 4.1231267339761594e-05}
{"train_loss": 0.15285596251487732, "global_step": 151179, "epoch": 1698, "lr": 4.123069658864009e-05}
{"train_loss": 0.13172794878482819, "global_step": 151180, "epoch": 1698, "lr": 4.1230125838697545e-05}
{"train_loss": 0.07803331315517426, "global_step": 151181, "epoch": 1698, "lr": 4.1229555089933966e-05}
{"train_loss": 0.10588346421718597, "global_step": 151182, "epoch": 1698, "lr": 4.1228984342349505e-05}
{"train_loss": 0.12571750581264496, "global_step": 151183, "epoch": 1698, "lr": 4.1228413595944174e-05}
{"train_loss": 0.07803753763437271, "global_step": 151184, "epoch": 1698, "lr": 4.122784285071809e-05}
{"train_loss": 0.11648446321487427, "global_step": 151185, "epoch": 1698, "lr": 4.1227272106671305e-05}
{"train_loss": 0.07522159814834595, "global_step": 151186, "epoch": 1698, "lr": 4.122670136380391e-05}
{"train_loss": 0.10929974168539047, "global_step": 151187, "epoch": 1698, "lr": 4.1226130622115976e-05}
{"train_loss": 0.04064672812819481, "global_step": 151188, "epoch": 1698, "lr": 4.122555988160759e-05}
{"train_loss": 0.11046136915683746, "global_step": 151189, "epoch": 1698, "lr": 4.122498914227882e-05}
{"train_loss": 0.0989617183804512, "global_step": 151190, "epoch": 1698, "lr": 4.122441840412974e-05}
{"train_loss": 0.09101186692714691, "global_step": 151191, "epoch": 1698, "lr": 4.122384766716042e-05}
{"train_loss": 0.20240922272205353, "global_step": 151192, "epoch": 1698, "lr": 4.1223276931370964e-05}
{"train_loss": 0.12225638329982758, "global_step": 151193, "epoch": 1698, "lr": 4.1222706196761414e-05}
{"train_loss": 0.10045237094163895, "global_step": 151194, "epoch": 1698, "lr": 4.1222135463331876e-05}
{"train_loss": 0.07628500461578369, "global_step": 151195, "epoch": 1698, "lr": 4.1221564731082404e-05}
{"train_loss": 0.07764414697885513, "global_step": 151196, "epoch": 1698, "lr": 4.12209940000131e-05}
{"train_loss": 0.05572722852230072, "global_step": 151197, "epoch": 1698, "lr": 4.122042327012401e-05}
{"train_loss": 0.07872030884027481, "global_step": 151198, "epoch": 1698, "lr": 4.121985254141523e-05}
{"train_loss": 0.11813095957040787, "global_step": 151199, "epoch": 1698, "lr": 4.121928181388684e-05}
{"train_loss": 0.0674433633685112, "global_step": 151200, "epoch": 1698, "lr": 4.1218711087538906e-05}
{"train_loss": 0.09884706139564514, "global_step": 151201, "epoch": 1698, "lr": 4.1218140362371524e-05}
{"train_loss": 0.139724001288414, "global_step": 151202, "epoch": 1698, "lr": 4.1217569638384736e-05}
{"train_loss": 0.08847907930612564, "global_step": 151203, "epoch": 1698, "lr": 4.121699891557866e-05}
{"train_loss": 0.11803332716226578, "global_step": 151204, "epoch": 1698, "lr": 4.121642819395334e-05}
{"train_loss": 0.054175976663827896, "global_step": 151205, "epoch": 1698, "lr": 4.121585747350886e-05}
{"train_loss": 0.055750343948602676, "global_step": 151206, "epoch": 1698, "lr": 4.1215286754245306e-05}
{"train_loss": 0.11350417137145996, "global_step": 151207, "epoch": 1698, "lr": 4.121471603616275e-05}
{"train_loss": 0.10411480069160461, "global_step": 151208, "epoch": 1698, "lr": 4.1214145319261265e-05}
{"train_loss": 0.10710424184799194, "global_step": 151209, "epoch": 1698, "lr": 4.121357460354094e-05}
{"train_loss": 0.10045127267164461, "global_step": 151210, "epoch": 1698, "lr": 4.121300388900183e-05, "val_loss": 6.426478862762451}
{"train_loss": 0.07436530292034149, "global_step": 151211, "epoch": 1699, "lr": 4.121243317564404e-05}
{"train_loss": 0.1450078785419464, "global_step": 151212, "epoch": 1699, "lr": 4.121186246346761e-05}
{"train_loss": 0.09648299962282181, "global_step": 151213, "epoch": 1699, "lr": 4.121129175247266e-05}
{"train_loss": 0.13151535391807556, "global_step": 151214, "epoch": 1699, "lr": 4.121072104265923e-05}
{"train_loss": 0.08225221931934357, "global_step": 151215, "epoch": 1699, "lr": 4.12101503340274e-05}
{"train_loss": 0.058974988758563995, "global_step": 151216, "epoch": 1699, "lr": 4.120957962657729e-05}
{"train_loss": 0.08779243379831314, "global_step": 151217, "epoch": 1699, "lr": 4.1209008920308914e-05}
{"train_loss": 0.07737236469984055, "global_step": 151218, "epoch": 1699, "lr": 4.1208438215222406e-05}
{"train_loss": 0.0968925729393959, "global_step": 151219, "epoch": 1699, "lr": 4.120786751131779e-05}
{"train_loss": 0.1355881541967392, "global_step": 151220, "epoch": 1699, "lr": 4.12072968085952e-05}
{"train_loss": 0.10545401275157928, "global_step": 151221, "epoch": 1699, "lr": 4.120672610705465e-05}
{"train_loss": 0.1544390469789505, "global_step": 151222, "epoch": 1699, "lr": 4.120615540669628e-05}
{"train_loss": 0.12257097661495209, "global_step": 151223, "epoch": 1699, "lr": 4.1205584707520116e-05}
{"train_loss": 0.11116765439510345, "global_step": 151224, "epoch": 1699, "lr": 4.1205014009526255e-05}
{"train_loss": 0.0858922079205513, "global_step": 151225, "epoch": 1699, "lr": 4.120444331271477e-05}
{"train_loss": 0.0765998512506485, "global_step": 151226, "epoch": 1699, "lr": 4.1203872617085746e-05}
{"train_loss": 0.08203989267349243, "global_step": 151227, "epoch": 1699, "lr": 4.120330192263926e-05}
{"train_loss": 0.08720623701810837, "global_step": 151228, "epoch": 1699, "lr": 4.120273122937537e-05}
{"train_loss": 0.11748956888914108, "global_step": 151229, "epoch": 1699, "lr": 4.1202160537294175e-05}
{"train_loss": 0.09804342687129974, "global_step": 151230, "epoch": 1699, "lr": 4.120158984639575e-05}
{"train_loss": 0.13852199912071228, "global_step": 151231, "epoch": 1699, "lr": 4.120101915668014e-05}
{"train_loss": 0.08683615177869797, "global_step": 151232, "epoch": 1699, "lr": 4.120044846814747e-05}
{"train_loss": 0.043354157358407974, "global_step": 151233, "epoch": 1699, "lr": 4.119987778079777e-05}
{"train_loss": 0.15289507806301117, "global_step": 151234, "epoch": 1699, "lr": 4.119930709463114e-05}
{"train_loss": 0.07368282973766327, "global_step": 151235, "epoch": 1699, "lr": 4.119873640964768e-05}
{"train_loss": 0.1046571135520935, "global_step": 151236, "epoch": 1699, "lr": 4.1198165725847414e-05}
{"train_loss": 0.07674617320299149, "global_step": 151237, "epoch": 1699, "lr": 4.1197595043230474e-05}
{"train_loss": 0.09584550559520721, "global_step": 151238, "epoch": 1699, "lr": 4.119702436179689e-05}
{"train_loss": 0.11555472761392593, "global_step": 151239, "epoch": 1699, "lr": 4.1196453681546774e-05}
{"train_loss": 0.12250485271215439, "global_step": 151240, "epoch": 1699, "lr": 4.119588300248018e-05}
{"train_loss": 0.1342524290084839, "global_step": 151241, "epoch": 1699, "lr": 4.1195312324597196e-05}
{"train_loss": 0.11039875447750092, "global_step": 151242, "epoch": 1699, "lr": 4.1194741647897886e-05}
{"train_loss": 0.0923556238412857, "global_step": 151243, "epoch": 1699, "lr": 4.119417097238235e-05}
{"train_loss": 0.1493292897939682, "global_step": 151244, "epoch": 1699, "lr": 4.119360029805064e-05}
{"train_loss": 0.11597773432731628, "global_step": 151245, "epoch": 1699, "lr": 4.119302962490285e-05}
{"train_loss": 0.1224510669708252, "global_step": 151246, "epoch": 1699, "lr": 4.1192458952939036e-05}
{"train_loss": 0.10950847715139389, "global_step": 151247, "epoch": 1699, "lr": 4.1191888282159294e-05}
{"train_loss": 0.04332534596323967, "global_step": 151248, "epoch": 1699, "lr": 4.1191317612563704e-05}
{"train_loss": 0.12559354305267334, "global_step": 151249, "epoch": 1699, "lr": 4.119074694415233e-05}
{"train_loss": 0.12017905712127686, "global_step": 151250, "epoch": 1699, "lr": 4.119017627692525e-05}
{"train_loss": 0.10466575622558594, "global_step": 151251, "epoch": 1699, "lr": 4.118960561088253e-05}
{"train_loss": 0.13234661519527435, "global_step": 151252, "epoch": 1699, "lr": 4.1189034946024286e-05}
{"train_loss": 0.05697716772556305, "global_step": 151253, "epoch": 1699, "lr": 4.118846428235055e-05}
{"train_loss": 0.1617356389760971, "global_step": 151254, "epoch": 1699, "lr": 4.1187893619861435e-05}
{"train_loss": 0.06828027963638306, "global_step": 151255, "epoch": 1699, "lr": 4.1187322958556976e-05}
{"train_loss": 0.1368962675333023, "global_step": 151256, "epoch": 1699, "lr": 4.1186752298437306e-05}
{"train_loss": 0.06418806314468384, "global_step": 151257, "epoch": 1699, "lr": 4.118618163950244e-05}
{"train_loss": 0.08672598749399185, "global_step": 151258, "epoch": 1699, "lr": 4.118561098175251e-05}
{"train_loss": 0.08058249950408936, "global_step": 151259, "epoch": 1699, "lr": 4.118504032518754e-05}
{"train_loss": 0.06152642145752907, "global_step": 151260, "epoch": 1699, "lr": 4.1184469669807655e-05}
{"train_loss": 0.10300680249929428, "global_step": 151261, "epoch": 1699, "lr": 4.118389901561289e-05}
{"train_loss": 0.06675561517477036, "global_step": 151262, "epoch": 1699, "lr": 4.1183328362603366e-05}
{"train_loss": 0.110943503677845, "global_step": 151263, "epoch": 1699, "lr": 4.118275771077913e-05}
{"train_loss": 0.07005169242620468, "global_step": 151264, "epoch": 1699, "lr": 4.1182187060140266e-05}
{"train_loss": 0.15358906984329224, "global_step": 151265, "epoch": 1699, "lr": 4.118161641068684e-05}
{"train_loss": 0.05844494700431824, "global_step": 151266, "epoch": 1699, "lr": 4.118104576241894e-05}
{"train_loss": 0.06372422724962234, "global_step": 151267, "epoch": 1699, "lr": 4.118047511533665e-05}
{"train_loss": 0.1414162963628769, "global_step": 151268, "epoch": 1699, "lr": 4.117990446944003e-05}
{"train_loss": 0.05403297394514084, "global_step": 151269, "epoch": 1699, "lr": 4.117933382472916e-05}
{"train_loss": 0.062232691794633865, "global_step": 151270, "epoch": 1699, "lr": 4.117876318120412e-05}
{"train_loss": 0.12721684575080872, "global_step": 151271, "epoch": 1699, "lr": 4.1178192538865e-05}
{"train_loss": 0.08727940917015076, "global_step": 151272, "epoch": 1699, "lr": 4.117762189771185e-05}
{"train_loss": 0.07606298476457596, "global_step": 151273, "epoch": 1699, "lr": 4.117705125774478e-05}
{"train_loss": 0.10754058510065079, "global_step": 151274, "epoch": 1699, "lr": 4.1176480618963824e-05}
{"train_loss": 0.10122983902692795, "global_step": 151275, "epoch": 1699, "lr": 4.117590998136911e-05}
{"train_loss": 0.12054864317178726, "global_step": 151276, "epoch": 1699, "lr": 4.1175339344960665e-05}
{"train_loss": 0.1106269434094429, "global_step": 151277, "epoch": 1699, "lr": 4.117476870973861e-05}
{"train_loss": 0.12996922433376312, "global_step": 151278, "epoch": 1699, "lr": 4.117419807570297e-05}
{"train_loss": 0.10565230995416641, "global_step": 151279, "epoch": 1699, "lr": 4.117362744285388e-05}
{"train_loss": 0.08113164454698563, "global_step": 151280, "epoch": 1699, "lr": 4.1173056811191365e-05}
{"train_loss": 0.07947545498609543, "global_step": 151281, "epoch": 1699, "lr": 4.1172486180715544e-05}
{"train_loss": 0.08037031441926956, "global_step": 151282, "epoch": 1699, "lr": 4.1171915551426464e-05}
{"train_loss": 0.16098356246948242, "global_step": 151283, "epoch": 1699, "lr": 4.117134492332422e-05}
{"train_loss": 0.06949896365404129, "global_step": 151284, "epoch": 1699, "lr": 4.117077429640888e-05}
{"train_loss": 0.14268673956394196, "global_step": 151285, "epoch": 1699, "lr": 4.117020367068051e-05}
{"train_loss": 0.0551479235291481, "global_step": 151286, "epoch": 1699, "lr": 4.11696330461392e-05}
{"train_loss": 0.0865621566772461, "global_step": 151287, "epoch": 1699, "lr": 4.1169062422785025e-05}
{"train_loss": 0.13142910599708557, "global_step": 151288, "epoch": 1699, "lr": 4.116849180061808e-05}
{"train_loss": 0.12330574542284012, "global_step": 151289, "epoch": 1699, "lr": 4.11679211796384e-05}
{"train_loss": 0.1002834290266037, "global_step": 151290, "epoch": 1699, "lr": 4.116735055984611e-05}
{"train_loss": 0.05559651181101799, "global_step": 151291, "epoch": 1699, "lr": 4.116677994124124e-05}
{"train_loss": 0.09665291011333466, "global_step": 151292, "epoch": 1699, "lr": 4.1166209323823905e-05}
{"train_loss": 0.06114226579666138, "global_step": 151293, "epoch": 1699, "lr": 4.116563870759415e-05}
{"train_loss": 0.07439720630645752, "global_step": 151294, "epoch": 1699, "lr": 4.116506809255209e-05}
{"train_loss": 0.09505432844161987, "global_step": 151295, "epoch": 1699, "lr": 4.116449747869775e-05}
{"train_loss": 0.12558650970458984, "global_step": 151296, "epoch": 1699, "lr": 4.1163926866031265e-05}
{"train_loss": 0.11608076095581055, "global_step": 151297, "epoch": 1699, "lr": 4.116335625455265e-05}
{"train_loss": 0.0629655197262764, "global_step": 151298, "epoch": 1699, "lr": 4.1162785644262046e-05}
{"train_loss": 0.09924344362670116, "global_step": 151299, "epoch": 1699, "lr": 4.1162215035159475e-05, "val_loss": 6.4666361808776855}
{"train_loss": 0.10910867899656296, "global_step": 151300, "epoch": 1700, "lr": 4.1161644427245054e-05}
{"train_loss": 0.13685357570648193, "global_step": 151301, "epoch": 1700, "lr": 4.116107382051884e-05}
{"train_loss": 0.08642637729644775, "global_step": 151302, "epoch": 1700, "lr": 4.11605032149809e-05}
{"train_loss": 0.15797728300094604, "global_step": 151303, "epoch": 1700, "lr": 4.115993261063132e-05}
{"train_loss": 0.08657295256853104, "global_step": 151304, "epoch": 1700, "lr": 4.1159362007470194e-05}
{"train_loss": 0.14130377769470215, "global_step": 151305, "epoch": 1700, "lr": 4.1158791405497574e-05}
{"train_loss": 0.0662258118391037, "global_step": 151306, "epoch": 1700, "lr": 4.1158220804713545e-05}
{"train_loss": 0.039443548768758774, "global_step": 151307, "epoch": 1700, "lr": 4.11576502051182e-05}
{"train_loss": 0.107136569917202, "global_step": 151308, "epoch": 1700, "lr": 4.115707960671158e-05}
{"train_loss": 0.09308533370494843, "global_step": 151309, "epoch": 1700, "lr": 4.1156509009493806e-05}
{"train_loss": 0.07314547896385193, "global_step": 151310, "epoch": 1700, "lr": 4.1155938413464904e-05}
{"train_loss": 0.07494354248046875, "global_step": 151311, "epoch": 1700, "lr": 4.1155367818625006e-05}
{"train_loss": 0.126139298081398, "global_step": 151312, "epoch": 1700, "lr": 4.115479722497414e-05}
{"train_loss": 0.1519646793603897, "global_step": 151313, "epoch": 1700, "lr": 4.1154226632512425e-05}
{"train_loss": 0.12093913555145264, "global_step": 151314, "epoch": 1700, "lr": 4.115365604123989e-05}
{"train_loss": 0.06819982081651688, "global_step": 151315, "epoch": 1700, "lr": 4.1153085451156666e-05}
{"train_loss": 0.08928874135017395, "global_step": 151316, "epoch": 1700, "lr": 4.115251486226278e-05}
{"train_loss": 0.26432904601097107, "global_step": 151317, "epoch": 1700, "lr": 4.1151944274558346e-05}
{"train_loss": 0.1087593361735344, "global_step": 151318, "epoch": 1700, "lr": 4.115137368804341e-05}
{"train_loss": 0.11688690632581711, "global_step": 151319, "epoch": 1700, "lr": 4.115080310271808e-05}
{"train_loss": 0.16304780542850494, "global_step": 151320, "epoch": 1700, "lr": 4.115023251858241e-05}
{"train_loss": 0.07762967795133591, "global_step": 151321, "epoch": 1700, "lr": 4.114966193563648e-05}
{"train_loss": 0.10263864696025848, "global_step": 151322, "epoch": 1700, "lr": 4.1149091353880375e-05}
{"train_loss": 0.13093721866607666, "global_step": 151323, "epoch": 1700, "lr": 4.114852077331416e-05}
{"train_loss": 0.14352083206176758, "global_step": 151324, "epoch": 1700, "lr": 4.114795019393792e-05}
{"train_loss": 0.15369506180286407, "global_step": 151325, "epoch": 1700, "lr": 4.114737961575173e-05}
{"train_loss": 0.10990183800458908, "global_step": 151326, "epoch": 1700, "lr": 4.1146809038755676e-05}
{"train_loss": 0.14050042629241943, "global_step": 151327, "epoch": 1700, "lr": 4.114623846294981e-05}
{"train_loss": 0.06467771530151367, "global_step": 151328, "epoch": 1700, "lr": 4.1145667888334244e-05}
{"train_loss": 0.12380000203847885, "global_step": 151329, "epoch": 1700, "lr": 4.114509731490901e-05}
{"train_loss": 0.1091386005282402, "global_step": 151330, "epoch": 1700, "lr": 4.1144526742674236e-05}
{"train_loss": 0.10276593267917633, "global_step": 151331, "epoch": 1700, "lr": 4.114395617162995e-05}
{"train_loss": 0.10620596259832382, "global_step": 151332, "epoch": 1700, "lr": 4.1143385601776274e-05}
{"train_loss": 0.0961371511220932, "global_step": 151333, "epoch": 1700, "lr": 4.114281503311324e-05}
{"train_loss": 0.06941350549459457, "global_step": 151334, "epoch": 1700, "lr": 4.114224446564097e-05}
{"train_loss": 0.13222038745880127, "global_step": 151335, "epoch": 1700, "lr": 4.114167389935949e-05}
{"train_loss": 0.11129786819219589, "global_step": 151336, "epoch": 1700, "lr": 4.114110333426893e-05}
{"train_loss": 0.07328111678361893, "global_step": 151337, "epoch": 1700, "lr": 4.1140532770369314e-05}
{"train_loss": 0.1215081661939621, "global_step": 151338, "epoch": 1700, "lr": 4.113996220766077e-05}
{"train_loss": 0.12283642590045929, "global_step": 151339, "epoch": 1700, "lr": 4.1139391646143346e-05}
{"train_loss": 0.11914931237697601, "global_step": 151340, "epoch": 1700, "lr": 4.113882108581711e-05}
{"train_loss": 0.08029012382030487, "global_step": 151341, "epoch": 1700, "lr": 4.113825052668216e-05}
{"train_loss": 0.09319330751895905, "global_step": 151342, "epoch": 1700, "lr": 4.113767996873855e-05}
{"train_loss": 0.12560489773750305, "global_step": 151343, "epoch": 1700, "lr": 4.11371094119864e-05}
{"train_loss": 0.051058992743492126, "global_step": 151344, "epoch": 1700, "lr": 4.113653885642572e-05}
{"train_loss": 0.07666467130184174, "global_step": 151345, "epoch": 1700, "lr": 4.113596830205666e-05}
{"train_loss": 0.0758015438914299, "global_step": 151346, "epoch": 1700, "lr": 4.1135397748879235e-05}
{"train_loss": 0.10944461077451706, "global_step": 151347, "epoch": 1700, "lr": 4.113482719689357e-05}
{"train_loss": 0.09829314798116684, "global_step": 151348, "epoch": 1700, "lr": 4.11342566460997e-05}
{"train_loss": 0.12803463637828827, "global_step": 151349, "epoch": 1700, "lr": 4.113368609649774e-05}
{"train_loss": 0.0936727225780487, "global_step": 151350, "epoch": 1700, "lr": 4.113311554808773e-05}
{"train_loss": 0.06344722956418991, "global_step": 151351, "epoch": 1700, "lr": 4.113254500086978e-05}
{"train_loss": 0.12213628739118576, "global_step": 151352, "epoch": 1700, "lr": 4.113197445484393e-05}
{"train_loss": 0.08125689625740051, "global_step": 151353, "epoch": 1700, "lr": 4.11314039100103e-05}
{"train_loss": 0.1445235311985016, "global_step": 151354, "epoch": 1700, "lr": 4.113083336636893e-05}
{"train_loss": 0.07136277854442596, "global_step": 151355, "epoch": 1700, "lr": 4.113026282391992e-05}
{"train_loss": 0.07307783514261246, "global_step": 151356, "epoch": 1700, "lr": 4.112969228266332e-05}
{"train_loss": 0.08199369162321091, "global_step": 151357, "epoch": 1700, "lr": 4.1129121742599256e-05}
{"train_loss": 0.059416599571704865, "global_step": 151358, "epoch": 1700, "lr": 4.112855120372775e-05}
{"train_loss": 0.10572602599859238, "global_step": 151359, "epoch": 1700, "lr": 4.112798066604892e-05}
{"train_loss": 0.09975414723157883, "global_step": 151360, "epoch": 1700, "lr": 4.112741012956281e-05}
{"train_loss": 0.1069663017988205, "global_step": 151361, "epoch": 1700, "lr": 4.112683959426951e-05}
{"train_loss": 0.05862417817115784, "global_step": 151362, "epoch": 1700, "lr": 4.112626906016911e-05}
{"train_loss": 0.09474197775125504, "global_step": 151363, "epoch": 1700, "lr": 4.112569852726166e-05}
{"train_loss": 0.11878461390733719, "global_step": 151364, "epoch": 1700, "lr": 4.112512799554728e-05}
{"train_loss": 0.05759096518158913, "global_step": 151365, "epoch": 1700, "lr": 4.112455746502598e-05}
{"train_loss": 0.09363910555839539, "global_step": 151366, "epoch": 1700, "lr": 4.1123986935697906e-05}
{"train_loss": 0.1390804797410965, "global_step": 151367, "epoch": 1700, "lr": 4.112341640756308e-05}
{"train_loss": 0.12531130015850067, "global_step": 151368, "epoch": 1700, "lr": 4.112284588062163e-05}
{"train_loss": 0.09886270016431808, "global_step": 151369, "epoch": 1700, "lr": 4.112227535487358e-05}
{"train_loss": 0.10264911502599716, "global_step": 151370, "epoch": 1700, "lr": 4.112170483031905e-05}
{"train_loss": 0.08054318279027939, "global_step": 151371, "epoch": 1700, "lr": 4.1121134306958074e-05}
{"train_loss": 0.10816888511180878, "global_step": 151372, "epoch": 1700, "lr": 4.112056378479078e-05}
{"train_loss": 0.09736885130405426, "global_step": 151373, "epoch": 1700, "lr": 4.1119993263817194e-05}
{"train_loss": 0.07201828062534332, "global_step": 151374, "epoch": 1700, "lr": 4.1119422744037446e-05}
{"train_loss": 0.13172945380210876, "global_step": 151375, "epoch": 1700, "lr": 4.111885222545155e-05}
{"train_loss": 0.0755290687084198, "global_step": 151376, "epoch": 1700, "lr": 4.1118281708059645e-05}
{"train_loss": 0.11831533908843994, "global_step": 151377, "epoch": 1700, "lr": 4.111771119186176e-05}
{"train_loss": 0.09679737687110901, "global_step": 151378, "epoch": 1700, "lr": 4.111714067685799e-05}
{"train_loss": 0.08707103878259659, "global_step": 151379, "epoch": 1700, "lr": 4.111657016304844e-05}
{"train_loss": 0.09165080636739731, "global_step": 151380, "epoch": 1700, "lr": 4.1115999650433125e-05}
{"train_loss": 0.11891572922468185, "global_step": 151381, "epoch": 1700, "lr": 4.1115429139012176e-05}
{"train_loss": 0.12819921970367432, "global_step": 151382, "epoch": 1700, "lr": 4.111485862878564e-05}
{"train_loss": 0.050167717039585114, "global_step": 151383, "epoch": 1700, "lr": 4.1114288119753614e-05}
{"train_loss": 0.1395529955625534, "global_step": 151384, "epoch": 1700, "lr": 4.111371761191614e-05}
{"train_loss": 0.08268795907497406, "global_step": 151385, "epoch": 1700, "lr": 4.111314710527335e-05}
{"train_loss": 0.10917622596025467, "global_step": 151386, "epoch": 1700, "lr": 4.111257659982526e-05}
{"train_loss": 0.06370031088590622, "global_step": 151387, "epoch": 1700, "lr": 4.1112006095572e-05}
{"train_loss": 0.10286298486289014, "global_step": 151388, "epoch": 1700, "lr": 4.11114355925136e-05, "train/sim_max_reward_0": 0.5301274317422227, "train/sim_max_reward_1": 0.5295404803308237, "train/sim_max_reward_2": 0.5150157021571775, "train/sim_max_reward_3": 0.6482337386561018, "train/sim_max_reward_4": 0.9248304702239084, "train/sim_max_reward_5": 0.9874694605716614, "test/sim_max_reward_4500000": 0.9395753192891877, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.4950391775314101, "test/sim_max_reward_4500004": 0.6698032466332643, "test/sim_max_reward_4500005": 0.9932339860852922, "test/sim_max_reward_4500006": 0.9529975197009276, "test/sim_max_reward_4500007": 0.955071858937563, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.5280946194131437, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.46835750411235477, "test/sim_max_reward_4500014": 0.36416716941859817, "test/sim_max_reward_4500015": 0.9309412646529277, "test/sim_max_reward_4500016": 0.9343841717093037, "test/sim_max_reward_4500017": 0.9290430477928245, "test/sim_max_reward_4500018": 0.9931422303447355, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9861285085123614, "test/sim_max_reward_4500023": 0.8182556182485358, "test/sim_max_reward_4500024": 0.39889255450650873, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.9704284149759285, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.8790116371242117, "test/sim_max_reward_4500029": 0.9591603427122158, "test/sim_max_reward_4500030": 0.6399198601862207, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6040829574306832, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.925533384271994, "test/sim_max_reward_4500035": 0.5529431142017825, "test/sim_max_reward_4500036": 0.39214049486620517, "test/sim_max_reward_4500037": 0.5302813817885699, "test/sim_max_reward_4500038": 0.1621596234081499, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9759630636129722, "test/sim_max_reward_4500041": 1.0, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9440168704736979, "test/sim_max_reward_4500044": 0.6564179443063579, "test/sim_max_reward_4500045": 0.48002496149312757, "test/sim_max_reward_4500046": 0.3981529156797759, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9780799601816628, "test/sim_max_reward_4500049": 0.8214519900630721, "train/mean_score": 0.6892028806136493, "test/mean_score": 0.6845379342733113, "val_loss": 6.33767032623291, "train_action_mse_error": 14.263181686401367}
{"train_loss": 0.09356191754341125, "global_step": 151389, "epoch": 1701, "lr": 4.111086509065018e-05}
{"train_loss": 0.12129942327737808, "global_step": 151390, "epoch": 1701, "lr": 4.1110294589981774e-05}
{"train_loss": 0.10569155216217041, "global_step": 151391, "epoch": 1701, "lr": 4.110972409050851e-05}
{"train_loss": 0.08243294060230255, "global_step": 151392, "epoch": 1701, "lr": 4.11091535922304e-05}
{"train_loss": 0.04569930210709572, "global_step": 151393, "epoch": 1701, "lr": 4.1108583095147585e-05}
{"train_loss": 0.14642220735549927, "global_step": 151394, "epoch": 1701, "lr": 4.110801259926009e-05}
{"train_loss": 0.06994916498661041, "global_step": 151395, "epoch": 1701, "lr": 4.1107442104568034e-05}
{"train_loss": 0.12582409381866455, "global_step": 151396, "epoch": 1701, "lr": 4.1106871611071454e-05}
{"train_loss": 0.08113952726125717, "global_step": 151397, "epoch": 1701, "lr": 4.110630111877046e-05}
{"train_loss": 0.05722178518772125, "global_step": 151398, "epoch": 1701, "lr": 4.1105730627665126e-05}
{"train_loss": 0.0881185457110405, "global_step": 151399, "epoch": 1701, "lr": 4.1105160137755494e-05}
{"train_loss": 0.053483784198760986, "global_step": 151400, "epoch": 1701, "lr": 4.110458964904169e-05}
{"train_loss": 0.08655665069818497, "global_step": 151401, "epoch": 1701, "lr": 4.110401916152374e-05}
{"train_loss": 0.11446291208267212, "global_step": 151402, "epoch": 1701, "lr": 4.1103448675201765e-05}
{"train_loss": 0.07530185580253601, "global_step": 151403, "epoch": 1701, "lr": 4.110287819007581e-05}
{"train_loss": 0.08888812363147736, "global_step": 151404, "epoch": 1701, "lr": 4.110230770614598e-05}
{"train_loss": 0.1039256900548935, "global_step": 151405, "epoch": 1701, "lr": 4.1101737223412314e-05}
{"train_loss": 0.0971098318696022, "global_step": 151406, "epoch": 1701, "lr": 4.110116674187494e-05}
{"train_loss": 0.0915452241897583, "global_step": 151407, "epoch": 1701, "lr": 4.110059626153388e-05}
{"train_loss": 0.09190300107002258, "global_step": 151408, "epoch": 1701, "lr": 4.1100025782389254e-05}
{"train_loss": 0.10964250564575195, "global_step": 151409, "epoch": 1701, "lr": 4.1099455304441096e-05}
{"train_loss": 0.09568300098180771, "global_step": 151410, "epoch": 1701, "lr": 4.109888482768954e-05}
{"train_loss": 0.2650330066680908, "global_step": 151411, "epoch": 1701, "lr": 4.1098314352134606e-05}
{"train_loss": 0.08602537214756012, "global_step": 151412, "epoch": 1701, "lr": 4.109774387777641e-05}
{"train_loss": 0.09482034295797348, "global_step": 151413, "epoch": 1701, "lr": 4.1097173404615e-05}
{"train_loss": 0.07744167745113373, "global_step": 151414, "epoch": 1701, "lr": 4.1096602932650476e-05}
{"train_loss": 0.11424218118190765, "global_step": 151415, "epoch": 1701, "lr": 4.10960324618829e-05}
{"train_loss": 0.0994926244020462, "global_step": 151416, "epoch": 1701, "lr": 4.109546199231236e-05}
{"train_loss": 0.10665445029735565, "global_step": 151417, "epoch": 1701, "lr": 4.109489152393893e-05}
{"train_loss": 0.08833611756563187, "global_step": 151418, "epoch": 1701, "lr": 4.109432105676267e-05}
{"train_loss": 0.1294003576040268, "global_step": 151419, "epoch": 1701, "lr": 4.109375059078369e-05}
{"train_loss": 0.08876269310712814, "global_step": 151420, "epoch": 1701, "lr": 4.1093180126002026e-05}
{"train_loss": 0.09933562576770782, "global_step": 151421, "epoch": 1701, "lr": 4.109260966241779e-05}
{"train_loss": 0.10221831500530243, "global_step": 151422, "epoch": 1701, "lr": 4.109203920003103e-05}
{"train_loss": 0.0938173159956932, "global_step": 151423, "epoch": 1701, "lr": 4.109146873884185e-05}
{"train_loss": 0.07838955521583557, "global_step": 151424, "epoch": 1701, "lr": 4.10908982788503e-05}
{"train_loss": 0.07709342986345291, "global_step": 151425, "epoch": 1701, "lr": 4.109032782005649e-05}
{"train_loss": 0.07851004600524902, "global_step": 151426, "epoch": 1701, "lr": 4.108975736246046e-05}
{"train_loss": 0.09394639730453491, "global_step": 151427, "epoch": 1701, "lr": 4.1089186906062315e-05}
{"train_loss": 0.07068762928247452, "global_step": 151428, "epoch": 1701, "lr": 4.10886164508621e-05}
{"train_loss": 0.043706268072128296, "global_step": 151429, "epoch": 1701, "lr": 4.1088045996859937e-05}
{"train_loss": 0.14077600836753845, "global_step": 151430, "epoch": 1701, "lr": 4.1087475544055856e-05}
{"train_loss": 0.08264382928609848, "global_step": 151431, "epoch": 1701, "lr": 4.108690509244998e-05}
{"train_loss": 0.06384588778018951, "global_step": 151432, "epoch": 1701, "lr": 4.1086334642042334e-05}
{"train_loss": 0.08080831170082092, "global_step": 151433, "epoch": 1701, "lr": 4.1085764192833034e-05}
{"train_loss": 0.1494518667459488, "global_step": 151434, "epoch": 1701, "lr": 4.1085193744822144e-05}
{"train_loss": 0.07903499156236649, "global_step": 151435, "epoch": 1701, "lr": 4.108462329800974e-05}
{"train_loss": 0.09162267297506332, "global_step": 151436, "epoch": 1701, "lr": 4.108405285239592e-05}
{"train_loss": 0.10748936235904694, "global_step": 151437, "epoch": 1701, "lr": 4.10834824079807e-05}
{"train_loss": 0.14179348945617676, "global_step": 151438, "epoch": 1701, "lr": 4.1082911964764235e-05}
{"train_loss": 0.08246445655822754, "global_step": 151439, "epoch": 1701, "lr": 4.1082341522746536e-05}
{"train_loss": 0.08791223913431168, "global_step": 151440, "epoch": 1701, "lr": 4.108177108192773e-05}
{"train_loss": 0.08409447222948074, "global_step": 151441, "epoch": 1701, "lr": 4.108120064230785e-05}
{"train_loss": 0.05436720699071884, "global_step": 151442, "epoch": 1701, "lr": 4.1080630203887013e-05}
{"train_loss": 0.1311528980731964, "global_step": 151443, "epoch": 1701, "lr": 4.1080059766665256e-05}
{"train_loss": 0.0585094690322876, "global_step": 151444, "epoch": 1701, "lr": 4.10794893306427e-05}
{"train_loss": 0.16131742298603058, "global_step": 151445, "epoch": 1701, "lr": 4.1078918895819375e-05}
{"train_loss": 0.09186234325170517, "global_step": 151446, "epoch": 1701, "lr": 4.107834846219541e-05}
{"train_loss": 0.10749082267284393, "global_step": 151447, "epoch": 1701, "lr": 4.107777802977081e-05}
{"train_loss": 0.053222231566905975, "global_step": 151448, "epoch": 1701, "lr": 4.107720759854573e-05}
{"train_loss": 0.12887445092201233, "global_step": 151449, "epoch": 1701, "lr": 4.1076637168520193e-05}
{"train_loss": 0.09879030287265778, "global_step": 151450, "epoch": 1701, "lr": 4.107606673969431e-05}
{"train_loss": 0.0320071280002594, "global_step": 151451, "epoch": 1701, "lr": 4.107549631206812e-05}
{"train_loss": 0.11508914828300476, "global_step": 151452, "epoch": 1701, "lr": 4.107492588564173e-05}
{"train_loss": 0.1197521984577179, "global_step": 151453, "epoch": 1701, "lr": 4.1074355460415205e-05}
{"train_loss": 0.08813778311014175, "global_step": 151454, "epoch": 1701, "lr": 4.107378503638863e-05}
{"train_loss": 0.07766517251729965, "global_step": 151455, "epoch": 1701, "lr": 4.107321461356208e-05}
{"train_loss": 0.16670358180999756, "global_step": 151456, "epoch": 1701, "lr": 4.107264419193561e-05}
{"train_loss": 0.05413015931844711, "global_step": 151457, "epoch": 1701, "lr": 4.107207377150934e-05}
{"train_loss": 0.07674853503704071, "global_step": 151458, "epoch": 1701, "lr": 4.107150335228329e-05}
{"train_loss": 0.17149685323238373, "global_step": 151459, "epoch": 1701, "lr": 4.1070932934257596e-05}
{"train_loss": 0.1262502819299698, "global_step": 151460, "epoch": 1701, "lr": 4.1070362517432283e-05}
{"train_loss": 0.09278695285320282, "global_step": 151461, "epoch": 1701, "lr": 4.106979210180747e-05}
{"train_loss": 0.08419022709131241, "global_step": 151462, "epoch": 1701, "lr": 4.1069221687383195e-05}
{"train_loss": 0.09834980964660645, "global_step": 151463, "epoch": 1701, "lr": 4.1068651274159575e-05}
{"train_loss": 0.09441814571619034, "global_step": 151464, "epoch": 1701, "lr": 4.106808086213665e-05}
{"train_loss": 0.04837565869092941, "global_step": 151465, "epoch": 1701, "lr": 4.106751045131453e-05}
{"train_loss": 0.11174647510051727, "global_step": 151466, "epoch": 1701, "lr": 4.1066940041693246e-05}
{"train_loss": 0.0767117291688919, "global_step": 151467, "epoch": 1701, "lr": 4.106636963327293e-05}
{"train_loss": 0.07085318118333817, "global_step": 151468, "epoch": 1701, "lr": 4.106579922605361e-05}
{"train_loss": 0.06776989251375198, "global_step": 151469, "epoch": 1701, "lr": 4.106522882003539e-05}
{"train_loss": 0.08213797956705093, "global_step": 151470, "epoch": 1701, "lr": 4.106465841521836e-05}
{"train_loss": 0.10132394731044769, "global_step": 151471, "epoch": 1701, "lr": 4.106408801160256e-05}
{"train_loss": 0.10029418021440506, "global_step": 151472, "epoch": 1701, "lr": 4.10635176091881e-05}
{"train_loss": 0.13180877268314362, "global_step": 151473, "epoch": 1701, "lr": 4.106294720797502e-05}
{"train_loss": 0.106336310505867, "global_step": 151474, "epoch": 1701, "lr": 4.1062376807963433e-05}
{"train_loss": 0.09567362815141678, "global_step": 151475, "epoch": 1701, "lr": 4.10618064091534e-05}
{"train_loss": 0.13449977338314056, "global_step": 151476, "epoch": 1701, "lr": 4.1061236011545e-05}
{"train_loss": 0.09674570303452149, "global_step": 151477, "epoch": 1701, "lr": 4.1060665615138293e-05, "val_loss": 6.493210792541504}
{"train_loss": 0.07224417477846146, "global_step": 151478, "epoch": 1702, "lr": 4.1060095219933395e-05}
{"train_loss": 0.07673781365156174, "global_step": 151479, "epoch": 1702, "lr": 4.105952482593033e-05}
{"train_loss": 0.06550271809101105, "global_step": 151480, "epoch": 1702, "lr": 4.105895443312923e-05}
{"train_loss": 0.047496333718299866, "global_step": 151481, "epoch": 1702, "lr": 4.105838404153011e-05}
{"train_loss": 0.08264360576868057, "global_step": 151482, "epoch": 1702, "lr": 4.105781365113311e-05}
{"train_loss": 0.094298355281353, "global_step": 151483, "epoch": 1702, "lr": 4.1057243261938266e-05}
{"train_loss": 0.07445667684078217, "global_step": 151484, "epoch": 1702, "lr": 4.105667287394568e-05}
{"train_loss": 0.12006888538599014, "global_step": 151485, "epoch": 1702, "lr": 4.1056102487155394e-05}
{"train_loss": 0.10950100421905518, "global_step": 151486, "epoch": 1702, "lr": 4.105553210156753e-05}
{"train_loss": 0.13783742487430573, "global_step": 151487, "epoch": 1702, "lr": 4.105496171718212e-05}
{"train_loss": 0.14187085628509521, "global_step": 151488, "epoch": 1702, "lr": 4.105439133399925e-05}
{"train_loss": 0.15884113311767578, "global_step": 151489, "epoch": 1702, "lr": 4.105382095201905e-05}
{"train_loss": 0.07775820791721344, "global_step": 151490, "epoch": 1702, "lr": 4.105325057124153e-05}
{"train_loss": 0.19080069661140442, "global_step": 151491, "epoch": 1702, "lr": 4.105268019166679e-05}
{"train_loss": 0.10068129003047943, "global_step": 151492, "epoch": 1702, "lr": 4.105210981329492e-05}
{"train_loss": 0.08088546246290207, "global_step": 151493, "epoch": 1702, "lr": 4.1051539436125974e-05}
{"train_loss": 0.19011105597019196, "global_step": 151494, "epoch": 1702, "lr": 4.105096906016004e-05}
{"train_loss": 0.08493123948574066, "global_step": 151495, "epoch": 1702, "lr": 4.1050398685397204e-05}
{"train_loss": 0.14600984752178192, "global_step": 151496, "epoch": 1702, "lr": 4.1049828311837515e-05}
{"train_loss": 0.05630618333816528, "global_step": 151497, "epoch": 1702, "lr": 4.104925793948109e-05}
{"train_loss": 0.10266274958848953, "global_step": 151498, "epoch": 1702, "lr": 4.104868756832796e-05}
{"train_loss": 0.08395352214574814, "global_step": 151499, "epoch": 1702, "lr": 4.1048117198378244e-05}
{"train_loss": 0.11445891857147217, "global_step": 151500, "epoch": 1702, "lr": 4.1047546829631986e-05}
{"train_loss": 0.08064901828765869, "global_step": 151501, "epoch": 1702, "lr": 4.104697646208929e-05}
{"train_loss": 0.11186591535806656, "global_step": 151502, "epoch": 1702, "lr": 4.10464060957502e-05}
{"train_loss": 0.05018636956810951, "global_step": 151503, "epoch": 1702, "lr": 4.1045835730614837e-05}
{"train_loss": 0.11140821874141693, "global_step": 151504, "epoch": 1702, "lr": 4.104526536668322e-05}
{"train_loss": 0.07538628578186035, "global_step": 151505, "epoch": 1702, "lr": 4.104469500395547e-05}
{"train_loss": 0.12782007455825806, "global_step": 151506, "epoch": 1702, "lr": 4.1044124642431666e-05}
{"train_loss": 0.13285845518112183, "global_step": 151507, "epoch": 1702, "lr": 4.1043554282111854e-05}
{"train_loss": 0.08564102649688721, "global_step": 151508, "epoch": 1702, "lr": 4.1042983922996136e-05}
{"train_loss": 0.08208636194467545, "global_step": 151509, "epoch": 1702, "lr": 4.104241356508458e-05}
{"train_loss": 0.08455508202314377, "global_step": 151510, "epoch": 1702, "lr": 4.104184320837726e-05}
{"train_loss": 0.1920785903930664, "global_step": 151511, "epoch": 1702, "lr": 4.1041272852874255e-05}
{"train_loss": 0.09018617123365402, "global_step": 151512, "epoch": 1702, "lr": 4.104070249857565e-05}
{"train_loss": 0.08753595501184464, "global_step": 151513, "epoch": 1702, "lr": 4.1040132145481495e-05}
{"train_loss": 0.11297992616891861, "global_step": 151514, "epoch": 1702, "lr": 4.103956179359191e-05}
{"train_loss": 0.1036120355129242, "global_step": 151515, "epoch": 1702, "lr": 4.1038991442906914e-05}
{"train_loss": 0.07920574396848679, "global_step": 151516, "epoch": 1702, "lr": 4.103842109342665e-05}
{"train_loss": 0.1670559048652649, "global_step": 151517, "epoch": 1702, "lr": 4.103785074515113e-05}
{"train_loss": 0.11611250042915344, "global_step": 151518, "epoch": 1702, "lr": 4.1037280398080494e-05}
{"train_loss": 0.14717553555965424, "global_step": 151519, "epoch": 1702, "lr": 4.103671005221476e-05}
{"train_loss": 0.053311217576265335, "global_step": 151520, "epoch": 1702, "lr": 4.103613970755404e-05}
{"train_loss": 0.10567981749773026, "global_step": 151521, "epoch": 1702, "lr": 4.10355693640984e-05}
{"train_loss": 0.11605317890644073, "global_step": 151522, "epoch": 1702, "lr": 4.1034999021847934e-05}
{"train_loss": 0.10144513845443726, "global_step": 151523, "epoch": 1702, "lr": 4.1034428680802675e-05}
{"train_loss": 0.09066072106361389, "global_step": 151524, "epoch": 1702, "lr": 4.103385834096273e-05}
{"train_loss": 0.15049205720424652, "global_step": 151525, "epoch": 1702, "lr": 4.1033288002328195e-05}
{"train_loss": 0.11065144836902618, "global_step": 151526, "epoch": 1702, "lr": 4.103271766489911e-05}
{"train_loss": 0.08694804459810257, "global_step": 151527, "epoch": 1702, "lr": 4.103214732867559e-05}
{"train_loss": 0.07789362967014313, "global_step": 151528, "epoch": 1702, "lr": 4.103157699365766e-05}
{"train_loss": 0.04913225769996643, "global_step": 151529, "epoch": 1702, "lr": 4.103100665984544e-05}
{"train_loss": 0.11099307984113693, "global_step": 151530, "epoch": 1702, "lr": 4.1030436327238986e-05}
{"train_loss": 0.07609177380800247, "global_step": 151531, "epoch": 1702, "lr": 4.1029865995838383e-05}
{"train_loss": 0.10338621586561203, "global_step": 151532, "epoch": 1702, "lr": 4.1029295665643705e-05}
{"train_loss": 0.13018712401390076, "global_step": 151533, "epoch": 1702, "lr": 4.1028725336655045e-05}
{"train_loss": 0.113033726811409, "global_step": 151534, "epoch": 1702, "lr": 4.102815500887244e-05}
{"train_loss": 0.0826820433139801, "global_step": 151535, "epoch": 1702, "lr": 4.1027584682296006e-05}
{"train_loss": 0.05761464685201645, "global_step": 151536, "epoch": 1702, "lr": 4.102701435692578e-05}
{"train_loss": 0.13548126816749573, "global_step": 151537, "epoch": 1702, "lr": 4.102644403276189e-05}
{"train_loss": 0.08648829162120819, "global_step": 151538, "epoch": 1702, "lr": 4.102587370980436e-05}
{"train_loss": 0.08723480254411697, "global_step": 151539, "epoch": 1702, "lr": 4.102530338805332e-05}
{"train_loss": 0.13593317568302155, "global_step": 151540, "epoch": 1702, "lr": 4.1024733067508794e-05}
{"train_loss": 0.10050874948501587, "global_step": 151541, "epoch": 1702, "lr": 4.102416274817088e-05}
{"train_loss": 0.11471953988075256, "global_step": 151542, "epoch": 1702, "lr": 4.102359243003968e-05}
{"train_loss": 0.12879520654678345, "global_step": 151543, "epoch": 1702, "lr": 4.102302211311523e-05}
{"train_loss": 0.08533560484647751, "global_step": 151544, "epoch": 1702, "lr": 4.102245179739765e-05}
{"train_loss": 0.09417949616909027, "global_step": 151545, "epoch": 1702, "lr": 4.102188148288697e-05}
{"train_loss": 0.1427672952413559, "global_step": 151546, "epoch": 1702, "lr": 4.10213111695833e-05}
{"train_loss": 0.06770510971546173, "global_step": 151547, "epoch": 1702, "lr": 4.10207408574867e-05}
{"train_loss": 0.16101950407028198, "global_step": 151548, "epoch": 1702, "lr": 4.1020170546597254e-05}
{"train_loss": 0.12200412154197693, "global_step": 151549, "epoch": 1702, "lr": 4.101960023691504e-05}
{"train_loss": 0.11179058253765106, "global_step": 151550, "epoch": 1702, "lr": 4.1019029928440126e-05}
{"train_loss": 0.07619759440422058, "global_step": 151551, "epoch": 1702, "lr": 4.101845962117259e-05}
{"train_loss": 0.03680071979761124, "global_step": 151552, "epoch": 1702, "lr": 4.101788931511252e-05}
{"train_loss": 0.06507798284292221, "global_step": 151553, "epoch": 1702, "lr": 4.101731901025998e-05}
{"train_loss": 0.10091527551412582, "global_step": 151554, "epoch": 1702, "lr": 4.101674870661506e-05}
{"train_loss": 0.08658266067504883, "global_step": 151555, "epoch": 1702, "lr": 4.101617840417782e-05}
{"train_loss": 0.045598581433296204, "global_step": 151556, "epoch": 1702, "lr": 4.101560810294835e-05}
{"train_loss": 0.05896846950054169, "global_step": 151557, "epoch": 1702, "lr": 4.1015037802926716e-05}
{"train_loss": 0.08731643110513687, "global_step": 151558, "epoch": 1702, "lr": 4.1014467504113015e-05}
{"train_loss": 0.09963443875312805, "global_step": 151559, "epoch": 1702, "lr": 4.101389720650729e-05}
{"train_loss": 0.07258788496255875, "global_step": 151560, "epoch": 1702, "lr": 4.1013326910109634e-05}
{"train_loss": 0.11382102966308594, "global_step": 151561, "epoch": 1702, "lr": 4.101275661492015e-05}
{"train_loss": 0.07669200003147125, "global_step": 151562, "epoch": 1702, "lr": 4.101218632093887e-05}
{"train_loss": 0.029596874490380287, "global_step": 151563, "epoch": 1702, "lr": 4.101161602816591e-05}
{"train_loss": 0.07376278191804886, "global_step": 151564, "epoch": 1702, "lr": 4.101104573660131e-05}
{"train_loss": 0.07129758596420288, "global_step": 151565, "epoch": 1702, "lr": 4.101047544624519e-05}
{"train_loss": 0.09866417193094666, "global_step": 151566, "epoch": 1702, "lr": 4.100990515709759e-05, "val_loss": 6.505527496337891}
{"train_loss": 0.07674195617437363, "global_step": 151567, "epoch": 1703, "lr": 4.100933486915859e-05}
{"train_loss": 0.050506625324487686, "global_step": 151568, "epoch": 1703, "lr": 4.1008764582428285e-05}
{"train_loss": 0.0908844918012619, "global_step": 151569, "epoch": 1703, "lr": 4.100819429690674e-05}
{"train_loss": 0.085713692009449, "global_step": 151570, "epoch": 1703, "lr": 4.1007624012594034e-05}
{"train_loss": 0.08523185551166534, "global_step": 151571, "epoch": 1703, "lr": 4.100705372949024e-05}
{"train_loss": 0.07011934369802475, "global_step": 151572, "epoch": 1703, "lr": 4.1006483447595446e-05}
{"train_loss": 0.10265609622001648, "global_step": 151573, "epoch": 1703, "lr": 4.100591316690973e-05}
{"train_loss": 0.10308276861906052, "global_step": 151574, "epoch": 1703, "lr": 4.100534288743313e-05}
{"train_loss": 0.17543765902519226, "global_step": 151575, "epoch": 1703, "lr": 4.1004772609165775e-05}
{"train_loss": 0.09818962216377258, "global_step": 151576, "epoch": 1703, "lr": 4.10042023321077e-05}
{"train_loss": 0.1157069131731987, "global_step": 151577, "epoch": 1703, "lr": 4.100363205625902e-05}
{"train_loss": 0.0639692023396492, "global_step": 151578, "epoch": 1703, "lr": 4.100306178161977e-05}
{"train_loss": 0.08715001493692398, "global_step": 151579, "epoch": 1703, "lr": 4.1002491508190056e-05}
{"train_loss": 0.09007245302200317, "global_step": 151580, "epoch": 1703, "lr": 4.100192123596996e-05}
{"train_loss": 0.06086127087473869, "global_step": 151581, "epoch": 1703, "lr": 4.100135096495953e-05}
{"train_loss": 0.06184845417737961, "global_step": 151582, "epoch": 1703, "lr": 4.100078069515888e-05}
{"train_loss": 0.07305994629859924, "global_step": 151583, "epoch": 1703, "lr": 4.100021042656804e-05}
{"train_loss": 0.14824868738651276, "global_step": 151584, "epoch": 1703, "lr": 4.099964015918714e-05}
{"train_loss": 0.10246159136295319, "global_step": 151585, "epoch": 1703, "lr": 4.09990698930162e-05}
{"train_loss": 0.11902246624231339, "global_step": 151586, "epoch": 1703, "lr": 4.099849962805534e-05}
{"train_loss": 0.13558726012706757, "global_step": 151587, "epoch": 1703, "lr": 4.0997929364304624e-05}
{"train_loss": 0.07005257159471512, "global_step": 151588, "epoch": 1703, "lr": 4.099735910176413e-05}
{"train_loss": 0.13983172178268433, "global_step": 151589, "epoch": 1703, "lr": 4.099678884043392e-05}
{"train_loss": 0.07088630646467209, "global_step": 151590, "epoch": 1703, "lr": 4.099621858031409e-05}
{"train_loss": 0.12567555904388428, "global_step": 151591, "epoch": 1703, "lr": 4.09956483214047e-05}
{"train_loss": 0.13406255841255188, "global_step": 151592, "epoch": 1703, "lr": 4.099507806370585e-05}
{"train_loss": 0.05406578630208969, "global_step": 151593, "epoch": 1703, "lr": 4.099450780721759e-05}
{"train_loss": 0.09263981878757477, "global_step": 151594, "epoch": 1703, "lr": 4.099393755194002e-05}
{"train_loss": 0.16792906820774078, "global_step": 151595, "epoch": 1703, "lr": 4.0993367297873186e-05}
{"train_loss": 0.04873419925570488, "global_step": 151596, "epoch": 1703, "lr": 4.099279704501718e-05}
{"train_loss": 0.12020587921142578, "global_step": 151597, "epoch": 1703, "lr": 4.099222679337211e-05}
{"train_loss": 0.08401822298765182, "global_step": 151598, "epoch": 1703, "lr": 4.0991656542938e-05}
{"train_loss": 0.08876259624958038, "global_step": 151599, "epoch": 1703, "lr": 4.099108629371497e-05}
{"train_loss": 0.10777128487825394, "global_step": 151600, "epoch": 1703, "lr": 4.099051604570307e-05}
{"train_loss": 0.13792555034160614, "global_step": 151601, "epoch": 1703, "lr": 4.0989945798902396e-05}
{"train_loss": 0.12486060708761215, "global_step": 151602, "epoch": 1703, "lr": 4.0989375553313e-05}
{"train_loss": 0.07823111116886139, "global_step": 151603, "epoch": 1703, "lr": 4.098880530893499e-05}
{"train_loss": 0.040745969861745834, "global_step": 151604, "epoch": 1703, "lr": 4.0988235065768406e-05}
{"train_loss": 0.11341868340969086, "global_step": 151605, "epoch": 1703, "lr": 4.098766482381336e-05}
{"train_loss": 0.09731899201869965, "global_step": 151606, "epoch": 1703, "lr": 4.0987094583069904e-05}
{"train_loss": 0.11322689801454544, "global_step": 151607, "epoch": 1703, "lr": 4.098652434353812e-05}
{"train_loss": 0.07982490956783295, "global_step": 151608, "epoch": 1703, "lr": 4.09859541052181e-05}
{"train_loss": 0.042318861931562424, "global_step": 151609, "epoch": 1703, "lr": 4.09853838681099e-05}
{"train_loss": 0.1466785967350006, "global_step": 151610, "epoch": 1703, "lr": 4.0984813632213604e-05}
{"train_loss": 0.05137808993458748, "global_step": 151611, "epoch": 1703, "lr": 4.0984243397529304e-05}
{"train_loss": 0.11387081444263458, "global_step": 151612, "epoch": 1703, "lr": 4.098367316405704e-05}
{"train_loss": 0.11516080796718597, "global_step": 151613, "epoch": 1703, "lr": 4.098310293179693e-05}
{"train_loss": 0.1963013857603073, "global_step": 151614, "epoch": 1703, "lr": 4.098253270074901e-05}
{"train_loss": 0.11974284797906876, "global_step": 151615, "epoch": 1703, "lr": 4.0981962470913386e-05}
{"train_loss": 0.05786379054188728, "global_step": 151616, "epoch": 1703, "lr": 4.098139224229015e-05}
{"train_loss": 0.0980975329875946, "global_step": 151617, "epoch": 1703, "lr": 4.0980822014879326e-05}
{"train_loss": 0.06941761821508408, "global_step": 151618, "epoch": 1703, "lr": 4.098025178868105e-05}
{"train_loss": 0.14138279855251312, "global_step": 151619, "epoch": 1703, "lr": 4.0979681563695335e-05}
{"train_loss": 0.07029927521944046, "global_step": 151620, "epoch": 1703, "lr": 4.097911133992232e-05}
{"train_loss": 0.051507554948329926, "global_step": 151621, "epoch": 1703, "lr": 4.097854111736203e-05}
{"train_loss": 0.08218206465244293, "global_step": 151622, "epoch": 1703, "lr": 4.09779708960146e-05}
{"train_loss": 0.09451594203710556, "global_step": 151623, "epoch": 1703, "lr": 4.097740067588003e-05}
{"train_loss": 0.024224840104579926, "global_step": 151624, "epoch": 1703, "lr": 4.097683045695848e-05}
{"train_loss": 0.08491731435060501, "global_step": 151625, "epoch": 1703, "lr": 4.097626023924996e-05}
{"train_loss": 0.10184824466705322, "global_step": 151626, "epoch": 1703, "lr": 4.097569002275458e-05}
{"train_loss": 0.06139019876718521, "global_step": 151627, "epoch": 1703, "lr": 4.09751198074724e-05}
{"train_loss": 0.06668419390916824, "global_step": 151628, "epoch": 1703, "lr": 4.097454959340351e-05}
{"train_loss": 0.09203656017780304, "global_step": 151629, "epoch": 1703, "lr": 4.097397938054798e-05}
{"train_loss": 0.10185442864894867, "global_step": 151630, "epoch": 1703, "lr": 4.09734091689059e-05}
{"train_loss": 0.10641621798276901, "global_step": 151631, "epoch": 1703, "lr": 4.0972838958477315e-05}
{"train_loss": 0.08762691169977188, "global_step": 151632, "epoch": 1703, "lr": 4.097226874926232e-05}
{"train_loss": 0.09428101778030396, "global_step": 151633, "epoch": 1703, "lr": 4.097169854126102e-05}
{"train_loss": 0.1791161298751831, "global_step": 151634, "epoch": 1703, "lr": 4.0971128334473444e-05}
{"train_loss": 0.14980854094028473, "global_step": 151635, "epoch": 1703, "lr": 4.097055812889971e-05}
{"train_loss": 0.1657337099313736, "global_step": 151636, "epoch": 1703, "lr": 4.0969987924539843e-05}
{"train_loss": 0.09312722831964493, "global_step": 151637, "epoch": 1703, "lr": 4.0969417721393974e-05}
{"train_loss": 0.10777506232261658, "global_step": 151638, "epoch": 1703, "lr": 4.096884751946215e-05}
{"train_loss": 0.09626416862010956, "global_step": 151639, "epoch": 1703, "lr": 4.0968277318744464e-05}
{"train_loss": 0.07676711678504944, "global_step": 151640, "epoch": 1703, "lr": 4.096770711924096e-05}
{"train_loss": 0.07830773293972015, "global_step": 151641, "epoch": 1703, "lr": 4.0967136920951766e-05}
{"train_loss": 0.11617226898670197, "global_step": 151642, "epoch": 1703, "lr": 4.096656672387691e-05}
{"train_loss": 0.034907493740320206, "global_step": 151643, "epoch": 1703, "lr": 4.09659965280165e-05}
{"train_loss": 0.08553728461265564, "global_step": 151644, "epoch": 1703, "lr": 4.09654263333706e-05}
{"train_loss": 0.1001725047826767, "global_step": 151645, "epoch": 1703, "lr": 4.096485613993929e-05}
{"train_loss": 0.11693330854177475, "global_step": 151646, "epoch": 1703, "lr": 4.0964285947722636e-05}
{"train_loss": 0.085832379758358, "global_step": 151647, "epoch": 1703, "lr": 4.096371575672073e-05}
{"train_loss": 0.1458260416984558, "global_step": 151648, "epoch": 1703, "lr": 4.096314556693364e-05}
{"train_loss": 0.0906505286693573, "global_step": 151649, "epoch": 1703, "lr": 4.096257537836146e-05}
{"train_loss": 0.05714377760887146, "global_step": 151650, "epoch": 1703, "lr": 4.096200519100423e-05}
{"train_loss": 0.11926466226577759, "global_step": 151651, "epoch": 1703, "lr": 4.0961435004862036e-05}
{"train_loss": 0.07430928945541382, "global_step": 151652, "epoch": 1703, "lr": 4.0960864819935e-05}
{"train_loss": 0.12867692112922668, "global_step": 151653, "epoch": 1703, "lr": 4.096029463622314e-05}
{"train_loss": 0.09596727043390274, "global_step": 151654, "epoch": 1703, "lr": 4.095972445372658e-05}
{"train_loss": 0.09764215777094444, "global_step": 151655, "epoch": 1703, "lr": 4.0959154272445356e-05, "val_loss": 6.617133140563965}
{"train_loss": 0.14226002991199493, "global_step": 151656, "epoch": 1704, "lr": 4.095858409237958e-05}
{"train_loss": 0.04931897297501564, "global_step": 151657, "epoch": 1704, "lr": 4.0958013913529295e-05}
{"train_loss": 0.0866779237985611, "global_step": 151658, "epoch": 1704, "lr": 4.095744373589462e-05}
{"train_loss": 0.08986084908246994, "global_step": 151659, "epoch": 1704, "lr": 4.0956873559475575e-05}
{"train_loss": 0.08623115718364716, "global_step": 151660, "epoch": 1704, "lr": 4.09563033842723e-05}
{"train_loss": 0.08006089180707932, "global_step": 151661, "epoch": 1704, "lr": 4.0955733210284815e-05}
{"train_loss": 0.12106987833976746, "global_step": 151662, "epoch": 1704, "lr": 4.095516303751324e-05}
{"train_loss": 0.12152113765478134, "global_step": 151663, "epoch": 1704, "lr": 4.095459286595762e-05}
{"train_loss": 0.09508716315031052, "global_step": 151664, "epoch": 1704, "lr": 4.095402269561805e-05}
{"train_loss": 0.10149852186441422, "global_step": 151665, "epoch": 1704, "lr": 4.09534525264946e-05}
{"train_loss": 0.12749481201171875, "global_step": 151666, "epoch": 1704, "lr": 4.095288235858735e-05}
{"train_loss": 0.0785088762640953, "global_step": 151667, "epoch": 1704, "lr": 4.095231219189637e-05}
{"train_loss": 0.11834545433521271, "global_step": 151668, "epoch": 1704, "lr": 4.0951742026421734e-05}
{"train_loss": 0.11032600700855255, "global_step": 151669, "epoch": 1704, "lr": 4.095117186216355e-05}
{"train_loss": 0.05039292573928833, "global_step": 151670, "epoch": 1704, "lr": 4.095060169912184e-05}
{"train_loss": 0.06514114886522293, "global_step": 151671, "epoch": 1704, "lr": 4.095003153729674e-05}
{"train_loss": 0.09855340421199799, "global_step": 151672, "epoch": 1704, "lr": 4.0949461376688276e-05}
{"train_loss": 0.1653212457895279, "global_step": 151673, "epoch": 1704, "lr": 4.094889121729657e-05}
{"train_loss": 0.10889533162117004, "global_step": 151674, "epoch": 1704, "lr": 4.0948321059121646e-05}
{"train_loss": 0.09939683228731155, "global_step": 151675, "epoch": 1704, "lr": 4.094775090216364e-05}
{"train_loss": 0.15461468696594238, "global_step": 151676, "epoch": 1704, "lr": 4.094718074642257e-05}
{"train_loss": 0.06024787575006485, "global_step": 151677, "epoch": 1704, "lr": 4.0946610591898566e-05}
{"train_loss": 0.06978471577167511, "global_step": 151678, "epoch": 1704, "lr": 4.094604043859166e-05}
{"train_loss": 0.15253232419490814, "global_step": 151679, "epoch": 1704, "lr": 4.094547028650197e-05}
{"train_loss": 0.11089050024747849, "global_step": 151680, "epoch": 1704, "lr": 4.0944900135629526e-05}
{"train_loss": 0.13259220123291016, "global_step": 151681, "epoch": 1704, "lr": 4.094432998597445e-05}
{"train_loss": 0.12697350978851318, "global_step": 151682, "epoch": 1704, "lr": 4.0943759837536794e-05}
{"train_loss": 0.14528732001781464, "global_step": 151683, "epoch": 1704, "lr": 4.0943189690316634e-05}
{"train_loss": 0.10097885876893997, "global_step": 151684, "epoch": 1704, "lr": 4.094261954431405e-05}
{"train_loss": 0.1186235249042511, "global_step": 151685, "epoch": 1704, "lr": 4.0942049399529134e-05}
{"train_loss": 0.12103232741355896, "global_step": 151686, "epoch": 1704, "lr": 4.0941479255961935e-05}
{"train_loss": 0.1549413502216339, "global_step": 151687, "epoch": 1704, "lr": 4.0940909113612534e-05}
{"train_loss": 0.09955134987831116, "global_step": 151688, "epoch": 1704, "lr": 4.094033897248104e-05}
{"train_loss": 0.05627990886569023, "global_step": 151689, "epoch": 1704, "lr": 4.093976883256749e-05}
{"train_loss": 0.07246285676956177, "global_step": 151690, "epoch": 1704, "lr": 4.0939198693872e-05}
{"train_loss": 0.11921980232000351, "global_step": 151691, "epoch": 1704, "lr": 4.0938628556394596e-05}
{"train_loss": 0.13447441160678864, "global_step": 151692, "epoch": 1704, "lr": 4.093805842013541e-05}
{"train_loss": 0.10533219575881958, "global_step": 151693, "epoch": 1704, "lr": 4.093748828509447e-05}
{"train_loss": 0.05920013412833214, "global_step": 151694, "epoch": 1704, "lr": 4.09369181512719e-05}
{"train_loss": 0.12799616158008575, "global_step": 151695, "epoch": 1704, "lr": 4.093634801866772e-05}
{"train_loss": 0.11104242503643036, "global_step": 151696, "epoch": 1704, "lr": 4.093577788728206e-05}
{"train_loss": 0.11316465586423874, "global_step": 151697, "epoch": 1704, "lr": 4.093520775711496e-05}
{"train_loss": 0.10538557916879654, "global_step": 151698, "epoch": 1704, "lr": 4.093463762816652e-05}
{"train_loss": 0.06486058980226517, "global_step": 151699, "epoch": 1704, "lr": 4.09340675004368e-05}
{"train_loss": 0.10007509589195251, "global_step": 151700, "epoch": 1704, "lr": 4.0933497373925895e-05}
{"train_loss": 0.1081468015909195, "global_step": 151701, "epoch": 1704, "lr": 4.093292724863385e-05}
{"train_loss": 0.07610516250133514, "global_step": 151702, "epoch": 1704, "lr": 4.0932357124560786e-05}
{"train_loss": 0.10691140592098236, "global_step": 151703, "epoch": 1704, "lr": 4.093178700170675e-05}
{"train_loss": 0.11966953426599503, "global_step": 151704, "epoch": 1704, "lr": 4.093121688007182e-05}
{"train_loss": 0.10484311729669571, "global_step": 151705, "epoch": 1704, "lr": 4.0930646759656074e-05}
{"train_loss": 0.059684474021196365, "global_step": 151706, "epoch": 1704, "lr": 4.093007664045959e-05}
{"train_loss": 0.10311434417963028, "global_step": 151707, "epoch": 1704, "lr": 4.0929506522482466e-05}
{"train_loss": 0.10050144046545029, "global_step": 151708, "epoch": 1704, "lr": 4.0928936405724735e-05}
{"train_loss": 0.056155089288949966, "global_step": 151709, "epoch": 1704, "lr": 4.092836629018652e-05}
{"train_loss": 0.08850287646055222, "global_step": 151710, "epoch": 1704, "lr": 4.092779617586785e-05}
{"train_loss": 0.12288064509630203, "global_step": 151711, "epoch": 1704, "lr": 4.0927226062768856e-05}
{"train_loss": 0.15934008359909058, "global_step": 151712, "epoch": 1704, "lr": 4.092665595088957e-05}
{"train_loss": 0.051157522946596146, "global_step": 151713, "epoch": 1704, "lr": 4.092608584023009e-05}
{"train_loss": 0.0915190652012825, "global_step": 151714, "epoch": 1704, "lr": 4.0925515730790476e-05}
{"train_loss": 0.05943979695439339, "global_step": 151715, "epoch": 1704, "lr": 4.092494562257083e-05}
{"train_loss": 0.10026005655527115, "global_step": 151716, "epoch": 1704, "lr": 4.0924375515571204e-05}
{"train_loss": 0.13921858370304108, "global_step": 151717, "epoch": 1704, "lr": 4.092380540979169e-05}
{"train_loss": 0.13165056705474854, "global_step": 151718, "epoch": 1704, "lr": 4.092323530523235e-05}
{"train_loss": 0.11460281163454056, "global_step": 151719, "epoch": 1704, "lr": 4.0922665201893296e-05}
{"train_loss": 0.10530298203229904, "global_step": 151720, "epoch": 1704, "lr": 4.092209509977455e-05}
{"train_loss": 0.12204896658658981, "global_step": 151721, "epoch": 1704, "lr": 4.092152499887624e-05}
{"train_loss": 0.0850345566868782, "global_step": 151722, "epoch": 1704, "lr": 4.092095489919841e-05}
{"train_loss": 0.17946679890155792, "global_step": 151723, "epoch": 1704, "lr": 4.092038480074114e-05}
{"train_loss": 0.11630111932754517, "global_step": 151724, "epoch": 1704, "lr": 4.091981470350453e-05}
{"train_loss": 0.14421121776103973, "global_step": 151725, "epoch": 1704, "lr": 4.0919244607488614e-05}
{"train_loss": 0.1232239156961441, "global_step": 151726, "epoch": 1704, "lr": 4.0918674512693526e-05}
{"train_loss": 0.08746644854545593, "global_step": 151727, "epoch": 1704, "lr": 4.091810441911928e-05}
{"train_loss": 0.14061854779720306, "global_step": 151728, "epoch": 1704, "lr": 4.091753432676602e-05}
{"train_loss": 0.056324392557144165, "global_step": 151729, "epoch": 1704, "lr": 4.091696423563376e-05}
{"train_loss": 0.10593891888856888, "global_step": 151730, "epoch": 1704, "lr": 4.091639414572262e-05}
{"train_loss": 0.10704715549945831, "global_step": 151731, "epoch": 1704, "lr": 4.0915824057032644e-05}
{"train_loss": 0.09030832350254059, "global_step": 151732, "epoch": 1704, "lr": 4.091525396956394e-05}
{"train_loss": 0.11010546237230301, "global_step": 151733, "epoch": 1704, "lr": 4.091468388331655e-05}
{"train_loss": 0.10805872082710266, "global_step": 151734, "epoch": 1704, "lr": 4.0914113798290596e-05}
{"train_loss": 0.08974352478981018, "global_step": 151735, "epoch": 1704, "lr": 4.09135437144861e-05}
{"train_loss": 0.0600762739777565, "global_step": 151736, "epoch": 1704, "lr": 4.091297363190319e-05}
{"train_loss": 0.10950233787298203, "global_step": 151737, "epoch": 1704, "lr": 4.091240355054191e-05}
{"train_loss": 0.12019486725330353, "global_step": 151738, "epoch": 1704, "lr": 4.091183347040236e-05}
{"train_loss": 0.044167254120111465, "global_step": 151739, "epoch": 1704, "lr": 4.0911263391484576e-05}
{"train_loss": 0.06720192730426788, "global_step": 151740, "epoch": 1704, "lr": 4.09106933137887e-05}
{"train_loss": 0.10135974735021591, "global_step": 151741, "epoch": 1704, "lr": 4.091012323731474e-05}
{"train_loss": 0.0806557908654213, "global_step": 151742, "epoch": 1704, "lr": 4.09095531620628e-05}
{"train_loss": 0.11190785467624664, "global_step": 151743, "epoch": 1704, "lr": 4.090898308803298e-05}
{"train_loss": 0.1033500638905536, "global_step": 151744, "epoch": 1704, "lr": 4.090841301522532e-05, "val_loss": 6.389397621154785}
{"train_loss": 0.15667925775051117, "global_step": 151745, "epoch": 1705, "lr": 4.090784294363993e-05}
{"train_loss": 0.11679307371377945, "global_step": 151746, "epoch": 1705, "lr": 4.0907272873276844e-05}
{"train_loss": 0.12831100821495056, "global_step": 151747, "epoch": 1705, "lr": 4.09067028041362e-05}
{"train_loss": 0.11295419186353683, "global_step": 151748, "epoch": 1705, "lr": 4.0906132736218004e-05}
{"train_loss": 0.08300825208425522, "global_step": 151749, "epoch": 1705, "lr": 4.090556266952239e-05}
{"train_loss": 0.04231476038694382, "global_step": 151750, "epoch": 1705, "lr": 4.0904992604049404e-05}
{"train_loss": 0.051263049244880676, "global_step": 151751, "epoch": 1705, "lr": 4.090442253979914e-05}
{"train_loss": 0.10282153636217117, "global_step": 151752, "epoch": 1705, "lr": 4.0903852476771646e-05}
{"train_loss": 0.10608814656734467, "global_step": 151753, "epoch": 1705, "lr": 4.090328241496704e-05}
{"train_loss": 0.08865860849618912, "global_step": 151754, "epoch": 1705, "lr": 4.090271235438536e-05}
{"train_loss": 0.07895360887050629, "global_step": 151755, "epoch": 1705, "lr": 4.0902142295026716e-05}
{"train_loss": 0.11128364503383636, "global_step": 151756, "epoch": 1705, "lr": 4.090157223689115e-05}
{"train_loss": 0.09035178273916245, "global_step": 151757, "epoch": 1705, "lr": 4.090100217997877e-05}
{"train_loss": 0.09782228618860245, "global_step": 151758, "epoch": 1705, "lr": 4.090043212428963e-05}
{"train_loss": 0.13463333249092102, "global_step": 151759, "epoch": 1705, "lr": 4.089986206982382e-05}
{"train_loss": 0.14950616657733917, "global_step": 151760, "epoch": 1705, "lr": 4.089929201658142e-05}
{"train_loss": 0.0769495889544487, "global_step": 151761, "epoch": 1705, "lr": 4.089872196456248e-05}
{"train_loss": 0.1449182778596878, "global_step": 151762, "epoch": 1705, "lr": 4.089815191376712e-05}
{"train_loss": 0.09735675901174545, "global_step": 151763, "epoch": 1705, "lr": 4.0897581864195374e-05}
{"train_loss": 0.1297885924577713, "global_step": 151764, "epoch": 1705, "lr": 4.089701181584735e-05}
{"train_loss": 0.06215805932879448, "global_step": 151765, "epoch": 1705, "lr": 4.08964417687231e-05}
{"train_loss": 0.06714460998773575, "global_step": 151766, "epoch": 1705, "lr": 4.0895871722822725e-05}
{"train_loss": 0.10678468644618988, "global_step": 151767, "epoch": 1705, "lr": 4.089530167814628e-05}
{"train_loss": 0.05092955380678177, "global_step": 151768, "epoch": 1705, "lr": 4.0894731634693854e-05}
{"train_loss": 0.08027923107147217, "global_step": 151769, "epoch": 1705, "lr": 4.0894161592465515e-05}
{"train_loss": 0.1315394639968872, "global_step": 151770, "epoch": 1705, "lr": 4.0893591551461356e-05}
{"train_loss": 0.06333521753549576, "global_step": 151771, "epoch": 1705, "lr": 4.089302151168143e-05}
{"train_loss": 0.0985775962471962, "global_step": 151772, "epoch": 1705, "lr": 4.089245147312584e-05}
{"train_loss": 0.1281946748495102, "global_step": 151773, "epoch": 1705, "lr": 4.089188143579463e-05}
{"train_loss": 0.09582384675741196, "global_step": 151774, "epoch": 1705, "lr": 4.0891311399687915e-05}
{"train_loss": 0.11178220063447952, "global_step": 151775, "epoch": 1705, "lr": 4.089074136480574e-05}
{"train_loss": 0.15157580375671387, "global_step": 151776, "epoch": 1705, "lr": 4.0890171331148206e-05}
{"train_loss": 0.04828832671046257, "global_step": 151777, "epoch": 1705, "lr": 4.0889601298715356e-05}
{"train_loss": 0.09499666839838028, "global_step": 151778, "epoch": 1705, "lr": 4.08890312675073e-05}
{"train_loss": 0.12716326117515564, "global_step": 151779, "epoch": 1705, "lr": 4.088846123752412e-05}
{"train_loss": 0.1375049501657486, "global_step": 151780, "epoch": 1705, "lr": 4.0887891208765846e-05}
{"train_loss": 0.11850982159376144, "global_step": 151781, "epoch": 1705, "lr": 4.088732118123261e-05}
{"train_loss": 0.08630912005901337, "global_step": 151782, "epoch": 1705, "lr": 4.0886751154924436e-05}
{"train_loss": 0.070313960313797, "global_step": 151783, "epoch": 1705, "lr": 4.088618112984145e-05}
{"train_loss": 0.12331634759902954, "global_step": 151784, "epoch": 1705, "lr": 4.088561110598369e-05}
{"train_loss": 0.10212620347738266, "global_step": 151785, "epoch": 1705, "lr": 4.088504108335126e-05}
{"train_loss": 0.12503251433372498, "global_step": 151786, "epoch": 1705, "lr": 4.0884471061944216e-05}
{"train_loss": 0.08612367510795593, "global_step": 151787, "epoch": 1705, "lr": 4.0883901041762654e-05}
{"train_loss": 0.07755673676729202, "global_step": 151788, "epoch": 1705, "lr": 4.0883331022806627e-05}
{"train_loss": 0.18031901121139526, "global_step": 151789, "epoch": 1705, "lr": 4.088276100507624e-05}
{"train_loss": 0.1053556501865387, "global_step": 151790, "epoch": 1705, "lr": 4.088219098857154e-05}
{"train_loss": 0.10839520394802094, "global_step": 151791, "epoch": 1705, "lr": 4.0881620973292634e-05}
{"train_loss": 0.08265869319438934, "global_step": 151792, "epoch": 1705, "lr": 4.0881050959239565e-05}
{"train_loss": 0.17274358868598938, "global_step": 151793, "epoch": 1705, "lr": 4.0880480946412444e-05}
{"train_loss": 0.16121719777584076, "global_step": 151794, "epoch": 1705, "lr": 4.087991093481132e-05}
{"train_loss": 0.15150071680545807, "global_step": 151795, "epoch": 1705, "lr": 4.0879340924436285e-05}
{"train_loss": 0.12464897334575653, "global_step": 151796, "epoch": 1705, "lr": 4.087877091528741e-05}
{"train_loss": 0.09165244549512863, "global_step": 151797, "epoch": 1705, "lr": 4.087820090736477e-05}
{"train_loss": 0.1329054981470108, "global_step": 151798, "epoch": 1705, "lr": 4.087763090066846e-05}
{"train_loss": 0.207069531083107, "global_step": 151799, "epoch": 1705, "lr": 4.087706089519852e-05}
{"train_loss": 0.10664995014667511, "global_step": 151800, "epoch": 1705, "lr": 4.087649089095507e-05}
{"train_loss": 0.09169704467058182, "global_step": 151801, "epoch": 1705, "lr": 4.087592088793815e-05}
{"train_loss": 0.0766289234161377, "global_step": 151802, "epoch": 1705, "lr": 4.087535088614786e-05}
{"train_loss": 0.054723434150218964, "global_step": 151803, "epoch": 1705, "lr": 4.087478088558425e-05}
{"train_loss": 0.10042725503444672, "global_step": 151804, "epoch": 1705, "lr": 4.087421088624744e-05}
{"train_loss": 0.06309811770915985, "global_step": 151805, "epoch": 1705, "lr": 4.0873640888137454e-05}
{"train_loss": 0.1026727482676506, "global_step": 151806, "epoch": 1705, "lr": 4.087307089125442e-05}
{"train_loss": 0.15921710431575775, "global_step": 151807, "epoch": 1705, "lr": 4.087250089559837e-05}
{"train_loss": 0.14741723239421844, "global_step": 151808, "epoch": 1705, "lr": 4.0871930901169425e-05}
{"train_loss": 0.11557844281196594, "global_step": 151809, "epoch": 1705, "lr": 4.087136090796762e-05}
{"train_loss": 0.1098015233874321, "global_step": 151810, "epoch": 1705, "lr": 4.087079091599306e-05}
{"train_loss": 0.12730804085731506, "global_step": 151811, "epoch": 1705, "lr": 4.087022092524579e-05}
{"train_loss": 0.1220291256904602, "global_step": 151812, "epoch": 1705, "lr": 4.086965093572593e-05}
{"train_loss": 0.09293588250875473, "global_step": 151813, "epoch": 1705, "lr": 4.0869080947433514e-05}
{"train_loss": 0.12594011425971985, "global_step": 151814, "epoch": 1705, "lr": 4.086851096036866e-05}
{"train_loss": 0.0720706582069397, "global_step": 151815, "epoch": 1705, "lr": 4.086794097453141e-05}
{"train_loss": 0.11008481681346893, "global_step": 151816, "epoch": 1705, "lr": 4.0867370989921855e-05}
{"train_loss": 0.08566752076148987, "global_step": 151817, "epoch": 1705, "lr": 4.086680100654007e-05}
{"train_loss": 0.0829932913184166, "global_step": 151818, "epoch": 1705, "lr": 4.0866231024386136e-05}
{"train_loss": 0.07145067304372787, "global_step": 151819, "epoch": 1705, "lr": 4.086566104346013e-05}
{"train_loss": 0.08231906592845917, "global_step": 151820, "epoch": 1705, "lr": 4.086509106376212e-05}
{"train_loss": 0.10945086926221848, "global_step": 151821, "epoch": 1705, "lr": 4.08645210852922e-05}
{"train_loss": 0.1115826740860939, "global_step": 151822, "epoch": 1705, "lr": 4.086395110805041e-05}
{"train_loss": 0.050061628222465515, "global_step": 151823, "epoch": 1705, "lr": 4.086338113203687e-05}
{"train_loss": 0.053382836282253265, "global_step": 151824, "epoch": 1705, "lr": 4.0862811157251614e-05}
{"train_loss": 0.0854463130235672, "global_step": 151825, "epoch": 1705, "lr": 4.086224118369477e-05}
{"train_loss": 0.12746943533420563, "global_step": 151826, "epoch": 1705, "lr": 4.086167121136636e-05}
{"train_loss": 0.10694509744644165, "global_step": 151827, "epoch": 1705, "lr": 4.086110124026651e-05}
{"train_loss": 0.10493883490562439, "global_step": 151828, "epoch": 1705, "lr": 4.0860531270395256e-05}
{"train_loss": 0.15193058550357819, "global_step": 151829, "epoch": 1705, "lr": 4.085996130175271e-05}
{"train_loss": 0.08850593119859695, "global_step": 151830, "epoch": 1705, "lr": 4.0859391334338906e-05}
{"train_loss": 0.11469825357198715, "global_step": 151831, "epoch": 1705, "lr": 4.085882136815397e-05}
{"train_loss": 0.11691766232252121, "global_step": 151832, "epoch": 1705, "lr": 4.0858251403197936e-05}
{"train_loss": 0.10540447919891122, "global_step": 151833, "epoch": 1705, "lr": 4.08576814394709e-05, "val_loss": 6.3163933753967285, "train_action_mse_error": 23.921875}
{"train_loss": 0.05873626843094826, "global_step": 151834, "epoch": 1706, "lr": 4.085711147697295e-05}
{"train_loss": 0.0994659885764122, "global_step": 151835, "epoch": 1706, "lr": 4.085654151570414e-05}
{"train_loss": 0.09445629268884659, "global_step": 151836, "epoch": 1706, "lr": 4.085597155566456e-05}
{"train_loss": 0.07941530644893646, "global_step": 151837, "epoch": 1706, "lr": 4.085540159685428e-05}
{"train_loss": 0.08497268706560135, "global_step": 151838, "epoch": 1706, "lr": 4.0854831639273396e-05}
{"train_loss": 0.10821082442998886, "global_step": 151839, "epoch": 1706, "lr": 4.0854261682921946e-05}
{"train_loss": 0.09238319844007492, "global_step": 151840, "epoch": 1706, "lr": 4.085369172780005e-05}
{"train_loss": 0.10533727705478668, "global_step": 151841, "epoch": 1706, "lr": 4.085312177390774e-05}
{"train_loss": 0.16411596536636353, "global_step": 151842, "epoch": 1706, "lr": 4.085255182124514e-05}
{"train_loss": 0.07861602306365967, "global_step": 151843, "epoch": 1706, "lr": 4.0851981869812275e-05}
{"train_loss": 0.13787151873111725, "global_step": 151844, "epoch": 1706, "lr": 4.085141191960928e-05}
{"train_loss": 0.06490927189588547, "global_step": 151845, "epoch": 1706, "lr": 4.085084197063618e-05}
{"train_loss": 0.10944182425737381, "global_step": 151846, "epoch": 1706, "lr": 4.0850272022893086e-05}
{"train_loss": 0.07534259557723999, "global_step": 151847, "epoch": 1706, "lr": 4.0849702076380046e-05}
{"train_loss": 0.12373784184455872, "global_step": 151848, "epoch": 1706, "lr": 4.0849132131097176e-05}
{"train_loss": 0.13931229710578918, "global_step": 151849, "epoch": 1706, "lr": 4.0848562187044495e-05}
{"train_loss": 0.11871804296970367, "global_step": 151850, "epoch": 1706, "lr": 4.084799224422212e-05}
{"train_loss": 0.14033925533294678, "global_step": 151851, "epoch": 1706, "lr": 4.084742230263015e-05}
{"train_loss": 0.11241079866886139, "global_step": 151852, "epoch": 1706, "lr": 4.084685236226862e-05}
{"train_loss": 0.09766309708356857, "global_step": 151853, "epoch": 1706, "lr": 4.084628242313761e-05}
{"train_loss": 0.09063266217708588, "global_step": 151854, "epoch": 1706, "lr": 4.084571248523721e-05}
{"train_loss": 0.11441211402416229, "global_step": 151855, "epoch": 1706, "lr": 4.084514254856749e-05}
{"train_loss": 0.17508834600448608, "global_step": 151856, "epoch": 1706, "lr": 4.0844572613128535e-05}
{"train_loss": 0.08664777874946594, "global_step": 151857, "epoch": 1706, "lr": 4.0844002678920424e-05}
{"train_loss": 0.06122653931379318, "global_step": 151858, "epoch": 1706, "lr": 4.08434327459432e-05}
{"train_loss": 0.08636382967233658, "global_step": 151859, "epoch": 1706, "lr": 4.0842862814196994e-05}
{"train_loss": 0.08813678473234177, "global_step": 151860, "epoch": 1706, "lr": 4.084229288368182e-05}
{"train_loss": 0.07170786708593369, "global_step": 151861, "epoch": 1706, "lr": 4.084172295439782e-05}
{"train_loss": 0.09669097512960434, "global_step": 151862, "epoch": 1706, "lr": 4.084115302634501e-05}
{"train_loss": 0.14453406631946564, "global_step": 151863, "epoch": 1706, "lr": 4.084058309952353e-05}
{"train_loss": 0.08983156085014343, "global_step": 151864, "epoch": 1706, "lr": 4.084001317393339e-05}
{"train_loss": 0.11358844488859177, "global_step": 151865, "epoch": 1706, "lr": 4.083944324957472e-05}
{"train_loss": 0.0630379244685173, "global_step": 151866, "epoch": 1706, "lr": 4.0838873326447554e-05}
{"train_loss": 0.09300777316093445, "global_step": 151867, "epoch": 1706, "lr": 4.0838303404552016e-05}
{"train_loss": 0.10991468280553818, "global_step": 151868, "epoch": 1706, "lr": 4.083773348388813e-05}
{"train_loss": 0.07536287605762482, "global_step": 151869, "epoch": 1706, "lr": 4.0837163564456e-05}
{"train_loss": 0.09620615094900131, "global_step": 151870, "epoch": 1706, "lr": 4.083659364625572e-05}
{"train_loss": 0.0706435814499855, "global_step": 151871, "epoch": 1706, "lr": 4.083602372928734e-05}
{"train_loss": 0.06473440676927567, "global_step": 151872, "epoch": 1706, "lr": 4.083545381355095e-05}
{"train_loss": 0.07203023880720139, "global_step": 151873, "epoch": 1706, "lr": 4.0834883899046616e-05}
{"train_loss": 0.08587006479501724, "global_step": 151874, "epoch": 1706, "lr": 4.083431398577442e-05}
{"train_loss": 0.09989455342292786, "global_step": 151875, "epoch": 1706, "lr": 4.0833744073734444e-05}
{"train_loss": 0.10166072100400925, "global_step": 151876, "epoch": 1706, "lr": 4.083317416292677e-05}
{"train_loss": 0.10731647908687592, "global_step": 151877, "epoch": 1706, "lr": 4.0832604253351434e-05}
{"train_loss": 0.08570948243141174, "global_step": 151878, "epoch": 1706, "lr": 4.083203434500857e-05}
{"train_loss": 0.06052413955330849, "global_step": 151879, "epoch": 1706, "lr": 4.0831464437898204e-05}
{"train_loss": 0.08616147935390472, "global_step": 151880, "epoch": 1706, "lr": 4.083089453202046e-05}
{"train_loss": 0.10367453098297119, "global_step": 151881, "epoch": 1706, "lr": 4.083032462737536e-05}
{"train_loss": 0.16238994896411896, "global_step": 151882, "epoch": 1706, "lr": 4.0829754723963045e-05}
{"train_loss": 0.10206866264343262, "global_step": 151883, "epoch": 1706, "lr": 4.0829184821783534e-05}
{"train_loss": 0.13238319754600525, "global_step": 151884, "epoch": 1706, "lr": 4.082861492083695e-05}
{"train_loss": 0.08473768830299377, "global_step": 151885, "epoch": 1706, "lr": 4.082804502112332e-05}
{"train_loss": 0.10645324736833572, "global_step": 151886, "epoch": 1706, "lr": 4.082747512264275e-05}
{"train_loss": 0.0626072809100151, "global_step": 151887, "epoch": 1706, "lr": 4.0826905225395326e-05}
{"train_loss": 0.03399256244301796, "global_step": 151888, "epoch": 1706, "lr": 4.08263353293811e-05}
{"train_loss": 0.05663599073886871, "global_step": 151889, "epoch": 1706, "lr": 4.082576543460019e-05}
{"train_loss": 0.08902142196893692, "global_step": 151890, "epoch": 1706, "lr": 4.0825195541052625e-05}
{"train_loss": 0.08756124973297119, "global_step": 151891, "epoch": 1706, "lr": 4.08246256487385e-05}
{"train_loss": 0.1306285858154297, "global_step": 151892, "epoch": 1706, "lr": 4.08240557576579e-05}
{"train_loss": 0.07397014647722244, "global_step": 151893, "epoch": 1706, "lr": 4.082348586781089e-05}
{"train_loss": 0.09282224625349045, "global_step": 151894, "epoch": 1706, "lr": 4.082291597919754e-05}
{"train_loss": 0.15457993745803833, "global_step": 151895, "epoch": 1706, "lr": 4.082234609181796e-05}
{"train_loss": 0.06467902660369873, "global_step": 151896, "epoch": 1706, "lr": 4.082177620567218e-05}
{"train_loss": 0.04250939190387726, "global_step": 151897, "epoch": 1706, "lr": 4.0821206320760316e-05}
{"train_loss": 0.11990365386009216, "global_step": 151898, "epoch": 1706, "lr": 4.0820636437082416e-05}
{"train_loss": 0.07661979645490646, "global_step": 151899, "epoch": 1706, "lr": 4.082006655463858e-05}
{"train_loss": 0.1251670867204666, "global_step": 151900, "epoch": 1706, "lr": 4.081949667342886e-05}
{"train_loss": 0.0697317123413086, "global_step": 151901, "epoch": 1706, "lr": 4.081892679345337e-05}
{"train_loss": 0.1689116507768631, "global_step": 151902, "epoch": 1706, "lr": 4.081835691471214e-05}
{"train_loss": 0.07588095217943192, "global_step": 151903, "epoch": 1706, "lr": 4.081778703720529e-05}
{"train_loss": 0.07497122138738632, "global_step": 151904, "epoch": 1706, "lr": 4.081721716093286e-05}
{"train_loss": 0.11159955710172653, "global_step": 151905, "epoch": 1706, "lr": 4.081664728589494e-05}
{"train_loss": 0.11700600385665894, "global_step": 151906, "epoch": 1706, "lr": 4.0816077412091626e-05}
{"train_loss": 0.09040530771017075, "global_step": 151907, "epoch": 1706, "lr": 4.0815507539522964e-05}
{"train_loss": 0.08325459808111191, "global_step": 151908, "epoch": 1706, "lr": 4.0814937668189065e-05}
{"train_loss": 0.04942599684000015, "global_step": 151909, "epoch": 1706, "lr": 4.081436779808997e-05}
{"train_loss": 0.06408271938562393, "global_step": 151910, "epoch": 1706, "lr": 4.0813797929225776e-05}
{"train_loss": 0.10400548577308655, "global_step": 151911, "epoch": 1706, "lr": 4.081322806159656e-05}
{"train_loss": 0.08729106187820435, "global_step": 151912, "epoch": 1706, "lr": 4.081265819520238e-05}
{"train_loss": 0.15692828595638275, "global_step": 151913, "epoch": 1706, "lr": 4.081208833004334e-05}
{"train_loss": 0.06974196434020996, "global_step": 151914, "epoch": 1706, "lr": 4.081151846611951e-05}
{"train_loss": 0.10616135597229004, "global_step": 151915, "epoch": 1706, "lr": 4.081094860343093e-05}
{"train_loss": 0.13883076608181, "global_step": 151916, "epoch": 1706, "lr": 4.081037874197773e-05}
{"train_loss": 0.09623531997203827, "global_step": 151917, "epoch": 1706, "lr": 4.080980888175995e-05}
{"train_loss": 0.05563506484031677, "global_step": 151918, "epoch": 1706, "lr": 4.0809239022777694e-05}
{"train_loss": 0.06937392055988312, "global_step": 151919, "epoch": 1706, "lr": 4.0808669165030996e-05}
{"train_loss": 0.07183711230754852, "global_step": 151920, "epoch": 1706, "lr": 4.080809930852e-05}
{"train_loss": 0.10344264656305313, "global_step": 151921, "epoch": 1706, "lr": 4.0807529453244705e-05}
{"train_loss": 0.09581363774584921, "global_step": 151922, "epoch": 1706, "lr": 4.080695959920523e-05, "val_loss": 6.712111473083496}
{"train_loss": 0.10248742997646332, "global_step": 151923, "epoch": 1707, "lr": 4.080638974640167e-05}
{"train_loss": 0.14966291189193726, "global_step": 151924, "epoch": 1707, "lr": 4.080581989483406e-05}
{"train_loss": 0.07631199061870575, "global_step": 151925, "epoch": 1707, "lr": 4.080525004450251e-05}
{"train_loss": 0.09496316313743591, "global_step": 151926, "epoch": 1707, "lr": 4.080468019540707e-05}
{"train_loss": 0.0722815990447998, "global_step": 151927, "epoch": 1707, "lr": 4.080411034754784e-05}
{"train_loss": 0.0889921635389328, "global_step": 151928, "epoch": 1707, "lr": 4.080354050092487e-05}
{"train_loss": 0.15077254176139832, "global_step": 151929, "epoch": 1707, "lr": 4.080297065553827e-05}
{"train_loss": 0.07457748800516129, "global_step": 151930, "epoch": 1707, "lr": 4.0802400811388094e-05}
{"train_loss": 0.03600623831152916, "global_step": 151931, "epoch": 1707, "lr": 4.080183096847442e-05}
{"train_loss": 0.12334862351417542, "global_step": 151932, "epoch": 1707, "lr": 4.080126112679733e-05}
{"train_loss": 0.10184730589389801, "global_step": 151933, "epoch": 1707, "lr": 4.080069128635689e-05}
{"train_loss": 0.07842192053794861, "global_step": 151934, "epoch": 1707, "lr": 4.080012144715319e-05}
{"train_loss": 0.051062848418951035, "global_step": 151935, "epoch": 1707, "lr": 4.0799551609186307e-05}
{"train_loss": 0.10219597071409225, "global_step": 151936, "epoch": 1707, "lr": 4.07989817724563e-05}
{"train_loss": 0.10076936334371567, "global_step": 151937, "epoch": 1707, "lr": 4.079841193696328e-05}
{"train_loss": 0.1158628761768341, "global_step": 151938, "epoch": 1707, "lr": 4.079784210270727e-05}
{"train_loss": 0.10816871374845505, "global_step": 151939, "epoch": 1707, "lr": 4.0797272269688406e-05}
{"train_loss": 0.09624746441841125, "global_step": 151940, "epoch": 1707, "lr": 4.079670243790671e-05}
{"train_loss": 0.06481912732124329, "global_step": 151941, "epoch": 1707, "lr": 4.079613260736229e-05}
{"train_loss": 0.07188583165407181, "global_step": 151942, "epoch": 1707, "lr": 4.0795562778055235e-05}
{"train_loss": 0.08274488151073456, "global_step": 151943, "epoch": 1707, "lr": 4.079499294998559e-05}
{"train_loss": 0.043850794434547424, "global_step": 151944, "epoch": 1707, "lr": 4.079442312315346e-05}
{"train_loss": 0.09981425106525421, "global_step": 151945, "epoch": 1707, "lr": 4.0793853297558894e-05}
{"train_loss": 0.0369124710559845, "global_step": 151946, "epoch": 1707, "lr": 4.079328347320199e-05}
{"train_loss": 0.07953332364559174, "global_step": 151947, "epoch": 1707, "lr": 4.079271365008281e-05}
{"train_loss": 0.1250331699848175, "global_step": 151948, "epoch": 1707, "lr": 4.079214382820145e-05}
{"train_loss": 0.08605992794036865, "global_step": 151949, "epoch": 1707, "lr": 4.079157400755797e-05}
{"train_loss": 0.1255728155374527, "global_step": 151950, "epoch": 1707, "lr": 4.0791004188152446e-05}
{"train_loss": 0.11586062610149384, "global_step": 151951, "epoch": 1707, "lr": 4.079043436998496e-05}
{"train_loss": 0.057831522077322006, "global_step": 151952, "epoch": 1707, "lr": 4.078986455305559e-05}
{"train_loss": 0.11411349475383759, "global_step": 151953, "epoch": 1707, "lr": 4.078929473736441e-05}
{"train_loss": 0.08496500551700592, "global_step": 151954, "epoch": 1707, "lr": 4.0788724922911504e-05}
{"train_loss": 0.08552762120962143, "global_step": 151955, "epoch": 1707, "lr": 4.078815510969693e-05}
{"train_loss": 0.1583043783903122, "global_step": 151956, "epoch": 1707, "lr": 4.0787585297720786e-05}
{"train_loss": 0.08973218500614166, "global_step": 151957, "epoch": 1707, "lr": 4.078701548698312e-05}
{"train_loss": 0.1394004076719284, "global_step": 151958, "epoch": 1707, "lr": 4.078644567748405e-05}
{"train_loss": 0.12891554832458496, "global_step": 151959, "epoch": 1707, "lr": 4.078587586922361e-05}
{"train_loss": 0.13741901516914368, "global_step": 151960, "epoch": 1707, "lr": 4.07853060622019e-05}
{"train_loss": 0.12969769537448883, "global_step": 151961, "epoch": 1707, "lr": 4.078473625641901e-05}
{"train_loss": 0.10267601907253265, "global_step": 151962, "epoch": 1707, "lr": 4.0784166451874976e-05}
{"train_loss": 0.16172201931476593, "global_step": 151963, "epoch": 1707, "lr": 4.078359664856992e-05}
{"train_loss": 0.0524735227227211, "global_step": 151964, "epoch": 1707, "lr": 4.078302684650388e-05}
{"train_loss": 0.08076196163892746, "global_step": 151965, "epoch": 1707, "lr": 4.078245704567698e-05}
{"train_loss": 0.10763376206159592, "global_step": 151966, "epoch": 1707, "lr": 4.078188724608923e-05}
{"train_loss": 0.11282367259263992, "global_step": 151967, "epoch": 1707, "lr": 4.078131744774077e-05}
{"train_loss": 0.0904826819896698, "global_step": 151968, "epoch": 1707, "lr": 4.078074765063164e-05}
{"train_loss": 0.1520203948020935, "global_step": 151969, "epoch": 1707, "lr": 4.0780177854761916e-05}
{"train_loss": 0.07803541421890259, "global_step": 151970, "epoch": 1707, "lr": 4.0779608060131696e-05}
{"train_loss": 0.07979379594326019, "global_step": 151971, "epoch": 1707, "lr": 4.0779038266741044e-05}
{"train_loss": 0.06730584800243378, "global_step": 151972, "epoch": 1707, "lr": 4.0778468474590035e-05}
{"train_loss": 0.11196725815534592, "global_step": 151973, "epoch": 1707, "lr": 4.077789868367876e-05}
{"train_loss": 0.08930899202823639, "global_step": 151974, "epoch": 1707, "lr": 4.077732889400726e-05}
{"train_loss": 0.062451183795928955, "global_step": 151975, "epoch": 1707, "lr": 4.077675910557567e-05}
{"train_loss": 0.09503956139087677, "global_step": 151976, "epoch": 1707, "lr": 4.077618931838399e-05}
{"train_loss": 0.07230900973081589, "global_step": 151977, "epoch": 1707, "lr": 4.0775619532432356e-05}
{"train_loss": 0.10557552427053452, "global_step": 151978, "epoch": 1707, "lr": 4.0775049747720844e-05}
{"train_loss": 0.08072622120380402, "global_step": 151979, "epoch": 1707, "lr": 4.077447996424949e-05}
{"train_loss": 0.06896263360977173, "global_step": 151980, "epoch": 1707, "lr": 4.0773910182018415e-05}
{"train_loss": 0.10632102936506271, "global_step": 151981, "epoch": 1707, "lr": 4.077334040102766e-05}
{"train_loss": 0.1860942840576172, "global_step": 151982, "epoch": 1707, "lr": 4.077277062127734e-05}
{"train_loss": 0.061913080513477325, "global_step": 151983, "epoch": 1707, "lr": 4.0772200842767475e-05}
{"train_loss": 0.0793885737657547, "global_step": 151984, "epoch": 1707, "lr": 4.0771631065498204e-05}
{"train_loss": 0.09406732767820358, "global_step": 151985, "epoch": 1707, "lr": 4.077106128946955e-05}
{"train_loss": 0.09817345440387726, "global_step": 151986, "epoch": 1707, "lr": 4.077049151468164e-05}
{"train_loss": 0.1026257649064064, "global_step": 151987, "epoch": 1707, "lr": 4.076992174113452e-05}
{"train_loss": 0.16710716485977173, "global_step": 151988, "epoch": 1707, "lr": 4.0769351968828264e-05}
{"train_loss": 0.09478062391281128, "global_step": 151989, "epoch": 1707, "lr": 4.0768782197762954e-05}
{"train_loss": 0.13786020874977112, "global_step": 151990, "epoch": 1707, "lr": 4.076821242793868e-05}
{"train_loss": 0.08125761151313782, "global_step": 151991, "epoch": 1707, "lr": 4.0767642659355485e-05}
{"train_loss": 0.10614340752363205, "global_step": 151992, "epoch": 1707, "lr": 4.0767072892013495e-05}
{"train_loss": 0.12815159559249878, "global_step": 151993, "epoch": 1707, "lr": 4.076650312591273e-05}
{"train_loss": 0.10604733228683472, "global_step": 151994, "epoch": 1707, "lr": 4.076593336105333e-05}
{"train_loss": 0.12796126306056976, "global_step": 151995, "epoch": 1707, "lr": 4.0765363597435316e-05}
{"train_loss": 0.12823866307735443, "global_step": 151996, "epoch": 1707, "lr": 4.0764793835058775e-05}
{"train_loss": 0.10419661551713943, "global_step": 151997, "epoch": 1707, "lr": 4.0764224073923815e-05}
{"train_loss": 0.07584654539823532, "global_step": 151998, "epoch": 1707, "lr": 4.0763654314030484e-05}
{"train_loss": 0.11081482470035553, "global_step": 151999, "epoch": 1707, "lr": 4.0763084555378876e-05}
{"train_loss": 0.13982799649238586, "global_step": 152000, "epoch": 1707, "lr": 4.076251479796905e-05}
{"train_loss": 0.111412413418293, "global_step": 152001, "epoch": 1707, "lr": 4.07619450418011e-05}
{"train_loss": 0.09408159554004669, "global_step": 152002, "epoch": 1707, "lr": 4.076137528687508e-05}
{"train_loss": 0.12657390534877777, "global_step": 152003, "epoch": 1707, "lr": 4.07608055331911e-05}
{"train_loss": 0.1356395184993744, "global_step": 152004, "epoch": 1707, "lr": 4.0760235780749196e-05}
{"train_loss": 0.1044345274567604, "global_step": 152005, "epoch": 1707, "lr": 4.0759666029549485e-05}
{"train_loss": 0.1361115574836731, "global_step": 152006, "epoch": 1707, "lr": 4.075909627959201e-05}
{"train_loss": 0.06169702112674713, "global_step": 152007, "epoch": 1707, "lr": 4.0758526530876874e-05}
{"train_loss": 0.1290176510810852, "global_step": 152008, "epoch": 1707, "lr": 4.0757956783404135e-05}
{"train_loss": 0.1407804936170578, "global_step": 152009, "epoch": 1707, "lr": 4.075738703717388e-05}
{"train_loss": 0.09600751101970673, "global_step": 152010, "epoch": 1707, "lr": 4.075681729218617e-05}
{"train_loss": 0.10162239408727443, "global_step": 152011, "epoch": 1707, "lr": 4.075624754844111e-05, "val_loss": 6.450213432312012}
{"train_loss": 0.07732350379228592, "global_step": 152012, "epoch": 1708, "lr": 4.075567780593874e-05}
{"train_loss": 0.10524159669876099, "global_step": 152013, "epoch": 1708, "lr": 4.075510806467916e-05}
{"train_loss": 0.11591458320617676, "global_step": 152014, "epoch": 1708, "lr": 4.075453832466246e-05}
{"train_loss": 0.07340971380472183, "global_step": 152015, "epoch": 1708, "lr": 4.075396858588868e-05}
{"train_loss": 0.051709096878767014, "global_step": 152016, "epoch": 1708, "lr": 4.075339884835794e-05}
{"train_loss": 0.16881653666496277, "global_step": 152017, "epoch": 1708, "lr": 4.0752829112070265e-05}
{"train_loss": 0.1284288763999939, "global_step": 152018, "epoch": 1708, "lr": 4.0752259377025783e-05}
{"train_loss": 0.10509061813354492, "global_step": 152019, "epoch": 1708, "lr": 4.075168964322452e-05}
{"train_loss": 0.0746590793132782, "global_step": 152020, "epoch": 1708, "lr": 4.075111991066661e-05}
{"train_loss": 0.09484116733074188, "global_step": 152021, "epoch": 1708, "lr": 4.075055017935208e-05}
{"train_loss": 0.09364014863967896, "global_step": 152022, "epoch": 1708, "lr": 4.074998044928104e-05}
{"train_loss": 0.07729458808898926, "global_step": 152023, "epoch": 1708, "lr": 4.074941072045353e-05}
{"train_loss": 0.09013403952121735, "global_step": 152024, "epoch": 1708, "lr": 4.0748840992869674e-05}
{"train_loss": 0.04711397364735603, "global_step": 152025, "epoch": 1708, "lr": 4.074827126652951e-05}
{"train_loss": 0.040445584803819656, "global_step": 152026, "epoch": 1708, "lr": 4.0747701541433135e-05}
{"train_loss": 0.05682380497455597, "global_step": 152027, "epoch": 1708, "lr": 4.0747131817580615e-05}
{"train_loss": 0.13479818403720856, "global_step": 152028, "epoch": 1708, "lr": 4.074656209497203e-05}
{"train_loss": 0.08280731737613678, "global_step": 152029, "epoch": 1708, "lr": 4.074599237360745e-05}
{"train_loss": 0.13097833096981049, "global_step": 152030, "epoch": 1708, "lr": 4.0745422653486974e-05}
{"train_loss": 0.05804739519953728, "global_step": 152031, "epoch": 1708, "lr": 4.074485293461066e-05}
{"train_loss": 0.08766302466392517, "global_step": 152032, "epoch": 1708, "lr": 4.0744283216978564e-05}
{"train_loss": 0.07534877955913544, "global_step": 152033, "epoch": 1708, "lr": 4.074371350059081e-05}
{"train_loss": 0.10551553964614868, "global_step": 152034, "epoch": 1708, "lr": 4.074314378544744e-05}
{"train_loss": 0.14864470064640045, "global_step": 152035, "epoch": 1708, "lr": 4.0742574071548555e-05}
{"train_loss": 0.10204873234033585, "global_step": 152036, "epoch": 1708, "lr": 4.0742004358894204e-05}
{"train_loss": 0.16588813066482544, "global_step": 152037, "epoch": 1708, "lr": 4.074143464748449e-05}
{"train_loss": 0.05861994996666908, "global_step": 152038, "epoch": 1708, "lr": 4.074086493731946e-05}
{"train_loss": 0.11216481775045395, "global_step": 152039, "epoch": 1708, "lr": 4.0740295228399236e-05}
{"train_loss": 0.1668063998222351, "global_step": 152040, "epoch": 1708, "lr": 4.073972552072384e-05}
{"train_loss": 0.10932880640029907, "global_step": 152041, "epoch": 1708, "lr": 4.073915581429339e-05}
{"train_loss": 0.09074215590953827, "global_step": 152042, "epoch": 1708, "lr": 4.073858610910793e-05}
{"train_loss": 0.07749779522418976, "global_step": 152043, "epoch": 1708, "lr": 4.073801640516758e-05}
{"train_loss": 0.09239774942398071, "global_step": 152044, "epoch": 1708, "lr": 4.073744670247236e-05}
{"train_loss": 0.09329306334257126, "global_step": 152045, "epoch": 1708, "lr": 4.073687700102241e-05}
{"train_loss": 0.12123408913612366, "global_step": 152046, "epoch": 1708, "lr": 4.073630730081776e-05}
{"train_loss": 0.08848060667514801, "global_step": 152047, "epoch": 1708, "lr": 4.073573760185849e-05}
{"train_loss": 0.12752869725227356, "global_step": 152048, "epoch": 1708, "lr": 4.07351679041447e-05}
{"train_loss": 0.08762907981872559, "global_step": 152049, "epoch": 1708, "lr": 4.073459820767644e-05}
{"train_loss": 0.10080410540103912, "global_step": 152050, "epoch": 1708, "lr": 4.073402851245382e-05}
{"train_loss": 0.12975038588047028, "global_step": 152051, "epoch": 1708, "lr": 4.0733458818476875e-05}
{"train_loss": 0.12846343219280243, "global_step": 152052, "epoch": 1708, "lr": 4.073288912574573e-05}
{"train_loss": 0.10659052431583405, "global_step": 152053, "epoch": 1708, "lr": 4.073231943426041e-05}
{"train_loss": 0.10223051905632019, "global_step": 152054, "epoch": 1708, "lr": 4.073174974402104e-05}
{"train_loss": 0.07563712447881699, "global_step": 152055, "epoch": 1708, "lr": 4.0731180055027644e-05}
{"train_loss": 0.07666002959012985, "global_step": 152056, "epoch": 1708, "lr": 4.073061036728036e-05}
{"train_loss": 0.14994704723358154, "global_step": 152057, "epoch": 1708, "lr": 4.0730040680779205e-05}
{"train_loss": 0.11789319664239883, "global_step": 152058, "epoch": 1708, "lr": 4.07294709955243e-05}
{"train_loss": 0.1233941987156868, "global_step": 152059, "epoch": 1708, "lr": 4.072890131151569e-05}
{"train_loss": 0.06488533318042755, "global_step": 152060, "epoch": 1708, "lr": 4.072833162875349e-05}
{"train_loss": 0.12436001747846603, "global_step": 152061, "epoch": 1708, "lr": 4.072776194723773e-05}
{"train_loss": 0.09779690951108932, "global_step": 152062, "epoch": 1708, "lr": 4.0727192266968525e-05}
{"train_loss": 0.09443825483322144, "global_step": 152063, "epoch": 1708, "lr": 4.0726622587945915e-05}
{"train_loss": 0.11742253601551056, "global_step": 152064, "epoch": 1708, "lr": 4.072605291017003e-05}
{"train_loss": 0.04491572827100754, "global_step": 152065, "epoch": 1708, "lr": 4.0725483233640894e-05}
{"train_loss": 0.0884270891547203, "global_step": 152066, "epoch": 1708, "lr": 4.07249135583586e-05}
{"train_loss": 0.08475332707166672, "global_step": 152067, "epoch": 1708, "lr": 4.072434388432324e-05}
{"train_loss": 0.09457498788833618, "global_step": 152068, "epoch": 1708, "lr": 4.072377421153486e-05}
{"train_loss": 0.09228379279375076, "global_step": 152069, "epoch": 1708, "lr": 4.0723204539993574e-05}
{"train_loss": 0.0463707409799099, "global_step": 152070, "epoch": 1708, "lr": 4.072263486969943e-05}
{"train_loss": 0.12953400611877441, "global_step": 152071, "epoch": 1708, "lr": 4.072206520065253e-05}
{"train_loss": 0.11218911409378052, "global_step": 152072, "epoch": 1708, "lr": 4.072149553285291e-05}
{"train_loss": 0.08102769404649734, "global_step": 152073, "epoch": 1708, "lr": 4.07209258663007e-05}
{"train_loss": 0.060068510472774506, "global_step": 152074, "epoch": 1708, "lr": 4.072035620099592e-05}
{"train_loss": 0.16082555055618286, "global_step": 152075, "epoch": 1708, "lr": 4.07197865369387e-05}
{"train_loss": 0.10414265096187592, "global_step": 152076, "epoch": 1708, "lr": 4.0719216874129076e-05}
{"train_loss": 0.11646102368831635, "global_step": 152077, "epoch": 1708, "lr": 4.071864721256715e-05}
{"train_loss": 0.12863653898239136, "global_step": 152078, "epoch": 1708, "lr": 4.0718077552252975e-05}
{"train_loss": 0.08003034442663193, "global_step": 152079, "epoch": 1708, "lr": 4.0717507893186665e-05}
{"train_loss": 0.039135806262493134, "global_step": 152080, "epoch": 1708, "lr": 4.0716938235368244e-05}
{"train_loss": 0.11930590867996216, "global_step": 152081, "epoch": 1708, "lr": 4.071636857879784e-05}
{"train_loss": 0.15695875883102417, "global_step": 152082, "epoch": 1708, "lr": 4.071579892347549e-05}
{"train_loss": 0.08115431666374207, "global_step": 152083, "epoch": 1708, "lr": 4.0715229269401314e-05}
{"train_loss": 0.1950768381357193, "global_step": 152084, "epoch": 1708, "lr": 4.071465961657534e-05}
{"train_loss": 0.10709577053785324, "global_step": 152085, "epoch": 1708, "lr": 4.071408996499767e-05}
{"train_loss": 0.08119481056928635, "global_step": 152086, "epoch": 1708, "lr": 4.071352031466838e-05}
{"train_loss": 0.1008114367723465, "global_step": 152087, "epoch": 1708, "lr": 4.071295066558754e-05}
{"train_loss": 0.08789119124412537, "global_step": 152088, "epoch": 1708, "lr": 4.0712381017755235e-05}
{"train_loss": 0.05940995737910271, "global_step": 152089, "epoch": 1708, "lr": 4.071181137117153e-05}
{"train_loss": 0.09488421678543091, "global_step": 152090, "epoch": 1708, "lr": 4.071124172583652e-05}
{"train_loss": 0.14393576979637146, "global_step": 152091, "epoch": 1708, "lr": 4.071067208175025e-05}
{"train_loss": 0.14331574738025665, "global_step": 152092, "epoch": 1708, "lr": 4.071010243891285e-05}
{"train_loss": 0.12565495073795319, "global_step": 152093, "epoch": 1708, "lr": 4.070953279732433e-05}
{"train_loss": 0.12662026286125183, "global_step": 152094, "epoch": 1708, "lr": 4.070896315698483e-05}
{"train_loss": 0.07507651299238205, "global_step": 152095, "epoch": 1708, "lr": 4.0708393517894374e-05}
{"train_loss": 0.06993478536605835, "global_step": 152096, "epoch": 1708, "lr": 4.0707823880053076e-05}
{"train_loss": 0.07236199080944061, "global_step": 152097, "epoch": 1708, "lr": 4.070725424346098e-05}
{"train_loss": 0.09977857768535614, "global_step": 152098, "epoch": 1708, "lr": 4.070668460811821e-05}
{"train_loss": 0.12553183734416962, "global_step": 152099, "epoch": 1708, "lr": 4.070611497402478e-05}
{"train_loss": 0.10007897722587157, "global_step": 152100, "epoch": 1708, "lr": 4.070554534118082e-05, "val_loss": 6.500217437744141}
{"train_loss": 0.05721880868077278, "global_step": 152101, "epoch": 1709, "lr": 4.070497570958638e-05}
{"train_loss": 0.13687404990196228, "global_step": 152102, "epoch": 1709, "lr": 4.0704406079241544e-05}
{"train_loss": 0.13841204345226288, "global_step": 152103, "epoch": 1709, "lr": 4.070383645014639e-05}
{"train_loss": 0.09648964554071426, "global_step": 152104, "epoch": 1709, "lr": 4.0703266822300975e-05}
{"train_loss": 0.15908066928386688, "global_step": 152105, "epoch": 1709, "lr": 4.070269719570542e-05}
{"train_loss": 0.1676657348871231, "global_step": 152106, "epoch": 1709, "lr": 4.0702127570359744e-05}
{"train_loss": 0.11165641248226166, "global_step": 152107, "epoch": 1709, "lr": 4.070155794626408e-05}
{"train_loss": 0.12549081444740295, "global_step": 152108, "epoch": 1709, "lr": 4.070098832341846e-05}
{"train_loss": 0.10168636590242386, "global_step": 152109, "epoch": 1709, "lr": 4.0700418701822996e-05}
{"train_loss": 0.1666429191827774, "global_step": 152110, "epoch": 1709, "lr": 4.0699849081477726e-05}
{"train_loss": 0.07918515801429749, "global_step": 152111, "epoch": 1709, "lr": 4.069927946238277e-05}
{"train_loss": 0.08168786764144897, "global_step": 152112, "epoch": 1709, "lr": 4.0698709844538166e-05}
{"train_loss": 0.19747507572174072, "global_step": 152113, "epoch": 1709, "lr": 4.069814022794402e-05}
{"train_loss": 0.12428384274244308, "global_step": 152114, "epoch": 1709, "lr": 4.0697570612600376e-05}
{"train_loss": 0.09784135967493057, "global_step": 152115, "epoch": 1709, "lr": 4.0697000998507355e-05}
{"train_loss": 0.08188294619321823, "global_step": 152116, "epoch": 1709, "lr": 4.069643138566499e-05}
{"train_loss": 0.11559274792671204, "global_step": 152117, "epoch": 1709, "lr": 4.069586177407339e-05}
{"train_loss": 0.11482338607311249, "global_step": 152118, "epoch": 1709, "lr": 4.06952921637326e-05}
{"train_loss": 0.07107959687709808, "global_step": 152119, "epoch": 1709, "lr": 4.069472255464274e-05}
{"train_loss": 0.08783906698226929, "global_step": 152120, "epoch": 1709, "lr": 4.069415294680383e-05}
{"train_loss": 0.09376706928014755, "global_step": 152121, "epoch": 1709, "lr": 4.0693583340216014e-05}
{"train_loss": 0.1156168282032013, "global_step": 152122, "epoch": 1709, "lr": 4.069301373487931e-05}
{"train_loss": 0.06684771180152893, "global_step": 152123, "epoch": 1709, "lr": 4.0692444130793804e-05}
{"train_loss": 0.13569584488868713, "global_step": 152124, "epoch": 1709, "lr": 4.069187452795962e-05}
{"train_loss": 0.09416841715574265, "global_step": 152125, "epoch": 1709, "lr": 4.0691304926376765e-05}
{"train_loss": 0.1410961151123047, "global_step": 152126, "epoch": 1709, "lr": 4.069073532604538e-05}
{"train_loss": 0.1066051721572876, "global_step": 152127, "epoch": 1709, "lr": 4.069016572696549e-05}
{"train_loss": 0.06931106746196747, "global_step": 152128, "epoch": 1709, "lr": 4.068959612913721e-05}
{"train_loss": 0.12819062173366547, "global_step": 152129, "epoch": 1709, "lr": 4.068902653256058e-05}
{"train_loss": 0.11618313193321228, "global_step": 152130, "epoch": 1709, "lr": 4.0688456937235723e-05}
{"train_loss": 0.10320716351270676, "global_step": 152131, "epoch": 1709, "lr": 4.0687887343162664e-05}
{"train_loss": 0.11353997141122818, "global_step": 152132, "epoch": 1709, "lr": 4.068731775034153e-05}
{"train_loss": 0.17545832693576813, "global_step": 152133, "epoch": 1709, "lr": 4.068674815877235e-05}
{"train_loss": 0.1406852751970291, "global_step": 152134, "epoch": 1709, "lr": 4.0686178568455246e-05}
{"train_loss": 0.11456702649593353, "global_step": 152135, "epoch": 1709, "lr": 4.068560897939025e-05}
{"train_loss": 0.1236962154507637, "global_step": 152136, "epoch": 1709, "lr": 4.068503939157748e-05}
{"train_loss": 0.07662870734930038, "global_step": 152137, "epoch": 1709, "lr": 4.068446980501697e-05}
{"train_loss": 0.0767684280872345, "global_step": 152138, "epoch": 1709, "lr": 4.0683900219708846e-05}
{"train_loss": 0.13152815401554108, "global_step": 152139, "epoch": 1709, "lr": 4.068333063565313e-05}
{"train_loss": 0.1305127888917923, "global_step": 152140, "epoch": 1709, "lr": 4.068276105284994e-05}
{"train_loss": 0.18021006882190704, "global_step": 152141, "epoch": 1709, "lr": 4.0682191471299354e-05}
{"train_loss": 0.07028935104608536, "global_step": 152142, "epoch": 1709, "lr": 4.068162189100141e-05}
{"train_loss": 0.08020882308483124, "global_step": 152143, "epoch": 1709, "lr": 4.0681052311956225e-05}
{"train_loss": 0.1346883773803711, "global_step": 152144, "epoch": 1709, "lr": 4.068048273416384e-05}
{"train_loss": 0.12509791553020477, "global_step": 152145, "epoch": 1709, "lr": 4.0679913157624374e-05}
{"train_loss": 0.13247105479240417, "global_step": 152146, "epoch": 1709, "lr": 4.067934358233786e-05}
{"train_loss": 0.07687989622354507, "global_step": 152147, "epoch": 1709, "lr": 4.067877400830441e-05}
{"train_loss": 0.08487574756145477, "global_step": 152148, "epoch": 1709, "lr": 4.067820443552407e-05}
{"train_loss": 0.21140336990356445, "global_step": 152149, "epoch": 1709, "lr": 4.067763486399695e-05}
{"train_loss": 0.16381126642227173, "global_step": 152150, "epoch": 1709, "lr": 4.067706529372309e-05}
{"train_loss": 0.16884174942970276, "global_step": 152151, "epoch": 1709, "lr": 4.06764957247026e-05}
{"train_loss": 0.0740189477801323, "global_step": 152152, "epoch": 1709, "lr": 4.0675926156935526e-05}
{"train_loss": 0.11909843236207962, "global_step": 152153, "epoch": 1709, "lr": 4.0675356590421985e-05}
{"train_loss": 0.0531952902674675, "global_step": 152154, "epoch": 1709, "lr": 4.0674787025162e-05}
{"train_loss": 0.10133858025074005, "global_step": 152155, "epoch": 1709, "lr": 4.06742174611557e-05}
{"train_loss": 0.06540872156620026, "global_step": 152156, "epoch": 1709, "lr": 4.0673647898403115e-05}
{"train_loss": 0.04891728609800339, "global_step": 152157, "epoch": 1709, "lr": 4.067307833690437e-05}
{"train_loss": 0.06859398633241653, "global_step": 152158, "epoch": 1709, "lr": 4.067250877665949e-05}
{"train_loss": 0.2022705078125, "global_step": 152159, "epoch": 1709, "lr": 4.067193921766859e-05}
{"train_loss": 0.12105707824230194, "global_step": 152160, "epoch": 1709, "lr": 4.067136965993173e-05}
{"train_loss": 0.08577658236026764, "global_step": 152161, "epoch": 1709, "lr": 4.0670800103449e-05}
{"train_loss": 0.11431339383125305, "global_step": 152162, "epoch": 1709, "lr": 4.067023054822047e-05}
{"train_loss": 0.12160158157348633, "global_step": 152163, "epoch": 1709, "lr": 4.0669660994246195e-05}
{"train_loss": 0.14330284297466278, "global_step": 152164, "epoch": 1709, "lr": 4.0669091441526293e-05}
{"train_loss": 0.17642256617546082, "global_step": 152165, "epoch": 1709, "lr": 4.066852189006079e-05}
{"train_loss": 0.14146362245082855, "global_step": 152166, "epoch": 1709, "lr": 4.066795233984982e-05}
{"train_loss": 0.10790408402681351, "global_step": 152167, "epoch": 1709, "lr": 4.0667382790893406e-05}
{"train_loss": 0.047838836908340454, "global_step": 152168, "epoch": 1709, "lr": 4.066681324319167e-05}
{"train_loss": 0.07168207317590714, "global_step": 152169, "epoch": 1709, "lr": 4.066624369674464e-05}
{"train_loss": 0.05252634733915329, "global_step": 152170, "epoch": 1709, "lr": 4.0665674151552444e-05}
{"train_loss": 0.04975526034832001, "global_step": 152171, "epoch": 1709, "lr": 4.066510460761512e-05}
{"train_loss": 0.1261346936225891, "global_step": 152172, "epoch": 1709, "lr": 4.066453506493277e-05}
{"train_loss": 0.13425669074058533, "global_step": 152173, "epoch": 1709, "lr": 4.066396552350545e-05}
{"train_loss": 0.09807075560092926, "global_step": 152174, "epoch": 1709, "lr": 4.066339598333326e-05}
{"train_loss": 0.11240594834089279, "global_step": 152175, "epoch": 1709, "lr": 4.066282644441624e-05}
{"train_loss": 0.1377728134393692, "global_step": 152176, "epoch": 1709, "lr": 4.06622569067545e-05}
{"train_loss": 0.07018056511878967, "global_step": 152177, "epoch": 1709, "lr": 4.066168737034811e-05}
{"train_loss": 0.1625652015209198, "global_step": 152178, "epoch": 1709, "lr": 4.066111783519714e-05}
{"train_loss": 0.11252051591873169, "global_step": 152179, "epoch": 1709, "lr": 4.0660548301301674e-05}
{"train_loss": 0.11807674169540405, "global_step": 152180, "epoch": 1709, "lr": 4.065997876866178e-05}
{"train_loss": 0.08767151832580566, "global_step": 152181, "epoch": 1709, "lr": 4.065940923727755e-05}
{"train_loss": 0.0920286774635315, "global_step": 152182, "epoch": 1709, "lr": 4.065883970714902e-05}
{"train_loss": 0.12778198719024658, "global_step": 152183, "epoch": 1709, "lr": 4.065827017827633e-05}
{"train_loss": 0.08300795406103134, "global_step": 152184, "epoch": 1709, "lr": 4.065770065065949e-05}
{"train_loss": 0.09705471247434616, "global_step": 152185, "epoch": 1709, "lr": 4.065713112429863e-05}
{"train_loss": 0.09552999585866928, "global_step": 152186, "epoch": 1709, "lr": 4.065656159919378e-05}
{"train_loss": 0.10194210708141327, "global_step": 152187, "epoch": 1709, "lr": 4.065599207534507e-05}
{"train_loss": 0.14301984012126923, "global_step": 152188, "epoch": 1709, "lr": 4.065542255275252e-05}
{"train_loss": 0.11209679893061017, "global_step": 152189, "epoch": 1709, "lr": 4.065485303141626e-05, "val_loss": 6.2635722160339355}
{"train_loss": 0.1720956712961197, "global_step": 152190, "epoch": 1710, "lr": 4.065428351133632e-05}
{"train_loss": 0.11830504983663559, "global_step": 152191, "epoch": 1710, "lr": 4.0653713992512823e-05}
{"train_loss": 0.09301398694515228, "global_step": 152192, "epoch": 1710, "lr": 4.065314447494579e-05}
{"train_loss": 0.07092776149511337, "global_step": 152193, "epoch": 1710, "lr": 4.065257495863536e-05}
{"train_loss": 0.07453259825706482, "global_step": 152194, "epoch": 1710, "lr": 4.065200544358154e-05}
{"train_loss": 0.10016936808824539, "global_step": 152195, "epoch": 1710, "lr": 4.065143592978447e-05}
{"train_loss": 0.07654988765716553, "global_step": 152196, "epoch": 1710, "lr": 4.0650866417244196e-05}
{"train_loss": 0.10181430727243423, "global_step": 152197, "epoch": 1710, "lr": 4.0650296905960794e-05}
{"train_loss": 0.05756566300988197, "global_step": 152198, "epoch": 1710, "lr": 4.064972739593435e-05}
{"train_loss": 0.07126926630735397, "global_step": 152199, "epoch": 1710, "lr": 4.064915788716494e-05}
{"train_loss": 0.09909319877624512, "global_step": 152200, "epoch": 1710, "lr": 4.064858837965264e-05}
{"train_loss": 0.06694948673248291, "global_step": 152201, "epoch": 1710, "lr": 4.064801887339751e-05}
{"train_loss": 0.11978008598089218, "global_step": 152202, "epoch": 1710, "lr": 4.0647449368399654e-05}
{"train_loss": 0.06701536476612091, "global_step": 152203, "epoch": 1710, "lr": 4.0646879864659115e-05}
{"train_loss": 0.0792137011885643, "global_step": 152204, "epoch": 1710, "lr": 4.064631036217601e-05}
{"train_loss": 0.1545577049255371, "global_step": 152205, "epoch": 1710, "lr": 4.0645740860950375e-05}
{"train_loss": 0.10282277315855026, "global_step": 152206, "epoch": 1710, "lr": 4.0645171360982324e-05}
{"train_loss": 0.1164855808019638, "global_step": 152207, "epoch": 1710, "lr": 4.06446018622719e-05}
{"train_loss": 0.08168113976716995, "global_step": 152208, "epoch": 1710, "lr": 4.064403236481921e-05}
{"train_loss": 0.07870381325483322, "global_step": 152209, "epoch": 1710, "lr": 4.06434628686243e-05}
{"train_loss": 0.05693197250366211, "global_step": 152210, "epoch": 1710, "lr": 4.0642893373687285e-05}
{"train_loss": 0.07096259295940399, "global_step": 152211, "epoch": 1710, "lr": 4.0642323880008196e-05}
{"train_loss": 0.08402194082736969, "global_step": 152212, "epoch": 1710, "lr": 4.064175438758715e-05}
{"train_loss": 0.10216397047042847, "global_step": 152213, "epoch": 1710, "lr": 4.0641184896424187e-05}
{"train_loss": 0.14311324059963226, "global_step": 152214, "epoch": 1710, "lr": 4.064061540651941e-05}
{"train_loss": 0.16385556757450104, "global_step": 152215, "epoch": 1710, "lr": 4.0640045917872895e-05}
{"train_loss": 0.0726204514503479, "global_step": 152216, "epoch": 1710, "lr": 4.06394764304847e-05}
{"train_loss": 0.1046251505613327, "global_step": 152217, "epoch": 1710, "lr": 4.063890694435492e-05}
{"train_loss": 0.09421837329864502, "global_step": 152218, "epoch": 1710, "lr": 4.0638337459483635e-05}
{"train_loss": 0.09952738881111145, "global_step": 152219, "epoch": 1710, "lr": 4.063776797587091e-05}
{"train_loss": 0.0925791785120964, "global_step": 152220, "epoch": 1710, "lr": 4.063719849351681e-05}
{"train_loss": 0.10089181363582611, "global_step": 152221, "epoch": 1710, "lr": 4.0636629012421436e-05}
{"train_loss": 0.07179386168718338, "global_step": 152222, "epoch": 1710, "lr": 4.063605953258484e-05}
{"train_loss": 0.0977991446852684, "global_step": 152223, "epoch": 1710, "lr": 4.063549005400713e-05}
{"train_loss": 0.09640029817819595, "global_step": 152224, "epoch": 1710, "lr": 4.063492057668834e-05}
{"train_loss": 0.15640679001808167, "global_step": 152225, "epoch": 1710, "lr": 4.0634351100628596e-05}
{"train_loss": 0.10583220422267914, "global_step": 152226, "epoch": 1710, "lr": 4.063378162582793e-05}
{"train_loss": 0.13215003907680511, "global_step": 152227, "epoch": 1710, "lr": 4.063321215228645e-05}
{"train_loss": 0.09473700076341629, "global_step": 152228, "epoch": 1710, "lr": 4.0632642680004206e-05}
{"train_loss": 0.0806436613202095, "global_step": 152229, "epoch": 1710, "lr": 4.063207320898131e-05}
{"train_loss": 0.119293712079525, "global_step": 152230, "epoch": 1710, "lr": 4.0631503739217797e-05}
{"train_loss": 0.10201781243085861, "global_step": 152231, "epoch": 1710, "lr": 4.063093427071376e-05}
{"train_loss": 0.09477991610765457, "global_step": 152232, "epoch": 1710, "lr": 4.063036480346931e-05}
{"train_loss": 0.13988272845745087, "global_step": 152233, "epoch": 1710, "lr": 4.062979533748447e-05}
{"train_loss": 0.11656013876199722, "global_step": 152234, "epoch": 1710, "lr": 4.0629225872759346e-05}
{"train_loss": 0.1423196941614151, "global_step": 152235, "epoch": 1710, "lr": 4.062865640929401e-05}
{"train_loss": 0.0925421193242073, "global_step": 152236, "epoch": 1710, "lr": 4.062808694708853e-05}
{"train_loss": 0.10204992443323135, "global_step": 152237, "epoch": 1710, "lr": 4.0627517486143e-05}
{"train_loss": 0.13571473956108093, "global_step": 152238, "epoch": 1710, "lr": 4.062694802645749e-05}
{"train_loss": 0.050894733518362045, "global_step": 152239, "epoch": 1710, "lr": 4.062637856803205e-05}
{"train_loss": 0.11077512055635452, "global_step": 152240, "epoch": 1710, "lr": 4.0625809110866805e-05}
{"train_loss": 0.052685242146253586, "global_step": 152241, "epoch": 1710, "lr": 4.062523965496179e-05}
{"train_loss": 0.15426695346832275, "global_step": 152242, "epoch": 1710, "lr": 4.062467020031711e-05}
{"train_loss": 0.07297050207853317, "global_step": 152243, "epoch": 1710, "lr": 4.0624100746932806e-05}
{"train_loss": 0.12101133167743683, "global_step": 152244, "epoch": 1710, "lr": 4.0623531294809004e-05}
{"train_loss": 0.08806999772787094, "global_step": 152245, "epoch": 1710, "lr": 4.062296184394573e-05}
{"train_loss": 0.08559531718492508, "global_step": 152246, "epoch": 1710, "lr": 4.0622392394343104e-05}
{"train_loss": 0.12681594491004944, "global_step": 152247, "epoch": 1710, "lr": 4.062182294600116e-05}
{"train_loss": 0.09250794351100922, "global_step": 152248, "epoch": 1710, "lr": 4.0621253498920026e-05}
{"train_loss": 0.17589125037193298, "global_step": 152249, "epoch": 1710, "lr": 4.062068405309973e-05}
{"train_loss": 0.060844726860523224, "global_step": 152250, "epoch": 1710, "lr": 4.062011460854036e-05}
{"train_loss": 0.06265129148960114, "global_step": 152251, "epoch": 1710, "lr": 4.061954516524203e-05}
{"train_loss": 0.07539771497249603, "global_step": 152252, "epoch": 1710, "lr": 4.061897572320477e-05}
{"train_loss": 0.10807378590106964, "global_step": 152253, "epoch": 1710, "lr": 4.0618406282428676e-05}
{"train_loss": 0.13790778815746307, "global_step": 152254, "epoch": 1710, "lr": 4.061783684291383e-05}
{"train_loss": 0.06762025505304337, "global_step": 152255, "epoch": 1710, "lr": 4.06172674046603e-05}
{"train_loss": 0.09354651719331741, "global_step": 152256, "epoch": 1710, "lr": 4.061669796766816e-05}
{"train_loss": 0.12620365619659424, "global_step": 152257, "epoch": 1710, "lr": 4.06161285319375e-05}
{"train_loss": 0.06886603683233261, "global_step": 152258, "epoch": 1710, "lr": 4.0615559097468365e-05}
{"train_loss": 0.09500718861818314, "global_step": 152259, "epoch": 1710, "lr": 4.0614989664260876e-05}
{"train_loss": 0.06794038414955139, "global_step": 152260, "epoch": 1710, "lr": 4.061442023231507e-05}
{"train_loss": 0.10841476172208786, "global_step": 152261, "epoch": 1710, "lr": 4.061385080163106e-05}
{"train_loss": 0.1253431886434555, "global_step": 152262, "epoch": 1710, "lr": 4.0613281372208884e-05}
{"train_loss": 0.1314731240272522, "global_step": 152263, "epoch": 1710, "lr": 4.0612711944048654e-05}
{"train_loss": 0.04071119800209999, "global_step": 152264, "epoch": 1710, "lr": 4.0612142517150415e-05}
{"train_loss": 0.11961640417575836, "global_step": 152265, "epoch": 1710, "lr": 4.0611573091514275e-05}
{"train_loss": 0.11812486499547958, "global_step": 152266, "epoch": 1710, "lr": 4.061100366714028e-05}
{"train_loss": 0.11353269964456558, "global_step": 152267, "epoch": 1710, "lr": 4.061043424402852e-05}
{"train_loss": 0.12090258300304413, "global_step": 152268, "epoch": 1710, "lr": 4.060986482217909e-05}
{"train_loss": 0.1119803935289383, "global_step": 152269, "epoch": 1710, "lr": 4.0609295401592026e-05}
{"train_loss": 0.05636635050177574, "global_step": 152270, "epoch": 1710, "lr": 4.060872598226745e-05}
{"train_loss": 0.14442725479602814, "global_step": 152271, "epoch": 1710, "lr": 4.0608156564205395e-05}
{"train_loss": 0.14044319093227386, "global_step": 152272, "epoch": 1710, "lr": 4.0607587147405984e-05}
{"train_loss": 0.07321662455797195, "global_step": 152273, "epoch": 1710, "lr": 4.0607017731869256e-05}
{"train_loss": 0.10843843966722488, "global_step": 152274, "epoch": 1710, "lr": 4.0606448317595304e-05}
{"train_loss": 0.10384674370288849, "global_step": 152275, "epoch": 1710, "lr": 4.060587890458419e-05}
{"train_loss": 0.067917101085186, "global_step": 152276, "epoch": 1710, "lr": 4.060530949283601e-05}
{"train_loss": 0.13827256858348846, "global_step": 152277, "epoch": 1710, "lr": 4.060474008235083e-05}
{"train_loss": 0.10080310739995388, "global_step": 152278, "epoch": 1710, "lr": 4.060417067312874e-05, "val_loss": 6.4503068923950195, "train_action_mse_error": 30.128456115722656}
{"train_loss": 0.15613411366939545, "global_step": 152279, "epoch": 1711, "lr": 4.060360126516978e-05}
{"train_loss": 0.08991653472185135, "global_step": 152280, "epoch": 1711, "lr": 4.0603031858474074e-05}
{"train_loss": 0.0915934145450592, "global_step": 152281, "epoch": 1711, "lr": 4.0602462453041656e-05}
{"train_loss": 0.07747851312160492, "global_step": 152282, "epoch": 1711, "lr": 4.060189304887264e-05}
{"train_loss": 0.14831946790218353, "global_step": 152283, "epoch": 1711, "lr": 4.060132364596707e-05}
{"train_loss": 0.09979847818613052, "global_step": 152284, "epoch": 1711, "lr": 4.060075424432505e-05}
{"train_loss": 0.11947469413280487, "global_step": 152285, "epoch": 1711, "lr": 4.060018484394663e-05}
{"train_loss": 0.07372206449508667, "global_step": 152286, "epoch": 1711, "lr": 4.059961544483189e-05}
{"train_loss": 0.12954986095428467, "global_step": 152287, "epoch": 1711, "lr": 4.0599046046980955e-05}
{"train_loss": 0.056949593126773834, "global_step": 152288, "epoch": 1711, "lr": 4.059847665039384e-05}
{"train_loss": 0.12311269342899323, "global_step": 152289, "epoch": 1711, "lr": 4.0597907255070656e-05}
{"train_loss": 0.10153992474079132, "global_step": 152290, "epoch": 1711, "lr": 4.059733786101145e-05}
{"train_loss": 0.10499245673418045, "global_step": 152291, "epoch": 1711, "lr": 4.059676846821634e-05}
{"train_loss": 0.10279116779565811, "global_step": 152292, "epoch": 1711, "lr": 4.059619907668537e-05}
{"train_loss": 0.07553255558013916, "global_step": 152293, "epoch": 1711, "lr": 4.059562968641862e-05}
{"train_loss": 0.09285929799079895, "global_step": 152294, "epoch": 1711, "lr": 4.059506029741619e-05}
{"train_loss": 0.09406664222478867, "global_step": 152295, "epoch": 1711, "lr": 4.059449090967813e-05}
{"train_loss": 0.09414608776569366, "global_step": 152296, "epoch": 1711, "lr": 4.0593921523204515e-05}
{"train_loss": 0.06524072587490082, "global_step": 152297, "epoch": 1711, "lr": 4.0593352137995455e-05}
{"train_loss": 0.09138954430818558, "global_step": 152298, "epoch": 1711, "lr": 4.059278275405098e-05}
{"train_loss": 0.1400827020406723, "global_step": 152299, "epoch": 1711, "lr": 4.059221337137121e-05}
{"train_loss": 0.14889152348041534, "global_step": 152300, "epoch": 1711, "lr": 4.059164398995619e-05}
{"train_loss": 0.13857929408550262, "global_step": 152301, "epoch": 1711, "lr": 4.059107460980602e-05}
{"train_loss": 0.07011447846889496, "global_step": 152302, "epoch": 1711, "lr": 4.059050523092075e-05}
{"train_loss": 0.11256294697523117, "global_step": 152303, "epoch": 1711, "lr": 4.058993585330047e-05}
{"train_loss": 0.1212191954255104, "global_step": 152304, "epoch": 1711, "lr": 4.058936647694528e-05}
{"train_loss": 0.07758469134569168, "global_step": 152305, "epoch": 1711, "lr": 4.0588797101855216e-05}
{"train_loss": 0.10571769624948502, "global_step": 152306, "epoch": 1711, "lr": 4.058822772803039e-05}
{"train_loss": 0.12657809257507324, "global_step": 152307, "epoch": 1711, "lr": 4.0587658355470844e-05}
{"train_loss": 0.13098779320716858, "global_step": 152308, "epoch": 1711, "lr": 4.058708898417669e-05}
{"train_loss": 0.05525229871273041, "global_step": 152309, "epoch": 1711, "lr": 4.058651961414797e-05}
{"train_loss": 0.0739608034491539, "global_step": 152310, "epoch": 1711, "lr": 4.05859502453848e-05}
{"train_loss": 0.10308416187763214, "global_step": 152311, "epoch": 1711, "lr": 4.058538087788721e-05}
{"train_loss": 0.12716829776763916, "global_step": 152312, "epoch": 1711, "lr": 4.0584811511655315e-05}
{"train_loss": 0.10994067043066025, "global_step": 152313, "epoch": 1711, "lr": 4.058424214668918e-05}
{"train_loss": 0.0813598781824112, "global_step": 152314, "epoch": 1711, "lr": 4.0583672782988867e-05}
{"train_loss": 0.07834099978208542, "global_step": 152315, "epoch": 1711, "lr": 4.058310342055448e-05}
{"train_loss": 0.11508134007453918, "global_step": 152316, "epoch": 1711, "lr": 4.0582534059386076e-05}
{"train_loss": 0.11043798178434372, "global_step": 152317, "epoch": 1711, "lr": 4.058196469948372e-05}
{"train_loss": 0.08329524844884872, "global_step": 152318, "epoch": 1711, "lr": 4.0581395340847525e-05}
{"train_loss": 0.05605163425207138, "global_step": 152319, "epoch": 1711, "lr": 4.058082598347752e-05}
{"train_loss": 0.04014008119702339, "global_step": 152320, "epoch": 1711, "lr": 4.058025662737384e-05}
{"train_loss": 0.10485636442899704, "global_step": 152321, "epoch": 1711, "lr": 4.05796872725365e-05}
{"train_loss": 0.08122219890356064, "global_step": 152322, "epoch": 1711, "lr": 4.0579117918965607e-05}
{"train_loss": 0.20612649619579315, "global_step": 152323, "epoch": 1711, "lr": 4.057854856666126e-05}
{"train_loss": 0.10733164846897125, "global_step": 152324, "epoch": 1711, "lr": 4.05779792156235e-05}
{"train_loss": 0.07703596353530884, "global_step": 152325, "epoch": 1711, "lr": 4.057740986585242e-05}
{"train_loss": 0.05084855854511261, "global_step": 152326, "epoch": 1711, "lr": 4.0576840517348083e-05}
{"train_loss": 0.17422373592853546, "global_step": 152327, "epoch": 1711, "lr": 4.057627117011059e-05}
{"train_loss": 0.11587613821029663, "global_step": 152328, "epoch": 1711, "lr": 4.0575701824139986e-05}
{"train_loss": 0.08005791157484055, "global_step": 152329, "epoch": 1711, "lr": 4.057513247943639e-05}
{"train_loss": 0.10383420437574387, "global_step": 152330, "epoch": 1711, "lr": 4.057456313599983e-05}
{"train_loss": 0.10289396345615387, "global_step": 152331, "epoch": 1711, "lr": 4.057399379383041e-05}
{"train_loss": 0.050865329802036285, "global_step": 152332, "epoch": 1711, "lr": 4.05734244529282e-05}
{"train_loss": 0.16577939689159393, "global_step": 152333, "epoch": 1711, "lr": 4.057285511329328e-05}
{"train_loss": 0.07588968425989151, "global_step": 152334, "epoch": 1711, "lr": 4.057228577492572e-05}
{"train_loss": 0.0792965367436409, "global_step": 152335, "epoch": 1711, "lr": 4.0571716437825625e-05}
{"train_loss": 0.057403285056352615, "global_step": 152336, "epoch": 1711, "lr": 4.057114710199301e-05}
{"train_loss": 0.11252625286579132, "global_step": 152337, "epoch": 1711, "lr": 4.057057776742802e-05}
{"train_loss": 0.19073104858398438, "global_step": 152338, "epoch": 1711, "lr": 4.057000843413068e-05}
{"train_loss": 0.09789583086967468, "global_step": 152339, "epoch": 1711, "lr": 4.0569439102101106e-05}
{"train_loss": 0.06741060316562653, "global_step": 152340, "epoch": 1711, "lr": 4.056886977133934e-05}
{"train_loss": 0.11728495359420776, "global_step": 152341, "epoch": 1711, "lr": 4.056830044184547e-05}
{"train_loss": 0.10293012857437134, "global_step": 152342, "epoch": 1711, "lr": 4.05677311136196e-05}
{"train_loss": 0.09741339832544327, "global_step": 152343, "epoch": 1711, "lr": 4.056716178666176e-05}
{"train_loss": 0.12504370510578156, "global_step": 152344, "epoch": 1711, "lr": 4.056659246097207e-05}
{"train_loss": 0.12502029538154602, "global_step": 152345, "epoch": 1711, "lr": 4.056602313655057e-05}
{"train_loss": 0.16618596017360687, "global_step": 152346, "epoch": 1711, "lr": 4.056545381339737e-05}
{"train_loss": 0.10723677277565002, "global_step": 152347, "epoch": 1711, "lr": 4.056488449151251e-05}
{"train_loss": 0.14437879621982574, "global_step": 152348, "epoch": 1711, "lr": 4.056431517089611e-05}
{"train_loss": 0.13297630846500397, "global_step": 152349, "epoch": 1711, "lr": 4.056374585154821e-05}
{"train_loss": 0.08298385143280029, "global_step": 152350, "epoch": 1711, "lr": 4.05631765334689e-05}
{"train_loss": 0.07945023477077484, "global_step": 152351, "epoch": 1711, "lr": 4.056260721665826e-05}
{"train_loss": 0.0950341746211052, "global_step": 152352, "epoch": 1711, "lr": 4.056203790111636e-05}
{"train_loss": 0.08324483782052994, "global_step": 152353, "epoch": 1711, "lr": 4.056146858684327e-05}
{"train_loss": 0.12849847972393036, "global_step": 152354, "epoch": 1711, "lr": 4.0560899273839095e-05}
{"train_loss": 0.10145255923271179, "global_step": 152355, "epoch": 1711, "lr": 4.056032996210387e-05}
{"train_loss": 0.1252240240573883, "global_step": 152356, "epoch": 1711, "lr": 4.0559760651637715e-05}
{"train_loss": 0.10738898813724518, "global_step": 152357, "epoch": 1711, "lr": 4.055919134244066e-05}
{"train_loss": 0.07195280492305756, "global_step": 152358, "epoch": 1711, "lr": 4.055862203451281e-05}
{"train_loss": 0.09043415635824203, "global_step": 152359, "epoch": 1711, "lr": 4.0558052727854255e-05}
{"train_loss": 0.07832566648721695, "global_step": 152360, "epoch": 1711, "lr": 4.055748342246504e-05}
{"train_loss": 0.17253080010414124, "global_step": 152361, "epoch": 1711, "lr": 4.055691411834528e-05}
{"train_loss": 0.13058851659297943, "global_step": 152362, "epoch": 1711, "lr": 4.0556344815494994e-05}
{"train_loss": 0.14608389139175415, "global_step": 152363, "epoch": 1711, "lr": 4.0555775513914324e-05}
{"train_loss": 0.07277620583772659, "global_step": 152364, "epoch": 1711, "lr": 4.055520621360329e-05}
{"train_loss": 0.10010547190904617, "global_step": 152365, "epoch": 1711, "lr": 4.055463691456202e-05}
{"train_loss": 0.10550755262374878, "global_step": 152366, "epoch": 1711, "lr": 4.0554067616790535e-05}
{"train_loss": 0.1043292772903871, "global_step": 152367, "epoch": 1711, "lr": 4.055349832028896e-05, "val_loss": 6.374935626983643}
{"train_loss": 0.07962672412395477, "global_step": 152368, "epoch": 1712, "lr": 4.055292902505735e-05}
{"train_loss": 0.10802583396434784, "global_step": 152369, "epoch": 1712, "lr": 4.0552359731095776e-05}
{"train_loss": 0.12475433945655823, "global_step": 152370, "epoch": 1712, "lr": 4.055179043840433e-05}
{"train_loss": 0.09163977205753326, "global_step": 152371, "epoch": 1712, "lr": 4.0551221146983076e-05}
{"train_loss": 0.10912702977657318, "global_step": 152372, "epoch": 1712, "lr": 4.055065185683209e-05}
{"train_loss": 0.0852050706744194, "global_step": 152373, "epoch": 1712, "lr": 4.055008256795146e-05}
{"train_loss": 0.07645575702190399, "global_step": 152374, "epoch": 1712, "lr": 4.054951328034126e-05}
{"train_loss": 0.05443907156586647, "global_step": 152375, "epoch": 1712, "lr": 4.054894399400157e-05}
{"train_loss": 0.09000368416309357, "global_step": 152376, "epoch": 1712, "lr": 4.054837470893243e-05}
{"train_loss": 0.12802691757678986, "global_step": 152377, "epoch": 1712, "lr": 4.054780542513395e-05}
{"train_loss": 0.13140594959259033, "global_step": 152378, "epoch": 1712, "lr": 4.054723614260624e-05}
{"train_loss": 0.07678792625665665, "global_step": 152379, "epoch": 1712, "lr": 4.05466668613493e-05}
{"train_loss": 0.09690157324075699, "global_step": 152380, "epoch": 1712, "lr": 4.0546097581363265e-05}
{"train_loss": 0.06491446495056152, "global_step": 152381, "epoch": 1712, "lr": 4.054552830264818e-05}
{"train_loss": 0.066031314432621, "global_step": 152382, "epoch": 1712, "lr": 4.054495902520416e-05}
{"train_loss": 0.09120845049619675, "global_step": 152383, "epoch": 1712, "lr": 4.0544389749031216e-05}
{"train_loss": 0.05736340209841728, "global_step": 152384, "epoch": 1712, "lr": 4.0543820474129495e-05}
{"train_loss": 0.10197082906961441, "global_step": 152385, "epoch": 1712, "lr": 4.0543251200499015e-05}
{"train_loss": 0.13128149509429932, "global_step": 152386, "epoch": 1712, "lr": 4.054268192813991e-05}
{"train_loss": 0.09719482809305191, "global_step": 152387, "epoch": 1712, "lr": 4.05421126570522e-05}
{"train_loss": 0.13050876557826996, "global_step": 152388, "epoch": 1712, "lr": 4.054154338723601e-05}
{"train_loss": 0.10863278061151505, "global_step": 152389, "epoch": 1712, "lr": 4.054097411869138e-05}
{"train_loss": 0.15942299365997314, "global_step": 152390, "epoch": 1712, "lr": 4.05404048514184e-05}
{"train_loss": 0.12016570568084717, "global_step": 152391, "epoch": 1712, "lr": 4.053983558541715e-05}
{"train_loss": 0.09978687018156052, "global_step": 152392, "epoch": 1712, "lr": 4.053926632068771e-05}
{"train_loss": 0.10682228207588196, "global_step": 152393, "epoch": 1712, "lr": 4.0538697057230144e-05}
{"train_loss": 0.06062104552984238, "global_step": 152394, "epoch": 1712, "lr": 4.053812779504452e-05}
{"train_loss": 0.10811956226825714, "global_step": 152395, "epoch": 1712, "lr": 4.053755853413096e-05}
{"train_loss": 0.13955427706241608, "global_step": 152396, "epoch": 1712, "lr": 4.053698927448948e-05}
{"train_loss": 0.07185167819261551, "global_step": 152397, "epoch": 1712, "lr": 4.053642001612022e-05}
{"train_loss": 0.09295008331537247, "global_step": 152398, "epoch": 1712, "lr": 4.053585075902319e-05}
{"train_loss": 0.07966343313455582, "global_step": 152399, "epoch": 1712, "lr": 4.053528150319852e-05}
{"train_loss": 0.10151880234479904, "global_step": 152400, "epoch": 1712, "lr": 4.0534712248646254e-05}
{"train_loss": 0.10324450582265854, "global_step": 152401, "epoch": 1712, "lr": 4.05341429953665e-05}
{"train_loss": 0.10617360472679138, "global_step": 152402, "epoch": 1712, "lr": 4.0533573743359286e-05}
{"train_loss": 0.14575408399105072, "global_step": 152403, "epoch": 1712, "lr": 4.053300449262475e-05}
{"train_loss": 0.0911312997341156, "global_step": 152404, "epoch": 1712, "lr": 4.0532435243162906e-05}
{"train_loss": 0.15476877987384796, "global_step": 152405, "epoch": 1712, "lr": 4.053186599497388e-05}
{"train_loss": 0.08658094704151154, "global_step": 152406, "epoch": 1712, "lr": 4.0531296748057715e-05}
{"train_loss": 0.09459688514471054, "global_step": 152407, "epoch": 1712, "lr": 4.0530727502414525e-05}
{"train_loss": 0.08449945598840714, "global_step": 152408, "epoch": 1712, "lr": 4.053015825804435e-05}
{"train_loss": 0.07652729004621506, "global_step": 152409, "epoch": 1712, "lr": 4.052958901494728e-05}
{"train_loss": 0.07528223842382431, "global_step": 152410, "epoch": 1712, "lr": 4.052901977312338e-05}
{"train_loss": 0.0746002197265625, "global_step": 152411, "epoch": 1712, "lr": 4.052845053257275e-05}
{"train_loss": 0.13450345396995544, "global_step": 152412, "epoch": 1712, "lr": 4.0527881293295455e-05}
{"train_loss": 0.15523260831832886, "global_step": 152413, "epoch": 1712, "lr": 4.052731205529155e-05}
{"train_loss": 0.1503060758113861, "global_step": 152414, "epoch": 1712, "lr": 4.052674281856117e-05}
{"train_loss": 0.11996027827262878, "global_step": 152415, "epoch": 1712, "lr": 4.052617358310432e-05}
{"train_loss": 0.07982689887285233, "global_step": 152416, "epoch": 1712, "lr": 4.0525604348921136e-05}
{"train_loss": 0.10477617383003235, "global_step": 152417, "epoch": 1712, "lr": 4.052503511601164e-05}
{"train_loss": 0.09086205810308456, "global_step": 152418, "epoch": 1712, "lr": 4.052446588437597e-05}
{"train_loss": 0.11778128147125244, "global_step": 152419, "epoch": 1712, "lr": 4.052389665401415e-05}
{"train_loss": 0.17628848552703857, "global_step": 152420, "epoch": 1712, "lr": 4.052332742492629e-05}
{"train_loss": 0.05940661579370499, "global_step": 152421, "epoch": 1712, "lr": 4.052275819711243e-05}
{"train_loss": 0.10634051263332367, "global_step": 152422, "epoch": 1712, "lr": 4.05221889705727e-05}
{"train_loss": 0.122494637966156, "global_step": 152423, "epoch": 1712, "lr": 4.0521619745307116e-05}
{"train_loss": 0.118761345744133, "global_step": 152424, "epoch": 1712, "lr": 4.052105052131581e-05}
{"train_loss": 0.11245293915271759, "global_step": 152425, "epoch": 1712, "lr": 4.052048129859881e-05}
{"train_loss": 0.14792561531066895, "global_step": 152426, "epoch": 1712, "lr": 4.051991207715624e-05}
{"train_loss": 0.053122904151678085, "global_step": 152427, "epoch": 1712, "lr": 4.0519342856988144e-05}
{"train_loss": 0.05244099348783493, "global_step": 152428, "epoch": 1712, "lr": 4.051877363809461e-05}
{"train_loss": 0.08682426810264587, "global_step": 152429, "epoch": 1712, "lr": 4.05182044204757e-05}
{"train_loss": 0.09333427995443344, "global_step": 152430, "epoch": 1712, "lr": 4.0517635204131493e-05}
{"train_loss": 0.09857655316591263, "global_step": 152431, "epoch": 1712, "lr": 4.05170659890621e-05}
{"train_loss": 0.14034008979797363, "global_step": 152432, "epoch": 1712, "lr": 4.0516496775267545e-05}
{"train_loss": 0.06088963896036148, "global_step": 152433, "epoch": 1712, "lr": 4.0515927562747965e-05}
{"train_loss": 0.06828427314758301, "global_step": 152434, "epoch": 1712, "lr": 4.0515358351503366e-05}
{"train_loss": 0.12539586424827576, "global_step": 152435, "epoch": 1712, "lr": 4.051478914153389e-05}
{"train_loss": 0.07567022740840912, "global_step": 152436, "epoch": 1712, "lr": 4.051421993283957e-05}
{"train_loss": 0.13995903730392456, "global_step": 152437, "epoch": 1712, "lr": 4.051365072542051e-05}
{"train_loss": 0.09323268383741379, "global_step": 152438, "epoch": 1712, "lr": 4.051308151927676e-05}
{"train_loss": 0.07054489105939865, "global_step": 152439, "epoch": 1712, "lr": 4.051251231440843e-05}
{"train_loss": 0.05994865670800209, "global_step": 152440, "epoch": 1712, "lr": 4.051194311081555e-05}
{"train_loss": 0.1320895403623581, "global_step": 152441, "epoch": 1712, "lr": 4.0511373908498245e-05}
{"train_loss": 0.08575209230184555, "global_step": 152442, "epoch": 1712, "lr": 4.051080470745655e-05}
{"train_loss": 0.10641184449195862, "global_step": 152443, "epoch": 1712, "lr": 4.051023550769059e-05}
{"train_loss": 0.1627921164035797, "global_step": 152444, "epoch": 1712, "lr": 4.050966630920038e-05}
{"train_loss": 0.10755444318056107, "global_step": 152445, "epoch": 1712, "lr": 4.050909711198606e-05}
{"train_loss": 0.07647912204265594, "global_step": 152446, "epoch": 1712, "lr": 4.050852791604767e-05}
{"train_loss": 0.13084998726844788, "global_step": 152447, "epoch": 1712, "lr": 4.0507958721385284e-05}
{"train_loss": 0.0631040707230568, "global_step": 152448, "epoch": 1712, "lr": 4.0507389527998985e-05}
{"train_loss": 0.0921420082449913, "global_step": 152449, "epoch": 1712, "lr": 4.050682033588885e-05}
{"train_loss": 0.1799941211938858, "global_step": 152450, "epoch": 1712, "lr": 4.0506251145054976e-05}
{"train_loss": 0.1002977192401886, "global_step": 152451, "epoch": 1712, "lr": 4.0505681955497395e-05}
{"train_loss": 0.08099187910556793, "global_step": 152452, "epoch": 1712, "lr": 4.050511276721624e-05}
{"train_loss": 0.10573495924472809, "global_step": 152453, "epoch": 1712, "lr": 4.050454358021152e-05}
{"train_loss": 0.12028171122074127, "global_step": 152454, "epoch": 1712, "lr": 4.050397439448338e-05}
{"train_loss": 0.07194239646196365, "global_step": 152455, "epoch": 1712, "lr": 4.0503405210031844e-05}
{"train_loss": 0.10208192551403902, "global_step": 152456, "epoch": 1712, "lr": 4.050283602685703e-05, "val_loss": 6.297026634216309}
{"train_loss": 0.08879847824573517, "global_step": 152457, "epoch": 1713, "lr": 4.050226684495897e-05}
{"train_loss": 0.06898007541894913, "global_step": 152458, "epoch": 1713, "lr": 4.0501697664337784e-05}
{"train_loss": 0.09706512093544006, "global_step": 152459, "epoch": 1713, "lr": 4.050112848499351e-05}
{"train_loss": 0.07991647720336914, "global_step": 152460, "epoch": 1713, "lr": 4.050055930692627e-05}
{"train_loss": 0.1111539900302887, "global_step": 152461, "epoch": 1713, "lr": 4.049999013013609e-05}
{"train_loss": 0.19259366393089294, "global_step": 152462, "epoch": 1713, "lr": 4.049942095462309e-05}
{"train_loss": 0.08338554948568344, "global_step": 152463, "epoch": 1713, "lr": 4.04988517803873e-05}
{"train_loss": 0.19223415851593018, "global_step": 152464, "epoch": 1713, "lr": 4.049828260742886e-05}
{"train_loss": 0.10627587884664536, "global_step": 152465, "epoch": 1713, "lr": 4.0497713435747784e-05}
{"train_loss": 0.11756586283445358, "global_step": 152466, "epoch": 1713, "lr": 4.0497144265344186e-05}
{"train_loss": 0.09756051748991013, "global_step": 152467, "epoch": 1713, "lr": 4.049657509621812e-05}
{"train_loss": 0.16340272128582, "global_step": 152468, "epoch": 1713, "lr": 4.049600592836967e-05}
{"train_loss": 0.20852385461330414, "global_step": 152469, "epoch": 1713, "lr": 4.049543676179894e-05}
{"train_loss": 0.09649533033370972, "global_step": 152470, "epoch": 1713, "lr": 4.049486759650596e-05}
{"train_loss": 0.07991313934326172, "global_step": 152471, "epoch": 1713, "lr": 4.0494298432490846e-05}
{"train_loss": 0.10780282318592072, "global_step": 152472, "epoch": 1713, "lr": 4.049372926975364e-05}
{"train_loss": 0.11087339371442795, "global_step": 152473, "epoch": 1713, "lr": 4.0493160108294454e-05}
{"train_loss": 0.14323386549949646, "global_step": 152474, "epoch": 1713, "lr": 4.049259094811333e-05}
{"train_loss": 0.08413400501012802, "global_step": 152475, "epoch": 1713, "lr": 4.0492021789210374e-05}
{"train_loss": 0.1065840944647789, "global_step": 152476, "epoch": 1713, "lr": 4.049145263158564e-05}
{"train_loss": 0.10998762398958206, "global_step": 152477, "epoch": 1713, "lr": 4.049088347523923e-05}
{"train_loss": 0.18539243936538696, "global_step": 152478, "epoch": 1713, "lr": 4.049031432017118e-05}
{"train_loss": 0.10048891603946686, "global_step": 152479, "epoch": 1713, "lr": 4.048974516638162e-05}
{"train_loss": 0.0927044078707695, "global_step": 152480, "epoch": 1713, "lr": 4.048917601387057e-05}
{"train_loss": 0.09150519967079163, "global_step": 152481, "epoch": 1713, "lr": 4.048860686263816e-05}
{"train_loss": 0.08521577715873718, "global_step": 152482, "epoch": 1713, "lr": 4.0488037712684425e-05}
{"train_loss": 0.10133840143680573, "global_step": 152483, "epoch": 1713, "lr": 4.0487468564009465e-05}
{"train_loss": 0.17594754695892334, "global_step": 152484, "epoch": 1713, "lr": 4.048689941661335e-05}
{"train_loss": 0.09160450100898743, "global_step": 152485, "epoch": 1713, "lr": 4.048633027049614e-05}
{"train_loss": 0.12391802668571472, "global_step": 152486, "epoch": 1713, "lr": 4.0485761125657945e-05}
{"train_loss": 0.2505938410758972, "global_step": 152487, "epoch": 1713, "lr": 4.048519198209881e-05}
{"train_loss": 0.1420845091342926, "global_step": 152488, "epoch": 1713, "lr": 4.0484622839818844e-05}
{"train_loss": 0.12641723453998566, "global_step": 152489, "epoch": 1713, "lr": 4.048405369881808e-05}
{"train_loss": 0.06806199252605438, "global_step": 152490, "epoch": 1713, "lr": 4.048348455909664e-05}
{"train_loss": 0.10702796280384064, "global_step": 152491, "epoch": 1713, "lr": 4.048291542065455e-05}
{"train_loss": 0.11552533507347107, "global_step": 152492, "epoch": 1713, "lr": 4.0482346283491956e-05}
{"train_loss": 0.16874633729457855, "global_step": 152493, "epoch": 1713, "lr": 4.048177714760886e-05}
{"train_loss": 0.1371435672044754, "global_step": 152494, "epoch": 1713, "lr": 4.04812080130054e-05}
{"train_loss": 0.08552756905555725, "global_step": 152495, "epoch": 1713, "lr": 4.0480638879681606e-05}
{"train_loss": 0.13975004851818085, "global_step": 152496, "epoch": 1713, "lr": 4.048006974763759e-05}
{"train_loss": 0.12425366044044495, "global_step": 152497, "epoch": 1713, "lr": 4.047950061687339e-05}
{"train_loss": 0.10090900957584381, "global_step": 152498, "epoch": 1713, "lr": 4.0478931487389135e-05}
{"train_loss": 0.15828357636928558, "global_step": 152499, "epoch": 1713, "lr": 4.047836235918484e-05}
{"train_loss": 0.09443097561597824, "global_step": 152500, "epoch": 1713, "lr": 4.0477793232260644e-05}
{"train_loss": 0.07425279915332794, "global_step": 152501, "epoch": 1713, "lr": 4.0477224106616564e-05}
{"train_loss": 0.14809919893741608, "global_step": 152502, "epoch": 1713, "lr": 4.047665498225273e-05}
{"train_loss": 0.12429308146238327, "global_step": 152503, "epoch": 1713, "lr": 4.047608585916917e-05}
{"train_loss": 0.1204466000199318, "global_step": 152504, "epoch": 1713, "lr": 4.0475516737366e-05}
{"train_loss": 0.09876780211925507, "global_step": 152505, "epoch": 1713, "lr": 4.047494761684328e-05}
{"train_loss": 0.05023296922445297, "global_step": 152506, "epoch": 1713, "lr": 4.047437849760108e-05}
{"train_loss": 0.07684111595153809, "global_step": 152507, "epoch": 1713, "lr": 4.04738093796395e-05}
{"train_loss": 0.15104693174362183, "global_step": 152508, "epoch": 1713, "lr": 4.047324026295858e-05}
{"train_loss": 0.18161797523498535, "global_step": 152509, "epoch": 1713, "lr": 4.0472671147558435e-05}
{"train_loss": 0.09503845125436783, "global_step": 152510, "epoch": 1713, "lr": 4.04721020334391e-05}
{"train_loss": 0.08075613528490067, "global_step": 152511, "epoch": 1713, "lr": 4.0471532920600694e-05}
{"train_loss": 0.13653819262981415, "global_step": 152512, "epoch": 1713, "lr": 4.047096380904326e-05}
{"train_loss": 0.07644827663898468, "global_step": 152513, "epoch": 1713, "lr": 4.047039469876691e-05}
{"train_loss": 0.10068132728338242, "global_step": 152514, "epoch": 1713, "lr": 4.046982558977168e-05}
{"train_loss": 0.10369294136762619, "global_step": 152515, "epoch": 1713, "lr": 4.046925648205768e-05}
{"train_loss": 0.10658300668001175, "global_step": 152516, "epoch": 1713, "lr": 4.046868737562495e-05}
{"train_loss": 0.18951375782489777, "global_step": 152517, "epoch": 1713, "lr": 4.046811827047362e-05}
{"train_loss": 0.09091610461473465, "global_step": 152518, "epoch": 1713, "lr": 4.04675491666037e-05}
{"train_loss": 0.21882805228233337, "global_step": 152519, "epoch": 1713, "lr": 4.046698006401534e-05}
{"train_loss": 0.1304495632648468, "global_step": 152520, "epoch": 1713, "lr": 4.046641096270855e-05}
{"train_loss": 0.07991635799407959, "global_step": 152521, "epoch": 1713, "lr": 4.046584186268344e-05}
{"train_loss": 0.12813077867031097, "global_step": 152522, "epoch": 1713, "lr": 4.046527276394008e-05}
{"train_loss": 0.08521701395511627, "global_step": 152523, "epoch": 1713, "lr": 4.046470366647855e-05}
{"train_loss": 0.11123191565275192, "global_step": 152524, "epoch": 1713, "lr": 4.0464134570298946e-05}
{"train_loss": 0.1105603575706482, "global_step": 152525, "epoch": 1713, "lr": 4.046356547540129e-05}
{"train_loss": 0.09503234177827835, "global_step": 152526, "epoch": 1713, "lr": 4.046299638178572e-05}
{"train_loss": 0.17924095690250397, "global_step": 152527, "epoch": 1713, "lr": 4.0462427289452254e-05}
{"train_loss": 0.14497067034244537, "global_step": 152528, "epoch": 1713, "lr": 4.046185819840103e-05}
{"train_loss": 0.11743145436048508, "global_step": 152529, "epoch": 1713, "lr": 4.046128910863206e-05}
{"train_loss": 0.10547586530447006, "global_step": 152530, "epoch": 1713, "lr": 4.0460720020145484e-05}
{"train_loss": 0.17277312278747559, "global_step": 152531, "epoch": 1713, "lr": 4.046015093294132e-05}
{"train_loss": 0.10807538777589798, "global_step": 152532, "epoch": 1713, "lr": 4.04595818470197e-05}
{"train_loss": 0.14437958598136902, "global_step": 152533, "epoch": 1713, "lr": 4.045901276238065e-05}
{"train_loss": 0.07815752923488617, "global_step": 152534, "epoch": 1713, "lr": 4.045844367902429e-05}
{"train_loss": 0.14952079951763153, "global_step": 152535, "epoch": 1713, "lr": 4.045787459695065e-05}
{"train_loss": 0.11887466907501221, "global_step": 152536, "epoch": 1713, "lr": 4.045730551615986e-05}
{"train_loss": 0.0733262151479721, "global_step": 152537, "epoch": 1713, "lr": 4.045673643665194e-05}
{"train_loss": 0.08162128925323486, "global_step": 152538, "epoch": 1713, "lr": 4.045616735842703e-05}
{"train_loss": 0.15258030593395233, "global_step": 152539, "epoch": 1713, "lr": 4.045559828148514e-05}
{"train_loss": 0.11085139960050583, "global_step": 152540, "epoch": 1713, "lr": 4.0455029205826395e-05}
{"train_loss": 0.09483903646469116, "global_step": 152541, "epoch": 1713, "lr": 4.045446013145085e-05}
{"train_loss": 0.09932098537683487, "global_step": 152542, "epoch": 1713, "lr": 4.045389105835859e-05}
{"train_loss": 0.06259843707084656, "global_step": 152543, "epoch": 1713, "lr": 4.0453321986549694e-05}
{"train_loss": 0.09111277014017105, "global_step": 152544, "epoch": 1713, "lr": 4.0452752916024217e-05}
{"train_loss": 0.11808395168084777, "global_step": 152545, "epoch": 1713, "lr": 4.045218384678227e-05, "val_loss": 6.438498497009277}
{"train_loss": 0.11340495198965073, "global_step": 152546, "epoch": 1714, "lr": 4.045161477882389e-05}
{"train_loss": 0.09661269932985306, "global_step": 152547, "epoch": 1714, "lr": 4.045104571214919e-05}
{"train_loss": 0.0990704894065857, "global_step": 152548, "epoch": 1714, "lr": 4.045047664675822e-05}
{"train_loss": 0.06046955659985542, "global_step": 152549, "epoch": 1714, "lr": 4.044990758265108e-05}
{"train_loss": 0.1336895376443863, "global_step": 152550, "epoch": 1714, "lr": 4.044933851982781e-05}
{"train_loss": 0.1296095997095108, "global_step": 152551, "epoch": 1714, "lr": 4.044876945828854e-05}
{"train_loss": 0.19282467663288116, "global_step": 152552, "epoch": 1714, "lr": 4.044820039803329e-05}
{"train_loss": 0.13041217625141144, "global_step": 152553, "epoch": 1714, "lr": 4.0447631339062187e-05}
{"train_loss": 0.17792604863643646, "global_step": 152554, "epoch": 1714, "lr": 4.044706228137526e-05}
{"train_loss": 0.08798589557409286, "global_step": 152555, "epoch": 1714, "lr": 4.044649322497263e-05}
{"train_loss": 0.17386843264102936, "global_step": 152556, "epoch": 1714, "lr": 4.0445924169854335e-05}
{"train_loss": 0.08770664781332016, "global_step": 152557, "epoch": 1714, "lr": 4.0445355116020475e-05}
{"train_loss": 0.13454340398311615, "global_step": 152558, "epoch": 1714, "lr": 4.044478606347112e-05}
{"train_loss": 0.04967597499489784, "global_step": 152559, "epoch": 1714, "lr": 4.044421701220635e-05}
{"train_loss": 0.10855992138385773, "global_step": 152560, "epoch": 1714, "lr": 4.044364796222624e-05}
{"train_loss": 0.11045388877391815, "global_step": 152561, "epoch": 1714, "lr": 4.044307891353086e-05}
{"train_loss": 0.07632328569889069, "global_step": 152562, "epoch": 1714, "lr": 4.04425098661203e-05}
{"train_loss": 0.0714859813451767, "global_step": 152563, "epoch": 1714, "lr": 4.044194081999462e-05}
{"train_loss": 0.10623949021100998, "global_step": 152564, "epoch": 1714, "lr": 4.044137177515391e-05}
{"train_loss": 0.08587463200092316, "global_step": 152565, "epoch": 1714, "lr": 4.044080273159823e-05}
{"train_loss": 0.0984203964471817, "global_step": 152566, "epoch": 1714, "lr": 4.044023368932769e-05}
{"train_loss": 0.09814902395009995, "global_step": 152567, "epoch": 1714, "lr": 4.0439664648342315e-05}
{"train_loss": 0.08472735434770584, "global_step": 152568, "epoch": 1714, "lr": 4.0439095608642236e-05}
{"train_loss": 0.10850633680820465, "global_step": 152569, "epoch": 1714, "lr": 4.043852657022748e-05}
{"train_loss": 0.10623885691165924, "global_step": 152570, "epoch": 1714, "lr": 4.0437957533098173e-05}
{"train_loss": 0.13225838541984558, "global_step": 152571, "epoch": 1714, "lr": 4.0437388497254345e-05}
{"train_loss": 0.08364254981279373, "global_step": 152572, "epoch": 1714, "lr": 4.0436819462696116e-05}
{"train_loss": 0.08520440012216568, "global_step": 152573, "epoch": 1714, "lr": 4.043625042942351e-05}
{"train_loss": 0.09420677274465561, "global_step": 152574, "epoch": 1714, "lr": 4.0435681397436666e-05}
{"train_loss": 0.09869101643562317, "global_step": 152575, "epoch": 1714, "lr": 4.04351123667356e-05}
{"train_loss": 0.08616921305656433, "global_step": 152576, "epoch": 1714, "lr": 4.043454333732043e-05}
{"train_loss": 0.05381440371274948, "global_step": 152577, "epoch": 1714, "lr": 4.043397430919121e-05}
{"train_loss": 0.04427410662174225, "global_step": 152578, "epoch": 1714, "lr": 4.043340528234804e-05}
{"train_loss": 0.09644715487957001, "global_step": 152579, "epoch": 1714, "lr": 4.0432836256790976e-05}
{"train_loss": 0.07302539795637131, "global_step": 152580, "epoch": 1714, "lr": 4.04322672325201e-05}
{"train_loss": 0.126872718334198, "global_step": 152581, "epoch": 1714, "lr": 4.0431698209535496e-05}
{"train_loss": 0.07918918132781982, "global_step": 152582, "epoch": 1714, "lr": 4.043112918783722e-05}
{"train_loss": 0.10934707522392273, "global_step": 152583, "epoch": 1714, "lr": 4.043056016742538e-05}
{"train_loss": 0.10746446996927261, "global_step": 152584, "epoch": 1714, "lr": 4.0429991148300005e-05}
{"train_loss": 0.120663121342659, "global_step": 152585, "epoch": 1714, "lr": 4.042942213046124e-05}
{"train_loss": 0.087930828332901, "global_step": 152586, "epoch": 1714, "lr": 4.042885311390908e-05}
{"train_loss": 0.0648614764213562, "global_step": 152587, "epoch": 1714, "lr": 4.042828409864368e-05}
{"train_loss": 0.10662300884723663, "global_step": 152588, "epoch": 1714, "lr": 4.0427715084665057e-05}
{"train_loss": 0.07753891497850418, "global_step": 152589, "epoch": 1714, "lr": 4.0427146071973335e-05}
{"train_loss": 0.1453038603067398, "global_step": 152590, "epoch": 1714, "lr": 4.0426577060568535e-05}
{"train_loss": 0.12151569128036499, "global_step": 152591, "epoch": 1714, "lr": 4.042600805045079e-05}
{"train_loss": 0.15517327189445496, "global_step": 152592, "epoch": 1714, "lr": 4.042543904162014e-05}
{"train_loss": 0.1097259446978569, "global_step": 152593, "epoch": 1714, "lr": 4.042487003407668e-05}
{"train_loss": 0.06863941997289658, "global_step": 152594, "epoch": 1714, "lr": 4.042430102782047e-05}
{"train_loss": 0.08966763317584991, "global_step": 152595, "epoch": 1714, "lr": 4.04237320228516e-05}
{"train_loss": 0.12793998420238495, "global_step": 152596, "epoch": 1714, "lr": 4.0423163019170146e-05}
{"train_loss": 0.07527321577072144, "global_step": 152597, "epoch": 1714, "lr": 4.0422594016776184e-05}
{"train_loss": 0.08999320864677429, "global_step": 152598, "epoch": 1714, "lr": 4.042202501566978e-05}
{"train_loss": 0.13343580067157745, "global_step": 152599, "epoch": 1714, "lr": 4.042145601585102e-05}
{"train_loss": 0.1356482207775116, "global_step": 152600, "epoch": 1714, "lr": 4.042088701731999e-05}
{"train_loss": 0.16178517043590546, "global_step": 152601, "epoch": 1714, "lr": 4.042031802007673e-05}
{"train_loss": 0.14570584893226624, "global_step": 152602, "epoch": 1714, "lr": 4.041974902412137e-05}
{"train_loss": 0.10417557507753372, "global_step": 152603, "epoch": 1714, "lr": 4.041918002945394e-05}
{"train_loss": 0.12869296967983246, "global_step": 152604, "epoch": 1714, "lr": 4.041861103607455e-05}
{"train_loss": 0.13061287999153137, "global_step": 152605, "epoch": 1714, "lr": 4.041804204398323e-05}
{"train_loss": 0.11218450963497162, "global_step": 152606, "epoch": 1714, "lr": 4.041747305318012e-05}
{"train_loss": 0.07878903299570084, "global_step": 152607, "epoch": 1714, "lr": 4.041690406366524e-05}
{"train_loss": 0.12473400682210922, "global_step": 152608, "epoch": 1714, "lr": 4.041633507543871e-05}
{"train_loss": 0.07219251990318298, "global_step": 152609, "epoch": 1714, "lr": 4.0415766088500564e-05}
{"train_loss": 0.0750119760632515, "global_step": 152610, "epoch": 1714, "lr": 4.041519710285093e-05}
{"train_loss": 0.10453224927186966, "global_step": 152611, "epoch": 1714, "lr": 4.0414628118489834e-05}
{"train_loss": 0.10275164991617203, "global_step": 152612, "epoch": 1714, "lr": 4.041405913541737e-05}
{"train_loss": 0.12240706384181976, "global_step": 152613, "epoch": 1714, "lr": 4.0413490153633645e-05}
{"train_loss": 0.20518705248832703, "global_step": 152614, "epoch": 1714, "lr": 4.041292117313869e-05}
{"train_loss": 0.07936558127403259, "global_step": 152615, "epoch": 1714, "lr": 4.0412352193932606e-05}
{"train_loss": 0.08352534472942352, "global_step": 152616, "epoch": 1714, "lr": 4.0411783216015463e-05}
{"train_loss": 0.1287415325641632, "global_step": 152617, "epoch": 1714, "lr": 4.041121423938734e-05}
{"train_loss": 0.0884263888001442, "global_step": 152618, "epoch": 1714, "lr": 4.041064526404832e-05}
{"train_loss": 0.10120736062526703, "global_step": 152619, "epoch": 1714, "lr": 4.041007628999846e-05}
{"train_loss": 0.13784782588481903, "global_step": 152620, "epoch": 1714, "lr": 4.040950731723785e-05}
{"train_loss": 0.1123141199350357, "global_step": 152621, "epoch": 1714, "lr": 4.0408938345766583e-05}
{"train_loss": 0.0895024985074997, "global_step": 152622, "epoch": 1714, "lr": 4.0408369375584695e-05}
{"train_loss": 0.06623310595750809, "global_step": 152623, "epoch": 1714, "lr": 4.0407800406692295e-05}
{"train_loss": 0.1410817950963974, "global_step": 152624, "epoch": 1714, "lr": 4.040723143908944e-05}
{"train_loss": 0.14787840843200684, "global_step": 152625, "epoch": 1714, "lr": 4.0406662472776234e-05}
{"train_loss": 0.1396200805902481, "global_step": 152626, "epoch": 1714, "lr": 4.0406093507752715e-05}
{"train_loss": 0.1362697184085846, "global_step": 152627, "epoch": 1714, "lr": 4.0405524544018996e-05}
{"train_loss": 0.10844836384057999, "global_step": 152628, "epoch": 1714, "lr": 4.040495558157512e-05}
{"train_loss": 0.11513587832450867, "global_step": 152629, "epoch": 1714, "lr": 4.04043866204212e-05}
{"train_loss": 0.14909221231937408, "global_step": 152630, "epoch": 1714, "lr": 4.0403817660557276e-05}
{"train_loss": 0.06977785378694534, "global_step": 152631, "epoch": 1714, "lr": 4.040324870198343e-05}
{"train_loss": 0.1773514747619629, "global_step": 152632, "epoch": 1714, "lr": 4.040267974469978e-05}
{"train_loss": 0.09890495240688324, "global_step": 152633, "epoch": 1714, "lr": 4.040211078870636e-05}
{"train_loss": 0.10778299020080084, "global_step": 152634, "epoch": 1714, "lr": 4.040154183400327e-05, "val_loss": 6.274326801300049}
{"train_loss": 0.09873776882886887, "global_step": 152635, "epoch": 1715, "lr": 4.040097288059057e-05}
{"train_loss": 0.0734834223985672, "global_step": 152636, "epoch": 1715, "lr": 4.0400403928468337e-05}
{"train_loss": 0.14578495919704437, "global_step": 152637, "epoch": 1715, "lr": 4.039983497763665e-05}
{"train_loss": 0.12834173440933228, "global_step": 152638, "epoch": 1715, "lr": 4.039926602809559e-05}
{"train_loss": 0.056041523814201355, "global_step": 152639, "epoch": 1715, "lr": 4.0398697079845244e-05}
{"train_loss": 0.11443112790584564, "global_step": 152640, "epoch": 1715, "lr": 4.039812813288568e-05}
{"train_loss": 0.07898388057947159, "global_step": 152641, "epoch": 1715, "lr": 4.039755918721694e-05}
{"train_loss": 0.09113418310880661, "global_step": 152642, "epoch": 1715, "lr": 4.039699024283916e-05}
{"train_loss": 0.10457119345664978, "global_step": 152643, "epoch": 1715, "lr": 4.0396421299752365e-05}
{"train_loss": 0.09315812587738037, "global_step": 152644, "epoch": 1715, "lr": 4.039585235795668e-05}
{"train_loss": 0.08954121172428131, "global_step": 152645, "epoch": 1715, "lr": 4.0395283417452134e-05}
{"train_loss": 0.1590481698513031, "global_step": 152646, "epoch": 1715, "lr": 4.039471447823884e-05}
{"train_loss": 0.061876215040683746, "global_step": 152647, "epoch": 1715, "lr": 4.039414554031684e-05}
{"train_loss": 0.09963502734899521, "global_step": 152648, "epoch": 1715, "lr": 4.0393576603686236e-05}
{"train_loss": 0.1275763213634491, "global_step": 152649, "epoch": 1715, "lr": 4.0393007668347115e-05}
{"train_loss": 0.10444043576717377, "global_step": 152650, "epoch": 1715, "lr": 4.0392438734299517e-05}
{"train_loss": 0.11603793501853943, "global_step": 152651, "epoch": 1715, "lr": 4.0391869801543555e-05}
{"train_loss": 0.12516885995864868, "global_step": 152652, "epoch": 1715, "lr": 4.039130087007927e-05}
{"train_loss": 0.12202306091785431, "global_step": 152653, "epoch": 1715, "lr": 4.039073193990678e-05}
{"train_loss": 0.06141693890094757, "global_step": 152654, "epoch": 1715, "lr": 4.039016301102613e-05}
{"train_loss": 0.0949995219707489, "global_step": 152655, "epoch": 1715, "lr": 4.03895940834374e-05}
{"train_loss": 0.06590002775192261, "global_step": 152656, "epoch": 1715, "lr": 4.038902515714068e-05}
{"train_loss": 0.09432488679885864, "global_step": 152657, "epoch": 1715, "lr": 4.0388456232136034e-05}
{"train_loss": 0.07612241059541702, "global_step": 152658, "epoch": 1715, "lr": 4.0387887308423545e-05}
{"train_loss": 0.06335657835006714, "global_step": 152659, "epoch": 1715, "lr": 4.038731838600329e-05}
{"train_loss": 0.12324322760105133, "global_step": 152660, "epoch": 1715, "lr": 4.0386749464875335e-05}
{"train_loss": 0.0997810959815979, "global_step": 152661, "epoch": 1715, "lr": 4.038618054503978e-05}
{"train_loss": 0.10991999506950378, "global_step": 152662, "epoch": 1715, "lr": 4.038561162649666e-05}
{"train_loss": 0.12793897092342377, "global_step": 152663, "epoch": 1715, "lr": 4.03850427092461e-05}
{"train_loss": 0.08481760323047638, "global_step": 152664, "epoch": 1715, "lr": 4.0384473793288136e-05}
{"train_loss": 0.06929407268762589, "global_step": 152665, "epoch": 1715, "lr": 4.038390487862288e-05}
{"train_loss": 0.11964830756187439, "global_step": 152666, "epoch": 1715, "lr": 4.0383335965250374e-05}
{"train_loss": 0.035550955682992935, "global_step": 152667, "epoch": 1715, "lr": 4.03827670531707e-05}
{"train_loss": 0.1632666438817978, "global_step": 152668, "epoch": 1715, "lr": 4.0382198142383984e-05}
{"train_loss": 0.10218621045351028, "global_step": 152669, "epoch": 1715, "lr": 4.038162923289023e-05}
{"train_loss": 0.038724713027477264, "global_step": 152670, "epoch": 1715, "lr": 4.038106032468958e-05}
{"train_loss": 0.11450496315956116, "global_step": 152671, "epoch": 1715, "lr": 4.038049141778205e-05}
{"train_loss": 0.09480597078800201, "global_step": 152672, "epoch": 1715, "lr": 4.0379922512167765e-05}
{"train_loss": 0.14303696155548096, "global_step": 152673, "epoch": 1715, "lr": 4.037935360784678e-05}
{"train_loss": 0.0790395438671112, "global_step": 152674, "epoch": 1715, "lr": 4.0378784704819175e-05}
{"train_loss": 0.09918871521949768, "global_step": 152675, "epoch": 1715, "lr": 4.0378215803085016e-05}
{"train_loss": 0.09411723166704178, "global_step": 152676, "epoch": 1715, "lr": 4.03776469026444e-05}
{"train_loss": 0.10243780165910721, "global_step": 152677, "epoch": 1715, "lr": 4.0377078003497375e-05}
{"train_loss": 0.10458600521087646, "global_step": 152678, "epoch": 1715, "lr": 4.037650910564406e-05}
{"train_loss": 0.06680846959352493, "global_step": 152679, "epoch": 1715, "lr": 4.0375940209084476e-05}
{"train_loss": 0.17494486272335052, "global_step": 152680, "epoch": 1715, "lr": 4.0375371313818755e-05}
{"train_loss": 0.08915649354457855, "global_step": 152681, "epoch": 1715, "lr": 4.037480241984693e-05}
{"train_loss": 0.12837277352809906, "global_step": 152682, "epoch": 1715, "lr": 4.037423352716911e-05}
{"train_loss": 0.0869065672159195, "global_step": 152683, "epoch": 1715, "lr": 4.037366463578533e-05}
{"train_loss": 0.14136822521686554, "global_step": 152684, "epoch": 1715, "lr": 4.037309574569571e-05}
{"train_loss": 0.1014188900589943, "global_step": 152685, "epoch": 1715, "lr": 4.037252685690032e-05}
{"train_loss": 0.14922736585140228, "global_step": 152686, "epoch": 1715, "lr": 4.037195796939921e-05}
{"train_loss": 0.08705000579357147, "global_step": 152687, "epoch": 1715, "lr": 4.037138908319249e-05}
{"train_loss": 0.07046803832054138, "global_step": 152688, "epoch": 1715, "lr": 4.03708201982802e-05}
{"train_loss": 0.11168613284826279, "global_step": 152689, "epoch": 1715, "lr": 4.037025131466246e-05}
{"train_loss": 0.04728153347969055, "global_step": 152690, "epoch": 1715, "lr": 4.03696824323393e-05}
{"train_loss": 0.10945546627044678, "global_step": 152691, "epoch": 1715, "lr": 4.0369113551310837e-05}
{"train_loss": 0.07967069000005722, "global_step": 152692, "epoch": 1715, "lr": 4.0368544671577125e-05}
{"train_loss": 0.05779223144054413, "global_step": 152693, "epoch": 1715, "lr": 4.0367975793138235e-05}
{"train_loss": 0.10312894731760025, "global_step": 152694, "epoch": 1715, "lr": 4.036740691599426e-05}
{"train_loss": 0.07343035936355591, "global_step": 152695, "epoch": 1715, "lr": 4.0366838040145274e-05}
{"train_loss": 0.09570886939764023, "global_step": 152696, "epoch": 1715, "lr": 4.0366269165591345e-05}
{"train_loss": 0.10223051905632019, "global_step": 152697, "epoch": 1715, "lr": 4.0365700292332556e-05}
{"train_loss": 0.0934762954711914, "global_step": 152698, "epoch": 1715, "lr": 4.036513142036897e-05}
{"train_loss": 0.09936288744211197, "global_step": 152699, "epoch": 1715, "lr": 4.036456254970069e-05}
{"train_loss": 0.15877537429332733, "global_step": 152700, "epoch": 1715, "lr": 4.036399368032776e-05}
{"train_loss": 0.11567950993776321, "global_step": 152701, "epoch": 1715, "lr": 4.03634248122503e-05}
{"train_loss": 0.055088333785533905, "global_step": 152702, "epoch": 1715, "lr": 4.036285594546832e-05}
{"train_loss": 0.08230023831129074, "global_step": 152703, "epoch": 1715, "lr": 4.0362287079981956e-05}
{"train_loss": 0.051703449338674545, "global_step": 152704, "epoch": 1715, "lr": 4.036171821579128e-05}
{"train_loss": 0.06640101224184036, "global_step": 152705, "epoch": 1715, "lr": 4.0361149352896324e-05}
{"train_loss": 0.13285954296588898, "global_step": 152706, "epoch": 1715, "lr": 4.036058049129723e-05}
{"train_loss": 0.1026957780122757, "global_step": 152707, "epoch": 1715, "lr": 4.0360011630994006e-05}
{"train_loss": 0.13803084194660187, "global_step": 152708, "epoch": 1715, "lr": 4.035944277198679e-05}
{"train_loss": 0.1470804065465927, "global_step": 152709, "epoch": 1715, "lr": 4.03588739142756e-05}
{"train_loss": 0.10112960636615753, "global_step": 152710, "epoch": 1715, "lr": 4.0358305057860566e-05}
{"train_loss": 0.10311165452003479, "global_step": 152711, "epoch": 1715, "lr": 4.0357736202741734e-05}
{"train_loss": 0.1314486712217331, "global_step": 152712, "epoch": 1715, "lr": 4.035716734891919e-05}
{"train_loss": 0.058271586894989014, "global_step": 152713, "epoch": 1715, "lr": 4.035659849639299e-05}
{"train_loss": 0.11035967618227005, "global_step": 152714, "epoch": 1715, "lr": 4.035602964516324e-05}
{"train_loss": 0.10584384202957153, "global_step": 152715, "epoch": 1715, "lr": 4.0355460795230005e-05}
{"train_loss": 0.11716780811548233, "global_step": 152716, "epoch": 1715, "lr": 4.035489194659336e-05}
{"train_loss": 0.11440937966108322, "global_step": 152717, "epoch": 1715, "lr": 4.035432309925338e-05}
{"train_loss": 0.11030463129281998, "global_step": 152718, "epoch": 1715, "lr": 4.035375425321016e-05}
{"train_loss": 0.06692751497030258, "global_step": 152719, "epoch": 1715, "lr": 4.035318540846373e-05}
{"train_loss": 0.08211254328489304, "global_step": 152720, "epoch": 1715, "lr": 4.035261656501422e-05}
{"train_loss": 0.10140500217676163, "global_step": 152721, "epoch": 1715, "lr": 4.035204772286167e-05}
{"train_loss": 0.12116429954767227, "global_step": 152722, "epoch": 1715, "lr": 4.035147888200616e-05}
{"train_loss": 0.1004407089412882, "global_step": 152723, "epoch": 1715, "lr": 4.0350910042447803e-05, "val_loss": 6.432656288146973, "train_action_mse_error": 11.354209899902344}
{"train_loss": 0.074152372777462, "global_step": 152724, "epoch": 1716, "lr": 4.035034120418663e-05}
{"train_loss": 0.11758299171924591, "global_step": 152725, "epoch": 1716, "lr": 4.0349772367222747e-05}
{"train_loss": 0.10255908966064453, "global_step": 152726, "epoch": 1716, "lr": 4.034920353155621e-05}
{"train_loss": 0.10132590681314468, "global_step": 152727, "epoch": 1716, "lr": 4.0348634697187115e-05}
{"train_loss": 0.11655167490243912, "global_step": 152728, "epoch": 1716, "lr": 4.034806586411551e-05}
{"train_loss": 0.10349655151367188, "global_step": 152729, "epoch": 1716, "lr": 4.034749703234152e-05}
{"train_loss": 0.11092236638069153, "global_step": 152730, "epoch": 1716, "lr": 4.0346928201865165e-05}
{"train_loss": 0.048303693532943726, "global_step": 152731, "epoch": 1716, "lr": 4.034635937268657e-05}
{"train_loss": 0.10871397703886032, "global_step": 152732, "epoch": 1716, "lr": 4.034579054480577e-05}
{"train_loss": 0.1305159628391266, "global_step": 152733, "epoch": 1716, "lr": 4.034522171822287e-05}
{"train_loss": 0.13923469185829163, "global_step": 152734, "epoch": 1716, "lr": 4.034465289293793e-05}
{"train_loss": 0.05967927724123001, "global_step": 152735, "epoch": 1716, "lr": 4.034408406895104e-05}
{"train_loss": 0.09440895915031433, "global_step": 152736, "epoch": 1716, "lr": 4.034351524626227e-05}
{"train_loss": 0.08708924055099487, "global_step": 152737, "epoch": 1716, "lr": 4.03429464248717e-05}
{"train_loss": 0.06211645528674126, "global_step": 152738, "epoch": 1716, "lr": 4.034237760477939e-05}
{"train_loss": 0.14303559064865112, "global_step": 152739, "epoch": 1716, "lr": 4.034180878598543e-05}
{"train_loss": 0.09057893604040146, "global_step": 152740, "epoch": 1716, "lr": 4.0341239968489916e-05}
{"train_loss": 0.12364871054887772, "global_step": 152741, "epoch": 1716, "lr": 4.034067115229288e-05}
{"train_loss": 0.12406478077173233, "global_step": 152742, "epoch": 1716, "lr": 4.034010233739445e-05}
{"train_loss": 0.12069247663021088, "global_step": 152743, "epoch": 1716, "lr": 4.033953352379465e-05}
{"train_loss": 0.09995707124471664, "global_step": 152744, "epoch": 1716, "lr": 4.03389647114936e-05}
{"train_loss": 0.1250295341014862, "global_step": 152745, "epoch": 1716, "lr": 4.0338395900491344e-05}
{"train_loss": 0.09545255452394485, "global_step": 152746, "epoch": 1716, "lr": 4.0337827090787986e-05}
{"train_loss": 0.10435976088047028, "global_step": 152747, "epoch": 1716, "lr": 4.033725828238357e-05}
{"train_loss": 0.08041966706514359, "global_step": 152748, "epoch": 1716, "lr": 4.033668947527822e-05}
{"train_loss": 0.16031929850578308, "global_step": 152749, "epoch": 1716, "lr": 4.033612066947196e-05}
{"train_loss": 0.07657913118600845, "global_step": 152750, "epoch": 1716, "lr": 4.033555186496491e-05}
{"train_loss": 0.11722595244646072, "global_step": 152751, "epoch": 1716, "lr": 4.033498306175712e-05}
{"train_loss": 0.1041853129863739, "global_step": 152752, "epoch": 1716, "lr": 4.033441425984867e-05}
{"train_loss": 0.08584677428007126, "global_step": 152753, "epoch": 1716, "lr": 4.033384545923965e-05}
{"train_loss": 0.11860170215368271, "global_step": 152754, "epoch": 1716, "lr": 4.0333276659930115e-05}
{"train_loss": 0.07709066569805145, "global_step": 152755, "epoch": 1716, "lr": 4.0332707861920154e-05}
{"train_loss": 0.1225043535232544, "global_step": 152756, "epoch": 1716, "lr": 4.0332139065209854e-05}
{"train_loss": 0.05587925761938095, "global_step": 152757, "epoch": 1716, "lr": 4.033157026979926e-05}
{"train_loss": 0.13293159008026123, "global_step": 152758, "epoch": 1716, "lr": 4.033100147568847e-05}
{"train_loss": 0.10191301256418228, "global_step": 152759, "epoch": 1716, "lr": 4.0330432682877586e-05}
{"train_loss": 0.145027756690979, "global_step": 152760, "epoch": 1716, "lr": 4.0329863891366625e-05}
{"train_loss": 0.06299017369747162, "global_step": 152761, "epoch": 1716, "lr": 4.0329295101155724e-05}
{"train_loss": 0.20158840715885162, "global_step": 152762, "epoch": 1716, "lr": 4.032872631224491e-05}
{"train_loss": 0.09304045885801315, "global_step": 152763, "epoch": 1716, "lr": 4.03281575246343e-05}
{"train_loss": 0.11504177749156952, "global_step": 152764, "epoch": 1716, "lr": 4.032758873832393e-05}
{"train_loss": 0.11803537607192993, "global_step": 152765, "epoch": 1716, "lr": 4.032701995331392e-05}
{"train_loss": 0.06419670581817627, "global_step": 152766, "epoch": 1716, "lr": 4.0326451169604294e-05}
{"train_loss": 0.11570779234170914, "global_step": 152767, "epoch": 1716, "lr": 4.032588238719519e-05}
{"train_loss": 0.10870406031608582, "global_step": 152768, "epoch": 1716, "lr": 4.032531360608663e-05}
{"train_loss": 0.1890423446893692, "global_step": 152769, "epoch": 1716, "lr": 4.032474482627874e-05}
{"train_loss": 0.1267564743757248, "global_step": 152770, "epoch": 1716, "lr": 4.032417604777156e-05}
{"train_loss": 0.0964447408914566, "global_step": 152771, "epoch": 1716, "lr": 4.0323607270565167e-05}
{"train_loss": 0.14485110342502594, "global_step": 152772, "epoch": 1716, "lr": 4.032303849465965e-05}
{"train_loss": 0.10027430951595306, "global_step": 152773, "epoch": 1716, "lr": 4.0322469720055086e-05}
{"train_loss": 0.09982295334339142, "global_step": 152774, "epoch": 1716, "lr": 4.0321900946751545e-05}
{"train_loss": 0.04813649505376816, "global_step": 152775, "epoch": 1716, "lr": 4.0321332174749094e-05}
{"train_loss": 0.08747585862874985, "global_step": 152776, "epoch": 1716, "lr": 4.032076340404784e-05}
{"train_loss": 0.16797959804534912, "global_step": 152777, "epoch": 1716, "lr": 4.0320194634647836e-05}
{"train_loss": 0.07425880432128906, "global_step": 152778, "epoch": 1716, "lr": 4.031962586654917e-05}
{"train_loss": 0.1177174374461174, "global_step": 152779, "epoch": 1716, "lr": 4.03190570997519e-05}
{"train_loss": 0.06358969956636429, "global_step": 152780, "epoch": 1716, "lr": 4.031848833425613e-05}
{"train_loss": 0.11159322410821915, "global_step": 152781, "epoch": 1716, "lr": 4.03179195700619e-05}
{"train_loss": 0.07454616576433182, "global_step": 152782, "epoch": 1716, "lr": 4.031735080716934e-05}
{"train_loss": 0.07541505247354507, "global_step": 152783, "epoch": 1716, "lr": 4.031678204557846e-05}
{"train_loss": 0.08290281146764755, "global_step": 152784, "epoch": 1716, "lr": 4.031621328528939e-05}
{"train_loss": 0.12155694514513016, "global_step": 152785, "epoch": 1716, "lr": 4.031564452630218e-05}
{"train_loss": 0.06000587344169617, "global_step": 152786, "epoch": 1716, "lr": 4.031507576861692e-05}
{"train_loss": 0.08974935114383698, "global_step": 152787, "epoch": 1716, "lr": 4.031450701223367e-05}
{"train_loss": 0.09496904164552689, "global_step": 152788, "epoch": 1716, "lr": 4.031393825715253e-05}
{"train_loss": 0.07005338370800018, "global_step": 152789, "epoch": 1716, "lr": 4.0313369503373555e-05}
{"train_loss": 0.09881191700696945, "global_step": 152790, "epoch": 1716, "lr": 4.0312800750896833e-05}
{"train_loss": 0.0822710171341896, "global_step": 152791, "epoch": 1716, "lr": 4.0312231999722435e-05}
{"train_loss": 0.07234051078557968, "global_step": 152792, "epoch": 1716, "lr": 4.0311663249850434e-05}
{"train_loss": 0.16071245074272156, "global_step": 152793, "epoch": 1716, "lr": 4.031109450128091e-05}
{"train_loss": 0.11502283811569214, "global_step": 152794, "epoch": 1716, "lr": 4.0310525754013945e-05}
{"train_loss": 0.10394445806741714, "global_step": 152795, "epoch": 1716, "lr": 4.0309957008049626e-05}
{"train_loss": 0.14736932516098022, "global_step": 152796, "epoch": 1716, "lr": 4.030938826338799e-05}
{"train_loss": 0.19170129299163818, "global_step": 152797, "epoch": 1716, "lr": 4.0308819520029164e-05}
{"train_loss": 0.1121058315038681, "global_step": 152798, "epoch": 1716, "lr": 4.030825077797318e-05}
{"train_loss": 0.1109309047460556, "global_step": 152799, "epoch": 1716, "lr": 4.030768203722014e-05}
{"train_loss": 0.12284555286169052, "global_step": 152800, "epoch": 1716, "lr": 4.030711329777011e-05}
{"train_loss": 0.08425736427307129, "global_step": 152801, "epoch": 1716, "lr": 4.030654455962318e-05}
{"train_loss": 0.03807814419269562, "global_step": 152802, "epoch": 1716, "lr": 4.030597582277941e-05}
{"train_loss": 0.13663563132286072, "global_step": 152803, "epoch": 1716, "lr": 4.0305407087238886e-05}
{"train_loss": 0.1152290478348732, "global_step": 152804, "epoch": 1716, "lr": 4.030483835300168e-05}
{"train_loss": 0.12602683901786804, "global_step": 152805, "epoch": 1716, "lr": 4.030426962006788e-05}
{"train_loss": 0.1359623521566391, "global_step": 152806, "epoch": 1716, "lr": 4.0303700888437536e-05}
{"train_loss": 0.0543287917971611, "global_step": 152807, "epoch": 1716, "lr": 4.030313215811076e-05}
{"train_loss": 0.09152227640151978, "global_step": 152808, "epoch": 1716, "lr": 4.0302563429087594e-05}
{"train_loss": 0.08852078020572662, "global_step": 152809, "epoch": 1716, "lr": 4.0301994701368134e-05}
{"train_loss": 0.1387258619070053, "global_step": 152810, "epoch": 1716, "lr": 4.0301425974952454e-05}
{"train_loss": 0.12560227513313293, "global_step": 152811, "epoch": 1716, "lr": 4.0300857249840616e-05}
{"train_loss": 0.10502763017174903, "global_step": 152812, "epoch": 1716, "lr": 4.0300288526032734e-05, "val_loss": 6.3754448890686035}
{"train_loss": 0.08743756264448166, "global_step": 152813, "epoch": 1717, "lr": 4.029971980352884e-05}
{"train_loss": 0.08704458922147751, "global_step": 152814, "epoch": 1717, "lr": 4.029915108232905e-05}
{"train_loss": 0.10458191484212875, "global_step": 152815, "epoch": 1717, "lr": 4.02985823624334e-05}
{"train_loss": 0.0709906592965126, "global_step": 152816, "epoch": 1717, "lr": 4.029801364384201e-05}
{"train_loss": 0.08921344578266144, "global_step": 152817, "epoch": 1717, "lr": 4.0297444926554906e-05}
{"train_loss": 0.12860575318336487, "global_step": 152818, "epoch": 1717, "lr": 4.029687621057222e-05}
{"train_loss": 0.13786740601062775, "global_step": 152819, "epoch": 1717, "lr": 4.029630749589398e-05}
{"train_loss": 0.08210783451795578, "global_step": 152820, "epoch": 1717, "lr": 4.02957387825203e-05}
{"train_loss": 0.12130651623010635, "global_step": 152821, "epoch": 1717, "lr": 4.029517007045122e-05}
{"train_loss": 0.06676732003688812, "global_step": 152822, "epoch": 1717, "lr": 4.029460135968686e-05}
{"train_loss": 0.1544787734746933, "global_step": 152823, "epoch": 1717, "lr": 4.0294032650227246e-05}
{"train_loss": 0.12858815491199493, "global_step": 152824, "epoch": 1717, "lr": 4.0293463942072506e-05}
{"train_loss": 0.15727393329143524, "global_step": 152825, "epoch": 1717, "lr": 4.0292895235222664e-05}
{"train_loss": 0.08820102363824844, "global_step": 152826, "epoch": 1717, "lr": 4.0292326529677856e-05}
{"train_loss": 0.12913347780704498, "global_step": 152827, "epoch": 1717, "lr": 4.029175782543811e-05}
{"train_loss": 0.11418412625789642, "global_step": 152828, "epoch": 1717, "lr": 4.029118912250352e-05}
{"train_loss": 0.14075727760791779, "global_step": 152829, "epoch": 1717, "lr": 4.0290620420874156e-05}
{"train_loss": 0.13988666236400604, "global_step": 152830, "epoch": 1717, "lr": 4.029005172055009e-05}
{"train_loss": 0.10760379582643509, "global_step": 152831, "epoch": 1717, "lr": 4.0289483021531436e-05}
{"train_loss": 0.11411301791667938, "global_step": 152832, "epoch": 1717, "lr": 4.0288914323818214e-05}
{"train_loss": 0.09455298632383347, "global_step": 152833, "epoch": 1717, "lr": 4.028834562741056e-05}
{"train_loss": 0.09353560954332352, "global_step": 152834, "epoch": 1717, "lr": 4.028777693230849e-05}
{"train_loss": 0.06907633692026138, "global_step": 152835, "epoch": 1717, "lr": 4.0287208238512136e-05}
{"train_loss": 0.15845945477485657, "global_step": 152836, "epoch": 1717, "lr": 4.028663954602153e-05}
{"train_loss": 0.10558503866195679, "global_step": 152837, "epoch": 1717, "lr": 4.0286070854836775e-05}
{"train_loss": 0.10407337546348572, "global_step": 152838, "epoch": 1717, "lr": 4.0285502164957936e-05}
{"train_loss": 0.0507403202354908, "global_step": 152839, "epoch": 1717, "lr": 4.02849334763851e-05}
{"train_loss": 0.08357588946819305, "global_step": 152840, "epoch": 1717, "lr": 4.028436478911832e-05}
{"train_loss": 0.13503263890743256, "global_step": 152841, "epoch": 1717, "lr": 4.0283796103157715e-05}
{"train_loss": 0.10407336056232452, "global_step": 152842, "epoch": 1717, "lr": 4.028322741850331e-05}
{"train_loss": 0.11066434532403946, "global_step": 152843, "epoch": 1717, "lr": 4.028265873515522e-05}
{"train_loss": 0.11230012029409409, "global_step": 152844, "epoch": 1717, "lr": 4.02820900531135e-05}
{"train_loss": 0.15140488743782043, "global_step": 152845, "epoch": 1717, "lr": 4.028152137237824e-05}
{"train_loss": 0.11250080168247223, "global_step": 152846, "epoch": 1717, "lr": 4.02809526929495e-05}
{"train_loss": 0.09191878885030746, "global_step": 152847, "epoch": 1717, "lr": 4.028038401482739e-05}
{"train_loss": 0.17420171201229095, "global_step": 152848, "epoch": 1717, "lr": 4.027981533801195e-05}
{"train_loss": 0.16738465428352356, "global_step": 152849, "epoch": 1717, "lr": 4.027924666250326e-05}
{"train_loss": 0.0737115889787674, "global_step": 152850, "epoch": 1717, "lr": 4.027867798830143e-05}
{"train_loss": 0.07375765591859818, "global_step": 152851, "epoch": 1717, "lr": 4.0278109315406495e-05}
{"train_loss": 0.07685954123735428, "global_step": 152852, "epoch": 1717, "lr": 4.027754064381856e-05}
{"train_loss": 0.0537300668656826, "global_step": 152853, "epoch": 1717, "lr": 4.027697197353768e-05}
{"train_loss": 0.09435498714447021, "global_step": 152854, "epoch": 1717, "lr": 4.027640330456396e-05}
{"train_loss": 0.0709744319319725, "global_step": 152855, "epoch": 1717, "lr": 4.0275834636897434e-05}
{"train_loss": 0.06109236180782318, "global_step": 152856, "epoch": 1717, "lr": 4.027526597053823e-05}
{"train_loss": 0.11675972491502762, "global_step": 152857, "epoch": 1717, "lr": 4.0274697305486384e-05}
{"train_loss": 0.06619438529014587, "global_step": 152858, "epoch": 1717, "lr": 4.0274128641741994e-05}
{"train_loss": 0.07995825260877609, "global_step": 152859, "epoch": 1717, "lr": 4.027355997930511e-05}
{"train_loss": 0.17166244983673096, "global_step": 152860, "epoch": 1717, "lr": 4.027299131817586e-05}
{"train_loss": 0.06453799456357956, "global_step": 152861, "epoch": 1717, "lr": 4.0272422658354255e-05}
{"train_loss": 0.14830929040908813, "global_step": 152862, "epoch": 1717, "lr": 4.027185399984042e-05}
{"train_loss": 0.111343152821064, "global_step": 152863, "epoch": 1717, "lr": 4.027128534263441e-05}
{"train_loss": 0.1348598450422287, "global_step": 152864, "epoch": 1717, "lr": 4.0270716686736323e-05}
{"train_loss": 0.06988435238599777, "global_step": 152865, "epoch": 1717, "lr": 4.027014803214619e-05}
{"train_loss": 0.12963594496250153, "global_step": 152866, "epoch": 1717, "lr": 4.026957937886414e-05}
{"train_loss": 0.1368870586156845, "global_step": 152867, "epoch": 1717, "lr": 4.026901072689023e-05}
{"train_loss": 0.0785725861787796, "global_step": 152868, "epoch": 1717, "lr": 4.026844207622452e-05}
{"train_loss": 0.12091685831546783, "global_step": 152869, "epoch": 1717, "lr": 4.0267873426867106e-05}
{"train_loss": 0.08440382033586502, "global_step": 152870, "epoch": 1717, "lr": 4.026730477881805e-05}
{"train_loss": 0.10523928701877594, "global_step": 152871, "epoch": 1717, "lr": 4.026673613207745e-05}
{"train_loss": 0.08330941945314407, "global_step": 152872, "epoch": 1717, "lr": 4.0266167486645345e-05}
{"train_loss": 0.13691551983356476, "global_step": 152873, "epoch": 1717, "lr": 4.0265598842521864e-05}
{"train_loss": 0.06729424744844437, "global_step": 152874, "epoch": 1717, "lr": 4.0265030199707035e-05}
{"train_loss": 0.06650634109973907, "global_step": 152875, "epoch": 1717, "lr": 4.026446155820097e-05}
{"train_loss": 0.06950860470533371, "global_step": 152876, "epoch": 1717, "lr": 4.02638929180037e-05}
{"train_loss": 0.10272372514009476, "global_step": 152877, "epoch": 1717, "lr": 4.0263324279115364e-05}
{"train_loss": 0.11002659797668457, "global_step": 152878, "epoch": 1717, "lr": 4.0262755641535985e-05}
{"train_loss": 0.1426229476928711, "global_step": 152879, "epoch": 1717, "lr": 4.026218700526567e-05}
{"train_loss": 0.15107429027557373, "global_step": 152880, "epoch": 1717, "lr": 4.026161837030447e-05}
{"train_loss": 0.14697052538394928, "global_step": 152881, "epoch": 1717, "lr": 4.02610497366525e-05}
{"train_loss": 0.06593149155378342, "global_step": 152882, "epoch": 1717, "lr": 4.02604811043098e-05}
{"train_loss": 0.0726112574338913, "global_step": 152883, "epoch": 1717, "lr": 4.0259912473276464e-05}
{"train_loss": 0.10486330091953278, "global_step": 152884, "epoch": 1717, "lr": 4.025934384355255e-05}
{"train_loss": 0.11665249615907669, "global_step": 152885, "epoch": 1717, "lr": 4.025877521513816e-05}
{"train_loss": 0.056335680186748505, "global_step": 152886, "epoch": 1717, "lr": 4.025820658803337e-05}
{"train_loss": 0.07890380918979645, "global_step": 152887, "epoch": 1717, "lr": 4.0257637962238224e-05}
{"train_loss": 0.11370953917503357, "global_step": 152888, "epoch": 1717, "lr": 4.025706933775284e-05}
{"train_loss": 0.15750300884246826, "global_step": 152889, "epoch": 1717, "lr": 4.025650071457725e-05}
{"train_loss": 0.08885524421930313, "global_step": 152890, "epoch": 1717, "lr": 4.025593209271158e-05}
{"train_loss": 0.06779711693525314, "global_step": 152891, "epoch": 1717, "lr": 4.025536347215586e-05}
{"train_loss": 0.07400038838386536, "global_step": 152892, "epoch": 1717, "lr": 4.025479485291021e-05}
{"train_loss": 0.18122072517871857, "global_step": 152893, "epoch": 1717, "lr": 4.0254226234974654e-05}
{"train_loss": 0.10135210305452347, "global_step": 152894, "epoch": 1717, "lr": 4.025365761834933e-05}
{"train_loss": 0.1274845004081726, "global_step": 152895, "epoch": 1717, "lr": 4.025308900303426e-05}
{"train_loss": 0.11265010386705399, "global_step": 152896, "epoch": 1717, "lr": 4.0252520389029564e-05}
{"train_loss": 0.076618991792202, "global_step": 152897, "epoch": 1717, "lr": 4.0251951776335274e-05}
{"train_loss": 0.10761157423257828, "global_step": 152898, "epoch": 1717, "lr": 4.025138316495152e-05}
{"train_loss": 0.11397787928581238, "global_step": 152899, "epoch": 1717, "lr": 4.0250814554878317e-05}
{"train_loss": 0.054304156452417374, "global_step": 152900, "epoch": 1717, "lr": 4.025024594611579e-05}
{"train_loss": 0.10499959845054016, "global_step": 152901, "epoch": 1717, "lr": 4.024967733866399e-05, "val_loss": 6.526222229003906}
{"train_loss": 0.09673568606376648, "global_step": 152902, "epoch": 1718, "lr": 4.024910873252301e-05}
{"train_loss": 0.14279192686080933, "global_step": 152903, "epoch": 1718, "lr": 4.0248540127692916e-05}
{"train_loss": 0.10211458057165146, "global_step": 152904, "epoch": 1718, "lr": 4.024797152417379e-05}
{"train_loss": 0.10263840854167938, "global_step": 152905, "epoch": 1718, "lr": 4.024740292196571e-05}
{"train_loss": 0.15985089540481567, "global_step": 152906, "epoch": 1718, "lr": 4.024683432106873e-05}
{"train_loss": 0.06855139881372452, "global_step": 152907, "epoch": 1718, "lr": 4.0246265721482965e-05}
{"train_loss": 0.1244470551609993, "global_step": 152908, "epoch": 1718, "lr": 4.024569712320845e-05}
{"train_loss": 0.09972654283046722, "global_step": 152909, "epoch": 1718, "lr": 4.02451285262453e-05}
{"train_loss": 0.0922580137848854, "global_step": 152910, "epoch": 1718, "lr": 4.024455993059356e-05}
{"train_loss": 0.06480453163385391, "global_step": 152911, "epoch": 1718, "lr": 4.0243991336253335e-05}
{"train_loss": 0.11395787447690964, "global_step": 152912, "epoch": 1718, "lr": 4.024342274322467e-05}
{"train_loss": 0.0830736756324768, "global_step": 152913, "epoch": 1718, "lr": 4.024285415150767e-05}
{"train_loss": 0.08382642269134521, "global_step": 152914, "epoch": 1718, "lr": 4.024228556110239e-05}
{"train_loss": 0.1334053874015808, "global_step": 152915, "epoch": 1718, "lr": 4.0241716972008933e-05}
{"train_loss": 0.060537777841091156, "global_step": 152916, "epoch": 1718, "lr": 4.024114838422733e-05}
{"train_loss": 0.11529377102851868, "global_step": 152917, "epoch": 1718, "lr": 4.024057979775772e-05}
{"train_loss": 0.14221930503845215, "global_step": 152918, "epoch": 1718, "lr": 4.024001121260011e-05}
{"train_loss": 0.1476302146911621, "global_step": 152919, "epoch": 1718, "lr": 4.023944262875464e-05}
{"train_loss": 0.1311911791563034, "global_step": 152920, "epoch": 1718, "lr": 4.023887404622133e-05}
{"train_loss": 0.13346615433692932, "global_step": 152921, "epoch": 1718, "lr": 4.02383054650003e-05}
{"train_loss": 0.08587773889303207, "global_step": 152922, "epoch": 1718, "lr": 4.0237736885091616e-05}
{"train_loss": 0.1333855837583542, "global_step": 152923, "epoch": 1718, "lr": 4.023716830649534e-05}
{"train_loss": 0.1109774112701416, "global_step": 152924, "epoch": 1718, "lr": 4.0236599729211577e-05}
{"train_loss": 0.10098955035209656, "global_step": 152925, "epoch": 1718, "lr": 4.023603115324036e-05}
{"train_loss": 0.06081898882985115, "global_step": 152926, "epoch": 1718, "lr": 4.0235462578581807e-05}
{"train_loss": 0.11998069286346436, "global_step": 152927, "epoch": 1718, "lr": 4.0234894005235965e-05}
{"train_loss": 0.15211831033229828, "global_step": 152928, "epoch": 1718, "lr": 4.0234325433202936e-05}
{"train_loss": 0.10381430387496948, "global_step": 152929, "epoch": 1718, "lr": 4.023375686248276e-05}
{"train_loss": 0.10901745408773422, "global_step": 152930, "epoch": 1718, "lr": 4.023318829307556e-05}
{"train_loss": 0.09273846447467804, "global_step": 152931, "epoch": 1718, "lr": 4.023261972498137e-05}
{"train_loss": 0.1173085942864418, "global_step": 152932, "epoch": 1718, "lr": 4.02320511582003e-05}
{"train_loss": 0.0997343584895134, "global_step": 152933, "epoch": 1718, "lr": 4.02314825927324e-05}
{"train_loss": 0.07121597230434418, "global_step": 152934, "epoch": 1718, "lr": 4.023091402857777e-05}
{"train_loss": 0.05288684740662575, "global_step": 152935, "epoch": 1718, "lr": 4.023034546573646e-05}
{"train_loss": 0.08260003477334976, "global_step": 152936, "epoch": 1718, "lr": 4.0229776904208586e-05}
{"train_loss": 0.09094411134719849, "global_step": 152937, "epoch": 1718, "lr": 4.022920834399417e-05}
{"train_loss": 0.18539386987686157, "global_step": 152938, "epoch": 1718, "lr": 4.022863978509333e-05}
{"train_loss": 0.04828668013215065, "global_step": 152939, "epoch": 1718, "lr": 4.022807122750613e-05}
{"train_loss": 0.10914406180381775, "global_step": 152940, "epoch": 1718, "lr": 4.022750267123264e-05}
{"train_loss": 0.08958081156015396, "global_step": 152941, "epoch": 1718, "lr": 4.022693411627296e-05}
{"train_loss": 0.10499254614114761, "global_step": 152942, "epoch": 1718, "lr": 4.022636556262713e-05}
{"train_loss": 0.07325921952724457, "global_step": 152943, "epoch": 1718, "lr": 4.0225797010295274e-05}
{"train_loss": 0.07767786830663681, "global_step": 152944, "epoch": 1718, "lr": 4.022522845927741e-05}
{"train_loss": 0.2075270116329193, "global_step": 152945, "epoch": 1718, "lr": 4.022465990957367e-05}
{"train_loss": 0.10581924021244049, "global_step": 152946, "epoch": 1718, "lr": 4.022409136118408e-05}
{"train_loss": 0.0894046202301979, "global_step": 152947, "epoch": 1718, "lr": 4.022352281410877e-05}
{"train_loss": 0.10753101110458374, "global_step": 152948, "epoch": 1718, "lr": 4.0222954268347754e-05}
{"train_loss": 0.11135746538639069, "global_step": 152949, "epoch": 1718, "lr": 4.022238572390118e-05}
{"train_loss": 0.0905810296535492, "global_step": 152950, "epoch": 1718, "lr": 4.022181718076906e-05}
{"train_loss": 0.15836387872695923, "global_step": 152951, "epoch": 1718, "lr": 4.0221248638951516e-05}
{"train_loss": 0.11080016195774078, "global_step": 152952, "epoch": 1718, "lr": 4.0220680098448584e-05}
{"train_loss": 0.13310576975345612, "global_step": 152953, "epoch": 1718, "lr": 4.022011155926039e-05}
{"train_loss": 0.10588548332452774, "global_step": 152954, "epoch": 1718, "lr": 4.021954302138695e-05}
{"train_loss": 0.06372149288654327, "global_step": 152955, "epoch": 1718, "lr": 4.02189744848284e-05}
{"train_loss": 0.07411585003137589, "global_step": 152956, "epoch": 1718, "lr": 4.021840594958477e-05}
{"train_loss": 0.06324049830436707, "global_step": 152957, "epoch": 1718, "lr": 4.0217837415656165e-05}
{"train_loss": 0.08920639753341675, "global_step": 152958, "epoch": 1718, "lr": 4.021726888304266e-05}
{"train_loss": 0.14436358213424683, "global_step": 152959, "epoch": 1718, "lr": 4.021670035174431e-05}
{"train_loss": 0.12249929457902908, "global_step": 152960, "epoch": 1718, "lr": 4.021613182176122e-05}
{"train_loss": 0.11256559193134308, "global_step": 152961, "epoch": 1718, "lr": 4.021556329309344e-05}
{"train_loss": 0.10603901743888855, "global_step": 152962, "epoch": 1718, "lr": 4.0214994765741065e-05}
{"train_loss": 0.11084458231925964, "global_step": 152963, "epoch": 1718, "lr": 4.021442623970416e-05}
{"train_loss": 0.08731279522180557, "global_step": 152964, "epoch": 1718, "lr": 4.021385771498282e-05}
{"train_loss": 0.08926825225353241, "global_step": 152965, "epoch": 1718, "lr": 4.021328919157709e-05}
{"train_loss": 0.11801347881555557, "global_step": 152966, "epoch": 1718, "lr": 4.021272066948708e-05}
{"train_loss": 0.06954348832368851, "global_step": 152967, "epoch": 1718, "lr": 4.021215214871283e-05}
{"train_loss": 0.07876954972743988, "global_step": 152968, "epoch": 1718, "lr": 4.0211583629254454e-05}
{"train_loss": 0.13197962939739227, "global_step": 152969, "epoch": 1718, "lr": 4.0211015111112e-05}
{"train_loss": 0.047741010785102844, "global_step": 152970, "epoch": 1718, "lr": 4.021044659428557e-05}
{"train_loss": 0.21187061071395874, "global_step": 152971, "epoch": 1718, "lr": 4.020987807877521e-05}
{"train_loss": 0.13922788202762604, "global_step": 152972, "epoch": 1718, "lr": 4.020930956458103e-05}
{"train_loss": 0.08937345445156097, "global_step": 152973, "epoch": 1718, "lr": 4.0208741051703066e-05}
{"train_loss": 0.11483095586299896, "global_step": 152974, "epoch": 1718, "lr": 4.020817254014144e-05}
{"train_loss": 0.11249256134033203, "global_step": 152975, "epoch": 1718, "lr": 4.0207604029896186e-05}
{"train_loss": 0.06312131881713867, "global_step": 152976, "epoch": 1718, "lr": 4.02070355209674e-05}
{"train_loss": 0.10775373876094818, "global_step": 152977, "epoch": 1718, "lr": 4.0206467013355174e-05}
{"train_loss": 0.10261041671037674, "global_step": 152978, "epoch": 1718, "lr": 4.020589850705957e-05}
{"train_loss": 0.14876212179660797, "global_step": 152979, "epoch": 1718, "lr": 4.020533000208065e-05}
{"train_loss": 0.06940209865570068, "global_step": 152980, "epoch": 1718, "lr": 4.020476149841852e-05}
{"train_loss": 0.07279795408248901, "global_step": 152981, "epoch": 1718, "lr": 4.0204192996073234e-05}
{"train_loss": 0.20016393065452576, "global_step": 152982, "epoch": 1718, "lr": 4.020362449504487e-05}
{"train_loss": 0.12141057848930359, "global_step": 152983, "epoch": 1718, "lr": 4.020305599533352e-05}
{"train_loss": 0.050715867429971695, "global_step": 152984, "epoch": 1718, "lr": 4.020248749693923e-05}
{"train_loss": 0.12699651718139648, "global_step": 152985, "epoch": 1718, "lr": 4.020191899986212e-05}
{"train_loss": 0.13198208808898926, "global_step": 152986, "epoch": 1718, "lr": 4.020135050410222e-05}
{"train_loss": 0.14082956314086914, "global_step": 152987, "epoch": 1718, "lr": 4.020078200965965e-05}
{"train_loss": 0.08805537968873978, "global_step": 152988, "epoch": 1718, "lr": 4.0200213516534444e-05}
{"train_loss": 0.1790267676115036, "global_step": 152989, "epoch": 1718, "lr": 4.0199645024726725e-05}
{"train_loss": 0.10782539794284306, "global_step": 152990, "epoch": 1718, "lr": 4.019907653423652e-05, "val_loss": 6.388188362121582}
{"train_loss": 0.13633807003498077, "global_step": 152991, "epoch": 1719, "lr": 4.019850804506395e-05}
{"train_loss": 0.09424279630184174, "global_step": 152992, "epoch": 1719, "lr": 4.019793955720905e-05}
{"train_loss": 0.09928691387176514, "global_step": 152993, "epoch": 1719, "lr": 4.019737107067192e-05}
{"train_loss": 0.06922270357608795, "global_step": 152994, "epoch": 1719, "lr": 4.0196802585452646e-05}
{"train_loss": 0.12241482734680176, "global_step": 152995, "epoch": 1719, "lr": 4.019623410155128e-05}
{"train_loss": 0.08546525985002518, "global_step": 152996, "epoch": 1719, "lr": 4.019566561896793e-05}
{"train_loss": 0.04988110810518265, "global_step": 152997, "epoch": 1719, "lr": 4.0195097137702644e-05}
{"train_loss": 0.1257888376712799, "global_step": 152998, "epoch": 1719, "lr": 4.0194528657755506e-05}
{"train_loss": 0.053263355046510696, "global_step": 152999, "epoch": 1719, "lr": 4.0193960179126596e-05}
{"train_loss": 0.06658779829740524, "global_step": 153000, "epoch": 1719, "lr": 4.019339170181599e-05}
{"train_loss": 0.0726579949259758, "global_step": 153001, "epoch": 1719, "lr": 4.019282322582376e-05}
{"train_loss": 0.09958799183368683, "global_step": 153002, "epoch": 1719, "lr": 4.0192254751149996e-05}
{"train_loss": 0.09504950791597366, "global_step": 153003, "epoch": 1719, "lr": 4.019168627779475e-05}
{"train_loss": 0.09779953211545944, "global_step": 153004, "epoch": 1719, "lr": 4.019111780575812e-05}
{"train_loss": 0.12713055312633514, "global_step": 153005, "epoch": 1719, "lr": 4.019054933504016e-05}
{"train_loss": 0.12857428193092346, "global_step": 153006, "epoch": 1719, "lr": 4.018998086564098e-05}
{"train_loss": 0.10489052534103394, "global_step": 153007, "epoch": 1719, "lr": 4.018941239756062e-05}
{"train_loss": 0.11297232657670975, "global_step": 153008, "epoch": 1719, "lr": 4.0188843930799195e-05}
{"train_loss": 0.14274857938289642, "global_step": 153009, "epoch": 1719, "lr": 4.018827546535673e-05}
{"train_loss": 0.07338099181652069, "global_step": 153010, "epoch": 1719, "lr": 4.018770700123337e-05}
{"train_loss": 0.11242739856243134, "global_step": 153011, "epoch": 1719, "lr": 4.0187138538429115e-05}
{"train_loss": 0.0712708905339241, "global_step": 153012, "epoch": 1719, "lr": 4.0186570076944084e-05}
{"train_loss": 0.07978206127882004, "global_step": 153013, "epoch": 1719, "lr": 4.018600161677838e-05}
{"train_loss": 0.1016557365655899, "global_step": 153014, "epoch": 1719, "lr": 4.018543315793202e-05}
{"train_loss": 0.194056898355484, "global_step": 153015, "epoch": 1719, "lr": 4.018486470040513e-05}
{"train_loss": 0.15096372365951538, "global_step": 153016, "epoch": 1719, "lr": 4.018429624419776e-05}
{"train_loss": 0.06661701947450638, "global_step": 153017, "epoch": 1719, "lr": 4.018372778930998e-05}
{"train_loss": 0.1280456781387329, "global_step": 153018, "epoch": 1719, "lr": 4.0183159335741884e-05}
{"train_loss": 0.15599925816059113, "global_step": 153019, "epoch": 1719, "lr": 4.018259088349355e-05}
{"train_loss": 0.08322247862815857, "global_step": 153020, "epoch": 1719, "lr": 4.018202243256504e-05}
{"train_loss": 0.09434029459953308, "global_step": 153021, "epoch": 1719, "lr": 4.018145398295645e-05}
{"train_loss": 0.0765744000673294, "global_step": 153022, "epoch": 1719, "lr": 4.018088553466782e-05}
{"train_loss": 0.07544895261526108, "global_step": 153023, "epoch": 1719, "lr": 4.018031708769927e-05}
{"train_loss": 0.0707785040140152, "global_step": 153024, "epoch": 1719, "lr": 4.017974864205084e-05}
{"train_loss": 0.13385455310344696, "global_step": 153025, "epoch": 1719, "lr": 4.0179180197722645e-05}
{"train_loss": 0.06471234560012817, "global_step": 153026, "epoch": 1719, "lr": 4.0178611754714716e-05}
{"train_loss": 0.09367059916257858, "global_step": 153027, "epoch": 1719, "lr": 4.017804331302718e-05}
{"train_loss": 0.12349030375480652, "global_step": 153028, "epoch": 1719, "lr": 4.017747487266005e-05}
{"train_loss": 0.07543303072452545, "global_step": 153029, "epoch": 1719, "lr": 4.017690643361345e-05}
{"train_loss": 0.06711135059595108, "global_step": 153030, "epoch": 1719, "lr": 4.017633799588747e-05}
{"train_loss": 0.07818607240915298, "global_step": 153031, "epoch": 1719, "lr": 4.017576955948213e-05}
{"train_loss": 0.0878266841173172, "global_step": 153032, "epoch": 1719, "lr": 4.0175201124397567e-05}
{"train_loss": 0.09574005752801895, "global_step": 153033, "epoch": 1719, "lr": 4.017463269063381e-05}
{"train_loss": 0.07978567481040955, "global_step": 153034, "epoch": 1719, "lr": 4.0174064258190964e-05}
{"train_loss": 0.08602587133646011, "global_step": 153035, "epoch": 1719, "lr": 4.017349582706909e-05}
{"train_loss": 0.05150950700044632, "global_step": 153036, "epoch": 1719, "lr": 4.017292739726827e-05}
{"train_loss": 0.1196187362074852, "global_step": 153037, "epoch": 1719, "lr": 4.017235896878858e-05}
{"train_loss": 0.06054891645908356, "global_step": 153038, "epoch": 1719, "lr": 4.0171790541630095e-05}
{"train_loss": 0.0926506370306015, "global_step": 153039, "epoch": 1719, "lr": 4.017122211579289e-05}
{"train_loss": 0.1598334014415741, "global_step": 153040, "epoch": 1719, "lr": 4.017065369127707e-05}
{"train_loss": 0.06919819861650467, "global_step": 153041, "epoch": 1719, "lr": 4.017008526808265e-05}
{"train_loss": 0.0932416319847107, "global_step": 153042, "epoch": 1719, "lr": 4.016951684620977e-05}
{"train_loss": 0.15044787526130676, "global_step": 153043, "epoch": 1719, "lr": 4.0168948425658456e-05}
{"train_loss": 0.10543840378522873, "global_step": 153044, "epoch": 1719, "lr": 4.016838000642883e-05}
{"train_loss": 0.048690639436244965, "global_step": 153045, "epoch": 1719, "lr": 4.016781158852092e-05}
{"train_loss": 0.11590446531772614, "global_step": 153046, "epoch": 1719, "lr": 4.016724317193485e-05}
{"train_loss": 0.13686366379261017, "global_step": 153047, "epoch": 1719, "lr": 4.016667475667065e-05}
{"train_loss": 0.0843551903963089, "global_step": 153048, "epoch": 1719, "lr": 4.016610634272843e-05}
{"train_loss": 0.04721691459417343, "global_step": 153049, "epoch": 1719, "lr": 4.016553793010827e-05}
{"train_loss": 0.10220344364643097, "global_step": 153050, "epoch": 1719, "lr": 4.0164969518810216e-05}
{"train_loss": 0.1661529839038849, "global_step": 153051, "epoch": 1719, "lr": 4.016440110883438e-05}
{"train_loss": 0.08354394882917404, "global_step": 153052, "epoch": 1719, "lr": 4.01638327001808e-05}
{"train_loss": 0.04965756833553314, "global_step": 153053, "epoch": 1719, "lr": 4.0163264292849604e-05}
{"train_loss": 0.10959797352552414, "global_step": 153054, "epoch": 1719, "lr": 4.016269588684082e-05}
{"train_loss": 0.15350741147994995, "global_step": 153055, "epoch": 1719, "lr": 4.016212748215453e-05}
{"train_loss": 0.11358354985713959, "global_step": 153056, "epoch": 1719, "lr": 4.016155907879084e-05}
{"train_loss": 0.18594487011432648, "global_step": 153057, "epoch": 1719, "lr": 4.01609906767498e-05}
{"train_loss": 0.13667818903923035, "global_step": 153058, "epoch": 1719, "lr": 4.016042227603149e-05}
{"train_loss": 0.06564019620418549, "global_step": 153059, "epoch": 1719, "lr": 4.015985387663601e-05}
{"train_loss": 0.11078669875860214, "global_step": 153060, "epoch": 1719, "lr": 4.0159285478563396e-05}
{"train_loss": 0.0658956840634346, "global_step": 153061, "epoch": 1719, "lr": 4.0158717081813766e-05}
{"train_loss": 0.0589301735162735, "global_step": 153062, "epoch": 1719, "lr": 4.0158148686387154e-05}
{"train_loss": 0.072390116751194, "global_step": 153063, "epoch": 1719, "lr": 4.015758029228368e-05}
{"train_loss": 0.06198617443442345, "global_step": 153064, "epoch": 1719, "lr": 4.015701189950338e-05}
{"train_loss": 0.12280937284231186, "global_step": 153065, "epoch": 1719, "lr": 4.015644350804635e-05}
{"train_loss": 0.12391822040081024, "global_step": 153066, "epoch": 1719, "lr": 4.015587511791268e-05}
{"train_loss": 0.08071733266115189, "global_step": 153067, "epoch": 1719, "lr": 4.015530672910242e-05}
{"train_loss": 0.06941637396812439, "global_step": 153068, "epoch": 1719, "lr": 4.015473834161568e-05}
{"train_loss": 0.08371876180171967, "global_step": 153069, "epoch": 1719, "lr": 4.015416995545249e-05}
{"train_loss": 0.06569898873567581, "global_step": 153070, "epoch": 1719, "lr": 4.015360157061298e-05}
{"train_loss": 0.0726795643568039, "global_step": 153071, "epoch": 1719, "lr": 4.015303318709717e-05}
{"train_loss": 0.10524684935808182, "global_step": 153072, "epoch": 1719, "lr": 4.015246480490519e-05}
{"train_loss": 0.08769568055868149, "global_step": 153073, "epoch": 1719, "lr": 4.015189642403706e-05}
{"train_loss": 0.06990446150302887, "global_step": 153074, "epoch": 1719, "lr": 4.015132804449292e-05}
{"train_loss": 0.20341609418392181, "global_step": 153075, "epoch": 1719, "lr": 4.01507596662728e-05}
{"train_loss": 0.06303922086954117, "global_step": 153076, "epoch": 1719, "lr": 4.015019128937679e-05}
{"train_loss": 0.1373283565044403, "global_step": 153077, "epoch": 1719, "lr": 4.014962291380496e-05}
{"train_loss": 0.07546457648277283, "global_step": 153078, "epoch": 1719, "lr": 4.0149054539557405e-05}
{"train_loss": 0.09799155350146668, "global_step": 153079, "epoch": 1719, "lr": 4.014848616663419e-05, "val_loss": 6.5209269523620605}
{"train_loss": 0.12667864561080933, "global_step": 153080, "epoch": 1720, "lr": 4.014791779503539e-05}
{"train_loss": 0.09966085106134415, "global_step": 153081, "epoch": 1720, "lr": 4.014734942476107e-05}
{"train_loss": 0.08475007116794586, "global_step": 153082, "epoch": 1720, "lr": 4.0146781055811334e-05}
{"train_loss": 0.08845780044794083, "global_step": 153083, "epoch": 1720, "lr": 4.014621268818623e-05}
{"train_loss": 0.05718230456113815, "global_step": 153084, "epoch": 1720, "lr": 4.014564432188584e-05}
{"train_loss": 0.11058032512664795, "global_step": 153085, "epoch": 1720, "lr": 4.014507595691027e-05}
{"train_loss": 0.07832902669906616, "global_step": 153086, "epoch": 1720, "lr": 4.014450759325955e-05}
{"train_loss": 0.11545708775520325, "global_step": 153087, "epoch": 1720, "lr": 4.0143939230933816e-05}
{"train_loss": 0.08223959058523178, "global_step": 153088, "epoch": 1720, "lr": 4.0143370869933076e-05}
{"train_loss": 0.1292775273323059, "global_step": 153089, "epoch": 1720, "lr": 4.014280251025747e-05}
{"train_loss": 0.05343988537788391, "global_step": 153090, "epoch": 1720, "lr": 4.014223415190701e-05}
{"train_loss": 0.1635851263999939, "global_step": 153091, "epoch": 1720, "lr": 4.014166579488184e-05}
{"train_loss": 0.10590120404958725, "global_step": 153092, "epoch": 1720, "lr": 4.0141097439181975e-05}
{"train_loss": 0.08430963009595871, "global_step": 153093, "epoch": 1720, "lr": 4.0140529084807546e-05}
{"train_loss": 0.19465351104736328, "global_step": 153094, "epoch": 1720, "lr": 4.013996073175858e-05}
{"train_loss": 0.1032467857003212, "global_step": 153095, "epoch": 1720, "lr": 4.013939238003519e-05}
{"train_loss": 0.15197792649269104, "global_step": 153096, "epoch": 1720, "lr": 4.013882402963743e-05}
{"train_loss": 0.08305221796035767, "global_step": 153097, "epoch": 1720, "lr": 4.0138255680565384e-05}
{"train_loss": 0.07780010998249054, "global_step": 153098, "epoch": 1720, "lr": 4.013768733281914e-05}
{"train_loss": 0.12642596662044525, "global_step": 153099, "epoch": 1720, "lr": 4.013711898639876e-05}
{"train_loss": 0.08790289610624313, "global_step": 153100, "epoch": 1720, "lr": 4.013655064130432e-05}
{"train_loss": 0.06777377426624298, "global_step": 153101, "epoch": 1720, "lr": 4.013598229753591e-05}
{"train_loss": 0.1330767124891281, "global_step": 153102, "epoch": 1720, "lr": 4.013541395509357e-05}
{"train_loss": 0.09512302279472351, "global_step": 153103, "epoch": 1720, "lr": 4.013484561397741e-05}
{"train_loss": 0.11579813808202744, "global_step": 153104, "epoch": 1720, "lr": 4.013427727418752e-05}
{"train_loss": 0.17430710792541504, "global_step": 153105, "epoch": 1720, "lr": 4.013370893572394e-05}
{"train_loss": 0.05066434293985367, "global_step": 153106, "epoch": 1720, "lr": 4.0133140598586786e-05}
{"train_loss": 0.1120913177728653, "global_step": 153107, "epoch": 1720, "lr": 4.013257226277608e-05}
{"train_loss": 0.12338883429765701, "global_step": 153108, "epoch": 1720, "lr": 4.013200392829196e-05}
{"train_loss": 0.0790400430560112, "global_step": 153109, "epoch": 1720, "lr": 4.0131435595134445e-05}
{"train_loss": 0.06107671558856964, "global_step": 153110, "epoch": 1720, "lr": 4.0130867263303656e-05}
{"train_loss": 0.18303680419921875, "global_step": 153111, "epoch": 1720, "lr": 4.013029893279964e-05}
{"train_loss": 0.09726248681545258, "global_step": 153112, "epoch": 1720, "lr": 4.01297306036225e-05}
{"train_loss": 0.14014261960983276, "global_step": 153113, "epoch": 1720, "lr": 4.012916227577229e-05}
{"train_loss": 0.1667795479297638, "global_step": 153114, "epoch": 1720, "lr": 4.012859394924908e-05}
{"train_loss": 0.13556356728076935, "global_step": 153115, "epoch": 1720, "lr": 4.0128025624052975e-05}
{"train_loss": 0.13395820558071136, "global_step": 153116, "epoch": 1720, "lr": 4.012745730018403e-05}
{"train_loss": 0.14150157570838928, "global_step": 153117, "epoch": 1720, "lr": 4.0126888977642335e-05}
{"train_loss": 0.13263647258281708, "global_step": 153118, "epoch": 1720, "lr": 4.0126320656427966e-05}
{"train_loss": 0.1391078680753708, "global_step": 153119, "epoch": 1720, "lr": 4.012575233654097e-05}
{"train_loss": 0.13506381213665009, "global_step": 153120, "epoch": 1720, "lr": 4.0125184017981455e-05}
{"train_loss": 0.12384772300720215, "global_step": 153121, "epoch": 1720, "lr": 4.01246157007495e-05}
{"train_loss": 0.07192374765872955, "global_step": 153122, "epoch": 1720, "lr": 4.0124047384845157e-05}
{"train_loss": 0.08307969570159912, "global_step": 153123, "epoch": 1720, "lr": 4.0123479070268524e-05}
{"train_loss": 0.11389926075935364, "global_step": 153124, "epoch": 1720, "lr": 4.012291075701966e-05}
{"train_loss": 0.0903102457523346, "global_step": 153125, "epoch": 1720, "lr": 4.0122342445098665e-05}
{"train_loss": 0.11775589734315872, "global_step": 153126, "epoch": 1720, "lr": 4.012177413450558e-05}
{"train_loss": 0.06906556338071823, "global_step": 153127, "epoch": 1720, "lr": 4.012120582524053e-05}
{"train_loss": 0.12279824167490005, "global_step": 153128, "epoch": 1720, "lr": 4.012063751730353e-05}
{"train_loss": 0.10359590500593185, "global_step": 153129, "epoch": 1720, "lr": 4.012006921069472e-05}
{"train_loss": 0.16051757335662842, "global_step": 153130, "epoch": 1720, "lr": 4.011950090541412e-05}
{"train_loss": 0.11431550234556198, "global_step": 153131, "epoch": 1720, "lr": 4.0118932601461854e-05}
{"train_loss": 0.2275347262620926, "global_step": 153132, "epoch": 1720, "lr": 4.0118364298837974e-05}
{"train_loss": 0.11046729236841202, "global_step": 153133, "epoch": 1720, "lr": 4.0117795997542554e-05}
{"train_loss": 0.11223820596933365, "global_step": 153134, "epoch": 1720, "lr": 4.0117227697575674e-05}
{"train_loss": 0.11227887123823166, "global_step": 153135, "epoch": 1720, "lr": 4.011665939893742e-05}
{"train_loss": 0.14810822904109955, "global_step": 153136, "epoch": 1720, "lr": 4.011609110162785e-05}
{"train_loss": 0.10948840528726578, "global_step": 153137, "epoch": 1720, "lr": 4.011552280564707e-05}
{"train_loss": 0.07528480142354965, "global_step": 153138, "epoch": 1720, "lr": 4.0114954510995115e-05}
{"train_loss": 0.0982324630022049, "global_step": 153139, "epoch": 1720, "lr": 4.0114386217672084e-05}
{"train_loss": 0.13086703419685364, "global_step": 153140, "epoch": 1720, "lr": 4.011381792567807e-05}
{"train_loss": 0.10060638189315796, "global_step": 153141, "epoch": 1720, "lr": 4.0113249635013114e-05}
{"train_loss": 0.09050070494413376, "global_step": 153142, "epoch": 1720, "lr": 4.011268134567733e-05}
{"train_loss": 0.15259195864200592, "global_step": 153143, "epoch": 1720, "lr": 4.011211305767076e-05}
{"train_loss": 0.12039745599031448, "global_step": 153144, "epoch": 1720, "lr": 4.0111544770993516e-05}
{"train_loss": 0.05578940361738205, "global_step": 153145, "epoch": 1720, "lr": 4.011097648564563e-05}
{"train_loss": 0.12293856590986252, "global_step": 153146, "epoch": 1720, "lr": 4.0110408201627225e-05}
{"train_loss": 0.1038355827331543, "global_step": 153147, "epoch": 1720, "lr": 4.010983991893833e-05}
{"train_loss": 0.11372478306293488, "global_step": 153148, "epoch": 1720, "lr": 4.0109271637579073e-05}
{"train_loss": 0.08425477892160416, "global_step": 153149, "epoch": 1720, "lr": 4.010870335754948e-05}
{"train_loss": 0.09036710858345032, "global_step": 153150, "epoch": 1720, "lr": 4.010813507884967e-05}
{"train_loss": 0.076156847178936, "global_step": 153151, "epoch": 1720, "lr": 4.01075668014797e-05}
{"train_loss": 0.10298921167850494, "global_step": 153152, "epoch": 1720, "lr": 4.0106998525439645e-05}
{"train_loss": 0.07828475534915924, "global_step": 153153, "epoch": 1720, "lr": 4.0106430250729574e-05}
{"train_loss": 0.14009304344654083, "global_step": 153154, "epoch": 1720, "lr": 4.010586197734958e-05}
{"train_loss": 0.05648946389555931, "global_step": 153155, "epoch": 1720, "lr": 4.010529370529973e-05}
{"train_loss": 0.09769397228956223, "global_step": 153156, "epoch": 1720, "lr": 4.010472543458009e-05}
{"train_loss": 0.09938377887010574, "global_step": 153157, "epoch": 1720, "lr": 4.0104157165190775e-05}
{"train_loss": 0.07919299602508545, "global_step": 153158, "epoch": 1720, "lr": 4.010358889713181e-05}
{"train_loss": 0.07496694475412369, "global_step": 153159, "epoch": 1720, "lr": 4.010302063040332e-05}
{"train_loss": 0.07876002788543701, "global_step": 153160, "epoch": 1720, "lr": 4.010245236500534e-05}
{"train_loss": 0.13532333076000214, "global_step": 153161, "epoch": 1720, "lr": 4.010188410093798e-05}
{"train_loss": 0.13869673013687134, "global_step": 153162, "epoch": 1720, "lr": 4.010131583820128e-05}
{"train_loss": 0.06727954745292664, "global_step": 153163, "epoch": 1720, "lr": 4.010074757679537e-05}
{"train_loss": 0.12197130918502808, "global_step": 153164, "epoch": 1720, "lr": 4.010017931672026e-05}
{"train_loss": 0.12356878817081451, "global_step": 153165, "epoch": 1720, "lr": 4.0099611057976086e-05}
{"train_loss": 0.05776339769363403, "global_step": 153166, "epoch": 1720, "lr": 4.0099042800562875e-05}
{"train_loss": 0.10924849659204483, "global_step": 153167, "epoch": 1720, "lr": 4.009847454448075e-05}
{"train_loss": 0.10860925674271048, "global_step": 153168, "epoch": 1720, "lr": 4.0097906289729745e-05, "val_loss": 6.36413049697876, "train_action_mse_error": 33.673431396484375}
{"train_loss": 0.08520665019750595, "global_step": 153169, "epoch": 1721, "lr": 4.0097338036309975e-05}
{"train_loss": 0.09600614756345749, "global_step": 153170, "epoch": 1721, "lr": 4.009676978422149e-05}
{"train_loss": 0.10046041011810303, "global_step": 153171, "epoch": 1721, "lr": 4.0096201533464366e-05}
{"train_loss": 0.10376806557178497, "global_step": 153172, "epoch": 1721, "lr": 4.009563328403869e-05}
{"train_loss": 0.1363898664712906, "global_step": 153173, "epoch": 1721, "lr": 4.009506503594454e-05}
{"train_loss": 0.1497391015291214, "global_step": 153174, "epoch": 1721, "lr": 4.009449678918198e-05}
{"train_loss": 0.11348695307970047, "global_step": 153175, "epoch": 1721, "lr": 4.009392854375109e-05}
{"train_loss": 0.10088058561086655, "global_step": 153176, "epoch": 1721, "lr": 4.0093360299651975e-05}
{"train_loss": 0.09982854127883911, "global_step": 153177, "epoch": 1721, "lr": 4.009279205688466e-05}
{"train_loss": 0.11974762380123138, "global_step": 153178, "epoch": 1721, "lr": 4.009222381544927e-05}
{"train_loss": 0.09598152339458466, "global_step": 153179, "epoch": 1721, "lr": 4.009165557534584e-05}
{"train_loss": 0.06494324654340744, "global_step": 153180, "epoch": 1721, "lr": 4.009108733657448e-05}
{"train_loss": 0.09635233879089355, "global_step": 153181, "epoch": 1721, "lr": 4.009051909913524e-05}
{"train_loss": 0.08574699610471725, "global_step": 153182, "epoch": 1721, "lr": 4.008995086302823e-05}
{"train_loss": 0.04300456866621971, "global_step": 153183, "epoch": 1721, "lr": 4.008938262825348e-05}
{"train_loss": 0.13692894577980042, "global_step": 153184, "epoch": 1721, "lr": 4.0088814394811116e-05}
{"train_loss": 0.1275332272052765, "global_step": 153185, "epoch": 1721, "lr": 4.0088246162701164e-05}
{"train_loss": 0.10663064569234848, "global_step": 153186, "epoch": 1721, "lr": 4.008767793192376e-05}
{"train_loss": 0.14135608077049255, "global_step": 153187, "epoch": 1721, "lr": 4.008710970247891e-05}
{"train_loss": 0.08585523068904877, "global_step": 153188, "epoch": 1721, "lr": 4.008654147436676e-05}
{"train_loss": 0.10363870114088058, "global_step": 153189, "epoch": 1721, "lr": 4.0085973247587325e-05}
{"train_loss": 0.11565107852220535, "global_step": 153190, "epoch": 1721, "lr": 4.008540502214073e-05}
{"train_loss": 0.12094714492559433, "global_step": 153191, "epoch": 1721, "lr": 4.0084836798027034e-05}
{"train_loss": 0.062200866639614105, "global_step": 153192, "epoch": 1721, "lr": 4.008426857524629e-05}
{"train_loss": 0.15894384682178497, "global_step": 153193, "epoch": 1721, "lr": 4.008370035379861e-05}
{"train_loss": 0.23369890451431274, "global_step": 153194, "epoch": 1721, "lr": 4.008313213368405e-05}
{"train_loss": 0.08358780294656754, "global_step": 153195, "epoch": 1721, "lr": 4.00825639149027e-05}
{"train_loss": 0.11126130074262619, "global_step": 153196, "epoch": 1721, "lr": 4.008199569745462e-05}
{"train_loss": 0.07404830306768417, "global_step": 153197, "epoch": 1721, "lr": 4.008142748133991e-05}
{"train_loss": 0.07892326265573502, "global_step": 153198, "epoch": 1721, "lr": 4.0080859266558604e-05}
{"train_loss": 0.10180506855249405, "global_step": 153199, "epoch": 1721, "lr": 4.008029105311083e-05}
{"train_loss": 0.06943123042583466, "global_step": 153200, "epoch": 1721, "lr": 4.007972284099662e-05}
{"train_loss": 0.05812330171465874, "global_step": 153201, "epoch": 1721, "lr": 4.0079154630216096e-05}
{"train_loss": 0.07400329411029816, "global_step": 153202, "epoch": 1721, "lr": 4.0078586420769276e-05}
{"train_loss": 0.07066862285137177, "global_step": 153203, "epoch": 1721, "lr": 4.0078018212656296e-05}
{"train_loss": 0.15320315957069397, "global_step": 153204, "epoch": 1721, "lr": 4.0077450005877196e-05}
{"train_loss": 0.04580972343683243, "global_step": 153205, "epoch": 1721, "lr": 4.0076881800432065e-05}
{"train_loss": 0.10535521805286407, "global_step": 153206, "epoch": 1721, "lr": 4.007631359632096e-05}
{"train_loss": 0.14728455245494843, "global_step": 153207, "epoch": 1721, "lr": 4.0075745393544e-05}
{"train_loss": 0.09551925957202911, "global_step": 153208, "epoch": 1721, "lr": 4.0075177192101214e-05}
{"train_loss": 0.09371503442525864, "global_step": 153209, "epoch": 1721, "lr": 4.0074608991992715e-05}
{"train_loss": 0.06411032378673553, "global_step": 153210, "epoch": 1721, "lr": 4.007404079321856e-05}
{"train_loss": 0.08061791211366653, "global_step": 153211, "epoch": 1721, "lr": 4.0073472595778816e-05}
{"train_loss": 0.1338409185409546, "global_step": 153212, "epoch": 1721, "lr": 4.007290439967359e-05}
{"train_loss": 0.1397678256034851, "global_step": 153213, "epoch": 1721, "lr": 4.0072336204902926e-05}
{"train_loss": 0.06875903904438019, "global_step": 153214, "epoch": 1721, "lr": 4.007176801146694e-05}
{"train_loss": 0.0833873301744461, "global_step": 153215, "epoch": 1721, "lr": 4.0071199819365654e-05}
{"train_loss": 0.1201590970158577, "global_step": 153216, "epoch": 1721, "lr": 4.007063162859919e-05}
{"train_loss": 0.13116954267024994, "global_step": 153217, "epoch": 1721, "lr": 4.0070063439167596e-05}
{"train_loss": 0.09218122065067291, "global_step": 153218, "epoch": 1721, "lr": 4.006949525107098e-05}
{"train_loss": 0.12602457404136658, "global_step": 153219, "epoch": 1721, "lr": 4.006892706430938e-05}
{"train_loss": 0.17482203245162964, "global_step": 153220, "epoch": 1721, "lr": 4.006835887888292e-05}
{"train_loss": 0.10361137986183167, "global_step": 153221, "epoch": 1721, "lr": 4.006779069479161e-05}
{"train_loss": 0.0799703449010849, "global_step": 153222, "epoch": 1721, "lr": 4.00672225120356e-05}
{"train_loss": 0.08913123607635498, "global_step": 153223, "epoch": 1721, "lr": 4.0066654330614904e-05}
{"train_loss": 0.08944398164749146, "global_step": 153224, "epoch": 1721, "lr": 4.006608615052965e-05}
{"train_loss": 0.10645190626382828, "global_step": 153225, "epoch": 1721, "lr": 4.006551797177986e-05}
{"train_loss": 0.11576668918132782, "global_step": 153226, "epoch": 1721, "lr": 4.006494979436567e-05}
{"train_loss": 0.07744492590427399, "global_step": 153227, "epoch": 1721, "lr": 4.00643816182871e-05}
{"train_loss": 0.11538621038198471, "global_step": 153228, "epoch": 1721, "lr": 4.006381344354427e-05}
{"train_loss": 0.09631026536226273, "global_step": 153229, "epoch": 1721, "lr": 4.006324527013724e-05}
{"train_loss": 0.10241932421922684, "global_step": 153230, "epoch": 1721, "lr": 4.006267709806607e-05}
{"train_loss": 0.06678208708763123, "global_step": 153231, "epoch": 1721, "lr": 4.006210892733087e-05}
{"train_loss": 0.15426692366600037, "global_step": 153232, "epoch": 1721, "lr": 4.006154075793168e-05}
{"train_loss": 0.07788420468568802, "global_step": 153233, "epoch": 1721, "lr": 4.006097258986862e-05}
{"train_loss": 0.10088451951742172, "global_step": 153234, "epoch": 1721, "lr": 4.006040442314171e-05}
{"train_loss": 0.0946340411901474, "global_step": 153235, "epoch": 1721, "lr": 4.005983625775108e-05}
{"train_loss": 0.06021404266357422, "global_step": 153236, "epoch": 1721, "lr": 4.0059268093696776e-05}
{"train_loss": 0.07999775558710098, "global_step": 153237, "epoch": 1721, "lr": 4.0058699930978887e-05}
{"train_loss": 0.13341392576694489, "global_step": 153238, "epoch": 1721, "lr": 4.005813176959747e-05}
{"train_loss": 0.14991630613803864, "global_step": 153239, "epoch": 1721, "lr": 4.005756360955264e-05}
{"train_loss": 0.06977733969688416, "global_step": 153240, "epoch": 1721, "lr": 4.0056995450844426e-05}
{"train_loss": 0.1262907087802887, "global_step": 153241, "epoch": 1721, "lr": 4.005642729347295e-05}
{"train_loss": 0.125907301902771, "global_step": 153242, "epoch": 1721, "lr": 4.005585913743824e-05}
{"train_loss": 0.10863497108221054, "global_step": 153243, "epoch": 1721, "lr": 4.005529098274043e-05}
{"train_loss": 0.11104881018400192, "global_step": 153244, "epoch": 1721, "lr": 4.005472282937953e-05}
{"train_loss": 0.12169849127531052, "global_step": 153245, "epoch": 1721, "lr": 4.005415467735568e-05}
{"train_loss": 0.06284770369529724, "global_step": 153246, "epoch": 1721, "lr": 4.0053586526668905e-05}
{"train_loss": 0.0653475895524025, "global_step": 153247, "epoch": 1721, "lr": 4.005301837731932e-05}
{"train_loss": 0.07048975676298141, "global_step": 153248, "epoch": 1721, "lr": 4.005245022930699e-05}
{"train_loss": 0.10135423392057419, "global_step": 153249, "epoch": 1721, "lr": 4.0051882082631966e-05}
{"train_loss": 0.08941494673490524, "global_step": 153250, "epoch": 1721, "lr": 4.005131393729437e-05}
{"train_loss": 0.08704541623592377, "global_step": 153251, "epoch": 1721, "lr": 4.005074579329423e-05}
{"train_loss": 0.131597101688385, "global_step": 153252, "epoch": 1721, "lr": 4.005017765063167e-05}
{"train_loss": 0.12608034908771515, "global_step": 153253, "epoch": 1721, "lr": 4.0049609509306716e-05}
{"train_loss": 0.12977507710456848, "global_step": 153254, "epoch": 1721, "lr": 4.00490413693195e-05}
{"train_loss": 0.07719305902719498, "global_step": 153255, "epoch": 1721, "lr": 4.0048473230670044e-05}
{"train_loss": 0.08700038492679596, "global_step": 153256, "epoch": 1721, "lr": 4.004790509335846e-05}
{"train_loss": 0.10267819740464178, "global_step": 153257, "epoch": 1721, "lr": 4.004733695738481e-05, "val_loss": 6.393478870391846}
{"train_loss": 0.10255405306816101, "global_step": 153258, "epoch": 1722, "lr": 4.004676882274919e-05}
{"train_loss": 0.08997489511966705, "global_step": 153259, "epoch": 1722, "lr": 4.004620068945163e-05}
{"train_loss": 0.07096344977617264, "global_step": 153260, "epoch": 1722, "lr": 4.004563255749226e-05}
{"train_loss": 0.09643680602312088, "global_step": 153261, "epoch": 1722, "lr": 4.0045064426871126e-05}
{"train_loss": 0.10681834816932678, "global_step": 153262, "epoch": 1722, "lr": 4.004449629758832e-05}
{"train_loss": 0.08757898211479187, "global_step": 153263, "epoch": 1722, "lr": 4.0043928169643895e-05}
{"train_loss": 0.0992901548743248, "global_step": 153264, "epoch": 1722, "lr": 4.004336004303796e-05}
{"train_loss": 0.15665452182292938, "global_step": 153265, "epoch": 1722, "lr": 4.004279191777056e-05}
{"train_loss": 0.14712582528591156, "global_step": 153266, "epoch": 1722, "lr": 4.004222379384178e-05}
{"train_loss": 0.09555726498365402, "global_step": 153267, "epoch": 1722, "lr": 4.004165567125173e-05}
{"train_loss": 0.09203511476516724, "global_step": 153268, "epoch": 1722, "lr": 4.004108755000043e-05}
{"train_loss": 0.12620776891708374, "global_step": 153269, "epoch": 1722, "lr": 4.0040519430088e-05}
{"train_loss": 0.095943383872509, "global_step": 153270, "epoch": 1722, "lr": 4.003995131151449e-05}
{"train_loss": 0.07267464697360992, "global_step": 153271, "epoch": 1722, "lr": 4.003938319428e-05}
{"train_loss": 0.13514716923236847, "global_step": 153272, "epoch": 1722, "lr": 4.003881507838457e-05}
{"train_loss": 0.05397013574838638, "global_step": 153273, "epoch": 1722, "lr": 4.003824696382833e-05}
{"train_loss": 0.07268723845481873, "global_step": 153274, "epoch": 1722, "lr": 4.003767885061131e-05}
{"train_loss": 0.1574055254459381, "global_step": 153275, "epoch": 1722, "lr": 4.003711073873361e-05}
{"train_loss": 0.08580370992422104, "global_step": 153276, "epoch": 1722, "lr": 4.003654262819528e-05}
{"train_loss": 0.08254417032003403, "global_step": 153277, "epoch": 1722, "lr": 4.0035974518996436e-05}
{"train_loss": 0.077259860932827, "global_step": 153278, "epoch": 1722, "lr": 4.0035406411137124e-05}
{"train_loss": 0.11101849377155304, "global_step": 153279, "epoch": 1722, "lr": 4.003483830461744e-05}
{"train_loss": 0.1033996194601059, "global_step": 153280, "epoch": 1722, "lr": 4.0034270199437426e-05}
{"train_loss": 0.06107374653220177, "global_step": 153281, "epoch": 1722, "lr": 4.003370209559722e-05}
{"train_loss": 0.09776023030281067, "global_step": 153282, "epoch": 1722, "lr": 4.0033133993096825e-05}
{"train_loss": 0.08074226975440979, "global_step": 153283, "epoch": 1722, "lr": 4.003256589193638e-05}
{"train_loss": 0.10058396309614182, "global_step": 153284, "epoch": 1722, "lr": 4.0031997792115916e-05}
{"train_loss": 0.05568847432732582, "global_step": 153285, "epoch": 1722, "lr": 4.003142969363554e-05}
{"train_loss": 0.06289421021938324, "global_step": 153286, "epoch": 1722, "lr": 4.003086159649533e-05}
{"train_loss": 0.06892309337854385, "global_step": 153287, "epoch": 1722, "lr": 4.003029350069532e-05}
{"train_loss": 0.1398787945508957, "global_step": 153288, "epoch": 1722, "lr": 4.002972540623564e-05}
{"train_loss": 0.12106410413980484, "global_step": 153289, "epoch": 1722, "lr": 4.0029157313116315e-05}
{"train_loss": 0.12997972965240479, "global_step": 153290, "epoch": 1722, "lr": 4.002858922133748e-05}
{"train_loss": 0.1314820498228073, "global_step": 153291, "epoch": 1722, "lr": 4.002802113089915e-05}
{"train_loss": 0.12345652282238007, "global_step": 153292, "epoch": 1722, "lr": 4.002745304180146e-05}
{"train_loss": 0.12767452001571655, "global_step": 153293, "epoch": 1722, "lr": 4.002688495404443e-05}
{"train_loss": 0.10057844966650009, "global_step": 153294, "epoch": 1722, "lr": 4.002631686762819e-05}
{"train_loss": 0.05332397669553757, "global_step": 153295, "epoch": 1722, "lr": 4.0025748782552764e-05}
{"train_loss": 0.07284649461507797, "global_step": 153296, "epoch": 1722, "lr": 4.002518069881828e-05}
{"train_loss": 0.0712389424443245, "global_step": 153297, "epoch": 1722, "lr": 4.002461261642476e-05}
{"train_loss": 0.08461533486843109, "global_step": 153298, "epoch": 1722, "lr": 4.002404453537234e-05}
{"train_loss": 0.13308213651180267, "global_step": 153299, "epoch": 1722, "lr": 4.002347645566104e-05}
{"train_loss": 0.10783594101667404, "global_step": 153300, "epoch": 1722, "lr": 4.002290837729098e-05}
{"train_loss": 0.08187048137187958, "global_step": 153301, "epoch": 1722, "lr": 4.00223403002622e-05}
{"train_loss": 0.0635770708322525, "global_step": 153302, "epoch": 1722, "lr": 4.002177222457481e-05}
{"train_loss": 0.14033497869968414, "global_step": 153303, "epoch": 1722, "lr": 4.002120415022886e-05}
{"train_loss": 0.10201619565486908, "global_step": 153304, "epoch": 1722, "lr": 4.002063607722445e-05}
{"train_loss": 0.15178409218788147, "global_step": 153305, "epoch": 1722, "lr": 4.002006800556163e-05}
{"train_loss": 0.18277929723262787, "global_step": 153306, "epoch": 1722, "lr": 4.00194999352405e-05}
{"train_loss": 0.08750240504741669, "global_step": 153307, "epoch": 1722, "lr": 4.001893186626114e-05}
{"train_loss": 0.1085350438952446, "global_step": 153308, "epoch": 1722, "lr": 4.0018363798623586e-05}
{"train_loss": 0.13818173110485077, "global_step": 153309, "epoch": 1722, "lr": 4.001779573232797e-05}
{"train_loss": 0.12971144914627075, "global_step": 153310, "epoch": 1722, "lr": 4.0017227667374306e-05}
{"train_loss": 0.09728068858385086, "global_step": 153311, "epoch": 1722, "lr": 4.001665960376273e-05}
{"train_loss": 0.08499681204557419, "global_step": 153312, "epoch": 1722, "lr": 4.0016091541493276e-05}
{"train_loss": 0.08732675015926361, "global_step": 153313, "epoch": 1722, "lr": 4.001552348056605e-05}
{"train_loss": 0.09727765619754791, "global_step": 153314, "epoch": 1722, "lr": 4.001495542098111e-05}
{"train_loss": 0.127559632062912, "global_step": 153315, "epoch": 1722, "lr": 4.0014387362738545e-05}
{"train_loss": 0.11158380657434464, "global_step": 153316, "epoch": 1722, "lr": 4.00138193058384e-05}
{"train_loss": 0.15349747240543365, "global_step": 153317, "epoch": 1722, "lr": 4.0013251250280805e-05}
{"train_loss": 0.058660492300987244, "global_step": 153318, "epoch": 1722, "lr": 4.001268319606578e-05}
{"train_loss": 0.08678332716226578, "global_step": 153319, "epoch": 1722, "lr": 4.001211514319343e-05}
{"train_loss": 0.10276815295219421, "global_step": 153320, "epoch": 1722, "lr": 4.001154709166385e-05}
{"train_loss": 0.052957821637392044, "global_step": 153321, "epoch": 1722, "lr": 4.0010979041477085e-05}
{"train_loss": 0.11353641748428345, "global_step": 153322, "epoch": 1722, "lr": 4.001041099263322e-05}
{"train_loss": 0.13240203261375427, "global_step": 153323, "epoch": 1722, "lr": 4.0009842945132345e-05}
{"train_loss": 0.15417887270450592, "global_step": 153324, "epoch": 1722, "lr": 4.000927489897451e-05}
{"train_loss": 0.12215229123830795, "global_step": 153325, "epoch": 1722, "lr": 4.000870685415981e-05}
{"train_loss": 0.06446084380149841, "global_step": 153326, "epoch": 1722, "lr": 4.0008138810688334e-05}
{"train_loss": 0.10588527470827103, "global_step": 153327, "epoch": 1722, "lr": 4.000757076856012e-05}
{"train_loss": 0.08762328326702118, "global_step": 153328, "epoch": 1722, "lr": 4.0007002727775286e-05}
{"train_loss": 0.12194546312093735, "global_step": 153329, "epoch": 1722, "lr": 4.000643468833387e-05}
{"train_loss": 0.08685921877622604, "global_step": 153330, "epoch": 1722, "lr": 4.000586665023598e-05}
{"train_loss": 0.07684438675642014, "global_step": 153331, "epoch": 1722, "lr": 4.000529861348166e-05}
{"train_loss": 0.08951486647129059, "global_step": 153332, "epoch": 1722, "lr": 4.0004730578071034e-05}
{"train_loss": 0.09544242918491364, "global_step": 153333, "epoch": 1722, "lr": 4.000416254400413e-05}
{"train_loss": 0.17013628780841827, "global_step": 153334, "epoch": 1722, "lr": 4.0003594511281055e-05}
{"train_loss": 0.09274154156446457, "global_step": 153335, "epoch": 1722, "lr": 4.000302647990186e-05}
{"train_loss": 0.09753888845443726, "global_step": 153336, "epoch": 1722, "lr": 4.000245844986665e-05}
{"train_loss": 0.1100696474313736, "global_step": 153337, "epoch": 1722, "lr": 4.000189042117548e-05}
{"train_loss": 0.08542830497026443, "global_step": 153338, "epoch": 1722, "lr": 4.000132239382843e-05}
{"train_loss": 0.12257567048072815, "global_step": 153339, "epoch": 1722, "lr": 4.0000754367825595e-05}
{"train_loss": 0.1280107945203781, "global_step": 153340, "epoch": 1722, "lr": 4.000018634316703e-05}
{"train_loss": 0.07383827865123749, "global_step": 153341, "epoch": 1722, "lr": 3.999961831985282e-05}
{"train_loss": 0.05844826623797417, "global_step": 153342, "epoch": 1722, "lr": 3.9999050297883034e-05}
{"train_loss": 0.10907880961894989, "global_step": 153343, "epoch": 1722, "lr": 3.999848227725776e-05}
{"train_loss": 0.1445801556110382, "global_step": 153344, "epoch": 1722, "lr": 3.999791425797707e-05}
{"train_loss": 0.05024058371782303, "global_step": 153345, "epoch": 1722, "lr": 3.9997346240041036e-05}
{"train_loss": 0.10243630593412378, "global_step": 153346, "epoch": 1722, "lr": 3.999677822344973e-05, "val_loss": 6.2516703605651855}
{"train_loss": 0.12419318407773972, "global_step": 153347, "epoch": 1723, "lr": 3.999621020820324e-05}
{"train_loss": 0.11332906037569046, "global_step": 153348, "epoch": 1723, "lr": 3.999564219430163e-05}
{"train_loss": 0.10308510810136795, "global_step": 153349, "epoch": 1723, "lr": 3.999507418174501e-05}
{"train_loss": 0.14731863141059875, "global_step": 153350, "epoch": 1723, "lr": 3.9994506170533395e-05}
{"train_loss": 0.07144191116094589, "global_step": 153351, "epoch": 1723, "lr": 3.9993938160666925e-05}
{"train_loss": 0.07883667945861816, "global_step": 153352, "epoch": 1723, "lr": 3.9993370152145624e-05}
{"train_loss": 0.13049554824829102, "global_step": 153353, "epoch": 1723, "lr": 3.999280214496962e-05}
{"train_loss": 0.07239711284637451, "global_step": 153354, "epoch": 1723, "lr": 3.999223413913894e-05}
{"train_loss": 0.10997273772954941, "global_step": 153355, "epoch": 1723, "lr": 3.9991666134653694e-05}
{"train_loss": 0.11304599046707153, "global_step": 153356, "epoch": 1723, "lr": 3.999109813151393e-05}
{"train_loss": 0.0501803457736969, "global_step": 153357, "epoch": 1723, "lr": 3.999053012971974e-05}
{"train_loss": 0.0453074648976326, "global_step": 153358, "epoch": 1723, "lr": 3.9989962129271225e-05}
{"train_loss": 0.13631248474121094, "global_step": 153359, "epoch": 1723, "lr": 3.998939413016843e-05}
{"train_loss": 0.10728632658720016, "global_step": 153360, "epoch": 1723, "lr": 3.998882613241143e-05}
{"train_loss": 0.05963771790266037, "global_step": 153361, "epoch": 1723, "lr": 3.998825813600032e-05}
{"train_loss": 0.10206311196088791, "global_step": 153362, "epoch": 1723, "lr": 3.998769014093516e-05}
{"train_loss": 0.05273718759417534, "global_step": 153363, "epoch": 1723, "lr": 3.9987122147216036e-05}
{"train_loss": 0.11163298040628433, "global_step": 153364, "epoch": 1723, "lr": 3.998655415484303e-05}
{"train_loss": 0.0807010680437088, "global_step": 153365, "epoch": 1723, "lr": 3.9985986163816194e-05}
{"train_loss": 0.05494772642850876, "global_step": 153366, "epoch": 1723, "lr": 3.9985418174135636e-05}
{"train_loss": 0.08118519932031631, "global_step": 153367, "epoch": 1723, "lr": 3.9984850185801394e-05}
{"train_loss": 0.09173033386468887, "global_step": 153368, "epoch": 1723, "lr": 3.998428219881359e-05}
{"train_loss": 0.08624633401632309, "global_step": 153369, "epoch": 1723, "lr": 3.998371421317225e-05}
{"train_loss": 0.03919926658272743, "global_step": 153370, "epoch": 1723, "lr": 3.998314622887751e-05}
{"train_loss": 0.14204609394073486, "global_step": 153371, "epoch": 1723, "lr": 3.998257824592939e-05}
{"train_loss": 0.13316696882247925, "global_step": 153372, "epoch": 1723, "lr": 3.998201026432801e-05}
{"train_loss": 0.1155235692858696, "global_step": 153373, "epoch": 1723, "lr": 3.99814422840734e-05}
{"train_loss": 0.07608973234891891, "global_step": 153374, "epoch": 1723, "lr": 3.998087430516567e-05}
{"train_loss": 0.08001817762851715, "global_step": 153375, "epoch": 1723, "lr": 3.9980306327604904e-05}
{"train_loss": 0.12318657338619232, "global_step": 153376, "epoch": 1723, "lr": 3.997973835139115e-05}
{"train_loss": 0.10242614150047302, "global_step": 153377, "epoch": 1723, "lr": 3.997917037652452e-05}
{"train_loss": 0.08599217981100082, "global_step": 153378, "epoch": 1723, "lr": 3.997860240300505e-05}
{"train_loss": 0.10317441076040268, "global_step": 153379, "epoch": 1723, "lr": 3.9978034430832835e-05}
{"train_loss": 0.07771054655313492, "global_step": 153380, "epoch": 1723, "lr": 3.9977466460007956e-05}
{"train_loss": 0.13509657979011536, "global_step": 153381, "epoch": 1723, "lr": 3.997689849053048e-05}
{"train_loss": 0.11358069628477097, "global_step": 153382, "epoch": 1723, "lr": 3.99763305224005e-05}
{"train_loss": 0.07452215254306793, "global_step": 153383, "epoch": 1723, "lr": 3.997576255561807e-05}
{"train_loss": 0.07642359286546707, "global_step": 153384, "epoch": 1723, "lr": 3.997519459018327e-05}
{"train_loss": 0.10866893827915192, "global_step": 153385, "epoch": 1723, "lr": 3.99746266260962e-05}
{"train_loss": 0.04703289270401001, "global_step": 153386, "epoch": 1723, "lr": 3.99740586633569e-05}
{"train_loss": 0.045408934354782104, "global_step": 153387, "epoch": 1723, "lr": 3.997349070196549e-05}
{"train_loss": 0.05140335112810135, "global_step": 153388, "epoch": 1723, "lr": 3.997292274192199e-05}
{"train_loss": 0.0625411719083786, "global_step": 153389, "epoch": 1723, "lr": 3.997235478322654e-05}
{"train_loss": 0.08575553447008133, "global_step": 153390, "epoch": 1723, "lr": 3.997178682587916e-05}
{"train_loss": 0.11070483177900314, "global_step": 153391, "epoch": 1723, "lr": 3.997121886987997e-05}
{"train_loss": 0.09154512733221054, "global_step": 153392, "epoch": 1723, "lr": 3.9970650915229016e-05}
{"train_loss": 0.14367759227752686, "global_step": 153393, "epoch": 1723, "lr": 3.997008296192637e-05}
{"train_loss": 0.06749532371759415, "global_step": 153394, "epoch": 1723, "lr": 3.996951500997215e-05}
{"train_loss": 0.09226039797067642, "global_step": 153395, "epoch": 1723, "lr": 3.996894705936639e-05}
{"train_loss": 0.04137786477804184, "global_step": 153396, "epoch": 1723, "lr": 3.9968379110109205e-05}
{"train_loss": 0.0781324952840805, "global_step": 153397, "epoch": 1723, "lr": 3.996781116220063e-05}
{"train_loss": 0.07315166294574738, "global_step": 153398, "epoch": 1723, "lr": 3.996724321564077e-05}
{"train_loss": 0.14304237067699432, "global_step": 153399, "epoch": 1723, "lr": 3.9966675270429685e-05}
{"train_loss": 0.1756763905286789, "global_step": 153400, "epoch": 1723, "lr": 3.996610732656747e-05}
{"train_loss": 0.08254310488700867, "global_step": 153401, "epoch": 1723, "lr": 3.996553938405418e-05}
{"train_loss": 0.12004710733890533, "global_step": 153402, "epoch": 1723, "lr": 3.996497144288991e-05}
{"train_loss": 0.10960080474615097, "global_step": 153403, "epoch": 1723, "lr": 3.996440350307471e-05}
{"train_loss": 0.0498531349003315, "global_step": 153404, "epoch": 1723, "lr": 3.996383556460869e-05}
{"train_loss": 0.043493498116731644, "global_step": 153405, "epoch": 1723, "lr": 3.9963267627491894e-05}
{"train_loss": 0.0889110416173935, "global_step": 153406, "epoch": 1723, "lr": 3.9962699691724435e-05}
{"train_loss": 0.06975133717060089, "global_step": 153407, "epoch": 1723, "lr": 3.996213175730635e-05}
{"train_loss": 0.08291857689619064, "global_step": 153408, "epoch": 1723, "lr": 3.9961563824237745e-05}
{"train_loss": 0.12319109588861465, "global_step": 153409, "epoch": 1723, "lr": 3.996099589251867e-05}
{"train_loss": 0.10673247277736664, "global_step": 153410, "epoch": 1723, "lr": 3.996042796214922e-05}
{"train_loss": 0.11235155165195465, "global_step": 153411, "epoch": 1723, "lr": 3.9959860033129485e-05}
{"train_loss": 0.09137757867574692, "global_step": 153412, "epoch": 1723, "lr": 3.9959292105459505e-05}
{"train_loss": 0.07082556188106537, "global_step": 153413, "epoch": 1723, "lr": 3.99587241791394e-05}
{"train_loss": 0.04840914160013199, "global_step": 153414, "epoch": 1723, "lr": 3.9958156254169196e-05}
{"train_loss": 0.12055744975805283, "global_step": 153415, "epoch": 1723, "lr": 3.995758833054902e-05}
{"train_loss": 0.1535765677690506, "global_step": 153416, "epoch": 1723, "lr": 3.99570204082789e-05}
{"train_loss": 0.08663853257894516, "global_step": 153417, "epoch": 1723, "lr": 3.9956452487358964e-05}
{"train_loss": 0.09039643406867981, "global_step": 153418, "epoch": 1723, "lr": 3.995588456778924e-05}
{"train_loss": 0.09867165982723236, "global_step": 153419, "epoch": 1723, "lr": 3.995531664956983e-05}
{"train_loss": 0.04167352244257927, "global_step": 153420, "epoch": 1723, "lr": 3.995474873270081e-05}
{"train_loss": 0.07840999960899353, "global_step": 153421, "epoch": 1723, "lr": 3.995418081718224e-05}
{"train_loss": 0.14907453954219818, "global_step": 153422, "epoch": 1723, "lr": 3.995361290301422e-05}
{"train_loss": 0.10988765954971313, "global_step": 153423, "epoch": 1723, "lr": 3.995304499019682e-05}
{"train_loss": 0.13446591794490814, "global_step": 153424, "epoch": 1723, "lr": 3.995247707873009e-05}
{"train_loss": 0.1108834445476532, "global_step": 153425, "epoch": 1723, "lr": 3.9951909168614143e-05}
{"train_loss": 0.08169739693403244, "global_step": 153426, "epoch": 1723, "lr": 3.995134125984902e-05}
{"train_loss": 0.08853336423635483, "global_step": 153427, "epoch": 1723, "lr": 3.995077335243484e-05}
{"train_loss": 0.10939355939626694, "global_step": 153428, "epoch": 1723, "lr": 3.9950205446371634e-05}
{"train_loss": 0.06797552853822708, "global_step": 153429, "epoch": 1723, "lr": 3.99496375416595e-05}
{"train_loss": 0.13847850263118744, "global_step": 153430, "epoch": 1723, "lr": 3.994906963829854e-05}
{"train_loss": 0.2448156327009201, "global_step": 153431, "epoch": 1723, "lr": 3.994850173628878e-05}
{"train_loss": 0.13147246837615967, "global_step": 153432, "epoch": 1723, "lr": 3.994793383563035e-05}
{"train_loss": 0.1350916028022766, "global_step": 153433, "epoch": 1723, "lr": 3.9947365936323265e-05}
{"train_loss": 0.10929220914840698, "global_step": 153434, "epoch": 1723, "lr": 3.994679803836766e-05}
{"train_loss": 0.09638595367582996, "global_step": 153435, "epoch": 1723, "lr": 3.9946230141763563e-05, "val_loss": 6.432972431182861}
{"train_loss": 0.09242064505815506, "global_step": 153436, "epoch": 1724, "lr": 3.994566224651109e-05}
{"train_loss": 0.0858689546585083, "global_step": 153437, "epoch": 1724, "lr": 3.9945094352610304e-05}
{"train_loss": 0.09822649508714676, "global_step": 153438, "epoch": 1724, "lr": 3.994452646006127e-05}
{"train_loss": 0.10685604810714722, "global_step": 153439, "epoch": 1724, "lr": 3.994395856886407e-05}
{"train_loss": 0.10786408931016922, "global_step": 153440, "epoch": 1724, "lr": 3.9943390679018786e-05}
{"train_loss": 0.07062526047229767, "global_step": 153441, "epoch": 1724, "lr": 3.994282279052548e-05}
{"train_loss": 0.09519651532173157, "global_step": 153442, "epoch": 1724, "lr": 3.994225490338426e-05}
{"train_loss": 0.11007513105869293, "global_step": 153443, "epoch": 1724, "lr": 3.994168701759516e-05}
{"train_loss": 0.1088511049747467, "global_step": 153444, "epoch": 1724, "lr": 3.99411191331583e-05}
{"train_loss": 0.06527101993560791, "global_step": 153445, "epoch": 1724, "lr": 3.99405512500737e-05}
{"train_loss": 0.10079827904701233, "global_step": 153446, "epoch": 1724, "lr": 3.993998336834148e-05}
{"train_loss": 0.05534062907099724, "global_step": 153447, "epoch": 1724, "lr": 3.993941548796173e-05}
{"train_loss": 0.08949514478445053, "global_step": 153448, "epoch": 1724, "lr": 3.993884760893448e-05}
{"train_loss": 0.06960470229387283, "global_step": 153449, "epoch": 1724, "lr": 3.993827973125984e-05}
{"train_loss": 0.053810328245162964, "global_step": 153450, "epoch": 1724, "lr": 3.9937711854937864e-05}
{"train_loss": 0.06796927005052567, "global_step": 153451, "epoch": 1724, "lr": 3.993714397996867e-05}
{"train_loss": 0.048630088567733765, "global_step": 153452, "epoch": 1724, "lr": 3.993657610635227e-05}
{"train_loss": 0.08322998881340027, "global_step": 153453, "epoch": 1724, "lr": 3.99360082340888e-05}
{"train_loss": 0.12628573179244995, "global_step": 153454, "epoch": 1724, "lr": 3.993544036317829e-05}
{"train_loss": 0.0895286500453949, "global_step": 153455, "epoch": 1724, "lr": 3.993487249362086e-05}
{"train_loss": 0.14783944189548492, "global_step": 153456, "epoch": 1724, "lr": 3.993430462541655e-05}
{"train_loss": 0.1804351508617401, "global_step": 153457, "epoch": 1724, "lr": 3.993373675856545e-05}
{"train_loss": 0.11988658457994461, "global_step": 153458, "epoch": 1724, "lr": 3.9933168893067634e-05}
{"train_loss": 0.12632504105567932, "global_step": 153459, "epoch": 1724, "lr": 3.993260102892319e-05}
{"train_loss": 0.06751873344182968, "global_step": 153460, "epoch": 1724, "lr": 3.993203316613218e-05}
{"train_loss": 0.04367345944046974, "global_step": 153461, "epoch": 1724, "lr": 3.993146530469469e-05}
{"train_loss": 0.07017726451158524, "global_step": 153462, "epoch": 1724, "lr": 3.993089744461078e-05}
{"train_loss": 0.11611117422580719, "global_step": 153463, "epoch": 1724, "lr": 3.993032958588055e-05}
{"train_loss": 0.08984280377626419, "global_step": 153464, "epoch": 1724, "lr": 3.992976172850405e-05}
{"train_loss": 0.10700885206460953, "global_step": 153465, "epoch": 1724, "lr": 3.9929193872481366e-05}
{"train_loss": 0.059119004756212234, "global_step": 153466, "epoch": 1724, "lr": 3.992862601781261e-05}
{"train_loss": 0.08660146594047546, "global_step": 153467, "epoch": 1724, "lr": 3.9928058164497796e-05}
{"train_loss": 0.09052539616823196, "global_step": 153468, "epoch": 1724, "lr": 3.992749031253706e-05}
{"train_loss": 0.1442565768957138, "global_step": 153469, "epoch": 1724, "lr": 3.9926922461930426e-05}
{"train_loss": 0.15753571689128876, "global_step": 153470, "epoch": 1724, "lr": 3.992635461267801e-05}
{"train_loss": 0.08857657760381699, "global_step": 153471, "epoch": 1724, "lr": 3.992578676477986e-05}
{"train_loss": 0.08997489511966705, "global_step": 153472, "epoch": 1724, "lr": 3.9925218918236086e-05}
{"train_loss": 0.15165667235851288, "global_step": 153473, "epoch": 1724, "lr": 3.992465107304672e-05}
{"train_loss": 0.10229319334030151, "global_step": 153474, "epoch": 1724, "lr": 3.9924083229211886e-05}
{"train_loss": 0.1892256885766983, "global_step": 153475, "epoch": 1724, "lr": 3.992351538673162e-05}
{"train_loss": 0.10085096210241318, "global_step": 153476, "epoch": 1724, "lr": 3.992294754560601e-05}
{"train_loss": 0.11047399789094925, "global_step": 153477, "epoch": 1724, "lr": 3.992237970583515e-05}
{"train_loss": 0.1292157620191574, "global_step": 153478, "epoch": 1724, "lr": 3.9921811867419104e-05}
{"train_loss": 0.1058168113231659, "global_step": 153479, "epoch": 1724, "lr": 3.992124403035793e-05}
{"train_loss": 0.10099001228809357, "global_step": 153480, "epoch": 1724, "lr": 3.992067619465175e-05}
{"train_loss": 0.10499865561723709, "global_step": 153481, "epoch": 1724, "lr": 3.992010836030058e-05}
{"train_loss": 0.14445534348487854, "global_step": 153482, "epoch": 1724, "lr": 3.991954052730456e-05}
{"train_loss": 0.12310482561588287, "global_step": 153483, "epoch": 1724, "lr": 3.9918972695663706e-05}
{"train_loss": 0.09529697895050049, "global_step": 153484, "epoch": 1724, "lr": 3.991840486537812e-05}
{"train_loss": 0.12507504224777222, "global_step": 153485, "epoch": 1724, "lr": 3.9917837036447916e-05}
{"train_loss": 0.10596071183681488, "global_step": 153486, "epoch": 1724, "lr": 3.991726920887311e-05}
{"train_loss": 0.11555487662553787, "global_step": 153487, "epoch": 1724, "lr": 3.9916701382653815e-05}
{"train_loss": 0.09554646909236908, "global_step": 153488, "epoch": 1724, "lr": 3.991613355779008e-05}
{"train_loss": 0.08162809908390045, "global_step": 153489, "epoch": 1724, "lr": 3.9915565734282025e-05}
{"train_loss": 0.058679018169641495, "global_step": 153490, "epoch": 1724, "lr": 3.991499791212968e-05}
{"train_loss": 0.12380166351795197, "global_step": 153491, "epoch": 1724, "lr": 3.991443009133315e-05}
{"train_loss": 0.18257103860378265, "global_step": 153492, "epoch": 1724, "lr": 3.991386227189249e-05}
{"train_loss": 0.07806525379419327, "global_step": 153493, "epoch": 1724, "lr": 3.99132944538078e-05}
{"train_loss": 0.13748124241828918, "global_step": 153494, "epoch": 1724, "lr": 3.991272663707915e-05}
{"train_loss": 0.17596429586410522, "global_step": 153495, "epoch": 1724, "lr": 3.99121588217066e-05}
{"train_loss": 0.11480025202035904, "global_step": 153496, "epoch": 1724, "lr": 3.9911591007690234e-05}
{"train_loss": 0.12547828257083893, "global_step": 153497, "epoch": 1724, "lr": 3.9911023195030137e-05}
{"train_loss": 0.08083266019821167, "global_step": 153498, "epoch": 1724, "lr": 3.991045538372638e-05}
{"train_loss": 0.16279982030391693, "global_step": 153499, "epoch": 1724, "lr": 3.9909887573779055e-05}
{"train_loss": 0.06697540730237961, "global_step": 153500, "epoch": 1724, "lr": 3.990931976518819e-05}
{"train_loss": 0.08000069856643677, "global_step": 153501, "epoch": 1724, "lr": 3.99087519579539e-05}
{"train_loss": 0.0701458677649498, "global_step": 153502, "epoch": 1724, "lr": 3.9908184152076276e-05}
{"train_loss": 0.15549525618553162, "global_step": 153503, "epoch": 1724, "lr": 3.990761634755535e-05}
{"train_loss": 0.08505848795175552, "global_step": 153504, "epoch": 1724, "lr": 3.9907048544391254e-05}
{"train_loss": 0.08049419522285461, "global_step": 153505, "epoch": 1724, "lr": 3.9906480742584e-05}
{"train_loss": 0.07748575508594513, "global_step": 153506, "epoch": 1724, "lr": 3.990591294213372e-05}
{"train_loss": 0.07597028464078903, "global_step": 153507, "epoch": 1724, "lr": 3.990534514304045e-05}
{"train_loss": 0.12788629531860352, "global_step": 153508, "epoch": 1724, "lr": 3.99047773453043e-05}
{"train_loss": 0.11759046465158463, "global_step": 153509, "epoch": 1724, "lr": 3.990420954892531e-05}
{"train_loss": 0.1320164054632187, "global_step": 153510, "epoch": 1724, "lr": 3.99036417539036e-05}
{"train_loss": 0.1356351375579834, "global_step": 153511, "epoch": 1724, "lr": 3.990307396023919e-05}
{"train_loss": 0.1347327083349228, "global_step": 153512, "epoch": 1724, "lr": 3.9902506167932226e-05}
{"train_loss": 0.16564227640628815, "global_step": 153513, "epoch": 1724, "lr": 3.990193837698273e-05}
{"train_loss": 0.17830996215343475, "global_step": 153514, "epoch": 1724, "lr": 3.990137058739079e-05}
{"train_loss": 0.08615147322416306, "global_step": 153515, "epoch": 1724, "lr": 3.990080279915649e-05}
{"train_loss": 0.1110190898180008, "global_step": 153516, "epoch": 1724, "lr": 3.990023501227991e-05}
{"train_loss": 0.09974176436662674, "global_step": 153517, "epoch": 1724, "lr": 3.989966722676111e-05}
{"train_loss": 0.09653652459383011, "global_step": 153518, "epoch": 1724, "lr": 3.989909944260018e-05}
{"train_loss": 0.13095179200172424, "global_step": 153519, "epoch": 1724, "lr": 3.98985316597972e-05}
{"train_loss": 0.12700609862804413, "global_step": 153520, "epoch": 1724, "lr": 3.989796387835222e-05}
{"train_loss": 0.16788633167743683, "global_step": 153521, "epoch": 1724, "lr": 3.989739609826536e-05}
{"train_loss": 0.10836874693632126, "global_step": 153522, "epoch": 1724, "lr": 3.989682831953666e-05}
{"train_loss": 0.13144399225711823, "global_step": 153523, "epoch": 1724, "lr": 3.9896260542166216e-05}
{"train_loss": 0.1074099587590507, "global_step": 153524, "epoch": 1724, "lr": 3.989569276615408e-05, "val_loss": 6.326467037200928}
{"train_loss": 0.08481787890195847, "global_step": 153525, "epoch": 1725, "lr": 3.989512499150037e-05}
{"train_loss": 0.1516827791929245, "global_step": 153526, "epoch": 1725, "lr": 3.989455721820511e-05}
{"train_loss": 0.0742882639169693, "global_step": 153527, "epoch": 1725, "lr": 3.989398944626843e-05}
{"train_loss": 0.06979458034038544, "global_step": 153528, "epoch": 1725, "lr": 3.989342167569035e-05}
{"train_loss": 0.19088336825370789, "global_step": 153529, "epoch": 1725, "lr": 3.989285390647101e-05}
{"train_loss": 0.10094506293535233, "global_step": 153530, "epoch": 1725, "lr": 3.989228613861042e-05}
{"train_loss": 0.10577159374952316, "global_step": 153531, "epoch": 1725, "lr": 3.9891718372108715e-05}
{"train_loss": 0.14403966069221497, "global_step": 153532, "epoch": 1725, "lr": 3.9891150606965926e-05}
{"train_loss": 0.1349116861820221, "global_step": 153533, "epoch": 1725, "lr": 3.9890582843182164e-05}
{"train_loss": 0.1242929995059967, "global_step": 153534, "epoch": 1725, "lr": 3.9890015080757484e-05}
{"train_loss": 0.1131940484046936, "global_step": 153535, "epoch": 1725, "lr": 3.988944731969197e-05}
{"train_loss": 0.1444130688905716, "global_step": 153536, "epoch": 1725, "lr": 3.988887955998569e-05}
{"train_loss": 0.07015533745288849, "global_step": 153537, "epoch": 1725, "lr": 3.988831180163872e-05}
{"train_loss": 0.062482800334692, "global_step": 153538, "epoch": 1725, "lr": 3.9887744044651174e-05}
{"train_loss": 0.05463330075144768, "global_step": 153539, "epoch": 1725, "lr": 3.9887176289023065e-05}
{"train_loss": 0.11694613844156265, "global_step": 153540, "epoch": 1725, "lr": 3.988660853475452e-05}
{"train_loss": 0.09922724962234497, "global_step": 153541, "epoch": 1725, "lr": 3.988604078184558e-05}
{"train_loss": 0.06684451550245285, "global_step": 153542, "epoch": 1725, "lr": 3.9885473030296365e-05}
{"train_loss": 0.10976778715848923, "global_step": 153543, "epoch": 1725, "lr": 3.98849052801069e-05}
{"train_loss": 0.06330368667840958, "global_step": 153544, "epoch": 1725, "lr": 3.988433753127731e-05}
{"train_loss": 0.17571668326854706, "global_step": 153545, "epoch": 1725, "lr": 3.9883769783807624e-05}
{"train_loss": 0.08088914304971695, "global_step": 153546, "epoch": 1725, "lr": 3.988320203769796e-05}
{"train_loss": 0.16939011216163635, "global_step": 153547, "epoch": 1725, "lr": 3.9882634292948354e-05}
{"train_loss": 0.1112750992178917, "global_step": 153548, "epoch": 1725, "lr": 3.9882066549558936e-05}
{"train_loss": 0.14658837020397186, "global_step": 153549, "epoch": 1725, "lr": 3.988149880752972e-05}
{"train_loss": 0.11458314210176468, "global_step": 153550, "epoch": 1725, "lr": 3.988093106686084e-05}
{"train_loss": 0.11571057885885239, "global_step": 153551, "epoch": 1725, "lr": 3.9880363327552314e-05}
{"train_loss": 0.1122819185256958, "global_step": 153552, "epoch": 1725, "lr": 3.987979558960428e-05}
{"train_loss": 0.07947146147489548, "global_step": 153553, "epoch": 1725, "lr": 3.987922785301678e-05}
{"train_loss": 0.11929595470428467, "global_step": 153554, "epoch": 1725, "lr": 3.987866011778988e-05}
{"train_loss": 0.14411985874176025, "global_step": 153555, "epoch": 1725, "lr": 3.987809238392368e-05}
{"train_loss": 0.06704164296388626, "global_step": 153556, "epoch": 1725, "lr": 3.987752465141823e-05}
{"train_loss": 0.1015111654996872, "global_step": 153557, "epoch": 1725, "lr": 3.987695692027365e-05}
{"train_loss": 0.13802692294120789, "global_step": 153558, "epoch": 1725, "lr": 3.9876389190489964e-05}
{"train_loss": 0.13854248821735382, "global_step": 153559, "epoch": 1725, "lr": 3.98758214620673e-05}
{"train_loss": 0.07824202626943588, "global_step": 153560, "epoch": 1725, "lr": 3.987525373500568e-05}
{"train_loss": 0.1091233417391777, "global_step": 153561, "epoch": 1725, "lr": 3.987468600930524e-05}
{"train_loss": 0.12046230584383011, "global_step": 153562, "epoch": 1725, "lr": 3.9874118284966e-05}
{"train_loss": 0.10761100798845291, "global_step": 153563, "epoch": 1725, "lr": 3.9873550561988085e-05}
{"train_loss": 0.070847287774086, "global_step": 153564, "epoch": 1725, "lr": 3.987298284037152e-05}
{"train_loss": 0.10421932488679886, "global_step": 153565, "epoch": 1725, "lr": 3.987241512011644e-05}
{"train_loss": 0.11602064222097397, "global_step": 153566, "epoch": 1725, "lr": 3.987184740122285e-05}
{"train_loss": 0.06118640676140785, "global_step": 153567, "epoch": 1725, "lr": 3.987127968369091e-05}
{"train_loss": 0.143045574426651, "global_step": 153568, "epoch": 1725, "lr": 3.987071196752062e-05}
{"train_loss": 0.12033937871456146, "global_step": 153569, "epoch": 1725, "lr": 3.9870144252712114e-05}
{"train_loss": 0.1217062920331955, "global_step": 153570, "epoch": 1725, "lr": 3.986957653926542e-05}
{"train_loss": 0.19949088990688324, "global_step": 153571, "epoch": 1725, "lr": 3.986900882718066e-05}
{"train_loss": 0.08382584154605865, "global_step": 153572, "epoch": 1725, "lr": 3.986844111645788e-05}
{"train_loss": 0.15303511917591095, "global_step": 153573, "epoch": 1725, "lr": 3.986787340709715e-05}
{"train_loss": 0.12271389365196228, "global_step": 153574, "epoch": 1725, "lr": 3.9867305699098586e-05}
{"train_loss": 0.09099046885967255, "global_step": 153575, "epoch": 1725, "lr": 3.986673799246222e-05}
{"train_loss": 0.11742215603590012, "global_step": 153576, "epoch": 1725, "lr": 3.9866170287188167e-05}
{"train_loss": 0.1270890086889267, "global_step": 153577, "epoch": 1725, "lr": 3.9865602583276454e-05}
{"train_loss": 0.1597174108028412, "global_step": 153578, "epoch": 1725, "lr": 3.986503488072721e-05}
{"train_loss": 0.1232718825340271, "global_step": 153579, "epoch": 1725, "lr": 3.986446717954048e-05}
{"train_loss": 0.10390478372573853, "global_step": 153580, "epoch": 1725, "lr": 3.986389947971636e-05}
{"train_loss": 0.15264379978179932, "global_step": 153581, "epoch": 1725, "lr": 3.9863331781254896e-05}
{"train_loss": 0.12873417139053345, "global_step": 153582, "epoch": 1725, "lr": 3.986276408415621e-05}
{"train_loss": 0.1492285132408142, "global_step": 153583, "epoch": 1725, "lr": 3.986219638842032e-05}
{"train_loss": 0.09578866511583328, "global_step": 153584, "epoch": 1725, "lr": 3.986162869404737e-05}
{"train_loss": 0.1048574447631836, "global_step": 153585, "epoch": 1725, "lr": 3.986106100103736e-05}
{"train_loss": 0.07718325406312943, "global_step": 153586, "epoch": 1725, "lr": 3.986049330939044e-05}
{"train_loss": 0.1524098813533783, "global_step": 153587, "epoch": 1725, "lr": 3.985992561910663e-05}
{"train_loss": 0.18836499750614166, "global_step": 153588, "epoch": 1725, "lr": 3.985935793018605e-05}
{"train_loss": 0.10975460708141327, "global_step": 153589, "epoch": 1725, "lr": 3.9858790242628743e-05}
{"train_loss": 0.07635398209095001, "global_step": 153590, "epoch": 1725, "lr": 3.985822255643481e-05}
{"train_loss": 0.13233380019664764, "global_step": 153591, "epoch": 1725, "lr": 3.9857654871604306e-05}
{"train_loss": 0.10089147835969925, "global_step": 153592, "epoch": 1725, "lr": 3.9857087188137306e-05}
{"train_loss": 0.13774268329143524, "global_step": 153593, "epoch": 1725, "lr": 3.985651950603392e-05}
{"train_loss": 0.1005692109465599, "global_step": 153594, "epoch": 1725, "lr": 3.985595182529418e-05}
{"train_loss": 0.057438626885414124, "global_step": 153595, "epoch": 1725, "lr": 3.9855384145918204e-05}
{"train_loss": 0.09310075640678406, "global_step": 153596, "epoch": 1725, "lr": 3.985481646790603e-05}
{"train_loss": 0.05187945067882538, "global_step": 153597, "epoch": 1725, "lr": 3.9854248791257766e-05}
{"train_loss": 0.09755288809537888, "global_step": 153598, "epoch": 1725, "lr": 3.9853681115973456e-05}
{"train_loss": 0.1278771162033081, "global_step": 153599, "epoch": 1725, "lr": 3.9853113442053216e-05}
{"train_loss": 0.0930267944931984, "global_step": 153600, "epoch": 1725, "lr": 3.9852545769497084e-05}
{"train_loss": 0.06626482307910919, "global_step": 153601, "epoch": 1725, "lr": 3.985197809830517e-05}
{"train_loss": 0.08922986686229706, "global_step": 153602, "epoch": 1725, "lr": 3.9851410428477515e-05}
{"train_loss": 0.09189043194055557, "global_step": 153603, "epoch": 1725, "lr": 3.985084276001423e-05}
{"train_loss": 0.07449976354837418, "global_step": 153604, "epoch": 1725, "lr": 3.9850275092915366e-05}
{"train_loss": 0.15380483865737915, "global_step": 153605, "epoch": 1725, "lr": 3.984970742718102e-05}
{"train_loss": 0.05252302810549736, "global_step": 153606, "epoch": 1725, "lr": 3.9849139762811246e-05}
{"train_loss": 0.07552836090326309, "global_step": 153607, "epoch": 1725, "lr": 3.984857209980614e-05}
{"train_loss": 0.10303696990013123, "global_step": 153608, "epoch": 1725, "lr": 3.984800443816576e-05}
{"train_loss": 0.11146320402622223, "global_step": 153609, "epoch": 1725, "lr": 3.984743677789021e-05}
{"train_loss": 0.07994622737169266, "global_step": 153610, "epoch": 1725, "lr": 3.984686911897953e-05}
{"train_loss": 0.09960968047380447, "global_step": 153611, "epoch": 1725, "lr": 3.984630146143381e-05}
{"train_loss": 0.15432021021842957, "global_step": 153612, "epoch": 1725, "lr": 3.9845733805253156e-05}
{"train_loss": 0.11044899164960625, "global_step": 153613, "epoch": 1725, "lr": 3.984516615043759e-05, "val_loss": 6.4357218742370605, "train_action_mse_error": 27.54769515991211}
{"train_loss": 0.1514337956905365, "global_step": 153614, "epoch": 1726, "lr": 3.984459849698725e-05}
{"train_loss": 0.10722632706165314, "global_step": 153615, "epoch": 1726, "lr": 3.9844030844902147e-05}
{"train_loss": 0.08920274674892426, "global_step": 153616, "epoch": 1726, "lr": 3.9843463194182416e-05}
{"train_loss": 0.1032220721244812, "global_step": 153617, "epoch": 1726, "lr": 3.984289554482808e-05}
{"train_loss": 0.06053125485777855, "global_step": 153618, "epoch": 1726, "lr": 3.9842327896839274e-05}
{"train_loss": 0.09900329262018204, "global_step": 153619, "epoch": 1726, "lr": 3.984176025021602e-05}
{"train_loss": 0.1077403798699379, "global_step": 153620, "epoch": 1726, "lr": 3.984119260495844e-05}
{"train_loss": 0.09761568158864975, "global_step": 153621, "epoch": 1726, "lr": 3.9840624961066565e-05}
{"train_loss": 0.11615561693906784, "global_step": 153622, "epoch": 1726, "lr": 3.984005731854052e-05}
{"train_loss": 0.06015770137310028, "global_step": 153623, "epoch": 1726, "lr": 3.983948967738032e-05}
{"train_loss": 0.09113020449876785, "global_step": 153624, "epoch": 1726, "lr": 3.983892203758611e-05}
{"train_loss": 0.07195454835891724, "global_step": 153625, "epoch": 1726, "lr": 3.983835439915791e-05}
{"train_loss": 0.08550956845283508, "global_step": 153626, "epoch": 1726, "lr": 3.983778676209583e-05}
{"train_loss": 0.11240989714860916, "global_step": 153627, "epoch": 1726, "lr": 3.9837219126399925e-05}
{"train_loss": 0.09034458547830582, "global_step": 153628, "epoch": 1726, "lr": 3.9836651492070296e-05}
{"train_loss": 0.07845093309879303, "global_step": 153629, "epoch": 1726, "lr": 3.983608385910701e-05}
{"train_loss": 0.08569027483463287, "global_step": 153630, "epoch": 1726, "lr": 3.9835516227510114e-05}
{"train_loss": 0.09254922717809677, "global_step": 153631, "epoch": 1726, "lr": 3.9834948597279745e-05}
{"train_loss": 0.11000552773475647, "global_step": 153632, "epoch": 1726, "lr": 3.983438096841591e-05}
{"train_loss": 0.09271891415119171, "global_step": 153633, "epoch": 1726, "lr": 3.9833813340918734e-05}
{"train_loss": 0.05996645614504814, "global_step": 153634, "epoch": 1726, "lr": 3.983324571478827e-05}
{"train_loss": 0.15761013329029083, "global_step": 153635, "epoch": 1726, "lr": 3.9832678090024615e-05}
{"train_loss": 0.049772851169109344, "global_step": 153636, "epoch": 1726, "lr": 3.9832110466627816e-05}
{"train_loss": 0.09900826960802078, "global_step": 153637, "epoch": 1726, "lr": 3.9831542844597986e-05}
{"train_loss": 0.05979280173778534, "global_step": 153638, "epoch": 1726, "lr": 3.983097522393516e-05}
{"train_loss": 0.06203193962574005, "global_step": 153639, "epoch": 1726, "lr": 3.983040760463945e-05}
{"train_loss": 0.11813165992498398, "global_step": 153640, "epoch": 1726, "lr": 3.9829839986710906e-05}
{"train_loss": 0.11528432369232178, "global_step": 153641, "epoch": 1726, "lr": 3.9829272370149635e-05}
{"train_loss": 0.10537834465503693, "global_step": 153642, "epoch": 1726, "lr": 3.982870475495567e-05}
{"train_loss": 0.07004152238368988, "global_step": 153643, "epoch": 1726, "lr": 3.9828137141129133e-05}
{"train_loss": 0.10497488081455231, "global_step": 153644, "epoch": 1726, "lr": 3.9827569528670064e-05}
{"train_loss": 0.12378701567649841, "global_step": 153645, "epoch": 1726, "lr": 3.982700191757857e-05}
{"train_loss": 0.08194702863693237, "global_step": 153646, "epoch": 1726, "lr": 3.982643430785469e-05}
{"train_loss": 0.1402798444032669, "global_step": 153647, "epoch": 1726, "lr": 3.982586669949853e-05}
{"train_loss": 0.05807304382324219, "global_step": 153648, "epoch": 1726, "lr": 3.9825299092510164e-05}
{"train_loss": 0.07741287350654602, "global_step": 153649, "epoch": 1726, "lr": 3.982473148688966e-05}
{"train_loss": 0.058710865676403046, "global_step": 153650, "epoch": 1726, "lr": 3.982416388263711e-05}
{"train_loss": 0.1007227823138237, "global_step": 153651, "epoch": 1726, "lr": 3.982359627975255e-05}
{"train_loss": 0.10506770014762878, "global_step": 153652, "epoch": 1726, "lr": 3.9823028678236104e-05}
{"train_loss": 0.1312599927186966, "global_step": 153653, "epoch": 1726, "lr": 3.982246107808781e-05}
{"train_loss": 0.05829082429409027, "global_step": 153654, "epoch": 1726, "lr": 3.982189347930779e-05}
{"train_loss": 0.11456169188022614, "global_step": 153655, "epoch": 1726, "lr": 3.9821325881896064e-05}
{"train_loss": 0.06755475699901581, "global_step": 153656, "epoch": 1726, "lr": 3.982075828585276e-05}
{"train_loss": 0.11379749327898026, "global_step": 153657, "epoch": 1726, "lr": 3.982019069117791e-05}
{"train_loss": 0.09263178706169128, "global_step": 153658, "epoch": 1726, "lr": 3.981962309787163e-05}
{"train_loss": 0.11249624937772751, "global_step": 153659, "epoch": 1726, "lr": 3.981905550593396e-05}
{"train_loss": 0.06192727014422417, "global_step": 153660, "epoch": 1726, "lr": 3.981848791536501e-05}
{"train_loss": 0.05823522061109543, "global_step": 153661, "epoch": 1726, "lr": 3.981792032616483e-05}
{"train_loss": 0.07037530094385147, "global_step": 153662, "epoch": 1726, "lr": 3.981735273833351e-05}
{"train_loss": 0.09585550427436829, "global_step": 153663, "epoch": 1726, "lr": 3.981678515187112e-05}
{"train_loss": 0.052927181124687195, "global_step": 153664, "epoch": 1726, "lr": 3.981621756677774e-05}
{"train_loss": 0.03956311196088791, "global_step": 153665, "epoch": 1726, "lr": 3.981564998305344e-05}
{"train_loss": 0.11007542163133621, "global_step": 153666, "epoch": 1726, "lr": 3.981508240069831e-05}
{"train_loss": 0.09266210347414017, "global_step": 153667, "epoch": 1726, "lr": 3.9814514819712416e-05}
{"train_loss": 0.0745587944984436, "global_step": 153668, "epoch": 1726, "lr": 3.9813947240095836e-05}
{"train_loss": 0.10706374794244766, "global_step": 153669, "epoch": 1726, "lr": 3.981337966184866e-05}
{"train_loss": 0.11276764422655106, "global_step": 153670, "epoch": 1726, "lr": 3.981281208497093e-05}
{"train_loss": 0.0925644263625145, "global_step": 153671, "epoch": 1726, "lr": 3.981224450946276e-05}
{"train_loss": 0.06927260011434555, "global_step": 153672, "epoch": 1726, "lr": 3.98116769353242e-05}
{"train_loss": 0.12831531465053558, "global_step": 153673, "epoch": 1726, "lr": 3.9811109362555337e-05}
{"train_loss": 0.06066058948636055, "global_step": 153674, "epoch": 1726, "lr": 3.981054179115624e-05}
{"train_loss": 0.07681792974472046, "global_step": 153675, "epoch": 1726, "lr": 3.9809974221127006e-05}
{"train_loss": 0.07334371656179428, "global_step": 153676, "epoch": 1726, "lr": 3.980940665246768e-05}
{"train_loss": 0.11682575941085815, "global_step": 153677, "epoch": 1726, "lr": 3.9808839085178376e-05}
{"train_loss": 0.09392092376947403, "global_step": 153678, "epoch": 1726, "lr": 3.980827151925912e-05}
{"train_loss": 0.10833342373371124, "global_step": 153679, "epoch": 1726, "lr": 3.980770395471005e-05}
{"train_loss": 0.10528941452503204, "global_step": 153680, "epoch": 1726, "lr": 3.980713639153118e-05}
{"train_loss": 0.10156912356615067, "global_step": 153681, "epoch": 1726, "lr": 3.980656882972264e-05}
{"train_loss": 0.11638850718736649, "global_step": 153682, "epoch": 1726, "lr": 3.980600126928447e-05}
{"train_loss": 0.11198629438877106, "global_step": 153683, "epoch": 1726, "lr": 3.980543371021676e-05}
{"train_loss": 0.05215805768966675, "global_step": 153684, "epoch": 1726, "lr": 3.980486615251958e-05}
{"train_loss": 0.11062810570001602, "global_step": 153685, "epoch": 1726, "lr": 3.9804298596193026e-05}
{"train_loss": 0.07288841158151627, "global_step": 153686, "epoch": 1726, "lr": 3.9803731041237146e-05}
{"train_loss": 0.11475809663534164, "global_step": 153687, "epoch": 1726, "lr": 3.980316348765204e-05}
{"train_loss": 0.0681934729218483, "global_step": 153688, "epoch": 1726, "lr": 3.980259593543778e-05}
{"train_loss": 0.1363174319267273, "global_step": 153689, "epoch": 1726, "lr": 3.980202838459442e-05}
{"train_loss": 0.11782708764076233, "global_step": 153690, "epoch": 1726, "lr": 3.980146083512207e-05}
{"train_loss": 0.057801973074674606, "global_step": 153691, "epoch": 1726, "lr": 3.980089328702077e-05}
{"train_loss": 0.09121918678283691, "global_step": 153692, "epoch": 1726, "lr": 3.9800325740290635e-05}
{"train_loss": 0.11613587290048599, "global_step": 153693, "epoch": 1726, "lr": 3.97997581949317e-05}
{"train_loss": 0.04978834465146065, "global_step": 153694, "epoch": 1726, "lr": 3.9799190650944084e-05}
{"train_loss": 0.12357909977436066, "global_step": 153695, "epoch": 1726, "lr": 3.979862310832783e-05}
{"train_loss": 0.1538882553577423, "global_step": 153696, "epoch": 1726, "lr": 3.979805556708304e-05}
{"train_loss": 0.05347665771842003, "global_step": 153697, "epoch": 1726, "lr": 3.9797488027209755e-05}
{"train_loss": 0.060237668454647064, "global_step": 153698, "epoch": 1726, "lr": 3.97969204887081e-05}
{"train_loss": 0.1239650622010231, "global_step": 153699, "epoch": 1726, "lr": 3.9796352951578097e-05}
{"train_loss": 0.047155026346445084, "global_step": 153700, "epoch": 1726, "lr": 3.979578541581985e-05}
{"train_loss": 0.090739406645298, "global_step": 153701, "epoch": 1726, "lr": 3.9795217881433464e-05}
{"train_loss": 0.09196608111764608, "global_step": 153702, "epoch": 1726, "lr": 3.979465034841897e-05, "val_loss": 6.422869682312012}
{"train_loss": 0.08091524988412857, "global_step": 153703, "epoch": 1727, "lr": 3.9794082816776466e-05}
{"train_loss": 0.09659148752689362, "global_step": 153704, "epoch": 1727, "lr": 3.979351528650602e-05}
{"train_loss": 0.08192121982574463, "global_step": 153705, "epoch": 1727, "lr": 3.979294775760771e-05}
{"train_loss": 0.08245395869016647, "global_step": 153706, "epoch": 1727, "lr": 3.979238023008161e-05}
{"train_loss": 0.11497590690851212, "global_step": 153707, "epoch": 1727, "lr": 3.979181270392782e-05}
{"train_loss": 0.08164063841104507, "global_step": 153708, "epoch": 1727, "lr": 3.979124517914637e-05}
{"train_loss": 0.10556145012378693, "global_step": 153709, "epoch": 1727, "lr": 3.979067765573738e-05}
{"train_loss": 0.19710572063922882, "global_step": 153710, "epoch": 1727, "lr": 3.9790110133700896e-05}
{"train_loss": 0.10364027321338654, "global_step": 153711, "epoch": 1727, "lr": 3.978954261303703e-05}
{"train_loss": 0.09123874455690384, "global_step": 153712, "epoch": 1727, "lr": 3.9788975093745805e-05}
{"train_loss": 0.10760919749736786, "global_step": 153713, "epoch": 1727, "lr": 3.9788407575827356e-05}
{"train_loss": 0.10203630477190018, "global_step": 153714, "epoch": 1727, "lr": 3.978784005928171e-05}
{"train_loss": 0.08660188317298889, "global_step": 153715, "epoch": 1727, "lr": 3.978727254410898e-05}
{"train_loss": 0.10817522555589676, "global_step": 153716, "epoch": 1727, "lr": 3.9786705030309214e-05}
{"train_loss": 0.08089939504861832, "global_step": 153717, "epoch": 1727, "lr": 3.978613751788251e-05}
{"train_loss": 0.1968049705028534, "global_step": 153718, "epoch": 1727, "lr": 3.978557000682893e-05}
{"train_loss": 0.10278541594743729, "global_step": 153719, "epoch": 1727, "lr": 3.978500249714854e-05}
{"train_loss": 0.06927382200956345, "global_step": 153720, "epoch": 1727, "lr": 3.978443498884146e-05}
{"train_loss": 0.0624033659696579, "global_step": 153721, "epoch": 1727, "lr": 3.978386748190773e-05}
{"train_loss": 0.06997275352478027, "global_step": 153722, "epoch": 1727, "lr": 3.9783299976347434e-05}
{"train_loss": 0.06550317257642746, "global_step": 153723, "epoch": 1727, "lr": 3.978273247216064e-05}
{"train_loss": 0.13788573443889618, "global_step": 153724, "epoch": 1727, "lr": 3.978216496934744e-05}
{"train_loss": 0.08161477744579315, "global_step": 153725, "epoch": 1727, "lr": 3.97815974679079e-05}
{"train_loss": 0.082266665995121, "global_step": 153726, "epoch": 1727, "lr": 3.978102996784212e-05}
{"train_loss": 0.1284712553024292, "global_step": 153727, "epoch": 1727, "lr": 3.978046246915013e-05}
{"train_loss": 0.06637714803218842, "global_step": 153728, "epoch": 1727, "lr": 3.977989497183205e-05}
{"train_loss": 0.11253733932971954, "global_step": 153729, "epoch": 1727, "lr": 3.977932747588792e-05}
{"train_loss": 0.10058100521564484, "global_step": 153730, "epoch": 1727, "lr": 3.977875998131785e-05}
{"train_loss": 0.12352611869573593, "global_step": 153731, "epoch": 1727, "lr": 3.977819248812188e-05}
{"train_loss": 0.07570195198059082, "global_step": 153732, "epoch": 1727, "lr": 3.977762499630014e-05}
{"train_loss": 0.08255085349082947, "global_step": 153733, "epoch": 1727, "lr": 3.977705750585264e-05}
{"train_loss": 0.058430805802345276, "global_step": 153734, "epoch": 1727, "lr": 3.977649001677952e-05}
{"train_loss": 0.09438149631023407, "global_step": 153735, "epoch": 1727, "lr": 3.977592252908081e-05}
{"train_loss": 0.13508285582065582, "global_step": 153736, "epoch": 1727, "lr": 3.977535504275661e-05}
{"train_loss": 0.12281642109155655, "global_step": 153737, "epoch": 1727, "lr": 3.9774787557806976e-05}
{"train_loss": 0.08695032447576523, "global_step": 153738, "epoch": 1727, "lr": 3.977422007423199e-05}
{"train_loss": 0.13748490810394287, "global_step": 153739, "epoch": 1727, "lr": 3.977365259203177e-05}
{"train_loss": 0.15089617669582367, "global_step": 153740, "epoch": 1727, "lr": 3.977308511120634e-05}
{"train_loss": 0.10122261941432953, "global_step": 153741, "epoch": 1727, "lr": 3.977251763175579e-05}
{"train_loss": 0.06834442913532257, "global_step": 153742, "epoch": 1727, "lr": 3.9771950153680204e-05}
{"train_loss": 0.1011621505022049, "global_step": 153743, "epoch": 1727, "lr": 3.977138267697965e-05}
{"train_loss": 0.05267271399497986, "global_step": 153744, "epoch": 1727, "lr": 3.9770815201654214e-05}
{"train_loss": 0.1116521880030632, "global_step": 153745, "epoch": 1727, "lr": 3.977024772770398e-05}
{"train_loss": 0.0872054249048233, "global_step": 153746, "epoch": 1727, "lr": 3.9769680255128985e-05}
{"train_loss": 0.18856240808963776, "global_step": 153747, "epoch": 1727, "lr": 3.976911278392935e-05}
{"train_loss": 0.12190404534339905, "global_step": 153748, "epoch": 1727, "lr": 3.976854531410512e-05}
{"train_loss": 0.0969565361738205, "global_step": 153749, "epoch": 1727, "lr": 3.97679778456564e-05}
{"train_loss": 0.08816863596439362, "global_step": 153750, "epoch": 1727, "lr": 3.9767410378583235e-05}
{"train_loss": 0.108937568962574, "global_step": 153751, "epoch": 1727, "lr": 3.9766842912885736e-05}
{"train_loss": 0.11040078103542328, "global_step": 153752, "epoch": 1727, "lr": 3.976627544856394e-05}
{"train_loss": 0.09519742429256439, "global_step": 153753, "epoch": 1727, "lr": 3.976570798561796e-05}
{"train_loss": 0.09235498309135437, "global_step": 153754, "epoch": 1727, "lr": 3.976514052404784e-05}
{"train_loss": 0.0888160988688469, "global_step": 153755, "epoch": 1727, "lr": 3.976457306385367e-05}
{"train_loss": 0.08965318650007248, "global_step": 153756, "epoch": 1727, "lr": 3.976400560503555e-05}
{"train_loss": 0.07352560013532639, "global_step": 153757, "epoch": 1727, "lr": 3.9763438147593516e-05}
{"train_loss": 0.12942935526371002, "global_step": 153758, "epoch": 1727, "lr": 3.9762870691527676e-05}
{"train_loss": 0.08260421454906464, "global_step": 153759, "epoch": 1727, "lr": 3.976230323683808e-05}
{"train_loss": 0.061184361577034, "global_step": 153760, "epoch": 1727, "lr": 3.9761735783524834e-05}
{"train_loss": 0.08715490251779556, "global_step": 153761, "epoch": 1727, "lr": 3.976116833158798e-05}
{"train_loss": 0.11280333250761032, "global_step": 153762, "epoch": 1727, "lr": 3.9760600881027625e-05}
{"train_loss": 0.134676992893219, "global_step": 153763, "epoch": 1727, "lr": 3.9760033431843826e-05}
{"train_loss": 0.05536903440952301, "global_step": 153764, "epoch": 1727, "lr": 3.975946598403667e-05}
{"train_loss": 0.10536260902881622, "global_step": 153765, "epoch": 1727, "lr": 3.975889853760623e-05}
{"train_loss": 0.08064080774784088, "global_step": 153766, "epoch": 1727, "lr": 3.975833109255259e-05}
{"train_loss": 0.13180211186408997, "global_step": 153767, "epoch": 1727, "lr": 3.9757763648875796e-05}
{"train_loss": 0.14816275238990784, "global_step": 153768, "epoch": 1727, "lr": 3.975719620657596e-05}
{"train_loss": 0.14191073179244995, "global_step": 153769, "epoch": 1727, "lr": 3.9756628765653133e-05}
{"train_loss": 0.08683709800243378, "global_step": 153770, "epoch": 1727, "lr": 3.975606132610742e-05}
{"train_loss": 0.07277172058820724, "global_step": 153771, "epoch": 1727, "lr": 3.975549388793886e-05}
{"train_loss": 0.060103464871644974, "global_step": 153772, "epoch": 1727, "lr": 3.975492645114757e-05}
{"train_loss": 0.11449216306209564, "global_step": 153773, "epoch": 1727, "lr": 3.975435901573358e-05}
{"train_loss": 0.09874320030212402, "global_step": 153774, "epoch": 1727, "lr": 3.9753791581697e-05}
{"train_loss": 0.08272665739059448, "global_step": 153775, "epoch": 1727, "lr": 3.975322414903791e-05}
{"train_loss": 0.18415819108486176, "global_step": 153776, "epoch": 1727, "lr": 3.9752656717756356e-05}
{"train_loss": 0.1501365602016449, "global_step": 153777, "epoch": 1727, "lr": 3.975208928785245e-05}
{"train_loss": 0.08914177119731903, "global_step": 153778, "epoch": 1727, "lr": 3.975152185932623e-05}
{"train_loss": 0.09334441274404526, "global_step": 153779, "epoch": 1727, "lr": 3.975095443217782e-05}
{"train_loss": 0.11983262002468109, "global_step": 153780, "epoch": 1727, "lr": 3.975038700640726e-05}
{"train_loss": 0.05373257026076317, "global_step": 153781, "epoch": 1727, "lr": 3.974981958201463e-05}
{"train_loss": 0.14307767152786255, "global_step": 153782, "epoch": 1727, "lr": 3.9749252159000015e-05}
{"train_loss": 0.11879584193229675, "global_step": 153783, "epoch": 1727, "lr": 3.974868473736348e-05}
{"train_loss": 0.1322236955165863, "global_step": 153784, "epoch": 1727, "lr": 3.974811731710512e-05}
{"train_loss": 0.10862410813570023, "global_step": 153785, "epoch": 1727, "lr": 3.974754989822501e-05}
{"train_loss": 0.042427707463502884, "global_step": 153786, "epoch": 1727, "lr": 3.97469824807232e-05}
{"train_loss": 0.0433744415640831, "global_step": 153787, "epoch": 1727, "lr": 3.974641506459979e-05}
{"train_loss": 0.10746397823095322, "global_step": 153788, "epoch": 1727, "lr": 3.9745847649854834e-05}
{"train_loss": 0.10140077024698257, "global_step": 153789, "epoch": 1727, "lr": 3.974528023648845e-05}
{"train_loss": 0.11984176933765411, "global_step": 153790, "epoch": 1727, "lr": 3.974471282450066e-05}
{"train_loss": 0.10218691478451986, "global_step": 153791, "epoch": 1727, "lr": 3.974414541389158e-05, "val_loss": 6.478821754455566}
{"train_loss": 0.1542174071073532, "global_step": 153792, "epoch": 1728, "lr": 3.974357800466128e-05}
{"train_loss": 0.07548511773347855, "global_step": 153793, "epoch": 1728, "lr": 3.974301059680982e-05}
{"train_loss": 0.06215924024581909, "global_step": 153794, "epoch": 1728, "lr": 3.974244319033731e-05}
{"train_loss": 0.14901037514209747, "global_step": 153795, "epoch": 1728, "lr": 3.9741875785243776e-05}
{"train_loss": 0.06229501590132713, "global_step": 153796, "epoch": 1728, "lr": 3.974130838152934e-05}
{"train_loss": 0.14259754121303558, "global_step": 153797, "epoch": 1728, "lr": 3.974074097919405e-05}
{"train_loss": 0.08724574744701385, "global_step": 153798, "epoch": 1728, "lr": 3.974017357823801e-05}
{"train_loss": 0.0698825940489769, "global_step": 153799, "epoch": 1728, "lr": 3.973960617866126e-05}
{"train_loss": 0.09783152490854263, "global_step": 153800, "epoch": 1728, "lr": 3.97390387804639e-05}
{"train_loss": 0.08215483278036118, "global_step": 153801, "epoch": 1728, "lr": 3.9738471383646e-05}
{"train_loss": 0.10207226872444153, "global_step": 153802, "epoch": 1728, "lr": 3.973790398820764e-05}
{"train_loss": 0.07805129885673523, "global_step": 153803, "epoch": 1728, "lr": 3.973733659414889e-05}
{"train_loss": 0.08024165779352188, "global_step": 153804, "epoch": 1728, "lr": 3.973676920146985e-05}
{"train_loss": 0.07392450422048569, "global_step": 153805, "epoch": 1728, "lr": 3.973620181017055e-05}
{"train_loss": 0.10113202035427094, "global_step": 153806, "epoch": 1728, "lr": 3.973563442025111e-05}
{"train_loss": 0.07183868438005447, "global_step": 153807, "epoch": 1728, "lr": 3.973506703171157e-05}
{"train_loss": 0.07947004586458206, "global_step": 153808, "epoch": 1728, "lr": 3.9734499644552047e-05}
{"train_loss": 0.11336446553468704, "global_step": 153809, "epoch": 1728, "lr": 3.9733932258772574e-05}
{"train_loss": 0.0845746323466301, "global_step": 153810, "epoch": 1728, "lr": 3.973336487437325e-05}
{"train_loss": 0.12206828594207764, "global_step": 153811, "epoch": 1728, "lr": 3.973279749135417e-05}
{"train_loss": 0.08969166874885559, "global_step": 153812, "epoch": 1728, "lr": 3.9732230109715365e-05}
{"train_loss": 0.036045774817466736, "global_step": 153813, "epoch": 1728, "lr": 3.973166272945696e-05}
{"train_loss": 0.07222875207662582, "global_step": 153814, "epoch": 1728, "lr": 3.973109535057899e-05}
{"train_loss": 0.07149427384138107, "global_step": 153815, "epoch": 1728, "lr": 3.973052797308157e-05}
{"train_loss": 0.08503057062625885, "global_step": 153816, "epoch": 1728, "lr": 3.972996059696473e-05}
{"train_loss": 0.09459687769412994, "global_step": 153817, "epoch": 1728, "lr": 3.9729393222228594e-05}
{"train_loss": 0.08118001371622086, "global_step": 153818, "epoch": 1728, "lr": 3.972882584887321e-05}
{"train_loss": 0.0840943455696106, "global_step": 153819, "epoch": 1728, "lr": 3.972825847689865e-05}
{"train_loss": 0.07428135722875595, "global_step": 153820, "epoch": 1728, "lr": 3.972769110630501e-05}
{"train_loss": 0.09223607927560806, "global_step": 153821, "epoch": 1728, "lr": 3.972712373709235e-05}
{"train_loss": 0.09532348066568375, "global_step": 153822, "epoch": 1728, "lr": 3.972655636926076e-05}
{"train_loss": 0.06379511952400208, "global_step": 153823, "epoch": 1728, "lr": 3.9725989002810315e-05}
{"train_loss": 0.054703254252672195, "global_step": 153824, "epoch": 1728, "lr": 3.972542163774107e-05}
{"train_loss": 0.09289935231208801, "global_step": 153825, "epoch": 1728, "lr": 3.972485427405313e-05}
{"train_loss": 0.0796283632516861, "global_step": 153826, "epoch": 1728, "lr": 3.9724286911746546e-05}
{"train_loss": 0.0765700414776802, "global_step": 153827, "epoch": 1728, "lr": 3.97237195508214e-05}
{"train_loss": 0.11293826997280121, "global_step": 153828, "epoch": 1728, "lr": 3.9723152191277795e-05}
{"train_loss": 0.05841220170259476, "global_step": 153829, "epoch": 1728, "lr": 3.972258483311577e-05}
{"train_loss": 0.1481838822364807, "global_step": 153830, "epoch": 1728, "lr": 3.972201747633544e-05}
{"train_loss": 0.0628032237291336, "global_step": 153831, "epoch": 1728, "lr": 3.972145012093683e-05}
{"train_loss": 0.12000476568937302, "global_step": 153832, "epoch": 1728, "lr": 3.9720882766920075e-05}
{"train_loss": 0.09914324432611465, "global_step": 153833, "epoch": 1728, "lr": 3.97203154142852e-05}
{"train_loss": 0.08930972963571548, "global_step": 153834, "epoch": 1728, "lr": 3.971974806303232e-05}
{"train_loss": 0.06961185485124588, "global_step": 153835, "epoch": 1728, "lr": 3.971918071316147e-05}
{"train_loss": 0.10806193202733994, "global_step": 153836, "epoch": 1728, "lr": 3.971861336467279e-05}
{"train_loss": 0.0796351358294487, "global_step": 153837, "epoch": 1728, "lr": 3.971804601756629e-05}
{"train_loss": 0.08209573477506638, "global_step": 153838, "epoch": 1728, "lr": 3.971747867184208e-05}
{"train_loss": 0.07334218919277191, "global_step": 153839, "epoch": 1728, "lr": 3.971691132750022e-05}
{"train_loss": 0.10932568460702896, "global_step": 153840, "epoch": 1728, "lr": 3.971634398454081e-05}
{"train_loss": 0.12727686762809753, "global_step": 153841, "epoch": 1728, "lr": 3.9715776642963904e-05}
{"train_loss": 0.06061974912881851, "global_step": 153842, "epoch": 1728, "lr": 3.9715209302769596e-05}
{"train_loss": 0.08907710760831833, "global_step": 153843, "epoch": 1728, "lr": 3.971464196395794e-05}
{"train_loss": 0.11728821694850922, "global_step": 153844, "epoch": 1728, "lr": 3.971407462652904e-05}
{"train_loss": 0.12346697598695755, "global_step": 153845, "epoch": 1728, "lr": 3.9713507290482933e-05}
{"train_loss": 0.08432909846305847, "global_step": 153846, "epoch": 1728, "lr": 3.971293995581973e-05}
{"train_loss": 0.14212733507156372, "global_step": 153847, "epoch": 1728, "lr": 3.971237262253951e-05}
{"train_loss": 0.10174903273582458, "global_step": 153848, "epoch": 1728, "lr": 3.9711805290642315e-05}
{"train_loss": 0.11026843637228012, "global_step": 153849, "epoch": 1728, "lr": 3.971123796012826e-05}
{"train_loss": 0.11654356122016907, "global_step": 153850, "epoch": 1728, "lr": 3.9710670630997394e-05}
{"train_loss": 0.06435335427522659, "global_step": 153851, "epoch": 1728, "lr": 3.9710103303249816e-05}
{"train_loss": 0.09781541675329208, "global_step": 153852, "epoch": 1728, "lr": 3.970953597688557e-05}
{"train_loss": 0.0880298763513565, "global_step": 153853, "epoch": 1728, "lr": 3.9708968651904774e-05}
{"train_loss": 0.06591752171516418, "global_step": 153854, "epoch": 1728, "lr": 3.970840132830745e-05}
{"train_loss": 0.137885183095932, "global_step": 153855, "epoch": 1728, "lr": 3.9707834006093744e-05}
{"train_loss": 0.12028786540031433, "global_step": 153856, "epoch": 1728, "lr": 3.970726668526367e-05}
{"train_loss": 0.07919958233833313, "global_step": 153857, "epoch": 1728, "lr": 3.9706699365817336e-05}
{"train_loss": 0.04832851514220238, "global_step": 153858, "epoch": 1728, "lr": 3.9706132047754806e-05}
{"train_loss": 0.07399517297744751, "global_step": 153859, "epoch": 1728, "lr": 3.9705564731076166e-05}
{"train_loss": 0.11529204994440079, "global_step": 153860, "epoch": 1728, "lr": 3.970499741578149e-05}
{"train_loss": 0.11111925542354584, "global_step": 153861, "epoch": 1728, "lr": 3.970443010187085e-05}
{"train_loss": 0.10083579272031784, "global_step": 153862, "epoch": 1728, "lr": 3.970386278934431e-05}
{"train_loss": 0.12356049567461014, "global_step": 153863, "epoch": 1728, "lr": 3.970329547820199e-05}
{"train_loss": 0.10375969856977463, "global_step": 153864, "epoch": 1728, "lr": 3.9702728168443904e-05}
{"train_loss": 0.0802186131477356, "global_step": 153865, "epoch": 1728, "lr": 3.9702160860070165e-05}
{"train_loss": 0.11965782940387726, "global_step": 153866, "epoch": 1728, "lr": 3.970159355308086e-05}
{"train_loss": 0.1267411857843399, "global_step": 153867, "epoch": 1728, "lr": 3.970102624747603e-05}
{"train_loss": 0.08676330745220184, "global_step": 153868, "epoch": 1728, "lr": 3.9700458943255804e-05}
{"train_loss": 0.06515119224786758, "global_step": 153869, "epoch": 1728, "lr": 3.9699891640420196e-05}
{"train_loss": 0.10788421332836151, "global_step": 153870, "epoch": 1728, "lr": 3.969932433896934e-05}
{"train_loss": 0.045377422124147415, "global_step": 153871, "epoch": 1728, "lr": 3.969875703890325e-05}
{"train_loss": 0.10300733894109726, "global_step": 153872, "epoch": 1728, "lr": 3.969818974022207e-05}
{"train_loss": 0.10080072283744812, "global_step": 153873, "epoch": 1728, "lr": 3.969762244292581e-05}
{"train_loss": 0.12469515204429626, "global_step": 153874, "epoch": 1728, "lr": 3.969705514701461e-05}
{"train_loss": 0.12631550431251526, "global_step": 153875, "epoch": 1728, "lr": 3.9696487852488496e-05}
{"train_loss": 0.12507931888103485, "global_step": 153876, "epoch": 1728, "lr": 3.969592055934758e-05}
{"train_loss": 0.0836426243185997, "global_step": 153877, "epoch": 1728, "lr": 3.969535326759191e-05}
{"train_loss": 0.08980914205312729, "global_step": 153878, "epoch": 1728, "lr": 3.969478597722157e-05}
{"train_loss": 0.06997184455394745, "global_step": 153879, "epoch": 1728, "lr": 3.969421868823665e-05}
{"train_loss": 0.09397523506973567, "global_step": 153880, "epoch": 1728, "lr": 3.969365140063721e-05, "val_loss": 6.519661903381348}
{"train_loss": 0.084641233086586, "global_step": 153881, "epoch": 1729, "lr": 3.969308411442334e-05}
{"train_loss": 0.08644840866327286, "global_step": 153882, "epoch": 1729, "lr": 3.96925168295951e-05}
{"train_loss": 0.07310391217470169, "global_step": 153883, "epoch": 1729, "lr": 3.9691949546152586e-05}
{"train_loss": 0.10319045186042786, "global_step": 153884, "epoch": 1729, "lr": 3.969138226409584e-05}
{"train_loss": 0.1046459972858429, "global_step": 153885, "epoch": 1729, "lr": 3.969081498342499e-05}
{"train_loss": 0.13909780979156494, "global_step": 153886, "epoch": 1729, "lr": 3.969024770414007e-05}
{"train_loss": 0.11414549499750137, "global_step": 153887, "epoch": 1729, "lr": 3.968968042624118e-05}
{"train_loss": 0.07971366494894028, "global_step": 153888, "epoch": 1729, "lr": 3.9689113149728376e-05}
{"train_loss": 0.18166369199752808, "global_step": 153889, "epoch": 1729, "lr": 3.9688545874601757e-05}
{"train_loss": 0.09064224362373352, "global_step": 153890, "epoch": 1729, "lr": 3.968797860086138e-05}
{"train_loss": 0.09124760329723358, "global_step": 153891, "epoch": 1729, "lr": 3.9687411328507335e-05}
{"train_loss": 0.0697605237364769, "global_step": 153892, "epoch": 1729, "lr": 3.9686844057539674e-05}
{"train_loss": 0.08121941238641739, "global_step": 153893, "epoch": 1729, "lr": 3.968627678795851e-05}
{"train_loss": 0.1427716612815857, "global_step": 153894, "epoch": 1729, "lr": 3.968570951976389e-05}
{"train_loss": 0.08522722870111465, "global_step": 153895, "epoch": 1729, "lr": 3.968514225295591e-05}
{"train_loss": 0.12063419818878174, "global_step": 153896, "epoch": 1729, "lr": 3.968457498753464e-05}
{"train_loss": 0.07321655750274658, "global_step": 153897, "epoch": 1729, "lr": 3.968400772350014e-05}
{"train_loss": 0.11217379570007324, "global_step": 153898, "epoch": 1729, "lr": 3.96834404608525e-05}
{"train_loss": 0.1022714152932167, "global_step": 153899, "epoch": 1729, "lr": 3.96828731995918e-05}
{"train_loss": 0.2045925408601761, "global_step": 153900, "epoch": 1729, "lr": 3.9682305939718116e-05}
{"train_loss": 0.08519074320793152, "global_step": 153901, "epoch": 1729, "lr": 3.968173868123151e-05}
{"train_loss": 0.10203257948160172, "global_step": 153902, "epoch": 1729, "lr": 3.968117142413208e-05}
{"train_loss": 0.09147228300571442, "global_step": 153903, "epoch": 1729, "lr": 3.9680604168419876e-05}
{"train_loss": 0.08694743365049362, "global_step": 153904, "epoch": 1729, "lr": 3.968003691409501e-05}
{"train_loss": 0.09847430884838104, "global_step": 153905, "epoch": 1729, "lr": 3.967946966115751e-05}
{"train_loss": 0.061256542801856995, "global_step": 153906, "epoch": 1729, "lr": 3.967890240960751e-05}
{"train_loss": 0.1258661448955536, "global_step": 153907, "epoch": 1729, "lr": 3.9678335159445026e-05}
{"train_loss": 0.12920251488685608, "global_step": 153908, "epoch": 1729, "lr": 3.967776791067019e-05}
{"train_loss": 0.07379478216171265, "global_step": 153909, "epoch": 1729, "lr": 3.9677200663283034e-05}
{"train_loss": 0.11572050303220749, "global_step": 153910, "epoch": 1729, "lr": 3.967663341728367e-05}
{"train_loss": 0.11702970415353775, "global_step": 153911, "epoch": 1729, "lr": 3.967606617267213e-05}
{"train_loss": 0.06653453409671783, "global_step": 153912, "epoch": 1729, "lr": 3.9675498929448546e-05}
{"train_loss": 0.0727267786860466, "global_step": 153913, "epoch": 1729, "lr": 3.9674931687612944e-05}
{"train_loss": 0.07484354823827744, "global_step": 153914, "epoch": 1729, "lr": 3.967436444716544e-05}
{"train_loss": 0.12615075707435608, "global_step": 153915, "epoch": 1729, "lr": 3.9673797208106086e-05}
{"train_loss": 0.07235930860042572, "global_step": 153916, "epoch": 1729, "lr": 3.967322997043495e-05}
{"train_loss": 0.054736580699682236, "global_step": 153917, "epoch": 1729, "lr": 3.967266273415214e-05}
{"train_loss": 0.12024933844804764, "global_step": 153918, "epoch": 1729, "lr": 3.9672095499257694e-05}
{"train_loss": 0.10758177191019058, "global_step": 153919, "epoch": 1729, "lr": 3.967152826575173e-05}
{"train_loss": 0.10467546433210373, "global_step": 153920, "epoch": 1729, "lr": 3.967096103363429e-05}
{"train_loss": 0.08530984818935394, "global_step": 153921, "epoch": 1729, "lr": 3.967039380290548e-05}
{"train_loss": 0.07890666276216507, "global_step": 153922, "epoch": 1729, "lr": 3.966982657356534e-05}
{"train_loss": 0.11644435673952103, "global_step": 153923, "epoch": 1729, "lr": 3.9669259345613984e-05}
{"train_loss": 0.07467295974493027, "global_step": 153924, "epoch": 1729, "lr": 3.966869211905145e-05}
{"train_loss": 0.08374890685081482, "global_step": 153925, "epoch": 1729, "lr": 3.966812489387785e-05}
{"train_loss": 0.06613674014806747, "global_step": 153926, "epoch": 1729, "lr": 3.966755767009323e-05}
{"train_loss": 0.08988061547279358, "global_step": 153927, "epoch": 1729, "lr": 3.96669904476977e-05}
{"train_loss": 0.10830521583557129, "global_step": 153928, "epoch": 1729, "lr": 3.96664232266913e-05}
{"train_loss": 0.10030146688222885, "global_step": 153929, "epoch": 1729, "lr": 3.9665856007074144e-05}
{"train_loss": 0.08446577936410904, "global_step": 153930, "epoch": 1729, "lr": 3.966528878884626e-05}
{"train_loss": 0.08271871507167816, "global_step": 153931, "epoch": 1729, "lr": 3.966472157200778e-05}
{"train_loss": 0.05293518677353859, "global_step": 153932, "epoch": 1729, "lr": 3.966415435655872e-05}
{"train_loss": 0.06785648316144943, "global_step": 153933, "epoch": 1729, "lr": 3.966358714249923e-05}
{"train_loss": 0.06970755010843277, "global_step": 153934, "epoch": 1729, "lr": 3.966301992982932e-05}
{"train_loss": 0.07796061784029007, "global_step": 153935, "epoch": 1729, "lr": 3.966245271854909e-05}
{"train_loss": 0.05302724614739418, "global_step": 153936, "epoch": 1729, "lr": 3.9661885508658626e-05}
{"train_loss": 0.076485276222229, "global_step": 153937, "epoch": 1729, "lr": 3.966131830015797e-05}
{"train_loss": 0.16015498340129852, "global_step": 153938, "epoch": 1729, "lr": 3.966075109304726e-05}
{"train_loss": 0.05370784550905228, "global_step": 153939, "epoch": 1729, "lr": 3.9660183887326505e-05}
{"train_loss": 0.0803486704826355, "global_step": 153940, "epoch": 1729, "lr": 3.965961668299584e-05}
{"train_loss": 0.12205559015274048, "global_step": 153941, "epoch": 1729, "lr": 3.965904948005529e-05}
{"train_loss": 0.10703261196613312, "global_step": 153942, "epoch": 1729, "lr": 3.965848227850498e-05}
{"train_loss": 0.09846162796020508, "global_step": 153943, "epoch": 1729, "lr": 3.965791507834493e-05}
{"train_loss": 0.11115438491106033, "global_step": 153944, "epoch": 1729, "lr": 3.965734787957528e-05}
{"train_loss": 0.14563366770744324, "global_step": 153945, "epoch": 1729, "lr": 3.965678068219604e-05}
{"train_loss": 0.07242457568645477, "global_step": 153946, "epoch": 1729, "lr": 3.965621348620735e-05}
{"train_loss": 0.10687252134084702, "global_step": 153947, "epoch": 1729, "lr": 3.965564629160923e-05}
{"train_loss": 0.09322244673967361, "global_step": 153948, "epoch": 1729, "lr": 3.965507909840181e-05}
{"train_loss": 0.02389860339462757, "global_step": 153949, "epoch": 1729, "lr": 3.965451190658511e-05}
{"train_loss": 0.1008247584104538, "global_step": 153950, "epoch": 1729, "lr": 3.9653944716159266e-05}
{"train_loss": 0.08251255750656128, "global_step": 153951, "epoch": 1729, "lr": 3.965337752712429e-05}
{"train_loss": 0.1158435121178627, "global_step": 153952, "epoch": 1729, "lr": 3.965281033948032e-05}
{"train_loss": 0.05172119289636612, "global_step": 153953, "epoch": 1729, "lr": 3.965224315322738e-05}
{"train_loss": 0.05308854579925537, "global_step": 153954, "epoch": 1729, "lr": 3.9651675968365575e-05}
{"train_loss": 0.04317944124341011, "global_step": 153955, "epoch": 1729, "lr": 3.965110878489499e-05}
{"train_loss": 0.10939517617225647, "global_step": 153956, "epoch": 1729, "lr": 3.965054160281567e-05}
{"train_loss": 0.07953446358442307, "global_step": 153957, "epoch": 1729, "lr": 3.964997442212773e-05}
{"train_loss": 0.06773430109024048, "global_step": 153958, "epoch": 1729, "lr": 3.96494072428312e-05}
{"train_loss": 0.11147961020469666, "global_step": 153959, "epoch": 1729, "lr": 3.96488400649262e-05}
{"train_loss": 0.09566447138786316, "global_step": 153960, "epoch": 1729, "lr": 3.964827288841277e-05}
{"train_loss": 0.11120016872882843, "global_step": 153961, "epoch": 1729, "lr": 3.964770571329103e-05}
{"train_loss": 0.11590341478586197, "global_step": 153962, "epoch": 1729, "lr": 3.9647138539561e-05}
{"train_loss": 0.08724396675825119, "global_step": 153963, "epoch": 1729, "lr": 3.964657136722281e-05}
{"train_loss": 0.07861950248479843, "global_step": 153964, "epoch": 1729, "lr": 3.9646004196276484e-05}
{"train_loss": 0.10492182523012161, "global_step": 153965, "epoch": 1729, "lr": 3.964543702672216e-05}
{"train_loss": 0.08911401778459549, "global_step": 153966, "epoch": 1729, "lr": 3.964486985855985e-05}
{"train_loss": 0.06301114708185196, "global_step": 153967, "epoch": 1729, "lr": 3.964430269178969e-05}
{"train_loss": 0.14072363078594208, "global_step": 153968, "epoch": 1729, "lr": 3.964373552641171e-05}
{"train_loss": 0.09471954089285953, "global_step": 153969, "epoch": 1729, "lr": 3.964316836242601e-05, "val_loss": 6.70200252532959}
{"train_loss": 0.07457753270864487, "global_step": 153970, "epoch": 1730, "lr": 3.9642601199832645e-05}
{"train_loss": 0.1085079163312912, "global_step": 153971, "epoch": 1730, "lr": 3.964203403863173e-05}
{"train_loss": 0.11616091430187225, "global_step": 153972, "epoch": 1730, "lr": 3.964146687882331e-05}
{"train_loss": 0.07017075270414352, "global_step": 153973, "epoch": 1730, "lr": 3.964089972040745e-05}
{"train_loss": 0.05214667692780495, "global_step": 153974, "epoch": 1730, "lr": 3.964033256338427e-05}
{"train_loss": 0.20821726322174072, "global_step": 153975, "epoch": 1730, "lr": 3.963976540775379e-05}
{"train_loss": 0.06201135367155075, "global_step": 153976, "epoch": 1730, "lr": 3.9639198253516154e-05}
{"train_loss": 0.08095167577266693, "global_step": 153977, "epoch": 1730, "lr": 3.963863110067137e-05}
{"train_loss": 0.11112058162689209, "global_step": 153978, "epoch": 1730, "lr": 3.963806394921957e-05}
{"train_loss": 0.1004616767168045, "global_step": 153979, "epoch": 1730, "lr": 3.963749679916079e-05}
{"train_loss": 0.22003915905952454, "global_step": 153980, "epoch": 1730, "lr": 3.963692965049513e-05}
{"train_loss": 0.16644027829170227, "global_step": 153981, "epoch": 1730, "lr": 3.963636250322265e-05}
{"train_loss": 0.09651732444763184, "global_step": 153982, "epoch": 1730, "lr": 3.963579535734345e-05}
{"train_loss": 0.0758875235915184, "global_step": 153983, "epoch": 1730, "lr": 3.963522821285758e-05}
{"train_loss": 0.16772004961967468, "global_step": 153984, "epoch": 1730, "lr": 3.963466106976513e-05}
{"train_loss": 0.07512842863798141, "global_step": 153985, "epoch": 1730, "lr": 3.963409392806617e-05}
{"train_loss": 0.1078011766076088, "global_step": 153986, "epoch": 1730, "lr": 3.96335267877608e-05}
{"train_loss": 0.08639785647392273, "global_step": 153987, "epoch": 1730, "lr": 3.9632959648849036e-05}
{"train_loss": 0.07222013920545578, "global_step": 153988, "epoch": 1730, "lr": 3.963239251133103e-05}
{"train_loss": 0.10728605091571808, "global_step": 153989, "epoch": 1730, "lr": 3.96318253752068e-05}
{"train_loss": 0.13869766891002655, "global_step": 153990, "epoch": 1730, "lr": 3.963125824047647e-05}
{"train_loss": 0.13195036351680756, "global_step": 153991, "epoch": 1730, "lr": 3.963069110714006e-05}
{"train_loss": 0.047605376690626144, "global_step": 153992, "epoch": 1730, "lr": 3.9630123975197694e-05}
{"train_loss": 0.0998825654387474, "global_step": 153993, "epoch": 1730, "lr": 3.962955684464944e-05}
{"train_loss": 0.12046414613723755, "global_step": 153994, "epoch": 1730, "lr": 3.9628989715495344e-05}
{"train_loss": 0.12697680294513702, "global_step": 153995, "epoch": 1730, "lr": 3.9628422587735523e-05}
{"train_loss": 0.08259192109107971, "global_step": 153996, "epoch": 1730, "lr": 3.9627855461370014e-05}
{"train_loss": 0.16697554290294647, "global_step": 153997, "epoch": 1730, "lr": 3.962728833639894e-05}
{"train_loss": 0.10183604061603546, "global_step": 153998, "epoch": 1730, "lr": 3.962672121282233e-05}
{"train_loss": 0.05788721516728401, "global_step": 153999, "epoch": 1730, "lr": 3.9626154090640295e-05}
{"train_loss": 0.09167157858610153, "global_step": 154000, "epoch": 1730, "lr": 3.962558696985288e-05}
{"train_loss": 0.08778255432844162, "global_step": 154001, "epoch": 1730, "lr": 3.96250198504602e-05}
{"train_loss": 0.11562549322843552, "global_step": 154002, "epoch": 1730, "lr": 3.962445273246228e-05}
{"train_loss": 0.11695734411478043, "global_step": 154003, "epoch": 1730, "lr": 3.9623885615859256e-05}
{"train_loss": 0.12243948876857758, "global_step": 154004, "epoch": 1730, "lr": 3.962331850065115e-05}
{"train_loss": 0.11372207850217819, "global_step": 154005, "epoch": 1730, "lr": 3.9622751386838086e-05}
{"train_loss": 0.06886887550354004, "global_step": 154006, "epoch": 1730, "lr": 3.962218427442009e-05}
{"train_loss": 0.07532531768083572, "global_step": 154007, "epoch": 1730, "lr": 3.96216171633973e-05}
{"train_loss": 0.17297695577144623, "global_step": 154008, "epoch": 1730, "lr": 3.962105005376972e-05}
{"train_loss": 0.06661739945411682, "global_step": 154009, "epoch": 1730, "lr": 3.962048294553748e-05}
{"train_loss": 0.11348937451839447, "global_step": 154010, "epoch": 1730, "lr": 3.961991583870064e-05}
{"train_loss": 0.10172797739505768, "global_step": 154011, "epoch": 1730, "lr": 3.961934873325928e-05}
{"train_loss": 0.13221034407615662, "global_step": 154012, "epoch": 1730, "lr": 3.9618781629213474e-05}
{"train_loss": 0.14461112022399902, "global_step": 154013, "epoch": 1730, "lr": 3.961821452656328e-05}
{"train_loss": 0.138805091381073, "global_step": 154014, "epoch": 1730, "lr": 3.961764742530881e-05}
{"train_loss": 0.056650400161743164, "global_step": 154015, "epoch": 1730, "lr": 3.96170803254501e-05}
{"train_loss": 0.06728015094995499, "global_step": 154016, "epoch": 1730, "lr": 3.961651322698727e-05}
{"train_loss": 0.05793757364153862, "global_step": 154017, "epoch": 1730, "lr": 3.9615946129920347e-05}
{"train_loss": 0.06144866347312927, "global_step": 154018, "epoch": 1730, "lr": 3.9615379034249455e-05}
{"train_loss": 0.12692934274673462, "global_step": 154019, "epoch": 1730, "lr": 3.9614811939974636e-05}
{"train_loss": 0.06167995557188988, "global_step": 154020, "epoch": 1730, "lr": 3.961424484709599e-05}
{"train_loss": 0.095415860414505, "global_step": 154021, "epoch": 1730, "lr": 3.961367775561356e-05}
{"train_loss": 0.15533509850502014, "global_step": 154022, "epoch": 1730, "lr": 3.9613110665527484e-05}
{"train_loss": 0.11474178731441498, "global_step": 154023, "epoch": 1730, "lr": 3.9612543576837757e-05}
{"train_loss": 0.08699986338615417, "global_step": 154024, "epoch": 1730, "lr": 3.961197648954453e-05}
{"train_loss": 0.11776228249073029, "global_step": 154025, "epoch": 1730, "lr": 3.961140940364782e-05}
{"train_loss": 0.06568893790245056, "global_step": 154026, "epoch": 1730, "lr": 3.961084231914775e-05}
{"train_loss": 0.09303853660821915, "global_step": 154027, "epoch": 1730, "lr": 3.961027523604436e-05}
{"train_loss": 0.05392489582300186, "global_step": 154028, "epoch": 1730, "lr": 3.960970815433774e-05}
{"train_loss": 0.09685532003641129, "global_step": 154029, "epoch": 1730, "lr": 3.960914107402798e-05}
{"train_loss": 0.12569521367549896, "global_step": 154030, "epoch": 1730, "lr": 3.960857399511514e-05}
{"train_loss": 0.08243831247091293, "global_step": 154031, "epoch": 1730, "lr": 3.960800691759932e-05}
{"train_loss": 0.1296197772026062, "global_step": 154032, "epoch": 1730, "lr": 3.960743984148054e-05}
{"train_loss": 0.13796208798885345, "global_step": 154033, "epoch": 1730, "lr": 3.960687276675895e-05}
{"train_loss": 0.09977742284536362, "global_step": 154034, "epoch": 1730, "lr": 3.9606305693434555e-05}
{"train_loss": 0.11281716823577881, "global_step": 154035, "epoch": 1730, "lr": 3.9605738621507494e-05}
{"train_loss": 0.06296872347593307, "global_step": 154036, "epoch": 1730, "lr": 3.960517155097779e-05}
{"train_loss": 0.11374147236347198, "global_step": 154037, "epoch": 1730, "lr": 3.960460448184556e-05}
{"train_loss": 0.08564149588346481, "global_step": 154038, "epoch": 1730, "lr": 3.9604037414110854e-05}
{"train_loss": 0.07629551738500595, "global_step": 154039, "epoch": 1730, "lr": 3.960347034777378e-05}
{"train_loss": 0.11322864145040512, "global_step": 154040, "epoch": 1730, "lr": 3.960290328283436e-05}
{"train_loss": 0.11886315792798996, "global_step": 154041, "epoch": 1730, "lr": 3.9602336219292734e-05}
{"train_loss": 0.078539177775383, "global_step": 154042, "epoch": 1730, "lr": 3.9601769157148924e-05}
{"train_loss": 0.09154842793941498, "global_step": 154043, "epoch": 1730, "lr": 3.9601202096403044e-05}
{"train_loss": 0.13015595078468323, "global_step": 154044, "epoch": 1730, "lr": 3.960063503705514e-05}
{"train_loss": 0.07758812606334686, "global_step": 154045, "epoch": 1730, "lr": 3.9600067979105316e-05}
{"train_loss": 0.08377789705991745, "global_step": 154046, "epoch": 1730, "lr": 3.959950092255363e-05}
{"train_loss": 0.06826312094926834, "global_step": 154047, "epoch": 1730, "lr": 3.9598933867400175e-05}
{"train_loss": 0.09297307580709457, "global_step": 154048, "epoch": 1730, "lr": 3.9598366813645006e-05}
{"train_loss": 0.06444007158279419, "global_step": 154049, "epoch": 1730, "lr": 3.9597799761288215e-05}
{"train_loss": 0.06526287645101547, "global_step": 154050, "epoch": 1730, "lr": 3.959723271032988e-05}
{"train_loss": 0.09631787985563278, "global_step": 154051, "epoch": 1730, "lr": 3.9596665660770054e-05}
{"train_loss": 0.10985489189624786, "global_step": 154052, "epoch": 1730, "lr": 3.9596098612608854e-05}
{"train_loss": 0.13936686515808105, "global_step": 154053, "epoch": 1730, "lr": 3.95955315658463e-05}
{"train_loss": 0.08852870762348175, "global_step": 154054, "epoch": 1730, "lr": 3.9594964520482526e-05}
{"train_loss": 0.11715935915708542, "global_step": 154055, "epoch": 1730, "lr": 3.959439747651756e-05}
{"train_loss": 0.11988117545843124, "global_step": 154056, "epoch": 1730, "lr": 3.959383043395152e-05}
{"train_loss": 0.12623050808906555, "global_step": 154057, "epoch": 1730, "lr": 3.959326339278445e-05}
{"train_loss": 0.10270885474393877, "global_step": 154058, "epoch": 1730, "lr": 3.959269635301645e-05, "val_loss": 6.483631610870361, "train_action_mse_error": 20.941913604736328}
{"train_loss": 0.11022717505693436, "global_step": 154059, "epoch": 1731, "lr": 3.959212931464757e-05}
{"train_loss": 0.10568501800298691, "global_step": 154060, "epoch": 1731, "lr": 3.959156227767792e-05}
{"train_loss": 0.09892246127128601, "global_step": 154061, "epoch": 1731, "lr": 3.9590995242107536e-05}
{"train_loss": 0.07770605385303497, "global_step": 154062, "epoch": 1731, "lr": 3.959042820793654e-05}
{"train_loss": 0.04140174388885498, "global_step": 154063, "epoch": 1731, "lr": 3.958986117516496e-05}
{"train_loss": 0.08058024197816849, "global_step": 154064, "epoch": 1731, "lr": 3.958929414379291e-05}
{"train_loss": 0.07098942250013351, "global_step": 154065, "epoch": 1731, "lr": 3.958872711382045e-05}
{"train_loss": 0.10721918195486069, "global_step": 154066, "epoch": 1731, "lr": 3.958816008524766e-05}
{"train_loss": 0.10265786945819855, "global_step": 154067, "epoch": 1731, "lr": 3.958759305807461e-05}
{"train_loss": 0.14227335155010223, "global_step": 154068, "epoch": 1731, "lr": 3.958702603230139e-05}
{"train_loss": 0.09803476184606552, "global_step": 154069, "epoch": 1731, "lr": 3.958645900792807e-05}
{"train_loss": 0.1284780502319336, "global_step": 154070, "epoch": 1731, "lr": 3.9585891984954704e-05}
{"train_loss": 0.07855065912008286, "global_step": 154071, "epoch": 1731, "lr": 3.9585324963381414e-05}
{"train_loss": 0.1455516219139099, "global_step": 154072, "epoch": 1731, "lr": 3.958475794320823e-05}
{"train_loss": 0.08287989348173141, "global_step": 154073, "epoch": 1731, "lr": 3.958419092443526e-05}
{"train_loss": 0.10362819582223892, "global_step": 154074, "epoch": 1731, "lr": 3.958362390706255e-05}
{"train_loss": 0.09961476922035217, "global_step": 154075, "epoch": 1731, "lr": 3.958305689109022e-05}
{"train_loss": 0.08637144416570663, "global_step": 154076, "epoch": 1731, "lr": 3.9582489876518306e-05}
{"train_loss": 0.07051078975200653, "global_step": 154077, "epoch": 1731, "lr": 3.958192286334691e-05}
{"train_loss": 0.10675371438264847, "global_step": 154078, "epoch": 1731, "lr": 3.958135585157608e-05}
{"train_loss": 0.10303623974323273, "global_step": 154079, "epoch": 1731, "lr": 3.958078884120593e-05}
{"train_loss": 0.07060062140226364, "global_step": 154080, "epoch": 1731, "lr": 3.958022183223649e-05}
{"train_loss": 0.12605199217796326, "global_step": 154081, "epoch": 1731, "lr": 3.957965482466787e-05}
{"train_loss": 0.09912536293268204, "global_step": 154082, "epoch": 1731, "lr": 3.957908781850016e-05}
{"train_loss": 0.06754759699106216, "global_step": 154083, "epoch": 1731, "lr": 3.95785208137334e-05}
{"train_loss": 0.09264984726905823, "global_step": 154084, "epoch": 1731, "lr": 3.957795381036768e-05}
{"train_loss": 0.13462920486927032, "global_step": 154085, "epoch": 1731, "lr": 3.9577386808403075e-05}
{"train_loss": 0.049568064510822296, "global_step": 154086, "epoch": 1731, "lr": 3.957681980783967e-05}
{"train_loss": 0.1333024650812149, "global_step": 154087, "epoch": 1731, "lr": 3.957625280867753e-05}
{"train_loss": 0.10777394473552704, "global_step": 154088, "epoch": 1731, "lr": 3.957568581091675e-05}
{"train_loss": 0.12762528657913208, "global_step": 154089, "epoch": 1731, "lr": 3.957511881455737e-05}
{"train_loss": 0.09633125364780426, "global_step": 154090, "epoch": 1731, "lr": 3.9574551819599507e-05}
{"train_loss": 0.12468349188566208, "global_step": 154091, "epoch": 1731, "lr": 3.95739848260432e-05}
{"train_loss": 0.08124087750911713, "global_step": 154092, "epoch": 1731, "lr": 3.957341783388856e-05}
{"train_loss": 0.13933350145816803, "global_step": 154093, "epoch": 1731, "lr": 3.957285084313563e-05}
{"train_loss": 0.10020130127668381, "global_step": 154094, "epoch": 1731, "lr": 3.9572283853784526e-05}
{"train_loss": 0.06147150322794914, "global_step": 154095, "epoch": 1731, "lr": 3.957171686583527e-05}
{"train_loss": 0.10575959831476212, "global_step": 154096, "epoch": 1731, "lr": 3.957114987928801e-05}
{"train_loss": 0.058619175106287, "global_step": 154097, "epoch": 1731, "lr": 3.9570582894142746e-05}
{"train_loss": 0.09074237197637558, "global_step": 154098, "epoch": 1731, "lr": 3.957001591039961e-05}
{"train_loss": 0.0665455237030983, "global_step": 154099, "epoch": 1731, "lr": 3.956944892805864e-05}
{"train_loss": 0.1768038272857666, "global_step": 154100, "epoch": 1731, "lr": 3.9568881947119934e-05}
{"train_loss": 0.09283130615949631, "global_step": 154101, "epoch": 1731, "lr": 3.956831496758359e-05}
{"train_loss": 0.09052305668592453, "global_step": 154102, "epoch": 1731, "lr": 3.956774798944963e-05}
{"train_loss": 0.09690897911787033, "global_step": 154103, "epoch": 1731, "lr": 3.956718101271818e-05}
{"train_loss": 0.16082854568958282, "global_step": 154104, "epoch": 1731, "lr": 3.956661403738928e-05}
{"train_loss": 0.1333068460226059, "global_step": 154105, "epoch": 1731, "lr": 3.956604706346303e-05}
{"train_loss": 0.1175379678606987, "global_step": 154106, "epoch": 1731, "lr": 3.9565480090939495e-05}
{"train_loss": 0.07643953710794449, "global_step": 154107, "epoch": 1731, "lr": 3.956491311981875e-05}
{"train_loss": 0.0889241173863411, "global_step": 154108, "epoch": 1731, "lr": 3.956434615010088e-05}
{"train_loss": 0.04656369611620903, "global_step": 154109, "epoch": 1731, "lr": 3.956377918178597e-05}
{"train_loss": 0.05300536006689072, "global_step": 154110, "epoch": 1731, "lr": 3.956321221487406e-05}
{"train_loss": 0.08180109411478043, "global_step": 154111, "epoch": 1731, "lr": 3.956264524936527e-05}
{"train_loss": 0.10745587199926376, "global_step": 154112, "epoch": 1731, "lr": 3.956207828525963e-05}
{"train_loss": 0.08550488203763962, "global_step": 154113, "epoch": 1731, "lr": 3.956151132255726e-05}
{"train_loss": 0.17320072650909424, "global_step": 154114, "epoch": 1731, "lr": 3.9560944361258207e-05}
{"train_loss": 0.11358918994665146, "global_step": 154115, "epoch": 1731, "lr": 3.9560377401362576e-05}
{"train_loss": 0.07128524035215378, "global_step": 154116, "epoch": 1731, "lr": 3.95598104428704e-05}
{"train_loss": 0.10219220817089081, "global_step": 154117, "epoch": 1731, "lr": 3.95592434857818e-05}
{"train_loss": 0.15004253387451172, "global_step": 154118, "epoch": 1731, "lr": 3.955867653009682e-05}
{"train_loss": 0.06274038553237915, "global_step": 154119, "epoch": 1731, "lr": 3.955810957581554e-05}
{"train_loss": 0.08028057217597961, "global_step": 154120, "epoch": 1731, "lr": 3.9557542622938065e-05}
{"train_loss": 0.08722197264432907, "global_step": 154121, "epoch": 1731, "lr": 3.9556975671464436e-05}
{"train_loss": 0.0789676159620285, "global_step": 154122, "epoch": 1731, "lr": 3.955640872139476e-05}
{"train_loss": 0.04772811383008957, "global_step": 154123, "epoch": 1731, "lr": 3.955584177272909e-05}
{"train_loss": 0.05603472888469696, "global_step": 154124, "epoch": 1731, "lr": 3.9555274825467515e-05}
{"train_loss": 0.09020508080720901, "global_step": 154125, "epoch": 1731, "lr": 3.95547078796101e-05}
{"train_loss": 0.13789713382720947, "global_step": 154126, "epoch": 1731, "lr": 3.9554140935156924e-05}
{"train_loss": 0.19393746554851532, "global_step": 154127, "epoch": 1731, "lr": 3.9553573992108075e-05}
{"train_loss": 0.11195023357868195, "global_step": 154128, "epoch": 1731, "lr": 3.9553007050463626e-05}
{"train_loss": 0.07240986824035645, "global_step": 154129, "epoch": 1731, "lr": 3.9552440110223625e-05}
{"train_loss": 0.19229862093925476, "global_step": 154130, "epoch": 1731, "lr": 3.9551873171388195e-05}
{"train_loss": 0.19569946825504303, "global_step": 154131, "epoch": 1731, "lr": 3.955130623395736e-05}
{"train_loss": 0.07313697040081024, "global_step": 154132, "epoch": 1731, "lr": 3.955073929793126e-05}
{"train_loss": 0.06933586299419403, "global_step": 154133, "epoch": 1731, "lr": 3.95501723633099e-05}
{"train_loss": 0.10985542088747025, "global_step": 154134, "epoch": 1731, "lr": 3.9549605430093426e-05}
{"train_loss": 0.22547587752342224, "global_step": 154135, "epoch": 1731, "lr": 3.954903849828186e-05}
{"train_loss": 0.07100266218185425, "global_step": 154136, "epoch": 1731, "lr": 3.9548471567875286e-05}
{"train_loss": 0.11753527075052261, "global_step": 154137, "epoch": 1731, "lr": 3.954790463887382e-05}
{"train_loss": 0.07519350200891495, "global_step": 154138, "epoch": 1731, "lr": 3.954733771127749e-05}
{"train_loss": 0.04060462862253189, "global_step": 154139, "epoch": 1731, "lr": 3.954677078508641e-05}
{"train_loss": 0.08159387856721878, "global_step": 154140, "epoch": 1731, "lr": 3.9546203860300627e-05}
{"train_loss": 0.11098933219909668, "global_step": 154141, "epoch": 1731, "lr": 3.954563693692024e-05}
{"train_loss": 0.08580992370843887, "global_step": 154142, "epoch": 1731, "lr": 3.954507001494531e-05}
{"train_loss": 0.05015791207551956, "global_step": 154143, "epoch": 1731, "lr": 3.954450309437592e-05}
{"train_loss": 0.13820871710777283, "global_step": 154144, "epoch": 1731, "lr": 3.954393617521214e-05}
{"train_loss": 0.0591190904378891, "global_step": 154145, "epoch": 1731, "lr": 3.954336925745405e-05}
{"train_loss": 0.08853316307067871, "global_step": 154146, "epoch": 1731, "lr": 3.954280234110173e-05}
{"train_loss": 0.09968460062414072, "global_step": 154147, "epoch": 1731, "lr": 3.9542235426155264e-05, "val_loss": 6.393703460693359}
{"train_loss": 0.15826299786567688, "global_step": 154148, "epoch": 1732, "lr": 3.954166851261469e-05}
{"train_loss": 0.10955092310905457, "global_step": 154149, "epoch": 1732, "lr": 3.954110160048014e-05}
{"train_loss": 0.11582356691360474, "global_step": 154150, "epoch": 1732, "lr": 3.954053468975163e-05}
{"train_loss": 0.09720509499311447, "global_step": 154151, "epoch": 1732, "lr": 3.953996778042929e-05}
{"train_loss": 0.06270056217908859, "global_step": 154152, "epoch": 1732, "lr": 3.953940087251316e-05}
{"train_loss": 0.11582432687282562, "global_step": 154153, "epoch": 1732, "lr": 3.953883396600335e-05}
{"train_loss": 0.10848511010408401, "global_step": 154154, "epoch": 1732, "lr": 3.953826706089989e-05}
{"train_loss": 0.05069899931550026, "global_step": 154155, "epoch": 1732, "lr": 3.953770015720289e-05}
{"train_loss": 0.08560546487569809, "global_step": 154156, "epoch": 1732, "lr": 3.953713325491243e-05}
{"train_loss": 0.09285615384578705, "global_step": 154157, "epoch": 1732, "lr": 3.9536566354028564e-05}
{"train_loss": 0.1016717180609703, "global_step": 154158, "epoch": 1732, "lr": 3.95359994545514e-05}
{"train_loss": 0.098463274538517, "global_step": 154159, "epoch": 1732, "lr": 3.953543255648097e-05}
{"train_loss": 0.13055220246315002, "global_step": 154160, "epoch": 1732, "lr": 3.9534865659817386e-05}
{"train_loss": 0.06830637156963348, "global_step": 154161, "epoch": 1732, "lr": 3.953429876456071e-05}
{"train_loss": 0.1359955072402954, "global_step": 154162, "epoch": 1732, "lr": 3.9533731870711024e-05}
{"train_loss": 0.03125205636024475, "global_step": 154163, "epoch": 1732, "lr": 3.953316497826839e-05}
{"train_loss": 0.09100096672773361, "global_step": 154164, "epoch": 1732, "lr": 3.95325980872329e-05}
{"train_loss": 0.1033889427781105, "global_step": 154165, "epoch": 1732, "lr": 3.953203119760462e-05}
{"train_loss": 0.1326126903295517, "global_step": 154166, "epoch": 1732, "lr": 3.953146430938365e-05}
{"train_loss": 0.038576964288949966, "global_step": 154167, "epoch": 1732, "lr": 3.9530897422570016e-05}
{"train_loss": 0.1189863383769989, "global_step": 154168, "epoch": 1732, "lr": 3.953033053716385e-05}
{"train_loss": 0.1373639702796936, "global_step": 154169, "epoch": 1732, "lr": 3.9529763653165186e-05}
{"train_loss": 0.1029239073395729, "global_step": 154170, "epoch": 1732, "lr": 3.952919677057413e-05}
{"train_loss": 0.10674823820590973, "global_step": 154171, "epoch": 1732, "lr": 3.952862988939073e-05}
{"train_loss": 0.11350890249013901, "global_step": 154172, "epoch": 1732, "lr": 3.952806300961509e-05}
{"train_loss": 0.05044974386692047, "global_step": 154173, "epoch": 1732, "lr": 3.9527496131247274e-05}
{"train_loss": 0.08677928149700165, "global_step": 154174, "epoch": 1732, "lr": 3.952692925428735e-05}
{"train_loss": 0.12062719464302063, "global_step": 154175, "epoch": 1732, "lr": 3.952636237873542e-05}
{"train_loss": 0.15671902894973755, "global_step": 154176, "epoch": 1732, "lr": 3.952579550459152e-05}
{"train_loss": 0.11918158829212189, "global_step": 154177, "epoch": 1732, "lr": 3.952522863185578e-05}
{"train_loss": 0.17557115852832794, "global_step": 154178, "epoch": 1732, "lr": 3.9524661760528215e-05}
{"train_loss": 0.1266908198595047, "global_step": 154179, "epoch": 1732, "lr": 3.952409489060896e-05}
{"train_loss": 0.1473771035671234, "global_step": 154180, "epoch": 1732, "lr": 3.9523528022098045e-05}
{"train_loss": 0.07573007047176361, "global_step": 154181, "epoch": 1732, "lr": 3.9522961154995565e-05}
{"train_loss": 0.10179010033607483, "global_step": 154182, "epoch": 1732, "lr": 3.95223942893016e-05}
{"train_loss": 0.11945411562919617, "global_step": 154183, "epoch": 1732, "lr": 3.9521827425016215e-05}
{"train_loss": 0.09550356864929199, "global_step": 154184, "epoch": 1732, "lr": 3.9521260562139495e-05}
{"train_loss": 0.11558077484369278, "global_step": 154185, "epoch": 1732, "lr": 3.952069370067153e-05}
{"train_loss": 0.07482220977544785, "global_step": 154186, "epoch": 1732, "lr": 3.952012684061235e-05}
{"train_loss": 0.13117077946662903, "global_step": 154187, "epoch": 1732, "lr": 3.951955998196209e-05}
{"train_loss": 0.07603888213634491, "global_step": 154188, "epoch": 1732, "lr": 3.951899312472077e-05}
{"train_loss": 0.2017470896244049, "global_step": 154189, "epoch": 1732, "lr": 3.951842626888852e-05}
{"train_loss": 0.07302293181419373, "global_step": 154190, "epoch": 1732, "lr": 3.951785941446536e-05}
{"train_loss": 0.08135426789522171, "global_step": 154191, "epoch": 1732, "lr": 3.9517292561451405e-05}
{"train_loss": 0.05885213613510132, "global_step": 154192, "epoch": 1732, "lr": 3.951672570984674e-05}
{"train_loss": 0.08692415058612823, "global_step": 154193, "epoch": 1732, "lr": 3.9516158859651405e-05}
{"train_loss": 0.0783751904964447, "global_step": 154194, "epoch": 1732, "lr": 3.951559201086551e-05}
{"train_loss": 0.12211926281452179, "global_step": 154195, "epoch": 1732, "lr": 3.95150251634891e-05}
{"train_loss": 0.08456255495548248, "global_step": 154196, "epoch": 1732, "lr": 3.951445831752228e-05}
{"train_loss": 0.0868695005774498, "global_step": 154197, "epoch": 1732, "lr": 3.95138914729651e-05}
{"train_loss": 0.08885319530963898, "global_step": 154198, "epoch": 1732, "lr": 3.9513324629817676e-05}
{"train_loss": 0.106478251516819, "global_step": 154199, "epoch": 1732, "lr": 3.951275778808004e-05}
{"train_loss": 0.09817394614219666, "global_step": 154200, "epoch": 1732, "lr": 3.951219094775228e-05}
{"train_loss": 0.09020711481571198, "global_step": 154201, "epoch": 1732, "lr": 3.9511624108834486e-05}
{"train_loss": 0.06927457451820374, "global_step": 154202, "epoch": 1732, "lr": 3.951105727132673e-05}
{"train_loss": 0.10377039760351181, "global_step": 154203, "epoch": 1732, "lr": 3.951049043522908e-05}
{"train_loss": 0.12040723115205765, "global_step": 154204, "epoch": 1732, "lr": 3.9509923600541625e-05}
{"train_loss": 0.11377444863319397, "global_step": 154205, "epoch": 1732, "lr": 3.950935676726442e-05}
{"train_loss": 0.0895174890756607, "global_step": 154206, "epoch": 1732, "lr": 3.950878993539756e-05}
{"train_loss": 0.07032806426286697, "global_step": 154207, "epoch": 1732, "lr": 3.9508223104941106e-05}
{"train_loss": 0.08395007252693176, "global_step": 154208, "epoch": 1732, "lr": 3.950765627589514e-05}
{"train_loss": 0.08814240247011185, "global_step": 154209, "epoch": 1732, "lr": 3.950708944825976e-05}
{"train_loss": 0.09753992408514023, "global_step": 154210, "epoch": 1732, "lr": 3.950652262203501e-05}
{"train_loss": 0.1185772642493248, "global_step": 154211, "epoch": 1732, "lr": 3.9505955797220994e-05}
{"train_loss": 0.13996170461177826, "global_step": 154212, "epoch": 1732, "lr": 3.950538897381776e-05}
{"train_loss": 0.07077381014823914, "global_step": 154213, "epoch": 1732, "lr": 3.950482215182541e-05}
{"train_loss": 0.07834409177303314, "global_step": 154214, "epoch": 1732, "lr": 3.9504255331244e-05}
{"train_loss": 0.10299571603536606, "global_step": 154215, "epoch": 1732, "lr": 3.9503688512073626e-05}
{"train_loss": 0.0733020082116127, "global_step": 154216, "epoch": 1732, "lr": 3.950312169431434e-05}
{"train_loss": 0.11949995160102844, "global_step": 154217, "epoch": 1732, "lr": 3.950255487796625e-05}
{"train_loss": 0.13593949377536774, "global_step": 154218, "epoch": 1732, "lr": 3.9501988063029385e-05}
{"train_loss": 0.04888324439525604, "global_step": 154219, "epoch": 1732, "lr": 3.950142124950388e-05}
{"train_loss": 0.12766632437705994, "global_step": 154220, "epoch": 1732, "lr": 3.950085443738977e-05}
{"train_loss": 0.09993669390678406, "global_step": 154221, "epoch": 1732, "lr": 3.9500287626687134e-05}
{"train_loss": 0.13884080946445465, "global_step": 154222, "epoch": 1732, "lr": 3.9499720817396056e-05}
{"train_loss": 0.09241744130849838, "global_step": 154223, "epoch": 1732, "lr": 3.949915400951663e-05}
{"train_loss": 0.10095822811126709, "global_step": 154224, "epoch": 1732, "lr": 3.94985872030489e-05}
{"train_loss": 0.0902564525604248, "global_step": 154225, "epoch": 1732, "lr": 3.949802039799296e-05}
{"train_loss": 0.06206487864255905, "global_step": 154226, "epoch": 1732, "lr": 3.9497453594348874e-05}
{"train_loss": 0.11670241504907608, "global_step": 154227, "epoch": 1732, "lr": 3.9496886792116724e-05}
{"train_loss": 0.08850402384996414, "global_step": 154228, "epoch": 1732, "lr": 3.949631999129661e-05}
{"train_loss": 0.13790273666381836, "global_step": 154229, "epoch": 1732, "lr": 3.949575319188857e-05}
{"train_loss": 0.11557606607675552, "global_step": 154230, "epoch": 1732, "lr": 3.9495186393892725e-05}
{"train_loss": 0.1249968633055687, "global_step": 154231, "epoch": 1732, "lr": 3.949461959730909e-05}
{"train_loss": 0.12480294704437256, "global_step": 154232, "epoch": 1732, "lr": 3.949405280213781e-05}
{"train_loss": 0.09597472846508026, "global_step": 154233, "epoch": 1732, "lr": 3.9493486008378896e-05}
{"train_loss": 0.0631101205945015, "global_step": 154234, "epoch": 1732, "lr": 3.9492919216032475e-05}
{"train_loss": 0.07200827449560165, "global_step": 154235, "epoch": 1732, "lr": 3.9492352425098586e-05}
{"train_loss": 0.10088792183760846, "global_step": 154236, "epoch": 1732, "lr": 3.9491785635577344e-05, "val_loss": 6.384932518005371}
{"train_loss": 0.10120315104722977, "global_step": 154237, "epoch": 1733, "lr": 3.949121884746878e-05}
{"train_loss": 0.09275595098733902, "global_step": 154238, "epoch": 1733, "lr": 3.9490652060773015e-05}
{"train_loss": 0.09192735701799393, "global_step": 154239, "epoch": 1733, "lr": 3.9490085275490096e-05}
{"train_loss": 0.12652109563350677, "global_step": 154240, "epoch": 1733, "lr": 3.94895184916201e-05}
{"train_loss": 0.14212587475776672, "global_step": 154241, "epoch": 1733, "lr": 3.9488951709163116e-05}
{"train_loss": 0.09389945864677429, "global_step": 154242, "epoch": 1733, "lr": 3.948838492811922e-05}
{"train_loss": 0.11542942374944687, "global_step": 154243, "epoch": 1733, "lr": 3.948781814848847e-05}
{"train_loss": 0.07587838917970657, "global_step": 154244, "epoch": 1733, "lr": 3.948725137027097e-05}
{"train_loss": 0.08846891671419144, "global_step": 154245, "epoch": 1733, "lr": 3.948668459346676e-05}
{"train_loss": 0.0839846208691597, "global_step": 154246, "epoch": 1733, "lr": 3.9486117818075934e-05}
{"train_loss": 0.14125078916549683, "global_step": 154247, "epoch": 1733, "lr": 3.9485551044098595e-05}
{"train_loss": 0.03809452801942825, "global_step": 154248, "epoch": 1733, "lr": 3.948498427153478e-05}
{"train_loss": 0.06733553856611252, "global_step": 154249, "epoch": 1733, "lr": 3.9484417500384596e-05}
{"train_loss": 0.1527530997991562, "global_step": 154250, "epoch": 1733, "lr": 3.948385073064808e-05}
{"train_loss": 0.06120630353689194, "global_step": 154251, "epoch": 1733, "lr": 3.9483283962325355e-05}
{"train_loss": 0.08692511916160583, "global_step": 154252, "epoch": 1733, "lr": 3.948271719541646e-05}
{"train_loss": 0.11659204959869385, "global_step": 154253, "epoch": 1733, "lr": 3.94821504299215e-05}
{"train_loss": 0.1430320143699646, "global_step": 154254, "epoch": 1733, "lr": 3.9481583665840516e-05}
{"train_loss": 0.11800165474414825, "global_step": 154255, "epoch": 1733, "lr": 3.948101690317363e-05}
{"train_loss": 0.1106424406170845, "global_step": 154256, "epoch": 1733, "lr": 3.9480450141920864e-05}
{"train_loss": 0.07819211483001709, "global_step": 154257, "epoch": 1733, "lr": 3.947988338208235e-05}
{"train_loss": 0.09205377101898193, "global_step": 154258, "epoch": 1733, "lr": 3.9479316623658125e-05}
{"train_loss": 0.13681365549564362, "global_step": 154259, "epoch": 1733, "lr": 3.947874986664828e-05}
{"train_loss": 0.060931894928216934, "global_step": 154260, "epoch": 1733, "lr": 3.947818311105289e-05}
{"train_loss": 0.08749418705701828, "global_step": 154261, "epoch": 1733, "lr": 3.947761635687203e-05}
{"train_loss": 0.06533670425415039, "global_step": 154262, "epoch": 1733, "lr": 3.947704960410577e-05}
{"train_loss": 0.10521361231803894, "global_step": 154263, "epoch": 1733, "lr": 3.9476482852754185e-05}
{"train_loss": 0.17713141441345215, "global_step": 154264, "epoch": 1733, "lr": 3.947591610281738e-05}
{"train_loss": 0.08210930228233337, "global_step": 154265, "epoch": 1733, "lr": 3.9475349354295395e-05}
{"train_loss": 0.06718898564577103, "global_step": 154266, "epoch": 1733, "lr": 3.947478260718833e-05}
{"train_loss": 0.09581246227025986, "global_step": 154267, "epoch": 1733, "lr": 3.947421586149624e-05}
{"train_loss": 0.11103890091180801, "global_step": 154268, "epoch": 1733, "lr": 3.947364911721924e-05}
{"train_loss": 0.116139255464077, "global_step": 154269, "epoch": 1733, "lr": 3.947308237435735e-05}
{"train_loss": 0.14242328703403473, "global_step": 154270, "epoch": 1733, "lr": 3.947251563291069e-05}
{"train_loss": 0.08210501074790955, "global_step": 154271, "epoch": 1733, "lr": 3.9471948892879305e-05}
{"train_loss": 0.09795936942100525, "global_step": 154272, "epoch": 1733, "lr": 3.947138215426332e-05}
{"train_loss": 0.0573844239115715, "global_step": 154273, "epoch": 1733, "lr": 3.947081541706274e-05}
{"train_loss": 0.15820151567459106, "global_step": 154274, "epoch": 1733, "lr": 3.947024868127772e-05}
{"train_loss": 0.11326257139444351, "global_step": 154275, "epoch": 1733, "lr": 3.946968194690826e-05}
{"train_loss": 0.13774043321609497, "global_step": 154276, "epoch": 1733, "lr": 3.946911521395451e-05}
{"train_loss": 0.09514467418193817, "global_step": 154277, "epoch": 1733, "lr": 3.946854848241649e-05}
{"train_loss": 0.09309583157300949, "global_step": 154278, "epoch": 1733, "lr": 3.9467981752294294e-05}
{"train_loss": 0.05379553884267807, "global_step": 154279, "epoch": 1733, "lr": 3.9467415023588006e-05}
{"train_loss": 0.10287657380104065, "global_step": 154280, "epoch": 1733, "lr": 3.9466848296297686e-05}
{"train_loss": 0.06317070871591568, "global_step": 154281, "epoch": 1733, "lr": 3.946628157042343e-05}
{"train_loss": 0.0856509879231453, "global_step": 154282, "epoch": 1733, "lr": 3.946571484596528e-05}
{"train_loss": 0.15755699574947357, "global_step": 154283, "epoch": 1733, "lr": 3.9465148122923373e-05}
{"train_loss": 0.10784509032964706, "global_step": 154284, "epoch": 1733, "lr": 3.946458140129772e-05}
{"train_loss": 0.14167869091033936, "global_step": 154285, "epoch": 1733, "lr": 3.9464014681088444e-05}
{"train_loss": 0.08471135050058365, "global_step": 154286, "epoch": 1733, "lr": 3.9463447962295586e-05}
{"train_loss": 0.119209885597229, "global_step": 154287, "epoch": 1733, "lr": 3.946288124491926e-05}
{"train_loss": 0.0641024261713028, "global_step": 154288, "epoch": 1733, "lr": 3.94623145289595e-05}
{"train_loss": 0.06242576986551285, "global_step": 154289, "epoch": 1733, "lr": 3.946174781441642e-05}
{"train_loss": 0.10521947592496872, "global_step": 154290, "epoch": 1733, "lr": 3.9461181101290066e-05}
{"train_loss": 0.09142518788576126, "global_step": 154291, "epoch": 1733, "lr": 3.9460614389580544e-05}
{"train_loss": 0.06781917810440063, "global_step": 154292, "epoch": 1733, "lr": 3.946004767928789e-05}
{"train_loss": 0.05762506276369095, "global_step": 154293, "epoch": 1733, "lr": 3.945948097041223e-05}
{"train_loss": 0.12251840531826019, "global_step": 154294, "epoch": 1733, "lr": 3.9458914262953593e-05}
{"train_loss": 0.09025125950574875, "global_step": 154295, "epoch": 1733, "lr": 3.94583475569121e-05}
{"train_loss": 0.07587912678718567, "global_step": 154296, "epoch": 1733, "lr": 3.945778085228778e-05}
{"train_loss": 0.11152905970811844, "global_step": 154297, "epoch": 1733, "lr": 3.9457214149080744e-05}
{"train_loss": 0.2210836559534073, "global_step": 154298, "epoch": 1733, "lr": 3.945664744729106e-05}
{"train_loss": 0.06859920173883438, "global_step": 154299, "epoch": 1733, "lr": 3.945608074691879e-05}
{"train_loss": 0.06444177031517029, "global_step": 154300, "epoch": 1733, "lr": 3.945551404796404e-05}
{"train_loss": 0.1588471382856369, "global_step": 154301, "epoch": 1733, "lr": 3.9454947350426843e-05}
{"train_loss": 0.10328882187604904, "global_step": 154302, "epoch": 1733, "lr": 3.945438065430732e-05}
{"train_loss": 0.06445979326963425, "global_step": 154303, "epoch": 1733, "lr": 3.945381395960551e-05}
{"train_loss": 0.14672976732254028, "global_step": 154304, "epoch": 1733, "lr": 3.945324726632152e-05}
{"train_loss": 0.10323776304721832, "global_step": 154305, "epoch": 1733, "lr": 3.94526805744554e-05}
{"train_loss": 0.14422951638698578, "global_step": 154306, "epoch": 1733, "lr": 3.945211388400726e-05}
{"train_loss": 0.08915300667285919, "global_step": 154307, "epoch": 1733, "lr": 3.9451547194977134e-05}
{"train_loss": 0.11277462542057037, "global_step": 154308, "epoch": 1733, "lr": 3.945098050736513e-05}
{"train_loss": 0.10767132043838501, "global_step": 154309, "epoch": 1733, "lr": 3.94504138211713e-05}
{"train_loss": 0.06590160727500916, "global_step": 154310, "epoch": 1733, "lr": 3.9449847136395755e-05}
{"train_loss": 0.11273703724145889, "global_step": 154311, "epoch": 1733, "lr": 3.944928045303853e-05}
{"train_loss": 0.1218840703368187, "global_step": 154312, "epoch": 1733, "lr": 3.944871377109974e-05}
{"train_loss": 0.131867453455925, "global_step": 154313, "epoch": 1733, "lr": 3.9448147090579414e-05}
{"train_loss": 0.13351558148860931, "global_step": 154314, "epoch": 1733, "lr": 3.944758041147768e-05}
{"train_loss": 0.17313827574253082, "global_step": 154315, "epoch": 1733, "lr": 3.944701373379459e-05}
{"train_loss": 0.04971638321876526, "global_step": 154316, "epoch": 1733, "lr": 3.9446447057530204e-05}
{"train_loss": 0.10981475561857224, "global_step": 154317, "epoch": 1733, "lr": 3.944588038268463e-05}
{"train_loss": 0.11294445395469666, "global_step": 154318, "epoch": 1733, "lr": 3.944531370925791e-05}
{"train_loss": 0.10676778107881546, "global_step": 154319, "epoch": 1733, "lr": 3.944474703725016e-05}
{"train_loss": 0.1421683430671692, "global_step": 154320, "epoch": 1733, "lr": 3.944418036666141e-05}
{"train_loss": 0.078197181224823, "global_step": 154321, "epoch": 1733, "lr": 3.9443613697491786e-05}
{"train_loss": 0.11253342777490616, "global_step": 154322, "epoch": 1733, "lr": 3.944304702974132e-05}
{"train_loss": 0.1099494993686676, "global_step": 154323, "epoch": 1733, "lr": 3.9442480363410134e-05}
{"train_loss": 0.08967012912034988, "global_step": 154324, "epoch": 1733, "lr": 3.9441913698498254e-05}
{"train_loss": 0.10413771679394701, "global_step": 154325, "epoch": 1733, "lr": 3.944134703500579e-05, "val_loss": 6.489520072937012}
{"train_loss": 0.08491288870573044, "global_step": 154326, "epoch": 1734, "lr": 3.94407803729328e-05}
{"train_loss": 0.11462468653917313, "global_step": 154327, "epoch": 1734, "lr": 3.944021371227938e-05}
{"train_loss": 0.09940887242555618, "global_step": 154328, "epoch": 1734, "lr": 3.9439647053045584e-05}
{"train_loss": 0.05970965325832367, "global_step": 154329, "epoch": 1734, "lr": 3.9439080395231514e-05}
{"train_loss": 0.07502514123916626, "global_step": 154330, "epoch": 1734, "lr": 3.943851373883721e-05}
{"train_loss": 0.06928742676973343, "global_step": 154331, "epoch": 1734, "lr": 3.943794708386279e-05}
{"train_loss": 0.09008364379405975, "global_step": 154332, "epoch": 1734, "lr": 3.943738043030829e-05}
{"train_loss": 0.10201800614595413, "global_step": 154333, "epoch": 1734, "lr": 3.943681377817382e-05}
{"train_loss": 0.1118241399526596, "global_step": 154334, "epoch": 1734, "lr": 3.943624712745943e-05}
{"train_loss": 0.1573161482810974, "global_step": 154335, "epoch": 1734, "lr": 3.9435680478165214e-05}
{"train_loss": 0.06837128102779388, "global_step": 154336, "epoch": 1734, "lr": 3.9435113830291246e-05}
{"train_loss": 0.11781885474920273, "global_step": 154337, "epoch": 1734, "lr": 3.943454718383759e-05}
{"train_loss": 0.10921237617731094, "global_step": 154338, "epoch": 1734, "lr": 3.943398053880435e-05}
{"train_loss": 0.09034325927495956, "global_step": 154339, "epoch": 1734, "lr": 3.943341389519155e-05}
{"train_loss": 0.047807883471250534, "global_step": 154340, "epoch": 1734, "lr": 3.943284725299933e-05}
{"train_loss": 0.10256806761026382, "global_step": 154341, "epoch": 1734, "lr": 3.943228061222771e-05}
{"train_loss": 0.05642261728644371, "global_step": 154342, "epoch": 1734, "lr": 3.943171397287681e-05}
{"train_loss": 0.07261350750923157, "global_step": 154343, "epoch": 1734, "lr": 3.943114733494667e-05}
{"train_loss": 0.0878705382347107, "global_step": 154344, "epoch": 1734, "lr": 3.94305806984374e-05}
{"train_loss": 0.06920461356639862, "global_step": 154345, "epoch": 1734, "lr": 3.9430014063349044e-05}
{"train_loss": 0.11089548468589783, "global_step": 154346, "epoch": 1734, "lr": 3.942944742968171e-05}
{"train_loss": 0.07729443907737732, "global_step": 154347, "epoch": 1734, "lr": 3.9428880797435444e-05}
{"train_loss": 0.12171618640422821, "global_step": 154348, "epoch": 1734, "lr": 3.9428314166610346e-05}
{"train_loss": 0.05317506566643715, "global_step": 154349, "epoch": 1734, "lr": 3.942774753720647e-05}
{"train_loss": 0.10059227049350739, "global_step": 154350, "epoch": 1734, "lr": 3.942718090922392e-05}
{"train_loss": 0.1409156322479248, "global_step": 154351, "epoch": 1734, "lr": 3.9426614282662746e-05}
{"train_loss": 0.07843246310949326, "global_step": 154352, "epoch": 1734, "lr": 3.942604765752305e-05}
{"train_loss": 0.0621366947889328, "global_step": 154353, "epoch": 1734, "lr": 3.942548103380486e-05}
{"train_loss": 0.08088566362857819, "global_step": 154354, "epoch": 1734, "lr": 3.9424914411508305e-05}
{"train_loss": 0.09323837608098984, "global_step": 154355, "epoch": 1734, "lr": 3.942434779063346e-05}
{"train_loss": 0.14273017644882202, "global_step": 154356, "epoch": 1734, "lr": 3.942378117118035e-05}
{"train_loss": 0.09652231633663177, "global_step": 154357, "epoch": 1734, "lr": 3.942321455314911e-05}
{"train_loss": 0.06290002912282944, "global_step": 154358, "epoch": 1734, "lr": 3.942264793653977e-05}
{"train_loss": 0.09207164496183395, "global_step": 154359, "epoch": 1734, "lr": 3.9422081321352446e-05}
{"train_loss": 0.0762636587023735, "global_step": 154360, "epoch": 1734, "lr": 3.9421514707587174e-05}
{"train_loss": 0.1202663704752922, "global_step": 154361, "epoch": 1734, "lr": 3.942094809524407e-05}
{"train_loss": 0.07602445781230927, "global_step": 154362, "epoch": 1734, "lr": 3.942038148432317e-05}
{"train_loss": 0.06190931424498558, "global_step": 154363, "epoch": 1734, "lr": 3.941981487482459e-05}
{"train_loss": 0.05124455690383911, "global_step": 154364, "epoch": 1734, "lr": 3.941924826674836e-05}
{"train_loss": 0.08950063586235046, "global_step": 154365, "epoch": 1734, "lr": 3.941868166009462e-05}
{"train_loss": 0.11531906574964523, "global_step": 154366, "epoch": 1734, "lr": 3.9418115054863383e-05}
{"train_loss": 0.12017396837472916, "global_step": 154367, "epoch": 1734, "lr": 3.9417548451054766e-05}
{"train_loss": 0.11001415550708771, "global_step": 154368, "epoch": 1734, "lr": 3.941698184866881e-05}
{"train_loss": 0.14081601798534393, "global_step": 154369, "epoch": 1734, "lr": 3.941641524770564e-05}
{"train_loss": 0.033408813178539276, "global_step": 154370, "epoch": 1734, "lr": 3.941584864816528e-05}
{"train_loss": 0.13176098465919495, "global_step": 154371, "epoch": 1734, "lr": 3.941528205004785e-05}
{"train_loss": 0.0882439911365509, "global_step": 154372, "epoch": 1734, "lr": 3.941471545335338e-05}
{"train_loss": 0.1975337266921997, "global_step": 154373, "epoch": 1734, "lr": 3.941414885808199e-05}
{"train_loss": 0.10830775648355484, "global_step": 154374, "epoch": 1734, "lr": 3.9413582264233746e-05}
{"train_loss": 0.10148593038320541, "global_step": 154375, "epoch": 1734, "lr": 3.9413015671808696e-05}
{"train_loss": 0.17458368837833405, "global_step": 154376, "epoch": 1734, "lr": 3.941244908080696e-05}
{"train_loss": 0.07525838166475296, "global_step": 154377, "epoch": 1734, "lr": 3.941188249122857e-05}
{"train_loss": 0.0862634927034378, "global_step": 154378, "epoch": 1734, "lr": 3.9411315903073644e-05}
{"train_loss": 0.071562759578228, "global_step": 154379, "epoch": 1734, "lr": 3.9410749316342214e-05}
{"train_loss": 0.08513683825731277, "global_step": 154380, "epoch": 1734, "lr": 3.941018273103441e-05}
{"train_loss": 0.14235834777355194, "global_step": 154381, "epoch": 1734, "lr": 3.940961614715024e-05}
{"train_loss": 0.15328359603881836, "global_step": 154382, "epoch": 1734, "lr": 3.940904956468985e-05}
{"train_loss": 0.10505098849534988, "global_step": 154383, "epoch": 1734, "lr": 3.940848298365326e-05}
{"train_loss": 0.08971671015024185, "global_step": 154384, "epoch": 1734, "lr": 3.9407916404040594e-05}
{"train_loss": 0.1558869481086731, "global_step": 154385, "epoch": 1734, "lr": 3.9407349825851884e-05}
{"train_loss": 0.10907556116580963, "global_step": 154386, "epoch": 1734, "lr": 3.940678324908725e-05}
{"train_loss": 0.05390424281358719, "global_step": 154387, "epoch": 1734, "lr": 3.940621667374672e-05}
{"train_loss": 0.09525768458843231, "global_step": 154388, "epoch": 1734, "lr": 3.940565009983041e-05}
{"train_loss": 0.10258939117193222, "global_step": 154389, "epoch": 1734, "lr": 3.940508352733837e-05}
{"train_loss": 0.06675919890403748, "global_step": 154390, "epoch": 1734, "lr": 3.9404516956270696e-05}
{"train_loss": 0.13689745962619781, "global_step": 154391, "epoch": 1734, "lr": 3.940395038662745e-05}
{"train_loss": 0.1069251224398613, "global_step": 154392, "epoch": 1734, "lr": 3.9403383818408725e-05}
{"train_loss": 0.06271117925643921, "global_step": 154393, "epoch": 1734, "lr": 3.940281725161459e-05}
{"train_loss": 0.09558545798063278, "global_step": 154394, "epoch": 1734, "lr": 3.940225068624509e-05}
{"train_loss": 0.08754079788923264, "global_step": 154395, "epoch": 1734, "lr": 3.940168412230035e-05}
{"train_loss": 0.09288305044174194, "global_step": 154396, "epoch": 1734, "lr": 3.940111755978042e-05}
{"train_loss": 0.13467583060264587, "global_step": 154397, "epoch": 1734, "lr": 3.940055099868538e-05}
{"train_loss": 0.08103231340646744, "global_step": 154398, "epoch": 1734, "lr": 3.93999844390153e-05}
{"train_loss": 0.10421866923570633, "global_step": 154399, "epoch": 1734, "lr": 3.939941788077028e-05}
{"train_loss": 0.0765124037861824, "global_step": 154400, "epoch": 1734, "lr": 3.939885132395035e-05}
{"train_loss": 0.0902429148554802, "global_step": 154401, "epoch": 1734, "lr": 3.939828476855565e-05}
{"train_loss": 0.1903030276298523, "global_step": 154402, "epoch": 1734, "lr": 3.9397718214586195e-05}
{"train_loss": 0.08649216592311859, "global_step": 154403, "epoch": 1734, "lr": 3.9397151662042106e-05}
{"train_loss": 0.07669951766729355, "global_step": 154404, "epoch": 1734, "lr": 3.939658511092342e-05}
{"train_loss": 0.14341357350349426, "global_step": 154405, "epoch": 1734, "lr": 3.939601856123025e-05}
{"train_loss": 0.09437691420316696, "global_step": 154406, "epoch": 1734, "lr": 3.939545201296265e-05}
{"train_loss": 0.0997130423784256, "global_step": 154407, "epoch": 1734, "lr": 3.939488546612071e-05}
{"train_loss": 0.0679047629237175, "global_step": 154408, "epoch": 1734, "lr": 3.9394318920704484e-05}
{"train_loss": 0.10087532550096512, "global_step": 154409, "epoch": 1734, "lr": 3.9393752376714065e-05}
{"train_loss": 0.09943359345197678, "global_step": 154410, "epoch": 1734, "lr": 3.9393185834149535e-05}
{"train_loss": 0.07968569546937943, "global_step": 154411, "epoch": 1734, "lr": 3.9392619293010955e-05}
{"train_loss": 0.09452804923057556, "global_step": 154412, "epoch": 1734, "lr": 3.9392052753298425e-05}
{"train_loss": 0.08710053563117981, "global_step": 154413, "epoch": 1734, "lr": 3.939148621501199e-05}
{"train_loss": 0.09704882140909687, "global_step": 154414, "epoch": 1734, "lr": 3.939091967815175e-05, "val_loss": 6.460049152374268}
{"train_loss": 0.09008770436048508, "global_step": 154415, "epoch": 1735, "lr": 3.939035314271775e-05}
{"train_loss": 0.08933502435684204, "global_step": 154416, "epoch": 1735, "lr": 3.938978660871011e-05}
{"train_loss": 0.058617860078811646, "global_step": 154417, "epoch": 1735, "lr": 3.938922007612886e-05}
{"train_loss": 0.05850423872470856, "global_step": 154418, "epoch": 1735, "lr": 3.938865354497413e-05}
{"train_loss": 0.13990147411823273, "global_step": 154419, "epoch": 1735, "lr": 3.9388087015245944e-05}
{"train_loss": 0.11742555350065231, "global_step": 154420, "epoch": 1735, "lr": 3.938752048694442e-05}
{"train_loss": 0.08552318066358566, "global_step": 154421, "epoch": 1735, "lr": 3.938695396006959e-05}
{"train_loss": 0.10193874686956406, "global_step": 154422, "epoch": 1735, "lr": 3.9386387434621574e-05}
{"train_loss": 0.10919185727834702, "global_step": 154423, "epoch": 1735, "lr": 3.938582091060041e-05}
{"train_loss": 0.09412931650876999, "global_step": 154424, "epoch": 1735, "lr": 3.938525438800622e-05}
{"train_loss": 0.10592149198055267, "global_step": 154425, "epoch": 1735, "lr": 3.9384687866839024e-05}
{"train_loss": 0.11676454544067383, "global_step": 154426, "epoch": 1735, "lr": 3.938412134709894e-05}
{"train_loss": 0.10261347144842148, "global_step": 154427, "epoch": 1735, "lr": 3.9383554828786036e-05}
{"train_loss": 0.16039931774139404, "global_step": 154428, "epoch": 1735, "lr": 3.938298831190038e-05}
{"train_loss": 0.14114819467067719, "global_step": 154429, "epoch": 1735, "lr": 3.938242179644205e-05}
{"train_loss": 0.08057534694671631, "global_step": 154430, "epoch": 1735, "lr": 3.938185528241113e-05}
{"train_loss": 0.10605818033218384, "global_step": 154431, "epoch": 1735, "lr": 3.9381288769807695e-05}
{"train_loss": 0.06237011402845383, "global_step": 154432, "epoch": 1735, "lr": 3.9380722258631794e-05}
{"train_loss": 0.1144777238368988, "global_step": 154433, "epoch": 1735, "lr": 3.938015574888355e-05}
{"train_loss": 0.08102598041296005, "global_step": 154434, "epoch": 1735, "lr": 3.9379589240562996e-05}
{"train_loss": 0.06713482737541199, "global_step": 154435, "epoch": 1735, "lr": 3.937902273367024e-05}
{"train_loss": 0.06277596205472946, "global_step": 154436, "epoch": 1735, "lr": 3.937845622820533e-05}
{"train_loss": 0.1370389312505722, "global_step": 154437, "epoch": 1735, "lr": 3.937788972416838e-05}
{"train_loss": 0.09016800671815872, "global_step": 154438, "epoch": 1735, "lr": 3.937732322155942e-05}
{"train_loss": 0.10130800306797028, "global_step": 154439, "epoch": 1735, "lr": 3.937675672037857e-05}
{"train_loss": 0.12589529156684875, "global_step": 154440, "epoch": 1735, "lr": 3.937619022062586e-05}
{"train_loss": 0.09724593907594681, "global_step": 154441, "epoch": 1735, "lr": 3.9375623722301416e-05}
{"train_loss": 0.0930405706167221, "global_step": 154442, "epoch": 1735, "lr": 3.937505722540528e-05}
{"train_loss": 0.08067701756954193, "global_step": 154443, "epoch": 1735, "lr": 3.937449072993754e-05}
{"train_loss": 0.11815302819013596, "global_step": 154444, "epoch": 1735, "lr": 3.937392423589826e-05}
{"train_loss": 0.15464861690998077, "global_step": 154445, "epoch": 1735, "lr": 3.937335774328754e-05}
{"train_loss": 0.14112699031829834, "global_step": 154446, "epoch": 1735, "lr": 3.937279125210544e-05}
{"train_loss": 0.08579528331756592, "global_step": 154447, "epoch": 1735, "lr": 3.937222476235203e-05}
{"train_loss": 0.08889412134885788, "global_step": 154448, "epoch": 1735, "lr": 3.93716582740274e-05}
{"train_loss": 0.05298657715320587, "global_step": 154449, "epoch": 1735, "lr": 3.937109178713162e-05}
{"train_loss": 0.08205532282590866, "global_step": 154450, "epoch": 1735, "lr": 3.9370525301664774e-05}
{"train_loss": 0.105522520840168, "global_step": 154451, "epoch": 1735, "lr": 3.936995881762692e-05}
{"train_loss": 0.11710870265960693, "global_step": 154452, "epoch": 1735, "lr": 3.936939233501817e-05}
{"train_loss": 0.11163230240345001, "global_step": 154453, "epoch": 1735, "lr": 3.936882585383854e-05}
{"train_loss": 0.10075855255126953, "global_step": 154454, "epoch": 1735, "lr": 3.936825937408817e-05}
{"train_loss": 0.037265755236148834, "global_step": 154455, "epoch": 1735, "lr": 3.9367692895767094e-05}
{"train_loss": 0.11454006284475327, "global_step": 154456, "epoch": 1735, "lr": 3.936712641887541e-05}
{"train_loss": 0.09805459529161453, "global_step": 154457, "epoch": 1735, "lr": 3.936655994341317e-05}
{"train_loss": 0.08618701249361038, "global_step": 154458, "epoch": 1735, "lr": 3.936599346938049e-05}
{"train_loss": 0.11721914261579514, "global_step": 154459, "epoch": 1735, "lr": 3.936542699677741e-05}
{"train_loss": 0.09471053630113602, "global_step": 154460, "epoch": 1735, "lr": 3.9364860525604026e-05}
{"train_loss": 0.12050765752792358, "global_step": 154461, "epoch": 1735, "lr": 3.936429405586038e-05}
{"train_loss": 0.11482454836368561, "global_step": 154462, "epoch": 1735, "lr": 3.93637275875466e-05}
{"train_loss": 0.055135030299425125, "global_step": 154463, "epoch": 1735, "lr": 3.936316112066273e-05}
{"train_loss": 0.11175284534692764, "global_step": 154464, "epoch": 1735, "lr": 3.936259465520885e-05}
{"train_loss": 0.08581115305423737, "global_step": 154465, "epoch": 1735, "lr": 3.936202819118506e-05}
{"train_loss": 0.07218417525291443, "global_step": 154466, "epoch": 1735, "lr": 3.93614617285914e-05}
{"train_loss": 0.07523848861455917, "global_step": 154467, "epoch": 1735, "lr": 3.9360895267427964e-05}
{"train_loss": 0.07215229421854019, "global_step": 154468, "epoch": 1735, "lr": 3.9360328807694826e-05}
{"train_loss": 0.11817894876003265, "global_step": 154469, "epoch": 1735, "lr": 3.9359762349392056e-05}
{"train_loss": 0.13922254741191864, "global_step": 154470, "epoch": 1735, "lr": 3.935919589251974e-05}
{"train_loss": 0.08726470172405243, "global_step": 154471, "epoch": 1735, "lr": 3.9358629437077965e-05}
{"train_loss": 0.09259224683046341, "global_step": 154472, "epoch": 1735, "lr": 3.9358062983066765e-05}
{"train_loss": 0.13307256996631622, "global_step": 154473, "epoch": 1735, "lr": 3.935749653048627e-05}
{"train_loss": 0.11889345198869705, "global_step": 154474, "epoch": 1735, "lr": 3.9356930079336505e-05}
{"train_loss": 0.04995468258857727, "global_step": 154475, "epoch": 1735, "lr": 3.93563636296176e-05}
{"train_loss": 0.07150935381650925, "global_step": 154476, "epoch": 1735, "lr": 3.935579718132957e-05}
{"train_loss": 0.13562408089637756, "global_step": 154477, "epoch": 1735, "lr": 3.9355230734472556e-05}
{"train_loss": 0.16680872440338135, "global_step": 154478, "epoch": 1735, "lr": 3.935466428904657e-05}
{"train_loss": 0.1562311053276062, "global_step": 154479, "epoch": 1735, "lr": 3.9354097845051746e-05}
{"train_loss": 0.10957691818475723, "global_step": 154480, "epoch": 1735, "lr": 3.935353140248811e-05}
{"train_loss": 0.1585305631160736, "global_step": 154481, "epoch": 1735, "lr": 3.935296496135577e-05}
{"train_loss": 0.14249135553836823, "global_step": 154482, "epoch": 1735, "lr": 3.93523985216548e-05}
{"train_loss": 0.05806026980280876, "global_step": 154483, "epoch": 1735, "lr": 3.935183208338526e-05}
{"train_loss": 0.10252219438552856, "global_step": 154484, "epoch": 1735, "lr": 3.935126564654726e-05}
{"train_loss": 0.14967887103557587, "global_step": 154485, "epoch": 1735, "lr": 3.935069921114084e-05}
{"train_loss": 0.1317431777715683, "global_step": 154486, "epoch": 1735, "lr": 3.935013277716608e-05}
{"train_loss": 0.08964243531227112, "global_step": 154487, "epoch": 1735, "lr": 3.934956634462307e-05}
{"train_loss": 0.07438162714242935, "global_step": 154488, "epoch": 1735, "lr": 3.9348999913511884e-05}
{"train_loss": 0.09854718297719955, "global_step": 154489, "epoch": 1735, "lr": 3.934843348383259e-05}
{"train_loss": 0.14711007475852966, "global_step": 154490, "epoch": 1735, "lr": 3.934786705558528e-05}
{"train_loss": 0.07450853288173676, "global_step": 154491, "epoch": 1735, "lr": 3.934730062877001e-05}
{"train_loss": 0.16462507843971252, "global_step": 154492, "epoch": 1735, "lr": 3.9346734203386875e-05}
{"train_loss": 0.14838439226150513, "global_step": 154493, "epoch": 1735, "lr": 3.9346167779435926e-05}
{"train_loss": 0.13424989581108093, "global_step": 154494, "epoch": 1735, "lr": 3.934560135691727e-05}
{"train_loss": 0.11583973467350006, "global_step": 154495, "epoch": 1735, "lr": 3.934503493583095e-05}
{"train_loss": 0.11857686936855316, "global_step": 154496, "epoch": 1735, "lr": 3.934446851617708e-05}
{"train_loss": 0.11200941354036331, "global_step": 154497, "epoch": 1735, "lr": 3.93439020979557e-05}
{"train_loss": 0.05979011207818985, "global_step": 154498, "epoch": 1735, "lr": 3.9343335681166924e-05}
{"train_loss": 0.11586278676986694, "global_step": 154499, "epoch": 1735, "lr": 3.934276926581078e-05}
{"train_loss": 0.14041480422019958, "global_step": 154500, "epoch": 1735, "lr": 3.934220285188737e-05}
{"train_loss": 0.11584730446338654, "global_step": 154501, "epoch": 1735, "lr": 3.93416364393968e-05}
{"train_loss": 0.12580208480358124, "global_step": 154502, "epoch": 1735, "lr": 3.9341070028339086e-05}
{"train_loss": 0.10552416766962308, "global_step": 154503, "epoch": 1735, "lr": 3.9340503618714365e-05, "val_loss": 6.355956077575684, "train_action_mse_error": 23.090267181396484}
{"train_loss": 0.08499781042337418, "global_step": 154504, "epoch": 1736, "lr": 3.9339937210522674e-05}
{"train_loss": 0.15987183153629303, "global_step": 154505, "epoch": 1736, "lr": 3.933937080376409e-05}
{"train_loss": 0.15079665184020996, "global_step": 154506, "epoch": 1736, "lr": 3.93388043984387e-05}
{"train_loss": 0.11720301955938339, "global_step": 154507, "epoch": 1736, "lr": 3.933823799454658e-05}
{"train_loss": 0.11172255128622055, "global_step": 154508, "epoch": 1736, "lr": 3.93376715920878e-05}
{"train_loss": 0.11686834692955017, "global_step": 154509, "epoch": 1736, "lr": 3.933710519106246e-05}
{"train_loss": 0.08549455553293228, "global_step": 154510, "epoch": 1736, "lr": 3.9336538791470584e-05}
{"train_loss": 0.10119816660881042, "global_step": 154511, "epoch": 1736, "lr": 3.933597239331231e-05}
{"train_loss": 0.11222650110721588, "global_step": 154512, "epoch": 1736, "lr": 3.933540599658766e-05}
{"train_loss": 0.09140380471944809, "global_step": 154513, "epoch": 1736, "lr": 3.933483960129675e-05}
{"train_loss": 0.06690364331007004, "global_step": 154514, "epoch": 1736, "lr": 3.933427320743963e-05}
{"train_loss": 0.14779628813266754, "global_step": 154515, "epoch": 1736, "lr": 3.9333706815016405e-05}
{"train_loss": 0.09752457588911057, "global_step": 154516, "epoch": 1736, "lr": 3.9333140424027106e-05}
{"train_loss": 0.08621012419462204, "global_step": 154517, "epoch": 1736, "lr": 3.933257403447184e-05}
{"train_loss": 0.09248026460409164, "global_step": 154518, "epoch": 1736, "lr": 3.93320076463507e-05}
{"train_loss": 0.10461478680372238, "global_step": 154519, "epoch": 1736, "lr": 3.933144125966372e-05}
{"train_loss": 0.0960058867931366, "global_step": 154520, "epoch": 1736, "lr": 3.9330874874411026e-05}
{"train_loss": 0.062184400856494904, "global_step": 154521, "epoch": 1736, "lr": 3.933030849059264e-05}
{"train_loss": 0.10662352293729782, "global_step": 154522, "epoch": 1736, "lr": 3.932974210820868e-05}
{"train_loss": 0.14061686396598816, "global_step": 154523, "epoch": 1736, "lr": 3.93291757272592e-05}
{"train_loss": 0.060424961149692535, "global_step": 154524, "epoch": 1736, "lr": 3.932860934774428e-05}
{"train_loss": 0.09180890768766403, "global_step": 154525, "epoch": 1736, "lr": 3.9328042969664e-05}
{"train_loss": 0.08020567893981934, "global_step": 154526, "epoch": 1736, "lr": 3.932747659301844e-05}
{"train_loss": 0.10165638476610184, "global_step": 154527, "epoch": 1736, "lr": 3.932691021780766e-05}
{"train_loss": 0.1645657867193222, "global_step": 154528, "epoch": 1736, "lr": 3.932634384403177e-05}
{"train_loss": 0.07609840482473373, "global_step": 154529, "epoch": 1736, "lr": 3.9325777471690794e-05}
{"train_loss": 0.09547696262598038, "global_step": 154530, "epoch": 1736, "lr": 3.9325211100784857e-05}
{"train_loss": 0.0908980518579483, "global_step": 154531, "epoch": 1736, "lr": 3.9324644731313996e-05}
{"train_loss": 0.0844595655798912, "global_step": 154532, "epoch": 1736, "lr": 3.932407836327833e-05}
{"train_loss": 0.11047276109457016, "global_step": 154533, "epoch": 1736, "lr": 3.9323511996677886e-05}
{"train_loss": 0.15450002253055573, "global_step": 154534, "epoch": 1736, "lr": 3.932294563151279e-05}
{"train_loss": 0.12919382750988007, "global_step": 154535, "epoch": 1736, "lr": 3.9322379267783074e-05}
{"train_loss": 0.18318398296833038, "global_step": 154536, "epoch": 1736, "lr": 3.932181290548883e-05}
{"train_loss": 0.08832322806119919, "global_step": 154537, "epoch": 1736, "lr": 3.932124654463016e-05}
{"train_loss": 0.09242978692054749, "global_step": 154538, "epoch": 1736, "lr": 3.93206801852071e-05}
{"train_loss": 0.10223466902971268, "global_step": 154539, "epoch": 1736, "lr": 3.932011382721976e-05}
{"train_loss": 0.13106586039066315, "global_step": 154540, "epoch": 1736, "lr": 3.9319547470668186e-05}
{"train_loss": 0.06602805852890015, "global_step": 154541, "epoch": 1736, "lr": 3.931898111555249e-05}
{"train_loss": 0.18273493647575378, "global_step": 154542, "epoch": 1736, "lr": 3.931841476187271e-05}
{"train_loss": 0.11154864728450775, "global_step": 154543, "epoch": 1736, "lr": 3.9317848409628944e-05}
{"train_loss": 0.11893656104803085, "global_step": 154544, "epoch": 1736, "lr": 3.9317282058821256e-05}
{"train_loss": 0.16156253218650818, "global_step": 154545, "epoch": 1736, "lr": 3.9316715709449735e-05}
{"train_loss": 0.09717289358377457, "global_step": 154546, "epoch": 1736, "lr": 3.931614936151446e-05}
{"train_loss": 0.10567038506269455, "global_step": 154547, "epoch": 1736, "lr": 3.9315583015015496e-05}
{"train_loss": 0.15906736254692078, "global_step": 154548, "epoch": 1736, "lr": 3.93150166699529e-05}
{"train_loss": 0.1201566606760025, "global_step": 154549, "epoch": 1736, "lr": 3.9314450326326796e-05}
{"train_loss": 0.049566928297281265, "global_step": 154550, "epoch": 1736, "lr": 3.9313883984137214e-05}
{"train_loss": 0.15626351535320282, "global_step": 154551, "epoch": 1736, "lr": 3.9313317643384267e-05}
{"train_loss": 0.07952294498682022, "global_step": 154552, "epoch": 1736, "lr": 3.9312751304067996e-05}
{"train_loss": 0.08587241917848587, "global_step": 154553, "epoch": 1736, "lr": 3.931218496618849e-05}
{"train_loss": 0.06384990364313126, "global_step": 154554, "epoch": 1736, "lr": 3.931161862974586e-05}
{"train_loss": 0.15498721599578857, "global_step": 154555, "epoch": 1736, "lr": 3.931105229474012e-05}
{"train_loss": 0.10103420168161392, "global_step": 154556, "epoch": 1736, "lr": 3.931048596117142e-05}
{"train_loss": 0.12507857382297516, "global_step": 154557, "epoch": 1736, "lr": 3.9309919629039755e-05}
{"train_loss": 0.07918469607830048, "global_step": 154558, "epoch": 1736, "lr": 3.930935329834527e-05}
{"train_loss": 0.15776345133781433, "global_step": 154559, "epoch": 1736, "lr": 3.930878696908798e-05}
{"train_loss": 0.0850384309887886, "global_step": 154560, "epoch": 1736, "lr": 3.9308220641268037e-05}
{"train_loss": 0.08191293478012085, "global_step": 154561, "epoch": 1736, "lr": 3.9307654314885444e-05}
{"train_loss": 0.13208307325839996, "global_step": 154562, "epoch": 1736, "lr": 3.9307087989940314e-05}
{"train_loss": 0.07817872613668442, "global_step": 154563, "epoch": 1736, "lr": 3.9306521666432714e-05}
{"train_loss": 0.07360044121742249, "global_step": 154564, "epoch": 1736, "lr": 3.930595534436272e-05}
{"train_loss": 0.1362413763999939, "global_step": 154565, "epoch": 1736, "lr": 3.930538902373041e-05}
{"train_loss": 0.12259003520011902, "global_step": 154566, "epoch": 1736, "lr": 3.9304822704535866e-05}
{"train_loss": 0.11375747621059418, "global_step": 154567, "epoch": 1736, "lr": 3.930425638677915e-05}
{"train_loss": 0.07014912366867065, "global_step": 154568, "epoch": 1736, "lr": 3.930369007046036e-05}
{"train_loss": 0.10049866139888763, "global_step": 154569, "epoch": 1736, "lr": 3.930312375557954e-05}
{"train_loss": 0.0968608558177948, "global_step": 154570, "epoch": 1736, "lr": 3.930255744213681e-05}
{"train_loss": 0.14497284591197968, "global_step": 154571, "epoch": 1736, "lr": 3.9301991130132185e-05}
{"train_loss": 0.1427827775478363, "global_step": 154572, "epoch": 1736, "lr": 3.930142481956579e-05}
{"train_loss": 0.10727830231189728, "global_step": 154573, "epoch": 1736, "lr": 3.930085851043771e-05}
{"train_loss": 0.09688553959131241, "global_step": 154574, "epoch": 1736, "lr": 3.9300292202747966e-05}
{"train_loss": 0.08397264778614044, "global_step": 154575, "epoch": 1736, "lr": 3.92997258964967e-05}
{"train_loss": 0.09004685282707214, "global_step": 154576, "epoch": 1736, "lr": 3.929915959168393e-05}
{"train_loss": 0.1311725229024887, "global_step": 154577, "epoch": 1736, "lr": 3.929859328830977e-05}
{"train_loss": 0.10144581645727158, "global_step": 154578, "epoch": 1736, "lr": 3.929802698637428e-05}
{"train_loss": 0.06377318501472473, "global_step": 154579, "epoch": 1736, "lr": 3.929746068587755e-05}
{"train_loss": 0.07135124504566193, "global_step": 154580, "epoch": 1736, "lr": 3.929689438681962e-05}
{"train_loss": 0.091395802795887, "global_step": 154581, "epoch": 1736, "lr": 3.929632808920062e-05}
{"train_loss": 0.12568923830986023, "global_step": 154582, "epoch": 1736, "lr": 3.9295761793020594e-05}
{"train_loss": 0.14835378527641296, "global_step": 154583, "epoch": 1736, "lr": 3.9295195498279616e-05}
{"train_loss": 0.11255985498428345, "global_step": 154584, "epoch": 1736, "lr": 3.929462920497776e-05}
{"train_loss": 0.11524935811758041, "global_step": 154585, "epoch": 1736, "lr": 3.929406291311513e-05}
{"train_loss": 0.06790760159492493, "global_step": 154586, "epoch": 1736, "lr": 3.9293496622691766e-05}
{"train_loss": 0.07906198501586914, "global_step": 154587, "epoch": 1736, "lr": 3.929293033370778e-05}
{"train_loss": 0.09792304784059525, "global_step": 154588, "epoch": 1736, "lr": 3.92923640461632e-05}
{"train_loss": 0.09978293627500534, "global_step": 154589, "epoch": 1736, "lr": 3.9291797760058134e-05}
{"train_loss": 0.09928908944129944, "global_step": 154590, "epoch": 1736, "lr": 3.929123147539268e-05}
{"train_loss": 0.09526415914297104, "global_step": 154591, "epoch": 1736, "lr": 3.929066519216687e-05}
{"train_loss": 0.10676902854877911, "global_step": 154592, "epoch": 1736, "lr": 3.929009891038082e-05, "val_loss": 6.391510963439941}
{"train_loss": 0.09465596824884415, "global_step": 154593, "epoch": 1737, "lr": 3.9289532630034556e-05}
{"train_loss": 0.06953691691160202, "global_step": 154594, "epoch": 1737, "lr": 3.9288966351128214e-05}
{"train_loss": 0.09889961034059525, "global_step": 154595, "epoch": 1737, "lr": 3.9288400073661815e-05}
{"train_loss": 0.08350135385990143, "global_step": 154596, "epoch": 1737, "lr": 3.928783379763548e-05}
{"train_loss": 0.10951162129640579, "global_step": 154597, "epoch": 1737, "lr": 3.928726752304925e-05}
{"train_loss": 0.11685937643051147, "global_step": 154598, "epoch": 1737, "lr": 3.9286701249903234e-05}
{"train_loss": 0.06157039478421211, "global_step": 154599, "epoch": 1737, "lr": 3.9286134978197466e-05}
{"train_loss": 0.11601033061742783, "global_step": 154600, "epoch": 1737, "lr": 3.9285568707932066e-05}
{"train_loss": 0.1270381510257721, "global_step": 154601, "epoch": 1737, "lr": 3.928500243910709e-05}
{"train_loss": 0.08686911314725876, "global_step": 154602, "epoch": 1737, "lr": 3.928443617172261e-05}
{"train_loss": 0.08918382972478867, "global_step": 154603, "epoch": 1737, "lr": 3.92838699057787e-05}
{"train_loss": 0.17218835651874542, "global_step": 154604, "epoch": 1737, "lr": 3.9283303641275456e-05}
{"train_loss": 0.1363026350736618, "global_step": 154605, "epoch": 1737, "lr": 3.9282737378212934e-05}
{"train_loss": 0.08572211116552353, "global_step": 154606, "epoch": 1737, "lr": 3.928217111659123e-05}
{"train_loss": 0.09286901354789734, "global_step": 154607, "epoch": 1737, "lr": 3.928160485641038e-05}
{"train_loss": 0.12093653529882431, "global_step": 154608, "epoch": 1737, "lr": 3.9281038597670496e-05}
{"train_loss": 0.08497576415538788, "global_step": 154609, "epoch": 1737, "lr": 3.9280472340371666e-05}
{"train_loss": 0.06176547333598137, "global_step": 154610, "epoch": 1737, "lr": 3.927990608451392e-05}
{"train_loss": 0.12140551954507828, "global_step": 154611, "epoch": 1737, "lr": 3.9279339830097386e-05}
{"train_loss": 0.13264767825603485, "global_step": 154612, "epoch": 1737, "lr": 3.9278773577122084e-05}
{"train_loss": 0.07708559185266495, "global_step": 154613, "epoch": 1737, "lr": 3.9278207325588154e-05}
{"train_loss": 0.07735565304756165, "global_step": 154614, "epoch": 1737, "lr": 3.927764107549561e-05}
{"train_loss": 0.13944011926651, "global_step": 154615, "epoch": 1737, "lr": 3.9277074826844574e-05}
{"train_loss": 0.0986422523856163, "global_step": 154616, "epoch": 1737, "lr": 3.9276508579635086e-05}
{"train_loss": 0.0995522141456604, "global_step": 154617, "epoch": 1737, "lr": 3.927594233386727e-05}
{"train_loss": 0.16204136610031128, "global_step": 154618, "epoch": 1737, "lr": 3.927537608954114e-05}
{"train_loss": 0.07745163142681122, "global_step": 154619, "epoch": 1737, "lr": 3.9274809846656834e-05}
{"train_loss": 0.08250638097524643, "global_step": 154620, "epoch": 1737, "lr": 3.927424360521438e-05}
{"train_loss": 0.05335129797458649, "global_step": 154621, "epoch": 1737, "lr": 3.927367736521388e-05}
{"train_loss": 0.08213219046592712, "global_step": 154622, "epoch": 1737, "lr": 3.9273111126655405e-05}
{"train_loss": 0.1474415510892868, "global_step": 154623, "epoch": 1737, "lr": 3.927254488953903e-05}
{"train_loss": 0.06185878813266754, "global_step": 154624, "epoch": 1737, "lr": 3.9271978653864826e-05}
{"train_loss": 0.0955096185207367, "global_step": 154625, "epoch": 1737, "lr": 3.9271412419632886e-05}
{"train_loss": 0.07924117892980576, "global_step": 154626, "epoch": 1737, "lr": 3.927084618684325e-05}
{"train_loss": 0.0939277708530426, "global_step": 154627, "epoch": 1737, "lr": 3.927027995549602e-05}
{"train_loss": 0.09114900976419449, "global_step": 154628, "epoch": 1737, "lr": 3.9269713725591296e-05}
{"train_loss": 0.1535002738237381, "global_step": 154629, "epoch": 1737, "lr": 3.92691474971291e-05}
{"train_loss": 0.14067338407039642, "global_step": 154630, "epoch": 1737, "lr": 3.9268581270109564e-05}
{"train_loss": 0.14719258248806, "global_step": 154631, "epoch": 1737, "lr": 3.926801504453271e-05}
{"train_loss": 0.09949135780334473, "global_step": 154632, "epoch": 1737, "lr": 3.926744882039866e-05}
{"train_loss": 0.09679514169692993, "global_step": 154633, "epoch": 1737, "lr": 3.926688259770746e-05}
{"train_loss": 0.03760608285665512, "global_step": 154634, "epoch": 1737, "lr": 3.92663163764592e-05}
{"train_loss": 0.09152545779943466, "global_step": 154635, "epoch": 1737, "lr": 3.926575015665395e-05}
{"train_loss": 0.10386356711387634, "global_step": 154636, "epoch": 1737, "lr": 3.92651839382918e-05}
{"train_loss": 0.05602102354168892, "global_step": 154637, "epoch": 1737, "lr": 3.92646177213728e-05}
{"train_loss": 0.09108618646860123, "global_step": 154638, "epoch": 1737, "lr": 3.926405150589706e-05}
{"train_loss": 0.04952554032206535, "global_step": 154639, "epoch": 1737, "lr": 3.926348529186462e-05}
{"train_loss": 0.08467895537614822, "global_step": 154640, "epoch": 1737, "lr": 3.926291907927559e-05}
{"train_loss": 0.1075783371925354, "global_step": 154641, "epoch": 1737, "lr": 3.926235286813002e-05}
{"train_loss": 0.18002517521381378, "global_step": 154642, "epoch": 1737, "lr": 3.926178665842799e-05}
{"train_loss": 0.0946488231420517, "global_step": 154643, "epoch": 1737, "lr": 3.926122045016958e-05}
{"train_loss": 0.11311998963356018, "global_step": 154644, "epoch": 1737, "lr": 3.9260654243354865e-05}
{"train_loss": 0.10649117082357407, "global_step": 154645, "epoch": 1737, "lr": 3.926008803798395e-05}
{"train_loss": 0.09019093960523605, "global_step": 154646, "epoch": 1737, "lr": 3.925952183405685e-05}
{"train_loss": 0.12628453969955444, "global_step": 154647, "epoch": 1737, "lr": 3.9258955631573713e-05}
{"train_loss": 0.05299897864460945, "global_step": 154648, "epoch": 1737, "lr": 3.925838943053455e-05}
{"train_loss": 0.0636586919426918, "global_step": 154649, "epoch": 1737, "lr": 3.9257823230939486e-05}
{"train_loss": 0.059358980506658554, "global_step": 154650, "epoch": 1737, "lr": 3.925725703278856e-05}
{"train_loss": 0.10194080322980881, "global_step": 154651, "epoch": 1737, "lr": 3.925669083608188e-05}
{"train_loss": 0.08375293016433716, "global_step": 154652, "epoch": 1737, "lr": 3.9256124640819494e-05}
{"train_loss": 0.1045156940817833, "global_step": 154653, "epoch": 1737, "lr": 3.925555844700151e-05}
{"train_loss": 0.1222122311592102, "global_step": 154654, "epoch": 1737, "lr": 3.925499225462795e-05}
{"train_loss": 0.1288023293018341, "global_step": 154655, "epoch": 1737, "lr": 3.9254426063698964e-05}
{"train_loss": 0.09544327110052109, "global_step": 154656, "epoch": 1737, "lr": 3.925385987421456e-05}
{"train_loss": 0.1424621194601059, "global_step": 154657, "epoch": 1737, "lr": 3.925329368617487e-05}
{"train_loss": 0.08525008708238602, "global_step": 154658, "epoch": 1737, "lr": 3.925272749957993e-05}
{"train_loss": 0.08627574890851974, "global_step": 154659, "epoch": 1737, "lr": 3.925216131442983e-05}
{"train_loss": 0.2267359048128128, "global_step": 154660, "epoch": 1737, "lr": 3.925159513072464e-05}
{"train_loss": 0.10313217341899872, "global_step": 154661, "epoch": 1737, "lr": 3.925102894846445e-05}
{"train_loss": 0.13609124720096588, "global_step": 154662, "epoch": 1737, "lr": 3.925046276764933e-05}
{"train_loss": 0.0585520900785923, "global_step": 154663, "epoch": 1737, "lr": 3.924989658827933e-05}
{"train_loss": 0.11210045963525772, "global_step": 154664, "epoch": 1737, "lr": 3.924933041035459e-05}
{"train_loss": 0.0343196876347065, "global_step": 154665, "epoch": 1737, "lr": 3.924876423387511e-05}
{"train_loss": 0.11071497946977615, "global_step": 154666, "epoch": 1737, "lr": 3.924819805884104e-05}
{"train_loss": 0.09704136848449707, "global_step": 154667, "epoch": 1737, "lr": 3.924763188525238e-05}
{"train_loss": 0.08344133198261261, "global_step": 154668, "epoch": 1737, "lr": 3.9247065713109276e-05}
{"train_loss": 0.08298566192388535, "global_step": 154669, "epoch": 1737, "lr": 3.9246499542411746e-05}
{"train_loss": 0.1053692102432251, "global_step": 154670, "epoch": 1737, "lr": 3.924593337315992e-05}
{"train_loss": 0.1307508647441864, "global_step": 154671, "epoch": 1737, "lr": 3.924536720535382e-05}
{"train_loss": 0.08503518998622894, "global_step": 154672, "epoch": 1737, "lr": 3.924480103899358e-05}
{"train_loss": 0.06867074966430664, "global_step": 154673, "epoch": 1737, "lr": 3.9244234874079217e-05}
{"train_loss": 0.09851732105016708, "global_step": 154674, "epoch": 1737, "lr": 3.924366871061086e-05}
{"train_loss": 0.10462656617164612, "global_step": 154675, "epoch": 1737, "lr": 3.9243102548588534e-05}
{"train_loss": 0.0639260783791542, "global_step": 154676, "epoch": 1737, "lr": 3.924253638801237e-05}
{"train_loss": 0.07670959085226059, "global_step": 154677, "epoch": 1737, "lr": 3.924197022888238e-05}
{"train_loss": 0.07173611223697662, "global_step": 154678, "epoch": 1737, "lr": 3.9241404071198714e-05}
{"train_loss": 0.12327306717634201, "global_step": 154679, "epoch": 1737, "lr": 3.92408379149614e-05}
{"train_loss": 0.06835494935512543, "global_step": 154680, "epoch": 1737, "lr": 3.92402717601705e-05}
{"train_loss": 0.099096761995487, "global_step": 154681, "epoch": 1737, "lr": 3.923970560682615e-05, "val_loss": 6.350991249084473}
{"train_loss": 0.09853608906269073, "global_step": 154682, "epoch": 1738, "lr": 3.923913945492836e-05}
{"train_loss": 0.06495566666126251, "global_step": 154683, "epoch": 1738, "lr": 3.923857330447726e-05}
{"train_loss": 0.10164700448513031, "global_step": 154684, "epoch": 1738, "lr": 3.9238007155472876e-05}
{"train_loss": 0.10408075898885727, "global_step": 154685, "epoch": 1738, "lr": 3.9237441007915344e-05}
{"train_loss": 0.06475716084241867, "global_step": 154686, "epoch": 1738, "lr": 3.923687486180467e-05}
{"train_loss": 0.10197898745536804, "global_step": 154687, "epoch": 1738, "lr": 3.9236308717141005e-05}
{"train_loss": 0.10707370936870575, "global_step": 154688, "epoch": 1738, "lr": 3.9235742573924354e-05}
{"train_loss": 0.06754595041275024, "global_step": 154689, "epoch": 1738, "lr": 3.923517643215485e-05}
{"train_loss": 0.09055274724960327, "global_step": 154690, "epoch": 1738, "lr": 3.923461029183253e-05}
{"train_loss": 0.10537314414978027, "global_step": 154691, "epoch": 1738, "lr": 3.9234044152957506e-05}
{"train_loss": 0.025452716276049614, "global_step": 154692, "epoch": 1738, "lr": 3.923347801552981e-05}
{"train_loss": 0.08813094347715378, "global_step": 154693, "epoch": 1738, "lr": 3.923291187954956e-05}
{"train_loss": 0.08466670662164688, "global_step": 154694, "epoch": 1738, "lr": 3.923234574501679e-05}
{"train_loss": 0.07823145389556885, "global_step": 154695, "epoch": 1738, "lr": 3.923177961193162e-05}
{"train_loss": 0.1245637834072113, "global_step": 154696, "epoch": 1738, "lr": 3.9231213480294095e-05}
{"train_loss": 0.07750025391578674, "global_step": 154697, "epoch": 1738, "lr": 3.923064735010432e-05}
{"train_loss": 0.08208229392766953, "global_step": 154698, "epoch": 1738, "lr": 3.923008122136234e-05}
{"train_loss": 0.08307899534702301, "global_step": 154699, "epoch": 1738, "lr": 3.9229515094068235e-05}
{"train_loss": 0.17022086679935455, "global_step": 154700, "epoch": 1738, "lr": 3.922894896822211e-05}
{"train_loss": 0.129702627658844, "global_step": 154701, "epoch": 1738, "lr": 3.9228382843824e-05}
{"train_loss": 0.0685831755399704, "global_step": 154702, "epoch": 1738, "lr": 3.9227816720874025e-05}
{"train_loss": 0.07325397431850433, "global_step": 154703, "epoch": 1738, "lr": 3.922725059937222e-05}
{"train_loss": 0.07619893550872803, "global_step": 154704, "epoch": 1738, "lr": 3.92266844793187e-05}
{"train_loss": 0.07207764685153961, "global_step": 154705, "epoch": 1738, "lr": 3.9226118360713497e-05}
{"train_loss": 0.13562093675136566, "global_step": 154706, "epoch": 1738, "lr": 3.9225552243556736e-05}
{"train_loss": 0.08562509715557098, "global_step": 154707, "epoch": 1738, "lr": 3.922498612784844e-05}
{"train_loss": 0.10667796432971954, "global_step": 154708, "epoch": 1738, "lr": 3.9224420013588744e-05}
{"train_loss": 0.11874695867300034, "global_step": 154709, "epoch": 1738, "lr": 3.922385390077766e-05}
{"train_loss": 0.04600030928850174, "global_step": 154710, "epoch": 1738, "lr": 3.922328778941533e-05}
{"train_loss": 0.10026489943265915, "global_step": 154711, "epoch": 1738, "lr": 3.9222721679501774e-05}
{"train_loss": 0.06825367361307144, "global_step": 154712, "epoch": 1738, "lr": 3.922215557103711e-05}
{"train_loss": 0.0811241939663887, "global_step": 154713, "epoch": 1738, "lr": 3.922158946402138e-05}
{"train_loss": 0.05126362293958664, "global_step": 154714, "epoch": 1738, "lr": 3.9221023358454686e-05}
{"train_loss": 0.08695428818464279, "global_step": 154715, "epoch": 1738, "lr": 3.922045725433708e-05}
{"train_loss": 0.09505674988031387, "global_step": 154716, "epoch": 1738, "lr": 3.921989115166867e-05}
{"train_loss": 0.13133712112903595, "global_step": 154717, "epoch": 1738, "lr": 3.921932505044952e-05}
{"train_loss": 0.14065483212471008, "global_step": 154718, "epoch": 1738, "lr": 3.9218758950679676e-05}
{"train_loss": 0.06044388934969902, "global_step": 154719, "epoch": 1738, "lr": 3.921819285235926e-05}
{"train_loss": 0.10617053508758545, "global_step": 154720, "epoch": 1738, "lr": 3.92176267554883e-05}
{"train_loss": 0.06989558041095734, "global_step": 154721, "epoch": 1738, "lr": 3.921706066006693e-05}
{"train_loss": 0.15269958972930908, "global_step": 154722, "epoch": 1738, "lr": 3.921649456609517e-05}
{"train_loss": 0.08881517499685287, "global_step": 154723, "epoch": 1738, "lr": 3.921592847357314e-05}
{"train_loss": 0.05903135612607002, "global_step": 154724, "epoch": 1738, "lr": 3.9215362382500875e-05}
{"train_loss": 0.09061028063297272, "global_step": 154725, "epoch": 1738, "lr": 3.92147962928785e-05}
{"train_loss": 0.03819020465016365, "global_step": 154726, "epoch": 1738, "lr": 3.921423020470604e-05}
{"train_loss": 0.12542465329170227, "global_step": 154727, "epoch": 1738, "lr": 3.9213664117983614e-05}
{"train_loss": 0.0889478400349617, "global_step": 154728, "epoch": 1738, "lr": 3.921309803271126e-05}
{"train_loss": 0.07035920768976212, "global_step": 154729, "epoch": 1738, "lr": 3.9212531948889095e-05}
{"train_loss": 0.1141880601644516, "global_step": 154730, "epoch": 1738, "lr": 3.921196586651715e-05}
{"train_loss": 0.09349459409713745, "global_step": 154731, "epoch": 1738, "lr": 3.921139978559555e-05}
{"train_loss": 0.0539327934384346, "global_step": 154732, "epoch": 1738, "lr": 3.9210833706124325e-05}
{"train_loss": 0.12203400582075119, "global_step": 154733, "epoch": 1738, "lr": 3.921026762810359e-05}
{"train_loss": 0.1035524532198906, "global_step": 154734, "epoch": 1738, "lr": 3.920970155153338e-05}
{"train_loss": 0.13089711964130402, "global_step": 154735, "epoch": 1738, "lr": 3.920913547641381e-05}
{"train_loss": 0.06756275147199631, "global_step": 154736, "epoch": 1738, "lr": 3.920856940274495e-05}
{"train_loss": 0.07375182211399078, "global_step": 154737, "epoch": 1738, "lr": 3.920800333052684e-05}
{"train_loss": 0.0645061656832695, "global_step": 154738, "epoch": 1738, "lr": 3.920743725975961e-05}
{"train_loss": 0.11102259159088135, "global_step": 154739, "epoch": 1738, "lr": 3.920687119044328e-05}
{"train_loss": 0.0843546986579895, "global_step": 154740, "epoch": 1738, "lr": 3.9206305122577985e-05}
{"train_loss": 0.07275145500898361, "global_step": 154741, "epoch": 1738, "lr": 3.920573905616374e-05}
{"train_loss": 0.09575837105512619, "global_step": 154742, "epoch": 1738, "lr": 3.920517299120068e-05}
{"train_loss": 0.1198941171169281, "global_step": 154743, "epoch": 1738, "lr": 3.920460692768883e-05}
{"train_loss": 0.10648796707391739, "global_step": 154744, "epoch": 1738, "lr": 3.920404086562831e-05}
{"train_loss": 0.07589279115200043, "global_step": 154745, "epoch": 1738, "lr": 3.920347480501915e-05}
{"train_loss": 0.11015500873327255, "global_step": 154746, "epoch": 1738, "lr": 3.9202908745861474e-05}
{"train_loss": 0.10642525553703308, "global_step": 154747, "epoch": 1738, "lr": 3.920234268815532e-05}
{"train_loss": 0.07311996817588806, "global_step": 154748, "epoch": 1738, "lr": 3.92017766319008e-05}
{"train_loss": 0.06400761753320694, "global_step": 154749, "epoch": 1738, "lr": 3.920121057709794e-05}
{"train_loss": 0.059241414070129395, "global_step": 154750, "epoch": 1738, "lr": 3.920064452374688e-05}
{"train_loss": 0.15154767036437988, "global_step": 154751, "epoch": 1738, "lr": 3.920007847184762e-05}
{"train_loss": 0.06683716177940369, "global_step": 154752, "epoch": 1738, "lr": 3.919951242140032e-05}
{"train_loss": 0.06338860839605331, "global_step": 154753, "epoch": 1738, "lr": 3.919894637240498e-05}
{"train_loss": 0.10640428960323334, "global_step": 154754, "epoch": 1738, "lr": 3.9198380324861726e-05}
{"train_loss": 0.06373950839042664, "global_step": 154755, "epoch": 1738, "lr": 3.919781427877063e-05}
{"train_loss": 0.024202080443501472, "global_step": 154756, "epoch": 1738, "lr": 3.919724823413173e-05}
{"train_loss": 0.10216858983039856, "global_step": 154757, "epoch": 1738, "lr": 3.9196682190945154e-05}
{"train_loss": 0.14727701246738434, "global_step": 154758, "epoch": 1738, "lr": 3.9196116149210925e-05}
{"train_loss": 0.1353686898946762, "global_step": 154759, "epoch": 1738, "lr": 3.9195550108929173e-05}
{"train_loss": 0.12339130789041519, "global_step": 154760, "epoch": 1738, "lr": 3.919498407009993e-05}
{"train_loss": 0.09143847972154617, "global_step": 154761, "epoch": 1738, "lr": 3.91944180327233e-05}
{"train_loss": 0.03623569756746292, "global_step": 154762, "epoch": 1738, "lr": 3.919385199679933e-05}
{"train_loss": 0.10873580724000931, "global_step": 154763, "epoch": 1738, "lr": 3.919328596232815e-05}
{"train_loss": 0.1264626383781433, "global_step": 154764, "epoch": 1738, "lr": 3.9192719929309764e-05}
{"train_loss": 0.08386760950088501, "global_step": 154765, "epoch": 1738, "lr": 3.919215389774431e-05}
{"train_loss": 0.0864892452955246, "global_step": 154766, "epoch": 1738, "lr": 3.919158786763182e-05}
{"train_loss": 0.07549671083688736, "global_step": 154767, "epoch": 1738, "lr": 3.9191021838972414e-05}
{"train_loss": 0.14434584975242615, "global_step": 154768, "epoch": 1738, "lr": 3.919045581176612e-05}
{"train_loss": 0.08774030953645706, "global_step": 154769, "epoch": 1738, "lr": 3.918988978601306e-05}
{"train_loss": 0.09141460266173555, "global_step": 154770, "epoch": 1738, "lr": 3.918932376171326e-05, "val_loss": 6.48250675201416}
{"train_loss": 0.08553439378738403, "global_step": 154771, "epoch": 1739, "lr": 3.9188757738866835e-05}
{"train_loss": 0.1307915598154068, "global_step": 154772, "epoch": 1739, "lr": 3.918819171747385e-05}
{"train_loss": 0.06617368757724762, "global_step": 154773, "epoch": 1739, "lr": 3.918762569753439e-05}
{"train_loss": 0.1082414910197258, "global_step": 154774, "epoch": 1739, "lr": 3.9187059679048524e-05}
{"train_loss": 0.14060494303703308, "global_step": 154775, "epoch": 1739, "lr": 3.91864936620163e-05}
{"train_loss": 0.11765443533658981, "global_step": 154776, "epoch": 1739, "lr": 3.9185927646437853e-05}
{"train_loss": 0.07039448618888855, "global_step": 154777, "epoch": 1739, "lr": 3.91853616323132e-05}
{"train_loss": 0.0852210596203804, "global_step": 154778, "epoch": 1739, "lr": 3.9184795619642464e-05}
{"train_loss": 0.08167310059070587, "global_step": 154779, "epoch": 1739, "lr": 3.918422960842568e-05}
{"train_loss": 0.12885278463363647, "global_step": 154780, "epoch": 1739, "lr": 3.9183663598662966e-05}
{"train_loss": 0.06292378157377243, "global_step": 154781, "epoch": 1739, "lr": 3.918309759035436e-05}
{"train_loss": 0.10202658921480179, "global_step": 154782, "epoch": 1739, "lr": 3.918253158349997e-05}
{"train_loss": 0.07691903412342072, "global_step": 154783, "epoch": 1739, "lr": 3.9181965578099834e-05}
{"train_loss": 0.056097373366355896, "global_step": 154784, "epoch": 1739, "lr": 3.9181399574154075e-05}
{"train_loss": 0.08225534111261368, "global_step": 154785, "epoch": 1739, "lr": 3.9180833571662725e-05}
{"train_loss": 0.10662709176540375, "global_step": 154786, "epoch": 1739, "lr": 3.9180267570625894e-05}
{"train_loss": 0.1217525452375412, "global_step": 154787, "epoch": 1739, "lr": 3.917970157104363e-05}
{"train_loss": 0.11680575460195541, "global_step": 154788, "epoch": 1739, "lr": 3.9179135572916045e-05}
{"train_loss": 0.10873472690582275, "global_step": 154789, "epoch": 1739, "lr": 3.917856957624317e-05}
{"train_loss": 0.05081801488995552, "global_step": 154790, "epoch": 1739, "lr": 3.9178003581025115e-05}
{"train_loss": 0.11246170103549957, "global_step": 154791, "epoch": 1739, "lr": 3.9177437587261945e-05}
{"train_loss": 0.10426129400730133, "global_step": 154792, "epoch": 1739, "lr": 3.9176871594953735e-05}
{"train_loss": 0.10725873708724976, "global_step": 154793, "epoch": 1739, "lr": 3.9176305604100564e-05}
{"train_loss": 0.09892509132623672, "global_step": 154794, "epoch": 1739, "lr": 3.9175739614702515e-05}
{"train_loss": 0.048810817301273346, "global_step": 154795, "epoch": 1739, "lr": 3.9175173626759654e-05}
{"train_loss": 0.06895778328180313, "global_step": 154796, "epoch": 1739, "lr": 3.9174607640272044e-05}
{"train_loss": 0.16569602489471436, "global_step": 154797, "epoch": 1739, "lr": 3.91740416552398e-05}
{"train_loss": 0.09478770941495895, "global_step": 154798, "epoch": 1739, "lr": 3.917347567166294e-05}
{"train_loss": 0.088480643928051, "global_step": 154799, "epoch": 1739, "lr": 3.917290968954161e-05}
{"train_loss": 0.07531803846359253, "global_step": 154800, "epoch": 1739, "lr": 3.917234370887582e-05}
{"train_loss": 0.09394949674606323, "global_step": 154801, "epoch": 1739, "lr": 3.9171777729665704e-05}
{"train_loss": 0.07207772135734558, "global_step": 154802, "epoch": 1739, "lr": 3.917121175191129e-05}
{"train_loss": 0.10178282856941223, "global_step": 154803, "epoch": 1739, "lr": 3.917064577561269e-05}
{"train_loss": 0.14153720438480377, "global_step": 154804, "epoch": 1739, "lr": 3.917007980076994e-05}
{"train_loss": 0.0488215871155262, "global_step": 154805, "epoch": 1739, "lr": 3.916951382738317e-05}
{"train_loss": 0.08564195781946182, "global_step": 154806, "epoch": 1739, "lr": 3.9168947855452406e-05}
{"train_loss": 0.07558926939964294, "global_step": 154807, "epoch": 1739, "lr": 3.916838188497774e-05}
{"train_loss": 0.0825299471616745, "global_step": 154808, "epoch": 1739, "lr": 3.916781591595928e-05}
{"train_loss": 0.09599783271551132, "global_step": 154809, "epoch": 1739, "lr": 3.9167249948397056e-05}
{"train_loss": 0.06967242062091827, "global_step": 154810, "epoch": 1739, "lr": 3.916668398229117e-05}
{"train_loss": 0.06910156458616257, "global_step": 154811, "epoch": 1739, "lr": 3.916611801764169e-05}
{"train_loss": 0.12693679332733154, "global_step": 154812, "epoch": 1739, "lr": 3.916555205444869e-05}
{"train_loss": 0.08108994364738464, "global_step": 154813, "epoch": 1739, "lr": 3.916498609271226e-05}
{"train_loss": 0.04868641123175621, "global_step": 154814, "epoch": 1739, "lr": 3.9164420132432465e-05}
{"train_loss": 0.1426670104265213, "global_step": 154815, "epoch": 1739, "lr": 3.9163854173609364e-05}
{"train_loss": 0.08466459810733795, "global_step": 154816, "epoch": 1739, "lr": 3.916328821624307e-05}
{"train_loss": 0.06407472491264343, "global_step": 154817, "epoch": 1739, "lr": 3.916272226033362e-05}
{"train_loss": 0.09980642050504684, "global_step": 154818, "epoch": 1739, "lr": 3.9162156305881136e-05}
{"train_loss": 0.11782960593700409, "global_step": 154819, "epoch": 1739, "lr": 3.916159035288564e-05}
{"train_loss": 0.07830827683210373, "global_step": 154820, "epoch": 1739, "lr": 3.916102440134726e-05}
{"train_loss": 0.11617013812065125, "global_step": 154821, "epoch": 1739, "lr": 3.916045845126603e-05}
{"train_loss": 0.09057025611400604, "global_step": 154822, "epoch": 1739, "lr": 3.915989250264206e-05}
{"train_loss": 0.042587414383888245, "global_step": 154823, "epoch": 1739, "lr": 3.91593265554754e-05}
{"train_loss": 0.0772857591509819, "global_step": 154824, "epoch": 1739, "lr": 3.915876060976614e-05}
{"train_loss": 0.09339576214551926, "global_step": 154825, "epoch": 1739, "lr": 3.915819466551434e-05}
{"train_loss": 0.08071532845497131, "global_step": 154826, "epoch": 1739, "lr": 3.915762872272009e-05}
{"train_loss": 0.054664041846990585, "global_step": 154827, "epoch": 1739, "lr": 3.9157062781383483e-05}
{"train_loss": 0.06843718141317368, "global_step": 154828, "epoch": 1739, "lr": 3.915649684150457e-05}
{"train_loss": 0.13072195649147034, "global_step": 154829, "epoch": 1739, "lr": 3.915593090308343e-05}
{"train_loss": 0.07590804249048233, "global_step": 154830, "epoch": 1739, "lr": 3.9155364966120145e-05}
{"train_loss": 0.08217711746692657, "global_step": 154831, "epoch": 1739, "lr": 3.915479903061479e-05}
{"train_loss": 0.1517256498336792, "global_step": 154832, "epoch": 1739, "lr": 3.915423309656744e-05}
{"train_loss": 0.09049151837825775, "global_step": 154833, "epoch": 1739, "lr": 3.915366716397818e-05}
{"train_loss": 0.1109098345041275, "global_step": 154834, "epoch": 1739, "lr": 3.915310123284706e-05}
{"train_loss": 0.1298336535692215, "global_step": 154835, "epoch": 1739, "lr": 3.915253530317419e-05}
{"train_loss": 0.10375239700078964, "global_step": 154836, "epoch": 1739, "lr": 3.915196937495961e-05}
{"train_loss": 0.15340179204940796, "global_step": 154837, "epoch": 1739, "lr": 3.915140344820343e-05}
{"train_loss": 0.1790958046913147, "global_step": 154838, "epoch": 1739, "lr": 3.9150837522905706e-05}
{"train_loss": 0.16222840547561646, "global_step": 154839, "epoch": 1739, "lr": 3.915027159906652e-05}
{"train_loss": 0.11212725937366486, "global_step": 154840, "epoch": 1739, "lr": 3.9149705676685944e-05}
{"train_loss": 0.051600612699985504, "global_step": 154841, "epoch": 1739, "lr": 3.914913975576408e-05}
{"train_loss": 0.08481568843126297, "global_step": 154842, "epoch": 1739, "lr": 3.9148573836300944e-05}
{"train_loss": 0.10365274548530579, "global_step": 154843, "epoch": 1739, "lr": 3.914800791829666e-05}
{"train_loss": 0.15857021510601044, "global_step": 154844, "epoch": 1739, "lr": 3.914744200175132e-05}
{"train_loss": 0.06773465126752853, "global_step": 154845, "epoch": 1739, "lr": 3.914687608666494e-05}
{"train_loss": 0.06352780014276505, "global_step": 154846, "epoch": 1739, "lr": 3.9146310173037666e-05}
{"train_loss": 0.10387600213289261, "global_step": 154847, "epoch": 1739, "lr": 3.914574426086952e-05}
{"train_loss": 0.08968807011842728, "global_step": 154848, "epoch": 1739, "lr": 3.9145178350160594e-05}
{"train_loss": 0.07778768986463547, "global_step": 154849, "epoch": 1739, "lr": 3.914461244091097e-05}
{"train_loss": 0.047171249985694885, "global_step": 154850, "epoch": 1739, "lr": 3.9144046533120723e-05}
{"train_loss": 0.07517150044441223, "global_step": 154851, "epoch": 1739, "lr": 3.914348062678992e-05}
{"train_loss": 0.14516551792621613, "global_step": 154852, "epoch": 1739, "lr": 3.914291472191866e-05}
{"train_loss": 0.13688552379608154, "global_step": 154853, "epoch": 1739, "lr": 3.914234881850698e-05}
{"train_loss": 0.07197155803442001, "global_step": 154854, "epoch": 1739, "lr": 3.9141782916555e-05}
{"train_loss": 0.06081680953502655, "global_step": 154855, "epoch": 1739, "lr": 3.914121701606275e-05}
{"train_loss": 0.11252933740615845, "global_step": 154856, "epoch": 1739, "lr": 3.914065111703036e-05}
{"train_loss": 0.14541207253932953, "global_step": 154857, "epoch": 1739, "lr": 3.914008521945785e-05}
{"train_loss": 0.0642104521393776, "global_step": 154858, "epoch": 1739, "lr": 3.9139519323345346e-05}
{"train_loss": 0.0965247443050481, "global_step": 154859, "epoch": 1739, "lr": 3.913895342869288e-05, "val_loss": 6.517917156219482}
{"train_loss": 0.11556665599346161, "global_step": 154860, "epoch": 1740, "lr": 3.9138387535500574e-05}
{"train_loss": 0.12067417055368423, "global_step": 154861, "epoch": 1740, "lr": 3.913782164376845e-05}
{"train_loss": 0.05269278958439827, "global_step": 154862, "epoch": 1740, "lr": 3.913725575349662e-05}
{"train_loss": 0.102867491543293, "global_step": 154863, "epoch": 1740, "lr": 3.913668986468518e-05}
{"train_loss": 0.10537263005971909, "global_step": 154864, "epoch": 1740, "lr": 3.913612397733415e-05}
{"train_loss": 0.08918934315443039, "global_step": 154865, "epoch": 1740, "lr": 3.913555809144366e-05}
{"train_loss": 0.06807418912649155, "global_step": 154866, "epoch": 1740, "lr": 3.913499220701374e-05}
{"train_loss": 0.13740219175815582, "global_step": 154867, "epoch": 1740, "lr": 3.91344263240445e-05}
{"train_loss": 0.045742250978946686, "global_step": 154868, "epoch": 1740, "lr": 3.913386044253601e-05}
{"train_loss": 0.14614248275756836, "global_step": 154869, "epoch": 1740, "lr": 3.9133294562488326e-05}
{"train_loss": 0.11236228793859482, "global_step": 154870, "epoch": 1740, "lr": 3.9132728683901546e-05}
{"train_loss": 0.10104671865701675, "global_step": 154871, "epoch": 1740, "lr": 3.913216280677575e-05}
{"train_loss": 0.08434788137674332, "global_step": 154872, "epoch": 1740, "lr": 3.913159693111097e-05}
{"train_loss": 0.0769447460770607, "global_step": 154873, "epoch": 1740, "lr": 3.9131031056907346e-05}
{"train_loss": 0.056964438408613205, "global_step": 154874, "epoch": 1740, "lr": 3.91304651841649e-05}
{"train_loss": 0.20696507394313812, "global_step": 154875, "epoch": 1740, "lr": 3.912989931288375e-05}
{"train_loss": 0.09526177495718002, "global_step": 154876, "epoch": 1740, "lr": 3.912933344306393e-05}
{"train_loss": 0.0800112932920456, "global_step": 154877, "epoch": 1740, "lr": 3.912876757470556e-05}
{"train_loss": 0.09571690112352371, "global_step": 154878, "epoch": 1740, "lr": 3.9128201707808675e-05}
{"train_loss": 0.0907764732837677, "global_step": 154879, "epoch": 1740, "lr": 3.912763584237339e-05}
{"train_loss": 0.1270311176776886, "global_step": 154880, "epoch": 1740, "lr": 3.912706997839974e-05}
{"train_loss": 0.09537936002016068, "global_step": 154881, "epoch": 1740, "lr": 3.912650411588783e-05}
{"train_loss": 0.08440398424863815, "global_step": 154882, "epoch": 1740, "lr": 3.9125938254837745e-05}
{"train_loss": 0.16786406934261322, "global_step": 154883, "epoch": 1740, "lr": 3.912537239524952e-05}
{"train_loss": 0.12483612447977066, "global_step": 154884, "epoch": 1740, "lr": 3.912480653712328e-05}
{"train_loss": 0.11283442378044128, "global_step": 154885, "epoch": 1740, "lr": 3.9124240680459077e-05}
{"train_loss": 0.06882001459598541, "global_step": 154886, "epoch": 1740, "lr": 3.912367482525697e-05}
{"train_loss": 0.07065082341432571, "global_step": 154887, "epoch": 1740, "lr": 3.912310897151706e-05}
{"train_loss": 0.06162956357002258, "global_step": 154888, "epoch": 1740, "lr": 3.912254311923942e-05}
{"train_loss": 0.14652302861213684, "global_step": 154889, "epoch": 1740, "lr": 3.912197726842412e-05}
{"train_loss": 0.05233060568571091, "global_step": 154890, "epoch": 1740, "lr": 3.912141141907125e-05}
{"train_loss": 0.07499773055315018, "global_step": 154891, "epoch": 1740, "lr": 3.912084557118084e-05}
{"train_loss": 0.12985718250274658, "global_step": 154892, "epoch": 1740, "lr": 3.9120279724753036e-05}
{"train_loss": 0.14385667443275452, "global_step": 154893, "epoch": 1740, "lr": 3.9119713879787847e-05}
{"train_loss": 0.11572878062725067, "global_step": 154894, "epoch": 1740, "lr": 3.91191480362854e-05}
{"train_loss": 0.07599499076604843, "global_step": 154895, "epoch": 1740, "lr": 3.9118582194245735e-05}
{"train_loss": 0.09124299138784409, "global_step": 154896, "epoch": 1740, "lr": 3.911801635366897e-05}
{"train_loss": 0.10164418816566467, "global_step": 154897, "epoch": 1740, "lr": 3.9117450514555124e-05}
{"train_loss": 0.09150831401348114, "global_step": 154898, "epoch": 1740, "lr": 3.911688467690431e-05}
{"train_loss": 0.0809079110622406, "global_step": 154899, "epoch": 1740, "lr": 3.9116318840716615e-05}
{"train_loss": 0.10261174291372299, "global_step": 154900, "epoch": 1740, "lr": 3.911575300599208e-05}
{"train_loss": 0.08567535132169724, "global_step": 154901, "epoch": 1740, "lr": 3.911518717273081e-05}
{"train_loss": 0.07921551167964935, "global_step": 154902, "epoch": 1740, "lr": 3.9114621340932866e-05}
{"train_loss": 0.06487150490283966, "global_step": 154903, "epoch": 1740, "lr": 3.911405551059835e-05}
{"train_loss": 0.09369182586669922, "global_step": 154904, "epoch": 1740, "lr": 3.911348968172729e-05}
{"train_loss": 0.09992754459381104, "global_step": 154905, "epoch": 1740, "lr": 3.91129238543198e-05}
{"train_loss": 0.1048276424407959, "global_step": 154906, "epoch": 1740, "lr": 3.911235802837594e-05}
{"train_loss": 0.11128561943769455, "global_step": 154907, "epoch": 1740, "lr": 3.91117922038958e-05}
{"train_loss": 0.17859286069869995, "global_step": 154908, "epoch": 1740, "lr": 3.9111226380879434e-05}
{"train_loss": 0.07664024829864502, "global_step": 154909, "epoch": 1740, "lr": 3.911066055932694e-05}
{"train_loss": 0.10357550531625748, "global_step": 154910, "epoch": 1740, "lr": 3.9110094739238376e-05}
{"train_loss": 0.10095218569040298, "global_step": 154911, "epoch": 1740, "lr": 3.9109528920613844e-05}
{"train_loss": 0.0532374382019043, "global_step": 154912, "epoch": 1740, "lr": 3.9108963103453385e-05}
{"train_loss": 0.08976853638887405, "global_step": 154913, "epoch": 1740, "lr": 3.910839728775711e-05}
{"train_loss": 0.0811530277132988, "global_step": 154914, "epoch": 1740, "lr": 3.910783147352506e-05}
{"train_loss": 0.09189657866954803, "global_step": 154915, "epoch": 1740, "lr": 3.910726566075734e-05}
{"train_loss": 0.1232520043849945, "global_step": 154916, "epoch": 1740, "lr": 3.910669984945401e-05}
{"train_loss": 0.08821696788072586, "global_step": 154917, "epoch": 1740, "lr": 3.9106134039615136e-05}
{"train_loss": 0.07441749423742294, "global_step": 154918, "epoch": 1740, "lr": 3.910556823124084e-05}
{"train_loss": 0.11738307029008865, "global_step": 154919, "epoch": 1740, "lr": 3.910500242433115e-05}
{"train_loss": 0.15026643872261047, "global_step": 154920, "epoch": 1740, "lr": 3.910443661888618e-05}
{"train_loss": 0.13428224623203278, "global_step": 154921, "epoch": 1740, "lr": 3.910387081490595e-05}
{"train_loss": 0.13033859431743622, "global_step": 154922, "epoch": 1740, "lr": 3.9103305012390604e-05}
{"train_loss": 0.13486939668655396, "global_step": 154923, "epoch": 1740, "lr": 3.910273921134017e-05}
{"train_loss": 0.10767385363578796, "global_step": 154924, "epoch": 1740, "lr": 3.910217341175475e-05}
{"train_loss": 0.17698590457439423, "global_step": 154925, "epoch": 1740, "lr": 3.9101607613634405e-05}
{"train_loss": 0.09546291828155518, "global_step": 154926, "epoch": 1740, "lr": 3.910104181697921e-05}
{"train_loss": 0.15725161135196686, "global_step": 154927, "epoch": 1740, "lr": 3.9100476021789256e-05}
{"train_loss": 0.06890466064214706, "global_step": 154928, "epoch": 1740, "lr": 3.9099910228064606e-05}
{"train_loss": 0.07920844852924347, "global_step": 154929, "epoch": 1740, "lr": 3.909934443580534e-05}
{"train_loss": 0.07456004619598389, "global_step": 154930, "epoch": 1740, "lr": 3.909877864501155e-05}
{"train_loss": 0.0735493153333664, "global_step": 154931, "epoch": 1740, "lr": 3.9098212855683265e-05}
{"train_loss": 0.14962603151798248, "global_step": 154932, "epoch": 1740, "lr": 3.909764706782062e-05}
{"train_loss": 0.14426563680171967, "global_step": 154933, "epoch": 1740, "lr": 3.9097081281423644e-05}
{"train_loss": 0.11188386380672455, "global_step": 154934, "epoch": 1740, "lr": 3.909651549649242e-05}
{"train_loss": 0.0608072429895401, "global_step": 154935, "epoch": 1740, "lr": 3.909594971302708e-05}
{"train_loss": 0.06675945222377777, "global_step": 154936, "epoch": 1740, "lr": 3.909538393102762e-05}
{"train_loss": 0.07212604582309723, "global_step": 154937, "epoch": 1740, "lr": 3.909481815049417e-05}
{"train_loss": 0.08836902678012848, "global_step": 154938, "epoch": 1740, "lr": 3.9094252371426785e-05}
{"train_loss": 0.09713315218687057, "global_step": 154939, "epoch": 1740, "lr": 3.909368659382555e-05}
{"train_loss": 0.09909254312515259, "global_step": 154940, "epoch": 1740, "lr": 3.909312081769052e-05}
{"train_loss": 0.12366136908531189, "global_step": 154941, "epoch": 1740, "lr": 3.909255504302181e-05}
{"train_loss": 0.10137015581130981, "global_step": 154942, "epoch": 1740, "lr": 3.909198926981945e-05}
{"train_loss": 0.09157925844192505, "global_step": 154943, "epoch": 1740, "lr": 3.909142349808356e-05}
{"train_loss": 0.06207055225968361, "global_step": 154944, "epoch": 1740, "lr": 3.909085772781418e-05}
{"train_loss": 0.07786254584789276, "global_step": 154945, "epoch": 1740, "lr": 3.909029195901142e-05}
{"train_loss": 0.12685084342956543, "global_step": 154946, "epoch": 1740, "lr": 3.908972619167531e-05}
{"train_loss": 0.11285851895809174, "global_step": 154947, "epoch": 1740, "lr": 3.908916042580597e-05}
{"train_loss": 0.10098856775445884, "global_step": 154948, "epoch": 1740, "lr": 3.908859466140346e-05, "val_loss": 6.5301971435546875, "train_action_mse_error": 23.93671989440918}
{"train_loss": 0.07322624325752258, "global_step": 154949, "epoch": 1741, "lr": 3.908802889846786e-05}
{"train_loss": 0.07854251563549042, "global_step": 154950, "epoch": 1741, "lr": 3.9087463136999224e-05}
{"train_loss": 0.10275167226791382, "global_step": 154951, "epoch": 1741, "lr": 3.908689737699767e-05}
{"train_loss": 0.128248393535614, "global_step": 154952, "epoch": 1741, "lr": 3.9086331618463224e-05}
{"train_loss": 0.1684388816356659, "global_step": 154953, "epoch": 1741, "lr": 3.9085765861395985e-05}
{"train_loss": 0.17985498905181885, "global_step": 154954, "epoch": 1741, "lr": 3.908520010579606e-05}
{"train_loss": 0.08407916128635406, "global_step": 154955, "epoch": 1741, "lr": 3.908463435166347e-05}
{"train_loss": 0.11109419912099838, "global_step": 154956, "epoch": 1741, "lr": 3.9084068598998336e-05}
{"train_loss": 0.07558923214673996, "global_step": 154957, "epoch": 1741, "lr": 3.90835028478007e-05}
{"train_loss": 0.11009034514427185, "global_step": 154958, "epoch": 1741, "lr": 3.908293709807067e-05}
{"train_loss": 0.1045856699347496, "global_step": 154959, "epoch": 1741, "lr": 3.9082371349808294e-05}
{"train_loss": 0.05768619105219841, "global_step": 154960, "epoch": 1741, "lr": 3.9081805603013675e-05}
{"train_loss": 0.1300320327281952, "global_step": 154961, "epoch": 1741, "lr": 3.908123985768686e-05}
{"train_loss": 0.06262970715761185, "global_step": 154962, "epoch": 1741, "lr": 3.9080674113827956e-05}
{"train_loss": 0.07694221287965775, "global_step": 154963, "epoch": 1741, "lr": 3.9080108371437006e-05}
{"train_loss": 0.10682765394449234, "global_step": 154964, "epoch": 1741, "lr": 3.9079542630514106e-05}
{"train_loss": 0.09776785224676132, "global_step": 154965, "epoch": 1741, "lr": 3.9078976891059325e-05}
{"train_loss": 0.08820796757936478, "global_step": 154966, "epoch": 1741, "lr": 3.907841115307275e-05}
{"train_loss": 0.09709128737449646, "global_step": 154967, "epoch": 1741, "lr": 3.907784541655445e-05}
{"train_loss": 0.13516660034656525, "global_step": 154968, "epoch": 1741, "lr": 3.9077279681504504e-05}
{"train_loss": 0.08711493760347366, "global_step": 154969, "epoch": 1741, "lr": 3.907671394792297e-05}
{"train_loss": 0.09581049531698227, "global_step": 154970, "epoch": 1741, "lr": 3.9076148215809936e-05}
{"train_loss": 0.06745133548974991, "global_step": 154971, "epoch": 1741, "lr": 3.9075582485165496e-05}
{"train_loss": 0.1047421246767044, "global_step": 154972, "epoch": 1741, "lr": 3.90750167559897e-05}
{"train_loss": 0.10509724915027618, "global_step": 154973, "epoch": 1741, "lr": 3.9074451028282646e-05}
{"train_loss": 0.09091851860284805, "global_step": 154974, "epoch": 1741, "lr": 3.9073885302044384e-05}
{"train_loss": 0.20925705134868622, "global_step": 154975, "epoch": 1741, "lr": 3.907331957727503e-05}
{"train_loss": 0.11119598150253296, "global_step": 154976, "epoch": 1741, "lr": 3.90727538539746e-05}
{"train_loss": 0.14218084514141083, "global_step": 154977, "epoch": 1741, "lr": 3.907218813214324e-05}
{"train_loss": 0.12176378071308136, "global_step": 154978, "epoch": 1741, "lr": 3.907162241178096e-05}
{"train_loss": 0.09746041893959045, "global_step": 154979, "epoch": 1741, "lr": 3.90710566928879e-05}
{"train_loss": 0.08417853713035583, "global_step": 154980, "epoch": 1741, "lr": 3.907049097546407e-05}
{"train_loss": 0.10762003809213638, "global_step": 154981, "epoch": 1741, "lr": 3.906992525950961e-05}
{"train_loss": 0.09059164673089981, "global_step": 154982, "epoch": 1741, "lr": 3.9069359545024546e-05}
{"train_loss": 0.1703425496816635, "global_step": 154983, "epoch": 1741, "lr": 3.9068793832008984e-05}
{"train_loss": 0.12141011655330658, "global_step": 154984, "epoch": 1741, "lr": 3.906822812046298e-05}
{"train_loss": 0.07066626101732254, "global_step": 154985, "epoch": 1741, "lr": 3.9067662410386624e-05}
{"train_loss": 0.09710714221000671, "global_step": 154986, "epoch": 1741, "lr": 3.906709670177999e-05}
{"train_loss": 0.11353489011526108, "global_step": 154987, "epoch": 1741, "lr": 3.9066530994643154e-05}
{"train_loss": 0.06091350317001343, "global_step": 154988, "epoch": 1741, "lr": 3.906596528897617e-05}
{"train_loss": 0.09073010832071304, "global_step": 154989, "epoch": 1741, "lr": 3.906539958477914e-05}
{"train_loss": 0.08928178995847702, "global_step": 154990, "epoch": 1741, "lr": 3.906483388205215e-05}
{"train_loss": 0.06600417196750641, "global_step": 154991, "epoch": 1741, "lr": 3.9064268180795236e-05}
{"train_loss": 0.08725782483816147, "global_step": 154992, "epoch": 1741, "lr": 3.906370248100853e-05}
{"train_loss": 0.08429212868213654, "global_step": 154993, "epoch": 1741, "lr": 3.906313678269205e-05}
{"train_loss": 0.07875888794660568, "global_step": 154994, "epoch": 1741, "lr": 3.906257108584591e-05}
{"train_loss": 0.11448005586862564, "global_step": 154995, "epoch": 1741, "lr": 3.906200539047016e-05}
{"train_loss": 0.12578131258487701, "global_step": 154996, "epoch": 1741, "lr": 3.9061439696564915e-05}
{"train_loss": 0.09142196923494339, "global_step": 154997, "epoch": 1741, "lr": 3.90608740041302e-05}
{"train_loss": 0.11114152520895004, "global_step": 154998, "epoch": 1741, "lr": 3.906030831316614e-05}
{"train_loss": 0.08226395398378372, "global_step": 154999, "epoch": 1741, "lr": 3.905974262367277e-05}
{"train_loss": 0.08056595921516418, "global_step": 155000, "epoch": 1741, "lr": 3.90591769356502e-05}
{"train_loss": 0.07496335357427597, "global_step": 155001, "epoch": 1741, "lr": 3.905861124909848e-05}
{"train_loss": 0.08095595985651016, "global_step": 155002, "epoch": 1741, "lr": 3.90580455640177e-05}
{"train_loss": 0.13609519600868225, "global_step": 155003, "epoch": 1741, "lr": 3.905747988040793e-05}
{"train_loss": 0.11661442369222641, "global_step": 155004, "epoch": 1741, "lr": 3.9056914198269244e-05}
{"train_loss": 0.09025174379348755, "global_step": 155005, "epoch": 1741, "lr": 3.9056348517601734e-05}
{"train_loss": 0.0461253896355629, "global_step": 155006, "epoch": 1741, "lr": 3.9055782838405467e-05}
{"train_loss": 0.0358203649520874, "global_step": 155007, "epoch": 1741, "lr": 3.9055217160680495e-05}
{"train_loss": 0.1368892937898636, "global_step": 155008, "epoch": 1741, "lr": 3.905465148442692e-05}
{"train_loss": 0.10390455275774002, "global_step": 155009, "epoch": 1741, "lr": 3.905408580964483e-05}
{"train_loss": 0.08471324294805527, "global_step": 155010, "epoch": 1741, "lr": 3.9053520136334265e-05}
{"train_loss": 0.08748754113912582, "global_step": 155011, "epoch": 1741, "lr": 3.9052954464495346e-05}
{"train_loss": 0.10211705416440964, "global_step": 155012, "epoch": 1741, "lr": 3.9052388794128096e-05}
{"train_loss": 0.08414208143949509, "global_step": 155013, "epoch": 1741, "lr": 3.905182312523264e-05}
{"train_loss": 0.05819614604115486, "global_step": 155014, "epoch": 1741, "lr": 3.905125745780902e-05}
{"train_loss": 0.13800516724586487, "global_step": 155015, "epoch": 1741, "lr": 3.905069179185734e-05}
{"train_loss": 0.08909191936254501, "global_step": 155016, "epoch": 1741, "lr": 3.9050126127377647e-05}
{"train_loss": 0.09949161112308502, "global_step": 155017, "epoch": 1741, "lr": 3.904956046437004e-05}
{"train_loss": 0.21347741782665253, "global_step": 155018, "epoch": 1741, "lr": 3.9048994802834574e-05}
{"train_loss": 0.08387336134910583, "global_step": 155019, "epoch": 1741, "lr": 3.904842914277136e-05}
{"train_loss": 0.05078783631324768, "global_step": 155020, "epoch": 1741, "lr": 3.9047863484180426e-05}
{"train_loss": 0.14594604074954987, "global_step": 155021, "epoch": 1741, "lr": 3.904729782706189e-05}
{"train_loss": 0.13968871533870697, "global_step": 155022, "epoch": 1741, "lr": 3.9046732171415814e-05}
{"train_loss": 0.08455356955528259, "global_step": 155023, "epoch": 1741, "lr": 3.9046166517242254e-05}
{"train_loss": 0.12652364373207092, "global_step": 155024, "epoch": 1741, "lr": 3.904560086454131e-05}
{"train_loss": 0.08439403772354126, "global_step": 155025, "epoch": 1741, "lr": 3.904503521331305e-05}
{"train_loss": 0.10932980477809906, "global_step": 155026, "epoch": 1741, "lr": 3.904446956355756e-05}
{"train_loss": 0.09810327738523483, "global_step": 155027, "epoch": 1741, "lr": 3.904390391527489e-05}
{"train_loss": 0.10815487802028656, "global_step": 155028, "epoch": 1741, "lr": 3.9043338268465165e-05}
{"train_loss": 0.13507893681526184, "global_step": 155029, "epoch": 1741, "lr": 3.9042772623128395e-05}
{"train_loss": 0.12795045971870422, "global_step": 155030, "epoch": 1741, "lr": 3.904220697926472e-05}
{"train_loss": 0.09940013289451599, "global_step": 155031, "epoch": 1741, "lr": 3.9041641336874166e-05}
{"train_loss": 0.09110943973064423, "global_step": 155032, "epoch": 1741, "lr": 3.904107569595684e-05}
{"train_loss": 0.08136053383350372, "global_step": 155033, "epoch": 1741, "lr": 3.90405100565128e-05}
{"train_loss": 0.05231354013085365, "global_step": 155034, "epoch": 1741, "lr": 3.903994441854214e-05}
{"train_loss": 0.16811175644397736, "global_step": 155035, "epoch": 1741, "lr": 3.903937878204491e-05}
{"train_loss": 0.1031232699751854, "global_step": 155036, "epoch": 1741, "lr": 3.903881314702122e-05}
{"train_loss": 0.10252617822771662, "global_step": 155037, "epoch": 1741, "lr": 3.903824751347112e-05, "val_loss": 6.428653717041016}
{"train_loss": 0.13988922536373138, "global_step": 155038, "epoch": 1742, "lr": 3.90376818813947e-05}
{"train_loss": 0.17103180289268494, "global_step": 155039, "epoch": 1742, "lr": 3.903711625079202e-05}
{"train_loss": 0.0989353358745575, "global_step": 155040, "epoch": 1742, "lr": 3.903655062166318e-05}
{"train_loss": 0.14759670197963715, "global_step": 155041, "epoch": 1742, "lr": 3.903598499400823e-05}
{"train_loss": 0.03387192264199257, "global_step": 155042, "epoch": 1742, "lr": 3.903541936782726e-05}
{"train_loss": 0.10703422874212265, "global_step": 155043, "epoch": 1742, "lr": 3.903485374312035e-05}
{"train_loss": 0.126081645488739, "global_step": 155044, "epoch": 1742, "lr": 3.903428811988755e-05}
{"train_loss": 0.10479671508073807, "global_step": 155045, "epoch": 1742, "lr": 3.903372249812898e-05}
{"train_loss": 0.06244682893157005, "global_step": 155046, "epoch": 1742, "lr": 3.9033156877844676e-05}
{"train_loss": 0.11268661171197891, "global_step": 155047, "epoch": 1742, "lr": 3.903259125903474e-05}
{"train_loss": 0.09009645134210587, "global_step": 155048, "epoch": 1742, "lr": 3.9032025641699235e-05}
{"train_loss": 0.0766211599111557, "global_step": 155049, "epoch": 1742, "lr": 3.903146002583825e-05}
{"train_loss": 0.11910506337881088, "global_step": 155050, "epoch": 1742, "lr": 3.9030894411451834e-05}
{"train_loss": 0.05714523419737816, "global_step": 155051, "epoch": 1742, "lr": 3.90303287985401e-05}
{"train_loss": 0.07330195605754852, "global_step": 155052, "epoch": 1742, "lr": 3.902976318710309e-05}
{"train_loss": 0.11290249228477478, "global_step": 155053, "epoch": 1742, "lr": 3.902919757714091e-05}
{"train_loss": 0.15027464926242828, "global_step": 155054, "epoch": 1742, "lr": 3.9028631968653596e-05}
{"train_loss": 0.11013992130756378, "global_step": 155055, "epoch": 1742, "lr": 3.902806636164127e-05}
{"train_loss": 0.1251748502254486, "global_step": 155056, "epoch": 1742, "lr": 3.902750075610397e-05}
{"train_loss": 0.09966804832220078, "global_step": 155057, "epoch": 1742, "lr": 3.902693515204181e-05}
{"train_loss": 0.14306607842445374, "global_step": 155058, "epoch": 1742, "lr": 3.902636954945482e-05}
{"train_loss": 0.12889260053634644, "global_step": 155059, "epoch": 1742, "lr": 3.902580394834313e-05}
{"train_loss": 0.09748002141714096, "global_step": 155060, "epoch": 1742, "lr": 3.902523834870677e-05}
{"train_loss": 0.09395736455917358, "global_step": 155061, "epoch": 1742, "lr": 3.902467275054583e-05}
{"train_loss": 0.1368805170059204, "global_step": 155062, "epoch": 1742, "lr": 3.90241071538604e-05}
{"train_loss": 0.16256313025951385, "global_step": 155063, "epoch": 1742, "lr": 3.902354155865052e-05}
{"train_loss": 0.06415119767189026, "global_step": 155064, "epoch": 1742, "lr": 3.9022975964916326e-05}
{"train_loss": 0.15615439414978027, "global_step": 155065, "epoch": 1742, "lr": 3.902241037265783e-05}
{"train_loss": 0.07960119843482971, "global_step": 155066, "epoch": 1742, "lr": 3.902184478187516e-05}
{"train_loss": 0.065967857837677, "global_step": 155067, "epoch": 1742, "lr": 3.902127919256835e-05}
{"train_loss": 0.10299904644489288, "global_step": 155068, "epoch": 1742, "lr": 3.902071360473751e-05}
{"train_loss": 0.1046910285949707, "global_step": 155069, "epoch": 1742, "lr": 3.902014801838268e-05}
{"train_loss": 0.14105075597763062, "global_step": 155070, "epoch": 1742, "lr": 3.901958243350399e-05}
{"train_loss": 0.10298866033554077, "global_step": 155071, "epoch": 1742, "lr": 3.9019016850101454e-05}
{"train_loss": 0.07270655781030655, "global_step": 155072, "epoch": 1742, "lr": 3.90184512681752e-05}
{"train_loss": 0.10710225999355316, "global_step": 155073, "epoch": 1742, "lr": 3.901788568772526e-05}
{"train_loss": 0.11458028852939606, "global_step": 155074, "epoch": 1742, "lr": 3.901732010875175e-05}
{"train_loss": 0.11953799426555634, "global_step": 155075, "epoch": 1742, "lr": 3.901675453125471e-05}
{"train_loss": 0.09033122658729553, "global_step": 155076, "epoch": 1742, "lr": 3.901618895523426e-05}
{"train_loss": 0.07269720733165741, "global_step": 155077, "epoch": 1742, "lr": 3.9015623380690416e-05}
{"train_loss": 0.08245328068733215, "global_step": 155078, "epoch": 1742, "lr": 3.901505780762331e-05}
{"train_loss": 0.07499965280294418, "global_step": 155079, "epoch": 1742, "lr": 3.9014492236033e-05}
{"train_loss": 0.08232969045639038, "global_step": 155080, "epoch": 1742, "lr": 3.901392666591953e-05}
{"train_loss": 0.1272372305393219, "global_step": 155081, "epoch": 1742, "lr": 3.901336109728303e-05}
{"train_loss": 0.16736651957035065, "global_step": 155082, "epoch": 1742, "lr": 3.901279553012353e-05}
{"train_loss": 0.07109487801790237, "global_step": 155083, "epoch": 1742, "lr": 3.901222996444115e-05}
{"train_loss": 0.10965076833963394, "global_step": 155084, "epoch": 1742, "lr": 3.9011664400235915e-05}
{"train_loss": 0.14404043555259705, "global_step": 155085, "epoch": 1742, "lr": 3.901109883750795e-05}
{"train_loss": 0.11474417895078659, "global_step": 155086, "epoch": 1742, "lr": 3.901053327625729e-05}
{"train_loss": 0.06995262950658798, "global_step": 155087, "epoch": 1742, "lr": 3.9009967716484046e-05}
{"train_loss": 0.10260476917028427, "global_step": 155088, "epoch": 1742, "lr": 3.900940215818826e-05}
{"train_loss": 0.08877977728843689, "global_step": 155089, "epoch": 1742, "lr": 3.900883660137005e-05}
{"train_loss": 0.08654885739088058, "global_step": 155090, "epoch": 1742, "lr": 3.900827104602944e-05}
{"train_loss": 0.10260702669620514, "global_step": 155091, "epoch": 1742, "lr": 3.9007705492166566e-05}
{"train_loss": 0.06835722923278809, "global_step": 155092, "epoch": 1742, "lr": 3.9007139939781444e-05}
{"train_loss": 0.1076345145702362, "global_step": 155093, "epoch": 1742, "lr": 3.90065743888742e-05}
{"train_loss": 0.12250661849975586, "global_step": 155094, "epoch": 1742, "lr": 3.900600883944486e-05}
{"train_loss": 0.09321397542953491, "global_step": 155095, "epoch": 1742, "lr": 3.9005443291493556e-05}
{"train_loss": 0.07556016743183136, "global_step": 155096, "epoch": 1742, "lr": 3.900487774502032e-05}
{"train_loss": 0.06610526889562607, "global_step": 155097, "epoch": 1742, "lr": 3.900431220002525e-05}
{"train_loss": 0.1121935322880745, "global_step": 155098, "epoch": 1742, "lr": 3.900374665650843e-05}
{"train_loss": 0.19988833367824554, "global_step": 155099, "epoch": 1742, "lr": 3.90031811144699e-05}
{"train_loss": 0.12394370883703232, "global_step": 155100, "epoch": 1742, "lr": 3.9002615573909775e-05}
{"train_loss": 0.09457544982433319, "global_step": 155101, "epoch": 1742, "lr": 3.900205003482809e-05}
{"train_loss": 0.15626537799835205, "global_step": 155102, "epoch": 1742, "lr": 3.9001484497224975e-05}
{"train_loss": 0.0951821431517601, "global_step": 155103, "epoch": 1742, "lr": 3.9000918961100445e-05}
{"train_loss": 0.07208777219057083, "global_step": 155104, "epoch": 1742, "lr": 3.900035342645464e-05}
{"train_loss": 0.10731277614831924, "global_step": 155105, "epoch": 1742, "lr": 3.899978789328758e-05}
{"train_loss": 0.0952051654458046, "global_step": 155106, "epoch": 1742, "lr": 3.899922236159938e-05}
{"train_loss": 0.13959293067455292, "global_step": 155107, "epoch": 1742, "lr": 3.899865683139008e-05}
{"train_loss": 0.1614259034395218, "global_step": 155108, "epoch": 1742, "lr": 3.89980913026598e-05}
{"train_loss": 0.10739410668611526, "global_step": 155109, "epoch": 1742, "lr": 3.8997525775408575e-05}
{"train_loss": 0.10297560691833496, "global_step": 155110, "epoch": 1742, "lr": 3.899696024963652e-05}
{"train_loss": 0.1242230087518692, "global_step": 155111, "epoch": 1742, "lr": 3.8996394725343656e-05}
{"train_loss": 0.08408772200345993, "global_step": 155112, "epoch": 1742, "lr": 3.899582920253013e-05}
{"train_loss": 0.06969992071390152, "global_step": 155113, "epoch": 1742, "lr": 3.8995263681195946e-05}
{"train_loss": 0.09713638573884964, "global_step": 155114, "epoch": 1742, "lr": 3.899469816134124e-05}
{"train_loss": 0.13876894116401672, "global_step": 155115, "epoch": 1742, "lr": 3.899413264296604e-05}
{"train_loss": 0.09530378878116608, "global_step": 155116, "epoch": 1742, "lr": 3.899356712607046e-05}
{"train_loss": 0.05063533037900925, "global_step": 155117, "epoch": 1742, "lr": 3.899300161065457e-05}
{"train_loss": 0.12670521438121796, "global_step": 155118, "epoch": 1742, "lr": 3.899243609671842e-05}
{"train_loss": 0.09230175614356995, "global_step": 155119, "epoch": 1742, "lr": 3.899187058426212e-05}
{"train_loss": 0.14495423436164856, "global_step": 155120, "epoch": 1742, "lr": 3.8991305073285705e-05}
{"train_loss": 0.1328621506690979, "global_step": 155121, "epoch": 1742, "lr": 3.8990739563789296e-05}
{"train_loss": 0.10170527547597885, "global_step": 155122, "epoch": 1742, "lr": 3.899017405577293e-05}
{"train_loss": 0.055256735533475876, "global_step": 155123, "epoch": 1742, "lr": 3.898960854923672e-05}
{"train_loss": 0.12536579370498657, "global_step": 155124, "epoch": 1742, "lr": 3.89890430441807e-05}
{"train_loss": 0.10506217926740646, "global_step": 155125, "epoch": 1742, "lr": 3.8988477540604994e-05}
{"train_loss": 0.10578285768795549, "global_step": 155126, "epoch": 1742, "lr": 3.8987912038509634e-05, "val_loss": 6.461122989654541}
{"train_loss": 0.14418169856071472, "global_step": 155127, "epoch": 1743, "lr": 3.898734653789473e-05}
{"train_loss": 0.13991372287273407, "global_step": 155128, "epoch": 1743, "lr": 3.898678103876033e-05}
{"train_loss": 0.0488310307264328, "global_step": 155129, "epoch": 1743, "lr": 3.898621554110654e-05}
{"train_loss": 0.05988398194313049, "global_step": 155130, "epoch": 1743, "lr": 3.898565004493339e-05}
{"train_loss": 0.13340319693088531, "global_step": 155131, "epoch": 1743, "lr": 3.8985084550241015e-05}
{"train_loss": 0.05376604199409485, "global_step": 155132, "epoch": 1743, "lr": 3.898451905702944e-05}
{"train_loss": 0.09871330857276917, "global_step": 155133, "epoch": 1743, "lr": 3.898395356529878e-05}
{"train_loss": 0.08416932821273804, "global_step": 155134, "epoch": 1743, "lr": 3.898338807504907e-05}
{"train_loss": 0.11340349167585373, "global_step": 155135, "epoch": 1743, "lr": 3.898282258628042e-05}
{"train_loss": 0.09921471774578094, "global_step": 155136, "epoch": 1743, "lr": 3.89822570989929e-05}
{"train_loss": 0.06199245527386665, "global_step": 155137, "epoch": 1743, "lr": 3.898169161318658e-05}
{"train_loss": 0.15204262733459473, "global_step": 155138, "epoch": 1743, "lr": 3.898112612886155e-05}
{"train_loss": 0.08522681891918182, "global_step": 155139, "epoch": 1743, "lr": 3.898056064601785e-05}
{"train_loss": 0.08232853561639786, "global_step": 155140, "epoch": 1743, "lr": 3.8979995164655605e-05}
{"train_loss": 0.05931016057729721, "global_step": 155141, "epoch": 1743, "lr": 3.8979429684774836e-05}
{"train_loss": 0.10373297333717346, "global_step": 155142, "epoch": 1743, "lr": 3.897886420637567e-05}
{"train_loss": 0.04740648716688156, "global_step": 155143, "epoch": 1743, "lr": 3.897829872945815e-05}
{"train_loss": 0.04302976280450821, "global_step": 155144, "epoch": 1743, "lr": 3.897773325402238e-05}
{"train_loss": 0.08127466589212418, "global_step": 155145, "epoch": 1743, "lr": 3.897716778006839e-05}
{"train_loss": 0.14094534516334534, "global_step": 155146, "epoch": 1743, "lr": 3.897660230759631e-05}
{"train_loss": 0.11222029477357864, "global_step": 155147, "epoch": 1743, "lr": 3.8976036836606174e-05}
{"train_loss": 0.10451919585466385, "global_step": 155148, "epoch": 1743, "lr": 3.8975471367098096e-05}
{"train_loss": 0.12179803103208542, "global_step": 155149, "epoch": 1743, "lr": 3.897490589907211e-05}
{"train_loss": 0.13074791431427002, "global_step": 155150, "epoch": 1743, "lr": 3.8974340432528335e-05}
{"train_loss": 0.1359139233827591, "global_step": 155151, "epoch": 1743, "lr": 3.89737749674668e-05}
{"train_loss": 0.053147803992033005, "global_step": 155152, "epoch": 1743, "lr": 3.897320950388762e-05}
{"train_loss": 0.08681850135326385, "global_step": 155153, "epoch": 1743, "lr": 3.8972644041790865e-05}
{"train_loss": 0.05990812927484512, "global_step": 155154, "epoch": 1743, "lr": 3.897207858117659e-05}
{"train_loss": 0.11785842478275299, "global_step": 155155, "epoch": 1743, "lr": 3.89715131220449e-05}
{"train_loss": 0.08657222986221313, "global_step": 155156, "epoch": 1743, "lr": 3.897094766439584e-05}
{"train_loss": 0.04823094233870506, "global_step": 155157, "epoch": 1743, "lr": 3.8970382208229524e-05}
{"train_loss": 0.06532571464776993, "global_step": 155158, "epoch": 1743, "lr": 3.896981675354597e-05}
{"train_loss": 0.13276948034763336, "global_step": 155159, "epoch": 1743, "lr": 3.8969251300345326e-05}
{"train_loss": 0.13050615787506104, "global_step": 155160, "epoch": 1743, "lr": 3.896868584862761e-05}
{"train_loss": 0.11789119243621826, "global_step": 155161, "epoch": 1743, "lr": 3.896812039839293e-05}
{"train_loss": 0.11061272770166397, "global_step": 155162, "epoch": 1743, "lr": 3.8967554949641336e-05}
{"train_loss": 0.06262005120515823, "global_step": 155163, "epoch": 1743, "lr": 3.896698950237295e-05}
{"train_loss": 0.09685449302196503, "global_step": 155164, "epoch": 1743, "lr": 3.8966424056587783e-05}
{"train_loss": 0.07359451055526733, "global_step": 155165, "epoch": 1743, "lr": 3.8965858612285976e-05}
{"train_loss": 0.09848706424236298, "global_step": 155166, "epoch": 1743, "lr": 3.896529316946755e-05}
{"train_loss": 0.08333651721477509, "global_step": 155167, "epoch": 1743, "lr": 3.8964727728132625e-05}
{"train_loss": 0.12765729427337646, "global_step": 155168, "epoch": 1743, "lr": 3.896416228828124e-05}
{"train_loss": 0.09259340167045593, "global_step": 155169, "epoch": 1743, "lr": 3.8963596849913506e-05}
{"train_loss": 0.1307596117258072, "global_step": 155170, "epoch": 1743, "lr": 3.8963031413029465e-05}
{"train_loss": 0.08292562514543533, "global_step": 155171, "epoch": 1743, "lr": 3.896246597762922e-05}
{"train_loss": 0.08454620093107224, "global_step": 155172, "epoch": 1743, "lr": 3.896190054371284e-05}
{"train_loss": 0.09693781286478043, "global_step": 155173, "epoch": 1743, "lr": 3.896133511128039e-05}
{"train_loss": 0.10534785687923431, "global_step": 155174, "epoch": 1743, "lr": 3.8960769680331954e-05}
{"train_loss": 0.1236162856221199, "global_step": 155175, "epoch": 1743, "lr": 3.8960204250867616e-05}
{"train_loss": 0.12230220437049866, "global_step": 155176, "epoch": 1743, "lr": 3.895963882288745e-05}
{"train_loss": 0.07817825675010681, "global_step": 155177, "epoch": 1743, "lr": 3.895907339639151e-05}
{"train_loss": 0.07699690759181976, "global_step": 155178, "epoch": 1743, "lr": 3.89585079713799e-05}
{"train_loss": 0.14454203844070435, "global_step": 155179, "epoch": 1743, "lr": 3.895794254785267e-05}
{"train_loss": 0.11756149679422379, "global_step": 155180, "epoch": 1743, "lr": 3.8957377125809923e-05}
{"train_loss": 0.11016232520341873, "global_step": 155181, "epoch": 1743, "lr": 3.89568117052517e-05}
{"train_loss": 0.04650903865695, "global_step": 155182, "epoch": 1743, "lr": 3.8956246286178135e-05}
{"train_loss": 0.13831332325935364, "global_step": 155183, "epoch": 1743, "lr": 3.895568086858924e-05}
{"train_loss": 0.07297586649656296, "global_step": 155184, "epoch": 1743, "lr": 3.895511545248513e-05}
{"train_loss": 0.08582548052072525, "global_step": 155185, "epoch": 1743, "lr": 3.895455003786587e-05}
{"train_loss": 0.09540330618619919, "global_step": 155186, "epoch": 1743, "lr": 3.8953984624731545e-05}
{"train_loss": 0.11264698952436447, "global_step": 155187, "epoch": 1743, "lr": 3.89534192130822e-05}
{"train_loss": 0.07604304701089859, "global_step": 155188, "epoch": 1743, "lr": 3.8952853802917936e-05}
{"train_loss": 0.0951598510146141, "global_step": 155189, "epoch": 1743, "lr": 3.895228839423884e-05}
{"train_loss": 0.05604353919625282, "global_step": 155190, "epoch": 1743, "lr": 3.895172298704497e-05}
{"train_loss": 0.10539691895246506, "global_step": 155191, "epoch": 1743, "lr": 3.895115758133641e-05}
{"train_loss": 0.13117384910583496, "global_step": 155192, "epoch": 1743, "lr": 3.895059217711322e-05}
{"train_loss": 0.06381788849830627, "global_step": 155193, "epoch": 1743, "lr": 3.89500267743755e-05}
{"train_loss": 0.07694724947214127, "global_step": 155194, "epoch": 1743, "lr": 3.894946137312331e-05}
{"train_loss": 0.1348968893289566, "global_step": 155195, "epoch": 1743, "lr": 3.894889597335674e-05}
{"train_loss": 0.06809774786233902, "global_step": 155196, "epoch": 1743, "lr": 3.894833057507583e-05}
{"train_loss": 0.10397031158208847, "global_step": 155197, "epoch": 1743, "lr": 3.8947765178280714e-05}
{"train_loss": 0.07391635328531265, "global_step": 155198, "epoch": 1743, "lr": 3.894719978297141e-05}
{"train_loss": 0.08543715626001358, "global_step": 155199, "epoch": 1743, "lr": 3.894663438914804e-05}
{"train_loss": 0.07850039005279541, "global_step": 155200, "epoch": 1743, "lr": 3.894606899681064e-05}
{"train_loss": 0.0939788743853569, "global_step": 155201, "epoch": 1743, "lr": 3.894550360595933e-05}
{"train_loss": 0.10079494118690491, "global_step": 155202, "epoch": 1743, "lr": 3.894493821659414e-05}
{"train_loss": 0.06617368757724762, "global_step": 155203, "epoch": 1743, "lr": 3.894437282871519e-05}
{"train_loss": 0.15930157899856567, "global_step": 155204, "epoch": 1743, "lr": 3.894380744232251e-05}
{"train_loss": 0.0766148567199707, "global_step": 155205, "epoch": 1743, "lr": 3.894324205741622e-05}
{"train_loss": 0.15526334941387177, "global_step": 155206, "epoch": 1743, "lr": 3.8942676673996364e-05}
{"train_loss": 0.12687255442142487, "global_step": 155207, "epoch": 1743, "lr": 3.894211129206302e-05}
{"train_loss": 0.07792391628026962, "global_step": 155208, "epoch": 1743, "lr": 3.8941545911616305e-05}
{"train_loss": 0.07199726998806, "global_step": 155209, "epoch": 1743, "lr": 3.8940980532656246e-05}
{"train_loss": 0.10092290490865707, "global_step": 155210, "epoch": 1743, "lr": 3.894041515518294e-05}
{"train_loss": 0.09612789750099182, "global_step": 155211, "epoch": 1743, "lr": 3.893984977919646e-05}
{"train_loss": 0.0874752551317215, "global_step": 155212, "epoch": 1743, "lr": 3.8939284404696883e-05}
{"train_loss": 0.06147771701216698, "global_step": 155213, "epoch": 1743, "lr": 3.893871903168429e-05}
{"train_loss": 0.08114515244960785, "global_step": 155214, "epoch": 1743, "lr": 3.893815366015876e-05}
{"train_loss": 0.09586484144242961, "global_step": 155215, "epoch": 1743, "lr": 3.893758829012033e-05, "val_loss": 6.617105007171631}
{"train_loss": 0.08093103021383286, "global_step": 155216, "epoch": 1744, "lr": 3.893702292156914e-05}
{"train_loss": 0.13348864018917084, "global_step": 155217, "epoch": 1744, "lr": 3.8936457554505204e-05}
{"train_loss": 0.14589139819145203, "global_step": 155218, "epoch": 1744, "lr": 3.8935892188928645e-05}
{"train_loss": 0.10124368220567703, "global_step": 155219, "epoch": 1744, "lr": 3.89353268248395e-05}
{"train_loss": 0.07400773465633392, "global_step": 155220, "epoch": 1744, "lr": 3.893476146223789e-05}
{"train_loss": 0.12594394385814667, "global_step": 155221, "epoch": 1744, "lr": 3.893419610112385e-05}
{"train_loss": 0.07773970812559128, "global_step": 155222, "epoch": 1744, "lr": 3.893363074149749e-05}
{"train_loss": 0.07936710119247437, "global_step": 155223, "epoch": 1744, "lr": 3.8933065383358844e-05}
{"train_loss": 0.10348083823919296, "global_step": 155224, "epoch": 1744, "lr": 3.893250002670801e-05}
{"train_loss": 0.15763626992702484, "global_step": 155225, "epoch": 1744, "lr": 3.8931934671545094e-05}
{"train_loss": 0.06351901590824127, "global_step": 155226, "epoch": 1744, "lr": 3.8931369317870126e-05}
{"train_loss": 0.07462729513645172, "global_step": 155227, "epoch": 1744, "lr": 3.893080396568321e-05}
{"train_loss": 0.09271718561649323, "global_step": 155228, "epoch": 1744, "lr": 3.8930238614984415e-05}
{"train_loss": 0.07075836509466171, "global_step": 155229, "epoch": 1744, "lr": 3.8929673265773806e-05}
{"train_loss": 0.10773080587387085, "global_step": 155230, "epoch": 1744, "lr": 3.8929107918051475e-05}
{"train_loss": 0.15525104105472565, "global_step": 155231, "epoch": 1744, "lr": 3.892854257181748e-05}
{"train_loss": 0.0956762433052063, "global_step": 155232, "epoch": 1744, "lr": 3.892797722707191e-05}
{"train_loss": 0.1462070792913437, "global_step": 155233, "epoch": 1744, "lr": 3.892741188381486e-05}
{"train_loss": 0.0797584280371666, "global_step": 155234, "epoch": 1744, "lr": 3.892684654204635e-05}
{"train_loss": 0.08800015598535538, "global_step": 155235, "epoch": 1744, "lr": 3.8926281201766516e-05}
{"train_loss": 0.09695029258728027, "global_step": 155236, "epoch": 1744, "lr": 3.892571586297539e-05}
{"train_loss": 0.10698208957910538, "global_step": 155237, "epoch": 1744, "lr": 3.892515052567309e-05}
{"train_loss": 0.07042265683412552, "global_step": 155238, "epoch": 1744, "lr": 3.892458518985964e-05}
{"train_loss": 0.10467003285884857, "global_step": 155239, "epoch": 1744, "lr": 3.8924019855535166e-05}
{"train_loss": 0.10279599577188492, "global_step": 155240, "epoch": 1744, "lr": 3.892345452269971e-05}
{"train_loss": 0.07159706205129623, "global_step": 155241, "epoch": 1744, "lr": 3.892288919135337e-05}
{"train_loss": 0.08561714738607407, "global_step": 155242, "epoch": 1744, "lr": 3.89223238614962e-05}
{"train_loss": 0.10127165168523788, "global_step": 155243, "epoch": 1744, "lr": 3.8921758533128286e-05}
{"train_loss": 0.1029045581817627, "global_step": 155244, "epoch": 1744, "lr": 3.892119320624973e-05}
{"train_loss": 0.08133954554796219, "global_step": 155245, "epoch": 1744, "lr": 3.892062788086056e-05}
{"train_loss": 0.05561596527695656, "global_step": 155246, "epoch": 1744, "lr": 3.892006255696089e-05}
{"train_loss": 0.07445967197418213, "global_step": 155247, "epoch": 1744, "lr": 3.8919497234550786e-05}
{"train_loss": 0.06471721827983856, "global_step": 155248, "epoch": 1744, "lr": 3.891893191363032e-05}
{"train_loss": 0.16174691915512085, "global_step": 155249, "epoch": 1744, "lr": 3.891836659419956e-05}
{"train_loss": 0.047993142157793045, "global_step": 155250, "epoch": 1744, "lr": 3.8917801276258594e-05}
{"train_loss": 0.07789931446313858, "global_step": 155251, "epoch": 1744, "lr": 3.89172359598075e-05}
{"train_loss": 0.1329491287469864, "global_step": 155252, "epoch": 1744, "lr": 3.891667064484634e-05}
{"train_loss": 0.10261306166648865, "global_step": 155253, "epoch": 1744, "lr": 3.8916105331375205e-05}
{"train_loss": 0.08319639414548874, "global_step": 155254, "epoch": 1744, "lr": 3.891554001939417e-05}
{"train_loss": 0.04742904752492905, "global_step": 155255, "epoch": 1744, "lr": 3.8914974708903285e-05}
{"train_loss": 0.16437600553035736, "global_step": 155256, "epoch": 1744, "lr": 3.891440939990267e-05}
{"train_loss": 0.09876138716936111, "global_step": 155257, "epoch": 1744, "lr": 3.8913844092392355e-05}
{"train_loss": 0.07603256404399872, "global_step": 155258, "epoch": 1744, "lr": 3.891327878637246e-05}
{"train_loss": 0.1385360211133957, "global_step": 155259, "epoch": 1744, "lr": 3.891271348184302e-05}
{"train_loss": 0.06416940689086914, "global_step": 155260, "epoch": 1744, "lr": 3.891214817880415e-05}
{"train_loss": 0.09950144588947296, "global_step": 155261, "epoch": 1744, "lr": 3.891158287725589e-05}
{"train_loss": 0.14437554776668549, "global_step": 155262, "epoch": 1744, "lr": 3.8911017577198324e-05}
{"train_loss": 0.11771135032176971, "global_step": 155263, "epoch": 1744, "lr": 3.891045227863156e-05}
{"train_loss": 0.08196567744016647, "global_step": 155264, "epoch": 1744, "lr": 3.890988698155563e-05}
{"train_loss": 0.056295931339263916, "global_step": 155265, "epoch": 1744, "lr": 3.890932168597066e-05}
{"train_loss": 0.10232318937778473, "global_step": 155266, "epoch": 1744, "lr": 3.8908756391876665e-05}
{"train_loss": 0.09187949448823929, "global_step": 155267, "epoch": 1744, "lr": 3.8908191099273776e-05}
{"train_loss": 0.1255279928445816, "global_step": 155268, "epoch": 1744, "lr": 3.890762580816204e-05}
{"train_loss": 0.15469762682914734, "global_step": 155269, "epoch": 1744, "lr": 3.890706051854153e-05}
{"train_loss": 0.10283869504928589, "global_step": 155270, "epoch": 1744, "lr": 3.890649523041234e-05}
{"train_loss": 0.041483934968709946, "global_step": 155271, "epoch": 1744, "lr": 3.890592994377453e-05}
{"train_loss": 0.13416039943695068, "global_step": 155272, "epoch": 1744, "lr": 3.8905364658628185e-05}
{"train_loss": 0.06054313853383064, "global_step": 155273, "epoch": 1744, "lr": 3.8904799374973386e-05}
{"train_loss": 0.0905299261212349, "global_step": 155274, "epoch": 1744, "lr": 3.8904234092810185e-05}
{"train_loss": 0.1404513120651245, "global_step": 155275, "epoch": 1744, "lr": 3.890366881213869e-05}
{"train_loss": 0.05727072432637215, "global_step": 155276, "epoch": 1744, "lr": 3.890310353295894e-05}
{"train_loss": 0.12357895076274872, "global_step": 155277, "epoch": 1744, "lr": 3.890253825527106e-05}
{"train_loss": 0.08507635444402695, "global_step": 155278, "epoch": 1744, "lr": 3.8901972979075075e-05}
{"train_loss": 0.11184920370578766, "global_step": 155279, "epoch": 1744, "lr": 3.890140770437108e-05}
{"train_loss": 0.09955327957868576, "global_step": 155280, "epoch": 1744, "lr": 3.8900842431159184e-05}
{"train_loss": 0.10932943969964981, "global_step": 155281, "epoch": 1744, "lr": 3.8900277159439405e-05}
{"train_loss": 0.08151713013648987, "global_step": 155282, "epoch": 1744, "lr": 3.8899711889211874e-05}
{"train_loss": 0.09525492042303085, "global_step": 155283, "epoch": 1744, "lr": 3.8899146620476626e-05}
{"train_loss": 0.06236790493130684, "global_step": 155284, "epoch": 1744, "lr": 3.889858135323377e-05}
{"train_loss": 0.057159680873155594, "global_step": 155285, "epoch": 1744, "lr": 3.889801608748334e-05}
{"train_loss": 0.15166862308979034, "global_step": 155286, "epoch": 1744, "lr": 3.8897450823225455e-05}
{"train_loss": 0.08724114298820496, "global_step": 155287, "epoch": 1744, "lr": 3.8896885560460176e-05}
{"train_loss": 0.18398219347000122, "global_step": 155288, "epoch": 1744, "lr": 3.889632029918756e-05}
{"train_loss": 0.11285343021154404, "global_step": 155289, "epoch": 1744, "lr": 3.88957550394077e-05}
{"train_loss": 0.057386357337236404, "global_step": 155290, "epoch": 1744, "lr": 3.8895189781120675e-05}
{"train_loss": 0.0500696562230587, "global_step": 155291, "epoch": 1744, "lr": 3.889462452432656e-05}
{"train_loss": 0.11084799468517303, "global_step": 155292, "epoch": 1744, "lr": 3.889405926902543e-05}
{"train_loss": 0.1705210953950882, "global_step": 155293, "epoch": 1744, "lr": 3.889349401521734e-05}
{"train_loss": 0.12287913262844086, "global_step": 155294, "epoch": 1744, "lr": 3.889292876290241e-05}
{"train_loss": 0.08312448859214783, "global_step": 155295, "epoch": 1744, "lr": 3.8892363512080666e-05}
{"train_loss": 0.1277480572462082, "global_step": 155296, "epoch": 1744, "lr": 3.889179826275223e-05}
{"train_loss": 0.11785615980625153, "global_step": 155297, "epoch": 1744, "lr": 3.8891233014917126e-05}
{"train_loss": 0.10952067375183105, "global_step": 155298, "epoch": 1744, "lr": 3.889066776857547e-05}
{"train_loss": 0.14581580460071564, "global_step": 155299, "epoch": 1744, "lr": 3.889010252372735e-05}
{"train_loss": 0.10834251344203949, "global_step": 155300, "epoch": 1744, "lr": 3.88895372803728e-05}
{"train_loss": 0.11627885699272156, "global_step": 155301, "epoch": 1744, "lr": 3.8888972038511935e-05}
{"train_loss": 0.11144672334194183, "global_step": 155302, "epoch": 1744, "lr": 3.888840679814479e-05}
{"train_loss": 0.13007843494415283, "global_step": 155303, "epoch": 1744, "lr": 3.8887841559271484e-05}
{"train_loss": 0.10066180646921812, "global_step": 155304, "epoch": 1744, "lr": 3.888727632189205e-05, "val_loss": 6.4164581298828125}
{"train_loss": 0.15113000571727753, "global_step": 155305, "epoch": 1745, "lr": 3.888671108600661e-05}
{"train_loss": 0.06195200979709625, "global_step": 155306, "epoch": 1745, "lr": 3.888614585161521e-05}
{"train_loss": 0.1080416813492775, "global_step": 155307, "epoch": 1745, "lr": 3.888558061871792e-05}
{"train_loss": 0.0830955058336258, "global_step": 155308, "epoch": 1745, "lr": 3.888501538731484e-05}
{"train_loss": 0.14689666032791138, "global_step": 155309, "epoch": 1745, "lr": 3.888445015740603e-05}
{"train_loss": 0.09831146895885468, "global_step": 155310, "epoch": 1745, "lr": 3.888388492899157e-05}
{"train_loss": 0.10765080153942108, "global_step": 155311, "epoch": 1745, "lr": 3.8883319702071545e-05}
{"train_loss": 0.03774026781320572, "global_step": 155312, "epoch": 1745, "lr": 3.8882754476646006e-05}
{"train_loss": 0.048744603991508484, "global_step": 155313, "epoch": 1745, "lr": 3.888218925271506e-05}
{"train_loss": 0.14431780576705933, "global_step": 155314, "epoch": 1745, "lr": 3.8881624030278754e-05}
{"train_loss": 0.07977603375911713, "global_step": 155315, "epoch": 1745, "lr": 3.888105880933717e-05}
{"train_loss": 0.19120502471923828, "global_step": 155316, "epoch": 1745, "lr": 3.888049358989042e-05}
{"train_loss": 0.06629214435815811, "global_step": 155317, "epoch": 1745, "lr": 3.887992837193853e-05}
{"train_loss": 0.05087229982018471, "global_step": 155318, "epoch": 1745, "lr": 3.8879363155481616e-05}
{"train_loss": 0.07248037308454514, "global_step": 155319, "epoch": 1745, "lr": 3.887879794051972e-05}
{"train_loss": 0.1032448336482048, "global_step": 155320, "epoch": 1745, "lr": 3.887823272705295e-05}
{"train_loss": 0.06254184991121292, "global_step": 155321, "epoch": 1745, "lr": 3.8877667515081344e-05}
{"train_loss": 0.10460607707500458, "global_step": 155322, "epoch": 1745, "lr": 3.887710230460502e-05}
{"train_loss": 0.07458100467920303, "global_step": 155323, "epoch": 1745, "lr": 3.8876537095624024e-05}
{"train_loss": 0.11436334252357483, "global_step": 155324, "epoch": 1745, "lr": 3.887597188813845e-05}
{"train_loss": 0.0759742334485054, "global_step": 155325, "epoch": 1745, "lr": 3.887540668214836e-05}
{"train_loss": 0.07126995921134949, "global_step": 155326, "epoch": 1745, "lr": 3.887484147765385e-05}
{"train_loss": 0.09301570802927017, "global_step": 155327, "epoch": 1745, "lr": 3.887427627465496e-05}
{"train_loss": 0.07475151866674423, "global_step": 155328, "epoch": 1745, "lr": 3.8873711073151795e-05}
{"train_loss": 0.07461141794919968, "global_step": 155329, "epoch": 1745, "lr": 3.8873145873144434e-05}
{"train_loss": 0.1257505863904953, "global_step": 155330, "epoch": 1745, "lr": 3.887258067463294e-05}
{"train_loss": 0.1147637888789177, "global_step": 155331, "epoch": 1745, "lr": 3.887201547761738e-05}
{"train_loss": 0.07933604717254639, "global_step": 155332, "epoch": 1745, "lr": 3.887145028209786e-05}
{"train_loss": 0.16003482043743134, "global_step": 155333, "epoch": 1745, "lr": 3.887088508807442e-05}
{"train_loss": 0.13066798448562622, "global_step": 155334, "epoch": 1745, "lr": 3.8870319895547144e-05}
{"train_loss": 0.06823284178972244, "global_step": 155335, "epoch": 1745, "lr": 3.886975470451615e-05}
{"train_loss": 0.09022939205169678, "global_step": 155336, "epoch": 1745, "lr": 3.886918951498145e-05}
{"train_loss": 0.09553144872188568, "global_step": 155337, "epoch": 1745, "lr": 3.886862432694318e-05}
{"train_loss": 0.0991329774260521, "global_step": 155338, "epoch": 1745, "lr": 3.886805914040137e-05}
{"train_loss": 0.08078248798847198, "global_step": 155339, "epoch": 1745, "lr": 3.886749395535613e-05}
{"train_loss": 0.06324800103902817, "global_step": 155340, "epoch": 1745, "lr": 3.88669287718075e-05}
{"train_loss": 0.0784946009516716, "global_step": 155341, "epoch": 1745, "lr": 3.88663635897556e-05}
{"train_loss": 0.1256331503391266, "global_step": 155342, "epoch": 1745, "lr": 3.886579840920045e-05}
{"train_loss": 0.12456893175840378, "global_step": 155343, "epoch": 1745, "lr": 3.8865233230142184e-05}
{"train_loss": 0.14550678431987762, "global_step": 155344, "epoch": 1745, "lr": 3.8864668052580844e-05}
{"train_loss": 0.0459582544863224, "global_step": 155345, "epoch": 1745, "lr": 3.886410287651651e-05}
{"train_loss": 0.1117263063788414, "global_step": 155346, "epoch": 1745, "lr": 3.886353770194927e-05}
{"train_loss": 0.15148887038230896, "global_step": 155347, "epoch": 1745, "lr": 3.8862972528879173e-05}
{"train_loss": 0.08185233920812607, "global_step": 155348, "epoch": 1745, "lr": 3.886240735730633e-05}
{"train_loss": 0.1659296303987503, "global_step": 155349, "epoch": 1745, "lr": 3.886184218723081e-05}
{"train_loss": 0.07701157033443451, "global_step": 155350, "epoch": 1745, "lr": 3.886127701865265e-05}
{"train_loss": 0.14375737309455872, "global_step": 155351, "epoch": 1745, "lr": 3.886071185157196e-05}
{"train_loss": 0.11197201162576675, "global_step": 155352, "epoch": 1745, "lr": 3.886014668598884e-05}
{"train_loss": 0.11235679686069489, "global_step": 155353, "epoch": 1745, "lr": 3.885958152190331e-05}
{"train_loss": 0.07863174378871918, "global_step": 155354, "epoch": 1745, "lr": 3.8859016359315495e-05}
{"train_loss": 0.13992291688919067, "global_step": 155355, "epoch": 1745, "lr": 3.885845119822543e-05}
{"train_loss": 0.05682193115353584, "global_step": 155356, "epoch": 1745, "lr": 3.885788603863323e-05}
{"train_loss": 0.09654947370290756, "global_step": 155357, "epoch": 1745, "lr": 3.885732088053893e-05}
{"train_loss": 0.09492862224578857, "global_step": 155358, "epoch": 1745, "lr": 3.885675572394264e-05}
{"train_loss": 0.1085195317864418, "global_step": 155359, "epoch": 1745, "lr": 3.885619056884441e-05}
{"train_loss": 0.16732774674892426, "global_step": 155360, "epoch": 1745, "lr": 3.885562541524436e-05}
{"train_loss": 0.12571197748184204, "global_step": 155361, "epoch": 1745, "lr": 3.88550602631425e-05}
{"train_loss": 0.11298679560422897, "global_step": 155362, "epoch": 1745, "lr": 3.885449511253897e-05}
{"train_loss": 0.09008597582578659, "global_step": 155363, "epoch": 1745, "lr": 3.8853929963433804e-05}
{"train_loss": 0.06560906767845154, "global_step": 155364, "epoch": 1745, "lr": 3.885336481582709e-05}
{"train_loss": 0.06369829922914505, "global_step": 155365, "epoch": 1745, "lr": 3.885279966971891e-05}
{"train_loss": 0.10487765073776245, "global_step": 155366, "epoch": 1745, "lr": 3.8852234525109334e-05}
{"train_loss": 0.12222716957330704, "global_step": 155367, "epoch": 1745, "lr": 3.885166938199844e-05}
{"train_loss": 0.09214413166046143, "global_step": 155368, "epoch": 1745, "lr": 3.885110424038629e-05}
{"train_loss": 0.12005177140235901, "global_step": 155369, "epoch": 1745, "lr": 3.8850539100272984e-05}
{"train_loss": 0.09443829953670502, "global_step": 155370, "epoch": 1745, "lr": 3.8849973961658584e-05}
{"train_loss": 0.12277109920978546, "global_step": 155371, "epoch": 1745, "lr": 3.8849408824543175e-05}
{"train_loss": 0.08857277780771255, "global_step": 155372, "epoch": 1745, "lr": 3.884884368892682e-05}
{"train_loss": 0.09848012030124664, "global_step": 155373, "epoch": 1745, "lr": 3.8848278554809615e-05}
{"train_loss": 0.05285383760929108, "global_step": 155374, "epoch": 1745, "lr": 3.88477134221916e-05}
{"train_loss": 0.07972199469804764, "global_step": 155375, "epoch": 1745, "lr": 3.88471482910729e-05}
{"train_loss": 0.07040184736251831, "global_step": 155376, "epoch": 1745, "lr": 3.884658316145354e-05}
{"train_loss": 0.11536666750907898, "global_step": 155377, "epoch": 1745, "lr": 3.884601803333365e-05}
{"train_loss": 0.06228043884038925, "global_step": 155378, "epoch": 1745, "lr": 3.884545290671325e-05}
{"train_loss": 0.05939578637480736, "global_step": 155379, "epoch": 1745, "lr": 3.8844887781592465e-05}
{"train_loss": 0.09776902943849564, "global_step": 155380, "epoch": 1745, "lr": 3.884432265797132e-05}
{"train_loss": 0.1249881237745285, "global_step": 155381, "epoch": 1745, "lr": 3.8843757535849954e-05}
{"train_loss": 0.06810694932937622, "global_step": 155382, "epoch": 1745, "lr": 3.884319241522838e-05}
{"train_loss": 0.1196579560637474, "global_step": 155383, "epoch": 1745, "lr": 3.884262729610673e-05}
{"train_loss": 0.08645876497030258, "global_step": 155384, "epoch": 1745, "lr": 3.8842062178485036e-05}
{"train_loss": 0.12193980813026428, "global_step": 155385, "epoch": 1745, "lr": 3.884149706236339e-05}
{"train_loss": 0.08404451608657837, "global_step": 155386, "epoch": 1745, "lr": 3.884093194774187e-05}
{"train_loss": 0.09131436794996262, "global_step": 155387, "epoch": 1745, "lr": 3.8840366834620556e-05}
{"train_loss": 0.06767375767230988, "global_step": 155388, "epoch": 1745, "lr": 3.883980172299951e-05}
{"train_loss": 0.10120441019535065, "global_step": 155389, "epoch": 1745, "lr": 3.883923661287882e-05}
{"train_loss": 0.08260759711265564, "global_step": 155390, "epoch": 1745, "lr": 3.8838671504258567e-05}
{"train_loss": 0.15191085636615753, "global_step": 155391, "epoch": 1745, "lr": 3.88381063971388e-05}
{"train_loss": 0.1321755051612854, "global_step": 155392, "epoch": 1745, "lr": 3.883754129151964e-05}
{"train_loss": 0.09891890751177006, "global_step": 155393, "epoch": 1745, "lr": 3.883697618740111e-05, "val_loss": 6.804654598236084, "train_action_mse_error": 5.809503555297852}
{"train_loss": 0.16907167434692383, "global_step": 155394, "epoch": 1746, "lr": 3.883641108478333e-05}
{"train_loss": 0.0941135361790657, "global_step": 155395, "epoch": 1746, "lr": 3.883584598366634e-05}
{"train_loss": 0.1715385615825653, "global_step": 155396, "epoch": 1746, "lr": 3.883528088405026e-05}
{"train_loss": 0.05975418537855148, "global_step": 155397, "epoch": 1746, "lr": 3.8834715785935116e-05}
{"train_loss": 0.09403996914625168, "global_step": 155398, "epoch": 1746, "lr": 3.883415068932103e-05}
{"train_loss": 0.1122419685125351, "global_step": 155399, "epoch": 1746, "lr": 3.883358559420804e-05}
{"train_loss": 0.11341098695993423, "global_step": 155400, "epoch": 1746, "lr": 3.883302050059625e-05}
{"train_loss": 0.09637641161680222, "global_step": 155401, "epoch": 1746, "lr": 3.883245540848571e-05}
{"train_loss": 0.08835787326097488, "global_step": 155402, "epoch": 1746, "lr": 3.8831890317876526e-05}
{"train_loss": 0.06353937089443207, "global_step": 155403, "epoch": 1746, "lr": 3.8831325228768756e-05}
{"train_loss": 0.06986124068498611, "global_step": 155404, "epoch": 1746, "lr": 3.883076014116247e-05}
{"train_loss": 0.09848422557115555, "global_step": 155405, "epoch": 1746, "lr": 3.883019505505776e-05}
{"train_loss": 0.09941449016332626, "global_step": 155406, "epoch": 1746, "lr": 3.882962997045468e-05}
{"train_loss": 0.11619158089160919, "global_step": 155407, "epoch": 1746, "lr": 3.882906488735334e-05}
{"train_loss": 0.07166735082864761, "global_step": 155408, "epoch": 1746, "lr": 3.882849980575377e-05}
{"train_loss": 0.07131511718034744, "global_step": 155409, "epoch": 1746, "lr": 3.88279347256561e-05}
{"train_loss": 0.0362708680331707, "global_step": 155410, "epoch": 1746, "lr": 3.8827369647060356e-05}
{"train_loss": 0.08683951199054718, "global_step": 155411, "epoch": 1746, "lr": 3.882680456996666e-05}
{"train_loss": 0.043637797236442566, "global_step": 155412, "epoch": 1746, "lr": 3.882623949437504e-05}
{"train_loss": 0.10465393215417862, "global_step": 155413, "epoch": 1746, "lr": 3.882567442028561e-05}
{"train_loss": 0.08285244554281235, "global_step": 155414, "epoch": 1746, "lr": 3.882510934769842e-05}
{"train_loss": 0.12971411645412445, "global_step": 155415, "epoch": 1746, "lr": 3.882454427661358e-05}
{"train_loss": 0.11635338515043259, "global_step": 155416, "epoch": 1746, "lr": 3.882397920703111e-05}
{"train_loss": 0.1047593355178833, "global_step": 155417, "epoch": 1746, "lr": 3.882341413895115e-05}
{"train_loss": 0.06556472927331924, "global_step": 155418, "epoch": 1746, "lr": 3.882284907237373e-05}
{"train_loss": 0.10536546260118484, "global_step": 155419, "epoch": 1746, "lr": 3.8822284007298956e-05}
{"train_loss": 0.09581523388624191, "global_step": 155420, "epoch": 1746, "lr": 3.882171894372687e-05}
{"train_loss": 0.06085177883505821, "global_step": 155421, "epoch": 1746, "lr": 3.882115388165759e-05}
{"train_loss": 0.10119442641735077, "global_step": 155422, "epoch": 1746, "lr": 3.882058882109116e-05}
{"train_loss": 0.1179148256778717, "global_step": 155423, "epoch": 1746, "lr": 3.8820023762027655e-05}
{"train_loss": 0.06253478676080704, "global_step": 155424, "epoch": 1746, "lr": 3.8819458704467173e-05}
{"train_loss": 0.1425207406282425, "global_step": 155425, "epoch": 1746, "lr": 3.881889364840977e-05}
{"train_loss": 0.051758985966444016, "global_step": 155426, "epoch": 1746, "lr": 3.881832859385554e-05}
{"train_loss": 0.08160922676324844, "global_step": 155427, "epoch": 1746, "lr": 3.8817763540804535e-05}
{"train_loss": 0.12525179982185364, "global_step": 155428, "epoch": 1746, "lr": 3.881719848925687e-05}
{"train_loss": 0.07230638712644577, "global_step": 155429, "epoch": 1746, "lr": 3.881663343921257e-05}
{"train_loss": 0.11654972285032272, "global_step": 155430, "epoch": 1746, "lr": 3.8816068390671756e-05}
{"train_loss": 0.09845442324876785, "global_step": 155431, "epoch": 1746, "lr": 3.8815503343634466e-05}
{"train_loss": 0.0670107901096344, "global_step": 155432, "epoch": 1746, "lr": 3.881493829810082e-05}
{"train_loss": 0.10932643711566925, "global_step": 155433, "epoch": 1746, "lr": 3.8814373254070844e-05}
{"train_loss": 0.10260885208845139, "global_step": 155434, "epoch": 1746, "lr": 3.881380821154467e-05}
{"train_loss": 0.08927496522665024, "global_step": 155435, "epoch": 1746, "lr": 3.8813243170522315e-05}
{"train_loss": 0.05810466408729553, "global_step": 155436, "epoch": 1746, "lr": 3.88126781310039e-05}
{"train_loss": 0.09693306684494019, "global_step": 155437, "epoch": 1746, "lr": 3.881211309298947e-05}
{"train_loss": 0.09247865527868271, "global_step": 155438, "epoch": 1746, "lr": 3.881154805647913e-05}
{"train_loss": 0.1254037469625473, "global_step": 155439, "epoch": 1746, "lr": 3.881098302147293e-05}
{"train_loss": 0.07495005428791046, "global_step": 155440, "epoch": 1746, "lr": 3.8810417987970966e-05}
{"train_loss": 0.07090739160776138, "global_step": 155441, "epoch": 1746, "lr": 3.8809852955973295e-05}
{"train_loss": 0.08868897706270218, "global_step": 155442, "epoch": 1746, "lr": 3.8809287925479995e-05}
{"train_loss": 0.08349460363388062, "global_step": 155443, "epoch": 1746, "lr": 3.880872289649118e-05}
{"train_loss": 0.07211374491453171, "global_step": 155444, "epoch": 1746, "lr": 3.880815786900687e-05}
{"train_loss": 0.19908128678798676, "global_step": 155445, "epoch": 1746, "lr": 3.880759284302718e-05}
{"train_loss": 0.06439658999443054, "global_step": 155446, "epoch": 1746, "lr": 3.880702781855215e-05}
{"train_loss": 0.13044272363185883, "global_step": 155447, "epoch": 1746, "lr": 3.880646279558191e-05}
{"train_loss": 0.12984172999858856, "global_step": 155448, "epoch": 1746, "lr": 3.880589777411648e-05}
{"train_loss": 0.09312116354703903, "global_step": 155449, "epoch": 1746, "lr": 3.880533275415599e-05}
{"train_loss": 0.14029061794281006, "global_step": 155450, "epoch": 1746, "lr": 3.8804767735700456e-05}
{"train_loss": 0.08220885694026947, "global_step": 155451, "epoch": 1746, "lr": 3.8804202718750006e-05}
{"train_loss": 0.09518076479434967, "global_step": 155452, "epoch": 1746, "lr": 3.880363770330467e-05}
{"train_loss": 0.10191912204027176, "global_step": 155453, "epoch": 1746, "lr": 3.880307268936458e-05}
{"train_loss": 0.14739196002483368, "global_step": 155454, "epoch": 1746, "lr": 3.880250767692975e-05}
{"train_loss": 0.10143879801034927, "global_step": 155455, "epoch": 1746, "lr": 3.8801942666000316e-05}
{"train_loss": 0.07590650767087936, "global_step": 155456, "epoch": 1746, "lr": 3.8801377656576296e-05}
{"train_loss": 0.08796878904104233, "global_step": 155457, "epoch": 1746, "lr": 3.880081264865781e-05}
{"train_loss": 0.0893012285232544, "global_step": 155458, "epoch": 1746, "lr": 3.880024764224491e-05}
{"train_loss": 0.09762740880250931, "global_step": 155459, "epoch": 1746, "lr": 3.8799682637337694e-05}
{"train_loss": 0.19975121319293976, "global_step": 155460, "epoch": 1746, "lr": 3.8799117633936214e-05}
{"train_loss": 0.09502645581960678, "global_step": 155461, "epoch": 1746, "lr": 3.879855263204055e-05}
{"train_loss": 0.1625349521636963, "global_step": 155462, "epoch": 1746, "lr": 3.87979876316508e-05}
{"train_loss": 0.07974027842283249, "global_step": 155463, "epoch": 1746, "lr": 3.879742263276701e-05}
{"train_loss": 0.12520717084407806, "global_step": 155464, "epoch": 1746, "lr": 3.879685763538928e-05}
{"train_loss": 0.11420956254005432, "global_step": 155465, "epoch": 1746, "lr": 3.879629263951766e-05}
{"train_loss": 0.12136788666248322, "global_step": 155466, "epoch": 1746, "lr": 3.8795727645152266e-05}
{"train_loss": 0.1767803281545639, "global_step": 155467, "epoch": 1746, "lr": 3.879516265229313e-05}
{"train_loss": 0.08438912034034729, "global_step": 155468, "epoch": 1746, "lr": 3.8794597660940366e-05}
{"train_loss": 0.06998182088136673, "global_step": 155469, "epoch": 1746, "lr": 3.879403267109401e-05}
{"train_loss": 0.05899927765130997, "global_step": 155470, "epoch": 1746, "lr": 3.879346768275418e-05}
{"train_loss": 0.06199502944946289, "global_step": 155471, "epoch": 1746, "lr": 3.8792902695920905e-05}
{"train_loss": 0.11358320713043213, "global_step": 155472, "epoch": 1746, "lr": 3.8792337710594316e-05}
{"train_loss": 0.08448272198438644, "global_step": 155473, "epoch": 1746, "lr": 3.879177272677444e-05}
{"train_loss": 0.15253989398479462, "global_step": 155474, "epoch": 1746, "lr": 3.879120774446139e-05}
{"train_loss": 0.09412533789873123, "global_step": 155475, "epoch": 1746, "lr": 3.879064276365521e-05}
{"train_loss": 0.0945499986410141, "global_step": 155476, "epoch": 1746, "lr": 3.8790077784356016e-05}
{"train_loss": 0.06928905099630356, "global_step": 155477, "epoch": 1746, "lr": 3.878951280656383e-05}
{"train_loss": 0.10257591307163239, "global_step": 155478, "epoch": 1746, "lr": 3.878894783027877e-05}
{"train_loss": 0.093017578125, "global_step": 155479, "epoch": 1746, "lr": 3.87883828555009e-05}
{"train_loss": 0.07760012149810791, "global_step": 155480, "epoch": 1746, "lr": 3.878781788223029e-05}
{"train_loss": 0.10511773824691772, "global_step": 155481, "epoch": 1746, "lr": 3.8787252910467044e-05}
{"train_loss": 0.09856474278180787, "global_step": 155482, "epoch": 1746, "lr": 3.878668794021118e-05, "val_loss": 6.514389991760254}
{"train_loss": 0.14750318229198456, "global_step": 155483, "epoch": 1747, "lr": 3.8786122971462844e-05}
{"train_loss": 0.07159632444381714, "global_step": 155484, "epoch": 1747, "lr": 3.878555800422205e-05}
{"train_loss": 0.13773241639137268, "global_step": 155485, "epoch": 1747, "lr": 3.878499303848892e-05}
{"train_loss": 0.1041603833436966, "global_step": 155486, "epoch": 1747, "lr": 3.878442807426349e-05}
{"train_loss": 0.1151493713259697, "global_step": 155487, "epoch": 1747, "lr": 3.878386311154588e-05}
{"train_loss": 0.19014419615268707, "global_step": 155488, "epoch": 1747, "lr": 3.8783298150336123e-05}
{"train_loss": 0.04287165403366089, "global_step": 155489, "epoch": 1747, "lr": 3.8782733190634334e-05}
{"train_loss": 0.11595483869314194, "global_step": 155490, "epoch": 1747, "lr": 3.878216823244055e-05}
{"train_loss": 0.13631656765937805, "global_step": 155491, "epoch": 1747, "lr": 3.8781603275754886e-05}
{"train_loss": 0.11920572072267532, "global_step": 155492, "epoch": 1747, "lr": 3.878103832057738e-05}
{"train_loss": 0.10928353667259216, "global_step": 155493, "epoch": 1747, "lr": 3.878047336690814e-05}
{"train_loss": 0.09928108751773834, "global_step": 155494, "epoch": 1747, "lr": 3.8779908414747216e-05}
{"train_loss": 0.1336504966020584, "global_step": 155495, "epoch": 1747, "lr": 3.877934346409472e-05}
{"train_loss": 0.03699706494808197, "global_step": 155496, "epoch": 1747, "lr": 3.8778778514950674e-05}
{"train_loss": 0.09636437892913818, "global_step": 155497, "epoch": 1747, "lr": 3.87782135673152e-05}
{"train_loss": 0.07832703739404678, "global_step": 155498, "epoch": 1747, "lr": 3.8777648621188353e-05}
{"train_loss": 0.11534463614225388, "global_step": 155499, "epoch": 1747, "lr": 3.877708367657023e-05}
{"train_loss": 0.07053816318511963, "global_step": 155500, "epoch": 1747, "lr": 3.8776518733460886e-05}
{"train_loss": 0.09515737742185593, "global_step": 155501, "epoch": 1747, "lr": 3.8775953791860386e-05}
{"train_loss": 0.13863591849803925, "global_step": 155502, "epoch": 1747, "lr": 3.8775388851768846e-05}
{"train_loss": 0.04057907685637474, "global_step": 155503, "epoch": 1747, "lr": 3.877482391318629e-05}
{"train_loss": 0.13958165049552917, "global_step": 155504, "epoch": 1747, "lr": 3.877425897611285e-05}
{"train_loss": 0.10176166892051697, "global_step": 155505, "epoch": 1747, "lr": 3.877369404054855e-05}
{"train_loss": 0.11157191544771194, "global_step": 155506, "epoch": 1747, "lr": 3.877312910649352e-05}
{"train_loss": 0.08639422804117203, "global_step": 155507, "epoch": 1747, "lr": 3.8772564173947774e-05}
{"train_loss": 0.10070869326591492, "global_step": 155508, "epoch": 1747, "lr": 3.877199924291144e-05}
{"train_loss": 0.14177305996418, "global_step": 155509, "epoch": 1747, "lr": 3.877143431338456e-05}
{"train_loss": 0.1356370896100998, "global_step": 155510, "epoch": 1747, "lr": 3.877086938536724e-05}
{"train_loss": 0.09430088102817535, "global_step": 155511, "epoch": 1747, "lr": 3.877030445885952e-05}
{"train_loss": 0.0792352631688118, "global_step": 155512, "epoch": 1747, "lr": 3.876973953386152e-05}
{"train_loss": 0.07968949526548386, "global_step": 155513, "epoch": 1747, "lr": 3.876917461037327e-05}
{"train_loss": 0.1262148767709732, "global_step": 155514, "epoch": 1747, "lr": 3.876860968839489e-05}
{"train_loss": 0.09044450521469116, "global_step": 155515, "epoch": 1747, "lr": 3.8768044767926405e-05}
{"train_loss": 0.1314314603805542, "global_step": 155516, "epoch": 1747, "lr": 3.876747984896793e-05}
{"train_loss": 0.08559711277484894, "global_step": 155517, "epoch": 1747, "lr": 3.876691493151953e-05}
{"train_loss": 0.1017945408821106, "global_step": 155518, "epoch": 1747, "lr": 3.8766350015581285e-05}
{"train_loss": 0.09188157320022583, "global_step": 155519, "epoch": 1747, "lr": 3.8765785101153284e-05}
{"train_loss": 0.1151030957698822, "global_step": 155520, "epoch": 1747, "lr": 3.8765220188235564e-05}
{"train_loss": 0.10118085145950317, "global_step": 155521, "epoch": 1747, "lr": 3.876465527682823e-05}
{"train_loss": 0.09268971532583237, "global_step": 155522, "epoch": 1747, "lr": 3.876409036693134e-05}
{"train_loss": 0.06901367008686066, "global_step": 155523, "epoch": 1747, "lr": 3.876352545854501e-05}
{"train_loss": 0.1157728061079979, "global_step": 155524, "epoch": 1747, "lr": 3.876296055166926e-05}
{"train_loss": 0.1290881633758545, "global_step": 155525, "epoch": 1747, "lr": 3.8762395646304204e-05}
{"train_loss": 0.13660506904125214, "global_step": 155526, "epoch": 1747, "lr": 3.87618307424499e-05}
{"train_loss": 0.058069467544555664, "global_step": 155527, "epoch": 1747, "lr": 3.876126584010644e-05}
{"train_loss": 0.11401735991239548, "global_step": 155528, "epoch": 1747, "lr": 3.876070093927388e-05}
{"train_loss": 0.09817610681056976, "global_step": 155529, "epoch": 1747, "lr": 3.876013603995232e-05}
{"train_loss": 0.07584773004055023, "global_step": 155530, "epoch": 1747, "lr": 3.8759571142141806e-05}
{"train_loss": 0.06448071449995041, "global_step": 155531, "epoch": 1747, "lr": 3.875900624584245e-05}
{"train_loss": 0.08720485866069794, "global_step": 155532, "epoch": 1747, "lr": 3.875844135105429e-05}
{"train_loss": 0.07185826450586319, "global_step": 155533, "epoch": 1747, "lr": 3.875787645777743e-05}
{"train_loss": 0.0926637202501297, "global_step": 155534, "epoch": 1747, "lr": 3.875731156601194e-05}
{"train_loss": 0.09175299108028412, "global_step": 155535, "epoch": 1747, "lr": 3.8756746675757885e-05}
{"train_loss": 0.09739813208580017, "global_step": 155536, "epoch": 1747, "lr": 3.875618178701536e-05}
{"train_loss": 0.06520287692546844, "global_step": 155537, "epoch": 1747, "lr": 3.8755616899784413e-05}
{"train_loss": 0.07242940366268158, "global_step": 155538, "epoch": 1747, "lr": 3.875505201406516e-05}
{"train_loss": 0.10449354350566864, "global_step": 155539, "epoch": 1747, "lr": 3.8754487129857626e-05}
{"train_loss": 0.05641663819551468, "global_step": 155540, "epoch": 1747, "lr": 3.875392224716193e-05}
{"train_loss": 0.12057365477085114, "global_step": 155541, "epoch": 1747, "lr": 3.875335736597812e-05}
{"train_loss": 0.0995897501707077, "global_step": 155542, "epoch": 1747, "lr": 3.8752792486306305e-05}
{"train_loss": 0.040118005126714706, "global_step": 155543, "epoch": 1747, "lr": 3.875222760814652e-05}
{"train_loss": 0.09882457554340363, "global_step": 155544, "epoch": 1747, "lr": 3.875166273149887e-05}
{"train_loss": 0.13043265044689178, "global_step": 155545, "epoch": 1747, "lr": 3.8751097856363416e-05}
{"train_loss": 0.1306452453136444, "global_step": 155546, "epoch": 1747, "lr": 3.875053298274025e-05}
{"train_loss": 0.11800547689199448, "global_step": 155547, "epoch": 1747, "lr": 3.874996811062942e-05}
{"train_loss": 0.06323941797018051, "global_step": 155548, "epoch": 1747, "lr": 3.8749403240031045e-05}
{"train_loss": 0.1355677843093872, "global_step": 155549, "epoch": 1747, "lr": 3.874883837094515e-05}
{"train_loss": 0.058291442692279816, "global_step": 155550, "epoch": 1747, "lr": 3.8748273503371856e-05}
{"train_loss": 0.10220476239919662, "global_step": 155551, "epoch": 1747, "lr": 3.8747708637311205e-05}
{"train_loss": 0.07464510202407837, "global_step": 155552, "epoch": 1747, "lr": 3.8747143772763294e-05}
{"train_loss": 0.0817704051733017, "global_step": 155553, "epoch": 1747, "lr": 3.874657890972819e-05}
{"train_loss": 0.0883786752820015, "global_step": 155554, "epoch": 1747, "lr": 3.8746014048205976e-05}
{"train_loss": 0.05769569054245949, "global_step": 155555, "epoch": 1747, "lr": 3.874544918819672e-05}
{"train_loss": 0.11643465608358383, "global_step": 155556, "epoch": 1747, "lr": 3.8744884329700506e-05}
{"train_loss": 0.08599685877561569, "global_step": 155557, "epoch": 1747, "lr": 3.8744319472717405e-05}
{"train_loss": 0.061786625534296036, "global_step": 155558, "epoch": 1747, "lr": 3.874375461724748e-05}
{"train_loss": 0.08579283207654953, "global_step": 155559, "epoch": 1747, "lr": 3.874318976329084e-05}
{"train_loss": 0.14279767870903015, "global_step": 155560, "epoch": 1747, "lr": 3.874262491084752e-05}
{"train_loss": 0.1440781205892563, "global_step": 155561, "epoch": 1747, "lr": 3.874206005991764e-05}
{"train_loss": 0.06972275674343109, "global_step": 155562, "epoch": 1747, "lr": 3.874149521050123e-05}
{"train_loss": 0.15618188679218292, "global_step": 155563, "epoch": 1747, "lr": 3.87409303625984e-05}
{"train_loss": 0.06365068256855011, "global_step": 155564, "epoch": 1747, "lr": 3.874036551620921e-05}
{"train_loss": 0.07962340861558914, "global_step": 155565, "epoch": 1747, "lr": 3.8739800671333755e-05}
{"train_loss": 0.10092073678970337, "global_step": 155566, "epoch": 1747, "lr": 3.8739235827972076e-05}
{"train_loss": 0.16967271268367767, "global_step": 155567, "epoch": 1747, "lr": 3.873867098612429e-05}
{"train_loss": 0.09398511797189713, "global_step": 155568, "epoch": 1747, "lr": 3.873810614579043e-05}
{"train_loss": 0.12868882715702057, "global_step": 155569, "epoch": 1747, "lr": 3.873754130697059e-05}
{"train_loss": 0.08207284659147263, "global_step": 155570, "epoch": 1747, "lr": 3.8736976469664875e-05}
{"train_loss": 0.1001421416575989, "global_step": 155571, "epoch": 1747, "lr": 3.873641163387333e-05, "val_loss": 6.437779426574707}
{"train_loss": 0.07500304281711578, "global_step": 155572, "epoch": 1748, "lr": 3.8735846799596034e-05}
{"train_loss": 0.1573439985513687, "global_step": 155573, "epoch": 1748, "lr": 3.873528196683307e-05}
{"train_loss": 0.0562933124601841, "global_step": 155574, "epoch": 1748, "lr": 3.873471713558451e-05}
{"train_loss": 0.08344196528196335, "global_step": 155575, "epoch": 1748, "lr": 3.8734152305850425e-05}
{"train_loss": 0.06771068274974823, "global_step": 155576, "epoch": 1748, "lr": 3.8733587477630904e-05}
{"train_loss": 0.07635462284088135, "global_step": 155577, "epoch": 1748, "lr": 3.8733022650926e-05}
{"train_loss": 0.08949936926364899, "global_step": 155578, "epoch": 1748, "lr": 3.8732457825735826e-05}
{"train_loss": 0.07735347747802734, "global_step": 155579, "epoch": 1748, "lr": 3.8731893002060406e-05}
{"train_loss": 0.11467617750167847, "global_step": 155580, "epoch": 1748, "lr": 3.8731328179899874e-05}
{"train_loss": 0.06987170875072479, "global_step": 155581, "epoch": 1748, "lr": 3.873076335925425e-05}
{"train_loss": 0.12888020277023315, "global_step": 155582, "epoch": 1748, "lr": 3.873019854012366e-05}
{"train_loss": 0.1285986304283142, "global_step": 155583, "epoch": 1748, "lr": 3.8729633722508146e-05}
{"train_loss": 0.11117181181907654, "global_step": 155584, "epoch": 1748, "lr": 3.872906890640781e-05}
{"train_loss": 0.10281869024038315, "global_step": 155585, "epoch": 1748, "lr": 3.872850409182269e-05}
{"train_loss": 0.042036354541778564, "global_step": 155586, "epoch": 1748, "lr": 3.872793927875291e-05}
{"train_loss": 0.1261863261461258, "global_step": 155587, "epoch": 1748, "lr": 3.87273744671985e-05}
{"train_loss": 0.12789656221866608, "global_step": 155588, "epoch": 1748, "lr": 3.8726809657159556e-05}
{"train_loss": 0.05982619896531105, "global_step": 155589, "epoch": 1748, "lr": 3.872624484863617e-05}
{"train_loss": 0.09675608575344086, "global_step": 155590, "epoch": 1748, "lr": 3.87256800416284e-05}
{"train_loss": 0.16055195033550262, "global_step": 155591, "epoch": 1748, "lr": 3.872511523613633e-05}
{"train_loss": 0.12181563675403595, "global_step": 155592, "epoch": 1748, "lr": 3.872455043216002e-05}
{"train_loss": 0.09042380750179291, "global_step": 155593, "epoch": 1748, "lr": 3.872398562969956e-05}
{"train_loss": 0.09667155891656876, "global_step": 155594, "epoch": 1748, "lr": 3.8723420828755024e-05}
{"train_loss": 0.10059650242328644, "global_step": 155595, "epoch": 1748, "lr": 3.8722856029326484e-05}
{"train_loss": 0.10649988055229187, "global_step": 155596, "epoch": 1748, "lr": 3.872229123141402e-05}
{"train_loss": 0.08854474127292633, "global_step": 155597, "epoch": 1748, "lr": 3.872172643501772e-05}
{"train_loss": 0.07214395701885223, "global_step": 155598, "epoch": 1748, "lr": 3.8721161640137626e-05}
{"train_loss": 0.13452211022377014, "global_step": 155599, "epoch": 1748, "lr": 3.872059684677385e-05}
{"train_loss": 0.09946280717849731, "global_step": 155600, "epoch": 1748, "lr": 3.872003205492644e-05}
{"train_loss": 0.08069045841693878, "global_step": 155601, "epoch": 1748, "lr": 3.871946726459549e-05}
{"train_loss": 0.07892151921987534, "global_step": 155602, "epoch": 1748, "lr": 3.871890247578106e-05}
{"train_loss": 0.06421589106321335, "global_step": 155603, "epoch": 1748, "lr": 3.871833768848325e-05}
{"train_loss": 0.08191782981157303, "global_step": 155604, "epoch": 1748, "lr": 3.871777290270211e-05}
{"train_loss": 0.09593784064054489, "global_step": 155605, "epoch": 1748, "lr": 3.871720811843773e-05}
{"train_loss": 0.0682285875082016, "global_step": 155606, "epoch": 1748, "lr": 3.8716643335690194e-05}
{"train_loss": 0.11146179586648941, "global_step": 155607, "epoch": 1748, "lr": 3.871607855445955e-05}
{"train_loss": 0.043897345662117004, "global_step": 155608, "epoch": 1748, "lr": 3.871551377474592e-05}
{"train_loss": 0.0726127028465271, "global_step": 155609, "epoch": 1748, "lr": 3.871494899654932e-05}
{"train_loss": 0.09336134791374207, "global_step": 155610, "epoch": 1748, "lr": 3.8714384219869885e-05}
{"train_loss": 0.12285726517438889, "global_step": 155611, "epoch": 1748, "lr": 3.871381944470765e-05}
{"train_loss": 0.1396237462759018, "global_step": 155612, "epoch": 1748, "lr": 3.87132546710627e-05}
{"train_loss": 0.07600201666355133, "global_step": 155613, "epoch": 1748, "lr": 3.871268989893513e-05}
{"train_loss": 0.11139655113220215, "global_step": 155614, "epoch": 1748, "lr": 3.8712125128325e-05}
{"train_loss": 0.07101896405220032, "global_step": 155615, "epoch": 1748, "lr": 3.8711560359232366e-05}
{"train_loss": 0.11371219158172607, "global_step": 155616, "epoch": 1748, "lr": 3.8710995591657354e-05}
{"train_loss": 0.10938483476638794, "global_step": 155617, "epoch": 1748, "lr": 3.871043082559999e-05}
{"train_loss": 0.09974252432584763, "global_step": 155618, "epoch": 1748, "lr": 3.870986606106038e-05}
{"train_loss": 0.07325530797243118, "global_step": 155619, "epoch": 1748, "lr": 3.870930129803858e-05}
{"train_loss": 0.08342406153678894, "global_step": 155620, "epoch": 1748, "lr": 3.8708736536534693e-05}
{"train_loss": 0.06411798298358917, "global_step": 155621, "epoch": 1748, "lr": 3.8708171776548765e-05}
{"train_loss": 0.10388623178005219, "global_step": 155622, "epoch": 1748, "lr": 3.87076070180809e-05}
{"train_loss": 0.05718132108449936, "global_step": 155623, "epoch": 1748, "lr": 3.870704226113114e-05}
{"train_loss": 0.09512721747159958, "global_step": 155624, "epoch": 1748, "lr": 3.870647750569958e-05}
{"train_loss": 0.08339428901672363, "global_step": 155625, "epoch": 1748, "lr": 3.870591275178632e-05}
{"train_loss": 0.17902928590774536, "global_step": 155626, "epoch": 1748, "lr": 3.8705347999391385e-05}
{"train_loss": 0.1440967619419098, "global_step": 155627, "epoch": 1748, "lr": 3.870478324851491e-05}
{"train_loss": 0.11536283791065216, "global_step": 155628, "epoch": 1748, "lr": 3.8704218499156906e-05}
{"train_loss": 0.08244238793849945, "global_step": 155629, "epoch": 1748, "lr": 3.870365375131752e-05}
{"train_loss": 0.0822671502828598, "global_step": 155630, "epoch": 1748, "lr": 3.8703089004996765e-05}
{"train_loss": 0.08794823288917542, "global_step": 155631, "epoch": 1748, "lr": 3.870252426019474e-05}
{"train_loss": 0.06425593048334122, "global_step": 155632, "epoch": 1748, "lr": 3.870195951691153e-05}
{"train_loss": 0.08773743361234665, "global_step": 155633, "epoch": 1748, "lr": 3.8701394775147205e-05}
{"train_loss": 0.08752468228340149, "global_step": 155634, "epoch": 1748, "lr": 3.870083003490184e-05}
{"train_loss": 0.08796758204698563, "global_step": 155635, "epoch": 1748, "lr": 3.8700265296175516e-05}
{"train_loss": 0.12285546213388443, "global_step": 155636, "epoch": 1748, "lr": 3.869970055896828e-05}
{"train_loss": 0.06287689507007599, "global_step": 155637, "epoch": 1748, "lr": 3.869913582328026e-05}
{"train_loss": 0.10381252318620682, "global_step": 155638, "epoch": 1748, "lr": 3.869857108911148e-05}
{"train_loss": 0.048901837319135666, "global_step": 155639, "epoch": 1748, "lr": 3.8698006356462057e-05}
{"train_loss": 0.08641406148672104, "global_step": 155640, "epoch": 1748, "lr": 3.869744162533203e-05}
{"train_loss": 0.12272408604621887, "global_step": 155641, "epoch": 1748, "lr": 3.8696876895721526e-05}
{"train_loss": 0.0677911713719368, "global_step": 155642, "epoch": 1748, "lr": 3.8696312167630555e-05}
{"train_loss": 0.07021500915288925, "global_step": 155643, "epoch": 1748, "lr": 3.869574744105923e-05}
{"train_loss": 0.0798587054014206, "global_step": 155644, "epoch": 1748, "lr": 3.869518271600765e-05}
{"train_loss": 0.10047681629657745, "global_step": 155645, "epoch": 1748, "lr": 3.8694617992475837e-05}
{"train_loss": 0.13671281933784485, "global_step": 155646, "epoch": 1748, "lr": 3.869405327046391e-05}
{"train_loss": 0.1097952127456665, "global_step": 155647, "epoch": 1748, "lr": 3.8693488549971925e-05}
{"train_loss": 0.11457720398902893, "global_step": 155648, "epoch": 1748, "lr": 3.8692923830999975e-05}
{"train_loss": 0.10415391623973846, "global_step": 155649, "epoch": 1748, "lr": 3.869235911354812e-05}
{"train_loss": 0.06202681362628937, "global_step": 155650, "epoch": 1748, "lr": 3.8691794397616433e-05}
{"train_loss": 0.043186865746974945, "global_step": 155651, "epoch": 1748, "lr": 3.869122968320499e-05}
{"train_loss": 0.1369183510541916, "global_step": 155652, "epoch": 1748, "lr": 3.8690664970313883e-05}
{"train_loss": 0.14739343523979187, "global_step": 155653, "epoch": 1748, "lr": 3.8690100258943174e-05}
{"train_loss": 0.10634548962116241, "global_step": 155654, "epoch": 1748, "lr": 3.868953554909296e-05}
{"train_loss": 0.09804036468267441, "global_step": 155655, "epoch": 1748, "lr": 3.868897084076327e-05}
{"train_loss": 0.0991898849606514, "global_step": 155656, "epoch": 1748, "lr": 3.868840613395424e-05}
{"train_loss": 0.0908280536532402, "global_step": 155657, "epoch": 1748, "lr": 3.868784142866589e-05}
{"train_loss": 0.1741722822189331, "global_step": 155658, "epoch": 1748, "lr": 3.8687276724898344e-05}
{"train_loss": 0.08743459731340408, "global_step": 155659, "epoch": 1748, "lr": 3.8686712022651636e-05}
{"train_loss": 0.09571834624315916, "global_step": 155660, "epoch": 1748, "lr": 3.868614732192586e-05, "val_loss": 6.583466053009033}
{"train_loss": 0.12651902437210083, "global_step": 155661, "epoch": 1749, "lr": 3.868558262272111e-05}
{"train_loss": 0.07723653316497803, "global_step": 155662, "epoch": 1749, "lr": 3.8685017925037424e-05}
{"train_loss": 0.1489545851945877, "global_step": 155663, "epoch": 1749, "lr": 3.8684453228874926e-05}
{"train_loss": 0.1024550125002861, "global_step": 155664, "epoch": 1749, "lr": 3.8683888534233646e-05}
{"train_loss": 0.10296548157930374, "global_step": 155665, "epoch": 1749, "lr": 3.86833238411137e-05}
{"train_loss": 0.1421058475971222, "global_step": 155666, "epoch": 1749, "lr": 3.868275914951512e-05}
{"train_loss": 0.05957449972629547, "global_step": 155667, "epoch": 1749, "lr": 3.868219445943803e-05}
{"train_loss": 0.1112661361694336, "global_step": 155668, "epoch": 1749, "lr": 3.868162977088247e-05}
{"train_loss": 0.08573207259178162, "global_step": 155669, "epoch": 1749, "lr": 3.868106508384852e-05}
{"train_loss": 0.09285751730203629, "global_step": 155670, "epoch": 1749, "lr": 3.868050039833626e-05}
{"train_loss": 0.05391556769609451, "global_step": 155671, "epoch": 1749, "lr": 3.8679935714345776e-05}
{"train_loss": 0.09588733315467834, "global_step": 155672, "epoch": 1749, "lr": 3.867937103187714e-05}
{"train_loss": 0.06453126668930054, "global_step": 155673, "epoch": 1749, "lr": 3.867880635093043e-05}
{"train_loss": 0.09521492570638657, "global_step": 155674, "epoch": 1749, "lr": 3.8678241671505695e-05}
{"train_loss": 0.08840607851743698, "global_step": 155675, "epoch": 1749, "lr": 3.867767699360306e-05}
{"train_loss": 0.08794336020946503, "global_step": 155676, "epoch": 1749, "lr": 3.867711231722254e-05}
{"train_loss": 0.08414293080568314, "global_step": 155677, "epoch": 1749, "lr": 3.867654764236428e-05}
{"train_loss": 0.09666907787322998, "global_step": 155678, "epoch": 1749, "lr": 3.867598296902829e-05}
{"train_loss": 0.10381446778774261, "global_step": 155679, "epoch": 1749, "lr": 3.8675418297214674e-05}
{"train_loss": 0.10795988142490387, "global_step": 155680, "epoch": 1749, "lr": 3.867485362692354e-05}
{"train_loss": 0.14382560551166534, "global_step": 155681, "epoch": 1749, "lr": 3.8674288958154905e-05}
{"train_loss": 0.06200253590941429, "global_step": 155682, "epoch": 1749, "lr": 3.86737242909089e-05}
{"train_loss": 0.17378853261470795, "global_step": 155683, "epoch": 1749, "lr": 3.867315962518555e-05}
{"train_loss": 0.1523556113243103, "global_step": 155684, "epoch": 1749, "lr": 3.867259496098498e-05}
{"train_loss": 0.09575355798006058, "global_step": 155685, "epoch": 1749, "lr": 3.867203029830721e-05}
{"train_loss": 0.10867134481668472, "global_step": 155686, "epoch": 1749, "lr": 3.8671465637152374e-05}
{"train_loss": 0.11257611215114594, "global_step": 155687, "epoch": 1749, "lr": 3.867090097752052e-05}
{"train_loss": 0.12197980284690857, "global_step": 155688, "epoch": 1749, "lr": 3.867033631941171e-05}
{"train_loss": 0.11244159936904907, "global_step": 155689, "epoch": 1749, "lr": 3.866977166282605e-05}
{"train_loss": 0.06237016245722771, "global_step": 155690, "epoch": 1749, "lr": 3.86692070077636e-05}
{"train_loss": 0.08020303398370743, "global_step": 155691, "epoch": 1749, "lr": 3.866864235422443e-05}
{"train_loss": 0.09963401407003403, "global_step": 155692, "epoch": 1749, "lr": 3.866807770220863e-05}
{"train_loss": 0.11558361351490021, "global_step": 155693, "epoch": 1749, "lr": 3.866751305171626e-05}
{"train_loss": 0.11570580303668976, "global_step": 155694, "epoch": 1749, "lr": 3.8666948402747414e-05}
{"train_loss": 0.08057542145252228, "global_step": 155695, "epoch": 1749, "lr": 3.8666383755302145e-05}
{"train_loss": 0.16796711087226868, "global_step": 155696, "epoch": 1749, "lr": 3.866581910938054e-05}
{"train_loss": 0.15264356136322021, "global_step": 155697, "epoch": 1749, "lr": 3.86652544649827e-05}
{"train_loss": 0.08984354138374329, "global_step": 155698, "epoch": 1749, "lr": 3.8664689822108656e-05}
{"train_loss": 0.10178179293870926, "global_step": 155699, "epoch": 1749, "lr": 3.866412518075853e-05}
{"train_loss": 0.04420767351984978, "global_step": 155700, "epoch": 1749, "lr": 3.866356054093235e-05}
{"train_loss": 0.09443177282810211, "global_step": 155701, "epoch": 1749, "lr": 3.866299590263024e-05}
{"train_loss": 0.10320164263248444, "global_step": 155702, "epoch": 1749, "lr": 3.866243126585222e-05}
{"train_loss": 0.07862124592065811, "global_step": 155703, "epoch": 1749, "lr": 3.866186663059843e-05}
{"train_loss": 0.06533417850732803, "global_step": 155704, "epoch": 1749, "lr": 3.866130199686889e-05}
{"train_loss": 0.12280969321727753, "global_step": 155705, "epoch": 1749, "lr": 3.8660737364663725e-05}
{"train_loss": 0.061996277421712875, "global_step": 155706, "epoch": 1749, "lr": 3.8660172733982966e-05}
{"train_loss": 0.1519603133201599, "global_step": 155707, "epoch": 1749, "lr": 3.865960810482672e-05}
{"train_loss": 0.10855995118618011, "global_step": 155708, "epoch": 1749, "lr": 3.865904347719505e-05}
{"train_loss": 0.1504191905260086, "global_step": 155709, "epoch": 1749, "lr": 3.865847885108802e-05}
{"train_loss": 0.09420724213123322, "global_step": 155710, "epoch": 1749, "lr": 3.865791422650573e-05}
{"train_loss": 0.06912670284509659, "global_step": 155711, "epoch": 1749, "lr": 3.865734960344825e-05}
{"train_loss": 0.11146999150514603, "global_step": 155712, "epoch": 1749, "lr": 3.865678498191564e-05}
{"train_loss": 0.07493200153112411, "global_step": 155713, "epoch": 1749, "lr": 3.8656220361908e-05}
{"train_loss": 0.10079703480005264, "global_step": 155714, "epoch": 1749, "lr": 3.865565574342538e-05}
{"train_loss": 0.08804918825626373, "global_step": 155715, "epoch": 1749, "lr": 3.865509112646786e-05}
{"train_loss": 0.09889168292284012, "global_step": 155716, "epoch": 1749, "lr": 3.865452651103555e-05}
{"train_loss": 0.14831681549549103, "global_step": 155717, "epoch": 1749, "lr": 3.8653961897128474e-05}
{"train_loss": 0.05702665075659752, "global_step": 155718, "epoch": 1749, "lr": 3.8653397284746764e-05}
{"train_loss": 0.0817047730088234, "global_step": 155719, "epoch": 1749, "lr": 3.865283267389044e-05}
{"train_loss": 0.13354970514774323, "global_step": 155720, "epoch": 1749, "lr": 3.8652268064559626e-05}
{"train_loss": 0.1418856382369995, "global_step": 155721, "epoch": 1749, "lr": 3.8651703456754354e-05}
{"train_loss": 0.08092664182186127, "global_step": 155722, "epoch": 1749, "lr": 3.8651138850474744e-05}
{"train_loss": 0.04838700592517853, "global_step": 155723, "epoch": 1749, "lr": 3.865057424572083e-05}
{"train_loss": 0.08705569803714752, "global_step": 155724, "epoch": 1749, "lr": 3.865000964249273e-05}
{"train_loss": 0.11999605596065521, "global_step": 155725, "epoch": 1749, "lr": 3.864944504079048e-05}
{"train_loss": 0.09399142116308212, "global_step": 155726, "epoch": 1749, "lr": 3.8648880440614187e-05}
{"train_loss": 0.07301390171051025, "global_step": 155727, "epoch": 1749, "lr": 3.8648315841963904e-05}
{"train_loss": 0.1295413225889206, "global_step": 155728, "epoch": 1749, "lr": 3.864775124483972e-05}
{"train_loss": 0.1191665306687355, "global_step": 155729, "epoch": 1749, "lr": 3.864718664924171e-05}
{"train_loss": 0.1366453319787979, "global_step": 155730, "epoch": 1749, "lr": 3.864662205516994e-05}
{"train_loss": 0.045469172298908234, "global_step": 155731, "epoch": 1749, "lr": 3.86460574626245e-05}
{"train_loss": 0.1687014400959015, "global_step": 155732, "epoch": 1749, "lr": 3.864549287160544e-05}
{"train_loss": 0.12467389553785324, "global_step": 155733, "epoch": 1749, "lr": 3.864492828211289e-05}
{"train_loss": 0.14763130247592926, "global_step": 155734, "epoch": 1749, "lr": 3.864436369414686e-05}
{"train_loss": 0.037583351135253906, "global_step": 155735, "epoch": 1749, "lr": 3.864379910770748e-05}
{"train_loss": 0.14324480295181274, "global_step": 155736, "epoch": 1749, "lr": 3.864323452279478e-05}
{"train_loss": 0.10003581643104553, "global_step": 155737, "epoch": 1749, "lr": 3.864266993940888e-05}
{"train_loss": 0.09265168011188507, "global_step": 155738, "epoch": 1749, "lr": 3.864210535754982e-05}
{"train_loss": 0.09996049851179123, "global_step": 155739, "epoch": 1749, "lr": 3.86415407772177e-05}
{"train_loss": 0.07053503394126892, "global_step": 155740, "epoch": 1749, "lr": 3.8640976198412575e-05}
{"train_loss": 0.06826882064342499, "global_step": 155741, "epoch": 1749, "lr": 3.864041162113455e-05}
{"train_loss": 0.11058831959962845, "global_step": 155742, "epoch": 1749, "lr": 3.8639847045383664e-05}
{"train_loss": 0.08260820806026459, "global_step": 155743, "epoch": 1749, "lr": 3.863928247116003e-05}
{"train_loss": 0.0834517776966095, "global_step": 155744, "epoch": 1749, "lr": 3.863871789846369e-05}
{"train_loss": 0.09592051059007645, "global_step": 155745, "epoch": 1749, "lr": 3.8638153327294754e-05}
{"train_loss": 0.07730001211166382, "global_step": 155746, "epoch": 1749, "lr": 3.863758875765326e-05}
{"train_loss": 0.10511799156665802, "global_step": 155747, "epoch": 1749, "lr": 3.8637024189539324e-05}
{"train_loss": 0.08025313168764114, "global_step": 155748, "epoch": 1749, "lr": 3.8636459622952986e-05}
{"train_loss": 0.10097516215081966, "global_step": 155749, "epoch": 1749, "lr": 3.863589505789434e-05, "val_loss": 6.185537815093994}
{"train_loss": 0.06652604043483734, "global_step": 155750, "epoch": 1750, "lr": 3.863533049436346e-05}
{"train_loss": 0.0582093708217144, "global_step": 155751, "epoch": 1750, "lr": 3.863476593236042e-05}
{"train_loss": 0.08578810095787048, "global_step": 155752, "epoch": 1750, "lr": 3.86342013718853e-05}
{"train_loss": 0.14889666438102722, "global_step": 155753, "epoch": 1750, "lr": 3.863363681293817e-05}
{"train_loss": 0.08098985254764557, "global_step": 155754, "epoch": 1750, "lr": 3.863307225551912e-05}
{"train_loss": 0.11391153186559677, "global_step": 155755, "epoch": 1750, "lr": 3.8632507699628193e-05}
{"train_loss": 0.0670444518327713, "global_step": 155756, "epoch": 1750, "lr": 3.8631943145265516e-05}
{"train_loss": 0.06267012655735016, "global_step": 155757, "epoch": 1750, "lr": 3.8631378592431104e-05}
{"train_loss": 0.09909863024950027, "global_step": 155758, "epoch": 1750, "lr": 3.86308140411251e-05}
{"train_loss": 0.10434575378894806, "global_step": 155759, "epoch": 1750, "lr": 3.863024949134751e-05}
{"train_loss": 0.10300327092409134, "global_step": 155760, "epoch": 1750, "lr": 3.8629684943098474e-05}
{"train_loss": 0.07406371086835861, "global_step": 155761, "epoch": 1750, "lr": 3.862912039637801e-05}
{"train_loss": 0.10195830464363098, "global_step": 155762, "epoch": 1750, "lr": 3.862855585118624e-05}
{"train_loss": 0.11432088166475296, "global_step": 155763, "epoch": 1750, "lr": 3.862799130752321e-05}
{"train_loss": 0.08651205152273178, "global_step": 155764, "epoch": 1750, "lr": 3.862742676538903e-05}
{"train_loss": 0.05956505611538887, "global_step": 155765, "epoch": 1750, "lr": 3.862686222478374e-05}
{"train_loss": 0.054710280150175095, "global_step": 155766, "epoch": 1750, "lr": 3.862629768570743e-05}
{"train_loss": 0.1135181337594986, "global_step": 155767, "epoch": 1750, "lr": 3.862573314816017e-05}
{"train_loss": 0.07132315635681152, "global_step": 155768, "epoch": 1750, "lr": 3.862516861214205e-05}
{"train_loss": 0.10889317095279694, "global_step": 155769, "epoch": 1750, "lr": 3.8624604077653135e-05}
{"train_loss": 0.1346452385187149, "global_step": 155770, "epoch": 1750, "lr": 3.862403954469349e-05}
{"train_loss": 0.11862476170063019, "global_step": 155771, "epoch": 1750, "lr": 3.8623475013263234e-05}
{"train_loss": 0.09044855087995529, "global_step": 155772, "epoch": 1750, "lr": 3.862291048336238e-05}
{"train_loss": 0.0825079157948494, "global_step": 155773, "epoch": 1750, "lr": 3.8622345954991065e-05}
{"train_loss": 0.13537268340587616, "global_step": 155774, "epoch": 1750, "lr": 3.8621781428149315e-05}
{"train_loss": 0.1238054633140564, "global_step": 155775, "epoch": 1750, "lr": 3.8621216902837246e-05}
{"train_loss": 0.07124050706624985, "global_step": 155776, "epoch": 1750, "lr": 3.86206523790549e-05}
{"train_loss": 0.11575734615325928, "global_step": 155777, "epoch": 1750, "lr": 3.8620087856802386e-05}
{"train_loss": 0.07874810695648193, "global_step": 155778, "epoch": 1750, "lr": 3.861952333607974e-05}
{"train_loss": 0.10362492501735687, "global_step": 155779, "epoch": 1750, "lr": 3.861895881688709e-05}
{"train_loss": 0.06765501201152802, "global_step": 155780, "epoch": 1750, "lr": 3.8618394299224444e-05}
{"train_loss": 0.10111692547798157, "global_step": 155781, "epoch": 1750, "lr": 3.8617829783091956e-05}
{"train_loss": 0.10562480986118317, "global_step": 155782, "epoch": 1750, "lr": 3.861726526848963e-05}
{"train_loss": 0.10258593410253525, "global_step": 155783, "epoch": 1750, "lr": 3.86167007554176e-05}
{"train_loss": 0.11292044818401337, "global_step": 155784, "epoch": 1750, "lr": 3.86161362438759e-05}
{"train_loss": 0.07993284612894058, "global_step": 155785, "epoch": 1750, "lr": 3.8615571733864627e-05}
{"train_loss": 0.07259175926446915, "global_step": 155786, "epoch": 1750, "lr": 3.861500722538385e-05}
{"train_loss": 0.08350574225187302, "global_step": 155787, "epoch": 1750, "lr": 3.861444271843363e-05}
{"train_loss": 0.10471373051404953, "global_step": 155788, "epoch": 1750, "lr": 3.861387821301409e-05}
{"train_loss": 0.06067802011966705, "global_step": 155789, "epoch": 1750, "lr": 3.861331370912525e-05}
{"train_loss": 0.08962336927652359, "global_step": 155790, "epoch": 1750, "lr": 3.861274920676723e-05}
{"train_loss": 0.07064638286828995, "global_step": 155791, "epoch": 1750, "lr": 3.8612184705940066e-05}
{"train_loss": 0.08872643113136292, "global_step": 155792, "epoch": 1750, "lr": 3.861162020664388e-05}
{"train_loss": 0.14723944664001465, "global_step": 155793, "epoch": 1750, "lr": 3.86110557088787e-05}
{"train_loss": 0.06958308815956116, "global_step": 155794, "epoch": 1750, "lr": 3.861049121264465e-05}
{"train_loss": 0.10720263421535492, "global_step": 155795, "epoch": 1750, "lr": 3.8609926717941754e-05}
{"train_loss": 0.11730263382196426, "global_step": 155796, "epoch": 1750, "lr": 3.860936222477013e-05}
{"train_loss": 0.14922943711280823, "global_step": 155797, "epoch": 1750, "lr": 3.860879773312983e-05}
{"train_loss": 0.1525220423936844, "global_step": 155798, "epoch": 1750, "lr": 3.8608233243020953e-05}
{"train_loss": 0.06635549664497375, "global_step": 155799, "epoch": 1750, "lr": 3.8607668754443544e-05}
{"train_loss": 0.08058629930019379, "global_step": 155800, "epoch": 1750, "lr": 3.8607104267397706e-05}
{"train_loss": 0.12682363390922546, "global_step": 155801, "epoch": 1750, "lr": 3.860653978188349e-05}
{"train_loss": 0.07791705429553986, "global_step": 155802, "epoch": 1750, "lr": 3.860597529790101e-05}
{"train_loss": 0.062100764364004135, "global_step": 155803, "epoch": 1750, "lr": 3.86054108154503e-05}
{"train_loss": 0.029370438307523727, "global_step": 155804, "epoch": 1750, "lr": 3.860484633453146e-05}
{"train_loss": 0.10210511088371277, "global_step": 155805, "epoch": 1750, "lr": 3.8604281855144555e-05}
{"train_loss": 0.12503576278686523, "global_step": 155806, "epoch": 1750, "lr": 3.8603717377289655e-05}
{"train_loss": 0.09518817812204361, "global_step": 155807, "epoch": 1750, "lr": 3.8603152900966874e-05}
{"train_loss": 0.056664515286684036, "global_step": 155808, "epoch": 1750, "lr": 3.860258842617623e-05}
{"train_loss": 0.07338391989469528, "global_step": 155809, "epoch": 1750, "lr": 3.8602023952917856e-05}
{"train_loss": 0.15134581923484802, "global_step": 155810, "epoch": 1750, "lr": 3.8601459481191775e-05}
{"train_loss": 0.09824799001216888, "global_step": 155811, "epoch": 1750, "lr": 3.860089501099812e-05}
{"train_loss": 0.08609361946582794, "global_step": 155812, "epoch": 1750, "lr": 3.8600330542336906e-05}
{"train_loss": 0.16151751577854156, "global_step": 155813, "epoch": 1750, "lr": 3.8599766075208255e-05}
{"train_loss": 0.14991319179534912, "global_step": 155814, "epoch": 1750, "lr": 3.859920160961221e-05}
{"train_loss": 0.11630763858556747, "global_step": 155815, "epoch": 1750, "lr": 3.859863714554889e-05}
{"train_loss": 0.15677623450756073, "global_step": 155816, "epoch": 1750, "lr": 3.8598072683018314e-05}
{"train_loss": 0.10277428478002548, "global_step": 155817, "epoch": 1750, "lr": 3.8597508222020617e-05}
{"train_loss": 0.10537251830101013, "global_step": 155818, "epoch": 1750, "lr": 3.8596943762555825e-05}
{"train_loss": 0.07063379138708115, "global_step": 155819, "epoch": 1750, "lr": 3.8596379304624054e-05}
{"train_loss": 0.09818029403686523, "global_step": 155820, "epoch": 1750, "lr": 3.859581484822534e-05}
{"train_loss": 0.0887431725859642, "global_step": 155821, "epoch": 1750, "lr": 3.859525039335979e-05}
{"train_loss": 0.07243058830499649, "global_step": 155822, "epoch": 1750, "lr": 3.8594685940027465e-05}
{"train_loss": 0.08829609304666519, "global_step": 155823, "epoch": 1750, "lr": 3.859412148822845e-05}
{"train_loss": 0.07082098722457886, "global_step": 155824, "epoch": 1750, "lr": 3.859355703796282e-05}
{"train_loss": 0.05767044052481651, "global_step": 155825, "epoch": 1750, "lr": 3.8592992589230636e-05}
{"train_loss": 0.10980571806430817, "global_step": 155826, "epoch": 1750, "lr": 3.8592428142032006e-05}
{"train_loss": 0.08232564479112625, "global_step": 155827, "epoch": 1750, "lr": 3.859186369636696e-05}
{"train_loss": 0.08665274828672409, "global_step": 155828, "epoch": 1750, "lr": 3.859129925223562e-05}
{"train_loss": 0.11768355220556259, "global_step": 155829, "epoch": 1750, "lr": 3.859073480963802e-05}
{"train_loss": 0.10535410046577454, "global_step": 155830, "epoch": 1750, "lr": 3.859017036857428e-05}
{"train_loss": 0.08238770067691803, "global_step": 155831, "epoch": 1750, "lr": 3.8589605929044426e-05}
{"train_loss": 0.06740189343690872, "global_step": 155832, "epoch": 1750, "lr": 3.8589041491048586e-05}
{"train_loss": 0.08640497922897339, "global_step": 155833, "epoch": 1750, "lr": 3.858847705458679e-05}
{"train_loss": 0.12433342635631561, "global_step": 155834, "epoch": 1750, "lr": 3.858791261965915e-05}
{"train_loss": 0.05781923606991768, "global_step": 155835, "epoch": 1750, "lr": 3.858734818626571e-05}
{"train_loss": 0.0816953182220459, "global_step": 155836, "epoch": 1750, "lr": 3.858678375440658e-05}
{"train_loss": 0.06669007241725922, "global_step": 155837, "epoch": 1750, "lr": 3.85862193240818e-05}
{"train_loss": 0.09493085457367843, "global_step": 155838, "epoch": 1750, "lr": 3.858565489529148e-05, "train/sim_max_reward_0": 0.5890932534024136, "train/sim_max_reward_1": 0.9980358941584758, "train/sim_max_reward_2": 0.42995628200739017, "train/sim_max_reward_3": 0.9988327637135781, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.6114512395329486, "test/sim_max_reward_4500000": 1.0, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9777951261128816, "test/sim_max_reward_4500003": 0.52630922800554, "test/sim_max_reward_4500004": 0.4540259935463356, "test/sim_max_reward_4500005": 0.34793695380419504, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9949422052758122, "test/sim_max_reward_4500008": 0.33974940053465785, "test/sim_max_reward_4500009": 0.2945177642204615, "test/sim_max_reward_4500010": 0.9909779013786447, "test/sim_max_reward_4500011": 0.9915320946582457, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.15212050373073882, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.8499513919906888, "test/sim_max_reward_4500016": 0.8488170591533126, "test/sim_max_reward_4500017": 0.18613864429731453, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9038337038977827, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9108471052162204, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6555468044347638, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9667236337545513, "test/sim_max_reward_4500026": 0.9650289845073476, "test/sim_max_reward_4500027": 0.977804136448476, "test/sim_max_reward_4500028": 0.9979562151324369, "test/sim_max_reward_4500029": 0.36150721750158166, "test/sim_max_reward_4500030": 0.41064278443054925, "test/sim_max_reward_4500031": 0.4507694508082629, "test/sim_max_reward_4500032": 0.5742454456132878, "test/sim_max_reward_4500033": 0.3827159328288871, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.9519351128035934, "test/sim_max_reward_4500036": 0.9970011654882476, "test/sim_max_reward_4500037": 0.9906398218607539, "test/sim_max_reward_4500038": 0.1631011866323135, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.589278682479054, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.7051985317161523, "test/sim_max_reward_4500044": 0.6903602046795541, "test/sim_max_reward_4500045": 0.6486850469518928, "test/sim_max_reward_4500046": 0.5358687578511272, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9144346660567174, "test/sim_max_reward_4500049": 0.972305859711075, "train/mean_score": 0.7712282388024677, "test/mean_score": 0.6534248943502691, "val_loss": 6.431739807128906, "train_action_mse_error": 20.836151123046875}
{"train_loss": 0.07150673866271973, "global_step": 155839, "epoch": 1751, "lr": 3.858509046803566e-05}
{"train_loss": 0.1272805780172348, "global_step": 155840, "epoch": 1751, "lr": 3.858452604231446e-05}
{"train_loss": 0.049915194511413574, "global_step": 155841, "epoch": 1751, "lr": 3.8583961618127906e-05}
{"train_loss": 0.12817983329296112, "global_step": 155842, "epoch": 1751, "lr": 3.8583397195476115e-05}
{"train_loss": 0.12528376281261444, "global_step": 155843, "epoch": 1751, "lr": 3.8582832774359154e-05}
{"train_loss": 0.13111260533332825, "global_step": 155844, "epoch": 1751, "lr": 3.858226835477707e-05}
{"train_loss": 0.08647605031728745, "global_step": 155845, "epoch": 1751, "lr": 3.858170393672999e-05}
{"train_loss": 0.09798764437437057, "global_step": 155846, "epoch": 1751, "lr": 3.858113952021793e-05}
{"train_loss": 0.0753161683678627, "global_step": 155847, "epoch": 1751, "lr": 3.858057510524102e-05}
{"train_loss": 0.1007741317152977, "global_step": 155848, "epoch": 1751, "lr": 3.858001069179929e-05}
{"train_loss": 0.06487661600112915, "global_step": 155849, "epoch": 1751, "lr": 3.857944627989286e-05}
{"train_loss": 0.11241964250802994, "global_step": 155850, "epoch": 1751, "lr": 3.8578881869521766e-05}
{"train_loss": 0.10603901743888855, "global_step": 155851, "epoch": 1751, "lr": 3.857831746068612e-05}
{"train_loss": 0.10208405554294586, "global_step": 155852, "epoch": 1751, "lr": 3.857775305338596e-05}
{"train_loss": 0.08491767197847366, "global_step": 155853, "epoch": 1751, "lr": 3.857718864762141e-05}
{"train_loss": 0.10127907991409302, "global_step": 155854, "epoch": 1751, "lr": 3.857662424339249e-05}
{"train_loss": 0.06988193839788437, "global_step": 155855, "epoch": 1751, "lr": 3.857605984069932e-05}
{"train_loss": 0.12865625321865082, "global_step": 155856, "epoch": 1751, "lr": 3.857549543954195e-05}
{"train_loss": 0.15344972908496857, "global_step": 155857, "epoch": 1751, "lr": 3.8574931039920473e-05}
{"train_loss": 0.13739781081676483, "global_step": 155858, "epoch": 1751, "lr": 3.857436664183495e-05}
{"train_loss": 0.05469591170549393, "global_step": 155859, "epoch": 1751, "lr": 3.857380224528546e-05}
{"train_loss": 0.08264835923910141, "global_step": 155860, "epoch": 1751, "lr": 3.8573237850272096e-05}
{"train_loss": 0.16941452026367188, "global_step": 155861, "epoch": 1751, "lr": 3.857267345679492e-05}
{"train_loss": 0.10033436119556427, "global_step": 155862, "epoch": 1751, "lr": 3.8572109064854014e-05}
{"train_loss": 0.12080858647823334, "global_step": 155863, "epoch": 1751, "lr": 3.857154467444943e-05}
{"train_loss": 0.13929200172424316, "global_step": 155864, "epoch": 1751, "lr": 3.8570980285581284e-05}
{"train_loss": 0.06106209009885788, "global_step": 155865, "epoch": 1751, "lr": 3.85704158982496e-05}
{"train_loss": 0.1510208696126938, "global_step": 155866, "epoch": 1751, "lr": 3.856985151245451e-05}
{"train_loss": 0.08228448778390884, "global_step": 155867, "epoch": 1751, "lr": 3.856928712819605e-05}
{"train_loss": 0.05071040987968445, "global_step": 155868, "epoch": 1751, "lr": 3.856872274547433e-05}
{"train_loss": 0.0899280309677124, "global_step": 155869, "epoch": 1751, "lr": 3.856815836428938e-05}
{"train_loss": 0.16504572331905365, "global_step": 155870, "epoch": 1751, "lr": 3.8567593984641324e-05}
{"train_loss": 0.07994955033063889, "global_step": 155871, "epoch": 1751, "lr": 3.85670296065302e-05}
{"train_loss": 0.07374552637338638, "global_step": 155872, "epoch": 1751, "lr": 3.856646522995611e-05}
{"train_loss": 0.10933977365493774, "global_step": 155873, "epoch": 1751, "lr": 3.8565900854919105e-05}
{"train_loss": 0.05389109253883362, "global_step": 155874, "epoch": 1751, "lr": 3.85653364814193e-05}
{"train_loss": 0.06943317502737045, "global_step": 155875, "epoch": 1751, "lr": 3.856477210945672e-05}
{"train_loss": 0.11922827363014221, "global_step": 155876, "epoch": 1751, "lr": 3.856420773903149e-05}
{"train_loss": 0.20801058411598206, "global_step": 155877, "epoch": 1751, "lr": 3.8563643370143636e-05}
{"train_loss": 0.07766677439212799, "global_step": 155878, "epoch": 1751, "lr": 3.856307900279328e-05}
{"train_loss": 0.15585705637931824, "global_step": 155879, "epoch": 1751, "lr": 3.856251463698047e-05}
{"train_loss": 0.10015468299388885, "global_step": 155880, "epoch": 1751, "lr": 3.85619502727053e-05}
{"train_loss": 0.13926556706428528, "global_step": 155881, "epoch": 1751, "lr": 3.856138590996784e-05}
{"train_loss": 0.10651805996894836, "global_step": 155882, "epoch": 1751, "lr": 3.8560821548768146e-05}
{"train_loss": 0.0664835199713707, "global_step": 155883, "epoch": 1751, "lr": 3.8560257189106326e-05}
{"train_loss": 0.09574466198682785, "global_step": 155884, "epoch": 1751, "lr": 3.855969283098242e-05}
{"train_loss": 0.11910150945186615, "global_step": 155885, "epoch": 1751, "lr": 3.855912847439655e-05}
{"train_loss": 0.06240154802799225, "global_step": 155886, "epoch": 1751, "lr": 3.855856411934874e-05}
{"train_loss": 0.13549289107322693, "global_step": 155887, "epoch": 1751, "lr": 3.8557999765839116e-05}
{"train_loss": 0.10926328599452972, "global_step": 155888, "epoch": 1751, "lr": 3.8557435413867705e-05}
{"train_loss": 0.15744571387767792, "global_step": 155889, "epoch": 1751, "lr": 3.855687106343463e-05}
{"train_loss": 0.056415632367134094, "global_step": 155890, "epoch": 1751, "lr": 3.855630671453992e-05}
{"train_loss": 0.07736097276210785, "global_step": 155891, "epoch": 1751, "lr": 3.85557423671837e-05}
{"train_loss": 0.09272801876068115, "global_step": 155892, "epoch": 1751, "lr": 3.855517802136599e-05}
{"train_loss": 0.09798358380794525, "global_step": 155893, "epoch": 1751, "lr": 3.8554613677086934e-05}
{"train_loss": 0.06737473607063293, "global_step": 155894, "epoch": 1751, "lr": 3.855404933434654e-05}
{"train_loss": 0.12182949483394623, "global_step": 155895, "epoch": 1751, "lr": 3.855348499314494e-05}
{"train_loss": 0.07259274274110794, "global_step": 155896, "epoch": 1751, "lr": 3.855292065348216e-05}
{"train_loss": 0.05556429177522659, "global_step": 155897, "epoch": 1751, "lr": 3.8552356315358315e-05}
{"train_loss": 0.07906829565763474, "global_step": 155898, "epoch": 1751, "lr": 3.855179197877346e-05}
{"train_loss": 0.10852378606796265, "global_step": 155899, "epoch": 1751, "lr": 3.8551227643727684e-05}
{"train_loss": 0.07861273735761642, "global_step": 155900, "epoch": 1751, "lr": 3.8550663310221055e-05}
{"train_loss": 0.09199224412441254, "global_step": 155901, "epoch": 1751, "lr": 3.855009897825364e-05}
{"train_loss": 0.11881403625011444, "global_step": 155902, "epoch": 1751, "lr": 3.8549534647825544e-05}
{"train_loss": 0.07108353823423386, "global_step": 155903, "epoch": 1751, "lr": 3.85489703189368e-05}
{"train_loss": 0.08419748395681381, "global_step": 155904, "epoch": 1751, "lr": 3.854840599158753e-05}
{"train_loss": 0.07417416572570801, "global_step": 155905, "epoch": 1751, "lr": 3.854784166577776e-05}
{"train_loss": 0.04383791610598564, "global_step": 155906, "epoch": 1751, "lr": 3.854727734150762e-05}
{"train_loss": 0.12006865441799164, "global_step": 155907, "epoch": 1751, "lr": 3.854671301877714e-05}
{"train_loss": 0.11409073323011398, "global_step": 155908, "epoch": 1751, "lr": 3.854614869758643e-05}
{"train_loss": 0.0646304339170456, "global_step": 155909, "epoch": 1751, "lr": 3.854558437793553e-05}
{"train_loss": 0.11079886555671692, "global_step": 155910, "epoch": 1751, "lr": 3.854502005982457e-05}
{"train_loss": 0.10685347765684128, "global_step": 155911, "epoch": 1751, "lr": 3.8544455743253556e-05}
{"train_loss": 0.12375723570585251, "global_step": 155912, "epoch": 1751, "lr": 3.854389142822263e-05}
{"train_loss": 0.08599782735109329, "global_step": 155913, "epoch": 1751, "lr": 3.854332711473181e-05}
{"train_loss": 0.11284026503562927, "global_step": 155914, "epoch": 1751, "lr": 3.854276280278122e-05}
{"train_loss": 0.1702549010515213, "global_step": 155915, "epoch": 1751, "lr": 3.854219849237091e-05}
{"train_loss": 0.08233985304832458, "global_step": 155916, "epoch": 1751, "lr": 3.854163418350096e-05}
{"train_loss": 0.10998599231243134, "global_step": 155917, "epoch": 1751, "lr": 3.854106987617145e-05}
{"train_loss": 0.05271482095122337, "global_step": 155918, "epoch": 1751, "lr": 3.8540505570382454e-05}
{"train_loss": 0.11807815730571747, "global_step": 155919, "epoch": 1751, "lr": 3.853994126613406e-05}
{"train_loss": 0.10661818087100983, "global_step": 155920, "epoch": 1751, "lr": 3.8539376963426306e-05}
{"train_loss": 0.08556368947029114, "global_step": 155921, "epoch": 1751, "lr": 3.8538812662259314e-05}
{"train_loss": 0.09987759590148926, "global_step": 155922, "epoch": 1751, "lr": 3.8538248362633115e-05}
{"train_loss": 0.06701160967350006, "global_step": 155923, "epoch": 1751, "lr": 3.853768406454784e-05}
{"train_loss": 0.08846819400787354, "global_step": 155924, "epoch": 1751, "lr": 3.8537119768003505e-05}
{"train_loss": 0.07941767573356628, "global_step": 155925, "epoch": 1751, "lr": 3.8536555473000237e-05}
{"train_loss": 0.1123310998082161, "global_step": 155926, "epoch": 1751, "lr": 3.8535991179538066e-05}
{"train_loss": 0.09933174552207583, "global_step": 155927, "epoch": 1751, "lr": 3.853542688761712e-05, "val_loss": 6.42123556137085}
{"train_loss": 0.14959150552749634, "global_step": 155928, "epoch": 1752, "lr": 3.853486259723742e-05}
{"train_loss": 0.05360637977719307, "global_step": 155929, "epoch": 1752, "lr": 3.853429830839909e-05}
{"train_loss": 0.06060744449496269, "global_step": 155930, "epoch": 1752, "lr": 3.8533734021102165e-05}
{"train_loss": 0.11738674342632294, "global_step": 155931, "epoch": 1752, "lr": 3.8533169735346755e-05}
{"train_loss": 0.06965531408786774, "global_step": 155932, "epoch": 1752, "lr": 3.8532605451132903e-05}
{"train_loss": 0.05828458070755005, "global_step": 155933, "epoch": 1752, "lr": 3.8532041168460706e-05}
{"train_loss": 0.09195466339588165, "global_step": 155934, "epoch": 1752, "lr": 3.8531476887330245e-05}
{"train_loss": 0.12478853762149811, "global_step": 155935, "epoch": 1752, "lr": 3.8530912607741586e-05}
{"train_loss": 0.13776324689388275, "global_step": 155936, "epoch": 1752, "lr": 3.8530348329694806e-05}
{"train_loss": 0.11764100939035416, "global_step": 155937, "epoch": 1752, "lr": 3.852978405318998e-05}
{"train_loss": 0.09174986928701401, "global_step": 155938, "epoch": 1752, "lr": 3.852921977822718e-05}
{"train_loss": 0.09952647984027863, "global_step": 155939, "epoch": 1752, "lr": 3.8528655504806496e-05}
{"train_loss": 0.07442577928304672, "global_step": 155940, "epoch": 1752, "lr": 3.8528091232928e-05}
{"train_loss": 0.1181560531258583, "global_step": 155941, "epoch": 1752, "lr": 3.852752696259174e-05}
{"train_loss": 0.12135480344295502, "global_step": 155942, "epoch": 1752, "lr": 3.852696269379784e-05}
{"train_loss": 0.09472326934337616, "global_step": 155943, "epoch": 1752, "lr": 3.852639842654632e-05}
{"train_loss": 0.045365143567323685, "global_step": 155944, "epoch": 1752, "lr": 3.8525834160837316e-05}
{"train_loss": 0.15400153398513794, "global_step": 155945, "epoch": 1752, "lr": 3.8525269896670844e-05}
{"train_loss": 0.07746198773384094, "global_step": 155946, "epoch": 1752, "lr": 3.852470563404704e-05}
{"train_loss": 0.06251674145460129, "global_step": 155947, "epoch": 1752, "lr": 3.8524141372965925e-05}
{"train_loss": 0.12269601225852966, "global_step": 155948, "epoch": 1752, "lr": 3.8523577113427614e-05}
{"train_loss": 0.11150330305099487, "global_step": 155949, "epoch": 1752, "lr": 3.8523012855432156e-05}
{"train_loss": 0.051397643983364105, "global_step": 155950, "epoch": 1752, "lr": 3.852244859897964e-05}
{"train_loss": 0.09422536939382553, "global_step": 155951, "epoch": 1752, "lr": 3.8521884344070154e-05}
{"train_loss": 0.142578125, "global_step": 155952, "epoch": 1752, "lr": 3.8521320090703745e-05}
{"train_loss": 0.16778934001922607, "global_step": 155953, "epoch": 1752, "lr": 3.852075583888052e-05}
{"train_loss": 0.11399711668491364, "global_step": 155954, "epoch": 1752, "lr": 3.852019158860053e-05}
{"train_loss": 0.05392890051007271, "global_step": 155955, "epoch": 1752, "lr": 3.851962733986387e-05}
{"train_loss": 0.07761705666780472, "global_step": 155956, "epoch": 1752, "lr": 3.8519063092670596e-05}
{"train_loss": 0.10128394514322281, "global_step": 155957, "epoch": 1752, "lr": 3.85184988470208e-05}
{"train_loss": 0.047053128480911255, "global_step": 155958, "epoch": 1752, "lr": 3.8517934602914554e-05}
{"train_loss": 0.08782154321670532, "global_step": 155959, "epoch": 1752, "lr": 3.8517370360351935e-05}
{"train_loss": 0.09138353914022446, "global_step": 155960, "epoch": 1752, "lr": 3.851680611933299e-05}
{"train_loss": 0.14298051595687866, "global_step": 155961, "epoch": 1752, "lr": 3.8516241879857854e-05}
{"train_loss": 0.08640887588262558, "global_step": 155962, "epoch": 1752, "lr": 3.8515677641926547e-05}
{"train_loss": 0.11178578436374664, "global_step": 155963, "epoch": 1752, "lr": 3.8515113405539184e-05}
{"train_loss": 0.11806781589984894, "global_step": 155964, "epoch": 1752, "lr": 3.85145491706958e-05}
{"train_loss": 0.10932181775569916, "global_step": 155965, "epoch": 1752, "lr": 3.8513984937396516e-05}
{"train_loss": 0.10871285945177078, "global_step": 155966, "epoch": 1752, "lr": 3.8513420705641374e-05}
{"train_loss": 0.10718711465597153, "global_step": 155967, "epoch": 1752, "lr": 3.8512856475430474e-05}
{"train_loss": 0.08667445182800293, "global_step": 155968, "epoch": 1752, "lr": 3.8512292246763864e-05}
{"train_loss": 0.1026800274848938, "global_step": 155969, "epoch": 1752, "lr": 3.851172801964163e-05}
{"train_loss": 0.07874970883131027, "global_step": 155970, "epoch": 1752, "lr": 3.851116379406389e-05}
{"train_loss": 0.08518379181623459, "global_step": 155971, "epoch": 1752, "lr": 3.851059957003065e-05}
{"train_loss": 0.048698946833610535, "global_step": 155972, "epoch": 1752, "lr": 3.851003534754205e-05}
{"train_loss": 0.08610832691192627, "global_step": 155973, "epoch": 1752, "lr": 3.8509471126598115e-05}
{"train_loss": 0.09138604998588562, "global_step": 155974, "epoch": 1752, "lr": 3.850890690719895e-05}
{"train_loss": 0.10503385961055756, "global_step": 155975, "epoch": 1752, "lr": 3.8508342689344625e-05}
{"train_loss": 0.08675047755241394, "global_step": 155976, "epoch": 1752, "lr": 3.85077784730352e-05}
{"train_loss": 0.07303400337696075, "global_step": 155977, "epoch": 1752, "lr": 3.8507214258270775e-05}
{"train_loss": 0.05791529268026352, "global_step": 155978, "epoch": 1752, "lr": 3.8506650045051424e-05}
{"train_loss": 0.16207562386989594, "global_step": 155979, "epoch": 1752, "lr": 3.850608583337719e-05}
{"train_loss": 0.07692763954401016, "global_step": 155980, "epoch": 1752, "lr": 3.8505521623248194e-05}
{"train_loss": 0.08790076524019241, "global_step": 155981, "epoch": 1752, "lr": 3.8504957414664465e-05}
{"train_loss": 0.06769304722547531, "global_step": 155982, "epoch": 1752, "lr": 3.850439320762613e-05}
{"train_loss": 0.143398717045784, "global_step": 155983, "epoch": 1752, "lr": 3.8503829002133224e-05}
{"train_loss": 0.08489714562892914, "global_step": 155984, "epoch": 1752, "lr": 3.8503264798185854e-05}
{"train_loss": 0.10147260129451752, "global_step": 155985, "epoch": 1752, "lr": 3.8502700595784056e-05}
{"train_loss": 0.12625135481357574, "global_step": 155986, "epoch": 1752, "lr": 3.8502136394927936e-05}
{"train_loss": 0.14217723906040192, "global_step": 155987, "epoch": 1752, "lr": 3.850157219561758e-05}
{"train_loss": 0.07859434187412262, "global_step": 155988, "epoch": 1752, "lr": 3.850100799785303e-05}
{"train_loss": 0.08226040005683899, "global_step": 155989, "epoch": 1752, "lr": 3.850044380163439e-05}
{"train_loss": 0.06916201114654541, "global_step": 155990, "epoch": 1752, "lr": 3.849987960696172e-05}
{"train_loss": 0.12064377218484879, "global_step": 155991, "epoch": 1752, "lr": 3.849931541383511e-05}
{"train_loss": 0.10714282095432281, "global_step": 155992, "epoch": 1752, "lr": 3.849875122225462e-05}
{"train_loss": 0.11425958573818207, "global_step": 155993, "epoch": 1752, "lr": 3.8498187032220335e-05}
{"train_loss": 0.052971001714468, "global_step": 155994, "epoch": 1752, "lr": 3.8497622843732327e-05}
{"train_loss": 0.10750497132539749, "global_step": 155995, "epoch": 1752, "lr": 3.8497058656790675e-05}
{"train_loss": 0.10051687806844711, "global_step": 155996, "epoch": 1752, "lr": 3.849649447139545e-05}
{"train_loss": 0.08704514056444168, "global_step": 155997, "epoch": 1752, "lr": 3.849593028754675e-05}
{"train_loss": 0.05502493679523468, "global_step": 155998, "epoch": 1752, "lr": 3.8495366105244594e-05}
{"train_loss": 0.11693479865789413, "global_step": 155999, "epoch": 1752, "lr": 3.8494801924489135e-05}
{"train_loss": 0.0323861800134182, "global_step": 156000, "epoch": 1752, "lr": 3.849423774528037e-05}
{"train_loss": 0.13185054063796997, "global_step": 156001, "epoch": 1752, "lr": 3.849367356761845e-05}
{"train_loss": 0.09354960918426514, "global_step": 156002, "epoch": 1752, "lr": 3.849310939150339e-05}
{"train_loss": 0.06908413022756577, "global_step": 156003, "epoch": 1752, "lr": 3.849254521693532e-05}
{"train_loss": 0.11073286831378937, "global_step": 156004, "epoch": 1752, "lr": 3.849198104391425e-05}
{"train_loss": 0.08537004888057709, "global_step": 156005, "epoch": 1752, "lr": 3.84914168724403e-05}
{"train_loss": 0.11165991425514221, "global_step": 156006, "epoch": 1752, "lr": 3.8490852702513565e-05}
{"train_loss": 0.09622327983379364, "global_step": 156007, "epoch": 1752, "lr": 3.8490288534134064e-05}
{"train_loss": 0.08551768213510513, "global_step": 156008, "epoch": 1752, "lr": 3.848972436730194e-05}
{"train_loss": 0.13356320559978485, "global_step": 156009, "epoch": 1752, "lr": 3.8489160202017197e-05}
{"train_loss": 0.0933924987912178, "global_step": 156010, "epoch": 1752, "lr": 3.8488596038279965e-05}
{"train_loss": 0.09862519055604935, "global_step": 156011, "epoch": 1752, "lr": 3.8488031876090304e-05}
{"train_loss": 0.07969959825277328, "global_step": 156012, "epoch": 1752, "lr": 3.848746771544828e-05}
{"train_loss": 0.11663149297237396, "global_step": 156013, "epoch": 1752, "lr": 3.848690355635397e-05}
{"train_loss": 0.1033192127943039, "global_step": 156014, "epoch": 1752, "lr": 3.848633939880746e-05}
{"train_loss": 0.08118594437837601, "global_step": 156015, "epoch": 1752, "lr": 3.848577524280883e-05}
{"train_loss": 0.09562275089909521, "global_step": 156016, "epoch": 1752, "lr": 3.8485211088358146e-05, "val_loss": 6.581635475158691}
{"train_loss": 0.12422237545251846, "global_step": 156017, "epoch": 1753, "lr": 3.848464693545547e-05}
{"train_loss": 0.16370850801467896, "global_step": 156018, "epoch": 1753, "lr": 3.848408278410091e-05}
{"train_loss": 0.08324356377124786, "global_step": 156019, "epoch": 1753, "lr": 3.8483518634294505e-05}
{"train_loss": 0.07313849031925201, "global_step": 156020, "epoch": 1753, "lr": 3.8482954486036373e-05}
{"train_loss": 0.08258022367954254, "global_step": 156021, "epoch": 1753, "lr": 3.8482390339326544e-05}
{"train_loss": 0.09545037150382996, "global_step": 156022, "epoch": 1753, "lr": 3.8481826194165145e-05}
{"train_loss": 0.13537366688251495, "global_step": 156023, "epoch": 1753, "lr": 3.848126205055219e-05}
{"train_loss": 0.08425693958997726, "global_step": 156024, "epoch": 1753, "lr": 3.84806979084878e-05}
{"train_loss": 0.11439237743616104, "global_step": 156025, "epoch": 1753, "lr": 3.848013376797206e-05}
{"train_loss": 0.0806020125746727, "global_step": 156026, "epoch": 1753, "lr": 3.8479569629005005e-05}
{"train_loss": 0.14200864732265472, "global_step": 156027, "epoch": 1753, "lr": 3.847900549158675e-05}
{"train_loss": 0.11005187034606934, "global_step": 156028, "epoch": 1753, "lr": 3.847844135571733e-05}
{"train_loss": 0.10667415708303452, "global_step": 156029, "epoch": 1753, "lr": 3.8477877221396865e-05}
{"train_loss": 0.09815362840890884, "global_step": 156030, "epoch": 1753, "lr": 3.84773130886254e-05}
{"train_loss": 0.14807584881782532, "global_step": 156031, "epoch": 1753, "lr": 3.8476748957403016e-05}
{"train_loss": 0.09082801640033722, "global_step": 156032, "epoch": 1753, "lr": 3.847618482772979e-05}
{"train_loss": 0.08641017973423004, "global_step": 156033, "epoch": 1753, "lr": 3.847562069960581e-05}
{"train_loss": 0.11029309034347534, "global_step": 156034, "epoch": 1753, "lr": 3.847505657303113e-05}
{"train_loss": 0.07418530434370041, "global_step": 156035, "epoch": 1753, "lr": 3.847449244800586e-05}
{"train_loss": 0.10485637187957764, "global_step": 156036, "epoch": 1753, "lr": 3.847392832453003e-05}
{"train_loss": 0.08321820944547653, "global_step": 156037, "epoch": 1753, "lr": 3.847336420260376e-05}
{"train_loss": 0.06985664367675781, "global_step": 156038, "epoch": 1753, "lr": 3.847280008222708e-05}
{"train_loss": 0.09210973232984543, "global_step": 156039, "epoch": 1753, "lr": 3.847223596340012e-05}
{"train_loss": 0.07632247358560562, "global_step": 156040, "epoch": 1753, "lr": 3.847167184612291e-05}
{"train_loss": 0.11267026513814926, "global_step": 156041, "epoch": 1753, "lr": 3.847110773039553e-05}
{"train_loss": 0.0792892649769783, "global_step": 156042, "epoch": 1753, "lr": 3.8470543616218104e-05}
{"train_loss": 0.16929548978805542, "global_step": 156043, "epoch": 1753, "lr": 3.8469979503590644e-05}
{"train_loss": 0.05422490835189819, "global_step": 156044, "epoch": 1753, "lr": 3.8469415392513274e-05}
{"train_loss": 0.047774624079465866, "global_step": 156045, "epoch": 1753, "lr": 3.8468851282986036e-05}
{"train_loss": 0.11771515756845474, "global_step": 156046, "epoch": 1753, "lr": 3.846828717500903e-05}
{"train_loss": 0.11247766017913818, "global_step": 156047, "epoch": 1753, "lr": 3.8467723068582314e-05}
{"train_loss": 0.11754919588565826, "global_step": 156048, "epoch": 1753, "lr": 3.846715896370599e-05}
{"train_loss": 0.06903750449419022, "global_step": 156049, "epoch": 1753, "lr": 3.84665948603801e-05}
{"train_loss": 0.1251579225063324, "global_step": 156050, "epoch": 1753, "lr": 3.846603075860474e-05}
{"train_loss": 0.09716783463954926, "global_step": 156051, "epoch": 1753, "lr": 3.846546665837998e-05}
{"train_loss": 0.07891270518302917, "global_step": 156052, "epoch": 1753, "lr": 3.8464902559705904e-05}
{"train_loss": 0.09677352756261826, "global_step": 156053, "epoch": 1753, "lr": 3.846433846258257e-05}
{"train_loss": 0.050826720893383026, "global_step": 156054, "epoch": 1753, "lr": 3.846377436701008e-05}
{"train_loss": 0.1580718457698822, "global_step": 156055, "epoch": 1753, "lr": 3.8463210272988475e-05}
{"train_loss": 0.13045690953731537, "global_step": 156056, "epoch": 1753, "lr": 3.846264618051787e-05}
{"train_loss": 0.07178301364183426, "global_step": 156057, "epoch": 1753, "lr": 3.846208208959831e-05}
{"train_loss": 0.09707299619913101, "global_step": 156058, "epoch": 1753, "lr": 3.846151800022989e-05}
{"train_loss": 0.07863204181194305, "global_step": 156059, "epoch": 1753, "lr": 3.8460953912412664e-05}
{"train_loss": 0.06030154973268509, "global_step": 156060, "epoch": 1753, "lr": 3.8460389826146726e-05}
{"train_loss": 0.08348909020423889, "global_step": 156061, "epoch": 1753, "lr": 3.845982574143216e-05}
{"train_loss": 0.12480518966913223, "global_step": 156062, "epoch": 1753, "lr": 3.845926165826901e-05}
{"train_loss": 0.09220204502344131, "global_step": 156063, "epoch": 1753, "lr": 3.84586975766574e-05}
{"train_loss": 0.09719003736972809, "global_step": 156064, "epoch": 1753, "lr": 3.8458133496597345e-05}
{"train_loss": 0.10878448188304901, "global_step": 156065, "epoch": 1753, "lr": 3.845756941808899e-05}
{"train_loss": 0.12607400119304657, "global_step": 156066, "epoch": 1753, "lr": 3.845700534113234e-05}
{"train_loss": 0.17061921954154968, "global_step": 156067, "epoch": 1753, "lr": 3.8456441265727536e-05}
{"train_loss": 0.10216808319091797, "global_step": 156068, "epoch": 1753, "lr": 3.845587719187459e-05}
{"train_loss": 0.06867556273937225, "global_step": 156069, "epoch": 1753, "lr": 3.8455313119573635e-05}
{"train_loss": 0.11212234944105148, "global_step": 156070, "epoch": 1753, "lr": 3.845474904882472e-05}
{"train_loss": 0.055432237684726715, "global_step": 156071, "epoch": 1753, "lr": 3.8454184979627917e-05}
{"train_loss": 0.1112266331911087, "global_step": 156072, "epoch": 1753, "lr": 3.845362091198331e-05}
{"train_loss": 0.09152662754058838, "global_step": 156073, "epoch": 1753, "lr": 3.845305684589097e-05}
{"train_loss": 0.09253712743520737, "global_step": 156074, "epoch": 1753, "lr": 3.845249278135098e-05}
{"train_loss": 0.11883124709129333, "global_step": 156075, "epoch": 1753, "lr": 3.845192871836342e-05}
{"train_loss": 0.09803066402673721, "global_step": 156076, "epoch": 1753, "lr": 3.845136465692833e-05}
{"train_loss": 0.11412588506937027, "global_step": 156077, "epoch": 1753, "lr": 3.8450800597045824e-05}
{"train_loss": 0.09686829149723053, "global_step": 156078, "epoch": 1753, "lr": 3.845023653871598e-05}
{"train_loss": 0.09313187748193741, "global_step": 156079, "epoch": 1753, "lr": 3.8449672481938845e-05}
{"train_loss": 0.08199384063482285, "global_step": 156080, "epoch": 1753, "lr": 3.844910842671453e-05}
{"train_loss": 0.1616315245628357, "global_step": 156081, "epoch": 1753, "lr": 3.844854437304307e-05}
{"train_loss": 0.08356954157352448, "global_step": 156082, "epoch": 1753, "lr": 3.844798032092459e-05}
{"train_loss": 0.12880603969097137, "global_step": 156083, "epoch": 1753, "lr": 3.8447416270359104e-05}
{"train_loss": 0.08770702034235, "global_step": 156084, "epoch": 1753, "lr": 3.844685222134675e-05}
{"train_loss": 0.09017712622880936, "global_step": 156085, "epoch": 1753, "lr": 3.8446288173887555e-05}
{"train_loss": 0.07898493111133575, "global_step": 156086, "epoch": 1753, "lr": 3.844572412798164e-05}
{"train_loss": 0.049974095076322556, "global_step": 156087, "epoch": 1753, "lr": 3.844516008362903e-05}
{"train_loss": 0.10052129626274109, "global_step": 156088, "epoch": 1753, "lr": 3.844459604082986e-05}
{"train_loss": 0.08003983646631241, "global_step": 156089, "epoch": 1753, "lr": 3.844403199958415e-05}
{"train_loss": 0.09372080117464066, "global_step": 156090, "epoch": 1753, "lr": 3.8443467959892005e-05}
{"train_loss": 0.06873669475317001, "global_step": 156091, "epoch": 1753, "lr": 3.844290392175349e-05}
{"train_loss": 0.06796418875455856, "global_step": 156092, "epoch": 1753, "lr": 3.844233988516869e-05}
{"train_loss": 0.08234618604183197, "global_step": 156093, "epoch": 1753, "lr": 3.8441775850137675e-05}
{"train_loss": 0.08524511754512787, "global_step": 156094, "epoch": 1753, "lr": 3.844121181666053e-05}
{"train_loss": 0.07176953554153442, "global_step": 156095, "epoch": 1753, "lr": 3.844064778473731e-05}
{"train_loss": 0.1229136660695076, "global_step": 156096, "epoch": 1753, "lr": 3.84400837543681e-05}
{"train_loss": 0.1081533208489418, "global_step": 156097, "epoch": 1753, "lr": 3.8439519725552995e-05}
{"train_loss": 0.08745191991329193, "global_step": 156098, "epoch": 1753, "lr": 3.8438955698292037e-05}
{"train_loss": 0.08987338095903397, "global_step": 156099, "epoch": 1753, "lr": 3.843839167258535e-05}
{"train_loss": 0.07017290592193604, "global_step": 156100, "epoch": 1753, "lr": 3.843782764843295e-05}
{"train_loss": 0.09101538360118866, "global_step": 156101, "epoch": 1753, "lr": 3.8437263625834966e-05}
{"train_loss": 0.1302792727947235, "global_step": 156102, "epoch": 1753, "lr": 3.843669960479144e-05}
{"train_loss": 0.09570643305778503, "global_step": 156103, "epoch": 1753, "lr": 3.843613558530247e-05}
{"train_loss": 0.06653446704149246, "global_step": 156104, "epoch": 1753, "lr": 3.8435571567368095e-05}
{"train_loss": 0.09741371175211468, "global_step": 156105, "epoch": 1753, "lr": 3.843500755098845e-05, "val_loss": 6.492442607879639}
{"train_loss": 0.09233983606100082, "global_step": 156106, "epoch": 1754, "lr": 3.8434443536163545e-05}
{"train_loss": 0.09067074954509735, "global_step": 156107, "epoch": 1754, "lr": 3.843387952289351e-05}
{"train_loss": 0.07374025136232376, "global_step": 156108, "epoch": 1754, "lr": 3.84333155111784e-05}
{"train_loss": 0.10967551916837692, "global_step": 156109, "epoch": 1754, "lr": 3.8432751501018276e-05}
{"train_loss": 0.17205855250358582, "global_step": 156110, "epoch": 1754, "lr": 3.843218749241324e-05}
{"train_loss": 0.0614863857626915, "global_step": 156111, "epoch": 1754, "lr": 3.8431623485363346e-05}
{"train_loss": 0.08144230395555496, "global_step": 156112, "epoch": 1754, "lr": 3.843105947986868e-05}
{"train_loss": 0.09009677171707153, "global_step": 156113, "epoch": 1754, "lr": 3.8430495475929304e-05}
{"train_loss": 0.0752238929271698, "global_step": 156114, "epoch": 1754, "lr": 3.842993147354534e-05}
{"train_loss": 0.08024294674396515, "global_step": 156115, "epoch": 1754, "lr": 3.842936747271681e-05}
{"train_loss": 0.0690070241689682, "global_step": 156116, "epoch": 1754, "lr": 3.8428803473443816e-05}
{"train_loss": 0.10608810931444168, "global_step": 156117, "epoch": 1754, "lr": 3.842823947572641e-05}
{"train_loss": 0.06232057139277458, "global_step": 156118, "epoch": 1754, "lr": 3.8427675479564715e-05}
{"train_loss": 0.07878722995519638, "global_step": 156119, "epoch": 1754, "lr": 3.842711148495875e-05}
{"train_loss": 0.10445573925971985, "global_step": 156120, "epoch": 1754, "lr": 3.842654749190865e-05}
{"train_loss": 0.11683201789855957, "global_step": 156121, "epoch": 1754, "lr": 3.842598350041443e-05}
{"train_loss": 0.18322689831256866, "global_step": 156122, "epoch": 1754, "lr": 3.842541951047622e-05}
{"train_loss": 0.09396679699420929, "global_step": 156123, "epoch": 1754, "lr": 3.842485552209405e-05}
{"train_loss": 0.07849632203578949, "global_step": 156124, "epoch": 1754, "lr": 3.8424291535268044e-05}
{"train_loss": 0.07571949064731598, "global_step": 156125, "epoch": 1754, "lr": 3.8423727549998215e-05}
{"train_loss": 0.041483670473098755, "global_step": 156126, "epoch": 1754, "lr": 3.842316356628471e-05}
{"train_loss": 0.10556170344352722, "global_step": 156127, "epoch": 1754, "lr": 3.842259958412755e-05}
{"train_loss": 0.11072196066379547, "global_step": 156128, "epoch": 1754, "lr": 3.842203560352683e-05}
{"train_loss": 0.05690972879528999, "global_step": 156129, "epoch": 1754, "lr": 3.8421471624482635e-05}
{"train_loss": 0.12838326394557953, "global_step": 156130, "epoch": 1754, "lr": 3.8420907646995027e-05}
{"train_loss": 0.05919114872813225, "global_step": 156131, "epoch": 1754, "lr": 3.842034367106409e-05}
{"train_loss": 0.1503618061542511, "global_step": 156132, "epoch": 1754, "lr": 3.841977969668988e-05}
{"train_loss": 0.09884309023618698, "global_step": 156133, "epoch": 1754, "lr": 3.8419215723872514e-05}
{"train_loss": 0.11764511466026306, "global_step": 156134, "epoch": 1754, "lr": 3.841865175261202e-05}
{"train_loss": 0.0690167173743248, "global_step": 156135, "epoch": 1754, "lr": 3.8418087782908516e-05}
{"train_loss": 0.17227528989315033, "global_step": 156136, "epoch": 1754, "lr": 3.841752381476205e-05}
{"train_loss": 0.10546359419822693, "global_step": 156137, "epoch": 1754, "lr": 3.841695984817272e-05}
{"train_loss": 0.10342555493116379, "global_step": 156138, "epoch": 1754, "lr": 3.841639588314056e-05}
{"train_loss": 0.0688299834728241, "global_step": 156139, "epoch": 1754, "lr": 3.84158319196657e-05}
{"train_loss": 0.09497545659542084, "global_step": 156140, "epoch": 1754, "lr": 3.841526795774818e-05}
{"train_loss": 0.08964107930660248, "global_step": 156141, "epoch": 1754, "lr": 3.84147039973881e-05}
{"train_loss": 0.08995714783668518, "global_step": 156142, "epoch": 1754, "lr": 3.841414003858549e-05}
{"train_loss": 0.09123395383358002, "global_step": 156143, "epoch": 1754, "lr": 3.841357608134049e-05}
{"train_loss": 0.16435644030570984, "global_step": 156144, "epoch": 1754, "lr": 3.841301212565312e-05}
{"train_loss": 0.0945938378572464, "global_step": 156145, "epoch": 1754, "lr": 3.8412448171523504e-05}
{"train_loss": 0.09196338057518005, "global_step": 156146, "epoch": 1754, "lr": 3.8411884218951674e-05}
{"train_loss": 0.13211405277252197, "global_step": 156147, "epoch": 1754, "lr": 3.8411320267937735e-05}
{"train_loss": 0.06480088084936142, "global_step": 156148, "epoch": 1754, "lr": 3.841075631848175e-05}
{"train_loss": 0.09689828008413315, "global_step": 156149, "epoch": 1754, "lr": 3.84101923705838e-05}
{"train_loss": 0.16160805523395538, "global_step": 156150, "epoch": 1754, "lr": 3.840962842424395e-05}
{"train_loss": 0.07748490571975708, "global_step": 156151, "epoch": 1754, "lr": 3.8409064479462274e-05}
{"train_loss": 0.08889775723218918, "global_step": 156152, "epoch": 1754, "lr": 3.840850053623888e-05}
{"train_loss": 0.07414738088846207, "global_step": 156153, "epoch": 1754, "lr": 3.840793659457381e-05}
{"train_loss": 0.04944143816828728, "global_step": 156154, "epoch": 1754, "lr": 3.840737265446717e-05}
{"train_loss": 0.12932007014751434, "global_step": 156155, "epoch": 1754, "lr": 3.8406808715918984e-05}
{"train_loss": 0.07008597999811172, "global_step": 156156, "epoch": 1754, "lr": 3.8406244778929396e-05}
{"train_loss": 0.04492386430501938, "global_step": 156157, "epoch": 1754, "lr": 3.8405680843498424e-05}
{"train_loss": 0.12210088223218918, "global_step": 156158, "epoch": 1754, "lr": 3.840511690962618e-05}
{"train_loss": 0.08109648525714874, "global_step": 156159, "epoch": 1754, "lr": 3.840455297731271e-05}
{"train_loss": 0.1204167976975441, "global_step": 156160, "epoch": 1754, "lr": 3.840398904655813e-05}
{"train_loss": 0.12281068414449692, "global_step": 156161, "epoch": 1754, "lr": 3.840342511736247e-05}
{"train_loss": 0.06921850144863129, "global_step": 156162, "epoch": 1754, "lr": 3.840286118972585e-05}
{"train_loss": 0.13422925770282745, "global_step": 156163, "epoch": 1754, "lr": 3.8402297263648304e-05}
{"train_loss": 0.052272651344537735, "global_step": 156164, "epoch": 1754, "lr": 3.840173333912994e-05}
{"train_loss": 0.08986937999725342, "global_step": 156165, "epoch": 1754, "lr": 3.840116941617081e-05}
{"train_loss": 0.059284962713718414, "global_step": 156166, "epoch": 1754, "lr": 3.840060549477102e-05}
{"train_loss": 0.08923423290252686, "global_step": 156167, "epoch": 1754, "lr": 3.840004157493061e-05}
{"train_loss": 0.12735241651535034, "global_step": 156168, "epoch": 1754, "lr": 3.8399477656649676e-05}
{"train_loss": 0.07296936213970184, "global_step": 156169, "epoch": 1754, "lr": 3.83989137399283e-05}
{"train_loss": 0.07135722041130066, "global_step": 156170, "epoch": 1754, "lr": 3.8398349824766533e-05}
{"train_loss": 0.1561477929353714, "global_step": 156171, "epoch": 1754, "lr": 3.839778591116449e-05}
{"train_loss": 0.10658536106348038, "global_step": 156172, "epoch": 1754, "lr": 3.8397221999122195e-05}
{"train_loss": 0.05594942346215248, "global_step": 156173, "epoch": 1754, "lr": 3.839665808863978e-05}
{"train_loss": 0.09847763180732727, "global_step": 156174, "epoch": 1754, "lr": 3.839609417971727e-05}
{"train_loss": 0.07976266741752625, "global_step": 156175, "epoch": 1754, "lr": 3.839553027235478e-05}
{"train_loss": 0.1029750406742096, "global_step": 156176, "epoch": 1754, "lr": 3.839496636655236e-05}
{"train_loss": 0.09119447320699692, "global_step": 156177, "epoch": 1754, "lr": 3.83944024623101e-05}
{"train_loss": 0.11070799827575684, "global_step": 156178, "epoch": 1754, "lr": 3.839383855962806e-05}
{"train_loss": 0.09598449617624283, "global_step": 156179, "epoch": 1754, "lr": 3.8393274658506343e-05}
{"train_loss": 0.06879644095897675, "global_step": 156180, "epoch": 1754, "lr": 3.8392710758944993e-05}
{"train_loss": 0.10048800706863403, "global_step": 156181, "epoch": 1754, "lr": 3.839214686094412e-05}
{"train_loss": 0.10866972804069519, "global_step": 156182, "epoch": 1754, "lr": 3.839158296450376e-05}
{"train_loss": 0.08081403374671936, "global_step": 156183, "epoch": 1754, "lr": 3.839101906962403e-05}
{"train_loss": 0.08915265649557114, "global_step": 156184, "epoch": 1754, "lr": 3.8390455176304976e-05}
{"train_loss": 0.09586375951766968, "global_step": 156185, "epoch": 1754, "lr": 3.83898912845467e-05}
{"train_loss": 0.07751045376062393, "global_step": 156186, "epoch": 1754, "lr": 3.838932739434924e-05}
{"train_loss": 0.11077482253313065, "global_step": 156187, "epoch": 1754, "lr": 3.8388763505712686e-05}
{"train_loss": 0.11095711588859558, "global_step": 156188, "epoch": 1754, "lr": 3.838819961863716e-05}
{"train_loss": 0.0851295217871666, "global_step": 156189, "epoch": 1754, "lr": 3.8387635733122656e-05}
{"train_loss": 0.1095941960811615, "global_step": 156190, "epoch": 1754, "lr": 3.838707184916932e-05}
{"train_loss": 0.09439598023891449, "global_step": 156191, "epoch": 1754, "lr": 3.8386507966777184e-05}
{"train_loss": 0.1790383905172348, "global_step": 156192, "epoch": 1754, "lr": 3.8385944085946354e-05}
{"train_loss": 0.09315004199743271, "global_step": 156193, "epoch": 1754, "lr": 3.838538020667688e-05}
{"train_loss": 0.096628776343351, "global_step": 156194, "epoch": 1754, "lr": 3.838481632896887e-05, "val_loss": 6.530172824859619}
{"train_loss": 0.07506833970546722, "global_step": 156195, "epoch": 1755, "lr": 3.8384252452822356e-05}
{"train_loss": 0.0784493163228035, "global_step": 156196, "epoch": 1755, "lr": 3.838368857823746e-05}
{"train_loss": 0.11535923928022385, "global_step": 156197, "epoch": 1755, "lr": 3.8383124705214215e-05}
{"train_loss": 0.10812048614025116, "global_step": 156198, "epoch": 1755, "lr": 3.8382560833752736e-05}
{"train_loss": 0.09265932440757751, "global_step": 156199, "epoch": 1755, "lr": 3.838199696385306e-05}
{"train_loss": 0.13201354444026947, "global_step": 156200, "epoch": 1755, "lr": 3.8381433095515306e-05}
{"train_loss": 0.06538926810026169, "global_step": 156201, "epoch": 1755, "lr": 3.8380869228739505e-05}
{"train_loss": 0.10985612124204636, "global_step": 156202, "epoch": 1755, "lr": 3.838030536352577e-05}
{"train_loss": 0.09623096138238907, "global_step": 156203, "epoch": 1755, "lr": 3.837974149987414e-05}
{"train_loss": 0.07345844060182571, "global_step": 156204, "epoch": 1755, "lr": 3.837917763778474e-05}
{"train_loss": 0.15174426138401031, "global_step": 156205, "epoch": 1755, "lr": 3.837861377725761e-05}
{"train_loss": 0.15018267929553986, "global_step": 156206, "epoch": 1755, "lr": 3.837804991829282e-05}
{"train_loss": 0.07934605330228806, "global_step": 156207, "epoch": 1755, "lr": 3.837748606089048e-05}
{"train_loss": 0.12611940503120422, "global_step": 156208, "epoch": 1755, "lr": 3.837692220505062e-05}
{"train_loss": 0.13473589718341827, "global_step": 156209, "epoch": 1755, "lr": 3.8376358350773365e-05}
{"train_loss": 0.13620561361312866, "global_step": 156210, "epoch": 1755, "lr": 3.837579449805875e-05}
{"train_loss": 0.09491882473230362, "global_step": 156211, "epoch": 1755, "lr": 3.837523064690688e-05}
{"train_loss": 0.1181451603770256, "global_step": 156212, "epoch": 1755, "lr": 3.83746667973178e-05}
{"train_loss": 0.09577323496341705, "global_step": 156213, "epoch": 1755, "lr": 3.837410294929164e-05}
{"train_loss": 0.06414245814085007, "global_step": 156214, "epoch": 1755, "lr": 3.83735391028284e-05}
{"train_loss": 0.052371278405189514, "global_step": 156215, "epoch": 1755, "lr": 3.8372975257928226e-05}
{"train_loss": 0.07362673431634903, "global_step": 156216, "epoch": 1755, "lr": 3.8372411414591134e-05}
{"train_loss": 0.09908480942249298, "global_step": 156217, "epoch": 1755, "lr": 3.8371847572817265e-05}
{"train_loss": 0.1096658930182457, "global_step": 156218, "epoch": 1755, "lr": 3.8371283732606626e-05}
{"train_loss": 0.08212192356586456, "global_step": 156219, "epoch": 1755, "lr": 3.837071989395936e-05}
{"train_loss": 0.11651433259248734, "global_step": 156220, "epoch": 1755, "lr": 3.837015605687547e-05}
{"train_loss": 0.09789776057004929, "global_step": 156221, "epoch": 1755, "lr": 3.8369592221355104e-05}
{"train_loss": 0.13509006798267365, "global_step": 156222, "epoch": 1755, "lr": 3.836902838739828e-05}
{"train_loss": 0.07115980982780457, "global_step": 156223, "epoch": 1755, "lr": 3.836846455500511e-05}
{"train_loss": 0.06735386699438095, "global_step": 156224, "epoch": 1755, "lr": 3.836790072417567e-05}
{"train_loss": 0.10660795122385025, "global_step": 156225, "epoch": 1755, "lr": 3.836733689491001e-05}
{"train_loss": 0.08710749447345734, "global_step": 156226, "epoch": 1755, "lr": 3.836677306720823e-05}
{"train_loss": 0.14964257180690765, "global_step": 156227, "epoch": 1755, "lr": 3.836620924107037e-05}
{"train_loss": 0.10908124595880508, "global_step": 156228, "epoch": 1755, "lr": 3.836564541649656e-05}
{"train_loss": 0.1030345931649208, "global_step": 156229, "epoch": 1755, "lr": 3.836508159348683e-05}
{"train_loss": 0.09822279214859009, "global_step": 156230, "epoch": 1755, "lr": 3.836451777204129e-05}
{"train_loss": 0.07161185145378113, "global_step": 156231, "epoch": 1755, "lr": 3.8363953952159974e-05}
{"train_loss": 0.07458774745464325, "global_step": 156232, "epoch": 1755, "lr": 3.836339013384301e-05}
{"train_loss": 0.12528781592845917, "global_step": 156233, "epoch": 1755, "lr": 3.8362826317090425e-05}
{"train_loss": 0.13332709670066833, "global_step": 156234, "epoch": 1755, "lr": 3.836226250190233e-05}
{"train_loss": 0.10010842233896255, "global_step": 156235, "epoch": 1755, "lr": 3.836169868827877e-05}
{"train_loss": 0.12369461357593536, "global_step": 156236, "epoch": 1755, "lr": 3.836113487621986e-05}
{"train_loss": 0.14311684668064117, "global_step": 156237, "epoch": 1755, "lr": 3.836057106572563e-05}
{"train_loss": 0.09236524999141693, "global_step": 156238, "epoch": 1755, "lr": 3.83600072567962e-05}
{"train_loss": 0.07365304231643677, "global_step": 156239, "epoch": 1755, "lr": 3.835944344943161e-05}
{"train_loss": 0.16173651814460754, "global_step": 156240, "epoch": 1755, "lr": 3.835887964363196e-05}
{"train_loss": 0.06048104166984558, "global_step": 156241, "epoch": 1755, "lr": 3.8358315839397314e-05}
{"train_loss": 0.08128534257411957, "global_step": 156242, "epoch": 1755, "lr": 3.835775203672775e-05}
{"train_loss": 0.07626328617334366, "global_step": 156243, "epoch": 1755, "lr": 3.8357188235623356e-05}
{"train_loss": 0.0893339216709137, "global_step": 156244, "epoch": 1755, "lr": 3.8356624436084176e-05}
{"train_loss": 0.0985829159617424, "global_step": 156245, "epoch": 1755, "lr": 3.835606063811033e-05}
{"train_loss": 0.057055551558732986, "global_step": 156246, "epoch": 1755, "lr": 3.835549684170184e-05}
{"train_loss": 0.16504740715026855, "global_step": 156247, "epoch": 1755, "lr": 3.8354933046858836e-05}
{"train_loss": 0.08589335530996323, "global_step": 156248, "epoch": 1755, "lr": 3.835436925358135e-05}
{"train_loss": 0.16107240319252014, "global_step": 156249, "epoch": 1755, "lr": 3.8353805461869494e-05}
{"train_loss": 0.09353333711624146, "global_step": 156250, "epoch": 1755, "lr": 3.8353241671723314e-05}
{"train_loss": 0.07199772447347641, "global_step": 156251, "epoch": 1755, "lr": 3.835267788314292e-05}
{"train_loss": 0.046527374535799026, "global_step": 156252, "epoch": 1755, "lr": 3.835211409612834e-05}
{"train_loss": 0.09188257902860641, "global_step": 156253, "epoch": 1755, "lr": 3.8351550310679695e-05}
{"train_loss": 0.08639217168092728, "global_step": 156254, "epoch": 1755, "lr": 3.8350986526797025e-05}
{"train_loss": 0.12921226024627686, "global_step": 156255, "epoch": 1755, "lr": 3.8350422744480444e-05}
{"train_loss": 0.12812714278697968, "global_step": 156256, "epoch": 1755, "lr": 3.8349858963729986e-05}
{"train_loss": 0.0681474506855011, "global_step": 156257, "epoch": 1755, "lr": 3.834929518454577e-05}
{"train_loss": 0.14593571424484253, "global_step": 156258, "epoch": 1755, "lr": 3.834873140692782e-05}
{"train_loss": 0.1363259106874466, "global_step": 156259, "epoch": 1755, "lr": 3.834816763087626e-05}
{"train_loss": 0.11431076377630234, "global_step": 156260, "epoch": 1755, "lr": 3.834760385639114e-05}
{"train_loss": 0.13668949902057648, "global_step": 156261, "epoch": 1755, "lr": 3.834704008347255e-05}
{"train_loss": 0.10795993357896805, "global_step": 156262, "epoch": 1755, "lr": 3.834647631212056e-05}
{"train_loss": 0.08039072155952454, "global_step": 156263, "epoch": 1755, "lr": 3.8345912542335234e-05}
{"train_loss": 0.10053157061338425, "global_step": 156264, "epoch": 1755, "lr": 3.834534877411668e-05}
{"train_loss": 0.08247663080692291, "global_step": 156265, "epoch": 1755, "lr": 3.834478500746492e-05}
{"train_loss": 0.03753559663891792, "global_step": 156266, "epoch": 1755, "lr": 3.8344221242380086e-05}
{"train_loss": 0.08019090443849564, "global_step": 156267, "epoch": 1755, "lr": 3.83436574788622e-05}
{"train_loss": 0.08751694858074188, "global_step": 156268, "epoch": 1755, "lr": 3.8343093716911403e-05}
{"train_loss": 0.05540047585964203, "global_step": 156269, "epoch": 1755, "lr": 3.8342529956527704e-05}
{"train_loss": 0.07485359907150269, "global_step": 156270, "epoch": 1755, "lr": 3.8341966197711246e-05}
{"train_loss": 0.11598870903253555, "global_step": 156271, "epoch": 1755, "lr": 3.834140244046203e-05}
{"train_loss": 0.10235217213630676, "global_step": 156272, "epoch": 1755, "lr": 3.83408386847802e-05}
{"train_loss": 0.10567755997180939, "global_step": 156273, "epoch": 1755, "lr": 3.8340274930665774e-05}
{"train_loss": 0.12356079369783401, "global_step": 156274, "epoch": 1755, "lr": 3.8339711178118884e-05}
{"train_loss": 0.15994209051132202, "global_step": 156275, "epoch": 1755, "lr": 3.833914742713956e-05}
{"train_loss": 0.1315414011478424, "global_step": 156276, "epoch": 1755, "lr": 3.833858367772791e-05}
{"train_loss": 0.09219446778297424, "global_step": 156277, "epoch": 1755, "lr": 3.8338019929883974e-05}
{"train_loss": 0.062075309455394745, "global_step": 156278, "epoch": 1755, "lr": 3.833745618360786e-05}
{"train_loss": 0.0750020295381546, "global_step": 156279, "epoch": 1755, "lr": 3.833689243889964e-05}
{"train_loss": 0.09127037227153778, "global_step": 156280, "epoch": 1755, "lr": 3.833632869575937e-05}
{"train_loss": 0.1487165093421936, "global_step": 156281, "epoch": 1755, "lr": 3.833576495418715e-05}
{"train_loss": 0.05704633519053459, "global_step": 156282, "epoch": 1755, "lr": 3.8335201214183036e-05}
{"train_loss": 0.09986129955629285, "global_step": 156283, "epoch": 1755, "lr": 3.833463747574713e-05, "val_loss": 6.563645839691162, "train_action_mse_error": 3.8038628101348877}
{"train_loss": 0.061215490102767944, "global_step": 156284, "epoch": 1756, "lr": 3.8334073738879466e-05}
{"train_loss": 0.08300988376140594, "global_step": 156285, "epoch": 1756, "lr": 3.8333510003580167e-05}
{"train_loss": 0.10621459037065506, "global_step": 156286, "epoch": 1756, "lr": 3.833294626984926e-05}
{"train_loss": 0.16390879452228546, "global_step": 156287, "epoch": 1756, "lr": 3.833238253768686e-05}
{"train_loss": 0.07549518346786499, "global_step": 156288, "epoch": 1756, "lr": 3.833181880709302e-05}
{"train_loss": 0.12557028234004974, "global_step": 156289, "epoch": 1756, "lr": 3.833125507806784e-05}
{"train_loss": 0.036983076483011246, "global_step": 156290, "epoch": 1756, "lr": 3.833069135061136e-05}
{"train_loss": 0.13137872517108917, "global_step": 156291, "epoch": 1756, "lr": 3.83301276247237e-05}
{"train_loss": 0.08180959522724152, "global_step": 156292, "epoch": 1756, "lr": 3.832956390040489e-05}
{"train_loss": 0.13207513093948364, "global_step": 156293, "epoch": 1756, "lr": 3.832900017765505e-05}
{"train_loss": 0.13164900243282318, "global_step": 156294, "epoch": 1756, "lr": 3.8328436456474216e-05}
{"train_loss": 0.0421021543443203, "global_step": 156295, "epoch": 1756, "lr": 3.8327872736862475e-05}
{"train_loss": 0.10778742283582687, "global_step": 156296, "epoch": 1756, "lr": 3.8327309018819926e-05}
{"train_loss": 0.06142323836684227, "global_step": 156297, "epoch": 1756, "lr": 3.832674530234663e-05}
{"train_loss": 0.09595905989408493, "global_step": 156298, "epoch": 1756, "lr": 3.832618158744265e-05}
{"train_loss": 0.12587687373161316, "global_step": 156299, "epoch": 1756, "lr": 3.8325617874108075e-05}
{"train_loss": 0.16366638243198395, "global_step": 156300, "epoch": 1756, "lr": 3.832505416234298e-05}
{"train_loss": 0.07163997739553452, "global_step": 156301, "epoch": 1756, "lr": 3.832449045214744e-05}
{"train_loss": 0.12055344134569168, "global_step": 156302, "epoch": 1756, "lr": 3.832392674352154e-05}
{"train_loss": 0.06898021697998047, "global_step": 156303, "epoch": 1756, "lr": 3.832336303646533e-05}
{"train_loss": 0.1860819160938263, "global_step": 156304, "epoch": 1756, "lr": 3.832279933097891e-05}
{"train_loss": 0.1270889788866043, "global_step": 156305, "epoch": 1756, "lr": 3.832223562706234e-05}
{"train_loss": 0.09188953042030334, "global_step": 156306, "epoch": 1756, "lr": 3.832167192471572e-05}
{"train_loss": 0.08274518698453903, "global_step": 156307, "epoch": 1756, "lr": 3.832110822393908e-05}
{"train_loss": 0.10976532101631165, "global_step": 156308, "epoch": 1756, "lr": 3.8320544524732554e-05}
{"train_loss": 0.09054465591907501, "global_step": 156309, "epoch": 1756, "lr": 3.831998082709616e-05}
{"train_loss": 0.11161988973617554, "global_step": 156310, "epoch": 1756, "lr": 3.831941713103003e-05}
{"train_loss": 0.10606913268566132, "global_step": 156311, "epoch": 1756, "lr": 3.831885343653419e-05}
{"train_loss": 0.08306853473186493, "global_step": 156312, "epoch": 1756, "lr": 3.831828974360876e-05}
{"train_loss": 0.11387789249420166, "global_step": 156313, "epoch": 1756, "lr": 3.8317726052253766e-05}
{"train_loss": 0.0889137014746666, "global_step": 156314, "epoch": 1756, "lr": 3.8317162362469314e-05}
{"train_loss": 0.11401797831058502, "global_step": 156315, "epoch": 1756, "lr": 3.8316598674255505e-05}
{"train_loss": 0.07384853810071945, "global_step": 156316, "epoch": 1756, "lr": 3.831603498761237e-05}
{"train_loss": 0.07293839007616043, "global_step": 156317, "epoch": 1756, "lr": 3.831547130254e-05}
{"train_loss": 0.09321161359548569, "global_step": 156318, "epoch": 1756, "lr": 3.831490761903848e-05}
{"train_loss": 0.10512856394052505, "global_step": 156319, "epoch": 1756, "lr": 3.831434393710787e-05}
{"train_loss": 0.09499406069517136, "global_step": 156320, "epoch": 1756, "lr": 3.831378025674826e-05}
{"train_loss": 0.0739317312836647, "global_step": 156321, "epoch": 1756, "lr": 3.8313216577959726e-05}
{"train_loss": 0.17054659128189087, "global_step": 156322, "epoch": 1756, "lr": 3.8312652900742316e-05}
{"train_loss": 0.10075092315673828, "global_step": 156323, "epoch": 1756, "lr": 3.831208922509615e-05}
{"train_loss": 0.11722303181886673, "global_step": 156324, "epoch": 1756, "lr": 3.831152555102127e-05}
{"train_loss": 0.12755638360977173, "global_step": 156325, "epoch": 1756, "lr": 3.831096187851777e-05}
{"train_loss": 0.13666751980781555, "global_step": 156326, "epoch": 1756, "lr": 3.8310398207585705e-05}
{"train_loss": 0.11134263128042221, "global_step": 156327, "epoch": 1756, "lr": 3.830983453822518e-05}
{"train_loss": 0.1303444802761078, "global_step": 156328, "epoch": 1756, "lr": 3.830927087043624e-05}
{"train_loss": 0.12864546477794647, "global_step": 156329, "epoch": 1756, "lr": 3.8308707204218994e-05}
{"train_loss": 0.10529470443725586, "global_step": 156330, "epoch": 1756, "lr": 3.8308143539573477e-05}
{"train_loss": 0.09113801270723343, "global_step": 156331, "epoch": 1756, "lr": 3.830757987649979e-05}
{"train_loss": 0.11862564086914062, "global_step": 156332, "epoch": 1756, "lr": 3.8307016214998024e-05}
{"train_loss": 0.11756911128759384, "global_step": 156333, "epoch": 1756, "lr": 3.8306452555068224e-05}
{"train_loss": 0.035617321729660034, "global_step": 156334, "epoch": 1756, "lr": 3.83058888967105e-05}
{"train_loss": 0.16731226444244385, "global_step": 156335, "epoch": 1756, "lr": 3.830532523992488e-05}
{"train_loss": 0.0772816613316536, "global_step": 156336, "epoch": 1756, "lr": 3.8304761584711485e-05}
{"train_loss": 0.12472110241651535, "global_step": 156337, "epoch": 1756, "lr": 3.830419793107036e-05}
{"train_loss": 0.1252158284187317, "global_step": 156338, "epoch": 1756, "lr": 3.83036342790016e-05}
{"train_loss": 0.15236209332942963, "global_step": 156339, "epoch": 1756, "lr": 3.830307062850527e-05}
{"train_loss": 0.1248769462108612, "global_step": 156340, "epoch": 1756, "lr": 3.8302506979581465e-05}
{"train_loss": 0.1285543590784073, "global_step": 156341, "epoch": 1756, "lr": 3.830194333223022e-05}
{"train_loss": 0.12770700454711914, "global_step": 156342, "epoch": 1756, "lr": 3.830137968645166e-05}
{"train_loss": 0.13208501040935516, "global_step": 156343, "epoch": 1756, "lr": 3.830081604224581e-05}
{"train_loss": 0.08633848279714584, "global_step": 156344, "epoch": 1756, "lr": 3.83002523996128e-05}
{"train_loss": 0.08376829326152802, "global_step": 156345, "epoch": 1756, "lr": 3.829968875855266e-05}
{"train_loss": 0.08070503175258636, "global_step": 156346, "epoch": 1756, "lr": 3.8299125119065506e-05}
{"train_loss": 0.0918596163392067, "global_step": 156347, "epoch": 1756, "lr": 3.829856148115136e-05}
{"train_loss": 0.13054096698760986, "global_step": 156348, "epoch": 1756, "lr": 3.8297997844810357e-05}
{"train_loss": 0.08894770592451096, "global_step": 156349, "epoch": 1756, "lr": 3.8297434210042524e-05}
{"train_loss": 0.04210001230239868, "global_step": 156350, "epoch": 1756, "lr": 3.829687057684796e-05}
{"train_loss": 0.12231504172086716, "global_step": 156351, "epoch": 1756, "lr": 3.829630694522676e-05}
{"train_loss": 0.08229485154151917, "global_step": 156352, "epoch": 1756, "lr": 3.829574331517896e-05}
{"train_loss": 0.0659179538488388, "global_step": 156353, "epoch": 1756, "lr": 3.829517968670467e-05}
{"train_loss": 0.07402076572179794, "global_step": 156354, "epoch": 1756, "lr": 3.829461605980394e-05}
{"train_loss": 0.09729811549186707, "global_step": 156355, "epoch": 1756, "lr": 3.8294052434476855e-05}
{"train_loss": 0.102219358086586, "global_step": 156356, "epoch": 1756, "lr": 3.82934888107235e-05}
{"train_loss": 0.10545853525400162, "global_step": 156357, "epoch": 1756, "lr": 3.829292518854394e-05}
{"train_loss": 0.06863672286272049, "global_step": 156358, "epoch": 1756, "lr": 3.8292361567938254e-05}
{"train_loss": 0.09666088968515396, "global_step": 156359, "epoch": 1756, "lr": 3.829179794890652e-05}
{"train_loss": 0.11965327709913254, "global_step": 156360, "epoch": 1756, "lr": 3.829123433144879e-05}
{"train_loss": 0.09005525708198547, "global_step": 156361, "epoch": 1756, "lr": 3.8290670715565194e-05}
{"train_loss": 0.07771088182926178, "global_step": 156362, "epoch": 1756, "lr": 3.829010710125575e-05}
{"train_loss": 0.1497829705476761, "global_step": 156363, "epoch": 1756, "lr": 3.828954348852057e-05}
{"train_loss": 0.06780286133289337, "global_step": 156364, "epoch": 1756, "lr": 3.82889798773597e-05}
{"train_loss": 0.07096417248249054, "global_step": 156365, "epoch": 1756, "lr": 3.8288416267773266e-05}
{"train_loss": 0.11912249028682709, "global_step": 156366, "epoch": 1756, "lr": 3.828785265976128e-05}
{"train_loss": 0.0650121346116066, "global_step": 156367, "epoch": 1756, "lr": 3.8287289053323846e-05}
{"train_loss": 0.10326963663101196, "global_step": 156368, "epoch": 1756, "lr": 3.828672544846108e-05}
{"train_loss": 0.04878535121679306, "global_step": 156369, "epoch": 1756, "lr": 3.8286161845172984e-05}
{"train_loss": 0.0926377922296524, "global_step": 156370, "epoch": 1756, "lr": 3.8285598243459705e-05}
{"train_loss": 0.06930742412805557, "global_step": 156371, "epoch": 1756, "lr": 3.8285034643321256e-05}
{"train_loss": 0.10125772948010583, "global_step": 156372, "epoch": 1756, "lr": 3.8284471044757766e-05, "val_loss": 6.59787130355835}
{"train_loss": 0.19544416666030884, "global_step": 156373, "epoch": 1757, "lr": 3.828390744776927e-05}
{"train_loss": 0.12108688801527023, "global_step": 156374, "epoch": 1757, "lr": 3.828334385235586e-05}
{"train_loss": 0.11791026592254639, "global_step": 156375, "epoch": 1757, "lr": 3.828278025851762e-05}
{"train_loss": 0.09866255521774292, "global_step": 156376, "epoch": 1757, "lr": 3.828221666625461e-05}
{"train_loss": 0.10831406712532043, "global_step": 156377, "epoch": 1757, "lr": 3.8281653075566914e-05}
{"train_loss": 0.08342121541500092, "global_step": 156378, "epoch": 1757, "lr": 3.828108948645462e-05}
{"train_loss": 0.09322373569011688, "global_step": 156379, "epoch": 1757, "lr": 3.828052589891776e-05}
{"train_loss": 0.10193478316068649, "global_step": 156380, "epoch": 1757, "lr": 3.827996231295648e-05}
{"train_loss": 0.09789276123046875, "global_step": 156381, "epoch": 1757, "lr": 3.827939872857078e-05}
{"train_loss": 0.13228385150432587, "global_step": 156382, "epoch": 1757, "lr": 3.8278835145760797e-05}
{"train_loss": 0.10944745689630508, "global_step": 156383, "epoch": 1757, "lr": 3.8278271564526564e-05}
{"train_loss": 0.09877797216176987, "global_step": 156384, "epoch": 1757, "lr": 3.827770798486819e-05}
{"train_loss": 0.11484351009130478, "global_step": 156385, "epoch": 1757, "lr": 3.8277144406785716e-05}
{"train_loss": 0.07033639401197433, "global_step": 156386, "epoch": 1757, "lr": 3.827658083027924e-05}
{"train_loss": 0.06136402487754822, "global_step": 156387, "epoch": 1757, "lr": 3.8276017255348853e-05}
{"train_loss": 0.04405101761221886, "global_step": 156388, "epoch": 1757, "lr": 3.8275453681994596e-05}
{"train_loss": 0.08134109526872635, "global_step": 156389, "epoch": 1757, "lr": 3.827489011021658e-05}
{"train_loss": 0.08519663661718369, "global_step": 156390, "epoch": 1757, "lr": 3.8274326540014836e-05}
{"train_loss": 0.07810412347316742, "global_step": 156391, "epoch": 1757, "lr": 3.8273762971389495e-05}
{"train_loss": 0.10826756060123444, "global_step": 156392, "epoch": 1757, "lr": 3.8273199404340585e-05}
{"train_loss": 0.10119643807411194, "global_step": 156393, "epoch": 1757, "lr": 3.8272635838868206e-05}
{"train_loss": 0.2042994499206543, "global_step": 156394, "epoch": 1757, "lr": 3.827207227497243e-05}
{"train_loss": 0.06046772375702858, "global_step": 156395, "epoch": 1757, "lr": 3.827150871265333e-05}
{"train_loss": 0.19081439077854156, "global_step": 156396, "epoch": 1757, "lr": 3.827094515191098e-05}
{"train_loss": 0.1752205789089203, "global_step": 156397, "epoch": 1757, "lr": 3.827038159274545e-05}
{"train_loss": 0.08788997679948807, "global_step": 156398, "epoch": 1757, "lr": 3.8269818035156834e-05}
{"train_loss": 0.08981380611658096, "global_step": 156399, "epoch": 1757, "lr": 3.826925447914521e-05}
{"train_loss": 0.11391782760620117, "global_step": 156400, "epoch": 1757, "lr": 3.826869092471062e-05}
{"train_loss": 0.09789122641086578, "global_step": 156401, "epoch": 1757, "lr": 3.826812737185318e-05}
{"train_loss": 0.06358567625284195, "global_step": 156402, "epoch": 1757, "lr": 3.826756382057293e-05}
{"train_loss": 0.04123394563794136, "global_step": 156403, "epoch": 1757, "lr": 3.826700027086998e-05}
{"train_loss": 0.07728119194507599, "global_step": 156404, "epoch": 1757, "lr": 3.826643672274437e-05}
{"train_loss": 0.102161705493927, "global_step": 156405, "epoch": 1757, "lr": 3.826587317619619e-05}
{"train_loss": 0.11629471182823181, "global_step": 156406, "epoch": 1757, "lr": 3.8265309631225546e-05}
{"train_loss": 0.1010812371969223, "global_step": 156407, "epoch": 1757, "lr": 3.8264746087832464e-05}
{"train_loss": 0.1502261906862259, "global_step": 156408, "epoch": 1757, "lr": 3.826418254601707e-05}
{"train_loss": 0.09311489015817642, "global_step": 156409, "epoch": 1757, "lr": 3.826361900577939e-05}
{"train_loss": 0.06732448935508728, "global_step": 156410, "epoch": 1757, "lr": 3.826305546711954e-05}
{"train_loss": 0.0820593312382698, "global_step": 156411, "epoch": 1757, "lr": 3.8262491930037565e-05}
{"train_loss": 0.11883950978517532, "global_step": 156412, "epoch": 1757, "lr": 3.826192839453358e-05}
{"train_loss": 0.09743139147758484, "global_step": 156413, "epoch": 1757, "lr": 3.8261364860607607e-05}
{"train_loss": 0.04164784029126167, "global_step": 156414, "epoch": 1757, "lr": 3.8260801328259766e-05}
{"train_loss": 0.12051136046648026, "global_step": 156415, "epoch": 1757, "lr": 3.826023779749011e-05}
{"train_loss": 0.08269503712654114, "global_step": 156416, "epoch": 1757, "lr": 3.825967426829873e-05}
{"train_loss": 0.06825047731399536, "global_step": 156417, "epoch": 1757, "lr": 3.8259110740685686e-05}
{"train_loss": 0.11410556733608246, "global_step": 156418, "epoch": 1757, "lr": 3.825854721465107e-05}
{"train_loss": 0.054347287863492966, "global_step": 156419, "epoch": 1757, "lr": 3.8257983690194936e-05}
{"train_loss": 0.04729337617754936, "global_step": 156420, "epoch": 1757, "lr": 3.825742016731739e-05}
{"train_loss": 0.1302761286497116, "global_step": 156421, "epoch": 1757, "lr": 3.8256856646018465e-05}
{"train_loss": 0.03313493728637695, "global_step": 156422, "epoch": 1757, "lr": 3.825629312629827e-05}
{"train_loss": 0.18219347298145294, "global_step": 156423, "epoch": 1757, "lr": 3.8255729608156896e-05}
{"train_loss": 0.08316558599472046, "global_step": 156424, "epoch": 1757, "lr": 3.825516609159437e-05}
{"train_loss": 0.13199320435523987, "global_step": 156425, "epoch": 1757, "lr": 3.825460257661082e-05}
{"train_loss": 0.09775690734386444, "global_step": 156426, "epoch": 1757, "lr": 3.8254039063206266e-05}
{"train_loss": 0.09990806132555008, "global_step": 156427, "epoch": 1757, "lr": 3.8253475551380834e-05}
{"train_loss": 0.06867027282714844, "global_step": 156428, "epoch": 1757, "lr": 3.825291204113457e-05}
{"train_loss": 0.05422539636492729, "global_step": 156429, "epoch": 1757, "lr": 3.825234853246757e-05}
{"train_loss": 0.1077948808670044, "global_step": 156430, "epoch": 1757, "lr": 3.8251785025379876e-05}
{"train_loss": 0.11163008958101273, "global_step": 156431, "epoch": 1757, "lr": 3.8251221519871616e-05}
{"train_loss": 0.051881443709135056, "global_step": 156432, "epoch": 1757, "lr": 3.825065801594282e-05}
{"train_loss": 0.10313597321510315, "global_step": 156433, "epoch": 1757, "lr": 3.825009451359358e-05}
{"train_loss": 0.13513398170471191, "global_step": 156434, "epoch": 1757, "lr": 3.824953101282397e-05}
{"train_loss": 0.08763066679239273, "global_step": 156435, "epoch": 1757, "lr": 3.824896751363407e-05}
{"train_loss": 0.047927286475896835, "global_step": 156436, "epoch": 1757, "lr": 3.8248404016023955e-05}
{"train_loss": 0.03562074527144432, "global_step": 156437, "epoch": 1757, "lr": 3.82478405199937e-05}
{"train_loss": 0.034753650426864624, "global_step": 156438, "epoch": 1757, "lr": 3.824727702554337e-05}
{"train_loss": 0.09959227591753006, "global_step": 156439, "epoch": 1757, "lr": 3.824671353267307e-05}
{"train_loss": 0.09284394979476929, "global_step": 156440, "epoch": 1757, "lr": 3.8246150041382835e-05}
{"train_loss": 0.08927600085735321, "global_step": 156441, "epoch": 1757, "lr": 3.824558655167275e-05}
{"train_loss": 0.10011905431747437, "global_step": 156442, "epoch": 1757, "lr": 3.824502306354294e-05}
{"train_loss": 0.06619589030742645, "global_step": 156443, "epoch": 1757, "lr": 3.824445957699341e-05}
{"train_loss": 0.12685228884220123, "global_step": 156444, "epoch": 1757, "lr": 3.82438960920243e-05}
{"train_loss": 0.04462394863367081, "global_step": 156445, "epoch": 1757, "lr": 3.824333260863563e-05}
{"train_loss": 0.1063244491815567, "global_step": 156446, "epoch": 1757, "lr": 3.824276912682752e-05}
{"train_loss": 0.08844558149576187, "global_step": 156447, "epoch": 1757, "lr": 3.824220564660001e-05}
{"train_loss": 0.1563422977924347, "global_step": 156448, "epoch": 1757, "lr": 3.82416421679532e-05}
{"train_loss": 0.15359431505203247, "global_step": 156449, "epoch": 1757, "lr": 3.824107869088715e-05}
{"train_loss": 0.12349610030651093, "global_step": 156450, "epoch": 1757, "lr": 3.8240515215401964e-05}
{"train_loss": 0.0697995126247406, "global_step": 156451, "epoch": 1757, "lr": 3.823995174149768e-05}
{"train_loss": 0.07066367566585541, "global_step": 156452, "epoch": 1757, "lr": 3.823938826917439e-05}
{"train_loss": 0.11618760228157043, "global_step": 156453, "epoch": 1757, "lr": 3.8238824798432185e-05}
{"train_loss": 0.1070740595459938, "global_step": 156454, "epoch": 1757, "lr": 3.823826132927111e-05}
{"train_loss": 0.13166655600070953, "global_step": 156455, "epoch": 1757, "lr": 3.8237697861691266e-05}
{"train_loss": 0.17022305727005005, "global_step": 156456, "epoch": 1757, "lr": 3.823713439569273e-05}
{"train_loss": 0.14267194271087646, "global_step": 156457, "epoch": 1757, "lr": 3.823657093127555e-05}
{"train_loss": 0.07920199632644653, "global_step": 156458, "epoch": 1757, "lr": 3.823600746843981e-05}
{"train_loss": 0.17606642842292786, "global_step": 156459, "epoch": 1757, "lr": 3.823544400718563e-05}
{"train_loss": 0.054634347558021545, "global_step": 156460, "epoch": 1757, "lr": 3.823488054751302e-05}
{"train_loss": 0.09942300776752193, "global_step": 156461, "epoch": 1757, "lr": 3.823431708942211e-05, "val_loss": 6.69884729385376}
{"train_loss": 0.09630052000284195, "global_step": 156462, "epoch": 1758, "lr": 3.823375363291293e-05}
{"train_loss": 0.08294213563203812, "global_step": 156463, "epoch": 1758, "lr": 3.82331901779856e-05}
{"train_loss": 0.07660667598247528, "global_step": 156464, "epoch": 1758, "lr": 3.823262672464015e-05}
{"train_loss": 0.06637880951166153, "global_step": 156465, "epoch": 1758, "lr": 3.8232063272876714e-05}
{"train_loss": 0.10893842577934265, "global_step": 156466, "epoch": 1758, "lr": 3.8231499822695304e-05}
{"train_loss": 0.06810081005096436, "global_step": 156467, "epoch": 1758, "lr": 3.8230936374096044e-05}
{"train_loss": 0.12393476814031601, "global_step": 156468, "epoch": 1758, "lr": 3.823037292707897e-05}
{"train_loss": 0.07304736971855164, "global_step": 156469, "epoch": 1758, "lr": 3.822980948164421e-05}
{"train_loss": 0.17312993109226227, "global_step": 156470, "epoch": 1758, "lr": 3.822924603779179e-05}
{"train_loss": 0.05114941671490669, "global_step": 156471, "epoch": 1758, "lr": 3.82286825955218e-05}
{"train_loss": 0.10282991826534271, "global_step": 156472, "epoch": 1758, "lr": 3.822811915483433e-05}
{"train_loss": 0.09869007766246796, "global_step": 156473, "epoch": 1758, "lr": 3.822755571572944e-05}
{"train_loss": 0.08303504437208176, "global_step": 156474, "epoch": 1758, "lr": 3.8226992278207216e-05}
{"train_loss": 0.07069042325019836, "global_step": 156475, "epoch": 1758, "lr": 3.822642884226774e-05}
{"train_loss": 0.09526965022087097, "global_step": 156476, "epoch": 1758, "lr": 3.8225865407911044e-05}
{"train_loss": 0.10210932791233063, "global_step": 156477, "epoch": 1758, "lr": 3.822530197513725e-05}
{"train_loss": 0.10251484811306, "global_step": 156478, "epoch": 1758, "lr": 3.8224738543946436e-05}
{"train_loss": 0.0965024009346962, "global_step": 156479, "epoch": 1758, "lr": 3.8224175114338644e-05}
{"train_loss": 0.1373642235994339, "global_step": 156480, "epoch": 1758, "lr": 3.822361168631399e-05}
{"train_loss": 0.12699046730995178, "global_step": 156481, "epoch": 1758, "lr": 3.82230482598725e-05}
{"train_loss": 0.17613495886325836, "global_step": 156482, "epoch": 1758, "lr": 3.822248483501431e-05}
{"train_loss": 0.14023804664611816, "global_step": 156483, "epoch": 1758, "lr": 3.8221921411739434e-05}
{"train_loss": 0.08972641080617905, "global_step": 156484, "epoch": 1758, "lr": 3.8221357990048e-05}
{"train_loss": 0.11549210548400879, "global_step": 156485, "epoch": 1758, "lr": 3.822079456994004e-05}
{"train_loss": 0.1144280880689621, "global_step": 156486, "epoch": 1758, "lr": 3.822023115141567e-05}
{"train_loss": 0.13684207201004028, "global_step": 156487, "epoch": 1758, "lr": 3.8219667734474926e-05}
{"train_loss": 0.09245721250772476, "global_step": 156488, "epoch": 1758, "lr": 3.821910431911793e-05}
{"train_loss": 0.07818534970283508, "global_step": 156489, "epoch": 1758, "lr": 3.821854090534471e-05}
{"train_loss": 0.12632973492145538, "global_step": 156490, "epoch": 1758, "lr": 3.821797749315537e-05}
{"train_loss": 0.06856241077184677, "global_step": 156491, "epoch": 1758, "lr": 3.8217414082549984e-05}
{"train_loss": 0.08173403143882751, "global_step": 156492, "epoch": 1758, "lr": 3.821685067352862e-05}
{"train_loss": 0.10116878896951675, "global_step": 156493, "epoch": 1758, "lr": 3.8216287266091355e-05}
{"train_loss": 0.09696413576602936, "global_step": 156494, "epoch": 1758, "lr": 3.8215723860238256e-05}
{"train_loss": 0.07237192243337631, "global_step": 156495, "epoch": 1758, "lr": 3.821516045596943e-05}
{"train_loss": 0.09689777344465256, "global_step": 156496, "epoch": 1758, "lr": 3.8214597053284915e-05}
{"train_loss": 0.11772220581769943, "global_step": 156497, "epoch": 1758, "lr": 3.821403365218482e-05}
{"train_loss": 0.0772460550069809, "global_step": 156498, "epoch": 1758, "lr": 3.821347025266919e-05}
{"train_loss": 0.09873995929956436, "global_step": 156499, "epoch": 1758, "lr": 3.8212906854738126e-05}
{"train_loss": 0.06927597522735596, "global_step": 156500, "epoch": 1758, "lr": 3.821234345839168e-05}
{"train_loss": 0.06120440363883972, "global_step": 156501, "epoch": 1758, "lr": 3.8211780063629966e-05}
{"train_loss": 0.07518786936998367, "global_step": 156502, "epoch": 1758, "lr": 3.821121667045301e-05}
{"train_loss": 0.12938570976257324, "global_step": 156503, "epoch": 1758, "lr": 3.821065327886093e-05}
{"train_loss": 0.1545739322900772, "global_step": 156504, "epoch": 1758, "lr": 3.8210089888853754e-05}
{"train_loss": 0.1298220008611679, "global_step": 156505, "epoch": 1758, "lr": 3.820952650043162e-05}
{"train_loss": 0.10884755849838257, "global_step": 156506, "epoch": 1758, "lr": 3.820896311359455e-05}
{"train_loss": 0.06851412355899811, "global_step": 156507, "epoch": 1758, "lr": 3.820839972834266e-05}
{"train_loss": 0.11020837724208832, "global_step": 156508, "epoch": 1758, "lr": 3.8207836344676e-05}
{"train_loss": 0.07658891379833221, "global_step": 156509, "epoch": 1758, "lr": 3.820727296259465e-05}
{"train_loss": 0.14243432879447937, "global_step": 156510, "epoch": 1758, "lr": 3.8206709582098686e-05}
{"train_loss": 0.03702964633703232, "global_step": 156511, "epoch": 1758, "lr": 3.820614620318819e-05}
{"train_loss": 0.05418771132826805, "global_step": 156512, "epoch": 1758, "lr": 3.820558282586323e-05}
{"train_loss": 0.039046019315719604, "global_step": 156513, "epoch": 1758, "lr": 3.820501945012388e-05}
{"train_loss": 0.053810928016901016, "global_step": 156514, "epoch": 1758, "lr": 3.820445607597024e-05}
{"train_loss": 0.1024363711476326, "global_step": 156515, "epoch": 1758, "lr": 3.820389270340234e-05}
{"train_loss": 0.12659485638141632, "global_step": 156516, "epoch": 1758, "lr": 3.820332933242031e-05}
{"train_loss": 0.14868590235710144, "global_step": 156517, "epoch": 1758, "lr": 3.820276596302418e-05}
{"train_loss": 0.05487629026174545, "global_step": 156518, "epoch": 1758, "lr": 3.820220259521407e-05}
{"train_loss": 0.14729706943035126, "global_step": 156519, "epoch": 1758, "lr": 3.8201639228989994e-05}
{"train_loss": 0.044109731912612915, "global_step": 156520, "epoch": 1758, "lr": 3.8201075864352095e-05}
{"train_loss": 0.09981070458889008, "global_step": 156521, "epoch": 1758, "lr": 3.82005125013004e-05}
{"train_loss": 0.10164675861597061, "global_step": 156522, "epoch": 1758, "lr": 3.819994913983501e-05}
{"train_loss": 0.10771165043115616, "global_step": 156523, "epoch": 1758, "lr": 3.819938577995598e-05}
{"train_loss": 0.0908622145652771, "global_step": 156524, "epoch": 1758, "lr": 3.819882242166342e-05}
{"train_loss": 0.1352110654115677, "global_step": 156525, "epoch": 1758, "lr": 3.819825906495736e-05}
{"train_loss": 0.054342761635780334, "global_step": 156526, "epoch": 1758, "lr": 3.8197695709837925e-05}
{"train_loss": 0.09635661542415619, "global_step": 156527, "epoch": 1758, "lr": 3.819713235630515e-05}
{"train_loss": 0.11311841756105423, "global_step": 156528, "epoch": 1758, "lr": 3.819656900435914e-05}
{"train_loss": 0.09479349106550217, "global_step": 156529, "epoch": 1758, "lr": 3.8196005653999944e-05}
{"train_loss": 0.08733335137367249, "global_step": 156530, "epoch": 1758, "lr": 3.819544230522766e-05}
{"train_loss": 0.07159432768821716, "global_step": 156531, "epoch": 1758, "lr": 3.819487895804235e-05}
{"train_loss": 0.11855052411556244, "global_step": 156532, "epoch": 1758, "lr": 3.8194315612444085e-05}
{"train_loss": 0.09590897709131241, "global_step": 156533, "epoch": 1758, "lr": 3.819375226843297e-05}
{"train_loss": 0.08572240173816681, "global_step": 156534, "epoch": 1758, "lr": 3.8193188926009045e-05}
{"train_loss": 0.07913106679916382, "global_step": 156535, "epoch": 1758, "lr": 3.819262558517241e-05}
{"train_loss": 0.1370318979024887, "global_step": 156536, "epoch": 1758, "lr": 3.819206224592313e-05}
{"train_loss": 0.12138397991657257, "global_step": 156537, "epoch": 1758, "lr": 3.819149890826129e-05}
{"train_loss": 0.11301085352897644, "global_step": 156538, "epoch": 1758, "lr": 3.819093557218695e-05}
{"train_loss": 0.1220124363899231, "global_step": 156539, "epoch": 1758, "lr": 3.81903722377002e-05}
{"train_loss": 0.12654952704906464, "global_step": 156540, "epoch": 1758, "lr": 3.81898089048011e-05}
{"train_loss": 0.09398332983255386, "global_step": 156541, "epoch": 1758, "lr": 3.8189245573489756e-05}
{"train_loss": 0.0895419791340828, "global_step": 156542, "epoch": 1758, "lr": 3.81886822437662e-05}
{"train_loss": 0.12546245753765106, "global_step": 156543, "epoch": 1758, "lr": 3.818811891563056e-05}
{"train_loss": 0.07097511738538742, "global_step": 156544, "epoch": 1758, "lr": 3.8187555589082857e-05}
{"train_loss": 0.12414562702178955, "global_step": 156545, "epoch": 1758, "lr": 3.818699226412322e-05}
{"train_loss": 0.07768341153860092, "global_step": 156546, "epoch": 1758, "lr": 3.818642894075167e-05}
{"train_loss": 0.11500716954469681, "global_step": 156547, "epoch": 1758, "lr": 3.818586561896833e-05}
{"train_loss": 0.11416083574295044, "global_step": 156548, "epoch": 1758, "lr": 3.8185302298773254e-05}
{"train_loss": 0.1289331465959549, "global_step": 156549, "epoch": 1758, "lr": 3.81847389801665e-05}
{"train_loss": 0.0993582107796428, "global_step": 156550, "epoch": 1758, "lr": 3.818417566314819e-05, "val_loss": 6.442248344421387}
{"train_loss": 0.10993853956460953, "global_step": 156551, "epoch": 1759, "lr": 3.818361234771836e-05}
{"train_loss": 0.0688248947262764, "global_step": 156552, "epoch": 1759, "lr": 3.818304903387711e-05}
{"train_loss": 0.07451529800891876, "global_step": 156553, "epoch": 1759, "lr": 3.818248572162448e-05}
{"train_loss": 0.0705779418349266, "global_step": 156554, "epoch": 1759, "lr": 3.81819224109606e-05}
{"train_loss": 0.10300462692975998, "global_step": 156555, "epoch": 1759, "lr": 3.8181359101885495e-05}
{"train_loss": 0.11063028126955032, "global_step": 156556, "epoch": 1759, "lr": 3.818079579439928e-05}
{"train_loss": 0.09239029884338379, "global_step": 156557, "epoch": 1759, "lr": 3.8180232488502e-05}
{"train_loss": 0.08351985365152359, "global_step": 156558, "epoch": 1759, "lr": 3.817966918419376e-05}
{"train_loss": 0.10892296582460403, "global_step": 156559, "epoch": 1759, "lr": 3.817910588147459e-05}
{"train_loss": 0.12575183808803558, "global_step": 156560, "epoch": 1759, "lr": 3.8178542580344626e-05}
{"train_loss": 0.05430454760789871, "global_step": 156561, "epoch": 1759, "lr": 3.8177979280803886e-05}
{"train_loss": 0.07977953553199768, "global_step": 156562, "epoch": 1759, "lr": 3.81774159828525e-05}
{"train_loss": 0.03178524598479271, "global_step": 156563, "epoch": 1759, "lr": 3.8176852686490496e-05}
{"train_loss": 0.06259053945541382, "global_step": 156564, "epoch": 1759, "lr": 3.817628939171798e-05}
{"train_loss": 0.10098201781511307, "global_step": 156565, "epoch": 1759, "lr": 3.817572609853501e-05}
{"train_loss": 0.18610666692256927, "global_step": 156566, "epoch": 1759, "lr": 3.8175162806941674e-05}
{"train_loss": 0.07133089751005173, "global_step": 156567, "epoch": 1759, "lr": 3.817459951693805e-05}
{"train_loss": 0.11066427826881409, "global_step": 156568, "epoch": 1759, "lr": 3.817403622852419e-05}
{"train_loss": 0.10197285562753677, "global_step": 156569, "epoch": 1759, "lr": 3.817347294170021e-05}
{"train_loss": 0.09007410705089569, "global_step": 156570, "epoch": 1759, "lr": 3.817290965646614e-05}
{"train_loss": 0.09701447188854218, "global_step": 156571, "epoch": 1759, "lr": 3.8172346372822093e-05}
{"train_loss": 0.02429487742483616, "global_step": 156572, "epoch": 1759, "lr": 3.817178309076812e-05}
{"train_loss": 0.12610165774822235, "global_step": 156573, "epoch": 1759, "lr": 3.8171219810304315e-05}
{"train_loss": 0.04985817149281502, "global_step": 156574, "epoch": 1759, "lr": 3.817065653143073e-05}
{"train_loss": 0.06305801123380661, "global_step": 156575, "epoch": 1759, "lr": 3.8170093254147475e-05}
{"train_loss": 0.08840595930814743, "global_step": 156576, "epoch": 1759, "lr": 3.8169529978454595e-05}
{"train_loss": 0.035244911909103394, "global_step": 156577, "epoch": 1759, "lr": 3.816896670435219e-05}
{"train_loss": 0.062390878796577454, "global_step": 156578, "epoch": 1759, "lr": 3.816840343184031e-05}
{"train_loss": 0.10514786094427109, "global_step": 156579, "epoch": 1759, "lr": 3.8167840160919055e-05}
{"train_loss": 0.18697945773601532, "global_step": 156580, "epoch": 1759, "lr": 3.816727689158847e-05}
{"train_loss": 0.10011190921068192, "global_step": 156581, "epoch": 1759, "lr": 3.8166713623848674e-05}
{"train_loss": 0.0876452699303627, "global_step": 156582, "epoch": 1759, "lr": 3.81661503576997e-05}
{"train_loss": 0.07169843465089798, "global_step": 156583, "epoch": 1759, "lr": 3.8165587093141664e-05}
{"train_loss": 0.08229229599237442, "global_step": 156584, "epoch": 1759, "lr": 3.8165023830174594e-05}
{"train_loss": 0.13061577081680298, "global_step": 156585, "epoch": 1759, "lr": 3.816446056879861e-05}
{"train_loss": 0.12231110781431198, "global_step": 156586, "epoch": 1759, "lr": 3.8163897309013775e-05}
{"train_loss": 0.04977301135659218, "global_step": 156587, "epoch": 1759, "lr": 3.816333405082014e-05}
{"train_loss": 0.1592232584953308, "global_step": 156588, "epoch": 1759, "lr": 3.8162770794217825e-05}
{"train_loss": 0.06494152545928955, "global_step": 156589, "epoch": 1759, "lr": 3.8162207539206854e-05}
{"train_loss": 0.12166547775268555, "global_step": 156590, "epoch": 1759, "lr": 3.8161644285787356e-05}
{"train_loss": 0.09026075899600983, "global_step": 156591, "epoch": 1759, "lr": 3.816108103395936e-05}
{"train_loss": 0.1414012610912323, "global_step": 156592, "epoch": 1759, "lr": 3.816051778372298e-05}
{"train_loss": 0.08666073530912399, "global_step": 156593, "epoch": 1759, "lr": 3.8159954535078256e-05}
{"train_loss": 0.09641481935977936, "global_step": 156594, "epoch": 1759, "lr": 3.81593912880253e-05}
{"train_loss": 0.11714117228984833, "global_step": 156595, "epoch": 1759, "lr": 3.8158828042564145e-05}
{"train_loss": 0.1403944492340088, "global_step": 156596, "epoch": 1759, "lr": 3.8158264798694923e-05}
{"train_loss": 0.14410993456840515, "global_step": 156597, "epoch": 1759, "lr": 3.815770155641765e-05}
{"train_loss": 0.07731293886899948, "global_step": 156598, "epoch": 1759, "lr": 3.815713831573246e-05}
{"train_loss": 0.15858770906925201, "global_step": 156599, "epoch": 1759, "lr": 3.815657507663936e-05}
{"train_loss": 0.14339222013950348, "global_step": 156600, "epoch": 1759, "lr": 3.8156011839138494e-05}
{"train_loss": 0.1070547103881836, "global_step": 156601, "epoch": 1759, "lr": 3.815544860322989e-05}
{"train_loss": 0.1438814252614975, "global_step": 156602, "epoch": 1759, "lr": 3.815488536891366e-05}
{"train_loss": 0.14883682131767273, "global_step": 156603, "epoch": 1759, "lr": 3.815432213618984e-05}
{"train_loss": 0.13736166059970856, "global_step": 156604, "epoch": 1759, "lr": 3.815375890505853e-05}
{"train_loss": 0.10573811829090118, "global_step": 156605, "epoch": 1759, "lr": 3.8153195675519826e-05}
{"train_loss": 0.1008085384964943, "global_step": 156606, "epoch": 1759, "lr": 3.815263244757375e-05}
{"train_loss": 0.07650324702262878, "global_step": 156607, "epoch": 1759, "lr": 3.815206922122044e-05}
{"train_loss": 0.049344029277563095, "global_step": 156608, "epoch": 1759, "lr": 3.8151505996459916e-05}
{"train_loss": 0.10170215368270874, "global_step": 156609, "epoch": 1759, "lr": 3.8150942773292294e-05}
{"train_loss": 0.11981087923049927, "global_step": 156610, "epoch": 1759, "lr": 3.815037955171761e-05}
{"train_loss": 0.07341842353343964, "global_step": 156611, "epoch": 1759, "lr": 3.814981633173599e-05}
{"train_loss": 0.07859086245298386, "global_step": 156612, "epoch": 1759, "lr": 3.814925311334746e-05}
{"train_loss": 0.14330336451530457, "global_step": 156613, "epoch": 1759, "lr": 3.814868989655214e-05}
{"train_loss": 0.08824891597032547, "global_step": 156614, "epoch": 1759, "lr": 3.8148126681350056e-05}
{"train_loss": 0.08663035929203033, "global_step": 156615, "epoch": 1759, "lr": 3.814756346774134e-05}
{"train_loss": 0.1557978242635727, "global_step": 156616, "epoch": 1759, "lr": 3.8147000255726015e-05}
{"train_loss": 0.12122022360563278, "global_step": 156617, "epoch": 1759, "lr": 3.814643704530421e-05}
{"train_loss": 0.15524831414222717, "global_step": 156618, "epoch": 1759, "lr": 3.8145873836475945e-05}
{"train_loss": 0.11883524805307388, "global_step": 156619, "epoch": 1759, "lr": 3.814531062924134e-05}
{"train_loss": 0.060814935714006424, "global_step": 156620, "epoch": 1759, "lr": 3.8144747423600434e-05}
{"train_loss": 0.11544085294008255, "global_step": 156621, "epoch": 1759, "lr": 3.8144184219553335e-05}
{"train_loss": 0.0971667543053627, "global_step": 156622, "epoch": 1759, "lr": 3.8143621017100114e-05}
{"train_loss": 0.1150692030787468, "global_step": 156623, "epoch": 1759, "lr": 3.8143057816240826e-05}
{"train_loss": 0.15489313006401062, "global_step": 156624, "epoch": 1759, "lr": 3.8142494616975566e-05}
{"train_loss": 0.06562487035989761, "global_step": 156625, "epoch": 1759, "lr": 3.81419314193044e-05}
{"train_loss": 0.12194167822599411, "global_step": 156626, "epoch": 1759, "lr": 3.814136822322742e-05}
{"train_loss": 0.08616423606872559, "global_step": 156627, "epoch": 1759, "lr": 3.8140805028744666e-05}
{"train_loss": 0.11662353575229645, "global_step": 156628, "epoch": 1759, "lr": 3.814024183585626e-05}
{"train_loss": 0.09907682240009308, "global_step": 156629, "epoch": 1759, "lr": 3.813967864456223e-05}
{"train_loss": 0.07256975769996643, "global_step": 156630, "epoch": 1759, "lr": 3.813911545486269e-05}
{"train_loss": 0.08196859061717987, "global_step": 156631, "epoch": 1759, "lr": 3.8138552266757695e-05}
{"train_loss": 0.168159618973732, "global_step": 156632, "epoch": 1759, "lr": 3.8137989080247335e-05}
{"train_loss": 0.11011391133069992, "global_step": 156633, "epoch": 1759, "lr": 3.813742589533167e-05}
{"train_loss": 0.12164703011512756, "global_step": 156634, "epoch": 1759, "lr": 3.813686271201079e-05}
{"train_loss": 0.14212265610694885, "global_step": 156635, "epoch": 1759, "lr": 3.813629953028476e-05}
{"train_loss": 0.10499772429466248, "global_step": 156636, "epoch": 1759, "lr": 3.813573635015367e-05}
{"train_loss": 0.09090634435415268, "global_step": 156637, "epoch": 1759, "lr": 3.813517317161756e-05}
{"train_loss": 0.0809171125292778, "global_step": 156638, "epoch": 1759, "lr": 3.813460999467656e-05}
{"train_loss": 0.1016733002419887, "global_step": 156639, "epoch": 1759, "lr": 3.813404681933069e-05, "val_loss": 6.406307220458984}
{"train_loss": 0.12099854648113251, "global_step": 156640, "epoch": 1760, "lr": 3.813348364558007e-05}
{"train_loss": 0.11687464267015457, "global_step": 156641, "epoch": 1760, "lr": 3.813292047342476e-05}
{"train_loss": 0.1683521419763565, "global_step": 156642, "epoch": 1760, "lr": 3.8132357302864826e-05}
{"train_loss": 0.06600594520568848, "global_step": 156643, "epoch": 1760, "lr": 3.8131794133900354e-05}
{"train_loss": 0.13012388348579407, "global_step": 156644, "epoch": 1760, "lr": 3.813123096653142e-05}
{"train_loss": 0.1484450250864029, "global_step": 156645, "epoch": 1760, "lr": 3.8130667800758115e-05}
{"train_loss": 0.0957721620798111, "global_step": 156646, "epoch": 1760, "lr": 3.813010463658046e-05}
{"train_loss": 0.1061234176158905, "global_step": 156647, "epoch": 1760, "lr": 3.81295414739986e-05}
{"train_loss": 0.09301778674125671, "global_step": 156648, "epoch": 1760, "lr": 3.8128978313012555e-05}
{"train_loss": 0.0701812207698822, "global_step": 156649, "epoch": 1760, "lr": 3.812841515362245e-05}
{"train_loss": 0.1106015294790268, "global_step": 156650, "epoch": 1760, "lr": 3.812785199582831e-05}
{"train_loss": 0.19050921499729156, "global_step": 156651, "epoch": 1760, "lr": 3.812728883963025e-05}
{"train_loss": 0.08384624123573303, "global_step": 156652, "epoch": 1760, "lr": 3.812672568502832e-05}
{"train_loss": 0.07739841938018799, "global_step": 156653, "epoch": 1760, "lr": 3.812616253202262e-05}
{"train_loss": 0.09828642755746841, "global_step": 156654, "epoch": 1760, "lr": 3.8125599380613195e-05}
{"train_loss": 0.1157083809375763, "global_step": 156655, "epoch": 1760, "lr": 3.812503623080016e-05}
{"train_loss": 0.07683365046977997, "global_step": 156656, "epoch": 1760, "lr": 3.812447308258355e-05}
{"train_loss": 0.0849374309182167, "global_step": 156657, "epoch": 1760, "lr": 3.812390993596347e-05}
{"train_loss": 0.11225630342960358, "global_step": 156658, "epoch": 1760, "lr": 3.8123346790939987e-05}
{"train_loss": 0.1246272400021553, "global_step": 156659, "epoch": 1760, "lr": 3.8122783647513165e-05}
{"train_loss": 0.12038129568099976, "global_step": 156660, "epoch": 1760, "lr": 3.81222205056831e-05}
{"train_loss": 0.17025910317897797, "global_step": 156661, "epoch": 1760, "lr": 3.812165736544985e-05}
{"train_loss": 0.08159938454627991, "global_step": 156662, "epoch": 1760, "lr": 3.812109422681351e-05}
{"train_loss": 0.0797296017408371, "global_step": 156663, "epoch": 1760, "lr": 3.8120531089774136e-05}
{"train_loss": 0.060170143842697144, "global_step": 156664, "epoch": 1760, "lr": 3.8119967954331826e-05}
{"train_loss": 0.1391233503818512, "global_step": 156665, "epoch": 1760, "lr": 3.8119404820486613e-05}
{"train_loss": 0.13864709436893463, "global_step": 156666, "epoch": 1760, "lr": 3.811884168823863e-05}
{"train_loss": 0.11972235143184662, "global_step": 156667, "epoch": 1760, "lr": 3.81182785575879e-05}
{"train_loss": 0.16270311176776886, "global_step": 156668, "epoch": 1760, "lr": 3.8117715428534544e-05}
{"train_loss": 0.11249171942472458, "global_step": 156669, "epoch": 1760, "lr": 3.8117152301078597e-05}
{"train_loss": 0.11209773272275925, "global_step": 156670, "epoch": 1760, "lr": 3.811658917522017e-05}
{"train_loss": 0.10982945561408997, "global_step": 156671, "epoch": 1760, "lr": 3.811602605095931e-05}
{"train_loss": 0.15041033923625946, "global_step": 156672, "epoch": 1760, "lr": 3.8115462928296117e-05}
{"train_loss": 0.12160065025091171, "global_step": 156673, "epoch": 1760, "lr": 3.811489980723064e-05}
{"train_loss": 0.06487450748682022, "global_step": 156674, "epoch": 1760, "lr": 3.811433668776299e-05}
{"train_loss": 0.07851998507976532, "global_step": 156675, "epoch": 1760, "lr": 3.81137735698932e-05}
{"train_loss": 0.10460011661052704, "global_step": 156676, "epoch": 1760, "lr": 3.811321045362137e-05}
{"train_loss": 0.09931302070617676, "global_step": 156677, "epoch": 1760, "lr": 3.811264733894759e-05}
{"train_loss": 0.11715298891067505, "global_step": 156678, "epoch": 1760, "lr": 3.811208422587192e-05}
{"train_loss": 0.07122774422168732, "global_step": 156679, "epoch": 1760, "lr": 3.811152111439443e-05}
{"train_loss": 0.08215582370758057, "global_step": 156680, "epoch": 1760, "lr": 3.8110958004515194e-05}
{"train_loss": 0.16209620237350464, "global_step": 156681, "epoch": 1760, "lr": 3.81103948962343e-05}
{"train_loss": 0.13166332244873047, "global_step": 156682, "epoch": 1760, "lr": 3.8109831789551825e-05}
{"train_loss": 0.0942888930439949, "global_step": 156683, "epoch": 1760, "lr": 3.810926868446784e-05}
{"train_loss": 0.09176904708147049, "global_step": 156684, "epoch": 1760, "lr": 3.81087055809824e-05}
{"train_loss": 0.11124784499406815, "global_step": 156685, "epoch": 1760, "lr": 3.810814247909563e-05}
{"train_loss": 0.09590481966733932, "global_step": 156686, "epoch": 1760, "lr": 3.810757937880754e-05}
{"train_loss": 0.12839330732822418, "global_step": 156687, "epoch": 1760, "lr": 3.810701628011827e-05}
{"train_loss": 0.09729840606451035, "global_step": 156688, "epoch": 1760, "lr": 3.810645318302784e-05}
{"train_loss": 0.10594209283590317, "global_step": 156689, "epoch": 1760, "lr": 3.810589008753638e-05}
{"train_loss": 0.09097671508789062, "global_step": 156690, "epoch": 1760, "lr": 3.8105326993643915e-05}
{"train_loss": 0.10099951922893524, "global_step": 156691, "epoch": 1760, "lr": 3.810476390135056e-05}
{"train_loss": 0.12612797319889069, "global_step": 156692, "epoch": 1760, "lr": 3.810420081065636e-05}
{"train_loss": 0.1588112860918045, "global_step": 156693, "epoch": 1760, "lr": 3.810363772156142e-05}
{"train_loss": 0.12686191499233246, "global_step": 156694, "epoch": 1760, "lr": 3.8103074634065784e-05}
{"train_loss": 0.09551582485437393, "global_step": 156695, "epoch": 1760, "lr": 3.810251154816954e-05}
{"train_loss": 0.05287837237119675, "global_step": 156696, "epoch": 1760, "lr": 3.81019484638728e-05}
{"train_loss": 0.15647198259830475, "global_step": 156697, "epoch": 1760, "lr": 3.8101385381175594e-05}
{"train_loss": 0.06912712752819061, "global_step": 156698, "epoch": 1760, "lr": 3.8100822300078e-05}
{"train_loss": 0.09612251073122025, "global_step": 156699, "epoch": 1760, "lr": 3.810025922058012e-05}
{"train_loss": 0.1705283224582672, "global_step": 156700, "epoch": 1760, "lr": 3.809969614268201e-05}
{"train_loss": 0.07051587104797363, "global_step": 156701, "epoch": 1760, "lr": 3.8099133066383754e-05}
{"train_loss": 0.1517094373703003, "global_step": 156702, "epoch": 1760, "lr": 3.809856999168543e-05}
{"train_loss": 0.15499798953533173, "global_step": 156703, "epoch": 1760, "lr": 3.809800691858708e-05}
{"train_loss": 0.09373440593481064, "global_step": 156704, "epoch": 1760, "lr": 3.809744384708884e-05}
{"train_loss": 0.09471704065799713, "global_step": 156705, "epoch": 1760, "lr": 3.809688077719073e-05}
{"train_loss": 0.1302219182252884, "global_step": 156706, "epoch": 1760, "lr": 3.809631770889287e-05}
{"train_loss": 0.14711900055408478, "global_step": 156707, "epoch": 1760, "lr": 3.809575464219529e-05}
{"train_loss": 0.061529021710157394, "global_step": 156708, "epoch": 1760, "lr": 3.809519157709812e-05}
{"train_loss": 0.10467006266117096, "global_step": 156709, "epoch": 1760, "lr": 3.8094628513601385e-05}
{"train_loss": 0.10418163239955902, "global_step": 156710, "epoch": 1760, "lr": 3.80940654517052e-05}
{"train_loss": 0.07652194797992706, "global_step": 156711, "epoch": 1760, "lr": 3.80935023914096e-05}
{"train_loss": 0.0985693484544754, "global_step": 156712, "epoch": 1760, "lr": 3.8092939332714687e-05}
{"train_loss": 0.08790884912014008, "global_step": 156713, "epoch": 1760, "lr": 3.8092376275620546e-05}
{"train_loss": 0.08445896953344345, "global_step": 156714, "epoch": 1760, "lr": 3.809181322012722e-05}
{"train_loss": 0.07994028925895691, "global_step": 156715, "epoch": 1760, "lr": 3.809125016623484e-05}
{"train_loss": 0.09327384829521179, "global_step": 156716, "epoch": 1760, "lr": 3.8090687113943416e-05}
{"train_loss": 0.12312661856412888, "global_step": 156717, "epoch": 1760, "lr": 3.809012406325307e-05}
{"train_loss": 0.06173519417643547, "global_step": 156718, "epoch": 1760, "lr": 3.8089561014163857e-05}
{"train_loss": 0.09369146078824997, "global_step": 156719, "epoch": 1760, "lr": 3.8088997966675856e-05}
{"train_loss": 0.12474952638149261, "global_step": 156720, "epoch": 1760, "lr": 3.808843492078914e-05}
{"train_loss": 0.12270447611808777, "global_step": 156721, "epoch": 1760, "lr": 3.8087871876503804e-05}
{"train_loss": 0.143976628780365, "global_step": 156722, "epoch": 1760, "lr": 3.808730883381988e-05}
{"train_loss": 0.14094869792461395, "global_step": 156723, "epoch": 1760, "lr": 3.8086745792737504e-05}
{"train_loss": 0.1366812139749527, "global_step": 156724, "epoch": 1760, "lr": 3.8086182753256695e-05}
{"train_loss": 0.10894033312797546, "global_step": 156725, "epoch": 1760, "lr": 3.808561971537757e-05}
{"train_loss": 0.10698199272155762, "global_step": 156726, "epoch": 1760, "lr": 3.808505667910018e-05}
{"train_loss": 0.13927017152309418, "global_step": 156727, "epoch": 1760, "lr": 3.8084493644424616e-05}
{"train_loss": 0.10961053495326739, "global_step": 156728, "epoch": 1760, "lr": 3.8083930611350934e-05, "val_loss": 6.537821292877197, "train_action_mse_error": 9.23406982421875}
{"train_loss": 0.04528870806097984, "global_step": 156729, "epoch": 1761, "lr": 3.8083367579879235e-05}
{"train_loss": 0.0642470121383667, "global_step": 156730, "epoch": 1761, "lr": 3.8082804550009574e-05}
{"train_loss": 0.10667945444583893, "global_step": 156731, "epoch": 1761, "lr": 3.8082241521742024e-05}
{"train_loss": 0.07038635015487671, "global_step": 156732, "epoch": 1761, "lr": 3.8081678495076695e-05}
{"train_loss": 0.08610934019088745, "global_step": 156733, "epoch": 1761, "lr": 3.808111547001362e-05}
{"train_loss": 0.1041693389415741, "global_step": 156734, "epoch": 1761, "lr": 3.808055244655291e-05}
{"train_loss": 0.0345928817987442, "global_step": 156735, "epoch": 1761, "lr": 3.807998942469462e-05}
{"train_loss": 0.1230870932340622, "global_step": 156736, "epoch": 1761, "lr": 3.807942640443882e-05}
{"train_loss": 0.10606921464204788, "global_step": 156737, "epoch": 1761, "lr": 3.8078863385785606e-05}
{"train_loss": 0.0756794810295105, "global_step": 156738, "epoch": 1761, "lr": 3.807830036873504e-05}
{"train_loss": 0.11255427449941635, "global_step": 156739, "epoch": 1761, "lr": 3.80777373532872e-05}
{"train_loss": 0.06925303488969803, "global_step": 156740, "epoch": 1761, "lr": 3.807717433944217e-05}
{"train_loss": 0.08376429229974747, "global_step": 156741, "epoch": 1761, "lr": 3.807661132720001e-05}
{"train_loss": 0.11077715456485748, "global_step": 156742, "epoch": 1761, "lr": 3.807604831656082e-05}
{"train_loss": 0.11052753776311874, "global_step": 156743, "epoch": 1761, "lr": 3.807548530752463e-05}
{"train_loss": 0.1426636278629303, "global_step": 156744, "epoch": 1761, "lr": 3.807492230009157e-05}
{"train_loss": 0.05983543023467064, "global_step": 156745, "epoch": 1761, "lr": 3.807435929426168e-05}
{"train_loss": 0.08939877152442932, "global_step": 156746, "epoch": 1761, "lr": 3.807379629003506e-05}
{"train_loss": 0.09074107557535172, "global_step": 156747, "epoch": 1761, "lr": 3.807323328741176e-05}
{"train_loss": 0.06633973866701126, "global_step": 156748, "epoch": 1761, "lr": 3.807267028639186e-05}
{"train_loss": 0.08966723084449768, "global_step": 156749, "epoch": 1761, "lr": 3.8072107286975464e-05}
{"train_loss": 0.09639495611190796, "global_step": 156750, "epoch": 1761, "lr": 3.807154428916261e-05}
{"train_loss": 0.10703150928020477, "global_step": 156751, "epoch": 1761, "lr": 3.807098129295341e-05}
{"train_loss": 0.13903595507144928, "global_step": 156752, "epoch": 1761, "lr": 3.807041829834791e-05}
{"train_loss": 0.1453401744365692, "global_step": 156753, "epoch": 1761, "lr": 3.8069855305346206e-05}
{"train_loss": 0.07330676913261414, "global_step": 156754, "epoch": 1761, "lr": 3.806929231394835e-05}
{"train_loss": 0.07983323931694031, "global_step": 156755, "epoch": 1761, "lr": 3.806872932415445e-05}
{"train_loss": 0.07541420310735703, "global_step": 156756, "epoch": 1761, "lr": 3.8068166335964556e-05}
{"train_loss": 0.08034330606460571, "global_step": 156757, "epoch": 1761, "lr": 3.806760334937875e-05}
{"train_loss": 0.142231285572052, "global_step": 156758, "epoch": 1761, "lr": 3.8067040364397114e-05}
{"train_loss": 0.10136716812849045, "global_step": 156759, "epoch": 1761, "lr": 3.806647738101971e-05}
{"train_loss": 0.0720967948436737, "global_step": 156760, "epoch": 1761, "lr": 3.8065914399246624e-05}
{"train_loss": 0.15603239834308624, "global_step": 156761, "epoch": 1761, "lr": 3.806535141907794e-05}
{"train_loss": 0.08518392592668533, "global_step": 156762, "epoch": 1761, "lr": 3.806478844051372e-05}
{"train_loss": 0.08442208915948868, "global_step": 156763, "epoch": 1761, "lr": 3.806422546355405e-05}
{"train_loss": 0.12708820402622223, "global_step": 156764, "epoch": 1761, "lr": 3.8063662488198975e-05}
{"train_loss": 0.17553704977035522, "global_step": 156765, "epoch": 1761, "lr": 3.806309951444862e-05}
{"train_loss": 0.09669385105371475, "global_step": 156766, "epoch": 1761, "lr": 3.806253654230302e-05}
{"train_loss": 0.08423546701669693, "global_step": 156767, "epoch": 1761, "lr": 3.806197357176226e-05}
{"train_loss": 0.1276981085538864, "global_step": 156768, "epoch": 1761, "lr": 3.8061410602826445e-05}
{"train_loss": 0.08458855003118515, "global_step": 156769, "epoch": 1761, "lr": 3.806084763549561e-05}
{"train_loss": 0.10961808264255524, "global_step": 156770, "epoch": 1761, "lr": 3.8060284669769865e-05}
{"train_loss": 0.12554717063903809, "global_step": 156771, "epoch": 1761, "lr": 3.805972170564925e-05}
{"train_loss": 0.10803113132715225, "global_step": 156772, "epoch": 1761, "lr": 3.805915874313388e-05}
{"train_loss": 0.06686773151159286, "global_step": 156773, "epoch": 1761, "lr": 3.8058595782223795e-05}
{"train_loss": 0.06904543936252594, "global_step": 156774, "epoch": 1761, "lr": 3.80580328229191e-05}
{"train_loss": 0.11329858005046844, "global_step": 156775, "epoch": 1761, "lr": 3.805746986521985e-05}
{"train_loss": 0.18551048636436462, "global_step": 156776, "epoch": 1761, "lr": 3.805690690912613e-05}
{"train_loss": 0.09690023958683014, "global_step": 156777, "epoch": 1761, "lr": 3.805634395463801e-05}
{"train_loss": 0.0621359683573246, "global_step": 156778, "epoch": 1761, "lr": 3.8055781001755564e-05}
{"train_loss": 0.09027452766895294, "global_step": 156779, "epoch": 1761, "lr": 3.805521805047888e-05}
{"train_loss": 0.11821692436933517, "global_step": 156780, "epoch": 1761, "lr": 3.805465510080804e-05}
{"train_loss": 0.10071847587823868, "global_step": 156781, "epoch": 1761, "lr": 3.8054092152743077e-05}
{"train_loss": 0.08207918703556061, "global_step": 156782, "epoch": 1761, "lr": 3.805352920628412e-05}
{"train_loss": 0.12296604365110397, "global_step": 156783, "epoch": 1761, "lr": 3.80529662614312e-05}
{"train_loss": 0.17499583959579468, "global_step": 156784, "epoch": 1761, "lr": 3.8052403318184433e-05}
{"train_loss": 0.06992091238498688, "global_step": 156785, "epoch": 1761, "lr": 3.805184037654385e-05}
{"train_loss": 0.1290709227323532, "global_step": 156786, "epoch": 1761, "lr": 3.8051277436509556e-05}
{"train_loss": 0.07242761552333832, "global_step": 156787, "epoch": 1761, "lr": 3.805071449808164e-05}
{"train_loss": 0.08935845643281937, "global_step": 156788, "epoch": 1761, "lr": 3.805015156126015e-05}
{"train_loss": 0.07222753763198853, "global_step": 156789, "epoch": 1761, "lr": 3.804958862604517e-05}
{"train_loss": 0.1115359216928482, "global_step": 156790, "epoch": 1761, "lr": 3.804902569243677e-05}
{"train_loss": 0.15287156403064728, "global_step": 156791, "epoch": 1761, "lr": 3.804846276043506e-05}
{"train_loss": 0.11006686091423035, "global_step": 156792, "epoch": 1761, "lr": 3.804789983004006e-05}
{"train_loss": 0.1028292179107666, "global_step": 156793, "epoch": 1761, "lr": 3.8047336901251895e-05}
{"train_loss": 0.08547137677669525, "global_step": 156794, "epoch": 1761, "lr": 3.804677397407061e-05}
{"train_loss": 0.153464674949646, "global_step": 156795, "epoch": 1761, "lr": 3.8046211048496294e-05}
{"train_loss": 0.10685313493013382, "global_step": 156796, "epoch": 1761, "lr": 3.8045648124529014e-05}
{"train_loss": 0.07823174446821213, "global_step": 156797, "epoch": 1761, "lr": 3.804508520216885e-05}
{"train_loss": 0.12089493870735168, "global_step": 156798, "epoch": 1761, "lr": 3.804452228141589e-05}
{"train_loss": 0.06445006281137466, "global_step": 156799, "epoch": 1761, "lr": 3.8043959362270196e-05}
{"train_loss": 0.07724284380674362, "global_step": 156800, "epoch": 1761, "lr": 3.8043396444731835e-05}
{"train_loss": 0.1196155995130539, "global_step": 156801, "epoch": 1761, "lr": 3.8042833528800905e-05}
{"train_loss": 0.08563677966594696, "global_step": 156802, "epoch": 1761, "lr": 3.8042270614477454e-05}
{"train_loss": 0.08868226408958435, "global_step": 156803, "epoch": 1761, "lr": 3.804170770176158e-05}
{"train_loss": 0.07184959948062897, "global_step": 156804, "epoch": 1761, "lr": 3.804114479065336e-05}
{"train_loss": 0.10919991880655289, "global_step": 156805, "epoch": 1761, "lr": 3.804058188115285e-05}
{"train_loss": 0.07135627418756485, "global_step": 156806, "epoch": 1761, "lr": 3.804001897326015e-05}
{"train_loss": 0.08650274574756622, "global_step": 156807, "epoch": 1761, "lr": 3.8039456066975316e-05}
{"train_loss": 0.06771086156368256, "global_step": 156808, "epoch": 1761, "lr": 3.803889316229845e-05}
{"train_loss": 0.08432425558567047, "global_step": 156809, "epoch": 1761, "lr": 3.803833025922958e-05}
{"train_loss": 0.032283980399370193, "global_step": 156810, "epoch": 1761, "lr": 3.803776735776884e-05}
{"train_loss": 0.09098681807518005, "global_step": 156811, "epoch": 1761, "lr": 3.803720445791625e-05}
{"train_loss": 0.17365296185016632, "global_step": 156812, "epoch": 1761, "lr": 3.803664155967194e-05}
{"train_loss": 0.09109213948249817, "global_step": 156813, "epoch": 1761, "lr": 3.803607866303594e-05}
{"train_loss": 0.12076336145401001, "global_step": 156814, "epoch": 1761, "lr": 3.803551576800834e-05}
{"train_loss": 0.09215161949396133, "global_step": 156815, "epoch": 1761, "lr": 3.8034952874589226e-05}
{"train_loss": 0.08116841316223145, "global_step": 156816, "epoch": 1761, "lr": 3.8034389982778664e-05}
{"train_loss": 0.0989119726118077, "global_step": 156817, "epoch": 1761, "lr": 3.803382709257673e-05, "val_loss": 6.3353753089904785}
{"train_loss": 0.09253071248531342, "global_step": 156818, "epoch": 1762, "lr": 3.803326420398352e-05}
{"train_loss": 0.16431540250778198, "global_step": 156819, "epoch": 1762, "lr": 3.803270131699907e-05}
{"train_loss": 0.10420399159193039, "global_step": 156820, "epoch": 1762, "lr": 3.8032138431623496e-05}
{"train_loss": 0.1316492110490799, "global_step": 156821, "epoch": 1762, "lr": 3.8031575547856826e-05}
{"train_loss": 0.08487854897975922, "global_step": 156822, "epoch": 1762, "lr": 3.803101266569918e-05}
{"train_loss": 0.09140868484973907, "global_step": 156823, "epoch": 1762, "lr": 3.8030449785150626e-05}
{"train_loss": 0.1467696726322174, "global_step": 156824, "epoch": 1762, "lr": 3.802988690621122e-05}
{"train_loss": 0.09353030472993851, "global_step": 156825, "epoch": 1762, "lr": 3.802932402888106e-05}
{"train_loss": 0.0446256548166275, "global_step": 156826, "epoch": 1762, "lr": 3.8028761153160204e-05}
{"train_loss": 0.1184912621974945, "global_step": 156827, "epoch": 1762, "lr": 3.8028198279048754e-05}
{"train_loss": 0.07796517759561539, "global_step": 156828, "epoch": 1762, "lr": 3.802763540654674e-05}
{"train_loss": 0.08002812415361404, "global_step": 156829, "epoch": 1762, "lr": 3.802707253565428e-05}
{"train_loss": 0.11393560469150543, "global_step": 156830, "epoch": 1762, "lr": 3.802650966637143e-05}
{"train_loss": 0.07479523867368698, "global_step": 156831, "epoch": 1762, "lr": 3.802594679869828e-05}
{"train_loss": 0.13791392743587494, "global_step": 156832, "epoch": 1762, "lr": 3.802538393263489e-05}
{"train_loss": 0.07595884799957275, "global_step": 156833, "epoch": 1762, "lr": 3.8024821068181334e-05}
{"train_loss": 0.10421819239854813, "global_step": 156834, "epoch": 1762, "lr": 3.80242582053377e-05}
{"train_loss": 0.05770838260650635, "global_step": 156835, "epoch": 1762, "lr": 3.802369534410405e-05}
{"train_loss": 0.07158150523900986, "global_step": 156836, "epoch": 1762, "lr": 3.8023132484480484e-05}
{"train_loss": 0.07793630659580231, "global_step": 156837, "epoch": 1762, "lr": 3.8022569626467065e-05}
{"train_loss": 0.05888976529240608, "global_step": 156838, "epoch": 1762, "lr": 3.802200677006384e-05}
{"train_loss": 0.08916923403739929, "global_step": 156839, "epoch": 1762, "lr": 3.802144391527092e-05}
{"train_loss": 0.07393781840801239, "global_step": 156840, "epoch": 1762, "lr": 3.802088106208838e-05}
{"train_loss": 0.11746402084827423, "global_step": 156841, "epoch": 1762, "lr": 3.802031821051627e-05}
{"train_loss": 0.0842045322060585, "global_step": 156842, "epoch": 1762, "lr": 3.801975536055471e-05}
{"train_loss": 0.14860446751117706, "global_step": 156843, "epoch": 1762, "lr": 3.801919251220372e-05}
{"train_loss": 0.12630979716777802, "global_step": 156844, "epoch": 1762, "lr": 3.801862966546342e-05}
{"train_loss": 0.12193842232227325, "global_step": 156845, "epoch": 1762, "lr": 3.8018066820333854e-05}
{"train_loss": 0.15740926563739777, "global_step": 156846, "epoch": 1762, "lr": 3.8017503976815134e-05}
{"train_loss": 0.0833180695772171, "global_step": 156847, "epoch": 1762, "lr": 3.8016941134907294e-05}
{"train_loss": 0.13573981821537018, "global_step": 156848, "epoch": 1762, "lr": 3.801637829461045e-05}
{"train_loss": 0.09555425494909286, "global_step": 156849, "epoch": 1762, "lr": 3.8015815455924636e-05}
{"train_loss": 0.11253306269645691, "global_step": 156850, "epoch": 1762, "lr": 3.801525261884997e-05}
{"train_loss": 0.0864696204662323, "global_step": 156851, "epoch": 1762, "lr": 3.8014689783386496e-05}
{"train_loss": 0.10375314205884933, "global_step": 156852, "epoch": 1762, "lr": 3.8014126949534305e-05}
{"train_loss": 0.12041587382555008, "global_step": 156853, "epoch": 1762, "lr": 3.801356411729347e-05}
{"train_loss": 0.05729624629020691, "global_step": 156854, "epoch": 1762, "lr": 3.8013001286664064e-05}
{"train_loss": 0.13417799770832062, "global_step": 156855, "epoch": 1762, "lr": 3.801243845764616e-05}
{"train_loss": 0.06681814789772034, "global_step": 156856, "epoch": 1762, "lr": 3.801187563023984e-05}
{"train_loss": 0.08344850689172745, "global_step": 156857, "epoch": 1762, "lr": 3.8011312804445173e-05}
{"train_loss": 0.12589657306671143, "global_step": 156858, "epoch": 1762, "lr": 3.801074998026224e-05}
{"train_loss": 0.13542643189430237, "global_step": 156859, "epoch": 1762, "lr": 3.801018715769112e-05}
{"train_loss": 0.056281764060258865, "global_step": 156860, "epoch": 1762, "lr": 3.800962433673187e-05}
{"train_loss": 0.15778885781764984, "global_step": 156861, "epoch": 1762, "lr": 3.8009061517384606e-05}
{"train_loss": 0.13049159944057465, "global_step": 156862, "epoch": 1762, "lr": 3.800849869964935e-05}
{"train_loss": 0.06349024921655655, "global_step": 156863, "epoch": 1762, "lr": 3.800793588352624e-05}
{"train_loss": 0.1427130103111267, "global_step": 156864, "epoch": 1762, "lr": 3.800737306901528e-05}
{"train_loss": 0.12949219346046448, "global_step": 156865, "epoch": 1762, "lr": 3.800681025611661e-05}
{"train_loss": 0.10330130159854889, "global_step": 156866, "epoch": 1762, "lr": 3.800624744483026e-05}
{"train_loss": 0.13281022012233734, "global_step": 156867, "epoch": 1762, "lr": 3.800568463515633e-05}
{"train_loss": 0.04376697912812233, "global_step": 156868, "epoch": 1762, "lr": 3.8005121827094887e-05}
{"train_loss": 0.07199284434318542, "global_step": 156869, "epoch": 1762, "lr": 3.800455902064602e-05}
{"train_loss": 0.07675737142562866, "global_step": 156870, "epoch": 1762, "lr": 3.800399621580978e-05}
{"train_loss": 0.13131000101566315, "global_step": 156871, "epoch": 1762, "lr": 3.800343341258627e-05}
{"train_loss": 0.09674391150474548, "global_step": 156872, "epoch": 1762, "lr": 3.8002870610975544e-05}
{"train_loss": 0.09189090877771378, "global_step": 156873, "epoch": 1762, "lr": 3.800230781097769e-05}
{"train_loss": 0.10066906362771988, "global_step": 156874, "epoch": 1762, "lr": 3.800174501259277e-05}
{"train_loss": 0.12261650711297989, "global_step": 156875, "epoch": 1762, "lr": 3.800118221582086e-05}
{"train_loss": 0.04628197103738785, "global_step": 156876, "epoch": 1762, "lr": 3.8000619420662076e-05}
{"train_loss": 0.13700196146965027, "global_step": 156877, "epoch": 1762, "lr": 3.800005662711644e-05}
{"train_loss": 0.13551363348960876, "global_step": 156878, "epoch": 1762, "lr": 3.799949383518406e-05}
{"train_loss": 0.08487515151500702, "global_step": 156879, "epoch": 1762, "lr": 3.799893104486499e-05}
{"train_loss": 0.10069343447685242, "global_step": 156880, "epoch": 1762, "lr": 3.7998368256159336e-05}
{"train_loss": 0.12754122912883759, "global_step": 156881, "epoch": 1762, "lr": 3.799780546906713e-05}
{"train_loss": 0.08507772535085678, "global_step": 156882, "epoch": 1762, "lr": 3.7997242683588505e-05}
{"train_loss": 0.1586643010377884, "global_step": 156883, "epoch": 1762, "lr": 3.7996679899723476e-05}
{"train_loss": 0.11665285378694534, "global_step": 156884, "epoch": 1762, "lr": 3.799611711747217e-05}
{"train_loss": 0.09789182990789413, "global_step": 156885, "epoch": 1762, "lr": 3.799555433683462e-05}
{"train_loss": 0.08820010721683502, "global_step": 156886, "epoch": 1762, "lr": 3.799499155781093e-05}
{"train_loss": 0.09014583379030228, "global_step": 156887, "epoch": 1762, "lr": 3.799442878040116e-05}
{"train_loss": 0.09966644644737244, "global_step": 156888, "epoch": 1762, "lr": 3.799386600460541e-05}
{"train_loss": 0.07764405012130737, "global_step": 156889, "epoch": 1762, "lr": 3.7993303230423715e-05}
{"train_loss": 0.104974165558815, "global_step": 156890, "epoch": 1762, "lr": 3.7992740457856195e-05}
{"train_loss": 0.10660076886415482, "global_step": 156891, "epoch": 1762, "lr": 3.7992177686902896e-05}
{"train_loss": 0.15199311077594757, "global_step": 156892, "epoch": 1762, "lr": 3.7991614917563906e-05}
{"train_loss": 0.12726357579231262, "global_step": 156893, "epoch": 1762, "lr": 3.799105214983929e-05}
{"train_loss": 0.06101801618933678, "global_step": 156894, "epoch": 1762, "lr": 3.7990489383729125e-05}
{"train_loss": 0.11091283708810806, "global_step": 156895, "epoch": 1762, "lr": 3.798992661923351e-05}
{"train_loss": 0.09050317108631134, "global_step": 156896, "epoch": 1762, "lr": 3.798936385635249e-05}
{"train_loss": 0.07794991135597229, "global_step": 156897, "epoch": 1762, "lr": 3.798880109508617e-05}
{"train_loss": 0.13376373052597046, "global_step": 156898, "epoch": 1762, "lr": 3.798823833543458e-05}
{"train_loss": 0.1096283569931984, "global_step": 156899, "epoch": 1762, "lr": 3.798767557739785e-05}
{"train_loss": 0.13171595335006714, "global_step": 156900, "epoch": 1762, "lr": 3.798711282097602e-05}
{"train_loss": 0.08745791763067245, "global_step": 156901, "epoch": 1762, "lr": 3.798655006616919e-05}
{"train_loss": 0.07687266170978546, "global_step": 156902, "epoch": 1762, "lr": 3.79859873129774e-05}
{"train_loss": 0.06093783676624298, "global_step": 156903, "epoch": 1762, "lr": 3.798542456140077e-05}
{"train_loss": 0.12633123993873596, "global_step": 156904, "epoch": 1762, "lr": 3.7984861811439325e-05}
{"train_loss": 0.12668001651763916, "global_step": 156905, "epoch": 1762, "lr": 3.79842990630932e-05}
{"train_loss": 0.10275828670919611, "global_step": 156906, "epoch": 1762, "lr": 3.798373631636242e-05, "val_loss": 6.533576965332031}
{"train_loss": 0.09404394030570984, "global_step": 156907, "epoch": 1763, "lr": 3.798317357124709e-05}
{"train_loss": 0.06403421610593796, "global_step": 156908, "epoch": 1763, "lr": 3.798261082774726e-05}
{"train_loss": 0.14041809737682343, "global_step": 156909, "epoch": 1763, "lr": 3.798204808586304e-05}
{"train_loss": 0.06598706543445587, "global_step": 156910, "epoch": 1763, "lr": 3.7981485345594485e-05}
{"train_loss": 0.11555055528879166, "global_step": 156911, "epoch": 1763, "lr": 3.798092260694166e-05}
{"train_loss": 0.14165426790714264, "global_step": 156912, "epoch": 1763, "lr": 3.798035986990467e-05}
{"train_loss": 0.0430670827627182, "global_step": 156913, "epoch": 1763, "lr": 3.7979797134483553e-05}
{"train_loss": 0.1363879293203354, "global_step": 156914, "epoch": 1763, "lr": 3.797923440067843e-05}
{"train_loss": 0.08811791241168976, "global_step": 156915, "epoch": 1763, "lr": 3.797867166848933e-05}
{"train_loss": 0.06754206866025925, "global_step": 156916, "epoch": 1763, "lr": 3.7978108937916374e-05}
{"train_loss": 0.13694894313812256, "global_step": 156917, "epoch": 1763, "lr": 3.797754620895959e-05}
{"train_loss": 0.10885089635848999, "global_step": 156918, "epoch": 1763, "lr": 3.79769834816191e-05}
{"train_loss": 0.07856878638267517, "global_step": 156919, "epoch": 1763, "lr": 3.797642075589494e-05}
{"train_loss": 0.0826965719461441, "global_step": 156920, "epoch": 1763, "lr": 3.7975858031787224e-05}
{"train_loss": 0.09797146916389465, "global_step": 156921, "epoch": 1763, "lr": 3.797529530929598e-05}
{"train_loss": 0.0968458503484726, "global_step": 156922, "epoch": 1763, "lr": 3.797473258842134e-05}
{"train_loss": 0.13800391554832458, "global_step": 156923, "epoch": 1763, "lr": 3.797416986916333e-05}
{"train_loss": 0.09303595870733261, "global_step": 156924, "epoch": 1763, "lr": 3.797360715152206e-05}
{"train_loss": 0.14057239890098572, "global_step": 156925, "epoch": 1763, "lr": 3.7973044435497576e-05}
{"train_loss": 0.13676252961158752, "global_step": 156926, "epoch": 1763, "lr": 3.7972481721089994e-05}
{"train_loss": 0.094575896859169, "global_step": 156927, "epoch": 1763, "lr": 3.797191900829934e-05}
{"train_loss": 0.113803431391716, "global_step": 156928, "epoch": 1763, "lr": 3.7971356297125734e-05}
{"train_loss": 0.10828016698360443, "global_step": 156929, "epoch": 1763, "lr": 3.797079358756922e-05}
{"train_loss": 0.09950516372919083, "global_step": 156930, "epoch": 1763, "lr": 3.797023087962988e-05}
{"train_loss": 0.11140795797109604, "global_step": 156931, "epoch": 1763, "lr": 3.796966817330782e-05}
{"train_loss": 0.08807075768709183, "global_step": 156932, "epoch": 1763, "lr": 3.7969105468603064e-05}
{"train_loss": 0.07828530669212341, "global_step": 156933, "epoch": 1763, "lr": 3.796854276551574e-05}
{"train_loss": 0.1105726957321167, "global_step": 156934, "epoch": 1763, "lr": 3.7967980064045874e-05}
{"train_loss": 0.12643089890480042, "global_step": 156935, "epoch": 1763, "lr": 3.796741736419359e-05}
{"train_loss": 0.15284192562103271, "global_step": 156936, "epoch": 1763, "lr": 3.7966854665958924e-05}
{"train_loss": 0.07217040657997131, "global_step": 156937, "epoch": 1763, "lr": 3.796629196934198e-05}
{"train_loss": 0.1395719051361084, "global_step": 156938, "epoch": 1763, "lr": 3.796572927434281e-05}
{"train_loss": 0.1386973261833191, "global_step": 156939, "epoch": 1763, "lr": 3.7965166580961515e-05}
{"train_loss": 0.06121445447206497, "global_step": 156940, "epoch": 1763, "lr": 3.796460388919813e-05}
{"train_loss": 0.1529940664768219, "global_step": 156941, "epoch": 1763, "lr": 3.796404119905279e-05}
{"train_loss": 0.11147840321063995, "global_step": 156942, "epoch": 1763, "lr": 3.796347851052551e-05}
{"train_loss": 0.0879519134759903, "global_step": 156943, "epoch": 1763, "lr": 3.7962915823616426e-05}
{"train_loss": 0.04307219013571739, "global_step": 156944, "epoch": 1763, "lr": 3.796235313832555e-05}
{"train_loss": 0.142636239528656, "global_step": 156945, "epoch": 1763, "lr": 3.7961790454653004e-05}
{"train_loss": 0.0664147287607193, "global_step": 156946, "epoch": 1763, "lr": 3.796122777259884e-05}
{"train_loss": 0.07919351011514664, "global_step": 156947, "epoch": 1763, "lr": 3.7960665092163156e-05}
{"train_loss": 0.0964643582701683, "global_step": 156948, "epoch": 1763, "lr": 3.7960102413346e-05}
{"train_loss": 0.08046330511569977, "global_step": 156949, "epoch": 1763, "lr": 3.7959539736147466e-05}
{"train_loss": 0.07890816777944565, "global_step": 156950, "epoch": 1763, "lr": 3.795897706056764e-05}
{"train_loss": 0.08672866970300674, "global_step": 156951, "epoch": 1763, "lr": 3.795841438660655e-05}
{"train_loss": 0.10115063935518265, "global_step": 156952, "epoch": 1763, "lr": 3.795785171426434e-05}
{"train_loss": 0.046702686697244644, "global_step": 156953, "epoch": 1763, "lr": 3.795728904354102e-05}
{"train_loss": 0.117914117872715, "global_step": 156954, "epoch": 1763, "lr": 3.795672637443672e-05}
{"train_loss": 0.13722847402095795, "global_step": 156955, "epoch": 1763, "lr": 3.7956163706951465e-05}
{"train_loss": 0.12236616760492325, "global_step": 156956, "epoch": 1763, "lr": 3.795560104108539e-05}
{"train_loss": 0.09831944853067398, "global_step": 156957, "epoch": 1763, "lr": 3.79550383768385e-05}
{"train_loss": 0.09495271742343903, "global_step": 156958, "epoch": 1763, "lr": 3.795447571421094e-05}
{"train_loss": 0.08994822204113007, "global_step": 156959, "epoch": 1763, "lr": 3.7953913053202734e-05}
{"train_loss": 0.05281992629170418, "global_step": 156960, "epoch": 1763, "lr": 3.7953350393813994e-05}
{"train_loss": 0.06780967116355896, "global_step": 156961, "epoch": 1763, "lr": 3.795278773604476e-05}
{"train_loss": 0.08551912009716034, "global_step": 156962, "epoch": 1763, "lr": 3.795222507989516e-05}
{"train_loss": 0.14378637075424194, "global_step": 156963, "epoch": 1763, "lr": 3.7951662425365195e-05}
{"train_loss": 0.05548207461833954, "global_step": 156964, "epoch": 1763, "lr": 3.795109977245501e-05}
{"train_loss": 0.08108577132225037, "global_step": 156965, "epoch": 1763, "lr": 3.795053712116464e-05}
{"train_loss": 0.08712827414274216, "global_step": 156966, "epoch": 1763, "lr": 3.794997447149417e-05}
{"train_loss": 0.08679851144552231, "global_step": 156967, "epoch": 1763, "lr": 3.7949411823443686e-05}
{"train_loss": 0.14922687411308289, "global_step": 156968, "epoch": 1763, "lr": 3.7948849177013256e-05}
{"train_loss": 0.08873546868562698, "global_step": 156969, "epoch": 1763, "lr": 3.794828653220297e-05}
{"train_loss": 0.09537111222743988, "global_step": 156970, "epoch": 1763, "lr": 3.794772388901287e-05}
{"train_loss": 0.12320080399513245, "global_step": 156971, "epoch": 1763, "lr": 3.794716124744307e-05}
{"train_loss": 0.12944632768630981, "global_step": 156972, "epoch": 1763, "lr": 3.79465986074936e-05}
{"train_loss": 0.08712023496627808, "global_step": 156973, "epoch": 1763, "lr": 3.7946035969164596e-05}
{"train_loss": 0.13347205519676208, "global_step": 156974, "epoch": 1763, "lr": 3.794547333245607e-05}
{"train_loss": 0.1002945825457573, "global_step": 156975, "epoch": 1763, "lr": 3.794491069736815e-05}
{"train_loss": 0.11310605704784393, "global_step": 156976, "epoch": 1763, "lr": 3.794434806390088e-05}
{"train_loss": 0.13867050409317017, "global_step": 156977, "epoch": 1763, "lr": 3.7943785432054355e-05}
{"train_loss": 0.11607394367456436, "global_step": 156978, "epoch": 1763, "lr": 3.794322280182862e-05}
{"train_loss": 0.1012490838766098, "global_step": 156979, "epoch": 1763, "lr": 3.79426601732238e-05}
{"train_loss": 0.12919650971889496, "global_step": 156980, "epoch": 1763, "lr": 3.794209754623992e-05}
{"train_loss": 0.0993640273809433, "global_step": 156981, "epoch": 1763, "lr": 3.794153492087709e-05}
{"train_loss": 0.09474273771047592, "global_step": 156982, "epoch": 1763, "lr": 3.7940972297135355e-05}
{"train_loss": 0.10568618774414062, "global_step": 156983, "epoch": 1763, "lr": 3.7940409675014845e-05}
{"train_loss": 0.08974363654851913, "global_step": 156984, "epoch": 1763, "lr": 3.793984705451556e-05}
{"train_loss": 0.05161512643098831, "global_step": 156985, "epoch": 1763, "lr": 3.793928443563763e-05}
{"train_loss": 0.08658257126808167, "global_step": 156986, "epoch": 1763, "lr": 3.793872181838113e-05}
{"train_loss": 0.09439930319786072, "global_step": 156987, "epoch": 1763, "lr": 3.793815920274611e-05}
{"train_loss": 0.09575651586055756, "global_step": 156988, "epoch": 1763, "lr": 3.793759658873267e-05}
{"train_loss": 0.1547258049249649, "global_step": 156989, "epoch": 1763, "lr": 3.793703397634085e-05}
{"train_loss": 0.09769541770219803, "global_step": 156990, "epoch": 1763, "lr": 3.793647136557077e-05}
{"train_loss": 0.10526344925165176, "global_step": 156991, "epoch": 1763, "lr": 3.7935908756422464e-05}
{"train_loss": 0.062122832983732224, "global_step": 156992, "epoch": 1763, "lr": 3.7935346148896054e-05}
{"train_loss": 0.11160296201705933, "global_step": 156993, "epoch": 1763, "lr": 3.793478354299157e-05}
{"train_loss": 0.09660754352807999, "global_step": 156994, "epoch": 1763, "lr": 3.793422093870912e-05}
{"train_loss": 0.10081040708536512, "global_step": 156995, "epoch": 1763, "lr": 3.793365833604875e-05, "val_loss": 6.527840614318848}
{"train_loss": 0.1574382781982422, "global_step": 156996, "epoch": 1764, "lr": 3.7933095735010574e-05}
{"train_loss": 0.05921756848692894, "global_step": 156997, "epoch": 1764, "lr": 3.793253313559462e-05}
{"train_loss": 0.09939588606357574, "global_step": 156998, "epoch": 1764, "lr": 3.793197053780102e-05}
{"train_loss": 0.09149599075317383, "global_step": 156999, "epoch": 1764, "lr": 3.793140794162979e-05}
{"train_loss": 0.08674395829439163, "global_step": 157000, "epoch": 1764, "lr": 3.793084534708106e-05}
{"train_loss": 0.08772751688957214, "global_step": 157001, "epoch": 1764, "lr": 3.793028275415486e-05}
{"train_loss": 0.15559718012809753, "global_step": 157002, "epoch": 1764, "lr": 3.79297201628513e-05}
{"train_loss": 0.03912273049354553, "global_step": 157003, "epoch": 1764, "lr": 3.7929157573170436e-05}
{"train_loss": 0.1417456716299057, "global_step": 157004, "epoch": 1764, "lr": 3.7928594985112353e-05}
{"train_loss": 0.10580727458000183, "global_step": 157005, "epoch": 1764, "lr": 3.792803239867712e-05}
{"train_loss": 0.11005715280771255, "global_step": 157006, "epoch": 1764, "lr": 3.792746981386482e-05}
{"train_loss": 0.05957163870334625, "global_step": 157007, "epoch": 1764, "lr": 3.7926907230675535e-05}
{"train_loss": 0.08427488803863525, "global_step": 157008, "epoch": 1764, "lr": 3.79263446491093e-05}
{"train_loss": 0.054373402148485184, "global_step": 157009, "epoch": 1764, "lr": 3.792578206916626e-05}
{"train_loss": 0.11150120943784714, "global_step": 157010, "epoch": 1764, "lr": 3.792521949084641e-05}
{"train_loss": 0.13046881556510925, "global_step": 157011, "epoch": 1764, "lr": 3.7924656914149906e-05}
{"train_loss": 0.07489380240440369, "global_step": 157012, "epoch": 1764, "lr": 3.792409433907675e-05}
{"train_loss": 0.11026167869567871, "global_step": 157013, "epoch": 1764, "lr": 3.792353176562708e-05}
{"train_loss": 0.08723534643650055, "global_step": 157014, "epoch": 1764, "lr": 3.792296919380092e-05}
{"train_loss": 0.09895557165145874, "global_step": 157015, "epoch": 1764, "lr": 3.792240662359839e-05}
{"train_loss": 0.057762425392866135, "global_step": 157016, "epoch": 1764, "lr": 3.792184405501953e-05}
{"train_loss": 0.10846853256225586, "global_step": 157017, "epoch": 1764, "lr": 3.792128148806445e-05}
{"train_loss": 0.12718795239925385, "global_step": 157018, "epoch": 1764, "lr": 3.7920718922733176e-05}
{"train_loss": 0.13462653756141663, "global_step": 157019, "epoch": 1764, "lr": 3.792015635902585e-05}
{"train_loss": 0.0905781164765358, "global_step": 157020, "epoch": 1764, "lr": 3.791959379694248e-05}
{"train_loss": 0.07963680475950241, "global_step": 157021, "epoch": 1764, "lr": 3.791903123648319e-05}
{"train_loss": 0.05888880416750908, "global_step": 157022, "epoch": 1764, "lr": 3.7918468677648034e-05}
{"train_loss": 0.09965258091688156, "global_step": 157023, "epoch": 1764, "lr": 3.7917906120437095e-05}
{"train_loss": 0.11465838551521301, "global_step": 157024, "epoch": 1764, "lr": 3.791734356485045e-05}
{"train_loss": 0.09515657275915146, "global_step": 157025, "epoch": 1764, "lr": 3.7916781010888165e-05}
{"train_loss": 0.06288816779851913, "global_step": 157026, "epoch": 1764, "lr": 3.7916218458550336e-05}
{"train_loss": 0.131907120347023, "global_step": 157027, "epoch": 1764, "lr": 3.7915655907837e-05}
{"train_loss": 0.08113770931959152, "global_step": 157028, "epoch": 1764, "lr": 3.7915093358748275e-05}
{"train_loss": 0.08758869767189026, "global_step": 157029, "epoch": 1764, "lr": 3.7914530811284206e-05}
{"train_loss": 0.08128926903009415, "global_step": 157030, "epoch": 1764, "lr": 3.7913968265444895e-05}
{"train_loss": 0.16367128491401672, "global_step": 157031, "epoch": 1764, "lr": 3.791340572123039e-05}
{"train_loss": 0.06468009203672409, "global_step": 157032, "epoch": 1764, "lr": 3.791284317864079e-05}
{"train_loss": 0.0812232494354248, "global_step": 157033, "epoch": 1764, "lr": 3.791228063767615e-05}
{"train_loss": 0.10474058240652084, "global_step": 157034, "epoch": 1764, "lr": 3.7911718098336575e-05}
{"train_loss": 0.09717307984828949, "global_step": 157035, "epoch": 1764, "lr": 3.79111555606221e-05}
{"train_loss": 0.1522226482629776, "global_step": 157036, "epoch": 1764, "lr": 3.791059302453284e-05}
{"train_loss": 0.10556349158287048, "global_step": 157037, "epoch": 1764, "lr": 3.791003049006884e-05}
{"train_loss": 0.06423402577638626, "global_step": 157038, "epoch": 1764, "lr": 3.79094679572302e-05}
{"train_loss": 0.0947701558470726, "global_step": 157039, "epoch": 1764, "lr": 3.7908905426016975e-05}
{"train_loss": 0.05717906728386879, "global_step": 157040, "epoch": 1764, "lr": 3.790834289642926e-05}
{"train_loss": 0.1045432984828949, "global_step": 157041, "epoch": 1764, "lr": 3.790778036846711e-05}
{"train_loss": 0.0977020189166069, "global_step": 157042, "epoch": 1764, "lr": 3.790721784213062e-05}
{"train_loss": 0.11626214534044266, "global_step": 157043, "epoch": 1764, "lr": 3.790665531741986e-05}
{"train_loss": 0.07998479157686234, "global_step": 157044, "epoch": 1764, "lr": 3.790609279433489e-05}
{"train_loss": 0.06516751646995544, "global_step": 157045, "epoch": 1764, "lr": 3.7905530272875825e-05}
{"train_loss": 0.06583306938409805, "global_step": 157046, "epoch": 1764, "lr": 3.790496775304268e-05}
{"train_loss": 0.10743096470832825, "global_step": 157047, "epoch": 1764, "lr": 3.790440523483559e-05}
{"train_loss": 0.12917393445968628, "global_step": 157048, "epoch": 1764, "lr": 3.790384271825459e-05}
{"train_loss": 0.07745558023452759, "global_step": 157049, "epoch": 1764, "lr": 3.7903280203299785e-05}
{"train_loss": 0.056683026254177094, "global_step": 157050, "epoch": 1764, "lr": 3.790271768997122e-05}
{"train_loss": 0.045585185289382935, "global_step": 157051, "epoch": 1764, "lr": 3.790215517826901e-05}
{"train_loss": 0.13248853385448456, "global_step": 157052, "epoch": 1764, "lr": 3.790159266819318e-05}
{"train_loss": 0.046413764357566833, "global_step": 157053, "epoch": 1764, "lr": 3.790103015974386e-05}
{"train_loss": 0.08091246336698532, "global_step": 157054, "epoch": 1764, "lr": 3.790046765292107e-05}
{"train_loss": 0.15311624109745026, "global_step": 157055, "epoch": 1764, "lr": 3.789990514772494e-05}
{"train_loss": 0.14815360307693481, "global_step": 157056, "epoch": 1764, "lr": 3.78993426441555e-05}
{"train_loss": 0.060872796922922134, "global_step": 157057, "epoch": 1764, "lr": 3.789878014221284e-05}
{"train_loss": 0.15209393203258514, "global_step": 157058, "epoch": 1764, "lr": 3.7898217641897074e-05}
{"train_loss": 0.07438936084508896, "global_step": 157059, "epoch": 1764, "lr": 3.7897655143208225e-05}
{"train_loss": 0.11553042382001877, "global_step": 157060, "epoch": 1764, "lr": 3.78970926461464e-05}
{"train_loss": 0.10757145285606384, "global_step": 157061, "epoch": 1764, "lr": 3.789653015071165e-05}
{"train_loss": 0.0889761671423912, "global_step": 157062, "epoch": 1764, "lr": 3.789596765690408e-05}
{"train_loss": 0.12810960412025452, "global_step": 157063, "epoch": 1764, "lr": 3.789540516472373e-05}
{"train_loss": 0.06781005859375, "global_step": 157064, "epoch": 1764, "lr": 3.789484267417072e-05}
{"train_loss": 0.13627633452415466, "global_step": 157065, "epoch": 1764, "lr": 3.7894280185245075e-05}
{"train_loss": 0.09325313568115234, "global_step": 157066, "epoch": 1764, "lr": 3.7893717697946915e-05}
{"train_loss": 0.04428275302052498, "global_step": 157067, "epoch": 1764, "lr": 3.7893155212276284e-05}
{"train_loss": 0.11478915810585022, "global_step": 157068, "epoch": 1764, "lr": 3.7892592728233285e-05}
{"train_loss": 0.11355991661548615, "global_step": 157069, "epoch": 1764, "lr": 3.789203024581796e-05}
{"train_loss": 0.10085465013980865, "global_step": 157070, "epoch": 1764, "lr": 3.7891467765030425e-05}
{"train_loss": 0.12983658909797668, "global_step": 157071, "epoch": 1764, "lr": 3.789090528587071e-05}
{"train_loss": 0.07296694815158844, "global_step": 157072, "epoch": 1764, "lr": 3.7890342808338944e-05}
{"train_loss": 0.09767743945121765, "global_step": 157073, "epoch": 1764, "lr": 3.7889780332435146e-05}
{"train_loss": 0.12028802186250687, "global_step": 157074, "epoch": 1764, "lr": 3.788921785815944e-05}
{"train_loss": 0.10167668759822845, "global_step": 157075, "epoch": 1764, "lr": 3.7888655385511866e-05}
{"train_loss": 0.1209915429353714, "global_step": 157076, "epoch": 1764, "lr": 3.7888092914492513e-05}
{"train_loss": 0.14907735586166382, "global_step": 157077, "epoch": 1764, "lr": 3.788753044510148e-05}
{"train_loss": 0.08644337952136993, "global_step": 157078, "epoch": 1764, "lr": 3.78869679773388e-05}
{"train_loss": 0.08966463059186935, "global_step": 157079, "epoch": 1764, "lr": 3.788640551120458e-05}
{"train_loss": 0.169786274433136, "global_step": 157080, "epoch": 1764, "lr": 3.7885843046698885e-05}
{"train_loss": 0.09870348870754242, "global_step": 157081, "epoch": 1764, "lr": 3.788528058382179e-05}
{"train_loss": 0.11484228074550629, "global_step": 157082, "epoch": 1764, "lr": 3.788471812257337e-05}
{"train_loss": 0.14982332289218903, "global_step": 157083, "epoch": 1764, "lr": 3.78841556629537e-05}
{"train_loss": 0.09942571797899985, "global_step": 157084, "epoch": 1764, "lr": 3.788359320496286e-05, "val_loss": 6.627671718597412}
{"train_loss": 0.1579902470111847, "global_step": 157085, "epoch": 1765, "lr": 3.7883030748600934e-05}
{"train_loss": 0.07464645057916641, "global_step": 157086, "epoch": 1765, "lr": 3.7882468293867965e-05}
{"train_loss": 0.11153998225927353, "global_step": 157087, "epoch": 1765, "lr": 3.788190584076408e-05}
{"train_loss": 0.09358920902013779, "global_step": 157088, "epoch": 1765, "lr": 3.788134338928929e-05}
{"train_loss": 0.11665420979261398, "global_step": 157089, "epoch": 1765, "lr": 3.788078093944373e-05}
{"train_loss": 0.09053181856870651, "global_step": 157090, "epoch": 1765, "lr": 3.788021849122744e-05}
{"train_loss": 0.09568820893764496, "global_step": 157091, "epoch": 1765, "lr": 3.7879656044640516e-05}
{"train_loss": 0.07152320444583893, "global_step": 157092, "epoch": 1765, "lr": 3.7879093599683005e-05}
{"train_loss": 0.1442490816116333, "global_step": 157093, "epoch": 1765, "lr": 3.787853115635501e-05}
{"train_loss": 0.08652937412261963, "global_step": 157094, "epoch": 1765, "lr": 3.787796871465662e-05}
{"train_loss": 0.14128567278385162, "global_step": 157095, "epoch": 1765, "lr": 3.7877406274587855e-05}
{"train_loss": 0.13763867318630219, "global_step": 157096, "epoch": 1765, "lr": 3.787684383614886e-05}
{"train_loss": 0.08439195901155472, "global_step": 157097, "epoch": 1765, "lr": 3.787628139933965e-05}
{"train_loss": 0.07247292995452881, "global_step": 157098, "epoch": 1765, "lr": 3.787571896416035e-05}
{"train_loss": 0.10660069435834885, "global_step": 157099, "epoch": 1765, "lr": 3.787515653061099e-05}
{"train_loss": 0.0918349176645279, "global_step": 157100, "epoch": 1765, "lr": 3.787459409869167e-05}
{"train_loss": 0.1265135556459427, "global_step": 157101, "epoch": 1765, "lr": 3.787403166840247e-05}
{"train_loss": 0.10982908308506012, "global_step": 157102, "epoch": 1765, "lr": 3.787346923974346e-05}
{"train_loss": 0.05118691176176071, "global_step": 157103, "epoch": 1765, "lr": 3.787290681271471e-05}
{"train_loss": 0.09554312378168106, "global_step": 157104, "epoch": 1765, "lr": 3.787234438731631e-05}
{"train_loss": 0.10129406303167343, "global_step": 157105, "epoch": 1765, "lr": 3.7871781963548306e-05}
{"train_loss": 0.06433972716331482, "global_step": 157106, "epoch": 1765, "lr": 3.787121954141081e-05}
{"train_loss": 0.06772296130657196, "global_step": 157107, "epoch": 1765, "lr": 3.787065712090387e-05}
{"train_loss": 0.152324378490448, "global_step": 157108, "epoch": 1765, "lr": 3.787009470202759e-05}
{"train_loss": 0.06300164014101028, "global_step": 157109, "epoch": 1765, "lr": 3.7869532284782e-05}
{"train_loss": 0.11675385385751724, "global_step": 157110, "epoch": 1765, "lr": 3.786896986916723e-05}
{"train_loss": 0.10162708908319473, "global_step": 157111, "epoch": 1765, "lr": 3.78684074551833e-05}
{"train_loss": 0.08784593641757965, "global_step": 157112, "epoch": 1765, "lr": 3.786784504283032e-05}
{"train_loss": 0.09789658337831497, "global_step": 157113, "epoch": 1765, "lr": 3.7867282632108386e-05}
{"train_loss": 0.07682286947965622, "global_step": 157114, "epoch": 1765, "lr": 3.786672022301753e-05}
{"train_loss": 0.11416558921337128, "global_step": 157115, "epoch": 1765, "lr": 3.7866157815557856e-05}
{"train_loss": 0.06975609809160233, "global_step": 157116, "epoch": 1765, "lr": 3.786559540972941e-05}
{"train_loss": 0.11453243345022202, "global_step": 157117, "epoch": 1765, "lr": 3.786503300553231e-05}
{"train_loss": 0.09843406826257706, "global_step": 157118, "epoch": 1765, "lr": 3.78644706029666e-05}
{"train_loss": 0.08589994162321091, "global_step": 157119, "epoch": 1765, "lr": 3.7863908202032364e-05}
{"train_loss": 0.0625942200422287, "global_step": 157120, "epoch": 1765, "lr": 3.786334580272967e-05}
{"train_loss": 0.11993308365345001, "global_step": 157121, "epoch": 1765, "lr": 3.786278340505861e-05}
{"train_loss": 0.11289776116609573, "global_step": 157122, "epoch": 1765, "lr": 3.786222100901925e-05}
{"train_loss": 0.09174343198537827, "global_step": 157123, "epoch": 1765, "lr": 3.786165861461167e-05}
{"train_loss": 0.0858563780784607, "global_step": 157124, "epoch": 1765, "lr": 3.786109622183593e-05}
{"train_loss": 0.07271285355091095, "global_step": 157125, "epoch": 1765, "lr": 3.7860533830692135e-05}
{"train_loss": 0.10529275238513947, "global_step": 157126, "epoch": 1765, "lr": 3.7859971441180316e-05}
{"train_loss": 0.1397695541381836, "global_step": 157127, "epoch": 1765, "lr": 3.78594090533006e-05}
{"train_loss": 0.10730337351560593, "global_step": 157128, "epoch": 1765, "lr": 3.7858846667053025e-05}
{"train_loss": 0.07539809495210648, "global_step": 157129, "epoch": 1765, "lr": 3.7858284282437674e-05}
{"train_loss": 0.10939448326826096, "global_step": 157130, "epoch": 1765, "lr": 3.785772189945465e-05}
{"train_loss": 0.05998361110687256, "global_step": 157131, "epoch": 1765, "lr": 3.785715951810399e-05}
{"train_loss": 0.12985584139823914, "global_step": 157132, "epoch": 1765, "lr": 3.785659713838581e-05}
{"train_loss": 0.10992154479026794, "global_step": 157133, "epoch": 1765, "lr": 3.785603476030013e-05}
{"train_loss": 0.07211798429489136, "global_step": 157134, "epoch": 1765, "lr": 3.785547238384709e-05}
{"train_loss": 0.1301201432943344, "global_step": 157135, "epoch": 1765, "lr": 3.7854910009026704e-05}
{"train_loss": 0.07005555927753448, "global_step": 157136, "epoch": 1765, "lr": 3.78543476358391e-05}
{"train_loss": 0.10919606685638428, "global_step": 157137, "epoch": 1765, "lr": 3.785378526428433e-05}
{"train_loss": 0.07331937551498413, "global_step": 157138, "epoch": 1765, "lr": 3.7853222894362464e-05}
{"train_loss": 0.13118992745876312, "global_step": 157139, "epoch": 1765, "lr": 3.785266052607358e-05}
{"train_loss": 0.11139166355133057, "global_step": 157140, "epoch": 1765, "lr": 3.785209815941776e-05}
{"train_loss": 0.1469877064228058, "global_step": 157141, "epoch": 1765, "lr": 3.7851535794395085e-05}
{"train_loss": 0.04980142414569855, "global_step": 157142, "epoch": 1765, "lr": 3.785097343100562e-05}
{"train_loss": 0.07160218060016632, "global_step": 157143, "epoch": 1765, "lr": 3.785041106924943e-05}
{"train_loss": 0.07517765462398529, "global_step": 157144, "epoch": 1765, "lr": 3.784984870912663e-05}
{"train_loss": 0.1339564472436905, "global_step": 157145, "epoch": 1765, "lr": 3.784928635063724e-05}
{"train_loss": 0.07048599421977997, "global_step": 157146, "epoch": 1765, "lr": 3.784872399378139e-05}
{"train_loss": 0.13765458762645721, "global_step": 157147, "epoch": 1765, "lr": 3.784816163855911e-05}
{"train_loss": 0.08848337829113007, "global_step": 157148, "epoch": 1765, "lr": 3.78475992849705e-05}
{"train_loss": 0.0728660523891449, "global_step": 157149, "epoch": 1765, "lr": 3.7847036933015645e-05}
{"train_loss": 0.09763786196708679, "global_step": 157150, "epoch": 1765, "lr": 3.784647458269459e-05}
{"train_loss": 0.05871254578232765, "global_step": 157151, "epoch": 1765, "lr": 3.784591223400745e-05}
{"train_loss": 0.055957306176424026, "global_step": 157152, "epoch": 1765, "lr": 3.784534988695426e-05}
{"train_loss": 0.15134865045547485, "global_step": 157153, "epoch": 1765, "lr": 3.784478754153514e-05}
{"train_loss": 0.19573023915290833, "global_step": 157154, "epoch": 1765, "lr": 3.784422519775011e-05}
{"train_loss": 0.08836918324232101, "global_step": 157155, "epoch": 1765, "lr": 3.7843662855599304e-05}
{"train_loss": 0.11131633818149567, "global_step": 157156, "epoch": 1765, "lr": 3.784310051508275e-05}
{"train_loss": 0.09799489378929138, "global_step": 157157, "epoch": 1765, "lr": 3.7842538176200545e-05}
{"train_loss": 0.09256584197282791, "global_step": 157158, "epoch": 1765, "lr": 3.784197583895277e-05}
{"train_loss": 0.09489008784294128, "global_step": 157159, "epoch": 1765, "lr": 3.784141350333949e-05}
{"train_loss": 0.10014160722494125, "global_step": 157160, "epoch": 1765, "lr": 3.784085116936078e-05}
{"train_loss": 0.07048254460096359, "global_step": 157161, "epoch": 1765, "lr": 3.784028883701673e-05}
{"train_loss": 0.10113977640867233, "global_step": 157162, "epoch": 1765, "lr": 3.783972650630739e-05}
{"train_loss": 0.09245548397302628, "global_step": 157163, "epoch": 1765, "lr": 3.783916417723287e-05}
{"train_loss": 0.05891348049044609, "global_step": 157164, "epoch": 1765, "lr": 3.78386018497932e-05}
{"train_loss": 0.12820972502231598, "global_step": 157165, "epoch": 1765, "lr": 3.783803952398851e-05}
{"train_loss": 0.15860286355018616, "global_step": 157166, "epoch": 1765, "lr": 3.783747719981882e-05}
{"train_loss": 0.19228148460388184, "global_step": 157167, "epoch": 1765, "lr": 3.7836914877284236e-05}
{"train_loss": 0.100545234978199, "global_step": 157168, "epoch": 1765, "lr": 3.783635255638485e-05}
{"train_loss": 0.11186101287603378, "global_step": 157169, "epoch": 1765, "lr": 3.78357902371207e-05}
{"train_loss": 0.08679541945457458, "global_step": 157170, "epoch": 1765, "lr": 3.78352279194919e-05}
{"train_loss": 0.12140394002199173, "global_step": 157171, "epoch": 1765, "lr": 3.783466560349848e-05}
{"train_loss": 0.05060335248708725, "global_step": 157172, "epoch": 1765, "lr": 3.7834103289140564e-05}
{"train_loss": 0.10176868339100581, "global_step": 157173, "epoch": 1765, "lr": 3.7833540976418184e-05, "val_loss": 6.454873561859131, "train_action_mse_error": 29.497920989990234}
{"train_loss": 0.11586110293865204, "global_step": 157174, "epoch": 1766, "lr": 3.783297866533146e-05}
{"train_loss": 0.0655956044793129, "global_step": 157175, "epoch": 1766, "lr": 3.783241635588043e-05}
{"train_loss": 0.13570356369018555, "global_step": 157176, "epoch": 1766, "lr": 3.783185404806518e-05}
{"train_loss": 0.06347475200891495, "global_step": 157177, "epoch": 1766, "lr": 3.78312917418858e-05}
{"train_loss": 0.07107081264257431, "global_step": 157178, "epoch": 1766, "lr": 3.783072943734235e-05}
{"train_loss": 0.1368044912815094, "global_step": 157179, "epoch": 1766, "lr": 3.78301671344349e-05}
{"train_loss": 0.1213461384177208, "global_step": 157180, "epoch": 1766, "lr": 3.7829604833163554e-05}
{"train_loss": 0.08267424255609512, "global_step": 157181, "epoch": 1766, "lr": 3.7829042533528355e-05}
{"train_loss": 0.09256650507450104, "global_step": 157182, "epoch": 1766, "lr": 3.782848023552941e-05}
{"train_loss": 0.12375156581401825, "global_step": 157183, "epoch": 1766, "lr": 3.7827917939166744e-05}
{"train_loss": 0.08701565861701965, "global_step": 157184, "epoch": 1766, "lr": 3.7827355644440474e-05}
{"train_loss": 0.1282689869403839, "global_step": 157185, "epoch": 1766, "lr": 3.78267933513507e-05}
{"train_loss": 0.08471643179655075, "global_step": 157186, "epoch": 1766, "lr": 3.782623105989743e-05}
{"train_loss": 0.07326429337263107, "global_step": 157187, "epoch": 1766, "lr": 3.78256687700808e-05}
{"train_loss": 0.11681200563907623, "global_step": 157188, "epoch": 1766, "lr": 3.782510648190084e-05}
{"train_loss": 0.11361756175756454, "global_step": 157189, "epoch": 1766, "lr": 3.782454419535767e-05}
{"train_loss": 0.1432332843542099, "global_step": 157190, "epoch": 1766, "lr": 3.782398191045131e-05}
{"train_loss": 0.10839168727397919, "global_step": 157191, "epoch": 1766, "lr": 3.78234196271819e-05}
{"train_loss": 0.1040380597114563, "global_step": 157192, "epoch": 1766, "lr": 3.782285734554946e-05}
{"train_loss": 0.06376020610332489, "global_step": 157193, "epoch": 1766, "lr": 3.78222950655541e-05}
{"train_loss": 0.06454090029001236, "global_step": 157194, "epoch": 1766, "lr": 3.782173278719588e-05}
{"train_loss": 0.11548460274934769, "global_step": 157195, "epoch": 1766, "lr": 3.782117051047488e-05}
{"train_loss": 0.12929806113243103, "global_step": 157196, "epoch": 1766, "lr": 3.782060823539118e-05}
{"train_loss": 0.05364986136555672, "global_step": 157197, "epoch": 1766, "lr": 3.782004596194485e-05}
{"train_loss": 0.09815987199544907, "global_step": 157198, "epoch": 1766, "lr": 3.7819483690135957e-05}
{"train_loss": 0.09564321488142014, "global_step": 157199, "epoch": 1766, "lr": 3.7818921419964596e-05}
{"train_loss": 0.055449362844228745, "global_step": 157200, "epoch": 1766, "lr": 3.781835915143083e-05}
{"train_loss": 0.06844094395637512, "global_step": 157201, "epoch": 1766, "lr": 3.781779688453474e-05}
{"train_loss": 0.0734032616019249, "global_step": 157202, "epoch": 1766, "lr": 3.7817234619276387e-05}
{"train_loss": 0.12553375959396362, "global_step": 157203, "epoch": 1766, "lr": 3.781667235565585e-05}
{"train_loss": 0.051000919193029404, "global_step": 157204, "epoch": 1766, "lr": 3.781611009367324e-05}
{"train_loss": 0.05879141017794609, "global_step": 157205, "epoch": 1766, "lr": 3.7815547833328584e-05}
{"train_loss": 0.15658998489379883, "global_step": 157206, "epoch": 1766, "lr": 3.7814985574622e-05}
{"train_loss": 0.0510261133313179, "global_step": 157207, "epoch": 1766, "lr": 3.781442331755352e-05}
{"train_loss": 0.09303024411201477, "global_step": 157208, "epoch": 1766, "lr": 3.781386106212326e-05}
{"train_loss": 0.09249986708164215, "global_step": 157209, "epoch": 1766, "lr": 3.781329880833127e-05}
{"train_loss": 0.063715361058712, "global_step": 157210, "epoch": 1766, "lr": 3.781273655617764e-05}
{"train_loss": 0.10219799727201462, "global_step": 157211, "epoch": 1766, "lr": 3.781217430566242e-05}
{"train_loss": 0.11536414176225662, "global_step": 157212, "epoch": 1766, "lr": 3.7811612056785734e-05}
{"train_loss": 0.07884518057107925, "global_step": 157213, "epoch": 1766, "lr": 3.78110498095476e-05}
{"train_loss": 0.06703470647335052, "global_step": 157214, "epoch": 1766, "lr": 3.7810487563948144e-05}
{"train_loss": 0.1106734499335289, "global_step": 157215, "epoch": 1766, "lr": 3.780992531998741e-05}
{"train_loss": 0.10499714314937592, "global_step": 157216, "epoch": 1766, "lr": 3.7809363077665484e-05}
{"train_loss": 0.1612715870141983, "global_step": 157217, "epoch": 1766, "lr": 3.780880083698244e-05}
{"train_loss": 0.12629112601280212, "global_step": 157218, "epoch": 1766, "lr": 3.780823859793835e-05}
{"train_loss": 0.11781415343284607, "global_step": 157219, "epoch": 1766, "lr": 3.78076763605333e-05}
{"train_loss": 0.11334318667650223, "global_step": 157220, "epoch": 1766, "lr": 3.780711412476735e-05}
{"train_loss": 0.1280534714460373, "global_step": 157221, "epoch": 1766, "lr": 3.780655189064059e-05}
{"train_loss": 0.11995714157819748, "global_step": 157222, "epoch": 1766, "lr": 3.780598965815309e-05}
{"train_loss": 0.07867222279310226, "global_step": 157223, "epoch": 1766, "lr": 3.780542742730494e-05}
{"train_loss": 0.11538227647542953, "global_step": 157224, "epoch": 1766, "lr": 3.780486519809617e-05}
{"train_loss": 0.04814967140555382, "global_step": 157225, "epoch": 1766, "lr": 3.780430297052692e-05}
{"train_loss": 0.08138945698738098, "global_step": 157226, "epoch": 1766, "lr": 3.7803740744597216e-05}
{"train_loss": 0.06739422678947449, "global_step": 157227, "epoch": 1766, "lr": 3.7803178520307156e-05}
{"train_loss": 0.12378204613924026, "global_step": 157228, "epoch": 1766, "lr": 3.7802616297656803e-05}
{"train_loss": 0.10084199905395508, "global_step": 157229, "epoch": 1766, "lr": 3.780205407664625e-05}
{"train_loss": 0.04926092177629471, "global_step": 157230, "epoch": 1766, "lr": 3.780149185727555e-05}
{"train_loss": 0.0821453407406807, "global_step": 157231, "epoch": 1766, "lr": 3.7800929639544805e-05}
{"train_loss": 0.14400792121887207, "global_step": 157232, "epoch": 1766, "lr": 3.7800367423454065e-05}
{"train_loss": 0.07970578968524933, "global_step": 157233, "epoch": 1766, "lr": 3.779980520900343e-05}
{"train_loss": 0.09020531922578812, "global_step": 157234, "epoch": 1766, "lr": 3.779924299619295e-05}
{"train_loss": 0.06851768493652344, "global_step": 157235, "epoch": 1766, "lr": 3.779868078502272e-05}
{"train_loss": 0.07780759036540985, "global_step": 157236, "epoch": 1766, "lr": 3.77981185754928e-05}
{"train_loss": 0.06622248888015747, "global_step": 157237, "epoch": 1766, "lr": 3.7797556367603285e-05}
{"train_loss": 0.12441600114107132, "global_step": 157238, "epoch": 1766, "lr": 3.7796994161354235e-05}
{"train_loss": 0.08219041675329208, "global_step": 157239, "epoch": 1766, "lr": 3.779643195674572e-05}
{"train_loss": 0.11428418755531311, "global_step": 157240, "epoch": 1766, "lr": 3.779586975377784e-05}
{"train_loss": 0.13203005492687225, "global_step": 157241, "epoch": 1766, "lr": 3.779530755245064e-05}
{"train_loss": 0.06992162019014359, "global_step": 157242, "epoch": 1766, "lr": 3.779474535276424e-05}
{"train_loss": 0.15740875899791718, "global_step": 157243, "epoch": 1766, "lr": 3.779418315471866e-05}
{"train_loss": 0.13133013248443604, "global_step": 157244, "epoch": 1766, "lr": 3.7793620958314027e-05}
{"train_loss": 0.0983167365193367, "global_step": 157245, "epoch": 1766, "lr": 3.779305876355037e-05}
{"train_loss": 0.0617571584880352, "global_step": 157246, "epoch": 1766, "lr": 3.77924965704278e-05}
{"train_loss": 0.10172681510448456, "global_step": 157247, "epoch": 1766, "lr": 3.779193437894638e-05}
{"train_loss": 0.07328791171312332, "global_step": 157248, "epoch": 1766, "lr": 3.7791372189106186e-05}
{"train_loss": 0.0738566443324089, "global_step": 157249, "epoch": 1766, "lr": 3.7790810000907286e-05}
{"train_loss": 0.13048915565013885, "global_step": 157250, "epoch": 1766, "lr": 3.779024781434978e-05}
{"train_loss": 0.05508057400584221, "global_step": 157251, "epoch": 1766, "lr": 3.77896856294337e-05}
{"train_loss": 0.04068605601787567, "global_step": 157252, "epoch": 1766, "lr": 3.778912344615917e-05}
{"train_loss": 0.09099695086479187, "global_step": 157253, "epoch": 1766, "lr": 3.778856126452624e-05}
{"train_loss": 0.12183038890361786, "global_step": 157254, "epoch": 1766, "lr": 3.778799908453499e-05}
{"train_loss": 0.08222737908363342, "global_step": 157255, "epoch": 1766, "lr": 3.778743690618549e-05}
{"train_loss": 0.14390534162521362, "global_step": 157256, "epoch": 1766, "lr": 3.77868747294778e-05}
{"train_loss": 0.07231278717517853, "global_step": 157257, "epoch": 1766, "lr": 3.7786312554412053e-05}
{"train_loss": 0.054449062794446945, "global_step": 157258, "epoch": 1766, "lr": 3.778575038098826e-05}
{"train_loss": 0.10032864660024643, "global_step": 157259, "epoch": 1766, "lr": 3.7785188209206546e-05}
{"train_loss": 0.07250047475099564, "global_step": 157260, "epoch": 1766, "lr": 3.778462603906694e-05}
{"train_loss": 0.0677654817700386, "global_step": 157261, "epoch": 1766, "lr": 3.778406387056956e-05}
{"train_loss": 0.09463943457335569, "global_step": 157262, "epoch": 1766, "lr": 3.7783501703714455e-05, "val_loss": 6.6493306159973145}
{"train_loss": 0.09950366616249084, "global_step": 157263, "epoch": 1767, "lr": 3.778293953850172e-05}
{"train_loss": 0.109145887196064, "global_step": 157264, "epoch": 1767, "lr": 3.7782377374931405e-05}
{"train_loss": 0.14846065640449524, "global_step": 157265, "epoch": 1767, "lr": 3.778181521300362e-05}
{"train_loss": 0.07683242857456207, "global_step": 157266, "epoch": 1767, "lr": 3.77812530527184e-05}
{"train_loss": 0.06532382220029831, "global_step": 157267, "epoch": 1767, "lr": 3.778069089407586e-05}
{"train_loss": 0.09359150379896164, "global_step": 157268, "epoch": 1767, "lr": 3.778012873707604e-05}
{"train_loss": 0.07129707932472229, "global_step": 157269, "epoch": 1767, "lr": 3.7779566581719054e-05}
{"train_loss": 0.12672708928585052, "global_step": 157270, "epoch": 1767, "lr": 3.7779004428004925e-05}
{"train_loss": 0.13199275732040405, "global_step": 157271, "epoch": 1767, "lr": 3.777844227593379e-05}
{"train_loss": 0.07907240837812424, "global_step": 157272, "epoch": 1767, "lr": 3.777788012550568e-05}
{"train_loss": 0.10142839699983597, "global_step": 157273, "epoch": 1767, "lr": 3.7777317976720686e-05}
{"train_loss": 0.10112342983484268, "global_step": 157274, "epoch": 1767, "lr": 3.777675582957889e-05}
{"train_loss": 0.19022119045257568, "global_step": 157275, "epoch": 1767, "lr": 3.777619368408034e-05}
{"train_loss": 0.09696328639984131, "global_step": 157276, "epoch": 1767, "lr": 3.777563154022516e-05}
{"train_loss": 0.08236303180456161, "global_step": 157277, "epoch": 1767, "lr": 3.7775069398013374e-05}
{"train_loss": 0.10495414584875107, "global_step": 157278, "epoch": 1767, "lr": 3.77745072574451e-05}
{"train_loss": 0.11160949617624283, "global_step": 157279, "epoch": 1767, "lr": 3.777394511852037e-05}
{"train_loss": 0.043780941516160965, "global_step": 157280, "epoch": 1767, "lr": 3.777338298123931e-05}
{"train_loss": 0.09192994236946106, "global_step": 157281, "epoch": 1767, "lr": 3.777282084560195e-05}
{"train_loss": 0.06455930322408676, "global_step": 157282, "epoch": 1767, "lr": 3.777225871160841e-05}
{"train_loss": 0.108748659491539, "global_step": 157283, "epoch": 1767, "lr": 3.7771696579258706e-05}
{"train_loss": 0.1182018369436264, "global_step": 157284, "epoch": 1767, "lr": 3.777113444855298e-05}
{"train_loss": 0.07762325555086136, "global_step": 157285, "epoch": 1767, "lr": 3.777057231949126e-05}
{"train_loss": 0.07275209575891495, "global_step": 157286, "epoch": 1767, "lr": 3.7770010192073645e-05}
{"train_loss": 0.10937424749135971, "global_step": 157287, "epoch": 1767, "lr": 3.776944806630019e-05}
{"train_loss": 0.10443970561027527, "global_step": 157288, "epoch": 1767, "lr": 3.7768885942171e-05}
{"train_loss": 0.11990253627300262, "global_step": 157289, "epoch": 1767, "lr": 3.776832381968612e-05}
{"train_loss": 0.07495987415313721, "global_step": 157290, "epoch": 1767, "lr": 3.776776169884565e-05}
{"train_loss": 0.09754251688718796, "global_step": 157291, "epoch": 1767, "lr": 3.7767199579649654e-05}
{"train_loss": 0.09940829873085022, "global_step": 157292, "epoch": 1767, "lr": 3.776663746209821e-05}
{"train_loss": 0.10464435815811157, "global_step": 157293, "epoch": 1767, "lr": 3.776607534619139e-05}
{"train_loss": 0.08290770649909973, "global_step": 157294, "epoch": 1767, "lr": 3.7765513231929254e-05}
{"train_loss": 0.06270430982112885, "global_step": 157295, "epoch": 1767, "lr": 3.7764951119311924e-05}
{"train_loss": 0.09090352058410645, "global_step": 157296, "epoch": 1767, "lr": 3.776438900833943e-05}
{"train_loss": 0.047415390610694885, "global_step": 157297, "epoch": 1767, "lr": 3.776382689901188e-05}
{"train_loss": 0.09311352670192719, "global_step": 157298, "epoch": 1767, "lr": 3.776326479132931e-05}
{"train_loss": 0.078766368329525, "global_step": 157299, "epoch": 1767, "lr": 3.776270268529185e-05}
{"train_loss": 0.06958365440368652, "global_step": 157300, "epoch": 1767, "lr": 3.776214058089952e-05}
{"train_loss": 0.11549636721611023, "global_step": 157301, "epoch": 1767, "lr": 3.776157847815244e-05}
{"train_loss": 0.13488350808620453, "global_step": 157302, "epoch": 1767, "lr": 3.776101637705065e-05}
{"train_loss": 0.12658807635307312, "global_step": 157303, "epoch": 1767, "lr": 3.776045427759426e-05}
{"train_loss": 0.07948168367147446, "global_step": 157304, "epoch": 1767, "lr": 3.775989217978331e-05}
{"train_loss": 0.09660139679908752, "global_step": 157305, "epoch": 1767, "lr": 3.7759330083617905e-05}
{"train_loss": 0.08608334511518478, "global_step": 157306, "epoch": 1767, "lr": 3.77587679890981e-05}
{"train_loss": 0.11138926446437836, "global_step": 157307, "epoch": 1767, "lr": 3.7758205896223993e-05}
{"train_loss": 0.1025577262043953, "global_step": 157308, "epoch": 1767, "lr": 3.775764380499563e-05}
{"train_loss": 0.09143824875354767, "global_step": 157309, "epoch": 1767, "lr": 3.775708171541312e-05}
{"train_loss": 0.055987272411584854, "global_step": 157310, "epoch": 1767, "lr": 3.775651962747651e-05}
{"train_loss": 0.10373126715421677, "global_step": 157311, "epoch": 1767, "lr": 3.7755957541185875e-05}
{"train_loss": 0.12027960270643234, "global_step": 157312, "epoch": 1767, "lr": 3.775539545654133e-05}
{"train_loss": 0.17717792093753815, "global_step": 157313, "epoch": 1767, "lr": 3.775483337354289e-05}
{"train_loss": 0.060496263206005096, "global_step": 157314, "epoch": 1767, "lr": 3.7754271292190696e-05}
{"train_loss": 0.11401073634624481, "global_step": 157315, "epoch": 1767, "lr": 3.775370921248477e-05}
{"train_loss": 0.09132043272256851, "global_step": 157316, "epoch": 1767, "lr": 3.775314713442522e-05}
{"train_loss": 0.1296325922012329, "global_step": 157317, "epoch": 1767, "lr": 3.7752585058012095e-05}
{"train_loss": 0.08063074201345444, "global_step": 157318, "epoch": 1767, "lr": 3.77520229832455e-05}
{"train_loss": 0.14909321069717407, "global_step": 157319, "epoch": 1767, "lr": 3.7751460910125494e-05}
{"train_loss": 0.11993462592363358, "global_step": 157320, "epoch": 1767, "lr": 3.775089883865216e-05}
{"train_loss": 0.13010184466838837, "global_step": 157321, "epoch": 1767, "lr": 3.775033676882554e-05}
{"train_loss": 0.15627291798591614, "global_step": 157322, "epoch": 1767, "lr": 3.774977470064578e-05}
{"train_loss": 0.08840031176805496, "global_step": 157323, "epoch": 1767, "lr": 3.774921263411288e-05}
{"train_loss": 0.09034430980682373, "global_step": 157324, "epoch": 1767, "lr": 3.7748650569226974e-05}
{"train_loss": 0.08384569734334946, "global_step": 157325, "epoch": 1767, "lr": 3.7748088505988086e-05}
{"train_loss": 0.0756300687789917, "global_step": 157326, "epoch": 1767, "lr": 3.774752644439634e-05}
{"train_loss": 0.06906820088624954, "global_step": 157327, "epoch": 1767, "lr": 3.774696438445178e-05}
{"train_loss": 0.10531727224588394, "global_step": 157328, "epoch": 1767, "lr": 3.77464023261545e-05}
{"train_loss": 0.12873584032058716, "global_step": 157329, "epoch": 1767, "lr": 3.774584026950455e-05}
{"train_loss": 0.05946343392133713, "global_step": 157330, "epoch": 1767, "lr": 3.774527821450203e-05}
{"train_loss": 0.09108182042837143, "global_step": 157331, "epoch": 1767, "lr": 3.774471616114702e-05}
{"train_loss": 0.07101941853761673, "global_step": 157332, "epoch": 1767, "lr": 3.774415410943957e-05}
{"train_loss": 0.07771507650613785, "global_step": 157333, "epoch": 1767, "lr": 3.774359205937978e-05}
{"train_loss": 0.14389392733573914, "global_step": 157334, "epoch": 1767, "lr": 3.77430300109677e-05}
{"train_loss": 0.046822674572467804, "global_step": 157335, "epoch": 1767, "lr": 3.774246796420344e-05}
{"train_loss": 0.13386580348014832, "global_step": 157336, "epoch": 1767, "lr": 3.774190591908703e-05}
{"train_loss": 0.09111235290765762, "global_step": 157337, "epoch": 1767, "lr": 3.7741343875618596e-05}
{"train_loss": 0.15699715912342072, "global_step": 157338, "epoch": 1767, "lr": 3.7740781833798163e-05}
{"train_loss": 0.12841570377349854, "global_step": 157339, "epoch": 1767, "lr": 3.774021979362587e-05}
{"train_loss": 0.10282628983259201, "global_step": 157340, "epoch": 1767, "lr": 3.773965775510171e-05}
{"train_loss": 0.16022463142871857, "global_step": 157341, "epoch": 1767, "lr": 3.7739095718225845e-05}
{"train_loss": 0.06353341788053513, "global_step": 157342, "epoch": 1767, "lr": 3.7738533682998276e-05}
{"train_loss": 0.11708855628967285, "global_step": 157343, "epoch": 1767, "lr": 3.7737971649419135e-05}
{"train_loss": 0.06758642941713333, "global_step": 157344, "epoch": 1767, "lr": 3.7737409617488455e-05}
{"train_loss": 0.21940653026103973, "global_step": 157345, "epoch": 1767, "lr": 3.773684758720635e-05}
{"train_loss": 0.2348633110523224, "global_step": 157346, "epoch": 1767, "lr": 3.773628555857286e-05}
{"train_loss": 0.10563122481107712, "global_step": 157347, "epoch": 1767, "lr": 3.773572353158809e-05}
{"train_loss": 0.10427346080541611, "global_step": 157348, "epoch": 1767, "lr": 3.7735161506252095e-05}
{"train_loss": 0.06349551677703857, "global_step": 157349, "epoch": 1767, "lr": 3.7734599482564957e-05}
{"train_loss": 0.11762213706970215, "global_step": 157350, "epoch": 1767, "lr": 3.773403746052677e-05}
{"train_loss": 0.102269625981872, "global_step": 157351, "epoch": 1767, "lr": 3.773347544013756e-05, "val_loss": 6.657754421234131}
{"train_loss": 0.11013578623533249, "global_step": 157352, "epoch": 1768, "lr": 3.773291342139746e-05}
{"train_loss": 0.09767456352710724, "global_step": 157353, "epoch": 1768, "lr": 3.773235140430651e-05}
{"train_loss": 0.07697537541389465, "global_step": 157354, "epoch": 1768, "lr": 3.773178938886481e-05}
{"train_loss": 0.06106456369161606, "global_step": 157355, "epoch": 1768, "lr": 3.77312273750724e-05}
{"train_loss": 0.08405342698097229, "global_step": 157356, "epoch": 1768, "lr": 3.77306653629294e-05}
{"train_loss": 0.11264168471097946, "global_step": 157357, "epoch": 1768, "lr": 3.773010335243583e-05}
{"train_loss": 0.08727619051933289, "global_step": 157358, "epoch": 1768, "lr": 3.772954134359183e-05}
{"train_loss": 0.06690321862697601, "global_step": 157359, "epoch": 1768, "lr": 3.772897933639742e-05}
{"train_loss": 0.10569317638874054, "global_step": 157360, "epoch": 1768, "lr": 3.7728417330852725e-05}
{"train_loss": 0.07324982434511185, "global_step": 157361, "epoch": 1768, "lr": 3.7727855326957766e-05}
{"train_loss": 0.214156374335289, "global_step": 157362, "epoch": 1768, "lr": 3.772729332471268e-05}
{"train_loss": 0.11544641852378845, "global_step": 157363, "epoch": 1768, "lr": 3.772673132411747e-05}
{"train_loss": 0.059332963079214096, "global_step": 157364, "epoch": 1768, "lr": 3.772616932517229e-05}
{"train_loss": 0.09799253195524216, "global_step": 157365, "epoch": 1768, "lr": 3.7725607327877146e-05}
{"train_loss": 0.10884460806846619, "global_step": 157366, "epoch": 1768, "lr": 3.772504533223216e-05}
{"train_loss": 0.11242140829563141, "global_step": 157367, "epoch": 1768, "lr": 3.772448333823739e-05}
{"train_loss": 0.07436928898096085, "global_step": 157368, "epoch": 1768, "lr": 3.772392134589291e-05}
{"train_loss": 0.07453490048646927, "global_step": 157369, "epoch": 1768, "lr": 3.7723359355198814e-05}
{"train_loss": 0.13118907809257507, "global_step": 157370, "epoch": 1768, "lr": 3.772279736615514e-05}
{"train_loss": 0.07091005891561508, "global_step": 157371, "epoch": 1768, "lr": 3.7722235378762e-05}
{"train_loss": 0.10988768935203552, "global_step": 157372, "epoch": 1768, "lr": 3.772167339301945e-05}
{"train_loss": 0.1612340658903122, "global_step": 157373, "epoch": 1768, "lr": 3.772111140892759e-05}
{"train_loss": 0.05687381327152252, "global_step": 157374, "epoch": 1768, "lr": 3.772054942648645e-05}
{"train_loss": 0.10304414480924606, "global_step": 157375, "epoch": 1768, "lr": 3.7719987445696156e-05}
{"train_loss": 0.08172643184661865, "global_step": 157376, "epoch": 1768, "lr": 3.771942546655674e-05}
{"train_loss": 0.1464935839176178, "global_step": 157377, "epoch": 1768, "lr": 3.7718863489068314e-05}
{"train_loss": 0.10962030291557312, "global_step": 157378, "epoch": 1768, "lr": 3.7718301513230914e-05}
{"train_loss": 0.1267424374818802, "global_step": 157379, "epoch": 1768, "lr": 3.7717739539044666e-05}
{"train_loss": 0.08289531618356705, "global_step": 157380, "epoch": 1768, "lr": 3.771717756650959e-05}
{"train_loss": 0.09612971544265747, "global_step": 157381, "epoch": 1768, "lr": 3.771661559562582e-05}
{"train_loss": 0.13718530535697937, "global_step": 157382, "epoch": 1768, "lr": 3.7716053626393375e-05}
{"train_loss": 0.049491219222545624, "global_step": 157383, "epoch": 1768, "lr": 3.771549165881237e-05}
{"train_loss": 0.05026514455676079, "global_step": 157384, "epoch": 1768, "lr": 3.7714929692882864e-05}
{"train_loss": 0.11779306083917618, "global_step": 157385, "epoch": 1768, "lr": 3.7714367728604936e-05}
{"train_loss": 0.14090284705162048, "global_step": 157386, "epoch": 1768, "lr": 3.771380576597867e-05}
{"train_loss": 0.18280887603759766, "global_step": 157387, "epoch": 1768, "lr": 3.771324380500413e-05}
{"train_loss": 0.07938165962696075, "global_step": 157388, "epoch": 1768, "lr": 3.77126818456814e-05}
{"train_loss": 0.12797920405864716, "global_step": 157389, "epoch": 1768, "lr": 3.771211988801053e-05}
{"train_loss": 0.11998197436332703, "global_step": 157390, "epoch": 1768, "lr": 3.7711557931991635e-05}
{"train_loss": 0.12167954444885254, "global_step": 157391, "epoch": 1768, "lr": 3.771099597762475e-05}
{"train_loss": 0.04420919716358185, "global_step": 157392, "epoch": 1768, "lr": 3.771043402491e-05}
{"train_loss": 0.07779102027416229, "global_step": 157393, "epoch": 1768, "lr": 3.770987207384741e-05}
{"train_loss": 0.12788012623786926, "global_step": 157394, "epoch": 1768, "lr": 3.770931012443709e-05}
{"train_loss": 0.1700020283460617, "global_step": 157395, "epoch": 1768, "lr": 3.770874817667909e-05}
{"train_loss": 0.14093177020549774, "global_step": 157396, "epoch": 1768, "lr": 3.7708186230573516e-05}
{"train_loss": 0.11381752789020538, "global_step": 157397, "epoch": 1768, "lr": 3.7707624286120415e-05}
{"train_loss": 0.1378263235092163, "global_step": 157398, "epoch": 1768, "lr": 3.770706234331989e-05}
{"train_loss": 0.04265943542122841, "global_step": 157399, "epoch": 1768, "lr": 3.770650040217197e-05}
{"train_loss": 0.06687788665294647, "global_step": 157400, "epoch": 1768, "lr": 3.770593846267679e-05}
{"train_loss": 0.07082471996545792, "global_step": 157401, "epoch": 1768, "lr": 3.770537652483438e-05}
{"train_loss": 0.11638383567333221, "global_step": 157402, "epoch": 1768, "lr": 3.770481458864484e-05}
{"train_loss": 0.09312711656093597, "global_step": 157403, "epoch": 1768, "lr": 3.770425265410823e-05}
{"train_loss": 0.08086752146482468, "global_step": 157404, "epoch": 1768, "lr": 3.770369072122464e-05}
{"train_loss": 0.07199010998010635, "global_step": 157405, "epoch": 1768, "lr": 3.770312878999414e-05}
{"train_loss": 0.10174078494310379, "global_step": 157406, "epoch": 1768, "lr": 3.770256686041681e-05}
{"train_loss": 0.12498755753040314, "global_step": 157407, "epoch": 1768, "lr": 3.770200493249272e-05}
{"train_loss": 0.06314819306135178, "global_step": 157408, "epoch": 1768, "lr": 3.770144300622193e-05}
{"train_loss": 0.13009096682071686, "global_step": 157409, "epoch": 1768, "lr": 3.770088108160454e-05}
{"train_loss": 0.08851998299360275, "global_step": 157410, "epoch": 1768, "lr": 3.770031915864061e-05}
{"train_loss": 0.16013385355472565, "global_step": 157411, "epoch": 1768, "lr": 3.769975723733024e-05}
{"train_loss": 0.1006016805768013, "global_step": 157412, "epoch": 1768, "lr": 3.7699195317673465e-05}
{"train_loss": 0.10864685475826263, "global_step": 157413, "epoch": 1768, "lr": 3.7698633399670404e-05}
{"train_loss": 0.08877024799585342, "global_step": 157414, "epoch": 1768, "lr": 3.76980714833211e-05}
{"train_loss": 0.09292283654212952, "global_step": 157415, "epoch": 1768, "lr": 3.7697509568625646e-05}
{"train_loss": 0.11131976544857025, "global_step": 157416, "epoch": 1768, "lr": 3.76969476555841e-05}
{"train_loss": 0.0760757103562355, "global_step": 157417, "epoch": 1768, "lr": 3.769638574419657e-05}
{"train_loss": 0.06892206519842148, "global_step": 157418, "epoch": 1768, "lr": 3.769582383446308e-05}
{"train_loss": 0.08806122094392776, "global_step": 157419, "epoch": 1768, "lr": 3.769526192638377e-05}
{"train_loss": 0.06609879434108734, "global_step": 157420, "epoch": 1768, "lr": 3.7694700019958654e-05}
{"train_loss": 0.15513625741004944, "global_step": 157421, "epoch": 1768, "lr": 3.7694138115187854e-05}
{"train_loss": 0.07975807785987854, "global_step": 157422, "epoch": 1768, "lr": 3.7693576212071424e-05}
{"train_loss": 0.13974052667617798, "global_step": 157423, "epoch": 1768, "lr": 3.769301431060944e-05}
{"train_loss": 0.06858016550540924, "global_step": 157424, "epoch": 1768, "lr": 3.769245241080198e-05}
{"train_loss": 0.06340236961841583, "global_step": 157425, "epoch": 1768, "lr": 3.769189051264912e-05}
{"train_loss": 0.07886447757482529, "global_step": 157426, "epoch": 1768, "lr": 3.7691328616150935e-05}
{"train_loss": 0.11564358323812485, "global_step": 157427, "epoch": 1768, "lr": 3.76907667213075e-05}
{"train_loss": 0.06516817957162857, "global_step": 157428, "epoch": 1768, "lr": 3.769020482811891e-05}
{"train_loss": 0.1430608034133911, "global_step": 157429, "epoch": 1768, "lr": 3.768964293658519e-05}
{"train_loss": 0.10196787118911743, "global_step": 157430, "epoch": 1768, "lr": 3.7689081046706474e-05}
{"train_loss": 0.053973447531461716, "global_step": 157431, "epoch": 1768, "lr": 3.7688519158482794e-05}
{"train_loss": 0.10622235387563705, "global_step": 157432, "epoch": 1768, "lr": 3.768795727191425e-05}
{"train_loss": 0.1347714066505432, "global_step": 157433, "epoch": 1768, "lr": 3.7687395387000904e-05}
{"train_loss": 0.06023101508617401, "global_step": 157434, "epoch": 1768, "lr": 3.768683350374285e-05}
{"train_loss": 0.14248371124267578, "global_step": 157435, "epoch": 1768, "lr": 3.7686271622140136e-05}
{"train_loss": 0.0629730150103569, "global_step": 157436, "epoch": 1768, "lr": 3.768570974219287e-05}
{"train_loss": 0.03636525943875313, "global_step": 157437, "epoch": 1768, "lr": 3.768514786390109e-05}
{"train_loss": 0.10480460524559021, "global_step": 157438, "epoch": 1768, "lr": 3.768458598726489e-05}
{"train_loss": 0.06594132632017136, "global_step": 157439, "epoch": 1768, "lr": 3.7684024112284374e-05}
{"train_loss": 0.09965433368689558, "global_step": 157440, "epoch": 1768, "lr": 3.7683462238959564e-05, "val_loss": 6.611464977264404}
{"train_loss": 0.09218120574951172, "global_step": 157441, "epoch": 1769, "lr": 3.768290036729059e-05}
{"train_loss": 0.057979993522167206, "global_step": 157442, "epoch": 1769, "lr": 3.768233849727748e-05}
{"train_loss": 0.06827818602323532, "global_step": 157443, "epoch": 1769, "lr": 3.768177662892033e-05}
{"train_loss": 0.06770891696214676, "global_step": 157444, "epoch": 1769, "lr": 3.768121476221922e-05}
{"train_loss": 0.08470235764980316, "global_step": 157445, "epoch": 1769, "lr": 3.7680652897174217e-05}
{"train_loss": 0.13244202733039856, "global_step": 157446, "epoch": 1769, "lr": 3.76800910337854e-05}
{"train_loss": 0.08166413754224777, "global_step": 157447, "epoch": 1769, "lr": 3.7679529172052866e-05}
{"train_loss": 0.1192554235458374, "global_step": 157448, "epoch": 1769, "lr": 3.767896731197664e-05}
{"train_loss": 0.1346949338912964, "global_step": 157449, "epoch": 1769, "lr": 3.7678405453556845e-05}
{"train_loss": 0.1242426186800003, "global_step": 157450, "epoch": 1769, "lr": 3.7677843596793514e-05}
{"train_loss": 0.1136573851108551, "global_step": 157451, "epoch": 1769, "lr": 3.7677281741686776e-05}
{"train_loss": 0.17869512736797333, "global_step": 157452, "epoch": 1769, "lr": 3.767671988823666e-05}
{"train_loss": 0.10403648763895035, "global_step": 157453, "epoch": 1769, "lr": 3.767615803644326e-05}
{"train_loss": 0.05996495112776756, "global_step": 157454, "epoch": 1769, "lr": 3.767559618630665e-05}
{"train_loss": 0.11698257923126221, "global_step": 157455, "epoch": 1769, "lr": 3.767503433782692e-05}
{"train_loss": 0.07746026664972305, "global_step": 157456, "epoch": 1769, "lr": 3.767447249100411e-05}
{"train_loss": 0.05506040155887604, "global_step": 157457, "epoch": 1769, "lr": 3.7673910645838315e-05}
{"train_loss": 0.0949825644493103, "global_step": 157458, "epoch": 1769, "lr": 3.767334880232964e-05}
{"train_loss": 0.09114429354667664, "global_step": 157459, "epoch": 1769, "lr": 3.767278696047811e-05}
{"train_loss": 0.14698004722595215, "global_step": 157460, "epoch": 1769, "lr": 3.767222512028384e-05}
{"train_loss": 0.13797691464424133, "global_step": 157461, "epoch": 1769, "lr": 3.7671663281746874e-05}
{"train_loss": 0.12964223325252533, "global_step": 157462, "epoch": 1769, "lr": 3.767110144486731e-05}
{"train_loss": 0.07389093190431595, "global_step": 157463, "epoch": 1769, "lr": 3.767053960964522e-05}
{"train_loss": 0.06538096815347672, "global_step": 157464, "epoch": 1769, "lr": 3.7669977776080664e-05}
{"train_loss": 0.11450162529945374, "global_step": 157465, "epoch": 1769, "lr": 3.766941594417374e-05}
{"train_loss": 0.09925918281078339, "global_step": 157466, "epoch": 1769, "lr": 3.7668854113924524e-05}
{"train_loss": 0.056874100118875504, "global_step": 157467, "epoch": 1769, "lr": 3.766829228533305e-05}
{"train_loss": 0.072783462703228, "global_step": 157468, "epoch": 1769, "lr": 3.766773045839945e-05}
{"train_loss": 0.11693612486124039, "global_step": 157469, "epoch": 1769, "lr": 3.766716863312375e-05}
{"train_loss": 0.16751743853092194, "global_step": 157470, "epoch": 1769, "lr": 3.766660680950608e-05}
{"train_loss": 0.08002741634845734, "global_step": 157471, "epoch": 1769, "lr": 3.766604498754645e-05}
{"train_loss": 0.08646111190319061, "global_step": 157472, "epoch": 1769, "lr": 3.7665483167244996e-05}
{"train_loss": 0.0626518502831459, "global_step": 157473, "epoch": 1769, "lr": 3.7664921348601746e-05}
{"train_loss": 0.16380588710308075, "global_step": 157474, "epoch": 1769, "lr": 3.7664359531616796e-05}
{"train_loss": 0.07599356025457382, "global_step": 157475, "epoch": 1769, "lr": 3.766379771629025e-05}
{"train_loss": 0.07957706600427628, "global_step": 157476, "epoch": 1769, "lr": 3.766323590262213e-05}
{"train_loss": 0.10011203587055206, "global_step": 157477, "epoch": 1769, "lr": 3.766267409061255e-05}
{"train_loss": 0.09942777454853058, "global_step": 157478, "epoch": 1769, "lr": 3.7662112280261565e-05}
{"train_loss": 0.16152416169643402, "global_step": 157479, "epoch": 1769, "lr": 3.766155047156928e-05}
{"train_loss": 0.08478851616382599, "global_step": 157480, "epoch": 1769, "lr": 3.7660988664535726e-05}
{"train_loss": 0.07067587971687317, "global_step": 157481, "epoch": 1769, "lr": 3.7660426859161016e-05}
{"train_loss": 0.12145105749368668, "global_step": 157482, "epoch": 1769, "lr": 3.7659865055445204e-05}
{"train_loss": 0.1340656727552414, "global_step": 157483, "epoch": 1769, "lr": 3.765930325338837e-05}
{"train_loss": 0.04445115849375725, "global_step": 157484, "epoch": 1769, "lr": 3.7658741452990596e-05}
{"train_loss": 0.06988446414470673, "global_step": 157485, "epoch": 1769, "lr": 3.765817965425197e-05}
{"train_loss": 0.09430575370788574, "global_step": 157486, "epoch": 1769, "lr": 3.7657617857172524e-05}
{"train_loss": 0.1318996399641037, "global_step": 157487, "epoch": 1769, "lr": 3.765705606175238e-05}
{"train_loss": 0.11476832628250122, "global_step": 157488, "epoch": 1769, "lr": 3.765649426799158e-05}
{"train_loss": 0.14289982616901398, "global_step": 157489, "epoch": 1769, "lr": 3.765593247589023e-05}
{"train_loss": 0.15089307725429535, "global_step": 157490, "epoch": 1769, "lr": 3.7655370685448365e-05}
{"train_loss": 0.12524019181728363, "global_step": 157491, "epoch": 1769, "lr": 3.765480889666611e-05}
{"train_loss": 0.1266038417816162, "global_step": 157492, "epoch": 1769, "lr": 3.7654247109543496e-05}
{"train_loss": 0.13196666538715363, "global_step": 157493, "epoch": 1769, "lr": 3.7653685324080614e-05}
{"train_loss": 0.09733042120933533, "global_step": 157494, "epoch": 1769, "lr": 3.765312354027757e-05}
{"train_loss": 0.09981034696102142, "global_step": 157495, "epoch": 1769, "lr": 3.7652561758134386e-05}
{"train_loss": 0.12198524177074432, "global_step": 157496, "epoch": 1769, "lr": 3.765199997765119e-05}
{"train_loss": 0.10343870520591736, "global_step": 157497, "epoch": 1769, "lr": 3.765143819882801e-05}
{"train_loss": 0.08639822900295258, "global_step": 157498, "epoch": 1769, "lr": 3.7650876421664965e-05}
{"train_loss": 0.07052774727344513, "global_step": 157499, "epoch": 1769, "lr": 3.76503146461621e-05}
{"train_loss": 0.06895851343870163, "global_step": 157500, "epoch": 1769, "lr": 3.764975287231949e-05}
{"train_loss": 0.040206875652074814, "global_step": 157501, "epoch": 1769, "lr": 3.764919110013723e-05}
{"train_loss": 0.13157224655151367, "global_step": 157502, "epoch": 1769, "lr": 3.764862932961539e-05}
{"train_loss": 0.1432889848947525, "global_step": 157503, "epoch": 1769, "lr": 3.764806756075403e-05}
{"train_loss": 0.053571417927742004, "global_step": 157504, "epoch": 1769, "lr": 3.764750579355325e-05}
{"train_loss": 0.05861625447869301, "global_step": 157505, "epoch": 1769, "lr": 3.7646944028013095e-05}
{"train_loss": 0.11102375388145447, "global_step": 157506, "epoch": 1769, "lr": 3.7646382264133675e-05}
{"train_loss": 0.08382980525493622, "global_step": 157507, "epoch": 1769, "lr": 3.764582050191503e-05}
{"train_loss": 0.13202962279319763, "global_step": 157508, "epoch": 1769, "lr": 3.7645258741357275e-05}
{"train_loss": 0.06306988000869751, "global_step": 157509, "epoch": 1769, "lr": 3.764469698246045e-05}
{"train_loss": 0.15717066824436188, "global_step": 157510, "epoch": 1769, "lr": 3.764413522522463e-05}
{"train_loss": 0.08304513990879059, "global_step": 157511, "epoch": 1769, "lr": 3.764357346964994e-05}
{"train_loss": 0.13375481963157654, "global_step": 157512, "epoch": 1769, "lr": 3.764301171573639e-05}
{"train_loss": 0.1117955893278122, "global_step": 157513, "epoch": 1769, "lr": 3.764244996348411e-05}
{"train_loss": 0.1434253752231598, "global_step": 157514, "epoch": 1769, "lr": 3.7641888212893136e-05}
{"train_loss": 0.21390420198440552, "global_step": 157515, "epoch": 1769, "lr": 3.764132646396358e-05}
{"train_loss": 0.09119721502065659, "global_step": 157516, "epoch": 1769, "lr": 3.764076471669547e-05}
{"train_loss": 0.10570035874843597, "global_step": 157517, "epoch": 1769, "lr": 3.764020297108894e-05}
{"train_loss": 0.08356436342000961, "global_step": 157518, "epoch": 1769, "lr": 3.7639641227144004e-05}
{"train_loss": 0.1018945574760437, "global_step": 157519, "epoch": 1769, "lr": 3.763907948486078e-05}
{"train_loss": 0.0619942843914032, "global_step": 157520, "epoch": 1769, "lr": 3.7638517744239335e-05}
{"train_loss": 0.1464102566242218, "global_step": 157521, "epoch": 1769, "lr": 3.763795600527973e-05}
{"train_loss": 0.14601005613803864, "global_step": 157522, "epoch": 1769, "lr": 3.7637394267982065e-05}
{"train_loss": 0.1363762468099594, "global_step": 157523, "epoch": 1769, "lr": 3.76368325323464e-05}
{"train_loss": 0.13010646402835846, "global_step": 157524, "epoch": 1769, "lr": 3.76362707983728e-05}
{"train_loss": 0.0990428626537323, "global_step": 157525, "epoch": 1769, "lr": 3.763570906606137e-05}
{"train_loss": 0.1671178638935089, "global_step": 157526, "epoch": 1769, "lr": 3.763514733541215e-05}
{"train_loss": 0.2054852545261383, "global_step": 157527, "epoch": 1769, "lr": 3.763458560642525e-05}
{"train_loss": 0.1331539899110794, "global_step": 157528, "epoch": 1769, "lr": 3.76340238791007e-05}
{"train_loss": 0.10672499946831318, "global_step": 157529, "epoch": 1769, "lr": 3.763346215343862e-05, "val_loss": 6.698069095611572}
{"train_loss": 0.06879421323537827, "global_step": 157530, "epoch": 1770, "lr": 3.763290042943909e-05}
{"train_loss": 0.07467862963676453, "global_step": 157531, "epoch": 1770, "lr": 3.763233870710214e-05}
{"train_loss": 0.0473644956946373, "global_step": 157532, "epoch": 1770, "lr": 3.7631776986427894e-05}
{"train_loss": 0.13970987498760223, "global_step": 157533, "epoch": 1770, "lr": 3.763121526741638e-05}
{"train_loss": 0.09249962866306305, "global_step": 157534, "epoch": 1770, "lr": 3.7630653550067726e-05}
{"train_loss": 0.11651136726140976, "global_step": 157535, "epoch": 1770, "lr": 3.7630091834381954e-05}
{"train_loss": 0.1793075054883957, "global_step": 157536, "epoch": 1770, "lr": 3.7629530120359185e-05}
{"train_loss": 0.07429113239049911, "global_step": 157537, "epoch": 1770, "lr": 3.762896840799947e-05}
{"train_loss": 0.14328689873218536, "global_step": 157538, "epoch": 1770, "lr": 3.762840669730288e-05}
{"train_loss": 0.09664570540189743, "global_step": 157539, "epoch": 1770, "lr": 3.762784498826951e-05}
{"train_loss": 0.0988091379404068, "global_step": 157540, "epoch": 1770, "lr": 3.7627283280899414e-05}
{"train_loss": 0.08912891149520874, "global_step": 157541, "epoch": 1770, "lr": 3.762672157519269e-05}
{"train_loss": 0.1277342587709427, "global_step": 157542, "epoch": 1770, "lr": 3.76261598711494e-05}
{"train_loss": 0.1225225180387497, "global_step": 157543, "epoch": 1770, "lr": 3.762559816876962e-05}
{"train_loss": 0.0992182120680809, "global_step": 157544, "epoch": 1770, "lr": 3.7625036468053444e-05}
{"train_loss": 0.09125210344791412, "global_step": 157545, "epoch": 1770, "lr": 3.7624474769000905e-05}
{"train_loss": 0.0796641930937767, "global_step": 157546, "epoch": 1770, "lr": 3.762391307161213e-05}
{"train_loss": 0.13783639669418335, "global_step": 157547, "epoch": 1770, "lr": 3.762335137588714e-05}
{"train_loss": 0.048706650733947754, "global_step": 157548, "epoch": 1770, "lr": 3.762278968182604e-05}
{"train_loss": 0.05422427877783775, "global_step": 157549, "epoch": 1770, "lr": 3.7622227989428924e-05}
{"train_loss": 0.07973986864089966, "global_step": 157550, "epoch": 1770, "lr": 3.762166629869584e-05}
{"train_loss": 0.07408549636602402, "global_step": 157551, "epoch": 1770, "lr": 3.762110460962688e-05}
{"train_loss": 0.09431968629360199, "global_step": 157552, "epoch": 1770, "lr": 3.76205429222221e-05}
{"train_loss": 0.12127742171287537, "global_step": 157553, "epoch": 1770, "lr": 3.761998123648161e-05}
{"train_loss": 0.06534061580896378, "global_step": 157554, "epoch": 1770, "lr": 3.761941955240543e-05}
{"train_loss": 0.09842096269130707, "global_step": 157555, "epoch": 1770, "lr": 3.7618857869993695e-05}
{"train_loss": 0.06490272283554077, "global_step": 157556, "epoch": 1770, "lr": 3.761829618924644e-05}
{"train_loss": 0.09973929077386856, "global_step": 157557, "epoch": 1770, "lr": 3.761773451016377e-05}
{"train_loss": 0.07662554085254669, "global_step": 157558, "epoch": 1770, "lr": 3.761717283274573e-05}
{"train_loss": 0.09573167562484741, "global_step": 157559, "epoch": 1770, "lr": 3.761661115699241e-05}
{"train_loss": 0.06165093183517456, "global_step": 157560, "epoch": 1770, "lr": 3.7616049482903886e-05}
{"train_loss": 0.1260422170162201, "global_step": 157561, "epoch": 1770, "lr": 3.761548781048024e-05}
{"train_loss": 0.06858336925506592, "global_step": 157562, "epoch": 1770, "lr": 3.7614926139721536e-05}
{"train_loss": 0.05693753436207771, "global_step": 157563, "epoch": 1770, "lr": 3.761436447062786e-05}
{"train_loss": 0.12212680280208588, "global_step": 157564, "epoch": 1770, "lr": 3.7613802803199264e-05}
{"train_loss": 0.1048952043056488, "global_step": 157565, "epoch": 1770, "lr": 3.7613241137435846e-05}
{"train_loss": 0.1348133534193039, "global_step": 157566, "epoch": 1770, "lr": 3.76126794733377e-05}
{"train_loss": 0.0966130793094635, "global_step": 157567, "epoch": 1770, "lr": 3.761211781090484e-05}
{"train_loss": 0.0821998119354248, "global_step": 157568, "epoch": 1770, "lr": 3.761155615013742e-05}
{"train_loss": 0.09253980964422226, "global_step": 157569, "epoch": 1770, "lr": 3.761099449103545e-05}
{"train_loss": 0.09133180230855942, "global_step": 157570, "epoch": 1770, "lr": 3.761043283359905e-05}
{"train_loss": 0.09817644208669662, "global_step": 157571, "epoch": 1770, "lr": 3.7609871177828246e-05}
{"train_loss": 0.0884522795677185, "global_step": 157572, "epoch": 1770, "lr": 3.760930952372318e-05}
{"train_loss": 0.11871270090341568, "global_step": 157573, "epoch": 1770, "lr": 3.760874787128386e-05}
{"train_loss": 0.16798053681850433, "global_step": 157574, "epoch": 1770, "lr": 3.760818622051042e-05}
{"train_loss": 0.15348488092422485, "global_step": 157575, "epoch": 1770, "lr": 3.7607624571402884e-05}
{"train_loss": 0.03498736023902893, "global_step": 157576, "epoch": 1770, "lr": 3.760706292396137e-05}
{"train_loss": 0.06525924056768417, "global_step": 157577, "epoch": 1770, "lr": 3.760650127818593e-05}
{"train_loss": 0.14494876563549042, "global_step": 157578, "epoch": 1770, "lr": 3.7605939634076645e-05}
{"train_loss": 0.0717291459441185, "global_step": 157579, "epoch": 1770, "lr": 3.760537799163358e-05}
{"train_loss": 0.09248976409435272, "global_step": 157580, "epoch": 1770, "lr": 3.7604816350856835e-05}
{"train_loss": 0.051932353526353836, "global_step": 157581, "epoch": 1770, "lr": 3.760425471174647e-05}
{"train_loss": 0.09465007483959198, "global_step": 157582, "epoch": 1770, "lr": 3.7603693074302565e-05}
{"train_loss": 0.09317294508218765, "global_step": 157583, "epoch": 1770, "lr": 3.7603131438525176e-05}
{"train_loss": 0.05478166416287422, "global_step": 157584, "epoch": 1770, "lr": 3.760256980441439e-05}
{"train_loss": 0.11230331659317017, "global_step": 157585, "epoch": 1770, "lr": 3.760200817197031e-05}
{"train_loss": 0.10091234743595123, "global_step": 157586, "epoch": 1770, "lr": 3.760144654119297e-05}
{"train_loss": 0.10310158878564835, "global_step": 157587, "epoch": 1770, "lr": 3.7600884912082486e-05}
{"train_loss": 0.1246161088347435, "global_step": 157588, "epoch": 1770, "lr": 3.7600323284638886e-05}
{"train_loss": 0.10981374233961105, "global_step": 157589, "epoch": 1770, "lr": 3.75997616588623e-05}
{"train_loss": 0.14822962880134583, "global_step": 157590, "epoch": 1770, "lr": 3.759920003475275e-05}
{"train_loss": 0.14024680852890015, "global_step": 157591, "epoch": 1770, "lr": 3.7598638412310357e-05}
{"train_loss": 0.17516200244426727, "global_step": 157592, "epoch": 1770, "lr": 3.7598076791535155e-05}
{"train_loss": 0.174224853515625, "global_step": 157593, "epoch": 1770, "lr": 3.759751517242726e-05}
{"train_loss": 0.12354069948196411, "global_step": 157594, "epoch": 1770, "lr": 3.759695355498671e-05}
{"train_loss": 0.057979434728622437, "global_step": 157595, "epoch": 1770, "lr": 3.759639193921362e-05}
{"train_loss": 0.08631640672683716, "global_step": 157596, "epoch": 1770, "lr": 3.759583032510803e-05}
{"train_loss": 0.0679827556014061, "global_step": 157597, "epoch": 1770, "lr": 3.759526871267003e-05}
{"train_loss": 0.07691240310668945, "global_step": 157598, "epoch": 1770, "lr": 3.75947071018997e-05}
{"train_loss": 0.09051371365785599, "global_step": 157599, "epoch": 1770, "lr": 3.7594145492797104e-05}
{"train_loss": 0.09871107339859009, "global_step": 157600, "epoch": 1770, "lr": 3.7593583885362325e-05}
{"train_loss": 0.10831833630800247, "global_step": 157601, "epoch": 1770, "lr": 3.759302227959543e-05}
{"train_loss": 0.10379301756620407, "global_step": 157602, "epoch": 1770, "lr": 3.7592460675496524e-05}
{"train_loss": 0.1419878453016281, "global_step": 157603, "epoch": 1770, "lr": 3.759189907306564e-05}
{"train_loss": 0.11761734634637833, "global_step": 157604, "epoch": 1770, "lr": 3.759133747230288e-05}
{"train_loss": 0.0742659717798233, "global_step": 157605, "epoch": 1770, "lr": 3.759077587320831e-05}
{"train_loss": 0.10566338896751404, "global_step": 157606, "epoch": 1770, "lr": 3.759021427578202e-05}
{"train_loss": 0.03834384307265282, "global_step": 157607, "epoch": 1770, "lr": 3.7589652680024056e-05}
{"train_loss": 0.17076358199119568, "global_step": 157608, "epoch": 1770, "lr": 3.758909108593453e-05}
{"train_loss": 0.06770646572113037, "global_step": 157609, "epoch": 1770, "lr": 3.7588529493513495e-05}
{"train_loss": 0.0874209925532341, "global_step": 157610, "epoch": 1770, "lr": 3.758796790276104e-05}
{"train_loss": 0.09284437447786331, "global_step": 157611, "epoch": 1770, "lr": 3.758740631367721e-05}
{"train_loss": 0.15626391768455505, "global_step": 157612, "epoch": 1770, "lr": 3.758684472626211e-05}
{"train_loss": 0.053745441138744354, "global_step": 157613, "epoch": 1770, "lr": 3.758628314051581e-05}
{"train_loss": 0.054197438061237335, "global_step": 157614, "epoch": 1770, "lr": 3.758572155643839e-05}
{"train_loss": 0.09254355728626251, "global_step": 157615, "epoch": 1770, "lr": 3.7585159974029913e-05}
{"train_loss": 0.1458580493927002, "global_step": 157616, "epoch": 1770, "lr": 3.758459839329046e-05}
{"train_loss": 0.06281702220439911, "global_step": 157617, "epoch": 1770, "lr": 3.75840368142201e-05}
{"train_loss": 0.09913554532306917, "global_step": 157618, "epoch": 1770, "lr": 3.758347523681892e-05, "val_loss": 6.576328277587891, "train_action_mse_error": 19.99081802368164}
{"train_loss": 0.060211434960365295, "global_step": 157619, "epoch": 1771, "lr": 3.7582913661086994e-05}
{"train_loss": 0.0916384756565094, "global_step": 157620, "epoch": 1771, "lr": 3.758235208702438e-05}
{"train_loss": 0.12515708804130554, "global_step": 157621, "epoch": 1771, "lr": 3.758179051463119e-05}
{"train_loss": 0.11673267185688019, "global_step": 157622, "epoch": 1771, "lr": 3.758122894390745e-05}
{"train_loss": 0.047473784536123276, "global_step": 157623, "epoch": 1771, "lr": 3.7580667374853286e-05}
{"train_loss": 0.11272770166397095, "global_step": 157624, "epoch": 1771, "lr": 3.758010580746873e-05}
{"train_loss": 0.14761680364608765, "global_step": 157625, "epoch": 1771, "lr": 3.757954424175389e-05}
{"train_loss": 0.06893625855445862, "global_step": 157626, "epoch": 1771, "lr": 3.757898267770882e-05}
{"train_loss": 0.10691796243190765, "global_step": 157627, "epoch": 1771, "lr": 3.757842111533363e-05}
{"train_loss": 0.0970304012298584, "global_step": 157628, "epoch": 1771, "lr": 3.757785955462834e-05}
{"train_loss": 0.04591882973909378, "global_step": 157629, "epoch": 1771, "lr": 3.7577297995593075e-05}
{"train_loss": 0.09985106438398361, "global_step": 157630, "epoch": 1771, "lr": 3.757673643822787e-05}
{"train_loss": 0.10713780671358109, "global_step": 157631, "epoch": 1771, "lr": 3.757617488253285e-05}
{"train_loss": 0.04985431954264641, "global_step": 157632, "epoch": 1771, "lr": 3.757561332850803e-05}
{"train_loss": 0.09183265268802643, "global_step": 157633, "epoch": 1771, "lr": 3.757505177615354e-05}
{"train_loss": 0.08881691098213196, "global_step": 157634, "epoch": 1771, "lr": 3.757449022546943e-05}
{"train_loss": 0.07257208973169327, "global_step": 157635, "epoch": 1771, "lr": 3.757392867645577e-05}
{"train_loss": 0.17036786675453186, "global_step": 157636, "epoch": 1771, "lr": 3.7573367129112654e-05}
{"train_loss": 0.05870363116264343, "global_step": 157637, "epoch": 1771, "lr": 3.757280558344013e-05}
{"train_loss": 0.1282425969839096, "global_step": 157638, "epoch": 1771, "lr": 3.757224403943831e-05}
{"train_loss": 0.09405362606048584, "global_step": 157639, "epoch": 1771, "lr": 3.757168249710723e-05}
{"train_loss": 0.1070280373096466, "global_step": 157640, "epoch": 1771, "lr": 3.757112095644701e-05}
{"train_loss": 0.05159667134284973, "global_step": 157641, "epoch": 1771, "lr": 3.757055941745768e-05}
{"train_loss": 0.07652749866247177, "global_step": 157642, "epoch": 1771, "lr": 3.756999788013936e-05}
{"train_loss": 0.08589109778404236, "global_step": 157643, "epoch": 1771, "lr": 3.756943634449208e-05}
{"train_loss": 0.05206236243247986, "global_step": 157644, "epoch": 1771, "lr": 3.7568874810515955e-05}
{"train_loss": 0.07685883343219757, "global_step": 157645, "epoch": 1771, "lr": 3.756831327821102e-05}
{"train_loss": 0.11110089719295502, "global_step": 157646, "epoch": 1771, "lr": 3.7567751747577404e-05}
{"train_loss": 0.0737227126955986, "global_step": 157647, "epoch": 1771, "lr": 3.7567190218615125e-05}
{"train_loss": 0.10472621768712997, "global_step": 157648, "epoch": 1771, "lr": 3.75666286913243e-05}
{"train_loss": 0.07898177951574326, "global_step": 157649, "epoch": 1771, "lr": 3.7566067165704983e-05}
{"train_loss": 0.06199478358030319, "global_step": 157650, "epoch": 1771, "lr": 3.756550564175727e-05}
{"train_loss": 0.04056486859917641, "global_step": 157651, "epoch": 1771, "lr": 3.7564944119481204e-05}
{"train_loss": 0.07577145099639893, "global_step": 157652, "epoch": 1771, "lr": 3.7564382598876905e-05}
{"train_loss": 0.11724735796451569, "global_step": 157653, "epoch": 1771, "lr": 3.7563821079944405e-05}
{"train_loss": 0.10782463103532791, "global_step": 157654, "epoch": 1771, "lr": 3.75632595626838e-05}
{"train_loss": 0.11506189405918121, "global_step": 157655, "epoch": 1771, "lr": 3.756269804709516e-05}
{"train_loss": 0.1365092396736145, "global_step": 157656, "epoch": 1771, "lr": 3.756213653317856e-05}
{"train_loss": 0.06426595151424408, "global_step": 157657, "epoch": 1771, "lr": 3.7561575020934095e-05}
{"train_loss": 0.12261921167373657, "global_step": 157658, "epoch": 1771, "lr": 3.756101351036181e-05}
{"train_loss": 0.08556180447340012, "global_step": 157659, "epoch": 1771, "lr": 3.756045200146181e-05}
{"train_loss": 0.1778491586446762, "global_step": 157660, "epoch": 1771, "lr": 3.755989049423414e-05}
{"train_loss": 0.08916434645652771, "global_step": 157661, "epoch": 1771, "lr": 3.755932898867891e-05}
{"train_loss": 0.12404873222112656, "global_step": 157662, "epoch": 1771, "lr": 3.7558767484796156e-05}
{"train_loss": 0.04578273370862007, "global_step": 157663, "epoch": 1771, "lr": 3.755820598258599e-05}
{"train_loss": 0.10242467373609543, "global_step": 157664, "epoch": 1771, "lr": 3.755764448204846e-05}
{"train_loss": 0.07505837827920914, "global_step": 157665, "epoch": 1771, "lr": 3.7557082983183665e-05}
{"train_loss": 0.11257605999708176, "global_step": 157666, "epoch": 1771, "lr": 3.755652148599165e-05}
{"train_loss": 0.0401788167655468, "global_step": 157667, "epoch": 1771, "lr": 3.7555959990472534e-05}
{"train_loss": 0.09514199197292328, "global_step": 157668, "epoch": 1771, "lr": 3.755539849662634e-05}
{"train_loss": 0.10246376693248749, "global_step": 157669, "epoch": 1771, "lr": 3.7554837004453194e-05}
{"train_loss": 0.06651829928159714, "global_step": 157670, "epoch": 1771, "lr": 3.755427551395313e-05}
{"train_loss": 0.056688833981752396, "global_step": 157671, "epoch": 1771, "lr": 3.755371402512626e-05}
{"train_loss": 0.06308767199516296, "global_step": 157672, "epoch": 1771, "lr": 3.755315253797263e-05}
{"train_loss": 0.10589750111103058, "global_step": 157673, "epoch": 1771, "lr": 3.755259105249234e-05}
{"train_loss": 0.10967430472373962, "global_step": 157674, "epoch": 1771, "lr": 3.755202956868544e-05}
{"train_loss": 0.04100855439901352, "global_step": 157675, "epoch": 1771, "lr": 3.7551468086552e-05}
{"train_loss": 0.09962024539709091, "global_step": 157676, "epoch": 1771, "lr": 3.755090660609215e-05}
{"train_loss": 0.09783727675676346, "global_step": 157677, "epoch": 1771, "lr": 3.75503451273059e-05}
{"train_loss": 0.051822636276483536, "global_step": 157678, "epoch": 1771, "lr": 3.7549783650193384e-05}
{"train_loss": 0.11824139207601547, "global_step": 157679, "epoch": 1771, "lr": 3.7549222174754616e-05}
{"train_loss": 0.076961949467659, "global_step": 157680, "epoch": 1771, "lr": 3.754866070098973e-05}
{"train_loss": 0.14443637430667877, "global_step": 157681, "epoch": 1771, "lr": 3.7548099228898746e-05}
{"train_loss": 0.15520428121089935, "global_step": 157682, "epoch": 1771, "lr": 3.75475377584818e-05}
{"train_loss": 0.06658265739679337, "global_step": 157683, "epoch": 1771, "lr": 3.754697628973891e-05}
{"train_loss": 0.11312782764434814, "global_step": 157684, "epoch": 1771, "lr": 3.7546414822670194e-05}
{"train_loss": 0.07138717174530029, "global_step": 157685, "epoch": 1771, "lr": 3.754585335727569e-05}
{"train_loss": 0.07587000727653503, "global_step": 157686, "epoch": 1771, "lr": 3.754529189355553e-05}
{"train_loss": 0.08605539798736572, "global_step": 157687, "epoch": 1771, "lr": 3.754473043150972e-05}
{"train_loss": 0.1628981977701187, "global_step": 157688, "epoch": 1771, "lr": 3.7544168971138397e-05}
{"train_loss": 0.06911943107843399, "global_step": 157689, "epoch": 1771, "lr": 3.754360751244158e-05}
{"train_loss": 0.07887887954711914, "global_step": 157690, "epoch": 1771, "lr": 3.7543046055419396e-05}
{"train_loss": 0.10662760585546494, "global_step": 157691, "epoch": 1771, "lr": 3.754248460007188e-05}
{"train_loss": 0.12371370196342468, "global_step": 157692, "epoch": 1771, "lr": 3.7541923146399125e-05}
{"train_loss": 0.11258766055107117, "global_step": 157693, "epoch": 1771, "lr": 3.7541361694401235e-05}
{"train_loss": 0.13580381870269775, "global_step": 157694, "epoch": 1771, "lr": 3.754080024407822e-05}
{"train_loss": 0.10875707864761353, "global_step": 157695, "epoch": 1771, "lr": 3.7540238795430224e-05}
{"train_loss": 0.05939086526632309, "global_step": 157696, "epoch": 1771, "lr": 3.7539677348457265e-05}
{"train_loss": 0.05300137773156166, "global_step": 157697, "epoch": 1771, "lr": 3.753911590315947e-05}
{"train_loss": 0.15405859053134918, "global_step": 157698, "epoch": 1771, "lr": 3.753855445953686e-05}
{"train_loss": 0.07164686173200607, "global_step": 157699, "epoch": 1771, "lr": 3.753799301758957e-05}
{"train_loss": 0.1100861132144928, "global_step": 157700, "epoch": 1771, "lr": 3.7537431577317615e-05}
{"train_loss": 0.07981615513563156, "global_step": 157701, "epoch": 1771, "lr": 3.753687013872113e-05}
{"train_loss": 0.054429780691862106, "global_step": 157702, "epoch": 1771, "lr": 3.7536308701800126e-05}
{"train_loss": 0.0707736611366272, "global_step": 157703, "epoch": 1771, "lr": 3.753574726655475e-05}
{"train_loss": 0.13590586185455322, "global_step": 157704, "epoch": 1771, "lr": 3.753518583298501e-05}
{"train_loss": 0.09561826288700104, "global_step": 157705, "epoch": 1771, "lr": 3.753462440109104e-05}
{"train_loss": 0.09252641350030899, "global_step": 157706, "epoch": 1771, "lr": 3.7534062970872864e-05}
{"train_loss": 0.09330941196740343, "global_step": 157707, "epoch": 1771, "lr": 3.7533501542330603e-05, "val_loss": 6.61578893661499}
{"train_loss": 0.06408991664648056, "global_step": 157708, "epoch": 1772, "lr": 3.753294011546429e-05}
{"train_loss": 0.14867542684078217, "global_step": 157709, "epoch": 1772, "lr": 3.753237869027404e-05}
{"train_loss": 0.15101301670074463, "global_step": 157710, "epoch": 1772, "lr": 3.7531817266759886e-05}
{"train_loss": 0.09135586023330688, "global_step": 157711, "epoch": 1772, "lr": 3.753125584492194e-05}
{"train_loss": 0.10447333008050919, "global_step": 157712, "epoch": 1772, "lr": 3.7530694424760284e-05}
{"train_loss": 0.08349531888961792, "global_step": 157713, "epoch": 1772, "lr": 3.753013300627494e-05}
{"train_loss": 0.10631860047578812, "global_step": 157714, "epoch": 1772, "lr": 3.7529571589466045e-05}
{"train_loss": 0.06809248775243759, "global_step": 157715, "epoch": 1772, "lr": 3.7529010174333624e-05}
{"train_loss": 0.08972514420747757, "global_step": 157716, "epoch": 1772, "lr": 3.75284487608778e-05}
{"train_loss": 0.10054942965507507, "global_step": 157717, "epoch": 1772, "lr": 3.75278873490986e-05}
{"train_loss": 0.09494023025035858, "global_step": 157718, "epoch": 1772, "lr": 3.752732593899615e-05}
{"train_loss": 0.11762848496437073, "global_step": 157719, "epoch": 1772, "lr": 3.752676453057047e-05}
{"train_loss": 0.07853709161281586, "global_step": 157720, "epoch": 1772, "lr": 3.752620312382169e-05}
{"train_loss": 0.15595999360084534, "global_step": 157721, "epoch": 1772, "lr": 3.752564171874984e-05}
{"train_loss": 0.08168794959783554, "global_step": 157722, "epoch": 1772, "lr": 3.752508031535503e-05}
{"train_loss": 0.10119260847568512, "global_step": 157723, "epoch": 1772, "lr": 3.752451891363731e-05}
{"train_loss": 0.06895900517702103, "global_step": 157724, "epoch": 1772, "lr": 3.752395751359678e-05}
{"train_loss": 0.09292399883270264, "global_step": 157725, "epoch": 1772, "lr": 3.752339611523348e-05}
{"train_loss": 0.08392205834388733, "global_step": 157726, "epoch": 1772, "lr": 3.7522834718547535e-05}
{"train_loss": 0.0815797746181488, "global_step": 157727, "epoch": 1772, "lr": 3.7522273323538963e-05}
{"train_loss": 0.16906118392944336, "global_step": 157728, "epoch": 1772, "lr": 3.752171193020789e-05}
{"train_loss": 0.1346069872379303, "global_step": 157729, "epoch": 1772, "lr": 3.752115053855436e-05}
{"train_loss": 0.12629114091396332, "global_step": 157730, "epoch": 1772, "lr": 3.752058914857847e-05}
{"train_loss": 0.1483752429485321, "global_step": 157731, "epoch": 1772, "lr": 3.752002776028029e-05}
{"train_loss": 0.0856621265411377, "global_step": 157732, "epoch": 1772, "lr": 3.751946637365987e-05}
{"train_loss": 0.10552491247653961, "global_step": 157733, "epoch": 1772, "lr": 3.751890498871732e-05}
{"train_loss": 0.08143404126167297, "global_step": 157734, "epoch": 1772, "lr": 3.751834360545269e-05}
{"train_loss": 0.04293197765946388, "global_step": 157735, "epoch": 1772, "lr": 3.7517782223866084e-05}
{"train_loss": 0.06082671508193016, "global_step": 157736, "epoch": 1772, "lr": 3.7517220843957535e-05}
{"train_loss": 0.1491112858057022, "global_step": 157737, "epoch": 1772, "lr": 3.751665946572717e-05}
{"train_loss": 0.10748464614152908, "global_step": 157738, "epoch": 1772, "lr": 3.751609808917501e-05}
{"train_loss": 0.08557765185832977, "global_step": 157739, "epoch": 1772, "lr": 3.7515536714301186e-05}
{"train_loss": 0.09215323626995087, "global_step": 157740, "epoch": 1772, "lr": 3.7514975341105725e-05}
{"train_loss": 0.114698126912117, "global_step": 157741, "epoch": 1772, "lr": 3.751441396958874e-05}
{"train_loss": 0.09787137806415558, "global_step": 157742, "epoch": 1772, "lr": 3.751385259975027e-05}
{"train_loss": 0.06334972381591797, "global_step": 157743, "epoch": 1772, "lr": 3.751329123159044e-05}
{"train_loss": 0.039749402552843094, "global_step": 157744, "epoch": 1772, "lr": 3.7512729865109264e-05}
{"train_loss": 0.12580430507659912, "global_step": 157745, "epoch": 1772, "lr": 3.751216850030688e-05}
{"train_loss": 0.07637352496385574, "global_step": 157746, "epoch": 1772, "lr": 3.7511607137183295e-05}
{"train_loss": 0.08512182533740997, "global_step": 157747, "epoch": 1772, "lr": 3.751104577573865e-05}
{"train_loss": 0.0837671309709549, "global_step": 157748, "epoch": 1772, "lr": 3.751048441597298e-05}
{"train_loss": 0.13153547048568726, "global_step": 157749, "epoch": 1772, "lr": 3.750992305788638e-05}
{"train_loss": 0.07239179313182831, "global_step": 157750, "epoch": 1772, "lr": 3.750936170147893e-05}
{"train_loss": 0.13842543959617615, "global_step": 157751, "epoch": 1772, "lr": 3.750880034675067e-05}
{"train_loss": 0.0960528701543808, "global_step": 157752, "epoch": 1772, "lr": 3.750823899370172e-05}
{"train_loss": 0.0695282369852066, "global_step": 157753, "epoch": 1772, "lr": 3.750767764233212e-05}
{"train_loss": 0.08081799745559692, "global_step": 157754, "epoch": 1772, "lr": 3.750711629264198e-05}
{"train_loss": 0.06275604665279388, "global_step": 157755, "epoch": 1772, "lr": 3.750655494463133e-05}
{"train_loss": 0.10590099543333054, "global_step": 157756, "epoch": 1772, "lr": 3.750599359830029e-05}
{"train_loss": 0.09139247983694077, "global_step": 157757, "epoch": 1772, "lr": 3.7505432253648904e-05}
{"train_loss": 0.09445037692785263, "global_step": 157758, "epoch": 1772, "lr": 3.750487091067728e-05}
{"train_loss": 0.09072622656822205, "global_step": 157759, "epoch": 1772, "lr": 3.750430956938545e-05}
{"train_loss": 0.1348842978477478, "global_step": 157760, "epoch": 1772, "lr": 3.750374822977353e-05}
{"train_loss": 0.036139488220214844, "global_step": 157761, "epoch": 1772, "lr": 3.750318689184157e-05}
{"train_loss": 0.11369241774082184, "global_step": 157762, "epoch": 1772, "lr": 3.750262555558966e-05}
{"train_loss": 0.12370643019676208, "global_step": 157763, "epoch": 1772, "lr": 3.750206422101786e-05}
{"train_loss": 0.08477433770895004, "global_step": 157764, "epoch": 1772, "lr": 3.750150288812626e-05}
{"train_loss": 0.11971785873174667, "global_step": 157765, "epoch": 1772, "lr": 3.750094155691494e-05}
{"train_loss": 0.08507183194160461, "global_step": 157766, "epoch": 1772, "lr": 3.750038022738396e-05}
{"train_loss": 0.11104536056518555, "global_step": 157767, "epoch": 1772, "lr": 3.74998188995334e-05}
{"train_loss": 0.0836382806301117, "global_step": 157768, "epoch": 1772, "lr": 3.7499257573363335e-05}
{"train_loss": 0.08460386842489243, "global_step": 157769, "epoch": 1772, "lr": 3.749869624887384e-05}
{"train_loss": 0.07866504043340683, "global_step": 157770, "epoch": 1772, "lr": 3.7498134926064996e-05}
{"train_loss": 0.07676815986633301, "global_step": 157771, "epoch": 1772, "lr": 3.7497573604936895e-05}
{"train_loss": 0.10571186244487762, "global_step": 157772, "epoch": 1772, "lr": 3.7497012285489565e-05}
{"train_loss": 0.07883695513010025, "global_step": 157773, "epoch": 1772, "lr": 3.749645096772313e-05}
{"train_loss": 0.054094791412353516, "global_step": 157774, "epoch": 1772, "lr": 3.7495889651637616e-05}
{"train_loss": 0.12923723459243774, "global_step": 157775, "epoch": 1772, "lr": 3.749532833723316e-05}
{"train_loss": 0.12417151778936386, "global_step": 157776, "epoch": 1772, "lr": 3.749476702450978e-05}
{"train_loss": 0.05636025220155716, "global_step": 157777, "epoch": 1772, "lr": 3.74942057134676e-05}
{"train_loss": 0.1631438285112381, "global_step": 157778, "epoch": 1772, "lr": 3.749364440410664e-05}
{"train_loss": 0.08756891638040543, "global_step": 157779, "epoch": 1772, "lr": 3.749308309642704e-05}
{"train_loss": 0.08321081101894379, "global_step": 157780, "epoch": 1772, "lr": 3.7492521790428817e-05}
{"train_loss": 0.0955357700586319, "global_step": 157781, "epoch": 1772, "lr": 3.749196048611209e-05}
{"train_loss": 0.058170441538095474, "global_step": 157782, "epoch": 1772, "lr": 3.74913991834769e-05}
{"train_loss": 0.1534610390663147, "global_step": 157783, "epoch": 1772, "lr": 3.749083788252334e-05}
{"train_loss": 0.12654684484004974, "global_step": 157784, "epoch": 1772, "lr": 3.74902765832515e-05}
{"train_loss": 0.0682377889752388, "global_step": 157785, "epoch": 1772, "lr": 3.748971528566143e-05}
{"train_loss": 0.12995487451553345, "global_step": 157786, "epoch": 1772, "lr": 3.748915398975322e-05}
{"train_loss": 0.14351023733615875, "global_step": 157787, "epoch": 1772, "lr": 3.748859269552694e-05}
{"train_loss": 0.14292459189891815, "global_step": 157788, "epoch": 1772, "lr": 3.748803140298266e-05}
{"train_loss": 0.08651788532733917, "global_step": 157789, "epoch": 1772, "lr": 3.748747011212047e-05}
{"train_loss": 0.0851326435804367, "global_step": 157790, "epoch": 1772, "lr": 3.748690882294044e-05}
{"train_loss": 0.07686840742826462, "global_step": 157791, "epoch": 1772, "lr": 3.748634753544263e-05}
{"train_loss": 0.11369463056325912, "global_step": 157792, "epoch": 1772, "lr": 3.748578624962714e-05}
{"train_loss": 0.07233394682407379, "global_step": 157793, "epoch": 1772, "lr": 3.748522496549401e-05}
{"train_loss": 0.10174567252397537, "global_step": 157794, "epoch": 1772, "lr": 3.748466368304337e-05}
{"train_loss": 0.09536189585924149, "global_step": 157795, "epoch": 1772, "lr": 3.748410240227525e-05}
{"train_loss": 0.09820761966906237, "global_step": 157796, "epoch": 1772, "lr": 3.748354112318975e-05, "val_loss": 6.550612926483154}
{"train_loss": 0.07952868938446045, "global_step": 157797, "epoch": 1773, "lr": 3.748297984578692e-05}
{"train_loss": 0.08214663714170456, "global_step": 157798, "epoch": 1773, "lr": 3.7482418570066856e-05}
{"train_loss": 0.15475647151470184, "global_step": 157799, "epoch": 1773, "lr": 3.748185729602962e-05}
{"train_loss": 0.037634365260601044, "global_step": 157800, "epoch": 1773, "lr": 3.7481296023675315e-05}
{"train_loss": 0.07082971930503845, "global_step": 157801, "epoch": 1773, "lr": 3.748073475300398e-05}
{"train_loss": 0.11886569857597351, "global_step": 157802, "epoch": 1773, "lr": 3.74801734840157e-05}
{"train_loss": 0.10746797919273376, "global_step": 157803, "epoch": 1773, "lr": 3.747961221671059e-05}
{"train_loss": 0.057586319744586945, "global_step": 157804, "epoch": 1773, "lr": 3.747905095108867e-05}
{"train_loss": 0.0915011391043663, "global_step": 157805, "epoch": 1773, "lr": 3.747848968715005e-05}
{"train_loss": 0.12249959260225296, "global_step": 157806, "epoch": 1773, "lr": 3.7477928424894785e-05}
{"train_loss": 0.11097171157598495, "global_step": 157807, "epoch": 1773, "lr": 3.747736716432296e-05}
{"train_loss": 0.11856532841920853, "global_step": 157808, "epoch": 1773, "lr": 3.747680590543466e-05}
{"train_loss": 0.06741634756326675, "global_step": 157809, "epoch": 1773, "lr": 3.7476244648229957e-05}
{"train_loss": 0.07038996368646622, "global_step": 157810, "epoch": 1773, "lr": 3.74756833927089e-05}
{"train_loss": 0.18732018768787384, "global_step": 157811, "epoch": 1773, "lr": 3.74751221388716e-05}
{"train_loss": 0.06845082342624664, "global_step": 157812, "epoch": 1773, "lr": 3.74745608867181e-05}
{"train_loss": 0.05342110991477966, "global_step": 157813, "epoch": 1773, "lr": 3.747399963624851e-05}
{"train_loss": 0.05819392204284668, "global_step": 157814, "epoch": 1773, "lr": 3.747343838746288e-05}
{"train_loss": 0.10705547779798508, "global_step": 157815, "epoch": 1773, "lr": 3.74728771403613e-05}
{"train_loss": 0.1479349583387375, "global_step": 157816, "epoch": 1773, "lr": 3.7472315894943826e-05}
{"train_loss": 0.08931075781583786, "global_step": 157817, "epoch": 1773, "lr": 3.747175465121057e-05}
{"train_loss": 0.07257801294326782, "global_step": 157818, "epoch": 1773, "lr": 3.747119340916155e-05}
{"train_loss": 0.09335092455148697, "global_step": 157819, "epoch": 1773, "lr": 3.747063216879688e-05}
{"train_loss": 0.10835547000169754, "global_step": 157820, "epoch": 1773, "lr": 3.7470070930116664e-05}
{"train_loss": 0.12209796905517578, "global_step": 157821, "epoch": 1773, "lr": 3.746950969312091e-05}
{"train_loss": 0.06106959283351898, "global_step": 157822, "epoch": 1773, "lr": 3.7468948457809755e-05}
{"train_loss": 0.10661815851926804, "global_step": 157823, "epoch": 1773, "lr": 3.7468387224183234e-05}
{"train_loss": 0.06977522373199463, "global_step": 157824, "epoch": 1773, "lr": 3.7467825992241434e-05}
{"train_loss": 0.08893013745546341, "global_step": 157825, "epoch": 1773, "lr": 3.746726476198443e-05}
{"train_loss": 0.045270152390003204, "global_step": 157826, "epoch": 1773, "lr": 3.74667035334123e-05}
{"train_loss": 0.11922624707221985, "global_step": 157827, "epoch": 1773, "lr": 3.746614230652513e-05}
{"train_loss": 0.12380771338939667, "global_step": 157828, "epoch": 1773, "lr": 3.746558108132299e-05}
{"train_loss": 0.09774859994649887, "global_step": 157829, "epoch": 1773, "lr": 3.7465019857805926e-05}
{"train_loss": 0.09947513788938522, "global_step": 157830, "epoch": 1773, "lr": 3.746445863597406e-05}
{"train_loss": 0.10518146306276321, "global_step": 157831, "epoch": 1773, "lr": 3.746389741582742e-05}
{"train_loss": 0.08266787230968475, "global_step": 157832, "epoch": 1773, "lr": 3.7463336197366127e-05}
{"train_loss": 0.07010989636182785, "global_step": 157833, "epoch": 1773, "lr": 3.746277498059022e-05}
{"train_loss": 0.06495247781276703, "global_step": 157834, "epoch": 1773, "lr": 3.746221376549981e-05}
{"train_loss": 0.16268625855445862, "global_step": 157835, "epoch": 1773, "lr": 3.7461652552094926e-05}
{"train_loss": 0.12694773077964783, "global_step": 157836, "epoch": 1773, "lr": 3.74610913403757e-05}
{"train_loss": 0.10874862223863602, "global_step": 157837, "epoch": 1773, "lr": 3.7460530130342155e-05}
{"train_loss": 0.10061965137720108, "global_step": 157838, "epoch": 1773, "lr": 3.7459968921994384e-05}
{"train_loss": 0.1303914338350296, "global_step": 157839, "epoch": 1773, "lr": 3.7459407715332485e-05}
{"train_loss": 0.14681464433670044, "global_step": 157840, "epoch": 1773, "lr": 3.74588465103565e-05}
{"train_loss": 0.09859943389892578, "global_step": 157841, "epoch": 1773, "lr": 3.7458285307066546e-05}
{"train_loss": 0.11601004004478455, "global_step": 157842, "epoch": 1773, "lr": 3.7457724105462654e-05}
{"train_loss": 0.103884257376194, "global_step": 157843, "epoch": 1773, "lr": 3.745716290554492e-05}
{"train_loss": 0.07613949477672577, "global_step": 157844, "epoch": 1773, "lr": 3.745660170731342e-05}
{"train_loss": 0.0795590952038765, "global_step": 157845, "epoch": 1773, "lr": 3.745604051076822e-05}
{"train_loss": 0.04321414232254028, "global_step": 157846, "epoch": 1773, "lr": 3.745547931590941e-05}
{"train_loss": 0.06568239629268646, "global_step": 157847, "epoch": 1773, "lr": 3.745491812273707e-05}
{"train_loss": 0.09608215093612671, "global_step": 157848, "epoch": 1773, "lr": 3.7454356931251236e-05}
{"train_loss": 0.11894594877958298, "global_step": 157849, "epoch": 1773, "lr": 3.7453795741452035e-05}
{"train_loss": 0.06192857027053833, "global_step": 157850, "epoch": 1773, "lr": 3.7453234553339494e-05}
{"train_loss": 0.12600551545619965, "global_step": 157851, "epoch": 1773, "lr": 3.7452673366913735e-05}
{"train_loss": 0.1387351006269455, "global_step": 157852, "epoch": 1773, "lr": 3.745211218217479e-05}
{"train_loss": 0.09032496064901352, "global_step": 157853, "epoch": 1773, "lr": 3.7451550999122784e-05}
{"train_loss": 0.08455697447061539, "global_step": 157854, "epoch": 1773, "lr": 3.7450989817757735e-05}
{"train_loss": 0.04407855495810509, "global_step": 157855, "epoch": 1773, "lr": 3.745042863807975e-05}
{"train_loss": 0.08488224446773529, "global_step": 157856, "epoch": 1773, "lr": 3.744986746008892e-05}
{"train_loss": 0.09505461901426315, "global_step": 157857, "epoch": 1773, "lr": 3.744930628378529e-05}
{"train_loss": 0.07171975076198578, "global_step": 157858, "epoch": 1773, "lr": 3.7448745109168956e-05}
{"train_loss": 0.09623393416404724, "global_step": 157859, "epoch": 1773, "lr": 3.7448183936239976e-05}
{"train_loss": 0.06220052018761635, "global_step": 157860, "epoch": 1773, "lr": 3.744762276499845e-05}
{"train_loss": 0.15195158123970032, "global_step": 157861, "epoch": 1773, "lr": 3.7447061595444425e-05}
{"train_loss": 0.0718412697315216, "global_step": 157862, "epoch": 1773, "lr": 3.7446500427578e-05}
{"train_loss": 0.05112307518720627, "global_step": 157863, "epoch": 1773, "lr": 3.744593926139923e-05}
{"train_loss": 0.14636464416980743, "global_step": 157864, "epoch": 1773, "lr": 3.74453780969082e-05}
{"train_loss": 0.06018734723329544, "global_step": 157865, "epoch": 1773, "lr": 3.7444816934104995e-05}
{"train_loss": 0.10582873225212097, "global_step": 157866, "epoch": 1773, "lr": 3.7444255772989684e-05}
{"train_loss": 0.06286761164665222, "global_step": 157867, "epoch": 1773, "lr": 3.744369461356232e-05}
{"train_loss": 0.05220219865441322, "global_step": 157868, "epoch": 1773, "lr": 3.744313345582302e-05}
{"train_loss": 0.13930486142635345, "global_step": 157869, "epoch": 1773, "lr": 3.7442572299771814e-05}
{"train_loss": 0.08605285733938217, "global_step": 157870, "epoch": 1773, "lr": 3.7442011145408826e-05}
{"train_loss": 0.05740552023053169, "global_step": 157871, "epoch": 1773, "lr": 3.7441449992734084e-05}
{"train_loss": 0.09140065312385559, "global_step": 157872, "epoch": 1773, "lr": 3.744088884174771e-05}
{"train_loss": 0.103786900639534, "global_step": 157873, "epoch": 1773, "lr": 3.744032769244973e-05}
{"train_loss": 0.11312112212181091, "global_step": 157874, "epoch": 1773, "lr": 3.743976654484025e-05}
{"train_loss": 0.06080411374568939, "global_step": 157875, "epoch": 1773, "lr": 3.7439205398919354e-05}
{"train_loss": 0.07379070669412613, "global_step": 157876, "epoch": 1773, "lr": 3.7438644254687084e-05}
{"train_loss": 0.11360689997673035, "global_step": 157877, "epoch": 1773, "lr": 3.743808311214356e-05}
{"train_loss": 0.11826583743095398, "global_step": 157878, "epoch": 1773, "lr": 3.743752197128881e-05}
{"train_loss": 0.07526431977748871, "global_step": 157879, "epoch": 1773, "lr": 3.7436960832122956e-05}
{"train_loss": 0.13196611404418945, "global_step": 157880, "epoch": 1773, "lr": 3.743639969464604e-05}
{"train_loss": 0.1196211725473404, "global_step": 157881, "epoch": 1773, "lr": 3.743583855885814e-05}
{"train_loss": 0.09850528836250305, "global_step": 157882, "epoch": 1773, "lr": 3.743527742475934e-05}
{"train_loss": 0.09971509873867035, "global_step": 157883, "epoch": 1773, "lr": 3.7434716292349715e-05}
{"train_loss": 0.1590128391981125, "global_step": 157884, "epoch": 1773, "lr": 3.743415516162934e-05}
{"train_loss": 0.09517165758971417, "global_step": 157885, "epoch": 1773, "lr": 3.74335940325983e-05, "val_loss": 6.622310161590576}
{"train_loss": 0.1491483598947525, "global_step": 157886, "epoch": 1774, "lr": 3.7433032905256637e-05}
{"train_loss": 0.06844833493232727, "global_step": 157887, "epoch": 1774, "lr": 3.7432471779604475e-05}
{"train_loss": 0.10181919485330582, "global_step": 157888, "epoch": 1774, "lr": 3.7431910655641845e-05}
{"train_loss": 0.07208056002855301, "global_step": 157889, "epoch": 1774, "lr": 3.743134953336886e-05}
{"train_loss": 0.09000325947999954, "global_step": 157890, "epoch": 1774, "lr": 3.743078841278555e-05}
{"train_loss": 0.11858919262886047, "global_step": 157891, "epoch": 1774, "lr": 3.743022729389202e-05}
{"train_loss": 0.1275988072156906, "global_step": 157892, "epoch": 1774, "lr": 3.742966617668837e-05}
{"train_loss": 0.09716552495956421, "global_step": 157893, "epoch": 1774, "lr": 3.742910506117462e-05}
{"train_loss": 0.08600397408008575, "global_step": 157894, "epoch": 1774, "lr": 3.7428543947350895e-05}
{"train_loss": 0.11537551879882812, "global_step": 157895, "epoch": 1774, "lr": 3.742798283521723e-05}
{"train_loss": 0.07867046445608139, "global_step": 157896, "epoch": 1774, "lr": 3.7427421724773734e-05}
{"train_loss": 0.07609283179044724, "global_step": 157897, "epoch": 1774, "lr": 3.742686061602046e-05}
{"train_loss": 0.09100255370140076, "global_step": 157898, "epoch": 1774, "lr": 3.742629950895751e-05}
{"train_loss": 0.16097623109817505, "global_step": 157899, "epoch": 1774, "lr": 3.742573840358491e-05}
{"train_loss": 0.1076543852686882, "global_step": 157900, "epoch": 1774, "lr": 3.742517729990279e-05}
{"train_loss": 0.08830159157514572, "global_step": 157901, "epoch": 1774, "lr": 3.74246161979112e-05}
{"train_loss": 0.1012205109000206, "global_step": 157902, "epoch": 1774, "lr": 3.742405509761021e-05}
{"train_loss": 0.13451029360294342, "global_step": 157903, "epoch": 1774, "lr": 3.74234939989999e-05}
{"train_loss": 0.04724550247192383, "global_step": 157904, "epoch": 1774, "lr": 3.7422932902080356e-05}
{"train_loss": 0.09311602264642715, "global_step": 157905, "epoch": 1774, "lr": 3.742237180685164e-05}
{"train_loss": 0.07936662435531616, "global_step": 157906, "epoch": 1774, "lr": 3.7421810713313843e-05}
{"train_loss": 0.13083279132843018, "global_step": 157907, "epoch": 1774, "lr": 3.742124962146701e-05}
{"train_loss": 0.06400617212057114, "global_step": 157908, "epoch": 1774, "lr": 3.7420688531311255e-05}
{"train_loss": 0.10080038756132126, "global_step": 157909, "epoch": 1774, "lr": 3.742012744284662e-05}
{"train_loss": 0.06976905465126038, "global_step": 157910, "epoch": 1774, "lr": 3.7419566356073187e-05}
{"train_loss": 0.048329632729291916, "global_step": 157911, "epoch": 1774, "lr": 3.741900527099107e-05}
{"train_loss": 0.06716126203536987, "global_step": 157912, "epoch": 1774, "lr": 3.741844418760029e-05}
{"train_loss": 0.10207680612802505, "global_step": 157913, "epoch": 1774, "lr": 3.7417883105900965e-05}
{"train_loss": 0.10486806929111481, "global_step": 157914, "epoch": 1774, "lr": 3.741732202589314e-05}
{"train_loss": 0.08607561886310577, "global_step": 157915, "epoch": 1774, "lr": 3.741676094757691e-05}
{"train_loss": 0.0695752277970314, "global_step": 157916, "epoch": 1774, "lr": 3.741619987095233e-05}
{"train_loss": 0.1065046638250351, "global_step": 157917, "epoch": 1774, "lr": 3.741563879601951e-05}
{"train_loss": 0.12598761916160583, "global_step": 157918, "epoch": 1774, "lr": 3.741507772277848e-05}
{"train_loss": 0.1115972101688385, "global_step": 157919, "epoch": 1774, "lr": 3.741451665122937e-05}
{"train_loss": 0.1423279047012329, "global_step": 157920, "epoch": 1774, "lr": 3.7413955581372205e-05}
{"train_loss": 0.12011278420686722, "global_step": 157921, "epoch": 1774, "lr": 3.7413394513207085e-05}
{"train_loss": 0.08046475052833557, "global_step": 157922, "epoch": 1774, "lr": 3.741283344673407e-05}
{"train_loss": 0.09959765523672104, "global_step": 157923, "epoch": 1774, "lr": 3.741227238195326e-05}
{"train_loss": 0.07739394158124924, "global_step": 157924, "epoch": 1774, "lr": 3.741171131886471e-05}
{"train_loss": 0.14702871441841125, "global_step": 157925, "epoch": 1774, "lr": 3.7411150257468514e-05}
{"train_loss": 0.10280423611402512, "global_step": 157926, "epoch": 1774, "lr": 3.741058919776471e-05}
{"train_loss": 0.061729397624731064, "global_step": 157927, "epoch": 1774, "lr": 3.741002813975343e-05}
{"train_loss": 0.08320444077253342, "global_step": 157928, "epoch": 1774, "lr": 3.740946708343469e-05}
{"train_loss": 0.08098649233579636, "global_step": 157929, "epoch": 1774, "lr": 3.74089060288086e-05}
{"train_loss": 0.10626846551895142, "global_step": 157930, "epoch": 1774, "lr": 3.7408344975875237e-05}
{"train_loss": 0.07479264587163925, "global_step": 157931, "epoch": 1774, "lr": 3.7407783924634655e-05}
{"train_loss": 0.08625688403844833, "global_step": 157932, "epoch": 1774, "lr": 3.7407222875086967e-05}
{"train_loss": 0.09460761398077011, "global_step": 157933, "epoch": 1774, "lr": 3.74066618272322e-05}
{"train_loss": 0.07779845595359802, "global_step": 157934, "epoch": 1774, "lr": 3.7406100781070474e-05}
{"train_loss": 0.0658770278096199, "global_step": 157935, "epoch": 1774, "lr": 3.7405539736601826e-05}
{"train_loss": 0.049144845455884933, "global_step": 157936, "epoch": 1774, "lr": 3.740497869382636e-05}
{"train_loss": 0.09546862542629242, "global_step": 157937, "epoch": 1774, "lr": 3.740441765274414e-05}
{"train_loss": 0.0751282274723053, "global_step": 157938, "epoch": 1774, "lr": 3.740385661335525e-05}
{"train_loss": 0.06896329671144485, "global_step": 157939, "epoch": 1774, "lr": 3.740329557565975e-05}
{"train_loss": 0.09690829366445541, "global_step": 157940, "epoch": 1774, "lr": 3.740273453965772e-05}
{"train_loss": 0.09144952148199081, "global_step": 157941, "epoch": 1774, "lr": 3.740217350534925e-05}
{"train_loss": 0.14362217485904694, "global_step": 157942, "epoch": 1774, "lr": 3.740161247273439e-05}
{"train_loss": 0.12452460080385208, "global_step": 157943, "epoch": 1774, "lr": 3.740105144181324e-05}
{"train_loss": 0.1242295578122139, "global_step": 157944, "epoch": 1774, "lr": 3.740049041258587e-05}
{"train_loss": 0.11383272707462311, "global_step": 157945, "epoch": 1774, "lr": 3.7399929385052335e-05}
{"train_loss": 0.0649024024605751, "global_step": 157946, "epoch": 1774, "lr": 3.739936835921272e-05}
{"train_loss": 0.07856927067041397, "global_step": 157947, "epoch": 1774, "lr": 3.739880733506713e-05}
{"train_loss": 0.0909276008605957, "global_step": 157948, "epoch": 1774, "lr": 3.739824631261559e-05}
{"train_loss": 0.07469023764133453, "global_step": 157949, "epoch": 1774, "lr": 3.739768529185823e-05}
{"train_loss": 0.11128753423690796, "global_step": 157950, "epoch": 1774, "lr": 3.739712427279508e-05}
{"train_loss": 0.06957349926233292, "global_step": 157951, "epoch": 1774, "lr": 3.739656325542625e-05}
{"train_loss": 0.06858828663825989, "global_step": 157952, "epoch": 1774, "lr": 3.739600223975177e-05}
{"train_loss": 0.07216712832450867, "global_step": 157953, "epoch": 1774, "lr": 3.739544122577178e-05}
{"train_loss": 0.09648879617452621, "global_step": 157954, "epoch": 1774, "lr": 3.739488021348628e-05}
{"train_loss": 0.10456902533769608, "global_step": 157955, "epoch": 1774, "lr": 3.739431920289542e-05}
{"train_loss": 0.0769210159778595, "global_step": 157956, "epoch": 1774, "lr": 3.7393758193999215e-05}
{"train_loss": 0.12135546654462814, "global_step": 157957, "epoch": 1774, "lr": 3.7393197186797794e-05}
{"train_loss": 0.08172585070133209, "global_step": 157958, "epoch": 1774, "lr": 3.739263618129118e-05}
{"train_loss": 0.11772102117538452, "global_step": 157959, "epoch": 1774, "lr": 3.739207517747948e-05}
{"train_loss": 0.09653790295124054, "global_step": 157960, "epoch": 1774, "lr": 3.739151417536276e-05}
{"train_loss": 0.10629522055387497, "global_step": 157961, "epoch": 1774, "lr": 3.7390953174941096e-05}
{"train_loss": 0.15105409920215607, "global_step": 157962, "epoch": 1774, "lr": 3.7390392176214564e-05}
{"train_loss": 0.07205255329608917, "global_step": 157963, "epoch": 1774, "lr": 3.738983117918326e-05}
{"train_loss": 0.08420609682798386, "global_step": 157964, "epoch": 1774, "lr": 3.73892701838472e-05}
{"train_loss": 0.06859350949525833, "global_step": 157965, "epoch": 1774, "lr": 3.7388709190206515e-05}
{"train_loss": 0.11451597511768341, "global_step": 157966, "epoch": 1774, "lr": 3.738814819826128e-05}
{"train_loss": 0.09314430505037308, "global_step": 157967, "epoch": 1774, "lr": 3.738758720801153e-05}
{"train_loss": 0.13195864856243134, "global_step": 157968, "epoch": 1774, "lr": 3.738702621945739e-05}
{"train_loss": 0.07927683740854263, "global_step": 157969, "epoch": 1774, "lr": 3.7386465232598896e-05}
{"train_loss": 0.1210300400853157, "global_step": 157970, "epoch": 1774, "lr": 3.7385904247436154e-05}
{"train_loss": 0.03433263674378395, "global_step": 157971, "epoch": 1774, "lr": 3.73853432639692e-05}
{"train_loss": 0.10600313544273376, "global_step": 157972, "epoch": 1774, "lr": 3.738478228219815e-05}
{"train_loss": 0.13862401247024536, "global_step": 157973, "epoch": 1774, "lr": 3.738422130212304e-05}
{"train_loss": 0.09554067873552943, "global_step": 157974, "epoch": 1774, "lr": 3.7383660323744005e-05, "val_loss": 6.61049222946167}
{"train_loss": 0.06883811950683594, "global_step": 157975, "epoch": 1775, "lr": 3.738309934706105e-05}
{"train_loss": 0.07118134945631027, "global_step": 157976, "epoch": 1775, "lr": 3.738253837207431e-05}
{"train_loss": 0.09299401193857193, "global_step": 157977, "epoch": 1775, "lr": 3.7381977398783816e-05}
{"train_loss": 0.13961823284626007, "global_step": 157978, "epoch": 1775, "lr": 3.738141642718967e-05}
{"train_loss": 0.08660534769296646, "global_step": 157979, "epoch": 1775, "lr": 3.738085545729193e-05}
{"train_loss": 0.09222102910280228, "global_step": 157980, "epoch": 1775, "lr": 3.738029448909068e-05}
{"train_loss": 0.15429231524467468, "global_step": 157981, "epoch": 1775, "lr": 3.737973352258601e-05}
{"train_loss": 0.09728623181581497, "global_step": 157982, "epoch": 1775, "lr": 3.737917255777796e-05}
{"train_loss": 0.048949792981147766, "global_step": 157983, "epoch": 1775, "lr": 3.737861159466665e-05}
{"train_loss": 0.06863848119974136, "global_step": 157984, "epoch": 1775, "lr": 3.73780506332521e-05}
{"train_loss": 0.08239442110061646, "global_step": 157985, "epoch": 1775, "lr": 3.737748967353445e-05}
{"train_loss": 0.09774502366781235, "global_step": 157986, "epoch": 1775, "lr": 3.737692871551372e-05}
{"train_loss": 0.09351976215839386, "global_step": 157987, "epoch": 1775, "lr": 3.7376367759190027e-05}
{"train_loss": 0.10006710886955261, "global_step": 157988, "epoch": 1775, "lr": 3.7375806804563405e-05}
{"train_loss": 0.09777908027172089, "global_step": 157989, "epoch": 1775, "lr": 3.737524585163398e-05}
{"train_loss": 0.09326006472110748, "global_step": 157990, "epoch": 1775, "lr": 3.7374684900401765e-05}
{"train_loss": 0.18723204731941223, "global_step": 157991, "epoch": 1775, "lr": 3.737412395086691e-05}
{"train_loss": 0.07392971217632294, "global_step": 157992, "epoch": 1775, "lr": 3.737356300302941e-05}
{"train_loss": 0.1508018672466278, "global_step": 157993, "epoch": 1775, "lr": 3.737300205688942e-05}
{"train_loss": 0.12710437178611755, "global_step": 157994, "epoch": 1775, "lr": 3.737244111244694e-05}
{"train_loss": 0.07971654832363129, "global_step": 157995, "epoch": 1775, "lr": 3.737188016970211e-05}
{"train_loss": 0.08321081101894379, "global_step": 157996, "epoch": 1775, "lr": 3.737131922865496e-05}
{"train_loss": 0.11135178059339523, "global_step": 157997, "epoch": 1775, "lr": 3.7370758289305584e-05}
{"train_loss": 0.07071800529956818, "global_step": 157998, "epoch": 1775, "lr": 3.737019735165406e-05}
{"train_loss": 0.06935808807611465, "global_step": 157999, "epoch": 1775, "lr": 3.736963641570046e-05}
{"train_loss": 0.11581908911466599, "global_step": 158000, "epoch": 1775, "lr": 3.7369075481444846e-05}
{"train_loss": 0.06484248489141464, "global_step": 158001, "epoch": 1775, "lr": 3.736851454888731e-05}
{"train_loss": 0.08416630327701569, "global_step": 158002, "epoch": 1775, "lr": 3.736795361802794e-05}
{"train_loss": 0.13536706566810608, "global_step": 158003, "epoch": 1775, "lr": 3.736739268886677e-05}
{"train_loss": 0.04389987885951996, "global_step": 158004, "epoch": 1775, "lr": 3.736683176140392e-05}
{"train_loss": 0.0898803099989891, "global_step": 158005, "epoch": 1775, "lr": 3.7366270835639425e-05}
{"train_loss": 0.21257053315639496, "global_step": 158006, "epoch": 1775, "lr": 3.736570991157341e-05}
{"train_loss": 0.03710097447037697, "global_step": 158007, "epoch": 1775, "lr": 3.736514898920589e-05}
{"train_loss": 0.10972949117422104, "global_step": 158008, "epoch": 1775, "lr": 3.7364588068537e-05}
{"train_loss": 0.09981467574834824, "global_step": 158009, "epoch": 1775, "lr": 3.7364027149566764e-05}
{"train_loss": 0.07749857008457184, "global_step": 158010, "epoch": 1775, "lr": 3.7363466232295305e-05}
{"train_loss": 0.08613757789134979, "global_step": 158011, "epoch": 1775, "lr": 3.736290531672265e-05}
{"train_loss": 0.11895307898521423, "global_step": 158012, "epoch": 1775, "lr": 3.736234440284892e-05}
{"train_loss": 0.11987419426441193, "global_step": 158013, "epoch": 1775, "lr": 3.7361783490674144e-05}
{"train_loss": 0.09594223648309708, "global_step": 158014, "epoch": 1775, "lr": 3.7361222580198444e-05}
{"train_loss": 0.08717681467533112, "global_step": 158015, "epoch": 1775, "lr": 3.736066167142185e-05}
{"train_loss": 0.05833518132567406, "global_step": 158016, "epoch": 1775, "lr": 3.736010076434449e-05}
{"train_loss": 0.08686909079551697, "global_step": 158017, "epoch": 1775, "lr": 3.7359539858966395e-05}
{"train_loss": 0.10103105753660202, "global_step": 158018, "epoch": 1775, "lr": 3.7358978955287647e-05}
{"train_loss": 0.12237303704023361, "global_step": 158019, "epoch": 1775, "lr": 3.735841805330835e-05}
{"train_loss": 0.04809291660785675, "global_step": 158020, "epoch": 1775, "lr": 3.735785715302854e-05}
{"train_loss": 0.07937660068273544, "global_step": 158021, "epoch": 1775, "lr": 3.7357296254448336e-05}
{"train_loss": 0.06282995641231537, "global_step": 158022, "epoch": 1775, "lr": 3.7356735357567766e-05}
{"train_loss": 0.08022832870483398, "global_step": 158023, "epoch": 1775, "lr": 3.7356174462386946e-05}
{"train_loss": 0.0970759466290474, "global_step": 158024, "epoch": 1775, "lr": 3.735561356890591e-05}
{"train_loss": 0.11644688993692398, "global_step": 158025, "epoch": 1775, "lr": 3.735505267712479e-05}
{"train_loss": 0.09967199712991714, "global_step": 158026, "epoch": 1775, "lr": 3.73544917870436e-05}
{"train_loss": 0.09263409674167633, "global_step": 158027, "epoch": 1775, "lr": 3.735393089866247e-05}
{"train_loss": 0.07051753997802734, "global_step": 158028, "epoch": 1775, "lr": 3.735337001198143e-05}
{"train_loss": 0.09013146907091141, "global_step": 158029, "epoch": 1775, "lr": 3.735280912700059e-05}
{"train_loss": 0.08933741599321365, "global_step": 158030, "epoch": 1775, "lr": 3.735224824371999e-05}
{"train_loss": 0.11988624930381775, "global_step": 158031, "epoch": 1775, "lr": 3.7351687362139756e-05}
{"train_loss": 0.11609809845685959, "global_step": 158032, "epoch": 1775, "lr": 3.73511264822599e-05}
{"train_loss": 0.096402607858181, "global_step": 158033, "epoch": 1775, "lr": 3.735056560408056e-05}
{"train_loss": 0.12056812644004822, "global_step": 158034, "epoch": 1775, "lr": 3.735000472760176e-05}
{"train_loss": 0.09741820394992828, "global_step": 158035, "epoch": 1775, "lr": 3.734944385282362e-05}
{"train_loss": 0.07931797951459885, "global_step": 158036, "epoch": 1775, "lr": 3.734888297974618e-05}
{"train_loss": 0.14847451448440552, "global_step": 158037, "epoch": 1775, "lr": 3.7348322108369514e-05}
{"train_loss": 0.048048894852399826, "global_step": 158038, "epoch": 1775, "lr": 3.734776123869374e-05}
{"train_loss": 0.10352212190628052, "global_step": 158039, "epoch": 1775, "lr": 3.7347200370718884e-05}
{"train_loss": 0.05418475717306137, "global_step": 158040, "epoch": 1775, "lr": 3.734663950444506e-05}
{"train_loss": 0.06649188697338104, "global_step": 158041, "epoch": 1775, "lr": 3.734607863987231e-05}
{"train_loss": 0.06662626564502716, "global_step": 158042, "epoch": 1775, "lr": 3.734551777700074e-05}
{"train_loss": 0.1840580850839615, "global_step": 158043, "epoch": 1775, "lr": 3.734495691583039e-05}
{"train_loss": 0.07195308059453964, "global_step": 158044, "epoch": 1775, "lr": 3.734439605636138e-05}
{"train_loss": 0.08177793771028519, "global_step": 158045, "epoch": 1775, "lr": 3.734383519859374e-05}
{"train_loss": 0.06318210065364838, "global_step": 158046, "epoch": 1775, "lr": 3.734327434252759e-05}
{"train_loss": 0.08468051254749298, "global_step": 158047, "epoch": 1775, "lr": 3.734271348816296e-05}
{"train_loss": 0.13604658842086792, "global_step": 158048, "epoch": 1775, "lr": 3.734215263549997e-05}
{"train_loss": 0.09898418933153152, "global_step": 158049, "epoch": 1775, "lr": 3.734159178453866e-05}
{"train_loss": 0.06496682018041611, "global_step": 158050, "epoch": 1775, "lr": 3.734103093527912e-05}
{"train_loss": 0.11285021901130676, "global_step": 158051, "epoch": 1775, "lr": 3.734047008772142e-05}
{"train_loss": 0.1204916313290596, "global_step": 158052, "epoch": 1775, "lr": 3.733990924186565e-05}
{"train_loss": 0.1417139768600464, "global_step": 158053, "epoch": 1775, "lr": 3.733934839771186e-05}
{"train_loss": 0.10910351574420929, "global_step": 158054, "epoch": 1775, "lr": 3.733878755526016e-05}
{"train_loss": 0.08638688176870346, "global_step": 158055, "epoch": 1775, "lr": 3.7338226714510596e-05}
{"train_loss": 0.08712726086378098, "global_step": 158056, "epoch": 1775, "lr": 3.7337665875463247e-05}
{"train_loss": 0.06760159134864807, "global_step": 158057, "epoch": 1775, "lr": 3.733710503811821e-05}
{"train_loss": 0.10732093453407288, "global_step": 158058, "epoch": 1775, "lr": 3.733654420247552e-05}
{"train_loss": 0.08625929802656174, "global_step": 158059, "epoch": 1775, "lr": 3.7335983368535304e-05}
{"train_loss": 0.07068295776844025, "global_step": 158060, "epoch": 1775, "lr": 3.7335422536297584e-05}
{"train_loss": 0.07893961668014526, "global_step": 158061, "epoch": 1775, "lr": 3.733486170576249e-05}
{"train_loss": 0.15435685217380524, "global_step": 158062, "epoch": 1775, "lr": 3.733430087693005e-05}
{"train_loss": 0.09555002691203289, "global_step": 158063, "epoch": 1775, "lr": 3.733374004980037e-05, "val_loss": 6.728046894073486, "train_action_mse_error": 14.622953414916992}
{"train_loss": 0.13166867196559906, "global_step": 158064, "epoch": 1776, "lr": 3.7333179224373505e-05}
{"train_loss": 0.10358666628599167, "global_step": 158065, "epoch": 1776, "lr": 3.733261840064956e-05}
{"train_loss": 0.08580721169710159, "global_step": 158066, "epoch": 1776, "lr": 3.733205757862855e-05}
{"train_loss": 0.09217210859060287, "global_step": 158067, "epoch": 1776, "lr": 3.733149675831063e-05}
{"train_loss": 0.10773059725761414, "global_step": 158068, "epoch": 1776, "lr": 3.733093593969581e-05}
{"train_loss": 0.08957560360431671, "global_step": 158069, "epoch": 1776, "lr": 3.733037512278421e-05}
{"train_loss": 0.1026228591799736, "global_step": 158070, "epoch": 1776, "lr": 3.732981430757587e-05}
{"train_loss": 0.11479096114635468, "global_step": 158071, "epoch": 1776, "lr": 3.73292534940709e-05}
{"train_loss": 0.058457277715206146, "global_step": 158072, "epoch": 1776, "lr": 3.732869268226934e-05}
{"train_loss": 0.2190665602684021, "global_step": 158073, "epoch": 1776, "lr": 3.732813187217129e-05}
{"train_loss": 0.0706920400261879, "global_step": 158074, "epoch": 1776, "lr": 3.7327571063776836e-05}
{"train_loss": 0.14044581353664398, "global_step": 158075, "epoch": 1776, "lr": 3.7327010257086e-05}
{"train_loss": 0.06483035534620285, "global_step": 158076, "epoch": 1776, "lr": 3.732644945209892e-05}
{"train_loss": 0.09113753587007523, "global_step": 158077, "epoch": 1776, "lr": 3.732588864881563e-05}
{"train_loss": 0.15291480720043182, "global_step": 158078, "epoch": 1776, "lr": 3.7325327847236236e-05}
{"train_loss": 0.09374627470970154, "global_step": 158079, "epoch": 1776, "lr": 3.732476704736078e-05}
{"train_loss": 0.08547966182231903, "global_step": 158080, "epoch": 1776, "lr": 3.732420624918937e-05}
{"train_loss": 0.07790767401456833, "global_step": 158081, "epoch": 1776, "lr": 3.732364545272205e-05}
{"train_loss": 0.07015619426965714, "global_step": 158082, "epoch": 1776, "lr": 3.732308465795893e-05}
{"train_loss": 0.07752152532339096, "global_step": 158083, "epoch": 1776, "lr": 3.732252386490004e-05}
{"train_loss": 0.13939987123012543, "global_step": 158084, "epoch": 1776, "lr": 3.7321963073545504e-05}
{"train_loss": 0.14195440709590912, "global_step": 158085, "epoch": 1776, "lr": 3.732140228389537e-05}
{"train_loss": 0.12747693061828613, "global_step": 158086, "epoch": 1776, "lr": 3.732084149594972e-05}
{"train_loss": 0.09283772110939026, "global_step": 158087, "epoch": 1776, "lr": 3.732028070970862e-05}
{"train_loss": 0.08968197554349899, "global_step": 158088, "epoch": 1776, "lr": 3.731971992517217e-05}
{"train_loss": 0.12698400020599365, "global_step": 158089, "epoch": 1776, "lr": 3.731915914234041e-05}
{"train_loss": 0.0814124122262001, "global_step": 158090, "epoch": 1776, "lr": 3.731859836121345e-05}
{"train_loss": 0.1559947431087494, "global_step": 158091, "epoch": 1776, "lr": 3.7318037581791334e-05}
{"train_loss": 0.1153116449713707, "global_step": 158092, "epoch": 1776, "lr": 3.731747680407417e-05}
{"train_loss": 0.08923619240522385, "global_step": 158093, "epoch": 1776, "lr": 3.731691602806202e-05}
{"train_loss": 0.09035730361938477, "global_step": 158094, "epoch": 1776, "lr": 3.7316355253754934e-05}
{"train_loss": 0.13055835664272308, "global_step": 158095, "epoch": 1776, "lr": 3.731579448115303e-05}
{"train_loss": 0.09574859589338303, "global_step": 158096, "epoch": 1776, "lr": 3.731523371025634e-05}
{"train_loss": 0.049566321074962616, "global_step": 158097, "epoch": 1776, "lr": 3.731467294106499e-05}
{"train_loss": 0.12177987396717072, "global_step": 158098, "epoch": 1776, "lr": 3.7314112173579e-05}
{"train_loss": 0.12458793818950653, "global_step": 158099, "epoch": 1776, "lr": 3.73135514077985e-05}
{"train_loss": 0.16652661561965942, "global_step": 158100, "epoch": 1776, "lr": 3.731299064372352e-05}
{"train_loss": 0.0661066323518753, "global_step": 158101, "epoch": 1776, "lr": 3.7312429881354165e-05}
{"train_loss": 0.09949919581413269, "global_step": 158102, "epoch": 1776, "lr": 3.731186912069048e-05}
{"train_loss": 0.17144927382469177, "global_step": 158103, "epoch": 1776, "lr": 3.731130836173259e-05}
{"train_loss": 0.06505084037780762, "global_step": 158104, "epoch": 1776, "lr": 3.731074760448051e-05}
{"train_loss": 0.0703423023223877, "global_step": 158105, "epoch": 1776, "lr": 3.731018684893437e-05}
{"train_loss": 0.13819590210914612, "global_step": 158106, "epoch": 1776, "lr": 3.73096260950942e-05}
{"train_loss": 0.11110862344503403, "global_step": 158107, "epoch": 1776, "lr": 3.730906534296011e-05}
{"train_loss": 0.1284770667552948, "global_step": 158108, "epoch": 1776, "lr": 3.730850459253215e-05}
{"train_loss": 0.12736386060714722, "global_step": 158109, "epoch": 1776, "lr": 3.7307943843810414e-05}
{"train_loss": 0.07907398045063019, "global_step": 158110, "epoch": 1776, "lr": 3.730738309679497e-05}
{"train_loss": 0.10363180190324783, "global_step": 158111, "epoch": 1776, "lr": 3.730682235148589e-05}
{"train_loss": 0.04329818859696388, "global_step": 158112, "epoch": 1776, "lr": 3.730626160788325e-05}
{"train_loss": 0.08897948265075684, "global_step": 158113, "epoch": 1776, "lr": 3.730570086598714e-05}
{"train_loss": 0.10978716611862183, "global_step": 158114, "epoch": 1776, "lr": 3.7305140125797626e-05}
{"train_loss": 0.03948844596743584, "global_step": 158115, "epoch": 1776, "lr": 3.7304579387314764e-05}
{"train_loss": 0.11417879909276962, "global_step": 158116, "epoch": 1776, "lr": 3.730401865053867e-05}
{"train_loss": 0.08912473917007446, "global_step": 158117, "epoch": 1776, "lr": 3.730345791546937e-05}
{"train_loss": 0.08193816244602203, "global_step": 158118, "epoch": 1776, "lr": 3.730289718210699e-05}
{"train_loss": 0.11875584721565247, "global_step": 158119, "epoch": 1776, "lr": 3.730233645045156e-05}
{"train_loss": 0.062433697283267975, "global_step": 158120, "epoch": 1776, "lr": 3.7301775720503194e-05}
{"train_loss": 0.15288832783699036, "global_step": 158121, "epoch": 1776, "lr": 3.730121499226192e-05}
{"train_loss": 0.059161942452192307, "global_step": 158122, "epoch": 1776, "lr": 3.730065426572788e-05}
{"train_loss": 0.11860102415084839, "global_step": 158123, "epoch": 1776, "lr": 3.7300093540901084e-05}
{"train_loss": 0.07264149188995361, "global_step": 158124, "epoch": 1776, "lr": 3.729953281778166e-05}
{"train_loss": 0.05494851991534233, "global_step": 158125, "epoch": 1776, "lr": 3.7298972096369635e-05}
{"train_loss": 0.10223943740129471, "global_step": 158126, "epoch": 1776, "lr": 3.729841137666513e-05}
{"train_loss": 0.11216738075017929, "global_step": 158127, "epoch": 1776, "lr": 3.729785065866817e-05}
{"train_loss": 0.1664978265762329, "global_step": 158128, "epoch": 1776, "lr": 3.7297289942378875e-05}
{"train_loss": 0.10989987105131149, "global_step": 158129, "epoch": 1776, "lr": 3.729672922779731e-05}
{"train_loss": 0.10654719173908234, "global_step": 158130, "epoch": 1776, "lr": 3.729616851492354e-05}
{"train_loss": 0.12221390008926392, "global_step": 158131, "epoch": 1776, "lr": 3.729560780375765e-05}
{"train_loss": 0.04641058295965195, "global_step": 158132, "epoch": 1776, "lr": 3.729504709429971e-05}
{"train_loss": 0.07613304257392883, "global_step": 158133, "epoch": 1776, "lr": 3.7294486386549796e-05}
{"train_loss": 0.09058442711830139, "global_step": 158134, "epoch": 1776, "lr": 3.729392568050797e-05}
{"train_loss": 0.11632175743579865, "global_step": 158135, "epoch": 1776, "lr": 3.729336497617435e-05}
{"train_loss": 0.0620286762714386, "global_step": 158136, "epoch": 1776, "lr": 3.7292804273548945e-05}
{"train_loss": 0.06008216366171837, "global_step": 158137, "epoch": 1776, "lr": 3.72922435726319e-05}
{"train_loss": 0.06398485600948334, "global_step": 158138, "epoch": 1776, "lr": 3.729168287342323e-05}
{"train_loss": 0.12312740832567215, "global_step": 158139, "epoch": 1776, "lr": 3.7291122175923064e-05}
{"train_loss": 0.09981806576251984, "global_step": 158140, "epoch": 1776, "lr": 3.7290561480131436e-05}
{"train_loss": 0.07345979660749435, "global_step": 158141, "epoch": 1776, "lr": 3.729000078604845e-05}
{"train_loss": 0.14575527608394623, "global_step": 158142, "epoch": 1776, "lr": 3.728944009367415e-05}
{"train_loss": 0.083944171667099, "global_step": 158143, "epoch": 1776, "lr": 3.7288879403008644e-05}
{"train_loss": 0.05951505899429321, "global_step": 158144, "epoch": 1776, "lr": 3.728831871405198e-05}
{"train_loss": 0.11820877343416214, "global_step": 158145, "epoch": 1776, "lr": 3.728775802680425e-05}
{"train_loss": 0.08454650640487671, "global_step": 158146, "epoch": 1776, "lr": 3.7287197341265545e-05}
{"train_loss": 0.11148740351200104, "global_step": 158147, "epoch": 1776, "lr": 3.7286636657435906e-05}
{"train_loss": 0.08070969581604004, "global_step": 158148, "epoch": 1776, "lr": 3.728607597531542e-05}
{"train_loss": 0.07795076072216034, "global_step": 158149, "epoch": 1776, "lr": 3.728551529490418e-05}
{"train_loss": 0.07113420963287354, "global_step": 158150, "epoch": 1776, "lr": 3.728495461620224e-05}
{"train_loss": 0.07838959991931915, "global_step": 158151, "epoch": 1776, "lr": 3.728439393920968e-05}
{"train_loss": 0.099852301891935, "global_step": 158152, "epoch": 1776, "lr": 3.728383326392659e-05, "val_loss": 6.647189140319824}
{"train_loss": 0.08134995400905609, "global_step": 158153, "epoch": 1777, "lr": 3.7283272590353015e-05}
{"train_loss": 0.11773216724395752, "global_step": 158154, "epoch": 1777, "lr": 3.7282711918489065e-05}
{"train_loss": 0.15657442808151245, "global_step": 158155, "epoch": 1777, "lr": 3.7282151248334786e-05}
{"train_loss": 0.13174296915531158, "global_step": 158156, "epoch": 1777, "lr": 3.728159057989028e-05}
{"train_loss": 0.1594342142343521, "global_step": 158157, "epoch": 1777, "lr": 3.7281029913155595e-05}
{"train_loss": 0.12141657620668411, "global_step": 158158, "epoch": 1777, "lr": 3.728046924813084e-05}
{"train_loss": 0.07996772229671478, "global_step": 158159, "epoch": 1777, "lr": 3.727990858481605e-05}
{"train_loss": 0.12320049107074738, "global_step": 158160, "epoch": 1777, "lr": 3.727934792321134e-05}
{"train_loss": 0.07081768661737442, "global_step": 158161, "epoch": 1777, "lr": 3.727878726331674e-05}
{"train_loss": 0.10502943396568298, "global_step": 158162, "epoch": 1777, "lr": 3.727822660513238e-05}
{"train_loss": 0.07456979155540466, "global_step": 158163, "epoch": 1777, "lr": 3.727766594865829e-05}
{"train_loss": 0.1319342851638794, "global_step": 158164, "epoch": 1777, "lr": 3.727710529389456e-05}
{"train_loss": 0.11883928626775742, "global_step": 158165, "epoch": 1777, "lr": 3.7276544640841284e-05}
{"train_loss": 0.10505707561969757, "global_step": 158166, "epoch": 1777, "lr": 3.7275983989498516e-05}
{"train_loss": 0.17096352577209473, "global_step": 158167, "epoch": 1777, "lr": 3.7275423339866334e-05}
{"train_loss": 0.061927780508995056, "global_step": 158168, "epoch": 1777, "lr": 3.727486269194482e-05}
{"train_loss": 0.13272836804389954, "global_step": 158169, "epoch": 1777, "lr": 3.727430204573404e-05}
{"train_loss": 0.07014711201190948, "global_step": 158170, "epoch": 1777, "lr": 3.727374140123408e-05}
{"train_loss": 0.055994968861341476, "global_step": 158171, "epoch": 1777, "lr": 3.7273180758445017e-05}
{"train_loss": 0.17692340910434723, "global_step": 158172, "epoch": 1777, "lr": 3.7272620117366905e-05}
{"train_loss": 0.14616043865680695, "global_step": 158173, "epoch": 1777, "lr": 3.727205947799985e-05}
{"train_loss": 0.105486661195755, "global_step": 158174, "epoch": 1777, "lr": 3.7271498840343886e-05}
{"train_loss": 0.17480885982513428, "global_step": 158175, "epoch": 1777, "lr": 3.727093820439914e-05}
{"train_loss": 0.15102918446063995, "global_step": 158176, "epoch": 1777, "lr": 3.7270377570165644e-05}
{"train_loss": 0.105289988219738, "global_step": 158177, "epoch": 1777, "lr": 3.726981693764351e-05}
{"train_loss": 0.11015330255031586, "global_step": 158178, "epoch": 1777, "lr": 3.7269256306832764e-05}
{"train_loss": 0.1416568011045456, "global_step": 158179, "epoch": 1777, "lr": 3.726869567773354e-05}
{"train_loss": 0.09487983584403992, "global_step": 158180, "epoch": 1777, "lr": 3.726813505034586e-05}
{"train_loss": 0.10020536184310913, "global_step": 158181, "epoch": 1777, "lr": 3.726757442466986e-05}
{"train_loss": 0.09796399623155594, "global_step": 158182, "epoch": 1777, "lr": 3.726701380070554e-05}
{"train_loss": 0.09661181271076202, "global_step": 158183, "epoch": 1777, "lr": 3.726645317845302e-05}
{"train_loss": 0.17704422771930695, "global_step": 158184, "epoch": 1777, "lr": 3.7265892557912396e-05}
{"train_loss": 0.12063203752040863, "global_step": 158185, "epoch": 1777, "lr": 3.7265331939083706e-05}
{"train_loss": 0.07797131687402725, "global_step": 158186, "epoch": 1777, "lr": 3.726477132196703e-05}
{"train_loss": 0.12156786024570465, "global_step": 158187, "epoch": 1777, "lr": 3.7264210706562456e-05}
{"train_loss": 0.08602388203144073, "global_step": 158188, "epoch": 1777, "lr": 3.726365009287005e-05}
{"train_loss": 0.10834754258394241, "global_step": 158189, "epoch": 1777, "lr": 3.7263089480889893e-05}
{"train_loss": 0.0689707100391388, "global_step": 158190, "epoch": 1777, "lr": 3.726252887062207e-05}
{"train_loss": 0.09736554324626923, "global_step": 158191, "epoch": 1777, "lr": 3.726196826206663e-05}
{"train_loss": 0.11650864779949188, "global_step": 158192, "epoch": 1777, "lr": 3.726140765522367e-05}
{"train_loss": 0.07057028263807297, "global_step": 158193, "epoch": 1777, "lr": 3.726084705009325e-05}
{"train_loss": 0.07843329012393951, "global_step": 158194, "epoch": 1777, "lr": 3.726028644667546e-05}
{"train_loss": 0.10305918753147125, "global_step": 158195, "epoch": 1777, "lr": 3.725972584497036e-05}
{"train_loss": 0.10578016191720963, "global_step": 158196, "epoch": 1777, "lr": 3.725916524497806e-05}
{"train_loss": 0.18647868931293488, "global_step": 158197, "epoch": 1777, "lr": 3.725860464669858e-05}
{"train_loss": 0.10405030846595764, "global_step": 158198, "epoch": 1777, "lr": 3.7258044050132047e-05}
{"train_loss": 0.14130009710788727, "global_step": 158199, "epoch": 1777, "lr": 3.7257483455278495e-05}
{"train_loss": 0.12424655258655548, "global_step": 158200, "epoch": 1777, "lr": 3.7256922862138024e-05}
{"train_loss": 0.13536407053470612, "global_step": 158201, "epoch": 1777, "lr": 3.725636227071072e-05}
{"train_loss": 0.09158248454332352, "global_step": 158202, "epoch": 1777, "lr": 3.725580168099661e-05}
{"train_loss": 0.10518556833267212, "global_step": 158203, "epoch": 1777, "lr": 3.725524109299584e-05}
{"train_loss": 0.0746251568198204, "global_step": 158204, "epoch": 1777, "lr": 3.725468050670844e-05}
{"train_loss": 0.08393099159002304, "global_step": 158205, "epoch": 1777, "lr": 3.7254119922134485e-05}
{"train_loss": 0.05955565720796585, "global_step": 158206, "epoch": 1777, "lr": 3.725355933927406e-05}
{"train_loss": 0.13511520624160767, "global_step": 158207, "epoch": 1777, "lr": 3.7252998758127236e-05}
{"train_loss": 0.1141556054353714, "global_step": 158208, "epoch": 1777, "lr": 3.7252438178694096e-05}
{"train_loss": 0.09754569828510284, "global_step": 158209, "epoch": 1777, "lr": 3.725187760097471e-05}
{"train_loss": 0.0996398776769638, "global_step": 158210, "epoch": 1777, "lr": 3.725131702496914e-05}
{"train_loss": 0.06463135033845901, "global_step": 158211, "epoch": 1777, "lr": 3.725075645067749e-05}
{"train_loss": 0.09045685827732086, "global_step": 158212, "epoch": 1777, "lr": 3.7250195878099816e-05}
{"train_loss": 0.13280704617500305, "global_step": 158213, "epoch": 1777, "lr": 3.72496353072362e-05}
{"train_loss": 0.15075579285621643, "global_step": 158214, "epoch": 1777, "lr": 3.724907473808671e-05}
{"train_loss": 0.11451182514429092, "global_step": 158215, "epoch": 1777, "lr": 3.7248514170651436e-05}
{"train_loss": 0.09058105200529099, "global_step": 158216, "epoch": 1777, "lr": 3.7247953604930424e-05}
{"train_loss": 0.0788659080862999, "global_step": 158217, "epoch": 1777, "lr": 3.724739304092379e-05}
{"train_loss": 0.06330417841672897, "global_step": 158218, "epoch": 1777, "lr": 3.724683247863157e-05}
{"train_loss": 0.09818964451551437, "global_step": 158219, "epoch": 1777, "lr": 3.724627191805386e-05}
{"train_loss": 0.05575897917151451, "global_step": 158220, "epoch": 1777, "lr": 3.724571135919075e-05}
{"train_loss": 0.07437054067850113, "global_step": 158221, "epoch": 1777, "lr": 3.724515080204228e-05}
{"train_loss": 0.06624921411275864, "global_step": 158222, "epoch": 1777, "lr": 3.7244590246608566e-05}
{"train_loss": 0.051208626478910446, "global_step": 158223, "epoch": 1777, "lr": 3.724402969288965e-05}
{"train_loss": 0.05241335928440094, "global_step": 158224, "epoch": 1777, "lr": 3.724346914088562e-05}
{"train_loss": 0.10680899024009705, "global_step": 158225, "epoch": 1777, "lr": 3.7242908590596545e-05}
{"train_loss": 0.10792221873998642, "global_step": 158226, "epoch": 1777, "lr": 3.724234804202251e-05}
{"train_loss": 0.07873524725437164, "global_step": 158227, "epoch": 1777, "lr": 3.7241787495163586e-05}
{"train_loss": 0.12259016931056976, "global_step": 158228, "epoch": 1777, "lr": 3.7241226950019844e-05}
{"train_loss": 0.09613925963640213, "global_step": 158229, "epoch": 1777, "lr": 3.724066640659137e-05}
{"train_loss": 0.10481680929660797, "global_step": 158230, "epoch": 1777, "lr": 3.724010586487824e-05}
{"train_loss": 0.10750319808721542, "global_step": 158231, "epoch": 1777, "lr": 3.72395453248805e-05}
{"train_loss": 0.08534171432256699, "global_step": 158232, "epoch": 1777, "lr": 3.7238984786598275e-05}
{"train_loss": 0.09732108563184738, "global_step": 158233, "epoch": 1777, "lr": 3.723842425003158e-05}
{"train_loss": 0.14750273525714874, "global_step": 158234, "epoch": 1777, "lr": 3.7237863715180556e-05}
{"train_loss": 0.09312330186367035, "global_step": 158235, "epoch": 1777, "lr": 3.723730318204522e-05}
{"train_loss": 0.11517661809921265, "global_step": 158236, "epoch": 1777, "lr": 3.7236742650625675e-05}
{"train_loss": 0.0451788604259491, "global_step": 158237, "epoch": 1777, "lr": 3.7236182120922015e-05}
{"train_loss": 0.07159103453159332, "global_step": 158238, "epoch": 1777, "lr": 3.723562159293427e-05}
{"train_loss": 0.07117856293916702, "global_step": 158239, "epoch": 1777, "lr": 3.7235061066662555e-05}
{"train_loss": 0.12394953519105911, "global_step": 158240, "epoch": 1777, "lr": 3.723450054210692e-05}
{"train_loss": 0.1041430625855253, "global_step": 158241, "epoch": 1777, "lr": 3.723394001926747e-05, "val_loss": 6.425482749938965}
{"train_loss": 0.14878034591674805, "global_step": 158242, "epoch": 1778, "lr": 3.723337949814424e-05}
{"train_loss": 0.07041531056165695, "global_step": 158243, "epoch": 1778, "lr": 3.723281897873735e-05}
{"train_loss": 0.07493461668491364, "global_step": 158244, "epoch": 1778, "lr": 3.7232258461046834e-05}
{"train_loss": 0.09659125655889511, "global_step": 158245, "epoch": 1778, "lr": 3.72316979450728e-05}
{"train_loss": 0.16095294058322906, "global_step": 158246, "epoch": 1778, "lr": 3.723113743081529e-05}
{"train_loss": 0.0665954202413559, "global_step": 158247, "epoch": 1778, "lr": 3.7230576918274405e-05}
{"train_loss": 0.1456528753042221, "global_step": 158248, "epoch": 1778, "lr": 3.723001640745022e-05}
{"train_loss": 0.09102459251880646, "global_step": 158249, "epoch": 1778, "lr": 3.7229455898342805e-05}
{"train_loss": 0.08591334521770477, "global_step": 158250, "epoch": 1778, "lr": 3.7228895390952216e-05}
{"train_loss": 0.06451677531003952, "global_step": 158251, "epoch": 1778, "lr": 3.722833488527857e-05}
{"train_loss": 0.10000486671924591, "global_step": 158252, "epoch": 1778, "lr": 3.722777438132189e-05}
{"train_loss": 0.14529432356357574, "global_step": 158253, "epoch": 1778, "lr": 3.722721387908231e-05}
{"train_loss": 0.14157243072986603, "global_step": 158254, "epoch": 1778, "lr": 3.722665337855985e-05}
{"train_loss": 0.16261178255081177, "global_step": 158255, "epoch": 1778, "lr": 3.722609287975461e-05}
{"train_loss": 0.12103967368602753, "global_step": 158256, "epoch": 1778, "lr": 3.7225532382666696e-05}
{"train_loss": 0.10383430868387222, "global_step": 158257, "epoch": 1778, "lr": 3.722497188729612e-05}
{"train_loss": 0.07677280902862549, "global_step": 158258, "epoch": 1778, "lr": 3.722441139364302e-05}
{"train_loss": 0.09462420642375946, "global_step": 158259, "epoch": 1778, "lr": 3.722385090170742e-05}
{"train_loss": 0.08503451198339462, "global_step": 158260, "epoch": 1778, "lr": 3.722329041148944e-05}
{"train_loss": 0.12877848744392395, "global_step": 158261, "epoch": 1778, "lr": 3.722272992298911e-05}
{"train_loss": 0.1453869342803955, "global_step": 158262, "epoch": 1778, "lr": 3.7222169436206544e-05}
{"train_loss": 0.049218714237213135, "global_step": 158263, "epoch": 1778, "lr": 3.72216089511418e-05}
{"train_loss": 0.13107964396476746, "global_step": 158264, "epoch": 1778, "lr": 3.7221048467794944e-05}
{"train_loss": 0.04813510179519653, "global_step": 158265, "epoch": 1778, "lr": 3.722048798616607e-05}
{"train_loss": 0.08330732583999634, "global_step": 158266, "epoch": 1778, "lr": 3.721992750625524e-05}
{"train_loss": 0.12163275480270386, "global_step": 158267, "epoch": 1778, "lr": 3.721936702806255e-05}
{"train_loss": 0.08814334869384766, "global_step": 158268, "epoch": 1778, "lr": 3.721880655158805e-05}
{"train_loss": 0.0982225313782692, "global_step": 158269, "epoch": 1778, "lr": 3.721824607683182e-05}
{"train_loss": 0.13164827227592468, "global_step": 158270, "epoch": 1778, "lr": 3.721768560379395e-05}
{"train_loss": 0.09211891144514084, "global_step": 158271, "epoch": 1778, "lr": 3.72171251324745e-05}
{"train_loss": 0.08435171097517014, "global_step": 158272, "epoch": 1778, "lr": 3.721656466287354e-05}
{"train_loss": 0.09858215600252151, "global_step": 158273, "epoch": 1778, "lr": 3.721600419499118e-05}
{"train_loss": 0.09180015325546265, "global_step": 158274, "epoch": 1778, "lr": 3.7215443728827453e-05}
{"train_loss": 0.08978860825300217, "global_step": 158275, "epoch": 1778, "lr": 3.721488326438248e-05}
{"train_loss": 0.10877971351146698, "global_step": 158276, "epoch": 1778, "lr": 3.7214322801656275e-05}
{"train_loss": 0.14280404150485992, "global_step": 158277, "epoch": 1778, "lr": 3.7213762340648986e-05}
{"train_loss": 0.09569184482097626, "global_step": 158278, "epoch": 1778, "lr": 3.721320188136062e-05}
{"train_loss": 0.11722707748413086, "global_step": 158279, "epoch": 1778, "lr": 3.721264142379131e-05}
{"train_loss": 0.12501315772533417, "global_step": 158280, "epoch": 1778, "lr": 3.7212080967941074e-05}
{"train_loss": 0.06771419197320938, "global_step": 158281, "epoch": 1778, "lr": 3.721152051381005e-05}
{"train_loss": 0.0685138925909996, "global_step": 158282, "epoch": 1778, "lr": 3.721096006139826e-05}
{"train_loss": 0.05139611288905144, "global_step": 158283, "epoch": 1778, "lr": 3.721039961070581e-05}
{"train_loss": 0.0685930997133255, "global_step": 158284, "epoch": 1778, "lr": 3.720983916173276e-05}
{"train_loss": 0.10396254062652588, "global_step": 158285, "epoch": 1778, "lr": 3.720927871447919e-05}
{"train_loss": 0.0859653502702713, "global_step": 158286, "epoch": 1778, "lr": 3.720871826894519e-05}
{"train_loss": 0.13189038634300232, "global_step": 158287, "epoch": 1778, "lr": 3.720815782513082e-05}
{"train_loss": 0.05419158935546875, "global_step": 158288, "epoch": 1778, "lr": 3.720759738303614e-05}
{"train_loss": 0.09888225793838501, "global_step": 158289, "epoch": 1778, "lr": 3.7207036942661256e-05}
{"train_loss": 0.1261204034090042, "global_step": 158290, "epoch": 1778, "lr": 3.720647650400622e-05}
{"train_loss": 0.060432519763708115, "global_step": 158291, "epoch": 1778, "lr": 3.720591606707111e-05}
{"train_loss": 0.09489427506923676, "global_step": 158292, "epoch": 1778, "lr": 3.720535563185603e-05}
{"train_loss": 0.19365651905536652, "global_step": 158293, "epoch": 1778, "lr": 3.7204795198361014e-05}
{"train_loss": 0.09988521784543991, "global_step": 158294, "epoch": 1778, "lr": 3.7204234766586174e-05}
{"train_loss": 0.047188855707645416, "global_step": 158295, "epoch": 1778, "lr": 3.7203674336531557e-05}
{"train_loss": 0.11918867379426956, "global_step": 158296, "epoch": 1778, "lr": 3.7203113908197276e-05}
{"train_loss": 0.09960363805294037, "global_step": 158297, "epoch": 1778, "lr": 3.720255348158335e-05}
{"train_loss": 0.06454874575138092, "global_step": 158298, "epoch": 1778, "lr": 3.72019930566899e-05}
{"train_loss": 0.09586908668279648, "global_step": 158299, "epoch": 1778, "lr": 3.7201432633516975e-05}
{"train_loss": 0.07497550547122955, "global_step": 158300, "epoch": 1778, "lr": 3.7200872212064675e-05}
{"train_loss": 0.12028595060110092, "global_step": 158301, "epoch": 1778, "lr": 3.720031179233305e-05}
{"train_loss": 0.1034926250576973, "global_step": 158302, "epoch": 1778, "lr": 3.71997513743222e-05}
{"train_loss": 0.07411330938339233, "global_step": 158303, "epoch": 1778, "lr": 3.719919095803218e-05}
{"train_loss": 0.11884471774101257, "global_step": 158304, "epoch": 1778, "lr": 3.719863054346306e-05}
{"train_loss": 0.05431830510497093, "global_step": 158305, "epoch": 1778, "lr": 3.719807013061495e-05}
{"train_loss": 0.10331632941961288, "global_step": 158306, "epoch": 1778, "lr": 3.71975097194879e-05}
{"train_loss": 0.04960177466273308, "global_step": 158307, "epoch": 1778, "lr": 3.7196949310081974e-05}
{"train_loss": 0.08417258411645889, "global_step": 158308, "epoch": 1778, "lr": 3.7196388902397286e-05}
{"train_loss": 0.06815104931592941, "global_step": 158309, "epoch": 1778, "lr": 3.719582849643386e-05}
{"train_loss": 0.12292281538248062, "global_step": 158310, "epoch": 1778, "lr": 3.719526809219179e-05}
{"train_loss": 0.07435332983732224, "global_step": 158311, "epoch": 1778, "lr": 3.71947076896712e-05}
{"train_loss": 0.05204971134662628, "global_step": 158312, "epoch": 1778, "lr": 3.7194147288872095e-05}
{"train_loss": 0.1545737087726593, "global_step": 158313, "epoch": 1778, "lr": 3.7193586889794603e-05}
{"train_loss": 0.10147969424724579, "global_step": 158314, "epoch": 1778, "lr": 3.719302649243875e-05}
{"train_loss": 0.1398012936115265, "global_step": 158315, "epoch": 1778, "lr": 3.7192466096804665e-05}
{"train_loss": 0.10859405249357224, "global_step": 158316, "epoch": 1778, "lr": 3.719190570289238e-05}
{"train_loss": 0.09105301648378372, "global_step": 158317, "epoch": 1778, "lr": 3.7191345310702006e-05}
{"train_loss": 0.09023960679769516, "global_step": 158318, "epoch": 1778, "lr": 3.719078492023358e-05}
{"train_loss": 0.11772098392248154, "global_step": 158319, "epoch": 1778, "lr": 3.719022453148721e-05}
{"train_loss": 0.13383539021015167, "global_step": 158320, "epoch": 1778, "lr": 3.7189664144462955e-05}
{"train_loss": 0.09565804898738861, "global_step": 158321, "epoch": 1778, "lr": 3.718910375916089e-05}
{"train_loss": 0.0939403548836708, "global_step": 158322, "epoch": 1778, "lr": 3.71885433755811e-05}
{"train_loss": 0.051092952489852905, "global_step": 158323, "epoch": 1778, "lr": 3.718798299372364e-05}
{"train_loss": 0.04918438568711281, "global_step": 158324, "epoch": 1778, "lr": 3.718742261358861e-05}
{"train_loss": 0.08183138072490692, "global_step": 158325, "epoch": 1778, "lr": 3.7186862235176086e-05}
{"train_loss": 0.0744224563241005, "global_step": 158326, "epoch": 1778, "lr": 3.71863018584861e-05}
{"train_loss": 0.10706017166376114, "global_step": 158327, "epoch": 1778, "lr": 3.7185741483518774e-05}
{"train_loss": 0.07571528106927872, "global_step": 158328, "epoch": 1778, "lr": 3.718518111027419e-05}
{"train_loss": 0.14119212329387665, "global_step": 158329, "epoch": 1778, "lr": 3.718462073875237e-05}
{"train_loss": 0.09784696050239412, "global_step": 158330, "epoch": 1778, "lr": 3.718406036895345e-05, "val_loss": 6.714236259460449}
{"train_loss": 0.06266411393880844, "global_step": 158331, "epoch": 1779, "lr": 3.718350000087745e-05}
{"train_loss": 0.12772221863269806, "global_step": 158332, "epoch": 1779, "lr": 3.71829396345245e-05}
{"train_loss": 0.07119744271039963, "global_step": 158333, "epoch": 1779, "lr": 3.718237926989462e-05}
{"train_loss": 0.11303921788930893, "global_step": 158334, "epoch": 1779, "lr": 3.7181818906987945e-05}
{"train_loss": 0.0976206585764885, "global_step": 158335, "epoch": 1779, "lr": 3.7181258545804485e-05}
{"train_loss": 0.08045125007629395, "global_step": 158336, "epoch": 1779, "lr": 3.718069818634438e-05}
{"train_loss": 0.10546311736106873, "global_step": 158337, "epoch": 1779, "lr": 3.7180137828607644e-05}
{"train_loss": 0.08077472448348999, "global_step": 158338, "epoch": 1779, "lr": 3.7179577472594406e-05}
{"train_loss": 0.11251062899827957, "global_step": 158339, "epoch": 1779, "lr": 3.717901711830471e-05}
{"train_loss": 0.0762854516506195, "global_step": 158340, "epoch": 1779, "lr": 3.717845676573863e-05}
{"train_loss": 0.10834413766860962, "global_step": 158341, "epoch": 1779, "lr": 3.7177896414896264e-05}
{"train_loss": 0.10208787024021149, "global_step": 158342, "epoch": 1779, "lr": 3.717733606577767e-05}
{"train_loss": 0.08257647603750229, "global_step": 158343, "epoch": 1779, "lr": 3.717677571838292e-05}
{"train_loss": 0.06731171905994415, "global_step": 158344, "epoch": 1779, "lr": 3.71762153727121e-05}
{"train_loss": 0.06580714136362076, "global_step": 158345, "epoch": 1779, "lr": 3.717565502876528e-05}
{"train_loss": 0.05642280727624893, "global_step": 158346, "epoch": 1779, "lr": 3.7175094686542524e-05}
{"train_loss": 0.0789821445941925, "global_step": 158347, "epoch": 1779, "lr": 3.717453434604394e-05}
{"train_loss": 0.0970408171415329, "global_step": 158348, "epoch": 1779, "lr": 3.717397400726957e-05}
{"train_loss": 0.06341741979122162, "global_step": 158349, "epoch": 1779, "lr": 3.717341367021952e-05}
{"train_loss": 0.0782085582613945, "global_step": 158350, "epoch": 1779, "lr": 3.717285333489383e-05}
{"train_loss": 0.06665878742933273, "global_step": 158351, "epoch": 1779, "lr": 3.717229300129261e-05}
{"train_loss": 0.0700775682926178, "global_step": 158352, "epoch": 1779, "lr": 3.71717326694159e-05}
{"train_loss": 0.0863899290561676, "global_step": 158353, "epoch": 1779, "lr": 3.717117233926382e-05}
{"train_loss": 0.06516857445240021, "global_step": 158354, "epoch": 1779, "lr": 3.717061201083639e-05}
{"train_loss": 0.043112095445394516, "global_step": 158355, "epoch": 1779, "lr": 3.7170051684133744e-05}
{"train_loss": 0.03931565210223198, "global_step": 158356, "epoch": 1779, "lr": 3.716949135915591e-05}
{"train_loss": 0.09203249961137772, "global_step": 158357, "epoch": 1779, "lr": 3.716893103590299e-05}
{"train_loss": 0.15188907086849213, "global_step": 158358, "epoch": 1779, "lr": 3.716837071437503e-05}
{"train_loss": 0.07581809908151627, "global_step": 158359, "epoch": 1779, "lr": 3.716781039457215e-05}
{"train_loss": 0.1602540761232376, "global_step": 158360, "epoch": 1779, "lr": 3.7167250076494396e-05}
{"train_loss": 0.10056471079587936, "global_step": 158361, "epoch": 1779, "lr": 3.716668976014185e-05}
{"train_loss": 0.07843460142612457, "global_step": 158362, "epoch": 1779, "lr": 3.7166129445514586e-05}
{"train_loss": 0.19166789948940277, "global_step": 158363, "epoch": 1779, "lr": 3.716556913261267e-05}
{"train_loss": 0.06296097487211227, "global_step": 158364, "epoch": 1779, "lr": 3.71650088214362e-05}
{"train_loss": 0.07736460864543915, "global_step": 158365, "epoch": 1779, "lr": 3.716444851198522e-05}
{"train_loss": 0.09017577022314072, "global_step": 158366, "epoch": 1779, "lr": 3.7163888204259846e-05}
{"train_loss": 0.08203058689832687, "global_step": 158367, "epoch": 1779, "lr": 3.71633278982601e-05}
{"train_loss": 0.13631413877010345, "global_step": 158368, "epoch": 1779, "lr": 3.716276759398612e-05}
{"train_loss": 0.09760846942663193, "global_step": 158369, "epoch": 1779, "lr": 3.716220729143793e-05}
{"train_loss": 0.05037274211645126, "global_step": 158370, "epoch": 1779, "lr": 3.716164699061564e-05}
{"train_loss": 0.08319257199764252, "global_step": 158371, "epoch": 1779, "lr": 3.716108669151929e-05}
{"train_loss": 0.12220126390457153, "global_step": 158372, "epoch": 1779, "lr": 3.7160526394148997e-05}
{"train_loss": 0.1279364377260208, "global_step": 158373, "epoch": 1779, "lr": 3.715996609850479e-05}
{"train_loss": 0.11259834468364716, "global_step": 158374, "epoch": 1779, "lr": 3.71594058045868e-05}
{"train_loss": 0.07310006022453308, "global_step": 158375, "epoch": 1779, "lr": 3.715884551239505e-05}
{"train_loss": 0.062035053968429565, "global_step": 158376, "epoch": 1779, "lr": 3.7158285221929644e-05}
{"train_loss": 0.05378919094800949, "global_step": 158377, "epoch": 1779, "lr": 3.715772493319064e-05}
{"train_loss": 0.12081713229417801, "global_step": 158378, "epoch": 1779, "lr": 3.7157164646178144e-05}
{"train_loss": 0.1280786544084549, "global_step": 158379, "epoch": 1779, "lr": 3.71566043608922e-05}
{"train_loss": 0.07151749730110168, "global_step": 158380, "epoch": 1779, "lr": 3.715604407733289e-05}
{"train_loss": 0.06287923455238342, "global_step": 158381, "epoch": 1779, "lr": 3.71554837955003e-05}
{"train_loss": 0.10109268873929977, "global_step": 158382, "epoch": 1779, "lr": 3.715492351539448e-05}
{"train_loss": 0.08434934169054031, "global_step": 158383, "epoch": 1779, "lr": 3.715436323701554e-05}
{"train_loss": 0.05799295753240585, "global_step": 158384, "epoch": 1779, "lr": 3.715380296036354e-05}
{"train_loss": 0.12602598965168, "global_step": 158385, "epoch": 1779, "lr": 3.715324268543855e-05}
{"train_loss": 0.06311000138521194, "global_step": 158386, "epoch": 1779, "lr": 3.715268241224065e-05}
{"train_loss": 0.16306355595588684, "global_step": 158387, "epoch": 1779, "lr": 3.715212214076993e-05}
{"train_loss": 0.13582272827625275, "global_step": 158388, "epoch": 1779, "lr": 3.715156187102642e-05}
{"train_loss": 0.1280459761619568, "global_step": 158389, "epoch": 1779, "lr": 3.715100160301026e-05}
{"train_loss": 0.054080989211797714, "global_step": 158390, "epoch": 1779, "lr": 3.715044133672146e-05}
{"train_loss": 0.05525975674390793, "global_step": 158391, "epoch": 1779, "lr": 3.7149881072160156e-05}
{"train_loss": 0.0700666606426239, "global_step": 158392, "epoch": 1779, "lr": 3.714932080932636e-05}
{"train_loss": 0.08000383526086807, "global_step": 158393, "epoch": 1779, "lr": 3.714876054822022e-05}
{"train_loss": 0.06340884417295456, "global_step": 158394, "epoch": 1779, "lr": 3.7148200288841736e-05}
{"train_loss": 0.13527671992778778, "global_step": 158395, "epoch": 1779, "lr": 3.714764003119105e-05}
{"train_loss": 0.10509441047906876, "global_step": 158396, "epoch": 1779, "lr": 3.714707977526819e-05}
{"train_loss": 0.1330406665802002, "global_step": 158397, "epoch": 1779, "lr": 3.7146519521073256e-05}
{"train_loss": 0.06905512511730194, "global_step": 158398, "epoch": 1779, "lr": 3.714595926860631e-05}
{"train_loss": 0.07520941644906998, "global_step": 158399, "epoch": 1779, "lr": 3.7145399017867424e-05}
{"train_loss": 0.10024239122867584, "global_step": 158400, "epoch": 1779, "lr": 3.7144838768856707e-05}
{"train_loss": 0.1075572744011879, "global_step": 158401, "epoch": 1779, "lr": 3.714427852157418e-05}
{"train_loss": 0.12773141264915466, "global_step": 158402, "epoch": 1779, "lr": 3.714371827601998e-05}
{"train_loss": 0.24668541550636292, "global_step": 158403, "epoch": 1779, "lr": 3.714315803219412e-05}
{"train_loss": 0.07108066976070404, "global_step": 158404, "epoch": 1779, "lr": 3.714259779009673e-05}
{"train_loss": 0.08677324652671814, "global_step": 158405, "epoch": 1779, "lr": 3.714203754972785e-05}
{"train_loss": 0.12811648845672607, "global_step": 158406, "epoch": 1779, "lr": 3.714147731108758e-05}
{"train_loss": 0.06647009402513504, "global_step": 158407, "epoch": 1779, "lr": 3.714091707417595e-05}
{"train_loss": 0.0891697108745575, "global_step": 158408, "epoch": 1779, "lr": 3.714035683899311e-05}
{"train_loss": 0.11463844031095505, "global_step": 158409, "epoch": 1779, "lr": 3.713979660553905e-05}
{"train_loss": 0.0812937542796135, "global_step": 158410, "epoch": 1779, "lr": 3.713923637381392e-05}
{"train_loss": 0.0429181344807148, "global_step": 158411, "epoch": 1779, "lr": 3.7138676143817736e-05}
{"train_loss": 0.05936465039849281, "global_step": 158412, "epoch": 1779, "lr": 3.7138115915550625e-05}
{"train_loss": 0.13080982863903046, "global_step": 158413, "epoch": 1779, "lr": 3.7137555689012614e-05}
{"train_loss": 0.1074693575501442, "global_step": 158414, "epoch": 1779, "lr": 3.713699546420383e-05}
{"train_loss": 0.07871724665164948, "global_step": 158415, "epoch": 1779, "lr": 3.713643524112429e-05}
{"train_loss": 0.06696749478578568, "global_step": 158416, "epoch": 1779, "lr": 3.713587501977412e-05}
{"train_loss": 0.07029589265584946, "global_step": 158417, "epoch": 1779, "lr": 3.7135314800153375e-05}
{"train_loss": 0.1071387529373169, "global_step": 158418, "epoch": 1779, "lr": 3.713475458226211e-05}
{"train_loss": 0.09262746116251089, "global_step": 158419, "epoch": 1779, "lr": 3.7134194366100445e-05, "val_loss": 6.579172611236572}
{"train_loss": 0.0795154869556427, "global_step": 158420, "epoch": 1780, "lr": 3.7133634151668404e-05}
{"train_loss": 0.09171433746814728, "global_step": 158421, "epoch": 1780, "lr": 3.713307393896612e-05}
{"train_loss": 0.08629560470581055, "global_step": 158422, "epoch": 1780, "lr": 3.713251372799361e-05}
{"train_loss": 0.12629112601280212, "global_step": 158423, "epoch": 1780, "lr": 3.7131953518751005e-05}
{"train_loss": 0.04975869506597519, "global_step": 158424, "epoch": 1780, "lr": 3.713139331123832e-05}
{"train_loss": 0.09741179645061493, "global_step": 158425, "epoch": 1780, "lr": 3.7130833105455694e-05}
{"train_loss": 0.15647603571414948, "global_step": 158426, "epoch": 1780, "lr": 3.713027290140314e-05}
{"train_loss": 0.104673370718956, "global_step": 158427, "epoch": 1780, "lr": 3.7129712699080785e-05}
{"train_loss": 0.11596496403217316, "global_step": 158428, "epoch": 1780, "lr": 3.712915249848867e-05}
{"train_loss": 0.08673014491796494, "global_step": 158429, "epoch": 1780, "lr": 3.712859229962691e-05}
{"train_loss": 0.08140856772661209, "global_step": 158430, "epoch": 1780, "lr": 3.712803210249552e-05}
{"train_loss": 0.061089884489774704, "global_step": 158431, "epoch": 1780, "lr": 3.712747190709464e-05}
{"train_loss": 0.14859524369239807, "global_step": 158432, "epoch": 1780, "lr": 3.712691171342428e-05}
{"train_loss": 0.058458756655454636, "global_step": 158433, "epoch": 1780, "lr": 3.712635152148458e-05}
{"train_loss": 0.09818216413259506, "global_step": 158434, "epoch": 1780, "lr": 3.712579133127556e-05}
{"train_loss": 0.081006720662117, "global_step": 158435, "epoch": 1780, "lr": 3.712523114279735e-05}
{"train_loss": 0.0944775938987732, "global_step": 158436, "epoch": 1780, "lr": 3.7124670956049976e-05}
{"train_loss": 0.12445098161697388, "global_step": 158437, "epoch": 1780, "lr": 3.7124110771033526e-05}
{"train_loss": 0.09676438570022583, "global_step": 158438, "epoch": 1780, "lr": 3.71235505877481e-05}
{"train_loss": 0.06992517411708832, "global_step": 158439, "epoch": 1780, "lr": 3.7122990406193745e-05}
{"train_loss": 0.09599891304969788, "global_step": 158440, "epoch": 1780, "lr": 3.7122430226370556e-05}
{"train_loss": 0.11863002926111221, "global_step": 158441, "epoch": 1780, "lr": 3.712187004827859e-05}
{"train_loss": 0.11395420134067535, "global_step": 158442, "epoch": 1780, "lr": 3.712130987191794e-05}
{"train_loss": 0.11628296226263046, "global_step": 158443, "epoch": 1780, "lr": 3.712074969728866e-05}
{"train_loss": 0.06304283440113068, "global_step": 158444, "epoch": 1780, "lr": 3.712018952439086e-05}
{"train_loss": 0.14749599993228912, "global_step": 158445, "epoch": 1780, "lr": 3.711962935322456e-05}
{"train_loss": 0.023284146562218666, "global_step": 158446, "epoch": 1780, "lr": 3.7119069183789904e-05}
{"train_loss": 0.08451720327138901, "global_step": 158447, "epoch": 1780, "lr": 3.711850901608691e-05}
{"train_loss": 0.13300682604312897, "global_step": 158448, "epoch": 1780, "lr": 3.711794885011569e-05}
{"train_loss": 0.06418311595916748, "global_step": 158449, "epoch": 1780, "lr": 3.711738868587629e-05}
{"train_loss": 0.07013864815235138, "global_step": 158450, "epoch": 1780, "lr": 3.711682852336882e-05}
{"train_loss": 0.1408795565366745, "global_step": 158451, "epoch": 1780, "lr": 3.711626836259331e-05}
{"train_loss": 0.10325372964143753, "global_step": 158452, "epoch": 1780, "lr": 3.711570820354988e-05}
{"train_loss": 0.11654773354530334, "global_step": 158453, "epoch": 1780, "lr": 3.711514804623857e-05}
{"train_loss": 0.08010273426771164, "global_step": 158454, "epoch": 1780, "lr": 3.711458789065948e-05}
{"train_loss": 0.11621619015932083, "global_step": 158455, "epoch": 1780, "lr": 3.711402773681269e-05}
{"train_loss": 0.06747278571128845, "global_step": 158456, "epoch": 1780, "lr": 3.711346758469824e-05}
{"train_loss": 0.1125023290514946, "global_step": 158457, "epoch": 1780, "lr": 3.7112907434316246e-05}
{"train_loss": 0.06614835560321808, "global_step": 158458, "epoch": 1780, "lr": 3.711234728566674e-05}
{"train_loss": 0.08445194363594055, "global_step": 158459, "epoch": 1780, "lr": 3.7111787138749855e-05}
{"train_loss": 0.07150933891534805, "global_step": 158460, "epoch": 1780, "lr": 3.7111226993565604e-05}
{"train_loss": 0.1017841100692749, "global_step": 158461, "epoch": 1780, "lr": 3.7110666850114105e-05}
{"train_loss": 0.12560565769672394, "global_step": 158462, "epoch": 1780, "lr": 3.711010670839541e-05}
{"train_loss": 0.09020106494426727, "global_step": 158463, "epoch": 1780, "lr": 3.7109546568409625e-05}
{"train_loss": 0.15800435841083527, "global_step": 158464, "epoch": 1780, "lr": 3.710898643015678e-05}
{"train_loss": 0.06613735109567642, "global_step": 158465, "epoch": 1780, "lr": 3.710842629363699e-05}
{"train_loss": 0.14778664708137512, "global_step": 158466, "epoch": 1780, "lr": 3.71078661588503e-05}
{"train_loss": 0.09378597140312195, "global_step": 158467, "epoch": 1780, "lr": 3.710730602579683e-05}
{"train_loss": 0.08557961881160736, "global_step": 158468, "epoch": 1780, "lr": 3.710674589447659e-05}
{"train_loss": 0.14564627408981323, "global_step": 158469, "epoch": 1780, "lr": 3.710618576488972e-05}
{"train_loss": 0.08061011880636215, "global_step": 158470, "epoch": 1780, "lr": 3.710562563703624e-05}
{"train_loss": 0.09204751998186111, "global_step": 158471, "epoch": 1780, "lr": 3.7105065510916277e-05}
{"train_loss": 0.12419457733631134, "global_step": 158472, "epoch": 1780, "lr": 3.7104505386529856e-05}
{"train_loss": 0.09268095344305038, "global_step": 158473, "epoch": 1780, "lr": 3.7103945263877094e-05}
{"train_loss": 0.09529642015695572, "global_step": 158474, "epoch": 1780, "lr": 3.710338514295804e-05}
{"train_loss": 0.09989053010940552, "global_step": 158475, "epoch": 1780, "lr": 3.710282502377278e-05}
{"train_loss": 0.06469842046499252, "global_step": 158476, "epoch": 1780, "lr": 3.7102264906321405e-05}
{"train_loss": 0.09807975590229034, "global_step": 158477, "epoch": 1780, "lr": 3.710170479060395e-05}
{"train_loss": 0.07128407061100006, "global_step": 158478, "epoch": 1780, "lr": 3.710114467662053e-05}
{"train_loss": 0.09451334923505783, "global_step": 158479, "epoch": 1780, "lr": 3.7100584564371186e-05}
{"train_loss": 0.08551178127527237, "global_step": 158480, "epoch": 1780, "lr": 3.710002445385604e-05}
{"train_loss": 0.11943328380584717, "global_step": 158481, "epoch": 1780, "lr": 3.7099464345075105e-05}
{"train_loss": 0.09583143889904022, "global_step": 158482, "epoch": 1780, "lr": 3.709890423802852e-05}
{"train_loss": 0.06820708513259888, "global_step": 158483, "epoch": 1780, "lr": 3.70983441327163e-05}
{"train_loss": 0.09752217680215836, "global_step": 158484, "epoch": 1780, "lr": 3.709778402913858e-05}
{"train_loss": 0.1347961127758026, "global_step": 158485, "epoch": 1780, "lr": 3.709722392729538e-05}
{"train_loss": 0.18140199780464172, "global_step": 158486, "epoch": 1780, "lr": 3.709666382718683e-05}
{"train_loss": 0.10323240607976913, "global_step": 158487, "epoch": 1780, "lr": 3.7096103728812945e-05}
{"train_loss": 0.10215259343385696, "global_step": 158488, "epoch": 1780, "lr": 3.709554363217386e-05}
{"train_loss": 0.08229023218154907, "global_step": 158489, "epoch": 1780, "lr": 3.7094983537269587e-05}
{"train_loss": 0.07914470881223679, "global_step": 158490, "epoch": 1780, "lr": 3.709442344410026e-05}
{"train_loss": 0.09800314903259277, "global_step": 158491, "epoch": 1780, "lr": 3.709386335266593e-05}
{"train_loss": 0.09557406604290009, "global_step": 158492, "epoch": 1780, "lr": 3.709330326296667e-05}
{"train_loss": 0.05811937898397446, "global_step": 158493, "epoch": 1780, "lr": 3.709274317500255e-05}
{"train_loss": 0.10181088000535965, "global_step": 158494, "epoch": 1780, "lr": 3.709218308877367e-05}
{"train_loss": 0.06077554449439049, "global_step": 158495, "epoch": 1780, "lr": 3.709162300428009e-05}
{"train_loss": 0.11857300996780396, "global_step": 158496, "epoch": 1780, "lr": 3.709106292152186e-05}
{"train_loss": 0.07526058703660965, "global_step": 158497, "epoch": 1780, "lr": 3.709050284049911e-05}
{"train_loss": 0.051002997905015945, "global_step": 158498, "epoch": 1780, "lr": 3.7089942761211854e-05}
{"train_loss": 0.10221908241510391, "global_step": 158499, "epoch": 1780, "lr": 3.708938268366022e-05}
{"train_loss": 0.09113291651010513, "global_step": 158500, "epoch": 1780, "lr": 3.708882260784424e-05}
{"train_loss": 0.07930003851652145, "global_step": 158501, "epoch": 1780, "lr": 3.708826253376404e-05}
{"train_loss": 0.058168355375528336, "global_step": 158502, "epoch": 1780, "lr": 3.708770246141964e-05}
{"train_loss": 0.09195727109909058, "global_step": 158503, "epoch": 1780, "lr": 3.708714239081116e-05}
{"train_loss": 0.057689789682626724, "global_step": 158504, "epoch": 1780, "lr": 3.708658232193863e-05}
{"train_loss": 0.15506282448768616, "global_step": 158505, "epoch": 1780, "lr": 3.708602225480218e-05}
{"train_loss": 0.08172480016946793, "global_step": 158506, "epoch": 1780, "lr": 3.708546218940183e-05}
{"train_loss": 0.0862714871764183, "global_step": 158507, "epoch": 1780, "lr": 3.708490212573771e-05}
{"train_loss": 0.09569281609624289, "global_step": 158508, "epoch": 1780, "lr": 3.7084342063809846e-05, "val_loss": 6.518562316894531, "train_action_mse_error": 21.297069549560547}
{"train_loss": 0.17068834602832794, "global_step": 158509, "epoch": 1781, "lr": 3.7083782003618345e-05}
{"train_loss": 0.09372380375862122, "global_step": 158510, "epoch": 1781, "lr": 3.7083221945163275e-05}
{"train_loss": 0.11699703335762024, "global_step": 158511, "epoch": 1781, "lr": 3.7082661888444696e-05}
{"train_loss": 0.10166336596012115, "global_step": 158512, "epoch": 1781, "lr": 3.70821018334627e-05}
{"train_loss": 0.13010701537132263, "global_step": 158513, "epoch": 1781, "lr": 3.708154178021737e-05}
{"train_loss": 0.09645204991102219, "global_step": 158514, "epoch": 1781, "lr": 3.708098172870877e-05}
{"train_loss": 0.0982794538140297, "global_step": 158515, "epoch": 1781, "lr": 3.708042167893695e-05}
{"train_loss": 0.1103803738951683, "global_step": 158516, "epoch": 1781, "lr": 3.707986163090203e-05}
{"train_loss": 0.07429048418998718, "global_step": 158517, "epoch": 1781, "lr": 3.707930158460405e-05}
{"train_loss": 0.08928616344928741, "global_step": 158518, "epoch": 1781, "lr": 3.707874154004313e-05}
{"train_loss": 0.09370098263025284, "global_step": 158519, "epoch": 1781, "lr": 3.7078181497219285e-05}
{"train_loss": 0.047199029475450516, "global_step": 158520, "epoch": 1781, "lr": 3.707762145613264e-05}
{"train_loss": 0.10014116019010544, "global_step": 158521, "epoch": 1781, "lr": 3.707706141678323e-05}
{"train_loss": 0.10019922256469727, "global_step": 158522, "epoch": 1781, "lr": 3.707650137917118e-05}
{"train_loss": 0.14054101705551147, "global_step": 158523, "epoch": 1781, "lr": 3.707594134329652e-05}
{"train_loss": 0.08225859701633453, "global_step": 158524, "epoch": 1781, "lr": 3.7075381309159344e-05}
{"train_loss": 0.10986718535423279, "global_step": 158525, "epoch": 1781, "lr": 3.707482127675972e-05}
{"train_loss": 0.1544197052717209, "global_step": 158526, "epoch": 1781, "lr": 3.7074261246097717e-05}
{"train_loss": 0.11764837056398392, "global_step": 158527, "epoch": 1781, "lr": 3.7073701217173454e-05}
{"train_loss": 0.07705038785934448, "global_step": 158528, "epoch": 1781, "lr": 3.707314118998696e-05}
{"train_loss": 0.1197965145111084, "global_step": 158529, "epoch": 1781, "lr": 3.707258116453832e-05}
{"train_loss": 0.08048982173204422, "global_step": 158530, "epoch": 1781, "lr": 3.707202114082762e-05}
{"train_loss": 0.07157229632139206, "global_step": 158531, "epoch": 1781, "lr": 3.707146111885493e-05}
{"train_loss": 0.08660998195409775, "global_step": 158532, "epoch": 1781, "lr": 3.7070901098620324e-05}
{"train_loss": 0.12669607996940613, "global_step": 158533, "epoch": 1781, "lr": 3.7070341080123884e-05}
{"train_loss": 0.08068845421075821, "global_step": 158534, "epoch": 1781, "lr": 3.706978106336566e-05}
{"train_loss": 0.06893801689147949, "global_step": 158535, "epoch": 1781, "lr": 3.706922104834577e-05}
{"train_loss": 0.0965656042098999, "global_step": 158536, "epoch": 1781, "lr": 3.706866103506424e-05}
{"train_loss": 0.0754203200340271, "global_step": 158537, "epoch": 1781, "lr": 3.70681010235212e-05}
{"train_loss": 0.06371892988681793, "global_step": 158538, "epoch": 1781, "lr": 3.7067541013716664e-05}
{"train_loss": 0.16102853417396545, "global_step": 158539, "epoch": 1781, "lr": 3.706698100565077e-05}
{"train_loss": 0.15941309928894043, "global_step": 158540, "epoch": 1781, "lr": 3.706642099932354e-05}
{"train_loss": 0.10038002580404282, "global_step": 158541, "epoch": 1781, "lr": 3.706586099473509e-05}
{"train_loss": 0.06556273996829987, "global_step": 158542, "epoch": 1781, "lr": 3.706530099188545e-05}
{"train_loss": 0.09855404496192932, "global_step": 158543, "epoch": 1781, "lr": 3.706474099077475e-05}
{"train_loss": 0.08165319263935089, "global_step": 158544, "epoch": 1781, "lr": 3.706418099140302e-05}
{"train_loss": 0.07297204434871674, "global_step": 158545, "epoch": 1781, "lr": 3.706362099377036e-05}
{"train_loss": 0.054372236132621765, "global_step": 158546, "epoch": 1781, "lr": 3.706306099787684e-05}
{"train_loss": 0.10352322459220886, "global_step": 158547, "epoch": 1781, "lr": 3.7062501003722536e-05}
{"train_loss": 0.07567434757947922, "global_step": 158548, "epoch": 1781, "lr": 3.7061941011307514e-05}
{"train_loss": 0.08115468174219131, "global_step": 158549, "epoch": 1781, "lr": 3.7061381020631865e-05}
{"train_loss": 0.059977587312459946, "global_step": 158550, "epoch": 1781, "lr": 3.706082103169565e-05}
{"train_loss": 0.10919317603111267, "global_step": 158551, "epoch": 1781, "lr": 3.706026104449895e-05}
{"train_loss": 0.13714756071567535, "global_step": 158552, "epoch": 1781, "lr": 3.705970105904185e-05}
{"train_loss": 0.09723694622516632, "global_step": 158553, "epoch": 1781, "lr": 3.70591410753244e-05}
{"train_loss": 0.07077964395284653, "global_step": 158554, "epoch": 1781, "lr": 3.705858109334671e-05}
{"train_loss": 0.07710198312997818, "global_step": 158555, "epoch": 1781, "lr": 3.705802111310881e-05}
{"train_loss": 0.10466521233320236, "global_step": 158556, "epoch": 1781, "lr": 3.7057461134610824e-05}
{"train_loss": 0.059528421610593796, "global_step": 158557, "epoch": 1781, "lr": 3.7056901157852786e-05}
{"train_loss": 0.07467970252037048, "global_step": 158558, "epoch": 1781, "lr": 3.705634118283481e-05}
{"train_loss": 0.15016885101795197, "global_step": 158559, "epoch": 1781, "lr": 3.7055781209556934e-05}
{"train_loss": 0.04534932225942612, "global_step": 158560, "epoch": 1781, "lr": 3.705522123801927e-05}
{"train_loss": 0.13370192050933838, "global_step": 158561, "epoch": 1781, "lr": 3.705466126822185e-05}
{"train_loss": 0.12430624663829803, "global_step": 158562, "epoch": 1781, "lr": 3.705410130016479e-05}
{"train_loss": 0.09946341812610626, "global_step": 158563, "epoch": 1781, "lr": 3.705354133384813e-05}
{"train_loss": 0.09239087253808975, "global_step": 158564, "epoch": 1781, "lr": 3.705298136927197e-05}
{"train_loss": 0.09503025561571121, "global_step": 158565, "epoch": 1781, "lr": 3.7052421406436395e-05}
{"train_loss": 0.08075904846191406, "global_step": 158566, "epoch": 1781, "lr": 3.705186144534145e-05}
{"train_loss": 0.07771019637584686, "global_step": 158567, "epoch": 1781, "lr": 3.705130148598723e-05}
{"train_loss": 0.05491857975721359, "global_step": 158568, "epoch": 1781, "lr": 3.70507415283738e-05}
{"train_loss": 0.07591260224580765, "global_step": 158569, "epoch": 1781, "lr": 3.705018157250125e-05}
{"train_loss": 0.07712184637784958, "global_step": 158570, "epoch": 1781, "lr": 3.7049621618369636e-05}
{"train_loss": 0.1387665867805481, "global_step": 158571, "epoch": 1781, "lr": 3.704906166597905e-05}
{"train_loss": 0.07643851637840271, "global_step": 158572, "epoch": 1781, "lr": 3.704850171532955e-05}
{"train_loss": 0.10256088525056839, "global_step": 158573, "epoch": 1781, "lr": 3.7047941766421235e-05}
{"train_loss": 0.09176143258810043, "global_step": 158574, "epoch": 1781, "lr": 3.7047381819254144e-05}
{"train_loss": 0.051649294793605804, "global_step": 158575, "epoch": 1781, "lr": 3.70468218738284e-05}
{"train_loss": 0.0724799782037735, "global_step": 158576, "epoch": 1781, "lr": 3.7046261930144035e-05}
{"train_loss": 0.11535592377185822, "global_step": 158577, "epoch": 1781, "lr": 3.704570198820115e-05}
{"train_loss": 0.05741623789072037, "global_step": 158578, "epoch": 1781, "lr": 3.70451420479998e-05}
{"train_loss": 0.11626800149679184, "global_step": 158579, "epoch": 1781, "lr": 3.7044582109540095e-05}
{"train_loss": 0.07410623878240585, "global_step": 158580, "epoch": 1781, "lr": 3.704402217282207e-05}
{"train_loss": 0.15415571630001068, "global_step": 158581, "epoch": 1781, "lr": 3.704346223784581e-05}
{"train_loss": 0.05057837814092636, "global_step": 158582, "epoch": 1781, "lr": 3.704290230461143e-05}
{"train_loss": 0.08756791800260544, "global_step": 158583, "epoch": 1781, "lr": 3.7042342373118946e-05}
{"train_loss": 0.07244580239057541, "global_step": 158584, "epoch": 1781, "lr": 3.704178244336848e-05}
{"train_loss": 0.08154532313346863, "global_step": 158585, "epoch": 1781, "lr": 3.704122251536006e-05}
{"train_loss": 0.13781963288784027, "global_step": 158586, "epoch": 1781, "lr": 3.7040662589093826e-05}
{"train_loss": 0.08327270299196243, "global_step": 158587, "epoch": 1781, "lr": 3.70401026645698e-05}
{"train_loss": 0.09940946102142334, "global_step": 158588, "epoch": 1781, "lr": 3.703954274178808e-05}
{"train_loss": 0.08322969824075699, "global_step": 158589, "epoch": 1781, "lr": 3.703898282074872e-05}
{"train_loss": 0.08828874677419662, "global_step": 158590, "epoch": 1781, "lr": 3.703842290145183e-05}
{"train_loss": 0.1586458832025528, "global_step": 158591, "epoch": 1781, "lr": 3.7037862983897455e-05}
{"train_loss": 0.07684545964002609, "global_step": 158592, "epoch": 1781, "lr": 3.7037303068085694e-05}
{"train_loss": 0.10997509956359863, "global_step": 158593, "epoch": 1781, "lr": 3.70367431540166e-05}
{"train_loss": 0.15187107026576996, "global_step": 158594, "epoch": 1781, "lr": 3.7036183241690257e-05}
{"train_loss": 0.11115490645170212, "global_step": 158595, "epoch": 1781, "lr": 3.7035623331106735e-05}
{"train_loss": 0.08086805045604706, "global_step": 158596, "epoch": 1781, "lr": 3.703506342226613e-05}
{"train_loss": 0.09564679542954048, "global_step": 158597, "epoch": 1781, "lr": 3.703450351516848e-05, "val_loss": 6.612674713134766}
{"train_loss": 0.10975033044815063, "global_step": 158598, "epoch": 1782, "lr": 3.703394360981391e-05}
{"train_loss": 0.1153034120798111, "global_step": 158599, "epoch": 1782, "lr": 3.7033383706202455e-05}
{"train_loss": 0.07324592769145966, "global_step": 158600, "epoch": 1782, "lr": 3.703282380433419e-05}
{"train_loss": 0.09803217649459839, "global_step": 158601, "epoch": 1782, "lr": 3.703226390420923e-05}
{"train_loss": 0.045245446264743805, "global_step": 158602, "epoch": 1782, "lr": 3.70317040058276e-05}
{"train_loss": 0.06210694462060928, "global_step": 158603, "epoch": 1782, "lr": 3.703114410918942e-05}
{"train_loss": 0.12950894236564636, "global_step": 158604, "epoch": 1782, "lr": 3.703058421429473e-05}
{"train_loss": 0.0684090182185173, "global_step": 158605, "epoch": 1782, "lr": 3.703002432114363e-05}
{"train_loss": 0.061585430055856705, "global_step": 158606, "epoch": 1782, "lr": 3.7029464429736186e-05}
{"train_loss": 0.09195192158222198, "global_step": 158607, "epoch": 1782, "lr": 3.702890454007246e-05}
{"train_loss": 0.141875222325325, "global_step": 158608, "epoch": 1782, "lr": 3.702834465215255e-05}
{"train_loss": 0.09226182103157043, "global_step": 158609, "epoch": 1782, "lr": 3.702778476597652e-05}
{"train_loss": 0.05345781147480011, "global_step": 158610, "epoch": 1782, "lr": 3.702722488154444e-05}
{"train_loss": 0.0811779797077179, "global_step": 158611, "epoch": 1782, "lr": 3.702666499885641e-05}
{"train_loss": 0.07061831653118134, "global_step": 158612, "epoch": 1782, "lr": 3.702610511791247e-05}
{"train_loss": 0.06927713751792908, "global_step": 158613, "epoch": 1782, "lr": 3.7025545238712726e-05}
{"train_loss": 0.12338030338287354, "global_step": 158614, "epoch": 1782, "lr": 3.702498536125721e-05}
{"train_loss": 0.09150715917348862, "global_step": 158615, "epoch": 1782, "lr": 3.702442548554606e-05}
{"train_loss": 0.1087619811296463, "global_step": 158616, "epoch": 1782, "lr": 3.702386561157929e-05}
{"train_loss": 0.09427723288536072, "global_step": 158617, "epoch": 1782, "lr": 3.7023305739357e-05}
{"train_loss": 0.11227718740701675, "global_step": 158618, "epoch": 1782, "lr": 3.70227458688793e-05}
{"train_loss": 0.048552531749010086, "global_step": 158619, "epoch": 1782, "lr": 3.7022186000146206e-05}
{"train_loss": 0.09292013943195343, "global_step": 158620, "epoch": 1782, "lr": 3.7021626133157844e-05}
{"train_loss": 0.03222312033176422, "global_step": 158621, "epoch": 1782, "lr": 3.702106626791424e-05}
{"train_loss": 0.06363587826490402, "global_step": 158622, "epoch": 1782, "lr": 3.702050640441552e-05}
{"train_loss": 0.10383736342191696, "global_step": 158623, "epoch": 1782, "lr": 3.701994654266172e-05}
{"train_loss": 0.0805722177028656, "global_step": 158624, "epoch": 1782, "lr": 3.7019386682652944e-05}
{"train_loss": 0.10077507048845291, "global_step": 158625, "epoch": 1782, "lr": 3.701882682438924e-05}
{"train_loss": 0.1503150463104248, "global_step": 158626, "epoch": 1782, "lr": 3.7018266967870706e-05}
{"train_loss": 0.12143091857433319, "global_step": 158627, "epoch": 1782, "lr": 3.70177071130974e-05}
{"train_loss": 0.10796701163053513, "global_step": 158628, "epoch": 1782, "lr": 3.701714726006941e-05}
{"train_loss": 0.11547347158193588, "global_step": 158629, "epoch": 1782, "lr": 3.70165874087868e-05}
{"train_loss": 0.1666707694530487, "global_step": 158630, "epoch": 1782, "lr": 3.701602755924967e-05}
{"train_loss": 0.08462892472743988, "global_step": 158631, "epoch": 1782, "lr": 3.7015467711458054e-05}
{"train_loss": 0.08132611215114594, "global_step": 158632, "epoch": 1782, "lr": 3.701490786541206e-05}
{"train_loss": 0.09914589673280716, "global_step": 158633, "epoch": 1782, "lr": 3.701434802111175e-05}
{"train_loss": 0.07029540091753006, "global_step": 158634, "epoch": 1782, "lr": 3.701378817855721e-05}
{"train_loss": 0.0587860606610775, "global_step": 158635, "epoch": 1782, "lr": 3.7013228337748496e-05}
{"train_loss": 0.06771277636289597, "global_step": 158636, "epoch": 1782, "lr": 3.7012668498685685e-05}
{"train_loss": 0.0817767083644867, "global_step": 158637, "epoch": 1782, "lr": 3.7012108661368895e-05}
{"train_loss": 0.06880106776952744, "global_step": 158638, "epoch": 1782, "lr": 3.701154882579814e-05}
{"train_loss": 0.05907810106873512, "global_step": 158639, "epoch": 1782, "lr": 3.701098899197355e-05}
{"train_loss": 0.10078112035989761, "global_step": 158640, "epoch": 1782, "lr": 3.701042915989514e-05}
{"train_loss": 0.06871474534273148, "global_step": 158641, "epoch": 1782, "lr": 3.700986932956306e-05}
{"train_loss": 0.06245678290724754, "global_step": 158642, "epoch": 1782, "lr": 3.7009309500977314e-05}
{"train_loss": 0.06662125140428543, "global_step": 158643, "epoch": 1782, "lr": 3.700874967413803e-05}
{"train_loss": 0.12107657641172409, "global_step": 158644, "epoch": 1782, "lr": 3.700818984904525e-05}
{"train_loss": 0.11085028946399689, "global_step": 158645, "epoch": 1782, "lr": 3.700763002569906e-05}
{"train_loss": 0.10061230510473251, "global_step": 158646, "epoch": 1782, "lr": 3.700707020409954e-05}
{"train_loss": 0.10478731989860535, "global_step": 158647, "epoch": 1782, "lr": 3.700651038424675e-05}
{"train_loss": 0.14351849257946014, "global_step": 158648, "epoch": 1782, "lr": 3.700595056614079e-05}
{"train_loss": 0.11602701991796494, "global_step": 158649, "epoch": 1782, "lr": 3.700539074978172e-05}
{"train_loss": 0.10517267882823944, "global_step": 158650, "epoch": 1782, "lr": 3.70048309351696e-05}
{"train_loss": 0.17738187313079834, "global_step": 158651, "epoch": 1782, "lr": 3.7004271122304546e-05}
{"train_loss": 0.10788297653198242, "global_step": 158652, "epoch": 1782, "lr": 3.700371131118659e-05}
{"train_loss": 0.14888258278369904, "global_step": 158653, "epoch": 1782, "lr": 3.7003151501815824e-05}
{"train_loss": 0.08446363359689713, "global_step": 158654, "epoch": 1782, "lr": 3.700259169419234e-05}
{"train_loss": 0.09533244371414185, "global_step": 158655, "epoch": 1782, "lr": 3.700203188831619e-05}
{"train_loss": 0.11666467040777206, "global_step": 158656, "epoch": 1782, "lr": 3.700147208418747e-05}
{"train_loss": 0.049213603138923645, "global_step": 158657, "epoch": 1782, "lr": 3.7000912281806235e-05}
{"train_loss": 0.07254020869731903, "global_step": 158658, "epoch": 1782, "lr": 3.700035248117258e-05}
{"train_loss": 0.06805771589279175, "global_step": 158659, "epoch": 1782, "lr": 3.699979268228655e-05}
{"train_loss": 0.11575707793235779, "global_step": 158660, "epoch": 1782, "lr": 3.699923288514826e-05}
{"train_loss": 0.16983847320079803, "global_step": 158661, "epoch": 1782, "lr": 3.699867308975774e-05}
{"train_loss": 0.0936986580491066, "global_step": 158662, "epoch": 1782, "lr": 3.6998113296115125e-05}
{"train_loss": 0.08067727088928223, "global_step": 158663, "epoch": 1782, "lr": 3.699755350422043e-05}
{"train_loss": 0.1467587947845459, "global_step": 158664, "epoch": 1782, "lr": 3.699699371407376e-05}
{"train_loss": 0.048486970365047455, "global_step": 158665, "epoch": 1782, "lr": 3.6996433925675187e-05}
{"train_loss": 0.14045968651771545, "global_step": 158666, "epoch": 1782, "lr": 3.6995874139024776e-05}
{"train_loss": 0.059169359505176544, "global_step": 158667, "epoch": 1782, "lr": 3.699531435412262e-05}
{"train_loss": 0.04676597937941551, "global_step": 158668, "epoch": 1782, "lr": 3.6994754570968804e-05}
{"train_loss": 0.08568179607391357, "global_step": 158669, "epoch": 1782, "lr": 3.699419478956335e-05}
{"train_loss": 0.07683372497558594, "global_step": 158670, "epoch": 1782, "lr": 3.699363500990639e-05}
{"train_loss": 0.07643112540245056, "global_step": 158671, "epoch": 1782, "lr": 3.699307523199796e-05}
{"train_loss": 0.12404212355613708, "global_step": 158672, "epoch": 1782, "lr": 3.6992515455838156e-05}
{"train_loss": 0.09652512520551682, "global_step": 158673, "epoch": 1782, "lr": 3.6991955681427074e-05}
{"train_loss": 0.15894140303134918, "global_step": 158674, "epoch": 1782, "lr": 3.699139590876473e-05}
{"train_loss": 0.14852386713027954, "global_step": 158675, "epoch": 1782, "lr": 3.699083613785126e-05}
{"train_loss": 0.14371943473815918, "global_step": 158676, "epoch": 1782, "lr": 3.69902763686867e-05}
{"train_loss": 0.13611754775047302, "global_step": 158677, "epoch": 1782, "lr": 3.698971660127115e-05}
{"train_loss": 0.03936523199081421, "global_step": 158678, "epoch": 1782, "lr": 3.698915683560466e-05}
{"train_loss": 0.09704598039388657, "global_step": 158679, "epoch": 1782, "lr": 3.698859707168734e-05}
{"train_loss": 0.05044376850128174, "global_step": 158680, "epoch": 1782, "lr": 3.698803730951922e-05}
{"train_loss": 0.12127187103033066, "global_step": 158681, "epoch": 1782, "lr": 3.6987477549100426e-05}
{"train_loss": 0.15278667211532593, "global_step": 158682, "epoch": 1782, "lr": 3.698691779043099e-05}
{"train_loss": 0.16123077273368835, "global_step": 158683, "epoch": 1782, "lr": 3.698635803351101e-05}
{"train_loss": 0.06692512333393097, "global_step": 158684, "epoch": 1782, "lr": 3.698579827834056e-05}
{"train_loss": 0.042301733046770096, "global_step": 158685, "epoch": 1782, "lr": 3.69852385249197e-05}
{"train_loss": 0.09568418265226182, "global_step": 158686, "epoch": 1782, "lr": 3.6984678773248515e-05, "val_loss": 6.613399982452393}
{"train_loss": 0.076788529753685, "global_step": 158687, "epoch": 1783, "lr": 3.6984119023327087e-05}
{"train_loss": 0.13721899688243866, "global_step": 158688, "epoch": 1783, "lr": 3.6983559275155486e-05}
{"train_loss": 0.14799761772155762, "global_step": 158689, "epoch": 1783, "lr": 3.69829995287338e-05}
{"train_loss": 0.07701750099658966, "global_step": 158690, "epoch": 1783, "lr": 3.6982439784062065e-05}
{"train_loss": 0.04454865679144859, "global_step": 158691, "epoch": 1783, "lr": 3.6981880041140374e-05}
{"train_loss": 0.07011236250400543, "global_step": 158692, "epoch": 1783, "lr": 3.698132029996885e-05}
{"train_loss": 0.07665051519870758, "global_step": 158693, "epoch": 1783, "lr": 3.698076056054749e-05}
{"train_loss": 0.09040380269289017, "global_step": 158694, "epoch": 1783, "lr": 3.6980200822876444e-05}
{"train_loss": 0.13558563590049744, "global_step": 158695, "epoch": 1783, "lr": 3.697964108695572e-05}
{"train_loss": 0.12688426673412323, "global_step": 158696, "epoch": 1783, "lr": 3.697908135278545e-05}
{"train_loss": 0.13057702779769897, "global_step": 158697, "epoch": 1783, "lr": 3.6978521620365666e-05}
{"train_loss": 0.09293264895677567, "global_step": 158698, "epoch": 1783, "lr": 3.6977961889696476e-05}
{"train_loss": 0.1453886479139328, "global_step": 158699, "epoch": 1783, "lr": 3.697740216077792e-05}
{"train_loss": 0.08125515282154083, "global_step": 158700, "epoch": 1783, "lr": 3.6976842433610114e-05}
{"train_loss": 0.10072137415409088, "global_step": 158701, "epoch": 1783, "lr": 3.69762827081931e-05}
{"train_loss": 0.12806251645088196, "global_step": 158702, "epoch": 1783, "lr": 3.697572298452698e-05}
{"train_loss": 0.08220765739679337, "global_step": 158703, "epoch": 1783, "lr": 3.69751632626118e-05}
{"train_loss": 0.08355486392974854, "global_step": 158704, "epoch": 1783, "lr": 3.697460354244766e-05}
{"train_loss": 0.08925414085388184, "global_step": 158705, "epoch": 1783, "lr": 3.697404382403462e-05}
{"train_loss": 0.04888151213526726, "global_step": 158706, "epoch": 1783, "lr": 3.697348410737277e-05}
{"train_loss": 0.13491623103618622, "global_step": 158707, "epoch": 1783, "lr": 3.6972924392462164e-05}
{"train_loss": 0.0834188386797905, "global_step": 158708, "epoch": 1783, "lr": 3.697236467930289e-05}
{"train_loss": 0.07216328382492065, "global_step": 158709, "epoch": 1783, "lr": 3.6971804967895037e-05}
{"train_loss": 0.15148687362670898, "global_step": 158710, "epoch": 1783, "lr": 3.697124525823865e-05}
{"train_loss": 0.09863366931676865, "global_step": 158711, "epoch": 1783, "lr": 3.6970685550333834e-05}
{"train_loss": 0.11849649250507355, "global_step": 158712, "epoch": 1783, "lr": 3.697012584418063e-05}
{"train_loss": 0.06860000640153885, "global_step": 158713, "epoch": 1783, "lr": 3.696956613977916e-05}
{"train_loss": 0.12184701859951019, "global_step": 158714, "epoch": 1783, "lr": 3.696900643712945e-05}
{"train_loss": 0.06062190234661102, "global_step": 158715, "epoch": 1783, "lr": 3.696844673623162e-05}
{"train_loss": 0.09382737427949905, "global_step": 158716, "epoch": 1783, "lr": 3.6967887037085705e-05}
{"train_loss": 0.12971727550029755, "global_step": 158717, "epoch": 1783, "lr": 3.696732733969182e-05}
{"train_loss": 0.20086424052715302, "global_step": 158718, "epoch": 1783, "lr": 3.696676764404999e-05}
{"train_loss": 0.12427496165037155, "global_step": 158719, "epoch": 1783, "lr": 3.6966207950160345e-05}
{"train_loss": 0.11611989885568619, "global_step": 158720, "epoch": 1783, "lr": 3.6965648258022914e-05}
{"train_loss": 0.20848585665225983, "global_step": 158721, "epoch": 1783, "lr": 3.6965088567637805e-05}
{"train_loss": 0.08172225952148438, "global_step": 158722, "epoch": 1783, "lr": 3.6964528879005085e-05}
{"train_loss": 0.0799730122089386, "global_step": 158723, "epoch": 1783, "lr": 3.6963969192124815e-05}
{"train_loss": 0.09369362890720367, "global_step": 158724, "epoch": 1783, "lr": 3.6963409506997084e-05}
{"train_loss": 0.11753565818071365, "global_step": 158725, "epoch": 1783, "lr": 3.696284982362197e-05}
{"train_loss": 0.08326328545808792, "global_step": 158726, "epoch": 1783, "lr": 3.6962290141999533e-05}
{"train_loss": 0.08643070608377457, "global_step": 158727, "epoch": 1783, "lr": 3.696173046212985e-05}
{"train_loss": 0.10365818440914154, "global_step": 158728, "epoch": 1783, "lr": 3.6961170784013024e-05}
{"train_loss": 0.08800909668207169, "global_step": 158729, "epoch": 1783, "lr": 3.696061110764909e-05}
{"train_loss": 0.13555367290973663, "global_step": 158730, "epoch": 1783, "lr": 3.696005143303816e-05}
{"train_loss": 0.07370498031377792, "global_step": 158731, "epoch": 1783, "lr": 3.6959491760180275e-05}
{"train_loss": 0.09367507696151733, "global_step": 158732, "epoch": 1783, "lr": 3.695893208907555e-05}
{"train_loss": 0.08540209382772446, "global_step": 158733, "epoch": 1783, "lr": 3.695837241972401e-05}
{"train_loss": 0.09721452742815018, "global_step": 158734, "epoch": 1783, "lr": 3.695781275212578e-05}
{"train_loss": 0.08142904937267303, "global_step": 158735, "epoch": 1783, "lr": 3.6957253086280894e-05}
{"train_loss": 0.0639454573392868, "global_step": 158736, "epoch": 1783, "lr": 3.695669342218947e-05}
{"train_loss": 0.10058682411909103, "global_step": 158737, "epoch": 1783, "lr": 3.695613375985153e-05}
{"train_loss": 0.0705995187163353, "global_step": 158738, "epoch": 1783, "lr": 3.695557409926721e-05}
{"train_loss": 0.10665536671876907, "global_step": 158739, "epoch": 1783, "lr": 3.695501444043653e-05}
{"train_loss": 0.07360541820526123, "global_step": 158740, "epoch": 1783, "lr": 3.695445478335962e-05}
{"train_loss": 0.08384207636117935, "global_step": 158741, "epoch": 1783, "lr": 3.6953895128036496e-05}
{"train_loss": 0.09832174330949783, "global_step": 158742, "epoch": 1783, "lr": 3.695333547446727e-05}
{"train_loss": 0.13713401556015015, "global_step": 158743, "epoch": 1783, "lr": 3.6952775822652006e-05}
{"train_loss": 0.10599380731582642, "global_step": 158744, "epoch": 1783, "lr": 3.6952216172590775e-05}
{"train_loss": 0.1219877228140831, "global_step": 158745, "epoch": 1783, "lr": 3.695165652428369e-05}
{"train_loss": 0.11890112608671188, "global_step": 158746, "epoch": 1783, "lr": 3.695109687773076e-05}
{"train_loss": 0.10890211910009384, "global_step": 158747, "epoch": 1783, "lr": 3.6950537232932115e-05}
{"train_loss": 0.10717709362506866, "global_step": 158748, "epoch": 1783, "lr": 3.69499775898878e-05}
{"train_loss": 0.08579084277153015, "global_step": 158749, "epoch": 1783, "lr": 3.694941794859792e-05}
{"train_loss": 0.08045990020036697, "global_step": 158750, "epoch": 1783, "lr": 3.694885830906251e-05}
{"train_loss": 0.09498611837625504, "global_step": 158751, "epoch": 1783, "lr": 3.694829867128169e-05}
{"train_loss": 0.05215843766927719, "global_step": 158752, "epoch": 1783, "lr": 3.694773903525549e-05}
{"train_loss": 0.08051986247301102, "global_step": 158753, "epoch": 1783, "lr": 3.694717940098402e-05}
{"train_loss": 0.10946110635995865, "global_step": 158754, "epoch": 1783, "lr": 3.6946619768467336e-05}
{"train_loss": 0.11485342681407928, "global_step": 158755, "epoch": 1783, "lr": 3.694606013770553e-05}
{"train_loss": 0.058179985731840134, "global_step": 158756, "epoch": 1783, "lr": 3.694550050869865e-05}
{"train_loss": 0.06480822712182999, "global_step": 158757, "epoch": 1783, "lr": 3.694494088144681e-05}
{"train_loss": 0.08592981845140457, "global_step": 158758, "epoch": 1783, "lr": 3.6944381255950046e-05}
{"train_loss": 0.07902897149324417, "global_step": 158759, "epoch": 1783, "lr": 3.6943821632208466e-05}
{"train_loss": 0.06688375025987625, "global_step": 158760, "epoch": 1783, "lr": 3.694326201022212e-05}
{"train_loss": 0.05619269236922264, "global_step": 158761, "epoch": 1783, "lr": 3.69427023899911e-05}
{"train_loss": 0.08721893280744553, "global_step": 158762, "epoch": 1783, "lr": 3.694214277151546e-05}
{"train_loss": 0.11307229101657867, "global_step": 158763, "epoch": 1783, "lr": 3.694158315479529e-05}
{"train_loss": 0.06744532287120819, "global_step": 158764, "epoch": 1783, "lr": 3.6941023539830686e-05}
{"train_loss": 0.1283521205186844, "global_step": 158765, "epoch": 1783, "lr": 3.694046392662168e-05}
{"train_loss": 0.06608983129262924, "global_step": 158766, "epoch": 1783, "lr": 3.6939904315168396e-05}
{"train_loss": 0.08165156841278076, "global_step": 158767, "epoch": 1783, "lr": 3.693934470547085e-05}
{"train_loss": 0.032470278441905975, "global_step": 158768, "epoch": 1783, "lr": 3.6938785097529174e-05}
{"train_loss": 0.13140912353992462, "global_step": 158769, "epoch": 1783, "lr": 3.69382254913434e-05}
{"train_loss": 0.13803565502166748, "global_step": 158770, "epoch": 1783, "lr": 3.693766588691365e-05}
{"train_loss": 0.06837932765483856, "global_step": 158771, "epoch": 1783, "lr": 3.693710628423994e-05}
{"train_loss": 0.1693875640630722, "global_step": 158772, "epoch": 1783, "lr": 3.6936546683322404e-05}
{"train_loss": 0.0452103465795517, "global_step": 158773, "epoch": 1783, "lr": 3.693598708416107e-05}
{"train_loss": 0.15204504132270813, "global_step": 158774, "epoch": 1783, "lr": 3.6935427486756054e-05}
{"train_loss": 0.09816861479135042, "global_step": 158775, "epoch": 1783, "lr": 3.6934867891107386e-05, "val_loss": 6.810370922088623}
{"train_loss": 0.055381838232278824, "global_step": 158776, "epoch": 1784, "lr": 3.693430829721519e-05}
{"train_loss": 0.0964471697807312, "global_step": 158777, "epoch": 1784, "lr": 3.6933748705079495e-05}
{"train_loss": 0.12134058028459549, "global_step": 158778, "epoch": 1784, "lr": 3.6933189114700415e-05}
{"train_loss": 0.05504663288593292, "global_step": 158779, "epoch": 1784, "lr": 3.6932629526078e-05}
{"train_loss": 0.06492070853710175, "global_step": 158780, "epoch": 1784, "lr": 3.6932069939212324e-05}
{"train_loss": 0.10099953413009644, "global_step": 158781, "epoch": 1784, "lr": 3.69315103541035e-05}
{"train_loss": 0.08275808393955231, "global_step": 158782, "epoch": 1784, "lr": 3.693095077075155e-05}
{"train_loss": 0.07290703058242798, "global_step": 158783, "epoch": 1784, "lr": 3.693039118915659e-05}
{"train_loss": 0.142774298787117, "global_step": 158784, "epoch": 1784, "lr": 3.692983160931866e-05}
{"train_loss": 0.12235917150974274, "global_step": 158785, "epoch": 1784, "lr": 3.692927203123788e-05}
{"train_loss": 0.0626731589436531, "global_step": 158786, "epoch": 1784, "lr": 3.6928712454914274e-05}
{"train_loss": 0.09171470999717712, "global_step": 158787, "epoch": 1784, "lr": 3.6928152880347965e-05}
{"train_loss": 0.10327968746423721, "global_step": 158788, "epoch": 1784, "lr": 3.692759330753899e-05}
{"train_loss": 0.10954000800848007, "global_step": 158789, "epoch": 1784, "lr": 3.692703373648746e-05}
{"train_loss": 0.07992364466190338, "global_step": 158790, "epoch": 1784, "lr": 3.692647416719341e-05}
{"train_loss": 0.06530529260635376, "global_step": 158791, "epoch": 1784, "lr": 3.692591459965696e-05}
{"train_loss": 0.08559716492891312, "global_step": 158792, "epoch": 1784, "lr": 3.6925355033878134e-05}
{"train_loss": 0.14813266694545746, "global_step": 158793, "epoch": 1784, "lr": 3.692479546985705e-05}
{"train_loss": 0.042879655957221985, "global_step": 158794, "epoch": 1784, "lr": 3.692423590759376e-05}
{"train_loss": 0.0961570143699646, "global_step": 158795, "epoch": 1784, "lr": 3.692367634708837e-05}
{"train_loss": 0.07143903523683548, "global_step": 158796, "epoch": 1784, "lr": 3.6923116788340905e-05}
{"train_loss": 0.09519369155168533, "global_step": 158797, "epoch": 1784, "lr": 3.6922557231351485e-05}
{"train_loss": 0.11200576275587082, "global_step": 158798, "epoch": 1784, "lr": 3.6921997676120166e-05}
{"train_loss": 0.09766262769699097, "global_step": 158799, "epoch": 1784, "lr": 3.6921438122647e-05}
{"train_loss": 0.09231748431921005, "global_step": 158800, "epoch": 1784, "lr": 3.692087857093212e-05}
{"train_loss": 0.1595260202884674, "global_step": 158801, "epoch": 1784, "lr": 3.692031902097555e-05}
{"train_loss": 0.12510421872138977, "global_step": 158802, "epoch": 1784, "lr": 3.6919759472777395e-05}
{"train_loss": 0.07293566316366196, "global_step": 158803, "epoch": 1784, "lr": 3.69191999263377e-05}
{"train_loss": 0.1228252500295639, "global_step": 158804, "epoch": 1784, "lr": 3.6918640381656585e-05}
{"train_loss": 0.0936577320098877, "global_step": 158805, "epoch": 1784, "lr": 3.6918080838734084e-05}
{"train_loss": 0.07292535156011581, "global_step": 158806, "epoch": 1784, "lr": 3.6917521297570295e-05}
{"train_loss": 0.15220069885253906, "global_step": 158807, "epoch": 1784, "lr": 3.691696175816527e-05}
{"train_loss": 0.10565317422151566, "global_step": 158808, "epoch": 1784, "lr": 3.691640222051912e-05}
{"train_loss": 0.06961969286203384, "global_step": 158809, "epoch": 1784, "lr": 3.691584268463189e-05}
{"train_loss": 0.09442297369241714, "global_step": 158810, "epoch": 1784, "lr": 3.691528315050367e-05}
{"train_loss": 0.08704150468111038, "global_step": 158811, "epoch": 1784, "lr": 3.6914723618134515e-05}
{"train_loss": 0.08254844695329666, "global_step": 158812, "epoch": 1784, "lr": 3.691416408752454e-05}
{"train_loss": 0.07640870660543442, "global_step": 158813, "epoch": 1784, "lr": 3.6913604558673776e-05}
{"train_loss": 0.07213691622018814, "global_step": 158814, "epoch": 1784, "lr": 3.6913045031582335e-05}
{"train_loss": 0.06843150407075882, "global_step": 158815, "epoch": 1784, "lr": 3.691248550625026e-05}
{"train_loss": 0.0902881771326065, "global_step": 158816, "epoch": 1784, "lr": 3.691192598267766e-05}
{"train_loss": 0.08494743704795837, "global_step": 158817, "epoch": 1784, "lr": 3.691136646086457e-05}
{"train_loss": 0.06795601546764374, "global_step": 158818, "epoch": 1784, "lr": 3.69108069408111e-05}
{"train_loss": 0.1263403445482254, "global_step": 158819, "epoch": 1784, "lr": 3.691024742251732e-05}
{"train_loss": 0.1411954015493393, "global_step": 158820, "epoch": 1784, "lr": 3.6909687905983284e-05}
{"train_loss": 0.1114932969212532, "global_step": 158821, "epoch": 1784, "lr": 3.690912839120909e-05}
{"train_loss": 0.14262655377388, "global_step": 158822, "epoch": 1784, "lr": 3.6908568878194795e-05}
{"train_loss": 0.07590425759553909, "global_step": 158823, "epoch": 1784, "lr": 3.690800936694049e-05}
{"train_loss": 0.07598280161619186, "global_step": 158824, "epoch": 1784, "lr": 3.6907449857446234e-05}
{"train_loss": 0.06385036557912827, "global_step": 158825, "epoch": 1784, "lr": 3.690689034971213e-05}
{"train_loss": 0.1176685318350792, "global_step": 158826, "epoch": 1784, "lr": 3.690633084373821e-05}
{"train_loss": 0.1605168581008911, "global_step": 158827, "epoch": 1784, "lr": 3.69057713395246e-05}
{"train_loss": 0.10636958479881287, "global_step": 158828, "epoch": 1784, "lr": 3.690521183707133e-05}
{"train_loss": 0.10431525111198425, "global_step": 158829, "epoch": 1784, "lr": 3.6904652336378516e-05}
{"train_loss": 0.11880527436733246, "global_step": 158830, "epoch": 1784, "lr": 3.6904092837446184e-05}
{"train_loss": 0.11867846548557281, "global_step": 158831, "epoch": 1784, "lr": 3.690353334027447e-05}
{"train_loss": 0.0647113248705864, "global_step": 158832, "epoch": 1784, "lr": 3.6902973844863386e-05}
{"train_loss": 0.08219791948795319, "global_step": 158833, "epoch": 1784, "lr": 3.6902414351213056e-05}
{"train_loss": 0.09220237284898758, "global_step": 158834, "epoch": 1784, "lr": 3.690185485932352e-05}
{"train_loss": 0.12016689777374268, "global_step": 158835, "epoch": 1784, "lr": 3.690129536919488e-05}
{"train_loss": 0.16099126636981964, "global_step": 158836, "epoch": 1784, "lr": 3.69007358808272e-05}
{"train_loss": 0.06889760494232178, "global_step": 158837, "epoch": 1784, "lr": 3.690017639422056e-05}
{"train_loss": 0.09749229997396469, "global_step": 158838, "epoch": 1784, "lr": 3.6899616909375045e-05}
{"train_loss": 0.06279943138360977, "global_step": 158839, "epoch": 1784, "lr": 3.6899057426290686e-05}
{"train_loss": 0.07701563835144043, "global_step": 158840, "epoch": 1784, "lr": 3.689849794496761e-05}
{"train_loss": 0.06946084648370743, "global_step": 158841, "epoch": 1784, "lr": 3.689793846540586e-05}
{"train_loss": 0.0874912217259407, "global_step": 158842, "epoch": 1784, "lr": 3.689737898760554e-05}
{"train_loss": 0.05332479998469353, "global_step": 158843, "epoch": 1784, "lr": 3.689681951156668e-05}
{"train_loss": 0.07995299994945526, "global_step": 158844, "epoch": 1784, "lr": 3.689626003728942e-05}
{"train_loss": 0.059932637959718704, "global_step": 158845, "epoch": 1784, "lr": 3.689570056477376e-05}
{"train_loss": 0.1405218243598938, "global_step": 158846, "epoch": 1784, "lr": 3.689514109401985e-05}
{"train_loss": 0.04025805741548538, "global_step": 158847, "epoch": 1784, "lr": 3.689458162502769e-05}
{"train_loss": 0.15055139362812042, "global_step": 158848, "epoch": 1784, "lr": 3.689402215779742e-05}
{"train_loss": 0.08347856998443604, "global_step": 158849, "epoch": 1784, "lr": 3.689346269232907e-05}
{"train_loss": 0.03455350548028946, "global_step": 158850, "epoch": 1784, "lr": 3.6892903228622766e-05}
{"train_loss": 0.11233691871166229, "global_step": 158851, "epoch": 1784, "lr": 3.689234376667852e-05}
{"train_loss": 0.10251142829656601, "global_step": 158852, "epoch": 1784, "lr": 3.689178430649646e-05}
{"train_loss": 0.05072268471121788, "global_step": 158853, "epoch": 1784, "lr": 3.6891224848076616e-05}
{"train_loss": 0.12293975800275803, "global_step": 158854, "epoch": 1784, "lr": 3.68906653914191e-05}
{"train_loss": 0.1022990420460701, "global_step": 158855, "epoch": 1784, "lr": 3.689010593652398e-05}
{"train_loss": 0.10113970935344696, "global_step": 158856, "epoch": 1784, "lr": 3.6889546483391314e-05}
{"train_loss": 0.10459861159324646, "global_step": 158857, "epoch": 1784, "lr": 3.68889870320212e-05}
{"train_loss": 0.1055857464671135, "global_step": 158858, "epoch": 1784, "lr": 3.68884275824137e-05}
{"train_loss": 0.12238550931215286, "global_step": 158859, "epoch": 1784, "lr": 3.688786813456889e-05}
{"train_loss": 0.08348052203655243, "global_step": 158860, "epoch": 1784, "lr": 3.6887308688486835e-05}
{"train_loss": 0.05446743592619896, "global_step": 158861, "epoch": 1784, "lr": 3.688674924416764e-05}
{"train_loss": 0.07658864557743073, "global_step": 158862, "epoch": 1784, "lr": 3.688618980161135e-05}
{"train_loss": 0.12268008291721344, "global_step": 158863, "epoch": 1784, "lr": 3.688563036081806e-05}
{"train_loss": 0.0951280662936441, "global_step": 158864, "epoch": 1784, "lr": 3.688507092178782e-05, "val_loss": 6.766031742095947}
{"train_loss": 0.07889512926340103, "global_step": 158865, "epoch": 1785, "lr": 3.688451148452075e-05}
{"train_loss": 0.1307549625635147, "global_step": 158866, "epoch": 1785, "lr": 3.6883952049016865e-05}
{"train_loss": 0.09171122312545776, "global_step": 158867, "epoch": 1785, "lr": 3.68833926152763e-05}
{"train_loss": 0.12317338585853577, "global_step": 158868, "epoch": 1785, "lr": 3.6882833183299084e-05}
{"train_loss": 0.09425573796033859, "global_step": 158869, "epoch": 1785, "lr": 3.6882273753085335e-05}
{"train_loss": 0.06720992177724838, "global_step": 158870, "epoch": 1785, "lr": 3.6881714324635075e-05}
{"train_loss": 0.13113148510456085, "global_step": 158871, "epoch": 1785, "lr": 3.688115489794842e-05}
{"train_loss": 0.12179987132549286, "global_step": 158872, "epoch": 1785, "lr": 3.6880595473025434e-05}
{"train_loss": 0.10409579426050186, "global_step": 158873, "epoch": 1785, "lr": 3.68800360498662e-05}
{"train_loss": 0.09950033575296402, "global_step": 158874, "epoch": 1785, "lr": 3.687947662847078e-05}
{"train_loss": 0.16917218267917633, "global_step": 158875, "epoch": 1785, "lr": 3.6878917208839256e-05}
{"train_loss": 0.051482293754816055, "global_step": 158876, "epoch": 1785, "lr": 3.6878357790971704e-05}
{"train_loss": 0.051237624138593674, "global_step": 158877, "epoch": 1785, "lr": 3.687779837486819e-05}
{"train_loss": 0.07519190013408661, "global_step": 158878, "epoch": 1785, "lr": 3.687723896052881e-05}
{"train_loss": 0.06450445204973221, "global_step": 158879, "epoch": 1785, "lr": 3.6876679547953604e-05}
{"train_loss": 0.08558192104101181, "global_step": 158880, "epoch": 1785, "lr": 3.687612013714269e-05}
{"train_loss": 0.10041726380586624, "global_step": 158881, "epoch": 1785, "lr": 3.687556072809609e-05}
{"train_loss": 0.05270593985915184, "global_step": 158882, "epoch": 1785, "lr": 3.687500132081395e-05}
{"train_loss": 0.09268810600042343, "global_step": 158883, "epoch": 1785, "lr": 3.6874441915296274e-05}
{"train_loss": 0.08470603078603745, "global_step": 158884, "epoch": 1785, "lr": 3.68738825115432e-05}
{"train_loss": 0.06467259675264359, "global_step": 158885, "epoch": 1785, "lr": 3.6873323109554734e-05}
{"train_loss": 0.09342686831951141, "global_step": 158886, "epoch": 1785, "lr": 3.687276370933103e-05}
{"train_loss": 0.058896295726299286, "global_step": 158887, "epoch": 1785, "lr": 3.6872204310872096e-05}
{"train_loss": 0.15389850735664368, "global_step": 158888, "epoch": 1785, "lr": 3.687164491417805e-05}
{"train_loss": 0.10838112235069275, "global_step": 158889, "epoch": 1785, "lr": 3.687108551924893e-05}
{"train_loss": 0.13248181343078613, "global_step": 158890, "epoch": 1785, "lr": 3.687052612608485e-05}
{"train_loss": 0.1466924548149109, "global_step": 158891, "epoch": 1785, "lr": 3.6869966734685856e-05}
{"train_loss": 0.11023450642824173, "global_step": 158892, "epoch": 1785, "lr": 3.686940734505204e-05}
{"train_loss": 0.10550955682992935, "global_step": 158893, "epoch": 1785, "lr": 3.686884795718348e-05}
{"train_loss": 0.1289842575788498, "global_step": 158894, "epoch": 1785, "lr": 3.686828857108023e-05}
{"train_loss": 0.13235029578208923, "global_step": 158895, "epoch": 1785, "lr": 3.6867729186742405e-05}
{"train_loss": 0.06514272838830948, "global_step": 158896, "epoch": 1785, "lr": 3.6867169804170024e-05}
{"train_loss": 0.12700510025024414, "global_step": 158897, "epoch": 1785, "lr": 3.686661042336321e-05}
{"train_loss": 0.09624002873897552, "global_step": 158898, "epoch": 1785, "lr": 3.6866051044322004e-05}
{"train_loss": 0.12284242361783981, "global_step": 158899, "epoch": 1785, "lr": 3.686549166704652e-05}
{"train_loss": 0.09681250900030136, "global_step": 158900, "epoch": 1785, "lr": 3.686493229153679e-05}
{"train_loss": 0.10450682044029236, "global_step": 158901, "epoch": 1785, "lr": 3.686437291779292e-05}
{"train_loss": 0.09813322871923447, "global_step": 158902, "epoch": 1785, "lr": 3.686381354581496e-05}
{"train_loss": 0.12508541345596313, "global_step": 158903, "epoch": 1785, "lr": 3.6863254175603026e-05}
{"train_loss": 0.07551603019237518, "global_step": 158904, "epoch": 1785, "lr": 3.686269480715715e-05}
{"train_loss": 0.0989527627825737, "global_step": 158905, "epoch": 1785, "lr": 3.686213544047744e-05}
{"train_loss": 0.11617706716060638, "global_step": 158906, "epoch": 1785, "lr": 3.686157607556393e-05}
{"train_loss": 0.07880229502916336, "global_step": 158907, "epoch": 1785, "lr": 3.686101671241672e-05}
{"train_loss": 0.08099577575922012, "global_step": 158908, "epoch": 1785, "lr": 3.686045735103592e-05}
{"train_loss": 0.11402211338281631, "global_step": 158909, "epoch": 1785, "lr": 3.6859897991421546e-05}
{"train_loss": 0.09503529965877533, "global_step": 158910, "epoch": 1785, "lr": 3.6859338633573704e-05}
{"train_loss": 0.1113002821803093, "global_step": 158911, "epoch": 1785, "lr": 3.685877927749247e-05}
{"train_loss": 0.07523903250694275, "global_step": 158912, "epoch": 1785, "lr": 3.685821992317791e-05}
{"train_loss": 0.04507201910018921, "global_step": 158913, "epoch": 1785, "lr": 3.685766057063009e-05}
{"train_loss": 0.12105195224285126, "global_step": 158914, "epoch": 1785, "lr": 3.685710121984911e-05}
{"train_loss": 0.031040187925100327, "global_step": 158915, "epoch": 1785, "lr": 3.685654187083503e-05}
{"train_loss": 0.16587963700294495, "global_step": 158916, "epoch": 1785, "lr": 3.685598252358794e-05}
{"train_loss": 0.09462939947843552, "global_step": 158917, "epoch": 1785, "lr": 3.685542317810788e-05}
{"train_loss": 0.14046625792980194, "global_step": 158918, "epoch": 1785, "lr": 3.685486383439497e-05}
{"train_loss": 0.08052689582109451, "global_step": 158919, "epoch": 1785, "lr": 3.6854304492449236e-05}
{"train_loss": 0.08279719203710556, "global_step": 158920, "epoch": 1785, "lr": 3.6853745152270814e-05}
{"train_loss": 0.09817297756671906, "global_step": 158921, "epoch": 1785, "lr": 3.6853185813859715e-05}
{"train_loss": 0.15084047615528107, "global_step": 158922, "epoch": 1785, "lr": 3.685262647721607e-05}
{"train_loss": 0.09183313697576523, "global_step": 158923, "epoch": 1785, "lr": 3.6852067142339905e-05}
{"train_loss": 0.12557019293308258, "global_step": 158924, "epoch": 1785, "lr": 3.685150780923135e-05}
{"train_loss": 0.1013287752866745, "global_step": 158925, "epoch": 1785, "lr": 3.685094847789041e-05}
{"train_loss": 0.14717116951942444, "global_step": 158926, "epoch": 1785, "lr": 3.685038914831721e-05}
{"train_loss": 0.0976727157831192, "global_step": 158927, "epoch": 1785, "lr": 3.684982982051184e-05}
{"train_loss": 0.12668004631996155, "global_step": 158928, "epoch": 1785, "lr": 3.684927049447432e-05}
{"train_loss": 0.09789136797189713, "global_step": 158929, "epoch": 1785, "lr": 3.684871117020479e-05}
{"train_loss": 0.07859282940626144, "global_step": 158930, "epoch": 1785, "lr": 3.6848151847703265e-05}
{"train_loss": 0.06804948300123215, "global_step": 158931, "epoch": 1785, "lr": 3.684759252696985e-05}
{"train_loss": 0.0819440484046936, "global_step": 158932, "epoch": 1785, "lr": 3.684703320800462e-05}
{"train_loss": 0.10153207927942276, "global_step": 158933, "epoch": 1785, "lr": 3.6846473890807644e-05}
{"train_loss": 0.06471141427755356, "global_step": 158934, "epoch": 1785, "lr": 3.6845914575379e-05}
{"train_loss": 0.10843809694051743, "global_step": 158935, "epoch": 1785, "lr": 3.684535526171877e-05}
{"train_loss": 0.06696580350399017, "global_step": 158936, "epoch": 1785, "lr": 3.6844795949827005e-05}
{"train_loss": 0.08911315351724625, "global_step": 158937, "epoch": 1785, "lr": 3.684423663970382e-05}
{"train_loss": 0.06289999932050705, "global_step": 158938, "epoch": 1785, "lr": 3.684367733134924e-05}
{"train_loss": 0.09129256755113602, "global_step": 158939, "epoch": 1785, "lr": 3.684311802476339e-05}
{"train_loss": 0.09038139879703522, "global_step": 158940, "epoch": 1785, "lr": 3.68425587199463e-05}
{"train_loss": 0.10405359417200089, "global_step": 158941, "epoch": 1785, "lr": 3.684199941689809e-05}
{"train_loss": 0.06052980571985245, "global_step": 158942, "epoch": 1785, "lr": 3.68414401156188e-05}
{"train_loss": 0.05892537534236908, "global_step": 158943, "epoch": 1785, "lr": 3.684088081610853e-05}
{"train_loss": 0.07688530534505844, "global_step": 158944, "epoch": 1785, "lr": 3.6840321518367316e-05}
{"train_loss": 0.09300890564918518, "global_step": 158945, "epoch": 1785, "lr": 3.683976222239527e-05}
{"train_loss": 0.05699681490659714, "global_step": 158946, "epoch": 1785, "lr": 3.6839202928192476e-05}
{"train_loss": 0.18870782852172852, "global_step": 158947, "epoch": 1785, "lr": 3.683864363575897e-05}
{"train_loss": 0.0792158916592598, "global_step": 158948, "epoch": 1785, "lr": 3.683808434509487e-05}
{"train_loss": 0.061952561140060425, "global_step": 158949, "epoch": 1785, "lr": 3.683752505620022e-05}
{"train_loss": 0.07441246509552002, "global_step": 158950, "epoch": 1785, "lr": 3.68369657690751e-05}
{"train_loss": 0.16779294610023499, "global_step": 158951, "epoch": 1785, "lr": 3.6836406483719586e-05}
{"train_loss": 0.258249968290329, "global_step": 158952, "epoch": 1785, "lr": 3.6835847200133766e-05}
{"train_loss": 0.09984493632329984, "global_step": 158953, "epoch": 1785, "lr": 3.68352879183177e-05, "val_loss": 6.723156929016113, "train_action_mse_error": 23.15321922302246}
{"train_loss": 0.11825469136238098, "global_step": 158954, "epoch": 1786, "lr": 3.6834728638271484e-05}
{"train_loss": 0.10222960263490677, "global_step": 158955, "epoch": 1786, "lr": 3.683416935999515e-05}
{"train_loss": 0.06551043689250946, "global_step": 158956, "epoch": 1786, "lr": 3.683361008348883e-05}
{"train_loss": 0.06673698872327805, "global_step": 158957, "epoch": 1786, "lr": 3.683305080875255e-05}
{"train_loss": 0.13306522369384766, "global_step": 158958, "epoch": 1786, "lr": 3.683249153578642e-05}
{"train_loss": 0.10248582065105438, "global_step": 158959, "epoch": 1786, "lr": 3.6831932264590483e-05}
{"train_loss": 0.07661771774291992, "global_step": 158960, "epoch": 1786, "lr": 3.683137299516486e-05}
{"train_loss": 0.12092772871255875, "global_step": 158961, "epoch": 1786, "lr": 3.683081372750957e-05}
{"train_loss": 0.12586162984371185, "global_step": 158962, "epoch": 1786, "lr": 3.6830254461624716e-05}
{"train_loss": 0.054918333888053894, "global_step": 158963, "epoch": 1786, "lr": 3.6829695197510394e-05}
{"train_loss": 0.1076958104968071, "global_step": 158964, "epoch": 1786, "lr": 3.682913593516664e-05}
{"train_loss": 0.1079595685005188, "global_step": 158965, "epoch": 1786, "lr": 3.6828576674593576e-05}
{"train_loss": 0.09087951481342316, "global_step": 158966, "epoch": 1786, "lr": 3.6828017415791215e-05}
{"train_loss": 0.10862957686185837, "global_step": 158967, "epoch": 1786, "lr": 3.6827458158759695e-05}
{"train_loss": 0.07879003882408142, "global_step": 158968, "epoch": 1786, "lr": 3.682689890349905e-05}
{"train_loss": 0.09680914878845215, "global_step": 158969, "epoch": 1786, "lr": 3.682633965000938e-05}
{"train_loss": 0.03686676174402237, "global_step": 158970, "epoch": 1786, "lr": 3.6825780398290736e-05}
{"train_loss": 0.11197801679372787, "global_step": 158971, "epoch": 1786, "lr": 3.68252211483432e-05}
{"train_loss": 0.094382144510746, "global_step": 158972, "epoch": 1786, "lr": 3.682466190016687e-05}
{"train_loss": 0.12373775988817215, "global_step": 158973, "epoch": 1786, "lr": 3.682410265376179e-05}
{"train_loss": 0.06515303999185562, "global_step": 158974, "epoch": 1786, "lr": 3.682354340912805e-05}
{"train_loss": 0.08259876817464828, "global_step": 158975, "epoch": 1786, "lr": 3.6822984166265736e-05}
{"train_loss": 0.10104335099458694, "global_step": 158976, "epoch": 1786, "lr": 3.6822424925174895e-05}
{"train_loss": 0.08503521233797073, "global_step": 158977, "epoch": 1786, "lr": 3.682186568585563e-05}
{"train_loss": 0.11313194036483765, "global_step": 158978, "epoch": 1786, "lr": 3.682130644830799e-05}
{"train_loss": 0.09037347137928009, "global_step": 158979, "epoch": 1786, "lr": 3.682074721253208e-05}
{"train_loss": 0.10551492869853973, "global_step": 158980, "epoch": 1786, "lr": 3.682018797852794e-05}
{"train_loss": 0.08407530188560486, "global_step": 158981, "epoch": 1786, "lr": 3.681962874629568e-05}
{"train_loss": 0.07813489437103271, "global_step": 158982, "epoch": 1786, "lr": 3.6819069515835363e-05}
{"train_loss": 0.13593865931034088, "global_step": 158983, "epoch": 1786, "lr": 3.681851028714704e-05}
{"train_loss": 0.08498793095350266, "global_step": 158984, "epoch": 1786, "lr": 3.6817951060230844e-05}
{"train_loss": 0.059834323823451996, "global_step": 158985, "epoch": 1786, "lr": 3.681739183508679e-05}
{"train_loss": 0.07356420159339905, "global_step": 158986, "epoch": 1786, "lr": 3.681683261171499e-05}
{"train_loss": 0.12395023554563522, "global_step": 158987, "epoch": 1786, "lr": 3.68162733901155e-05}
{"train_loss": 0.05670812353491783, "global_step": 158988, "epoch": 1786, "lr": 3.68157141702884e-05}
{"train_loss": 0.06428632140159607, "global_step": 158989, "epoch": 1786, "lr": 3.681515495223376e-05}
{"train_loss": 0.09078054875135422, "global_step": 158990, "epoch": 1786, "lr": 3.681459573595167e-05}
{"train_loss": 0.07060413062572479, "global_step": 158991, "epoch": 1786, "lr": 3.681403652144219e-05}
{"train_loss": 0.0829775333404541, "global_step": 158992, "epoch": 1786, "lr": 3.681347730870542e-05}
{"train_loss": 0.09708858281373978, "global_step": 158993, "epoch": 1786, "lr": 3.681291809774139e-05}
{"train_loss": 0.1689973771572113, "global_step": 158994, "epoch": 1786, "lr": 3.681235888855023e-05}
{"train_loss": 0.11204801499843597, "global_step": 158995, "epoch": 1786, "lr": 3.6811799681131966e-05}
{"train_loss": 0.05542640760540962, "global_step": 158996, "epoch": 1786, "lr": 3.681124047548671e-05}
{"train_loss": 0.05366532504558563, "global_step": 158997, "epoch": 1786, "lr": 3.68106812716145e-05}
{"train_loss": 0.059132903814315796, "global_step": 158998, "epoch": 1786, "lr": 3.681012206951544e-05}
{"train_loss": 0.08153000473976135, "global_step": 158999, "epoch": 1786, "lr": 3.680956286918962e-05}
{"train_loss": 0.2021234780550003, "global_step": 159000, "epoch": 1786, "lr": 3.680900367063706e-05}
{"train_loss": 0.08424017578363419, "global_step": 159001, "epoch": 1786, "lr": 3.6808444473857903e-05}
{"train_loss": 0.15477964282035828, "global_step": 159002, "epoch": 1786, "lr": 3.6807885278852167e-05}
{"train_loss": 0.11609083414077759, "global_step": 159003, "epoch": 1786, "lr": 3.680732608561997e-05}
{"train_loss": 0.14716453850269318, "global_step": 159004, "epoch": 1786, "lr": 3.680676689416135e-05}
{"train_loss": 0.15267780423164368, "global_step": 159005, "epoch": 1786, "lr": 3.6806207704476414e-05}
{"train_loss": 0.11474549025297165, "global_step": 159006, "epoch": 1786, "lr": 3.680564851656521e-05}
{"train_loss": 0.07886244356632233, "global_step": 159007, "epoch": 1786, "lr": 3.6805089330427825e-05}
{"train_loss": 0.08922835439443588, "global_step": 159008, "epoch": 1786, "lr": 3.6804530146064345e-05}
{"train_loss": 0.0500820055603981, "global_step": 159009, "epoch": 1786, "lr": 3.6803970963474824e-05}
{"train_loss": 0.11317117512226105, "global_step": 159010, "epoch": 1786, "lr": 3.6803411782659355e-05}
{"train_loss": 0.12761104106903076, "global_step": 159011, "epoch": 1786, "lr": 3.680285260361802e-05}
{"train_loss": 0.1015506237745285, "global_step": 159012, "epoch": 1786, "lr": 3.6802293426350855e-05}
{"train_loss": 0.08716700226068497, "global_step": 159013, "epoch": 1786, "lr": 3.680173425085798e-05}
{"train_loss": 0.17076866328716278, "global_step": 159014, "epoch": 1786, "lr": 3.680117507713943e-05}
{"train_loss": 0.08888547867536545, "global_step": 159015, "epoch": 1786, "lr": 3.680061590519532e-05}
{"train_loss": 0.08641280978918076, "global_step": 159016, "epoch": 1786, "lr": 3.680005673502569e-05}
{"train_loss": 0.08901120722293854, "global_step": 159017, "epoch": 1786, "lr": 3.6799497566630636e-05}
{"train_loss": 0.08765452355146408, "global_step": 159018, "epoch": 1786, "lr": 3.679893840001024e-05}
{"train_loss": 0.0936661884188652, "global_step": 159019, "epoch": 1786, "lr": 3.679837923516455e-05}
{"train_loss": 0.11356374621391296, "global_step": 159020, "epoch": 1786, "lr": 3.679782007209368e-05}
{"train_loss": 0.094304658472538, "global_step": 159021, "epoch": 1786, "lr": 3.679726091079766e-05}
{"train_loss": 0.1475166231393814, "global_step": 159022, "epoch": 1786, "lr": 3.679670175127661e-05}
{"train_loss": 0.11200536042451859, "global_step": 159023, "epoch": 1786, "lr": 3.6796142593530555e-05}
{"train_loss": 0.09462496638298035, "global_step": 159024, "epoch": 1786, "lr": 3.679558343755962e-05}
{"train_loss": 0.08004151284694672, "global_step": 159025, "epoch": 1786, "lr": 3.679502428336385e-05}
{"train_loss": 0.09594197571277618, "global_step": 159026, "epoch": 1786, "lr": 3.6794465130943324e-05}
{"train_loss": 0.09451913833618164, "global_step": 159027, "epoch": 1786, "lr": 3.6793905980298125e-05}
{"train_loss": 0.07259999215602875, "global_step": 159028, "epoch": 1786, "lr": 3.679334683142833e-05}
{"train_loss": 0.10986397415399551, "global_step": 159029, "epoch": 1786, "lr": 3.6792787684333994e-05}
{"train_loss": 0.1135922223329544, "global_step": 159030, "epoch": 1786, "lr": 3.6792228539015216e-05}
{"train_loss": 0.05890993773937225, "global_step": 159031, "epoch": 1786, "lr": 3.6791669395472064e-05}
{"train_loss": 0.05726943165063858, "global_step": 159032, "epoch": 1786, "lr": 3.679111025370462e-05}
{"train_loss": 0.056126102805137634, "global_step": 159033, "epoch": 1786, "lr": 3.6790551113712925e-05}
{"train_loss": 0.07436160743236542, "global_step": 159034, "epoch": 1786, "lr": 3.678999197549708e-05}
{"train_loss": 0.10582515597343445, "global_step": 159035, "epoch": 1786, "lr": 3.678943283905719e-05}
{"train_loss": 0.0877961590886116, "global_step": 159036, "epoch": 1786, "lr": 3.6788873704393265e-05}
{"train_loss": 0.12260867655277252, "global_step": 159037, "epoch": 1786, "lr": 3.678831457150544e-05}
{"train_loss": 0.08336641639471054, "global_step": 159038, "epoch": 1786, "lr": 3.678775544039376e-05}
{"train_loss": 0.09814929217100143, "global_step": 159039, "epoch": 1786, "lr": 3.67871963110583e-05}
{"train_loss": 0.1304503232240677, "global_step": 159040, "epoch": 1786, "lr": 3.6786637183499136e-05}
{"train_loss": 0.08077143877744675, "global_step": 159041, "epoch": 1786, "lr": 3.678607805771637e-05}
{"train_loss": 0.0970359304815196, "global_step": 159042, "epoch": 1786, "lr": 3.678551893371003e-05, "val_loss": 6.4978108406066895}
{"train_loss": 0.06627564132213593, "global_step": 159043, "epoch": 1787, "lr": 3.678495981148023e-05}
{"train_loss": 0.0594385527074337, "global_step": 159044, "epoch": 1787, "lr": 3.6784400691027015e-05}
{"train_loss": 0.17254866659641266, "global_step": 159045, "epoch": 1787, "lr": 3.67838415723505e-05}
{"train_loss": 0.0568542554974556, "global_step": 159046, "epoch": 1787, "lr": 3.678328245545072e-05}
{"train_loss": 0.1347471922636032, "global_step": 159047, "epoch": 1787, "lr": 3.6782723340327774e-05}
{"train_loss": 0.11482487618923187, "global_step": 159048, "epoch": 1787, "lr": 3.6782164226981726e-05}
{"train_loss": 0.10842367261648178, "global_step": 159049, "epoch": 1787, "lr": 3.678160511541265e-05}
{"train_loss": 0.10886084288358688, "global_step": 159050, "epoch": 1787, "lr": 3.678104600562063e-05}
{"train_loss": 0.10099844634532928, "global_step": 159051, "epoch": 1787, "lr": 3.6780486897605754e-05}
{"train_loss": 0.0638139545917511, "global_step": 159052, "epoch": 1787, "lr": 3.677992779136805e-05}
{"train_loss": 0.06629079580307007, "global_step": 159053, "epoch": 1787, "lr": 3.677936868690763e-05}
{"train_loss": 0.1134810745716095, "global_step": 159054, "epoch": 1787, "lr": 3.677880958422458e-05}
{"train_loss": 0.10532327741384506, "global_step": 159055, "epoch": 1787, "lr": 3.677825048331894e-05}
{"train_loss": 0.11193902045488358, "global_step": 159056, "epoch": 1787, "lr": 3.677769138419083e-05}
{"train_loss": 0.1473713219165802, "global_step": 159057, "epoch": 1787, "lr": 3.6777132286840265e-05}
{"train_loss": 0.10138821601867676, "global_step": 159058, "epoch": 1787, "lr": 3.677657319126738e-05}
{"train_loss": 0.08642854541540146, "global_step": 159059, "epoch": 1787, "lr": 3.67760140974722e-05}
{"train_loss": 0.0620395801961422, "global_step": 159060, "epoch": 1787, "lr": 3.677545500545485e-05}
{"train_loss": 0.11349726468324661, "global_step": 159061, "epoch": 1787, "lr": 3.677489591521535e-05}
{"train_loss": 0.06090496852993965, "global_step": 159062, "epoch": 1787, "lr": 3.677433682675383e-05}
{"train_loss": 0.10794194787740707, "global_step": 159063, "epoch": 1787, "lr": 3.677377774007032e-05}
{"train_loss": 0.12002846598625183, "global_step": 159064, "epoch": 1787, "lr": 3.677321865516493e-05}
{"train_loss": 0.09381496906280518, "global_step": 159065, "epoch": 1787, "lr": 3.677265957203771e-05}
{"train_loss": 0.09749926626682281, "global_step": 159066, "epoch": 1787, "lr": 3.6772100490688754e-05}
{"train_loss": 0.10514145344495773, "global_step": 159067, "epoch": 1787, "lr": 3.677154141111812e-05}
{"train_loss": 0.07794634997844696, "global_step": 159068, "epoch": 1787, "lr": 3.6770982333325884e-05}
{"train_loss": 0.09747426956892014, "global_step": 159069, "epoch": 1787, "lr": 3.677042325731214e-05}
{"train_loss": 0.06665349751710892, "global_step": 159070, "epoch": 1787, "lr": 3.6769864183076966e-05}
{"train_loss": 0.0908016785979271, "global_step": 159071, "epoch": 1787, "lr": 3.676930511062039e-05}
{"train_loss": 0.07961633056402206, "global_step": 159072, "epoch": 1787, "lr": 3.676874603994252e-05}
{"train_loss": 0.09799943119287491, "global_step": 159073, "epoch": 1787, "lr": 3.676818697104345e-05}
{"train_loss": 0.10071014612913132, "global_step": 159074, "epoch": 1787, "lr": 3.676762790392322e-05}
{"train_loss": 0.08347315341234207, "global_step": 159075, "epoch": 1787, "lr": 3.676706883858194e-05}
{"train_loss": 0.07588580250740051, "global_step": 159076, "epoch": 1787, "lr": 3.6766509775019655e-05}
{"train_loss": 0.12202846258878708, "global_step": 159077, "epoch": 1787, "lr": 3.6765950713236466e-05}
{"train_loss": 0.07534768432378769, "global_step": 159078, "epoch": 1787, "lr": 3.6765391653232406e-05}
{"train_loss": 0.11051064729690552, "global_step": 159079, "epoch": 1787, "lr": 3.67648325950076e-05}
{"train_loss": 0.07183166593313217, "global_step": 159080, "epoch": 1787, "lr": 3.676427353856209e-05}
{"train_loss": 0.09373339265584946, "global_step": 159081, "epoch": 1787, "lr": 3.676371448389598e-05}
{"train_loss": 0.08687259256839752, "global_step": 159082, "epoch": 1787, "lr": 3.67631554310093e-05}
{"train_loss": 0.03705539554357529, "global_step": 159083, "epoch": 1787, "lr": 3.676259637990217e-05}
{"train_loss": 0.1367933452129364, "global_step": 159084, "epoch": 1787, "lr": 3.676203733057465e-05}
{"train_loss": 0.06360208988189697, "global_step": 159085, "epoch": 1787, "lr": 3.6761478283026806e-05}
{"train_loss": 0.06963172554969788, "global_step": 159086, "epoch": 1787, "lr": 3.676091923725872e-05}
{"train_loss": 0.087009496986866, "global_step": 159087, "epoch": 1787, "lr": 3.6760360193270474e-05}
{"train_loss": 0.08507779240608215, "global_step": 159088, "epoch": 1787, "lr": 3.675980115106212e-05}
{"train_loss": 0.0811496451497078, "global_step": 159089, "epoch": 1787, "lr": 3.675924211063375e-05}
{"train_loss": 0.1696517914533615, "global_step": 159090, "epoch": 1787, "lr": 3.675868307198546e-05}
{"train_loss": 0.21965831518173218, "global_step": 159091, "epoch": 1787, "lr": 3.675812403511728e-05}
{"train_loss": 0.08984041213989258, "global_step": 159092, "epoch": 1787, "lr": 3.6757565000029334e-05}
{"train_loss": 0.11520341783761978, "global_step": 159093, "epoch": 1787, "lr": 3.6757005966721645e-05}
{"train_loss": 0.10865194350481033, "global_step": 159094, "epoch": 1787, "lr": 3.6756446935194344e-05}
{"train_loss": 0.08317295461893082, "global_step": 159095, "epoch": 1787, "lr": 3.675588790544745e-05}
{"train_loss": 0.09063069522380829, "global_step": 159096, "epoch": 1787, "lr": 3.6755328877481085e-05}
{"train_loss": 0.08410186320543289, "global_step": 159097, "epoch": 1787, "lr": 3.675476985129528e-05}
{"train_loss": 0.13366524875164032, "global_step": 159098, "epoch": 1787, "lr": 3.675421082689017e-05}
{"train_loss": 0.09858909249305725, "global_step": 159099, "epoch": 1787, "lr": 3.6753651804265765e-05}
{"train_loss": 0.14723826944828033, "global_step": 159100, "epoch": 1787, "lr": 3.675309278342219e-05}
{"train_loss": 0.18298247456550598, "global_step": 159101, "epoch": 1787, "lr": 3.6752533764359483e-05}
{"train_loss": 0.11800834536552429, "global_step": 159102, "epoch": 1787, "lr": 3.675197474707775e-05}
{"train_loss": 0.12248321622610092, "global_step": 159103, "epoch": 1787, "lr": 3.6751415731577054e-05}
{"train_loss": 0.1162232980132103, "global_step": 159104, "epoch": 1787, "lr": 3.675085671785746e-05}
{"train_loss": 0.09785022586584091, "global_step": 159105, "epoch": 1787, "lr": 3.675029770591905e-05}
{"train_loss": 0.07625788450241089, "global_step": 159106, "epoch": 1787, "lr": 3.6749738695761905e-05}
{"train_loss": 0.09053832292556763, "global_step": 159107, "epoch": 1787, "lr": 3.67491796873861e-05}
{"train_loss": 0.12140300124883652, "global_step": 159108, "epoch": 1787, "lr": 3.674862068079169e-05}
{"train_loss": 0.10996048152446747, "global_step": 159109, "epoch": 1787, "lr": 3.674806167597879e-05}
{"train_loss": 0.1582046002149582, "global_step": 159110, "epoch": 1787, "lr": 3.6747502672947435e-05}
{"train_loss": 0.099818155169487, "global_step": 159111, "epoch": 1787, "lr": 3.674694367169773e-05}
{"train_loss": 0.11316212266683578, "global_step": 159112, "epoch": 1787, "lr": 3.6746384672229715e-05}
{"train_loss": 0.06121089309453964, "global_step": 159113, "epoch": 1787, "lr": 3.674582567454351e-05}
{"train_loss": 0.11664029955863953, "global_step": 159114, "epoch": 1787, "lr": 3.674526667863915e-05}
{"train_loss": 0.08692137897014618, "global_step": 159115, "epoch": 1787, "lr": 3.674470768451675e-05}
{"train_loss": 0.09494659304618835, "global_step": 159116, "epoch": 1787, "lr": 3.6744148692176335e-05}
{"train_loss": 0.15919886529445648, "global_step": 159117, "epoch": 1787, "lr": 3.674358970161803e-05}
{"train_loss": 0.11482588201761246, "global_step": 159118, "epoch": 1787, "lr": 3.674303071284187e-05}
{"train_loss": 0.07291451841592789, "global_step": 159119, "epoch": 1787, "lr": 3.674247172584796e-05}
{"train_loss": 0.10678592324256897, "global_step": 159120, "epoch": 1787, "lr": 3.6741912740636354e-05}
{"train_loss": 0.07331068813800812, "global_step": 159121, "epoch": 1787, "lr": 3.6741353757207155e-05}
{"train_loss": 0.16423024237155914, "global_step": 159122, "epoch": 1787, "lr": 3.6740794775560405e-05}
{"train_loss": 0.09687130898237228, "global_step": 159123, "epoch": 1787, "lr": 3.674023579569619e-05}
{"train_loss": 0.05342208594083786, "global_step": 159124, "epoch": 1787, "lr": 3.6739676817614596e-05}
{"train_loss": 0.07948693633079529, "global_step": 159125, "epoch": 1787, "lr": 3.673911784131567e-05}
{"train_loss": 0.10790760815143585, "global_step": 159126, "epoch": 1787, "lr": 3.6738558866799545e-05}
{"train_loss": 0.060810089111328125, "global_step": 159127, "epoch": 1787, "lr": 3.673799989406623e-05}
{"train_loss": 0.10600137710571289, "global_step": 159128, "epoch": 1787, "lr": 3.673744092311585e-05}
{"train_loss": 0.07672739773988724, "global_step": 159129, "epoch": 1787, "lr": 3.673688195394844e-05}
{"train_loss": 0.11499382555484772, "global_step": 159130, "epoch": 1787, "lr": 3.673632298656411e-05}
{"train_loss": 0.10057944013329034, "global_step": 159131, "epoch": 1787, "lr": 3.67357640209629e-05, "val_loss": 6.737691402435303}
{"train_loss": 0.14982649683952332, "global_step": 159132, "epoch": 1788, "lr": 3.6735205057144936e-05}
{"train_loss": 0.0708715096116066, "global_step": 159133, "epoch": 1788, "lr": 3.6734646095110224e-05}
{"train_loss": 0.08300577849149704, "global_step": 159134, "epoch": 1788, "lr": 3.6734087134858916e-05}
{"train_loss": 0.1120728999376297, "global_step": 159135, "epoch": 1788, "lr": 3.673352817639102e-05}
{"train_loss": 0.0952056273818016, "global_step": 159136, "epoch": 1788, "lr": 3.6732969219706656e-05}
{"train_loss": 0.05126432701945305, "global_step": 159137, "epoch": 1788, "lr": 3.673241026480586e-05}
{"train_loss": 0.10404010117053986, "global_step": 159138, "epoch": 1788, "lr": 3.673185131168876e-05}
{"train_loss": 0.08131101727485657, "global_step": 159139, "epoch": 1788, "lr": 3.673129236035537e-05}
{"train_loss": 0.06634747982025146, "global_step": 159140, "epoch": 1788, "lr": 3.673073341080583e-05}
{"train_loss": 0.045526448637247086, "global_step": 159141, "epoch": 1788, "lr": 3.673017446304016e-05}
{"train_loss": 0.08594495058059692, "global_step": 159142, "epoch": 1788, "lr": 3.6729615517058465e-05}
{"train_loss": 0.11290909349918365, "global_step": 159143, "epoch": 1788, "lr": 3.67290565728608e-05}
{"train_loss": 0.08226723223924637, "global_step": 159144, "epoch": 1788, "lr": 3.672849763044725e-05}
{"train_loss": 0.07651275396347046, "global_step": 159145, "epoch": 1788, "lr": 3.672793868981791e-05}
{"train_loss": 0.09269271045923233, "global_step": 159146, "epoch": 1788, "lr": 3.672737975097281e-05}
{"train_loss": 0.08816360682249069, "global_step": 159147, "epoch": 1788, "lr": 3.6726820813912086e-05}
{"train_loss": 0.08756410330533981, "global_step": 159148, "epoch": 1788, "lr": 3.672626187863575e-05}
{"train_loss": 0.08256026357412338, "global_step": 159149, "epoch": 1788, "lr": 3.6725702945143925e-05}
{"train_loss": 0.11726715415716171, "global_step": 159150, "epoch": 1788, "lr": 3.672514401343665e-05}
{"train_loss": 0.11414085328578949, "global_step": 159151, "epoch": 1788, "lr": 3.672458508351404e-05}
{"train_loss": 0.10911357402801514, "global_step": 159152, "epoch": 1788, "lr": 3.6724026155376126e-05}
{"train_loss": 0.1372213214635849, "global_step": 159153, "epoch": 1788, "lr": 3.672346722902302e-05}
{"train_loss": 0.07666783034801483, "global_step": 159154, "epoch": 1788, "lr": 3.6722908304454774e-05}
{"train_loss": 0.18138940632343292, "global_step": 159155, "epoch": 1788, "lr": 3.672234938167148e-05}
{"train_loss": 0.1808534860610962, "global_step": 159156, "epoch": 1788, "lr": 3.672179046067319e-05}
{"train_loss": 0.12231400609016418, "global_step": 159157, "epoch": 1788, "lr": 3.672123154146001e-05}
{"train_loss": 0.10194069147109985, "global_step": 159158, "epoch": 1788, "lr": 3.672067262403198e-05}
{"train_loss": 0.08813793957233429, "global_step": 159159, "epoch": 1788, "lr": 3.6720113708389215e-05}
{"train_loss": 0.09932269155979156, "global_step": 159160, "epoch": 1788, "lr": 3.6719554794531745e-05}
{"train_loss": 0.12988793849945068, "global_step": 159161, "epoch": 1788, "lr": 3.6718995882459685e-05}
{"train_loss": 0.0888185128569603, "global_step": 159162, "epoch": 1788, "lr": 3.6718436972173096e-05}
{"train_loss": 0.04505462571978569, "global_step": 159163, "epoch": 1788, "lr": 3.671787806367204e-05}
{"train_loss": 0.09208600223064423, "global_step": 159164, "epoch": 1788, "lr": 3.671731915695662e-05}
{"train_loss": 0.12092071771621704, "global_step": 159165, "epoch": 1788, "lr": 3.671676025202687e-05}
{"train_loss": 0.08143524825572968, "global_step": 159166, "epoch": 1788, "lr": 3.671620134888292e-05}
{"train_loss": 0.06458877772092819, "global_step": 159167, "epoch": 1788, "lr": 3.671564244752478e-05}
{"train_loss": 0.09239228069782257, "global_step": 159168, "epoch": 1788, "lr": 3.671508354795259e-05}
{"train_loss": 0.09376703947782516, "global_step": 159169, "epoch": 1788, "lr": 3.671452465016638e-05}
{"train_loss": 0.0846245139837265, "global_step": 159170, "epoch": 1788, "lr": 3.6713965754166246e-05}
{"train_loss": 0.1014571562409401, "global_step": 159171, "epoch": 1788, "lr": 3.671340685995225e-05}
{"train_loss": 0.07077706605195999, "global_step": 159172, "epoch": 1788, "lr": 3.671284796752449e-05}
{"train_loss": 0.11823948472738266, "global_step": 159173, "epoch": 1788, "lr": 3.6712289076883e-05}
{"train_loss": 0.09026060253381729, "global_step": 159174, "epoch": 1788, "lr": 3.671173018802791e-05}
{"train_loss": 0.13036899268627167, "global_step": 159175, "epoch": 1788, "lr": 3.671117130095924e-05}
{"train_loss": 0.10136719793081284, "global_step": 159176, "epoch": 1788, "lr": 3.6710612415677116e-05}
{"train_loss": 0.10303527861833572, "global_step": 159177, "epoch": 1788, "lr": 3.671005353218156e-05}
{"train_loss": 0.07311698794364929, "global_step": 159178, "epoch": 1788, "lr": 3.6709494650472706e-05}
{"train_loss": 0.11346910148859024, "global_step": 159179, "epoch": 1788, "lr": 3.670893577055057e-05}
{"train_loss": 0.13062773644924164, "global_step": 159180, "epoch": 1788, "lr": 3.670837689241527e-05}
{"train_loss": 0.1664760410785675, "global_step": 159181, "epoch": 1788, "lr": 3.670781801606688e-05}
{"train_loss": 0.1394224464893341, "global_step": 159182, "epoch": 1788, "lr": 3.6707259141505434e-05}
{"train_loss": 0.09828945249319077, "global_step": 159183, "epoch": 1788, "lr": 3.6706700268731054e-05}
{"train_loss": 0.12277165800333023, "global_step": 159184, "epoch": 1788, "lr": 3.6706141397743785e-05}
{"train_loss": 0.043185099959373474, "global_step": 159185, "epoch": 1788, "lr": 3.670558252854373e-05}
{"train_loss": 0.13395237922668457, "global_step": 159186, "epoch": 1788, "lr": 3.6705023661130924e-05}
{"train_loss": 0.08787158131599426, "global_step": 159187, "epoch": 1788, "lr": 3.670446479550548e-05}
{"train_loss": 0.12482519447803497, "global_step": 159188, "epoch": 1788, "lr": 3.670390593166745e-05}
{"train_loss": 0.13432827591896057, "global_step": 159189, "epoch": 1788, "lr": 3.670334706961693e-05}
{"train_loss": 0.08728381246328354, "global_step": 159190, "epoch": 1788, "lr": 3.670278820935397e-05}
{"train_loss": 0.04997624084353447, "global_step": 159191, "epoch": 1788, "lr": 3.670222935087868e-05}
{"train_loss": 0.10274692624807358, "global_step": 159192, "epoch": 1788, "lr": 3.670167049419108e-05}
{"train_loss": 0.12454558908939362, "global_step": 159193, "epoch": 1788, "lr": 3.670111163929131e-05}
{"train_loss": 0.059626247733831406, "global_step": 159194, "epoch": 1788, "lr": 3.670055278617939e-05}
{"train_loss": 0.09894472360610962, "global_step": 159195, "epoch": 1788, "lr": 3.6699993934855436e-05}
{"train_loss": 0.13044323027133942, "global_step": 159196, "epoch": 1788, "lr": 3.6699435085319486e-05}
{"train_loss": 0.0983964055776596, "global_step": 159197, "epoch": 1788, "lr": 3.6698876237571664e-05}
{"train_loss": 0.09172438085079193, "global_step": 159198, "epoch": 1788, "lr": 3.669831739161199e-05}
{"train_loss": 0.09962323307991028, "global_step": 159199, "epoch": 1788, "lr": 3.669775854744057e-05}
{"train_loss": 0.12990355491638184, "global_step": 159200, "epoch": 1788, "lr": 3.669719970505748e-05}
{"train_loss": 0.06405191868543625, "global_step": 159201, "epoch": 1788, "lr": 3.6696640864462784e-05}
{"train_loss": 0.0609082467854023, "global_step": 159202, "epoch": 1788, "lr": 3.6696082025656574e-05}
{"train_loss": 0.09078327566385269, "global_step": 159203, "epoch": 1788, "lr": 3.66955231886389e-05}
{"train_loss": 0.08643440157175064, "global_step": 159204, "epoch": 1788, "lr": 3.669496435340987e-05}
{"train_loss": 0.10653558373451233, "global_step": 159205, "epoch": 1788, "lr": 3.669440551996951e-05}
{"train_loss": 0.09224718064069748, "global_step": 159206, "epoch": 1788, "lr": 3.669384668831796e-05}
{"train_loss": 0.05016893520951271, "global_step": 159207, "epoch": 1788, "lr": 3.669328785845522e-05}
{"train_loss": 0.08464239537715912, "global_step": 159208, "epoch": 1788, "lr": 3.669272903038144e-05}
{"train_loss": 0.09317158907651901, "global_step": 159209, "epoch": 1788, "lr": 3.669217020409664e-05}
{"train_loss": 0.08831196278333664, "global_step": 159210, "epoch": 1788, "lr": 3.669161137960093e-05}
{"train_loss": 0.13650023937225342, "global_step": 159211, "epoch": 1788, "lr": 3.6691052556894355e-05}
{"train_loss": 0.07499227672815323, "global_step": 159212, "epoch": 1788, "lr": 3.669049373597702e-05}
{"train_loss": 0.13237422704696655, "global_step": 159213, "epoch": 1788, "lr": 3.668993491684897e-05}
{"train_loss": 0.061467040330171585, "global_step": 159214, "epoch": 1788, "lr": 3.668937609951032e-05}
{"train_loss": 0.11188700795173645, "global_step": 159215, "epoch": 1788, "lr": 3.6688817283961094e-05}
{"train_loss": 0.08583281189203262, "global_step": 159216, "epoch": 1788, "lr": 3.668825847020141e-05}
{"train_loss": 0.06558889895677567, "global_step": 159217, "epoch": 1788, "lr": 3.6687699658231324e-05}
{"train_loss": 0.11546747386455536, "global_step": 159218, "epoch": 1788, "lr": 3.668714084805092e-05}
{"train_loss": 0.09085502475500107, "global_step": 159219, "epoch": 1788, "lr": 3.6686582039660264e-05}
{"train_loss": 0.09773704158456138, "global_step": 159220, "epoch": 1788, "lr": 3.668602323305943e-05, "val_loss": 6.828068256378174}
{"train_loss": 0.1372256726026535, "global_step": 159221, "epoch": 1789, "lr": 3.668546442824851e-05}
{"train_loss": 0.10663529485464096, "global_step": 159222, "epoch": 1789, "lr": 3.668490562522755e-05}
{"train_loss": 0.13786345720291138, "global_step": 159223, "epoch": 1789, "lr": 3.6684346823996656e-05}
{"train_loss": 0.10686904937028885, "global_step": 159224, "epoch": 1789, "lr": 3.6683788024555875e-05}
{"train_loss": 0.17194372415542603, "global_step": 159225, "epoch": 1789, "lr": 3.6683229226905314e-05}
{"train_loss": 0.10185565054416656, "global_step": 159226, "epoch": 1789, "lr": 3.6682670431045016e-05}
{"train_loss": 0.09811004996299744, "global_step": 159227, "epoch": 1789, "lr": 3.668211163697509e-05}
{"train_loss": 0.12603206932544708, "global_step": 159228, "epoch": 1789, "lr": 3.668155284469556e-05}
{"train_loss": 0.12172200530767441, "global_step": 159229, "epoch": 1789, "lr": 3.6680994054206565e-05}
{"train_loss": 0.06466732174158096, "global_step": 159230, "epoch": 1789, "lr": 3.668043526550812e-05}
{"train_loss": 0.06411309540271759, "global_step": 159231, "epoch": 1789, "lr": 3.6679876478600356e-05}
{"train_loss": 0.11886173486709595, "global_step": 159232, "epoch": 1789, "lr": 3.6679317693483295e-05}
{"train_loss": 0.06883179396390915, "global_step": 159233, "epoch": 1789, "lr": 3.6678758910157065e-05}
{"train_loss": 0.11855580657720566, "global_step": 159234, "epoch": 1789, "lr": 3.6678200128621686e-05}
{"train_loss": 0.14898478984832764, "global_step": 159235, "epoch": 1789, "lr": 3.667764134887727e-05}
{"train_loss": 0.10174179822206497, "global_step": 159236, "epoch": 1789, "lr": 3.667708257092388e-05}
{"train_loss": 0.1308564841747284, "global_step": 159237, "epoch": 1789, "lr": 3.6676523794761605e-05}
{"train_loss": 0.0777963399887085, "global_step": 159238, "epoch": 1789, "lr": 3.667596502039051e-05}
{"train_loss": 0.11347750574350357, "global_step": 159239, "epoch": 1789, "lr": 3.667540624781065e-05}
{"train_loss": 0.09983275830745697, "global_step": 159240, "epoch": 1789, "lr": 3.667484747702215e-05}
{"train_loss": 0.05234362185001373, "global_step": 159241, "epoch": 1789, "lr": 3.667428870802502e-05}
{"train_loss": 0.049990132451057434, "global_step": 159242, "epoch": 1789, "lr": 3.66737299408194e-05}
{"train_loss": 0.08555704355239868, "global_step": 159243, "epoch": 1789, "lr": 3.6673171175405304e-05}
{"train_loss": 0.09828363358974457, "global_step": 159244, "epoch": 1789, "lr": 3.667261241178286e-05}
{"train_loss": 0.10778117924928665, "global_step": 159245, "epoch": 1789, "lr": 3.6672053649952106e-05}
{"train_loss": 0.1325199007987976, "global_step": 159246, "epoch": 1789, "lr": 3.667149488991315e-05}
{"train_loss": 0.12629489600658417, "global_step": 159247, "epoch": 1789, "lr": 3.667093613166603e-05}
{"train_loss": 0.14549535512924194, "global_step": 159248, "epoch": 1789, "lr": 3.667037737521086e-05}
{"train_loss": 0.08930575847625732, "global_step": 159249, "epoch": 1789, "lr": 3.666981862054767e-05}
{"train_loss": 0.09327088296413422, "global_step": 159250, "epoch": 1789, "lr": 3.666925986767659e-05}
{"train_loss": 0.05496634542942047, "global_step": 159251, "epoch": 1789, "lr": 3.6668701116597635e-05}
{"train_loss": 0.060875508934259415, "global_step": 159252, "epoch": 1789, "lr": 3.666814236731093e-05}
{"train_loss": 0.1131390631198883, "global_step": 159253, "epoch": 1789, "lr": 3.666758361981652e-05}
{"train_loss": 0.08085279911756516, "global_step": 159254, "epoch": 1789, "lr": 3.66670248741145e-05}
{"train_loss": 0.09367674589157104, "global_step": 159255, "epoch": 1789, "lr": 3.6666466130204935e-05}
{"train_loss": 0.04796808212995529, "global_step": 159256, "epoch": 1789, "lr": 3.666590738808791e-05}
{"train_loss": 0.14465390145778656, "global_step": 159257, "epoch": 1789, "lr": 3.666534864776349e-05}
{"train_loss": 0.13339129090309143, "global_step": 159258, "epoch": 1789, "lr": 3.666478990923172e-05}
{"train_loss": 0.052903804928064346, "global_step": 159259, "epoch": 1789, "lr": 3.666423117249275e-05}
{"train_loss": 0.08817694336175919, "global_step": 159260, "epoch": 1789, "lr": 3.6663672437546574e-05}
{"train_loss": 0.07459145039319992, "global_step": 159261, "epoch": 1789, "lr": 3.666311370439334e-05}
{"train_loss": 0.0794903114438057, "global_step": 159262, "epoch": 1789, "lr": 3.666255497303305e-05}
{"train_loss": 0.10725447535514832, "global_step": 159263, "epoch": 1789, "lr": 3.666199624346584e-05}
{"train_loss": 0.10354398190975189, "global_step": 159264, "epoch": 1789, "lr": 3.666143751569175e-05}
{"train_loss": 0.0878504291176796, "global_step": 159265, "epoch": 1789, "lr": 3.6660878789710894e-05}
{"train_loss": 0.09714744240045547, "global_step": 159266, "epoch": 1789, "lr": 3.666032006552329e-05}
{"train_loss": 0.04423754662275314, "global_step": 159267, "epoch": 1789, "lr": 3.665976134312906e-05}
{"train_loss": 0.09874669462442398, "global_step": 159268, "epoch": 1789, "lr": 3.665920262252825e-05}
{"train_loss": 0.05570346862077713, "global_step": 159269, "epoch": 1789, "lr": 3.6658643903720965e-05}
{"train_loss": 0.11321119219064713, "global_step": 159270, "epoch": 1789, "lr": 3.6658085186707244e-05}
{"train_loss": 0.12120236456394196, "global_step": 159271, "epoch": 1789, "lr": 3.665752647148718e-05}
{"train_loss": 0.14466270804405212, "global_step": 159272, "epoch": 1789, "lr": 3.6656967758060865e-05}
{"train_loss": 0.06648586690425873, "global_step": 159273, "epoch": 1789, "lr": 3.665640904642835e-05}
{"train_loss": 0.07718345522880554, "global_step": 159274, "epoch": 1789, "lr": 3.665585033658972e-05}
{"train_loss": 0.11502891778945923, "global_step": 159275, "epoch": 1789, "lr": 3.665529162854505e-05}
{"train_loss": 0.08144716173410416, "global_step": 159276, "epoch": 1789, "lr": 3.665473292229441e-05}
{"train_loss": 0.08521812409162521, "global_step": 159277, "epoch": 1789, "lr": 3.6654174217837874e-05}
{"train_loss": 0.10677681863307953, "global_step": 159278, "epoch": 1789, "lr": 3.665361551517554e-05}
{"train_loss": 0.11526905000209808, "global_step": 159279, "epoch": 1789, "lr": 3.6653056814307446e-05}
{"train_loss": 0.09504179656505585, "global_step": 159280, "epoch": 1789, "lr": 3.66524981152337e-05}
{"train_loss": 0.10427320003509521, "global_step": 159281, "epoch": 1789, "lr": 3.665193941795435e-05}
{"train_loss": 0.08126989752054214, "global_step": 159282, "epoch": 1789, "lr": 3.66513807224695e-05}
{"train_loss": 0.1173994168639183, "global_step": 159283, "epoch": 1789, "lr": 3.665082202877918e-05}
{"train_loss": 0.09197815507650375, "global_step": 159284, "epoch": 1789, "lr": 3.6650263336883525e-05}
{"train_loss": 0.059587590396404266, "global_step": 159285, "epoch": 1789, "lr": 3.6649704646782554e-05}
{"train_loss": 0.06614900380373001, "global_step": 159286, "epoch": 1789, "lr": 3.664914595847639e-05}
{"train_loss": 0.08533712476491928, "global_step": 159287, "epoch": 1789, "lr": 3.664858727196507e-05}
{"train_loss": 0.14577847719192505, "global_step": 159288, "epoch": 1789, "lr": 3.6648028587248676e-05}
{"train_loss": 0.10145729780197144, "global_step": 159289, "epoch": 1789, "lr": 3.664746990432732e-05}
{"train_loss": 0.10791824758052826, "global_step": 159290, "epoch": 1789, "lr": 3.664691122320102e-05}
{"train_loss": 0.08911899477243423, "global_step": 159291, "epoch": 1789, "lr": 3.664635254386991e-05}
{"train_loss": 0.09576107561588287, "global_step": 159292, "epoch": 1789, "lr": 3.6645793866334024e-05}
{"train_loss": 0.09788352251052856, "global_step": 159293, "epoch": 1789, "lr": 3.664523519059344e-05}
{"train_loss": 0.1177152544260025, "global_step": 159294, "epoch": 1789, "lr": 3.664467651664825e-05}
{"train_loss": 0.06630662828683853, "global_step": 159295, "epoch": 1789, "lr": 3.664411784449851e-05}
{"train_loss": 0.04785934463143349, "global_step": 159296, "epoch": 1789, "lr": 3.664355917414431e-05}
{"train_loss": 0.06816233694553375, "global_step": 159297, "epoch": 1789, "lr": 3.664300050558574e-05}
{"train_loss": 0.06167249754071236, "global_step": 159298, "epoch": 1789, "lr": 3.6642441838822825e-05}
{"train_loss": 0.09831863641738892, "global_step": 159299, "epoch": 1789, "lr": 3.6641883173855693e-05}
{"train_loss": 0.10048893839120865, "global_step": 159300, "epoch": 1789, "lr": 3.664132451068438e-05}
{"train_loss": 0.10308362543582916, "global_step": 159301, "epoch": 1789, "lr": 3.6640765849309e-05}
{"train_loss": 0.06874304264783859, "global_step": 159302, "epoch": 1789, "lr": 3.664020718972958e-05}
{"train_loss": 0.05786697566509247, "global_step": 159303, "epoch": 1789, "lr": 3.6639648531946244e-05}
{"train_loss": 0.1011272519826889, "global_step": 159304, "epoch": 1789, "lr": 3.6639089875959024e-05}
{"train_loss": 0.12620437145233154, "global_step": 159305, "epoch": 1789, "lr": 3.663853122176804e-05}
{"train_loss": 0.11251053214073181, "global_step": 159306, "epoch": 1789, "lr": 3.663797256937332e-05}
{"train_loss": 0.09769325703382492, "global_step": 159307, "epoch": 1789, "lr": 3.6637413918774955e-05}
{"train_loss": 0.1002003401517868, "global_step": 159308, "epoch": 1789, "lr": 3.663685526997306e-05}
{"train_loss": 0.09660733633496787, "global_step": 159309, "epoch": 1789, "lr": 3.663629662296765e-05, "val_loss": 6.801963806152344}
{"train_loss": 0.07406234741210938, "global_step": 159310, "epoch": 1790, "lr": 3.663573797775884e-05}
{"train_loss": 0.13365159928798676, "global_step": 159311, "epoch": 1790, "lr": 3.663517933434669e-05}
{"train_loss": 0.06498408317565918, "global_step": 159312, "epoch": 1790, "lr": 3.663462069273128e-05}
{"train_loss": 0.08989395946264267, "global_step": 159313, "epoch": 1790, "lr": 3.663406205291267e-05}
{"train_loss": 0.060155220329761505, "global_step": 159314, "epoch": 1790, "lr": 3.6633503414890957e-05}
{"train_loss": 0.07501507550477982, "global_step": 159315, "epoch": 1790, "lr": 3.663294477866621e-05}
{"train_loss": 0.12353615462779999, "global_step": 159316, "epoch": 1790, "lr": 3.66323861442385e-05}
{"train_loss": 0.07949142903089523, "global_step": 159317, "epoch": 1790, "lr": 3.6631827511607884e-05}
{"train_loss": 0.08006066828966141, "global_step": 159318, "epoch": 1790, "lr": 3.6631268880774486e-05}
{"train_loss": 0.12566310167312622, "global_step": 159319, "epoch": 1790, "lr": 3.663071025173832e-05}
{"train_loss": 0.07434646040201187, "global_step": 159320, "epoch": 1790, "lr": 3.663015162449952e-05}
{"train_loss": 0.14650759100914001, "global_step": 159321, "epoch": 1790, "lr": 3.662959299905811e-05}
{"train_loss": 0.07763572037220001, "global_step": 159322, "epoch": 1790, "lr": 3.662903437541421e-05}
{"train_loss": 0.10836157202720642, "global_step": 159323, "epoch": 1790, "lr": 3.662847575356785e-05}
{"train_loss": 0.05678824707865715, "global_step": 159324, "epoch": 1790, "lr": 3.6627917133519154e-05}
{"train_loss": 0.06276197731494904, "global_step": 159325, "epoch": 1790, "lr": 3.6627358515268146e-05}
{"train_loss": 0.1375967413187027, "global_step": 159326, "epoch": 1790, "lr": 3.662679989881493e-05}
{"train_loss": 0.08067288994789124, "global_step": 159327, "epoch": 1790, "lr": 3.6626241284159595e-05}
{"train_loss": 0.11468011140823364, "global_step": 159328, "epoch": 1790, "lr": 3.662568267130218e-05}
{"train_loss": 0.14355194568634033, "global_step": 159329, "epoch": 1790, "lr": 3.66251240602428e-05}
{"train_loss": 0.1429913192987442, "global_step": 159330, "epoch": 1790, "lr": 3.6624565450981496e-05}
{"train_loss": 0.142784982919693, "global_step": 159331, "epoch": 1790, "lr": 3.662400684351836e-05}
{"train_loss": 0.11347119510173798, "global_step": 159332, "epoch": 1790, "lr": 3.662344823785346e-05}
{"train_loss": 0.11307167261838913, "global_step": 159333, "epoch": 1790, "lr": 3.6622889633986876e-05}
{"train_loss": 0.07956396043300629, "global_step": 159334, "epoch": 1790, "lr": 3.662233103191868e-05}
{"train_loss": 0.07019523531198502, "global_step": 159335, "epoch": 1790, "lr": 3.6621772431648965e-05}
{"train_loss": 0.06827519834041595, "global_step": 159336, "epoch": 1790, "lr": 3.662121383317776e-05}
{"train_loss": 0.099395252764225, "global_step": 159337, "epoch": 1790, "lr": 3.6620655236505196e-05}
{"train_loss": 0.0787142738699913, "global_step": 159338, "epoch": 1790, "lr": 3.66200966416313e-05}
{"train_loss": 0.08090715110301971, "global_step": 159339, "epoch": 1790, "lr": 3.661953804855619e-05}
{"train_loss": 0.08067887276411057, "global_step": 159340, "epoch": 1790, "lr": 3.6618979457279904e-05}
{"train_loss": 0.13082605600357056, "global_step": 159341, "epoch": 1790, "lr": 3.6618420867802544e-05}
{"train_loss": 0.09242143481969833, "global_step": 159342, "epoch": 1790, "lr": 3.661786228012416e-05}
{"train_loss": 0.11960525810718536, "global_step": 159343, "epoch": 1790, "lr": 3.661730369424483e-05}
{"train_loss": 0.09804034233093262, "global_step": 159344, "epoch": 1790, "lr": 3.661674511016468e-05}
{"train_loss": 0.0794810801744461, "global_step": 159345, "epoch": 1790, "lr": 3.661618652788371e-05}
{"train_loss": 0.06278440356254578, "global_step": 159346, "epoch": 1790, "lr": 3.661562794740206e-05}
{"train_loss": 0.0614769384264946, "global_step": 159347, "epoch": 1790, "lr": 3.6615069368719754e-05}
{"train_loss": 0.08438336104154587, "global_step": 159348, "epoch": 1790, "lr": 3.6614510791836906e-05}
{"train_loss": 0.1349075436592102, "global_step": 159349, "epoch": 1790, "lr": 3.6613952216753564e-05}
{"train_loss": 0.13132774829864502, "global_step": 159350, "epoch": 1790, "lr": 3.6613393643469814e-05}
{"train_loss": 0.14710675179958344, "global_step": 159351, "epoch": 1790, "lr": 3.661283507198573e-05}
{"train_loss": 0.078633151948452, "global_step": 159352, "epoch": 1790, "lr": 3.661227650230138e-05}
{"train_loss": 0.05735567957162857, "global_step": 159353, "epoch": 1790, "lr": 3.661171793441686e-05}
{"train_loss": 0.06149515509605408, "global_step": 159354, "epoch": 1790, "lr": 3.661115936833224e-05}
{"train_loss": 0.07878723740577698, "global_step": 159355, "epoch": 1790, "lr": 3.6610600804047555e-05}
{"train_loss": 0.12241434305906296, "global_step": 159356, "epoch": 1790, "lr": 3.661004224156294e-05}
{"train_loss": 0.07861480861902237, "global_step": 159357, "epoch": 1790, "lr": 3.660948368087842e-05}
{"train_loss": 0.10314138233661652, "global_step": 159358, "epoch": 1790, "lr": 3.6608925121994106e-05}
{"train_loss": 0.06563139706850052, "global_step": 159359, "epoch": 1790, "lr": 3.6608366564910044e-05}
{"train_loss": 0.14759129285812378, "global_step": 159360, "epoch": 1790, "lr": 3.660780800962634e-05}
{"train_loss": 0.12520970404148102, "global_step": 159361, "epoch": 1790, "lr": 3.6607249456143036e-05}
{"train_loss": 0.08766555041074753, "global_step": 159362, "epoch": 1790, "lr": 3.6606690904460225e-05}
{"train_loss": 0.122691310942173, "global_step": 159363, "epoch": 1790, "lr": 3.6606132354578e-05}
{"train_loss": 0.09472313523292542, "global_step": 159364, "epoch": 1790, "lr": 3.6605573806496404e-05}
{"train_loss": 0.07074941694736481, "global_step": 159365, "epoch": 1790, "lr": 3.6605015260215535e-05}
{"train_loss": 0.08306396007537842, "global_step": 159366, "epoch": 1790, "lr": 3.6604456715735445e-05}
{"train_loss": 0.06166040897369385, "global_step": 159367, "epoch": 1790, "lr": 3.660389817305624e-05}
{"train_loss": 0.06279079616069794, "global_step": 159368, "epoch": 1790, "lr": 3.660333963217797e-05}
{"train_loss": 0.159795880317688, "global_step": 159369, "epoch": 1790, "lr": 3.6602781093100714e-05}
{"train_loss": 0.08079704642295837, "global_step": 159370, "epoch": 1790, "lr": 3.660222255582455e-05}
{"train_loss": 0.11150823533535004, "global_step": 159371, "epoch": 1790, "lr": 3.660166402034956e-05}
{"train_loss": 0.12137454748153687, "global_step": 159372, "epoch": 1790, "lr": 3.660110548667581e-05}
{"train_loss": 0.08188256621360779, "global_step": 159373, "epoch": 1790, "lr": 3.660054695480337e-05}
{"train_loss": 0.14817258715629578, "global_step": 159374, "epoch": 1790, "lr": 3.6599988424732334e-05}
{"train_loss": 0.12515854835510254, "global_step": 159375, "epoch": 1790, "lr": 3.6599429896462775e-05}
{"train_loss": 0.09441765397787094, "global_step": 159376, "epoch": 1790, "lr": 3.659887136999474e-05}
{"train_loss": 0.13993854820728302, "global_step": 159377, "epoch": 1790, "lr": 3.6598312845328335e-05}
{"train_loss": 0.09247205406427383, "global_step": 159378, "epoch": 1790, "lr": 3.6597754322463605e-05}
{"train_loss": 0.08071506023406982, "global_step": 159379, "epoch": 1790, "lr": 3.6597195801400655e-05}
{"train_loss": 0.07148580253124237, "global_step": 159380, "epoch": 1790, "lr": 3.6596637282139555e-05}
{"train_loss": 0.07987045496702194, "global_step": 159381, "epoch": 1790, "lr": 3.6596078764680365e-05}
{"train_loss": 0.04696718230843544, "global_step": 159382, "epoch": 1790, "lr": 3.659552024902318e-05}
{"train_loss": 0.1468401551246643, "global_step": 159383, "epoch": 1790, "lr": 3.659496173516805e-05}
{"train_loss": 0.14067378640174866, "global_step": 159384, "epoch": 1790, "lr": 3.659440322311508e-05}
{"train_loss": 0.17195744812488556, "global_step": 159385, "epoch": 1790, "lr": 3.659384471286431e-05}
{"train_loss": 0.07373741269111633, "global_step": 159386, "epoch": 1790, "lr": 3.659328620441585e-05}
{"train_loss": 0.06974688917398453, "global_step": 159387, "epoch": 1790, "lr": 3.659272769776974e-05}
{"train_loss": 0.07150604575872421, "global_step": 159388, "epoch": 1790, "lr": 3.6592169192926106e-05}
{"train_loss": 0.11622089892625809, "global_step": 159389, "epoch": 1790, "lr": 3.6591610689884976e-05}
{"train_loss": 0.08365313708782196, "global_step": 159390, "epoch": 1790, "lr": 3.659105218864643e-05}
{"train_loss": 0.09667875617742538, "global_step": 159391, "epoch": 1790, "lr": 3.659049368921057e-05}
{"train_loss": 0.1515807956457138, "global_step": 159392, "epoch": 1790, "lr": 3.658993519157745e-05}
{"train_loss": 0.060866452753543854, "global_step": 159393, "epoch": 1790, "lr": 3.6589376695747144e-05}
{"train_loss": 0.07934225350618362, "global_step": 159394, "epoch": 1790, "lr": 3.6588818201719745e-05}
{"train_loss": 0.13928577303886414, "global_step": 159395, "epoch": 1790, "lr": 3.65882597094953e-05}
{"train_loss": 0.13325533270835876, "global_step": 159396, "epoch": 1790, "lr": 3.658770121907391e-05}
{"train_loss": 0.10147689282894135, "global_step": 159397, "epoch": 1790, "lr": 3.658714273045563e-05}
{"train_loss": 0.0987279647019472, "global_step": 159398, "epoch": 1790, "lr": 3.658658424364054e-05, "val_loss": 6.65529727935791, "train_action_mse_error": 22.3009090423584}
{"train_loss": 0.14925137162208557, "global_step": 159399, "epoch": 1791, "lr": 3.658602575862874e-05}
{"train_loss": 0.11101388186216354, "global_step": 159400, "epoch": 1791, "lr": 3.658546727542027e-05}
{"train_loss": 0.15537545084953308, "global_step": 159401, "epoch": 1791, "lr": 3.6584908794015235e-05}
{"train_loss": 0.09265623241662979, "global_step": 159402, "epoch": 1791, "lr": 3.658435031441368e-05}
{"train_loss": 0.11614538729190826, "global_step": 159403, "epoch": 1791, "lr": 3.658379183661571e-05}
{"train_loss": 0.07937432825565338, "global_step": 159404, "epoch": 1791, "lr": 3.658323336062137e-05}
{"train_loss": 0.08016224950551987, "global_step": 159405, "epoch": 1791, "lr": 3.6582674886430774e-05}
{"train_loss": 0.2131236493587494, "global_step": 159406, "epoch": 1791, "lr": 3.658211641404395e-05}
{"train_loss": 0.1156768798828125, "global_step": 159407, "epoch": 1791, "lr": 3.658155794346101e-05}
{"train_loss": 0.06558860093355179, "global_step": 159408, "epoch": 1791, "lr": 3.658099947468202e-05}
{"train_loss": 0.10389002412557602, "global_step": 159409, "epoch": 1791, "lr": 3.658044100770704e-05}
{"train_loss": 0.12349256128072739, "global_step": 159410, "epoch": 1791, "lr": 3.657988254253616e-05}
{"train_loss": 0.10759618878364563, "global_step": 159411, "epoch": 1791, "lr": 3.6579324079169455e-05}
{"train_loss": 0.10304325073957443, "global_step": 159412, "epoch": 1791, "lr": 3.657876561760699e-05}
{"train_loss": 0.14267122745513916, "global_step": 159413, "epoch": 1791, "lr": 3.657820715784886e-05}
{"train_loss": 0.07833684235811234, "global_step": 159414, "epoch": 1791, "lr": 3.65776486998951e-05}
{"train_loss": 0.08966224640607834, "global_step": 159415, "epoch": 1791, "lr": 3.657709024374582e-05}
{"train_loss": 0.08780869096517563, "global_step": 159416, "epoch": 1791, "lr": 3.6576531789401106e-05}
{"train_loss": 0.042831066995859146, "global_step": 159417, "epoch": 1791, "lr": 3.6575973336860995e-05}
{"train_loss": 0.06749413907527924, "global_step": 159418, "epoch": 1791, "lr": 3.657541488612559e-05}
{"train_loss": 0.07059790194034576, "global_step": 159419, "epoch": 1791, "lr": 3.657485643719495e-05}
{"train_loss": 0.06835506856441498, "global_step": 159420, "epoch": 1791, "lr": 3.657429799006917e-05}
{"train_loss": 0.09976620227098465, "global_step": 159421, "epoch": 1791, "lr": 3.6573739544748286e-05}
{"train_loss": 0.14112642407417297, "global_step": 159422, "epoch": 1791, "lr": 3.657318110123244e-05}
{"train_loss": 0.07435882836580276, "global_step": 159423, "epoch": 1791, "lr": 3.657262265952162e-05}
{"train_loss": 0.1844983696937561, "global_step": 159424, "epoch": 1791, "lr": 3.657206421961598e-05}
{"train_loss": 0.05199792981147766, "global_step": 159425, "epoch": 1791, "lr": 3.657150578151554e-05}
{"train_loss": 0.1422674059867859, "global_step": 159426, "epoch": 1791, "lr": 3.657094734522042e-05}
{"train_loss": 0.07834932953119278, "global_step": 159427, "epoch": 1791, "lr": 3.6570388910730664e-05}
{"train_loss": 0.058084357529878616, "global_step": 159428, "epoch": 1791, "lr": 3.656983047804635e-05}
{"train_loss": 0.0491517148911953, "global_step": 159429, "epoch": 1791, "lr": 3.6569272047167564e-05}
{"train_loss": 0.12007011473178864, "global_step": 159430, "epoch": 1791, "lr": 3.656871361809437e-05}
{"train_loss": 0.0743970051407814, "global_step": 159431, "epoch": 1791, "lr": 3.656815519082686e-05}
{"train_loss": 0.08989109843969345, "global_step": 159432, "epoch": 1791, "lr": 3.6567596765365094e-05}
{"train_loss": 0.09674054384231567, "global_step": 159433, "epoch": 1791, "lr": 3.656703834170914e-05}
{"train_loss": 0.08658348768949509, "global_step": 159434, "epoch": 1791, "lr": 3.656647991985908e-05}
{"train_loss": 0.06795772165060043, "global_step": 159435, "epoch": 1791, "lr": 3.656592149981501e-05}
{"train_loss": 0.09754923731088638, "global_step": 159436, "epoch": 1791, "lr": 3.656536308157698e-05}
{"train_loss": 0.09595748037099838, "global_step": 159437, "epoch": 1791, "lr": 3.656480466514508e-05}
{"train_loss": 0.14097289741039276, "global_step": 159438, "epoch": 1791, "lr": 3.656424625051936e-05}
{"train_loss": 0.08292243629693985, "global_step": 159439, "epoch": 1791, "lr": 3.656368783769993e-05}
{"train_loss": 0.06951787322759628, "global_step": 159440, "epoch": 1791, "lr": 3.6563129426686845e-05}
{"train_loss": 0.07091150432825089, "global_step": 159441, "epoch": 1791, "lr": 3.6562571017480194e-05}
{"train_loss": 0.11035825312137604, "global_step": 159442, "epoch": 1791, "lr": 3.656201261008002e-05}
{"train_loss": 0.07402309030294418, "global_step": 159443, "epoch": 1791, "lr": 3.656145420448644e-05}
{"train_loss": 0.058754365891218185, "global_step": 159444, "epoch": 1791, "lr": 3.656089580069949e-05}
{"train_loss": 0.13392998278141022, "global_step": 159445, "epoch": 1791, "lr": 3.6560337398719277e-05}
{"train_loss": 0.06816783547401428, "global_step": 159446, "epoch": 1791, "lr": 3.6559778998545866e-05}
{"train_loss": 0.07946628332138062, "global_step": 159447, "epoch": 1791, "lr": 3.6559220600179325e-05}
{"train_loss": 0.045029547065496445, "global_step": 159448, "epoch": 1791, "lr": 3.655866220361972e-05}
{"train_loss": 0.06664631515741348, "global_step": 159449, "epoch": 1791, "lr": 3.6558103808867153e-05}
{"train_loss": 0.17355410754680634, "global_step": 159450, "epoch": 1791, "lr": 3.655754541592168e-05}
{"train_loss": 0.08996398746967316, "global_step": 159451, "epoch": 1791, "lr": 3.65569870247834e-05}
{"train_loss": 0.06031819432973862, "global_step": 159452, "epoch": 1791, "lr": 3.6556428635452334e-05}
{"train_loss": 0.0846150666475296, "global_step": 159453, "epoch": 1791, "lr": 3.6555870247928595e-05}
{"train_loss": 0.04766802117228508, "global_step": 159454, "epoch": 1791, "lr": 3.655531186221228e-05}
{"train_loss": 0.06559929996728897, "global_step": 159455, "epoch": 1791, "lr": 3.655475347830342e-05}
{"train_loss": 0.10644679516553879, "global_step": 159456, "epoch": 1791, "lr": 3.655419509620213e-05}
{"train_loss": 0.029007770121097565, "global_step": 159457, "epoch": 1791, "lr": 3.6553636715908446e-05}
{"train_loss": 0.10169903188943863, "global_step": 159458, "epoch": 1791, "lr": 3.655307833742248e-05}
{"train_loss": 0.06039927527308464, "global_step": 159459, "epoch": 1791, "lr": 3.655251996074427e-05}
{"train_loss": 0.08662474900484085, "global_step": 159460, "epoch": 1791, "lr": 3.655196158587393e-05}
{"train_loss": 0.07888807356357574, "global_step": 159461, "epoch": 1791, "lr": 3.6551403212811484e-05}
{"train_loss": 0.08777706325054169, "global_step": 159462, "epoch": 1791, "lr": 3.655084484155707e-05}
{"train_loss": 0.13401927053928375, "global_step": 159463, "epoch": 1791, "lr": 3.655028647211071e-05}
{"train_loss": 0.09413932263851166, "global_step": 159464, "epoch": 1791, "lr": 3.654972810447251e-05}
{"train_loss": 0.12212824076414108, "global_step": 159465, "epoch": 1791, "lr": 3.654916973864254e-05}
{"train_loss": 0.047153785824775696, "global_step": 159466, "epoch": 1791, "lr": 3.654861137462086e-05}
{"train_loss": 0.08238810300827026, "global_step": 159467, "epoch": 1791, "lr": 3.654805301240756e-05}
{"train_loss": 0.026940248906612396, "global_step": 159468, "epoch": 1791, "lr": 3.654749465200271e-05}
{"train_loss": 0.1349010020494461, "global_step": 159469, "epoch": 1791, "lr": 3.654693629340638e-05}
{"train_loss": 0.07033281773328781, "global_step": 159470, "epoch": 1791, "lr": 3.6546377936618645e-05}
{"train_loss": 0.08634278178215027, "global_step": 159471, "epoch": 1791, "lr": 3.6545819581639605e-05}
{"train_loss": 0.05358145385980606, "global_step": 159472, "epoch": 1791, "lr": 3.65452612284693e-05}
{"train_loss": 0.0964050218462944, "global_step": 159473, "epoch": 1791, "lr": 3.654470287710783e-05}
{"train_loss": 0.09046629816293716, "global_step": 159474, "epoch": 1791, "lr": 3.654414452755525e-05}
{"train_loss": 0.11709605902433395, "global_step": 159475, "epoch": 1791, "lr": 3.654358617981166e-05}
{"train_loss": 0.04237348586320877, "global_step": 159476, "epoch": 1791, "lr": 3.65430278338771e-05}
{"train_loss": 0.0818260908126831, "global_step": 159477, "epoch": 1791, "lr": 3.65424694897517e-05}
{"train_loss": 0.12502747774124146, "global_step": 159478, "epoch": 1791, "lr": 3.6541911147435456e-05}
{"train_loss": 0.05843256786465645, "global_step": 159479, "epoch": 1791, "lr": 3.654135280692853e-05}
{"train_loss": 0.0680154338479042, "global_step": 159480, "epoch": 1791, "lr": 3.654079446823093e-05}
{"train_loss": 0.09217407554388046, "global_step": 159481, "epoch": 1791, "lr": 3.654023613134277e-05}
{"train_loss": 0.1318853795528412, "global_step": 159482, "epoch": 1791, "lr": 3.653967779626409e-05}
{"train_loss": 0.04678814113140106, "global_step": 159483, "epoch": 1791, "lr": 3.6539119462995015e-05}
{"train_loss": 0.06850829720497131, "global_step": 159484, "epoch": 1791, "lr": 3.653856113153557e-05}
{"train_loss": 0.13495834171772003, "global_step": 159485, "epoch": 1791, "lr": 3.653800280188586e-05}
{"train_loss": 0.14758744835853577, "global_step": 159486, "epoch": 1791, "lr": 3.6537444474045954e-05}
{"train_loss": 0.09212686548407158, "global_step": 159487, "epoch": 1791, "lr": 3.6536886148015925e-05, "val_loss": 6.757650852203369}
{"train_loss": 0.11916989088058472, "global_step": 159488, "epoch": 1792, "lr": 3.6536327823795844e-05}
{"train_loss": 0.03692266717553139, "global_step": 159489, "epoch": 1792, "lr": 3.6535769501385776e-05}
{"train_loss": 0.1120493933558464, "global_step": 159490, "epoch": 1792, "lr": 3.653521118078583e-05}
{"train_loss": 0.0975918099284172, "global_step": 159491, "epoch": 1792, "lr": 3.6534652861996045e-05}
{"train_loss": 0.12574544548988342, "global_step": 159492, "epoch": 1792, "lr": 3.653409454501653e-05}
{"train_loss": 0.0965043380856514, "global_step": 159493, "epoch": 1792, "lr": 3.653353622984732e-05}
{"train_loss": 0.045709386467933655, "global_step": 159494, "epoch": 1792, "lr": 3.653297791648853e-05}
{"train_loss": 0.15090058743953705, "global_step": 159495, "epoch": 1792, "lr": 3.6532419604940206e-05}
{"train_loss": 0.0899231806397438, "global_step": 159496, "epoch": 1792, "lr": 3.653186129520245e-05}
{"train_loss": 0.15098543465137482, "global_step": 159497, "epoch": 1792, "lr": 3.65313029872753e-05}
{"train_loss": 0.0939609631896019, "global_step": 159498, "epoch": 1792, "lr": 3.653074468115887e-05}
{"train_loss": 0.069915272295475, "global_step": 159499, "epoch": 1792, "lr": 3.65301863768532e-05}
{"train_loss": 0.15413978695869446, "global_step": 159500, "epoch": 1792, "lr": 3.65296280743584e-05}
{"train_loss": 0.08314529061317444, "global_step": 159501, "epoch": 1792, "lr": 3.6529069773674504e-05}
{"train_loss": 0.07784339785575867, "global_step": 159502, "epoch": 1792, "lr": 3.652851147480164e-05}
{"train_loss": 0.08647184818983078, "global_step": 159503, "epoch": 1792, "lr": 3.6527953177739826e-05}
{"train_loss": 0.09889919310808182, "global_step": 159504, "epoch": 1792, "lr": 3.652739488248919e-05}
{"train_loss": 0.08778466284275055, "global_step": 159505, "epoch": 1792, "lr": 3.652683658904977e-05}
{"train_loss": 0.07967208325862885, "global_step": 159506, "epoch": 1792, "lr": 3.652627829742164e-05}
{"train_loss": 0.1267428994178772, "global_step": 159507, "epoch": 1792, "lr": 3.652572000760491e-05}
{"train_loss": 0.10826457291841507, "global_step": 159508, "epoch": 1792, "lr": 3.652516171959961e-05}
{"train_loss": 0.07889346033334732, "global_step": 159509, "epoch": 1792, "lr": 3.6524603433405855e-05}
{"train_loss": 0.14941637217998505, "global_step": 159510, "epoch": 1792, "lr": 3.652404514902369e-05}
{"train_loss": 0.08183345943689346, "global_step": 159511, "epoch": 1792, "lr": 3.652348686645322e-05}
{"train_loss": 0.1214447170495987, "global_step": 159512, "epoch": 1792, "lr": 3.652292858569448e-05}
{"train_loss": 0.12446921318769455, "global_step": 159513, "epoch": 1792, "lr": 3.6522370306747594e-05}
{"train_loss": 0.0761217549443245, "global_step": 159514, "epoch": 1792, "lr": 3.652181202961259e-05}
{"train_loss": 0.076128289103508, "global_step": 159515, "epoch": 1792, "lr": 3.652125375428958e-05}
{"train_loss": 0.06336044520139694, "global_step": 159516, "epoch": 1792, "lr": 3.65206954807786e-05}
{"train_loss": 0.13488957285881042, "global_step": 159517, "epoch": 1792, "lr": 3.652013720907978e-05}
{"train_loss": 0.08224139362573624, "global_step": 159518, "epoch": 1792, "lr": 3.6519578939193135e-05}
{"train_loss": 0.06887490302324295, "global_step": 159519, "epoch": 1792, "lr": 3.6519020671118795e-05}
{"train_loss": 0.10135221481323242, "global_step": 159520, "epoch": 1792, "lr": 3.651846240485678e-05}
{"train_loss": 0.15451575815677643, "global_step": 159521, "epoch": 1792, "lr": 3.651790414040722e-05}
{"train_loss": 0.10365656763315201, "global_step": 159522, "epoch": 1792, "lr": 3.651734587777014e-05}
{"train_loss": 0.13026969134807587, "global_step": 159523, "epoch": 1792, "lr": 3.6516787616945656e-05}
{"train_loss": 0.11977314949035645, "global_step": 159524, "epoch": 1792, "lr": 3.6516229357933817e-05}
{"train_loss": 0.1289615035057068, "global_step": 159525, "epoch": 1792, "lr": 3.65156711007347e-05}
{"train_loss": 0.06604267656803131, "global_step": 159526, "epoch": 1792, "lr": 3.651511284534841e-05}
{"train_loss": 0.1222376823425293, "global_step": 159527, "epoch": 1792, "lr": 3.6514554591774966e-05}
{"train_loss": 0.047204770147800446, "global_step": 159528, "epoch": 1792, "lr": 3.651399634001451e-05}
{"train_loss": 0.07874992489814758, "global_step": 159529, "epoch": 1792, "lr": 3.651343809006705e-05}
{"train_loss": 0.16498824954032898, "global_step": 159530, "epoch": 1792, "lr": 3.651287984193272e-05}
{"train_loss": 0.11443185061216354, "global_step": 159531, "epoch": 1792, "lr": 3.651232159561155e-05}
{"train_loss": 0.06424426287412643, "global_step": 159532, "epoch": 1792, "lr": 3.651176335110366e-05}
{"train_loss": 0.062218647450208664, "global_step": 159533, "epoch": 1792, "lr": 3.651120510840906e-05}
{"train_loss": 0.07391281425952911, "global_step": 159534, "epoch": 1792, "lr": 3.65106468675279e-05}
{"train_loss": 0.0357222855091095, "global_step": 159535, "epoch": 1792, "lr": 3.651008862846019e-05}
{"train_loss": 0.08793986588716507, "global_step": 159536, "epoch": 1792, "lr": 3.6509530391206065e-05}
{"train_loss": 0.11356168985366821, "global_step": 159537, "epoch": 1792, "lr": 3.650897215576554e-05}
{"train_loss": 0.1384412795305252, "global_step": 159538, "epoch": 1792, "lr": 3.650841392213875e-05}
{"train_loss": 0.11729849129915237, "global_step": 159539, "epoch": 1792, "lr": 3.650785569032571e-05}
{"train_loss": 0.061128366738557816, "global_step": 159540, "epoch": 1792, "lr": 3.650729746032654e-05}
{"train_loss": 0.11609559506177902, "global_step": 159541, "epoch": 1792, "lr": 3.6506739232141286e-05}
{"train_loss": 0.059231098741292953, "global_step": 159542, "epoch": 1792, "lr": 3.6506181005770044e-05}
{"train_loss": 0.11249276995658875, "global_step": 159543, "epoch": 1792, "lr": 3.65056227812129e-05}
{"train_loss": 0.04778015986084938, "global_step": 159544, "epoch": 1792, "lr": 3.650506455846988e-05}
{"train_loss": 0.0486680343747139, "global_step": 159545, "epoch": 1792, "lr": 3.650450633754112e-05}
{"train_loss": 0.07624596357345581, "global_step": 159546, "epoch": 1792, "lr": 3.650394811842664e-05}
{"train_loss": 0.07376903295516968, "global_step": 159547, "epoch": 1792, "lr": 3.650338990112655e-05}
{"train_loss": 0.10577109456062317, "global_step": 159548, "epoch": 1792, "lr": 3.6502831685640906e-05}
{"train_loss": 0.13643531501293182, "global_step": 159549, "epoch": 1792, "lr": 3.6502273471969805e-05}
{"train_loss": 0.03213169425725937, "global_step": 159550, "epoch": 1792, "lr": 3.650171526011329e-05}
{"train_loss": 0.06306102871894836, "global_step": 159551, "epoch": 1792, "lr": 3.650115705007148e-05}
{"train_loss": 0.12139762938022614, "global_step": 159552, "epoch": 1792, "lr": 3.6500598841844405e-05}
{"train_loss": 0.06550806760787964, "global_step": 159553, "epoch": 1792, "lr": 3.650004063543217e-05}
{"train_loss": 0.10842407494783401, "global_step": 159554, "epoch": 1792, "lr": 3.649948243083483e-05}
{"train_loss": 0.10603927075862885, "global_step": 159555, "epoch": 1792, "lr": 3.649892422805249e-05}
{"train_loss": 0.062101978808641434, "global_step": 159556, "epoch": 1792, "lr": 3.649836602708518e-05}
{"train_loss": 0.13815937936306, "global_step": 159557, "epoch": 1792, "lr": 3.6497807827933026e-05}
{"train_loss": 0.034826021641492844, "global_step": 159558, "epoch": 1792, "lr": 3.6497249630596054e-05}
{"train_loss": 0.04087155684828758, "global_step": 159559, "epoch": 1792, "lr": 3.649669143507438e-05}
{"train_loss": 0.10642732679843903, "global_step": 159560, "epoch": 1792, "lr": 3.6496133241368043e-05}
{"train_loss": 0.11394289880990982, "global_step": 159561, "epoch": 1792, "lr": 3.649557504947715e-05}
{"train_loss": 0.08058938384056091, "global_step": 159562, "epoch": 1792, "lr": 3.649501685940176e-05}
{"train_loss": 0.12564168870449066, "global_step": 159563, "epoch": 1792, "lr": 3.649445867114194e-05}
{"train_loss": 0.0962752252817154, "global_step": 159564, "epoch": 1792, "lr": 3.64939004846978e-05}
{"train_loss": 0.09052726626396179, "global_step": 159565, "epoch": 1792, "lr": 3.649334230006936e-05}
{"train_loss": 0.07837651669979095, "global_step": 159566, "epoch": 1792, "lr": 3.649278411725675e-05}
{"train_loss": 0.10788442194461823, "global_step": 159567, "epoch": 1792, "lr": 3.6492225936260005e-05}
{"train_loss": 0.10739295929670334, "global_step": 159568, "epoch": 1792, "lr": 3.6491667757079225e-05}
{"train_loss": 0.058584488928318024, "global_step": 159569, "epoch": 1792, "lr": 3.6491109579714463e-05}
{"train_loss": 0.09549374878406525, "global_step": 159570, "epoch": 1792, "lr": 3.649055140416583e-05}
{"train_loss": 0.14751394093036652, "global_step": 159571, "epoch": 1792, "lr": 3.648999323043335e-05}
{"train_loss": 0.059487234801054, "global_step": 159572, "epoch": 1792, "lr": 3.6489435058517155e-05}
{"train_loss": 0.12563998997211456, "global_step": 159573, "epoch": 1792, "lr": 3.648887688841726e-05}
{"train_loss": 0.07825886458158493, "global_step": 159574, "epoch": 1792, "lr": 3.6488318720133795e-05}
{"train_loss": 0.08404488861560822, "global_step": 159575, "epoch": 1792, "lr": 3.6487760553666796e-05}
{"train_loss": 0.09510373838999298, "global_step": 159576, "epoch": 1792, "lr": 3.648720238901637e-05, "val_loss": 6.749534606933594}
{"train_loss": 0.06717410683631897, "global_step": 159577, "epoch": 1793, "lr": 3.6486644226182555e-05}
{"train_loss": 0.1053743064403534, "global_step": 159578, "epoch": 1793, "lr": 3.648608606516546e-05}
{"train_loss": 0.07104969024658203, "global_step": 159579, "epoch": 1793, "lr": 3.648552790596513e-05}
{"train_loss": 0.15074633061885834, "global_step": 159580, "epoch": 1793, "lr": 3.6484969748581666e-05}
{"train_loss": 0.08183091133832932, "global_step": 159581, "epoch": 1793, "lr": 3.648441159301514e-05}
{"train_loss": 0.08287135511636734, "global_step": 159582, "epoch": 1793, "lr": 3.648385343926561e-05}
{"train_loss": 0.053624339401721954, "global_step": 159583, "epoch": 1793, "lr": 3.6483295287333176e-05}
{"train_loss": 0.10632304102182388, "global_step": 159584, "epoch": 1793, "lr": 3.648273713721788e-05}
{"train_loss": 0.08765912055969238, "global_step": 159585, "epoch": 1793, "lr": 3.648217898891983e-05}
{"train_loss": 0.0885232612490654, "global_step": 159586, "epoch": 1793, "lr": 3.648162084243907e-05}
{"train_loss": 0.15712983906269073, "global_step": 159587, "epoch": 1793, "lr": 3.648106269777571e-05}
{"train_loss": 0.11311058700084686, "global_step": 159588, "epoch": 1793, "lr": 3.648050455492978e-05}
{"train_loss": 0.08291149884462357, "global_step": 159589, "epoch": 1793, "lr": 3.647994641390141e-05}
{"train_loss": 0.057286933064460754, "global_step": 159590, "epoch": 1793, "lr": 3.647938827469063e-05}
{"train_loss": 0.1087440773844719, "global_step": 159591, "epoch": 1793, "lr": 3.6478830137297545e-05}
{"train_loss": 0.10897915065288544, "global_step": 159592, "epoch": 1793, "lr": 3.647827200172219e-05}
{"train_loss": 0.06812198460102081, "global_step": 159593, "epoch": 1793, "lr": 3.64777138679647e-05}
{"train_loss": 0.07645789533853531, "global_step": 159594, "epoch": 1793, "lr": 3.6477155736025086e-05}
{"train_loss": 0.08099407702684402, "global_step": 159595, "epoch": 1793, "lr": 3.647659760590348e-05}
{"train_loss": 0.08138583600521088, "global_step": 159596, "epoch": 1793, "lr": 3.647603947759991e-05}
{"train_loss": 0.1153656393289566, "global_step": 159597, "epoch": 1793, "lr": 3.647548135111448e-05}
{"train_loss": 0.08032415807247162, "global_step": 159598, "epoch": 1793, "lr": 3.6474923226447253e-05}
{"train_loss": 0.10544361174106598, "global_step": 159599, "epoch": 1793, "lr": 3.647436510359831e-05}
{"train_loss": 0.10197177529335022, "global_step": 159600, "epoch": 1793, "lr": 3.647380698256774e-05}
{"train_loss": 0.1381540447473526, "global_step": 159601, "epoch": 1793, "lr": 3.647324886335558e-05}
{"train_loss": 0.11925669014453888, "global_step": 159602, "epoch": 1793, "lr": 3.647269074596194e-05}
{"train_loss": 0.07373474538326263, "global_step": 159603, "epoch": 1793, "lr": 3.647213263038687e-05}
{"train_loss": 0.13916093111038208, "global_step": 159604, "epoch": 1793, "lr": 3.647157451663047e-05}
{"train_loss": 0.0685572400689125, "global_step": 159605, "epoch": 1793, "lr": 3.647101640469278e-05}
{"train_loss": 0.08789794892072678, "global_step": 159606, "epoch": 1793, "lr": 3.647045829457392e-05}
{"train_loss": 0.07404524087905884, "global_step": 159607, "epoch": 1793, "lr": 3.6469900186273924e-05}
{"train_loss": 0.04824135825037956, "global_step": 159608, "epoch": 1793, "lr": 3.6469342079792904e-05}
{"train_loss": 0.07413306087255478, "global_step": 159609, "epoch": 1793, "lr": 3.64687839751309e-05}
{"train_loss": 0.06756916642189026, "global_step": 159610, "epoch": 1793, "lr": 3.646822587228802e-05}
{"train_loss": 0.06281433254480362, "global_step": 159611, "epoch": 1793, "lr": 3.6467667771264305e-05}
{"train_loss": 0.10722462832927704, "global_step": 159612, "epoch": 1793, "lr": 3.646710967205986e-05}
{"train_loss": 0.0909329280257225, "global_step": 159613, "epoch": 1793, "lr": 3.646655157467473e-05}
{"train_loss": 0.06616324186325073, "global_step": 159614, "epoch": 1793, "lr": 3.646599347910903e-05}
{"train_loss": 0.14242354035377502, "global_step": 159615, "epoch": 1793, "lr": 3.646543538536279e-05}
{"train_loss": 0.17283567786216736, "global_step": 159616, "epoch": 1793, "lr": 3.646487729343612e-05}
{"train_loss": 0.07627792656421661, "global_step": 159617, "epoch": 1793, "lr": 3.646431920332908e-05}
{"train_loss": 0.09007640928030014, "global_step": 159618, "epoch": 1793, "lr": 3.6463761115041746e-05}
{"train_loss": 0.08143782615661621, "global_step": 159619, "epoch": 1793, "lr": 3.646320302857419e-05}
{"train_loss": 0.10073337703943253, "global_step": 159620, "epoch": 1793, "lr": 3.64626449439265e-05}
{"train_loss": 0.07942243665456772, "global_step": 159621, "epoch": 1793, "lr": 3.646208686109875e-05}
{"train_loss": 0.08189667016267776, "global_step": 159622, "epoch": 1793, "lr": 3.6461528780090984e-05}
{"train_loss": 0.07916390150785446, "global_step": 159623, "epoch": 1793, "lr": 3.646097070090333e-05}
{"train_loss": 0.07448739558458328, "global_step": 159624, "epoch": 1793, "lr": 3.64604126235358e-05}
{"train_loss": 0.1444951444864273, "global_step": 159625, "epoch": 1793, "lr": 3.645985454798853e-05}
{"train_loss": 0.13838353753089905, "global_step": 159626, "epoch": 1793, "lr": 3.645929647426155e-05}
{"train_loss": 0.09088891744613647, "global_step": 159627, "epoch": 1793, "lr": 3.645873840235496e-05}
{"train_loss": 0.13748237490653992, "global_step": 159628, "epoch": 1793, "lr": 3.645818033226882e-05}
{"train_loss": 0.1138521283864975, "global_step": 159629, "epoch": 1793, "lr": 3.645762226400323e-05}
{"train_loss": 0.1111079677939415, "global_step": 159630, "epoch": 1793, "lr": 3.645706419755823e-05}
{"train_loss": 0.06339786946773529, "global_step": 159631, "epoch": 1793, "lr": 3.645650613293393e-05}
{"train_loss": 0.10708851367235184, "global_step": 159632, "epoch": 1793, "lr": 3.6455948070130366e-05}
{"train_loss": 0.10815085470676422, "global_step": 159633, "epoch": 1793, "lr": 3.645539000914764e-05}
{"train_loss": 0.08602140098810196, "global_step": 159634, "epoch": 1793, "lr": 3.645483194998584e-05}
{"train_loss": 0.064236581325531, "global_step": 159635, "epoch": 1793, "lr": 3.645427389264501e-05}
{"train_loss": 0.09741275757551193, "global_step": 159636, "epoch": 1793, "lr": 3.645371583712524e-05}
{"train_loss": 0.13793446123600006, "global_step": 159637, "epoch": 1793, "lr": 3.6453157783426595e-05}
{"train_loss": 0.07122930139303207, "global_step": 159638, "epoch": 1793, "lr": 3.645259973154916e-05}
{"train_loss": 0.07326629757881165, "global_step": 159639, "epoch": 1793, "lr": 3.645204168149302e-05}
{"train_loss": 0.12450596690177917, "global_step": 159640, "epoch": 1793, "lr": 3.645148363325823e-05}
{"train_loss": 0.1335500031709671, "global_step": 159641, "epoch": 1793, "lr": 3.6450925586844865e-05}
{"train_loss": 0.08479268848896027, "global_step": 159642, "epoch": 1793, "lr": 3.6450367542253026e-05}
{"train_loss": 0.14557547867298126, "global_step": 159643, "epoch": 1793, "lr": 3.644980949948275e-05}
{"train_loss": 0.06387301534414291, "global_step": 159644, "epoch": 1793, "lr": 3.6449251458534153e-05}
{"train_loss": 0.14293628931045532, "global_step": 159645, "epoch": 1793, "lr": 3.644869341940726e-05}
{"train_loss": 0.07322647422552109, "global_step": 159646, "epoch": 1793, "lr": 3.644813538210221e-05}
{"train_loss": 0.036278460174798965, "global_step": 159647, "epoch": 1793, "lr": 3.644757734661901e-05}
{"train_loss": 0.0980684831738472, "global_step": 159648, "epoch": 1793, "lr": 3.64470193129578e-05}
{"train_loss": 0.04351208731532097, "global_step": 159649, "epoch": 1793, "lr": 3.644646128111859e-05}
{"train_loss": 0.10833396762609482, "global_step": 159650, "epoch": 1793, "lr": 3.644590325110152e-05}
{"train_loss": 0.12208391726016998, "global_step": 159651, "epoch": 1793, "lr": 3.6445345222906605e-05}
{"train_loss": 0.1484546959400177, "global_step": 159652, "epoch": 1793, "lr": 3.644478719653395e-05}
{"train_loss": 0.09929757565259933, "global_step": 159653, "epoch": 1793, "lr": 3.644422917198365e-05}
{"train_loss": 0.06716340780258179, "global_step": 159654, "epoch": 1793, "lr": 3.644367114925575e-05}
{"train_loss": 0.0823177844285965, "global_step": 159655, "epoch": 1793, "lr": 3.6443113128350326e-05}
{"train_loss": 0.1555849015712738, "global_step": 159656, "epoch": 1793, "lr": 3.6442555109267466e-05}
{"train_loss": 0.12127599120140076, "global_step": 159657, "epoch": 1793, "lr": 3.644199709200724e-05}
{"train_loss": 0.08721422404050827, "global_step": 159658, "epoch": 1793, "lr": 3.644143907656972e-05}
{"train_loss": 0.14733633399009705, "global_step": 159659, "epoch": 1793, "lr": 3.644088106295499e-05}
{"train_loss": 0.07102201879024506, "global_step": 159660, "epoch": 1793, "lr": 3.644032305116311e-05}
{"train_loss": 0.07255999743938446, "global_step": 159661, "epoch": 1793, "lr": 3.643976504119417e-05}
{"train_loss": 0.08341051638126373, "global_step": 159662, "epoch": 1793, "lr": 3.643920703304822e-05}
{"train_loss": 0.15642157196998596, "global_step": 159663, "epoch": 1793, "lr": 3.643864902672538e-05}
{"train_loss": 0.042766936123371124, "global_step": 159664, "epoch": 1793, "lr": 3.643809102222567e-05}
{"train_loss": 0.09643850889935922, "global_step": 159665, "epoch": 1793, "lr": 3.643753301954922e-05, "val_loss": 6.761104106903076}
{"train_loss": 0.1300385296344757, "global_step": 159666, "epoch": 1794, "lr": 3.643697501869606e-05}
{"train_loss": 0.14106756448745728, "global_step": 159667, "epoch": 1794, "lr": 3.64364170196663e-05}
{"train_loss": 0.06362837553024292, "global_step": 159668, "epoch": 1794, "lr": 3.6435859022459976e-05}
{"train_loss": 0.1254490613937378, "global_step": 159669, "epoch": 1794, "lr": 3.6435301027077186e-05}
{"train_loss": 0.08798582851886749, "global_step": 159670, "epoch": 1794, "lr": 3.6434743033518025e-05}
{"train_loss": 0.07760205864906311, "global_step": 159671, "epoch": 1794, "lr": 3.643418504178253e-05}
{"train_loss": 0.08894184976816177, "global_step": 159672, "epoch": 1794, "lr": 3.643362705187081e-05}
{"train_loss": 0.04961053654551506, "global_step": 159673, "epoch": 1794, "lr": 3.643306906378292e-05}
{"train_loss": 0.09711691737174988, "global_step": 159674, "epoch": 1794, "lr": 3.6432511077518935e-05}
{"train_loss": 0.06627405434846878, "global_step": 159675, "epoch": 1794, "lr": 3.643195309307893e-05}
{"train_loss": 0.13982270658016205, "global_step": 159676, "epoch": 1794, "lr": 3.643139511046299e-05}
{"train_loss": 0.05872487276792526, "global_step": 159677, "epoch": 1794, "lr": 3.643083712967118e-05}
{"train_loss": 0.10677704215049744, "global_step": 159678, "epoch": 1794, "lr": 3.6430279150703585e-05}
{"train_loss": 0.11180280894041061, "global_step": 159679, "epoch": 1794, "lr": 3.642972117356026e-05}
{"train_loss": 0.07456129044294357, "global_step": 159680, "epoch": 1794, "lr": 3.642916319824131e-05}
{"train_loss": 0.07283330708742142, "global_step": 159681, "epoch": 1794, "lr": 3.642860522474677e-05}
{"train_loss": 0.13334989547729492, "global_step": 159682, "epoch": 1794, "lr": 3.642804725307676e-05}
{"train_loss": 0.10120996087789536, "global_step": 159683, "epoch": 1794, "lr": 3.642748928323132e-05}
{"train_loss": 0.06944457441568375, "global_step": 159684, "epoch": 1794, "lr": 3.642693131521056e-05}
{"train_loss": 0.06344982981681824, "global_step": 159685, "epoch": 1794, "lr": 3.6426373349014505e-05}
{"train_loss": 0.11566828191280365, "global_step": 159686, "epoch": 1794, "lr": 3.642581538464328e-05}
{"train_loss": 0.050426531583070755, "global_step": 159687, "epoch": 1794, "lr": 3.642525742209693e-05}
{"train_loss": 0.07356686145067215, "global_step": 159688, "epoch": 1794, "lr": 3.642469946137552e-05}
{"train_loss": 0.0693369060754776, "global_step": 159689, "epoch": 1794, "lr": 3.642414150247918e-05}
{"train_loss": 0.10293687880039215, "global_step": 159690, "epoch": 1794, "lr": 3.642358354540791e-05}
{"train_loss": 0.14788149297237396, "global_step": 159691, "epoch": 1794, "lr": 3.6423025590161866e-05}
{"train_loss": 0.12982338666915894, "global_step": 159692, "epoch": 1794, "lr": 3.6422467636741055e-05}
{"train_loss": 0.10600456595420837, "global_step": 159693, "epoch": 1794, "lr": 3.642190968514558e-05}
{"train_loss": 0.11973389983177185, "global_step": 159694, "epoch": 1794, "lr": 3.642135173537552e-05}
{"train_loss": 0.09187640994787216, "global_step": 159695, "epoch": 1794, "lr": 3.642079378743094e-05}
{"train_loss": 0.06762924045324326, "global_step": 159696, "epoch": 1794, "lr": 3.642023584131192e-05}
{"train_loss": 0.1325100064277649, "global_step": 159697, "epoch": 1794, "lr": 3.641967789701855e-05}
{"train_loss": 0.13486571609973907, "global_step": 159698, "epoch": 1794, "lr": 3.641911995455086e-05}
{"train_loss": 0.0548919215798378, "global_step": 159699, "epoch": 1794, "lr": 3.641856201390897e-05}
{"train_loss": 0.09755458682775497, "global_step": 159700, "epoch": 1794, "lr": 3.641800407509293e-05}
{"train_loss": 0.11030915379524231, "global_step": 159701, "epoch": 1794, "lr": 3.641744613810284e-05}
{"train_loss": 0.07978399842977524, "global_step": 159702, "epoch": 1794, "lr": 3.6416888202938734e-05}
{"train_loss": 0.13420115411281586, "global_step": 159703, "epoch": 1794, "lr": 3.641633026960074e-05}
{"train_loss": 0.053692881017923355, "global_step": 159704, "epoch": 1794, "lr": 3.641577233808887e-05}
{"train_loss": 0.12433288246393204, "global_step": 159705, "epoch": 1794, "lr": 3.641521440840327e-05}
{"train_loss": 0.12392393499612808, "global_step": 159706, "epoch": 1794, "lr": 3.641465648054395e-05}
{"train_loss": 0.11635339260101318, "global_step": 159707, "epoch": 1794, "lr": 3.6414098554511015e-05}
{"train_loss": 0.09549126029014587, "global_step": 159708, "epoch": 1794, "lr": 3.641354063030457e-05}
{"train_loss": 0.046230267733335495, "global_step": 159709, "epoch": 1794, "lr": 3.641298270792463e-05}
{"train_loss": 0.06387931853532791, "global_step": 159710, "epoch": 1794, "lr": 3.641242478737131e-05}
{"train_loss": 0.1255137324333191, "global_step": 159711, "epoch": 1794, "lr": 3.641186686864467e-05}
{"train_loss": 0.10285493731498718, "global_step": 159712, "epoch": 1794, "lr": 3.64113089517448e-05}
{"train_loss": 0.09980428963899612, "global_step": 159713, "epoch": 1794, "lr": 3.6410751036671756e-05}
{"train_loss": 0.11196514964103699, "global_step": 159714, "epoch": 1794, "lr": 3.641019312342561e-05}
{"train_loss": 0.12950171530246735, "global_step": 159715, "epoch": 1794, "lr": 3.640963521200646e-05}
{"train_loss": 0.09513191878795624, "global_step": 159716, "epoch": 1794, "lr": 3.6409077302414364e-05}
{"train_loss": 0.13439784944057465, "global_step": 159717, "epoch": 1794, "lr": 3.640851939464941e-05}
{"train_loss": 0.08513215184211731, "global_step": 159718, "epoch": 1794, "lr": 3.6407961488711666e-05}
{"train_loss": 0.08273950219154358, "global_step": 159719, "epoch": 1794, "lr": 3.640740358460119e-05}
{"train_loss": 0.11596797406673431, "global_step": 159720, "epoch": 1794, "lr": 3.640684568231809e-05}
{"train_loss": 0.1579609364271164, "global_step": 159721, "epoch": 1794, "lr": 3.640628778186241e-05}
{"train_loss": 0.0759982317686081, "global_step": 159722, "epoch": 1794, "lr": 3.640572988323425e-05}
{"train_loss": 0.06297574192285538, "global_step": 159723, "epoch": 1794, "lr": 3.640517198643366e-05}
{"train_loss": 0.11243965476751328, "global_step": 159724, "epoch": 1794, "lr": 3.640461409146073e-05}
{"train_loss": 0.10889222472906113, "global_step": 159725, "epoch": 1794, "lr": 3.640405619831555e-05}
{"train_loss": 0.07594513148069382, "global_step": 159726, "epoch": 1794, "lr": 3.640349830699816e-05}
{"train_loss": 0.06765927374362946, "global_step": 159727, "epoch": 1794, "lr": 3.640294041750868e-05}
{"train_loss": 0.041899874806404114, "global_step": 159728, "epoch": 1794, "lr": 3.640238252984713e-05}
{"train_loss": 0.07736954838037491, "global_step": 159729, "epoch": 1794, "lr": 3.640182464401364e-05}
{"train_loss": 0.05935371667146683, "global_step": 159730, "epoch": 1794, "lr": 3.6401266760008234e-05}
{"train_loss": 0.10409189760684967, "global_step": 159731, "epoch": 1794, "lr": 3.640070887783104e-05}
{"train_loss": 0.07036697864532471, "global_step": 159732, "epoch": 1794, "lr": 3.6400150997482086e-05}
{"train_loss": 0.05030554160475731, "global_step": 159733, "epoch": 1794, "lr": 3.6399593118961465e-05}
{"train_loss": 0.06944803148508072, "global_step": 159734, "epoch": 1794, "lr": 3.6399035242269255e-05}
{"train_loss": 0.1336931735277176, "global_step": 159735, "epoch": 1794, "lr": 3.639847736740554e-05}
{"train_loss": 0.07274331152439117, "global_step": 159736, "epoch": 1794, "lr": 3.639791949437037e-05}
{"train_loss": 0.08515509963035583, "global_step": 159737, "epoch": 1794, "lr": 3.639736162316385e-05}
{"train_loss": 0.11211881041526794, "global_step": 159738, "epoch": 1794, "lr": 3.6396803753786016e-05}
{"train_loss": 0.11342158168554306, "global_step": 159739, "epoch": 1794, "lr": 3.639624588623699e-05}
{"train_loss": 0.0954662561416626, "global_step": 159740, "epoch": 1794, "lr": 3.63956880205168e-05}
{"train_loss": 0.10090824216604233, "global_step": 159741, "epoch": 1794, "lr": 3.639513015662557e-05}
{"train_loss": 0.15950298309326172, "global_step": 159742, "epoch": 1794, "lr": 3.639457229456332e-05}
{"train_loss": 0.04860460385680199, "global_step": 159743, "epoch": 1794, "lr": 3.639401443433015e-05}
{"train_loss": 0.050470609217882156, "global_step": 159744, "epoch": 1794, "lr": 3.639345657592617e-05}
{"train_loss": 0.09612524509429932, "global_step": 159745, "epoch": 1794, "lr": 3.63928987193514e-05}
{"train_loss": 0.07587066292762756, "global_step": 159746, "epoch": 1794, "lr": 3.6392340864605964e-05}
{"train_loss": 0.1283150613307953, "global_step": 159747, "epoch": 1794, "lr": 3.639178301168988e-05}
{"train_loss": 0.08903086930513382, "global_step": 159748, "epoch": 1794, "lr": 3.639122516060329e-05}
{"train_loss": 0.07055249065160751, "global_step": 159749, "epoch": 1794, "lr": 3.6390667311346204e-05}
{"train_loss": 0.1446147859096527, "global_step": 159750, "epoch": 1794, "lr": 3.6390109463918746e-05}
{"train_loss": 0.1063799262046814, "global_step": 159751, "epoch": 1794, "lr": 3.638955161832097e-05}
{"train_loss": 0.06846294552087784, "global_step": 159752, "epoch": 1794, "lr": 3.638899377455295e-05}
{"train_loss": 0.1185796707868576, "global_step": 159753, "epoch": 1794, "lr": 3.638843593261476e-05}
{"train_loss": 0.09567598190702749, "global_step": 159754, "epoch": 1794, "lr": 3.638787809250648e-05, "val_loss": 6.8132710456848145}
{"train_loss": 0.07132679224014282, "global_step": 159755, "epoch": 1795, "lr": 3.638732025422818e-05}
{"train_loss": 0.11401998996734619, "global_step": 159756, "epoch": 1795, "lr": 3.638676241777995e-05}
{"train_loss": 0.09135770797729492, "global_step": 159757, "epoch": 1795, "lr": 3.638620458316184e-05}
{"train_loss": 0.12181452661752701, "global_step": 159758, "epoch": 1795, "lr": 3.638564675037396e-05}
{"train_loss": 0.12260110676288605, "global_step": 159759, "epoch": 1795, "lr": 3.6385088919416336e-05}
{"train_loss": 0.08749325573444366, "global_step": 159760, "epoch": 1795, "lr": 3.638453109028908e-05}
{"train_loss": 0.06967616081237793, "global_step": 159761, "epoch": 1795, "lr": 3.638397326299227e-05}
{"train_loss": 0.018805239349603653, "global_step": 159762, "epoch": 1795, "lr": 3.638341543752595e-05}
{"train_loss": 0.0639115497469902, "global_step": 159763, "epoch": 1795, "lr": 3.6382857613890235e-05}
{"train_loss": 0.11871012300252914, "global_step": 159764, "epoch": 1795, "lr": 3.6382299792085164e-05}
{"train_loss": 0.09333411604166031, "global_step": 159765, "epoch": 1795, "lr": 3.6381741972110836e-05}
{"train_loss": 0.11306264251470566, "global_step": 159766, "epoch": 1795, "lr": 3.6381184153967306e-05}
{"train_loss": 0.11424136161804199, "global_step": 159767, "epoch": 1795, "lr": 3.6380626337654675e-05}
{"train_loss": 0.14565399289131165, "global_step": 159768, "epoch": 1795, "lr": 3.638006852317298e-05}
{"train_loss": 0.09713660925626755, "global_step": 159769, "epoch": 1795, "lr": 3.6379510710522346e-05}
{"train_loss": 0.09496483206748962, "global_step": 159770, "epoch": 1795, "lr": 3.637895289970281e-05}
{"train_loss": 0.10612418502569199, "global_step": 159771, "epoch": 1795, "lr": 3.6378395090714454e-05}
{"train_loss": 0.08512523025274277, "global_step": 159772, "epoch": 1795, "lr": 3.637783728355736e-05}
{"train_loss": 0.10626094788312912, "global_step": 159773, "epoch": 1795, "lr": 3.6377279478231594e-05}
{"train_loss": 0.09742335975170135, "global_step": 159774, "epoch": 1795, "lr": 3.637672167473725e-05}
{"train_loss": 0.05323820561170578, "global_step": 159775, "epoch": 1795, "lr": 3.637616387307439e-05}
{"train_loss": 0.07737810164690018, "global_step": 159776, "epoch": 1795, "lr": 3.637560607324306e-05}
{"train_loss": 0.1125865951180458, "global_step": 159777, "epoch": 1795, "lr": 3.63750482752434e-05}
{"train_loss": 0.09511362016201019, "global_step": 159778, "epoch": 1795, "lr": 3.637449047907542e-05}
{"train_loss": 0.16391398012638092, "global_step": 159779, "epoch": 1795, "lr": 3.637393268473922e-05}
{"train_loss": 0.07096754759550095, "global_step": 159780, "epoch": 1795, "lr": 3.637337489223491e-05}
{"train_loss": 0.06431354582309723, "global_step": 159781, "epoch": 1795, "lr": 3.637281710156251e-05}
{"train_loss": 0.1202324852347374, "global_step": 159782, "epoch": 1795, "lr": 3.637225931272214e-05}
{"train_loss": 0.11833519488573074, "global_step": 159783, "epoch": 1795, "lr": 3.6371701525713816e-05}
{"train_loss": 0.07591398805379868, "global_step": 159784, "epoch": 1795, "lr": 3.6371143740537696e-05}
{"train_loss": 0.1136774942278862, "global_step": 159785, "epoch": 1795, "lr": 3.637058595719377e-05}
{"train_loss": 0.10212276875972748, "global_step": 159786, "epoch": 1795, "lr": 3.6370028175682184e-05}
{"train_loss": 0.07130982726812363, "global_step": 159787, "epoch": 1795, "lr": 3.6369470396002956e-05}
{"train_loss": 0.06913633644580841, "global_step": 159788, "epoch": 1795, "lr": 3.636891261815621e-05}
{"train_loss": 0.060847122222185135, "global_step": 159789, "epoch": 1795, "lr": 3.6368354842141986e-05}
{"train_loss": 0.0674033910036087, "global_step": 159790, "epoch": 1795, "lr": 3.636779706796037e-05}
{"train_loss": 0.09128358215093613, "global_step": 159791, "epoch": 1795, "lr": 3.636723929561144e-05}
{"train_loss": 0.09564296156167984, "global_step": 159792, "epoch": 1795, "lr": 3.636668152509527e-05}
{"train_loss": 0.15600043535232544, "global_step": 159793, "epoch": 1795, "lr": 3.636612375641193e-05}
{"train_loss": 0.1155099868774414, "global_step": 159794, "epoch": 1795, "lr": 3.6365565989561515e-05}
{"train_loss": 0.0735844150185585, "global_step": 159795, "epoch": 1795, "lr": 3.6365008224544054e-05}
{"train_loss": 0.11197485774755478, "global_step": 159796, "epoch": 1795, "lr": 3.636445046135965e-05}
{"train_loss": 0.07017070055007935, "global_step": 159797, "epoch": 1795, "lr": 3.636389270000841e-05}
{"train_loss": 0.05483430624008179, "global_step": 159798, "epoch": 1795, "lr": 3.636333494049035e-05}
{"train_loss": 0.16875670850276947, "global_step": 159799, "epoch": 1795, "lr": 3.6362777182805594e-05}
{"train_loss": 0.11811506748199463, "global_step": 159800, "epoch": 1795, "lr": 3.636221942695418e-05}
{"train_loss": 0.07136176526546478, "global_step": 159801, "epoch": 1795, "lr": 3.636166167293622e-05}
{"train_loss": 0.07323146611452103, "global_step": 159802, "epoch": 1795, "lr": 3.636110392075175e-05}
{"train_loss": 0.039523646235466, "global_step": 159803, "epoch": 1795, "lr": 3.6360546170400875e-05}
{"train_loss": 0.06595882773399353, "global_step": 159804, "epoch": 1795, "lr": 3.635998842188364e-05}
{"train_loss": 0.13266900181770325, "global_step": 159805, "epoch": 1795, "lr": 3.635943067520017e-05}
{"train_loss": 0.07142901420593262, "global_step": 159806, "epoch": 1795, "lr": 3.6358872930350474e-05}
{"train_loss": 0.09197285771369934, "global_step": 159807, "epoch": 1795, "lr": 3.6358315187334694e-05}
{"train_loss": 0.12492342293262482, "global_step": 159808, "epoch": 1795, "lr": 3.635775744615285e-05}
{"train_loss": 0.03937249630689621, "global_step": 159809, "epoch": 1795, "lr": 3.6357199706805046e-05}
{"train_loss": 0.06640826910734177, "global_step": 159810, "epoch": 1795, "lr": 3.635664196929135e-05}
{"train_loss": 0.053883764892816544, "global_step": 159811, "epoch": 1795, "lr": 3.6356084233611835e-05}
{"train_loss": 0.11490403115749359, "global_step": 159812, "epoch": 1795, "lr": 3.635552649976658e-05}
{"train_loss": 0.1090032160282135, "global_step": 159813, "epoch": 1795, "lr": 3.6354968767755663e-05}
{"train_loss": 0.04802974313497543, "global_step": 159814, "epoch": 1795, "lr": 3.635441103757914e-05}
{"train_loss": 0.10346267372369766, "global_step": 159815, "epoch": 1795, "lr": 3.6353853309237095e-05}
{"train_loss": 0.07903394103050232, "global_step": 159816, "epoch": 1795, "lr": 3.6353295582729634e-05}
{"train_loss": 0.11752462387084961, "global_step": 159817, "epoch": 1795, "lr": 3.6352737858056786e-05}
{"train_loss": 0.10619363933801651, "global_step": 159818, "epoch": 1795, "lr": 3.635218013521866e-05}
{"train_loss": 0.06771320849657059, "global_step": 159819, "epoch": 1795, "lr": 3.6351622414215306e-05}
{"train_loss": 0.06892701238393784, "global_step": 159820, "epoch": 1795, "lr": 3.6351064695046823e-05}
{"train_loss": 0.06046702712774277, "global_step": 159821, "epoch": 1795, "lr": 3.635050697771325e-05}
{"train_loss": 0.0609586238861084, "global_step": 159822, "epoch": 1795, "lr": 3.634994926221471e-05}
{"train_loss": 0.12420692294836044, "global_step": 159823, "epoch": 1795, "lr": 3.634939154855124e-05}
{"train_loss": 0.07384621351957321, "global_step": 159824, "epoch": 1795, "lr": 3.634883383672294e-05}
{"train_loss": 0.12918831408023834, "global_step": 159825, "epoch": 1795, "lr": 3.634827612672985e-05}
{"train_loss": 0.12981466948986053, "global_step": 159826, "epoch": 1795, "lr": 3.63477184185721e-05}
{"train_loss": 0.08846655488014221, "global_step": 159827, "epoch": 1795, "lr": 3.6347160712249704e-05}
{"train_loss": 0.0907868817448616, "global_step": 159828, "epoch": 1795, "lr": 3.634660300776278e-05}
{"train_loss": 0.10595939308404922, "global_step": 159829, "epoch": 1795, "lr": 3.634604530511139e-05}
{"train_loss": 0.09478215873241425, "global_step": 159830, "epoch": 1795, "lr": 3.6345487604295596e-05}
{"train_loss": 0.08915559947490692, "global_step": 159831, "epoch": 1795, "lr": 3.634492990531549e-05}
{"train_loss": 0.11847241222858429, "global_step": 159832, "epoch": 1795, "lr": 3.634437220817115e-05}
{"train_loss": 0.054855234920978546, "global_step": 159833, "epoch": 1795, "lr": 3.634381451286263e-05}
{"train_loss": 0.10053904354572296, "global_step": 159834, "epoch": 1795, "lr": 3.6343256819390015e-05}
{"train_loss": 0.07568614929914474, "global_step": 159835, "epoch": 1795, "lr": 3.6342699127753396e-05}
{"train_loss": 0.07722175121307373, "global_step": 159836, "epoch": 1795, "lr": 3.6342141437952827e-05}
{"train_loss": 0.10070666670799255, "global_step": 159837, "epoch": 1795, "lr": 3.634158374998841e-05}
{"train_loss": 0.12047025561332703, "global_step": 159838, "epoch": 1795, "lr": 3.6341026063860166e-05}
{"train_loss": 0.08850541710853577, "global_step": 159839, "epoch": 1795, "lr": 3.634046837956823e-05}
{"train_loss": 0.12086085230112076, "global_step": 159840, "epoch": 1795, "lr": 3.633991069711264e-05}
{"train_loss": 0.12484806030988693, "global_step": 159841, "epoch": 1795, "lr": 3.633935301649349e-05}
{"train_loss": 0.08430531620979309, "global_step": 159842, "epoch": 1795, "lr": 3.633879533771084e-05}
{"train_loss": 0.09255661467990178, "global_step": 159843, "epoch": 1795, "lr": 3.633823766076478e-05, "val_loss": 6.837348461151123, "train_action_mse_error": 24.666776657104492}
{"train_loss": 0.13080692291259766, "global_step": 159844, "epoch": 1796, "lr": 3.633767998565537e-05}
{"train_loss": 0.1388692408800125, "global_step": 159845, "epoch": 1796, "lr": 3.633712231238271e-05}
{"train_loss": 0.15955036878585815, "global_step": 159846, "epoch": 1796, "lr": 3.633656464094683e-05}
{"train_loss": 0.10628538578748703, "global_step": 159847, "epoch": 1796, "lr": 3.6336006971347855e-05}
{"train_loss": 0.060300979763269424, "global_step": 159848, "epoch": 1796, "lr": 3.6335449303585836e-05}
{"train_loss": 0.1199929267168045, "global_step": 159849, "epoch": 1796, "lr": 3.633489163766084e-05}
{"train_loss": 0.08167517185211182, "global_step": 159850, "epoch": 1796, "lr": 3.633433397357295e-05}
{"train_loss": 0.0946478322148323, "global_step": 159851, "epoch": 1796, "lr": 3.633377631132223e-05}
{"train_loss": 0.06465225666761398, "global_step": 159852, "epoch": 1796, "lr": 3.633321865090879e-05}
{"train_loss": 0.06826166063547134, "global_step": 159853, "epoch": 1796, "lr": 3.633266099233268e-05}
{"train_loss": 0.16698585450649261, "global_step": 159854, "epoch": 1796, "lr": 3.633210333559398e-05}
{"train_loss": 0.10007268190383911, "global_step": 159855, "epoch": 1796, "lr": 3.633154568069274e-05}
{"train_loss": 0.1303386390209198, "global_step": 159856, "epoch": 1796, "lr": 3.633098802762908e-05}
{"train_loss": 0.06226957216858864, "global_step": 159857, "epoch": 1796, "lr": 3.633043037640303e-05}
{"train_loss": 0.09047240763902664, "global_step": 159858, "epoch": 1796, "lr": 3.632987272701471e-05}
{"train_loss": 0.08866900950670242, "global_step": 159859, "epoch": 1796, "lr": 3.632931507946416e-05}
{"train_loss": 0.07025923579931259, "global_step": 159860, "epoch": 1796, "lr": 3.632875743375148e-05}
{"train_loss": 0.08641456067562103, "global_step": 159861, "epoch": 1796, "lr": 3.63281997898767e-05}
{"train_loss": 0.07650631666183472, "global_step": 159862, "epoch": 1796, "lr": 3.632764214783996e-05}
{"train_loss": 0.06895928084850311, "global_step": 159863, "epoch": 1796, "lr": 3.632708450764128e-05}
{"train_loss": 0.07040485739707947, "global_step": 159864, "epoch": 1796, "lr": 3.632652686928078e-05}
{"train_loss": 0.11197519302368164, "global_step": 159865, "epoch": 1796, "lr": 3.632596923275849e-05}
{"train_loss": 0.11458610743284225, "global_step": 159866, "epoch": 1796, "lr": 3.632541159807452e-05}
{"train_loss": 0.11891795694828033, "global_step": 159867, "epoch": 1796, "lr": 3.6324853965228934e-05}
{"train_loss": 0.13116084039211273, "global_step": 159868, "epoch": 1796, "lr": 3.6324296334221796e-05}
{"train_loss": 0.14428292214870453, "global_step": 159869, "epoch": 1796, "lr": 3.632373870505319e-05}
{"train_loss": 0.10762949287891388, "global_step": 159870, "epoch": 1796, "lr": 3.632318107772319e-05}
{"train_loss": 0.06739796698093414, "global_step": 159871, "epoch": 1796, "lr": 3.632262345223187e-05}
{"train_loss": 0.14507293701171875, "global_step": 159872, "epoch": 1796, "lr": 3.632206582857931e-05}
{"train_loss": 0.10898284614086151, "global_step": 159873, "epoch": 1796, "lr": 3.6321508206765584e-05}
{"train_loss": 0.04990840703248978, "global_step": 159874, "epoch": 1796, "lr": 3.632095058679076e-05}
{"train_loss": 0.07952477037906647, "global_step": 159875, "epoch": 1796, "lr": 3.632039296865493e-05}
{"train_loss": 0.07243725657463074, "global_step": 159876, "epoch": 1796, "lr": 3.631983535235814e-05}
{"train_loss": 0.11043121665716171, "global_step": 159877, "epoch": 1796, "lr": 3.631927773790049e-05}
{"train_loss": 0.08564755320549011, "global_step": 159878, "epoch": 1796, "lr": 3.631872012528204e-05}
{"train_loss": 0.11881989240646362, "global_step": 159879, "epoch": 1796, "lr": 3.6318162514502886e-05}
{"train_loss": 0.1116112768650055, "global_step": 159880, "epoch": 1796, "lr": 3.631760490556306e-05}
{"train_loss": 0.09525559097528458, "global_step": 159881, "epoch": 1796, "lr": 3.631704729846269e-05}
{"train_loss": 0.07506895065307617, "global_step": 159882, "epoch": 1796, "lr": 3.631648969320182e-05}
{"train_loss": 0.17972585558891296, "global_step": 159883, "epoch": 1796, "lr": 3.6315932089780535e-05}
{"train_loss": 0.0697348415851593, "global_step": 159884, "epoch": 1796, "lr": 3.6315374488198895e-05}
{"train_loss": 0.16225501894950867, "global_step": 159885, "epoch": 1796, "lr": 3.6314816888456996e-05}
{"train_loss": 0.11405867338180542, "global_step": 159886, "epoch": 1796, "lr": 3.63142592905549e-05}
{"train_loss": 0.08548576384782791, "global_step": 159887, "epoch": 1796, "lr": 3.6313701694492675e-05}
{"train_loss": 0.046776264905929565, "global_step": 159888, "epoch": 1796, "lr": 3.631314410027042e-05}
{"train_loss": 0.10748805850744247, "global_step": 159889, "epoch": 1796, "lr": 3.631258650788819e-05}
{"train_loss": 0.1397208273410797, "global_step": 159890, "epoch": 1796, "lr": 3.631202891734608e-05}
{"train_loss": 0.10849838703870773, "global_step": 159891, "epoch": 1796, "lr": 3.631147132864413e-05}
{"train_loss": 0.14224474132061005, "global_step": 159892, "epoch": 1796, "lr": 3.631091374178245e-05}
{"train_loss": 0.08302739262580872, "global_step": 159893, "epoch": 1796, "lr": 3.631035615676108e-05}
{"train_loss": 0.05574826896190643, "global_step": 159894, "epoch": 1796, "lr": 3.630979857358015e-05}
{"train_loss": 0.09744128584861755, "global_step": 159895, "epoch": 1796, "lr": 3.630924099223967e-05}
{"train_loss": 0.05470938980579376, "global_step": 159896, "epoch": 1796, "lr": 3.630868341273976e-05}
{"train_loss": 0.10671747475862503, "global_step": 159897, "epoch": 1796, "lr": 3.630812583508047e-05}
{"train_loss": 0.14040392637252808, "global_step": 159898, "epoch": 1796, "lr": 3.63075682592619e-05}
{"train_loss": 0.08625803887844086, "global_step": 159899, "epoch": 1796, "lr": 3.6307010685284095e-05}
{"train_loss": 0.07784443348646164, "global_step": 159900, "epoch": 1796, "lr": 3.630645311314716e-05}
{"train_loss": 0.06647765636444092, "global_step": 159901, "epoch": 1796, "lr": 3.630589554285115e-05}
{"train_loss": 0.1097649484872818, "global_step": 159902, "epoch": 1796, "lr": 3.6305337974396156e-05}
{"train_loss": 0.047540780156850815, "global_step": 159903, "epoch": 1796, "lr": 3.6304780407782224e-05}
{"train_loss": 0.08473308384418488, "global_step": 159904, "epoch": 1796, "lr": 3.6304222843009463e-05}
{"train_loss": 0.11980389058589935, "global_step": 159905, "epoch": 1796, "lr": 3.630366528007792e-05}
{"train_loss": 0.11825800687074661, "global_step": 159906, "epoch": 1796, "lr": 3.630310771898768e-05}
{"train_loss": 0.06457307189702988, "global_step": 159907, "epoch": 1796, "lr": 3.630255015973884e-05}
{"train_loss": 0.08394864946603775, "global_step": 159908, "epoch": 1796, "lr": 3.630199260233143e-05}
{"train_loss": 0.11071747541427612, "global_step": 159909, "epoch": 1796, "lr": 3.630143504676558e-05}
{"train_loss": 0.0927557498216629, "global_step": 159910, "epoch": 1796, "lr": 3.63008774930413e-05}
{"train_loss": 0.10231883823871613, "global_step": 159911, "epoch": 1796, "lr": 3.6300319941158725e-05}
{"train_loss": 0.105146624147892, "global_step": 159912, "epoch": 1796, "lr": 3.629976239111789e-05}
{"train_loss": 0.0926814153790474, "global_step": 159913, "epoch": 1796, "lr": 3.62992048429189e-05}
{"train_loss": 0.10848074406385422, "global_step": 159914, "epoch": 1796, "lr": 3.6298647296561793e-05}
{"train_loss": 0.09150438755750656, "global_step": 159915, "epoch": 1796, "lr": 3.6298089752046693e-05}
{"train_loss": 0.07195606827735901, "global_step": 159916, "epoch": 1796, "lr": 3.629753220937363e-05}
{"train_loss": 0.07261960208415985, "global_step": 159917, "epoch": 1796, "lr": 3.629697466854271e-05}
{"train_loss": 0.14341259002685547, "global_step": 159918, "epoch": 1796, "lr": 3.6296417129553974e-05}
{"train_loss": 0.11041104048490524, "global_step": 159919, "epoch": 1796, "lr": 3.6295859592407546e-05}
{"train_loss": 0.06614791601896286, "global_step": 159920, "epoch": 1796, "lr": 3.629530205710344e-05}
{"train_loss": 0.10567014664411545, "global_step": 159921, "epoch": 1796, "lr": 3.6294744523641796e-05}
{"train_loss": 0.10966097563505173, "global_step": 159922, "epoch": 1796, "lr": 3.629418699202263e-05}
{"train_loss": 0.11490258574485779, "global_step": 159923, "epoch": 1796, "lr": 3.629362946224605e-05}
{"train_loss": 0.086866095662117, "global_step": 159924, "epoch": 1796, "lr": 3.629307193431214e-05}
{"train_loss": 0.05535545200109482, "global_step": 159925, "epoch": 1796, "lr": 3.629251440822094e-05}
{"train_loss": 0.061310846358537674, "global_step": 159926, "epoch": 1796, "lr": 3.6291956883972566e-05}
{"train_loss": 0.08333545923233032, "global_step": 159927, "epoch": 1796, "lr": 3.6291399361567045e-05}
{"train_loss": 0.16109532117843628, "global_step": 159928, "epoch": 1796, "lr": 3.6290841841004506e-05}
{"train_loss": 0.09659542143344879, "global_step": 159929, "epoch": 1796, "lr": 3.629028432228497e-05}
{"train_loss": 0.07913047820329666, "global_step": 159930, "epoch": 1796, "lr": 3.628972680540856e-05}
{"train_loss": 0.10035071521997452, "global_step": 159931, "epoch": 1796, "lr": 3.628916929037531e-05}
{"train_loss": 0.09847586066200492, "global_step": 159932, "epoch": 1796, "lr": 3.628861177718533e-05, "val_loss": 6.722408771514893}
{"train_loss": 0.08577345311641693, "global_step": 159933, "epoch": 1797, "lr": 3.6288054265838665e-05}
{"train_loss": 0.1285814493894577, "global_step": 159934, "epoch": 1797, "lr": 3.6287496756335414e-05}
{"train_loss": 0.06786203384399414, "global_step": 159935, "epoch": 1797, "lr": 3.628693924867563e-05}
{"train_loss": 0.12320180982351303, "global_step": 159936, "epoch": 1797, "lr": 3.628638174285942e-05}
{"train_loss": 0.12214662879705429, "global_step": 159937, "epoch": 1797, "lr": 3.628582423888682e-05}
{"train_loss": 0.11095298081636429, "global_step": 159938, "epoch": 1797, "lr": 3.628526673675794e-05}
{"train_loss": 0.07159450650215149, "global_step": 159939, "epoch": 1797, "lr": 3.628470923647282e-05}
{"train_loss": 0.11175262182950974, "global_step": 159940, "epoch": 1797, "lr": 3.628415173803158e-05}
{"train_loss": 0.08477199822664261, "global_step": 159941, "epoch": 1797, "lr": 3.6283594241434236e-05}
{"train_loss": 0.08609263598918915, "global_step": 159942, "epoch": 1797, "lr": 3.628303674668091e-05}
{"train_loss": 0.06942760199308395, "global_step": 159943, "epoch": 1797, "lr": 3.628247925377168e-05}
{"train_loss": 0.0845038890838623, "global_step": 159944, "epoch": 1797, "lr": 3.6281921762706575e-05}
{"train_loss": 0.09725464880466461, "global_step": 159945, "epoch": 1797, "lr": 3.6281364273485725e-05}
{"train_loss": 0.04059545323252678, "global_step": 159946, "epoch": 1797, "lr": 3.628080678610916e-05}
{"train_loss": 0.14173880219459534, "global_step": 159947, "epoch": 1797, "lr": 3.6280249300576986e-05}
{"train_loss": 0.10259787738323212, "global_step": 159948, "epoch": 1797, "lr": 3.627969181688924e-05}
{"train_loss": 0.057688646018505096, "global_step": 159949, "epoch": 1797, "lr": 3.627913433504605e-05}
{"train_loss": 0.06779097765684128, "global_step": 159950, "epoch": 1797, "lr": 3.627857685504744e-05}
{"train_loss": 0.10568039119243622, "global_step": 159951, "epoch": 1797, "lr": 3.6278019376893537e-05}
{"train_loss": 0.11933080852031708, "global_step": 159952, "epoch": 1797, "lr": 3.627746190058436e-05}
{"train_loss": 0.09418187290430069, "global_step": 159953, "epoch": 1797, "lr": 3.627690442612003e-05}
{"train_loss": 0.08049867302179337, "global_step": 159954, "epoch": 1797, "lr": 3.627634695350059e-05}
{"train_loss": 0.11394015699625015, "global_step": 159955, "epoch": 1797, "lr": 3.627578948272614e-05}
{"train_loss": 0.11035530269145966, "global_step": 159956, "epoch": 1797, "lr": 3.6275232013796726e-05}
{"train_loss": 0.0769762173295021, "global_step": 159957, "epoch": 1797, "lr": 3.6274674546712465e-05}
{"train_loss": 0.07432691007852554, "global_step": 159958, "epoch": 1797, "lr": 3.6274117081473374e-05}
{"train_loss": 0.055548664182424545, "global_step": 159959, "epoch": 1797, "lr": 3.627355961807959e-05}
{"train_loss": 0.11234048008918762, "global_step": 159960, "epoch": 1797, "lr": 3.6273002156531135e-05}
{"train_loss": 0.0845184251666069, "global_step": 159961, "epoch": 1797, "lr": 3.6272444696828116e-05}
{"train_loss": 0.04627741500735283, "global_step": 159962, "epoch": 1797, "lr": 3.6271887238970605e-05}
{"train_loss": 0.04766722023487091, "global_step": 159963, "epoch": 1797, "lr": 3.627132978295867e-05}
{"train_loss": 0.1036856472492218, "global_step": 159964, "epoch": 1797, "lr": 3.627077232879239e-05}
{"train_loss": 0.07246869802474976, "global_step": 159965, "epoch": 1797, "lr": 3.627021487647183e-05}
{"train_loss": 0.11015148460865021, "global_step": 159966, "epoch": 1797, "lr": 3.6269657425997086e-05}
{"train_loss": 0.13687767088413239, "global_step": 159967, "epoch": 1797, "lr": 3.6269099977368196e-05}
{"train_loss": 0.1141943708062172, "global_step": 159968, "epoch": 1797, "lr": 3.626854253058528e-05}
{"train_loss": 0.04675474390387535, "global_step": 159969, "epoch": 1797, "lr": 3.626798508564837e-05}
{"train_loss": 0.07587880641222, "global_step": 159970, "epoch": 1797, "lr": 3.6267427642557586e-05}
{"train_loss": 0.09856278449296951, "global_step": 159971, "epoch": 1797, "lr": 3.626687020131296e-05}
{"train_loss": 0.08457222580909729, "global_step": 159972, "epoch": 1797, "lr": 3.62663127619146e-05}
{"train_loss": 0.07055623829364777, "global_step": 159973, "epoch": 1797, "lr": 3.626575532436255e-05}
{"train_loss": 0.07224826514720917, "global_step": 159974, "epoch": 1797, "lr": 3.6265197888656924e-05}
{"train_loss": 0.121226966381073, "global_step": 159975, "epoch": 1797, "lr": 3.626464045479775e-05}
{"train_loss": 0.11252909898757935, "global_step": 159976, "epoch": 1797, "lr": 3.6264083022785157e-05}
{"train_loss": 0.11510033160448074, "global_step": 159977, "epoch": 1797, "lr": 3.6263525592619166e-05}
{"train_loss": 0.14825065433979034, "global_step": 159978, "epoch": 1797, "lr": 3.626296816429988e-05}
{"train_loss": 0.058540236204862595, "global_step": 159979, "epoch": 1797, "lr": 3.6262410737827376e-05}
{"train_loss": 0.07932613044977188, "global_step": 159980, "epoch": 1797, "lr": 3.626185331320172e-05}
{"train_loss": 0.0828157365322113, "global_step": 159981, "epoch": 1797, "lr": 3.6261295890422994e-05}
{"train_loss": 0.09179399162530899, "global_step": 159982, "epoch": 1797, "lr": 3.626073846949127e-05}
{"train_loss": 0.10358067601919174, "global_step": 159983, "epoch": 1797, "lr": 3.626018105040663e-05}
{"train_loss": 0.148318812251091, "global_step": 159984, "epoch": 1797, "lr": 3.625962363316911e-05}
{"train_loss": 0.08391670882701874, "global_step": 159985, "epoch": 1797, "lr": 3.625906621777886e-05}
{"train_loss": 0.1193212941288948, "global_step": 159986, "epoch": 1797, "lr": 3.6258508804235876e-05}
{"train_loss": 0.09284064173698425, "global_step": 159987, "epoch": 1797, "lr": 3.625795139254029e-05}
{"train_loss": 0.07140421867370605, "global_step": 159988, "epoch": 1797, "lr": 3.6257393982692135e-05}
{"train_loss": 0.08300445228815079, "global_step": 159989, "epoch": 1797, "lr": 3.6256836574691526e-05}
{"train_loss": 0.10079453885555267, "global_step": 159990, "epoch": 1797, "lr": 3.62562791685385e-05}
{"train_loss": 0.05307380482554436, "global_step": 159991, "epoch": 1797, "lr": 3.625572176423316e-05}
{"train_loss": 0.1137368306517601, "global_step": 159992, "epoch": 1797, "lr": 3.6255164361775564e-05}
{"train_loss": 0.0936216488480568, "global_step": 159993, "epoch": 1797, "lr": 3.625460696116581e-05}
{"train_loss": 0.037380050867795944, "global_step": 159994, "epoch": 1797, "lr": 3.6254049562403934e-05}
{"train_loss": 0.13484162092208862, "global_step": 159995, "epoch": 1797, "lr": 3.625349216549006e-05}
{"train_loss": 0.11330468952655792, "global_step": 159996, "epoch": 1797, "lr": 3.625293477042421e-05}
{"train_loss": 0.06825421750545502, "global_step": 159997, "epoch": 1797, "lr": 3.6252377377206495e-05}
{"train_loss": 0.16272643208503723, "global_step": 159998, "epoch": 1797, "lr": 3.625181998583698e-05}
{"train_loss": 0.07224962115287781, "global_step": 159999, "epoch": 1797, "lr": 3.625126259631574e-05}
{"train_loss": 0.11715900897979736, "global_step": 160000, "epoch": 1797, "lr": 3.625070520864285e-05}
{"train_loss": 0.04693910479545593, "global_step": 160001, "epoch": 1797, "lr": 3.625014782281839e-05}
{"train_loss": 0.15824133157730103, "global_step": 160002, "epoch": 1797, "lr": 3.624959043884244e-05}
{"train_loss": 0.14350435137748718, "global_step": 160003, "epoch": 1797, "lr": 3.6249033056715044e-05}
{"train_loss": 0.1999405026435852, "global_step": 160004, "epoch": 1797, "lr": 3.624847567643631e-05}
{"train_loss": 0.07788302004337311, "global_step": 160005, "epoch": 1797, "lr": 3.624791829800629e-05}
{"train_loss": 0.10719407349824905, "global_step": 160006, "epoch": 1797, "lr": 3.624736092142509e-05}
{"train_loss": 0.06758203357458115, "global_step": 160007, "epoch": 1797, "lr": 3.624680354669274e-05}
{"train_loss": 0.03539981320500374, "global_step": 160008, "epoch": 1797, "lr": 3.624624617380936e-05}
{"train_loss": 0.11170970648527145, "global_step": 160009, "epoch": 1797, "lr": 3.624568880277498e-05}
{"train_loss": 0.09290821850299835, "global_step": 160010, "epoch": 1797, "lr": 3.624513143358973e-05}
{"train_loss": 0.09380887448787689, "global_step": 160011, "epoch": 1797, "lr": 3.624457406625363e-05}
{"train_loss": 0.07645994424819946, "global_step": 160012, "epoch": 1797, "lr": 3.6244016700766794e-05}
{"train_loss": 0.07851247489452362, "global_step": 160013, "epoch": 1797, "lr": 3.624345933712927e-05}
{"train_loss": 0.10152849555015564, "global_step": 160014, "epoch": 1797, "lr": 3.624290197534114e-05}
{"train_loss": 0.10304798930883408, "global_step": 160015, "epoch": 1797, "lr": 3.624234461540251e-05}
{"train_loss": 0.0859118402004242, "global_step": 160016, "epoch": 1797, "lr": 3.624178725731342e-05}
{"train_loss": 0.03568105399608612, "global_step": 160017, "epoch": 1797, "lr": 3.624122990107395e-05}
{"train_loss": 0.14570027589797974, "global_step": 160018, "epoch": 1797, "lr": 3.6240672546684175e-05}
{"train_loss": 0.2231801599264145, "global_step": 160019, "epoch": 1797, "lr": 3.624011519414417e-05}
{"train_loss": 0.05522899702191353, "global_step": 160020, "epoch": 1797, "lr": 3.623955784345402e-05}
{"train_loss": 0.09635554609841175, "global_step": 160021, "epoch": 1797, "lr": 3.6239000494613814e-05, "val_loss": 6.678478240966797}
{"train_loss": 0.08679842948913574, "global_step": 160022, "epoch": 1798, "lr": 3.623844314762357e-05}
{"train_loss": 0.10256461799144745, "global_step": 160023, "epoch": 1798, "lr": 3.623788580248343e-05}
{"train_loss": 0.12285137176513672, "global_step": 160024, "epoch": 1798, "lr": 3.623732845919342e-05}
{"train_loss": 0.09857674688100815, "global_step": 160025, "epoch": 1798, "lr": 3.6236771117753645e-05}
{"train_loss": 0.053372398018836975, "global_step": 160026, "epoch": 1798, "lr": 3.6236213778164154e-05}
{"train_loss": 0.08933274447917938, "global_step": 160027, "epoch": 1798, "lr": 3.6235656440425056e-05}
{"train_loss": 0.07313668727874756, "global_step": 160028, "epoch": 1798, "lr": 3.623509910453639e-05}
{"train_loss": 0.10745218396186829, "global_step": 160029, "epoch": 1798, "lr": 3.6234541770498254e-05}
{"train_loss": 0.11245105415582657, "global_step": 160030, "epoch": 1798, "lr": 3.6233984438310706e-05}
{"train_loss": 0.05234838277101517, "global_step": 160031, "epoch": 1798, "lr": 3.623342710797385e-05}
{"train_loss": 0.08123322576284409, "global_step": 160032, "epoch": 1798, "lr": 3.623286977948772e-05}
{"train_loss": 0.08723253756761551, "global_step": 160033, "epoch": 1798, "lr": 3.623231245285241e-05}
{"train_loss": 0.09854647517204285, "global_step": 160034, "epoch": 1798, "lr": 3.6231755128068025e-05}
{"train_loss": 0.06166408956050873, "global_step": 160035, "epoch": 1798, "lr": 3.6231197805134596e-05}
{"train_loss": 0.0787125825881958, "global_step": 160036, "epoch": 1798, "lr": 3.6230640484052225e-05}
{"train_loss": 0.11271757632493973, "global_step": 160037, "epoch": 1798, "lr": 3.6230083164820974e-05}
{"train_loss": 0.1522100865840912, "global_step": 160038, "epoch": 1798, "lr": 3.622952584744091e-05}
{"train_loss": 0.08584298938512802, "global_step": 160039, "epoch": 1798, "lr": 3.622896853191212e-05}
{"train_loss": 0.12148293107748032, "global_step": 160040, "epoch": 1798, "lr": 3.6228411218234695e-05}
{"train_loss": 0.1396619975566864, "global_step": 160041, "epoch": 1798, "lr": 3.6227853906408673e-05}
{"train_loss": 0.10281466692686081, "global_step": 160042, "epoch": 1798, "lr": 3.6227296596434164e-05}
{"train_loss": 0.10114691406488419, "global_step": 160043, "epoch": 1798, "lr": 3.622673928831121e-05}
{"train_loss": 0.08086828142404556, "global_step": 160044, "epoch": 1798, "lr": 3.622618198203992e-05}
{"train_loss": 0.10583218187093735, "global_step": 160045, "epoch": 1798, "lr": 3.622562467762033e-05}
{"train_loss": 0.04891344904899597, "global_step": 160046, "epoch": 1798, "lr": 3.622506737505256e-05}
{"train_loss": 0.0841042622923851, "global_step": 160047, "epoch": 1798, "lr": 3.6224510074336634e-05}
{"train_loss": 0.14972075819969177, "global_step": 160048, "epoch": 1798, "lr": 3.622395277547269e-05}
{"train_loss": 0.10700114816427231, "global_step": 160049, "epoch": 1798, "lr": 3.622339547846074e-05}
{"train_loss": 0.10129030793905258, "global_step": 160050, "epoch": 1798, "lr": 3.622283818330088e-05}
{"train_loss": 0.06837017089128494, "global_step": 160051, "epoch": 1798, "lr": 3.6222280889993224e-05}
{"train_loss": 0.09178958088159561, "global_step": 160052, "epoch": 1798, "lr": 3.622172359853778e-05}
{"train_loss": 0.08592328429222107, "global_step": 160053, "epoch": 1798, "lr": 3.6221166308934686e-05}
{"train_loss": 0.08588388562202454, "global_step": 160054, "epoch": 1798, "lr": 3.6220609021183984e-05}
{"train_loss": 0.11178506165742874, "global_step": 160055, "epoch": 1798, "lr": 3.622005173528574e-05}
{"train_loss": 0.06982528418302536, "global_step": 160056, "epoch": 1798, "lr": 3.6219494451240056e-05}
{"train_loss": 0.0727110356092453, "global_step": 160057, "epoch": 1798, "lr": 3.621893716904699e-05}
{"train_loss": 0.09763272106647491, "global_step": 160058, "epoch": 1798, "lr": 3.621837988870661e-05}
{"train_loss": 0.09415247291326523, "global_step": 160059, "epoch": 1798, "lr": 3.621782261021902e-05}
{"train_loss": 0.0962221547961235, "global_step": 160060, "epoch": 1798, "lr": 3.621726533358425e-05}
{"train_loss": 0.11083552241325378, "global_step": 160061, "epoch": 1798, "lr": 3.621670805880242e-05}
{"train_loss": 0.12481050193309784, "global_step": 160062, "epoch": 1798, "lr": 3.621615078587357e-05}
{"train_loss": 0.05049954727292061, "global_step": 160063, "epoch": 1798, "lr": 3.6215593514797806e-05}
{"train_loss": 0.15614724159240723, "global_step": 160064, "epoch": 1798, "lr": 3.621503624557518e-05}
{"train_loss": 0.13256002962589264, "global_step": 160065, "epoch": 1798, "lr": 3.621447897820578e-05}
{"train_loss": 0.0389576181769371, "global_step": 160066, "epoch": 1798, "lr": 3.621392171268966e-05}
{"train_loss": 0.05827799439430237, "global_step": 160067, "epoch": 1798, "lr": 3.6213364449026934e-05}
{"train_loss": 0.10350178182125092, "global_step": 160068, "epoch": 1798, "lr": 3.6212807187217625e-05}
{"train_loss": 0.08160832524299622, "global_step": 160069, "epoch": 1798, "lr": 3.621224992726184e-05}
{"train_loss": 0.11227596551179886, "global_step": 160070, "epoch": 1798, "lr": 3.621169266915967e-05}
{"train_loss": 0.15964311361312866, "global_step": 160071, "epoch": 1798, "lr": 3.621113541291116e-05}
{"train_loss": 0.09747859090566635, "global_step": 160072, "epoch": 1798, "lr": 3.62105781585164e-05}
{"train_loss": 0.11129890382289886, "global_step": 160073, "epoch": 1798, "lr": 3.621002090597544e-05}
{"train_loss": 0.17352871596813202, "global_step": 160074, "epoch": 1798, "lr": 3.62094636552884e-05}
{"train_loss": 0.1078854650259018, "global_step": 160075, "epoch": 1798, "lr": 3.620890640645532e-05}
{"train_loss": 0.1165112778544426, "global_step": 160076, "epoch": 1798, "lr": 3.6208349159476284e-05}
{"train_loss": 0.1136777251958847, "global_step": 160077, "epoch": 1798, "lr": 3.620779191435137e-05}
{"train_loss": 0.10931076109409332, "global_step": 160078, "epoch": 1798, "lr": 3.620723467108065e-05}
{"train_loss": 0.10440915822982788, "global_step": 160079, "epoch": 1798, "lr": 3.6206677429664195e-05}
{"train_loss": 0.10517311841249466, "global_step": 160080, "epoch": 1798, "lr": 3.620612019010209e-05}
{"train_loss": 0.12089775502681732, "global_step": 160081, "epoch": 1798, "lr": 3.620556295239439e-05}
{"train_loss": 0.09830892086029053, "global_step": 160082, "epoch": 1798, "lr": 3.620500571654121e-05}
{"train_loss": 0.11154835671186447, "global_step": 160083, "epoch": 1798, "lr": 3.620444848254257e-05}
{"train_loss": 0.0984681025147438, "global_step": 160084, "epoch": 1798, "lr": 3.620389125039859e-05}
{"train_loss": 0.09287223219871521, "global_step": 160085, "epoch": 1798, "lr": 3.620333402010932e-05}
{"train_loss": 0.09637091308832169, "global_step": 160086, "epoch": 1798, "lr": 3.620277679167486e-05}
{"train_loss": 0.08626032620668411, "global_step": 160087, "epoch": 1798, "lr": 3.620221956509524e-05}
{"train_loss": 0.15142765641212463, "global_step": 160088, "epoch": 1798, "lr": 3.620166234037057e-05}
{"train_loss": 0.10954964905977249, "global_step": 160089, "epoch": 1798, "lr": 3.620110511750094e-05}
{"train_loss": 0.10839772969484329, "global_step": 160090, "epoch": 1798, "lr": 3.620054789648638e-05}
{"train_loss": 0.07535511255264282, "global_step": 160091, "epoch": 1798, "lr": 3.6199990677327015e-05}
{"train_loss": 0.08833940327167511, "global_step": 160092, "epoch": 1798, "lr": 3.619943346002287e-05}
{"train_loss": 0.11322419345378876, "global_step": 160093, "epoch": 1798, "lr": 3.619887624457406e-05}
{"train_loss": 0.07372792810201645, "global_step": 160094, "epoch": 1798, "lr": 3.619831903098063e-05}
{"train_loss": 0.09080750495195389, "global_step": 160095, "epoch": 1798, "lr": 3.619776181924267e-05}
{"train_loss": 0.13374197483062744, "global_step": 160096, "epoch": 1798, "lr": 3.6197204609360256e-05}
{"train_loss": 0.14982186257839203, "global_step": 160097, "epoch": 1798, "lr": 3.619664740133346e-05}
{"train_loss": 0.11603011190891266, "global_step": 160098, "epoch": 1798, "lr": 3.619609019516236e-05}
{"train_loss": 0.11769262701272964, "global_step": 160099, "epoch": 1798, "lr": 3.619553299084702e-05}
{"train_loss": 0.07013952732086182, "global_step": 160100, "epoch": 1798, "lr": 3.6194975788387516e-05}
{"train_loss": 0.07989202439785004, "global_step": 160101, "epoch": 1798, "lr": 3.6194418587783954e-05}
{"train_loss": 0.071186363697052, "global_step": 160102, "epoch": 1798, "lr": 3.619386138903636e-05}
{"train_loss": 0.11041706800460815, "global_step": 160103, "epoch": 1798, "lr": 3.619330419214485e-05}
{"train_loss": 0.04744521155953407, "global_step": 160104, "epoch": 1798, "lr": 3.619274699710946e-05}
{"train_loss": 0.11060378700494766, "global_step": 160105, "epoch": 1798, "lr": 3.6192189803930285e-05}
{"train_loss": 0.06500016897916794, "global_step": 160106, "epoch": 1798, "lr": 3.6191632612607426e-05}
{"train_loss": 0.06466162204742432, "global_step": 160107, "epoch": 1798, "lr": 3.619107542314092e-05}
{"train_loss": 0.06145986169576645, "global_step": 160108, "epoch": 1798, "lr": 3.619051823553087e-05}
{"train_loss": 0.05221832916140556, "global_step": 160109, "epoch": 1798, "lr": 3.618996104977731e-05}
{"train_loss": 0.09784944426645054, "global_step": 160110, "epoch": 1798, "lr": 3.618940386588037e-05, "val_loss": 6.587347507476807}
{"train_loss": 0.09949298948049545, "global_step": 160111, "epoch": 1799, "lr": 3.618884668384007e-05}
{"train_loss": 0.09160779416561127, "global_step": 160112, "epoch": 1799, "lr": 3.6188289503656535e-05}
{"train_loss": 0.12570853531360626, "global_step": 160113, "epoch": 1799, "lr": 3.618773232532981e-05}
{"train_loss": 0.07184047996997833, "global_step": 160114, "epoch": 1799, "lr": 3.618717514885997e-05}
{"train_loss": 0.2068614363670349, "global_step": 160115, "epoch": 1799, "lr": 3.61866179742471e-05}
{"train_loss": 0.08635088801383972, "global_step": 160116, "epoch": 1799, "lr": 3.618606080149127e-05}
{"train_loss": 0.08615930378437042, "global_step": 160117, "epoch": 1799, "lr": 3.618550363059255e-05}
{"train_loss": 0.12379439920186996, "global_step": 160118, "epoch": 1799, "lr": 3.618494646155104e-05}
{"train_loss": 0.09251369535923004, "global_step": 160119, "epoch": 1799, "lr": 3.6184389294366774e-05}
{"train_loss": 0.10056649148464203, "global_step": 160120, "epoch": 1799, "lr": 3.6183832129039864e-05}
{"train_loss": 0.08319962024688721, "global_step": 160121, "epoch": 1799, "lr": 3.618327496557035e-05}
{"train_loss": 0.09243062138557434, "global_step": 160122, "epoch": 1799, "lr": 3.618271780395834e-05}
{"train_loss": 0.0897284746170044, "global_step": 160123, "epoch": 1799, "lr": 3.6182160644203886e-05}
{"train_loss": 0.07439398020505905, "global_step": 160124, "epoch": 1799, "lr": 3.618160348630707e-05}
{"train_loss": 0.13081666827201843, "global_step": 160125, "epoch": 1799, "lr": 3.618104633026799e-05}
{"train_loss": 0.14390844106674194, "global_step": 160126, "epoch": 1799, "lr": 3.618048917608667e-05}
{"train_loss": 0.06527212262153625, "global_step": 160127, "epoch": 1799, "lr": 3.617993202376325e-05}
{"train_loss": 0.12118933349847794, "global_step": 160128, "epoch": 1799, "lr": 3.617937487329774e-05}
{"train_loss": 0.08733375370502472, "global_step": 160129, "epoch": 1799, "lr": 3.6178817724690266e-05}
{"train_loss": 0.08131782710552216, "global_step": 160130, "epoch": 1799, "lr": 3.617826057794086e-05}
{"train_loss": 0.08582226186990738, "global_step": 160131, "epoch": 1799, "lr": 3.6177703433049646e-05}
{"train_loss": 0.11460752040147781, "global_step": 160132, "epoch": 1799, "lr": 3.617714629001665e-05}
{"train_loss": 0.05487972870469093, "global_step": 160133, "epoch": 1799, "lr": 3.6176589148841976e-05}
{"train_loss": 0.057176776230335236, "global_step": 160134, "epoch": 1799, "lr": 3.6176032009525687e-05}
{"train_loss": 0.05734322965145111, "global_step": 160135, "epoch": 1799, "lr": 3.617547487206787e-05}
{"train_loss": 0.09059128910303116, "global_step": 160136, "epoch": 1799, "lr": 3.617491773646858e-05}
{"train_loss": 0.07128745317459106, "global_step": 160137, "epoch": 1799, "lr": 3.617436060272792e-05}
{"train_loss": 0.10796672105789185, "global_step": 160138, "epoch": 1799, "lr": 3.617380347084593e-05}
{"train_loss": 0.0788777619600296, "global_step": 160139, "epoch": 1799, "lr": 3.617324634082272e-05}
{"train_loss": 0.061873238533735275, "global_step": 160140, "epoch": 1799, "lr": 3.617268921265833e-05}
{"train_loss": 0.09398873150348663, "global_step": 160141, "epoch": 1799, "lr": 3.617213208635285e-05}
{"train_loss": 0.04021347314119339, "global_step": 160142, "epoch": 1799, "lr": 3.617157496190638e-05}
{"train_loss": 0.08413335680961609, "global_step": 160143, "epoch": 1799, "lr": 3.617101783931896e-05}
{"train_loss": 0.10731500387191772, "global_step": 160144, "epoch": 1799, "lr": 3.617046071859069e-05}
{"train_loss": 0.12241993844509125, "global_step": 160145, "epoch": 1799, "lr": 3.616990359972162e-05}
{"train_loss": 0.08457153290510178, "global_step": 160146, "epoch": 1799, "lr": 3.616934648271186e-05}
{"train_loss": 0.08697793632745743, "global_step": 160147, "epoch": 1799, "lr": 3.616878936756143e-05}
{"train_loss": 0.05987028405070305, "global_step": 160148, "epoch": 1799, "lr": 3.6168232254270466e-05}
{"train_loss": 0.09436672180891037, "global_step": 160149, "epoch": 1799, "lr": 3.6167675142838995e-05}
{"train_loss": 0.0763150304555893, "global_step": 160150, "epoch": 1799, "lr": 3.6167118033267135e-05}
{"train_loss": 0.11578966677188873, "global_step": 160151, "epoch": 1799, "lr": 3.616656092555492e-05}
{"train_loss": 0.11543772369623184, "global_step": 160152, "epoch": 1799, "lr": 3.616600381970244e-05}
{"train_loss": 0.12153784185647964, "global_step": 160153, "epoch": 1799, "lr": 3.616544671570978e-05}
{"train_loss": 0.08757724612951279, "global_step": 160154, "epoch": 1799, "lr": 3.6164889613577e-05}
{"train_loss": 0.08002050966024399, "global_step": 160155, "epoch": 1799, "lr": 3.616433251330419e-05}
{"train_loss": 0.13537314534187317, "global_step": 160156, "epoch": 1799, "lr": 3.616377541489142e-05}
{"train_loss": 0.0811501145362854, "global_step": 160157, "epoch": 1799, "lr": 3.616321831833874e-05}
{"train_loss": 0.1496174931526184, "global_step": 160158, "epoch": 1799, "lr": 3.6162661223646275e-05}
{"train_loss": 0.08213034272193909, "global_step": 160159, "epoch": 1799, "lr": 3.616210413081404e-05}
{"train_loss": 0.0676780715584755, "global_step": 160160, "epoch": 1799, "lr": 3.616154703984216e-05}
{"train_loss": 0.08151210099458694, "global_step": 160161, "epoch": 1799, "lr": 3.6160989950730696e-05}
{"train_loss": 0.10015583038330078, "global_step": 160162, "epoch": 1799, "lr": 3.6160432863479706e-05}
{"train_loss": 0.1586833894252777, "global_step": 160163, "epoch": 1799, "lr": 3.6159875778089294e-05}
{"train_loss": 0.13829880952835083, "global_step": 160164, "epoch": 1799, "lr": 3.6159318694559496e-05}
{"train_loss": 0.13761480152606964, "global_step": 160165, "epoch": 1799, "lr": 3.615876161289043e-05}
{"train_loss": 0.1865050047636032, "global_step": 160166, "epoch": 1799, "lr": 3.615820453308213e-05}
{"train_loss": 0.07193250954151154, "global_step": 160167, "epoch": 1799, "lr": 3.615764745513471e-05}
{"train_loss": 0.14584125578403473, "global_step": 160168, "epoch": 1799, "lr": 3.6157090379048205e-05}
{"train_loss": 0.126206174492836, "global_step": 160169, "epoch": 1799, "lr": 3.615653330482274e-05}
{"train_loss": 0.09012684971094131, "global_step": 160170, "epoch": 1799, "lr": 3.615597623245834e-05}
{"train_loss": 0.10289476811885834, "global_step": 160171, "epoch": 1799, "lr": 3.61554191619551e-05}
{"train_loss": 0.13112829625606537, "global_step": 160172, "epoch": 1799, "lr": 3.6154862093313105e-05}
{"train_loss": 0.14464972913265228, "global_step": 160173, "epoch": 1799, "lr": 3.615430502653241e-05}
{"train_loss": 0.12852612137794495, "global_step": 160174, "epoch": 1799, "lr": 3.61537479616131e-05}
{"train_loss": 0.062114130705595016, "global_step": 160175, "epoch": 1799, "lr": 3.6153190898555255e-05}
{"train_loss": 0.1263677179813385, "global_step": 160176, "epoch": 1799, "lr": 3.615263383735894e-05}
{"train_loss": 0.06830528378486633, "global_step": 160177, "epoch": 1799, "lr": 3.615207677802422e-05}
{"train_loss": 0.04982052370905876, "global_step": 160178, "epoch": 1799, "lr": 3.6151519720551216e-05}
{"train_loss": 0.08269163221120834, "global_step": 160179, "epoch": 1799, "lr": 3.615096266493995e-05}
{"train_loss": 0.1138993352651596, "global_step": 160180, "epoch": 1799, "lr": 3.6150405611190534e-05}
{"train_loss": 0.09596738964319229, "global_step": 160181, "epoch": 1799, "lr": 3.614984855930301e-05}
{"train_loss": 0.09550999104976654, "global_step": 160182, "epoch": 1799, "lr": 3.614929150927749e-05}
{"train_loss": 0.15783604979515076, "global_step": 160183, "epoch": 1799, "lr": 3.6148734461114004e-05}
{"train_loss": 0.051480211317539215, "global_step": 160184, "epoch": 1799, "lr": 3.614817741481268e-05}
{"train_loss": 0.15660414099693298, "global_step": 160185, "epoch": 1799, "lr": 3.614762037037354e-05}
{"train_loss": 0.09024450927972794, "global_step": 160186, "epoch": 1799, "lr": 3.61470633277967e-05}
{"train_loss": 0.07293727248907089, "global_step": 160187, "epoch": 1799, "lr": 3.614650628708221e-05}
{"train_loss": 0.08991678804159164, "global_step": 160188, "epoch": 1799, "lr": 3.614594924823016e-05}
{"train_loss": 0.06862196326255798, "global_step": 160189, "epoch": 1799, "lr": 3.6145392211240604e-05}
{"train_loss": 0.0747767984867096, "global_step": 160190, "epoch": 1799, "lr": 3.614483517611364e-05}
{"train_loss": 0.09290356189012527, "global_step": 160191, "epoch": 1799, "lr": 3.614427814284934e-05}
{"train_loss": 0.0653749629855156, "global_step": 160192, "epoch": 1799, "lr": 3.614372111144777e-05}
{"train_loss": 0.0567835308611393, "global_step": 160193, "epoch": 1799, "lr": 3.614316408190901e-05}
{"train_loss": 0.08539260178804398, "global_step": 160194, "epoch": 1799, "lr": 3.614260705423312e-05}
{"train_loss": 0.1286792904138565, "global_step": 160195, "epoch": 1799, "lr": 3.61420500284202e-05}
{"train_loss": 0.09560278058052063, "global_step": 160196, "epoch": 1799, "lr": 3.6141493004470295e-05}
{"train_loss": 0.07809540629386902, "global_step": 160197, "epoch": 1799, "lr": 3.614093598238352e-05}
{"train_loss": 0.07717211544513702, "global_step": 160198, "epoch": 1799, "lr": 3.614037896215991e-05}
{"train_loss": 0.09750754059700484, "global_step": 160199, "epoch": 1799, "lr": 3.613982194379958e-05, "val_loss": 6.809102535247803}
{"train_loss": 0.06914325058460236, "global_step": 160200, "epoch": 1800, "lr": 3.613926492730255e-05}
{"train_loss": 0.10802342742681503, "global_step": 160201, "epoch": 1800, "lr": 3.6138707912668954e-05}
{"train_loss": 0.08137310296297073, "global_step": 160202, "epoch": 1800, "lr": 3.6138150899898824e-05}
{"train_loss": 0.13825680315494537, "global_step": 160203, "epoch": 1800, "lr": 3.613759388899227e-05}
{"train_loss": 0.1327032893896103, "global_step": 160204, "epoch": 1800, "lr": 3.613703687994931e-05}
{"train_loss": 0.06642094254493713, "global_step": 160205, "epoch": 1800, "lr": 3.61364798727701e-05}
{"train_loss": 0.09694603830575943, "global_step": 160206, "epoch": 1800, "lr": 3.613592286745464e-05}
{"train_loss": 0.07924024760723114, "global_step": 160207, "epoch": 1800, "lr": 3.613536586400306e-05}
{"train_loss": 0.11113189905881882, "global_step": 160208, "epoch": 1800, "lr": 3.6134808862415384e-05}
{"train_loss": 0.10603272169828415, "global_step": 160209, "epoch": 1800, "lr": 3.6134251862691744e-05}
{"train_loss": 0.10278249531984329, "global_step": 160210, "epoch": 1800, "lr": 3.6133694864832165e-05}
{"train_loss": 0.10081866383552551, "global_step": 160211, "epoch": 1800, "lr": 3.613313786883675e-05}
{"train_loss": 0.1267542839050293, "global_step": 160212, "epoch": 1800, "lr": 3.613258087470556e-05}
{"train_loss": 0.0893922671675682, "global_step": 160213, "epoch": 1800, "lr": 3.613202388243868e-05}
{"train_loss": 0.05567288398742676, "global_step": 160214, "epoch": 1800, "lr": 3.613146689203618e-05}
{"train_loss": 0.0954945832490921, "global_step": 160215, "epoch": 1800, "lr": 3.613090990349812e-05}
{"train_loss": 0.079283706843853, "global_step": 160216, "epoch": 1800, "lr": 3.613035291682462e-05}
{"train_loss": 0.07743881642818451, "global_step": 160217, "epoch": 1800, "lr": 3.612979593201569e-05}
{"train_loss": 0.11347275972366333, "global_step": 160218, "epoch": 1800, "lr": 3.612923894907147e-05}
{"train_loss": 0.07838980853557587, "global_step": 160219, "epoch": 1800, "lr": 3.612868196799198e-05}
{"train_loss": 0.07919192314147949, "global_step": 160220, "epoch": 1800, "lr": 3.612812498877735e-05}
{"train_loss": 0.17379166185855865, "global_step": 160221, "epoch": 1800, "lr": 3.6127568011427595e-05}
{"train_loss": 0.058371830731630325, "global_step": 160222, "epoch": 1800, "lr": 3.612701103594284e-05}
{"train_loss": 0.10407643020153046, "global_step": 160223, "epoch": 1800, "lr": 3.612645406232312e-05}
{"train_loss": 0.04983513057231903, "global_step": 160224, "epoch": 1800, "lr": 3.612589709056855e-05}
{"train_loss": 0.09056523442268372, "global_step": 160225, "epoch": 1800, "lr": 3.612534012067916e-05}
{"train_loss": 0.15226005017757416, "global_step": 160226, "epoch": 1800, "lr": 3.612478315265507e-05}
{"train_loss": 0.09180723130702972, "global_step": 160227, "epoch": 1800, "lr": 3.6124226186496326e-05}
{"train_loss": 0.13250672817230225, "global_step": 160228, "epoch": 1800, "lr": 3.612366922220302e-05}
{"train_loss": 0.10521722584962845, "global_step": 160229, "epoch": 1800, "lr": 3.6123112259775206e-05}
{"train_loss": 0.08480913192033768, "global_step": 160230, "epoch": 1800, "lr": 3.6122555299212975e-05}
{"train_loss": 0.10259819030761719, "global_step": 160231, "epoch": 1800, "lr": 3.612199834051639e-05}
{"train_loss": 0.11134205013513565, "global_step": 160232, "epoch": 1800, "lr": 3.612144138368554e-05}
{"train_loss": 0.04561831057071686, "global_step": 160233, "epoch": 1800, "lr": 3.612088442872049e-05}
{"train_loss": 0.06741316616535187, "global_step": 160234, "epoch": 1800, "lr": 3.612032747562132e-05}
{"train_loss": 0.11476568877696991, "global_step": 160235, "epoch": 1800, "lr": 3.611977052438811e-05}
{"train_loss": 0.07412151992321014, "global_step": 160236, "epoch": 1800, "lr": 3.611921357502091e-05}
{"train_loss": 0.10176802426576614, "global_step": 160237, "epoch": 1800, "lr": 3.611865662751983e-05}
{"train_loss": 0.12133894860744476, "global_step": 160238, "epoch": 1800, "lr": 3.611809968188491e-05}
{"train_loss": 0.027642402797937393, "global_step": 160239, "epoch": 1800, "lr": 3.611754273811626e-05}
{"train_loss": 0.07351847738027573, "global_step": 160240, "epoch": 1800, "lr": 3.611698579621392e-05}
{"train_loss": 0.10036543011665344, "global_step": 160241, "epoch": 1800, "lr": 3.611642885617801e-05}
{"train_loss": 0.0808105617761612, "global_step": 160242, "epoch": 1800, "lr": 3.611587191800854e-05}
{"train_loss": 0.09039726853370667, "global_step": 160243, "epoch": 1800, "lr": 3.611531498170565e-05}
{"train_loss": 0.11889228224754333, "global_step": 160244, "epoch": 1800, "lr": 3.611475804726937e-05}
{"train_loss": 0.094598188996315, "global_step": 160245, "epoch": 1800, "lr": 3.611420111469981e-05}
{"train_loss": 0.062188878655433655, "global_step": 160246, "epoch": 1800, "lr": 3.6113644183997e-05}
{"train_loss": 0.06706692278385162, "global_step": 160247, "epoch": 1800, "lr": 3.611308725516107e-05}
{"train_loss": 0.05664798617362976, "global_step": 160248, "epoch": 1800, "lr": 3.611253032819205e-05}
{"train_loss": 0.08982551842927933, "global_step": 160249, "epoch": 1800, "lr": 3.611197340309004e-05}
{"train_loss": 0.08892756700515747, "global_step": 160250, "epoch": 1800, "lr": 3.61114164798551e-05}
{"train_loss": 0.09611427038908005, "global_step": 160251, "epoch": 1800, "lr": 3.61108595584873e-05}
{"train_loss": 0.08679406344890594, "global_step": 160252, "epoch": 1800, "lr": 3.6110302638986745e-05}
{"train_loss": 0.09984888881444931, "global_step": 160253, "epoch": 1800, "lr": 3.610974572135347e-05}
{"train_loss": 0.05569298565387726, "global_step": 160254, "epoch": 1800, "lr": 3.610918880558759e-05}
{"train_loss": 0.06559672206640244, "global_step": 160255, "epoch": 1800, "lr": 3.610863189168914e-05}
{"train_loss": 0.08023841679096222, "global_step": 160256, "epoch": 1800, "lr": 3.610807497965824e-05}
{"train_loss": 0.15890148282051086, "global_step": 160257, "epoch": 1800, "lr": 3.6107518069494915e-05}
{"train_loss": 0.1330491602420807, "global_step": 160258, "epoch": 1800, "lr": 3.6106961161199286e-05}
{"train_loss": 0.10970906913280487, "global_step": 160259, "epoch": 1800, "lr": 3.6106404254771384e-05}
{"train_loss": 0.08161541074514389, "global_step": 160260, "epoch": 1800, "lr": 3.610584735021133e-05}
{"train_loss": 0.10483431816101074, "global_step": 160261, "epoch": 1800, "lr": 3.610529044751916e-05}
{"train_loss": 0.07820494472980499, "global_step": 160262, "epoch": 1800, "lr": 3.6104733546694977e-05}
{"train_loss": 0.0925663635134697, "global_step": 160263, "epoch": 1800, "lr": 3.610417664773883e-05}
{"train_loss": 0.10965234041213989, "global_step": 160264, "epoch": 1800, "lr": 3.6103619750650826e-05}
{"train_loss": 0.11704622954130173, "global_step": 160265, "epoch": 1800, "lr": 3.6103062855430994e-05}
{"train_loss": 0.07615718990564346, "global_step": 160266, "epoch": 1800, "lr": 3.6102505962079465e-05}
{"train_loss": 0.1036047488451004, "global_step": 160267, "epoch": 1800, "lr": 3.610194907059627e-05}
{"train_loss": 0.06064971163868904, "global_step": 160268, "epoch": 1800, "lr": 3.610139218098149e-05}
{"train_loss": 0.10664686560630798, "global_step": 160269, "epoch": 1800, "lr": 3.6100835293235224e-05}
{"train_loss": 0.08078745007514954, "global_step": 160270, "epoch": 1800, "lr": 3.6100278407357515e-05}
{"train_loss": 0.06756086647510529, "global_step": 160271, "epoch": 1800, "lr": 3.6099721523348475e-05}
{"train_loss": 0.051976609975099564, "global_step": 160272, "epoch": 1800, "lr": 3.609916464120813e-05}
{"train_loss": 0.1330176442861557, "global_step": 160273, "epoch": 1800, "lr": 3.6098607760936615e-05}
{"train_loss": 0.07075582444667816, "global_step": 160274, "epoch": 1800, "lr": 3.6098050882533944e-05}
{"train_loss": 0.07762718200683594, "global_step": 160275, "epoch": 1800, "lr": 3.609749400600024e-05}
{"train_loss": 0.08003084361553192, "global_step": 160276, "epoch": 1800, "lr": 3.609693713133555e-05}
{"train_loss": 0.051551397889852524, "global_step": 160277, "epoch": 1800, "lr": 3.609638025853996e-05}
{"train_loss": 0.07697470486164093, "global_step": 160278, "epoch": 1800, "lr": 3.6095823387613536e-05}
{"train_loss": 0.04723740369081497, "global_step": 160279, "epoch": 1800, "lr": 3.609526651855638e-05}
{"train_loss": 0.06700228154659271, "global_step": 160280, "epoch": 1800, "lr": 3.6094709651368516e-05}
{"train_loss": 0.09519364684820175, "global_step": 160281, "epoch": 1800, "lr": 3.609415278605007e-05}
{"train_loss": 0.14121311902999878, "global_step": 160282, "epoch": 1800, "lr": 3.609359592260109e-05}
{"train_loss": 0.09608301520347595, "global_step": 160283, "epoch": 1800, "lr": 3.609303906102166e-05}
{"train_loss": 0.12090648710727692, "global_step": 160284, "epoch": 1800, "lr": 3.609248220131184e-05}
{"train_loss": 0.07958915829658508, "global_step": 160285, "epoch": 1800, "lr": 3.609192534347173e-05}
{"train_loss": 0.0475832037627697, "global_step": 160286, "epoch": 1800, "lr": 3.609136848750139e-05}
{"train_loss": 0.06607559323310852, "global_step": 160287, "epoch": 1800, "lr": 3.609081163340088e-05}
{"train_loss": 0.09062024180808764, "global_step": 160288, "epoch": 1800, "lr": 3.6090254781170304e-05, "train/sim_max_reward_0": 0.9373543968917487, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.6612782597539636, "train/sim_max_reward_3": 1.0, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9157012694228144, "test/sim_max_reward_4500000": 0.9870099746478899, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.18843814418085308, "test/sim_max_reward_4500004": 1.0, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9719356541776368, "test/sim_max_reward_4500008": 0.18812966486392488, "test/sim_max_reward_4500009": 0.05111122995105206, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5356824504929221, "test/sim_max_reward_4500014": 0.4169603965718288, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 1.0, "test/sim_max_reward_4500017": 0.5665687237355056, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9751416609739273, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.869616922165706, "test/sim_max_reward_4500022": 0.9748466741667339, "test/sim_max_reward_4500023": 0.7068115074513067, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.1989861605525087, "test/sim_max_reward_4500027": 0.9670214532023338, "test/sim_max_reward_4500028": 0.48119429052383683, "test/sim_max_reward_4500029": 0.5431570939800685, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.9056323538423907, "test/sim_max_reward_4500033": 0.9147904285122959, "test/sim_max_reward_4500034": 0.45986213199763715, "test/sim_max_reward_4500035": 0.5736106083874611, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.7866113836373545, "test/sim_max_reward_4500038": 0.16275924742216422, "test/sim_max_reward_4500039": 0.9743617858598161, "test/sim_max_reward_4500040": 0.8966723050372655, "test/sim_max_reward_4500041": 0.44302314711918317, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9759927870855968, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.44351369063684404, "test/sim_max_reward_4500046": 0.44335148933290225, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.9190556543447544, "test/mean_score": 0.592055867210179, "val_loss": 6.736750602722168, "train_action_mse_error": 16.494308471679688}
{"train_loss": 0.10639539361000061, "global_step": 160289, "epoch": 1801, "lr": 3.608969793080971e-05}
{"train_loss": 0.09437980502843857, "global_step": 160290, "epoch": 1801, "lr": 3.6089141082319214e-05}
{"train_loss": 0.15815570950508118, "global_step": 160291, "epoch": 1801, "lr": 3.608858423569883e-05}
{"train_loss": 0.09836137294769287, "global_step": 160292, "epoch": 1801, "lr": 3.60880273909487e-05}
{"train_loss": 0.0524071604013443, "global_step": 160293, "epoch": 1801, "lr": 3.6087470548068834e-05}
{"train_loss": 0.07377277314662933, "global_step": 160294, "epoch": 1801, "lr": 3.608691370705937e-05}
{"train_loss": 0.13682031631469727, "global_step": 160295, "epoch": 1801, "lr": 3.608635686792032e-05}
{"train_loss": 0.09037887305021286, "global_step": 160296, "epoch": 1801, "lr": 3.608580003065181e-05}
{"train_loss": 0.11808095127344131, "global_step": 160297, "epoch": 1801, "lr": 3.6085243195253884e-05}
{"train_loss": 0.07601168751716614, "global_step": 160298, "epoch": 1801, "lr": 3.608468636172664e-05}
{"train_loss": 0.04214835911989212, "global_step": 160299, "epoch": 1801, "lr": 3.608412953007012e-05}
{"train_loss": 0.05674837902188301, "global_step": 160300, "epoch": 1801, "lr": 3.608357270028444e-05}
{"train_loss": 0.09688561409711838, "global_step": 160301, "epoch": 1801, "lr": 3.608301587236963e-05}
{"train_loss": 0.06204354763031006, "global_step": 160302, "epoch": 1801, "lr": 3.608245904632582e-05}
{"train_loss": 0.07436750829219818, "global_step": 160303, "epoch": 1801, "lr": 3.608190222215302e-05}
{"train_loss": 0.08729352056980133, "global_step": 160304, "epoch": 1801, "lr": 3.608134539985136e-05}
{"train_loss": 0.07346197962760925, "global_step": 160305, "epoch": 1801, "lr": 3.6080788579420886e-05}
{"train_loss": 0.09280446171760559, "global_step": 160306, "epoch": 1801, "lr": 3.608023176086168e-05}
{"train_loss": 0.0638383999466896, "global_step": 160307, "epoch": 1801, "lr": 3.607967494417384e-05}
{"train_loss": 0.12296018749475479, "global_step": 160308, "epoch": 1801, "lr": 3.607911812935739e-05}
{"train_loss": 0.07316706329584122, "global_step": 160309, "epoch": 1801, "lr": 3.6078561316412455e-05}
{"train_loss": 0.045901164412498474, "global_step": 160310, "epoch": 1801, "lr": 3.607800450533906e-05}
{"train_loss": 0.11139070987701416, "global_step": 160311, "epoch": 1801, "lr": 3.607744769613733e-05}
{"train_loss": 0.05506063252687454, "global_step": 160312, "epoch": 1801, "lr": 3.6076890888807304e-05}
{"train_loss": 0.07696972787380219, "global_step": 160313, "epoch": 1801, "lr": 3.6076334083349084e-05}
{"train_loss": 0.07849843055009842, "global_step": 160314, "epoch": 1801, "lr": 3.607577727976271e-05}
{"train_loss": 0.13624092936515808, "global_step": 160315, "epoch": 1801, "lr": 3.60752204780483e-05}
{"train_loss": 0.05934770777821541, "global_step": 160316, "epoch": 1801, "lr": 3.607466367820589e-05}
{"train_loss": 0.06885834783315659, "global_step": 160317, "epoch": 1801, "lr": 3.607410688023559e-05}
{"train_loss": 0.059873148798942566, "global_step": 160318, "epoch": 1801, "lr": 3.607355008413744e-05}
{"train_loss": 0.04687143489718437, "global_step": 160319, "epoch": 1801, "lr": 3.607299328991155e-05}
{"train_loss": 0.12111670523881912, "global_step": 160320, "epoch": 1801, "lr": 3.607243649755795e-05}
{"train_loss": 0.13986711204051971, "global_step": 160321, "epoch": 1801, "lr": 3.6071879707076764e-05}
{"train_loss": 0.11042520403862, "global_step": 160322, "epoch": 1801, "lr": 3.607132291846803e-05}
{"train_loss": 0.12505565583705902, "global_step": 160323, "epoch": 1801, "lr": 3.607076613173184e-05}
{"train_loss": 0.09537273645401001, "global_step": 160324, "epoch": 1801, "lr": 3.607020934686827e-05}
{"train_loss": 0.12543606758117676, "global_step": 160325, "epoch": 1801, "lr": 3.606965256387738e-05}
{"train_loss": 0.07085704803466797, "global_step": 160326, "epoch": 1801, "lr": 3.606909578275928e-05}
{"train_loss": 0.07694867253303528, "global_step": 160327, "epoch": 1801, "lr": 3.606853900351399e-05}
{"train_loss": 0.08735410124063492, "global_step": 160328, "epoch": 1801, "lr": 3.606798222614164e-05}
{"train_loss": 0.09265381842851639, "global_step": 160329, "epoch": 1801, "lr": 3.606742545064226e-05}
{"train_loss": 0.08125327527523041, "global_step": 160330, "epoch": 1801, "lr": 3.606686867701596e-05}
{"train_loss": 0.1499319076538086, "global_step": 160331, "epoch": 1801, "lr": 3.6066311905262784e-05}
{"train_loss": 0.06213981658220291, "global_step": 160332, "epoch": 1801, "lr": 3.606575513538285e-05}
{"train_loss": 0.056048158556222916, "global_step": 160333, "epoch": 1801, "lr": 3.6065198367376176e-05}
{"train_loss": 0.1361817866563797, "global_step": 160334, "epoch": 1801, "lr": 3.606464160124289e-05}
{"train_loss": 0.05906063690781593, "global_step": 160335, "epoch": 1801, "lr": 3.606408483698301e-05}
{"train_loss": 0.09673252701759338, "global_step": 160336, "epoch": 1801, "lr": 3.606352807459669e-05}
{"train_loss": 0.04867730662226677, "global_step": 160337, "epoch": 1801, "lr": 3.6062971314083914e-05}
{"train_loss": 0.20376788079738617, "global_step": 160338, "epoch": 1801, "lr": 3.606241455544484e-05}
{"train_loss": 0.12009898573160172, "global_step": 160339, "epoch": 1801, "lr": 3.6061857798679465e-05}
{"train_loss": 0.15368737280368805, "global_step": 160340, "epoch": 1801, "lr": 3.606130104378794e-05}
{"train_loss": 0.10208356380462646, "global_step": 160341, "epoch": 1801, "lr": 3.6060744290770275e-05}
{"train_loss": 0.031221497803926468, "global_step": 160342, "epoch": 1801, "lr": 3.6060187539626586e-05}
{"train_loss": 0.0857737734913826, "global_step": 160343, "epoch": 1801, "lr": 3.6059630790356934e-05}
{"train_loss": 0.10778038203716278, "global_step": 160344, "epoch": 1801, "lr": 3.605907404296139e-05}
{"train_loss": 0.049643225967884064, "global_step": 160345, "epoch": 1801, "lr": 3.605851729744005e-05}
{"train_loss": 0.10656128823757172, "global_step": 160346, "epoch": 1801, "lr": 3.6057960553792956e-05}
{"train_loss": 0.047775957733392715, "global_step": 160347, "epoch": 1801, "lr": 3.6057403812020217e-05}
{"train_loss": 0.13122907280921936, "global_step": 160348, "epoch": 1801, "lr": 3.6056847072121866e-05}
{"train_loss": 0.13767927885055542, "global_step": 160349, "epoch": 1801, "lr": 3.605629033409802e-05}
{"train_loss": 0.10119818896055222, "global_step": 160350, "epoch": 1801, "lr": 3.605573359794872e-05}
{"train_loss": 0.051234498620033264, "global_step": 160351, "epoch": 1801, "lr": 3.6055176863674076e-05}
{"train_loss": 0.09320365637540817, "global_step": 160352, "epoch": 1801, "lr": 3.605462013127412e-05}
{"train_loss": 0.1507904976606369, "global_step": 160353, "epoch": 1801, "lr": 3.605406340074898e-05}
{"train_loss": 0.05720391124486923, "global_step": 160354, "epoch": 1801, "lr": 3.6053506672098674e-05}
{"train_loss": 0.05607256665825844, "global_step": 160355, "epoch": 1801, "lr": 3.6052949945323324e-05}
{"train_loss": 0.06906219571828842, "global_step": 160356, "epoch": 1801, "lr": 3.6052393220422965e-05}
{"train_loss": 0.07433780282735825, "global_step": 160357, "epoch": 1801, "lr": 3.6051836497397714e-05}
{"train_loss": 0.10474025458097458, "global_step": 160358, "epoch": 1801, "lr": 3.6051279776247604e-05}
{"train_loss": 0.11167732626199722, "global_step": 160359, "epoch": 1801, "lr": 3.605072305697274e-05}
{"train_loss": 0.12645672261714935, "global_step": 160360, "epoch": 1801, "lr": 3.6050166339573174e-05}
{"train_loss": 0.08749192208051682, "global_step": 160361, "epoch": 1801, "lr": 3.604960962404901e-05}
{"train_loss": 0.05100138112902641, "global_step": 160362, "epoch": 1801, "lr": 3.604905291040029e-05}
{"train_loss": 0.0918019488453865, "global_step": 160363, "epoch": 1801, "lr": 3.604849619862713e-05}
{"train_loss": 0.11599420011043549, "global_step": 160364, "epoch": 1801, "lr": 3.6047939488729565e-05}
{"train_loss": 0.07823579758405685, "global_step": 160365, "epoch": 1801, "lr": 3.604738278070767e-05}
{"train_loss": 0.16350024938583374, "global_step": 160366, "epoch": 1801, "lr": 3.604682607456156e-05}
{"train_loss": 0.1257745772600174, "global_step": 160367, "epoch": 1801, "lr": 3.6046269370291264e-05}
{"train_loss": 0.08081464469432831, "global_step": 160368, "epoch": 1801, "lr": 3.604571266789689e-05}
{"train_loss": 0.16305911540985107, "global_step": 160369, "epoch": 1801, "lr": 3.604515596737848e-05}
{"train_loss": 0.17584022879600525, "global_step": 160370, "epoch": 1801, "lr": 3.604459926873616e-05}
{"train_loss": 0.16429242491722107, "global_step": 160371, "epoch": 1801, "lr": 3.604404257196995e-05}
{"train_loss": 0.1118026003241539, "global_step": 160372, "epoch": 1801, "lr": 3.6043485877079964e-05}
{"train_loss": 0.07414701581001282, "global_step": 160373, "epoch": 1801, "lr": 3.604292918406624e-05}
{"train_loss": 0.1040748581290245, "global_step": 160374, "epoch": 1801, "lr": 3.60423724929289e-05}
{"train_loss": 0.1185338944196701, "global_step": 160375, "epoch": 1801, "lr": 3.604181580366797e-05}
{"train_loss": 0.08161720633506775, "global_step": 160376, "epoch": 1801, "lr": 3.604125911628357e-05}
{"train_loss": 0.09508272277170353, "global_step": 160377, "epoch": 1801, "lr": 3.6040702430775716e-05, "val_loss": 6.65085506439209}
{"train_loss": 0.08308196067810059, "global_step": 160378, "epoch": 1802, "lr": 3.604014574714455e-05}
{"train_loss": 0.14349675178527832, "global_step": 160379, "epoch": 1802, "lr": 3.6039589065390106e-05}
{"train_loss": 0.10019206255674362, "global_step": 160380, "epoch": 1802, "lr": 3.603903238551247e-05}
{"train_loss": 0.09387041628360748, "global_step": 160381, "epoch": 1802, "lr": 3.603847570751172e-05}
{"train_loss": 0.08758629113435745, "global_step": 160382, "epoch": 1802, "lr": 3.6037919031387925e-05}
{"train_loss": 0.08948793262243271, "global_step": 160383, "epoch": 1802, "lr": 3.603736235714117e-05}
{"train_loss": 0.06991998851299286, "global_step": 160384, "epoch": 1802, "lr": 3.60368056847715e-05}
{"train_loss": 0.11509586870670319, "global_step": 160385, "epoch": 1802, "lr": 3.603624901427904e-05}
{"train_loss": 0.11574097722768784, "global_step": 160386, "epoch": 1802, "lr": 3.603569234566381e-05}
{"train_loss": 0.0790359228849411, "global_step": 160387, "epoch": 1802, "lr": 3.6035135678925936e-05}
{"train_loss": 0.08930175006389618, "global_step": 160388, "epoch": 1802, "lr": 3.603457901406544e-05}
{"train_loss": 0.06792567670345306, "global_step": 160389, "epoch": 1802, "lr": 3.6034022351082445e-05}
{"train_loss": 0.1259394735097885, "global_step": 160390, "epoch": 1802, "lr": 3.603346568997698e-05}
{"train_loss": 0.09037069231271744, "global_step": 160391, "epoch": 1802, "lr": 3.603290903074918e-05}
{"train_loss": 0.060970015823841095, "global_step": 160392, "epoch": 1802, "lr": 3.603235237339906e-05}
{"train_loss": 0.08056411147117615, "global_step": 160393, "epoch": 1802, "lr": 3.6031795717926745e-05}
{"train_loss": 0.12160531431436539, "global_step": 160394, "epoch": 1802, "lr": 3.6031239064332256e-05}
{"train_loss": 0.10310593247413635, "global_step": 160395, "epoch": 1802, "lr": 3.603068241261569e-05}
{"train_loss": 0.08888988941907883, "global_step": 160396, "epoch": 1802, "lr": 3.6030125762777165e-05}
{"train_loss": 0.11624249815940857, "global_step": 160397, "epoch": 1802, "lr": 3.60295691148167e-05}
{"train_loss": 0.07051756978034973, "global_step": 160398, "epoch": 1802, "lr": 3.602901246873439e-05}
{"train_loss": 0.13520190119743347, "global_step": 160399, "epoch": 1802, "lr": 3.602845582453031e-05}
{"train_loss": 0.13262945413589478, "global_step": 160400, "epoch": 1802, "lr": 3.602789918220454e-05}
{"train_loss": 0.12003915756940842, "global_step": 160401, "epoch": 1802, "lr": 3.602734254175714e-05}
{"train_loss": 0.05692348629236221, "global_step": 160402, "epoch": 1802, "lr": 3.602678590318821e-05}
{"train_loss": 0.08837667852640152, "global_step": 160403, "epoch": 1802, "lr": 3.602622926649778e-05}
{"train_loss": 0.09240131080150604, "global_step": 160404, "epoch": 1802, "lr": 3.602567263168598e-05}
{"train_loss": 0.09552943706512451, "global_step": 160405, "epoch": 1802, "lr": 3.6025115998752834e-05}
{"train_loss": 0.12449580430984497, "global_step": 160406, "epoch": 1802, "lr": 3.602455936769846e-05}
{"train_loss": 0.10878465324640274, "global_step": 160407, "epoch": 1802, "lr": 3.60240027385229e-05}
{"train_loss": 0.04148418828845024, "global_step": 160408, "epoch": 1802, "lr": 3.602344611122626e-05}
{"train_loss": 0.06590814888477325, "global_step": 160409, "epoch": 1802, "lr": 3.602288948580857e-05}
{"train_loss": 0.06362177431583405, "global_step": 160410, "epoch": 1802, "lr": 3.6022332862269955e-05}
{"train_loss": 0.15299397706985474, "global_step": 160411, "epoch": 1802, "lr": 3.602177624061045e-05}
{"train_loss": 0.04877561703324318, "global_step": 160412, "epoch": 1802, "lr": 3.602121962083017e-05}
{"train_loss": 0.08985832333564758, "global_step": 160413, "epoch": 1802, "lr": 3.6020663002929146e-05}
{"train_loss": 0.09902859479188919, "global_step": 160414, "epoch": 1802, "lr": 3.602010638690747e-05}
{"train_loss": 0.06067603453993797, "global_step": 160415, "epoch": 1802, "lr": 3.601954977276524e-05}
{"train_loss": 0.11775296181440353, "global_step": 160416, "epoch": 1802, "lr": 3.601899316050249e-05}
{"train_loss": 0.07116745412349701, "global_step": 160417, "epoch": 1802, "lr": 3.6018436550119336e-05}
{"train_loss": 0.09551133960485458, "global_step": 160418, "epoch": 1802, "lr": 3.6017879941615826e-05}
{"train_loss": 0.07528531551361084, "global_step": 160419, "epoch": 1802, "lr": 3.6017323334992045e-05}
{"train_loss": 0.11595209687948227, "global_step": 160420, "epoch": 1802, "lr": 3.601676673024806e-05}
{"train_loss": 0.08438172191381454, "global_step": 160421, "epoch": 1802, "lr": 3.601621012738395e-05}
{"train_loss": 0.06912244111299515, "global_step": 160422, "epoch": 1802, "lr": 3.601565352639979e-05}
{"train_loss": 0.10940206795930862, "global_step": 160423, "epoch": 1802, "lr": 3.6015096927295664e-05}
{"train_loss": 0.11191447824239731, "global_step": 160424, "epoch": 1802, "lr": 3.6014540330071624e-05}
{"train_loss": 0.07736030220985413, "global_step": 160425, "epoch": 1802, "lr": 3.6013983734727774e-05}
{"train_loss": 0.07992889732122421, "global_step": 160426, "epoch": 1802, "lr": 3.601342714126415e-05}
{"train_loss": 0.12051092833280563, "global_step": 160427, "epoch": 1802, "lr": 3.601287054968088e-05}
{"train_loss": 0.06710793823003769, "global_step": 160428, "epoch": 1802, "lr": 3.601231395997797e-05}
{"train_loss": 0.0822707936167717, "global_step": 160429, "epoch": 1802, "lr": 3.6011757372155576e-05}
{"train_loss": 0.07054171711206436, "global_step": 160430, "epoch": 1802, "lr": 3.6011200786213696e-05}
{"train_loss": 0.03487033024430275, "global_step": 160431, "epoch": 1802, "lr": 3.601064420215245e-05}
{"train_loss": 0.0783163458108902, "global_step": 160432, "epoch": 1802, "lr": 3.601008761997191e-05}
{"train_loss": 0.08378269523382187, "global_step": 160433, "epoch": 1802, "lr": 3.6009531039672136e-05}
{"train_loss": 0.1336316019296646, "global_step": 160434, "epoch": 1802, "lr": 3.600897446125322e-05}
{"train_loss": 0.09410843998193741, "global_step": 160435, "epoch": 1802, "lr": 3.600841788471522e-05}
{"train_loss": 0.1170925498008728, "global_step": 160436, "epoch": 1802, "lr": 3.6007861310058224e-05}
{"train_loss": 0.11724796891212463, "global_step": 160437, "epoch": 1802, "lr": 3.600730473728229e-05}
{"train_loss": 0.18780145049095154, "global_step": 160438, "epoch": 1802, "lr": 3.600674816638751e-05}
{"train_loss": 0.09389760345220566, "global_step": 160439, "epoch": 1802, "lr": 3.600619159737395e-05}
{"train_loss": 0.07958358526229858, "global_step": 160440, "epoch": 1802, "lr": 3.6005635030241694e-05}
{"train_loss": 0.09136106818914413, "global_step": 160441, "epoch": 1802, "lr": 3.60050784649908e-05}
{"train_loss": 0.08923576027154922, "global_step": 160442, "epoch": 1802, "lr": 3.600452190162137e-05}
{"train_loss": 0.04702671617269516, "global_step": 160443, "epoch": 1802, "lr": 3.600396534013344e-05}
{"train_loss": 0.08642019331455231, "global_step": 160444, "epoch": 1802, "lr": 3.600340878052712e-05}
{"train_loss": 0.11258441209793091, "global_step": 160445, "epoch": 1802, "lr": 3.600285222280245e-05}
{"train_loss": 0.05617177486419678, "global_step": 160446, "epoch": 1802, "lr": 3.6002295666959556e-05}
{"train_loss": 0.040732257068157196, "global_step": 160447, "epoch": 1802, "lr": 3.6001739112998455e-05}
{"train_loss": 0.06738711148500443, "global_step": 160448, "epoch": 1802, "lr": 3.600118256091928e-05}
{"train_loss": 0.05760498717427254, "global_step": 160449, "epoch": 1802, "lr": 3.600062601072204e-05}
{"train_loss": 0.11948001384735107, "global_step": 160450, "epoch": 1802, "lr": 3.6000069462406856e-05}
{"train_loss": 0.06675873696804047, "global_step": 160451, "epoch": 1802, "lr": 3.599951291597381e-05}
{"train_loss": 0.08458689600229263, "global_step": 160452, "epoch": 1802, "lr": 3.5998956371422936e-05}
{"train_loss": 0.14919538795948029, "global_step": 160453, "epoch": 1802, "lr": 3.599839982875435e-05}
{"train_loss": 0.05702593922615051, "global_step": 160454, "epoch": 1802, "lr": 3.5997843287968095e-05}
{"train_loss": 0.07696717977523804, "global_step": 160455, "epoch": 1802, "lr": 3.599728674906427e-05}
{"train_loss": 0.12676088511943817, "global_step": 160456, "epoch": 1802, "lr": 3.599673021204294e-05}
{"train_loss": 0.06353798508644104, "global_step": 160457, "epoch": 1802, "lr": 3.5996173676904176e-05}
{"train_loss": 0.13756899535655975, "global_step": 160458, "epoch": 1802, "lr": 3.599561714364805e-05}
{"train_loss": 0.04376130923628807, "global_step": 160459, "epoch": 1802, "lr": 3.5995060612274657e-05}
{"train_loss": 0.08320063352584839, "global_step": 160460, "epoch": 1802, "lr": 3.599450408278404e-05}
{"train_loss": 0.11294271796941757, "global_step": 160461, "epoch": 1802, "lr": 3.599394755517631e-05}
{"train_loss": 0.1120527908205986, "global_step": 160462, "epoch": 1802, "lr": 3.599339102945151e-05}
{"train_loss": 0.1197570189833641, "global_step": 160463, "epoch": 1802, "lr": 3.5992834505609726e-05}
{"train_loss": 0.07278287410736084, "global_step": 160464, "epoch": 1802, "lr": 3.5992277983651035e-05}
{"train_loss": 0.1119566261768341, "global_step": 160465, "epoch": 1802, "lr": 3.599172146357552e-05}
{"train_loss": 0.09290362082505493, "global_step": 160466, "epoch": 1802, "lr": 3.599116494538324e-05, "val_loss": 6.697198390960693}
{"train_loss": 0.11514467000961304, "global_step": 160467, "epoch": 1803, "lr": 3.5990608429074286e-05}
{"train_loss": 0.12415024638175964, "global_step": 160468, "epoch": 1803, "lr": 3.5990051914648706e-05}
{"train_loss": 0.12683521211147308, "global_step": 160469, "epoch": 1803, "lr": 3.598949540210659e-05}
{"train_loss": 0.0669136643409729, "global_step": 160470, "epoch": 1803, "lr": 3.598893889144804e-05}
{"train_loss": 0.09279493242502213, "global_step": 160471, "epoch": 1803, "lr": 3.598838238267309e-05}
{"train_loss": 0.076717808842659, "global_step": 160472, "epoch": 1803, "lr": 3.5987825875781845e-05}
{"train_loss": 0.08914192020893097, "global_step": 160473, "epoch": 1803, "lr": 3.598726937077434e-05}
{"train_loss": 0.07398667931556702, "global_step": 160474, "epoch": 1803, "lr": 3.598671286765071e-05}
{"train_loss": 0.10389513522386551, "global_step": 160475, "epoch": 1803, "lr": 3.598615636641098e-05}
{"train_loss": 0.09434808045625687, "global_step": 160476, "epoch": 1803, "lr": 3.598559986705524e-05}
{"train_loss": 0.084722138941288, "global_step": 160477, "epoch": 1803, "lr": 3.598504336958357e-05}
{"train_loss": 0.09377240389585495, "global_step": 160478, "epoch": 1803, "lr": 3.598448687399603e-05}
{"train_loss": 0.07742441445589066, "global_step": 160479, "epoch": 1803, "lr": 3.598393038029271e-05}
{"train_loss": 0.16060426831245422, "global_step": 160480, "epoch": 1803, "lr": 3.598337388847369e-05}
{"train_loss": 0.036917101591825485, "global_step": 160481, "epoch": 1803, "lr": 3.5982817398539014e-05}
{"train_loss": 0.028426362201571465, "global_step": 160482, "epoch": 1803, "lr": 3.598226091048879e-05}
{"train_loss": 0.13653169572353363, "global_step": 160483, "epoch": 1803, "lr": 3.598170442432308e-05}
{"train_loss": 0.06533076614141464, "global_step": 160484, "epoch": 1803, "lr": 3.598114794004196e-05}
{"train_loss": 0.1687234789133072, "global_step": 160485, "epoch": 1803, "lr": 3.59805914576455e-05}
{"train_loss": 0.06186985969543457, "global_step": 160486, "epoch": 1803, "lr": 3.5980034977133756e-05}
{"train_loss": 0.10815659165382385, "global_step": 160487, "epoch": 1803, "lr": 3.5979478498506865e-05}
{"train_loss": 0.11429613828659058, "global_step": 160488, "epoch": 1803, "lr": 3.597892202176484e-05}
{"train_loss": 0.11883702129125595, "global_step": 160489, "epoch": 1803, "lr": 3.597836554690779e-05}
{"train_loss": 0.13153928518295288, "global_step": 160490, "epoch": 1803, "lr": 3.597780907393576e-05}
{"train_loss": 0.11553626507520676, "global_step": 160491, "epoch": 1803, "lr": 3.597725260284887e-05}
{"train_loss": 0.07110194861888885, "global_step": 160492, "epoch": 1803, "lr": 3.5976696133647135e-05}
{"train_loss": 0.08465572446584702, "global_step": 160493, "epoch": 1803, "lr": 3.597613966633069e-05}
{"train_loss": 0.12436926364898682, "global_step": 160494, "epoch": 1803, "lr": 3.5975583200899574e-05}
{"train_loss": 0.09131357073783875, "global_step": 160495, "epoch": 1803, "lr": 3.597502673735387e-05}
{"train_loss": 0.05829915404319763, "global_step": 160496, "epoch": 1803, "lr": 3.597447027569364e-05}
{"train_loss": 0.07985042780637741, "global_step": 160497, "epoch": 1803, "lr": 3.597391381591899e-05}
{"train_loss": 0.13302582502365112, "global_step": 160498, "epoch": 1803, "lr": 3.597335735802997e-05}
{"train_loss": 0.07801828533411026, "global_step": 160499, "epoch": 1803, "lr": 3.5972800902026665e-05}
{"train_loss": 0.07261264324188232, "global_step": 160500, "epoch": 1803, "lr": 3.597224444790913e-05}
{"train_loss": 0.07837711274623871, "global_step": 160501, "epoch": 1803, "lr": 3.597168799567747e-05}
{"train_loss": 0.0757928118109703, "global_step": 160502, "epoch": 1803, "lr": 3.5971131545331734e-05}
{"train_loss": 0.0956774577498436, "global_step": 160503, "epoch": 1803, "lr": 3.597057509687203e-05}
{"train_loss": 0.12230794876813889, "global_step": 160504, "epoch": 1803, "lr": 3.597001865029839e-05}
{"train_loss": 0.10171964019536972, "global_step": 160505, "epoch": 1803, "lr": 3.5969462205610896e-05}
{"train_loss": 0.09630021452903748, "global_step": 160506, "epoch": 1803, "lr": 3.596890576280967e-05}
{"train_loss": 0.09402135014533997, "global_step": 160507, "epoch": 1803, "lr": 3.596834932189473e-05}
{"train_loss": 0.05859219655394554, "global_step": 160508, "epoch": 1803, "lr": 3.5967792882866205e-05}
{"train_loss": 0.09789750725030899, "global_step": 160509, "epoch": 1803, "lr": 3.596723644572411e-05}
{"train_loss": 0.16528058052062988, "global_step": 160510, "epoch": 1803, "lr": 3.596668001046856e-05}
{"train_loss": 0.14662432670593262, "global_step": 160511, "epoch": 1803, "lr": 3.596612357709961e-05}
{"train_loss": 0.08765517175197601, "global_step": 160512, "epoch": 1803, "lr": 3.596556714561736e-05}
{"train_loss": 0.07988890260457993, "global_step": 160513, "epoch": 1803, "lr": 3.5965010716021866e-05}
{"train_loss": 0.09333822131156921, "global_step": 160514, "epoch": 1803, "lr": 3.5964454288313195e-05}
{"train_loss": 0.1407260298728943, "global_step": 160515, "epoch": 1803, "lr": 3.5963897862491436e-05}
{"train_loss": 0.10061277449131012, "global_step": 160516, "epoch": 1803, "lr": 3.5963341438556655e-05}
{"train_loss": 0.10365112870931625, "global_step": 160517, "epoch": 1803, "lr": 3.5962785016508936e-05}
{"train_loss": 0.07848747819662094, "global_step": 160518, "epoch": 1803, "lr": 3.596222859634836e-05}
{"train_loss": 0.07082074880599976, "global_step": 160519, "epoch": 1803, "lr": 3.596167217807498e-05}
{"train_loss": 0.1262967586517334, "global_step": 160520, "epoch": 1803, "lr": 3.5961115761688894e-05}
{"train_loss": 0.10880795866250992, "global_step": 160521, "epoch": 1803, "lr": 3.5960559347190135e-05}
{"train_loss": 0.09208985418081284, "global_step": 160522, "epoch": 1803, "lr": 3.596000293457882e-05}
{"train_loss": 0.11846098303794861, "global_step": 160523, "epoch": 1803, "lr": 3.595944652385503e-05}
{"train_loss": 0.12493962794542313, "global_step": 160524, "epoch": 1803, "lr": 3.595889011501879e-05}
{"train_loss": 0.09907476603984833, "global_step": 160525, "epoch": 1803, "lr": 3.595833370807025e-05}
{"train_loss": 0.09313293546438217, "global_step": 160526, "epoch": 1803, "lr": 3.59577773030094e-05}
{"train_loss": 0.11427042633295059, "global_step": 160527, "epoch": 1803, "lr": 3.595722089983639e-05}
{"train_loss": 0.12099140882492065, "global_step": 160528, "epoch": 1803, "lr": 3.595666449855123e-05}
{"train_loss": 0.1456233114004135, "global_step": 160529, "epoch": 1803, "lr": 3.5956108099154044e-05}
{"train_loss": 0.1588791012763977, "global_step": 160530, "epoch": 1803, "lr": 3.5955551701644876e-05}
{"train_loss": 0.10292389243841171, "global_step": 160531, "epoch": 1803, "lr": 3.5954995306023826e-05}
{"train_loss": 0.08564439415931702, "global_step": 160532, "epoch": 1803, "lr": 3.595443891229094e-05}
{"train_loss": 0.12752962112426758, "global_step": 160533, "epoch": 1803, "lr": 3.595388252044632e-05}
{"train_loss": 0.08330089598894119, "global_step": 160534, "epoch": 1803, "lr": 3.595332613049003e-05}
{"train_loss": 0.04515673965215683, "global_step": 160535, "epoch": 1803, "lr": 3.595276974242214e-05}
{"train_loss": 0.12468710541725159, "global_step": 160536, "epoch": 1803, "lr": 3.595221335624274e-05}
{"train_loss": 0.11025480180978775, "global_step": 160537, "epoch": 1803, "lr": 3.595165697195187e-05}
{"train_loss": 0.15654419362545013, "global_step": 160538, "epoch": 1803, "lr": 3.5951100589549645e-05}
{"train_loss": 0.17839328944683075, "global_step": 160539, "epoch": 1803, "lr": 3.595054420903613e-05}
{"train_loss": 0.10435888916254044, "global_step": 160540, "epoch": 1803, "lr": 3.594998783041138e-05}
{"train_loss": 0.12577350437641144, "global_step": 160541, "epoch": 1803, "lr": 3.594943145367548e-05}
{"train_loss": 0.08734183013439178, "global_step": 160542, "epoch": 1803, "lr": 3.5948875078828514e-05}
{"train_loss": 0.088358074426651, "global_step": 160543, "epoch": 1803, "lr": 3.594831870587054e-05}
{"train_loss": 0.06705960631370544, "global_step": 160544, "epoch": 1803, "lr": 3.594776233480166e-05}
{"train_loss": 0.1323966383934021, "global_step": 160545, "epoch": 1803, "lr": 3.5947205965621924e-05}
{"train_loss": 0.03526407107710838, "global_step": 160546, "epoch": 1803, "lr": 3.594664959833143e-05}
{"train_loss": 0.09887606650590897, "global_step": 160547, "epoch": 1803, "lr": 3.594609323293021e-05}
{"train_loss": 0.05331852287054062, "global_step": 160548, "epoch": 1803, "lr": 3.594553686941839e-05}
{"train_loss": 0.1188281700015068, "global_step": 160549, "epoch": 1803, "lr": 3.5944980507796e-05}
{"train_loss": 0.11833960562944412, "global_step": 160550, "epoch": 1803, "lr": 3.594442414806316e-05}
{"train_loss": 0.08550336211919785, "global_step": 160551, "epoch": 1803, "lr": 3.5943867790219896e-05}
{"train_loss": 0.13147439062595367, "global_step": 160552, "epoch": 1803, "lr": 3.5943311434266333e-05}
{"train_loss": 0.04995311424136162, "global_step": 160553, "epoch": 1803, "lr": 3.594275508020251e-05}
{"train_loss": 0.09829085320234299, "global_step": 160554, "epoch": 1803, "lr": 3.594219872802851e-05}
{"train_loss": 0.09966971627907471, "global_step": 160555, "epoch": 1803, "lr": 3.594164237774441e-05, "val_loss": 6.9422783851623535}
{"train_loss": 0.05257251113653183, "global_step": 160556, "epoch": 1804, "lr": 3.594108602935028e-05}
{"train_loss": 0.1209874227643013, "global_step": 160557, "epoch": 1804, "lr": 3.59405296828462e-05}
{"train_loss": 0.11092213541269302, "global_step": 160558, "epoch": 1804, "lr": 3.593997333823225e-05}
{"train_loss": 0.08797883987426758, "global_step": 160559, "epoch": 1804, "lr": 3.59394169955085e-05}
{"train_loss": 0.08329927176237106, "global_step": 160560, "epoch": 1804, "lr": 3.593886065467502e-05}
{"train_loss": 0.07961123436689377, "global_step": 160561, "epoch": 1804, "lr": 3.5938304315731896e-05}
{"train_loss": 0.05198271945118904, "global_step": 160562, "epoch": 1804, "lr": 3.593774797867918e-05}
{"train_loss": 0.07374666631221771, "global_step": 160563, "epoch": 1804, "lr": 3.5937191643516986e-05}
{"train_loss": 0.13080991804599762, "global_step": 160564, "epoch": 1804, "lr": 3.593663531024534e-05}
{"train_loss": 0.17207840085029602, "global_step": 160565, "epoch": 1804, "lr": 3.5936078978864365e-05}
{"train_loss": 0.13491810858249664, "global_step": 160566, "epoch": 1804, "lr": 3.593552264937409e-05}
{"train_loss": 0.15832588076591492, "global_step": 160567, "epoch": 1804, "lr": 3.593496632177464e-05}
{"train_loss": 0.09747882187366486, "global_step": 160568, "epoch": 1804, "lr": 3.5934409996066033e-05}
{"train_loss": 0.042369309812784195, "global_step": 160569, "epoch": 1804, "lr": 3.59338536722484e-05}
{"train_loss": 0.07220035791397095, "global_step": 160570, "epoch": 1804, "lr": 3.593329735032176e-05}
{"train_loss": 0.07517915964126587, "global_step": 160571, "epoch": 1804, "lr": 3.593274103028624e-05}
{"train_loss": 0.12184647470712662, "global_step": 160572, "epoch": 1804, "lr": 3.593218471214189e-05}
{"train_loss": 0.05669713020324707, "global_step": 160573, "epoch": 1804, "lr": 3.593162839588878e-05}
{"train_loss": 0.09142735600471497, "global_step": 160574, "epoch": 1804, "lr": 3.5931072081527e-05}
{"train_loss": 0.09466644376516342, "global_step": 160575, "epoch": 1804, "lr": 3.59305157690566e-05}
{"train_loss": 0.07414579391479492, "global_step": 160576, "epoch": 1804, "lr": 3.592995945847768e-05}
{"train_loss": 0.06542646884918213, "global_step": 160577, "epoch": 1804, "lr": 3.592940314979029e-05}
{"train_loss": 0.07087036222219467, "global_step": 160578, "epoch": 1804, "lr": 3.5928846842994545e-05}
{"train_loss": 0.08950860053300858, "global_step": 160579, "epoch": 1804, "lr": 3.592829053809047e-05}
{"train_loss": 0.04396119341254234, "global_step": 160580, "epoch": 1804, "lr": 3.592773423507819e-05}
{"train_loss": 0.10181660205125809, "global_step": 160581, "epoch": 1804, "lr": 3.592717793395773e-05}
{"train_loss": 0.04178076609969139, "global_step": 160582, "epoch": 1804, "lr": 3.592662163472922e-05}
{"train_loss": 0.10540735721588135, "global_step": 160583, "epoch": 1804, "lr": 3.5926065337392677e-05}
{"train_loss": 0.11989181488752365, "global_step": 160584, "epoch": 1804, "lr": 3.592550904194823e-05}
{"train_loss": 0.11086046695709229, "global_step": 160585, "epoch": 1804, "lr": 3.592495274839589e-05}
{"train_loss": 0.055245302617549896, "global_step": 160586, "epoch": 1804, "lr": 3.59243964567358e-05}
{"train_loss": 0.07703397423028946, "global_step": 160587, "epoch": 1804, "lr": 3.592384016696798e-05}
{"train_loss": 0.04745905101299286, "global_step": 160588, "epoch": 1804, "lr": 3.592328387909255e-05}
{"train_loss": 0.1195402517914772, "global_step": 160589, "epoch": 1804, "lr": 3.592272759310955e-05}
{"train_loss": 0.12555396556854248, "global_step": 160590, "epoch": 1804, "lr": 3.592217130901907e-05}
{"train_loss": 0.034237176179885864, "global_step": 160591, "epoch": 1804, "lr": 3.592161502682119e-05}
{"train_loss": 0.09506554901599884, "global_step": 160592, "epoch": 1804, "lr": 3.592105874651597e-05}
{"train_loss": 0.17166441679000854, "global_step": 160593, "epoch": 1804, "lr": 3.592050246810349e-05}
{"train_loss": 0.12648116052150726, "global_step": 160594, "epoch": 1804, "lr": 3.591994619158383e-05}
{"train_loss": 0.08045221120119095, "global_step": 160595, "epoch": 1804, "lr": 3.591938991695706e-05}
{"train_loss": 0.09511170536279678, "global_step": 160596, "epoch": 1804, "lr": 3.591883364422325e-05}
{"train_loss": 0.09324007481336594, "global_step": 160597, "epoch": 1804, "lr": 3.591827737338249e-05}
{"train_loss": 0.07098591327667236, "global_step": 160598, "epoch": 1804, "lr": 3.591772110443484e-05}
{"train_loss": 0.10794452577829361, "global_step": 160599, "epoch": 1804, "lr": 3.5917164837380394e-05}
{"train_loss": 0.049523524940013885, "global_step": 160600, "epoch": 1804, "lr": 3.5916608572219196e-05}
{"train_loss": 0.07560352981090546, "global_step": 160601, "epoch": 1804, "lr": 3.591605230895135e-05}
{"train_loss": 0.0842798575758934, "global_step": 160602, "epoch": 1804, "lr": 3.591549604757691e-05}
{"train_loss": 0.04664582759141922, "global_step": 160603, "epoch": 1804, "lr": 3.591493978809598e-05}
{"train_loss": 0.14814145863056183, "global_step": 160604, "epoch": 1804, "lr": 3.591438353050859e-05}
{"train_loss": 0.08925413340330124, "global_step": 160605, "epoch": 1804, "lr": 3.591382727481486e-05}
{"train_loss": 0.03891386836767197, "global_step": 160606, "epoch": 1804, "lr": 3.591327102101482e-05}
{"train_loss": 0.10855230689048767, "global_step": 160607, "epoch": 1804, "lr": 3.59127147691086e-05}
{"train_loss": 0.10606882721185684, "global_step": 160608, "epoch": 1804, "lr": 3.591215851909621e-05}
{"train_loss": 0.06658565998077393, "global_step": 160609, "epoch": 1804, "lr": 3.591160227097778e-05}
{"train_loss": 0.05658334121108055, "global_step": 160610, "epoch": 1804, "lr": 3.591104602475337e-05}
{"train_loss": 0.17049188911914825, "global_step": 160611, "epoch": 1804, "lr": 3.591048978042303e-05}
{"train_loss": 0.11085487902164459, "global_step": 160612, "epoch": 1804, "lr": 3.590993353798686e-05}
{"train_loss": 0.13117459416389465, "global_step": 160613, "epoch": 1804, "lr": 3.5909377297444926e-05}
{"train_loss": 0.09683298319578171, "global_step": 160614, "epoch": 1804, "lr": 3.5908821058797315e-05}
{"train_loss": 0.1537567526102066, "global_step": 160615, "epoch": 1804, "lr": 3.590826482204407e-05}
{"train_loss": 0.08056648820638657, "global_step": 160616, "epoch": 1804, "lr": 3.5907708587185306e-05}
{"train_loss": 0.0711912214756012, "global_step": 160617, "epoch": 1804, "lr": 3.590715235422106e-05}
{"train_loss": 0.09041309356689453, "global_step": 160618, "epoch": 1804, "lr": 3.590659612315146e-05}
{"train_loss": 0.046752482652664185, "global_step": 160619, "epoch": 1804, "lr": 3.590603989397651e-05}
{"train_loss": 0.12182177603244781, "global_step": 160620, "epoch": 1804, "lr": 3.5905483666696343e-05}
{"train_loss": 0.1213863268494606, "global_step": 160621, "epoch": 1804, "lr": 3.590492744131099e-05}
{"train_loss": 0.16288979351520538, "global_step": 160622, "epoch": 1804, "lr": 3.5904371217820576e-05}
{"train_loss": 0.19624650478363037, "global_step": 160623, "epoch": 1804, "lr": 3.590381499622513e-05}
{"train_loss": 0.12233702838420868, "global_step": 160624, "epoch": 1804, "lr": 3.590325877652475e-05}
{"train_loss": 0.10956320911645889, "global_step": 160625, "epoch": 1804, "lr": 3.590270255871949e-05}
{"train_loss": 0.13781195878982544, "global_step": 160626, "epoch": 1804, "lr": 3.590214634280947e-05}
{"train_loss": 0.04448460042476654, "global_step": 160627, "epoch": 1804, "lr": 3.59015901287947e-05}
{"train_loss": 0.12607142329216003, "global_step": 160628, "epoch": 1804, "lr": 3.590103391667532e-05}
{"train_loss": 0.11527884006500244, "global_step": 160629, "epoch": 1804, "lr": 3.590047770645136e-05}
{"train_loss": 0.06743770837783813, "global_step": 160630, "epoch": 1804, "lr": 3.5899921498122914e-05}
{"train_loss": 0.0808805525302887, "global_step": 160631, "epoch": 1804, "lr": 3.589936529169005e-05}
{"train_loss": 0.06502223759889603, "global_step": 160632, "epoch": 1804, "lr": 3.5898809087152826e-05}
{"train_loss": 0.08795548975467682, "global_step": 160633, "epoch": 1804, "lr": 3.589825288451136e-05}
{"train_loss": 0.07478184252977371, "global_step": 160634, "epoch": 1804, "lr": 3.589769668376568e-05}
{"train_loss": 0.08547624200582504, "global_step": 160635, "epoch": 1804, "lr": 3.5897140484915915e-05}
{"train_loss": 0.13318921625614166, "global_step": 160636, "epoch": 1804, "lr": 3.589658428796208e-05}
{"train_loss": 0.07871674001216888, "global_step": 160637, "epoch": 1804, "lr": 3.589602809290429e-05}
{"train_loss": 0.10027677565813065, "global_step": 160638, "epoch": 1804, "lr": 3.58954718997426e-05}
{"train_loss": 0.10392080247402191, "global_step": 160639, "epoch": 1804, "lr": 3.5894915708477095e-05}
{"train_loss": 0.09266047924757004, "global_step": 160640, "epoch": 1804, "lr": 3.589435951910784e-05}
{"train_loss": 0.1532619148492813, "global_step": 160641, "epoch": 1804, "lr": 3.5893803331634934e-05}
{"train_loss": 0.12483643740415573, "global_step": 160642, "epoch": 1804, "lr": 3.589324714605841e-05}
{"train_loss": 0.11705697327852249, "global_step": 160643, "epoch": 1804, "lr": 3.589269096237839e-05}
{"train_loss": 0.0959916820100854, "global_step": 160644, "epoch": 1804, "lr": 3.589213478059491e-05, "val_loss": 6.764102458953857}
{"train_loss": 0.026369428262114525, "global_step": 160645, "epoch": 1805, "lr": 3.5891578600708075e-05}
{"train_loss": 0.11168675124645233, "global_step": 160646, "epoch": 1805, "lr": 3.5891022422717924e-05}
{"train_loss": 0.07858540862798691, "global_step": 160647, "epoch": 1805, "lr": 3.589046624662458e-05}
{"train_loss": 0.1231893002986908, "global_step": 160648, "epoch": 1805, "lr": 3.588991007242807e-05}
{"train_loss": 0.1172012984752655, "global_step": 160649, "epoch": 1805, "lr": 3.588935390012849e-05}
{"train_loss": 0.13291610777378082, "global_step": 160650, "epoch": 1805, "lr": 3.588879772972593e-05}
{"train_loss": 0.06516528129577637, "global_step": 160651, "epoch": 1805, "lr": 3.5888241561220434e-05}
{"train_loss": 0.12809863686561584, "global_step": 160652, "epoch": 1805, "lr": 3.58876853946121e-05}
{"train_loss": 0.07430636882781982, "global_step": 160653, "epoch": 1805, "lr": 3.588712922990098e-05}
{"train_loss": 0.08020788431167603, "global_step": 160654, "epoch": 1805, "lr": 3.5886573067087194e-05}
{"train_loss": 0.08415598422288895, "global_step": 160655, "epoch": 1805, "lr": 3.588601690617076e-05}
{"train_loss": 0.12390118837356567, "global_step": 160656, "epoch": 1805, "lr": 3.5885460747151796e-05}
{"train_loss": 0.053385064005851746, "global_step": 160657, "epoch": 1805, "lr": 3.588490459003034e-05}
{"train_loss": 0.10349306464195251, "global_step": 160658, "epoch": 1805, "lr": 3.588434843480651e-05}
{"train_loss": 0.11449311673641205, "global_step": 160659, "epoch": 1805, "lr": 3.588379228148033e-05}
{"train_loss": 0.10620532929897308, "global_step": 160660, "epoch": 1805, "lr": 3.588323613005193e-05}
{"train_loss": 0.050863638520240784, "global_step": 160661, "epoch": 1805, "lr": 3.588267998052133e-05}
{"train_loss": 0.09561058133840561, "global_step": 160662, "epoch": 1805, "lr": 3.588212383288866e-05}
{"train_loss": 0.129124253988266, "global_step": 160663, "epoch": 1805, "lr": 3.5881567687153935e-05}
{"train_loss": 0.09572754800319672, "global_step": 160664, "epoch": 1805, "lr": 3.588101154331729e-05}
{"train_loss": 0.053977787494659424, "global_step": 160665, "epoch": 1805, "lr": 3.588045540137875e-05}
{"train_loss": 0.07895932346582413, "global_step": 160666, "epoch": 1805, "lr": 3.5879899261338435e-05}
{"train_loss": 0.06640752404928207, "global_step": 160667, "epoch": 1805, "lr": 3.587934312319636e-05}
{"train_loss": 0.05294207111001015, "global_step": 160668, "epoch": 1805, "lr": 3.587878698695266e-05}
{"train_loss": 0.11447049677371979, "global_step": 160669, "epoch": 1805, "lr": 3.587823085260739e-05}
{"train_loss": 0.09560837596654892, "global_step": 160670, "epoch": 1805, "lr": 3.5877674720160604e-05}
{"train_loss": 0.06448127329349518, "global_step": 160671, "epoch": 1805, "lr": 3.58771185896124e-05}
{"train_loss": 0.07386548817157745, "global_step": 160672, "epoch": 1805, "lr": 3.5876562460962835e-05}
{"train_loss": 0.09726749360561371, "global_step": 160673, "epoch": 1805, "lr": 3.587600633421202e-05}
{"train_loss": 0.06516680866479874, "global_step": 160674, "epoch": 1805, "lr": 3.587545020935997e-05}
{"train_loss": 0.04708113521337509, "global_step": 160675, "epoch": 1805, "lr": 3.587489408640682e-05}
{"train_loss": 0.09027782827615738, "global_step": 160676, "epoch": 1805, "lr": 3.58743379653526e-05}
{"train_loss": 0.10160766541957855, "global_step": 160677, "epoch": 1805, "lr": 3.587378184619742e-05}
{"train_loss": 0.058126941323280334, "global_step": 160678, "epoch": 1805, "lr": 3.587322572894132e-05}
{"train_loss": 0.0545814074575901, "global_step": 160679, "epoch": 1805, "lr": 3.5872669613584414e-05}
{"train_loss": 0.05255461856722832, "global_step": 160680, "epoch": 1805, "lr": 3.587211350012673e-05}
{"train_loss": 0.10066452622413635, "global_step": 160681, "epoch": 1805, "lr": 3.5871557388568395e-05}
{"train_loss": 0.1020401120185852, "global_step": 160682, "epoch": 1805, "lr": 3.5871001278909435e-05}
{"train_loss": 0.07019112259149551, "global_step": 160683, "epoch": 1805, "lr": 3.587044517114996e-05}
{"train_loss": 0.05328909680247307, "global_step": 160684, "epoch": 1805, "lr": 3.586988906529002e-05}
{"train_loss": 0.12993338704109192, "global_step": 160685, "epoch": 1805, "lr": 3.58693329613297e-05}
{"train_loss": 0.14505112171173096, "global_step": 160686, "epoch": 1805, "lr": 3.586877685926909e-05}
{"train_loss": 0.1284780651330948, "global_step": 160687, "epoch": 1805, "lr": 3.586822075910824e-05}
{"train_loss": 0.060814619064331055, "global_step": 160688, "epoch": 1805, "lr": 3.5867664660847256e-05}
{"train_loss": 0.08372736722230911, "global_step": 160689, "epoch": 1805, "lr": 3.5867108564486164e-05}
{"train_loss": 0.12295115739107132, "global_step": 160690, "epoch": 1805, "lr": 3.586655247002509e-05}
{"train_loss": 0.09974567592144012, "global_step": 160691, "epoch": 1805, "lr": 3.586599637746406e-05}
{"train_loss": 0.0871797502040863, "global_step": 160692, "epoch": 1805, "lr": 3.58654402868032e-05}
{"train_loss": 0.09351768344640732, "global_step": 160693, "epoch": 1805, "lr": 3.5864884198042536e-05}
{"train_loss": 0.0647624209523201, "global_step": 160694, "epoch": 1805, "lr": 3.586432811118219e-05}
{"train_loss": 0.07276854664087296, "global_step": 160695, "epoch": 1805, "lr": 3.58637720262222e-05}
{"train_loss": 0.08389949053525925, "global_step": 160696, "epoch": 1805, "lr": 3.586321594316266e-05}
{"train_loss": 0.09883370995521545, "global_step": 160697, "epoch": 1805, "lr": 3.5862659862003624e-05}
{"train_loss": 0.10068590193986893, "global_step": 160698, "epoch": 1805, "lr": 3.58621037827452e-05}
{"train_loss": 0.10669643431901932, "global_step": 160699, "epoch": 1805, "lr": 3.586154770538742e-05}
{"train_loss": 0.08993695676326752, "global_step": 160700, "epoch": 1805, "lr": 3.586099162993041e-05}
{"train_loss": 0.11906006932258606, "global_step": 160701, "epoch": 1805, "lr": 3.58604355563742e-05}
{"train_loss": 0.05739854648709297, "global_step": 160702, "epoch": 1805, "lr": 3.585987948471889e-05}
{"train_loss": 0.11758003383874893, "global_step": 160703, "epoch": 1805, "lr": 3.585932341496453e-05}
{"train_loss": 0.09435059130191803, "global_step": 160704, "epoch": 1805, "lr": 3.585876734711123e-05}
{"train_loss": 0.0848345011472702, "global_step": 160705, "epoch": 1805, "lr": 3.585821128115904e-05}
{"train_loss": 0.11701223254203796, "global_step": 160706, "epoch": 1805, "lr": 3.585765521710805e-05}
{"train_loss": 0.06078571826219559, "global_step": 160707, "epoch": 1805, "lr": 3.585709915495832e-05}
{"train_loss": 0.14343953132629395, "global_step": 160708, "epoch": 1805, "lr": 3.5856543094709914e-05}
{"train_loss": 0.07759604603052139, "global_step": 160709, "epoch": 1805, "lr": 3.585598703636295e-05}
{"train_loss": 0.08597313612699509, "global_step": 160710, "epoch": 1805, "lr": 3.585543097991746e-05}
{"train_loss": 0.07542811334133148, "global_step": 160711, "epoch": 1805, "lr": 3.5854874925373544e-05}
{"train_loss": 0.08040965348482132, "global_step": 160712, "epoch": 1805, "lr": 3.585431887273125e-05}
{"train_loss": 0.09777404367923737, "global_step": 160713, "epoch": 1805, "lr": 3.58537628219907e-05}
{"train_loss": 0.10624466836452484, "global_step": 160714, "epoch": 1805, "lr": 3.585320677315191e-05}
{"train_loss": 0.11052485555410385, "global_step": 160715, "epoch": 1805, "lr": 3.5852650726215007e-05}
{"train_loss": 0.10155186802148819, "global_step": 160716, "epoch": 1805, "lr": 3.585209468118002e-05}
{"train_loss": 0.0614984966814518, "global_step": 160717, "epoch": 1805, "lr": 3.585153863804707e-05}
{"train_loss": 0.09772778302431107, "global_step": 160718, "epoch": 1805, "lr": 3.5850982596816186e-05}
{"train_loss": 0.1346137374639511, "global_step": 160719, "epoch": 1805, "lr": 3.585042655748748e-05}
{"train_loss": 0.0267450250685215, "global_step": 160720, "epoch": 1805, "lr": 3.5849870520061e-05}
{"train_loss": 0.10344960540533066, "global_step": 160721, "epoch": 1805, "lr": 3.584931448453686e-05}
{"train_loss": 0.10586746782064438, "global_step": 160722, "epoch": 1805, "lr": 3.584875845091507e-05}
{"train_loss": 0.06882428377866745, "global_step": 160723, "epoch": 1805, "lr": 3.584820241919575e-05}
{"train_loss": 0.15147671103477478, "global_step": 160724, "epoch": 1805, "lr": 3.5847646389378984e-05}
{"train_loss": 0.058930132538080215, "global_step": 160725, "epoch": 1805, "lr": 3.584709036146482e-05}
{"train_loss": 0.05544412136077881, "global_step": 160726, "epoch": 1805, "lr": 3.5846534335453354e-05}
{"train_loss": 0.0821278989315033, "global_step": 160727, "epoch": 1805, "lr": 3.584597831134463e-05}
{"train_loss": 0.10806389153003693, "global_step": 160728, "epoch": 1805, "lr": 3.5845422289138756e-05}
{"train_loss": 0.07540412247180939, "global_step": 160729, "epoch": 1805, "lr": 3.5844866268835776e-05}
{"train_loss": 0.17059089243412018, "global_step": 160730, "epoch": 1805, "lr": 3.58443102504358e-05}
{"train_loss": 0.04199492931365967, "global_step": 160731, "epoch": 1805, "lr": 3.584375423393886e-05}
{"train_loss": 0.0908300057053566, "global_step": 160732, "epoch": 1805, "lr": 3.584319821934509e-05}
{"train_loss": 0.09026212971448229, "global_step": 160733, "epoch": 1805, "lr": 3.58426422066545e-05, "val_loss": 6.843083381652832, "train_action_mse_error": 6.625659942626953}
{"train_loss": 0.07609772682189941, "global_step": 160734, "epoch": 1806, "lr": 3.584208619586721e-05}
{"train_loss": 0.16373951733112335, "global_step": 160735, "epoch": 1806, "lr": 3.584153018698326e-05}
{"train_loss": 0.10058575123548508, "global_step": 160736, "epoch": 1806, "lr": 3.5840974180002775e-05}
{"train_loss": 0.07068945467472076, "global_step": 160737, "epoch": 1806, "lr": 3.584041817492577e-05}
{"train_loss": 0.08044375479221344, "global_step": 160738, "epoch": 1806, "lr": 3.583986217175237e-05}
{"train_loss": 0.11130832135677338, "global_step": 160739, "epoch": 1806, "lr": 3.58393061704826e-05}
{"train_loss": 0.05985885113477707, "global_step": 160740, "epoch": 1806, "lr": 3.5838750171116576e-05}
{"train_loss": 0.14443066716194153, "global_step": 160741, "epoch": 1806, "lr": 3.583819417365436e-05}
{"train_loss": 0.07539527118206024, "global_step": 160742, "epoch": 1806, "lr": 3.583763817809603e-05}
{"train_loss": 0.0467316135764122, "global_step": 160743, "epoch": 1806, "lr": 3.583708218444166e-05}
{"train_loss": 0.10104693472385406, "global_step": 160744, "epoch": 1806, "lr": 3.583652619269131e-05}
{"train_loss": 0.10680819302797318, "global_step": 160745, "epoch": 1806, "lr": 3.583597020284509e-05}
{"train_loss": 0.09301996976137161, "global_step": 160746, "epoch": 1806, "lr": 3.5835414214903024e-05}
{"train_loss": 0.17347076535224915, "global_step": 160747, "epoch": 1806, "lr": 3.583485822886523e-05}
{"train_loss": 0.11453112959861755, "global_step": 160748, "epoch": 1806, "lr": 3.5834302244731755e-05}
{"train_loss": 0.0965019017457962, "global_step": 160749, "epoch": 1806, "lr": 3.58337462625027e-05}
{"train_loss": 0.10921549052000046, "global_step": 160750, "epoch": 1806, "lr": 3.583319028217811e-05}
{"train_loss": 0.104783795773983, "global_step": 160751, "epoch": 1806, "lr": 3.583263430375809e-05}
{"train_loss": 0.10045044869184494, "global_step": 160752, "epoch": 1806, "lr": 3.583207832724268e-05}
{"train_loss": 0.03181996941566467, "global_step": 160753, "epoch": 1806, "lr": 3.5831522352632e-05}
{"train_loss": 0.10876155644655228, "global_step": 160754, "epoch": 1806, "lr": 3.583096637992607e-05}
{"train_loss": 0.07030194997787476, "global_step": 160755, "epoch": 1806, "lr": 3.583041040912502e-05}
{"train_loss": 0.04117223247885704, "global_step": 160756, "epoch": 1806, "lr": 3.5829854440228876e-05}
{"train_loss": 0.08561912924051285, "global_step": 160757, "epoch": 1806, "lr": 3.582929847323776e-05}
{"train_loss": 0.11269622296094894, "global_step": 160758, "epoch": 1806, "lr": 3.5828742508151694e-05}
{"train_loss": 0.16795974969863892, "global_step": 160759, "epoch": 1806, "lr": 3.58281865449708e-05}
{"train_loss": 0.0984068289399147, "global_step": 160760, "epoch": 1806, "lr": 3.5827630583695125e-05}
{"train_loss": 0.08806261420249939, "global_step": 160761, "epoch": 1806, "lr": 3.582707462432476e-05}
{"train_loss": 0.07476749271154404, "global_step": 160762, "epoch": 1806, "lr": 3.5826518666859765e-05}
{"train_loss": 0.08289783447980881, "global_step": 160763, "epoch": 1806, "lr": 3.582596271130023e-05}
{"train_loss": 0.03990935906767845, "global_step": 160764, "epoch": 1806, "lr": 3.582540675764621e-05}
{"train_loss": 0.07285956293344498, "global_step": 160765, "epoch": 1806, "lr": 3.58248508058978e-05}
{"train_loss": 0.13806988298892975, "global_step": 160766, "epoch": 1806, "lr": 3.5824294856055073e-05}
{"train_loss": 0.09796315431594849, "global_step": 160767, "epoch": 1806, "lr": 3.582373890811808e-05}
{"train_loss": 0.08954105526208878, "global_step": 160768, "epoch": 1806, "lr": 3.5823182962086934e-05}
{"train_loss": 0.1394903063774109, "global_step": 160769, "epoch": 1806, "lr": 3.5822627017961664e-05}
{"train_loss": 0.06493350863456726, "global_step": 160770, "epoch": 1806, "lr": 3.582207107574239e-05}
{"train_loss": 0.13456721603870392, "global_step": 160771, "epoch": 1806, "lr": 3.582151513542914e-05}
{"train_loss": 0.10290412604808807, "global_step": 160772, "epoch": 1806, "lr": 3.582095919702204e-05}
{"train_loss": 0.10170220583677292, "global_step": 160773, "epoch": 1806, "lr": 3.582040326052112e-05}
{"train_loss": 0.10416653752326965, "global_step": 160774, "epoch": 1806, "lr": 3.5819847325926495e-05}
{"train_loss": 0.05589636415243149, "global_step": 160775, "epoch": 1806, "lr": 3.5819291393238197e-05}
{"train_loss": 0.10321281850337982, "global_step": 160776, "epoch": 1806, "lr": 3.581873546245632e-05}
{"train_loss": 0.08955467492341995, "global_step": 160777, "epoch": 1806, "lr": 3.581817953358096e-05}
{"train_loss": 0.07357367873191833, "global_step": 160778, "epoch": 1806, "lr": 3.581762360661216e-05}
{"train_loss": 0.15434153378009796, "global_step": 160779, "epoch": 1806, "lr": 3.581706768155003e-05}
{"train_loss": 0.057634662836790085, "global_step": 160780, "epoch": 1806, "lr": 3.58165117583946e-05}
{"train_loss": 0.07026327401399612, "global_step": 160781, "epoch": 1806, "lr": 3.581595583714597e-05}
{"train_loss": 0.08929269760847092, "global_step": 160782, "epoch": 1806, "lr": 3.5815399917804224e-05}
{"train_loss": 0.08288826793432236, "global_step": 160783, "epoch": 1806, "lr": 3.581484400036941e-05}
{"train_loss": 0.08631926029920578, "global_step": 160784, "epoch": 1806, "lr": 3.5814288084841626e-05}
{"train_loss": 0.135916069149971, "global_step": 160785, "epoch": 1806, "lr": 3.581373217122095e-05}
{"train_loss": 0.1031896099448204, "global_step": 160786, "epoch": 1806, "lr": 3.581317625950742e-05}
{"train_loss": 0.09721183776855469, "global_step": 160787, "epoch": 1806, "lr": 3.5812620349701155e-05}
{"train_loss": 0.0379110649228096, "global_step": 160788, "epoch": 1806, "lr": 3.5812064441802184e-05}
{"train_loss": 0.10450900346040726, "global_step": 160789, "epoch": 1806, "lr": 3.5811508535810636e-05}
{"train_loss": 0.0650220587849617, "global_step": 160790, "epoch": 1806, "lr": 3.581095263172654e-05}
{"train_loss": 0.08256284892559052, "global_step": 160791, "epoch": 1806, "lr": 3.581039672955001e-05}
{"train_loss": 0.08553856611251831, "global_step": 160792, "epoch": 1806, "lr": 3.5809840829281075e-05}
{"train_loss": 0.08482580631971359, "global_step": 160793, "epoch": 1806, "lr": 3.580928493091985e-05}
{"train_loss": 0.08711978793144226, "global_step": 160794, "epoch": 1806, "lr": 3.5808729034466375e-05}
{"train_loss": 0.06589271128177643, "global_step": 160795, "epoch": 1806, "lr": 3.580817313992074e-05}
{"train_loss": 0.10273738950490952, "global_step": 160796, "epoch": 1806, "lr": 3.5807617247283055e-05}
{"train_loss": 0.10697069764137268, "global_step": 160797, "epoch": 1806, "lr": 3.580706135655333e-05}
{"train_loss": 0.0522129125893116, "global_step": 160798, "epoch": 1806, "lr": 3.5806505467731706e-05}
{"train_loss": 0.07992769777774811, "global_step": 160799, "epoch": 1806, "lr": 3.580594958081821e-05}
{"train_loss": 0.07701513171195984, "global_step": 160800, "epoch": 1806, "lr": 3.580539369581293e-05}
{"train_loss": 0.07189924269914627, "global_step": 160801, "epoch": 1806, "lr": 3.580483781271594e-05}
{"train_loss": 0.08288467675447464, "global_step": 160802, "epoch": 1806, "lr": 3.580428193152731e-05}
{"train_loss": 0.07587293535470963, "global_step": 160803, "epoch": 1806, "lr": 3.5803726052247135e-05}
{"train_loss": 0.08258041739463806, "global_step": 160804, "epoch": 1806, "lr": 3.5803170174875476e-05}
{"train_loss": 0.09419254213571548, "global_step": 160805, "epoch": 1806, "lr": 3.580261429941239e-05}
{"train_loss": 0.05841695889830589, "global_step": 160806, "epoch": 1806, "lr": 3.580205842585799e-05}
{"train_loss": 0.07533791661262512, "global_step": 160807, "epoch": 1806, "lr": 3.580150255421231e-05}
{"train_loss": 0.06940068304538727, "global_step": 160808, "epoch": 1806, "lr": 3.580094668447547e-05}
{"train_loss": 0.08857709169387817, "global_step": 160809, "epoch": 1806, "lr": 3.5800390816647486e-05}
{"train_loss": 0.07513841986656189, "global_step": 160810, "epoch": 1806, "lr": 3.57998349507285e-05}
{"train_loss": 0.05847300961613655, "global_step": 160811, "epoch": 1806, "lr": 3.579927908671853e-05}
{"train_loss": 0.08088773488998413, "global_step": 160812, "epoch": 1806, "lr": 3.579872322461767e-05}
{"train_loss": 0.10152899473905563, "global_step": 160813, "epoch": 1806, "lr": 3.579816736442602e-05}
{"train_loss": 0.13676804304122925, "global_step": 160814, "epoch": 1806, "lr": 3.579761150614361e-05}
{"train_loss": 0.10547874122858047, "global_step": 160815, "epoch": 1806, "lr": 3.5797055649770575e-05}
{"train_loss": 0.11980431526899338, "global_step": 160816, "epoch": 1806, "lr": 3.5796499795306914e-05}
{"train_loss": 0.07648342847824097, "global_step": 160817, "epoch": 1806, "lr": 3.579594394275277e-05}
{"train_loss": 0.06798393279314041, "global_step": 160818, "epoch": 1806, "lr": 3.579538809210817e-05}
{"train_loss": 0.10149117559194565, "global_step": 160819, "epoch": 1806, "lr": 3.579483224337321e-05}
{"train_loss": 0.11218401789665222, "global_step": 160820, "epoch": 1806, "lr": 3.5794276396547966e-05}
{"train_loss": 0.13490891456604004, "global_step": 160821, "epoch": 1806, "lr": 3.57937205516325e-05}
{"train_loss": 0.09251593915599116, "global_step": 160822, "epoch": 1806, "lr": 3.5793164708626906e-05, "val_loss": 6.808901309967041}
{"train_loss": 0.09374958276748657, "global_step": 160823, "epoch": 1807, "lr": 3.5792608867531255e-05}
{"train_loss": 0.06926903873682022, "global_step": 160824, "epoch": 1807, "lr": 3.5792053028345596e-05}
{"train_loss": 0.07467704266309738, "global_step": 160825, "epoch": 1807, "lr": 3.5791497191070036e-05}
{"train_loss": 0.09755333513021469, "global_step": 160826, "epoch": 1807, "lr": 3.5790941355704624e-05}
{"train_loss": 0.13006456196308136, "global_step": 160827, "epoch": 1807, "lr": 3.579038552224946e-05}
{"train_loss": 0.0961824581027031, "global_step": 160828, "epoch": 1807, "lr": 3.578982969070458e-05}
{"train_loss": 0.09812455624341965, "global_step": 160829, "epoch": 1807, "lr": 3.578927386107012e-05}
{"train_loss": 0.06900253891944885, "global_step": 160830, "epoch": 1807, "lr": 3.578871803334608e-05}
{"train_loss": 0.18507008254528046, "global_step": 160831, "epoch": 1807, "lr": 3.5788162207532586e-05}
{"train_loss": 0.11363844573497772, "global_step": 160832, "epoch": 1807, "lr": 3.5787606383629725e-05}
{"train_loss": 0.11986149847507477, "global_step": 160833, "epoch": 1807, "lr": 3.578705056163751e-05}
{"train_loss": 0.08623749762773514, "global_step": 160834, "epoch": 1807, "lr": 3.5786494741556085e-05}
{"train_loss": 0.1944681853055954, "global_step": 160835, "epoch": 1807, "lr": 3.5785938923385476e-05}
{"train_loss": 0.1140560582280159, "global_step": 160836, "epoch": 1807, "lr": 3.5785383107125786e-05}
{"train_loss": 0.10853605717420578, "global_step": 160837, "epoch": 1807, "lr": 3.578482729277707e-05}
{"train_loss": 0.08240221440792084, "global_step": 160838, "epoch": 1807, "lr": 3.578427148033942e-05}
{"train_loss": 0.1316804140806198, "global_step": 160839, "epoch": 1807, "lr": 3.578371566981289e-05}
{"train_loss": 0.1060914397239685, "global_step": 160840, "epoch": 1807, "lr": 3.578315986119757e-05}
{"train_loss": 0.10930667072534561, "global_step": 160841, "epoch": 1807, "lr": 3.578260405449352e-05}
{"train_loss": 0.06636640429496765, "global_step": 160842, "epoch": 1807, "lr": 3.578204824970085e-05}
{"train_loss": 0.13331250846385956, "global_step": 160843, "epoch": 1807, "lr": 3.5781492446819584e-05}
{"train_loss": 0.06923484057188034, "global_step": 160844, "epoch": 1807, "lr": 3.578093664584985e-05}
{"train_loss": 0.07337062805891037, "global_step": 160845, "epoch": 1807, "lr": 3.5780380846791664e-05}
{"train_loss": 0.03987763077020645, "global_step": 160846, "epoch": 1807, "lr": 3.5779825049645164e-05}
{"train_loss": 0.09112595021724701, "global_step": 160847, "epoch": 1807, "lr": 3.5779269254410364e-05}
{"train_loss": 0.06056930124759674, "global_step": 160848, "epoch": 1807, "lr": 3.57787134610874e-05}
{"train_loss": 0.08329804241657257, "global_step": 160849, "epoch": 1807, "lr": 3.577815766967628e-05}
{"train_loss": 0.11694211512804031, "global_step": 160850, "epoch": 1807, "lr": 3.577760188017713e-05}
{"train_loss": 0.12789814174175262, "global_step": 160851, "epoch": 1807, "lr": 3.577704609259001e-05}
{"train_loss": 0.054087113589048386, "global_step": 160852, "epoch": 1807, "lr": 3.577649030691498e-05}
{"train_loss": 0.09935101121664047, "global_step": 160853, "epoch": 1807, "lr": 3.577593452315216e-05}
{"train_loss": 0.08128409087657928, "global_step": 160854, "epoch": 1807, "lr": 3.5775378741301554e-05}
{"train_loss": 0.08693884313106537, "global_step": 160855, "epoch": 1807, "lr": 3.5774822961363295e-05}
{"train_loss": 0.07840266823768616, "global_step": 160856, "epoch": 1807, "lr": 3.577426718333744e-05}
{"train_loss": 0.05166451260447502, "global_step": 160857, "epoch": 1807, "lr": 3.5773711407224055e-05}
{"train_loss": 0.09672174602746964, "global_step": 160858, "epoch": 1807, "lr": 3.5773155633023224e-05}
{"train_loss": 0.08652401715517044, "global_step": 160859, "epoch": 1807, "lr": 3.5772599860735014e-05}
{"train_loss": 0.0489172525703907, "global_step": 160860, "epoch": 1807, "lr": 3.577204409035951e-05}
{"train_loss": 0.1216738149523735, "global_step": 160861, "epoch": 1807, "lr": 3.577148832189678e-05}
{"train_loss": 0.06019557639956474, "global_step": 160862, "epoch": 1807, "lr": 3.57709325553469e-05}
{"train_loss": 0.09113828837871552, "global_step": 160863, "epoch": 1807, "lr": 3.5770376790709945e-05}
{"train_loss": 0.0778580904006958, "global_step": 160864, "epoch": 1807, "lr": 3.5769821027985984e-05}
{"train_loss": 0.04119360074400902, "global_step": 160865, "epoch": 1807, "lr": 3.576926526717511e-05}
{"train_loss": 0.030733175575733185, "global_step": 160866, "epoch": 1807, "lr": 3.5768709508277366e-05}
{"train_loss": 0.09705089777708054, "global_step": 160867, "epoch": 1807, "lr": 3.5768153751292846e-05}
{"train_loss": 0.12121398746967316, "global_step": 160868, "epoch": 1807, "lr": 3.5767597996221644e-05}
{"train_loss": 0.12633372843265533, "global_step": 160869, "epoch": 1807, "lr": 3.5767042243063795e-05}
{"train_loss": 0.10227303206920624, "global_step": 160870, "epoch": 1807, "lr": 3.5766486491819414e-05}
{"train_loss": 0.13420863449573517, "global_step": 160871, "epoch": 1807, "lr": 3.576593074248853e-05}
{"train_loss": 0.06986093521118164, "global_step": 160872, "epoch": 1807, "lr": 3.5765374995071264e-05}
{"train_loss": 0.07696295529603958, "global_step": 160873, "epoch": 1807, "lr": 3.576481924956766e-05}
{"train_loss": 0.08092319965362549, "global_step": 160874, "epoch": 1807, "lr": 3.576426350597781e-05}
{"train_loss": 0.10273852199316025, "global_step": 160875, "epoch": 1807, "lr": 3.5763707764301766e-05}
{"train_loss": 0.05243723839521408, "global_step": 160876, "epoch": 1807, "lr": 3.576315202453965e-05}
{"train_loss": 0.0650905966758728, "global_step": 160877, "epoch": 1807, "lr": 3.576259628669148e-05}
{"train_loss": 0.06502193957567215, "global_step": 160878, "epoch": 1807, "lr": 3.576204055075736e-05}
{"train_loss": 0.06003357842564583, "global_step": 160879, "epoch": 1807, "lr": 3.576148481673735e-05}
{"train_loss": 0.07385027408599854, "global_step": 160880, "epoch": 1807, "lr": 3.576092908463156e-05}
{"train_loss": 0.05643551051616669, "global_step": 160881, "epoch": 1807, "lr": 3.576037335444002e-05}
{"train_loss": 0.10733287036418915, "global_step": 160882, "epoch": 1807, "lr": 3.575981762616284e-05}
{"train_loss": 0.05653415620326996, "global_step": 160883, "epoch": 1807, "lr": 3.575926189980006e-05}
{"train_loss": 0.1006329208612442, "global_step": 160884, "epoch": 1807, "lr": 3.575870617535179e-05}
{"train_loss": 0.14090801775455475, "global_step": 160885, "epoch": 1807, "lr": 3.575815045281806e-05}
{"train_loss": 0.0648588016629219, "global_step": 160886, "epoch": 1807, "lr": 3.5757594732198985e-05}
{"train_loss": 0.09746488928794861, "global_step": 160887, "epoch": 1807, "lr": 3.575703901349464e-05}
{"train_loss": 0.18021664023399353, "global_step": 160888, "epoch": 1807, "lr": 3.575648329670508e-05}
{"train_loss": 0.11005418002605438, "global_step": 160889, "epoch": 1807, "lr": 3.5755927581830396e-05}
{"train_loss": 0.0807991698384285, "global_step": 160890, "epoch": 1807, "lr": 3.5755371868870636e-05}
{"train_loss": 0.10014790296554565, "global_step": 160891, "epoch": 1807, "lr": 3.5754816157825924e-05}
{"train_loss": 0.0817846730351448, "global_step": 160892, "epoch": 1807, "lr": 3.5754260448696264e-05}
{"train_loss": 0.02972285822033882, "global_step": 160893, "epoch": 1807, "lr": 3.57537047414818e-05}
{"train_loss": 0.05749417468905449, "global_step": 160894, "epoch": 1807, "lr": 3.575314903618256e-05}
{"train_loss": 0.15341350436210632, "global_step": 160895, "epoch": 1807, "lr": 3.575259333279865e-05}
{"train_loss": 0.14192849397659302, "global_step": 160896, "epoch": 1807, "lr": 3.5752037631330126e-05}
{"train_loss": 0.09372051805257797, "global_step": 160897, "epoch": 1807, "lr": 3.575148193177706e-05}
{"train_loss": 0.07744795083999634, "global_step": 160898, "epoch": 1807, "lr": 3.575092623413954e-05}
{"train_loss": 0.0625704750418663, "global_step": 160899, "epoch": 1807, "lr": 3.575037053841763e-05}
{"train_loss": 0.08404962718486786, "global_step": 160900, "epoch": 1807, "lr": 3.574981484461141e-05}
{"train_loss": 0.07377124577760696, "global_step": 160901, "epoch": 1807, "lr": 3.5749259152720956e-05}
{"train_loss": 0.09405391663312912, "global_step": 160902, "epoch": 1807, "lr": 3.574870346274634e-05}
{"train_loss": 0.10925053805112839, "global_step": 160903, "epoch": 1807, "lr": 3.574814777468762e-05}
{"train_loss": 0.06603889912366867, "global_step": 160904, "epoch": 1807, "lr": 3.5747592088544916e-05}
{"train_loss": 0.12538683414459229, "global_step": 160905, "epoch": 1807, "lr": 3.574703640431825e-05}
{"train_loss": 0.055734314024448395, "global_step": 160906, "epoch": 1807, "lr": 3.574648072200774e-05}
{"train_loss": 0.11377160251140594, "global_step": 160907, "epoch": 1807, "lr": 3.574592504161343e-05}
{"train_loss": 0.09289221465587616, "global_step": 160908, "epoch": 1807, "lr": 3.574536936313542e-05}
{"train_loss": 0.11001800000667572, "global_step": 160909, "epoch": 1807, "lr": 3.5744813686573756e-05}
{"train_loss": 0.11547934263944626, "global_step": 160910, "epoch": 1807, "lr": 3.5744258011928536e-05}
{"train_loss": 0.09191877336314555, "global_step": 160911, "epoch": 1807, "lr": 3.574370233919981e-05, "val_loss": 6.762791156768799}
{"train_loss": 0.13313567638397217, "global_step": 160912, "epoch": 1808, "lr": 3.574314666838771e-05}
{"train_loss": 0.09037624299526215, "global_step": 160913, "epoch": 1808, "lr": 3.574259099949223e-05}
{"train_loss": 0.06675540655851364, "global_step": 160914, "epoch": 1808, "lr": 3.574203533251351e-05}
{"train_loss": 0.060600653290748596, "global_step": 160915, "epoch": 1808, "lr": 3.574147966745159e-05}
{"train_loss": 0.05601247400045395, "global_step": 160916, "epoch": 1808, "lr": 3.5740924004306554e-05}
{"train_loss": 0.0871296226978302, "global_step": 160917, "epoch": 1808, "lr": 3.574036834307847e-05}
{"train_loss": 0.0951797291636467, "global_step": 160918, "epoch": 1808, "lr": 3.5739812683767434e-05}
{"train_loss": 0.11672622710466385, "global_step": 160919, "epoch": 1808, "lr": 3.5739257026373497e-05}
{"train_loss": 0.10604166984558105, "global_step": 160920, "epoch": 1808, "lr": 3.5738701370896754e-05}
{"train_loss": 0.1077638790011406, "global_step": 160921, "epoch": 1808, "lr": 3.573814571733725e-05}
{"train_loss": 0.0562351830303669, "global_step": 160922, "epoch": 1808, "lr": 3.573759006569508e-05}
{"train_loss": 0.0782223790884018, "global_step": 160923, "epoch": 1808, "lr": 3.573703441597033e-05}
{"train_loss": 0.06336215883493423, "global_step": 160924, "epoch": 1808, "lr": 3.573647876816305e-05}
{"train_loss": 0.12764360010623932, "global_step": 160925, "epoch": 1808, "lr": 3.573592312227334e-05}
{"train_loss": 0.05327614024281502, "global_step": 160926, "epoch": 1808, "lr": 3.573536747830125e-05}
{"train_loss": 0.08092372864484787, "global_step": 160927, "epoch": 1808, "lr": 3.573481183624688e-05}
{"train_loss": 0.042218729853630066, "global_step": 160928, "epoch": 1808, "lr": 3.573425619611027e-05}
{"train_loss": 0.11007843911647797, "global_step": 160929, "epoch": 1808, "lr": 3.5733700557891535e-05}
{"train_loss": 0.06658439338207245, "global_step": 160930, "epoch": 1808, "lr": 3.573314492159072e-05}
{"train_loss": 0.05696617066860199, "global_step": 160931, "epoch": 1808, "lr": 3.573258928720792e-05}
{"train_loss": 0.10771382600069046, "global_step": 160932, "epoch": 1808, "lr": 3.573203365474318e-05}
{"train_loss": 0.1275021880865097, "global_step": 160933, "epoch": 1808, "lr": 3.5731478024196616e-05}
{"train_loss": 0.07905779778957367, "global_step": 160934, "epoch": 1808, "lr": 3.573092239556827e-05}
{"train_loss": 0.10008165240287781, "global_step": 160935, "epoch": 1808, "lr": 3.573036676885822e-05}
{"train_loss": 0.06180785596370697, "global_step": 160936, "epoch": 1808, "lr": 3.5729811144066564e-05}
{"train_loss": 0.10842489451169968, "global_step": 160937, "epoch": 1808, "lr": 3.572925552119335e-05}
{"train_loss": 0.1308087706565857, "global_step": 160938, "epoch": 1808, "lr": 3.572869990023867e-05}
{"train_loss": 0.1425902545452118, "global_step": 160939, "epoch": 1808, "lr": 3.572814428120258e-05}
{"train_loss": 0.10863146930932999, "global_step": 160940, "epoch": 1808, "lr": 3.572758866408519e-05}
{"train_loss": 0.06177925318479538, "global_step": 160941, "epoch": 1808, "lr": 3.572703304888653e-05}
{"train_loss": 0.13220013678073883, "global_step": 160942, "epoch": 1808, "lr": 3.572647743560671e-05}
{"train_loss": 0.14121249318122864, "global_step": 160943, "epoch": 1808, "lr": 3.5725921824245776e-05}
{"train_loss": 0.09400484710931778, "global_step": 160944, "epoch": 1808, "lr": 3.572536621480384e-05}
{"train_loss": 0.10789847373962402, "global_step": 160945, "epoch": 1808, "lr": 3.572481060728093e-05}
{"train_loss": 0.07245969772338867, "global_step": 160946, "epoch": 1808, "lr": 3.572425500167716e-05}
{"train_loss": 0.11031975597143173, "global_step": 160947, "epoch": 1808, "lr": 3.572369939799258e-05}
{"train_loss": 0.10525202006101608, "global_step": 160948, "epoch": 1808, "lr": 3.5723143796227296e-05}
{"train_loss": 0.06454090029001236, "global_step": 160949, "epoch": 1808, "lr": 3.572258819638133e-05}
{"train_loss": 0.14903026819229126, "global_step": 160950, "epoch": 1808, "lr": 3.572203259845483e-05}
{"train_loss": 0.0970136821269989, "global_step": 160951, "epoch": 1808, "lr": 3.572147700244779e-05}
{"train_loss": 0.06180759519338608, "global_step": 160952, "epoch": 1808, "lr": 3.572092140836034e-05}
{"train_loss": 0.043106559664011, "global_step": 160953, "epoch": 1808, "lr": 3.5720365816192534e-05}
{"train_loss": 0.0775868147611618, "global_step": 160954, "epoch": 1808, "lr": 3.571981022594446e-05}
{"train_loss": 0.07519238442182541, "global_step": 160955, "epoch": 1808, "lr": 3.571925463761617e-05}
{"train_loss": 0.05591791868209839, "global_step": 160956, "epoch": 1808, "lr": 3.571869905120776e-05}
{"train_loss": 0.12242235988378525, "global_step": 160957, "epoch": 1808, "lr": 3.57181434667193e-05}
{"train_loss": 0.08415812253952026, "global_step": 160958, "epoch": 1808, "lr": 3.5717587884150835e-05}
{"train_loss": 0.07690150290727615, "global_step": 160959, "epoch": 1808, "lr": 3.571703230350251e-05}
{"train_loss": 0.08375383913516998, "global_step": 160960, "epoch": 1808, "lr": 3.5716476724774316e-05}
{"train_loss": 0.08150206506252289, "global_step": 160961, "epoch": 1808, "lr": 3.571592114796639e-05}
{"train_loss": 0.09869498014450073, "global_step": 160962, "epoch": 1808, "lr": 3.571536557307877e-05}
{"train_loss": 0.11731158941984177, "global_step": 160963, "epoch": 1808, "lr": 3.5714810000111564e-05}
{"train_loss": 0.13245050609111786, "global_step": 160964, "epoch": 1808, "lr": 3.571425442906481e-05}
{"train_loss": 0.0501069575548172, "global_step": 160965, "epoch": 1808, "lr": 3.5713698859938624e-05}
{"train_loss": 0.12665458023548126, "global_step": 160966, "epoch": 1808, "lr": 3.571314329273303e-05}
{"train_loss": 0.11864453554153442, "global_step": 160967, "epoch": 1808, "lr": 3.571258772744815e-05}
{"train_loss": 0.12891508638858795, "global_step": 160968, "epoch": 1808, "lr": 3.571203216408402e-05}
{"train_loss": 0.1390223205089569, "global_step": 160969, "epoch": 1808, "lr": 3.5711476602640746e-05}
{"train_loss": 0.12604749202728271, "global_step": 160970, "epoch": 1808, "lr": 3.571092104311838e-05}
{"train_loss": 0.058062706142663956, "global_step": 160971, "epoch": 1808, "lr": 3.571036548551703e-05}
{"train_loss": 0.1015813872218132, "global_step": 160972, "epoch": 1808, "lr": 3.5709809929836726e-05}
{"train_loss": 0.20924882590770721, "global_step": 160973, "epoch": 1808, "lr": 3.570925437607757e-05}
{"train_loss": 0.08138895779848099, "global_step": 160974, "epoch": 1808, "lr": 3.570869882423963e-05}
{"train_loss": 0.08347573131322861, "global_step": 160975, "epoch": 1808, "lr": 3.570814327432297e-05}
{"train_loss": 0.07747562974691391, "global_step": 160976, "epoch": 1808, "lr": 3.570758772632769e-05}
{"train_loss": 0.10336922109127045, "global_step": 160977, "epoch": 1808, "lr": 3.570703218025384e-05}
{"train_loss": 0.09529316425323486, "global_step": 160978, "epoch": 1808, "lr": 3.570647663610152e-05}
{"train_loss": 0.11523713916540146, "global_step": 160979, "epoch": 1808, "lr": 3.570592109387077e-05}
{"train_loss": 0.06575234979391098, "global_step": 160980, "epoch": 1808, "lr": 3.570536555356171e-05}
{"train_loss": 0.1370425969362259, "global_step": 160981, "epoch": 1808, "lr": 3.570481001517436e-05}
{"train_loss": 0.1273505836725235, "global_step": 160982, "epoch": 1808, "lr": 3.570425447870885e-05}
{"train_loss": 0.07979083806276321, "global_step": 160983, "epoch": 1808, "lr": 3.5703698944165207e-05}
{"train_loss": 0.06217031925916672, "global_step": 160984, "epoch": 1808, "lr": 3.570314341154355e-05}
{"train_loss": 0.10204453766345978, "global_step": 160985, "epoch": 1808, "lr": 3.570258788084391e-05}
{"train_loss": 0.12276049703359604, "global_step": 160986, "epoch": 1808, "lr": 3.5702032352066404e-05}
{"train_loss": 0.1083073616027832, "global_step": 160987, "epoch": 1808, "lr": 3.5701476825211066e-05}
{"train_loss": 0.16228356957435608, "global_step": 160988, "epoch": 1808, "lr": 3.570092130027801e-05}
{"train_loss": 0.07164428383111954, "global_step": 160989, "epoch": 1808, "lr": 3.570036577726727e-05}
{"train_loss": 0.05604506656527519, "global_step": 160990, "epoch": 1808, "lr": 3.5699810256178956e-05}
{"train_loss": 0.10031171143054962, "global_step": 160991, "epoch": 1808, "lr": 3.569925473701312e-05}
{"train_loss": 0.10386595129966736, "global_step": 160992, "epoch": 1808, "lr": 3.5698699219769855e-05}
{"train_loss": 0.12163590639829636, "global_step": 160993, "epoch": 1808, "lr": 3.5698143704449215e-05}
{"train_loss": 0.04138638451695442, "global_step": 160994, "epoch": 1808, "lr": 3.569758819105128e-05}
{"train_loss": 0.07839877903461456, "global_step": 160995, "epoch": 1808, "lr": 3.569703267957615e-05}
{"train_loss": 0.06313207745552063, "global_step": 160996, "epoch": 1808, "lr": 3.5696477170023856e-05}
{"train_loss": 0.11731097102165222, "global_step": 160997, "epoch": 1808, "lr": 3.569592166239452e-05}
{"train_loss": 0.10213035345077515, "global_step": 160998, "epoch": 1808, "lr": 3.569536615668817e-05}
{"train_loss": 0.09522271901369095, "global_step": 160999, "epoch": 1808, "lr": 3.569481065290493e-05}
{"train_loss": 0.09590188455715608, "global_step": 161000, "epoch": 1808, "lr": 3.569425515104482e-05, "val_loss": 6.718753814697266}
{"train_loss": 0.08684968203306198, "global_step": 161001, "epoch": 1809, "lr": 3.569369965110797e-05}
{"train_loss": 0.08297150582075119, "global_step": 161002, "epoch": 1809, "lr": 3.56931441530944e-05}
{"train_loss": 0.07327266037464142, "global_step": 161003, "epoch": 1809, "lr": 3.569258865700424e-05}
{"train_loss": 0.10305017232894897, "global_step": 161004, "epoch": 1809, "lr": 3.569203316283751e-05}
{"train_loss": 0.07700168341398239, "global_step": 161005, "epoch": 1809, "lr": 3.5691477670594344e-05}
{"train_loss": 0.08506151288747787, "global_step": 161006, "epoch": 1809, "lr": 3.5690922180274755e-05}
{"train_loss": 0.1298893243074417, "global_step": 161007, "epoch": 1809, "lr": 3.5690366691878866e-05}
{"train_loss": 0.057458486407995224, "global_step": 161008, "epoch": 1809, "lr": 3.568981120540671e-05}
{"train_loss": 0.05482759699225426, "global_step": 161009, "epoch": 1809, "lr": 3.568925572085842e-05}
{"train_loss": 0.06805209815502167, "global_step": 161010, "epoch": 1809, "lr": 3.5688700238233996e-05}
{"train_loss": 0.169671431183815, "global_step": 161011, "epoch": 1809, "lr": 3.5688144757533584e-05}
{"train_loss": 0.08665355294942856, "global_step": 161012, "epoch": 1809, "lr": 3.568758927875722e-05}
{"train_loss": 0.08557207882404327, "global_step": 161013, "epoch": 1809, "lr": 3.568703380190496e-05}
{"train_loss": 0.08691348135471344, "global_step": 161014, "epoch": 1809, "lr": 3.5686478326976935e-05}
{"train_loss": 0.08219220489263535, "global_step": 161015, "epoch": 1809, "lr": 3.5685922853973166e-05}
{"train_loss": 0.1424431949853897, "global_step": 161016, "epoch": 1809, "lr": 3.568536738289377e-05}
{"train_loss": 0.15004447102546692, "global_step": 161017, "epoch": 1809, "lr": 3.568481191373878e-05}
{"train_loss": 0.07391902804374695, "global_step": 161018, "epoch": 1809, "lr": 3.5684256446508315e-05}
{"train_loss": 0.0751606822013855, "global_step": 161019, "epoch": 1809, "lr": 3.568370098120241e-05}
{"train_loss": 0.09478496015071869, "global_step": 161020, "epoch": 1809, "lr": 3.568314551782117e-05}
{"train_loss": 0.12644250690937042, "global_step": 161021, "epoch": 1809, "lr": 3.568259005636464e-05}
{"train_loss": 0.0757058709859848, "global_step": 161022, "epoch": 1809, "lr": 3.568203459683293e-05}
{"train_loss": 0.05612976476550102, "global_step": 161023, "epoch": 1809, "lr": 3.5681479139226084e-05}
{"train_loss": 0.17086642980575562, "global_step": 161024, "epoch": 1809, "lr": 3.56809236835442e-05}
{"train_loss": 0.07084392756223679, "global_step": 161025, "epoch": 1809, "lr": 3.568036822978732e-05}
{"train_loss": 0.09385232627391815, "global_step": 161026, "epoch": 1809, "lr": 3.5679812777955564e-05}
{"train_loss": 0.06700067967176437, "global_step": 161027, "epoch": 1809, "lr": 3.567925732804896e-05}
{"train_loss": 0.09883996844291687, "global_step": 161028, "epoch": 1809, "lr": 3.567870188006763e-05}
{"train_loss": 0.09782706201076508, "global_step": 161029, "epoch": 1809, "lr": 3.56781464340116e-05}
{"train_loss": 0.0817485898733139, "global_step": 161030, "epoch": 1809, "lr": 3.567759098988098e-05}
{"train_loss": 0.055791597813367844, "global_step": 161031, "epoch": 1809, "lr": 3.5677035547675844e-05}
{"train_loss": 0.13618028163909912, "global_step": 161032, "epoch": 1809, "lr": 3.567648010739623e-05}
{"train_loss": 0.1356711983680725, "global_step": 161033, "epoch": 1809, "lr": 3.567592466904226e-05}
{"train_loss": 0.1508660912513733, "global_step": 161034, "epoch": 1809, "lr": 3.567536923261397e-05}
{"train_loss": 0.07855455577373505, "global_step": 161035, "epoch": 1809, "lr": 3.567481379811148e-05}
{"train_loss": 0.20324669778347015, "global_step": 161036, "epoch": 1809, "lr": 3.567425836553481e-05}
{"train_loss": 0.0675809234380722, "global_step": 161037, "epoch": 1809, "lr": 3.5673702934884076e-05}
{"train_loss": 0.0685109943151474, "global_step": 161038, "epoch": 1809, "lr": 3.567314750615933e-05}
{"train_loss": 0.07830630242824554, "global_step": 161039, "epoch": 1809, "lr": 3.5672592079360664e-05}
{"train_loss": 0.06143657863140106, "global_step": 161040, "epoch": 1809, "lr": 3.567203665448813e-05}
{"train_loss": 0.05401841178536415, "global_step": 161041, "epoch": 1809, "lr": 3.567148123154183e-05}
{"train_loss": 0.14116300642490387, "global_step": 161042, "epoch": 1809, "lr": 3.56709258105218e-05}
{"train_loss": 0.0998331606388092, "global_step": 161043, "epoch": 1809, "lr": 3.5670370391428177e-05}
{"train_loss": 0.12324783205986023, "global_step": 161044, "epoch": 1809, "lr": 3.566981497426096e-05}
{"train_loss": 0.060269810259342194, "global_step": 161045, "epoch": 1809, "lr": 3.5669259559020294e-05}
{"train_loss": 0.08038213104009628, "global_step": 161046, "epoch": 1809, "lr": 3.56687041457062e-05}
{"train_loss": 0.09646415710449219, "global_step": 161047, "epoch": 1809, "lr": 3.56681487343188e-05}
{"train_loss": 0.053683631122112274, "global_step": 161048, "epoch": 1809, "lr": 3.5667593324858106e-05}
{"train_loss": 0.09218216687440872, "global_step": 161049, "epoch": 1809, "lr": 3.5667037917324244e-05}
{"train_loss": 0.05283690243959427, "global_step": 161050, "epoch": 1809, "lr": 3.5666482511717295e-05}
{"train_loss": 0.07539688795804977, "global_step": 161051, "epoch": 1809, "lr": 3.566592710803728e-05}
{"train_loss": 0.051793172955513, "global_step": 161052, "epoch": 1809, "lr": 3.5665371706284335e-05}
{"train_loss": 0.09831738471984863, "global_step": 161053, "epoch": 1809, "lr": 3.566481630645848e-05}
{"train_loss": 0.14298982918262482, "global_step": 161054, "epoch": 1809, "lr": 3.566426090855984e-05}
{"train_loss": 0.11506123840808868, "global_step": 161055, "epoch": 1809, "lr": 3.566370551258844e-05}
{"train_loss": 0.08021258562803268, "global_step": 161056, "epoch": 1809, "lr": 3.5663150118544407e-05}
{"train_loss": 0.08546920865774155, "global_step": 161057, "epoch": 1809, "lr": 3.5662594726427765e-05}
{"train_loss": 0.08444242179393768, "global_step": 161058, "epoch": 1809, "lr": 3.566203933623863e-05}
{"train_loss": 0.16373322904109955, "global_step": 161059, "epoch": 1809, "lr": 3.5661483947977046e-05}
{"train_loss": 0.05799764767289162, "global_step": 161060, "epoch": 1809, "lr": 3.566092856164311e-05}
{"train_loss": 0.1495581418275833, "global_step": 161061, "epoch": 1809, "lr": 3.566037317723688e-05}
{"train_loss": 0.1564590334892273, "global_step": 161062, "epoch": 1809, "lr": 3.565981779475845e-05}
{"train_loss": 0.08223003149032593, "global_step": 161063, "epoch": 1809, "lr": 3.565926241420787e-05}
{"train_loss": 0.08558513224124908, "global_step": 161064, "epoch": 1809, "lr": 3.565870703558524e-05}
{"train_loss": 0.10708814114332199, "global_step": 161065, "epoch": 1809, "lr": 3.56581516588906e-05}
{"train_loss": 0.1167018711566925, "global_step": 161066, "epoch": 1809, "lr": 3.565759628412406e-05}
{"train_loss": 0.09501739591360092, "global_step": 161067, "epoch": 1809, "lr": 3.5657040911285686e-05}
{"train_loss": 0.08694557100534439, "global_step": 161068, "epoch": 1809, "lr": 3.565648554037555e-05}
{"train_loss": 0.11337858438491821, "global_step": 161069, "epoch": 1809, "lr": 3.565593017139372e-05}
{"train_loss": 0.1250336468219757, "global_step": 161070, "epoch": 1809, "lr": 3.565537480434027e-05}
{"train_loss": 0.08100571483373642, "global_step": 161071, "epoch": 1809, "lr": 3.565481943921529e-05}
{"train_loss": 0.0921565592288971, "global_step": 161072, "epoch": 1809, "lr": 3.565426407601883e-05}
{"train_loss": 0.15169686079025269, "global_step": 161073, "epoch": 1809, "lr": 3.565370871475101e-05}
{"train_loss": 0.10230185836553574, "global_step": 161074, "epoch": 1809, "lr": 3.565315335541184e-05}
{"train_loss": 0.1728319227695465, "global_step": 161075, "epoch": 1809, "lr": 3.565259799800145e-05}
{"train_loss": 0.08355244249105453, "global_step": 161076, "epoch": 1809, "lr": 3.565204264251987e-05}
{"train_loss": 0.09852059185504913, "global_step": 161077, "epoch": 1809, "lr": 3.565148728896722e-05}
{"train_loss": 0.10880885273218155, "global_step": 161078, "epoch": 1809, "lr": 3.565093193734354e-05}
{"train_loss": 0.06933590769767761, "global_step": 161079, "epoch": 1809, "lr": 3.5650376587648936e-05}
{"train_loss": 0.10463372617959976, "global_step": 161080, "epoch": 1809, "lr": 3.564982123988344e-05}
{"train_loss": 0.0782657265663147, "global_step": 161081, "epoch": 1809, "lr": 3.564926589404717e-05}
{"train_loss": 0.06767229735851288, "global_step": 161082, "epoch": 1809, "lr": 3.5648710550140165e-05}
{"train_loss": 0.14485031366348267, "global_step": 161083, "epoch": 1809, "lr": 3.5648155208162527e-05}
{"train_loss": 0.08217174559831619, "global_step": 161084, "epoch": 1809, "lr": 3.5647599868114315e-05}
{"train_loss": 0.048056840896606445, "global_step": 161085, "epoch": 1809, "lr": 3.5647044529995596e-05}
{"train_loss": 0.07430412620306015, "global_step": 161086, "epoch": 1809, "lr": 3.564648919380647e-05}
{"train_loss": 0.08485903590917587, "global_step": 161087, "epoch": 1809, "lr": 3.564593385954701e-05}
{"train_loss": 0.03533457964658737, "global_step": 161088, "epoch": 1809, "lr": 3.5645378527217276e-05}
{"train_loss": 0.09575845778323291, "global_step": 161089, "epoch": 1809, "lr": 3.5644823196817325e-05, "val_loss": 6.773340702056885}
{"train_loss": 0.13982819020748138, "global_step": 161090, "epoch": 1810, "lr": 3.564426786834728e-05}
{"train_loss": 0.07712292671203613, "global_step": 161091, "epoch": 1810, "lr": 3.5643712541807164e-05}
{"train_loss": 0.09343498945236206, "global_step": 161092, "epoch": 1810, "lr": 3.564315721719709e-05}
{"train_loss": 0.06795438379049301, "global_step": 161093, "epoch": 1810, "lr": 3.5642601894517105e-05}
{"train_loss": 0.12434717267751694, "global_step": 161094, "epoch": 1810, "lr": 3.564204657376732e-05}
{"train_loss": 0.08126100897789001, "global_step": 161095, "epoch": 1810, "lr": 3.564149125494776e-05}
{"train_loss": 0.10071417689323425, "global_step": 161096, "epoch": 1810, "lr": 3.564093593805855e-05}
{"train_loss": 0.12205395102500916, "global_step": 161097, "epoch": 1810, "lr": 3.564038062309972e-05}
{"train_loss": 0.11848493665456772, "global_step": 161098, "epoch": 1810, "lr": 3.5639825310071385e-05}
{"train_loss": 0.0778418779373169, "global_step": 161099, "epoch": 1810, "lr": 3.5639269998973587e-05}
{"train_loss": 0.15702860057353973, "global_step": 161100, "epoch": 1810, "lr": 3.563871468980643e-05}
{"train_loss": 0.11852019280195236, "global_step": 161101, "epoch": 1810, "lr": 3.563815938256995e-05}
{"train_loss": 0.08347508311271667, "global_step": 161102, "epoch": 1810, "lr": 3.563760407726426e-05}
{"train_loss": 0.10804345458745956, "global_step": 161103, "epoch": 1810, "lr": 3.563704877388941e-05}
{"train_loss": 0.1082547977566719, "global_step": 161104, "epoch": 1810, "lr": 3.5636493472445485e-05}
{"train_loss": 0.18336021900177002, "global_step": 161105, "epoch": 1810, "lr": 3.563593817293257e-05}
{"train_loss": 0.08692348748445511, "global_step": 161106, "epoch": 1810, "lr": 3.563538287535072e-05}
{"train_loss": 0.07378814369440079, "global_step": 161107, "epoch": 1810, "lr": 3.563482757970001e-05}
{"train_loss": 0.07894636690616608, "global_step": 161108, "epoch": 1810, "lr": 3.563427228598054e-05}
{"train_loss": 0.0698079839348793, "global_step": 161109, "epoch": 1810, "lr": 3.5633716994192367e-05}
{"train_loss": 0.1406697928905487, "global_step": 161110, "epoch": 1810, "lr": 3.5633161704335546e-05}
{"train_loss": 0.09974121302366257, "global_step": 161111, "epoch": 1810, "lr": 3.5632606416410184e-05}
{"train_loss": 0.09185301512479782, "global_step": 161112, "epoch": 1810, "lr": 3.5632051130416335e-05}
{"train_loss": 0.10003825277090073, "global_step": 161113, "epoch": 1810, "lr": 3.56314958463541e-05}
{"train_loss": 0.06750588864088058, "global_step": 161114, "epoch": 1810, "lr": 3.5630940564223514e-05}
{"train_loss": 0.1937686651945114, "global_step": 161115, "epoch": 1810, "lr": 3.563038528402469e-05}
{"train_loss": 0.07765817642211914, "global_step": 161116, "epoch": 1810, "lr": 3.562983000575767e-05}
{"train_loss": 0.07169386744499207, "global_step": 161117, "epoch": 1810, "lr": 3.5629274729422555e-05}
{"train_loss": 0.1271800994873047, "global_step": 161118, "epoch": 1810, "lr": 3.5628719455019396e-05}
{"train_loss": 0.09963968396186829, "global_step": 161119, "epoch": 1810, "lr": 3.56281641825483e-05}
{"train_loss": 0.049233801662921906, "global_step": 161120, "epoch": 1810, "lr": 3.5627608912009305e-05}
{"train_loss": 0.06009780243039131, "global_step": 161121, "epoch": 1810, "lr": 3.5627053643402494e-05}
{"train_loss": 0.09028320014476776, "global_step": 161122, "epoch": 1810, "lr": 3.5626498376727976e-05}
{"train_loss": 0.09069705009460449, "global_step": 161123, "epoch": 1810, "lr": 3.562594311198579e-05}
{"train_loss": 0.1085495725274086, "global_step": 161124, "epoch": 1810, "lr": 3.562538784917602e-05}
{"train_loss": 0.09278832376003265, "global_step": 161125, "epoch": 1810, "lr": 3.5624832588298737e-05}
{"train_loss": 0.04832267016172409, "global_step": 161126, "epoch": 1810, "lr": 3.562427732935403e-05}
{"train_loss": 0.03910689800977707, "global_step": 161127, "epoch": 1810, "lr": 3.5623722072341956e-05}
{"train_loss": 0.06411847472190857, "global_step": 161128, "epoch": 1810, "lr": 3.5623166817262614e-05}
{"train_loss": 0.07937202602624893, "global_step": 161129, "epoch": 1810, "lr": 3.562261156411604e-05}
{"train_loss": 0.08055900037288666, "global_step": 161130, "epoch": 1810, "lr": 3.562205631290235e-05}
{"train_loss": 0.08558286726474762, "global_step": 161131, "epoch": 1810, "lr": 3.5621501063621573e-05}
{"train_loss": 0.08438495546579361, "global_step": 161132, "epoch": 1810, "lr": 3.562094581627384e-05}
{"train_loss": 0.06516337394714355, "global_step": 161133, "epoch": 1810, "lr": 3.562039057085917e-05}
{"train_loss": 0.05539771914482117, "global_step": 161134, "epoch": 1810, "lr": 3.561983532737768e-05}
{"train_loss": 0.0899757444858551, "global_step": 161135, "epoch": 1810, "lr": 3.561928008582941e-05}
{"train_loss": 0.07271422445774078, "global_step": 161136, "epoch": 1810, "lr": 3.561872484621448e-05}
{"train_loss": 0.060870859771966934, "global_step": 161137, "epoch": 1810, "lr": 3.561816960853291e-05}
{"train_loss": 0.1236744076013565, "global_step": 161138, "epoch": 1810, "lr": 3.5617614372784825e-05}
{"train_loss": 0.1285095512866974, "global_step": 161139, "epoch": 1810, "lr": 3.561705913897025e-05}
{"train_loss": 0.11061248928308487, "global_step": 161140, "epoch": 1810, "lr": 3.561650390708929e-05}
{"train_loss": 0.14964228868484497, "global_step": 161141, "epoch": 1810, "lr": 3.561594867714204e-05}
{"train_loss": 0.09995128214359283, "global_step": 161142, "epoch": 1810, "lr": 3.561539344912854e-05}
{"train_loss": 0.08179410547018051, "global_step": 161143, "epoch": 1810, "lr": 3.561483822304887e-05}
{"train_loss": 0.061413705348968506, "global_step": 161144, "epoch": 1810, "lr": 3.5614282998903115e-05}
{"train_loss": 0.08740490674972534, "global_step": 161145, "epoch": 1810, "lr": 3.5613727776691335e-05}
{"train_loss": 0.13630066812038422, "global_step": 161146, "epoch": 1810, "lr": 3.561317255641362e-05}
{"train_loss": 0.09139355272054672, "global_step": 161147, "epoch": 1810, "lr": 3.561261733807005e-05}
{"train_loss": 0.11346478015184402, "global_step": 161148, "epoch": 1810, "lr": 3.5612062121660664e-05}
{"train_loss": 0.08604352176189423, "global_step": 161149, "epoch": 1810, "lr": 3.561150690718559e-05}
{"train_loss": 0.0911712795495987, "global_step": 161150, "epoch": 1810, "lr": 3.561095169464484e-05}
{"train_loss": 0.10147224366664886, "global_step": 161151, "epoch": 1810, "lr": 3.561039648403855e-05}
{"train_loss": 0.05872195214033127, "global_step": 161152, "epoch": 1810, "lr": 3.560984127536675e-05}
{"train_loss": 0.10719341784715652, "global_step": 161153, "epoch": 1810, "lr": 3.560928606862954e-05}
{"train_loss": 0.08991796523332596, "global_step": 161154, "epoch": 1810, "lr": 3.560873086382698e-05}
{"train_loss": 0.06362010538578033, "global_step": 161155, "epoch": 1810, "lr": 3.560817566095916e-05}
{"train_loss": 0.1024104654788971, "global_step": 161156, "epoch": 1810, "lr": 3.560762046002614e-05}
{"train_loss": 0.06099263206124306, "global_step": 161157, "epoch": 1810, "lr": 3.5607065261027986e-05}
{"train_loss": 0.09551099687814713, "global_step": 161158, "epoch": 1810, "lr": 3.5606510063964804e-05}
{"train_loss": 0.12265270203351974, "global_step": 161159, "epoch": 1810, "lr": 3.5605954868836644e-05}
{"train_loss": 0.050151046365499496, "global_step": 161160, "epoch": 1810, "lr": 3.56053996756436e-05}
{"train_loss": 0.055316630750894547, "global_step": 161161, "epoch": 1810, "lr": 3.560484448438573e-05}
{"train_loss": 0.0390465222299099, "global_step": 161162, "epoch": 1810, "lr": 3.5604289295063104e-05}
{"train_loss": 0.045106079429388046, "global_step": 161163, "epoch": 1810, "lr": 3.560373410767581e-05}
{"train_loss": 0.06390836834907532, "global_step": 161164, "epoch": 1810, "lr": 3.5603178922223924e-05}
{"train_loss": 0.10038554668426514, "global_step": 161165, "epoch": 1810, "lr": 3.560262373870751e-05}
{"train_loss": 0.06581971049308777, "global_step": 161166, "epoch": 1810, "lr": 3.560206855712665e-05}
{"train_loss": 0.12174694985151291, "global_step": 161167, "epoch": 1810, "lr": 3.560151337748141e-05}
{"train_loss": 0.06855833530426025, "global_step": 161168, "epoch": 1810, "lr": 3.5600958199771886e-05}
{"train_loss": 0.07490293681621552, "global_step": 161169, "epoch": 1810, "lr": 3.560040302399812e-05}
{"train_loss": 0.08190136402845383, "global_step": 161170, "epoch": 1810, "lr": 3.559984785016022e-05}
{"train_loss": 0.06792690604925156, "global_step": 161171, "epoch": 1810, "lr": 3.5599292678258225e-05}
{"train_loss": 0.09977715462446213, "global_step": 161172, "epoch": 1810, "lr": 3.559873750829226e-05}
{"train_loss": 0.08363761007785797, "global_step": 161173, "epoch": 1810, "lr": 3.5598182340262335e-05}
{"train_loss": 0.08846419304609299, "global_step": 161174, "epoch": 1810, "lr": 3.559762717416859e-05}
{"train_loss": 0.06559012830257416, "global_step": 161175, "epoch": 1810, "lr": 3.559707201001104e-05}
{"train_loss": 0.05133679881691933, "global_step": 161176, "epoch": 1810, "lr": 3.559651684778979e-05}
{"train_loss": 0.10365033894777298, "global_step": 161177, "epoch": 1810, "lr": 3.559596168750493e-05}
{"train_loss": 0.09060299982515614, "global_step": 161178, "epoch": 1810, "lr": 3.55954065291565e-05, "val_loss": 6.736195087432861, "train_action_mse_error": 5.292823791503906}
{"train_loss": 0.08464490622282028, "global_step": 161179, "epoch": 1811, "lr": 3.559485137274462e-05}
{"train_loss": 0.15637321770191193, "global_step": 161180, "epoch": 1811, "lr": 3.559429621826932e-05}
{"train_loss": 0.09196027368307114, "global_step": 161181, "epoch": 1811, "lr": 3.559374106573069e-05}
{"train_loss": 0.06644890457391739, "global_step": 161182, "epoch": 1811, "lr": 3.559318591512881e-05}
{"train_loss": 0.08400982618331909, "global_step": 161183, "epoch": 1811, "lr": 3.559263076646374e-05}
{"train_loss": 0.06759124994277954, "global_step": 161184, "epoch": 1811, "lr": 3.559207561973558e-05}
{"train_loss": 0.10173564404249191, "global_step": 161185, "epoch": 1811, "lr": 3.55915204749444e-05}
{"train_loss": 0.039530303329229355, "global_step": 161186, "epoch": 1811, "lr": 3.5590965332090234e-05}
{"train_loss": 0.15947474539279938, "global_step": 161187, "epoch": 1811, "lr": 3.5590410191173206e-05}
{"train_loss": 0.1309860199689865, "global_step": 161188, "epoch": 1811, "lr": 3.558985505219336e-05}
{"train_loss": 0.15243792533874512, "global_step": 161189, "epoch": 1811, "lr": 3.5589299915150794e-05}
{"train_loss": 0.08118969202041626, "global_step": 161190, "epoch": 1811, "lr": 3.558874478004556e-05}
{"train_loss": 0.13541150093078613, "global_step": 161191, "epoch": 1811, "lr": 3.5588189646877756e-05}
{"train_loss": 0.08468088507652283, "global_step": 161192, "epoch": 1811, "lr": 3.5587634515647425e-05}
{"train_loss": 0.0655144453048706, "global_step": 161193, "epoch": 1811, "lr": 3.558707938635466e-05}
{"train_loss": 0.053312111645936966, "global_step": 161194, "epoch": 1811, "lr": 3.558652425899956e-05}
{"train_loss": 0.07176906615495682, "global_step": 161195, "epoch": 1811, "lr": 3.558596913358216e-05}
{"train_loss": 0.16056086122989655, "global_step": 161196, "epoch": 1811, "lr": 3.558541401010257e-05}
{"train_loss": 0.11564328521490097, "global_step": 161197, "epoch": 1811, "lr": 3.5584858888560815e-05}
{"train_loss": 0.07452666759490967, "global_step": 161198, "epoch": 1811, "lr": 3.558430376895703e-05}
{"train_loss": 0.07551471143960953, "global_step": 161199, "epoch": 1811, "lr": 3.558374865129125e-05}
{"train_loss": 0.11026003211736679, "global_step": 161200, "epoch": 1811, "lr": 3.558319353556355e-05}
{"train_loss": 0.04457050934433937, "global_step": 161201, "epoch": 1811, "lr": 3.5582638421774016e-05}
{"train_loss": 0.10996758192777634, "global_step": 161202, "epoch": 1811, "lr": 3.5582083309922735e-05}
{"train_loss": 0.07189598679542542, "global_step": 161203, "epoch": 1811, "lr": 3.558152820000975e-05}
{"train_loss": 0.11222916096448898, "global_step": 161204, "epoch": 1811, "lr": 3.5580973092035174e-05}
{"train_loss": 0.0497797429561615, "global_step": 161205, "epoch": 1811, "lr": 3.558041798599903e-05}
{"train_loss": 0.047298137098550797, "global_step": 161206, "epoch": 1811, "lr": 3.557986288190145e-05}
{"train_loss": 0.08814459294080734, "global_step": 161207, "epoch": 1811, "lr": 3.557930777974247e-05}
{"train_loss": 0.07262038439512253, "global_step": 161208, "epoch": 1811, "lr": 3.5578752679522174e-05}
{"train_loss": 0.055444300174713135, "global_step": 161209, "epoch": 1811, "lr": 3.557819758124064e-05}
{"train_loss": 0.06502190977334976, "global_step": 161210, "epoch": 1811, "lr": 3.557764248489795e-05}
{"train_loss": 0.10846059024333954, "global_step": 161211, "epoch": 1811, "lr": 3.5577087390494155e-05}
{"train_loss": 0.11366085708141327, "global_step": 161212, "epoch": 1811, "lr": 3.5576532298029344e-05}
{"train_loss": 0.106332927942276, "global_step": 161213, "epoch": 1811, "lr": 3.557597720750361e-05}
{"train_loss": 0.1069650948047638, "global_step": 161214, "epoch": 1811, "lr": 3.557542211891699e-05}
{"train_loss": 0.09627600759267807, "global_step": 161215, "epoch": 1811, "lr": 3.55748670322696e-05}
{"train_loss": 0.07889734208583832, "global_step": 161216, "epoch": 1811, "lr": 3.557431194756148e-05}
{"train_loss": 0.08435861766338348, "global_step": 161217, "epoch": 1811, "lr": 3.557375686479273e-05}
{"train_loss": 0.08919372409582138, "global_step": 161218, "epoch": 1811, "lr": 3.557320178396339e-05}
{"train_loss": 0.0698758214712143, "global_step": 161219, "epoch": 1811, "lr": 3.5572646705073584e-05}
{"train_loss": 0.10183079540729523, "global_step": 161220, "epoch": 1811, "lr": 3.557209162812335e-05}
{"train_loss": 0.10159096121788025, "global_step": 161221, "epoch": 1811, "lr": 3.557153655311277e-05}
{"train_loss": 0.06906908005475998, "global_step": 161222, "epoch": 1811, "lr": 3.557098148004191e-05}
{"train_loss": 0.058570653200149536, "global_step": 161223, "epoch": 1811, "lr": 3.557042640891087e-05}
{"train_loss": 0.09431598335504532, "global_step": 161224, "epoch": 1811, "lr": 3.556987133971971e-05}
{"train_loss": 0.13476479053497314, "global_step": 161225, "epoch": 1811, "lr": 3.556931627246851e-05}
{"train_loss": 0.08040174096822739, "global_step": 161226, "epoch": 1811, "lr": 3.556876120715732e-05}
{"train_loss": 0.07900529354810715, "global_step": 161227, "epoch": 1811, "lr": 3.556820614378626e-05}
{"train_loss": 0.07662304490804672, "global_step": 161228, "epoch": 1811, "lr": 3.5567651082355354e-05}
{"train_loss": 0.07006409764289856, "global_step": 161229, "epoch": 1811, "lr": 3.556709602286472e-05}
{"train_loss": 0.06959924846887589, "global_step": 161230, "epoch": 1811, "lr": 3.55665409653144e-05}
{"train_loss": 0.10949523001909256, "global_step": 161231, "epoch": 1811, "lr": 3.556598590970447e-05}
{"train_loss": 0.06983213126659393, "global_step": 161232, "epoch": 1811, "lr": 3.556543085603505e-05}
{"train_loss": 0.09366198629140854, "global_step": 161233, "epoch": 1811, "lr": 3.556487580430615e-05}
{"train_loss": 0.10753944516181946, "global_step": 161234, "epoch": 1811, "lr": 3.556432075451791e-05}
{"train_loss": 0.045100778341293335, "global_step": 161235, "epoch": 1811, "lr": 3.556376570667034e-05}
{"train_loss": 0.13873562216758728, "global_step": 161236, "epoch": 1811, "lr": 3.556321066076358e-05}
{"train_loss": 0.07931378483772278, "global_step": 161237, "epoch": 1811, "lr": 3.5562655616797636e-05}
{"train_loss": 0.12182746082544327, "global_step": 161238, "epoch": 1811, "lr": 3.556210057477264e-05}
{"train_loss": 0.11427553743124008, "global_step": 161239, "epoch": 1811, "lr": 3.556154553468864e-05}
{"train_loss": 0.09614380449056625, "global_step": 161240, "epoch": 1811, "lr": 3.556099049654571e-05}
{"train_loss": 0.09086257219314575, "global_step": 161241, "epoch": 1811, "lr": 3.556043546034393e-05}
{"train_loss": 0.1266021579504013, "global_step": 161242, "epoch": 1811, "lr": 3.5559880426083366e-05}
{"train_loss": 0.07042104005813599, "global_step": 161243, "epoch": 1811, "lr": 3.555932539376411e-05}
{"train_loss": 0.08849966526031494, "global_step": 161244, "epoch": 1811, "lr": 3.5558770363386235e-05}
{"train_loss": 0.09843514859676361, "global_step": 161245, "epoch": 1811, "lr": 3.555821533494978e-05}
{"train_loss": 0.1310529112815857, "global_step": 161246, "epoch": 1811, "lr": 3.555766030845488e-05}
{"train_loss": 0.03361304849386215, "global_step": 161247, "epoch": 1811, "lr": 3.555710528390154e-05}
{"train_loss": 0.12234491109848022, "global_step": 161248, "epoch": 1811, "lr": 3.5556550261289885e-05}
{"train_loss": 0.12593871355056763, "global_step": 161249, "epoch": 1811, "lr": 3.5555995240619986e-05}
{"train_loss": 0.1462852954864502, "global_step": 161250, "epoch": 1811, "lr": 3.555544022189189e-05}
{"train_loss": 0.06814205646514893, "global_step": 161251, "epoch": 1811, "lr": 3.5554885205105714e-05}
{"train_loss": 0.05607551336288452, "global_step": 161252, "epoch": 1811, "lr": 3.5554330190261484e-05}
{"train_loss": 0.057095326483249664, "global_step": 161253, "epoch": 1811, "lr": 3.5553775177359324e-05}
{"train_loss": 0.056491147726774216, "global_step": 161254, "epoch": 1811, "lr": 3.555322016639926e-05}
{"train_loss": 0.09572203457355499, "global_step": 161255, "epoch": 1811, "lr": 3.5552665157381405e-05}
{"train_loss": 0.06453465670347214, "global_step": 161256, "epoch": 1811, "lr": 3.5552110150305795e-05}
{"train_loss": 0.11088179051876068, "global_step": 161257, "epoch": 1811, "lr": 3.5551555145172554e-05}
{"train_loss": 0.06464029848575592, "global_step": 161258, "epoch": 1811, "lr": 3.555100014198172e-05}
{"train_loss": 0.11813786625862122, "global_step": 161259, "epoch": 1811, "lr": 3.555044514073338e-05}
{"train_loss": 0.0835399404168129, "global_step": 161260, "epoch": 1811, "lr": 3.55498901414276e-05}
{"train_loss": 0.07608763873577118, "global_step": 161261, "epoch": 1811, "lr": 3.5549335144064466e-05}
{"train_loss": 0.08530806005001068, "global_step": 161262, "epoch": 1811, "lr": 3.554878014864405e-05}
{"train_loss": 0.10992226004600525, "global_step": 161263, "epoch": 1811, "lr": 3.554822515516643e-05}
{"train_loss": 0.10307132452726364, "global_step": 161264, "epoch": 1811, "lr": 3.554767016363166e-05}
{"train_loss": 0.12256467342376709, "global_step": 161265, "epoch": 1811, "lr": 3.5547115174039844e-05}
{"train_loss": 0.05165039002895355, "global_step": 161266, "epoch": 1811, "lr": 3.554656018639102e-05}
{"train_loss": 0.09066456434934327, "global_step": 161267, "epoch": 1811, "lr": 3.554600520068529e-05, "val_loss": 6.781318187713623}
{"train_loss": 0.09134747087955475, "global_step": 161268, "epoch": 1812, "lr": 3.554545021692274e-05}
{"train_loss": 0.1123807355761528, "global_step": 161269, "epoch": 1812, "lr": 3.554489523510341e-05}
{"train_loss": 0.07595700770616531, "global_step": 161270, "epoch": 1812, "lr": 3.554434025522741e-05}
{"train_loss": 0.08937689661979675, "global_step": 161271, "epoch": 1812, "lr": 3.5543785277294784e-05}
{"train_loss": 0.0844009593129158, "global_step": 161272, "epoch": 1812, "lr": 3.554323030130564e-05}
{"train_loss": 0.07053031772375107, "global_step": 161273, "epoch": 1812, "lr": 3.554267532726001e-05}
{"train_loss": 0.18907292187213898, "global_step": 161274, "epoch": 1812, "lr": 3.5542120355158e-05}
{"train_loss": 0.10970498621463776, "global_step": 161275, "epoch": 1812, "lr": 3.554156538499967e-05}
{"train_loss": 0.06948131322860718, "global_step": 161276, "epoch": 1812, "lr": 3.5541010416785114e-05}
{"train_loss": 0.14913585782051086, "global_step": 161277, "epoch": 1812, "lr": 3.554045545051439e-05}
{"train_loss": 0.08919335156679153, "global_step": 161278, "epoch": 1812, "lr": 3.553990048618757e-05}
{"train_loss": 0.04381173104047775, "global_step": 161279, "epoch": 1812, "lr": 3.553934552380474e-05}
{"train_loss": 0.07336550205945969, "global_step": 161280, "epoch": 1812, "lr": 3.5538790563365955e-05}
{"train_loss": 0.043549440801143646, "global_step": 161281, "epoch": 1812, "lr": 3.553823560487131e-05}
{"train_loss": 0.08091450482606888, "global_step": 161282, "epoch": 1812, "lr": 3.5537680648320884e-05}
{"train_loss": 0.06494752317667007, "global_step": 161283, "epoch": 1812, "lr": 3.553712569371472e-05}
{"train_loss": 0.065376877784729, "global_step": 161284, "epoch": 1812, "lr": 3.5536570741052916e-05}
{"train_loss": 0.0895652249455452, "global_step": 161285, "epoch": 1812, "lr": 3.5536015790335565e-05}
{"train_loss": 0.048024412244558334, "global_step": 161286, "epoch": 1812, "lr": 3.5535460841562696e-05}
{"train_loss": 0.1345096081495285, "global_step": 161287, "epoch": 1812, "lr": 3.5534905894734425e-05}
{"train_loss": 0.09996645152568817, "global_step": 161288, "epoch": 1812, "lr": 3.5534350949850793e-05}
{"train_loss": 0.05367701128125191, "global_step": 161289, "epoch": 1812, "lr": 3.5533796006911915e-05}
{"train_loss": 0.09959480166435242, "global_step": 161290, "epoch": 1812, "lr": 3.5533241065917824e-05}
{"train_loss": 0.06038612127304077, "global_step": 161291, "epoch": 1812, "lr": 3.5532686126868616e-05}
{"train_loss": 0.07933524250984192, "global_step": 161292, "epoch": 1812, "lr": 3.553213118976436e-05}
{"train_loss": 0.13823498785495758, "global_step": 161293, "epoch": 1812, "lr": 3.5531576254605146e-05}
{"train_loss": 0.08940386027097702, "global_step": 161294, "epoch": 1812, "lr": 3.553102132139101e-05}
{"train_loss": 0.11227009445428848, "global_step": 161295, "epoch": 1812, "lr": 3.553046639012208e-05}
{"train_loss": 0.08386921137571335, "global_step": 161296, "epoch": 1812, "lr": 3.5529911460798395e-05}
{"train_loss": 0.061054542660713196, "global_step": 161297, "epoch": 1812, "lr": 3.552935653342003e-05}
{"train_loss": 0.12823531031608582, "global_step": 161298, "epoch": 1812, "lr": 3.5528801607987063e-05}
{"train_loss": 0.13520081341266632, "global_step": 161299, "epoch": 1812, "lr": 3.552824668449958e-05}
{"train_loss": 0.08479437977075577, "global_step": 161300, "epoch": 1812, "lr": 3.552769176295764e-05}
{"train_loss": 0.11497464776039124, "global_step": 161301, "epoch": 1812, "lr": 3.552713684336134e-05}
{"train_loss": 0.08087878674268723, "global_step": 161302, "epoch": 1812, "lr": 3.552658192571072e-05}
{"train_loss": 0.1311560720205307, "global_step": 161303, "epoch": 1812, "lr": 3.552602701000588e-05}
{"train_loss": 0.07696662843227386, "global_step": 161304, "epoch": 1812, "lr": 3.5525472096246904e-05}
{"train_loss": 0.144846573472023, "global_step": 161305, "epoch": 1812, "lr": 3.552491718443383e-05}
{"train_loss": 0.11378989368677139, "global_step": 161306, "epoch": 1812, "lr": 3.552436227456678e-05}
{"train_loss": 0.11034850776195526, "global_step": 161307, "epoch": 1812, "lr": 3.552380736664578e-05}
{"train_loss": 0.08509234338998795, "global_step": 161308, "epoch": 1812, "lr": 3.5523252460670945e-05}
{"train_loss": 0.05264439433813095, "global_step": 161309, "epoch": 1812, "lr": 3.5522697556642314e-05}
{"train_loss": 0.08540797233581543, "global_step": 161310, "epoch": 1812, "lr": 3.552214265456e-05}
{"train_loss": 0.05341525003314018, "global_step": 161311, "epoch": 1812, "lr": 3.552158775442404e-05}
{"train_loss": 0.07646284252405167, "global_step": 161312, "epoch": 1812, "lr": 3.5521032856234546e-05}
{"train_loss": 0.06573674827814102, "global_step": 161313, "epoch": 1812, "lr": 3.552047795999155e-05}
{"train_loss": 0.10729684680700302, "global_step": 161314, "epoch": 1812, "lr": 3.551992306569517e-05}
{"train_loss": 0.10992317646741867, "global_step": 161315, "epoch": 1812, "lr": 3.5519368173345445e-05}
{"train_loss": 0.08906149864196777, "global_step": 161316, "epoch": 1812, "lr": 3.551881328294248e-05}
{"train_loss": 0.036761701107025146, "global_step": 161317, "epoch": 1812, "lr": 3.551825839448632e-05}
{"train_loss": 0.1256171017885208, "global_step": 161318, "epoch": 1812, "lr": 3.551770350797705e-05}
{"train_loss": 0.14122353494167328, "global_step": 161319, "epoch": 1812, "lr": 3.551714862341476e-05}
{"train_loss": 0.10781744122505188, "global_step": 161320, "epoch": 1812, "lr": 3.55165937407995e-05}
{"train_loss": 0.04983000457286835, "global_step": 161321, "epoch": 1812, "lr": 3.551603886013137e-05}
{"train_loss": 0.12913259863853455, "global_step": 161322, "epoch": 1812, "lr": 3.5515483981410424e-05}
{"train_loss": 0.06812410801649094, "global_step": 161323, "epoch": 1812, "lr": 3.551492910463676e-05}
{"train_loss": 0.15935280919075012, "global_step": 161324, "epoch": 1812, "lr": 3.551437422981041e-05}
{"train_loss": 0.05917724594473839, "global_step": 161325, "epoch": 1812, "lr": 3.55138193569315e-05}
{"train_loss": 0.08854767680168152, "global_step": 161326, "epoch": 1812, "lr": 3.551326448600007e-05}
{"train_loss": 0.10469477623701096, "global_step": 161327, "epoch": 1812, "lr": 3.551270961701622e-05}
{"train_loss": 0.09263558685779572, "global_step": 161328, "epoch": 1812, "lr": 3.5512154749979976e-05}
{"train_loss": 0.10821191966533661, "global_step": 161329, "epoch": 1812, "lr": 3.5511599884891476e-05}
{"train_loss": 0.10000638663768768, "global_step": 161330, "epoch": 1812, "lr": 3.5511045021750745e-05}
{"train_loss": 0.05292055755853653, "global_step": 161331, "epoch": 1812, "lr": 3.5510490160557905e-05}
{"train_loss": 0.1467241793870926, "global_step": 161332, "epoch": 1812, "lr": 3.550993530131297e-05}
{"train_loss": 0.10561927407979965, "global_step": 161333, "epoch": 1812, "lr": 3.5509380444016075e-05}
{"train_loss": 0.09827101230621338, "global_step": 161334, "epoch": 1812, "lr": 3.550882558866724e-05}
{"train_loss": 0.11433140933513641, "global_step": 161335, "epoch": 1812, "lr": 3.55082707352666e-05}
{"train_loss": 0.08291112631559372, "global_step": 161336, "epoch": 1812, "lr": 3.5507715883814174e-05}
{"train_loss": 0.1434127390384674, "global_step": 161337, "epoch": 1812, "lr": 3.5507161034310063e-05}
{"train_loss": 0.061917293816804886, "global_step": 161338, "epoch": 1812, "lr": 3.5506606186754346e-05}
{"train_loss": 0.10622070729732513, "global_step": 161339, "epoch": 1812, "lr": 3.550605134114707e-05}
{"train_loss": 0.08693339675664902, "global_step": 161340, "epoch": 1812, "lr": 3.5505496497488345e-05}
{"train_loss": 0.10777948051691055, "global_step": 161341, "epoch": 1812, "lr": 3.550494165577822e-05}
{"train_loss": 0.034846123307943344, "global_step": 161342, "epoch": 1812, "lr": 3.550438681601679e-05}
{"train_loss": 0.10574857890605927, "global_step": 161343, "epoch": 1812, "lr": 3.5503831978204106e-05}
{"train_loss": 0.08714137226343155, "global_step": 161344, "epoch": 1812, "lr": 3.550327714234026e-05}
{"train_loss": 0.16014906764030457, "global_step": 161345, "epoch": 1812, "lr": 3.550272230842532e-05}
{"train_loss": 0.06342769414186478, "global_step": 161346, "epoch": 1812, "lr": 3.550216747645938e-05}
{"train_loss": 0.06483633816242218, "global_step": 161347, "epoch": 1812, "lr": 3.5501612646442466e-05}
{"train_loss": 0.06230740621685982, "global_step": 161348, "epoch": 1812, "lr": 3.550105781837471e-05}
{"train_loss": 0.0654277577996254, "global_step": 161349, "epoch": 1812, "lr": 3.550050299225614e-05}
{"train_loss": 0.07305940240621567, "global_step": 161350, "epoch": 1812, "lr": 3.549994816808687e-05}
{"train_loss": 0.14692829549312592, "global_step": 161351, "epoch": 1812, "lr": 3.549939334586694e-05}
{"train_loss": 0.10103394091129303, "global_step": 161352, "epoch": 1812, "lr": 3.549883852559645e-05}
{"train_loss": 0.09668673574924469, "global_step": 161353, "epoch": 1812, "lr": 3.549828370727545e-05}
{"train_loss": 0.039478011429309845, "global_step": 161354, "epoch": 1812, "lr": 3.549772889090405e-05}
{"train_loss": 0.1010243147611618, "global_step": 161355, "epoch": 1812, "lr": 3.549717407648229e-05}
{"train_loss": 0.09342707547076633, "global_step": 161356, "epoch": 1812, "lr": 3.549661926401026e-05, "val_loss": 6.843028545379639}
{"train_loss": 0.12346909195184708, "global_step": 161357, "epoch": 1813, "lr": 3.549606445348803e-05}
{"train_loss": 0.05539422109723091, "global_step": 161358, "epoch": 1813, "lr": 3.549550964491567e-05}
{"train_loss": 0.09803566336631775, "global_step": 161359, "epoch": 1813, "lr": 3.549495483829329e-05}
{"train_loss": 0.04899377003312111, "global_step": 161360, "epoch": 1813, "lr": 3.54944000336209e-05}
{"train_loss": 0.06382522732019424, "global_step": 161361, "epoch": 1813, "lr": 3.5493845230898646e-05}
{"train_loss": 0.09248746931552887, "global_step": 161362, "epoch": 1813, "lr": 3.549329043012654e-05}
{"train_loss": 0.06783607602119446, "global_step": 161363, "epoch": 1813, "lr": 3.549273563130471e-05}
{"train_loss": 0.047227174043655396, "global_step": 161364, "epoch": 1813, "lr": 3.5492180834433174e-05}
{"train_loss": 0.06791073083877563, "global_step": 161365, "epoch": 1813, "lr": 3.549162603951207e-05}
{"train_loss": 0.09590748697519302, "global_step": 161366, "epoch": 1813, "lr": 3.549107124654141e-05}
{"train_loss": 0.07891162484884262, "global_step": 161367, "epoch": 1813, "lr": 3.549051645552133e-05}
{"train_loss": 0.13755567371845245, "global_step": 161368, "epoch": 1813, "lr": 3.5489961666451846e-05}
{"train_loss": 0.1256331205368042, "global_step": 161369, "epoch": 1813, "lr": 3.5489406879333074e-05}
{"train_loss": 0.09843206405639648, "global_step": 161370, "epoch": 1813, "lr": 3.548885209416506e-05}
{"train_loss": 0.06167704239487648, "global_step": 161371, "epoch": 1813, "lr": 3.548829731094792e-05}
{"train_loss": 0.05670632794499397, "global_step": 161372, "epoch": 1813, "lr": 3.548774252968167e-05}
{"train_loss": 0.16042013466358185, "global_step": 161373, "epoch": 1813, "lr": 3.548718775036645e-05}
{"train_loss": 0.04340825229883194, "global_step": 161374, "epoch": 1813, "lr": 3.548663297300228e-05}
{"train_loss": 0.08438027650117874, "global_step": 161375, "epoch": 1813, "lr": 3.5486078197589245e-05}
{"train_loss": 0.09533814340829849, "global_step": 161376, "epoch": 1813, "lr": 3.548552342412745e-05}
{"train_loss": 0.0745122954249382, "global_step": 161377, "epoch": 1813, "lr": 3.5484968652616926e-05}
{"train_loss": 0.09638461470603943, "global_step": 161378, "epoch": 1813, "lr": 3.54844138830578e-05}
{"train_loss": 0.08949348330497742, "global_step": 161379, "epoch": 1813, "lr": 3.548385911545009e-05}
{"train_loss": 0.04287843778729439, "global_step": 161380, "epoch": 1813, "lr": 3.548330434979393e-05}
{"train_loss": 0.12717914581298828, "global_step": 161381, "epoch": 1813, "lr": 3.548274958608933e-05}
{"train_loss": 0.15400780737400055, "global_step": 161382, "epoch": 1813, "lr": 3.548219482433642e-05}
{"train_loss": 0.05161149054765701, "global_step": 161383, "epoch": 1813, "lr": 3.5481640064535236e-05}
{"train_loss": 0.08519178628921509, "global_step": 161384, "epoch": 1813, "lr": 3.548108530668589e-05}
{"train_loss": 0.12904639542102814, "global_step": 161385, "epoch": 1813, "lr": 3.5480530550788414e-05}
{"train_loss": 0.10458621382713318, "global_step": 161386, "epoch": 1813, "lr": 3.547997579684292e-05}
{"train_loss": 0.11235792189836502, "global_step": 161387, "epoch": 1813, "lr": 3.547942104484945e-05}
{"train_loss": 0.0830833911895752, "global_step": 161388, "epoch": 1813, "lr": 3.5478866294808113e-05}
{"train_loss": 0.10415230691432953, "global_step": 161389, "epoch": 1813, "lr": 3.5478311546718946e-05}
{"train_loss": 0.07448362559080124, "global_step": 161390, "epoch": 1813, "lr": 3.547775680058206e-05}
{"train_loss": 0.07403364032506943, "global_step": 161391, "epoch": 1813, "lr": 3.54772020563975e-05}
{"train_loss": 0.10097179561853409, "global_step": 161392, "epoch": 1813, "lr": 3.547664731416537e-05}
{"train_loss": 0.1333564966917038, "global_step": 161393, "epoch": 1813, "lr": 3.547609257388571e-05}
{"train_loss": 0.06228012591600418, "global_step": 161394, "epoch": 1813, "lr": 3.547553783555861e-05}
{"train_loss": 0.07961845397949219, "global_step": 161395, "epoch": 1813, "lr": 3.5474983099184165e-05}
{"train_loss": 0.08348960429430008, "global_step": 161396, "epoch": 1813, "lr": 3.5474428364762416e-05}
{"train_loss": 0.1553618460893631, "global_step": 161397, "epoch": 1813, "lr": 3.5473873632293465e-05}
{"train_loss": 0.08364056795835495, "global_step": 161398, "epoch": 1813, "lr": 3.547331890177735e-05}
{"train_loss": 0.11025654524564743, "global_step": 161399, "epoch": 1813, "lr": 3.5472764173214204e-05}
{"train_loss": 0.08757113665342331, "global_step": 161400, "epoch": 1813, "lr": 3.5472209446604044e-05}
{"train_loss": 0.0891755223274231, "global_step": 161401, "epoch": 1813, "lr": 3.547165472194699e-05}
{"train_loss": 0.044426944106817245, "global_step": 161402, "epoch": 1813, "lr": 3.547109999924307e-05}
{"train_loss": 0.15267004072666168, "global_step": 161403, "epoch": 1813, "lr": 3.547054527849241e-05}
{"train_loss": 0.1563308835029602, "global_step": 161404, "epoch": 1813, "lr": 3.546999055969503e-05}
{"train_loss": 0.0614396333694458, "global_step": 161405, "epoch": 1813, "lr": 3.546943584285105e-05}
{"train_loss": 0.05569733679294586, "global_step": 161406, "epoch": 1813, "lr": 3.546888112796052e-05}
{"train_loss": 0.10872773826122284, "global_step": 161407, "epoch": 1813, "lr": 3.546832641502354e-05}
{"train_loss": 0.07349971681833267, "global_step": 161408, "epoch": 1813, "lr": 3.546777170404013e-05}
{"train_loss": 0.06165679171681404, "global_step": 161409, "epoch": 1813, "lr": 3.546721699501043e-05}
{"train_loss": 0.09255056828260422, "global_step": 161410, "epoch": 1813, "lr": 3.546666228793447e-05}
{"train_loss": 0.1185222938656807, "global_step": 161411, "epoch": 1813, "lr": 3.546610758281235e-05}
{"train_loss": 0.0883389338850975, "global_step": 161412, "epoch": 1813, "lr": 3.5465552879644146e-05}
{"train_loss": 0.09232107549905777, "global_step": 161413, "epoch": 1813, "lr": 3.54649981784299e-05}
{"train_loss": 0.11679372191429138, "global_step": 161414, "epoch": 1813, "lr": 3.5464443479169726e-05}
{"train_loss": 0.0961356833577156, "global_step": 161415, "epoch": 1813, "lr": 3.546388878186366e-05}
{"train_loss": 0.11242763698101044, "global_step": 161416, "epoch": 1813, "lr": 3.5463334086511815e-05}
{"train_loss": 0.10413073003292084, "global_step": 161417, "epoch": 1813, "lr": 3.546277939311423e-05}
{"train_loss": 0.12564301490783691, "global_step": 161418, "epoch": 1813, "lr": 3.546222470167102e-05}
{"train_loss": 0.12035781145095825, "global_step": 161419, "epoch": 1813, "lr": 3.5461670012182215e-05}
{"train_loss": 0.10330769419670105, "global_step": 161420, "epoch": 1813, "lr": 3.546111532464793e-05}
{"train_loss": 0.1332647055387497, "global_step": 161421, "epoch": 1813, "lr": 3.54605606390682e-05}
{"train_loss": 0.08689220994710922, "global_step": 161422, "epoch": 1813, "lr": 3.546000595544314e-05}
{"train_loss": 0.06679947674274445, "global_step": 161423, "epoch": 1813, "lr": 3.545945127377278e-05}
{"train_loss": 0.1286676675081253, "global_step": 161424, "epoch": 1813, "lr": 3.545889659405725e-05}
{"train_loss": 0.09867704659700394, "global_step": 161425, "epoch": 1813, "lr": 3.5458341916296564e-05}
{"train_loss": 0.07591112703084946, "global_step": 161426, "epoch": 1813, "lr": 3.545778724049086e-05}
{"train_loss": 0.1332090049982071, "global_step": 161427, "epoch": 1813, "lr": 3.545723256664015e-05}
{"train_loss": 0.10919919610023499, "global_step": 161428, "epoch": 1813, "lr": 3.5456677894744564e-05}
{"train_loss": 0.098098985850811, "global_step": 161429, "epoch": 1813, "lr": 3.545612322480412e-05}
{"train_loss": 0.10589653253555298, "global_step": 161430, "epoch": 1813, "lr": 3.545556855681894e-05}
{"train_loss": 0.144366055727005, "global_step": 161431, "epoch": 1813, "lr": 3.545501389078909e-05}
{"train_loss": 0.09965036064386368, "global_step": 161432, "epoch": 1813, "lr": 3.545445922671462e-05}
{"train_loss": 0.09241826832294464, "global_step": 161433, "epoch": 1813, "lr": 3.545390456459564e-05}
{"train_loss": 0.05386554077267647, "global_step": 161434, "epoch": 1813, "lr": 3.545334990443218e-05}
{"train_loss": 0.10089626908302307, "global_step": 161435, "epoch": 1813, "lr": 3.545279524622436e-05}
{"train_loss": 0.09326479583978653, "global_step": 161436, "epoch": 1813, "lr": 3.545224058997222e-05}
{"train_loss": 0.1291038990020752, "global_step": 161437, "epoch": 1813, "lr": 3.545168593567586e-05}
{"train_loss": 0.04946362227201462, "global_step": 161438, "epoch": 1813, "lr": 3.5451131283335335e-05}
{"train_loss": 0.027352135628461838, "global_step": 161439, "epoch": 1813, "lr": 3.545057663295074e-05}
{"train_loss": 0.10680819302797318, "global_step": 161440, "epoch": 1813, "lr": 3.545002198452212e-05}
{"train_loss": 0.0907316729426384, "global_step": 161441, "epoch": 1813, "lr": 3.544946733804959e-05}
{"train_loss": 0.09812449663877487, "global_step": 161442, "epoch": 1813, "lr": 3.5448912693533174e-05}
{"train_loss": 0.1346665322780609, "global_step": 161443, "epoch": 1813, "lr": 3.5448358050973e-05}
{"train_loss": 0.09754735231399536, "global_step": 161444, "epoch": 1813, "lr": 3.54478034103691e-05}
{"train_loss": 0.09487287503447425, "global_step": 161445, "epoch": 1813, "lr": 3.544724877172158e-05, "val_loss": 6.841782569885254}
{"train_loss": 0.07860708236694336, "global_step": 161446, "epoch": 1814, "lr": 3.544669413503048e-05}
{"train_loss": 0.10324857383966446, "global_step": 161447, "epoch": 1814, "lr": 3.544613950029592e-05}
{"train_loss": 0.07376650720834732, "global_step": 161448, "epoch": 1814, "lr": 3.5445584867517925e-05}
{"train_loss": 0.09546748548746109, "global_step": 161449, "epoch": 1814, "lr": 3.54450302366966e-05}
{"train_loss": 0.1083264872431755, "global_step": 161450, "epoch": 1814, "lr": 3.5444475607832016e-05}
{"train_loss": 0.05459006503224373, "global_step": 161451, "epoch": 1814, "lr": 3.544392098092425e-05}
{"train_loss": 0.07701723277568817, "global_step": 161452, "epoch": 1814, "lr": 3.5443366355973376e-05}
{"train_loss": 0.0655977800488472, "global_step": 161453, "epoch": 1814, "lr": 3.544281173297945e-05}
{"train_loss": 0.06053275614976883, "global_step": 161454, "epoch": 1814, "lr": 3.5442257111942573e-05}
{"train_loss": 0.08883760869503021, "global_step": 161455, "epoch": 1814, "lr": 3.544170249286279e-05}
{"train_loss": 0.1685715615749359, "global_step": 161456, "epoch": 1814, "lr": 3.544114787574021e-05}
{"train_loss": 0.05582130700349808, "global_step": 161457, "epoch": 1814, "lr": 3.5440593260574875e-05}
{"train_loss": 0.11774075031280518, "global_step": 161458, "epoch": 1814, "lr": 3.544003864736689e-05}
{"train_loss": 0.11232607066631317, "global_step": 161459, "epoch": 1814, "lr": 3.54394840361163e-05}
{"train_loss": 0.12034101784229279, "global_step": 161460, "epoch": 1814, "lr": 3.543892942682321e-05}
{"train_loss": 0.06437420099973679, "global_step": 161461, "epoch": 1814, "lr": 3.5438374819487664e-05}
{"train_loss": 0.08352866023778915, "global_step": 161462, "epoch": 1814, "lr": 3.5437820214109766e-05}
{"train_loss": 0.09267769008874893, "global_step": 161463, "epoch": 1814, "lr": 3.543726561068955e-05}
{"train_loss": 0.07918526232242584, "global_step": 161464, "epoch": 1814, "lr": 3.5436711009227146e-05}
{"train_loss": 0.08471640944480896, "global_step": 161465, "epoch": 1814, "lr": 3.5436156409722574e-05}
{"train_loss": 0.09972209483385086, "global_step": 161466, "epoch": 1814, "lr": 3.543560181217594e-05}
{"train_loss": 0.08069786429405212, "global_step": 161467, "epoch": 1814, "lr": 3.5435047216587325e-05}
{"train_loss": 0.04314495995640755, "global_step": 161468, "epoch": 1814, "lr": 3.5434492622956784e-05}
{"train_loss": 0.08223692327737808, "global_step": 161469, "epoch": 1814, "lr": 3.543393803128439e-05}
{"train_loss": 0.0556110143661499, "global_step": 161470, "epoch": 1814, "lr": 3.5433383441570234e-05}
{"train_loss": 0.05290398374199867, "global_step": 161471, "epoch": 1814, "lr": 3.543282885381439e-05}
{"train_loss": 0.08940573036670685, "global_step": 161472, "epoch": 1814, "lr": 3.54322742680169e-05}
{"train_loss": 0.08463741093873978, "global_step": 161473, "epoch": 1814, "lr": 3.543171968417789e-05}
{"train_loss": 0.10858535021543503, "global_step": 161474, "epoch": 1814, "lr": 3.543116510229738e-05}
{"train_loss": 0.049551479518413544, "global_step": 161475, "epoch": 1814, "lr": 3.54306105223755e-05}
{"train_loss": 0.04817318916320801, "global_step": 161476, "epoch": 1814, "lr": 3.5430055944412275e-05}
{"train_loss": 0.08343317359685898, "global_step": 161477, "epoch": 1814, "lr": 3.542950136840782e-05}
{"train_loss": 0.09642722457647324, "global_step": 161478, "epoch": 1814, "lr": 3.542894679436217e-05}
{"train_loss": 0.11187316477298737, "global_step": 161479, "epoch": 1814, "lr": 3.5428392222275444e-05}
{"train_loss": 0.13192912936210632, "global_step": 161480, "epoch": 1814, "lr": 3.542783765214767e-05}
{"train_loss": 0.05059286206960678, "global_step": 161481, "epoch": 1814, "lr": 3.5427283083978965e-05}
{"train_loss": 0.04637029021978378, "global_step": 161482, "epoch": 1814, "lr": 3.542672851776936e-05}
{"train_loss": 0.08470971882343292, "global_step": 161483, "epoch": 1814, "lr": 3.5426173953518984e-05}
{"train_loss": 0.056569717824459076, "global_step": 161484, "epoch": 1814, "lr": 3.5425619391227856e-05}
{"train_loss": 0.08892206102609634, "global_step": 161485, "epoch": 1814, "lr": 3.5425064830896084e-05}
{"train_loss": 0.03753453493118286, "global_step": 161486, "epoch": 1814, "lr": 3.542451027252374e-05}
{"train_loss": 0.07820986211299896, "global_step": 161487, "epoch": 1814, "lr": 3.542395571611089e-05}
{"train_loss": 0.1134888082742691, "global_step": 161488, "epoch": 1814, "lr": 3.542340116165761e-05}
{"train_loss": 0.06606052815914154, "global_step": 161489, "epoch": 1814, "lr": 3.542284660916397e-05}
{"train_loss": 0.07108984142541885, "global_step": 161490, "epoch": 1814, "lr": 3.5422292058630075e-05}
{"train_loss": 0.09971603006124496, "global_step": 161491, "epoch": 1814, "lr": 3.542173751005595e-05}
{"train_loss": 0.06062023714184761, "global_step": 161492, "epoch": 1814, "lr": 3.542118296344171e-05}
{"train_loss": 0.09410357475280762, "global_step": 161493, "epoch": 1814, "lr": 3.5420628418787395e-05}
{"train_loss": 0.12137286365032196, "global_step": 161494, "epoch": 1814, "lr": 3.542007387609312e-05}
{"train_loss": 0.09587299823760986, "global_step": 161495, "epoch": 1814, "lr": 3.541951933535892e-05}
{"train_loss": 0.12537983059883118, "global_step": 161496, "epoch": 1814, "lr": 3.541896479658491e-05}
{"train_loss": 0.06711570173501968, "global_step": 161497, "epoch": 1814, "lr": 3.541841025977112e-05}
{"train_loss": 0.10612227022647858, "global_step": 161498, "epoch": 1814, "lr": 3.5417855724917666e-05}
{"train_loss": 0.10114236921072006, "global_step": 161499, "epoch": 1814, "lr": 3.5417301192024585e-05}
{"train_loss": 0.14114484190940857, "global_step": 161500, "epoch": 1814, "lr": 3.5416746661091985e-05}
{"train_loss": 0.07604023814201355, "global_step": 161501, "epoch": 1814, "lr": 3.541619213211991e-05}
{"train_loss": 0.06653618812561035, "global_step": 161502, "epoch": 1814, "lr": 3.5415637605108453e-05}
{"train_loss": 0.07310964912176132, "global_step": 161503, "epoch": 1814, "lr": 3.54150830800577e-05}
{"train_loss": 0.09456238895654678, "global_step": 161504, "epoch": 1814, "lr": 3.541452855696771e-05}
{"train_loss": 0.1320997178554535, "global_step": 161505, "epoch": 1814, "lr": 3.5413974035838546e-05}
{"train_loss": 0.09807589650154114, "global_step": 161506, "epoch": 1814, "lr": 3.5413419516670305e-05}
{"train_loss": 0.09185733646154404, "global_step": 161507, "epoch": 1814, "lr": 3.541286499946305e-05}
{"train_loss": 0.10727791488170624, "global_step": 161508, "epoch": 1814, "lr": 3.5412310484216854e-05}
{"train_loss": 0.10558833181858063, "global_step": 161509, "epoch": 1814, "lr": 3.5411755970931806e-05}
{"train_loss": 0.11829535663127899, "global_step": 161510, "epoch": 1814, "lr": 3.541120145960796e-05}
{"train_loss": 0.08902376890182495, "global_step": 161511, "epoch": 1814, "lr": 3.5410646950245405e-05}
{"train_loss": 0.08633939921855927, "global_step": 161512, "epoch": 1814, "lr": 3.54100924428442e-05}
{"train_loss": 0.11125189065933228, "global_step": 161513, "epoch": 1814, "lr": 3.5409537937404446e-05}
{"train_loss": 0.0795515924692154, "global_step": 161514, "epoch": 1814, "lr": 3.540898343392618e-05}
{"train_loss": 0.05635737255215645, "global_step": 161515, "epoch": 1814, "lr": 3.5408428932409526e-05}
{"train_loss": 0.09483462572097778, "global_step": 161516, "epoch": 1814, "lr": 3.54078744328545e-05}
{"train_loss": 0.05604325234889984, "global_step": 161517, "epoch": 1814, "lr": 3.540731993526124e-05}
{"train_loss": 0.08786772191524506, "global_step": 161518, "epoch": 1814, "lr": 3.540676543962976e-05}
{"train_loss": 0.0714389830827713, "global_step": 161519, "epoch": 1814, "lr": 3.540621094596018e-05}
{"train_loss": 0.1266700178384781, "global_step": 161520, "epoch": 1814, "lr": 3.5405656454252547e-05}
{"train_loss": 0.09802556037902832, "global_step": 161521, "epoch": 1814, "lr": 3.540510196450694e-05}
{"train_loss": 0.08110303431749344, "global_step": 161522, "epoch": 1814, "lr": 3.540454747672346e-05}
{"train_loss": 0.08908949792385101, "global_step": 161523, "epoch": 1814, "lr": 3.5403992990902144e-05}
{"train_loss": 0.12059611827135086, "global_step": 161524, "epoch": 1814, "lr": 3.5403438507043095e-05}
{"train_loss": 0.08156256377696991, "global_step": 161525, "epoch": 1814, "lr": 3.540288402514637e-05}
{"train_loss": 0.10304634273052216, "global_step": 161526, "epoch": 1814, "lr": 3.5402329545212046e-05}
{"train_loss": 0.11316513270139694, "global_step": 161527, "epoch": 1814, "lr": 3.540177506724021e-05}
{"train_loss": 0.15185298025608063, "global_step": 161528, "epoch": 1814, "lr": 3.540122059123093e-05}
{"train_loss": 0.13053421676158905, "global_step": 161529, "epoch": 1814, "lr": 3.540066611718426e-05}
{"train_loss": 0.10325895994901657, "global_step": 161530, "epoch": 1814, "lr": 3.540011164510031e-05}
{"train_loss": 0.08253689855337143, "global_step": 161531, "epoch": 1814, "lr": 3.539955717497912e-05}
{"train_loss": 0.0670061707496643, "global_step": 161532, "epoch": 1814, "lr": 3.539900270682081e-05}
{"train_loss": 0.06801707297563553, "global_step": 161533, "epoch": 1814, "lr": 3.5398448240625385e-05}
{"train_loss": 0.0890134344023935, "global_step": 161534, "epoch": 1814, "lr": 3.5397893776393e-05, "val_loss": 6.804374694824219}
{"train_loss": 0.15709947049617767, "global_step": 161535, "epoch": 1815, "lr": 3.539733931412365e-05}
{"train_loss": 0.0666322112083435, "global_step": 161536, "epoch": 1815, "lr": 3.539678485381749e-05}
{"train_loss": 0.13705016672611237, "global_step": 161537, "epoch": 1815, "lr": 3.539623039547453e-05}
{"train_loss": 0.14016784727573395, "global_step": 161538, "epoch": 1815, "lr": 3.539567593909487e-05}
{"train_loss": 0.09356855601072311, "global_step": 161539, "epoch": 1815, "lr": 3.5395121484678595e-05}
{"train_loss": 0.12166079133749008, "global_step": 161540, "epoch": 1815, "lr": 3.5394567032225746e-05}
{"train_loss": 0.09481906145811081, "global_step": 161541, "epoch": 1815, "lr": 3.539401258173645e-05}
{"train_loss": 0.13112562894821167, "global_step": 161542, "epoch": 1815, "lr": 3.5393458133210735e-05}
{"train_loss": 0.15497195720672607, "global_step": 161543, "epoch": 1815, "lr": 3.5392903686648695e-05}
{"train_loss": 0.14064128696918488, "global_step": 161544, "epoch": 1815, "lr": 3.5392349242050396e-05}
{"train_loss": 0.10058091580867767, "global_step": 161545, "epoch": 1815, "lr": 3.5391794799415925e-05}
{"train_loss": 0.09231044352054596, "global_step": 161546, "epoch": 1815, "lr": 3.539124035874534e-05}
{"train_loss": 0.05953659489750862, "global_step": 161547, "epoch": 1815, "lr": 3.539068592003874e-05}
{"train_loss": 0.12962010502815247, "global_step": 161548, "epoch": 1815, "lr": 3.5390131483296166e-05}
{"train_loss": 0.10332835465669632, "global_step": 161549, "epoch": 1815, "lr": 3.538957704851773e-05}
{"train_loss": 0.10657045245170593, "global_step": 161550, "epoch": 1815, "lr": 3.538902261570346e-05}
{"train_loss": 0.16993774473667145, "global_step": 161551, "epoch": 1815, "lr": 3.538846818485348e-05}
{"train_loss": 0.10075213015079498, "global_step": 161552, "epoch": 1815, "lr": 3.538791375596783e-05}
{"train_loss": 0.06843308359384537, "global_step": 161553, "epoch": 1815, "lr": 3.538735932904662e-05}
{"train_loss": 0.10914973169565201, "global_step": 161554, "epoch": 1815, "lr": 3.538680490408987e-05}
{"train_loss": 0.10099111497402191, "global_step": 161555, "epoch": 1815, "lr": 3.5386250481097706e-05}
{"train_loss": 0.09838537126779556, "global_step": 161556, "epoch": 1815, "lr": 3.5385696060070165e-05}
{"train_loss": 0.0980897843837738, "global_step": 161557, "epoch": 1815, "lr": 3.538514164100735e-05}
{"train_loss": 0.11107520759105682, "global_step": 161558, "epoch": 1815, "lr": 3.538458722390933e-05}
{"train_loss": 0.0726836621761322, "global_step": 161559, "epoch": 1815, "lr": 3.5384032808776155e-05}
{"train_loss": 0.09575676918029785, "global_step": 161560, "epoch": 1815, "lr": 3.538347839560795e-05}
{"train_loss": 0.12102929502725601, "global_step": 161561, "epoch": 1815, "lr": 3.538292398440472e-05}
{"train_loss": 0.12012255191802979, "global_step": 161562, "epoch": 1815, "lr": 3.538236957516662e-05}
{"train_loss": 0.1426844745874405, "global_step": 161563, "epoch": 1815, "lr": 3.538181516789366e-05}
{"train_loss": 0.05121150612831116, "global_step": 161564, "epoch": 1815, "lr": 3.5381260762585934e-05}
{"train_loss": 0.11818455159664154, "global_step": 161565, "epoch": 1815, "lr": 3.538070635924352e-05}
{"train_loss": 0.07867039740085602, "global_step": 161566, "epoch": 1815, "lr": 3.53801519578665e-05}
{"train_loss": 0.17570845782756805, "global_step": 161567, "epoch": 1815, "lr": 3.5379597558454935e-05}
{"train_loss": 0.08334453403949738, "global_step": 161568, "epoch": 1815, "lr": 3.537904316100892e-05}
{"train_loss": 0.10415990650653839, "global_step": 161569, "epoch": 1815, "lr": 3.537848876552848e-05}
{"train_loss": 0.09203572571277618, "global_step": 161570, "epoch": 1815, "lr": 3.537793437201376e-05}
{"train_loss": 0.08614396303892136, "global_step": 161571, "epoch": 1815, "lr": 3.537737998046477e-05}
{"train_loss": 0.09055551886558533, "global_step": 161572, "epoch": 1815, "lr": 3.537682559088164e-05}
{"train_loss": 0.06306289881467819, "global_step": 161573, "epoch": 1815, "lr": 3.537627120326439e-05}
{"train_loss": 0.12048192322254181, "global_step": 161574, "epoch": 1815, "lr": 3.5375716817613146e-05}
{"train_loss": 0.07618489116430283, "global_step": 161575, "epoch": 1815, "lr": 3.5375162433927934e-05}
{"train_loss": 0.08343775570392609, "global_step": 161576, "epoch": 1815, "lr": 3.5374608052208855e-05}
{"train_loss": 0.06389693170785904, "global_step": 161577, "epoch": 1815, "lr": 3.537405367245601e-05}
{"train_loss": 0.04170124605298042, "global_step": 161578, "epoch": 1815, "lr": 3.537349929466941e-05}
{"train_loss": 0.06218661740422249, "global_step": 161579, "epoch": 1815, "lr": 3.53729449188492e-05}
{"train_loss": 0.07502881437540054, "global_step": 161580, "epoch": 1815, "lr": 3.537239054499539e-05}
{"train_loss": 0.11298149079084396, "global_step": 161581, "epoch": 1815, "lr": 3.537183617310811e-05}
{"train_loss": 0.056116070598363876, "global_step": 161582, "epoch": 1815, "lr": 3.53712818031874e-05}
{"train_loss": 0.07935018837451935, "global_step": 161583, "epoch": 1815, "lr": 3.537072743523333e-05}
{"train_loss": 0.07093080878257751, "global_step": 161584, "epoch": 1815, "lr": 3.5370173069245996e-05}
{"train_loss": 0.08123103529214859, "global_step": 161585, "epoch": 1815, "lr": 3.536961870522546e-05}
{"train_loss": 0.07786378264427185, "global_step": 161586, "epoch": 1815, "lr": 3.536906434317181e-05}
{"train_loss": 0.11349480599164963, "global_step": 161587, "epoch": 1815, "lr": 3.536850998308511e-05}
{"train_loss": 0.09588809311389923, "global_step": 161588, "epoch": 1815, "lr": 3.536795562496542e-05}
{"train_loss": 0.0684710368514061, "global_step": 161589, "epoch": 1815, "lr": 3.536740126881285e-05}
{"train_loss": 0.11999765038490295, "global_step": 161590, "epoch": 1815, "lr": 3.536684691462744e-05}
{"train_loss": 0.05657053738832474, "global_step": 161591, "epoch": 1815, "lr": 3.536629256240929e-05}
{"train_loss": 0.09937852621078491, "global_step": 161592, "epoch": 1815, "lr": 3.5365738212158446e-05}
{"train_loss": 0.05880209058523178, "global_step": 161593, "epoch": 1815, "lr": 3.5365183863875e-05}
{"train_loss": 0.0996021181344986, "global_step": 161594, "epoch": 1815, "lr": 3.536462951755905e-05}
{"train_loss": 0.098307766020298, "global_step": 161595, "epoch": 1815, "lr": 3.536407517321062e-05}
{"train_loss": 0.09684539586305618, "global_step": 161596, "epoch": 1815, "lr": 3.5363520830829847e-05}
{"train_loss": 0.09517344832420349, "global_step": 161597, "epoch": 1815, "lr": 3.5362966490416735e-05}
{"train_loss": 0.11409592628479004, "global_step": 161598, "epoch": 1815, "lr": 3.536241215197142e-05}
{"train_loss": 0.14837199449539185, "global_step": 161599, "epoch": 1815, "lr": 3.5361857815493926e-05}
{"train_loss": 0.06684049218893051, "global_step": 161600, "epoch": 1815, "lr": 3.536130348098438e-05}
{"train_loss": 0.08566519618034363, "global_step": 161601, "epoch": 1815, "lr": 3.5360749148442804e-05}
{"train_loss": 0.17126110196113586, "global_step": 161602, "epoch": 1815, "lr": 3.536019481786931e-05}
{"train_loss": 0.06988491117954254, "global_step": 161603, "epoch": 1815, "lr": 3.535964048926396e-05}
{"train_loss": 0.08571337908506393, "global_step": 161604, "epoch": 1815, "lr": 3.535908616262682e-05}
{"train_loss": 0.15286792814731598, "global_step": 161605, "epoch": 1815, "lr": 3.5358531837957974e-05}
{"train_loss": 0.08647166937589645, "global_step": 161606, "epoch": 1815, "lr": 3.5357977515257514e-05}
{"train_loss": 0.0504717193543911, "global_step": 161607, "epoch": 1815, "lr": 3.5357423194525464e-05}
{"train_loss": 0.07295572757720947, "global_step": 161608, "epoch": 1815, "lr": 3.535686887576195e-05}
{"train_loss": 0.09778673946857452, "global_step": 161609, "epoch": 1815, "lr": 3.5356314558967014e-05}
{"train_loss": 0.12773825228214264, "global_step": 161610, "epoch": 1815, "lr": 3.5355760244140755e-05}
{"train_loss": 0.1720035970211029, "global_step": 161611, "epoch": 1815, "lr": 3.5355205931283225e-05}
{"train_loss": 0.12134305387735367, "global_step": 161612, "epoch": 1815, "lr": 3.5354651620394505e-05}
{"train_loss": 0.12678906321525574, "global_step": 161613, "epoch": 1815, "lr": 3.5354097311474695e-05}
{"train_loss": 0.10345464944839478, "global_step": 161614, "epoch": 1815, "lr": 3.5353543004523824e-05}
{"train_loss": 0.09747772663831711, "global_step": 161615, "epoch": 1815, "lr": 3.535298869954201e-05}
{"train_loss": 0.13384461402893066, "global_step": 161616, "epoch": 1815, "lr": 3.535243439652929e-05}
{"train_loss": 0.08968573808670044, "global_step": 161617, "epoch": 1815, "lr": 3.535188009548578e-05}
{"train_loss": 0.04862518236041069, "global_step": 161618, "epoch": 1815, "lr": 3.5351325796411504e-05}
{"train_loss": 0.12750203907489777, "global_step": 161619, "epoch": 1815, "lr": 3.535077149930659e-05}
{"train_loss": 0.07148299366235733, "global_step": 161620, "epoch": 1815, "lr": 3.535021720417108e-05}
{"train_loss": 0.11969227343797684, "global_step": 161621, "epoch": 1815, "lr": 3.5349662911005044e-05}
{"train_loss": 0.1343579739332199, "global_step": 161622, "epoch": 1815, "lr": 3.534910861980857e-05}
{"train_loss": 0.10099359392450097, "global_step": 161623, "epoch": 1815, "lr": 3.534855433058174e-05, "val_loss": 6.806649208068848, "train_action_mse_error": 23.66380500793457}
{"train_loss": 0.1031501442193985, "global_step": 161624, "epoch": 1816, "lr": 3.534800004332461e-05}
{"train_loss": 0.08567804098129272, "global_step": 161625, "epoch": 1816, "lr": 3.534744575803727e-05}
{"train_loss": 0.08177300542593002, "global_step": 161626, "epoch": 1816, "lr": 3.534689147471977e-05}
{"train_loss": 0.14224830269813538, "global_step": 161627, "epoch": 1816, "lr": 3.5346337193372227e-05}
{"train_loss": 0.06065027415752411, "global_step": 161628, "epoch": 1816, "lr": 3.534578291399466e-05}
{"train_loss": 0.12266907095909119, "global_step": 161629, "epoch": 1816, "lr": 3.534522863658718e-05}
{"train_loss": 0.07244870066642761, "global_step": 161630, "epoch": 1816, "lr": 3.5344674361149874e-05}
{"train_loss": 0.08009817451238632, "global_step": 161631, "epoch": 1816, "lr": 3.5344120087682785e-05}
{"train_loss": 0.13514046370983124, "global_step": 161632, "epoch": 1816, "lr": 3.5343565816186016e-05}
{"train_loss": 0.05818087235093117, "global_step": 161633, "epoch": 1816, "lr": 3.53430115466596e-05}
{"train_loss": 0.08534195274114609, "global_step": 161634, "epoch": 1816, "lr": 3.534245727910366e-05}
{"train_loss": 0.09599018096923828, "global_step": 161635, "epoch": 1816, "lr": 3.534190301351823e-05}
{"train_loss": 0.14416563510894775, "global_step": 161636, "epoch": 1816, "lr": 3.5341348749903426e-05}
{"train_loss": 0.0293596014380455, "global_step": 161637, "epoch": 1816, "lr": 3.534079448825928e-05}
{"train_loss": 0.11454766243696213, "global_step": 161638, "epoch": 1816, "lr": 3.5340240228585895e-05}
{"train_loss": 0.054575853049755096, "global_step": 161639, "epoch": 1816, "lr": 3.5339685970883335e-05}
{"train_loss": 0.10623739659786224, "global_step": 161640, "epoch": 1816, "lr": 3.533913171515167e-05}
{"train_loss": 0.07129013538360596, "global_step": 161641, "epoch": 1816, "lr": 3.533857746139099e-05}
{"train_loss": 0.09850060939788818, "global_step": 161642, "epoch": 1816, "lr": 3.533802320960135e-05}
{"train_loss": 0.10270313173532486, "global_step": 161643, "epoch": 1816, "lr": 3.5337468959782835e-05}
{"train_loss": 0.0481991209089756, "global_step": 161644, "epoch": 1816, "lr": 3.533691471193553e-05}
{"train_loss": 0.09334646910429001, "global_step": 161645, "epoch": 1816, "lr": 3.5336360466059476e-05}
{"train_loss": 0.10767173767089844, "global_step": 161646, "epoch": 1816, "lr": 3.5335806222154785e-05}
{"train_loss": 0.08585049211978912, "global_step": 161647, "epoch": 1816, "lr": 3.5335251980221505e-05}
{"train_loss": 0.0705912709236145, "global_step": 161648, "epoch": 1816, "lr": 3.533469774025972e-05}
{"train_loss": 0.08036373555660248, "global_step": 161649, "epoch": 1816, "lr": 3.5334143502269527e-05}
{"train_loss": 0.05062229931354523, "global_step": 161650, "epoch": 1816, "lr": 3.5333589266250956e-05}
{"train_loss": 0.09365624934434891, "global_step": 161651, "epoch": 1816, "lr": 3.533303503220413e-05}
{"train_loss": 0.03184371069073677, "global_step": 161652, "epoch": 1816, "lr": 3.533248080012907e-05}
{"train_loss": 0.08241062611341476, "global_step": 161653, "epoch": 1816, "lr": 3.533192657002591e-05}
{"train_loss": 0.09074661135673523, "global_step": 161654, "epoch": 1816, "lr": 3.5331372341894666e-05}
{"train_loss": 0.09002940356731415, "global_step": 161655, "epoch": 1816, "lr": 3.5330818115735465e-05}
{"train_loss": 0.08124349266290665, "global_step": 161656, "epoch": 1816, "lr": 3.533026389154833e-05}
{"train_loss": 0.12110251933336258, "global_step": 161657, "epoch": 1816, "lr": 3.5329709669333397e-05}
{"train_loss": 0.12691378593444824, "global_step": 161658, "epoch": 1816, "lr": 3.532915544909068e-05}
{"train_loss": 0.18024569749832153, "global_step": 161659, "epoch": 1816, "lr": 3.532860123082029e-05}
{"train_loss": 0.07359334081411362, "global_step": 161660, "epoch": 1816, "lr": 3.532804701452228e-05}
{"train_loss": 0.06439348310232162, "global_step": 161661, "epoch": 1816, "lr": 3.532749280019674e-05}
{"train_loss": 0.1170850545167923, "global_step": 161662, "epoch": 1816, "lr": 3.5326938587843745e-05}
{"train_loss": 0.08163896948099136, "global_step": 161663, "epoch": 1816, "lr": 3.5326384377463364e-05}
{"train_loss": 0.07335958629846573, "global_step": 161664, "epoch": 1816, "lr": 3.532583016905566e-05}
{"train_loss": 0.08253225684165955, "global_step": 161665, "epoch": 1816, "lr": 3.532527596262072e-05}
{"train_loss": 0.09646579623222351, "global_step": 161666, "epoch": 1816, "lr": 3.532472175815863e-05}
{"train_loss": 0.050017230212688446, "global_step": 161667, "epoch": 1816, "lr": 3.532416755566944e-05}
{"train_loss": 0.10755476355552673, "global_step": 161668, "epoch": 1816, "lr": 3.532361335515325e-05}
{"train_loss": 0.11036788672208786, "global_step": 161669, "epoch": 1816, "lr": 3.5323059156610105e-05}
{"train_loss": 0.1048775464296341, "global_step": 161670, "epoch": 1816, "lr": 3.532250496004012e-05}
{"train_loss": 0.05847887322306633, "global_step": 161671, "epoch": 1816, "lr": 3.532195076544333e-05}
{"train_loss": 0.053205203264951706, "global_step": 161672, "epoch": 1816, "lr": 3.532139657281983e-05}
{"train_loss": 0.0526885949075222, "global_step": 161673, "epoch": 1816, "lr": 3.532084238216967e-05}
{"train_loss": 0.10673557966947556, "global_step": 161674, "epoch": 1816, "lr": 3.5320288193492976e-05}
{"train_loss": 0.11838852614164352, "global_step": 161675, "epoch": 1816, "lr": 3.531973400678976e-05}
{"train_loss": 0.057865679264068604, "global_step": 161676, "epoch": 1816, "lr": 3.531917982206015e-05}
{"train_loss": 0.0892522856593132, "global_step": 161677, "epoch": 1816, "lr": 3.5318625639304174e-05}
{"train_loss": 0.08511816710233688, "global_step": 161678, "epoch": 1816, "lr": 3.531807145852196e-05}
{"train_loss": 0.060825176537036896, "global_step": 161679, "epoch": 1816, "lr": 3.531751727971353e-05}
{"train_loss": 0.049602314829826355, "global_step": 161680, "epoch": 1816, "lr": 3.531696310287899e-05}
{"train_loss": 0.1314416080713272, "global_step": 161681, "epoch": 1816, "lr": 3.53164089280184e-05}
{"train_loss": 0.08658257871866226, "global_step": 161682, "epoch": 1816, "lr": 3.5315854755131844e-05}
{"train_loss": 0.09581868350505829, "global_step": 161683, "epoch": 1816, "lr": 3.531530058421939e-05}
{"train_loss": 0.08906269073486328, "global_step": 161684, "epoch": 1816, "lr": 3.5314746415281106e-05}
{"train_loss": 0.08348656445741653, "global_step": 161685, "epoch": 1816, "lr": 3.531419224831709e-05}
{"train_loss": 0.0907507836818695, "global_step": 161686, "epoch": 1816, "lr": 3.531363808332739e-05}
{"train_loss": 0.06774034351110458, "global_step": 161687, "epoch": 1816, "lr": 3.5313083920312105e-05}
{"train_loss": 0.08545967936515808, "global_step": 161688, "epoch": 1816, "lr": 3.5312529759271285e-05}
{"train_loss": 0.11824613809585571, "global_step": 161689, "epoch": 1816, "lr": 3.5311975600205034e-05}
{"train_loss": 0.0764356181025505, "global_step": 161690, "epoch": 1816, "lr": 3.531142144311339e-05}
{"train_loss": 0.13219332695007324, "global_step": 161691, "epoch": 1816, "lr": 3.531086728799646e-05}
{"train_loss": 0.0868673101067543, "global_step": 161692, "epoch": 1816, "lr": 3.531031313485428e-05}
{"train_loss": 0.0617162249982357, "global_step": 161693, "epoch": 1816, "lr": 3.530975898368698e-05}
{"train_loss": 0.09718450158834457, "global_step": 161694, "epoch": 1816, "lr": 3.530920483449458e-05}
{"train_loss": 0.06290961802005768, "global_step": 161695, "epoch": 1816, "lr": 3.53086506872772e-05}
{"train_loss": 0.12370367348194122, "global_step": 161696, "epoch": 1816, "lr": 3.530809654203487e-05}
{"train_loss": 0.04683784395456314, "global_step": 161697, "epoch": 1816, "lr": 3.530754239876771e-05}
{"train_loss": 0.04425017163157463, "global_step": 161698, "epoch": 1816, "lr": 3.530698825747577e-05}
{"train_loss": 0.08014792203903198, "global_step": 161699, "epoch": 1816, "lr": 3.5306434118159115e-05}
{"train_loss": 0.09907559305429459, "global_step": 161700, "epoch": 1816, "lr": 3.5305879980817837e-05}
{"train_loss": 0.12881766259670258, "global_step": 161701, "epoch": 1816, "lr": 3.5305325845451995e-05}
{"train_loss": 0.10742160677909851, "global_step": 161702, "epoch": 1816, "lr": 3.530477171206168e-05}
{"train_loss": 0.09103319048881531, "global_step": 161703, "epoch": 1816, "lr": 3.5304217580646947e-05}
{"train_loss": 0.07922465354204178, "global_step": 161704, "epoch": 1816, "lr": 3.530366345120791e-05}
{"train_loss": 0.13054011762142181, "global_step": 161705, "epoch": 1816, "lr": 3.530310932374459e-05}
{"train_loss": 0.08228866755962372, "global_step": 161706, "epoch": 1816, "lr": 3.53025551982571e-05}
{"train_loss": 0.0795140340924263, "global_step": 161707, "epoch": 1816, "lr": 3.530200107474549e-05}
{"train_loss": 0.06673499941825867, "global_step": 161708, "epoch": 1816, "lr": 3.530144695320987e-05}
{"train_loss": 0.15029174089431763, "global_step": 161709, "epoch": 1816, "lr": 3.530089283365027e-05}
{"train_loss": 0.050978630781173706, "global_step": 161710, "epoch": 1816, "lr": 3.53003387160668e-05}
{"train_loss": 0.11695265024900436, "global_step": 161711, "epoch": 1816, "lr": 3.5299784600459515e-05}
{"train_loss": 0.08967226182811716, "global_step": 161712, "epoch": 1816, "lr": 3.5299230486828505e-05, "val_loss": 6.8897318840026855}
{"train_loss": 0.0782313123345375, "global_step": 161713, "epoch": 1817, "lr": 3.529867637517381e-05}
{"train_loss": 0.0574444904923439, "global_step": 161714, "epoch": 1817, "lr": 3.529812226549555e-05}
{"train_loss": 0.10371779650449753, "global_step": 161715, "epoch": 1817, "lr": 3.5297568157793766e-05}
{"train_loss": 0.05375413969159126, "global_step": 161716, "epoch": 1817, "lr": 3.5297014052068555e-05}
{"train_loss": 0.10220880061388016, "global_step": 161717, "epoch": 1817, "lr": 3.529645994831998e-05}
{"train_loss": 0.05696482211351395, "global_step": 161718, "epoch": 1817, "lr": 3.5295905846548115e-05}
{"train_loss": 0.057510849088430405, "global_step": 161719, "epoch": 1817, "lr": 3.5295351746753035e-05}
{"train_loss": 0.10905465483665466, "global_step": 161720, "epoch": 1817, "lr": 3.52947976489348e-05}
{"train_loss": 0.11038454622030258, "global_step": 161721, "epoch": 1817, "lr": 3.529424355309352e-05}
{"train_loss": 0.10972531139850616, "global_step": 161722, "epoch": 1817, "lr": 3.529368945922924e-05}
{"train_loss": 0.10355071723461151, "global_step": 161723, "epoch": 1817, "lr": 3.529313536734205e-05}
{"train_loss": 0.02387082949280739, "global_step": 161724, "epoch": 1817, "lr": 3.529258127743201e-05}
{"train_loss": 0.0639960765838623, "global_step": 161725, "epoch": 1817, "lr": 3.529202718949923e-05}
{"train_loss": 0.11659950762987137, "global_step": 161726, "epoch": 1817, "lr": 3.5291473103543725e-05}
{"train_loss": 0.04997781664133072, "global_step": 161727, "epoch": 1817, "lr": 3.5290919019565624e-05}
{"train_loss": 0.0665396973490715, "global_step": 161728, "epoch": 1817, "lr": 3.5290364937564965e-05}
{"train_loss": 0.1435501128435135, "global_step": 161729, "epoch": 1817, "lr": 3.528981085754185e-05}
{"train_loss": 0.1257794350385666, "global_step": 161730, "epoch": 1817, "lr": 3.528925677949633e-05}
{"train_loss": 0.08887467533349991, "global_step": 161731, "epoch": 1817, "lr": 3.52887027034285e-05}
{"train_loss": 0.162422314286232, "global_step": 161732, "epoch": 1817, "lr": 3.5288148629338415e-05}
{"train_loss": 0.10736899822950363, "global_step": 161733, "epoch": 1817, "lr": 3.528759455722617e-05}
{"train_loss": 0.1099817082285881, "global_step": 161734, "epoch": 1817, "lr": 3.5287040487091804e-05}
{"train_loss": 0.1147616058588028, "global_step": 161735, "epoch": 1817, "lr": 3.528648641893545e-05}
{"train_loss": 0.04388601332902908, "global_step": 161736, "epoch": 1817, "lr": 3.528593235275713e-05}
{"train_loss": 0.10787802189588547, "global_step": 161737, "epoch": 1817, "lr": 3.528537828855694e-05}
{"train_loss": 0.09251455217599869, "global_step": 161738, "epoch": 1817, "lr": 3.528482422633495e-05}
{"train_loss": 0.11814308166503906, "global_step": 161739, "epoch": 1817, "lr": 3.528427016609123e-05}
{"train_loss": 0.07017101347446442, "global_step": 161740, "epoch": 1817, "lr": 3.528371610782588e-05}
{"train_loss": 0.059857890009880066, "global_step": 161741, "epoch": 1817, "lr": 3.528316205153892e-05}
{"train_loss": 0.06158232316374779, "global_step": 161742, "epoch": 1817, "lr": 3.52826079972305e-05}
{"train_loss": 0.08743815124034882, "global_step": 161743, "epoch": 1817, "lr": 3.528205394490063e-05}
{"train_loss": 0.06487134844064713, "global_step": 161744, "epoch": 1817, "lr": 3.528149989454942e-05}
{"train_loss": 0.0955272987484932, "global_step": 161745, "epoch": 1817, "lr": 3.5280945846176915e-05}
{"train_loss": 0.0828709825873375, "global_step": 161746, "epoch": 1817, "lr": 3.528039179978323e-05}
{"train_loss": 0.08882755041122437, "global_step": 161747, "epoch": 1817, "lr": 3.527983775536839e-05}
{"train_loss": 0.048720408231019974, "global_step": 161748, "epoch": 1817, "lr": 3.5279283712932534e-05}
{"train_loss": 0.06354977190494537, "global_step": 161749, "epoch": 1817, "lr": 3.527872967247566e-05}
{"train_loss": 0.10181665420532227, "global_step": 161750, "epoch": 1817, "lr": 3.527817563399791e-05}
{"train_loss": 0.08904891461133957, "global_step": 161751, "epoch": 1817, "lr": 3.527762159749931e-05}
{"train_loss": 0.11102643609046936, "global_step": 161752, "epoch": 1817, "lr": 3.527706756297998e-05}
{"train_loss": 0.10190337151288986, "global_step": 161753, "epoch": 1817, "lr": 3.527651353043994e-05}
{"train_loss": 0.10492301732301712, "global_step": 161754, "epoch": 1817, "lr": 3.527595949987931e-05}
{"train_loss": 0.04650932550430298, "global_step": 161755, "epoch": 1817, "lr": 3.527540547129814e-05}
{"train_loss": 0.09792386740446091, "global_step": 161756, "epoch": 1817, "lr": 3.527485144469651e-05}
{"train_loss": 0.06668310612440109, "global_step": 161757, "epoch": 1817, "lr": 3.527429742007451e-05}
{"train_loss": 0.07421193271875381, "global_step": 161758, "epoch": 1817, "lr": 3.5273743397432194e-05}
{"train_loss": 0.1388763189315796, "global_step": 161759, "epoch": 1817, "lr": 3.527318937676964e-05}
{"train_loss": 0.07020880281925201, "global_step": 161760, "epoch": 1817, "lr": 3.5272635358086927e-05}
{"train_loss": 0.09489977359771729, "global_step": 161761, "epoch": 1817, "lr": 3.527208134138414e-05}
{"train_loss": 0.1424313336610794, "global_step": 161762, "epoch": 1817, "lr": 3.527152732666132e-05}
{"train_loss": 0.08057712018489838, "global_step": 161763, "epoch": 1817, "lr": 3.527097331391859e-05}
{"train_loss": 0.07826192677021027, "global_step": 161764, "epoch": 1817, "lr": 3.527041930315598e-05}
{"train_loss": 0.07762167602777481, "global_step": 161765, "epoch": 1817, "lr": 3.526986529437359e-05}
{"train_loss": 0.06501387059688568, "global_step": 161766, "epoch": 1817, "lr": 3.5269311287571475e-05}
{"train_loss": 0.10324589908123016, "global_step": 161767, "epoch": 1817, "lr": 3.526875728274974e-05}
{"train_loss": 0.06475061923265457, "global_step": 161768, "epoch": 1817, "lr": 3.526820327990843e-05}
{"train_loss": 0.1505117118358612, "global_step": 161769, "epoch": 1817, "lr": 3.526764927904764e-05}
{"train_loss": 0.08899584412574768, "global_step": 161770, "epoch": 1817, "lr": 3.5267095280167416e-05}
{"train_loss": 0.06906519830226898, "global_step": 161771, "epoch": 1817, "lr": 3.526654128326787e-05}
{"train_loss": 0.09362845867872238, "global_step": 161772, "epoch": 1817, "lr": 3.5265987288349045e-05}
{"train_loss": 0.15662914514541626, "global_step": 161773, "epoch": 1817, "lr": 3.526543329541104e-05}
{"train_loss": 0.13132211565971375, "global_step": 161774, "epoch": 1817, "lr": 3.5264879304453915e-05}
{"train_loss": 0.09427233785390854, "global_step": 161775, "epoch": 1817, "lr": 3.5264325315477735e-05}
{"train_loss": 0.11956766247749329, "global_step": 161776, "epoch": 1817, "lr": 3.526377132848261e-05}
{"train_loss": 0.09190411120653152, "global_step": 161777, "epoch": 1817, "lr": 3.526321734346857e-05}
{"train_loss": 0.10008113831281662, "global_step": 161778, "epoch": 1817, "lr": 3.5262663360435724e-05}
{"train_loss": 0.06366223841905594, "global_step": 161779, "epoch": 1817, "lr": 3.5262109379384126e-05}
{"train_loss": 0.08166676014661789, "global_step": 161780, "epoch": 1817, "lr": 3.526155540031386e-05}
{"train_loss": 0.08131136000156403, "global_step": 161781, "epoch": 1817, "lr": 3.5261001423224995e-05}
{"train_loss": 0.07811617106199265, "global_step": 161782, "epoch": 1817, "lr": 3.526044744811762e-05}
{"train_loss": 0.052547287195920944, "global_step": 161783, "epoch": 1817, "lr": 3.5259893474991776e-05}
{"train_loss": 0.07893608510494232, "global_step": 161784, "epoch": 1817, "lr": 3.525933950384759e-05}
{"train_loss": 0.11142273247241974, "global_step": 161785, "epoch": 1817, "lr": 3.5258785534685077e-05}
{"train_loss": 0.10909238457679749, "global_step": 161786, "epoch": 1817, "lr": 3.525823156750436e-05}
{"train_loss": 0.053244203329086304, "global_step": 161787, "epoch": 1817, "lr": 3.525767760230548e-05}
{"train_loss": 0.07683101296424866, "global_step": 161788, "epoch": 1817, "lr": 3.525712363908854e-05}
{"train_loss": 0.06238133832812309, "global_step": 161789, "epoch": 1817, "lr": 3.5256569677853585e-05}
{"train_loss": 0.12717527151107788, "global_step": 161790, "epoch": 1817, "lr": 3.525601571860072e-05}
{"train_loss": 0.06263555586338043, "global_step": 161791, "epoch": 1817, "lr": 3.5255461761329986e-05}
{"train_loss": 0.15359048545360565, "global_step": 161792, "epoch": 1817, "lr": 3.525490780604149e-05}
{"train_loss": 0.08607146143913269, "global_step": 161793, "epoch": 1817, "lr": 3.525435385273529e-05}
{"train_loss": 0.14164499938488007, "global_step": 161794, "epoch": 1817, "lr": 3.5253799901411455e-05}
{"train_loss": 0.08183545619249344, "global_step": 161795, "epoch": 1817, "lr": 3.5253245952070094e-05}
{"train_loss": 0.06600035727024078, "global_step": 161796, "epoch": 1817, "lr": 3.525269200471122e-05}
{"train_loss": 0.12148693203926086, "global_step": 161797, "epoch": 1817, "lr": 3.525213805933496e-05}
{"train_loss": 0.08252312988042831, "global_step": 161798, "epoch": 1817, "lr": 3.525158411594136e-05}
{"train_loss": 0.10176260024309158, "global_step": 161799, "epoch": 1817, "lr": 3.5251030174530516e-05}
{"train_loss": 0.10381148010492325, "global_step": 161800, "epoch": 1817, "lr": 3.525047623510247e-05}
{"train_loss": 0.09123879156253312, "global_step": 161801, "epoch": 1817, "lr": 3.524992229765734e-05, "val_loss": 6.8148980140686035}
{"train_loss": 0.07067540287971497, "global_step": 161802, "epoch": 1818, "lr": 3.5249368362195165e-05}
{"train_loss": 0.048835836350917816, "global_step": 161803, "epoch": 1818, "lr": 3.524881442871605e-05}
{"train_loss": 0.05713236331939697, "global_step": 161804, "epoch": 1818, "lr": 3.524826049722003e-05}
{"train_loss": 0.041641511023044586, "global_step": 161805, "epoch": 1818, "lr": 3.5247706567707225e-05}
{"train_loss": 0.12583215534687042, "global_step": 161806, "epoch": 1818, "lr": 3.524715264017766e-05}
{"train_loss": 0.09540954977273941, "global_step": 161807, "epoch": 1818, "lr": 3.524659871463146e-05}
{"train_loss": 0.0752064660191536, "global_step": 161808, "epoch": 1818, "lr": 3.5246044791068655e-05}
{"train_loss": 0.09951002895832062, "global_step": 161809, "epoch": 1818, "lr": 3.5245490869489355e-05}
{"train_loss": 0.06771630793809891, "global_step": 161810, "epoch": 1818, "lr": 3.52449369498936e-05}
{"train_loss": 0.06601881980895996, "global_step": 161811, "epoch": 1818, "lr": 3.5244383032281494e-05}
{"train_loss": 0.19826740026474, "global_step": 161812, "epoch": 1818, "lr": 3.52438291166531e-05}
{"train_loss": 0.14568336308002472, "global_step": 161813, "epoch": 1818, "lr": 3.5243275203008495e-05}
{"train_loss": 0.07978265732526779, "global_step": 161814, "epoch": 1818, "lr": 3.524272129134777e-05}
{"train_loss": 0.10116739571094513, "global_step": 161815, "epoch": 1818, "lr": 3.5242167381670955e-05}
{"train_loss": 0.0790524035692215, "global_step": 161816, "epoch": 1818, "lr": 3.5241613473978164e-05}
{"train_loss": 0.07133671641349792, "global_step": 161817, "epoch": 1818, "lr": 3.524105956826944e-05}
{"train_loss": 0.08509119600057602, "global_step": 161818, "epoch": 1818, "lr": 3.5240505664544906e-05}
{"train_loss": 0.07293295860290527, "global_step": 161819, "epoch": 1818, "lr": 3.523995176280458e-05}
{"train_loss": 0.10023811459541321, "global_step": 161820, "epoch": 1818, "lr": 3.523939786304858e-05}
{"train_loss": 0.0873132050037384, "global_step": 161821, "epoch": 1818, "lr": 3.523884396527695e-05}
{"train_loss": 0.10196530818939209, "global_step": 161822, "epoch": 1818, "lr": 3.523829006948979e-05}
{"train_loss": 0.10467293858528137, "global_step": 161823, "epoch": 1818, "lr": 3.5237736175687144e-05}
{"train_loss": 0.14082388579845428, "global_step": 161824, "epoch": 1818, "lr": 3.523718228386912e-05}
{"train_loss": 0.07873599976301193, "global_step": 161825, "epoch": 1818, "lr": 3.523662839403577e-05}
{"train_loss": 0.10198026895523071, "global_step": 161826, "epoch": 1818, "lr": 3.5236074506187176e-05}
{"train_loss": 0.11377614736557007, "global_step": 161827, "epoch": 1818, "lr": 3.5235520620323406e-05}
{"train_loss": 0.13655604422092438, "global_step": 161828, "epoch": 1818, "lr": 3.523496673644456e-05}
{"train_loss": 0.1087745949625969, "global_step": 161829, "epoch": 1818, "lr": 3.523441285455067e-05}
{"train_loss": 0.09298630803823471, "global_step": 161830, "epoch": 1818, "lr": 3.523385897464184e-05}
{"train_loss": 0.09709372371435165, "global_step": 161831, "epoch": 1818, "lr": 3.523330509671814e-05}
{"train_loss": 0.09019329398870468, "global_step": 161832, "epoch": 1818, "lr": 3.5232751220779644e-05}
{"train_loss": 0.04602571204304695, "global_step": 161833, "epoch": 1818, "lr": 3.523219734682643e-05}
{"train_loss": 0.04390058293938637, "global_step": 161834, "epoch": 1818, "lr": 3.523164347485856e-05}
{"train_loss": 0.16994726657867432, "global_step": 161835, "epoch": 1818, "lr": 3.523108960487612e-05}
{"train_loss": 0.16406503319740295, "global_step": 161836, "epoch": 1818, "lr": 3.523053573687917e-05}
{"train_loss": 0.05156417191028595, "global_step": 161837, "epoch": 1818, "lr": 3.522998187086781e-05}
{"train_loss": 0.07711879909038544, "global_step": 161838, "epoch": 1818, "lr": 3.522942800684208e-05}
{"train_loss": 0.15603435039520264, "global_step": 161839, "epoch": 1818, "lr": 3.522887414480209e-05}
{"train_loss": 0.05920588970184326, "global_step": 161840, "epoch": 1818, "lr": 3.522832028474788e-05}
{"train_loss": 0.08737551420927048, "global_step": 161841, "epoch": 1818, "lr": 3.522776642667957e-05}
{"train_loss": 0.05939445272088051, "global_step": 161842, "epoch": 1818, "lr": 3.522721257059717e-05}
{"train_loss": 0.08702191710472107, "global_step": 161843, "epoch": 1818, "lr": 3.5226658716500824e-05}
{"train_loss": 0.10182029008865356, "global_step": 161844, "epoch": 1818, "lr": 3.522610486439055e-05}
{"train_loss": 0.12020401656627655, "global_step": 161845, "epoch": 1818, "lr": 3.5225551014266464e-05}
{"train_loss": 0.0965922474861145, "global_step": 161846, "epoch": 1818, "lr": 3.522499716612861e-05}
{"train_loss": 0.10450799763202667, "global_step": 161847, "epoch": 1818, "lr": 3.522444331997708e-05}
{"train_loss": 0.13994336128234863, "global_step": 161848, "epoch": 1818, "lr": 3.5223889475811945e-05}
{"train_loss": 0.08093095570802689, "global_step": 161849, "epoch": 1818, "lr": 3.522333563363328e-05}
{"train_loss": 0.1008404791355133, "global_step": 161850, "epoch": 1818, "lr": 3.522278179344115e-05}
{"train_loss": 0.08000703901052475, "global_step": 161851, "epoch": 1818, "lr": 3.522222795523564e-05}
{"train_loss": 0.11678041517734528, "global_step": 161852, "epoch": 1818, "lr": 3.5221674119016834e-05}
{"train_loss": 0.0671398937702179, "global_step": 161853, "epoch": 1818, "lr": 3.522112028478477e-05}
{"train_loss": 0.0774778500199318, "global_step": 161854, "epoch": 1818, "lr": 3.522056645253957e-05}
{"train_loss": 0.042757127434015274, "global_step": 161855, "epoch": 1818, "lr": 3.522001262228126e-05}
{"train_loss": 0.1006006971001625, "global_step": 161856, "epoch": 1818, "lr": 3.521945879400997e-05}
{"train_loss": 0.08107557892799377, "global_step": 161857, "epoch": 1818, "lr": 3.5218904967725717e-05}
{"train_loss": 0.08967011421918869, "global_step": 161858, "epoch": 1818, "lr": 3.521835114342862e-05}
{"train_loss": 0.1176256388425827, "global_step": 161859, "epoch": 1818, "lr": 3.521779732111872e-05}
{"train_loss": 0.05984221398830414, "global_step": 161860, "epoch": 1818, "lr": 3.521724350079612e-05}
{"train_loss": 0.08791317790746689, "global_step": 161861, "epoch": 1818, "lr": 3.5216689682460864e-05}
{"train_loss": 0.1425684243440628, "global_step": 161862, "epoch": 1818, "lr": 3.521613586611307e-05}
{"train_loss": 0.10468848049640656, "global_step": 161863, "epoch": 1818, "lr": 3.521558205175276e-05}
{"train_loss": 0.09524037688970566, "global_step": 161864, "epoch": 1818, "lr": 3.521502823938005e-05}
{"train_loss": 0.14831046760082245, "global_step": 161865, "epoch": 1818, "lr": 3.521447442899499e-05}
{"train_loss": 0.09036678075790405, "global_step": 161866, "epoch": 1818, "lr": 3.521392062059767e-05}
{"train_loss": 0.11020714044570923, "global_step": 161867, "epoch": 1818, "lr": 3.521336681418815e-05}
{"train_loss": 0.0551302395761013, "global_step": 161868, "epoch": 1818, "lr": 3.521281300976652e-05}
{"train_loss": 0.12891049683094025, "global_step": 161869, "epoch": 1818, "lr": 3.521225920733285e-05}
{"train_loss": 0.04833675175905228, "global_step": 161870, "epoch": 1818, "lr": 3.5211705406887205e-05}
{"train_loss": 0.13439339399337769, "global_step": 161871, "epoch": 1818, "lr": 3.521115160842968e-05}
{"train_loss": 0.056748729199171066, "global_step": 161872, "epoch": 1818, "lr": 3.5210597811960325e-05}
{"train_loss": 0.15791438519954681, "global_step": 161873, "epoch": 1818, "lr": 3.5210044017479227e-05}
{"train_loss": 0.06018766015768051, "global_step": 161874, "epoch": 1818, "lr": 3.520949022498645e-05}
{"train_loss": 0.08287709951400757, "global_step": 161875, "epoch": 1818, "lr": 3.520893643448209e-05}
{"train_loss": 0.1316811591386795, "global_step": 161876, "epoch": 1818, "lr": 3.5208382645966195e-05}
{"train_loss": 0.09740006923675537, "global_step": 161877, "epoch": 1818, "lr": 3.5207828859438865e-05}
{"train_loss": 0.09769083559513092, "global_step": 161878, "epoch": 1818, "lr": 3.520727507490015e-05}
{"train_loss": 0.09541123360395432, "global_step": 161879, "epoch": 1818, "lr": 3.520672129235015e-05}
{"train_loss": 0.08514803647994995, "global_step": 161880, "epoch": 1818, "lr": 3.5206167511788906e-05}
{"train_loss": 0.06179321929812431, "global_step": 161881, "epoch": 1818, "lr": 3.5205613733216537e-05}
{"train_loss": 0.08019233494997025, "global_step": 161882, "epoch": 1818, "lr": 3.520505995663307e-05}
{"train_loss": 0.14270742237567902, "global_step": 161883, "epoch": 1818, "lr": 3.5204506182038596e-05}
{"train_loss": 0.10229908674955368, "global_step": 161884, "epoch": 1818, "lr": 3.520395240943323e-05}
{"train_loss": 0.09487749636173248, "global_step": 161885, "epoch": 1818, "lr": 3.5203398638817e-05}
{"train_loss": 0.053583115339279175, "global_step": 161886, "epoch": 1818, "lr": 3.520284487018998e-05}
{"train_loss": 0.10643787682056427, "global_step": 161887, "epoch": 1818, "lr": 3.520229110355227e-05}
{"train_loss": 0.06087072566151619, "global_step": 161888, "epoch": 1818, "lr": 3.520173733890392e-05}
{"train_loss": 0.07489894330501556, "global_step": 161889, "epoch": 1818, "lr": 3.520118357624502e-05}
{"train_loss": 0.09452475381366322, "global_step": 161890, "epoch": 1818, "lr": 3.520062981557566e-05, "val_loss": 6.860655784606934}
{"train_loss": 0.08255290240049362, "global_step": 161891, "epoch": 1819, "lr": 3.520007605689587e-05}
{"train_loss": 0.08655346930027008, "global_step": 161892, "epoch": 1819, "lr": 3.519952230020577e-05}
{"train_loss": 0.10076773166656494, "global_step": 161893, "epoch": 1819, "lr": 3.5198968545505385e-05}
{"train_loss": 0.1120409443974495, "global_step": 161894, "epoch": 1819, "lr": 3.519841479279485e-05}
{"train_loss": 0.10710697621107101, "global_step": 161895, "epoch": 1819, "lr": 3.519786104207418e-05}
{"train_loss": 0.06536053121089935, "global_step": 161896, "epoch": 1819, "lr": 3.51973072933435e-05}
{"train_loss": 0.11620509624481201, "global_step": 161897, "epoch": 1819, "lr": 3.519675354660284e-05}
{"train_loss": 0.06854419410228729, "global_step": 161898, "epoch": 1819, "lr": 3.5196199801852316e-05}
{"train_loss": 0.06298661231994629, "global_step": 161899, "epoch": 1819, "lr": 3.5195646059091964e-05}
{"train_loss": 0.11134929955005646, "global_step": 161900, "epoch": 1819, "lr": 3.5195092318321896e-05}
{"train_loss": 0.06903278827667236, "global_step": 161901, "epoch": 1819, "lr": 3.5194538579542145e-05}
{"train_loss": 0.07735510915517807, "global_step": 161902, "epoch": 1819, "lr": 3.5193984842752814e-05}
{"train_loss": 0.08937722444534302, "global_step": 161903, "epoch": 1819, "lr": 3.519343110795399e-05}
{"train_loss": 0.08194959163665771, "global_step": 161904, "epoch": 1819, "lr": 3.519287737514572e-05}
{"train_loss": 0.09228118509054184, "global_step": 161905, "epoch": 1819, "lr": 3.519232364432808e-05}
{"train_loss": 0.10874460637569427, "global_step": 161906, "epoch": 1819, "lr": 3.5191769915501154e-05}
{"train_loss": 0.12359356880187988, "global_step": 161907, "epoch": 1819, "lr": 3.519121618866502e-05}
{"train_loss": 0.10291071981191635, "global_step": 161908, "epoch": 1819, "lr": 3.519066246381974e-05}
{"train_loss": 0.11190033704042435, "global_step": 161909, "epoch": 1819, "lr": 3.51901087409654e-05}
{"train_loss": 0.11175655573606491, "global_step": 161910, "epoch": 1819, "lr": 3.518955502010206e-05}
{"train_loss": 0.06503263115882874, "global_step": 161911, "epoch": 1819, "lr": 3.5189001301229826e-05}
{"train_loss": 0.11418119817972183, "global_step": 161912, "epoch": 1819, "lr": 3.518844758434872e-05}
{"train_loss": 0.07015371322631836, "global_step": 161913, "epoch": 1819, "lr": 3.518789386945888e-05}
{"train_loss": 0.24646885693073273, "global_step": 161914, "epoch": 1819, "lr": 3.518734015656031e-05}
{"train_loss": 0.0925104022026062, "global_step": 161915, "epoch": 1819, "lr": 3.518678644565315e-05}
{"train_loss": 0.10243301838636398, "global_step": 161916, "epoch": 1819, "lr": 3.5186232736737426e-05}
{"train_loss": 0.05179750546813011, "global_step": 161917, "epoch": 1819, "lr": 3.5185679029813256e-05}
{"train_loss": 0.12465666234493256, "global_step": 161918, "epoch": 1819, "lr": 3.5185125324880664e-05}
{"train_loss": 0.09200853854417801, "global_step": 161919, "epoch": 1819, "lr": 3.518457162193976e-05}
{"train_loss": 0.06868939101696014, "global_step": 161920, "epoch": 1819, "lr": 3.518401792099062e-05}
{"train_loss": 0.06255662441253662, "global_step": 161921, "epoch": 1819, "lr": 3.518346422203329e-05}
{"train_loss": 0.11875371634960175, "global_step": 161922, "epoch": 1819, "lr": 3.518291052506789e-05}
{"train_loss": 0.10149664431810379, "global_step": 161923, "epoch": 1819, "lr": 3.518235683009444e-05}
{"train_loss": 0.06722178310155869, "global_step": 161924, "epoch": 1819, "lr": 3.518180313711307e-05}
{"train_loss": 0.09079767018556595, "global_step": 161925, "epoch": 1819, "lr": 3.5181249446123805e-05}
{"train_loss": 0.14636293053627014, "global_step": 161926, "epoch": 1819, "lr": 3.518069575712675e-05}
{"train_loss": 0.10850943624973297, "global_step": 161927, "epoch": 1819, "lr": 3.5180142070121955e-05}
{"train_loss": 0.08974575996398926, "global_step": 161928, "epoch": 1819, "lr": 3.5179588385109525e-05}
{"train_loss": 0.1328253298997879, "global_step": 161929, "epoch": 1819, "lr": 3.517903470208951e-05}
{"train_loss": 0.06097292900085449, "global_step": 161930, "epoch": 1819, "lr": 3.517848102106201e-05}
{"train_loss": 0.14796534180641174, "global_step": 161931, "epoch": 1819, "lr": 3.517792734202706e-05}
{"train_loss": 0.055736057460308075, "global_step": 161932, "epoch": 1819, "lr": 3.5177373664984776e-05}
{"train_loss": 0.10114911198616028, "global_step": 161933, "epoch": 1819, "lr": 3.5176819989935195e-05}
{"train_loss": 0.07501530647277832, "global_step": 161934, "epoch": 1819, "lr": 3.5176266316878425e-05}
{"train_loss": 0.12711146473884583, "global_step": 161935, "epoch": 1819, "lr": 3.5175712645814506e-05}
{"train_loss": 0.09400749206542969, "global_step": 161936, "epoch": 1819, "lr": 3.5175158976743555e-05}
{"train_loss": 0.1540566086769104, "global_step": 161937, "epoch": 1819, "lr": 3.5174605309665604e-05}
{"train_loss": 0.06805074214935303, "global_step": 161938, "epoch": 1819, "lr": 3.517405164458074e-05}
{"train_loss": 0.1365245133638382, "global_step": 161939, "epoch": 1819, "lr": 3.517349798148907e-05}
{"train_loss": 0.0895802229642868, "global_step": 161940, "epoch": 1819, "lr": 3.517294432039062e-05}
{"train_loss": 0.08746202290058136, "global_step": 161941, "epoch": 1819, "lr": 3.5172390661285515e-05}
{"train_loss": 0.11492083221673965, "global_step": 161942, "epoch": 1819, "lr": 3.517183700417377e-05}
{"train_loss": 0.10858599841594696, "global_step": 161943, "epoch": 1819, "lr": 3.5171283349055515e-05}
{"train_loss": 0.0722481980919838, "global_step": 161944, "epoch": 1819, "lr": 3.517072969593079e-05}
{"train_loss": 0.05636720731854439, "global_step": 161945, "epoch": 1819, "lr": 3.5170176044799686e-05}
{"train_loss": 0.0869080126285553, "global_step": 161946, "epoch": 1819, "lr": 3.516962239566226e-05}
{"train_loss": 0.09115484356880188, "global_step": 161947, "epoch": 1819, "lr": 3.5169068748518605e-05}
{"train_loss": 0.06548243016004562, "global_step": 161948, "epoch": 1819, "lr": 3.5168515103368785e-05}
{"train_loss": 0.07389587163925171, "global_step": 161949, "epoch": 1819, "lr": 3.5167961460212895e-05}
{"train_loss": 0.07004241645336151, "global_step": 161950, "epoch": 1819, "lr": 3.516740781905096e-05}
{"train_loss": 0.061155181378126144, "global_step": 161951, "epoch": 1819, "lr": 3.516685417988311e-05}
{"train_loss": 0.08144885301589966, "global_step": 161952, "epoch": 1819, "lr": 3.516630054270938e-05}
{"train_loss": 0.058827999979257584, "global_step": 161953, "epoch": 1819, "lr": 3.516574690752988e-05}
{"train_loss": 0.06705956906080246, "global_step": 161954, "epoch": 1819, "lr": 3.516519327434464e-05}
{"train_loss": 0.07135344296693802, "global_step": 161955, "epoch": 1819, "lr": 3.516463964315378e-05}
{"train_loss": 0.08221855014562607, "global_step": 161956, "epoch": 1819, "lr": 3.5164086013957334e-05}
{"train_loss": 0.08956273645162582, "global_step": 161957, "epoch": 1819, "lr": 3.51635323867554e-05}
{"train_loss": 0.07648750394582748, "global_step": 161958, "epoch": 1819, "lr": 3.516297876154806e-05}
{"train_loss": 0.04665127024054527, "global_step": 161959, "epoch": 1819, "lr": 3.516242513833536e-05}
{"train_loss": 0.06653708964586258, "global_step": 161960, "epoch": 1819, "lr": 3.516187151711741e-05}
{"train_loss": 0.051037780940532684, "global_step": 161961, "epoch": 1819, "lr": 3.516131789789425e-05}
{"train_loss": 0.15510129928588867, "global_step": 161962, "epoch": 1819, "lr": 3.5160764280665985e-05}
{"train_loss": 0.10497085750102997, "global_step": 161963, "epoch": 1819, "lr": 3.516021066543267e-05}
{"train_loss": 0.06263227760791779, "global_step": 161964, "epoch": 1819, "lr": 3.5159657052194384e-05}
{"train_loss": 0.09765903651714325, "global_step": 161965, "epoch": 1819, "lr": 3.515910344095119e-05}
{"train_loss": 0.11684460192918777, "global_step": 161966, "epoch": 1819, "lr": 3.515854983170319e-05}
{"train_loss": 0.110556460916996, "global_step": 161967, "epoch": 1819, "lr": 3.515799622445043e-05}
{"train_loss": 0.047546420246362686, "global_step": 161968, "epoch": 1819, "lr": 3.515744261919301e-05}
{"train_loss": 0.05379980802536011, "global_step": 161969, "epoch": 1819, "lr": 3.5156889015930974e-05}
{"train_loss": 0.05589195340871811, "global_step": 161970, "epoch": 1819, "lr": 3.515633541466443e-05}
{"train_loss": 0.059971168637275696, "global_step": 161971, "epoch": 1819, "lr": 3.5155781815393415e-05}
{"train_loss": 0.044568680226802826, "global_step": 161972, "epoch": 1819, "lr": 3.515522821811804e-05}
{"train_loss": 0.06303375214338303, "global_step": 161973, "epoch": 1819, "lr": 3.515467462283835e-05}
{"train_loss": 0.08398847281932831, "global_step": 161974, "epoch": 1819, "lr": 3.515412102955443e-05}
{"train_loss": 0.09738045930862427, "global_step": 161975, "epoch": 1819, "lr": 3.5153567438266385e-05}
{"train_loss": 0.0894588828086853, "global_step": 161976, "epoch": 1819, "lr": 3.5153013848974234e-05}
{"train_loss": 0.052278440445661545, "global_step": 161977, "epoch": 1819, "lr": 3.51524602616781e-05}
{"train_loss": 0.08465393632650375, "global_step": 161978, "epoch": 1819, "lr": 3.515190667637802e-05}
{"train_loss": 0.08948650148310018, "global_step": 161979, "epoch": 1819, "lr": 3.51513530930741e-05, "val_loss": 6.893360614776611}
{"train_loss": 0.08557320386171341, "global_step": 161980, "epoch": 1820, "lr": 3.515079951176638e-05}
{"train_loss": 0.08458375930786133, "global_step": 161981, "epoch": 1820, "lr": 3.5150245932454984e-05}
{"train_loss": 0.08622907102108002, "global_step": 161982, "epoch": 1820, "lr": 3.514969235513994e-05}
{"train_loss": 0.0797976702451706, "global_step": 161983, "epoch": 1820, "lr": 3.514913877982134e-05}
{"train_loss": 0.15559817850589752, "global_step": 161984, "epoch": 1820, "lr": 3.514858520649925e-05}
{"train_loss": 0.07446630299091339, "global_step": 161985, "epoch": 1820, "lr": 3.514803163517376e-05}
{"train_loss": 0.14426547288894653, "global_step": 161986, "epoch": 1820, "lr": 3.5147478065844935e-05}
{"train_loss": 0.09990299493074417, "global_step": 161987, "epoch": 1820, "lr": 3.514692449851286e-05}
{"train_loss": 0.08071085810661316, "global_step": 161988, "epoch": 1820, "lr": 3.514637093317758e-05}
{"train_loss": 0.07670146226882935, "global_step": 161989, "epoch": 1820, "lr": 3.514581736983921e-05}
{"train_loss": 0.09155263751745224, "global_step": 161990, "epoch": 1820, "lr": 3.514526380849778e-05}
{"train_loss": 0.1129927709698677, "global_step": 161991, "epoch": 1820, "lr": 3.514471024915341e-05}
{"train_loss": 0.05012594908475876, "global_step": 161992, "epoch": 1820, "lr": 3.514415669180613e-05}
{"train_loss": 0.056298211216926575, "global_step": 161993, "epoch": 1820, "lr": 3.514360313645604e-05}
{"train_loss": 0.09379749745130539, "global_step": 161994, "epoch": 1820, "lr": 3.514304958310324e-05}
{"train_loss": 0.02375752478837967, "global_step": 161995, "epoch": 1820, "lr": 3.5142496031747744e-05}
{"train_loss": 0.04811583459377289, "global_step": 161996, "epoch": 1820, "lr": 3.514194248238968e-05}
{"train_loss": 0.07331976294517517, "global_step": 161997, "epoch": 1820, "lr": 3.514138893502908e-05}
{"train_loss": 0.08915921300649643, "global_step": 161998, "epoch": 1820, "lr": 3.5140835389666065e-05}
{"train_loss": 0.052343204617500305, "global_step": 161999, "epoch": 1820, "lr": 3.514028184630066e-05}
{"train_loss": 0.1441333293914795, "global_step": 162000, "epoch": 1820, "lr": 3.513972830493298e-05}
{"train_loss": 0.09587506949901581, "global_step": 162001, "epoch": 1820, "lr": 3.5139174765563075e-05}
{"train_loss": 0.06861738115549088, "global_step": 162002, "epoch": 1820, "lr": 3.5138621228191035e-05}
{"train_loss": 0.14736279845237732, "global_step": 162003, "epoch": 1820, "lr": 3.5138067692816916e-05}
{"train_loss": 0.07879769057035446, "global_step": 162004, "epoch": 1820, "lr": 3.513751415944081e-05}
{"train_loss": 0.05840795114636421, "global_step": 162005, "epoch": 1820, "lr": 3.513696062806278e-05}
{"train_loss": 0.08712662011384964, "global_step": 162006, "epoch": 1820, "lr": 3.513640709868291e-05}
{"train_loss": 0.05842173472046852, "global_step": 162007, "epoch": 1820, "lr": 3.5135853571301254e-05}
{"train_loss": 0.13303782045841217, "global_step": 162008, "epoch": 1820, "lr": 3.5135300045917916e-05}
{"train_loss": 0.07655473053455353, "global_step": 162009, "epoch": 1820, "lr": 3.5134746522532945e-05}
{"train_loss": 0.06551209837198257, "global_step": 162010, "epoch": 1820, "lr": 3.513419300114642e-05}
{"train_loss": 0.08105538040399551, "global_step": 162011, "epoch": 1820, "lr": 3.513363948175844e-05}
{"train_loss": 0.09029808640480042, "global_step": 162012, "epoch": 1820, "lr": 3.513308596436905e-05}
{"train_loss": 0.07101540267467499, "global_step": 162013, "epoch": 1820, "lr": 3.513253244897834e-05}
{"train_loss": 0.03487708792090416, "global_step": 162014, "epoch": 1820, "lr": 3.5131978935586374e-05}
{"train_loss": 0.10855825245380402, "global_step": 162015, "epoch": 1820, "lr": 3.513142542419325e-05}
{"train_loss": 0.044562119990587234, "global_step": 162016, "epoch": 1820, "lr": 3.5130871914799005e-05}
{"train_loss": 0.11289890855550766, "global_step": 162017, "epoch": 1820, "lr": 3.5130318407403746e-05}
{"train_loss": 0.07550227642059326, "global_step": 162018, "epoch": 1820, "lr": 3.512976490200752e-05}
{"train_loss": 0.05382516607642174, "global_step": 162019, "epoch": 1820, "lr": 3.512921139861043e-05}
{"train_loss": 0.11734626442193985, "global_step": 162020, "epoch": 1820, "lr": 3.5128657897212524e-05}
{"train_loss": 0.08110260218381882, "global_step": 162021, "epoch": 1820, "lr": 3.51281043978139e-05}
{"train_loss": 0.10701929032802582, "global_step": 162022, "epoch": 1820, "lr": 3.512755090041462e-05}
{"train_loss": 0.04922034963965416, "global_step": 162023, "epoch": 1820, "lr": 3.5126997405014756e-05}
{"train_loss": 0.06408124417066574, "global_step": 162024, "epoch": 1820, "lr": 3.512644391161438e-05}
{"train_loss": 0.09094516932964325, "global_step": 162025, "epoch": 1820, "lr": 3.512589042021358e-05}
{"train_loss": 0.1525057554244995, "global_step": 162026, "epoch": 1820, "lr": 3.5125336930812426e-05}
{"train_loss": 0.07881246507167816, "global_step": 162027, "epoch": 1820, "lr": 3.5124783443410994e-05}
{"train_loss": 0.12491929531097412, "global_step": 162028, "epoch": 1820, "lr": 3.512422995800933e-05}
{"train_loss": 0.05784689635038376, "global_step": 162029, "epoch": 1820, "lr": 3.5123676474607534e-05}
{"train_loss": 0.1139497309923172, "global_step": 162030, "epoch": 1820, "lr": 3.5123122993205704e-05}
{"train_loss": 0.049028147011995316, "global_step": 162031, "epoch": 1820, "lr": 3.512256951380386e-05}
{"train_loss": 0.10228561609983444, "global_step": 162032, "epoch": 1820, "lr": 3.512201603640213e-05}
{"train_loss": 0.1094261109828949, "global_step": 162033, "epoch": 1820, "lr": 3.512146256100055e-05}
{"train_loss": 0.09974674880504608, "global_step": 162034, "epoch": 1820, "lr": 3.512090908759922e-05}
{"train_loss": 0.1429489552974701, "global_step": 162035, "epoch": 1820, "lr": 3.512035561619818e-05}
{"train_loss": 0.12082785367965698, "global_step": 162036, "epoch": 1820, "lr": 3.511980214679755e-05}
{"train_loss": 0.15812711417675018, "global_step": 162037, "epoch": 1820, "lr": 3.511924867939737e-05}
{"train_loss": 0.04363676905632019, "global_step": 162038, "epoch": 1820, "lr": 3.511869521399773e-05}
{"train_loss": 0.0747748389840126, "global_step": 162039, "epoch": 1820, "lr": 3.511814175059869e-05}
{"train_loss": 0.11377569288015366, "global_step": 162040, "epoch": 1820, "lr": 3.5117588289200356e-05}
{"train_loss": 0.1145937368273735, "global_step": 162041, "epoch": 1820, "lr": 3.511703482980276e-05}
{"train_loss": 0.08388672769069672, "global_step": 162042, "epoch": 1820, "lr": 3.511648137240601e-05}
{"train_loss": 0.1501576453447342, "global_step": 162043, "epoch": 1820, "lr": 3.511592791701016e-05}
{"train_loss": 0.04480401799082756, "global_step": 162044, "epoch": 1820, "lr": 3.511537446361529e-05}
{"train_loss": 0.10242775827646255, "global_step": 162045, "epoch": 1820, "lr": 3.511482101222148e-05}
{"train_loss": 0.07673626393079758, "global_step": 162046, "epoch": 1820, "lr": 3.5114267562828796e-05}
{"train_loss": 0.04149102047085762, "global_step": 162047, "epoch": 1820, "lr": 3.511371411543734e-05}
{"train_loss": 0.05972001329064369, "global_step": 162048, "epoch": 1820, "lr": 3.511316067004713e-05}
{"train_loss": 0.06083285063505173, "global_step": 162049, "epoch": 1820, "lr": 3.51126072266583e-05}
{"train_loss": 0.06350241601467133, "global_step": 162050, "epoch": 1820, "lr": 3.511205378527088e-05}
{"train_loss": 0.0907096341252327, "global_step": 162051, "epoch": 1820, "lr": 3.511150034588498e-05}
{"train_loss": 0.05420129373669624, "global_step": 162052, "epoch": 1820, "lr": 3.511094690850064e-05}
{"train_loss": 0.0592869371175766, "global_step": 162053, "epoch": 1820, "lr": 3.511039347311797e-05}
{"train_loss": 0.0762382447719574, "global_step": 162054, "epoch": 1820, "lr": 3.510984003973701e-05}
{"train_loss": 0.10339552164077759, "global_step": 162055, "epoch": 1820, "lr": 3.510928660835786e-05}
{"train_loss": 0.17927135527133942, "global_step": 162056, "epoch": 1820, "lr": 3.510873317898058e-05}
{"train_loss": 0.0671476423740387, "global_step": 162057, "epoch": 1820, "lr": 3.510817975160526e-05}
{"train_loss": 0.05476725101470947, "global_step": 162058, "epoch": 1820, "lr": 3.510762632623195e-05}
{"train_loss": 0.15019294619560242, "global_step": 162059, "epoch": 1820, "lr": 3.510707290286075e-05}
{"train_loss": 0.16959957778453827, "global_step": 162060, "epoch": 1820, "lr": 3.510651948149171e-05}
{"train_loss": 0.12033423781394958, "global_step": 162061, "epoch": 1820, "lr": 3.5105966062124926e-05}
{"train_loss": 0.0759163424372673, "global_step": 162062, "epoch": 1820, "lr": 3.510541264476046e-05}
{"train_loss": 0.0995657742023468, "global_step": 162063, "epoch": 1820, "lr": 3.510485922939839e-05}
{"train_loss": 0.10501837730407715, "global_step": 162064, "epoch": 1820, "lr": 3.510430581603879e-05}
{"train_loss": 0.10023908317089081, "global_step": 162065, "epoch": 1820, "lr": 3.510375240468172e-05}
{"train_loss": 0.10142405331134796, "global_step": 162066, "epoch": 1820, "lr": 3.51031989953273e-05}
{"train_loss": 0.068138487637043, "global_step": 162067, "epoch": 1820, "lr": 3.5102645587975545e-05}
{"train_loss": 0.0894967286188281, "global_step": 162068, "epoch": 1820, "lr": 3.510209218262658e-05, "val_loss": 6.977118492126465, "train_action_mse_error": 8.182168006896973}
{"train_loss": 0.10569041967391968, "global_step": 162069, "epoch": 1821, "lr": 3.510153877928044e-05}
{"train_loss": 0.062484387308359146, "global_step": 162070, "epoch": 1821, "lr": 3.510098537793723e-05}
{"train_loss": 0.12760645151138306, "global_step": 162071, "epoch": 1821, "lr": 3.510043197859699e-05}
{"train_loss": 0.06879272311925888, "global_step": 162072, "epoch": 1821, "lr": 3.5099878581259845e-05}
{"train_loss": 0.08563819527626038, "global_step": 162073, "epoch": 1821, "lr": 3.509932518592581e-05}
{"train_loss": 0.07669245451688766, "global_step": 162074, "epoch": 1821, "lr": 3.509877179259502e-05}
{"train_loss": 0.09696704149246216, "global_step": 162075, "epoch": 1821, "lr": 3.509821840126749e-05}
{"train_loss": 0.0848926529288292, "global_step": 162076, "epoch": 1821, "lr": 3.5097665011943346e-05}
{"train_loss": 0.10502529889345169, "global_step": 162077, "epoch": 1821, "lr": 3.509711162462262e-05}
{"train_loss": 0.11576822400093079, "global_step": 162078, "epoch": 1821, "lr": 3.509655823930542e-05}
{"train_loss": 0.10153163969516754, "global_step": 162079, "epoch": 1821, "lr": 3.5096004855991806e-05}
{"train_loss": 0.10186968743801117, "global_step": 162080, "epoch": 1821, "lr": 3.509545147468184e-05}
{"train_loss": 0.10467053204774857, "global_step": 162081, "epoch": 1821, "lr": 3.5094898095375615e-05}
{"train_loss": 0.10116316378116608, "global_step": 162082, "epoch": 1821, "lr": 3.5094344718073205e-05}
{"train_loss": 0.085565946996212, "global_step": 162083, "epoch": 1821, "lr": 3.509379134277467e-05}
{"train_loss": 0.06787605583667755, "global_step": 162084, "epoch": 1821, "lr": 3.5093237969480084e-05}
{"train_loss": 0.09174710512161255, "global_step": 162085, "epoch": 1821, "lr": 3.5092684598189555e-05}
{"train_loss": 0.14587630331516266, "global_step": 162086, "epoch": 1821, "lr": 3.509213122890311e-05}
{"train_loss": 0.09525684267282486, "global_step": 162087, "epoch": 1821, "lr": 3.509157786162087e-05}
{"train_loss": 0.09551623463630676, "global_step": 162088, "epoch": 1821, "lr": 3.509102449634286e-05}
{"train_loss": 0.07293963432312012, "global_step": 162089, "epoch": 1821, "lr": 3.50904711330692e-05}
{"train_loss": 0.15944072604179382, "global_step": 162090, "epoch": 1821, "lr": 3.508991777179993e-05}
{"train_loss": 0.1239331066608429, "global_step": 162091, "epoch": 1821, "lr": 3.508936441253516e-05}
{"train_loss": 0.160375714302063, "global_step": 162092, "epoch": 1821, "lr": 3.5088811055274916e-05}
{"train_loss": 0.13686417043209076, "global_step": 162093, "epoch": 1821, "lr": 3.5088257700019324e-05}
{"train_loss": 0.11086401343345642, "global_step": 162094, "epoch": 1821, "lr": 3.5087704346768415e-05}
{"train_loss": 0.050481412559747696, "global_step": 162095, "epoch": 1821, "lr": 3.50871509955223e-05}
{"train_loss": 0.07305161654949188, "global_step": 162096, "epoch": 1821, "lr": 3.508659764628102e-05}
{"train_loss": 0.08458525687456131, "global_step": 162097, "epoch": 1821, "lr": 3.508604429904467e-05}
{"train_loss": 0.13918443024158478, "global_step": 162098, "epoch": 1821, "lr": 3.508549095381333e-05}
{"train_loss": 0.14206556975841522, "global_step": 162099, "epoch": 1821, "lr": 3.5084937610587056e-05}
{"train_loss": 0.10673267394304276, "global_step": 162100, "epoch": 1821, "lr": 3.508438426936592e-05}
{"train_loss": 0.1942053884267807, "global_step": 162101, "epoch": 1821, "lr": 3.5083830930150007e-05}
{"train_loss": 0.12501877546310425, "global_step": 162102, "epoch": 1821, "lr": 3.5083277592939415e-05}
{"train_loss": 0.07156454026699066, "global_step": 162103, "epoch": 1821, "lr": 3.508272425773417e-05}
{"train_loss": 0.1345745027065277, "global_step": 162104, "epoch": 1821, "lr": 3.508217092453439e-05}
{"train_loss": 0.12848590314388275, "global_step": 162105, "epoch": 1821, "lr": 3.508161759334011e-05}
{"train_loss": 0.11489153653383255, "global_step": 162106, "epoch": 1821, "lr": 3.508106426415144e-05}
{"train_loss": 0.08196321874856949, "global_step": 162107, "epoch": 1821, "lr": 3.5080510936968423e-05}
{"train_loss": 0.10012099146842957, "global_step": 162108, "epoch": 1821, "lr": 3.5079957611791175e-05}
{"train_loss": 0.04803336039185524, "global_step": 162109, "epoch": 1821, "lr": 3.507940428861972e-05}
{"train_loss": 0.06812195479869843, "global_step": 162110, "epoch": 1821, "lr": 3.507885096745418e-05}
{"train_loss": 0.054565321654081345, "global_step": 162111, "epoch": 1821, "lr": 3.507829764829458e-05}
{"train_loss": 0.09117808192968369, "global_step": 162112, "epoch": 1821, "lr": 3.507774433114105e-05}
{"train_loss": 0.10502266883850098, "global_step": 162113, "epoch": 1821, "lr": 3.507719101599361e-05}
{"train_loss": 0.10255498439073563, "global_step": 162114, "epoch": 1821, "lr": 3.5076637702852385e-05}
{"train_loss": 0.08510099351406097, "global_step": 162115, "epoch": 1821, "lr": 3.50760843917174e-05}
{"train_loss": 0.11619557440280914, "global_step": 162116, "epoch": 1821, "lr": 3.5075531082588766e-05}
{"train_loss": 0.06298348307609558, "global_step": 162117, "epoch": 1821, "lr": 3.5074977775466547e-05}
{"train_loss": 0.11515801399946213, "global_step": 162118, "epoch": 1821, "lr": 3.5074424470350805e-05}
{"train_loss": 0.06440463662147522, "global_step": 162119, "epoch": 1821, "lr": 3.507387116724163e-05}
{"train_loss": 0.12166265398263931, "global_step": 162120, "epoch": 1821, "lr": 3.5073317866139085e-05}
{"train_loss": 0.09077578037977219, "global_step": 162121, "epoch": 1821, "lr": 3.507276456704327e-05}
{"train_loss": 0.12359991669654846, "global_step": 162122, "epoch": 1821, "lr": 3.5072211269954214e-05}
{"train_loss": 0.07805448025465012, "global_step": 162123, "epoch": 1821, "lr": 3.507165797487204e-05}
{"train_loss": 0.10795406252145767, "global_step": 162124, "epoch": 1821, "lr": 3.5071104681796783e-05}
{"train_loss": 0.06420457363128662, "global_step": 162125, "epoch": 1821, "lr": 3.507055139072855e-05}
{"train_loss": 0.06595585495233536, "global_step": 162126, "epoch": 1821, "lr": 3.506999810166738e-05}
{"train_loss": 0.10266274958848953, "global_step": 162127, "epoch": 1821, "lr": 3.506944481461339e-05}
{"train_loss": 0.09752662479877472, "global_step": 162128, "epoch": 1821, "lr": 3.5068891529566606e-05}
{"train_loss": 0.06108293682336807, "global_step": 162129, "epoch": 1821, "lr": 3.506833824652715e-05}
{"train_loss": 0.08853938430547714, "global_step": 162130, "epoch": 1821, "lr": 3.506778496549505e-05}
{"train_loss": 0.18659068644046783, "global_step": 162131, "epoch": 1821, "lr": 3.506723168647043e-05}
{"train_loss": 0.09162461757659912, "global_step": 162132, "epoch": 1821, "lr": 3.506667840945332e-05}
{"train_loss": 0.07298702746629715, "global_step": 162133, "epoch": 1821, "lr": 3.506612513444383e-05}
{"train_loss": 0.07765844464302063, "global_step": 162134, "epoch": 1821, "lr": 3.506557186144199e-05}
{"train_loss": 0.09301441162824631, "global_step": 162135, "epoch": 1821, "lr": 3.506501859044794e-05}
{"train_loss": 0.17216506600379944, "global_step": 162136, "epoch": 1821, "lr": 3.506446532146168e-05}
{"train_loss": 0.05921536684036255, "global_step": 162137, "epoch": 1821, "lr": 3.506391205448334e-05}
{"train_loss": 0.11517225205898285, "global_step": 162138, "epoch": 1821, "lr": 3.506335878951298e-05}
{"train_loss": 0.06784457713365555, "global_step": 162139, "epoch": 1821, "lr": 3.5062805526550645e-05}
{"train_loss": 0.10798852145671844, "global_step": 162140, "epoch": 1821, "lr": 3.5062252265596464e-05}
{"train_loss": 0.07041475176811218, "global_step": 162141, "epoch": 1821, "lr": 3.506169900665045e-05}
{"train_loss": 0.06003689393401146, "global_step": 162142, "epoch": 1821, "lr": 3.5061145749712736e-05}
{"train_loss": 0.0738186165690422, "global_step": 162143, "epoch": 1821, "lr": 3.506059249478335e-05}
{"train_loss": 0.07269127666950226, "global_step": 162144, "epoch": 1821, "lr": 3.5060039241862404e-05}
{"train_loss": 0.05739776790142059, "global_step": 162145, "epoch": 1821, "lr": 3.505948599094993e-05}
{"train_loss": 0.09116718918085098, "global_step": 162146, "epoch": 1821, "lr": 3.5058932742046044e-05}
{"train_loss": 0.10293567180633545, "global_step": 162147, "epoch": 1821, "lr": 3.505837949515079e-05}
{"train_loss": 0.12602540850639343, "global_step": 162148, "epoch": 1821, "lr": 3.505782625026427e-05}
{"train_loss": 0.12668152153491974, "global_step": 162149, "epoch": 1821, "lr": 3.505727300738653e-05}
{"train_loss": 0.1202259361743927, "global_step": 162150, "epoch": 1821, "lr": 3.505671976651767e-05}
{"train_loss": 0.06379275023937225, "global_step": 162151, "epoch": 1821, "lr": 3.505616652765773e-05}
{"train_loss": 0.14174677431583405, "global_step": 162152, "epoch": 1821, "lr": 3.5055613290806834e-05}
{"train_loss": 0.18235906958580017, "global_step": 162153, "epoch": 1821, "lr": 3.5055060055965005e-05}
{"train_loss": 0.09855937957763672, "global_step": 162154, "epoch": 1821, "lr": 3.505450682313236e-05}
{"train_loss": 0.0907413586974144, "global_step": 162155, "epoch": 1821, "lr": 3.5053953592308935e-05}
{"train_loss": 0.0982152670621872, "global_step": 162156, "epoch": 1821, "lr": 3.505340036349484e-05}
{"train_loss": 0.10000518738721194, "global_step": 162157, "epoch": 1821, "lr": 3.505284713669013e-05, "val_loss": 6.802487850189209}
{"train_loss": 0.10268771648406982, "global_step": 162158, "epoch": 1822, "lr": 3.505229391189487e-05}
{"train_loss": 0.12370489537715912, "global_step": 162159, "epoch": 1822, "lr": 3.505174068910917e-05}
{"train_loss": 0.10583965480327606, "global_step": 162160, "epoch": 1822, "lr": 3.505118746833306e-05}
{"train_loss": 0.06796800345182419, "global_step": 162161, "epoch": 1822, "lr": 3.505063424956665e-05}
{"train_loss": 0.04893580451607704, "global_step": 162162, "epoch": 1822, "lr": 3.5050081032809984e-05}
{"train_loss": 0.10999184101819992, "global_step": 162163, "epoch": 1822, "lr": 3.504952781806317e-05}
{"train_loss": 0.08383994549512863, "global_step": 162164, "epoch": 1822, "lr": 3.5048974605326246e-05}
{"train_loss": 0.09083665162324905, "global_step": 162165, "epoch": 1822, "lr": 3.504842139459933e-05}
{"train_loss": 0.14860470592975616, "global_step": 162166, "epoch": 1822, "lr": 3.504786818588245e-05}
{"train_loss": 0.12312474846839905, "global_step": 162167, "epoch": 1822, "lr": 3.5047314979175715e-05}
{"train_loss": 0.11177908629179001, "global_step": 162168, "epoch": 1822, "lr": 3.504676177447917e-05}
{"train_loss": 0.08354882150888443, "global_step": 162169, "epoch": 1822, "lr": 3.5046208571792923e-05}
{"train_loss": 0.10234000533819199, "global_step": 162170, "epoch": 1822, "lr": 3.5045655371117016e-05}
{"train_loss": 0.13154184818267822, "global_step": 162171, "epoch": 1822, "lr": 3.504510217245156e-05}
{"train_loss": 0.07897061109542847, "global_step": 162172, "epoch": 1822, "lr": 3.504454897579658e-05}
{"train_loss": 0.07435184717178345, "global_step": 162173, "epoch": 1822, "lr": 3.504399578115219e-05}
{"train_loss": 0.05812973529100418, "global_step": 162174, "epoch": 1822, "lr": 3.5043442588518463e-05}
{"train_loss": 0.06952260434627533, "global_step": 162175, "epoch": 1822, "lr": 3.504288939789545e-05}
{"train_loss": 0.06446731835603714, "global_step": 162176, "epoch": 1822, "lr": 3.5042336209283254e-05}
{"train_loss": 0.16411714255809784, "global_step": 162177, "epoch": 1822, "lr": 3.504178302268192e-05}
{"train_loss": 0.15835300087928772, "global_step": 162178, "epoch": 1822, "lr": 3.5041229838091544e-05}
{"train_loss": 0.1278688609600067, "global_step": 162179, "epoch": 1822, "lr": 3.504067665551217e-05}
{"train_loss": 0.1254967302083969, "global_step": 162180, "epoch": 1822, "lr": 3.504012347494393e-05}
{"train_loss": 0.10575619339942932, "global_step": 162181, "epoch": 1822, "lr": 3.503957029638683e-05}
{"train_loss": 0.09064197540283203, "global_step": 162182, "epoch": 1822, "lr": 3.5039017119841e-05}
{"train_loss": 0.10717793554067612, "global_step": 162183, "epoch": 1822, "lr": 3.5038463945306474e-05}
{"train_loss": 0.06500258296728134, "global_step": 162184, "epoch": 1822, "lr": 3.503791077278337e-05}
{"train_loss": 0.15346023440361023, "global_step": 162185, "epoch": 1822, "lr": 3.503735760227171e-05}
{"train_loss": 0.0690910667181015, "global_step": 162186, "epoch": 1822, "lr": 3.503680443377161e-05}
{"train_loss": 0.12076395750045776, "global_step": 162187, "epoch": 1822, "lr": 3.503625126728312e-05}
{"train_loss": 0.08420776575803757, "global_step": 162188, "epoch": 1822, "lr": 3.5035698102806335e-05}
{"train_loss": 0.12334055453538895, "global_step": 162189, "epoch": 1822, "lr": 3.5035144940341304e-05}
{"train_loss": 0.07666095346212387, "global_step": 162190, "epoch": 1822, "lr": 3.503459177988813e-05}
{"train_loss": 0.05343944579362869, "global_step": 162191, "epoch": 1822, "lr": 3.503403862144686e-05}
{"train_loss": 0.06332404166460037, "global_step": 162192, "epoch": 1822, "lr": 3.503348546501758e-05}
{"train_loss": 0.1571904569864273, "global_step": 162193, "epoch": 1822, "lr": 3.503293231060038e-05}
{"train_loss": 0.07230257242918015, "global_step": 162194, "epoch": 1822, "lr": 3.5032379158195305e-05}
{"train_loss": 0.04472627490758896, "global_step": 162195, "epoch": 1822, "lr": 3.503182600780246e-05}
{"train_loss": 0.11830142885446548, "global_step": 162196, "epoch": 1822, "lr": 3.503127285942189e-05}
{"train_loss": 0.13630801439285278, "global_step": 162197, "epoch": 1822, "lr": 3.5030719713053696e-05}
{"train_loss": 0.1367737352848053, "global_step": 162198, "epoch": 1822, "lr": 3.5030166568697917e-05}
{"train_loss": 0.14114323258399963, "global_step": 162199, "epoch": 1822, "lr": 3.502961342635468e-05}
{"train_loss": 0.08549505472183228, "global_step": 162200, "epoch": 1822, "lr": 3.5029060286023995e-05}
{"train_loss": 0.06475088745355606, "global_step": 162201, "epoch": 1822, "lr": 3.5028507147706e-05}
{"train_loss": 0.06054205819964409, "global_step": 162202, "epoch": 1822, "lr": 3.502795401140072e-05}
{"train_loss": 0.11572234332561493, "global_step": 162203, "epoch": 1822, "lr": 3.5027400877108264e-05}
{"train_loss": 0.09205889701843262, "global_step": 162204, "epoch": 1822, "lr": 3.5026847744828675e-05}
{"train_loss": 0.0836353749036789, "global_step": 162205, "epoch": 1822, "lr": 3.502629461456206e-05}
{"train_loss": 0.09234211593866348, "global_step": 162206, "epoch": 1822, "lr": 3.502574148630846e-05}
{"train_loss": 0.10585423558950424, "global_step": 162207, "epoch": 1822, "lr": 3.502518836006799e-05}
{"train_loss": 0.07360531389713287, "global_step": 162208, "epoch": 1822, "lr": 3.502463523584068e-05}
{"train_loss": 0.084803007543087, "global_step": 162209, "epoch": 1822, "lr": 3.5024082113626644e-05}
{"train_loss": 0.09568338841199875, "global_step": 162210, "epoch": 1822, "lr": 3.502352899342591e-05}
{"train_loss": 0.0966293215751648, "global_step": 162211, "epoch": 1822, "lr": 3.50229758752386e-05}
{"train_loss": 0.06283749639987946, "global_step": 162212, "epoch": 1822, "lr": 3.5022422759064766e-05}
{"train_loss": 0.05560417100787163, "global_step": 162213, "epoch": 1822, "lr": 3.502186964490448e-05}
{"train_loss": 0.05785488337278366, "global_step": 162214, "epoch": 1822, "lr": 3.502131653275784e-05}
{"train_loss": 0.0540890246629715, "global_step": 162215, "epoch": 1822, "lr": 3.502076342262487e-05}
{"train_loss": 0.08403845131397247, "global_step": 162216, "epoch": 1822, "lr": 3.5020210314505706e-05}
{"train_loss": 0.10888677090406418, "global_step": 162217, "epoch": 1822, "lr": 3.501965720840037e-05}
{"train_loss": 0.06432415544986725, "global_step": 162218, "epoch": 1822, "lr": 3.501910410430897e-05}
{"train_loss": 0.11742675304412842, "global_step": 162219, "epoch": 1822, "lr": 3.5018551002231555e-05}
{"train_loss": 0.08676416426897049, "global_step": 162220, "epoch": 1822, "lr": 3.5017997902168234e-05}
{"train_loss": 0.07319480925798416, "global_step": 162221, "epoch": 1822, "lr": 3.501744480411904e-05}
{"train_loss": 0.06407447904348373, "global_step": 162222, "epoch": 1822, "lr": 3.5016891708084086e-05}
{"train_loss": 0.1195157915353775, "global_step": 162223, "epoch": 1822, "lr": 3.5016338614063416e-05}
{"train_loss": 0.1491624414920807, "global_step": 162224, "epoch": 1822, "lr": 3.501578552205713e-05}
{"train_loss": 0.06823034584522247, "global_step": 162225, "epoch": 1822, "lr": 3.501523243206527e-05}
{"train_loss": 0.06788235902786255, "global_step": 162226, "epoch": 1822, "lr": 3.5014679344087956e-05}
{"train_loss": 0.076170913875103, "global_step": 162227, "epoch": 1822, "lr": 3.50141262581252e-05}
{"train_loss": 0.09848257154226303, "global_step": 162228, "epoch": 1822, "lr": 3.501357317417713e-05}
{"train_loss": 0.04874264821410179, "global_step": 162229, "epoch": 1822, "lr": 3.501302009224381e-05}
{"train_loss": 0.09618102759122849, "global_step": 162230, "epoch": 1822, "lr": 3.50124670123253e-05}
{"train_loss": 0.09095881879329681, "global_step": 162231, "epoch": 1822, "lr": 3.501191393442168e-05}
{"train_loss": 0.09867151081562042, "global_step": 162232, "epoch": 1822, "lr": 3.5011360858533024e-05}
{"train_loss": 0.09941043704748154, "global_step": 162233, "epoch": 1822, "lr": 3.5010807784659426e-05}
{"train_loss": 0.129330113530159, "global_step": 162234, "epoch": 1822, "lr": 3.501025471280092e-05}
{"train_loss": 0.08135449141263962, "global_step": 162235, "epoch": 1822, "lr": 3.500970164295762e-05}
{"train_loss": 0.11309131234884262, "global_step": 162236, "epoch": 1822, "lr": 3.500914857512956e-05}
{"train_loss": 0.11674273014068604, "global_step": 162237, "epoch": 1822, "lr": 3.5008595509316854e-05}
{"train_loss": 0.06245212256908417, "global_step": 162238, "epoch": 1822, "lr": 3.500804244551955e-05}
{"train_loss": 0.056805990636348724, "global_step": 162239, "epoch": 1822, "lr": 3.500748938373774e-05}
{"train_loss": 0.07926114648580551, "global_step": 162240, "epoch": 1822, "lr": 3.5006936323971485e-05}
{"train_loss": 0.05271918699145317, "global_step": 162241, "epoch": 1822, "lr": 3.5006383266220874e-05}
{"train_loss": 0.08070220053195953, "global_step": 162242, "epoch": 1822, "lr": 3.500583021048596e-05}
{"train_loss": 0.08811325579881668, "global_step": 162243, "epoch": 1822, "lr": 3.5005277156766844e-05}
{"train_loss": 0.08959419280290604, "global_step": 162244, "epoch": 1822, "lr": 3.500472410506356e-05}
{"train_loss": 0.08398846536874771, "global_step": 162245, "epoch": 1822, "lr": 3.5004171055376236e-05}
{"train_loss": 0.09388364889146236, "global_step": 162246, "epoch": 1822, "lr": 3.50036180077049e-05, "val_loss": 6.906123638153076}
{"train_loss": 0.09376179426908493, "global_step": 162247, "epoch": 1823, "lr": 3.5003064962049646e-05}
{"train_loss": 0.08105086535215378, "global_step": 162248, "epoch": 1823, "lr": 3.5002511918410556e-05}
{"train_loss": 0.08685558289289474, "global_step": 162249, "epoch": 1823, "lr": 3.5001958876787697e-05}
{"train_loss": 0.10822641849517822, "global_step": 162250, "epoch": 1823, "lr": 3.5001405837181135e-05}
{"train_loss": 0.12837259471416473, "global_step": 162251, "epoch": 1823, "lr": 3.500085279959096e-05}
{"train_loss": 0.06588892638683319, "global_step": 162252, "epoch": 1823, "lr": 3.500029976401722e-05}
{"train_loss": 0.13009940087795258, "global_step": 162253, "epoch": 1823, "lr": 3.499974673046002e-05}
{"train_loss": 0.11353038251399994, "global_step": 162254, "epoch": 1823, "lr": 3.499919369891943e-05}
{"train_loss": 0.11864755302667618, "global_step": 162255, "epoch": 1823, "lr": 3.499864066939549e-05}
{"train_loss": 0.09759394824504852, "global_step": 162256, "epoch": 1823, "lr": 3.4998087641888334e-05}
{"train_loss": 0.13009782135486603, "global_step": 162257, "epoch": 1823, "lr": 3.499753461639796e-05}
{"train_loss": 0.1084773913025856, "global_step": 162258, "epoch": 1823, "lr": 3.4996981592924525e-05}
{"train_loss": 0.08926361054182053, "global_step": 162259, "epoch": 1823, "lr": 3.499642857146803e-05}
{"train_loss": 0.06833665072917938, "global_step": 162260, "epoch": 1823, "lr": 3.499587555202861e-05}
{"train_loss": 0.1419476419687271, "global_step": 162261, "epoch": 1823, "lr": 3.499532253460629e-05}
{"train_loss": 0.07305457442998886, "global_step": 162262, "epoch": 1823, "lr": 3.4994769519201186e-05}
{"train_loss": 0.08892665803432465, "global_step": 162263, "epoch": 1823, "lr": 3.499421650581333e-05}
{"train_loss": 0.11216121166944504, "global_step": 162264, "epoch": 1823, "lr": 3.499366349444282e-05}
{"train_loss": 0.11044035106897354, "global_step": 162265, "epoch": 1823, "lr": 3.4993110485089754e-05}
{"train_loss": 0.07566212862730026, "global_step": 162266, "epoch": 1823, "lr": 3.499255747775415e-05}
{"train_loss": 0.0748470276594162, "global_step": 162267, "epoch": 1823, "lr": 3.499200447243615e-05}
{"train_loss": 0.11045899242162704, "global_step": 162268, "epoch": 1823, "lr": 3.499145146913577e-05}
{"train_loss": 0.07546193152666092, "global_step": 162269, "epoch": 1823, "lr": 3.4990898467853104e-05}
{"train_loss": 0.05235942453145981, "global_step": 162270, "epoch": 1823, "lr": 3.4990345468588234e-05}
{"train_loss": 0.1445305198431015, "global_step": 162271, "epoch": 1823, "lr": 3.4989792471341234e-05}
{"train_loss": 0.08896559476852417, "global_step": 162272, "epoch": 1823, "lr": 3.4989239476112165e-05}
{"train_loss": 0.0640854462981224, "global_step": 162273, "epoch": 1823, "lr": 3.498868648290112e-05}
{"train_loss": 0.12792856991291046, "global_step": 162274, "epoch": 1823, "lr": 3.498813349170815e-05}
{"train_loss": 0.08494367450475693, "global_step": 162275, "epoch": 1823, "lr": 3.4987580502533354e-05}
{"train_loss": 0.09815306961536407, "global_step": 162276, "epoch": 1823, "lr": 3.4987027515376786e-05}
{"train_loss": 0.08113722503185272, "global_step": 162277, "epoch": 1823, "lr": 3.498647453023854e-05}
{"train_loss": 0.08239804208278656, "global_step": 162278, "epoch": 1823, "lr": 3.498592154711866e-05}
{"train_loss": 0.09287121891975403, "global_step": 162279, "epoch": 1823, "lr": 3.498536856601726e-05}
{"train_loss": 0.09228674322366714, "global_step": 162280, "epoch": 1823, "lr": 3.498481558693438e-05}
{"train_loss": 0.06972309201955795, "global_step": 162281, "epoch": 1823, "lr": 3.4984262609870125e-05}
{"train_loss": 0.13628585636615753, "global_step": 162282, "epoch": 1823, "lr": 3.4983709634824536e-05}
{"train_loss": 0.08078442513942719, "global_step": 162283, "epoch": 1823, "lr": 3.49831566617977e-05}
{"train_loss": 0.05421065911650658, "global_step": 162284, "epoch": 1823, "lr": 3.498260369078972e-05}
{"train_loss": 0.114823117852211, "global_step": 162285, "epoch": 1823, "lr": 3.498205072180062e-05}
{"train_loss": 0.13583184778690338, "global_step": 162286, "epoch": 1823, "lr": 3.498149775483052e-05}
{"train_loss": 0.09302180260419846, "global_step": 162287, "epoch": 1823, "lr": 3.498094478987946e-05}
{"train_loss": 0.04645925015211105, "global_step": 162288, "epoch": 1823, "lr": 3.498039182694754e-05}
{"train_loss": 0.05468830093741417, "global_step": 162289, "epoch": 1823, "lr": 3.497983886603482e-05}
{"train_loss": 0.12301851809024811, "global_step": 162290, "epoch": 1823, "lr": 3.497928590714138e-05}
{"train_loss": 0.038392070680856705, "global_step": 162291, "epoch": 1823, "lr": 3.497873295026729e-05}
{"train_loss": 0.0864163264632225, "global_step": 162292, "epoch": 1823, "lr": 3.4978179995412634e-05}
{"train_loss": 0.10419101268053055, "global_step": 162293, "epoch": 1823, "lr": 3.497762704257746e-05}
{"train_loss": 0.11918129771947861, "global_step": 162294, "epoch": 1823, "lr": 3.497707409176189e-05}
{"train_loss": 0.061894774436950684, "global_step": 162295, "epoch": 1823, "lr": 3.497652114296593e-05}
{"train_loss": 0.09242033213376999, "global_step": 162296, "epoch": 1823, "lr": 3.497596819618972e-05}
{"train_loss": 0.1026768907904625, "global_step": 162297, "epoch": 1823, "lr": 3.49754152514333e-05}
{"train_loss": 0.07830733060836792, "global_step": 162298, "epoch": 1823, "lr": 3.497486230869676e-05}
{"train_loss": 0.0959877297282219, "global_step": 162299, "epoch": 1823, "lr": 3.4974309367980155e-05}
{"train_loss": 0.11100845038890839, "global_step": 162300, "epoch": 1823, "lr": 3.497375642928356e-05}
{"train_loss": 0.10611789673566818, "global_step": 162301, "epoch": 1823, "lr": 3.497320349260709e-05}
{"train_loss": 0.05226331576704979, "global_step": 162302, "epoch": 1823, "lr": 3.497265055795076e-05}
{"train_loss": 0.08653143048286438, "global_step": 162303, "epoch": 1823, "lr": 3.49720976253147e-05}
{"train_loss": 0.13853634893894196, "global_step": 162304, "epoch": 1823, "lr": 3.4971544694698946e-05}
{"train_loss": 0.06659381091594696, "global_step": 162305, "epoch": 1823, "lr": 3.4970991766103594e-05}
{"train_loss": 0.07174430787563324, "global_step": 162306, "epoch": 1823, "lr": 3.497043883952871e-05}
{"train_loss": 0.086381696164608, "global_step": 162307, "epoch": 1823, "lr": 3.496988591497435e-05}
{"train_loss": 0.07402054965496063, "global_step": 162308, "epoch": 1823, "lr": 3.496933299244062e-05}
{"train_loss": 0.10499785840511322, "global_step": 162309, "epoch": 1823, "lr": 3.496878007192757e-05}
{"train_loss": 0.11459609866142273, "global_step": 162310, "epoch": 1823, "lr": 3.49682271534353e-05}
{"train_loss": 0.050162896513938904, "global_step": 162311, "epoch": 1823, "lr": 3.496767423696387e-05}
{"train_loss": 0.0909574031829834, "global_step": 162312, "epoch": 1823, "lr": 3.496712132251333e-05}
{"train_loss": 0.07252271473407745, "global_step": 162313, "epoch": 1823, "lr": 3.496656841008379e-05}
{"train_loss": 0.16372810304164886, "global_step": 162314, "epoch": 1823, "lr": 3.4966015499675305e-05}
{"train_loss": 0.10058700293302536, "global_step": 162315, "epoch": 1823, "lr": 3.496546259128797e-05}
{"train_loss": 0.09765742719173431, "global_step": 162316, "epoch": 1823, "lr": 3.496490968492182e-05}
{"train_loss": 0.11023440957069397, "global_step": 162317, "epoch": 1823, "lr": 3.496435678057699e-05}
{"train_loss": 0.060787785798311234, "global_step": 162318, "epoch": 1823, "lr": 3.4963803878253484e-05}
{"train_loss": 0.1146068274974823, "global_step": 162319, "epoch": 1823, "lr": 3.496325097795142e-05}
{"train_loss": 0.1646737903356552, "global_step": 162320, "epoch": 1823, "lr": 3.496269807967088e-05}
{"train_loss": 0.06748635321855545, "global_step": 162321, "epoch": 1823, "lr": 3.49621451834119e-05}
{"train_loss": 0.06310256570577621, "global_step": 162322, "epoch": 1823, "lr": 3.496159228917461e-05}
{"train_loss": 0.05407936871051788, "global_step": 162323, "epoch": 1823, "lr": 3.496103939695901e-05}
{"train_loss": 0.06663703918457031, "global_step": 162324, "epoch": 1823, "lr": 3.4960486506765244e-05}
{"train_loss": 0.10256519168615341, "global_step": 162325, "epoch": 1823, "lr": 3.495993361859335e-05}
{"train_loss": 0.08016635477542877, "global_step": 162326, "epoch": 1823, "lr": 3.495938073244341e-05}
{"train_loss": 0.06551550328731537, "global_step": 162327, "epoch": 1823, "lr": 3.495882784831549e-05}
{"train_loss": 0.05820893496274948, "global_step": 162328, "epoch": 1823, "lr": 3.495827496620968e-05}
{"train_loss": 0.06864003837108612, "global_step": 162329, "epoch": 1823, "lr": 3.4957722086126045e-05}
{"train_loss": 0.11546140909194946, "global_step": 162330, "epoch": 1823, "lr": 3.495716920806467e-05}
{"train_loss": 0.09250645339488983, "global_step": 162331, "epoch": 1823, "lr": 3.4956616332025606e-05}
{"train_loss": 0.10032960772514343, "global_step": 162332, "epoch": 1823, "lr": 3.495606345800895e-05}
{"train_loss": 0.10228418558835983, "global_step": 162333, "epoch": 1823, "lr": 3.495551058601476e-05}
{"train_loss": 0.07002721726894379, "global_step": 162334, "epoch": 1823, "lr": 3.495495771604313e-05}
{"train_loss": 0.09254941365189767, "global_step": 162335, "epoch": 1823, "lr": 3.49544048480941e-05, "val_loss": 6.801942348480225}
{"train_loss": 0.05874263495206833, "global_step": 162336, "epoch": 1824, "lr": 3.4953851982167794e-05}
{"train_loss": 0.09028401970863342, "global_step": 162337, "epoch": 1824, "lr": 3.495329911826424e-05}
{"train_loss": 0.06473532319068909, "global_step": 162338, "epoch": 1824, "lr": 3.495274625638353e-05}
{"train_loss": 0.11969062685966492, "global_step": 162339, "epoch": 1824, "lr": 3.495219339652576e-05}
{"train_loss": 0.10076000541448593, "global_step": 162340, "epoch": 1824, "lr": 3.495164053869096e-05}
{"train_loss": 0.12058866024017334, "global_step": 162341, "epoch": 1824, "lr": 3.495108768287925e-05}
{"train_loss": 0.09185061603784561, "global_step": 162342, "epoch": 1824, "lr": 3.4950534829090665e-05}
{"train_loss": 0.09440886229276657, "global_step": 162343, "epoch": 1824, "lr": 3.494998197732532e-05}
{"train_loss": 0.15349681675434113, "global_step": 162344, "epoch": 1824, "lr": 3.494942912758325e-05}
{"train_loss": 0.06991089880466461, "global_step": 162345, "epoch": 1824, "lr": 3.494887627986455e-05}
{"train_loss": 0.061057448387145996, "global_step": 162346, "epoch": 1824, "lr": 3.4948323434169283e-05}
{"train_loss": 0.050360001623630524, "global_step": 162347, "epoch": 1824, "lr": 3.494777059049754e-05}
{"train_loss": 0.08029327541589737, "global_step": 162348, "epoch": 1824, "lr": 3.4947217748849384e-05}
{"train_loss": 0.11942823976278305, "global_step": 162349, "epoch": 1824, "lr": 3.4946664909224895e-05}
{"train_loss": 0.048336125910282135, "global_step": 162350, "epoch": 1824, "lr": 3.4946112071624124e-05}
{"train_loss": 0.10225705057382584, "global_step": 162351, "epoch": 1824, "lr": 3.494555923604719e-05}
{"train_loss": 0.07890910655260086, "global_step": 162352, "epoch": 1824, "lr": 3.494500640249413e-05}
{"train_loss": 0.10645151883363724, "global_step": 162353, "epoch": 1824, "lr": 3.494445357096504e-05}
{"train_loss": 0.061567552387714386, "global_step": 162354, "epoch": 1824, "lr": 3.4943900741459956e-05}
{"train_loss": 0.07773500680923462, "global_step": 162355, "epoch": 1824, "lr": 3.4943347913979e-05}
{"train_loss": 0.10229851305484772, "global_step": 162356, "epoch": 1824, "lr": 3.494279508852223e-05}
{"train_loss": 0.0691954493522644, "global_step": 162357, "epoch": 1824, "lr": 3.494224226508971e-05}
{"train_loss": 0.14947378635406494, "global_step": 162358, "epoch": 1824, "lr": 3.494168944368154e-05}
{"train_loss": 0.0672341138124466, "global_step": 162359, "epoch": 1824, "lr": 3.494113662429774e-05}
{"train_loss": 0.08829227834939957, "global_step": 162360, "epoch": 1824, "lr": 3.494058380693846e-05}
{"train_loss": 0.07897942513227463, "global_step": 162361, "epoch": 1824, "lr": 3.4940030991603714e-05}
{"train_loss": 0.098027803003788, "global_step": 162362, "epoch": 1824, "lr": 3.493947817829361e-05}
{"train_loss": 0.1769367754459381, "global_step": 162363, "epoch": 1824, "lr": 3.493892536700819e-05}
{"train_loss": 0.10731486976146698, "global_step": 162364, "epoch": 1824, "lr": 3.4938372557747575e-05}
{"train_loss": 0.08981316536664963, "global_step": 162365, "epoch": 1824, "lr": 3.4937819750511796e-05}
{"train_loss": 0.044723957777023315, "global_step": 162366, "epoch": 1824, "lr": 3.493726694530095e-05}
{"train_loss": 0.16308459639549255, "global_step": 162367, "epoch": 1824, "lr": 3.493671414211509e-05}
{"train_loss": 0.13323336839675903, "global_step": 162368, "epoch": 1824, "lr": 3.493616134095432e-05}
{"train_loss": 0.11252500116825104, "global_step": 162369, "epoch": 1824, "lr": 3.493560854181869e-05}
{"train_loss": 0.101804718375206, "global_step": 162370, "epoch": 1824, "lr": 3.49350557447083e-05}
{"train_loss": 0.08996450155973434, "global_step": 162371, "epoch": 1824, "lr": 3.493450294962319e-05}
{"train_loss": 0.12679851055145264, "global_step": 162372, "epoch": 1824, "lr": 3.493395015656347e-05}
{"train_loss": 0.1566585898399353, "global_step": 162373, "epoch": 1824, "lr": 3.493339736552917e-05}
{"train_loss": 0.11581210792064667, "global_step": 162374, "epoch": 1824, "lr": 3.49328445765204e-05}
{"train_loss": 0.07400978356599808, "global_step": 162375, "epoch": 1824, "lr": 3.493229178953724e-05}
{"train_loss": 0.15428569912910461, "global_step": 162376, "epoch": 1824, "lr": 3.4931739004579734e-05}
{"train_loss": 0.08826250582933426, "global_step": 162377, "epoch": 1824, "lr": 3.4931186221647994e-05}
{"train_loss": 0.11867770552635193, "global_step": 162378, "epoch": 1824, "lr": 3.4930633440742045e-05}
{"train_loss": 0.07602181285619736, "global_step": 162379, "epoch": 1824, "lr": 3.493008066186201e-05}
{"train_loss": 0.0964026004076004, "global_step": 162380, "epoch": 1824, "lr": 3.492952788500792e-05}
{"train_loss": 0.1417311280965805, "global_step": 162381, "epoch": 1824, "lr": 3.49289751101799e-05}
{"train_loss": 0.10185487568378448, "global_step": 162382, "epoch": 1824, "lr": 3.492842233737797e-05}
{"train_loss": 0.12442508339881897, "global_step": 162383, "epoch": 1824, "lr": 3.492786956660225e-05}
{"train_loss": 0.10243337601423264, "global_step": 162384, "epoch": 1824, "lr": 3.492731679785278e-05}
{"train_loss": 0.1234997808933258, "global_step": 162385, "epoch": 1824, "lr": 3.492676403112966e-05}
{"train_loss": 0.13166463375091553, "global_step": 162386, "epoch": 1824, "lr": 3.492621126643295e-05}
{"train_loss": 0.08489860594272614, "global_step": 162387, "epoch": 1824, "lr": 3.4925658503762723e-05}
{"train_loss": 0.10565172880887985, "global_step": 162388, "epoch": 1824, "lr": 3.4925105743119056e-05}
{"train_loss": 0.13719476759433746, "global_step": 162389, "epoch": 1824, "lr": 3.4924552984502034e-05}
{"train_loss": 0.04925410449504852, "global_step": 162390, "epoch": 1824, "lr": 3.492400022791171e-05}
{"train_loss": 0.11442367732524872, "global_step": 162391, "epoch": 1824, "lr": 3.492344747334816e-05}
{"train_loss": 0.054958976805210114, "global_step": 162392, "epoch": 1824, "lr": 3.492289472081149e-05}
{"train_loss": 0.07505957782268524, "global_step": 162393, "epoch": 1824, "lr": 3.492234197030174e-05}
{"train_loss": 0.16581808030605316, "global_step": 162394, "epoch": 1824, "lr": 3.492178922181902e-05}
{"train_loss": 0.16957660019397736, "global_step": 162395, "epoch": 1824, "lr": 3.492123647536335e-05}
{"train_loss": 0.10717403888702393, "global_step": 162396, "epoch": 1824, "lr": 3.4920683730934864e-05}
{"train_loss": 0.06931126117706299, "global_step": 162397, "epoch": 1824, "lr": 3.492013098853359e-05}
{"train_loss": 0.06967902183532715, "global_step": 162398, "epoch": 1824, "lr": 3.491957824815964e-05}
{"train_loss": 0.08495578169822693, "global_step": 162399, "epoch": 1824, "lr": 3.491902550981304e-05}
{"train_loss": 0.10550849884748459, "global_step": 162400, "epoch": 1824, "lr": 3.4918472773493924e-05}
{"train_loss": 0.0990644246339798, "global_step": 162401, "epoch": 1824, "lr": 3.4917920039202306e-05}
{"train_loss": 0.0609886571764946, "global_step": 162402, "epoch": 1824, "lr": 3.491736730693832e-05}
{"train_loss": 0.08575204014778137, "global_step": 162403, "epoch": 1824, "lr": 3.491681457670199e-05}
{"train_loss": 0.037328679114580154, "global_step": 162404, "epoch": 1824, "lr": 3.491626184849341e-05}
{"train_loss": 0.07001976668834686, "global_step": 162405, "epoch": 1824, "lr": 3.491570912231267e-05}
{"train_loss": 0.10112974792718887, "global_step": 162406, "epoch": 1824, "lr": 3.491515639815982e-05}
{"train_loss": 0.10419584065675735, "global_step": 162407, "epoch": 1824, "lr": 3.491460367603494e-05}
{"train_loss": 0.08098728209733963, "global_step": 162408, "epoch": 1824, "lr": 3.4914050955938116e-05}
{"train_loss": 0.0919058695435524, "global_step": 162409, "epoch": 1824, "lr": 3.49134982378694e-05}
{"train_loss": 0.026275906711816788, "global_step": 162410, "epoch": 1824, "lr": 3.4912945521828876e-05}
{"train_loss": 0.1293088048696518, "global_step": 162411, "epoch": 1824, "lr": 3.491239280781664e-05}
{"train_loss": 0.06638125330209732, "global_step": 162412, "epoch": 1824, "lr": 3.4911840095832735e-05}
{"train_loss": 0.10235510021448135, "global_step": 162413, "epoch": 1824, "lr": 3.491128738587727e-05}
{"train_loss": 0.10093840211629868, "global_step": 162414, "epoch": 1824, "lr": 3.4910734677950274e-05}
{"train_loss": 0.12479998916387558, "global_step": 162415, "epoch": 1824, "lr": 3.491018197205187e-05}
{"train_loss": 0.10673516988754272, "global_step": 162416, "epoch": 1824, "lr": 3.4909629268182084e-05}
{"train_loss": 0.20653574168682098, "global_step": 162417, "epoch": 1824, "lr": 3.4909076566341037e-05}
{"train_loss": 0.08805663883686066, "global_step": 162418, "epoch": 1824, "lr": 3.490852386652876e-05}
{"train_loss": 0.08050855249166489, "global_step": 162419, "epoch": 1824, "lr": 3.4907971168745374e-05}
{"train_loss": 0.14297479391098022, "global_step": 162420, "epoch": 1824, "lr": 3.49074184729909e-05}
{"train_loss": 0.1116061583161354, "global_step": 162421, "epoch": 1824, "lr": 3.490686577926546e-05}
{"train_loss": 0.0987459346652031, "global_step": 162422, "epoch": 1824, "lr": 3.4906313087569096e-05}
{"train_loss": 0.06489796936511993, "global_step": 162423, "epoch": 1824, "lr": 3.49057603979019e-05}
{"train_loss": 0.0993910168161553, "global_step": 162424, "epoch": 1824, "lr": 3.4905207710263945e-05, "val_loss": 6.791927337646484}
{"train_loss": 0.06866544485092163, "global_step": 162425, "epoch": 1825, "lr": 3.490465502465529e-05}
{"train_loss": 0.06185056269168854, "global_step": 162426, "epoch": 1825, "lr": 3.490410234107602e-05}
{"train_loss": 0.08022210747003555, "global_step": 162427, "epoch": 1825, "lr": 3.49035496595262e-05}
{"train_loss": 0.07187145203351974, "global_step": 162428, "epoch": 1825, "lr": 3.490299698000594e-05}
{"train_loss": 0.08612211048603058, "global_step": 162429, "epoch": 1825, "lr": 3.490244430251527e-05}
{"train_loss": 0.0902470052242279, "global_step": 162430, "epoch": 1825, "lr": 3.49018916270543e-05}
{"train_loss": 0.14093872904777527, "global_step": 162431, "epoch": 1825, "lr": 3.4901338953623066e-05}
{"train_loss": 0.13646069169044495, "global_step": 162432, "epoch": 1825, "lr": 3.490078628222167e-05}
{"train_loss": 0.15129879117012024, "global_step": 162433, "epoch": 1825, "lr": 3.4900233612850176e-05}
{"train_loss": 0.09667273610830307, "global_step": 162434, "epoch": 1825, "lr": 3.489968094550868e-05}
{"train_loss": 0.07322835177183151, "global_step": 162435, "epoch": 1825, "lr": 3.4899128280197214e-05}
{"train_loss": 0.17453943192958832, "global_step": 162436, "epoch": 1825, "lr": 3.48985756169159e-05}
{"train_loss": 0.13448773324489594, "global_step": 162437, "epoch": 1825, "lr": 3.489802295566476e-05}
{"train_loss": 0.08167082071304321, "global_step": 162438, "epoch": 1825, "lr": 3.489747029644392e-05}
{"train_loss": 0.07062935829162598, "global_step": 162439, "epoch": 1825, "lr": 3.4896917639253415e-05}
{"train_loss": 0.09437300264835358, "global_step": 162440, "epoch": 1825, "lr": 3.489636498409335e-05}
{"train_loss": 0.08295327425003052, "global_step": 162441, "epoch": 1825, "lr": 3.489581233096377e-05}
{"train_loss": 0.09574844688177109, "global_step": 162442, "epoch": 1825, "lr": 3.489525967986478e-05}
{"train_loss": 0.10940729081630707, "global_step": 162443, "epoch": 1825, "lr": 3.4894707030796426e-05}
{"train_loss": 0.08731677383184433, "global_step": 162444, "epoch": 1825, "lr": 3.4894154383758796e-05}
{"train_loss": 0.15566763281822205, "global_step": 162445, "epoch": 1825, "lr": 3.4893601738751965e-05}
{"train_loss": 0.1016051322221756, "global_step": 162446, "epoch": 1825, "lr": 3.4893049095776e-05}
{"train_loss": 0.0898350179195404, "global_step": 162447, "epoch": 1825, "lr": 3.489249645483099e-05}
{"train_loss": 0.15100599825382233, "global_step": 162448, "epoch": 1825, "lr": 3.489194381591699e-05}
{"train_loss": 0.12338828295469284, "global_step": 162449, "epoch": 1825, "lr": 3.489139117903409e-05}
{"train_loss": 0.12324665486812592, "global_step": 162450, "epoch": 1825, "lr": 3.489083854418236e-05}
{"train_loss": 0.11222042143344879, "global_step": 162451, "epoch": 1825, "lr": 3.489028591136187e-05}
{"train_loss": 0.0978136956691742, "global_step": 162452, "epoch": 1825, "lr": 3.488973328057269e-05}
{"train_loss": 0.11857680976390839, "global_step": 162453, "epoch": 1825, "lr": 3.488918065181492e-05}
{"train_loss": 0.05855584517121315, "global_step": 162454, "epoch": 1825, "lr": 3.4888628025088594e-05}
{"train_loss": 0.11165563017129898, "global_step": 162455, "epoch": 1825, "lr": 3.488807540039383e-05}
{"train_loss": 0.13051408529281616, "global_step": 162456, "epoch": 1825, "lr": 3.488752277773065e-05}
{"train_loss": 0.11241422593593597, "global_step": 162457, "epoch": 1825, "lr": 3.488697015709919e-05}
{"train_loss": 0.11119624972343445, "global_step": 162458, "epoch": 1825, "lr": 3.4886417538499473e-05}
{"train_loss": 0.09099376201629639, "global_step": 162459, "epoch": 1825, "lr": 3.4885864921931604e-05}
{"train_loss": 0.07104158401489258, "global_step": 162460, "epoch": 1825, "lr": 3.488531230739564e-05}
{"train_loss": 0.09195715934038162, "global_step": 162461, "epoch": 1825, "lr": 3.4884759694891664e-05}
{"train_loss": 0.07413005828857422, "global_step": 162462, "epoch": 1825, "lr": 3.488420708441975e-05}
{"train_loss": 0.09866807609796524, "global_step": 162463, "epoch": 1825, "lr": 3.488365447597996e-05}
{"train_loss": 0.07457654178142548, "global_step": 162464, "epoch": 1825, "lr": 3.4883101869572386e-05}
{"train_loss": 0.04139357805252075, "global_step": 162465, "epoch": 1825, "lr": 3.488254926519708e-05}
{"train_loss": 0.07097189128398895, "global_step": 162466, "epoch": 1825, "lr": 3.4881996662854156e-05}
{"train_loss": 0.14820919930934906, "global_step": 162467, "epoch": 1825, "lr": 3.488144406254365e-05}
{"train_loss": 0.05452781543135643, "global_step": 162468, "epoch": 1825, "lr": 3.488089146426565e-05}
{"train_loss": 0.11193013191223145, "global_step": 162469, "epoch": 1825, "lr": 3.4880338868020224e-05}
{"train_loss": 0.07338552922010422, "global_step": 162470, "epoch": 1825, "lr": 3.487978627380747e-05}
{"train_loss": 0.05332318693399429, "global_step": 162471, "epoch": 1825, "lr": 3.487923368162742e-05}
{"train_loss": 0.05350572243332863, "global_step": 162472, "epoch": 1825, "lr": 3.4878681091480205e-05}
{"train_loss": 0.056032683700323105, "global_step": 162473, "epoch": 1825, "lr": 3.487812850336584e-05}
{"train_loss": 0.11363934725522995, "global_step": 162474, "epoch": 1825, "lr": 3.487757591728444e-05}
{"train_loss": 0.08400675654411316, "global_step": 162475, "epoch": 1825, "lr": 3.4877023333236044e-05}
{"train_loss": 0.09970787912607193, "global_step": 162476, "epoch": 1825, "lr": 3.487647075122078e-05}
{"train_loss": 0.15248577296733856, "global_step": 162477, "epoch": 1825, "lr": 3.4875918171238666e-05}
{"train_loss": 0.05499732866883278, "global_step": 162478, "epoch": 1825, "lr": 3.4875365593289814e-05}
{"train_loss": 0.10546580702066422, "global_step": 162479, "epoch": 1825, "lr": 3.487481301737427e-05}
{"train_loss": 0.08191349357366562, "global_step": 162480, "epoch": 1825, "lr": 3.487426044349214e-05}
{"train_loss": 0.12085938453674316, "global_step": 162481, "epoch": 1825, "lr": 3.487370787164348e-05}
{"train_loss": 0.074005126953125, "global_step": 162482, "epoch": 1825, "lr": 3.4873155301828346e-05}
{"train_loss": 0.09039797633886337, "global_step": 162483, "epoch": 1825, "lr": 3.487260273404686e-05}
{"train_loss": 0.05704345554113388, "global_step": 162484, "epoch": 1825, "lr": 3.487205016829904e-05}
{"train_loss": 0.1151122897863388, "global_step": 162485, "epoch": 1825, "lr": 3.487149760458501e-05}
{"train_loss": 0.14235153794288635, "global_step": 162486, "epoch": 1825, "lr": 3.487094504290481e-05}
{"train_loss": 0.08692332357168198, "global_step": 162487, "epoch": 1825, "lr": 3.487039248325854e-05}
{"train_loss": 0.08661476522684097, "global_step": 162488, "epoch": 1825, "lr": 3.486983992564624e-05}
{"train_loss": 0.04015336558222771, "global_step": 162489, "epoch": 1825, "lr": 3.486928737006803e-05}
{"train_loss": 0.14099222421646118, "global_step": 162490, "epoch": 1825, "lr": 3.486873481652394e-05}
{"train_loss": 0.09612169861793518, "global_step": 162491, "epoch": 1825, "lr": 3.486818226501408e-05}
{"train_loss": 0.08031049370765686, "global_step": 162492, "epoch": 1825, "lr": 3.48676297155385e-05}
{"train_loss": 0.06040677800774574, "global_step": 162493, "epoch": 1825, "lr": 3.486707716809729e-05}
{"train_loss": 0.09387458115816116, "global_step": 162494, "epoch": 1825, "lr": 3.48665246226905e-05}
{"train_loss": 0.09036742895841599, "global_step": 162495, "epoch": 1825, "lr": 3.4865972079318243e-05}
{"train_loss": 0.046205852180719376, "global_step": 162496, "epoch": 1825, "lr": 3.486541953798055e-05}
{"train_loss": 0.11983753740787506, "global_step": 162497, "epoch": 1825, "lr": 3.486486699867754e-05}
{"train_loss": 0.11039793491363525, "global_step": 162498, "epoch": 1825, "lr": 3.486431446140924e-05}
{"train_loss": 0.11488214880228043, "global_step": 162499, "epoch": 1825, "lr": 3.486376192617577e-05}
{"train_loss": 0.0874161496758461, "global_step": 162500, "epoch": 1825, "lr": 3.486320939297717e-05}
{"train_loss": 0.1418127417564392, "global_step": 162501, "epoch": 1825, "lr": 3.4862656861813524e-05}
{"train_loss": 0.11493881046772003, "global_step": 162502, "epoch": 1825, "lr": 3.4862104332684926e-05}
{"train_loss": 0.085903100669384, "global_step": 162503, "epoch": 1825, "lr": 3.486155180559141e-05}
{"train_loss": 0.08026523888111115, "global_step": 162504, "epoch": 1825, "lr": 3.4860999280533105e-05}
{"train_loss": 0.02989831008017063, "global_step": 162505, "epoch": 1825, "lr": 3.486044675751002e-05}
{"train_loss": 0.08091440051794052, "global_step": 162506, "epoch": 1825, "lr": 3.485989423652229e-05}
{"train_loss": 0.08050012588500977, "global_step": 162507, "epoch": 1825, "lr": 3.485934171756994e-05}
{"train_loss": 0.16660012304782867, "global_step": 162508, "epoch": 1825, "lr": 3.485878920065309e-05}
{"train_loss": 0.06353077292442322, "global_step": 162509, "epoch": 1825, "lr": 3.4858236685771775e-05}
{"train_loss": 0.09974168241024017, "global_step": 162510, "epoch": 1825, "lr": 3.48576841729261e-05}
{"train_loss": 0.1461414396762848, "global_step": 162511, "epoch": 1825, "lr": 3.485713166211611e-05}
{"train_loss": 0.0980103388428688, "global_step": 162512, "epoch": 1825, "lr": 3.485657915334191e-05}
{"train_loss": 0.09623182617295324, "global_step": 162513, "epoch": 1825, "lr": 3.4856026646603536e-05, "val_loss": 6.734996318817139, "train_action_mse_error": 18.92542266845703}
{"train_loss": 0.09373819828033447, "global_step": 162514, "epoch": 1826, "lr": 3.48554741419011e-05}
{"train_loss": 0.09530921280384064, "global_step": 162515, "epoch": 1826, "lr": 3.485492163923465e-05}
{"train_loss": 0.17514659464359283, "global_step": 162516, "epoch": 1826, "lr": 3.4854369138604295e-05}
{"train_loss": 0.07054536789655685, "global_step": 162517, "epoch": 1826, "lr": 3.485381664001005e-05}
{"train_loss": 0.13186830282211304, "global_step": 162518, "epoch": 1826, "lr": 3.4853264143452045e-05}
{"train_loss": 0.09827346354722977, "global_step": 162519, "epoch": 1826, "lr": 3.485271164893035e-05}
{"train_loss": 0.09638107568025589, "global_step": 162520, "epoch": 1826, "lr": 3.485215915644499e-05}
{"train_loss": 0.07113087922334671, "global_step": 162521, "epoch": 1826, "lr": 3.48516066659961e-05}
{"train_loss": 0.07074595242738724, "global_step": 162522, "epoch": 1826, "lr": 3.4851054177583707e-05}
{"train_loss": 0.11349602788686752, "global_step": 162523, "epoch": 1826, "lr": 3.4850501691207925e-05}
{"train_loss": 0.09522798657417297, "global_step": 162524, "epoch": 1826, "lr": 3.484994920686879e-05}
{"train_loss": 0.08717552572488785, "global_step": 162525, "epoch": 1826, "lr": 3.484939672456641e-05}
{"train_loss": 0.14033271372318268, "global_step": 162526, "epoch": 1826, "lr": 3.484884424430083e-05}
{"train_loss": 0.05713437870144844, "global_step": 162527, "epoch": 1826, "lr": 3.484829176607215e-05}
{"train_loss": 0.07930092513561249, "global_step": 162528, "epoch": 1826, "lr": 3.484773928988041e-05}
{"train_loss": 0.0839981958270073, "global_step": 162529, "epoch": 1826, "lr": 3.484718681572574e-05}
{"train_loss": 0.0716060921549797, "global_step": 162530, "epoch": 1826, "lr": 3.484663434360815e-05}
{"train_loss": 0.06657681614160538, "global_step": 162531, "epoch": 1826, "lr": 3.484608187352777e-05}
{"train_loss": 0.08879882842302322, "global_step": 162532, "epoch": 1826, "lr": 3.4845529405484636e-05}
{"train_loss": 0.0813262015581131, "global_step": 162533, "epoch": 1826, "lr": 3.4844976939478845e-05}
{"train_loss": 0.05343944579362869, "global_step": 162534, "epoch": 1826, "lr": 3.484442447551045e-05}
{"train_loss": 0.048756323754787445, "global_step": 162535, "epoch": 1826, "lr": 3.484387201357955e-05}
{"train_loss": 0.10619660466909409, "global_step": 162536, "epoch": 1826, "lr": 3.484331955368619e-05}
{"train_loss": 0.07257209718227386, "global_step": 162537, "epoch": 1826, "lr": 3.4842767095830476e-05}
{"train_loss": 0.08513075858354568, "global_step": 162538, "epoch": 1826, "lr": 3.484221464001247e-05}
{"train_loss": 0.06299116462469101, "global_step": 162539, "epoch": 1826, "lr": 3.4841662186232226e-05}
{"train_loss": 0.08583710342645645, "global_step": 162540, "epoch": 1826, "lr": 3.484110973448986e-05}
{"train_loss": 0.08817664533853531, "global_step": 162541, "epoch": 1826, "lr": 3.484055728478539e-05}
{"train_loss": 0.053315240889787674, "global_step": 162542, "epoch": 1826, "lr": 3.484000483711896e-05}
{"train_loss": 0.048564959317445755, "global_step": 162543, "epoch": 1826, "lr": 3.483945239149057e-05}
{"train_loss": 0.08589474856853485, "global_step": 162544, "epoch": 1826, "lr": 3.483889994790036e-05}
{"train_loss": 0.14979872107505798, "global_step": 162545, "epoch": 1826, "lr": 3.483834750634836e-05}
{"train_loss": 0.07855286449193954, "global_step": 162546, "epoch": 1826, "lr": 3.483779506683467e-05}
{"train_loss": 0.08849924057722092, "global_step": 162547, "epoch": 1826, "lr": 3.483724262935934e-05}
{"train_loss": 0.08331528306007385, "global_step": 162548, "epoch": 1826, "lr": 3.4836690193922465e-05}
{"train_loss": 0.14300400018692017, "global_step": 162549, "epoch": 1826, "lr": 3.48361377605241e-05}
{"train_loss": 0.11311689019203186, "global_step": 162550, "epoch": 1826, "lr": 3.483558532916436e-05}
{"train_loss": 0.04925492778420448, "global_step": 162551, "epoch": 1826, "lr": 3.483503289984326e-05}
{"train_loss": 0.11638668924570084, "global_step": 162552, "epoch": 1826, "lr": 3.483448047256093e-05}
{"train_loss": 0.07317622750997543, "global_step": 162553, "epoch": 1826, "lr": 3.483392804731739e-05}
{"train_loss": 0.07114697247743607, "global_step": 162554, "epoch": 1826, "lr": 3.4833375624112764e-05}
{"train_loss": 0.09106311202049255, "global_step": 162555, "epoch": 1826, "lr": 3.48328232029471e-05}
{"train_loss": 0.17842288315296173, "global_step": 162556, "epoch": 1826, "lr": 3.4832270783820484e-05}
{"train_loss": 0.074091337621212, "global_step": 162557, "epoch": 1826, "lr": 3.483171836673299e-05}
{"train_loss": 0.07422975450754166, "global_step": 162558, "epoch": 1826, "lr": 3.483116595168467e-05}
{"train_loss": 0.028643149882555008, "global_step": 162559, "epoch": 1826, "lr": 3.4830613538675624e-05}
{"train_loss": 0.08049654960632324, "global_step": 162560, "epoch": 1826, "lr": 3.4830061127705914e-05}
{"train_loss": 0.05126851052045822, "global_step": 162561, "epoch": 1826, "lr": 3.482950871877563e-05}
{"train_loss": 0.13389042019844055, "global_step": 162562, "epoch": 1826, "lr": 3.482895631188481e-05}
{"train_loss": 0.06606419384479523, "global_step": 162563, "epoch": 1826, "lr": 3.482840390703357e-05}
{"train_loss": 0.18542729318141937, "global_step": 162564, "epoch": 1826, "lr": 3.4827851504221954e-05}
{"train_loss": 0.07925773411989212, "global_step": 162565, "epoch": 1826, "lr": 3.482729910345007e-05}
{"train_loss": 0.06477741152048111, "global_step": 162566, "epoch": 1826, "lr": 3.482674670471794e-05}
{"train_loss": 0.11006876081228256, "global_step": 162567, "epoch": 1826, "lr": 3.482619430802569e-05}
{"train_loss": 0.04634585976600647, "global_step": 162568, "epoch": 1826, "lr": 3.482564191337337e-05}
{"train_loss": 0.038033001124858856, "global_step": 162569, "epoch": 1826, "lr": 3.482508952076106e-05}
{"train_loss": 0.10544297099113464, "global_step": 162570, "epoch": 1826, "lr": 3.4824537130188815e-05}
{"train_loss": 0.05613555759191513, "global_step": 162571, "epoch": 1826, "lr": 3.4823984741656757e-05}
{"train_loss": 0.15503600239753723, "global_step": 162572, "epoch": 1826, "lr": 3.4823432355164896e-05}
{"train_loss": 0.08453868329524994, "global_step": 162573, "epoch": 1826, "lr": 3.482287997071335e-05}
{"train_loss": 0.0781559944152832, "global_step": 162574, "epoch": 1826, "lr": 3.4822327588302194e-05}
{"train_loss": 0.07748134434223175, "global_step": 162575, "epoch": 1826, "lr": 3.482177520793148e-05}
{"train_loss": 0.05465542525053024, "global_step": 162576, "epoch": 1826, "lr": 3.4821222829601305e-05}
{"train_loss": 0.06527899205684662, "global_step": 162577, "epoch": 1826, "lr": 3.482067045331172e-05}
{"train_loss": 0.07806821167469025, "global_step": 162578, "epoch": 1826, "lr": 3.4820118079062826e-05}
{"train_loss": 0.05637246370315552, "global_step": 162579, "epoch": 1826, "lr": 3.4819565706854665e-05}
{"train_loss": 0.08856583386659622, "global_step": 162580, "epoch": 1826, "lr": 3.481901333668734e-05}
{"train_loss": 0.10533164441585541, "global_step": 162581, "epoch": 1826, "lr": 3.4818460968560894e-05}
{"train_loss": 0.10033882409334183, "global_step": 162582, "epoch": 1826, "lr": 3.4817908602475446e-05}
{"train_loss": 0.10431836545467377, "global_step": 162583, "epoch": 1826, "lr": 3.481735623843103e-05}
{"train_loss": 0.16402281820774078, "global_step": 162584, "epoch": 1826, "lr": 3.4816803876427745e-05}
{"train_loss": 0.07409188151359558, "global_step": 162585, "epoch": 1826, "lr": 3.4816251516465634e-05}
{"train_loss": 0.12785769999027252, "global_step": 162586, "epoch": 1826, "lr": 3.4815699158544825e-05}
{"train_loss": 0.1272117644548416, "global_step": 162587, "epoch": 1826, "lr": 3.481514680266533e-05}
{"train_loss": 0.09778320789337158, "global_step": 162588, "epoch": 1826, "lr": 3.481459444882728e-05}
{"train_loss": 0.05248773843050003, "global_step": 162589, "epoch": 1826, "lr": 3.4814042097030705e-05}
{"train_loss": 0.04351354390382767, "global_step": 162590, "epoch": 1826, "lr": 3.4813489747275707e-05}
{"train_loss": 0.15527381002902985, "global_step": 162591, "epoch": 1826, "lr": 3.481293739956234e-05}
{"train_loss": 0.1395452320575714, "global_step": 162592, "epoch": 1826, "lr": 3.4812385053890694e-05}
{"train_loss": 0.11477624624967575, "global_step": 162593, "epoch": 1826, "lr": 3.481183271026084e-05}
{"train_loss": 0.09102143347263336, "global_step": 162594, "epoch": 1826, "lr": 3.481128036867285e-05}
{"train_loss": 0.07698120176792145, "global_step": 162595, "epoch": 1826, "lr": 3.481072802912679e-05}
{"train_loss": 0.08117182552814484, "global_step": 162596, "epoch": 1826, "lr": 3.481017569162275e-05}
{"train_loss": 0.062354955822229385, "global_step": 162597, "epoch": 1826, "lr": 3.48096233561608e-05}
{"train_loss": 0.1522098332643509, "global_step": 162598, "epoch": 1826, "lr": 3.4809071022741004e-05}
{"train_loss": 0.06892521679401398, "global_step": 162599, "epoch": 1826, "lr": 3.480851869136345e-05}
{"train_loss": 0.1009465679526329, "global_step": 162600, "epoch": 1826, "lr": 3.480796636202819e-05}
{"train_loss": 0.09288091212511063, "global_step": 162601, "epoch": 1826, "lr": 3.480741403473534e-05}
{"train_loss": 0.09063559092497558, "global_step": 162602, "epoch": 1826, "lr": 3.4806861709484914e-05, "val_loss": 6.94919490814209}
{"train_loss": 0.06692380458116531, "global_step": 162603, "epoch": 1827, "lr": 3.480630938627705e-05}
{"train_loss": 0.11204041540622711, "global_step": 162604, "epoch": 1827, "lr": 3.4805757065111775e-05}
{"train_loss": 0.07305066287517548, "global_step": 162605, "epoch": 1827, "lr": 3.4805204745989195e-05}
{"train_loss": 0.08283570408821106, "global_step": 162606, "epoch": 1827, "lr": 3.4804652428909354e-05}
{"train_loss": 0.061488352715969086, "global_step": 162607, "epoch": 1827, "lr": 3.480410011387236e-05}
{"train_loss": 0.06595088541507721, "global_step": 162608, "epoch": 1827, "lr": 3.480354780087825e-05}
{"train_loss": 0.0509958378970623, "global_step": 162609, "epoch": 1827, "lr": 3.4802995489927114e-05}
{"train_loss": 0.08120083808898926, "global_step": 162610, "epoch": 1827, "lr": 3.480244318101906e-05}
{"train_loss": 0.1151311993598938, "global_step": 162611, "epoch": 1827, "lr": 3.480189087415411e-05}
{"train_loss": 0.038425229489803314, "global_step": 162612, "epoch": 1827, "lr": 3.480133856933237e-05}
{"train_loss": 0.0791541039943695, "global_step": 162613, "epoch": 1827, "lr": 3.4800786266553905e-05}
{"train_loss": 0.08393105119466782, "global_step": 162614, "epoch": 1827, "lr": 3.4800233965818784e-05}
{"train_loss": 0.0945122167468071, "global_step": 162615, "epoch": 1827, "lr": 3.4799681667127086e-05}
{"train_loss": 0.04523700475692749, "global_step": 162616, "epoch": 1827, "lr": 3.47991293704789e-05}
{"train_loss": 0.07315926253795624, "global_step": 162617, "epoch": 1827, "lr": 3.479857707587426e-05}
{"train_loss": 0.07384566217660904, "global_step": 162618, "epoch": 1827, "lr": 3.479802478331329e-05}
{"train_loss": 0.16803762316703796, "global_step": 162619, "epoch": 1827, "lr": 3.479747249279603e-05}
{"train_loss": 0.09839212894439697, "global_step": 162620, "epoch": 1827, "lr": 3.479692020432256e-05}
{"train_loss": 0.08507943153381348, "global_step": 162621, "epoch": 1827, "lr": 3.479636791789296e-05}
{"train_loss": 0.0871591567993164, "global_step": 162622, "epoch": 1827, "lr": 3.479581563350731e-05}
{"train_loss": 0.08457661420106888, "global_step": 162623, "epoch": 1827, "lr": 3.4795263351165665e-05}
{"train_loss": 0.05852130800485611, "global_step": 162624, "epoch": 1827, "lr": 3.479471107086813e-05}
{"train_loss": 0.08360946923494339, "global_step": 162625, "epoch": 1827, "lr": 3.479415879261474e-05}
{"train_loss": 0.10635501146316528, "global_step": 162626, "epoch": 1827, "lr": 3.4793606516405605e-05}
{"train_loss": 0.0621219202876091, "global_step": 162627, "epoch": 1827, "lr": 3.4793054242240776e-05}
{"train_loss": 0.0748598724603653, "global_step": 162628, "epoch": 1827, "lr": 3.4792501970120325e-05}
{"train_loss": 0.06425434350967407, "global_step": 162629, "epoch": 1827, "lr": 3.479194970004436e-05}
{"train_loss": 0.07799085974693298, "global_step": 162630, "epoch": 1827, "lr": 3.479139743201293e-05}
{"train_loss": 0.14040111005306244, "global_step": 162631, "epoch": 1827, "lr": 3.47908451660261e-05}
{"train_loss": 0.10299447178840637, "global_step": 162632, "epoch": 1827, "lr": 3.479029290208396e-05}
{"train_loss": 0.036314211785793304, "global_step": 162633, "epoch": 1827, "lr": 3.478974064018658e-05}
{"train_loss": 0.13980233669281006, "global_step": 162634, "epoch": 1827, "lr": 3.478918838033404e-05}
{"train_loss": 0.16779093444347382, "global_step": 162635, "epoch": 1827, "lr": 3.4788636122526405e-05}
{"train_loss": 0.077601358294487, "global_step": 162636, "epoch": 1827, "lr": 3.478808386676374e-05}
{"train_loss": 0.08072824031114578, "global_step": 162637, "epoch": 1827, "lr": 3.478753161304616e-05}
{"train_loss": 0.08695860952138901, "global_step": 162638, "epoch": 1827, "lr": 3.478697936137367e-05}
{"train_loss": 0.0931534543633461, "global_step": 162639, "epoch": 1827, "lr": 3.4786427111746424e-05}
{"train_loss": 0.12366802245378494, "global_step": 162640, "epoch": 1827, "lr": 3.478587486416442e-05}
{"train_loss": 0.17956747114658356, "global_step": 162641, "epoch": 1827, "lr": 3.47853226186278e-05}
{"train_loss": 0.06616540998220444, "global_step": 162642, "epoch": 1827, "lr": 3.4784770375136585e-05}
{"train_loss": 0.05672222003340721, "global_step": 162643, "epoch": 1827, "lr": 3.4784218133690904e-05}
{"train_loss": 0.12443729490041733, "global_step": 162644, "epoch": 1827, "lr": 3.478366589429077e-05}
{"train_loss": 0.1183871179819107, "global_step": 162645, "epoch": 1827, "lr": 3.478311365693628e-05}
{"train_loss": 0.16519612073898315, "global_step": 162646, "epoch": 1827, "lr": 3.4782561421627544e-05}
{"train_loss": 0.09281367808580399, "global_step": 162647, "epoch": 1827, "lr": 3.478200918836459e-05}
{"train_loss": 0.0615641288459301, "global_step": 162648, "epoch": 1827, "lr": 3.478145695714752e-05}
{"train_loss": 0.08169611543416977, "global_step": 162649, "epoch": 1827, "lr": 3.478090472797639e-05}
{"train_loss": 0.08985244482755661, "global_step": 162650, "epoch": 1827, "lr": 3.4780352500851285e-05}
{"train_loss": 0.03275587782263756, "global_step": 162651, "epoch": 1827, "lr": 3.477980027577228e-05}
{"train_loss": 0.16509197652339935, "global_step": 162652, "epoch": 1827, "lr": 3.477924805273944e-05}
{"train_loss": 0.08617259562015533, "global_step": 162653, "epoch": 1827, "lr": 3.477869583175285e-05}
{"train_loss": 0.05623074248433113, "global_step": 162654, "epoch": 1827, "lr": 3.477814361281258e-05}
{"train_loss": 0.08839619904756546, "global_step": 162655, "epoch": 1827, "lr": 3.477759139591868e-05}
{"train_loss": 0.11071133613586426, "global_step": 162656, "epoch": 1827, "lr": 3.477703918107128e-05}
{"train_loss": 0.06272780895233154, "global_step": 162657, "epoch": 1827, "lr": 3.4776486968270396e-05}
{"train_loss": 0.06893344968557358, "global_step": 162658, "epoch": 1827, "lr": 3.477593475751615e-05}
{"train_loss": 0.09403389692306519, "global_step": 162659, "epoch": 1827, "lr": 3.4775382548808574e-05}
{"train_loss": 0.05923215299844742, "global_step": 162660, "epoch": 1827, "lr": 3.477483034214778e-05}
{"train_loss": 0.04129248112440109, "global_step": 162661, "epoch": 1827, "lr": 3.47742781375338e-05}
{"train_loss": 0.15301203727722168, "global_step": 162662, "epoch": 1827, "lr": 3.477372593496676e-05}
{"train_loss": 0.13690191507339478, "global_step": 162663, "epoch": 1827, "lr": 3.4773173734446684e-05}
{"train_loss": 0.09009411185979843, "global_step": 162664, "epoch": 1827, "lr": 3.477262153597367e-05}
{"train_loss": 0.11532585322856903, "global_step": 162665, "epoch": 1827, "lr": 3.477206933954781e-05}
{"train_loss": 0.11346418410539627, "global_step": 162666, "epoch": 1827, "lr": 3.477151714516914e-05}
{"train_loss": 0.09336981177330017, "global_step": 162667, "epoch": 1827, "lr": 3.477096495283777e-05}
{"train_loss": 0.06372558325529099, "global_step": 162668, "epoch": 1827, "lr": 3.477041276255376e-05}
{"train_loss": 0.08332893997430801, "global_step": 162669, "epoch": 1827, "lr": 3.476986057431717e-05}
{"train_loss": 0.08012162894010544, "global_step": 162670, "epoch": 1827, "lr": 3.476930838812809e-05}
{"train_loss": 0.1029975637793541, "global_step": 162671, "epoch": 1827, "lr": 3.47687562039866e-05}
{"train_loss": 0.10656310617923737, "global_step": 162672, "epoch": 1827, "lr": 3.4768204021892754e-05}
{"train_loss": 0.10966029763221741, "global_step": 162673, "epoch": 1827, "lr": 3.476765184184665e-05}
{"train_loss": 0.05542204901576042, "global_step": 162674, "epoch": 1827, "lr": 3.476709966384833e-05}
{"train_loss": 0.07975052297115326, "global_step": 162675, "epoch": 1827, "lr": 3.4766547487897905e-05}
{"train_loss": 0.07214019447565079, "global_step": 162676, "epoch": 1827, "lr": 3.476599531399542e-05}
{"train_loss": 0.06279724091291428, "global_step": 162677, "epoch": 1827, "lr": 3.4765443142140973e-05}
{"train_loss": 0.1267712116241455, "global_step": 162678, "epoch": 1827, "lr": 3.476489097233461e-05}
{"train_loss": 0.062036290764808655, "global_step": 162679, "epoch": 1827, "lr": 3.476433880457644e-05}
{"train_loss": 0.15256546437740326, "global_step": 162680, "epoch": 1827, "lr": 3.4763786638866505e-05}
{"train_loss": 0.10680445283651352, "global_step": 162681, "epoch": 1827, "lr": 3.476323447520489e-05}
{"train_loss": 0.107745461165905, "global_step": 162682, "epoch": 1827, "lr": 3.476268231359169e-05}
{"train_loss": 0.14710800349712372, "global_step": 162683, "epoch": 1827, "lr": 3.476213015402694e-05}
{"train_loss": 0.07789978384971619, "global_step": 162684, "epoch": 1827, "lr": 3.4761577996510755e-05}
{"train_loss": 0.10762175917625427, "global_step": 162685, "epoch": 1827, "lr": 3.4761025841043174e-05}
{"train_loss": 0.141829252243042, "global_step": 162686, "epoch": 1827, "lr": 3.4760473687624314e-05}
{"train_loss": 0.07981106638908386, "global_step": 162687, "epoch": 1827, "lr": 3.4759921536254204e-05}
{"train_loss": 0.08763950318098068, "global_step": 162688, "epoch": 1827, "lr": 3.475936938693294e-05}
{"train_loss": 0.10679155588150024, "global_step": 162689, "epoch": 1827, "lr": 3.475881723966059e-05}
{"train_loss": 0.09284330904483795, "global_step": 162690, "epoch": 1827, "lr": 3.4758265094437235e-05}
{"train_loss": 0.09240580584560887, "global_step": 162691, "epoch": 1827, "lr": 3.475771295126294e-05, "val_loss": 6.847858905792236}
{"train_loss": 0.08052351325750351, "global_step": 162692, "epoch": 1828, "lr": 3.47571608101378e-05}
{"train_loss": 0.09004680067300797, "global_step": 162693, "epoch": 1828, "lr": 3.475660867106185e-05}
{"train_loss": 0.1312905251979828, "global_step": 162694, "epoch": 1828, "lr": 3.475605653403521e-05}
{"train_loss": 0.06772720813751221, "global_step": 162695, "epoch": 1828, "lr": 3.475550439905792e-05}
{"train_loss": 0.10620291531085968, "global_step": 162696, "epoch": 1828, "lr": 3.475495226613007e-05}
{"train_loss": 0.05201023444533348, "global_step": 162697, "epoch": 1828, "lr": 3.475440013525172e-05}
{"train_loss": 0.12623535096645355, "global_step": 162698, "epoch": 1828, "lr": 3.4753848006422973e-05}
{"train_loss": 0.08378472924232483, "global_step": 162699, "epoch": 1828, "lr": 3.475329587964387e-05}
{"train_loss": 0.09526220709085464, "global_step": 162700, "epoch": 1828, "lr": 3.47527437549145e-05}
{"train_loss": 0.10464818775653839, "global_step": 162701, "epoch": 1828, "lr": 3.475219163223495e-05}
{"train_loss": 0.10999366641044617, "global_step": 162702, "epoch": 1828, "lr": 3.475163951160527e-05}
{"train_loss": 0.09748465567827225, "global_step": 162703, "epoch": 1828, "lr": 3.475108739302557e-05}
{"train_loss": 0.16731800138950348, "global_step": 162704, "epoch": 1828, "lr": 3.475053527649587e-05}
{"train_loss": 0.057573102414608, "global_step": 162705, "epoch": 1828, "lr": 3.47499831620163e-05}
{"train_loss": 0.08258890360593796, "global_step": 162706, "epoch": 1828, "lr": 3.47494310495869e-05}
{"train_loss": 0.05300167202949524, "global_step": 162707, "epoch": 1828, "lr": 3.4748878939207745e-05}
{"train_loss": 0.1605011522769928, "global_step": 162708, "epoch": 1828, "lr": 3.474832683087892e-05}
{"train_loss": 0.13372983038425446, "global_step": 162709, "epoch": 1828, "lr": 3.474777472460051e-05}
{"train_loss": 0.09142036736011505, "global_step": 162710, "epoch": 1828, "lr": 3.474722262037256e-05}
{"train_loss": 0.0976729467511177, "global_step": 162711, "epoch": 1828, "lr": 3.474667051819517e-05}
{"train_loss": 0.07929053902626038, "global_step": 162712, "epoch": 1828, "lr": 3.47461184180684e-05}
{"train_loss": 0.07420384138822556, "global_step": 162713, "epoch": 1828, "lr": 3.474556631999234e-05}
{"train_loss": 0.15796200931072235, "global_step": 162714, "epoch": 1828, "lr": 3.474501422396703e-05}
{"train_loss": 0.10533329844474792, "global_step": 162715, "epoch": 1828, "lr": 3.474446212999258e-05}
{"train_loss": 0.11067145317792892, "global_step": 162716, "epoch": 1828, "lr": 3.474391003806905e-05}
{"train_loss": 0.08126524090766907, "global_step": 162717, "epoch": 1828, "lr": 3.474335794819652e-05}
{"train_loss": 0.08366292715072632, "global_step": 162718, "epoch": 1828, "lr": 3.474280586037504e-05}
{"train_loss": 0.05644598230719566, "global_step": 162719, "epoch": 1828, "lr": 3.474225377460471e-05}
{"train_loss": 0.0544007308781147, "global_step": 162720, "epoch": 1828, "lr": 3.474170169088562e-05}
{"train_loss": 0.10784199833869934, "global_step": 162721, "epoch": 1828, "lr": 3.4741149609217794e-05}
{"train_loss": 0.08396385610103607, "global_step": 162722, "epoch": 1828, "lr": 3.474059752960136e-05}
{"train_loss": 0.06707944720983505, "global_step": 162723, "epoch": 1828, "lr": 3.474004545203634e-05}
{"train_loss": 0.07522828876972198, "global_step": 162724, "epoch": 1828, "lr": 3.4739493376522865e-05}
{"train_loss": 0.09326060861349106, "global_step": 162725, "epoch": 1828, "lr": 3.4738941303060954e-05}
{"train_loss": 0.12561337649822235, "global_step": 162726, "epoch": 1828, "lr": 3.473838923165073e-05}
{"train_loss": 0.08544354140758514, "global_step": 162727, "epoch": 1828, "lr": 3.4737837162292223e-05}
{"train_loss": 0.08683789521455765, "global_step": 162728, "epoch": 1828, "lr": 3.473728509498554e-05}
{"train_loss": 0.08651752024888992, "global_step": 162729, "epoch": 1828, "lr": 3.4736733029730734e-05}
{"train_loss": 0.1041300892829895, "global_step": 162730, "epoch": 1828, "lr": 3.473618096652789e-05}
{"train_loss": 0.03889628127217293, "global_step": 162731, "epoch": 1828, "lr": 3.473562890537708e-05}
{"train_loss": 0.07344416528940201, "global_step": 162732, "epoch": 1828, "lr": 3.47350768462784e-05}
{"train_loss": 0.07528708875179291, "global_step": 162733, "epoch": 1828, "lr": 3.473452478923187e-05}
{"train_loss": 0.08817113935947418, "global_step": 162734, "epoch": 1828, "lr": 3.473397273423762e-05}
{"train_loss": 0.1304865926504135, "global_step": 162735, "epoch": 1828, "lr": 3.473342068129568e-05}
{"train_loss": 0.10958479344844818, "global_step": 162736, "epoch": 1828, "lr": 3.473286863040615e-05}
{"train_loss": 0.09207786619663239, "global_step": 162737, "epoch": 1828, "lr": 3.473231658156912e-05}
{"train_loss": 0.12357897311449051, "global_step": 162738, "epoch": 1828, "lr": 3.473176453478462e-05}
{"train_loss": 0.08487454056739807, "global_step": 162739, "epoch": 1828, "lr": 3.473121249005277e-05}
{"train_loss": 0.07551246881484985, "global_step": 162740, "epoch": 1828, "lr": 3.4730660447373604e-05}
{"train_loss": 0.06404423713684082, "global_step": 162741, "epoch": 1828, "lr": 3.473010840674723e-05}
{"train_loss": 0.0479942262172699, "global_step": 162742, "epoch": 1828, "lr": 3.4729556368173685e-05}
{"train_loss": 0.08339841663837433, "global_step": 162743, "epoch": 1828, "lr": 3.472900433165309e-05}
{"train_loss": 0.08640625327825546, "global_step": 162744, "epoch": 1828, "lr": 3.4728452297185474e-05}
{"train_loss": 0.05502897500991821, "global_step": 162745, "epoch": 1828, "lr": 3.4727900264770956e-05}
{"train_loss": 0.0577266588807106, "global_step": 162746, "epoch": 1828, "lr": 3.472734823440957e-05}
{"train_loss": 0.0831860601902008, "global_step": 162747, "epoch": 1828, "lr": 3.472679620610141e-05}
{"train_loss": 0.11534470319747925, "global_step": 162748, "epoch": 1828, "lr": 3.472624417984655e-05}
{"train_loss": 0.09953806549310684, "global_step": 162749, "epoch": 1828, "lr": 3.4725692155645045e-05}
{"train_loss": 0.066938117146492, "global_step": 162750, "epoch": 1828, "lr": 3.4725140133496994e-05}
{"train_loss": 0.061853379011154175, "global_step": 162751, "epoch": 1828, "lr": 3.472458811340247e-05}
{"train_loss": 0.14140290021896362, "global_step": 162752, "epoch": 1828, "lr": 3.472403609536152e-05}
{"train_loss": 0.07558012753725052, "global_step": 162753, "epoch": 1828, "lr": 3.4723484079374255e-05}
{"train_loss": 0.09392668306827545, "global_step": 162754, "epoch": 1828, "lr": 3.4722932065440715e-05}
{"train_loss": 0.04443282261490822, "global_step": 162755, "epoch": 1828, "lr": 3.472238005356099e-05}
{"train_loss": 0.03821127489209175, "global_step": 162756, "epoch": 1828, "lr": 3.472182804373517e-05}
{"train_loss": 0.1285375952720642, "global_step": 162757, "epoch": 1828, "lr": 3.4721276035963294e-05}
{"train_loss": 0.06208407133817673, "global_step": 162758, "epoch": 1828, "lr": 3.472072403024548e-05}
{"train_loss": 0.10188835114240646, "global_step": 162759, "epoch": 1828, "lr": 3.472017202658176e-05}
{"train_loss": 0.08650852739810944, "global_step": 162760, "epoch": 1828, "lr": 3.4719620024972246e-05}
{"train_loss": 0.10721265524625778, "global_step": 162761, "epoch": 1828, "lr": 3.471906802541697e-05}
{"train_loss": 0.10706986486911774, "global_step": 162762, "epoch": 1828, "lr": 3.471851602791605e-05}
{"train_loss": 0.058843228965997696, "global_step": 162763, "epoch": 1828, "lr": 3.471796403246952e-05}
{"train_loss": 0.10107820481061935, "global_step": 162764, "epoch": 1828, "lr": 3.47174120390775e-05}
{"train_loss": 0.13244600594043732, "global_step": 162765, "epoch": 1828, "lr": 3.471686004774002e-05}
{"train_loss": 0.03892401605844498, "global_step": 162766, "epoch": 1828, "lr": 3.471630805845717e-05}
{"train_loss": 0.07720877230167389, "global_step": 162767, "epoch": 1828, "lr": 3.471575607122903e-05}
{"train_loss": 0.13223285973072052, "global_step": 162768, "epoch": 1828, "lr": 3.471520408605566e-05}
{"train_loss": 0.058319851756095886, "global_step": 162769, "epoch": 1828, "lr": 3.471465210293716e-05}
{"train_loss": 0.11320337653160095, "global_step": 162770, "epoch": 1828, "lr": 3.4714100121873594e-05}
{"train_loss": 0.08775604516267776, "global_step": 162771, "epoch": 1828, "lr": 3.4713548142865005e-05}
{"train_loss": 0.05774909630417824, "global_step": 162772, "epoch": 1828, "lr": 3.47129961659115e-05}
{"train_loss": 0.09721668064594269, "global_step": 162773, "epoch": 1828, "lr": 3.4712444191013164e-05}
{"train_loss": 0.05892443284392357, "global_step": 162774, "epoch": 1828, "lr": 3.471189221817003e-05}
{"train_loss": 0.0996689423918724, "global_step": 162775, "epoch": 1828, "lr": 3.4711340247382224e-05}
{"train_loss": 0.06753115355968475, "global_step": 162776, "epoch": 1828, "lr": 3.4710788278649764e-05}
{"train_loss": 0.08426786214113235, "global_step": 162777, "epoch": 1828, "lr": 3.4710236311972775e-05}
{"train_loss": 0.07643842697143555, "global_step": 162778, "epoch": 1828, "lr": 3.470968434735129e-05}
{"train_loss": 0.10648103803396225, "global_step": 162779, "epoch": 1828, "lr": 3.470913238478543e-05}
{"train_loss": 0.08869711209214136, "global_step": 162780, "epoch": 1828, "lr": 3.4708580424275205e-05, "val_loss": 7.1623687744140625}
{"train_loss": 0.11274763941764832, "global_step": 162781, "epoch": 1829, "lr": 3.470802846582075e-05}
{"train_loss": 0.08758912235498428, "global_step": 162782, "epoch": 1829, "lr": 3.4707476509422095e-05}
{"train_loss": 0.08644644170999527, "global_step": 162783, "epoch": 1829, "lr": 3.4706924555079356e-05}
{"train_loss": 0.0966600552201271, "global_step": 162784, "epoch": 1829, "lr": 3.470637260279257e-05}
{"train_loss": 0.05786556378006935, "global_step": 162785, "epoch": 1829, "lr": 3.470582065256183e-05}
{"train_loss": 0.120207279920578, "global_step": 162786, "epoch": 1829, "lr": 3.47052687043872e-05}
{"train_loss": 0.05061052739620209, "global_step": 162787, "epoch": 1829, "lr": 3.4704716758268764e-05}
{"train_loss": 0.08237046748399734, "global_step": 162788, "epoch": 1829, "lr": 3.47041648142066e-05}
{"train_loss": 0.05885470658540726, "global_step": 162789, "epoch": 1829, "lr": 3.470361287220077e-05}
{"train_loss": 0.07497088611125946, "global_step": 162790, "epoch": 1829, "lr": 3.470306093225134e-05}
{"train_loss": 0.11538158357143402, "global_step": 162791, "epoch": 1829, "lr": 3.4702508994358394e-05}
{"train_loss": 0.09572986513376236, "global_step": 162792, "epoch": 1829, "lr": 3.470195705852204e-05}
{"train_loss": 0.08588661253452301, "global_step": 162793, "epoch": 1829, "lr": 3.470140512474229e-05}
{"train_loss": 0.1586063802242279, "global_step": 162794, "epoch": 1829, "lr": 3.4700853193019276e-05}
{"train_loss": 0.10669122636318207, "global_step": 162795, "epoch": 1829, "lr": 3.470030126335302e-05}
{"train_loss": 0.09221972525119781, "global_step": 162796, "epoch": 1829, "lr": 3.469974933574365e-05}
{"train_loss": 0.14799651503562927, "global_step": 162797, "epoch": 1829, "lr": 3.469919741019118e-05}
{"train_loss": 0.07866666465997696, "global_step": 162798, "epoch": 1829, "lr": 3.469864548669575e-05}
{"train_loss": 0.14432020485401154, "global_step": 162799, "epoch": 1829, "lr": 3.4698093565257374e-05}
{"train_loss": 0.07537302374839783, "global_step": 162800, "epoch": 1829, "lr": 3.469754164587617e-05}
{"train_loss": 0.046377994120121, "global_step": 162801, "epoch": 1829, "lr": 3.4696989728552174e-05}
{"train_loss": 0.09039855748414993, "global_step": 162802, "epoch": 1829, "lr": 3.469643781328551e-05}
{"train_loss": 0.08865304291248322, "global_step": 162803, "epoch": 1829, "lr": 3.469588590007621e-05}
{"train_loss": 0.07809380441904068, "global_step": 162804, "epoch": 1829, "lr": 3.469533398892436e-05}
{"train_loss": 0.047080203890800476, "global_step": 162805, "epoch": 1829, "lr": 3.469478207983003e-05}
{"train_loss": 0.11757246404886246, "global_step": 162806, "epoch": 1829, "lr": 3.469423017279331e-05}
{"train_loss": 0.06834807246923447, "global_step": 162807, "epoch": 1829, "lr": 3.469367826781426e-05}
{"train_loss": 0.0664563998579979, "global_step": 162808, "epoch": 1829, "lr": 3.469312636489295e-05}
{"train_loss": 0.1075093001127243, "global_step": 162809, "epoch": 1829, "lr": 3.469257446402948e-05}
{"train_loss": 0.06770749390125275, "global_step": 162810, "epoch": 1829, "lr": 3.469202256522389e-05}
{"train_loss": 0.12527728080749512, "global_step": 162811, "epoch": 1829, "lr": 3.469147066847629e-05}
{"train_loss": 0.06631460785865784, "global_step": 162812, "epoch": 1829, "lr": 3.469091877378672e-05}
{"train_loss": 0.057809874415397644, "global_step": 162813, "epoch": 1829, "lr": 3.469036688115528e-05}
{"train_loss": 0.09402774274349213, "global_step": 162814, "epoch": 1829, "lr": 3.468981499058201e-05}
{"train_loss": 0.10108936578035355, "global_step": 162815, "epoch": 1829, "lr": 3.4689263102067035e-05}
{"train_loss": 0.09320524334907532, "global_step": 162816, "epoch": 1829, "lr": 3.468871121561039e-05}
{"train_loss": 0.0929047167301178, "global_step": 162817, "epoch": 1829, "lr": 3.4688159331212175e-05}
{"train_loss": 0.09990246593952179, "global_step": 162818, "epoch": 1829, "lr": 3.468760744887243e-05}
{"train_loss": 0.09411093592643738, "global_step": 162819, "epoch": 1829, "lr": 3.468705556859126e-05}
{"train_loss": 0.09630665183067322, "global_step": 162820, "epoch": 1829, "lr": 3.4686503690368724e-05}
{"train_loss": 0.05967488884925842, "global_step": 162821, "epoch": 1829, "lr": 3.468595181420491e-05}
{"train_loss": 0.09520380198955536, "global_step": 162822, "epoch": 1829, "lr": 3.468539994009987e-05}
{"train_loss": 0.08245399594306946, "global_step": 162823, "epoch": 1829, "lr": 3.4684848068053705e-05}
{"train_loss": 0.10011083632707596, "global_step": 162824, "epoch": 1829, "lr": 3.468429619806647e-05}
{"train_loss": 0.0727330669760704, "global_step": 162825, "epoch": 1829, "lr": 3.468374433013825e-05}
{"train_loss": 0.09496255218982697, "global_step": 162826, "epoch": 1829, "lr": 3.46831924642691e-05}
{"train_loss": 0.06815337389707565, "global_step": 162827, "epoch": 1829, "lr": 3.46826406004591e-05}
{"train_loss": 0.0655948743224144, "global_step": 162828, "epoch": 1829, "lr": 3.468208873870836e-05}
{"train_loss": 0.10526825487613678, "global_step": 162829, "epoch": 1829, "lr": 3.4681536879016906e-05}
{"train_loss": 0.0925552248954773, "global_step": 162830, "epoch": 1829, "lr": 3.4680985021384845e-05}
{"train_loss": 0.10510580986738205, "global_step": 162831, "epoch": 1829, "lr": 3.4680433165812233e-05}
{"train_loss": 0.0950007364153862, "global_step": 162832, "epoch": 1829, "lr": 3.4679881312299165e-05}
{"train_loss": 0.07569518685340881, "global_step": 162833, "epoch": 1829, "lr": 3.4679329460845674e-05}
{"train_loss": 0.06808727979660034, "global_step": 162834, "epoch": 1829, "lr": 3.467877761145188e-05}
{"train_loss": 0.062316033989191055, "global_step": 162835, "epoch": 1829, "lr": 3.4678225764117825e-05}
{"train_loss": 0.060735058039426804, "global_step": 162836, "epoch": 1829, "lr": 3.467767391884361e-05}
{"train_loss": 0.08698978275060654, "global_step": 162837, "epoch": 1829, "lr": 3.4677122075629275e-05}
{"train_loss": 0.11585406213998795, "global_step": 162838, "epoch": 1829, "lr": 3.467657023447494e-05}
{"train_loss": 0.05614318698644638, "global_step": 162839, "epoch": 1829, "lr": 3.467601839538063e-05}
{"train_loss": 0.08803561329841614, "global_step": 162840, "epoch": 1829, "lr": 3.467546655834646e-05}
{"train_loss": 0.12376248836517334, "global_step": 162841, "epoch": 1829, "lr": 3.467491472337247e-05}
{"train_loss": 0.0945928692817688, "global_step": 162842, "epoch": 1829, "lr": 3.4674362890458765e-05}
{"train_loss": 0.20591787993907928, "global_step": 162843, "epoch": 1829, "lr": 3.46738110596054e-05}
{"train_loss": 0.0515148863196373, "global_step": 162844, "epoch": 1829, "lr": 3.467325923081246e-05}
{"train_loss": 0.10516050457954407, "global_step": 162845, "epoch": 1829, "lr": 3.4672707404080005e-05}
{"train_loss": 0.11278244853019714, "global_step": 162846, "epoch": 1829, "lr": 3.4672155579408104e-05}
{"train_loss": 0.0987975224852562, "global_step": 162847, "epoch": 1829, "lr": 3.467160375679688e-05}
{"train_loss": 0.10105693340301514, "global_step": 162848, "epoch": 1829, "lr": 3.467105193624634e-05}
{"train_loss": 0.09552191942930222, "global_step": 162849, "epoch": 1829, "lr": 3.467050011775661e-05}
{"train_loss": 0.11554315686225891, "global_step": 162850, "epoch": 1829, "lr": 3.466994830132773e-05}
{"train_loss": 0.06892276555299759, "global_step": 162851, "epoch": 1829, "lr": 3.46693964869598e-05}
{"train_loss": 0.0929095447063446, "global_step": 162852, "epoch": 1829, "lr": 3.4668844674652875e-05}
{"train_loss": 0.10642167180776596, "global_step": 162853, "epoch": 1829, "lr": 3.4668292864407056e-05}
{"train_loss": 0.0729091465473175, "global_step": 162854, "epoch": 1829, "lr": 3.4667741056222366e-05}
{"train_loss": 0.08226023614406586, "global_step": 162855, "epoch": 1829, "lr": 3.466718925009894e-05}
{"train_loss": 0.07579892873764038, "global_step": 162856, "epoch": 1829, "lr": 3.46666374460368e-05}
{"train_loss": 0.1344725340604782, "global_step": 162857, "epoch": 1829, "lr": 3.466608564403607e-05}
{"train_loss": 0.10857830941677094, "global_step": 162858, "epoch": 1829, "lr": 3.466553384409678e-05}
{"train_loss": 0.09927843511104584, "global_step": 162859, "epoch": 1829, "lr": 3.466498204621903e-05}
{"train_loss": 0.0798727422952652, "global_step": 162860, "epoch": 1829, "lr": 3.466443025040287e-05}
{"train_loss": 0.0735628753900528, "global_step": 162861, "epoch": 1829, "lr": 3.4663878456648416e-05}
{"train_loss": 0.1032826229929924, "global_step": 162862, "epoch": 1829, "lr": 3.4663326664955695e-05}
{"train_loss": 0.1650659292936325, "global_step": 162863, "epoch": 1829, "lr": 3.466277487532481e-05}
{"train_loss": 0.06940881907939911, "global_step": 162864, "epoch": 1829, "lr": 3.466222308775584e-05}
{"train_loss": 0.12699604034423828, "global_step": 162865, "epoch": 1829, "lr": 3.466167130224882e-05}
{"train_loss": 0.07009334862232208, "global_step": 162866, "epoch": 1829, "lr": 3.466111951880388e-05}
{"train_loss": 0.161874920129776, "global_step": 162867, "epoch": 1829, "lr": 3.4660567737421044e-05}
{"train_loss": 0.07662872225046158, "global_step": 162868, "epoch": 1829, "lr": 3.4660015958100425e-05}
{"train_loss": 0.09210676323162036, "global_step": 162869, "epoch": 1829, "lr": 3.465946418084206e-05, "val_loss": 6.929399013519287}
{"train_loss": 0.08396027982234955, "global_step": 162870, "epoch": 1830, "lr": 3.4658912405646065e-05}
{"train_loss": 0.09000828862190247, "global_step": 162871, "epoch": 1830, "lr": 3.465836063251248e-05}
{"train_loss": 0.0642237663269043, "global_step": 162872, "epoch": 1830, "lr": 3.46578088614414e-05}
{"train_loss": 0.10616657882928848, "global_step": 162873, "epoch": 1830, "lr": 3.465725709243287e-05}
{"train_loss": 0.07612256705760956, "global_step": 162874, "epoch": 1830, "lr": 3.465670532548701e-05}
{"train_loss": 0.1054617166519165, "global_step": 162875, "epoch": 1830, "lr": 3.4656153560603843e-05}
{"train_loss": 0.0915842279791832, "global_step": 162876, "epoch": 1830, "lr": 3.46556017977835e-05}
{"train_loss": 0.07974117994308472, "global_step": 162877, "epoch": 1830, "lr": 3.4655050037025995e-05}
{"train_loss": 0.08574041724205017, "global_step": 162878, "epoch": 1830, "lr": 3.465449827833146e-05}
{"train_loss": 0.11610502004623413, "global_step": 162879, "epoch": 1830, "lr": 3.465394652169991e-05}
{"train_loss": 0.046887099742889404, "global_step": 162880, "epoch": 1830, "lr": 3.4653394767131475e-05}
{"train_loss": 0.06523241102695465, "global_step": 162881, "epoch": 1830, "lr": 3.46528430146262e-05}
{"train_loss": 0.06505099684000015, "global_step": 162882, "epoch": 1830, "lr": 3.4652291264184144e-05}
{"train_loss": 0.09641270339488983, "global_step": 162883, "epoch": 1830, "lr": 3.465173951580543e-05}
{"train_loss": 0.10037828236818314, "global_step": 162884, "epoch": 1830, "lr": 3.4651187769490065e-05}
{"train_loss": 0.09711280465126038, "global_step": 162885, "epoch": 1830, "lr": 3.4650636025238195e-05}
{"train_loss": 0.04633583873510361, "global_step": 162886, "epoch": 1830, "lr": 3.4650084283049834e-05}
{"train_loss": 0.09885751456022263, "global_step": 162887, "epoch": 1830, "lr": 3.464953254292511e-05}
{"train_loss": 0.12322327494621277, "global_step": 162888, "epoch": 1830, "lr": 3.464898080486404e-05}
{"train_loss": 0.06859459728002548, "global_step": 162889, "epoch": 1830, "lr": 3.464842906886675e-05}
{"train_loss": 0.1270267218351364, "global_step": 162890, "epoch": 1830, "lr": 3.4647877334933275e-05}
{"train_loss": 0.06034303084015846, "global_step": 162891, "epoch": 1830, "lr": 3.464732560306373e-05}
{"train_loss": 0.17882414162158966, "global_step": 162892, "epoch": 1830, "lr": 3.464677387325813e-05}
{"train_loss": 0.09744272381067276, "global_step": 162893, "epoch": 1830, "lr": 3.4646222145516615e-05}
{"train_loss": 0.10241415351629257, "global_step": 162894, "epoch": 1830, "lr": 3.464567041983921e-05}
{"train_loss": 0.09210371226072311, "global_step": 162895, "epoch": 1830, "lr": 3.464511869622602e-05}
{"train_loss": 0.13129816949367523, "global_step": 162896, "epoch": 1830, "lr": 3.4644566974677096e-05}
{"train_loss": 0.0990322083234787, "global_step": 162897, "epoch": 1830, "lr": 3.464401525519254e-05}
{"train_loss": 0.09312253445386887, "global_step": 162898, "epoch": 1830, "lr": 3.464346353777238e-05}
{"train_loss": 0.10301919281482697, "global_step": 162899, "epoch": 1830, "lr": 3.464291182241674e-05}
{"train_loss": 0.06336959451436996, "global_step": 162900, "epoch": 1830, "lr": 3.4642360109125677e-05}
{"train_loss": 0.10209789872169495, "global_step": 162901, "epoch": 1830, "lr": 3.464180839789924e-05}
{"train_loss": 0.10455139726400375, "global_step": 162902, "epoch": 1830, "lr": 3.464125668873755e-05}
{"train_loss": 0.0744968131184578, "global_step": 162903, "epoch": 1830, "lr": 3.464070498164064e-05}
{"train_loss": 0.06348222494125366, "global_step": 162904, "epoch": 1830, "lr": 3.464015327660861e-05}
{"train_loss": 0.0764680951833725, "global_step": 162905, "epoch": 1830, "lr": 3.4639601573641506e-05}
{"train_loss": 0.06269660592079163, "global_step": 162906, "epoch": 1830, "lr": 3.463904987273944e-05}
{"train_loss": 0.09797462821006775, "global_step": 162907, "epoch": 1830, "lr": 3.463849817390245e-05}
{"train_loss": 0.12349303066730499, "global_step": 162908, "epoch": 1830, "lr": 3.4637946477130645e-05}
{"train_loss": 0.09141483902931213, "global_step": 162909, "epoch": 1830, "lr": 3.463739478242406e-05}
{"train_loss": 0.08572743833065033, "global_step": 162910, "epoch": 1830, "lr": 3.4636843089782814e-05}
{"train_loss": 0.06106162443757057, "global_step": 162911, "epoch": 1830, "lr": 3.4636291399206934e-05}
{"train_loss": 0.07628196477890015, "global_step": 162912, "epoch": 1830, "lr": 3.463573971069654e-05}
{"train_loss": 0.07208921015262604, "global_step": 162913, "epoch": 1830, "lr": 3.463518802425166e-05}
{"train_loss": 0.0887487530708313, "global_step": 162914, "epoch": 1830, "lr": 3.463463633987242e-05}
{"train_loss": 0.06459273397922516, "global_step": 162915, "epoch": 1830, "lr": 3.463408465755883e-05}
{"train_loss": 0.09852983057498932, "global_step": 162916, "epoch": 1830, "lr": 3.463353297731103e-05}
{"train_loss": 0.10031434148550034, "global_step": 162917, "epoch": 1830, "lr": 3.463298129912904e-05}
{"train_loss": 0.10631430894136429, "global_step": 162918, "epoch": 1830, "lr": 3.463242962301298e-05}
{"train_loss": 0.13477598130702972, "global_step": 162919, "epoch": 1830, "lr": 3.46318779489629e-05}
{"train_loss": 0.06801268458366394, "global_step": 162920, "epoch": 1830, "lr": 3.463132627697886e-05}
{"train_loss": 0.12366281449794769, "global_step": 162921, "epoch": 1830, "lr": 3.463077460706098e-05}
{"train_loss": 0.06875766068696976, "global_step": 162922, "epoch": 1830, "lr": 3.463022293920927e-05}
{"train_loss": 0.07660697400569916, "global_step": 162923, "epoch": 1830, "lr": 3.462967127342387e-05}
{"train_loss": 0.06514068692922592, "global_step": 162924, "epoch": 1830, "lr": 3.4629119609704794e-05}
{"train_loss": 0.06707850843667984, "global_step": 162925, "epoch": 1830, "lr": 3.462856794805217e-05}
{"train_loss": 0.18531033396720886, "global_step": 162926, "epoch": 1830, "lr": 3.4628016288466033e-05}
{"train_loss": 0.0896376371383667, "global_step": 162927, "epoch": 1830, "lr": 3.462746463094649e-05}
{"train_loss": 0.08616092056035995, "global_step": 162928, "epoch": 1830, "lr": 3.462691297549357e-05}
{"train_loss": 0.07353459298610687, "global_step": 162929, "epoch": 1830, "lr": 3.46263613221074e-05}
{"train_loss": 0.08094486594200134, "global_step": 162930, "epoch": 1830, "lr": 3.4625809670788004e-05}
{"train_loss": 0.06461929529905319, "global_step": 162931, "epoch": 1830, "lr": 3.462525802153551e-05}
{"train_loss": 0.08530517667531967, "global_step": 162932, "epoch": 1830, "lr": 3.4624706374349935e-05}
{"train_loss": 0.1895725429058075, "global_step": 162933, "epoch": 1830, "lr": 3.46241547292314e-05}
{"train_loss": 0.07859337329864502, "global_step": 162934, "epoch": 1830, "lr": 3.462360308617994e-05}
{"train_loss": 0.07033533602952957, "global_step": 162935, "epoch": 1830, "lr": 3.462305144519566e-05}
{"train_loss": 0.09913668781518936, "global_step": 162936, "epoch": 1830, "lr": 3.462249980627863e-05}
{"train_loss": 0.08305618911981583, "global_step": 162937, "epoch": 1830, "lr": 3.462194816942891e-05}
{"train_loss": 0.1379091739654541, "global_step": 162938, "epoch": 1830, "lr": 3.462139653464658e-05}
{"train_loss": 0.12797211110591888, "global_step": 162939, "epoch": 1830, "lr": 3.462084490193172e-05}
{"train_loss": 0.0717596709728241, "global_step": 162940, "epoch": 1830, "lr": 3.462029327128441e-05}
{"train_loss": 0.10694466531276703, "global_step": 162941, "epoch": 1830, "lr": 3.4619741642704696e-05}
{"train_loss": 0.11032140254974365, "global_step": 162942, "epoch": 1830, "lr": 3.461919001619269e-05}
{"train_loss": 0.11506948620080948, "global_step": 162943, "epoch": 1830, "lr": 3.461863839174842e-05}
{"train_loss": 0.20715612173080444, "global_step": 162944, "epoch": 1830, "lr": 3.461808676937201e-05}
{"train_loss": 0.09320744127035141, "global_step": 162945, "epoch": 1830, "lr": 3.4617535149063495e-05}
{"train_loss": 0.07917889952659607, "global_step": 162946, "epoch": 1830, "lr": 3.4616983530822976e-05}
{"train_loss": 0.06941995024681091, "global_step": 162947, "epoch": 1830, "lr": 3.4616431914650504e-05}
{"train_loss": 0.08426403254270554, "global_step": 162948, "epoch": 1830, "lr": 3.461588030054618e-05}
{"train_loss": 0.09200846403837204, "global_step": 162949, "epoch": 1830, "lr": 3.461532868851005e-05}
{"train_loss": 0.060663361102342606, "global_step": 162950, "epoch": 1830, "lr": 3.461477707854222e-05}
{"train_loss": 0.06246994808316231, "global_step": 162951, "epoch": 1830, "lr": 3.461422547064272e-05}
{"train_loss": 0.07070878893136978, "global_step": 162952, "epoch": 1830, "lr": 3.461367386481168e-05}
{"train_loss": 0.0793551504611969, "global_step": 162953, "epoch": 1830, "lr": 3.461312226104911e-05}
{"train_loss": 0.09851296991109848, "global_step": 162954, "epoch": 1830, "lr": 3.461257065935514e-05}
{"train_loss": 0.1396479308605194, "global_step": 162955, "epoch": 1830, "lr": 3.4612019059729814e-05}
{"train_loss": 0.15445834398269653, "global_step": 162956, "epoch": 1830, "lr": 3.4611467462173216e-05}
{"train_loss": 0.12770040333271027, "global_step": 162957, "epoch": 1830, "lr": 3.4610915866685425e-05}
{"train_loss": 0.09382907476987731, "global_step": 162958, "epoch": 1830, "lr": 3.46103642732665e-05, "val_loss": 6.90378475189209, "train_action_mse_error": 10.835341453552246}
{"train_loss": 0.07634581625461578, "global_step": 162959, "epoch": 1831, "lr": 3.460981268191654e-05}
{"train_loss": 0.1054588183760643, "global_step": 162960, "epoch": 1831, "lr": 3.460926109263558e-05}
{"train_loss": 0.09587269276380539, "global_step": 162961, "epoch": 1831, "lr": 3.4608709505423744e-05}
{"train_loss": 0.06538372486829758, "global_step": 162962, "epoch": 1831, "lr": 3.460815792028105e-05}
{"train_loss": 0.0745188519358635, "global_step": 162963, "epoch": 1831, "lr": 3.460760633720763e-05}
{"train_loss": 0.12291739881038666, "global_step": 162964, "epoch": 1831, "lr": 3.460705475620351e-05}
{"train_loss": 0.056858498603105545, "global_step": 162965, "epoch": 1831, "lr": 3.460650317726879e-05}
{"train_loss": 0.0624973364174366, "global_step": 162966, "epoch": 1831, "lr": 3.460595160040353e-05}
{"train_loss": 0.07242093980312347, "global_step": 162967, "epoch": 1831, "lr": 3.460540002560783e-05}
{"train_loss": 0.07985588908195496, "global_step": 162968, "epoch": 1831, "lr": 3.4604848452881726e-05}
{"train_loss": 0.09054974466562271, "global_step": 162969, "epoch": 1831, "lr": 3.4604296882225336e-05}
{"train_loss": 0.05453721061348915, "global_step": 162970, "epoch": 1831, "lr": 3.460374531363868e-05}
{"train_loss": 0.08859175443649292, "global_step": 162971, "epoch": 1831, "lr": 3.4603193747121896e-05}
{"train_loss": 0.1527746468782425, "global_step": 162972, "epoch": 1831, "lr": 3.4602642182674996e-05}
{"train_loss": 0.12801967561244965, "global_step": 162973, "epoch": 1831, "lr": 3.46020906202981e-05}
{"train_loss": 0.07217485457658768, "global_step": 162974, "epoch": 1831, "lr": 3.460153905999126e-05}
{"train_loss": 0.1096934825181961, "global_step": 162975, "epoch": 1831, "lr": 3.4600987501754556e-05}
{"train_loss": 0.09115730971097946, "global_step": 162976, "epoch": 1831, "lr": 3.460043594558806e-05}
{"train_loss": 0.1511305868625641, "global_step": 162977, "epoch": 1831, "lr": 3.4599884391491855e-05}
{"train_loss": 0.046705231070518494, "global_step": 162978, "epoch": 1831, "lr": 3.459933283946601e-05}
{"train_loss": 0.07058360427618027, "global_step": 162979, "epoch": 1831, "lr": 3.459878128951058e-05}
{"train_loss": 0.08505594730377197, "global_step": 162980, "epoch": 1831, "lr": 3.4598229741625677e-05}
{"train_loss": 0.07780187577009201, "global_step": 162981, "epoch": 1831, "lr": 3.459767819581133e-05}
{"train_loss": 0.11141032725572586, "global_step": 162982, "epoch": 1831, "lr": 3.4597126652067666e-05}
{"train_loss": 0.10007461160421371, "global_step": 162983, "epoch": 1831, "lr": 3.4596575110394704e-05}
{"train_loss": 0.129430890083313, "global_step": 162984, "epoch": 1831, "lr": 3.459602357079256e-05}
{"train_loss": 0.08658332377672195, "global_step": 162985, "epoch": 1831, "lr": 3.459547203326128e-05}
{"train_loss": 0.053693462163209915, "global_step": 162986, "epoch": 1831, "lr": 3.459492049780097e-05}
{"train_loss": 0.1398058384656906, "global_step": 162987, "epoch": 1831, "lr": 3.4594368964411664e-05}
{"train_loss": 0.16505767405033112, "global_step": 162988, "epoch": 1831, "lr": 3.4593817433093476e-05}
{"train_loss": 0.11070393025875092, "global_step": 162989, "epoch": 1831, "lr": 3.459326590384645e-05}
{"train_loss": 0.08918257057666779, "global_step": 162990, "epoch": 1831, "lr": 3.459271437667066e-05}
{"train_loss": 0.11385849118232727, "global_step": 162991, "epoch": 1831, "lr": 3.459216285156622e-05}
{"train_loss": 0.0942831039428711, "global_step": 162992, "epoch": 1831, "lr": 3.4591611328533166e-05}
{"train_loss": 0.07111837714910507, "global_step": 162993, "epoch": 1831, "lr": 3.459105980757158e-05}
{"train_loss": 0.08779000490903854, "global_step": 162994, "epoch": 1831, "lr": 3.4590508288681536e-05}
{"train_loss": 0.03611241281032562, "global_step": 162995, "epoch": 1831, "lr": 3.458995677186311e-05}
{"train_loss": 0.03806135430932045, "global_step": 162996, "epoch": 1831, "lr": 3.458940525711638e-05}
{"train_loss": 0.0626017227768898, "global_step": 162997, "epoch": 1831, "lr": 3.458885374444142e-05}
{"train_loss": 0.05273576080799103, "global_step": 162998, "epoch": 1831, "lr": 3.4588302233838286e-05}
{"train_loss": 0.15237407386302948, "global_step": 162999, "epoch": 1831, "lr": 3.458775072530709e-05}
{"train_loss": 0.04510454833507538, "global_step": 163000, "epoch": 1831, "lr": 3.458719921884785e-05}
{"train_loss": 0.11509992927312851, "global_step": 163001, "epoch": 1831, "lr": 3.458664771446071e-05}
{"train_loss": 0.11825119704008102, "global_step": 163002, "epoch": 1831, "lr": 3.458609621214568e-05}
{"train_loss": 0.058413852006196976, "global_step": 163003, "epoch": 1831, "lr": 3.4585544711902884e-05}
{"train_loss": 0.08227425068616867, "global_step": 163004, "epoch": 1831, "lr": 3.4584993213732355e-05}
{"train_loss": 0.07247234880924225, "global_step": 163005, "epoch": 1831, "lr": 3.4584441717634196e-05}
{"train_loss": 0.06713791936635971, "global_step": 163006, "epoch": 1831, "lr": 3.4583890223608464e-05}
{"train_loss": 0.13897354900836945, "global_step": 163007, "epoch": 1831, "lr": 3.4583338731655244e-05}
{"train_loss": 0.0476815439760685, "global_step": 163008, "epoch": 1831, "lr": 3.458278724177459e-05}
{"train_loss": 0.07732459157705307, "global_step": 163009, "epoch": 1831, "lr": 3.458223575396661e-05}
{"train_loss": 0.08064267039299011, "global_step": 163010, "epoch": 1831, "lr": 3.4581684268231364e-05}
{"train_loss": 0.05755636468529701, "global_step": 163011, "epoch": 1831, "lr": 3.458113278456892e-05}
{"train_loss": 0.11348842084407806, "global_step": 163012, "epoch": 1831, "lr": 3.458058130297935e-05}
{"train_loss": 0.12281995266675949, "global_step": 163013, "epoch": 1831, "lr": 3.4580029823462734e-05}
{"train_loss": 0.07967229187488556, "global_step": 163014, "epoch": 1831, "lr": 3.457947834601914e-05}
{"train_loss": 0.0811142548918724, "global_step": 163015, "epoch": 1831, "lr": 3.457892687064865e-05}
{"train_loss": 0.12055107206106186, "global_step": 163016, "epoch": 1831, "lr": 3.457837539735136e-05}
{"train_loss": 0.08304542303085327, "global_step": 163017, "epoch": 1831, "lr": 3.457782392612729e-05}
{"train_loss": 0.10533590614795685, "global_step": 163018, "epoch": 1831, "lr": 3.4577272456976554e-05}
{"train_loss": 0.060244299471378326, "global_step": 163019, "epoch": 1831, "lr": 3.45767209898992e-05}
{"train_loss": 0.11737200617790222, "global_step": 163020, "epoch": 1831, "lr": 3.457616952489534e-05}
{"train_loss": 0.07467299699783325, "global_step": 163021, "epoch": 1831, "lr": 3.457561806196501e-05}
{"train_loss": 0.12552478909492493, "global_step": 163022, "epoch": 1831, "lr": 3.4575066601108316e-05}
{"train_loss": 0.09700389951467514, "global_step": 163023, "epoch": 1831, "lr": 3.45745151423253e-05}
{"train_loss": 0.21244393289089203, "global_step": 163024, "epoch": 1831, "lr": 3.457396368561607e-05}
{"train_loss": 0.06757421791553497, "global_step": 163025, "epoch": 1831, "lr": 3.457341223098066e-05}
{"train_loss": 0.061445921659469604, "global_step": 163026, "epoch": 1831, "lr": 3.457286077841917e-05}
{"train_loss": 0.12016730010509491, "global_step": 163027, "epoch": 1831, "lr": 3.4572309327931684e-05}
{"train_loss": 0.04754316806793213, "global_step": 163028, "epoch": 1831, "lr": 3.457175787951825e-05}
{"train_loss": 0.09233144670724869, "global_step": 163029, "epoch": 1831, "lr": 3.457120643317897e-05}
{"train_loss": 0.07551407814025879, "global_step": 163030, "epoch": 1831, "lr": 3.4570654988913905e-05}
{"train_loss": 0.10905150324106216, "global_step": 163031, "epoch": 1831, "lr": 3.457010354672312e-05}
{"train_loss": 0.09394559264183044, "global_step": 163032, "epoch": 1831, "lr": 3.4569552106606697e-05}
{"train_loss": 0.07205747067928314, "global_step": 163033, "epoch": 1831, "lr": 3.4569000668564705e-05}
{"train_loss": 0.08995243161916733, "global_step": 163034, "epoch": 1831, "lr": 3.456844923259722e-05}
{"train_loss": 0.04379963502287865, "global_step": 163035, "epoch": 1831, "lr": 3.456789779870434e-05}
{"train_loss": 0.06828032433986664, "global_step": 163036, "epoch": 1831, "lr": 3.4567346366886096e-05}
{"train_loss": 0.09785112738609314, "global_step": 163037, "epoch": 1831, "lr": 3.45667949371426e-05}
{"train_loss": 0.14905792474746704, "global_step": 163038, "epoch": 1831, "lr": 3.45662435094739e-05}
{"train_loss": 0.12695233523845673, "global_step": 163039, "epoch": 1831, "lr": 3.4565692083880085e-05}
{"train_loss": 0.06897760182619095, "global_step": 163040, "epoch": 1831, "lr": 3.4565140660361205e-05}
{"train_loss": 0.11399569362401962, "global_step": 163041, "epoch": 1831, "lr": 3.456458923891739e-05}
{"train_loss": 0.09727213531732559, "global_step": 163042, "epoch": 1831, "lr": 3.4564037819548643e-05}
{"train_loss": 0.09561716020107269, "global_step": 163043, "epoch": 1831, "lr": 3.45634864022551e-05}
{"train_loss": 0.06515361368656158, "global_step": 163044, "epoch": 1831, "lr": 3.456293498703679e-05}
{"train_loss": 0.09721118956804276, "global_step": 163045, "epoch": 1831, "lr": 3.45623835738938e-05}
{"train_loss": 0.05230889469385147, "global_step": 163046, "epoch": 1831, "lr": 3.456183216282623e-05}
{"train_loss": 0.09053196929646341, "global_step": 163047, "epoch": 1831, "lr": 3.456128075383411e-05, "val_loss": 6.740813255310059}
{"train_loss": 0.10162386298179626, "global_step": 163048, "epoch": 1832, "lr": 3.4560729346917565e-05}
{"train_loss": 0.16032451391220093, "global_step": 163049, "epoch": 1832, "lr": 3.456017794207663e-05}
{"train_loss": 0.08026231080293655, "global_step": 163050, "epoch": 1832, "lr": 3.4559626539311385e-05}
{"train_loss": 0.08751248568296432, "global_step": 163051, "epoch": 1832, "lr": 3.455907513862192e-05}
{"train_loss": 0.0674457997083664, "global_step": 163052, "epoch": 1832, "lr": 3.455852374000829e-05}
{"train_loss": 0.0562850683927536, "global_step": 163053, "epoch": 1832, "lr": 3.4557972343470586e-05}
{"train_loss": 0.04688476026058197, "global_step": 163054, "epoch": 1832, "lr": 3.455742094900887e-05}
{"train_loss": 0.09106946736574173, "global_step": 163055, "epoch": 1832, "lr": 3.455686955662322e-05}
{"train_loss": 0.09188875555992126, "global_step": 163056, "epoch": 1832, "lr": 3.455631816631372e-05}
{"train_loss": 0.07673429697751999, "global_step": 163057, "epoch": 1832, "lr": 3.455576677808042e-05}
{"train_loss": 0.09712979942560196, "global_step": 163058, "epoch": 1832, "lr": 3.4555215391923424e-05}
{"train_loss": 0.12538912892341614, "global_step": 163059, "epoch": 1832, "lr": 3.455466400784277e-05}
{"train_loss": 0.08465729653835297, "global_step": 163060, "epoch": 1832, "lr": 3.4554112625838575e-05}
{"train_loss": 0.16087430715560913, "global_step": 163061, "epoch": 1832, "lr": 3.455356124591088e-05}
{"train_loss": 0.0427212230861187, "global_step": 163062, "epoch": 1832, "lr": 3.455300986805976e-05}
{"train_loss": 0.15002527832984924, "global_step": 163063, "epoch": 1832, "lr": 3.455245849228532e-05}
{"train_loss": 0.13022863864898682, "global_step": 163064, "epoch": 1832, "lr": 3.45519071185876e-05}
{"train_loss": 0.05626361072063446, "global_step": 163065, "epoch": 1832, "lr": 3.45513557469667e-05}
{"train_loss": 0.04373134672641754, "global_step": 163066, "epoch": 1832, "lr": 3.455080437742266e-05}
{"train_loss": 0.12207786738872528, "global_step": 163067, "epoch": 1832, "lr": 3.455025300995561e-05}
{"train_loss": 0.15119686722755432, "global_step": 163068, "epoch": 1832, "lr": 3.454970164456556e-05}
{"train_loss": 0.11287690699100494, "global_step": 163069, "epoch": 1832, "lr": 3.4549150281252636e-05}
{"train_loss": 0.11302413046360016, "global_step": 163070, "epoch": 1832, "lr": 3.4548598920016875e-05}
{"train_loss": 0.08927250653505325, "global_step": 163071, "epoch": 1832, "lr": 3.454804756085838e-05}
{"train_loss": 0.13482782244682312, "global_step": 163072, "epoch": 1832, "lr": 3.45474962037772e-05}
{"train_loss": 0.13381555676460266, "global_step": 163073, "epoch": 1832, "lr": 3.4546944848773424e-05}
{"train_loss": 0.08409908413887024, "global_step": 163074, "epoch": 1832, "lr": 3.4546393495847124e-05}
{"train_loss": 0.07499293982982635, "global_step": 163075, "epoch": 1832, "lr": 3.454584214499838e-05}
{"train_loss": 0.12922222912311554, "global_step": 163076, "epoch": 1832, "lr": 3.454529079622724e-05}
{"train_loss": 0.08561872690916061, "global_step": 163077, "epoch": 1832, "lr": 3.4544739449533824e-05}
{"train_loss": 0.07602868974208832, "global_step": 163078, "epoch": 1832, "lr": 3.454418810491815e-05}
{"train_loss": 0.08480817824602127, "global_step": 163079, "epoch": 1832, "lr": 3.454363676238035e-05}
{"train_loss": 0.08121027797460556, "global_step": 163080, "epoch": 1832, "lr": 3.4543085421920444e-05}
{"train_loss": 0.1481231153011322, "global_step": 163081, "epoch": 1832, "lr": 3.454253408353853e-05}
{"train_loss": 0.10464758425951004, "global_step": 163082, "epoch": 1832, "lr": 3.4541982747234705e-05}
{"train_loss": 0.041021574288606644, "global_step": 163083, "epoch": 1832, "lr": 3.454143141300901e-05}
{"train_loss": 0.0880340039730072, "global_step": 163084, "epoch": 1832, "lr": 3.4540880080861545e-05}
{"train_loss": 0.06681802123785019, "global_step": 163085, "epoch": 1832, "lr": 3.454032875079235e-05}
{"train_loss": 0.08395105600357056, "global_step": 163086, "epoch": 1832, "lr": 3.453977742280154e-05}
{"train_loss": 0.10390447825193405, "global_step": 163087, "epoch": 1832, "lr": 3.4539226096889145e-05}
{"train_loss": 0.10191135853528976, "global_step": 163088, "epoch": 1832, "lr": 3.453867477305529e-05}
{"train_loss": 0.08883962780237198, "global_step": 163089, "epoch": 1832, "lr": 3.453812345130001e-05}
{"train_loss": 0.1362244337797165, "global_step": 163090, "epoch": 1832, "lr": 3.453757213162339e-05}
{"train_loss": 0.05001959204673767, "global_step": 163091, "epoch": 1832, "lr": 3.45370208140255e-05}
{"train_loss": 0.09862707555294037, "global_step": 163092, "epoch": 1832, "lr": 3.453646949850642e-05}
{"train_loss": 0.09989355504512787, "global_step": 163093, "epoch": 1832, "lr": 3.453591818506623e-05}
{"train_loss": 0.06185334920883179, "global_step": 163094, "epoch": 1832, "lr": 3.4535366873705e-05}
{"train_loss": 0.08332516252994537, "global_step": 163095, "epoch": 1832, "lr": 3.4534815564422784e-05}
{"train_loss": 0.05553624778985977, "global_step": 163096, "epoch": 1832, "lr": 3.4534264257219695e-05}
{"train_loss": 0.1060180813074112, "global_step": 163097, "epoch": 1832, "lr": 3.453371295209577e-05}
{"train_loss": 0.12387639284133911, "global_step": 163098, "epoch": 1832, "lr": 3.45331616490511e-05}
{"train_loss": 0.1397666484117508, "global_step": 163099, "epoch": 1832, "lr": 3.453261034808576e-05}
{"train_loss": 0.10511292517185211, "global_step": 163100, "epoch": 1832, "lr": 3.453205904919981e-05}
{"train_loss": 0.07696640491485596, "global_step": 163101, "epoch": 1832, "lr": 3.4531507752393365e-05}
{"train_loss": 0.15308956801891327, "global_step": 163102, "epoch": 1832, "lr": 3.453095645766644e-05}
{"train_loss": 0.07674097269773483, "global_step": 163103, "epoch": 1832, "lr": 3.4530405165019166e-05}
{"train_loss": 0.0832965224981308, "global_step": 163104, "epoch": 1832, "lr": 3.452985387445157e-05}
{"train_loss": 0.1410311907529831, "global_step": 163105, "epoch": 1832, "lr": 3.4529302585963766e-05}
{"train_loss": 0.14349041879177094, "global_step": 163106, "epoch": 1832, "lr": 3.452875129955578e-05}
{"train_loss": 0.10464433580636978, "global_step": 163107, "epoch": 1832, "lr": 3.4528200015227754e-05}
{"train_loss": 0.04854600504040718, "global_step": 163108, "epoch": 1832, "lr": 3.45276487329797e-05}
{"train_loss": 0.049297600984573364, "global_step": 163109, "epoch": 1832, "lr": 3.452709745281171e-05}
{"train_loss": 0.07952055335044861, "global_step": 163110, "epoch": 1832, "lr": 3.452654617472387e-05}
{"train_loss": 0.08182176202535629, "global_step": 163111, "epoch": 1832, "lr": 3.452599489871625e-05}
{"train_loss": 0.06298947334289551, "global_step": 163112, "epoch": 1832, "lr": 3.452544362478892e-05}
{"train_loss": 0.08740133047103882, "global_step": 163113, "epoch": 1832, "lr": 3.452489235294196e-05}
{"train_loss": 0.12681995332241058, "global_step": 163114, "epoch": 1832, "lr": 3.4524341083175426e-05}
{"train_loss": 0.12660957872867584, "global_step": 163115, "epoch": 1832, "lr": 3.4523789815489426e-05}
{"train_loss": 0.09995775669813156, "global_step": 163116, "epoch": 1832, "lr": 3.452323854988399e-05}
{"train_loss": 0.13092657923698425, "global_step": 163117, "epoch": 1832, "lr": 3.452268728635922e-05}
{"train_loss": 0.11105281859636307, "global_step": 163118, "epoch": 1832, "lr": 3.4522136024915194e-05}
{"train_loss": 0.09843053668737411, "global_step": 163119, "epoch": 1832, "lr": 3.452158476555197e-05}
{"train_loss": 0.07678288966417313, "global_step": 163120, "epoch": 1832, "lr": 3.452103350826965e-05}
{"train_loss": 0.10774257779121399, "global_step": 163121, "epoch": 1832, "lr": 3.4520482253068264e-05}
{"train_loss": 0.0948261246085167, "global_step": 163122, "epoch": 1832, "lr": 3.451993099994794e-05}
{"train_loss": 0.11817633360624313, "global_step": 163123, "epoch": 1832, "lr": 3.451937974890869e-05}
{"train_loss": 0.05585315451025963, "global_step": 163124, "epoch": 1832, "lr": 3.451882849995065e-05}
{"train_loss": 0.09223859757184982, "global_step": 163125, "epoch": 1832, "lr": 3.4518277253073836e-05}
{"train_loss": 0.06059027090668678, "global_step": 163126, "epoch": 1832, "lr": 3.451772600827838e-05}
{"train_loss": 0.055895332247018814, "global_step": 163127, "epoch": 1832, "lr": 3.4517174765564306e-05}
{"train_loss": 0.12212230265140533, "global_step": 163128, "epoch": 1832, "lr": 3.4516623524931725e-05}
{"train_loss": 0.11746072769165039, "global_step": 163129, "epoch": 1832, "lr": 3.4516072286380684e-05}
{"train_loss": 0.058601029217243195, "global_step": 163130, "epoch": 1832, "lr": 3.451552104991127e-05}
{"train_loss": 0.09405092895030975, "global_step": 163131, "epoch": 1832, "lr": 3.451496981552356e-05}
{"train_loss": 0.0647982507944107, "global_step": 163132, "epoch": 1832, "lr": 3.451441858321762e-05}
{"train_loss": 0.08925186842679977, "global_step": 163133, "epoch": 1832, "lr": 3.4513867352993525e-05}
{"train_loss": 0.11854367703199387, "global_step": 163134, "epoch": 1832, "lr": 3.451331612485136e-05}
{"train_loss": 0.12178505957126617, "global_step": 163135, "epoch": 1832, "lr": 3.4512764898791174e-05}
{"train_loss": 0.09588748174771834, "global_step": 163136, "epoch": 1832, "lr": 3.451221367481306e-05, "val_loss": 6.795382499694824}
{"train_loss": 0.08925488591194153, "global_step": 163137, "epoch": 1833, "lr": 3.45116624529171e-05}
{"train_loss": 0.1056007593870163, "global_step": 163138, "epoch": 1833, "lr": 3.451111123310336e-05}
{"train_loss": 0.10511703789234161, "global_step": 163139, "epoch": 1833, "lr": 3.451056001537192e-05}
{"train_loss": 0.08696488291025162, "global_step": 163140, "epoch": 1833, "lr": 3.451000879972282e-05}
{"train_loss": 0.05066074803471565, "global_step": 163141, "epoch": 1833, "lr": 3.4509457586156186e-05}
{"train_loss": 0.07780057936906815, "global_step": 163142, "epoch": 1833, "lr": 3.450890637467205e-05}
{"train_loss": 0.03943212330341339, "global_step": 163143, "epoch": 1833, "lr": 3.4508355165270514e-05}
{"train_loss": 0.05398542806506157, "global_step": 163144, "epoch": 1833, "lr": 3.450780395795163e-05}
{"train_loss": 0.05163630098104477, "global_step": 163145, "epoch": 1833, "lr": 3.45072527527155e-05}
{"train_loss": 0.08300091326236725, "global_step": 163146, "epoch": 1833, "lr": 3.450670154956217e-05}
{"train_loss": 0.05618225038051605, "global_step": 163147, "epoch": 1833, "lr": 3.450615034849173e-05}
{"train_loss": 0.07879804074764252, "global_step": 163148, "epoch": 1833, "lr": 3.4505599149504244e-05}
{"train_loss": 0.05342855304479599, "global_step": 163149, "epoch": 1833, "lr": 3.450504795259979e-05}
{"train_loss": 0.09019261598587036, "global_step": 163150, "epoch": 1833, "lr": 3.450449675777844e-05}
{"train_loss": 0.08658499270677567, "global_step": 163151, "epoch": 1833, "lr": 3.4503945565040293e-05}
{"train_loss": 0.10707035660743713, "global_step": 163152, "epoch": 1833, "lr": 3.450339437438538e-05}
{"train_loss": 0.07218629121780396, "global_step": 163153, "epoch": 1833, "lr": 3.450284318581378e-05}
{"train_loss": 0.07253780215978622, "global_step": 163154, "epoch": 1833, "lr": 3.450229199932562e-05}
{"train_loss": 0.09332640469074249, "global_step": 163155, "epoch": 1833, "lr": 3.450174081492091e-05}
{"train_loss": 0.08117923140525818, "global_step": 163156, "epoch": 1833, "lr": 3.4501189632599785e-05}
{"train_loss": 0.05338740348815918, "global_step": 163157, "epoch": 1833, "lr": 3.450063845236225e-05}
{"train_loss": 0.0644257515668869, "global_step": 163158, "epoch": 1833, "lr": 3.450008727420845e-05}
{"train_loss": 0.08918995410203934, "global_step": 163159, "epoch": 1833, "lr": 3.44995360981384e-05}
{"train_loss": 0.09665841609239578, "global_step": 163160, "epoch": 1833, "lr": 3.449898492415221e-05}
{"train_loss": 0.1410224288702011, "global_step": 163161, "epoch": 1833, "lr": 3.449843375224993e-05}
{"train_loss": 0.11231635510921478, "global_step": 163162, "epoch": 1833, "lr": 3.4497882582431663e-05}
{"train_loss": 0.09763742983341217, "global_step": 163163, "epoch": 1833, "lr": 3.4497331414697455e-05}
{"train_loss": 0.06482349336147308, "global_step": 163164, "epoch": 1833, "lr": 3.44967802490474e-05}
{"train_loss": 0.08101870864629745, "global_step": 163165, "epoch": 1833, "lr": 3.449622908548155e-05}
{"train_loss": 0.1117430180311203, "global_step": 163166, "epoch": 1833, "lr": 3.449567792400001e-05}
{"train_loss": 0.07811227440834045, "global_step": 163167, "epoch": 1833, "lr": 3.449512676460284e-05}
{"train_loss": 0.0902937576174736, "global_step": 163168, "epoch": 1833, "lr": 3.449457560729009e-05}
{"train_loss": 0.07683609426021576, "global_step": 163169, "epoch": 1833, "lr": 3.449402445206187e-05}
{"train_loss": 0.04759341850876808, "global_step": 163170, "epoch": 1833, "lr": 3.4493473298918236e-05}
{"train_loss": 0.07071886956691742, "global_step": 163171, "epoch": 1833, "lr": 3.449292214785926e-05}
{"train_loss": 0.12933756411075592, "global_step": 163172, "epoch": 1833, "lr": 3.449237099888502e-05}
{"train_loss": 0.04135245829820633, "global_step": 163173, "epoch": 1833, "lr": 3.449181985199561e-05}
{"train_loss": 0.07908237725496292, "global_step": 163174, "epoch": 1833, "lr": 3.449126870719106e-05}
{"train_loss": 0.11918949335813522, "global_step": 163175, "epoch": 1833, "lr": 3.4490717564471496e-05}
{"train_loss": 0.09029948711395264, "global_step": 163176, "epoch": 1833, "lr": 3.449016642383695e-05}
{"train_loss": 0.07525032013654709, "global_step": 163177, "epoch": 1833, "lr": 3.4489615285287524e-05}
{"train_loss": 0.05922620743513107, "global_step": 163178, "epoch": 1833, "lr": 3.448906414882326e-05}
{"train_loss": 0.08617608994245529, "global_step": 163179, "epoch": 1833, "lr": 3.448851301444428e-05}
{"train_loss": 0.03796824812889099, "global_step": 163180, "epoch": 1833, "lr": 3.44879618821506e-05}
{"train_loss": 0.13524191081523895, "global_step": 163181, "epoch": 1833, "lr": 3.448741075194235e-05}
{"train_loss": 0.08147819340229034, "global_step": 163182, "epoch": 1833, "lr": 3.448685962381956e-05}
{"train_loss": 0.060157615691423416, "global_step": 163183, "epoch": 1833, "lr": 3.448630849778234e-05}
{"train_loss": 0.03868113458156586, "global_step": 163184, "epoch": 1833, "lr": 3.4485757373830724e-05}
{"train_loss": 0.07582062482833862, "global_step": 163185, "epoch": 1833, "lr": 3.448520625196483e-05}
{"train_loss": 0.039267800748348236, "global_step": 163186, "epoch": 1833, "lr": 3.4484655132184705e-05}
{"train_loss": 0.1135496124625206, "global_step": 163187, "epoch": 1833, "lr": 3.4484104014490425e-05}
{"train_loss": 0.06479661911725998, "global_step": 163188, "epoch": 1833, "lr": 3.4483552898882075e-05}
{"train_loss": 0.07733020186424255, "global_step": 163189, "epoch": 1833, "lr": 3.4483001785359704e-05}
{"train_loss": 0.13938267529010773, "global_step": 163190, "epoch": 1833, "lr": 3.4482450673923425e-05}
{"train_loss": 0.1308416724205017, "global_step": 163191, "epoch": 1833, "lr": 3.448189956457328e-05}
{"train_loss": 0.1007939875125885, "global_step": 163192, "epoch": 1833, "lr": 3.448134845730937e-05}
{"train_loss": 0.08195357024669647, "global_step": 163193, "epoch": 1833, "lr": 3.448079735213173e-05}
{"train_loss": 0.05285535007715225, "global_step": 163194, "epoch": 1833, "lr": 3.448024624904048e-05}
{"train_loss": 0.09292816370725632, "global_step": 163195, "epoch": 1833, "lr": 3.447969514803565e-05}
{"train_loss": 0.0856521874666214, "global_step": 163196, "epoch": 1833, "lr": 3.447914404911736e-05}
{"train_loss": 0.06567936390638351, "global_step": 163197, "epoch": 1833, "lr": 3.447859295228564e-05}
{"train_loss": 0.07828610390424728, "global_step": 163198, "epoch": 1833, "lr": 3.44780418575406e-05}
{"train_loss": 0.12303540855646133, "global_step": 163199, "epoch": 1833, "lr": 3.4477490764882274e-05}
{"train_loss": 0.09732784330844879, "global_step": 163200, "epoch": 1833, "lr": 3.447693967431079e-05}
{"train_loss": 0.10047556459903717, "global_step": 163201, "epoch": 1833, "lr": 3.447638858582616e-05}
{"train_loss": 0.15201720595359802, "global_step": 163202, "epoch": 1833, "lr": 3.447583749942852e-05}
{"train_loss": 0.09006695449352264, "global_step": 163203, "epoch": 1833, "lr": 3.44752864151179e-05}
{"train_loss": 0.05993648245930672, "global_step": 163204, "epoch": 1833, "lr": 3.44747353328944e-05}
{"train_loss": 0.10495463013648987, "global_step": 163205, "epoch": 1833, "lr": 3.4474184252758066e-05}
{"train_loss": 0.08074285835027695, "global_step": 163206, "epoch": 1833, "lr": 3.4473633174709e-05}
{"train_loss": 0.0743975043296814, "global_step": 163207, "epoch": 1833, "lr": 3.447308209874726e-05}
{"train_loss": 0.07639234513044357, "global_step": 163208, "epoch": 1833, "lr": 3.4472531024872916e-05}
{"train_loss": 0.08990960568189621, "global_step": 163209, "epoch": 1833, "lr": 3.4471979953086074e-05}
{"train_loss": 0.07489296793937683, "global_step": 163210, "epoch": 1833, "lr": 3.447142888338676e-05}
{"train_loss": 0.06083030253648758, "global_step": 163211, "epoch": 1833, "lr": 3.447087781577509e-05}
{"train_loss": 0.09214135259389877, "global_step": 163212, "epoch": 1833, "lr": 3.447032675025112e-05}
{"train_loss": 0.08285146951675415, "global_step": 163213, "epoch": 1833, "lr": 3.446977568681492e-05}
{"train_loss": 0.10077907145023346, "global_step": 163214, "epoch": 1833, "lr": 3.446922462546657e-05}
{"train_loss": 0.08167950063943863, "global_step": 163215, "epoch": 1833, "lr": 3.446867356620615e-05}
{"train_loss": 0.08036968111991882, "global_step": 163216, "epoch": 1833, "lr": 3.446812250903371e-05}
{"train_loss": 0.050436753779649734, "global_step": 163217, "epoch": 1833, "lr": 3.446757145394937e-05}
{"train_loss": 0.0389973521232605, "global_step": 163218, "epoch": 1833, "lr": 3.4467020400953144e-05}
{"train_loss": 0.0848584920167923, "global_step": 163219, "epoch": 1833, "lr": 3.446646935004516e-05}
{"train_loss": 0.12465032935142517, "global_step": 163220, "epoch": 1833, "lr": 3.4465918301225456e-05}
{"train_loss": 0.09980955719947815, "global_step": 163221, "epoch": 1833, "lr": 3.446536725449414e-05}
{"train_loss": 0.08794352412223816, "global_step": 163222, "epoch": 1833, "lr": 3.446481620985123e-05}
{"train_loss": 0.04991902410984039, "global_step": 163223, "epoch": 1833, "lr": 3.446426516729688e-05}
{"train_loss": 0.06714422255754471, "global_step": 163224, "epoch": 1833, "lr": 3.44637141268311e-05}
{"train_loss": 0.0821963447365868, "global_step": 163225, "epoch": 1833, "lr": 3.446316308845398e-05, "val_loss": 7.012966632843018}
{"train_loss": 0.11157095432281494, "global_step": 163226, "epoch": 1834, "lr": 3.4462612052165605e-05}
{"train_loss": 0.13453318178653717, "global_step": 163227, "epoch": 1834, "lr": 3.446206101796603e-05}
{"train_loss": 0.10370168089866638, "global_step": 163228, "epoch": 1834, "lr": 3.446150998585536e-05}
{"train_loss": 0.075834721326828, "global_step": 163229, "epoch": 1834, "lr": 3.446095895583363e-05}
{"train_loss": 0.14052297174930573, "global_step": 163230, "epoch": 1834, "lr": 3.446040792790096e-05}
{"train_loss": 0.04399954155087471, "global_step": 163231, "epoch": 1834, "lr": 3.445985690205737e-05}
{"train_loss": 0.10569452494382858, "global_step": 163232, "epoch": 1834, "lr": 3.445930587830299e-05}
{"train_loss": 0.08089184015989304, "global_step": 163233, "epoch": 1834, "lr": 3.4458754856637844e-05}
{"train_loss": 0.09406425058841705, "global_step": 163234, "epoch": 1834, "lr": 3.4458203837062044e-05}
{"train_loss": 0.09392999857664108, "global_step": 163235, "epoch": 1834, "lr": 3.445765281957564e-05}
{"train_loss": 0.05347922816872597, "global_step": 163236, "epoch": 1834, "lr": 3.445710180417873e-05}
{"train_loss": 0.11573833227157593, "global_step": 163237, "epoch": 1834, "lr": 3.445655079087135e-05}
{"train_loss": 0.06696036458015442, "global_step": 163238, "epoch": 1834, "lr": 3.4455999779653625e-05}
{"train_loss": 0.0699053555727005, "global_step": 163239, "epoch": 1834, "lr": 3.445544877052557e-05}
{"train_loss": 0.04270675405859947, "global_step": 163240, "epoch": 1834, "lr": 3.445489776348732e-05}
{"train_loss": 0.10966504365205765, "global_step": 163241, "epoch": 1834, "lr": 3.4454346758538894e-05}
{"train_loss": 0.058985982090234756, "global_step": 163242, "epoch": 1834, "lr": 3.4453795755680415e-05}
{"train_loss": 0.10743401199579239, "global_step": 163243, "epoch": 1834, "lr": 3.4453244754911925e-05}
{"train_loss": 0.0782092735171318, "global_step": 163244, "epoch": 1834, "lr": 3.445269375623349e-05}
{"train_loss": 0.07919147610664368, "global_step": 163245, "epoch": 1834, "lr": 3.445214275964523e-05}
{"train_loss": 0.09780541062355042, "global_step": 163246, "epoch": 1834, "lr": 3.4451591765147165e-05}
{"train_loss": 0.11561044305562973, "global_step": 163247, "epoch": 1834, "lr": 3.445104077273942e-05}
{"train_loss": 0.08111000806093216, "global_step": 163248, "epoch": 1834, "lr": 3.4450489782422014e-05}
{"train_loss": 0.07765712589025497, "global_step": 163249, "epoch": 1834, "lr": 3.444993879419507e-05}
{"train_loss": 0.12623295187950134, "global_step": 163250, "epoch": 1834, "lr": 3.444938780805863e-05}
{"train_loss": 0.09362101554870605, "global_step": 163251, "epoch": 1834, "lr": 3.44488368240128e-05}
{"train_loss": 0.06515081226825714, "global_step": 163252, "epoch": 1834, "lr": 3.4448285842057604e-05}
{"train_loss": 0.09623680263757706, "global_step": 163253, "epoch": 1834, "lr": 3.444773486219317e-05}
{"train_loss": 0.06544289737939835, "global_step": 163254, "epoch": 1834, "lr": 3.444718388441953e-05}
{"train_loss": 0.08595195412635803, "global_step": 163255, "epoch": 1834, "lr": 3.44466329087368e-05}
{"train_loss": 0.06645719707012177, "global_step": 163256, "epoch": 1834, "lr": 3.444608193514501e-05}
{"train_loss": 0.08820575475692749, "global_step": 163257, "epoch": 1834, "lr": 3.4445530963644274e-05}
{"train_loss": 0.06511452794075012, "global_step": 163258, "epoch": 1834, "lr": 3.4444979994234625e-05}
{"train_loss": 0.11281609535217285, "global_step": 163259, "epoch": 1834, "lr": 3.4444429026916184e-05}
{"train_loss": 0.08675860613584518, "global_step": 163260, "epoch": 1834, "lr": 3.444387806168897e-05}
{"train_loss": 0.08577471226453781, "global_step": 163261, "epoch": 1834, "lr": 3.4443327098553113e-05}
{"train_loss": 0.05304298922419548, "global_step": 163262, "epoch": 1834, "lr": 3.444277613750865e-05}
{"train_loss": 0.0372130461037159, "global_step": 163263, "epoch": 1834, "lr": 3.444222517855566e-05}
{"train_loss": 0.09884393215179443, "global_step": 163264, "epoch": 1834, "lr": 3.444167422169424e-05}
{"train_loss": 0.15553484857082367, "global_step": 163265, "epoch": 1834, "lr": 3.444112326692442e-05}
{"train_loss": 0.0747290551662445, "global_step": 163266, "epoch": 1834, "lr": 3.444057231424633e-05}
{"train_loss": 0.16719280183315277, "global_step": 163267, "epoch": 1834, "lr": 3.444002136366e-05}
{"train_loss": 0.09056595712900162, "global_step": 163268, "epoch": 1834, "lr": 3.4439470415165534e-05}
{"train_loss": 0.06794992089271545, "global_step": 163269, "epoch": 1834, "lr": 3.443891946876297e-05}
{"train_loss": 0.15184538066387177, "global_step": 163270, "epoch": 1834, "lr": 3.443836852445243e-05}
{"train_loss": 0.06713519245386124, "global_step": 163271, "epoch": 1834, "lr": 3.443781758223393e-05}
{"train_loss": 0.08320251107215881, "global_step": 163272, "epoch": 1834, "lr": 3.44372666421076e-05}
{"train_loss": 0.10035402327775955, "global_step": 163273, "epoch": 1834, "lr": 3.4436715704073474e-05}
{"train_loss": 0.08679763227701187, "global_step": 163274, "epoch": 1834, "lr": 3.443616476813166e-05}
{"train_loss": 0.06455446034669876, "global_step": 163275, "epoch": 1834, "lr": 3.4435613834282196e-05}
{"train_loss": 0.10513072460889816, "global_step": 163276, "epoch": 1834, "lr": 3.4435062902525194e-05}
{"train_loss": 0.06876367330551147, "global_step": 163277, "epoch": 1834, "lr": 3.443451197286068e-05}
{"train_loss": 0.10746852308511734, "global_step": 163278, "epoch": 1834, "lr": 3.443396104528879e-05}
{"train_loss": 0.0829409584403038, "global_step": 163279, "epoch": 1834, "lr": 3.443341011980953e-05}
{"train_loss": 0.06260338425636292, "global_step": 163280, "epoch": 1834, "lr": 3.4432859196423025e-05}
{"train_loss": 0.12122607976198196, "global_step": 163281, "epoch": 1834, "lr": 3.443230827512933e-05}
{"train_loss": 0.1189684122800827, "global_step": 163282, "epoch": 1834, "lr": 3.443175735592853e-05}
{"train_loss": 0.09421421587467194, "global_step": 163283, "epoch": 1834, "lr": 3.44312064388207e-05}
{"train_loss": 0.10995866358280182, "global_step": 163284, "epoch": 1834, "lr": 3.443065552380587e-05}
{"train_loss": 0.08004869520664215, "global_step": 163285, "epoch": 1834, "lr": 3.443010461088418e-05}
{"train_loss": 0.09951832890510559, "global_step": 163286, "epoch": 1834, "lr": 3.442955370005565e-05}
{"train_loss": 0.05915429815649986, "global_step": 163287, "epoch": 1834, "lr": 3.4429002791320396e-05}
{"train_loss": 0.06669549643993378, "global_step": 163288, "epoch": 1834, "lr": 3.442845188467845e-05}
{"train_loss": 0.09302257001399994, "global_step": 163289, "epoch": 1834, "lr": 3.4427900980129934e-05}
{"train_loss": 0.11425377428531647, "global_step": 163290, "epoch": 1834, "lr": 3.442735007767487e-05}
{"train_loss": 0.043202515691518784, "global_step": 163291, "epoch": 1834, "lr": 3.442679917731338e-05}
{"train_loss": 0.12708492577075958, "global_step": 163292, "epoch": 1834, "lr": 3.442624827904551e-05}
{"train_loss": 0.06633679568767548, "global_step": 163293, "epoch": 1834, "lr": 3.442569738287135e-05}
{"train_loss": 0.10750938951969147, "global_step": 163294, "epoch": 1834, "lr": 3.442514648879095e-05}
{"train_loss": 0.14929352700710297, "global_step": 163295, "epoch": 1834, "lr": 3.4424595596804414e-05}
{"train_loss": 0.07793673872947693, "global_step": 163296, "epoch": 1834, "lr": 3.4424044706911785e-05}
{"train_loss": 0.08808368444442749, "global_step": 163297, "epoch": 1834, "lr": 3.442349381911317e-05}
{"train_loss": 0.06259920448064804, "global_step": 163298, "epoch": 1834, "lr": 3.442294293340861e-05}
{"train_loss": 0.09262017160654068, "global_step": 163299, "epoch": 1834, "lr": 3.44223920497982e-05}
{"train_loss": 0.1298196017742157, "global_step": 163300, "epoch": 1834, "lr": 3.4421841168282016e-05}
{"train_loss": 0.056683555245399475, "global_step": 163301, "epoch": 1834, "lr": 3.442129028886012e-05}
{"train_loss": 0.13448011875152588, "global_step": 163302, "epoch": 1834, "lr": 3.442073941153261e-05}
{"train_loss": 0.1415884643793106, "global_step": 163303, "epoch": 1834, "lr": 3.4420188536299514e-05}
{"train_loss": 0.1923246532678604, "global_step": 163304, "epoch": 1834, "lr": 3.441963766316096e-05}
{"train_loss": 0.06514402478933334, "global_step": 163305, "epoch": 1834, "lr": 3.4419086792116974e-05}
{"train_loss": 0.06489972025156021, "global_step": 163306, "epoch": 1834, "lr": 3.441853592316767e-05}
{"train_loss": 0.04991897568106651, "global_step": 163307, "epoch": 1834, "lr": 3.441798505631308e-05}
{"train_loss": 0.0907515212893486, "global_step": 163308, "epoch": 1834, "lr": 3.441743419155332e-05}
{"train_loss": 0.1095849797129631, "global_step": 163309, "epoch": 1834, "lr": 3.441688332888844e-05}
{"train_loss": 0.08641017228364944, "global_step": 163310, "epoch": 1834, "lr": 3.441633246831853e-05}
{"train_loss": 0.06845877319574356, "global_step": 163311, "epoch": 1834, "lr": 3.4415781609843635e-05}
{"train_loss": 0.05088847503066063, "global_step": 163312, "epoch": 1834, "lr": 3.4415230753463865e-05}
{"train_loss": 0.06800131499767303, "global_step": 163313, "epoch": 1834, "lr": 3.441467989917927e-05}
{"train_loss": 0.09030333621783203, "global_step": 163314, "epoch": 1834, "lr": 3.441412904698994e-05, "val_loss": 6.887457847595215}
{"train_loss": 0.05367092043161392, "global_step": 163315, "epoch": 1835, "lr": 3.4413578196895924e-05}
{"train_loss": 0.057133883237838745, "global_step": 163316, "epoch": 1835, "lr": 3.441302734889732e-05}
{"train_loss": 0.09308432042598724, "global_step": 163317, "epoch": 1835, "lr": 3.4412476502994196e-05}
{"train_loss": 0.10162001848220825, "global_step": 163318, "epoch": 1835, "lr": 3.441192565918663e-05}
{"train_loss": 0.07032732665538788, "global_step": 163319, "epoch": 1835, "lr": 3.441137481747468e-05}
{"train_loss": 0.09323840588331223, "global_step": 163320, "epoch": 1835, "lr": 3.4410823977858435e-05}
{"train_loss": 0.08015871047973633, "global_step": 163321, "epoch": 1835, "lr": 3.441027314033798e-05}
{"train_loss": 0.09157244116067886, "global_step": 163322, "epoch": 1835, "lr": 3.4409722304913345e-05}
{"train_loss": 0.10925190895795822, "global_step": 163323, "epoch": 1835, "lr": 3.4409171471584655e-05}
{"train_loss": 0.08493457734584808, "global_step": 163324, "epoch": 1835, "lr": 3.4408620640351944e-05}
{"train_loss": 0.13870227336883545, "global_step": 163325, "epoch": 1835, "lr": 3.440806981121533e-05}
{"train_loss": 0.062702976167202, "global_step": 163326, "epoch": 1835, "lr": 3.4407518984174833e-05}
{"train_loss": 0.09356185048818588, "global_step": 163327, "epoch": 1835, "lr": 3.4406968159230576e-05}
{"train_loss": 0.08061116188764572, "global_step": 163328, "epoch": 1835, "lr": 3.440641733638259e-05}
{"train_loss": 0.09623842686414719, "global_step": 163329, "epoch": 1835, "lr": 3.4405866515631004e-05}
{"train_loss": 0.1606745421886444, "global_step": 163330, "epoch": 1835, "lr": 3.4405315696975824e-05}
{"train_loss": 0.07383643835783005, "global_step": 163331, "epoch": 1835, "lr": 3.440476488041719e-05}
{"train_loss": 0.0718153640627861, "global_step": 163332, "epoch": 1835, "lr": 3.440421406595511e-05}
{"train_loss": 0.12062809616327286, "global_step": 163333, "epoch": 1835, "lr": 3.440366325358973e-05}
{"train_loss": 0.08969233930110931, "global_step": 163334, "epoch": 1835, "lr": 3.440311244332106e-05}
{"train_loss": 0.08460458368062973, "global_step": 163335, "epoch": 1835, "lr": 3.440256163514921e-05}
{"train_loss": 0.09680361300706863, "global_step": 163336, "epoch": 1835, "lr": 3.440201082907425e-05}
{"train_loss": 0.055694129317998886, "global_step": 163337, "epoch": 1835, "lr": 3.440146002509624e-05}
{"train_loss": 0.13173094391822815, "global_step": 163338, "epoch": 1835, "lr": 3.4400909223215275e-05}
{"train_loss": 0.07780560106039047, "global_step": 163339, "epoch": 1835, "lr": 3.440035842343141e-05}
{"train_loss": 0.05655984580516815, "global_step": 163340, "epoch": 1835, "lr": 3.4399807625744744e-05}
{"train_loss": 0.10250360518693924, "global_step": 163341, "epoch": 1835, "lr": 3.4399256830155315e-05}
{"train_loss": 0.10964880883693695, "global_step": 163342, "epoch": 1835, "lr": 3.439870603666323e-05}
{"train_loss": 0.0591139942407608, "global_step": 163343, "epoch": 1835, "lr": 3.439815524526853e-05}
{"train_loss": 0.0833468809723854, "global_step": 163344, "epoch": 1835, "lr": 3.439760445597132e-05}
{"train_loss": 0.11296951025724411, "global_step": 163345, "epoch": 1835, "lr": 3.439705366877165e-05}
{"train_loss": 0.07884196192026138, "global_step": 163346, "epoch": 1835, "lr": 3.439650288366963e-05}
{"train_loss": 0.10007377713918686, "global_step": 163347, "epoch": 1835, "lr": 3.439595210066529e-05}
{"train_loss": 0.13667410612106323, "global_step": 163348, "epoch": 1835, "lr": 3.4395401319758734e-05}
{"train_loss": 0.12630492448806763, "global_step": 163349, "epoch": 1835, "lr": 3.439485054095001e-05}
{"train_loss": 0.09339658170938492, "global_step": 163350, "epoch": 1835, "lr": 3.439429976423923e-05}
{"train_loss": 0.0776250958442688, "global_step": 163351, "epoch": 1835, "lr": 3.4393748989626425e-05}
{"train_loss": 0.08921696245670319, "global_step": 163352, "epoch": 1835, "lr": 3.439319821711171e-05}
{"train_loss": 0.09379962086677551, "global_step": 163353, "epoch": 1835, "lr": 3.439264744669512e-05}
{"train_loss": 0.07164229452610016, "global_step": 163354, "epoch": 1835, "lr": 3.439209667837676e-05}
{"train_loss": 0.10727377235889435, "global_step": 163355, "epoch": 1835, "lr": 3.439154591215668e-05}
{"train_loss": 0.14635153114795685, "global_step": 163356, "epoch": 1835, "lr": 3.439099514803499e-05}
{"train_loss": 0.07720918953418732, "global_step": 163357, "epoch": 1835, "lr": 3.439044438601172e-05}
{"train_loss": 0.06464763730764389, "global_step": 163358, "epoch": 1835, "lr": 3.438989362608698e-05}
{"train_loss": 0.13640648126602173, "global_step": 163359, "epoch": 1835, "lr": 3.438934286826083e-05}
{"train_loss": 0.07036016881465912, "global_step": 163360, "epoch": 1835, "lr": 3.438879211253332e-05}
{"train_loss": 0.10780894756317139, "global_step": 163361, "epoch": 1835, "lr": 3.438824135890457e-05}
{"train_loss": 0.08210106194019318, "global_step": 163362, "epoch": 1835, "lr": 3.438769060737461e-05}
{"train_loss": 0.09888496994972229, "global_step": 163363, "epoch": 1835, "lr": 3.438713985794356e-05}
{"train_loss": 0.2121196836233139, "global_step": 163364, "epoch": 1835, "lr": 3.438658911061144e-05}
{"train_loss": 0.10909685492515564, "global_step": 163365, "epoch": 1835, "lr": 3.438603836537838e-05}
{"train_loss": 0.04225137084722519, "global_step": 163366, "epoch": 1835, "lr": 3.43854876222444e-05}
{"train_loss": 0.09736256301403046, "global_step": 163367, "epoch": 1835, "lr": 3.438493688120963e-05}
{"train_loss": 0.09380314499139786, "global_step": 163368, "epoch": 1835, "lr": 3.438438614227408e-05}
{"train_loss": 0.10449597984552383, "global_step": 163369, "epoch": 1835, "lr": 3.4383835405437894e-05}
{"train_loss": 0.0894213616847992, "global_step": 163370, "epoch": 1835, "lr": 3.438328467070109e-05}
{"train_loss": 0.03561670705676079, "global_step": 163371, "epoch": 1835, "lr": 3.438273393806375e-05}
{"train_loss": 0.09011033177375793, "global_step": 163372, "epoch": 1835, "lr": 3.438218320752599e-05}
{"train_loss": 0.12930095195770264, "global_step": 163373, "epoch": 1835, "lr": 3.438163247908784e-05}
{"train_loss": 0.10718049854040146, "global_step": 163374, "epoch": 1835, "lr": 3.43810817527494e-05}
{"train_loss": 0.06115325912833214, "global_step": 163375, "epoch": 1835, "lr": 3.4380531028510735e-05}
{"train_loss": 0.21963751316070557, "global_step": 163376, "epoch": 1835, "lr": 3.43799803063719e-05}
{"train_loss": 0.09619186818599701, "global_step": 163377, "epoch": 1835, "lr": 3.4379429586333e-05}
{"train_loss": 0.11553537100553513, "global_step": 163378, "epoch": 1835, "lr": 3.43788788683941e-05}
{"train_loss": 0.18207553029060364, "global_step": 163379, "epoch": 1835, "lr": 3.437832815255525e-05}
{"train_loss": 0.09687657654285431, "global_step": 163380, "epoch": 1835, "lr": 3.4377777438816564e-05}
{"train_loss": 0.09949812293052673, "global_step": 163381, "epoch": 1835, "lr": 3.4377226727178074e-05}
{"train_loss": 0.10006143897771835, "global_step": 163382, "epoch": 1835, "lr": 3.4376676017639895e-05}
{"train_loss": 0.04599956423044205, "global_step": 163383, "epoch": 1835, "lr": 3.4376125310202066e-05}
{"train_loss": 0.0423988439142704, "global_step": 163384, "epoch": 1835, "lr": 3.437557460486469e-05}
{"train_loss": 0.09057566523551941, "global_step": 163385, "epoch": 1835, "lr": 3.437502390162781e-05}
{"train_loss": 0.10066280514001846, "global_step": 163386, "epoch": 1835, "lr": 3.437447320049154e-05}
{"train_loss": 0.10820630192756653, "global_step": 163387, "epoch": 1835, "lr": 3.437392250145591e-05}
{"train_loss": 0.07290641218423843, "global_step": 163388, "epoch": 1835, "lr": 3.437337180452104e-05}
{"train_loss": 0.10795728117227554, "global_step": 163389, "epoch": 1835, "lr": 3.437282110968695e-05}
{"train_loss": 0.11391475051641464, "global_step": 163390, "epoch": 1835, "lr": 3.437227041695375e-05}
{"train_loss": 0.10982973873615265, "global_step": 163391, "epoch": 1835, "lr": 3.437171972632153e-05}
{"train_loss": 0.05458620563149452, "global_step": 163392, "epoch": 1835, "lr": 3.437116903779033e-05}
{"train_loss": 0.08332555741071701, "global_step": 163393, "epoch": 1835, "lr": 3.4370618351360226e-05}
{"train_loss": 0.12089825421571732, "global_step": 163394, "epoch": 1835, "lr": 3.437006766703131e-05}
{"train_loss": 0.11876694113016129, "global_step": 163395, "epoch": 1835, "lr": 3.4369516984803656e-05}
{"train_loss": 0.12296388298273087, "global_step": 163396, "epoch": 1835, "lr": 3.436896630467732e-05}
{"train_loss": 0.09199127554893494, "global_step": 163397, "epoch": 1835, "lr": 3.436841562665239e-05}
{"train_loss": 0.10830778628587723, "global_step": 163398, "epoch": 1835, "lr": 3.436786495072893e-05}
{"train_loss": 0.10055854916572571, "global_step": 163399, "epoch": 1835, "lr": 3.4367314276907034e-05}
{"train_loss": 0.11801432818174362, "global_step": 163400, "epoch": 1835, "lr": 3.436676360518675e-05}
{"train_loss": 0.09135285764932632, "global_step": 163401, "epoch": 1835, "lr": 3.4366212935568174e-05}
{"train_loss": 0.060010526329278946, "global_step": 163402, "epoch": 1835, "lr": 3.436566226805135e-05}
{"train_loss": 0.0963225941477197, "global_step": 163403, "epoch": 1835, "lr": 3.43651116026364e-05, "val_loss": 6.911185264587402, "train_action_mse_error": 31.708541870117188}
{"train_loss": 0.09690305590629578, "global_step": 163404, "epoch": 1836, "lr": 3.436456093932334e-05}
{"train_loss": 0.05163314938545227, "global_step": 163405, "epoch": 1836, "lr": 3.43640102781123e-05}
{"train_loss": 0.08038199692964554, "global_step": 163406, "epoch": 1836, "lr": 3.4363459619003314e-05}
{"train_loss": 0.09176148474216461, "global_step": 163407, "epoch": 1836, "lr": 3.436290896199646e-05}
{"train_loss": 0.08133654296398163, "global_step": 163408, "epoch": 1836, "lr": 3.436235830709185e-05}
{"train_loss": 0.09954002499580383, "global_step": 163409, "epoch": 1836, "lr": 3.43618076542895e-05}
{"train_loss": 0.07519997656345367, "global_step": 163410, "epoch": 1836, "lr": 3.4361257003589544e-05}
{"train_loss": 0.10845977813005447, "global_step": 163411, "epoch": 1836, "lr": 3.4360706354992004e-05}
{"train_loss": 0.07989973574876785, "global_step": 163412, "epoch": 1836, "lr": 3.4360155708496996e-05}
{"train_loss": 0.054557278752326965, "global_step": 163413, "epoch": 1836, "lr": 3.435960506410456e-05}
{"train_loss": 0.05981140211224556, "global_step": 163414, "epoch": 1836, "lr": 3.435905442181479e-05}
{"train_loss": 0.09643461555242538, "global_step": 163415, "epoch": 1836, "lr": 3.435850378162775e-05}
{"train_loss": 0.12100106477737427, "global_step": 163416, "epoch": 1836, "lr": 3.435795314354352e-05}
{"train_loss": 0.11524087935686111, "global_step": 163417, "epoch": 1836, "lr": 3.4357402507562176e-05}
{"train_loss": 0.11025042086839676, "global_step": 163418, "epoch": 1836, "lr": 3.43568518736838e-05}
{"train_loss": 0.04994044452905655, "global_step": 163419, "epoch": 1836, "lr": 3.4356301241908427e-05}
{"train_loss": 0.13493944704532623, "global_step": 163420, "epoch": 1836, "lr": 3.4355750612236184e-05}
{"train_loss": 0.04430123791098595, "global_step": 163421, "epoch": 1836, "lr": 3.43551999846671e-05}
{"train_loss": 0.13078588247299194, "global_step": 163422, "epoch": 1836, "lr": 3.435464935920129e-05}
{"train_loss": 0.136557474732399, "global_step": 163423, "epoch": 1836, "lr": 3.435409873583878e-05}
{"train_loss": 0.08457695692777634, "global_step": 163424, "epoch": 1836, "lr": 3.435354811457969e-05}
{"train_loss": 0.05299399420619011, "global_step": 163425, "epoch": 1836, "lr": 3.435299749542407e-05}
{"train_loss": 0.09866819530725479, "global_step": 163426, "epoch": 1836, "lr": 3.435244687837198e-05}
{"train_loss": 0.12114088237285614, "global_step": 163427, "epoch": 1836, "lr": 3.435189626342354e-05}
{"train_loss": 0.07861415296792984, "global_step": 163428, "epoch": 1836, "lr": 3.435134565057878e-05}
{"train_loss": 0.14316704869270325, "global_step": 163429, "epoch": 1836, "lr": 3.435079503983781e-05}
{"train_loss": 0.12046154588460922, "global_step": 163430, "epoch": 1836, "lr": 3.4350244431200664e-05}
{"train_loss": 0.10120212286710739, "global_step": 163431, "epoch": 1836, "lr": 3.434969382466746e-05}
{"train_loss": 0.07499448210000992, "global_step": 163432, "epoch": 1836, "lr": 3.434914322023823e-05}
{"train_loss": 0.07530751079320908, "global_step": 163433, "epoch": 1836, "lr": 3.434859261791307e-05}
{"train_loss": 0.1101146787405014, "global_step": 163434, "epoch": 1836, "lr": 3.434804201769206e-05}
{"train_loss": 0.10336005687713623, "global_step": 163435, "epoch": 1836, "lr": 3.434749141957525e-05}
{"train_loss": 0.1501014232635498, "global_step": 163436, "epoch": 1836, "lr": 3.434694082356274e-05}
{"train_loss": 0.15316618978977203, "global_step": 163437, "epoch": 1836, "lr": 3.43463902296546e-05}
{"train_loss": 0.11318621039390564, "global_step": 163438, "epoch": 1836, "lr": 3.434583963785088e-05}
{"train_loss": 0.09767761826515198, "global_step": 163439, "epoch": 1836, "lr": 3.434528904815168e-05}
{"train_loss": 0.08362799882888794, "global_step": 163440, "epoch": 1836, "lr": 3.4344738460557056e-05}
{"train_loss": 0.052448540925979614, "global_step": 163441, "epoch": 1836, "lr": 3.4344187875067115e-05}
{"train_loss": 0.07397565245628357, "global_step": 163442, "epoch": 1836, "lr": 3.4343637291681876e-05}
{"train_loss": 0.09400001913309097, "global_step": 163443, "epoch": 1836, "lr": 3.434308671040145e-05}
{"train_loss": 0.08653557300567627, "global_step": 163444, "epoch": 1836, "lr": 3.434253613122592e-05}
{"train_loss": 0.08656147122383118, "global_step": 163445, "epoch": 1836, "lr": 3.434198555415533e-05}
{"train_loss": 0.06737887114286423, "global_step": 163446, "epoch": 1836, "lr": 3.434143497918979e-05}
{"train_loss": 0.08951269835233688, "global_step": 163447, "epoch": 1836, "lr": 3.434088440632933e-05}
{"train_loss": 0.07643373310565948, "global_step": 163448, "epoch": 1836, "lr": 3.4340333835574066e-05}
{"train_loss": 0.06091577187180519, "global_step": 163449, "epoch": 1836, "lr": 3.433978326692404e-05}
{"train_loss": 0.08827053010463715, "global_step": 163450, "epoch": 1836, "lr": 3.4339232700379356e-05}
{"train_loss": 0.07361743599176407, "global_step": 163451, "epoch": 1836, "lr": 3.433868213594006e-05}
{"train_loss": 0.07749480754137039, "global_step": 163452, "epoch": 1836, "lr": 3.433813157360623e-05}
{"train_loss": 0.09451937675476074, "global_step": 163453, "epoch": 1836, "lr": 3.433758101337796e-05}
{"train_loss": 0.08639460057020187, "global_step": 163454, "epoch": 1836, "lr": 3.4337030455255315e-05}
{"train_loss": 0.05408095195889473, "global_step": 163455, "epoch": 1836, "lr": 3.433647989923835e-05}
{"train_loss": 0.06547548621892929, "global_step": 163456, "epoch": 1836, "lr": 3.433592934532718e-05}
{"train_loss": 0.05145782232284546, "global_step": 163457, "epoch": 1836, "lr": 3.433537879352182e-05}
{"train_loss": 0.09538328647613525, "global_step": 163458, "epoch": 1836, "lr": 3.43348282438224e-05}
{"train_loss": 0.04812964051961899, "global_step": 163459, "epoch": 1836, "lr": 3.4334277696228954e-05}
{"train_loss": 0.1615961492061615, "global_step": 163460, "epoch": 1836, "lr": 3.433372715074159e-05}
{"train_loss": 0.10457338392734528, "global_step": 163461, "epoch": 1836, "lr": 3.433317660736035e-05}
{"train_loss": 0.12040695548057556, "global_step": 163462, "epoch": 1836, "lr": 3.433262606608532e-05}
{"train_loss": 0.07497908174991608, "global_step": 163463, "epoch": 1836, "lr": 3.43320755269166e-05}
{"train_loss": 0.0780893862247467, "global_step": 163464, "epoch": 1836, "lr": 3.433152498985422e-05}
{"train_loss": 0.09560249745845795, "global_step": 163465, "epoch": 1836, "lr": 3.43309744548983e-05}
{"train_loss": 0.1291804015636444, "global_step": 163466, "epoch": 1836, "lr": 3.433042392204886e-05}
{"train_loss": 0.10958582907915115, "global_step": 163467, "epoch": 1836, "lr": 3.432987339130603e-05}
{"train_loss": 0.14099761843681335, "global_step": 163468, "epoch": 1836, "lr": 3.4329322862669836e-05}
{"train_loss": 0.1458815634250641, "global_step": 163469, "epoch": 1836, "lr": 3.4328772336140396e-05}
{"train_loss": 0.12603537738323212, "global_step": 163470, "epoch": 1836, "lr": 3.432822181171775e-05}
{"train_loss": 0.08125199377536774, "global_step": 163471, "epoch": 1836, "lr": 3.432767128940199e-05}
{"train_loss": 0.04730907455086708, "global_step": 163472, "epoch": 1836, "lr": 3.4327120769193174e-05}
{"train_loss": 0.11461635679006577, "global_step": 163473, "epoch": 1836, "lr": 3.4326570251091386e-05}
{"train_loss": 0.09280769526958466, "global_step": 163474, "epoch": 1836, "lr": 3.4326019735096705e-05}
{"train_loss": 0.033371470868587494, "global_step": 163475, "epoch": 1836, "lr": 3.4325469221209204e-05}
{"train_loss": 0.10295877605676651, "global_step": 163476, "epoch": 1836, "lr": 3.432491870942893e-05}
{"train_loss": 0.07052092999219894, "global_step": 163477, "epoch": 1836, "lr": 3.432436819975601e-05}
{"train_loss": 0.11403137445449829, "global_step": 163478, "epoch": 1836, "lr": 3.4323817692190464e-05}
{"train_loss": 0.09507285803556442, "global_step": 163479, "epoch": 1836, "lr": 3.432326718673241e-05}
{"train_loss": 0.11681126803159714, "global_step": 163480, "epoch": 1836, "lr": 3.432271668338187e-05}
{"train_loss": 0.10175982117652893, "global_step": 163481, "epoch": 1836, "lr": 3.432216618213897e-05}
{"train_loss": 0.1618972271680832, "global_step": 163482, "epoch": 1836, "lr": 3.432161568300377e-05}
{"train_loss": 0.11528841406106949, "global_step": 163483, "epoch": 1836, "lr": 3.432106518597632e-05}
{"train_loss": 0.07422526925802231, "global_step": 163484, "epoch": 1836, "lr": 3.432051469105673e-05}
{"train_loss": 0.09655343741178513, "global_step": 163485, "epoch": 1836, "lr": 3.4319964198245044e-05}
{"train_loss": 0.11364829540252686, "global_step": 163486, "epoch": 1836, "lr": 3.431941370754136e-05}
{"train_loss": 0.17913879454135895, "global_step": 163487, "epoch": 1836, "lr": 3.431886321894572e-05}
{"train_loss": 0.12246375530958176, "global_step": 163488, "epoch": 1836, "lr": 3.431831273245824e-05}
{"train_loss": 0.1120486631989479, "global_step": 163489, "epoch": 1836, "lr": 3.4317762248078964e-05}
{"train_loss": 0.08695746213197708, "global_step": 163490, "epoch": 1836, "lr": 3.4317211765807964e-05}
{"train_loss": 0.11061422526836395, "global_step": 163491, "epoch": 1836, "lr": 3.431666128564534e-05}
{"train_loss": 0.09621758658564493, "global_step": 163492, "epoch": 1836, "lr": 3.431611080759113e-05, "val_loss": 7.001155853271484}
{"train_loss": 0.06340005993843079, "global_step": 163493, "epoch": 1837, "lr": 3.431556033164545e-05}
{"train_loss": 0.048028312623500824, "global_step": 163494, "epoch": 1837, "lr": 3.431500985780834e-05}
{"train_loss": 0.10680459439754486, "global_step": 163495, "epoch": 1837, "lr": 3.431445938607988e-05}
{"train_loss": 0.13108313083648682, "global_step": 163496, "epoch": 1837, "lr": 3.4313908916460164e-05}
{"train_loss": 0.12402452528476715, "global_step": 163497, "epoch": 1837, "lr": 3.431335844894923e-05}
{"train_loss": 0.11048982292413712, "global_step": 163498, "epoch": 1837, "lr": 3.431280798354718e-05}
{"train_loss": 0.07439183443784714, "global_step": 163499, "epoch": 1837, "lr": 3.4312257520254097e-05}
{"train_loss": 0.13578662276268005, "global_step": 163500, "epoch": 1837, "lr": 3.431170705907002e-05}
{"train_loss": 0.049890656024217606, "global_step": 163501, "epoch": 1837, "lr": 3.431115659999507e-05}
{"train_loss": 0.08796993643045425, "global_step": 163502, "epoch": 1837, "lr": 3.4310606143029265e-05}
{"train_loss": 0.09467379748821259, "global_step": 163503, "epoch": 1837, "lr": 3.431005568817272e-05}
{"train_loss": 0.12933672964572906, "global_step": 163504, "epoch": 1837, "lr": 3.43095052354255e-05}
{"train_loss": 0.11102140694856644, "global_step": 163505, "epoch": 1837, "lr": 3.430895478478768e-05}
{"train_loss": 0.06848353892564774, "global_step": 163506, "epoch": 1837, "lr": 3.4308404336259316e-05}
{"train_loss": 0.08518674969673157, "global_step": 163507, "epoch": 1837, "lr": 3.430785388984051e-05}
{"train_loss": 0.11852844059467316, "global_step": 163508, "epoch": 1837, "lr": 3.430730344553131e-05}
{"train_loss": 0.07065457105636597, "global_step": 163509, "epoch": 1837, "lr": 3.4306753003331814e-05}
{"train_loss": 0.08624667674303055, "global_step": 163510, "epoch": 1837, "lr": 3.430620256324207e-05}
{"train_loss": 0.05017789453268051, "global_step": 163511, "epoch": 1837, "lr": 3.4305652125262174e-05}
{"train_loss": 0.051170602440834045, "global_step": 163512, "epoch": 1837, "lr": 3.4305101689392195e-05}
{"train_loss": 0.2312137633562088, "global_step": 163513, "epoch": 1837, "lr": 3.4304551255632195e-05}
{"train_loss": 0.07429108768701553, "global_step": 163514, "epoch": 1837, "lr": 3.430400082398226e-05}
{"train_loss": 0.07650104910135269, "global_step": 163515, "epoch": 1837, "lr": 3.430345039444247e-05}
{"train_loss": 0.09105837345123291, "global_step": 163516, "epoch": 1837, "lr": 3.430289996701287e-05}
{"train_loss": 0.10939714312553406, "global_step": 163517, "epoch": 1837, "lr": 3.4302349541693554e-05}
{"train_loss": 0.10631980746984482, "global_step": 163518, "epoch": 1837, "lr": 3.4301799118484623e-05}
{"train_loss": 0.09573575109243393, "global_step": 163519, "epoch": 1837, "lr": 3.430124869738609e-05}
{"train_loss": 0.09951920062303543, "global_step": 163520, "epoch": 1837, "lr": 3.430069827839809e-05}
{"train_loss": 0.06044367700815201, "global_step": 163521, "epoch": 1837, "lr": 3.430014786152065e-05}
{"train_loss": 0.09384094178676605, "global_step": 163522, "epoch": 1837, "lr": 3.4299597446753885e-05}
{"train_loss": 0.060784827917814255, "global_step": 163523, "epoch": 1837, "lr": 3.429904703409783e-05}
{"train_loss": 0.05198787525296211, "global_step": 163524, "epoch": 1837, "lr": 3.429849662355259e-05}
{"train_loss": 0.058722708374261856, "global_step": 163525, "epoch": 1837, "lr": 3.429794621511821e-05}
{"train_loss": 0.10655273497104645, "global_step": 163526, "epoch": 1837, "lr": 3.42973958087948e-05}
{"train_loss": 0.09245414286851883, "global_step": 163527, "epoch": 1837, "lr": 3.429684540458239e-05}
{"train_loss": 0.07645805180072784, "global_step": 163528, "epoch": 1837, "lr": 3.42962950024811e-05}
{"train_loss": 0.05069540813565254, "global_step": 163529, "epoch": 1837, "lr": 3.4295744602490976e-05}
{"train_loss": 0.12889757752418518, "global_step": 163530, "epoch": 1837, "lr": 3.429519420461209e-05}
{"train_loss": 0.09087824076414108, "global_step": 163531, "epoch": 1837, "lr": 3.429464380884453e-05}
{"train_loss": 0.13065887987613678, "global_step": 163532, "epoch": 1837, "lr": 3.429409341518835e-05}
{"train_loss": 0.045502133667469025, "global_step": 163533, "epoch": 1837, "lr": 3.429354302364366e-05}
{"train_loss": 0.08112241327762604, "global_step": 163534, "epoch": 1837, "lr": 3.429299263421049e-05}
{"train_loss": 0.05603126436471939, "global_step": 163535, "epoch": 1837, "lr": 3.4292442246888954e-05}
{"train_loss": 0.07488101720809937, "global_step": 163536, "epoch": 1837, "lr": 3.429189186167909e-05}
{"train_loss": 0.10487920790910721, "global_step": 163537, "epoch": 1837, "lr": 3.429134147858101e-05}
{"train_loss": 0.05411899834871292, "global_step": 163538, "epoch": 1837, "lr": 3.4290791097594746e-05}
{"train_loss": 0.08969229459762573, "global_step": 163539, "epoch": 1837, "lr": 3.4290240718720414e-05}
{"train_loss": 0.09214885532855988, "global_step": 163540, "epoch": 1837, "lr": 3.428969034195805e-05}
{"train_loss": 0.10686615109443665, "global_step": 163541, "epoch": 1837, "lr": 3.428913996730776e-05}
{"train_loss": 0.10682857781648636, "global_step": 163542, "epoch": 1837, "lr": 3.428858959476959e-05}
{"train_loss": 0.05346207693219185, "global_step": 163543, "epoch": 1837, "lr": 3.428803922434365e-05}
{"train_loss": 0.0756511464715004, "global_step": 163544, "epoch": 1837, "lr": 3.428748885602996e-05}
{"train_loss": 0.04830881580710411, "global_step": 163545, "epoch": 1837, "lr": 3.428693848982866e-05}
{"train_loss": 0.12149015069007874, "global_step": 163546, "epoch": 1837, "lr": 3.4286388125739755e-05}
{"train_loss": 0.12000247836112976, "global_step": 163547, "epoch": 1837, "lr": 3.4285837763763385e-05}
{"train_loss": 0.07993802428245544, "global_step": 163548, "epoch": 1837, "lr": 3.4285287403899576e-05}
{"train_loss": 0.06933492422103882, "global_step": 163549, "epoch": 1837, "lr": 3.428473704614842e-05}
{"train_loss": 0.07825890183448792, "global_step": 163550, "epoch": 1837, "lr": 3.428418669050999e-05}
{"train_loss": 0.07335244119167328, "global_step": 163551, "epoch": 1837, "lr": 3.428363633698436e-05}
{"train_loss": 0.08067796379327774, "global_step": 163552, "epoch": 1837, "lr": 3.4283085985571606e-05}
{"train_loss": 0.14230242371559143, "global_step": 163553, "epoch": 1837, "lr": 3.428253563627178e-05}
{"train_loss": 0.08321861922740936, "global_step": 163554, "epoch": 1837, "lr": 3.4281985289085007e-05}
{"train_loss": 0.06308785825967789, "global_step": 163555, "epoch": 1837, "lr": 3.42814349440113e-05}
{"train_loss": 0.11481255292892456, "global_step": 163556, "epoch": 1837, "lr": 3.4280884601050793e-05}
{"train_loss": 0.08222749829292297, "global_step": 163557, "epoch": 1837, "lr": 3.4280334260203504e-05}
{"train_loss": 0.11271483451128006, "global_step": 163558, "epoch": 1837, "lr": 3.4279783921469556e-05}
{"train_loss": 0.1141917034983635, "global_step": 163559, "epoch": 1837, "lr": 3.4279233584848975e-05}
{"train_loss": 0.11172353476285934, "global_step": 163560, "epoch": 1837, "lr": 3.427868325034188e-05}
{"train_loss": 0.07269122451543808, "global_step": 163561, "epoch": 1837, "lr": 3.42781329179483e-05}
{"train_loss": 0.09732850641012192, "global_step": 163562, "epoch": 1837, "lr": 3.4277582587668365e-05}
{"train_loss": 0.14874054491519928, "global_step": 163563, "epoch": 1837, "lr": 3.427703225950209e-05}
{"train_loss": 0.0982644185423851, "global_step": 163564, "epoch": 1837, "lr": 3.42764819334496e-05}
{"train_loss": 0.09356564283370972, "global_step": 163565, "epoch": 1837, "lr": 3.4275931609510916e-05}
{"train_loss": 0.04189571738243103, "global_step": 163566, "epoch": 1837, "lr": 3.427538128768617e-05}
{"train_loss": 0.11368894577026367, "global_step": 163567, "epoch": 1837, "lr": 3.427483096797539e-05}
{"train_loss": 0.14076581597328186, "global_step": 163568, "epoch": 1837, "lr": 3.4274280650378673e-05}
{"train_loss": 0.1090988740324974, "global_step": 163569, "epoch": 1837, "lr": 3.427373033489609e-05}
{"train_loss": 0.08591283857822418, "global_step": 163570, "epoch": 1837, "lr": 3.4273180021527696e-05}
{"train_loss": 0.11386867612600327, "global_step": 163571, "epoch": 1837, "lr": 3.427262971027361e-05}
{"train_loss": 0.08451207727193832, "global_step": 163572, "epoch": 1837, "lr": 3.427207940113385e-05}
{"train_loss": 0.0732492059469223, "global_step": 163573, "epoch": 1837, "lr": 3.4271529094108537e-05}
{"train_loss": 0.08677402138710022, "global_step": 163574, "epoch": 1837, "lr": 3.4270978789197704e-05}
{"train_loss": 0.0760609582066536, "global_step": 163575, "epoch": 1837, "lr": 3.427042848640147e-05}
{"train_loss": 0.15942049026489258, "global_step": 163576, "epoch": 1837, "lr": 3.426987818571987e-05}
{"train_loss": 0.06392781436443329, "global_step": 163577, "epoch": 1837, "lr": 3.4269327887153e-05}
{"train_loss": 0.12679950892925262, "global_step": 163578, "epoch": 1837, "lr": 3.426877759070092e-05}
{"train_loss": 0.12573790550231934, "global_step": 163579, "epoch": 1837, "lr": 3.426822729636373e-05}
{"train_loss": 0.09474261850118637, "global_step": 163580, "epoch": 1837, "lr": 3.426767700414145e-05}
{"train_loss": 0.09205480083153489, "global_step": 163581, "epoch": 1837, "lr": 3.4267126714034226e-05, "val_loss": 6.999309062957764}
{"train_loss": 0.06696083396673203, "global_step": 163582, "epoch": 1838, "lr": 3.426657642604207e-05}
{"train_loss": 0.09569676965475082, "global_step": 163583, "epoch": 1838, "lr": 3.42660261401651e-05}
{"train_loss": 0.07315175235271454, "global_step": 163584, "epoch": 1838, "lr": 3.426547585640335e-05}
{"train_loss": 0.0650968849658966, "global_step": 163585, "epoch": 1838, "lr": 3.426492557475694e-05}
{"train_loss": 0.08161642402410507, "global_step": 163586, "epoch": 1838, "lr": 3.42643752952259e-05}
{"train_loss": 0.08966775238513947, "global_step": 163587, "epoch": 1838, "lr": 3.426382501781032e-05}
{"train_loss": 0.05983072891831398, "global_step": 163588, "epoch": 1838, "lr": 3.426327474251029e-05}
{"train_loss": 0.09695181995630264, "global_step": 163589, "epoch": 1838, "lr": 3.426272446932586e-05}
{"train_loss": 0.10108068585395813, "global_step": 163590, "epoch": 1838, "lr": 3.426217419825713e-05}
{"train_loss": 0.12656712532043457, "global_step": 163591, "epoch": 1838, "lr": 3.426162392930414e-05}
{"train_loss": 0.08665464073419571, "global_step": 163592, "epoch": 1838, "lr": 3.4261073662467005e-05}
{"train_loss": 0.10892539471387863, "global_step": 163593, "epoch": 1838, "lr": 3.4260523397745745e-05}
{"train_loss": 0.09616288542747498, "global_step": 163594, "epoch": 1838, "lr": 3.42599731351405e-05}
{"train_loss": 0.13439764082431793, "global_step": 163595, "epoch": 1838, "lr": 3.425942287465128e-05}
{"train_loss": 0.1450606733560562, "global_step": 163596, "epoch": 1838, "lr": 3.4258872616278216e-05}
{"train_loss": 0.07726582884788513, "global_step": 163597, "epoch": 1838, "lr": 3.4258322360021336e-05}
{"train_loss": 0.05651954934000969, "global_step": 163598, "epoch": 1838, "lr": 3.425777210588075e-05}
{"train_loss": 0.07865199446678162, "global_step": 163599, "epoch": 1838, "lr": 3.4257221853856494e-05}
{"train_loss": 0.09279293566942215, "global_step": 163600, "epoch": 1838, "lr": 3.4256671603948675e-05}
{"train_loss": 0.07922716438770294, "global_step": 163601, "epoch": 1838, "lr": 3.425612135615735e-05}
{"train_loss": 0.0784466341137886, "global_step": 163602, "epoch": 1838, "lr": 3.4255571110482605e-05}
{"train_loss": 0.06248299032449722, "global_step": 163603, "epoch": 1838, "lr": 3.425502086692449e-05}
{"train_loss": 0.11149606108665466, "global_step": 163604, "epoch": 1838, "lr": 3.425447062548312e-05}
{"train_loss": 0.09718931466341019, "global_step": 163605, "epoch": 1838, "lr": 3.425392038615853e-05}
{"train_loss": 0.0961860865354538, "global_step": 163606, "epoch": 1838, "lr": 3.425337014895081e-05}
{"train_loss": 0.12015426903963089, "global_step": 163607, "epoch": 1838, "lr": 3.425281991386003e-05}
{"train_loss": 0.08439528942108154, "global_step": 163608, "epoch": 1838, "lr": 3.4252269680886255e-05}
{"train_loss": 0.11266113072633743, "global_step": 163609, "epoch": 1838, "lr": 3.4251719450029596e-05}
{"train_loss": 0.1079220250248909, "global_step": 163610, "epoch": 1838, "lr": 3.425116922129008e-05}
{"train_loss": 0.19609588384628296, "global_step": 163611, "epoch": 1838, "lr": 3.425061899466782e-05}
{"train_loss": 0.04726621136069298, "global_step": 163612, "epoch": 1838, "lr": 3.4250068770162856e-05}
{"train_loss": 0.09368635714054108, "global_step": 163613, "epoch": 1838, "lr": 3.424951854777529e-05}
{"train_loss": 0.08377386629581451, "global_step": 163614, "epoch": 1838, "lr": 3.424896832750517e-05}
{"train_loss": 0.0547841377556324, "global_step": 163615, "epoch": 1838, "lr": 3.424841810935261e-05}
{"train_loss": 0.10006530582904816, "global_step": 163616, "epoch": 1838, "lr": 3.424786789331763e-05}
{"train_loss": 0.032177358865737915, "global_step": 163617, "epoch": 1838, "lr": 3.4247317679400356e-05}
{"train_loss": 0.04757414013147354, "global_step": 163618, "epoch": 1838, "lr": 3.4246767467600813e-05}
{"train_loss": 0.1065436378121376, "global_step": 163619, "epoch": 1838, "lr": 3.424621725791913e-05}
{"train_loss": 0.07551884651184082, "global_step": 163620, "epoch": 1838, "lr": 3.424566705035532e-05}
{"train_loss": 0.051931023597717285, "global_step": 163621, "epoch": 1838, "lr": 3.424511684490952e-05}
{"train_loss": 0.059068311005830765, "global_step": 163622, "epoch": 1838, "lr": 3.4244566641581746e-05}
{"train_loss": 0.09588896483182907, "global_step": 163623, "epoch": 1838, "lr": 3.424401644037212e-05}
{"train_loss": 0.041651614010334015, "global_step": 163624, "epoch": 1838, "lr": 3.424346624128066e-05}
{"train_loss": 0.08541037887334824, "global_step": 163625, "epoch": 1838, "lr": 3.42429160443075e-05}
{"train_loss": 0.1302557736635208, "global_step": 163626, "epoch": 1838, "lr": 3.42423658494527e-05}
{"train_loss": 0.06001623719930649, "global_step": 163627, "epoch": 1838, "lr": 3.4241815656716295e-05}
{"train_loss": 0.15339979529380798, "global_step": 163628, "epoch": 1838, "lr": 3.4241265466098397e-05}
{"train_loss": 0.05682886764407158, "global_step": 163629, "epoch": 1838, "lr": 3.424071527759906e-05}
{"train_loss": 0.066127710044384, "global_step": 163630, "epoch": 1838, "lr": 3.424016509121838e-05}
{"train_loss": 0.0767187550663948, "global_step": 163631, "epoch": 1838, "lr": 3.423961490695641e-05}
{"train_loss": 0.0574333593249321, "global_step": 163632, "epoch": 1838, "lr": 3.423906472481324e-05}
{"train_loss": 0.07742621749639511, "global_step": 163633, "epoch": 1838, "lr": 3.423851454478892e-05}
{"train_loss": 0.11153937876224518, "global_step": 163634, "epoch": 1838, "lr": 3.4237964366883554e-05}
{"train_loss": 0.04946921393275261, "global_step": 163635, "epoch": 1838, "lr": 3.4237414191097196e-05}
{"train_loss": 0.08168812841176987, "global_step": 163636, "epoch": 1838, "lr": 3.423686401742993e-05}
{"train_loss": 0.1313585788011551, "global_step": 163637, "epoch": 1838, "lr": 3.4236313845881814e-05}
{"train_loss": 0.05326814949512482, "global_step": 163638, "epoch": 1838, "lr": 3.423576367645295e-05}
{"train_loss": 0.09080922603607178, "global_step": 163639, "epoch": 1838, "lr": 3.4235213509143376e-05}
{"train_loss": 0.07043048739433289, "global_step": 163640, "epoch": 1838, "lr": 3.423466334395321e-05}
{"train_loss": 0.09122279286384583, "global_step": 163641, "epoch": 1838, "lr": 3.423411318088248e-05}
{"train_loss": 0.0596841424703598, "global_step": 163642, "epoch": 1838, "lr": 3.423356301993129e-05}
{"train_loss": 0.13493414223194122, "global_step": 163643, "epoch": 1838, "lr": 3.423301286109969e-05}
{"train_loss": 0.09027338773012161, "global_step": 163644, "epoch": 1838, "lr": 3.423246270438779e-05}
{"train_loss": 0.09077418595552444, "global_step": 163645, "epoch": 1838, "lr": 3.423191254979564e-05}
{"train_loss": 0.09181931614875793, "global_step": 163646, "epoch": 1838, "lr": 3.423136239732331e-05}
{"train_loss": 0.05971422791481018, "global_step": 163647, "epoch": 1838, "lr": 3.42308122469709e-05}
{"train_loss": 0.10981743782758713, "global_step": 163648, "epoch": 1838, "lr": 3.4230262098738433e-05}
{"train_loss": 0.140679270029068, "global_step": 163649, "epoch": 1838, "lr": 3.4229711952626045e-05}
{"train_loss": 0.12153280526399612, "global_step": 163650, "epoch": 1838, "lr": 3.422916180863375e-05}
{"train_loss": 0.09479457885026932, "global_step": 163651, "epoch": 1838, "lr": 3.4228611666761686e-05}
{"train_loss": 0.1406332403421402, "global_step": 163652, "epoch": 1838, "lr": 3.422806152700986e-05}
{"train_loss": 0.11022936552762985, "global_step": 163653, "epoch": 1838, "lr": 3.422751138937841e-05}
{"train_loss": 0.08369360119104385, "global_step": 163654, "epoch": 1838, "lr": 3.422696125386735e-05}
{"train_loss": 0.09473444521427155, "global_step": 163655, "epoch": 1838, "lr": 3.422641112047681e-05}
{"train_loss": 0.1085004210472107, "global_step": 163656, "epoch": 1838, "lr": 3.422586098920681e-05}
{"train_loss": 0.06234537437558174, "global_step": 163657, "epoch": 1838, "lr": 3.4225310860057475e-05}
{"train_loss": 0.06005927547812462, "global_step": 163658, "epoch": 1838, "lr": 3.422476073302883e-05}
{"train_loss": 0.15435539186000824, "global_step": 163659, "epoch": 1838, "lr": 3.4224210608121e-05}
{"train_loss": 0.0975656732916832, "global_step": 163660, "epoch": 1838, "lr": 3.422366048533401e-05}
{"train_loss": 0.07329215854406357, "global_step": 163661, "epoch": 1838, "lr": 3.422311036466797e-05}
{"train_loss": 0.0847492441534996, "global_step": 163662, "epoch": 1838, "lr": 3.422256024612293e-05}
{"train_loss": 0.06750942021608353, "global_step": 163663, "epoch": 1838, "lr": 3.4222010129698985e-05}
{"train_loss": 0.11073728650808334, "global_step": 163664, "epoch": 1838, "lr": 3.42214600153962e-05}
{"train_loss": 0.0670921579003334, "global_step": 163665, "epoch": 1838, "lr": 3.422090990321464e-05}
{"train_loss": 0.07822169363498688, "global_step": 163666, "epoch": 1838, "lr": 3.422035979315439e-05}
{"train_loss": 0.06590383499860764, "global_step": 163667, "epoch": 1838, "lr": 3.421980968521551e-05}
{"train_loss": 0.10115661472082138, "global_step": 163668, "epoch": 1838, "lr": 3.42192595793981e-05}
{"train_loss": 0.08195061981678009, "global_step": 163669, "epoch": 1838, "lr": 3.4218709475702203e-05}
{"train_loss": 0.08937400173437729, "global_step": 163670, "epoch": 1838, "lr": 3.4218159374127926e-05, "val_loss": 6.915329456329346}
{"train_loss": 0.08846216648817062, "global_step": 163671, "epoch": 1839, "lr": 3.421760927467531e-05}
{"train_loss": 0.06974826753139496, "global_step": 163672, "epoch": 1839, "lr": 3.4217059177344456e-05}
{"train_loss": 0.07923303544521332, "global_step": 163673, "epoch": 1839, "lr": 3.4216509082135414e-05}
{"train_loss": 0.1019437164068222, "global_step": 163674, "epoch": 1839, "lr": 3.421595898904828e-05}
{"train_loss": 0.08065339922904968, "global_step": 163675, "epoch": 1839, "lr": 3.421540889808311e-05}
{"train_loss": 0.07301880419254303, "global_step": 163676, "epoch": 1839, "lr": 3.421485880924e-05}
{"train_loss": 0.08219864219427109, "global_step": 163677, "epoch": 1839, "lr": 3.4214308722518986e-05}
{"train_loss": 0.055415861308574677, "global_step": 163678, "epoch": 1839, "lr": 3.421375863792019e-05}
{"train_loss": 0.03862956911325455, "global_step": 163679, "epoch": 1839, "lr": 3.421320855544364e-05}
{"train_loss": 0.08026739954948425, "global_step": 163680, "epoch": 1839, "lr": 3.421265847508944e-05}
{"train_loss": 0.10587604343891144, "global_step": 163681, "epoch": 1839, "lr": 3.4212108396857656e-05}
{"train_loss": 0.10193736851215363, "global_step": 163682, "epoch": 1839, "lr": 3.4211558320748366e-05}
{"train_loss": 0.11362560838460922, "global_step": 163683, "epoch": 1839, "lr": 3.4211008246761646e-05}
{"train_loss": 0.1521284282207489, "global_step": 163684, "epoch": 1839, "lr": 3.421045817489755e-05}
{"train_loss": 0.13074380159378052, "global_step": 163685, "epoch": 1839, "lr": 3.4209908105156174e-05}
{"train_loss": 0.07439734786748886, "global_step": 163686, "epoch": 1839, "lr": 3.4209358037537576e-05}
{"train_loss": 0.12957949936389923, "global_step": 163687, "epoch": 1839, "lr": 3.420880797204185e-05}
{"train_loss": 0.09929288923740387, "global_step": 163688, "epoch": 1839, "lr": 3.420825790866904e-05}
{"train_loss": 0.13895951211452484, "global_step": 163689, "epoch": 1839, "lr": 3.420770784741925e-05}
{"train_loss": 0.08964719623327255, "global_step": 163690, "epoch": 1839, "lr": 3.420715778829253e-05}
{"train_loss": 0.07739299535751343, "global_step": 163691, "epoch": 1839, "lr": 3.420660773128898e-05}
{"train_loss": 0.08566340059041977, "global_step": 163692, "epoch": 1839, "lr": 3.420605767640864e-05}
{"train_loss": 0.12474971264600754, "global_step": 163693, "epoch": 1839, "lr": 3.4205507623651625e-05}
{"train_loss": 0.1181197315454483, "global_step": 163694, "epoch": 1839, "lr": 3.420495757301796e-05}
{"train_loss": 0.07337819784879684, "global_step": 163695, "epoch": 1839, "lr": 3.420440752450778e-05}
{"train_loss": 0.077338807284832, "global_step": 163696, "epoch": 1839, "lr": 3.420385747812109e-05}
{"train_loss": 0.085783950984478, "global_step": 163697, "epoch": 1839, "lr": 3.4203307433858025e-05}
{"train_loss": 0.0851876363158226, "global_step": 163698, "epoch": 1839, "lr": 3.420275739171862e-05}
{"train_loss": 0.09860683977603912, "global_step": 163699, "epoch": 1839, "lr": 3.420220735170296e-05}
{"train_loss": 0.14598040282726288, "global_step": 163700, "epoch": 1839, "lr": 3.420165731381113e-05}
{"train_loss": 0.04385948181152344, "global_step": 163701, "epoch": 1839, "lr": 3.4201107278043186e-05}
{"train_loss": 0.07175380736589432, "global_step": 163702, "epoch": 1839, "lr": 3.420055724439923e-05}
{"train_loss": 0.08262472599744797, "global_step": 163703, "epoch": 1839, "lr": 3.420000721287929e-05}
{"train_loss": 0.07031823694705963, "global_step": 163704, "epoch": 1839, "lr": 3.419945718348348e-05}
{"train_loss": 0.06328267604112625, "global_step": 163705, "epoch": 1839, "lr": 3.419890715621186e-05}
{"train_loss": 0.0782434269785881, "global_step": 163706, "epoch": 1839, "lr": 3.419835713106451e-05}
{"train_loss": 0.09542801231145859, "global_step": 163707, "epoch": 1839, "lr": 3.419780710804148e-05}
{"train_loss": 0.04120706021785736, "global_step": 163708, "epoch": 1839, "lr": 3.4197257087142886e-05}
{"train_loss": 0.06369318068027496, "global_step": 163709, "epoch": 1839, "lr": 3.419670706836875e-05}
{"train_loss": 0.0610826313495636, "global_step": 163710, "epoch": 1839, "lr": 3.41961570517192e-05}
{"train_loss": 0.12579451501369476, "global_step": 163711, "epoch": 1839, "lr": 3.419560703719427e-05}
{"train_loss": 0.10170061886310577, "global_step": 163712, "epoch": 1839, "lr": 3.419505702479406e-05}
{"train_loss": 0.09535854309797287, "global_step": 163713, "epoch": 1839, "lr": 3.4194507014518604e-05}
{"train_loss": 0.0948963612318039, "global_step": 163714, "epoch": 1839, "lr": 3.4193957006368035e-05}
{"train_loss": 0.0662078857421875, "global_step": 163715, "epoch": 1839, "lr": 3.419340700034238e-05}
{"train_loss": 0.09666869789361954, "global_step": 163716, "epoch": 1839, "lr": 3.4192856996441735e-05}
{"train_loss": 0.11030738800764084, "global_step": 163717, "epoch": 1839, "lr": 3.419230699466616e-05}
{"train_loss": 0.06751764565706253, "global_step": 163718, "epoch": 1839, "lr": 3.419175699501574e-05}
{"train_loss": 0.0707501694560051, "global_step": 163719, "epoch": 1839, "lr": 3.419120699749056e-05}
{"train_loss": 0.04892009496688843, "global_step": 163720, "epoch": 1839, "lr": 3.419065700209066e-05}
{"train_loss": 0.1280144900083542, "global_step": 163721, "epoch": 1839, "lr": 3.419010700881615e-05}
{"train_loss": 0.08152008801698685, "global_step": 163722, "epoch": 1839, "lr": 3.418955701766707e-05}
{"train_loss": 0.081394724547863, "global_step": 163723, "epoch": 1839, "lr": 3.418900702864353e-05}
{"train_loss": 0.07716093957424164, "global_step": 163724, "epoch": 1839, "lr": 3.418845704174557e-05}
{"train_loss": 0.07020171731710434, "global_step": 163725, "epoch": 1839, "lr": 3.418790705697329e-05}
{"train_loss": 0.07694485038518906, "global_step": 163726, "epoch": 1839, "lr": 3.4187357074326734e-05}
{"train_loss": 0.08172951638698578, "global_step": 163727, "epoch": 1839, "lr": 3.418680709380602e-05}
{"train_loss": 0.044887613505125046, "global_step": 163728, "epoch": 1839, "lr": 3.418625711541118e-05}
{"train_loss": 0.08912060409784317, "global_step": 163729, "epoch": 1839, "lr": 3.418570713914232e-05}
{"train_loss": 0.12290900200605392, "global_step": 163730, "epoch": 1839, "lr": 3.418515716499948e-05}
{"train_loss": 0.08425116539001465, "global_step": 163731, "epoch": 1839, "lr": 3.418460719298277e-05}
{"train_loss": 0.02949243038892746, "global_step": 163732, "epoch": 1839, "lr": 3.418405722309223e-05}
{"train_loss": 0.08241879940032959, "global_step": 163733, "epoch": 1839, "lr": 3.4183507255327966e-05}
{"train_loss": 0.10695131123065948, "global_step": 163734, "epoch": 1839, "lr": 3.418295728969002e-05}
{"train_loss": 0.13692344725131989, "global_step": 163735, "epoch": 1839, "lr": 3.4182407326178496e-05}
{"train_loss": 0.09152404218912125, "global_step": 163736, "epoch": 1839, "lr": 3.418185736479345e-05}
{"train_loss": 0.14010214805603027, "global_step": 163737, "epoch": 1839, "lr": 3.418130740553497e-05}
{"train_loss": 0.12015368789434433, "global_step": 163738, "epoch": 1839, "lr": 3.418075744840311e-05}
{"train_loss": 0.1438957154750824, "global_step": 163739, "epoch": 1839, "lr": 3.418020749339795e-05}
{"train_loss": 0.11236602067947388, "global_step": 163740, "epoch": 1839, "lr": 3.4179657540519584e-05}
{"train_loss": 0.06061431020498276, "global_step": 163741, "epoch": 1839, "lr": 3.4179107589768054e-05}
{"train_loss": 0.06602395325899124, "global_step": 163742, "epoch": 1839, "lr": 3.417855764114347e-05}
{"train_loss": 0.08579020947217941, "global_step": 163743, "epoch": 1839, "lr": 3.417800769464586e-05}
{"train_loss": 0.09044817835092545, "global_step": 163744, "epoch": 1839, "lr": 3.4177457750275356e-05}
{"train_loss": 0.05551481619477272, "global_step": 163745, "epoch": 1839, "lr": 3.4176907808031966e-05}
{"train_loss": 0.08033549785614014, "global_step": 163746, "epoch": 1839, "lr": 3.417635786791583e-05}
{"train_loss": 0.051120415329933167, "global_step": 163747, "epoch": 1839, "lr": 3.4175807929926964e-05}
{"train_loss": 0.07748787105083466, "global_step": 163748, "epoch": 1839, "lr": 3.417525799406548e-05}
{"train_loss": 0.08750937879085541, "global_step": 163749, "epoch": 1839, "lr": 3.417470806033143e-05}
{"train_loss": 0.10715048015117645, "global_step": 163750, "epoch": 1839, "lr": 3.417415812872491e-05}
{"train_loss": 0.11835472285747528, "global_step": 163751, "epoch": 1839, "lr": 3.4173608199245964e-05}
{"train_loss": 0.09787580370903015, "global_step": 163752, "epoch": 1839, "lr": 3.417305827189469e-05}
{"train_loss": 0.08426962792873383, "global_step": 163753, "epoch": 1839, "lr": 3.417250834667117e-05}
{"train_loss": 0.06521574407815933, "global_step": 163754, "epoch": 1839, "lr": 3.417195842357544e-05}
{"train_loss": 0.07766150683164597, "global_step": 163755, "epoch": 1839, "lr": 3.4171408502607624e-05}
{"train_loss": 0.08432712405920029, "global_step": 163756, "epoch": 1839, "lr": 3.417085858376776e-05}
{"train_loss": 0.11197205632925034, "global_step": 163757, "epoch": 1839, "lr": 3.417030866705593e-05}
{"train_loss": 0.05699969828128815, "global_step": 163758, "epoch": 1839, "lr": 3.4169758752472206e-05}
{"train_loss": 0.08797233622897876, "global_step": 163759, "epoch": 1839, "lr": 3.416920884001666e-05, "val_loss": 7.016875743865967}
{"train_loss": 0.08968527615070343, "global_step": 163760, "epoch": 1840, "lr": 3.416865892968938e-05}
{"train_loss": 0.1206878274679184, "global_step": 163761, "epoch": 1840, "lr": 3.416810902149044e-05}
{"train_loss": 0.10092992335557938, "global_step": 163762, "epoch": 1840, "lr": 3.416755911541988e-05}
{"train_loss": 0.05474928021430969, "global_step": 163763, "epoch": 1840, "lr": 3.4167009211477826e-05}
{"train_loss": 0.09499169141054153, "global_step": 163764, "epoch": 1840, "lr": 3.41664593096643e-05}
{"train_loss": 0.08027248829603195, "global_step": 163765, "epoch": 1840, "lr": 3.416590940997942e-05}
{"train_loss": 0.12260773032903671, "global_step": 163766, "epoch": 1840, "lr": 3.4165359512423225e-05}
{"train_loss": 0.08958170562982559, "global_step": 163767, "epoch": 1840, "lr": 3.416480961699582e-05}
{"train_loss": 0.15005648136138916, "global_step": 163768, "epoch": 1840, "lr": 3.416425972369724e-05}
{"train_loss": 0.13331130146980286, "global_step": 163769, "epoch": 1840, "lr": 3.41637098325276e-05}
{"train_loss": 0.045998360961675644, "global_step": 163770, "epoch": 1840, "lr": 3.416315994348695e-05}
{"train_loss": 0.06343766301870346, "global_step": 163771, "epoch": 1840, "lr": 3.416261005657536e-05}
{"train_loss": 0.07018201798200607, "global_step": 163772, "epoch": 1840, "lr": 3.416206017179293e-05}
{"train_loss": 0.12854373455047607, "global_step": 163773, "epoch": 1840, "lr": 3.4161510289139706e-05}
{"train_loss": 0.11589447408914566, "global_step": 163774, "epoch": 1840, "lr": 3.416096040861579e-05}
{"train_loss": 0.06862293928861618, "global_step": 163775, "epoch": 1840, "lr": 3.416041053022124e-05}
{"train_loss": 0.1381167322397232, "global_step": 163776, "epoch": 1840, "lr": 3.415986065395612e-05}
{"train_loss": 0.09015023708343506, "global_step": 163777, "epoch": 1840, "lr": 3.4159310779820514e-05}
{"train_loss": 0.11735718697309494, "global_step": 163778, "epoch": 1840, "lr": 3.41587609078145e-05}
{"train_loss": 0.11176188290119171, "global_step": 163779, "epoch": 1840, "lr": 3.415821103793814e-05}
{"train_loss": 0.06671754270792007, "global_step": 163780, "epoch": 1840, "lr": 3.415766117019154e-05}
{"train_loss": 0.048551131039857864, "global_step": 163781, "epoch": 1840, "lr": 3.415711130457471e-05}
{"train_loss": 0.13520042598247528, "global_step": 163782, "epoch": 1840, "lr": 3.41565614410878e-05}
{"train_loss": 0.14467942714691162, "global_step": 163783, "epoch": 1840, "lr": 3.4156011579730816e-05}
{"train_loss": 0.12703944742679596, "global_step": 163784, "epoch": 1840, "lr": 3.4155461720503886e-05}
{"train_loss": 0.05731890723109245, "global_step": 163785, "epoch": 1840, "lr": 3.415491186340704e-05}
{"train_loss": 0.05487513169646263, "global_step": 163786, "epoch": 1840, "lr": 3.41543620084404e-05}
{"train_loss": 0.05396982654929161, "global_step": 163787, "epoch": 1840, "lr": 3.415381215560398e-05}
{"train_loss": 0.14704053103923798, "global_step": 163788, "epoch": 1840, "lr": 3.41532623048979e-05}
{"train_loss": 0.07275180518627167, "global_step": 163789, "epoch": 1840, "lr": 3.415271245632223e-05}
{"train_loss": 0.09915079176425934, "global_step": 163790, "epoch": 1840, "lr": 3.415216260987703e-05}
{"train_loss": 0.10377174615859985, "global_step": 163791, "epoch": 1840, "lr": 3.415161276556238e-05}
{"train_loss": 0.08825791627168655, "global_step": 163792, "epoch": 1840, "lr": 3.415106292337834e-05}
{"train_loss": 0.08041162043809891, "global_step": 163793, "epoch": 1840, "lr": 3.415051308332502e-05}
{"train_loss": 0.1348794549703598, "global_step": 163794, "epoch": 1840, "lr": 3.414996324540246e-05}
{"train_loss": 0.08208012580871582, "global_step": 163795, "epoch": 1840, "lr": 3.4149413409610745e-05}
{"train_loss": 0.09846003353595734, "global_step": 163796, "epoch": 1840, "lr": 3.414886357594994e-05}
{"train_loss": 0.0778040811419487, "global_step": 163797, "epoch": 1840, "lr": 3.414831374442013e-05}
{"train_loss": 0.07210398465394974, "global_step": 163798, "epoch": 1840, "lr": 3.414776391502139e-05}
{"train_loss": 0.0867391973733902, "global_step": 163799, "epoch": 1840, "lr": 3.41472140877538e-05}
{"train_loss": 0.07130765169858932, "global_step": 163800, "epoch": 1840, "lr": 3.41466642626174e-05}
{"train_loss": 0.13786450028419495, "global_step": 163801, "epoch": 1840, "lr": 3.4146114439612315e-05}
{"train_loss": 0.08354798704385757, "global_step": 163802, "epoch": 1840, "lr": 3.414556461873857e-05}
{"train_loss": 0.10853173583745956, "global_step": 163803, "epoch": 1840, "lr": 3.414501479999627e-05}
{"train_loss": 0.1511635184288025, "global_step": 163804, "epoch": 1840, "lr": 3.414446498338547e-05}
{"train_loss": 0.07122565060853958, "global_step": 163805, "epoch": 1840, "lr": 3.414391516890627e-05}
{"train_loss": 0.08131163567304611, "global_step": 163806, "epoch": 1840, "lr": 3.414336535655871e-05}
{"train_loss": 0.10806065797805786, "global_step": 163807, "epoch": 1840, "lr": 3.4142815546342874e-05}
{"train_loss": 0.1750871241092682, "global_step": 163808, "epoch": 1840, "lr": 3.414226573825887e-05}
{"train_loss": 0.09071960300207138, "global_step": 163809, "epoch": 1840, "lr": 3.414171593230672e-05}
{"train_loss": 0.05244482308626175, "global_step": 163810, "epoch": 1840, "lr": 3.414116612848655e-05}
{"train_loss": 0.06902837008237839, "global_step": 163811, "epoch": 1840, "lr": 3.414061632679838e-05}
{"train_loss": 0.07190260291099548, "global_step": 163812, "epoch": 1840, "lr": 3.4140066527242335e-05}
{"train_loss": 0.11662615835666656, "global_step": 163813, "epoch": 1840, "lr": 3.413951672981845e-05}
{"train_loss": 0.1138477623462677, "global_step": 163814, "epoch": 1840, "lr": 3.413896693452682e-05}
{"train_loss": 0.08976873755455017, "global_step": 163815, "epoch": 1840, "lr": 3.4138417141367513e-05}
{"train_loss": 0.08149553090333939, "global_step": 163816, "epoch": 1840, "lr": 3.41378673503406e-05}
{"train_loss": 0.058275312185287476, "global_step": 163817, "epoch": 1840, "lr": 3.4137317561446156e-05}
{"train_loss": 0.0607757493853569, "global_step": 163818, "epoch": 1840, "lr": 3.413676777468426e-05}
{"train_loss": 0.11883486062288284, "global_step": 163819, "epoch": 1840, "lr": 3.4136217990054975e-05}
{"train_loss": 0.050094202160835266, "global_step": 163820, "epoch": 1840, "lr": 3.4135668207558395e-05}
{"train_loss": 0.10770364105701447, "global_step": 163821, "epoch": 1840, "lr": 3.413511842719457e-05}
{"train_loss": 0.07275420427322388, "global_step": 163822, "epoch": 1840, "lr": 3.413456864896359e-05}
{"train_loss": 0.07893436402082443, "global_step": 163823, "epoch": 1840, "lr": 3.413401887286551e-05}
{"train_loss": 0.11008244752883911, "global_step": 163824, "epoch": 1840, "lr": 3.4133469098900424e-05}
{"train_loss": 0.08349836617708206, "global_step": 163825, "epoch": 1840, "lr": 3.4132919327068415e-05}
{"train_loss": 0.10154154151678085, "global_step": 163826, "epoch": 1840, "lr": 3.413236955736952e-05}
{"train_loss": 0.09226212650537491, "global_step": 163827, "epoch": 1840, "lr": 3.413181978980385e-05}
{"train_loss": 0.07130388915538788, "global_step": 163828, "epoch": 1840, "lr": 3.4131270024371456e-05}
{"train_loss": 0.09798870235681534, "global_step": 163829, "epoch": 1840, "lr": 3.413072026107244e-05}
{"train_loss": 0.12992313504219055, "global_step": 163830, "epoch": 1840, "lr": 3.413017049990682e-05}
{"train_loss": 0.11146929115056992, "global_step": 163831, "epoch": 1840, "lr": 3.4129620740874736e-05}
{"train_loss": 0.08201460540294647, "global_step": 163832, "epoch": 1840, "lr": 3.4129070983976216e-05}
{"train_loss": 0.07937715202569962, "global_step": 163833, "epoch": 1840, "lr": 3.412852122921135e-05}
{"train_loss": 0.046702541410923004, "global_step": 163834, "epoch": 1840, "lr": 3.412797147658021e-05}
{"train_loss": 0.11993908137083054, "global_step": 163835, "epoch": 1840, "lr": 3.4127421726082875e-05}
{"train_loss": 0.11492957919836044, "global_step": 163836, "epoch": 1840, "lr": 3.412687197771941e-05}
{"train_loss": 0.07272068411111832, "global_step": 163837, "epoch": 1840, "lr": 3.412632223148991e-05}
{"train_loss": 0.05874716490507126, "global_step": 163838, "epoch": 1840, "lr": 3.41257724873944e-05}
{"train_loss": 0.07552168518304825, "global_step": 163839, "epoch": 1840, "lr": 3.412522274543302e-05}
{"train_loss": 0.0586930550634861, "global_step": 163840, "epoch": 1840, "lr": 3.412467300560579e-05}
{"train_loss": 0.12685178220272064, "global_step": 163841, "epoch": 1840, "lr": 3.412412326791282e-05}
{"train_loss": 0.1384073793888092, "global_step": 163842, "epoch": 1840, "lr": 3.412357353235415e-05}
{"train_loss": 0.15115094184875488, "global_step": 163843, "epoch": 1840, "lr": 3.4123023798929866e-05}
{"train_loss": 0.08011480420827866, "global_step": 163844, "epoch": 1840, "lr": 3.412247406764007e-05}
{"train_loss": 0.07721303403377533, "global_step": 163845, "epoch": 1840, "lr": 3.4121924338484796e-05}
{"train_loss": 0.09632532298564911, "global_step": 163846, "epoch": 1840, "lr": 3.412137461146415e-05}
{"train_loss": 0.11145974695682526, "global_step": 163847, "epoch": 1840, "lr": 3.4120824886578184e-05}
{"train_loss": 0.09494286725360356, "global_step": 163848, "epoch": 1840, "lr": 3.4120275163826984e-05, "val_loss": 6.873176097869873, "train_action_mse_error": 28.55854606628418}
{"train_loss": 0.08931922912597656, "global_step": 163849, "epoch": 1841, "lr": 3.411972544321061e-05}
{"train_loss": 0.08087123930454254, "global_step": 163850, "epoch": 1841, "lr": 3.411917572472917e-05}
{"train_loss": 0.08071084320545197, "global_step": 163851, "epoch": 1841, "lr": 3.411862600838269e-05}
{"train_loss": 0.1417686641216278, "global_step": 163852, "epoch": 1841, "lr": 3.411807629417127e-05}
{"train_loss": 0.13330481946468353, "global_step": 163853, "epoch": 1841, "lr": 3.411752658209499e-05}
{"train_loss": 0.16529449820518494, "global_step": 163854, "epoch": 1841, "lr": 3.411697687215391e-05}
{"train_loss": 0.043675970286130905, "global_step": 163855, "epoch": 1841, "lr": 3.411642716434811e-05}
{"train_loss": 0.0957556664943695, "global_step": 163856, "epoch": 1841, "lr": 3.411587745867766e-05}
{"train_loss": 0.0677897185087204, "global_step": 163857, "epoch": 1841, "lr": 3.411532775514264e-05}
{"train_loss": 0.12815575301647186, "global_step": 163858, "epoch": 1841, "lr": 3.411477805374312e-05}
{"train_loss": 0.069358691573143, "global_step": 163859, "epoch": 1841, "lr": 3.411422835447917e-05}
{"train_loss": 0.09473896771669388, "global_step": 163860, "epoch": 1841, "lr": 3.411367865735088e-05}
{"train_loss": 0.10233312100172043, "global_step": 163861, "epoch": 1841, "lr": 3.41131289623583e-05}
{"train_loss": 0.07661859691143036, "global_step": 163862, "epoch": 1841, "lr": 3.41125792695015e-05}
{"train_loss": 0.07318232208490372, "global_step": 163863, "epoch": 1841, "lr": 3.41120295787806e-05}
{"train_loss": 0.14121487736701965, "global_step": 163864, "epoch": 1841, "lr": 3.411147989019563e-05}
{"train_loss": 0.13544268906116486, "global_step": 163865, "epoch": 1841, "lr": 3.4110930203746695e-05}
{"train_loss": 0.08924040198326111, "global_step": 163866, "epoch": 1841, "lr": 3.4110380519433824e-05}
{"train_loss": 0.1203763410449028, "global_step": 163867, "epoch": 1841, "lr": 3.410983083725714e-05}
{"train_loss": 0.07142340391874313, "global_step": 163868, "epoch": 1841, "lr": 3.4109281157216684e-05}
{"train_loss": 0.0909116268157959, "global_step": 163869, "epoch": 1841, "lr": 3.410873147931256e-05}
{"train_loss": 0.09659000486135483, "global_step": 163870, "epoch": 1841, "lr": 3.410818180354479e-05}
{"train_loss": 0.11722631007432938, "global_step": 163871, "epoch": 1841, "lr": 3.4107632129913515e-05}
{"train_loss": 0.06428799778223038, "global_step": 163872, "epoch": 1841, "lr": 3.410708245841876e-05}
{"train_loss": 0.10878320038318634, "global_step": 163873, "epoch": 1841, "lr": 3.410653278906062e-05}
{"train_loss": 0.07184328883886337, "global_step": 163874, "epoch": 1841, "lr": 3.410598312183916e-05}
{"train_loss": 0.07583831995725632, "global_step": 163875, "epoch": 1841, "lr": 3.410543345675445e-05}
{"train_loss": 0.06174759194254875, "global_step": 163876, "epoch": 1841, "lr": 3.4104883793806575e-05}
{"train_loss": 0.1296035349369049, "global_step": 163877, "epoch": 1841, "lr": 3.410433413299562e-05}
{"train_loss": 0.0761137306690216, "global_step": 163878, "epoch": 1841, "lr": 3.410378447432162e-05}
{"train_loss": 0.11011654883623123, "global_step": 163879, "epoch": 1841, "lr": 3.410323481778467e-05}
{"train_loss": 0.10233807563781738, "global_step": 163880, "epoch": 1841, "lr": 3.410268516338486e-05}
{"train_loss": 0.1388217955827713, "global_step": 163881, "epoch": 1841, "lr": 3.410213551112224e-05}
{"train_loss": 0.09292639791965485, "global_step": 163882, "epoch": 1841, "lr": 3.410158586099691e-05}
{"train_loss": 0.07013550400733948, "global_step": 163883, "epoch": 1841, "lr": 3.4101036213008914e-05}
{"train_loss": 0.08260958641767502, "global_step": 163884, "epoch": 1841, "lr": 3.410048656715835e-05}
{"train_loss": 0.06581178307533264, "global_step": 163885, "epoch": 1841, "lr": 3.4099936923445274e-05}
{"train_loss": 0.0951676070690155, "global_step": 163886, "epoch": 1841, "lr": 3.4099387281869775e-05}
{"train_loss": 0.05456661060452461, "global_step": 163887, "epoch": 1841, "lr": 3.4098837642431916e-05}
{"train_loss": 0.07925480604171753, "global_step": 163888, "epoch": 1841, "lr": 3.409828800513178e-05}
{"train_loss": 0.11686181277036667, "global_step": 163889, "epoch": 1841, "lr": 3.409773836996942e-05}
{"train_loss": 0.0503857284784317, "global_step": 163890, "epoch": 1841, "lr": 3.409718873694494e-05}
{"train_loss": 0.06189752742648125, "global_step": 163891, "epoch": 1841, "lr": 3.40966391060584e-05}
{"train_loss": 0.10216730833053589, "global_step": 163892, "epoch": 1841, "lr": 3.4096089477309866e-05}
{"train_loss": 0.0521911196410656, "global_step": 163893, "epoch": 1841, "lr": 3.409553985069942e-05}
{"train_loss": 0.09095582365989685, "global_step": 163894, "epoch": 1841, "lr": 3.409499022622713e-05}
{"train_loss": 0.06744490563869476, "global_step": 163895, "epoch": 1841, "lr": 3.409444060389309e-05}
{"train_loss": 0.0896264985203743, "global_step": 163896, "epoch": 1841, "lr": 3.409389098369735e-05}
{"train_loss": 0.07071033120155334, "global_step": 163897, "epoch": 1841, "lr": 3.4093341365639986e-05}
{"train_loss": 0.11964519321918488, "global_step": 163898, "epoch": 1841, "lr": 3.409279174972107e-05}
{"train_loss": 0.09242372214794159, "global_step": 163899, "epoch": 1841, "lr": 3.409224213594071e-05}
{"train_loss": 0.05734169855713844, "global_step": 163900, "epoch": 1841, "lr": 3.409169252429893e-05}
{"train_loss": 0.11786903440952301, "global_step": 163901, "epoch": 1841, "lr": 3.4091142914795845e-05}
{"train_loss": 0.09930012375116348, "global_step": 163902, "epoch": 1841, "lr": 3.40905933074315e-05}
{"train_loss": 0.1459072232246399, "global_step": 163903, "epoch": 1841, "lr": 3.4090043702206e-05}
{"train_loss": 0.04400693625211716, "global_step": 163904, "epoch": 1841, "lr": 3.408949409911937e-05}
{"train_loss": 0.10866893827915192, "global_step": 163905, "epoch": 1841, "lr": 3.408894449817175e-05}
{"train_loss": 0.07781535387039185, "global_step": 163906, "epoch": 1841, "lr": 3.408839489936314e-05}
{"train_loss": 0.05760209262371063, "global_step": 163907, "epoch": 1841, "lr": 3.408784530269368e-05}
{"train_loss": 0.0918903723359108, "global_step": 163908, "epoch": 1841, "lr": 3.4087295708163394e-05}
{"train_loss": 0.12473412603139877, "global_step": 163909, "epoch": 1841, "lr": 3.40867461157724e-05}
{"train_loss": 0.08606787025928497, "global_step": 163910, "epoch": 1841, "lr": 3.408619652552074e-05}
{"train_loss": 0.09719114750623703, "global_step": 163911, "epoch": 1841, "lr": 3.4085646937408486e-05}
{"train_loss": 0.08655290305614471, "global_step": 163912, "epoch": 1841, "lr": 3.4085097351435733e-05}
{"train_loss": 0.0847301110625267, "global_step": 163913, "epoch": 1841, "lr": 3.4084547767602545e-05}
{"train_loss": 0.04655250906944275, "global_step": 163914, "epoch": 1841, "lr": 3.4083998185908996e-05}
{"train_loss": 0.0461064875125885, "global_step": 163915, "epoch": 1841, "lr": 3.408344860635514e-05}
{"train_loss": 0.11620305478572845, "global_step": 163916, "epoch": 1841, "lr": 3.40828990289411e-05}
{"train_loss": 0.09965040534734726, "global_step": 163917, "epoch": 1841, "lr": 3.4082349453666894e-05}
{"train_loss": 0.1382603645324707, "global_step": 163918, "epoch": 1841, "lr": 3.408179988053265e-05}
{"train_loss": 0.12596093118190765, "global_step": 163919, "epoch": 1841, "lr": 3.408125030953839e-05}
{"train_loss": 0.10029896348714828, "global_step": 163920, "epoch": 1841, "lr": 3.408070074068423e-05}
{"train_loss": 0.10546278953552246, "global_step": 163921, "epoch": 1841, "lr": 3.4080151173970205e-05}
{"train_loss": 0.09886731952428818, "global_step": 163922, "epoch": 1841, "lr": 3.4079601609396425e-05}
{"train_loss": 0.10141459107398987, "global_step": 163923, "epoch": 1841, "lr": 3.407905204696294e-05}
{"train_loss": 0.10285475850105286, "global_step": 163924, "epoch": 1841, "lr": 3.407850248666985e-05}
{"train_loss": 0.0645933523774147, "global_step": 163925, "epoch": 1841, "lr": 3.407795292851719e-05}
{"train_loss": 0.09680283069610596, "global_step": 163926, "epoch": 1841, "lr": 3.407740337250507e-05}
{"train_loss": 0.09195762127637863, "global_step": 163927, "epoch": 1841, "lr": 3.4076853818633536e-05}
{"train_loss": 0.07503437250852585, "global_step": 163928, "epoch": 1841, "lr": 3.407630426690269e-05}
{"train_loss": 0.04109484702348709, "global_step": 163929, "epoch": 1841, "lr": 3.407575471731258e-05}
{"train_loss": 0.07076936215162277, "global_step": 163930, "epoch": 1841, "lr": 3.407520516986329e-05}
{"train_loss": 0.16060346364974976, "global_step": 163931, "epoch": 1841, "lr": 3.40746556245549e-05}
{"train_loss": 0.12276731431484222, "global_step": 163932, "epoch": 1841, "lr": 3.407410608138748e-05}
{"train_loss": 0.11271986365318298, "global_step": 163933, "epoch": 1841, "lr": 3.4073556540361096e-05}
{"train_loss": 0.16055309772491455, "global_step": 163934, "epoch": 1841, "lr": 3.407300700147582e-05}
{"train_loss": 0.04086358845233917, "global_step": 163935, "epoch": 1841, "lr": 3.4072457464731754e-05}
{"train_loss": 0.08989714086055756, "global_step": 163936, "epoch": 1841, "lr": 3.407190793012893e-05}
{"train_loss": 0.09336630162898074, "global_step": 163937, "epoch": 1841, "lr": 3.407135839766747e-05, "val_loss": 6.7641682624816895}
{"train_loss": 0.07823941856622696, "global_step": 163938, "epoch": 1842, "lr": 3.40708088673474e-05}
{"train_loss": 0.09716945141553879, "global_step": 163939, "epoch": 1842, "lr": 3.407025933916883e-05}
{"train_loss": 0.051993515342473984, "global_step": 163940, "epoch": 1842, "lr": 3.4069709813131803e-05}
{"train_loss": 0.07485414296388626, "global_step": 163941, "epoch": 1842, "lr": 3.4069160289236435e-05}
{"train_loss": 0.08090510964393616, "global_step": 163942, "epoch": 1842, "lr": 3.4068610767482745e-05}
{"train_loss": 0.091128408908844, "global_step": 163943, "epoch": 1842, "lr": 3.406806124787086e-05}
{"train_loss": 0.1320134550333023, "global_step": 163944, "epoch": 1842, "lr": 3.4067511730400815e-05}
{"train_loss": 0.13246507942676544, "global_step": 163945, "epoch": 1842, "lr": 3.406696221507271e-05}
{"train_loss": 0.049041997641325, "global_step": 163946, "epoch": 1842, "lr": 3.40664127018866e-05}
{"train_loss": 0.11329792439937592, "global_step": 163947, "epoch": 1842, "lr": 3.4065863190842575e-05}
{"train_loss": 0.0692131370306015, "global_step": 163948, "epoch": 1842, "lr": 3.406531368194069e-05}
{"train_loss": 0.046193927526474, "global_step": 163949, "epoch": 1842, "lr": 3.406476417518103e-05}
{"train_loss": 0.11591214686632156, "global_step": 163950, "epoch": 1842, "lr": 3.406421467056368e-05}
{"train_loss": 0.12876740097999573, "global_step": 163951, "epoch": 1842, "lr": 3.4063665168088675e-05}
{"train_loss": 0.037118468433618546, "global_step": 163952, "epoch": 1842, "lr": 3.4063115667756144e-05}
{"train_loss": 0.045148737728595734, "global_step": 163953, "epoch": 1842, "lr": 3.406256616956611e-05}
{"train_loss": 0.10472264140844345, "global_step": 163954, "epoch": 1842, "lr": 3.40620166735187e-05}
{"train_loss": 0.08829411119222641, "global_step": 163955, "epoch": 1842, "lr": 3.406146717961393e-05}
{"train_loss": 0.06742943823337555, "global_step": 163956, "epoch": 1842, "lr": 3.406091768785192e-05}
{"train_loss": 0.05938079208135605, "global_step": 163957, "epoch": 1842, "lr": 3.406036819823271e-05}
{"train_loss": 0.09023801982402802, "global_step": 163958, "epoch": 1842, "lr": 3.405981871075641e-05}
{"train_loss": 0.10149664431810379, "global_step": 163959, "epoch": 1842, "lr": 3.405926922542304e-05}
{"train_loss": 0.1163415014743805, "global_step": 163960, "epoch": 1842, "lr": 3.4058719742232744e-05}
{"train_loss": 0.1282416582107544, "global_step": 163961, "epoch": 1842, "lr": 3.4058170261185526e-05}
{"train_loss": 0.05572563037276268, "global_step": 163962, "epoch": 1842, "lr": 3.405762078228152e-05}
{"train_loss": 0.07012273371219635, "global_step": 163963, "epoch": 1842, "lr": 3.405707130552075e-05}
{"train_loss": 0.07040117681026459, "global_step": 163964, "epoch": 1842, "lr": 3.405652183090334e-05}
{"train_loss": 0.10355480760335922, "global_step": 163965, "epoch": 1842, "lr": 3.405597235842931e-05}
{"train_loss": 0.10089359432458878, "global_step": 163966, "epoch": 1842, "lr": 3.405542288809878e-05}
{"train_loss": 0.08394593000411987, "global_step": 163967, "epoch": 1842, "lr": 3.4054873419911784e-05}
{"train_loss": 0.034128692001104355, "global_step": 163968, "epoch": 1842, "lr": 3.405432395386843e-05}
{"train_loss": 0.09880999475717545, "global_step": 163969, "epoch": 1842, "lr": 3.405377448996877e-05}
{"train_loss": 0.059002116322517395, "global_step": 163970, "epoch": 1842, "lr": 3.405322502821288e-05}
{"train_loss": 0.06726425141096115, "global_step": 163971, "epoch": 1842, "lr": 3.405267556860086e-05}
{"train_loss": 0.08906331658363342, "global_step": 163972, "epoch": 1842, "lr": 3.4052126111132745e-05}
{"train_loss": 0.14494964480400085, "global_step": 163973, "epoch": 1842, "lr": 3.4051576655808644e-05}
{"train_loss": 0.09959601610898972, "global_step": 163974, "epoch": 1842, "lr": 3.4051027202628594e-05}
{"train_loss": 0.07091742008924484, "global_step": 163975, "epoch": 1842, "lr": 3.4050477751592716e-05}
{"train_loss": 0.08768541365861893, "global_step": 163976, "epoch": 1842, "lr": 3.4049928302701024e-05}
{"train_loss": 0.056998610496520996, "global_step": 163977, "epoch": 1842, "lr": 3.4049378855953654e-05}
{"train_loss": 0.10543075203895569, "global_step": 163978, "epoch": 1842, "lr": 3.4048829411350625e-05}
{"train_loss": 0.14035646617412567, "global_step": 163979, "epoch": 1842, "lr": 3.404827996889206e-05}
{"train_loss": 0.03131350874900818, "global_step": 163980, "epoch": 1842, "lr": 3.4047730528577986e-05}
{"train_loss": 0.08039851486682892, "global_step": 163981, "epoch": 1842, "lr": 3.4047181090408525e-05}
{"train_loss": 0.058633629232645035, "global_step": 163982, "epoch": 1842, "lr": 3.40466316543837e-05}
{"train_loss": 0.09102824330329895, "global_step": 163983, "epoch": 1842, "lr": 3.404608222050364e-05}
{"train_loss": 0.07897856831550598, "global_step": 163984, "epoch": 1842, "lr": 3.404553278876836e-05}
{"train_loss": 0.08842078596353531, "global_step": 163985, "epoch": 1842, "lr": 3.404498335917799e-05}
{"train_loss": 0.14173561334609985, "global_step": 163986, "epoch": 1842, "lr": 3.4044433931732555e-05}
{"train_loss": 0.05703664571046829, "global_step": 163987, "epoch": 1842, "lr": 3.4043884506432174e-05}
{"train_loss": 0.058246880769729614, "global_step": 163988, "epoch": 1842, "lr": 3.404333508327689e-05}
{"train_loss": 0.059239234775304794, "global_step": 163989, "epoch": 1842, "lr": 3.404278566226676e-05}
{"train_loss": 0.14105044305324554, "global_step": 163990, "epoch": 1842, "lr": 3.404223624340192e-05}
{"train_loss": 0.1147594228386879, "global_step": 163991, "epoch": 1842, "lr": 3.404168682668238e-05}
{"train_loss": 0.07160710543394089, "global_step": 163992, "epoch": 1842, "lr": 3.4041137412108264e-05}
{"train_loss": 0.058462295681238174, "global_step": 163993, "epoch": 1842, "lr": 3.40405879996796e-05}
{"train_loss": 0.04805688560009003, "global_step": 163994, "epoch": 1842, "lr": 3.40400385893965e-05}
{"train_loss": 0.11732026189565659, "global_step": 163995, "epoch": 1842, "lr": 3.403948918125901e-05}
{"train_loss": 0.07070524990558624, "global_step": 163996, "epoch": 1842, "lr": 3.403893977526724e-05}
{"train_loss": 0.07790593057870865, "global_step": 163997, "epoch": 1842, "lr": 3.4038390371421205e-05}
{"train_loss": 0.03985533118247986, "global_step": 163998, "epoch": 1842, "lr": 3.4037840969721046e-05}
{"train_loss": 0.09891811013221741, "global_step": 163999, "epoch": 1842, "lr": 3.403729157016679e-05}
{"train_loss": 0.081051766872406, "global_step": 164000, "epoch": 1842, "lr": 3.403674217275853e-05}
{"train_loss": 0.054137974977493286, "global_step": 164001, "epoch": 1842, "lr": 3.403619277749632e-05}
{"train_loss": 0.11483798921108246, "global_step": 164002, "epoch": 1842, "lr": 3.403564338438027e-05}
{"train_loss": 0.04786974936723709, "global_step": 164003, "epoch": 1842, "lr": 3.4035093993410415e-05}
{"train_loss": 0.137925386428833, "global_step": 164004, "epoch": 1842, "lr": 3.403454460458687e-05}
{"train_loss": 0.0727534294128418, "global_step": 164005, "epoch": 1842, "lr": 3.403399521790966e-05}
{"train_loss": 0.1096712201833725, "global_step": 164006, "epoch": 1842, "lr": 3.40334458333789e-05}
{"train_loss": 0.10511545091867447, "global_step": 164007, "epoch": 1842, "lr": 3.4032896450994656e-05}
{"train_loss": 0.0794939398765564, "global_step": 164008, "epoch": 1842, "lr": 3.403234707075697e-05}
{"train_loss": 0.03771365061402321, "global_step": 164009, "epoch": 1842, "lr": 3.403179769266597e-05}
{"train_loss": 0.10187698155641556, "global_step": 164010, "epoch": 1842, "lr": 3.403124831672167e-05}
{"train_loss": 0.12843982875347137, "global_step": 164011, "epoch": 1842, "lr": 3.40306989429242e-05}
{"train_loss": 0.023227931931614876, "global_step": 164012, "epoch": 1842, "lr": 3.4030149571273584e-05}
{"train_loss": 0.15876524150371552, "global_step": 164013, "epoch": 1842, "lr": 3.402960020176995e-05}
{"train_loss": 0.11832084506750107, "global_step": 164014, "epoch": 1842, "lr": 3.4029050834413314e-05}
{"train_loss": 0.11017913371324539, "global_step": 164015, "epoch": 1842, "lr": 3.402850146920379e-05}
{"train_loss": 0.118733249604702, "global_step": 164016, "epoch": 1842, "lr": 3.402795210614142e-05}
{"train_loss": 0.06393925100564957, "global_step": 164017, "epoch": 1842, "lr": 3.4027402745226334e-05}
{"train_loss": 0.049447983503341675, "global_step": 164018, "epoch": 1842, "lr": 3.4026853386458534e-05}
{"train_loss": 0.04575781896710396, "global_step": 164019, "epoch": 1842, "lr": 3.402630402983816e-05}
{"train_loss": 0.13045060634613037, "global_step": 164020, "epoch": 1842, "lr": 3.402575467536523e-05}
{"train_loss": 0.048306480050086975, "global_step": 164021, "epoch": 1842, "lr": 3.4025205323039854e-05}
{"train_loss": 0.07292575389146805, "global_step": 164022, "epoch": 1842, "lr": 3.402465597286209e-05}
{"train_loss": 0.05363411456346512, "global_step": 164023, "epoch": 1842, "lr": 3.402410662483202e-05}
{"train_loss": 0.11845789849758148, "global_step": 164024, "epoch": 1842, "lr": 3.40235572789497e-05}
{"train_loss": 0.11184792220592499, "global_step": 164025, "epoch": 1842, "lr": 3.402300793521523e-05}
{"train_loss": 0.08539873732023695, "global_step": 164026, "epoch": 1842, "lr": 3.402245859362869e-05, "val_loss": 7.041652202606201}
{"train_loss": 0.05851992964744568, "global_step": 164027, "epoch": 1843, "lr": 3.4021909254190107e-05}
{"train_loss": 0.11514291167259216, "global_step": 164028, "epoch": 1843, "lr": 3.40213599168996e-05}
{"train_loss": 0.10949918627738953, "global_step": 164029, "epoch": 1843, "lr": 3.402081058175721e-05}
{"train_loss": 0.14594440162181854, "global_step": 164030, "epoch": 1843, "lr": 3.402026124876305e-05}
{"train_loss": 0.06478167325258255, "global_step": 164031, "epoch": 1843, "lr": 3.401971191791715e-05}
{"train_loss": 0.1468418687582016, "global_step": 164032, "epoch": 1843, "lr": 3.401916258921962e-05}
{"train_loss": 0.0717218890786171, "global_step": 164033, "epoch": 1843, "lr": 3.401861326267051e-05}
{"train_loss": 0.05553406849503517, "global_step": 164034, "epoch": 1843, "lr": 3.4018063938269906e-05}
{"train_loss": 0.04431017115712166, "global_step": 164035, "epoch": 1843, "lr": 3.401751461601787e-05}
{"train_loss": 0.16500256955623627, "global_step": 164036, "epoch": 1843, "lr": 3.4016965295914505e-05}
{"train_loss": 0.06571248173713684, "global_step": 164037, "epoch": 1843, "lr": 3.4016415977959836e-05}
{"train_loss": 0.08013977110385895, "global_step": 164038, "epoch": 1843, "lr": 3.4015866662153994e-05}
{"train_loss": 0.06845271587371826, "global_step": 164039, "epoch": 1843, "lr": 3.4015317348497e-05}
{"train_loss": 0.07468673586845398, "global_step": 164040, "epoch": 1843, "lr": 3.401476803698898e-05}
{"train_loss": 0.10313262045383453, "global_step": 164041, "epoch": 1843, "lr": 3.4014218727629945e-05}
{"train_loss": 0.18902736902236938, "global_step": 164042, "epoch": 1843, "lr": 3.401366942042002e-05}
{"train_loss": 0.15652796626091003, "global_step": 164043, "epoch": 1843, "lr": 3.401312011535926e-05}
{"train_loss": 0.0851147249341011, "global_step": 164044, "epoch": 1843, "lr": 3.401257081244774e-05}
{"train_loss": 0.12322767823934555, "global_step": 164045, "epoch": 1843, "lr": 3.401202151168555e-05}
{"train_loss": 0.12347211688756943, "global_step": 164046, "epoch": 1843, "lr": 3.401147221307273e-05}
{"train_loss": 0.10361693799495697, "global_step": 164047, "epoch": 1843, "lr": 3.401092291660939e-05}
{"train_loss": 0.09487824141979218, "global_step": 164048, "epoch": 1843, "lr": 3.4010373622295565e-05}
{"train_loss": 0.13032707571983337, "global_step": 164049, "epoch": 1843, "lr": 3.4009824330131376e-05}
{"train_loss": 0.026835015043616295, "global_step": 164050, "epoch": 1843, "lr": 3.400927504011685e-05}
{"train_loss": 0.16646729409694672, "global_step": 164051, "epoch": 1843, "lr": 3.4008725752252095e-05}
{"train_loss": 0.07948726415634155, "global_step": 164052, "epoch": 1843, "lr": 3.4008176466537153e-05}
{"train_loss": 0.1146174967288971, "global_step": 164053, "epoch": 1843, "lr": 3.4007627182972144e-05}
{"train_loss": 0.08622143417596817, "global_step": 164054, "epoch": 1843, "lr": 3.4007077901557094e-05}
{"train_loss": 0.12122738361358643, "global_step": 164055, "epoch": 1843, "lr": 3.400652862229211e-05}
{"train_loss": 0.05220160260796547, "global_step": 164056, "epoch": 1843, "lr": 3.4005979345177244e-05}
{"train_loss": 0.08481819182634354, "global_step": 164057, "epoch": 1843, "lr": 3.400543007021259e-05}
{"train_loss": 0.0863669291138649, "global_step": 164058, "epoch": 1843, "lr": 3.4004880797398194e-05}
{"train_loss": 0.05771113932132721, "global_step": 164059, "epoch": 1843, "lr": 3.400433152673417e-05}
{"train_loss": 0.16887244582176208, "global_step": 164060, "epoch": 1843, "lr": 3.400378225822054e-05}
{"train_loss": 0.026567276567220688, "global_step": 164061, "epoch": 1843, "lr": 3.400323299185742e-05}
{"train_loss": 0.05467625707387924, "global_step": 164062, "epoch": 1843, "lr": 3.4002683727644865e-05}
{"train_loss": 0.04278388246893883, "global_step": 164063, "epoch": 1843, "lr": 3.4002134465582966e-05}
{"train_loss": 0.09042632579803467, "global_step": 164064, "epoch": 1843, "lr": 3.400158520567179e-05}
{"train_loss": 0.09072651714086533, "global_step": 164065, "epoch": 1843, "lr": 3.4001035947911384e-05}
{"train_loss": 0.07032643258571625, "global_step": 164066, "epoch": 1843, "lr": 3.400048669230187e-05}
{"train_loss": 0.10757099092006683, "global_step": 164067, "epoch": 1843, "lr": 3.399993743884327e-05}
{"train_loss": 0.07721075415611267, "global_step": 164068, "epoch": 1843, "lr": 3.399938818753571e-05}
{"train_loss": 0.12278162688016891, "global_step": 164069, "epoch": 1843, "lr": 3.3998838938379204e-05}
{"train_loss": 0.10236925631761551, "global_step": 164070, "epoch": 1843, "lr": 3.399828969137389e-05}
{"train_loss": 0.09536214172840118, "global_step": 164071, "epoch": 1843, "lr": 3.3997740446519785e-05}
{"train_loss": 0.11653178930282593, "global_step": 164072, "epoch": 1843, "lr": 3.3997191203817014e-05}
{"train_loss": 0.10756544023752213, "global_step": 164073, "epoch": 1843, "lr": 3.3996641963265605e-05}
{"train_loss": 0.1639818549156189, "global_step": 164074, "epoch": 1843, "lr": 3.3996092724865666e-05}
{"train_loss": 0.14762191474437714, "global_step": 164075, "epoch": 1843, "lr": 3.3995543488617245e-05}
{"train_loss": 0.06327108293771744, "global_step": 164076, "epoch": 1843, "lr": 3.3994994254520444e-05}
{"train_loss": 0.1074640303850174, "global_step": 164077, "epoch": 1843, "lr": 3.39944450225753e-05}
{"train_loss": 0.07685332745313644, "global_step": 164078, "epoch": 1843, "lr": 3.399389579278193e-05}
{"train_loss": 0.09368836879730225, "global_step": 164079, "epoch": 1843, "lr": 3.399334656514037e-05}
{"train_loss": 0.07364597916603088, "global_step": 164080, "epoch": 1843, "lr": 3.3992797339650714e-05}
{"train_loss": 0.09932220727205276, "global_step": 164081, "epoch": 1843, "lr": 3.399224811631303e-05}
{"train_loss": 0.10466374456882477, "global_step": 164082, "epoch": 1843, "lr": 3.39916988951274e-05}
{"train_loss": 0.07614389061927795, "global_step": 164083, "epoch": 1843, "lr": 3.399114967609388e-05}
{"train_loss": 0.08609777688980103, "global_step": 164084, "epoch": 1843, "lr": 3.399060045921256e-05}
{"train_loss": 0.0865154042840004, "global_step": 164085, "epoch": 1843, "lr": 3.399005124448352e-05}
{"train_loss": 0.11312302947044373, "global_step": 164086, "epoch": 1843, "lr": 3.3989502031906805e-05}
{"train_loss": 0.12700143456459045, "global_step": 164087, "epoch": 1843, "lr": 3.398895282148252e-05}
{"train_loss": 0.17151035368442535, "global_step": 164088, "epoch": 1843, "lr": 3.398840361321071e-05}
{"train_loss": 0.08762125670909882, "global_step": 164089, "epoch": 1843, "lr": 3.398785440709148e-05}
{"train_loss": 0.09439731389284134, "global_step": 164090, "epoch": 1843, "lr": 3.3987305203124877e-05}
{"train_loss": 0.09404446184635162, "global_step": 164091, "epoch": 1843, "lr": 3.3986756001311e-05}
{"train_loss": 0.1127590611577034, "global_step": 164092, "epoch": 1843, "lr": 3.398620680164989e-05}
{"train_loss": 0.14835672080516815, "global_step": 164093, "epoch": 1843, "lr": 3.398565760414165e-05}
{"train_loss": 0.10225147008895874, "global_step": 164094, "epoch": 1843, "lr": 3.398510840878633e-05}
{"train_loss": 0.07914119958877563, "global_step": 164095, "epoch": 1843, "lr": 3.398455921558404e-05}
{"train_loss": 0.08712244033813477, "global_step": 164096, "epoch": 1843, "lr": 3.398401002453481e-05}
{"train_loss": 0.06074431911110878, "global_step": 164097, "epoch": 1843, "lr": 3.398346083563873e-05}
{"train_loss": 0.08409406244754791, "global_step": 164098, "epoch": 1843, "lr": 3.39829116488959e-05}
{"train_loss": 0.10352623462677002, "global_step": 164099, "epoch": 1843, "lr": 3.398236246430636e-05}
{"train_loss": 0.07676585018634796, "global_step": 164100, "epoch": 1843, "lr": 3.39818132818702e-05}
{"train_loss": 0.11605003476142883, "global_step": 164101, "epoch": 1843, "lr": 3.3981264101587495e-05}
{"train_loss": 0.1869998276233673, "global_step": 164102, "epoch": 1843, "lr": 3.398071492345831e-05}
{"train_loss": 0.17939786612987518, "global_step": 164103, "epoch": 1843, "lr": 3.398016574748272e-05}
{"train_loss": 0.11278426647186279, "global_step": 164104, "epoch": 1843, "lr": 3.397961657366081e-05}
{"train_loss": 0.10196536034345627, "global_step": 164105, "epoch": 1843, "lr": 3.3979067401992625e-05}
{"train_loss": 0.0862852931022644, "global_step": 164106, "epoch": 1843, "lr": 3.3978518232478285e-05}
{"train_loss": 0.09778042137622833, "global_step": 164107, "epoch": 1843, "lr": 3.397796906511781e-05}
{"train_loss": 0.15526096522808075, "global_step": 164108, "epoch": 1843, "lr": 3.397741989991133e-05}
{"train_loss": 0.08069824427366257, "global_step": 164109, "epoch": 1843, "lr": 3.3976870736858865e-05}
{"train_loss": 0.1248171404004097, "global_step": 164110, "epoch": 1843, "lr": 3.3976321575960537e-05}
{"train_loss": 0.09319333732128143, "global_step": 164111, "epoch": 1843, "lr": 3.3975772417216376e-05}
{"train_loss": 0.15212824940681458, "global_step": 164112, "epoch": 1843, "lr": 3.39752232606265e-05}
{"train_loss": 0.11434148252010345, "global_step": 164113, "epoch": 1843, "lr": 3.397467410619094e-05}
{"train_loss": 0.08176210522651672, "global_step": 164114, "epoch": 1843, "lr": 3.3974124953909806e-05}
{"train_loss": 0.10125329427086235, "global_step": 164115, "epoch": 1843, "lr": 3.3973575803783134e-05, "val_loss": 6.853416442871094}
{"train_loss": 0.06696700304746628, "global_step": 164116, "epoch": 1844, "lr": 3.3973026655811024e-05}
{"train_loss": 0.11462850868701935, "global_step": 164117, "epoch": 1844, "lr": 3.397247750999356e-05}
{"train_loss": 0.16802695393562317, "global_step": 164118, "epoch": 1844, "lr": 3.3971928366330795e-05}
{"train_loss": 0.08646763116121292, "global_step": 164119, "epoch": 1844, "lr": 3.3971379224822806e-05}
{"train_loss": 0.10423002392053604, "global_step": 164120, "epoch": 1844, "lr": 3.397083008546968e-05}
{"train_loss": 0.07532550394535065, "global_step": 164121, "epoch": 1844, "lr": 3.397028094827147e-05}
{"train_loss": 0.0878177136182785, "global_step": 164122, "epoch": 1844, "lr": 3.396973181322827e-05}
{"train_loss": 0.13163183629512787, "global_step": 164123, "epoch": 1844, "lr": 3.396918268034014e-05}
{"train_loss": 0.15411832928657532, "global_step": 164124, "epoch": 1844, "lr": 3.396863354960714e-05}
{"train_loss": 0.07459277659654617, "global_step": 164125, "epoch": 1844, "lr": 3.3968084421029396e-05}
{"train_loss": 0.10638835281133652, "global_step": 164126, "epoch": 1844, "lr": 3.3967535294606914e-05}
{"train_loss": 0.08823106437921524, "global_step": 164127, "epoch": 1844, "lr": 3.3966986170339834e-05}
{"train_loss": 0.1579159051179886, "global_step": 164128, "epoch": 1844, "lr": 3.396643704822816e-05}
{"train_loss": 0.06965750455856323, "global_step": 164129, "epoch": 1844, "lr": 3.396588792827203e-05}
{"train_loss": 0.04636676609516144, "global_step": 164130, "epoch": 1844, "lr": 3.3965338810471484e-05}
{"train_loss": 0.10073795169591904, "global_step": 164131, "epoch": 1844, "lr": 3.396478969482661e-05}
{"train_loss": 0.05752138793468475, "global_step": 164132, "epoch": 1844, "lr": 3.396424058133746e-05}
{"train_loss": 0.11167294532060623, "global_step": 164133, "epoch": 1844, "lr": 3.396369147000411e-05}
{"train_loss": 0.06635314226150513, "global_step": 164134, "epoch": 1844, "lr": 3.3963142360826686e-05}
{"train_loss": 0.13486167788505554, "global_step": 164135, "epoch": 1844, "lr": 3.396259325380519e-05}
{"train_loss": 0.10667473822832108, "global_step": 164136, "epoch": 1844, "lr": 3.3962044148939744e-05}
{"train_loss": 0.0928175151348114, "global_step": 164137, "epoch": 1844, "lr": 3.3961495046230404e-05}
{"train_loss": 0.12118399143218994, "global_step": 164138, "epoch": 1844, "lr": 3.3960945945677244e-05}
{"train_loss": 0.09705861657857895, "global_step": 164139, "epoch": 1844, "lr": 3.396039684728033e-05}
{"train_loss": 0.153061643242836, "global_step": 164140, "epoch": 1844, "lr": 3.3959847751039755e-05}
{"train_loss": 0.10984787344932556, "global_step": 164141, "epoch": 1844, "lr": 3.395929865695558e-05}
{"train_loss": 0.06774452328681946, "global_step": 164142, "epoch": 1844, "lr": 3.3958749565027894e-05}
{"train_loss": 0.06126376986503601, "global_step": 164143, "epoch": 1844, "lr": 3.395820047525674e-05}
{"train_loss": 0.06370281428098679, "global_step": 164144, "epoch": 1844, "lr": 3.395765138764223e-05}
{"train_loss": 0.13044556975364685, "global_step": 164145, "epoch": 1844, "lr": 3.39571023021844e-05}
{"train_loss": 0.1197967380285263, "global_step": 164146, "epoch": 1844, "lr": 3.395655321888336e-05}
{"train_loss": 0.061653006821870804, "global_step": 164147, "epoch": 1844, "lr": 3.395600413773914e-05}
{"train_loss": 0.10284971445798874, "global_step": 164148, "epoch": 1844, "lr": 3.395545505875186e-05}
{"train_loss": 0.09409092366695404, "global_step": 164149, "epoch": 1844, "lr": 3.395490598192156e-05}
{"train_loss": 0.0955008864402771, "global_step": 164150, "epoch": 1844, "lr": 3.3954356907248344e-05}
{"train_loss": 0.07810785621404648, "global_step": 164151, "epoch": 1844, "lr": 3.395380783473225e-05}
{"train_loss": 0.09420213848352432, "global_step": 164152, "epoch": 1844, "lr": 3.395325876437337e-05}
{"train_loss": 0.04909190535545349, "global_step": 164153, "epoch": 1844, "lr": 3.39527096961718e-05}
{"train_loss": 0.09149612486362457, "global_step": 164154, "epoch": 1844, "lr": 3.395216063012757e-05}
{"train_loss": 0.07131720334291458, "global_step": 164155, "epoch": 1844, "lr": 3.3951611566240806e-05}
{"train_loss": 0.08102075755596161, "global_step": 164156, "epoch": 1844, "lr": 3.395106250451153e-05}
{"train_loss": 0.13217779994010925, "global_step": 164157, "epoch": 1844, "lr": 3.395051344493985e-05}
{"train_loss": 0.1511317938566208, "global_step": 164158, "epoch": 1844, "lr": 3.394996438752581e-05}
{"train_loss": 0.04537281021475792, "global_step": 164159, "epoch": 1844, "lr": 3.3949415332269516e-05}
{"train_loss": 0.09788557142019272, "global_step": 164160, "epoch": 1844, "lr": 3.3948866279171024e-05}
{"train_loss": 0.09548838436603546, "global_step": 164161, "epoch": 1844, "lr": 3.394831722823042e-05}
{"train_loss": 0.09920084476470947, "global_step": 164162, "epoch": 1844, "lr": 3.3947768179447744e-05}
{"train_loss": 0.1075030043721199, "global_step": 164163, "epoch": 1844, "lr": 3.3947219132823126e-05}
{"train_loss": 0.10537479817867279, "global_step": 164164, "epoch": 1844, "lr": 3.3946670088356584e-05}
{"train_loss": 0.15109610557556152, "global_step": 164165, "epoch": 1844, "lr": 3.394612104604824e-05}
{"train_loss": 0.0716056153178215, "global_step": 164166, "epoch": 1844, "lr": 3.394557200589812e-05}
{"train_loss": 0.11620484292507172, "global_step": 164167, "epoch": 1844, "lr": 3.3945022967906346e-05}
{"train_loss": 0.08536756783723831, "global_step": 164168, "epoch": 1844, "lr": 3.394447393207294e-05}
{"train_loss": 0.13093183934688568, "global_step": 164169, "epoch": 1844, "lr": 3.3943924898398005e-05}
{"train_loss": 0.11944460868835449, "global_step": 164170, "epoch": 1844, "lr": 3.394337586688164e-05}
{"train_loss": 0.08180160820484161, "global_step": 164171, "epoch": 1844, "lr": 3.394282683752387e-05}
{"train_loss": 0.1011984795331955, "global_step": 164172, "epoch": 1844, "lr": 3.3942277810324815e-05}
{"train_loss": 0.09934985637664795, "global_step": 164173, "epoch": 1844, "lr": 3.39417287852845e-05}
{"train_loss": 0.13246168196201324, "global_step": 164174, "epoch": 1844, "lr": 3.394117976240304e-05}
{"train_loss": 0.15836158394813538, "global_step": 164175, "epoch": 1844, "lr": 3.394063074168049e-05}
{"train_loss": 0.1596052348613739, "global_step": 164176, "epoch": 1844, "lr": 3.3940081723116925e-05}
{"train_loss": 0.15047016739845276, "global_step": 164177, "epoch": 1844, "lr": 3.3939532706712416e-05}
{"train_loss": 0.06272611021995544, "global_step": 164178, "epoch": 1844, "lr": 3.393898369246705e-05}
{"train_loss": 0.09418267011642456, "global_step": 164179, "epoch": 1844, "lr": 3.393843468038088e-05}
{"train_loss": 0.04319615289568901, "global_step": 164180, "epoch": 1844, "lr": 3.393788567045401e-05}
{"train_loss": 0.08478456735610962, "global_step": 164181, "epoch": 1844, "lr": 3.3937336662686475e-05}
{"train_loss": 0.09022373706102371, "global_step": 164182, "epoch": 1844, "lr": 3.393678765707838e-05}
{"train_loss": 0.08674280345439911, "global_step": 164183, "epoch": 1844, "lr": 3.393623865362978e-05}
{"train_loss": 0.07950114458799362, "global_step": 164184, "epoch": 1844, "lr": 3.3935689652340775e-05}
{"train_loss": 0.08129195123910904, "global_step": 164185, "epoch": 1844, "lr": 3.39351406532114e-05}
{"train_loss": 0.16941580176353455, "global_step": 164186, "epoch": 1844, "lr": 3.3934591656241765e-05}
{"train_loss": 0.09677553176879883, "global_step": 164187, "epoch": 1844, "lr": 3.3934042661431913e-05}
{"train_loss": 0.08978645503520966, "global_step": 164188, "epoch": 1844, "lr": 3.3933493668781935e-05}
{"train_loss": 0.06749671697616577, "global_step": 164189, "epoch": 1844, "lr": 3.393294467829192e-05}
{"train_loss": 0.07327494770288467, "global_step": 164190, "epoch": 1844, "lr": 3.39323956899619e-05}
{"train_loss": 0.14093010127544403, "global_step": 164191, "epoch": 1844, "lr": 3.3931846703792e-05}
{"train_loss": 0.09052974730730057, "global_step": 164192, "epoch": 1844, "lr": 3.393129771978224e-05}
{"train_loss": 0.06593217700719833, "global_step": 164193, "epoch": 1844, "lr": 3.393074873793275e-05}
{"train_loss": 0.11582581698894501, "global_step": 164194, "epoch": 1844, "lr": 3.3930199758243565e-05}
{"train_loss": 0.06615245342254639, "global_step": 164195, "epoch": 1844, "lr": 3.392965078071476e-05}
{"train_loss": 0.05844196677207947, "global_step": 164196, "epoch": 1844, "lr": 3.392910180534643e-05}
{"train_loss": 0.061491724103689194, "global_step": 164197, "epoch": 1844, "lr": 3.392855283213862e-05}
{"train_loss": 0.11685024201869965, "global_step": 164198, "epoch": 1844, "lr": 3.392800386109143e-05}
{"train_loss": 0.08222295343875885, "global_step": 164199, "epoch": 1844, "lr": 3.392745489220493e-05}
{"train_loss": 0.09602302312850952, "global_step": 164200, "epoch": 1844, "lr": 3.392690592547918e-05}
{"train_loss": 0.11517520993947983, "global_step": 164201, "epoch": 1844, "lr": 3.392635696091427e-05}
{"train_loss": 0.07251088321208954, "global_step": 164202, "epoch": 1844, "lr": 3.392580799851025e-05}
{"train_loss": 0.1403263509273529, "global_step": 164203, "epoch": 1844, "lr": 3.392525903826722e-05}
{"train_loss": 0.09800826967432258, "global_step": 164204, "epoch": 1844, "lr": 3.392471008018523e-05, "val_loss": 6.786181926727295}
{"train_loss": 0.05789772421121597, "global_step": 164205, "epoch": 1845, "lr": 3.392416112426437e-05}
{"train_loss": 0.06008952856063843, "global_step": 164206, "epoch": 1845, "lr": 3.3923612170504725e-05}
{"train_loss": 0.10796000063419342, "global_step": 164207, "epoch": 1845, "lr": 3.3923063218906334e-05}
{"train_loss": 0.13050894439220428, "global_step": 164208, "epoch": 1845, "lr": 3.392251426946932e-05}
{"train_loss": 0.12264314293861389, "global_step": 164209, "epoch": 1845, "lr": 3.3921965322193704e-05}
{"train_loss": 0.03996005281805992, "global_step": 164210, "epoch": 1845, "lr": 3.39214163770796e-05}
{"train_loss": 0.07674027234315872, "global_step": 164211, "epoch": 1845, "lr": 3.3920867434127044e-05}
{"train_loss": 0.11270762234926224, "global_step": 164212, "epoch": 1845, "lr": 3.3920318493336156e-05}
{"train_loss": 0.12055686861276627, "global_step": 164213, "epoch": 1845, "lr": 3.391976955470697e-05}
{"train_loss": 0.10915785282850266, "global_step": 164214, "epoch": 1845, "lr": 3.391922061823959e-05}
{"train_loss": 0.11102516204118729, "global_step": 164215, "epoch": 1845, "lr": 3.391867168393406e-05}
{"train_loss": 0.10840218514204025, "global_step": 164216, "epoch": 1845, "lr": 3.391812275179047e-05}
{"train_loss": 0.12606501579284668, "global_step": 164217, "epoch": 1845, "lr": 3.3917573821808904e-05}
{"train_loss": 0.09762021154165268, "global_step": 164218, "epoch": 1845, "lr": 3.3917024893989404e-05}
{"train_loss": 0.06731254607439041, "global_step": 164219, "epoch": 1845, "lr": 3.391647596833208e-05}
{"train_loss": 0.05196412280201912, "global_step": 164220, "epoch": 1845, "lr": 3.391592704483699e-05}
{"train_loss": 0.06985023617744446, "global_step": 164221, "epoch": 1845, "lr": 3.391537812350419e-05}
{"train_loss": 0.058443453162908554, "global_step": 164222, "epoch": 1845, "lr": 3.39148292043338e-05}
{"train_loss": 0.0749838724732399, "global_step": 164223, "epoch": 1845, "lr": 3.391428028732584e-05}
{"train_loss": 0.09987430274486542, "global_step": 164224, "epoch": 1845, "lr": 3.3913731372480406e-05}
{"train_loss": 0.08703338354825974, "global_step": 164225, "epoch": 1845, "lr": 3.39131824597976e-05}
{"train_loss": 0.10540540516376495, "global_step": 164226, "epoch": 1845, "lr": 3.391263354927744e-05}
{"train_loss": 0.10622310638427734, "global_step": 164227, "epoch": 1845, "lr": 3.391208464092006e-05}
{"train_loss": 0.15382938086986542, "global_step": 164228, "epoch": 1845, "lr": 3.3911535734725484e-05}
{"train_loss": 0.08460161834955215, "global_step": 164229, "epoch": 1845, "lr": 3.391098683069383e-05}
{"train_loss": 0.0832938477396965, "global_step": 164230, "epoch": 1845, "lr": 3.391043792882511e-05}
{"train_loss": 0.09904889017343521, "global_step": 164231, "epoch": 1845, "lr": 3.390988902911947e-05}
{"train_loss": 0.0498674176633358, "global_step": 164232, "epoch": 1845, "lr": 3.3909340131576926e-05}
{"train_loss": 0.13466979563236237, "global_step": 164233, "epoch": 1845, "lr": 3.39087912361976e-05}
{"train_loss": 0.04894406348466873, "global_step": 164234, "epoch": 1845, "lr": 3.390824234298152e-05}
{"train_loss": 0.10444532334804535, "global_step": 164235, "epoch": 1845, "lr": 3.390769345192879e-05}
{"train_loss": 0.05115192010998726, "global_step": 164236, "epoch": 1845, "lr": 3.390714456303947e-05}
{"train_loss": 0.0683433786034584, "global_step": 164237, "epoch": 1845, "lr": 3.390659567631364e-05}
{"train_loss": 0.07838205248117447, "global_step": 164238, "epoch": 1845, "lr": 3.390604679175137e-05}
{"train_loss": 0.16347460448741913, "global_step": 164239, "epoch": 1845, "lr": 3.3905497909352754e-05}
{"train_loss": 0.11945033073425293, "global_step": 164240, "epoch": 1845, "lr": 3.390494902911782e-05}
{"train_loss": 0.06993315368890762, "global_step": 164241, "epoch": 1845, "lr": 3.390440015104669e-05}
{"train_loss": 0.0907355472445488, "global_step": 164242, "epoch": 1845, "lr": 3.39038512751394e-05}
{"train_loss": 0.04779937490820885, "global_step": 164243, "epoch": 1845, "lr": 3.390330240139603e-05}
{"train_loss": 0.06623654812574387, "global_step": 164244, "epoch": 1845, "lr": 3.39027535298167e-05}
{"train_loss": 0.0709780752658844, "global_step": 164245, "epoch": 1845, "lr": 3.390220466040142e-05}
{"train_loss": 0.09030888974666595, "global_step": 164246, "epoch": 1845, "lr": 3.390165579315032e-05}
{"train_loss": 0.06440335512161255, "global_step": 164247, "epoch": 1845, "lr": 3.390110692806342e-05}
{"train_loss": 0.08853907138109207, "global_step": 164248, "epoch": 1845, "lr": 3.3900558065140835e-05}
{"train_loss": 0.08607515692710876, "global_step": 164249, "epoch": 1845, "lr": 3.390000920438261e-05}
{"train_loss": 0.041002944111824036, "global_step": 164250, "epoch": 1845, "lr": 3.3899460345788855e-05}
{"train_loss": 0.07003234326839447, "global_step": 164251, "epoch": 1845, "lr": 3.3898911489359596e-05}
{"train_loss": 0.07821615040302277, "global_step": 164252, "epoch": 1845, "lr": 3.389836263509495e-05}
{"train_loss": 0.08517144620418549, "global_step": 164253, "epoch": 1845, "lr": 3.389781378299497e-05}
{"train_loss": 0.06166769564151764, "global_step": 164254, "epoch": 1845, "lr": 3.3897264933059734e-05}
{"train_loss": 0.1031823381781578, "global_step": 164255, "epoch": 1845, "lr": 3.389671608528931e-05}
{"train_loss": 0.12597866356372833, "global_step": 164256, "epoch": 1845, "lr": 3.3896167239683774e-05}
{"train_loss": 0.07993078231811523, "global_step": 164257, "epoch": 1845, "lr": 3.389561839624321e-05}
{"train_loss": 0.0763234943151474, "global_step": 164258, "epoch": 1845, "lr": 3.389506955496768e-05}
{"train_loss": 0.0781882256269455, "global_step": 164259, "epoch": 1845, "lr": 3.389452071585726e-05}
{"train_loss": 0.09030909836292267, "global_step": 164260, "epoch": 1845, "lr": 3.389397187891201e-05}
{"train_loss": 0.10586856305599213, "global_step": 164261, "epoch": 1845, "lr": 3.389342304413204e-05}
{"train_loss": 0.04838170111179352, "global_step": 164262, "epoch": 1845, "lr": 3.389287421151739e-05}
{"train_loss": 0.0963704064488411, "global_step": 164263, "epoch": 1845, "lr": 3.3892325381068165e-05}
{"train_loss": 0.05871447175741196, "global_step": 164264, "epoch": 1845, "lr": 3.38917765527844e-05}
{"train_loss": 0.08295133709907532, "global_step": 164265, "epoch": 1845, "lr": 3.389122772666621e-05}
{"train_loss": 0.09054629504680634, "global_step": 164266, "epoch": 1845, "lr": 3.3890678902713624e-05}
{"train_loss": 0.058000922203063965, "global_step": 164267, "epoch": 1845, "lr": 3.3890130080926766e-05}
{"train_loss": 0.052475959062576294, "global_step": 164268, "epoch": 1845, "lr": 3.3889581261305656e-05}
{"train_loss": 0.042276185005903244, "global_step": 164269, "epoch": 1845, "lr": 3.388903244385043e-05}
{"train_loss": 0.12271412461996078, "global_step": 164270, "epoch": 1845, "lr": 3.38884836285611e-05}
{"train_loss": 0.058249834924936295, "global_step": 164271, "epoch": 1845, "lr": 3.388793481543778e-05}
{"train_loss": 0.09191007167100906, "global_step": 164272, "epoch": 1845, "lr": 3.3887386004480525e-05}
{"train_loss": 0.11510152369737625, "global_step": 164273, "epoch": 1845, "lr": 3.3886837195689415e-05}
{"train_loss": 0.1212993860244751, "global_step": 164274, "epoch": 1845, "lr": 3.388628838906452e-05}
{"train_loss": 0.16438625752925873, "global_step": 164275, "epoch": 1845, "lr": 3.3885739584605924e-05}
{"train_loss": 0.0766172781586647, "global_step": 164276, "epoch": 1845, "lr": 3.388519078231369e-05}
{"train_loss": 0.11598348617553711, "global_step": 164277, "epoch": 1845, "lr": 3.3884641982187906e-05}
{"train_loss": 0.06701789051294327, "global_step": 164278, "epoch": 1845, "lr": 3.388409318422861e-05}
{"train_loss": 0.08850931376218796, "global_step": 164279, "epoch": 1845, "lr": 3.388354438843591e-05}
{"train_loss": 0.12876129150390625, "global_step": 164280, "epoch": 1845, "lr": 3.388299559480989e-05}
{"train_loss": 0.058564696460962296, "global_step": 164281, "epoch": 1845, "lr": 3.388244680335058e-05}
{"train_loss": 0.047457899898290634, "global_step": 164282, "epoch": 1845, "lr": 3.38818980140581e-05}
{"train_loss": 0.11266487836837769, "global_step": 164283, "epoch": 1845, "lr": 3.3881349226932484e-05}
{"train_loss": 0.14605391025543213, "global_step": 164284, "epoch": 1845, "lr": 3.3880800441973846e-05}
{"train_loss": 0.046670351177453995, "global_step": 164285, "epoch": 1845, "lr": 3.3880251659182216e-05}
{"train_loss": 0.06259728968143463, "global_step": 164286, "epoch": 1845, "lr": 3.38797028785577e-05}
{"train_loss": 0.10661907494068146, "global_step": 164287, "epoch": 1845, "lr": 3.3879154100100355e-05}
{"train_loss": 0.10077584534883499, "global_step": 164288, "epoch": 1845, "lr": 3.387860532381028e-05}
{"train_loss": 0.08563212305307388, "global_step": 164289, "epoch": 1845, "lr": 3.387805654968751e-05}
{"train_loss": 0.07865184545516968, "global_step": 164290, "epoch": 1845, "lr": 3.3877507777732155e-05}
{"train_loss": 0.07241968810558319, "global_step": 164291, "epoch": 1845, "lr": 3.387695900794426e-05}
{"train_loss": 0.07236848026514053, "global_step": 164292, "epoch": 1845, "lr": 3.387641024032392e-05}
{"train_loss": 0.08717738506331872, "global_step": 164293, "epoch": 1845, "lr": 3.387586147487119e-05, "val_loss": 6.941695213317871, "train_action_mse_error": 20.579601287841797}
{"train_loss": 0.09833110868930817, "global_step": 164294, "epoch": 1846, "lr": 3.387531271158616e-05}
{"train_loss": 0.13242842257022858, "global_step": 164295, "epoch": 1846, "lr": 3.38747639504689e-05}
{"train_loss": 0.07985831797122955, "global_step": 164296, "epoch": 1846, "lr": 3.3874215191519465e-05}
{"train_loss": 0.0621427521109581, "global_step": 164297, "epoch": 1846, "lr": 3.387366643473797e-05}
{"train_loss": 0.17101404070854187, "global_step": 164298, "epoch": 1846, "lr": 3.387311768012444e-05}
{"train_loss": 0.14529885351657867, "global_step": 164299, "epoch": 1846, "lr": 3.387256892767899e-05}
{"train_loss": 0.027654366567730904, "global_step": 164300, "epoch": 1846, "lr": 3.3872020177401664e-05}
{"train_loss": 0.10266856849193573, "global_step": 164301, "epoch": 1846, "lr": 3.3871471429292564e-05}
{"train_loss": 0.0486057810485363, "global_step": 164302, "epoch": 1846, "lr": 3.387092268335173e-05}
{"train_loss": 0.0764794647693634, "global_step": 164303, "epoch": 1846, "lr": 3.387037393957927e-05}
{"train_loss": 0.09974777698516846, "global_step": 164304, "epoch": 1846, "lr": 3.386982519797523e-05}
{"train_loss": 0.08152849972248077, "global_step": 164305, "epoch": 1846, "lr": 3.386927645853971e-05}
{"train_loss": 0.09855980426073074, "global_step": 164306, "epoch": 1846, "lr": 3.386872772127275e-05}
{"train_loss": 0.15458723902702332, "global_step": 164307, "epoch": 1846, "lr": 3.386817898617446e-05}
{"train_loss": 0.10964886844158173, "global_step": 164308, "epoch": 1846, "lr": 3.3867630253244883e-05}
{"train_loss": 0.0864495038986206, "global_step": 164309, "epoch": 1846, "lr": 3.386708152248412e-05}
{"train_loss": 0.08389902859926224, "global_step": 164310, "epoch": 1846, "lr": 3.3866532793892214e-05}
{"train_loss": 0.0717504471540451, "global_step": 164311, "epoch": 1846, "lr": 3.386598406746928e-05}
{"train_loss": 0.09681610763072968, "global_step": 164312, "epoch": 1846, "lr": 3.386543534321536e-05}
{"train_loss": 0.06140093132853508, "global_step": 164313, "epoch": 1846, "lr": 3.386488662113052e-05}
{"train_loss": 0.14925704896450043, "global_step": 164314, "epoch": 1846, "lr": 3.386433790121486e-05}
{"train_loss": 0.11621103435754776, "global_step": 164315, "epoch": 1846, "lr": 3.386378918346843e-05}
{"train_loss": 0.030482126399874687, "global_step": 164316, "epoch": 1846, "lr": 3.3863240467891345e-05}
{"train_loss": 0.054815709590911865, "global_step": 164317, "epoch": 1846, "lr": 3.386269175448362e-05}
{"train_loss": 0.13865157961845398, "global_step": 164318, "epoch": 1846, "lr": 3.386214304324539e-05}
{"train_loss": 0.035744063556194305, "global_step": 164319, "epoch": 1846, "lr": 3.3861594334176674e-05}
{"train_loss": 0.13356353342533112, "global_step": 164320, "epoch": 1846, "lr": 3.3861045627277584e-05}
{"train_loss": 0.08659201860427856, "global_step": 164321, "epoch": 1846, "lr": 3.3860496922548165e-05}
{"train_loss": 0.10588788241147995, "global_step": 164322, "epoch": 1846, "lr": 3.3859948219988526e-05}
{"train_loss": 0.08867179602384567, "global_step": 164323, "epoch": 1846, "lr": 3.3859399519598694e-05}
{"train_loss": 0.08355333656072617, "global_step": 164324, "epoch": 1846, "lr": 3.38588508213788e-05}
{"train_loss": 0.04622736573219299, "global_step": 164325, "epoch": 1846, "lr": 3.385830212532886e-05}
{"train_loss": 0.0756022036075592, "global_step": 164326, "epoch": 1846, "lr": 3.385775343144899e-05}
{"train_loss": 0.1338738203048706, "global_step": 164327, "epoch": 1846, "lr": 3.385720473973923e-05}
{"train_loss": 0.12312439829111099, "global_step": 164328, "epoch": 1846, "lr": 3.3856656050199696e-05}
{"train_loss": 0.09255719929933548, "global_step": 164329, "epoch": 1846, "lr": 3.385610736283042e-05}
{"train_loss": 0.053969115018844604, "global_step": 164330, "epoch": 1846, "lr": 3.385555867763151e-05}
{"train_loss": 0.14095069468021393, "global_step": 164331, "epoch": 1846, "lr": 3.385500999460301e-05}
{"train_loss": 0.04248736798763275, "global_step": 164332, "epoch": 1846, "lr": 3.3854461313745e-05}
{"train_loss": 0.10689148306846619, "global_step": 164333, "epoch": 1846, "lr": 3.385391263505758e-05}
{"train_loss": 0.08246544003486633, "global_step": 164334, "epoch": 1846, "lr": 3.38533639585408e-05}
{"train_loss": 0.09004770219326019, "global_step": 164335, "epoch": 1846, "lr": 3.3852815284194736e-05}
{"train_loss": 0.08512289077043533, "global_step": 164336, "epoch": 1846, "lr": 3.3852266612019456e-05}
{"train_loss": 0.08531609177589417, "global_step": 164337, "epoch": 1846, "lr": 3.385171794201506e-05}
{"train_loss": 0.11339277774095535, "global_step": 164338, "epoch": 1846, "lr": 3.3851169274181584e-05}
{"train_loss": 0.07655427604913712, "global_step": 164339, "epoch": 1846, "lr": 3.3850620608519134e-05}
{"train_loss": 0.07293680310249329, "global_step": 164340, "epoch": 1846, "lr": 3.3850071945027765e-05}
{"train_loss": 0.06713386625051498, "global_step": 164341, "epoch": 1846, "lr": 3.3849523283707565e-05}
{"train_loss": 0.1536065638065338, "global_step": 164342, "epoch": 1846, "lr": 3.384897462455858e-05}
{"train_loss": 0.05655280873179436, "global_step": 164343, "epoch": 1846, "lr": 3.384842596758093e-05}
{"train_loss": 0.06801559031009674, "global_step": 164344, "epoch": 1846, "lr": 3.384787731277464e-05}
{"train_loss": 0.11294959485530853, "global_step": 164345, "epoch": 1846, "lr": 3.384732866013983e-05}
{"train_loss": 0.07464753091335297, "global_step": 164346, "epoch": 1846, "lr": 3.3846780009676526e-05}
{"train_loss": 0.10569995641708374, "global_step": 164347, "epoch": 1846, "lr": 3.3846231361384845e-05}
{"train_loss": 0.09221174567937851, "global_step": 164348, "epoch": 1846, "lr": 3.384568271526482e-05}
{"train_loss": 0.10916160047054291, "global_step": 164349, "epoch": 1846, "lr": 3.384513407131657e-05}
{"train_loss": 0.07118566334247589, "global_step": 164350, "epoch": 1846, "lr": 3.3844585429540135e-05}
{"train_loss": 0.05671306326985359, "global_step": 164351, "epoch": 1846, "lr": 3.384403678993558e-05}
{"train_loss": 0.1030629426240921, "global_step": 164352, "epoch": 1846, "lr": 3.384348815250302e-05}
{"train_loss": 0.09936855733394623, "global_step": 164353, "epoch": 1846, "lr": 3.384293951724249e-05}
{"train_loss": 0.06347369402647018, "global_step": 164354, "epoch": 1846, "lr": 3.3842390884154104e-05}
{"train_loss": 0.12053313851356506, "global_step": 164355, "epoch": 1846, "lr": 3.384184225323789e-05}
{"train_loss": 0.047352537512779236, "global_step": 164356, "epoch": 1846, "lr": 3.384129362449396e-05}
{"train_loss": 0.11095038056373596, "global_step": 164357, "epoch": 1846, "lr": 3.384074499792235e-05}
{"train_loss": 0.14108878374099731, "global_step": 164358, "epoch": 1846, "lr": 3.3840196373523184e-05}
{"train_loss": 0.1363653689622879, "global_step": 164359, "epoch": 1846, "lr": 3.3839647751296475e-05}
{"train_loss": 0.09928666055202484, "global_step": 164360, "epoch": 1846, "lr": 3.3839099131242344e-05}
{"train_loss": 0.07128827273845673, "global_step": 164361, "epoch": 1846, "lr": 3.3838550513360846e-05}
{"train_loss": 0.10735709965229034, "global_step": 164362, "epoch": 1846, "lr": 3.383800189765207e-05}
{"train_loss": 0.10672623664140701, "global_step": 164363, "epoch": 1846, "lr": 3.3837453284116054e-05}
{"train_loss": 0.06496249884366989, "global_step": 164364, "epoch": 1846, "lr": 3.383690467275292e-05}
{"train_loss": 0.12201229482889175, "global_step": 164365, "epoch": 1846, "lr": 3.3836356063562694e-05}
{"train_loss": 0.1052020788192749, "global_step": 164366, "epoch": 1846, "lr": 3.38358074565455e-05}
{"train_loss": 0.07022005319595337, "global_step": 164367, "epoch": 1846, "lr": 3.3835258851701355e-05}
{"train_loss": 0.09218376129865646, "global_step": 164368, "epoch": 1846, "lr": 3.383471024903039e-05}
{"train_loss": 0.13283482193946838, "global_step": 164369, "epoch": 1846, "lr": 3.3834161648532635e-05}
{"train_loss": 0.11666981130838394, "global_step": 164370, "epoch": 1846, "lr": 3.383361305020817e-05}
{"train_loss": 0.09329435229301453, "global_step": 164371, "epoch": 1846, "lr": 3.383306445405711e-05}
{"train_loss": 0.09416563808917999, "global_step": 164372, "epoch": 1846, "lr": 3.383251586007946e-05}
{"train_loss": 0.06570176780223846, "global_step": 164373, "epoch": 1846, "lr": 3.3831967268275365e-05}
{"train_loss": 0.173628568649292, "global_step": 164374, "epoch": 1846, "lr": 3.3831418678644836e-05}
{"train_loss": 0.134053036570549, "global_step": 164375, "epoch": 1846, "lr": 3.3830870091188e-05}
{"train_loss": 0.09634143114089966, "global_step": 164376, "epoch": 1846, "lr": 3.3830321505904886e-05}
{"train_loss": 0.10135497897863388, "global_step": 164377, "epoch": 1846, "lr": 3.382977292279561e-05}
{"train_loss": 0.09906509518623352, "global_step": 164378, "epoch": 1846, "lr": 3.3829224341860204e-05}
{"train_loss": 0.046175941824913025, "global_step": 164379, "epoch": 1846, "lr": 3.382867576309878e-05}
{"train_loss": 0.07348853349685669, "global_step": 164380, "epoch": 1846, "lr": 3.382812718651137e-05}
{"train_loss": 0.08575443923473358, "global_step": 164381, "epoch": 1846, "lr": 3.38275786120981e-05}
{"train_loss": 0.09370114177130581, "global_step": 164382, "epoch": 1846, "lr": 3.382703003985899e-05, "val_loss": 6.846869468688965}
{"train_loss": 0.06708839535713196, "global_step": 164383, "epoch": 1847, "lr": 3.382648146979416e-05}
{"train_loss": 0.09653905034065247, "global_step": 164384, "epoch": 1847, "lr": 3.382593290190363e-05}
{"train_loss": 0.06871863454580307, "global_step": 164385, "epoch": 1847, "lr": 3.382538433618755e-05}
{"train_loss": 0.15093719959259033, "global_step": 164386, "epoch": 1847, "lr": 3.3824835772645915e-05}
{"train_loss": 0.04861246049404144, "global_step": 164387, "epoch": 1847, "lr": 3.382428721127884e-05}
{"train_loss": 0.14494463801383972, "global_step": 164388, "epoch": 1847, "lr": 3.3823738652086414e-05}
{"train_loss": 0.06967197358608246, "global_step": 164389, "epoch": 1847, "lr": 3.382319009506866e-05}
{"train_loss": 0.112681083381176, "global_step": 164390, "epoch": 1847, "lr": 3.382264154022571e-05}
{"train_loss": 0.07746879756450653, "global_step": 164391, "epoch": 1847, "lr": 3.382209298755758e-05}
{"train_loss": 0.08426252007484436, "global_step": 164392, "epoch": 1847, "lr": 3.382154443706439e-05}
{"train_loss": 0.0760665014386177, "global_step": 164393, "epoch": 1847, "lr": 3.382099588874619e-05}
{"train_loss": 0.06497938185930252, "global_step": 164394, "epoch": 1847, "lr": 3.382044734260307e-05}
{"train_loss": 0.11699679493904114, "global_step": 164395, "epoch": 1847, "lr": 3.381989879863507e-05}
{"train_loss": 0.06657689064741135, "global_step": 164396, "epoch": 1847, "lr": 3.381935025684231e-05}
{"train_loss": 0.09101007133722305, "global_step": 164397, "epoch": 1847, "lr": 3.381880171722482e-05}
{"train_loss": 0.0847250446677208, "global_step": 164398, "epoch": 1847, "lr": 3.381825317978272e-05}
{"train_loss": 0.11544690281152725, "global_step": 164399, "epoch": 1847, "lr": 3.381770464451604e-05}
{"train_loss": 0.07684700191020966, "global_step": 164400, "epoch": 1847, "lr": 3.381715611142489e-05}
{"train_loss": 0.10800237953662872, "global_step": 164401, "epoch": 1847, "lr": 3.381660758050931e-05}
{"train_loss": 0.10688012838363647, "global_step": 164402, "epoch": 1847, "lr": 3.38160590517694e-05}
{"train_loss": 0.09557799249887466, "global_step": 164403, "epoch": 1847, "lr": 3.381551052520521e-05}
{"train_loss": 0.059045225381851196, "global_step": 164404, "epoch": 1847, "lr": 3.381496200081685e-05}
{"train_loss": 0.06262931227684021, "global_step": 164405, "epoch": 1847, "lr": 3.381441347860435e-05}
{"train_loss": 0.09797041118144989, "global_step": 164406, "epoch": 1847, "lr": 3.3813864958567806e-05}
{"train_loss": 0.0903344601392746, "global_step": 164407, "epoch": 1847, "lr": 3.381331644070731e-05}
{"train_loss": 0.08201400935649872, "global_step": 164408, "epoch": 1847, "lr": 3.3812767925022895e-05}
{"train_loss": 0.08288390189409256, "global_step": 164409, "epoch": 1847, "lr": 3.381221941151467e-05}
{"train_loss": 0.07280832529067993, "global_step": 164410, "epoch": 1847, "lr": 3.381167090018268e-05}
{"train_loss": 0.04024374485015869, "global_step": 164411, "epoch": 1847, "lr": 3.381112239102703e-05}
{"train_loss": 0.07813301682472229, "global_step": 164412, "epoch": 1847, "lr": 3.3810573884047756e-05}
{"train_loss": 0.08609514683485031, "global_step": 164413, "epoch": 1847, "lr": 3.381002537924498e-05}
{"train_loss": 0.061114370822906494, "global_step": 164414, "epoch": 1847, "lr": 3.380947687661873e-05}
{"train_loss": 0.13450570404529572, "global_step": 164415, "epoch": 1847, "lr": 3.380892837616911e-05}
{"train_loss": 0.07397463172674179, "global_step": 164416, "epoch": 1847, "lr": 3.380837987789617e-05}
{"train_loss": 0.10218802839517593, "global_step": 164417, "epoch": 1847, "lr": 3.3807831381800016e-05}
{"train_loss": 0.10267698764801025, "global_step": 164418, "epoch": 1847, "lr": 3.380728288788068e-05}
{"train_loss": 0.09647900611162186, "global_step": 164419, "epoch": 1847, "lr": 3.380673439613827e-05}
{"train_loss": 0.08039887249469757, "global_step": 164420, "epoch": 1847, "lr": 3.380618590657283e-05}
{"train_loss": 0.07850669324398041, "global_step": 164421, "epoch": 1847, "lr": 3.380563741918448e-05}
{"train_loss": 0.11151763796806335, "global_step": 164422, "epoch": 1847, "lr": 3.3805088933973235e-05}
{"train_loss": 0.0699540302157402, "global_step": 164423, "epoch": 1847, "lr": 3.3804540450939213e-05}
{"train_loss": 0.07065586000680923, "global_step": 164424, "epoch": 1847, "lr": 3.380399197008247e-05}
{"train_loss": 0.10625279694795609, "global_step": 164425, "epoch": 1847, "lr": 3.380344349140309e-05}
{"train_loss": 0.05124666541814804, "global_step": 164426, "epoch": 1847, "lr": 3.3802895014901126e-05}
{"train_loss": 0.08453383296728134, "global_step": 164427, "epoch": 1847, "lr": 3.380234654057668e-05}
{"train_loss": 0.054853033274412155, "global_step": 164428, "epoch": 1847, "lr": 3.380179806842981e-05}
{"train_loss": 0.08188610523939133, "global_step": 164429, "epoch": 1847, "lr": 3.3801249598460576e-05}
{"train_loss": 0.049495112150907516, "global_step": 164430, "epoch": 1847, "lr": 3.380070113066909e-05}
{"train_loss": 0.09637080132961273, "global_step": 164431, "epoch": 1847, "lr": 3.3800152665055377e-05}
{"train_loss": 0.12355132400989532, "global_step": 164432, "epoch": 1847, "lr": 3.3799604201619556e-05}
{"train_loss": 0.12649311125278473, "global_step": 164433, "epoch": 1847, "lr": 3.379905574036166e-05}
{"train_loss": 0.14915035665035248, "global_step": 164434, "epoch": 1847, "lr": 3.379850728128181e-05}
{"train_loss": 0.12767399847507477, "global_step": 164435, "epoch": 1847, "lr": 3.379795882438003e-05}
{"train_loss": 0.06338982284069061, "global_step": 164436, "epoch": 1847, "lr": 3.379741036965644e-05}
{"train_loss": 0.13069067895412445, "global_step": 164437, "epoch": 1847, "lr": 3.379686191711106e-05}
{"train_loss": 0.10532563179731369, "global_step": 164438, "epoch": 1847, "lr": 3.3796313466744023e-05}
{"train_loss": 0.07531667500734329, "global_step": 164439, "epoch": 1847, "lr": 3.3795765018555344e-05}
{"train_loss": 0.13206423819065094, "global_step": 164440, "epoch": 1847, "lr": 3.379521657254516e-05}
{"train_loss": 0.09981928765773773, "global_step": 164441, "epoch": 1847, "lr": 3.379466812871349e-05}
{"train_loss": 0.06051678583025932, "global_step": 164442, "epoch": 1847, "lr": 3.3794119687060435e-05}
{"train_loss": 0.07363343238830566, "global_step": 164443, "epoch": 1847, "lr": 3.3793571247586065e-05}
{"train_loss": 0.11213500052690506, "global_step": 164444, "epoch": 1847, "lr": 3.3793022810290454e-05}
{"train_loss": 0.05053107067942619, "global_step": 164445, "epoch": 1847, "lr": 3.379247437517367e-05}
{"train_loss": 0.02834520861506462, "global_step": 164446, "epoch": 1847, "lr": 3.3791925942235784e-05}
{"train_loss": 0.07895871996879578, "global_step": 164447, "epoch": 1847, "lr": 3.3791377511476895e-05}
{"train_loss": 0.06403598934412003, "global_step": 164448, "epoch": 1847, "lr": 3.3790829082897035e-05}
{"train_loss": 0.1255047619342804, "global_step": 164449, "epoch": 1847, "lr": 3.379028065649632e-05}
{"train_loss": 0.12314087152481079, "global_step": 164450, "epoch": 1847, "lr": 3.3789732232274785e-05}
{"train_loss": 0.13497142493724823, "global_step": 164451, "epoch": 1847, "lr": 3.378918381023255e-05}
{"train_loss": 0.07967473566532135, "global_step": 164452, "epoch": 1847, "lr": 3.378863539036963e-05}
{"train_loss": 0.15296344459056854, "global_step": 164453, "epoch": 1847, "lr": 3.3788086972686156e-05}
{"train_loss": 0.10101614892482758, "global_step": 164454, "epoch": 1847, "lr": 3.3787538557182154e-05}
{"train_loss": 0.07209939509630203, "global_step": 164455, "epoch": 1847, "lr": 3.3786990143857744e-05}
{"train_loss": 0.09821335226297379, "global_step": 164456, "epoch": 1847, "lr": 3.378644173271295e-05}
{"train_loss": 0.11233358085155487, "global_step": 164457, "epoch": 1847, "lr": 3.37858933237479e-05}
{"train_loss": 0.0860951617360115, "global_step": 164458, "epoch": 1847, "lr": 3.378534491696261e-05}
{"train_loss": 0.08131811022758484, "global_step": 164459, "epoch": 1847, "lr": 3.378479651235718e-05}
{"train_loss": 0.03574739396572113, "global_step": 164460, "epoch": 1847, "lr": 3.3784248109931716e-05}
{"train_loss": 0.1020171120762825, "global_step": 164461, "epoch": 1847, "lr": 3.378369970968625e-05}
{"train_loss": 0.1080596074461937, "global_step": 164462, "epoch": 1847, "lr": 3.3783151311620865e-05}
{"train_loss": 0.04251684248447418, "global_step": 164463, "epoch": 1847, "lr": 3.378260291573564e-05}
{"train_loss": 0.10425686836242676, "global_step": 164464, "epoch": 1847, "lr": 3.3782054522030644e-05}
{"train_loss": 0.15008896589279175, "global_step": 164465, "epoch": 1847, "lr": 3.3781506130505956e-05}
{"train_loss": 0.0791185274720192, "global_step": 164466, "epoch": 1847, "lr": 3.378095774116166e-05}
{"train_loss": 0.09742453694343567, "global_step": 164467, "epoch": 1847, "lr": 3.378040935399779e-05}
{"train_loss": 0.1744905710220337, "global_step": 164468, "epoch": 1847, "lr": 3.377986096901446e-05}
{"train_loss": 0.07184547185897827, "global_step": 164469, "epoch": 1847, "lr": 3.377931258621172e-05}
{"train_loss": 0.06484371423721313, "global_step": 164470, "epoch": 1847, "lr": 3.377876420558967e-05}
{"train_loss": 0.09094280109144329, "global_step": 164471, "epoch": 1847, "lr": 3.377821582714835e-05, "val_loss": 6.739120006561279}
{"train_loss": 0.10568614304065704, "global_step": 164472, "epoch": 1848, "lr": 3.377766745088787e-05}
{"train_loss": 0.09776373952627182, "global_step": 164473, "epoch": 1848, "lr": 3.3777119076808274e-05}
{"train_loss": 0.06606335192918777, "global_step": 164474, "epoch": 1848, "lr": 3.377657070490966e-05}
{"train_loss": 0.096786268055439, "global_step": 164475, "epoch": 1848, "lr": 3.377602233519206e-05}
{"train_loss": 0.056810468435287476, "global_step": 164476, "epoch": 1848, "lr": 3.3775473967655604e-05}
{"train_loss": 0.0895126461982727, "global_step": 164477, "epoch": 1848, "lr": 3.377492560230031e-05}
{"train_loss": 0.0815141424536705, "global_step": 164478, "epoch": 1848, "lr": 3.377437723912629e-05}
{"train_loss": 0.07861708849668503, "global_step": 164479, "epoch": 1848, "lr": 3.377382887813362e-05}
{"train_loss": 0.11605270206928253, "global_step": 164480, "epoch": 1848, "lr": 3.377328051932236e-05}
{"train_loss": 0.06510383635759354, "global_step": 164481, "epoch": 1848, "lr": 3.3772732162692576e-05}
{"train_loss": 0.07230768352746964, "global_step": 164482, "epoch": 1848, "lr": 3.377218380824435e-05}
{"train_loss": 0.05809487774968147, "global_step": 164483, "epoch": 1848, "lr": 3.3771635455977754e-05}
{"train_loss": 0.06549036502838135, "global_step": 164484, "epoch": 1848, "lr": 3.377108710589287e-05}
{"train_loss": 0.06299129128456116, "global_step": 164485, "epoch": 1848, "lr": 3.3770538757989776e-05}
{"train_loss": 0.08639557659626007, "global_step": 164486, "epoch": 1848, "lr": 3.3769990412268516e-05}
{"train_loss": 0.08565159142017365, "global_step": 164487, "epoch": 1848, "lr": 3.37694420687292e-05}
{"train_loss": 0.1096983328461647, "global_step": 164488, "epoch": 1848, "lr": 3.3768893727371866e-05}
{"train_loss": 0.07611832022666931, "global_step": 164489, "epoch": 1848, "lr": 3.3768345388196626e-05}
{"train_loss": 0.12957648932933807, "global_step": 164490, "epoch": 1848, "lr": 3.376779705120351e-05}
{"train_loss": 0.14076606929302216, "global_step": 164491, "epoch": 1848, "lr": 3.376724871639264e-05}
{"train_loss": 0.06022809073328972, "global_step": 164492, "epoch": 1848, "lr": 3.376670038376405e-05}
{"train_loss": 0.0971345603466034, "global_step": 164493, "epoch": 1848, "lr": 3.376615205331785e-05}
{"train_loss": 0.06213809549808502, "global_step": 164494, "epoch": 1848, "lr": 3.376560372505406e-05}
{"train_loss": 0.08327604085206985, "global_step": 164495, "epoch": 1848, "lr": 3.376505539897281e-05}
{"train_loss": 0.09675503522157669, "global_step": 164496, "epoch": 1848, "lr": 3.3764507075074134e-05}
{"train_loss": 0.10955418646335602, "global_step": 164497, "epoch": 1848, "lr": 3.3763958753358126e-05}
{"train_loss": 0.10288289189338684, "global_step": 164498, "epoch": 1848, "lr": 3.376341043382487e-05}
{"train_loss": 0.10397885739803314, "global_step": 164499, "epoch": 1848, "lr": 3.376286211647443e-05}
{"train_loss": 0.09436850249767303, "global_step": 164500, "epoch": 1848, "lr": 3.376231380130685e-05}
{"train_loss": 0.07606109976768494, "global_step": 164501, "epoch": 1848, "lr": 3.376176548832225e-05}
{"train_loss": 0.12220971286296844, "global_step": 164502, "epoch": 1848, "lr": 3.3761217177520675e-05}
{"train_loss": 0.08992363512516022, "global_step": 164503, "epoch": 1848, "lr": 3.37606688689022e-05}
{"train_loss": 0.07928318530321121, "global_step": 164504, "epoch": 1848, "lr": 3.3760120562466926e-05}
{"train_loss": 0.0931532010436058, "global_step": 164505, "epoch": 1848, "lr": 3.375957225821488e-05}
{"train_loss": 0.12128130346536636, "global_step": 164506, "epoch": 1848, "lr": 3.375902395614618e-05}
{"train_loss": 0.1232135146856308, "global_step": 164507, "epoch": 1848, "lr": 3.375847565626087e-05}
{"train_loss": 0.1319396197795868, "global_step": 164508, "epoch": 1848, "lr": 3.3757927358559046e-05}
{"train_loss": 0.16196128726005554, "global_step": 164509, "epoch": 1848, "lr": 3.375737906304076e-05}
{"train_loss": 0.1099700927734375, "global_step": 164510, "epoch": 1848, "lr": 3.375683076970611e-05}
{"train_loss": 0.09270606189966202, "global_step": 164511, "epoch": 1848, "lr": 3.3756282478555134e-05}
{"train_loss": 0.11082141846418381, "global_step": 164512, "epoch": 1848, "lr": 3.375573418958795e-05}
{"train_loss": 0.0812620297074318, "global_step": 164513, "epoch": 1848, "lr": 3.375518590280459e-05}
{"train_loss": 0.04460703581571579, "global_step": 164514, "epoch": 1848, "lr": 3.375463761820515e-05}
{"train_loss": 0.06776522099971771, "global_step": 164515, "epoch": 1848, "lr": 3.375408933578972e-05}
{"train_loss": 0.09534479677677155, "global_step": 164516, "epoch": 1848, "lr": 3.375354105555833e-05}
{"train_loss": 0.10044772177934647, "global_step": 164517, "epoch": 1848, "lr": 3.37529927775111e-05}
{"train_loss": 0.08939158171415329, "global_step": 164518, "epoch": 1848, "lr": 3.375244450164807e-05}
{"train_loss": 0.09179550409317017, "global_step": 164519, "epoch": 1848, "lr": 3.375189622796933e-05}
{"train_loss": 0.07450762391090393, "global_step": 164520, "epoch": 1848, "lr": 3.3751347956474956e-05}
{"train_loss": 0.08776853233575821, "global_step": 164521, "epoch": 1848, "lr": 3.3750799687165005e-05}
{"train_loss": 0.13905984163284302, "global_step": 164522, "epoch": 1848, "lr": 3.3750251420039566e-05}
{"train_loss": 0.1238241046667099, "global_step": 164523, "epoch": 1848, "lr": 3.3749703155098714e-05}
{"train_loss": 0.1456560343503952, "global_step": 164524, "epoch": 1848, "lr": 3.3749154892342494e-05}
{"train_loss": 0.14367054402828217, "global_step": 164525, "epoch": 1848, "lr": 3.374860663177103e-05}
{"train_loss": 0.18714085221290588, "global_step": 164526, "epoch": 1848, "lr": 3.374805837338434e-05}
{"train_loss": 0.13292296230793, "global_step": 164527, "epoch": 1848, "lr": 3.374751011718255e-05}
{"train_loss": 0.08293595910072327, "global_step": 164528, "epoch": 1848, "lr": 3.3746961863165686e-05}
{"train_loss": 0.03958402946591377, "global_step": 164529, "epoch": 1848, "lr": 3.3746413611333875e-05}
{"train_loss": 0.04227094352245331, "global_step": 164530, "epoch": 1848, "lr": 3.374586536168713e-05}
{"train_loss": 0.06400879472494125, "global_step": 164531, "epoch": 1848, "lr": 3.3745317114225575e-05}
{"train_loss": 0.06803444027900696, "global_step": 164532, "epoch": 1848, "lr": 3.374476886894925e-05}
{"train_loss": 0.08745457977056503, "global_step": 164533, "epoch": 1848, "lr": 3.3744220625858234e-05}
{"train_loss": 0.09796649217605591, "global_step": 164534, "epoch": 1848, "lr": 3.374367238495264e-05}
{"train_loss": 0.07222054153680801, "global_step": 164535, "epoch": 1848, "lr": 3.3743124146232484e-05}
{"train_loss": 0.09755318611860275, "global_step": 164536, "epoch": 1848, "lr": 3.374257590969789e-05}
{"train_loss": 0.12047852575778961, "global_step": 164537, "epoch": 1848, "lr": 3.37420276753489e-05}
{"train_loss": 0.09554169327020645, "global_step": 164538, "epoch": 1848, "lr": 3.374147944318559e-05}
{"train_loss": 0.08656204491853714, "global_step": 164539, "epoch": 1848, "lr": 3.374093121320805e-05}
{"train_loss": 0.05674183368682861, "global_step": 164540, "epoch": 1848, "lr": 3.374038298541633e-05}
{"train_loss": 0.10212956368923187, "global_step": 164541, "epoch": 1848, "lr": 3.373983475981053e-05}
{"train_loss": 0.09011179953813553, "global_step": 164542, "epoch": 1848, "lr": 3.37392865363907e-05}
{"train_loss": 0.1162538155913353, "global_step": 164543, "epoch": 1848, "lr": 3.373873831515694e-05}
{"train_loss": 0.08122175931930542, "global_step": 164544, "epoch": 1848, "lr": 3.37381900961093e-05}
{"train_loss": 0.16678522527217865, "global_step": 164545, "epoch": 1848, "lr": 3.373764187924785e-05}
{"train_loss": 0.059924088418483734, "global_step": 164546, "epoch": 1848, "lr": 3.3737093664572706e-05}
{"train_loss": 0.0841711014509201, "global_step": 164547, "epoch": 1848, "lr": 3.373654545208388e-05}
{"train_loss": 0.08408276736736298, "global_step": 164548, "epoch": 1848, "lr": 3.37359972417815e-05}
{"train_loss": 0.14390230178833008, "global_step": 164549, "epoch": 1848, "lr": 3.373544903366559e-05}
{"train_loss": 0.07284249365329742, "global_step": 164550, "epoch": 1848, "lr": 3.373490082773626e-05}
{"train_loss": 0.1249789148569107, "global_step": 164551, "epoch": 1848, "lr": 3.3734352623993596e-05}
{"train_loss": 0.10607395321130753, "global_step": 164552, "epoch": 1848, "lr": 3.373380442243762e-05}
{"train_loss": 0.14759522676467896, "global_step": 164553, "epoch": 1848, "lr": 3.373325622306847e-05}
{"train_loss": 0.17003290355205536, "global_step": 164554, "epoch": 1848, "lr": 3.373270802588615e-05}
{"train_loss": 0.08027799427509308, "global_step": 164555, "epoch": 1848, "lr": 3.37321598308908e-05}
{"train_loss": 0.058705639094114304, "global_step": 164556, "epoch": 1848, "lr": 3.373161163808244e-05}
{"train_loss": 0.09598846733570099, "global_step": 164557, "epoch": 1848, "lr": 3.373106344746118e-05}
{"train_loss": 0.06957133114337921, "global_step": 164558, "epoch": 1848, "lr": 3.373051525902707e-05}
{"train_loss": 0.10115013271570206, "global_step": 164559, "epoch": 1848, "lr": 3.372996707278021e-05}
{"train_loss": 0.09514662008104699, "global_step": 164560, "epoch": 1848, "lr": 3.372941888872064e-05, "val_loss": 6.887449741363525}
{"train_loss": 0.10905496031045914, "global_step": 164561, "epoch": 1849, "lr": 3.372887070684846e-05}
{"train_loss": 0.1358964592218399, "global_step": 164562, "epoch": 1849, "lr": 3.3728322527163735e-05}
{"train_loss": 0.12971056997776031, "global_step": 164563, "epoch": 1849, "lr": 3.3727774349666546e-05}
{"train_loss": 0.06745174527168274, "global_step": 164564, "epoch": 1849, "lr": 3.3727226174356944e-05}
{"train_loss": 0.05592837929725647, "global_step": 164565, "epoch": 1849, "lr": 3.372667800123503e-05}
{"train_loss": 0.06938265264034271, "global_step": 164566, "epoch": 1849, "lr": 3.3726129830300856e-05}
{"train_loss": 0.0758989155292511, "global_step": 164567, "epoch": 1849, "lr": 3.372558166155452e-05}
{"train_loss": 0.12089647352695465, "global_step": 164568, "epoch": 1849, "lr": 3.372503349499606e-05}
{"train_loss": 0.04551757872104645, "global_step": 164569, "epoch": 1849, "lr": 3.3724485330625566e-05}
{"train_loss": 0.10310081392526627, "global_step": 164570, "epoch": 1849, "lr": 3.3723937168443146e-05}
{"train_loss": 0.08922333270311356, "global_step": 164571, "epoch": 1849, "lr": 3.372338900844883e-05}
{"train_loss": 0.09060832858085632, "global_step": 164572, "epoch": 1849, "lr": 3.3722840850642715e-05}
{"train_loss": 0.08589135110378265, "global_step": 164573, "epoch": 1849, "lr": 3.372229269502484e-05}
{"train_loss": 0.10578787326812744, "global_step": 164574, "epoch": 1849, "lr": 3.372174454159533e-05}
{"train_loss": 0.1060163602232933, "global_step": 164575, "epoch": 1849, "lr": 3.372119639035421e-05}
{"train_loss": 0.062606580555439, "global_step": 164576, "epoch": 1849, "lr": 3.372064824130161e-05}
{"train_loss": 0.06728030741214752, "global_step": 164577, "epoch": 1849, "lr": 3.372010009443755e-05}
{"train_loss": 0.09120246022939682, "global_step": 164578, "epoch": 1849, "lr": 3.371955194976213e-05}
{"train_loss": 0.13042622804641724, "global_step": 164579, "epoch": 1849, "lr": 3.371900380727541e-05}
{"train_loss": 0.06687701493501663, "global_step": 164580, "epoch": 1849, "lr": 3.3718455666977476e-05}
{"train_loss": 0.0829601138830185, "global_step": 164581, "epoch": 1849, "lr": 3.3717907528868396e-05}
{"train_loss": 0.10530739277601242, "global_step": 164582, "epoch": 1849, "lr": 3.371735939294826e-05}
{"train_loss": 0.08617327362298965, "global_step": 164583, "epoch": 1849, "lr": 3.37168112592171e-05}
{"train_loss": 0.07282901555299759, "global_step": 164584, "epoch": 1849, "lr": 3.371626312767504e-05}
{"train_loss": 0.13751812279224396, "global_step": 164585, "epoch": 1849, "lr": 3.371571499832211e-05}
{"train_loss": 0.09391091018915176, "global_step": 164586, "epoch": 1849, "lr": 3.3715166871158396e-05}
{"train_loss": 0.07040343433618546, "global_step": 164587, "epoch": 1849, "lr": 3.3714618746184005e-05}
{"train_loss": 0.05617240071296692, "global_step": 164588, "epoch": 1849, "lr": 3.3714070623398964e-05}
{"train_loss": 0.10007890313863754, "global_step": 164589, "epoch": 1849, "lr": 3.3713522502803394e-05}
{"train_loss": 0.10906464606523514, "global_step": 164590, "epoch": 1849, "lr": 3.371297438439732e-05}
{"train_loss": 0.07934312522411346, "global_step": 164591, "epoch": 1849, "lr": 3.3712426268180854e-05}
{"train_loss": 0.07473039627075195, "global_step": 164592, "epoch": 1849, "lr": 3.371187815415403e-05}
{"train_loss": 0.11205662786960602, "global_step": 164593, "epoch": 1849, "lr": 3.371133004231697e-05}
{"train_loss": 0.12586016952991486, "global_step": 164594, "epoch": 1849, "lr": 3.371078193266971e-05}
{"train_loss": 0.09887129068374634, "global_step": 164595, "epoch": 1849, "lr": 3.371023382521235e-05}
{"train_loss": 0.09303499013185501, "global_step": 164596, "epoch": 1849, "lr": 3.3709685719944936e-05}
{"train_loss": 0.09349130839109421, "global_step": 164597, "epoch": 1849, "lr": 3.3709137616867566e-05}
{"train_loss": 0.07864897698163986, "global_step": 164598, "epoch": 1849, "lr": 3.37085895159803e-05}
{"train_loss": 0.0647447481751442, "global_step": 164599, "epoch": 1849, "lr": 3.370804141728321e-05}
{"train_loss": 0.09942620247602463, "global_step": 164600, "epoch": 1849, "lr": 3.370749332077638e-05}
{"train_loss": 0.09495961666107178, "global_step": 164601, "epoch": 1849, "lr": 3.3706945226459893e-05}
{"train_loss": 0.12922203540802002, "global_step": 164602, "epoch": 1849, "lr": 3.3706397134333786e-05}
{"train_loss": 0.06959960609674454, "global_step": 164603, "epoch": 1849, "lr": 3.370584904439817e-05}
{"train_loss": 0.07954496145248413, "global_step": 164604, "epoch": 1849, "lr": 3.370530095665309e-05}
{"train_loss": 0.06949439644813538, "global_step": 164605, "epoch": 1849, "lr": 3.3704752871098636e-05}
{"train_loss": 0.07199762016534805, "global_step": 164606, "epoch": 1849, "lr": 3.3704204787734894e-05}
{"train_loss": 0.06561990082263947, "global_step": 164607, "epoch": 1849, "lr": 3.3703656706561905e-05}
{"train_loss": 0.08820013701915741, "global_step": 164608, "epoch": 1849, "lr": 3.3703108627579776e-05}
{"train_loss": 0.08277744054794312, "global_step": 164609, "epoch": 1849, "lr": 3.370256055078855e-05}
{"train_loss": 0.08219971507787704, "global_step": 164610, "epoch": 1849, "lr": 3.370201247618834e-05}
{"train_loss": 0.09318774193525314, "global_step": 164611, "epoch": 1849, "lr": 3.370146440377917e-05}
{"train_loss": 0.06593604385852814, "global_step": 164612, "epoch": 1849, "lr": 3.370091633356115e-05}
{"train_loss": 0.11553128808736801, "global_step": 164613, "epoch": 1849, "lr": 3.3700368265534345e-05}
{"train_loss": 0.11318068206310272, "global_step": 164614, "epoch": 1849, "lr": 3.3699820199698826e-05}
{"train_loss": 0.1282498687505722, "global_step": 164615, "epoch": 1849, "lr": 3.369927213605467e-05}
{"train_loss": 0.04634958878159523, "global_step": 164616, "epoch": 1849, "lr": 3.369872407460195e-05}
{"train_loss": 0.07866322994232178, "global_step": 164617, "epoch": 1849, "lr": 3.369817601534073e-05}
{"train_loss": 0.14354392886161804, "global_step": 164618, "epoch": 1849, "lr": 3.3697627958271095e-05}
{"train_loss": 0.13648313283920288, "global_step": 164619, "epoch": 1849, "lr": 3.369707990339311e-05}
{"train_loss": 0.10087282210588455, "global_step": 164620, "epoch": 1849, "lr": 3.369653185070688e-05}
{"train_loss": 0.06727384030818939, "global_step": 164621, "epoch": 1849, "lr": 3.369598380021242e-05}
{"train_loss": 0.10715040564537048, "global_step": 164622, "epoch": 1849, "lr": 3.369543575190985e-05}
{"train_loss": 0.07857757806777954, "global_step": 164623, "epoch": 1849, "lr": 3.3694887705799215e-05}
{"train_loss": 0.08310364931821823, "global_step": 164624, "epoch": 1849, "lr": 3.369433966188061e-05}
{"train_loss": 0.05639885365962982, "global_step": 164625, "epoch": 1849, "lr": 3.3693791620154116e-05}
{"train_loss": 0.08605628460645676, "global_step": 164626, "epoch": 1849, "lr": 3.369324358061978e-05}
{"train_loss": 0.15429754555225372, "global_step": 164627, "epoch": 1849, "lr": 3.36926955432777e-05}
{"train_loss": 0.08180733770132065, "global_step": 164628, "epoch": 1849, "lr": 3.369214750812793e-05}
{"train_loss": 0.05417178198695183, "global_step": 164629, "epoch": 1849, "lr": 3.369159947517056e-05}
{"train_loss": 0.09622165560722351, "global_step": 164630, "epoch": 1849, "lr": 3.369105144440564e-05}
{"train_loss": 0.07688285410404205, "global_step": 164631, "epoch": 1849, "lr": 3.369050341583328e-05}
{"train_loss": 0.05269855260848999, "global_step": 164632, "epoch": 1849, "lr": 3.368995538945351e-05}
{"train_loss": 0.06139618903398514, "global_step": 164633, "epoch": 1849, "lr": 3.3689407365266454e-05}
{"train_loss": 0.06213265657424927, "global_step": 164634, "epoch": 1849, "lr": 3.368885934327214e-05}
{"train_loss": 0.11158985644578934, "global_step": 164635, "epoch": 1849, "lr": 3.3688311323470665e-05}
{"train_loss": 0.0941554456949234, "global_step": 164636, "epoch": 1849, "lr": 3.36877633058621e-05}
{"train_loss": 0.09420251846313477, "global_step": 164637, "epoch": 1849, "lr": 3.3687215290446506e-05}
{"train_loss": 0.0917489305138588, "global_step": 164638, "epoch": 1849, "lr": 3.368666727722398e-05}
{"train_loss": 0.09275759011507034, "global_step": 164639, "epoch": 1849, "lr": 3.368611926619458e-05}
{"train_loss": 0.08442984521389008, "global_step": 164640, "epoch": 1849, "lr": 3.368557125735837e-05}
{"train_loss": 0.04906816780567169, "global_step": 164641, "epoch": 1849, "lr": 3.368502325071543e-05}
{"train_loss": 0.07661625742912292, "global_step": 164642, "epoch": 1849, "lr": 3.3684475246265875e-05}
{"train_loss": 0.07209086418151855, "global_step": 164643, "epoch": 1849, "lr": 3.368392724400971e-05}
{"train_loss": 0.13579770922660828, "global_step": 164644, "epoch": 1849, "lr": 3.368337924394707e-05}
{"train_loss": 0.11134360730648041, "global_step": 164645, "epoch": 1849, "lr": 3.368283124607797e-05}
{"train_loss": 0.09224608540534973, "global_step": 164646, "epoch": 1849, "lr": 3.3682283250402544e-05}
{"train_loss": 0.06423736363649368, "global_step": 164647, "epoch": 1849, "lr": 3.368173525692081e-05}
{"train_loss": 0.14332109689712524, "global_step": 164648, "epoch": 1849, "lr": 3.3681187265632896e-05}
{"train_loss": 0.08993823097997837, "global_step": 164649, "epoch": 1849, "lr": 3.368063927653882e-05, "val_loss": 6.932070255279541}
{"train_loss": 0.1177060604095459, "global_step": 164650, "epoch": 1850, "lr": 3.3680091289638707e-05}
{"train_loss": 0.13926203548908234, "global_step": 164651, "epoch": 1850, "lr": 3.367954330493258e-05}
{"train_loss": 0.1485317051410675, "global_step": 164652, "epoch": 1850, "lr": 3.367899532242057e-05}
{"train_loss": 0.11259409785270691, "global_step": 164653, "epoch": 1850, "lr": 3.3678447342102704e-05}
{"train_loss": 0.0827755555510521, "global_step": 164654, "epoch": 1850, "lr": 3.367789936397907e-05}
{"train_loss": 0.07871674001216888, "global_step": 164655, "epoch": 1850, "lr": 3.367735138804975e-05}
{"train_loss": 0.12889157235622406, "global_step": 164656, "epoch": 1850, "lr": 3.3676803414314806e-05}
{"train_loss": 0.1135428175330162, "global_step": 164657, "epoch": 1850, "lr": 3.3676255442774315e-05}
{"train_loss": 0.06207702308893204, "global_step": 164658, "epoch": 1850, "lr": 3.367570747342835e-05}
{"train_loss": 0.09424661099910736, "global_step": 164659, "epoch": 1850, "lr": 3.3675159506277e-05}
{"train_loss": 0.10856690257787704, "global_step": 164660, "epoch": 1850, "lr": 3.367461154132031e-05}
{"train_loss": 0.05668952688574791, "global_step": 164661, "epoch": 1850, "lr": 3.367406357855838e-05}
{"train_loss": 0.1515587419271469, "global_step": 164662, "epoch": 1850, "lr": 3.367351561799127e-05}
{"train_loss": 0.06399271637201309, "global_step": 164663, "epoch": 1850, "lr": 3.3672967659619065e-05}
{"train_loss": 0.1274900734424591, "global_step": 164664, "epoch": 1850, "lr": 3.367241970344181e-05}
{"train_loss": 0.10501214861869812, "global_step": 164665, "epoch": 1850, "lr": 3.367187174945963e-05}
{"train_loss": 0.10084962844848633, "global_step": 164666, "epoch": 1850, "lr": 3.3671323797672536e-05}
{"train_loss": 0.05447527766227722, "global_step": 164667, "epoch": 1850, "lr": 3.367077584808066e-05}
{"train_loss": 0.1453029066324234, "global_step": 164668, "epoch": 1850, "lr": 3.367022790068402e-05}
{"train_loss": 0.09422974288463593, "global_step": 164669, "epoch": 1850, "lr": 3.366967995548275e-05}
{"train_loss": 0.06565909087657928, "global_step": 164670, "epoch": 1850, "lr": 3.366913201247687e-05}
{"train_loss": 0.023038126528263092, "global_step": 164671, "epoch": 1850, "lr": 3.3668584071666497e-05}
{"train_loss": 0.04698658362030983, "global_step": 164672, "epoch": 1850, "lr": 3.366803613305166e-05}
{"train_loss": 0.1925962269306183, "global_step": 164673, "epoch": 1850, "lr": 3.366748819663248e-05}
{"train_loss": 0.06502983719110489, "global_step": 164674, "epoch": 1850, "lr": 3.366694026240899e-05}
{"train_loss": 0.0834236741065979, "global_step": 164675, "epoch": 1850, "lr": 3.366639233038129e-05}
{"train_loss": 0.14504018425941467, "global_step": 164676, "epoch": 1850, "lr": 3.366584440054944e-05}
{"train_loss": 0.13002432882785797, "global_step": 164677, "epoch": 1850, "lr": 3.366529647291351e-05}
{"train_loss": 0.08941949903964996, "global_step": 164678, "epoch": 1850, "lr": 3.36647485474736e-05}
{"train_loss": 0.06711672246456146, "global_step": 164679, "epoch": 1850, "lr": 3.366420062422975e-05}
{"train_loss": 0.09569250792264938, "global_step": 164680, "epoch": 1850, "lr": 3.3663652703182056e-05}
{"train_loss": 0.0819932371377945, "global_step": 164681, "epoch": 1850, "lr": 3.366310478433058e-05}
{"train_loss": 0.0595492422580719, "global_step": 164682, "epoch": 1850, "lr": 3.366255686767541e-05}
{"train_loss": 0.08543998003005981, "global_step": 164683, "epoch": 1850, "lr": 3.36620089532166e-05}
{"train_loss": 0.07930227369070053, "global_step": 164684, "epoch": 1850, "lr": 3.366146104095424e-05}
{"train_loss": 0.05572133511304855, "global_step": 164685, "epoch": 1850, "lr": 3.366091313088839e-05}
{"train_loss": 0.10281534492969513, "global_step": 164686, "epoch": 1850, "lr": 3.366036522301914e-05}
{"train_loss": 0.063029944896698, "global_step": 164687, "epoch": 1850, "lr": 3.3659817317346544e-05}
{"train_loss": 0.040876325219869614, "global_step": 164688, "epoch": 1850, "lr": 3.3659269413870706e-05}
{"train_loss": 0.1415514051914215, "global_step": 164689, "epoch": 1850, "lr": 3.365872151259166e-05}
{"train_loss": 0.11551640182733536, "global_step": 164690, "epoch": 1850, "lr": 3.3658173613509516e-05}
{"train_loss": 0.11558480560779572, "global_step": 164691, "epoch": 1850, "lr": 3.3657625716624306e-05}
{"train_loss": 0.050681211054325104, "global_step": 164692, "epoch": 1850, "lr": 3.3657077821936164e-05}
{"train_loss": 0.09661214798688889, "global_step": 164693, "epoch": 1850, "lr": 3.365652992944511e-05}
{"train_loss": 0.12797537446022034, "global_step": 164694, "epoch": 1850, "lr": 3.3655982039151233e-05}
{"train_loss": 0.10844993591308594, "global_step": 164695, "epoch": 1850, "lr": 3.3655434151054614e-05}
{"train_loss": 0.16024750471115112, "global_step": 164696, "epoch": 1850, "lr": 3.365488626515532e-05}
{"train_loss": 0.06715038418769836, "global_step": 164697, "epoch": 1850, "lr": 3.365433838145343e-05}
{"train_loss": 0.09770755469799042, "global_step": 164698, "epoch": 1850, "lr": 3.365379049994901e-05}
{"train_loss": 0.1390073001384735, "global_step": 164699, "epoch": 1850, "lr": 3.3653242620642155e-05}
{"train_loss": 0.07330134510993958, "global_step": 164700, "epoch": 1850, "lr": 3.3652694743532894e-05}
{"train_loss": 0.05956730246543884, "global_step": 164701, "epoch": 1850, "lr": 3.365214686862136e-05}
{"train_loss": 0.10991078615188599, "global_step": 164702, "epoch": 1850, "lr": 3.3651598995907576e-05}
{"train_loss": 0.14316363632678986, "global_step": 164703, "epoch": 1850, "lr": 3.365105112539165e-05}
{"train_loss": 0.11071097105741501, "global_step": 164704, "epoch": 1850, "lr": 3.365050325707362e-05}
{"train_loss": 0.1031520739197731, "global_step": 164705, "epoch": 1850, "lr": 3.36499553909536e-05}
{"train_loss": 0.05344220995903015, "global_step": 164706, "epoch": 1850, "lr": 3.364940752703163e-05}
{"train_loss": 0.06928987056016922, "global_step": 164707, "epoch": 1850, "lr": 3.3648859665307805e-05}
{"train_loss": 0.16927945613861084, "global_step": 164708, "epoch": 1850, "lr": 3.364831180578218e-05}
{"train_loss": 0.09363117069005966, "global_step": 164709, "epoch": 1850, "lr": 3.364776394845487e-05}
{"train_loss": 0.09420087188482285, "global_step": 164710, "epoch": 1850, "lr": 3.364721609332588e-05}
{"train_loss": 0.10724345594644547, "global_step": 164711, "epoch": 1850, "lr": 3.364666824039536e-05}
{"train_loss": 0.13408903777599335, "global_step": 164712, "epoch": 1850, "lr": 3.364612038966333e-05}
{"train_loss": 0.10524345934391022, "global_step": 164713, "epoch": 1850, "lr": 3.364557254112986e-05}
{"train_loss": 0.09602086246013641, "global_step": 164714, "epoch": 1850, "lr": 3.364502469479508e-05}
{"train_loss": 0.06884327530860901, "global_step": 164715, "epoch": 1850, "lr": 3.3644476850658994e-05}
{"train_loss": 0.04809267446398735, "global_step": 164716, "epoch": 1850, "lr": 3.364392900872173e-05}
{"train_loss": 0.09892070293426514, "global_step": 164717, "epoch": 1850, "lr": 3.364338116898333e-05}
{"train_loss": 0.08427582681179047, "global_step": 164718, "epoch": 1850, "lr": 3.364283333144389e-05}
{"train_loss": 0.14783687889575958, "global_step": 164719, "epoch": 1850, "lr": 3.364228549610346e-05}
{"train_loss": 0.16431282460689545, "global_step": 164720, "epoch": 1850, "lr": 3.364173766296214e-05}
{"train_loss": 0.052608657628297806, "global_step": 164721, "epoch": 1850, "lr": 3.364118983201996e-05}
{"train_loss": 0.07506994158029556, "global_step": 164722, "epoch": 1850, "lr": 3.364064200327705e-05}
{"train_loss": 0.06002579629421234, "global_step": 164723, "epoch": 1850, "lr": 3.364009417673344e-05}
{"train_loss": 0.12173382192850113, "global_step": 164724, "epoch": 1850, "lr": 3.3639546352389234e-05}
{"train_loss": 0.14907881617546082, "global_step": 164725, "epoch": 1850, "lr": 3.363899853024448e-05}
{"train_loss": 0.11670350283384323, "global_step": 164726, "epoch": 1850, "lr": 3.3638450710299275e-05}
{"train_loss": 0.14476782083511353, "global_step": 164727, "epoch": 1850, "lr": 3.363790289255366e-05}
{"train_loss": 0.18146800994873047, "global_step": 164728, "epoch": 1850, "lr": 3.363735507700775e-05}
{"train_loss": 0.05846960470080376, "global_step": 164729, "epoch": 1850, "lr": 3.3636807263661585e-05}
{"train_loss": 0.11779733002185822, "global_step": 164730, "epoch": 1850, "lr": 3.3636259452515266e-05}
{"train_loss": 0.08846374601125717, "global_step": 164731, "epoch": 1850, "lr": 3.363571164356885e-05}
{"train_loss": 0.05538460239768028, "global_step": 164732, "epoch": 1850, "lr": 3.363516383682239e-05}
{"train_loss": 0.12699280679225922, "global_step": 164733, "epoch": 1850, "lr": 3.363461603227601e-05}
{"train_loss": 0.09835886210203171, "global_step": 164734, "epoch": 1850, "lr": 3.363406822992974e-05}
{"train_loss": 0.2069038301706314, "global_step": 164735, "epoch": 1850, "lr": 3.363352042978369e-05}
{"train_loss": 0.07068970799446106, "global_step": 164736, "epoch": 1850, "lr": 3.363297263183789e-05}
{"train_loss": 0.0940771996974945, "global_step": 164737, "epoch": 1850, "lr": 3.363242483609246e-05}
{"train_loss": 0.100547181128451, "global_step": 164738, "epoch": 1850, "lr": 3.363187704254742e-05, "train/sim_max_reward_0": 0.5392096504394934, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.40236478930611147, "train/sim_max_reward_3": 0.6978777380456415, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.4037480902725165, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.994057591476559, "test/sim_max_reward_4500003": 0.541289884988439, "test/sim_max_reward_4500004": 0.9966115563109712, "test/sim_max_reward_4500005": 0.9892570550379681, "test/sim_max_reward_4500006": 0.9900580617381014, "test/sim_max_reward_4500007": 0.8427671809677236, "test/sim_max_reward_4500008": 0.6480265168303441, "test/sim_max_reward_4500009": 0.050364693121764545, "test/sim_max_reward_4500010": 0.982532133032773, "test/sim_max_reward_4500011": 0.9772045425429144, "test/sim_max_reward_4500012": 0.8767557658712115, "test/sim_max_reward_4500013": 0.2877840788035654, "test/sim_max_reward_4500014": 0.9699062502160049, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.017544158064586216, "test/sim_max_reward_4500017": 0.44063648060471555, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9380610841571477, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9827120500215357, "test/sim_max_reward_4500022": 0.9338534251513463, "test/sim_max_reward_4500023": 0.6342705277969539, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.0, "test/sim_max_reward_4500026": 0.7076587006720209, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6310887220934096, "test/sim_max_reward_4500029": 0.14430973631014002, "test/sim_max_reward_4500030": 0.9950571900230574, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.9766386362004821, "test/sim_max_reward_4500033": 0.5759081802370427, "test/sim_max_reward_4500034": 0.9486981591911937, "test/sim_max_reward_4500035": 0.49792527967001365, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.9472163645727677, "test/sim_max_reward_4500038": 0.16200576649891585, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6454460626807735, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.5533361754309549, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.6683306172053727, "test/sim_max_reward_4500046": 0.5151090737056843, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.995090457803661, "test/sim_max_reward_4500049": 0.48640715402542634, "train/mean_score": 0.7732420296318745, "test/mean_score": 0.6189533480665612, "val_loss": 6.829163074493408, "train_action_mse_error": 9.412918090820312}
{"train_loss": 0.07404284179210663, "global_step": 164739, "epoch": 1851, "lr": 3.363132925120291e-05}
{"train_loss": 0.05203721672296524, "global_step": 164740, "epoch": 1851, "lr": 3.363078146205894e-05}
{"train_loss": 0.14347848296165466, "global_step": 164741, "epoch": 1851, "lr": 3.363023367511563e-05}
{"train_loss": 0.08421945571899414, "global_step": 164742, "epoch": 1851, "lr": 3.362968589037302e-05}
{"train_loss": 0.06890489906072617, "global_step": 164743, "epoch": 1851, "lr": 3.362913810783122e-05}
{"train_loss": 0.09921753406524658, "global_step": 164744, "epoch": 1851, "lr": 3.3628590327490264e-05}
{"train_loss": 0.08317872136831284, "global_step": 164745, "epoch": 1851, "lr": 3.362804254935026e-05}
{"train_loss": 0.09584657102823257, "global_step": 164746, "epoch": 1851, "lr": 3.362749477341125e-05}
{"train_loss": 0.07032765448093414, "global_step": 164747, "epoch": 1851, "lr": 3.362694699967333e-05}
{"train_loss": 0.1203286200761795, "global_step": 164748, "epoch": 1851, "lr": 3.3626399228136565e-05}
{"train_loss": 0.12893930077552795, "global_step": 164749, "epoch": 1851, "lr": 3.362585145880105e-05}
{"train_loss": 0.09010789543390274, "global_step": 164750, "epoch": 1851, "lr": 3.362530369166682e-05}
{"train_loss": 0.11305628716945648, "global_step": 164751, "epoch": 1851, "lr": 3.3624755926733964e-05}
{"train_loss": 0.045858290046453476, "global_step": 164752, "epoch": 1851, "lr": 3.362420816400258e-05}
{"train_loss": 0.09963113814592361, "global_step": 164753, "epoch": 1851, "lr": 3.36236604034727e-05}
{"train_loss": 0.07336267083883286, "global_step": 164754, "epoch": 1851, "lr": 3.362311264514444e-05}
{"train_loss": 0.08285927772521973, "global_step": 164755, "epoch": 1851, "lr": 3.3622564889017837e-05}
{"train_loss": 0.05731384828686714, "global_step": 164756, "epoch": 1851, "lr": 3.3622017135093e-05}
{"train_loss": 0.09351706504821777, "global_step": 164757, "epoch": 1851, "lr": 3.362146938336995e-05}
{"train_loss": 0.0996059849858284, "global_step": 164758, "epoch": 1851, "lr": 3.362092163384883e-05}
{"train_loss": 0.11744395643472672, "global_step": 164759, "epoch": 1851, "lr": 3.362037388652965e-05}
{"train_loss": 0.06609403342008591, "global_step": 164760, "epoch": 1851, "lr": 3.361982614141253e-05}
{"train_loss": 0.0862463191151619, "global_step": 164761, "epoch": 1851, "lr": 3.3619278398497504e-05}
{"train_loss": 0.10645825415849686, "global_step": 164762, "epoch": 1851, "lr": 3.361873065778469e-05}
{"train_loss": 0.08788944780826569, "global_step": 164763, "epoch": 1851, "lr": 3.3618182919274124e-05}
{"train_loss": 0.09369844943284988, "global_step": 164764, "epoch": 1851, "lr": 3.3617635182965904e-05}
{"train_loss": 0.05806056782603264, "global_step": 164765, "epoch": 1851, "lr": 3.361708744886008e-05}
{"train_loss": 0.0674939975142479, "global_step": 164766, "epoch": 1851, "lr": 3.361653971695675e-05}
{"train_loss": 0.08709261566400528, "global_step": 164767, "epoch": 1851, "lr": 3.361599198725596e-05}
{"train_loss": 0.10917270183563232, "global_step": 164768, "epoch": 1851, "lr": 3.361544425975781e-05}
{"train_loss": 0.07506675273180008, "global_step": 164769, "epoch": 1851, "lr": 3.361489653446237e-05}
{"train_loss": 0.11846470832824707, "global_step": 164770, "epoch": 1851, "lr": 3.36143488113697e-05}
{"train_loss": 0.13088713586330414, "global_step": 164771, "epoch": 1851, "lr": 3.361380109047989e-05}
{"train_loss": 0.06795288622379303, "global_step": 164772, "epoch": 1851, "lr": 3.361325337179299e-05}
{"train_loss": 0.12574447691440582, "global_step": 164773, "epoch": 1851, "lr": 3.3612705655309104e-05}
{"train_loss": 0.07404154539108276, "global_step": 164774, "epoch": 1851, "lr": 3.361215794102828e-05}
{"train_loss": 0.08442219346761703, "global_step": 164775, "epoch": 1851, "lr": 3.3611610228950616e-05}
{"train_loss": 0.06934764236211777, "global_step": 164776, "epoch": 1851, "lr": 3.3611062519076146e-05}
{"train_loss": 0.09242436289787292, "global_step": 164777, "epoch": 1851, "lr": 3.361051481140499e-05}
{"train_loss": 0.07742074877023697, "global_step": 164778, "epoch": 1851, "lr": 3.360996710593719e-05}
{"train_loss": 0.08570803701877594, "global_step": 164779, "epoch": 1851, "lr": 3.3609419402672845e-05}
{"train_loss": 0.08189570903778076, "global_step": 164780, "epoch": 1851, "lr": 3.360887170161199e-05}
{"train_loss": 0.09621401876211166, "global_step": 164781, "epoch": 1851, "lr": 3.360832400275474e-05}
{"train_loss": 0.08054258674383163, "global_step": 164782, "epoch": 1851, "lr": 3.3607776306101134e-05}
{"train_loss": 0.05656491965055466, "global_step": 164783, "epoch": 1851, "lr": 3.360722861165129e-05}
{"train_loss": 0.1131301000714302, "global_step": 164784, "epoch": 1851, "lr": 3.3606680919405235e-05}
{"train_loss": 0.08752955496311188, "global_step": 164785, "epoch": 1851, "lr": 3.360613322936307e-05}
{"train_loss": 0.09237851947546005, "global_step": 164786, "epoch": 1851, "lr": 3.360558554152484e-05}
{"train_loss": 0.2030838429927826, "global_step": 164787, "epoch": 1851, "lr": 3.3605037855890664e-05}
{"train_loss": 0.11931163817644119, "global_step": 164788, "epoch": 1851, "lr": 3.3604490172460576e-05}
{"train_loss": 0.08486247807741165, "global_step": 164789, "epoch": 1851, "lr": 3.360394249123468e-05}
{"train_loss": 0.08767074346542358, "global_step": 164790, "epoch": 1851, "lr": 3.360339481221303e-05}
{"train_loss": 0.07675992697477341, "global_step": 164791, "epoch": 1851, "lr": 3.360284713539569e-05}
{"train_loss": 0.1456243097782135, "global_step": 164792, "epoch": 1851, "lr": 3.360229946078276e-05}
{"train_loss": 0.07873544842004776, "global_step": 164793, "epoch": 1851, "lr": 3.360175178837429e-05}
{"train_loss": 0.09210014343261719, "global_step": 164794, "epoch": 1851, "lr": 3.360120411817038e-05}
{"train_loss": 0.06681012362241745, "global_step": 164795, "epoch": 1851, "lr": 3.3600656450171074e-05}
{"train_loss": 0.09823250025510788, "global_step": 164796, "epoch": 1851, "lr": 3.360010878437647e-05}
{"train_loss": 0.13681833446025848, "global_step": 164797, "epoch": 1851, "lr": 3.359956112078662e-05}
{"train_loss": 0.0776982307434082, "global_step": 164798, "epoch": 1851, "lr": 3.359901345940163e-05}
{"train_loss": 0.06428293883800507, "global_step": 164799, "epoch": 1851, "lr": 3.359846580022153e-05}
{"train_loss": 0.08417203277349472, "global_step": 164800, "epoch": 1851, "lr": 3.359791814324643e-05}
{"train_loss": 0.07229773700237274, "global_step": 164801, "epoch": 1851, "lr": 3.3597370488476375e-05}
{"train_loss": 0.07176462560892105, "global_step": 164802, "epoch": 1851, "lr": 3.3596822835911476e-05}
{"train_loss": 0.09564489126205444, "global_step": 164803, "epoch": 1851, "lr": 3.359627518555176e-05}
{"train_loss": 0.10452047735452652, "global_step": 164804, "epoch": 1851, "lr": 3.359572753739734e-05}
{"train_loss": 0.09775689989328384, "global_step": 164805, "epoch": 1851, "lr": 3.359517989144827e-05}
{"train_loss": 0.0966925248503685, "global_step": 164806, "epoch": 1851, "lr": 3.359463224770463e-05}
{"train_loss": 0.10982237756252289, "global_step": 164807, "epoch": 1851, "lr": 3.359408460616649e-05}
{"train_loss": 0.08011693507432938, "global_step": 164808, "epoch": 1851, "lr": 3.359353696683393e-05}
{"train_loss": 0.08910562843084335, "global_step": 164809, "epoch": 1851, "lr": 3.3592989329707024e-05}
{"train_loss": 0.06521882861852646, "global_step": 164810, "epoch": 1851, "lr": 3.359244169478583e-05}
{"train_loss": 0.06866665184497833, "global_step": 164811, "epoch": 1851, "lr": 3.359189406207044e-05}
{"train_loss": 0.12849509716033936, "global_step": 164812, "epoch": 1851, "lr": 3.35913464315609e-05}
{"train_loss": 0.06377182900905609, "global_step": 164813, "epoch": 1851, "lr": 3.359079880325733e-05}
{"train_loss": 0.059459127485752106, "global_step": 164814, "epoch": 1851, "lr": 3.3590251177159755e-05}
{"train_loss": 0.05674643814563751, "global_step": 164815, "epoch": 1851, "lr": 3.358970355326829e-05}
{"train_loss": 0.06684659421443939, "global_step": 164816, "epoch": 1851, "lr": 3.3589155931582974e-05}
{"train_loss": 0.1100539118051529, "global_step": 164817, "epoch": 1851, "lr": 3.358860831210391e-05}
{"train_loss": 0.11637890338897705, "global_step": 164818, "epoch": 1851, "lr": 3.358806069483114e-05}
{"train_loss": 0.11915141344070435, "global_step": 164819, "epoch": 1851, "lr": 3.358751307976478e-05}
{"train_loss": 0.07027217745780945, "global_step": 164820, "epoch": 1851, "lr": 3.358696546690485e-05}
{"train_loss": 0.08898928761482239, "global_step": 164821, "epoch": 1851, "lr": 3.358641785625147e-05}
{"train_loss": 0.12856675684452057, "global_step": 164822, "epoch": 1851, "lr": 3.3585870247804684e-05}
{"train_loss": 0.12051685154438019, "global_step": 164823, "epoch": 1851, "lr": 3.3585322641564586e-05}
{"train_loss": 0.09200824052095413, "global_step": 164824, "epoch": 1851, "lr": 3.358477503753124e-05}
{"train_loss": 0.1005808413028717, "global_step": 164825, "epoch": 1851, "lr": 3.3584227435704725e-05}
{"train_loss": 0.11372876167297363, "global_step": 164826, "epoch": 1851, "lr": 3.3583679836085105e-05}
{"train_loss": 0.09267547839645589, "global_step": 164827, "epoch": 1851, "lr": 3.3583132238672466e-05, "val_loss": 7.0587005615234375}
{"train_loss": 0.07515136897563934, "global_step": 164828, "epoch": 1852, "lr": 3.358258464346688e-05}
{"train_loss": 0.06400446593761444, "global_step": 164829, "epoch": 1852, "lr": 3.35820370504684e-05}
{"train_loss": 0.05389624089002609, "global_step": 164830, "epoch": 1852, "lr": 3.358148945967713e-05}
{"train_loss": 0.14444127678871155, "global_step": 164831, "epoch": 1852, "lr": 3.358094187109311e-05}
{"train_loss": 0.07586723566055298, "global_step": 164832, "epoch": 1852, "lr": 3.358039428471644e-05}
{"train_loss": 0.07764089852571487, "global_step": 164833, "epoch": 1852, "lr": 3.3579846700547184e-05}
{"train_loss": 0.06221798062324524, "global_step": 164834, "epoch": 1852, "lr": 3.3579299118585435e-05}
{"train_loss": 0.0789349302649498, "global_step": 164835, "epoch": 1852, "lr": 3.3578751538831224e-05}
{"train_loss": 0.1194719597697258, "global_step": 164836, "epoch": 1852, "lr": 3.357820396128467e-05}
{"train_loss": 0.13888515532016754, "global_step": 164837, "epoch": 1852, "lr": 3.357765638594581e-05}
{"train_loss": 0.10787616670131683, "global_step": 164838, "epoch": 1852, "lr": 3.357710881281475e-05}
{"train_loss": 0.10550761967897415, "global_step": 164839, "epoch": 1852, "lr": 3.357656124189152e-05}
{"train_loss": 0.10115431994199753, "global_step": 164840, "epoch": 1852, "lr": 3.357601367317623e-05}
{"train_loss": 0.12290637195110321, "global_step": 164841, "epoch": 1852, "lr": 3.357546610666897e-05}
{"train_loss": 0.12287478893995285, "global_step": 164842, "epoch": 1852, "lr": 3.357491854236977e-05}
{"train_loss": 0.10158725082874298, "global_step": 164843, "epoch": 1852, "lr": 3.357437098027873e-05}
{"train_loss": 0.0702543705701828, "global_step": 164844, "epoch": 1852, "lr": 3.357382342039591e-05}
{"train_loss": 0.07308127731084824, "global_step": 164845, "epoch": 1852, "lr": 3.3573275862721384e-05}
{"train_loss": 0.14860379695892334, "global_step": 164846, "epoch": 1852, "lr": 3.357272830725524e-05}
{"train_loss": 0.11511269956827164, "global_step": 164847, "epoch": 1852, "lr": 3.357218075399755e-05}
{"train_loss": 0.04262220114469528, "global_step": 164848, "epoch": 1852, "lr": 3.357163320294836e-05}
{"train_loss": 0.10856468975543976, "global_step": 164849, "epoch": 1852, "lr": 3.3571085654107783e-05}
{"train_loss": 0.08249296247959137, "global_step": 164850, "epoch": 1852, "lr": 3.357053810747586e-05}
{"train_loss": 0.07580092549324036, "global_step": 164851, "epoch": 1852, "lr": 3.356999056305269e-05}
{"train_loss": 0.0672278106212616, "global_step": 164852, "epoch": 1852, "lr": 3.3569443020838307e-05}
{"train_loss": 0.07897655665874481, "global_step": 164853, "epoch": 1852, "lr": 3.356889548083285e-05}
{"train_loss": 0.0788157656788826, "global_step": 164854, "epoch": 1852, "lr": 3.356834794303632e-05}
{"train_loss": 0.054704755544662476, "global_step": 164855, "epoch": 1852, "lr": 3.3567800407448856e-05}
{"train_loss": 0.10054203867912292, "global_step": 164856, "epoch": 1852, "lr": 3.356725287407048e-05}
{"train_loss": 0.06910249590873718, "global_step": 164857, "epoch": 1852, "lr": 3.35667053429013e-05}
{"train_loss": 0.0750320628285408, "global_step": 164858, "epoch": 1852, "lr": 3.3566157813941355e-05}
{"train_loss": 0.0713251531124115, "global_step": 164859, "epoch": 1852, "lr": 3.356561028719074e-05}
{"train_loss": 0.11663487553596497, "global_step": 164860, "epoch": 1852, "lr": 3.356506276264956e-05}
{"train_loss": 0.08971376717090607, "global_step": 164861, "epoch": 1852, "lr": 3.356451524031783e-05}
{"train_loss": 0.14901886880397797, "global_step": 164862, "epoch": 1852, "lr": 3.356396772019567e-05}
{"train_loss": 0.10292357951402664, "global_step": 164863, "epoch": 1852, "lr": 3.3563420202283115e-05}
{"train_loss": 0.07086088508367538, "global_step": 164864, "epoch": 1852, "lr": 3.356287268658027e-05}
{"train_loss": 0.07286091893911362, "global_step": 164865, "epoch": 1852, "lr": 3.3562325173087196e-05}
{"train_loss": 0.14554782211780548, "global_step": 164866, "epoch": 1852, "lr": 3.356177766180397e-05}
{"train_loss": 0.059931788593530655, "global_step": 164867, "epoch": 1852, "lr": 3.356123015273065e-05}
{"train_loss": 0.06344226002693176, "global_step": 164868, "epoch": 1852, "lr": 3.3560682645867335e-05}
{"train_loss": 0.06805341690778732, "global_step": 164869, "epoch": 1852, "lr": 3.356013514121407e-05}
{"train_loss": 0.07602953165769577, "global_step": 164870, "epoch": 1852, "lr": 3.3559587638770965e-05}
{"train_loss": 0.08321601897478104, "global_step": 164871, "epoch": 1852, "lr": 3.355904013853805e-05}
{"train_loss": 0.11694197356700897, "global_step": 164872, "epoch": 1852, "lr": 3.355849264051545e-05}
{"train_loss": 0.08434471487998962, "global_step": 164873, "epoch": 1852, "lr": 3.355794514470318e-05}
{"train_loss": 0.09504447877407074, "global_step": 164874, "epoch": 1852, "lr": 3.355739765110137e-05}
{"train_loss": 0.09360940009355545, "global_step": 164875, "epoch": 1852, "lr": 3.355685015971004e-05}
{"train_loss": 0.10478094220161438, "global_step": 164876, "epoch": 1852, "lr": 3.3556302670529314e-05}
{"train_loss": 0.09024042636156082, "global_step": 164877, "epoch": 1852, "lr": 3.3555755183559234e-05}
{"train_loss": 0.07585697621107101, "global_step": 164878, "epoch": 1852, "lr": 3.355520769879987e-05}
{"train_loss": 0.13607212901115417, "global_step": 164879, "epoch": 1852, "lr": 3.355466021625133e-05}
{"train_loss": 0.12892045080661774, "global_step": 164880, "epoch": 1852, "lr": 3.3554112735913653e-05}
{"train_loss": 0.12913967669010162, "global_step": 164881, "epoch": 1852, "lr": 3.3553565257786937e-05}
{"train_loss": 0.07253174483776093, "global_step": 164882, "epoch": 1852, "lr": 3.355301778187123e-05}
{"train_loss": 0.0474468357861042, "global_step": 164883, "epoch": 1852, "lr": 3.355247030816663e-05}
{"train_loss": 0.1132308766245842, "global_step": 164884, "epoch": 1852, "lr": 3.355192283667319e-05}
{"train_loss": 0.07207642495632172, "global_step": 164885, "epoch": 1852, "lr": 3.3551375367391004e-05}
{"train_loss": 0.12895908951759338, "global_step": 164886, "epoch": 1852, "lr": 3.355082790032013e-05}
{"train_loss": 0.12289989739656448, "global_step": 164887, "epoch": 1852, "lr": 3.3550280435460656e-05}
{"train_loss": 0.10164094716310501, "global_step": 164888, "epoch": 1852, "lr": 3.3549732972812626e-05}
{"train_loss": 0.13308311998844147, "global_step": 164889, "epoch": 1852, "lr": 3.354918551237615e-05}
{"train_loss": 0.07404877990484238, "global_step": 164890, "epoch": 1852, "lr": 3.3548638054151274e-05}
{"train_loss": 0.11263878643512726, "global_step": 164891, "epoch": 1852, "lr": 3.35480905981381e-05}
{"train_loss": 0.11801266670227051, "global_step": 164892, "epoch": 1852, "lr": 3.3547543144336665e-05}
{"train_loss": 0.09152232110500336, "global_step": 164893, "epoch": 1852, "lr": 3.354699569274708e-05}
{"train_loss": 0.10342206060886383, "global_step": 164894, "epoch": 1852, "lr": 3.354644824336938e-05}
{"train_loss": 0.14553070068359375, "global_step": 164895, "epoch": 1852, "lr": 3.354590079620366e-05}
{"train_loss": 0.09622615575790405, "global_step": 164896, "epoch": 1852, "lr": 3.354535335125002e-05}
{"train_loss": 0.18876580893993378, "global_step": 164897, "epoch": 1852, "lr": 3.3544805908508476e-05}
{"train_loss": 0.08927639573812485, "global_step": 164898, "epoch": 1852, "lr": 3.354425846797916e-05}
{"train_loss": 0.13562394678592682, "global_step": 164899, "epoch": 1852, "lr": 3.3543711029662096e-05}
{"train_loss": 0.09788474440574646, "global_step": 164900, "epoch": 1852, "lr": 3.3543163593557396e-05}
{"train_loss": 0.11541301012039185, "global_step": 164901, "epoch": 1852, "lr": 3.354261615966511e-05}
{"train_loss": 0.09733620285987854, "global_step": 164902, "epoch": 1852, "lr": 3.3542068727985324e-05}
{"train_loss": 0.09617195278406143, "global_step": 164903, "epoch": 1852, "lr": 3.3541521298518095e-05}
{"train_loss": 0.18073582649230957, "global_step": 164904, "epoch": 1852, "lr": 3.354097387126351e-05}
{"train_loss": 0.08781532198190689, "global_step": 164905, "epoch": 1852, "lr": 3.354042644622165e-05}
{"train_loss": 0.11619368195533752, "global_step": 164906, "epoch": 1852, "lr": 3.3539879023392587e-05}
{"train_loss": 0.1290775090456009, "global_step": 164907, "epoch": 1852, "lr": 3.353933160277636e-05}
{"train_loss": 0.09579987078905106, "global_step": 164908, "epoch": 1852, "lr": 3.353878418437309e-05}
{"train_loss": 0.09998327493667603, "global_step": 164909, "epoch": 1852, "lr": 3.3538236768182815e-05}
{"train_loss": 0.061479054391384125, "global_step": 164910, "epoch": 1852, "lr": 3.353768935420564e-05}
{"train_loss": 0.08058567345142365, "global_step": 164911, "epoch": 1852, "lr": 3.35371419424416e-05}
{"train_loss": 0.07010582834482193, "global_step": 164912, "epoch": 1852, "lr": 3.353659453289081e-05}
{"train_loss": 0.11250689625740051, "global_step": 164913, "epoch": 1852, "lr": 3.353604712555331e-05}
{"train_loss": 0.1043587177991867, "global_step": 164914, "epoch": 1852, "lr": 3.3535499720429184e-05}
{"train_loss": 0.06880761682987213, "global_step": 164915, "epoch": 1852, "lr": 3.353495231751853e-05}
{"train_loss": 0.09706504362520207, "global_step": 164916, "epoch": 1852, "lr": 3.353440491682138e-05, "val_loss": 6.923308849334717}
{"train_loss": 0.048544690012931824, "global_step": 164917, "epoch": 1853, "lr": 3.353385751833785e-05}
{"train_loss": 0.05798788368701935, "global_step": 164918, "epoch": 1853, "lr": 3.353331012206797e-05}
{"train_loss": 0.09893503785133362, "global_step": 164919, "epoch": 1853, "lr": 3.353276272801186e-05}
{"train_loss": 0.059528641402721405, "global_step": 164920, "epoch": 1853, "lr": 3.353221533616956e-05}
{"train_loss": 0.10146235674619675, "global_step": 164921, "epoch": 1853, "lr": 3.353166794654115e-05}
{"train_loss": 0.13264986872673035, "global_step": 164922, "epoch": 1853, "lr": 3.35311205591267e-05}
{"train_loss": 0.17338627576828003, "global_step": 164923, "epoch": 1853, "lr": 3.35305731739263e-05}
{"train_loss": 0.0633171871304512, "global_step": 164924, "epoch": 1853, "lr": 3.3530025790940014e-05}
{"train_loss": 0.07086677104234695, "global_step": 164925, "epoch": 1853, "lr": 3.3529478410167923e-05}
{"train_loss": 0.08693675696849823, "global_step": 164926, "epoch": 1853, "lr": 3.352893103161007e-05}
{"train_loss": 0.05464131757616997, "global_step": 164927, "epoch": 1853, "lr": 3.352838365526657e-05}
{"train_loss": 0.09673871845006943, "global_step": 164928, "epoch": 1853, "lr": 3.3527836281137465e-05}
{"train_loss": 0.09005782008171082, "global_step": 164929, "epoch": 1853, "lr": 3.352728890922286e-05}
{"train_loss": 0.08347229659557343, "global_step": 164930, "epoch": 1853, "lr": 3.352674153952279e-05}
{"train_loss": 0.0852840319275856, "global_step": 164931, "epoch": 1853, "lr": 3.352619417203734e-05}
{"train_loss": 0.1086539551615715, "global_step": 164932, "epoch": 1853, "lr": 3.352564680676662e-05}
{"train_loss": 0.111765556037426, "global_step": 164933, "epoch": 1853, "lr": 3.352509944371066e-05}
{"train_loss": 0.10787488520145416, "global_step": 164934, "epoch": 1853, "lr": 3.3524552082869566e-05}
{"train_loss": 0.07253240793943405, "global_step": 164935, "epoch": 1853, "lr": 3.3524004724243376e-05}
{"train_loss": 0.07661095261573792, "global_step": 164936, "epoch": 1853, "lr": 3.3523457367832197e-05}
{"train_loss": 0.14342358708381653, "global_step": 164937, "epoch": 1853, "lr": 3.352291001363607e-05}
{"train_loss": 0.09093859791755676, "global_step": 164938, "epoch": 1853, "lr": 3.352236266165511e-05}
{"train_loss": 0.09959431737661362, "global_step": 164939, "epoch": 1853, "lr": 3.352181531188936e-05}
{"train_loss": 0.07800617069005966, "global_step": 164940, "epoch": 1853, "lr": 3.3521267964338896e-05}
{"train_loss": 0.09984198212623596, "global_step": 164941, "epoch": 1853, "lr": 3.35207206190038e-05}
{"train_loss": 0.043417949229478836, "global_step": 164942, "epoch": 1853, "lr": 3.352017327588414e-05}
{"train_loss": 0.05848897621035576, "global_step": 164943, "epoch": 1853, "lr": 3.351962593497999e-05}
{"train_loss": 0.04999929293990135, "global_step": 164944, "epoch": 1853, "lr": 3.351907859629144e-05}
{"train_loss": 0.10011627525091171, "global_step": 164945, "epoch": 1853, "lr": 3.351853125981852e-05}
{"train_loss": 0.0846768245100975, "global_step": 164946, "epoch": 1853, "lr": 3.3517983925561365e-05}
{"train_loss": 0.06467630714178085, "global_step": 164947, "epoch": 1853, "lr": 3.351743659351999e-05}
{"train_loss": 0.06895191967487335, "global_step": 164948, "epoch": 1853, "lr": 3.351688926369452e-05}
{"train_loss": 0.14301785826683044, "global_step": 164949, "epoch": 1853, "lr": 3.351634193608497e-05}
{"train_loss": 0.0825280249118805, "global_step": 164950, "epoch": 1853, "lr": 3.351579461069146e-05}
{"train_loss": 0.05656103789806366, "global_step": 164951, "epoch": 1853, "lr": 3.351524728751406e-05}
{"train_loss": 0.11649232357740402, "global_step": 164952, "epoch": 1853, "lr": 3.351469996655282e-05}
{"train_loss": 0.0535515733063221, "global_step": 164953, "epoch": 1853, "lr": 3.351415264780785e-05}
{"train_loss": 0.11251315474510193, "global_step": 164954, "epoch": 1853, "lr": 3.351360533127918e-05}
{"train_loss": 0.06506933271884918, "global_step": 164955, "epoch": 1853, "lr": 3.3513058016966925e-05}
{"train_loss": 0.09431736916303635, "global_step": 164956, "epoch": 1853, "lr": 3.3512510704871114e-05}
{"train_loss": 0.07289281487464905, "global_step": 164957, "epoch": 1853, "lr": 3.351196339499187e-05}
{"train_loss": 0.05265258997678757, "global_step": 164958, "epoch": 1853, "lr": 3.351141608732923e-05}
{"train_loss": 0.06338577717542648, "global_step": 164959, "epoch": 1853, "lr": 3.351086878188328e-05}
{"train_loss": 0.1008598580956459, "global_step": 164960, "epoch": 1853, "lr": 3.351032147865409e-05}
{"train_loss": 0.13567191362380981, "global_step": 164961, "epoch": 1853, "lr": 3.350977417764174e-05}
{"train_loss": 0.08560439944267273, "global_step": 164962, "epoch": 1853, "lr": 3.35092268788463e-05}
{"train_loss": 0.10044962167739868, "global_step": 164963, "epoch": 1853, "lr": 3.350867958226785e-05}
{"train_loss": 0.1000446155667305, "global_step": 164964, "epoch": 1853, "lr": 3.350813228790644e-05}
{"train_loss": 0.11783868819475174, "global_step": 164965, "epoch": 1853, "lr": 3.3507584995762175e-05}
{"train_loss": 0.08579123765230179, "global_step": 164966, "epoch": 1853, "lr": 3.35070377058351e-05}
{"train_loss": 0.0937625840306282, "global_step": 164967, "epoch": 1853, "lr": 3.35064904181253e-05}
{"train_loss": 0.07590837776660919, "global_step": 164968, "epoch": 1853, "lr": 3.3505943132632876e-05}
{"train_loss": 0.14118681848049164, "global_step": 164969, "epoch": 1853, "lr": 3.350539584935785e-05}
{"train_loss": 0.07228520512580872, "global_step": 164970, "epoch": 1853, "lr": 3.350484856830035e-05}
{"train_loss": 0.06383991241455078, "global_step": 164971, "epoch": 1853, "lr": 3.350430128946039e-05}
{"train_loss": 0.0569872111082077, "global_step": 164972, "epoch": 1853, "lr": 3.350375401283811e-05}
{"train_loss": 0.04784242808818817, "global_step": 164973, "epoch": 1853, "lr": 3.350320673843352e-05}
{"train_loss": 0.09459079802036285, "global_step": 164974, "epoch": 1853, "lr": 3.3502659466246746e-05}
{"train_loss": 0.07795985788106918, "global_step": 164975, "epoch": 1853, "lr": 3.3502112196277814e-05}
{"train_loss": 0.05513322725892067, "global_step": 164976, "epoch": 1853, "lr": 3.3501564928526854e-05}
{"train_loss": 0.10651486366987228, "global_step": 164977, "epoch": 1853, "lr": 3.350101766299388e-05}
{"train_loss": 0.22256995737552643, "global_step": 164978, "epoch": 1853, "lr": 3.350047039967901e-05}
{"train_loss": 0.10254042595624924, "global_step": 164979, "epoch": 1853, "lr": 3.349992313858229e-05}
{"train_loss": 0.06502455472946167, "global_step": 164980, "epoch": 1853, "lr": 3.349937587970381e-05}
{"train_loss": 0.16170859336853027, "global_step": 164981, "epoch": 1853, "lr": 3.349882862304363e-05}
{"train_loss": 0.11333691328763962, "global_step": 164982, "epoch": 1853, "lr": 3.349828136860185e-05}
{"train_loss": 0.07279035449028015, "global_step": 164983, "epoch": 1853, "lr": 3.34977341163785e-05}
{"train_loss": 0.057052500545978546, "global_step": 164984, "epoch": 1853, "lr": 3.3497186866373694e-05}
{"train_loss": 0.10587016493082047, "global_step": 164985, "epoch": 1853, "lr": 3.349663961858748e-05}
{"train_loss": 0.11195200681686401, "global_step": 164986, "epoch": 1853, "lr": 3.349609237301994e-05}
{"train_loss": 0.12135304510593414, "global_step": 164987, "epoch": 1853, "lr": 3.349554512967117e-05}
{"train_loss": 0.11095397174358368, "global_step": 164988, "epoch": 1853, "lr": 3.349499788854119e-05}
{"train_loss": 0.11359674483537674, "global_step": 164989, "epoch": 1853, "lr": 3.3494450649630136e-05}
{"train_loss": 0.08462333679199219, "global_step": 164990, "epoch": 1853, "lr": 3.349390341293803e-05}
{"train_loss": 0.13297073543071747, "global_step": 164991, "epoch": 1853, "lr": 3.3493356178465e-05}
{"train_loss": 0.07992122322320938, "global_step": 164992, "epoch": 1853, "lr": 3.349280894621105e-05}
{"train_loss": 0.06968632340431213, "global_step": 164993, "epoch": 1853, "lr": 3.349226171617632e-05}
{"train_loss": 0.08422597497701645, "global_step": 164994, "epoch": 1853, "lr": 3.3491714488360825e-05}
{"train_loss": 0.06113476678729057, "global_step": 164995, "epoch": 1853, "lr": 3.34911672627647e-05}
{"train_loss": 0.0920846089720726, "global_step": 164996, "epoch": 1853, "lr": 3.349062003938796e-05}
{"train_loss": 0.0794389396905899, "global_step": 164997, "epoch": 1853, "lr": 3.349007281823072e-05}
{"train_loss": 0.1293834149837494, "global_step": 164998, "epoch": 1853, "lr": 3.3489525599293036e-05}
{"train_loss": 0.10870599001646042, "global_step": 164999, "epoch": 1853, "lr": 3.3488978382574984e-05}
{"train_loss": 0.07553305476903915, "global_step": 165000, "epoch": 1853, "lr": 3.348843116807664e-05}
{"train_loss": 0.14336183667182922, "global_step": 165001, "epoch": 1853, "lr": 3.3487883955798065e-05}
{"train_loss": 0.16904914379119873, "global_step": 165002, "epoch": 1853, "lr": 3.348733674573935e-05}
{"train_loss": 0.08384931832551956, "global_step": 165003, "epoch": 1853, "lr": 3.348678953790057e-05}
{"train_loss": 0.10763750225305557, "global_step": 165004, "epoch": 1853, "lr": 3.348624233228177e-05}
{"train_loss": 0.09380989149212837, "global_step": 165005, "epoch": 1853, "lr": 3.3485695128883046e-05, "val_loss": 6.811689853668213}
{"train_loss": 0.09550108760595322, "global_step": 165006, "epoch": 1854, "lr": 3.3485147927704485e-05}
{"train_loss": 0.06478890031576157, "global_step": 165007, "epoch": 1854, "lr": 3.348460072874612e-05}
{"train_loss": 0.0700119361281395, "global_step": 165008, "epoch": 1854, "lr": 3.348405353200808e-05}
{"train_loss": 0.1443088948726654, "global_step": 165009, "epoch": 1854, "lr": 3.348350633749038e-05}
{"train_loss": 0.06023595854640007, "global_step": 165010, "epoch": 1854, "lr": 3.348295914519314e-05}
{"train_loss": 0.06056996062397957, "global_step": 165011, "epoch": 1854, "lr": 3.34824119551164e-05}
{"train_loss": 0.07718927413225174, "global_step": 165012, "epoch": 1854, "lr": 3.348186476726027e-05}
{"train_loss": 0.06831042468547821, "global_step": 165013, "epoch": 1854, "lr": 3.3481317581624776e-05}
{"train_loss": 0.1523115634918213, "global_step": 165014, "epoch": 1854, "lr": 3.348077039821004e-05}
{"train_loss": 0.08529305458068848, "global_step": 165015, "epoch": 1854, "lr": 3.348022321701609e-05}
{"train_loss": 0.04515903443098068, "global_step": 165016, "epoch": 1854, "lr": 3.347967603804305e-05}
{"train_loss": 0.10919070988893509, "global_step": 165017, "epoch": 1854, "lr": 3.347912886129095e-05}
{"train_loss": 0.06254330277442932, "global_step": 165018, "epoch": 1854, "lr": 3.347858168675988e-05}
{"train_loss": 0.0991450771689415, "global_step": 165019, "epoch": 1854, "lr": 3.347803451444992e-05}
{"train_loss": 0.07326759397983551, "global_step": 165020, "epoch": 1854, "lr": 3.347748734436113e-05}
{"train_loss": 0.08507966250181198, "global_step": 165021, "epoch": 1854, "lr": 3.34769401764936e-05}
{"train_loss": 0.057428162544965744, "global_step": 165022, "epoch": 1854, "lr": 3.347639301084737e-05}
{"train_loss": 0.07714735716581345, "global_step": 165023, "epoch": 1854, "lr": 3.347584584742257e-05}
{"train_loss": 0.14064617455005646, "global_step": 165024, "epoch": 1854, "lr": 3.347529868621921e-05}
{"train_loss": 0.1258409321308136, "global_step": 165025, "epoch": 1854, "lr": 3.3474751527237426e-05}
{"train_loss": 0.0814950093626976, "global_step": 165026, "epoch": 1854, "lr": 3.347420437047723e-05}
{"train_loss": 0.06698192656040192, "global_step": 165027, "epoch": 1854, "lr": 3.3473657215938744e-05}
{"train_loss": 0.06973826885223389, "global_step": 165028, "epoch": 1854, "lr": 3.347311006362202e-05}
{"train_loss": 0.1039501503109932, "global_step": 165029, "epoch": 1854, "lr": 3.347256291352714e-05}
{"train_loss": 0.05920560657978058, "global_step": 165030, "epoch": 1854, "lr": 3.347201576565416e-05}
{"train_loss": 0.0651177316904068, "global_step": 165031, "epoch": 1854, "lr": 3.347146862000318e-05}
{"train_loss": 0.06903136521577835, "global_step": 165032, "epoch": 1854, "lr": 3.347092147657425e-05}
{"train_loss": 0.07099691033363342, "global_step": 165033, "epoch": 1854, "lr": 3.347037433536746e-05}
{"train_loss": 0.17291772365570068, "global_step": 165034, "epoch": 1854, "lr": 3.3469827196382864e-05}
{"train_loss": 0.0654527023434639, "global_step": 165035, "epoch": 1854, "lr": 3.346928005962057e-05}
{"train_loss": 0.060874924063682556, "global_step": 165036, "epoch": 1854, "lr": 3.346873292508061e-05}
{"train_loss": 0.12665636837482452, "global_step": 165037, "epoch": 1854, "lr": 3.3468185792763084e-05}
{"train_loss": 0.1009727194905281, "global_step": 165038, "epoch": 1854, "lr": 3.3467638662668056e-05}
{"train_loss": 0.053309038281440735, "global_step": 165039, "epoch": 1854, "lr": 3.346709153479561e-05}
{"train_loss": 0.06698031723499298, "global_step": 165040, "epoch": 1854, "lr": 3.34665444091458e-05}
{"train_loss": 0.08413388580083847, "global_step": 165041, "epoch": 1854, "lr": 3.3465997285718714e-05}
{"train_loss": 0.07865098118782043, "global_step": 165042, "epoch": 1854, "lr": 3.3465450164514435e-05}
{"train_loss": 0.10770811140537262, "global_step": 165043, "epoch": 1854, "lr": 3.3464903045533015e-05}
{"train_loss": 0.12940983474254608, "global_step": 165044, "epoch": 1854, "lr": 3.346435592877455e-05}
{"train_loss": 0.1357559710741043, "global_step": 165045, "epoch": 1854, "lr": 3.3463808814239076e-05}
{"train_loss": 0.06890536099672318, "global_step": 165046, "epoch": 1854, "lr": 3.346326170192672e-05}
{"train_loss": 0.07599660009145737, "global_step": 165047, "epoch": 1854, "lr": 3.34627145918375e-05}
{"train_loss": 0.09480861574411392, "global_step": 165048, "epoch": 1854, "lr": 3.346216748397154e-05}
{"train_loss": 0.12805092334747314, "global_step": 165049, "epoch": 1854, "lr": 3.3461620378328865e-05}
{"train_loss": 0.10784891247749329, "global_step": 165050, "epoch": 1854, "lr": 3.34610732749096e-05}
{"train_loss": 0.13294215500354767, "global_step": 165051, "epoch": 1854, "lr": 3.3460526173713766e-05}
{"train_loss": 0.0981135368347168, "global_step": 165052, "epoch": 1854, "lr": 3.3459979074741493e-05}
{"train_loss": 0.060938481241464615, "global_step": 165053, "epoch": 1854, "lr": 3.34594319779928e-05}
{"train_loss": 0.07208788394927979, "global_step": 165054, "epoch": 1854, "lr": 3.34588848834678e-05}
{"train_loss": 0.06403131037950516, "global_step": 165055, "epoch": 1854, "lr": 3.345833779116654e-05}
{"train_loss": 0.07713352888822556, "global_step": 165056, "epoch": 1854, "lr": 3.345779070108911e-05}
{"train_loss": 0.07188507914543152, "global_step": 165057, "epoch": 1854, "lr": 3.345724361323558e-05}
{"train_loss": 0.06429779529571533, "global_step": 165058, "epoch": 1854, "lr": 3.345669652760601e-05}
{"train_loss": 0.08071519434452057, "global_step": 165059, "epoch": 1854, "lr": 3.3456149444200504e-05}
{"train_loss": 0.0812385231256485, "global_step": 165060, "epoch": 1854, "lr": 3.34556023630191e-05}
{"train_loss": 0.07845098525285721, "global_step": 165061, "epoch": 1854, "lr": 3.345505528406191e-05}
{"train_loss": 0.04271402955055237, "global_step": 165062, "epoch": 1854, "lr": 3.345450820732896e-05}
{"train_loss": 0.12499850988388062, "global_step": 165063, "epoch": 1854, "lr": 3.345396113282037e-05}
{"train_loss": 0.15312504768371582, "global_step": 165064, "epoch": 1854, "lr": 3.3453414060536176e-05}
{"train_loss": 0.08846896141767502, "global_step": 165065, "epoch": 1854, "lr": 3.3452866990476486e-05}
{"train_loss": 0.05355656147003174, "global_step": 165066, "epoch": 1854, "lr": 3.345231992264134e-05}
{"train_loss": 0.09669524431228638, "global_step": 165067, "epoch": 1854, "lr": 3.345177285703085e-05}
{"train_loss": 0.0902177095413208, "global_step": 165068, "epoch": 1854, "lr": 3.345122579364504e-05}
{"train_loss": 0.05554455891251564, "global_step": 165069, "epoch": 1854, "lr": 3.3450678732484035e-05}
{"train_loss": 0.08940496295690536, "global_step": 165070, "epoch": 1854, "lr": 3.345013167354787e-05}
{"train_loss": 0.15615253150463104, "global_step": 165071, "epoch": 1854, "lr": 3.344958461683665e-05}
{"train_loss": 0.11607811599969864, "global_step": 165072, "epoch": 1854, "lr": 3.3449037562350404e-05}
{"train_loss": 0.058762963861227036, "global_step": 165073, "epoch": 1854, "lr": 3.344849051008926e-05}
{"train_loss": 0.15923543274402618, "global_step": 165074, "epoch": 1854, "lr": 3.344794346005325e-05}
{"train_loss": 0.04778408259153366, "global_step": 165075, "epoch": 1854, "lr": 3.344739641224247e-05}
{"train_loss": 0.059643860906362534, "global_step": 165076, "epoch": 1854, "lr": 3.344684936665698e-05}
{"train_loss": 0.08039043843746185, "global_step": 165077, "epoch": 1854, "lr": 3.3446302323296854e-05}
{"train_loss": 0.08353587985038757, "global_step": 165078, "epoch": 1854, "lr": 3.344575528216219e-05}
{"train_loss": 0.09682468324899673, "global_step": 165079, "epoch": 1854, "lr": 3.3445208243253014e-05}
{"train_loss": 0.14746546745300293, "global_step": 165080, "epoch": 1854, "lr": 3.344466120656946e-05}
{"train_loss": 0.08179330080747604, "global_step": 165081, "epoch": 1854, "lr": 3.344411417211154e-05}
{"train_loss": 0.0719296783208847, "global_step": 165082, "epoch": 1854, "lr": 3.3443567139879376e-05}
{"train_loss": 0.039048388600349426, "global_step": 165083, "epoch": 1854, "lr": 3.344302010987301e-05}
{"train_loss": 0.06040630489587784, "global_step": 165084, "epoch": 1854, "lr": 3.3442473082092546e-05}
{"train_loss": 0.07145517319440842, "global_step": 165085, "epoch": 1854, "lr": 3.344192605653802e-05}
{"train_loss": 0.08611464500427246, "global_step": 165086, "epoch": 1854, "lr": 3.3441379033209543e-05}
{"train_loss": 0.06848423182964325, "global_step": 165087, "epoch": 1854, "lr": 3.344083201210715e-05}
{"train_loss": 0.04776336997747421, "global_step": 165088, "epoch": 1854, "lr": 3.344028499323096e-05}
{"train_loss": 0.05683888494968414, "global_step": 165089, "epoch": 1854, "lr": 3.3439737976581e-05}
{"train_loss": 0.08150893449783325, "global_step": 165090, "epoch": 1854, "lr": 3.3439190962157385e-05}
{"train_loss": 0.10337119549512863, "global_step": 165091, "epoch": 1854, "lr": 3.343864394996015e-05}
{"train_loss": 0.0827593132853508, "global_step": 165092, "epoch": 1854, "lr": 3.34380969399894e-05}
{"train_loss": 0.03926096484065056, "global_step": 165093, "epoch": 1854, "lr": 3.3437549932245194e-05}
{"train_loss": 0.08619763191496388, "global_step": 165094, "epoch": 1854, "lr": 3.34370029267276e-05, "val_loss": 6.961448669433594}
{"train_loss": 0.07566992938518524, "global_step": 165095, "epoch": 1855, "lr": 3.343645592343671e-05}
{"train_loss": 0.11240527033805847, "global_step": 165096, "epoch": 1855, "lr": 3.343590892237257e-05}
{"train_loss": 0.09039026498794556, "global_step": 165097, "epoch": 1855, "lr": 3.343536192353529e-05}
{"train_loss": 0.07138188928365707, "global_step": 165098, "epoch": 1855, "lr": 3.3434814926924904e-05}
{"train_loss": 0.06812490522861481, "global_step": 165099, "epoch": 1855, "lr": 3.343426793254153e-05}
{"train_loss": 0.15804260969161987, "global_step": 165100, "epoch": 1855, "lr": 3.343372094038519e-05}
{"train_loss": 0.07243829220533371, "global_step": 165101, "epoch": 1855, "lr": 3.343317395045601e-05}
{"train_loss": 0.06401511281728745, "global_step": 165102, "epoch": 1855, "lr": 3.343262696275401e-05}
{"train_loss": 0.08779577910900116, "global_step": 165103, "epoch": 1855, "lr": 3.3432079977279324e-05}
{"train_loss": 0.13653886318206787, "global_step": 165104, "epoch": 1855, "lr": 3.3431532994031965e-05}
{"train_loss": 0.09276875108480453, "global_step": 165105, "epoch": 1855, "lr": 3.343098601301205e-05}
{"train_loss": 0.126344695687294, "global_step": 165106, "epoch": 1855, "lr": 3.343043903421963e-05}
{"train_loss": 0.054226312786340714, "global_step": 165107, "epoch": 1855, "lr": 3.3429892057654796e-05}
{"train_loss": 0.2141861468553543, "global_step": 165108, "epoch": 1855, "lr": 3.342934508331759e-05}
{"train_loss": 0.06840500235557556, "global_step": 165109, "epoch": 1855, "lr": 3.342879811120812e-05}
{"train_loss": 0.06937966495752335, "global_step": 165110, "epoch": 1855, "lr": 3.3428251141326443e-05}
{"train_loss": 0.1112985834479332, "global_step": 165111, "epoch": 1855, "lr": 3.342770417367265e-05}
{"train_loss": 0.18860021233558655, "global_step": 165112, "epoch": 1855, "lr": 3.342715720824677e-05}
{"train_loss": 0.07392530888319016, "global_step": 165113, "epoch": 1855, "lr": 3.3426610245048926e-05}
{"train_loss": 0.04912896454334259, "global_step": 165114, "epoch": 1855, "lr": 3.342606328407919e-05}
{"train_loss": 0.09282355010509491, "global_step": 165115, "epoch": 1855, "lr": 3.342551632533759e-05}
{"train_loss": 0.13929040729999542, "global_step": 165116, "epoch": 1855, "lr": 3.342496936882424e-05}
{"train_loss": 0.06951980292797089, "global_step": 165117, "epoch": 1855, "lr": 3.342442241453919e-05}
{"train_loss": 0.07522151619195938, "global_step": 165118, "epoch": 1855, "lr": 3.342387546248254e-05}
{"train_loss": 0.10584849119186401, "global_step": 165119, "epoch": 1855, "lr": 3.342332851265433e-05}
{"train_loss": 0.08126456290483475, "global_step": 165120, "epoch": 1855, "lr": 3.342278156505467e-05}
{"train_loss": 0.07481499016284943, "global_step": 165121, "epoch": 1855, "lr": 3.34222346196836e-05}
{"train_loss": 0.0765870213508606, "global_step": 165122, "epoch": 1855, "lr": 3.3421687676541224e-05}
{"train_loss": 0.09696899354457855, "global_step": 165123, "epoch": 1855, "lr": 3.3421140735627576e-05}
{"train_loss": 0.07078414410352707, "global_step": 165124, "epoch": 1855, "lr": 3.342059379694278e-05}
{"train_loss": 0.1301463395357132, "global_step": 165125, "epoch": 1855, "lr": 3.342004686048686e-05}
{"train_loss": 0.1050785705447197, "global_step": 165126, "epoch": 1855, "lr": 3.3419499926259935e-05}
{"train_loss": 0.13661904633045197, "global_step": 165127, "epoch": 1855, "lr": 3.3418952994262034e-05}
{"train_loss": 0.1041632741689682, "global_step": 165128, "epoch": 1855, "lr": 3.3418406064493266e-05}
{"train_loss": 0.13769939541816711, "global_step": 165129, "epoch": 1855, "lr": 3.3417859136953686e-05}
{"train_loss": 0.09591764211654663, "global_step": 165130, "epoch": 1855, "lr": 3.341731221164338e-05}
{"train_loss": 0.1428457796573639, "global_step": 165131, "epoch": 1855, "lr": 3.34167652885624e-05}
{"train_loss": 0.10916147381067276, "global_step": 165132, "epoch": 1855, "lr": 3.341621836771085e-05}
{"train_loss": 0.1003357321023941, "global_step": 165133, "epoch": 1855, "lr": 3.3415671449088784e-05}
{"train_loss": 0.14962953329086304, "global_step": 165134, "epoch": 1855, "lr": 3.3415124532696266e-05}
{"train_loss": 0.10203269124031067, "global_step": 165135, "epoch": 1855, "lr": 3.34145776185334e-05}
{"train_loss": 0.07743434607982635, "global_step": 165136, "epoch": 1855, "lr": 3.341403070660023e-05}
{"train_loss": 0.0592811144888401, "global_step": 165137, "epoch": 1855, "lr": 3.341348379689685e-05}
{"train_loss": 0.0959366038441658, "global_step": 165138, "epoch": 1855, "lr": 3.341293688942331e-05}
{"train_loss": 0.1246945708990097, "global_step": 165139, "epoch": 1855, "lr": 3.341238998417972e-05}
{"train_loss": 0.08452056348323822, "global_step": 165140, "epoch": 1855, "lr": 3.3411843081166114e-05}
{"train_loss": 0.0683077722787857, "global_step": 165141, "epoch": 1855, "lr": 3.3411296180382604e-05}
{"train_loss": 0.12009675055742264, "global_step": 165142, "epoch": 1855, "lr": 3.341074928182921e-05}
{"train_loss": 0.04838758707046509, "global_step": 165143, "epoch": 1855, "lr": 3.341020238550607e-05}
{"train_loss": 0.10156042128801346, "global_step": 165144, "epoch": 1855, "lr": 3.340965549141321e-05}
{"train_loss": 0.13930341601371765, "global_step": 165145, "epoch": 1855, "lr": 3.340910859955073e-05}
{"train_loss": 0.05013997107744217, "global_step": 165146, "epoch": 1855, "lr": 3.340856170991867e-05}
{"train_loss": 0.10830805450677872, "global_step": 165147, "epoch": 1855, "lr": 3.340801482251716e-05}
{"train_loss": 0.05957910045981407, "global_step": 165148, "epoch": 1855, "lr": 3.340746793734621e-05}
{"train_loss": 0.06696286052465439, "global_step": 165149, "epoch": 1855, "lr": 3.340692105440594e-05}
{"train_loss": 0.17398355901241302, "global_step": 165150, "epoch": 1855, "lr": 3.34063741736964e-05}
{"train_loss": 0.06210629269480705, "global_step": 165151, "epoch": 1855, "lr": 3.340582729521767e-05}
{"train_loss": 0.05067092925310135, "global_step": 165152, "epoch": 1855, "lr": 3.340528041896985e-05}
{"train_loss": 0.048115603625774384, "global_step": 165153, "epoch": 1855, "lr": 3.340473354495296e-05}
{"train_loss": 0.07305676490068436, "global_step": 165154, "epoch": 1855, "lr": 3.340418667316711e-05}
{"train_loss": 0.06172017380595207, "global_step": 165155, "epoch": 1855, "lr": 3.340363980361235e-05}
{"train_loss": 0.09078028798103333, "global_step": 165156, "epoch": 1855, "lr": 3.340309293628879e-05}
{"train_loss": 0.09937772899866104, "global_step": 165157, "epoch": 1855, "lr": 3.340254607119647e-05}
{"train_loss": 0.0896376296877861, "global_step": 165158, "epoch": 1855, "lr": 3.3401999208335485e-05}
{"train_loss": 0.11923781782388687, "global_step": 165159, "epoch": 1855, "lr": 3.3401452347705884e-05}
{"train_loss": 0.09086154401302338, "global_step": 165160, "epoch": 1855, "lr": 3.340090548930777e-05}
{"train_loss": 0.0538288876414299, "global_step": 165161, "epoch": 1855, "lr": 3.3400358633141195e-05}
{"train_loss": 0.09378141909837723, "global_step": 165162, "epoch": 1855, "lr": 3.339981177920625e-05}
{"train_loss": 0.06963402032852173, "global_step": 165163, "epoch": 1855, "lr": 3.339926492750298e-05}
{"train_loss": 0.0942232757806778, "global_step": 165164, "epoch": 1855, "lr": 3.3398718078031486e-05}
{"train_loss": 0.07301760464906693, "global_step": 165165, "epoch": 1855, "lr": 3.339817123079182e-05}
{"train_loss": 0.11703953891992569, "global_step": 165166, "epoch": 1855, "lr": 3.3397624385784096e-05}
{"train_loss": 0.1017351821064949, "global_step": 165167, "epoch": 1855, "lr": 3.339707754300833e-05}
{"train_loss": 0.08869370073080063, "global_step": 165168, "epoch": 1855, "lr": 3.339653070246464e-05}
{"train_loss": 0.20179057121276855, "global_step": 165169, "epoch": 1855, "lr": 3.339598386415308e-05}
{"train_loss": 0.12295281887054443, "global_step": 165170, "epoch": 1855, "lr": 3.3395437028073725e-05}
{"train_loss": 0.07670653611421585, "global_step": 165171, "epoch": 1855, "lr": 3.339489019422666e-05}
{"train_loss": 0.09632119536399841, "global_step": 165172, "epoch": 1855, "lr": 3.3394343362611936e-05}
{"train_loss": 0.06800560653209686, "global_step": 165173, "epoch": 1855, "lr": 3.339379653322966e-05}
{"train_loss": 0.13017480075359344, "global_step": 165174, "epoch": 1855, "lr": 3.339324970607986e-05}
{"train_loss": 0.07239487767219543, "global_step": 165175, "epoch": 1855, "lr": 3.3392702881162664e-05}
{"train_loss": 0.11133790016174316, "global_step": 165176, "epoch": 1855, "lr": 3.3392156058478086e-05}
{"train_loss": 0.06192706152796745, "global_step": 165177, "epoch": 1855, "lr": 3.339160923802626e-05}
{"train_loss": 0.10141263902187347, "global_step": 165178, "epoch": 1855, "lr": 3.3391062419807203e-05}
{"train_loss": 0.09829139709472656, "global_step": 165179, "epoch": 1855, "lr": 3.3390515603821036e-05}
{"train_loss": 0.0550214946269989, "global_step": 165180, "epoch": 1855, "lr": 3.3389968790067796e-05}
{"train_loss": 0.07099497318267822, "global_step": 165181, "epoch": 1855, "lr": 3.3389421978547586e-05}
{"train_loss": 0.08136507123708725, "global_step": 165182, "epoch": 1855, "lr": 3.3388875169260445e-05}
{"train_loss": 0.09542593971062242, "global_step": 165183, "epoch": 1855, "lr": 3.33883283622065e-05, "val_loss": 6.862624645233154, "train_action_mse_error": 25.650390625}
{"train_loss": 0.1048312783241272, "global_step": 165184, "epoch": 1856, "lr": 3.3387781557385756e-05}
{"train_loss": 0.08704569190740585, "global_step": 165185, "epoch": 1856, "lr": 3.338723475479834e-05}
{"train_loss": 0.06495560705661774, "global_step": 165186, "epoch": 1856, "lr": 3.338668795444431e-05}
{"train_loss": 0.0589858815073967, "global_step": 165187, "epoch": 1856, "lr": 3.338614115632373e-05}
{"train_loss": 0.09366049617528915, "global_step": 165188, "epoch": 1856, "lr": 3.338559436043669e-05}
{"train_loss": 0.12370283156633377, "global_step": 165189, "epoch": 1856, "lr": 3.338504756678325e-05}
{"train_loss": 0.0617092028260231, "global_step": 165190, "epoch": 1856, "lr": 3.3384500775363496e-05}
{"train_loss": 0.05690235272049904, "global_step": 165191, "epoch": 1856, "lr": 3.338395398617747e-05}
{"train_loss": 0.07880059629678726, "global_step": 165192, "epoch": 1856, "lr": 3.33834071992253e-05}
{"train_loss": 0.1693136990070343, "global_step": 165193, "epoch": 1856, "lr": 3.3382860414507e-05}
{"train_loss": 0.11730344593524933, "global_step": 165194, "epoch": 1856, "lr": 3.3382313632022696e-05}
{"train_loss": 0.07470456510782242, "global_step": 165195, "epoch": 1856, "lr": 3.338176685177241e-05}
{"train_loss": 0.05480610579252243, "global_step": 165196, "epoch": 1856, "lr": 3.338122007375627e-05}
{"train_loss": 0.051647886633872986, "global_step": 165197, "epoch": 1856, "lr": 3.338067329797431e-05}
{"train_loss": 0.023276671767234802, "global_step": 165198, "epoch": 1856, "lr": 3.338012652442662e-05}
{"train_loss": 0.10473345220088959, "global_step": 165199, "epoch": 1856, "lr": 3.3379579753113254e-05}
{"train_loss": 0.1435474157333374, "global_step": 165200, "epoch": 1856, "lr": 3.337903298403433e-05}
{"train_loss": 0.07555580139160156, "global_step": 165201, "epoch": 1856, "lr": 3.337848621718987e-05}
{"train_loss": 0.13302502036094666, "global_step": 165202, "epoch": 1856, "lr": 3.3377939452579974e-05}
{"train_loss": 0.08874139934778214, "global_step": 165203, "epoch": 1856, "lr": 3.3377392690204704e-05}
{"train_loss": 0.09941241890192032, "global_step": 165204, "epoch": 1856, "lr": 3.337684593006415e-05}
{"train_loss": 0.06947020441293716, "global_step": 165205, "epoch": 1856, "lr": 3.3376299172158374e-05}
{"train_loss": 0.15528465807437897, "global_step": 165206, "epoch": 1856, "lr": 3.337575241648746e-05}
{"train_loss": 0.09011947363615036, "global_step": 165207, "epoch": 1856, "lr": 3.337520566305147e-05}
{"train_loss": 0.0747680589556694, "global_step": 165208, "epoch": 1856, "lr": 3.337465891185048e-05}
{"train_loss": 0.10810430347919464, "global_step": 165209, "epoch": 1856, "lr": 3.337411216288456e-05}
{"train_loss": 0.19167879223823547, "global_step": 165210, "epoch": 1856, "lr": 3.3373565416153785e-05}
{"train_loss": 0.07707635313272476, "global_step": 165211, "epoch": 1856, "lr": 3.3373018671658255e-05}
{"train_loss": 0.08012068271636963, "global_step": 165212, "epoch": 1856, "lr": 3.3372471929397984e-05}
{"train_loss": 0.09039925783872604, "global_step": 165213, "epoch": 1856, "lr": 3.337192518937311e-05}
{"train_loss": 0.06294260174036026, "global_step": 165214, "epoch": 1856, "lr": 3.337137845158366e-05}
{"train_loss": 0.05627428740262985, "global_step": 165215, "epoch": 1856, "lr": 3.337083171602975e-05}
{"train_loss": 0.042998213320970535, "global_step": 165216, "epoch": 1856, "lr": 3.33702849827114e-05}
{"train_loss": 0.11310426890850067, "global_step": 165217, "epoch": 1856, "lr": 3.336973825162873e-05}
{"train_loss": 0.11267303675413132, "global_step": 165218, "epoch": 1856, "lr": 3.3369191522781784e-05}
{"train_loss": 0.06922744959592819, "global_step": 165219, "epoch": 1856, "lr": 3.336864479617067e-05}
{"train_loss": 0.048167791217565536, "global_step": 165220, "epoch": 1856, "lr": 3.3368098071795405e-05}
{"train_loss": 0.07916954159736633, "global_step": 165221, "epoch": 1856, "lr": 3.33675513496561e-05}
{"train_loss": 0.0933883786201477, "global_step": 165222, "epoch": 1856, "lr": 3.336700462975286e-05}
{"train_loss": 0.15914790332317352, "global_step": 165223, "epoch": 1856, "lr": 3.33664579120857e-05}
{"train_loss": 0.06618021428585052, "global_step": 165224, "epoch": 1856, "lr": 3.336591119665472e-05}
{"train_loss": 0.09844879060983658, "global_step": 165225, "epoch": 1856, "lr": 3.336536448345999e-05}
{"train_loss": 0.09498544782400131, "global_step": 165226, "epoch": 1856, "lr": 3.3364817772501586e-05}
{"train_loss": 0.18449664115905762, "global_step": 165227, "epoch": 1856, "lr": 3.336427106377958e-05}
{"train_loss": 0.06307698786258698, "global_step": 165228, "epoch": 1856, "lr": 3.336372435729405e-05}
{"train_loss": 0.056980542838573456, "global_step": 165229, "epoch": 1856, "lr": 3.336317765304505e-05}
{"train_loss": 0.03823141008615494, "global_step": 165230, "epoch": 1856, "lr": 3.336263095103269e-05}
{"train_loss": 0.07445432245731354, "global_step": 165231, "epoch": 1856, "lr": 3.3362084251256996e-05}
{"train_loss": 0.15914864838123322, "global_step": 165232, "epoch": 1856, "lr": 3.3361537553718086e-05}
{"train_loss": 0.10460785031318665, "global_step": 165233, "epoch": 1856, "lr": 3.3360990858416e-05}
{"train_loss": 0.17189839482307434, "global_step": 165234, "epoch": 1856, "lr": 3.336044416535084e-05}
{"train_loss": 0.14596116542816162, "global_step": 165235, "epoch": 1856, "lr": 3.335989747452265e-05}
{"train_loss": 0.08870668709278107, "global_step": 165236, "epoch": 1856, "lr": 3.3359350785931535e-05}
{"train_loss": 0.14141352474689484, "global_step": 165237, "epoch": 1856, "lr": 3.335880409957753e-05}
{"train_loss": 0.11502518504858017, "global_step": 165238, "epoch": 1856, "lr": 3.335825741546076e-05}
{"train_loss": 0.04744242504239082, "global_step": 165239, "epoch": 1856, "lr": 3.335771073358125e-05}
{"train_loss": 0.09334076195955276, "global_step": 165240, "epoch": 1856, "lr": 3.335716405393908e-05}
{"train_loss": 0.08578003942966461, "global_step": 165241, "epoch": 1856, "lr": 3.335661737653437e-05}
{"train_loss": 0.059027109295129776, "global_step": 165242, "epoch": 1856, "lr": 3.3356070701367135e-05}
{"train_loss": 0.07218186557292938, "global_step": 165243, "epoch": 1856, "lr": 3.335552402843749e-05}
{"train_loss": 0.09632137417793274, "global_step": 165244, "epoch": 1856, "lr": 3.335497735774549e-05}
{"train_loss": 0.06749580055475235, "global_step": 165245, "epoch": 1856, "lr": 3.33544306892912e-05}
{"train_loss": 0.0655210018157959, "global_step": 165246, "epoch": 1856, "lr": 3.335388402307471e-05}
{"train_loss": 0.1418445110321045, "global_step": 165247, "epoch": 1856, "lr": 3.335333735909609e-05}
{"train_loss": 0.06917121261358261, "global_step": 165248, "epoch": 1856, "lr": 3.3352790697355405e-05}
{"train_loss": 0.1256176382303238, "global_step": 165249, "epoch": 1856, "lr": 3.3352244037852746e-05}
{"train_loss": 0.0833895355463028, "global_step": 165250, "epoch": 1856, "lr": 3.335169738058815e-05}
{"train_loss": 0.06135990098118782, "global_step": 165251, "epoch": 1856, "lr": 3.335115072556175e-05}
{"train_loss": 0.0910673588514328, "global_step": 165252, "epoch": 1856, "lr": 3.335060407277356e-05}
{"train_loss": 0.05636972561478615, "global_step": 165253, "epoch": 1856, "lr": 3.335005742222369e-05}
{"train_loss": 0.091352678835392, "global_step": 165254, "epoch": 1856, "lr": 3.334951077391218e-05}
{"train_loss": 0.08447091281414032, "global_step": 165255, "epoch": 1856, "lr": 3.3348964127839156e-05}
{"train_loss": 0.143466055393219, "global_step": 165256, "epoch": 1856, "lr": 3.3348417484004635e-05}
{"train_loss": 0.09846805036067963, "global_step": 165257, "epoch": 1856, "lr": 3.334787084240874e-05}
{"train_loss": 0.13800515234470367, "global_step": 165258, "epoch": 1856, "lr": 3.33473242030515e-05}
{"train_loss": 0.06461425125598907, "global_step": 165259, "epoch": 1856, "lr": 3.334677756593301e-05}
{"train_loss": 0.16646598279476166, "global_step": 165260, "epoch": 1856, "lr": 3.3346230931053355e-05}
{"train_loss": 0.07241877913475037, "global_step": 165261, "epoch": 1856, "lr": 3.3345684298412585e-05}
{"train_loss": 0.09199947118759155, "global_step": 165262, "epoch": 1856, "lr": 3.33451376680108e-05}
{"train_loss": 0.04654202610254288, "global_step": 165263, "epoch": 1856, "lr": 3.334459103984805e-05}
{"train_loss": 0.09072205424308777, "global_step": 165264, "epoch": 1856, "lr": 3.334404441392441e-05}
{"train_loss": 0.07122011482715607, "global_step": 165265, "epoch": 1856, "lr": 3.334349779023997e-05}
{"train_loss": 0.1074954941868782, "global_step": 165266, "epoch": 1856, "lr": 3.3342951168794796e-05}
{"train_loss": 0.0844266414642334, "global_step": 165267, "epoch": 1856, "lr": 3.334240454958895e-05}
{"train_loss": 0.047835975885391235, "global_step": 165268, "epoch": 1856, "lr": 3.334185793262253e-05}
{"train_loss": 0.1143912822008133, "global_step": 165269, "epoch": 1856, "lr": 3.3341311317895574e-05}
{"train_loss": 0.11028586328029633, "global_step": 165270, "epoch": 1856, "lr": 3.334076470540819e-05}
{"train_loss": 0.12404261529445648, "global_step": 165271, "epoch": 1856, "lr": 3.3340218095160424e-05}
{"train_loss": 0.09351165580113283, "global_step": 165272, "epoch": 1856, "lr": 3.333967148715238e-05, "val_loss": 6.826596260070801}
{"train_loss": 0.09762781113386154, "global_step": 165273, "epoch": 1857, "lr": 3.33391248813841e-05}
{"train_loss": 0.0794026330113411, "global_step": 165274, "epoch": 1857, "lr": 3.333857827785568e-05}
{"train_loss": 0.13818378746509552, "global_step": 165275, "epoch": 1857, "lr": 3.333803167656717e-05}
{"train_loss": 0.07100656628608704, "global_step": 165276, "epoch": 1857, "lr": 3.333748507751866e-05}
{"train_loss": 0.052772555500268936, "global_step": 165277, "epoch": 1857, "lr": 3.3336938480710246e-05}
{"train_loss": 0.06905791908502579, "global_step": 165278, "epoch": 1857, "lr": 3.3336391886141946e-05}
{"train_loss": 0.1095968708395958, "global_step": 165279, "epoch": 1857, "lr": 3.333584529381389e-05}
{"train_loss": 0.030919501557946205, "global_step": 165280, "epoch": 1857, "lr": 3.333529870372611e-05}
{"train_loss": 0.06137371063232422, "global_step": 165281, "epoch": 1857, "lr": 3.333475211587872e-05}
{"train_loss": 0.10486050695180893, "global_step": 165282, "epoch": 1857, "lr": 3.333420553027175e-05}
{"train_loss": 0.056567732244729996, "global_step": 165283, "epoch": 1857, "lr": 3.33336589469053e-05}
{"train_loss": 0.07801054418087006, "global_step": 165284, "epoch": 1857, "lr": 3.333311236577943e-05}
{"train_loss": 0.06517638266086578, "global_step": 165285, "epoch": 1857, "lr": 3.333256578689422e-05}
{"train_loss": 0.08736798167228699, "global_step": 165286, "epoch": 1857, "lr": 3.3332019210249757e-05}
{"train_loss": 0.05691259726881981, "global_step": 165287, "epoch": 1857, "lr": 3.333147263584609e-05}
{"train_loss": 0.12041472643613815, "global_step": 165288, "epoch": 1857, "lr": 3.33309260636833e-05}
{"train_loss": 0.10519348084926605, "global_step": 165289, "epoch": 1857, "lr": 3.3330379493761476e-05}
{"train_loss": 0.1203685998916626, "global_step": 165290, "epoch": 1857, "lr": 3.332983292608066e-05}
{"train_loss": 0.10330092906951904, "global_step": 165291, "epoch": 1857, "lr": 3.332928636064097e-05}
{"train_loss": 0.07855289429426193, "global_step": 165292, "epoch": 1857, "lr": 3.3328739797442426e-05}
{"train_loss": 0.09957734495401382, "global_step": 165293, "epoch": 1857, "lr": 3.332819323648515e-05}
{"train_loss": 0.0800304189324379, "global_step": 165294, "epoch": 1857, "lr": 3.332764667776918e-05}
{"train_loss": 0.06958068907260895, "global_step": 165295, "epoch": 1857, "lr": 3.3327100121294605e-05}
{"train_loss": 0.09013545513153076, "global_step": 165296, "epoch": 1857, "lr": 3.332655356706151e-05}
{"train_loss": 0.1116969957947731, "global_step": 165297, "epoch": 1857, "lr": 3.332600701506995e-05}
{"train_loss": 0.1089237630367279, "global_step": 165298, "epoch": 1857, "lr": 3.332546046532002e-05}
{"train_loss": 0.14643442630767822, "global_step": 165299, "epoch": 1857, "lr": 3.332491391781176e-05}
{"train_loss": 0.067982017993927, "global_step": 165300, "epoch": 1857, "lr": 3.3324367372545275e-05}
{"train_loss": 0.11910317093133926, "global_step": 165301, "epoch": 1857, "lr": 3.332382082952062e-05}
{"train_loss": 0.09350329637527466, "global_step": 165302, "epoch": 1857, "lr": 3.3323274288737874e-05}
{"train_loss": 0.1323574036359787, "global_step": 165303, "epoch": 1857, "lr": 3.332272775019711e-05}
{"train_loss": 0.03761851787567139, "global_step": 165304, "epoch": 1857, "lr": 3.33221812138984e-05}
{"train_loss": 0.10750368982553482, "global_step": 165305, "epoch": 1857, "lr": 3.3321634679841826e-05}
{"train_loss": 0.053370166569948196, "global_step": 165306, "epoch": 1857, "lr": 3.332108814802746e-05}
{"train_loss": 0.07247348129749298, "global_step": 165307, "epoch": 1857, "lr": 3.3320541618455346e-05}
{"train_loss": 0.04536443576216698, "global_step": 165308, "epoch": 1857, "lr": 3.331999509112561e-05}
{"train_loss": 0.11504305899143219, "global_step": 165309, "epoch": 1857, "lr": 3.3319448566038274e-05}
{"train_loss": 0.0857391208410263, "global_step": 165310, "epoch": 1857, "lr": 3.331890204319346e-05}
{"train_loss": 0.11304552108049393, "global_step": 165311, "epoch": 1857, "lr": 3.3318355522591186e-05}
{"train_loss": 0.11464837938547134, "global_step": 165312, "epoch": 1857, "lr": 3.331780900423156e-05}
{"train_loss": 0.0655631348490715, "global_step": 165313, "epoch": 1857, "lr": 3.331726248811468e-05}
{"train_loss": 0.08492708951234818, "global_step": 165314, "epoch": 1857, "lr": 3.331671597424056e-05}
{"train_loss": 0.11233831942081451, "global_step": 165315, "epoch": 1857, "lr": 3.331616946260933e-05}
{"train_loss": 0.0761343464255333, "global_step": 165316, "epoch": 1857, "lr": 3.331562295322101e-05}
{"train_loss": 0.06032560393214226, "global_step": 165317, "epoch": 1857, "lr": 3.331507644607573e-05}
{"train_loss": 0.07985130697488785, "global_step": 165318, "epoch": 1857, "lr": 3.3314529941173504e-05}
{"train_loss": 0.09213872998952866, "global_step": 165319, "epoch": 1857, "lr": 3.331398343851446e-05}
{"train_loss": 0.09755824506282806, "global_step": 165320, "epoch": 1857, "lr": 3.331343693809865e-05}
{"train_loss": 0.10822468996047974, "global_step": 165321, "epoch": 1857, "lr": 3.331289043992613e-05}
{"train_loss": 0.15215115249156952, "global_step": 165322, "epoch": 1857, "lr": 3.331234394399699e-05}
{"train_loss": 0.10859216004610062, "global_step": 165323, "epoch": 1857, "lr": 3.331179745031131e-05}
{"train_loss": 0.07117218524217606, "global_step": 165324, "epoch": 1857, "lr": 3.331125095886916e-05}
{"train_loss": 0.09990877658128738, "global_step": 165325, "epoch": 1857, "lr": 3.3310704469670605e-05}
{"train_loss": 0.0785706639289856, "global_step": 165326, "epoch": 1857, "lr": 3.331015798271571e-05}
{"train_loss": 0.08013276755809784, "global_step": 165327, "epoch": 1857, "lr": 3.330961149800458e-05}
{"train_loss": 0.1081383228302002, "global_step": 165328, "epoch": 1857, "lr": 3.330906501553724e-05}
{"train_loss": 0.12338785827159882, "global_step": 165329, "epoch": 1857, "lr": 3.330851853531383e-05}
{"train_loss": 0.12670168280601501, "global_step": 165330, "epoch": 1857, "lr": 3.330797205733435e-05}
{"train_loss": 0.08242671191692352, "global_step": 165331, "epoch": 1857, "lr": 3.330742558159892e-05}
{"train_loss": 0.06769386678934097, "global_step": 165332, "epoch": 1857, "lr": 3.3306879108107615e-05}
{"train_loss": 0.09253748506307602, "global_step": 165333, "epoch": 1857, "lr": 3.330633263686048e-05}
{"train_loss": 0.06454688310623169, "global_step": 165334, "epoch": 1857, "lr": 3.330578616785763e-05}
{"train_loss": 0.06816326081752777, "global_step": 165335, "epoch": 1857, "lr": 3.330523970109909e-05}
{"train_loss": 0.09423144906759262, "global_step": 165336, "epoch": 1857, "lr": 3.330469323658497e-05}
{"train_loss": 0.09794779866933823, "global_step": 165337, "epoch": 1857, "lr": 3.330414677431533e-05}
{"train_loss": 0.1417378932237625, "global_step": 165338, "epoch": 1857, "lr": 3.3303600314290244e-05}
{"train_loss": 0.07446736097335815, "global_step": 165339, "epoch": 1857, "lr": 3.3303053856509783e-05}
{"train_loss": 0.12427083402872086, "global_step": 165340, "epoch": 1857, "lr": 3.330250740097403e-05}
{"train_loss": 0.14261575043201447, "global_step": 165341, "epoch": 1857, "lr": 3.3301960947683045e-05}
{"train_loss": 0.15512293577194214, "global_step": 165342, "epoch": 1857, "lr": 3.33014144966369e-05}
{"train_loss": 0.05373574048280716, "global_step": 165343, "epoch": 1857, "lr": 3.330086804783569e-05}
{"train_loss": 0.10627484321594238, "global_step": 165344, "epoch": 1857, "lr": 3.330032160127946e-05}
{"train_loss": 0.11559896916151047, "global_step": 165345, "epoch": 1857, "lr": 3.3299775156968305e-05}
{"train_loss": 0.14401324093341827, "global_step": 165346, "epoch": 1857, "lr": 3.329922871490231e-05}
{"train_loss": 0.0969349592924118, "global_step": 165347, "epoch": 1857, "lr": 3.32986822750815e-05}
{"train_loss": 0.06523267179727554, "global_step": 165348, "epoch": 1857, "lr": 3.329813583750598e-05}
{"train_loss": 0.1311168372631073, "global_step": 165349, "epoch": 1857, "lr": 3.3297589402175846e-05}
{"train_loss": 0.09219349175691605, "global_step": 165350, "epoch": 1857, "lr": 3.3297042969091125e-05}
{"train_loss": 0.09322980046272278, "global_step": 165351, "epoch": 1857, "lr": 3.329649653825193e-05}
{"train_loss": 0.12558063864707947, "global_step": 165352, "epoch": 1857, "lr": 3.3295950109658306e-05}
{"train_loss": 0.11621425300836563, "global_step": 165353, "epoch": 1857, "lr": 3.329540368331035e-05}
{"train_loss": 0.057567473500967026, "global_step": 165354, "epoch": 1857, "lr": 3.329485725920811e-05}
{"train_loss": 0.12333564460277557, "global_step": 165355, "epoch": 1857, "lr": 3.3294310837351684e-05}
{"train_loss": 0.06360882520675659, "global_step": 165356, "epoch": 1857, "lr": 3.3293764417741116e-05}
{"train_loss": 0.057895589619874954, "global_step": 165357, "epoch": 1857, "lr": 3.329321800037653e-05}
{"train_loss": 0.0726439505815506, "global_step": 165358, "epoch": 1857, "lr": 3.329267158525793e-05}
{"train_loss": 0.060044147074222565, "global_step": 165359, "epoch": 1857, "lr": 3.3292125172385456e-05}
{"train_loss": 0.11187207698822021, "global_step": 165360, "epoch": 1857, "lr": 3.329157876175914e-05}
{"train_loss": 0.09269834722109725, "global_step": 165361, "epoch": 1857, "lr": 3.329103235337907e-05, "val_loss": 6.785762310028076}
{"train_loss": 0.036430198699235916, "global_step": 165362, "epoch": 1858, "lr": 3.329048594724532e-05}
{"train_loss": 0.05417517572641373, "global_step": 165363, "epoch": 1858, "lr": 3.328993954335795e-05}
{"train_loss": 0.0823962613940239, "global_step": 165364, "epoch": 1858, "lr": 3.328939314171705e-05}
{"train_loss": 0.07147235423326492, "global_step": 165365, "epoch": 1858, "lr": 3.32888467423227e-05}
{"train_loss": 0.06177353486418724, "global_step": 165366, "epoch": 1858, "lr": 3.3288300345174936e-05}
{"train_loss": 0.09084772318601608, "global_step": 165367, "epoch": 1858, "lr": 3.328775395027386e-05}
{"train_loss": 0.0723952054977417, "global_step": 165368, "epoch": 1858, "lr": 3.328720755761956e-05}
{"train_loss": 0.08594609051942825, "global_step": 165369, "epoch": 1858, "lr": 3.328666116721208e-05}
{"train_loss": 0.06182427331805229, "global_step": 165370, "epoch": 1858, "lr": 3.328611477905151e-05}
{"train_loss": 0.06215474009513855, "global_step": 165371, "epoch": 1858, "lr": 3.3285568393137906e-05}
{"train_loss": 0.12474256753921509, "global_step": 165372, "epoch": 1858, "lr": 3.328502200947137e-05}
{"train_loss": 0.03927236050367355, "global_step": 165373, "epoch": 1858, "lr": 3.328447562805194e-05}
{"train_loss": 0.08938605338335037, "global_step": 165374, "epoch": 1858, "lr": 3.328392924887973e-05}
{"train_loss": 0.057374950498342514, "global_step": 165375, "epoch": 1858, "lr": 3.328338287195477e-05}
{"train_loss": 0.057966213673353195, "global_step": 165376, "epoch": 1858, "lr": 3.328283649727718e-05}
{"train_loss": 0.11906486749649048, "global_step": 165377, "epoch": 1858, "lr": 3.328229012484699e-05}
{"train_loss": 0.11876139789819717, "global_step": 165378, "epoch": 1858, "lr": 3.328174375466431e-05}
{"train_loss": 0.06190618872642517, "global_step": 165379, "epoch": 1858, "lr": 3.3281197386729176e-05}
{"train_loss": 0.15074095129966736, "global_step": 165380, "epoch": 1858, "lr": 3.3280651021041695e-05}
{"train_loss": 0.11786681413650513, "global_step": 165381, "epoch": 1858, "lr": 3.328010465760192e-05}
{"train_loss": 0.08628244698047638, "global_step": 165382, "epoch": 1858, "lr": 3.327955829640993e-05}
{"train_loss": 0.09439637511968613, "global_step": 165383, "epoch": 1858, "lr": 3.3279011937465804e-05}
{"train_loss": 0.14006398618221283, "global_step": 165384, "epoch": 1858, "lr": 3.327846558076962e-05}
{"train_loss": 0.07903432101011276, "global_step": 165385, "epoch": 1858, "lr": 3.327791922632142e-05}
{"train_loss": 0.07943571358919144, "global_step": 165386, "epoch": 1858, "lr": 3.32773728741213e-05}
{"train_loss": 0.09669603407382965, "global_step": 165387, "epoch": 1858, "lr": 3.3276826524169355e-05}
{"train_loss": 0.05665065348148346, "global_step": 165388, "epoch": 1858, "lr": 3.327628017646562e-05}
{"train_loss": 0.08857504278421402, "global_step": 165389, "epoch": 1858, "lr": 3.32757338310102e-05}
{"train_loss": 0.12910376489162445, "global_step": 165390, "epoch": 1858, "lr": 3.327518748780314e-05}
{"train_loss": 0.0962386280298233, "global_step": 165391, "epoch": 1858, "lr": 3.327464114684454e-05}
{"train_loss": 0.08311918377876282, "global_step": 165392, "epoch": 1858, "lr": 3.327409480813444e-05}
{"train_loss": 0.07876589149236679, "global_step": 165393, "epoch": 1858, "lr": 3.327354847167296e-05}
{"train_loss": 0.09428320825099945, "global_step": 165394, "epoch": 1858, "lr": 3.327300213746012e-05}
{"train_loss": 0.06873106211423874, "global_step": 165395, "epoch": 1858, "lr": 3.327245580549605e-05}
{"train_loss": 0.08265615254640579, "global_step": 165396, "epoch": 1858, "lr": 3.327190947578077e-05}
{"train_loss": 0.08792988210916519, "global_step": 165397, "epoch": 1858, "lr": 3.32713631483144e-05}
{"train_loss": 0.09296878427267075, "global_step": 165398, "epoch": 1858, "lr": 3.327081682309698e-05}
{"train_loss": 0.10390301048755646, "global_step": 165399, "epoch": 1858, "lr": 3.3270270500128595e-05}
{"train_loss": 0.14528724551200867, "global_step": 165400, "epoch": 1858, "lr": 3.326972417940932e-05}
{"train_loss": 0.07523763924837112, "global_step": 165401, "epoch": 1858, "lr": 3.326917786093922e-05}
{"train_loss": 0.0823420062661171, "global_step": 165402, "epoch": 1858, "lr": 3.3268631544718384e-05}
{"train_loss": 0.09457617998123169, "global_step": 165403, "epoch": 1858, "lr": 3.326808523074686e-05}
{"train_loss": 0.06950818747282028, "global_step": 165404, "epoch": 1858, "lr": 3.3267538919024754e-05}
{"train_loss": 0.17375607788562775, "global_step": 165405, "epoch": 1858, "lr": 3.326699260955212e-05}
{"train_loss": 0.10630180686712265, "global_step": 165406, "epoch": 1858, "lr": 3.3266446302329033e-05}
{"train_loss": 0.07951132208108902, "global_step": 165407, "epoch": 1858, "lr": 3.3265899997355566e-05}
{"train_loss": 0.09043717384338379, "global_step": 165408, "epoch": 1858, "lr": 3.3265353694631804e-05}
{"train_loss": 0.09026317298412323, "global_step": 165409, "epoch": 1858, "lr": 3.3264807394157796e-05}
{"train_loss": 0.06760238856077194, "global_step": 165410, "epoch": 1858, "lr": 3.326426109593366e-05}
{"train_loss": 0.07630966603755951, "global_step": 165411, "epoch": 1858, "lr": 3.32637147999594e-05}
{"train_loss": 0.05982329323887825, "global_step": 165412, "epoch": 1858, "lr": 3.3263168506235166e-05}
{"train_loss": 0.11941207945346832, "global_step": 165413, "epoch": 1858, "lr": 3.3262622214760964e-05}
{"train_loss": 0.057833168655633926, "global_step": 165414, "epoch": 1858, "lr": 3.326207592553693e-05}
{"train_loss": 0.0417545884847641, "global_step": 165415, "epoch": 1858, "lr": 3.3261529638563075e-05}
{"train_loss": 0.1010327935218811, "global_step": 165416, "epoch": 1858, "lr": 3.326098335383953e-05}
{"train_loss": 0.10509096831083298, "global_step": 165417, "epoch": 1858, "lr": 3.3260437071366326e-05}
{"train_loss": 0.08605508506298065, "global_step": 165418, "epoch": 1858, "lr": 3.3259890791143554e-05}
{"train_loss": 0.13270483911037445, "global_step": 165419, "epoch": 1858, "lr": 3.3259344513171295e-05}
{"train_loss": 0.06993594765663147, "global_step": 165420, "epoch": 1858, "lr": 3.325879823744961e-05}
{"train_loss": 0.047189489006996155, "global_step": 165421, "epoch": 1858, "lr": 3.3258251963978564e-05}
{"train_loss": 0.05272296443581581, "global_step": 165422, "epoch": 1858, "lr": 3.325770569275824e-05}
{"train_loss": 0.0664597675204277, "global_step": 165423, "epoch": 1858, "lr": 3.325715942378873e-05}
{"train_loss": 0.09756095707416534, "global_step": 165424, "epoch": 1858, "lr": 3.325661315707007e-05}
{"train_loss": 0.07554943859577179, "global_step": 165425, "epoch": 1858, "lr": 3.325606689260238e-05}
{"train_loss": 0.11203930526971817, "global_step": 165426, "epoch": 1858, "lr": 3.3255520630385685e-05}
{"train_loss": 0.07271969318389893, "global_step": 165427, "epoch": 1858, "lr": 3.3254974370420106e-05}
{"train_loss": 0.0988064557313919, "global_step": 165428, "epoch": 1858, "lr": 3.325442811270566e-05}
{"train_loss": 0.10403060168027878, "global_step": 165429, "epoch": 1858, "lr": 3.3253881857242484e-05}
{"train_loss": 0.06006919965147972, "global_step": 165430, "epoch": 1858, "lr": 3.32533356040306e-05}
{"train_loss": 0.0621875636279583, "global_step": 165431, "epoch": 1858, "lr": 3.325278935307011e-05}
{"train_loss": 0.043428853154182434, "global_step": 165432, "epoch": 1858, "lr": 3.3252243104361063e-05}
{"train_loss": 0.09085120260715485, "global_step": 165433, "epoch": 1858, "lr": 3.3251696857903566e-05}
{"train_loss": 0.04978266730904579, "global_step": 165434, "epoch": 1858, "lr": 3.325115061369766e-05}
{"train_loss": 0.09671392291784286, "global_step": 165435, "epoch": 1858, "lr": 3.325060437174345e-05}
{"train_loss": 0.05775833502411842, "global_step": 165436, "epoch": 1858, "lr": 3.3250058132040985e-05}
{"train_loss": 0.07287291437387466, "global_step": 165437, "epoch": 1858, "lr": 3.3249511894590335e-05}
{"train_loss": 0.03715694695711136, "global_step": 165438, "epoch": 1858, "lr": 3.324896565939159e-05}
{"train_loss": 0.05740128830075264, "global_step": 165439, "epoch": 1858, "lr": 3.324841942644481e-05}
{"train_loss": 0.11111359298229218, "global_step": 165440, "epoch": 1858, "lr": 3.3247873195750104e-05}
{"train_loss": 0.0961209312081337, "global_step": 165441, "epoch": 1858, "lr": 3.324732696730748e-05}
{"train_loss": 0.10250285267829895, "global_step": 165442, "epoch": 1858, "lr": 3.324678074111708e-05}
{"train_loss": 0.09338008612394333, "global_step": 165443, "epoch": 1858, "lr": 3.324623451717892e-05}
{"train_loss": 0.06512387096881866, "global_step": 165444, "epoch": 1858, "lr": 3.324568829549312e-05}
{"train_loss": 0.10629163682460785, "global_step": 165445, "epoch": 1858, "lr": 3.324514207605972e-05}
{"train_loss": 0.07706296443939209, "global_step": 165446, "epoch": 1858, "lr": 3.324459585887882e-05}
{"train_loss": 0.14486216008663177, "global_step": 165447, "epoch": 1858, "lr": 3.324404964395046e-05}
{"train_loss": 0.0678263008594513, "global_step": 165448, "epoch": 1858, "lr": 3.3243503431274756e-05}
{"train_loss": 0.1008404791355133, "global_step": 165449, "epoch": 1858, "lr": 3.3242957220851734e-05}
{"train_loss": 0.08545018644647652, "global_step": 165450, "epoch": 1858, "lr": 3.3242411012681515e-05, "val_loss": 7.046607971191406}
{"train_loss": 0.11593671143054962, "global_step": 165451, "epoch": 1859, "lr": 3.324186480676413e-05}
{"train_loss": 0.06254970282316208, "global_step": 165452, "epoch": 1859, "lr": 3.324131860309969e-05}
{"train_loss": 0.09475255757570267, "global_step": 165453, "epoch": 1859, "lr": 3.324077240168824e-05}
{"train_loss": 0.0774741917848587, "global_step": 165454, "epoch": 1859, "lr": 3.324022620252987e-05}
{"train_loss": 0.08882554620504379, "global_step": 165455, "epoch": 1859, "lr": 3.323968000562465e-05}
{"train_loss": 0.08113480359315872, "global_step": 165456, "epoch": 1859, "lr": 3.323913381097264e-05}
{"train_loss": 0.05780179053544998, "global_step": 165457, "epoch": 1859, "lr": 3.323858761857393e-05}
{"train_loss": 0.07041548937559128, "global_step": 165458, "epoch": 1859, "lr": 3.323804142842858e-05}
{"train_loss": 0.08457355946302414, "global_step": 165459, "epoch": 1859, "lr": 3.3237495240536684e-05}
{"train_loss": 0.024111859500408173, "global_step": 165460, "epoch": 1859, "lr": 3.3236949054898287e-05}
{"train_loss": 0.09361256659030914, "global_step": 165461, "epoch": 1859, "lr": 3.323640287151349e-05}
{"train_loss": 0.09885667264461517, "global_step": 165462, "epoch": 1859, "lr": 3.323585669038235e-05}
{"train_loss": 0.06520774215459824, "global_step": 165463, "epoch": 1859, "lr": 3.323531051150495e-05}
{"train_loss": 0.09651242941617966, "global_step": 165464, "epoch": 1859, "lr": 3.3234764334881344e-05}
{"train_loss": 0.09566630423069, "global_step": 165465, "epoch": 1859, "lr": 3.323421816051164e-05}
{"train_loss": 0.06829856336116791, "global_step": 165466, "epoch": 1859, "lr": 3.3233671988395876e-05}
{"train_loss": 0.07072308659553528, "global_step": 165467, "epoch": 1859, "lr": 3.323312581853415e-05}
{"train_loss": 0.05930720642209053, "global_step": 165468, "epoch": 1859, "lr": 3.323257965092651e-05}
{"train_loss": 0.12192083150148392, "global_step": 165469, "epoch": 1859, "lr": 3.323203348557307e-05}
{"train_loss": 0.10676036030054092, "global_step": 165470, "epoch": 1859, "lr": 3.323148732247385e-05}
{"train_loss": 0.08033422380685806, "global_step": 165471, "epoch": 1859, "lr": 3.323094116162898e-05}
{"train_loss": 0.06262429058551788, "global_step": 165472, "epoch": 1859, "lr": 3.323039500303848e-05}
{"train_loss": 0.10450094938278198, "global_step": 165473, "epoch": 1859, "lr": 3.322984884670247e-05}
{"train_loss": 0.07446856796741486, "global_step": 165474, "epoch": 1859, "lr": 3.322930269262099e-05}
{"train_loss": 0.0897202119231224, "global_step": 165475, "epoch": 1859, "lr": 3.3228756540794124e-05}
{"train_loss": 0.05601085349917412, "global_step": 165476, "epoch": 1859, "lr": 3.322821039122197e-05}
{"train_loss": 0.07203359156847, "global_step": 165477, "epoch": 1859, "lr": 3.322766424390455e-05}
{"train_loss": 0.11667381227016449, "global_step": 165478, "epoch": 1859, "lr": 3.322711809884199e-05}
{"train_loss": 0.072642482817173, "global_step": 165479, "epoch": 1859, "lr": 3.3226571956034316e-05}
{"train_loss": 0.07825785130262375, "global_step": 165480, "epoch": 1859, "lr": 3.3226025815481646e-05}
{"train_loss": 0.12545464932918549, "global_step": 165481, "epoch": 1859, "lr": 3.3225479677184015e-05}
{"train_loss": 0.10578589886426926, "global_step": 165482, "epoch": 1859, "lr": 3.322493354114153e-05}
{"train_loss": 0.0581163614988327, "global_step": 165483, "epoch": 1859, "lr": 3.322438740735423e-05}
{"train_loss": 0.04861319065093994, "global_step": 165484, "epoch": 1859, "lr": 3.322384127582223e-05}
{"train_loss": 0.08149027079343796, "global_step": 165485, "epoch": 1859, "lr": 3.3223295146545554e-05}
{"train_loss": 0.08167194575071335, "global_step": 165486, "epoch": 1859, "lr": 3.322274901952432e-05}
{"train_loss": 0.09863708913326263, "global_step": 165487, "epoch": 1859, "lr": 3.322220289475857e-05}
{"train_loss": 0.04388117417693138, "global_step": 165488, "epoch": 1859, "lr": 3.32216567722484e-05}
{"train_loss": 0.07810370624065399, "global_step": 165489, "epoch": 1859, "lr": 3.322111065199387e-05}
{"train_loss": 0.041975490748882294, "global_step": 165490, "epoch": 1859, "lr": 3.322056453399507e-05}
{"train_loss": 0.06578999012708664, "global_step": 165491, "epoch": 1859, "lr": 3.322001841825203e-05}
{"train_loss": 0.08412855863571167, "global_step": 165492, "epoch": 1859, "lr": 3.3219472304764887e-05}
{"train_loss": 0.107027068734169, "global_step": 165493, "epoch": 1859, "lr": 3.321892619353366e-05}
{"train_loss": 0.12653112411499023, "global_step": 165494, "epoch": 1859, "lr": 3.321838008455845e-05}
{"train_loss": 0.048216838389635086, "global_step": 165495, "epoch": 1859, "lr": 3.321783397783933e-05}
{"train_loss": 0.12974584102630615, "global_step": 165496, "epoch": 1859, "lr": 3.321728787337636e-05}
{"train_loss": 0.07199755311012268, "global_step": 165497, "epoch": 1859, "lr": 3.3216741771169636e-05}
{"train_loss": 0.08044028282165527, "global_step": 165498, "epoch": 1859, "lr": 3.3216195671219186e-05}
{"train_loss": 0.09948823601007462, "global_step": 165499, "epoch": 1859, "lr": 3.321564957352514e-05}
{"train_loss": 0.08483073115348816, "global_step": 165500, "epoch": 1859, "lr": 3.3215103478087535e-05}
{"train_loss": 0.03306939825415611, "global_step": 165501, "epoch": 1859, "lr": 3.321455738490646e-05}
{"train_loss": 0.10727079212665558, "global_step": 165502, "epoch": 1859, "lr": 3.3214011293981976e-05}
{"train_loss": 0.04771075397729874, "global_step": 165503, "epoch": 1859, "lr": 3.3213465205314173e-05}
{"train_loss": 0.07339607924222946, "global_step": 165504, "epoch": 1859, "lr": 3.3212919118903104e-05}
{"train_loss": 0.06682740896940231, "global_step": 165505, "epoch": 1859, "lr": 3.321237303474887e-05}
{"train_loss": 0.1113440990447998, "global_step": 165506, "epoch": 1859, "lr": 3.321182695285151e-05}
{"train_loss": 0.06419462710618973, "global_step": 165507, "epoch": 1859, "lr": 3.3211280873211134e-05}
{"train_loss": 0.0700574591755867, "global_step": 165508, "epoch": 1859, "lr": 3.3210734795827776e-05}
{"train_loss": 0.09022893756628036, "global_step": 165509, "epoch": 1859, "lr": 3.321018872070155e-05}
{"train_loss": 0.09697600454092026, "global_step": 165510, "epoch": 1859, "lr": 3.3209642647832485e-05}
{"train_loss": 0.08336472511291504, "global_step": 165511, "epoch": 1859, "lr": 3.320909657722071e-05}
{"train_loss": 0.15458402037620544, "global_step": 165512, "epoch": 1859, "lr": 3.320855050886623e-05}
{"train_loss": 0.04335184395313263, "global_step": 165513, "epoch": 1859, "lr": 3.320800444276918e-05}
{"train_loss": 0.15833546221256256, "global_step": 165514, "epoch": 1859, "lr": 3.320745837892962e-05}
{"train_loss": 0.04965733736753464, "global_step": 165515, "epoch": 1859, "lr": 3.3206912317347584e-05}
{"train_loss": 0.08367056399583817, "global_step": 165516, "epoch": 1859, "lr": 3.3206366258023205e-05}
{"train_loss": 0.05619506165385246, "global_step": 165517, "epoch": 1859, "lr": 3.3205820200956494e-05}
{"train_loss": 0.10942565649747849, "global_step": 165518, "epoch": 1859, "lr": 3.320527414614758e-05}
{"train_loss": 0.09185333549976349, "global_step": 165519, "epoch": 1859, "lr": 3.320472809359649e-05}
{"train_loss": 0.11094782501459122, "global_step": 165520, "epoch": 1859, "lr": 3.320418204330335e-05}
{"train_loss": 0.1082928255200386, "global_step": 165521, "epoch": 1859, "lr": 3.320363599526817e-05}
{"train_loss": 0.037021979689598083, "global_step": 165522, "epoch": 1859, "lr": 3.320308994949107e-05}
{"train_loss": 0.08246518671512604, "global_step": 165523, "epoch": 1859, "lr": 3.3202543905972105e-05}
{"train_loss": 0.08401249349117279, "global_step": 165524, "epoch": 1859, "lr": 3.3201997864711375e-05}
{"train_loss": 0.1281673163175583, "global_step": 165525, "epoch": 1859, "lr": 3.32014518257089e-05}
{"train_loss": 0.10870904475450516, "global_step": 165526, "epoch": 1859, "lr": 3.3200905788964817e-05}
{"train_loss": 0.14524222910404205, "global_step": 165527, "epoch": 1859, "lr": 3.320035975447914e-05}
{"train_loss": 0.04484326392412186, "global_step": 165528, "epoch": 1859, "lr": 3.319981372225199e-05}
{"train_loss": 0.13626912236213684, "global_step": 165529, "epoch": 1859, "lr": 3.31992676922834e-05}
{"train_loss": 0.06433278322219849, "global_step": 165530, "epoch": 1859, "lr": 3.3198721664573474e-05}
{"train_loss": 0.0912027582526207, "global_step": 165531, "epoch": 1859, "lr": 3.3198175639122275e-05}
{"train_loss": 0.07086606323719025, "global_step": 165532, "epoch": 1859, "lr": 3.319762961592988e-05}
{"train_loss": 0.11819734424352646, "global_step": 165533, "epoch": 1859, "lr": 3.319708359499637e-05}
{"train_loss": 0.048750557005405426, "global_step": 165534, "epoch": 1859, "lr": 3.319653757632179e-05}
{"train_loss": 0.08863271772861481, "global_step": 165535, "epoch": 1859, "lr": 3.3195991559906234e-05}
{"train_loss": 0.04078548774123192, "global_step": 165536, "epoch": 1859, "lr": 3.3195445545749766e-05}
{"train_loss": 0.11033753305673599, "global_step": 165537, "epoch": 1859, "lr": 3.319489953385248e-05}
{"train_loss": 0.10603532195091248, "global_step": 165538, "epoch": 1859, "lr": 3.3194353524214413e-05}
{"train_loss": 0.08436062090684858, "global_step": 165539, "epoch": 1859, "lr": 3.319380751683569e-05, "val_loss": 7.033646583557129}
{"train_loss": 0.06602077186107635, "global_step": 165540, "epoch": 1860, "lr": 3.3193261511716327e-05}
{"train_loss": 0.04394908994436264, "global_step": 165541, "epoch": 1860, "lr": 3.319271550885645e-05}
{"train_loss": 0.09609684348106384, "global_step": 165542, "epoch": 1860, "lr": 3.319216950825609e-05}
{"train_loss": 0.09466613084077835, "global_step": 165543, "epoch": 1860, "lr": 3.319162350991535e-05}
{"train_loss": 0.07203347980976105, "global_step": 165544, "epoch": 1860, "lr": 3.319107751383427e-05}
{"train_loss": 0.10122253000736237, "global_step": 165545, "epoch": 1860, "lr": 3.3190531520012966e-05}
{"train_loss": 0.05538539960980415, "global_step": 165546, "epoch": 1860, "lr": 3.318998552845148e-05}
{"train_loss": 0.061976149678230286, "global_step": 165547, "epoch": 1860, "lr": 3.3189439539149906e-05}
{"train_loss": 0.08684799075126648, "global_step": 165548, "epoch": 1860, "lr": 3.318889355210829e-05}
{"train_loss": 0.11730962991714478, "global_step": 165549, "epoch": 1860, "lr": 3.3188347567326734e-05}
{"train_loss": 0.09566725790500641, "global_step": 165550, "epoch": 1860, "lr": 3.31878015848053e-05}
{"train_loss": 0.03686534985899925, "global_step": 165551, "epoch": 1860, "lr": 3.3187255604544054e-05}
{"train_loss": 0.11224967241287231, "global_step": 165552, "epoch": 1860, "lr": 3.318670962654309e-05}
{"train_loss": 0.08813517540693283, "global_step": 165553, "epoch": 1860, "lr": 3.3186163650802454e-05}
{"train_loss": 0.07369321584701538, "global_step": 165554, "epoch": 1860, "lr": 3.318561767732224e-05}
{"train_loss": 0.08356496691703796, "global_step": 165555, "epoch": 1860, "lr": 3.318507170610251e-05}
{"train_loss": 0.1181531548500061, "global_step": 165556, "epoch": 1860, "lr": 3.318452573714336e-05}
{"train_loss": 0.054285526275634766, "global_step": 165557, "epoch": 1860, "lr": 3.318397977044482e-05}
{"train_loss": 0.127362459897995, "global_step": 165558, "epoch": 1860, "lr": 3.318343380600702e-05}
{"train_loss": 0.05921795964241028, "global_step": 165559, "epoch": 1860, "lr": 3.3182887843829965e-05}
{"train_loss": 0.07390705496072769, "global_step": 165560, "epoch": 1860, "lr": 3.318234188391379e-05}
{"train_loss": 0.041847113519907, "global_step": 165561, "epoch": 1860, "lr": 3.318179592625853e-05}
{"train_loss": 0.04893156886100769, "global_step": 165562, "epoch": 1860, "lr": 3.318124997086429e-05}
{"train_loss": 0.0932096540927887, "global_step": 165563, "epoch": 1860, "lr": 3.31807040177311e-05}
{"train_loss": 0.17983010411262512, "global_step": 165564, "epoch": 1860, "lr": 3.318015806685909e-05}
{"train_loss": 0.08327533304691315, "global_step": 165565, "epoch": 1860, "lr": 3.317961211824828e-05}
{"train_loss": 0.062404219061136246, "global_step": 165566, "epoch": 1860, "lr": 3.317906617189877e-05}
{"train_loss": 0.08101458102464676, "global_step": 165567, "epoch": 1860, "lr": 3.317852022781064e-05}
{"train_loss": 0.07496065646409988, "global_step": 165568, "epoch": 1860, "lr": 3.3177974285983946e-05}
{"train_loss": 0.11581309884786606, "global_step": 165569, "epoch": 1860, "lr": 3.317742834641877e-05}
{"train_loss": 0.062288038432598114, "global_step": 165570, "epoch": 1860, "lr": 3.317688240911519e-05}
{"train_loss": 0.06966736167669296, "global_step": 165571, "epoch": 1860, "lr": 3.3176336474073256e-05}
{"train_loss": 0.1073049008846283, "global_step": 165572, "epoch": 1860, "lr": 3.317579054129307e-05}
{"train_loss": 0.06420744955539703, "global_step": 165573, "epoch": 1860, "lr": 3.31752446107747e-05}
{"train_loss": 0.09584641456604004, "global_step": 165574, "epoch": 1860, "lr": 3.31746986825182e-05}
{"train_loss": 0.08537483960390091, "global_step": 165575, "epoch": 1860, "lr": 3.317415275652367e-05}
{"train_loss": 0.0720396414399147, "global_step": 165576, "epoch": 1860, "lr": 3.3173606832791145e-05}
{"train_loss": 0.09294585883617401, "global_step": 165577, "epoch": 1860, "lr": 3.3173060911320756e-05}
{"train_loss": 0.14391325414180756, "global_step": 165578, "epoch": 1860, "lr": 3.3172514992112526e-05}
{"train_loss": 0.04826747253537178, "global_step": 165579, "epoch": 1860, "lr": 3.317196907516655e-05}
{"train_loss": 0.07732365280389786, "global_step": 165580, "epoch": 1860, "lr": 3.317142316048289e-05}
{"train_loss": 0.08046940714120865, "global_step": 165581, "epoch": 1860, "lr": 3.317087724806165e-05}
{"train_loss": 0.09600189328193665, "global_step": 165582, "epoch": 1860, "lr": 3.317033133790285e-05}
{"train_loss": 0.11761796474456787, "global_step": 165583, "epoch": 1860, "lr": 3.316978543000662e-05}
{"train_loss": 0.1292831003665924, "global_step": 165584, "epoch": 1860, "lr": 3.3169239524372986e-05}
{"train_loss": 0.09406483173370361, "global_step": 165585, "epoch": 1860, "lr": 3.3168693621002036e-05}
{"train_loss": 0.04495944082736969, "global_step": 165586, "epoch": 1860, "lr": 3.316814771989388e-05}
{"train_loss": 0.05676264315843582, "global_step": 165587, "epoch": 1860, "lr": 3.3167601821048545e-05}
{"train_loss": 0.09834787994623184, "global_step": 165588, "epoch": 1860, "lr": 3.3167055924466124e-05}
{"train_loss": 0.046343881636857986, "global_step": 165589, "epoch": 1860, "lr": 3.3166510030146684e-05}
{"train_loss": 0.13185837864875793, "global_step": 165590, "epoch": 1860, "lr": 3.3165964138090306e-05}
{"train_loss": 0.07204703986644745, "global_step": 165591, "epoch": 1860, "lr": 3.316541824829705e-05}
{"train_loss": 0.09992194175720215, "global_step": 165592, "epoch": 1860, "lr": 3.316487236076702e-05}
{"train_loss": 0.12350426614284515, "global_step": 165593, "epoch": 1860, "lr": 3.316432647550024e-05}
{"train_loss": 0.05413222312927246, "global_step": 165594, "epoch": 1860, "lr": 3.316378059249684e-05}
{"train_loss": 0.07350121438503265, "global_step": 165595, "epoch": 1860, "lr": 3.316323471175683e-05}
{"train_loss": 0.05578242987394333, "global_step": 165596, "epoch": 1860, "lr": 3.316268883328035e-05}
{"train_loss": 0.07198373228311539, "global_step": 165597, "epoch": 1860, "lr": 3.316214295706742e-05}
{"train_loss": 0.09248659759759903, "global_step": 165598, "epoch": 1860, "lr": 3.3161597083118153e-05}
{"train_loss": 0.07670161873102188, "global_step": 165599, "epoch": 1860, "lr": 3.316105121143258e-05}
{"train_loss": 0.1733514666557312, "global_step": 165600, "epoch": 1860, "lr": 3.3160505342010826e-05}
{"train_loss": 0.07552149146795273, "global_step": 165601, "epoch": 1860, "lr": 3.3159959474852907e-05}
{"train_loss": 0.10863806307315826, "global_step": 165602, "epoch": 1860, "lr": 3.315941360995893e-05}
{"train_loss": 0.06675659120082855, "global_step": 165603, "epoch": 1860, "lr": 3.315886774732899e-05}
{"train_loss": 0.09240435808897018, "global_step": 165604, "epoch": 1860, "lr": 3.315832188696311e-05}
{"train_loss": 0.10179831087589264, "global_step": 165605, "epoch": 1860, "lr": 3.315777602886141e-05}
{"train_loss": 0.14886707067489624, "global_step": 165606, "epoch": 1860, "lr": 3.315723017302393e-05}
{"train_loss": 0.09167887270450592, "global_step": 165607, "epoch": 1860, "lr": 3.315668431945076e-05}
{"train_loss": 0.13113032281398773, "global_step": 165608, "epoch": 1860, "lr": 3.315613846814196e-05}
{"train_loss": 0.1425698846578598, "global_step": 165609, "epoch": 1860, "lr": 3.315559261909762e-05}
{"train_loss": 0.10394202172756195, "global_step": 165610, "epoch": 1860, "lr": 3.315504677231781e-05}
{"train_loss": 0.0827106386423111, "global_step": 165611, "epoch": 1860, "lr": 3.3154500927802596e-05}
{"train_loss": 0.04387384653091431, "global_step": 165612, "epoch": 1860, "lr": 3.3153955085552034e-05}
{"train_loss": 0.07923838496208191, "global_step": 165613, "epoch": 1860, "lr": 3.315340924556625e-05}
{"train_loss": 0.16640634834766388, "global_step": 165614, "epoch": 1860, "lr": 3.315286340784525e-05}
{"train_loss": 0.06402862817049026, "global_step": 165615, "epoch": 1860, "lr": 3.3152317572389166e-05}
{"train_loss": 0.09412588179111481, "global_step": 165616, "epoch": 1860, "lr": 3.3151771739198034e-05}
{"train_loss": 0.06951938569545746, "global_step": 165617, "epoch": 1860, "lr": 3.315122590827196e-05}
{"train_loss": 0.10523071140050888, "global_step": 165618, "epoch": 1860, "lr": 3.315068007961097e-05}
{"train_loss": 0.09692901372909546, "global_step": 165619, "epoch": 1860, "lr": 3.31501342532152e-05}
{"train_loss": 0.04022838547825813, "global_step": 165620, "epoch": 1860, "lr": 3.314958842908466e-05}
{"train_loss": 0.08708323538303375, "global_step": 165621, "epoch": 1860, "lr": 3.3149042607219447e-05}
{"train_loss": 0.05313890427350998, "global_step": 165622, "epoch": 1860, "lr": 3.314849678761967e-05}
{"train_loss": 0.06216726824641228, "global_step": 165623, "epoch": 1860, "lr": 3.314795097028535e-05}
{"train_loss": 0.08968359231948853, "global_step": 165624, "epoch": 1860, "lr": 3.3147405155216606e-05}
{"train_loss": 0.14946512877941132, "global_step": 165625, "epoch": 1860, "lr": 3.314685934241347e-05}
{"train_loss": 0.047070086002349854, "global_step": 165626, "epoch": 1860, "lr": 3.314631353187603e-05}
{"train_loss": 0.16670134663581848, "global_step": 165627, "epoch": 1860, "lr": 3.314576772360437e-05}
{"train_loss": 0.08917228341772315, "global_step": 165628, "epoch": 1860, "lr": 3.314522191759856e-05, "val_loss": 6.8835129737854, "train_action_mse_error": 19.413738250732422}
{"train_loss": 0.07834034413099289, "global_step": 165629, "epoch": 1861, "lr": 3.3144676113858666e-05}
{"train_loss": 0.08934799581766129, "global_step": 165630, "epoch": 1861, "lr": 3.314413031238477e-05}
{"train_loss": 0.08526092022657394, "global_step": 165631, "epoch": 1861, "lr": 3.3143584513176926e-05}
{"train_loss": 0.07888586074113846, "global_step": 165632, "epoch": 1861, "lr": 3.314303871623524e-05}
{"train_loss": 0.126428484916687, "global_step": 165633, "epoch": 1861, "lr": 3.3142492921559746e-05}
{"train_loss": 0.037038654088974, "global_step": 165634, "epoch": 1861, "lr": 3.314194712915055e-05}
{"train_loss": 0.08229506015777588, "global_step": 165635, "epoch": 1861, "lr": 3.31414013390077e-05}
{"train_loss": 0.11576122045516968, "global_step": 165636, "epoch": 1861, "lr": 3.314085555113131e-05}
{"train_loss": 0.11081444472074509, "global_step": 165637, "epoch": 1861, "lr": 3.3140309765521405e-05}
{"train_loss": 0.06320361793041229, "global_step": 165638, "epoch": 1861, "lr": 3.3139763982178094e-05}
{"train_loss": 0.10587329417467117, "global_step": 165639, "epoch": 1861, "lr": 3.3139218201101416e-05}
{"train_loss": 0.09782467782497406, "global_step": 165640, "epoch": 1861, "lr": 3.313867242229147e-05}
{"train_loss": 0.07394075393676758, "global_step": 165641, "epoch": 1861, "lr": 3.3138126645748346e-05}
{"train_loss": 0.1330432891845703, "global_step": 165642, "epoch": 1861, "lr": 3.313758087147207e-05}
{"train_loss": 0.1221412643790245, "global_step": 165643, "epoch": 1861, "lr": 3.313703509946276e-05}
{"train_loss": 0.061526767909526825, "global_step": 165644, "epoch": 1861, "lr": 3.313648932972046e-05}
{"train_loss": 0.09126757830381393, "global_step": 165645, "epoch": 1861, "lr": 3.313594356224526e-05}
{"train_loss": 0.06606212258338928, "global_step": 165646, "epoch": 1861, "lr": 3.313539779703722e-05}
{"train_loss": 0.09420007467269897, "global_step": 165647, "epoch": 1861, "lr": 3.313485203409642e-05}
{"train_loss": 0.08407092094421387, "global_step": 165648, "epoch": 1861, "lr": 3.3134306273422935e-05}
{"train_loss": 0.07947038114070892, "global_step": 165649, "epoch": 1861, "lr": 3.313376051501686e-05}
{"train_loss": 0.09866105020046234, "global_step": 165650, "epoch": 1861, "lr": 3.3133214758878205e-05}
{"train_loss": 0.07422386109828949, "global_step": 165651, "epoch": 1861, "lr": 3.3132669005007114e-05}
{"train_loss": 0.11624214053153992, "global_step": 165652, "epoch": 1861, "lr": 3.3132123253403605e-05}
{"train_loss": 0.11114925146102905, "global_step": 165653, "epoch": 1861, "lr": 3.31315775040678e-05}
{"train_loss": 0.12813706696033478, "global_step": 165654, "epoch": 1861, "lr": 3.313103175699973e-05}
{"train_loss": 0.08126915246248245, "global_step": 165655, "epoch": 1861, "lr": 3.31304860121995e-05}
{"train_loss": 0.07830359041690826, "global_step": 165656, "epoch": 1861, "lr": 3.312994026966716e-05}
{"train_loss": 0.07013735920190811, "global_step": 165657, "epoch": 1861, "lr": 3.3129394529402794e-05}
{"train_loss": 0.09007574617862701, "global_step": 165658, "epoch": 1861, "lr": 3.312884879140648e-05}
{"train_loss": 0.09838952869176865, "global_step": 165659, "epoch": 1861, "lr": 3.312830305567828e-05}
{"train_loss": 0.10918230563402176, "global_step": 165660, "epoch": 1861, "lr": 3.312775732221829e-05}
{"train_loss": 0.10722015798091888, "global_step": 165661, "epoch": 1861, "lr": 3.312721159102655e-05}
{"train_loss": 0.12539005279541016, "global_step": 165662, "epoch": 1861, "lr": 3.3126665862103175e-05}
{"train_loss": 0.09873031079769135, "global_step": 165663, "epoch": 1861, "lr": 3.3126120135448194e-05}
{"train_loss": 0.0817183330655098, "global_step": 165664, "epoch": 1861, "lr": 3.312557441106171e-05}
{"train_loss": 0.09602916240692139, "global_step": 165665, "epoch": 1861, "lr": 3.3125028688943784e-05}
{"train_loss": 0.11391128599643707, "global_step": 165666, "epoch": 1861, "lr": 3.312448296909449e-05}
{"train_loss": 0.06217894330620766, "global_step": 165667, "epoch": 1861, "lr": 3.31239372515139e-05}
{"train_loss": 0.08427327871322632, "global_step": 165668, "epoch": 1861, "lr": 3.312339153620211e-05}
{"train_loss": 0.08572166413068771, "global_step": 165669, "epoch": 1861, "lr": 3.3122845823159145e-05}
{"train_loss": 0.1337674856185913, "global_step": 165670, "epoch": 1861, "lr": 3.312230011238514e-05}
{"train_loss": 0.12210052460432053, "global_step": 165671, "epoch": 1861, "lr": 3.31217544038801e-05}
{"train_loss": 0.1376447230577469, "global_step": 165672, "epoch": 1861, "lr": 3.312120869764417e-05}
{"train_loss": 0.10201796144247055, "global_step": 165673, "epoch": 1861, "lr": 3.3120662993677355e-05}
{"train_loss": 0.0794457271695137, "global_step": 165674, "epoch": 1861, "lr": 3.3120117291979794e-05}
{"train_loss": 0.0688655748963356, "global_step": 165675, "epoch": 1861, "lr": 3.311957159255149e-05}
{"train_loss": 0.08450274169445038, "global_step": 165676, "epoch": 1861, "lr": 3.3119025895392566e-05}
{"train_loss": 0.08541855216026306, "global_step": 165677, "epoch": 1861, "lr": 3.31184802005031e-05}
{"train_loss": 0.06642754375934601, "global_step": 165678, "epoch": 1861, "lr": 3.3117934507883134e-05}
{"train_loss": 0.09310215711593628, "global_step": 165679, "epoch": 1861, "lr": 3.311738881753277e-05}
{"train_loss": 0.1182517260313034, "global_step": 165680, "epoch": 1861, "lr": 3.3116843129452044e-05}
{"train_loss": 0.09761595726013184, "global_step": 165681, "epoch": 1861, "lr": 3.311629744364108e-05}
{"train_loss": 0.11079853773117065, "global_step": 165682, "epoch": 1861, "lr": 3.3115751760099914e-05}
{"train_loss": 0.08548548817634583, "global_step": 165683, "epoch": 1861, "lr": 3.311520607882863e-05}
{"train_loss": 0.05542527884244919, "global_step": 165684, "epoch": 1861, "lr": 3.3114660399827294e-05}
{"train_loss": 0.08128397911787033, "global_step": 165685, "epoch": 1861, "lr": 3.311411472309599e-05}
{"train_loss": 0.09614137560129166, "global_step": 165686, "epoch": 1861, "lr": 3.3113569048634785e-05}
{"train_loss": 0.14240506291389465, "global_step": 165687, "epoch": 1861, "lr": 3.311302337644376e-05}
{"train_loss": 0.11291010677814484, "global_step": 165688, "epoch": 1861, "lr": 3.311247770652298e-05}
{"train_loss": 0.08553176373243332, "global_step": 165689, "epoch": 1861, "lr": 3.311193203887253e-05}
{"train_loss": 0.11076737940311432, "global_step": 165690, "epoch": 1861, "lr": 3.311138637349246e-05}
{"train_loss": 0.09921377152204514, "global_step": 165691, "epoch": 1861, "lr": 3.3110840710382874e-05}
{"train_loss": 0.0764218345284462, "global_step": 165692, "epoch": 1861, "lr": 3.3110295049543804e-05}
{"train_loss": 0.10352519154548645, "global_step": 165693, "epoch": 1861, "lr": 3.310974939097536e-05}
{"train_loss": 0.14537972211837769, "global_step": 165694, "epoch": 1861, "lr": 3.310920373467762e-05}
{"train_loss": 0.06946072727441788, "global_step": 165695, "epoch": 1861, "lr": 3.3108658080650624e-05}
{"train_loss": 0.13261006772518158, "global_step": 165696, "epoch": 1861, "lr": 3.310811242889448e-05}
{"train_loss": 0.05611757934093475, "global_step": 165697, "epoch": 1861, "lr": 3.310756677940922e-05}
{"train_loss": 0.09824224561452866, "global_step": 165698, "epoch": 1861, "lr": 3.310702113219497e-05}
{"train_loss": 0.08102073520421982, "global_step": 165699, "epoch": 1861, "lr": 3.3106475487251756e-05}
{"train_loss": 0.11282635480165482, "global_step": 165700, "epoch": 1861, "lr": 3.310592984457968e-05}
{"train_loss": 0.06603650003671646, "global_step": 165701, "epoch": 1861, "lr": 3.310538420417879e-05}
{"train_loss": 0.06237303465604782, "global_step": 165702, "epoch": 1861, "lr": 3.31048385660492e-05}
{"train_loss": 0.07198868691921234, "global_step": 165703, "epoch": 1861, "lr": 3.310429293019094e-05}
{"train_loss": 0.12301509827375412, "global_step": 165704, "epoch": 1861, "lr": 3.310374729660411e-05}
{"train_loss": 0.05221238359808922, "global_step": 165705, "epoch": 1861, "lr": 3.3103201665288775e-05}
{"train_loss": 0.1411193460226059, "global_step": 165706, "epoch": 1861, "lr": 3.3102656036245004e-05}
{"train_loss": 0.09071790426969528, "global_step": 165707, "epoch": 1861, "lr": 3.310211040947288e-05}
{"train_loss": 0.05721340328454971, "global_step": 165708, "epoch": 1861, "lr": 3.3101564784972486e-05}
{"train_loss": 0.12054023891687393, "global_step": 165709, "epoch": 1861, "lr": 3.3101019162743854e-05}
{"train_loss": 0.10466795414686203, "global_step": 165710, "epoch": 1861, "lr": 3.31004735427871e-05}
{"train_loss": 0.04710187017917633, "global_step": 165711, "epoch": 1861, "lr": 3.309992792510226e-05}
{"train_loss": 0.07812202721834183, "global_step": 165712, "epoch": 1861, "lr": 3.3099382309689436e-05}
{"train_loss": 0.11620155721902847, "global_step": 165713, "epoch": 1861, "lr": 3.309883669654871e-05}
{"train_loss": 0.062293704599142075, "global_step": 165714, "epoch": 1861, "lr": 3.3098291085680124e-05}
{"train_loss": 0.08366075903177261, "global_step": 165715, "epoch": 1861, "lr": 3.309774547708378e-05}
{"train_loss": 0.0705728828907013, "global_step": 165716, "epoch": 1861, "lr": 3.309719987075972e-05}
{"train_loss": 0.09274752057167922, "global_step": 165717, "epoch": 1861, "lr": 3.309665426670806e-05, "val_loss": 7.1014227867126465}
{"train_loss": 0.1446973979473114, "global_step": 165718, "epoch": 1862, "lr": 3.3096108664928824e-05}
{"train_loss": 0.10278284549713135, "global_step": 165719, "epoch": 1862, "lr": 3.3095563065422136e-05}
{"train_loss": 0.06726349890232086, "global_step": 165720, "epoch": 1862, "lr": 3.3095017468188016e-05}
{"train_loss": 0.10597643256187439, "global_step": 165721, "epoch": 1862, "lr": 3.3094471873226594e-05}
{"train_loss": 0.07863728702068329, "global_step": 165722, "epoch": 1862, "lr": 3.30939262805379e-05}
{"train_loss": 0.11733675748109818, "global_step": 165723, "epoch": 1862, "lr": 3.3093380690122014e-05}
{"train_loss": 0.10109900683164597, "global_step": 165724, "epoch": 1862, "lr": 3.3092835101979025e-05}
{"train_loss": 0.11942828446626663, "global_step": 165725, "epoch": 1862, "lr": 3.3092289516109e-05}
{"train_loss": 0.17858591675758362, "global_step": 165726, "epoch": 1862, "lr": 3.309174393251201e-05}
{"train_loss": 0.12354255467653275, "global_step": 165727, "epoch": 1862, "lr": 3.3091198351188134e-05}
{"train_loss": 0.03997262194752693, "global_step": 165728, "epoch": 1862, "lr": 3.309065277213742e-05}
{"train_loss": 0.07418327778577805, "global_step": 165729, "epoch": 1862, "lr": 3.3090107195359966e-05}
{"train_loss": 0.08901534974575043, "global_step": 165730, "epoch": 1862, "lr": 3.308956162085586e-05}
{"train_loss": 0.07439900934696198, "global_step": 165731, "epoch": 1862, "lr": 3.308901604862513e-05}
{"train_loss": 0.12866170704364777, "global_step": 165732, "epoch": 1862, "lr": 3.30884704786679e-05}
{"train_loss": 0.06457553803920746, "global_step": 165733, "epoch": 1862, "lr": 3.3087924910984204e-05}
{"train_loss": 0.13451725244522095, "global_step": 165734, "epoch": 1862, "lr": 3.308737934557414e-05}
{"train_loss": 0.07155627757310867, "global_step": 165735, "epoch": 1862, "lr": 3.308683378243777e-05}
{"train_loss": 0.07296250015497208, "global_step": 165736, "epoch": 1862, "lr": 3.308628822157517e-05}
{"train_loss": 0.11477483808994293, "global_step": 165737, "epoch": 1862, "lr": 3.30857426629864e-05}
{"train_loss": 0.07275520265102386, "global_step": 165738, "epoch": 1862, "lr": 3.308519710667156e-05}
{"train_loss": 0.06815830618143082, "global_step": 165739, "epoch": 1862, "lr": 3.30846515526307e-05}
{"train_loss": 0.1156846135854721, "global_step": 165740, "epoch": 1862, "lr": 3.308410600086392e-05}
{"train_loss": 0.05969502031803131, "global_step": 165741, "epoch": 1862, "lr": 3.308356045137126e-05}
{"train_loss": 0.07900064438581467, "global_step": 165742, "epoch": 1862, "lr": 3.308301490415281e-05}
{"train_loss": 0.0741400271654129, "global_step": 165743, "epoch": 1862, "lr": 3.3082469359208644e-05}
{"train_loss": 0.11357371509075165, "global_step": 165744, "epoch": 1862, "lr": 3.3081923816538835e-05}
{"train_loss": 0.07484102994203568, "global_step": 165745, "epoch": 1862, "lr": 3.308137827614345e-05}
{"train_loss": 0.11936049908399582, "global_step": 165746, "epoch": 1862, "lr": 3.308083273802259e-05}
{"train_loss": 0.14020498096942902, "global_step": 165747, "epoch": 1862, "lr": 3.3080287202176275e-05}
{"train_loss": 0.08489061892032623, "global_step": 165748, "epoch": 1862, "lr": 3.3079741668604617e-05}
{"train_loss": 0.0685981810092926, "global_step": 165749, "epoch": 1862, "lr": 3.307919613730769e-05}
{"train_loss": 0.09911849349737167, "global_step": 165750, "epoch": 1862, "lr": 3.3078650608285546e-05}
{"train_loss": 0.09203790128231049, "global_step": 165751, "epoch": 1862, "lr": 3.307810508153829e-05}
{"train_loss": 0.11222957074642181, "global_step": 165752, "epoch": 1862, "lr": 3.3077559557065954e-05}
{"train_loss": 0.15431995689868927, "global_step": 165753, "epoch": 1862, "lr": 3.307701403486866e-05}
{"train_loss": 0.07487668097019196, "global_step": 165754, "epoch": 1862, "lr": 3.307646851494643e-05}
{"train_loss": 0.040903400629758835, "global_step": 165755, "epoch": 1862, "lr": 3.307592299729938e-05}
{"train_loss": 0.06916455179452896, "global_step": 165756, "epoch": 1862, "lr": 3.307537748192755e-05}
{"train_loss": 0.09885291755199432, "global_step": 165757, "epoch": 1862, "lr": 3.3074831968831056e-05}
{"train_loss": 0.12916122376918793, "global_step": 165758, "epoch": 1862, "lr": 3.3074286458009916e-05}
{"train_loss": 0.0833243876695633, "global_step": 165759, "epoch": 1862, "lr": 3.3073740949464246e-05}
{"train_loss": 0.07765450328588486, "global_step": 165760, "epoch": 1862, "lr": 3.307319544319411e-05}
{"train_loss": 0.03215064853429794, "global_step": 165761, "epoch": 1862, "lr": 3.307264993919957e-05}
{"train_loss": 0.056058384478092194, "global_step": 165762, "epoch": 1862, "lr": 3.307210443748071e-05}
{"train_loss": 0.04638427868485451, "global_step": 165763, "epoch": 1862, "lr": 3.3071558938037586e-05}
{"train_loss": 0.13598211109638214, "global_step": 165764, "epoch": 1862, "lr": 3.30710134408703e-05}
{"train_loss": 0.12167108058929443, "global_step": 165765, "epoch": 1862, "lr": 3.307046794597891e-05}
{"train_loss": 0.07000527530908585, "global_step": 165766, "epoch": 1862, "lr": 3.3069922453363475e-05}
{"train_loss": 0.060878969728946686, "global_step": 165767, "epoch": 1862, "lr": 3.306937696302408e-05}
{"train_loss": 0.08235009014606476, "global_step": 165768, "epoch": 1862, "lr": 3.306883147496081e-05}
{"train_loss": 0.07736774533987045, "global_step": 165769, "epoch": 1862, "lr": 3.306828598917373e-05}
{"train_loss": 0.05693632364273071, "global_step": 165770, "epoch": 1862, "lr": 3.3067740505662926e-05}
{"train_loss": 0.03636366128921509, "global_step": 165771, "epoch": 1862, "lr": 3.306719502442843e-05}
{"train_loss": 0.06407416611909866, "global_step": 165772, "epoch": 1862, "lr": 3.306664954547037e-05}
{"train_loss": 0.15214867889881134, "global_step": 165773, "epoch": 1862, "lr": 3.306610406878877e-05}
{"train_loss": 0.09974785149097443, "global_step": 165774, "epoch": 1862, "lr": 3.3065558594383743e-05}
{"train_loss": 0.13170519471168518, "global_step": 165775, "epoch": 1862, "lr": 3.3065013122255326e-05}
{"train_loss": 0.1564025580883026, "global_step": 165776, "epoch": 1862, "lr": 3.306446765240364e-05}
{"train_loss": 0.07201018184423447, "global_step": 165777, "epoch": 1862, "lr": 3.30639221848287e-05}
{"train_loss": 0.11057605594396591, "global_step": 165778, "epoch": 1862, "lr": 3.3063376719530637e-05}
{"train_loss": 0.09369558840990067, "global_step": 165779, "epoch": 1862, "lr": 3.3062831256509484e-05}
{"train_loss": 0.10477863252162933, "global_step": 165780, "epoch": 1862, "lr": 3.3062285795765325e-05}
{"train_loss": 0.10237821936607361, "global_step": 165781, "epoch": 1862, "lr": 3.3061740337298235e-05}
{"train_loss": 0.09057825058698654, "global_step": 165782, "epoch": 1862, "lr": 3.3061194881108295e-05}
{"train_loss": 0.04143171384930611, "global_step": 165783, "epoch": 1862, "lr": 3.306064942719556e-05}
{"train_loss": 0.08663298934698105, "global_step": 165784, "epoch": 1862, "lr": 3.306010397556011e-05}
{"train_loss": 0.09539425373077393, "global_step": 165785, "epoch": 1862, "lr": 3.305955852620204e-05}
{"train_loss": 0.11535318940877914, "global_step": 165786, "epoch": 1862, "lr": 3.305901307912138e-05}
{"train_loss": 0.08780607581138611, "global_step": 165787, "epoch": 1862, "lr": 3.3058467634318266e-05}
{"train_loss": 0.0866648405790329, "global_step": 165788, "epoch": 1862, "lr": 3.30579221917927e-05}
{"train_loss": 0.10349459946155548, "global_step": 165789, "epoch": 1862, "lr": 3.3057376751544816e-05}
{"train_loss": 0.09837502241134644, "global_step": 165790, "epoch": 1862, "lr": 3.305683131357464e-05}
{"train_loss": 0.09738301485776901, "global_step": 165791, "epoch": 1862, "lr": 3.305628587788229e-05}
{"train_loss": 0.07779455929994583, "global_step": 165792, "epoch": 1862, "lr": 3.3055740444467785e-05}
{"train_loss": 0.07428177446126938, "global_step": 165793, "epoch": 1862, "lr": 3.305519501333126e-05}
{"train_loss": 0.10868719965219498, "global_step": 165794, "epoch": 1862, "lr": 3.305464958447273e-05}
{"train_loss": 0.09312106668949127, "global_step": 165795, "epoch": 1862, "lr": 3.3054104157892325e-05}
{"train_loss": 0.10895919799804688, "global_step": 165796, "epoch": 1862, "lr": 3.3053558733590065e-05}
{"train_loss": 0.1355658769607544, "global_step": 165797, "epoch": 1862, "lr": 3.305301331156606e-05}
{"train_loss": 0.06595536321401596, "global_step": 165798, "epoch": 1862, "lr": 3.305246789182037e-05}
{"train_loss": 0.12130507826805115, "global_step": 165799, "epoch": 1862, "lr": 3.305192247435307e-05}
{"train_loss": 0.06691092997789383, "global_step": 165800, "epoch": 1862, "lr": 3.3051377059164225e-05}
{"train_loss": 0.07493038475513458, "global_step": 165801, "epoch": 1862, "lr": 3.305083164625392e-05}
{"train_loss": 0.06920917332172394, "global_step": 165802, "epoch": 1862, "lr": 3.305028623562223e-05}
{"train_loss": 0.04999760538339615, "global_step": 165803, "epoch": 1862, "lr": 3.30497408272692e-05}
{"train_loss": 0.09471999108791351, "global_step": 165804, "epoch": 1862, "lr": 3.3049195421194954e-05}
{"train_loss": 0.042520880699157715, "global_step": 165805, "epoch": 1862, "lr": 3.304865001739952e-05}
{"train_loss": 0.09157025018769703, "global_step": 165806, "epoch": 1862, "lr": 3.3048104615883e-05, "val_loss": 6.875964641571045}
{"train_loss": 0.12184024602174759, "global_step": 165807, "epoch": 1863, "lr": 3.304755921664544e-05}
{"train_loss": 0.09161265939474106, "global_step": 165808, "epoch": 1863, "lr": 3.304701381968694e-05}
{"train_loss": 0.0447419099509716, "global_step": 165809, "epoch": 1863, "lr": 3.304646842500756e-05}
{"train_loss": 0.08353061974048615, "global_step": 165810, "epoch": 1863, "lr": 3.3045923032607385e-05}
{"train_loss": 0.1443619579076767, "global_step": 165811, "epoch": 1863, "lr": 3.3045377642486455e-05}
{"train_loss": 0.0688696801662445, "global_step": 165812, "epoch": 1863, "lr": 3.30448322546449e-05}
{"train_loss": 0.08052871376276016, "global_step": 165813, "epoch": 1863, "lr": 3.304428686908273e-05}
{"train_loss": 0.11209160089492798, "global_step": 165814, "epoch": 1863, "lr": 3.3043741485800076e-05}
{"train_loss": 0.09017059952020645, "global_step": 165815, "epoch": 1863, "lr": 3.304319610479696e-05}
{"train_loss": 0.1535860300064087, "global_step": 165816, "epoch": 1863, "lr": 3.3042650726073496e-05}
{"train_loss": 0.07425849139690399, "global_step": 165817, "epoch": 1863, "lr": 3.304210534962973e-05}
{"train_loss": 0.08200813084840775, "global_step": 165818, "epoch": 1863, "lr": 3.3041559975465755e-05}
{"train_loss": 0.07600647956132889, "global_step": 165819, "epoch": 1863, "lr": 3.3041014603581634e-05}
{"train_loss": 0.06893643736839294, "global_step": 165820, "epoch": 1863, "lr": 3.304046923397743e-05}
{"train_loss": 0.09956087917089462, "global_step": 165821, "epoch": 1863, "lr": 3.3039923866653244e-05}
{"train_loss": 0.1722792088985443, "global_step": 165822, "epoch": 1863, "lr": 3.303937850160912e-05}
{"train_loss": 0.07738816738128662, "global_step": 165823, "epoch": 1863, "lr": 3.303883313884516e-05}
{"train_loss": 0.08751729130744934, "global_step": 165824, "epoch": 1863, "lr": 3.303828777836141e-05}
{"train_loss": 0.1385062336921692, "global_step": 165825, "epoch": 1863, "lr": 3.3037742420157966e-05}
{"train_loss": 0.12461558729410172, "global_step": 165826, "epoch": 1863, "lr": 3.3037197064234874e-05}
{"train_loss": 0.07748964428901672, "global_step": 165827, "epoch": 1863, "lr": 3.303665171059225e-05}
{"train_loss": 0.09468437731266022, "global_step": 165828, "epoch": 1863, "lr": 3.303610635923011e-05}
{"train_loss": 0.16946597397327423, "global_step": 165829, "epoch": 1863, "lr": 3.3035561010148586e-05}
{"train_loss": 0.11138034611940384, "global_step": 165830, "epoch": 1863, "lr": 3.30350156633477e-05}
{"train_loss": 0.12695668637752533, "global_step": 165831, "epoch": 1863, "lr": 3.3034470318827574e-05}
{"train_loss": 0.07996253669261932, "global_step": 165832, "epoch": 1863, "lr": 3.303392497658824e-05}
{"train_loss": 0.09985984861850739, "global_step": 165833, "epoch": 1863, "lr": 3.3033379636629804e-05}
{"train_loss": 0.09156432747840881, "global_step": 165834, "epoch": 1863, "lr": 3.3032834298952296e-05}
{"train_loss": 0.061380237340927124, "global_step": 165835, "epoch": 1863, "lr": 3.303228896355585e-05}
{"train_loss": 0.08765749633312225, "global_step": 165836, "epoch": 1863, "lr": 3.303174363044048e-05}
{"train_loss": 0.05676639825105667, "global_step": 165837, "epoch": 1863, "lr": 3.3031198299606305e-05}
{"train_loss": 0.06525009125471115, "global_step": 165838, "epoch": 1863, "lr": 3.303065297105337e-05}
{"train_loss": 0.0783427432179451, "global_step": 165839, "epoch": 1863, "lr": 3.303010764478174e-05}
{"train_loss": 0.05590147152543068, "global_step": 165840, "epoch": 1863, "lr": 3.302956232079153e-05}
{"train_loss": 0.049479078501462936, "global_step": 165841, "epoch": 1863, "lr": 3.302901699908276e-05}
{"train_loss": 0.19387483596801758, "global_step": 165842, "epoch": 1863, "lr": 3.302847167965556e-05}
{"train_loss": 0.1142423003911972, "global_step": 165843, "epoch": 1863, "lr": 3.302792636250996e-05}
{"train_loss": 0.04616781324148178, "global_step": 165844, "epoch": 1863, "lr": 3.302738104764607e-05}
{"train_loss": 0.08688192069530487, "global_step": 165845, "epoch": 1863, "lr": 3.302683573506391e-05}
{"train_loss": 0.09207987040281296, "global_step": 165846, "epoch": 1863, "lr": 3.302629042476361e-05}
{"train_loss": 0.07431824505329132, "global_step": 165847, "epoch": 1863, "lr": 3.30257451167452e-05}
{"train_loss": 0.07413055747747421, "global_step": 165848, "epoch": 1863, "lr": 3.3025199811008786e-05}
{"train_loss": 0.10359761118888855, "global_step": 165849, "epoch": 1863, "lr": 3.302465450755442e-05}
{"train_loss": 0.1267145276069641, "global_step": 165850, "epoch": 1863, "lr": 3.302410920638219e-05}
{"train_loss": 0.12087653577327728, "global_step": 165851, "epoch": 1863, "lr": 3.3023563907492145e-05}
{"train_loss": 0.08917514979839325, "global_step": 165852, "epoch": 1863, "lr": 3.302301861088439e-05}
{"train_loss": 0.06907322257757187, "global_step": 165853, "epoch": 1863, "lr": 3.302247331655898e-05}
{"train_loss": 0.0978371649980545, "global_step": 165854, "epoch": 1863, "lr": 3.302192802451599e-05}
{"train_loss": 0.12784509360790253, "global_step": 165855, "epoch": 1863, "lr": 3.302138273475549e-05}
{"train_loss": 0.05619616061449051, "global_step": 165856, "epoch": 1863, "lr": 3.3020837447277566e-05}
{"train_loss": 0.08517225086688995, "global_step": 165857, "epoch": 1863, "lr": 3.3020292162082286e-05}
{"train_loss": 0.1755666434764862, "global_step": 165858, "epoch": 1863, "lr": 3.3019746879169714e-05}
{"train_loss": 0.09814752638339996, "global_step": 165859, "epoch": 1863, "lr": 3.3019201598539935e-05}
{"train_loss": 0.03425389528274536, "global_step": 165860, "epoch": 1863, "lr": 3.3018656320193005e-05}
{"train_loss": 0.07959622889757156, "global_step": 165861, "epoch": 1863, "lr": 3.301811104412903e-05}
{"train_loss": 0.0532173290848732, "global_step": 165862, "epoch": 1863, "lr": 3.301756577034804e-05}
{"train_loss": 0.09084866940975189, "global_step": 165863, "epoch": 1863, "lr": 3.301702049885015e-05}
{"train_loss": 0.06132861226797104, "global_step": 165864, "epoch": 1863, "lr": 3.30164752296354e-05}
{"train_loss": 0.08107172697782516, "global_step": 165865, "epoch": 1863, "lr": 3.30159299627039e-05}
{"train_loss": 0.10141569375991821, "global_step": 165866, "epoch": 1863, "lr": 3.301538469805567e-05}
{"train_loss": 0.07810591161251068, "global_step": 165867, "epoch": 1863, "lr": 3.301483943569084e-05}
{"train_loss": 0.10528977960348129, "global_step": 165868, "epoch": 1863, "lr": 3.301429417560944e-05}
{"train_loss": 0.05385635048151016, "global_step": 165869, "epoch": 1863, "lr": 3.301374891781158e-05}
{"train_loss": 0.08958138525485992, "global_step": 165870, "epoch": 1863, "lr": 3.30132036622973e-05}
{"train_loss": 0.09843118488788605, "global_step": 165871, "epoch": 1863, "lr": 3.3012658409066696e-05}
{"train_loss": 0.06126770004630089, "global_step": 165872, "epoch": 1863, "lr": 3.301211315811982e-05}
{"train_loss": 0.06569213420152664, "global_step": 165873, "epoch": 1863, "lr": 3.301156790945677e-05}
{"train_loss": 0.124050073325634, "global_step": 165874, "epoch": 1863, "lr": 3.3011022663077604e-05}
{"train_loss": 0.1787584275007248, "global_step": 165875, "epoch": 1863, "lr": 3.30104774189824e-05}
{"train_loss": 0.04878111183643341, "global_step": 165876, "epoch": 1863, "lr": 3.3009932177171244e-05}
{"train_loss": 0.14972162246704102, "global_step": 165877, "epoch": 1863, "lr": 3.300938693764417e-05}
{"train_loss": 0.11561642587184906, "global_step": 165878, "epoch": 1863, "lr": 3.300884170040129e-05}
{"train_loss": 0.08730050176382065, "global_step": 165879, "epoch": 1863, "lr": 3.3008296465442656e-05}
{"train_loss": 0.08577360957860947, "global_step": 165880, "epoch": 1863, "lr": 3.3007751232768366e-05}
{"train_loss": 0.07968967407941818, "global_step": 165881, "epoch": 1863, "lr": 3.300720600237846e-05}
{"train_loss": 0.09389941394329071, "global_step": 165882, "epoch": 1863, "lr": 3.300666077427305e-05}
{"train_loss": 0.19441570341587067, "global_step": 165883, "epoch": 1863, "lr": 3.3006115548452156e-05}
{"train_loss": 0.14000257849693298, "global_step": 165884, "epoch": 1863, "lr": 3.3005570324915915e-05}
{"train_loss": 0.10109269618988037, "global_step": 165885, "epoch": 1863, "lr": 3.300502510366434e-05}
{"train_loss": 0.0561501607298851, "global_step": 165886, "epoch": 1863, "lr": 3.3004479884697556e-05}
{"train_loss": 0.0677097961306572, "global_step": 165887, "epoch": 1863, "lr": 3.300393466801559e-05}
{"train_loss": 0.06371630728244781, "global_step": 165888, "epoch": 1863, "lr": 3.3003389453618554e-05}
{"train_loss": 0.06463451683521271, "global_step": 165889, "epoch": 1863, "lr": 3.300284424150649e-05}
{"train_loss": 0.11022695899009705, "global_step": 165890, "epoch": 1863, "lr": 3.3002299031679505e-05}
{"train_loss": 0.047338731586933136, "global_step": 165891, "epoch": 1863, "lr": 3.300175382413764e-05}
{"train_loss": 0.09611701965332031, "global_step": 165892, "epoch": 1863, "lr": 3.3001208618881e-05}
{"train_loss": 0.0990346297621727, "global_step": 165893, "epoch": 1863, "lr": 3.300066341590961e-05}
{"train_loss": 0.06616067886352539, "global_step": 165894, "epoch": 1863, "lr": 3.300011821522359e-05}
{"train_loss": 0.09354089510239912, "global_step": 165895, "epoch": 1863, "lr": 3.299957301682301e-05, "val_loss": 7.0028791427612305}
{"train_loss": 0.04273057356476784, "global_step": 165896, "epoch": 1864, "lr": 3.299902782070791e-05}
{"train_loss": 0.07580841332674026, "global_step": 165897, "epoch": 1864, "lr": 3.29984826268784e-05}
{"train_loss": 0.07604030519723892, "global_step": 165898, "epoch": 1864, "lr": 3.2997937435334514e-05}
{"train_loss": 0.04296623915433884, "global_step": 165899, "epoch": 1864, "lr": 3.2997392246076375e-05}
{"train_loss": 0.06276226043701172, "global_step": 165900, "epoch": 1864, "lr": 3.2996847059104005e-05}
{"train_loss": 0.14050140976905823, "global_step": 165901, "epoch": 1864, "lr": 3.2996301874417526e-05}
{"train_loss": 0.07558004558086395, "global_step": 165902, "epoch": 1864, "lr": 3.2995756692016966e-05}
{"train_loss": 0.1107756495475769, "global_step": 165903, "epoch": 1864, "lr": 3.299521151190243e-05}
{"train_loss": 0.07053123414516449, "global_step": 165904, "epoch": 1864, "lr": 3.299466633407397e-05}
{"train_loss": 0.056575506925582886, "global_step": 165905, "epoch": 1864, "lr": 3.29941211585317e-05}
{"train_loss": 0.08254152536392212, "global_step": 165906, "epoch": 1864, "lr": 3.299357598527563e-05}
{"train_loss": 0.09446809440851212, "global_step": 165907, "epoch": 1864, "lr": 3.299303081430588e-05}
{"train_loss": 0.09103575348854065, "global_step": 165908, "epoch": 1864, "lr": 3.29924856456225e-05}
{"train_loss": 0.038686979562044144, "global_step": 165909, "epoch": 1864, "lr": 3.2991940479225595e-05}
{"train_loss": 0.08829522877931595, "global_step": 165910, "epoch": 1864, "lr": 3.299139531511518e-05}
{"train_loss": 0.06057881936430931, "global_step": 165911, "epoch": 1864, "lr": 3.2990850153291394e-05}
{"train_loss": 0.06416112929582596, "global_step": 165912, "epoch": 1864, "lr": 3.299030499375427e-05}
{"train_loss": 0.08513764292001724, "global_step": 165913, "epoch": 1864, "lr": 3.2989759836503896e-05}
{"train_loss": 0.08980387449264526, "global_step": 165914, "epoch": 1864, "lr": 3.298921468154034e-05}
{"train_loss": 0.08748802542686462, "global_step": 165915, "epoch": 1864, "lr": 3.298866952886368e-05}
{"train_loss": 0.09991943091154099, "global_step": 165916, "epoch": 1864, "lr": 3.2988124378473995e-05}
{"train_loss": 0.06608506292104721, "global_step": 165917, "epoch": 1864, "lr": 3.298757923037132e-05}
{"train_loss": 0.07579594850540161, "global_step": 165918, "epoch": 1864, "lr": 3.298703408455579e-05}
{"train_loss": 0.04441511631011963, "global_step": 165919, "epoch": 1864, "lr": 3.298648894102742e-05}
{"train_loss": 0.08905429393053055, "global_step": 165920, "epoch": 1864, "lr": 3.2985943799786333e-05}
{"train_loss": 0.08607597649097443, "global_step": 165921, "epoch": 1864, "lr": 3.298539866083255e-05}
{"train_loss": 0.08068154752254486, "global_step": 165922, "epoch": 1864, "lr": 3.29848535241662e-05}
{"train_loss": 0.09445322304964066, "global_step": 165923, "epoch": 1864, "lr": 3.2984308389787314e-05}
{"train_loss": 0.09895812720060349, "global_step": 165924, "epoch": 1864, "lr": 3.298376325769599e-05}
{"train_loss": 0.0594002902507782, "global_step": 165925, "epoch": 1864, "lr": 3.298321812789228e-05}
{"train_loss": 0.0860811099410057, "global_step": 165926, "epoch": 1864, "lr": 3.2982673000376284e-05}
{"train_loss": 0.03124593198299408, "global_step": 165927, "epoch": 1864, "lr": 3.298212787514804e-05}
{"train_loss": 0.10947108268737793, "global_step": 165928, "epoch": 1864, "lr": 3.298158275220766e-05}
{"train_loss": 0.11444254964590073, "global_step": 165929, "epoch": 1864, "lr": 3.298103763155518e-05}
{"train_loss": 0.13742360472679138, "global_step": 165930, "epoch": 1864, "lr": 3.2980492513190705e-05}
{"train_loss": 0.07408594340085983, "global_step": 165931, "epoch": 1864, "lr": 3.2979947397114295e-05}
{"train_loss": 0.0688614770770073, "global_step": 165932, "epoch": 1864, "lr": 3.297940228332602e-05}
{"train_loss": 0.0843929648399353, "global_step": 165933, "epoch": 1864, "lr": 3.297885717182596e-05}
{"train_loss": 0.08325062692165375, "global_step": 165934, "epoch": 1864, "lr": 3.297831206261418e-05}
{"train_loss": 0.09212770313024521, "global_step": 165935, "epoch": 1864, "lr": 3.297776695569078e-05}
{"train_loss": 0.09705445915460587, "global_step": 165936, "epoch": 1864, "lr": 3.297722185105578e-05}
{"train_loss": 0.10722887516021729, "global_step": 165937, "epoch": 1864, "lr": 3.2976676748709315e-05}
{"train_loss": 0.08180805295705795, "global_step": 165938, "epoch": 1864, "lr": 3.2976131648651396e-05}
{"train_loss": 0.06815987825393677, "global_step": 165939, "epoch": 1864, "lr": 3.297558655088216e-05}
{"train_loss": 0.1037074476480484, "global_step": 165940, "epoch": 1864, "lr": 3.297504145540162e-05}
{"train_loss": 0.10476267337799072, "global_step": 165941, "epoch": 1864, "lr": 3.2974496362209906e-05}
{"train_loss": 0.049398161470890045, "global_step": 165942, "epoch": 1864, "lr": 3.2973951271307044e-05}
{"train_loss": 0.07231494784355164, "global_step": 165943, "epoch": 1864, "lr": 3.2973406182693136e-05}
{"train_loss": 0.06959261000156403, "global_step": 165944, "epoch": 1864, "lr": 3.297286109636824e-05}
{"train_loss": 0.1157892569899559, "global_step": 165945, "epoch": 1864, "lr": 3.297231601233244e-05}
{"train_loss": 0.10344169288873672, "global_step": 165946, "epoch": 1864, "lr": 3.29717709305858e-05}
{"train_loss": 0.12057137489318848, "global_step": 165947, "epoch": 1864, "lr": 3.297122585112839e-05}
{"train_loss": 0.05512082949280739, "global_step": 165948, "epoch": 1864, "lr": 3.297068077396031e-05}
{"train_loss": 0.0845441222190857, "global_step": 165949, "epoch": 1864, "lr": 3.297013569908161e-05}
{"train_loss": 0.04196381941437721, "global_step": 165950, "epoch": 1864, "lr": 3.296959062649236e-05}
{"train_loss": 0.051591068506240845, "global_step": 165951, "epoch": 1864, "lr": 3.2969045556192644e-05}
{"train_loss": 0.07814475893974304, "global_step": 165952, "epoch": 1864, "lr": 3.296850048818253e-05}
{"train_loss": 0.0766337588429451, "global_step": 165953, "epoch": 1864, "lr": 3.29679554224621e-05}
{"train_loss": 0.13643020391464233, "global_step": 165954, "epoch": 1864, "lr": 3.296741035903143e-05}
{"train_loss": 0.05120182782411575, "global_step": 165955, "epoch": 1864, "lr": 3.296686529789056e-05}
{"train_loss": 0.05468754097819328, "global_step": 165956, "epoch": 1864, "lr": 3.296632023903961e-05}
{"train_loss": 0.08213527500629425, "global_step": 165957, "epoch": 1864, "lr": 3.296577518247861e-05}
{"train_loss": 0.04531648010015488, "global_step": 165958, "epoch": 1864, "lr": 3.296523012820767e-05}
{"train_loss": 0.05557418614625931, "global_step": 165959, "epoch": 1864, "lr": 3.296468507622683e-05}
{"train_loss": 0.08763536810874939, "global_step": 165960, "epoch": 1864, "lr": 3.296414002653621e-05}
{"train_loss": 0.03976486623287201, "global_step": 165961, "epoch": 1864, "lr": 3.2963594979135814e-05}
{"train_loss": 0.10659012943506241, "global_step": 165962, "epoch": 1864, "lr": 3.296304993402579e-05}
{"train_loss": 0.11328136175870895, "global_step": 165963, "epoch": 1864, "lr": 3.296250489120615e-05}
{"train_loss": 0.06434226781129837, "global_step": 165964, "epoch": 1864, "lr": 3.2961959850677013e-05}
{"train_loss": 0.13322040438652039, "global_step": 165965, "epoch": 1864, "lr": 3.296141481243842e-05}
{"train_loss": 0.12883608043193817, "global_step": 165966, "epoch": 1864, "lr": 3.296086977649045e-05}
{"train_loss": 0.09799856692552567, "global_step": 165967, "epoch": 1864, "lr": 3.2960324742833206e-05}
{"train_loss": 0.07474131882190704, "global_step": 165968, "epoch": 1864, "lr": 3.295977971146673e-05}
{"train_loss": 0.12352324277162552, "global_step": 165969, "epoch": 1864, "lr": 3.29592346823911e-05}
{"train_loss": 0.07000494748353958, "global_step": 165970, "epoch": 1864, "lr": 3.29586896556064e-05}
{"train_loss": 0.13429909944534302, "global_step": 165971, "epoch": 1864, "lr": 3.295814463111269e-05}
{"train_loss": 0.0912364050745964, "global_step": 165972, "epoch": 1864, "lr": 3.295759960891005e-05}
{"train_loss": 0.10932973027229309, "global_step": 165973, "epoch": 1864, "lr": 3.295705458899857e-05}
{"train_loss": 0.11678893864154816, "global_step": 165974, "epoch": 1864, "lr": 3.295650957137828e-05}
{"train_loss": 0.09269336611032486, "global_step": 165975, "epoch": 1864, "lr": 3.29559645560493e-05}
{"train_loss": 0.11094214767217636, "global_step": 165976, "epoch": 1864, "lr": 3.2955419543011665e-05}
{"train_loss": 0.09887087345123291, "global_step": 165977, "epoch": 1864, "lr": 3.2954874532265476e-05}
{"train_loss": 0.08146114647388458, "global_step": 165978, "epoch": 1864, "lr": 3.295432952381079e-05}
{"train_loss": 0.0543852299451828, "global_step": 165979, "epoch": 1864, "lr": 3.29537845176477e-05}
{"train_loss": 0.0689956396818161, "global_step": 165980, "epoch": 1864, "lr": 3.2953239513776245e-05}
{"train_loss": 0.09518023580312729, "global_step": 165981, "epoch": 1864, "lr": 3.2952694512196545e-05}
{"train_loss": 0.12957927584648132, "global_step": 165982, "epoch": 1864, "lr": 3.2952149512908624e-05}
{"train_loss": 0.11557981371879578, "global_step": 165983, "epoch": 1864, "lr": 3.2951604515912585e-05}
{"train_loss": 0.08472740633434124, "global_step": 165984, "epoch": 1864, "lr": 3.29510595212085e-05, "val_loss": 6.8671183586120605}
{"train_loss": 0.08850892633199692, "global_step": 165985, "epoch": 1865, "lr": 3.295051452879643e-05}
{"train_loss": 0.13142237067222595, "global_step": 165986, "epoch": 1865, "lr": 3.2949969538676474e-05}
{"train_loss": 0.11290855705738068, "global_step": 165987, "epoch": 1865, "lr": 3.294942455084867e-05}
{"train_loss": 0.12446306645870209, "global_step": 165988, "epoch": 1865, "lr": 3.294887956531312e-05}
{"train_loss": 0.09205649793148041, "global_step": 165989, "epoch": 1865, "lr": 3.294833458206987e-05}
{"train_loss": 0.06215668469667435, "global_step": 165990, "epoch": 1865, "lr": 3.2947789601119015e-05}
{"train_loss": 0.10798531025648117, "global_step": 165991, "epoch": 1865, "lr": 3.294724462246063e-05}
{"train_loss": 0.04901114106178284, "global_step": 165992, "epoch": 1865, "lr": 3.294669964609478e-05}
{"train_loss": 0.16531680524349213, "global_step": 165993, "epoch": 1865, "lr": 3.294615467202152e-05}
{"train_loss": 0.06946361064910889, "global_step": 165994, "epoch": 1865, "lr": 3.294560970024096e-05}
{"train_loss": 0.09466411918401718, "global_step": 165995, "epoch": 1865, "lr": 3.294506473075314e-05}
{"train_loss": 0.09650862216949463, "global_step": 165996, "epoch": 1865, "lr": 3.294451976355817e-05}
{"train_loss": 0.08363136649131775, "global_step": 165997, "epoch": 1865, "lr": 3.2943974798656074e-05}
{"train_loss": 0.08910820633172989, "global_step": 165998, "epoch": 1865, "lr": 3.294342983604698e-05}
{"train_loss": 0.0411306694149971, "global_step": 165999, "epoch": 1865, "lr": 3.2942884875730914e-05}
{"train_loss": 0.12381471693515778, "global_step": 166000, "epoch": 1865, "lr": 3.294233991770798e-05}
{"train_loss": 0.0640445202589035, "global_step": 166001, "epoch": 1865, "lr": 3.2941794961978226e-05}
{"train_loss": 0.0588187612593174, "global_step": 166002, "epoch": 1865, "lr": 3.2941250008541744e-05}
{"train_loss": 0.0856601670384407, "global_step": 166003, "epoch": 1865, "lr": 3.294070505739862e-05}
{"train_loss": 0.10355989634990692, "global_step": 166004, "epoch": 1865, "lr": 3.294016010854889e-05}
{"train_loss": 0.1247464120388031, "global_step": 166005, "epoch": 1865, "lr": 3.2939615161992666e-05}
{"train_loss": 0.17124611139297485, "global_step": 166006, "epoch": 1865, "lr": 3.293907021773e-05}
{"train_loss": 0.05446004867553711, "global_step": 166007, "epoch": 1865, "lr": 3.293852527576096e-05}
{"train_loss": 0.1358717381954193, "global_step": 166008, "epoch": 1865, "lr": 3.2937980336085625e-05}
{"train_loss": 0.13515040278434753, "global_step": 166009, "epoch": 1865, "lr": 3.2937435398704076e-05}
{"train_loss": 0.07477197796106339, "global_step": 166010, "epoch": 1865, "lr": 3.2936890463616374e-05}
{"train_loss": 0.07205606997013092, "global_step": 166011, "epoch": 1865, "lr": 3.293634553082262e-05}
{"train_loss": 0.06634096056222916, "global_step": 166012, "epoch": 1865, "lr": 3.293580060032284e-05}
{"train_loss": 0.11430379748344421, "global_step": 166013, "epoch": 1865, "lr": 3.293525567211715e-05}
{"train_loss": 0.13320577144622803, "global_step": 166014, "epoch": 1865, "lr": 3.29347107462056e-05}
{"train_loss": 0.07093087583780289, "global_step": 166015, "epoch": 1865, "lr": 3.293416582258827e-05}
{"train_loss": 0.08375229686498642, "global_step": 166016, "epoch": 1865, "lr": 3.293362090126523e-05}
{"train_loss": 0.06398287415504456, "global_step": 166017, "epoch": 1865, "lr": 3.293307598223657e-05}
{"train_loss": 0.09830641001462936, "global_step": 166018, "epoch": 1865, "lr": 3.2932531065502333e-05}
{"train_loss": 0.050354864448308945, "global_step": 166019, "epoch": 1865, "lr": 3.293198615106263e-05}
{"train_loss": 0.1199502944946289, "global_step": 166020, "epoch": 1865, "lr": 3.2931441238917494e-05}
{"train_loss": 0.12644483149051666, "global_step": 166021, "epoch": 1865, "lr": 3.293089632906701e-05}
{"train_loss": 0.07555016130208969, "global_step": 166022, "epoch": 1865, "lr": 3.293035142151129e-05}
{"train_loss": 0.08916544914245605, "global_step": 166023, "epoch": 1865, "lr": 3.2929806516250346e-05}
{"train_loss": 0.08661957085132599, "global_step": 166024, "epoch": 1865, "lr": 3.29292616132843e-05}
{"train_loss": 0.1330871880054474, "global_step": 166025, "epoch": 1865, "lr": 3.292871671261321e-05}
{"train_loss": 0.08428069204092026, "global_step": 166026, "epoch": 1865, "lr": 3.2928171814237134e-05}
{"train_loss": 0.11669149994850159, "global_step": 166027, "epoch": 1865, "lr": 3.292762691815616e-05}
{"train_loss": 0.0875631645321846, "global_step": 166028, "epoch": 1865, "lr": 3.292708202437036e-05}
{"train_loss": 0.08455464243888855, "global_step": 166029, "epoch": 1865, "lr": 3.292653713287981e-05}
{"train_loss": 0.07724089920520782, "global_step": 166030, "epoch": 1865, "lr": 3.292599224368457e-05}
{"train_loss": 0.0496097132563591, "global_step": 166031, "epoch": 1865, "lr": 3.2925447356784735e-05}
{"train_loss": 0.10401308536529541, "global_step": 166032, "epoch": 1865, "lr": 3.2924902472180366e-05}
{"train_loss": 0.06238166615366936, "global_step": 166033, "epoch": 1865, "lr": 3.292435758987151e-05}
{"train_loss": 0.06553316861391068, "global_step": 166034, "epoch": 1865, "lr": 3.292381270985829e-05}
{"train_loss": 0.054694660007953644, "global_step": 166035, "epoch": 1865, "lr": 3.2923267832140746e-05}
{"train_loss": 0.07786662131547928, "global_step": 166036, "epoch": 1865, "lr": 3.292272295671897e-05}
{"train_loss": 0.09363554418087006, "global_step": 166037, "epoch": 1865, "lr": 3.292217808359301e-05}
{"train_loss": 0.09221335500478745, "global_step": 166038, "epoch": 1865, "lr": 3.292163321276295e-05}
{"train_loss": 0.10217712819576263, "global_step": 166039, "epoch": 1865, "lr": 3.29210883442289e-05}
{"train_loss": 0.07035290449857712, "global_step": 166040, "epoch": 1865, "lr": 3.2920543477990865e-05}
{"train_loss": 0.0796063244342804, "global_step": 166041, "epoch": 1865, "lr": 3.2919998614048984e-05}
{"train_loss": 0.04999992996454239, "global_step": 166042, "epoch": 1865, "lr": 3.291945375240329e-05}
{"train_loss": 0.129069522023201, "global_step": 166043, "epoch": 1865, "lr": 3.291890889305387e-05}
{"train_loss": 0.09597327560186386, "global_step": 166044, "epoch": 1865, "lr": 3.291836403600078e-05}
{"train_loss": 0.04903113842010498, "global_step": 166045, "epoch": 1865, "lr": 3.2917819181244136e-05}
{"train_loss": 0.12676510214805603, "global_step": 166046, "epoch": 1865, "lr": 3.291727432878396e-05}
{"train_loss": 0.08188007026910782, "global_step": 166047, "epoch": 1865, "lr": 3.291672947862035e-05}
{"train_loss": 0.09548912197351456, "global_step": 166048, "epoch": 1865, "lr": 3.291618463075339e-05}
{"train_loss": 0.07416228950023651, "global_step": 166049, "epoch": 1865, "lr": 3.291563978518313e-05}
{"train_loss": 0.1723787635564804, "global_step": 166050, "epoch": 1865, "lr": 3.2915094941909666e-05}
{"train_loss": 0.07896123826503754, "global_step": 166051, "epoch": 1865, "lr": 3.291455010093306e-05}
{"train_loss": 0.07547201961278915, "global_step": 166052, "epoch": 1865, "lr": 3.291400526225336e-05}
{"train_loss": 0.07365281134843826, "global_step": 166053, "epoch": 1865, "lr": 3.291346042587069e-05}
{"train_loss": 0.08713142573833466, "global_step": 166054, "epoch": 1865, "lr": 3.291291559178507e-05}
{"train_loss": 0.06408676505088806, "global_step": 166055, "epoch": 1865, "lr": 3.291237075999663e-05}
{"train_loss": 0.15851451456546783, "global_step": 166056, "epoch": 1865, "lr": 3.291182593050539e-05}
{"train_loss": 0.08200609683990479, "global_step": 166057, "epoch": 1865, "lr": 3.291128110331144e-05}
{"train_loss": 0.08103179186582565, "global_step": 166058, "epoch": 1865, "lr": 3.2910736278414886e-05}
{"train_loss": 0.08731044083833694, "global_step": 166059, "epoch": 1865, "lr": 3.291019145581576e-05}
{"train_loss": 0.08018761873245239, "global_step": 166060, "epoch": 1865, "lr": 3.290964663551417e-05}
{"train_loss": 0.06525357812643051, "global_step": 166061, "epoch": 1865, "lr": 3.2909101817510145e-05}
{"train_loss": 0.08097021281719208, "global_step": 166062, "epoch": 1865, "lr": 3.29085570018038e-05}
{"train_loss": 0.06416967511177063, "global_step": 166063, "epoch": 1865, "lr": 3.2908012188395174e-05}
{"train_loss": 0.02556527964770794, "global_step": 166064, "epoch": 1865, "lr": 3.290746737728438e-05}
{"train_loss": 0.09283480793237686, "global_step": 166065, "epoch": 1865, "lr": 3.290692256847145e-05}
{"train_loss": 0.12212404608726501, "global_step": 166066, "epoch": 1865, "lr": 3.290637776195649e-05}
{"train_loss": 0.0800967738032341, "global_step": 166067, "epoch": 1865, "lr": 3.290583295773955e-05}
{"train_loss": 0.1099184900522232, "global_step": 166068, "epoch": 1865, "lr": 3.2905288155820715e-05}
{"train_loss": 0.13328693807125092, "global_step": 166069, "epoch": 1865, "lr": 3.290474335620005e-05}
{"train_loss": 0.1273604780435562, "global_step": 166070, "epoch": 1865, "lr": 3.2904198558877655e-05}
{"train_loss": 0.0660448744893074, "global_step": 166071, "epoch": 1865, "lr": 3.290365376385356e-05}
{"train_loss": 0.08898177742958069, "global_step": 166072, "epoch": 1865, "lr": 3.290310897112787e-05}
{"train_loss": 0.09119155368861857, "global_step": 166073, "epoch": 1865, "lr": 3.2902564180700636e-05, "val_loss": 7.004725456237793, "train_action_mse_error": 25.91888427734375}
{"train_loss": 0.19440139830112457, "global_step": 166074, "epoch": 1866, "lr": 3.2902019392571944e-05}
{"train_loss": 0.09438693523406982, "global_step": 166075, "epoch": 1866, "lr": 3.290147460674189e-05}
{"train_loss": 0.09343675523996353, "global_step": 166076, "epoch": 1866, "lr": 3.29009298232105e-05}
{"train_loss": 0.08735986053943634, "global_step": 166077, "epoch": 1866, "lr": 3.290038504197789e-05}
{"train_loss": 0.11620038002729416, "global_step": 166078, "epoch": 1866, "lr": 3.289984026304409e-05}
{"train_loss": 0.12665776908397675, "global_step": 166079, "epoch": 1866, "lr": 3.289929548640923e-05}
{"train_loss": 0.07711518555879593, "global_step": 166080, "epoch": 1866, "lr": 3.2898750712073325e-05}
{"train_loss": 0.07873724400997162, "global_step": 166081, "epoch": 1866, "lr": 3.28982059400365e-05}
{"train_loss": 0.0966285839676857, "global_step": 166082, "epoch": 1866, "lr": 3.2897661170298774e-05}
{"train_loss": 0.134043350815773, "global_step": 166083, "epoch": 1866, "lr": 3.289711640286027e-05}
{"train_loss": 0.10631097108125687, "global_step": 166084, "epoch": 1866, "lr": 3.289657163772104e-05}
{"train_loss": 0.16699063777923584, "global_step": 166085, "epoch": 1866, "lr": 3.289602687488115e-05}
{"train_loss": 0.10473596304655075, "global_step": 166086, "epoch": 1866, "lr": 3.289548211434068e-05}
{"train_loss": 0.0941309854388237, "global_step": 166087, "epoch": 1866, "lr": 3.289493735609971e-05}
{"train_loss": 0.04159151017665863, "global_step": 166088, "epoch": 1866, "lr": 3.2894392600158296e-05}
{"train_loss": 0.07794182747602463, "global_step": 166089, "epoch": 1866, "lr": 3.2893847846516544e-05}
{"train_loss": 0.058269817382097244, "global_step": 166090, "epoch": 1866, "lr": 3.289330309517448e-05}
{"train_loss": 0.07734118402004242, "global_step": 166091, "epoch": 1866, "lr": 3.2892758346132224e-05}
{"train_loss": 0.15286004543304443, "global_step": 166092, "epoch": 1866, "lr": 3.289221359938981e-05}
{"train_loss": 0.1152789518237114, "global_step": 166093, "epoch": 1866, "lr": 3.289166885494732e-05}
{"train_loss": 0.06956490129232407, "global_step": 166094, "epoch": 1866, "lr": 3.289112411280487e-05}
{"train_loss": 0.1254461407661438, "global_step": 166095, "epoch": 1866, "lr": 3.289057937296247e-05}
{"train_loss": 0.08750956505537033, "global_step": 166096, "epoch": 1866, "lr": 3.2890034635420256e-05}
{"train_loss": 0.07751420140266418, "global_step": 166097, "epoch": 1866, "lr": 3.288948990017823e-05}
{"train_loss": 0.08280158042907715, "global_step": 166098, "epoch": 1866, "lr": 3.288894516723654e-05}
{"train_loss": 0.07565096020698547, "global_step": 166099, "epoch": 1866, "lr": 3.288840043659519e-05}
{"train_loss": 0.11605224758386612, "global_step": 166100, "epoch": 1866, "lr": 3.2887855708254315e-05}
{"train_loss": 0.06074203923344612, "global_step": 166101, "epoch": 1866, "lr": 3.288731098221393e-05}
{"train_loss": 0.15734069049358368, "global_step": 166102, "epoch": 1866, "lr": 3.288676625847417e-05}
{"train_loss": 0.08263756334781647, "global_step": 166103, "epoch": 1866, "lr": 3.288622153703506e-05}
{"train_loss": 0.07746292650699615, "global_step": 166104, "epoch": 1866, "lr": 3.2885676817896684e-05}
{"train_loss": 0.09425210952758789, "global_step": 166105, "epoch": 1866, "lr": 3.2885132101059124e-05}
{"train_loss": 0.06617814302444458, "global_step": 166106, "epoch": 1866, "lr": 3.288458738652245e-05}
{"train_loss": 0.06859313696622849, "global_step": 166107, "epoch": 1866, "lr": 3.288404267428673e-05}
{"train_loss": 0.08711149543523788, "global_step": 166108, "epoch": 1866, "lr": 3.288349796435206e-05}
{"train_loss": 0.0774497464299202, "global_step": 166109, "epoch": 1866, "lr": 3.288295325671847e-05}
{"train_loss": 0.07678654044866562, "global_step": 166110, "epoch": 1866, "lr": 3.2882408551386055e-05}
{"train_loss": 0.08196325600147247, "global_step": 166111, "epoch": 1866, "lr": 3.288186384835492e-05}
{"train_loss": 0.09882860630750656, "global_step": 166112, "epoch": 1866, "lr": 3.288131914762508e-05}
{"train_loss": 0.10961775481700897, "global_step": 166113, "epoch": 1866, "lr": 3.2880774449196664e-05}
{"train_loss": 0.09509968012571335, "global_step": 166114, "epoch": 1866, "lr": 3.28802297530697e-05}
{"train_loss": 0.07548706978559494, "global_step": 166115, "epoch": 1866, "lr": 3.2879685059244305e-05}
{"train_loss": 0.1007830798625946, "global_step": 166116, "epoch": 1866, "lr": 3.2879140367720493e-05}
{"train_loss": 0.07454055547714233, "global_step": 166117, "epoch": 1866, "lr": 3.28785956784984e-05}
{"train_loss": 0.11432869732379913, "global_step": 166118, "epoch": 1866, "lr": 3.287805099157806e-05}
{"train_loss": 0.10082896053791046, "global_step": 166119, "epoch": 1866, "lr": 3.2877506306959564e-05}
{"train_loss": 0.06394688785076141, "global_step": 166120, "epoch": 1866, "lr": 3.287696162464297e-05}
{"train_loss": 0.06129595264792442, "global_step": 166121, "epoch": 1866, "lr": 3.287641694462837e-05}
{"train_loss": 0.06195239722728729, "global_step": 166122, "epoch": 1866, "lr": 3.287587226691582e-05}
{"train_loss": 0.07988301664590836, "global_step": 166123, "epoch": 1866, "lr": 3.28753275915054e-05}
{"train_loss": 0.0957007184624672, "global_step": 166124, "epoch": 1866, "lr": 3.2874782918397184e-05}
{"train_loss": 0.06882251054048538, "global_step": 166125, "epoch": 1866, "lr": 3.287423824759125e-05}
{"train_loss": 0.11660612374544144, "global_step": 166126, "epoch": 1866, "lr": 3.287369357908766e-05}
{"train_loss": 0.10934748500585556, "global_step": 166127, "epoch": 1866, "lr": 3.2873148912886506e-05}
{"train_loss": 0.05895426496863365, "global_step": 166128, "epoch": 1866, "lr": 3.287260424898783e-05}
{"train_loss": 0.09936939179897308, "global_step": 166129, "epoch": 1866, "lr": 3.2872059587391715e-05}
{"train_loss": 0.06359031796455383, "global_step": 166130, "epoch": 1866, "lr": 3.2871514928098266e-05}
{"train_loss": 0.1572515368461609, "global_step": 166131, "epoch": 1866, "lr": 3.287097027110751e-05}
{"train_loss": 0.1091398373246193, "global_step": 166132, "epoch": 1866, "lr": 3.287042561641958e-05}
{"train_loss": 0.11194978654384613, "global_step": 166133, "epoch": 1866, "lr": 3.286988096403447e-05}
{"train_loss": 0.11654537916183472, "global_step": 166134, "epoch": 1866, "lr": 3.2869336313952334e-05}
{"train_loss": 0.10932794958353043, "global_step": 166135, "epoch": 1866, "lr": 3.286879166617318e-05}
{"train_loss": 0.09548130631446838, "global_step": 166136, "epoch": 1866, "lr": 3.286824702069712e-05}
{"train_loss": 0.09480825066566467, "global_step": 166137, "epoch": 1866, "lr": 3.286770237752421e-05}
{"train_loss": 0.08456305414438248, "global_step": 166138, "epoch": 1866, "lr": 3.2867157736654544e-05}
{"train_loss": 0.12057530134916306, "global_step": 166139, "epoch": 1866, "lr": 3.2866613098088154e-05}
{"train_loss": 0.08145846426486969, "global_step": 166140, "epoch": 1866, "lr": 3.2866068461825164e-05}
{"train_loss": 0.053894590586423874, "global_step": 166141, "epoch": 1866, "lr": 3.286552382786561e-05}
{"train_loss": 0.07301370054483414, "global_step": 166142, "epoch": 1866, "lr": 3.286497919620958e-05}
{"train_loss": 0.0753995031118393, "global_step": 166143, "epoch": 1866, "lr": 3.286443456685714e-05}
{"train_loss": 0.10307129472494125, "global_step": 166144, "epoch": 1866, "lr": 3.286388993980837e-05}
{"train_loss": 0.10353139787912369, "global_step": 166145, "epoch": 1866, "lr": 3.286334531506334e-05}
{"train_loss": 0.09099097549915314, "global_step": 166146, "epoch": 1866, "lr": 3.2862800692622134e-05}
{"train_loss": 0.08060126006603241, "global_step": 166147, "epoch": 1866, "lr": 3.2862256072484795e-05}
{"train_loss": 0.06168830767273903, "global_step": 166148, "epoch": 1866, "lr": 3.286171145465142e-05}
{"train_loss": 0.08654409646987915, "global_step": 166149, "epoch": 1866, "lr": 3.2861166839122103e-05}
{"train_loss": 0.07692565023899078, "global_step": 166150, "epoch": 1866, "lr": 3.286062222589687e-05}
{"train_loss": 0.11513300240039825, "global_step": 166151, "epoch": 1866, "lr": 3.2860077614975835e-05}
{"train_loss": 0.17427566647529602, "global_step": 166152, "epoch": 1866, "lr": 3.285953300635904e-05}
{"train_loss": 0.07385671883821487, "global_step": 166153, "epoch": 1866, "lr": 3.2858988400046585e-05}
{"train_loss": 0.12931415438652039, "global_step": 166154, "epoch": 1866, "lr": 3.2858443796038516e-05}
{"train_loss": 0.12314886599779129, "global_step": 166155, "epoch": 1866, "lr": 3.285789919433493e-05}
{"train_loss": 0.09893716871738434, "global_step": 166156, "epoch": 1866, "lr": 3.285735459493588e-05}
{"train_loss": 0.11289123445749283, "global_step": 166157, "epoch": 1866, "lr": 3.285680999784146e-05}
{"train_loss": 0.043433938175439835, "global_step": 166158, "epoch": 1866, "lr": 3.285626540305172e-05}
{"train_loss": 0.1015286073088646, "global_step": 166159, "epoch": 1866, "lr": 3.285572081056677e-05}
{"train_loss": 0.1039484292268753, "global_step": 166160, "epoch": 1866, "lr": 3.2855176220386634e-05}
{"train_loss": 0.11969122290611267, "global_step": 166161, "epoch": 1866, "lr": 3.285463163251143e-05}
{"train_loss": 0.09530327171924409, "global_step": 166162, "epoch": 1866, "lr": 3.285408704694121e-05, "val_loss": 6.904907703399658}
{"train_loss": 0.12172766774892807, "global_step": 166163, "epoch": 1867, "lr": 3.285354246367604e-05}
{"train_loss": 0.0605936273932457, "global_step": 166164, "epoch": 1867, "lr": 3.285299788271601e-05}
{"train_loss": 0.11087334901094437, "global_step": 166165, "epoch": 1867, "lr": 3.285245330406117e-05}
{"train_loss": 0.1378885954618454, "global_step": 166166, "epoch": 1867, "lr": 3.2851908727711624e-05}
{"train_loss": 0.11058776825666428, "global_step": 166167, "epoch": 1867, "lr": 3.285136415366741e-05}
{"train_loss": 0.1543164700269699, "global_step": 166168, "epoch": 1867, "lr": 3.285081958192865e-05}
{"train_loss": 0.09843723475933075, "global_step": 166169, "epoch": 1867, "lr": 3.285027501249536e-05}
{"train_loss": 0.10238520056009293, "global_step": 166170, "epoch": 1867, "lr": 3.284973044536767e-05}
{"train_loss": 0.12031201273202896, "global_step": 166171, "epoch": 1867, "lr": 3.28491858805456e-05}
{"train_loss": 0.10009817779064178, "global_step": 166172, "epoch": 1867, "lr": 3.284864131802927e-05}
{"train_loss": 0.12123958766460419, "global_step": 166173, "epoch": 1867, "lr": 3.284809675781871e-05}
{"train_loss": 0.09844336658716202, "global_step": 166174, "epoch": 1867, "lr": 3.2847552199914035e-05}
{"train_loss": 0.1253581941127777, "global_step": 166175, "epoch": 1867, "lr": 3.2847007644315274e-05}
{"train_loss": 0.063446544110775, "global_step": 166176, "epoch": 1867, "lr": 3.2846463091022544e-05}
{"train_loss": 0.07862409949302673, "global_step": 166177, "epoch": 1867, "lr": 3.2845918540035884e-05}
{"train_loss": 0.15061478316783905, "global_step": 166178, "epoch": 1867, "lr": 3.2845373991355396e-05}
{"train_loss": 0.07550930976867676, "global_step": 166179, "epoch": 1867, "lr": 3.2844829444981116e-05}
{"train_loss": 0.05313742533326149, "global_step": 166180, "epoch": 1867, "lr": 3.284428490091316e-05}
{"train_loss": 0.0543651357293129, "global_step": 166181, "epoch": 1867, "lr": 3.284374035915157e-05}
{"train_loss": 0.0946061909198761, "global_step": 166182, "epoch": 1867, "lr": 3.284319581969644e-05}
{"train_loss": 0.07280801981687546, "global_step": 166183, "epoch": 1867, "lr": 3.284265128254783e-05}
{"train_loss": 0.11100941151380539, "global_step": 166184, "epoch": 1867, "lr": 3.28421067477058e-05}
{"train_loss": 0.03834162652492523, "global_step": 166185, "epoch": 1867, "lr": 3.284156221517046e-05}
{"train_loss": 0.09281624853610992, "global_step": 166186, "epoch": 1867, "lr": 3.284101768494185e-05}
{"train_loss": 0.09810776263475418, "global_step": 166187, "epoch": 1867, "lr": 3.284047315702007e-05}
{"train_loss": 0.1024520993232727, "global_step": 166188, "epoch": 1867, "lr": 3.2839928631405166e-05}
{"train_loss": 0.07172174006700516, "global_step": 166189, "epoch": 1867, "lr": 3.283938410809724e-05}
{"train_loss": 0.06847594678401947, "global_step": 166190, "epoch": 1867, "lr": 3.2838839587096325e-05}
{"train_loss": 0.14732477068901062, "global_step": 166191, "epoch": 1867, "lr": 3.2838295068402545e-05}
{"train_loss": 0.058693237602710724, "global_step": 166192, "epoch": 1867, "lr": 3.283775055201592e-05}
{"train_loss": 0.08022989332675934, "global_step": 166193, "epoch": 1867, "lr": 3.283720603793658e-05}
{"train_loss": 0.08880974352359772, "global_step": 166194, "epoch": 1867, "lr": 3.2836661526164544e-05}
{"train_loss": 0.1123371422290802, "global_step": 166195, "epoch": 1867, "lr": 3.283611701669993e-05}
{"train_loss": 0.10674618929624557, "global_step": 166196, "epoch": 1867, "lr": 3.2835572509542764e-05}
{"train_loss": 0.0487854965031147, "global_step": 166197, "epoch": 1867, "lr": 3.283502800469318e-05}
{"train_loss": 0.10461792349815369, "global_step": 166198, "epoch": 1867, "lr": 3.283448350215119e-05}
{"train_loss": 0.11637317389249802, "global_step": 166199, "epoch": 1867, "lr": 3.2833939001916914e-05}
{"train_loss": 0.07148337364196777, "global_step": 166200, "epoch": 1867, "lr": 3.28333945039904e-05}
{"train_loss": 0.09852975606918335, "global_step": 166201, "epoch": 1867, "lr": 3.2832850008371706e-05}
{"train_loss": 0.08923742920160294, "global_step": 166202, "epoch": 1867, "lr": 3.2832305515060956e-05}
{"train_loss": 0.07980232685804367, "global_step": 166203, "epoch": 1867, "lr": 3.2831761024058175e-05}
{"train_loss": 0.11460302770137787, "global_step": 166204, "epoch": 1867, "lr": 3.283121653536346e-05}
{"train_loss": 0.11004474014043808, "global_step": 166205, "epoch": 1867, "lr": 3.2830672048976875e-05}
{"train_loss": 0.05813983455300331, "global_step": 166206, "epoch": 1867, "lr": 3.283012756489851e-05}
{"train_loss": 0.089627206325531, "global_step": 166207, "epoch": 1867, "lr": 3.28295830831284e-05}
{"train_loss": 0.07334166765213013, "global_step": 166208, "epoch": 1867, "lr": 3.282903860366667e-05}
{"train_loss": 0.06966402381658554, "global_step": 166209, "epoch": 1867, "lr": 3.2828494126513345e-05}
{"train_loss": 0.0850181058049202, "global_step": 166210, "epoch": 1867, "lr": 3.2827949651668544e-05}
{"train_loss": 0.13122297823429108, "global_step": 166211, "epoch": 1867, "lr": 3.282740517913229e-05}
{"train_loss": 0.1046065241098404, "global_step": 166212, "epoch": 1867, "lr": 3.28268607089047e-05}
{"train_loss": 0.03897859528660774, "global_step": 166213, "epoch": 1867, "lr": 3.282631624098582e-05}
{"train_loss": 0.13383226096630096, "global_step": 166214, "epoch": 1867, "lr": 3.282577177537574e-05}
{"train_loss": 0.041515279561281204, "global_step": 166215, "epoch": 1867, "lr": 3.2825227312074516e-05}
{"train_loss": 0.07897259294986725, "global_step": 166216, "epoch": 1867, "lr": 3.282468285108225e-05}
{"train_loss": 0.10919375717639923, "global_step": 166217, "epoch": 1867, "lr": 3.2824138392398964e-05}
{"train_loss": 0.06509749591350555, "global_step": 166218, "epoch": 1867, "lr": 3.2823593936024786e-05}
{"train_loss": 0.06901656836271286, "global_step": 166219, "epoch": 1867, "lr": 3.282304948195977e-05}
{"train_loss": 0.10135423392057419, "global_step": 166220, "epoch": 1867, "lr": 3.282250503020397e-05}
{"train_loss": 0.11732804775238037, "global_step": 166221, "epoch": 1867, "lr": 3.2821960580757484e-05}
{"train_loss": 0.06201298162341118, "global_step": 166222, "epoch": 1867, "lr": 3.282141613362036e-05}
{"train_loss": 0.12196094542741776, "global_step": 166223, "epoch": 1867, "lr": 3.282087168879272e-05}
{"train_loss": 0.08474046736955643, "global_step": 166224, "epoch": 1867, "lr": 3.2820327246274574e-05}
{"train_loss": 0.0807439312338829, "global_step": 166225, "epoch": 1867, "lr": 3.281978280606605e-05}
{"train_loss": 0.05134038254618645, "global_step": 166226, "epoch": 1867, "lr": 3.281923836816717e-05}
{"train_loss": 0.04959024116396904, "global_step": 166227, "epoch": 1867, "lr": 3.281869393257806e-05}
{"train_loss": 0.03277778625488281, "global_step": 166228, "epoch": 1867, "lr": 3.281814949929875e-05}
{"train_loss": 0.09692392498254776, "global_step": 166229, "epoch": 1867, "lr": 3.281760506832934e-05}
{"train_loss": 0.09797828644514084, "global_step": 166230, "epoch": 1867, "lr": 3.281706063966988e-05}
{"train_loss": 0.08214456588029861, "global_step": 166231, "epoch": 1867, "lr": 3.2816516213320474e-05}
{"train_loss": 0.08371999114751816, "global_step": 166232, "epoch": 1867, "lr": 3.281597178928116e-05}
{"train_loss": 0.11425583064556122, "global_step": 166233, "epoch": 1867, "lr": 3.2815427367552056e-05}
{"train_loss": 0.043547481298446655, "global_step": 166234, "epoch": 1867, "lr": 3.2814882948133174e-05}
{"train_loss": 0.04631347954273224, "global_step": 166235, "epoch": 1867, "lr": 3.281433853102465e-05}
{"train_loss": 0.08235380053520203, "global_step": 166236, "epoch": 1867, "lr": 3.281379411622651e-05}
{"train_loss": 0.03305086866021156, "global_step": 166237, "epoch": 1867, "lr": 3.281324970373885e-05}
{"train_loss": 0.07602225244045258, "global_step": 166238, "epoch": 1867, "lr": 3.281270529356176e-05}
{"train_loss": 0.05967696011066437, "global_step": 166239, "epoch": 1867, "lr": 3.2812160885695266e-05}
{"train_loss": 0.08209758251905441, "global_step": 166240, "epoch": 1867, "lr": 3.281161648013949e-05}
{"train_loss": 0.06069822609424591, "global_step": 166241, "epoch": 1867, "lr": 3.281107207689445e-05}
{"train_loss": 0.0572916641831398, "global_step": 166242, "epoch": 1867, "lr": 3.281052767596028e-05}
{"train_loss": 0.10904596745967865, "global_step": 166243, "epoch": 1867, "lr": 3.280998327733701e-05}
{"train_loss": 0.06897777318954468, "global_step": 166244, "epoch": 1867, "lr": 3.280943888102475e-05}
{"train_loss": 0.06386303901672363, "global_step": 166245, "epoch": 1867, "lr": 3.280889448702352e-05}
{"train_loss": 0.056854259222745895, "global_step": 166246, "epoch": 1867, "lr": 3.280835009533345e-05}
{"train_loss": 0.09598693251609802, "global_step": 166247, "epoch": 1867, "lr": 3.280780570595456e-05}
{"train_loss": 0.1344318985939026, "global_step": 166248, "epoch": 1867, "lr": 3.2807261318886986e-05}
{"train_loss": 0.060956623405218124, "global_step": 166249, "epoch": 1867, "lr": 3.2806716934130746e-05}
{"train_loss": 0.1101689338684082, "global_step": 166250, "epoch": 1867, "lr": 3.280617255168594e-05}
{"train_loss": 0.08688274385888925, "global_step": 166251, "epoch": 1867, "lr": 3.280562817155262e-05, "val_loss": 7.0247802734375}
{"train_loss": 0.07964060455560684, "global_step": 166252, "epoch": 1868, "lr": 3.2805083793730894e-05}
{"train_loss": 0.11224157363176346, "global_step": 166253, "epoch": 1868, "lr": 3.28045394182208e-05}
{"train_loss": 0.08587262034416199, "global_step": 166254, "epoch": 1868, "lr": 3.2803995045022436e-05}
{"train_loss": 0.06697703897953033, "global_step": 166255, "epoch": 1868, "lr": 3.280345067413585e-05}
{"train_loss": 0.07114893198013306, "global_step": 166256, "epoch": 1868, "lr": 3.2802906305561146e-05}
{"train_loss": 0.12211081385612488, "global_step": 166257, "epoch": 1868, "lr": 3.2802361939298386e-05}
{"train_loss": 0.04930933192372322, "global_step": 166258, "epoch": 1868, "lr": 3.2801817575347626e-05}
{"train_loss": 0.05879521369934082, "global_step": 166259, "epoch": 1868, "lr": 3.280127321370896e-05}
{"train_loss": 0.054141879081726074, "global_step": 166260, "epoch": 1868, "lr": 3.2800728854382445e-05}
{"train_loss": 0.03815505653619766, "global_step": 166261, "epoch": 1868, "lr": 3.280018449736817e-05}
{"train_loss": 0.11164189130067825, "global_step": 166262, "epoch": 1868, "lr": 3.2799640142666186e-05}
{"train_loss": 0.12687735259532928, "global_step": 166263, "epoch": 1868, "lr": 3.2799095790276606e-05}
{"train_loss": 0.15517392754554749, "global_step": 166264, "epoch": 1868, "lr": 3.279855144019945e-05}
{"train_loss": 0.11262116581201553, "global_step": 166265, "epoch": 1868, "lr": 3.279800709243485e-05}
{"train_loss": 0.09940749406814575, "global_step": 166266, "epoch": 1868, "lr": 3.279746274698282e-05}
{"train_loss": 0.07101438194513321, "global_step": 166267, "epoch": 1868, "lr": 3.279691840384348e-05}
{"train_loss": 0.05814317613840103, "global_step": 166268, "epoch": 1868, "lr": 3.279637406301687e-05}
{"train_loss": 0.1334647685289383, "global_step": 166269, "epoch": 1868, "lr": 3.2795829724503094e-05}
{"train_loss": 0.10101312398910522, "global_step": 166270, "epoch": 1868, "lr": 3.27952853883022e-05}
{"train_loss": 0.0582120381295681, "global_step": 166271, "epoch": 1868, "lr": 3.2794741054414276e-05}
{"train_loss": 0.09619273990392685, "global_step": 166272, "epoch": 1868, "lr": 3.279419672283937e-05}
{"train_loss": 0.037219755351543427, "global_step": 166273, "epoch": 1868, "lr": 3.27936523935776e-05}
{"train_loss": 0.06761203706264496, "global_step": 166274, "epoch": 1868, "lr": 3.279310806662899e-05}
{"train_loss": 0.08506280928850174, "global_step": 166275, "epoch": 1868, "lr": 3.279256374199365e-05}
{"train_loss": 0.03510896861553192, "global_step": 166276, "epoch": 1868, "lr": 3.279201941967164e-05}
{"train_loss": 0.1798783838748932, "global_step": 166277, "epoch": 1868, "lr": 3.279147509966303e-05}
{"train_loss": 0.08988821506500244, "global_step": 166278, "epoch": 1868, "lr": 3.279093078196791e-05}
{"train_loss": 0.04854198917746544, "global_step": 166279, "epoch": 1868, "lr": 3.279038646658632e-05}
{"train_loss": 0.12888476252555847, "global_step": 166280, "epoch": 1868, "lr": 3.278984215351837e-05}
{"train_loss": 0.07308640331029892, "global_step": 166281, "epoch": 1868, "lr": 3.27892978427641e-05}
{"train_loss": 0.06686470657587051, "global_step": 166282, "epoch": 1868, "lr": 3.2788753534323614e-05}
{"train_loss": 0.06525151431560516, "global_step": 166283, "epoch": 1868, "lr": 3.2788209228196954e-05}
{"train_loss": 0.10486973077058792, "global_step": 166284, "epoch": 1868, "lr": 3.2787664924384226e-05}
{"train_loss": 0.1287296563386917, "global_step": 166285, "epoch": 1868, "lr": 3.278712062288547e-05}
{"train_loss": 0.08508969098329544, "global_step": 166286, "epoch": 1868, "lr": 3.278657632370079e-05}
{"train_loss": 0.09623689949512482, "global_step": 166287, "epoch": 1868, "lr": 3.278603202683023e-05}
{"train_loss": 0.10453300178050995, "global_step": 166288, "epoch": 1868, "lr": 3.27854877322739e-05}
{"train_loss": 0.08953318744897842, "global_step": 166289, "epoch": 1868, "lr": 3.278494344003183e-05}
{"train_loss": 0.10379529744386673, "global_step": 166290, "epoch": 1868, "lr": 3.278439915010413e-05}
{"train_loss": 0.1013883650302887, "global_step": 166291, "epoch": 1868, "lr": 3.278385486249085e-05}
{"train_loss": 0.07022179663181305, "global_step": 166292, "epoch": 1868, "lr": 3.2783310577192064e-05}
{"train_loss": 0.02654060162603855, "global_step": 166293, "epoch": 1868, "lr": 3.278276629420787e-05}
{"train_loss": 0.07504602521657944, "global_step": 166294, "epoch": 1868, "lr": 3.278222201353831e-05}
{"train_loss": 0.06374765932559967, "global_step": 166295, "epoch": 1868, "lr": 3.278167773518347e-05}
{"train_loss": 0.08738541603088379, "global_step": 166296, "epoch": 1868, "lr": 3.278113345914343e-05}
{"train_loss": 0.09575244784355164, "global_step": 166297, "epoch": 1868, "lr": 3.278058918541826e-05}
{"train_loss": 0.052386894822120667, "global_step": 166298, "epoch": 1868, "lr": 3.2780044914008013e-05}
{"train_loss": 0.0706634670495987, "global_step": 166299, "epoch": 1868, "lr": 3.27795006449128e-05}
{"train_loss": 0.034737780690193176, "global_step": 166300, "epoch": 1868, "lr": 3.277895637813266e-05}
{"train_loss": 0.0913902297616005, "global_step": 166301, "epoch": 1868, "lr": 3.277841211366769e-05}
{"train_loss": 0.14660699665546417, "global_step": 166302, "epoch": 1868, "lr": 3.277786785151793e-05}
{"train_loss": 0.09378675371408463, "global_step": 166303, "epoch": 1868, "lr": 3.2777323591683507e-05}
{"train_loss": 0.05843295156955719, "global_step": 166304, "epoch": 1868, "lr": 3.2776779334164435e-05}
{"train_loss": 0.05710848420858383, "global_step": 166305, "epoch": 1868, "lr": 3.277623507896084e-05}
{"train_loss": 0.07215587049722672, "global_step": 166306, "epoch": 1868, "lr": 3.277569082607275e-05}
{"train_loss": 0.03026588447391987, "global_step": 166307, "epoch": 1868, "lr": 3.277514657550027e-05}
{"train_loss": 0.0748869851231575, "global_step": 166308, "epoch": 1868, "lr": 3.2774602327243456e-05}
{"train_loss": 0.11051012575626373, "global_step": 166309, "epoch": 1868, "lr": 3.27740580813024e-05}
{"train_loss": 0.08070538938045502, "global_step": 166310, "epoch": 1868, "lr": 3.277351383767713e-05}
{"train_loss": 0.061243705451488495, "global_step": 166311, "epoch": 1868, "lr": 3.277296959636778e-05}
{"train_loss": 0.07322709262371063, "global_step": 166312, "epoch": 1868, "lr": 3.277242535737438e-05}
{"train_loss": 0.04558848589658737, "global_step": 166313, "epoch": 1868, "lr": 3.277188112069702e-05}
{"train_loss": 0.12575991451740265, "global_step": 166314, "epoch": 1868, "lr": 3.277133688633578e-05}
{"train_loss": 0.1361784189939499, "global_step": 166315, "epoch": 1868, "lr": 3.2770792654290715e-05}
{"train_loss": 0.0769340768456459, "global_step": 166316, "epoch": 1868, "lr": 3.277024842456192e-05}
{"train_loss": 0.10547192394733429, "global_step": 166317, "epoch": 1868, "lr": 3.276970419714943e-05}
{"train_loss": 0.11420900374650955, "global_step": 166318, "epoch": 1868, "lr": 3.276915997205337e-05}
{"train_loss": 0.08283233642578125, "global_step": 166319, "epoch": 1868, "lr": 3.276861574927377e-05}
{"train_loss": 0.07674138993024826, "global_step": 166320, "epoch": 1868, "lr": 3.276807152881073e-05}
{"train_loss": 0.08311847597360611, "global_step": 166321, "epoch": 1868, "lr": 3.276752731066429e-05}
{"train_loss": 0.11077864468097687, "global_step": 166322, "epoch": 1868, "lr": 3.276698309483458e-05}
{"train_loss": 0.08984125405550003, "global_step": 166323, "epoch": 1868, "lr": 3.276643888132161e-05}
{"train_loss": 0.12676382064819336, "global_step": 166324, "epoch": 1868, "lr": 3.27658946701255e-05}
{"train_loss": 0.10476522147655487, "global_step": 166325, "epoch": 1868, "lr": 3.276535046124629e-05}
{"train_loss": 0.12139669805765152, "global_step": 166326, "epoch": 1868, "lr": 3.276480625468409e-05}
{"train_loss": 0.07338850945234299, "global_step": 166327, "epoch": 1868, "lr": 3.276426205043893e-05}
{"train_loss": 0.08685442060232162, "global_step": 166328, "epoch": 1868, "lr": 3.27637178485109e-05}
{"train_loss": 0.07036619633436203, "global_step": 166329, "epoch": 1868, "lr": 3.27631736489001e-05}
{"train_loss": 0.084522545337677, "global_step": 166330, "epoch": 1868, "lr": 3.2762629451606576e-05}
{"train_loss": 0.0938836857676506, "global_step": 166331, "epoch": 1868, "lr": 3.2762085256630405e-05}
{"train_loss": 0.08790891617536545, "global_step": 166332, "epoch": 1868, "lr": 3.276154106397166e-05}
{"train_loss": 0.08498256653547287, "global_step": 166333, "epoch": 1868, "lr": 3.2760996873630415e-05}
{"train_loss": 0.059202782809734344, "global_step": 166334, "epoch": 1868, "lr": 3.2760452685606744e-05}
{"train_loss": 0.06494583189487457, "global_step": 166335, "epoch": 1868, "lr": 3.275990849990074e-05}
{"train_loss": 0.08694314956665039, "global_step": 166336, "epoch": 1868, "lr": 3.275936431651242e-05}
{"train_loss": 0.03449591249227524, "global_step": 166337, "epoch": 1868, "lr": 3.275882013544193e-05}
{"train_loss": 0.12718725204467773, "global_step": 166338, "epoch": 1868, "lr": 3.275827595668927e-05}
{"train_loss": 0.07320339232683182, "global_step": 166339, "epoch": 1868, "lr": 3.2757731780254575e-05}
{"train_loss": 0.08465495699325974, "global_step": 166340, "epoch": 1868, "lr": 3.275718760613787e-05, "val_loss": 6.957510948181152}
{"train_loss": 0.06921374797821045, "global_step": 166341, "epoch": 1869, "lr": 3.2756643434339286e-05}
{"train_loss": 0.07119030505418777, "global_step": 166342, "epoch": 1869, "lr": 3.275609926485883e-05}
{"train_loss": 0.09027551859617233, "global_step": 166343, "epoch": 1869, "lr": 3.275555509769662e-05}
{"train_loss": 0.05467422306537628, "global_step": 166344, "epoch": 1869, "lr": 3.27550109328527e-05}
{"train_loss": 0.09823227673768997, "global_step": 166345, "epoch": 1869, "lr": 3.275446677032719e-05}
{"train_loss": 0.138985276222229, "global_step": 166346, "epoch": 1869, "lr": 3.27539226101201e-05}
{"train_loss": 0.09019432216882706, "global_step": 166347, "epoch": 1869, "lr": 3.275337845223154e-05}
{"train_loss": 0.08974600583314896, "global_step": 166348, "epoch": 1869, "lr": 3.2752834296661596e-05}
{"train_loss": 0.07773855328559875, "global_step": 166349, "epoch": 1869, "lr": 3.2752290143410316e-05}
{"train_loss": 0.09412244707345963, "global_step": 166350, "epoch": 1869, "lr": 3.2751745992477764e-05}
{"train_loss": 0.06415104866027832, "global_step": 166351, "epoch": 1869, "lr": 3.275120184386405e-05}
{"train_loss": 0.06889057159423828, "global_step": 166352, "epoch": 1869, "lr": 3.275065769756922e-05}
{"train_loss": 0.06946063786745071, "global_step": 166353, "epoch": 1869, "lr": 3.275011355359335e-05}
{"train_loss": 0.04990164190530777, "global_step": 166354, "epoch": 1869, "lr": 3.2749569411936535e-05}
{"train_loss": 0.07361684739589691, "global_step": 166355, "epoch": 1869, "lr": 3.274902527259881e-05}
{"train_loss": 0.11335058510303497, "global_step": 166356, "epoch": 1869, "lr": 3.2748481135580276e-05}
{"train_loss": 0.05028792470693588, "global_step": 166357, "epoch": 1869, "lr": 3.274793700088099e-05}
{"train_loss": 0.07493019849061966, "global_step": 166358, "epoch": 1869, "lr": 3.274739286850105e-05}
{"train_loss": 0.16456446051597595, "global_step": 166359, "epoch": 1869, "lr": 3.274684873844049e-05}
{"train_loss": 0.06359604001045227, "global_step": 166360, "epoch": 1869, "lr": 3.274630461069943e-05}
{"train_loss": 0.09871397912502289, "global_step": 166361, "epoch": 1869, "lr": 3.27457604852779e-05}
{"train_loss": 0.13216632604599, "global_step": 166362, "epoch": 1869, "lr": 3.2745216362176004e-05}
{"train_loss": 0.0865427777171135, "global_step": 166363, "epoch": 1869, "lr": 3.2744672241393794e-05}
{"train_loss": 0.080217644572258, "global_step": 166364, "epoch": 1869, "lr": 3.274412812293134e-05}
{"train_loss": 0.08992274105548859, "global_step": 166365, "epoch": 1869, "lr": 3.274358400678876e-05}
{"train_loss": 0.061741285026073456, "global_step": 166366, "epoch": 1869, "lr": 3.274303989296607e-05}
{"train_loss": 0.054730355739593506, "global_step": 166367, "epoch": 1869, "lr": 3.274249578146339e-05}
{"train_loss": 0.09894078969955444, "global_step": 166368, "epoch": 1869, "lr": 3.274195167228076e-05}
{"train_loss": 0.0908270999789238, "global_step": 166369, "epoch": 1869, "lr": 3.274140756541826e-05}
{"train_loss": 0.10214634239673615, "global_step": 166370, "epoch": 1869, "lr": 3.274086346087597e-05}
{"train_loss": 0.06658133119344711, "global_step": 166371, "epoch": 1869, "lr": 3.274031935865396e-05}
{"train_loss": 0.05101213604211807, "global_step": 166372, "epoch": 1869, "lr": 3.27397752587523e-05}
{"train_loss": 0.0664830133318901, "global_step": 166373, "epoch": 1869, "lr": 3.273923116117108e-05}
{"train_loss": 0.0640592947602272, "global_step": 166374, "epoch": 1869, "lr": 3.273868706591035e-05}
{"train_loss": 0.11260607838630676, "global_step": 166375, "epoch": 1869, "lr": 3.2738142972970205e-05}
{"train_loss": 0.060935042798519135, "global_step": 166376, "epoch": 1869, "lr": 3.273759888235068e-05}
{"train_loss": 0.08329672366380692, "global_step": 166377, "epoch": 1869, "lr": 3.27370547940519e-05}
{"train_loss": 0.0656389594078064, "global_step": 166378, "epoch": 1869, "lr": 3.27365107080739e-05}
{"train_loss": 0.09407296031713486, "global_step": 166379, "epoch": 1869, "lr": 3.273596662441678e-05}
{"train_loss": 0.05764077603816986, "global_step": 166380, "epoch": 1869, "lr": 3.273542254308057e-05}
{"train_loss": 0.09263455867767334, "global_step": 166381, "epoch": 1869, "lr": 3.273487846406541e-05}
{"train_loss": 0.07919628918170929, "global_step": 166382, "epoch": 1869, "lr": 3.27343343873713e-05}
{"train_loss": 0.11815563589334488, "global_step": 166383, "epoch": 1869, "lr": 3.273379031299835e-05}
{"train_loss": 0.05971862003207207, "global_step": 166384, "epoch": 1869, "lr": 3.273324624094666e-05}
{"train_loss": 0.14969663321971893, "global_step": 166385, "epoch": 1869, "lr": 3.273270217121624e-05}
{"train_loss": 0.11252237111330032, "global_step": 166386, "epoch": 1869, "lr": 3.273215810380723e-05}
{"train_loss": 0.10592962056398392, "global_step": 166387, "epoch": 1869, "lr": 3.2731614038719646e-05}
{"train_loss": 0.11300718784332275, "global_step": 166388, "epoch": 1869, "lr": 3.273106997595361e-05}
{"train_loss": 0.12712526321411133, "global_step": 166389, "epoch": 1869, "lr": 3.273052591550916e-05}
{"train_loss": 0.12704534828662872, "global_step": 166390, "epoch": 1869, "lr": 3.272998185738638e-05}
{"train_loss": 0.06066248565912247, "global_step": 166391, "epoch": 1869, "lr": 3.272943780158535e-05}
{"train_loss": 0.07912536710500717, "global_step": 166392, "epoch": 1869, "lr": 3.2728893748106125e-05}
{"train_loss": 0.1469041109085083, "global_step": 166393, "epoch": 1869, "lr": 3.272834969694879e-05}
{"train_loss": 0.1318764090538025, "global_step": 166394, "epoch": 1869, "lr": 3.2727805648113435e-05}
{"train_loss": 0.102158322930336, "global_step": 166395, "epoch": 1869, "lr": 3.272726160160009e-05}
{"train_loss": 0.07495468109846115, "global_step": 166396, "epoch": 1869, "lr": 3.272671755740888e-05}
{"train_loss": 0.08845703303813934, "global_step": 166397, "epoch": 1869, "lr": 3.2726173515539825e-05}
{"train_loss": 0.1595412641763687, "global_step": 166398, "epoch": 1869, "lr": 3.272562947599306e-05}
{"train_loss": 0.06388288736343384, "global_step": 166399, "epoch": 1869, "lr": 3.272508543876859e-05}
{"train_loss": 0.09226755797863007, "global_step": 166400, "epoch": 1869, "lr": 3.272454140386654e-05}
{"train_loss": 0.09324630349874496, "global_step": 166401, "epoch": 1869, "lr": 3.272399737128695e-05}
{"train_loss": 0.12919855117797852, "global_step": 166402, "epoch": 1869, "lr": 3.272345334102991e-05}
{"train_loss": 0.08988424390554428, "global_step": 166403, "epoch": 1869, "lr": 3.272290931309551e-05}
{"train_loss": 0.06751654297113419, "global_step": 166404, "epoch": 1869, "lr": 3.272236528748378e-05}
{"train_loss": 0.14736272394657135, "global_step": 166405, "epoch": 1869, "lr": 3.272182126419485e-05}
{"train_loss": 0.11097677052021027, "global_step": 166406, "epoch": 1869, "lr": 3.272127724322873e-05}
{"train_loss": 0.14880523085594177, "global_step": 166407, "epoch": 1869, "lr": 3.2720733224585545e-05}
{"train_loss": 0.15981066226959229, "global_step": 166408, "epoch": 1869, "lr": 3.272018920826533e-05}
{"train_loss": 0.13679973781108856, "global_step": 166409, "epoch": 1869, "lr": 3.2719645194268186e-05}
{"train_loss": 0.07205405086278915, "global_step": 166410, "epoch": 1869, "lr": 3.271910118259417e-05}
{"train_loss": 0.08850841969251633, "global_step": 166411, "epoch": 1869, "lr": 3.2718557173243355e-05}
{"train_loss": 0.17858147621154785, "global_step": 166412, "epoch": 1869, "lr": 3.271801316621583e-05}
{"train_loss": 0.07147549837827682, "global_step": 166413, "epoch": 1869, "lr": 3.271746916151167e-05}
{"train_loss": 0.10842885076999664, "global_step": 166414, "epoch": 1869, "lr": 3.271692515913091e-05}
{"train_loss": 0.07963430136442184, "global_step": 166415, "epoch": 1869, "lr": 3.271638115907366e-05}
{"train_loss": 0.05953635275363922, "global_step": 166416, "epoch": 1869, "lr": 3.271583716133997e-05}
{"train_loss": 0.1113293319940567, "global_step": 166417, "epoch": 1869, "lr": 3.271529316592994e-05}
{"train_loss": 0.06447510421276093, "global_step": 166418, "epoch": 1869, "lr": 3.271474917284362e-05}
{"train_loss": 0.07823816686868668, "global_step": 166419, "epoch": 1869, "lr": 3.271420518208108e-05}
{"train_loss": 0.08645718544721603, "global_step": 166420, "epoch": 1869, "lr": 3.271366119364243e-05}
{"train_loss": 0.11567037552595139, "global_step": 166421, "epoch": 1869, "lr": 3.27131172075277e-05}
{"train_loss": 0.10449089854955673, "global_step": 166422, "epoch": 1869, "lr": 3.271257322373699e-05}
{"train_loss": 0.07236823439598083, "global_step": 166423, "epoch": 1869, "lr": 3.271202924227035e-05}
{"train_loss": 0.05570356920361519, "global_step": 166424, "epoch": 1869, "lr": 3.271148526312789e-05}
{"train_loss": 0.038262248039245605, "global_step": 166425, "epoch": 1869, "lr": 3.2710941286309636e-05}
{"train_loss": 0.09618638455867767, "global_step": 166426, "epoch": 1869, "lr": 3.271039731181571e-05}
{"train_loss": 0.05503883212804794, "global_step": 166427, "epoch": 1869, "lr": 3.2709853339646146e-05}
{"train_loss": 0.06946681439876556, "global_step": 166428, "epoch": 1869, "lr": 3.270930936980103e-05}
{"train_loss": 0.09154950885960225, "global_step": 166429, "epoch": 1869, "lr": 3.270876540228044e-05, "val_loss": 7.051222801208496}
{"train_loss": 0.07629788666963577, "global_step": 166430, "epoch": 1870, "lr": 3.270822143708444e-05}
{"train_loss": 0.05028177425265312, "global_step": 166431, "epoch": 1870, "lr": 3.270767747421312e-05}
{"train_loss": 0.06719405949115753, "global_step": 166432, "epoch": 1870, "lr": 3.2707133513666543e-05}
{"train_loss": 0.10843530297279358, "global_step": 166433, "epoch": 1870, "lr": 3.270658955544477e-05}
{"train_loss": 0.16489574313163757, "global_step": 166434, "epoch": 1870, "lr": 3.270604559954789e-05}
{"train_loss": 0.10185054689645767, "global_step": 166435, "epoch": 1870, "lr": 3.270550164597596e-05}
{"train_loss": 0.07856889069080353, "global_step": 166436, "epoch": 1870, "lr": 3.2704957694729086e-05}
{"train_loss": 0.07807577401399612, "global_step": 166437, "epoch": 1870, "lr": 3.27044137458073e-05}
{"train_loss": 0.1014641523361206, "global_step": 166438, "epoch": 1870, "lr": 3.270386979921069e-05}
{"train_loss": 0.08047070354223251, "global_step": 166439, "epoch": 1870, "lr": 3.270332585493936e-05}
{"train_loss": 0.08667466789484024, "global_step": 166440, "epoch": 1870, "lr": 3.270278191299333e-05}
{"train_loss": 0.0875779315829277, "global_step": 166441, "epoch": 1870, "lr": 3.270223797337273e-05}
{"train_loss": 0.11308635026216507, "global_step": 166442, "epoch": 1870, "lr": 3.270169403607758e-05}
{"train_loss": 0.0888620913028717, "global_step": 166443, "epoch": 1870, "lr": 3.270115010110799e-05}
{"train_loss": 0.06193186342716217, "global_step": 166444, "epoch": 1870, "lr": 3.2700606168464e-05}
{"train_loss": 0.052246931940317154, "global_step": 166445, "epoch": 1870, "lr": 3.270006223814573e-05}
{"train_loss": 0.10356208682060242, "global_step": 166446, "epoch": 1870, "lr": 3.269951831015321e-05}
{"train_loss": 0.08117891103029251, "global_step": 166447, "epoch": 1870, "lr": 3.269897438448654e-05}
{"train_loss": 0.08582724630832672, "global_step": 166448, "epoch": 1870, "lr": 3.269843046114577e-05}
{"train_loss": 0.05332108587026596, "global_step": 166449, "epoch": 1870, "lr": 3.269788654013098e-05}
{"train_loss": 0.08834674209356308, "global_step": 166450, "epoch": 1870, "lr": 3.269734262144226e-05}
{"train_loss": 0.07156737148761749, "global_step": 166451, "epoch": 1870, "lr": 3.269679870507968e-05}
{"train_loss": 0.05313923582434654, "global_step": 166452, "epoch": 1870, "lr": 3.269625479104328e-05}
{"train_loss": 0.10409097373485565, "global_step": 166453, "epoch": 1870, "lr": 3.269571087933318e-05}
{"train_loss": 0.06276963651180267, "global_step": 166454, "epoch": 1870, "lr": 3.269516696994941e-05}
{"train_loss": 0.07400030642747879, "global_step": 166455, "epoch": 1870, "lr": 3.269462306289206e-05}
{"train_loss": 0.0569891631603241, "global_step": 166456, "epoch": 1870, "lr": 3.2694079158161236e-05}
{"train_loss": 0.06727053225040436, "global_step": 166457, "epoch": 1870, "lr": 3.269353525575696e-05}
{"train_loss": 0.09560571610927582, "global_step": 166458, "epoch": 1870, "lr": 3.269299135567934e-05}
{"train_loss": 0.1006288230419159, "global_step": 166459, "epoch": 1870, "lr": 3.269244745792842e-05}
{"train_loss": 0.060820359736680984, "global_step": 166460, "epoch": 1870, "lr": 3.2691903562504314e-05}
{"train_loss": 0.09165281802415848, "global_step": 166461, "epoch": 1870, "lr": 3.2691359669407046e-05}
{"train_loss": 0.13722382485866547, "global_step": 166462, "epoch": 1870, "lr": 3.2690815778636734e-05}
{"train_loss": 0.07879720628261566, "global_step": 166463, "epoch": 1870, "lr": 3.269027189019341e-05}
{"train_loss": 0.036549463868141174, "global_step": 166464, "epoch": 1870, "lr": 3.268972800407719e-05}
{"train_loss": 0.0456584133207798, "global_step": 166465, "epoch": 1870, "lr": 3.268918412028812e-05}
{"train_loss": 0.0887458547949791, "global_step": 166466, "epoch": 1870, "lr": 3.268864023882627e-05}
{"train_loss": 0.11183352768421173, "global_step": 166467, "epoch": 1870, "lr": 3.2688096359691727e-05}
{"train_loss": 0.07959403097629547, "global_step": 166468, "epoch": 1870, "lr": 3.2687552482884556e-05}
{"train_loss": 0.07876526564359665, "global_step": 166469, "epoch": 1870, "lr": 3.2687008608404836e-05}
{"train_loss": 0.12717345356941223, "global_step": 166470, "epoch": 1870, "lr": 3.268646473625264e-05}
{"train_loss": 0.08341766148805618, "global_step": 166471, "epoch": 1870, "lr": 3.268592086642802e-05}
{"train_loss": 0.10140098631381989, "global_step": 166472, "epoch": 1870, "lr": 3.2685376998931086e-05}
{"train_loss": 0.08594554662704468, "global_step": 166473, "epoch": 1870, "lr": 3.268483313376188e-05}
{"train_loss": 0.10666991770267487, "global_step": 166474, "epoch": 1870, "lr": 3.268428927092048e-05}
{"train_loss": 0.08011143654584885, "global_step": 166475, "epoch": 1870, "lr": 3.268374541040698e-05}
{"train_loss": 0.08671163767576218, "global_step": 166476, "epoch": 1870, "lr": 3.268320155222143e-05}
{"train_loss": 0.12088102847337723, "global_step": 166477, "epoch": 1870, "lr": 3.2682657696363926e-05}
{"train_loss": 0.09611956030130386, "global_step": 166478, "epoch": 1870, "lr": 3.2682113842834516e-05}
{"train_loss": 0.055492643266916275, "global_step": 166479, "epoch": 1870, "lr": 3.26815699916333e-05}
{"train_loss": 0.050787851214408875, "global_step": 166480, "epoch": 1870, "lr": 3.2681026142760314e-05}
{"train_loss": 0.1513480544090271, "global_step": 166481, "epoch": 1870, "lr": 3.268048229621567e-05}
{"train_loss": 0.07170270383358002, "global_step": 166482, "epoch": 1870, "lr": 3.2679938451999416e-05}
{"train_loss": 0.10084465146064758, "global_step": 166483, "epoch": 1870, "lr": 3.267939461011164e-05}
{"train_loss": 0.04066639393568039, "global_step": 166484, "epoch": 1870, "lr": 3.2678850770552404e-05}
{"train_loss": 0.08970507234334946, "global_step": 166485, "epoch": 1870, "lr": 3.267830693332179e-05}
{"train_loss": 0.13448566198349, "global_step": 166486, "epoch": 1870, "lr": 3.2677763098419854e-05}
{"train_loss": 0.07701315730810165, "global_step": 166487, "epoch": 1870, "lr": 3.267721926584669e-05}
{"train_loss": 0.03956640139222145, "global_step": 166488, "epoch": 1870, "lr": 3.267667543560236e-05}
{"train_loss": 0.13440760970115662, "global_step": 166489, "epoch": 1870, "lr": 3.267613160768695e-05}
{"train_loss": 0.1399219036102295, "global_step": 166490, "epoch": 1870, "lr": 3.267558778210051e-05}
{"train_loss": 0.15901829302310944, "global_step": 166491, "epoch": 1870, "lr": 3.267504395884312e-05}
{"train_loss": 0.07955268025398254, "global_step": 166492, "epoch": 1870, "lr": 3.267450013791489e-05}
{"train_loss": 0.05438512936234474, "global_step": 166493, "epoch": 1870, "lr": 3.267395631931582e-05}
{"train_loss": 0.11988894641399384, "global_step": 166494, "epoch": 1870, "lr": 3.267341250304606e-05}
{"train_loss": 0.09275155514478683, "global_step": 166495, "epoch": 1870, "lr": 3.267286868910563e-05}
{"train_loss": 0.08219203352928162, "global_step": 166496, "epoch": 1870, "lr": 3.2672324877494645e-05}
{"train_loss": 0.09974587708711624, "global_step": 166497, "epoch": 1870, "lr": 3.267178106821312e-05}
{"train_loss": 0.10573101788759232, "global_step": 166498, "epoch": 1870, "lr": 3.267123726126119e-05}
{"train_loss": 0.11269290745258331, "global_step": 166499, "epoch": 1870, "lr": 3.267069345663888e-05}
{"train_loss": 0.15220481157302856, "global_step": 166500, "epoch": 1870, "lr": 3.2670149654346306e-05}
{"train_loss": 0.07924994826316833, "global_step": 166501, "epoch": 1870, "lr": 3.26696058543835e-05}
{"train_loss": 0.06378988921642303, "global_step": 166502, "epoch": 1870, "lr": 3.2669062056750575e-05}
{"train_loss": 0.0942986011505127, "global_step": 166503, "epoch": 1870, "lr": 3.266851826144756e-05}
{"train_loss": 0.18582801520824432, "global_step": 166504, "epoch": 1870, "lr": 3.266797446847457e-05}
{"train_loss": 0.12171908468008041, "global_step": 166505, "epoch": 1870, "lr": 3.266743067783166e-05}
{"train_loss": 0.10920208692550659, "global_step": 166506, "epoch": 1870, "lr": 3.2666886889518886e-05}
{"train_loss": 0.10617614537477493, "global_step": 166507, "epoch": 1870, "lr": 3.2666343103536354e-05}
{"train_loss": 0.07265811413526535, "global_step": 166508, "epoch": 1870, "lr": 3.266579931988411e-05}
{"train_loss": 0.06734031438827515, "global_step": 166509, "epoch": 1870, "lr": 3.266525553856225e-05}
{"train_loss": 0.11200007796287537, "global_step": 166510, "epoch": 1870, "lr": 3.266471175957081e-05}
{"train_loss": 0.11052881181240082, "global_step": 166511, "epoch": 1870, "lr": 3.2664167982909915e-05}
{"train_loss": 0.1163179874420166, "global_step": 166512, "epoch": 1870, "lr": 3.266362420857959e-05}
{"train_loss": 0.06410206854343414, "global_step": 166513, "epoch": 1870, "lr": 3.266308043657995e-05}
{"train_loss": 0.08001061528921127, "global_step": 166514, "epoch": 1870, "lr": 3.2662536666911025e-05}
{"train_loss": 0.09378114342689514, "global_step": 166515, "epoch": 1870, "lr": 3.2661992899572935e-05}
{"train_loss": 0.06367939710617065, "global_step": 166516, "epoch": 1870, "lr": 3.266144913456571e-05}
{"train_loss": 0.09677483886480331, "global_step": 166517, "epoch": 1870, "lr": 3.266090537188946e-05}
{"train_loss": 0.09110244690032487, "global_step": 166518, "epoch": 1870, "lr": 3.266036161154422e-05, "val_loss": 7.130810260772705, "train_action_mse_error": 26.78598403930664}
{"train_loss": 0.1152842715382576, "global_step": 166519, "epoch": 1871, "lr": 3.2659817853530105e-05}
{"train_loss": 0.10087758302688599, "global_step": 166520, "epoch": 1871, "lr": 3.265927409784714e-05}
{"train_loss": 0.1154583990573883, "global_step": 166521, "epoch": 1871, "lr": 3.265873034449545e-05}
{"train_loss": 0.051385242491960526, "global_step": 166522, "epoch": 1871, "lr": 3.265818659347506e-05}
{"train_loss": 0.08414024859666824, "global_step": 166523, "epoch": 1871, "lr": 3.2657642844786086e-05}
{"train_loss": 0.044178761541843414, "global_step": 166524, "epoch": 1871, "lr": 3.265709909842857e-05}
{"train_loss": 0.06880320608615875, "global_step": 166525, "epoch": 1871, "lr": 3.2656555354402595e-05}
{"train_loss": 0.08002590388059616, "global_step": 166526, "epoch": 1871, "lr": 3.265601161270824e-05}
{"train_loss": 0.07787859439849854, "global_step": 166527, "epoch": 1871, "lr": 3.265546787334557e-05}
{"train_loss": 0.056323256343603134, "global_step": 166528, "epoch": 1871, "lr": 3.2654924136314656e-05}
{"train_loss": 0.034098245203495026, "global_step": 166529, "epoch": 1871, "lr": 3.265438040161557e-05}
{"train_loss": 0.08020259439945221, "global_step": 166530, "epoch": 1871, "lr": 3.26538366692484e-05}
{"train_loss": 0.06825492531061172, "global_step": 166531, "epoch": 1871, "lr": 3.265329293921321e-05}
{"train_loss": 0.10332324355840683, "global_step": 166532, "epoch": 1871, "lr": 3.2652749211510084e-05}
{"train_loss": 0.08431147038936615, "global_step": 166533, "epoch": 1871, "lr": 3.265220548613906e-05}
{"train_loss": 0.0800662711262703, "global_step": 166534, "epoch": 1871, "lr": 3.265166176310026e-05}
{"train_loss": 0.06564555317163467, "global_step": 166535, "epoch": 1871, "lr": 3.2651118042393713e-05}
{"train_loss": 0.09657079726457596, "global_step": 166536, "epoch": 1871, "lr": 3.2650574324019536e-05}
{"train_loss": 0.09721715748310089, "global_step": 166537, "epoch": 1871, "lr": 3.2650030607977755e-05}
{"train_loss": 0.06776808202266693, "global_step": 166538, "epoch": 1871, "lr": 3.264948689426848e-05}
{"train_loss": 0.07264024019241333, "global_step": 166539, "epoch": 1871, "lr": 3.2648943182891757e-05}
{"train_loss": 0.039743464440107346, "global_step": 166540, "epoch": 1871, "lr": 3.264839947384769e-05}
{"train_loss": 0.11289834976196289, "global_step": 166541, "epoch": 1871, "lr": 3.2647855767136314e-05}
{"train_loss": 0.1051725521683693, "global_step": 166542, "epoch": 1871, "lr": 3.264731206275774e-05}
{"train_loss": 0.10617908835411072, "global_step": 166543, "epoch": 1871, "lr": 3.2646768360712016e-05}
{"train_loss": 0.06098596379160881, "global_step": 166544, "epoch": 1871, "lr": 3.264622466099922e-05}
{"train_loss": 0.09384629875421524, "global_step": 166545, "epoch": 1871, "lr": 3.2645680963619426e-05}
{"train_loss": 0.05215892568230629, "global_step": 166546, "epoch": 1871, "lr": 3.2645137268572706e-05}
{"train_loss": 0.07695869356393814, "global_step": 166547, "epoch": 1871, "lr": 3.264459357585915e-05}
{"train_loss": 0.1054900586605072, "global_step": 166548, "epoch": 1871, "lr": 3.26440498854788e-05}
{"train_loss": 0.08158125728368759, "global_step": 166549, "epoch": 1871, "lr": 3.264350619743176e-05}
{"train_loss": 0.11405804008245468, "global_step": 166550, "epoch": 1871, "lr": 3.264296251171808e-05}
{"train_loss": 0.12377635389566422, "global_step": 166551, "epoch": 1871, "lr": 3.264241882833785e-05}
{"train_loss": 0.051215190440416336, "global_step": 166552, "epoch": 1871, "lr": 3.264187514729112e-05}
{"train_loss": 0.11478754878044128, "global_step": 166553, "epoch": 1871, "lr": 3.264133146857801e-05}
{"train_loss": 0.10745257884263992, "global_step": 166554, "epoch": 1871, "lr": 3.2640787792198527e-05}
{"train_loss": 0.1334063857793808, "global_step": 166555, "epoch": 1871, "lr": 3.26402441181528e-05}
{"train_loss": 0.07460128515958786, "global_step": 166556, "epoch": 1871, "lr": 3.263970044644086e-05}
{"train_loss": 0.07527025789022446, "global_step": 166557, "epoch": 1871, "lr": 3.263915677706283e-05}
{"train_loss": 0.0857628658413887, "global_step": 166558, "epoch": 1871, "lr": 3.263861311001872e-05}
{"train_loss": 0.12458404898643494, "global_step": 166559, "epoch": 1871, "lr": 3.263806944530867e-05}
{"train_loss": 0.027135111391544342, "global_step": 166560, "epoch": 1871, "lr": 3.2637525782932694e-05}
{"train_loss": 0.06822428852319717, "global_step": 166561, "epoch": 1871, "lr": 3.263698212289091e-05}
{"train_loss": 0.06393080204725266, "global_step": 166562, "epoch": 1871, "lr": 3.263643846518336e-05}
{"train_loss": 0.0838456004858017, "global_step": 166563, "epoch": 1871, "lr": 3.263589480981013e-05}
{"train_loss": 0.16686764359474182, "global_step": 166564, "epoch": 1871, "lr": 3.26353511567713e-05}
{"train_loss": 0.02557722106575966, "global_step": 166565, "epoch": 1871, "lr": 3.263480750606692e-05}
{"train_loss": 0.11926306039094925, "global_step": 166566, "epoch": 1871, "lr": 3.26342638576971e-05}
{"train_loss": 0.07484181970357895, "global_step": 166567, "epoch": 1871, "lr": 3.2633720211661865e-05}
{"train_loss": 0.07241883128881454, "global_step": 166568, "epoch": 1871, "lr": 3.2633176567961347e-05}
{"train_loss": 0.07794156670570374, "global_step": 166569, "epoch": 1871, "lr": 3.263263292659555e-05}
{"train_loss": 0.08229213953018188, "global_step": 166570, "epoch": 1871, "lr": 3.263208928756462e-05}
{"train_loss": 0.05174756795167923, "global_step": 166571, "epoch": 1871, "lr": 3.2631545650868564e-05}
{"train_loss": 0.0929023027420044, "global_step": 166572, "epoch": 1871, "lr": 3.263100201650751e-05}
{"train_loss": 0.12061599642038345, "global_step": 166573, "epoch": 1871, "lr": 3.2630458384481486e-05}
{"train_loss": 0.10375174134969711, "global_step": 166574, "epoch": 1871, "lr": 3.26299147547906e-05}
{"train_loss": 0.11245465278625488, "global_step": 166575, "epoch": 1871, "lr": 3.262937112743489e-05}
{"train_loss": 0.09398355334997177, "global_step": 166576, "epoch": 1871, "lr": 3.2628827502414485e-05}
{"train_loss": 0.1544778048992157, "global_step": 166577, "epoch": 1871, "lr": 3.262828387972938e-05}
{"train_loss": 0.08917169272899628, "global_step": 166578, "epoch": 1871, "lr": 3.2627740259379726e-05}
{"train_loss": 0.0664609968662262, "global_step": 166579, "epoch": 1871, "lr": 3.2627196641365535e-05}
{"train_loss": 0.11722952127456665, "global_step": 166580, "epoch": 1871, "lr": 3.262665302568693e-05}
{"train_loss": 0.19737258553504944, "global_step": 166581, "epoch": 1871, "lr": 3.262610941234394e-05}
{"train_loss": 0.06144469976425171, "global_step": 166582, "epoch": 1871, "lr": 3.262556580133666e-05}
{"train_loss": 0.0400562658905983, "global_step": 166583, "epoch": 1871, "lr": 3.262502219266517e-05}
{"train_loss": 0.11619783937931061, "global_step": 166584, "epoch": 1871, "lr": 3.262447858632952e-05}
{"train_loss": 0.09739341586828232, "global_step": 166585, "epoch": 1871, "lr": 3.2623934982329816e-05}
{"train_loss": 0.13215918838977814, "global_step": 166586, "epoch": 1871, "lr": 3.262339138066609e-05}
{"train_loss": 0.14620763063430786, "global_step": 166587, "epoch": 1871, "lr": 3.262284778133846e-05}
{"train_loss": 0.06820482760667801, "global_step": 166588, "epoch": 1871, "lr": 3.2622304184346955e-05}
{"train_loss": 0.0797303095459938, "global_step": 166589, "epoch": 1871, "lr": 3.26217605896917e-05}
{"train_loss": 0.09020369499921799, "global_step": 166590, "epoch": 1871, "lr": 3.26212169973727e-05}
{"train_loss": 0.1011534184217453, "global_step": 166591, "epoch": 1871, "lr": 3.2620673407390095e-05}
{"train_loss": 0.08444862812757492, "global_step": 166592, "epoch": 1871, "lr": 3.2620129819743906e-05}
{"train_loss": 0.047465987503528595, "global_step": 166593, "epoch": 1871, "lr": 3.2619586234434254e-05}
{"train_loss": 0.0694097951054573, "global_step": 166594, "epoch": 1871, "lr": 3.261904265146116e-05}
{"train_loss": 0.13102005422115326, "global_step": 166595, "epoch": 1871, "lr": 3.2618499070824745e-05}
{"train_loss": 0.06500248610973358, "global_step": 166596, "epoch": 1871, "lr": 3.261795549252504e-05}
{"train_loss": 0.10650106519460678, "global_step": 166597, "epoch": 1871, "lr": 3.261741191656216e-05}
{"train_loss": 0.16816461086273193, "global_step": 166598, "epoch": 1871, "lr": 3.261686834293614e-05}
{"train_loss": 0.10283888131380081, "global_step": 166599, "epoch": 1871, "lr": 3.261632477164709e-05}
{"train_loss": 0.10446127504110336, "global_step": 166600, "epoch": 1871, "lr": 3.2615781202695054e-05}
{"train_loss": 0.09079979360103607, "global_step": 166601, "epoch": 1871, "lr": 3.26152376360801e-05}
{"train_loss": 0.1590200513601303, "global_step": 166602, "epoch": 1871, "lr": 3.2614694071802334e-05}
{"train_loss": 0.10173243284225464, "global_step": 166603, "epoch": 1871, "lr": 3.26141505098618e-05}
{"train_loss": 0.10938488692045212, "global_step": 166604, "epoch": 1871, "lr": 3.261360695025859e-05}
{"train_loss": 0.08119335770606995, "global_step": 166605, "epoch": 1871, "lr": 3.2613063392992754e-05}
{"train_loss": 0.0798657163977623, "global_step": 166606, "epoch": 1871, "lr": 3.2612519838064405e-05}
{"train_loss": 0.09128459078374873, "global_step": 166607, "epoch": 1871, "lr": 3.261197628547356e-05, "val_loss": 7.024694919586182}
{"train_loss": 0.15756633877754211, "global_step": 166608, "epoch": 1872, "lr": 3.261143273522035e-05}
{"train_loss": 0.08414161205291748, "global_step": 166609, "epoch": 1872, "lr": 3.2610889187304795e-05}
{"train_loss": 0.13547734916210175, "global_step": 166610, "epoch": 1872, "lr": 3.2610345641727026e-05}
{"train_loss": 0.08636302500963211, "global_step": 166611, "epoch": 1872, "lr": 3.2609802098487055e-05}
{"train_loss": 0.06145143881440163, "global_step": 166612, "epoch": 1872, "lr": 3.2609258557584996e-05}
{"train_loss": 0.09058079868555069, "global_step": 166613, "epoch": 1872, "lr": 3.260871501902091e-05}
{"train_loss": 0.10543931275606155, "global_step": 166614, "epoch": 1872, "lr": 3.260817148279487e-05}
{"train_loss": 0.08389666676521301, "global_step": 166615, "epoch": 1872, "lr": 3.2607627948906945e-05}
{"train_loss": 0.05056636407971382, "global_step": 166616, "epoch": 1872, "lr": 3.260708441735723e-05}
{"train_loss": 0.07143368571996689, "global_step": 166617, "epoch": 1872, "lr": 3.2606540888145754e-05}
{"train_loss": 0.08883124589920044, "global_step": 166618, "epoch": 1872, "lr": 3.2605997361272635e-05}
{"train_loss": 0.056629158556461334, "global_step": 166619, "epoch": 1872, "lr": 3.260545383673792e-05}
{"train_loss": 0.10781153291463852, "global_step": 166620, "epoch": 1872, "lr": 3.260491031454169e-05}
{"train_loss": 0.19086317718029022, "global_step": 166621, "epoch": 1872, "lr": 3.260436679468403e-05}
{"train_loss": 0.0953768864274025, "global_step": 166622, "epoch": 1872, "lr": 3.260382327716498e-05}
{"train_loss": 0.10240649431943893, "global_step": 166623, "epoch": 1872, "lr": 3.2603279761984655e-05}
{"train_loss": 0.034527137875556946, "global_step": 166624, "epoch": 1872, "lr": 3.2602736249143083e-05}
{"train_loss": 0.0378783643245697, "global_step": 166625, "epoch": 1872, "lr": 3.260219273864039e-05}
{"train_loss": 0.08940677344799042, "global_step": 166626, "epoch": 1872, "lr": 3.260164923047659e-05}
{"train_loss": 0.060057133436203, "global_step": 166627, "epoch": 1872, "lr": 3.260110572465181e-05}
{"train_loss": 0.11930727958679199, "global_step": 166628, "epoch": 1872, "lr": 3.2600562221166086e-05}
{"train_loss": 0.059448376297950745, "global_step": 166629, "epoch": 1872, "lr": 3.260001872001952e-05}
{"train_loss": 0.09464038163423538, "global_step": 166630, "epoch": 1872, "lr": 3.259947522121215e-05}
{"train_loss": 0.08538895100355148, "global_step": 166631, "epoch": 1872, "lr": 3.2598931724744085e-05}
{"train_loss": 0.08766161650419235, "global_step": 166632, "epoch": 1872, "lr": 3.259838823061536e-05}
{"train_loss": 0.05114596709609032, "global_step": 166633, "epoch": 1872, "lr": 3.25978447388261e-05}
{"train_loss": 0.08373177796602249, "global_step": 166634, "epoch": 1872, "lr": 3.2597301249376315e-05}
{"train_loss": 0.07444094866514206, "global_step": 166635, "epoch": 1872, "lr": 3.259675776226614e-05}
{"train_loss": 0.07545749098062515, "global_step": 166636, "epoch": 1872, "lr": 3.259621427749559e-05}
{"train_loss": 0.1466645449399948, "global_step": 166637, "epoch": 1872, "lr": 3.259567079506478e-05}
{"train_loss": 0.056895650923252106, "global_step": 166638, "epoch": 1872, "lr": 3.259512731497378e-05}
{"train_loss": 0.04220684990286827, "global_step": 166639, "epoch": 1872, "lr": 3.259458383722264e-05}
{"train_loss": 0.0902012437582016, "global_step": 166640, "epoch": 1872, "lr": 3.259404036181146e-05}
{"train_loss": 0.09211377054452896, "global_step": 166641, "epoch": 1872, "lr": 3.259349688874028e-05}
{"train_loss": 0.05253018066287041, "global_step": 166642, "epoch": 1872, "lr": 3.259295341800922e-05}
{"train_loss": 0.10493526607751846, "global_step": 166643, "epoch": 1872, "lr": 3.2592409949618295e-05}
{"train_loss": 0.058039527386426926, "global_step": 166644, "epoch": 1872, "lr": 3.259186648356763e-05}
{"train_loss": 0.0805794969201088, "global_step": 166645, "epoch": 1872, "lr": 3.2591323019857255e-05}
{"train_loss": 0.0674986019730568, "global_step": 166646, "epoch": 1872, "lr": 3.2590779558487287e-05}
{"train_loss": 0.09701661765575409, "global_step": 166647, "epoch": 1872, "lr": 3.259023609945776e-05}
{"train_loss": 0.07890249788761139, "global_step": 166648, "epoch": 1872, "lr": 3.258969264276878e-05}
{"train_loss": 0.08044189214706421, "global_step": 166649, "epoch": 1872, "lr": 3.258914918842038e-05}
{"train_loss": 0.10610499978065491, "global_step": 166650, "epoch": 1872, "lr": 3.258860573641268e-05}
{"train_loss": 0.04710360988974571, "global_step": 166651, "epoch": 1872, "lr": 3.2588062286745715e-05}
{"train_loss": 0.042651426047086716, "global_step": 166652, "epoch": 1872, "lr": 3.258751883941959e-05}
{"train_loss": 0.04678570479154587, "global_step": 166653, "epoch": 1872, "lr": 3.2586975394434326e-05}
{"train_loss": 0.08431312441825867, "global_step": 166654, "epoch": 1872, "lr": 3.258643195179007e-05}
{"train_loss": 0.08986780047416687, "global_step": 166655, "epoch": 1872, "lr": 3.258588851148683e-05}
{"train_loss": 0.0645304024219513, "global_step": 166656, "epoch": 1872, "lr": 3.258534507352471e-05}
{"train_loss": 0.09947246313095093, "global_step": 166657, "epoch": 1872, "lr": 3.258480163790378e-05}
{"train_loss": 0.05496193468570709, "global_step": 166658, "epoch": 1872, "lr": 3.2584258204624116e-05}
{"train_loss": 0.10823342949151993, "global_step": 166659, "epoch": 1872, "lr": 3.25837147736858e-05}
{"train_loss": 0.0847863033413887, "global_step": 166660, "epoch": 1872, "lr": 3.258317134508886e-05}
{"train_loss": 0.08826097846031189, "global_step": 166661, "epoch": 1872, "lr": 3.258262791883343e-05}
{"train_loss": 0.11954431235790253, "global_step": 166662, "epoch": 1872, "lr": 3.258208449491953e-05}
{"train_loss": 0.08855581283569336, "global_step": 166663, "epoch": 1872, "lr": 3.258154107334727e-05}
{"train_loss": 0.0949593260884285, "global_step": 166664, "epoch": 1872, "lr": 3.25809976541167e-05}
{"train_loss": 0.09971671551465988, "global_step": 166665, "epoch": 1872, "lr": 3.258045423722791e-05}
{"train_loss": 0.13723896443843842, "global_step": 166666, "epoch": 1872, "lr": 3.257991082268096e-05}
{"train_loss": 0.0723068043589592, "global_step": 166667, "epoch": 1872, "lr": 3.257936741047594e-05}
{"train_loss": 0.0851757600903511, "global_step": 166668, "epoch": 1872, "lr": 3.257882400061289e-05}
{"train_loss": 0.07729475945234299, "global_step": 166669, "epoch": 1872, "lr": 3.257828059309193e-05}
{"train_loss": 0.11098545789718628, "global_step": 166670, "epoch": 1872, "lr": 3.257773718791308e-05}
{"train_loss": 0.09233932942152023, "global_step": 166671, "epoch": 1872, "lr": 3.2577193785076476e-05}
{"train_loss": 0.13791483640670776, "global_step": 166672, "epoch": 1872, "lr": 3.257665038458212e-05}
{"train_loss": 0.08448789268732071, "global_step": 166673, "epoch": 1872, "lr": 3.2576106986430146e-05}
{"train_loss": 0.08898857980966568, "global_step": 166674, "epoch": 1872, "lr": 3.257556359062059e-05}
{"train_loss": 0.059150949120521545, "global_step": 166675, "epoch": 1872, "lr": 3.257502019715354e-05}
{"train_loss": 0.04693465679883957, "global_step": 166676, "epoch": 1872, "lr": 3.2574476806029064e-05}
{"train_loss": 0.08861155807971954, "global_step": 166677, "epoch": 1872, "lr": 3.2573933417247244e-05}
{"train_loss": 0.09424880146980286, "global_step": 166678, "epoch": 1872, "lr": 3.257339003080815e-05}
{"train_loss": 0.0745219737291336, "global_step": 166679, "epoch": 1872, "lr": 3.257284664671184e-05}
{"train_loss": 0.06545946002006531, "global_step": 166680, "epoch": 1872, "lr": 3.257230326495841e-05}
{"train_loss": 0.09888967126607895, "global_step": 166681, "epoch": 1872, "lr": 3.25717598855479e-05}
{"train_loss": 0.1582842618227005, "global_step": 166682, "epoch": 1872, "lr": 3.257121650848043e-05}
{"train_loss": 0.127589151263237, "global_step": 166683, "epoch": 1872, "lr": 3.2570673133756026e-05}
{"train_loss": 0.07706767320632935, "global_step": 166684, "epoch": 1872, "lr": 3.2570129761374804e-05}
{"train_loss": 0.09880126267671585, "global_step": 166685, "epoch": 1872, "lr": 3.256958639133679e-05}
{"train_loss": 0.06734207272529602, "global_step": 166686, "epoch": 1872, "lr": 3.256904302364211e-05}
{"train_loss": 0.08771426230669022, "global_step": 166687, "epoch": 1872, "lr": 3.256849965829079e-05}
{"train_loss": 0.036716606467962265, "global_step": 166688, "epoch": 1872, "lr": 3.256795629528293e-05}
{"train_loss": 0.11755703389644623, "global_step": 166689, "epoch": 1872, "lr": 3.256741293461859e-05}
{"train_loss": 0.1508709192276001, "global_step": 166690, "epoch": 1872, "lr": 3.256686957629786e-05}
{"train_loss": 0.07404717803001404, "global_step": 166691, "epoch": 1872, "lr": 3.256632622032079e-05}
{"train_loss": 0.07356089353561401, "global_step": 166692, "epoch": 1872, "lr": 3.256578286668747e-05}
{"train_loss": 0.08726052194833755, "global_step": 166693, "epoch": 1872, "lr": 3.2565239515397964e-05}
{"train_loss": 0.09157314151525497, "global_step": 166694, "epoch": 1872, "lr": 3.2564696166452355e-05}
{"train_loss": 0.08275483548641205, "global_step": 166695, "epoch": 1872, "lr": 3.2564152819850716e-05}
{"train_loss": 0.08668795441476147, "global_step": 166696, "epoch": 1872, "lr": 3.25636094755931e-05, "val_loss": 7.143988609313965}
{"train_loss": 0.11082496494054794, "global_step": 166697, "epoch": 1873, "lr": 3.256306613367961e-05}
{"train_loss": 0.06597176939249039, "global_step": 166698, "epoch": 1873, "lr": 3.256252279411029e-05}
{"train_loss": 0.050530340522527695, "global_step": 166699, "epoch": 1873, "lr": 3.256197945688524e-05}
{"train_loss": 0.07491482049226761, "global_step": 166700, "epoch": 1873, "lr": 3.25614361220045e-05}
{"train_loss": 0.08167871087789536, "global_step": 166701, "epoch": 1873, "lr": 3.2560892789468175e-05}
{"train_loss": 0.06715817749500275, "global_step": 166702, "epoch": 1873, "lr": 3.2560349459276315e-05}
{"train_loss": 0.0731910914182663, "global_step": 166703, "epoch": 1873, "lr": 3.255980613142902e-05}
{"train_loss": 0.07043841481208801, "global_step": 166704, "epoch": 1873, "lr": 3.2559262805926324e-05}
{"train_loss": 0.11554659157991409, "global_step": 166705, "epoch": 1873, "lr": 3.255871948276835e-05}
{"train_loss": 0.06003323942422867, "global_step": 166706, "epoch": 1873, "lr": 3.255817616195512e-05}
{"train_loss": 0.05741884931921959, "global_step": 166707, "epoch": 1873, "lr": 3.2557632843486755e-05}
{"train_loss": 0.11083538830280304, "global_step": 166708, "epoch": 1873, "lr": 3.255708952736327e-05}
{"train_loss": 0.06600570678710938, "global_step": 166709, "epoch": 1873, "lr": 3.2556546213584784e-05}
{"train_loss": 0.02975120209157467, "global_step": 166710, "epoch": 1873, "lr": 3.255600290215137e-05}
{"train_loss": 0.07416095584630966, "global_step": 166711, "epoch": 1873, "lr": 3.2555459593063086e-05}
{"train_loss": 0.08332639187574387, "global_step": 166712, "epoch": 1873, "lr": 3.2554916286320005e-05}
{"train_loss": 0.08141803741455078, "global_step": 166713, "epoch": 1873, "lr": 3.255437298192221e-05}
{"train_loss": 0.07191669195890427, "global_step": 166714, "epoch": 1873, "lr": 3.2553829679869754e-05}
{"train_loss": 0.07595078647136688, "global_step": 166715, "epoch": 1873, "lr": 3.2553286380162726e-05}
{"train_loss": 0.01570791006088257, "global_step": 166716, "epoch": 1873, "lr": 3.2552743082801206e-05}
{"train_loss": 0.11457724869251251, "global_step": 166717, "epoch": 1873, "lr": 3.255219978778525e-05}
{"train_loss": 0.09190242737531662, "global_step": 166718, "epoch": 1873, "lr": 3.2551656495114944e-05}
{"train_loss": 0.06132319197058678, "global_step": 166719, "epoch": 1873, "lr": 3.255111320479034e-05}
{"train_loss": 0.08743767440319061, "global_step": 166720, "epoch": 1873, "lr": 3.255056991681154e-05}
{"train_loss": 0.09701035171747208, "global_step": 166721, "epoch": 1873, "lr": 3.25500266311786e-05}
{"train_loss": 0.11511678993701935, "global_step": 166722, "epoch": 1873, "lr": 3.2549483347891594e-05}
{"train_loss": 0.11984545737504959, "global_step": 166723, "epoch": 1873, "lr": 3.2548940066950594e-05}
{"train_loss": 0.09517424553632736, "global_step": 166724, "epoch": 1873, "lr": 3.254839678835569e-05}
{"train_loss": 0.09535126388072968, "global_step": 166725, "epoch": 1873, "lr": 3.2547853512106916e-05}
{"train_loss": 0.10401058942079544, "global_step": 166726, "epoch": 1873, "lr": 3.25473102382044e-05}
{"train_loss": 0.08347876369953156, "global_step": 166727, "epoch": 1873, "lr": 3.254676696664816e-05}
{"train_loss": 0.11857030540704727, "global_step": 166728, "epoch": 1873, "lr": 3.2546223697438296e-05}
{"train_loss": 0.06969628483057022, "global_step": 166729, "epoch": 1873, "lr": 3.25456804305749e-05}
{"train_loss": 0.08270348608493805, "global_step": 166730, "epoch": 1873, "lr": 3.2545137166058006e-05}
{"train_loss": 0.1113380566239357, "global_step": 166731, "epoch": 1873, "lr": 3.2544593903887724e-05}
{"train_loss": 0.0915428027510643, "global_step": 166732, "epoch": 1873, "lr": 3.2544050644064096e-05}
{"train_loss": 0.11849471926689148, "global_step": 166733, "epoch": 1873, "lr": 3.2543507386587216e-05}
{"train_loss": 0.05867578089237213, "global_step": 166734, "epoch": 1873, "lr": 3.254296413145714e-05}
{"train_loss": 0.05915023759007454, "global_step": 166735, "epoch": 1873, "lr": 3.2542420878673955e-05}
{"train_loss": 0.16386011242866516, "global_step": 166736, "epoch": 1873, "lr": 3.2541877628237715e-05}
{"train_loss": 0.052456069737672806, "global_step": 166737, "epoch": 1873, "lr": 3.2541334380148535e-05}
{"train_loss": 0.06363874673843384, "global_step": 166738, "epoch": 1873, "lr": 3.254079113440643e-05}
{"train_loss": 0.08739201724529266, "global_step": 166739, "epoch": 1873, "lr": 3.254024789101152e-05}
{"train_loss": 0.0626753568649292, "global_step": 166740, "epoch": 1873, "lr": 3.2539704649963855e-05}
{"train_loss": 0.10087854415178299, "global_step": 166741, "epoch": 1873, "lr": 3.253916141126352e-05}
{"train_loss": 0.08661719411611557, "global_step": 166742, "epoch": 1873, "lr": 3.2538618174910565e-05}
{"train_loss": 0.07380669564008713, "global_step": 166743, "epoch": 1873, "lr": 3.2538074940905104e-05}
{"train_loss": 0.13645926117897034, "global_step": 166744, "epoch": 1873, "lr": 3.2537531709247165e-05}
{"train_loss": 0.07641903311014175, "global_step": 166745, "epoch": 1873, "lr": 3.2536988479936834e-05}
{"train_loss": 0.04689854383468628, "global_step": 166746, "epoch": 1873, "lr": 3.253644525297422e-05}
{"train_loss": 0.056418243795633316, "global_step": 166747, "epoch": 1873, "lr": 3.2535902028359354e-05}
{"train_loss": 0.05228489264845848, "global_step": 166748, "epoch": 1873, "lr": 3.2535358806092333e-05}
{"train_loss": 0.0982004776597023, "global_step": 166749, "epoch": 1873, "lr": 3.2534815586173207e-05}
{"train_loss": 0.06629300117492676, "global_step": 166750, "epoch": 1873, "lr": 3.2534272368602075e-05}
{"train_loss": 0.08162081986665726, "global_step": 166751, "epoch": 1873, "lr": 3.253372915337899e-05}
{"train_loss": 0.09113288670778275, "global_step": 166752, "epoch": 1873, "lr": 3.253318594050404e-05}
{"train_loss": 0.08110477775335312, "global_step": 166753, "epoch": 1873, "lr": 3.253264272997728e-05}
{"train_loss": 0.10118818283081055, "global_step": 166754, "epoch": 1873, "lr": 3.2532099521798794e-05}
{"train_loss": 0.06973394006490707, "global_step": 166755, "epoch": 1873, "lr": 3.253155631596866e-05}
{"train_loss": 0.04457201063632965, "global_step": 166756, "epoch": 1873, "lr": 3.253101311248695e-05}
{"train_loss": 0.06983024626970291, "global_step": 166757, "epoch": 1873, "lr": 3.2530469911353714e-05}
{"train_loss": 0.08137373626232147, "global_step": 166758, "epoch": 1873, "lr": 3.252992671256906e-05}
{"train_loss": 0.037032511085271835, "global_step": 166759, "epoch": 1873, "lr": 3.252938351613303e-05}
{"train_loss": 0.11853385716676712, "global_step": 166760, "epoch": 1873, "lr": 3.252884032204573e-05}
{"train_loss": 0.08494831621646881, "global_step": 166761, "epoch": 1873, "lr": 3.252829713030719e-05}
{"train_loss": 0.050074152648448944, "global_step": 166762, "epoch": 1873, "lr": 3.252775394091754e-05}
{"train_loss": 0.047533560544252396, "global_step": 166763, "epoch": 1873, "lr": 3.252721075387679e-05}
{"train_loss": 0.15164919197559357, "global_step": 166764, "epoch": 1873, "lr": 3.2526667569185044e-05}
{"train_loss": 0.10726235806941986, "global_step": 166765, "epoch": 1873, "lr": 3.252612438684239e-05}
{"train_loss": 0.07123000174760818, "global_step": 166766, "epoch": 1873, "lr": 3.2525581206848875e-05}
{"train_loss": 0.13546425104141235, "global_step": 166767, "epoch": 1873, "lr": 3.25250380292046e-05}
{"train_loss": 0.09958096593618393, "global_step": 166768, "epoch": 1873, "lr": 3.25244948539096e-05}
{"train_loss": 0.08920148015022278, "global_step": 166769, "epoch": 1873, "lr": 3.2523951680963995e-05}
{"train_loss": 0.15360292792320251, "global_step": 166770, "epoch": 1873, "lr": 3.252340851036781e-05}
{"train_loss": 0.07046731561422348, "global_step": 166771, "epoch": 1873, "lr": 3.2522865342121145e-05}
{"train_loss": 0.11838266253471375, "global_step": 166772, "epoch": 1873, "lr": 3.252232217622407e-05}
{"train_loss": 0.12451809644699097, "global_step": 166773, "epoch": 1873, "lr": 3.2521779012676655e-05}
{"train_loss": 0.06873244792222977, "global_step": 166774, "epoch": 1873, "lr": 3.252123585147897e-05}
{"train_loss": 0.07013177126646042, "global_step": 166775, "epoch": 1873, "lr": 3.2520692692631106e-05}
{"train_loss": 0.1458788961172104, "global_step": 166776, "epoch": 1873, "lr": 3.2520149536133103e-05}
{"train_loss": 0.039214521646499634, "global_step": 166777, "epoch": 1873, "lr": 3.251960638198507e-05}
{"train_loss": 0.11384198069572449, "global_step": 166778, "epoch": 1873, "lr": 3.251906323018704e-05}
{"train_loss": 0.029565783217549324, "global_step": 166779, "epoch": 1873, "lr": 3.251852008073913e-05}
{"train_loss": 0.07312776893377304, "global_step": 166780, "epoch": 1873, "lr": 3.251797693364138e-05}
{"train_loss": 0.11683696508407593, "global_step": 166781, "epoch": 1873, "lr": 3.2517433788893885e-05}
{"train_loss": 0.09141034632921219, "global_step": 166782, "epoch": 1873, "lr": 3.251689064649669e-05}
{"train_loss": 0.06125083565711975, "global_step": 166783, "epoch": 1873, "lr": 3.2516347506449884e-05}
{"train_loss": 0.08271078765392303, "global_step": 166784, "epoch": 1873, "lr": 3.2515804368753566e-05}
{"train_loss": 0.084712782895632, "global_step": 166785, "epoch": 1873, "lr": 3.2515261233407766e-05, "val_loss": 7.026007652282715}
{"train_loss": 0.08465075492858887, "global_step": 166786, "epoch": 1874, "lr": 3.25147181004126e-05}
{"train_loss": 0.05960206314921379, "global_step": 166787, "epoch": 1874, "lr": 3.2514174969768086e-05}
{"train_loss": 0.06174679845571518, "global_step": 166788, "epoch": 1874, "lr": 3.2513631841474356e-05}
{"train_loss": 0.06351937353610992, "global_step": 166789, "epoch": 1874, "lr": 3.251308871553144e-05}
{"train_loss": 0.0845910981297493, "global_step": 166790, "epoch": 1874, "lr": 3.251254559193943e-05}
{"train_loss": 0.06380575895309448, "global_step": 166791, "epoch": 1874, "lr": 3.25120024706984e-05}
{"train_loss": 0.08966239541769028, "global_step": 166792, "epoch": 1874, "lr": 3.25114593518084e-05}
{"train_loss": 0.04201219603419304, "global_step": 166793, "epoch": 1874, "lr": 3.251091623526954e-05}
{"train_loss": 0.06351809948682785, "global_step": 166794, "epoch": 1874, "lr": 3.2510373121081875e-05}
{"train_loss": 0.09660604596138, "global_step": 166795, "epoch": 1874, "lr": 3.250983000924546e-05}
{"train_loss": 0.07053864747285843, "global_step": 166796, "epoch": 1874, "lr": 3.250928689976041e-05}
{"train_loss": 0.04569819197058678, "global_step": 166797, "epoch": 1874, "lr": 3.250874379262675e-05}
{"train_loss": 0.1435515284538269, "global_step": 166798, "epoch": 1874, "lr": 3.250820068784459e-05}
{"train_loss": 0.04880273714661598, "global_step": 166799, "epoch": 1874, "lr": 3.2507657585413977e-05}
{"train_loss": 0.07094507664442062, "global_step": 166800, "epoch": 1874, "lr": 3.250711448533499e-05}
{"train_loss": 0.10753853619098663, "global_step": 166801, "epoch": 1874, "lr": 3.250657138760773e-05}
{"train_loss": 0.07730009406805038, "global_step": 166802, "epoch": 1874, "lr": 3.250602829223223e-05}
{"train_loss": 0.08209484815597534, "global_step": 166803, "epoch": 1874, "lr": 3.25054851992086e-05}
{"train_loss": 0.07225096225738525, "global_step": 166804, "epoch": 1874, "lr": 3.250494210853687e-05}
{"train_loss": 0.10064300894737244, "global_step": 166805, "epoch": 1874, "lr": 3.250439902021716e-05}
{"train_loss": 0.0965387225151062, "global_step": 166806, "epoch": 1874, "lr": 3.2503855934249506e-05}
{"train_loss": 0.11634376645088196, "global_step": 166807, "epoch": 1874, "lr": 3.2503312850634013e-05}
{"train_loss": 0.0457569882273674, "global_step": 166808, "epoch": 1874, "lr": 3.2502769769370724e-05}
{"train_loss": 0.13783565163612366, "global_step": 166809, "epoch": 1874, "lr": 3.250222669045973e-05}
{"train_loss": 0.07275289297103882, "global_step": 166810, "epoch": 1874, "lr": 3.250168361390109e-05}
{"train_loss": 0.13878722488880157, "global_step": 166811, "epoch": 1874, "lr": 3.2501140539694885e-05}
{"train_loss": 0.10024680197238922, "global_step": 166812, "epoch": 1874, "lr": 3.250059746784119e-05}
{"train_loss": 0.10455095767974854, "global_step": 166813, "epoch": 1874, "lr": 3.25000543983401e-05}
{"train_loss": 0.0656493529677391, "global_step": 166814, "epoch": 1874, "lr": 3.2499511331191623e-05}
{"train_loss": 0.07845968753099442, "global_step": 166815, "epoch": 1874, "lr": 3.249896826639591e-05}
{"train_loss": 0.06966762244701385, "global_step": 166816, "epoch": 1874, "lr": 3.2498425203952966e-05}
{"train_loss": 0.07309833914041519, "global_step": 166817, "epoch": 1874, "lr": 3.2497882143862925e-05}
{"train_loss": 0.1561610996723175, "global_step": 166818, "epoch": 1874, "lr": 3.24973390861258e-05}
{"train_loss": 0.07570306956768036, "global_step": 166819, "epoch": 1874, "lr": 3.2496796030741705e-05}
{"train_loss": 0.07691094279289246, "global_step": 166820, "epoch": 1874, "lr": 3.2496252977710715e-05}
{"train_loss": 0.15557783842086792, "global_step": 166821, "epoch": 1874, "lr": 3.249570992703287e-05}
{"train_loss": 0.11928921192884445, "global_step": 166822, "epoch": 1874, "lr": 3.249516687870829e-05}
{"train_loss": 0.17596065998077393, "global_step": 166823, "epoch": 1874, "lr": 3.249462383273699e-05}
{"train_loss": 0.13871005177497864, "global_step": 166824, "epoch": 1874, "lr": 3.2494080789119106e-05}
{"train_loss": 0.0801912173628807, "global_step": 166825, "epoch": 1874, "lr": 3.249353774785466e-05}
{"train_loss": 0.13541221618652344, "global_step": 166826, "epoch": 1874, "lr": 3.249299470894376e-05}
{"train_loss": 0.14363358914852142, "global_step": 166827, "epoch": 1874, "lr": 3.2492451672386456e-05}
{"train_loss": 0.10570453107357025, "global_step": 166828, "epoch": 1874, "lr": 3.249190863818282e-05}
{"train_loss": 0.09691709280014038, "global_step": 166829, "epoch": 1874, "lr": 3.2491365606332946e-05}
{"train_loss": 0.08165644109249115, "global_step": 166830, "epoch": 1874, "lr": 3.249082257683689e-05}
{"train_loss": 0.11602097749710083, "global_step": 166831, "epoch": 1874, "lr": 3.249027954969472e-05}
{"train_loss": 0.09839704632759094, "global_step": 166832, "epoch": 1874, "lr": 3.248973652490653e-05}
{"train_loss": 0.10538754612207413, "global_step": 166833, "epoch": 1874, "lr": 3.2489193502472373e-05}
{"train_loss": 0.06430341303348541, "global_step": 166834, "epoch": 1874, "lr": 3.248865048239235e-05}
{"train_loss": 0.1360633373260498, "global_step": 166835, "epoch": 1874, "lr": 3.248810746466649e-05}
{"train_loss": 0.08991415053606033, "global_step": 166836, "epoch": 1874, "lr": 3.248756444929489e-05}
{"train_loss": 0.11099663376808167, "global_step": 166837, "epoch": 1874, "lr": 3.248702143627764e-05}
{"train_loss": 0.14603669941425323, "global_step": 166838, "epoch": 1874, "lr": 3.2486478425614786e-05}
{"train_loss": 0.16390544176101685, "global_step": 166839, "epoch": 1874, "lr": 3.2485935417306424e-05}
{"train_loss": 0.09055078029632568, "global_step": 166840, "epoch": 1874, "lr": 3.2485392411352596e-05}
{"train_loss": 0.07169153541326523, "global_step": 166841, "epoch": 1874, "lr": 3.248484940775341e-05}
{"train_loss": 0.09113270789384842, "global_step": 166842, "epoch": 1874, "lr": 3.248430640650891e-05}
{"train_loss": 0.06035372242331505, "global_step": 166843, "epoch": 1874, "lr": 3.24837634076192e-05}
{"train_loss": 0.05791078135371208, "global_step": 166844, "epoch": 1874, "lr": 3.248322041108431e-05}
{"train_loss": 0.12763909995555878, "global_step": 166845, "epoch": 1874, "lr": 3.248267741690436e-05}
{"train_loss": 0.04317092522978783, "global_step": 166846, "epoch": 1874, "lr": 3.248213442507938e-05}
{"train_loss": 0.07865568995475769, "global_step": 166847, "epoch": 1874, "lr": 3.248159143560948e-05}
{"train_loss": 0.06817218661308289, "global_step": 166848, "epoch": 1874, "lr": 3.248104844849471e-05}
{"train_loss": 0.07007882744073868, "global_step": 166849, "epoch": 1874, "lr": 3.248050546373516e-05}
{"train_loss": 0.059471797198057175, "global_step": 166850, "epoch": 1874, "lr": 3.247996248133088e-05}
{"train_loss": 0.09495995938777924, "global_step": 166851, "epoch": 1874, "lr": 3.247941950128196e-05}
{"train_loss": 0.080841563642025, "global_step": 166852, "epoch": 1874, "lr": 3.247887652358846e-05}
{"train_loss": 0.05724477395415306, "global_step": 166853, "epoch": 1874, "lr": 3.2478333548250485e-05}
{"train_loss": 0.08014646172523499, "global_step": 166854, "epoch": 1874, "lr": 3.247779057526805e-05}
{"train_loss": 0.10060147941112518, "global_step": 166855, "epoch": 1874, "lr": 3.247724760464127e-05}
{"train_loss": 0.05396181344985962, "global_step": 166856, "epoch": 1874, "lr": 3.2476704636370234e-05}
{"train_loss": 0.06731041520833969, "global_step": 166857, "epoch": 1874, "lr": 3.247616167045496e-05}
{"train_loss": 0.10283331573009491, "global_step": 166858, "epoch": 1874, "lr": 3.247561870689558e-05}
{"train_loss": 0.09718286246061325, "global_step": 166859, "epoch": 1874, "lr": 3.247507574569213e-05}
{"train_loss": 0.06186474859714508, "global_step": 166860, "epoch": 1874, "lr": 3.24745327868447e-05}
{"train_loss": 0.10369466245174408, "global_step": 166861, "epoch": 1874, "lr": 3.2473989830353335e-05}
{"train_loss": 0.1203593760728836, "global_step": 166862, "epoch": 1874, "lr": 3.247344687621815e-05}
{"train_loss": 0.07467545568943024, "global_step": 166863, "epoch": 1874, "lr": 3.247290392443918e-05}
{"train_loss": 0.04678457975387573, "global_step": 166864, "epoch": 1874, "lr": 3.247236097501654e-05}
{"train_loss": 0.05665416643023491, "global_step": 166865, "epoch": 1874, "lr": 3.247181802795024e-05}
{"train_loss": 0.04811063036322594, "global_step": 166866, "epoch": 1874, "lr": 3.247127508324043e-05}
{"train_loss": 0.09768351167440414, "global_step": 166867, "epoch": 1874, "lr": 3.2470732140887114e-05}
{"train_loss": 0.12495283037424088, "global_step": 166868, "epoch": 1874, "lr": 3.247018920089041e-05}
{"train_loss": 0.0799180120229721, "global_step": 166869, "epoch": 1874, "lr": 3.246964626325036e-05}
{"train_loss": 0.09131699800491333, "global_step": 166870, "epoch": 1874, "lr": 3.246910332796706e-05}
{"train_loss": 0.09385872632265091, "global_step": 166871, "epoch": 1874, "lr": 3.246856039504058e-05}
{"train_loss": 0.05914858356118202, "global_step": 166872, "epoch": 1874, "lr": 3.246801746447098e-05}
{"train_loss": 0.13596133887767792, "global_step": 166873, "epoch": 1874, "lr": 3.246747453625836e-05}
{"train_loss": 0.09019594220967775, "global_step": 166874, "epoch": 1874, "lr": 3.246693161040275e-05, "val_loss": 6.930141925811768}
{"train_loss": 0.08529775589704514, "global_step": 166875, "epoch": 1875, "lr": 3.246638868690427e-05}
{"train_loss": 0.04912849888205528, "global_step": 166876, "epoch": 1875, "lr": 3.246584576576295e-05}
{"train_loss": 0.05762543901801109, "global_step": 166877, "epoch": 1875, "lr": 3.2465302846978894e-05}
{"train_loss": 0.10025622695684433, "global_step": 166878, "epoch": 1875, "lr": 3.246475993055216e-05}
{"train_loss": 0.12704993784427643, "global_step": 166879, "epoch": 1875, "lr": 3.2464217016482834e-05}
{"train_loss": 0.07574530690908432, "global_step": 166880, "epoch": 1875, "lr": 3.246367410477096e-05}
{"train_loss": 0.0904049500823021, "global_step": 166881, "epoch": 1875, "lr": 3.2463131195416656e-05}
{"train_loss": 0.09646747261285782, "global_step": 166882, "epoch": 1875, "lr": 3.246258828841995e-05}
{"train_loss": 0.08792144060134888, "global_step": 166883, "epoch": 1875, "lr": 3.246204538378096e-05}
{"train_loss": 0.07324164360761642, "global_step": 166884, "epoch": 1875, "lr": 3.246150248149971e-05}
{"train_loss": 0.06477236747741699, "global_step": 166885, "epoch": 1875, "lr": 3.246095958157631e-05}
{"train_loss": 0.10735947638750076, "global_step": 166886, "epoch": 1875, "lr": 3.246041668401082e-05}
{"train_loss": 0.07121910899877548, "global_step": 166887, "epoch": 1875, "lr": 3.2459873788803305e-05}
{"train_loss": 0.08320560306310654, "global_step": 166888, "epoch": 1875, "lr": 3.2459330895953855e-05}
{"train_loss": 0.14311663806438446, "global_step": 166889, "epoch": 1875, "lr": 3.2458788005462524e-05}
{"train_loss": 0.06888650357723236, "global_step": 166890, "epoch": 1875, "lr": 3.245824511732941e-05}
{"train_loss": 0.0775497704744339, "global_step": 166891, "epoch": 1875, "lr": 3.245770223155455e-05}
{"train_loss": 0.11298444122076035, "global_step": 166892, "epoch": 1875, "lr": 3.245715934813806e-05}
{"train_loss": 0.09989173710346222, "global_step": 166893, "epoch": 1875, "lr": 3.245661646707997e-05}
{"train_loss": 0.05556170269846916, "global_step": 166894, "epoch": 1875, "lr": 3.24560735883804e-05}
{"train_loss": 0.07657473534345627, "global_step": 166895, "epoch": 1875, "lr": 3.245553071203937e-05}
{"train_loss": 0.051704056560993195, "global_step": 166896, "epoch": 1875, "lr": 3.2454987838057005e-05}
{"train_loss": 0.08684244006872177, "global_step": 166897, "epoch": 1875, "lr": 3.2454444966433326e-05}
{"train_loss": 0.047943126410245895, "global_step": 166898, "epoch": 1875, "lr": 3.245390209716846e-05}
{"train_loss": 0.0977758914232254, "global_step": 166899, "epoch": 1875, "lr": 3.245335923026243e-05}
{"train_loss": 0.17183242738246918, "global_step": 166900, "epoch": 1875, "lr": 3.245281636571536e-05}
{"train_loss": 0.11530643701553345, "global_step": 166901, "epoch": 1875, "lr": 3.245227350352727e-05}
{"train_loss": 0.12972962856292725, "global_step": 166902, "epoch": 1875, "lr": 3.245173064369827e-05}
{"train_loss": 0.10443071275949478, "global_step": 166903, "epoch": 1875, "lr": 3.2451187786228406e-05}
{"train_loss": 0.037541601806879044, "global_step": 166904, "epoch": 1875, "lr": 3.2450644931117784e-05}
{"train_loss": 0.07323762029409409, "global_step": 166905, "epoch": 1875, "lr": 3.245010207836645e-05}
{"train_loss": 0.08695909380912781, "global_step": 166906, "epoch": 1875, "lr": 3.244955922797449e-05}
{"train_loss": 0.12723605334758759, "global_step": 166907, "epoch": 1875, "lr": 3.244901637994197e-05}
{"train_loss": 0.13867385685443878, "global_step": 166908, "epoch": 1875, "lr": 3.2448473534268964e-05}
{"train_loss": 0.08823025971651077, "global_step": 166909, "epoch": 1875, "lr": 3.244793069095555e-05}
{"train_loss": 0.07821820676326752, "global_step": 166910, "epoch": 1875, "lr": 3.2447387850001786e-05}
{"train_loss": 0.07817544788122177, "global_step": 166911, "epoch": 1875, "lr": 3.244684501140778e-05}
{"train_loss": 0.05772843211889267, "global_step": 166912, "epoch": 1875, "lr": 3.244630217517356e-05}
{"train_loss": 0.06527712196111679, "global_step": 166913, "epoch": 1875, "lr": 3.244575934129924e-05}
{"train_loss": 0.04607902839779854, "global_step": 166914, "epoch": 1875, "lr": 3.244521650978485e-05}
{"train_loss": 0.1391286849975586, "global_step": 166915, "epoch": 1875, "lr": 3.244467368063051e-05}
{"train_loss": 0.07423675060272217, "global_step": 166916, "epoch": 1875, "lr": 3.2444130853836254e-05}
{"train_loss": 0.09801837056875229, "global_step": 166917, "epoch": 1875, "lr": 3.244358802940218e-05}
{"train_loss": 0.10874372720718384, "global_step": 166918, "epoch": 1875, "lr": 3.244304520732834e-05}
{"train_loss": 0.05402173846960068, "global_step": 166919, "epoch": 1875, "lr": 3.244250238761484e-05}
{"train_loss": 0.1128726452589035, "global_step": 166920, "epoch": 1875, "lr": 3.244195957026171e-05}
{"train_loss": 0.1060371845960617, "global_step": 166921, "epoch": 1875, "lr": 3.244141675526906e-05}
{"train_loss": 0.0922698900103569, "global_step": 166922, "epoch": 1875, "lr": 3.244087394263693e-05}
{"train_loss": 0.10015419125556946, "global_step": 166923, "epoch": 1875, "lr": 3.244033113236543e-05}
{"train_loss": 0.0984259694814682, "global_step": 166924, "epoch": 1875, "lr": 3.243978832445461e-05}
{"train_loss": 0.05636350065469742, "global_step": 166925, "epoch": 1875, "lr": 3.2439245518904536e-05}
{"train_loss": 0.10408959537744522, "global_step": 166926, "epoch": 1875, "lr": 3.24387027157153e-05}
{"train_loss": 0.12051770091056824, "global_step": 166927, "epoch": 1875, "lr": 3.243815991488695e-05}
{"train_loss": 0.1046380028128624, "global_step": 166928, "epoch": 1875, "lr": 3.24376171164196e-05}
{"train_loss": 0.1150456890463829, "global_step": 166929, "epoch": 1875, "lr": 3.243707432031328e-05}
{"train_loss": 0.10581736266613007, "global_step": 166930, "epoch": 1875, "lr": 3.243653152656809e-05}
{"train_loss": 0.09414349496364594, "global_step": 166931, "epoch": 1875, "lr": 3.243598873518409e-05}
{"train_loss": 0.0891646146774292, "global_step": 166932, "epoch": 1875, "lr": 3.243544594616136e-05}
{"train_loss": 0.08893407881259918, "global_step": 166933, "epoch": 1875, "lr": 3.243490315949996e-05}
{"train_loss": 0.18584147095680237, "global_step": 166934, "epoch": 1875, "lr": 3.243436037519999e-05}
{"train_loss": 0.0995006412267685, "global_step": 166935, "epoch": 1875, "lr": 3.24338175932615e-05}
{"train_loss": 0.16030707955360413, "global_step": 166936, "epoch": 1875, "lr": 3.243327481368457e-05}
{"train_loss": 0.1426515430212021, "global_step": 166937, "epoch": 1875, "lr": 3.243273203646926e-05}
{"train_loss": 0.08715824037790298, "global_step": 166938, "epoch": 1875, "lr": 3.2432189261615674e-05}
{"train_loss": 0.07195228338241577, "global_step": 166939, "epoch": 1875, "lr": 3.2431646489123846e-05}
{"train_loss": 0.12659646570682526, "global_step": 166940, "epoch": 1875, "lr": 3.243110371899389e-05}
{"train_loss": 0.08208617568016052, "global_step": 166941, "epoch": 1875, "lr": 3.243056095122584e-05}
{"train_loss": 0.05437422916293144, "global_step": 166942, "epoch": 1875, "lr": 3.2430018185819806e-05}
{"train_loss": 0.07223334908485413, "global_step": 166943, "epoch": 1875, "lr": 3.242947542277583e-05}
{"train_loss": 0.051678355783224106, "global_step": 166944, "epoch": 1875, "lr": 3.2428932662093994e-05}
{"train_loss": 0.10209310054779053, "global_step": 166945, "epoch": 1875, "lr": 3.242838990377438e-05}
{"train_loss": 0.09102775156497955, "global_step": 166946, "epoch": 1875, "lr": 3.2427847147817044e-05}
{"train_loss": 0.10134775191545486, "global_step": 166947, "epoch": 1875, "lr": 3.242730439422209e-05}
{"train_loss": 0.07444306463003159, "global_step": 166948, "epoch": 1875, "lr": 3.242676164298955e-05}
{"train_loss": 0.06700529903173447, "global_step": 166949, "epoch": 1875, "lr": 3.242621889411953e-05}
{"train_loss": 0.07409194856882095, "global_step": 166950, "epoch": 1875, "lr": 3.2425676147612084e-05}
{"train_loss": 0.11799561232328415, "global_step": 166951, "epoch": 1875, "lr": 3.24251334034673e-05}
{"train_loss": 0.06570076197385788, "global_step": 166952, "epoch": 1875, "lr": 3.2424590661685225e-05}
{"train_loss": 0.0675811842083931, "global_step": 166953, "epoch": 1875, "lr": 3.2424047922265975e-05}
{"train_loss": 0.17994588613510132, "global_step": 166954, "epoch": 1875, "lr": 3.2423505185209574e-05}
{"train_loss": 0.06518435478210449, "global_step": 166955, "epoch": 1875, "lr": 3.242296245051614e-05}
{"train_loss": 0.05928045138716698, "global_step": 166956, "epoch": 1875, "lr": 3.2422419718185706e-05}
{"train_loss": 0.06945262849330902, "global_step": 166957, "epoch": 1875, "lr": 3.242187698821838e-05}
{"train_loss": 0.09159880876541138, "global_step": 166958, "epoch": 1875, "lr": 3.24213342606142e-05}
{"train_loss": 0.04068850353360176, "global_step": 166959, "epoch": 1875, "lr": 3.2420791535373275e-05}
{"train_loss": 0.08263549208641052, "global_step": 166960, "epoch": 1875, "lr": 3.242024881249564e-05}
{"train_loss": 0.1689046025276184, "global_step": 166961, "epoch": 1875, "lr": 3.241970609198141e-05}
{"train_loss": 0.09674343466758728, "global_step": 166962, "epoch": 1875, "lr": 3.241916337383062e-05}
{"train_loss": 0.09176590226674347, "global_step": 166963, "epoch": 1875, "lr": 3.2418620658043366e-05, "val_loss": 7.086670398712158, "train_action_mse_error": 13.576519012451172}
{"train_loss": 0.12679140269756317, "global_step": 166964, "epoch": 1876, "lr": 3.241807794461972e-05}
{"train_loss": 0.09573464095592499, "global_step": 166965, "epoch": 1876, "lr": 3.2417535233559735e-05}
{"train_loss": 0.10671674460172653, "global_step": 166966, "epoch": 1876, "lr": 3.2416992524863515e-05}
{"train_loss": 0.0924241691827774, "global_step": 166967, "epoch": 1876, "lr": 3.24164498185311e-05}
{"train_loss": 0.07369329035282135, "global_step": 166968, "epoch": 1876, "lr": 3.24159071145626e-05}
{"train_loss": 0.07397568970918655, "global_step": 166969, "epoch": 1876, "lr": 3.2415364412958035e-05}
{"train_loss": 0.13769598305225372, "global_step": 166970, "epoch": 1876, "lr": 3.2414821713717535e-05}
{"train_loss": 0.11386919021606445, "global_step": 166971, "epoch": 1876, "lr": 3.241427901684113e-05}
{"train_loss": 0.1023353859782219, "global_step": 166972, "epoch": 1876, "lr": 3.241373632232894e-05}
{"train_loss": 0.0969172865152359, "global_step": 166973, "epoch": 1876, "lr": 3.241319363018098e-05}
{"train_loss": 0.06979815661907196, "global_step": 166974, "epoch": 1876, "lr": 3.241265094039737e-05}
{"train_loss": 0.039583683013916016, "global_step": 166975, "epoch": 1876, "lr": 3.241210825297815e-05}
{"train_loss": 0.13229049742221832, "global_step": 166976, "epoch": 1876, "lr": 3.241156556792342e-05}
{"train_loss": 0.09424757212400436, "global_step": 166977, "epoch": 1876, "lr": 3.241102288523323e-05}
{"train_loss": 0.04203871265053749, "global_step": 166978, "epoch": 1876, "lr": 3.241048020490768e-05}
{"train_loss": 0.024947453290224075, "global_step": 166979, "epoch": 1876, "lr": 3.24099375269468e-05}
{"train_loss": 0.09033431857824326, "global_step": 166980, "epoch": 1876, "lr": 3.240939485135071e-05}
{"train_loss": 0.08579280227422714, "global_step": 166981, "epoch": 1876, "lr": 3.240885217811945e-05}
{"train_loss": 0.03917595371603966, "global_step": 166982, "epoch": 1876, "lr": 3.240830950725312e-05}
{"train_loss": 0.0673607587814331, "global_step": 166983, "epoch": 1876, "lr": 3.240776683875178e-05}
{"train_loss": 0.09815249592065811, "global_step": 166984, "epoch": 1876, "lr": 3.240722417261548e-05}
{"train_loss": 0.15176382660865784, "global_step": 166985, "epoch": 1876, "lr": 3.240668150884434e-05}
{"train_loss": 0.040349628776311874, "global_step": 166986, "epoch": 1876, "lr": 3.240613884743838e-05}
{"train_loss": 0.09155281633138657, "global_step": 166987, "epoch": 1876, "lr": 3.240559618839772e-05}
{"train_loss": 0.06836774945259094, "global_step": 166988, "epoch": 1876, "lr": 3.240505353172241e-05}
{"train_loss": 0.09677114337682724, "global_step": 166989, "epoch": 1876, "lr": 3.240451087741253e-05}
{"train_loss": 0.17475463449954987, "global_step": 166990, "epoch": 1876, "lr": 3.240396822546813e-05}
{"train_loss": 0.07625982165336609, "global_step": 166991, "epoch": 1876, "lr": 3.240342557588932e-05}
{"train_loss": 0.11243551969528198, "global_step": 166992, "epoch": 1876, "lr": 3.240288292867614e-05}
{"train_loss": 0.05718819051980972, "global_step": 166993, "epoch": 1876, "lr": 3.24023402838287e-05}
{"train_loss": 0.0734299048781395, "global_step": 166994, "epoch": 1876, "lr": 3.240179764134703e-05}
{"train_loss": 0.08143175393342972, "global_step": 166995, "epoch": 1876, "lr": 3.240125500123125e-05}
{"train_loss": 0.08882756531238556, "global_step": 166996, "epoch": 1876, "lr": 3.240071236348137e-05}
{"train_loss": 0.056305523961782455, "global_step": 166997, "epoch": 1876, "lr": 3.240016972809753e-05}
{"train_loss": 0.1204279363155365, "global_step": 166998, "epoch": 1876, "lr": 3.2399627095079754e-05}
{"train_loss": 0.08429847657680511, "global_step": 166999, "epoch": 1876, "lr": 3.2399084464428134e-05}
{"train_loss": 0.05232369527220726, "global_step": 167000, "epoch": 1876, "lr": 3.239854183614276e-05}
{"train_loss": 0.0709880143404007, "global_step": 167001, "epoch": 1876, "lr": 3.239799921022367e-05}
{"train_loss": 0.11655090004205704, "global_step": 167002, "epoch": 1876, "lr": 3.239745658667097e-05}
{"train_loss": 0.11398404091596603, "global_step": 167003, "epoch": 1876, "lr": 3.23969139654847e-05}
{"train_loss": 0.07577534765005112, "global_step": 167004, "epoch": 1876, "lr": 3.2396371346664965e-05}
{"train_loss": 0.10888569802045822, "global_step": 167005, "epoch": 1876, "lr": 3.239582873021181e-05}
{"train_loss": 0.1921042650938034, "global_step": 167006, "epoch": 1876, "lr": 3.2395286116125344e-05}
{"train_loss": 0.057692792266607285, "global_step": 167007, "epoch": 1876, "lr": 3.239474350440559e-05}
{"train_loss": 0.09222885966300964, "global_step": 167008, "epoch": 1876, "lr": 3.239420089505267e-05}
{"train_loss": 0.06684897840023041, "global_step": 167009, "epoch": 1876, "lr": 3.2393658288066615e-05}
{"train_loss": 0.07969565689563751, "global_step": 167010, "epoch": 1876, "lr": 3.239311568344754e-05}
{"train_loss": 0.09828118234872818, "global_step": 167011, "epoch": 1876, "lr": 3.239257308119548e-05}
{"train_loss": 0.08920799195766449, "global_step": 167012, "epoch": 1876, "lr": 3.2392030481310534e-05}
{"train_loss": 0.1450783610343933, "global_step": 167013, "epoch": 1876, "lr": 3.239148788379275e-05}
{"train_loss": 0.060067515820264816, "global_step": 167014, "epoch": 1876, "lr": 3.239094528864223e-05}
{"train_loss": 0.08253573626279831, "global_step": 167015, "epoch": 1876, "lr": 3.239040269585902e-05}
{"train_loss": 0.029278045520186424, "global_step": 167016, "epoch": 1876, "lr": 3.2389860105443225e-05}
{"train_loss": 0.0794617310166359, "global_step": 167017, "epoch": 1876, "lr": 3.2389317517394875e-05}
{"train_loss": 0.06344678997993469, "global_step": 167018, "epoch": 1876, "lr": 3.2388774931714076e-05}
{"train_loss": 0.07070761919021606, "global_step": 167019, "epoch": 1876, "lr": 3.238823234840089e-05}
{"train_loss": 0.05142275243997574, "global_step": 167020, "epoch": 1876, "lr": 3.2387689767455406e-05}
{"train_loss": 0.08969522267580032, "global_step": 167021, "epoch": 1876, "lr": 3.238714718887768e-05}
{"train_loss": 0.0736098513007164, "global_step": 167022, "epoch": 1876, "lr": 3.238660461266777e-05}
{"train_loss": 0.07305873930454254, "global_step": 167023, "epoch": 1876, "lr": 3.238606203882578e-05}
{"train_loss": 0.10021039098501205, "global_step": 167024, "epoch": 1876, "lr": 3.238551946735176e-05}
{"train_loss": 0.14299136400222778, "global_step": 167025, "epoch": 1876, "lr": 3.238497689824581e-05}
{"train_loss": 0.13926444947719574, "global_step": 167026, "epoch": 1876, "lr": 3.238443433150796e-05}
{"train_loss": 0.14065347611904144, "global_step": 167027, "epoch": 1876, "lr": 3.238389176713833e-05}
{"train_loss": 0.06687448173761368, "global_step": 167028, "epoch": 1876, "lr": 3.238334920513695e-05}
{"train_loss": 0.10727277398109436, "global_step": 167029, "epoch": 1876, "lr": 3.2382806645503935e-05}
{"train_loss": 0.05623139813542366, "global_step": 167030, "epoch": 1876, "lr": 3.238226408823932e-05}
{"train_loss": 0.07169298082590103, "global_step": 167031, "epoch": 1876, "lr": 3.238172153334321e-05}
{"train_loss": 0.08440601080656052, "global_step": 167032, "epoch": 1876, "lr": 3.238117898081565e-05}
{"train_loss": 0.08456412702798843, "global_step": 167033, "epoch": 1876, "lr": 3.238063643065674e-05}
{"train_loss": 0.09874100983142853, "global_step": 167034, "epoch": 1876, "lr": 3.2380093882866526e-05}
{"train_loss": 0.03654955327510834, "global_step": 167035, "epoch": 1876, "lr": 3.23795513374451e-05}
{"train_loss": 0.11546916514635086, "global_step": 167036, "epoch": 1876, "lr": 3.237900879439252e-05}
{"train_loss": 0.0875009149312973, "global_step": 167037, "epoch": 1876, "lr": 3.2378466253708874e-05}
{"train_loss": 0.07542171329259872, "global_step": 167038, "epoch": 1876, "lr": 3.2377923715394225e-05}
{"train_loss": 0.07483142614364624, "global_step": 167039, "epoch": 1876, "lr": 3.2377381179448654e-05}
{"train_loss": 0.1290026605129242, "global_step": 167040, "epoch": 1876, "lr": 3.237683864587224e-05}
{"train_loss": 0.06681376695632935, "global_step": 167041, "epoch": 1876, "lr": 3.237629611466502e-05}
{"train_loss": 0.04518769681453705, "global_step": 167042, "epoch": 1876, "lr": 3.2375753585827116e-05}
{"train_loss": 0.08893968909978867, "global_step": 167043, "epoch": 1876, "lr": 3.2375211059358556e-05}
{"train_loss": 0.08957794308662415, "global_step": 167044, "epoch": 1876, "lr": 3.237466853525946e-05}
{"train_loss": 0.09785918891429901, "global_step": 167045, "epoch": 1876, "lr": 3.2374126013529844e-05}
{"train_loss": 0.053373195230960846, "global_step": 167046, "epoch": 1876, "lr": 3.2373583494169834e-05}
{"train_loss": 0.06569890677928925, "global_step": 167047, "epoch": 1876, "lr": 3.237304097717947e-05}
{"train_loss": 0.09691362828016281, "global_step": 167048, "epoch": 1876, "lr": 3.237249846255884e-05}
{"train_loss": 0.13174790143966675, "global_step": 167049, "epoch": 1876, "lr": 3.2371955950308006e-05}
{"train_loss": 0.05718447268009186, "global_step": 167050, "epoch": 1876, "lr": 3.237141344042707e-05}
{"train_loss": 0.07247917354106903, "global_step": 167051, "epoch": 1876, "lr": 3.2370870932916054e-05}
{"train_loss": 0.08722936280406593, "global_step": 167052, "epoch": 1876, "lr": 3.2370328427775085e-05, "val_loss": 7.166207790374756}
{"train_loss": 0.09518741816282272, "global_step": 167053, "epoch": 1877, "lr": 3.236978592500418e-05}
{"train_loss": 0.06113380566239357, "global_step": 167054, "epoch": 1877, "lr": 3.236924342460347e-05}
{"train_loss": 0.1018030047416687, "global_step": 167055, "epoch": 1877, "lr": 3.236870092657298e-05}
{"train_loss": 0.09892988950014114, "global_step": 167056, "epoch": 1877, "lr": 3.236815843091282e-05}
{"train_loss": 0.09638914465904236, "global_step": 167057, "epoch": 1877, "lr": 3.236761593762304e-05}
{"train_loss": 0.10118070989847183, "global_step": 167058, "epoch": 1877, "lr": 3.236707344670372e-05}
{"train_loss": 0.18306046724319458, "global_step": 167059, "epoch": 1877, "lr": 3.2366530958154926e-05}
{"train_loss": 0.14468270540237427, "global_step": 167060, "epoch": 1877, "lr": 3.236598847197675e-05}
{"train_loss": 0.05980810150504112, "global_step": 167061, "epoch": 1877, "lr": 3.236544598816925e-05}
{"train_loss": 0.040584124624729156, "global_step": 167062, "epoch": 1877, "lr": 3.236490350673249e-05}
{"train_loss": 0.07180699706077576, "global_step": 167063, "epoch": 1877, "lr": 3.236436102766657e-05}
{"train_loss": 0.15864242613315582, "global_step": 167064, "epoch": 1877, "lr": 3.236381855097152e-05}
{"train_loss": 0.12317566573619843, "global_step": 167065, "epoch": 1877, "lr": 3.2363276076647466e-05}
{"train_loss": 0.06995312124490738, "global_step": 167066, "epoch": 1877, "lr": 3.236273360469443e-05}
{"train_loss": 0.09282469749450684, "global_step": 167067, "epoch": 1877, "lr": 3.236219113511254e-05}
{"train_loss": 0.09313442558050156, "global_step": 167068, "epoch": 1877, "lr": 3.2361648667901806e-05}
{"train_loss": 0.05895273759961128, "global_step": 167069, "epoch": 1877, "lr": 3.236110620306237e-05}
{"train_loss": 0.052326276898384094, "global_step": 167070, "epoch": 1877, "lr": 3.236056374059423e-05}
{"train_loss": 0.07421064376831055, "global_step": 167071, "epoch": 1877, "lr": 3.236002128049752e-05}
{"train_loss": 0.0525035597383976, "global_step": 167072, "epoch": 1877, "lr": 3.235947882277228e-05}
{"train_loss": 0.08335884660482407, "global_step": 167073, "epoch": 1877, "lr": 3.235893636741858e-05}
{"train_loss": 0.0699944794178009, "global_step": 167074, "epoch": 1877, "lr": 3.2358393914436535e-05}
{"train_loss": 0.10953963547945023, "global_step": 167075, "epoch": 1877, "lr": 3.2357851463826185e-05}
{"train_loss": 0.057507410645484924, "global_step": 167076, "epoch": 1877, "lr": 3.2357309015587596e-05}
{"train_loss": 0.09952200204133987, "global_step": 167077, "epoch": 1877, "lr": 3.2356766569720854e-05}
{"train_loss": 0.059007734060287476, "global_step": 167078, "epoch": 1877, "lr": 3.235622412622603e-05}
{"train_loss": 0.08546452224254608, "global_step": 167079, "epoch": 1877, "lr": 3.23556816851032e-05}
{"train_loss": 0.04628213495016098, "global_step": 167080, "epoch": 1877, "lr": 3.235513924635244e-05}
{"train_loss": 0.06864375621080399, "global_step": 167081, "epoch": 1877, "lr": 3.23545968099738e-05}
{"train_loss": 0.057613812386989594, "global_step": 167082, "epoch": 1877, "lr": 3.235405437596738e-05}
{"train_loss": 0.10533472895622253, "global_step": 167083, "epoch": 1877, "lr": 3.235351194433323e-05}
{"train_loss": 0.12386056780815125, "global_step": 167084, "epoch": 1877, "lr": 3.235296951507146e-05}
{"train_loss": 0.10483153909444809, "global_step": 167085, "epoch": 1877, "lr": 3.235242708818209e-05}
{"train_loss": 0.060478247702121735, "global_step": 167086, "epoch": 1877, "lr": 3.235188466366524e-05}
{"train_loss": 0.09456028044223785, "global_step": 167087, "epoch": 1877, "lr": 3.235134224152095e-05}
{"train_loss": 0.08734685182571411, "global_step": 167088, "epoch": 1877, "lr": 3.2350799821749325e-05}
{"train_loss": 0.07514383643865585, "global_step": 167089, "epoch": 1877, "lr": 3.2350257404350405e-05}
{"train_loss": 0.06274176388978958, "global_step": 167090, "epoch": 1877, "lr": 3.234971498932426e-05}
{"train_loss": 0.059347715228796005, "global_step": 167091, "epoch": 1877, "lr": 3.234917257667102e-05}
{"train_loss": 0.10648662596940994, "global_step": 167092, "epoch": 1877, "lr": 3.234863016639069e-05}
{"train_loss": 0.07765834033489227, "global_step": 167093, "epoch": 1877, "lr": 3.234808775848339e-05}
{"train_loss": 0.056677550077438354, "global_step": 167094, "epoch": 1877, "lr": 3.2347545352949174e-05}
{"train_loss": 0.03852559253573418, "global_step": 167095, "epoch": 1877, "lr": 3.2347002949788106e-05}
{"train_loss": 0.07438009977340698, "global_step": 167096, "epoch": 1877, "lr": 3.2346460549000275e-05}
{"train_loss": 0.13812372088432312, "global_step": 167097, "epoch": 1877, "lr": 3.234591815058574e-05}
{"train_loss": 0.031490031629800797, "global_step": 167098, "epoch": 1877, "lr": 3.234537575454458e-05}
{"train_loss": 0.05734984204173088, "global_step": 167099, "epoch": 1877, "lr": 3.234483336087689e-05}
{"train_loss": 0.07524384558200836, "global_step": 167100, "epoch": 1877, "lr": 3.234429096958269e-05}
{"train_loss": 0.1060730516910553, "global_step": 167101, "epoch": 1877, "lr": 3.234374858066212e-05}
{"train_loss": 0.06768770515918732, "global_step": 167102, "epoch": 1877, "lr": 3.2343206194115186e-05}
{"train_loss": 0.08222169429063797, "global_step": 167103, "epoch": 1877, "lr": 3.2342663809942015e-05}
{"train_loss": 0.07633816450834274, "global_step": 167104, "epoch": 1877, "lr": 3.2342121428142636e-05}
{"train_loss": 0.12397757172584534, "global_step": 167105, "epoch": 1877, "lr": 3.2341579048717166e-05}
{"train_loss": 0.05594077333807945, "global_step": 167106, "epoch": 1877, "lr": 3.234103667166564e-05}
{"train_loss": 0.07517209649085999, "global_step": 167107, "epoch": 1877, "lr": 3.234049429698815e-05}
{"train_loss": 0.09347937256097794, "global_step": 167108, "epoch": 1877, "lr": 3.2339951924684765e-05}
{"train_loss": 0.036532286554574966, "global_step": 167109, "epoch": 1877, "lr": 3.2339409554755544e-05}
{"train_loss": 0.07950685173273087, "global_step": 167110, "epoch": 1877, "lr": 3.23388671872006e-05}
{"train_loss": 0.04691287875175476, "global_step": 167111, "epoch": 1877, "lr": 3.233832482201996e-05}
{"train_loss": 0.06893523782491684, "global_step": 167112, "epoch": 1877, "lr": 3.233778245921374e-05}
{"train_loss": 0.10048853605985641, "global_step": 167113, "epoch": 1877, "lr": 3.2337240098781974e-05}
{"train_loss": 0.11072558164596558, "global_step": 167114, "epoch": 1877, "lr": 3.233669774072475e-05}
{"train_loss": 0.03649543225765228, "global_step": 167115, "epoch": 1877, "lr": 3.233615538504215e-05}
{"train_loss": 0.0645277351140976, "global_step": 167116, "epoch": 1877, "lr": 3.2335613031734237e-05}
{"train_loss": 0.07047280669212341, "global_step": 167117, "epoch": 1877, "lr": 3.233507068080108e-05}
{"train_loss": 0.07930050045251846, "global_step": 167118, "epoch": 1877, "lr": 3.233452833224276e-05}
{"train_loss": 0.07656173408031464, "global_step": 167119, "epoch": 1877, "lr": 3.233398598605934e-05}
{"train_loss": 0.05572163686156273, "global_step": 167120, "epoch": 1877, "lr": 3.233344364225092e-05}
{"train_loss": 0.08366940915584564, "global_step": 167121, "epoch": 1877, "lr": 3.233290130081753e-05}
{"train_loss": 0.10809464752674103, "global_step": 167122, "epoch": 1877, "lr": 3.233235896175929e-05}
{"train_loss": 0.07294860482215881, "global_step": 167123, "epoch": 1877, "lr": 3.233181662507623e-05}
{"train_loss": 0.09190250188112259, "global_step": 167124, "epoch": 1877, "lr": 3.2331274290768456e-05}
{"train_loss": 0.14951017498970032, "global_step": 167125, "epoch": 1877, "lr": 3.2330731958836e-05}
{"train_loss": 0.06496847420930862, "global_step": 167126, "epoch": 1877, "lr": 3.233018962927897e-05}
{"train_loss": 0.10075503587722778, "global_step": 167127, "epoch": 1877, "lr": 3.2329647302097457e-05}
{"train_loss": 0.07072402536869049, "global_step": 167128, "epoch": 1877, "lr": 3.232910497729148e-05}
{"train_loss": 0.11397602409124374, "global_step": 167129, "epoch": 1877, "lr": 3.232856265486116e-05}
{"train_loss": 0.0782582238316536, "global_step": 167130, "epoch": 1877, "lr": 3.232802033480653e-05}
{"train_loss": 0.08151875436306, "global_step": 167131, "epoch": 1877, "lr": 3.23274780171277e-05}
{"train_loss": 0.12007739394903183, "global_step": 167132, "epoch": 1877, "lr": 3.232693570182472e-05}
{"train_loss": 0.05864444002509117, "global_step": 167133, "epoch": 1877, "lr": 3.2326393388897665e-05}
{"train_loss": 0.10470099002122879, "global_step": 167134, "epoch": 1877, "lr": 3.232585107834661e-05}
{"train_loss": 0.07876958698034286, "global_step": 167135, "epoch": 1877, "lr": 3.232530877017163e-05}
{"train_loss": 0.10015816241502762, "global_step": 167136, "epoch": 1877, "lr": 3.23247664643728e-05}
{"train_loss": 0.09875667840242386, "global_step": 167137, "epoch": 1877, "lr": 3.2324224160950196e-05}
{"train_loss": 0.08974441885948181, "global_step": 167138, "epoch": 1877, "lr": 3.232368185990386e-05}
{"train_loss": 0.06938648223876953, "global_step": 167139, "epoch": 1877, "lr": 3.2323139561233916e-05}
{"train_loss": 0.12709695100784302, "global_step": 167140, "epoch": 1877, "lr": 3.232259726494039e-05}
{"train_loss": 0.0839857699328594, "global_step": 167141, "epoch": 1877, "lr": 3.23220549710234e-05, "val_loss": 7.12805700302124}
{"train_loss": 0.10200019180774689, "global_step": 167142, "epoch": 1878, "lr": 3.232151267948297e-05}
{"train_loss": 0.09464850276708603, "global_step": 167143, "epoch": 1878, "lr": 3.2320970390319215e-05}
{"train_loss": 0.10859454423189163, "global_step": 167144, "epoch": 1878, "lr": 3.232042810353217e-05}
{"train_loss": 0.08268103748559952, "global_step": 167145, "epoch": 1878, "lr": 3.231988581912193e-05}
{"train_loss": 0.08083245903253555, "global_step": 167146, "epoch": 1878, "lr": 3.231934353708859e-05}
{"train_loss": 0.05243831127882004, "global_step": 167147, "epoch": 1878, "lr": 3.231880125743217e-05}
{"train_loss": 0.15716858208179474, "global_step": 167148, "epoch": 1878, "lr": 3.231825898015279e-05}
{"train_loss": 0.08788766711950302, "global_step": 167149, "epoch": 1878, "lr": 3.2317716705250495e-05}
{"train_loss": 0.06404045224189758, "global_step": 167150, "epoch": 1878, "lr": 3.231717443272538e-05}
{"train_loss": 0.07168331742286682, "global_step": 167151, "epoch": 1878, "lr": 3.23166321625775e-05}
{"train_loss": 0.048887938261032104, "global_step": 167152, "epoch": 1878, "lr": 3.231608989480693e-05}
{"train_loss": 0.13668540120124817, "global_step": 167153, "epoch": 1878, "lr": 3.231554762941375e-05}
{"train_loss": 0.0417468324303627, "global_step": 167154, "epoch": 1878, "lr": 3.2315005366398025e-05}
{"train_loss": 0.11329352110624313, "global_step": 167155, "epoch": 1878, "lr": 3.231446310575983e-05}
{"train_loss": 0.11005173623561859, "global_step": 167156, "epoch": 1878, "lr": 3.231392084749926e-05}
{"train_loss": 0.10350707173347473, "global_step": 167157, "epoch": 1878, "lr": 3.2313378591616336e-05}
{"train_loss": 0.07619859278202057, "global_step": 167158, "epoch": 1878, "lr": 3.231283633811119e-05}
{"train_loss": 0.1278499960899353, "global_step": 167159, "epoch": 1878, "lr": 3.231229408698385e-05}
{"train_loss": 0.13041743636131287, "global_step": 167160, "epoch": 1878, "lr": 3.231175183823441e-05}
{"train_loss": 0.08924473077058792, "global_step": 167161, "epoch": 1878, "lr": 3.231120959186294e-05}
{"train_loss": 0.1549954116344452, "global_step": 167162, "epoch": 1878, "lr": 3.2310667347869525e-05}
{"train_loss": 0.05908547341823578, "global_step": 167163, "epoch": 1878, "lr": 3.23101251062542e-05}
{"train_loss": 0.10621555149555206, "global_step": 167164, "epoch": 1878, "lr": 3.230958286701707e-05}
{"train_loss": 0.13572685420513153, "global_step": 167165, "epoch": 1878, "lr": 3.230904063015822e-05}
{"train_loss": 0.10603617131710052, "global_step": 167166, "epoch": 1878, "lr": 3.230849839567768e-05}
{"train_loss": 0.07519451528787613, "global_step": 167167, "epoch": 1878, "lr": 3.230795616357557e-05}
{"train_loss": 0.06660345196723938, "global_step": 167168, "epoch": 1878, "lr": 3.230741393385192e-05}
{"train_loss": 0.15879814326763153, "global_step": 167169, "epoch": 1878, "lr": 3.2306871706506835e-05}
{"train_loss": 0.10483598709106445, "global_step": 167170, "epoch": 1878, "lr": 3.230632948154038e-05}
{"train_loss": 0.10437710583209991, "global_step": 167171, "epoch": 1878, "lr": 3.2305787258952605e-05}
{"train_loss": 0.03892882168292999, "global_step": 167172, "epoch": 1878, "lr": 3.230524503874361e-05}
{"train_loss": 0.08466081321239471, "global_step": 167173, "epoch": 1878, "lr": 3.230470282091346e-05}
{"train_loss": 0.10958719253540039, "global_step": 167174, "epoch": 1878, "lr": 3.230416060546223e-05}
{"train_loss": 0.10360033065080643, "global_step": 167175, "epoch": 1878, "lr": 3.230361839238999e-05}
{"train_loss": 0.1304013431072235, "global_step": 167176, "epoch": 1878, "lr": 3.230307618169681e-05}
{"train_loss": 0.16253452003002167, "global_step": 167177, "epoch": 1878, "lr": 3.230253397338277e-05}
{"train_loss": 0.0354292057454586, "global_step": 167178, "epoch": 1878, "lr": 3.2301991767447926e-05}
{"train_loss": 0.08022921532392502, "global_step": 167179, "epoch": 1878, "lr": 3.230144956389238e-05}
{"train_loss": 0.1161033883690834, "global_step": 167180, "epoch": 1878, "lr": 3.230090736271617e-05}
{"train_loss": 0.07539878785610199, "global_step": 167181, "epoch": 1878, "lr": 3.2300365163919386e-05}
{"train_loss": 0.08913987874984741, "global_step": 167182, "epoch": 1878, "lr": 3.2299822967502126e-05}
{"train_loss": 0.09569036215543747, "global_step": 167183, "epoch": 1878, "lr": 3.229928077346441e-05}
{"train_loss": 0.0798109769821167, "global_step": 167184, "epoch": 1878, "lr": 3.229873858180637e-05}
{"train_loss": 0.06746537983417511, "global_step": 167185, "epoch": 1878, "lr": 3.229819639252802e-05}
{"train_loss": 0.04019705578684807, "global_step": 167186, "epoch": 1878, "lr": 3.229765420562949e-05}
{"train_loss": 0.08533649891614914, "global_step": 167187, "epoch": 1878, "lr": 3.2297112021110804e-05}
{"train_loss": 0.09007243067026138, "global_step": 167188, "epoch": 1878, "lr": 3.2296569838972074e-05}
{"train_loss": 0.10661918669939041, "global_step": 167189, "epoch": 1878, "lr": 3.2296027659213336e-05}
{"train_loss": 0.16070441901683807, "global_step": 167190, "epoch": 1878, "lr": 3.22954854818347e-05}
{"train_loss": 0.052836958318948746, "global_step": 167191, "epoch": 1878, "lr": 3.22949433068362e-05}
{"train_loss": 0.08390162885189056, "global_step": 167192, "epoch": 1878, "lr": 3.2294401134217944e-05}
{"train_loss": 0.14871670305728912, "global_step": 167193, "epoch": 1878, "lr": 3.2293858963979985e-05}
{"train_loss": 0.06775050610303879, "global_step": 167194, "epoch": 1878, "lr": 3.22933167961224e-05}
{"train_loss": 0.04354409873485565, "global_step": 167195, "epoch": 1878, "lr": 3.229277463064526e-05}
{"train_loss": 0.11860091239213943, "global_step": 167196, "epoch": 1878, "lr": 3.229223246754866e-05}
{"train_loss": 0.08675312250852585, "global_step": 167197, "epoch": 1878, "lr": 3.2291690306832624e-05}
{"train_loss": 0.06492502242326736, "global_step": 167198, "epoch": 1878, "lr": 3.2291148148497274e-05}
{"train_loss": 0.06389443576335907, "global_step": 167199, "epoch": 1878, "lr": 3.229060599254264e-05}
{"train_loss": 0.07746654748916626, "global_step": 167200, "epoch": 1878, "lr": 3.229006383896883e-05}
{"train_loss": 0.14496955275535583, "global_step": 167201, "epoch": 1878, "lr": 3.228952168777592e-05}
{"train_loss": 0.09277059882879257, "global_step": 167202, "epoch": 1878, "lr": 3.2288979538963945e-05}
{"train_loss": 0.08663888275623322, "global_step": 167203, "epoch": 1878, "lr": 3.228843739253302e-05}
{"train_loss": 0.09860440343618393, "global_step": 167204, "epoch": 1878, "lr": 3.228789524848319e-05}
{"train_loss": 0.03653773292899132, "global_step": 167205, "epoch": 1878, "lr": 3.2287353106814544e-05}
{"train_loss": 0.08282949775457382, "global_step": 167206, "epoch": 1878, "lr": 3.2286810967527136e-05}
{"train_loss": 0.11806492507457733, "global_step": 167207, "epoch": 1878, "lr": 3.228626883062107e-05}
{"train_loss": 0.10924625396728516, "global_step": 167208, "epoch": 1878, "lr": 3.228572669609637e-05}
{"train_loss": 0.14957831799983978, "global_step": 167209, "epoch": 1878, "lr": 3.2285184563953164e-05}
{"train_loss": 0.06911958009004593, "global_step": 167210, "epoch": 1878, "lr": 3.2284642434191495e-05}
{"train_loss": 0.07772955298423767, "global_step": 167211, "epoch": 1878, "lr": 3.2284100306811425e-05}
{"train_loss": 0.080202117562294, "global_step": 167212, "epoch": 1878, "lr": 3.2283558181813056e-05}
{"train_loss": 0.05407823622226715, "global_step": 167213, "epoch": 1878, "lr": 3.228301605919644e-05}
{"train_loss": 0.04980893433094025, "global_step": 167214, "epoch": 1878, "lr": 3.228247393896166e-05}
{"train_loss": 0.07588336616754532, "global_step": 167215, "epoch": 1878, "lr": 3.228193182110879e-05}
{"train_loss": 0.12460442632436752, "global_step": 167216, "epoch": 1878, "lr": 3.228138970563788e-05}
{"train_loss": 0.056703243404626846, "global_step": 167217, "epoch": 1878, "lr": 3.2280847592549026e-05}
{"train_loss": 0.11938509345054626, "global_step": 167218, "epoch": 1878, "lr": 3.228030548184231e-05}
{"train_loss": 0.08996519446372986, "global_step": 167219, "epoch": 1878, "lr": 3.227976337351778e-05}
{"train_loss": 0.04552751034498215, "global_step": 167220, "epoch": 1878, "lr": 3.227922126757553e-05}
{"train_loss": 0.06471951305866241, "global_step": 167221, "epoch": 1878, "lr": 3.2278679164015607e-05}
{"train_loss": 0.09843119233846664, "global_step": 167222, "epoch": 1878, "lr": 3.227813706283812e-05}
{"train_loss": 0.09640417248010635, "global_step": 167223, "epoch": 1878, "lr": 3.22775949640431e-05}
{"train_loss": 0.10983437299728394, "global_step": 167224, "epoch": 1878, "lr": 3.2277052867630666e-05}
{"train_loss": 0.07178100198507309, "global_step": 167225, "epoch": 1878, "lr": 3.227651077360084e-05}
{"train_loss": 0.14140915870666504, "global_step": 167226, "epoch": 1878, "lr": 3.227596868195374e-05}
{"train_loss": 0.09140365570783615, "global_step": 167227, "epoch": 1878, "lr": 3.227542659268941e-05}
{"train_loss": 0.08805284649133682, "global_step": 167228, "epoch": 1878, "lr": 3.227488450580794e-05}
{"train_loss": 0.08879747986793518, "global_step": 167229, "epoch": 1878, "lr": 3.227434242130939e-05}
{"train_loss": 0.0928413002762232, "global_step": 167230, "epoch": 1878, "lr": 3.2273800339193845e-05, "val_loss": 7.008275032043457}
{"train_loss": 0.05948345363140106, "global_step": 167231, "epoch": 1879, "lr": 3.2273258259461364e-05}
{"train_loss": 0.11722716689109802, "global_step": 167232, "epoch": 1879, "lr": 3.227271618211203e-05}
{"train_loss": 0.05097460374236107, "global_step": 167233, "epoch": 1879, "lr": 3.2272174107145915e-05}
{"train_loss": 0.08266168087720871, "global_step": 167234, "epoch": 1879, "lr": 3.22716320345631e-05}
{"train_loss": 0.07251685857772827, "global_step": 167235, "epoch": 1879, "lr": 3.227108996436362e-05}
{"train_loss": 0.0855681300163269, "global_step": 167236, "epoch": 1879, "lr": 3.227054789654759e-05}
{"train_loss": 0.13618510961532593, "global_step": 167237, "epoch": 1879, "lr": 3.227000583111508e-05}
{"train_loss": 0.05513972043991089, "global_step": 167238, "epoch": 1879, "lr": 3.226946376806613e-05}
{"train_loss": 0.04897484555840492, "global_step": 167239, "epoch": 1879, "lr": 3.2268921707400856e-05}
{"train_loss": 0.05743946507573128, "global_step": 167240, "epoch": 1879, "lr": 3.2268379649119295e-05}
{"train_loss": 0.11289876699447632, "global_step": 167241, "epoch": 1879, "lr": 3.226783759322155e-05}
{"train_loss": 0.06967207789421082, "global_step": 167242, "epoch": 1879, "lr": 3.226729553970765e-05}
{"train_loss": 0.06677134335041046, "global_step": 167243, "epoch": 1879, "lr": 3.226675348857773e-05}
{"train_loss": 0.10712837427854538, "global_step": 167244, "epoch": 1879, "lr": 3.226621143983181e-05}
{"train_loss": 0.03134826198220253, "global_step": 167245, "epoch": 1879, "lr": 3.2265669393469986e-05}
{"train_loss": 0.04200273007154465, "global_step": 167246, "epoch": 1879, "lr": 3.226512734949232e-05}
{"train_loss": 0.08138252049684525, "global_step": 167247, "epoch": 1879, "lr": 3.2264585307898904e-05}
{"train_loss": 0.07528014481067657, "global_step": 167248, "epoch": 1879, "lr": 3.2264043268689794e-05}
{"train_loss": 0.08348707109689713, "global_step": 167249, "epoch": 1879, "lr": 3.226350123186506e-05}
{"train_loss": 0.09413061290979385, "global_step": 167250, "epoch": 1879, "lr": 3.226295919742479e-05}
{"train_loss": 0.08367770910263062, "global_step": 167251, "epoch": 1879, "lr": 3.226241716536904e-05}
{"train_loss": 0.11982009559869766, "global_step": 167252, "epoch": 1879, "lr": 3.22618751356979e-05}
{"train_loss": 0.06153922155499458, "global_step": 167253, "epoch": 1879, "lr": 3.226133310841142e-05}
{"train_loss": 0.04942560940980911, "global_step": 167254, "epoch": 1879, "lr": 3.2260791083509713e-05}
{"train_loss": 0.09242033213376999, "global_step": 167255, "epoch": 1879, "lr": 3.22602490609928e-05}
{"train_loss": 0.06505660712718964, "global_step": 167256, "epoch": 1879, "lr": 3.2259707040860806e-05}
{"train_loss": 0.09277696907520294, "global_step": 167257, "epoch": 1879, "lr": 3.225916502311375e-05}
{"train_loss": 0.06943876296281815, "global_step": 167258, "epoch": 1879, "lr": 3.225862300775177e-05}
{"train_loss": 0.11098092049360275, "global_step": 167259, "epoch": 1879, "lr": 3.225808099477487e-05}
{"train_loss": 0.055336352437734604, "global_step": 167260, "epoch": 1879, "lr": 3.2257538984183165e-05}
{"train_loss": 0.09352301806211472, "global_step": 167261, "epoch": 1879, "lr": 3.225699697597671e-05}
{"train_loss": 0.09286917746067047, "global_step": 167262, "epoch": 1879, "lr": 3.22564549701556e-05}
{"train_loss": 0.06457290798425674, "global_step": 167263, "epoch": 1879, "lr": 3.225591296671988e-05}
{"train_loss": 0.04543709754943848, "global_step": 167264, "epoch": 1879, "lr": 3.225537096566965e-05}
{"train_loss": 0.1246475949883461, "global_step": 167265, "epoch": 1879, "lr": 3.225482896700495e-05}
{"train_loss": 0.08873742818832397, "global_step": 167266, "epoch": 1879, "lr": 3.22542869707259e-05}
{"train_loss": 0.17503829300403595, "global_step": 167267, "epoch": 1879, "lr": 3.225374497683252e-05}
{"train_loss": 0.13317856192588806, "global_step": 167268, "epoch": 1879, "lr": 3.2253202985324923e-05}
{"train_loss": 0.07101277261972427, "global_step": 167269, "epoch": 1879, "lr": 3.225266099620316e-05}
{"train_loss": 0.09292016178369522, "global_step": 167270, "epoch": 1879, "lr": 3.225211900946731e-05}
{"train_loss": 0.10766235738992691, "global_step": 167271, "epoch": 1879, "lr": 3.2251577025117446e-05}
{"train_loss": 0.13373561203479767, "global_step": 167272, "epoch": 1879, "lr": 3.2251035043153634e-05}
{"train_loss": 0.09649781882762909, "global_step": 167273, "epoch": 1879, "lr": 3.225049306357597e-05}
{"train_loss": 0.061299487948417664, "global_step": 167274, "epoch": 1879, "lr": 3.224995108638449e-05}
{"train_loss": 0.08789674192667007, "global_step": 167275, "epoch": 1879, "lr": 3.224940911157931e-05}
{"train_loss": 0.060915470123291016, "global_step": 167276, "epoch": 1879, "lr": 3.2248867139160464e-05}
{"train_loss": 0.05191049724817276, "global_step": 167277, "epoch": 1879, "lr": 3.224832516912806e-05}
{"train_loss": 0.062116459012031555, "global_step": 167278, "epoch": 1879, "lr": 3.224778320148213e-05}
{"train_loss": 0.02990019880235195, "global_step": 167279, "epoch": 1879, "lr": 3.224724123622279e-05}
{"train_loss": 0.06486788392066956, "global_step": 167280, "epoch": 1879, "lr": 3.224669927335007e-05}
{"train_loss": 0.0799710676074028, "global_step": 167281, "epoch": 1879, "lr": 3.2246157312864086e-05}
{"train_loss": 0.11165421456098557, "global_step": 167282, "epoch": 1879, "lr": 3.224561535476487e-05}
{"train_loss": 0.07944255322217941, "global_step": 167283, "epoch": 1879, "lr": 3.2245073399052536e-05}
{"train_loss": 0.14155994355678558, "global_step": 167284, "epoch": 1879, "lr": 3.224453144572711e-05}
{"train_loss": 0.12180361151695251, "global_step": 167285, "epoch": 1879, "lr": 3.224398949478872e-05}
{"train_loss": 0.11220032721757889, "global_step": 167286, "epoch": 1879, "lr": 3.224344754623739e-05}
{"train_loss": 0.14335091412067413, "global_step": 167287, "epoch": 1879, "lr": 3.224290560007322e-05}
{"train_loss": 0.09959948062896729, "global_step": 167288, "epoch": 1879, "lr": 3.224236365629627e-05}
{"train_loss": 0.06267169862985611, "global_step": 167289, "epoch": 1879, "lr": 3.224182171490662e-05}
{"train_loss": 0.12282586097717285, "global_step": 167290, "epoch": 1879, "lr": 3.224127977590434e-05}
{"train_loss": 0.06282899528741837, "global_step": 167291, "epoch": 1879, "lr": 3.224073783928949e-05}
{"train_loss": 0.09649680554866791, "global_step": 167292, "epoch": 1879, "lr": 3.224019590506218e-05}
{"train_loss": 0.1755312830209732, "global_step": 167293, "epoch": 1879, "lr": 3.223965397322244e-05}
{"train_loss": 0.14617884159088135, "global_step": 167294, "epoch": 1879, "lr": 3.223911204377038e-05}
{"train_loss": 0.09958096593618393, "global_step": 167295, "epoch": 1879, "lr": 3.2238570116706034e-05}
{"train_loss": 0.10879334807395935, "global_step": 167296, "epoch": 1879, "lr": 3.223802819202953e-05}
{"train_loss": 0.07420480251312256, "global_step": 167297, "epoch": 1879, "lr": 3.223748626974086e-05}
{"train_loss": 0.06090492010116577, "global_step": 167298, "epoch": 1879, "lr": 3.223694434984018e-05}
{"train_loss": 0.13523998856544495, "global_step": 167299, "epoch": 1879, "lr": 3.223640243232751e-05}
{"train_loss": 0.10522473603487015, "global_step": 167300, "epoch": 1879, "lr": 3.223586051720295e-05}
{"train_loss": 0.0713101252913475, "global_step": 167301, "epoch": 1879, "lr": 3.223531860446654e-05}
{"train_loss": 0.10026367008686066, "global_step": 167302, "epoch": 1879, "lr": 3.22347766941184e-05}
{"train_loss": 0.0658826008439064, "global_step": 167303, "epoch": 1879, "lr": 3.2234234786158556e-05}
{"train_loss": 0.07371820509433746, "global_step": 167304, "epoch": 1879, "lr": 3.223369288058713e-05}
{"train_loss": 0.1057400330901146, "global_step": 167305, "epoch": 1879, "lr": 3.223315097740414e-05}
{"train_loss": 0.11423923820257187, "global_step": 167306, "epoch": 1879, "lr": 3.223260907660971e-05}
{"train_loss": 0.1173144057393074, "global_step": 167307, "epoch": 1879, "lr": 3.2232067178203876e-05}
{"train_loss": 0.07390492409467697, "global_step": 167308, "epoch": 1879, "lr": 3.2231525282186715e-05}
{"train_loss": 0.10763295739889145, "global_step": 167309, "epoch": 1879, "lr": 3.223098338855833e-05}
{"train_loss": 0.1144007220864296, "global_step": 167310, "epoch": 1879, "lr": 3.2230441497318754e-05}
{"train_loss": 0.10960053652524948, "global_step": 167311, "epoch": 1879, "lr": 3.2229899608468095e-05}
{"train_loss": 0.11274530738592148, "global_step": 167312, "epoch": 1879, "lr": 3.222935772200639e-05}
{"train_loss": 0.10014238953590393, "global_step": 167313, "epoch": 1879, "lr": 3.2228815837933746e-05}
{"train_loss": 0.08007042855024338, "global_step": 167314, "epoch": 1879, "lr": 3.222827395625021e-05}
{"train_loss": 0.12338526546955109, "global_step": 167315, "epoch": 1879, "lr": 3.222773207695589e-05}
{"train_loss": 0.142278254032135, "global_step": 167316, "epoch": 1879, "lr": 3.22271902000508e-05}
{"train_loss": 0.09752456843852997, "global_step": 167317, "epoch": 1879, "lr": 3.222664832553508e-05}
{"train_loss": 0.13732187449932098, "global_step": 167318, "epoch": 1879, "lr": 3.2226106453408743e-05}
{"train_loss": 0.09139702006588492, "global_step": 167319, "epoch": 1879, "lr": 3.2225564583671905e-05, "val_loss": 7.147754192352295}
{"train_loss": 0.07564160227775574, "global_step": 167320, "epoch": 1880, "lr": 3.2225022716324614e-05}
{"train_loss": 0.13171762228012085, "global_step": 167321, "epoch": 1880, "lr": 3.222448085136697e-05}
{"train_loss": 0.10734599828720093, "global_step": 167322, "epoch": 1880, "lr": 3.222393898879901e-05}
{"train_loss": 0.06947970390319824, "global_step": 167323, "epoch": 1880, "lr": 3.2223397128620834e-05}
{"train_loss": 0.09060826897621155, "global_step": 167324, "epoch": 1880, "lr": 3.22228552708325e-05}
{"train_loss": 0.08959663659334183, "global_step": 167325, "epoch": 1880, "lr": 3.22223134154341e-05}
{"train_loss": 0.0849827229976654, "global_step": 167326, "epoch": 1880, "lr": 3.222177156242568e-05}
{"train_loss": 0.19092534482479095, "global_step": 167327, "epoch": 1880, "lr": 3.2221229711807324e-05}
{"train_loss": 0.062440160661935806, "global_step": 167328, "epoch": 1880, "lr": 3.222068786357911e-05}
{"train_loss": 0.12813448905944824, "global_step": 167329, "epoch": 1880, "lr": 3.2220146017741104e-05}
{"train_loss": 0.15818069875240326, "global_step": 167330, "epoch": 1880, "lr": 3.22196041742934e-05}
{"train_loss": 0.05865393579006195, "global_step": 167331, "epoch": 1880, "lr": 3.221906233323603e-05}
{"train_loss": 0.12561260163784027, "global_step": 167332, "epoch": 1880, "lr": 3.221852049456912e-05}
{"train_loss": 0.11173327267169952, "global_step": 167333, "epoch": 1880, "lr": 3.221797865829268e-05}
{"train_loss": 0.1235428974032402, "global_step": 167334, "epoch": 1880, "lr": 3.221743682440684e-05}
{"train_loss": 0.06218116730451584, "global_step": 167335, "epoch": 1880, "lr": 3.221689499291163e-05}
{"train_loss": 0.11855793744325638, "global_step": 167336, "epoch": 1880, "lr": 3.221635316380716e-05}
{"train_loss": 0.09129588305950165, "global_step": 167337, "epoch": 1880, "lr": 3.221581133709347e-05}
{"train_loss": 0.05058056861162186, "global_step": 167338, "epoch": 1880, "lr": 3.2215269512770665e-05}
{"train_loss": 0.07170035690069199, "global_step": 167339, "epoch": 1880, "lr": 3.221472769083877e-05}
{"train_loss": 0.10563353449106216, "global_step": 167340, "epoch": 1880, "lr": 3.221418587129792e-05}
{"train_loss": 0.13896246254444122, "global_step": 167341, "epoch": 1880, "lr": 3.221364405414813e-05}
{"train_loss": 0.0899655893445015, "global_step": 167342, "epoch": 1880, "lr": 3.2213102239389526e-05}
{"train_loss": 0.1376636028289795, "global_step": 167343, "epoch": 1880, "lr": 3.2212560427022134e-05}
{"train_loss": 0.16249892115592957, "global_step": 167344, "epoch": 1880, "lr": 3.2212018617046045e-05}
{"train_loss": 0.0649632066488266, "global_step": 167345, "epoch": 1880, "lr": 3.2211476809461347e-05}
{"train_loss": 0.06558997929096222, "global_step": 167346, "epoch": 1880, "lr": 3.2210935004268086e-05}
{"train_loss": 0.17068003118038177, "global_step": 167347, "epoch": 1880, "lr": 3.2210393201466366e-05}
{"train_loss": 0.12368801981210709, "global_step": 167348, "epoch": 1880, "lr": 3.220985140105623e-05}
{"train_loss": 0.11478962004184723, "global_step": 167349, "epoch": 1880, "lr": 3.220930960303776e-05}
{"train_loss": 0.049443669617176056, "global_step": 167350, "epoch": 1880, "lr": 3.220876780741103e-05}
{"train_loss": 0.10180750489234924, "global_step": 167351, "epoch": 1880, "lr": 3.2208226014176125e-05}
{"train_loss": 0.08224032074213028, "global_step": 167352, "epoch": 1880, "lr": 3.220768422333309e-05}
{"train_loss": 0.08716558665037155, "global_step": 167353, "epoch": 1880, "lr": 3.2207142434882036e-05}
{"train_loss": 0.04418080300092697, "global_step": 167354, "epoch": 1880, "lr": 3.220660064882299e-05}
{"train_loss": 0.09770290553569794, "global_step": 167355, "epoch": 1880, "lr": 3.220605886515608e-05}
{"train_loss": 0.10579222440719604, "global_step": 167356, "epoch": 1880, "lr": 3.220551708388132e-05}
{"train_loss": 0.035550110042095184, "global_step": 167357, "epoch": 1880, "lr": 3.220497530499884e-05}
{"train_loss": 0.12106636166572571, "global_step": 167358, "epoch": 1880, "lr": 3.220443352850866e-05}
{"train_loss": 0.07590866833925247, "global_step": 167359, "epoch": 1880, "lr": 3.2203891754410884e-05}
{"train_loss": 0.17201030254364014, "global_step": 167360, "epoch": 1880, "lr": 3.220334998270558e-05}
{"train_loss": 0.11302267014980316, "global_step": 167361, "epoch": 1880, "lr": 3.220280821339282e-05}
{"train_loss": 0.12060194462537766, "global_step": 167362, "epoch": 1880, "lr": 3.220226644647266e-05}
{"train_loss": 0.11473114043474197, "global_step": 167363, "epoch": 1880, "lr": 3.220172468194521e-05}
{"train_loss": 0.11488266289234161, "global_step": 167364, "epoch": 1880, "lr": 3.220118291981052e-05}
{"train_loss": 0.10360652953386307, "global_step": 167365, "epoch": 1880, "lr": 3.2200641160068644e-05}
{"train_loss": 0.10994747281074524, "global_step": 167366, "epoch": 1880, "lr": 3.22000994027197e-05}
{"train_loss": 0.06030244007706642, "global_step": 167367, "epoch": 1880, "lr": 3.2199557647763715e-05}
{"train_loss": 0.09588845074176788, "global_step": 167368, "epoch": 1880, "lr": 3.21990158952008e-05}
{"train_loss": 0.08020219206809998, "global_step": 167369, "epoch": 1880, "lr": 3.219847414503099e-05}
{"train_loss": 0.057456403970718384, "global_step": 167370, "epoch": 1880, "lr": 3.2197932397254407e-05}
{"train_loss": 0.0938914343714714, "global_step": 167371, "epoch": 1880, "lr": 3.2197390651871064e-05}
{"train_loss": 0.09070488065481186, "global_step": 167372, "epoch": 1880, "lr": 3.2196848908881085e-05}
{"train_loss": 0.11388792842626572, "global_step": 167373, "epoch": 1880, "lr": 3.219630716828451e-05}
{"train_loss": 0.10321702808141708, "global_step": 167374, "epoch": 1880, "lr": 3.219576543008144e-05}
{"train_loss": 0.09648522734642029, "global_step": 167375, "epoch": 1880, "lr": 3.2195223694271906e-05}
{"train_loss": 0.10742141306400299, "global_step": 167376, "epoch": 1880, "lr": 3.2194681960856035e-05}
{"train_loss": 0.09263388067483902, "global_step": 167377, "epoch": 1880, "lr": 3.219414022983385e-05}
{"train_loss": 0.08506138622760773, "global_step": 167378, "epoch": 1880, "lr": 3.2193598501205466e-05}
{"train_loss": 0.0979425460100174, "global_step": 167379, "epoch": 1880, "lr": 3.219305677497092e-05}
{"train_loss": 0.15229220688343048, "global_step": 167380, "epoch": 1880, "lr": 3.2192515051130305e-05}
{"train_loss": 0.1161300539970398, "global_step": 167381, "epoch": 1880, "lr": 3.2191973329683694e-05}
{"train_loss": 0.08662835508584976, "global_step": 167382, "epoch": 1880, "lr": 3.219143161063115e-05}
{"train_loss": 0.059888582676649094, "global_step": 167383, "epoch": 1880, "lr": 3.219088989397276e-05}
{"train_loss": 0.08567266166210175, "global_step": 167384, "epoch": 1880, "lr": 3.219034817970857e-05}
{"train_loss": 0.05014844983816147, "global_step": 167385, "epoch": 1880, "lr": 3.2189806467838704e-05}
{"train_loss": 0.09108021855354309, "global_step": 167386, "epoch": 1880, "lr": 3.218926475836316e-05}
{"train_loss": 0.05527574568986893, "global_step": 167387, "epoch": 1880, "lr": 3.218872305128209e-05}
{"train_loss": 0.1081356331706047, "global_step": 167388, "epoch": 1880, "lr": 3.218818134659549e-05}
{"train_loss": 0.13027460873126984, "global_step": 167389, "epoch": 1880, "lr": 3.218763964430351e-05}
{"train_loss": 0.10893774032592773, "global_step": 167390, "epoch": 1880, "lr": 3.2187097944406155e-05}
{"train_loss": 0.07248059660196304, "global_step": 167391, "epoch": 1880, "lr": 3.2186556246903545e-05}
{"train_loss": 0.11268346756696701, "global_step": 167392, "epoch": 1880, "lr": 3.218601455179572e-05}
{"train_loss": 0.1308479607105255, "global_step": 167393, "epoch": 1880, "lr": 3.218547285908279e-05}
{"train_loss": 0.08687956631183624, "global_step": 167394, "epoch": 1880, "lr": 3.2184931168764785e-05}
{"train_loss": 0.11459112167358398, "global_step": 167395, "epoch": 1880, "lr": 3.218438948084182e-05}
{"train_loss": 0.12235143035650253, "global_step": 167396, "epoch": 1880, "lr": 3.2183847795313926e-05}
{"train_loss": 0.0531780906021595, "global_step": 167397, "epoch": 1880, "lr": 3.218330611218122e-05}
{"train_loss": 0.08398424834012985, "global_step": 167398, "epoch": 1880, "lr": 3.218276443144372e-05}
{"train_loss": 0.16118505597114563, "global_step": 167399, "epoch": 1880, "lr": 3.2182222753101546e-05}
{"train_loss": 0.13585534691810608, "global_step": 167400, "epoch": 1880, "lr": 3.218168107715476e-05}
{"train_loss": 0.11542148888111115, "global_step": 167401, "epoch": 1880, "lr": 3.218113940360342e-05}
{"train_loss": 0.06251197308301926, "global_step": 167402, "epoch": 1880, "lr": 3.2180597732447624e-05}
{"train_loss": 0.07442562282085419, "global_step": 167403, "epoch": 1880, "lr": 3.218005606368741e-05}
{"train_loss": 0.06813453882932663, "global_step": 167404, "epoch": 1880, "lr": 3.21795143973229e-05}
{"train_loss": 0.10891950875520706, "global_step": 167405, "epoch": 1880, "lr": 3.21789727333541e-05}
{"train_loss": 0.09562283009290695, "global_step": 167406, "epoch": 1880, "lr": 3.217843107178115e-05}
{"train_loss": 0.15731273591518402, "global_step": 167407, "epoch": 1880, "lr": 3.2177889412604066e-05}
{"train_loss": 0.09991425819946138, "global_step": 167408, "epoch": 1880, "lr": 3.217734775582297e-05, "val_loss": 6.92806339263916, "train_action_mse_error": 12.884649276733398}
{"train_loss": 0.13226570188999176, "global_step": 167409, "epoch": 1881, "lr": 3.21768061014379e-05}
{"train_loss": 0.12573795020580292, "global_step": 167410, "epoch": 1881, "lr": 3.217626444944896e-05}
{"train_loss": 0.062271904200315475, "global_step": 167411, "epoch": 1881, "lr": 3.217572279985617e-05}
{"train_loss": 0.16249525547027588, "global_step": 167412, "epoch": 1881, "lr": 3.2175181152659675e-05}
{"train_loss": 0.11491993069648743, "global_step": 167413, "epoch": 1881, "lr": 3.217463950785947e-05}
{"train_loss": 0.0696079209446907, "global_step": 167414, "epoch": 1881, "lr": 3.2174097865455696e-05}
{"train_loss": 0.09462904930114746, "global_step": 167415, "epoch": 1881, "lr": 3.217355622544839e-05}
{"train_loss": 0.08204156160354614, "global_step": 167416, "epoch": 1881, "lr": 3.2173014587837637e-05}
{"train_loss": 0.06564406305551529, "global_step": 167417, "epoch": 1881, "lr": 3.217247295262349e-05}
{"train_loss": 0.10493041574954987, "global_step": 167418, "epoch": 1881, "lr": 3.2171931319806045e-05}
{"train_loss": 0.06278668344020844, "global_step": 167419, "epoch": 1881, "lr": 3.217138968938537e-05}
{"train_loss": 0.08845438063144684, "global_step": 167420, "epoch": 1881, "lr": 3.2170848061361525e-05}
{"train_loss": 0.11832929402589798, "global_step": 167421, "epoch": 1881, "lr": 3.21703064357346e-05}
{"train_loss": 0.058986712247133255, "global_step": 167422, "epoch": 1881, "lr": 3.216976481250466e-05}
{"train_loss": 0.0792420282959938, "global_step": 167423, "epoch": 1881, "lr": 3.2169223191671785e-05}
{"train_loss": 0.08953415602445602, "global_step": 167424, "epoch": 1881, "lr": 3.216868157323603e-05}
{"train_loss": 0.08746613562107086, "global_step": 167425, "epoch": 1881, "lr": 3.2168139957197484e-05}
{"train_loss": 0.04716420918703079, "global_step": 167426, "epoch": 1881, "lr": 3.2167598343556206e-05}
{"train_loss": 0.09242141246795654, "global_step": 167427, "epoch": 1881, "lr": 3.216705673231229e-05}
{"train_loss": 0.09329113364219666, "global_step": 167428, "epoch": 1881, "lr": 3.216651512346578e-05}
{"train_loss": 0.07559513300657272, "global_step": 167429, "epoch": 1881, "lr": 3.216597351701678e-05}
{"train_loss": 0.1061476320028305, "global_step": 167430, "epoch": 1881, "lr": 3.216543191296533e-05}
{"train_loss": 0.08422993123531342, "global_step": 167431, "epoch": 1881, "lr": 3.2164890311311544e-05}
{"train_loss": 0.03980238735675812, "global_step": 167432, "epoch": 1881, "lr": 3.216434871205544e-05}
{"train_loss": 0.09314271807670593, "global_step": 167433, "epoch": 1881, "lr": 3.2163807115197156e-05}
{"train_loss": 0.10224675387144089, "global_step": 167434, "epoch": 1881, "lr": 3.21632655207367e-05}
{"train_loss": 0.09214231371879578, "global_step": 167435, "epoch": 1881, "lr": 3.2162723928674176e-05}
{"train_loss": 0.13934046030044556, "global_step": 167436, "epoch": 1881, "lr": 3.216218233900968e-05}
{"train_loss": 0.1048746332526207, "global_step": 167437, "epoch": 1881, "lr": 3.2161640751743255e-05}
{"train_loss": 0.09667293727397919, "global_step": 167438, "epoch": 1881, "lr": 3.2161099166874974e-05}
{"train_loss": 0.11313795298337936, "global_step": 167439, "epoch": 1881, "lr": 3.216055758440492e-05}
{"train_loss": 0.11660046130418777, "global_step": 167440, "epoch": 1881, "lr": 3.216001600433316e-05}
{"train_loss": 0.05903904139995575, "global_step": 167441, "epoch": 1881, "lr": 3.215947442665977e-05}
{"train_loss": 0.038637466728687286, "global_step": 167442, "epoch": 1881, "lr": 3.215893285138483e-05}
{"train_loss": 0.0798000618815422, "global_step": 167443, "epoch": 1881, "lr": 3.215839127850838e-05}
{"train_loss": 0.16394008696079254, "global_step": 167444, "epoch": 1881, "lr": 3.215784970803054e-05}
{"train_loss": 0.09654074907302856, "global_step": 167445, "epoch": 1881, "lr": 3.2157308139951344e-05}
{"train_loss": 0.025671927258372307, "global_step": 167446, "epoch": 1881, "lr": 3.215676657427089e-05}
{"train_loss": 0.06224449723958969, "global_step": 167447, "epoch": 1881, "lr": 3.215622501098923e-05}
{"train_loss": 0.07116241008043289, "global_step": 167448, "epoch": 1881, "lr": 3.2155683450106465e-05}
{"train_loss": 0.07258529216051102, "global_step": 167449, "epoch": 1881, "lr": 3.215514189162263e-05}
{"train_loss": 0.0921001210808754, "global_step": 167450, "epoch": 1881, "lr": 3.2154600335537844e-05}
{"train_loss": 0.043610695749521255, "global_step": 167451, "epoch": 1881, "lr": 3.215405878185213e-05}
{"train_loss": 0.10763636231422424, "global_step": 167452, "epoch": 1881, "lr": 3.2153517230565603e-05}
{"train_loss": 0.0733983963727951, "global_step": 167453, "epoch": 1881, "lr": 3.2152975681678306e-05}
{"train_loss": 0.07401613146066666, "global_step": 167454, "epoch": 1881, "lr": 3.2152434135190314e-05}
{"train_loss": 0.06673745810985565, "global_step": 167455, "epoch": 1881, "lr": 3.215189259110174e-05}
{"train_loss": 0.08582606911659241, "global_step": 167456, "epoch": 1881, "lr": 3.215135104941261e-05}
{"train_loss": 0.08186903595924377, "global_step": 167457, "epoch": 1881, "lr": 3.215080951012301e-05}
{"train_loss": 0.10423282533884048, "global_step": 167458, "epoch": 1881, "lr": 3.215026797323302e-05}
{"train_loss": 0.07319623976945877, "global_step": 167459, "epoch": 1881, "lr": 3.214972643874271e-05}
{"train_loss": 0.08615526556968689, "global_step": 167460, "epoch": 1881, "lr": 3.214918490665215e-05}
{"train_loss": 0.05273277685046196, "global_step": 167461, "epoch": 1881, "lr": 3.214864337696143e-05}
{"train_loss": 0.10386514663696289, "global_step": 167462, "epoch": 1881, "lr": 3.214810184967058e-05}
{"train_loss": 0.049580249935388565, "global_step": 167463, "epoch": 1881, "lr": 3.214756032477972e-05}
{"train_loss": 0.09452822804450989, "global_step": 167464, "epoch": 1881, "lr": 3.214701880228889e-05}
{"train_loss": 0.0597764253616333, "global_step": 167465, "epoch": 1881, "lr": 3.214647728219819e-05}
{"train_loss": 0.07083936780691147, "global_step": 167466, "epoch": 1881, "lr": 3.214593576450766e-05}
{"train_loss": 0.0626780092716217, "global_step": 167467, "epoch": 1881, "lr": 3.214539424921742e-05}
{"train_loss": 0.08939188718795776, "global_step": 167468, "epoch": 1881, "lr": 3.214485273632748e-05}
{"train_loss": 0.07856597751379013, "global_step": 167469, "epoch": 1881, "lr": 3.214431122583798e-05}
{"train_loss": 0.08748380094766617, "global_step": 167470, "epoch": 1881, "lr": 3.214376971774893e-05}
{"train_loss": 0.10175041109323502, "global_step": 167471, "epoch": 1881, "lr": 3.214322821206043e-05}
{"train_loss": 0.045325785875320435, "global_step": 167472, "epoch": 1881, "lr": 3.214268670877258e-05}
{"train_loss": 0.06365638226270676, "global_step": 167473, "epoch": 1881, "lr": 3.214214520788541e-05}
{"train_loss": 0.0697602778673172, "global_step": 167474, "epoch": 1881, "lr": 3.214160370939902e-05}
{"train_loss": 0.06830593198537827, "global_step": 167475, "epoch": 1881, "lr": 3.2141062213313476e-05}
{"train_loss": 0.1355726569890976, "global_step": 167476, "epoch": 1881, "lr": 3.2140520719628855e-05}
{"train_loss": 0.08583004027605057, "global_step": 167477, "epoch": 1881, "lr": 3.2139979228345205e-05}
{"train_loss": 0.06963834911584854, "global_step": 167478, "epoch": 1881, "lr": 3.213943773946263e-05}
{"train_loss": 0.11769917607307434, "global_step": 167479, "epoch": 1881, "lr": 3.213889625298119e-05}
{"train_loss": 0.07566718757152557, "global_step": 167480, "epoch": 1881, "lr": 3.213835476890096e-05}
{"train_loss": 0.1014380007982254, "global_step": 167481, "epoch": 1881, "lr": 3.2137813287222e-05}
{"train_loss": 0.07941891998052597, "global_step": 167482, "epoch": 1881, "lr": 3.2137271807944405e-05}
{"train_loss": 0.12232696264982224, "global_step": 167483, "epoch": 1881, "lr": 3.213673033106821e-05}
{"train_loss": 0.05265925079584122, "global_step": 167484, "epoch": 1881, "lr": 3.213618885659354e-05}
{"train_loss": 0.12942922115325928, "global_step": 167485, "epoch": 1881, "lr": 3.213564738452043e-05}
{"train_loss": 0.1271704137325287, "global_step": 167486, "epoch": 1881, "lr": 3.213510591484898e-05}
{"train_loss": 0.09222997725009918, "global_step": 167487, "epoch": 1881, "lr": 3.213456444757923e-05}
{"train_loss": 0.050285957753658295, "global_step": 167488, "epoch": 1881, "lr": 3.213402298271128e-05}
{"train_loss": 0.09115572273731232, "global_step": 167489, "epoch": 1881, "lr": 3.213348152024518e-05}
{"train_loss": 0.06280254572629929, "global_step": 167490, "epoch": 1881, "lr": 3.2132940060181015e-05}
{"train_loss": 0.0701047033071518, "global_step": 167491, "epoch": 1881, "lr": 3.2132398602518874e-05}
{"train_loss": 0.07651133090257645, "global_step": 167492, "epoch": 1881, "lr": 3.21318571472588e-05}
{"train_loss": 0.09647360444068909, "global_step": 167493, "epoch": 1881, "lr": 3.21313156944009e-05}
{"train_loss": 0.09082192182540894, "global_step": 167494, "epoch": 1881, "lr": 3.2130774243945214e-05}
{"train_loss": 0.14006668329238892, "global_step": 167495, "epoch": 1881, "lr": 3.213023279589182e-05}
{"train_loss": 0.13053804636001587, "global_step": 167496, "epoch": 1881, "lr": 3.212969135024081e-05}
{"train_loss": 0.08740219944732243, "global_step": 167497, "epoch": 1881, "lr": 3.2129149906992246e-05, "val_loss": 7.119405269622803}
{"train_loss": 0.1063741073012352, "global_step": 167498, "epoch": 1882, "lr": 3.212860846614619e-05}
{"train_loss": 0.04031264781951904, "global_step": 167499, "epoch": 1882, "lr": 3.2128067027702746e-05}
{"train_loss": 0.11140190064907074, "global_step": 167500, "epoch": 1882, "lr": 3.212752559166194e-05}
{"train_loss": 0.07423035055398941, "global_step": 167501, "epoch": 1882, "lr": 3.2126984158023885e-05}
{"train_loss": 0.12349406629800797, "global_step": 167502, "epoch": 1882, "lr": 3.212644272678863e-05}
{"train_loss": 0.07727641612291336, "global_step": 167503, "epoch": 1882, "lr": 3.212590129795627e-05}
{"train_loss": 0.11792165040969849, "global_step": 167504, "epoch": 1882, "lr": 3.212535987152685e-05}
{"train_loss": 0.12698635458946228, "global_step": 167505, "epoch": 1882, "lr": 3.212481844750047e-05}
{"train_loss": 0.05913180857896805, "global_step": 167506, "epoch": 1882, "lr": 3.212427702587717e-05}
{"train_loss": 0.04390601068735123, "global_step": 167507, "epoch": 1882, "lr": 3.212373560665706e-05}
{"train_loss": 0.03895198553800583, "global_step": 167508, "epoch": 1882, "lr": 3.21231941898402e-05}
{"train_loss": 0.05101573467254639, "global_step": 167509, "epoch": 1882, "lr": 3.212265277542664e-05}
{"train_loss": 0.062175653874874115, "global_step": 167510, "epoch": 1882, "lr": 3.2122111363416496e-05}
{"train_loss": 0.06909071654081345, "global_step": 167511, "epoch": 1882, "lr": 3.212156995380979e-05}
{"train_loss": 0.07354021817445755, "global_step": 167512, "epoch": 1882, "lr": 3.212102854660665e-05}
{"train_loss": 0.08704166114330292, "global_step": 167513, "epoch": 1882, "lr": 3.21204871418071e-05}
{"train_loss": 0.1069798544049263, "global_step": 167514, "epoch": 1882, "lr": 3.211994573941124e-05}
{"train_loss": 0.07985211908817291, "global_step": 167515, "epoch": 1882, "lr": 3.2119404339419126e-05}
{"train_loss": 0.13627338409423828, "global_step": 167516, "epoch": 1882, "lr": 3.2118862941830846e-05}
{"train_loss": 0.01394724752753973, "global_step": 167517, "epoch": 1882, "lr": 3.2118321546646466e-05}
{"train_loss": 0.09447196125984192, "global_step": 167518, "epoch": 1882, "lr": 3.211778015386606e-05}
{"train_loss": 0.11390040814876556, "global_step": 167519, "epoch": 1882, "lr": 3.2117238763489703e-05}
{"train_loss": 0.10539057850837708, "global_step": 167520, "epoch": 1882, "lr": 3.211669737551748e-05}
{"train_loss": 0.1246914491057396, "global_step": 167521, "epoch": 1882, "lr": 3.211615598994942e-05}
{"train_loss": 0.12045920640230179, "global_step": 167522, "epoch": 1882, "lr": 3.211561460678565e-05}
{"train_loss": 0.060066014528274536, "global_step": 167523, "epoch": 1882, "lr": 3.2115073226026195e-05}
{"train_loss": 0.09064090251922607, "global_step": 167524, "epoch": 1882, "lr": 3.211453184767117e-05}
{"train_loss": 0.08377817273139954, "global_step": 167525, "epoch": 1882, "lr": 3.211399047172061e-05}
{"train_loss": 0.13439041376113892, "global_step": 167526, "epoch": 1882, "lr": 3.2113449098174606e-05}
{"train_loss": 0.14558187127113342, "global_step": 167527, "epoch": 1882, "lr": 3.211290772703325e-05}
{"train_loss": 0.05634282901883125, "global_step": 167528, "epoch": 1882, "lr": 3.211236635829657e-05}
{"train_loss": 0.10172074288129807, "global_step": 167529, "epoch": 1882, "lr": 3.211182499196469e-05}
{"train_loss": 0.05981883034110069, "global_step": 167530, "epoch": 1882, "lr": 3.211128362803764e-05}
{"train_loss": 0.11299794912338257, "global_step": 167531, "epoch": 1882, "lr": 3.211074226651553e-05}
{"train_loss": 0.11016099900007248, "global_step": 167532, "epoch": 1882, "lr": 3.211020090739838e-05}
{"train_loss": 0.08094317466020584, "global_step": 167533, "epoch": 1882, "lr": 3.210965955068633e-05}
{"train_loss": 0.0892789289355278, "global_step": 167534, "epoch": 1882, "lr": 3.21091181963794e-05}
{"train_loss": 0.06565510481595993, "global_step": 167535, "epoch": 1882, "lr": 3.2108576844477687e-05}
{"train_loss": 0.09757878631353378, "global_step": 167536, "epoch": 1882, "lr": 3.2108035494981254e-05}
{"train_loss": 0.1469150185585022, "global_step": 167537, "epoch": 1882, "lr": 3.210749414789018e-05}
{"train_loss": 0.09258091449737549, "global_step": 167538, "epoch": 1882, "lr": 3.2106952803204527e-05}
{"train_loss": 0.061967555433511734, "global_step": 167539, "epoch": 1882, "lr": 3.2106411460924394e-05}
{"train_loss": 0.10497577488422394, "global_step": 167540, "epoch": 1882, "lr": 3.210587012104982e-05}
{"train_loss": 0.07730511575937271, "global_step": 167541, "epoch": 1882, "lr": 3.21053287835809e-05}
{"train_loss": 0.09818833321332932, "global_step": 167542, "epoch": 1882, "lr": 3.210478744851769e-05}
{"train_loss": 0.12715452909469604, "global_step": 167543, "epoch": 1882, "lr": 3.210424611586029e-05}
{"train_loss": 0.10152631253004074, "global_step": 167544, "epoch": 1882, "lr": 3.210370478560875e-05}
{"train_loss": 0.08211266249418259, "global_step": 167545, "epoch": 1882, "lr": 3.210316345776313e-05}
{"train_loss": 0.05039865896105766, "global_step": 167546, "epoch": 1882, "lr": 3.210262213232355e-05}
{"train_loss": 0.04326114058494568, "global_step": 167547, "epoch": 1882, "lr": 3.210208080929003e-05}
{"train_loss": 0.07176962494850159, "global_step": 167548, "epoch": 1882, "lr": 3.210153948866269e-05}
{"train_loss": 0.07961519807577133, "global_step": 167549, "epoch": 1882, "lr": 3.210099817044156e-05}
{"train_loss": 0.06190985068678856, "global_step": 167550, "epoch": 1882, "lr": 3.210045685462675e-05}
{"train_loss": 0.08935388922691345, "global_step": 167551, "epoch": 1882, "lr": 3.20999155412183e-05}
{"train_loss": 0.05627177283167839, "global_step": 167552, "epoch": 1882, "lr": 3.209937423021632e-05}
{"train_loss": 0.06330399215221405, "global_step": 167553, "epoch": 1882, "lr": 3.2098832921620845e-05}
{"train_loss": 0.061867889016866684, "global_step": 167554, "epoch": 1882, "lr": 3.209829161543197e-05}
{"train_loss": 0.06739997118711472, "global_step": 167555, "epoch": 1882, "lr": 3.2097750311649747e-05}
{"train_loss": 0.07626412808895111, "global_step": 167556, "epoch": 1882, "lr": 3.209720901027428e-05}
{"train_loss": 0.038431406021118164, "global_step": 167557, "epoch": 1882, "lr": 3.209666771130562e-05}
{"train_loss": 0.09783262759447098, "global_step": 167558, "epoch": 1882, "lr": 3.2096126414743845e-05}
{"train_loss": 0.038441289216279984, "global_step": 167559, "epoch": 1882, "lr": 3.2095585120589015e-05}
{"train_loss": 0.06510721147060394, "global_step": 167560, "epoch": 1882, "lr": 3.2095043828841235e-05}
{"train_loss": 0.0849752202630043, "global_step": 167561, "epoch": 1882, "lr": 3.209450253950054e-05}
{"train_loss": 0.060583412647247314, "global_step": 167562, "epoch": 1882, "lr": 3.209396125256702e-05}
{"train_loss": 0.06459848582744598, "global_step": 167563, "epoch": 1882, "lr": 3.209341996804077e-05}
{"train_loss": 0.12373033165931702, "global_step": 167564, "epoch": 1882, "lr": 3.2092878685921825e-05}
{"train_loss": 0.08552881330251694, "global_step": 167565, "epoch": 1882, "lr": 3.209233740621028e-05}
{"train_loss": 0.06056738644838333, "global_step": 167566, "epoch": 1882, "lr": 3.2091796128906196e-05}
{"train_loss": 0.11071165651082993, "global_step": 167567, "epoch": 1882, "lr": 3.209125485400967e-05}
{"train_loss": 0.08269181847572327, "global_step": 167568, "epoch": 1882, "lr": 3.209071358152073e-05}
{"train_loss": 0.028440168127417564, "global_step": 167569, "epoch": 1882, "lr": 3.2090172311439496e-05}
{"train_loss": 0.07366030663251877, "global_step": 167570, "epoch": 1882, "lr": 3.2089631043766006e-05}
{"train_loss": 0.05021055042743683, "global_step": 167571, "epoch": 1882, "lr": 3.2089089778500366e-05}
{"train_loss": 0.083391472697258, "global_step": 167572, "epoch": 1882, "lr": 3.208854851564261e-05}
{"train_loss": 0.13991890847682953, "global_step": 167573, "epoch": 1882, "lr": 3.208800725519284e-05}
{"train_loss": 0.08155341446399689, "global_step": 167574, "epoch": 1882, "lr": 3.2087465997151124e-05}
{"train_loss": 0.0760694295167923, "global_step": 167575, "epoch": 1882, "lr": 3.208692474151752e-05}
{"train_loss": 0.06694690138101578, "global_step": 167576, "epoch": 1882, "lr": 3.208638348829211e-05}
{"train_loss": 0.05511001870036125, "global_step": 167577, "epoch": 1882, "lr": 3.208584223747498e-05}
{"train_loss": 0.0713820829987526, "global_step": 167578, "epoch": 1882, "lr": 3.208530098906617e-05}
{"train_loss": 0.10214121639728546, "global_step": 167579, "epoch": 1882, "lr": 3.2084759743065795e-05}
{"train_loss": 0.08959897607564926, "global_step": 167580, "epoch": 1882, "lr": 3.2084218499473885e-05}
{"train_loss": 0.07598356157541275, "global_step": 167581, "epoch": 1882, "lr": 3.208367725829054e-05}
{"train_loss": 0.06352247297763824, "global_step": 167582, "epoch": 1882, "lr": 3.208313601951584e-05}
{"train_loss": 0.026687031611800194, "global_step": 167583, "epoch": 1882, "lr": 3.208259478314981e-05}
{"train_loss": 0.08353286236524582, "global_step": 167584, "epoch": 1882, "lr": 3.208205354919259e-05}
{"train_loss": 0.05266254022717476, "global_step": 167585, "epoch": 1882, "lr": 3.20815123176442e-05}
{"train_loss": 0.08198573527167018, "global_step": 167586, "epoch": 1882, "lr": 3.208097108850475e-05, "val_loss": 7.1757988929748535}
{"train_loss": 0.07761962711811066, "global_step": 167587, "epoch": 1883, "lr": 3.208042986177428e-05}
{"train_loss": 0.08554936200380325, "global_step": 167588, "epoch": 1883, "lr": 3.207988863745289e-05}
{"train_loss": 0.08775834739208221, "global_step": 167589, "epoch": 1883, "lr": 3.207934741554063e-05}
{"train_loss": 0.09685282409191132, "global_step": 167590, "epoch": 1883, "lr": 3.20788061960376e-05}
{"train_loss": 0.056732725352048874, "global_step": 167591, "epoch": 1883, "lr": 3.2078264978943844e-05}
{"train_loss": 0.12517213821411133, "global_step": 167592, "epoch": 1883, "lr": 3.207772376425945e-05}
{"train_loss": 0.10474194586277008, "global_step": 167593, "epoch": 1883, "lr": 3.2077182551984486e-05}
{"train_loss": 0.07494468986988068, "global_step": 167594, "epoch": 1883, "lr": 3.2076641342119016e-05}
{"train_loss": 0.08571766316890717, "global_step": 167595, "epoch": 1883, "lr": 3.207610013466314e-05}
{"train_loss": 0.10420320183038712, "global_step": 167596, "epoch": 1883, "lr": 3.207555892961691e-05}
{"train_loss": 0.11202602833509445, "global_step": 167597, "epoch": 1883, "lr": 3.207501772698039e-05}
{"train_loss": 0.07958999276161194, "global_step": 167598, "epoch": 1883, "lr": 3.207447652675367e-05}
{"train_loss": 0.13490848243236542, "global_step": 167599, "epoch": 1883, "lr": 3.207393532893683e-05}
{"train_loss": 0.08459343016147614, "global_step": 167600, "epoch": 1883, "lr": 3.207339413352991e-05}
{"train_loss": 0.06711962819099426, "global_step": 167601, "epoch": 1883, "lr": 3.207285294053303e-05}
{"train_loss": 0.08089519292116165, "global_step": 167602, "epoch": 1883, "lr": 3.207231174994622e-05}
{"train_loss": 0.10683492571115494, "global_step": 167603, "epoch": 1883, "lr": 3.2071770561769576e-05}
{"train_loss": 0.08918295800685883, "global_step": 167604, "epoch": 1883, "lr": 3.207122937600315e-05}
{"train_loss": 0.14956973493099213, "global_step": 167605, "epoch": 1883, "lr": 3.2070688192647044e-05}
{"train_loss": 0.08908416330814362, "global_step": 167606, "epoch": 1883, "lr": 3.2070147011701304e-05}
{"train_loss": 0.069366954267025, "global_step": 167607, "epoch": 1883, "lr": 3.206960583316603e-05}
{"train_loss": 0.12320178002119064, "global_step": 167608, "epoch": 1883, "lr": 3.206906465704126e-05}
{"train_loss": 0.0942520871758461, "global_step": 167609, "epoch": 1883, "lr": 3.2068523483327106e-05}
{"train_loss": 0.08242306858301163, "global_step": 167610, "epoch": 1883, "lr": 3.2067982312023617e-05}
{"train_loss": 0.1137588694691658, "global_step": 167611, "epoch": 1883, "lr": 3.206744114313086e-05}
{"train_loss": 0.09036217629909515, "global_step": 167612, "epoch": 1883, "lr": 3.206689997664891e-05}
{"train_loss": 0.06820924580097198, "global_step": 167613, "epoch": 1883, "lr": 3.2066358812577856e-05}
{"train_loss": 0.13973233103752136, "global_step": 167614, "epoch": 1883, "lr": 3.206581765091776e-05}
{"train_loss": 0.06361570209264755, "global_step": 167615, "epoch": 1883, "lr": 3.206527649166872e-05}
{"train_loss": 0.07616522163152695, "global_step": 167616, "epoch": 1883, "lr": 3.206473533483074e-05}
{"train_loss": 0.10408709198236465, "global_step": 167617, "epoch": 1883, "lr": 3.206419418040396e-05}
{"train_loss": 0.09241215139627457, "global_step": 167618, "epoch": 1883, "lr": 3.206365302838844e-05}
{"train_loss": 0.07138819992542267, "global_step": 167619, "epoch": 1883, "lr": 3.206311187878422e-05}
{"train_loss": 0.10924174636602402, "global_step": 167620, "epoch": 1883, "lr": 3.206257073159142e-05}
{"train_loss": 0.09305548667907715, "global_step": 167621, "epoch": 1883, "lr": 3.206202958681007e-05}
{"train_loss": 0.11024940758943558, "global_step": 167622, "epoch": 1883, "lr": 3.206148844444028e-05}
{"train_loss": 0.1121004968881607, "global_step": 167623, "epoch": 1883, "lr": 3.206094730448208e-05}
{"train_loss": 0.05465490370988846, "global_step": 167624, "epoch": 1883, "lr": 3.2060406166935596e-05}
{"train_loss": 0.08520317077636719, "global_step": 167625, "epoch": 1883, "lr": 3.205986503180085e-05}
{"train_loss": 0.08184225857257843, "global_step": 167626, "epoch": 1883, "lr": 3.2059323899077954e-05}
{"train_loss": 0.053497686982154846, "global_step": 167627, "epoch": 1883, "lr": 3.2058782768766936e-05}
{"train_loss": 0.07490226626396179, "global_step": 167628, "epoch": 1883, "lr": 3.205824164086792e-05}
{"train_loss": 0.07780727744102478, "global_step": 167629, "epoch": 1883, "lr": 3.2057700515380954e-05}
{"train_loss": 0.06991592794656754, "global_step": 167630, "epoch": 1883, "lr": 3.2057159392306105e-05}
{"train_loss": 0.0831473246216774, "global_step": 167631, "epoch": 1883, "lr": 3.205661827164345e-05}
{"train_loss": 0.07814153283834457, "global_step": 167632, "epoch": 1883, "lr": 3.205607715339307e-05}
{"train_loss": 0.04230794310569763, "global_step": 167633, "epoch": 1883, "lr": 3.2055536037555025e-05}
{"train_loss": 0.08869808167219162, "global_step": 167634, "epoch": 1883, "lr": 3.205499492412939e-05}
{"train_loss": 0.048883866518735886, "global_step": 167635, "epoch": 1883, "lr": 3.205445381311626e-05}
{"train_loss": 0.0825294554233551, "global_step": 167636, "epoch": 1883, "lr": 3.2053912704515664e-05}
{"train_loss": 0.07701979577541351, "global_step": 167637, "epoch": 1883, "lr": 3.205337159832773e-05}
{"train_loss": 0.12276241183280945, "global_step": 167638, "epoch": 1883, "lr": 3.205283049455248e-05}
{"train_loss": 0.06333008408546448, "global_step": 167639, "epoch": 1883, "lr": 3.2052289393190026e-05}
{"train_loss": 0.09524182975292206, "global_step": 167640, "epoch": 1883, "lr": 3.20517482942404e-05}
{"train_loss": 0.03582645207643509, "global_step": 167641, "epoch": 1883, "lr": 3.205120719770372e-05}
{"train_loss": 0.062016382813453674, "global_step": 167642, "epoch": 1883, "lr": 3.2050666103580015e-05}
{"train_loss": 0.10295503586530685, "global_step": 167643, "epoch": 1883, "lr": 3.20501250118694e-05}
{"train_loss": 0.08014825731515884, "global_step": 167644, "epoch": 1883, "lr": 3.2049583922571914e-05}
{"train_loss": 0.17348669469356537, "global_step": 167645, "epoch": 1883, "lr": 3.204904283568765e-05}
{"train_loss": 0.11494220793247223, "global_step": 167646, "epoch": 1883, "lr": 3.204850175121666e-05}
{"train_loss": 0.05358701944351196, "global_step": 167647, "epoch": 1883, "lr": 3.204796066915906e-05}
{"train_loss": 0.09785313159227371, "global_step": 167648, "epoch": 1883, "lr": 3.204741958951486e-05}
{"train_loss": 0.03775312751531601, "global_step": 167649, "epoch": 1883, "lr": 3.2046878512284184e-05}
{"train_loss": 0.0826830118894577, "global_step": 167650, "epoch": 1883, "lr": 3.2046337437467086e-05}
{"train_loss": 0.10532957315444946, "global_step": 167651, "epoch": 1883, "lr": 3.204579636506363e-05}
{"train_loss": 0.053555361926555634, "global_step": 167652, "epoch": 1883, "lr": 3.204525529507391e-05}
{"train_loss": 0.0619964636862278, "global_step": 167653, "epoch": 1883, "lr": 3.2044714227497966e-05}
{"train_loss": 0.09626178443431854, "global_step": 167654, "epoch": 1883, "lr": 3.204417316233591e-05}
{"train_loss": 0.06375648826360703, "global_step": 167655, "epoch": 1883, "lr": 3.2043632099587785e-05}
{"train_loss": 0.08863607794046402, "global_step": 167656, "epoch": 1883, "lr": 3.204309103925369e-05}
{"train_loss": 0.04585117846727371, "global_step": 167657, "epoch": 1883, "lr": 3.204254998133366e-05}
{"train_loss": 0.054858021438121796, "global_step": 167658, "epoch": 1883, "lr": 3.204200892582781e-05}
{"train_loss": 0.08657795190811157, "global_step": 167659, "epoch": 1883, "lr": 3.204146787273617e-05}
{"train_loss": 0.08038446307182312, "global_step": 167660, "epoch": 1883, "lr": 3.204092682205887e-05}
{"train_loss": 0.03114200010895729, "global_step": 167661, "epoch": 1883, "lr": 3.204038577379592e-05}
{"train_loss": 0.07620999962091446, "global_step": 167662, "epoch": 1883, "lr": 3.2039844727947445e-05}
{"train_loss": 0.1288735419511795, "global_step": 167663, "epoch": 1883, "lr": 3.2039303684513466e-05}
{"train_loss": 0.1028800830245018, "global_step": 167664, "epoch": 1883, "lr": 3.2038762643494106e-05}
{"train_loss": 0.06800501048564911, "global_step": 167665, "epoch": 1883, "lr": 3.2038221604889394e-05}
{"train_loss": 0.0674290806055069, "global_step": 167666, "epoch": 1883, "lr": 3.203768056869945e-05}
{"train_loss": 0.078559011220932, "global_step": 167667, "epoch": 1883, "lr": 3.20371395349243e-05}
{"train_loss": 0.07738304883241653, "global_step": 167668, "epoch": 1883, "lr": 3.2036598503564054e-05}
{"train_loss": 0.08449232578277588, "global_step": 167669, "epoch": 1883, "lr": 3.203605747461876e-05}
{"train_loss": 0.07026425004005432, "global_step": 167670, "epoch": 1883, "lr": 3.2035516448088505e-05}
{"train_loss": 0.060241155326366425, "global_step": 167671, "epoch": 1883, "lr": 3.203497542397335e-05}
{"train_loss": 0.0796046331524849, "global_step": 167672, "epoch": 1883, "lr": 3.203443440227336e-05}
{"train_loss": 0.04320700839161873, "global_step": 167673, "epoch": 1883, "lr": 3.203389338298864e-05}
{"train_loss": 0.12088865786790848, "global_step": 167674, "epoch": 1883, "lr": 3.2033352366119244e-05}
{"train_loss": 0.08508695096949513, "global_step": 167675, "epoch": 1883, "lr": 3.203281135166524e-05, "val_loss": 7.215887546539307}
{"train_loss": 0.10207751393318176, "global_step": 167676, "epoch": 1884, "lr": 3.20322703396267e-05}
{"train_loss": 0.11702810972929001, "global_step": 167677, "epoch": 1884, "lr": 3.203172933000372e-05}
{"train_loss": 0.164666548371315, "global_step": 167678, "epoch": 1884, "lr": 3.203118832279633e-05}
{"train_loss": 0.03959769010543823, "global_step": 167679, "epoch": 1884, "lr": 3.203064731800465e-05}
{"train_loss": 0.1191510558128357, "global_step": 167680, "epoch": 1884, "lr": 3.203010631562871e-05}
{"train_loss": 0.07054261118173599, "global_step": 167681, "epoch": 1884, "lr": 3.202956531566863e-05}
{"train_loss": 0.10456341505050659, "global_step": 167682, "epoch": 1884, "lr": 3.2029024318124426e-05}
{"train_loss": 0.06631738692522049, "global_step": 167683, "epoch": 1884, "lr": 3.202848332299623e-05}
{"train_loss": 0.10340312868356705, "global_step": 167684, "epoch": 1884, "lr": 3.202794233028405e-05}
{"train_loss": 0.0723656713962555, "global_step": 167685, "epoch": 1884, "lr": 3.202740133998803e-05}
{"train_loss": 0.14801515638828278, "global_step": 167686, "epoch": 1884, "lr": 3.202686035210818e-05}
{"train_loss": 0.05400580167770386, "global_step": 167687, "epoch": 1884, "lr": 3.202631936664462e-05}
{"train_loss": 0.0833681970834732, "global_step": 167688, "epoch": 1884, "lr": 3.202577838359739e-05}
{"train_loss": 0.09989173710346222, "global_step": 167689, "epoch": 1884, "lr": 3.2025237402966574e-05}
{"train_loss": 0.076624795794487, "global_step": 167690, "epoch": 1884, "lr": 3.202469642475226e-05}
{"train_loss": 0.0868256539106369, "global_step": 167691, "epoch": 1884, "lr": 3.202415544895449e-05}
{"train_loss": 0.11622240394353867, "global_step": 167692, "epoch": 1884, "lr": 3.2023614475573365e-05}
{"train_loss": 0.10146012157201767, "global_step": 167693, "epoch": 1884, "lr": 3.202307350460893e-05}
{"train_loss": 0.055184073746204376, "global_step": 167694, "epoch": 1884, "lr": 3.2022532536061295e-05}
{"train_loss": 0.11658579856157303, "global_step": 167695, "epoch": 1884, "lr": 3.202199156993049e-05}
{"train_loss": 0.0693909153342247, "global_step": 167696, "epoch": 1884, "lr": 3.202145060621663e-05}
{"train_loss": 0.05417769402265549, "global_step": 167697, "epoch": 1884, "lr": 3.202090964491975e-05}
{"train_loss": 0.08955300599336624, "global_step": 167698, "epoch": 1884, "lr": 3.202036868603995e-05}
{"train_loss": 0.07608753442764282, "global_step": 167699, "epoch": 1884, "lr": 3.2019827729577276e-05}
{"train_loss": 0.0960170328617096, "global_step": 167700, "epoch": 1884, "lr": 3.201928677553184e-05}
{"train_loss": 0.08727722615003586, "global_step": 167701, "epoch": 1884, "lr": 3.201874582390367e-05}
{"train_loss": 0.04014644771814346, "global_step": 167702, "epoch": 1884, "lr": 3.2018204874692885e-05}
{"train_loss": 0.07146486639976501, "global_step": 167703, "epoch": 1884, "lr": 3.201766392789951e-05}
{"train_loss": 0.06416159868240356, "global_step": 167704, "epoch": 1884, "lr": 3.201712298352366e-05}
{"train_loss": 0.09336847066879272, "global_step": 167705, "epoch": 1884, "lr": 3.201658204156537e-05}
{"train_loss": 0.08072162419557571, "global_step": 167706, "epoch": 1884, "lr": 3.201604110202475e-05}
{"train_loss": 0.04913175478577614, "global_step": 167707, "epoch": 1884, "lr": 3.201550016490184e-05}
{"train_loss": 0.060464877635240555, "global_step": 167708, "epoch": 1884, "lr": 3.2014959230196727e-05}
{"train_loss": 0.051389940083026886, "global_step": 167709, "epoch": 1884, "lr": 3.2014418297909496e-05}
{"train_loss": 0.04630730673670769, "global_step": 167710, "epoch": 1884, "lr": 3.201387736804019e-05}
{"train_loss": 0.046770934015512466, "global_step": 167711, "epoch": 1884, "lr": 3.201333644058892e-05}
{"train_loss": 0.09128905832767487, "global_step": 167712, "epoch": 1884, "lr": 3.2012795515555717e-05}
{"train_loss": 0.13575425744056702, "global_step": 167713, "epoch": 1884, "lr": 3.2012254592940694e-05}
{"train_loss": 0.11700217425823212, "global_step": 167714, "epoch": 1884, "lr": 3.201171367274388e-05}
{"train_loss": 0.06055856868624687, "global_step": 167715, "epoch": 1884, "lr": 3.20111727549654e-05}
{"train_loss": 0.05242493748664856, "global_step": 167716, "epoch": 1884, "lr": 3.2010631839605274e-05}
{"train_loss": 0.11500705778598785, "global_step": 167717, "epoch": 1884, "lr": 3.201009092666362e-05}
{"train_loss": 0.0432630255818367, "global_step": 167718, "epoch": 1884, "lr": 3.2009550016140466e-05}
{"train_loss": 0.10137300938367844, "global_step": 167719, "epoch": 1884, "lr": 3.200900910803594e-05}
{"train_loss": 0.04803730919957161, "global_step": 167720, "epoch": 1884, "lr": 3.200846820235005e-05}
{"train_loss": 0.12425898015499115, "global_step": 167721, "epoch": 1884, "lr": 3.200792729908293e-05}
{"train_loss": 0.058372486382722855, "global_step": 167722, "epoch": 1884, "lr": 3.2007386398234596e-05}
{"train_loss": 0.12754219770431519, "global_step": 167723, "epoch": 1884, "lr": 3.2006845499805183e-05}
{"train_loss": 0.10121671855449677, "global_step": 167724, "epoch": 1884, "lr": 3.20063046037947e-05}
{"train_loss": 0.08587870746850967, "global_step": 167725, "epoch": 1884, "lr": 3.200576371020326e-05}
{"train_loss": 0.10270459949970245, "global_step": 167726, "epoch": 1884, "lr": 3.200522281903094e-05}
{"train_loss": 0.0388106070458889, "global_step": 167727, "epoch": 1884, "lr": 3.200468193027778e-05}
{"train_loss": 0.08195885270833969, "global_step": 167728, "epoch": 1884, "lr": 3.200414104394389e-05}
{"train_loss": 0.06833358108997345, "global_step": 167729, "epoch": 1884, "lr": 3.2003600160029304e-05}
{"train_loss": 0.06814074516296387, "global_step": 167730, "epoch": 1884, "lr": 3.200305927853413e-05}
{"train_loss": 0.08434370160102844, "global_step": 167731, "epoch": 1884, "lr": 3.200251839945841e-05}
{"train_loss": 0.03035126067698002, "global_step": 167732, "epoch": 1884, "lr": 3.2001977522802254e-05}
{"train_loss": 0.0679408460855484, "global_step": 167733, "epoch": 1884, "lr": 3.200143664856569e-05}
{"train_loss": 0.16397704184055328, "global_step": 167734, "epoch": 1884, "lr": 3.200089577674883e-05}
{"train_loss": 0.09991942346096039, "global_step": 167735, "epoch": 1884, "lr": 3.200035490735171e-05}
{"train_loss": 0.1148400828242302, "global_step": 167736, "epoch": 1884, "lr": 3.1999814040374445e-05}
{"train_loss": 0.07549526542425156, "global_step": 167737, "epoch": 1884, "lr": 3.199927317581707e-05}
{"train_loss": 0.0856216698884964, "global_step": 167738, "epoch": 1884, "lr": 3.199873231367969e-05}
{"train_loss": 0.13045130670070648, "global_step": 167739, "epoch": 1884, "lr": 3.199819145396234e-05}
{"train_loss": 0.058611221611499786, "global_step": 167740, "epoch": 1884, "lr": 3.199765059666513e-05}
{"train_loss": 0.11636810749769211, "global_step": 167741, "epoch": 1884, "lr": 3.19971097417881e-05}
{"train_loss": 0.07991578429937363, "global_step": 167742, "epoch": 1884, "lr": 3.199656888933137e-05}
{"train_loss": 0.21206961572170258, "global_step": 167743, "epoch": 1884, "lr": 3.199602803929494e-05}
{"train_loss": 0.0632346123456955, "global_step": 167744, "epoch": 1884, "lr": 3.1995487191678955e-05}
{"train_loss": 0.07642509043216705, "global_step": 167745, "epoch": 1884, "lr": 3.1994946346483455e-05}
{"train_loss": 0.07851264625787735, "global_step": 167746, "epoch": 1884, "lr": 3.1994405503708505e-05}
{"train_loss": 0.06168201565742493, "global_step": 167747, "epoch": 1884, "lr": 3.19938646633542e-05}
{"train_loss": 0.07629549503326416, "global_step": 167748, "epoch": 1884, "lr": 3.1993323825420583e-05}
{"train_loss": 0.09455899149179459, "global_step": 167749, "epoch": 1884, "lr": 3.199278298990777e-05}
{"train_loss": 0.03631962463259697, "global_step": 167750, "epoch": 1884, "lr": 3.199224215681577e-05}
{"train_loss": 0.09809865057468414, "global_step": 167751, "epoch": 1884, "lr": 3.199170132614473e-05}
{"train_loss": 0.04530097544193268, "global_step": 167752, "epoch": 1884, "lr": 3.199116049789466e-05}
{"train_loss": 0.057894013822078705, "global_step": 167753, "epoch": 1884, "lr": 3.199061967206568e-05}
{"train_loss": 0.02535015530884266, "global_step": 167754, "epoch": 1884, "lr": 3.1990078848657824e-05}
{"train_loss": 0.0641930028796196, "global_step": 167755, "epoch": 1884, "lr": 3.19895380276712e-05}
{"train_loss": 0.17415858805179596, "global_step": 167756, "epoch": 1884, "lr": 3.1988997209105845e-05}
{"train_loss": 0.07081442326307297, "global_step": 167757, "epoch": 1884, "lr": 3.198845639296186e-05}
{"train_loss": 0.129961758852005, "global_step": 167758, "epoch": 1884, "lr": 3.19879155792393e-05}
{"train_loss": 0.10175029188394547, "global_step": 167759, "epoch": 1884, "lr": 3.198737476793826e-05}
{"train_loss": 0.20587284862995148, "global_step": 167760, "epoch": 1884, "lr": 3.198683395905878e-05}
{"train_loss": 0.10990867018699646, "global_step": 167761, "epoch": 1884, "lr": 3.198629315260096e-05}
{"train_loss": 0.042645782232284546, "global_step": 167762, "epoch": 1884, "lr": 3.1985752348564856e-05}
{"train_loss": 0.062029168009757996, "global_step": 167763, "epoch": 1884, "lr": 3.198521154695056e-05}
{"train_loss": 0.08647605425186372, "global_step": 167764, "epoch": 1884, "lr": 3.198467074775813e-05, "val_loss": 7.41274356842041}
{"train_loss": 0.08061520755290985, "global_step": 167765, "epoch": 1885, "lr": 3.198412995098764e-05}
{"train_loss": 0.05767560005187988, "global_step": 167766, "epoch": 1885, "lr": 3.1983589156639174e-05}
{"train_loss": 0.16828341782093048, "global_step": 167767, "epoch": 1885, "lr": 3.198304836471276e-05}
{"train_loss": 0.0903097540140152, "global_step": 167768, "epoch": 1885, "lr": 3.198250757520855e-05}
{"train_loss": 0.0788547694683075, "global_step": 167769, "epoch": 1885, "lr": 3.1981966788126536e-05}
{"train_loss": 0.072840616106987, "global_step": 167770, "epoch": 1885, "lr": 3.1981426003466855e-05}
{"train_loss": 0.1155194640159607, "global_step": 167771, "epoch": 1885, "lr": 3.1980885221229525e-05}
{"train_loss": 0.10105744749307632, "global_step": 167772, "epoch": 1885, "lr": 3.1980344441414676e-05}
{"train_loss": 0.080793596804142, "global_step": 167773, "epoch": 1885, "lr": 3.197980366402232e-05}
{"train_loss": 0.08969210833311081, "global_step": 167774, "epoch": 1885, "lr": 3.197926288905259e-05}
{"train_loss": 0.0921456590294838, "global_step": 167775, "epoch": 1885, "lr": 3.19787221165055e-05}
{"train_loss": 0.13656730949878693, "global_step": 167776, "epoch": 1885, "lr": 3.197818134638117e-05}
{"train_loss": 0.031928617507219315, "global_step": 167777, "epoch": 1885, "lr": 3.1977640578679635e-05}
{"train_loss": 0.09545572102069855, "global_step": 167778, "epoch": 1885, "lr": 3.197709981340101e-05}
{"train_loss": 0.10389438271522522, "global_step": 167779, "epoch": 1885, "lr": 3.1976559050545325e-05}
{"train_loss": 0.09279768168926239, "global_step": 167780, "epoch": 1885, "lr": 3.197601829011268e-05}
{"train_loss": 0.07690512388944626, "global_step": 167781, "epoch": 1885, "lr": 3.197547753210314e-05}
{"train_loss": 0.09749562293291092, "global_step": 167782, "epoch": 1885, "lr": 3.197493677651677e-05}
{"train_loss": 0.08775532245635986, "global_step": 167783, "epoch": 1885, "lr": 3.197439602335366e-05}
{"train_loss": 0.0960329994559288, "global_step": 167784, "epoch": 1885, "lr": 3.1973855272613874e-05}
{"train_loss": 0.1394650638103485, "global_step": 167785, "epoch": 1885, "lr": 3.197331452429748e-05}
{"train_loss": 0.07485383749008179, "global_step": 167786, "epoch": 1885, "lr": 3.1972773778404555e-05}
{"train_loss": 0.081862673163414, "global_step": 167787, "epoch": 1885, "lr": 3.197223303493517e-05}
{"train_loss": 0.06997904181480408, "global_step": 167788, "epoch": 1885, "lr": 3.1971692293889386e-05}
{"train_loss": 0.04732105880975723, "global_step": 167789, "epoch": 1885, "lr": 3.1971151555267315e-05}
{"train_loss": 0.06596040725708008, "global_step": 167790, "epoch": 1885, "lr": 3.197061081906898e-05}
{"train_loss": 0.0993988886475563, "global_step": 167791, "epoch": 1885, "lr": 3.197007008529449e-05}
{"train_loss": 0.08927030116319656, "global_step": 167792, "epoch": 1885, "lr": 3.196952935394389e-05}
{"train_loss": 0.11879602074623108, "global_step": 167793, "epoch": 1885, "lr": 3.196898862501728e-05}
{"train_loss": 0.06346132606267929, "global_step": 167794, "epoch": 1885, "lr": 3.196844789851471e-05}
{"train_loss": 0.14584100246429443, "global_step": 167795, "epoch": 1885, "lr": 3.196790717443628e-05}
{"train_loss": 0.06396692991256714, "global_step": 167796, "epoch": 1885, "lr": 3.196736645278202e-05}
{"train_loss": 0.08564215153455734, "global_step": 167797, "epoch": 1885, "lr": 3.1966825733552056e-05}
{"train_loss": 0.0750202164053917, "global_step": 167798, "epoch": 1885, "lr": 3.19662850167464e-05}
{"train_loss": 0.030089065432548523, "global_step": 167799, "epoch": 1885, "lr": 3.196574430236517e-05}
{"train_loss": 0.06384250521659851, "global_step": 167800, "epoch": 1885, "lr": 3.196520359040844e-05}
{"train_loss": 0.07904445379972458, "global_step": 167801, "epoch": 1885, "lr": 3.1964662880876256e-05}
{"train_loss": 0.05685695633292198, "global_step": 167802, "epoch": 1885, "lr": 3.196412217376871e-05}
{"train_loss": 0.07931431382894516, "global_step": 167803, "epoch": 1885, "lr": 3.196358146908586e-05}
{"train_loss": 0.08822409063577652, "global_step": 167804, "epoch": 1885, "lr": 3.19630407668278e-05}
{"train_loss": 0.11704237014055252, "global_step": 167805, "epoch": 1885, "lr": 3.196250006699457e-05}
{"train_loss": 0.07877883315086365, "global_step": 167806, "epoch": 1885, "lr": 3.1961959369586283e-05}
{"train_loss": 0.09760186076164246, "global_step": 167807, "epoch": 1885, "lr": 3.196141867460297e-05}
{"train_loss": 0.15022042393684387, "global_step": 167808, "epoch": 1885, "lr": 3.1960877982044745e-05}
{"train_loss": 0.11776959151029587, "global_step": 167809, "epoch": 1885, "lr": 3.196033729191164e-05}
{"train_loss": 0.10009071230888367, "global_step": 167810, "epoch": 1885, "lr": 3.195979660420376e-05}
{"train_loss": 0.06497394293546677, "global_step": 167811, "epoch": 1885, "lr": 3.195925591892116e-05}
{"train_loss": 0.14051492512226105, "global_step": 167812, "epoch": 1885, "lr": 3.195871523606393e-05}
{"train_loss": 0.08804819732904434, "global_step": 167813, "epoch": 1885, "lr": 3.195817455563211e-05}
{"train_loss": 0.07493223994970322, "global_step": 167814, "epoch": 1885, "lr": 3.1957633877625816e-05}
{"train_loss": 0.07535875588655472, "global_step": 167815, "epoch": 1885, "lr": 3.195709320204507e-05}
{"train_loss": 0.07413674145936966, "global_step": 167816, "epoch": 1885, "lr": 3.195655252888998e-05}
{"train_loss": 0.10045609623193741, "global_step": 167817, "epoch": 1885, "lr": 3.1956011858160636e-05}
{"train_loss": 0.16700753569602966, "global_step": 167818, "epoch": 1885, "lr": 3.195547118985707e-05}
{"train_loss": 0.05241359397768974, "global_step": 167819, "epoch": 1885, "lr": 3.1954930523979375e-05}
{"train_loss": 0.10460101813077927, "global_step": 167820, "epoch": 1885, "lr": 3.1954389860527615e-05}
{"train_loss": 0.06353411823511124, "global_step": 167821, "epoch": 1885, "lr": 3.195384919950187e-05}
{"train_loss": 0.051765650510787964, "global_step": 167822, "epoch": 1885, "lr": 3.1953308540902216e-05}
{"train_loss": 0.052585668861866, "global_step": 167823, "epoch": 1885, "lr": 3.195276788472872e-05}
{"train_loss": 0.06443024426698685, "global_step": 167824, "epoch": 1885, "lr": 3.1952227230981435e-05}
{"train_loss": 0.10291208326816559, "global_step": 167825, "epoch": 1885, "lr": 3.1951686579660475e-05}
{"train_loss": 0.11374622583389282, "global_step": 167826, "epoch": 1885, "lr": 3.195114593076588e-05}
{"train_loss": 0.08904781192541122, "global_step": 167827, "epoch": 1885, "lr": 3.1950605284297744e-05}
{"train_loss": 0.05747292563319206, "global_step": 167828, "epoch": 1885, "lr": 3.1950064640256104e-05}
{"train_loss": 0.07260224223136902, "global_step": 167829, "epoch": 1885, "lr": 3.194952399864109e-05}
{"train_loss": 0.06514386832714081, "global_step": 167830, "epoch": 1885, "lr": 3.194898335945271e-05}
{"train_loss": 0.08460771292448044, "global_step": 167831, "epoch": 1885, "lr": 3.19484427226911e-05}
{"train_loss": 0.08857323974370956, "global_step": 167832, "epoch": 1885, "lr": 3.194790208835628e-05}
{"train_loss": 0.14042741060256958, "global_step": 167833, "epoch": 1885, "lr": 3.194736145644836e-05}
{"train_loss": 0.09351399540901184, "global_step": 167834, "epoch": 1885, "lr": 3.1946820826967386e-05}
{"train_loss": 0.14614823460578918, "global_step": 167835, "epoch": 1885, "lr": 3.194628019991343e-05}
{"train_loss": 0.11760878562927246, "global_step": 167836, "epoch": 1885, "lr": 3.1945739575286605e-05}
{"train_loss": 0.06563930958509445, "global_step": 167837, "epoch": 1885, "lr": 3.194519895308694e-05}
{"train_loss": 0.1110866367816925, "global_step": 167838, "epoch": 1885, "lr": 3.194465833331452e-05}
{"train_loss": 0.05963609740138054, "global_step": 167839, "epoch": 1885, "lr": 3.194411771596943e-05}
{"train_loss": 0.07434863597154617, "global_step": 167840, "epoch": 1885, "lr": 3.194357710105173e-05}
{"train_loss": 0.10284781455993652, "global_step": 167841, "epoch": 1885, "lr": 3.1943036488561496e-05}
{"train_loss": 0.107091523706913, "global_step": 167842, "epoch": 1885, "lr": 3.194249587849881e-05}
{"train_loss": 0.09462623298168182, "global_step": 167843, "epoch": 1885, "lr": 3.194195527086371e-05}
{"train_loss": 0.063310407102108, "global_step": 167844, "epoch": 1885, "lr": 3.194141466565632e-05}
{"train_loss": 0.05203770846128464, "global_step": 167845, "epoch": 1885, "lr": 3.194087406287667e-05}
{"train_loss": 0.07387902587652206, "global_step": 167846, "epoch": 1885, "lr": 3.194033346252486e-05}
{"train_loss": 0.07316399365663528, "global_step": 167847, "epoch": 1885, "lr": 3.193979286460094e-05}
{"train_loss": 0.08509491384029388, "global_step": 167848, "epoch": 1885, "lr": 3.193925226910501e-05}
{"train_loss": 0.08170287311077118, "global_step": 167849, "epoch": 1885, "lr": 3.193871167603712e-05}
{"train_loss": 0.052185434848070145, "global_step": 167850, "epoch": 1885, "lr": 3.193817108539735e-05}
{"train_loss": 0.06298340857028961, "global_step": 167851, "epoch": 1885, "lr": 3.193763049718577e-05}
{"train_loss": 0.13858667016029358, "global_step": 167852, "epoch": 1885, "lr": 3.193708991140244e-05}
{"train_loss": 0.08847206429149328, "global_step": 167853, "epoch": 1885, "lr": 3.1936549328047474e-05, "val_loss": 7.116662502288818, "train_action_mse_error": 21.189411163330078}
{"train_loss": 0.11712206900119781, "global_step": 167854, "epoch": 1886, "lr": 3.193600874712091e-05}
{"train_loss": 0.1182192713022232, "global_step": 167855, "epoch": 1886, "lr": 3.193546816862283e-05}
{"train_loss": 0.04411279782652855, "global_step": 167856, "epoch": 1886, "lr": 3.193492759255331e-05}
{"train_loss": 0.09992988407611847, "global_step": 167857, "epoch": 1886, "lr": 3.193438701891242e-05}
{"train_loss": 0.05857308581471443, "global_step": 167858, "epoch": 1886, "lr": 3.193384644770022e-05}
{"train_loss": 0.08219324797391891, "global_step": 167859, "epoch": 1886, "lr": 3.19333058789168e-05}
{"train_loss": 0.06446864455938339, "global_step": 167860, "epoch": 1886, "lr": 3.193276531256223e-05}
{"train_loss": 0.08513706177473068, "global_step": 167861, "epoch": 1886, "lr": 3.193222474863658e-05}
{"train_loss": 0.10032685846090317, "global_step": 167862, "epoch": 1886, "lr": 3.1931684187139916e-05}
{"train_loss": 0.09780770540237427, "global_step": 167863, "epoch": 1886, "lr": 3.193114362807233e-05}
{"train_loss": 0.05848948284983635, "global_step": 167864, "epoch": 1886, "lr": 3.193060307143386e-05}
{"train_loss": 0.11647695302963257, "global_step": 167865, "epoch": 1886, "lr": 3.193006251722463e-05}
{"train_loss": 0.07199641317129135, "global_step": 167866, "epoch": 1886, "lr": 3.192952196544465e-05}
{"train_loss": 0.053923167288303375, "global_step": 167867, "epoch": 1886, "lr": 3.192898141609405e-05}
{"train_loss": 0.07224202901124954, "global_step": 167868, "epoch": 1886, "lr": 3.192844086917286e-05}
{"train_loss": 0.060872867703437805, "global_step": 167869, "epoch": 1886, "lr": 3.192790032468118e-05}
{"train_loss": 0.10065525770187378, "global_step": 167870, "epoch": 1886, "lr": 3.192735978261906e-05}
{"train_loss": 0.12559890747070312, "global_step": 167871, "epoch": 1886, "lr": 3.192681924298659e-05}
{"train_loss": 0.0913008451461792, "global_step": 167872, "epoch": 1886, "lr": 3.192627870578386e-05}
{"train_loss": 0.07629784196615219, "global_step": 167873, "epoch": 1886, "lr": 3.19257381710109e-05}
{"train_loss": 0.08308633416891098, "global_step": 167874, "epoch": 1886, "lr": 3.1925197638667825e-05}
{"train_loss": 0.09210435301065445, "global_step": 167875, "epoch": 1886, "lr": 3.1924657108754665e-05}
{"train_loss": 0.08672495186328888, "global_step": 167876, "epoch": 1886, "lr": 3.1924116581271534e-05}
{"train_loss": 0.07173440605401993, "global_step": 167877, "epoch": 1886, "lr": 3.192357605621847e-05}
{"train_loss": 0.13088716566562653, "global_step": 167878, "epoch": 1886, "lr": 3.1923035533595576e-05}
{"train_loss": 0.09229724109172821, "global_step": 167879, "epoch": 1886, "lr": 3.19224950134029e-05}
{"train_loss": 0.08641090244054794, "global_step": 167880, "epoch": 1886, "lr": 3.192195449564053e-05}
{"train_loss": 0.11891665309667587, "global_step": 167881, "epoch": 1886, "lr": 3.1921413980308524e-05}
{"train_loss": 0.09063445031642914, "global_step": 167882, "epoch": 1886, "lr": 3.192087346740698e-05}
{"train_loss": 0.11314880847930908, "global_step": 167883, "epoch": 1886, "lr": 3.192033295693593e-05}
{"train_loss": 0.12268438935279846, "global_step": 167884, "epoch": 1886, "lr": 3.191979244889549e-05}
{"train_loss": 0.044097308069467545, "global_step": 167885, "epoch": 1886, "lr": 3.191925194328569e-05}
{"train_loss": 0.05424528196454048, "global_step": 167886, "epoch": 1886, "lr": 3.191871144010665e-05}
{"train_loss": 0.054908256977796555, "global_step": 167887, "epoch": 1886, "lr": 3.191817093935841e-05}
{"train_loss": 0.09898699074983597, "global_step": 167888, "epoch": 1886, "lr": 3.191763044104103e-05}
{"train_loss": 0.0821308046579361, "global_step": 167889, "epoch": 1886, "lr": 3.191708994515464e-05}
{"train_loss": 0.035142190754413605, "global_step": 167890, "epoch": 1886, "lr": 3.191654945169926e-05}
{"train_loss": 0.11549447476863861, "global_step": 167891, "epoch": 1886, "lr": 3.191600896067499e-05}
{"train_loss": 0.05200155824422836, "global_step": 167892, "epoch": 1886, "lr": 3.1915468472081875e-05}
{"train_loss": 0.06748098880052567, "global_step": 167893, "epoch": 1886, "lr": 3.191492798592003e-05}
{"train_loss": 0.07241921871900558, "global_step": 167894, "epoch": 1886, "lr": 3.191438750218947e-05}
{"train_loss": 0.06725287437438965, "global_step": 167895, "epoch": 1886, "lr": 3.191384702089033e-05}
{"train_loss": 0.07563380151987076, "global_step": 167896, "epoch": 1886, "lr": 3.191330654202264e-05}
{"train_loss": 0.05775294825434685, "global_step": 167897, "epoch": 1886, "lr": 3.1912766065586484e-05}
{"train_loss": 0.066689133644104, "global_step": 167898, "epoch": 1886, "lr": 3.191222559158194e-05}
{"train_loss": 0.0687827467918396, "global_step": 167899, "epoch": 1886, "lr": 3.191168512000907e-05}
{"train_loss": 0.10505744814872742, "global_step": 167900, "epoch": 1886, "lr": 3.1911144650867964e-05}
{"train_loss": 0.10074254870414734, "global_step": 167901, "epoch": 1886, "lr": 3.191060418415869e-05}
{"train_loss": 0.07701373845338821, "global_step": 167902, "epoch": 1886, "lr": 3.191006371988129e-05}
{"train_loss": 0.07748878002166748, "global_step": 167903, "epoch": 1886, "lr": 3.190952325803589e-05}
{"train_loss": 0.16847138106822968, "global_step": 167904, "epoch": 1886, "lr": 3.190898279862251e-05}
{"train_loss": 0.11858761310577393, "global_step": 167905, "epoch": 1886, "lr": 3.190844234164126e-05}
{"train_loss": 0.06746416538953781, "global_step": 167906, "epoch": 1886, "lr": 3.190790188709219e-05}
{"train_loss": 0.05664435029029846, "global_step": 167907, "epoch": 1886, "lr": 3.190736143497537e-05}
{"train_loss": 0.06110282242298126, "global_step": 167908, "epoch": 1886, "lr": 3.1906820985290915e-05}
{"train_loss": 0.12451109290122986, "global_step": 167909, "epoch": 1886, "lr": 3.190628053803885e-05}
{"train_loss": 0.05727405101060867, "global_step": 167910, "epoch": 1886, "lr": 3.1905740093219274e-05}
{"train_loss": 0.0821952298283577, "global_step": 167911, "epoch": 1886, "lr": 3.190519965083224e-05}
{"train_loss": 0.08352242410182953, "global_step": 167912, "epoch": 1886, "lr": 3.190465921087784e-05}
{"train_loss": 0.09724670648574829, "global_step": 167913, "epoch": 1886, "lr": 3.190411877335613e-05}
{"train_loss": 0.1638193130493164, "global_step": 167914, "epoch": 1886, "lr": 3.1903578338267206e-05}
{"train_loss": 0.11979030817747116, "global_step": 167915, "epoch": 1886, "lr": 3.1903037905611114e-05}
{"train_loss": 0.09807252883911133, "global_step": 167916, "epoch": 1886, "lr": 3.190249747538794e-05}
{"train_loss": 0.10109429061412811, "global_step": 167917, "epoch": 1886, "lr": 3.190195704759775e-05}
{"train_loss": 0.10720368474721909, "global_step": 167918, "epoch": 1886, "lr": 3.1901416622240625e-05}
{"train_loss": 0.1068444475531578, "global_step": 167919, "epoch": 1886, "lr": 3.190087619931663e-05}
{"train_loss": 0.07164227217435837, "global_step": 167920, "epoch": 1886, "lr": 3.190033577882586e-05}
{"train_loss": 0.04046483337879181, "global_step": 167921, "epoch": 1886, "lr": 3.189979536076835e-05}
{"train_loss": 0.10217557102441788, "global_step": 167922, "epoch": 1886, "lr": 3.18992549451442e-05}
{"train_loss": 0.1091969758272171, "global_step": 167923, "epoch": 1886, "lr": 3.189871453195347e-05}
{"train_loss": 0.05258385092020035, "global_step": 167924, "epoch": 1886, "lr": 3.189817412119625e-05}
{"train_loss": 0.07661914080381393, "global_step": 167925, "epoch": 1886, "lr": 3.189763371287257e-05}
{"train_loss": 0.051886413246393204, "global_step": 167926, "epoch": 1886, "lr": 3.189709330698254e-05}
{"train_loss": 0.03869258612394333, "global_step": 167927, "epoch": 1886, "lr": 3.189655290352625e-05}
{"train_loss": 0.0645611435174942, "global_step": 167928, "epoch": 1886, "lr": 3.1896012502503726e-05}
{"train_loss": 0.0759480893611908, "global_step": 167929, "epoch": 1886, "lr": 3.189547210391508e-05}
{"train_loss": 0.12180089950561523, "global_step": 167930, "epoch": 1886, "lr": 3.1894931707760344e-05}
{"train_loss": 0.09466136991977692, "global_step": 167931, "epoch": 1886, "lr": 3.189439131403964e-05}
{"train_loss": 0.13680477440357208, "global_step": 167932, "epoch": 1886, "lr": 3.189385092275299e-05}
{"train_loss": 0.04550442472100258, "global_step": 167933, "epoch": 1886, "lr": 3.189331053390051e-05}
{"train_loss": 0.11857745796442032, "global_step": 167934, "epoch": 1886, "lr": 3.189277014748225e-05}
{"train_loss": 0.08341433107852936, "global_step": 167935, "epoch": 1886, "lr": 3.1892229763498286e-05}
{"train_loss": 0.14923857152462006, "global_step": 167936, "epoch": 1886, "lr": 3.1891689381948686e-05}
{"train_loss": 0.07721243798732758, "global_step": 167937, "epoch": 1886, "lr": 3.189114900283353e-05}
{"train_loss": 0.05624743551015854, "global_step": 167938, "epoch": 1886, "lr": 3.189060862615288e-05}
{"train_loss": 0.06339152902364731, "global_step": 167939, "epoch": 1886, "lr": 3.189006825190684e-05}
{"train_loss": 0.11928766965866089, "global_step": 167940, "epoch": 1886, "lr": 3.1889527880095436e-05}
{"train_loss": 0.06625392287969589, "global_step": 167941, "epoch": 1886, "lr": 3.1888987510718783e-05}
{"train_loss": 0.0861814118754328, "global_step": 167942, "epoch": 1886, "lr": 3.188844714377691e-05, "val_loss": 7.09507417678833}
{"train_loss": 0.0654774084687233, "global_step": 167943, "epoch": 1887, "lr": 3.188790677926992e-05}
{"train_loss": 0.07782132923603058, "global_step": 167944, "epoch": 1887, "lr": 3.18873664171979e-05}
{"train_loss": 0.07179708033800125, "global_step": 167945, "epoch": 1887, "lr": 3.188682605756088e-05}
{"train_loss": 0.07065945118665695, "global_step": 167946, "epoch": 1887, "lr": 3.188628570035898e-05}
{"train_loss": 0.09589377790689468, "global_step": 167947, "epoch": 1887, "lr": 3.188574534559222e-05}
{"train_loss": 0.11266224086284637, "global_step": 167948, "epoch": 1887, "lr": 3.188520499326073e-05}
{"train_loss": 0.11368177831172943, "global_step": 167949, "epoch": 1887, "lr": 3.188466464336453e-05}
{"train_loss": 0.05800360068678856, "global_step": 167950, "epoch": 1887, "lr": 3.188412429590374e-05}
{"train_loss": 0.11624272167682648, "global_step": 167951, "epoch": 1887, "lr": 3.188358395087838e-05}
{"train_loss": 0.11315087974071503, "global_step": 167952, "epoch": 1887, "lr": 3.188304360828858e-05}
{"train_loss": 0.1341005265712738, "global_step": 167953, "epoch": 1887, "lr": 3.1882503268134365e-05}
{"train_loss": 0.06665167212486267, "global_step": 167954, "epoch": 1887, "lr": 3.188196293041583e-05}
{"train_loss": 0.07315871119499207, "global_step": 167955, "epoch": 1887, "lr": 3.188142259513305e-05}
{"train_loss": 0.14470750093460083, "global_step": 167956, "epoch": 1887, "lr": 3.188088226228609e-05}
{"train_loss": 0.09488102793693542, "global_step": 167957, "epoch": 1887, "lr": 3.188034193187502e-05}
{"train_loss": 0.07244977355003357, "global_step": 167958, "epoch": 1887, "lr": 3.187980160389993e-05}
{"train_loss": 0.06798704713582993, "global_step": 167959, "epoch": 1887, "lr": 3.187926127836086e-05}
{"train_loss": 0.06744232773780823, "global_step": 167960, "epoch": 1887, "lr": 3.1878720955257925e-05}
{"train_loss": 0.0898650512099266, "global_step": 167961, "epoch": 1887, "lr": 3.187818063459115e-05}
{"train_loss": 0.14025403559207916, "global_step": 167962, "epoch": 1887, "lr": 3.187764031636063e-05}
{"train_loss": 0.07364468276500702, "global_step": 167963, "epoch": 1887, "lr": 3.187710000056646e-05}
{"train_loss": 0.12788769602775574, "global_step": 167964, "epoch": 1887, "lr": 3.1876559687208686e-05}
{"train_loss": 0.12649184465408325, "global_step": 167965, "epoch": 1887, "lr": 3.18760193762874e-05}
{"train_loss": 0.07064642757177353, "global_step": 167966, "epoch": 1887, "lr": 3.1875479067802635e-05}
{"train_loss": 0.19026578962802887, "global_step": 167967, "epoch": 1887, "lr": 3.187493876175452e-05}
{"train_loss": 0.09344811737537384, "global_step": 167968, "epoch": 1887, "lr": 3.187439845814308e-05}
{"train_loss": 0.09284026175737381, "global_step": 167969, "epoch": 1887, "lr": 3.187385815696842e-05}
{"train_loss": 0.10431774705648422, "global_step": 167970, "epoch": 1887, "lr": 3.187331785823058e-05}
{"train_loss": 0.13293123245239258, "global_step": 167971, "epoch": 1887, "lr": 3.1872777561929675e-05}
{"train_loss": 0.07443149387836456, "global_step": 167972, "epoch": 1887, "lr": 3.187223726806574e-05}
{"train_loss": 0.11361662298440933, "global_step": 167973, "epoch": 1887, "lr": 3.1871696976638865e-05}
{"train_loss": 0.14926737546920776, "global_step": 167974, "epoch": 1887, "lr": 3.1871156687649116e-05}
{"train_loss": 0.06422004103660583, "global_step": 167975, "epoch": 1887, "lr": 3.187061640109658e-05}
{"train_loss": 0.08073580265045166, "global_step": 167976, "epoch": 1887, "lr": 3.187007611698131e-05}
{"train_loss": 0.06324384361505508, "global_step": 167977, "epoch": 1887, "lr": 3.186953583530338e-05}
{"train_loss": 0.13187511265277863, "global_step": 167978, "epoch": 1887, "lr": 3.1868995556062875e-05}
{"train_loss": 0.07729927450418472, "global_step": 167979, "epoch": 1887, "lr": 3.1868455279259856e-05}
{"train_loss": 0.09079833328723907, "global_step": 167980, "epoch": 1887, "lr": 3.186791500489442e-05}
{"train_loss": 0.16366390883922577, "global_step": 167981, "epoch": 1887, "lr": 3.1867374732966606e-05}
{"train_loss": 0.0919807180762291, "global_step": 167982, "epoch": 1887, "lr": 3.186683446347651e-05}
{"train_loss": 0.09078922122716904, "global_step": 167983, "epoch": 1887, "lr": 3.1866294196424196e-05}
{"train_loss": 0.07760334759950638, "global_step": 167984, "epoch": 1887, "lr": 3.1865753931809745e-05}
{"train_loss": 0.053207337856292725, "global_step": 167985, "epoch": 1887, "lr": 3.18652136696332e-05}
{"train_loss": 0.06968551874160767, "global_step": 167986, "epoch": 1887, "lr": 3.1864673409894686e-05}
{"train_loss": 0.07381345331668854, "global_step": 167987, "epoch": 1887, "lr": 3.186413315259422e-05}
{"train_loss": 0.10031551867723465, "global_step": 167988, "epoch": 1887, "lr": 3.186359289773192e-05}
{"train_loss": 0.1062033399939537, "global_step": 167989, "epoch": 1887, "lr": 3.186305264530782e-05}
{"train_loss": 0.13639196753501892, "global_step": 167990, "epoch": 1887, "lr": 3.186251239532203e-05}
{"train_loss": 0.06984329968690872, "global_step": 167991, "epoch": 1887, "lr": 3.1861972147774585e-05}
{"train_loss": 0.16876114904880524, "global_step": 167992, "epoch": 1887, "lr": 3.18614319026656e-05}
{"train_loss": 0.05748957768082619, "global_step": 167993, "epoch": 1887, "lr": 3.186089165999511e-05}
{"train_loss": 0.07101733982563019, "global_step": 167994, "epoch": 1887, "lr": 3.1860351419763206e-05}
{"train_loss": 0.07243732362985611, "global_step": 167995, "epoch": 1887, "lr": 3.185981118196996e-05}
{"train_loss": 0.10160607844591141, "global_step": 167996, "epoch": 1887, "lr": 3.1859270946615426e-05}
{"train_loss": 0.0757441595196724, "global_step": 167997, "epoch": 1887, "lr": 3.1858730713699705e-05}
{"train_loss": 0.03424486517906189, "global_step": 167998, "epoch": 1887, "lr": 3.185819048322284e-05}
{"train_loss": 0.10260427743196487, "global_step": 167999, "epoch": 1887, "lr": 3.185765025518495e-05}
{"train_loss": 0.08811749517917633, "global_step": 168000, "epoch": 1887, "lr": 3.1857110029586045e-05}
{"train_loss": 0.116281658411026, "global_step": 168001, "epoch": 1887, "lr": 3.185656980642626e-05}
{"train_loss": 0.06782890111207962, "global_step": 168002, "epoch": 1887, "lr": 3.185602958570562e-05}
{"train_loss": 0.09043847769498825, "global_step": 168003, "epoch": 1887, "lr": 3.185548936742423e-05}
{"train_loss": 0.02676592767238617, "global_step": 168004, "epoch": 1887, "lr": 3.185494915158212e-05}
{"train_loss": 0.10439075529575348, "global_step": 168005, "epoch": 1887, "lr": 3.1854408938179425e-05}
{"train_loss": 0.08218032866716385, "global_step": 168006, "epoch": 1887, "lr": 3.1853868727216155e-05}
{"train_loss": 0.07018743455410004, "global_step": 168007, "epoch": 1887, "lr": 3.1853328518692435e-05}
{"train_loss": 0.07803739607334137, "global_step": 168008, "epoch": 1887, "lr": 3.185278831260829e-05}
{"train_loss": 0.07389618456363678, "global_step": 168009, "epoch": 1887, "lr": 3.185224810896385e-05}
{"train_loss": 0.08180839568376541, "global_step": 168010, "epoch": 1887, "lr": 3.185170790775912e-05}
{"train_loss": 0.10353540629148483, "global_step": 168011, "epoch": 1887, "lr": 3.185116770899423e-05}
{"train_loss": 0.08785469830036163, "global_step": 168012, "epoch": 1887, "lr": 3.1850627512669216e-05}
{"train_loss": 0.09445606917142868, "global_step": 168013, "epoch": 1887, "lr": 3.185008731878417e-05}
{"train_loss": 0.10860420763492584, "global_step": 168014, "epoch": 1887, "lr": 3.1849547127339155e-05}
{"train_loss": 0.12191230803728104, "global_step": 168015, "epoch": 1887, "lr": 3.184900693833424e-05}
{"train_loss": 0.10479484498500824, "global_step": 168016, "epoch": 1887, "lr": 3.184846675176952e-05}
{"train_loss": 0.07296398282051086, "global_step": 168017, "epoch": 1887, "lr": 3.184792656764503e-05}
{"train_loss": 0.043379396200180054, "global_step": 168018, "epoch": 1887, "lr": 3.18473863859609e-05}
{"train_loss": 0.046661633998155594, "global_step": 168019, "epoch": 1887, "lr": 3.184684620671714e-05}
{"train_loss": 0.06635230034589767, "global_step": 168020, "epoch": 1887, "lr": 3.1846306029913866e-05}
{"train_loss": 0.07623773068189621, "global_step": 168021, "epoch": 1887, "lr": 3.1845765855551114e-05}
{"train_loss": 0.07821597903966904, "global_step": 168022, "epoch": 1887, "lr": 3.1845225683628995e-05}
{"train_loss": 0.07447908818721771, "global_step": 168023, "epoch": 1887, "lr": 3.184468551414755e-05}
{"train_loss": 0.0677889958024025, "global_step": 168024, "epoch": 1887, "lr": 3.184414534710689e-05}
{"train_loss": 0.09917058795690536, "global_step": 168025, "epoch": 1887, "lr": 3.184360518250704e-05}
{"train_loss": 0.09884472191333771, "global_step": 168026, "epoch": 1887, "lr": 3.1843065020348114e-05}
{"train_loss": 0.06367241591215134, "global_step": 168027, "epoch": 1887, "lr": 3.1842524860630155e-05}
{"train_loss": 0.10214100033044815, "global_step": 168028, "epoch": 1887, "lr": 3.1841984703353256e-05}
{"train_loss": 0.07570742070674896, "global_step": 168029, "epoch": 1887, "lr": 3.1841444548517466e-05}
{"train_loss": 0.04801306501030922, "global_step": 168030, "epoch": 1887, "lr": 3.184090439612289e-05}
{"train_loss": 0.08990466017066763, "global_step": 168031, "epoch": 1887, "lr": 3.1840364246169576e-05, "val_loss": 7.133728981018066}
{"train_loss": 0.05857539549469948, "global_step": 168032, "epoch": 1888, "lr": 3.18398240986576e-05}
{"train_loss": 0.05535915493965149, "global_step": 168033, "epoch": 1888, "lr": 3.183928395358704e-05}
{"train_loss": 0.06123798340559006, "global_step": 168034, "epoch": 1888, "lr": 3.183874381095796e-05}
{"train_loss": 0.09309110790491104, "global_step": 168035, "epoch": 1888, "lr": 3.183820367077045e-05}
{"train_loss": 0.06964602321386337, "global_step": 168036, "epoch": 1888, "lr": 3.1837663533024566e-05}
{"train_loss": 0.0960719957947731, "global_step": 168037, "epoch": 1888, "lr": 3.1837123397720394e-05}
{"train_loss": 0.0958452820777893, "global_step": 168038, "epoch": 1888, "lr": 3.1836583264857986e-05}
{"train_loss": 0.11973923444747925, "global_step": 168039, "epoch": 1888, "lr": 3.1836043134437447e-05}
{"train_loss": 0.05616221949458122, "global_step": 168040, "epoch": 1888, "lr": 3.183550300645881e-05}
{"train_loss": 0.04875505343079567, "global_step": 168041, "epoch": 1888, "lr": 3.183496288092219e-05}
{"train_loss": 0.06641973555088043, "global_step": 168042, "epoch": 1888, "lr": 3.183442275782761e-05}
{"train_loss": 0.1311413198709488, "global_step": 168043, "epoch": 1888, "lr": 3.1833882637175206e-05}
{"train_loss": 0.09434864670038223, "global_step": 168044, "epoch": 1888, "lr": 3.183334251896498e-05}
{"train_loss": 0.06866651773452759, "global_step": 168045, "epoch": 1888, "lr": 3.183280240319707e-05}
{"train_loss": 0.12101418524980545, "global_step": 168046, "epoch": 1888, "lr": 3.183226228987149e-05}
{"train_loss": 0.0915433019399643, "global_step": 168047, "epoch": 1888, "lr": 3.1831722178988364e-05}
{"train_loss": 0.11405231058597565, "global_step": 168048, "epoch": 1888, "lr": 3.1831182070547726e-05}
{"train_loss": 0.04972586780786514, "global_step": 168049, "epoch": 1888, "lr": 3.183064196454968e-05}
{"train_loss": 0.07430518418550491, "global_step": 168050, "epoch": 1888, "lr": 3.1830101860994274e-05}
{"train_loss": 0.10691386461257935, "global_step": 168051, "epoch": 1888, "lr": 3.182956175988159e-05}
{"train_loss": 0.0654500275850296, "global_step": 168052, "epoch": 1888, "lr": 3.18290216612117e-05}
{"train_loss": 0.07644609361886978, "global_step": 168053, "epoch": 1888, "lr": 3.182848156498466e-05}
{"train_loss": 0.11049625277519226, "global_step": 168054, "epoch": 1888, "lr": 3.1827941471200586e-05}
{"train_loss": 0.0655898004770279, "global_step": 168055, "epoch": 1888, "lr": 3.1827401379859503e-05}
{"train_loss": 0.09695176780223846, "global_step": 168056, "epoch": 1888, "lr": 3.182686129096152e-05}
{"train_loss": 0.09282173216342926, "global_step": 168057, "epoch": 1888, "lr": 3.182632120450668e-05}
{"train_loss": 0.09416750818490982, "global_step": 168058, "epoch": 1888, "lr": 3.182578112049509e-05}
{"train_loss": 0.05458352342247963, "global_step": 168059, "epoch": 1888, "lr": 3.182524103892678e-05}
{"train_loss": 0.06343844532966614, "global_step": 168060, "epoch": 1888, "lr": 3.182470095980186e-05}
{"train_loss": 0.09425453096628189, "global_step": 168061, "epoch": 1888, "lr": 3.1824160883120376e-05}
{"train_loss": 0.08023034781217575, "global_step": 168062, "epoch": 1888, "lr": 3.182362080888243e-05}
{"train_loss": 0.10877273231744766, "global_step": 168063, "epoch": 1888, "lr": 3.182308073708805e-05}
{"train_loss": 0.06962434947490692, "global_step": 168064, "epoch": 1888, "lr": 3.182254066773736e-05}
{"train_loss": 0.11732446402311325, "global_step": 168065, "epoch": 1888, "lr": 3.182200060083039e-05}
{"train_loss": 0.06421715021133423, "global_step": 168066, "epoch": 1888, "lr": 3.182146053636724e-05}
{"train_loss": 0.04113398492336273, "global_step": 168067, "epoch": 1888, "lr": 3.1820920474347965e-05}
{"train_loss": 0.046397026628255844, "global_step": 168068, "epoch": 1888, "lr": 3.1820380414772664e-05}
{"train_loss": 0.11346348375082016, "global_step": 168069, "epoch": 1888, "lr": 3.181984035764138e-05}
{"train_loss": 0.04080411046743393, "global_step": 168070, "epoch": 1888, "lr": 3.1819300302954184e-05}
{"train_loss": 0.10158000886440277, "global_step": 168071, "epoch": 1888, "lr": 3.181876025071118e-05}
{"train_loss": 0.12181288748979568, "global_step": 168072, "epoch": 1888, "lr": 3.181822020091241e-05}
{"train_loss": 0.15079481899738312, "global_step": 168073, "epoch": 1888, "lr": 3.181768015355797e-05}
{"train_loss": 0.08539219945669174, "global_step": 168074, "epoch": 1888, "lr": 3.1817140108647907e-05}
{"train_loss": 0.10190615057945251, "global_step": 168075, "epoch": 1888, "lr": 3.1816600066182326e-05}
{"train_loss": 0.05112241581082344, "global_step": 168076, "epoch": 1888, "lr": 3.181606002616126e-05}
{"train_loss": 0.15663796663284302, "global_step": 168077, "epoch": 1888, "lr": 3.181551998858483e-05}
{"train_loss": 0.05893000215291977, "global_step": 168078, "epoch": 1888, "lr": 3.181497995345305e-05}
{"train_loss": 0.07886943221092224, "global_step": 168079, "epoch": 1888, "lr": 3.181443992076605e-05}
{"train_loss": 0.10276374220848083, "global_step": 168080, "epoch": 1888, "lr": 3.181389989052386e-05}
{"train_loss": 0.10408453643321991, "global_step": 168081, "epoch": 1888, "lr": 3.181335986272659e-05}
{"train_loss": 0.05580848455429077, "global_step": 168082, "epoch": 1888, "lr": 3.181281983737427e-05}
{"train_loss": 0.04766487702727318, "global_step": 168083, "epoch": 1888, "lr": 3.1812279814467013e-05}
{"train_loss": 0.13768380880355835, "global_step": 168084, "epoch": 1888, "lr": 3.1811739794004855e-05}
{"train_loss": 0.11448382586240768, "global_step": 168085, "epoch": 1888, "lr": 3.181119977598791e-05}
{"train_loss": 0.026048239320516586, "global_step": 168086, "epoch": 1888, "lr": 3.18106597604162e-05}
{"train_loss": 0.059804946184158325, "global_step": 168087, "epoch": 1888, "lr": 3.1810119747289845e-05}
{"train_loss": 0.05614657700061798, "global_step": 168088, "epoch": 1888, "lr": 3.18095797366089e-05}
{"train_loss": 0.1060003936290741, "global_step": 168089, "epoch": 1888, "lr": 3.1809039728373416e-05}
{"train_loss": 0.10666029900312424, "global_step": 168090, "epoch": 1888, "lr": 3.18084997225835e-05}
{"train_loss": 0.08650866150856018, "global_step": 168091, "epoch": 1888, "lr": 3.18079597192392e-05}
{"train_loss": 0.07252988964319229, "global_step": 168092, "epoch": 1888, "lr": 3.180741971834061e-05}
{"train_loss": 0.12629356980323792, "global_step": 168093, "epoch": 1888, "lr": 3.180687971988777e-05}
{"train_loss": 0.052817150950431824, "global_step": 168094, "epoch": 1888, "lr": 3.18063397238808e-05}
{"train_loss": 0.06810857355594635, "global_step": 168095, "epoch": 1888, "lr": 3.1805799730319725e-05}
{"train_loss": 0.1262809932231903, "global_step": 168096, "epoch": 1888, "lr": 3.180525973920466e-05}
{"train_loss": 0.057591620832681656, "global_step": 168097, "epoch": 1888, "lr": 3.180471975053563e-05}
{"train_loss": 0.09022635221481323, "global_step": 168098, "epoch": 1888, "lr": 3.180417976431276e-05}
{"train_loss": 0.0635291039943695, "global_step": 168099, "epoch": 1888, "lr": 3.1803639780536065e-05}
{"train_loss": 0.09389275312423706, "global_step": 168100, "epoch": 1888, "lr": 3.1803099799205684e-05}
{"train_loss": 0.07514921575784683, "global_step": 168101, "epoch": 1888, "lr": 3.180255982032162e-05}
{"train_loss": 0.08269280940294266, "global_step": 168102, "epoch": 1888, "lr": 3.1802019843884014e-05}
{"train_loss": 0.07883848994970322, "global_step": 168103, "epoch": 1888, "lr": 3.180147986989287e-05}
{"train_loss": 0.05288229137659073, "global_step": 168104, "epoch": 1888, "lr": 3.180093989834833e-05}
{"train_loss": 0.13435141742229462, "global_step": 168105, "epoch": 1888, "lr": 3.18003999292504e-05}
{"train_loss": 0.11518997699022293, "global_step": 168106, "epoch": 1888, "lr": 3.17998599625992e-05}
{"train_loss": 0.10875040292739868, "global_step": 168107, "epoch": 1888, "lr": 3.1799319998394785e-05}
{"train_loss": 0.056141600012779236, "global_step": 168108, "epoch": 1888, "lr": 3.1798780036637234e-05}
{"train_loss": 0.10066071897745132, "global_step": 168109, "epoch": 1888, "lr": 3.179824007732662e-05}
{"train_loss": 0.0752536803483963, "global_step": 168110, "epoch": 1888, "lr": 3.1797700120462995e-05}
{"train_loss": 0.10600907355546951, "global_step": 168111, "epoch": 1888, "lr": 3.1797160166046455e-05}
{"train_loss": 0.07744818180799484, "global_step": 168112, "epoch": 1888, "lr": 3.179662021407706e-05}
{"train_loss": 0.15270911157131195, "global_step": 168113, "epoch": 1888, "lr": 3.179608026455491e-05}
{"train_loss": 0.13803981244564056, "global_step": 168114, "epoch": 1888, "lr": 3.1795540317480024e-05}
{"train_loss": 0.07810540497303009, "global_step": 168115, "epoch": 1888, "lr": 3.179500037285252e-05}
{"train_loss": 0.08824122697114944, "global_step": 168116, "epoch": 1888, "lr": 3.179446043067245e-05}
{"train_loss": 0.10662411153316498, "global_step": 168117, "epoch": 1888, "lr": 3.179392049093991e-05}
{"train_loss": 0.11937075108289719, "global_step": 168118, "epoch": 1888, "lr": 3.179338055365493e-05}
{"train_loss": 0.050840117037296295, "global_step": 168119, "epoch": 1888, "lr": 3.179284061881764e-05}
{"train_loss": 0.08750902781828065, "global_step": 168120, "epoch": 1888, "lr": 3.1792300686428044e-05, "val_loss": 7.088855743408203}
{"train_loss": 0.05195556581020355, "global_step": 168121, "epoch": 1889, "lr": 3.179176075648629e-05}
{"train_loss": 0.04381508752703667, "global_step": 168122, "epoch": 1889, "lr": 3.1791220828992375e-05}
{"train_loss": 0.12141525745391846, "global_step": 168123, "epoch": 1889, "lr": 3.179068090394643e-05}
{"train_loss": 0.17067235708236694, "global_step": 168124, "epoch": 1889, "lr": 3.17901409813485e-05}
{"train_loss": 0.1205667033791542, "global_step": 168125, "epoch": 1889, "lr": 3.178960106119866e-05}
{"train_loss": 0.0999077782034874, "global_step": 168126, "epoch": 1889, "lr": 3.1789061143496994e-05}
{"train_loss": 0.10140427947044373, "global_step": 168127, "epoch": 1889, "lr": 3.1788521228243564e-05}
{"train_loss": 0.11580275744199753, "global_step": 168128, "epoch": 1889, "lr": 3.1787981315438454e-05}
{"train_loss": 0.12625372409820557, "global_step": 168129, "epoch": 1889, "lr": 3.178744140508171e-05}
{"train_loss": 0.1337263137102127, "global_step": 168130, "epoch": 1889, "lr": 3.178690149717344e-05}
{"train_loss": 0.05912800133228302, "global_step": 168131, "epoch": 1889, "lr": 3.178636159171368e-05}
{"train_loss": 0.11764940619468689, "global_step": 168132, "epoch": 1889, "lr": 3.178582168870254e-05}
{"train_loss": 0.08928795903921127, "global_step": 168133, "epoch": 1889, "lr": 3.1785281788140065e-05}
{"train_loss": 0.05390283092856407, "global_step": 168134, "epoch": 1889, "lr": 3.178474189002635e-05}
{"train_loss": 0.09046508371829987, "global_step": 168135, "epoch": 1889, "lr": 3.178420199436143e-05}
{"train_loss": 0.10550347715616226, "global_step": 168136, "epoch": 1889, "lr": 3.1783662101145425e-05}
{"train_loss": 0.08274875581264496, "global_step": 168137, "epoch": 1889, "lr": 3.1783122210378363e-05}
{"train_loss": 0.10988324135541916, "global_step": 168138, "epoch": 1889, "lr": 3.178258232206036e-05}
{"train_loss": 0.07045134156942368, "global_step": 168139, "epoch": 1889, "lr": 3.1782042436191444e-05}
{"train_loss": 0.07152340561151505, "global_step": 168140, "epoch": 1889, "lr": 3.1781502552771736e-05}
{"train_loss": 0.10295239835977554, "global_step": 168141, "epoch": 1889, "lr": 3.178096267180126e-05}
{"train_loss": 0.13709868490695953, "global_step": 168142, "epoch": 1889, "lr": 3.178042279328012e-05}
{"train_loss": 0.10576006770133972, "global_step": 168143, "epoch": 1889, "lr": 3.177988291720838e-05}
{"train_loss": 0.05784785374999046, "global_step": 168144, "epoch": 1889, "lr": 3.177934304358612e-05}
{"train_loss": 0.104505755007267, "global_step": 168145, "epoch": 1889, "lr": 3.17788031724134e-05}
{"train_loss": 0.10257939994335175, "global_step": 168146, "epoch": 1889, "lr": 3.177826330369029e-05}
{"train_loss": 0.11969934403896332, "global_step": 168147, "epoch": 1889, "lr": 3.177772343741689e-05}
{"train_loss": 0.13718055188655853, "global_step": 168148, "epoch": 1889, "lr": 3.177718357359323e-05}
{"train_loss": 0.12063290923833847, "global_step": 168149, "epoch": 1889, "lr": 3.177664371221942e-05}
{"train_loss": 0.09981754422187805, "global_step": 168150, "epoch": 1889, "lr": 3.177610385329551e-05}
{"train_loss": 0.1299521028995514, "global_step": 168151, "epoch": 1889, "lr": 3.17755639968216e-05}
{"train_loss": 0.07804346829652786, "global_step": 168152, "epoch": 1889, "lr": 3.177502414279772e-05}
{"train_loss": 0.08011148869991302, "global_step": 168153, "epoch": 1889, "lr": 3.177448429122398e-05}
{"train_loss": 0.07915101945400238, "global_step": 168154, "epoch": 1889, "lr": 3.177394444210042e-05}
{"train_loss": 0.09339776635169983, "global_step": 168155, "epoch": 1889, "lr": 3.177340459542716e-05}
{"train_loss": 0.10075882077217102, "global_step": 168156, "epoch": 1889, "lr": 3.177286475120421e-05}
{"train_loss": 0.07294391840696335, "global_step": 168157, "epoch": 1889, "lr": 3.17723249094317e-05}
{"train_loss": 0.16825470328330994, "global_step": 168158, "epoch": 1889, "lr": 3.177178507010967e-05}
{"train_loss": 0.10669554769992828, "global_step": 168159, "epoch": 1889, "lr": 3.1771245233238215e-05}
{"train_loss": 0.12140627205371857, "global_step": 168160, "epoch": 1889, "lr": 3.177070539881737e-05}
{"train_loss": 0.07229401171207428, "global_step": 168161, "epoch": 1889, "lr": 3.177016556684724e-05}
{"train_loss": 0.059195321053266525, "global_step": 168162, "epoch": 1889, "lr": 3.176962573732789e-05}
{"train_loss": 0.14889903366565704, "global_step": 168163, "epoch": 1889, "lr": 3.17690859102594e-05}
{"train_loss": 0.07849854975938797, "global_step": 168164, "epoch": 1889, "lr": 3.1768546085641826e-05}
{"train_loss": 0.11901336163282394, "global_step": 168165, "epoch": 1889, "lr": 3.176800626347525e-05}
{"train_loss": 0.09747350215911865, "global_step": 168166, "epoch": 1889, "lr": 3.176746644375976e-05}
{"train_loss": 0.12266519665718079, "global_step": 168167, "epoch": 1889, "lr": 3.1766926626495384e-05}
{"train_loss": 0.07371576875448227, "global_step": 168168, "epoch": 1889, "lr": 3.176638681168225e-05}
{"train_loss": 0.10942965000867844, "global_step": 168169, "epoch": 1889, "lr": 3.1765846999320377e-05}
{"train_loss": 0.07287610322237015, "global_step": 168170, "epoch": 1889, "lr": 3.176530718940989e-05}
{"train_loss": 0.06574532389640808, "global_step": 168171, "epoch": 1889, "lr": 3.17647673819508e-05}
{"train_loss": 0.054937008768320084, "global_step": 168172, "epoch": 1889, "lr": 3.176422757694325e-05}
{"train_loss": 0.04711730778217316, "global_step": 168173, "epoch": 1889, "lr": 3.1763687774387255e-05}
{"train_loss": 0.06426602602005005, "global_step": 168174, "epoch": 1889, "lr": 3.1763147974282927e-05}
{"train_loss": 0.09877319633960724, "global_step": 168175, "epoch": 1889, "lr": 3.176260817663029e-05}
{"train_loss": 0.07688909769058228, "global_step": 168176, "epoch": 1889, "lr": 3.176206838142949e-05}
{"train_loss": 0.07597791403532028, "global_step": 168177, "epoch": 1889, "lr": 3.176152858868052e-05}
{"train_loss": 0.09699207544326782, "global_step": 168178, "epoch": 1889, "lr": 3.1760988798383525e-05}
{"train_loss": 0.07207031548023224, "global_step": 168179, "epoch": 1889, "lr": 3.1760449010538515e-05}
{"train_loss": 0.08311201632022858, "global_step": 168180, "epoch": 1889, "lr": 3.17599092251456e-05}
{"train_loss": 0.09240784496068954, "global_step": 168181, "epoch": 1889, "lr": 3.175936944220485e-05}
{"train_loss": 0.11963067948818207, "global_step": 168182, "epoch": 1889, "lr": 3.175882966171632e-05}
{"train_loss": 0.10809659957885742, "global_step": 168183, "epoch": 1889, "lr": 3.175828988368009e-05}
{"train_loss": 0.10179238021373749, "global_step": 168184, "epoch": 1889, "lr": 3.1757750108096243e-05}
{"train_loss": 0.05687685310840607, "global_step": 168185, "epoch": 1889, "lr": 3.175721033496485e-05}
{"train_loss": 0.09728152304887772, "global_step": 168186, "epoch": 1889, "lr": 3.175667056428597e-05}
{"train_loss": 0.10609295219182968, "global_step": 168187, "epoch": 1889, "lr": 3.175613079605969e-05}
{"train_loss": 0.06560774147510529, "global_step": 168188, "epoch": 1889, "lr": 3.1755591030286056e-05}
{"train_loss": 0.11800853163003922, "global_step": 168189, "epoch": 1889, "lr": 3.175505126696519e-05}
{"train_loss": 0.07809536159038544, "global_step": 168190, "epoch": 1889, "lr": 3.175451150609711e-05}
{"train_loss": 0.07698746025562286, "global_step": 168191, "epoch": 1889, "lr": 3.1753971747681934e-05}
{"train_loss": 0.061298493295907974, "global_step": 168192, "epoch": 1889, "lr": 3.17534319917197e-05}
{"train_loss": 0.05867752805352211, "global_step": 168193, "epoch": 1889, "lr": 3.1752892238210503e-05}
{"train_loss": 0.1133403331041336, "global_step": 168194, "epoch": 1889, "lr": 3.175235248715439e-05}
{"train_loss": 0.05959634110331535, "global_step": 168195, "epoch": 1889, "lr": 3.1751812738551475e-05}
{"train_loss": 0.10559068620204926, "global_step": 168196, "epoch": 1889, "lr": 3.1751272992401784e-05}
{"train_loss": 0.05304091423749924, "global_step": 168197, "epoch": 1889, "lr": 3.175073324870541e-05}
{"train_loss": 0.09779731184244156, "global_step": 168198, "epoch": 1889, "lr": 3.1750193507462454e-05}
{"train_loss": 0.0833100751042366, "global_step": 168199, "epoch": 1889, "lr": 3.174965376867295e-05}
{"train_loss": 0.05554911866784096, "global_step": 168200, "epoch": 1889, "lr": 3.174911403233698e-05}
{"train_loss": 0.09841375797986984, "global_step": 168201, "epoch": 1889, "lr": 3.174857429845461e-05}
{"train_loss": 0.09955769032239914, "global_step": 168202, "epoch": 1889, "lr": 3.174803456702594e-05}
{"train_loss": 0.07137394696474075, "global_step": 168203, "epoch": 1889, "lr": 3.1747494838051014e-05}
{"train_loss": 0.06350845843553543, "global_step": 168204, "epoch": 1889, "lr": 3.174695511152993e-05}
{"train_loss": 0.09335432201623917, "global_step": 168205, "epoch": 1889, "lr": 3.174641538746272e-05}
{"train_loss": 0.08427859097719193, "global_step": 168206, "epoch": 1889, "lr": 3.1745875665849505e-05}
{"train_loss": 0.0781412422657013, "global_step": 168207, "epoch": 1889, "lr": 3.1745335946690315e-05}
{"train_loss": 0.0608813650906086, "global_step": 168208, "epoch": 1889, "lr": 3.1744796229985265e-05}
{"train_loss": 0.09263987454135766, "global_step": 168209, "epoch": 1889, "lr": 3.1744256515734374e-05, "val_loss": 6.8978352546691895}
{"train_loss": 0.05458221584558487, "global_step": 168210, "epoch": 1890, "lr": 3.174371680393777e-05}
{"train_loss": 0.12310294806957245, "global_step": 168211, "epoch": 1890, "lr": 3.174317709459548e-05}
{"train_loss": 0.11052396893501282, "global_step": 168212, "epoch": 1890, "lr": 3.1742637387707625e-05}
{"train_loss": 0.07995613664388657, "global_step": 168213, "epoch": 1890, "lr": 3.1742097683274225e-05}
{"train_loss": 0.0727599710226059, "global_step": 168214, "epoch": 1890, "lr": 3.17415579812954e-05}
{"train_loss": 0.07072695344686508, "global_step": 168215, "epoch": 1890, "lr": 3.174101828177117e-05}
{"train_loss": 0.10772956907749176, "global_step": 168216, "epoch": 1890, "lr": 3.174047858470165e-05}
{"train_loss": 0.107358917593956, "global_step": 168217, "epoch": 1890, "lr": 3.1739938890086915e-05}
{"train_loss": 0.16050104796886444, "global_step": 168218, "epoch": 1890, "lr": 3.1739399197927e-05}
{"train_loss": 0.07988860458135605, "global_step": 168219, "epoch": 1890, "lr": 3.173885950822202e-05}
{"train_loss": 0.13855355978012085, "global_step": 168220, "epoch": 1890, "lr": 3.1738319820972026e-05}
{"train_loss": 0.040600862354040146, "global_step": 168221, "epoch": 1890, "lr": 3.1737780136177085e-05}
{"train_loss": 0.05953327566385269, "global_step": 168222, "epoch": 1890, "lr": 3.173724045383728e-05}
{"train_loss": 0.046679168939590454, "global_step": 168223, "epoch": 1890, "lr": 3.1736700773952676e-05}
{"train_loss": 0.07922696322202682, "global_step": 168224, "epoch": 1890, "lr": 3.173616109652335e-05}
{"train_loss": 0.0924018919467926, "global_step": 168225, "epoch": 1890, "lr": 3.173562142154939e-05}
{"train_loss": 0.18092073500156403, "global_step": 168226, "epoch": 1890, "lr": 3.1735081749030824e-05}
{"train_loss": 0.08290892839431763, "global_step": 168227, "epoch": 1890, "lr": 3.173454207896779e-05}
{"train_loss": 0.09662559628486633, "global_step": 168228, "epoch": 1890, "lr": 3.1734002411360294e-05}
{"train_loss": 0.10230612754821777, "global_step": 168229, "epoch": 1890, "lr": 3.1733462746208454e-05}
{"train_loss": 0.09363608807325363, "global_step": 168230, "epoch": 1890, "lr": 3.173292308351231e-05}
{"train_loss": 0.10338324308395386, "global_step": 168231, "epoch": 1890, "lr": 3.173238342327198e-05}
{"train_loss": 0.13292628526687622, "global_step": 168232, "epoch": 1890, "lr": 3.173184376548748e-05}
{"train_loss": 0.05954733118414879, "global_step": 168233, "epoch": 1890, "lr": 3.173130411015891e-05}
{"train_loss": 0.05673482269048691, "global_step": 168234, "epoch": 1890, "lr": 3.173076445728637e-05}
{"train_loss": 0.08688269555568695, "global_step": 168235, "epoch": 1890, "lr": 3.1730224806869886e-05}
{"train_loss": 0.08638232946395874, "global_step": 168236, "epoch": 1890, "lr": 3.172968515890957e-05}
{"train_loss": 0.14568351209163666, "global_step": 168237, "epoch": 1890, "lr": 3.172914551340545e-05}
{"train_loss": 0.09914073348045349, "global_step": 168238, "epoch": 1890, "lr": 3.1728605870357646e-05}
{"train_loss": 0.10713759064674377, "global_step": 168239, "epoch": 1890, "lr": 3.17280662297662e-05}
{"train_loss": 0.09064920246601105, "global_step": 168240, "epoch": 1890, "lr": 3.172752659163119e-05}
{"train_loss": 0.08663950860500336, "global_step": 168241, "epoch": 1890, "lr": 3.1726986955952695e-05}
{"train_loss": 0.07620498538017273, "global_step": 168242, "epoch": 1890, "lr": 3.172644732273078e-05}
{"train_loss": 0.06979509443044662, "global_step": 168243, "epoch": 1890, "lr": 3.172590769196552e-05}
{"train_loss": 0.04038368538022041, "global_step": 168244, "epoch": 1890, "lr": 3.1725368063657005e-05}
{"train_loss": 0.10119923949241638, "global_step": 168245, "epoch": 1890, "lr": 3.172482843780527e-05}
{"train_loss": 0.17912867665290833, "global_step": 168246, "epoch": 1890, "lr": 3.172428881441043e-05}
{"train_loss": 0.09565439075231552, "global_step": 168247, "epoch": 1890, "lr": 3.172374919347251e-05}
{"train_loss": 0.07694024592638016, "global_step": 168248, "epoch": 1890, "lr": 3.172320957499163e-05}
{"train_loss": 0.15419936180114746, "global_step": 168249, "epoch": 1890, "lr": 3.172266995896783e-05}
{"train_loss": 0.06124041974544525, "global_step": 168250, "epoch": 1890, "lr": 3.172213034540121e-05}
{"train_loss": 0.072914257645607, "global_step": 168251, "epoch": 1890, "lr": 3.172159073429181e-05}
{"train_loss": 0.12440630048513412, "global_step": 168252, "epoch": 1890, "lr": 3.172105112563971e-05}
{"train_loss": 0.09608437120914459, "global_step": 168253, "epoch": 1890, "lr": 3.172051151944502e-05}
{"train_loss": 0.1254090666770935, "global_step": 168254, "epoch": 1890, "lr": 3.1719971915707765e-05}
{"train_loss": 0.07831937819719315, "global_step": 168255, "epoch": 1890, "lr": 3.1719432314428057e-05}
{"train_loss": 0.07059615850448608, "global_step": 168256, "epoch": 1890, "lr": 3.1718892715605916e-05}
{"train_loss": 0.05759935453534126, "global_step": 168257, "epoch": 1890, "lr": 3.1718353119241486e-05}
{"train_loss": 0.15072689950466156, "global_step": 168258, "epoch": 1890, "lr": 3.171781352533478e-05}
{"train_loss": 0.07787206768989563, "global_step": 168259, "epoch": 1890, "lr": 3.171727393388589e-05}
{"train_loss": 0.07003921270370483, "global_step": 168260, "epoch": 1890, "lr": 3.1716734344894896e-05}
{"train_loss": 0.08498801290988922, "global_step": 168261, "epoch": 1890, "lr": 3.1716194758361855e-05}
{"train_loss": 0.1176709532737732, "global_step": 168262, "epoch": 1890, "lr": 3.171565517428685e-05}
{"train_loss": 0.08930522203445435, "global_step": 168263, "epoch": 1890, "lr": 3.171511559266997e-05}
{"train_loss": 0.08270976692438126, "global_step": 168264, "epoch": 1890, "lr": 3.171457601351124e-05}
{"train_loss": 0.07479999959468842, "global_step": 168265, "epoch": 1890, "lr": 3.17140364368108e-05}
{"train_loss": 0.1089005097746849, "global_step": 168266, "epoch": 1890, "lr": 3.1713496862568646e-05}
{"train_loss": 0.11874517053365707, "global_step": 168267, "epoch": 1890, "lr": 3.1712957290784915e-05}
{"train_loss": 0.05252740532159805, "global_step": 168268, "epoch": 1890, "lr": 3.171241772145964e-05}
{"train_loss": 0.11137785017490387, "global_step": 168269, "epoch": 1890, "lr": 3.17118781545929e-05}
{"train_loss": 0.07750815153121948, "global_step": 168270, "epoch": 1890, "lr": 3.17113385901848e-05}
{"train_loss": 0.08915502578020096, "global_step": 168271, "epoch": 1890, "lr": 3.1710799028235374e-05}
{"train_loss": 0.15331928431987762, "global_step": 168272, "epoch": 1890, "lr": 3.171025946874472e-05}
{"train_loss": 0.11023007333278656, "global_step": 168273, "epoch": 1890, "lr": 3.1709719911712875e-05}
{"train_loss": 0.04979154095053673, "global_step": 168274, "epoch": 1890, "lr": 3.170918035713995e-05}
{"train_loss": 0.0914837047457695, "global_step": 168275, "epoch": 1890, "lr": 3.1708640805026e-05}
{"train_loss": 0.08899279683828354, "global_step": 168276, "epoch": 1890, "lr": 3.1708101255371114e-05}
{"train_loss": 0.08178562670946121, "global_step": 168277, "epoch": 1890, "lr": 3.1707561708175334e-05}
{"train_loss": 0.07243851572275162, "global_step": 168278, "epoch": 1890, "lr": 3.170702216343876e-05}
{"train_loss": 0.06605204939842224, "global_step": 168279, "epoch": 1890, "lr": 3.170648262116145e-05}
{"train_loss": 0.09695366024971008, "global_step": 168280, "epoch": 1890, "lr": 3.1705943081343476e-05}
{"train_loss": 0.09962791204452515, "global_step": 168281, "epoch": 1890, "lr": 3.170540354398492e-05}
{"train_loss": 0.08863610029220581, "global_step": 168282, "epoch": 1890, "lr": 3.1704864009085855e-05}
{"train_loss": 0.09210895001888275, "global_step": 168283, "epoch": 1890, "lr": 3.170432447664633e-05}
{"train_loss": 0.05778482183814049, "global_step": 168284, "epoch": 1890, "lr": 3.170378494666646e-05}
{"train_loss": 0.09608832001686096, "global_step": 168285, "epoch": 1890, "lr": 3.1703245419146266e-05}
{"train_loss": 0.07385344803333282, "global_step": 168286, "epoch": 1890, "lr": 3.170270589408587e-05}
{"train_loss": 0.049342766404151917, "global_step": 168287, "epoch": 1890, "lr": 3.17021663714853e-05}
{"train_loss": 0.04989277571439743, "global_step": 168288, "epoch": 1890, "lr": 3.1701626851344656e-05}
{"train_loss": 0.08164998143911362, "global_step": 168289, "epoch": 1890, "lr": 3.170108733366402e-05}
{"train_loss": 0.09223254024982452, "global_step": 168290, "epoch": 1890, "lr": 3.170054781844344e-05}
{"train_loss": 0.09387359023094177, "global_step": 168291, "epoch": 1890, "lr": 3.170000830568301e-05}
{"train_loss": 0.07023690640926361, "global_step": 168292, "epoch": 1890, "lr": 3.169946879538277e-05}
{"train_loss": 0.10739235579967499, "global_step": 168293, "epoch": 1890, "lr": 3.169892928754283e-05}
{"train_loss": 0.1417197436094284, "global_step": 168294, "epoch": 1890, "lr": 3.169838978216324e-05}
{"train_loss": 0.10544312000274658, "global_step": 168295, "epoch": 1890, "lr": 3.169785027924409e-05}
{"train_loss": 0.08284571021795273, "global_step": 168296, "epoch": 1890, "lr": 3.169731077878543e-05}
{"train_loss": 0.12618719041347504, "global_step": 168297, "epoch": 1890, "lr": 3.169677128078734e-05}
{"train_loss": 0.09241867617944653, "global_step": 168298, "epoch": 1890, "lr": 3.1696231785249905e-05, "val_loss": 6.982697486877441, "train_action_mse_error": 4.315543174743652}
{"train_loss": 0.07371724396944046, "global_step": 168299, "epoch": 1891, "lr": 3.169569229217318e-05}
{"train_loss": 0.11897328495979309, "global_step": 168300, "epoch": 1891, "lr": 3.169515280155725e-05}
{"train_loss": 0.02648516185581684, "global_step": 168301, "epoch": 1891, "lr": 3.169461331340219e-05}
{"train_loss": 0.077572762966156, "global_step": 168302, "epoch": 1891, "lr": 3.169407382770805e-05}
{"train_loss": 0.09399688988924026, "global_step": 168303, "epoch": 1891, "lr": 3.169353434447494e-05}
{"train_loss": 0.11034086346626282, "global_step": 168304, "epoch": 1891, "lr": 3.169299486370289e-05}
{"train_loss": 0.054971203207969666, "global_step": 168305, "epoch": 1891, "lr": 3.1692455385392014e-05}
{"train_loss": 0.0483359731733799, "global_step": 168306, "epoch": 1891, "lr": 3.169191590954234e-05}
{"train_loss": 0.06154564768075943, "global_step": 168307, "epoch": 1891, "lr": 3.169137643615397e-05}
{"train_loss": 0.06838306039571762, "global_step": 168308, "epoch": 1891, "lr": 3.1690836965226986e-05}
{"train_loss": 0.057241760194301605, "global_step": 168309, "epoch": 1891, "lr": 3.1690297496761434e-05}
{"train_loss": 0.07590434700250626, "global_step": 168310, "epoch": 1891, "lr": 3.168975803075741e-05}
{"train_loss": 0.10953271389007568, "global_step": 168311, "epoch": 1891, "lr": 3.168921856721496e-05}
{"train_loss": 0.09799450635910034, "global_step": 168312, "epoch": 1891, "lr": 3.168867910613419e-05}
{"train_loss": 0.07590341567993164, "global_step": 168313, "epoch": 1891, "lr": 3.1688139647515135e-05}
{"train_loss": 0.12560884654521942, "global_step": 168314, "epoch": 1891, "lr": 3.16876001913579e-05}
{"train_loss": 0.08121482282876968, "global_step": 168315, "epoch": 1891, "lr": 3.168706073766254e-05}
{"train_loss": 0.06398390233516693, "global_step": 168316, "epoch": 1891, "lr": 3.168652128642913e-05}
{"train_loss": 0.11101838946342468, "global_step": 168317, "epoch": 1891, "lr": 3.1685981837657745e-05}
{"train_loss": 0.09593471139669418, "global_step": 168318, "epoch": 1891, "lr": 3.168544239134846e-05}
{"train_loss": 0.050074152648448944, "global_step": 168319, "epoch": 1891, "lr": 3.168490294750134e-05}
{"train_loss": 0.08075468987226486, "global_step": 168320, "epoch": 1891, "lr": 3.168436350611646e-05}
{"train_loss": 0.08559297770261765, "global_step": 168321, "epoch": 1891, "lr": 3.168382406719389e-05}
{"train_loss": 0.06715460866689682, "global_step": 168322, "epoch": 1891, "lr": 3.168328463073373e-05}
{"train_loss": 0.07796348631381989, "global_step": 168323, "epoch": 1891, "lr": 3.1682745196735996e-05}
{"train_loss": 0.06207224354147911, "global_step": 168324, "epoch": 1891, "lr": 3.16822057652008e-05}
{"train_loss": 0.10084725171327591, "global_step": 168325, "epoch": 1891, "lr": 3.168166633612823e-05}
{"train_loss": 0.07511647790670395, "global_step": 168326, "epoch": 1891, "lr": 3.168112690951831e-05}
{"train_loss": 0.10237004607915878, "global_step": 168327, "epoch": 1891, "lr": 3.1680587485371164e-05}
{"train_loss": 0.07208849489688873, "global_step": 168328, "epoch": 1891, "lr": 3.168004806368682e-05}
{"train_loss": 0.09474054723978043, "global_step": 168329, "epoch": 1891, "lr": 3.1679508644465386e-05}
{"train_loss": 0.07301800698041916, "global_step": 168330, "epoch": 1891, "lr": 3.1678969227706906e-05}
{"train_loss": 0.08016040921211243, "global_step": 168331, "epoch": 1891, "lr": 3.167842981341148e-05}
{"train_loss": 0.06917960941791534, "global_step": 168332, "epoch": 1891, "lr": 3.167789040157914e-05}
{"train_loss": 0.07381287217140198, "global_step": 168333, "epoch": 1891, "lr": 3.167735099221002e-05}
{"train_loss": 0.13382048904895782, "global_step": 168334, "epoch": 1891, "lr": 3.167681158530412e-05}
{"train_loss": 0.0847180187702179, "global_step": 168335, "epoch": 1891, "lr": 3.167627218086159e-05}
{"train_loss": 0.08126033842563629, "global_step": 168336, "epoch": 1891, "lr": 3.167573277888243e-05}
{"train_loss": 0.10667271912097931, "global_step": 168337, "epoch": 1891, "lr": 3.167519337936676e-05}
{"train_loss": 0.09274326264858246, "global_step": 168338, "epoch": 1891, "lr": 3.167465398231463e-05}
{"train_loss": 0.08292020112276077, "global_step": 168339, "epoch": 1891, "lr": 3.167411458772612e-05}
{"train_loss": 0.053491391241550446, "global_step": 168340, "epoch": 1891, "lr": 3.16735751956013e-05}
{"train_loss": 0.058985400944948196, "global_step": 168341, "epoch": 1891, "lr": 3.167303580594026e-05}
{"train_loss": 0.12483453005552292, "global_step": 168342, "epoch": 1891, "lr": 3.167249641874303e-05}
{"train_loss": 0.0808277353644371, "global_step": 168343, "epoch": 1891, "lr": 3.1671957034009715e-05}
{"train_loss": 0.10469158738851547, "global_step": 168344, "epoch": 1891, "lr": 3.16714176517404e-05}
{"train_loss": 0.1959262192249298, "global_step": 168345, "epoch": 1891, "lr": 3.167087827193512e-05}
{"train_loss": 0.17101210355758667, "global_step": 168346, "epoch": 1891, "lr": 3.167033889459398e-05}
{"train_loss": 0.03764224052429199, "global_step": 168347, "epoch": 1891, "lr": 3.1669799519717024e-05}
{"train_loss": 0.07121599465608597, "global_step": 168348, "epoch": 1891, "lr": 3.166926014730437e-05}
{"train_loss": 0.0859924703836441, "global_step": 168349, "epoch": 1891, "lr": 3.166872077735603e-05}
{"train_loss": 0.08181764930486679, "global_step": 168350, "epoch": 1891, "lr": 3.1668181409872126e-05}
{"train_loss": 0.03561271354556084, "global_step": 168351, "epoch": 1891, "lr": 3.1667642044852695e-05}
{"train_loss": 0.07054560631513596, "global_step": 168352, "epoch": 1891, "lr": 3.166710268229785e-05}
{"train_loss": 0.08375980705022812, "global_step": 168353, "epoch": 1891, "lr": 3.166656332220762e-05}
{"train_loss": 0.09932816028594971, "global_step": 168354, "epoch": 1891, "lr": 3.166602396458211e-05}
{"train_loss": 0.07804270088672638, "global_step": 168355, "epoch": 1891, "lr": 3.166548460942137e-05}
{"train_loss": 0.07473593950271606, "global_step": 168356, "epoch": 1891, "lr": 3.166494525672549e-05}
{"train_loss": 0.11443588882684708, "global_step": 168357, "epoch": 1891, "lr": 3.166440590649452e-05}
{"train_loss": 0.07854592055082321, "global_step": 168358, "epoch": 1891, "lr": 3.166386655872856e-05}
{"train_loss": 0.10953816771507263, "global_step": 168359, "epoch": 1891, "lr": 3.1663327213427664e-05}
{"train_loss": 0.06813102215528488, "global_step": 168360, "epoch": 1891, "lr": 3.1662787870591904e-05}
{"train_loss": 0.09534097462892532, "global_step": 168361, "epoch": 1891, "lr": 3.1662248530221376e-05}
{"train_loss": 0.10751692950725555, "global_step": 168362, "epoch": 1891, "lr": 3.166170919231612e-05}
{"train_loss": 0.07212843000888824, "global_step": 168363, "epoch": 1891, "lr": 3.166116985687624e-05}
{"train_loss": 0.1505943238735199, "global_step": 168364, "epoch": 1891, "lr": 3.1660630523901765e-05}
{"train_loss": 0.08397863060235977, "global_step": 168365, "epoch": 1891, "lr": 3.166009119339283e-05}
{"train_loss": 0.06374865770339966, "global_step": 168366, "epoch": 1891, "lr": 3.165955186534944e-05}
{"train_loss": 0.11801181733608246, "global_step": 168367, "epoch": 1891, "lr": 3.165901253977173e-05}
{"train_loss": 0.09978469461202621, "global_step": 168368, "epoch": 1891, "lr": 3.165847321665971e-05}
{"train_loss": 0.0912107303738594, "global_step": 168369, "epoch": 1891, "lr": 3.165793389601351e-05}
{"train_loss": 0.10400126874446869, "global_step": 168370, "epoch": 1891, "lr": 3.1657394577833154e-05}
{"train_loss": 0.07362187653779984, "global_step": 168371, "epoch": 1891, "lr": 3.1656855262118766e-05}
{"train_loss": 0.07663898169994354, "global_step": 168372, "epoch": 1891, "lr": 3.165631594887036e-05}
{"train_loss": 0.07492545247077942, "global_step": 168373, "epoch": 1891, "lr": 3.165577663808807e-05}
{"train_loss": 0.13135966658592224, "global_step": 168374, "epoch": 1891, "lr": 3.1655237329771925e-05}
{"train_loss": 0.08122099190950394, "global_step": 168375, "epoch": 1891, "lr": 3.1654698023922e-05}
{"train_loss": 0.058992695063352585, "global_step": 168376, "epoch": 1891, "lr": 3.1654158720538386e-05}
{"train_loss": 0.07042808085680008, "global_step": 168377, "epoch": 1891, "lr": 3.165361941962114e-05}
{"train_loss": 0.14643117785453796, "global_step": 168378, "epoch": 1891, "lr": 3.1653080121170356e-05}
{"train_loss": 0.08676083385944366, "global_step": 168379, "epoch": 1891, "lr": 3.165254082518607e-05}
{"train_loss": 0.0921679139137268, "global_step": 168380, "epoch": 1891, "lr": 3.1652001531668395e-05}
{"train_loss": 0.05267352610826492, "global_step": 168381, "epoch": 1891, "lr": 3.165146224061737e-05}
{"train_loss": 0.0708073079586029, "global_step": 168382, "epoch": 1891, "lr": 3.1650922952033095e-05}
{"train_loss": 0.09170561283826828, "global_step": 168383, "epoch": 1891, "lr": 3.165038366591562e-05}
{"train_loss": 0.11045590788125992, "global_step": 168384, "epoch": 1891, "lr": 3.164984438226504e-05}
{"train_loss": 0.09353385120630264, "global_step": 168385, "epoch": 1891, "lr": 3.16493051010814e-05}
{"train_loss": 0.06462123990058899, "global_step": 168386, "epoch": 1891, "lr": 3.16487658223648e-05}
{"train_loss": 0.08619870038263583, "global_step": 168387, "epoch": 1891, "lr": 3.164822654611528e-05, "val_loss": 7.101418495178223}
{"train_loss": 0.12094567716121674, "global_step": 168388, "epoch": 1892, "lr": 3.164768727233296e-05}
{"train_loss": 0.16327357292175293, "global_step": 168389, "epoch": 1892, "lr": 3.164714800101786e-05}
{"train_loss": 0.07791829109191895, "global_step": 168390, "epoch": 1892, "lr": 3.1646608732170094e-05}
{"train_loss": 0.06755417585372925, "global_step": 168391, "epoch": 1892, "lr": 3.164606946578971e-05}
{"train_loss": 0.05783404782414436, "global_step": 168392, "epoch": 1892, "lr": 3.1645530201876794e-05}
{"train_loss": 0.09159006178379059, "global_step": 168393, "epoch": 1892, "lr": 3.1644990940431407e-05}
{"train_loss": 0.10431468486785889, "global_step": 168394, "epoch": 1892, "lr": 3.1644451681453637e-05}
{"train_loss": 0.16926005482673645, "global_step": 168395, "epoch": 1892, "lr": 3.164391242494353e-05}
{"train_loss": 0.08157181739807129, "global_step": 168396, "epoch": 1892, "lr": 3.164337317090118e-05}
{"train_loss": 0.07979018241167068, "global_step": 168397, "epoch": 1892, "lr": 3.1642833919326665e-05}
{"train_loss": 0.0693591982126236, "global_step": 168398, "epoch": 1892, "lr": 3.164229467022003e-05}
{"train_loss": 0.07374342530965805, "global_step": 168399, "epoch": 1892, "lr": 3.1641755423581386e-05}
{"train_loss": 0.11427991837263107, "global_step": 168400, "epoch": 1892, "lr": 3.164121617941077e-05}
{"train_loss": 0.12429042160511017, "global_step": 168401, "epoch": 1892, "lr": 3.164067693770828e-05}
{"train_loss": 0.09232360124588013, "global_step": 168402, "epoch": 1892, "lr": 3.1640137698473956e-05}
{"train_loss": 0.15570464730262756, "global_step": 168403, "epoch": 1892, "lr": 3.163959846170792e-05}
{"train_loss": 0.07360309362411499, "global_step": 168404, "epoch": 1892, "lr": 3.163905922741019e-05}
{"train_loss": 0.06444205343723297, "global_step": 168405, "epoch": 1892, "lr": 3.163851999558089e-05}
{"train_loss": 0.0791393369436264, "global_step": 168406, "epoch": 1892, "lr": 3.163798076622004e-05}
{"train_loss": 0.10455217212438583, "global_step": 168407, "epoch": 1892, "lr": 3.163744153932776e-05}
{"train_loss": 0.07779048383235931, "global_step": 168408, "epoch": 1892, "lr": 3.163690231490409e-05}
{"train_loss": 0.04839666187763214, "global_step": 168409, "epoch": 1892, "lr": 3.1636363092949126e-05}
{"train_loss": 0.15484869480133057, "global_step": 168410, "epoch": 1892, "lr": 3.163582387346291e-05}
{"train_loss": 0.04544735327363014, "global_step": 168411, "epoch": 1892, "lr": 3.163528465644556e-05}
{"train_loss": 0.05771533399820328, "global_step": 168412, "epoch": 1892, "lr": 3.163474544189711e-05}
{"train_loss": 0.08117327094078064, "global_step": 168413, "epoch": 1892, "lr": 3.163420622981764e-05}
{"train_loss": 0.08647595345973969, "global_step": 168414, "epoch": 1892, "lr": 3.1633667020207236e-05}
{"train_loss": 0.08077502250671387, "global_step": 168415, "epoch": 1892, "lr": 3.1633127813065944e-05}
{"train_loss": 0.062125179916620255, "global_step": 168416, "epoch": 1892, "lr": 3.1632588608393876e-05}
{"train_loss": 0.14689555764198303, "global_step": 168417, "epoch": 1892, "lr": 3.1632049406191064e-05}
{"train_loss": 0.1396518051624298, "global_step": 168418, "epoch": 1892, "lr": 3.163151020645762e-05}
{"train_loss": 0.046645816415548325, "global_step": 168419, "epoch": 1892, "lr": 3.163097100919358e-05}
{"train_loss": 0.0702696293592453, "global_step": 168420, "epoch": 1892, "lr": 3.1630431814399044e-05}
{"train_loss": 0.1205458715558052, "global_step": 168421, "epoch": 1892, "lr": 3.162989262207405e-05}
{"train_loss": 0.146465003490448, "global_step": 168422, "epoch": 1892, "lr": 3.162935343221872e-05}
{"train_loss": 0.11415514349937439, "global_step": 168423, "epoch": 1892, "lr": 3.1628814244833074e-05}
{"train_loss": 0.10605823248624802, "global_step": 168424, "epoch": 1892, "lr": 3.1628275059917235e-05}
{"train_loss": 0.07830064743757248, "global_step": 168425, "epoch": 1892, "lr": 3.162773587747122e-05}
{"train_loss": 0.06918378919363022, "global_step": 168426, "epoch": 1892, "lr": 3.1627196697495166e-05}
{"train_loss": 0.05989738926291466, "global_step": 168427, "epoch": 1892, "lr": 3.162665751998909e-05}
{"train_loss": 0.08983832597732544, "global_step": 168428, "epoch": 1892, "lr": 3.16261183449531e-05}
{"train_loss": 0.055126454681158066, "global_step": 168429, "epoch": 1892, "lr": 3.162557917238723e-05}
{"train_loss": 0.09650254994630814, "global_step": 168430, "epoch": 1892, "lr": 3.162504000229161e-05}
{"train_loss": 0.0907679945230484, "global_step": 168431, "epoch": 1892, "lr": 3.162450083466626e-05}
{"train_loss": 0.06300909072160721, "global_step": 168432, "epoch": 1892, "lr": 3.1623961669511275e-05}
{"train_loss": 0.08606668561697006, "global_step": 168433, "epoch": 1892, "lr": 3.162342250682673e-05}
{"train_loss": 0.09779078513383865, "global_step": 168434, "epoch": 1892, "lr": 3.162288334661268e-05}
{"train_loss": 0.11443093419075012, "global_step": 168435, "epoch": 1892, "lr": 3.1622344188869215e-05}
{"train_loss": 0.06616900116205215, "global_step": 168436, "epoch": 1892, "lr": 3.16218050335964e-05}
{"train_loss": 0.11888626217842102, "global_step": 168437, "epoch": 1892, "lr": 3.162126588079432e-05}
{"train_loss": 0.05750026926398277, "global_step": 168438, "epoch": 1892, "lr": 3.162072673046301e-05}
{"train_loss": 0.07795194536447525, "global_step": 168439, "epoch": 1892, "lr": 3.16201875826026e-05}
{"train_loss": 0.04958627372980118, "global_step": 168440, "epoch": 1892, "lr": 3.1619648437213114e-05}
{"train_loss": 0.08160430938005447, "global_step": 168441, "epoch": 1892, "lr": 3.161910929429465e-05}
{"train_loss": 0.09370310604572296, "global_step": 168442, "epoch": 1892, "lr": 3.161857015384726e-05}
{"train_loss": 0.09655570983886719, "global_step": 168443, "epoch": 1892, "lr": 3.161803101587105e-05}
{"train_loss": 0.08970525115728378, "global_step": 168444, "epoch": 1892, "lr": 3.1617491880366054e-05}
{"train_loss": 0.13688160479068756, "global_step": 168445, "epoch": 1892, "lr": 3.161695274733238e-05}
{"train_loss": 0.07767276465892792, "global_step": 168446, "epoch": 1892, "lr": 3.1616413616770056e-05}
{"train_loss": 0.08706093579530716, "global_step": 168447, "epoch": 1892, "lr": 3.161587448867921e-05}
{"train_loss": 0.09261760860681534, "global_step": 168448, "epoch": 1892, "lr": 3.161533536305986e-05}
{"train_loss": 0.11895374208688736, "global_step": 168449, "epoch": 1892, "lr": 3.161479623991213e-05}
{"train_loss": 0.10771144926548004, "global_step": 168450, "epoch": 1892, "lr": 3.1614257119236044e-05}
{"train_loss": 0.042160965502262115, "global_step": 168451, "epoch": 1892, "lr": 3.161371800103171e-05}
{"train_loss": 0.08174650371074677, "global_step": 168452, "epoch": 1892, "lr": 3.161317888529919e-05}
{"train_loss": 0.09999200701713562, "global_step": 168453, "epoch": 1892, "lr": 3.161263977203854e-05}
{"train_loss": 0.10247471183538437, "global_step": 168454, "epoch": 1892, "lr": 3.1612100661249865e-05}
{"train_loss": 0.09277818351984024, "global_step": 168455, "epoch": 1892, "lr": 3.1611561552933195e-05}
{"train_loss": 0.11885219067335129, "global_step": 168456, "epoch": 1892, "lr": 3.161102244708865e-05}
{"train_loss": 0.06840318441390991, "global_step": 168457, "epoch": 1892, "lr": 3.161048334371626e-05}
{"train_loss": 0.09125126153230667, "global_step": 168458, "epoch": 1892, "lr": 3.1609944242816134e-05}
{"train_loss": 0.09404301643371582, "global_step": 168459, "epoch": 1892, "lr": 3.16094051443883e-05}
{"train_loss": 0.04488439857959747, "global_step": 168460, "epoch": 1892, "lr": 3.160886604843289e-05}
{"train_loss": 0.06772267818450928, "global_step": 168461, "epoch": 1892, "lr": 3.1608326954949917e-05}
{"train_loss": 0.09347421675920486, "global_step": 168462, "epoch": 1892, "lr": 3.160778786393951e-05}
{"train_loss": 0.07021720707416534, "global_step": 168463, "epoch": 1892, "lr": 3.160724877540168e-05}
{"train_loss": 0.09927123785018921, "global_step": 168464, "epoch": 1892, "lr": 3.1606709689336556e-05}
{"train_loss": 0.08815421164035797, "global_step": 168465, "epoch": 1892, "lr": 3.160617060574417e-05}
{"train_loss": 0.09112423658370972, "global_step": 168466, "epoch": 1892, "lr": 3.160563152462462e-05}
{"train_loss": 0.07808520644903183, "global_step": 168467, "epoch": 1892, "lr": 3.160509244597795e-05}
{"train_loss": 0.07146112620830536, "global_step": 168468, "epoch": 1892, "lr": 3.160455336980428e-05}
{"train_loss": 0.10988476127386093, "global_step": 168469, "epoch": 1892, "lr": 3.1604014296103625e-05}
{"train_loss": 0.12755745649337769, "global_step": 168470, "epoch": 1892, "lr": 3.16034752248761e-05}
{"train_loss": 0.10030236840248108, "global_step": 168471, "epoch": 1892, "lr": 3.160293615612178e-05}
{"train_loss": 0.07808069884777069, "global_step": 168472, "epoch": 1892, "lr": 3.1602397089840684e-05}
{"train_loss": 0.026368621736764908, "global_step": 168473, "epoch": 1892, "lr": 3.160185802603295e-05}
{"train_loss": 0.1095639243721962, "global_step": 168474, "epoch": 1892, "lr": 3.160131896469861e-05}
{"train_loss": 0.13895641267299652, "global_step": 168475, "epoch": 1892, "lr": 3.160077990583776e-05}
{"train_loss": 0.09111915160430951, "global_step": 168476, "epoch": 1892, "lr": 3.1600240849450446e-05, "val_loss": 7.0688982009887695}
{"train_loss": 0.07068952918052673, "global_step": 168477, "epoch": 1893, "lr": 3.159970179553677e-05}
{"train_loss": 0.09981817752122879, "global_step": 168478, "epoch": 1893, "lr": 3.159916274409677e-05}
{"train_loss": 0.03896322846412659, "global_step": 168479, "epoch": 1893, "lr": 3.159862369513056e-05}
{"train_loss": 0.05529811978340149, "global_step": 168480, "epoch": 1893, "lr": 3.1598084648638174e-05}
{"train_loss": 0.04595198482275009, "global_step": 168481, "epoch": 1893, "lr": 3.1597545604619714e-05}
{"train_loss": 0.10694000869989395, "global_step": 168482, "epoch": 1893, "lr": 3.159700656307523e-05}
{"train_loss": 0.06841312348842621, "global_step": 168483, "epoch": 1893, "lr": 3.1596467524004815e-05}
{"train_loss": 0.06363489478826523, "global_step": 168484, "epoch": 1893, "lr": 3.159592848740851e-05}
{"train_loss": 0.06401831656694412, "global_step": 168485, "epoch": 1893, "lr": 3.159538945328643e-05}
{"train_loss": 0.10728287696838379, "global_step": 168486, "epoch": 1893, "lr": 3.159485042163861e-05}
{"train_loss": 0.08040818572044373, "global_step": 168487, "epoch": 1893, "lr": 3.159431139246515e-05}
{"train_loss": 0.06846322119235992, "global_step": 168488, "epoch": 1893, "lr": 3.1593772365766105e-05}
{"train_loss": 0.059617266058921814, "global_step": 168489, "epoch": 1893, "lr": 3.159323334154155e-05}
{"train_loss": 0.07622212916612625, "global_step": 168490, "epoch": 1893, "lr": 3.1592694319791574e-05}
{"train_loss": 0.05593464896082878, "global_step": 168491, "epoch": 1893, "lr": 3.159215530051622e-05}
{"train_loss": 0.08438674360513687, "global_step": 168492, "epoch": 1893, "lr": 3.159161628371559e-05}
{"train_loss": 0.059405308216810226, "global_step": 168493, "epoch": 1893, "lr": 3.159107726938972e-05}
{"train_loss": 0.0598989762365818, "global_step": 168494, "epoch": 1893, "lr": 3.159053825753874e-05}
{"train_loss": 0.07604847103357315, "global_step": 168495, "epoch": 1893, "lr": 3.1589999248162654e-05}
{"train_loss": 0.04019872844219208, "global_step": 168496, "epoch": 1893, "lr": 3.158946024126159e-05}
{"train_loss": 0.07695368677377701, "global_step": 168497, "epoch": 1893, "lr": 3.1588921236835585e-05}
{"train_loss": 0.06387685239315033, "global_step": 168498, "epoch": 1893, "lr": 3.158838223488474e-05}
{"train_loss": 0.050990987569093704, "global_step": 168499, "epoch": 1893, "lr": 3.15878432354091e-05}
{"train_loss": 0.05127251520752907, "global_step": 168500, "epoch": 1893, "lr": 3.158730423840876e-05}
{"train_loss": 0.16049572825431824, "global_step": 168501, "epoch": 1893, "lr": 3.158676524388376e-05}
{"train_loss": 0.12993170320987701, "global_step": 168502, "epoch": 1893, "lr": 3.158622625183423e-05}
{"train_loss": 0.06575258076190948, "global_step": 168503, "epoch": 1893, "lr": 3.158568726226018e-05}
{"train_loss": 0.08654134720563889, "global_step": 168504, "epoch": 1893, "lr": 3.1585148275161724e-05}
{"train_loss": 0.0807209387421608, "global_step": 168505, "epoch": 1893, "lr": 3.158460929053891e-05}
{"train_loss": 0.08301158249378204, "global_step": 168506, "epoch": 1893, "lr": 3.158407030839182e-05}
{"train_loss": 0.10988827049732208, "global_step": 168507, "epoch": 1893, "lr": 3.158353132872054e-05}
{"train_loss": 0.041206248104572296, "global_step": 168508, "epoch": 1893, "lr": 3.158299235152512e-05}
{"train_loss": 0.08203344792127609, "global_step": 168509, "epoch": 1893, "lr": 3.158245337680565e-05}
{"train_loss": 0.12425132840871811, "global_step": 168510, "epoch": 1893, "lr": 3.158191440456218e-05}
{"train_loss": 0.11625613272190094, "global_step": 168511, "epoch": 1893, "lr": 3.1581375434794826e-05}
{"train_loss": 0.13707399368286133, "global_step": 168512, "epoch": 1893, "lr": 3.15808364675036e-05}
{"train_loss": 0.13691553473472595, "global_step": 168513, "epoch": 1893, "lr": 3.158029750268863e-05}
{"train_loss": 0.07193663716316223, "global_step": 168514, "epoch": 1893, "lr": 3.157975854034994e-05}
{"train_loss": 0.05925259366631508, "global_step": 168515, "epoch": 1893, "lr": 3.157921958048765e-05}
{"train_loss": 0.0767270028591156, "global_step": 168516, "epoch": 1893, "lr": 3.157868062310179e-05}
{"train_loss": 0.06017289310693741, "global_step": 168517, "epoch": 1893, "lr": 3.157814166819247e-05}
{"train_loss": 0.08425837755203247, "global_step": 168518, "epoch": 1893, "lr": 3.157760271575973e-05}
{"train_loss": 0.13303667306900024, "global_step": 168519, "epoch": 1893, "lr": 3.157706376580367e-05}
{"train_loss": 0.05538232997059822, "global_step": 168520, "epoch": 1893, "lr": 3.1576524818324335e-05}
{"train_loss": 0.0793038085103035, "global_step": 168521, "epoch": 1893, "lr": 3.1575985873321824e-05}
{"train_loss": 0.0996677502989769, "global_step": 168522, "epoch": 1893, "lr": 3.157544693079618e-05}
{"train_loss": 0.061234522610902786, "global_step": 168523, "epoch": 1893, "lr": 3.1574907990747503e-05}
{"train_loss": 0.12004142999649048, "global_step": 168524, "epoch": 1893, "lr": 3.1574369053175863e-05}
{"train_loss": 0.105641670525074, "global_step": 168525, "epoch": 1893, "lr": 3.1573830118081314e-05}
{"train_loss": 0.05616457387804985, "global_step": 168526, "epoch": 1893, "lr": 3.157329118546395e-05}
{"train_loss": 0.1471255123615265, "global_step": 168527, "epoch": 1893, "lr": 3.1572752255323824e-05}
{"train_loss": 0.07280421257019043, "global_step": 168528, "epoch": 1893, "lr": 3.1572213327661026e-05}
{"train_loss": 0.10541107505559921, "global_step": 168529, "epoch": 1893, "lr": 3.15716744024756e-05}
{"train_loss": 0.11073561012744904, "global_step": 168530, "epoch": 1893, "lr": 3.157113547976766e-05}
{"train_loss": 0.07816600054502487, "global_step": 168531, "epoch": 1893, "lr": 3.1570596559537234e-05}
{"train_loss": 0.06564639508724213, "global_step": 168532, "epoch": 1893, "lr": 3.157005764178444e-05}
{"train_loss": 0.07930805534124374, "global_step": 168533, "epoch": 1893, "lr": 3.1569518726509306e-05}
{"train_loss": 0.03303009271621704, "global_step": 168534, "epoch": 1893, "lr": 3.156897981371194e-05}
{"train_loss": 0.08984467387199402, "global_step": 168535, "epoch": 1893, "lr": 3.156844090339239e-05}
{"train_loss": 0.09966550022363663, "global_step": 168536, "epoch": 1893, "lr": 3.156790199555075e-05}
{"train_loss": 0.06647171080112457, "global_step": 168537, "epoch": 1893, "lr": 3.1567363090187065e-05}
{"train_loss": 0.09943459928035736, "global_step": 168538, "epoch": 1893, "lr": 3.1566824187301444e-05}
{"train_loss": 0.09983466565608978, "global_step": 168539, "epoch": 1893, "lr": 3.156628528689392e-05}
{"train_loss": 0.08547144383192062, "global_step": 168540, "epoch": 1893, "lr": 3.15657463889646e-05}
{"train_loss": 0.07499929517507553, "global_step": 168541, "epoch": 1893, "lr": 3.1565207493513514e-05}
{"train_loss": 0.05795763432979584, "global_step": 168542, "epoch": 1893, "lr": 3.156466860054078e-05}
{"train_loss": 0.04883502423763275, "global_step": 168543, "epoch": 1893, "lr": 3.156412971004645e-05}
{"train_loss": 0.07815834879875183, "global_step": 168544, "epoch": 1893, "lr": 3.156359082203061e-05}
{"train_loss": 0.053624898195266724, "global_step": 168545, "epoch": 1893, "lr": 3.15630519364933e-05}
{"train_loss": 0.09499195963144302, "global_step": 168546, "epoch": 1893, "lr": 3.156251305343463e-05}
{"train_loss": 0.05791519582271576, "global_step": 168547, "epoch": 1893, "lr": 3.156197417285466e-05}
{"train_loss": 0.050070300698280334, "global_step": 168548, "epoch": 1893, "lr": 3.156143529475343e-05}
{"train_loss": 0.07602600008249283, "global_step": 168549, "epoch": 1893, "lr": 3.156089641913107e-05}
{"train_loss": 0.0993901863694191, "global_step": 168550, "epoch": 1893, "lr": 3.15603575459876e-05}
{"train_loss": 0.09138835966587067, "global_step": 168551, "epoch": 1893, "lr": 3.1559818675323124e-05}
{"train_loss": 0.1161949634552002, "global_step": 168552, "epoch": 1893, "lr": 3.15592798071377e-05}
{"train_loss": 0.0965227261185646, "global_step": 168553, "epoch": 1893, "lr": 3.1558740941431424e-05}
{"train_loss": 0.09494171291589737, "global_step": 168554, "epoch": 1893, "lr": 3.155820207820433e-05}
{"train_loss": 0.08235722035169601, "global_step": 168555, "epoch": 1893, "lr": 3.155766321745652e-05}
{"train_loss": 0.157846599817276, "global_step": 168556, "epoch": 1893, "lr": 3.155712435918805e-05}
{"train_loss": 0.10900718718767166, "global_step": 168557, "epoch": 1893, "lr": 3.1556585503399014e-05}
{"train_loss": 0.10484720766544342, "global_step": 168558, "epoch": 1893, "lr": 3.155604665008945e-05}
{"train_loss": 0.11936907470226288, "global_step": 168559, "epoch": 1893, "lr": 3.155550779925948e-05}
{"train_loss": 0.06155776605010033, "global_step": 168560, "epoch": 1893, "lr": 3.155496895090912e-05}
{"train_loss": 0.06806152313947678, "global_step": 168561, "epoch": 1893, "lr": 3.1554430105038475e-05}
{"train_loss": 0.08027543127536774, "global_step": 168562, "epoch": 1893, "lr": 3.155389126164763e-05}
{"train_loss": 0.07732560485601425, "global_step": 168563, "epoch": 1893, "lr": 3.1553352420736624e-05}
{"train_loss": 0.06170126795768738, "global_step": 168564, "epoch": 1893, "lr": 3.155281358230555e-05}
{"train_loss": 0.082639674792129, "global_step": 168565, "epoch": 1893, "lr": 3.155227474635447e-05, "val_loss": 6.989834785461426}
{"train_loss": 0.05979164317250252, "global_step": 168566, "epoch": 1894, "lr": 3.1551735912883464e-05}
{"train_loss": 0.07480517029762268, "global_step": 168567, "epoch": 1894, "lr": 3.1551197081892604e-05}
{"train_loss": 0.10026282072067261, "global_step": 168568, "epoch": 1894, "lr": 3.155065825338197e-05}
{"train_loss": 0.15354570746421814, "global_step": 168569, "epoch": 1894, "lr": 3.15501194273516e-05}
{"train_loss": 0.14629262685775757, "global_step": 168570, "epoch": 1894, "lr": 3.154958060380162e-05}
{"train_loss": 0.1027052029967308, "global_step": 168571, "epoch": 1894, "lr": 3.1549041782732045e-05}
{"train_loss": 0.05059979110956192, "global_step": 168572, "epoch": 1894, "lr": 3.1548502964143014e-05}
{"train_loss": 0.07113675028085709, "global_step": 168573, "epoch": 1894, "lr": 3.1547964148034524e-05}
{"train_loss": 0.06021459400653839, "global_step": 168574, "epoch": 1894, "lr": 3.154742533440671e-05}
{"train_loss": 0.11252705752849579, "global_step": 168575, "epoch": 1894, "lr": 3.15468865232596e-05}
{"train_loss": 0.04951076582074165, "global_step": 168576, "epoch": 1894, "lr": 3.154634771459331e-05}
{"train_loss": 0.055305033922195435, "global_step": 168577, "epoch": 1894, "lr": 3.1545808908407864e-05}
{"train_loss": 0.08785586804151535, "global_step": 168578, "epoch": 1894, "lr": 3.1545270104703364e-05}
{"train_loss": 0.11248571425676346, "global_step": 168579, "epoch": 1894, "lr": 3.154473130347989e-05}
{"train_loss": 0.10161058604717255, "global_step": 168580, "epoch": 1894, "lr": 3.1544192504737494e-05}
{"train_loss": 0.06905112415552139, "global_step": 168581, "epoch": 1894, "lr": 3.154365370847626e-05}
{"train_loss": 0.03967677429318428, "global_step": 168582, "epoch": 1894, "lr": 3.154311491469626e-05}
{"train_loss": 0.03493773564696312, "global_step": 168583, "epoch": 1894, "lr": 3.154257612339756e-05}
{"train_loss": 0.06363241374492645, "global_step": 168584, "epoch": 1894, "lr": 3.1542037334580235e-05}
{"train_loss": 0.10960055142641068, "global_step": 168585, "epoch": 1894, "lr": 3.154149854824435e-05}
{"train_loss": 0.14039185643196106, "global_step": 168586, "epoch": 1894, "lr": 3.154095976439e-05}
{"train_loss": 0.12379054725170135, "global_step": 168587, "epoch": 1894, "lr": 3.1540420983017245e-05}
{"train_loss": 0.06835586577653885, "global_step": 168588, "epoch": 1894, "lr": 3.1539882204126135e-05}
{"train_loss": 0.06999998539686203, "global_step": 168589, "epoch": 1894, "lr": 3.1539343427716784e-05}
{"train_loss": 0.08561260998249054, "global_step": 168590, "epoch": 1894, "lr": 3.1538804653789226e-05}
{"train_loss": 0.11338433623313904, "global_step": 168591, "epoch": 1894, "lr": 3.153826588234356e-05}
{"train_loss": 0.07227518409490585, "global_step": 168592, "epoch": 1894, "lr": 3.153772711337985e-05}
{"train_loss": 0.07758446037769318, "global_step": 168593, "epoch": 1894, "lr": 3.153718834689817e-05}
{"train_loss": 0.06444323062896729, "global_step": 168594, "epoch": 1894, "lr": 3.153664958289857e-05}
{"train_loss": 0.12098325788974762, "global_step": 168595, "epoch": 1894, "lr": 3.153611082138117e-05}
{"train_loss": 0.09373877942562103, "global_step": 168596, "epoch": 1894, "lr": 3.1535572062345996e-05}
{"train_loss": 0.06535926461219788, "global_step": 168597, "epoch": 1894, "lr": 3.153503330579313e-05}
{"train_loss": 0.09868048131465912, "global_step": 168598, "epoch": 1894, "lr": 3.153449455172268e-05}
{"train_loss": 0.07119474560022354, "global_step": 168599, "epoch": 1894, "lr": 3.1533955800134675e-05}
{"train_loss": 0.078549325466156, "global_step": 168600, "epoch": 1894, "lr": 3.153341705102922e-05}
{"train_loss": 0.055993225425481796, "global_step": 168601, "epoch": 1894, "lr": 3.1532878304406366e-05}
{"train_loss": 0.09221458435058594, "global_step": 168602, "epoch": 1894, "lr": 3.153233956026619e-05}
{"train_loss": 0.14700832962989807, "global_step": 168603, "epoch": 1894, "lr": 3.153180081860876e-05}
{"train_loss": 0.05613710358738899, "global_step": 168604, "epoch": 1894, "lr": 3.153126207943416e-05}
{"train_loss": 0.08266808092594147, "global_step": 168605, "epoch": 1894, "lr": 3.153072334274245e-05}
{"train_loss": 0.07314199209213257, "global_step": 168606, "epoch": 1894, "lr": 3.153018460853373e-05}
{"train_loss": 0.1388138383626938, "global_step": 168607, "epoch": 1894, "lr": 3.1529645876808026e-05}
{"train_loss": 0.04697965085506439, "global_step": 168608, "epoch": 1894, "lr": 3.1529107147565454e-05}
{"train_loss": 0.06922131031751633, "global_step": 168609, "epoch": 1894, "lr": 3.152856842080605e-05}
{"train_loss": 0.1109500378370285, "global_step": 168610, "epoch": 1894, "lr": 3.1528029696529924e-05}
{"train_loss": 0.06432442367076874, "global_step": 168611, "epoch": 1894, "lr": 3.1527490974737115e-05}
{"train_loss": 0.09479358792304993, "global_step": 168612, "epoch": 1894, "lr": 3.152695225542772e-05}
{"train_loss": 0.08764152228832245, "global_step": 168613, "epoch": 1894, "lr": 3.152641353860178e-05}
{"train_loss": 0.09696535766124725, "global_step": 168614, "epoch": 1894, "lr": 3.152587482425941e-05}
{"train_loss": 0.23066453635692596, "global_step": 168615, "epoch": 1894, "lr": 3.152533611240066e-05}
{"train_loss": 0.1328493356704712, "global_step": 168616, "epoch": 1894, "lr": 3.1524797403025586e-05}
{"train_loss": 0.06145261973142624, "global_step": 168617, "epoch": 1894, "lr": 3.15242586961343e-05}
{"train_loss": 0.08304495364427567, "global_step": 168618, "epoch": 1894, "lr": 3.152371999172684e-05}
{"train_loss": 0.0752619206905365, "global_step": 168619, "epoch": 1894, "lr": 3.15231812898033e-05}
{"train_loss": 0.050929270684719086, "global_step": 168620, "epoch": 1894, "lr": 3.152264259036374e-05}
{"train_loss": 0.08498742431402206, "global_step": 168621, "epoch": 1894, "lr": 3.152210389340823e-05}
{"train_loss": 0.12332955002784729, "global_step": 168622, "epoch": 1894, "lr": 3.152156519893685e-05}
{"train_loss": 0.03302343562245369, "global_step": 168623, "epoch": 1894, "lr": 3.152102650694967e-05}
{"train_loss": 0.09674081206321716, "global_step": 168624, "epoch": 1894, "lr": 3.1520487817446766e-05}
{"train_loss": 0.07240993529558182, "global_step": 168625, "epoch": 1894, "lr": 3.151994913042821e-05}
{"train_loss": 0.0787026584148407, "global_step": 168626, "epoch": 1894, "lr": 3.151941044589406e-05}
{"train_loss": 0.08561261743307114, "global_step": 168627, "epoch": 1894, "lr": 3.151887176384441e-05}
{"train_loss": 0.12675891816616058, "global_step": 168628, "epoch": 1894, "lr": 3.1518333084279314e-05}
{"train_loss": 0.13136833906173706, "global_step": 168629, "epoch": 1894, "lr": 3.151779440719887e-05}
{"train_loss": 0.141618549823761, "global_step": 168630, "epoch": 1894, "lr": 3.151725573260311e-05}
{"train_loss": 0.13275282084941864, "global_step": 168631, "epoch": 1894, "lr": 3.1516717060492155e-05}
{"train_loss": 0.09561339020729065, "global_step": 168632, "epoch": 1894, "lr": 3.151617839086602e-05}
{"train_loss": 0.07849340885877609, "global_step": 168633, "epoch": 1894, "lr": 3.151563972372482e-05}
{"train_loss": 0.05636424571275711, "global_step": 168634, "epoch": 1894, "lr": 3.151510105906863e-05}
{"train_loss": 0.07091695815324783, "global_step": 168635, "epoch": 1894, "lr": 3.1514562396897495e-05}
{"train_loss": 0.09074337780475616, "global_step": 168636, "epoch": 1894, "lr": 3.151402373721152e-05}
{"train_loss": 0.07970226556062698, "global_step": 168637, "epoch": 1894, "lr": 3.151348508001074e-05}
{"train_loss": 0.061196405440568924, "global_step": 168638, "epoch": 1894, "lr": 3.1512946425295263e-05}
{"train_loss": 0.17898355424404144, "global_step": 168639, "epoch": 1894, "lr": 3.1512407773065144e-05}
{"train_loss": 0.18961063027381897, "global_step": 168640, "epoch": 1894, "lr": 3.151186912332045e-05}
{"train_loss": 0.0501277856528759, "global_step": 168641, "epoch": 1894, "lr": 3.1511330476061254e-05}
{"train_loss": 0.10564586520195007, "global_step": 168642, "epoch": 1894, "lr": 3.151079183128765e-05}
{"train_loss": 0.056799232959747314, "global_step": 168643, "epoch": 1894, "lr": 3.151025318899968e-05}
{"train_loss": 0.03586761653423309, "global_step": 168644, "epoch": 1894, "lr": 3.1509714549197446e-05}
{"train_loss": 0.08405283093452454, "global_step": 168645, "epoch": 1894, "lr": 3.150917591188098e-05}
{"train_loss": 0.08061179518699646, "global_step": 168646, "epoch": 1894, "lr": 3.150863727705041e-05}
{"train_loss": 0.09536623954772949, "global_step": 168647, "epoch": 1894, "lr": 3.150809864470575e-05}
{"train_loss": 0.052834585309028625, "global_step": 168648, "epoch": 1894, "lr": 3.150756001484713e-05}
{"train_loss": 0.06224729120731354, "global_step": 168649, "epoch": 1894, "lr": 3.1507021387474564e-05}
{"train_loss": 0.06935521215200424, "global_step": 168650, "epoch": 1894, "lr": 3.1506482762588165e-05}
{"train_loss": 0.11204187572002411, "global_step": 168651, "epoch": 1894, "lr": 3.1505944140188005e-05}
{"train_loss": 0.1001993790268898, "global_step": 168652, "epoch": 1894, "lr": 3.150540552027412e-05}
{"train_loss": 0.10657749325037003, "global_step": 168653, "epoch": 1894, "lr": 3.150486690284664e-05}
{"train_loss": 0.08960874315895391, "global_step": 168654, "epoch": 1894, "lr": 3.150432828790558e-05, "val_loss": 7.02382755279541}
{"train_loss": 0.04270066320896149, "global_step": 168655, "epoch": 1895, "lr": 3.150378967545105e-05}
{"train_loss": 0.08594142645597458, "global_step": 168656, "epoch": 1895, "lr": 3.150325106548311e-05}
{"train_loss": 0.10006902366876602, "global_step": 168657, "epoch": 1895, "lr": 3.150271245800184e-05}
{"train_loss": 0.08261963725090027, "global_step": 168658, "epoch": 1895, "lr": 3.150217385300729e-05}
{"train_loss": 0.07191473990678787, "global_step": 168659, "epoch": 1895, "lr": 3.150163525049956e-05}
{"train_loss": 0.08182341605424881, "global_step": 168660, "epoch": 1895, "lr": 3.15010966504787e-05}
{"train_loss": 0.11345621943473816, "global_step": 168661, "epoch": 1895, "lr": 3.150055805294479e-05}
{"train_loss": 0.08998651057481766, "global_step": 168662, "epoch": 1895, "lr": 3.150001945789791e-05}
{"train_loss": 0.10773277282714844, "global_step": 168663, "epoch": 1895, "lr": 3.1499480865338126e-05}
{"train_loss": 0.07996195554733276, "global_step": 168664, "epoch": 1895, "lr": 3.149894227526551e-05}
{"train_loss": 0.037216342985630035, "global_step": 168665, "epoch": 1895, "lr": 3.149840368768014e-05}
{"train_loss": 0.08196413516998291, "global_step": 168666, "epoch": 1895, "lr": 3.149786510258208e-05}
{"train_loss": 0.09977665543556213, "global_step": 168667, "epoch": 1895, "lr": 3.1497326519971414e-05}
{"train_loss": 0.08984799683094025, "global_step": 168668, "epoch": 1895, "lr": 3.149678793984819e-05}
{"train_loss": 0.12466175854206085, "global_step": 168669, "epoch": 1895, "lr": 3.14962493622125e-05}
{"train_loss": 0.12254621833562851, "global_step": 168670, "epoch": 1895, "lr": 3.149571078706442e-05}
{"train_loss": 0.09824449568986893, "global_step": 168671, "epoch": 1895, "lr": 3.1495172214404015e-05}
{"train_loss": 0.1077856719493866, "global_step": 168672, "epoch": 1895, "lr": 3.149463364423138e-05}
{"train_loss": 0.07468003779649734, "global_step": 168673, "epoch": 1895, "lr": 3.1494095076546524e-05}
{"train_loss": 0.1475803703069687, "global_step": 168674, "epoch": 1895, "lr": 3.1493556511349595e-05}
{"train_loss": 0.09686706960201263, "global_step": 168675, "epoch": 1895, "lr": 3.149301794864061e-05}
{"train_loss": 0.14594730734825134, "global_step": 168676, "epoch": 1895, "lr": 3.149247938841969e-05}
{"train_loss": 0.14524784684181213, "global_step": 168677, "epoch": 1895, "lr": 3.149194083068685e-05}
{"train_loss": 0.06693769991397858, "global_step": 168678, "epoch": 1895, "lr": 3.149140227544222e-05}
{"train_loss": 0.06714919954538345, "global_step": 168679, "epoch": 1895, "lr": 3.149086372268584e-05}
{"train_loss": 0.06584988534450531, "global_step": 168680, "epoch": 1895, "lr": 3.149032517241778e-05}
{"train_loss": 0.12224538624286652, "global_step": 168681, "epoch": 1895, "lr": 3.148978662463812e-05}
{"train_loss": 0.05238083004951477, "global_step": 168682, "epoch": 1895, "lr": 3.148924807934693e-05}
{"train_loss": 0.0688791498541832, "global_step": 168683, "epoch": 1895, "lr": 3.148870953654429e-05}
{"train_loss": 0.11254207044839859, "global_step": 168684, "epoch": 1895, "lr": 3.148817099623028e-05}
{"train_loss": 0.08761079609394073, "global_step": 168685, "epoch": 1895, "lr": 3.148763245840493e-05}
{"train_loss": 0.06850408017635345, "global_step": 168686, "epoch": 1895, "lr": 3.1487093923068364e-05}
{"train_loss": 0.09627789258956909, "global_step": 168687, "epoch": 1895, "lr": 3.148655539022062e-05}
{"train_loss": 0.07561945170164108, "global_step": 168688, "epoch": 1895, "lr": 3.1486016859861776e-05}
{"train_loss": 0.08187238872051239, "global_step": 168689, "epoch": 1895, "lr": 3.148547833199194e-05}
{"train_loss": 0.11132242530584335, "global_step": 168690, "epoch": 1895, "lr": 3.148493980661113e-05}
{"train_loss": 0.12275766581296921, "global_step": 168691, "epoch": 1895, "lr": 3.148440128371947e-05}
{"train_loss": 0.13968756794929504, "global_step": 168692, "epoch": 1895, "lr": 3.1483862763316976e-05}
{"train_loss": 0.11595406383275986, "global_step": 168693, "epoch": 1895, "lr": 3.148332424540378e-05}
{"train_loss": 0.12886671721935272, "global_step": 168694, "epoch": 1895, "lr": 3.14827857299799e-05}
{"train_loss": 0.07886780053377151, "global_step": 168695, "epoch": 1895, "lr": 3.148224721704546e-05}
{"train_loss": 0.11507556587457657, "global_step": 168696, "epoch": 1895, "lr": 3.148170870660048e-05}
{"train_loss": 0.042315296828746796, "global_step": 168697, "epoch": 1895, "lr": 3.148117019864508e-05}
{"train_loss": 0.1258750855922699, "global_step": 168698, "epoch": 1895, "lr": 3.1480631693179295e-05}
{"train_loss": 0.08060436695814133, "global_step": 168699, "epoch": 1895, "lr": 3.148009319020322e-05}
{"train_loss": 0.08855345100164413, "global_step": 168700, "epoch": 1895, "lr": 3.147955468971692e-05}
{"train_loss": 0.11600612103939056, "global_step": 168701, "epoch": 1895, "lr": 3.1479016191720465e-05}
{"train_loss": 0.109979048371315, "global_step": 168702, "epoch": 1895, "lr": 3.147847769621394e-05}
{"train_loss": 0.07439112663269043, "global_step": 168703, "epoch": 1895, "lr": 3.1477939203197416e-05}
{"train_loss": 0.05289805307984352, "global_step": 168704, "epoch": 1895, "lr": 3.147740071267093e-05}
{"train_loss": 0.11235179007053375, "global_step": 168705, "epoch": 1895, "lr": 3.147686222463459e-05}
{"train_loss": 0.11329459398984909, "global_step": 168706, "epoch": 1895, "lr": 3.147632373908847e-05}
{"train_loss": 0.05412716045975685, "global_step": 168707, "epoch": 1895, "lr": 3.147578525603262e-05}
{"train_loss": 0.09645899385213852, "global_step": 168708, "epoch": 1895, "lr": 3.147524677546714e-05}
{"train_loss": 0.06069674342870712, "global_step": 168709, "epoch": 1895, "lr": 3.147470829739207e-05}
{"train_loss": 0.08019649982452393, "global_step": 168710, "epoch": 1895, "lr": 3.147416982180752e-05}
{"train_loss": 0.05377625301480293, "global_step": 168711, "epoch": 1895, "lr": 3.147363134871352e-05}
{"train_loss": 0.060438431799411774, "global_step": 168712, "epoch": 1895, "lr": 3.1473092878110186e-05}
{"train_loss": 0.1048734188079834, "global_step": 168713, "epoch": 1895, "lr": 3.1472554409997545e-05}
{"train_loss": 0.13418099284172058, "global_step": 168714, "epoch": 1895, "lr": 3.147201594437572e-05}
{"train_loss": 0.1655990034341812, "global_step": 168715, "epoch": 1895, "lr": 3.147147748124473e-05}
{"train_loss": 0.0479358434677124, "global_step": 168716, "epoch": 1895, "lr": 3.1470939020604704e-05}
{"train_loss": 0.10119114071130753, "global_step": 168717, "epoch": 1895, "lr": 3.1470400562455663e-05}
{"train_loss": 0.14350798726081848, "global_step": 168718, "epoch": 1895, "lr": 3.1469862106797706e-05}
{"train_loss": 0.10503050684928894, "global_step": 168719, "epoch": 1895, "lr": 3.14693236536309e-05}
{"train_loss": 0.08561578392982483, "global_step": 168720, "epoch": 1895, "lr": 3.146878520295532e-05}
{"train_loss": 0.11366450041532516, "global_step": 168721, "epoch": 1895, "lr": 3.146824675477103e-05}
{"train_loss": 0.07252088934183121, "global_step": 168722, "epoch": 1895, "lr": 3.146770830907813e-05}
{"train_loss": 0.10390748828649521, "global_step": 168723, "epoch": 1895, "lr": 3.1467169865876636e-05}
{"train_loss": 0.072206512093544, "global_step": 168724, "epoch": 1895, "lr": 3.146663142516666e-05}
{"train_loss": 0.062357816845178604, "global_step": 168725, "epoch": 1895, "lr": 3.1466092986948294e-05}
{"train_loss": 0.07471806555986404, "global_step": 168726, "epoch": 1895, "lr": 3.1465554551221564e-05}
{"train_loss": 0.10961643606424332, "global_step": 168727, "epoch": 1895, "lr": 3.146501611798658e-05}
{"train_loss": 0.1031084656715393, "global_step": 168728, "epoch": 1895, "lr": 3.146447768724339e-05}
{"train_loss": 0.12423870712518692, "global_step": 168729, "epoch": 1895, "lr": 3.146393925899208e-05}
{"train_loss": 0.060623906552791595, "global_step": 168730, "epoch": 1895, "lr": 3.146340083323271e-05}
{"train_loss": 0.05245043709874153, "global_step": 168731, "epoch": 1895, "lr": 3.146286240996538e-05}
{"train_loss": 0.0896512120962143, "global_step": 168732, "epoch": 1895, "lr": 3.146232398919011e-05}
{"train_loss": 0.06875600665807724, "global_step": 168733, "epoch": 1895, "lr": 3.146178557090703e-05}
{"train_loss": 0.07661812752485275, "global_step": 168734, "epoch": 1895, "lr": 3.1461247155116166e-05}
{"train_loss": 0.05729726329445839, "global_step": 168735, "epoch": 1895, "lr": 3.1460708741817635e-05}
{"train_loss": 0.04888760298490524, "global_step": 168736, "epoch": 1895, "lr": 3.1460170331011476e-05}
{"train_loss": 0.06491097062826157, "global_step": 168737, "epoch": 1895, "lr": 3.1459631922697775e-05}
{"train_loss": 0.0892074778676033, "global_step": 168738, "epoch": 1895, "lr": 3.145909351687659e-05}
{"train_loss": 0.16361694037914276, "global_step": 168739, "epoch": 1895, "lr": 3.145855511354801e-05}
{"train_loss": 0.06223686784505844, "global_step": 168740, "epoch": 1895, "lr": 3.14580167127121e-05}
{"train_loss": 0.11087077111005783, "global_step": 168741, "epoch": 1895, "lr": 3.145747831436893e-05}
{"train_loss": 0.12233898788690567, "global_step": 168742, "epoch": 1895, "lr": 3.145693991851858e-05}
{"train_loss": 0.09201798060636841, "global_step": 168743, "epoch": 1895, "lr": 3.1456401525161114e-05, "val_loss": 7.107117652893066, "train_action_mse_error": 19.8227596282959}
{"train_loss": 0.09922943264245987, "global_step": 168744, "epoch": 1896, "lr": 3.1455863134296623e-05}
{"train_loss": 0.11878462135791779, "global_step": 168745, "epoch": 1896, "lr": 3.145532474592514e-05}
{"train_loss": 0.11985812336206436, "global_step": 168746, "epoch": 1896, "lr": 3.1454786360046795e-05}
{"train_loss": 0.08009980618953705, "global_step": 168747, "epoch": 1896, "lr": 3.14542479766616e-05}
{"train_loss": 0.08385314047336578, "global_step": 168748, "epoch": 1896, "lr": 3.1453709595769675e-05}
{"train_loss": 0.057266365736722946, "global_step": 168749, "epoch": 1896, "lr": 3.145317121737105e-05}
{"train_loss": 0.08953544497489929, "global_step": 168750, "epoch": 1896, "lr": 3.1452632841465845e-05}
{"train_loss": 0.05687505751848221, "global_step": 168751, "epoch": 1896, "lr": 3.145209446805409e-05}
{"train_loss": 0.048057831823825836, "global_step": 168752, "epoch": 1896, "lr": 3.145155609713589e-05}
{"train_loss": 0.1063278466463089, "global_step": 168753, "epoch": 1896, "lr": 3.145101772871129e-05}
{"train_loss": 0.11774542182683945, "global_step": 168754, "epoch": 1896, "lr": 3.1450479362780385e-05}
{"train_loss": 0.0439780168235302, "global_step": 168755, "epoch": 1896, "lr": 3.144994099934322e-05}
{"train_loss": 0.08487261086702347, "global_step": 168756, "epoch": 1896, "lr": 3.14494026383999e-05}
{"train_loss": 0.06727713346481323, "global_step": 168757, "epoch": 1896, "lr": 3.144886427995047e-05}
{"train_loss": 0.03809220716357231, "global_step": 168758, "epoch": 1896, "lr": 3.144832592399503e-05}
{"train_loss": 0.09428291022777557, "global_step": 168759, "epoch": 1896, "lr": 3.144778757053362e-05}
{"train_loss": 0.05504542216658592, "global_step": 168760, "epoch": 1896, "lr": 3.144724921956633e-05}
{"train_loss": 0.08558113873004913, "global_step": 168761, "epoch": 1896, "lr": 3.144671087109324e-05}
{"train_loss": 0.07132803648710251, "global_step": 168762, "epoch": 1896, "lr": 3.1446172525114404e-05}
{"train_loss": 0.09076318889856339, "global_step": 168763, "epoch": 1896, "lr": 3.144563418162992e-05}
{"train_loss": 0.061264894902706146, "global_step": 168764, "epoch": 1896, "lr": 3.144509584063983e-05}
{"train_loss": 0.08003917336463928, "global_step": 168765, "epoch": 1896, "lr": 3.1444557502144234e-05}
{"train_loss": 0.05987512692809105, "global_step": 168766, "epoch": 1896, "lr": 3.1444019166143166e-05}
{"train_loss": 0.061197757720947266, "global_step": 168767, "epoch": 1896, "lr": 3.144348083263675e-05}
{"train_loss": 0.04986400529742241, "global_step": 168768, "epoch": 1896, "lr": 3.144294250162502e-05}
{"train_loss": 0.07426096498966217, "global_step": 168769, "epoch": 1896, "lr": 3.144240417310806e-05}
{"train_loss": 0.19899944961071014, "global_step": 168770, "epoch": 1896, "lr": 3.1441865847085936e-05}
{"train_loss": 0.1023072600364685, "global_step": 168771, "epoch": 1896, "lr": 3.144132752355875e-05}
{"train_loss": 0.11346342414617538, "global_step": 168772, "epoch": 1896, "lr": 3.144078920252653e-05}
{"train_loss": 0.11592505127191544, "global_step": 168773, "epoch": 1896, "lr": 3.144025088398939e-05}
{"train_loss": 0.061175357550382614, "global_step": 168774, "epoch": 1896, "lr": 3.143971256794736e-05}
{"train_loss": 0.1253347396850586, "global_step": 168775, "epoch": 1896, "lr": 3.143917425440054e-05}
{"train_loss": 0.12673518061637878, "global_step": 168776, "epoch": 1896, "lr": 3.1438635943349006e-05}
{"train_loss": 0.06233799085021019, "global_step": 168777, "epoch": 1896, "lr": 3.14380976347928e-05}
{"train_loss": 0.1893889158964157, "global_step": 168778, "epoch": 1896, "lr": 3.143755932873205e-05}
{"train_loss": 0.06325091421604156, "global_step": 168779, "epoch": 1896, "lr": 3.143702102516676e-05}
{"train_loss": 0.04820161685347557, "global_step": 168780, "epoch": 1896, "lr": 3.143648272409706e-05}
{"train_loss": 0.059315916150808334, "global_step": 168781, "epoch": 1896, "lr": 3.143594442552298e-05}
{"train_loss": 0.07209516316652298, "global_step": 168782, "epoch": 1896, "lr": 3.143540612944463e-05}
{"train_loss": 0.11738254129886627, "global_step": 168783, "epoch": 1896, "lr": 3.1434867835862035e-05}
{"train_loss": 0.06881239265203476, "global_step": 168784, "epoch": 1896, "lr": 3.143432954477533e-05}
{"train_loss": 0.11303328722715378, "global_step": 168785, "epoch": 1896, "lr": 3.143379125618453e-05}
{"train_loss": 0.11769536882638931, "global_step": 168786, "epoch": 1896, "lr": 3.143325297008974e-05}
{"train_loss": 0.12004977464675903, "global_step": 168787, "epoch": 1896, "lr": 3.143271468649102e-05}
{"train_loss": 0.09588848054409027, "global_step": 168788, "epoch": 1896, "lr": 3.143217640538845e-05}
{"train_loss": 0.13127903640270233, "global_step": 168789, "epoch": 1896, "lr": 3.143163812678208e-05}
{"train_loss": 0.10963136702775955, "global_step": 168790, "epoch": 1896, "lr": 3.143109985067203e-05}
{"train_loss": 0.07214615494012833, "global_step": 168791, "epoch": 1896, "lr": 3.1430561577058316e-05}
{"train_loss": 0.09524060040712357, "global_step": 168792, "epoch": 1896, "lr": 3.143002330594106e-05}
{"train_loss": 0.07241646945476532, "global_step": 168793, "epoch": 1896, "lr": 3.1429485037320285e-05}
{"train_loss": 0.08991027623414993, "global_step": 168794, "epoch": 1896, "lr": 3.142894677119611e-05}
{"train_loss": 0.0901886448264122, "global_step": 168795, "epoch": 1896, "lr": 3.142840850756858e-05}
{"train_loss": 0.11826442182064056, "global_step": 168796, "epoch": 1896, "lr": 3.142787024643776e-05}
{"train_loss": 0.09951683133840561, "global_step": 168797, "epoch": 1896, "lr": 3.142733198780376e-05}
{"train_loss": 0.04309559240937233, "global_step": 168798, "epoch": 1896, "lr": 3.14267937316666e-05}
{"train_loss": 0.14493806660175323, "global_step": 168799, "epoch": 1896, "lr": 3.1426255478026416e-05}
{"train_loss": 0.13295631110668182, "global_step": 168800, "epoch": 1896, "lr": 3.142571722688322e-05}
{"train_loss": 0.06878001242876053, "global_step": 168801, "epoch": 1896, "lr": 3.142517897823712e-05}
{"train_loss": 0.0898393914103508, "global_step": 168802, "epoch": 1896, "lr": 3.142464073208817e-05}
{"train_loss": 0.06118934974074364, "global_step": 168803, "epoch": 1896, "lr": 3.1424102488436475e-05}
{"train_loss": 0.07555055618286133, "global_step": 168804, "epoch": 1896, "lr": 3.142356424728206e-05}
{"train_loss": 0.11353690922260284, "global_step": 168805, "epoch": 1896, "lr": 3.142302600862503e-05}
{"train_loss": 0.07517998665571213, "global_step": 168806, "epoch": 1896, "lr": 3.1422487772465436e-05}
{"train_loss": 0.027820460498332977, "global_step": 168807, "epoch": 1896, "lr": 3.1421949538803376e-05}
{"train_loss": 0.11264881491661072, "global_step": 168808, "epoch": 1896, "lr": 3.142141130763889e-05}
{"train_loss": 0.09718869626522064, "global_step": 168809, "epoch": 1896, "lr": 3.1420873078972096e-05}
{"train_loss": 0.1252794712781906, "global_step": 168810, "epoch": 1896, "lr": 3.142033485280301e-05}
{"train_loss": 0.10375247150659561, "global_step": 168811, "epoch": 1896, "lr": 3.141979662913176e-05}
{"train_loss": 0.07865291833877563, "global_step": 168812, "epoch": 1896, "lr": 3.141925840795837e-05}
{"train_loss": 0.08508403599262238, "global_step": 168813, "epoch": 1896, "lr": 3.1418720189282955e-05}
{"train_loss": 0.10851595550775528, "global_step": 168814, "epoch": 1896, "lr": 3.141818197310555e-05}
{"train_loss": 0.12484405934810638, "global_step": 168815, "epoch": 1896, "lr": 3.141764375942624e-05}
{"train_loss": 0.10274682939052582, "global_step": 168816, "epoch": 1896, "lr": 3.141710554824513e-05}
{"train_loss": 0.10538415610790253, "global_step": 168817, "epoch": 1896, "lr": 3.141656733956223e-05}
{"train_loss": 0.10639386624097824, "global_step": 168818, "epoch": 1896, "lr": 3.141602913337767e-05}
{"train_loss": 0.07210665941238403, "global_step": 168819, "epoch": 1896, "lr": 3.141549092969148e-05}
{"train_loss": 0.04324965178966522, "global_step": 168820, "epoch": 1896, "lr": 3.141495272850376e-05}
{"train_loss": 0.09395020455121994, "global_step": 168821, "epoch": 1896, "lr": 3.1414414529814575e-05}
{"train_loss": 0.1030830517411232, "global_step": 168822, "epoch": 1896, "lr": 3.1413876333624e-05}
{"train_loss": 0.09391509741544724, "global_step": 168823, "epoch": 1896, "lr": 3.141333813993208e-05}
{"train_loss": 0.12256141006946564, "global_step": 168824, "epoch": 1896, "lr": 3.141279994873894e-05}
{"train_loss": 0.08173759281635284, "global_step": 168825, "epoch": 1896, "lr": 3.1412261760044606e-05}
{"train_loss": 0.11659795045852661, "global_step": 168826, "epoch": 1896, "lr": 3.1411723573849174e-05}
{"train_loss": 0.0696435496211052, "global_step": 168827, "epoch": 1896, "lr": 3.14111853901527e-05}
{"train_loss": 0.06852960586547852, "global_step": 168828, "epoch": 1896, "lr": 3.141064720895527e-05}
{"train_loss": 0.0467534065246582, "global_step": 168829, "epoch": 1896, "lr": 3.141010903025694e-05}
{"train_loss": 0.07900188863277435, "global_step": 168830, "epoch": 1896, "lr": 3.140957085405782e-05}
{"train_loss": 0.0651547983288765, "global_step": 168831, "epoch": 1896, "lr": 3.1409032680357935e-05}
{"train_loss": 0.09009015108092447, "global_step": 168832, "epoch": 1896, "lr": 3.1408494509157386e-05, "val_loss": 7.1775078773498535}
{"train_loss": 0.05701566860079765, "global_step": 168833, "epoch": 1897, "lr": 3.1407956340456255e-05}
{"train_loss": 0.053576260805130005, "global_step": 168834, "epoch": 1897, "lr": 3.140741817425457e-05}
{"train_loss": 0.09248558431863785, "global_step": 168835, "epoch": 1897, "lr": 3.1406880010552456e-05}
{"train_loss": 0.11074984818696976, "global_step": 168836, "epoch": 1897, "lr": 3.140634184934993e-05}
{"train_loss": 0.15356676280498505, "global_step": 168837, "epoch": 1897, "lr": 3.140580369064712e-05}
{"train_loss": 0.11910317093133926, "global_step": 168838, "epoch": 1897, "lr": 3.1405265534444064e-05}
{"train_loss": 0.14051567018032074, "global_step": 168839, "epoch": 1897, "lr": 3.140472738074085e-05}
{"train_loss": 0.11997925490140915, "global_step": 168840, "epoch": 1897, "lr": 3.140418922953753e-05}
{"train_loss": 0.07616939395666122, "global_step": 168841, "epoch": 1897, "lr": 3.140365108083421e-05}
{"train_loss": 0.08949263393878937, "global_step": 168842, "epoch": 1897, "lr": 3.1403112934630917e-05}
{"train_loss": 0.05498802289366722, "global_step": 168843, "epoch": 1897, "lr": 3.1402574790927774e-05}
{"train_loss": 0.13630414009094238, "global_step": 168844, "epoch": 1897, "lr": 3.140203664972481e-05}
{"train_loss": 0.08478046208620071, "global_step": 168845, "epoch": 1897, "lr": 3.1401498511022135e-05}
{"train_loss": 0.1123935803771019, "global_step": 168846, "epoch": 1897, "lr": 3.1400960374819774e-05}
{"train_loss": 0.1273903250694275, "global_step": 168847, "epoch": 1897, "lr": 3.140042224111785e-05}
{"train_loss": 0.05751077085733414, "global_step": 168848, "epoch": 1897, "lr": 3.1399884109916396e-05}
{"train_loss": 0.0650099366903305, "global_step": 168849, "epoch": 1897, "lr": 3.139934598121552e-05}
{"train_loss": 0.0739394873380661, "global_step": 168850, "epoch": 1897, "lr": 3.139880785501526e-05}
{"train_loss": 0.04674162715673447, "global_step": 168851, "epoch": 1897, "lr": 3.13982697313157e-05}
{"train_loss": 0.09545720368623734, "global_step": 168852, "epoch": 1897, "lr": 3.1397731610116935e-05}
{"train_loss": 0.06475795060396194, "global_step": 168853, "epoch": 1897, "lr": 3.1397193491419e-05}
{"train_loss": 0.06478403508663177, "global_step": 168854, "epoch": 1897, "lr": 3.1396655375222e-05}
{"train_loss": 0.08418221771717072, "global_step": 168855, "epoch": 1897, "lr": 3.139611726152598e-05}
{"train_loss": 0.08369403332471848, "global_step": 168856, "epoch": 1897, "lr": 3.139557915033104e-05}
{"train_loss": 0.07622437924146652, "global_step": 168857, "epoch": 1897, "lr": 3.139504104163722e-05}
{"train_loss": 0.046943169087171555, "global_step": 168858, "epoch": 1897, "lr": 3.1394502935444625e-05}
{"train_loss": 0.0576629713177681, "global_step": 168859, "epoch": 1897, "lr": 3.13939648317533e-05}
{"train_loss": 0.1347721666097641, "global_step": 168860, "epoch": 1897, "lr": 3.139342673056335e-05}
{"train_loss": 0.07871145755052567, "global_step": 168861, "epoch": 1897, "lr": 3.13928886318748e-05}
{"train_loss": 0.07848457247018814, "global_step": 168862, "epoch": 1897, "lr": 3.139235053568777e-05}
{"train_loss": 0.05696029216051102, "global_step": 168863, "epoch": 1897, "lr": 3.1391812442002296e-05}
{"train_loss": 0.13831660151481628, "global_step": 168864, "epoch": 1897, "lr": 3.139127435081848e-05}
{"train_loss": 0.08743438869714737, "global_step": 168865, "epoch": 1897, "lr": 3.1390736262136364e-05}
{"train_loss": 0.06546246260404587, "global_step": 168866, "epoch": 1897, "lr": 3.139019817595606e-05}
{"train_loss": 0.13144488632678986, "global_step": 168867, "epoch": 1897, "lr": 3.13896600922776e-05}
{"train_loss": 0.08269249647855759, "global_step": 168868, "epoch": 1897, "lr": 3.1389122011101066e-05}
{"train_loss": 0.13130861520767212, "global_step": 168869, "epoch": 1897, "lr": 3.138858393242655e-05}
{"train_loss": 0.08912583440542221, "global_step": 168870, "epoch": 1897, "lr": 3.1388045856254115e-05}
{"train_loss": 0.13887009024620056, "global_step": 168871, "epoch": 1897, "lr": 3.138750778258383e-05}
{"train_loss": 0.0952368900179863, "global_step": 168872, "epoch": 1897, "lr": 3.138696971141576e-05}
{"train_loss": 0.1713329404592514, "global_step": 168873, "epoch": 1897, "lr": 3.138643164275e-05}
{"train_loss": 0.11574382334947586, "global_step": 168874, "epoch": 1897, "lr": 3.138589357658659e-05}
{"train_loss": 0.10617279261350632, "global_step": 168875, "epoch": 1897, "lr": 3.1385355512925636e-05}
{"train_loss": 0.05388559028506279, "global_step": 168876, "epoch": 1897, "lr": 3.1384817451767175e-05}
{"train_loss": 0.07073568552732468, "global_step": 168877, "epoch": 1897, "lr": 3.138427939311132e-05}
{"train_loss": 0.07264707237482071, "global_step": 168878, "epoch": 1897, "lr": 3.13837413369581e-05}
{"train_loss": 0.05412698909640312, "global_step": 168879, "epoch": 1897, "lr": 3.138320328330763e-05}
{"train_loss": 0.11983349919319153, "global_step": 168880, "epoch": 1897, "lr": 3.138266523215994e-05}
{"train_loss": 0.1489601880311966, "global_step": 168881, "epoch": 1897, "lr": 3.138212718351514e-05}
{"train_loss": 0.039634399116039276, "global_step": 168882, "epoch": 1897, "lr": 3.138158913737327e-05}
{"train_loss": 0.1369485855102539, "global_step": 168883, "epoch": 1897, "lr": 3.138105109373444e-05}
{"train_loss": 0.042738959193229675, "global_step": 168884, "epoch": 1897, "lr": 3.1380513052598684e-05}
{"train_loss": 0.056862592697143555, "global_step": 168885, "epoch": 1897, "lr": 3.13799750139661e-05}
{"train_loss": 0.10813609510660172, "global_step": 168886, "epoch": 1897, "lr": 3.137943697783674e-05}
{"train_loss": 0.12245718389749527, "global_step": 168887, "epoch": 1897, "lr": 3.137889894421069e-05}
{"train_loss": 0.13557977974414825, "global_step": 168888, "epoch": 1897, "lr": 3.137836091308803e-05}
{"train_loss": 0.08798041939735413, "global_step": 168889, "epoch": 1897, "lr": 3.137782288446882e-05}
{"train_loss": 0.09641850739717484, "global_step": 168890, "epoch": 1897, "lr": 3.137728485835314e-05}
{"train_loss": 0.05621090531349182, "global_step": 168891, "epoch": 1897, "lr": 3.137674683474104e-05}
{"train_loss": 0.06553807109594345, "global_step": 168892, "epoch": 1897, "lr": 3.137620881363263e-05}
{"train_loss": 0.12455011159181595, "global_step": 168893, "epoch": 1897, "lr": 3.137567079502794e-05}
{"train_loss": 0.06500319391489029, "global_step": 168894, "epoch": 1897, "lr": 3.1375132778927085e-05}
{"train_loss": 0.07517923414707184, "global_step": 168895, "epoch": 1897, "lr": 3.1374594765330094e-05}
{"train_loss": 0.07195914536714554, "global_step": 168896, "epoch": 1897, "lr": 3.137405675423709e-05}
{"train_loss": 0.10359632968902588, "global_step": 168897, "epoch": 1897, "lr": 3.137351874564809e-05}
{"train_loss": 0.08028295636177063, "global_step": 168898, "epoch": 1897, "lr": 3.1372980739563214e-05}
{"train_loss": 0.11401226371526718, "global_step": 168899, "epoch": 1897, "lr": 3.137244273598249e-05}
{"train_loss": 0.09153754264116287, "global_step": 168900, "epoch": 1897, "lr": 3.137190473490604e-05}
{"train_loss": 0.1042424812912941, "global_step": 168901, "epoch": 1897, "lr": 3.137136673633388e-05}
{"train_loss": 0.1334920972585678, "global_step": 168902, "epoch": 1897, "lr": 3.137082874026615e-05}
{"train_loss": 0.08309132605791092, "global_step": 168903, "epoch": 1897, "lr": 3.1370290746702856e-05}
{"train_loss": 0.09641806036233902, "global_step": 168904, "epoch": 1897, "lr": 3.136975275564411e-05}
{"train_loss": 0.15162968635559082, "global_step": 168905, "epoch": 1897, "lr": 3.1369214767089974e-05}
{"train_loss": 0.08392202854156494, "global_step": 168906, "epoch": 1897, "lr": 3.136867678104052e-05}
{"train_loss": 0.05956206098198891, "global_step": 168907, "epoch": 1897, "lr": 3.136813879749582e-05}
{"train_loss": 0.0896296501159668, "global_step": 168908, "epoch": 1897, "lr": 3.1367600816455947e-05}
{"train_loss": 0.04433164373040199, "global_step": 168909, "epoch": 1897, "lr": 3.136706283792097e-05}
{"train_loss": 0.1085820123553276, "global_step": 168910, "epoch": 1897, "lr": 3.1366524861890967e-05}
{"train_loss": 0.10228515416383743, "global_step": 168911, "epoch": 1897, "lr": 3.136598688836602e-05}
{"train_loss": 0.037392813712358475, "global_step": 168912, "epoch": 1897, "lr": 3.136544891734616e-05}
{"train_loss": 0.11552944779396057, "global_step": 168913, "epoch": 1897, "lr": 3.136491094883152e-05}
{"train_loss": 0.10351511090993881, "global_step": 168914, "epoch": 1897, "lr": 3.136437298282211e-05}
{"train_loss": 0.10788363963365555, "global_step": 168915, "epoch": 1897, "lr": 3.136383501931806e-05}
{"train_loss": 0.07266608625650406, "global_step": 168916, "epoch": 1897, "lr": 3.1363297058319394e-05}
{"train_loss": 0.09327777475118637, "global_step": 168917, "epoch": 1897, "lr": 3.136275909982622e-05}
{"train_loss": 0.06296027451753616, "global_step": 168918, "epoch": 1897, "lr": 3.136222114383859e-05}
{"train_loss": 0.07674558460712433, "global_step": 168919, "epoch": 1897, "lr": 3.136168319035659e-05}
{"train_loss": 0.06246403232216835, "global_step": 168920, "epoch": 1897, "lr": 3.136114523938026e-05}
{"train_loss": 0.09098807550715597, "global_step": 168921, "epoch": 1897, "lr": 3.136060729090973e-05, "val_loss": 7.302163124084473}
{"train_loss": 0.09409177303314209, "global_step": 168922, "epoch": 1898, "lr": 3.1360069344945006e-05}
{"train_loss": 0.07080632448196411, "global_step": 168923, "epoch": 1898, "lr": 3.1359531401486205e-05}
{"train_loss": 0.05172525346279144, "global_step": 168924, "epoch": 1898, "lr": 3.13589934605334e-05}
{"train_loss": 0.0503130704164505, "global_step": 168925, "epoch": 1898, "lr": 3.135845552208665e-05}
{"train_loss": 0.06106870248913765, "global_step": 168926, "epoch": 1898, "lr": 3.135791758614603e-05}
{"train_loss": 0.08320549875497818, "global_step": 168927, "epoch": 1898, "lr": 3.13573796527116e-05}
{"train_loss": 0.09064082056283951, "global_step": 168928, "epoch": 1898, "lr": 3.135684172178345e-05}
{"train_loss": 0.09618408232927322, "global_step": 168929, "epoch": 1898, "lr": 3.135630379336164e-05}
{"train_loss": 0.12384642660617828, "global_step": 168930, "epoch": 1898, "lr": 3.135576586744626e-05}
{"train_loss": 0.0794682428240776, "global_step": 168931, "epoch": 1898, "lr": 3.1355227944037344e-05}
{"train_loss": 0.0974886566400528, "global_step": 168932, "epoch": 1898, "lr": 3.1354690023135024e-05}
{"train_loss": 0.08592633903026581, "global_step": 168933, "epoch": 1898, "lr": 3.135415210473931e-05}
{"train_loss": 0.08258325606584549, "global_step": 168934, "epoch": 1898, "lr": 3.135361418885032e-05}
{"train_loss": 0.1441962569952011, "global_step": 168935, "epoch": 1898, "lr": 3.1353076275468095e-05}
{"train_loss": 0.10940364748239517, "global_step": 168936, "epoch": 1898, "lr": 3.135253836459274e-05}
{"train_loss": 0.04722671955823898, "global_step": 168937, "epoch": 1898, "lr": 3.135200045622429e-05}
{"train_loss": 0.1280050426721573, "global_step": 168938, "epoch": 1898, "lr": 3.135146255036285e-05}
{"train_loss": 0.04450136795639992, "global_step": 168939, "epoch": 1898, "lr": 3.135092464700847e-05}
{"train_loss": 0.08794326335191727, "global_step": 168940, "epoch": 1898, "lr": 3.135038674616124e-05}
{"train_loss": 0.0662895068526268, "global_step": 168941, "epoch": 1898, "lr": 3.134984884782121e-05}
{"train_loss": 0.07537753880023956, "global_step": 168942, "epoch": 1898, "lr": 3.134931095198846e-05}
{"train_loss": 0.0725516751408577, "global_step": 168943, "epoch": 1898, "lr": 3.1348773058663096e-05}
{"train_loss": 0.0966028943657875, "global_step": 168944, "epoch": 1898, "lr": 3.134823516784515e-05}
{"train_loss": 0.04641911759972572, "global_step": 168945, "epoch": 1898, "lr": 3.1347697279534706e-05}
{"train_loss": 0.11451282352209091, "global_step": 168946, "epoch": 1898, "lr": 3.1347159393731825e-05}
{"train_loss": 0.0553378164768219, "global_step": 168947, "epoch": 1898, "lr": 3.134662151043661e-05}
{"train_loss": 0.09666857123374939, "global_step": 168948, "epoch": 1898, "lr": 3.1346083629649105e-05}
{"train_loss": 0.03958455100655556, "global_step": 168949, "epoch": 1898, "lr": 3.13455457513694e-05}
{"train_loss": 0.08631846308708191, "global_step": 168950, "epoch": 1898, "lr": 3.1345007875597546e-05}
{"train_loss": 0.044975556433200836, "global_step": 168951, "epoch": 1898, "lr": 3.134447000233364e-05}
{"train_loss": 0.08371800184249878, "global_step": 168952, "epoch": 1898, "lr": 3.134393213157772e-05}
{"train_loss": 0.08399838209152222, "global_step": 168953, "epoch": 1898, "lr": 3.134339426332991e-05}
{"train_loss": 0.08003111183643341, "global_step": 168954, "epoch": 1898, "lr": 3.134285639759023e-05}
{"train_loss": 0.10751663148403168, "global_step": 168955, "epoch": 1898, "lr": 3.1342318534358794e-05}
{"train_loss": 0.06801898032426834, "global_step": 168956, "epoch": 1898, "lr": 3.134178067363564e-05}
{"train_loss": 0.1199110820889473, "global_step": 168957, "epoch": 1898, "lr": 3.134124281542088e-05}
{"train_loss": 0.11438485980033875, "global_step": 168958, "epoch": 1898, "lr": 3.134070495971453e-05}
{"train_loss": 0.08952583372592926, "global_step": 168959, "epoch": 1898, "lr": 3.13401671065167e-05}
{"train_loss": 0.059128955006599426, "global_step": 168960, "epoch": 1898, "lr": 3.133962925582748e-05}
{"train_loss": 0.1192631796002388, "global_step": 168961, "epoch": 1898, "lr": 3.133909140764689e-05}
{"train_loss": 0.13057728111743927, "global_step": 168962, "epoch": 1898, "lr": 3.1338553561975064e-05}
{"train_loss": 0.07581064850091934, "global_step": 168963, "epoch": 1898, "lr": 3.133801571881203e-05}
{"train_loss": 0.0706101655960083, "global_step": 168964, "epoch": 1898, "lr": 3.1337477878157864e-05}
{"train_loss": 0.07203643769025803, "global_step": 168965, "epoch": 1898, "lr": 3.1336940040012655e-05}
{"train_loss": 0.09976188093423843, "global_step": 168966, "epoch": 1898, "lr": 3.1336402204376466e-05}
{"train_loss": 0.054965950548648834, "global_step": 168967, "epoch": 1898, "lr": 3.133586437124937e-05}
{"train_loss": 0.06941505521535873, "global_step": 168968, "epoch": 1898, "lr": 3.1335326540631446e-05}
{"train_loss": 0.07313995808362961, "global_step": 168969, "epoch": 1898, "lr": 3.133478871252274e-05}
{"train_loss": 0.09350569546222687, "global_step": 168970, "epoch": 1898, "lr": 3.133425088692337e-05}
{"train_loss": 0.08513881266117096, "global_step": 168971, "epoch": 1898, "lr": 3.1333713063833355e-05}
{"train_loss": 0.03814804181456566, "global_step": 168972, "epoch": 1898, "lr": 3.133317524325282e-05}
{"train_loss": 0.142981618642807, "global_step": 168973, "epoch": 1898, "lr": 3.133263742518179e-05}
{"train_loss": 0.07610426843166351, "global_step": 168974, "epoch": 1898, "lr": 3.133209960962038e-05}
{"train_loss": 0.09690561145544052, "global_step": 168975, "epoch": 1898, "lr": 3.133156179656862e-05}
{"train_loss": 0.09568841755390167, "global_step": 168976, "epoch": 1898, "lr": 3.133102398602662e-05}
{"train_loss": 0.09697023034095764, "global_step": 168977, "epoch": 1898, "lr": 3.133048617799442e-05}
{"train_loss": 0.0856136679649353, "global_step": 168978, "epoch": 1898, "lr": 3.132994837247212e-05}
{"train_loss": 0.07335606962442398, "global_step": 168979, "epoch": 1898, "lr": 3.132941056945978e-05}
{"train_loss": 0.13683831691741943, "global_step": 168980, "epoch": 1898, "lr": 3.132887276895747e-05}
{"train_loss": 0.09978718310594559, "global_step": 168981, "epoch": 1898, "lr": 3.132833497096528e-05}
{"train_loss": 0.0659409761428833, "global_step": 168982, "epoch": 1898, "lr": 3.132779717548325e-05}
{"train_loss": 0.09002311527729034, "global_step": 168983, "epoch": 1898, "lr": 3.132725938251149e-05}
{"train_loss": 0.09494497627019882, "global_step": 168984, "epoch": 1898, "lr": 3.132672159205003e-05}
{"train_loss": 0.07062660902738571, "global_step": 168985, "epoch": 1898, "lr": 3.132618380409898e-05}
{"train_loss": 0.03831319138407707, "global_step": 168986, "epoch": 1898, "lr": 3.132564601865839e-05}
{"train_loss": 0.07895389199256897, "global_step": 168987, "epoch": 1898, "lr": 3.132510823572835e-05}
{"train_loss": 0.12860508263111115, "global_step": 168988, "epoch": 1898, "lr": 3.13245704553089e-05}
{"train_loss": 0.09714360535144806, "global_step": 168989, "epoch": 1898, "lr": 3.132403267740016e-05}
{"train_loss": 0.09835504740476608, "global_step": 168990, "epoch": 1898, "lr": 3.132349490200215e-05}
{"train_loss": 0.12418022751808167, "global_step": 168991, "epoch": 1898, "lr": 3.1322957129114994e-05}
{"train_loss": 0.07110732793807983, "global_step": 168992, "epoch": 1898, "lr": 3.1322419358738716e-05}
{"train_loss": 0.0940147116780281, "global_step": 168993, "epoch": 1898, "lr": 3.132188159087343e-05}
{"train_loss": 0.11216895282268524, "global_step": 168994, "epoch": 1898, "lr": 3.132134382551917e-05}
{"train_loss": 0.14586225152015686, "global_step": 168995, "epoch": 1898, "lr": 3.132080606267603e-05}
{"train_loss": 0.09137433022260666, "global_step": 168996, "epoch": 1898, "lr": 3.132026830234409e-05}
{"train_loss": 0.09513430297374725, "global_step": 168997, "epoch": 1898, "lr": 3.1319730544523404e-05}
{"train_loss": 0.059152115136384964, "global_step": 168998, "epoch": 1898, "lr": 3.1319192789214066e-05}
{"train_loss": 0.05230437591671944, "global_step": 168999, "epoch": 1898, "lr": 3.131865503641612e-05}
{"train_loss": 0.05342863127589226, "global_step": 169000, "epoch": 1898, "lr": 3.1318117286129675e-05}
{"train_loss": 0.04493490606546402, "global_step": 169001, "epoch": 1898, "lr": 3.1317579538354764e-05}
{"train_loss": 0.08541112393140793, "global_step": 169002, "epoch": 1898, "lr": 3.131704179309148e-05}
{"train_loss": 0.07524808496236801, "global_step": 169003, "epoch": 1898, "lr": 3.131650405033989e-05}
{"train_loss": 0.13053922355175018, "global_step": 169004, "epoch": 1898, "lr": 3.131596631010007e-05}
{"train_loss": 0.09523004293441772, "global_step": 169005, "epoch": 1898, "lr": 3.131542857237209e-05}
{"train_loss": 0.18083231151103973, "global_step": 169006, "epoch": 1898, "lr": 3.131489083715602e-05}
{"train_loss": 0.08389764279127121, "global_step": 169007, "epoch": 1898, "lr": 3.131435310445194e-05}
{"train_loss": 0.11678480356931686, "global_step": 169008, "epoch": 1898, "lr": 3.131381537425993e-05}
{"train_loss": 0.0943431705236435, "global_step": 169009, "epoch": 1898, "lr": 3.131327764658003e-05}
{"train_loss": 0.08683385842301872, "global_step": 169010, "epoch": 1898, "lr": 3.131273992141235e-05, "val_loss": 7.263571262359619}
{"train_loss": 0.0523078516125679, "global_step": 169011, "epoch": 1899, "lr": 3.131220219875692e-05}
{"train_loss": 0.10644225031137466, "global_step": 169012, "epoch": 1899, "lr": 3.1311664478613855e-05}
{"train_loss": 0.1256949007511139, "global_step": 169013, "epoch": 1899, "lr": 3.131112676098319e-05}
{"train_loss": 0.12173008918762207, "global_step": 169014, "epoch": 1899, "lr": 3.1310589045865026e-05}
{"train_loss": 0.10268143564462662, "global_step": 169015, "epoch": 1899, "lr": 3.131005133325944e-05}
{"train_loss": 0.10824252665042877, "global_step": 169016, "epoch": 1899, "lr": 3.130951362316647e-05}
{"train_loss": 0.07902152091264725, "global_step": 169017, "epoch": 1899, "lr": 3.130897591558622e-05}
{"train_loss": 0.11927424371242523, "global_step": 169018, "epoch": 1899, "lr": 3.130843821051874e-05}
{"train_loss": 0.15694880485534668, "global_step": 169019, "epoch": 1899, "lr": 3.1307900507964126e-05}
{"train_loss": 0.09186610579490662, "global_step": 169020, "epoch": 1899, "lr": 3.130736280792243e-05}
{"train_loss": 0.06670920550823212, "global_step": 169021, "epoch": 1899, "lr": 3.1306825110393734e-05}
{"train_loss": 0.07059928774833679, "global_step": 169022, "epoch": 1899, "lr": 3.1306287415378106e-05}
{"train_loss": 0.09003084897994995, "global_step": 169023, "epoch": 1899, "lr": 3.130574972287562e-05}
{"train_loss": 0.07579044997692108, "global_step": 169024, "epoch": 1899, "lr": 3.1305212032886345e-05}
{"train_loss": 0.06727762520313263, "global_step": 169025, "epoch": 1899, "lr": 3.1304674345410353e-05}
{"train_loss": 0.09624651074409485, "global_step": 169026, "epoch": 1899, "lr": 3.130413666044773e-05}
{"train_loss": 0.0929766297340393, "global_step": 169027, "epoch": 1899, "lr": 3.130359897799854e-05}
{"train_loss": 0.04734499007463455, "global_step": 169028, "epoch": 1899, "lr": 3.130306129806284e-05}
{"train_loss": 0.050851479172706604, "global_step": 169029, "epoch": 1899, "lr": 3.1302523620640736e-05}
{"train_loss": 0.11805470287799835, "global_step": 169030, "epoch": 1899, "lr": 3.1301985945732254e-05}
{"train_loss": 0.08669358491897583, "global_step": 169031, "epoch": 1899, "lr": 3.1301448273337494e-05}
{"train_loss": 0.09024526923894882, "global_step": 169032, "epoch": 1899, "lr": 3.1300910603456546e-05}
{"train_loss": 0.08501099050045013, "global_step": 169033, "epoch": 1899, "lr": 3.130037293608944e-05}
{"train_loss": 0.07209000736474991, "global_step": 169034, "epoch": 1899, "lr": 3.12998352712363e-05}
{"train_loss": 0.059891145676374435, "global_step": 169035, "epoch": 1899, "lr": 3.129929760889714e-05}
{"train_loss": 0.04535490274429321, "global_step": 169036, "epoch": 1899, "lr": 3.1298759949072086e-05}
{"train_loss": 0.1475948691368103, "global_step": 169037, "epoch": 1899, "lr": 3.129822229176116e-05}
{"train_loss": 0.07390780746936798, "global_step": 169038, "epoch": 1899, "lr": 3.129768463696449e-05}
{"train_loss": 0.0998111143708229, "global_step": 169039, "epoch": 1899, "lr": 3.129714698468209e-05}
{"train_loss": 0.1200299859046936, "global_step": 169040, "epoch": 1899, "lr": 3.129660933491408e-05}
{"train_loss": 0.06275741755962372, "global_step": 169041, "epoch": 1899, "lr": 3.1296071687660514e-05}
{"train_loss": 0.10344161838293076, "global_step": 169042, "epoch": 1899, "lr": 3.129553404292145e-05}
{"train_loss": 0.08584226667881012, "global_step": 169043, "epoch": 1899, "lr": 3.129499640069698e-05}
{"train_loss": 0.11431559175252914, "global_step": 169044, "epoch": 1899, "lr": 3.129445876098717e-05}
{"train_loss": 0.08549028635025024, "global_step": 169045, "epoch": 1899, "lr": 3.129392112379209e-05}
{"train_loss": 0.04974491149187088, "global_step": 169046, "epoch": 1899, "lr": 3.1293383489111824e-05}
{"train_loss": 0.07140099257230759, "global_step": 169047, "epoch": 1899, "lr": 3.129284585694642e-05}
{"train_loss": 0.07634912431240082, "global_step": 169048, "epoch": 1899, "lr": 3.129230822729597e-05}
{"train_loss": 0.058332763612270355, "global_step": 169049, "epoch": 1899, "lr": 3.129177060016054e-05}
{"train_loss": 0.10354283452033997, "global_step": 169050, "epoch": 1899, "lr": 3.1291232975540195e-05}
{"train_loss": 0.07525839656591415, "global_step": 169051, "epoch": 1899, "lr": 3.1290695353435036e-05}
{"train_loss": 0.07698573172092438, "global_step": 169052, "epoch": 1899, "lr": 3.1290157733845096e-05}
{"train_loss": 0.10791978985071182, "global_step": 169053, "epoch": 1899, "lr": 3.128962011677048e-05}
{"train_loss": 0.07409112900495529, "global_step": 169054, "epoch": 1899, "lr": 3.128908250221123e-05}
{"train_loss": 0.10583741962909698, "global_step": 169055, "epoch": 1899, "lr": 3.128854489016746e-05}
{"train_loss": 0.147642120718956, "global_step": 169056, "epoch": 1899, "lr": 3.1288007280639195e-05}
{"train_loss": 0.10359952598810196, "global_step": 169057, "epoch": 1899, "lr": 3.128746967362654e-05}
{"train_loss": 0.055720530450344086, "global_step": 169058, "epoch": 1899, "lr": 3.128693206912955e-05}
{"train_loss": 0.10472966730594635, "global_step": 169059, "epoch": 1899, "lr": 3.1286394467148316e-05}
{"train_loss": 0.06933683902025223, "global_step": 169060, "epoch": 1899, "lr": 3.12858568676829e-05}
{"train_loss": 0.04507081210613251, "global_step": 169061, "epoch": 1899, "lr": 3.128531927073336e-05}
{"train_loss": 0.07106602191925049, "global_step": 169062, "epoch": 1899, "lr": 3.128478167629978e-05}
{"train_loss": 0.08776459097862244, "global_step": 169063, "epoch": 1899, "lr": 3.1284244084382244e-05}
{"train_loss": 0.08669252693653107, "global_step": 169064, "epoch": 1899, "lr": 3.12837064949808e-05}
{"train_loss": 0.09371736645698547, "global_step": 169065, "epoch": 1899, "lr": 3.128316890809555e-05}
{"train_loss": 0.0659291073679924, "global_step": 169066, "epoch": 1899, "lr": 3.128263132372653e-05}
{"train_loss": 0.1307060420513153, "global_step": 169067, "epoch": 1899, "lr": 3.128209374187385e-05}
{"train_loss": 0.04659900441765785, "global_step": 169068, "epoch": 1899, "lr": 3.128155616253755e-05}
{"train_loss": 0.12560057640075684, "global_step": 169069, "epoch": 1899, "lr": 3.128101858571771e-05}
{"train_loss": 0.07977259159088135, "global_step": 169070, "epoch": 1899, "lr": 3.1280481011414434e-05}
{"train_loss": 0.15760034322738647, "global_step": 169071, "epoch": 1899, "lr": 3.127994343962775e-05}
{"train_loss": 0.08312463760375977, "global_step": 169072, "epoch": 1899, "lr": 3.127940587035776e-05}
{"train_loss": 0.07408218085765839, "global_step": 169073, "epoch": 1899, "lr": 3.1278868303604515e-05}
{"train_loss": 0.06670378148555756, "global_step": 169074, "epoch": 1899, "lr": 3.127833073936811e-05}
{"train_loss": 0.11020264774560928, "global_step": 169075, "epoch": 1899, "lr": 3.127779317764859e-05}
{"train_loss": 0.08462178707122803, "global_step": 169076, "epoch": 1899, "lr": 3.1277255618446066e-05}
{"train_loss": 0.06009163707494736, "global_step": 169077, "epoch": 1899, "lr": 3.1276718061760554e-05}
{"train_loss": 0.040452588349580765, "global_step": 169078, "epoch": 1899, "lr": 3.127618050759219e-05}
{"train_loss": 0.11971147358417511, "global_step": 169079, "epoch": 1899, "lr": 3.127564295594101e-05}
{"train_loss": 0.07354853302240372, "global_step": 169080, "epoch": 1899, "lr": 3.1275105406807084e-05}
{"train_loss": 0.11276944726705551, "global_step": 169081, "epoch": 1899, "lr": 3.127456786019049e-05}
{"train_loss": 0.08459194004535675, "global_step": 169082, "epoch": 1899, "lr": 3.1274030316091304e-05}
{"train_loss": 0.0622854121029377, "global_step": 169083, "epoch": 1899, "lr": 3.12734927745096e-05}
{"train_loss": 0.08932404965162277, "global_step": 169084, "epoch": 1899, "lr": 3.127295523544545e-05}
{"train_loss": 0.10731853544712067, "global_step": 169085, "epoch": 1899, "lr": 3.1272417698898915e-05}
{"train_loss": 0.06347920745611191, "global_step": 169086, "epoch": 1899, "lr": 3.127188016487006e-05}
{"train_loss": 0.04228054732084274, "global_step": 169087, "epoch": 1899, "lr": 3.1271342633359e-05}
{"train_loss": 0.05324094370007515, "global_step": 169088, "epoch": 1899, "lr": 3.127080510436576e-05}
{"train_loss": 0.09791348874568939, "global_step": 169089, "epoch": 1899, "lr": 3.127026757789046e-05}
{"train_loss": 0.08800557255744934, "global_step": 169090, "epoch": 1899, "lr": 3.1269730053933114e-05}
{"train_loss": 0.05822769179940224, "global_step": 169091, "epoch": 1899, "lr": 3.126919253249384e-05}
{"train_loss": 0.07669658213853836, "global_step": 169092, "epoch": 1899, "lr": 3.126865501357268e-05}
{"train_loss": 0.11171062290668488, "global_step": 169093, "epoch": 1899, "lr": 3.126811749716975e-05}
{"train_loss": 0.09704841673374176, "global_step": 169094, "epoch": 1899, "lr": 3.1267579983285056e-05}
{"train_loss": 0.09018310904502869, "global_step": 169095, "epoch": 1899, "lr": 3.126704247191874e-05}
{"train_loss": 0.08852404356002808, "global_step": 169096, "epoch": 1899, "lr": 3.126650496307082e-05}
{"train_loss": 0.07360532879829407, "global_step": 169097, "epoch": 1899, "lr": 3.126596745674141e-05}
{"train_loss": 0.05305119603872299, "global_step": 169098, "epoch": 1899, "lr": 3.1265429952930544e-05}
{"train_loss": 0.08639211123913862, "global_step": 169099, "epoch": 1899, "lr": 3.126489245163833e-05, "val_loss": 7.376956462860107}
{"train_loss": 0.0856667309999466, "global_step": 169100, "epoch": 1900, "lr": 3.126435495286481e-05}
{"train_loss": 0.0595066100358963, "global_step": 169101, "epoch": 1900, "lr": 3.126381745661007e-05}
{"train_loss": 0.08513541519641876, "global_step": 169102, "epoch": 1900, "lr": 3.126327996287418e-05}
{"train_loss": 0.1281001716852188, "global_step": 169103, "epoch": 1900, "lr": 3.126274247165722e-05}
{"train_loss": 0.060369085520505905, "global_step": 169104, "epoch": 1900, "lr": 3.1262204982959256e-05}
{"train_loss": 0.06665664166212082, "global_step": 169105, "epoch": 1900, "lr": 3.126166749678035e-05}
{"train_loss": 0.06043680012226105, "global_step": 169106, "epoch": 1900, "lr": 3.1261130013120595e-05}
{"train_loss": 0.07606475055217743, "global_step": 169107, "epoch": 1900, "lr": 3.126059253198004e-05}
{"train_loss": 0.07012083381414413, "global_step": 169108, "epoch": 1900, "lr": 3.1260055053358786e-05}
{"train_loss": 0.15473006665706635, "global_step": 169109, "epoch": 1900, "lr": 3.125951757725687e-05}
{"train_loss": 0.061551716178655624, "global_step": 169110, "epoch": 1900, "lr": 3.1258980103674405e-05}
{"train_loss": 0.0749969333410263, "global_step": 169111, "epoch": 1900, "lr": 3.125844263261142e-05}
{"train_loss": 0.08995966613292694, "global_step": 169112, "epoch": 1900, "lr": 3.125790516406804e-05}
{"train_loss": 0.08395011723041534, "global_step": 169113, "epoch": 1900, "lr": 3.125736769804427e-05}
{"train_loss": 0.09112267196178436, "global_step": 169114, "epoch": 1900, "lr": 3.125683023454025e-05}
{"train_loss": 0.10268307477235794, "global_step": 169115, "epoch": 1900, "lr": 3.1256292773555996e-05}
{"train_loss": 0.06090359762310982, "global_step": 169116, "epoch": 1900, "lr": 3.1255755315091625e-05}
{"train_loss": 0.06345058232545853, "global_step": 169117, "epoch": 1900, "lr": 3.125521785914718e-05}
{"train_loss": 0.04110393673181534, "global_step": 169118, "epoch": 1900, "lr": 3.125468040572273e-05}
{"train_loss": 0.08521624654531479, "global_step": 169119, "epoch": 1900, "lr": 3.1254142954818375e-05}
{"train_loss": 0.12162484228610992, "global_step": 169120, "epoch": 1900, "lr": 3.125360550643418e-05}
{"train_loss": 0.10365109145641327, "global_step": 169121, "epoch": 1900, "lr": 3.125306806057019e-05}
{"train_loss": 0.09475313127040863, "global_step": 169122, "epoch": 1900, "lr": 3.125253061722649e-05}
{"train_loss": 0.06542467325925827, "global_step": 169123, "epoch": 1900, "lr": 3.1251993176403196e-05}
{"train_loss": 0.05736500769853592, "global_step": 169124, "epoch": 1900, "lr": 3.12514557381003e-05}
{"train_loss": 0.08325806260108948, "global_step": 169125, "epoch": 1900, "lr": 3.1250918302317956e-05}
{"train_loss": 0.07941033691167831, "global_step": 169126, "epoch": 1900, "lr": 3.125038086905617e-05}
{"train_loss": 0.12072587013244629, "global_step": 169127, "epoch": 1900, "lr": 3.124984343831506e-05}
{"train_loss": 0.08665606379508972, "global_step": 169128, "epoch": 1900, "lr": 3.1249306010094656e-05}
{"train_loss": 0.1287638247013092, "global_step": 169129, "epoch": 1900, "lr": 3.1248768584395085e-05}
{"train_loss": 0.11747437715530396, "global_step": 169130, "epoch": 1900, "lr": 3.124823116121636e-05}
{"train_loss": 0.04990279674530029, "global_step": 169131, "epoch": 1900, "lr": 3.1247693740558605e-05}
{"train_loss": 0.07595732808113098, "global_step": 169132, "epoch": 1900, "lr": 3.1247156322421845e-05}
{"train_loss": 0.14166437089443207, "global_step": 169133, "epoch": 1900, "lr": 3.12466189068062e-05}
{"train_loss": 0.10332021862268448, "global_step": 169134, "epoch": 1900, "lr": 3.12460814937117e-05}
{"train_loss": 0.07080666720867157, "global_step": 169135, "epoch": 1900, "lr": 3.124554408313846e-05}
{"train_loss": 0.07035045325756073, "global_step": 169136, "epoch": 1900, "lr": 3.12450066750865e-05}
{"train_loss": 0.08811372518539429, "global_step": 169137, "epoch": 1900, "lr": 3.124446926955594e-05}
{"train_loss": 0.06971649080514908, "global_step": 169138, "epoch": 1900, "lr": 3.124393186654683e-05}
{"train_loss": 0.09185072034597397, "global_step": 169139, "epoch": 1900, "lr": 3.1243394466059235e-05}
{"train_loss": 0.0803581178188324, "global_step": 169140, "epoch": 1900, "lr": 3.124285706809324e-05}
{"train_loss": 0.10586678236722946, "global_step": 169141, "epoch": 1900, "lr": 3.1242319672648904e-05}
{"train_loss": 0.07145009934902191, "global_step": 169142, "epoch": 1900, "lr": 3.1241782279726336e-05}
{"train_loss": 0.08905793726444244, "global_step": 169143, "epoch": 1900, "lr": 3.1241244889325555e-05}
{"train_loss": 0.05872678384184837, "global_step": 169144, "epoch": 1900, "lr": 3.124070750144668e-05}
{"train_loss": 0.05091860890388489, "global_step": 169145, "epoch": 1900, "lr": 3.124017011608975e-05}
{"train_loss": 0.08182036876678467, "global_step": 169146, "epoch": 1900, "lr": 3.1239632733254866e-05}
{"train_loss": 0.06565380096435547, "global_step": 169147, "epoch": 1900, "lr": 3.123909535294206e-05}
{"train_loss": 0.12374265491962433, "global_step": 169148, "epoch": 1900, "lr": 3.1238557975151456e-05}
{"train_loss": 0.06137583777308464, "global_step": 169149, "epoch": 1900, "lr": 3.1238020599883075e-05}
{"train_loss": 0.07346438616514206, "global_step": 169150, "epoch": 1900, "lr": 3.1237483227137035e-05}
{"train_loss": 0.10558675229549408, "global_step": 169151, "epoch": 1900, "lr": 3.123694585691337e-05}
{"train_loss": 0.0747385248541832, "global_step": 169152, "epoch": 1900, "lr": 3.123640848921218e-05}
{"train_loss": 0.07898177206516266, "global_step": 169153, "epoch": 1900, "lr": 3.123587112403351e-05}
{"train_loss": 0.11609029769897461, "global_step": 169154, "epoch": 1900, "lr": 3.123533376137747e-05}
{"train_loss": 0.06875432282686234, "global_step": 169155, "epoch": 1900, "lr": 3.123479640124409e-05}
{"train_loss": 0.07142149657011032, "global_step": 169156, "epoch": 1900, "lr": 3.123425904363349e-05}
{"train_loss": 0.09788523614406586, "global_step": 169157, "epoch": 1900, "lr": 3.12337216885457e-05}
{"train_loss": 0.10808145999908447, "global_step": 169158, "epoch": 1900, "lr": 3.1233184335980796e-05}
{"train_loss": 0.08826850354671478, "global_step": 169159, "epoch": 1900, "lr": 3.1232646985938885e-05}
{"train_loss": 0.08977969735860825, "global_step": 169160, "epoch": 1900, "lr": 3.1232109638419994e-05}
{"train_loss": 0.14379073679447174, "global_step": 169161, "epoch": 1900, "lr": 3.123157229342423e-05}
{"train_loss": 0.07293964177370071, "global_step": 169162, "epoch": 1900, "lr": 3.1231034950951644e-05}
{"train_loss": 0.10321438312530518, "global_step": 169163, "epoch": 1900, "lr": 3.123049761100233e-05}
{"train_loss": 0.108650803565979, "global_step": 169164, "epoch": 1900, "lr": 3.122996027357632e-05}
{"train_loss": 0.08774672448635101, "global_step": 169165, "epoch": 1900, "lr": 3.122942293867375e-05}
{"train_loss": 0.04061771184206009, "global_step": 169166, "epoch": 1900, "lr": 3.1228885606294624e-05}
{"train_loss": 0.03276490420103073, "global_step": 169167, "epoch": 1900, "lr": 3.122834827643907e-05}
{"train_loss": 0.06931172311306, "global_step": 169168, "epoch": 1900, "lr": 3.122781094910711e-05}
{"train_loss": 0.06270795315504074, "global_step": 169169, "epoch": 1900, "lr": 3.122727362429887e-05}
{"train_loss": 0.1486378312110901, "global_step": 169170, "epoch": 1900, "lr": 3.122673630201437e-05}
{"train_loss": 0.06415368616580963, "global_step": 169171, "epoch": 1900, "lr": 3.1226198982253725e-05}
{"train_loss": 0.09772179275751114, "global_step": 169172, "epoch": 1900, "lr": 3.122566166501697e-05}
{"train_loss": 0.057749222964048386, "global_step": 169173, "epoch": 1900, "lr": 3.122512435030422e-05}
{"train_loss": 0.08684495091438293, "global_step": 169174, "epoch": 1900, "lr": 3.12245870381155e-05}
{"train_loss": 0.06347993016242981, "global_step": 169175, "epoch": 1900, "lr": 3.122404972845092e-05}
{"train_loss": 0.11730071157217026, "global_step": 169176, "epoch": 1900, "lr": 3.122351242131053e-05}
{"train_loss": 0.060155075043439865, "global_step": 169177, "epoch": 1900, "lr": 3.12229751166944e-05}
{"train_loss": 0.08802391588687897, "global_step": 169178, "epoch": 1900, "lr": 3.122243781460263e-05}
{"train_loss": 0.07981322705745697, "global_step": 169179, "epoch": 1900, "lr": 3.122190051503526e-05}
{"train_loss": 0.07865265011787415, "global_step": 169180, "epoch": 1900, "lr": 3.12213632179924e-05}
{"train_loss": 0.10475108027458191, "global_step": 169181, "epoch": 1900, "lr": 3.122082592347407e-05}
{"train_loss": 0.09922163188457489, "global_step": 169182, "epoch": 1900, "lr": 3.12202886314804e-05}
{"train_loss": 0.09389366954565048, "global_step": 169183, "epoch": 1900, "lr": 3.121975134201141e-05}
{"train_loss": 0.11316613107919693, "global_step": 169184, "epoch": 1900, "lr": 3.121921405506722e-05}
{"train_loss": 0.10852626711130142, "global_step": 169185, "epoch": 1900, "lr": 3.121867677064785e-05}
{"train_loss": 0.07626963406801224, "global_step": 169186, "epoch": 1900, "lr": 3.121813948875342e-05}
{"train_loss": 0.07523039728403091, "global_step": 169187, "epoch": 1900, "lr": 3.121760220938396e-05}
{"train_loss": 0.08552167803216516, "global_step": 169188, "epoch": 1900, "lr": 3.1217064932539586e-05, "train/sim_max_reward_0": 0.6560683243046848, "train/sim_max_reward_1": 0.9714806534765513, "train/sim_max_reward_2": 0.9973160429674732, "train/sim_max_reward_3": 0.7203495191412668, "train/sim_max_reward_4": 0.9971893850847616, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.5098125077970224, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.5456285802654659, "test/sim_max_reward_4500004": 0.5456430554653924, "test/sim_max_reward_4500005": 0.4141290590186184, "test/sim_max_reward_4500006": 0.9113756497878739, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.15627730935052675, "test/sim_max_reward_4500009": 0.051179531668772704, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5239348622171851, "test/sim_max_reward_4500014": 0.9514456267792135, "test/sim_max_reward_4500015": 0.9311228023089088, "test/sim_max_reward_4500016": 0.984230662859383, "test/sim_max_reward_4500017": 0.14596138275427456, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.4131811299962409, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.8892423473920047, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.511133445844703, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.15341720422226657, "test/sim_max_reward_4500026": 0.9932637843814577, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6666963206002368, "test/sim_max_reward_4500029": 0.06094108118478406, "test/sim_max_reward_4500030": 0.8535931959554207, "test/sim_max_reward_4500031": 0.5052765711588503, "test/sim_max_reward_4500032": 0.6948806080616959, "test/sim_max_reward_4500033": 0.9865430306634498, "test/sim_max_reward_4500034": 0.4945954377205839, "test/sim_max_reward_4500035": 0.33084514300640283, "test/sim_max_reward_4500036": 0.09534004210944988, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.16275429925084317, "test/sim_max_reward_4500039": 0.9775246015364619, "test/sim_max_reward_4500040": 0.9966619555481705, "test/sim_max_reward_4500041": 0.9676138594987764, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.8865209823228123, "test/sim_max_reward_4500046": 0.4663214584601816, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9073320418175064, "test/sim_max_reward_4500049": 0.9935320911994336, "train/mean_score": 0.8904006541624563, "test/mean_score": 0.5735590332440875, "val_loss": 7.162790775299072, "train_action_mse_error": 25.131717681884766}
{"train_loss": 0.07482900470495224, "global_step": 169189, "epoch": 1901, "lr": 3.1216527658220334e-05}
{"train_loss": 0.05765087902545929, "global_step": 169190, "epoch": 1901, "lr": 3.121599038642631e-05}
{"train_loss": 0.1094091385602951, "global_step": 169191, "epoch": 1901, "lr": 3.121545311715754e-05}
{"train_loss": 0.08222349733114243, "global_step": 169192, "epoch": 1901, "lr": 3.121491585041415e-05}
{"train_loss": 0.07878203690052032, "global_step": 169193, "epoch": 1901, "lr": 3.121437858619617e-05}
{"train_loss": 0.062265440821647644, "global_step": 169194, "epoch": 1901, "lr": 3.1213841324503704e-05}
{"train_loss": 0.13267114758491516, "global_step": 169195, "epoch": 1901, "lr": 3.1213304065336793e-05}
{"train_loss": 0.08551442623138428, "global_step": 169196, "epoch": 1901, "lr": 3.121276680869551e-05}
{"train_loss": 0.074569471180439, "global_step": 169197, "epoch": 1901, "lr": 3.1212229554579974e-05}
{"train_loss": 0.09534070640802383, "global_step": 169198, "epoch": 1901, "lr": 3.12116923029902e-05}
{"train_loss": 0.06435449421405792, "global_step": 169199, "epoch": 1901, "lr": 3.1211155053926306e-05}
{"train_loss": 0.1028905138373375, "global_step": 169200, "epoch": 1901, "lr": 3.121061780738832e-05}
{"train_loss": 0.07667893171310425, "global_step": 169201, "epoch": 1901, "lr": 3.121008056337636e-05}
{"train_loss": 0.05271738022565842, "global_step": 169202, "epoch": 1901, "lr": 3.1209543321890454e-05}
{"train_loss": 0.11646635085344315, "global_step": 169203, "epoch": 1901, "lr": 3.1209006082930714e-05}
{"train_loss": 0.11181733012199402, "global_step": 169204, "epoch": 1901, "lr": 3.120846884649718e-05}
{"train_loss": 0.08925285190343857, "global_step": 169205, "epoch": 1901, "lr": 3.1207931612589956e-05}
{"train_loss": 0.044458694756031036, "global_step": 169206, "epoch": 1901, "lr": 3.1207394381209074e-05}
{"train_loss": 0.1068933829665184, "global_step": 169207, "epoch": 1901, "lr": 3.1206857152354654e-05}
{"train_loss": 0.11175372451543808, "global_step": 169208, "epoch": 1901, "lr": 3.1206319926026716e-05}
{"train_loss": 0.07256222516298294, "global_step": 169209, "epoch": 1901, "lr": 3.120578270222538e-05}
{"train_loss": 0.07029953598976135, "global_step": 169210, "epoch": 1901, "lr": 3.120524548095068e-05}
{"train_loss": 0.13060563802719116, "global_step": 169211, "epoch": 1901, "lr": 3.120470826220273e-05}
{"train_loss": 0.09561646729707718, "global_step": 169212, "epoch": 1901, "lr": 3.120417104598156e-05}
{"train_loss": 0.10194867104291916, "global_step": 169213, "epoch": 1901, "lr": 3.120363383228726e-05}
{"train_loss": 0.1398271918296814, "global_step": 169214, "epoch": 1901, "lr": 3.120309662111992e-05}
{"train_loss": 0.08974029123783112, "global_step": 169215, "epoch": 1901, "lr": 3.1202559412479566e-05}
{"train_loss": 0.07674433290958405, "global_step": 169216, "epoch": 1901, "lr": 3.120202220636632e-05}
{"train_loss": 0.08015955984592438, "global_step": 169217, "epoch": 1901, "lr": 3.1201485002780215e-05}
{"train_loss": 0.08806207776069641, "global_step": 169218, "epoch": 1901, "lr": 3.120094780172137e-05}
{"train_loss": 0.0746360719203949, "global_step": 169219, "epoch": 1901, "lr": 3.12004106031898e-05}
{"train_loss": 0.09697320312261581, "global_step": 169220, "epoch": 1901, "lr": 3.119987340718562e-05}
{"train_loss": 0.1253558248281479, "global_step": 169221, "epoch": 1901, "lr": 3.119933621370888e-05}
{"train_loss": 0.05406144633889198, "global_step": 169222, "epoch": 1901, "lr": 3.1198799022759674e-05}
{"train_loss": 0.11811498552560806, "global_step": 169223, "epoch": 1901, "lr": 3.119826183433805e-05}
{"train_loss": 0.08135315775871277, "global_step": 169224, "epoch": 1901, "lr": 3.1197724648444094e-05}
{"train_loss": 0.08789820969104767, "global_step": 169225, "epoch": 1901, "lr": 3.119718746507786e-05}
{"train_loss": 0.10943805426359177, "global_step": 169226, "epoch": 1901, "lr": 3.119665028423946e-05}
{"train_loss": 0.05335023254156113, "global_step": 169227, "epoch": 1901, "lr": 3.119611310592892e-05}
{"train_loss": 0.08689774572849274, "global_step": 169228, "epoch": 1901, "lr": 3.119557593014635e-05}
{"train_loss": 0.06810164451599121, "global_step": 169229, "epoch": 1901, "lr": 3.1195038756891795e-05}
{"train_loss": 0.11590704321861267, "global_step": 169230, "epoch": 1901, "lr": 3.119450158616535e-05}
{"train_loss": 0.086203433573246, "global_step": 169231, "epoch": 1901, "lr": 3.119396441796705e-05}
{"train_loss": 0.10367466509342194, "global_step": 169232, "epoch": 1901, "lr": 3.119342725229701e-05}
{"train_loss": 0.0896206796169281, "global_step": 169233, "epoch": 1901, "lr": 3.119289008915529e-05}
{"train_loss": 0.06930182874202728, "global_step": 169234, "epoch": 1901, "lr": 3.1192352928541944e-05}
{"train_loss": 0.08996149152517319, "global_step": 169235, "epoch": 1901, "lr": 3.1191815770457065e-05}
{"train_loss": 0.0929337814450264, "global_step": 169236, "epoch": 1901, "lr": 3.119127861490071e-05}
{"train_loss": 0.07348184287548065, "global_step": 169237, "epoch": 1901, "lr": 3.119074146187297e-05}
{"train_loss": 0.05509977415204048, "global_step": 169238, "epoch": 1901, "lr": 3.119020431137389e-05}
{"train_loss": 0.10516093671321869, "global_step": 169239, "epoch": 1901, "lr": 3.118966716340357e-05}
{"train_loss": 0.12327760457992554, "global_step": 169240, "epoch": 1901, "lr": 3.118913001796206e-05}
{"train_loss": 0.05864943563938141, "global_step": 169241, "epoch": 1901, "lr": 3.118859287504946e-05}
{"train_loss": 0.10190754383802414, "global_step": 169242, "epoch": 1901, "lr": 3.11880557346658e-05}
{"train_loss": 0.1109476387500763, "global_step": 169243, "epoch": 1901, "lr": 3.11875185968112e-05}
{"train_loss": 0.1296044886112213, "global_step": 169244, "epoch": 1901, "lr": 3.118698146148569e-05}
{"train_loss": 0.0932517871260643, "global_step": 169245, "epoch": 1901, "lr": 3.118644432868937e-05}
{"train_loss": 0.1081138476729393, "global_step": 169246, "epoch": 1901, "lr": 3.1185907198422294e-05}
{"train_loss": 0.06488662958145142, "global_step": 169247, "epoch": 1901, "lr": 3.118537007068456e-05}
{"train_loss": 0.08665142208337784, "global_step": 169248, "epoch": 1901, "lr": 3.118483294547621e-05}
{"train_loss": 0.04810024052858353, "global_step": 169249, "epoch": 1901, "lr": 3.118429582279734e-05}
{"train_loss": 0.10524120926856995, "global_step": 169250, "epoch": 1901, "lr": 3.1183758702648005e-05}
{"train_loss": 0.11364992707967758, "global_step": 169251, "epoch": 1901, "lr": 3.118322158502828e-05}
{"train_loss": 0.05022037401795387, "global_step": 169252, "epoch": 1901, "lr": 3.118268446993825e-05}
{"train_loss": 0.10958091914653778, "global_step": 169253, "epoch": 1901, "lr": 3.1182147357377975e-05}
{"train_loss": 0.10924261808395386, "global_step": 169254, "epoch": 1901, "lr": 3.1181610247347546e-05}
{"train_loss": 0.10470303893089294, "global_step": 169255, "epoch": 1901, "lr": 3.1181073139846994e-05}
{"train_loss": 0.06998573243618011, "global_step": 169256, "epoch": 1901, "lr": 3.118053603487644e-05}
{"train_loss": 0.08234906941652298, "global_step": 169257, "epoch": 1901, "lr": 3.1179998932435916e-05}
{"train_loss": 0.11688780784606934, "global_step": 169258, "epoch": 1901, "lr": 3.117946183252553e-05}
{"train_loss": 0.12334153056144714, "global_step": 169259, "epoch": 1901, "lr": 3.1178924735145316e-05}
{"train_loss": 0.13908244669437408, "global_step": 169260, "epoch": 1901, "lr": 3.1178387640295384e-05}
{"train_loss": 0.06493088603019714, "global_step": 169261, "epoch": 1901, "lr": 3.1177850547975776e-05}
{"train_loss": 0.038669146597385406, "global_step": 169262, "epoch": 1901, "lr": 3.117731345818659e-05}
{"train_loss": 0.07507745176553726, "global_step": 169263, "epoch": 1901, "lr": 3.117677637092786e-05}
{"train_loss": 0.09500957280397415, "global_step": 169264, "epoch": 1901, "lr": 3.117623928619971e-05}
{"train_loss": 0.09793047606945038, "global_step": 169265, "epoch": 1901, "lr": 3.1175702204002165e-05}
{"train_loss": 0.06790942698717117, "global_step": 169266, "epoch": 1901, "lr": 3.117516512433534e-05}
{"train_loss": 0.07967078685760498, "global_step": 169267, "epoch": 1901, "lr": 3.117462804719926e-05}
{"train_loss": 0.1008649468421936, "global_step": 169268, "epoch": 1901, "lr": 3.117409097259403e-05}
{"train_loss": 0.07157155871391296, "global_step": 169269, "epoch": 1901, "lr": 3.117355390051972e-05}
{"train_loss": 0.08682047575712204, "global_step": 169270, "epoch": 1901, "lr": 3.1173016830976386e-05}
{"train_loss": 0.06259773671627045, "global_step": 169271, "epoch": 1901, "lr": 3.117247976396412e-05}
{"train_loss": 0.0777185782790184, "global_step": 169272, "epoch": 1901, "lr": 3.117194269948298e-05}
{"train_loss": 0.1001693531870842, "global_step": 169273, "epoch": 1901, "lr": 3.117140563753306e-05}
{"train_loss": 0.07767901569604874, "global_step": 169274, "epoch": 1901, "lr": 3.117086857811439e-05}
{"train_loss": 0.08870761841535568, "global_step": 169275, "epoch": 1901, "lr": 3.117033152122708e-05}
{"train_loss": 0.10037089884281158, "global_step": 169276, "epoch": 1901, "lr": 3.1169794466871175e-05}
{"train_loss": 0.08918033201205597, "global_step": 169277, "epoch": 1901, "lr": 3.1169257415046783e-05, "val_loss": 6.939677715301514}
{"train_loss": 0.06072651967406273, "global_step": 169278, "epoch": 1902, "lr": 3.1168720365753936e-05}
{"train_loss": 0.04979837313294411, "global_step": 169279, "epoch": 1902, "lr": 3.116818331899274e-05}
{"train_loss": 0.13051243126392365, "global_step": 169280, "epoch": 1902, "lr": 3.116764627476324e-05}
{"train_loss": 0.09972530603408813, "global_step": 169281, "epoch": 1902, "lr": 3.116710923306554e-05}
{"train_loss": 0.06692562252283096, "global_step": 169282, "epoch": 1902, "lr": 3.116657219389967e-05}
{"train_loss": 0.08862046152353287, "global_step": 169283, "epoch": 1902, "lr": 3.1166035157265736e-05}
{"train_loss": 0.06584316492080688, "global_step": 169284, "epoch": 1902, "lr": 3.116549812316379e-05}
{"train_loss": 0.10940717905759811, "global_step": 169285, "epoch": 1902, "lr": 3.1164961091593905e-05}
{"train_loss": 0.07871770113706589, "global_step": 169286, "epoch": 1902, "lr": 3.116442406255619e-05}
{"train_loss": 0.09357728064060211, "global_step": 169287, "epoch": 1902, "lr": 3.116388703605068e-05}
{"train_loss": 0.07398243248462677, "global_step": 169288, "epoch": 1902, "lr": 3.116335001207745e-05}
{"train_loss": 0.07568833976984024, "global_step": 169289, "epoch": 1902, "lr": 3.1162812990636574e-05}
{"train_loss": 0.0908508375287056, "global_step": 169290, "epoch": 1902, "lr": 3.116227597172813e-05}
{"train_loss": 0.0905921682715416, "global_step": 169291, "epoch": 1902, "lr": 3.116173895535219e-05}
{"train_loss": 0.060359273105859756, "global_step": 169292, "epoch": 1902, "lr": 3.1161201941508846e-05}
{"train_loss": 0.1466178297996521, "global_step": 169293, "epoch": 1902, "lr": 3.1160664930198114e-05}
{"train_loss": 0.16425779461860657, "global_step": 169294, "epoch": 1902, "lr": 3.116012792142013e-05}
{"train_loss": 0.07066833972930908, "global_step": 169295, "epoch": 1902, "lr": 3.115959091517491e-05}
{"train_loss": 0.07165594398975372, "global_step": 169296, "epoch": 1902, "lr": 3.115905391146258e-05}
{"train_loss": 0.09031800180673599, "global_step": 169297, "epoch": 1902, "lr": 3.1158516910283154e-05}
{"train_loss": 0.13471968472003937, "global_step": 169298, "epoch": 1902, "lr": 3.115797991163677e-05}
{"train_loss": 0.1051456406712532, "global_step": 169299, "epoch": 1902, "lr": 3.115744291552344e-05}
{"train_loss": 0.12223712354898453, "global_step": 169300, "epoch": 1902, "lr": 3.115690592194328e-05}
{"train_loss": 0.09168093651533127, "global_step": 169301, "epoch": 1902, "lr": 3.115636893089632e-05}
{"train_loss": 0.08351583778858185, "global_step": 169302, "epoch": 1902, "lr": 3.115583194238268e-05}
{"train_loss": 0.05374691262841225, "global_step": 169303, "epoch": 1902, "lr": 3.11552949564024e-05}
{"train_loss": 0.12808923423290253, "global_step": 169304, "epoch": 1902, "lr": 3.115475797295555e-05}
{"train_loss": 0.09643003344535828, "global_step": 169305, "epoch": 1902, "lr": 3.115422099204224e-05}
{"train_loss": 0.05724383890628815, "global_step": 169306, "epoch": 1902, "lr": 3.11536840136625e-05}
{"train_loss": 0.050982095301151276, "global_step": 169307, "epoch": 1902, "lr": 3.115314703781641e-05}
{"train_loss": 0.04504901543259621, "global_step": 169308, "epoch": 1902, "lr": 3.1152610064504064e-05}
{"train_loss": 0.04995964840054512, "global_step": 169309, "epoch": 1902, "lr": 3.1152073093725506e-05}
{"train_loss": 0.10769139975309372, "global_step": 169310, "epoch": 1902, "lr": 3.1151536125480836e-05}
{"train_loss": 0.1595100462436676, "global_step": 169311, "epoch": 1902, "lr": 3.115099915977011e-05}
{"train_loss": 0.05105471611022949, "global_step": 169312, "epoch": 1902, "lr": 3.1150462196593395e-05}
{"train_loss": 0.11072596907615662, "global_step": 169313, "epoch": 1902, "lr": 3.1149925235950785e-05}
{"train_loss": 0.14110010862350464, "global_step": 169314, "epoch": 1902, "lr": 3.1149388277842315e-05}
{"train_loss": 0.08238057792186737, "global_step": 169315, "epoch": 1902, "lr": 3.11488513222681e-05}
{"train_loss": 0.08575451374053955, "global_step": 169316, "epoch": 1902, "lr": 3.114831436922817e-05}
{"train_loss": 0.052834540605545044, "global_step": 169317, "epoch": 1902, "lr": 3.1147777418722644e-05}
{"train_loss": 0.062455110251903534, "global_step": 169318, "epoch": 1902, "lr": 3.1147240470751545e-05}
{"train_loss": 0.05851743370294571, "global_step": 169319, "epoch": 1902, "lr": 3.1146703525315e-05}
{"train_loss": 0.14566463232040405, "global_step": 169320, "epoch": 1902, "lr": 3.114616658241303e-05}
{"train_loss": 0.11377186328172684, "global_step": 169321, "epoch": 1902, "lr": 3.114562964204574e-05}
{"train_loss": 0.04317086935043335, "global_step": 169322, "epoch": 1902, "lr": 3.1145092704213176e-05}
{"train_loss": 0.1231222003698349, "global_step": 169323, "epoch": 1902, "lr": 3.114455576891542e-05}
{"train_loss": 0.14180290699005127, "global_step": 169324, "epoch": 1902, "lr": 3.1144018836152566e-05}
{"train_loss": 0.09263366460800171, "global_step": 169325, "epoch": 1902, "lr": 3.114348190592467e-05}
{"train_loss": 0.08786236494779587, "global_step": 169326, "epoch": 1902, "lr": 3.1142944978231795e-05}
{"train_loss": 0.07083874195814133, "global_step": 169327, "epoch": 1902, "lr": 3.114240805307401e-05}
{"train_loss": 0.0787479355931282, "global_step": 169328, "epoch": 1902, "lr": 3.114187113045142e-05}
{"train_loss": 0.07592637836933136, "global_step": 169329, "epoch": 1902, "lr": 3.114133421036406e-05}
{"train_loss": 0.0932171419262886, "global_step": 169330, "epoch": 1902, "lr": 3.114079729281204e-05}
{"train_loss": 0.05262899398803711, "global_step": 169331, "epoch": 1902, "lr": 3.1140260377795385e-05}
{"train_loss": 0.09562116861343384, "global_step": 169332, "epoch": 1902, "lr": 3.113972346531421e-05}
{"train_loss": 0.07419556379318237, "global_step": 169333, "epoch": 1902, "lr": 3.1139186555368546e-05}
{"train_loss": 0.05815116688609123, "global_step": 169334, "epoch": 1902, "lr": 3.1138649647958515e-05}
{"train_loss": 0.07386177033185959, "global_step": 169335, "epoch": 1902, "lr": 3.1138112743084134e-05}
{"train_loss": 0.06858876347541809, "global_step": 169336, "epoch": 1902, "lr": 3.113757584074553e-05}
{"train_loss": 0.08764277398586273, "global_step": 169337, "epoch": 1902, "lr": 3.113703894094273e-05}
{"train_loss": 0.11178899556398392, "global_step": 169338, "epoch": 1902, "lr": 3.113650204367585e-05}
{"train_loss": 0.10264943540096283, "global_step": 169339, "epoch": 1902, "lr": 3.1135965148944915e-05}
{"train_loss": 0.07570993155241013, "global_step": 169340, "epoch": 1902, "lr": 3.113542825675001e-05}
{"train_loss": 0.057790085673332214, "global_step": 169341, "epoch": 1902, "lr": 3.1134891367091246e-05}
{"train_loss": 0.07920504361391068, "global_step": 169342, "epoch": 1902, "lr": 3.113435447996864e-05}
{"train_loss": 0.12761786580085754, "global_step": 169343, "epoch": 1902, "lr": 3.1133817595382315e-05}
{"train_loss": 0.07103323191404343, "global_step": 169344, "epoch": 1902, "lr": 3.11332807133323e-05}
{"train_loss": 0.07790070027112961, "global_step": 169345, "epoch": 1902, "lr": 3.11327438338187e-05}
{"train_loss": 0.07499716430902481, "global_step": 169346, "epoch": 1902, "lr": 3.1132206956841556e-05}
{"train_loss": 0.0777752697467804, "global_step": 169347, "epoch": 1902, "lr": 3.113167008240097e-05}
{"train_loss": 0.08331874012947083, "global_step": 169348, "epoch": 1902, "lr": 3.1131133210497e-05}
{"train_loss": 0.0577312633395195, "global_step": 169349, "epoch": 1902, "lr": 3.1130596341129725e-05}
{"train_loss": 0.12612420320510864, "global_step": 169350, "epoch": 1902, "lr": 3.113005947429919e-05}
{"train_loss": 0.09696399420499802, "global_step": 169351, "epoch": 1902, "lr": 3.112952261000551e-05}
{"train_loss": 0.11707603186368942, "global_step": 169352, "epoch": 1902, "lr": 3.112898574824872e-05}
{"train_loss": 0.04881712794303894, "global_step": 169353, "epoch": 1902, "lr": 3.112844888902892e-05}
{"train_loss": 0.07513643056154251, "global_step": 169354, "epoch": 1902, "lr": 3.112791203234615e-05}
{"train_loss": 0.09853284806013107, "global_step": 169355, "epoch": 1902, "lr": 3.1127375178200534e-05}
{"train_loss": 0.12935082614421844, "global_step": 169356, "epoch": 1902, "lr": 3.112683832659208e-05}
{"train_loss": 0.07543815672397614, "global_step": 169357, "epoch": 1902, "lr": 3.112630147752092e-05}
{"train_loss": 0.059423744678497314, "global_step": 169358, "epoch": 1902, "lr": 3.112576463098708e-05}
{"train_loss": 0.050928227603435516, "global_step": 169359, "epoch": 1902, "lr": 3.112522778699064e-05}
{"train_loss": 0.10434673726558685, "global_step": 169360, "epoch": 1902, "lr": 3.112469094553171e-05}
{"train_loss": 0.07463636249303818, "global_step": 169361, "epoch": 1902, "lr": 3.1124154106610315e-05}
{"train_loss": 0.08334353566169739, "global_step": 169362, "epoch": 1902, "lr": 3.112361727022657e-05}
{"train_loss": 0.17247219383716583, "global_step": 169363, "epoch": 1902, "lr": 3.1123080436380504e-05}
{"train_loss": 0.10948464274406433, "global_step": 169364, "epoch": 1902, "lr": 3.112254360507222e-05}
{"train_loss": 0.10025428980588913, "global_step": 169365, "epoch": 1902, "lr": 3.112200677630178e-05}
{"train_loss": 0.08835146186894245, "global_step": 169366, "epoch": 1902, "lr": 3.112146995006925e-05, "val_loss": 6.962600231170654}
{"train_loss": 0.11288311332464218, "global_step": 169367, "epoch": 1903, "lr": 3.112093312637472e-05}
{"train_loss": 0.1333191692829132, "global_step": 169368, "epoch": 1903, "lr": 3.1120396305218236e-05}
{"train_loss": 0.08645422756671906, "global_step": 169369, "epoch": 1903, "lr": 3.11198594865999e-05}
{"train_loss": 0.11546233296394348, "global_step": 169370, "epoch": 1903, "lr": 3.111932267051977e-05}
{"train_loss": 0.1344408094882965, "global_step": 169371, "epoch": 1903, "lr": 3.11187858569779e-05}
{"train_loss": 0.08592696487903595, "global_step": 169372, "epoch": 1903, "lr": 3.111824904597439e-05}
{"train_loss": 0.1352594941854477, "global_step": 169373, "epoch": 1903, "lr": 3.111771223750929e-05}
{"train_loss": 0.061242371797561646, "global_step": 169374, "epoch": 1903, "lr": 3.111717543158269e-05}
{"train_loss": 0.07210880517959595, "global_step": 169375, "epoch": 1903, "lr": 3.111663862819465e-05}
{"train_loss": 0.058472562581300735, "global_step": 169376, "epoch": 1903, "lr": 3.1116101827345246e-05}
{"train_loss": 0.07618936151266098, "global_step": 169377, "epoch": 1903, "lr": 3.111556502903456e-05}
{"train_loss": 0.140827015042305, "global_step": 169378, "epoch": 1903, "lr": 3.1115028233262654e-05}
{"train_loss": 0.05628354474902153, "global_step": 169379, "epoch": 1903, "lr": 3.1114491440029615e-05}
{"train_loss": 0.11307259649038315, "global_step": 169380, "epoch": 1903, "lr": 3.1113954649335486e-05}
{"train_loss": 0.06718358397483826, "global_step": 169381, "epoch": 1903, "lr": 3.111341786118037e-05}
{"train_loss": 0.02533891797065735, "global_step": 169382, "epoch": 1903, "lr": 3.1112881075564303e-05}
{"train_loss": 0.13821333646774292, "global_step": 169383, "epoch": 1903, "lr": 3.1112344292487406e-05}
{"train_loss": 0.05558403208851814, "global_step": 169384, "epoch": 1903, "lr": 3.111180751194971e-05}
{"train_loss": 0.045208755880594254, "global_step": 169385, "epoch": 1903, "lr": 3.1111270733951304e-05}
{"train_loss": 0.061472609639167786, "global_step": 169386, "epoch": 1903, "lr": 3.111073395849225e-05}
{"train_loss": 0.07119414955377579, "global_step": 169387, "epoch": 1903, "lr": 3.111019718557263e-05}
{"train_loss": 0.052677638828754425, "global_step": 169388, "epoch": 1903, "lr": 3.110966041519252e-05}
{"train_loss": 0.05439704284071922, "global_step": 169389, "epoch": 1903, "lr": 3.110912364735199e-05}
{"train_loss": 0.08948872238397598, "global_step": 169390, "epoch": 1903, "lr": 3.1108586882051096e-05}
{"train_loss": 0.11827564984560013, "global_step": 169391, "epoch": 1903, "lr": 3.110805011928993e-05}
{"train_loss": 0.09016882628202438, "global_step": 169392, "epoch": 1903, "lr": 3.1107513359068544e-05}
{"train_loss": 0.04833628982305527, "global_step": 169393, "epoch": 1903, "lr": 3.1106976601387036e-05}
{"train_loss": 0.07828228920698166, "global_step": 169394, "epoch": 1903, "lr": 3.1106439846245455e-05}
{"train_loss": 0.05210064351558685, "global_step": 169395, "epoch": 1903, "lr": 3.110590309364388e-05}
{"train_loss": 0.17008772492408752, "global_step": 169396, "epoch": 1903, "lr": 3.1105366343582404e-05}
{"train_loss": 0.14997483789920807, "global_step": 169397, "epoch": 1903, "lr": 3.1104829596061056e-05}
{"train_loss": 0.08653800189495087, "global_step": 169398, "epoch": 1903, "lr": 3.110429285107997e-05}
{"train_loss": 0.10771702975034714, "global_step": 169399, "epoch": 1903, "lr": 3.110375610863914e-05}
{"train_loss": 0.10926296561956406, "global_step": 169400, "epoch": 1903, "lr": 3.110321936873871e-05}
{"train_loss": 0.1088002622127533, "global_step": 169401, "epoch": 1903, "lr": 3.1102682631378705e-05}
{"train_loss": 0.14945577085018158, "global_step": 169402, "epoch": 1903, "lr": 3.110214589655923e-05}
{"train_loss": 0.10043682903051376, "global_step": 169403, "epoch": 1903, "lr": 3.110160916428033e-05}
{"train_loss": 0.07978936284780502, "global_step": 169404, "epoch": 1903, "lr": 3.1101072434542094e-05}
{"train_loss": 0.11026789993047714, "global_step": 169405, "epoch": 1903, "lr": 3.110053570734458e-05}
{"train_loss": 0.1227211207151413, "global_step": 169406, "epoch": 1903, "lr": 3.109999898268788e-05}
{"train_loss": 0.11895529180765152, "global_step": 169407, "epoch": 1903, "lr": 3.1099462260572046e-05}
{"train_loss": 0.06306400150060654, "global_step": 169408, "epoch": 1903, "lr": 3.109892554099717e-05}
{"train_loss": 0.09394842386245728, "global_step": 169409, "epoch": 1903, "lr": 3.10983888239633e-05}
{"train_loss": 0.1117396429181099, "global_step": 169410, "epoch": 1903, "lr": 3.1097852109470526e-05}
{"train_loss": 0.04091189056634903, "global_step": 169411, "epoch": 1903, "lr": 3.109731539751891e-05}
{"train_loss": 0.07773598283529282, "global_step": 169412, "epoch": 1903, "lr": 3.1096778688108516e-05}
{"train_loss": 0.09953649342060089, "global_step": 169413, "epoch": 1903, "lr": 3.1096241981239464e-05}
{"train_loss": 0.09089283645153046, "global_step": 169414, "epoch": 1903, "lr": 3.109570527691177e-05}
{"train_loss": 0.1394922435283661, "global_step": 169415, "epoch": 1903, "lr": 3.1095168575125545e-05}
{"train_loss": 0.10034844279289246, "global_step": 169416, "epoch": 1903, "lr": 3.109463187588082e-05}
{"train_loss": 0.06398991495370865, "global_step": 169417, "epoch": 1903, "lr": 3.1094095179177715e-05}
{"train_loss": 0.09672359377145767, "global_step": 169418, "epoch": 1903, "lr": 3.109355848501626e-05}
{"train_loss": 0.08008256554603577, "global_step": 169419, "epoch": 1903, "lr": 3.109302179339656e-05}
{"train_loss": 0.07519017159938812, "global_step": 169420, "epoch": 1903, "lr": 3.109248510431866e-05}
{"train_loss": 0.11113409698009491, "global_step": 169421, "epoch": 1903, "lr": 3.1091948417782654e-05}
{"train_loss": 0.05064147710800171, "global_step": 169422, "epoch": 1903, "lr": 3.109141173378861e-05}
{"train_loss": 0.07781326770782471, "global_step": 169423, "epoch": 1903, "lr": 3.109087505233658e-05}
{"train_loss": 0.06309697777032852, "global_step": 169424, "epoch": 1903, "lr": 3.1090338373426665e-05}
{"train_loss": 0.09882588684558868, "global_step": 169425, "epoch": 1903, "lr": 3.108980169705891e-05}
{"train_loss": 0.1343451738357544, "global_step": 169426, "epoch": 1903, "lr": 3.108926502323341e-05}
{"train_loss": 0.08220870792865753, "global_step": 169427, "epoch": 1903, "lr": 3.1088728351950236e-05}
{"train_loss": 0.11955467611551285, "global_step": 169428, "epoch": 1903, "lr": 3.108819168320943e-05}
{"train_loss": 0.05159466713666916, "global_step": 169429, "epoch": 1903, "lr": 3.1087655017011106e-05}
{"train_loss": 0.08698797225952148, "global_step": 169430, "epoch": 1903, "lr": 3.10871183533553e-05}
{"train_loss": 0.07022322714328766, "global_step": 169431, "epoch": 1903, "lr": 3.10865816922421e-05}
{"train_loss": 0.07679189741611481, "global_step": 169432, "epoch": 1903, "lr": 3.1086045033671596e-05}
{"train_loss": 0.0893377810716629, "global_step": 169433, "epoch": 1903, "lr": 3.1085508377643827e-05}
{"train_loss": 0.12440259754657745, "global_step": 169434, "epoch": 1903, "lr": 3.10849717241589e-05}
{"train_loss": 0.06686756014823914, "global_step": 169435, "epoch": 1903, "lr": 3.108443507321684e-05}
{"train_loss": 0.12981994450092316, "global_step": 169436, "epoch": 1903, "lr": 3.108389842481778e-05}
{"train_loss": 0.09670063108205795, "global_step": 169437, "epoch": 1903, "lr": 3.108336177896173e-05}
{"train_loss": 0.10385437309741974, "global_step": 169438, "epoch": 1903, "lr": 3.108282513564881e-05}
{"train_loss": 0.07800814509391785, "global_step": 169439, "epoch": 1903, "lr": 3.108228849487906e-05}
{"train_loss": 0.06816992163658142, "global_step": 169440, "epoch": 1903, "lr": 3.108175185665258e-05}
{"train_loss": 0.14811934530735016, "global_step": 169441, "epoch": 1903, "lr": 3.1081215220969426e-05}
{"train_loss": 0.10895118862390518, "global_step": 169442, "epoch": 1903, "lr": 3.1080678587829666e-05}
{"train_loss": 0.0710354745388031, "global_step": 169443, "epoch": 1903, "lr": 3.108014195723338e-05}
{"train_loss": 0.08441416919231415, "global_step": 169444, "epoch": 1903, "lr": 3.107960532918064e-05}
{"train_loss": 0.06523866951465607, "global_step": 169445, "epoch": 1903, "lr": 3.107906870367151e-05}
{"train_loss": 0.05882425978779793, "global_step": 169446, "epoch": 1903, "lr": 3.107853208070608e-05}
{"train_loss": 0.03302213177084923, "global_step": 169447, "epoch": 1903, "lr": 3.1077995460284394e-05}
{"train_loss": 0.07831797748804092, "global_step": 169448, "epoch": 1903, "lr": 3.107745884240656e-05}
{"train_loss": 0.07487767934799194, "global_step": 169449, "epoch": 1903, "lr": 3.1076922227072614e-05}
{"train_loss": 0.07492995262145996, "global_step": 169450, "epoch": 1903, "lr": 3.1076385614282644e-05}
{"train_loss": 0.07476074248552322, "global_step": 169451, "epoch": 1903, "lr": 3.107584900403674e-05}
{"train_loss": 0.060058414936065674, "global_step": 169452, "epoch": 1903, "lr": 3.107531239633494e-05}
{"train_loss": 0.09126073122024536, "global_step": 169453, "epoch": 1903, "lr": 3.1074775791177346e-05}
{"train_loss": 0.08532416820526123, "global_step": 169454, "epoch": 1903, "lr": 3.107423918856401e-05}
{"train_loss": 0.08957452992542406, "global_step": 169455, "epoch": 1903, "lr": 3.1073702588495025e-05, "val_loss": 7.067691802978516}
{"train_loss": 0.16338105499744415, "global_step": 169456, "epoch": 1904, "lr": 3.107316599097043e-05}
{"train_loss": 0.05565192177891731, "global_step": 169457, "epoch": 1904, "lr": 3.107262939599034e-05}
{"train_loss": 0.0646391212940216, "global_step": 169458, "epoch": 1904, "lr": 3.107209280355478e-05}
{"train_loss": 0.08429604023694992, "global_step": 169459, "epoch": 1904, "lr": 3.107155621366387e-05}
{"train_loss": 0.11375326663255692, "global_step": 169460, "epoch": 1904, "lr": 3.107101962631765e-05}
{"train_loss": 0.07382625341415405, "global_step": 169461, "epoch": 1904, "lr": 3.1070483041516195e-05}
{"train_loss": 0.11727458238601685, "global_step": 169462, "epoch": 1904, "lr": 3.106994645925959e-05}
{"train_loss": 0.10186392813920975, "global_step": 169463, "epoch": 1904, "lr": 3.1069409879547895e-05}
{"train_loss": 0.11132147908210754, "global_step": 169464, "epoch": 1904, "lr": 3.106887330238118e-05}
{"train_loss": 0.10306544601917267, "global_step": 169465, "epoch": 1904, "lr": 3.106833672775954e-05}
{"train_loss": 0.11201288551092148, "global_step": 169466, "epoch": 1904, "lr": 3.1067800155683024e-05}
{"train_loss": 0.18614579737186432, "global_step": 169467, "epoch": 1904, "lr": 3.1067263586151704e-05}
{"train_loss": 0.051540665328502655, "global_step": 169468, "epoch": 1904, "lr": 3.106672701916568e-05}
{"train_loss": 0.14041602611541748, "global_step": 169469, "epoch": 1904, "lr": 3.106619045472497e-05}
{"train_loss": 0.1225922703742981, "global_step": 169470, "epoch": 1904, "lr": 3.1065653892829716e-05}
{"train_loss": 0.06106390058994293, "global_step": 169471, "epoch": 1904, "lr": 3.106511733347993e-05}
{"train_loss": 0.10619021952152252, "global_step": 169472, "epoch": 1904, "lr": 3.106458077667572e-05}
{"train_loss": 0.07918388396501541, "global_step": 169473, "epoch": 1904, "lr": 3.106404422241714e-05}
{"train_loss": 0.07209060341119766, "global_step": 169474, "epoch": 1904, "lr": 3.106350767070429e-05}
{"train_loss": 0.08375412225723267, "global_step": 169475, "epoch": 1904, "lr": 3.1062971121537184e-05}
{"train_loss": 0.08086918294429779, "global_step": 169476, "epoch": 1904, "lr": 3.106243457491596e-05}
{"train_loss": 0.10540517419576645, "global_step": 169477, "epoch": 1904, "lr": 3.106189803084064e-05}
{"train_loss": 0.04473856836557388, "global_step": 169478, "epoch": 1904, "lr": 3.106136148931134e-05}
{"train_loss": 0.10138032585382462, "global_step": 169479, "epoch": 1904, "lr": 3.1060824950328085e-05}
{"train_loss": 0.12572818994522095, "global_step": 169480, "epoch": 1904, "lr": 3.106028841389099e-05}
{"train_loss": 0.10206551104784012, "global_step": 169481, "epoch": 1904, "lr": 3.1059751880000096e-05}
{"train_loss": 0.09354723244905472, "global_step": 169482, "epoch": 1904, "lr": 3.1059215348655495e-05}
{"train_loss": 0.08491486310958862, "global_step": 169483, "epoch": 1904, "lr": 3.105867881985725e-05}
{"train_loss": 0.10107506066560745, "global_step": 169484, "epoch": 1904, "lr": 3.105814229360543e-05}
{"train_loss": 0.06837372481822968, "global_step": 169485, "epoch": 1904, "lr": 3.105760576990012e-05}
{"train_loss": 0.052772797644138336, "global_step": 169486, "epoch": 1904, "lr": 3.105706924874137e-05}
{"train_loss": 0.0876954197883606, "global_step": 169487, "epoch": 1904, "lr": 3.1056532730129286e-05}
{"train_loss": 0.12039657682180405, "global_step": 169488, "epoch": 1904, "lr": 3.10559962140639e-05}
{"train_loss": 0.07841320335865021, "global_step": 169489, "epoch": 1904, "lr": 3.1055459700545317e-05}
{"train_loss": 0.029161473736166954, "global_step": 169490, "epoch": 1904, "lr": 3.105492318957359e-05}
{"train_loss": 0.052749693393707275, "global_step": 169491, "epoch": 1904, "lr": 3.105438668114881e-05}
{"train_loss": 0.08481624722480774, "global_step": 169492, "epoch": 1904, "lr": 3.105385017527102e-05}
{"train_loss": 0.030349181964993477, "global_step": 169493, "epoch": 1904, "lr": 3.105331367194033e-05}
{"train_loss": 0.09552381932735443, "global_step": 169494, "epoch": 1904, "lr": 3.105277717115677e-05}
{"train_loss": 0.12631559371948242, "global_step": 169495, "epoch": 1904, "lr": 3.105224067292045e-05}
{"train_loss": 0.08134787529706955, "global_step": 169496, "epoch": 1904, "lr": 3.10517041772314e-05}
{"train_loss": 0.07917916774749756, "global_step": 169497, "epoch": 1904, "lr": 3.105116768408975e-05}
{"train_loss": 0.11609943211078644, "global_step": 169498, "epoch": 1904, "lr": 3.105063119349552e-05}
{"train_loss": 0.10681214928627014, "global_step": 169499, "epoch": 1904, "lr": 3.105009470544881e-05}
{"train_loss": 0.10115472227334976, "global_step": 169500, "epoch": 1904, "lr": 3.104955821994968e-05}
{"train_loss": 0.0570191964507103, "global_step": 169501, "epoch": 1904, "lr": 3.104902173699821e-05}
{"train_loss": 0.12061398476362228, "global_step": 169502, "epoch": 1904, "lr": 3.1048485256594454e-05}
{"train_loss": 0.09393520653247833, "global_step": 169503, "epoch": 1904, "lr": 3.10479487787385e-05}
{"train_loss": 0.11779485642910004, "global_step": 169504, "epoch": 1904, "lr": 3.1047412303430434e-05}
{"train_loss": 0.09077199548482895, "global_step": 169505, "epoch": 1904, "lr": 3.10468758306703e-05}
{"train_loss": 0.09641727060079575, "global_step": 169506, "epoch": 1904, "lr": 3.104633936045819e-05}
{"train_loss": 0.06888185441493988, "global_step": 169507, "epoch": 1904, "lr": 3.104580289279416e-05}
{"train_loss": 0.10372816026210785, "global_step": 169508, "epoch": 1904, "lr": 3.1045266427678305e-05}
{"train_loss": 0.055985283106565475, "global_step": 169509, "epoch": 1904, "lr": 3.1044729965110665e-05}
{"train_loss": 0.05376671999692917, "global_step": 169510, "epoch": 1904, "lr": 3.104419350509135e-05}
{"train_loss": 0.12330488860607147, "global_step": 169511, "epoch": 1904, "lr": 3.104365704762039e-05}
{"train_loss": 0.08604571968317032, "global_step": 169512, "epoch": 1904, "lr": 3.10431205926979e-05}
{"train_loss": 0.08678939193487167, "global_step": 169513, "epoch": 1904, "lr": 3.104258414032391e-05}
{"train_loss": 0.10174080729484558, "global_step": 169514, "epoch": 1904, "lr": 3.1042047690498535e-05}
{"train_loss": 0.12240693718194962, "global_step": 169515, "epoch": 1904, "lr": 3.104151124322181e-05}
{"train_loss": 0.0956215187907219, "global_step": 169516, "epoch": 1904, "lr": 3.104097479849384e-05}
{"train_loss": 0.054693225771188736, "global_step": 169517, "epoch": 1904, "lr": 3.1040438356314654e-05}
{"train_loss": 0.10553109645843506, "global_step": 169518, "epoch": 1904, "lr": 3.1039901916684375e-05}
{"train_loss": 0.10584420710802078, "global_step": 169519, "epoch": 1904, "lr": 3.103936547960304e-05}
{"train_loss": 0.03832991048693657, "global_step": 169520, "epoch": 1904, "lr": 3.103882904507073e-05}
{"train_loss": 0.09763339906930923, "global_step": 169521, "epoch": 1904, "lr": 3.1038292613087525e-05}
{"train_loss": 0.14835791289806366, "global_step": 169522, "epoch": 1904, "lr": 3.103775618365347e-05}
{"train_loss": 0.12434239685535431, "global_step": 169523, "epoch": 1904, "lr": 3.1037219756768685e-05}
{"train_loss": 0.06047994643449783, "global_step": 169524, "epoch": 1904, "lr": 3.103668333243319e-05}
{"train_loss": 0.038750167936086655, "global_step": 169525, "epoch": 1904, "lr": 3.10361469106471e-05}
{"train_loss": 0.07993216812610626, "global_step": 169526, "epoch": 1904, "lr": 3.103561049141045e-05}
{"train_loss": 0.0723557099699974, "global_step": 169527, "epoch": 1904, "lr": 3.1035074074723355e-05}
{"train_loss": 0.04954967647790909, "global_step": 169528, "epoch": 1904, "lr": 3.103453766058585e-05}
{"train_loss": 0.094453364610672, "global_step": 169529, "epoch": 1904, "lr": 3.103400124899803e-05}
{"train_loss": 0.07399661093950272, "global_step": 169530, "epoch": 1904, "lr": 3.103346483995993e-05}
{"train_loss": 0.055900607258081436, "global_step": 169531, "epoch": 1904, "lr": 3.103292843347169e-05}
{"train_loss": 0.16791963577270508, "global_step": 169532, "epoch": 1904, "lr": 3.103239202953331e-05}
{"train_loss": 0.03706565126776695, "global_step": 169533, "epoch": 1904, "lr": 3.103185562814491e-05}
{"train_loss": 0.057926397770643234, "global_step": 169534, "epoch": 1904, "lr": 3.103131922930653e-05}
{"train_loss": 0.055478375405073166, "global_step": 169535, "epoch": 1904, "lr": 3.103078283301828e-05}
{"train_loss": 0.14911045134067535, "global_step": 169536, "epoch": 1904, "lr": 3.103024643928019e-05}
{"train_loss": 0.07190258800983429, "global_step": 169537, "epoch": 1904, "lr": 3.102971004809236e-05}
{"train_loss": 0.08579880744218826, "global_step": 169538, "epoch": 1904, "lr": 3.102917365945486e-05}
{"train_loss": 0.11914686858654022, "global_step": 169539, "epoch": 1904, "lr": 3.1028637273367745e-05}
{"train_loss": 0.06225290894508362, "global_step": 169540, "epoch": 1904, "lr": 3.102810088983111e-05}
{"train_loss": 0.07892660051584244, "global_step": 169541, "epoch": 1904, "lr": 3.1027564508845e-05}
{"train_loss": 0.08043461292982101, "global_step": 169542, "epoch": 1904, "lr": 3.102702813040953e-05}
{"train_loss": 0.07311994582414627, "global_step": 169543, "epoch": 1904, "lr": 3.102649175452471e-05}
{"train_loss": 0.09041192755103111, "global_step": 169544, "epoch": 1904, "lr": 3.102595538119067e-05, "val_loss": 7.061726093292236}
{"train_loss": 0.0645211935043335, "global_step": 169545, "epoch": 1905, "lr": 3.1025419010407455e-05}
{"train_loss": 0.14270366728305817, "global_step": 169546, "epoch": 1905, "lr": 3.102488264217515e-05}
{"train_loss": 0.08047649264335632, "global_step": 169547, "epoch": 1905, "lr": 3.102434627649379e-05}
{"train_loss": 0.09975013881921768, "global_step": 169548, "epoch": 1905, "lr": 3.1023809913363505e-05}
{"train_loss": 0.09425503760576248, "global_step": 169549, "epoch": 1905, "lr": 3.102327355278432e-05}
{"train_loss": 0.07902340590953827, "global_step": 169550, "epoch": 1905, "lr": 3.102273719475633e-05}
{"train_loss": 0.06094053015112877, "global_step": 169551, "epoch": 1905, "lr": 3.1022200839279596e-05}
{"train_loss": 0.06717517226934433, "global_step": 169552, "epoch": 1905, "lr": 3.102166448635422e-05}
{"train_loss": 0.08091054856777191, "global_step": 169553, "epoch": 1905, "lr": 3.102112813598021e-05}
{"train_loss": 0.09623104333877563, "global_step": 169554, "epoch": 1905, "lr": 3.1020591788157714e-05}
{"train_loss": 0.09472104907035828, "global_step": 169555, "epoch": 1905, "lr": 3.102005544288674e-05}
{"train_loss": 0.07209453731775284, "global_step": 169556, "epoch": 1905, "lr": 3.1019519100167415e-05}
{"train_loss": 0.10903725773096085, "global_step": 169557, "epoch": 1905, "lr": 3.101898275999977e-05}
{"train_loss": 0.09329705685377121, "global_step": 169558, "epoch": 1905, "lr": 3.101844642238388e-05}
{"train_loss": 0.053331900388002396, "global_step": 169559, "epoch": 1905, "lr": 3.1017910087319855e-05}
{"train_loss": 0.0725894421339035, "global_step": 169560, "epoch": 1905, "lr": 3.101737375480772e-05}
{"train_loss": 0.060707684606313705, "global_step": 169561, "epoch": 1905, "lr": 3.101683742484758e-05}
{"train_loss": 0.06896393746137619, "global_step": 169562, "epoch": 1905, "lr": 3.1016301097439485e-05}
{"train_loss": 0.056114401668310165, "global_step": 169563, "epoch": 1905, "lr": 3.101576477258353e-05}
{"train_loss": 0.15090112388134003, "global_step": 169564, "epoch": 1905, "lr": 3.101522845027975e-05}
{"train_loss": 0.1304170936346054, "global_step": 169565, "epoch": 1905, "lr": 3.101469213052827e-05}
{"train_loss": 0.10970136523246765, "global_step": 169566, "epoch": 1905, "lr": 3.101415581332911e-05}
{"train_loss": 0.06662137806415558, "global_step": 169567, "epoch": 1905, "lr": 3.101361949868239e-05}
{"train_loss": 0.09590993821620941, "global_step": 169568, "epoch": 1905, "lr": 3.1013083186588134e-05}
{"train_loss": 0.05720527842640877, "global_step": 169569, "epoch": 1905, "lr": 3.101254687704645e-05}
{"train_loss": 0.09148523211479187, "global_step": 169570, "epoch": 1905, "lr": 3.101201057005738e-05}
{"train_loss": 0.09081082791090012, "global_step": 169571, "epoch": 1905, "lr": 3.101147426562104e-05}
{"train_loss": 0.05601448193192482, "global_step": 169572, "epoch": 1905, "lr": 3.101093796373745e-05}
{"train_loss": 0.1394430696964264, "global_step": 169573, "epoch": 1905, "lr": 3.1010401664406735e-05}
{"train_loss": 0.09520795941352844, "global_step": 169574, "epoch": 1905, "lr": 3.100986536762892e-05}
{"train_loss": 0.07991763204336166, "global_step": 169575, "epoch": 1905, "lr": 3.100932907340411e-05}
{"train_loss": 0.03856724873185158, "global_step": 169576, "epoch": 1905, "lr": 3.100879278173236e-05}
{"train_loss": 0.0833439752459526, "global_step": 169577, "epoch": 1905, "lr": 3.100825649261373e-05}
{"train_loss": 0.08662690222263336, "global_step": 169578, "epoch": 1905, "lr": 3.100772020604833e-05}
{"train_loss": 0.056449443101882935, "global_step": 169579, "epoch": 1905, "lr": 3.10071839220362e-05}
{"train_loss": 0.0856635719537735, "global_step": 169580, "epoch": 1905, "lr": 3.100664764057743e-05}
{"train_loss": 0.11137774586677551, "global_step": 169581, "epoch": 1905, "lr": 3.1006111361672075e-05}
{"train_loss": 0.09284766018390656, "global_step": 169582, "epoch": 1905, "lr": 3.100557508532023e-05}
{"train_loss": 0.09447688609361649, "global_step": 169583, "epoch": 1905, "lr": 3.1005038811521934e-05}
{"train_loss": 0.04090701416134834, "global_step": 169584, "epoch": 1905, "lr": 3.1004502540277306e-05}
{"train_loss": 0.06024724245071411, "global_step": 169585, "epoch": 1905, "lr": 3.1003966271586366e-05}
{"train_loss": 0.09035434573888779, "global_step": 169586, "epoch": 1905, "lr": 3.100343000544923e-05}
{"train_loss": 0.06340713798999786, "global_step": 169587, "epoch": 1905, "lr": 3.1002893741865935e-05}
{"train_loss": 0.08210674673318863, "global_step": 169588, "epoch": 1905, "lr": 3.1002357480836585e-05}
{"train_loss": 0.06399749219417572, "global_step": 169589, "epoch": 1905, "lr": 3.1001821222361215e-05}
{"train_loss": 0.06803465634584427, "global_step": 169590, "epoch": 1905, "lr": 3.100128496643995e-05}
{"train_loss": 0.06725151836872101, "global_step": 169591, "epoch": 1905, "lr": 3.1000748713072804e-05}
{"train_loss": 0.11411860585212708, "global_step": 169592, "epoch": 1905, "lr": 3.100021246225989e-05}
{"train_loss": 0.10223230719566345, "global_step": 169593, "epoch": 1905, "lr": 3.099967621400125e-05}
{"train_loss": 0.11658602952957153, "global_step": 169594, "epoch": 1905, "lr": 3.099913996829699e-05}
{"train_loss": 0.04860648140311241, "global_step": 169595, "epoch": 1905, "lr": 3.099860372514715e-05}
{"train_loss": 0.09426166862249374, "global_step": 169596, "epoch": 1905, "lr": 3.099806748455182e-05}
{"train_loss": 0.06453853100538254, "global_step": 169597, "epoch": 1905, "lr": 3.099753124651108e-05}
{"train_loss": 0.046683862805366516, "global_step": 169598, "epoch": 1905, "lr": 3.0996995011024975e-05}
{"train_loss": 0.1183473989367485, "global_step": 169599, "epoch": 1905, "lr": 3.0996458778093605e-05}
{"train_loss": 0.035493746399879456, "global_step": 169600, "epoch": 1905, "lr": 3.0995922547717016e-05}
{"train_loss": 0.09920990467071533, "global_step": 169601, "epoch": 1905, "lr": 3.0995386319895314e-05}
{"train_loss": 0.12207391858100891, "global_step": 169602, "epoch": 1905, "lr": 3.099485009462852e-05}
{"train_loss": 0.09155752509832382, "global_step": 169603, "epoch": 1905, "lr": 3.099431387191677e-05}
{"train_loss": 0.07670269906520844, "global_step": 169604, "epoch": 1905, "lr": 3.099377765176008e-05}
{"train_loss": 0.0881270319223404, "global_step": 169605, "epoch": 1905, "lr": 3.099324143415856e-05}
{"train_loss": 0.12053216248750687, "global_step": 169606, "epoch": 1905, "lr": 3.0992705219112244e-05}
{"train_loss": 0.06069124490022659, "global_step": 169607, "epoch": 1905, "lr": 3.099216900662125e-05}
{"train_loss": 0.06076066941022873, "global_step": 169608, "epoch": 1905, "lr": 3.099163279668561e-05}
{"train_loss": 0.09781602025032043, "global_step": 169609, "epoch": 1905, "lr": 3.099109658930544e-05}
{"train_loss": 0.11642089486122131, "global_step": 169610, "epoch": 1905, "lr": 3.0990560384480755e-05}
{"train_loss": 0.08484315872192383, "global_step": 169611, "epoch": 1905, "lr": 3.099002418221168e-05}
{"train_loss": 0.08430689573287964, "global_step": 169612, "epoch": 1905, "lr": 3.0989487982498245e-05}
{"train_loss": 0.06412995606660843, "global_step": 169613, "epoch": 1905, "lr": 3.0988951785340546e-05}
{"train_loss": 0.09967130422592163, "global_step": 169614, "epoch": 1905, "lr": 3.0988415590738664e-05}
{"train_loss": 0.11949572712182999, "global_step": 169615, "epoch": 1905, "lr": 3.0987879398692646e-05}
{"train_loss": 0.05713334679603577, "global_step": 169616, "epoch": 1905, "lr": 3.098734320920259e-05}
{"train_loss": 0.06427560746669769, "global_step": 169617, "epoch": 1905, "lr": 3.0986807022268535e-05}
{"train_loss": 0.08231566101312637, "global_step": 169618, "epoch": 1905, "lr": 3.098627083789059e-05}
{"train_loss": 0.061604566872119904, "global_step": 169619, "epoch": 1905, "lr": 3.098573465606879e-05}
{"train_loss": 0.08537643402814865, "global_step": 169620, "epoch": 1905, "lr": 3.0985198476803246e-05}
{"train_loss": 0.06292857229709625, "global_step": 169621, "epoch": 1905, "lr": 3.098466230009399e-05}
{"train_loss": 0.14384087920188904, "global_step": 169622, "epoch": 1905, "lr": 3.098412612594114e-05}
{"train_loss": 0.14365176856517792, "global_step": 169623, "epoch": 1905, "lr": 3.098358995434472e-05}
{"train_loss": 0.0940786823630333, "global_step": 169624, "epoch": 1905, "lr": 3.098305378530484e-05}
{"train_loss": 0.16554392874240875, "global_step": 169625, "epoch": 1905, "lr": 3.098251761882154e-05}
{"train_loss": 0.07916075736284256, "global_step": 169626, "epoch": 1905, "lr": 3.098198145489493e-05}
{"train_loss": 0.06558596342802048, "global_step": 169627, "epoch": 1905, "lr": 3.0981445293525035e-05}
{"train_loss": 0.05832721292972565, "global_step": 169628, "epoch": 1905, "lr": 3.098090913471197e-05}
{"train_loss": 0.07323343306779861, "global_step": 169629, "epoch": 1905, "lr": 3.0980372978455776e-05}
{"train_loss": 0.09306477755308151, "global_step": 169630, "epoch": 1905, "lr": 3.097983682475655e-05}
{"train_loss": 0.05435187742114067, "global_step": 169631, "epoch": 1905, "lr": 3.097930067361435e-05}
{"train_loss": 0.08195865899324417, "global_step": 169632, "epoch": 1905, "lr": 3.0978764525029256e-05}
{"train_loss": 0.08536812303106436, "global_step": 169633, "epoch": 1905, "lr": 3.097822837900133e-05, "val_loss": 7.370055198669434, "train_action_mse_error": 33.500728607177734}
{"train_loss": 0.08259569853544235, "global_step": 169634, "epoch": 1906, "lr": 3.0977692235530655e-05}
{"train_loss": 0.1287216693162918, "global_step": 169635, "epoch": 1906, "lr": 3.097715609461731e-05}
{"train_loss": 0.052772603929042816, "global_step": 169636, "epoch": 1906, "lr": 3.097661995626132e-05}
{"train_loss": 0.07544906437397003, "global_step": 169637, "epoch": 1906, "lr": 3.097608382046283e-05}
{"train_loss": 0.14945156872272491, "global_step": 169638, "epoch": 1906, "lr": 3.097554768722184e-05}
{"train_loss": 0.0635952353477478, "global_step": 169639, "epoch": 1906, "lr": 3.097501155653848e-05}
{"train_loss": 0.05965116620063782, "global_step": 169640, "epoch": 1906, "lr": 3.0974475428412776e-05}
{"train_loss": 0.06347750127315521, "global_step": 169641, "epoch": 1906, "lr": 3.0973939302844845e-05}
{"train_loss": 0.07403632253408432, "global_step": 169642, "epoch": 1906, "lr": 3.0973403179834716e-05}
{"train_loss": 0.07745061814785004, "global_step": 169643, "epoch": 1906, "lr": 3.09728670593825e-05}
{"train_loss": 0.09750759601593018, "global_step": 169644, "epoch": 1906, "lr": 3.0972330941488224e-05}
{"train_loss": 0.042330820113420486, "global_step": 169645, "epoch": 1906, "lr": 3.097179482615202e-05}
{"train_loss": 0.08010990917682648, "global_step": 169646, "epoch": 1906, "lr": 3.09712587133739e-05}
{"train_loss": 0.06866948306560516, "global_step": 169647, "epoch": 1906, "lr": 3.097072260315398e-05}
{"train_loss": 0.04370260611176491, "global_step": 169648, "epoch": 1906, "lr": 3.097018649549229e-05}
{"train_loss": 0.16202355921268463, "global_step": 169649, "epoch": 1906, "lr": 3.096965039038894e-05}
{"train_loss": 0.14023160934448242, "global_step": 169650, "epoch": 1906, "lr": 3.0969114287843994e-05}
{"train_loss": 0.10206563025712967, "global_step": 169651, "epoch": 1906, "lr": 3.096857818785751e-05}
{"train_loss": 0.10947316884994507, "global_step": 169652, "epoch": 1906, "lr": 3.096804209042957e-05}
{"train_loss": 0.13045746088027954, "global_step": 169653, "epoch": 1906, "lr": 3.0967505995560254e-05}
{"train_loss": 0.08859796822071075, "global_step": 169654, "epoch": 1906, "lr": 3.0966969903249635e-05}
{"train_loss": 0.10851459205150604, "global_step": 169655, "epoch": 1906, "lr": 3.096643381349774e-05}
{"train_loss": 0.08391136676073074, "global_step": 169656, "epoch": 1906, "lr": 3.0965897726304704e-05}
{"train_loss": 0.14373436570167542, "global_step": 169657, "epoch": 1906, "lr": 3.0965361641670555e-05}
{"train_loss": 0.07468291372060776, "global_step": 169658, "epoch": 1906, "lr": 3.09648255595954e-05}
{"train_loss": 0.04807046055793762, "global_step": 169659, "epoch": 1906, "lr": 3.096428948007927e-05}
{"train_loss": 0.10605703294277191, "global_step": 169660, "epoch": 1906, "lr": 3.096375340312227e-05}
{"train_loss": 0.08405175060033798, "global_step": 169661, "epoch": 1906, "lr": 3.0963217328724456e-05}
{"train_loss": 0.06041015684604645, "global_step": 169662, "epoch": 1906, "lr": 3.0962681256885925e-05}
{"train_loss": 0.14214448630809784, "global_step": 169663, "epoch": 1906, "lr": 3.096214518760671e-05}
{"train_loss": 0.07231317460536957, "global_step": 169664, "epoch": 1906, "lr": 3.096160912088691e-05}
{"train_loss": 0.0403437502682209, "global_step": 169665, "epoch": 1906, "lr": 3.096107305672657e-05}
{"train_loss": 0.10155387222766876, "global_step": 169666, "epoch": 1906, "lr": 3.09605369951258e-05}
{"train_loss": 0.09409498423337936, "global_step": 169667, "epoch": 1906, "lr": 3.096000093608466e-05}
{"train_loss": 0.068501316010952, "global_step": 169668, "epoch": 1906, "lr": 3.095946487960321e-05}
{"train_loss": 0.10919134318828583, "global_step": 169669, "epoch": 1906, "lr": 3.095892882568152e-05}
{"train_loss": 0.10313067585229874, "global_step": 169670, "epoch": 1906, "lr": 3.095839277431968e-05}
{"train_loss": 0.04688984155654907, "global_step": 169671, "epoch": 1906, "lr": 3.0957856725517746e-05}
{"train_loss": 0.09408982843160629, "global_step": 169672, "epoch": 1906, "lr": 3.0957320679275806e-05}
{"train_loss": 0.09925710409879684, "global_step": 169673, "epoch": 1906, "lr": 3.095678463559392e-05}
{"train_loss": 0.09210743755102158, "global_step": 169674, "epoch": 1906, "lr": 3.095624859447215e-05}
{"train_loss": 0.11648818850517273, "global_step": 169675, "epoch": 1906, "lr": 3.095571255591059e-05}
{"train_loss": 0.09478937089443207, "global_step": 169676, "epoch": 1906, "lr": 3.095517651990929e-05}
{"train_loss": 0.08405550569295883, "global_step": 169677, "epoch": 1906, "lr": 3.0954640486468355e-05}
{"train_loss": 0.14277960360050201, "global_step": 169678, "epoch": 1906, "lr": 3.095410445558781e-05}
{"train_loss": 0.11133789271116257, "global_step": 169679, "epoch": 1906, "lr": 3.095356842726778e-05}
{"train_loss": 0.056999869644641876, "global_step": 169680, "epoch": 1906, "lr": 3.095303240150829e-05}
{"train_loss": 0.04582245647907257, "global_step": 169681, "epoch": 1906, "lr": 3.0952496378309446e-05}
{"train_loss": 0.12226101756095886, "global_step": 169682, "epoch": 1906, "lr": 3.095196035767129e-05}
{"train_loss": 0.07081766426563263, "global_step": 169683, "epoch": 1906, "lr": 3.0951424339593935e-05}
{"train_loss": 0.09253162145614624, "global_step": 169684, "epoch": 1906, "lr": 3.095088832407741e-05}
{"train_loss": 0.052971795201301575, "global_step": 169685, "epoch": 1906, "lr": 3.0950352311121805e-05}
{"train_loss": 0.06328339129686356, "global_step": 169686, "epoch": 1906, "lr": 3.0949816300727207e-05}
{"train_loss": 0.1065201386809349, "global_step": 169687, "epoch": 1906, "lr": 3.094928029289367e-05}
{"train_loss": 0.05919470265507698, "global_step": 169688, "epoch": 1906, "lr": 3.094874428762127e-05}
{"train_loss": 0.06228905916213989, "global_step": 169689, "epoch": 1906, "lr": 3.0948208284910074e-05}
{"train_loss": 0.1015130802989006, "global_step": 169690, "epoch": 1906, "lr": 3.094767228476016e-05}
{"train_loss": 0.12029008567333221, "global_step": 169691, "epoch": 1906, "lr": 3.0947136287171605e-05}
{"train_loss": 0.07433976978063583, "global_step": 169692, "epoch": 1906, "lr": 3.094660029214448e-05}
{"train_loss": 0.07223884016275406, "global_step": 169693, "epoch": 1906, "lr": 3.0946064299678834e-05}
{"train_loss": 0.056871190667152405, "global_step": 169694, "epoch": 1906, "lr": 3.094552830977478e-05}
{"train_loss": 0.13312366604804993, "global_step": 169695, "epoch": 1906, "lr": 3.094499232243234e-05}
{"train_loss": 0.0693386122584343, "global_step": 169696, "epoch": 1906, "lr": 3.094445633765165e-05}
{"train_loss": 0.07365459948778152, "global_step": 169697, "epoch": 1906, "lr": 3.094392035543271e-05}
{"train_loss": 0.15533019602298737, "global_step": 169698, "epoch": 1906, "lr": 3.094338437577565e-05}
{"train_loss": 0.08572069555521011, "global_step": 169699, "epoch": 1906, "lr": 3.09428483986805e-05}
{"train_loss": 0.10187839716672897, "global_step": 169700, "epoch": 1906, "lr": 3.0942312424147374e-05}
{"train_loss": 0.06990985572338104, "global_step": 169701, "epoch": 1906, "lr": 3.0941776452176304e-05}
{"train_loss": 0.1446811854839325, "global_step": 169702, "epoch": 1906, "lr": 3.0941240482767394e-05}
{"train_loss": 0.08736984431743622, "global_step": 169703, "epoch": 1906, "lr": 3.0940704515920685e-05}
{"train_loss": 0.15029746294021606, "global_step": 169704, "epoch": 1906, "lr": 3.094016855163626e-05}
{"train_loss": 0.06085357069969177, "global_step": 169705, "epoch": 1906, "lr": 3.0939632589914224e-05}
{"train_loss": 0.17742741107940674, "global_step": 169706, "epoch": 1906, "lr": 3.0939096630754616e-05}
{"train_loss": 0.06185641139745712, "global_step": 169707, "epoch": 1906, "lr": 3.0938560674157507e-05}
{"train_loss": 0.09168749302625656, "global_step": 169708, "epoch": 1906, "lr": 3.093802472012298e-05}
{"train_loss": 0.053217388689517975, "global_step": 169709, "epoch": 1906, "lr": 3.09374887686511e-05}
{"train_loss": 0.09236086905002594, "global_step": 169710, "epoch": 1906, "lr": 3.0936952819741946e-05}
{"train_loss": 0.09223773330450058, "global_step": 169711, "epoch": 1906, "lr": 3.093641687339558e-05}
{"train_loss": 0.0704055055975914, "global_step": 169712, "epoch": 1906, "lr": 3.093588092961209e-05}
{"train_loss": 0.08815927803516388, "global_step": 169713, "epoch": 1906, "lr": 3.093534498839154e-05}
{"train_loss": 0.07098609954118729, "global_step": 169714, "epoch": 1906, "lr": 3.093480904973398e-05}
{"train_loss": 0.03746970742940903, "global_step": 169715, "epoch": 1906, "lr": 3.0934273113639525e-05}
{"train_loss": 0.07421260327100754, "global_step": 169716, "epoch": 1906, "lr": 3.093373718010821e-05}
{"train_loss": 0.12756094336509705, "global_step": 169717, "epoch": 1906, "lr": 3.093320124914014e-05}
{"train_loss": 0.15855303406715393, "global_step": 169718, "epoch": 1906, "lr": 3.093266532073534e-05}
{"train_loss": 0.09918028116226196, "global_step": 169719, "epoch": 1906, "lr": 3.093212939489394e-05}
{"train_loss": 0.11604691296815872, "global_step": 169720, "epoch": 1906, "lr": 3.093159347161596e-05}
{"train_loss": 0.11654102057218552, "global_step": 169721, "epoch": 1906, "lr": 3.0931057550901496e-05}
{"train_loss": 0.09235703032673075, "global_step": 169722, "epoch": 1906, "lr": 3.093052163275063e-05, "val_loss": 7.046032428741455}
{"train_loss": 0.1023283526301384, "global_step": 169723, "epoch": 1907, "lr": 3.092998571716341e-05}
{"train_loss": 0.055913519114255905, "global_step": 169724, "epoch": 1907, "lr": 3.092944980413994e-05}
{"train_loss": 0.08673131465911865, "global_step": 169725, "epoch": 1907, "lr": 3.0928913893680255e-05}
{"train_loss": 0.1286642700433731, "global_step": 169726, "epoch": 1907, "lr": 3.092837798578447e-05}
{"train_loss": 0.056497540324926376, "global_step": 169727, "epoch": 1907, "lr": 3.092784208045262e-05}
{"train_loss": 0.0803937166929245, "global_step": 169728, "epoch": 1907, "lr": 3.092730617768478e-05}
{"train_loss": 0.16935178637504578, "global_step": 169729, "epoch": 1907, "lr": 3.0926770277481036e-05}
{"train_loss": 0.07664401829242706, "global_step": 169730, "epoch": 1907, "lr": 3.092623437984146e-05}
{"train_loss": 0.10635493695735931, "global_step": 169731, "epoch": 1907, "lr": 3.092569848476611e-05}
{"train_loss": 0.10336635261774063, "global_step": 169732, "epoch": 1907, "lr": 3.092516259225509e-05}
{"train_loss": 0.08563794195652008, "global_step": 169733, "epoch": 1907, "lr": 3.092462670230842e-05}
{"train_loss": 0.07921299338340759, "global_step": 169734, "epoch": 1907, "lr": 3.092409081492622e-05}
{"train_loss": 0.0684896856546402, "global_step": 169735, "epoch": 1907, "lr": 3.0923554930108527e-05}
{"train_loss": 0.0787808820605278, "global_step": 169736, "epoch": 1907, "lr": 3.092301904785545e-05}
{"train_loss": 0.08814742416143417, "global_step": 169737, "epoch": 1907, "lr": 3.092248316816702e-05}
{"train_loss": 0.0717126727104187, "global_step": 169738, "epoch": 1907, "lr": 3.0921947291043355e-05}
{"train_loss": 0.1038673147559166, "global_step": 169739, "epoch": 1907, "lr": 3.092141141648447e-05}
{"train_loss": 0.07269024848937988, "global_step": 169740, "epoch": 1907, "lr": 3.092087554449048e-05}
{"train_loss": 0.050810571759939194, "global_step": 169741, "epoch": 1907, "lr": 3.0920339675061464e-05}
{"train_loss": 0.030497906729578972, "global_step": 169742, "epoch": 1907, "lr": 3.0919803808197445e-05}
{"train_loss": 0.09720628708600998, "global_step": 169743, "epoch": 1907, "lr": 3.0919267943898555e-05}
{"train_loss": 0.09157640486955643, "global_step": 169744, "epoch": 1907, "lr": 3.091873208216481e-05}
{"train_loss": 0.09384152293205261, "global_step": 169745, "epoch": 1907, "lr": 3.0918196222996345e-05}
{"train_loss": 0.09193552285432816, "global_step": 169746, "epoch": 1907, "lr": 3.091766036639317e-05}
{"train_loss": 0.12302683293819427, "global_step": 169747, "epoch": 1907, "lr": 3.091712451235539e-05}
{"train_loss": 0.0937279686331749, "global_step": 169748, "epoch": 1907, "lr": 3.0916588660883066e-05}
{"train_loss": 0.07183980941772461, "global_step": 169749, "epoch": 1907, "lr": 3.0916052811976275e-05}
{"train_loss": 0.07583523541688919, "global_step": 169750, "epoch": 1907, "lr": 3.091551696563509e-05}
{"train_loss": 0.11886255443096161, "global_step": 169751, "epoch": 1907, "lr": 3.0914981121859594e-05}
{"train_loss": 0.05511517450213432, "global_step": 169752, "epoch": 1907, "lr": 3.091444528064982e-05}
{"train_loss": 0.0747331753373146, "global_step": 169753, "epoch": 1907, "lr": 3.091390944200589e-05}
{"train_loss": 0.0479639396071434, "global_step": 169754, "epoch": 1907, "lr": 3.091337360592783e-05}
{"train_loss": 0.12835848331451416, "global_step": 169755, "epoch": 1907, "lr": 3.0912837772415756e-05}
{"train_loss": 0.11681382358074188, "global_step": 169756, "epoch": 1907, "lr": 3.09123019414697e-05}
{"train_loss": 0.05523964762687683, "global_step": 169757, "epoch": 1907, "lr": 3.0911766113089745e-05}
{"train_loss": 0.10054009407758713, "global_step": 169758, "epoch": 1907, "lr": 3.091123028727599e-05}
{"train_loss": 0.13125887513160706, "global_step": 169759, "epoch": 1907, "lr": 3.091069446402848e-05}
{"train_loss": 0.06475260853767395, "global_step": 169760, "epoch": 1907, "lr": 3.0910158643347306e-05}
{"train_loss": 0.07025812566280365, "global_step": 169761, "epoch": 1907, "lr": 3.090962282523251e-05}
{"train_loss": 0.08166155219078064, "global_step": 169762, "epoch": 1907, "lr": 3.09090870096842e-05}
{"train_loss": 0.1344795823097229, "global_step": 169763, "epoch": 1907, "lr": 3.090855119670242e-05}
{"train_loss": 0.058644410222768784, "global_step": 169764, "epoch": 1907, "lr": 3.0908015386287265e-05}
{"train_loss": 0.11667024344205856, "global_step": 169765, "epoch": 1907, "lr": 3.0907479578438784e-05}
{"train_loss": 0.08912724256515503, "global_step": 169766, "epoch": 1907, "lr": 3.090694377315706e-05}
{"train_loss": 0.08496486395597458, "global_step": 169767, "epoch": 1907, "lr": 3.090640797044216e-05}
{"train_loss": 0.08596673607826233, "global_step": 169768, "epoch": 1907, "lr": 3.090587217029417e-05}
{"train_loss": 0.06418367475271225, "global_step": 169769, "epoch": 1907, "lr": 3.090533637271315e-05}
{"train_loss": 0.0716252475976944, "global_step": 169770, "epoch": 1907, "lr": 3.090480057769919e-05}
{"train_loss": 0.09122064709663391, "global_step": 169771, "epoch": 1907, "lr": 3.090426478525231e-05}
{"train_loss": 0.08746284246444702, "global_step": 169772, "epoch": 1907, "lr": 3.090372899537265e-05}
{"train_loss": 0.14691218733787537, "global_step": 169773, "epoch": 1907, "lr": 3.090319320806023e-05}
{"train_loss": 0.11811341345310211, "global_step": 169774, "epoch": 1907, "lr": 3.0902657423315165e-05}
{"train_loss": 0.12865759432315826, "global_step": 169775, "epoch": 1907, "lr": 3.090212164113749e-05}
{"train_loss": 0.1014583632349968, "global_step": 169776, "epoch": 1907, "lr": 3.090158586152728e-05}
{"train_loss": 0.13174140453338623, "global_step": 169777, "epoch": 1907, "lr": 3.090105008448465e-05}
{"train_loss": 0.07822345197200775, "global_step": 169778, "epoch": 1907, "lr": 3.090051431000961e-05}
{"train_loss": 0.12686775624752045, "global_step": 169779, "epoch": 1907, "lr": 3.089997853810229e-05}
{"train_loss": 0.09750095009803772, "global_step": 169780, "epoch": 1907, "lr": 3.089944276876271e-05}
{"train_loss": 0.0743185356259346, "global_step": 169781, "epoch": 1907, "lr": 3.0898907001991e-05}
{"train_loss": 0.10182704776525497, "global_step": 169782, "epoch": 1907, "lr": 3.089837123778716e-05}
{"train_loss": 0.06404756009578705, "global_step": 169783, "epoch": 1907, "lr": 3.0897835476151335e-05}
{"train_loss": 0.13161814212799072, "global_step": 169784, "epoch": 1907, "lr": 3.089729971708355e-05}
{"train_loss": 0.0929444283246994, "global_step": 169785, "epoch": 1907, "lr": 3.0896763960583884e-05}
{"train_loss": 0.08703526854515076, "global_step": 169786, "epoch": 1907, "lr": 3.0896228206652425e-05}
{"train_loss": 0.0727502629160881, "global_step": 169787, "epoch": 1907, "lr": 3.089569245528923e-05}
{"train_loss": 0.08245019614696503, "global_step": 169788, "epoch": 1907, "lr": 3.0895156706494374e-05}
{"train_loss": 0.10681804269552231, "global_step": 169789, "epoch": 1907, "lr": 3.089462096026795e-05}
{"train_loss": 0.07808203995227814, "global_step": 169790, "epoch": 1907, "lr": 3.089408521660998e-05}
{"train_loss": 0.07722566276788712, "global_step": 169791, "epoch": 1907, "lr": 3.08935494755206e-05}
{"train_loss": 0.037553030997514725, "global_step": 169792, "epoch": 1907, "lr": 3.089301373699982e-05}
{"train_loss": 0.0635770857334137, "global_step": 169793, "epoch": 1907, "lr": 3.089247800104774e-05}
{"train_loss": 0.07799050956964493, "global_step": 169794, "epoch": 1907, "lr": 3.089194226766447e-05}
{"train_loss": 0.05953359976410866, "global_step": 169795, "epoch": 1907, "lr": 3.089140653685e-05}
{"train_loss": 0.08685808628797531, "global_step": 169796, "epoch": 1907, "lr": 3.089087080860448e-05}
{"train_loss": 0.1130155622959137, "global_step": 169797, "epoch": 1907, "lr": 3.0890335082927934e-05}
{"train_loss": 0.0468176007270813, "global_step": 169798, "epoch": 1907, "lr": 3.0889799359820464e-05}
{"train_loss": 0.06950920820236206, "global_step": 169799, "epoch": 1907, "lr": 3.0889263639282106e-05}
{"train_loss": 0.0855288878083229, "global_step": 169800, "epoch": 1907, "lr": 3.088872792131298e-05}
{"train_loss": 0.09160116314888, "global_step": 169801, "epoch": 1907, "lr": 3.0888192205913105e-05}
{"train_loss": 0.10251946747303009, "global_step": 169802, "epoch": 1907, "lr": 3.0887656493082604e-05}
{"train_loss": 0.1018892228603363, "global_step": 169803, "epoch": 1907, "lr": 3.088712078282151e-05}
{"train_loss": 0.07581255584955215, "global_step": 169804, "epoch": 1907, "lr": 3.088658507512992e-05}
{"train_loss": 0.063792884349823, "global_step": 169805, "epoch": 1907, "lr": 3.088604937000789e-05}
{"train_loss": 0.07770492136478424, "global_step": 169806, "epoch": 1907, "lr": 3.0885513667455494e-05}
{"train_loss": 0.09664876013994217, "global_step": 169807, "epoch": 1907, "lr": 3.088497796747281e-05}
{"train_loss": 0.06514196842908859, "global_step": 169808, "epoch": 1907, "lr": 3.088444227005991e-05}
{"train_loss": 0.08464736491441727, "global_step": 169809, "epoch": 1907, "lr": 3.0883906575216856e-05}
{"train_loss": 0.07766293734312057, "global_step": 169810, "epoch": 1907, "lr": 3.088337088294375e-05}
{"train_loss": 0.08770340917569198, "global_step": 169811, "epoch": 1907, "lr": 3.088283519324061e-05, "val_loss": 6.9858927726745605}
{"train_loss": 0.0780157819390297, "global_step": 169812, "epoch": 1908, "lr": 3.088229950610755e-05}
{"train_loss": 0.07841692119836807, "global_step": 169813, "epoch": 1908, "lr": 3.088176382154465e-05}
{"train_loss": 0.09383047372102737, "global_step": 169814, "epoch": 1908, "lr": 3.088122813955193e-05}
{"train_loss": 0.07836975157260895, "global_step": 169815, "epoch": 1908, "lr": 3.0880692460129536e-05}
{"train_loss": 0.0785747691988945, "global_step": 169816, "epoch": 1908, "lr": 3.088015678327747e-05}
{"train_loss": 0.09672209620475769, "global_step": 169817, "epoch": 1908, "lr": 3.0879621108995846e-05}
{"train_loss": 0.0735996663570404, "global_step": 169818, "epoch": 1908, "lr": 3.0879085437284716e-05}
{"train_loss": 0.0771317109465599, "global_step": 169819, "epoch": 1908, "lr": 3.0878549768144174e-05}
{"train_loss": 0.09042336791753769, "global_step": 169820, "epoch": 1908, "lr": 3.087801410157426e-05}
{"train_loss": 0.059976283460855484, "global_step": 169821, "epoch": 1908, "lr": 3.087747843757508e-05}
{"train_loss": 0.14773276448249817, "global_step": 169822, "epoch": 1908, "lr": 3.0876942776146675e-05}
{"train_loss": 0.10465673357248306, "global_step": 169823, "epoch": 1908, "lr": 3.0876407117289154e-05}
{"train_loss": 0.10051115602254868, "global_step": 169824, "epoch": 1908, "lr": 3.087587146100256e-05}
{"train_loss": 0.0712040364742279, "global_step": 169825, "epoch": 1908, "lr": 3.0875335807286964e-05}
{"train_loss": 0.06038933992385864, "global_step": 169826, "epoch": 1908, "lr": 3.0874800156142444e-05}
{"train_loss": 0.0598057359457016, "global_step": 169827, "epoch": 1908, "lr": 3.0874264507569076e-05}
{"train_loss": 0.07255852967500687, "global_step": 169828, "epoch": 1908, "lr": 3.0873728861566934e-05}
{"train_loss": 0.06002739816904068, "global_step": 169829, "epoch": 1908, "lr": 3.08731932181361e-05}
{"train_loss": 0.09552175551652908, "global_step": 169830, "epoch": 1908, "lr": 3.08726575772766e-05}
{"train_loss": 0.07859405130147934, "global_step": 169831, "epoch": 1908, "lr": 3.087212193898854e-05}
{"train_loss": 0.07374285161495209, "global_step": 169832, "epoch": 1908, "lr": 3.0871586303272015e-05}
{"train_loss": 0.07304397225379944, "global_step": 169833, "epoch": 1908, "lr": 3.087105067012705e-05}
{"train_loss": 0.08399646729230881, "global_step": 169834, "epoch": 1908, "lr": 3.0870515039553765e-05}
{"train_loss": 0.08313415199518204, "global_step": 169835, "epoch": 1908, "lr": 3.0869979411552184e-05}
{"train_loss": 0.08250942081212997, "global_step": 169836, "epoch": 1908, "lr": 3.086944378612241e-05}
{"train_loss": 0.04602355882525444, "global_step": 169837, "epoch": 1908, "lr": 3.0868908163264496e-05}
{"train_loss": 0.13795578479766846, "global_step": 169838, "epoch": 1908, "lr": 3.0868372542978545e-05}
{"train_loss": 0.09001438319683075, "global_step": 169839, "epoch": 1908, "lr": 3.0867836925264586e-05}
{"train_loss": 0.060750946402549744, "global_step": 169840, "epoch": 1908, "lr": 3.0867301310122734e-05}
{"train_loss": 0.06231914833188057, "global_step": 169841, "epoch": 1908, "lr": 3.086676569755302e-05}
{"train_loss": 0.025745732709765434, "global_step": 169842, "epoch": 1908, "lr": 3.086623008755556e-05}
{"train_loss": 0.10830004513263702, "global_step": 169843, "epoch": 1908, "lr": 3.086569448013038e-05}
{"train_loss": 0.10027510672807693, "global_step": 169844, "epoch": 1908, "lr": 3.086515887527758e-05}
{"train_loss": 0.07838588207960129, "global_step": 169845, "epoch": 1908, "lr": 3.086462327299723e-05}
{"train_loss": 0.11921315640211105, "global_step": 169846, "epoch": 1908, "lr": 3.08640876732894e-05}
{"train_loss": 0.04757702723145485, "global_step": 169847, "epoch": 1908, "lr": 3.086355207615415e-05}
{"train_loss": 0.08385733515024185, "global_step": 169848, "epoch": 1908, "lr": 3.086301648159156e-05}
{"train_loss": 0.13698714971542358, "global_step": 169849, "epoch": 1908, "lr": 3.086248088960172e-05}
{"train_loss": 0.11086380481719971, "global_step": 169850, "epoch": 1908, "lr": 3.0861945300184675e-05}
{"train_loss": 0.10971113294363022, "global_step": 169851, "epoch": 1908, "lr": 3.086140971334052e-05}
{"train_loss": 0.07347918301820755, "global_step": 169852, "epoch": 1908, "lr": 3.086087412906929e-05}
{"train_loss": 0.0823110044002533, "global_step": 169853, "epoch": 1908, "lr": 3.086033854737111e-05}
{"train_loss": 0.032774873077869415, "global_step": 169854, "epoch": 1908, "lr": 3.085980296824601e-05}
{"train_loss": 0.0631500855088234, "global_step": 169855, "epoch": 1908, "lr": 3.085926739169409e-05}
{"train_loss": 0.05556598678231239, "global_step": 169856, "epoch": 1908, "lr": 3.085873181771538e-05}
{"train_loss": 0.10792645066976547, "global_step": 169857, "epoch": 1908, "lr": 3.0858196246310016e-05}
{"train_loss": 0.11456480622291565, "global_step": 169858, "epoch": 1908, "lr": 3.0857660677478e-05}
{"train_loss": 0.05949248746037483, "global_step": 169859, "epoch": 1908, "lr": 3.085712511121947e-05}
{"train_loss": 0.045338280498981476, "global_step": 169860, "epoch": 1908, "lr": 3.0856589547534445e-05}
{"train_loss": 0.06941966712474823, "global_step": 169861, "epoch": 1908, "lr": 3.085605398642303e-05}
{"train_loss": 0.04508005827665329, "global_step": 169862, "epoch": 1908, "lr": 3.085551842788528e-05}
{"train_loss": 0.18344387412071228, "global_step": 169863, "epoch": 1908, "lr": 3.085498287192127e-05}
{"train_loss": 0.08322709798812866, "global_step": 169864, "epoch": 1908, "lr": 3.085444731853108e-05}
{"train_loss": 0.07281727343797684, "global_step": 169865, "epoch": 1908, "lr": 3.085391176771477e-05}
{"train_loss": 0.05432066321372986, "global_step": 169866, "epoch": 1908, "lr": 3.0853376219472416e-05}
{"train_loss": 0.061833612620830536, "global_step": 169867, "epoch": 1908, "lr": 3.0852840673804084e-05}
{"train_loss": 0.09585890173912048, "global_step": 169868, "epoch": 1908, "lr": 3.085230513070988e-05}
{"train_loss": 0.05111442506313324, "global_step": 169869, "epoch": 1908, "lr": 3.085176959018983e-05}
{"train_loss": 0.07414232939481735, "global_step": 169870, "epoch": 1908, "lr": 3.085123405224404e-05}
{"train_loss": 0.0885406956076622, "global_step": 169871, "epoch": 1908, "lr": 3.085069851687255e-05}
{"train_loss": 0.08100741356611252, "global_step": 169872, "epoch": 1908, "lr": 3.085016298407547e-05}
{"train_loss": 0.07517008483409882, "global_step": 169873, "epoch": 1908, "lr": 3.0849627453852834e-05}
{"train_loss": 0.10174166411161423, "global_step": 169874, "epoch": 1908, "lr": 3.0849091926204756e-05}
{"train_loss": 0.06927147507667542, "global_step": 169875, "epoch": 1908, "lr": 3.084855640113126e-05}
{"train_loss": 0.11277597397565842, "global_step": 169876, "epoch": 1908, "lr": 3.0848020878632455e-05}
{"train_loss": 0.08334188908338547, "global_step": 169877, "epoch": 1908, "lr": 3.084748535870839e-05}
{"train_loss": 0.02870357781648636, "global_step": 169878, "epoch": 1908, "lr": 3.084694984135916e-05}
{"train_loss": 0.053960319608449936, "global_step": 169879, "epoch": 1908, "lr": 3.084641432658481e-05}
{"train_loss": 0.08298709988594055, "global_step": 169880, "epoch": 1908, "lr": 3.084587881438544e-05}
{"train_loss": 0.11813604086637497, "global_step": 169881, "epoch": 1908, "lr": 3.084534330476111e-05}
{"train_loss": 0.08117062598466873, "global_step": 169882, "epoch": 1908, "lr": 3.0844807797711885e-05}
{"train_loss": 0.11982592940330505, "global_step": 169883, "epoch": 1908, "lr": 3.0844272293237835e-05}
{"train_loss": 0.08583501726388931, "global_step": 169884, "epoch": 1908, "lr": 3.0843736791339036e-05}
{"train_loss": 0.04971888288855553, "global_step": 169885, "epoch": 1908, "lr": 3.084320129201558e-05}
{"train_loss": 0.08714265376329422, "global_step": 169886, "epoch": 1908, "lr": 3.08426657952675e-05}
{"train_loss": 0.13186468183994293, "global_step": 169887, "epoch": 1908, "lr": 3.084213030109492e-05}
{"train_loss": 0.08920576423406601, "global_step": 169888, "epoch": 1908, "lr": 3.084159480949785e-05}
{"train_loss": 0.06108605116605759, "global_step": 169889, "epoch": 1908, "lr": 3.084105932047642e-05}
{"train_loss": 0.09469768404960632, "global_step": 169890, "epoch": 1908, "lr": 3.084052383403065e-05}
{"train_loss": 0.03371499478816986, "global_step": 169891, "epoch": 1908, "lr": 3.083998835016067e-05}
{"train_loss": 0.06030691787600517, "global_step": 169892, "epoch": 1908, "lr": 3.083945286886649e-05}
{"train_loss": 0.056033164262771606, "global_step": 169893, "epoch": 1908, "lr": 3.0838917390148234e-05}
{"train_loss": 0.07799064368009567, "global_step": 169894, "epoch": 1908, "lr": 3.0838381914005946e-05}
{"train_loss": 0.07275673002004623, "global_step": 169895, "epoch": 1908, "lr": 3.083784644043971e-05}
{"train_loss": 0.06733841449022293, "global_step": 169896, "epoch": 1908, "lr": 3.083731096944957e-05}
{"train_loss": 0.0884072333574295, "global_step": 169897, "epoch": 1908, "lr": 3.083677550103564e-05}
{"train_loss": 0.0885479748249054, "global_step": 169898, "epoch": 1908, "lr": 3.0836240035197964e-05}
{"train_loss": 0.04611789807677269, "global_step": 169899, "epoch": 1908, "lr": 3.083570457193663e-05}
{"train_loss": 0.08044347005864877, "global_step": 169900, "epoch": 1908, "lr": 3.0835169111251694e-05, "val_loss": 7.184210777282715}
{"train_loss": 0.0545215979218483, "global_step": 169901, "epoch": 1909, "lr": 3.083463365314324e-05}
{"train_loss": 0.08024805039167404, "global_step": 169902, "epoch": 1909, "lr": 3.083409819761133e-05}
{"train_loss": 0.08021588623523712, "global_step": 169903, "epoch": 1909, "lr": 3.083356274465603e-05}
{"train_loss": 0.08067446947097778, "global_step": 169904, "epoch": 1909, "lr": 3.083302729427745e-05}
{"train_loss": 0.04997628927230835, "global_step": 169905, "epoch": 1909, "lr": 3.083249184647562e-05}
{"train_loss": 0.05242157727479935, "global_step": 169906, "epoch": 1909, "lr": 3.083195640125064e-05}
{"train_loss": 0.1267702728509903, "global_step": 169907, "epoch": 1909, "lr": 3.0831420958602554e-05}
{"train_loss": 0.08566395938396454, "global_step": 169908, "epoch": 1909, "lr": 3.083088551853147e-05}
{"train_loss": 0.09027859568595886, "global_step": 169909, "epoch": 1909, "lr": 3.0830350081037416e-05}
{"train_loss": 0.04930557683110237, "global_step": 169910, "epoch": 1909, "lr": 3.0829814646120506e-05}
{"train_loss": 0.03397294506430626, "global_step": 169911, "epoch": 1909, "lr": 3.0829279213780784e-05}
{"train_loss": 0.1173233836889267, "global_step": 169912, "epoch": 1909, "lr": 3.082874378401834e-05}
{"train_loss": 0.13054674863815308, "global_step": 169913, "epoch": 1909, "lr": 3.082820835683322e-05}
{"train_loss": 0.11713499575853348, "global_step": 169914, "epoch": 1909, "lr": 3.082767293222554e-05}
{"train_loss": 0.07091940939426422, "global_step": 169915, "epoch": 1909, "lr": 3.082713751019532e-05}
{"train_loss": 0.05147305876016617, "global_step": 169916, "epoch": 1909, "lr": 3.0826602090742686e-05}
{"train_loss": 0.09883053600788116, "global_step": 169917, "epoch": 1909, "lr": 3.082606667386765e-05}
{"train_loss": 0.12532323598861694, "global_step": 169918, "epoch": 1909, "lr": 3.082553125957035e-05}
{"train_loss": 0.0541037879884243, "global_step": 169919, "epoch": 1909, "lr": 3.08249958478508e-05}
{"train_loss": 0.04880310967564583, "global_step": 169920, "epoch": 1909, "lr": 3.0824460438709094e-05}
{"train_loss": 0.11101128160953522, "global_step": 169921, "epoch": 1909, "lr": 3.082392503214532e-05}
{"train_loss": 0.07699299603700638, "global_step": 169922, "epoch": 1909, "lr": 3.0823389628159515e-05}
{"train_loss": 0.13116499781608582, "global_step": 169923, "epoch": 1909, "lr": 3.08228542267518e-05}
{"train_loss": 0.10367180407047272, "global_step": 169924, "epoch": 1909, "lr": 3.082231882792219e-05}
{"train_loss": 0.05024755746126175, "global_step": 169925, "epoch": 1909, "lr": 3.0821783431670816e-05}
{"train_loss": 0.06733334809541702, "global_step": 169926, "epoch": 1909, "lr": 3.082124803799769e-05}
{"train_loss": 0.1183624267578125, "global_step": 169927, "epoch": 1909, "lr": 3.0820712646902935e-05}
{"train_loss": 0.10997836291790009, "global_step": 169928, "epoch": 1909, "lr": 3.082017725838658e-05}
{"train_loss": 0.04427662864327431, "global_step": 169929, "epoch": 1909, "lr": 3.081964187244874e-05}
{"train_loss": 0.10432775318622589, "global_step": 169930, "epoch": 1909, "lr": 3.081910648908944e-05}
{"train_loss": 0.09211760759353638, "global_step": 169931, "epoch": 1909, "lr": 3.0818571108308806e-05}
{"train_loss": 0.08544234931468964, "global_step": 169932, "epoch": 1909, "lr": 3.081803573010686e-05}
{"train_loss": 0.05025488883256912, "global_step": 169933, "epoch": 1909, "lr": 3.0817500354483716e-05}
{"train_loss": 0.023130325600504875, "global_step": 169934, "epoch": 1909, "lr": 3.08169649814394e-05}
{"train_loss": 0.05145098268985748, "global_step": 169935, "epoch": 1909, "lr": 3.081642961097403e-05}
{"train_loss": 0.11810018122196198, "global_step": 169936, "epoch": 1909, "lr": 3.081589424308764e-05}
{"train_loss": 0.06303680688142776, "global_step": 169937, "epoch": 1909, "lr": 3.081535887778034e-05}
{"train_loss": 0.09850174188613892, "global_step": 169938, "epoch": 1909, "lr": 3.081482351505216e-05}
{"train_loss": 0.08767068386077881, "global_step": 169939, "epoch": 1909, "lr": 3.0814288154903205e-05}
{"train_loss": 0.06276597827672958, "global_step": 169940, "epoch": 1909, "lr": 3.081375279733354e-05}
{"train_loss": 0.07666455954313278, "global_step": 169941, "epoch": 1909, "lr": 3.081321744234321e-05}
{"train_loss": 0.09366945177316666, "global_step": 169942, "epoch": 1909, "lr": 3.081268208993234e-05}
{"train_loss": 0.057750340551137924, "global_step": 169943, "epoch": 1909, "lr": 3.081214674010094e-05}
{"train_loss": 0.06463263928890228, "global_step": 169944, "epoch": 1909, "lr": 3.081161139284914e-05}
{"train_loss": 0.14243802428245544, "global_step": 169945, "epoch": 1909, "lr": 3.0811076048176966e-05}
{"train_loss": 0.10572710633277893, "global_step": 169946, "epoch": 1909, "lr": 3.0810540706084526e-05}
{"train_loss": 0.0671776533126831, "global_step": 169947, "epoch": 1909, "lr": 3.0810005366571856e-05}
{"train_loss": 0.11678241193294525, "global_step": 169948, "epoch": 1909, "lr": 3.080947002963907e-05}
{"train_loss": 0.07887721061706543, "global_step": 169949, "epoch": 1909, "lr": 3.0808934695286184e-05}
{"train_loss": 0.045280516147613525, "global_step": 169950, "epoch": 1909, "lr": 3.0808399363513336e-05}
{"train_loss": 0.032745908945798874, "global_step": 169951, "epoch": 1909, "lr": 3.0807864034320544e-05}
{"train_loss": 0.07170076668262482, "global_step": 169952, "epoch": 1909, "lr": 3.0807328707707915e-05}
{"train_loss": 0.06512366235256195, "global_step": 169953, "epoch": 1909, "lr": 3.0806793383675485e-05}
{"train_loss": 0.06718740612268448, "global_step": 169954, "epoch": 1909, "lr": 3.080625806222337e-05}
{"train_loss": 0.11747705936431885, "global_step": 169955, "epoch": 1909, "lr": 3.0805722743351604e-05}
{"train_loss": 0.08580955117940903, "global_step": 169956, "epoch": 1909, "lr": 3.080518742706029e-05}
{"train_loss": 0.10597977042198181, "global_step": 169957, "epoch": 1909, "lr": 3.080465211334947e-05}
{"train_loss": 0.08146093040704727, "global_step": 169958, "epoch": 1909, "lr": 3.0804116802219235e-05}
{"train_loss": 0.09575556218624115, "global_step": 169959, "epoch": 1909, "lr": 3.0803581493669666e-05}
{"train_loss": 0.0748409554362297, "global_step": 169960, "epoch": 1909, "lr": 3.08030461877008e-05}
{"train_loss": 0.04160701110959053, "global_step": 169961, "epoch": 1909, "lr": 3.0802510884312754e-05}
{"train_loss": 0.09032116830348969, "global_step": 169962, "epoch": 1909, "lr": 3.0801975583505547e-05}
{"train_loss": 0.05824142321944237, "global_step": 169963, "epoch": 1909, "lr": 3.080144028527931e-05}
{"train_loss": 0.07180581241846085, "global_step": 169964, "epoch": 1909, "lr": 3.0800904989634064e-05}
{"train_loss": 0.05243955925107002, "global_step": 169965, "epoch": 1909, "lr": 3.080036969656992e-05}
{"train_loss": 0.09617023169994354, "global_step": 169966, "epoch": 1909, "lr": 3.079983440608691e-05}
{"train_loss": 0.08281011134386063, "global_step": 169967, "epoch": 1909, "lr": 3.079929911818515e-05}
{"train_loss": 0.09446772187948227, "global_step": 169968, "epoch": 1909, "lr": 3.079876383286467e-05}
{"train_loss": 0.09568548202514648, "global_step": 169969, "epoch": 1909, "lr": 3.079822855012559e-05}
{"train_loss": 0.1321629136800766, "global_step": 169970, "epoch": 1909, "lr": 3.079769326996792e-05}
{"train_loss": 0.11869706213474274, "global_step": 169971, "epoch": 1909, "lr": 3.079715799239179e-05}
{"train_loss": 0.051677361130714417, "global_step": 169972, "epoch": 1909, "lr": 3.0796622717397234e-05}
{"train_loss": 0.123264841735363, "global_step": 169973, "epoch": 1909, "lr": 3.0796087444984353e-05}
{"train_loss": 0.10429079830646515, "global_step": 169974, "epoch": 1909, "lr": 3.079555217515319e-05}
{"train_loss": 0.10033638030290604, "global_step": 169975, "epoch": 1909, "lr": 3.0795016907903835e-05}
{"train_loss": 0.03875237703323364, "global_step": 169976, "epoch": 1909, "lr": 3.079448164323636e-05}
{"train_loss": 0.05005834996700287, "global_step": 169977, "epoch": 1909, "lr": 3.079394638115083e-05}
{"train_loss": 0.07703942060470581, "global_step": 169978, "epoch": 1909, "lr": 3.0793411121647334e-05}
{"train_loss": 0.10522256791591644, "global_step": 169979, "epoch": 1909, "lr": 3.07928758647259e-05}
{"train_loss": 0.06827087700366974, "global_step": 169980, "epoch": 1909, "lr": 3.079234061038666e-05}
{"train_loss": 0.0963030755519867, "global_step": 169981, "epoch": 1909, "lr": 3.0791805358629636e-05}
{"train_loss": 0.06299746036529541, "global_step": 169982, "epoch": 1909, "lr": 3.079127010945493e-05}
{"train_loss": 0.0777532085776329, "global_step": 169983, "epoch": 1909, "lr": 3.079073486286259e-05}
{"train_loss": 0.04169345274567604, "global_step": 169984, "epoch": 1909, "lr": 3.079019961885272e-05}
{"train_loss": 0.1155477985739708, "global_step": 169985, "epoch": 1909, "lr": 3.0789664377425355e-05}
{"train_loss": 0.0754886195063591, "global_step": 169986, "epoch": 1909, "lr": 3.07891291385806e-05}
{"train_loss": 0.1027659922838211, "global_step": 169987, "epoch": 1909, "lr": 3.0788593902318496e-05}
{"train_loss": 0.05226423218846321, "global_step": 169988, "epoch": 1909, "lr": 3.078805866863916e-05}
{"train_loss": 0.08083660662006796, "global_step": 169989, "epoch": 1909, "lr": 3.0787523437542607e-05, "val_loss": 7.2018961906433105}
{"train_loss": 0.1260477900505066, "global_step": 169990, "epoch": 1910, "lr": 3.078698820902896e-05}
{"train_loss": 0.11504437774419785, "global_step": 169991, "epoch": 1910, "lr": 3.078645298309825e-05}
{"train_loss": 0.06054741144180298, "global_step": 169992, "epoch": 1910, "lr": 3.078591775975058e-05}
{"train_loss": 0.0865851640701294, "global_step": 169993, "epoch": 1910, "lr": 3.078538253898599e-05}
{"train_loss": 0.08116810023784637, "global_step": 169994, "epoch": 1910, "lr": 3.078484732080458e-05}
{"train_loss": 0.031666528433561325, "global_step": 169995, "epoch": 1910, "lr": 3.078431210520642e-05}
{"train_loss": 0.05139034613966942, "global_step": 169996, "epoch": 1910, "lr": 3.0783776892191576e-05}
{"train_loss": 0.07772958278656006, "global_step": 169997, "epoch": 1910, "lr": 3.078324168176012e-05}
{"train_loss": 0.057766254991292953, "global_step": 169998, "epoch": 1910, "lr": 3.078270647391211e-05}
{"train_loss": 0.055160041898489, "global_step": 169999, "epoch": 1910, "lr": 3.078217126864765e-05}
{"train_loss": 0.06961773335933685, "global_step": 170000, "epoch": 1910, "lr": 3.0781636065966765e-05}
{"train_loss": 0.06825721263885498, "global_step": 170001, "epoch": 1910, "lr": 3.078110086586958e-05}
{"train_loss": 0.07105150818824768, "global_step": 170002, "epoch": 1910, "lr": 3.078056566835613e-05}
{"train_loss": 0.09359239041805267, "global_step": 170003, "epoch": 1910, "lr": 3.078003047342652e-05}
{"train_loss": 0.05609256029129028, "global_step": 170004, "epoch": 1910, "lr": 3.077949528108076e-05}
{"train_loss": 0.10259237140417099, "global_step": 170005, "epoch": 1910, "lr": 3.0778960091318997e-05}
{"train_loss": 0.11815519630908966, "global_step": 170006, "epoch": 1910, "lr": 3.077842490414125e-05}
{"train_loss": 0.05752198025584221, "global_step": 170007, "epoch": 1910, "lr": 3.077788971954762e-05}
{"train_loss": 0.07106345891952515, "global_step": 170008, "epoch": 1910, "lr": 3.077735453753816e-05}
{"train_loss": 0.07718872278928757, "global_step": 170009, "epoch": 1910, "lr": 3.0776819358112965e-05}
{"train_loss": 0.07459200173616409, "global_step": 170010, "epoch": 1910, "lr": 3.077628418127207e-05}
{"train_loss": 0.04076714068651199, "global_step": 170011, "epoch": 1910, "lr": 3.0775749007015586e-05}
{"train_loss": 0.07848997414112091, "global_step": 170012, "epoch": 1910, "lr": 3.0775213835343566e-05}
{"train_loss": 0.11419368535280228, "global_step": 170013, "epoch": 1910, "lr": 3.077467866625608e-05}
{"train_loss": 0.08330008387565613, "global_step": 170014, "epoch": 1910, "lr": 3.0774143499753205e-05}
{"train_loss": 0.09326672554016113, "global_step": 170015, "epoch": 1910, "lr": 3.077360833583501e-05}
{"train_loss": 0.09863734990358353, "global_step": 170016, "epoch": 1910, "lr": 3.0773073174501585e-05}
{"train_loss": 0.03647662326693535, "global_step": 170017, "epoch": 1910, "lr": 3.0772538015752964e-05}
{"train_loss": 0.07739181816577911, "global_step": 170018, "epoch": 1910, "lr": 3.0772002859589264e-05}
{"train_loss": 0.07671815156936646, "global_step": 170019, "epoch": 1910, "lr": 3.0771467706010505e-05}
{"train_loss": 0.054682034999132156, "global_step": 170020, "epoch": 1910, "lr": 3.0770932555016815e-05}
{"train_loss": 0.08945154398679733, "global_step": 170021, "epoch": 1910, "lr": 3.077039740660822e-05}
{"train_loss": 0.07054776698350906, "global_step": 170022, "epoch": 1910, "lr": 3.076986226078482e-05}
{"train_loss": 0.07371141761541367, "global_step": 170023, "epoch": 1910, "lr": 3.076932711754667e-05}
{"train_loss": 0.06161130592226982, "global_step": 170024, "epoch": 1910, "lr": 3.0768791976893863e-05}
{"train_loss": 0.11587931960821152, "global_step": 170025, "epoch": 1910, "lr": 3.076825683882644e-05}
{"train_loss": 0.13129562139511108, "global_step": 170026, "epoch": 1910, "lr": 3.076772170334451e-05}
{"train_loss": 0.07159218192100525, "global_step": 170027, "epoch": 1910, "lr": 3.07671865704481e-05}
{"train_loss": 0.09109758585691452, "global_step": 170028, "epoch": 1910, "lr": 3.076665144013734e-05}
{"train_loss": 0.06039527431130409, "global_step": 170029, "epoch": 1910, "lr": 3.076611631241224e-05}
{"train_loss": 0.08160339295864105, "global_step": 170030, "epoch": 1910, "lr": 3.076558118727291e-05}
{"train_loss": 0.058039162307977676, "global_step": 170031, "epoch": 1910, "lr": 3.0765046064719416e-05}
{"train_loss": 0.11300192773342133, "global_step": 170032, "epoch": 1910, "lr": 3.0764510944751825e-05}
{"train_loss": 0.11752191185951233, "global_step": 170033, "epoch": 1910, "lr": 3.076397582737022e-05}
{"train_loss": 0.11188580095767975, "global_step": 170034, "epoch": 1910, "lr": 3.076344071257465e-05}
{"train_loss": 0.039638884365558624, "global_step": 170035, "epoch": 1910, "lr": 3.076290560036522e-05}
{"train_loss": 0.04996609315276146, "global_step": 170036, "epoch": 1910, "lr": 3.076237049074197e-05}
{"train_loss": 0.12348902225494385, "global_step": 170037, "epoch": 1910, "lr": 3.076183538370499e-05}
{"train_loss": 0.08021427690982819, "global_step": 170038, "epoch": 1910, "lr": 3.0761300279254335e-05}
{"train_loss": 0.13728339970111847, "global_step": 170039, "epoch": 1910, "lr": 3.076076517739011e-05}
{"train_loss": 0.07988506555557251, "global_step": 170040, "epoch": 1910, "lr": 3.0760230078112334e-05}
{"train_loss": 0.050227031111717224, "global_step": 170041, "epoch": 1910, "lr": 3.075969498142114e-05}
{"train_loss": 0.07709643244743347, "global_step": 170042, "epoch": 1910, "lr": 3.075915988731655e-05}
{"train_loss": 0.16909073293209076, "global_step": 170043, "epoch": 1910, "lr": 3.0758624795798675e-05}
{"train_loss": 0.06410505622625351, "global_step": 170044, "epoch": 1910, "lr": 3.0758089706867544e-05}
{"train_loss": 0.18859423696994781, "global_step": 170045, "epoch": 1910, "lr": 3.0757554620523285e-05}
{"train_loss": 0.11613480001688004, "global_step": 170046, "epoch": 1910, "lr": 3.0757019536765906e-05}
{"train_loss": 0.08102500438690186, "global_step": 170047, "epoch": 1910, "lr": 3.075648445559552e-05}
{"train_loss": 0.020586051046848297, "global_step": 170048, "epoch": 1910, "lr": 3.0755949377012205e-05}
{"train_loss": 0.08235232532024384, "global_step": 170049, "epoch": 1910, "lr": 3.0755414301016006e-05}
{"train_loss": 0.10014291852712631, "global_step": 170050, "epoch": 1910, "lr": 3.075487922760701e-05}
{"train_loss": 0.046846434473991394, "global_step": 170051, "epoch": 1910, "lr": 3.075434415678529e-05}
{"train_loss": 0.06361925601959229, "global_step": 170052, "epoch": 1910, "lr": 3.075380908855091e-05}
{"train_loss": 0.0834294855594635, "global_step": 170053, "epoch": 1910, "lr": 3.0753274022903945e-05}
{"train_loss": 0.08785144239664078, "global_step": 170054, "epoch": 1910, "lr": 3.075273895984447e-05}
{"train_loss": 0.09903647750616074, "global_step": 170055, "epoch": 1910, "lr": 3.075220389937256e-05}
{"train_loss": 0.05817495658993721, "global_step": 170056, "epoch": 1910, "lr": 3.075166884148828e-05}
{"train_loss": 0.07567747682332993, "global_step": 170057, "epoch": 1910, "lr": 3.075113378619169e-05}
{"train_loss": 0.07071416825056076, "global_step": 170058, "epoch": 1910, "lr": 3.075059873348289e-05}
{"train_loss": 0.08144272118806839, "global_step": 170059, "epoch": 1910, "lr": 3.0750063683361924e-05}
{"train_loss": 0.1029590368270874, "global_step": 170060, "epoch": 1910, "lr": 3.07495286358289e-05}
{"train_loss": 0.08754155039787292, "global_step": 170061, "epoch": 1910, "lr": 3.0748993590883835e-05}
{"train_loss": 0.09228982031345367, "global_step": 170062, "epoch": 1910, "lr": 3.0748458548526865e-05}
{"train_loss": 0.04885280877351761, "global_step": 170063, "epoch": 1910, "lr": 3.0747923508758e-05}
{"train_loss": 0.0742899626493454, "global_step": 170064, "epoch": 1910, "lr": 3.074738847157737e-05}
{"train_loss": 0.06690807640552521, "global_step": 170065, "epoch": 1910, "lr": 3.0746853436985e-05}
{"train_loss": 0.0450301356613636, "global_step": 170066, "epoch": 1910, "lr": 3.074631840498098e-05}
{"train_loss": 0.04267663508653641, "global_step": 170067, "epoch": 1910, "lr": 3.0745783375565396e-05}
{"train_loss": 0.05629091337323189, "global_step": 170068, "epoch": 1910, "lr": 3.074524834873829e-05}
{"train_loss": 0.14284774661064148, "global_step": 170069, "epoch": 1910, "lr": 3.074471332449977e-05}
{"train_loss": 0.09449058771133423, "global_step": 170070, "epoch": 1910, "lr": 3.074417830284988e-05}
{"train_loss": 0.11363105475902557, "global_step": 170071, "epoch": 1910, "lr": 3.0743643283788705e-05}
{"train_loss": 0.18032066524028778, "global_step": 170072, "epoch": 1910, "lr": 3.074310826731631e-05}
{"train_loss": 0.06861594319343567, "global_step": 170073, "epoch": 1910, "lr": 3.074257325343276e-05}
{"train_loss": 0.058089423924684525, "global_step": 170074, "epoch": 1910, "lr": 3.0742038242138146e-05}
{"train_loss": 0.11163056641817093, "global_step": 170075, "epoch": 1910, "lr": 3.074150323343254e-05}
{"train_loss": 0.10034088045358658, "global_step": 170076, "epoch": 1910, "lr": 3.074096822731598e-05}
{"train_loss": 0.0699426680803299, "global_step": 170077, "epoch": 1910, "lr": 3.074043322378859e-05}
{"train_loss": 0.08287803548273076, "global_step": 170078, "epoch": 1910, "lr": 3.0739898222850385e-05, "val_loss": 7.206421852111816, "train_action_mse_error": 26.644962310791016}
{"train_loss": 0.18363140523433685, "global_step": 170079, "epoch": 1911, "lr": 3.0739363224501486e-05}
{"train_loss": 0.08097384870052338, "global_step": 170080, "epoch": 1911, "lr": 3.073882822874193e-05}
{"train_loss": 0.05460560321807861, "global_step": 170081, "epoch": 1911, "lr": 3.073829323557183e-05}
{"train_loss": 0.10277697443962097, "global_step": 170082, "epoch": 1911, "lr": 3.07377582449912e-05}
{"train_loss": 0.11038763076066971, "global_step": 170083, "epoch": 1911, "lr": 3.073722325700016e-05}
{"train_loss": 0.08425484597682953, "global_step": 170084, "epoch": 1911, "lr": 3.073668827159876e-05}
{"train_loss": 0.07368940860033035, "global_step": 170085, "epoch": 1911, "lr": 3.0736153288787064e-05}
{"train_loss": 0.05635529011487961, "global_step": 170086, "epoch": 1911, "lr": 3.0735618308565185e-05}
{"train_loss": 0.10207010060548782, "global_step": 170087, "epoch": 1911, "lr": 3.073508333093315e-05}
{"train_loss": 0.0852673202753067, "global_step": 170088, "epoch": 1911, "lr": 3.073454835589106e-05}
{"train_loss": 0.10490541905164719, "global_step": 170089, "epoch": 1911, "lr": 3.073401338343897e-05}
{"train_loss": 0.039726465940475464, "global_step": 170090, "epoch": 1911, "lr": 3.0733478413576955e-05}
{"train_loss": 0.09118364751338959, "global_step": 170091, "epoch": 1911, "lr": 3.07329434463051e-05}
{"train_loss": 0.06948346644639969, "global_step": 170092, "epoch": 1911, "lr": 3.073240848162345e-05}
{"train_loss": 0.0768287405371666, "global_step": 170093, "epoch": 1911, "lr": 3.073187351953209e-05}
{"train_loss": 0.08360278606414795, "global_step": 170094, "epoch": 1911, "lr": 3.073133856003112e-05}
{"train_loss": 0.07179258018732071, "global_step": 170095, "epoch": 1911, "lr": 3.073080360312056e-05}
{"train_loss": 0.10592535883188248, "global_step": 170096, "epoch": 1911, "lr": 3.073026864880053e-05}
{"train_loss": 0.057604338973760605, "global_step": 170097, "epoch": 1911, "lr": 3.072973369707106e-05}
{"train_loss": 0.052358485758304596, "global_step": 170098, "epoch": 1911, "lr": 3.0729198747932266e-05}
{"train_loss": 0.09101493656635284, "global_step": 170099, "epoch": 1911, "lr": 3.0728663801384175e-05}
{"train_loss": 0.06507690250873566, "global_step": 170100, "epoch": 1911, "lr": 3.072812885742689e-05}
{"train_loss": 0.09845985472202301, "global_step": 170101, "epoch": 1911, "lr": 3.072759391606046e-05}
{"train_loss": 0.028975535184144974, "global_step": 170102, "epoch": 1911, "lr": 3.0727058977284976e-05}
{"train_loss": 0.09023477882146835, "global_step": 170103, "epoch": 1911, "lr": 3.0726524041100516e-05}
{"train_loss": 0.056967202574014664, "global_step": 170104, "epoch": 1911, "lr": 3.072598910750712e-05}
{"train_loss": 0.0411713570356369, "global_step": 170105, "epoch": 1911, "lr": 3.072545417650491e-05}
{"train_loss": 0.06252247840166092, "global_step": 170106, "epoch": 1911, "lr": 3.07249192480939e-05}
{"train_loss": 0.10398579388856888, "global_step": 170107, "epoch": 1911, "lr": 3.072438432227421e-05}
{"train_loss": 0.11586903780698776, "global_step": 170108, "epoch": 1911, "lr": 3.072384939904588e-05}
{"train_loss": 0.0893024355173111, "global_step": 170109, "epoch": 1911, "lr": 3.0723314478408995e-05}
{"train_loss": 0.08273501694202423, "global_step": 170110, "epoch": 1911, "lr": 3.072277956036363e-05}
{"train_loss": 0.09066580981016159, "global_step": 170111, "epoch": 1911, "lr": 3.072224464490985e-05}
{"train_loss": 0.15144728124141693, "global_step": 170112, "epoch": 1911, "lr": 3.072170973204772e-05}
{"train_loss": 0.12537844479084015, "global_step": 170113, "epoch": 1911, "lr": 3.072117482177734e-05}
{"train_loss": 0.08241813629865646, "global_step": 170114, "epoch": 1911, "lr": 3.0720639914098745e-05}
{"train_loss": 0.07243456691503525, "global_step": 170115, "epoch": 1911, "lr": 3.072010500901205e-05}
{"train_loss": 0.07954344898462296, "global_step": 170116, "epoch": 1911, "lr": 3.071957010651727e-05}
{"train_loss": 0.07765521854162216, "global_step": 170117, "epoch": 1911, "lr": 3.071903520661453e-05}
{"train_loss": 0.08374568074941635, "global_step": 170118, "epoch": 1911, "lr": 3.071850030930386e-05}
{"train_loss": 0.04016856849193573, "global_step": 170119, "epoch": 1911, "lr": 3.0717965414585376e-05}
{"train_loss": 0.11525128781795502, "global_step": 170120, "epoch": 1911, "lr": 3.0717430522459104e-05}
{"train_loss": 0.09449807554483414, "global_step": 170121, "epoch": 1911, "lr": 3.0716895632925146e-05}
{"train_loss": 0.13160142302513123, "global_step": 170122, "epoch": 1911, "lr": 3.071636074598358e-05}
{"train_loss": 0.08069881051778793, "global_step": 170123, "epoch": 1911, "lr": 3.0715825861634437e-05}
{"train_loss": 0.08785638213157654, "global_step": 170124, "epoch": 1911, "lr": 3.0715290979877854e-05}
{"train_loss": 0.042368099093437195, "global_step": 170125, "epoch": 1911, "lr": 3.071475610071383e-05}
{"train_loss": 0.07297533005475998, "global_step": 170126, "epoch": 1911, "lr": 3.0714221224142505e-05}
{"train_loss": 0.042896222323179245, "global_step": 170127, "epoch": 1911, "lr": 3.0713686350163893e-05}
{"train_loss": 0.043854158371686935, "global_step": 170128, "epoch": 1911, "lr": 3.0713151478778104e-05}
{"train_loss": 0.10413509607315063, "global_step": 170129, "epoch": 1911, "lr": 3.071261660998519e-05}
{"train_loss": 0.10354378819465637, "global_step": 170130, "epoch": 1911, "lr": 3.071208174378522e-05}
{"train_loss": 0.11106114089488983, "global_step": 170131, "epoch": 1911, "lr": 3.071154688017829e-05}
{"train_loss": 0.13928523659706116, "global_step": 170132, "epoch": 1911, "lr": 3.071101201916446e-05}
{"train_loss": 0.06600373238325119, "global_step": 170133, "epoch": 1911, "lr": 3.071047716074379e-05}
{"train_loss": 0.07914524525403976, "global_step": 170134, "epoch": 1911, "lr": 3.0709942304916374e-05}
{"train_loss": 0.08786880970001221, "global_step": 170135, "epoch": 1911, "lr": 3.070940745168225e-05}
{"train_loss": 0.06299149245023727, "global_step": 170136, "epoch": 1911, "lr": 3.070887260104154e-05}
{"train_loss": 0.06797093898057938, "global_step": 170137, "epoch": 1911, "lr": 3.070833775299425e-05}
{"train_loss": 0.06701010465621948, "global_step": 170138, "epoch": 1911, "lr": 3.07078029075405e-05}
{"train_loss": 0.05633281543850899, "global_step": 170139, "epoch": 1911, "lr": 3.070726806468037e-05}
{"train_loss": 0.05470835790038109, "global_step": 170140, "epoch": 1911, "lr": 3.070673322441389e-05}
{"train_loss": 0.1296483278274536, "global_step": 170141, "epoch": 1911, "lr": 3.070619838674118e-05}
{"train_loss": 0.054460324347019196, "global_step": 170142, "epoch": 1911, "lr": 3.0705663551662264e-05}
{"train_loss": 0.07518119364976883, "global_step": 170143, "epoch": 1911, "lr": 3.070512871917725e-05}
{"train_loss": 0.08608157932758331, "global_step": 170144, "epoch": 1911, "lr": 3.070459388928618e-05}
{"train_loss": 0.0407881923019886, "global_step": 170145, "epoch": 1911, "lr": 3.070405906198917e-05}
{"train_loss": 0.0535985603928566, "global_step": 170146, "epoch": 1911, "lr": 3.070352423728624e-05}
{"train_loss": 0.05831725895404816, "global_step": 170147, "epoch": 1911, "lr": 3.07029894151775e-05}
{"train_loss": 0.10058315098285675, "global_step": 170148, "epoch": 1911, "lr": 3.0702454595663e-05}
{"train_loss": 0.07882185280323029, "global_step": 170149, "epoch": 1911, "lr": 3.070191977874282e-05}
{"train_loss": 0.05677950382232666, "global_step": 170150, "epoch": 1911, "lr": 3.0701384964417036e-05}
{"train_loss": 0.10471082478761673, "global_step": 170151, "epoch": 1911, "lr": 3.070085015268572e-05}
{"train_loss": 0.0864025354385376, "global_step": 170152, "epoch": 1911, "lr": 3.0700315343548916e-05}
{"train_loss": 0.07056650519371033, "global_step": 170153, "epoch": 1911, "lr": 3.069978053700675e-05}
{"train_loss": 0.08989214152097702, "global_step": 170154, "epoch": 1911, "lr": 3.069924573305923e-05}
{"train_loss": 0.12266844511032104, "global_step": 170155, "epoch": 1911, "lr": 3.06987109317065e-05}
{"train_loss": 0.10453872382640839, "global_step": 170156, "epoch": 1911, "lr": 3.069817613294856e-05}
{"train_loss": 0.163487508893013, "global_step": 170157, "epoch": 1911, "lr": 3.069764133678552e-05}
{"train_loss": 0.09328445047140121, "global_step": 170158, "epoch": 1911, "lr": 3.0697106543217466e-05}
{"train_loss": 0.10101982206106186, "global_step": 170159, "epoch": 1911, "lr": 3.0696571752244425e-05}
{"train_loss": 0.06379737704992294, "global_step": 170160, "epoch": 1911, "lr": 3.069603696386652e-05}
{"train_loss": 0.09505259245634079, "global_step": 170161, "epoch": 1911, "lr": 3.069550217808378e-05}
{"train_loss": 0.0979711189866066, "global_step": 170162, "epoch": 1911, "lr": 3.069496739489631e-05}
{"train_loss": 0.09619744122028351, "global_step": 170163, "epoch": 1911, "lr": 3.069443261430415e-05}
{"train_loss": 0.09159322082996368, "global_step": 170164, "epoch": 1911, "lr": 3.06938978363074e-05}
{"train_loss": 0.07883044332265854, "global_step": 170165, "epoch": 1911, "lr": 3.0693363060906106e-05}
{"train_loss": 0.06138071417808533, "global_step": 170166, "epoch": 1911, "lr": 3.069282828810037e-05}
{"train_loss": 0.08397724044122053, "global_step": 170167, "epoch": 1911, "lr": 3.069229351789025e-05, "val_loss": 7.239826679229736}
{"train_loss": 0.09892839193344116, "global_step": 170168, "epoch": 1912, "lr": 3.06917587502758e-05}
{"train_loss": 0.06461413949728012, "global_step": 170169, "epoch": 1912, "lr": 3.0691223985257115e-05}
{"train_loss": 0.07024022936820984, "global_step": 170170, "epoch": 1912, "lr": 3.0690689222834254e-05}
{"train_loss": 0.1564505249261856, "global_step": 170171, "epoch": 1912, "lr": 3.0690154463007305e-05}
{"train_loss": 0.1028565764427185, "global_step": 170172, "epoch": 1912, "lr": 3.068961970577634e-05}
{"train_loss": 0.08302280306816101, "global_step": 170173, "epoch": 1912, "lr": 3.068908495114139e-05}
{"train_loss": 0.11629419028759003, "global_step": 170174, "epoch": 1912, "lr": 3.068855019910256e-05}
{"train_loss": 0.15281717479228973, "global_step": 170175, "epoch": 1912, "lr": 3.0688015449659945e-05}
{"train_loss": 0.0826662927865982, "global_step": 170176, "epoch": 1912, "lr": 3.0687480702813566e-05}
{"train_loss": 0.056687552481889725, "global_step": 170177, "epoch": 1912, "lr": 3.068694595856354e-05}
{"train_loss": 0.06993098556995392, "global_step": 170178, "epoch": 1912, "lr": 3.06864112169099e-05}
{"train_loss": 0.11125709116458893, "global_step": 170179, "epoch": 1912, "lr": 3.068587647785276e-05}
{"train_loss": 0.05230337008833885, "global_step": 170180, "epoch": 1912, "lr": 3.068534174139214e-05}
{"train_loss": 0.12024594843387604, "global_step": 170181, "epoch": 1912, "lr": 3.0684807007528165e-05}
{"train_loss": 0.1101287454366684, "global_step": 170182, "epoch": 1912, "lr": 3.068427227626086e-05}
{"train_loss": 0.05673739314079285, "global_step": 170183, "epoch": 1912, "lr": 3.068373754759034e-05}
{"train_loss": 0.054932065308094025, "global_step": 170184, "epoch": 1912, "lr": 3.0683202821516646e-05}
{"train_loss": 0.15321356058120728, "global_step": 170185, "epoch": 1912, "lr": 3.0682668098039865e-05}
{"train_loss": 0.0890699103474617, "global_step": 170186, "epoch": 1912, "lr": 3.068213337716006e-05}
{"train_loss": 0.11310736835002899, "global_step": 170187, "epoch": 1912, "lr": 3.0681598658877306e-05}
{"train_loss": 0.11201103031635284, "global_step": 170188, "epoch": 1912, "lr": 3.068106394319168e-05}
{"train_loss": 0.11541571468114853, "global_step": 170189, "epoch": 1912, "lr": 3.068052923010325e-05}
{"train_loss": 0.07385259866714478, "global_step": 170190, "epoch": 1912, "lr": 3.067999451961208e-05}
{"train_loss": 0.09593966603279114, "global_step": 170191, "epoch": 1912, "lr": 3.067945981171826e-05}
{"train_loss": 0.0988965705037117, "global_step": 170192, "epoch": 1912, "lr": 3.067892510642183e-05}
{"train_loss": 0.08595947176218033, "global_step": 170193, "epoch": 1912, "lr": 3.067839040372289e-05}
{"train_loss": 0.10205480456352234, "global_step": 170194, "epoch": 1912, "lr": 3.0677855703621525e-05}
{"train_loss": 0.0841766819357872, "global_step": 170195, "epoch": 1912, "lr": 3.067732100611776e-05}
{"train_loss": 0.07444033026695251, "global_step": 170196, "epoch": 1912, "lr": 3.067678631121172e-05}
{"train_loss": 0.10029477626085281, "global_step": 170197, "epoch": 1912, "lr": 3.067625161890343e-05}
{"train_loss": 0.12838783860206604, "global_step": 170198, "epoch": 1912, "lr": 3.0675716929193e-05}
{"train_loss": 0.09925314038991928, "global_step": 170199, "epoch": 1912, "lr": 3.067518224208046e-05}
{"train_loss": 0.062367528676986694, "global_step": 170200, "epoch": 1912, "lr": 3.067464755756593e-05}
{"train_loss": 0.06021355837583542, "global_step": 170201, "epoch": 1912, "lr": 3.067411287564944e-05}
{"train_loss": 0.06727347522974014, "global_step": 170202, "epoch": 1912, "lr": 3.06735781963311e-05}
{"train_loss": 0.06547791510820389, "global_step": 170203, "epoch": 1912, "lr": 3.0673043519610945e-05}
{"train_loss": 0.07199379801750183, "global_step": 170204, "epoch": 1912, "lr": 3.067250884548908e-05}
{"train_loss": 0.0604490265250206, "global_step": 170205, "epoch": 1912, "lr": 3.067197417396555e-05}
{"train_loss": 0.06988754868507385, "global_step": 170206, "epoch": 1912, "lr": 3.067143950504043e-05}
{"train_loss": 0.059043947607278824, "global_step": 170207, "epoch": 1912, "lr": 3.067090483871381e-05}
{"train_loss": 0.04601746052503586, "global_step": 170208, "epoch": 1912, "lr": 3.067037017498575e-05}
{"train_loss": 0.03736497461795807, "global_step": 170209, "epoch": 1912, "lr": 3.0669835513856323e-05}
{"train_loss": 0.07878724485635757, "global_step": 170210, "epoch": 1912, "lr": 3.066930085532561e-05}
{"train_loss": 0.05346154421567917, "global_step": 170211, "epoch": 1912, "lr": 3.066876619939365e-05}
{"train_loss": 0.10319046676158905, "global_step": 170212, "epoch": 1912, "lr": 3.066823154606054e-05}
{"train_loss": 0.05452505499124527, "global_step": 170213, "epoch": 1912, "lr": 3.0667696895326375e-05}
{"train_loss": 0.049461349844932556, "global_step": 170214, "epoch": 1912, "lr": 3.066716224719118e-05}
{"train_loss": 0.05386567860841751, "global_step": 170215, "epoch": 1912, "lr": 3.0666627601655074e-05}
{"train_loss": 0.05220622941851616, "global_step": 170216, "epoch": 1912, "lr": 3.0666092958718076e-05}
{"train_loss": 0.07768888026475906, "global_step": 170217, "epoch": 1912, "lr": 3.066555831838031e-05}
{"train_loss": 0.13241690397262573, "global_step": 170218, "epoch": 1912, "lr": 3.0665023680641803e-05}
{"train_loss": 0.0902949795126915, "global_step": 170219, "epoch": 1912, "lr": 3.066448904550267e-05}
{"train_loss": 0.08709195256233215, "global_step": 170220, "epoch": 1912, "lr": 3.0663954412962935e-05}
{"train_loss": 0.059750016778707504, "global_step": 170221, "epoch": 1912, "lr": 3.066341978302273e-05}
{"train_loss": 0.10974539816379547, "global_step": 170222, "epoch": 1912, "lr": 3.066288515568206e-05}
{"train_loss": 0.10261065512895584, "global_step": 170223, "epoch": 1912, "lr": 3.066235053094105e-05}
{"train_loss": 0.05488458648324013, "global_step": 170224, "epoch": 1912, "lr": 3.066181590879974e-05}
{"train_loss": 0.051216140389442444, "global_step": 170225, "epoch": 1912, "lr": 3.066128128925822e-05}
{"train_loss": 0.14690473675727844, "global_step": 170226, "epoch": 1912, "lr": 3.0660746672316554e-05}
{"train_loss": 0.055598095059394836, "global_step": 170227, "epoch": 1912, "lr": 3.066021205797481e-05}
{"train_loss": 0.0673016905784607, "global_step": 170228, "epoch": 1912, "lr": 3.065967744623307e-05}
{"train_loss": 0.08874398469924927, "global_step": 170229, "epoch": 1912, "lr": 3.065914283709139e-05}
{"train_loss": 0.10863551497459412, "global_step": 170230, "epoch": 1912, "lr": 3.0658608230549866e-05}
{"train_loss": 0.12775126099586487, "global_step": 170231, "epoch": 1912, "lr": 3.0658073626608543e-05}
{"train_loss": 0.0840396136045456, "global_step": 170232, "epoch": 1912, "lr": 3.0657539025267524e-05}
{"train_loss": 0.09876355528831482, "global_step": 170233, "epoch": 1912, "lr": 3.065700442652684e-05}
{"train_loss": 0.08485767990350723, "global_step": 170234, "epoch": 1912, "lr": 3.065646983038661e-05}
{"train_loss": 0.1152733787894249, "global_step": 170235, "epoch": 1912, "lr": 3.0655935236846866e-05}
{"train_loss": 0.10311536490917206, "global_step": 170236, "epoch": 1912, "lr": 3.06554006459077e-05}
{"train_loss": 0.08376098424196243, "global_step": 170237, "epoch": 1912, "lr": 3.0654866057569174e-05}
{"train_loss": 0.11094511300325394, "global_step": 170238, "epoch": 1912, "lr": 3.065433147183139e-05}
{"train_loss": 0.15286362171173096, "global_step": 170239, "epoch": 1912, "lr": 3.065379688869436e-05}
{"train_loss": 0.07966750860214233, "global_step": 170240, "epoch": 1912, "lr": 3.0653262308158226e-05}
{"train_loss": 0.06251096725463867, "global_step": 170241, "epoch": 1912, "lr": 3.0652727730223e-05}
{"train_loss": 0.03204652667045593, "global_step": 170242, "epoch": 1912, "lr": 3.065219315488879e-05}
{"train_loss": 0.08103843033313751, "global_step": 170243, "epoch": 1912, "lr": 3.065165858215565e-05}
{"train_loss": 0.05371019244194031, "global_step": 170244, "epoch": 1912, "lr": 3.065112401202367e-05}
{"train_loss": 0.09412968158721924, "global_step": 170245, "epoch": 1912, "lr": 3.0650589444492896e-05}
{"train_loss": 0.06232134625315666, "global_step": 170246, "epoch": 1912, "lr": 3.065005487956343e-05}
{"train_loss": 0.07942990213632584, "global_step": 170247, "epoch": 1912, "lr": 3.064952031723531e-05}
{"train_loss": 0.11266755312681198, "global_step": 170248, "epoch": 1912, "lr": 3.064898575750863e-05}
{"train_loss": 0.04476827383041382, "global_step": 170249, "epoch": 1912, "lr": 3.0648451200383474e-05}
{"train_loss": 0.04384494200348854, "global_step": 170250, "epoch": 1912, "lr": 3.064791664585988e-05}
{"train_loss": 0.08564841747283936, "global_step": 170251, "epoch": 1912, "lr": 3.064738209393795e-05}
{"train_loss": 0.0696762427687645, "global_step": 170252, "epoch": 1912, "lr": 3.064684754461773e-05}
{"train_loss": 0.06709425151348114, "global_step": 170253, "epoch": 1912, "lr": 3.0646312997899325e-05}
{"train_loss": 0.11692390590906143, "global_step": 170254, "epoch": 1912, "lr": 3.064577845378277e-05}
{"train_loss": 0.0407855287194252, "global_step": 170255, "epoch": 1912, "lr": 3.0645243912268166e-05}
{"train_loss": 0.08560010037395392, "global_step": 170256, "epoch": 1912, "lr": 3.064470937335556e-05, "val_loss": 7.3155517578125}
{"train_loss": 0.09387530386447906, "global_step": 170257, "epoch": 1913, "lr": 3.0644174837045056e-05}
{"train_loss": 0.09272295981645584, "global_step": 170258, "epoch": 1913, "lr": 3.064364030333668e-05}
{"train_loss": 0.07958564162254333, "global_step": 170259, "epoch": 1913, "lr": 3.0643105772230565e-05}
{"train_loss": 0.05352157726883888, "global_step": 170260, "epoch": 1913, "lr": 3.0642571243726715e-05}
{"train_loss": 0.031804442405700684, "global_step": 170261, "epoch": 1913, "lr": 3.064203671782526e-05}
{"train_loss": 0.11838517338037491, "global_step": 170262, "epoch": 1913, "lr": 3.064150219452624e-05}
{"train_loss": 0.08292806148529053, "global_step": 170263, "epoch": 1913, "lr": 3.064096767382972e-05}
{"train_loss": 0.07047160714864731, "global_step": 170264, "epoch": 1913, "lr": 3.0640433155735806e-05}
{"train_loss": 0.11842860281467438, "global_step": 170265, "epoch": 1913, "lr": 3.063989864024453e-05}
{"train_loss": 0.10452967137098312, "global_step": 170266, "epoch": 1913, "lr": 3.063936412735601e-05}
{"train_loss": 0.10934358835220337, "global_step": 170267, "epoch": 1913, "lr": 3.0638829617070266e-05}
{"train_loss": 0.1549607515335083, "global_step": 170268, "epoch": 1913, "lr": 3.063829510938741e-05}
{"train_loss": 0.10859736800193787, "global_step": 170269, "epoch": 1913, "lr": 3.063776060430749e-05}
{"train_loss": 0.11244150251150131, "global_step": 170270, "epoch": 1913, "lr": 3.0637226101830595e-05}
{"train_loss": 0.047515928745269775, "global_step": 170271, "epoch": 1913, "lr": 3.063669160195678e-05}
{"train_loss": 0.10052244365215302, "global_step": 170272, "epoch": 1913, "lr": 3.0636157104686136e-05}
{"train_loss": 0.05985301360487938, "global_step": 170273, "epoch": 1913, "lr": 3.063562261001871e-05}
{"train_loss": 0.0968405157327652, "global_step": 170274, "epoch": 1913, "lr": 3.063508811795461e-05}
{"train_loss": 0.07309293746948242, "global_step": 170275, "epoch": 1913, "lr": 3.0634553628493864e-05}
{"train_loss": 0.11058608442544937, "global_step": 170276, "epoch": 1913, "lr": 3.063401914163659e-05}
{"train_loss": 0.06699366122484207, "global_step": 170277, "epoch": 1913, "lr": 3.063348465738281e-05}
{"train_loss": 0.06815715879201889, "global_step": 170278, "epoch": 1913, "lr": 3.063295017573265e-05}
{"train_loss": 0.08679051697254181, "global_step": 170279, "epoch": 1913, "lr": 3.0632415696686127e-05}
{"train_loss": 0.08469360321760178, "global_step": 170280, "epoch": 1913, "lr": 3.063188122024336e-05}
{"train_loss": 0.03973924741148949, "global_step": 170281, "epoch": 1913, "lr": 3.063134674640439e-05}
{"train_loss": 0.04564154893159866, "global_step": 170282, "epoch": 1913, "lr": 3.063081227516931e-05}
{"train_loss": 0.060167569667100906, "global_step": 170283, "epoch": 1913, "lr": 3.063027780653817e-05}
{"train_loss": 0.07982651889324188, "global_step": 170284, "epoch": 1913, "lr": 3.0629743340511046e-05}
{"train_loss": 0.05985872820019722, "global_step": 170285, "epoch": 1913, "lr": 3.062920887708804e-05}
{"train_loss": 0.1080903634428978, "global_step": 170286, "epoch": 1913, "lr": 3.0628674416269174e-05}
{"train_loss": 0.13375131785869598, "global_step": 170287, "epoch": 1913, "lr": 3.062813995805457e-05}
{"train_loss": 0.12474098056554794, "global_step": 170288, "epoch": 1913, "lr": 3.0627605502444256e-05}
{"train_loss": 0.10476177930831909, "global_step": 170289, "epoch": 1913, "lr": 3.062707104943835e-05}
{"train_loss": 0.0828499048948288, "global_step": 170290, "epoch": 1913, "lr": 3.062653659903687e-05}
{"train_loss": 0.06814849376678467, "global_step": 170291, "epoch": 1913, "lr": 3.062600215123994e-05}
{"train_loss": 0.03200014308094978, "global_step": 170292, "epoch": 1913, "lr": 3.062546770604759e-05}
{"train_loss": 0.11993826180696487, "global_step": 170293, "epoch": 1913, "lr": 3.0624933263459924e-05}
{"train_loss": 0.12587252259254456, "global_step": 170294, "epoch": 1913, "lr": 3.062439882347699e-05}
{"train_loss": 0.1164429560303688, "global_step": 170295, "epoch": 1913, "lr": 3.062386438609888e-05}
{"train_loss": 0.11110620945692062, "global_step": 170296, "epoch": 1913, "lr": 3.062332995132563e-05}
{"train_loss": 0.11273478716611862, "global_step": 170297, "epoch": 1913, "lr": 3.062279551915736e-05}
{"train_loss": 0.11861708760261536, "global_step": 170298, "epoch": 1913, "lr": 3.062226108959411e-05}
{"train_loss": 0.0532601960003376, "global_step": 170299, "epoch": 1913, "lr": 3.062172666263597e-05}
{"train_loss": 0.11441589891910553, "global_step": 170300, "epoch": 1913, "lr": 3.0621192238282995e-05}
{"train_loss": 0.0865415558218956, "global_step": 170301, "epoch": 1913, "lr": 3.062065781653527e-05}
{"train_loss": 0.10817582160234451, "global_step": 170302, "epoch": 1913, "lr": 3.0620123397392864e-05}
{"train_loss": 0.10227886587381363, "global_step": 170303, "epoch": 1913, "lr": 3.0619588980855826e-05}
{"train_loss": 0.09854788333177567, "global_step": 170304, "epoch": 1913, "lr": 3.0619054566924275e-05}
{"train_loss": 0.05672793462872505, "global_step": 170305, "epoch": 1913, "lr": 3.061852015559823e-05}
{"train_loss": 0.09935179352760315, "global_step": 170306, "epoch": 1913, "lr": 3.061798574687781e-05}
{"train_loss": 0.060049429535865784, "global_step": 170307, "epoch": 1913, "lr": 3.061745134076305e-05}
{"train_loss": 0.09164337068796158, "global_step": 170308, "epoch": 1913, "lr": 3.061691693725405e-05}
{"train_loss": 0.09715824574232101, "global_step": 170309, "epoch": 1913, "lr": 3.0616382536350855e-05}
{"train_loss": 0.10921075195074081, "global_step": 170310, "epoch": 1913, "lr": 3.0615848138053574e-05}
{"train_loss": 0.07568267732858658, "global_step": 170311, "epoch": 1913, "lr": 3.061531374236223e-05}
{"train_loss": 0.11457359790802002, "global_step": 170312, "epoch": 1913, "lr": 3.061477934927694e-05}
{"train_loss": 0.060484711080789566, "global_step": 170313, "epoch": 1913, "lr": 3.061424495879774e-05}
{"train_loss": 0.11828746646642685, "global_step": 170314, "epoch": 1913, "lr": 3.061371057092474e-05}
{"train_loss": 0.09826400130987167, "global_step": 170315, "epoch": 1913, "lr": 3.061317618565797e-05}
{"train_loss": 0.08422494679689407, "global_step": 170316, "epoch": 1913, "lr": 3.0612641802997545e-05}
{"train_loss": 0.06785111874341965, "global_step": 170317, "epoch": 1913, "lr": 3.061210742294348e-05}
{"train_loss": 0.07522490620613098, "global_step": 170318, "epoch": 1913, "lr": 3.061157304549591e-05}
{"train_loss": 0.09725625067949295, "global_step": 170319, "epoch": 1913, "lr": 3.061103867065486e-05}
{"train_loss": 0.04551798850297928, "global_step": 170320, "epoch": 1913, "lr": 3.0610504298420426e-05}
{"train_loss": 0.03627493605017662, "global_step": 170321, "epoch": 1913, "lr": 3.060996992879269e-05}
{"train_loss": 0.06573555618524551, "global_step": 170322, "epoch": 1913, "lr": 3.060943556177168e-05}
{"train_loss": 0.10869209468364716, "global_step": 170323, "epoch": 1913, "lr": 3.060890119735752e-05}
{"train_loss": 0.05468033626675606, "global_step": 170324, "epoch": 1913, "lr": 3.060836683555023e-05}
{"train_loss": 0.06503535062074661, "global_step": 170325, "epoch": 1913, "lr": 3.060783247634993e-05}
{"train_loss": 0.10245177149772644, "global_step": 170326, "epoch": 1913, "lr": 3.060729811975666e-05}
{"train_loss": 0.04108208790421486, "global_step": 170327, "epoch": 1913, "lr": 3.0606763765770516e-05}
{"train_loss": 0.03931816294789314, "global_step": 170328, "epoch": 1913, "lr": 3.060622941439153e-05}
{"train_loss": 0.09488631784915924, "global_step": 170329, "epoch": 1913, "lr": 3.0605695065619836e-05}
{"train_loss": 0.08323042094707489, "global_step": 170330, "epoch": 1913, "lr": 3.0605160719455434e-05}
{"train_loss": 0.05962170287966728, "global_step": 170331, "epoch": 1913, "lr": 3.060462637589846e-05}
{"train_loss": 0.1096939668059349, "global_step": 170332, "epoch": 1913, "lr": 3.060409203494893e-05}
{"train_loss": 0.1407311111688614, "global_step": 170333, "epoch": 1913, "lr": 3.060355769660697e-05}
{"train_loss": 0.1520906537771225, "global_step": 170334, "epoch": 1913, "lr": 3.060302336087261e-05}
{"train_loss": 0.11726178973913193, "global_step": 170335, "epoch": 1913, "lr": 3.060248902774595e-05}
{"train_loss": 0.09508159756660461, "global_step": 170336, "epoch": 1913, "lr": 3.060195469722702e-05}
{"train_loss": 0.11431490629911423, "global_step": 170337, "epoch": 1913, "lr": 3.0601420369315956e-05}
{"train_loss": 0.05440916866064072, "global_step": 170338, "epoch": 1913, "lr": 3.060088604401277e-05}
{"train_loss": 0.07824999839067459, "global_step": 170339, "epoch": 1913, "lr": 3.060035172131757e-05}
{"train_loss": 0.11053585261106491, "global_step": 170340, "epoch": 1913, "lr": 3.0599817401230425e-05}
{"train_loss": 0.08833207190036774, "global_step": 170341, "epoch": 1913, "lr": 3.0599283083751376e-05}
{"train_loss": 0.0675300806760788, "global_step": 170342, "epoch": 1913, "lr": 3.059874876888054e-05}
{"train_loss": 0.08653192222118378, "global_step": 170343, "epoch": 1913, "lr": 3.059821445661794e-05}
{"train_loss": 0.11506439745426178, "global_step": 170344, "epoch": 1913, "lr": 3.059768014696369e-05}
{"train_loss": 0.08838554019673486, "global_step": 170345, "epoch": 1913, "lr": 3.059714583991784e-05, "val_loss": 7.160267353057861}
{"train_loss": 0.12352398037910461, "global_step": 170346, "epoch": 1914, "lr": 3.059661153548048e-05}
{"train_loss": 0.13244590163230896, "global_step": 170347, "epoch": 1914, "lr": 3.059607723365165e-05}
{"train_loss": 0.0727989599108696, "global_step": 170348, "epoch": 1914, "lr": 3.059554293443146e-05}
{"train_loss": 0.12319856882095337, "global_step": 170349, "epoch": 1914, "lr": 3.0595008637819946e-05}
{"train_loss": 0.07757514715194702, "global_step": 170350, "epoch": 1914, "lr": 3.059447434381721e-05}
{"train_loss": 0.05564379319548607, "global_step": 170351, "epoch": 1914, "lr": 3.059394005242329e-05}
{"train_loss": 0.08172868192195892, "global_step": 170352, "epoch": 1914, "lr": 3.05934057636383e-05}
{"train_loss": 0.044764891266822815, "global_step": 170353, "epoch": 1914, "lr": 3.059287147746228e-05}
{"train_loss": 0.1344444751739502, "global_step": 170354, "epoch": 1914, "lr": 3.059233719389532e-05}
{"train_loss": 0.06274975091218948, "global_step": 170355, "epoch": 1914, "lr": 3.059180291293747e-05}
{"train_loss": 0.12332852184772491, "global_step": 170356, "epoch": 1914, "lr": 3.059126863458882e-05}
{"train_loss": 0.07359150797128677, "global_step": 170357, "epoch": 1914, "lr": 3.0590734358849436e-05}
{"train_loss": 0.10221614688634872, "global_step": 170358, "epoch": 1914, "lr": 3.05902000857194e-05}
{"train_loss": 0.12853069603443146, "global_step": 170359, "epoch": 1914, "lr": 3.058966581519878e-05}
{"train_loss": 0.10579875111579895, "global_step": 170360, "epoch": 1914, "lr": 3.0589131547287625e-05}
{"train_loss": 0.07933391630649567, "global_step": 170361, "epoch": 1914, "lr": 3.058859728198604e-05}
{"train_loss": 0.04397136718034744, "global_step": 170362, "epoch": 1914, "lr": 3.058806301929407e-05}
{"train_loss": 0.0551934689283371, "global_step": 170363, "epoch": 1914, "lr": 3.0587528759211814e-05}
{"train_loss": 0.1026851087808609, "global_step": 170364, "epoch": 1914, "lr": 3.058699450173931e-05}
{"train_loss": 0.06107374653220177, "global_step": 170365, "epoch": 1914, "lr": 3.0586460246876656e-05}
{"train_loss": 0.07899629324674606, "global_step": 170366, "epoch": 1914, "lr": 3.058592599462391e-05}
{"train_loss": 0.1244521215558052, "global_step": 170367, "epoch": 1914, "lr": 3.058539174498116e-05}
{"train_loss": 0.09752510488033295, "global_step": 170368, "epoch": 1914, "lr": 3.058485749794846e-05}
{"train_loss": 0.14756111800670624, "global_step": 170369, "epoch": 1914, "lr": 3.05843232535259e-05}
{"train_loss": 0.07703428715467453, "global_step": 170370, "epoch": 1914, "lr": 3.0583789011713524e-05}
{"train_loss": 0.10202020406723022, "global_step": 170371, "epoch": 1914, "lr": 3.058325477251144e-05}
{"train_loss": 0.07612057775259018, "global_step": 170372, "epoch": 1914, "lr": 3.058272053591967e-05}
{"train_loss": 0.09462695568799973, "global_step": 170373, "epoch": 1914, "lr": 3.0582186301938353e-05}
{"train_loss": 0.10808735340833664, "global_step": 170374, "epoch": 1914, "lr": 3.05816520705675e-05}
{"train_loss": 0.1054905578494072, "global_step": 170375, "epoch": 1914, "lr": 3.058111784180722e-05}
{"train_loss": 0.056663092225790024, "global_step": 170376, "epoch": 1914, "lr": 3.058058361565756e-05}
{"train_loss": 0.09701070189476013, "global_step": 170377, "epoch": 1914, "lr": 3.058004939211861e-05}
{"train_loss": 0.07824952900409698, "global_step": 170378, "epoch": 1914, "lr": 3.057951517119045e-05}
{"train_loss": 0.10373606532812119, "global_step": 170379, "epoch": 1914, "lr": 3.0578980952873114e-05}
{"train_loss": 0.07133638113737106, "global_step": 170380, "epoch": 1914, "lr": 3.0578446737166714e-05}
{"train_loss": 0.10750439018011093, "global_step": 170381, "epoch": 1914, "lr": 3.0577912524071296e-05}
{"train_loss": 0.08954597264528275, "global_step": 170382, "epoch": 1914, "lr": 3.057737831358695e-05}
{"train_loss": 0.06619024276733398, "global_step": 170383, "epoch": 1914, "lr": 3.057684410571372e-05}
{"train_loss": 0.07085589319467545, "global_step": 170384, "epoch": 1914, "lr": 3.057630990045172e-05}
{"train_loss": 0.07187354564666748, "global_step": 170385, "epoch": 1914, "lr": 3.057577569780098e-05}
{"train_loss": 0.08882405608892441, "global_step": 170386, "epoch": 1914, "lr": 3.0575241497761605e-05}
{"train_loss": 0.06455176323652267, "global_step": 170387, "epoch": 1914, "lr": 3.057470730033364e-05}
{"train_loss": 0.03738832473754883, "global_step": 170388, "epoch": 1914, "lr": 3.057417310551718e-05}
{"train_loss": 0.07041256129741669, "global_step": 170389, "epoch": 1914, "lr": 3.057363891331227e-05}
{"train_loss": 0.10313592851161957, "global_step": 170390, "epoch": 1914, "lr": 3.0573104723719016e-05}
{"train_loss": 0.12598542869091034, "global_step": 170391, "epoch": 1914, "lr": 3.0572570536737454e-05}
{"train_loss": 0.08119522780179977, "global_step": 170392, "epoch": 1914, "lr": 3.057203635236768e-05}
{"train_loss": 0.06768932938575745, "global_step": 170393, "epoch": 1914, "lr": 3.057150217060976e-05}
{"train_loss": 0.04135455563664436, "global_step": 170394, "epoch": 1914, "lr": 3.057096799146376e-05}
{"train_loss": 0.07009771466255188, "global_step": 170395, "epoch": 1914, "lr": 3.0570433814929767e-05}
{"train_loss": 0.07200848311185837, "global_step": 170396, "epoch": 1914, "lr": 3.056989964100784e-05}
{"train_loss": 0.11847075074911118, "global_step": 170397, "epoch": 1914, "lr": 3.056936546969804e-05}
{"train_loss": 0.06343100965023041, "global_step": 170398, "epoch": 1914, "lr": 3.0568831301000465e-05}
{"train_loss": 0.09558402001857758, "global_step": 170399, "epoch": 1914, "lr": 3.056829713491518e-05}
{"train_loss": 0.13315390050411224, "global_step": 170400, "epoch": 1914, "lr": 3.056776297144223e-05}
{"train_loss": 0.08989408612251282, "global_step": 170401, "epoch": 1914, "lr": 3.056722881058173e-05}
{"train_loss": 0.09546104818582535, "global_step": 170402, "epoch": 1914, "lr": 3.056669465233371e-05}
{"train_loss": 0.04686960205435753, "global_step": 170403, "epoch": 1914, "lr": 3.056616049669827e-05}
{"train_loss": 0.05944797024130821, "global_step": 170404, "epoch": 1914, "lr": 3.056562634367547e-05}
{"train_loss": 0.0639166310429573, "global_step": 170405, "epoch": 1914, "lr": 3.05650921932654e-05}
{"train_loss": 0.06930004805326462, "global_step": 170406, "epoch": 1914, "lr": 3.056455804546809e-05}
{"train_loss": 0.0840340256690979, "global_step": 170407, "epoch": 1914, "lr": 3.0564023900283656e-05}
{"train_loss": 0.13840538263320923, "global_step": 170408, "epoch": 1914, "lr": 3.056348975771214e-05}
{"train_loss": 0.03966683894395828, "global_step": 170409, "epoch": 1914, "lr": 3.056295561775364e-05}
{"train_loss": 0.05076894164085388, "global_step": 170410, "epoch": 1914, "lr": 3.05624214804082e-05}
{"train_loss": 0.04935087263584137, "global_step": 170411, "epoch": 1914, "lr": 3.05618873456759e-05}
{"train_loss": 0.0548480749130249, "global_step": 170412, "epoch": 1914, "lr": 3.056135321355684e-05}
{"train_loss": 0.04231136292219162, "global_step": 170413, "epoch": 1914, "lr": 3.0560819084051065e-05}
{"train_loss": 0.03722605109214783, "global_step": 170414, "epoch": 1914, "lr": 3.0560284957158644e-05}
{"train_loss": 0.07129126787185669, "global_step": 170415, "epoch": 1914, "lr": 3.055975083287965e-05}
{"train_loss": 0.10620855540037155, "global_step": 170416, "epoch": 1914, "lr": 3.055921671121417e-05}
{"train_loss": 0.053667813539505005, "global_step": 170417, "epoch": 1914, "lr": 3.0558682592162265e-05}
{"train_loss": 0.09744241833686829, "global_step": 170418, "epoch": 1914, "lr": 3.055814847572401e-05}
{"train_loss": 0.07765233516693115, "global_step": 170419, "epoch": 1914, "lr": 3.055761436189947e-05}
{"train_loss": 0.06891100853681564, "global_step": 170420, "epoch": 1914, "lr": 3.055708025068873e-05}
{"train_loss": 0.05214495584368706, "global_step": 170421, "epoch": 1914, "lr": 3.0556546142091834e-05}
{"train_loss": 0.10550587624311447, "global_step": 170422, "epoch": 1914, "lr": 3.05560120361089e-05}
{"train_loss": 0.07900926470756531, "global_step": 170423, "epoch": 1914, "lr": 3.055547793273995e-05}
{"train_loss": 0.038551345467567444, "global_step": 170424, "epoch": 1914, "lr": 3.05549438319851e-05}
{"train_loss": 0.09755215793848038, "global_step": 170425, "epoch": 1914, "lr": 3.055440973384438e-05}
{"train_loss": 0.06444539874792099, "global_step": 170426, "epoch": 1914, "lr": 3.0553875638317894e-05}
{"train_loss": 0.052281998097896576, "global_step": 170427, "epoch": 1914, "lr": 3.05533415454057e-05}
{"train_loss": 0.09602733701467514, "global_step": 170428, "epoch": 1914, "lr": 3.055280745510786e-05}
{"train_loss": 0.06266028434038162, "global_step": 170429, "epoch": 1914, "lr": 3.0552273367424476e-05}
{"train_loss": 0.06908535957336426, "global_step": 170430, "epoch": 1914, "lr": 3.055173928235559e-05}
{"train_loss": 0.08451934903860092, "global_step": 170431, "epoch": 1914, "lr": 3.05512051999013e-05}
{"train_loss": 0.10190792381763458, "global_step": 170432, "epoch": 1914, "lr": 3.055067112006166e-05}
{"train_loss": 0.08750643581151962, "global_step": 170433, "epoch": 1914, "lr": 3.055013704283674e-05}
{"train_loss": 0.08190000094891933, "global_step": 170434, "epoch": 1914, "lr": 3.054960296822661e-05, "val_loss": 7.224175453186035}
{"train_loss": 0.05506256967782974, "global_step": 170435, "epoch": 1915, "lr": 3.054906889623136e-05}
{"train_loss": 0.11217731982469559, "global_step": 170436, "epoch": 1915, "lr": 3.054853482685105e-05}
{"train_loss": 0.08069057762622833, "global_step": 170437, "epoch": 1915, "lr": 3.054800076008576e-05}
{"train_loss": 0.095899797976017, "global_step": 170438, "epoch": 1915, "lr": 3.0547466695935536e-05}
{"train_loss": 0.10465503484010696, "global_step": 170439, "epoch": 1915, "lr": 3.054693263440048e-05}
{"train_loss": 0.06810713559389114, "global_step": 170440, "epoch": 1915, "lr": 3.054639857548064e-05}
{"train_loss": 0.04765363410115242, "global_step": 170441, "epoch": 1915, "lr": 3.0545864519176116e-05}
{"train_loss": 0.06760456413030624, "global_step": 170442, "epoch": 1915, "lr": 3.0545330465486945e-05}
{"train_loss": 0.08303818851709366, "global_step": 170443, "epoch": 1915, "lr": 3.0544796414413236e-05}
{"train_loss": 0.07866021990776062, "global_step": 170444, "epoch": 1915, "lr": 3.054426236595502e-05}
{"train_loss": 0.094826839864254, "global_step": 170445, "epoch": 1915, "lr": 3.054372832011242e-05}
{"train_loss": 0.07792550325393677, "global_step": 170446, "epoch": 1915, "lr": 3.054319427688545e-05}
{"train_loss": 0.08856234699487686, "global_step": 170447, "epoch": 1915, "lr": 3.0542660236274215e-05}
{"train_loss": 0.06807694584131241, "global_step": 170448, "epoch": 1915, "lr": 3.0542126198278796e-05}
{"train_loss": 0.060628484934568405, "global_step": 170449, "epoch": 1915, "lr": 3.054159216289923e-05}
{"train_loss": 0.09962297976016998, "global_step": 170450, "epoch": 1915, "lr": 3.054105813013564e-05}
{"train_loss": 0.11960422247648239, "global_step": 170451, "epoch": 1915, "lr": 3.0540524099988054e-05}
{"train_loss": 0.12338002771139145, "global_step": 170452, "epoch": 1915, "lr": 3.0539990072456555e-05}
{"train_loss": 0.10322028398513794, "global_step": 170453, "epoch": 1915, "lr": 3.0539456047541216e-05}
{"train_loss": 0.1348578929901123, "global_step": 170454, "epoch": 1915, "lr": 3.053892202524211e-05}
{"train_loss": 0.04305356368422508, "global_step": 170455, "epoch": 1915, "lr": 3.053838800555931e-05}
{"train_loss": 0.02743292599916458, "global_step": 170456, "epoch": 1915, "lr": 3.05378539884929e-05}
{"train_loss": 0.06225040927529335, "global_step": 170457, "epoch": 1915, "lr": 3.053731997404291e-05}
{"train_loss": 0.10241572558879852, "global_step": 170458, "epoch": 1915, "lr": 3.0536785962209466e-05}
{"train_loss": 0.093642458319664, "global_step": 170459, "epoch": 1915, "lr": 3.0536251952992595e-05}
{"train_loss": 0.04228563979268074, "global_step": 170460, "epoch": 1915, "lr": 3.0535717946392405e-05}
{"train_loss": 0.11501458287239075, "global_step": 170461, "epoch": 1915, "lr": 3.053518394240893e-05}
{"train_loss": 0.08304006606340408, "global_step": 170462, "epoch": 1915, "lr": 3.053464994104228e-05}
{"train_loss": 0.06649087369441986, "global_step": 170463, "epoch": 1915, "lr": 3.05341159422925e-05}
{"train_loss": 0.14307478070259094, "global_step": 170464, "epoch": 1915, "lr": 3.0533581946159685e-05}
{"train_loss": 0.14505024254322052, "global_step": 170465, "epoch": 1915, "lr": 3.053304795264387e-05}
{"train_loss": 0.09605969488620758, "global_step": 170466, "epoch": 1915, "lr": 3.0532513961745144e-05}
{"train_loss": 0.07112976163625717, "global_step": 170467, "epoch": 1915, "lr": 3.0531979973463613e-05}
{"train_loss": 0.05372917279601097, "global_step": 170468, "epoch": 1915, "lr": 3.05314459877993e-05}
{"train_loss": 0.0795220136642456, "global_step": 170469, "epoch": 1915, "lr": 3.053091200475231e-05}
{"train_loss": 0.03980056941509247, "global_step": 170470, "epoch": 1915, "lr": 3.05303780243227e-05}
{"train_loss": 0.0684048980474472, "global_step": 170471, "epoch": 1915, "lr": 3.052984404651054e-05}
{"train_loss": 0.09651445597410202, "global_step": 170472, "epoch": 1915, "lr": 3.0529310071315905e-05}
{"train_loss": 0.055733174085617065, "global_step": 170473, "epoch": 1915, "lr": 3.052877609873887e-05}
{"train_loss": 0.06804747134447098, "global_step": 170474, "epoch": 1915, "lr": 3.052824212877949e-05}
{"train_loss": 0.06578270345926285, "global_step": 170475, "epoch": 1915, "lr": 3.052770816143787e-05}
{"train_loss": 0.11724162846803665, "global_step": 170476, "epoch": 1915, "lr": 3.052717419671405e-05}
{"train_loss": 0.11761588603258133, "global_step": 170477, "epoch": 1915, "lr": 3.0526640234608126e-05}
{"train_loss": 0.0963313952088356, "global_step": 170478, "epoch": 1915, "lr": 3.052610627512014e-05}
{"train_loss": 0.11209194362163544, "global_step": 170479, "epoch": 1915, "lr": 3.05255723182502e-05}
{"train_loss": 0.0710689127445221, "global_step": 170480, "epoch": 1915, "lr": 3.0525038363998335e-05}
{"train_loss": 0.0770389512181282, "global_step": 170481, "epoch": 1915, "lr": 3.0524504412364675e-05}
{"train_loss": 0.146111398935318, "global_step": 170482, "epoch": 1915, "lr": 3.052397046334923e-05}
{"train_loss": 0.06354755163192749, "global_step": 170483, "epoch": 1915, "lr": 3.05234365169521e-05}
{"train_loss": 0.11421231180429459, "global_step": 170484, "epoch": 1915, "lr": 3.0522902573173376e-05}
{"train_loss": 0.10697758942842484, "global_step": 170485, "epoch": 1915, "lr": 3.052236863201309e-05}
{"train_loss": 0.05825384706258774, "global_step": 170486, "epoch": 1915, "lr": 3.052183469347135e-05}
{"train_loss": 0.06743783503770828, "global_step": 170487, "epoch": 1915, "lr": 3.05213007575482e-05}
{"train_loss": 0.07487033307552338, "global_step": 170488, "epoch": 1915, "lr": 3.052076682424374e-05}
{"train_loss": 0.1209331676363945, "global_step": 170489, "epoch": 1915, "lr": 3.052023289355801e-05}
{"train_loss": 0.030263816937804222, "global_step": 170490, "epoch": 1915, "lr": 3.051969896549111e-05}
{"train_loss": 0.11603792011737823, "global_step": 170491, "epoch": 1915, "lr": 3.0519165040043085e-05}
{"train_loss": 0.08059042692184448, "global_step": 170492, "epoch": 1915, "lr": 3.0518631117214026e-05}
{"train_loss": 0.07305184751749039, "global_step": 170493, "epoch": 1915, "lr": 3.0518097197004e-05}
{"train_loss": 0.05539019778370857, "global_step": 170494, "epoch": 1915, "lr": 3.0517563279413096e-05}
{"train_loss": 0.05509587377309799, "global_step": 170495, "epoch": 1915, "lr": 3.0517029364441342e-05}
{"train_loss": 0.08465767651796341, "global_step": 170496, "epoch": 1915, "lr": 3.0516495452088855e-05}
{"train_loss": 0.0635165125131607, "global_step": 170497, "epoch": 1915, "lr": 3.0515961542355664e-05}
{"train_loss": 0.053081873804330826, "global_step": 170498, "epoch": 1915, "lr": 3.051542763524189e-05}
{"train_loss": 0.08047899603843689, "global_step": 170499, "epoch": 1915, "lr": 3.0514893730747564e-05}
{"train_loss": 0.1106201559305191, "global_step": 170500, "epoch": 1915, "lr": 3.0514359828872786e-05}
{"train_loss": 0.0987587571144104, "global_step": 170501, "epoch": 1915, "lr": 3.051382592961759e-05}
{"train_loss": 0.10544295608997345, "global_step": 170502, "epoch": 1915, "lr": 3.0513292032982077e-05}
{"train_loss": 0.11774369329214096, "global_step": 170503, "epoch": 1915, "lr": 3.0512758138966335e-05}
{"train_loss": 0.12337436527013779, "global_step": 170504, "epoch": 1915, "lr": 3.0512224247570398e-05}
{"train_loss": 0.0477510541677475, "global_step": 170505, "epoch": 1915, "lr": 3.0511690358794375e-05}
{"train_loss": 0.06184202432632446, "global_step": 170506, "epoch": 1915, "lr": 3.0511156472638296e-05}
{"train_loss": 0.09395148605108261, "global_step": 170507, "epoch": 1915, "lr": 3.051062258910228e-05}
{"train_loss": 0.10322947055101395, "global_step": 170508, "epoch": 1915, "lr": 3.051008870818634e-05}
{"train_loss": 0.12421080470085144, "global_step": 170509, "epoch": 1915, "lr": 3.0509554829890607e-05}
{"train_loss": 0.08360065519809723, "global_step": 170510, "epoch": 1915, "lr": 3.0509020954215118e-05}
{"train_loss": 0.1190999373793602, "global_step": 170511, "epoch": 1915, "lr": 3.0508487081159955e-05}
{"train_loss": 0.08825521171092987, "global_step": 170512, "epoch": 1915, "lr": 3.0507953210725193e-05}
{"train_loss": 0.04039247706532478, "global_step": 170513, "epoch": 1915, "lr": 3.050741934291089e-05}
{"train_loss": 0.12915845215320587, "global_step": 170514, "epoch": 1915, "lr": 3.050688547771713e-05}
{"train_loss": 0.052581705152988434, "global_step": 170515, "epoch": 1915, "lr": 3.0506351615143992e-05}
{"train_loss": 0.11170361936092377, "global_step": 170516, "epoch": 1915, "lr": 3.0505817755191518e-05}
{"train_loss": 0.08303270488977432, "global_step": 170517, "epoch": 1915, "lr": 3.0505283897859814e-05}
{"train_loss": 0.11950348317623138, "global_step": 170518, "epoch": 1915, "lr": 3.0504750043148922e-05}
{"train_loss": 0.07178483158349991, "global_step": 170519, "epoch": 1915, "lr": 3.0504216191058932e-05}
{"train_loss": 0.0731179416179657, "global_step": 170520, "epoch": 1915, "lr": 3.0503682341589927e-05}
{"train_loss": 0.07864266633987427, "global_step": 170521, "epoch": 1915, "lr": 3.0503148494741947e-05}
{"train_loss": 0.07052945345640182, "global_step": 170522, "epoch": 1915, "lr": 3.05026146505151e-05}
{"train_loss": 0.08528191179790523, "global_step": 170523, "epoch": 1915, "lr": 3.0502080808909416e-05, "val_loss": 7.115211486816406, "train_action_mse_error": 18.94463539123535}
{"train_loss": 0.12547577917575836, "global_step": 170524, "epoch": 1916, "lr": 3.0501546969925016e-05}
{"train_loss": 0.047321710735559464, "global_step": 170525, "epoch": 1916, "lr": 3.0501013133561917e-05}
{"train_loss": 0.05506625398993492, "global_step": 170526, "epoch": 1916, "lr": 3.050047929982025e-05}
{"train_loss": 0.13648730516433716, "global_step": 170527, "epoch": 1916, "lr": 3.0499945468700032e-05}
{"train_loss": 0.0886150449514389, "global_step": 170528, "epoch": 1916, "lr": 3.049941164020137e-05}
{"train_loss": 0.07348933070898056, "global_step": 170529, "epoch": 1916, "lr": 3.0498877814324324e-05}
{"train_loss": 0.05985084921121597, "global_step": 170530, "epoch": 1916, "lr": 3.0498343991068962e-05}
{"train_loss": 0.10170179605484009, "global_step": 170531, "epoch": 1916, "lr": 3.049781017043536e-05}
{"train_loss": 0.13975465297698975, "global_step": 170532, "epoch": 1916, "lr": 3.0497276352423594e-05}
{"train_loss": 0.10404564440250397, "global_step": 170533, "epoch": 1916, "lr": 3.049674253703373e-05}
{"train_loss": 0.07712884992361069, "global_step": 170534, "epoch": 1916, "lr": 3.0496208724265845e-05}
{"train_loss": 0.0832194983959198, "global_step": 170535, "epoch": 1916, "lr": 3.0495674914119992e-05}
{"train_loss": 0.11998241394758224, "global_step": 170536, "epoch": 1916, "lr": 3.0495141106596282e-05}
{"train_loss": 0.08855321258306503, "global_step": 170537, "epoch": 1916, "lr": 3.049460730169473e-05}
{"train_loss": 0.06447763741016388, "global_step": 170538, "epoch": 1916, "lr": 3.0494073499415453e-05}
{"train_loss": 0.10137948393821716, "global_step": 170539, "epoch": 1916, "lr": 3.049353969975852e-05}
{"train_loss": 0.12305860966444016, "global_step": 170540, "epoch": 1916, "lr": 3.0493005902723982e-05}
{"train_loss": 0.08326295018196106, "global_step": 170541, "epoch": 1916, "lr": 3.0492472108311936e-05}
{"train_loss": 0.037371233105659485, "global_step": 170542, "epoch": 1916, "lr": 3.0491938316522422e-05}
{"train_loss": 0.06532759219408035, "global_step": 170543, "epoch": 1916, "lr": 3.049140452735555e-05}
{"train_loss": 0.038262154906988144, "global_step": 170544, "epoch": 1916, "lr": 3.049087074081135e-05}
{"train_loss": 0.13939924538135529, "global_step": 170545, "epoch": 1916, "lr": 3.0490336956889933e-05}
{"train_loss": 0.07665827870368958, "global_step": 170546, "epoch": 1916, "lr": 3.048980317559133e-05}
{"train_loss": 0.0641394779086113, "global_step": 170547, "epoch": 1916, "lr": 3.048926939691566e-05}
{"train_loss": 0.08977500349283218, "global_step": 170548, "epoch": 1916, "lr": 3.0488735620862957e-05}
{"train_loss": 0.0961863100528717, "global_step": 170549, "epoch": 1916, "lr": 3.0488201847433306e-05}
{"train_loss": 0.08609842509031296, "global_step": 170550, "epoch": 1916, "lr": 3.0487668076626787e-05}
{"train_loss": 0.07878673821687698, "global_step": 170551, "epoch": 1916, "lr": 3.048713430844345e-05}
{"train_loss": 0.10572157055139542, "global_step": 170552, "epoch": 1916, "lr": 3.0486600542883386e-05}
{"train_loss": 0.05630635842680931, "global_step": 170553, "epoch": 1916, "lr": 3.048606677994667e-05}
{"train_loss": 0.0586077943444252, "global_step": 170554, "epoch": 1916, "lr": 3.0485533019633345e-05}
{"train_loss": 0.06342040747404099, "global_step": 170555, "epoch": 1916, "lr": 3.0484999261943503e-05}
{"train_loss": 0.09594416618347168, "global_step": 170556, "epoch": 1916, "lr": 3.0484465506877235e-05}
{"train_loss": 0.09265480190515518, "global_step": 170557, "epoch": 1916, "lr": 3.048393175443457e-05}
{"train_loss": 0.10650242120027542, "global_step": 170558, "epoch": 1916, "lr": 3.0483398004615626e-05}
{"train_loss": 0.07402840256690979, "global_step": 170559, "epoch": 1916, "lr": 3.048286425742043e-05}
{"train_loss": 0.06036034971475601, "global_step": 170560, "epoch": 1916, "lr": 3.0482330512849088e-05}
{"train_loss": 0.09488823264837265, "global_step": 170561, "epoch": 1916, "lr": 3.048179677090165e-05}
{"train_loss": 0.17708562314510345, "global_step": 170562, "epoch": 1916, "lr": 3.048126303157821e-05}
{"train_loss": 0.10973846167325974, "global_step": 170563, "epoch": 1916, "lr": 3.0480729294878807e-05}
{"train_loss": 0.0659584254026413, "global_step": 170564, "epoch": 1916, "lr": 3.0480195560803553e-05}
{"train_loss": 0.1260518878698349, "global_step": 170565, "epoch": 1916, "lr": 3.047966182935248e-05}
{"train_loss": 0.0695120170712471, "global_step": 170566, "epoch": 1916, "lr": 3.04791281005257e-05}
{"train_loss": 0.07423384487628937, "global_step": 170567, "epoch": 1916, "lr": 3.047859437432325e-05}
{"train_loss": 0.10621422529220581, "global_step": 170568, "epoch": 1916, "lr": 3.047806065074521e-05}
{"train_loss": 0.13583794236183167, "global_step": 170569, "epoch": 1916, "lr": 3.0477526929791666e-05}
{"train_loss": 0.14094002544879913, "global_step": 170570, "epoch": 1916, "lr": 3.0476993211462677e-05}
{"train_loss": 0.09110565483570099, "global_step": 170571, "epoch": 1916, "lr": 3.0476459495758318e-05}
{"train_loss": 0.08276242762804031, "global_step": 170572, "epoch": 1916, "lr": 3.0475925782678673e-05}
{"train_loss": 0.12492557615041733, "global_step": 170573, "epoch": 1916, "lr": 3.047539207222378e-05}
{"train_loss": 0.043016862124204636, "global_step": 170574, "epoch": 1916, "lr": 3.0474858364393733e-05}
{"train_loss": 0.06852544099092484, "global_step": 170575, "epoch": 1916, "lr": 3.0474324659188625e-05}
{"train_loss": 0.08194699138402939, "global_step": 170576, "epoch": 1916, "lr": 3.0473790956608482e-05}
{"train_loss": 0.0478421151638031, "global_step": 170577, "epoch": 1916, "lr": 3.047325725665342e-05}
{"train_loss": 0.07347782701253891, "global_step": 170578, "epoch": 1916, "lr": 3.047272355932348e-05}
{"train_loss": 0.10354204475879669, "global_step": 170579, "epoch": 1916, "lr": 3.047218986461875e-05}
{"train_loss": 0.06170647218823433, "global_step": 170580, "epoch": 1916, "lr": 3.0471656172539288e-05}
{"train_loss": 0.07603268325328827, "global_step": 170581, "epoch": 1916, "lr": 3.047112248308519e-05}
{"train_loss": 0.06543689221143723, "global_step": 170582, "epoch": 1916, "lr": 3.047058879625649e-05}
{"train_loss": 0.0762234628200531, "global_step": 170583, "epoch": 1916, "lr": 3.0470055112053296e-05}
{"train_loss": 0.08143304288387299, "global_step": 170584, "epoch": 1916, "lr": 3.0469521430475657e-05}
{"train_loss": 0.06911502778530121, "global_step": 170585, "epoch": 1916, "lr": 3.0468987751523664e-05}
{"train_loss": 0.09143267571926117, "global_step": 170586, "epoch": 1916, "lr": 3.046845407519737e-05}
{"train_loss": 0.07911207526922226, "global_step": 170587, "epoch": 1916, "lr": 3.046792040149685e-05}
{"train_loss": 0.041266582906246185, "global_step": 170588, "epoch": 1916, "lr": 3.0467386730422186e-05}
{"train_loss": 0.08487687259912491, "global_step": 170589, "epoch": 1916, "lr": 3.0466853061973443e-05}
{"train_loss": 0.11951272189617157, "global_step": 170590, "epoch": 1916, "lr": 3.0466319396150694e-05}
{"train_loss": 0.05619814619421959, "global_step": 170591, "epoch": 1916, "lr": 3.0465785732954023e-05}
{"train_loss": 0.05274457484483719, "global_step": 170592, "epoch": 1916, "lr": 3.0465252072383465e-05}
{"train_loss": 0.09928586333990097, "global_step": 170593, "epoch": 1916, "lr": 3.046471841443911e-05}
{"train_loss": 0.04261702671647072, "global_step": 170594, "epoch": 1916, "lr": 3.0464184759121063e-05}
{"train_loss": 0.05961025133728981, "global_step": 170595, "epoch": 1916, "lr": 3.0463651106429348e-05}
{"train_loss": 0.07806114107370377, "global_step": 170596, "epoch": 1916, "lr": 3.0463117456364075e-05}
{"train_loss": 0.07151539623737335, "global_step": 170597, "epoch": 1916, "lr": 3.0462583808925283e-05}
{"train_loss": 0.07712912559509277, "global_step": 170598, "epoch": 1916, "lr": 3.046205016411307e-05}
{"train_loss": 0.09863972663879395, "global_step": 170599, "epoch": 1916, "lr": 3.046151652192748e-05}
{"train_loss": 0.0960625559091568, "global_step": 170600, "epoch": 1916, "lr": 3.046098288236862e-05}
{"train_loss": 0.07718360424041748, "global_step": 170601, "epoch": 1916, "lr": 3.0460449245436522e-05}
{"train_loss": 0.1019970178604126, "global_step": 170602, "epoch": 1916, "lr": 3.04599156111313e-05}
{"train_loss": 0.04418610408902168, "global_step": 170603, "epoch": 1916, "lr": 3.0459381979452982e-05}
{"train_loss": 0.08951237797737122, "global_step": 170604, "epoch": 1916, "lr": 3.0458848350401693e-05}
{"train_loss": 0.10050847381353378, "global_step": 170605, "epoch": 1916, "lr": 3.0458314723977455e-05}
{"train_loss": 0.06828003376722336, "global_step": 170606, "epoch": 1916, "lr": 3.0457781100180355e-05}
{"train_loss": 0.09345097094774246, "global_step": 170607, "epoch": 1916, "lr": 3.045724747901047e-05}
{"train_loss": 0.06622421741485596, "global_step": 170608, "epoch": 1916, "lr": 3.0456713860467874e-05}
{"train_loss": 0.12220944464206696, "global_step": 170609, "epoch": 1916, "lr": 3.045618024455263e-05}
{"train_loss": 0.12310934066772461, "global_step": 170610, "epoch": 1916, "lr": 3.0455646631264805e-05}
{"train_loss": 0.06586825847625732, "global_step": 170611, "epoch": 1916, "lr": 3.0455113020604505e-05}
{"train_loss": 0.0854550443505973, "global_step": 170612, "epoch": 1916, "lr": 3.0454579412571748e-05, "val_loss": 7.229586124420166}
{"train_loss": 0.11415441334247589, "global_step": 170613, "epoch": 1917, "lr": 3.0454045807166664e-05}
{"train_loss": 0.11229422688484192, "global_step": 170614, "epoch": 1917, "lr": 3.045351220438927e-05}
{"train_loss": 0.08323242515325546, "global_step": 170615, "epoch": 1917, "lr": 3.0452978604239685e-05}
{"train_loss": 0.09958222508430481, "global_step": 170616, "epoch": 1917, "lr": 3.0452445006717933e-05}
{"train_loss": 0.08514507114887238, "global_step": 170617, "epoch": 1917, "lr": 3.045191141182414e-05}
{"train_loss": 0.09781066328287125, "global_step": 170618, "epoch": 1917, "lr": 3.0451377819558324e-05}
{"train_loss": 0.08163946866989136, "global_step": 170619, "epoch": 1917, "lr": 3.0450844229920606e-05}
{"train_loss": 0.13672852516174316, "global_step": 170620, "epoch": 1917, "lr": 3.0450310642911014e-05}
{"train_loss": 0.061438675969839096, "global_step": 170621, "epoch": 1917, "lr": 3.0449777058529656e-05}
{"train_loss": 0.09275001287460327, "global_step": 170622, "epoch": 1917, "lr": 3.0449243476776568e-05}
{"train_loss": 0.13931140303611755, "global_step": 170623, "epoch": 1917, "lr": 3.044870989765186e-05}
{"train_loss": 0.14508412778377533, "global_step": 170624, "epoch": 1917, "lr": 3.0448176321155563e-05}
{"train_loss": 0.11181774735450745, "global_step": 170625, "epoch": 1917, "lr": 3.0447642747287795e-05}
{"train_loss": 0.03848046064376831, "global_step": 170626, "epoch": 1917, "lr": 3.0447109176048595e-05}
{"train_loss": 0.16163471341133118, "global_step": 170627, "epoch": 1917, "lr": 3.0446575607438032e-05}
{"train_loss": 0.1275404542684555, "global_step": 170628, "epoch": 1917, "lr": 3.0446042041456203e-05}
{"train_loss": 0.059943974018096924, "global_step": 170629, "epoch": 1917, "lr": 3.0445508478103146e-05}
{"train_loss": 0.06639501452445984, "global_step": 170630, "epoch": 1917, "lr": 3.0444974917378976e-05}
{"train_loss": 0.054235778748989105, "global_step": 170631, "epoch": 1917, "lr": 3.0444441359283714e-05}
{"train_loss": 0.06983034312725067, "global_step": 170632, "epoch": 1917, "lr": 3.0443907803817483e-05}
{"train_loss": 0.08675829321146011, "global_step": 170633, "epoch": 1917, "lr": 3.044337425098031e-05}
{"train_loss": 0.09073857218027115, "global_step": 170634, "epoch": 1917, "lr": 3.044284070077231e-05}
{"train_loss": 0.13062457740306854, "global_step": 170635, "epoch": 1917, "lr": 3.0442307153193505e-05}
{"train_loss": 0.08770830184221268, "global_step": 170636, "epoch": 1917, "lr": 3.044177360824402e-05}
{"train_loss": 0.07716190069913864, "global_step": 170637, "epoch": 1917, "lr": 3.0441240065923882e-05}
{"train_loss": 0.09781577438116074, "global_step": 170638, "epoch": 1917, "lr": 3.04407065262332e-05}
{"train_loss": 0.08446428179740906, "global_step": 170639, "epoch": 1917, "lr": 3.0440172989172e-05}
{"train_loss": 0.04937169700860977, "global_step": 170640, "epoch": 1917, "lr": 3.0439639454740398e-05}
{"train_loss": 0.06084248423576355, "global_step": 170641, "epoch": 1917, "lr": 3.0439105922938438e-05}
{"train_loss": 0.08744604885578156, "global_step": 170642, "epoch": 1917, "lr": 3.0438572393766217e-05}
{"train_loss": 0.097211554646492, "global_step": 170643, "epoch": 1917, "lr": 3.0438038867223773e-05}
{"train_loss": 0.08096697181463242, "global_step": 170644, "epoch": 1917, "lr": 3.0437505343311213e-05}
{"train_loss": 0.08028022944927216, "global_step": 170645, "epoch": 1917, "lr": 3.0436971822028583e-05}
{"train_loss": 0.07401253283023834, "global_step": 170646, "epoch": 1917, "lr": 3.043643830337596e-05}
{"train_loss": 0.10963016003370285, "global_step": 170647, "epoch": 1917, "lr": 3.043590478735343e-05}
{"train_loss": 0.04099975526332855, "global_step": 170648, "epoch": 1917, "lr": 3.0435371273961037e-05}
{"train_loss": 0.07485885173082352, "global_step": 170649, "epoch": 1917, "lr": 3.0434837763198898e-05}
{"train_loss": 0.07295838743448257, "global_step": 170650, "epoch": 1917, "lr": 3.0434304255067027e-05}
{"train_loss": 0.06887828558683395, "global_step": 170651, "epoch": 1917, "lr": 3.0433770749565548e-05}
{"train_loss": 0.08880524337291718, "global_step": 170652, "epoch": 1917, "lr": 3.043323724669449e-05}
{"train_loss": 0.08400309830904007, "global_step": 170653, "epoch": 1917, "lr": 3.043270374645396e-05}
{"train_loss": 0.12017993628978729, "global_step": 170654, "epoch": 1917, "lr": 3.0432170248844007e-05}
{"train_loss": 0.07729887217283249, "global_step": 170655, "epoch": 1917, "lr": 3.0431636753864724e-05}
{"train_loss": 0.06943900138139725, "global_step": 170656, "epoch": 1917, "lr": 3.0431103261516146e-05}
{"train_loss": 0.06881886720657349, "global_step": 170657, "epoch": 1917, "lr": 3.0430569771798388e-05}
{"train_loss": 0.10617763549089432, "global_step": 170658, "epoch": 1917, "lr": 3.0430036284711484e-05}
{"train_loss": 0.12038473784923553, "global_step": 170659, "epoch": 1917, "lr": 3.042950280025554e-05}
{"train_loss": 0.05874297767877579, "global_step": 170660, "epoch": 1917, "lr": 3.042896931843059e-05}
{"train_loss": 0.04899563267827034, "global_step": 170661, "epoch": 1917, "lr": 3.042843583923675e-05}
{"train_loss": 0.09437740594148636, "global_step": 170662, "epoch": 1917, "lr": 3.042790236267405e-05}
{"train_loss": 0.05880081653594971, "global_step": 170663, "epoch": 1917, "lr": 3.0427368888742596e-05}
{"train_loss": 0.07555992156267166, "global_step": 170664, "epoch": 1917, "lr": 3.0426835417442435e-05}
{"train_loss": 0.04978390783071518, "global_step": 170665, "epoch": 1917, "lr": 3.042630194877364e-05}
{"train_loss": 0.0845438539981842, "global_step": 170666, "epoch": 1917, "lr": 3.0425768482736306e-05}
{"train_loss": 0.05924452096223831, "global_step": 170667, "epoch": 1917, "lr": 3.0425235019330468e-05}
{"train_loss": 0.07740882784128189, "global_step": 170668, "epoch": 1917, "lr": 3.0424701558556246e-05}
{"train_loss": 0.09488668292760849, "global_step": 170669, "epoch": 1917, "lr": 3.0424168100413657e-05}
{"train_loss": 0.056960564106702805, "global_step": 170670, "epoch": 1917, "lr": 3.042363464490282e-05}
{"train_loss": 0.04500085487961769, "global_step": 170671, "epoch": 1917, "lr": 3.042310119202377e-05}
{"train_loss": 0.07381480187177658, "global_step": 170672, "epoch": 1917, "lr": 3.042256774177662e-05}
{"train_loss": 0.08411377668380737, "global_step": 170673, "epoch": 1917, "lr": 3.0422034294161387e-05}
{"train_loss": 0.08075297623872757, "global_step": 170674, "epoch": 1917, "lr": 3.0421500849178197e-05}
{"train_loss": 0.08036195486783981, "global_step": 170675, "epoch": 1917, "lr": 3.0420967406827084e-05}
{"train_loss": 0.03993629291653633, "global_step": 170676, "epoch": 1917, "lr": 3.0420433967108143e-05}
{"train_loss": 0.07384973764419556, "global_step": 170677, "epoch": 1917, "lr": 3.0419900530021424e-05}
{"train_loss": 0.044826485216617584, "global_step": 170678, "epoch": 1917, "lr": 3.0419367095567026e-05}
{"train_loss": 0.1687406599521637, "global_step": 170679, "epoch": 1917, "lr": 3.0418833663744983e-05}
{"train_loss": 0.06519164144992828, "global_step": 170680, "epoch": 1917, "lr": 3.0418300234555413e-05}
{"train_loss": 0.08876702934503555, "global_step": 170681, "epoch": 1917, "lr": 3.0417766807998348e-05}
{"train_loss": 0.05405110865831375, "global_step": 170682, "epoch": 1917, "lr": 3.0417233384073877e-05}
{"train_loss": 0.10617277026176453, "global_step": 170683, "epoch": 1917, "lr": 3.0416699962782087e-05}
{"train_loss": 0.06893938034772873, "global_step": 170684, "epoch": 1917, "lr": 3.041616654412301e-05}
{"train_loss": 0.07130357623100281, "global_step": 170685, "epoch": 1917, "lr": 3.041563312809676e-05}
{"train_loss": 0.07219871878623962, "global_step": 170686, "epoch": 1917, "lr": 3.0415099714703372e-05}
{"train_loss": 0.08780240267515182, "global_step": 170687, "epoch": 1917, "lr": 3.0414566303942956e-05}
{"train_loss": 0.04756496846675873, "global_step": 170688, "epoch": 1917, "lr": 3.0414032895815538e-05}
{"train_loss": 0.07919250428676605, "global_step": 170689, "epoch": 1917, "lr": 3.041349949032124e-05}
{"train_loss": 0.14887765049934387, "global_step": 170690, "epoch": 1917, "lr": 3.041296608746008e-05}
{"train_loss": 0.03693784773349762, "global_step": 170691, "epoch": 1917, "lr": 3.0412432687232184e-05}
{"train_loss": 0.08610889315605164, "global_step": 170692, "epoch": 1917, "lr": 3.041189928963758e-05}
{"train_loss": 0.07926354557275772, "global_step": 170693, "epoch": 1917, "lr": 3.0411365894676375e-05}
{"train_loss": 0.051376815885305405, "global_step": 170694, "epoch": 1917, "lr": 3.04108325023486e-05}
{"train_loss": 0.08788812905550003, "global_step": 170695, "epoch": 1917, "lr": 3.0410299112654372e-05}
{"train_loss": 0.07043579965829849, "global_step": 170696, "epoch": 1917, "lr": 3.040976572559372e-05}
{"train_loss": 0.032595593482255936, "global_step": 170697, "epoch": 1917, "lr": 3.040923234116676e-05}
{"train_loss": 0.08154109120368958, "global_step": 170698, "epoch": 1917, "lr": 3.0408698959373514e-05}
{"train_loss": 0.1080498918890953, "global_step": 170699, "epoch": 1917, "lr": 3.040816558021411e-05}
{"train_loss": 0.11218814551830292, "global_step": 170700, "epoch": 1917, "lr": 3.0407632203688558e-05}
{"train_loss": 0.08281798949569799, "global_step": 170701, "epoch": 1917, "lr": 3.0407098829796976e-05, "val_loss": 7.1962690353393555}
{"train_loss": 0.10491891205310822, "global_step": 170702, "epoch": 1918, "lr": 3.0406565458539428e-05}
{"train_loss": 0.06350687146186829, "global_step": 170703, "epoch": 1918, "lr": 3.040603208991596e-05}
{"train_loss": 0.16037452220916748, "global_step": 170704, "epoch": 1918, "lr": 3.0405498723926683e-05}
{"train_loss": 0.03688018396496773, "global_step": 170705, "epoch": 1918, "lr": 3.040496536057163e-05}
{"train_loss": 0.08925826102495193, "global_step": 170706, "epoch": 1918, "lr": 3.0404431999850906e-05}
{"train_loss": 0.06521809101104736, "global_step": 170707, "epoch": 1918, "lr": 3.040389864176455e-05}
{"train_loss": 0.06473395228385925, "global_step": 170708, "epoch": 1918, "lr": 3.040336528631267e-05}
{"train_loss": 0.09105649590492249, "global_step": 170709, "epoch": 1918, "lr": 3.04028319334953e-05}
{"train_loss": 0.03907628357410431, "global_step": 170710, "epoch": 1918, "lr": 3.040229858331255e-05}
{"train_loss": 0.07416431605815887, "global_step": 170711, "epoch": 1918, "lr": 3.040176523576445e-05}
{"train_loss": 0.06981127709150314, "global_step": 170712, "epoch": 1918, "lr": 3.0401231890851118e-05}
{"train_loss": 0.11401696503162384, "global_step": 170713, "epoch": 1918, "lr": 3.0400698548572577e-05}
{"train_loss": 0.10413658618927002, "global_step": 170714, "epoch": 1918, "lr": 3.0400165208928945e-05}
{"train_loss": 0.08018828183412552, "global_step": 170715, "epoch": 1918, "lr": 3.0399631871920253e-05}
{"train_loss": 0.11992718279361725, "global_step": 170716, "epoch": 1918, "lr": 3.039909853754661e-05}
{"train_loss": 0.05905573070049286, "global_step": 170717, "epoch": 1918, "lr": 3.039856520580805e-05}
{"train_loss": 0.09578040987253189, "global_step": 170718, "epoch": 1918, "lr": 3.039803187670468e-05}
{"train_loss": 0.06099323555827141, "global_step": 170719, "epoch": 1918, "lr": 3.0397498550236542e-05}
{"train_loss": 0.09541818499565125, "global_step": 170720, "epoch": 1918, "lr": 3.039696522640373e-05}
{"train_loss": 0.06478491425514221, "global_step": 170721, "epoch": 1918, "lr": 3.0396431905206314e-05}
{"train_loss": 0.06854165345430374, "global_step": 170722, "epoch": 1918, "lr": 3.0395898586644346e-05}
{"train_loss": 0.05028925836086273, "global_step": 170723, "epoch": 1918, "lr": 3.0395365270717925e-05}
{"train_loss": 0.08635342121124268, "global_step": 170724, "epoch": 1918, "lr": 3.0394831957427084e-05}
{"train_loss": 0.03186386823654175, "global_step": 170725, "epoch": 1918, "lr": 3.0394298646771945e-05}
{"train_loss": 0.1462448090314865, "global_step": 170726, "epoch": 1918, "lr": 3.0393765338752533e-05}
{"train_loss": 0.11765211820602417, "global_step": 170727, "epoch": 1918, "lr": 3.0393232033368958e-05}
{"train_loss": 0.10865209996700287, "global_step": 170728, "epoch": 1918, "lr": 3.039269873062125e-05}
{"train_loss": 0.09879381209611893, "global_step": 170729, "epoch": 1918, "lr": 3.0392165430509535e-05}
{"train_loss": 0.08727724850177765, "global_step": 170730, "epoch": 1918, "lr": 3.0391632133033832e-05}
{"train_loss": 0.05376623198390007, "global_step": 170731, "epoch": 1918, "lr": 3.039109883819425e-05}
{"train_loss": 0.049651455134153366, "global_step": 170732, "epoch": 1918, "lr": 3.0390565545990822e-05}
{"train_loss": 0.08953602612018585, "global_step": 170733, "epoch": 1918, "lr": 3.0390032256423673e-05}
{"train_loss": 0.12968692183494568, "global_step": 170734, "epoch": 1918, "lr": 3.0389498969492826e-05}
{"train_loss": 0.07230652868747711, "global_step": 170735, "epoch": 1918, "lr": 3.0388965685198383e-05}
{"train_loss": 0.06201741844415665, "global_step": 170736, "epoch": 1918, "lr": 3.038843240354039e-05}
{"train_loss": 0.09381774812936783, "global_step": 170737, "epoch": 1918, "lr": 3.0387899124518944e-05}
{"train_loss": 0.05095363408327103, "global_step": 170738, "epoch": 1918, "lr": 3.0387365848134104e-05}
{"train_loss": 0.1243910863995552, "global_step": 170739, "epoch": 1918, "lr": 3.0386832574385943e-05}
{"train_loss": 0.04247162491083145, "global_step": 170740, "epoch": 1918, "lr": 3.0386299303274534e-05}
{"train_loss": 0.10835640132427216, "global_step": 170741, "epoch": 1918, "lr": 3.0385766034799945e-05}
{"train_loss": 0.03411610424518585, "global_step": 170742, "epoch": 1918, "lr": 3.038523276896227e-05}
{"train_loss": 0.09415531158447266, "global_step": 170743, "epoch": 1918, "lr": 3.038469950576153e-05}
{"train_loss": 0.10131880640983582, "global_step": 170744, "epoch": 1918, "lr": 3.0384166245197854e-05}
{"train_loss": 0.05810052528977394, "global_step": 170745, "epoch": 1918, "lr": 3.0383632987271272e-05}
{"train_loss": 0.09718901664018631, "global_step": 170746, "epoch": 1918, "lr": 3.038309973198188e-05}
{"train_loss": 0.10519925504922867, "global_step": 170747, "epoch": 1918, "lr": 3.0382566479329734e-05}
{"train_loss": 0.0688791498541832, "global_step": 170748, "epoch": 1918, "lr": 3.038203322931492e-05}
{"train_loss": 0.11918315291404724, "global_step": 170749, "epoch": 1918, "lr": 3.038149998193749e-05}
{"train_loss": 0.09232134371995926, "global_step": 170750, "epoch": 1918, "lr": 3.0380966737197548e-05}
{"train_loss": 0.06136083975434303, "global_step": 170751, "epoch": 1918, "lr": 3.038043349509513e-05}
{"train_loss": 0.034544989466667175, "global_step": 170752, "epoch": 1918, "lr": 3.0379900255630332e-05}
{"train_loss": 0.12110692262649536, "global_step": 170753, "epoch": 1918, "lr": 3.0379367018803205e-05}
{"train_loss": 0.05088059976696968, "global_step": 170754, "epoch": 1918, "lr": 3.0378833784613854e-05}
{"train_loss": 0.10363154113292694, "global_step": 170755, "epoch": 1918, "lr": 3.0378300553062306e-05}
{"train_loss": 0.08980685472488403, "global_step": 170756, "epoch": 1918, "lr": 3.0377767324148666e-05}
{"train_loss": 0.07106967270374298, "global_step": 170757, "epoch": 1918, "lr": 3.0377234097872998e-05}
{"train_loss": 0.0835634097456932, "global_step": 170758, "epoch": 1918, "lr": 3.0376700874235377e-05}
{"train_loss": 0.051199886947870255, "global_step": 170759, "epoch": 1918, "lr": 3.0376167653235867e-05}
{"train_loss": 0.08842598646879196, "global_step": 170760, "epoch": 1918, "lr": 3.0375634434874532e-05}
{"train_loss": 0.07673674821853638, "global_step": 170761, "epoch": 1918, "lr": 3.037510121915147e-05}
{"train_loss": 0.07218543440103531, "global_step": 170762, "epoch": 1918, "lr": 3.0374568006066716e-05}
{"train_loss": 0.08804577589035034, "global_step": 170763, "epoch": 1918, "lr": 3.037403479562039e-05}
{"train_loss": 0.13057804107666016, "global_step": 170764, "epoch": 1918, "lr": 3.0373501587812503e-05}
{"train_loss": 0.10970617830753326, "global_step": 170765, "epoch": 1918, "lr": 3.0372968382643185e-05}
{"train_loss": 0.09189053624868393, "global_step": 170766, "epoch": 1918, "lr": 3.037243518011247e-05}
{"train_loss": 0.11448932439088821, "global_step": 170767, "epoch": 1918, "lr": 3.037190198022045e-05}
{"train_loss": 0.09509778767824173, "global_step": 170768, "epoch": 1918, "lr": 3.0371368782967174e-05}
{"train_loss": 0.10451342165470123, "global_step": 170769, "epoch": 1918, "lr": 3.0370835588352754e-05}
{"train_loss": 0.07758764922618866, "global_step": 170770, "epoch": 1918, "lr": 3.0370302396377204e-05}
{"train_loss": 0.09801216423511505, "global_step": 170771, "epoch": 1918, "lr": 3.0369769207040653e-05}
{"train_loss": 0.060696739703416824, "global_step": 170772, "epoch": 1918, "lr": 3.036923602034313e-05}
{"train_loss": 0.0710134506225586, "global_step": 170773, "epoch": 1918, "lr": 3.0368702836284717e-05}
{"train_loss": 0.10937455296516418, "global_step": 170774, "epoch": 1918, "lr": 3.036816965486552e-05}
{"train_loss": 0.11591806262731552, "global_step": 170775, "epoch": 1918, "lr": 3.036763647608557e-05}
{"train_loss": 0.08297628909349442, "global_step": 170776, "epoch": 1918, "lr": 3.0367103299944955e-05}
{"train_loss": 0.05780135095119476, "global_step": 170777, "epoch": 1918, "lr": 3.036657012644374e-05}
{"train_loss": 0.09284336864948273, "global_step": 170778, "epoch": 1918, "lr": 3.0366036955582e-05}
{"train_loss": 0.08749746531248093, "global_step": 170779, "epoch": 1918, "lr": 3.0365503787359806e-05}
{"train_loss": 0.09887022525072098, "global_step": 170780, "epoch": 1918, "lr": 3.0364970621777243e-05}
{"train_loss": 0.07043149322271347, "global_step": 170781, "epoch": 1918, "lr": 3.036443745883435e-05}
{"train_loss": 0.11791668832302094, "global_step": 170782, "epoch": 1918, "lr": 3.0363904298531243e-05}
{"train_loss": 0.07340606302022934, "global_step": 170783, "epoch": 1918, "lr": 3.0363371140867947e-05}
{"train_loss": 0.101250559091568, "global_step": 170784, "epoch": 1918, "lr": 3.0362837985844572e-05}
{"train_loss": 0.06720323115587234, "global_step": 170785, "epoch": 1918, "lr": 3.036230483346116e-05}
{"train_loss": 0.1096985787153244, "global_step": 170786, "epoch": 1918, "lr": 3.036177168371781e-05}
{"train_loss": 0.07371615618467331, "global_step": 170787, "epoch": 1918, "lr": 3.0361238536614566e-05}
{"train_loss": 0.08562101423740387, "global_step": 170788, "epoch": 1918, "lr": 3.0360705392151534e-05}
{"train_loss": 0.1440146267414093, "global_step": 170789, "epoch": 1918, "lr": 3.0360172250328744e-05}
{"train_loss": 0.08602189728885554, "global_step": 170790, "epoch": 1918, "lr": 3.035963911114631e-05, "val_loss": 7.365749835968018}
{"train_loss": 0.10430988669395447, "global_step": 170791, "epoch": 1919, "lr": 3.0359105974604263e-05}
{"train_loss": 0.14856547117233276, "global_step": 170792, "epoch": 1919, "lr": 3.03585728407027e-05}
{"train_loss": 0.04654340073466301, "global_step": 170793, "epoch": 1919, "lr": 3.0358039709441698e-05}
{"train_loss": 0.10866422951221466, "global_step": 170794, "epoch": 1919, "lr": 3.0357506580821315e-05}
{"train_loss": 0.03675178810954094, "global_step": 170795, "epoch": 1919, "lr": 3.035697345484162e-05}
{"train_loss": 0.09303819388151169, "global_step": 170796, "epoch": 1919, "lr": 3.03564403315027e-05}
{"train_loss": 0.05019456148147583, "global_step": 170797, "epoch": 1919, "lr": 3.0355907210804614e-05}
{"train_loss": 0.05233495309948921, "global_step": 170798, "epoch": 1919, "lr": 3.035537409274743e-05}
{"train_loss": 0.06055004522204399, "global_step": 170799, "epoch": 1919, "lr": 3.035484097733124e-05}
{"train_loss": 0.13425101339817047, "global_step": 170800, "epoch": 1919, "lr": 3.0354307864556088e-05}
{"train_loss": 0.1276538223028183, "global_step": 170801, "epoch": 1919, "lr": 3.035377475442207e-05}
{"train_loss": 0.05007519945502281, "global_step": 170802, "epoch": 1919, "lr": 3.035324164692923e-05}
{"train_loss": 0.08812477439641953, "global_step": 170803, "epoch": 1919, "lr": 3.0352708542077678e-05}
{"train_loss": 0.08776643127202988, "global_step": 170804, "epoch": 1919, "lr": 3.0352175439867447e-05}
{"train_loss": 0.06550461053848267, "global_step": 170805, "epoch": 1919, "lr": 3.035164234029865e-05}
{"train_loss": 0.04286205768585205, "global_step": 170806, "epoch": 1919, "lr": 3.0351109243371312e-05}
{"train_loss": 0.07146929204463959, "global_step": 170807, "epoch": 1919, "lr": 3.0350576149085545e-05}
{"train_loss": 0.062227122485637665, "global_step": 170808, "epoch": 1919, "lr": 3.0350043057441384e-05}
{"train_loss": 0.07385076582431793, "global_step": 170809, "epoch": 1919, "lr": 3.0349509968438928e-05}
{"train_loss": 0.08238077163696289, "global_step": 170810, "epoch": 1919, "lr": 3.034897688207825e-05}
{"train_loss": 0.0970032662153244, "global_step": 170811, "epoch": 1919, "lr": 3.0348443798359395e-05}
{"train_loss": 0.06751738488674164, "global_step": 170812, "epoch": 1919, "lr": 3.034791071728248e-05}
{"train_loss": 0.0955236405134201, "global_step": 170813, "epoch": 1919, "lr": 3.034737763884753e-05}
{"train_loss": 0.06888982653617859, "global_step": 170814, "epoch": 1919, "lr": 3.0346844563054638e-05}
{"train_loss": 0.15691061317920685, "global_step": 170815, "epoch": 1919, "lr": 3.0346311489903878e-05}
{"train_loss": 0.1145179346203804, "global_step": 170816, "epoch": 1919, "lr": 3.034577841939531e-05}
{"train_loss": 0.09376820176839828, "global_step": 170817, "epoch": 1919, "lr": 3.0345245351529016e-05}
{"train_loss": 0.1253305971622467, "global_step": 170818, "epoch": 1919, "lr": 3.034471228630508e-05}
{"train_loss": 0.08112925291061401, "global_step": 170819, "epoch": 1919, "lr": 3.0344179223723525e-05}
{"train_loss": 0.0658678486943245, "global_step": 170820, "epoch": 1919, "lr": 3.0343646163784477e-05}
{"train_loss": 0.08823899179697037, "global_step": 170821, "epoch": 1919, "lr": 3.0343113106487976e-05}
{"train_loss": 0.05963341146707535, "global_step": 170822, "epoch": 1919, "lr": 3.034258005183412e-05}
{"train_loss": 0.12461014091968536, "global_step": 170823, "epoch": 1919, "lr": 3.0342046999822943e-05}
{"train_loss": 0.12579117715358734, "global_step": 170824, "epoch": 1919, "lr": 3.034151395045456e-05}
{"train_loss": 0.16285806894302368, "global_step": 170825, "epoch": 1919, "lr": 3.0340980903729e-05}
{"train_loss": 0.07430703938007355, "global_step": 170826, "epoch": 1919, "lr": 3.034044785964637e-05}
{"train_loss": 0.060471270233392715, "global_step": 170827, "epoch": 1919, "lr": 3.0339914818206718e-05}
{"train_loss": 0.16534844040870667, "global_step": 170828, "epoch": 1919, "lr": 3.0339381779410114e-05}
{"train_loss": 0.0971047431230545, "global_step": 170829, "epoch": 1919, "lr": 3.0338848743256666e-05}
{"train_loss": 0.1248757541179657, "global_step": 170830, "epoch": 1919, "lr": 3.03383157097464e-05}
{"train_loss": 0.0580623596906662, "global_step": 170831, "epoch": 1919, "lr": 3.0337782678879428e-05}
{"train_loss": 0.11832980066537857, "global_step": 170832, "epoch": 1919, "lr": 3.033724965065579e-05}
{"train_loss": 0.030558783560991287, "global_step": 170833, "epoch": 1919, "lr": 3.0336716625075574e-05}
{"train_loss": 0.13477444648742676, "global_step": 170834, "epoch": 1919, "lr": 3.0336183602138845e-05}
{"train_loss": 0.10346432775259018, "global_step": 170835, "epoch": 1919, "lr": 3.033565058184567e-05}
{"train_loss": 0.16760601103305817, "global_step": 170836, "epoch": 1919, "lr": 3.0335117564196137e-05}
{"train_loss": 0.12951979041099548, "global_step": 170837, "epoch": 1919, "lr": 3.033458454919031e-05}
{"train_loss": 0.13397681713104248, "global_step": 170838, "epoch": 1919, "lr": 3.033405153682824e-05}
{"train_loss": 0.07680003345012665, "global_step": 170839, "epoch": 1919, "lr": 3.033351852711004e-05}
{"train_loss": 0.08516135811805725, "global_step": 170840, "epoch": 1919, "lr": 3.0332985520035734e-05}
{"train_loss": 0.06082553043961525, "global_step": 170841, "epoch": 1919, "lr": 3.0332452515605442e-05}
{"train_loss": 0.10282377153635025, "global_step": 170842, "epoch": 1919, "lr": 3.033191951381919e-05}
{"train_loss": 0.08618725091218948, "global_step": 170843, "epoch": 1919, "lr": 3.0331386514677094e-05}
{"train_loss": 0.08670954406261444, "global_step": 170844, "epoch": 1919, "lr": 3.033085351817918e-05}
{"train_loss": 0.11066215485334396, "global_step": 170845, "epoch": 1919, "lr": 3.0330320524325562e-05}
{"train_loss": 0.14142835140228271, "global_step": 170846, "epoch": 1919, "lr": 3.0329787533116276e-05}
{"train_loss": 0.09445775300264359, "global_step": 170847, "epoch": 1919, "lr": 3.0329254544551412e-05}
{"train_loss": 0.0702679231762886, "global_step": 170848, "epoch": 1919, "lr": 3.0328721558631057e-05}
{"train_loss": 0.16327238082885742, "global_step": 170849, "epoch": 1919, "lr": 3.032818857535525e-05}
{"train_loss": 0.09384080022573471, "global_step": 170850, "epoch": 1919, "lr": 3.032765559472409e-05}
{"train_loss": 0.11106238514184952, "global_step": 170851, "epoch": 1919, "lr": 3.0327122616737635e-05}
{"train_loss": 0.04694655165076256, "global_step": 170852, "epoch": 1919, "lr": 3.0326589641395952e-05}
{"train_loss": 0.12474332749843597, "global_step": 170853, "epoch": 1919, "lr": 3.0326056668699126e-05}
{"train_loss": 0.08228925615549088, "global_step": 170854, "epoch": 1919, "lr": 3.0325523698647218e-05}
{"train_loss": 0.10767409950494766, "global_step": 170855, "epoch": 1919, "lr": 3.0324990731240305e-05}
{"train_loss": 0.10241805016994476, "global_step": 170856, "epoch": 1919, "lr": 3.0324457766478454e-05}
{"train_loss": 0.06384909898042679, "global_step": 170857, "epoch": 1919, "lr": 3.0323924804361747e-05}
{"train_loss": 0.048173632472753525, "global_step": 170858, "epoch": 1919, "lr": 3.032339184489025e-05}
{"train_loss": 0.06892020255327225, "global_step": 170859, "epoch": 1919, "lr": 3.0322858888064022e-05}
{"train_loss": 0.0728096216917038, "global_step": 170860, "epoch": 1919, "lr": 3.0322325933883157e-05}
{"train_loss": 0.09327195584774017, "global_step": 170861, "epoch": 1919, "lr": 3.0321792982347697e-05}
{"train_loss": 0.09675920754671097, "global_step": 170862, "epoch": 1919, "lr": 3.0321260033457755e-05}
{"train_loss": 0.049439627677202225, "global_step": 170863, "epoch": 1919, "lr": 3.032072708721336e-05}
{"train_loss": 0.05152615159749985, "global_step": 170864, "epoch": 1919, "lr": 3.03201941436146e-05}
{"train_loss": 0.09356822073459625, "global_step": 170865, "epoch": 1919, "lr": 3.0319661202661574e-05}
{"train_loss": 0.05400475859642029, "global_step": 170866, "epoch": 1919, "lr": 3.0319128264354306e-05}
{"train_loss": 0.06878519803285599, "global_step": 170867, "epoch": 1919, "lr": 3.031859532869291e-05}
{"train_loss": 0.09246839582920074, "global_step": 170868, "epoch": 1919, "lr": 3.0318062395677428e-05}
{"train_loss": 0.06776417791843414, "global_step": 170869, "epoch": 1919, "lr": 3.031752946530796e-05}
{"train_loss": 0.07521902024745941, "global_step": 170870, "epoch": 1919, "lr": 3.031699653758453e-05}
{"train_loss": 0.1471005380153656, "global_step": 170871, "epoch": 1919, "lr": 3.0316463612507274e-05}
{"train_loss": 0.06760067492723465, "global_step": 170872, "epoch": 1919, "lr": 3.031593069007621e-05}
{"train_loss": 0.06959164887666702, "global_step": 170873, "epoch": 1919, "lr": 3.0315397770291433e-05}
{"train_loss": 0.09034060686826706, "global_step": 170874, "epoch": 1919, "lr": 3.031486485315301e-05}
{"train_loss": 0.08563794940710068, "global_step": 170875, "epoch": 1919, "lr": 3.031433193866101e-05}
{"train_loss": 0.09481866657733917, "global_step": 170876, "epoch": 1919, "lr": 3.0313799026815516e-05}
{"train_loss": 0.09681030362844467, "global_step": 170877, "epoch": 1919, "lr": 3.031326611761659e-05}
{"train_loss": 0.08454149961471558, "global_step": 170878, "epoch": 1919, "lr": 3.0312733211064292e-05}
{"train_loss": 0.09064277625653182, "global_step": 170879, "epoch": 1919, "lr": 3.0312200307158732e-05, "val_loss": 7.351597309112549}
{"train_loss": 0.06949540972709656, "global_step": 170880, "epoch": 1920, "lr": 3.031166740589993e-05}
{"train_loss": 0.08137966692447662, "global_step": 170881, "epoch": 1920, "lr": 3.0311134507288007e-05}
{"train_loss": 0.10237013548612595, "global_step": 170882, "epoch": 1920, "lr": 3.031060161132299e-05}
{"train_loss": 0.06524169445037842, "global_step": 170883, "epoch": 1920, "lr": 3.0310068718004976e-05}
{"train_loss": 0.09162360429763794, "global_step": 170884, "epoch": 1920, "lr": 3.0309535827334052e-05}
{"train_loss": 0.11449440568685532, "global_step": 170885, "epoch": 1920, "lr": 3.030900293931025e-05}
{"train_loss": 0.05595315247774124, "global_step": 170886, "epoch": 1920, "lr": 3.030847005393368e-05}
{"train_loss": 0.10691767185926437, "global_step": 170887, "epoch": 1920, "lr": 3.0307937171204383e-05}
{"train_loss": 0.07413062453269958, "global_step": 170888, "epoch": 1920, "lr": 3.030740429112246e-05}
{"train_loss": 0.07732459902763367, "global_step": 170889, "epoch": 1920, "lr": 3.0306871413687944e-05}
{"train_loss": 0.07947628200054169, "global_step": 170890, "epoch": 1920, "lr": 3.0306338538900954e-05}
{"train_loss": 0.07942880690097809, "global_step": 170891, "epoch": 1920, "lr": 3.030580566676152e-05}
{"train_loss": 0.07310297340154648, "global_step": 170892, "epoch": 1920, "lr": 3.0305272797269734e-05}
{"train_loss": 0.06221326068043709, "global_step": 170893, "epoch": 1920, "lr": 3.030473993042567e-05}
{"train_loss": 0.13722802698612213, "global_step": 170894, "epoch": 1920, "lr": 3.0304207066229384e-05}
{"train_loss": 0.08945740014314651, "global_step": 170895, "epoch": 1920, "lr": 3.0303674204680966e-05}
{"train_loss": 0.07178473472595215, "global_step": 170896, "epoch": 1920, "lr": 3.0303141345780482e-05}
{"train_loss": 0.07676303386688232, "global_step": 170897, "epoch": 1920, "lr": 3.0302608489527983e-05}
{"train_loss": 0.0950707495212555, "global_step": 170898, "epoch": 1920, "lr": 3.0302075635923576e-05}
{"train_loss": 0.07868551462888718, "global_step": 170899, "epoch": 1920, "lr": 3.0301542784967302e-05}
{"train_loss": 0.08243642747402191, "global_step": 170900, "epoch": 1920, "lr": 3.0301009936659243e-05}
{"train_loss": 0.03060196340084076, "global_step": 170901, "epoch": 1920, "lr": 3.030047709099949e-05}
{"train_loss": 0.12498442828655243, "global_step": 170902, "epoch": 1920, "lr": 3.0299944247988077e-05}
{"train_loss": 0.03737959265708923, "global_step": 170903, "epoch": 1920, "lr": 3.0299411407625123e-05}
{"train_loss": 0.05916851758956909, "global_step": 170904, "epoch": 1920, "lr": 3.0298878569910644e-05}
{"train_loss": 0.10421258211135864, "global_step": 170905, "epoch": 1920, "lr": 3.0298345734844762e-05}
{"train_loss": 0.07766158878803253, "global_step": 170906, "epoch": 1920, "lr": 3.0297812902427512e-05}
{"train_loss": 0.04044950753450394, "global_step": 170907, "epoch": 1920, "lr": 3.0297280072658995e-05}
{"train_loss": 0.06830228865146637, "global_step": 170908, "epoch": 1920, "lr": 3.029674724553926e-05}
{"train_loss": 0.05673724412918091, "global_step": 170909, "epoch": 1920, "lr": 3.0296214421068397e-05}
{"train_loss": 0.07752196490764618, "global_step": 170910, "epoch": 1920, "lr": 3.0295681599246462e-05}
{"train_loss": 0.11635953187942505, "global_step": 170911, "epoch": 1920, "lr": 3.0295148780073528e-05}
{"train_loss": 0.03884821757674217, "global_step": 170912, "epoch": 1920, "lr": 3.029461596354968e-05}
{"train_loss": 0.12332332134246826, "global_step": 170913, "epoch": 1920, "lr": 3.029408314967497e-05}
{"train_loss": 0.05519760027527809, "global_step": 170914, "epoch": 1920, "lr": 3.0293550338449484e-05}
{"train_loss": 0.10589435696601868, "global_step": 170915, "epoch": 1920, "lr": 3.02930175298733e-05}
{"train_loss": 0.09116724878549576, "global_step": 170916, "epoch": 1920, "lr": 3.0292484723946464e-05}
{"train_loss": 0.053060438483953476, "global_step": 170917, "epoch": 1920, "lr": 3.0291951920669082e-05}
{"train_loss": 0.07102271914482117, "global_step": 170918, "epoch": 1920, "lr": 3.0291419120041182e-05}
{"train_loss": 0.054517343640327454, "global_step": 170919, "epoch": 1920, "lr": 3.0290886322062862e-05}
{"train_loss": 0.051147446036338806, "global_step": 170920, "epoch": 1920, "lr": 3.0290353526734217e-05}
{"train_loss": 0.056584976613521576, "global_step": 170921, "epoch": 1920, "lr": 3.0289820734055263e-05}
{"train_loss": 0.08708427846431732, "global_step": 170922, "epoch": 1920, "lr": 3.0289287944026134e-05}
{"train_loss": 0.11332585662603378, "global_step": 170923, "epoch": 1920, "lr": 3.0288755156646842e-05}
{"train_loss": 0.08537862449884415, "global_step": 170924, "epoch": 1920, "lr": 3.0288222371917513e-05}
{"train_loss": 0.06993700563907623, "global_step": 170925, "epoch": 1920, "lr": 3.0287689589838164e-05}
{"train_loss": 0.06597869098186493, "global_step": 170926, "epoch": 1920, "lr": 3.0287156810408923e-05}
{"train_loss": 0.08103040605783463, "global_step": 170927, "epoch": 1920, "lr": 3.028662403362981e-05}
{"train_loss": 0.08011529594659805, "global_step": 170928, "epoch": 1920, "lr": 3.0286091259500944e-05}
{"train_loss": 0.08677232265472412, "global_step": 170929, "epoch": 1920, "lr": 3.0285558488022358e-05}
{"train_loss": 0.04502194747328758, "global_step": 170930, "epoch": 1920, "lr": 3.0285025719194138e-05}
{"train_loss": 0.10449318587779999, "global_step": 170931, "epoch": 1920, "lr": 3.0284492953016362e-05}
{"train_loss": 0.13058672845363617, "global_step": 170932, "epoch": 1920, "lr": 3.028396018948909e-05}
{"train_loss": 0.07365097105503082, "global_step": 170933, "epoch": 1920, "lr": 3.02834274286124e-05}
{"train_loss": 0.1078125461935997, "global_step": 170934, "epoch": 1920, "lr": 3.0282894670386375e-05}
{"train_loss": 0.055075086653232574, "global_step": 170935, "epoch": 1920, "lr": 3.0282361914811053e-05}
{"train_loss": 0.0990225151181221, "global_step": 170936, "epoch": 1920, "lr": 3.0281829161886527e-05}
{"train_loss": 0.13984447717666626, "global_step": 170937, "epoch": 1920, "lr": 3.0281296411612893e-05}
{"train_loss": 0.04171053320169449, "global_step": 170938, "epoch": 1920, "lr": 3.028076366399017e-05}
{"train_loss": 0.05267411842942238, "global_step": 170939, "epoch": 1920, "lr": 3.0280230919018483e-05}
{"train_loss": 0.07004441320896149, "global_step": 170940, "epoch": 1920, "lr": 3.0279698176697858e-05}
{"train_loss": 0.09722409397363663, "global_step": 170941, "epoch": 1920, "lr": 3.0279165437028405e-05}
{"train_loss": 0.09208546578884125, "global_step": 170942, "epoch": 1920, "lr": 3.0278632700010158e-05}
{"train_loss": 0.04877851530909538, "global_step": 170943, "epoch": 1920, "lr": 3.0278099965643236e-05}
{"train_loss": 0.09421845525503159, "global_step": 170944, "epoch": 1920, "lr": 3.027756723392765e-05}
{"train_loss": 0.0760551393032074, "global_step": 170945, "epoch": 1920, "lr": 3.0277034504863534e-05}
{"train_loss": 0.06603945046663284, "global_step": 170946, "epoch": 1920, "lr": 3.0276501778450912e-05}
{"train_loss": 0.11564436554908752, "global_step": 170947, "epoch": 1920, "lr": 3.027596905468989e-05}
{"train_loss": 0.0561722032725811, "global_step": 170948, "epoch": 1920, "lr": 3.0275436333580514e-05}
{"train_loss": 0.05962168052792549, "global_step": 170949, "epoch": 1920, "lr": 3.0274903615122864e-05}
{"train_loss": 0.13099291920661926, "global_step": 170950, "epoch": 1920, "lr": 3.027437089931702e-05}
{"train_loss": 0.11401967704296112, "global_step": 170951, "epoch": 1920, "lr": 3.0273838186163038e-05}
{"train_loss": 0.07309451699256897, "global_step": 170952, "epoch": 1920, "lr": 3.0273305475660995e-05}
{"train_loss": 0.06930600851774216, "global_step": 170953, "epoch": 1920, "lr": 3.0272772767810985e-05}
{"train_loss": 0.10971219837665558, "global_step": 170954, "epoch": 1920, "lr": 3.0272240062613032e-05}
{"train_loss": 0.062315862625837326, "global_step": 170955, "epoch": 1920, "lr": 3.0271707360067247e-05}
{"train_loss": 0.120221346616745, "global_step": 170956, "epoch": 1920, "lr": 3.0271174660173703e-05}
{"train_loss": 0.14235474169254303, "global_step": 170957, "epoch": 1920, "lr": 3.0270641962932435e-05}
{"train_loss": 0.14598581194877625, "global_step": 170958, "epoch": 1920, "lr": 3.0270109268343566e-05}
{"train_loss": 0.096269890666008, "global_step": 170959, "epoch": 1920, "lr": 3.026957657640712e-05}
{"train_loss": 0.13810327649116516, "global_step": 170960, "epoch": 1920, "lr": 3.0269043887123205e-05}
{"train_loss": 0.13828861713409424, "global_step": 170961, "epoch": 1920, "lr": 3.0268511200491855e-05}
{"train_loss": 0.09230615198612213, "global_step": 170962, "epoch": 1920, "lr": 3.026797851651319e-05}
{"train_loss": 0.0964265838265419, "global_step": 170963, "epoch": 1920, "lr": 3.026744583518723e-05}
{"train_loss": 0.05515411123633385, "global_step": 170964, "epoch": 1920, "lr": 3.0266913156514093e-05}
{"train_loss": 0.09827867895364761, "global_step": 170965, "epoch": 1920, "lr": 3.0266380480493806e-05}
{"train_loss": 0.0660492405295372, "global_step": 170966, "epoch": 1920, "lr": 3.026584780712649e-05}
{"train_loss": 0.10615326464176178, "global_step": 170967, "epoch": 1920, "lr": 3.0265315136412163e-05}
{"train_loss": 0.08398164612021339, "global_step": 170968, "epoch": 1920, "lr": 3.026478246835095e-05, "val_loss": 7.281038761138916, "train_action_mse_error": 18.794239044189453}
{"train_loss": 0.085770383477211, "global_step": 170969, "epoch": 1921, "lr": 3.0264249802942885e-05}
{"train_loss": 0.1464831382036209, "global_step": 170970, "epoch": 1921, "lr": 3.0263717140188052e-05}
{"train_loss": 0.06704092025756836, "global_step": 170971, "epoch": 1921, "lr": 3.0263184480086516e-05}
{"train_loss": 0.10342418402433395, "global_step": 170972, "epoch": 1921, "lr": 3.026265182263836e-05}
{"train_loss": 0.15851932764053345, "global_step": 170973, "epoch": 1921, "lr": 3.026211916784365e-05}
{"train_loss": 0.0839335024356842, "global_step": 170974, "epoch": 1921, "lr": 3.0261586515702444e-05}
{"train_loss": 0.11275763809680939, "global_step": 170975, "epoch": 1921, "lr": 3.026105386621485e-05}
{"train_loss": 0.0898946225643158, "global_step": 170976, "epoch": 1921, "lr": 3.0260521219380893e-05}
{"train_loss": 0.10303563624620438, "global_step": 170977, "epoch": 1921, "lr": 3.0259988575200693e-05}
{"train_loss": 0.08231965452432632, "global_step": 170978, "epoch": 1921, "lr": 3.0259455933674274e-05}
{"train_loss": 0.059896357357501984, "global_step": 170979, "epoch": 1921, "lr": 3.025892329480175e-05}
{"train_loss": 0.07079323381185532, "global_step": 170980, "epoch": 1921, "lr": 3.0258390658583154e-05}
{"train_loss": 0.07135583460330963, "global_step": 170981, "epoch": 1921, "lr": 3.0257858025018594e-05}
{"train_loss": 0.06208980083465576, "global_step": 170982, "epoch": 1921, "lr": 3.025732539410811e-05}
{"train_loss": 0.1564069390296936, "global_step": 170983, "epoch": 1921, "lr": 3.0256792765851805e-05}
{"train_loss": 0.1030334085226059, "global_step": 170984, "epoch": 1921, "lr": 3.0256260140249714e-05}
{"train_loss": 0.06365706026554108, "global_step": 170985, "epoch": 1921, "lr": 3.0255727517301947e-05}
{"train_loss": 0.06556680798530579, "global_step": 170986, "epoch": 1921, "lr": 3.0255194897008536e-05}
{"train_loss": 0.07612378895282745, "global_step": 170987, "epoch": 1921, "lr": 3.0254662279369595e-05}
{"train_loss": 0.05576859042048454, "global_step": 170988, "epoch": 1921, "lr": 3.0254129664385168e-05}
{"train_loss": 0.07755963504314423, "global_step": 170989, "epoch": 1921, "lr": 3.0253597052055328e-05}
{"train_loss": 0.06052431836724281, "global_step": 170990, "epoch": 1921, "lr": 3.0253064442380153e-05}
{"train_loss": 0.12743785977363586, "global_step": 170991, "epoch": 1921, "lr": 3.02525318353597e-05}
{"train_loss": 0.07858322560787201, "global_step": 170992, "epoch": 1921, "lr": 3.0251999230994075e-05}
{"train_loss": 0.08782431483268738, "global_step": 170993, "epoch": 1921, "lr": 3.0251466629283308e-05}
{"train_loss": 0.0879904106259346, "global_step": 170994, "epoch": 1921, "lr": 3.025093403022751e-05}
{"train_loss": 0.06603153795003891, "global_step": 170995, "epoch": 1921, "lr": 3.0250401433826715e-05}
{"train_loss": 0.07313989102840424, "global_step": 170996, "epoch": 1921, "lr": 3.0249868840081032e-05}
{"train_loss": 0.06580528616905212, "global_step": 170997, "epoch": 1921, "lr": 3.0249336248990495e-05}
{"train_loss": 0.10136295109987259, "global_step": 170998, "epoch": 1921, "lr": 3.0248803660555214e-05}
{"train_loss": 0.11326378583908081, "global_step": 170999, "epoch": 1921, "lr": 3.0248271074775215e-05}
{"train_loss": 0.05133243277668953, "global_step": 171000, "epoch": 1921, "lr": 3.0247738491650624e-05}
{"train_loss": 0.08842673897743225, "global_step": 171001, "epoch": 1921, "lr": 3.0247205911181454e-05}
{"train_loss": 0.08302891254425049, "global_step": 171002, "epoch": 1921, "lr": 3.0246673333367837e-05}
{"train_loss": 0.05079112946987152, "global_step": 171003, "epoch": 1921, "lr": 3.0246140758209784e-05}
{"train_loss": 0.08090145140886307, "global_step": 171004, "epoch": 1921, "lr": 3.024560818570742e-05}
{"train_loss": 0.0954648032784462, "global_step": 171005, "epoch": 1921, "lr": 3.0245075615860774e-05}
{"train_loss": 0.06690143793821335, "global_step": 171006, "epoch": 1921, "lr": 3.024454304866996e-05}
{"train_loss": 0.10645099729299545, "global_step": 171007, "epoch": 1921, "lr": 3.024401048413501e-05}
{"train_loss": 0.10946595668792725, "global_step": 171008, "epoch": 1921, "lr": 3.024347792225602e-05}
{"train_loss": 0.07783898711204529, "global_step": 171009, "epoch": 1921, "lr": 3.0242945363033047e-05}
{"train_loss": 0.03704598918557167, "global_step": 171010, "epoch": 1921, "lr": 3.0242412806466157e-05}
{"train_loss": 0.10799311846494675, "global_step": 171011, "epoch": 1921, "lr": 3.0241880252555465e-05}
{"train_loss": 0.0893513560295105, "global_step": 171012, "epoch": 1921, "lr": 3.0241347701300983e-05}
{"train_loss": 0.08042179048061371, "global_step": 171013, "epoch": 1921, "lr": 3.024081515270283e-05}
{"train_loss": 0.08260050415992737, "global_step": 171014, "epoch": 1921, "lr": 3.024028260676104e-05}
{"train_loss": 0.06517763435840607, "global_step": 171015, "epoch": 1921, "lr": 3.0239750063475725e-05}
{"train_loss": 0.08184662461280823, "global_step": 171016, "epoch": 1921, "lr": 3.023921752284692e-05}
{"train_loss": 0.05926613137125969, "global_step": 171017, "epoch": 1921, "lr": 3.023868498487472e-05}
{"train_loss": 0.08817028999328613, "global_step": 171018, "epoch": 1921, "lr": 3.0238152449559176e-05}
{"train_loss": 0.12507160007953644, "global_step": 171019, "epoch": 1921, "lr": 3.0237619916900384e-05}
{"train_loss": 0.11284700036048889, "global_step": 171020, "epoch": 1921, "lr": 3.023708738689839e-05}
{"train_loss": 0.06398316472768784, "global_step": 171021, "epoch": 1921, "lr": 3.0236554859553296e-05}
{"train_loss": 0.13311079144477844, "global_step": 171022, "epoch": 1921, "lr": 3.023602233486513e-05}
{"train_loss": 0.07461677491664886, "global_step": 171023, "epoch": 1921, "lr": 3.0235489812834028e-05}
{"train_loss": 0.06723649054765701, "global_step": 171024, "epoch": 1921, "lr": 3.0234957293459986e-05}
{"train_loss": 0.09309832006692886, "global_step": 171025, "epoch": 1921, "lr": 3.0234424776743143e-05}
{"train_loss": 0.05567237734794617, "global_step": 171026, "epoch": 1921, "lr": 3.0233892262683528e-05}
{"train_loss": 0.15876257419586182, "global_step": 171027, "epoch": 1921, "lr": 3.023335975128121e-05}
{"train_loss": 0.07280389964580536, "global_step": 171028, "epoch": 1921, "lr": 3.02328272425363e-05}
{"train_loss": 0.08104494214057922, "global_step": 171029, "epoch": 1921, "lr": 3.023229473644883e-05}
{"train_loss": 0.13958045840263367, "global_step": 171030, "epoch": 1921, "lr": 3.02317622330189e-05}
{"train_loss": 0.055141910910606384, "global_step": 171031, "epoch": 1921, "lr": 3.023122973224655e-05}
{"train_loss": 0.08273746818304062, "global_step": 171032, "epoch": 1921, "lr": 3.0230697234131894e-05}
{"train_loss": 0.04891275241971016, "global_step": 171033, "epoch": 1921, "lr": 3.023016473867496e-05}
{"train_loss": 0.02576282061636448, "global_step": 171034, "epoch": 1921, "lr": 3.022963224587585e-05}
{"train_loss": 0.11858717352151871, "global_step": 171035, "epoch": 1921, "lr": 3.0229099755734612e-05}
{"train_loss": 0.14194892346858978, "global_step": 171036, "epoch": 1921, "lr": 3.022856726825135e-05}
{"train_loss": 0.06451322883367538, "global_step": 171037, "epoch": 1921, "lr": 3.0228034783426095e-05}
{"train_loss": 0.10269346833229065, "global_step": 171038, "epoch": 1921, "lr": 3.0227502301258964e-05}
{"train_loss": 0.10512702912092209, "global_step": 171039, "epoch": 1921, "lr": 3.0226969821749985e-05}
{"train_loss": 0.07979825884103775, "global_step": 171040, "epoch": 1921, "lr": 3.022643734489926e-05}
{"train_loss": 0.07202864438295364, "global_step": 171041, "epoch": 1921, "lr": 3.0225904870706833e-05}
{"train_loss": 0.04694182425737381, "global_step": 171042, "epoch": 1921, "lr": 3.0225372399172818e-05}
{"train_loss": 0.07595338672399521, "global_step": 171043, "epoch": 1921, "lr": 3.0224839930297237e-05}
{"train_loss": 0.1018386036157608, "global_step": 171044, "epoch": 1921, "lr": 3.0224307464080204e-05}
{"train_loss": 0.06542252749204636, "global_step": 171045, "epoch": 1921, "lr": 3.0223775000521764e-05}
{"train_loss": 0.07041323184967041, "global_step": 171046, "epoch": 1921, "lr": 3.0223242539621987e-05}
{"train_loss": 0.04010288044810295, "global_step": 171047, "epoch": 1921, "lr": 3.0222710081380966e-05}
{"train_loss": 0.15247274935245514, "global_step": 171048, "epoch": 1921, "lr": 3.0222177625798753e-05}
{"train_loss": 0.048597127199172974, "global_step": 171049, "epoch": 1921, "lr": 3.0221645172875435e-05}
{"train_loss": 0.06440038979053497, "global_step": 171050, "epoch": 1921, "lr": 3.0221112722611065e-05}
{"train_loss": 0.063863605260849, "global_step": 171051, "epoch": 1921, "lr": 3.022058027500574e-05}
{"train_loss": 0.07439284771680832, "global_step": 171052, "epoch": 1921, "lr": 3.02200478300595e-05}
{"train_loss": 0.14473530650138855, "global_step": 171053, "epoch": 1921, "lr": 3.0219515387772458e-05}
{"train_loss": 0.10940991342067719, "global_step": 171054, "epoch": 1921, "lr": 3.021898294814463e-05}
{"train_loss": 0.060373254120349884, "global_step": 171055, "epoch": 1921, "lr": 3.0218450511176142e-05}
{"train_loss": 0.056442514061927795, "global_step": 171056, "epoch": 1921, "lr": 3.0217918076867024e-05}
{"train_loss": 0.0853193069734935, "global_step": 171057, "epoch": 1921, "lr": 3.021738564521739e-05, "val_loss": 7.2819132804870605}
{"train_loss": 0.06026042252779007, "global_step": 171058, "epoch": 1922, "lr": 3.0216853216227257e-05}
{"train_loss": 0.07608399540185928, "global_step": 171059, "epoch": 1922, "lr": 3.021632078989676e-05}
{"train_loss": 0.08549084514379501, "global_step": 171060, "epoch": 1922, "lr": 3.0215788366225906e-05}
{"train_loss": 0.11874009668827057, "global_step": 171061, "epoch": 1922, "lr": 3.0215255945214825e-05}
{"train_loss": 0.047741372138261795, "global_step": 171062, "epoch": 1922, "lr": 3.0214723526863543e-05}
{"train_loss": 0.07508593052625656, "global_step": 171063, "epoch": 1922, "lr": 3.021419111117215e-05}
{"train_loss": 0.06223307177424431, "global_step": 171064, "epoch": 1922, "lr": 3.021365869814074e-05}
{"train_loss": 0.047966182231903076, "global_step": 171065, "epoch": 1922, "lr": 3.0213126287769332e-05}
{"train_loss": 0.06324711441993713, "global_step": 171066, "epoch": 1922, "lr": 3.021259388005806e-05}
{"train_loss": 0.08955661952495575, "global_step": 171067, "epoch": 1922, "lr": 3.0212061475006937e-05}
{"train_loss": 0.09988635778427124, "global_step": 171068, "epoch": 1922, "lr": 3.021152907261608e-05}
{"train_loss": 0.04343332722783089, "global_step": 171069, "epoch": 1922, "lr": 3.0210996672885523e-05}
{"train_loss": 0.05126522481441498, "global_step": 171070, "epoch": 1922, "lr": 3.0210464275815374e-05}
{"train_loss": 0.08103884011507034, "global_step": 171071, "epoch": 1922, "lr": 3.0209931881405675e-05}
{"train_loss": 0.07901614159345627, "global_step": 171072, "epoch": 1922, "lr": 3.020939948965652e-05}
{"train_loss": 0.09566164761781693, "global_step": 171073, "epoch": 1922, "lr": 3.020886710056795e-05}
{"train_loss": 0.16166028380393982, "global_step": 171074, "epoch": 1922, "lr": 3.0208334714140084e-05}
{"train_loss": 0.12521137297153473, "global_step": 171075, "epoch": 1922, "lr": 3.0207802330372937e-05}
{"train_loss": 0.07238080352544785, "global_step": 171076, "epoch": 1922, "lr": 3.0207269949266636e-05}
{"train_loss": 0.1101551204919815, "global_step": 171077, "epoch": 1922, "lr": 3.020673757082121e-05}
{"train_loss": 0.06279290467500687, "global_step": 171078, "epoch": 1922, "lr": 3.020620519503676e-05}
{"train_loss": 0.08782680332660675, "global_step": 171079, "epoch": 1922, "lr": 3.0205672821913322e-05}
{"train_loss": 0.08416539430618286, "global_step": 171080, "epoch": 1922, "lr": 3.020514045145102e-05}
{"train_loss": 0.06416596472263336, "global_step": 171081, "epoch": 1922, "lr": 3.0204608083649864e-05}
{"train_loss": 0.08179818838834763, "global_step": 171082, "epoch": 1922, "lr": 3.0204075718509978e-05}
{"train_loss": 0.05895213410258293, "global_step": 171083, "epoch": 1922, "lr": 3.0203543356031404e-05}
{"train_loss": 0.09562699496746063, "global_step": 171084, "epoch": 1922, "lr": 3.0203010996214232e-05}
{"train_loss": 0.05934131145477295, "global_step": 171085, "epoch": 1922, "lr": 3.020247863905853e-05}
{"train_loss": 0.061027057468891144, "global_step": 171086, "epoch": 1922, "lr": 3.0201946284564336e-05}
{"train_loss": 0.07591476291418076, "global_step": 171087, "epoch": 1922, "lr": 3.0201413932731774e-05}
{"train_loss": 0.07668276131153107, "global_step": 171088, "epoch": 1922, "lr": 3.0200881583560874e-05}
{"train_loss": 0.04476534202694893, "global_step": 171089, "epoch": 1922, "lr": 3.020034923705174e-05}
{"train_loss": 0.05680040270090103, "global_step": 171090, "epoch": 1922, "lr": 3.0199816893204414e-05}
{"train_loss": 0.047378797084093094, "global_step": 171091, "epoch": 1922, "lr": 3.0199284552018992e-05}
{"train_loss": 0.06751026958227158, "global_step": 171092, "epoch": 1922, "lr": 3.0198752213495517e-05}
{"train_loss": 0.0925547257065773, "global_step": 171093, "epoch": 1922, "lr": 3.0198219877634103e-05}
{"train_loss": 0.09730321168899536, "global_step": 171094, "epoch": 1922, "lr": 3.0197687544434772e-05}
{"train_loss": 0.06316094845533371, "global_step": 171095, "epoch": 1922, "lr": 3.0197155213897642e-05}
{"train_loss": 0.11364002525806427, "global_step": 171096, "epoch": 1922, "lr": 3.0196622886022753e-05}
{"train_loss": 0.10862581431865692, "global_step": 171097, "epoch": 1922, "lr": 3.0196090560810197e-05}
{"train_loss": 0.09714668989181519, "global_step": 171098, "epoch": 1922, "lr": 3.0195558238260012e-05}
{"train_loss": 0.10576212406158447, "global_step": 171099, "epoch": 1922, "lr": 3.0195025918372322e-05}
{"train_loss": 0.07357244193553925, "global_step": 171100, "epoch": 1922, "lr": 3.0194493601147144e-05}
{"train_loss": 0.07398685067892075, "global_step": 171101, "epoch": 1922, "lr": 3.019396128658458e-05}
{"train_loss": 0.06699118763208389, "global_step": 171102, "epoch": 1922, "lr": 3.01934289746847e-05}
{"train_loss": 0.07927978783845901, "global_step": 171103, "epoch": 1922, "lr": 3.019289666544758e-05}
{"train_loss": 0.09666097164154053, "global_step": 171104, "epoch": 1922, "lr": 3.0192364358873288e-05}
{"train_loss": 0.10877200961112976, "global_step": 171105, "epoch": 1922, "lr": 3.0191832054961866e-05}
{"train_loss": 0.051944758743047714, "global_step": 171106, "epoch": 1922, "lr": 3.019129975371344e-05}
{"train_loss": 0.07672121375799179, "global_step": 171107, "epoch": 1922, "lr": 3.0190767455128027e-05}
{"train_loss": 0.07010921835899353, "global_step": 171108, "epoch": 1922, "lr": 3.0190235159205748e-05}
{"train_loss": 0.061790358275175095, "global_step": 171109, "epoch": 1922, "lr": 3.0189702865946623e-05}
{"train_loss": 0.09441238641738892, "global_step": 171110, "epoch": 1922, "lr": 3.018917057535078e-05}
{"train_loss": 0.11663763225078583, "global_step": 171111, "epoch": 1922, "lr": 3.0188638287418237e-05}
{"train_loss": 0.07605189830064774, "global_step": 171112, "epoch": 1922, "lr": 3.018810600214911e-05}
{"train_loss": 0.044810958206653595, "global_step": 171113, "epoch": 1922, "lr": 3.0187573719543428e-05}
{"train_loss": 0.0831049308180809, "global_step": 171114, "epoch": 1922, "lr": 3.0187041439601304e-05}
{"train_loss": 0.040696315467357635, "global_step": 171115, "epoch": 1922, "lr": 3.0186509162322773e-05}
{"train_loss": 0.09261517226696014, "global_step": 171116, "epoch": 1922, "lr": 3.018597688770795e-05}
{"train_loss": 0.07340718060731888, "global_step": 171117, "epoch": 1922, "lr": 3.0185444615756857e-05}
{"train_loss": 0.0968775674700737, "global_step": 171118, "epoch": 1922, "lr": 3.0184912346469597e-05}
{"train_loss": 0.10484468936920166, "global_step": 171119, "epoch": 1922, "lr": 3.0184380079846242e-05}
{"train_loss": 0.07422280311584473, "global_step": 171120, "epoch": 1922, "lr": 3.0183847815886847e-05}
{"train_loss": 0.1052064299583435, "global_step": 171121, "epoch": 1922, "lr": 3.0183315554591496e-05}
{"train_loss": 0.09352909028530121, "global_step": 171122, "epoch": 1922, "lr": 3.0182783295960253e-05}
{"train_loss": 0.10957629233598709, "global_step": 171123, "epoch": 1922, "lr": 3.018225103999321e-05}
{"train_loss": 0.04475487768650055, "global_step": 171124, "epoch": 1922, "lr": 3.0181718786690393e-05}
{"train_loss": 0.1070667952299118, "global_step": 171125, "epoch": 1922, "lr": 3.018118653605193e-05}
{"train_loss": 0.050897322595119476, "global_step": 171126, "epoch": 1922, "lr": 3.018065428807784e-05}
{"train_loss": 0.07041068375110626, "global_step": 171127, "epoch": 1922, "lr": 3.0180122042768245e-05}
{"train_loss": 0.10726558417081833, "global_step": 171128, "epoch": 1922, "lr": 3.0179589800123166e-05}
{"train_loss": 0.05107865110039711, "global_step": 171129, "epoch": 1922, "lr": 3.017905756014272e-05}
{"train_loss": 0.032601743936538696, "global_step": 171130, "epoch": 1922, "lr": 3.017852532282694e-05}
{"train_loss": 0.04810640960931778, "global_step": 171131, "epoch": 1922, "lr": 3.017799308817594e-05}
{"train_loss": 0.03270338475704193, "global_step": 171132, "epoch": 1922, "lr": 3.0177460856189737e-05}
{"train_loss": 0.11325983703136444, "global_step": 171133, "epoch": 1922, "lr": 3.0176928626868463e-05}
{"train_loss": 0.09293121844530106, "global_step": 171134, "epoch": 1922, "lr": 3.017639640021213e-05}
{"train_loss": 0.10201248526573181, "global_step": 171135, "epoch": 1922, "lr": 3.0175864176220863e-05}
{"train_loss": 0.05350542813539505, "global_step": 171136, "epoch": 1922, "lr": 3.0175331954894692e-05}
{"train_loss": 0.055825959891080856, "global_step": 171137, "epoch": 1922, "lr": 3.017479973623371e-05}
{"train_loss": 0.08036403357982635, "global_step": 171138, "epoch": 1922, "lr": 3.0174267520237987e-05}
{"train_loss": 0.04337165132164955, "global_step": 171139, "epoch": 1922, "lr": 3.0173735306907597e-05}
{"train_loss": 0.10607978701591492, "global_step": 171140, "epoch": 1922, "lr": 3.0173203096242603e-05}
{"train_loss": 0.04899429529905319, "global_step": 171141, "epoch": 1922, "lr": 3.0172670888243083e-05}
{"train_loss": 0.046725042164325714, "global_step": 171142, "epoch": 1922, "lr": 3.0172138682909112e-05}
{"train_loss": 0.09183945506811142, "global_step": 171143, "epoch": 1922, "lr": 3.0171606480240733e-05}
{"train_loss": 0.0651523694396019, "global_step": 171144, "epoch": 1922, "lr": 3.0171074280238064e-05}
{"train_loss": 0.07535966485738754, "global_step": 171145, "epoch": 1922, "lr": 3.017054208290113e-05}
{"train_loss": 0.07849093329872978, "global_step": 171146, "epoch": 1922, "lr": 3.0170009888230043e-05, "val_loss": 7.1667022705078125}
{"train_loss": 0.07704270631074905, "global_step": 171147, "epoch": 1923, "lr": 3.0169477696224847e-05}
{"train_loss": 0.06924937665462494, "global_step": 171148, "epoch": 1923, "lr": 3.0168945506885632e-05}
{"train_loss": 0.04899117350578308, "global_step": 171149, "epoch": 1923, "lr": 3.016841332021244e-05}
{"train_loss": 0.05194853991270065, "global_step": 171150, "epoch": 1923, "lr": 3.016788113620539e-05}
{"train_loss": 0.11423520743846893, "global_step": 171151, "epoch": 1923, "lr": 3.0167348954864503e-05}
{"train_loss": 0.09638489782810211, "global_step": 171152, "epoch": 1923, "lr": 3.0166816776189898e-05}
{"train_loss": 0.11199086159467697, "global_step": 171153, "epoch": 1923, "lr": 3.01662846001816e-05}
{"train_loss": 0.11278856545686722, "global_step": 171154, "epoch": 1923, "lr": 3.01657524268397e-05}
{"train_loss": 0.08837603777647018, "global_step": 171155, "epoch": 1923, "lr": 3.01652202561643e-05}
{"train_loss": 0.08600646257400513, "global_step": 171156, "epoch": 1923, "lr": 3.0164688088155428e-05}
{"train_loss": 0.0702965185046196, "global_step": 171157, "epoch": 1923, "lr": 3.0164155922813175e-05}
{"train_loss": 0.12955883145332336, "global_step": 171158, "epoch": 1923, "lr": 3.016362376013761e-05}
{"train_loss": 0.07382989674806595, "global_step": 171159, "epoch": 1923, "lr": 3.01630916001288e-05}
{"train_loss": 0.11014049500226974, "global_step": 171160, "epoch": 1923, "lr": 3.016255944278683e-05}
{"train_loss": 0.04498860612511635, "global_step": 171161, "epoch": 1923, "lr": 3.016202728811176e-05}
{"train_loss": 0.07710464298725128, "global_step": 171162, "epoch": 1923, "lr": 3.016149513610365e-05}
{"train_loss": 0.06039554998278618, "global_step": 171163, "epoch": 1923, "lr": 3.016096298676261e-05}
{"train_loss": 0.10338343679904938, "global_step": 171164, "epoch": 1923, "lr": 3.016043084008866e-05}
{"train_loss": 0.12446459382772446, "global_step": 171165, "epoch": 1923, "lr": 3.015989869608192e-05}
{"train_loss": 0.04954136162996292, "global_step": 171166, "epoch": 1923, "lr": 3.0159366554742418e-05}
{"train_loss": 0.06941007822751999, "global_step": 171167, "epoch": 1923, "lr": 3.015883441607027e-05}
{"train_loss": 0.11202607303857803, "global_step": 171168, "epoch": 1923, "lr": 3.01583022800655e-05}
{"train_loss": 0.07104789465665817, "global_step": 171169, "epoch": 1923, "lr": 3.0157770146728227e-05}
{"train_loss": 0.1219867542386055, "global_step": 171170, "epoch": 1923, "lr": 3.0157238016058488e-05}
{"train_loss": 0.0597832091152668, "global_step": 171171, "epoch": 1923, "lr": 3.0156705888056375e-05}
{"train_loss": 0.05920122563838959, "global_step": 171172, "epoch": 1923, "lr": 3.0156173762721934e-05}
{"train_loss": 0.07654502987861633, "global_step": 171173, "epoch": 1923, "lr": 3.0155641640055255e-05}
{"train_loss": 0.07128770649433136, "global_step": 171174, "epoch": 1923, "lr": 3.015510952005643e-05}
{"train_loss": 0.09470528364181519, "global_step": 171175, "epoch": 1923, "lr": 3.0154577402725492e-05}
{"train_loss": 0.05946226045489311, "global_step": 171176, "epoch": 1923, "lr": 3.015404528806254e-05}
{"train_loss": 0.03209874406456947, "global_step": 171177, "epoch": 1923, "lr": 3.0153513176067627e-05}
{"train_loss": 0.11380574107170105, "global_step": 171178, "epoch": 1923, "lr": 3.015298106674083e-05}
{"train_loss": 0.05769891291856766, "global_step": 171179, "epoch": 1923, "lr": 3.015244896008223e-05}
{"train_loss": 0.10771626979112625, "global_step": 171180, "epoch": 1923, "lr": 3.01519168560919e-05}
{"train_loss": 0.08235493302345276, "global_step": 171181, "epoch": 1923, "lr": 3.015138475476988e-05}
{"train_loss": 0.06900624930858612, "global_step": 171182, "epoch": 1923, "lr": 3.0150852656116286e-05}
{"train_loss": 0.08330660313367844, "global_step": 171183, "epoch": 1923, "lr": 3.0150320560131144e-05}
{"train_loss": 0.08588839322328568, "global_step": 171184, "epoch": 1923, "lr": 3.0149788466814576e-05}
{"train_loss": 0.08310113847255707, "global_step": 171185, "epoch": 1923, "lr": 3.0149256376166603e-05}
{"train_loss": 0.050318025052547455, "global_step": 171186, "epoch": 1923, "lr": 3.014872428818734e-05}
{"train_loss": 0.15272720158100128, "global_step": 171187, "epoch": 1923, "lr": 3.0148192202876824e-05}
{"train_loss": 0.0800967738032341, "global_step": 171188, "epoch": 1923, "lr": 3.0147660120235155e-05}
{"train_loss": 0.07367850095033646, "global_step": 171189, "epoch": 1923, "lr": 3.0147128040262374e-05}
{"train_loss": 0.05966438353061676, "global_step": 171190, "epoch": 1923, "lr": 3.0146595962958568e-05}
{"train_loss": 0.05627943575382233, "global_step": 171191, "epoch": 1923, "lr": 3.0146063888323833e-05}
{"train_loss": 0.0678182989358902, "global_step": 171192, "epoch": 1923, "lr": 3.0145531816358197e-05}
{"train_loss": 0.06848059594631195, "global_step": 171193, "epoch": 1923, "lr": 3.0144999747061776e-05}
{"train_loss": 0.11497218161821365, "global_step": 171194, "epoch": 1923, "lr": 3.0144467680434596e-05}
{"train_loss": 0.08539248257875443, "global_step": 171195, "epoch": 1923, "lr": 3.014393561647676e-05}
{"train_loss": 0.08095239102840424, "global_step": 171196, "epoch": 1923, "lr": 3.0143403555188333e-05}
{"train_loss": 0.04186488315463066, "global_step": 171197, "epoch": 1923, "lr": 3.0142871496569375e-05}
{"train_loss": 0.09617691487073898, "global_step": 171198, "epoch": 1923, "lr": 3.014233944061997e-05}
{"train_loss": 0.06599361449480057, "global_step": 171199, "epoch": 1923, "lr": 3.0141807387340183e-05}
{"train_loss": 0.09076111018657684, "global_step": 171200, "epoch": 1923, "lr": 3.014127533673009e-05}
{"train_loss": 0.08770128339529037, "global_step": 171201, "epoch": 1923, "lr": 3.014074328878977e-05}
{"train_loss": 0.0906912162899971, "global_step": 171202, "epoch": 1923, "lr": 3.014021124351927e-05}
{"train_loss": 0.0798519179224968, "global_step": 171203, "epoch": 1923, "lr": 3.0139679200918686e-05}
{"train_loss": 0.040990013629198074, "global_step": 171204, "epoch": 1923, "lr": 3.0139147160988067e-05}
{"train_loss": 0.10504277795553207, "global_step": 171205, "epoch": 1923, "lr": 3.0138615123727516e-05}
{"train_loss": 0.041016120463609695, "global_step": 171206, "epoch": 1923, "lr": 3.0138083089137064e-05}
{"train_loss": 0.09905160218477249, "global_step": 171207, "epoch": 1923, "lr": 3.013755105721683e-05}
{"train_loss": 0.08711418509483337, "global_step": 171208, "epoch": 1923, "lr": 3.0137019027966834e-05}
{"train_loss": 0.10316064953804016, "global_step": 171209, "epoch": 1923, "lr": 3.0136487001387182e-05}
{"train_loss": 0.09464345872402191, "global_step": 171210, "epoch": 1923, "lr": 3.0135954977477947e-05}
{"train_loss": 0.02657226100564003, "global_step": 171211, "epoch": 1923, "lr": 3.013542295623918e-05}
{"train_loss": 0.06304555386304855, "global_step": 171212, "epoch": 1923, "lr": 3.013489093767098e-05}
{"train_loss": 0.06790551543235779, "global_step": 171213, "epoch": 1923, "lr": 3.0134358921773375e-05}
{"train_loss": 0.07838788628578186, "global_step": 171214, "epoch": 1923, "lr": 3.0133826908546493e-05}
{"train_loss": 0.08035969734191895, "global_step": 171215, "epoch": 1923, "lr": 3.013329489799036e-05}
{"train_loss": 0.06625295430421829, "global_step": 171216, "epoch": 1923, "lr": 3.0132762890105064e-05}
{"train_loss": 0.1089615523815155, "global_step": 171217, "epoch": 1923, "lr": 3.0132230884890676e-05}
{"train_loss": 0.05852903425693512, "global_step": 171218, "epoch": 1923, "lr": 3.013169888234727e-05}
{"train_loss": 0.07434949278831482, "global_step": 171219, "epoch": 1923, "lr": 3.0131166882474914e-05}
{"train_loss": 0.07108084857463837, "global_step": 171220, "epoch": 1923, "lr": 3.013063488527369e-05}
{"train_loss": 0.07227817177772522, "global_step": 171221, "epoch": 1923, "lr": 3.013010289074364e-05}
{"train_loss": 0.06945433467626572, "global_step": 171222, "epoch": 1923, "lr": 3.0129570898884874e-05}
{"train_loss": 0.07390396296977997, "global_step": 171223, "epoch": 1923, "lr": 3.0129038909697427e-05}
{"train_loss": 0.106714628636837, "global_step": 171224, "epoch": 1923, "lr": 3.012850692318141e-05}
{"train_loss": 0.13027675449848175, "global_step": 171225, "epoch": 1923, "lr": 3.012797493933685e-05}
{"train_loss": 0.07306505739688873, "global_step": 171226, "epoch": 1923, "lr": 3.0127442958163864e-05}
{"train_loss": 0.08416780829429626, "global_step": 171227, "epoch": 1923, "lr": 3.0126910979662483e-05}
{"train_loss": 0.05001010745763779, "global_step": 171228, "epoch": 1923, "lr": 3.012637900383279e-05}
{"train_loss": 0.042899586260318756, "global_step": 171229, "epoch": 1923, "lr": 3.0125847030674886e-05}
{"train_loss": 0.0735776275396347, "global_step": 171230, "epoch": 1923, "lr": 3.0125315060188798e-05}
{"train_loss": 0.1033320426940918, "global_step": 171231, "epoch": 1923, "lr": 3.0124783092374642e-05}
{"train_loss": 0.06402724236249924, "global_step": 171232, "epoch": 1923, "lr": 3.012425112723245e-05}
{"train_loss": 0.03439275920391083, "global_step": 171233, "epoch": 1923, "lr": 3.012371916476232e-05}
{"train_loss": 0.04018130153417587, "global_step": 171234, "epoch": 1923, "lr": 3.012318720496431e-05}
{"train_loss": 0.07865041243226341, "global_step": 171235, "epoch": 1923, "lr": 3.0122655247838495e-05, "val_loss": 7.118033409118652}
{"train_loss": 0.0518936961889267, "global_step": 171236, "epoch": 1924, "lr": 3.0122123293384952e-05}
{"train_loss": 0.08855149149894714, "global_step": 171237, "epoch": 1924, "lr": 3.0121591341603738e-05}
{"train_loss": 0.09458091855049133, "global_step": 171238, "epoch": 1924, "lr": 3.012105939249494e-05}
{"train_loss": 0.07560313493013382, "global_step": 171239, "epoch": 1924, "lr": 3.0120527446058634e-05}
{"train_loss": 0.09869459271430969, "global_step": 171240, "epoch": 1924, "lr": 3.011999550229485e-05}
{"train_loss": 0.08590611815452576, "global_step": 171241, "epoch": 1924, "lr": 3.011946356120372e-05}
{"train_loss": 0.11954697966575623, "global_step": 171242, "epoch": 1924, "lr": 3.0118931622785263e-05}
{"train_loss": 0.060743678361177444, "global_step": 171243, "epoch": 1924, "lr": 3.01183996870396e-05}
{"train_loss": 0.1076248437166214, "global_step": 171244, "epoch": 1924, "lr": 3.0117867753966745e-05}
{"train_loss": 0.1578749269247055, "global_step": 171245, "epoch": 1924, "lr": 3.0117335823566807e-05}
{"train_loss": 0.08886676281690598, "global_step": 171246, "epoch": 1924, "lr": 3.0116803895839874e-05}
{"train_loss": 0.11594174802303314, "global_step": 171247, "epoch": 1924, "lr": 3.0116271970785964e-05}
{"train_loss": 0.06521552801132202, "global_step": 171248, "epoch": 1924, "lr": 3.0115740048405205e-05}
{"train_loss": 0.053499288856983185, "global_step": 171249, "epoch": 1924, "lr": 3.0115208128697625e-05}
{"train_loss": 0.0962180569767952, "global_step": 171250, "epoch": 1924, "lr": 3.0114676211663328e-05}
{"train_loss": 0.07646206766366959, "global_step": 171251, "epoch": 1924, "lr": 3.0114144297302348e-05}
{"train_loss": 0.06225278973579407, "global_step": 171252, "epoch": 1924, "lr": 3.0113612385614807e-05}
{"train_loss": 0.06427173316478729, "global_step": 171253, "epoch": 1924, "lr": 3.0113080476600737e-05}
{"train_loss": 0.06884153932332993, "global_step": 171254, "epoch": 1924, "lr": 3.0112548570260214e-05}
{"train_loss": 0.06526248157024384, "global_step": 171255, "epoch": 1924, "lr": 3.0112016666593323e-05}
{"train_loss": 0.10342874377965927, "global_step": 171256, "epoch": 1924, "lr": 3.011148476560013e-05}
{"train_loss": 0.1261296570301056, "global_step": 171257, "epoch": 1924, "lr": 3.01109528672807e-05}
{"train_loss": 0.0587804950773716, "global_step": 171258, "epoch": 1924, "lr": 3.0110420971635123e-05}
{"train_loss": 0.07248111069202423, "global_step": 171259, "epoch": 1924, "lr": 3.0109889078663433e-05}
{"train_loss": 0.07928044348955154, "global_step": 171260, "epoch": 1924, "lr": 3.010935718836575e-05}
{"train_loss": 0.05462721362709999, "global_step": 171261, "epoch": 1924, "lr": 3.0108825300742105e-05}
{"train_loss": 0.12641796469688416, "global_step": 171262, "epoch": 1924, "lr": 3.01082934157926e-05}
{"train_loss": 0.14911742508411407, "global_step": 171263, "epoch": 1924, "lr": 3.0107761533517275e-05}
{"train_loss": 0.04819875210523605, "global_step": 171264, "epoch": 1924, "lr": 3.010722965391622e-05}
{"train_loss": 0.12070970237255096, "global_step": 171265, "epoch": 1924, "lr": 3.0106697776989522e-05}
{"train_loss": 0.11467891931533813, "global_step": 171266, "epoch": 1924, "lr": 3.0106165902737223e-05}
{"train_loss": 0.05316615104675293, "global_step": 171267, "epoch": 1924, "lr": 3.0105634031159423e-05}
{"train_loss": 0.08730988949537277, "global_step": 171268, "epoch": 1924, "lr": 3.0105102162256154e-05}
{"train_loss": 0.03819340467453003, "global_step": 171269, "epoch": 1924, "lr": 3.0104570296027533e-05}
{"train_loss": 0.099720299243927, "global_step": 171270, "epoch": 1924, "lr": 3.0104038432473595e-05}
{"train_loss": 0.08044922351837158, "global_step": 171271, "epoch": 1924, "lr": 3.0103506571594442e-05}
{"train_loss": 0.03465164825320244, "global_step": 171272, "epoch": 1924, "lr": 3.0102974713390124e-05}
{"train_loss": 0.05616970360279083, "global_step": 171273, "epoch": 1924, "lr": 3.010244285786072e-05}
{"train_loss": 0.1522890031337738, "global_step": 171274, "epoch": 1924, "lr": 3.0101911005006294e-05}
{"train_loss": 0.0703817754983902, "global_step": 171275, "epoch": 1924, "lr": 3.0101379154826925e-05}
{"train_loss": 0.11531747877597809, "global_step": 171276, "epoch": 1924, "lr": 3.010084730732269e-05}
{"train_loss": 0.08990615606307983, "global_step": 171277, "epoch": 1924, "lr": 3.0100315462493655e-05}
{"train_loss": 0.049736399203538895, "global_step": 171278, "epoch": 1924, "lr": 3.0099783620339872e-05}
{"train_loss": 0.04160567373037338, "global_step": 171279, "epoch": 1924, "lr": 3.0099251780861447e-05}
{"train_loss": 0.08107569068670273, "global_step": 171280, "epoch": 1924, "lr": 3.0098719944058422e-05}
{"train_loss": 0.05378585308790207, "global_step": 171281, "epoch": 1924, "lr": 3.009818810993088e-05}
{"train_loss": 0.1531357318162918, "global_step": 171282, "epoch": 1924, "lr": 3.0097656278478915e-05}
{"train_loss": 0.05504170060157776, "global_step": 171283, "epoch": 1924, "lr": 3.009712444970255e-05}
{"train_loss": 0.11632522940635681, "global_step": 171284, "epoch": 1924, "lr": 3.0096592623601916e-05}
{"train_loss": 0.14903524518013, "global_step": 171285, "epoch": 1924, "lr": 3.009606080017702e-05}
{"train_loss": 0.11097698658704758, "global_step": 171286, "epoch": 1924, "lr": 3.0095528979428002e-05}
{"train_loss": 0.10397008061408997, "global_step": 171287, "epoch": 1924, "lr": 3.0094997161354864e-05}
{"train_loss": 0.06407646089792252, "global_step": 171288, "epoch": 1924, "lr": 3.0094465345957735e-05}
{"train_loss": 0.08237424492835999, "global_step": 171289, "epoch": 1924, "lr": 3.009393353323664e-05}
{"train_loss": 0.12335291504859924, "global_step": 171290, "epoch": 1924, "lr": 3.0093401723191694e-05}
{"train_loss": 0.11195738613605499, "global_step": 171291, "epoch": 1924, "lr": 3.0092869915822942e-05}
{"train_loss": 0.11478187888860703, "global_step": 171292, "epoch": 1924, "lr": 3.009233811113046e-05}
{"train_loss": 0.1560807228088379, "global_step": 171293, "epoch": 1924, "lr": 3.0091806309114312e-05}
{"train_loss": 0.08520978689193726, "global_step": 171294, "epoch": 1924, "lr": 3.0091274509774585e-05}
{"train_loss": 0.09290564060211182, "global_step": 171295, "epoch": 1924, "lr": 3.009074271311134e-05}
{"train_loss": 0.07281291484832764, "global_step": 171296, "epoch": 1924, "lr": 3.0090210919124666e-05}
{"train_loss": 0.1095685362815857, "global_step": 171297, "epoch": 1924, "lr": 3.0089679127814602e-05}
{"train_loss": 0.0371248759329319, "global_step": 171298, "epoch": 1924, "lr": 3.0089147339181257e-05}
{"train_loss": 0.07823839783668518, "global_step": 171299, "epoch": 1924, "lr": 3.0088615553224658e-05}
{"train_loss": 0.050697892904281616, "global_step": 171300, "epoch": 1924, "lr": 3.0088083769944907e-05}
{"train_loss": 0.07464426755905151, "global_step": 171301, "epoch": 1924, "lr": 3.0087551989342088e-05}
{"train_loss": 0.09698852151632309, "global_step": 171302, "epoch": 1924, "lr": 3.008702021141624e-05}
{"train_loss": 0.08160847425460815, "global_step": 171303, "epoch": 1924, "lr": 3.008648843616746e-05}
{"train_loss": 0.08649761229753494, "global_step": 171304, "epoch": 1924, "lr": 3.0085956663595795e-05}
{"train_loss": 0.10665407031774521, "global_step": 171305, "epoch": 1924, "lr": 3.0085424893701353e-05}
{"train_loss": 0.07811026275157928, "global_step": 171306, "epoch": 1924, "lr": 3.008489312648416e-05}
{"train_loss": 0.07225284725427628, "global_step": 171307, "epoch": 1924, "lr": 3.0084361361944323e-05}
{"train_loss": 0.11282671988010406, "global_step": 171308, "epoch": 1924, "lr": 3.008382960008189e-05}
{"train_loss": 0.09292881190776825, "global_step": 171309, "epoch": 1924, "lr": 3.008329784089696e-05}
{"train_loss": 0.06598801165819168, "global_step": 171310, "epoch": 1924, "lr": 3.008276608438957e-05}
{"train_loss": 0.07666272670030594, "global_step": 171311, "epoch": 1924, "lr": 3.0082234330559833e-05}
{"train_loss": 0.15288536250591278, "global_step": 171312, "epoch": 1924, "lr": 3.0081702579407777e-05}
{"train_loss": 0.14498098194599152, "global_step": 171313, "epoch": 1924, "lr": 3.0081170830933504e-05}
{"train_loss": 0.08675450831651688, "global_step": 171314, "epoch": 1924, "lr": 3.0080639085137065e-05}
{"train_loss": 0.1121448203921318, "global_step": 171315, "epoch": 1924, "lr": 3.0080107342018547e-05}
{"train_loss": 0.11848912388086319, "global_step": 171316, "epoch": 1924, "lr": 3.0079575601578013e-05}
{"train_loss": 0.06763243675231934, "global_step": 171317, "epoch": 1924, "lr": 3.007904386381553e-05}
{"train_loss": 0.0689888447523117, "global_step": 171318, "epoch": 1924, "lr": 3.0078512128731196e-05}
{"train_loss": 0.1476045399904251, "global_step": 171319, "epoch": 1924, "lr": 3.0077980396325047e-05}
{"train_loss": 0.05720590054988861, "global_step": 171320, "epoch": 1924, "lr": 3.0077448666597185e-05}
{"train_loss": 0.10882770270109177, "global_step": 171321, "epoch": 1924, "lr": 3.007691693954765e-05}
{"train_loss": 0.08470432460308075, "global_step": 171322, "epoch": 1924, "lr": 3.0076385215176545e-05}
{"train_loss": 0.06407417356967926, "global_step": 171323, "epoch": 1924, "lr": 3.007585349348391e-05}
{"train_loss": 0.08987942195675347, "global_step": 171324, "epoch": 1924, "lr": 3.0075321774469855e-05, "val_loss": 7.076879978179932}
{"train_loss": 0.13001473248004913, "global_step": 171325, "epoch": 1925, "lr": 3.0074790058134417e-05}
{"train_loss": 0.10604430735111237, "global_step": 171326, "epoch": 1925, "lr": 3.007425834447769e-05}
{"train_loss": 0.11570436507463455, "global_step": 171327, "epoch": 1925, "lr": 3.007372663349972e-05}
{"train_loss": 0.10471606254577637, "global_step": 171328, "epoch": 1925, "lr": 3.0073194925200616e-05}
{"train_loss": 0.08733925223350525, "global_step": 171329, "epoch": 1925, "lr": 3.0072663219580406e-05}
{"train_loss": 0.1359381079673767, "global_step": 171330, "epoch": 1925, "lr": 3.0072131516639202e-05}
{"train_loss": 0.07360578328371048, "global_step": 171331, "epoch": 1925, "lr": 3.007159981637705e-05}
{"train_loss": 0.08903966099023819, "global_step": 171332, "epoch": 1925, "lr": 3.0071068118794026e-05}
{"train_loss": 0.08020613342523575, "global_step": 171333, "epoch": 1925, "lr": 3.0070536423890206e-05}
{"train_loss": 0.08093935996294022, "global_step": 171334, "epoch": 1925, "lr": 3.0070004731665657e-05}
{"train_loss": 0.13473041355609894, "global_step": 171335, "epoch": 1925, "lr": 3.0069473042120455e-05}
{"train_loss": 0.08200355619192123, "global_step": 171336, "epoch": 1925, "lr": 3.0068941355254666e-05}
{"train_loss": 0.0778408795595169, "global_step": 171337, "epoch": 1925, "lr": 3.0068409671068378e-05}
{"train_loss": 0.05796441063284874, "global_step": 171338, "epoch": 1925, "lr": 3.0067877989561627e-05}
{"train_loss": 0.07495434582233429, "global_step": 171339, "epoch": 1925, "lr": 3.0067346310734533e-05}
{"train_loss": 0.15986353158950806, "global_step": 171340, "epoch": 1925, "lr": 3.0066814634587116e-05}
{"train_loss": 0.05548439174890518, "global_step": 171341, "epoch": 1925, "lr": 3.0066282961119498e-05}
{"train_loss": 0.046028219163417816, "global_step": 171342, "epoch": 1925, "lr": 3.0065751290331702e-05}
{"train_loss": 0.12203619629144669, "global_step": 171343, "epoch": 1925, "lr": 3.006521962222384e-05}
{"train_loss": 0.07389009743928909, "global_step": 171344, "epoch": 1925, "lr": 3.006468795679595e-05}
{"train_loss": 0.05880925804376602, "global_step": 171345, "epoch": 1925, "lr": 3.006415629404814e-05}
{"train_loss": 0.07513909041881561, "global_step": 171346, "epoch": 1925, "lr": 3.0063624633980438e-05}
{"train_loss": 0.10798023641109467, "global_step": 171347, "epoch": 1925, "lr": 3.006309297659296e-05}
{"train_loss": 0.1446790099143982, "global_step": 171348, "epoch": 1925, "lr": 3.006256132188574e-05}
{"train_loss": 0.11316636949777603, "global_step": 171349, "epoch": 1925, "lr": 3.0062029669858883e-05}
{"train_loss": 0.14412762224674225, "global_step": 171350, "epoch": 1925, "lr": 3.006149802051243e-05}
{"train_loss": 0.09593519568443298, "global_step": 171351, "epoch": 1925, "lr": 3.0060966373846468e-05}
{"train_loss": 0.028424136340618134, "global_step": 171352, "epoch": 1925, "lr": 3.0060434729861064e-05}
{"train_loss": 0.08809774369001389, "global_step": 171353, "epoch": 1925, "lr": 3.0059903088556297e-05}
{"train_loss": 0.050422269850969315, "global_step": 171354, "epoch": 1925, "lr": 3.005937144993223e-05}
{"train_loss": 0.09528759121894836, "global_step": 171355, "epoch": 1925, "lr": 3.0058839813988926e-05}
{"train_loss": 0.12028442323207855, "global_step": 171356, "epoch": 1925, "lr": 3.0058308180726492e-05}
{"train_loss": 0.07946432381868362, "global_step": 171357, "epoch": 1925, "lr": 3.005777655014495e-05}
{"train_loss": 0.08819151669740677, "global_step": 171358, "epoch": 1925, "lr": 3.0057244922244422e-05}
{"train_loss": 0.1045008972287178, "global_step": 171359, "epoch": 1925, "lr": 3.005671329702493e-05}
{"train_loss": 0.06409262865781784, "global_step": 171360, "epoch": 1925, "lr": 3.0056181674486593e-05}
{"train_loss": 0.08832836151123047, "global_step": 171361, "epoch": 1925, "lr": 3.0055650054629435e-05}
{"train_loss": 0.13838474452495575, "global_step": 171362, "epoch": 1925, "lr": 3.0055118437453576e-05}
{"train_loss": 0.13033290207386017, "global_step": 171363, "epoch": 1925, "lr": 3.005458682295904e-05}
{"train_loss": 0.10462649166584015, "global_step": 171364, "epoch": 1925, "lr": 3.0054055211145948e-05}
{"train_loss": 0.101538747549057, "global_step": 171365, "epoch": 1925, "lr": 3.005352360201432e-05}
{"train_loss": 0.10431940853595734, "global_step": 171366, "epoch": 1925, "lr": 3.005299199556427e-05}
{"train_loss": 0.1097341924905777, "global_step": 171367, "epoch": 1925, "lr": 3.005246039179584e-05}
{"train_loss": 0.11969676613807678, "global_step": 171368, "epoch": 1925, "lr": 3.0051928790709128e-05}
{"train_loss": 0.07469765841960907, "global_step": 171369, "epoch": 1925, "lr": 3.0051397192304187e-05}
{"train_loss": 0.03534727916121483, "global_step": 171370, "epoch": 1925, "lr": 3.0050865596581085e-05}
{"train_loss": 0.05960739031434059, "global_step": 171371, "epoch": 1925, "lr": 3.0050334003539905e-05}
{"train_loss": 0.042683329433202744, "global_step": 171372, "epoch": 1925, "lr": 3.00498024131807e-05}
{"train_loss": 0.08863293379545212, "global_step": 171373, "epoch": 1925, "lr": 3.0049270825503584e-05}
{"train_loss": 0.12227605283260345, "global_step": 171374, "epoch": 1925, "lr": 3.004873924050857e-05}
{"train_loss": 0.06532442569732666, "global_step": 171375, "epoch": 1925, "lr": 3.004820765819579e-05}
{"train_loss": 0.05932179465889931, "global_step": 171376, "epoch": 1925, "lr": 3.0047676078565258e-05}
{"train_loss": 0.04000161960721016, "global_step": 171377, "epoch": 1925, "lr": 3.00471445016171e-05}
{"train_loss": 0.0943618193268776, "global_step": 171378, "epoch": 1925, "lr": 3.004661292735133e-05}
{"train_loss": 0.13641512393951416, "global_step": 171379, "epoch": 1925, "lr": 3.0046081355768073e-05}
{"train_loss": 0.08927324414253235, "global_step": 171380, "epoch": 1925, "lr": 3.004554978686736e-05}
{"train_loss": 0.08059889078140259, "global_step": 171381, "epoch": 1925, "lr": 3.00450182206493e-05}
{"train_loss": 0.08524613827466965, "global_step": 171382, "epoch": 1925, "lr": 3.0044486657113925e-05}
{"train_loss": 0.10533471405506134, "global_step": 171383, "epoch": 1925, "lr": 3.0043955096261343e-05}
{"train_loss": 0.05911216884851456, "global_step": 171384, "epoch": 1925, "lr": 3.0043423538091585e-05}
{"train_loss": 0.060277074575424194, "global_step": 171385, "epoch": 1925, "lr": 3.004289198260477e-05}
{"train_loss": 0.09674124419689178, "global_step": 171386, "epoch": 1925, "lr": 3.0042360429800926e-05}
{"train_loss": 0.08765971660614014, "global_step": 171387, "epoch": 1925, "lr": 3.0041828879680163e-05}
{"train_loss": 0.12458628416061401, "global_step": 171388, "epoch": 1925, "lr": 3.0041297332242522e-05}
{"train_loss": 0.06265527009963989, "global_step": 171389, "epoch": 1925, "lr": 3.0040765787488086e-05}
{"train_loss": 0.12406180799007416, "global_step": 171390, "epoch": 1925, "lr": 3.004023424541692e-05}
{"train_loss": 0.13043393194675446, "global_step": 171391, "epoch": 1925, "lr": 3.0039702706029092e-05}
{"train_loss": 0.07582899183034897, "global_step": 171392, "epoch": 1925, "lr": 3.0039171169324708e-05}
{"train_loss": 0.042003583163022995, "global_step": 171393, "epoch": 1925, "lr": 3.0038639635303795e-05}
{"train_loss": 0.10525614768266678, "global_step": 171394, "epoch": 1925, "lr": 3.0038108103966462e-05}
{"train_loss": 0.061690803617239, "global_step": 171395, "epoch": 1925, "lr": 3.0037576575312743e-05}
{"train_loss": 0.06117789074778557, "global_step": 171396, "epoch": 1925, "lr": 3.003704504934275e-05}
{"train_loss": 0.0990195944905281, "global_step": 171397, "epoch": 1925, "lr": 3.0036513526056502e-05}
{"train_loss": 0.12244996428489685, "global_step": 171398, "epoch": 1925, "lr": 3.0035982005454132e-05}
{"train_loss": 0.0867300033569336, "global_step": 171399, "epoch": 1925, "lr": 3.0035450487535655e-05}
{"train_loss": 0.07447659969329834, "global_step": 171400, "epoch": 1925, "lr": 3.0034918972301197e-05}
{"train_loss": 0.0817832201719284, "global_step": 171401, "epoch": 1925, "lr": 3.0034387459750774e-05}
{"train_loss": 0.10621072351932526, "global_step": 171402, "epoch": 1925, "lr": 3.0033855949884503e-05}
{"train_loss": 0.11338046193122864, "global_step": 171403, "epoch": 1925, "lr": 3.0033324442702422e-05}
{"train_loss": 0.10310326516628265, "global_step": 171404, "epoch": 1925, "lr": 3.0032792938204636e-05}
{"train_loss": 0.07915002107620239, "global_step": 171405, "epoch": 1925, "lr": 3.0032261436391175e-05}
{"train_loss": 0.06329664587974548, "global_step": 171406, "epoch": 1925, "lr": 3.0031729937262155e-05}
{"train_loss": 0.06169682368636131, "global_step": 171407, "epoch": 1925, "lr": 3.003119844081761e-05}
{"train_loss": 0.09498528391122818, "global_step": 171408, "epoch": 1925, "lr": 3.0030666947057622e-05}
{"train_loss": 0.07891426235437393, "global_step": 171409, "epoch": 1925, "lr": 3.003013545598229e-05}
{"train_loss": 0.06613626331090927, "global_step": 171410, "epoch": 1925, "lr": 3.0029603967591634e-05}
{"train_loss": 0.09503389894962311, "global_step": 171411, "epoch": 1925, "lr": 3.0029072481885785e-05}
{"train_loss": 0.0968426764011383, "global_step": 171412, "epoch": 1925, "lr": 3.0028540998864757e-05}
{"train_loss": 0.09030105135916325, "global_step": 171413, "epoch": 1925, "lr": 3.002800951852867e-05, "val_loss": 7.022668361663818, "train_action_mse_error": 12.424936294555664}
{"train_loss": 0.12309619039297104, "global_step": 171414, "epoch": 1926, "lr": 3.0027478040877556e-05}
{"train_loss": 0.06299515068531036, "global_step": 171415, "epoch": 1926, "lr": 3.002694656591152e-05}
{"train_loss": 0.10089845955371857, "global_step": 171416, "epoch": 1926, "lr": 3.00264150936306e-05}
{"train_loss": 0.05476793646812439, "global_step": 171417, "epoch": 1926, "lr": 3.0025883624034902e-05}
{"train_loss": 0.06594676524400711, "global_step": 171418, "epoch": 1926, "lr": 3.0025352157124464e-05}
{"train_loss": 0.1318427175283432, "global_step": 171419, "epoch": 1926, "lr": 3.002482069289939e-05}
{"train_loss": 0.12293940037488937, "global_step": 171420, "epoch": 1926, "lr": 3.0024289231359724e-05}
{"train_loss": 0.06126084551215172, "global_step": 171421, "epoch": 1926, "lr": 3.0023757772505562e-05}
{"train_loss": 0.05149379000067711, "global_step": 171422, "epoch": 1926, "lr": 3.0023226316336945e-05}
{"train_loss": 0.08209098130464554, "global_step": 171423, "epoch": 1926, "lr": 3.0022694862853984e-05}
{"train_loss": 0.0512542724609375, "global_step": 171424, "epoch": 1926, "lr": 3.00221634120567e-05}
{"train_loss": 0.0918981283903122, "global_step": 171425, "epoch": 1926, "lr": 3.0021631963945217e-05}
{"train_loss": 0.09177903831005096, "global_step": 171426, "epoch": 1926, "lr": 3.0021100518519563e-05}
{"train_loss": 0.08623756468296051, "global_step": 171427, "epoch": 1926, "lr": 3.0020569075779837e-05}
{"train_loss": 0.09745608270168304, "global_step": 171428, "epoch": 1926, "lr": 3.0020037635726116e-05}
{"train_loss": 0.09254970401525497, "global_step": 171429, "epoch": 1926, "lr": 3.0019506198358432e-05}
{"train_loss": 0.09370492398738861, "global_step": 171430, "epoch": 1926, "lr": 3.0018974763676903e-05}
{"train_loss": 0.08644787967205048, "global_step": 171431, "epoch": 1926, "lr": 3.0018443331681563e-05}
{"train_loss": 0.06958399713039398, "global_step": 171432, "epoch": 1926, "lr": 3.001791190237252e-05}
{"train_loss": 0.09410922229290009, "global_step": 171433, "epoch": 1926, "lr": 3.00173804757498e-05}
{"train_loss": 0.05490244925022125, "global_step": 171434, "epoch": 1926, "lr": 3.001684905181352e-05}
{"train_loss": 0.04187664017081261, "global_step": 171435, "epoch": 1926, "lr": 3.0016317630563712e-05}
{"train_loss": 0.07785006612539291, "global_step": 171436, "epoch": 1926, "lr": 3.0015786212000485e-05}
{"train_loss": 0.11086387932300568, "global_step": 171437, "epoch": 1926, "lr": 3.0015254796123883e-05}
{"train_loss": 0.11140511929988861, "global_step": 171438, "epoch": 1926, "lr": 3.0014723382933996e-05}
{"train_loss": 0.07409045100212097, "global_step": 171439, "epoch": 1926, "lr": 3.001419197243086e-05}
{"train_loss": 0.09438513964414597, "global_step": 171440, "epoch": 1926, "lr": 3.0013660564614603e-05}
{"train_loss": 0.09042783081531525, "global_step": 171441, "epoch": 1926, "lr": 3.0013129159485243e-05}
{"train_loss": 0.0768701434135437, "global_step": 171442, "epoch": 1926, "lr": 3.0012597757042893e-05}
{"train_loss": 0.07376037538051605, "global_step": 171443, "epoch": 1926, "lr": 3.0012066357287588e-05}
{"train_loss": 0.07473539561033249, "global_step": 171444, "epoch": 1926, "lr": 3.0011534960219428e-05}
{"train_loss": 0.06667012721300125, "global_step": 171445, "epoch": 1926, "lr": 3.001100356583847e-05}
{"train_loss": 0.08509083837270737, "global_step": 171446, "epoch": 1926, "lr": 3.0010472174144787e-05}
{"train_loss": 0.03329836204648018, "global_step": 171447, "epoch": 1926, "lr": 3.000994078513847e-05}
{"train_loss": 0.11259923875331879, "global_step": 171448, "epoch": 1926, "lr": 3.0009409398819543e-05}
{"train_loss": 0.061921581625938416, "global_step": 171449, "epoch": 1926, "lr": 3.0008878015188134e-05}
{"train_loss": 0.12103895843029022, "global_step": 171450, "epoch": 1926, "lr": 3.0008346634244266e-05}
{"train_loss": 0.06941888481378555, "global_step": 171451, "epoch": 1926, "lr": 3.000781525598806e-05}
{"train_loss": 0.049442265182733536, "global_step": 171452, "epoch": 1926, "lr": 3.000728388041953e-05}
{"train_loss": 0.09113951772451401, "global_step": 171453, "epoch": 1926, "lr": 3.0006752507538792e-05}
{"train_loss": 0.14493270218372345, "global_step": 171454, "epoch": 1926, "lr": 3.0006221137345893e-05}
{"train_loss": 0.10085315257310867, "global_step": 171455, "epoch": 1926, "lr": 3.0005689769840927e-05}
{"train_loss": 0.10640891641378403, "global_step": 171456, "epoch": 1926, "lr": 3.0005158405023936e-05}
{"train_loss": 0.08245598524808884, "global_step": 171457, "epoch": 1926, "lr": 3.0004627042895027e-05}
{"train_loss": 0.07627374678850174, "global_step": 171458, "epoch": 1926, "lr": 3.0004095683454235e-05}
{"train_loss": 0.09651260077953339, "global_step": 171459, "epoch": 1926, "lr": 3.000356432670167e-05}
{"train_loss": 0.10942401736974716, "global_step": 171460, "epoch": 1926, "lr": 3.0003032972637357e-05}
{"train_loss": 0.07177245616912842, "global_step": 171461, "epoch": 1926, "lr": 3.0002501621261413e-05}
{"train_loss": 0.03680802509188652, "global_step": 171462, "epoch": 1926, "lr": 3.000197027257387e-05}
{"train_loss": 0.07285391539335251, "global_step": 171463, "epoch": 1926, "lr": 3.000143892657483e-05}
{"train_loss": 0.07668725401163101, "global_step": 171464, "epoch": 1926, "lr": 3.0000907583264348e-05}
{"train_loss": 0.04611631855368614, "global_step": 171465, "epoch": 1926, "lr": 3.0000376242642504e-05}
{"train_loss": 0.12703454494476318, "global_step": 171466, "epoch": 1926, "lr": 2.9999844904709373e-05}
{"train_loss": 0.07516907155513763, "global_step": 171467, "epoch": 1926, "lr": 2.9999313569465005e-05}
{"train_loss": 0.09956800192594528, "global_step": 171468, "epoch": 1926, "lr": 2.9998782236909497e-05}
{"train_loss": 0.06153322383761406, "global_step": 171469, "epoch": 1926, "lr": 2.9998250907042897e-05}
{"train_loss": 0.0790724977850914, "global_step": 171470, "epoch": 1926, "lr": 2.9997719579865303e-05}
{"train_loss": 0.09660325199365616, "global_step": 171471, "epoch": 1926, "lr": 2.9997188255376753e-05}
{"train_loss": 0.07851351052522659, "global_step": 171472, "epoch": 1926, "lr": 2.999665693357736e-05}
{"train_loss": 0.10459816455841064, "global_step": 171473, "epoch": 1926, "lr": 2.9996125614467154e-05}
{"train_loss": 0.08313220739364624, "global_step": 171474, "epoch": 1926, "lr": 2.9995594298046237e-05}
{"train_loss": 0.1277400106191635, "global_step": 171475, "epoch": 1926, "lr": 2.999506298431465e-05}
{"train_loss": 0.11330639570951462, "global_step": 171476, "epoch": 1926, "lr": 2.9994531673272507e-05}
{"train_loss": 0.14314602315425873, "global_step": 171477, "epoch": 1926, "lr": 2.999400036491984e-05}
{"train_loss": 0.05960479751229286, "global_step": 171478, "epoch": 1926, "lr": 2.9993469059256746e-05}
{"train_loss": 0.12801969051361084, "global_step": 171479, "epoch": 1926, "lr": 2.9992937756283267e-05}
{"train_loss": 0.07110930234193802, "global_step": 171480, "epoch": 1926, "lr": 2.999240645599952e-05}
{"train_loss": 0.11487732082605362, "global_step": 171481, "epoch": 1926, "lr": 2.9991875158405524e-05}
{"train_loss": 0.11012313514947891, "global_step": 171482, "epoch": 1926, "lr": 2.9991343863501392e-05}
{"train_loss": 0.14066408574581146, "global_step": 171483, "epoch": 1926, "lr": 2.9990812571287176e-05}
{"train_loss": 0.060980796813964844, "global_step": 171484, "epoch": 1926, "lr": 2.9990281281762956e-05}
{"train_loss": 0.07846836000680923, "global_step": 171485, "epoch": 1926, "lr": 2.9989749994928806e-05}
{"train_loss": 0.0803036317229271, "global_step": 171486, "epoch": 1926, "lr": 2.9989218710784767e-05}
{"train_loss": 0.048920776695013046, "global_step": 171487, "epoch": 1926, "lr": 2.998868742933096e-05}
{"train_loss": 0.06304527074098587, "global_step": 171488, "epoch": 1926, "lr": 2.9988156150567403e-05}
{"train_loss": 0.08365975320339203, "global_step": 171489, "epoch": 1926, "lr": 2.998762487449422e-05}
{"train_loss": 0.07017095386981964, "global_step": 171490, "epoch": 1926, "lr": 2.998709360111144e-05}
{"train_loss": 0.046615369617938995, "global_step": 171491, "epoch": 1926, "lr": 2.998656233041916e-05}
{"train_loss": 0.07524164766073227, "global_step": 171492, "epoch": 1926, "lr": 2.998603106241743e-05}
{"train_loss": 0.05509718880057335, "global_step": 171493, "epoch": 1926, "lr": 2.9985499797106352e-05}
{"train_loss": 0.10718139261007309, "global_step": 171494, "epoch": 1926, "lr": 2.9984968534485963e-05}
{"train_loss": 0.06211114302277565, "global_step": 171495, "epoch": 1926, "lr": 2.9984437274556367e-05}
{"train_loss": 0.07078853249549866, "global_step": 171496, "epoch": 1926, "lr": 2.9983906017317605e-05}
{"train_loss": 0.06254246085882187, "global_step": 171497, "epoch": 1926, "lr": 2.9983374762769778e-05}
{"train_loss": 0.10126516968011856, "global_step": 171498, "epoch": 1926, "lr": 2.998284351091292e-05}
{"train_loss": 0.03631368651986122, "global_step": 171499, "epoch": 1926, "lr": 2.9982312261747142e-05}
{"train_loss": 0.11928866803646088, "global_step": 171500, "epoch": 1926, "lr": 2.998178101527249e-05}
{"train_loss": 0.09884687513113022, "global_step": 171501, "epoch": 1926, "lr": 2.998124977148905e-05}
{"train_loss": 0.08447318092993136, "global_step": 171502, "epoch": 1926, "lr": 2.998071853039689e-05, "val_loss": 7.182399272918701}
{"train_loss": 0.057620201259851456, "global_step": 171503, "epoch": 1927, "lr": 2.998018729199607e-05}
{"train_loss": 0.0789099857211113, "global_step": 171504, "epoch": 1927, "lr": 2.9979656056286688e-05}
{"train_loss": 0.08036375045776367, "global_step": 171505, "epoch": 1927, "lr": 2.997912482326877e-05}
{"train_loss": 0.08563271909952164, "global_step": 171506, "epoch": 1927, "lr": 2.997859359294244e-05}
{"train_loss": 0.10795730352401733, "global_step": 171507, "epoch": 1927, "lr": 2.9978062365307712e-05}
{"train_loss": 0.08066560328006744, "global_step": 171508, "epoch": 1927, "lr": 2.9977531140364724e-05}
{"train_loss": 0.07568684220314026, "global_step": 171509, "epoch": 1927, "lr": 2.9976999918113484e-05}
{"train_loss": 0.0857788473367691, "global_step": 171510, "epoch": 1927, "lr": 2.9976468698554115e-05}
{"train_loss": 0.048658337444067, "global_step": 171511, "epoch": 1927, "lr": 2.997593748168665e-05}
{"train_loss": 0.059897273778915405, "global_step": 171512, "epoch": 1927, "lr": 2.9975406267511196e-05}
{"train_loss": 0.05253695696592331, "global_step": 171513, "epoch": 1927, "lr": 2.997487505602778e-05}
{"train_loss": 0.13586728274822235, "global_step": 171514, "epoch": 1927, "lr": 2.997434384723652e-05}
{"train_loss": 0.07511690258979797, "global_step": 171515, "epoch": 1927, "lr": 2.9973812641137443e-05}
{"train_loss": 0.046844206750392914, "global_step": 171516, "epoch": 1927, "lr": 2.997328143773066e-05}
{"train_loss": 0.07631375640630722, "global_step": 171517, "epoch": 1927, "lr": 2.9972750237016212e-05}
{"train_loss": 0.1312975138425827, "global_step": 171518, "epoch": 1927, "lr": 2.997221903899419e-05}
{"train_loss": 0.06345698982477188, "global_step": 171519, "epoch": 1927, "lr": 2.9971687843664664e-05}
{"train_loss": 0.1552678346633911, "global_step": 171520, "epoch": 1927, "lr": 2.99711566510277e-05}
{"train_loss": 0.1277182400226593, "global_step": 171521, "epoch": 1927, "lr": 2.9970625461083363e-05}
{"train_loss": 0.08137333393096924, "global_step": 171522, "epoch": 1927, "lr": 2.9970094273831735e-05}
{"train_loss": 0.04734458029270172, "global_step": 171523, "epoch": 1927, "lr": 2.996956308927289e-05}
{"train_loss": 0.06096242740750313, "global_step": 171524, "epoch": 1927, "lr": 2.996903190740688e-05}
{"train_loss": 0.08482376486063004, "global_step": 171525, "epoch": 1927, "lr": 2.9968500728233804e-05}
{"train_loss": 0.10138695687055588, "global_step": 171526, "epoch": 1927, "lr": 2.9967969551753705e-05}
{"train_loss": 0.09424862265586853, "global_step": 171527, "epoch": 1927, "lr": 2.9967438377966685e-05}
{"train_loss": 0.06677223742008209, "global_step": 171528, "epoch": 1927, "lr": 2.9966907206872773e-05}
{"train_loss": 0.09002441167831421, "global_step": 171529, "epoch": 1927, "lr": 2.99663760384721e-05}
{"train_loss": 0.11467347294092178, "global_step": 171530, "epoch": 1927, "lr": 2.996584487276467e-05}
{"train_loss": 0.08921055495738983, "global_step": 171531, "epoch": 1927, "lr": 2.9965313709750616e-05}
{"train_loss": 0.0589817613363266, "global_step": 171532, "epoch": 1927, "lr": 2.996478254942996e-05}
{"train_loss": 0.08822090923786163, "global_step": 171533, "epoch": 1927, "lr": 2.996425139180281e-05}
{"train_loss": 0.06707634031772614, "global_step": 171534, "epoch": 1927, "lr": 2.9963720236869203e-05}
{"train_loss": 0.16847455501556396, "global_step": 171535, "epoch": 1927, "lr": 2.9963189084629235e-05}
{"train_loss": 0.07642705738544464, "global_step": 171536, "epoch": 1927, "lr": 2.9962657935082987e-05}
{"train_loss": 0.08591428399085999, "global_step": 171537, "epoch": 1927, "lr": 2.9962126788230504e-05}
{"train_loss": 0.04855300113558769, "global_step": 171538, "epoch": 1927, "lr": 2.9961595644071877e-05}
{"train_loss": 0.0698482096195221, "global_step": 171539, "epoch": 1927, "lr": 2.996106450260716e-05}
{"train_loss": 0.06316887587308884, "global_step": 171540, "epoch": 1927, "lr": 2.996053336383644e-05}
{"train_loss": 0.08837603032588959, "global_step": 171541, "epoch": 1927, "lr": 2.996000222775978e-05}
{"train_loss": 0.0697033479809761, "global_step": 171542, "epoch": 1927, "lr": 2.9959471094377255e-05}
{"train_loss": 0.10311248898506165, "global_step": 171543, "epoch": 1927, "lr": 2.9958939963688936e-05}
{"train_loss": 0.10658341646194458, "global_step": 171544, "epoch": 1927, "lr": 2.9958408835694896e-05}
{"train_loss": 0.1180635616183281, "global_step": 171545, "epoch": 1927, "lr": 2.9957877710395186e-05}
{"train_loss": 0.12407345324754715, "global_step": 171546, "epoch": 1927, "lr": 2.9957346587789926e-05}
{"train_loss": 0.12952709197998047, "global_step": 171547, "epoch": 1927, "lr": 2.9956815467879125e-05}
{"train_loss": 0.05652695521712303, "global_step": 171548, "epoch": 1927, "lr": 2.9956284350662912e-05}
{"train_loss": 0.038646720349788666, "global_step": 171549, "epoch": 1927, "lr": 2.9955753236141303e-05}
{"train_loss": 0.14665953814983368, "global_step": 171550, "epoch": 1927, "lr": 2.995522212431443e-05}
{"train_loss": 0.04745728150010109, "global_step": 171551, "epoch": 1927, "lr": 2.99546910151823e-05}
{"train_loss": 0.0909113734960556, "global_step": 171552, "epoch": 1927, "lr": 2.995415990874505e-05}
{"train_loss": 0.06292637437582016, "global_step": 171553, "epoch": 1927, "lr": 2.9953628805002692e-05}
{"train_loss": 0.15475791692733765, "global_step": 171554, "epoch": 1927, "lr": 2.9953097703955325e-05}
{"train_loss": 0.08612443506717682, "global_step": 171555, "epoch": 1927, "lr": 2.9952566605603043e-05}
{"train_loss": 0.06862199306488037, "global_step": 171556, "epoch": 1927, "lr": 2.9952035509945875e-05}
{"train_loss": 0.0869465321302414, "global_step": 171557, "epoch": 1927, "lr": 2.995150441698393e-05}
{"train_loss": 0.06365957856178284, "global_step": 171558, "epoch": 1927, "lr": 2.9950973326717248e-05}
{"train_loss": 0.0885142907500267, "global_step": 171559, "epoch": 1927, "lr": 2.9950442239145914e-05}
{"train_loss": 0.1250014454126358, "global_step": 171560, "epoch": 1927, "lr": 2.9949911154269998e-05}
{"train_loss": 0.06504850089550018, "global_step": 171561, "epoch": 1927, "lr": 2.994938007208957e-05}
{"train_loss": 0.0809083878993988, "global_step": 171562, "epoch": 1927, "lr": 2.9948848992604712e-05}
{"train_loss": 0.09535421431064606, "global_step": 171563, "epoch": 1927, "lr": 2.9948317915815494e-05}
{"train_loss": 0.07079637795686722, "global_step": 171564, "epoch": 1927, "lr": 2.994778684172196e-05}
{"train_loss": 0.09495832026004791, "global_step": 171565, "epoch": 1927, "lr": 2.994725577032421e-05}
{"train_loss": 0.17736288905143738, "global_step": 171566, "epoch": 1927, "lr": 2.9946724701622305e-05}
{"train_loss": 0.1339678019285202, "global_step": 171567, "epoch": 1927, "lr": 2.9946193635616328e-05}
{"train_loss": 0.0715659111738205, "global_step": 171568, "epoch": 1927, "lr": 2.994566257230632e-05}
{"train_loss": 0.07387582212686539, "global_step": 171569, "epoch": 1927, "lr": 2.9945131511692397e-05}
{"train_loss": 0.09124299883842468, "global_step": 171570, "epoch": 1927, "lr": 2.9944600453774586e-05}
{"train_loss": 0.03853354603052139, "global_step": 171571, "epoch": 1927, "lr": 2.994406939855298e-05}
{"train_loss": 0.07340073585510254, "global_step": 171572, "epoch": 1927, "lr": 2.9943538346027666e-05}
{"train_loss": 0.08074746280908585, "global_step": 171573, "epoch": 1927, "lr": 2.9943007296198687e-05}
{"train_loss": 0.12705527245998383, "global_step": 171574, "epoch": 1927, "lr": 2.9942476249066136e-05}
{"train_loss": 0.09703221172094345, "global_step": 171575, "epoch": 1927, "lr": 2.9941945204630063e-05}
{"train_loss": 0.08962039649486542, "global_step": 171576, "epoch": 1927, "lr": 2.9941414162890568e-05}
{"train_loss": 0.04863107204437256, "global_step": 171577, "epoch": 1927, "lr": 2.994088312384769e-05}
{"train_loss": 0.08550948649644852, "global_step": 171578, "epoch": 1927, "lr": 2.9940352087501523e-05}
{"train_loss": 0.10775234550237656, "global_step": 171579, "epoch": 1927, "lr": 2.9939821053852124e-05}
{"train_loss": 0.0554378405213356, "global_step": 171580, "epoch": 1927, "lr": 2.993929002289958e-05}
{"train_loss": 0.05564219877123833, "global_step": 171581, "epoch": 1927, "lr": 2.9938758994643946e-05}
{"train_loss": 0.07642178982496262, "global_step": 171582, "epoch": 1927, "lr": 2.993822796908532e-05}
{"train_loss": 0.09557078778743744, "global_step": 171583, "epoch": 1927, "lr": 2.9937696946223726e-05}
{"train_loss": 0.09333959966897964, "global_step": 171584, "epoch": 1927, "lr": 2.993716592605929e-05}
{"train_loss": 0.04796924814581871, "global_step": 171585, "epoch": 1927, "lr": 2.9936634908592036e-05}
{"train_loss": 0.13353636860847473, "global_step": 171586, "epoch": 1927, "lr": 2.9936103893822076e-05}
{"train_loss": 0.1195412203669548, "global_step": 171587, "epoch": 1927, "lr": 2.993557288174944e-05}
{"train_loss": 0.14817796647548676, "global_step": 171588, "epoch": 1927, "lr": 2.9935041872374248e-05}
{"train_loss": 0.12338145822286606, "global_step": 171589, "epoch": 1927, "lr": 2.993451086569652e-05}
{"train_loss": 0.1096435934305191, "global_step": 171590, "epoch": 1927, "lr": 2.9933979861716356e-05}
{"train_loss": 0.08840147556548708, "global_step": 171591, "epoch": 1927, "lr": 2.9933448860433844e-05, "val_loss": 7.181080341339111}
{"train_loss": 0.07419408857822418, "global_step": 171592, "epoch": 1928, "lr": 2.993291786184901e-05}
{"train_loss": 0.12280957400798798, "global_step": 171593, "epoch": 1928, "lr": 2.993238686596197e-05}
{"train_loss": 0.0699225515127182, "global_step": 171594, "epoch": 1928, "lr": 2.993185587277276e-05}
{"train_loss": 0.16210807859897614, "global_step": 171595, "epoch": 1928, "lr": 2.9931324882281493e-05}
{"train_loss": 0.07391315698623657, "global_step": 171596, "epoch": 1928, "lr": 2.9930793894488197e-05}
{"train_loss": 0.09000617265701294, "global_step": 171597, "epoch": 1928, "lr": 2.993026290939296e-05}
{"train_loss": 0.08998633921146393, "global_step": 171598, "epoch": 1928, "lr": 2.9929731926995863e-05}
{"train_loss": 0.11151183396577835, "global_step": 171599, "epoch": 1928, "lr": 2.992920094729696e-05}
{"train_loss": 0.15350839495658875, "global_step": 171600, "epoch": 1928, "lr": 2.9928669970296335e-05}
{"train_loss": 0.14569208025932312, "global_step": 171601, "epoch": 1928, "lr": 2.9928138995994065e-05}
{"train_loss": 0.08135329186916351, "global_step": 171602, "epoch": 1928, "lr": 2.9927608024390198e-05}
{"train_loss": 0.10736963897943497, "global_step": 171603, "epoch": 1928, "lr": 2.9927077055484833e-05}
{"train_loss": 0.08117348700761795, "global_step": 171604, "epoch": 1928, "lr": 2.9926546089278006e-05}
{"train_loss": 0.0632266104221344, "global_step": 171605, "epoch": 1928, "lr": 2.9926015125769836e-05}
{"train_loss": 0.17693591117858887, "global_step": 171606, "epoch": 1928, "lr": 2.9925484164960337e-05}
{"train_loss": 0.04073663800954819, "global_step": 171607, "epoch": 1928, "lr": 2.9924953206849647e-05}
{"train_loss": 0.06656444072723389, "global_step": 171608, "epoch": 1928, "lr": 2.9924422251437774e-05}
{"train_loss": 0.02992374822497368, "global_step": 171609, "epoch": 1928, "lr": 2.9923891298724816e-05}
{"train_loss": 0.08541510999202728, "global_step": 171610, "epoch": 1928, "lr": 2.992336034871087e-05}
{"train_loss": 0.10419397801160812, "global_step": 171611, "epoch": 1928, "lr": 2.9922829401395968e-05}
{"train_loss": 0.08364702016115189, "global_step": 171612, "epoch": 1928, "lr": 2.9922298456780207e-05}
{"train_loss": 0.11969978362321854, "global_step": 171613, "epoch": 1928, "lr": 2.9921767514863636e-05}
{"train_loss": 0.13386785984039307, "global_step": 171614, "epoch": 1928, "lr": 2.992123657564636e-05}
{"train_loss": 0.08128469437360764, "global_step": 171615, "epoch": 1928, "lr": 2.992070563912841e-05}
{"train_loss": 0.11011973768472672, "global_step": 171616, "epoch": 1928, "lr": 2.9920174705309882e-05}
{"train_loss": 0.10465288162231445, "global_step": 171617, "epoch": 1928, "lr": 2.9919643774190843e-05}
{"train_loss": 0.060628119856119156, "global_step": 171618, "epoch": 1928, "lr": 2.991911284577136e-05}
{"train_loss": 0.06749061495065689, "global_step": 171619, "epoch": 1928, "lr": 2.9918581920051514e-05}
{"train_loss": 0.06508826464414597, "global_step": 171620, "epoch": 1928, "lr": 2.9918050997031373e-05}
{"train_loss": 0.14412781596183777, "global_step": 171621, "epoch": 1928, "lr": 2.991752007671098e-05}
{"train_loss": 0.12017758190631866, "global_step": 171622, "epoch": 1928, "lr": 2.9916989159090465e-05}
{"train_loss": 0.06379164010286331, "global_step": 171623, "epoch": 1928, "lr": 2.9916458244169836e-05}
{"train_loss": 0.10457843542098999, "global_step": 171624, "epoch": 1928, "lr": 2.9915927331949224e-05}
{"train_loss": 0.08266375213861465, "global_step": 171625, "epoch": 1928, "lr": 2.9915396422428643e-05}
{"train_loss": 0.04083981737494469, "global_step": 171626, "epoch": 1928, "lr": 2.991486551560819e-05}
{"train_loss": 0.09313313663005829, "global_step": 171627, "epoch": 1928, "lr": 2.9914334611487964e-05}
{"train_loss": 0.07245572656393051, "global_step": 171628, "epoch": 1928, "lr": 2.991380371006799e-05}
{"train_loss": 0.06439200788736343, "global_step": 171629, "epoch": 1928, "lr": 2.991327281134838e-05}
{"train_loss": 0.061400871723890305, "global_step": 171630, "epoch": 1928, "lr": 2.9912741915329168e-05}
{"train_loss": 0.07910667359828949, "global_step": 171631, "epoch": 1928, "lr": 2.991221102201046e-05}
{"train_loss": 0.062420547008514404, "global_step": 171632, "epoch": 1928, "lr": 2.991168013139229e-05}
{"train_loss": 0.08590200543403625, "global_step": 171633, "epoch": 1928, "lr": 2.991114924347477e-05}
{"train_loss": 0.07601255178451538, "global_step": 171634, "epoch": 1928, "lr": 2.9910618358257942e-05}
{"train_loss": 0.10582359880208969, "global_step": 171635, "epoch": 1928, "lr": 2.9910087475741887e-05}
{"train_loss": 0.12651024758815765, "global_step": 171636, "epoch": 1928, "lr": 2.9909556595926674e-05}
{"train_loss": 0.075723797082901, "global_step": 171637, "epoch": 1928, "lr": 2.990902571881238e-05}
{"train_loss": 0.09620968252420425, "global_step": 171638, "epoch": 1928, "lr": 2.990849484439907e-05}
{"train_loss": 0.10869454592466354, "global_step": 171639, "epoch": 1928, "lr": 2.9907963972686827e-05}
{"train_loss": 0.07405992597341537, "global_step": 171640, "epoch": 1928, "lr": 2.9907433103675696e-05}
{"train_loss": 0.04916761815547943, "global_step": 171641, "epoch": 1928, "lr": 2.990690223736578e-05}
{"train_loss": 0.1559634804725647, "global_step": 171642, "epoch": 1928, "lr": 2.9906371373757126e-05}
{"train_loss": 0.07837792485952377, "global_step": 171643, "epoch": 1928, "lr": 2.9905840512849827e-05}
{"train_loss": 0.1519528180360794, "global_step": 171644, "epoch": 1928, "lr": 2.990530965464392e-05}
{"train_loss": 0.06633497774600983, "global_step": 171645, "epoch": 1928, "lr": 2.990477879913951e-05}
{"train_loss": 0.08678357303142548, "global_step": 171646, "epoch": 1928, "lr": 2.9904247946336674e-05}
{"train_loss": 0.10705477744340897, "global_step": 171647, "epoch": 1928, "lr": 2.990371709623544e-05}
{"train_loss": 0.03445485234260559, "global_step": 171648, "epoch": 1928, "lr": 2.9903186248835935e-05}
{"train_loss": 0.10404011607170105, "global_step": 171649, "epoch": 1928, "lr": 2.990265540413818e-05}
{"train_loss": 0.13287729024887085, "global_step": 171650, "epoch": 1928, "lr": 2.9902124562142287e-05}
{"train_loss": 0.07766199856996536, "global_step": 171651, "epoch": 1928, "lr": 2.9901593722848287e-05}
{"train_loss": 0.0896201804280281, "global_step": 171652, "epoch": 1928, "lr": 2.9901062886256292e-05}
{"train_loss": 0.1311080902814865, "global_step": 171653, "epoch": 1928, "lr": 2.9900532052366342e-05}
{"train_loss": 0.1345670223236084, "global_step": 171654, "epoch": 1928, "lr": 2.9900001221178525e-05}
{"train_loss": 0.09621714800596237, "global_step": 171655, "epoch": 1928, "lr": 2.9899470392692906e-05}
{"train_loss": 0.12787143886089325, "global_step": 171656, "epoch": 1928, "lr": 2.9898939566909556e-05}
{"train_loss": 0.10805098712444305, "global_step": 171657, "epoch": 1928, "lr": 2.9898408743828556e-05}
{"train_loss": 0.06010642647743225, "global_step": 171658, "epoch": 1928, "lr": 2.9897877923449967e-05}
{"train_loss": 0.07240626960992813, "global_step": 171659, "epoch": 1928, "lr": 2.989734710577386e-05}
{"train_loss": 0.07655450701713562, "global_step": 171660, "epoch": 1928, "lr": 2.989681629080032e-05}
{"train_loss": 0.0627971962094307, "global_step": 171661, "epoch": 1928, "lr": 2.989628547852939e-05}
{"train_loss": 0.07860257476568222, "global_step": 171662, "epoch": 1928, "lr": 2.9895754668961163e-05}
{"train_loss": 0.1058807224035263, "global_step": 171663, "epoch": 1928, "lr": 2.989522386209572e-05}
{"train_loss": 0.08489768952131271, "global_step": 171664, "epoch": 1928, "lr": 2.9894693057933097e-05}
{"train_loss": 0.08391920477151871, "global_step": 171665, "epoch": 1928, "lr": 2.9894162256473413e-05}
{"train_loss": 0.06244064122438431, "global_step": 171666, "epoch": 1928, "lr": 2.98936314577167e-05}
{"train_loss": 0.10080307722091675, "global_step": 171667, "epoch": 1928, "lr": 2.9893100661663053e-05}
{"train_loss": 0.08179182559251785, "global_step": 171668, "epoch": 1928, "lr": 2.9892569868312513e-05}
{"train_loss": 0.09180562198162079, "global_step": 171669, "epoch": 1928, "lr": 2.98920390776652e-05}
{"train_loss": 0.06896693259477615, "global_step": 171670, "epoch": 1928, "lr": 2.9891508289721133e-05}
{"train_loss": 0.08075068891048431, "global_step": 171671, "epoch": 1928, "lr": 2.989097750448043e-05}
{"train_loss": 0.16403605043888092, "global_step": 171672, "epoch": 1928, "lr": 2.989044672194311e-05}
{"train_loss": 0.04584317281842232, "global_step": 171673, "epoch": 1928, "lr": 2.9889915942109303e-05}
{"train_loss": 0.07663270086050034, "global_step": 171674, "epoch": 1928, "lr": 2.9889385164979044e-05}
{"train_loss": 0.08867165446281433, "global_step": 171675, "epoch": 1928, "lr": 2.9888854390552413e-05}
{"train_loss": 0.12185521423816681, "global_step": 171676, "epoch": 1928, "lr": 2.9888323618829474e-05}
{"train_loss": 0.06658479571342468, "global_step": 171677, "epoch": 1928, "lr": 2.9887792849810304e-05}
{"train_loss": 0.11297136545181274, "global_step": 171678, "epoch": 1928, "lr": 2.9887262083494975e-05}
{"train_loss": 0.08932331204414368, "global_step": 171679, "epoch": 1928, "lr": 2.9886731319883576e-05}
{"train_loss": 0.09222388815846336, "global_step": 171680, "epoch": 1928, "lr": 2.9886200558976136e-05, "val_loss": 7.189910888671875}
{"train_loss": 0.07091273367404938, "global_step": 171681, "epoch": 1929, "lr": 2.9885669800772754e-05}
{"train_loss": 0.055095963180065155, "global_step": 171682, "epoch": 1929, "lr": 2.9885139045273514e-05}
{"train_loss": 0.05276435613632202, "global_step": 171683, "epoch": 1929, "lr": 2.9884608292478454e-05}
{"train_loss": 0.11330165714025497, "global_step": 171684, "epoch": 1929, "lr": 2.988407754238768e-05}
{"train_loss": 0.07913537323474884, "global_step": 171685, "epoch": 1929, "lr": 2.9883546795001226e-05}
{"train_loss": 0.07202967256307602, "global_step": 171686, "epoch": 1929, "lr": 2.9883016050319214e-05}
{"train_loss": 0.13757382333278656, "global_step": 171687, "epoch": 1929, "lr": 2.9882485308341656e-05}
{"train_loss": 0.04796101897954941, "global_step": 171688, "epoch": 1929, "lr": 2.9881954569068678e-05}
{"train_loss": 0.05754769593477249, "global_step": 171689, "epoch": 1929, "lr": 2.98814238325003e-05}
{"train_loss": 0.09258008748292923, "global_step": 171690, "epoch": 1929, "lr": 2.9880893098636642e-05}
{"train_loss": 0.08845046907663345, "global_step": 171691, "epoch": 1929, "lr": 2.9880362367477738e-05}
{"train_loss": 0.048579808324575424, "global_step": 171692, "epoch": 1929, "lr": 2.987983163902369e-05}
{"train_loss": 0.07567343860864639, "global_step": 171693, "epoch": 1929, "lr": 2.9879300913274543e-05}
{"train_loss": 0.06588877737522125, "global_step": 171694, "epoch": 1929, "lr": 2.9878770190230377e-05}
{"train_loss": 0.06461963802576065, "global_step": 171695, "epoch": 1929, "lr": 2.9878239469891268e-05}
{"train_loss": 0.057047102600336075, "global_step": 171696, "epoch": 1929, "lr": 2.9877708752257284e-05}
{"train_loss": 0.027150804176926613, "global_step": 171697, "epoch": 1929, "lr": 2.9877178037328497e-05}
{"train_loss": 0.12103987485170364, "global_step": 171698, "epoch": 1929, "lr": 2.9876647325104967e-05}
{"train_loss": 0.061583105474710464, "global_step": 171699, "epoch": 1929, "lr": 2.9876116615586792e-05}
{"train_loss": 0.10662033408880234, "global_step": 171700, "epoch": 1929, "lr": 2.9875585908774017e-05}
{"train_loss": 0.09116720408201218, "global_step": 171701, "epoch": 1929, "lr": 2.9875055204666736e-05}
{"train_loss": 0.10118422657251358, "global_step": 171702, "epoch": 1929, "lr": 2.9874524503264994e-05}
{"train_loss": 0.07060129940509796, "global_step": 171703, "epoch": 1929, "lr": 2.9873993804568902e-05}
{"train_loss": 0.11760182678699493, "global_step": 171704, "epoch": 1929, "lr": 2.987346310857847e-05}
{"train_loss": 0.10482519865036011, "global_step": 171705, "epoch": 1929, "lr": 2.987293241529384e-05}
{"train_loss": 0.07138308137655258, "global_step": 171706, "epoch": 1929, "lr": 2.987240172471503e-05}
{"train_loss": 0.09556179493665695, "global_step": 171707, "epoch": 1929, "lr": 2.9871871036842137e-05}
{"train_loss": 0.04865790158510208, "global_step": 171708, "epoch": 1929, "lr": 2.9871340351675214e-05}
{"train_loss": 0.08797388523817062, "global_step": 171709, "epoch": 1929, "lr": 2.9870809669214362e-05}
{"train_loss": 0.07873661816120148, "global_step": 171710, "epoch": 1929, "lr": 2.9870278989459617e-05}
{"train_loss": 0.10572519153356552, "global_step": 171711, "epoch": 1929, "lr": 2.986974831241108e-05}
{"train_loss": 0.09389730542898178, "global_step": 171712, "epoch": 1929, "lr": 2.9869217638068813e-05}
{"train_loss": 0.05385233461856842, "global_step": 171713, "epoch": 1929, "lr": 2.9868686966432872e-05}
{"train_loss": 0.06967797875404358, "global_step": 171714, "epoch": 1929, "lr": 2.9868156297503348e-05}
{"train_loss": 0.11686341464519501, "global_step": 171715, "epoch": 1929, "lr": 2.9867625631280307e-05}
{"train_loss": 0.08083236217498779, "global_step": 171716, "epoch": 1929, "lr": 2.986709496776381e-05}
{"train_loss": 0.10138262063264847, "global_step": 171717, "epoch": 1929, "lr": 2.986656430695393e-05}
{"train_loss": 0.03636957332491875, "global_step": 171718, "epoch": 1929, "lr": 2.9866033648850767e-05}
{"train_loss": 0.11230272054672241, "global_step": 171719, "epoch": 1929, "lr": 2.9865502993454352e-05}
{"train_loss": 0.09294848889112473, "global_step": 171720, "epoch": 1929, "lr": 2.9864972340764786e-05}
{"train_loss": 0.07602395117282867, "global_step": 171721, "epoch": 1929, "lr": 2.9864441690782118e-05}
{"train_loss": 0.1024819016456604, "global_step": 171722, "epoch": 1929, "lr": 2.986391104350645e-05}
{"train_loss": 0.07901116460561752, "global_step": 171723, "epoch": 1929, "lr": 2.986338039893781e-05}
{"train_loss": 0.08032872527837753, "global_step": 171724, "epoch": 1929, "lr": 2.9862849757076316e-05}
{"train_loss": 0.06393339484930038, "global_step": 171725, "epoch": 1929, "lr": 2.9862319117921988e-05}
{"train_loss": 0.07674214243888855, "global_step": 171726, "epoch": 1929, "lr": 2.9861788481474957e-05}
{"train_loss": 0.12396065890789032, "global_step": 171727, "epoch": 1929, "lr": 2.9861257847735236e-05}
{"train_loss": 0.09511683881282806, "global_step": 171728, "epoch": 1929, "lr": 2.9860727216702945e-05}
{"train_loss": 0.08476570248603821, "global_step": 171729, "epoch": 1929, "lr": 2.986019658837812e-05}
{"train_loss": 0.10137121379375458, "global_step": 171730, "epoch": 1929, "lr": 2.985966596276086e-05}
{"train_loss": 0.07003246247768402, "global_step": 171731, "epoch": 1929, "lr": 2.9859135339851207e-05}
{"train_loss": 0.08777759969234467, "global_step": 171732, "epoch": 1929, "lr": 2.9858604719649258e-05}
{"train_loss": 0.07443039119243622, "global_step": 171733, "epoch": 1929, "lr": 2.985807410215507e-05}
{"train_loss": 0.12720143795013428, "global_step": 171734, "epoch": 1929, "lr": 2.9857543487368722e-05}
{"train_loss": 0.09163831919431686, "global_step": 171735, "epoch": 1929, "lr": 2.9857012875290273e-05}
{"train_loss": 0.08149760961532593, "global_step": 171736, "epoch": 1929, "lr": 2.98564822659198e-05}
{"train_loss": 0.08618620038032532, "global_step": 171737, "epoch": 1929, "lr": 2.9855951659257397e-05}
{"train_loss": 0.07591436058282852, "global_step": 171738, "epoch": 1929, "lr": 2.9855421055303096e-05}
{"train_loss": 0.09463109821081161, "global_step": 171739, "epoch": 1929, "lr": 2.9854890454057006e-05}
{"train_loss": 0.10548023134469986, "global_step": 171740, "epoch": 1929, "lr": 2.9854359855519158e-05}
{"train_loss": 0.14245381951332092, "global_step": 171741, "epoch": 1929, "lr": 2.985382925968967e-05}
{"train_loss": 0.045459944754838943, "global_step": 171742, "epoch": 1929, "lr": 2.985329866656857e-05}
{"train_loss": 0.0697169080376625, "global_step": 171743, "epoch": 1929, "lr": 2.985276807615597e-05}
{"train_loss": 0.13926973938941956, "global_step": 171744, "epoch": 1929, "lr": 2.9852237488451896e-05}
{"train_loss": 0.06680132448673248, "global_step": 171745, "epoch": 1929, "lr": 2.9851706903456462e-05}
{"train_loss": 0.08185996860265732, "global_step": 171746, "epoch": 1929, "lr": 2.98511763211697e-05}
{"train_loss": 0.03386666625738144, "global_step": 171747, "epoch": 1929, "lr": 2.985064574159172e-05}
{"train_loss": 0.06553947180509567, "global_step": 171748, "epoch": 1929, "lr": 2.985011516472256e-05}
{"train_loss": 0.0910949632525444, "global_step": 171749, "epoch": 1929, "lr": 2.984958459056233e-05}
{"train_loss": 0.07734722644090652, "global_step": 171750, "epoch": 1929, "lr": 2.984905401911106e-05}
{"train_loss": 0.11918259412050247, "global_step": 171751, "epoch": 1929, "lr": 2.984852345036884e-05}
{"train_loss": 0.09180086851119995, "global_step": 171752, "epoch": 1929, "lr": 2.9847992884335747e-05}
{"train_loss": 0.03397171571850777, "global_step": 171753, "epoch": 1929, "lr": 2.984746232101183e-05}
{"train_loss": 0.10459309816360474, "global_step": 171754, "epoch": 1929, "lr": 2.9846931760397196e-05}
{"train_loss": 0.05572397634387016, "global_step": 171755, "epoch": 1929, "lr": 2.984640120249188e-05}
{"train_loss": 0.06907349824905396, "global_step": 171756, "epoch": 1929, "lr": 2.984587064729598e-05}
{"train_loss": 0.028532590717077255, "global_step": 171757, "epoch": 1929, "lr": 2.984534009480955e-05}
{"train_loss": 0.07297933846712112, "global_step": 171758, "epoch": 1929, "lr": 2.984480954503268e-05}
{"train_loss": 0.0817028060555458, "global_step": 171759, "epoch": 1929, "lr": 2.9844278997965414e-05}
{"train_loss": 0.07821760326623917, "global_step": 171760, "epoch": 1929, "lr": 2.9843748453607856e-05}
{"train_loss": 0.06804674118757248, "global_step": 171761, "epoch": 1929, "lr": 2.984321791196004e-05}
{"train_loss": 0.08392595499753952, "global_step": 171762, "epoch": 1929, "lr": 2.984268737302208e-05}
{"train_loss": 0.07460261881351471, "global_step": 171763, "epoch": 1929, "lr": 2.984215683679401e-05}
{"train_loss": 0.08176383376121521, "global_step": 171764, "epoch": 1929, "lr": 2.9841626303275925e-05}
{"train_loss": 0.09927518665790558, "global_step": 171765, "epoch": 1929, "lr": 2.9841095772467876e-05}
{"train_loss": 0.0600275918841362, "global_step": 171766, "epoch": 1929, "lr": 2.9840565244369965e-05}
{"train_loss": 0.0903199091553688, "global_step": 171767, "epoch": 1929, "lr": 2.9840034718982225e-05}
{"train_loss": 0.08313926309347153, "global_step": 171768, "epoch": 1929, "lr": 2.9839504196304756e-05}
{"train_loss": 0.0824665477645866, "global_step": 171769, "epoch": 1929, "lr": 2.983897367633761e-05, "val_loss": 7.139646053314209}
{"train_loss": 0.13634708523750305, "global_step": 171770, "epoch": 1930, "lr": 2.9838443159080888e-05}
{"train_loss": 0.08076140284538269, "global_step": 171771, "epoch": 1930, "lr": 2.983791264453463e-05}
{"train_loss": 0.0793396383523941, "global_step": 171772, "epoch": 1930, "lr": 2.983738213269891e-05}
{"train_loss": 0.05333207547664642, "global_step": 171773, "epoch": 1930, "lr": 2.9836851623573826e-05}
{"train_loss": 0.10747965425252914, "global_step": 171774, "epoch": 1930, "lr": 2.9836321117159417e-05}
{"train_loss": 0.049625881016254425, "global_step": 171775, "epoch": 1930, "lr": 2.9835790613455785e-05}
{"train_loss": 0.1373605579137802, "global_step": 171776, "epoch": 1930, "lr": 2.983526011246296e-05}
{"train_loss": 0.11998354643583298, "global_step": 171777, "epoch": 1930, "lr": 2.9834729614181066e-05}
{"train_loss": 0.10723844915628433, "global_step": 171778, "epoch": 1930, "lr": 2.9834199118610122e-05}
{"train_loss": 0.08347909897565842, "global_step": 171779, "epoch": 1930, "lr": 2.983366862575025e-05}
{"train_loss": 0.10106614232063293, "global_step": 171780, "epoch": 1930, "lr": 2.983313813560147e-05}
{"train_loss": 0.07966096699237823, "global_step": 171781, "epoch": 1930, "lr": 2.98326076481639e-05}
{"train_loss": 0.13337597250938416, "global_step": 171782, "epoch": 1930, "lr": 2.983207716343757e-05}
{"train_loss": 0.08221454173326492, "global_step": 171783, "epoch": 1930, "lr": 2.9831546681422597e-05}
{"train_loss": 0.06603508442640305, "global_step": 171784, "epoch": 1930, "lr": 2.9831016202119e-05}
{"train_loss": 0.10863813757896423, "global_step": 171785, "epoch": 1930, "lr": 2.9830485725526895e-05}
{"train_loss": 0.08679261058568954, "global_step": 171786, "epoch": 1930, "lr": 2.9829955251646324e-05}
{"train_loss": 0.06904704868793488, "global_step": 171787, "epoch": 1930, "lr": 2.9829424780477383e-05}
{"train_loss": 0.06957175582647324, "global_step": 171788, "epoch": 1930, "lr": 2.9828894312020112e-05}
{"train_loss": 0.07531500607728958, "global_step": 171789, "epoch": 1930, "lr": 2.9828363846274608e-05}
{"train_loss": 0.07891138643026352, "global_step": 171790, "epoch": 1930, "lr": 2.9827833383240944e-05}
{"train_loss": 0.10982169955968857, "global_step": 171791, "epoch": 1930, "lr": 2.9827302922919165e-05}
{"train_loss": 0.06530028581619263, "global_step": 171792, "epoch": 1930, "lr": 2.982677246530938e-05}
{"train_loss": 0.11437249928712845, "global_step": 171793, "epoch": 1930, "lr": 2.982624201041162e-05}
{"train_loss": 0.11834736913442612, "global_step": 171794, "epoch": 1930, "lr": 2.9825711558226e-05}
{"train_loss": 0.1079494059085846, "global_step": 171795, "epoch": 1930, "lr": 2.9825181108752537e-05}
{"train_loss": 0.07868098467588425, "global_step": 171796, "epoch": 1930, "lr": 2.982465066199136e-05}
{"train_loss": 0.09063229709863663, "global_step": 171797, "epoch": 1930, "lr": 2.9824120217942485e-05}
{"train_loss": 0.05888756737112999, "global_step": 171798, "epoch": 1930, "lr": 2.9823589776606037e-05}
{"train_loss": 0.12379291653633118, "global_step": 171799, "epoch": 1930, "lr": 2.982305933798204e-05}
{"train_loss": 0.13952675461769104, "global_step": 171800, "epoch": 1930, "lr": 2.982252890207061e-05}
{"train_loss": 0.0685269832611084, "global_step": 171801, "epoch": 1930, "lr": 2.982199846887177e-05}
{"train_loss": 0.04836711287498474, "global_step": 171802, "epoch": 1930, "lr": 2.9821468038385637e-05}
{"train_loss": 0.09250801801681519, "global_step": 171803, "epoch": 1930, "lr": 2.9820937610612244e-05}
{"train_loss": 0.08003953099250793, "global_step": 171804, "epoch": 1930, "lr": 2.98204071855517e-05}
{"train_loss": 0.06938941776752472, "global_step": 171805, "epoch": 1930, "lr": 2.9819876763204034e-05}
{"train_loss": 0.05387349799275398, "global_step": 171806, "epoch": 1930, "lr": 2.9819346343569364e-05}
{"train_loss": 0.13727471232414246, "global_step": 171807, "epoch": 1930, "lr": 2.9818815926647713e-05}
{"train_loss": 0.07236208766698837, "global_step": 171808, "epoch": 1930, "lr": 2.9818285512439193e-05}
{"train_loss": 0.07185883820056915, "global_step": 171809, "epoch": 1930, "lr": 2.9817755100943855e-05}
{"train_loss": 0.07927358895540237, "global_step": 171810, "epoch": 1930, "lr": 2.9817224692161767e-05}
{"train_loss": 0.07660567760467529, "global_step": 171811, "epoch": 1930, "lr": 2.9816694286093017e-05}
{"train_loss": 0.09701799601316452, "global_step": 171812, "epoch": 1930, "lr": 2.981616388273766e-05}
{"train_loss": 0.04054757580161095, "global_step": 171813, "epoch": 1930, "lr": 2.9815633482095785e-05}
{"train_loss": 0.08730967342853546, "global_step": 171814, "epoch": 1930, "lr": 2.9815103084167428e-05}
{"train_loss": 0.1395285725593567, "global_step": 171815, "epoch": 1930, "lr": 2.9814572688952707e-05}
{"train_loss": 0.09395171701908112, "global_step": 171816, "epoch": 1930, "lr": 2.981404229645165e-05}
{"train_loss": 0.13847360014915466, "global_step": 171817, "epoch": 1930, "lr": 2.981351190666437e-05}
{"train_loss": 0.10391182452440262, "global_step": 171818, "epoch": 1930, "lr": 2.98129815195909e-05}
{"train_loss": 0.06957056373357773, "global_step": 171819, "epoch": 1930, "lr": 2.981245113523135e-05}
{"train_loss": 0.0810449942946434, "global_step": 171820, "epoch": 1930, "lr": 2.9811920753585748e-05}
{"train_loss": 0.01925119198858738, "global_step": 171821, "epoch": 1930, "lr": 2.9811390374654202e-05}
{"train_loss": 0.13072113692760468, "global_step": 171822, "epoch": 1930, "lr": 2.9810859998436757e-05}
{"train_loss": 0.06595896184444427, "global_step": 171823, "epoch": 1930, "lr": 2.981032962493351e-05}
{"train_loss": 0.07018715143203735, "global_step": 171824, "epoch": 1930, "lr": 2.98097992541445e-05}
{"train_loss": 0.08615730702877045, "global_step": 171825, "epoch": 1930, "lr": 2.9809268886069825e-05}
{"train_loss": 0.03607865050435066, "global_step": 171826, "epoch": 1930, "lr": 2.980873852070955e-05}
{"train_loss": 0.07756688445806503, "global_step": 171827, "epoch": 1930, "lr": 2.9808208158063743e-05}
{"train_loss": 0.09597431123256683, "global_step": 171828, "epoch": 1930, "lr": 2.980767779813248e-05}
{"train_loss": 0.05586406588554382, "global_step": 171829, "epoch": 1930, "lr": 2.9807147440915818e-05}
{"train_loss": 0.09076815098524094, "global_step": 171830, "epoch": 1930, "lr": 2.9806617086413857e-05}
{"train_loss": 0.09529316425323486, "global_step": 171831, "epoch": 1930, "lr": 2.9806086734626624e-05}
{"train_loss": 0.10180030763149261, "global_step": 171832, "epoch": 1930, "lr": 2.9805556385554247e-05}
{"train_loss": 0.06840310245752335, "global_step": 171833, "epoch": 1930, "lr": 2.980502603919674e-05}
{"train_loss": 0.056574683636426926, "global_step": 171834, "epoch": 1930, "lr": 2.9804495695554226e-05}
{"train_loss": 0.057085733860731125, "global_step": 171835, "epoch": 1930, "lr": 2.9803965354626724e-05}
{"train_loss": 0.07439778745174408, "global_step": 171836, "epoch": 1930, "lr": 2.980343501641436e-05}
{"train_loss": 0.10174129158258438, "global_step": 171837, "epoch": 1930, "lr": 2.9802904680917155e-05}
{"train_loss": 0.07744325697422028, "global_step": 171838, "epoch": 1930, "lr": 2.980237434813522e-05}
{"train_loss": 0.05117979645729065, "global_step": 171839, "epoch": 1930, "lr": 2.9801844018068592e-05}
{"train_loss": 0.09383022785186768, "global_step": 171840, "epoch": 1930, "lr": 2.9801313690717382e-05}
{"train_loss": 0.07255066186189651, "global_step": 171841, "epoch": 1930, "lr": 2.9800783366081614e-05}
{"train_loss": 0.11190467327833176, "global_step": 171842, "epoch": 1930, "lr": 2.980025304416141e-05}
{"train_loss": 0.052954740822315216, "global_step": 171843, "epoch": 1930, "lr": 2.9799722724956792e-05}
{"train_loss": 0.07917654514312744, "global_step": 171844, "epoch": 1930, "lr": 2.979919240846787e-05}
{"train_loss": 0.07562161237001419, "global_step": 171845, "epoch": 1930, "lr": 2.9798662094694695e-05}
{"train_loss": 0.057779401540756226, "global_step": 171846, "epoch": 1930, "lr": 2.979813178363734e-05}
{"train_loss": 0.042827654629945755, "global_step": 171847, "epoch": 1930, "lr": 2.97976014752959e-05}
{"train_loss": 0.09967375546693802, "global_step": 171848, "epoch": 1930, "lr": 2.97970711696704e-05}
{"train_loss": 0.08449067920446396, "global_step": 171849, "epoch": 1930, "lr": 2.979654086676096e-05}
{"train_loss": 0.08599615842103958, "global_step": 171850, "epoch": 1930, "lr": 2.979601056656761e-05}
{"train_loss": 0.08476395159959793, "global_step": 171851, "epoch": 1930, "lr": 2.979548026909046e-05}
{"train_loss": 0.04051404818892479, "global_step": 171852, "epoch": 1930, "lr": 2.9794949974329533e-05}
{"train_loss": 0.09749399870634079, "global_step": 171853, "epoch": 1930, "lr": 2.9794419682284957e-05}
{"train_loss": 0.140231654047966, "global_step": 171854, "epoch": 1930, "lr": 2.9793889392956754e-05}
{"train_loss": 0.09308328479528427, "global_step": 171855, "epoch": 1930, "lr": 2.9793359106345037e-05}
{"train_loss": 0.12604059278964996, "global_step": 171856, "epoch": 1930, "lr": 2.9792828822449836e-05}
{"train_loss": 0.10457396507263184, "global_step": 171857, "epoch": 1930, "lr": 2.979229854127126e-05}
{"train_loss": 0.08721416721936692, "global_step": 171858, "epoch": 1930, "lr": 2.979176826280935e-05, "val_loss": 7.135758876800537, "train_action_mse_error": 24.668621063232422}
{"train_loss": 0.10637468844652176, "global_step": 171859, "epoch": 1931, "lr": 2.9791237987064202e-05}
{"train_loss": 0.0616762712597847, "global_step": 171860, "epoch": 1931, "lr": 2.979070771403586e-05}
{"train_loss": 0.05883157253265381, "global_step": 171861, "epoch": 1931, "lr": 2.9790177443724432e-05}
{"train_loss": 0.11854878813028336, "global_step": 171862, "epoch": 1931, "lr": 2.9789647176129943e-05}
{"train_loss": 0.0382540188729763, "global_step": 171863, "epoch": 1931, "lr": 2.9789116911252508e-05}
{"train_loss": 0.09365567564964294, "global_step": 171864, "epoch": 1931, "lr": 2.9788586649092176e-05}
{"train_loss": 0.09605584293603897, "global_step": 171865, "epoch": 1931, "lr": 2.9788056389649023e-05}
{"train_loss": 0.14635686576366425, "global_step": 171866, "epoch": 1931, "lr": 2.9787526132923125e-05}
{"train_loss": 0.08745722472667694, "global_step": 171867, "epoch": 1931, "lr": 2.978699587891453e-05}
{"train_loss": 0.07847412675619125, "global_step": 171868, "epoch": 1931, "lr": 2.978646562762335e-05}
{"train_loss": 0.08023319393396378, "global_step": 171869, "epoch": 1931, "lr": 2.9785935379049614e-05}
{"train_loss": 0.09829413890838623, "global_step": 171870, "epoch": 1931, "lr": 2.9785405133193422e-05}
{"train_loss": 0.13012966513633728, "global_step": 171871, "epoch": 1931, "lr": 2.9784874890054827e-05}
{"train_loss": 0.08993322402238846, "global_step": 171872, "epoch": 1931, "lr": 2.978434464963392e-05}
{"train_loss": 0.1855885535478592, "global_step": 171873, "epoch": 1931, "lr": 2.9783814411930756e-05}
{"train_loss": 0.06432660669088364, "global_step": 171874, "epoch": 1931, "lr": 2.978328417694542e-05}
{"train_loss": 0.11135730147361755, "global_step": 171875, "epoch": 1931, "lr": 2.9782753944677954e-05}
{"train_loss": 0.054848168045282364, "global_step": 171876, "epoch": 1931, "lr": 2.9782223715128478e-05}
{"train_loss": 0.03797940909862518, "global_step": 171877, "epoch": 1931, "lr": 2.9781693488297012e-05}
{"train_loss": 0.11138807982206345, "global_step": 171878, "epoch": 1931, "lr": 2.9781163264183675e-05}
{"train_loss": 0.10318415611982346, "global_step": 171879, "epoch": 1931, "lr": 2.9780633042788487e-05}
{"train_loss": 0.11999483406543732, "global_step": 171880, "epoch": 1931, "lr": 2.9780102824111557e-05}
{"train_loss": 0.0768362432718277, "global_step": 171881, "epoch": 1931, "lr": 2.977957260815295e-05}
{"train_loss": 0.07338552922010422, "global_step": 171882, "epoch": 1931, "lr": 2.9779042394912732e-05}
{"train_loss": 0.06883835047483444, "global_step": 171883, "epoch": 1931, "lr": 2.977851218439097e-05}
{"train_loss": 0.050277043133974075, "global_step": 171884, "epoch": 1931, "lr": 2.9777981976587748e-05}
{"train_loss": 0.09898817539215088, "global_step": 171885, "epoch": 1931, "lr": 2.9777451771503128e-05}
{"train_loss": 0.07528678327798843, "global_step": 171886, "epoch": 1931, "lr": 2.9776921569137186e-05}
{"train_loss": 0.0674053281545639, "global_step": 171887, "epoch": 1931, "lr": 2.9776391369489992e-05}
{"train_loss": 0.07442478835582733, "global_step": 171888, "epoch": 1931, "lr": 2.9775861172561605e-05}
{"train_loss": 0.07998353987932205, "global_step": 171889, "epoch": 1931, "lr": 2.9775330978352122e-05}
{"train_loss": 0.09629372507333755, "global_step": 171890, "epoch": 1931, "lr": 2.9774800786861578e-05}
{"train_loss": 0.08381223678588867, "global_step": 171891, "epoch": 1931, "lr": 2.977427059809008e-05}
{"train_loss": 0.0817394033074379, "global_step": 171892, "epoch": 1931, "lr": 2.9773740412037675e-05}
{"train_loss": 0.08896280825138092, "global_step": 171893, "epoch": 1931, "lr": 2.977321022870446e-05}
{"train_loss": 0.043008312582969666, "global_step": 171894, "epoch": 1931, "lr": 2.9772680048090474e-05}
{"train_loss": 0.060851871967315674, "global_step": 171895, "epoch": 1931, "lr": 2.977214987019582e-05}
{"train_loss": 0.09175878763198853, "global_step": 171896, "epoch": 1931, "lr": 2.9771619695020535e-05}
{"train_loss": 0.13850122690200806, "global_step": 171897, "epoch": 1931, "lr": 2.9771089522564732e-05}
{"train_loss": 0.11024009436368942, "global_step": 171898, "epoch": 1931, "lr": 2.9770559352828437e-05}
{"train_loss": 0.10042785108089447, "global_step": 171899, "epoch": 1931, "lr": 2.9770029185811745e-05}
{"train_loss": 0.06245790794491768, "global_step": 171900, "epoch": 1931, "lr": 2.9769499021514746e-05}
{"train_loss": 0.14364522695541382, "global_step": 171901, "epoch": 1931, "lr": 2.976896885993748e-05}
{"train_loss": 0.05672486871480942, "global_step": 171902, "epoch": 1931, "lr": 2.9768438701080024e-05}
{"train_loss": 0.11542859673500061, "global_step": 171903, "epoch": 1931, "lr": 2.976790854494247e-05}
{"train_loss": 0.06528384983539581, "global_step": 171904, "epoch": 1931, "lr": 2.9767378391524858e-05}
{"train_loss": 0.0362541526556015, "global_step": 171905, "epoch": 1931, "lr": 2.976684824082728e-05}
{"train_loss": 0.07476432621479034, "global_step": 171906, "epoch": 1931, "lr": 2.976631809284982e-05}
{"train_loss": 0.07880616933107376, "global_step": 171907, "epoch": 1931, "lr": 2.9765787947592506e-05}
{"train_loss": 0.08617538958787918, "global_step": 171908, "epoch": 1931, "lr": 2.9765257805055458e-05}
{"train_loss": 0.05736829340457916, "global_step": 171909, "epoch": 1931, "lr": 2.97647276652387e-05}
{"train_loss": 0.07826043665409088, "global_step": 171910, "epoch": 1931, "lr": 2.976419752814235e-05}
{"train_loss": 0.06879693269729614, "global_step": 171911, "epoch": 1931, "lr": 2.9763667393766438e-05}
{"train_loss": 0.09839959442615509, "global_step": 171912, "epoch": 1931, "lr": 2.9763137262111074e-05}
{"train_loss": 0.07070060819387436, "global_step": 171913, "epoch": 1931, "lr": 2.976260713317629e-05}
{"train_loss": 0.18784816563129425, "global_step": 171914, "epoch": 1931, "lr": 2.9762077006962195e-05}
{"train_loss": 0.10529656708240509, "global_step": 171915, "epoch": 1931, "lr": 2.9761546883468827e-05}
{"train_loss": 0.09337341785430908, "global_step": 171916, "epoch": 1931, "lr": 2.9761016762696264e-05}
{"train_loss": 0.09731954336166382, "global_step": 171917, "epoch": 1931, "lr": 2.9760486644644615e-05}
{"train_loss": 0.0843236893415451, "global_step": 171918, "epoch": 1931, "lr": 2.9759956529313893e-05}
{"train_loss": 0.03880613297224045, "global_step": 171919, "epoch": 1931, "lr": 2.975942641670422e-05}
{"train_loss": 0.06204390153288841, "global_step": 171920, "epoch": 1931, "lr": 2.975889630681564e-05}
{"train_loss": 0.10021210461854935, "global_step": 171921, "epoch": 1931, "lr": 2.9758366199648224e-05}
{"train_loss": 0.08699903637170792, "global_step": 171922, "epoch": 1931, "lr": 2.975783609520205e-05}
{"train_loss": 0.060889728367328644, "global_step": 171923, "epoch": 1931, "lr": 2.9757305993477193e-05}
{"train_loss": 0.05736033618450165, "global_step": 171924, "epoch": 1931, "lr": 2.9756775894473722e-05}
{"train_loss": 0.06609820574522018, "global_step": 171925, "epoch": 1931, "lr": 2.9756245798191707e-05}
{"train_loss": 0.07240349799394608, "global_step": 171926, "epoch": 1931, "lr": 2.9755715704631203e-05}
{"train_loss": 0.15183773636817932, "global_step": 171927, "epoch": 1931, "lr": 2.9755185613792313e-05}
{"train_loss": 0.06783171743154526, "global_step": 171928, "epoch": 1931, "lr": 2.9754655525675073e-05}
{"train_loss": 0.06814819574356079, "global_step": 171929, "epoch": 1931, "lr": 2.975412544027959e-05}
{"train_loss": 0.09695491194725037, "global_step": 171930, "epoch": 1931, "lr": 2.97535953576059e-05}
{"train_loss": 0.053662240505218506, "global_step": 171931, "epoch": 1931, "lr": 2.975306527765412e-05}
{"train_loss": 0.0850554034113884, "global_step": 171932, "epoch": 1931, "lr": 2.9752535200424264e-05}
{"train_loss": 0.06358510255813599, "global_step": 171933, "epoch": 1931, "lr": 2.975200512591646e-05}
{"train_loss": 0.10382427275180817, "global_step": 171934, "epoch": 1931, "lr": 2.9751475054130728e-05}
{"train_loss": 0.10362517833709717, "global_step": 171935, "epoch": 1931, "lr": 2.9750944985067165e-05}
{"train_loss": 0.0397234670817852, "global_step": 171936, "epoch": 1931, "lr": 2.9750414918725854e-05}
{"train_loss": 0.14317841827869415, "global_step": 171937, "epoch": 1931, "lr": 2.9749884855106836e-05}
{"train_loss": 0.048742763698101044, "global_step": 171938, "epoch": 1931, "lr": 2.9749354794210225e-05}
{"train_loss": 0.09514078497886658, "global_step": 171939, "epoch": 1931, "lr": 2.974882473603605e-05}
{"train_loss": 0.057559918612241745, "global_step": 171940, "epoch": 1931, "lr": 2.9748294680584398e-05}
{"train_loss": 0.05593413487076759, "global_step": 171941, "epoch": 1931, "lr": 2.9747764627855347e-05}
{"train_loss": 0.1279853731393814, "global_step": 171942, "epoch": 1931, "lr": 2.9747234577848958e-05}
{"train_loss": 0.06346966326236725, "global_step": 171943, "epoch": 1931, "lr": 2.9746704530565313e-05}
{"train_loss": 0.07990482449531555, "global_step": 171944, "epoch": 1931, "lr": 2.974617448600448e-05}
{"train_loss": 0.0979863852262497, "global_step": 171945, "epoch": 1931, "lr": 2.974564444416651e-05}
{"train_loss": 0.0807628482580185, "global_step": 171946, "epoch": 1931, "lr": 2.974511440505151e-05}
{"train_loss": 0.0851674998241864, "global_step": 171947, "epoch": 1931, "lr": 2.9744584368659512e-05, "val_loss": 7.145047664642334}
{"train_loss": 0.048849545419216156, "global_step": 171948, "epoch": 1932, "lr": 2.9744054334990624e-05}
{"train_loss": 0.03917425125837326, "global_step": 171949, "epoch": 1932, "lr": 2.974352430404489e-05}
{"train_loss": 0.05757607892155647, "global_step": 171950, "epoch": 1932, "lr": 2.9742994275822405e-05}
{"train_loss": 0.07433557510375977, "global_step": 171951, "epoch": 1932, "lr": 2.974246425032321e-05}
{"train_loss": 0.07284937053918839, "global_step": 171952, "epoch": 1932, "lr": 2.9741934227547395e-05}
{"train_loss": 0.06626841425895691, "global_step": 171953, "epoch": 1932, "lr": 2.974140420749505e-05}
{"train_loss": 0.04934662580490112, "global_step": 171954, "epoch": 1932, "lr": 2.9740874190166202e-05}
{"train_loss": 0.09931951761245728, "global_step": 171955, "epoch": 1932, "lr": 2.974034417556097e-05}
{"train_loss": 0.08073468506336212, "global_step": 171956, "epoch": 1932, "lr": 2.973981416367938e-05}
{"train_loss": 0.09939532727003098, "global_step": 171957, "epoch": 1932, "lr": 2.9739284154521552e-05}
{"train_loss": 0.10062503814697266, "global_step": 171958, "epoch": 1932, "lr": 2.973875414808751e-05}
{"train_loss": 0.10007976740598679, "global_step": 171959, "epoch": 1932, "lr": 2.9738224144377347e-05}
{"train_loss": 0.05698619410395622, "global_step": 171960, "epoch": 1932, "lr": 2.9737694143391136e-05}
{"train_loss": 0.08748267590999603, "global_step": 171961, "epoch": 1932, "lr": 2.973716414512895e-05}
{"train_loss": 0.10182677954435349, "global_step": 171962, "epoch": 1932, "lr": 2.973663414959085e-05}
{"train_loss": 0.0730065256357193, "global_step": 171963, "epoch": 1932, "lr": 2.9736104156776924e-05}
{"train_loss": 0.10929665714502335, "global_step": 171964, "epoch": 1932, "lr": 2.9735574166687206e-05}
{"train_loss": 0.06627925485372543, "global_step": 171965, "epoch": 1932, "lr": 2.9735044179321824e-05}
{"train_loss": 0.04187324270606041, "global_step": 171966, "epoch": 1932, "lr": 2.9734514194680792e-05}
{"train_loss": 0.04783537611365318, "global_step": 171967, "epoch": 1932, "lr": 2.973398421276422e-05}
{"train_loss": 0.062343426048755646, "global_step": 171968, "epoch": 1932, "lr": 2.9733454233572154e-05}
{"train_loss": 0.03211810812354088, "global_step": 171969, "epoch": 1932, "lr": 2.9732924257104705e-05}
{"train_loss": 0.10616261512041092, "global_step": 171970, "epoch": 1932, "lr": 2.9732394283361886e-05}
{"train_loss": 0.0758156105875969, "global_step": 171971, "epoch": 1932, "lr": 2.9731864312343803e-05}
{"train_loss": 0.04525509849190712, "global_step": 171972, "epoch": 1932, "lr": 2.9731334344050554e-05}
{"train_loss": 0.10361859202384949, "global_step": 171973, "epoch": 1932, "lr": 2.9730804378482145e-05}
{"train_loss": 0.11022692173719406, "global_step": 171974, "epoch": 1932, "lr": 2.9730274415638716e-05}
{"train_loss": 0.06571172177791595, "global_step": 171975, "epoch": 1932, "lr": 2.972974445552027e-05}
{"train_loss": 0.08245899528265, "global_step": 171976, "epoch": 1932, "lr": 2.9729214498126945e-05}
{"train_loss": 0.08242201060056686, "global_step": 171977, "epoch": 1932, "lr": 2.9728684543458763e-05}
{"train_loss": 0.06489389389753342, "global_step": 171978, "epoch": 1932, "lr": 2.972815459151581e-05}
{"train_loss": 0.09803923964500427, "global_step": 171979, "epoch": 1932, "lr": 2.9727624642298167e-05}
{"train_loss": 0.163829043507576, "global_step": 171980, "epoch": 1932, "lr": 2.9727094695805896e-05}
{"train_loss": 0.08416581898927689, "global_step": 171981, "epoch": 1932, "lr": 2.9726564752039065e-05}
{"train_loss": 0.058789849281311035, "global_step": 171982, "epoch": 1932, "lr": 2.9726034810997773e-05}
{"train_loss": 0.1334332972764969, "global_step": 171983, "epoch": 1932, "lr": 2.972550487268203e-05}
{"train_loss": 0.06396320462226868, "global_step": 171984, "epoch": 1932, "lr": 2.9724974937091977e-05}
{"train_loss": 0.10508143901824951, "global_step": 171985, "epoch": 1932, "lr": 2.9724445004227636e-05}
{"train_loss": 0.10289527475833893, "global_step": 171986, "epoch": 1932, "lr": 2.9723915074089114e-05}
{"train_loss": 0.10714922100305557, "global_step": 171987, "epoch": 1932, "lr": 2.972338514667644e-05}
{"train_loss": 0.10416519641876221, "global_step": 171988, "epoch": 1932, "lr": 2.972285522198973e-05}
{"train_loss": 0.07230032235383987, "global_step": 171989, "epoch": 1932, "lr": 2.9722325300029018e-05}
{"train_loss": 0.0861782506108284, "global_step": 171990, "epoch": 1932, "lr": 2.9721795380794392e-05}
{"train_loss": 0.1297486275434494, "global_step": 171991, "epoch": 1932, "lr": 2.9721265464285945e-05}
{"train_loss": 0.06205124035477638, "global_step": 171992, "epoch": 1932, "lr": 2.9720735550503708e-05}
{"train_loss": 0.05076803267002106, "global_step": 171993, "epoch": 1932, "lr": 2.9720205639447785e-05}
{"train_loss": 0.0740278959274292, "global_step": 171994, "epoch": 1932, "lr": 2.9719675731118214e-05}
{"train_loss": 0.09159128367900848, "global_step": 171995, "epoch": 1932, "lr": 2.971914582551511e-05}
{"train_loss": 0.06211135908961296, "global_step": 171996, "epoch": 1932, "lr": 2.9718615922638503e-05}
{"train_loss": 0.05134846270084381, "global_step": 171997, "epoch": 1932, "lr": 2.971808602248849e-05}
{"train_loss": 0.12423981726169586, "global_step": 171998, "epoch": 1932, "lr": 2.9717556125065127e-05}
{"train_loss": 0.07719457894563675, "global_step": 171999, "epoch": 1932, "lr": 2.9717026230368494e-05}
{"train_loss": 0.12345201522111893, "global_step": 172000, "epoch": 1932, "lr": 2.971649633839866e-05}
{"train_loss": 0.06369541585445404, "global_step": 172001, "epoch": 1932, "lr": 2.9715966449155692e-05}
{"train_loss": 0.11624941974878311, "global_step": 172002, "epoch": 1932, "lr": 2.9715436562639665e-05}
{"train_loss": 0.0969204530119896, "global_step": 172003, "epoch": 1932, "lr": 2.971490667885066e-05}
{"train_loss": 0.12550723552703857, "global_step": 172004, "epoch": 1932, "lr": 2.9714376797788724e-05}
{"train_loss": 0.06202516704797745, "global_step": 172005, "epoch": 1932, "lr": 2.971384691945396e-05}
{"train_loss": 0.046263132244348526, "global_step": 172006, "epoch": 1932, "lr": 2.9713317043846402e-05}
{"train_loss": 0.09368319064378738, "global_step": 172007, "epoch": 1932, "lr": 2.971278717096614e-05}
{"train_loss": 0.06401503831148148, "global_step": 172008, "epoch": 1932, "lr": 2.9712257300813274e-05}
{"train_loss": 0.05828361213207245, "global_step": 172009, "epoch": 1932, "lr": 2.9711727433387815e-05}
{"train_loss": 0.07052889466285706, "global_step": 172010, "epoch": 1932, "lr": 2.9711197568689897e-05}
{"train_loss": 0.07700137048959732, "global_step": 172011, "epoch": 1932, "lr": 2.9710667706719537e-05}
{"train_loss": 0.13947442173957825, "global_step": 172012, "epoch": 1932, "lr": 2.971013784747685e-05}
{"train_loss": 0.05271463096141815, "global_step": 172013, "epoch": 1932, "lr": 2.970960799096187e-05}
{"train_loss": 0.11631452292203903, "global_step": 172014, "epoch": 1932, "lr": 2.9709078137174705e-05}
{"train_loss": 0.09145928174257278, "global_step": 172015, "epoch": 1932, "lr": 2.9708548286115383e-05}
{"train_loss": 0.1193976178765297, "global_step": 172016, "epoch": 1932, "lr": 2.9708018437784025e-05}
{"train_loss": 0.06921016424894333, "global_step": 172017, "epoch": 1932, "lr": 2.9707488592180666e-05}
{"train_loss": 0.057188667356967926, "global_step": 172018, "epoch": 1932, "lr": 2.9706958749305386e-05}
{"train_loss": 0.12504640221595764, "global_step": 172019, "epoch": 1932, "lr": 2.9706428909158258e-05}
{"train_loss": 0.11463405191898346, "global_step": 172020, "epoch": 1932, "lr": 2.970589907173935e-05}
{"train_loss": 0.0974004790186882, "global_step": 172021, "epoch": 1932, "lr": 2.970536923704874e-05}
{"train_loss": 0.122600257396698, "global_step": 172022, "epoch": 1932, "lr": 2.970483940508651e-05}
{"train_loss": 0.09679564088582993, "global_step": 172023, "epoch": 1932, "lr": 2.9704309575852697e-05}
{"train_loss": 0.08280378580093384, "global_step": 172024, "epoch": 1932, "lr": 2.9703779749347403e-05}
{"train_loss": 0.1372518390417099, "global_step": 172025, "epoch": 1932, "lr": 2.9703249925570677e-05}
{"train_loss": 0.08302243798971176, "global_step": 172026, "epoch": 1932, "lr": 2.9702720104522596e-05}
{"train_loss": 0.048913512378931046, "global_step": 172027, "epoch": 1932, "lr": 2.970219028620326e-05}
{"train_loss": 0.049288567155599594, "global_step": 172028, "epoch": 1932, "lr": 2.9701660470612693e-05}
{"train_loss": 0.08339951932430267, "global_step": 172029, "epoch": 1932, "lr": 2.970113065775102e-05}
{"train_loss": 0.12195338308811188, "global_step": 172030, "epoch": 1932, "lr": 2.9700600847618255e-05}
{"train_loss": 0.09436605870723724, "global_step": 172031, "epoch": 1932, "lr": 2.9700071040214517e-05}
{"train_loss": 0.06757204234600067, "global_step": 172032, "epoch": 1932, "lr": 2.9699541235539837e-05}
{"train_loss": 0.10071056336164474, "global_step": 172033, "epoch": 1932, "lr": 2.9699011433594327e-05}
{"train_loss": 0.119674913585186, "global_step": 172034, "epoch": 1932, "lr": 2.9698481634378016e-05}
{"train_loss": 0.03942456468939781, "global_step": 172035, "epoch": 1932, "lr": 2.969795183789102e-05}
{"train_loss": 0.08471207287204399, "global_step": 172036, "epoch": 1932, "lr": 2.9697422044133373e-05, "val_loss": 7.106564998626709}
{"train_loss": 0.07853066176176071, "global_step": 172037, "epoch": 1933, "lr": 2.969689225310517e-05}
{"train_loss": 0.0965278297662735, "global_step": 172038, "epoch": 1933, "lr": 2.969636246480646e-05}
{"train_loss": 0.13017421960830688, "global_step": 172039, "epoch": 1933, "lr": 2.9695832679237335e-05}
{"train_loss": 0.06767210364341736, "global_step": 172040, "epoch": 1933, "lr": 2.9695302896397847e-05}
{"train_loss": 0.09325636923313141, "global_step": 172041, "epoch": 1933, "lr": 2.9694773116288098e-05}
{"train_loss": 0.13879603147506714, "global_step": 172042, "epoch": 1933, "lr": 2.9694243338908116e-05}
{"train_loss": 0.08351638168096542, "global_step": 172043, "epoch": 1933, "lr": 2.9693713564257997e-05}
{"train_loss": 0.10287311673164368, "global_step": 172044, "epoch": 1933, "lr": 2.969318379233783e-05}
{"train_loss": 0.09072674065828323, "global_step": 172045, "epoch": 1933, "lr": 2.9692654023147647e-05}
{"train_loss": 0.09730660915374756, "global_step": 172046, "epoch": 1933, "lr": 2.969212425668756e-05}
{"train_loss": 0.09862088412046432, "global_step": 172047, "epoch": 1933, "lr": 2.9691594492957596e-05}
{"train_loss": 0.17159996926784515, "global_step": 172048, "epoch": 1933, "lr": 2.9691064731957873e-05}
{"train_loss": 0.10236632078886032, "global_step": 172049, "epoch": 1933, "lr": 2.969053497368842e-05}
{"train_loss": 0.06769711524248123, "global_step": 172050, "epoch": 1933, "lr": 2.969000521814934e-05}
{"train_loss": 0.12567511200904846, "global_step": 172051, "epoch": 1933, "lr": 2.9689475465340676e-05}
{"train_loss": 0.07776740193367004, "global_step": 172052, "epoch": 1933, "lr": 2.9688945715262538e-05}
{"train_loss": 0.13522468507289886, "global_step": 172053, "epoch": 1933, "lr": 2.968841596791495e-05}
{"train_loss": 0.1891283392906189, "global_step": 172054, "epoch": 1933, "lr": 2.9687886223298027e-05}
{"train_loss": 0.05820150300860405, "global_step": 172055, "epoch": 1933, "lr": 2.968735648141181e-05}
{"train_loss": 0.08214929699897766, "global_step": 172056, "epoch": 1933, "lr": 2.968682674225638e-05}
{"train_loss": 0.11270948499441147, "global_step": 172057, "epoch": 1933, "lr": 2.9686297005831807e-05}
{"train_loss": 0.03152734041213989, "global_step": 172058, "epoch": 1933, "lr": 2.968576727213817e-05}
{"train_loss": 0.0644160807132721, "global_step": 172059, "epoch": 1933, "lr": 2.9685237541175533e-05}
{"train_loss": 0.0998556837439537, "global_step": 172060, "epoch": 1933, "lr": 2.9684707812943974e-05}
{"train_loss": 0.09556200355291367, "global_step": 172061, "epoch": 1933, "lr": 2.9684178087443537e-05}
{"train_loss": 0.06803464889526367, "global_step": 172062, "epoch": 1933, "lr": 2.968364836467432e-05}
{"train_loss": 0.12414689362049103, "global_step": 172063, "epoch": 1933, "lr": 2.9683118644636403e-05}
{"train_loss": 0.10706336051225662, "global_step": 172064, "epoch": 1933, "lr": 2.9682588927329824e-05}
{"train_loss": 0.10321778804063797, "global_step": 172065, "epoch": 1933, "lr": 2.96820592127547e-05}
{"train_loss": 0.08244708925485611, "global_step": 172066, "epoch": 1933, "lr": 2.9681529500911053e-05}
{"train_loss": 0.0841309130191803, "global_step": 172067, "epoch": 1933, "lr": 2.9680999791798992e-05}
{"train_loss": 0.09703733026981354, "global_step": 172068, "epoch": 1933, "lr": 2.9680470085418556e-05}
{"train_loss": 0.07982613146305084, "global_step": 172069, "epoch": 1933, "lr": 2.9679940381769855e-05}
{"train_loss": 0.07748210430145264, "global_step": 172070, "epoch": 1933, "lr": 2.9679410680852916e-05}
{"train_loss": 0.06157226860523224, "global_step": 172071, "epoch": 1933, "lr": 2.9678880982667852e-05}
{"train_loss": 0.1588970124721527, "global_step": 172072, "epoch": 1933, "lr": 2.967835128721469e-05}
{"train_loss": 0.08747061342000961, "global_step": 172073, "epoch": 1933, "lr": 2.9677821594493553e-05}
{"train_loss": 0.061203401535749435, "global_step": 172074, "epoch": 1933, "lr": 2.9677291904504478e-05}
{"train_loss": 0.12483160942792892, "global_step": 172075, "epoch": 1933, "lr": 2.9676762217247535e-05}
{"train_loss": 0.10397206246852875, "global_step": 172076, "epoch": 1933, "lr": 2.967623253272281e-05}
{"train_loss": 0.060167815536260605, "global_step": 172077, "epoch": 1933, "lr": 2.9675702850930366e-05}
{"train_loss": 0.07747960090637207, "global_step": 172078, "epoch": 1933, "lr": 2.9675173171870276e-05}
{"train_loss": 0.11030144989490509, "global_step": 172079, "epoch": 1933, "lr": 2.96746434955426e-05}
{"train_loss": 0.06301820278167725, "global_step": 172080, "epoch": 1933, "lr": 2.967411382194744e-05}
{"train_loss": 0.08039409667253494, "global_step": 172081, "epoch": 1933, "lr": 2.9673584151084833e-05}
{"train_loss": 0.07924800366163254, "global_step": 172082, "epoch": 1933, "lr": 2.9673054482954876e-05}
{"train_loss": 0.06305941194295883, "global_step": 172083, "epoch": 1933, "lr": 2.9672524817557613e-05}
{"train_loss": 0.09633629769086838, "global_step": 172084, "epoch": 1933, "lr": 2.967199515489315e-05}
{"train_loss": 0.062326960265636444, "global_step": 172085, "epoch": 1933, "lr": 2.9671465494961527e-05}
{"train_loss": 0.0402211956679821, "global_step": 172086, "epoch": 1933, "lr": 2.9670935837762836e-05}
{"train_loss": 0.0927296057343483, "global_step": 172087, "epoch": 1933, "lr": 2.967040618329713e-05}
{"train_loss": 0.05266239121556282, "global_step": 172088, "epoch": 1933, "lr": 2.9669876531564504e-05}
{"train_loss": 0.09930764138698578, "global_step": 172089, "epoch": 1933, "lr": 2.9669346882564997e-05}
{"train_loss": 0.1354888677597046, "global_step": 172090, "epoch": 1933, "lr": 2.966881723629872e-05}
{"train_loss": 0.07663058489561081, "global_step": 172091, "epoch": 1933, "lr": 2.9668287592765704e-05}
{"train_loss": 0.11582887172698975, "global_step": 172092, "epoch": 1933, "lr": 2.966775795196606e-05}
{"train_loss": 0.18560287356376648, "global_step": 172093, "epoch": 1933, "lr": 2.966722831389982e-05}
{"train_loss": 0.09330210089683533, "global_step": 172094, "epoch": 1933, "lr": 2.966669867856708e-05}
{"train_loss": 0.06433815509080887, "global_step": 172095, "epoch": 1933, "lr": 2.96661690459679e-05}
{"train_loss": 0.10233864933252335, "global_step": 172096, "epoch": 1933, "lr": 2.9665639416102363e-05}
{"train_loss": 0.08662266284227371, "global_step": 172097, "epoch": 1933, "lr": 2.9665109788970526e-05}
{"train_loss": 0.0628073439002037, "global_step": 172098, "epoch": 1933, "lr": 2.966458016457246e-05}
{"train_loss": 0.06637657433748245, "global_step": 172099, "epoch": 1933, "lr": 2.966405054290826e-05}
{"train_loss": 0.1285693496465683, "global_step": 172100, "epoch": 1933, "lr": 2.9663520923977965e-05}
{"train_loss": 0.12283927947282791, "global_step": 172101, "epoch": 1933, "lr": 2.9662991307781686e-05}
{"train_loss": 0.08159170299768448, "global_step": 172102, "epoch": 1933, "lr": 2.966246169431944e-05}
{"train_loss": 0.05598695948719978, "global_step": 172103, "epoch": 1933, "lr": 2.966193208359135e-05}
{"train_loss": 0.0966980829834938, "global_step": 172104, "epoch": 1933, "lr": 2.966140247559745e-05}
{"train_loss": 0.06383999437093735, "global_step": 172105, "epoch": 1933, "lr": 2.9660872870337842e-05}
{"train_loss": 0.04710070788860321, "global_step": 172106, "epoch": 1933, "lr": 2.9660343267812562e-05}
{"train_loss": 0.05823361873626709, "global_step": 172107, "epoch": 1933, "lr": 2.965981366802173e-05}
{"train_loss": 0.081581249833107, "global_step": 172108, "epoch": 1933, "lr": 2.9659284070965354e-05}
{"train_loss": 0.06006253510713577, "global_step": 172109, "epoch": 1933, "lr": 2.9658754476643568e-05}
{"train_loss": 0.09452375769615173, "global_step": 172110, "epoch": 1933, "lr": 2.9658224885056398e-05}
{"train_loss": 0.09148764610290527, "global_step": 172111, "epoch": 1933, "lr": 2.965769529620394e-05}
{"train_loss": 0.07305264472961426, "global_step": 172112, "epoch": 1933, "lr": 2.9657165710086245e-05}
{"train_loss": 0.12422595918178558, "global_step": 172113, "epoch": 1933, "lr": 2.965663612670342e-05}
{"train_loss": 0.09432390332221985, "global_step": 172114, "epoch": 1933, "lr": 2.9656106546055495e-05}
{"train_loss": 0.08580514043569565, "global_step": 172115, "epoch": 1933, "lr": 2.9655576968142557e-05}
{"train_loss": 0.05358903482556343, "global_step": 172116, "epoch": 1933, "lr": 2.9655047392964685e-05}
{"train_loss": 0.10928421467542648, "global_step": 172117, "epoch": 1933, "lr": 2.9654517820521933e-05}
{"train_loss": 0.059888120740652084, "global_step": 172118, "epoch": 1933, "lr": 2.9653988250814396e-05}
{"train_loss": 0.09845831990242004, "global_step": 172119, "epoch": 1933, "lr": 2.9653458683842118e-05}
{"train_loss": 0.0845985934138298, "global_step": 172120, "epoch": 1933, "lr": 2.9652929119605204e-05}
{"train_loss": 0.06534598767757416, "global_step": 172121, "epoch": 1933, "lr": 2.965239955810368e-05}
{"train_loss": 0.04392453283071518, "global_step": 172122, "epoch": 1933, "lr": 2.965186999933768e-05}
{"train_loss": 0.06623449176549911, "global_step": 172123, "epoch": 1933, "lr": 2.9651340443307203e-05}
{"train_loss": 0.10632839798927307, "global_step": 172124, "epoch": 1933, "lr": 2.9650810890012376e-05}
{"train_loss": 0.09000650843542614, "global_step": 172125, "epoch": 1933, "lr": 2.9650281339453235e-05, "val_loss": 7.17905855178833}
{"train_loss": 0.05768837407231331, "global_step": 172126, "epoch": 1934, "lr": 2.9649751791629882e-05}
{"train_loss": 0.09048577398061752, "global_step": 172127, "epoch": 1934, "lr": 2.964922224654235e-05}
{"train_loss": 0.10310973227024078, "global_step": 172128, "epoch": 1934, "lr": 2.9648692704190755e-05}
{"train_loss": 0.09033184498548508, "global_step": 172129, "epoch": 1934, "lr": 2.9648163164575128e-05}
{"train_loss": 0.07434666901826859, "global_step": 172130, "epoch": 1934, "lr": 2.964763362769557e-05}
{"train_loss": 0.06734856963157654, "global_step": 172131, "epoch": 1934, "lr": 2.9647104093552126e-05}
{"train_loss": 0.067677803337574, "global_step": 172132, "epoch": 1934, "lr": 2.9646574562144902e-05}
{"train_loss": 0.059755951166152954, "global_step": 172133, "epoch": 1934, "lr": 2.9646045033473936e-05}
{"train_loss": 0.14949539303779602, "global_step": 172134, "epoch": 1934, "lr": 2.96455155075393e-05}
{"train_loss": 0.03670162707567215, "global_step": 172135, "epoch": 1934, "lr": 2.9644985984341095e-05}
{"train_loss": 0.0946369394659996, "global_step": 172136, "epoch": 1934, "lr": 2.9644456463879356e-05}
{"train_loss": 0.11534901708364487, "global_step": 172137, "epoch": 1934, "lr": 2.9643926946154186e-05}
{"train_loss": 0.1277843713760376, "global_step": 172138, "epoch": 1934, "lr": 2.9643397431165632e-05}
{"train_loss": 0.14052164554595947, "global_step": 172139, "epoch": 1934, "lr": 2.9642867918913786e-05}
{"train_loss": 0.06335684657096863, "global_step": 172140, "epoch": 1934, "lr": 2.9642338409398694e-05}
{"train_loss": 0.04531489312648773, "global_step": 172141, "epoch": 1934, "lr": 2.964180890262046e-05}
{"train_loss": 0.06015341356396675, "global_step": 172142, "epoch": 1934, "lr": 2.9641279398579115e-05}
{"train_loss": 0.1108483076095581, "global_step": 172143, "epoch": 1934, "lr": 2.964074989727477e-05}
{"train_loss": 0.07409478724002838, "global_step": 172144, "epoch": 1934, "lr": 2.9640220398707464e-05}
{"train_loss": 0.09585124254226685, "global_step": 172145, "epoch": 1934, "lr": 2.9639690902877297e-05}
{"train_loss": 0.10771416127681732, "global_step": 172146, "epoch": 1934, "lr": 2.963916140978431e-05}
{"train_loss": 0.10916052758693695, "global_step": 172147, "epoch": 1934, "lr": 2.9638631919428605e-05}
{"train_loss": 0.060492128133773804, "global_step": 172148, "epoch": 1934, "lr": 2.9638102431810223e-05}
{"train_loss": 0.08495286107063293, "global_step": 172149, "epoch": 1934, "lr": 2.9637572946929264e-05}
{"train_loss": 0.10968717187643051, "global_step": 172150, "epoch": 1934, "lr": 2.9637043464785767e-05}
{"train_loss": 0.0739218071103096, "global_step": 172151, "epoch": 1934, "lr": 2.9636513985379844e-05}
{"train_loss": 0.07860119640827179, "global_step": 172152, "epoch": 1934, "lr": 2.9635984508711534e-05}
{"train_loss": 0.09643162786960602, "global_step": 172153, "epoch": 1934, "lr": 2.9635455034780902e-05}
{"train_loss": 0.10006319731473923, "global_step": 172154, "epoch": 1934, "lr": 2.9634925563588057e-05}
{"train_loss": 0.08572158217430115, "global_step": 172155, "epoch": 1934, "lr": 2.9634396095133033e-05}
{"train_loss": 0.07612846791744232, "global_step": 172156, "epoch": 1934, "lr": 2.963386662941593e-05}
{"train_loss": 0.06087256968021393, "global_step": 172157, "epoch": 1934, "lr": 2.963333716643679e-05}
{"train_loss": 0.09054812788963318, "global_step": 172158, "epoch": 1934, "lr": 2.963280770619572e-05}
{"train_loss": 0.06657585501670837, "global_step": 172159, "epoch": 1934, "lr": 2.9632278248692747e-05}
{"train_loss": 0.07053225487470627, "global_step": 172160, "epoch": 1934, "lr": 2.9631748793927984e-05}
{"train_loss": 0.05792297422885895, "global_step": 172161, "epoch": 1934, "lr": 2.9631219341901468e-05}
{"train_loss": 0.0636925920844078, "global_step": 172162, "epoch": 1934, "lr": 2.9630689892613305e-05}
{"train_loss": 0.06631731986999512, "global_step": 172163, "epoch": 1934, "lr": 2.9630160446063525e-05}
{"train_loss": 0.07130099833011627, "global_step": 172164, "epoch": 1934, "lr": 2.9629631002252244e-05}
{"train_loss": 0.05959726497530937, "global_step": 172165, "epoch": 1934, "lr": 2.9629101561179494e-05}
{"train_loss": 0.08567138016223907, "global_step": 172166, "epoch": 1934, "lr": 2.962857212284538e-05}
{"train_loss": 0.13685451447963715, "global_step": 172167, "epoch": 1934, "lr": 2.962804268724994e-05}
{"train_loss": 0.05801994726061821, "global_step": 172168, "epoch": 1934, "lr": 2.962751325439328e-05}
{"train_loss": 0.06908705085515976, "global_step": 172169, "epoch": 1934, "lr": 2.9626983824275428e-05}
{"train_loss": 0.07604913413524628, "global_step": 172170, "epoch": 1934, "lr": 2.9626454396896497e-05}
{"train_loss": 0.1668170988559723, "global_step": 172171, "epoch": 1934, "lr": 2.9625924972256546e-05}
{"train_loss": 0.03352730721235275, "global_step": 172172, "epoch": 1934, "lr": 2.9625395550355618e-05}
{"train_loss": 0.07976411283016205, "global_step": 172173, "epoch": 1934, "lr": 2.9624866131193828e-05}
{"train_loss": 0.08155152946710587, "global_step": 172174, "epoch": 1934, "lr": 2.962433671477121e-05}
{"train_loss": 0.11150384694337845, "global_step": 172175, "epoch": 1934, "lr": 2.9623807301087868e-05}
{"train_loss": 0.10011713951826096, "global_step": 172176, "epoch": 1934, "lr": 2.962327789014384e-05}
{"train_loss": 0.05833731219172478, "global_step": 172177, "epoch": 1934, "lr": 2.9622748481939238e-05}
{"train_loss": 0.08305540680885315, "global_step": 172178, "epoch": 1934, "lr": 2.9622219076474083e-05}
{"train_loss": 0.054739564657211304, "global_step": 172179, "epoch": 1934, "lr": 2.96216896737485e-05}
{"train_loss": 0.10207797586917877, "global_step": 172180, "epoch": 1934, "lr": 2.9621160273762506e-05}
{"train_loss": 0.064936563372612, "global_step": 172181, "epoch": 1934, "lr": 2.9620630876516214e-05}
{"train_loss": 0.0972011461853981, "global_step": 172182, "epoch": 1934, "lr": 2.9620101482009666e-05}
{"train_loss": 0.0767422765493393, "global_step": 172183, "epoch": 1934, "lr": 2.9619572090242965e-05}
{"train_loss": 0.06394943594932556, "global_step": 172184, "epoch": 1934, "lr": 2.9619042701216144e-05}
{"train_loss": 0.10072343051433563, "global_step": 172185, "epoch": 1934, "lr": 2.9618513314929313e-05}
{"train_loss": 0.11240259557962418, "global_step": 172186, "epoch": 1934, "lr": 2.9617983931382508e-05}
{"train_loss": 0.03870426118373871, "global_step": 172187, "epoch": 1934, "lr": 2.9617454550575834e-05}
{"train_loss": 0.10347849875688553, "global_step": 172188, "epoch": 1934, "lr": 2.9616925172509325e-05}
{"train_loss": 0.059792596846818924, "global_step": 172189, "epoch": 1934, "lr": 2.9616395797183083e-05}
{"train_loss": 0.061620984226465225, "global_step": 172190, "epoch": 1934, "lr": 2.9615866424597182e-05}
{"train_loss": 0.10472190380096436, "global_step": 172191, "epoch": 1934, "lr": 2.9615337054751653e-05}
{"train_loss": 0.08894090354442596, "global_step": 172192, "epoch": 1934, "lr": 2.9614807687646618e-05}
{"train_loss": 0.052062734961509705, "global_step": 172193, "epoch": 1934, "lr": 2.9614278323282096e-05}
{"train_loss": 0.10502257198095322, "global_step": 172194, "epoch": 1934, "lr": 2.9613748961658215e-05}
{"train_loss": 0.08108580112457275, "global_step": 172195, "epoch": 1934, "lr": 2.961321960277499e-05}
{"train_loss": 0.10803235322237015, "global_step": 172196, "epoch": 1934, "lr": 2.9612690246632546e-05}
{"train_loss": 0.06345657259225845, "global_step": 172197, "epoch": 1934, "lr": 2.96121608932309e-05}
{"train_loss": 0.10096930712461472, "global_step": 172198, "epoch": 1934, "lr": 2.9611631542570174e-05}
{"train_loss": 0.0629974976181984, "global_step": 172199, "epoch": 1934, "lr": 2.9611102194650398e-05}
{"train_loss": 0.06574897468090057, "global_step": 172200, "epoch": 1934, "lr": 2.9610572849471678e-05}
{"train_loss": 0.03339241445064545, "global_step": 172201, "epoch": 1934, "lr": 2.9610043507034048e-05}
{"train_loss": 0.07866225391626358, "global_step": 172202, "epoch": 1934, "lr": 2.960951416733762e-05}
{"train_loss": 0.04624800384044647, "global_step": 172203, "epoch": 1934, "lr": 2.9608984830382424e-05}
{"train_loss": 0.09099840372800827, "global_step": 172204, "epoch": 1934, "lr": 2.9608455496168573e-05}
{"train_loss": 0.08280232548713684, "global_step": 172205, "epoch": 1934, "lr": 2.96079261646961e-05}
{"train_loss": 0.103915274143219, "global_step": 172206, "epoch": 1934, "lr": 2.9607396835965096e-05}
{"train_loss": 0.1381150782108307, "global_step": 172207, "epoch": 1934, "lr": 2.9606867509975626e-05}
{"train_loss": 0.048812348395586014, "global_step": 172208, "epoch": 1934, "lr": 2.9606338186727768e-05}
{"train_loss": 0.09046120196580887, "global_step": 172209, "epoch": 1934, "lr": 2.96058088662216e-05}
{"train_loss": 0.05119937285780907, "global_step": 172210, "epoch": 1934, "lr": 2.9605279548457166e-05}
{"train_loss": 0.13015608489513397, "global_step": 172211, "epoch": 1934, "lr": 2.960475023343457e-05}
{"train_loss": 0.06449569761753082, "global_step": 172212, "epoch": 1934, "lr": 2.9604220921153848e-05}
{"train_loss": 0.05216022953391075, "global_step": 172213, "epoch": 1934, "lr": 2.960369161161511e-05}
{"train_loss": 0.08202412071522702, "global_step": 172214, "epoch": 1934, "lr": 2.9603162304818383e-05, "val_loss": 7.457366943359375}
{"train_loss": 0.07989145070314407, "global_step": 172215, "epoch": 1935, "lr": 2.9602633000763787e-05}
{"train_loss": 0.14534400403499603, "global_step": 172216, "epoch": 1935, "lr": 2.9602103699451345e-05}
{"train_loss": 0.08485285937786102, "global_step": 172217, "epoch": 1935, "lr": 2.9601574400881172e-05}
{"train_loss": 0.08330842107534409, "global_step": 172218, "epoch": 1935, "lr": 2.9601045105053305e-05}
{"train_loss": 0.07718265801668167, "global_step": 172219, "epoch": 1935, "lr": 2.960051581196784e-05}
{"train_loss": 0.08050704002380371, "global_step": 172220, "epoch": 1935, "lr": 2.9599986521624823e-05}
{"train_loss": 0.05371009558439255, "global_step": 172221, "epoch": 1935, "lr": 2.9599457234024354e-05}
{"train_loss": 0.07120361179113388, "global_step": 172222, "epoch": 1935, "lr": 2.9598927949166472e-05}
{"train_loss": 0.1467980146408081, "global_step": 172223, "epoch": 1935, "lr": 2.9598398667051286e-05}
{"train_loss": 0.15004302561283112, "global_step": 172224, "epoch": 1935, "lr": 2.9597869387678823e-05}
{"train_loss": 0.13519282639026642, "global_step": 172225, "epoch": 1935, "lr": 2.9597340111049192e-05}
{"train_loss": 0.08774770796298981, "global_step": 172226, "epoch": 1935, "lr": 2.9596810837162447e-05}
{"train_loss": 0.04404689371585846, "global_step": 172227, "epoch": 1935, "lr": 2.959628156601867e-05}
{"train_loss": 0.053242310881614685, "global_step": 172228, "epoch": 1935, "lr": 2.959575229761792e-05}
{"train_loss": 0.09135103970766068, "global_step": 172229, "epoch": 1935, "lr": 2.9595223031960268e-05}
{"train_loss": 0.10836819559335709, "global_step": 172230, "epoch": 1935, "lr": 2.95946937690458e-05}
{"train_loss": 0.1247510090470314, "global_step": 172231, "epoch": 1935, "lr": 2.9594164508874562e-05}
{"train_loss": 0.13022513687610626, "global_step": 172232, "epoch": 1935, "lr": 2.9593635251446657e-05}
{"train_loss": 0.07522828131914139, "global_step": 172233, "epoch": 1935, "lr": 2.9593105996762118e-05}
{"train_loss": 0.0466434620320797, "global_step": 172234, "epoch": 1935, "lr": 2.9592576744821054e-05}
{"train_loss": 0.14866192638874054, "global_step": 172235, "epoch": 1935, "lr": 2.959204749562351e-05}
{"train_loss": 0.11553735285997391, "global_step": 172236, "epoch": 1935, "lr": 2.959151824916958e-05}
{"train_loss": 0.07684501260519028, "global_step": 172237, "epoch": 1935, "lr": 2.9590989005459302e-05}
{"train_loss": 0.08393507450819016, "global_step": 172238, "epoch": 1935, "lr": 2.9590459764492783e-05}
{"train_loss": 0.12190759927034378, "global_step": 172239, "epoch": 1935, "lr": 2.958993052627006e-05}
{"train_loss": 0.043009065091609955, "global_step": 172240, "epoch": 1935, "lr": 2.958940129079124e-05}
{"train_loss": 0.05796743184328079, "global_step": 172241, "epoch": 1935, "lr": 2.958887205805636e-05}
{"train_loss": 0.07051245868206024, "global_step": 172242, "epoch": 1935, "lr": 2.958834282806553e-05}
{"train_loss": 0.055274300277233124, "global_step": 172243, "epoch": 1935, "lr": 2.9587813600818776e-05}
{"train_loss": 0.07874888181686401, "global_step": 172244, "epoch": 1935, "lr": 2.9587284376316203e-05}
{"train_loss": 0.13716495037078857, "global_step": 172245, "epoch": 1935, "lr": 2.9586755154557864e-05}
{"train_loss": 0.08335878700017929, "global_step": 172246, "epoch": 1935, "lr": 2.9586225935543845e-05}
{"train_loss": 0.12841682136058807, "global_step": 172247, "epoch": 1935, "lr": 2.95856967192742e-05}
{"train_loss": 0.12017710506916046, "global_step": 172248, "epoch": 1935, "lr": 2.9585167505749023e-05}
{"train_loss": 0.09307334572076797, "global_step": 172249, "epoch": 1935, "lr": 2.958463829496837e-05}
{"train_loss": 0.10537773370742798, "global_step": 172250, "epoch": 1935, "lr": 2.9584109086932297e-05}
{"train_loss": 0.17172597348690033, "global_step": 172251, "epoch": 1935, "lr": 2.958357988164091e-05}
{"train_loss": 0.07547841966152191, "global_step": 172252, "epoch": 1935, "lr": 2.9583050679094247e-05}
{"train_loss": 0.09318641573190689, "global_step": 172253, "epoch": 1935, "lr": 2.958252147929241e-05}
{"train_loss": 0.0646686851978302, "global_step": 172254, "epoch": 1935, "lr": 2.9581992282235437e-05}
{"train_loss": 0.10501300543546677, "global_step": 172255, "epoch": 1935, "lr": 2.9581463087923433e-05}
{"train_loss": 0.09663832932710648, "global_step": 172256, "epoch": 1935, "lr": 2.958093389635643e-05}
{"train_loss": 0.11664985120296478, "global_step": 172257, "epoch": 1935, "lr": 2.958040470753455e-05}
{"train_loss": 0.07222505658864975, "global_step": 172258, "epoch": 1935, "lr": 2.9579875521457812e-05}
{"train_loss": 0.15195930004119873, "global_step": 172259, "epoch": 1935, "lr": 2.957934633812633e-05}
{"train_loss": 0.05216594785451889, "global_step": 172260, "epoch": 1935, "lr": 2.9578817157540134e-05}
{"train_loss": 0.14866259694099426, "global_step": 172261, "epoch": 1935, "lr": 2.9578287979699315e-05}
{"train_loss": 0.1061878651380539, "global_step": 172262, "epoch": 1935, "lr": 2.957775880460397e-05}
{"train_loss": 0.0777590274810791, "global_step": 172263, "epoch": 1935, "lr": 2.9577229632254145e-05}
{"train_loss": 0.10518774390220642, "global_step": 172264, "epoch": 1935, "lr": 2.9576700462649898e-05}
{"train_loss": 0.0596502423286438, "global_step": 172265, "epoch": 1935, "lr": 2.9576171295791322e-05}
{"train_loss": 0.09314649552106857, "global_step": 172266, "epoch": 1935, "lr": 2.9575642131678482e-05}
{"train_loss": 0.04922904074192047, "global_step": 172267, "epoch": 1935, "lr": 2.9575112970311447e-05}
{"train_loss": 0.10752788186073303, "global_step": 172268, "epoch": 1935, "lr": 2.9574583811690294e-05}
{"train_loss": 0.047699544578790665, "global_step": 172269, "epoch": 1935, "lr": 2.9574054655815076e-05}
{"train_loss": 0.13277879357337952, "global_step": 172270, "epoch": 1935, "lr": 2.9573525502685894e-05}
{"train_loss": 0.10643742978572845, "global_step": 172271, "epoch": 1935, "lr": 2.9572996352302784e-05}
{"train_loss": 0.12308071553707123, "global_step": 172272, "epoch": 1935, "lr": 2.9572467204665855e-05}
{"train_loss": 0.12612847983837128, "global_step": 172273, "epoch": 1935, "lr": 2.9571938059775133e-05}
{"train_loss": 0.05640806257724762, "global_step": 172274, "epoch": 1935, "lr": 2.957140891763074e-05}
{"train_loss": 0.09922700375318527, "global_step": 172275, "epoch": 1935, "lr": 2.9570879778232707e-05}
{"train_loss": 0.049432218074798584, "global_step": 172276, "epoch": 1935, "lr": 2.9570350641581136e-05}
{"train_loss": 0.08777101337909698, "global_step": 172277, "epoch": 1935, "lr": 2.956982150767606e-05}
{"train_loss": 0.10654740035533905, "global_step": 172278, "epoch": 1935, "lr": 2.9569292376517593e-05}
{"train_loss": 0.13379669189453125, "global_step": 172279, "epoch": 1935, "lr": 2.956876324810577e-05}
{"train_loss": 0.15024691820144653, "global_step": 172280, "epoch": 1935, "lr": 2.9568234122440675e-05}
{"train_loss": 0.057423967868089676, "global_step": 172281, "epoch": 1935, "lr": 2.95677049995224e-05}
{"train_loss": 0.10629185289144516, "global_step": 172282, "epoch": 1935, "lr": 2.956717587935099e-05}
{"train_loss": 0.1172112226486206, "global_step": 172283, "epoch": 1935, "lr": 2.956664676192652e-05}
{"train_loss": 0.05460028350353241, "global_step": 172284, "epoch": 1935, "lr": 2.956611764724907e-05}
{"train_loss": 0.13737307488918304, "global_step": 172285, "epoch": 1935, "lr": 2.956558853531871e-05}
{"train_loss": 0.09333394467830658, "global_step": 172286, "epoch": 1935, "lr": 2.95650594261355e-05}
{"train_loss": 0.10459768772125244, "global_step": 172287, "epoch": 1935, "lr": 2.9564530319699535e-05}
{"train_loss": 0.06201021373271942, "global_step": 172288, "epoch": 1935, "lr": 2.9564001216010843e-05}
{"train_loss": 0.09977389872074127, "global_step": 172289, "epoch": 1935, "lr": 2.9563472115069547e-05}
{"train_loss": 0.07980316132307053, "global_step": 172290, "epoch": 1935, "lr": 2.956294301687567e-05}
{"train_loss": 0.07257799059152603, "global_step": 172291, "epoch": 1935, "lr": 2.9562413921429334e-05}
{"train_loss": 0.12087084352970123, "global_step": 172292, "epoch": 1935, "lr": 2.9561884828730553e-05}
{"train_loss": 0.07302316278219223, "global_step": 172293, "epoch": 1935, "lr": 2.956135573877945e-05}
{"train_loss": 0.09707130491733551, "global_step": 172294, "epoch": 1935, "lr": 2.9560826651576057e-05}
{"train_loss": 0.0997786819934845, "global_step": 172295, "epoch": 1935, "lr": 2.9560297567120475e-05}
{"train_loss": 0.06568561494350433, "global_step": 172296, "epoch": 1935, "lr": 2.955976848541275e-05}
{"train_loss": 0.10580844432115555, "global_step": 172297, "epoch": 1935, "lr": 2.9559239406452954e-05}
{"train_loss": 0.09865383803844452, "global_step": 172298, "epoch": 1935, "lr": 2.95587103302412e-05}
{"train_loss": 0.05390683189034462, "global_step": 172299, "epoch": 1935, "lr": 2.9558181256777505e-05}
{"train_loss": 0.07235148549079895, "global_step": 172300, "epoch": 1935, "lr": 2.955765218606198e-05}
{"train_loss": 0.09334193915128708, "global_step": 172301, "epoch": 1935, "lr": 2.9557123118094677e-05}
{"train_loss": 0.1498412936925888, "global_step": 172302, "epoch": 1935, "lr": 2.9556594052875664e-05}
{"train_loss": 0.09561020939537648, "global_step": 172303, "epoch": 1935, "lr": 2.9556064990405018e-05, "val_loss": 7.25648832321167, "train_action_mse_error": 27.625167846679688}
{"train_loss": 0.07045195251703262, "global_step": 172304, "epoch": 1936, "lr": 2.955553593068281e-05}
{"train_loss": 0.07333476841449738, "global_step": 172305, "epoch": 1936, "lr": 2.9555006873709112e-05}
{"train_loss": 0.06593034416437149, "global_step": 172306, "epoch": 1936, "lr": 2.9554477819484007e-05}
{"train_loss": 0.049443986266851425, "global_step": 172307, "epoch": 1936, "lr": 2.9553948768007528e-05}
{"train_loss": 0.09299428015947342, "global_step": 172308, "epoch": 1936, "lr": 2.9553419719279795e-05}
{"train_loss": 0.07257838547229767, "global_step": 172309, "epoch": 1936, "lr": 2.955289067330083e-05}
{"train_loss": 0.1352376639842987, "global_step": 172310, "epoch": 1936, "lr": 2.955236163007076e-05}
{"train_loss": 0.10706494003534317, "global_step": 172311, "epoch": 1936, "lr": 2.955183258958959e-05}
{"train_loss": 0.04679287225008011, "global_step": 172312, "epoch": 1936, "lr": 2.9551303551857457e-05}
{"train_loss": 0.10215157270431519, "global_step": 172313, "epoch": 1936, "lr": 2.9550774516874386e-05}
{"train_loss": 0.13140970468521118, "global_step": 172314, "epoch": 1936, "lr": 2.9550245484640475e-05}
{"train_loss": 0.0960598811507225, "global_step": 172315, "epoch": 1936, "lr": 2.9549716455155773e-05}
{"train_loss": 0.07049787789583206, "global_step": 172316, "epoch": 1936, "lr": 2.9549187428420357e-05}
{"train_loss": 0.06749472767114639, "global_step": 172317, "epoch": 1936, "lr": 2.9548658404434325e-05}
{"train_loss": 0.08130587637424469, "global_step": 172318, "epoch": 1936, "lr": 2.9548129383197704e-05}
{"train_loss": 0.05529937148094177, "global_step": 172319, "epoch": 1936, "lr": 2.954760036471061e-05}
{"train_loss": 0.08778314292430878, "global_step": 172320, "epoch": 1936, "lr": 2.9547071348973083e-05}
{"train_loss": 0.11690026521682739, "global_step": 172321, "epoch": 1936, "lr": 2.95465423359852e-05}
{"train_loss": 0.09693276882171631, "global_step": 172322, "epoch": 1936, "lr": 2.9546013325747035e-05}
{"train_loss": 0.0998590961098671, "global_step": 172323, "epoch": 1936, "lr": 2.9545484318258664e-05}
{"train_loss": 0.05884444713592529, "global_step": 172324, "epoch": 1936, "lr": 2.954495531352015e-05}
{"train_loss": 0.11484318226575851, "global_step": 172325, "epoch": 1936, "lr": 2.9544426311531575e-05}
{"train_loss": 0.05822335556149483, "global_step": 172326, "epoch": 1936, "lr": 2.954389731229299e-05}
{"train_loss": 0.09309540688991547, "global_step": 172327, "epoch": 1936, "lr": 2.9543368315804486e-05}
{"train_loss": 0.11779841780662537, "global_step": 172328, "epoch": 1936, "lr": 2.954283932206612e-05}
{"train_loss": 0.06557828933000565, "global_step": 172329, "epoch": 1936, "lr": 2.9542310331077983e-05}
{"train_loss": 0.08938662707805634, "global_step": 172330, "epoch": 1936, "lr": 2.9541781342840113e-05}
{"train_loss": 0.07387587428092957, "global_step": 172331, "epoch": 1936, "lr": 2.9541252357352623e-05}
{"train_loss": 0.0880618542432785, "global_step": 172332, "epoch": 1936, "lr": 2.9540723374615543e-05}
{"train_loss": 0.042753223329782486, "global_step": 172333, "epoch": 1936, "lr": 2.9540194394628963e-05}
{"train_loss": 0.10617420822381973, "global_step": 172334, "epoch": 1936, "lr": 2.9539665417392976e-05}
{"train_loss": 0.07387641817331314, "global_step": 172335, "epoch": 1936, "lr": 2.953913644290761e-05}
{"train_loss": 0.0565752312541008, "global_step": 172336, "epoch": 1936, "lr": 2.9538607471172974e-05}
{"train_loss": 0.10178136080503464, "global_step": 172337, "epoch": 1936, "lr": 2.9538078502189105e-05}
{"train_loss": 0.0868171751499176, "global_step": 172338, "epoch": 1936, "lr": 2.9537549535956117e-05}
{"train_loss": 0.06579485535621643, "global_step": 172339, "epoch": 1936, "lr": 2.9537020572474038e-05}
{"train_loss": 0.08688956499099731, "global_step": 172340, "epoch": 1936, "lr": 2.9536491611742966e-05}
{"train_loss": 0.06722215563058853, "global_step": 172341, "epoch": 1936, "lr": 2.9535962653762955e-05}
{"train_loss": 0.09116519242525101, "global_step": 172342, "epoch": 1936, "lr": 2.9535433698534088e-05}
{"train_loss": 0.11505087465047836, "global_step": 172343, "epoch": 1936, "lr": 2.9534904746056437e-05}
{"train_loss": 0.09686042368412018, "global_step": 172344, "epoch": 1936, "lr": 2.953437579633006e-05}
{"train_loss": 0.07390766590833664, "global_step": 172345, "epoch": 1936, "lr": 2.953384684935504e-05}
{"train_loss": 0.03859875723719597, "global_step": 172346, "epoch": 1936, "lr": 2.953331790513146e-05}
{"train_loss": 0.09237447381019592, "global_step": 172347, "epoch": 1936, "lr": 2.9532788963659352e-05}
{"train_loss": 0.14228029549121857, "global_step": 172348, "epoch": 1936, "lr": 2.9532260024938828e-05}
{"train_loss": 0.1160394549369812, "global_step": 172349, "epoch": 1936, "lr": 2.9531731088969927e-05}
{"train_loss": 0.10337609797716141, "global_step": 172350, "epoch": 1936, "lr": 2.9531202155752745e-05}
{"train_loss": 0.05540289729833603, "global_step": 172351, "epoch": 1936, "lr": 2.9530673225287336e-05}
{"train_loss": 0.08198904246091843, "global_step": 172352, "epoch": 1936, "lr": 2.9530144297573775e-05}
{"train_loss": 0.09728561341762543, "global_step": 172353, "epoch": 1936, "lr": 2.9529615372612152e-05}
{"train_loss": 0.12704698741436005, "global_step": 172354, "epoch": 1936, "lr": 2.952908645040251e-05}
{"train_loss": 0.11784085631370544, "global_step": 172355, "epoch": 1936, "lr": 2.9528557530944944e-05}
{"train_loss": 0.09526423364877701, "global_step": 172356, "epoch": 1936, "lr": 2.95280286142395e-05}
{"train_loss": 0.0654035210609436, "global_step": 172357, "epoch": 1936, "lr": 2.9527499700286287e-05}
{"train_loss": 0.09084265679121017, "global_step": 172358, "epoch": 1936, "lr": 2.9526970789085328e-05}
{"train_loss": 0.05833488330245018, "global_step": 172359, "epoch": 1936, "lr": 2.9526441880636734e-05}
{"train_loss": 0.11909395456314087, "global_step": 172360, "epoch": 1936, "lr": 2.9525912974940558e-05}
{"train_loss": 0.06401984393596649, "global_step": 172361, "epoch": 1936, "lr": 2.9525384071996865e-05}
{"train_loss": 0.08851025998592377, "global_step": 172362, "epoch": 1936, "lr": 2.9524855171805742e-05}
{"train_loss": 0.04533032700419426, "global_step": 172363, "epoch": 1936, "lr": 2.952432627436725e-05}
{"train_loss": 0.060766566544771194, "global_step": 172364, "epoch": 1936, "lr": 2.952379737968146e-05}
{"train_loss": 0.061471112072467804, "global_step": 172365, "epoch": 1936, "lr": 2.9523268487748457e-05}
{"train_loss": 0.1161939799785614, "global_step": 172366, "epoch": 1936, "lr": 2.9522739598568285e-05}
{"train_loss": 0.12666508555412292, "global_step": 172367, "epoch": 1936, "lr": 2.9522210712141047e-05}
{"train_loss": 0.08112727105617523, "global_step": 172368, "epoch": 1936, "lr": 2.9521681828466778e-05}
{"train_loss": 0.06973406672477722, "global_step": 172369, "epoch": 1936, "lr": 2.9521152947545593e-05}
{"train_loss": 0.09836550056934357, "global_step": 172370, "epoch": 1936, "lr": 2.9520624069377517e-05}
{"train_loss": 0.10781576484441757, "global_step": 172371, "epoch": 1936, "lr": 2.9520095193962642e-05}
{"train_loss": 0.094658762216568, "global_step": 172372, "epoch": 1936, "lr": 2.9519566321301062e-05}
{"train_loss": 0.05096156895160675, "global_step": 172373, "epoch": 1936, "lr": 2.9519037451392807e-05}
{"train_loss": 0.05808328092098236, "global_step": 172374, "epoch": 1936, "lr": 2.951850858423799e-05}
{"train_loss": 0.07306737452745438, "global_step": 172375, "epoch": 1936, "lr": 2.9517979719836636e-05}
{"train_loss": 0.12492512911558151, "global_step": 172376, "epoch": 1936, "lr": 2.9517450858188862e-05}
{"train_loss": 0.0824577584862709, "global_step": 172377, "epoch": 1936, "lr": 2.95169219992947e-05}
{"train_loss": 0.10653732717037201, "global_step": 172378, "epoch": 1936, "lr": 2.9516393143154252e-05}
{"train_loss": 0.07615768164396286, "global_step": 172379, "epoch": 1936, "lr": 2.951586428976757e-05}
{"train_loss": 0.06344743072986603, "global_step": 172380, "epoch": 1936, "lr": 2.951533543913473e-05}
{"train_loss": 0.08385930210351944, "global_step": 172381, "epoch": 1936, "lr": 2.9514806591255794e-05}
{"train_loss": 0.08991696685552597, "global_step": 172382, "epoch": 1936, "lr": 2.9514277746130852e-05}
{"train_loss": 0.158779576420784, "global_step": 172383, "epoch": 1936, "lr": 2.9513748903759963e-05}
{"train_loss": 0.05938053876161575, "global_step": 172384, "epoch": 1936, "lr": 2.9513220064143215e-05}
{"train_loss": 0.08281642198562622, "global_step": 172385, "epoch": 1936, "lr": 2.9512691227280642e-05}
{"train_loss": 0.08199942111968994, "global_step": 172386, "epoch": 1936, "lr": 2.951216239317236e-05}
{"train_loss": 0.08288061618804932, "global_step": 172387, "epoch": 1936, "lr": 2.9511633561818398e-05}
{"train_loss": 0.06624950468540192, "global_step": 172388, "epoch": 1936, "lr": 2.951110473321884e-05}
{"train_loss": 0.053903818130493164, "global_step": 172389, "epoch": 1936, "lr": 2.951057590737379e-05}
{"train_loss": 0.07188843190670013, "global_step": 172390, "epoch": 1936, "lr": 2.951004708428327e-05}
{"train_loss": 0.08410315215587616, "global_step": 172391, "epoch": 1936, "lr": 2.95095182639474e-05}
{"train_loss": 0.0853665905136071, "global_step": 172392, "epoch": 1936, "lr": 2.95089894463662e-05, "val_loss": 7.351062297821045}
{"train_loss": 0.07731876522302628, "global_step": 172393, "epoch": 1937, "lr": 2.950846063153979e-05}
{"train_loss": 0.09728123247623444, "global_step": 172394, "epoch": 1937, "lr": 2.9507931819468192e-05}
{"train_loss": 0.09475575387477875, "global_step": 172395, "epoch": 1937, "lr": 2.9507403010151534e-05}
{"train_loss": 0.08837806433439255, "global_step": 172396, "epoch": 1937, "lr": 2.950687420358983e-05}
{"train_loss": 0.0900273248553276, "global_step": 172397, "epoch": 1937, "lr": 2.95063453997832e-05}
{"train_loss": 0.12428467720746994, "global_step": 172398, "epoch": 1937, "lr": 2.950581659873167e-05}
{"train_loss": 0.15575039386749268, "global_step": 172399, "epoch": 1937, "lr": 2.9505287800435345e-05}
{"train_loss": 0.06238193064928055, "global_step": 172400, "epoch": 1937, "lr": 2.950475900489428e-05}
{"train_loss": 0.06990842521190643, "global_step": 172401, "epoch": 1937, "lr": 2.9504230212108553e-05}
{"train_loss": 0.04148055613040924, "global_step": 172402, "epoch": 1937, "lr": 2.9503701422078235e-05}
{"train_loss": 0.0637086033821106, "global_step": 172403, "epoch": 1937, "lr": 2.9503172634803393e-05}
{"train_loss": 0.038021381944417953, "global_step": 172404, "epoch": 1937, "lr": 2.950264385028409e-05}
{"train_loss": 0.06184716150164604, "global_step": 172405, "epoch": 1937, "lr": 2.950211506852042e-05}
{"train_loss": 0.04402568191289902, "global_step": 172406, "epoch": 1937, "lr": 2.950158628951243e-05}
{"train_loss": 0.08467410504817963, "global_step": 172407, "epoch": 1937, "lr": 2.9501057513260194e-05}
{"train_loss": 0.06138818711042404, "global_step": 172408, "epoch": 1937, "lr": 2.950052873976381e-05}
{"train_loss": 0.09464655071496964, "global_step": 172409, "epoch": 1937, "lr": 2.9499999969023316e-05}
{"train_loss": 0.06748664379119873, "global_step": 172410, "epoch": 1937, "lr": 2.9499471201038813e-05}
{"train_loss": 0.060382042080163956, "global_step": 172411, "epoch": 1937, "lr": 2.9498942435810338e-05}
{"train_loss": 0.04861261323094368, "global_step": 172412, "epoch": 1937, "lr": 2.9498413673338e-05}
{"train_loss": 0.10281682759523392, "global_step": 172413, "epoch": 1937, "lr": 2.9497884913621833e-05}
{"train_loss": 0.08947674930095673, "global_step": 172414, "epoch": 1937, "lr": 2.949735615666194e-05}
{"train_loss": 0.07446303218603134, "global_step": 172415, "epoch": 1937, "lr": 2.949682740245836e-05}
{"train_loss": 0.06217207759618759, "global_step": 172416, "epoch": 1937, "lr": 2.94962986510112e-05}
{"train_loss": 0.05913389101624489, "global_step": 172417, "epoch": 1937, "lr": 2.9495769902320507e-05}
{"train_loss": 0.07698819041252136, "global_step": 172418, "epoch": 1937, "lr": 2.9495241156386355e-05}
{"train_loss": 0.1364155262708664, "global_step": 172419, "epoch": 1937, "lr": 2.9494712413208825e-05}
{"train_loss": 0.07082057744264603, "global_step": 172420, "epoch": 1937, "lr": 2.9494183672787966e-05}
{"train_loss": 0.057967010885477066, "global_step": 172421, "epoch": 1937, "lr": 2.9493654935123875e-05}
{"train_loss": 0.06129719316959381, "global_step": 172422, "epoch": 1937, "lr": 2.9493126200216625e-05}
{"train_loss": 0.0820394977927208, "global_step": 172423, "epoch": 1937, "lr": 2.9492597468066247e-05}
{"train_loss": 0.041034504771232605, "global_step": 172424, "epoch": 1937, "lr": 2.949206873867284e-05}
{"train_loss": 0.07168889045715332, "global_step": 172425, "epoch": 1937, "lr": 2.94915400120365e-05}
{"train_loss": 0.08267684280872345, "global_step": 172426, "epoch": 1937, "lr": 2.9491011288157244e-05}
{"train_loss": 0.07168661803007126, "global_step": 172427, "epoch": 1937, "lr": 2.9490482567035194e-05}
{"train_loss": 0.10349802672863007, "global_step": 172428, "epoch": 1937, "lr": 2.9489953848670383e-05}
{"train_loss": 0.11931975185871124, "global_step": 172429, "epoch": 1937, "lr": 2.948942513306292e-05}
{"train_loss": 0.07977592200040817, "global_step": 172430, "epoch": 1937, "lr": 2.9488896420212825e-05}
{"train_loss": 0.08203732967376709, "global_step": 172431, "epoch": 1937, "lr": 2.9488367710120222e-05}
{"train_loss": 0.028738081455230713, "global_step": 172432, "epoch": 1937, "lr": 2.9487839002785135e-05}
{"train_loss": 0.07312462478876114, "global_step": 172433, "epoch": 1937, "lr": 2.9487310298207682e-05}
{"train_loss": 0.0835634246468544, "global_step": 172434, "epoch": 1937, "lr": 2.948678159638789e-05}
{"train_loss": 0.09091013669967651, "global_step": 172435, "epoch": 1937, "lr": 2.9486252897325872e-05}
{"train_loss": 0.07954147458076477, "global_step": 172436, "epoch": 1937, "lr": 2.9485724201021664e-05}
{"train_loss": 0.036156997084617615, "global_step": 172437, "epoch": 1937, "lr": 2.9485195507475348e-05}
{"train_loss": 0.05262457951903343, "global_step": 172438, "epoch": 1937, "lr": 2.9484666816687002e-05}
{"train_loss": 0.11818862706422806, "global_step": 172439, "epoch": 1937, "lr": 2.948413812865669e-05}
{"train_loss": 0.1091003343462944, "global_step": 172440, "epoch": 1937, "lr": 2.9483609443384487e-05}
{"train_loss": 0.06944422423839569, "global_step": 172441, "epoch": 1937, "lr": 2.9483080760870473e-05}
{"train_loss": 0.07772206515073776, "global_step": 172442, "epoch": 1937, "lr": 2.948255208111469e-05}
{"train_loss": 0.08619269728660583, "global_step": 172443, "epoch": 1937, "lr": 2.948202340411722e-05}
{"train_loss": 0.07071994245052338, "global_step": 172444, "epoch": 1937, "lr": 2.9481494729878166e-05}
{"train_loss": 0.09067993611097336, "global_step": 172445, "epoch": 1937, "lr": 2.9480966058397563e-05}
{"train_loss": 0.07080570608377457, "global_step": 172446, "epoch": 1937, "lr": 2.9480437389675503e-05}
{"train_loss": 0.11778255552053452, "global_step": 172447, "epoch": 1937, "lr": 2.9479908723712035e-05}
{"train_loss": 0.10467946529388428, "global_step": 172448, "epoch": 1937, "lr": 2.9479380060507256e-05}
{"train_loss": 0.10493918508291245, "global_step": 172449, "epoch": 1937, "lr": 2.9478851400061212e-05}
{"train_loss": 0.11723829060792923, "global_step": 172450, "epoch": 1937, "lr": 2.9478322742374002e-05}
{"train_loss": 0.10164462774991989, "global_step": 172451, "epoch": 1937, "lr": 2.9477794087445663e-05}
{"train_loss": 0.11596403270959854, "global_step": 172452, "epoch": 1937, "lr": 2.9477265435276304e-05}
{"train_loss": 0.04026508703827858, "global_step": 172453, "epoch": 1937, "lr": 2.9476736785865954e-05}
{"train_loss": 0.1401062160730362, "global_step": 172454, "epoch": 1937, "lr": 2.9476208139214733e-05}
{"train_loss": 0.08752724528312683, "global_step": 172455, "epoch": 1937, "lr": 2.947567949532266e-05}
{"train_loss": 0.1162940040230751, "global_step": 172456, "epoch": 1937, "lr": 2.9475150854189852e-05}
{"train_loss": 0.12215840071439743, "global_step": 172457, "epoch": 1937, "lr": 2.947462221581635e-05}
{"train_loss": 0.06413691490888596, "global_step": 172458, "epoch": 1937, "lr": 2.9474093580202238e-05}
{"train_loss": 0.06756623834371567, "global_step": 172459, "epoch": 1937, "lr": 2.9473564947347588e-05}
{"train_loss": 0.08413252979516983, "global_step": 172460, "epoch": 1937, "lr": 2.9473036317252446e-05}
{"train_loss": 0.10531967133283615, "global_step": 172461, "epoch": 1937, "lr": 2.947250768991693e-05}
{"train_loss": 0.12198585271835327, "global_step": 172462, "epoch": 1937, "lr": 2.9471979065341064e-05}
{"train_loss": 0.08856461942195892, "global_step": 172463, "epoch": 1937, "lr": 2.947145044352496e-05}
{"train_loss": 0.05849621444940567, "global_step": 172464, "epoch": 1937, "lr": 2.947092182446865e-05}
{"train_loss": 0.07089313119649887, "global_step": 172465, "epoch": 1937, "lr": 2.947039320817224e-05}
{"train_loss": 0.06937648355960846, "global_step": 172466, "epoch": 1937, "lr": 2.9469864594635776e-05}
{"train_loss": 0.13419826328754425, "global_step": 172467, "epoch": 1937, "lr": 2.9469335983859348e-05}
{"train_loss": 0.06030839681625366, "global_step": 172468, "epoch": 1937, "lr": 2.9468807375843004e-05}
{"train_loss": 0.05023546889424324, "global_step": 172469, "epoch": 1937, "lr": 2.9468278770586845e-05}
{"train_loss": 0.18855279684066772, "global_step": 172470, "epoch": 1937, "lr": 2.9467750168090913e-05}
{"train_loss": 0.06483884155750275, "global_step": 172471, "epoch": 1937, "lr": 2.94672215683553e-05}
{"train_loss": 0.0988205224275589, "global_step": 172472, "epoch": 1937, "lr": 2.9466692971380056e-05}
{"train_loss": 0.048746414482593536, "global_step": 172473, "epoch": 1937, "lr": 2.9466164377165283e-05}
{"train_loss": 0.07603470236063004, "global_step": 172474, "epoch": 1937, "lr": 2.9465635785711015e-05}
{"train_loss": 0.04234429448843002, "global_step": 172475, "epoch": 1937, "lr": 2.9465107197017362e-05}
{"train_loss": 0.08323170989751816, "global_step": 172476, "epoch": 1937, "lr": 2.9464578611084363e-05}
{"train_loss": 0.04737352207303047, "global_step": 172477, "epoch": 1937, "lr": 2.9464050027912105e-05}
{"train_loss": 0.07078903913497925, "global_step": 172478, "epoch": 1937, "lr": 2.9463521447500653e-05}
{"train_loss": 0.07820242643356323, "global_step": 172479, "epoch": 1937, "lr": 2.946299286985007e-05}
{"train_loss": 0.11685517430305481, "global_step": 172480, "epoch": 1937, "lr": 2.9462464294960456e-05}
{"train_loss": 0.08246515339679933, "global_step": 172481, "epoch": 1937, "lr": 2.9461935722831845e-05, "val_loss": 7.409488201141357}
{"train_loss": 0.11911088228225708, "global_step": 172482, "epoch": 1938, "lr": 2.9461407153464343e-05}
{"train_loss": 0.1128535345196724, "global_step": 172483, "epoch": 1938, "lr": 2.9460878586857986e-05}
{"train_loss": 0.06435085833072662, "global_step": 172484, "epoch": 1938, "lr": 2.9460350023012882e-05}
{"train_loss": 0.06535254418849945, "global_step": 172485, "epoch": 1938, "lr": 2.9459821461929072e-05}
{"train_loss": 0.03857576847076416, "global_step": 172486, "epoch": 1938, "lr": 2.945929290360665e-05}
{"train_loss": 0.09292832016944885, "global_step": 172487, "epoch": 1938, "lr": 2.9458764348045663e-05}
{"train_loss": 0.06610915064811707, "global_step": 172488, "epoch": 1938, "lr": 2.9458235795246208e-05}
{"train_loss": 0.0929533839225769, "global_step": 172489, "epoch": 1938, "lr": 2.9457707245208327e-05}
{"train_loss": 0.07541666179895401, "global_step": 172490, "epoch": 1938, "lr": 2.9457178697932126e-05}
{"train_loss": 0.07717736065387726, "global_step": 172491, "epoch": 1938, "lr": 2.9456650153417635e-05}
{"train_loss": 0.051795173436403275, "global_step": 172492, "epoch": 1938, "lr": 2.9456121611664966e-05}
{"train_loss": 0.0937521681189537, "global_step": 172493, "epoch": 1938, "lr": 2.945559307267416e-05}
{"train_loss": 0.13929130136966705, "global_step": 172494, "epoch": 1938, "lr": 2.945506453644531e-05}
{"train_loss": 0.053151633590459824, "global_step": 172495, "epoch": 1938, "lr": 2.9454536002978465e-05}
{"train_loss": 0.12047265470027924, "global_step": 172496, "epoch": 1938, "lr": 2.9454007472273715e-05}
{"train_loss": 0.04291653260588646, "global_step": 172497, "epoch": 1938, "lr": 2.9453478944331115e-05}
{"train_loss": 0.04902472347021103, "global_step": 172498, "epoch": 1938, "lr": 2.9452950419150738e-05}
{"train_loss": 0.10109766572713852, "global_step": 172499, "epoch": 1938, "lr": 2.9452421896732684e-05}
{"train_loss": 0.06630239635705948, "global_step": 172500, "epoch": 1938, "lr": 2.9451893377076977e-05}
{"train_loss": 0.07594713568687439, "global_step": 172501, "epoch": 1938, "lr": 2.945136486018374e-05}
{"train_loss": 0.04599470645189285, "global_step": 172502, "epoch": 1938, "lr": 2.945083634605299e-05}
{"train_loss": 0.06093853339552879, "global_step": 172503, "epoch": 1938, "lr": 2.9450307834684844e-05}
{"train_loss": 0.06439302861690521, "global_step": 172504, "epoch": 1938, "lr": 2.9449779326079336e-05}
{"train_loss": 0.04871617257595062, "global_step": 172505, "epoch": 1938, "lr": 2.9449250820236575e-05}
{"train_loss": 0.0878206267952919, "global_step": 172506, "epoch": 1938, "lr": 2.944872231715659e-05}
{"train_loss": 0.0986514762043953, "global_step": 172507, "epoch": 1938, "lr": 2.9448193816839497e-05}
{"train_loss": 0.07857518643140793, "global_step": 172508, "epoch": 1938, "lr": 2.9447665319285322e-05}
{"train_loss": 0.055356215685606, "global_step": 172509, "epoch": 1938, "lr": 2.9447136824494176e-05}
{"train_loss": 0.042209308594465256, "global_step": 172510, "epoch": 1938, "lr": 2.9446608332466096e-05}
{"train_loss": 0.14778830111026764, "global_step": 172511, "epoch": 1938, "lr": 2.944607984320118e-05}
{"train_loss": 0.09939945489168167, "global_step": 172512, "epoch": 1938, "lr": 2.9445551356699473e-05}
{"train_loss": 0.08117032051086426, "global_step": 172513, "epoch": 1938, "lr": 2.9445022872961087e-05}
{"train_loss": 0.11782728135585785, "global_step": 172514, "epoch": 1938, "lr": 2.9444494391986054e-05}
{"train_loss": 0.07044487446546555, "global_step": 172515, "epoch": 1938, "lr": 2.9443965913774446e-05}
{"train_loss": 0.0949903354048729, "global_step": 172516, "epoch": 1938, "lr": 2.944343743832637e-05}
{"train_loss": 0.06821561604738235, "global_step": 172517, "epoch": 1938, "lr": 2.9442908965641847e-05}
{"train_loss": 0.16194462776184082, "global_step": 172518, "epoch": 1938, "lr": 2.9442380495721e-05}
{"train_loss": 0.07423568516969681, "global_step": 172519, "epoch": 1938, "lr": 2.9441852028563854e-05}
{"train_loss": 0.03502557799220085, "global_step": 172520, "epoch": 1938, "lr": 2.944132356417052e-05}
{"train_loss": 0.043651141226291656, "global_step": 172521, "epoch": 1938, "lr": 2.9440795102541023e-05}
{"train_loss": 0.10050763189792633, "global_step": 172522, "epoch": 1938, "lr": 2.9440266643675485e-05}
{"train_loss": 0.0757969543337822, "global_step": 172523, "epoch": 1938, "lr": 2.9439738187573934e-05}
{"train_loss": 0.13057269155979156, "global_step": 172524, "epoch": 1938, "lr": 2.9439209734236482e-05}
{"train_loss": 0.04557643458247185, "global_step": 172525, "epoch": 1938, "lr": 2.943868128366315e-05}
{"train_loss": 0.08745753765106201, "global_step": 172526, "epoch": 1938, "lr": 2.943815283585406e-05}
{"train_loss": 0.11098896712064743, "global_step": 172527, "epoch": 1938, "lr": 2.943762439080924e-05}
{"train_loss": 0.06419423967599869, "global_step": 172528, "epoch": 1938, "lr": 2.9437095948528802e-05}
{"train_loss": 0.06716819107532501, "global_step": 172529, "epoch": 1938, "lr": 2.943656750901278e-05}
{"train_loss": 0.10097965598106384, "global_step": 172530, "epoch": 1938, "lr": 2.943603907226128e-05}
{"train_loss": 0.054953668266534805, "global_step": 172531, "epoch": 1938, "lr": 2.9435510638274322e-05}
{"train_loss": 0.06792991608381271, "global_step": 172532, "epoch": 1938, "lr": 2.9434982207052043e-05}
{"train_loss": 0.070498988032341, "global_step": 172533, "epoch": 1938, "lr": 2.9434453778594463e-05}
{"train_loss": 0.06619768589735031, "global_step": 172534, "epoch": 1938, "lr": 2.943392535290166e-05}
{"train_loss": 0.09875726699829102, "global_step": 172535, "epoch": 1938, "lr": 2.9433396929973733e-05}
{"train_loss": 0.03295833617448807, "global_step": 172536, "epoch": 1938, "lr": 2.943286850981072e-05}
{"train_loss": 0.052628885954618454, "global_step": 172537, "epoch": 1938, "lr": 2.9432340092412723e-05}
{"train_loss": 0.07599955052137375, "global_step": 172538, "epoch": 1938, "lr": 2.9431811677779784e-05}
{"train_loss": 0.08209265768527985, "global_step": 172539, "epoch": 1938, "lr": 2.9431283265912e-05}
{"train_loss": 0.08203349262475967, "global_step": 172540, "epoch": 1938, "lr": 2.943075485680942e-05}
{"train_loss": 0.07707159221172333, "global_step": 172541, "epoch": 1938, "lr": 2.943022645047213e-05}
{"train_loss": 0.06643422693014145, "global_step": 172542, "epoch": 1938, "lr": 2.9429698046900182e-05}
{"train_loss": 0.07050672173500061, "global_step": 172543, "epoch": 1938, "lr": 2.942916964609368e-05}
{"train_loss": 0.10500667244195938, "global_step": 172544, "epoch": 1938, "lr": 2.9428641248052657e-05}
{"train_loss": 0.08287519216537476, "global_step": 172545, "epoch": 1938, "lr": 2.9428112852777223e-05}
{"train_loss": 0.06476939469575882, "global_step": 172546, "epoch": 1938, "lr": 2.942758446026741e-05}
{"train_loss": 0.09834297001361847, "global_step": 172547, "epoch": 1938, "lr": 2.9427056070523317e-05}
{"train_loss": 0.11240845173597336, "global_step": 172548, "epoch": 1938, "lr": 2.9426527683544998e-05}
{"train_loss": 0.06849600374698639, "global_step": 172549, "epoch": 1938, "lr": 2.942599929933254e-05}
{"train_loss": 0.100912906229496, "global_step": 172550, "epoch": 1938, "lr": 2.9425470917886e-05}
{"train_loss": 0.10906396806240082, "global_step": 172551, "epoch": 1938, "lr": 2.942494253920545e-05}
{"train_loss": 0.12978523969650269, "global_step": 172552, "epoch": 1938, "lr": 2.9424414163290985e-05}
{"train_loss": 0.09921208024024963, "global_step": 172553, "epoch": 1938, "lr": 2.9423885790142637e-05}
{"train_loss": 0.06428791582584381, "global_step": 172554, "epoch": 1938, "lr": 2.9423357419760512e-05}
{"train_loss": 0.04151397943496704, "global_step": 172555, "epoch": 1938, "lr": 2.9422829052144647e-05}
{"train_loss": 0.06239442154765129, "global_step": 172556, "epoch": 1938, "lr": 2.9422300687295152e-05}
{"train_loss": 0.05179440602660179, "global_step": 172557, "epoch": 1938, "lr": 2.942177232521206e-05}
{"train_loss": 0.11576715111732483, "global_step": 172558, "epoch": 1938, "lr": 2.9421243965895475e-05}
{"train_loss": 0.08618637919425964, "global_step": 172559, "epoch": 1938, "lr": 2.9420715609345438e-05}
{"train_loss": 0.0686502680182457, "global_step": 172560, "epoch": 1938, "lr": 2.9420187255562056e-05}
{"train_loss": 0.1013256385922432, "global_step": 172561, "epoch": 1938, "lr": 2.9419658904545357e-05}
{"train_loss": 0.09388916194438934, "global_step": 172562, "epoch": 1938, "lr": 2.941913055629545e-05}
{"train_loss": 0.07081484794616699, "global_step": 172563, "epoch": 1938, "lr": 2.9418602210812374e-05}
{"train_loss": 0.08439768105745316, "global_step": 172564, "epoch": 1938, "lr": 2.941807386809623e-05}
{"train_loss": 0.05758921429514885, "global_step": 172565, "epoch": 1938, "lr": 2.9417545528147063e-05}
{"train_loss": 0.10663725435733795, "global_step": 172566, "epoch": 1938, "lr": 2.9417017190964973e-05}
{"train_loss": 0.05960216745734215, "global_step": 172567, "epoch": 1938, "lr": 2.9416488856549995e-05}
{"train_loss": 0.09609653055667877, "global_step": 172568, "epoch": 1938, "lr": 2.9415960524902243e-05}
{"train_loss": 0.07946322858333588, "global_step": 172569, "epoch": 1938, "lr": 2.9415432196021737e-05}
{"train_loss": 0.07979743793774187, "global_step": 172570, "epoch": 1938, "lr": 2.941490386990859e-05, "val_loss": 7.045375347137451}
{"train_loss": 0.06419634819030762, "global_step": 172571, "epoch": 1939, "lr": 2.9414375546562855e-05}
{"train_loss": 0.08288109302520752, "global_step": 172572, "epoch": 1939, "lr": 2.9413847225984607e-05}
{"train_loss": 0.0985860526561737, "global_step": 172573, "epoch": 1939, "lr": 2.9413318908173924e-05}
{"train_loss": 0.1473926603794098, "global_step": 172574, "epoch": 1939, "lr": 2.9412790593130857e-05}
{"train_loss": 0.07287339866161346, "global_step": 172575, "epoch": 1939, "lr": 2.9412262280855507e-05}
{"train_loss": 0.12232082337141037, "global_step": 172576, "epoch": 1939, "lr": 2.94117339713479e-05}
{"train_loss": 0.07766557484865189, "global_step": 172577, "epoch": 1939, "lr": 2.9411205664608167e-05}
{"train_loss": 0.08465509861707687, "global_step": 172578, "epoch": 1939, "lr": 2.941067736063632e-05}
{"train_loss": 0.08415115624666214, "global_step": 172579, "epoch": 1939, "lr": 2.941014905943247e-05}
{"train_loss": 0.09665682166814804, "global_step": 172580, "epoch": 1939, "lr": 2.9409620760996658e-05}
{"train_loss": 0.058283731341362, "global_step": 172581, "epoch": 1939, "lr": 2.9409092465328996e-05}
{"train_loss": 0.10379309207201004, "global_step": 172582, "epoch": 1939, "lr": 2.9408564172429508e-05}
{"train_loss": 0.07563692331314087, "global_step": 172583, "epoch": 1939, "lr": 2.9408035882298313e-05}
{"train_loss": 0.08648911118507385, "global_step": 172584, "epoch": 1939, "lr": 2.9407507594935425e-05}
{"train_loss": 0.08202425390481949, "global_step": 172585, "epoch": 1939, "lr": 2.9406979310340977e-05}
{"train_loss": 0.06633488088846207, "global_step": 172586, "epoch": 1939, "lr": 2.940645102851498e-05}
{"train_loss": 0.10086305439472198, "global_step": 172587, "epoch": 1939, "lr": 2.9405922749457553e-05}
{"train_loss": 0.06611399352550507, "global_step": 172588, "epoch": 1939, "lr": 2.9405394473168747e-05}
{"train_loss": 0.09113316982984543, "global_step": 172589, "epoch": 1939, "lr": 2.9404866199648634e-05}
{"train_loss": 0.04465750977396965, "global_step": 172590, "epoch": 1939, "lr": 2.9404337928897286e-05}
{"train_loss": 0.08941730111837387, "global_step": 172591, "epoch": 1939, "lr": 2.9403809660914773e-05}
{"train_loss": 0.043812450021505356, "global_step": 172592, "epoch": 1939, "lr": 2.9403281395701177e-05}
{"train_loss": 0.145907461643219, "global_step": 172593, "epoch": 1939, "lr": 2.940275313325654e-05}
{"train_loss": 0.0998741164803505, "global_step": 172594, "epoch": 1939, "lr": 2.940222487358097e-05}
{"train_loss": 0.11556726694107056, "global_step": 172595, "epoch": 1939, "lr": 2.94016966166745e-05}
{"train_loss": 0.08015915751457214, "global_step": 172596, "epoch": 1939, "lr": 2.9401168362537245e-05}
{"train_loss": 0.12388581782579422, "global_step": 172597, "epoch": 1939, "lr": 2.9400640111169226e-05}
{"train_loss": 0.09215475618839264, "global_step": 172598, "epoch": 1939, "lr": 2.940011186257056e-05}
{"train_loss": 0.07606961578130722, "global_step": 172599, "epoch": 1939, "lr": 2.9399583616741276e-05}
{"train_loss": 0.09769030660390854, "global_step": 172600, "epoch": 1939, "lr": 2.9399055373681495e-05}
{"train_loss": 0.07513183355331421, "global_step": 172601, "epoch": 1939, "lr": 2.9398527133391233e-05}
{"train_loss": 0.10287587344646454, "global_step": 172602, "epoch": 1939, "lr": 2.9397998895870614e-05}
{"train_loss": 0.0940534844994545, "global_step": 172603, "epoch": 1939, "lr": 2.939747066111966e-05}
{"train_loss": 0.045341819524765015, "global_step": 172604, "epoch": 1939, "lr": 2.9396942429138487e-05}
{"train_loss": 0.0659223198890686, "global_step": 172605, "epoch": 1939, "lr": 2.9396414199927118e-05}
{"train_loss": 0.06633403897285461, "global_step": 172606, "epoch": 1939, "lr": 2.9395885973485666e-05}
{"train_loss": 0.07683950662612915, "global_step": 172607, "epoch": 1939, "lr": 2.9395357749814177e-05}
{"train_loss": 0.08558236062526703, "global_step": 172608, "epoch": 1939, "lr": 2.9394829528912744e-05}
{"train_loss": 0.07318325340747833, "global_step": 172609, "epoch": 1939, "lr": 2.9394301310781413e-05}
{"train_loss": 0.12766002118587494, "global_step": 172610, "epoch": 1939, "lr": 2.9393773095420273e-05}
{"train_loss": 0.1209154725074768, "global_step": 172611, "epoch": 1939, "lr": 2.9393244882829402e-05}
{"train_loss": 0.06588149070739746, "global_step": 172612, "epoch": 1939, "lr": 2.9392716673008837e-05}
{"train_loss": 0.08926473557949066, "global_step": 172613, "epoch": 1939, "lr": 2.9392188465958682e-05}
{"train_loss": 0.07718678563833237, "global_step": 172614, "epoch": 1939, "lr": 2.9391660261678987e-05}
{"train_loss": 0.10804714262485504, "global_step": 172615, "epoch": 1939, "lr": 2.9391132060169847e-05}
{"train_loss": 0.08600864559412003, "global_step": 172616, "epoch": 1939, "lr": 2.93906038614313e-05}
{"train_loss": 0.07395169138908386, "global_step": 172617, "epoch": 1939, "lr": 2.9390075665463446e-05}
{"train_loss": 0.0720653235912323, "global_step": 172618, "epoch": 1939, "lr": 2.9389547472266332e-05}
{"train_loss": 0.11865097284317017, "global_step": 172619, "epoch": 1939, "lr": 2.938901928184007e-05}
{"train_loss": 0.06167047098278999, "global_step": 172620, "epoch": 1939, "lr": 2.9388491094184667e-05}
{"train_loss": 0.06521248072385788, "global_step": 172621, "epoch": 1939, "lr": 2.938796290930026e-05}
{"train_loss": 0.0794573500752449, "global_step": 172622, "epoch": 1939, "lr": 2.9387434727186868e-05}
{"train_loss": 0.08277341723442078, "global_step": 172623, "epoch": 1939, "lr": 2.9386906547844604e-05}
{"train_loss": 0.08904896676540375, "global_step": 172624, "epoch": 1939, "lr": 2.9386378371273504e-05}
{"train_loss": 0.16993699967861176, "global_step": 172625, "epoch": 1939, "lr": 2.9385850197473662e-05}
{"train_loss": 0.11058809608221054, "global_step": 172626, "epoch": 1939, "lr": 2.938532202644514e-05}
{"train_loss": 0.07935063540935516, "global_step": 172627, "epoch": 1939, "lr": 2.938479385818801e-05}
{"train_loss": 0.07619098573923111, "global_step": 172628, "epoch": 1939, "lr": 2.9384265692702338e-05}
{"train_loss": 0.046497080475091934, "global_step": 172629, "epoch": 1939, "lr": 2.938373752998821e-05}
{"train_loss": 0.06948627531528473, "global_step": 172630, "epoch": 1939, "lr": 2.9383209370045694e-05}
{"train_loss": 0.043567486107349396, "global_step": 172631, "epoch": 1939, "lr": 2.938268121287483e-05}
{"train_loss": 0.09160672128200531, "global_step": 172632, "epoch": 1939, "lr": 2.9382153058475735e-05}
{"train_loss": 0.050614453852176666, "global_step": 172633, "epoch": 1939, "lr": 2.9381624906848438e-05}
{"train_loss": 0.07392745465040207, "global_step": 172634, "epoch": 1939, "lr": 2.9381096757993055e-05}
{"train_loss": 0.11253616213798523, "global_step": 172635, "epoch": 1939, "lr": 2.938056861190961e-05}
{"train_loss": 0.10896052420139313, "global_step": 172636, "epoch": 1939, "lr": 2.9380040468598213e-05}
{"train_loss": 0.13300031423568726, "global_step": 172637, "epoch": 1939, "lr": 2.93795123280589e-05}
{"train_loss": 0.06727396696805954, "global_step": 172638, "epoch": 1939, "lr": 2.937898419029178e-05}
{"train_loss": 0.09479391574859619, "global_step": 172639, "epoch": 1939, "lr": 2.9378456055296882e-05}
{"train_loss": 0.0860392302274704, "global_step": 172640, "epoch": 1939, "lr": 2.9377927923074323e-05}
{"train_loss": 0.1238308772444725, "global_step": 172641, "epoch": 1939, "lr": 2.9377399793624127e-05}
{"train_loss": 0.0680641233921051, "global_step": 172642, "epoch": 1939, "lr": 2.9376871666946394e-05}
{"train_loss": 0.0620831698179245, "global_step": 172643, "epoch": 1939, "lr": 2.9376343543041208e-05}
{"train_loss": 0.15552587807178497, "global_step": 172644, "epoch": 1939, "lr": 2.9375815421908604e-05}
{"train_loss": 0.08502906560897827, "global_step": 172645, "epoch": 1939, "lr": 2.937528730354867e-05}
{"train_loss": 0.04442095011472702, "global_step": 172646, "epoch": 1939, "lr": 2.9374759187961486e-05}
{"train_loss": 0.09005588293075562, "global_step": 172647, "epoch": 1939, "lr": 2.937423107514711e-05}
{"train_loss": 0.11124169081449509, "global_step": 172648, "epoch": 1939, "lr": 2.9373702965105622e-05}
{"train_loss": 0.049911510199308395, "global_step": 172649, "epoch": 1939, "lr": 2.9373174857837094e-05}
{"train_loss": 0.06660383194684982, "global_step": 172650, "epoch": 1939, "lr": 2.9372646753341572e-05}
{"train_loss": 0.1112903580069542, "global_step": 172651, "epoch": 1939, "lr": 2.9372118651619162e-05}
{"train_loss": 0.06219375133514404, "global_step": 172652, "epoch": 1939, "lr": 2.9371590552669904e-05}
{"train_loss": 0.06326030194759369, "global_step": 172653, "epoch": 1939, "lr": 2.9371062456493902e-05}
{"train_loss": 0.07337825000286102, "global_step": 172654, "epoch": 1939, "lr": 2.937053436309119e-05}
{"train_loss": 0.15872961282730103, "global_step": 172655, "epoch": 1939, "lr": 2.937000627246188e-05}
{"train_loss": 0.0686536505818367, "global_step": 172656, "epoch": 1939, "lr": 2.9369478184605996e-05}
{"train_loss": 0.0879552811384201, "global_step": 172657, "epoch": 1939, "lr": 2.9368950099523664e-05}
{"train_loss": 0.21057850122451782, "global_step": 172658, "epoch": 1939, "lr": 2.9368422017214893e-05}
{"train_loss": 0.0888226092112868, "global_step": 172659, "epoch": 1939, "lr": 2.9367893937679813e-05, "val_loss": 7.249932289123535}
{"train_loss": 0.0920504704117775, "global_step": 172660, "epoch": 1940, "lr": 2.936736586091845e-05}
{"train_loss": 0.09686992317438126, "global_step": 172661, "epoch": 1940, "lr": 2.936683778693089e-05}
{"train_loss": 0.060650747269392014, "global_step": 172662, "epoch": 1940, "lr": 2.936630971571722e-05}
{"train_loss": 0.0675860345363617, "global_step": 172663, "epoch": 1940, "lr": 2.93657816472775e-05}
{"train_loss": 0.11734580248594284, "global_step": 172664, "epoch": 1940, "lr": 2.9365253581611795e-05}
{"train_loss": 0.05028710141777992, "global_step": 172665, "epoch": 1940, "lr": 2.936472551872018e-05}
{"train_loss": 0.10326813906431198, "global_step": 172666, "epoch": 1940, "lr": 2.9364197458602727e-05}
{"train_loss": 0.13590992987155914, "global_step": 172667, "epoch": 1940, "lr": 2.9363669401259507e-05}
{"train_loss": 0.10594984889030457, "global_step": 172668, "epoch": 1940, "lr": 2.936314134669059e-05}
{"train_loss": 0.16109098494052887, "global_step": 172669, "epoch": 1940, "lr": 2.936261329489604e-05}
{"train_loss": 0.10085131973028183, "global_step": 172670, "epoch": 1940, "lr": 2.9362085245875952e-05}
{"train_loss": 0.12664560973644257, "global_step": 172671, "epoch": 1940, "lr": 2.9361557199630364e-05}
{"train_loss": 0.06418630480766296, "global_step": 172672, "epoch": 1940, "lr": 2.936102915615937e-05}
{"train_loss": 0.08772516995668411, "global_step": 172673, "epoch": 1940, "lr": 2.9360501115463022e-05}
{"train_loss": 0.13532447814941406, "global_step": 172674, "epoch": 1940, "lr": 2.9359973077541424e-05}
{"train_loss": 0.10467103123664856, "global_step": 172675, "epoch": 1940, "lr": 2.9359445042394607e-05}
{"train_loss": 0.1462932676076889, "global_step": 172676, "epoch": 1940, "lr": 2.935891701002268e-05}
{"train_loss": 0.07506278157234192, "global_step": 172677, "epoch": 1940, "lr": 2.935838898042568e-05}
{"train_loss": 0.11010406166315079, "global_step": 172678, "epoch": 1940, "lr": 2.9357860953603678e-05}
{"train_loss": 0.06749638170003891, "global_step": 172679, "epoch": 1940, "lr": 2.935733292955679e-05}
{"train_loss": 0.06681787967681885, "global_step": 172680, "epoch": 1940, "lr": 2.935680490828504e-05}
{"train_loss": 0.05499453842639923, "global_step": 172681, "epoch": 1940, "lr": 2.935627688978853e-05}
{"train_loss": 0.12631507217884064, "global_step": 172682, "epoch": 1940, "lr": 2.9355748874067314e-05}
{"train_loss": 0.08322804421186447, "global_step": 172683, "epoch": 1940, "lr": 2.9355220861121458e-05}
{"train_loss": 0.11404091119766235, "global_step": 172684, "epoch": 1940, "lr": 2.9354692850951044e-05}
{"train_loss": 0.11665765196084976, "global_step": 172685, "epoch": 1940, "lr": 2.9354164843556144e-05}
{"train_loss": 0.08464702218770981, "global_step": 172686, "epoch": 1940, "lr": 2.9353636838936815e-05}
{"train_loss": 0.09150736033916473, "global_step": 172687, "epoch": 1940, "lr": 2.9353108837093145e-05}
{"train_loss": 0.053000446408987045, "global_step": 172688, "epoch": 1940, "lr": 2.9352580838025195e-05}
{"train_loss": 0.0937756597995758, "global_step": 172689, "epoch": 1940, "lr": 2.935205284173305e-05}
{"train_loss": 0.0672723725438118, "global_step": 172690, "epoch": 1940, "lr": 2.9351524848216748e-05}
{"train_loss": 0.06624016165733337, "global_step": 172691, "epoch": 1940, "lr": 2.935099685747641e-05}
{"train_loss": 0.05313274636864662, "global_step": 172692, "epoch": 1940, "lr": 2.935046886951205e-05}
{"train_loss": 0.08460693061351776, "global_step": 172693, "epoch": 1940, "lr": 2.934994088432379e-05}
{"train_loss": 0.08620632439851761, "global_step": 172694, "epoch": 1940, "lr": 2.934941290191166e-05}
{"train_loss": 0.09359381347894669, "global_step": 172695, "epoch": 1940, "lr": 2.9348884922275764e-05}
{"train_loss": 0.10039091855287552, "global_step": 172696, "epoch": 1940, "lr": 2.934835694541615e-05}
{"train_loss": 0.20441336929798126, "global_step": 172697, "epoch": 1940, "lr": 2.934782897133289e-05}
{"train_loss": 0.09244001656770706, "global_step": 172698, "epoch": 1940, "lr": 2.9347301000026084e-05}
{"train_loss": 0.0838891789317131, "global_step": 172699, "epoch": 1940, "lr": 2.9346773031495766e-05}
{"train_loss": 0.05881691351532936, "global_step": 172700, "epoch": 1940, "lr": 2.934624506574204e-05}
{"train_loss": 0.06605005264282227, "global_step": 172701, "epoch": 1940, "lr": 2.9345717102764936e-05}
{"train_loss": 0.07236830145120621, "global_step": 172702, "epoch": 1940, "lr": 2.9345189142564567e-05}
{"train_loss": 0.07724906504154205, "global_step": 172703, "epoch": 1940, "lr": 2.9344661185140982e-05}
{"train_loss": 0.027155151590704918, "global_step": 172704, "epoch": 1940, "lr": 2.9344133230494253e-05}
{"train_loss": 0.07552323490381241, "global_step": 172705, "epoch": 1940, "lr": 2.9343605278624453e-05}
{"train_loss": 0.10694485157728195, "global_step": 172706, "epoch": 1940, "lr": 2.9343077329531653e-05}
{"train_loss": 0.10239863395690918, "global_step": 172707, "epoch": 1940, "lr": 2.9342549383215923e-05}
{"train_loss": 0.10997000336647034, "global_step": 172708, "epoch": 1940, "lr": 2.9342021439677346e-05}
{"train_loss": 0.10495231300592422, "global_step": 172709, "epoch": 1940, "lr": 2.9341493498915966e-05}
{"train_loss": 0.03707846254110336, "global_step": 172710, "epoch": 1940, "lr": 2.934096556093189e-05}
{"train_loss": 0.07060261815786362, "global_step": 172711, "epoch": 1940, "lr": 2.9340437625725148e-05}
{"train_loss": 0.05032571777701378, "global_step": 172712, "epoch": 1940, "lr": 2.9339909693295852e-05}
{"train_loss": 0.058048542588949203, "global_step": 172713, "epoch": 1940, "lr": 2.9339381763644035e-05}
{"train_loss": 0.12679319083690643, "global_step": 172714, "epoch": 1940, "lr": 2.9338853836769787e-05}
{"train_loss": 0.11337529867887497, "global_step": 172715, "epoch": 1940, "lr": 2.933832591267319e-05}
{"train_loss": 0.06991682946681976, "global_step": 172716, "epoch": 1940, "lr": 2.9337797991354294e-05}
{"train_loss": 0.0716637670993805, "global_step": 172717, "epoch": 1940, "lr": 2.9337270072813194e-05}
{"train_loss": 0.05575395002961159, "global_step": 172718, "epoch": 1940, "lr": 2.933674215704993e-05}
{"train_loss": 0.12294196337461472, "global_step": 172719, "epoch": 1940, "lr": 2.933621424406461e-05}
{"train_loss": 0.09390006959438324, "global_step": 172720, "epoch": 1940, "lr": 2.9335686333857253e-05}
{"train_loss": 0.09695825725793839, "global_step": 172721, "epoch": 1940, "lr": 2.933515842642799e-05}
{"train_loss": 0.09416897594928741, "global_step": 172722, "epoch": 1940, "lr": 2.933463052177685e-05}
{"train_loss": 0.07216492295265198, "global_step": 172723, "epoch": 1940, "lr": 2.933410261990392e-05}
{"train_loss": 0.0844544842839241, "global_step": 172724, "epoch": 1940, "lr": 2.9333574720809265e-05}
{"train_loss": 0.08473236113786697, "global_step": 172725, "epoch": 1940, "lr": 2.933304682449296e-05}
{"train_loss": 0.13462644815444946, "global_step": 172726, "epoch": 1940, "lr": 2.9332518930955078e-05}
{"train_loss": 0.06358786672353745, "global_step": 172727, "epoch": 1940, "lr": 2.9331991040195693e-05}
{"train_loss": 0.055053308606147766, "global_step": 172728, "epoch": 1940, "lr": 2.9331463152214854e-05}
{"train_loss": 0.092107392847538, "global_step": 172729, "epoch": 1940, "lr": 2.933093526701266e-05}
{"train_loss": 0.12728255987167358, "global_step": 172730, "epoch": 1940, "lr": 2.933040738458916e-05}
{"train_loss": 0.029572853818535805, "global_step": 172731, "epoch": 1940, "lr": 2.9329879504944445e-05}
{"train_loss": 0.07121846079826355, "global_step": 172732, "epoch": 1940, "lr": 2.932935162807856e-05}
{"train_loss": 0.09790966659784317, "global_step": 172733, "epoch": 1940, "lr": 2.9328823753991598e-05}
{"train_loss": 0.08255736529827118, "global_step": 172734, "epoch": 1940, "lr": 2.9328295882683637e-05}
{"train_loss": 0.05769277736544609, "global_step": 172735, "epoch": 1940, "lr": 2.9327768014154717e-05}
{"train_loss": 0.08974748104810715, "global_step": 172736, "epoch": 1940, "lr": 2.9327240148404945e-05}
{"train_loss": 0.08460825681686401, "global_step": 172737, "epoch": 1940, "lr": 2.9326712285434353e-05}
{"train_loss": 0.08649251610040665, "global_step": 172738, "epoch": 1940, "lr": 2.9326184425243052e-05}
{"train_loss": 0.08480256795883179, "global_step": 172739, "epoch": 1940, "lr": 2.9325656567831073e-05}
{"train_loss": 0.08580835908651352, "global_step": 172740, "epoch": 1940, "lr": 2.9325128713198534e-05}
{"train_loss": 0.10263825207948685, "global_step": 172741, "epoch": 1940, "lr": 2.9324600861345463e-05}
{"train_loss": 0.11895863711833954, "global_step": 172742, "epoch": 1940, "lr": 2.9324073012271948e-05}
{"train_loss": 0.07093919813632965, "global_step": 172743, "epoch": 1940, "lr": 2.9323545165978062e-05}
{"train_loss": 0.05618317797780037, "global_step": 172744, "epoch": 1940, "lr": 2.9323017322463874e-05}
{"train_loss": 0.0383959524333477, "global_step": 172745, "epoch": 1940, "lr": 2.9322489481729452e-05}
{"train_loss": 0.10565381497144699, "global_step": 172746, "epoch": 1940, "lr": 2.9321961643774877e-05}
{"train_loss": 0.1327970325946808, "global_step": 172747, "epoch": 1940, "lr": 2.9321433808600197e-05}
{"train_loss": 0.08858666313665636, "global_step": 172748, "epoch": 1940, "lr": 2.9320905976205515e-05, "val_loss": 7.171473979949951, "train_action_mse_error": 7.140137672424316}
{"train_loss": 0.09035839140415192, "global_step": 172749, "epoch": 1941, "lr": 2.9320378146590867e-05}
{"train_loss": 0.07385865598917007, "global_step": 172750, "epoch": 1941, "lr": 2.9319850319756357e-05}
{"train_loss": 0.0765267014503479, "global_step": 172751, "epoch": 1941, "lr": 2.931932249570203e-05}
{"train_loss": 0.0745181068778038, "global_step": 172752, "epoch": 1941, "lr": 2.9318794674427963e-05}
{"train_loss": 0.05522551015019417, "global_step": 172753, "epoch": 1941, "lr": 2.931826685593425e-05}
{"train_loss": 0.08540711551904678, "global_step": 172754, "epoch": 1941, "lr": 2.9317739040220925e-05}
{"train_loss": 0.09323155879974365, "global_step": 172755, "epoch": 1941, "lr": 2.93172112272881e-05}
{"train_loss": 0.09481494128704071, "global_step": 172756, "epoch": 1941, "lr": 2.9316683417135805e-05}
{"train_loss": 0.06452298164367676, "global_step": 172757, "epoch": 1941, "lr": 2.931615560976415e-05}
{"train_loss": 0.06940538436174393, "global_step": 172758, "epoch": 1941, "lr": 2.9315627805173163e-05}
{"train_loss": 0.07826271653175354, "global_step": 172759, "epoch": 1941, "lr": 2.931510000336296e-05}
{"train_loss": 0.08490028232336044, "global_step": 172760, "epoch": 1941, "lr": 2.931457220433358e-05}
{"train_loss": 0.10838214308023453, "global_step": 172761, "epoch": 1941, "lr": 2.9314044408085097e-05}
{"train_loss": 0.09201981872320175, "global_step": 172762, "epoch": 1941, "lr": 2.9313516614617596e-05}
{"train_loss": 0.0808732658624649, "global_step": 172763, "epoch": 1941, "lr": 2.9312988823931136e-05}
{"train_loss": 0.09837367385625839, "global_step": 172764, "epoch": 1941, "lr": 2.9312461036025796e-05}
{"train_loss": 0.10477328300476074, "global_step": 172765, "epoch": 1941, "lr": 2.9311933250901652e-05}
{"train_loss": 0.06968776136636734, "global_step": 172766, "epoch": 1941, "lr": 2.9311405468558744e-05}
{"train_loss": 0.07137774676084518, "global_step": 172767, "epoch": 1941, "lr": 2.931087768899719e-05}
{"train_loss": 0.12155502289533615, "global_step": 172768, "epoch": 1941, "lr": 2.9310349912217015e-05}
{"train_loss": 0.13255208730697632, "global_step": 172769, "epoch": 1941, "lr": 2.9309822138218313e-05}
{"train_loss": 0.10508053749799728, "global_step": 172770, "epoch": 1941, "lr": 2.9309294367001166e-05}
{"train_loss": 0.06378384679555893, "global_step": 172771, "epoch": 1941, "lr": 2.9308766598565617e-05}
{"train_loss": 0.1548001766204834, "global_step": 172772, "epoch": 1941, "lr": 2.9308238832911772e-05}
{"train_loss": 0.03947633504867554, "global_step": 172773, "epoch": 1941, "lr": 2.930771107003966e-05}
{"train_loss": 0.07972721755504608, "global_step": 172774, "epoch": 1941, "lr": 2.93071833099494e-05}
{"train_loss": 0.10481566935777664, "global_step": 172775, "epoch": 1941, "lr": 2.930665555264101e-05}
{"train_loss": 0.10309290885925293, "global_step": 172776, "epoch": 1941, "lr": 2.9306127798114614e-05}
{"train_loss": 0.04836905002593994, "global_step": 172777, "epoch": 1941, "lr": 2.930560004637023e-05}
{"train_loss": 0.09680486470460892, "global_step": 172778, "epoch": 1941, "lr": 2.930507229740798e-05}
{"train_loss": 0.12426163256168365, "global_step": 172779, "epoch": 1941, "lr": 2.9304544551227904e-05}
{"train_loss": 0.07445777207612991, "global_step": 172780, "epoch": 1941, "lr": 2.930401680783007e-05}
{"train_loss": 0.13967055082321167, "global_step": 172781, "epoch": 1941, "lr": 2.9303489067214573e-05}
{"train_loss": 0.08638092875480652, "global_step": 172782, "epoch": 1941, "lr": 2.930296132938146e-05}
{"train_loss": 0.07976938039064407, "global_step": 172783, "epoch": 1941, "lr": 2.9302433594330814e-05}
{"train_loss": 0.1012900173664093, "global_step": 172784, "epoch": 1941, "lr": 2.9301905862062705e-05}
{"train_loss": 0.05217742547392845, "global_step": 172785, "epoch": 1941, "lr": 2.9301378132577194e-05}
{"train_loss": 0.10136951506137848, "global_step": 172786, "epoch": 1941, "lr": 2.930085040587438e-05}
{"train_loss": 0.09353835135698318, "global_step": 172787, "epoch": 1941, "lr": 2.930032268195429e-05}
{"train_loss": 0.06779786199331284, "global_step": 172788, "epoch": 1941, "lr": 2.9299794960817018e-05}
{"train_loss": 0.09437041729688644, "global_step": 172789, "epoch": 1941, "lr": 2.9299267242462658e-05}
{"train_loss": 0.04204300045967102, "global_step": 172790, "epoch": 1941, "lr": 2.929873952689124e-05}
{"train_loss": 0.08927527815103531, "global_step": 172791, "epoch": 1941, "lr": 2.929821181410287e-05}
{"train_loss": 0.04627401381731033, "global_step": 172792, "epoch": 1941, "lr": 2.929768410409759e-05}
{"train_loss": 0.1207277923822403, "global_step": 172793, "epoch": 1941, "lr": 2.9297156396875503e-05}
{"train_loss": 0.06231511011719704, "global_step": 172794, "epoch": 1941, "lr": 2.929662869243664e-05}
{"train_loss": 0.11167199164628983, "global_step": 172795, "epoch": 1941, "lr": 2.9296100990781112e-05}
{"train_loss": 0.10338347405195236, "global_step": 172796, "epoch": 1941, "lr": 2.929557329190895e-05}
{"train_loss": 0.05651390925049782, "global_step": 172797, "epoch": 1941, "lr": 2.929504559582027e-05}
{"train_loss": 0.12117625027894974, "global_step": 172798, "epoch": 1941, "lr": 2.9294517902515105e-05}
{"train_loss": 0.06897863000631332, "global_step": 172799, "epoch": 1941, "lr": 2.929399021199354e-05}
{"train_loss": 0.10592225193977356, "global_step": 172800, "epoch": 1941, "lr": 2.9293462524255643e-05}
{"train_loss": 0.06452218443155289, "global_step": 172801, "epoch": 1941, "lr": 2.929293483930149e-05}
{"train_loss": 0.059995315968990326, "global_step": 172802, "epoch": 1941, "lr": 2.929240715713115e-05}
{"train_loss": 0.02698408253490925, "global_step": 172803, "epoch": 1941, "lr": 2.9291879477744694e-05}
{"train_loss": 0.08391374349594116, "global_step": 172804, "epoch": 1941, "lr": 2.929135180114219e-05}
{"train_loss": 0.078269824385643, "global_step": 172805, "epoch": 1941, "lr": 2.9290824127323702e-05}
{"train_loss": 0.10996556282043457, "global_step": 172806, "epoch": 1941, "lr": 2.9290296456289336e-05}
{"train_loss": 0.056520335376262665, "global_step": 172807, "epoch": 1941, "lr": 2.9289768788039113e-05}
{"train_loss": 0.1037793830037117, "global_step": 172808, "epoch": 1941, "lr": 2.9289241122573148e-05}
{"train_loss": 0.09992306679487228, "global_step": 172809, "epoch": 1941, "lr": 2.9288713459891472e-05}
{"train_loss": 0.07530411332845688, "global_step": 172810, "epoch": 1941, "lr": 2.9288185799994193e-05}
{"train_loss": 0.09324801713228226, "global_step": 172811, "epoch": 1941, "lr": 2.9287658142881345e-05}
{"train_loss": 0.10830756276845932, "global_step": 172812, "epoch": 1941, "lr": 2.928713048855305e-05}
{"train_loss": 0.08231860399246216, "global_step": 172813, "epoch": 1941, "lr": 2.9286602837009312e-05}
{"train_loss": 0.13127806782722473, "global_step": 172814, "epoch": 1941, "lr": 2.9286075188250273e-05}
{"train_loss": 0.06401539593935013, "global_step": 172815, "epoch": 1941, "lr": 2.928554754227594e-05}
{"train_loss": 0.13060727715492249, "global_step": 172816, "epoch": 1941, "lr": 2.928501989908643e-05}
{"train_loss": 0.07381188124418259, "global_step": 172817, "epoch": 1941, "lr": 2.9284492258681784e-05}
{"train_loss": 0.07819236814975739, "global_step": 172818, "epoch": 1941, "lr": 2.92839646210621e-05}
{"train_loss": 0.0916040688753128, "global_step": 172819, "epoch": 1941, "lr": 2.928343698622743e-05}
{"train_loss": 0.06263355910778046, "global_step": 172820, "epoch": 1941, "lr": 2.9282909354177847e-05}
{"train_loss": 0.04340718314051628, "global_step": 172821, "epoch": 1941, "lr": 2.9282381724913423e-05}
{"train_loss": 0.1517065316438675, "global_step": 172822, "epoch": 1941, "lr": 2.9281854098434226e-05}
{"train_loss": 0.06042497232556343, "global_step": 172823, "epoch": 1941, "lr": 2.9281326474740335e-05}
{"train_loss": 0.09250260144472122, "global_step": 172824, "epoch": 1941, "lr": 2.928079885383181e-05}
{"train_loss": 0.09339108318090439, "global_step": 172825, "epoch": 1941, "lr": 2.9280271235708746e-05}
{"train_loss": 0.09927532821893692, "global_step": 172826, "epoch": 1941, "lr": 2.927974362037118e-05}
{"train_loss": 0.05090351775288582, "global_step": 172827, "epoch": 1941, "lr": 2.9279216007819215e-05}
{"train_loss": 0.05141695216298103, "global_step": 172828, "epoch": 1941, "lr": 2.927868839805289e-05}
{"train_loss": 0.07964682579040527, "global_step": 172829, "epoch": 1941, "lr": 2.9278160791072314e-05}
{"train_loss": 0.12399096041917801, "global_step": 172830, "epoch": 1941, "lr": 2.9277633186877517e-05}
{"train_loss": 0.06893432885408401, "global_step": 172831, "epoch": 1941, "lr": 2.9277105585468606e-05}
{"train_loss": 0.03146044537425041, "global_step": 172832, "epoch": 1941, "lr": 2.9276577986845616e-05}
{"train_loss": 0.054221656173467636, "global_step": 172833, "epoch": 1941, "lr": 2.927605039100866e-05}
{"train_loss": 0.14562976360321045, "global_step": 172834, "epoch": 1941, "lr": 2.9275522797957767e-05}
{"train_loss": 0.07153360545635223, "global_step": 172835, "epoch": 1941, "lr": 2.9274995207693044e-05}
{"train_loss": 0.09054314345121384, "global_step": 172836, "epoch": 1941, "lr": 2.927446762021453e-05}
{"train_loss": 0.0857833850124244, "global_step": 172837, "epoch": 1941, "lr": 2.9273940035522323e-05, "val_loss": 7.157495975494385}
{"train_loss": 0.08045724034309387, "global_step": 172838, "epoch": 1942, "lr": 2.9273412453616477e-05}
{"train_loss": 0.08774177730083466, "global_step": 172839, "epoch": 1942, "lr": 2.927288487449707e-05}
{"train_loss": 0.0638182982802391, "global_step": 172840, "epoch": 1942, "lr": 2.9272357298164166e-05}
{"train_loss": 0.03310910612344742, "global_step": 172841, "epoch": 1942, "lr": 2.9271829724617837e-05}
{"train_loss": 0.07945907860994339, "global_step": 172842, "epoch": 1942, "lr": 2.9271302153858176e-05}
{"train_loss": 0.10466461628675461, "global_step": 172843, "epoch": 1942, "lr": 2.927077458588522e-05}
{"train_loss": 0.11327864974737167, "global_step": 172844, "epoch": 1942, "lr": 2.9270247020699064e-05}
{"train_loss": 0.0670805349946022, "global_step": 172845, "epoch": 1942, "lr": 2.926971945829976e-05}
{"train_loss": 0.058794379234313965, "global_step": 172846, "epoch": 1942, "lr": 2.9269191898687408e-05}
{"train_loss": 0.08920171856880188, "global_step": 172847, "epoch": 1942, "lr": 2.926866434186204e-05}
{"train_loss": 0.05129401758313179, "global_step": 172848, "epoch": 1942, "lr": 2.9268136787823762e-05}
{"train_loss": 0.06188042461872101, "global_step": 172849, "epoch": 1942, "lr": 2.9267609236572618e-05}
{"train_loss": 0.03751441836357117, "global_step": 172850, "epoch": 1942, "lr": 2.926708168810871e-05}
{"train_loss": 0.06859243661165237, "global_step": 172851, "epoch": 1942, "lr": 2.926655414243207e-05}
{"train_loss": 0.05460990220308304, "global_step": 172852, "epoch": 1942, "lr": 2.9266026599542807e-05}
{"train_loss": 0.11661356687545776, "global_step": 172853, "epoch": 1942, "lr": 2.9265499059440947e-05}
{"train_loss": 0.05311998352408409, "global_step": 172854, "epoch": 1942, "lr": 2.9264971522126615e-05}
{"train_loss": 0.11116372048854828, "global_step": 172855, "epoch": 1942, "lr": 2.9264443987599842e-05}
{"train_loss": 0.11407061666250229, "global_step": 172856, "epoch": 1942, "lr": 2.9263916455860725e-05}
{"train_loss": 0.12742570042610168, "global_step": 172857, "epoch": 1942, "lr": 2.926338892690931e-05}
{"train_loss": 0.07423855364322662, "global_step": 172858, "epoch": 1942, "lr": 2.9262861400745677e-05}
{"train_loss": 0.10603592544794083, "global_step": 172859, "epoch": 1942, "lr": 2.9262333877369906e-05}
{"train_loss": 0.1304663121700287, "global_step": 172860, "epoch": 1942, "lr": 2.9261806356782053e-05}
{"train_loss": 0.042707812041044235, "global_step": 172861, "epoch": 1942, "lr": 2.9261278838982208e-05}
{"train_loss": 0.0973442792892456, "global_step": 172862, "epoch": 1942, "lr": 2.9260751323970426e-05}
{"train_loss": 0.09516657143831253, "global_step": 172863, "epoch": 1942, "lr": 2.926022381174679e-05}
{"train_loss": 0.08772194385528564, "global_step": 172864, "epoch": 1942, "lr": 2.9259696302311347e-05}
{"train_loss": 0.09932751208543777, "global_step": 172865, "epoch": 1942, "lr": 2.9259168795664206e-05}
{"train_loss": 0.05861542746424675, "global_step": 172866, "epoch": 1942, "lr": 2.92586412918054e-05}
{"train_loss": 0.06536518037319183, "global_step": 172867, "epoch": 1942, "lr": 2.925811379073503e-05}
{"train_loss": 0.0715608149766922, "global_step": 172868, "epoch": 1942, "lr": 2.9257586292453142e-05}
{"train_loss": 0.08371373265981674, "global_step": 172869, "epoch": 1942, "lr": 2.9257058796959836e-05}
{"train_loss": 0.11133845150470734, "global_step": 172870, "epoch": 1942, "lr": 2.925653130425514e-05}
{"train_loss": 0.14807304739952087, "global_step": 172871, "epoch": 1942, "lr": 2.9256003814339177e-05}
{"train_loss": 0.07999975979328156, "global_step": 172872, "epoch": 1942, "lr": 2.9255476327211974e-05}
{"train_loss": 0.05568500608205795, "global_step": 172873, "epoch": 1942, "lr": 2.925494884287363e-05}
{"train_loss": 0.1247267946600914, "global_step": 172874, "epoch": 1942, "lr": 2.925442136132419e-05}
{"train_loss": 0.12560902535915375, "global_step": 172875, "epoch": 1942, "lr": 2.9253893882563766e-05}
{"train_loss": 0.1126362532377243, "global_step": 172876, "epoch": 1942, "lr": 2.9253366406592387e-05}
{"train_loss": 0.03533698618412018, "global_step": 172877, "epoch": 1942, "lr": 2.9252838933410138e-05}
{"train_loss": 0.08080383390188217, "global_step": 172878, "epoch": 1942, "lr": 2.925231146301709e-05}
{"train_loss": 0.07798178493976593, "global_step": 172879, "epoch": 1942, "lr": 2.9251783995413313e-05}
{"train_loss": 0.1006808802485466, "global_step": 172880, "epoch": 1942, "lr": 2.9251256530598896e-05}
{"train_loss": 0.06312741339206696, "global_step": 172881, "epoch": 1942, "lr": 2.925072906857388e-05}
{"train_loss": 0.07194021344184875, "global_step": 172882, "epoch": 1942, "lr": 2.9250201609338368e-05}
{"train_loss": 0.08627907931804657, "global_step": 172883, "epoch": 1942, "lr": 2.9249674152892392e-05}
{"train_loss": 0.08004774153232574, "global_step": 172884, "epoch": 1942, "lr": 2.9249146699236062e-05}
{"train_loss": 0.06947803497314453, "global_step": 172885, "epoch": 1942, "lr": 2.9248619248369415e-05}
{"train_loss": 0.08886395394802094, "global_step": 172886, "epoch": 1942, "lr": 2.924809180029256e-05}
{"train_loss": 0.0659794881939888, "global_step": 172887, "epoch": 1942, "lr": 2.9247564355005517e-05}
{"train_loss": 0.05517415329813957, "global_step": 172888, "epoch": 1942, "lr": 2.924703691250842e-05}
{"train_loss": 0.14293958246707916, "global_step": 172889, "epoch": 1942, "lr": 2.9246509472801275e-05}
{"train_loss": 0.16508831083774567, "global_step": 172890, "epoch": 1942, "lr": 2.924598203588421e-05}
{"train_loss": 0.14028804004192352, "global_step": 172891, "epoch": 1942, "lr": 2.9245454601757248e-05}
{"train_loss": 0.06256366521120071, "global_step": 172892, "epoch": 1942, "lr": 2.92449271704205e-05}
{"train_loss": 0.07617222517728806, "global_step": 172893, "epoch": 1942, "lr": 2.9244399741873995e-05}
{"train_loss": 0.0742206871509552, "global_step": 172894, "epoch": 1942, "lr": 2.924387231611786e-05}
{"train_loss": 0.10912273824214935, "global_step": 172895, "epoch": 1942, "lr": 2.9243344893152115e-05}
{"train_loss": 0.10364692658185959, "global_step": 172896, "epoch": 1942, "lr": 2.9242817472976835e-05}
{"train_loss": 0.10870715230703354, "global_step": 172897, "epoch": 1942, "lr": 2.924229005559213e-05}
{"train_loss": 0.11956965178251266, "global_step": 172898, "epoch": 1942, "lr": 2.924176264099803e-05}
{"train_loss": 0.09712579101324081, "global_step": 172899, "epoch": 1942, "lr": 2.9241235229194637e-05}
{"train_loss": 0.03784206882119179, "global_step": 172900, "epoch": 1942, "lr": 2.9240707820181988e-05}
{"train_loss": 0.14991280436515808, "global_step": 172901, "epoch": 1942, "lr": 2.9240180413960194e-05}
{"train_loss": 0.07829270511865616, "global_step": 172902, "epoch": 1942, "lr": 2.923965301052928e-05}
{"train_loss": 0.049775395542383194, "global_step": 172903, "epoch": 1942, "lr": 2.9239125609889368e-05}
{"train_loss": 0.07437757402658463, "global_step": 172904, "epoch": 1942, "lr": 2.923859821204048e-05}
{"train_loss": 0.07429295033216476, "global_step": 172905, "epoch": 1942, "lr": 2.9238070816982727e-05}
{"train_loss": 0.08126718550920486, "global_step": 172906, "epoch": 1942, "lr": 2.923754342471614e-05}
{"train_loss": 0.19181707501411438, "global_step": 172907, "epoch": 1942, "lr": 2.9237016035240838e-05}
{"train_loss": 0.08343912661075592, "global_step": 172908, "epoch": 1942, "lr": 2.923648864855685e-05}
{"train_loss": 0.10157855600118637, "global_step": 172909, "epoch": 1942, "lr": 2.9235961264664273e-05}
{"train_loss": 0.02983897738158703, "global_step": 172910, "epoch": 1942, "lr": 2.9235433883563156e-05}
{"train_loss": 0.020977869629859924, "global_step": 172911, "epoch": 1942, "lr": 2.92349065052536e-05}
{"train_loss": 0.05473541095852852, "global_step": 172912, "epoch": 1942, "lr": 2.9234379129735634e-05}
{"train_loss": 0.02019641175866127, "global_step": 172913, "epoch": 1942, "lr": 2.923385175700938e-05}
{"train_loss": 0.07386549562215805, "global_step": 172914, "epoch": 1942, "lr": 2.9233324387074852e-05}
{"train_loss": 0.09949695318937302, "global_step": 172915, "epoch": 1942, "lr": 2.9232797019932163e-05}
{"train_loss": 0.07716300338506699, "global_step": 172916, "epoch": 1942, "lr": 2.9232269655581383e-05}
{"train_loss": 0.09378568083047867, "global_step": 172917, "epoch": 1942, "lr": 2.9231742294022553e-05}
{"train_loss": 0.07737938314676285, "global_step": 172918, "epoch": 1942, "lr": 2.923121493525578e-05}
{"train_loss": 0.08220706135034561, "global_step": 172919, "epoch": 1942, "lr": 2.9230687579281095e-05}
{"train_loss": 0.10080242902040482, "global_step": 172920, "epoch": 1942, "lr": 2.9230160226098614e-05}
{"train_loss": 0.05335129052400589, "global_step": 172921, "epoch": 1942, "lr": 2.922963287570837e-05}
{"train_loss": 0.10177437961101532, "global_step": 172922, "epoch": 1942, "lr": 2.9229105528110455e-05}
{"train_loss": 0.08390256017446518, "global_step": 172923, "epoch": 1942, "lr": 2.9228578183304923e-05}
{"train_loss": 0.04952887445688248, "global_step": 172924, "epoch": 1942, "lr": 2.9228050841291875e-05}
{"train_loss": 0.06068592146039009, "global_step": 172925, "epoch": 1942, "lr": 2.922752350207134e-05}
{"train_loss": 0.08391581766642212, "global_step": 172926, "epoch": 1942, "lr": 2.9226996165643438e-05, "val_loss": 7.218010902404785}
{"train_loss": 0.08272819221019745, "global_step": 172927, "epoch": 1943, "lr": 2.9226468832008187e-05}
{"train_loss": 0.04336369037628174, "global_step": 172928, "epoch": 1943, "lr": 2.9225941501165698e-05}
{"train_loss": 0.06566651165485382, "global_step": 172929, "epoch": 1943, "lr": 2.9225414173116016e-05}
{"train_loss": 0.10022953897714615, "global_step": 172930, "epoch": 1943, "lr": 2.9224886847859245e-05}
{"train_loss": 0.055226169526576996, "global_step": 172931, "epoch": 1943, "lr": 2.9224359525395417e-05}
{"train_loss": 0.09510287642478943, "global_step": 172932, "epoch": 1943, "lr": 2.9223832205724622e-05}
{"train_loss": 0.08196914196014404, "global_step": 172933, "epoch": 1943, "lr": 2.9223304888846935e-05}
{"train_loss": 0.104781374335289, "global_step": 172934, "epoch": 1943, "lr": 2.9222777574762417e-05}
{"train_loss": 0.0871821716427803, "global_step": 172935, "epoch": 1943, "lr": 2.9222250263471163e-05}
{"train_loss": 0.09058895707130432, "global_step": 172936, "epoch": 1943, "lr": 2.9221722954973197e-05}
{"train_loss": 0.0997048169374466, "global_step": 172937, "epoch": 1943, "lr": 2.9221195649268634e-05}
{"train_loss": 0.10394053161144257, "global_step": 172938, "epoch": 1943, "lr": 2.9220668346357517e-05}
{"train_loss": 0.030072100460529327, "global_step": 172939, "epoch": 1943, "lr": 2.9220141046239936e-05}
{"train_loss": 0.08477341383695602, "global_step": 172940, "epoch": 1943, "lr": 2.921961374891594e-05}
{"train_loss": 0.09690753370523453, "global_step": 172941, "epoch": 1943, "lr": 2.9219086454385635e-05}
{"train_loss": 0.11539899557828903, "global_step": 172942, "epoch": 1943, "lr": 2.921855916264905e-05}
{"train_loss": 0.11555387824773788, "global_step": 172943, "epoch": 1943, "lr": 2.9218031873706298e-05}
{"train_loss": 0.06289464235305786, "global_step": 172944, "epoch": 1943, "lr": 2.9217504587557408e-05}
{"train_loss": 0.05769987776875496, "global_step": 172945, "epoch": 1943, "lr": 2.9216977304202486e-05}
{"train_loss": 0.1327628642320633, "global_step": 172946, "epoch": 1943, "lr": 2.9216450023641574e-05}
{"train_loss": 0.0791347399353981, "global_step": 172947, "epoch": 1943, "lr": 2.9215922745874773e-05}
{"train_loss": 0.1015794426202774, "global_step": 172948, "epoch": 1943, "lr": 2.9215395470902117e-05}
{"train_loss": 0.1013241559267044, "global_step": 172949, "epoch": 1943, "lr": 2.921486819872372e-05}
{"train_loss": 0.08743942528963089, "global_step": 172950, "epoch": 1943, "lr": 2.921434092933961e-05}
{"train_loss": 0.06796111166477203, "global_step": 172951, "epoch": 1943, "lr": 2.9213813662749888e-05}
{"train_loss": 0.10131209343671799, "global_step": 172952, "epoch": 1943, "lr": 2.921328639895462e-05}
{"train_loss": 0.10456354916095734, "global_step": 172953, "epoch": 1943, "lr": 2.921275913795387e-05}
{"train_loss": 0.058151960372924805, "global_step": 172954, "epoch": 1943, "lr": 2.921223187974772e-05}
{"train_loss": 0.11070504784584045, "global_step": 172955, "epoch": 1943, "lr": 2.9211704624336213e-05}
{"train_loss": 0.0666116252541542, "global_step": 172956, "epoch": 1943, "lr": 2.9211177371719456e-05}
{"train_loss": 0.046401672065258026, "global_step": 172957, "epoch": 1943, "lr": 2.921065012189749e-05}
{"train_loss": 0.06047987565398216, "global_step": 172958, "epoch": 1943, "lr": 2.921012287487041e-05}
{"train_loss": 0.044022586196660995, "global_step": 172959, "epoch": 1943, "lr": 2.9209595630638264e-05}
{"train_loss": 0.06568024307489395, "global_step": 172960, "epoch": 1943, "lr": 2.9209068389201143e-05}
{"train_loss": 0.05682932212948799, "global_step": 172961, "epoch": 1943, "lr": 2.92085411505591e-05}
{"train_loss": 0.07320765405893326, "global_step": 172962, "epoch": 1943, "lr": 2.920801391471223e-05}
{"train_loss": 0.04271208494901657, "global_step": 172963, "epoch": 1943, "lr": 2.920748668166057e-05}
{"train_loss": 0.10078704357147217, "global_step": 172964, "epoch": 1943, "lr": 2.920695945140423e-05}
{"train_loss": 0.08880932629108429, "global_step": 172965, "epoch": 1943, "lr": 2.9206432223943242e-05}
{"train_loss": 0.1492437720298767, "global_step": 172966, "epoch": 1943, "lr": 2.9205904999277716e-05}
{"train_loss": 0.08306287229061127, "global_step": 172967, "epoch": 1943, "lr": 2.9205377777407683e-05}
{"train_loss": 0.123191699385643, "global_step": 172968, "epoch": 1943, "lr": 2.920485055833324e-05}
{"train_loss": 0.07793295383453369, "global_step": 172969, "epoch": 1943, "lr": 2.9204323342054453e-05}
{"train_loss": 0.05407204106450081, "global_step": 172970, "epoch": 1943, "lr": 2.920379612857139e-05}
{"train_loss": 0.055697258561849594, "global_step": 172971, "epoch": 1943, "lr": 2.920326891788413e-05}
{"train_loss": 0.05738076940178871, "global_step": 172972, "epoch": 1943, "lr": 2.920274170999273e-05}
{"train_loss": 0.05086103826761246, "global_step": 172973, "epoch": 1943, "lr": 2.920221450489728e-05}
{"train_loss": 0.05678239464759827, "global_step": 172974, "epoch": 1943, "lr": 2.9201687302597817e-05}
{"train_loss": 0.09323079138994217, "global_step": 172975, "epoch": 1943, "lr": 2.9201160103094462e-05}
{"train_loss": 0.04685668274760246, "global_step": 172976, "epoch": 1943, "lr": 2.9200632906387226e-05}
{"train_loss": 0.09433096647262573, "global_step": 172977, "epoch": 1943, "lr": 2.9200105712476233e-05}
{"train_loss": 0.09637729823589325, "global_step": 172978, "epoch": 1943, "lr": 2.919957852136152e-05}
{"train_loss": 0.08557546138763428, "global_step": 172979, "epoch": 1943, "lr": 2.9199051333043183e-05}
{"train_loss": 0.0894317477941513, "global_step": 172980, "epoch": 1943, "lr": 2.9198524147521268e-05}
{"train_loss": 0.09731408953666687, "global_step": 172981, "epoch": 1943, "lr": 2.919799696479587e-05}
{"train_loss": 0.05928069353103638, "global_step": 172982, "epoch": 1943, "lr": 2.9197469784867026e-05}
{"train_loss": 0.08346005529165268, "global_step": 172983, "epoch": 1943, "lr": 2.9196942607734857e-05}
{"train_loss": 0.051465634256601334, "global_step": 172984, "epoch": 1943, "lr": 2.919641543339938e-05}
{"train_loss": 0.08551295101642609, "global_step": 172985, "epoch": 1943, "lr": 2.9195888261860717e-05}
{"train_loss": 0.06013280898332596, "global_step": 172986, "epoch": 1943, "lr": 2.919536109311889e-05}
{"train_loss": 0.0884360522031784, "global_step": 172987, "epoch": 1943, "lr": 2.9194833927173998e-05}
{"train_loss": 0.16349457204341888, "global_step": 172988, "epoch": 1943, "lr": 2.9194306764026113e-05}
{"train_loss": 0.05419975519180298, "global_step": 172989, "epoch": 1943, "lr": 2.9193779603675297e-05}
{"train_loss": 0.04053879901766777, "global_step": 172990, "epoch": 1943, "lr": 2.919325244612162e-05}
{"train_loss": 0.08988465368747711, "global_step": 172991, "epoch": 1943, "lr": 2.9192725291365162e-05}
{"train_loss": 0.09043896943330765, "global_step": 172992, "epoch": 1943, "lr": 2.919219813940599e-05}
{"train_loss": 0.045689914375543594, "global_step": 172993, "epoch": 1943, "lr": 2.9191670990244164e-05}
{"train_loss": 0.07219184190034866, "global_step": 172994, "epoch": 1943, "lr": 2.9191143843879782e-05}
{"train_loss": 0.15543724596500397, "global_step": 172995, "epoch": 1943, "lr": 2.9190616700312877e-05}
{"train_loss": 0.10023801773786545, "global_step": 172996, "epoch": 1943, "lr": 2.9190089559543558e-05}
{"train_loss": 0.10593648999929428, "global_step": 172997, "epoch": 1943, "lr": 2.9189562421571857e-05}
{"train_loss": 0.09038960188627243, "global_step": 172998, "epoch": 1943, "lr": 2.9189035286397893e-05}
{"train_loss": 0.08056732267141342, "global_step": 172999, "epoch": 1943, "lr": 2.91885081540217e-05}
{"train_loss": 0.08993162959814072, "global_step": 173000, "epoch": 1943, "lr": 2.9187981024443345e-05}
{"train_loss": 0.08642853051424026, "global_step": 173001, "epoch": 1943, "lr": 2.918745389766293e-05}
{"train_loss": 0.08318211883306503, "global_step": 173002, "epoch": 1943, "lr": 2.9186926773680488e-05}
{"train_loss": 0.046046797186136246, "global_step": 173003, "epoch": 1943, "lr": 2.9186399652496133e-05}
{"train_loss": 0.056808747351169586, "global_step": 173004, "epoch": 1943, "lr": 2.9185872534109908e-05}
{"train_loss": 0.11326276510953903, "global_step": 173005, "epoch": 1943, "lr": 2.918534541852187e-05}
{"train_loss": 0.03442464396357536, "global_step": 173006, "epoch": 1943, "lr": 2.918481830573211e-05}
{"train_loss": 0.024079926311969757, "global_step": 173007, "epoch": 1943, "lr": 2.918429119574072e-05}
{"train_loss": 0.08691795915365219, "global_step": 173008, "epoch": 1943, "lr": 2.9183764088547728e-05}
{"train_loss": 0.0933416485786438, "global_step": 173009, "epoch": 1943, "lr": 2.918323698415324e-05}
{"train_loss": 0.1033397912979126, "global_step": 173010, "epoch": 1943, "lr": 2.918270988255729e-05}
{"train_loss": 0.04134107381105423, "global_step": 173011, "epoch": 1943, "lr": 2.9182182783760003e-05}
{"train_loss": 0.14604732394218445, "global_step": 173012, "epoch": 1943, "lr": 2.9181655687761387e-05}
{"train_loss": 0.07038822025060654, "global_step": 173013, "epoch": 1943, "lr": 2.9181128594561568e-05}
{"train_loss": 0.06326666474342346, "global_step": 173014, "epoch": 1943, "lr": 2.9180601504160576e-05}
{"train_loss": 0.08115851247076238, "global_step": 173015, "epoch": 1943, "lr": 2.9180074416558512e-05, "val_loss": 7.389462947845459}
{"train_loss": 0.06918053328990936, "global_step": 173016, "epoch": 1944, "lr": 2.9179547331755415e-05}
{"train_loss": 0.05141327902674675, "global_step": 173017, "epoch": 1944, "lr": 2.9179020249751392e-05}
{"train_loss": 0.07714475691318512, "global_step": 173018, "epoch": 1944, "lr": 2.9178493170546485e-05}
{"train_loss": 0.08129870146512985, "global_step": 173019, "epoch": 1944, "lr": 2.9177966094140785e-05}
{"train_loss": 0.11263629049062729, "global_step": 173020, "epoch": 1944, "lr": 2.9177439020534335e-05}
{"train_loss": 0.0748061016201973, "global_step": 173021, "epoch": 1944, "lr": 2.9176911949727247e-05}
{"train_loss": 0.12444545328617096, "global_step": 173022, "epoch": 1944, "lr": 2.9176384881719553e-05}
{"train_loss": 0.06930218636989594, "global_step": 173023, "epoch": 1944, "lr": 2.9175857816511337e-05}
{"train_loss": 0.07095558196306229, "global_step": 173024, "epoch": 1944, "lr": 2.9175330754102693e-05}
{"train_loss": 0.09115592390298843, "global_step": 173025, "epoch": 1944, "lr": 2.9174803694493652e-05}
{"train_loss": 0.07533959299325943, "global_step": 173026, "epoch": 1944, "lr": 2.9174276637684327e-05}
{"train_loss": 0.08101115375757217, "global_step": 173027, "epoch": 1944, "lr": 2.9173749583674743e-05}
{"train_loss": 0.08920152485370636, "global_step": 173028, "epoch": 1944, "lr": 2.9173222532465016e-05}
{"train_loss": 0.08808885514736176, "global_step": 173029, "epoch": 1944, "lr": 2.9172695484055174e-05}
{"train_loss": 0.08915245532989502, "global_step": 173030, "epoch": 1944, "lr": 2.9172168438445335e-05}
{"train_loss": 0.07176560908555984, "global_step": 173031, "epoch": 1944, "lr": 2.917164139563552e-05}
{"train_loss": 0.06715176999568939, "global_step": 173032, "epoch": 1944, "lr": 2.917111435562584e-05}
{"train_loss": 0.13653665781021118, "global_step": 173033, "epoch": 1944, "lr": 2.917058731841633e-05}
{"train_loss": 0.07893326878547668, "global_step": 173034, "epoch": 1944, "lr": 2.9170060284007106e-05}
{"train_loss": 0.08443669974803925, "global_step": 173035, "epoch": 1944, "lr": 2.9169533252398196e-05}
{"train_loss": 0.09314146637916565, "global_step": 173036, "epoch": 1944, "lr": 2.9169006223589703e-05}
{"train_loss": 0.10966906696557999, "global_step": 173037, "epoch": 1944, "lr": 2.9168479197581666e-05}
{"train_loss": 0.03808087483048439, "global_step": 173038, "epoch": 1944, "lr": 2.916795217437419e-05}
{"train_loss": 0.06117994338274002, "global_step": 173039, "epoch": 1944, "lr": 2.916742515396733e-05}
{"train_loss": 0.0482320636510849, "global_step": 173040, "epoch": 1944, "lr": 2.916689813636114e-05}
{"train_loss": 0.06012918800115585, "global_step": 173041, "epoch": 1944, "lr": 2.916637112155572e-05}
{"train_loss": 0.05910385027527809, "global_step": 173042, "epoch": 1944, "lr": 2.916584410955111e-05}
{"train_loss": 0.08982712775468826, "global_step": 173043, "epoch": 1944, "lr": 2.9165317100347417e-05}
{"train_loss": 0.09060772508382797, "global_step": 173044, "epoch": 1944, "lr": 2.9164790093944673e-05}
{"train_loss": 0.08460365980863571, "global_step": 173045, "epoch": 1944, "lr": 2.9164263090342993e-05}
{"train_loss": 0.07556306570768356, "global_step": 173046, "epoch": 1944, "lr": 2.9163736089542403e-05}
{"train_loss": 0.04893882945179939, "global_step": 173047, "epoch": 1944, "lr": 2.9163209091543008e-05}
{"train_loss": 0.10529865324497223, "global_step": 173048, "epoch": 1944, "lr": 2.9162682096344857e-05}
{"train_loss": 0.08117195963859558, "global_step": 173049, "epoch": 1944, "lr": 2.9162155103948043e-05}
{"train_loss": 0.11307401210069656, "global_step": 173050, "epoch": 1944, "lr": 2.91616281143526e-05}
{"train_loss": 0.03794051334261894, "global_step": 173051, "epoch": 1944, "lr": 2.916110112755865e-05}
{"train_loss": 0.11330657452344894, "global_step": 173052, "epoch": 1944, "lr": 2.9160574143566212e-05}
{"train_loss": 0.054544806480407715, "global_step": 173053, "epoch": 1944, "lr": 2.9160047162375397e-05}
{"train_loss": 0.029210228472948074, "global_step": 173054, "epoch": 1944, "lr": 2.9159520183986245e-05}
{"train_loss": 0.10010942071676254, "global_step": 173055, "epoch": 1944, "lr": 2.915899320839886e-05}
{"train_loss": 0.0450124554336071, "global_step": 173056, "epoch": 1944, "lr": 2.9158466235613276e-05}
{"train_loss": 0.09609776735305786, "global_step": 173057, "epoch": 1944, "lr": 2.9157939265629597e-05}
{"train_loss": 0.06757237017154694, "global_step": 173058, "epoch": 1944, "lr": 2.915741229844786e-05}
{"train_loss": 0.058379024267196655, "global_step": 173059, "epoch": 1944, "lr": 2.9156885334068162e-05}
{"train_loss": 0.04828475043177605, "global_step": 173060, "epoch": 1944, "lr": 2.9156358372490578e-05}
{"train_loss": 0.10058043152093887, "global_step": 173061, "epoch": 1944, "lr": 2.915583141371515e-05}
{"train_loss": 0.05934716388583183, "global_step": 173062, "epoch": 1944, "lr": 2.9155304457741995e-05}
{"train_loss": 0.07423073053359985, "global_step": 173063, "epoch": 1944, "lr": 2.9154777504571124e-05}
{"train_loss": 0.08050734549760818, "global_step": 173064, "epoch": 1944, "lr": 2.9154250554202666e-05}
{"train_loss": 0.051878057420253754, "global_step": 173065, "epoch": 1944, "lr": 2.915372360663664e-05}
{"train_loss": 0.0675196647644043, "global_step": 173066, "epoch": 1944, "lr": 2.915319666187316e-05}
{"train_loss": 0.08436301350593567, "global_step": 173067, "epoch": 1944, "lr": 2.9152669719912267e-05}
{"train_loss": 0.12407660484313965, "global_step": 173068, "epoch": 1944, "lr": 2.915214278075406e-05}
{"train_loss": 0.09401395171880722, "global_step": 173069, "epoch": 1944, "lr": 2.915161584439857e-05}
{"train_loss": 0.06289009004831314, "global_step": 173070, "epoch": 1944, "lr": 2.9151088910845912e-05}
{"train_loss": 0.07066984474658966, "global_step": 173071, "epoch": 1944, "lr": 2.915056198009612e-05}
{"train_loss": 0.08637537807226181, "global_step": 173072, "epoch": 1944, "lr": 2.91500350521493e-05}
{"train_loss": 0.0687917023897171, "global_step": 173073, "epoch": 1944, "lr": 2.914950812700548e-05}
{"train_loss": 0.08300864696502686, "global_step": 173074, "epoch": 1944, "lr": 2.9148981204664776e-05}
{"train_loss": 0.06232507899403572, "global_step": 173075, "epoch": 1944, "lr": 2.914845428512722e-05}
{"train_loss": 0.04274239391088486, "global_step": 173076, "epoch": 1944, "lr": 2.9147927368392914e-05}
{"train_loss": 0.08015821129083633, "global_step": 173077, "epoch": 1944, "lr": 2.914740045446192e-05}
{"train_loss": 0.07220304757356644, "global_step": 173078, "epoch": 1944, "lr": 2.9146873543334285e-05}
{"train_loss": 0.09077662229537964, "global_step": 173079, "epoch": 1944, "lr": 2.9146346635010113e-05}
{"train_loss": 0.07178113609552383, "global_step": 173080, "epoch": 1944, "lr": 2.914581972948945e-05}
{"train_loss": 0.054175835102796555, "global_step": 173081, "epoch": 1944, "lr": 2.9145292826772385e-05}
{"train_loss": 0.05938868597149849, "global_step": 173082, "epoch": 1944, "lr": 2.9144765926858976e-05}
{"train_loss": 0.06876172125339508, "global_step": 173083, "epoch": 1944, "lr": 2.914423902974931e-05}
{"train_loss": 0.06530506163835526, "global_step": 173084, "epoch": 1944, "lr": 2.9143712135443424e-05}
{"train_loss": 0.04737783968448639, "global_step": 173085, "epoch": 1944, "lr": 2.9143185243941435e-05}
{"train_loss": 0.06926129013299942, "global_step": 173086, "epoch": 1944, "lr": 2.9142658355243368e-05}
{"train_loss": 0.10035500675439835, "global_step": 173087, "epoch": 1944, "lr": 2.9142131469349343e-05}
{"train_loss": 0.12117455154657364, "global_step": 173088, "epoch": 1944, "lr": 2.914160458625938e-05}
{"train_loss": 0.07764750719070435, "global_step": 173089, "epoch": 1944, "lr": 2.9141077705973598e-05}
{"train_loss": 0.0664541944861412, "global_step": 173090, "epoch": 1944, "lr": 2.914055082849202e-05}
{"train_loss": 0.0586550347507, "global_step": 173091, "epoch": 1944, "lr": 2.9140023953814754e-05}
{"train_loss": 0.04003972187638283, "global_step": 173092, "epoch": 1944, "lr": 2.9139497081941848e-05}
{"train_loss": 0.139967679977417, "global_step": 173093, "epoch": 1944, "lr": 2.9138970212873394e-05}
{"train_loss": 0.09874178469181061, "global_step": 173094, "epoch": 1944, "lr": 2.913844334660944e-05}
{"train_loss": 0.06951393187046051, "global_step": 173095, "epoch": 1944, "lr": 2.9137916483150065e-05}
{"train_loss": 0.09936696290969849, "global_step": 173096, "epoch": 1944, "lr": 2.9137389622495358e-05}
{"train_loss": 0.13605202734470367, "global_step": 173097, "epoch": 1944, "lr": 2.913686276464536e-05}
{"train_loss": 0.07590541988611221, "global_step": 173098, "epoch": 1944, "lr": 2.9136335909600178e-05}
{"train_loss": 0.10638301074504852, "global_step": 173099, "epoch": 1944, "lr": 2.9135809057359836e-05}
{"train_loss": 0.12515750527381897, "global_step": 173100, "epoch": 1944, "lr": 2.9135282207924453e-05}
{"train_loss": 0.063749760389328, "global_step": 173101, "epoch": 1944, "lr": 2.913475536129405e-05}
{"train_loss": 0.09768634289503098, "global_step": 173102, "epoch": 1944, "lr": 2.9134228517468758e-05}
{"train_loss": 0.1032019853591919, "global_step": 173103, "epoch": 1944, "lr": 2.9133701676448587e-05}
{"train_loss": 0.07868409135870719, "global_step": 173104, "epoch": 1944, "lr": 2.9133174838233657e-05, "val_loss": 7.305641174316406}
{"train_loss": 0.1497231274843216, "global_step": 173105, "epoch": 1945, "lr": 2.9132648002824e-05}
{"train_loss": 0.10437647253274918, "global_step": 173106, "epoch": 1945, "lr": 2.9132121170219717e-05}
{"train_loss": 0.059727512300014496, "global_step": 173107, "epoch": 1945, "lr": 2.9131594340420855e-05}
{"train_loss": 0.08082307875156403, "global_step": 173108, "epoch": 1945, "lr": 2.9131067513427506e-05}
{"train_loss": 0.07532273232936859, "global_step": 173109, "epoch": 1945, "lr": 2.9130540689239716e-05}
{"train_loss": 0.0869244709610939, "global_step": 173110, "epoch": 1945, "lr": 2.913001386785759e-05}
{"train_loss": 0.0564117468893528, "global_step": 173111, "epoch": 1945, "lr": 2.9129487049281157e-05}
{"train_loss": 0.04879974573850632, "global_step": 173112, "epoch": 1945, "lr": 2.9128960233510537e-05}
{"train_loss": 0.21612648665905, "global_step": 173113, "epoch": 1945, "lr": 2.9128433420545747e-05}
{"train_loss": 0.027696965262293816, "global_step": 173114, "epoch": 1945, "lr": 2.9127906610386906e-05}
{"train_loss": 0.05361062288284302, "global_step": 173115, "epoch": 1945, "lr": 2.9127379803034066e-05}
{"train_loss": 0.04258604720234871, "global_step": 173116, "epoch": 1945, "lr": 2.9126852998487276e-05}
{"train_loss": 0.1284663826227188, "global_step": 173117, "epoch": 1945, "lr": 2.9126326196746645e-05}
{"train_loss": 0.08660408109426498, "global_step": 173118, "epoch": 1945, "lr": 2.9125799397812204e-05}
{"train_loss": 0.09373515099287033, "global_step": 173119, "epoch": 1945, "lr": 2.912527260168406e-05}
{"train_loss": 0.06292197108268738, "global_step": 173120, "epoch": 1945, "lr": 2.9124745808362253e-05}
{"train_loss": 0.1144065335392952, "global_step": 173121, "epoch": 1945, "lr": 2.9124219017846893e-05}
{"train_loss": 0.12984339892864227, "global_step": 173122, "epoch": 1945, "lr": 2.9123692230138e-05}
{"train_loss": 0.058137450367212296, "global_step": 173123, "epoch": 1945, "lr": 2.91231654452357e-05}
{"train_loss": 0.1078575998544693, "global_step": 173124, "epoch": 1945, "lr": 2.9122638663140013e-05}
{"train_loss": 0.08185426890850067, "global_step": 173125, "epoch": 1945, "lr": 2.9122111883851054e-05}
{"train_loss": 0.07938859611749649, "global_step": 173126, "epoch": 1945, "lr": 2.9121585107368848e-05}
{"train_loss": 0.06705241650342941, "global_step": 173127, "epoch": 1945, "lr": 2.9121058333693512e-05}
{"train_loss": 0.06503534317016602, "global_step": 173128, "epoch": 1945, "lr": 2.912053156282508e-05}
{"train_loss": 0.10227663069963455, "global_step": 173129, "epoch": 1945, "lr": 2.9120004794763655e-05}
{"train_loss": 0.03713454678654671, "global_step": 173130, "epoch": 1945, "lr": 2.9119478029509267e-05}
{"train_loss": 0.07099337130784988, "global_step": 173131, "epoch": 1945, "lr": 2.9118951267062034e-05}
{"train_loss": 0.14955410361289978, "global_step": 173132, "epoch": 1945, "lr": 2.9118424507421983e-05}
{"train_loss": 0.10200833529233932, "global_step": 173133, "epoch": 1945, "lr": 2.9117897750589206e-05}
{"train_loss": 0.10271024703979492, "global_step": 173134, "epoch": 1945, "lr": 2.9117370996563798e-05}
{"train_loss": 0.041993170976638794, "global_step": 173135, "epoch": 1945, "lr": 2.9116844245345776e-05}
{"train_loss": 0.07067210227251053, "global_step": 173136, "epoch": 1945, "lr": 2.9116317496935264e-05}
{"train_loss": 0.10314040631055832, "global_step": 173137, "epoch": 1945, "lr": 2.9115790751332288e-05}
{"train_loss": 0.03999767452478409, "global_step": 173138, "epoch": 1945, "lr": 2.9115264008536958e-05}
{"train_loss": 0.10605277866125107, "global_step": 173139, "epoch": 1945, "lr": 2.9114737268549308e-05}
{"train_loss": 0.11192598938941956, "global_step": 173140, "epoch": 1945, "lr": 2.9114210531369447e-05}
{"train_loss": 0.1219407320022583, "global_step": 173141, "epoch": 1945, "lr": 2.9113683796997406e-05}
{"train_loss": 0.0652804896235466, "global_step": 173142, "epoch": 1945, "lr": 2.9113157065433295e-05}
{"train_loss": 0.0847388505935669, "global_step": 173143, "epoch": 1945, "lr": 2.911263033667715e-05}
{"train_loss": 0.08439327776432037, "global_step": 173144, "epoch": 1945, "lr": 2.9112103610729068e-05}
{"train_loss": 0.1051735058426857, "global_step": 173145, "epoch": 1945, "lr": 2.9111576887589097e-05}
{"train_loss": 0.04690314084291458, "global_step": 173146, "epoch": 1945, "lr": 2.9111050167257338e-05}
{"train_loss": 0.12270143628120422, "global_step": 173147, "epoch": 1945, "lr": 2.911052344973382e-05}
{"train_loss": 0.08350125700235367, "global_step": 173148, "epoch": 1945, "lr": 2.9109996735018664e-05}
{"train_loss": 0.08952400088310242, "global_step": 173149, "epoch": 1945, "lr": 2.9109470023111885e-05}
{"train_loss": 0.09947231411933899, "global_step": 173150, "epoch": 1945, "lr": 2.9108943314013616e-05}
{"train_loss": 0.057710956782102585, "global_step": 173151, "epoch": 1945, "lr": 2.9108416607723864e-05}
{"train_loss": 0.11146030575037003, "global_step": 173152, "epoch": 1945, "lr": 2.9107889904242758e-05}
{"train_loss": 0.07483262568712234, "global_step": 173153, "epoch": 1945, "lr": 2.9107363203570316e-05}
{"train_loss": 0.055256959050893784, "global_step": 173154, "epoch": 1945, "lr": 2.9106836505706658e-05}
{"train_loss": 0.021332571282982826, "global_step": 173155, "epoch": 1945, "lr": 2.9106309810651826e-05}
{"train_loss": 0.036346301436424255, "global_step": 173156, "epoch": 1945, "lr": 2.9105783118405884e-05}
{"train_loss": 0.057330965995788574, "global_step": 173157, "epoch": 1945, "lr": 2.910525642896893e-05}
{"train_loss": 0.0504283644258976, "global_step": 173158, "epoch": 1945, "lr": 2.9104729742341e-05}
{"train_loss": 0.07942257821559906, "global_step": 173159, "epoch": 1945, "lr": 2.91042030585222e-05}
{"train_loss": 0.04753594473004341, "global_step": 173160, "epoch": 1945, "lr": 2.9103676377512567e-05}
{"train_loss": 0.12301399558782578, "global_step": 173161, "epoch": 1945, "lr": 2.9103149699312214e-05}
{"train_loss": 0.05454518273472786, "global_step": 173162, "epoch": 1945, "lr": 2.910262302392117e-05}
{"train_loss": 0.08084014803171158, "global_step": 173163, "epoch": 1945, "lr": 2.9102096351339535e-05}
{"train_loss": 0.09720616042613983, "global_step": 173164, "epoch": 1945, "lr": 2.9101569681567353e-05}
{"train_loss": 0.067439965903759, "global_step": 173165, "epoch": 1945, "lr": 2.9101043014604734e-05}
{"train_loss": 0.05228941887617111, "global_step": 173166, "epoch": 1945, "lr": 2.91005163504517e-05}
{"train_loss": 0.06512993574142456, "global_step": 173167, "epoch": 1945, "lr": 2.909998968910837e-05}
{"train_loss": 0.08602943271398544, "global_step": 173168, "epoch": 1945, "lr": 2.909946303057477e-05}
{"train_loss": 0.047930240631103516, "global_step": 173169, "epoch": 1945, "lr": 2.9098936374850994e-05}
{"train_loss": 0.06736644357442856, "global_step": 173170, "epoch": 1945, "lr": 2.9098409721937125e-05}
{"train_loss": 0.03284844011068344, "global_step": 173171, "epoch": 1945, "lr": 2.909788307183321e-05}
{"train_loss": 0.06962626427412033, "global_step": 173172, "epoch": 1945, "lr": 2.9097356424539347e-05}
{"train_loss": 0.09754179418087006, "global_step": 173173, "epoch": 1945, "lr": 2.9096829780055567e-05}
{"train_loss": 0.0900086909532547, "global_step": 173174, "epoch": 1945, "lr": 2.909630313838198e-05}
{"train_loss": 0.07637856900691986, "global_step": 173175, "epoch": 1945, "lr": 2.9095776499518624e-05}
{"train_loss": 0.050695065408945084, "global_step": 173176, "epoch": 1945, "lr": 2.9095249863465613e-05}
{"train_loss": 0.057228364050388336, "global_step": 173177, "epoch": 1945, "lr": 2.909472323022296e-05}
{"train_loss": 0.07735102623701096, "global_step": 173178, "epoch": 1945, "lr": 2.9094196599790792e-05}
{"train_loss": 0.09816807508468628, "global_step": 173179, "epoch": 1945, "lr": 2.9093669972169137e-05}
{"train_loss": 0.10323086380958557, "global_step": 173180, "epoch": 1945, "lr": 2.9093143347358097e-05}
{"train_loss": 0.08875381946563721, "global_step": 173181, "epoch": 1945, "lr": 2.9092616725357708e-05}
{"train_loss": 0.10902240872383118, "global_step": 173182, "epoch": 1945, "lr": 2.9092090106168084e-05}
{"train_loss": 0.07520138472318649, "global_step": 173183, "epoch": 1945, "lr": 2.909156348978926e-05}
{"train_loss": 0.12877723574638367, "global_step": 173184, "epoch": 1945, "lr": 2.9091036876221335e-05}
{"train_loss": 0.08365990966558456, "global_step": 173185, "epoch": 1945, "lr": 2.9090510265464343e-05}
{"train_loss": 0.05343569815158844, "global_step": 173186, "epoch": 1945, "lr": 2.9089983657518383e-05}
{"train_loss": 0.09714872390031815, "global_step": 173187, "epoch": 1945, "lr": 2.9089457052383535e-05}
{"train_loss": 0.08142682164907455, "global_step": 173188, "epoch": 1945, "lr": 2.9088930450059858e-05}
{"train_loss": 0.03530338779091835, "global_step": 173189, "epoch": 1945, "lr": 2.9088403850547398e-05}
{"train_loss": 0.12436630576848984, "global_step": 173190, "epoch": 1945, "lr": 2.9087877253846264e-05}
{"train_loss": 0.12944552302360535, "global_step": 173191, "epoch": 1945, "lr": 2.908735065995649e-05}
{"train_loss": 0.07374422997236252, "global_step": 173192, "epoch": 1945, "lr": 2.9086824068878195e-05}
{"train_loss": 0.08148347347807348, "global_step": 173193, "epoch": 1945, "lr": 2.9086297480611412e-05, "val_loss": 7.380185604095459, "train_action_mse_error": 14.043085098266602}
{"train_loss": 0.05472642555832863, "global_step": 173194, "epoch": 1946, "lr": 2.9085770895156206e-05}
{"train_loss": 0.07511293143033981, "global_step": 173195, "epoch": 1946, "lr": 2.908524431251268e-05}
{"train_loss": 0.0654938742518425, "global_step": 173196, "epoch": 1946, "lr": 2.9084717732680876e-05}
{"train_loss": 0.1378873586654663, "global_step": 173197, "epoch": 1946, "lr": 2.908419115566089e-05}
{"train_loss": 0.07948526740074158, "global_step": 173198, "epoch": 1946, "lr": 2.9083664581452762e-05}
{"train_loss": 0.14516352117061615, "global_step": 173199, "epoch": 1946, "lr": 2.90831380100566e-05}
{"train_loss": 0.07340960949659348, "global_step": 173200, "epoch": 1946, "lr": 2.908261144147244e-05}
{"train_loss": 0.1087803766131401, "global_step": 173201, "epoch": 1946, "lr": 2.9082084875700378e-05}
{"train_loss": 0.10953047126531601, "global_step": 173202, "epoch": 1946, "lr": 2.9081558312740466e-05}
{"train_loss": 0.07922303676605225, "global_step": 173203, "epoch": 1946, "lr": 2.908103175259279e-05}
{"train_loss": 0.07912903279066086, "global_step": 173204, "epoch": 1946, "lr": 2.9080505195257402e-05}
{"train_loss": 0.12745782732963562, "global_step": 173205, "epoch": 1946, "lr": 2.9079978640734386e-05}
{"train_loss": 0.0651354193687439, "global_step": 173206, "epoch": 1946, "lr": 2.9079452089023827e-05}
{"train_loss": 0.1547444611787796, "global_step": 173207, "epoch": 1946, "lr": 2.9078925540125767e-05}
{"train_loss": 0.11619328707456589, "global_step": 173208, "epoch": 1946, "lr": 2.9078398994040313e-05}
{"train_loss": 0.11701071262359619, "global_step": 173209, "epoch": 1946, "lr": 2.9077872450767484e-05}
{"train_loss": 0.0763162150979042, "global_step": 173210, "epoch": 1946, "lr": 2.9077345910307406e-05}
{"train_loss": 0.09514330327510834, "global_step": 173211, "epoch": 1946, "lr": 2.90768193726601e-05}
{"train_loss": 0.08773928135633469, "global_step": 173212, "epoch": 1946, "lr": 2.9076292837825682e-05}
{"train_loss": 0.13035418093204498, "global_step": 173213, "epoch": 1946, "lr": 2.907576630580418e-05}
{"train_loss": 0.05919189751148224, "global_step": 173214, "epoch": 1946, "lr": 2.907523977659571e-05}
{"train_loss": 0.09593671560287476, "global_step": 173215, "epoch": 1946, "lr": 2.90747132502003e-05}
{"train_loss": 0.03778204321861267, "global_step": 173216, "epoch": 1946, "lr": 2.907418672661805e-05}
{"train_loss": 0.05944426357746124, "global_step": 173217, "epoch": 1946, "lr": 2.9073660205849006e-05}
{"train_loss": 0.09981576353311539, "global_step": 173218, "epoch": 1946, "lr": 2.907313368789328e-05}
{"train_loss": 0.09565545618534088, "global_step": 173219, "epoch": 1946, "lr": 2.9072607172750888e-05}
{"train_loss": 0.11863919347524643, "global_step": 173220, "epoch": 1946, "lr": 2.907208066042195e-05}
{"train_loss": 0.04597412422299385, "global_step": 173221, "epoch": 1946, "lr": 2.9071554150906498e-05}
{"train_loss": 0.07206374406814575, "global_step": 173222, "epoch": 1946, "lr": 2.9071027644204617e-05}
{"train_loss": 0.06006103381514549, "global_step": 173223, "epoch": 1946, "lr": 2.907050114031641e-05}
{"train_loss": 0.019856195896863937, "global_step": 173224, "epoch": 1946, "lr": 2.906997463924189e-05}
{"train_loss": 0.04428771883249283, "global_step": 173225, "epoch": 1946, "lr": 2.906944814098118e-05}
{"train_loss": 0.056506093591451645, "global_step": 173226, "epoch": 1946, "lr": 2.906892164553433e-05}
{"train_loss": 0.08751433342695236, "global_step": 173227, "epoch": 1946, "lr": 2.9068395152901384e-05}
{"train_loss": 0.12909582257270813, "global_step": 173228, "epoch": 1946, "lr": 2.906786866308246e-05}
{"train_loss": 0.07502710819244385, "global_step": 173229, "epoch": 1946, "lr": 2.9067342176077584e-05}
{"train_loss": 0.09323498606681824, "global_step": 173230, "epoch": 1946, "lr": 2.9066815691886874e-05}
{"train_loss": 0.057975757867097855, "global_step": 173231, "epoch": 1946, "lr": 2.9066289210510367e-05}
{"train_loss": 0.0703786313533783, "global_step": 173232, "epoch": 1946, "lr": 2.9065762731948127e-05}
{"train_loss": 0.0817369893193245, "global_step": 173233, "epoch": 1946, "lr": 2.9065236256200256e-05}
{"train_loss": 0.061469078063964844, "global_step": 173234, "epoch": 1946, "lr": 2.9064709783266797e-05}
{"train_loss": 0.06372244656085968, "global_step": 173235, "epoch": 1946, "lr": 2.9064183313147842e-05}
{"train_loss": 0.07621683180332184, "global_step": 173236, "epoch": 1946, "lr": 2.906365684584344e-05}
{"train_loss": 0.05738409608602524, "global_step": 173237, "epoch": 1946, "lr": 2.9063130381353693e-05}
{"train_loss": 0.07179111242294312, "global_step": 173238, "epoch": 1946, "lr": 2.9062603919678633e-05}
{"train_loss": 0.10221655666828156, "global_step": 173239, "epoch": 1946, "lr": 2.9062077460818373e-05}
{"train_loss": 0.11625859141349792, "global_step": 173240, "epoch": 1946, "lr": 2.9061551004772935e-05}
{"train_loss": 0.028469569981098175, "global_step": 173241, "epoch": 1946, "lr": 2.906102455154242e-05}
{"train_loss": 0.07237137109041214, "global_step": 173242, "epoch": 1946, "lr": 2.9060498101126917e-05}
{"train_loss": 0.05841072276234627, "global_step": 173243, "epoch": 1946, "lr": 2.905997165352645e-05}
{"train_loss": 0.06750304251909256, "global_step": 173244, "epoch": 1946, "lr": 2.9059445208741137e-05}
{"train_loss": 0.04990292713046074, "global_step": 173245, "epoch": 1946, "lr": 2.9058918766771005e-05}
{"train_loss": 0.0522993765771389, "global_step": 173246, "epoch": 1946, "lr": 2.905839232761617e-05}
{"train_loss": 0.06936333328485489, "global_step": 173247, "epoch": 1946, "lr": 2.905786589127666e-05}
{"train_loss": 0.04192734137177467, "global_step": 173248, "epoch": 1946, "lr": 2.9057339457752573e-05}
{"train_loss": 0.08671212196350098, "global_step": 173249, "epoch": 1946, "lr": 2.905681302704396e-05}
{"train_loss": 0.07987423986196518, "global_step": 173250, "epoch": 1946, "lr": 2.9056286599150927e-05}
{"train_loss": 0.19069255888462067, "global_step": 173251, "epoch": 1946, "lr": 2.9055760174073498e-05}
{"train_loss": 0.09722677618265152, "global_step": 173252, "epoch": 1946, "lr": 2.9055233751811783e-05}
{"train_loss": 0.10238852351903915, "global_step": 173253, "epoch": 1946, "lr": 2.9054707332365822e-05}
{"train_loss": 0.10221187025308609, "global_step": 173254, "epoch": 1946, "lr": 2.9054180915735717e-05}
{"train_loss": 0.08516348153352737, "global_step": 173255, "epoch": 1946, "lr": 2.9053654501921503e-05}
{"train_loss": 0.11999671906232834, "global_step": 173256, "epoch": 1946, "lr": 2.9053128090923286e-05}
{"train_loss": 0.08108899742364883, "global_step": 173257, "epoch": 1946, "lr": 2.905260168274111e-05}
{"train_loss": 0.07464105635881424, "global_step": 173258, "epoch": 1946, "lr": 2.9052075277375068e-05}
{"train_loss": 0.07773563265800476, "global_step": 173259, "epoch": 1946, "lr": 2.9051548874825195e-05}
{"train_loss": 0.09325891733169556, "global_step": 173260, "epoch": 1946, "lr": 2.9051022475091595e-05}
{"train_loss": 0.06161538138985634, "global_step": 173261, "epoch": 1946, "lr": 2.905049607817435e-05}
{"train_loss": 0.0640481486916542, "global_step": 173262, "epoch": 1946, "lr": 2.9049969684073485e-05}
{"train_loss": 0.0689920037984848, "global_step": 173263, "epoch": 1946, "lr": 2.9049443292789126e-05}
{"train_loss": 0.1248488500714302, "global_step": 173264, "epoch": 1946, "lr": 2.904891690432128e-05}
{"train_loss": 0.08464904129505157, "global_step": 173265, "epoch": 1946, "lr": 2.9048390518670082e-05}
{"train_loss": 0.12890946865081787, "global_step": 173266, "epoch": 1946, "lr": 2.904786413583557e-05}
{"train_loss": 0.09163201600313187, "global_step": 173267, "epoch": 1946, "lr": 2.9047337755817795e-05}
{"train_loss": 0.0722280740737915, "global_step": 173268, "epoch": 1946, "lr": 2.904681137861687e-05}
{"train_loss": 0.10475241392850876, "global_step": 173269, "epoch": 1946, "lr": 2.9046285004232833e-05}
{"train_loss": 0.07707186043262482, "global_step": 173270, "epoch": 1946, "lr": 2.9045758632665775e-05}
{"train_loss": 0.06306501477956772, "global_step": 173271, "epoch": 1946, "lr": 2.9045232263915766e-05}
{"train_loss": 0.065861277282238, "global_step": 173272, "epoch": 1946, "lr": 2.9044705897982853e-05}
{"train_loss": 0.04643223062157631, "global_step": 173273, "epoch": 1946, "lr": 2.9044179534867145e-05}
{"train_loss": 0.0713334009051323, "global_step": 173274, "epoch": 1946, "lr": 2.9043653174568665e-05}
{"train_loss": 0.06675324589014053, "global_step": 173275, "epoch": 1946, "lr": 2.9043126817087535e-05}
{"train_loss": 0.09169615060091019, "global_step": 173276, "epoch": 1946, "lr": 2.9042600462423775e-05}
{"train_loss": 0.04001671448349953, "global_step": 173277, "epoch": 1946, "lr": 2.9042074110577488e-05}
{"train_loss": 0.08212847262620926, "global_step": 173278, "epoch": 1946, "lr": 2.904154776154876e-05}
{"train_loss": 0.11800432205200195, "global_step": 173279, "epoch": 1946, "lr": 2.9041021415337614e-05}
{"train_loss": 0.04759492725133896, "global_step": 173280, "epoch": 1946, "lr": 2.9040495071944174e-05}
{"train_loss": 0.10560746490955353, "global_step": 173281, "epoch": 1946, "lr": 2.903996873136846e-05}
{"train_loss": 0.08255240224887815, "global_step": 173282, "epoch": 1946, "lr": 2.903944239361059e-05, "val_loss": 7.118244647979736}
{"train_loss": 0.08013497292995453, "global_step": 173283, "epoch": 1947, "lr": 2.9038916058670585e-05}
{"train_loss": 0.040201585739851, "global_step": 173284, "epoch": 1947, "lr": 2.9038389726548565e-05}
{"train_loss": 0.09269894659519196, "global_step": 173285, "epoch": 1947, "lr": 2.9037863397244565e-05}
{"train_loss": 0.08216345310211182, "global_step": 173286, "epoch": 1947, "lr": 2.903733707075868e-05}
{"train_loss": 0.06143760308623314, "global_step": 173287, "epoch": 1947, "lr": 2.9036810747090954e-05}
{"train_loss": 0.04186287894845009, "global_step": 173288, "epoch": 1947, "lr": 2.903628442624149e-05}
{"train_loss": 0.06946627795696259, "global_step": 173289, "epoch": 1947, "lr": 2.9035758108210325e-05}
{"train_loss": 0.061131179332733154, "global_step": 173290, "epoch": 1947, "lr": 2.9035231792997562e-05}
{"train_loss": 0.07908258587121964, "global_step": 173291, "epoch": 1947, "lr": 2.903470548060324e-05}
{"train_loss": 0.10242222249507904, "global_step": 173292, "epoch": 1947, "lr": 2.903417917102747e-05}
{"train_loss": 0.08322405070066452, "global_step": 173293, "epoch": 1947, "lr": 2.9033652864270276e-05}
{"train_loss": 0.08230821043252945, "global_step": 173294, "epoch": 1947, "lr": 2.903312656033177e-05}
{"train_loss": 0.087359718978405, "global_step": 173295, "epoch": 1947, "lr": 2.903260025921199e-05}
{"train_loss": 0.08783280849456787, "global_step": 173296, "epoch": 1947, "lr": 2.9032073960911017e-05}
{"train_loss": 0.05711081996560097, "global_step": 173297, "epoch": 1947, "lr": 2.9031547665428948e-05}
{"train_loss": 0.08023881912231445, "global_step": 173298, "epoch": 1947, "lr": 2.903102137276581e-05}
{"train_loss": 0.09624601900577545, "global_step": 173299, "epoch": 1947, "lr": 2.9030495082921716e-05}
{"train_loss": 0.05519988760352135, "global_step": 173300, "epoch": 1947, "lr": 2.9029968795896695e-05}
{"train_loss": 0.08845023810863495, "global_step": 173301, "epoch": 1947, "lr": 2.9029442511690864e-05}
{"train_loss": 0.11412157863378525, "global_step": 173302, "epoch": 1947, "lr": 2.9028916230304243e-05}
{"train_loss": 0.11166156083345413, "global_step": 173303, "epoch": 1947, "lr": 2.902838995173695e-05}
{"train_loss": 0.12212255597114563, "global_step": 173304, "epoch": 1947, "lr": 2.9027863675989035e-05}
{"train_loss": 0.07897251844406128, "global_step": 173305, "epoch": 1947, "lr": 2.902733740306055e-05}
{"train_loss": 0.09573465585708618, "global_step": 173306, "epoch": 1947, "lr": 2.9026811132951604e-05}
{"train_loss": 0.08190155029296875, "global_step": 173307, "epoch": 1947, "lr": 2.9026284865662224e-05}
{"train_loss": 0.07293452322483063, "global_step": 173308, "epoch": 1947, "lr": 2.9025758601192528e-05}
{"train_loss": 0.08861217647790909, "global_step": 173309, "epoch": 1947, "lr": 2.9025232339542564e-05}
{"train_loss": 0.06275279819965363, "global_step": 173310, "epoch": 1947, "lr": 2.902470608071237e-05}
{"train_loss": 0.06484194844961166, "global_step": 173311, "epoch": 1947, "lr": 2.902417982470208e-05}
{"train_loss": 0.07475502043962479, "global_step": 173312, "epoch": 1947, "lr": 2.902365357151171e-05}
{"train_loss": 0.056591521948575974, "global_step": 173313, "epoch": 1947, "lr": 2.9023127321141358e-05}
{"train_loss": 0.047681089490652084, "global_step": 173314, "epoch": 1947, "lr": 2.9022601073591112e-05}
{"train_loss": 0.06370958685874939, "global_step": 173315, "epoch": 1947, "lr": 2.9022074828860997e-05}
{"train_loss": 0.060710202902555466, "global_step": 173316, "epoch": 1947, "lr": 2.902154858695113e-05}
{"train_loss": 0.1131325289607048, "global_step": 173317, "epoch": 1947, "lr": 2.9021022347861536e-05}
{"train_loss": 0.061847757548093796, "global_step": 173318, "epoch": 1947, "lr": 2.9020496111592332e-05}
{"train_loss": 0.138393834233284, "global_step": 173319, "epoch": 1947, "lr": 2.9019969878143548e-05}
{"train_loss": 0.12529896199703217, "global_step": 173320, "epoch": 1947, "lr": 2.901944364751529e-05}
{"train_loss": 0.05785020813345909, "global_step": 173321, "epoch": 1947, "lr": 2.9018917419707604e-05}
{"train_loss": 0.1351684182882309, "global_step": 173322, "epoch": 1947, "lr": 2.9018391194720574e-05}
{"train_loss": 0.09471377730369568, "global_step": 173323, "epoch": 1947, "lr": 2.9017864972554255e-05}
{"train_loss": 0.09157917648553848, "global_step": 173324, "epoch": 1947, "lr": 2.9017338753208746e-05}
{"train_loss": 0.11179421842098236, "global_step": 173325, "epoch": 1947, "lr": 2.901681253668408e-05}
{"train_loss": 0.0978948175907135, "global_step": 173326, "epoch": 1947, "lr": 2.901628632298036e-05}
{"train_loss": 0.10698667168617249, "global_step": 173327, "epoch": 1947, "lr": 2.9015760112097633e-05}
{"train_loss": 0.07695814222097397, "global_step": 173328, "epoch": 1947, "lr": 2.9015233904035998e-05}
{"train_loss": 0.06488234549760818, "global_step": 173329, "epoch": 1947, "lr": 2.9014707698795494e-05}
{"train_loss": 0.05565665289759636, "global_step": 173330, "epoch": 1947, "lr": 2.9014181496376224e-05}
{"train_loss": 0.044878389686346054, "global_step": 173331, "epoch": 1947, "lr": 2.901365529677822e-05}
{"train_loss": 0.1352614164352417, "global_step": 173332, "epoch": 1947, "lr": 2.9013129100001575e-05}
{"train_loss": 0.08591858297586441, "global_step": 173333, "epoch": 1947, "lr": 2.9012602906046383e-05}
{"train_loss": 0.12026375532150269, "global_step": 173334, "epoch": 1947, "lr": 2.901207671491266e-05}
{"train_loss": 0.07846062630414963, "global_step": 173335, "epoch": 1947, "lr": 2.901155052660054e-05}
{"train_loss": 0.038105543702840805, "global_step": 173336, "epoch": 1947, "lr": 2.9011024341110034e-05}
{"train_loss": 0.10838384926319122, "global_step": 173337, "epoch": 1947, "lr": 2.9010498158441257e-05}
{"train_loss": 0.06990429759025574, "global_step": 173338, "epoch": 1947, "lr": 2.9009971978594253e-05}
{"train_loss": 0.11223935335874557, "global_step": 173339, "epoch": 1947, "lr": 2.900944580156911e-05}
{"train_loss": 0.13367772102355957, "global_step": 173340, "epoch": 1947, "lr": 2.9008919627365884e-05}
{"train_loss": 0.07396242022514343, "global_step": 173341, "epoch": 1947, "lr": 2.9008393455984662e-05}
{"train_loss": 0.0636645033955574, "global_step": 173342, "epoch": 1947, "lr": 2.9007867287425516e-05}
{"train_loss": 0.09321178495883942, "global_step": 173343, "epoch": 1947, "lr": 2.900734112168848e-05}
{"train_loss": 0.0843886286020279, "global_step": 173344, "epoch": 1947, "lr": 2.900681495877367e-05}
{"train_loss": 0.0704248696565628, "global_step": 173345, "epoch": 1947, "lr": 2.900628879868112e-05}
{"train_loss": 0.0902857780456543, "global_step": 173346, "epoch": 1947, "lr": 2.9005762641410938e-05}
{"train_loss": 0.06958802044391632, "global_step": 173347, "epoch": 1947, "lr": 2.9005236486963178e-05}
{"train_loss": 0.11912067234516144, "global_step": 173348, "epoch": 1947, "lr": 2.900471033533788e-05}
{"train_loss": 0.06127655878663063, "global_step": 173349, "epoch": 1947, "lr": 2.9004184186535156e-05}
{"train_loss": 0.10249240696430206, "global_step": 173350, "epoch": 1947, "lr": 2.900365804055507e-05}
{"train_loss": 0.0691712498664856, "global_step": 173351, "epoch": 1947, "lr": 2.9003131897397674e-05}
{"train_loss": 0.07392143458127975, "global_step": 173352, "epoch": 1947, "lr": 2.9002605757063072e-05}
{"train_loss": 0.07432573288679123, "global_step": 173353, "epoch": 1947, "lr": 2.9002079619551293e-05}
{"train_loss": 0.09247618913650513, "global_step": 173354, "epoch": 1947, "lr": 2.9001553484862444e-05}
{"train_loss": 0.06954357028007507, "global_step": 173355, "epoch": 1947, "lr": 2.9001027352996562e-05}
{"train_loss": 0.047253407537937164, "global_step": 173356, "epoch": 1947, "lr": 2.9000501223953757e-05}
{"train_loss": 0.045794080942869186, "global_step": 173357, "epoch": 1947, "lr": 2.8999975097734055e-05}
{"train_loss": 0.06349048763513565, "global_step": 173358, "epoch": 1947, "lr": 2.8999448974337575e-05}
{"train_loss": 0.06470557302236557, "global_step": 173359, "epoch": 1947, "lr": 2.8998922853764343e-05}
{"train_loss": 0.05935750529170036, "global_step": 173360, "epoch": 1947, "lr": 2.8998396736014466e-05}
{"train_loss": 0.07772690057754517, "global_step": 173361, "epoch": 1947, "lr": 2.8997870621087986e-05}
{"train_loss": 0.10924402624368668, "global_step": 173362, "epoch": 1947, "lr": 2.8997344508985003e-05}
{"train_loss": 0.04868946224451065, "global_step": 173363, "epoch": 1947, "lr": 2.899681839970555e-05}
{"train_loss": 0.06551710516214371, "global_step": 173364, "epoch": 1947, "lr": 2.899629229324974e-05}
{"train_loss": 0.12043235450983047, "global_step": 173365, "epoch": 1947, "lr": 2.8995766189617597e-05}
{"train_loss": 0.09504902362823486, "global_step": 173366, "epoch": 1947, "lr": 2.899524008880924e-05}
{"train_loss": 0.04093281924724579, "global_step": 173367, "epoch": 1947, "lr": 2.8994713990824696e-05}
{"train_loss": 0.1179012805223465, "global_step": 173368, "epoch": 1947, "lr": 2.899418789566406e-05}
{"train_loss": 0.07303214073181152, "global_step": 173369, "epoch": 1947, "lr": 2.8993661803327423e-05}
{"train_loss": 0.04373486712574959, "global_step": 173370, "epoch": 1947, "lr": 2.89931357138148e-05}
{"train_loss": 0.08129439448540131, "global_step": 173371, "epoch": 1947, "lr": 2.8992609627126327e-05, "val_loss": 7.405384540557861}
{"train_loss": 0.033824220299720764, "global_step": 173372, "epoch": 1948, "lr": 2.8992083543262016e-05}
{"train_loss": 0.15267068147659302, "global_step": 173373, "epoch": 1948, "lr": 2.8991557462221984e-05}
{"train_loss": 0.07371516525745392, "global_step": 173374, "epoch": 1948, "lr": 2.899103138400626e-05}
{"train_loss": 0.08542903512716293, "global_step": 173375, "epoch": 1948, "lr": 2.8990505308614957e-05}
{"train_loss": 0.0536080040037632, "global_step": 173376, "epoch": 1948, "lr": 2.8989979236048104e-05}
{"train_loss": 0.04778290167450905, "global_step": 173377, "epoch": 1948, "lr": 2.8989453166305807e-05}
{"train_loss": 0.0522783026099205, "global_step": 173378, "epoch": 1948, "lr": 2.898892709938811e-05}
{"train_loss": 0.10746829211711884, "global_step": 173379, "epoch": 1948, "lr": 2.898840103529511e-05}
{"train_loss": 0.10115116834640503, "global_step": 173380, "epoch": 1948, "lr": 2.8987874974026845e-05}
{"train_loss": 0.08356965333223343, "global_step": 173381, "epoch": 1948, "lr": 2.8987348915583423e-05}
{"train_loss": 0.013328212313354015, "global_step": 173382, "epoch": 1948, "lr": 2.8986822859964903e-05}
{"train_loss": 0.04586624354124069, "global_step": 173383, "epoch": 1948, "lr": 2.898629680717132e-05}
{"train_loss": 0.07824941724538803, "global_step": 173384, "epoch": 1948, "lr": 2.8985770757202797e-05}
{"train_loss": 0.0447368323802948, "global_step": 173385, "epoch": 1948, "lr": 2.8985244710059363e-05}
{"train_loss": 0.0387759655714035, "global_step": 173386, "epoch": 1948, "lr": 2.8984718665741123e-05}
{"train_loss": 0.05742267891764641, "global_step": 173387, "epoch": 1948, "lr": 2.8984192624248108e-05}
{"train_loss": 0.06573192775249481, "global_step": 173388, "epoch": 1948, "lr": 2.8983666585580438e-05}
{"train_loss": 0.07133408635854721, "global_step": 173389, "epoch": 1948, "lr": 2.8983140549738142e-05}
{"train_loss": 0.12756766378879547, "global_step": 173390, "epoch": 1948, "lr": 2.8982614516721317e-05}
{"train_loss": 0.03650951012969017, "global_step": 173391, "epoch": 1948, "lr": 2.8982088486530008e-05}
{"train_loss": 0.08315333724021912, "global_step": 173392, "epoch": 1948, "lr": 2.8981562459164323e-05}
{"train_loss": 0.1172475516796112, "global_step": 173393, "epoch": 1948, "lr": 2.898103643462429e-05}
{"train_loss": 0.067960225045681, "global_step": 173394, "epoch": 1948, "lr": 2.898051041291001e-05}
{"train_loss": 0.0887971818447113, "global_step": 173395, "epoch": 1948, "lr": 2.8979984394021536e-05}
{"train_loss": 0.08375858515501022, "global_step": 173396, "epoch": 1948, "lr": 2.8979458377958967e-05}
{"train_loss": 0.15899457037448883, "global_step": 173397, "epoch": 1948, "lr": 2.8978932364722327e-05}
{"train_loss": 0.04550662636756897, "global_step": 173398, "epoch": 1948, "lr": 2.8978406354311737e-05}
{"train_loss": 0.10370875895023346, "global_step": 173399, "epoch": 1948, "lr": 2.8977880346727226e-05}
{"train_loss": 0.08283675462007523, "global_step": 173400, "epoch": 1948, "lr": 2.8977354341968897e-05}
{"train_loss": 0.10316862165927887, "global_step": 173401, "epoch": 1948, "lr": 2.8976828340036793e-05}
{"train_loss": 0.08370369672775269, "global_step": 173402, "epoch": 1948, "lr": 2.8976302340931017e-05}
{"train_loss": 0.072760671377182, "global_step": 173403, "epoch": 1948, "lr": 2.8975776344651596e-05}
{"train_loss": 0.04055339843034744, "global_step": 173404, "epoch": 1948, "lr": 2.897525035119863e-05}
{"train_loss": 0.10294660925865173, "global_step": 173405, "epoch": 1948, "lr": 2.8974724360572203e-05}
{"train_loss": 0.13189195096492767, "global_step": 173406, "epoch": 1948, "lr": 2.8974198372772355e-05}
{"train_loss": 0.09874146431684494, "global_step": 173407, "epoch": 1948, "lr": 2.8973672387799178e-05}
{"train_loss": 0.09096595644950867, "global_step": 173408, "epoch": 1948, "lr": 2.8973146405652723e-05}
{"train_loss": 0.09034860879182816, "global_step": 173409, "epoch": 1948, "lr": 2.8972620426333084e-05}
{"train_loss": 0.06550288200378418, "global_step": 173410, "epoch": 1948, "lr": 2.897209444984031e-05}
{"train_loss": 0.10444720834493637, "global_step": 173411, "epoch": 1948, "lr": 2.8971568476174497e-05}
{"train_loss": 0.15972495079040527, "global_step": 173412, "epoch": 1948, "lr": 2.8971042505335678e-05}
{"train_loss": 0.15581320226192474, "global_step": 173413, "epoch": 1948, "lr": 2.8970516537323973e-05}
{"train_loss": 0.0776815190911293, "global_step": 173414, "epoch": 1948, "lr": 2.89699905721394e-05}
{"train_loss": 0.14423535764217377, "global_step": 173415, "epoch": 1948, "lr": 2.8969464609782077e-05}
{"train_loss": 0.10115829855203629, "global_step": 173416, "epoch": 1948, "lr": 2.8968938650252032e-05}
{"train_loss": 0.04610845819115639, "global_step": 173417, "epoch": 1948, "lr": 2.8968412693549373e-05}
{"train_loss": 0.06227487698197365, "global_step": 173418, "epoch": 1948, "lr": 2.8967886739674143e-05}
{"train_loss": 0.10725844651460648, "global_step": 173419, "epoch": 1948, "lr": 2.896736078862644e-05}
{"train_loss": 0.09063646197319031, "global_step": 173420, "epoch": 1948, "lr": 2.8966834840406322e-05}
{"train_loss": 0.12128382176160812, "global_step": 173421, "epoch": 1948, "lr": 2.8966308895013833e-05}
{"train_loss": 0.053924962878227234, "global_step": 173422, "epoch": 1948, "lr": 2.8965782952449088e-05}
{"train_loss": 0.16371910274028778, "global_step": 173423, "epoch": 1948, "lr": 2.8965257012712122e-05}
{"train_loss": 0.10778042674064636, "global_step": 173424, "epoch": 1948, "lr": 2.896473107580303e-05}
{"train_loss": 0.14007781445980072, "global_step": 173425, "epoch": 1948, "lr": 2.8964205141721863e-05}
{"train_loss": 0.11278285086154938, "global_step": 173426, "epoch": 1948, "lr": 2.896367921046872e-05}
{"train_loss": 0.08267220109701157, "global_step": 173427, "epoch": 1948, "lr": 2.896315328204363e-05}
{"train_loss": 0.09315680712461472, "global_step": 173428, "epoch": 1948, "lr": 2.8962627356446714e-05}
{"train_loss": 0.12206678092479706, "global_step": 173429, "epoch": 1948, "lr": 2.896210143367799e-05}
{"train_loss": 0.08618085086345673, "global_step": 173430, "epoch": 1948, "lr": 2.8961575513737584e-05}
{"train_loss": 0.08773568272590637, "global_step": 173431, "epoch": 1948, "lr": 2.8961049596625512e-05}
{"train_loss": 0.0898890569806099, "global_step": 173432, "epoch": 1948, "lr": 2.896052368234189e-05}
{"train_loss": 0.04773823171854019, "global_step": 173433, "epoch": 1948, "lr": 2.8959997770886748e-05}
{"train_loss": 0.14194080233573914, "global_step": 173434, "epoch": 1948, "lr": 2.89594718622602e-05}
{"train_loss": 0.05947064980864525, "global_step": 173435, "epoch": 1948, "lr": 2.895894595646227e-05}
{"train_loss": 0.05784904584288597, "global_step": 173436, "epoch": 1948, "lr": 2.895842005349308e-05}
{"train_loss": 0.11472341418266296, "global_step": 173437, "epoch": 1948, "lr": 2.8957894153352648e-05}
{"train_loss": 0.11725286394357681, "global_step": 173438, "epoch": 1948, "lr": 2.8957368256041096e-05}
{"train_loss": 0.0574696883559227, "global_step": 173439, "epoch": 1948, "lr": 2.8956842361558445e-05}
{"train_loss": 0.05852539464831352, "global_step": 173440, "epoch": 1948, "lr": 2.89563164699048e-05}
{"train_loss": 0.10289330035448074, "global_step": 173441, "epoch": 1948, "lr": 2.8955790581080232e-05}
{"train_loss": 0.04824531823396683, "global_step": 173442, "epoch": 1948, "lr": 2.8955264695084783e-05}
{"train_loss": 0.08109071850776672, "global_step": 173443, "epoch": 1948, "lr": 2.895473881191856e-05}
{"train_loss": 0.04961014166474342, "global_step": 173444, "epoch": 1948, "lr": 2.8954212931581605e-05}
{"train_loss": 0.06535196304321289, "global_step": 173445, "epoch": 1948, "lr": 2.895368705407401e-05}
{"train_loss": 0.10543369501829147, "global_step": 173446, "epoch": 1948, "lr": 2.8953161179395814e-05}
{"train_loss": 0.10856705158948898, "global_step": 173447, "epoch": 1948, "lr": 2.8952635307547137e-05}
{"train_loss": 0.1094096451997757, "global_step": 173448, "epoch": 1948, "lr": 2.8952109438527996e-05}
{"train_loss": 0.10132687538862228, "global_step": 173449, "epoch": 1948, "lr": 2.8951583572338508e-05}
{"train_loss": 0.1060650572180748, "global_step": 173450, "epoch": 1948, "lr": 2.8951057708978707e-05}
{"train_loss": 0.0972028523683548, "global_step": 173451, "epoch": 1948, "lr": 2.8950531848448697e-05}
{"train_loss": 0.08895698934793472, "global_step": 173452, "epoch": 1948, "lr": 2.895000599074851e-05}
{"train_loss": 0.112469881772995, "global_step": 173453, "epoch": 1948, "lr": 2.894948013587826e-05}
{"train_loss": 0.13018080592155457, "global_step": 173454, "epoch": 1948, "lr": 2.8948954283837977e-05}
{"train_loss": 0.08548285067081451, "global_step": 173455, "epoch": 1948, "lr": 2.8948428434627773e-05}
{"train_loss": 0.08518481254577637, "global_step": 173456, "epoch": 1948, "lr": 2.8947902588247665e-05}
{"train_loss": 0.16533741354942322, "global_step": 173457, "epoch": 1948, "lr": 2.8947376744697785e-05}
{"train_loss": 0.06311371922492981, "global_step": 173458, "epoch": 1948, "lr": 2.8946850903978175e-05}
{"train_loss": 0.11025570333003998, "global_step": 173459, "epoch": 1948, "lr": 2.8946325066088874e-05}
{"train_loss": 0.08907850366085768, "global_step": 173460, "epoch": 1948, "lr": 2.8945799231030012e-05, "val_loss": 7.2420501708984375}
{"train_loss": 0.05842912197113037, "global_step": 173461, "epoch": 1949, "lr": 2.894527339880161e-05}
{"train_loss": 0.0851898044347763, "global_step": 173462, "epoch": 1949, "lr": 2.8944747569403773e-05}
{"train_loss": 0.07390043884515762, "global_step": 173463, "epoch": 1949, "lr": 2.894422174283654e-05}
{"train_loss": 0.11130546778440475, "global_step": 173464, "epoch": 1949, "lr": 2.8943695919100023e-05}
{"train_loss": 0.11258405447006226, "global_step": 173465, "epoch": 1949, "lr": 2.894317009819425e-05}
{"train_loss": 0.06277155876159668, "global_step": 173466, "epoch": 1949, "lr": 2.894264428011933e-05}
{"train_loss": 0.09266511350870132, "global_step": 173467, "epoch": 1949, "lr": 2.8942118464875295e-05}
{"train_loss": 0.0694044753909111, "global_step": 173468, "epoch": 1949, "lr": 2.8941592652462258e-05}
{"train_loss": 0.05358681082725525, "global_step": 173469, "epoch": 1949, "lr": 2.8941066842880236e-05}
{"train_loss": 0.10427026450634003, "global_step": 173470, "epoch": 1949, "lr": 2.8940541036129364e-05}
{"train_loss": 0.08092671632766724, "global_step": 173471, "epoch": 1949, "lr": 2.894001523220965e-05}
{"train_loss": 0.08669720590114594, "global_step": 173472, "epoch": 1949, "lr": 2.8939489431121226e-05}
{"train_loss": 0.059249937534332275, "global_step": 173473, "epoch": 1949, "lr": 2.8938963632864103e-05}
{"train_loss": 0.11378847062587738, "global_step": 173474, "epoch": 1949, "lr": 2.8938437837438397e-05}
{"train_loss": 0.1054854691028595, "global_step": 173475, "epoch": 1949, "lr": 2.8937912044844146e-05}
{"train_loss": 0.08375374972820282, "global_step": 173476, "epoch": 1949, "lr": 2.8937386255081435e-05}
{"train_loss": 0.1029757633805275, "global_step": 173477, "epoch": 1949, "lr": 2.893686046815035e-05}
{"train_loss": 0.0656774565577507, "global_step": 173478, "epoch": 1949, "lr": 2.8936334684050936e-05}
{"train_loss": 0.08591970801353455, "global_step": 173479, "epoch": 1949, "lr": 2.893580890278329e-05}
{"train_loss": 0.13377885520458221, "global_step": 173480, "epoch": 1949, "lr": 2.893528312434745e-05}
{"train_loss": 0.06469843536615372, "global_step": 173481, "epoch": 1949, "lr": 2.8934757348743524e-05}
{"train_loss": 0.06467092037200928, "global_step": 173482, "epoch": 1949, "lr": 2.8934231575971548e-05}
{"train_loss": 0.04932266101241112, "global_step": 173483, "epoch": 1949, "lr": 2.893370580603162e-05}
{"train_loss": 0.10513991117477417, "global_step": 173484, "epoch": 1949, "lr": 2.893318003892378e-05}
{"train_loss": 0.06344106793403625, "global_step": 173485, "epoch": 1949, "lr": 2.8932654274648142e-05}
{"train_loss": 0.05730830132961273, "global_step": 173486, "epoch": 1949, "lr": 2.8932128513204727e-05}
{"train_loss": 0.062060534954071045, "global_step": 173487, "epoch": 1949, "lr": 2.8931602754593658e-05}
{"train_loss": 0.12961237132549286, "global_step": 173488, "epoch": 1949, "lr": 2.8931076998814942e-05}
{"train_loss": 0.1827586442232132, "global_step": 173489, "epoch": 1949, "lr": 2.893055124586872e-05}
{"train_loss": 0.0909377783536911, "global_step": 173490, "epoch": 1949, "lr": 2.893002549575501e-05}
{"train_loss": 0.12192895263433456, "global_step": 173491, "epoch": 1949, "lr": 2.892949974847391e-05}
{"train_loss": 0.07633651793003082, "global_step": 173492, "epoch": 1949, "lr": 2.8928974004025466e-05}
{"train_loss": 0.04227486997842789, "global_step": 173493, "epoch": 1949, "lr": 2.892844826240979e-05}
{"train_loss": 0.07441503554582596, "global_step": 173494, "epoch": 1949, "lr": 2.8927922523626906e-05}
{"train_loss": 0.07605738937854767, "global_step": 173495, "epoch": 1949, "lr": 2.892739678767692e-05}
{"train_loss": 0.07183205336332321, "global_step": 173496, "epoch": 1949, "lr": 2.892687105455988e-05}
{"train_loss": 0.08891841024160385, "global_step": 173497, "epoch": 1949, "lr": 2.8926345324275872e-05}
{"train_loss": 0.15021558105945587, "global_step": 173498, "epoch": 1949, "lr": 2.8925819596824967e-05}
{"train_loss": 0.058249350637197495, "global_step": 173499, "epoch": 1949, "lr": 2.8925293872207203e-05}
{"train_loss": 0.08928816765546799, "global_step": 173500, "epoch": 1949, "lr": 2.8924768150422704e-05}
{"train_loss": 0.09855988621711731, "global_step": 173501, "epoch": 1949, "lr": 2.8924242431471494e-05}
{"train_loss": 0.07012586295604706, "global_step": 173502, "epoch": 1949, "lr": 2.8923716715353676e-05}
{"train_loss": 0.04228714480996132, "global_step": 173503, "epoch": 1949, "lr": 2.8923191002069295e-05}
{"train_loss": 0.0999244749546051, "global_step": 173504, "epoch": 1949, "lr": 2.8922665291618445e-05}
{"train_loss": 0.09182695299386978, "global_step": 173505, "epoch": 1949, "lr": 2.8922139584001172e-05}
{"train_loss": 0.08627469092607498, "global_step": 173506, "epoch": 1949, "lr": 2.892161387921758e-05}
{"train_loss": 0.09992567449808121, "global_step": 173507, "epoch": 1949, "lr": 2.8921088177267696e-05}
{"train_loss": 0.10017810016870499, "global_step": 173508, "epoch": 1949, "lr": 2.8920562478151635e-05}
{"train_loss": 0.09885118156671524, "global_step": 173509, "epoch": 1949, "lr": 2.892003678186943e-05}
{"train_loss": 0.10294771194458008, "global_step": 173510, "epoch": 1949, "lr": 2.8919511088421193e-05}
{"train_loss": 0.15659557282924652, "global_step": 173511, "epoch": 1949, "lr": 2.891898539780694e-05}
{"train_loss": 0.10510705411434174, "global_step": 173512, "epoch": 1949, "lr": 2.8918459710026803e-05}
{"train_loss": 0.05273739993572235, "global_step": 173513, "epoch": 1949, "lr": 2.8917934025080794e-05}
{"train_loss": 0.10210572183132172, "global_step": 173514, "epoch": 1949, "lr": 2.891740834296902e-05}
{"train_loss": 0.09959394484758377, "global_step": 173515, "epoch": 1949, "lr": 2.891688266369156e-05}
{"train_loss": 0.09299466758966446, "global_step": 173516, "epoch": 1949, "lr": 2.891635698724845e-05}
{"train_loss": 0.07520965486764908, "global_step": 173517, "epoch": 1949, "lr": 2.8915831313639797e-05}
{"train_loss": 0.10434232652187347, "global_step": 173518, "epoch": 1949, "lr": 2.891530564286563e-05}
{"train_loss": 0.10625317692756653, "global_step": 173519, "epoch": 1949, "lr": 2.8914779974926065e-05}
{"train_loss": 0.11177580058574677, "global_step": 173520, "epoch": 1949, "lr": 2.891425430982113e-05}
{"train_loss": 0.10296773165464401, "global_step": 173521, "epoch": 1949, "lr": 2.8913728647550943e-05}
{"train_loss": 0.0766763761639595, "global_step": 173522, "epoch": 1949, "lr": 2.8913202988115522e-05}
{"train_loss": 0.047891877591609955, "global_step": 173523, "epoch": 1949, "lr": 2.8912677331514986e-05}
{"train_loss": 0.07803639024496078, "global_step": 173524, "epoch": 1949, "lr": 2.8912151677749367e-05}
{"train_loss": 0.09154536575078964, "global_step": 173525, "epoch": 1949, "lr": 2.891162602681876e-05}
{"train_loss": 0.0422464981675148, "global_step": 173526, "epoch": 1949, "lr": 2.891110037872322e-05}
{"train_loss": 0.09421779960393906, "global_step": 173527, "epoch": 1949, "lr": 2.891057473346284e-05}
{"train_loss": 0.11608666181564331, "global_step": 173528, "epoch": 1949, "lr": 2.891004909103766e-05}
{"train_loss": 0.09997936338186264, "global_step": 173529, "epoch": 1949, "lr": 2.8909523451447783e-05}
{"train_loss": 0.2219320833683014, "global_step": 173530, "epoch": 1949, "lr": 2.8908997814693252e-05}
{"train_loss": 0.10402286052703857, "global_step": 173531, "epoch": 1949, "lr": 2.890847218077416e-05}
{"train_loss": 0.11097511649131775, "global_step": 173532, "epoch": 1949, "lr": 2.8907946549690546e-05}
{"train_loss": 0.07798963040113449, "global_step": 173533, "epoch": 1949, "lr": 2.890742092144253e-05}
{"train_loss": 0.10839322209358215, "global_step": 173534, "epoch": 1949, "lr": 2.8906895296030123e-05}
{"train_loss": 0.04007274657487869, "global_step": 173535, "epoch": 1949, "lr": 2.8906369673453453e-05}
{"train_loss": 0.08976998180150986, "global_step": 173536, "epoch": 1949, "lr": 2.8905844053712566e-05}
{"train_loss": 0.08958417922258377, "global_step": 173537, "epoch": 1949, "lr": 2.890531843680751e-05}
{"train_loss": 0.08572947978973389, "global_step": 173538, "epoch": 1949, "lr": 2.8904792822738395e-05}
{"train_loss": 0.09577200561761856, "global_step": 173539, "epoch": 1949, "lr": 2.8904267211505255e-05}
{"train_loss": 0.16017144918441772, "global_step": 173540, "epoch": 1949, "lr": 2.8903741603108202e-05}
{"train_loss": 0.11569298803806305, "global_step": 173541, "epoch": 1949, "lr": 2.8903215997547262e-05}
{"train_loss": 0.089785635471344, "global_step": 173542, "epoch": 1949, "lr": 2.890269039482254e-05}
{"train_loss": 0.11638979613780975, "global_step": 173543, "epoch": 1949, "lr": 2.8902164794934083e-05}
{"train_loss": 0.1252819150686264, "global_step": 173544, "epoch": 1949, "lr": 2.8901639197881992e-05}
{"train_loss": 0.06459707766771317, "global_step": 173545, "epoch": 1949, "lr": 2.8901113603666296e-05}
{"train_loss": 0.07230020314455032, "global_step": 173546, "epoch": 1949, "lr": 2.890058801228711e-05}
{"train_loss": 0.07451216131448746, "global_step": 173547, "epoch": 1949, "lr": 2.8900062423744455e-05}
{"train_loss": 0.09506283700466156, "global_step": 173548, "epoch": 1949, "lr": 2.889953683803846e-05}
{"train_loss": 0.09104901602428951, "global_step": 173549, "epoch": 1949, "lr": 2.889901125516914e-05, "val_loss": 7.24465799331665}
{"train_loss": 0.10527776926755905, "global_step": 173550, "epoch": 1950, "lr": 2.8898485675136593e-05}
{"train_loss": 0.1200353130698204, "global_step": 173551, "epoch": 1950, "lr": 2.8897960097940906e-05}
{"train_loss": 0.07027314603328705, "global_step": 173552, "epoch": 1950, "lr": 2.8897434523582112e-05}
{"train_loss": 0.060138069093227386, "global_step": 173553, "epoch": 1950, "lr": 2.889690895206032e-05}
{"train_loss": 0.09372398257255554, "global_step": 173554, "epoch": 1950, "lr": 2.889638338337556e-05}
{"train_loss": 0.06564319133758545, "global_step": 173555, "epoch": 1950, "lr": 2.889585781752795e-05}
{"train_loss": 0.05298258736729622, "global_step": 173556, "epoch": 1950, "lr": 2.8895332254517505e-05}
{"train_loss": 0.06660544127225876, "global_step": 173557, "epoch": 1950, "lr": 2.8894806694344355e-05}
{"train_loss": 0.10799435526132584, "global_step": 173558, "epoch": 1950, "lr": 2.8894281137008518e-05}
{"train_loss": 0.08855956047773361, "global_step": 173559, "epoch": 1950, "lr": 2.8893755582510102e-05}
{"train_loss": 0.09796805679798126, "global_step": 173560, "epoch": 1950, "lr": 2.889323003084915e-05}
{"train_loss": 0.10454526543617249, "global_step": 173561, "epoch": 1950, "lr": 2.8892704482025762e-05}
{"train_loss": 0.06586349010467529, "global_step": 173562, "epoch": 1950, "lr": 2.889217893603997e-05}
{"train_loss": 0.0926288440823555, "global_step": 173563, "epoch": 1950, "lr": 2.8891653392891894e-05}
{"train_loss": 0.09172496199607849, "global_step": 173564, "epoch": 1950, "lr": 2.8891127852581558e-05}
{"train_loss": 0.04653947800397873, "global_step": 173565, "epoch": 1950, "lr": 2.8890602315109067e-05}
{"train_loss": 0.04259125143289566, "global_step": 173566, "epoch": 1950, "lr": 2.889007678047446e-05}
{"train_loss": 0.05634520575404167, "global_step": 173567, "epoch": 1950, "lr": 2.8889551248677826e-05}
{"train_loss": 0.055369988083839417, "global_step": 173568, "epoch": 1950, "lr": 2.8889025719719253e-05}
{"train_loss": 0.0889904797077179, "global_step": 173569, "epoch": 1950, "lr": 2.8888500193598795e-05}
{"train_loss": 0.05860327556729317, "global_step": 173570, "epoch": 1950, "lr": 2.8887974670316498e-05}
{"train_loss": 0.10257419943809509, "global_step": 173571, "epoch": 1950, "lr": 2.8887449149872476e-05}
{"train_loss": 0.1075209304690361, "global_step": 173572, "epoch": 1950, "lr": 2.888692363226676e-05}
{"train_loss": 0.07021108269691467, "global_step": 173573, "epoch": 1950, "lr": 2.8886398117499458e-05}
{"train_loss": 0.06717737764120102, "global_step": 173574, "epoch": 1950, "lr": 2.888587260557063e-05}
{"train_loss": 0.09726552665233612, "global_step": 173575, "epoch": 1950, "lr": 2.888534709648031e-05}
{"train_loss": 0.09979411214590073, "global_step": 173576, "epoch": 1950, "lr": 2.8884821590228623e-05}
{"train_loss": 0.054532259702682495, "global_step": 173577, "epoch": 1950, "lr": 2.8884296086815588e-05}
{"train_loss": 0.05208759382367134, "global_step": 173578, "epoch": 1950, "lr": 2.8883770586241333e-05}
{"train_loss": 0.15706197917461395, "global_step": 173579, "epoch": 1950, "lr": 2.8883245088505868e-05}
{"train_loss": 0.0880332812666893, "global_step": 173580, "epoch": 1950, "lr": 2.8882719593609318e-05}
{"train_loss": 0.07740527391433716, "global_step": 173581, "epoch": 1950, "lr": 2.8882194101551708e-05}
{"train_loss": 0.11568224430084229, "global_step": 173582, "epoch": 1950, "lr": 2.8881668612333146e-05}
{"train_loss": 0.0886949747800827, "global_step": 173583, "epoch": 1950, "lr": 2.888114312595367e-05}
{"train_loss": 0.10549668222665787, "global_step": 173584, "epoch": 1950, "lr": 2.8880617642413388e-05}
{"train_loss": 0.12129663676023483, "global_step": 173585, "epoch": 1950, "lr": 2.8880092161712325e-05}
{"train_loss": 0.049041371792554855, "global_step": 173586, "epoch": 1950, "lr": 2.8879566683850577e-05}
{"train_loss": 0.05624203756451607, "global_step": 173587, "epoch": 1950, "lr": 2.887904120882824e-05}
{"train_loss": 0.0920272246003151, "global_step": 173588, "epoch": 1950, "lr": 2.8878515736645332e-05}
{"train_loss": 0.10302697867155075, "global_step": 173589, "epoch": 1950, "lr": 2.8877990267301975e-05}
{"train_loss": 0.05645079165697098, "global_step": 173590, "epoch": 1950, "lr": 2.8877464800798193e-05}
{"train_loss": 0.05641699582338333, "global_step": 173591, "epoch": 1950, "lr": 2.8876939337134097e-05}
{"train_loss": 0.05697540193796158, "global_step": 173592, "epoch": 1950, "lr": 2.8876413876309715e-05}
{"train_loss": 0.05548270419239998, "global_step": 173593, "epoch": 1950, "lr": 2.887588841832517e-05}
{"train_loss": 0.09259410202503204, "global_step": 173594, "epoch": 1950, "lr": 2.887536296318048e-05}
{"train_loss": 0.050174158066511154, "global_step": 173595, "epoch": 1950, "lr": 2.8874837510875763e-05}
{"train_loss": 0.10499605536460876, "global_step": 173596, "epoch": 1950, "lr": 2.8874312061411046e-05}
{"train_loss": 0.09019271284341812, "global_step": 173597, "epoch": 1950, "lr": 2.8873786614786437e-05}
{"train_loss": 0.10870706290006638, "global_step": 173598, "epoch": 1950, "lr": 2.887326117100197e-05}
{"train_loss": 0.08888130635023117, "global_step": 173599, "epoch": 1950, "lr": 2.8872735730057753e-05}
{"train_loss": 0.07705415785312653, "global_step": 173600, "epoch": 1950, "lr": 2.8872210291953827e-05}
{"train_loss": 0.09787872433662415, "global_step": 173601, "epoch": 1950, "lr": 2.887168485669029e-05}
{"train_loss": 0.1059688851237297, "global_step": 173602, "epoch": 1950, "lr": 2.887115942426718e-05}
{"train_loss": 0.1002984493970871, "global_step": 173603, "epoch": 1950, "lr": 2.887063399468458e-05}
{"train_loss": 0.06645987182855606, "global_step": 173604, "epoch": 1950, "lr": 2.887010856794259e-05}
{"train_loss": 0.08215292543172836, "global_step": 173605, "epoch": 1950, "lr": 2.8869583144041235e-05}
{"train_loss": 0.10304344445466995, "global_step": 173606, "epoch": 1950, "lr": 2.886905772298063e-05}
{"train_loss": 0.12002797424793243, "global_step": 173607, "epoch": 1950, "lr": 2.88685323047608e-05}
{"train_loss": 0.07048913091421127, "global_step": 173608, "epoch": 1950, "lr": 2.886800688938186e-05}
{"train_loss": 0.07334107160568237, "global_step": 173609, "epoch": 1950, "lr": 2.886748147684386e-05}
{"train_loss": 0.09689495712518692, "global_step": 173610, "epoch": 1950, "lr": 2.8866956067146846e-05}
{"train_loss": 0.022873708978295326, "global_step": 173611, "epoch": 1950, "lr": 2.886643066029093e-05}
{"train_loss": 0.11223840713500977, "global_step": 173612, "epoch": 1950, "lr": 2.886590525627615e-05}
{"train_loss": 0.08414489030838013, "global_step": 173613, "epoch": 1950, "lr": 2.886537985510261e-05}
{"train_loss": 0.07636401057243347, "global_step": 173614, "epoch": 1950, "lr": 2.8864854456770363e-05}
{"train_loss": 0.04095900058746338, "global_step": 173615, "epoch": 1950, "lr": 2.8864329061279456e-05}
{"train_loss": 0.08215387910604477, "global_step": 173616, "epoch": 1950, "lr": 2.8863803668630002e-05}
{"train_loss": 0.047539472579956055, "global_step": 173617, "epoch": 1950, "lr": 2.886327827882204e-05}
{"train_loss": 0.08588483929634094, "global_step": 173618, "epoch": 1950, "lr": 2.886275289185566e-05}
{"train_loss": 0.08820294588804245, "global_step": 173619, "epoch": 1950, "lr": 2.886222750773091e-05}
{"train_loss": 0.09478334337472916, "global_step": 173620, "epoch": 1950, "lr": 2.8861702126447898e-05}
{"train_loss": 0.09366224706172943, "global_step": 173621, "epoch": 1950, "lr": 2.8861176748006652e-05}
{"train_loss": 0.054306626319885254, "global_step": 173622, "epoch": 1950, "lr": 2.8860651372407267e-05}
{"train_loss": 0.06991303712129593, "global_step": 173623, "epoch": 1950, "lr": 2.8860125999649823e-05}
{"train_loss": 0.1298089176416397, "global_step": 173624, "epoch": 1950, "lr": 2.8859600629734363e-05}
{"train_loss": 0.11426602303981781, "global_step": 173625, "epoch": 1950, "lr": 2.8859075262660984e-05}
{"train_loss": 0.054617881774902344, "global_step": 173626, "epoch": 1950, "lr": 2.8858549898429733e-05}
{"train_loss": 0.11226369440555573, "global_step": 173627, "epoch": 1950, "lr": 2.8858024537040708e-05}
{"train_loss": 0.08313503116369247, "global_step": 173628, "epoch": 1950, "lr": 2.8857499178493942e-05}
{"train_loss": 0.09120990335941315, "global_step": 173629, "epoch": 1950, "lr": 2.885697382278955e-05}
{"train_loss": 0.07913899421691895, "global_step": 173630, "epoch": 1950, "lr": 2.8856448469927565e-05}
{"train_loss": 0.11045828461647034, "global_step": 173631, "epoch": 1950, "lr": 2.885592311990809e-05}
{"train_loss": 0.11308227479457855, "global_step": 173632, "epoch": 1950, "lr": 2.8855397772731153e-05}
{"train_loss": 0.16651523113250732, "global_step": 173633, "epoch": 1950, "lr": 2.885487242839687e-05}
{"train_loss": 0.16138072311878204, "global_step": 173634, "epoch": 1950, "lr": 2.885434708690528e-05}
{"train_loss": 0.09423518180847168, "global_step": 173635, "epoch": 1950, "lr": 2.8853821748256484e-05}
{"train_loss": 0.06587208807468414, "global_step": 173636, "epoch": 1950, "lr": 2.885329641245051e-05}
{"train_loss": 0.1670324206352234, "global_step": 173637, "epoch": 1950, "lr": 2.8852771079487472e-05}
{"train_loss": 0.0860167679199007, "global_step": 173638, "epoch": 1950, "lr": 2.88522457493674e-05, "train/sim_max_reward_0": 0.6527865946456034, "train/sim_max_reward_1": 0.46754255802569544, "train/sim_max_reward_2": 0.39016978673275865, "train/sim_max_reward_3": 0.6825637537341946, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9400995064639462, "test/sim_max_reward_4500000": 0.4130794010812449, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.46202637260149354, "test/sim_max_reward_4500004": 0.7153165013127498, "test/sim_max_reward_4500005": 0.9132408668109532, "test/sim_max_reward_4500006": 0.965208035803526, "test/sim_max_reward_4500007": 0.5202920816300597, "test/sim_max_reward_4500008": 0.12786473677523436, "test/sim_max_reward_4500009": 1.0, "test/sim_max_reward_4500010": 0.9767349570539855, "test/sim_max_reward_4500011": 0.9837585843029091, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5776991968736183, "test/sim_max_reward_4500014": 0.965592063718704, "test/sim_max_reward_4500015": 0.9850943031497218, "test/sim_max_reward_4500016": 0.7130224718481626, "test/sim_max_reward_4500017": 1.0, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9693682847976811, "test/sim_max_reward_4500023": 0.6263955776818865, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9701982276045067, "test/sim_max_reward_4500026": 0.9149967781911682, "test/sim_max_reward_4500027": 0.9714585006876606, "test/sim_max_reward_4500028": 0.6414723042939019, "test/sim_max_reward_4500029": 0.7052495432338589, "test/sim_max_reward_4500030": 0.9885332257314272, "test/sim_max_reward_4500031": 0.4736282506756247, "test/sim_max_reward_4500032": 0.1752284719642417, "test/sim_max_reward_4500033": 0.9865791387258039, "test/sim_max_reward_4500034": 0.927139041237333, "test/sim_max_reward_4500035": 0.38859139562600603, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.16224294614832174, "test/sim_max_reward_4500039": 0.9770571853140805, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.9779515840493186, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.49764038487288376, "test/sim_max_reward_4500044": 0.5138312622844873, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.3639928685473635, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.9856408925168755, "train/mean_score": 0.6888603666003664, "test/mean_score": 0.6707225087429358, "val_loss": 7.212029457092285, "train_action_mse_error": 18.588619232177734}
{"train_loss": 0.08471495658159256, "global_step": 173639, "epoch": 1951, "lr": 2.885172042209041e-05}
{"train_loss": 0.0659014955163002, "global_step": 173640, "epoch": 1951, "lr": 2.8851195097656525e-05}
{"train_loss": 0.09023284912109375, "global_step": 173641, "epoch": 1951, "lr": 2.885066977606584e-05}
{"train_loss": 0.10979879647493362, "global_step": 173642, "epoch": 1951, "lr": 2.8850144457318452e-05}
{"train_loss": 0.07091853022575378, "global_step": 173643, "epoch": 1951, "lr": 2.8849619141414374e-05}
{"train_loss": 0.08197709918022156, "global_step": 173644, "epoch": 1951, "lr": 2.8849093828353735e-05}
{"train_loss": 0.046117447316646576, "global_step": 173645, "epoch": 1951, "lr": 2.8848568518136555e-05}
{"train_loss": 0.0720100924372673, "global_step": 173646, "epoch": 1951, "lr": 2.8848043210762942e-05}
{"train_loss": 0.056786347180604935, "global_step": 173647, "epoch": 1951, "lr": 2.884751790623296e-05}
{"train_loss": 0.0733504667878151, "global_step": 173648, "epoch": 1951, "lr": 2.884699260454664e-05}
{"train_loss": 0.08445656299591064, "global_step": 173649, "epoch": 1951, "lr": 2.884646730570411e-05}
{"train_loss": 0.07845193147659302, "global_step": 173650, "epoch": 1951, "lr": 2.8845942009705394e-05}
{"train_loss": 0.12788322567939758, "global_step": 173651, "epoch": 1951, "lr": 2.8845416716550612e-05}
{"train_loss": 0.08127022534608841, "global_step": 173652, "epoch": 1951, "lr": 2.8844891426239794e-05}
{"train_loss": 0.08295505493879318, "global_step": 173653, "epoch": 1951, "lr": 2.8844366138773004e-05}
{"train_loss": 0.06339171528816223, "global_step": 173654, "epoch": 1951, "lr": 2.884384085415036e-05}
{"train_loss": 0.123357392847538, "global_step": 173655, "epoch": 1951, "lr": 2.8843315572371875e-05}
{"train_loss": 0.17191582918167114, "global_step": 173656, "epoch": 1951, "lr": 2.8842790293437672e-05}
{"train_loss": 0.06626532226800919, "global_step": 173657, "epoch": 1951, "lr": 2.884226501734778e-05}
{"train_loss": 0.09463999420404434, "global_step": 173658, "epoch": 1951, "lr": 2.884173974410228e-05}
{"train_loss": 0.10222989320755005, "global_step": 173659, "epoch": 1951, "lr": 2.8841214473701278e-05}
{"train_loss": 0.038270194083452225, "global_step": 173660, "epoch": 1951, "lr": 2.8840689206144793e-05}
{"train_loss": 0.11541396379470825, "global_step": 173661, "epoch": 1951, "lr": 2.884016394143294e-05}
{"train_loss": 0.11344598233699799, "global_step": 173662, "epoch": 1951, "lr": 2.883963867956575e-05}
{"train_loss": 0.0856667309999466, "global_step": 173663, "epoch": 1951, "lr": 2.883911342054333e-05}
{"train_loss": 0.089541494846344, "global_step": 173664, "epoch": 1951, "lr": 2.883858816436572e-05}
{"train_loss": 0.08494589477777481, "global_step": 173665, "epoch": 1951, "lr": 2.8838062911033016e-05}
{"train_loss": 0.07533702254295349, "global_step": 173666, "epoch": 1951, "lr": 2.883753766054526e-05}
{"train_loss": 0.0665699690580368, "global_step": 173667, "epoch": 1951, "lr": 2.883701241290256e-05}
{"train_loss": 0.11663641035556793, "global_step": 173668, "epoch": 1951, "lr": 2.8836487168104942e-05}
{"train_loss": 0.15650832653045654, "global_step": 173669, "epoch": 1951, "lr": 2.8835961926152523e-05}
{"train_loss": 0.06045189127326012, "global_step": 173670, "epoch": 1951, "lr": 2.8835436687045337e-05}
{"train_loss": 0.09513227641582489, "global_step": 173671, "epoch": 1951, "lr": 2.8834911450783475e-05}
{"train_loss": 0.1027228832244873, "global_step": 173672, "epoch": 1951, "lr": 2.8834386217366992e-05}
{"train_loss": 0.041965894401073456, "global_step": 173673, "epoch": 1951, "lr": 2.8833860986795985e-05}
{"train_loss": 0.052214931696653366, "global_step": 173674, "epoch": 1951, "lr": 2.883333575907049e-05}
{"train_loss": 0.08936922997236252, "global_step": 173675, "epoch": 1951, "lr": 2.8832810534190612e-05}
{"train_loss": 0.09069651365280151, "global_step": 173676, "epoch": 1951, "lr": 2.8832285312156382e-05}
{"train_loss": 0.10375498235225677, "global_step": 173677, "epoch": 1951, "lr": 2.8831760092967898e-05}
{"train_loss": 0.10840494930744171, "global_step": 173678, "epoch": 1951, "lr": 2.883123487662524e-05}
{"train_loss": 0.07455688714981079, "global_step": 173679, "epoch": 1951, "lr": 2.8830709663128453e-05}
{"train_loss": 0.07192022353410721, "global_step": 173680, "epoch": 1951, "lr": 2.8830184452477628e-05}
{"train_loss": 0.08443420380353928, "global_step": 173681, "epoch": 1951, "lr": 2.8829659244672813e-05}
{"train_loss": 0.07532696425914764, "global_step": 173682, "epoch": 1951, "lr": 2.8829134039714113e-05}
{"train_loss": 0.11654400825500488, "global_step": 173683, "epoch": 1951, "lr": 2.8828608837601556e-05}
{"train_loss": 0.08033856004476547, "global_step": 173684, "epoch": 1951, "lr": 2.8828083638335257e-05}
{"train_loss": 0.07692255824804306, "global_step": 173685, "epoch": 1951, "lr": 2.8827558441915266e-05}
{"train_loss": 0.07618304342031479, "global_step": 173686, "epoch": 1951, "lr": 2.8827033248341628e-05}
{"train_loss": 0.08545099198818207, "global_step": 173687, "epoch": 1951, "lr": 2.8826508057614454e-05}
{"train_loss": 0.05102516710758209, "global_step": 173688, "epoch": 1951, "lr": 2.882598286973378e-05}
{"train_loss": 0.05911693722009659, "global_step": 173689, "epoch": 1951, "lr": 2.882545768469972e-05}
{"train_loss": 0.06491106003522873, "global_step": 173690, "epoch": 1951, "lr": 2.8824932502512315e-05}
{"train_loss": 0.06725143641233444, "global_step": 173691, "epoch": 1951, "lr": 2.882440732317162e-05}
{"train_loss": 0.1000140979886055, "global_step": 173692, "epoch": 1951, "lr": 2.8823882146677745e-05}
{"train_loss": 0.1003027856349945, "global_step": 173693, "epoch": 1951, "lr": 2.8823356973030723e-05}
{"train_loss": 0.12343215197324753, "global_step": 173694, "epoch": 1951, "lr": 2.8822831802230636e-05}
{"train_loss": 0.10641604661941528, "global_step": 173695, "epoch": 1951, "lr": 2.8822306634277587e-05}
{"train_loss": 0.05256449431180954, "global_step": 173696, "epoch": 1951, "lr": 2.8821781469171604e-05}
{"train_loss": 0.07003644853830338, "global_step": 173697, "epoch": 1951, "lr": 2.882125630691279e-05}
{"train_loss": 0.12163995951414108, "global_step": 173698, "epoch": 1951, "lr": 2.8820731147501177e-05}
{"train_loss": 0.04455339163541794, "global_step": 173699, "epoch": 1951, "lr": 2.882020599093688e-05}
{"train_loss": 0.08633916825056076, "global_step": 173700, "epoch": 1951, "lr": 2.881968083721993e-05}
{"train_loss": 0.050482116639614105, "global_step": 173701, "epoch": 1951, "lr": 2.8819155686350442e-05}
{"train_loss": 0.0527619794011116, "global_step": 173702, "epoch": 1951, "lr": 2.881863053832843e-05}
{"train_loss": 0.07623619586229324, "global_step": 173703, "epoch": 1951, "lr": 2.8818105393154017e-05}
{"train_loss": 0.08238214999437332, "global_step": 173704, "epoch": 1951, "lr": 2.8817580250827237e-05}
{"train_loss": 0.08344325423240662, "global_step": 173705, "epoch": 1951, "lr": 2.881705511134819e-05}
{"train_loss": 0.0983261838555336, "global_step": 173706, "epoch": 1951, "lr": 2.881652997471691e-05}
{"train_loss": 0.1444239616394043, "global_step": 173707, "epoch": 1951, "lr": 2.8816004840933518e-05}
{"train_loss": 0.05863174423575401, "global_step": 173708, "epoch": 1951, "lr": 2.8815479709998027e-05}
{"train_loss": 0.11612656712532043, "global_step": 173709, "epoch": 1951, "lr": 2.881495458191056e-05}
{"train_loss": 0.07491427659988403, "global_step": 173710, "epoch": 1951, "lr": 2.8814429456671145e-05}
{"train_loss": 0.04707220196723938, "global_step": 173711, "epoch": 1951, "lr": 2.8813904334279895e-05}
{"train_loss": 0.13631117343902588, "global_step": 173712, "epoch": 1951, "lr": 2.881337921473683e-05}
{"train_loss": 0.10990145802497864, "global_step": 173713, "epoch": 1951, "lr": 2.8812854098042052e-05}
{"train_loss": 0.11654879152774811, "global_step": 173714, "epoch": 1951, "lr": 2.8812328984195647e-05}
{"train_loss": 0.10103966295719147, "global_step": 173715, "epoch": 1951, "lr": 2.8811803873197645e-05}
{"train_loss": 0.07206054031848907, "global_step": 173716, "epoch": 1951, "lr": 2.8811278765048157e-05}
{"train_loss": 0.10754197090864182, "global_step": 173717, "epoch": 1951, "lr": 2.8810753659747214e-05}
{"train_loss": 0.09779070317745209, "global_step": 173718, "epoch": 1951, "lr": 2.8810228557294934e-05}
{"train_loss": 0.11378943920135498, "global_step": 173719, "epoch": 1951, "lr": 2.8809703457691334e-05}
{"train_loss": 0.049834709614515305, "global_step": 173720, "epoch": 1951, "lr": 2.8809178360936533e-05}
{"train_loss": 0.09968019276857376, "global_step": 173721, "epoch": 1951, "lr": 2.880865326703056e-05}
{"train_loss": 0.08482035994529724, "global_step": 173722, "epoch": 1951, "lr": 2.8808128175973527e-05}
{"train_loss": 0.10381433367729187, "global_step": 173723, "epoch": 1951, "lr": 2.8807603087765454e-05}
{"train_loss": 0.06671220064163208, "global_step": 173724, "epoch": 1951, "lr": 2.8807078002406472e-05}
{"train_loss": 0.11848605424165726, "global_step": 173725, "epoch": 1951, "lr": 2.8806552919896614e-05}
{"train_loss": 0.09640124440193176, "global_step": 173726, "epoch": 1951, "lr": 2.8806027840235937e-05}
{"train_loss": 0.08669364757919579, "global_step": 173727, "epoch": 1951, "lr": 2.8805502763424554e-05, "val_loss": 7.275534152984619}
{"train_loss": 0.09927275776863098, "global_step": 173728, "epoch": 1952, "lr": 2.8804977689462488e-05}
{"train_loss": 0.0766587108373642, "global_step": 173729, "epoch": 1952, "lr": 2.8804452618349854e-05}
{"train_loss": 0.20205415785312653, "global_step": 173730, "epoch": 1952, "lr": 2.880392755008669e-05}
{"train_loss": 0.0896071270108223, "global_step": 173731, "epoch": 1952, "lr": 2.880340248467309e-05}
{"train_loss": 0.06496664881706238, "global_step": 173732, "epoch": 1952, "lr": 2.8802877422109103e-05}
{"train_loss": 0.06302428245544434, "global_step": 173733, "epoch": 1952, "lr": 2.880235236239483e-05}
{"train_loss": 0.11841252446174622, "global_step": 173734, "epoch": 1952, "lr": 2.88018273055303e-05}
{"train_loss": 0.09573892503976822, "global_step": 173735, "epoch": 1952, "lr": 2.8801302251515626e-05}
{"train_loss": 0.10187067091464996, "global_step": 173736, "epoch": 1952, "lr": 2.880077720035084e-05}
{"train_loss": 0.08961028605699539, "global_step": 173737, "epoch": 1952, "lr": 2.8800252152036046e-05}
{"train_loss": 0.054753631353378296, "global_step": 173738, "epoch": 1952, "lr": 2.8799727106571285e-05}
{"train_loss": 0.07627488672733307, "global_step": 173739, "epoch": 1952, "lr": 2.879920206395666e-05}
{"train_loss": 0.13973815739154816, "global_step": 173740, "epoch": 1952, "lr": 2.8798677024192206e-05}
{"train_loss": 0.06137239187955856, "global_step": 173741, "epoch": 1952, "lr": 2.8798151987278026e-05}
{"train_loss": 0.0704076886177063, "global_step": 173742, "epoch": 1952, "lr": 2.8797626953214163e-05}
{"train_loss": 0.04864996299147606, "global_step": 173743, "epoch": 1952, "lr": 2.8797101922000715e-05}
{"train_loss": 0.11563041061162949, "global_step": 173744, "epoch": 1952, "lr": 2.8796576893637722e-05}
{"train_loss": 0.06547112762928009, "global_step": 173745, "epoch": 1952, "lr": 2.879605186812529e-05}
{"train_loss": 0.0883711576461792, "global_step": 173746, "epoch": 1952, "lr": 2.879552684546345e-05}
{"train_loss": 0.08824864774942398, "global_step": 173747, "epoch": 1952, "lr": 2.879500182565231e-05}
{"train_loss": 0.09392818808555603, "global_step": 173748, "epoch": 1952, "lr": 2.8794476808691907e-05}
{"train_loss": 0.05457479879260063, "global_step": 173749, "epoch": 1952, "lr": 2.879395179458233e-05}
{"train_loss": 0.05668271705508232, "global_step": 173750, "epoch": 1952, "lr": 2.879342678332366e-05}
{"train_loss": 0.07178003340959549, "global_step": 173751, "epoch": 1952, "lr": 2.879290177491595e-05}
{"train_loss": 0.08480977267026901, "global_step": 173752, "epoch": 1952, "lr": 2.8792376769359287e-05}
{"train_loss": 0.08570828288793564, "global_step": 173753, "epoch": 1952, "lr": 2.879185176665371e-05}
{"train_loss": 0.10066394507884979, "global_step": 173754, "epoch": 1952, "lr": 2.8791326766799325e-05}
{"train_loss": 0.11367572844028473, "global_step": 173755, "epoch": 1952, "lr": 2.879080176979617e-05}
{"train_loss": 0.06279829889535904, "global_step": 173756, "epoch": 1952, "lr": 2.879027677564436e-05}
{"train_loss": 0.12327300757169724, "global_step": 173757, "epoch": 1952, "lr": 2.8789751784343916e-05}
{"train_loss": 0.10775738954544067, "global_step": 173758, "epoch": 1952, "lr": 2.8789226795894952e-05}
{"train_loss": 0.05198630690574646, "global_step": 173759, "epoch": 1952, "lr": 2.878870181029749e-05}
{"train_loss": 0.09276584535837173, "global_step": 173760, "epoch": 1952, "lr": 2.878817682755166e-05}
{"train_loss": 0.09153197705745697, "global_step": 173761, "epoch": 1952, "lr": 2.878765184765747e-05}
{"train_loss": 0.14342477917671204, "global_step": 173762, "epoch": 1952, "lr": 2.878712687061505e-05}
{"train_loss": 0.10756028443574905, "global_step": 173763, "epoch": 1952, "lr": 2.8786601896424437e-05}
{"train_loss": 0.07447165250778198, "global_step": 173764, "epoch": 1952, "lr": 2.878607692508569e-05}
{"train_loss": 0.09829708188772202, "global_step": 173765, "epoch": 1952, "lr": 2.8785551956598916e-05}
{"train_loss": 0.07698066532611847, "global_step": 173766, "epoch": 1952, "lr": 2.8785026990964148e-05}
{"train_loss": 0.13846302032470703, "global_step": 173767, "epoch": 1952, "lr": 2.878450202818149e-05}
{"train_loss": 0.15673798322677612, "global_step": 173768, "epoch": 1952, "lr": 2.878397706825098e-05}
{"train_loss": 0.10654488950967789, "global_step": 173769, "epoch": 1952, "lr": 2.8783452111172727e-05}
{"train_loss": 0.058763839304447174, "global_step": 173770, "epoch": 1952, "lr": 2.878292715694676e-05}
{"train_loss": 0.05143317952752113, "global_step": 173771, "epoch": 1952, "lr": 2.8782402205573188e-05}
{"train_loss": 0.12378650903701782, "global_step": 173772, "epoch": 1952, "lr": 2.8781877257052036e-05}
{"train_loss": 0.04569478705525398, "global_step": 173773, "epoch": 1952, "lr": 2.8781352311383437e-05}
{"train_loss": 0.059605129063129425, "global_step": 173774, "epoch": 1952, "lr": 2.8780827368567398e-05}
{"train_loss": 0.08574292808771133, "global_step": 173775, "epoch": 1952, "lr": 2.8780302428604033e-05}
{"train_loss": 0.09866596758365631, "global_step": 173776, "epoch": 1952, "lr": 2.8779777491493387e-05}
{"train_loss": 0.09313306212425232, "global_step": 173777, "epoch": 1952, "lr": 2.877925255723556e-05}
{"train_loss": 0.1268586814403534, "global_step": 173778, "epoch": 1952, "lr": 2.877872762583058e-05}
{"train_loss": 0.07359257340431213, "global_step": 173779, "epoch": 1952, "lr": 2.8778202697278566e-05}
{"train_loss": 0.04616896063089371, "global_step": 173780, "epoch": 1952, "lr": 2.877767777157954e-05}
{"train_loss": 0.05582228675484657, "global_step": 173781, "epoch": 1952, "lr": 2.8777152848733618e-05}
{"train_loss": 0.07591508328914642, "global_step": 173782, "epoch": 1952, "lr": 2.877662792874083e-05}
{"train_loss": 0.09297221153974533, "global_step": 173783, "epoch": 1952, "lr": 2.8776103011601284e-05}
{"train_loss": 0.07555384933948517, "global_step": 173784, "epoch": 1952, "lr": 2.8775578097315015e-05}
{"train_loss": 0.07277034968137741, "global_step": 173785, "epoch": 1952, "lr": 2.8775053185882105e-05}
{"train_loss": 0.10987162590026855, "global_step": 173786, "epoch": 1952, "lr": 2.877452827730266e-05}
{"train_loss": 0.06109406426548958, "global_step": 173787, "epoch": 1952, "lr": 2.8774003371576695e-05}
{"train_loss": 0.05978322774171829, "global_step": 173788, "epoch": 1952, "lr": 2.8773478468704324e-05}
{"train_loss": 0.059587135910987854, "global_step": 173789, "epoch": 1952, "lr": 2.8772953568685583e-05}
{"train_loss": 0.11306419968605042, "global_step": 173790, "epoch": 1952, "lr": 2.877242867152058e-05}
{"train_loss": 0.05871778354048729, "global_step": 173791, "epoch": 1952, "lr": 2.877190377720934e-05}
{"train_loss": 0.08285871148109436, "global_step": 173792, "epoch": 1952, "lr": 2.8771378885751986e-05}
{"train_loss": 0.08434562385082245, "global_step": 173793, "epoch": 1952, "lr": 2.8770853997148538e-05}
{"train_loss": 0.1361270248889923, "global_step": 173794, "epoch": 1952, "lr": 2.8770329111399114e-05}
{"train_loss": 0.11147745698690414, "global_step": 173795, "epoch": 1952, "lr": 2.8769804228503737e-05}
{"train_loss": 0.07735797017812729, "global_step": 173796, "epoch": 1952, "lr": 2.876927934846252e-05}
{"train_loss": 0.07286308705806732, "global_step": 173797, "epoch": 1952, "lr": 2.87687544712755e-05}
{"train_loss": 0.07145361602306366, "global_step": 173798, "epoch": 1952, "lr": 2.8768229596942776e-05}
{"train_loss": 0.08372682332992554, "global_step": 173799, "epoch": 1952, "lr": 2.876770472546439e-05}
{"train_loss": 0.03753085434436798, "global_step": 173800, "epoch": 1952, "lr": 2.8767179856840444e-05}
{"train_loss": 0.07466832548379898, "global_step": 173801, "epoch": 1952, "lr": 2.876665499107099e-05}
{"train_loss": 0.09196635335683823, "global_step": 173802, "epoch": 1952, "lr": 2.876613012815609e-05}
{"train_loss": 0.07747539132833481, "global_step": 173803, "epoch": 1952, "lr": 2.876560526809583e-05}
{"train_loss": 0.14187969267368317, "global_step": 173804, "epoch": 1952, "lr": 2.8765080410890267e-05}
{"train_loss": 0.05006725713610649, "global_step": 173805, "epoch": 1952, "lr": 2.8764555556539497e-05}
{"train_loss": 0.05874829366803169, "global_step": 173806, "epoch": 1952, "lr": 2.8764030705043554e-05}
{"train_loss": 0.04615464061498642, "global_step": 173807, "epoch": 1952, "lr": 2.8763505856402545e-05}
{"train_loss": 0.07962216436862946, "global_step": 173808, "epoch": 1952, "lr": 2.8762981010616508e-05}
{"train_loss": 0.07924073189496994, "global_step": 173809, "epoch": 1952, "lr": 2.8762456167685548e-05}
{"train_loss": 0.07609830051660538, "global_step": 173810, "epoch": 1952, "lr": 2.8761931327609692e-05}
{"train_loss": 0.09580420702695847, "global_step": 173811, "epoch": 1952, "lr": 2.8761406490389066e-05}
{"train_loss": 0.07057823985815048, "global_step": 173812, "epoch": 1952, "lr": 2.8760881656023676e-05}
{"train_loss": 0.09331446886062622, "global_step": 173813, "epoch": 1952, "lr": 2.8760356824513658e-05}
{"train_loss": 0.08967502415180206, "global_step": 173814, "epoch": 1952, "lr": 2.8759831995859028e-05}
{"train_loss": 0.08278589695692062, "global_step": 173815, "epoch": 1952, "lr": 2.8759307170059903e-05}
{"train_loss": 0.08619065105580212, "global_step": 173816, "epoch": 1952, "lr": 2.8758782347116305e-05, "val_loss": 7.24536657333374}
{"train_loss": 0.08501234650611877, "global_step": 173817, "epoch": 1953, "lr": 2.875825752702835e-05}
{"train_loss": 0.08443578332662582, "global_step": 173818, "epoch": 1953, "lr": 2.875773270979607e-05}
{"train_loss": 0.07318437844514847, "global_step": 173819, "epoch": 1953, "lr": 2.8757207895419568e-05}
{"train_loss": 0.10667236149311066, "global_step": 173820, "epoch": 1953, "lr": 2.8756683083898893e-05}
{"train_loss": 0.09018111974000931, "global_step": 173821, "epoch": 1953, "lr": 2.8756158275234113e-05}
{"train_loss": 0.09009237587451935, "global_step": 173822, "epoch": 1953, "lr": 2.8755633469425337e-05}
{"train_loss": 0.06605169922113419, "global_step": 173823, "epoch": 1953, "lr": 2.8755108666472574e-05}
{"train_loss": 0.06995640695095062, "global_step": 173824, "epoch": 1953, "lr": 2.8754583866375962e-05}
{"train_loss": 0.07835875451564789, "global_step": 173825, "epoch": 1953, "lr": 2.8754059069135508e-05}
{"train_loss": 0.05325813964009285, "global_step": 173826, "epoch": 1953, "lr": 2.8753534274751337e-05}
{"train_loss": 0.10401175171136856, "global_step": 173827, "epoch": 1953, "lr": 2.875300948322347e-05}
{"train_loss": 0.08450126647949219, "global_step": 173828, "epoch": 1953, "lr": 2.875248469455203e-05}
{"train_loss": 0.044673480093479156, "global_step": 173829, "epoch": 1953, "lr": 2.8751959908737024e-05}
{"train_loss": 0.15316511690616608, "global_step": 173830, "epoch": 1953, "lr": 2.8751435125778592e-05}
{"train_loss": 0.10633458942174911, "global_step": 173831, "epoch": 1953, "lr": 2.8750910345676745e-05}
{"train_loss": 0.12965905666351318, "global_step": 173832, "epoch": 1953, "lr": 2.8750385568431604e-05}
{"train_loss": 0.05931168794631958, "global_step": 173833, "epoch": 1953, "lr": 2.8749860794043187e-05}
{"train_loss": 0.08229326456785202, "global_step": 173834, "epoch": 1953, "lr": 2.874933602251162e-05}
{"train_loss": 0.09288671612739563, "global_step": 173835, "epoch": 1953, "lr": 2.8748811253836914e-05}
{"train_loss": 0.08091497421264648, "global_step": 173836, "epoch": 1953, "lr": 2.8748286488019204e-05}
{"train_loss": 0.09799255430698395, "global_step": 173837, "epoch": 1953, "lr": 2.87477617250585e-05}
{"train_loss": 0.09675019234418869, "global_step": 173838, "epoch": 1953, "lr": 2.8747236964954926e-05}
{"train_loss": 0.09989175945520401, "global_step": 173839, "epoch": 1953, "lr": 2.8746712207708504e-05}
{"train_loss": 0.08220642060041428, "global_step": 173840, "epoch": 1953, "lr": 2.8746187453319352e-05}
{"train_loss": 0.1120690256357193, "global_step": 173841, "epoch": 1953, "lr": 2.8745662701787513e-05}
{"train_loss": 0.08956199139356613, "global_step": 173842, "epoch": 1953, "lr": 2.8745137953113034e-05}
{"train_loss": 0.07626889646053314, "global_step": 173843, "epoch": 1953, "lr": 2.8744613207296038e-05}
{"train_loss": 0.06245783716440201, "global_step": 173844, "epoch": 1953, "lr": 2.874408846433655e-05}
{"train_loss": 0.06322330236434937, "global_step": 173845, "epoch": 1953, "lr": 2.874356372423468e-05}
{"train_loss": 0.07917509227991104, "global_step": 173846, "epoch": 1953, "lr": 2.8743038986990457e-05}
{"train_loss": 0.048858705908060074, "global_step": 173847, "epoch": 1953, "lr": 2.8742514252603998e-05}
{"train_loss": 0.04799157753586769, "global_step": 173848, "epoch": 1953, "lr": 2.8741989521075323e-05}
{"train_loss": 0.1161973774433136, "global_step": 173849, "epoch": 1953, "lr": 2.8741464792404547e-05}
{"train_loss": 0.06999564170837402, "global_step": 173850, "epoch": 1953, "lr": 2.8740940066591715e-05}
{"train_loss": 0.1125762015581131, "global_step": 173851, "epoch": 1953, "lr": 2.874041534363691e-05}
{"train_loss": 0.12842124700546265, "global_step": 173852, "epoch": 1953, "lr": 2.8739890623540182e-05}
{"train_loss": 0.10252119600772858, "global_step": 173853, "epoch": 1953, "lr": 2.873936590630164e-05}
{"train_loss": 0.06517747044563293, "global_step": 173854, "epoch": 1953, "lr": 2.8738841191921307e-05}
{"train_loss": 0.07989725470542908, "global_step": 173855, "epoch": 1953, "lr": 2.8738316480399296e-05}
{"train_loss": 0.06017323210835457, "global_step": 173856, "epoch": 1953, "lr": 2.8737791771735645e-05}
{"train_loss": 0.06753307580947876, "global_step": 173857, "epoch": 1953, "lr": 2.873726706593044e-05}
{"train_loss": 0.06261715292930603, "global_step": 173858, "epoch": 1953, "lr": 2.873674236298376e-05}
{"train_loss": 0.07371978461742401, "global_step": 173859, "epoch": 1953, "lr": 2.8736217662895653e-05}
{"train_loss": 0.053426921367645264, "global_step": 173860, "epoch": 1953, "lr": 2.8735692965666222e-05}
{"train_loss": 0.06820023059844971, "global_step": 173861, "epoch": 1953, "lr": 2.87351682712955e-05}
{"train_loss": 0.062221840023994446, "global_step": 173862, "epoch": 1953, "lr": 2.8734643579783592e-05}
{"train_loss": 0.08015996217727661, "global_step": 173863, "epoch": 1953, "lr": 2.8734118891130535e-05}
{"train_loss": 0.1250382959842682, "global_step": 173864, "epoch": 1953, "lr": 2.8733594205336435e-05}
{"train_loss": 0.05970381945371628, "global_step": 173865, "epoch": 1953, "lr": 2.873306952240132e-05}
{"train_loss": 0.1040586456656456, "global_step": 173866, "epoch": 1953, "lr": 2.8732544842325304e-05}
{"train_loss": 0.042183008044958115, "global_step": 173867, "epoch": 1953, "lr": 2.873202016510842e-05}
{"train_loss": 0.13262403011322021, "global_step": 173868, "epoch": 1953, "lr": 2.873149549075078e-05}
{"train_loss": 0.05735527724027634, "global_step": 173869, "epoch": 1953, "lr": 2.8730970819252412e-05}
{"train_loss": 0.07835759967565536, "global_step": 173870, "epoch": 1953, "lr": 2.8730446150613423e-05}
{"train_loss": 0.06881082057952881, "global_step": 173871, "epoch": 1953, "lr": 2.8729921484833844e-05}
{"train_loss": 0.05917687341570854, "global_step": 173872, "epoch": 1953, "lr": 2.8729396821913794e-05}
{"train_loss": 0.07546623796224594, "global_step": 173873, "epoch": 1953, "lr": 2.8728872161853294e-05}
{"train_loss": 0.04720890522003174, "global_step": 173874, "epoch": 1953, "lr": 2.8728347504652458e-05}
{"train_loss": 0.06586122512817383, "global_step": 173875, "epoch": 1953, "lr": 2.8727822850311314e-05}
{"train_loss": 0.08641017228364944, "global_step": 173876, "epoch": 1953, "lr": 2.8727298198829976e-05}
{"train_loss": 0.13608935475349426, "global_step": 173877, "epoch": 1953, "lr": 2.8726773550208476e-05}
{"train_loss": 0.11121798306703568, "global_step": 173878, "epoch": 1953, "lr": 2.872624890444692e-05}
{"train_loss": 0.11547181755304337, "global_step": 173879, "epoch": 1953, "lr": 2.872572426154536e-05}
{"train_loss": 0.08993247151374817, "global_step": 173880, "epoch": 1953, "lr": 2.872519962150385e-05}
{"train_loss": 0.04332646355032921, "global_step": 173881, "epoch": 1953, "lr": 2.87246749843225e-05}
{"train_loss": 0.12177590280771255, "global_step": 173882, "epoch": 1953, "lr": 2.8724150350001334e-05}
{"train_loss": 0.08162391930818558, "global_step": 173883, "epoch": 1953, "lr": 2.8723625718540464e-05}
{"train_loss": 0.1520622968673706, "global_step": 173884, "epoch": 1953, "lr": 2.872310108993993e-05}
{"train_loss": 0.05503707006573677, "global_step": 173885, "epoch": 1953, "lr": 2.8722576464199828e-05}
{"train_loss": 0.1155114695429802, "global_step": 173886, "epoch": 1953, "lr": 2.8722051841320198e-05}
{"train_loss": 0.058842405676841736, "global_step": 173887, "epoch": 1953, "lr": 2.8721527221301154e-05}
{"train_loss": 0.05536167323589325, "global_step": 173888, "epoch": 1953, "lr": 2.8721002604142716e-05}
{"train_loss": 0.05507825314998627, "global_step": 173889, "epoch": 1953, "lr": 2.8720477989845007e-05}
{"train_loss": 0.09778683632612228, "global_step": 173890, "epoch": 1953, "lr": 2.8719953378408036e-05}
{"train_loss": 0.09619665145874023, "global_step": 173891, "epoch": 1953, "lr": 2.8719428769831947e-05}
{"train_loss": 0.09988623857498169, "global_step": 173892, "epoch": 1953, "lr": 2.8718904164116735e-05}
{"train_loss": 0.05318094417452812, "global_step": 173893, "epoch": 1953, "lr": 2.8718379561262532e-05}
{"train_loss": 0.05527127534151077, "global_step": 173894, "epoch": 1953, "lr": 2.871785496126937e-05}
{"train_loss": 0.11550202965736389, "global_step": 173895, "epoch": 1953, "lr": 2.871733036413733e-05}
{"train_loss": 0.06533803790807724, "global_step": 173896, "epoch": 1953, "lr": 2.87168057698665e-05}
{"train_loss": 0.04093736410140991, "global_step": 173897, "epoch": 1953, "lr": 2.8716281178456916e-05}
{"train_loss": 0.0841677188873291, "global_step": 173898, "epoch": 1953, "lr": 2.8715756589908693e-05}
{"train_loss": 0.12652763724327087, "global_step": 173899, "epoch": 1953, "lr": 2.871523200422186e-05}
{"train_loss": 0.09147442132234573, "global_step": 173900, "epoch": 1953, "lr": 2.871470742139652e-05}
{"train_loss": 0.039504408836364746, "global_step": 173901, "epoch": 1953, "lr": 2.8714182841432712e-05}
{"train_loss": 0.06921955198049545, "global_step": 173902, "epoch": 1953, "lr": 2.8713658264330534e-05}
{"train_loss": 0.09997407346963882, "global_step": 173903, "epoch": 1953, "lr": 2.8713133690090032e-05}
{"train_loss": 0.16193433105945587, "global_step": 173904, "epoch": 1953, "lr": 2.871260911871131e-05}
{"train_loss": 0.0842565582290794, "global_step": 173905, "epoch": 1953, "lr": 2.8712084550194397e-05, "val_loss": 7.326107978820801}
{"train_loss": 0.053554974496364594, "global_step": 173906, "epoch": 1954, "lr": 2.8711559984539404e-05}
{"train_loss": 0.06697456538677216, "global_step": 173907, "epoch": 1954, "lr": 2.8711035421746367e-05}
{"train_loss": 0.10804138332605362, "global_step": 173908, "epoch": 1954, "lr": 2.8710510861815386e-05}
{"train_loss": 0.08805781602859497, "global_step": 173909, "epoch": 1954, "lr": 2.8709986304746496e-05}
{"train_loss": 0.06341046094894409, "global_step": 173910, "epoch": 1954, "lr": 2.870946175053981e-05}
{"train_loss": 0.03894961625337601, "global_step": 173911, "epoch": 1954, "lr": 2.8708937199195363e-05}
{"train_loss": 0.15244056284427643, "global_step": 173912, "epoch": 1954, "lr": 2.8708412650713255e-05}
{"train_loss": 0.09762011468410492, "global_step": 173913, "epoch": 1954, "lr": 2.8707888105093526e-05}
{"train_loss": 0.06545843929052353, "global_step": 173914, "epoch": 1954, "lr": 2.8707363562336274e-05}
{"train_loss": 0.06386418640613556, "global_step": 173915, "epoch": 1954, "lr": 2.8706839022441546e-05}
{"train_loss": 0.0961378812789917, "global_step": 173916, "epoch": 1954, "lr": 2.8706314485409437e-05}
{"train_loss": 0.09012661874294281, "global_step": 173917, "epoch": 1954, "lr": 2.8705789951240008e-05}
{"train_loss": 0.0844687670469284, "global_step": 173918, "epoch": 1954, "lr": 2.870526541993331e-05}
{"train_loss": 0.11033803969621658, "global_step": 173919, "epoch": 1954, "lr": 2.8704740891489447e-05}
{"train_loss": 0.08796527981758118, "global_step": 173920, "epoch": 1954, "lr": 2.8704216365908444e-05}
{"train_loss": 0.057534150779247284, "global_step": 173921, "epoch": 1954, "lr": 2.870369184319043e-05}
{"train_loss": 0.02383391745388508, "global_step": 173922, "epoch": 1954, "lr": 2.8703167323335417e-05}
{"train_loss": 0.10283765941858292, "global_step": 173923, "epoch": 1954, "lr": 2.8702642806343528e-05}
{"train_loss": 0.1019851416349411, "global_step": 173924, "epoch": 1954, "lr": 2.8702118292214786e-05}
{"train_loss": 0.0856052115559578, "global_step": 173925, "epoch": 1954, "lr": 2.870159378094931e-05}
{"train_loss": 0.09271199256181717, "global_step": 173926, "epoch": 1954, "lr": 2.8701069272547122e-05}
{"train_loss": 0.07244426012039185, "global_step": 173927, "epoch": 1954, "lr": 2.8700544767008335e-05}
{"train_loss": 0.08663473278284073, "global_step": 173928, "epoch": 1954, "lr": 2.8700020264332983e-05}
{"train_loss": 0.03520584851503372, "global_step": 173929, "epoch": 1954, "lr": 2.8699495764521166e-05}
{"train_loss": 0.11594756692647934, "global_step": 173930, "epoch": 1954, "lr": 2.8698971267572926e-05}
{"train_loss": 0.06353435665369034, "global_step": 173931, "epoch": 1954, "lr": 2.8698446773488358e-05}
{"train_loss": 0.07444354891777039, "global_step": 173932, "epoch": 1954, "lr": 2.8697922282267535e-05}
{"train_loss": 0.15181151032447815, "global_step": 173933, "epoch": 1954, "lr": 2.8697397793910496e-05}
{"train_loss": 0.12015658617019653, "global_step": 173934, "epoch": 1954, "lr": 2.869687330841736e-05}
{"train_loss": 0.10959906131029129, "global_step": 173935, "epoch": 1954, "lr": 2.8696348825788144e-05}
{"train_loss": 0.100674569606781, "global_step": 173936, "epoch": 1954, "lr": 2.869582434602297e-05}
{"train_loss": 0.08358418196439743, "global_step": 173937, "epoch": 1954, "lr": 2.8695299869121856e-05}
{"train_loss": 0.04948357120156288, "global_step": 173938, "epoch": 1954, "lr": 2.8694775395084922e-05}
{"train_loss": 0.12030669301748276, "global_step": 173939, "epoch": 1954, "lr": 2.8694250923912192e-05}
{"train_loss": 0.10937488079071045, "global_step": 173940, "epoch": 1954, "lr": 2.8693726455603786e-05}
{"train_loss": 0.080995112657547, "global_step": 173941, "epoch": 1954, "lr": 2.8693201990159724e-05}
{"train_loss": 0.14063681662082672, "global_step": 173942, "epoch": 1954, "lr": 2.8692677527580124e-05}
{"train_loss": 0.08267055451869965, "global_step": 173943, "epoch": 1954, "lr": 2.8692153067865013e-05}
{"train_loss": 0.12830917537212372, "global_step": 173944, "epoch": 1954, "lr": 2.8691628611014502e-05}
{"train_loss": 0.12301674485206604, "global_step": 173945, "epoch": 1954, "lr": 2.8691104157028624e-05}
{"train_loss": 0.062303368002176285, "global_step": 173946, "epoch": 1954, "lr": 2.869057970590749e-05}
{"train_loss": 0.05923740193247795, "global_step": 173947, "epoch": 1954, "lr": 2.869005525765112e-05}
{"train_loss": 0.07753798365592957, "global_step": 173948, "epoch": 1954, "lr": 2.8689530812259618e-05}
{"train_loss": 0.10292340070009232, "global_step": 173949, "epoch": 1954, "lr": 2.8689006369733062e-05}
{"train_loss": 0.06481988728046417, "global_step": 173950, "epoch": 1954, "lr": 2.868848193007152e-05}
{"train_loss": 0.046020012348890305, "global_step": 173951, "epoch": 1954, "lr": 2.8687957493275023e-05}
{"train_loss": 0.09782800823450089, "global_step": 173952, "epoch": 1954, "lr": 2.8687433059343694e-05}
{"train_loss": 0.044527534395456314, "global_step": 173953, "epoch": 1954, "lr": 2.8686908628277564e-05}
{"train_loss": 0.05810213461518288, "global_step": 173954, "epoch": 1954, "lr": 2.868638420007673e-05}
{"train_loss": 0.06586485356092453, "global_step": 173955, "epoch": 1954, "lr": 2.868585977474124e-05}
{"train_loss": 0.07036988437175751, "global_step": 173956, "epoch": 1954, "lr": 2.868533535227119e-05}
{"train_loss": 0.04903900623321533, "global_step": 173957, "epoch": 1954, "lr": 2.868481093266664e-05}
{"train_loss": 0.06280534714460373, "global_step": 173958, "epoch": 1954, "lr": 2.868428651592764e-05}
{"train_loss": 0.1105836033821106, "global_step": 173959, "epoch": 1954, "lr": 2.868376210205429e-05}
{"train_loss": 0.04660218209028244, "global_step": 173960, "epoch": 1954, "lr": 2.8683237691046637e-05}
{"train_loss": 0.11725985258817673, "global_step": 173961, "epoch": 1954, "lr": 2.8682713282904784e-05}
{"train_loss": 0.11785019189119339, "global_step": 173962, "epoch": 1954, "lr": 2.868218887762875e-05}
{"train_loss": 0.060427479445934296, "global_step": 173963, "epoch": 1954, "lr": 2.8681664475218662e-05}
{"train_loss": 0.09438350051641464, "global_step": 173964, "epoch": 1954, "lr": 2.8681140075674545e-05}
{"train_loss": 0.10391348600387573, "global_step": 173965, "epoch": 1954, "lr": 2.8680615678996507e-05}
{"train_loss": 0.04451695829629898, "global_step": 173966, "epoch": 1954, "lr": 2.8680091285184584e-05}
{"train_loss": 0.06497538089752197, "global_step": 173967, "epoch": 1954, "lr": 2.867956689423886e-05}
{"train_loss": 0.0918002724647522, "global_step": 173968, "epoch": 1954, "lr": 2.8679042506159424e-05}
{"train_loss": 0.07690480351448059, "global_step": 173969, "epoch": 1954, "lr": 2.867851812094632e-05}
{"train_loss": 0.09448261559009552, "global_step": 173970, "epoch": 1954, "lr": 2.867799373859964e-05}
{"train_loss": 0.1048702672123909, "global_step": 173971, "epoch": 1954, "lr": 2.867746935911943e-05}
{"train_loss": 0.08585231006145477, "global_step": 173972, "epoch": 1954, "lr": 2.8676944982505794e-05}
{"train_loss": 0.07662025839090347, "global_step": 173973, "epoch": 1954, "lr": 2.8676420608758758e-05}
{"train_loss": 0.05494159087538719, "global_step": 173974, "epoch": 1954, "lr": 2.8675896237878443e-05}
{"train_loss": 0.05627535656094551, "global_step": 173975, "epoch": 1954, "lr": 2.8675371869864877e-05}
{"train_loss": 0.07892832159996033, "global_step": 173976, "epoch": 1954, "lr": 2.8674847504718162e-05}
{"train_loss": 0.09210958331823349, "global_step": 173977, "epoch": 1954, "lr": 2.867432314243833e-05}
{"train_loss": 0.07068046927452087, "global_step": 173978, "epoch": 1954, "lr": 2.86737987830255e-05}
{"train_loss": 0.12737545371055603, "global_step": 173979, "epoch": 1954, "lr": 2.86732744264797e-05}
{"train_loss": 0.08033248037099838, "global_step": 173980, "epoch": 1954, "lr": 2.8672750072801036e-05}
{"train_loss": 0.07100789248943329, "global_step": 173981, "epoch": 1954, "lr": 2.8672225721989537e-05}
{"train_loss": 0.05706142261624336, "global_step": 173982, "epoch": 1954, "lr": 2.8671701374045323e-05}
{"train_loss": 0.07560208439826965, "global_step": 173983, "epoch": 1954, "lr": 2.8671177028968416e-05}
{"train_loss": 0.049431849271059036, "global_step": 173984, "epoch": 1954, "lr": 2.867065268675891e-05}
{"train_loss": 0.04718030244112015, "global_step": 173985, "epoch": 1954, "lr": 2.867012834741689e-05}
{"train_loss": 0.15978063642978668, "global_step": 173986, "epoch": 1954, "lr": 2.8669604010942397e-05}
{"train_loss": 0.06698682904243469, "global_step": 173987, "epoch": 1954, "lr": 2.8669079677335543e-05}
{"train_loss": 0.13421612977981567, "global_step": 173988, "epoch": 1954, "lr": 2.8668555346596336e-05}
{"train_loss": 0.04196888580918312, "global_step": 173989, "epoch": 1954, "lr": 2.8668031018724912e-05}
{"train_loss": 0.04426770657300949, "global_step": 173990, "epoch": 1954, "lr": 2.8667506693721308e-05}
{"train_loss": 0.10165976732969284, "global_step": 173991, "epoch": 1954, "lr": 2.8666982371585577e-05}
{"train_loss": 0.052647609263658524, "global_step": 173992, "epoch": 1954, "lr": 2.866645805231783e-05}
{"train_loss": 0.06935863196849823, "global_step": 173993, "epoch": 1954, "lr": 2.8665933735918098e-05}
{"train_loss": 0.08324329707813397, "global_step": 173994, "epoch": 1954, "lr": 2.866540942238649e-05, "val_loss": 7.383326530456543}
{"train_loss": 0.12298376858234406, "global_step": 173995, "epoch": 1955, "lr": 2.8664885111723063e-05}
{"train_loss": 0.06748464703559875, "global_step": 173996, "epoch": 1955, "lr": 2.8664360803927857e-05}
{"train_loss": 0.04957997426390648, "global_step": 173997, "epoch": 1955, "lr": 2.866383649900099e-05}
{"train_loss": 0.028899874538183212, "global_step": 173998, "epoch": 1955, "lr": 2.866331219694249e-05}
{"train_loss": 0.12147630006074905, "global_step": 173999, "epoch": 1955, "lr": 2.8662787897752464e-05}
{"train_loss": 0.11226102709770203, "global_step": 174000, "epoch": 1955, "lr": 2.866226360143095e-05}
{"train_loss": 0.06994404643774033, "global_step": 174001, "epoch": 1955, "lr": 2.8661739307978053e-05}
{"train_loss": 0.10048363357782364, "global_step": 174002, "epoch": 1955, "lr": 2.8661215017393805e-05}
{"train_loss": 0.05294063687324524, "global_step": 174003, "epoch": 1955, "lr": 2.8660690729678295e-05}
{"train_loss": 0.05368957296013832, "global_step": 174004, "epoch": 1955, "lr": 2.866016644483162e-05}
{"train_loss": 0.07194855809211731, "global_step": 174005, "epoch": 1955, "lr": 2.8659642162853804e-05}
{"train_loss": 0.10514054447412491, "global_step": 174006, "epoch": 1955, "lr": 2.8659117883744956e-05}
{"train_loss": 0.051018234342336655, "global_step": 174007, "epoch": 1955, "lr": 2.8658593607505113e-05}
{"train_loss": 0.056666210293769836, "global_step": 174008, "epoch": 1955, "lr": 2.865806933413438e-05}
{"train_loss": 0.07603296637535095, "global_step": 174009, "epoch": 1955, "lr": 2.865754506363279e-05}
{"train_loss": 0.05635616183280945, "global_step": 174010, "epoch": 1955, "lr": 2.8657020796000454e-05}
{"train_loss": 0.0557987242937088, "global_step": 174011, "epoch": 1955, "lr": 2.86564965312374e-05}
{"train_loss": 0.07630858570337296, "global_step": 174012, "epoch": 1955, "lr": 2.865597226934374e-05}
{"train_loss": 0.10908415168523788, "global_step": 174013, "epoch": 1955, "lr": 2.8655448010319502e-05}
{"train_loss": 0.10998456925153732, "global_step": 174014, "epoch": 1955, "lr": 2.8654923754164802e-05}
{"train_loss": 0.07650033384561539, "global_step": 174015, "epoch": 1955, "lr": 2.865439950087967e-05}
{"train_loss": 0.09085793048143387, "global_step": 174016, "epoch": 1955, "lr": 2.865387525046421e-05}
{"train_loss": 0.09963357448577881, "global_step": 174017, "epoch": 1955, "lr": 2.8653351002918454e-05}
{"train_loss": 0.061229679733514786, "global_step": 174018, "epoch": 1955, "lr": 2.8652826758242518e-05}
{"train_loss": 0.0682489350438118, "global_step": 174019, "epoch": 1955, "lr": 2.8652302516436424e-05}
{"train_loss": 0.08390168845653534, "global_step": 174020, "epoch": 1955, "lr": 2.8651778277500295e-05}
{"train_loss": 0.0935286208987236, "global_step": 174021, "epoch": 1955, "lr": 2.865125404143415e-05}
{"train_loss": 0.06984306871891022, "global_step": 174022, "epoch": 1955, "lr": 2.8650729808238085e-05}
{"train_loss": 0.03284028172492981, "global_step": 174023, "epoch": 1955, "lr": 2.8650205577912192e-05}
{"train_loss": 0.07822100073099136, "global_step": 174024, "epoch": 1955, "lr": 2.8649681350456496e-05}
{"train_loss": 0.12118727713823318, "global_step": 174025, "epoch": 1955, "lr": 2.864915712587111e-05}
{"train_loss": 0.09256768226623535, "global_step": 174026, "epoch": 1955, "lr": 2.864863290415607e-05}
{"train_loss": 0.07323884218931198, "global_step": 174027, "epoch": 1955, "lr": 2.864810868531147e-05}
{"train_loss": 0.11622952669858932, "global_step": 174028, "epoch": 1955, "lr": 2.864758446933738e-05}
{"train_loss": 0.10071155428886414, "global_step": 174029, "epoch": 1955, "lr": 2.864706025623384e-05}
{"train_loss": 0.11079481244087219, "global_step": 174030, "epoch": 1955, "lr": 2.864653604600096e-05}
{"train_loss": 0.09582363814115524, "global_step": 174031, "epoch": 1955, "lr": 2.8646011838638776e-05}
{"train_loss": 0.07551990449428558, "global_step": 174032, "epoch": 1955, "lr": 2.8645487634147394e-05}
{"train_loss": 0.11536893248558044, "global_step": 174033, "epoch": 1955, "lr": 2.8644963432526868e-05}
{"train_loss": 0.08245298266410828, "global_step": 174034, "epoch": 1955, "lr": 2.8644439233777244e-05}
{"train_loss": 0.06958315521478653, "global_step": 174035, "epoch": 1955, "lr": 2.864391503789864e-05}
{"train_loss": 0.055244386196136475, "global_step": 174036, "epoch": 1955, "lr": 2.8643390844891072e-05}
{"train_loss": 0.03608071058988571, "global_step": 174037, "epoch": 1955, "lr": 2.8642866654754662e-05}
{"train_loss": 0.09581920504570007, "global_step": 174038, "epoch": 1955, "lr": 2.8642342467489447e-05}
{"train_loss": 0.07630223035812378, "global_step": 174039, "epoch": 1955, "lr": 2.86418182830955e-05}
{"train_loss": 0.09222659468650818, "global_step": 174040, "epoch": 1955, "lr": 2.8641294101572925e-05}
{"train_loss": 0.1231125220656395, "global_step": 174041, "epoch": 1955, "lr": 2.864076992292174e-05}
{"train_loss": 0.06637591123580933, "global_step": 174042, "epoch": 1955, "lr": 2.864024574714207e-05}
{"train_loss": 0.1076548770070076, "global_step": 174043, "epoch": 1955, "lr": 2.8639721574233935e-05}
{"train_loss": 0.14243841171264648, "global_step": 174044, "epoch": 1955, "lr": 2.8639197404197458e-05}
{"train_loss": 0.12995579838752747, "global_step": 174045, "epoch": 1955, "lr": 2.8638673237032653e-05}
{"train_loss": 0.10141465812921524, "global_step": 174046, "epoch": 1955, "lr": 2.863814907273964e-05}
{"train_loss": 0.13653403520584106, "global_step": 174047, "epoch": 1955, "lr": 2.8637624911318446e-05}
{"train_loss": 0.08210207521915436, "global_step": 174048, "epoch": 1955, "lr": 2.8637100752769186e-05}
{"train_loss": 0.1058509424328804, "global_step": 174049, "epoch": 1955, "lr": 2.863657659709188e-05}
{"train_loss": 0.08275500684976578, "global_step": 174050, "epoch": 1955, "lr": 2.8636052444286654e-05}
{"train_loss": 0.08458767086267471, "global_step": 174051, "epoch": 1955, "lr": 2.8635528294353524e-05}
{"train_loss": 0.07026854902505875, "global_step": 174052, "epoch": 1955, "lr": 2.863500414729261e-05}
{"train_loss": 0.08615996688604355, "global_step": 174053, "epoch": 1955, "lr": 2.8634480003103936e-05}
{"train_loss": 0.08896804600954056, "global_step": 174054, "epoch": 1955, "lr": 2.8633955861787622e-05}
{"train_loss": 0.0934225544333458, "global_step": 174055, "epoch": 1955, "lr": 2.863343172334369e-05}
{"train_loss": 0.05484141409397125, "global_step": 174056, "epoch": 1955, "lr": 2.863290758777225e-05}
{"train_loss": 0.07441997528076172, "global_step": 174057, "epoch": 1955, "lr": 2.863238345507334e-05}
{"train_loss": 0.07713344693183899, "global_step": 174058, "epoch": 1955, "lr": 2.8631859325247046e-05}
{"train_loss": 0.047290511429309845, "global_step": 174059, "epoch": 1955, "lr": 2.8631335198293453e-05}
{"train_loss": 0.0688837394118309, "global_step": 174060, "epoch": 1955, "lr": 2.8630811074212592e-05}
{"train_loss": 0.1041676327586174, "global_step": 174061, "epoch": 1955, "lr": 2.8630286953004593e-05}
{"train_loss": 0.09431745857000351, "global_step": 174062, "epoch": 1955, "lr": 2.862976283466946e-05}
{"train_loss": 0.12368649244308472, "global_step": 174063, "epoch": 1955, "lr": 2.8629238719207318e-05}
{"train_loss": 0.07094739377498627, "global_step": 174064, "epoch": 1955, "lr": 2.862871460661819e-05}
{"train_loss": 0.1039976254105568, "global_step": 174065, "epoch": 1955, "lr": 2.8628190496902196e-05}
{"train_loss": 0.07789213955402374, "global_step": 174066, "epoch": 1955, "lr": 2.862766639005936e-05}
{"train_loss": 0.0864470899105072, "global_step": 174067, "epoch": 1955, "lr": 2.8627142286089793e-05}
{"train_loss": 0.0764404758810997, "global_step": 174068, "epoch": 1955, "lr": 2.8626618184993548e-05}
{"train_loss": 0.08450218290090561, "global_step": 174069, "epoch": 1955, "lr": 2.8626094086770672e-05}
{"train_loss": 0.10920503735542297, "global_step": 174070, "epoch": 1955, "lr": 2.862556999142128e-05}
{"train_loss": 0.04638444259762764, "global_step": 174071, "epoch": 1955, "lr": 2.8625045898945402e-05}
{"train_loss": 0.07472187280654907, "global_step": 174072, "epoch": 1955, "lr": 2.862452180934314e-05}
{"train_loss": 0.050379350781440735, "global_step": 174073, "epoch": 1955, "lr": 2.862399772261455e-05}
{"train_loss": 0.10716309398412704, "global_step": 174074, "epoch": 1955, "lr": 2.8623473638759686e-05}
{"train_loss": 0.11266747862100601, "global_step": 174075, "epoch": 1955, "lr": 2.8622949557778633e-05}
{"train_loss": 0.04970424622297287, "global_step": 174076, "epoch": 1955, "lr": 2.862242547967149e-05}
{"train_loss": 0.04580192267894745, "global_step": 174077, "epoch": 1955, "lr": 2.8621901404438273e-05}
{"train_loss": 0.05325041711330414, "global_step": 174078, "epoch": 1955, "lr": 2.86213773320791e-05}
{"train_loss": 0.05058097094297409, "global_step": 174079, "epoch": 1955, "lr": 2.862085326259401e-05}
{"train_loss": 0.060858845710754395, "global_step": 174080, "epoch": 1955, "lr": 2.8620329195983102e-05}
{"train_loss": 0.06791158020496368, "global_step": 174081, "epoch": 1955, "lr": 2.8619805132246414e-05}
{"train_loss": 0.10238275676965714, "global_step": 174082, "epoch": 1955, "lr": 2.8619281071384047e-05}
{"train_loss": 0.08274632198422142, "global_step": 174083, "epoch": 1955, "lr": 2.8618757013396036e-05, "val_loss": 7.354440212249756, "train_action_mse_error": 8.436762809753418}
{"train_loss": 0.06513949483633041, "global_step": 174084, "epoch": 1956, "lr": 2.86182329582825e-05}
{"train_loss": 0.09599312394857407, "global_step": 174085, "epoch": 1956, "lr": 2.861770890604346e-05}
{"train_loss": 0.03957240656018257, "global_step": 174086, "epoch": 1956, "lr": 2.8617184856679023e-05}
{"train_loss": 0.10889022797346115, "global_step": 174087, "epoch": 1956, "lr": 2.8616660810189234e-05}
{"train_loss": 0.11845215409994125, "global_step": 174088, "epoch": 1956, "lr": 2.8616136766574186e-05}
{"train_loss": 0.09791351109743118, "global_step": 174089, "epoch": 1956, "lr": 2.8615612725833917e-05}
{"train_loss": 0.05223405733704567, "global_step": 174090, "epoch": 1956, "lr": 2.861508868796854e-05}
{"train_loss": 0.08267354965209961, "global_step": 174091, "epoch": 1956, "lr": 2.861456465297809e-05}
{"train_loss": 0.08784420043230057, "global_step": 174092, "epoch": 1956, "lr": 2.8614040620862665e-05}
{"train_loss": 0.12637707591056824, "global_step": 174093, "epoch": 1956, "lr": 2.8613516591622302e-05}
{"train_loss": 0.09731539338827133, "global_step": 174094, "epoch": 1956, "lr": 2.861299256525709e-05}
{"train_loss": 0.050822578370571136, "global_step": 174095, "epoch": 1956, "lr": 2.8612468541767125e-05}
{"train_loss": 0.1486807018518448, "global_step": 174096, "epoch": 1956, "lr": 2.861194452115243e-05}
{"train_loss": 0.06983979791402817, "global_step": 174097, "epoch": 1956, "lr": 2.861142050341312e-05}
{"train_loss": 0.0717954933643341, "global_step": 174098, "epoch": 1956, "lr": 2.861089648854922e-05}
{"train_loss": 0.08915530890226364, "global_step": 174099, "epoch": 1956, "lr": 2.8610372476560848e-05}
{"train_loss": 0.058482229709625244, "global_step": 174100, "epoch": 1956, "lr": 2.860984846744803e-05}
{"train_loss": 0.03115125559270382, "global_step": 174101, "epoch": 1956, "lr": 2.8609324461210873e-05}
{"train_loss": 0.06714271008968353, "global_step": 174102, "epoch": 1956, "lr": 2.8608800457849416e-05}
{"train_loss": 0.10359057784080505, "global_step": 174103, "epoch": 1956, "lr": 2.860827645736377e-05}
{"train_loss": 0.04987911134958267, "global_step": 174104, "epoch": 1956, "lr": 2.8607752459753952e-05}
{"train_loss": 0.09990998357534409, "global_step": 174105, "epoch": 1956, "lr": 2.8607228465020085e-05}
{"train_loss": 0.09493035078048706, "global_step": 174106, "epoch": 1956, "lr": 2.8606704473162216e-05}
{"train_loss": 0.08099107444286346, "global_step": 174107, "epoch": 1956, "lr": 2.860618048418039e-05}
{"train_loss": 0.08854368329048157, "global_step": 174108, "epoch": 1956, "lr": 2.8605656498074722e-05}
{"train_loss": 0.06948729604482651, "global_step": 174109, "epoch": 1956, "lr": 2.8605132514845246e-05}
{"train_loss": 0.0371495820581913, "global_step": 174110, "epoch": 1956, "lr": 2.8604608534492073e-05}
{"train_loss": 0.12631972134113312, "global_step": 174111, "epoch": 1956, "lr": 2.860408455701522e-05}
{"train_loss": 0.06418123841285706, "global_step": 174112, "epoch": 1956, "lr": 2.860356058241482e-05}
{"train_loss": 0.07183053344488144, "global_step": 174113, "epoch": 1956, "lr": 2.8603036610690882e-05}
{"train_loss": 0.07094638049602509, "global_step": 174114, "epoch": 1956, "lr": 2.8602512641843528e-05}
{"train_loss": 0.06790736317634583, "global_step": 174115, "epoch": 1956, "lr": 2.860198867587278e-05}
{"train_loss": 0.06872377544641495, "global_step": 174116, "epoch": 1956, "lr": 2.860146471277876e-05}
{"train_loss": 0.06806318461894989, "global_step": 174117, "epoch": 1956, "lr": 2.8600940752561488e-05}
{"train_loss": 0.03441315144300461, "global_step": 174118, "epoch": 1956, "lr": 2.860041679522108e-05}
{"train_loss": 0.06386565417051315, "global_step": 174119, "epoch": 1956, "lr": 2.859989284075757e-05}
{"train_loss": 0.0954941064119339, "global_step": 174120, "epoch": 1956, "lr": 2.8599368889171053e-05}
{"train_loss": 0.16273514926433563, "global_step": 174121, "epoch": 1956, "lr": 2.8598844940461578e-05}
{"train_loss": 0.08638159930706024, "global_step": 174122, "epoch": 1956, "lr": 2.8598320994629246e-05}
{"train_loss": 0.12216196954250336, "global_step": 174123, "epoch": 1956, "lr": 2.859779705167409e-05}
{"train_loss": 0.04547776281833649, "global_step": 174124, "epoch": 1956, "lr": 2.8597273111596212e-05}
{"train_loss": 0.08056192845106125, "global_step": 174125, "epoch": 1956, "lr": 2.859674917439566e-05}
{"train_loss": 0.04405764117836952, "global_step": 174126, "epoch": 1956, "lr": 2.8596225240072526e-05}
{"train_loss": 0.08496672660112381, "global_step": 174127, "epoch": 1956, "lr": 2.8595701308626853e-05}
{"train_loss": 0.056083012372255325, "global_step": 174128, "epoch": 1956, "lr": 2.8595177380058748e-05}
{"train_loss": 0.056771181523799896, "global_step": 174129, "epoch": 1956, "lr": 2.8594653454368236e-05}
{"train_loss": 0.09351865947246552, "global_step": 174130, "epoch": 1956, "lr": 2.8594129531555418e-05}
{"train_loss": 0.09942816197872162, "global_step": 174131, "epoch": 1956, "lr": 2.8593605611620377e-05}
{"train_loss": 0.07124406099319458, "global_step": 174132, "epoch": 1956, "lr": 2.859308169456314e-05}
{"train_loss": 0.10912518948316574, "global_step": 174133, "epoch": 1956, "lr": 2.8592557780383828e-05}
{"train_loss": 0.08252773433923721, "global_step": 174134, "epoch": 1956, "lr": 2.8592033869082462e-05}
{"train_loss": 0.025672608986496925, "global_step": 174135, "epoch": 1956, "lr": 2.8591509960659158e-05}
{"train_loss": 0.04003604128956795, "global_step": 174136, "epoch": 1956, "lr": 2.8590986055113943e-05}
{"train_loss": 0.08598847687244415, "global_step": 174137, "epoch": 1956, "lr": 2.859046215244693e-05}
{"train_loss": 0.06001600623130798, "global_step": 174138, "epoch": 1956, "lr": 2.858993825265815e-05}
{"train_loss": 0.1998596042394638, "global_step": 174139, "epoch": 1956, "lr": 2.858941435574771e-05}
{"train_loss": 0.09359966218471527, "global_step": 174140, "epoch": 1956, "lr": 2.8588890461715646e-05}
{"train_loss": 0.06775448471307755, "global_step": 174141, "epoch": 1956, "lr": 2.8588366570562053e-05}
{"train_loss": 0.055863410234451294, "global_step": 174142, "epoch": 1956, "lr": 2.8587842682286986e-05}
{"train_loss": 0.07647522538900375, "global_step": 174143, "epoch": 1956, "lr": 2.8587318796890537e-05}
{"train_loss": 0.1472809612751007, "global_step": 174144, "epoch": 1956, "lr": 2.858679491437276e-05}
{"train_loss": 0.08149178326129913, "global_step": 174145, "epoch": 1956, "lr": 2.8586271034733713e-05}
{"train_loss": 0.09584146738052368, "global_step": 174146, "epoch": 1956, "lr": 2.8585747157973496e-05}
{"train_loss": 0.0929490476846695, "global_step": 174147, "epoch": 1956, "lr": 2.8585223284092145e-05}
{"train_loss": 0.06785938143730164, "global_step": 174148, "epoch": 1956, "lr": 2.8584699413089777e-05}
{"train_loss": 0.09766258299350739, "global_step": 174149, "epoch": 1956, "lr": 2.8584175544966407e-05}
{"train_loss": 0.16645145416259766, "global_step": 174150, "epoch": 1956, "lr": 2.8583651679722156e-05}
{"train_loss": 0.10502227395772934, "global_step": 174151, "epoch": 1956, "lr": 2.8583127817357048e-05}
{"train_loss": 0.10994453728199005, "global_step": 174152, "epoch": 1956, "lr": 2.85826039578712e-05}
{"train_loss": 0.041477423161268234, "global_step": 174153, "epoch": 1956, "lr": 2.8582080101264646e-05}
{"train_loss": 0.08640529960393906, "global_step": 174154, "epoch": 1956, "lr": 2.858155624753749e-05}
{"train_loss": 0.05474463850259781, "global_step": 174155, "epoch": 1956, "lr": 2.8581032396689755e-05}
{"train_loss": 0.04325517639517784, "global_step": 174156, "epoch": 1956, "lr": 2.8580508548721564e-05}
{"train_loss": 0.0726948082447052, "global_step": 174157, "epoch": 1956, "lr": 2.8579984703632933e-05}
{"train_loss": 0.07895080745220184, "global_step": 174158, "epoch": 1956, "lr": 2.857946086142399e-05}
{"train_loss": 0.10048646479845047, "global_step": 174159, "epoch": 1956, "lr": 2.857893702209476e-05}
{"train_loss": 0.048685841262340546, "global_step": 174160, "epoch": 1956, "lr": 2.857841318564535e-05}
{"train_loss": 0.0533628948032856, "global_step": 174161, "epoch": 1956, "lr": 2.8577889352075787e-05}
{"train_loss": 0.06959966570138931, "global_step": 174162, "epoch": 1956, "lr": 2.857736552138619e-05}
{"train_loss": 0.05517428368330002, "global_step": 174163, "epoch": 1956, "lr": 2.8576841693576584e-05}
{"train_loss": 0.09026488661766052, "global_step": 174164, "epoch": 1956, "lr": 2.8576317868647073e-05}
{"train_loss": 0.10504373908042908, "global_step": 174165, "epoch": 1956, "lr": 2.8575794046597704e-05}
{"train_loss": 0.08102557808160782, "global_step": 174166, "epoch": 1956, "lr": 2.857527022742856e-05}
{"train_loss": 0.09822695702314377, "global_step": 174167, "epoch": 1956, "lr": 2.8574746411139725e-05}
{"train_loss": 0.1143837496638298, "global_step": 174168, "epoch": 1956, "lr": 2.8574222597731237e-05}
{"train_loss": 0.05076214298605919, "global_step": 174169, "epoch": 1956, "lr": 2.8573698787203197e-05}
{"train_loss": 0.07011198997497559, "global_step": 174170, "epoch": 1956, "lr": 2.857317497955565e-05}
{"train_loss": 0.07630756497383118, "global_step": 174171, "epoch": 1956, "lr": 2.857265117478869e-05}
{"train_loss": 0.08162726846973548, "global_step": 174172, "epoch": 1956, "lr": 2.8572127372902363e-05, "val_loss": 7.258330345153809}
{"train_loss": 0.1321939378976822, "global_step": 174173, "epoch": 1957, "lr": 2.8571603573896773e-05}
{"train_loss": 0.14814393222332, "global_step": 174174, "epoch": 1957, "lr": 2.8571079777771948e-05}
{"train_loss": 0.0313289575278759, "global_step": 174175, "epoch": 1957, "lr": 2.857055598452799e-05}
{"train_loss": 0.0704350695014, "global_step": 174176, "epoch": 1957, "lr": 2.8570032194164953e-05}
{"train_loss": 0.09356342256069183, "global_step": 174177, "epoch": 1957, "lr": 2.856950840668293e-05}
{"train_loss": 0.09276396036148071, "global_step": 174178, "epoch": 1957, "lr": 2.8568984622081952e-05}
{"train_loss": 0.08165842294692993, "global_step": 174179, "epoch": 1957, "lr": 2.856846084036214e-05}
{"train_loss": 0.08424714207649231, "global_step": 174180, "epoch": 1957, "lr": 2.856793706152351e-05}
{"train_loss": 0.08483497053384781, "global_step": 174181, "epoch": 1957, "lr": 2.8567413285566187e-05}
{"train_loss": 0.12885664403438568, "global_step": 174182, "epoch": 1957, "lr": 2.8566889512490192e-05}
{"train_loss": 0.048830870538949966, "global_step": 174183, "epoch": 1957, "lr": 2.856636574229563e-05}
{"train_loss": 0.06785452365875244, "global_step": 174184, "epoch": 1957, "lr": 2.8565841974982565e-05}
{"train_loss": 0.06150352954864502, "global_step": 174185, "epoch": 1957, "lr": 2.856531821055104e-05}
{"train_loss": 0.0497128963470459, "global_step": 174186, "epoch": 1957, "lr": 2.8564794449001166e-05}
{"train_loss": 0.08239913731813431, "global_step": 174187, "epoch": 1957, "lr": 2.856427069033297e-05}
{"train_loss": 0.07809549570083618, "global_step": 174188, "epoch": 1957, "lr": 2.8563746934546577e-05}
{"train_loss": 0.1176401674747467, "global_step": 174189, "epoch": 1957, "lr": 2.8563223181642008e-05}
{"train_loss": 0.10962065309286118, "global_step": 174190, "epoch": 1957, "lr": 2.856269943161936e-05}
{"train_loss": 0.10731061547994614, "global_step": 174191, "epoch": 1957, "lr": 2.8562175684478686e-05}
{"train_loss": 0.07592172920703888, "global_step": 174192, "epoch": 1957, "lr": 2.856165194022008e-05}
{"train_loss": 0.09714086353778839, "global_step": 174193, "epoch": 1957, "lr": 2.856112819884358e-05}
{"train_loss": 0.10915378481149673, "global_step": 174194, "epoch": 1957, "lr": 2.8560604460349295e-05}
{"train_loss": 0.06777265667915344, "global_step": 174195, "epoch": 1957, "lr": 2.856008072473726e-05}
{"train_loss": 0.06793101131916046, "global_step": 174196, "epoch": 1957, "lr": 2.8559556992007576e-05}
{"train_loss": 0.11578701436519623, "global_step": 174197, "epoch": 1957, "lr": 2.8559033262160274e-05}
{"train_loss": 0.11292623728513718, "global_step": 174198, "epoch": 1957, "lr": 2.8558509535195476e-05}
{"train_loss": 0.07784193009138107, "global_step": 174199, "epoch": 1957, "lr": 2.8557985811113203e-05}
{"train_loss": 0.07077036052942276, "global_step": 174200, "epoch": 1957, "lr": 2.8557462089913565e-05}
{"train_loss": 0.10040722042322159, "global_step": 174201, "epoch": 1957, "lr": 2.8556938371596598e-05}
{"train_loss": 0.06355758756399155, "global_step": 174202, "epoch": 1957, "lr": 2.855641465616239e-05}
{"train_loss": 0.09893465042114258, "global_step": 174203, "epoch": 1957, "lr": 2.855589094361103e-05}
{"train_loss": 0.07465536147356033, "global_step": 174204, "epoch": 1957, "lr": 2.855536723394255e-05}
{"train_loss": 0.08567316830158234, "global_step": 174205, "epoch": 1957, "lr": 2.8554843527157054e-05}
{"train_loss": 0.060828231275081635, "global_step": 174206, "epoch": 1957, "lr": 2.855431982325458e-05}
{"train_loss": 0.09608246386051178, "global_step": 174207, "epoch": 1957, "lr": 2.8553796122235243e-05}
{"train_loss": 0.06430602818727493, "global_step": 174208, "epoch": 1957, "lr": 2.855327242409906e-05}
{"train_loss": 0.09625180810689926, "global_step": 174209, "epoch": 1957, "lr": 2.8552748728846145e-05}
{"train_loss": 0.09039638936519623, "global_step": 174210, "epoch": 1957, "lr": 2.8552225036476545e-05}
{"train_loss": 0.022251950576901436, "global_step": 174211, "epoch": 1957, "lr": 2.8551701346990346e-05}
{"train_loss": 0.11991088092327118, "global_step": 174212, "epoch": 1957, "lr": 2.855117766038759e-05}
{"train_loss": 0.1102740541100502, "global_step": 174213, "epoch": 1957, "lr": 2.8550653976668384e-05}
{"train_loss": 0.10163529217243195, "global_step": 174214, "epoch": 1957, "lr": 2.8550130295832767e-05}
{"train_loss": 0.05257660895586014, "global_step": 174215, "epoch": 1957, "lr": 2.8549606617880843e-05}
{"train_loss": 0.116280697286129, "global_step": 174216, "epoch": 1957, "lr": 2.8549082942812643e-05}
{"train_loss": 0.13600872457027435, "global_step": 174217, "epoch": 1957, "lr": 2.8548559270628278e-05}
{"train_loss": 0.1303921788930893, "global_step": 174218, "epoch": 1957, "lr": 2.8548035601327772e-05}
{"train_loss": 0.15151052176952362, "global_step": 174219, "epoch": 1957, "lr": 2.8547511934911248e-05}
{"train_loss": 0.035767726600170135, "global_step": 174220, "epoch": 1957, "lr": 2.8546988271378728e-05}
{"train_loss": 0.03538094833493233, "global_step": 174221, "epoch": 1957, "lr": 2.854646461073032e-05}
{"train_loss": 0.03511835262179375, "global_step": 174222, "epoch": 1957, "lr": 2.8545940952966083e-05}
{"train_loss": 0.11140937358140945, "global_step": 174223, "epoch": 1957, "lr": 2.8545417298086062e-05}
{"train_loss": 0.055069807916879654, "global_step": 174224, "epoch": 1957, "lr": 2.8544893646090365e-05}
{"train_loss": 0.08004526048898697, "global_step": 174225, "epoch": 1957, "lr": 2.8544369996979025e-05}
{"train_loss": 0.08606447279453278, "global_step": 174226, "epoch": 1957, "lr": 2.8543846350752158e-05}
{"train_loss": 0.10188372433185577, "global_step": 174227, "epoch": 1957, "lr": 2.8543322707409786e-05}
{"train_loss": 0.07908263802528381, "global_step": 174228, "epoch": 1957, "lr": 2.8542799066952026e-05}
{"train_loss": 0.1059284433722496, "global_step": 174229, "epoch": 1957, "lr": 2.85422754293789e-05}
{"train_loss": 0.058104678988456726, "global_step": 174230, "epoch": 1957, "lr": 2.854175179469053e-05}
{"train_loss": 0.04878862202167511, "global_step": 174231, "epoch": 1957, "lr": 2.8541228162886937e-05}
{"train_loss": 0.039894092828035355, "global_step": 174232, "epoch": 1957, "lr": 2.8540704533968232e-05}
{"train_loss": 0.10405688732862473, "global_step": 174233, "epoch": 1957, "lr": 2.8540180907934444e-05}
{"train_loss": 0.11273608356714249, "global_step": 174234, "epoch": 1957, "lr": 2.8539657284785694e-05}
{"train_loss": 0.051379989832639694, "global_step": 174235, "epoch": 1957, "lr": 2.8539133664522005e-05}
{"train_loss": 0.12121498584747314, "global_step": 174236, "epoch": 1957, "lr": 2.8538610047143488e-05}
{"train_loss": 0.11757884174585342, "global_step": 174237, "epoch": 1957, "lr": 2.8538086432650168e-05}
{"train_loss": 0.1414339393377304, "global_step": 174238, "epoch": 1957, "lr": 2.8537562821042153e-05}
{"train_loss": 0.0775502473115921, "global_step": 174239, "epoch": 1957, "lr": 2.8537039212319506e-05}
{"train_loss": 0.07733730971813202, "global_step": 174240, "epoch": 1957, "lr": 2.8536515606482283e-05}
{"train_loss": 0.06743781268596649, "global_step": 174241, "epoch": 1957, "lr": 2.8535992003530576e-05}
{"train_loss": 0.06337640434503555, "global_step": 174242, "epoch": 1957, "lr": 2.8535468403464426e-05}
{"train_loss": 0.09621285647153854, "global_step": 174243, "epoch": 1957, "lr": 2.8534944806283948e-05}
{"train_loss": 0.06459808349609375, "global_step": 174244, "epoch": 1957, "lr": 2.8534421211989158e-05}
{"train_loss": 0.08156809955835342, "global_step": 174245, "epoch": 1957, "lr": 2.8533897620580176e-05}
{"train_loss": 0.08515816926956177, "global_step": 174246, "epoch": 1957, "lr": 2.8533374032057025e-05}
{"train_loss": 0.04540850222110748, "global_step": 174247, "epoch": 1957, "lr": 2.8532850446419823e-05}
{"train_loss": 0.04875735566020012, "global_step": 174248, "epoch": 1957, "lr": 2.8532326863668602e-05}
{"train_loss": 0.08202244341373444, "global_step": 174249, "epoch": 1957, "lr": 2.853180328380346e-05}
{"train_loss": 0.06929729878902435, "global_step": 174250, "epoch": 1957, "lr": 2.8531279706824442e-05}
{"train_loss": 0.059473585337400436, "global_step": 174251, "epoch": 1957, "lr": 2.853075613273165e-05}
{"train_loss": 0.12802986800670624, "global_step": 174252, "epoch": 1957, "lr": 2.8530232561525118e-05}
{"train_loss": 0.1060086339712143, "global_step": 174253, "epoch": 1957, "lr": 2.852970899320495e-05}
{"train_loss": 0.05518222972750664, "global_step": 174254, "epoch": 1957, "lr": 2.852918542777118e-05}
{"train_loss": 0.11227677017450333, "global_step": 174255, "epoch": 1957, "lr": 2.8528661865223925e-05}
{"train_loss": 0.0876002237200737, "global_step": 174256, "epoch": 1957, "lr": 2.852813830556321e-05}
{"train_loss": 0.0672544613480568, "global_step": 174257, "epoch": 1957, "lr": 2.852761474878914e-05}
{"train_loss": 0.13565349578857422, "global_step": 174258, "epoch": 1957, "lr": 2.8527091194901755e-05}
{"train_loss": 0.09275692701339722, "global_step": 174259, "epoch": 1957, "lr": 2.852656764390116e-05}
{"train_loss": 0.10475945472717285, "global_step": 174260, "epoch": 1957, "lr": 2.8526044095787403e-05}
{"train_loss": 0.08573238871824206, "global_step": 174261, "epoch": 1957, "lr": 2.8525520550560538e-05, "val_loss": 7.257717132568359}
{"train_loss": 0.09657320380210876, "global_step": 174262, "epoch": 1958, "lr": 2.852499700822068e-05}
{"train_loss": 0.02767792157828808, "global_step": 174263, "epoch": 1958, "lr": 2.8524473468767847e-05}
{"train_loss": 0.11939676105976105, "global_step": 174264, "epoch": 1958, "lr": 2.852394993220216e-05}
{"train_loss": 0.10961076617240906, "global_step": 174265, "epoch": 1958, "lr": 2.8523426398523645e-05}
{"train_loss": 0.08407528698444366, "global_step": 174266, "epoch": 1958, "lr": 2.8522902867732416e-05}
{"train_loss": 0.07719273120164871, "global_step": 174267, "epoch": 1958, "lr": 2.8522379339828503e-05}
{"train_loss": 0.056439995765686035, "global_step": 174268, "epoch": 1958, "lr": 2.8521855814812016e-05}
{"train_loss": 0.14413438737392426, "global_step": 174269, "epoch": 1958, "lr": 2.8521332292682977e-05}
{"train_loss": 0.12891262769699097, "global_step": 174270, "epoch": 1958, "lr": 2.8520808773441503e-05}
{"train_loss": 0.053436193615198135, "global_step": 174271, "epoch": 1958, "lr": 2.852028525708763e-05}
{"train_loss": 0.08213117718696594, "global_step": 174272, "epoch": 1958, "lr": 2.8519761743621454e-05}
{"train_loss": 0.08729168772697449, "global_step": 174273, "epoch": 1958, "lr": 2.8519238233043023e-05}
{"train_loss": 0.0678090900182724, "global_step": 174274, "epoch": 1958, "lr": 2.8518714725352436e-05}
{"train_loss": 0.09802339226007462, "global_step": 174275, "epoch": 1958, "lr": 2.8518191220549727e-05}
{"train_loss": 0.06637644022703171, "global_step": 174276, "epoch": 1958, "lr": 2.8517667718634983e-05}
{"train_loss": 0.0709010660648346, "global_step": 174277, "epoch": 1958, "lr": 2.8517144219608293e-05}
{"train_loss": 0.049000125378370285, "global_step": 174278, "epoch": 1958, "lr": 2.85166207234697e-05}
{"train_loss": 0.06497540324926376, "global_step": 174279, "epoch": 1958, "lr": 2.85160972302193e-05}
{"train_loss": 0.07518827170133591, "global_step": 174280, "epoch": 1958, "lr": 2.851557373985713e-05}
{"train_loss": 0.10000920295715332, "global_step": 174281, "epoch": 1958, "lr": 2.8515050252383303e-05}
{"train_loss": 0.07787712663412094, "global_step": 174282, "epoch": 1958, "lr": 2.851452676779784e-05}
{"train_loss": 0.14702367782592773, "global_step": 174283, "epoch": 1958, "lr": 2.8514003286100858e-05}
{"train_loss": 0.13932551443576813, "global_step": 174284, "epoch": 1958, "lr": 2.8513479807292387e-05}
{"train_loss": 0.06552540510892868, "global_step": 174285, "epoch": 1958, "lr": 2.8512956331372543e-05}
{"train_loss": 0.05902661010622978, "global_step": 174286, "epoch": 1958, "lr": 2.851243285834134e-05}
{"train_loss": 0.08517970889806747, "global_step": 174287, "epoch": 1958, "lr": 2.8511909388198905e-05}
{"train_loss": 0.12093387544155121, "global_step": 174288, "epoch": 1958, "lr": 2.8511385920945254e-05}
{"train_loss": 0.09750313311815262, "global_step": 174289, "epoch": 1958, "lr": 2.851086245658051e-05}
{"train_loss": 0.10514921694993973, "global_step": 174290, "epoch": 1958, "lr": 2.8510338995104702e-05}
{"train_loss": 0.06082621216773987, "global_step": 174291, "epoch": 1958, "lr": 2.8509815536517937e-05}
{"train_loss": 0.1186143159866333, "global_step": 174292, "epoch": 1958, "lr": 2.8509292080820245e-05}
{"train_loss": 0.054233111441135406, "global_step": 174293, "epoch": 1958, "lr": 2.8508768628011735e-05}
{"train_loss": 0.06736729294061661, "global_step": 174294, "epoch": 1958, "lr": 2.850824517809244e-05}
{"train_loss": 0.11472760885953903, "global_step": 174295, "epoch": 1958, "lr": 2.8507721731062464e-05}
{"train_loss": 0.03853049874305725, "global_step": 174296, "epoch": 1958, "lr": 2.850719828692185e-05}
{"train_loss": 0.09997580200433731, "global_step": 174297, "epoch": 1958, "lr": 2.85066748456707e-05}
{"train_loss": 0.07125964015722275, "global_step": 174298, "epoch": 1958, "lr": 2.850615140730904e-05}
{"train_loss": 0.0987534448504448, "global_step": 174299, "epoch": 1958, "lr": 2.850562797183699e-05}
{"train_loss": 0.06718264520168304, "global_step": 174300, "epoch": 1958, "lr": 2.85051045392546e-05}
{"train_loss": 0.09642593562602997, "global_step": 174301, "epoch": 1958, "lr": 2.8504581109561912e-05}
{"train_loss": 0.1281076818704605, "global_step": 174302, "epoch": 1958, "lr": 2.8504057682759045e-05}
{"train_loss": 0.04446278512477875, "global_step": 174303, "epoch": 1958, "lr": 2.850353425884602e-05}
{"train_loss": 0.07242172956466675, "global_step": 174304, "epoch": 1958, "lr": 2.8503010837822953e-05}
{"train_loss": 0.03685056045651436, "global_step": 174305, "epoch": 1958, "lr": 2.850248741968988e-05}
{"train_loss": 0.09000067412853241, "global_step": 174306, "epoch": 1958, "lr": 2.85019640044469e-05}
{"train_loss": 0.11996994912624359, "global_step": 174307, "epoch": 1958, "lr": 2.8501440592094052e-05}
{"train_loss": 0.06937003880739212, "global_step": 174308, "epoch": 1958, "lr": 2.8500917182631437e-05}
{"train_loss": 0.08248824626207352, "global_step": 174309, "epoch": 1958, "lr": 2.8500393776059098e-05}
{"train_loss": 0.101401686668396, "global_step": 174310, "epoch": 1958, "lr": 2.849987037237714e-05}
{"train_loss": 0.14226539433002472, "global_step": 174311, "epoch": 1958, "lr": 2.8499346971585594e-05}
{"train_loss": 0.045919034630060196, "global_step": 174312, "epoch": 1958, "lr": 2.849882357368454e-05}
{"train_loss": 0.09060223400592804, "global_step": 174313, "epoch": 1958, "lr": 2.8498300178674087e-05}
{"train_loss": 0.053915902972221375, "global_step": 174314, "epoch": 1958, "lr": 2.8497776786554247e-05}
{"train_loss": 0.043793145567178726, "global_step": 174315, "epoch": 1958, "lr": 2.849725339732514e-05}
{"train_loss": 0.05344511941075325, "global_step": 174316, "epoch": 1958, "lr": 2.8496730010986807e-05}
{"train_loss": 0.0603472925722599, "global_step": 174317, "epoch": 1958, "lr": 2.8496206627539334e-05}
{"train_loss": 0.12918446958065033, "global_step": 174318, "epoch": 1958, "lr": 2.849568324698276e-05}
{"train_loss": 0.1112971231341362, "global_step": 174319, "epoch": 1958, "lr": 2.849515986931721e-05}
{"train_loss": 0.054164763540029526, "global_step": 174320, "epoch": 1958, "lr": 2.8494636494542702e-05}
{"train_loss": 0.0744575634598732, "global_step": 174321, "epoch": 1958, "lr": 2.849411312265935e-05}
{"train_loss": 0.045845311135053635, "global_step": 174322, "epoch": 1958, "lr": 2.8493589753667173e-05}
{"train_loss": 0.039189498871564865, "global_step": 174323, "epoch": 1958, "lr": 2.84930663875663e-05}
{"train_loss": 0.053878169506788254, "global_step": 174324, "epoch": 1958, "lr": 2.8492543024356744e-05}
{"train_loss": 0.07819615304470062, "global_step": 174325, "epoch": 1958, "lr": 2.8492019664038628e-05}
{"train_loss": 0.10653682798147202, "global_step": 174326, "epoch": 1958, "lr": 2.8491496306611977e-05}
{"train_loss": 0.09579606354236603, "global_step": 174327, "epoch": 1958, "lr": 2.8490972952076907e-05}
{"train_loss": 0.14484558999538422, "global_step": 174328, "epoch": 1958, "lr": 2.8490449600433438e-05}
{"train_loss": 0.09084607660770416, "global_step": 174329, "epoch": 1958, "lr": 2.8489926251681665e-05}
{"train_loss": 0.09881473332643509, "global_step": 174330, "epoch": 1958, "lr": 2.8489402905821683e-05}
{"train_loss": 0.09804937243461609, "global_step": 174331, "epoch": 1958, "lr": 2.8488879562853522e-05}
{"train_loss": 0.05081562697887421, "global_step": 174332, "epoch": 1958, "lr": 2.8488356222777278e-05}
{"train_loss": 0.05712161958217621, "global_step": 174333, "epoch": 1958, "lr": 2.848783288559302e-05}
{"train_loss": 0.10882668197154999, "global_step": 174334, "epoch": 1958, "lr": 2.8487309551300788e-05}
{"train_loss": 0.08000592142343521, "global_step": 174335, "epoch": 1958, "lr": 2.8486786219900695e-05}
{"train_loss": 0.05875694379210472, "global_step": 174336, "epoch": 1958, "lr": 2.8486262891392772e-05}
{"train_loss": 0.1467292755842209, "global_step": 174337, "epoch": 1958, "lr": 2.848573956577713e-05}
{"train_loss": 0.0748520940542221, "global_step": 174338, "epoch": 1958, "lr": 2.848521624305382e-05}
{"train_loss": 0.042631421238183975, "global_step": 174339, "epoch": 1958, "lr": 2.848469292322289e-05}
{"train_loss": 0.10095377266407013, "global_step": 174340, "epoch": 1958, "lr": 2.8484169606284445e-05}
{"train_loss": 0.0769021064043045, "global_step": 174341, "epoch": 1958, "lr": 2.8483646292238525e-05}
{"train_loss": 0.1045365259051323, "global_step": 174342, "epoch": 1958, "lr": 2.8483122981085237e-05}
{"train_loss": 0.09307408332824707, "global_step": 174343, "epoch": 1958, "lr": 2.8482599672824607e-05}
{"train_loss": 0.10559041798114777, "global_step": 174344, "epoch": 1958, "lr": 2.8482076367456757e-05}
{"train_loss": 0.04635203629732132, "global_step": 174345, "epoch": 1958, "lr": 2.8481553064981705e-05}
{"train_loss": 0.06412069499492645, "global_step": 174346, "epoch": 1958, "lr": 2.8481029765399557e-05}
{"train_loss": 0.06081284582614899, "global_step": 174347, "epoch": 1958, "lr": 2.8480506468710366e-05}
{"train_loss": 0.07790151238441467, "global_step": 174348, "epoch": 1958, "lr": 2.8479983174914203e-05}
{"train_loss": 0.06599519401788712, "global_step": 174349, "epoch": 1958, "lr": 2.8479459884011162e-05}
{"train_loss": 0.08285471924653884, "global_step": 174350, "epoch": 1958, "lr": 2.8478936596001284e-05, "val_loss": 7.4748406410217285}
{"train_loss": 0.0514586940407753, "global_step": 174351, "epoch": 1959, "lr": 2.847841331088466e-05}
{"train_loss": 0.04699043929576874, "global_step": 174352, "epoch": 1959, "lr": 2.8477890028661335e-05}
{"train_loss": 0.10365327447652817, "global_step": 174353, "epoch": 1959, "lr": 2.847736674933142e-05}
{"train_loss": 0.08322352916002274, "global_step": 174354, "epoch": 1959, "lr": 2.8476843472894933e-05}
{"train_loss": 0.062456972897052765, "global_step": 174355, "epoch": 1959, "lr": 2.8476320199351998e-05}
{"train_loss": 0.07012077420949936, "global_step": 174356, "epoch": 1959, "lr": 2.8475796928702636e-05}
{"train_loss": 0.0999814048409462, "global_step": 174357, "epoch": 1959, "lr": 2.8475273660946966e-05}
{"train_loss": 0.10605636239051819, "global_step": 174358, "epoch": 1959, "lr": 2.8474750396085004e-05}
{"train_loss": 0.08614587783813477, "global_step": 174359, "epoch": 1959, "lr": 2.8474227134116876e-05}
{"train_loss": 0.0496644489467144, "global_step": 174360, "epoch": 1959, "lr": 2.8473703875042602e-05}
{"train_loss": 0.05359296128153801, "global_step": 174361, "epoch": 1959, "lr": 2.84731806188623e-05}
{"train_loss": 0.05988795682787895, "global_step": 174362, "epoch": 1959, "lr": 2.8472657365575993e-05}
{"train_loss": 0.1226755678653717, "global_step": 174363, "epoch": 1959, "lr": 2.84721341151838e-05}
{"train_loss": 0.0936206802725792, "global_step": 174364, "epoch": 1959, "lr": 2.847161086768574e-05}
{"train_loss": 0.06472885608673096, "global_step": 174365, "epoch": 1959, "lr": 2.8471087623081915e-05}
{"train_loss": 0.10899614542722702, "global_step": 174366, "epoch": 1959, "lr": 2.847056438137241e-05}
{"train_loss": 0.12321851402521133, "global_step": 174367, "epoch": 1959, "lr": 2.8470041142557247e-05}
{"train_loss": 0.10520682483911514, "global_step": 174368, "epoch": 1959, "lr": 2.8469517906636557e-05}
{"train_loss": 0.04393020272254944, "global_step": 174369, "epoch": 1959, "lr": 2.846899467361035e-05}
{"train_loss": 0.07979973405599594, "global_step": 174370, "epoch": 1959, "lr": 2.846847144347874e-05}
{"train_loss": 0.06094133481383324, "global_step": 174371, "epoch": 1959, "lr": 2.8467948216241793e-05}
{"train_loss": 0.05295940861105919, "global_step": 174372, "epoch": 1959, "lr": 2.846742499189954e-05}
{"train_loss": 0.0435737781226635, "global_step": 174373, "epoch": 1959, "lr": 2.84669017704521e-05}
{"train_loss": 0.09400434792041779, "global_step": 174374, "epoch": 1959, "lr": 2.8466378551899505e-05}
{"train_loss": 0.06929474323987961, "global_step": 174375, "epoch": 1959, "lr": 2.846585533624186e-05}
{"train_loss": 0.03966839984059334, "global_step": 174376, "epoch": 1959, "lr": 2.8465332123479217e-05}
{"train_loss": 0.09677091985940933, "global_step": 174377, "epoch": 1959, "lr": 2.846480891361163e-05}
{"train_loss": 0.07720157504081726, "global_step": 174378, "epoch": 1959, "lr": 2.846428570663921e-05}
{"train_loss": 0.07868176698684692, "global_step": 174379, "epoch": 1959, "lr": 2.846376250256198e-05}
{"train_loss": 0.08474096655845642, "global_step": 174380, "epoch": 1959, "lr": 2.8463239301380058e-05}
{"train_loss": 0.09359412640333176, "global_step": 174381, "epoch": 1959, "lr": 2.846271610309347e-05}
{"train_loss": 0.09106187522411346, "global_step": 174382, "epoch": 1959, "lr": 2.8462192907702324e-05}
{"train_loss": 0.112225241959095, "global_step": 174383, "epoch": 1959, "lr": 2.8461669715206656e-05}
{"train_loss": 0.09868562966585159, "global_step": 174384, "epoch": 1959, "lr": 2.8461146525606553e-05}
{"train_loss": 0.09182541817426682, "global_step": 174385, "epoch": 1959, "lr": 2.846062333890211e-05}
{"train_loss": 0.08775488287210464, "global_step": 174386, "epoch": 1959, "lr": 2.846010015509335e-05}
{"train_loss": 0.08685298264026642, "global_step": 174387, "epoch": 1959, "lr": 2.845957697418039e-05}
{"train_loss": 0.046544481068849564, "global_step": 174388, "epoch": 1959, "lr": 2.845905379616326e-05}
{"train_loss": 0.13111087679862976, "global_step": 174389, "epoch": 1959, "lr": 2.8458530621042057e-05}
{"train_loss": 0.10027611255645752, "global_step": 174390, "epoch": 1959, "lr": 2.8458007448816833e-05}
{"train_loss": 0.10347264260053635, "global_step": 174391, "epoch": 1959, "lr": 2.845748427948768e-05}
{"train_loss": 0.08746812492609024, "global_step": 174392, "epoch": 1959, "lr": 2.8456961113054638e-05}
{"train_loss": 0.06365683674812317, "global_step": 174393, "epoch": 1959, "lr": 2.845643794951781e-05}
{"train_loss": 0.11894045025110245, "global_step": 174394, "epoch": 1959, "lr": 2.8455914788877242e-05}
{"train_loss": 0.06599385291337967, "global_step": 174395, "epoch": 1959, "lr": 2.845539163113303e-05}
{"train_loss": 0.13899961113929749, "global_step": 174396, "epoch": 1959, "lr": 2.84548684762852e-05}
{"train_loss": 0.025980144739151, "global_step": 174397, "epoch": 1959, "lr": 2.8454345324333882e-05}
{"train_loss": 0.07911791652441025, "global_step": 174398, "epoch": 1959, "lr": 2.8453822175279088e-05}
{"train_loss": 0.06688491255044937, "global_step": 174399, "epoch": 1959, "lr": 2.845329902912094e-05}
{"train_loss": 0.05780097097158432, "global_step": 174400, "epoch": 1959, "lr": 2.8452775885859455e-05}
{"train_loss": 0.12197474390268326, "global_step": 174401, "epoch": 1959, "lr": 2.8452252745494755e-05}
{"train_loss": 0.11229506880044937, "global_step": 174402, "epoch": 1959, "lr": 2.8451729608026873e-05}
{"train_loss": 0.1395261585712433, "global_step": 174403, "epoch": 1959, "lr": 2.845120647345589e-05}
{"train_loss": 0.08035764843225479, "global_step": 174404, "epoch": 1959, "lr": 2.8450683341781893e-05}
{"train_loss": 0.05077120289206505, "global_step": 174405, "epoch": 1959, "lr": 2.845016021300493e-05}
{"train_loss": 0.05768571048974991, "global_step": 174406, "epoch": 1959, "lr": 2.844963708712509e-05}
{"train_loss": 0.03902113437652588, "global_step": 174407, "epoch": 1959, "lr": 2.8449113964142415e-05}
{"train_loss": 0.04834664985537529, "global_step": 174408, "epoch": 1959, "lr": 2.8448590844057022e-05}
{"train_loss": 0.07562873512506485, "global_step": 174409, "epoch": 1959, "lr": 2.8448067726868926e-05}
{"train_loss": 0.06594385206699371, "global_step": 174410, "epoch": 1959, "lr": 2.8447544612578248e-05}
{"train_loss": 0.09734075516462326, "global_step": 174411, "epoch": 1959, "lr": 2.8447021501185032e-05}
{"train_loss": 0.05813179910182953, "global_step": 174412, "epoch": 1959, "lr": 2.8446498392689337e-05}
{"train_loss": 0.08221957832574844, "global_step": 174413, "epoch": 1959, "lr": 2.8445975287091263e-05}
{"train_loss": 0.06002601608633995, "global_step": 174414, "epoch": 1959, "lr": 2.8445452184390842e-05}
{"train_loss": 0.06865105032920837, "global_step": 174415, "epoch": 1959, "lr": 2.8444929084588195e-05}
{"train_loss": 0.07333818823099136, "global_step": 174416, "epoch": 1959, "lr": 2.844440598768336e-05}
{"train_loss": 0.044716138392686844, "global_step": 174417, "epoch": 1959, "lr": 2.8443882893676393e-05}
{"train_loss": 0.11642322689294815, "global_step": 174418, "epoch": 1959, "lr": 2.8443359802567404e-05}
{"train_loss": 0.1156415119767189, "global_step": 174419, "epoch": 1959, "lr": 2.844283671435642e-05}
{"train_loss": 0.061780642718076706, "global_step": 174420, "epoch": 1959, "lr": 2.844231362904354e-05}
{"train_loss": 0.1028192937374115, "global_step": 174421, "epoch": 1959, "lr": 2.844179054662884e-05}
{"train_loss": 0.07692693918943405, "global_step": 174422, "epoch": 1959, "lr": 2.844126746711236e-05}
{"train_loss": 0.08431001007556915, "global_step": 174423, "epoch": 1959, "lr": 2.844074439049421e-05}
{"train_loss": 0.08135943114757538, "global_step": 174424, "epoch": 1959, "lr": 2.844022131677442e-05}
{"train_loss": 0.1172683909535408, "global_step": 174425, "epoch": 1959, "lr": 2.8439698245953095e-05}
{"train_loss": 0.07404691725969315, "global_step": 174426, "epoch": 1959, "lr": 2.8439175178030275e-05}
{"train_loss": 0.10452976077795029, "global_step": 174427, "epoch": 1959, "lr": 2.843865211300607e-05}
{"train_loss": 0.07493788748979568, "global_step": 174428, "epoch": 1959, "lr": 2.8438129050880497e-05}
{"train_loss": 0.05793257802724838, "global_step": 174429, "epoch": 1959, "lr": 2.8437605991653676e-05}
{"train_loss": 0.0597669892013073, "global_step": 174430, "epoch": 1959, "lr": 2.843708293532564e-05}
{"train_loss": 0.062342505902051926, "global_step": 174431, "epoch": 1959, "lr": 2.843655988189649e-05}
{"train_loss": 0.09365484118461609, "global_step": 174432, "epoch": 1959, "lr": 2.8436036831366265e-05}
{"train_loss": 0.07756492495536804, "global_step": 174433, "epoch": 1959, "lr": 2.8435513783735068e-05}
{"train_loss": 0.1020219475030899, "global_step": 174434, "epoch": 1959, "lr": 2.8434990739002942e-05}
{"train_loss": 0.05925429239869118, "global_step": 174435, "epoch": 1959, "lr": 2.843446769716998e-05}
{"train_loss": 0.05910402908921242, "global_step": 174436, "epoch": 1959, "lr": 2.843394465823622e-05}
{"train_loss": 0.09393653273582458, "global_step": 174437, "epoch": 1959, "lr": 2.8433421622201782e-05}
{"train_loss": 0.059444110840559006, "global_step": 174438, "epoch": 1959, "lr": 2.8432898589066683e-05}
{"train_loss": 0.08059917591261061, "global_step": 174439, "epoch": 1959, "lr": 2.843237555883102e-05, "val_loss": 7.224456787109375}
{"train_loss": 0.04188133403658867, "global_step": 174440, "epoch": 1960, "lr": 2.843185253149488e-05}
{"train_loss": 0.070957712829113, "global_step": 174441, "epoch": 1960, "lr": 2.8431329507058292e-05}
{"train_loss": 0.07583294063806534, "global_step": 174442, "epoch": 1960, "lr": 2.843080648552137e-05}
{"train_loss": 0.1049632579088211, "global_step": 174443, "epoch": 1960, "lr": 2.8430283466884145e-05}
{"train_loss": 0.09915940463542938, "global_step": 174444, "epoch": 1960, "lr": 2.8429760451146726e-05}
{"train_loss": 0.11351849883794785, "global_step": 174445, "epoch": 1960, "lr": 2.8429237438309142e-05}
{"train_loss": 0.03087787516415119, "global_step": 174446, "epoch": 1960, "lr": 2.8428714428371507e-05}
{"train_loss": 0.07456301152706146, "global_step": 174447, "epoch": 1960, "lr": 2.8428191421333838e-05}
{"train_loss": 0.11730508506298065, "global_step": 174448, "epoch": 1960, "lr": 2.8427668417196264e-05}
{"train_loss": 0.05855561047792435, "global_step": 174449, "epoch": 1960, "lr": 2.8427145415958823e-05}
{"train_loss": 0.09071440249681473, "global_step": 174450, "epoch": 1960, "lr": 2.8426622417621575e-05}
{"train_loss": 0.0353081077337265, "global_step": 174451, "epoch": 1960, "lr": 2.8426099422184622e-05}
{"train_loss": 0.09169678390026093, "global_step": 174452, "epoch": 1960, "lr": 2.8425576429647995e-05}
{"train_loss": 0.0703757107257843, "global_step": 174453, "epoch": 1960, "lr": 2.8425053440011805e-05}
{"train_loss": 0.056172337383031845, "global_step": 174454, "epoch": 1960, "lr": 2.842453045327611e-05}
{"train_loss": 0.10795407742261887, "global_step": 174455, "epoch": 1960, "lr": 2.842400746944095e-05}
{"train_loss": 0.08404092490673065, "global_step": 174456, "epoch": 1960, "lr": 2.8423484488506424e-05}
{"train_loss": 0.07939627021551132, "global_step": 174457, "epoch": 1960, "lr": 2.842296151047261e-05}
{"train_loss": 0.06940673291683197, "global_step": 174458, "epoch": 1960, "lr": 2.8422438535339546e-05}
{"train_loss": 0.04565076529979706, "global_step": 174459, "epoch": 1960, "lr": 2.8421915563107348e-05}
{"train_loss": 0.11190915107727051, "global_step": 174460, "epoch": 1960, "lr": 2.8421392593776042e-05}
{"train_loss": 0.07104282081127167, "global_step": 174461, "epoch": 1960, "lr": 2.8420869627345726e-05}
{"train_loss": 0.10874440521001816, "global_step": 174462, "epoch": 1960, "lr": 2.842034666381645e-05}
{"train_loss": 0.10010917484760284, "global_step": 174463, "epoch": 1960, "lr": 2.841982370318831e-05}
{"train_loss": 0.13844726979732513, "global_step": 174464, "epoch": 1960, "lr": 2.8419300745461348e-05}
{"train_loss": 0.05979117006063461, "global_step": 174465, "epoch": 1960, "lr": 2.8418777790635664e-05}
{"train_loss": 0.13708196580410004, "global_step": 174466, "epoch": 1960, "lr": 2.841825483871129e-05}
{"train_loss": 0.10648585855960846, "global_step": 174467, "epoch": 1960, "lr": 2.8417731889688343e-05}
{"train_loss": 0.08966002613306046, "global_step": 174468, "epoch": 1960, "lr": 2.8417208943566848e-05}
{"train_loss": 0.07944321632385254, "global_step": 174469, "epoch": 1960, "lr": 2.8416686000346914e-05}
{"train_loss": 0.08746213465929031, "global_step": 174470, "epoch": 1960, "lr": 2.8416163060028577e-05}
{"train_loss": 0.06490100175142288, "global_step": 174471, "epoch": 1960, "lr": 2.841564012261194e-05}
{"train_loss": 0.05089616775512695, "global_step": 174472, "epoch": 1960, "lr": 2.841511718809704e-05}
{"train_loss": 0.10780476033687592, "global_step": 174473, "epoch": 1960, "lr": 2.8414594256483983e-05}
{"train_loss": 0.0919208750128746, "global_step": 174474, "epoch": 1960, "lr": 2.8414071327772803e-05}
{"train_loss": 0.1358826756477356, "global_step": 174475, "epoch": 1960, "lr": 2.8413548401963584e-05}
{"train_loss": 0.04929210990667343, "global_step": 174476, "epoch": 1960, "lr": 2.841302547905642e-05}
{"train_loss": 0.09444110840559006, "global_step": 174477, "epoch": 1960, "lr": 2.8412502559051345e-05}
{"train_loss": 0.08831235766410828, "global_step": 174478, "epoch": 1960, "lr": 2.8411979641948467e-05}
{"train_loss": 0.10705297440290451, "global_step": 174479, "epoch": 1960, "lr": 2.8411456727747808e-05}
{"train_loss": 0.1816612184047699, "global_step": 174480, "epoch": 1960, "lr": 2.841093381644949e-05}
{"train_loss": 0.054301053285598755, "global_step": 174481, "epoch": 1960, "lr": 2.8410410908053542e-05}
{"train_loss": 0.08185164630413055, "global_step": 174482, "epoch": 1960, "lr": 2.840988800256007e-05}
{"train_loss": 0.08588691800832748, "global_step": 174483, "epoch": 1960, "lr": 2.8409365099969106e-05}
{"train_loss": 0.08500572293996811, "global_step": 174484, "epoch": 1960, "lr": 2.840884220028076e-05}
{"train_loss": 0.13256268203258514, "global_step": 174485, "epoch": 1960, "lr": 2.840831930349506e-05}
{"train_loss": 0.0981714203953743, "global_step": 174486, "epoch": 1960, "lr": 2.8407796409612114e-05}
{"train_loss": 0.10227315872907639, "global_step": 174487, "epoch": 1960, "lr": 2.8407273518631983e-05}
{"train_loss": 0.04542921483516693, "global_step": 174488, "epoch": 1960, "lr": 2.8406750630554708e-05}
{"train_loss": 0.05010523647069931, "global_step": 174489, "epoch": 1960, "lr": 2.8406227745380398e-05}
{"train_loss": 0.1149468794465065, "global_step": 174490, "epoch": 1960, "lr": 2.8405704863109096e-05}
{"train_loss": 0.07073625177145004, "global_step": 174491, "epoch": 1960, "lr": 2.8405181983740902e-05}
{"train_loss": 0.06761837750673294, "global_step": 174492, "epoch": 1960, "lr": 2.8404659107275842e-05}
{"train_loss": 0.08721798658370972, "global_step": 174493, "epoch": 1960, "lr": 2.840413623371404e-05}
{"train_loss": 0.08048669248819351, "global_step": 174494, "epoch": 1960, "lr": 2.8403613363055515e-05}
{"train_loss": 0.10684535652399063, "global_step": 174495, "epoch": 1960, "lr": 2.8403090495300376e-05}
{"train_loss": 0.11247865855693817, "global_step": 174496, "epoch": 1960, "lr": 2.8402567630448663e-05}
{"train_loss": 0.06143205612897873, "global_step": 174497, "epoch": 1960, "lr": 2.8402044768500485e-05}
{"train_loss": 0.10307887941598892, "global_step": 174498, "epoch": 1960, "lr": 2.8401521909455863e-05}
{"train_loss": 0.10316652804613113, "global_step": 174499, "epoch": 1960, "lr": 2.840099905331492e-05}
{"train_loss": 0.07987670600414276, "global_step": 174500, "epoch": 1960, "lr": 2.8400476200077674e-05}
{"train_loss": 0.06558909267187119, "global_step": 174501, "epoch": 1960, "lr": 2.839995334974424e-05}
{"train_loss": 0.09940503537654877, "global_step": 174502, "epoch": 1960, "lr": 2.839943050231465e-05}
{"train_loss": 0.06055840104818344, "global_step": 174503, "epoch": 1960, "lr": 2.839890765778902e-05}
{"train_loss": 0.08122996985912323, "global_step": 174504, "epoch": 1960, "lr": 2.8398384816167367e-05}
{"train_loss": 0.03535321354866028, "global_step": 174505, "epoch": 1960, "lr": 2.83978619774498e-05}
{"train_loss": 0.11562806367874146, "global_step": 174506, "epoch": 1960, "lr": 2.839733914163637e-05}
{"train_loss": 0.0599311925470829, "global_step": 174507, "epoch": 1960, "lr": 2.839681630872717e-05}
{"train_loss": 0.13318641483783722, "global_step": 174508, "epoch": 1960, "lr": 2.8396293478722234e-05}
{"train_loss": 0.07265204936265945, "global_step": 174509, "epoch": 1960, "lr": 2.839577065162168e-05}
{"train_loss": 0.07459638267755508, "global_step": 174510, "epoch": 1960, "lr": 2.839524782742552e-05}
{"train_loss": 0.04233330488204956, "global_step": 174511, "epoch": 1960, "lr": 2.839472500613386e-05}
{"train_loss": 0.09341304749250412, "global_step": 174512, "epoch": 1960, "lr": 2.839420218774679e-05}
{"train_loss": 0.03882978856563568, "global_step": 174513, "epoch": 1960, "lr": 2.8393679372264338e-05}
{"train_loss": 0.057305172085762024, "global_step": 174514, "epoch": 1960, "lr": 2.8393156559686606e-05}
{"train_loss": 0.07274924963712692, "global_step": 174515, "epoch": 1960, "lr": 2.839263375001363e-05}
{"train_loss": 0.03892539069056511, "global_step": 174516, "epoch": 1960, "lr": 2.839211094324552e-05}
{"train_loss": 0.13392876088619232, "global_step": 174517, "epoch": 1960, "lr": 2.8391588139382312e-05}
{"train_loss": 0.0662003830075264, "global_step": 174518, "epoch": 1960, "lr": 2.839106533842411e-05}
{"train_loss": 0.08340632170438766, "global_step": 174519, "epoch": 1960, "lr": 2.8390542540370947e-05}
{"train_loss": 0.10207509994506836, "global_step": 174520, "epoch": 1960, "lr": 2.8390019745222927e-05}
{"train_loss": 0.10704267024993896, "global_step": 174521, "epoch": 1960, "lr": 2.8389496952980095e-05}
{"train_loss": 0.07924780249595642, "global_step": 174522, "epoch": 1960, "lr": 2.8388974163642545e-05}
{"train_loss": 0.07227898389101028, "global_step": 174523, "epoch": 1960, "lr": 2.8388451377210318e-05}
{"train_loss": 0.07216382771730423, "global_step": 174524, "epoch": 1960, "lr": 2.8387928593683515e-05}
{"train_loss": 0.08537279814481735, "global_step": 174525, "epoch": 1960, "lr": 2.8387405813062175e-05}
{"train_loss": 0.07660039514303207, "global_step": 174526, "epoch": 1960, "lr": 2.83868830353464e-05}
{"train_loss": 0.05575866997241974, "global_step": 174527, "epoch": 1960, "lr": 2.8386360260536256e-05}
{"train_loss": 0.08343854188667925, "global_step": 174528, "epoch": 1960, "lr": 2.8385837488631772e-05, "val_loss": 7.227214336395264, "train_action_mse_error": 14.407257080078125}
{"train_loss": 0.06344174593687057, "global_step": 174529, "epoch": 1961, "lr": 2.8385314719633072e-05}
{"train_loss": 0.06462160497903824, "global_step": 174530, "epoch": 1961, "lr": 2.8384791953540186e-05}
{"train_loss": 0.08348708599805832, "global_step": 174531, "epoch": 1961, "lr": 2.838426919035322e-05}
{"train_loss": 0.029849324375391006, "global_step": 174532, "epoch": 1961, "lr": 2.83837464300722e-05}
{"train_loss": 0.07230788469314575, "global_step": 174533, "epoch": 1961, "lr": 2.8383223672697252e-05}
{"train_loss": 0.1280708909034729, "global_step": 174534, "epoch": 1961, "lr": 2.838270091822839e-05}
{"train_loss": 0.10387030988931656, "global_step": 174535, "epoch": 1961, "lr": 2.838217816666573e-05}
{"train_loss": 0.05507408082485199, "global_step": 174536, "epoch": 1961, "lr": 2.8381655418009302e-05}
{"train_loss": 0.10810725390911102, "global_step": 174537, "epoch": 1961, "lr": 2.838113267225922e-05}
{"train_loss": 0.09457524865865707, "global_step": 174538, "epoch": 1961, "lr": 2.838060992941551e-05}
{"train_loss": 0.10891389101743698, "global_step": 174539, "epoch": 1961, "lr": 2.8380087189478276e-05}
{"train_loss": 0.1089540645480156, "global_step": 174540, "epoch": 1961, "lr": 2.8379564452447572e-05}
{"train_loss": 0.047341227531433105, "global_step": 174541, "epoch": 1961, "lr": 2.8379041718323478e-05}
{"train_loss": 0.10554992407560349, "global_step": 174542, "epoch": 1961, "lr": 2.837851898710604e-05}
{"train_loss": 0.04011840745806694, "global_step": 174543, "epoch": 1961, "lr": 2.837799625879537e-05}
{"train_loss": 0.12409144639968872, "global_step": 174544, "epoch": 1961, "lr": 2.8377473533391497e-05}
{"train_loss": 0.09779747575521469, "global_step": 174545, "epoch": 1961, "lr": 2.837695081089453e-05}
{"train_loss": 0.0911269560456276, "global_step": 174546, "epoch": 1961, "lr": 2.8376428091304495e-05}
{"train_loss": 0.04420292750000954, "global_step": 174547, "epoch": 1961, "lr": 2.837590537462148e-05}
{"train_loss": 0.07142861932516098, "global_step": 174548, "epoch": 1961, "lr": 2.837538266084559e-05}
{"train_loss": 0.09097302705049515, "global_step": 174549, "epoch": 1961, "lr": 2.837485994997684e-05}
{"train_loss": 0.08718522638082504, "global_step": 174550, "epoch": 1961, "lr": 2.8374337242015353e-05}
{"train_loss": 0.10551057755947113, "global_step": 174551, "epoch": 1961, "lr": 2.8373814536961152e-05}
{"train_loss": 0.08093290776014328, "global_step": 174552, "epoch": 1961, "lr": 2.837329183481434e-05}
{"train_loss": 0.06271678954362869, "global_step": 174553, "epoch": 1961, "lr": 2.837276913557496e-05}
{"train_loss": 0.07639329135417938, "global_step": 174554, "epoch": 1961, "lr": 2.8372246439243123e-05}
{"train_loss": 0.12937545776367188, "global_step": 174555, "epoch": 1961, "lr": 2.8371723745818847e-05}
{"train_loss": 0.06775597482919693, "global_step": 174556, "epoch": 1961, "lr": 2.8371201055302256e-05}
{"train_loss": 0.07720030844211578, "global_step": 174557, "epoch": 1961, "lr": 2.8370678367693366e-05}
{"train_loss": 0.07695701718330383, "global_step": 174558, "epoch": 1961, "lr": 2.8370155682992293e-05}
{"train_loss": 0.07278946787118912, "global_step": 174559, "epoch": 1961, "lr": 2.8369633001199074e-05}
{"train_loss": 0.05024731159210205, "global_step": 174560, "epoch": 1961, "lr": 2.8369110322313817e-05}
{"train_loss": 0.06875543296337128, "global_step": 174561, "epoch": 1961, "lr": 2.8368587646336543e-05}
{"train_loss": 0.13099446892738342, "global_step": 174562, "epoch": 1961, "lr": 2.836806497326737e-05}
{"train_loss": 0.10984478890895844, "global_step": 174563, "epoch": 1961, "lr": 2.8367542303106333e-05}
{"train_loss": 0.12078341096639633, "global_step": 174564, "epoch": 1961, "lr": 2.8367019635853532e-05}
{"train_loss": 0.09758039563894272, "global_step": 174565, "epoch": 1961, "lr": 2.8366496971509022e-05}
{"train_loss": 0.09323429316282272, "global_step": 174566, "epoch": 1961, "lr": 2.836597431007285e-05}
{"train_loss": 0.06352192908525467, "global_step": 174567, "epoch": 1961, "lr": 2.8365451651545133e-05}
{"train_loss": 0.10350289195775986, "global_step": 174568, "epoch": 1961, "lr": 2.8364928995925893e-05}
{"train_loss": 0.05228205770254135, "global_step": 174569, "epoch": 1961, "lr": 2.836440634321525e-05}
{"train_loss": 0.08543851971626282, "global_step": 174570, "epoch": 1961, "lr": 2.836388369341323e-05}
{"train_loss": 0.10085500031709671, "global_step": 174571, "epoch": 1961, "lr": 2.836336104651993e-05}
{"train_loss": 0.07626768201589584, "global_step": 174572, "epoch": 1961, "lr": 2.8362838402535406e-05}
{"train_loss": 0.08850443363189697, "global_step": 174573, "epoch": 1961, "lr": 2.8362315761459745e-05}
{"train_loss": 0.05094786360859871, "global_step": 174574, "epoch": 1961, "lr": 2.836179312329299e-05}
{"train_loss": 0.07079660147428513, "global_step": 174575, "epoch": 1961, "lr": 2.836127048803525e-05}
{"train_loss": 0.11528903245925903, "global_step": 174576, "epoch": 1961, "lr": 2.8360747855686554e-05}
{"train_loss": 0.08249366283416748, "global_step": 174577, "epoch": 1961, "lr": 2.836022522624701e-05}
{"train_loss": 0.10449092835187912, "global_step": 174578, "epoch": 1961, "lr": 2.8359702599716655e-05}
{"train_loss": 0.07295770198106766, "global_step": 174579, "epoch": 1961, "lr": 2.8359179976095585e-05}
{"train_loss": 0.08948998898267746, "global_step": 174580, "epoch": 1961, "lr": 2.8358657355383843e-05}
{"train_loss": 0.07069768011569977, "global_step": 174581, "epoch": 1961, "lr": 2.8358134737581544e-05}
{"train_loss": 0.09142998605966568, "global_step": 174582, "epoch": 1961, "lr": 2.8357612122688705e-05}
{"train_loss": 0.06674589216709137, "global_step": 174583, "epoch": 1961, "lr": 2.8357089510705413e-05}
{"train_loss": 0.07449405640363693, "global_step": 174584, "epoch": 1961, "lr": 2.8356566901631775e-05}
{"train_loss": 0.11626564711332321, "global_step": 174585, "epoch": 1961, "lr": 2.835604429546781e-05}
{"train_loss": 0.06616023927927017, "global_step": 174586, "epoch": 1961, "lr": 2.835552169221363e-05}
{"train_loss": 0.08895039558410645, "global_step": 174587, "epoch": 1961, "lr": 2.835499909186927e-05}
{"train_loss": 0.09071852266788483, "global_step": 174588, "epoch": 1961, "lr": 2.8354476494434834e-05}
{"train_loss": 0.04985504969954491, "global_step": 174589, "epoch": 1961, "lr": 2.8353953899910356e-05}
{"train_loss": 0.09528586268424988, "global_step": 174590, "epoch": 1961, "lr": 2.835343130829594e-05}
{"train_loss": 0.09210202842950821, "global_step": 174591, "epoch": 1961, "lr": 2.8352908719591632e-05}
{"train_loss": 0.06691782176494598, "global_step": 174592, "epoch": 1961, "lr": 2.8352386133797526e-05}
{"train_loss": 0.10146290808916092, "global_step": 174593, "epoch": 1961, "lr": 2.8351863550913655e-05}
{"train_loss": 0.03913957625627518, "global_step": 174594, "epoch": 1961, "lr": 2.8351340970940132e-05}
{"train_loss": 0.05065547674894333, "global_step": 174595, "epoch": 1961, "lr": 2.8350818393876987e-05}
{"train_loss": 0.05408554896712303, "global_step": 174596, "epoch": 1961, "lr": 2.835029581972434e-05}
{"train_loss": 0.04033856838941574, "global_step": 174597, "epoch": 1961, "lr": 2.8349773248482204e-05}
{"train_loss": 0.07634567469358444, "global_step": 174598, "epoch": 1961, "lr": 2.8349250680150695e-05}
{"train_loss": 0.08281975239515305, "global_step": 174599, "epoch": 1961, "lr": 2.834872811472985e-05}
{"train_loss": 0.08586149662733078, "global_step": 174600, "epoch": 1961, "lr": 2.8348205552219777e-05}
{"train_loss": 0.06347423046827316, "global_step": 174601, "epoch": 1961, "lr": 2.8347682992620493e-05}
{"train_loss": 0.0931391641497612, "global_step": 174602, "epoch": 1961, "lr": 2.834716043593213e-05}
{"train_loss": 0.14902350306510925, "global_step": 174603, "epoch": 1961, "lr": 2.8346637882154726e-05}
{"train_loss": 0.07351108640432358, "global_step": 174604, "epoch": 1961, "lr": 2.8346115331288325e-05}
{"train_loss": 0.09489518404006958, "global_step": 174605, "epoch": 1961, "lr": 2.8345592783333054e-05}
{"train_loss": 0.11579769104719162, "global_step": 174606, "epoch": 1961, "lr": 2.834507023828893e-05}
{"train_loss": 0.05935521051287651, "global_step": 174607, "epoch": 1961, "lr": 2.8344547696156065e-05}
{"train_loss": 0.06308615207672119, "global_step": 174608, "epoch": 1961, "lr": 2.8344025156934494e-05}
{"train_loss": 0.06366956233978271, "global_step": 174609, "epoch": 1961, "lr": 2.834350262062433e-05}
{"train_loss": 0.08871570974588394, "global_step": 174610, "epoch": 1961, "lr": 2.834298008722559e-05}
{"train_loss": 0.09680179506540298, "global_step": 174611, "epoch": 1961, "lr": 2.8342457556738396e-05}
{"train_loss": 0.058647193014621735, "global_step": 174612, "epoch": 1961, "lr": 2.8341935029162768e-05}
{"train_loss": 0.10728970170021057, "global_step": 174613, "epoch": 1961, "lr": 2.8341412504498833e-05}
{"train_loss": 0.08563017845153809, "global_step": 174614, "epoch": 1961, "lr": 2.834088998274661e-05}
{"train_loss": 0.10737506300210953, "global_step": 174615, "epoch": 1961, "lr": 2.8340367463906203e-05}
{"train_loss": 0.07579929381608963, "global_step": 174616, "epoch": 1961, "lr": 2.8339844947977655e-05}
{"train_loss": 0.08258870888626978, "global_step": 174617, "epoch": 1961, "lr": 2.833932243496107e-05, "val_loss": 7.434876918792725}
{"train_loss": 0.14985699951648712, "global_step": 174618, "epoch": 1962, "lr": 2.833879992485648e-05}
{"train_loss": 0.0538046658039093, "global_step": 174619, "epoch": 1962, "lr": 2.833827741766397e-05}
{"train_loss": 0.05964578315615654, "global_step": 174620, "epoch": 1962, "lr": 2.833775491338364e-05}
{"train_loss": 0.0873996764421463, "global_step": 174621, "epoch": 1962, "lr": 2.8337232412015512e-05}
{"train_loss": 0.09727922081947327, "global_step": 174622, "epoch": 1962, "lr": 2.83367099135597e-05}
{"train_loss": 0.03376610949635506, "global_step": 174623, "epoch": 1962, "lr": 2.8336187418016234e-05}
{"train_loss": 0.08783288300037384, "global_step": 174624, "epoch": 1962, "lr": 2.8335664925385218e-05}
{"train_loss": 0.06063530594110489, "global_step": 174625, "epoch": 1962, "lr": 2.8335142435666694e-05}
{"train_loss": 0.12188193947076797, "global_step": 174626, "epoch": 1962, "lr": 2.8334619948860765e-05}
{"train_loss": 0.06902176886796951, "global_step": 174627, "epoch": 1962, "lr": 2.8334097464967463e-05}
{"train_loss": 0.08406780660152435, "global_step": 174628, "epoch": 1962, "lr": 2.83335749839869e-05}
{"train_loss": 0.09494705498218536, "global_step": 174629, "epoch": 1962, "lr": 2.83330525059191e-05}
{"train_loss": 0.10458774119615555, "global_step": 174630, "epoch": 1962, "lr": 2.833253003076417e-05}
{"train_loss": 0.06094102934002876, "global_step": 174631, "epoch": 1962, "lr": 2.8332007558522156e-05}
{"train_loss": 0.06659895181655884, "global_step": 174632, "epoch": 1962, "lr": 2.833148508919316e-05}
{"train_loss": 0.1019003838300705, "global_step": 174633, "epoch": 1962, "lr": 2.8330962622777213e-05}
{"train_loss": 0.09482645243406296, "global_step": 174634, "epoch": 1962, "lr": 2.8330440159274417e-05}
{"train_loss": 0.10072548687458038, "global_step": 174635, "epoch": 1962, "lr": 2.8329917698684814e-05}
{"train_loss": 0.09408405423164368, "global_step": 174636, "epoch": 1962, "lr": 2.8329395241008504e-05}
{"train_loss": 0.08189284056425095, "global_step": 174637, "epoch": 1962, "lr": 2.8328872786245532e-05}
{"train_loss": 0.08871413767337799, "global_step": 174638, "epoch": 1962, "lr": 2.832835033439599e-05}
{"train_loss": 0.08339481800794601, "global_step": 174639, "epoch": 1962, "lr": 2.8327827885459922e-05}
{"train_loss": 0.11437136679887772, "global_step": 174640, "epoch": 1962, "lr": 2.8327305439437436e-05}
{"train_loss": 0.05007031932473183, "global_step": 174641, "epoch": 1962, "lr": 2.832678299632855e-05}
{"train_loss": 0.02740740403532982, "global_step": 174642, "epoch": 1962, "lr": 2.8326260556133387e-05}
{"train_loss": 0.08669516444206238, "global_step": 174643, "epoch": 1962, "lr": 2.8325738118852003e-05}
{"train_loss": 0.05394269526004791, "global_step": 174644, "epoch": 1962, "lr": 2.832521568448443e-05}
{"train_loss": 0.07189049571752548, "global_step": 174645, "epoch": 1962, "lr": 2.8324693253030792e-05}
{"train_loss": 0.15471100807189941, "global_step": 174646, "epoch": 1962, "lr": 2.8324170824491114e-05}
{"train_loss": 0.14925719797611237, "global_step": 174647, "epoch": 1962, "lr": 2.832364839886551e-05}
{"train_loss": 0.08216053247451782, "global_step": 174648, "epoch": 1962, "lr": 2.8323125976153997e-05}
{"train_loss": 0.07573415338993073, "global_step": 174649, "epoch": 1962, "lr": 2.832260355635671e-05}
{"train_loss": 0.15038089454174042, "global_step": 174650, "epoch": 1962, "lr": 2.8322081139473654e-05}
{"train_loss": 0.07705671340227127, "global_step": 174651, "epoch": 1962, "lr": 2.8321558725504948e-05}
{"train_loss": 0.11221341788768768, "global_step": 174652, "epoch": 1962, "lr": 2.832103631445063e-05}
{"train_loss": 0.0474100336432457, "global_step": 174653, "epoch": 1962, "lr": 2.832051390631081e-05}
{"train_loss": 0.07146085053682327, "global_step": 174654, "epoch": 1962, "lr": 2.8319991501085503e-05}
{"train_loss": 0.058386046439409256, "global_step": 174655, "epoch": 1962, "lr": 2.831946909877483e-05}
{"train_loss": 0.11570277065038681, "global_step": 174656, "epoch": 1962, "lr": 2.8318946699378825e-05}
{"train_loss": 0.054956864565610886, "global_step": 174657, "epoch": 1962, "lr": 2.831842430289757e-05}
{"train_loss": 0.06338687241077423, "global_step": 174658, "epoch": 1962, "lr": 2.8317901909331158e-05}
{"train_loss": 0.11135756224393845, "global_step": 174659, "epoch": 1962, "lr": 2.8317379518679623e-05}
{"train_loss": 0.104610875248909, "global_step": 174660, "epoch": 1962, "lr": 2.831685713094307e-05}
{"train_loss": 0.11315654963254929, "global_step": 174661, "epoch": 1962, "lr": 2.831633474612153e-05}
{"train_loss": 0.04308392480015755, "global_step": 174662, "epoch": 1962, "lr": 2.831581236421512e-05}
{"train_loss": 0.12971746921539307, "global_step": 174663, "epoch": 1962, "lr": 2.8315289985223858e-05}
{"train_loss": 0.07185331732034683, "global_step": 174664, "epoch": 1962, "lr": 2.8314767609147862e-05}
{"train_loss": 0.08007821440696716, "global_step": 174665, "epoch": 1962, "lr": 2.8314245235987168e-05}
{"train_loss": 0.1097947508096695, "global_step": 174666, "epoch": 1962, "lr": 2.8313722865741875e-05}
{"train_loss": 0.06160255894064903, "global_step": 174667, "epoch": 1962, "lr": 2.8313200498412016e-05}
{"train_loss": 0.0521586574614048, "global_step": 174668, "epoch": 1962, "lr": 2.8312678133997706e-05}
{"train_loss": 0.07463418692350388, "global_step": 174669, "epoch": 1962, "lr": 2.8312155772498967e-05}
{"train_loss": 0.051944345235824585, "global_step": 174670, "epoch": 1962, "lr": 2.8311633413915916e-05}
{"train_loss": 0.05340654402971268, "global_step": 174671, "epoch": 1962, "lr": 2.8311111058248586e-05}
{"train_loss": 0.053886160254478455, "global_step": 174672, "epoch": 1962, "lr": 2.8310588705497076e-05}
{"train_loss": 0.07375123351812363, "global_step": 174673, "epoch": 1962, "lr": 2.8310066355661425e-05}
{"train_loss": 0.07931109517812729, "global_step": 174674, "epoch": 1962, "lr": 2.830954400874173e-05}
{"train_loss": 0.06683450192213058, "global_step": 174675, "epoch": 1962, "lr": 2.830902166473806e-05}
{"train_loss": 0.07977431267499924, "global_step": 174676, "epoch": 1962, "lr": 2.8308499323650485e-05}
{"train_loss": 0.08852187544107437, "global_step": 174677, "epoch": 1962, "lr": 2.8307976985479044e-05}
{"train_loss": 0.11034554243087769, "global_step": 174678, "epoch": 1962, "lr": 2.8307454650223853e-05}
{"train_loss": 0.06789849698543549, "global_step": 174679, "epoch": 1962, "lr": 2.830693231788494e-05}
{"train_loss": 0.0581696480512619, "global_step": 174680, "epoch": 1962, "lr": 2.8306409988462413e-05}
{"train_loss": 0.10065174847841263, "global_step": 174681, "epoch": 1962, "lr": 2.830588766195633e-05}
{"train_loss": 0.07775458693504333, "global_step": 174682, "epoch": 1962, "lr": 2.8305365338366734e-05}
{"train_loss": 0.05115871503949165, "global_step": 174683, "epoch": 1962, "lr": 2.8304843017693737e-05}
{"train_loss": 0.12382492423057556, "global_step": 174684, "epoch": 1962, "lr": 2.830432069993737e-05}
{"train_loss": 0.13721692562103271, "global_step": 174685, "epoch": 1962, "lr": 2.8303798385097742e-05}
{"train_loss": 0.10123801231384277, "global_step": 174686, "epoch": 1962, "lr": 2.8303276073174883e-05}
{"train_loss": 0.09054851531982422, "global_step": 174687, "epoch": 1962, "lr": 2.8302753764168904e-05}
{"train_loss": 0.04976081848144531, "global_step": 174688, "epoch": 1962, "lr": 2.8302231458079836e-05}
{"train_loss": 0.15362998843193054, "global_step": 174689, "epoch": 1962, "lr": 2.8301709154907786e-05}
{"train_loss": 0.04210680350661278, "global_step": 174690, "epoch": 1962, "lr": 2.8301186854652796e-05}
{"train_loss": 0.060496583580970764, "global_step": 174691, "epoch": 1962, "lr": 2.830066455731496e-05}
{"train_loss": 0.06796417385339737, "global_step": 174692, "epoch": 1962, "lr": 2.8300142262894312e-05}
{"train_loss": 0.03009633719921112, "global_step": 174693, "epoch": 1962, "lr": 2.8299619971390944e-05}
{"train_loss": 0.08449899405241013, "global_step": 174694, "epoch": 1962, "lr": 2.8299097682804955e-05}
{"train_loss": 0.12378048896789551, "global_step": 174695, "epoch": 1962, "lr": 2.8298575397136366e-05}
{"train_loss": 0.08403636515140533, "global_step": 174696, "epoch": 1962, "lr": 2.8298053114385292e-05}
{"train_loss": 0.050098344683647156, "global_step": 174697, "epoch": 1962, "lr": 2.8297530834551754e-05}
{"train_loss": 0.0853501409292221, "global_step": 174698, "epoch": 1962, "lr": 2.8297008557635873e-05}
{"train_loss": 0.08104394376277924, "global_step": 174699, "epoch": 1962, "lr": 2.8296486283637667e-05}
{"train_loss": 0.05900948867201805, "global_step": 174700, "epoch": 1962, "lr": 2.8295964012557263e-05}
{"train_loss": 0.06285269558429718, "global_step": 174701, "epoch": 1962, "lr": 2.829544174439468e-05}
{"train_loss": 0.11723434925079346, "global_step": 174702, "epoch": 1962, "lr": 2.829491947915003e-05}
{"train_loss": 0.06684209406375885, "global_step": 174703, "epoch": 1962, "lr": 2.8294397216823345e-05}
{"train_loss": 0.09116262942552567, "global_step": 174704, "epoch": 1962, "lr": 2.8293874957414723e-05}
{"train_loss": 0.08895751088857651, "global_step": 174705, "epoch": 1962, "lr": 2.8293352700924214e-05}
{"train_loss": 0.08292442941096391, "global_step": 174706, "epoch": 1962, "lr": 2.829283044735192e-05, "val_loss": 7.482929229736328}
{"train_loss": 0.06374786049127579, "global_step": 174707, "epoch": 1963, "lr": 2.829230819669787e-05}
{"train_loss": 0.14465948939323425, "global_step": 174708, "epoch": 1963, "lr": 2.8291785948962173e-05}
{"train_loss": 0.05569717660546303, "global_step": 174709, "epoch": 1963, "lr": 2.829126370414486e-05}
{"train_loss": 0.10095228999853134, "global_step": 174710, "epoch": 1963, "lr": 2.8290741462246027e-05}
{"train_loss": 0.08763982355594635, "global_step": 174711, "epoch": 1963, "lr": 2.8290219223265756e-05}
{"train_loss": 0.07377370446920395, "global_step": 174712, "epoch": 1963, "lr": 2.8289696987204073e-05}
{"train_loss": 0.060318153351545334, "global_step": 174713, "epoch": 1963, "lr": 2.828917475406111e-05}
{"train_loss": 0.06661349534988403, "global_step": 174714, "epoch": 1963, "lr": 2.828865252383689e-05}
{"train_loss": 0.06471775472164154, "global_step": 174715, "epoch": 1963, "lr": 2.828813029653148e-05}
{"train_loss": 0.09716653823852539, "global_step": 174716, "epoch": 1963, "lr": 2.828760807214499e-05}
{"train_loss": 0.046566206961870193, "global_step": 174717, "epoch": 1963, "lr": 2.8287085850677443e-05}
{"train_loss": 0.06118268147110939, "global_step": 174718, "epoch": 1963, "lr": 2.8286563632128955e-05}
{"train_loss": 0.10667043924331665, "global_step": 174719, "epoch": 1963, "lr": 2.8286041416499577e-05}
{"train_loss": 0.06815219670534134, "global_step": 174720, "epoch": 1963, "lr": 2.8285519203789345e-05}
{"train_loss": 0.08925670385360718, "global_step": 174721, "epoch": 1963, "lr": 2.828499699399839e-05}
{"train_loss": 0.08496777713298798, "global_step": 174722, "epoch": 1963, "lr": 2.8284474787126724e-05}
{"train_loss": 0.08566374331712723, "global_step": 174723, "epoch": 1963, "lr": 2.8283952583174477e-05}
{"train_loss": 0.0759478211402893, "global_step": 174724, "epoch": 1963, "lr": 2.828343038214166e-05}
{"train_loss": 0.07669194787740707, "global_step": 174725, "epoch": 1963, "lr": 2.828290818402839e-05}
{"train_loss": 0.14036738872528076, "global_step": 174726, "epoch": 1963, "lr": 2.8282385988834704e-05}
{"train_loss": 0.11462617665529251, "global_step": 174727, "epoch": 1963, "lr": 2.8281863796560703e-05}
{"train_loss": 0.0896870344877243, "global_step": 174728, "epoch": 1963, "lr": 2.8281341607206426e-05}
{"train_loss": 0.1288282722234726, "global_step": 174729, "epoch": 1963, "lr": 2.8280819420771952e-05}
{"train_loss": 0.0861121267080307, "global_step": 174730, "epoch": 1963, "lr": 2.828029723725738e-05}
{"train_loss": 0.0944683700799942, "global_step": 174731, "epoch": 1963, "lr": 2.827977505666273e-05}
{"train_loss": 0.05130709335207939, "global_step": 174732, "epoch": 1963, "lr": 2.8279252878988127e-05}
{"train_loss": 0.14753606915473938, "global_step": 174733, "epoch": 1963, "lr": 2.827873070423359e-05}
{"train_loss": 0.05106032267212868, "global_step": 174734, "epoch": 1963, "lr": 2.827820853239924e-05}
{"train_loss": 0.09952902048826218, "global_step": 174735, "epoch": 1963, "lr": 2.827768636348509e-05}
{"train_loss": 0.055493682622909546, "global_step": 174736, "epoch": 1963, "lr": 2.8277164197491267e-05}
{"train_loss": 0.08381199836730957, "global_step": 174737, "epoch": 1963, "lr": 2.8276642034417792e-05}
{"train_loss": 0.05320091173052788, "global_step": 174738, "epoch": 1963, "lr": 2.8276119874264783e-05}
{"train_loss": 0.05974402278661728, "global_step": 174739, "epoch": 1963, "lr": 2.827559771703226e-05}
{"train_loss": 0.05954572185873985, "global_step": 174740, "epoch": 1963, "lr": 2.827507556272034e-05}
{"train_loss": 0.07657146453857422, "global_step": 174741, "epoch": 1963, "lr": 2.8274553411329045e-05}
{"train_loss": 0.09737154096364975, "global_step": 174742, "epoch": 1963, "lr": 2.8274031262858503e-05}
{"train_loss": 0.07450935244560242, "global_step": 174743, "epoch": 1963, "lr": 2.827350911730872e-05}
{"train_loss": 0.162784144282341, "global_step": 174744, "epoch": 1963, "lr": 2.8272986974679827e-05}
{"train_loss": 0.0809461921453476, "global_step": 174745, "epoch": 1963, "lr": 2.8272464834971844e-05}
{"train_loss": 0.11271070688962936, "global_step": 174746, "epoch": 1963, "lr": 2.827194269818486e-05}
{"train_loss": 0.08325468748807907, "global_step": 174747, "epoch": 1963, "lr": 2.8271420564318974e-05}
{"train_loss": 0.03985058516263962, "global_step": 174748, "epoch": 1963, "lr": 2.827089843337421e-05}
{"train_loss": 0.06287036836147308, "global_step": 174749, "epoch": 1963, "lr": 2.8270376305350675e-05}
{"train_loss": 0.0838954970240593, "global_step": 174750, "epoch": 1963, "lr": 2.826985418024841e-05}
{"train_loss": 0.077787384390831, "global_step": 174751, "epoch": 1963, "lr": 2.826933205806751e-05}
{"train_loss": 0.08708532899618149, "global_step": 174752, "epoch": 1963, "lr": 2.826880993880804e-05}
{"train_loss": 0.1386953443288803, "global_step": 174753, "epoch": 1963, "lr": 2.826828782247004e-05}
{"train_loss": 0.09960055351257324, "global_step": 174754, "epoch": 1963, "lr": 2.8267765709053623e-05}
{"train_loss": 0.10280165821313858, "global_step": 174755, "epoch": 1963, "lr": 2.8267243598558824e-05}
{"train_loss": 0.06432851403951645, "global_step": 174756, "epoch": 1963, "lr": 2.8266721490985747e-05}
{"train_loss": 0.07567014545202255, "global_step": 174757, "epoch": 1963, "lr": 2.8266199386334425e-05}
{"train_loss": 0.07963055372238159, "global_step": 174758, "epoch": 1963, "lr": 2.8265677284604964e-05}
{"train_loss": 0.049826040863990784, "global_step": 174759, "epoch": 1963, "lr": 2.826515518579742e-05}
{"train_loss": 0.09910117834806442, "global_step": 174760, "epoch": 1963, "lr": 2.826463308991184e-05}
{"train_loss": 0.10640872269868851, "global_step": 174761, "epoch": 1963, "lr": 2.826411099694834e-05}
{"train_loss": 0.04597671702504158, "global_step": 174762, "epoch": 1963, "lr": 2.8263588906906947e-05}
{"train_loss": 0.1262664645910263, "global_step": 174763, "epoch": 1963, "lr": 2.826306681978776e-05}
{"train_loss": 0.07250300794839859, "global_step": 174764, "epoch": 1963, "lr": 2.8262544735590824e-05}
{"train_loss": 0.09163624048233032, "global_step": 174765, "epoch": 1963, "lr": 2.8262022654316223e-05}
{"train_loss": 0.06951753795146942, "global_step": 174766, "epoch": 1963, "lr": 2.826150057596405e-05}
{"train_loss": 0.04773930460214615, "global_step": 174767, "epoch": 1963, "lr": 2.826097850053433e-05}
{"train_loss": 0.05182580277323723, "global_step": 174768, "epoch": 1963, "lr": 2.8260456428027183e-05}
{"train_loss": 0.12038589268922806, "global_step": 174769, "epoch": 1963, "lr": 2.825993435844262e-05}
{"train_loss": 0.04032917693257332, "global_step": 174770, "epoch": 1963, "lr": 2.8259412291780772e-05}
{"train_loss": 0.08587461709976196, "global_step": 174771, "epoch": 1963, "lr": 2.8258890228041663e-05}
{"train_loss": 0.09603802114725113, "global_step": 174772, "epoch": 1963, "lr": 2.82583681672254e-05}
{"train_loss": 0.10395140200853348, "global_step": 174773, "epoch": 1963, "lr": 2.825784610933201e-05}
{"train_loss": 0.07466500997543335, "global_step": 174774, "epoch": 1963, "lr": 2.8257324054361613e-05}
{"train_loss": 0.06362437456846237, "global_step": 174775, "epoch": 1963, "lr": 2.8256802002314232e-05}
{"train_loss": 0.03804480656981468, "global_step": 174776, "epoch": 1963, "lr": 2.825627995318998e-05}
{"train_loss": 0.07992115616798401, "global_step": 174777, "epoch": 1963, "lr": 2.8255757906988884e-05}
{"train_loss": 0.04722418636083603, "global_step": 174778, "epoch": 1963, "lr": 2.825523586371106e-05}
{"train_loss": 0.09070669859647751, "global_step": 174779, "epoch": 1963, "lr": 2.8254713823356528e-05}
{"train_loss": 0.1004888266324997, "global_step": 174780, "epoch": 1963, "lr": 2.825419178592541e-05}
{"train_loss": 0.08035764843225479, "global_step": 174781, "epoch": 1963, "lr": 2.825366975141773e-05}
{"train_loss": 0.08020047843456268, "global_step": 174782, "epoch": 1963, "lr": 2.8253147719833594e-05}
{"train_loss": 0.09851745516061783, "global_step": 174783, "epoch": 1963, "lr": 2.8252625691173047e-05}
{"train_loss": 0.06160043179988861, "global_step": 174784, "epoch": 1963, "lr": 2.825210366543616e-05}
{"train_loss": 0.06366098672151566, "global_step": 174785, "epoch": 1963, "lr": 2.8251581642623037e-05}
{"train_loss": 0.14875829219818115, "global_step": 174786, "epoch": 1963, "lr": 2.8251059622733706e-05}
{"train_loss": 0.08386654406785965, "global_step": 174787, "epoch": 1963, "lr": 2.8250537605768268e-05}
{"train_loss": 0.10845311731100082, "global_step": 174788, "epoch": 1963, "lr": 2.8250015591726764e-05}
{"train_loss": 0.09598291665315628, "global_step": 174789, "epoch": 1963, "lr": 2.8249493580609303e-05}
{"train_loss": 0.06482300907373428, "global_step": 174790, "epoch": 1963, "lr": 2.824897157241591e-05}
{"train_loss": 0.11457941681146622, "global_step": 174791, "epoch": 1963, "lr": 2.8248449567146695e-05}
{"train_loss": 0.0502738393843174, "global_step": 174792, "epoch": 1963, "lr": 2.824792756480172e-05}
{"train_loss": 0.07818193733692169, "global_step": 174793, "epoch": 1963, "lr": 2.8247405565381015e-05}
{"train_loss": 0.06838064640760422, "global_step": 174794, "epoch": 1963, "lr": 2.8246883568884708e-05}
{"train_loss": 0.08268913835956809, "global_step": 174795, "epoch": 1963, "lr": 2.824636157531282e-05, "val_loss": 7.2002272605896}
{"train_loss": 0.09422507137060165, "global_step": 174796, "epoch": 1964, "lr": 2.8245839584665467e-05}
{"train_loss": 0.12881554663181305, "global_step": 174797, "epoch": 1964, "lr": 2.8245317596942687e-05}
{"train_loss": 0.08839942514896393, "global_step": 174798, "epoch": 1964, "lr": 2.824479561214455e-05}
{"train_loss": 0.09865403175354004, "global_step": 174799, "epoch": 1964, "lr": 2.8244273630271144e-05}
{"train_loss": 0.0812106728553772, "global_step": 174800, "epoch": 1964, "lr": 2.824375165132252e-05}
{"train_loss": 0.0985722541809082, "global_step": 174801, "epoch": 1964, "lr": 2.8243229675298756e-05}
{"train_loss": 0.07954732328653336, "global_step": 174802, "epoch": 1964, "lr": 2.8242707702199938e-05}
{"train_loss": 0.09707680344581604, "global_step": 174803, "epoch": 1964, "lr": 2.8242185732026105e-05}
{"train_loss": 0.09190086275339127, "global_step": 174804, "epoch": 1964, "lr": 2.824166376477737e-05}
{"train_loss": 0.11338499188423157, "global_step": 174805, "epoch": 1964, "lr": 2.8241141800453757e-05}
{"train_loss": 0.07150772958993912, "global_step": 174806, "epoch": 1964, "lr": 2.8240619839055378e-05}
{"train_loss": 0.06550969183444977, "global_step": 174807, "epoch": 1964, "lr": 2.824009788058226e-05}
{"train_loss": 0.09203804284334183, "global_step": 174808, "epoch": 1964, "lr": 2.8239575925034512e-05}
{"train_loss": 0.11881197988986969, "global_step": 174809, "epoch": 1964, "lr": 2.8239053972412177e-05}
{"train_loss": 0.08235812187194824, "global_step": 174810, "epoch": 1964, "lr": 2.823853202271535e-05}
{"train_loss": 0.10425496846437454, "global_step": 174811, "epoch": 1964, "lr": 2.8238010075944066e-05}
{"train_loss": 0.09167993068695068, "global_step": 174812, "epoch": 1964, "lr": 2.823748813209844e-05}
{"train_loss": 0.046114224940538406, "global_step": 174813, "epoch": 1964, "lr": 2.8236966191178492e-05}
{"train_loss": 0.053579144179821014, "global_step": 174814, "epoch": 1964, "lr": 2.8236444253184348e-05}
{"train_loss": 0.08906305581331253, "global_step": 174815, "epoch": 1964, "lr": 2.823592231811602e-05}
{"train_loss": 0.0774705782532692, "global_step": 174816, "epoch": 1964, "lr": 2.8235400385973633e-05}
{"train_loss": 0.11188653856515884, "global_step": 174817, "epoch": 1964, "lr": 2.8234878456757207e-05}
{"train_loss": 0.0708077922463417, "global_step": 174818, "epoch": 1964, "lr": 2.823435653046686e-05}
{"train_loss": 0.06724999845027924, "global_step": 174819, "epoch": 1964, "lr": 2.8233834607102615e-05}
{"train_loss": 0.07591741532087326, "global_step": 174820, "epoch": 1964, "lr": 2.823331268666457e-05}
{"train_loss": 0.07974230498075485, "global_step": 174821, "epoch": 1964, "lr": 2.8232790769152806e-05}
{"train_loss": 0.11163005232810974, "global_step": 174822, "epoch": 1964, "lr": 2.8232268854567363e-05}
{"train_loss": 0.05281088501214981, "global_step": 174823, "epoch": 1964, "lr": 2.8231746942908344e-05}
{"train_loss": 0.11475074291229248, "global_step": 174824, "epoch": 1964, "lr": 2.823122503417578e-05}
{"train_loss": 0.04054657369852066, "global_step": 174825, "epoch": 1964, "lr": 2.8230703128369783e-05}
{"train_loss": 0.080861896276474, "global_step": 174826, "epoch": 1964, "lr": 2.8230181225490383e-05}
{"train_loss": 0.09979482740163803, "global_step": 174827, "epoch": 1964, "lr": 2.8229659325537696e-05}
{"train_loss": 0.08705585449934006, "global_step": 174828, "epoch": 1964, "lr": 2.822913742851173e-05}
{"train_loss": 0.05914849415421486, "global_step": 174829, "epoch": 1964, "lr": 2.822861553441263e-05}
{"train_loss": 0.07900812476873398, "global_step": 174830, "epoch": 1964, "lr": 2.8228093643240418e-05}
{"train_loss": 0.09235701709985733, "global_step": 174831, "epoch": 1964, "lr": 2.8227571754995148e-05}
{"train_loss": 0.030841456726193428, "global_step": 174832, "epoch": 1964, "lr": 2.822704986967694e-05}
{"train_loss": 0.06969504803419113, "global_step": 174833, "epoch": 1964, "lr": 2.8226527987285823e-05}
{"train_loss": 0.055913813412189484, "global_step": 174834, "epoch": 1964, "lr": 2.82260061078219e-05}
{"train_loss": 0.0771993026137352, "global_step": 174835, "epoch": 1964, "lr": 2.8225484231285225e-05}
{"train_loss": 0.12587608397006989, "global_step": 174836, "epoch": 1964, "lr": 2.8224962357675855e-05}
{"train_loss": 0.07938830554485321, "global_step": 174837, "epoch": 1964, "lr": 2.8224440486993864e-05}
{"train_loss": 0.07141973823308945, "global_step": 174838, "epoch": 1964, "lr": 2.822391861923936e-05}
{"train_loss": 0.09569936990737915, "global_step": 174839, "epoch": 1964, "lr": 2.8223396754412358e-05}
{"train_loss": 0.10183557868003845, "global_step": 174840, "epoch": 1964, "lr": 2.8222874892512978e-05}
{"train_loss": 0.14807839691638947, "global_step": 174841, "epoch": 1964, "lr": 2.822235303354125e-05}
{"train_loss": 0.08455301076173782, "global_step": 174842, "epoch": 1964, "lr": 2.8221831177497275e-05}
{"train_loss": 0.08290378004312515, "global_step": 174843, "epoch": 1964, "lr": 2.8221309324381094e-05}
{"train_loss": 0.07758845388889313, "global_step": 174844, "epoch": 1964, "lr": 2.8220787474192812e-05}
{"train_loss": 0.07151083648204803, "global_step": 174845, "epoch": 1964, "lr": 2.8220265626932458e-05}
{"train_loss": 0.10207835584878922, "global_step": 174846, "epoch": 1964, "lr": 2.8219743782600145e-05}
{"train_loss": 0.03833063691854477, "global_step": 174847, "epoch": 1964, "lr": 2.82192219411959e-05}
{"train_loss": 0.12979058921337128, "global_step": 174848, "epoch": 1964, "lr": 2.8218700102719836e-05}
{"train_loss": 0.07643262296915054, "global_step": 174849, "epoch": 1964, "lr": 2.8218178267171986e-05}
{"train_loss": 0.07946551591157913, "global_step": 174850, "epoch": 1964, "lr": 2.8217656434552455e-05}
{"train_loss": 0.06088532507419586, "global_step": 174851, "epoch": 1964, "lr": 2.8217134604861272e-05}
{"train_loss": 0.07107919454574585, "global_step": 174852, "epoch": 1964, "lr": 2.821661277809855e-05}
{"train_loss": 0.07621787488460541, "global_step": 174853, "epoch": 1964, "lr": 2.8216090954264318e-05}
{"train_loss": 0.057819120585918427, "global_step": 174854, "epoch": 1964, "lr": 2.8215569133358687e-05}
{"train_loss": 0.08111260086297989, "global_step": 174855, "epoch": 1964, "lr": 2.821504731538169e-05}
{"train_loss": 0.08012722432613373, "global_step": 174856, "epoch": 1964, "lr": 2.821452550033341e-05}
{"train_loss": 0.05266764014959335, "global_step": 174857, "epoch": 1964, "lr": 2.8214003688213947e-05}
{"train_loss": 0.05964281037449837, "global_step": 174858, "epoch": 1964, "lr": 2.8213481879023323e-05}
{"train_loss": 0.06992179900407791, "global_step": 174859, "epoch": 1964, "lr": 2.821296007276165e-05}
{"train_loss": 0.09472057968378067, "global_step": 174860, "epoch": 1964, "lr": 2.8212438269428965e-05}
{"train_loss": 0.06134383752942085, "global_step": 174861, "epoch": 1964, "lr": 2.821191646902537e-05}
{"train_loss": 0.0660848543047905, "global_step": 174862, "epoch": 1964, "lr": 2.8211394671550888e-05}
{"train_loss": 0.06354907900094986, "global_step": 174863, "epoch": 1964, "lr": 2.8210872877005646e-05}
{"train_loss": 0.04734731465578079, "global_step": 174864, "epoch": 1964, "lr": 2.8210351085389662e-05}
{"train_loss": 0.060231395065784454, "global_step": 174865, "epoch": 1964, "lr": 2.820982929670306e-05}
{"train_loss": 0.09405405074357986, "global_step": 174866, "epoch": 1964, "lr": 2.8209307510945853e-05}
{"train_loss": 0.11496879905462265, "global_step": 174867, "epoch": 1964, "lr": 2.820878572811816e-05}
{"train_loss": 0.09234963357448578, "global_step": 174868, "epoch": 1964, "lr": 2.8208263948220016e-05}
{"train_loss": 0.06167355552315712, "global_step": 174869, "epoch": 1964, "lr": 2.8207742171251518e-05}
{"train_loss": 0.04748079925775528, "global_step": 174870, "epoch": 1964, "lr": 2.820722039721273e-05}
{"train_loss": 0.07103408128023148, "global_step": 174871, "epoch": 1964, "lr": 2.8206698626103688e-05}
{"train_loss": 0.08115952461957932, "global_step": 174872, "epoch": 1964, "lr": 2.8206176857924516e-05}
{"train_loss": 0.076459139585495, "global_step": 174873, "epoch": 1964, "lr": 2.820565509267523e-05}
{"train_loss": 0.13255895674228668, "global_step": 174874, "epoch": 1964, "lr": 2.820513333035596e-05}
{"train_loss": 0.11024079471826553, "global_step": 174875, "epoch": 1964, "lr": 2.820461157096671e-05}
{"train_loss": 0.13078728318214417, "global_step": 174876, "epoch": 1964, "lr": 2.8204089814507617e-05}
{"train_loss": 0.09956453740596771, "global_step": 174877, "epoch": 1964, "lr": 2.8203568060978693e-05}
{"train_loss": 0.05577770993113518, "global_step": 174878, "epoch": 1964, "lr": 2.8203046310380055e-05}
{"train_loss": 0.08719839155673981, "global_step": 174879, "epoch": 1964, "lr": 2.820252456271173e-05}
{"train_loss": 0.11202381551265717, "global_step": 174880, "epoch": 1964, "lr": 2.8202002817973826e-05}
{"train_loss": 0.041184909641742706, "global_step": 174881, "epoch": 1964, "lr": 2.8201481076166387e-05}
{"train_loss": 0.09505992382764816, "global_step": 174882, "epoch": 1964, "lr": 2.8200959337289507e-05}
{"train_loss": 0.06280405819416046, "global_step": 174883, "epoch": 1964, "lr": 2.820043760134322e-05}
{"train_loss": 0.08268920844944005, "global_step": 174884, "epoch": 1964, "lr": 2.8199915868327642e-05, "val_loss": 7.191686153411865}
{"train_loss": 0.07024438679218292, "global_step": 174885, "epoch": 1965, "lr": 2.81993941382428e-05}
{"train_loss": 0.06233561411499977, "global_step": 174886, "epoch": 1965, "lr": 2.8198872411088805e-05}
{"train_loss": 0.08157967776060104, "global_step": 174887, "epoch": 1965, "lr": 2.8198350686865682e-05}
{"train_loss": 0.06127622351050377, "global_step": 174888, "epoch": 1965, "lr": 2.8197828965573548e-05}
{"train_loss": 0.0690111368894577, "global_step": 174889, "epoch": 1965, "lr": 2.8197307247212428e-05}
{"train_loss": 0.11057556420564651, "global_step": 174890, "epoch": 1965, "lr": 2.8196785531782432e-05}
{"train_loss": 0.05574426054954529, "global_step": 174891, "epoch": 1965, "lr": 2.8196263819283598e-05}
{"train_loss": 0.07171284407377243, "global_step": 174892, "epoch": 1965, "lr": 2.819574210971601e-05}
{"train_loss": 0.12387512624263763, "global_step": 174893, "epoch": 1965, "lr": 2.819522040307976e-05}
{"train_loss": 0.07862665504217148, "global_step": 174894, "epoch": 1965, "lr": 2.819469869937487e-05}
{"train_loss": 0.060143548995256424, "global_step": 174895, "epoch": 1965, "lr": 2.8194176998601463e-05}
{"train_loss": 0.10866504907608032, "global_step": 174896, "epoch": 1965, "lr": 2.8193655300759557e-05}
{"train_loss": 0.0877620205283165, "global_step": 174897, "epoch": 1965, "lr": 2.8193133605849277e-05}
{"train_loss": 0.06970232725143433, "global_step": 174898, "epoch": 1965, "lr": 2.819261191387064e-05}
{"train_loss": 0.04079294204711914, "global_step": 174899, "epoch": 1965, "lr": 2.8192090224823757e-05}
{"train_loss": 0.07599093019962311, "global_step": 174900, "epoch": 1965, "lr": 2.8191568538708667e-05}
{"train_loss": 0.08738918602466583, "global_step": 174901, "epoch": 1965, "lr": 2.819104685552547e-05}
{"train_loss": 0.11173871159553528, "global_step": 174902, "epoch": 1965, "lr": 2.8190525175274206e-05}
{"train_loss": 0.1191396415233612, "global_step": 174903, "epoch": 1965, "lr": 2.8190003497954977e-05}
{"train_loss": 0.060030072927474976, "global_step": 174904, "epoch": 1965, "lr": 2.818948182356782e-05}
{"train_loss": 0.060078248381614685, "global_step": 174905, "epoch": 1965, "lr": 2.8188960152112838e-05}
{"train_loss": 0.0705040991306305, "global_step": 174906, "epoch": 1965, "lr": 2.8188438483590063e-05}
{"train_loss": 0.07547152787446976, "global_step": 174907, "epoch": 1965, "lr": 2.8187916817999604e-05}
{"train_loss": 0.058623235672712326, "global_step": 174908, "epoch": 1965, "lr": 2.8187395155341523e-05}
{"train_loss": 0.08668840676546097, "global_step": 174909, "epoch": 1965, "lr": 2.8186873495615856e-05}
{"train_loss": 0.0664362832903862, "global_step": 174910, "epoch": 1965, "lr": 2.8186351838822716e-05}
{"train_loss": 0.09114153683185577, "global_step": 174911, "epoch": 1965, "lr": 2.8185830184962135e-05}
{"train_loss": 0.08524255454540253, "global_step": 174912, "epoch": 1965, "lr": 2.8185308534034227e-05}
{"train_loss": 0.0566842295229435, "global_step": 174913, "epoch": 1965, "lr": 2.8184786886039018e-05}
{"train_loss": 0.06731456518173218, "global_step": 174914, "epoch": 1965, "lr": 2.8184265240976616e-05}
{"train_loss": 0.0668860599398613, "global_step": 174915, "epoch": 1965, "lr": 2.8183743598847056e-05}
{"train_loss": 0.05638647451996803, "global_step": 174916, "epoch": 1965, "lr": 2.8183221959650442e-05}
{"train_loss": 0.09520605206489563, "global_step": 174917, "epoch": 1965, "lr": 2.8182700323386812e-05}
{"train_loss": 0.11347471922636032, "global_step": 174918, "epoch": 1965, "lr": 2.818217869005627e-05}
{"train_loss": 0.0755709633231163, "global_step": 174919, "epoch": 1965, "lr": 2.8181657059658852e-05}
{"train_loss": 0.09592771530151367, "global_step": 174920, "epoch": 1965, "lr": 2.818113543219466e-05}
{"train_loss": 0.12058211117982864, "global_step": 174921, "epoch": 1965, "lr": 2.8180613807663725e-05}
{"train_loss": 0.09994836151599884, "global_step": 174922, "epoch": 1965, "lr": 2.8180092186066165e-05}
{"train_loss": 0.07502274960279465, "global_step": 174923, "epoch": 1965, "lr": 2.817957056740201e-05}
{"train_loss": 0.11368827521800995, "global_step": 174924, "epoch": 1965, "lr": 2.817904895167136e-05}
{"train_loss": 0.08063489943742752, "global_step": 174925, "epoch": 1965, "lr": 2.817852733887425e-05}
{"train_loss": 0.08437982201576233, "global_step": 174926, "epoch": 1965, "lr": 2.8178005729010792e-05}
{"train_loss": 0.0825268030166626, "global_step": 174927, "epoch": 1965, "lr": 2.817748412208102e-05}
{"train_loss": 0.11883559823036194, "global_step": 174928, "epoch": 1965, "lr": 2.8176962518085016e-05}
{"train_loss": 0.06808115541934967, "global_step": 174929, "epoch": 1965, "lr": 2.8176440917022873e-05}
{"train_loss": 0.07524022459983826, "global_step": 174930, "epoch": 1965, "lr": 2.817591931889462e-05}
{"train_loss": 0.1084882915019989, "global_step": 174931, "epoch": 1965, "lr": 2.817539772370037e-05}
{"train_loss": 0.11567551642656326, "global_step": 174932, "epoch": 1965, "lr": 2.8174876131440155e-05}
{"train_loss": 0.09125673025846481, "global_step": 174933, "epoch": 1965, "lr": 2.817435454211408e-05}
{"train_loss": 0.07767953723669052, "global_step": 174934, "epoch": 1965, "lr": 2.8173832955722174e-05}
{"train_loss": 0.0787339061498642, "global_step": 174935, "epoch": 1965, "lr": 2.8173311372264553e-05}
{"train_loss": 0.08734911680221558, "global_step": 174936, "epoch": 1965, "lr": 2.8172789791741244e-05}
{"train_loss": 0.08941727131605148, "global_step": 174937, "epoch": 1965, "lr": 2.8172268214152352e-05}
{"train_loss": 0.07815738767385483, "global_step": 174938, "epoch": 1965, "lr": 2.817174663949792e-05}
{"train_loss": 0.06864852458238602, "global_step": 174939, "epoch": 1965, "lr": 2.8171225067778052e-05}
{"train_loss": 0.06538838893175125, "global_step": 174940, "epoch": 1965, "lr": 2.8170703498992767e-05}
{"train_loss": 0.1000167727470398, "global_step": 174941, "epoch": 1965, "lr": 2.8170181933142193e-05}
{"train_loss": 0.10362333804368973, "global_step": 174942, "epoch": 1965, "lr": 2.8169660370226347e-05}
{"train_loss": 0.11230134963989258, "global_step": 174943, "epoch": 1965, "lr": 2.816913881024535e-05}
{"train_loss": 0.12442890554666519, "global_step": 174944, "epoch": 1965, "lr": 2.8168617253199225e-05}
{"train_loss": 0.05543280765414238, "global_step": 174945, "epoch": 1965, "lr": 2.816809569908808e-05}
{"train_loss": 0.09448438137769699, "global_step": 174946, "epoch": 1965, "lr": 2.816757414791197e-05}
{"train_loss": 0.06800635159015656, "global_step": 174947, "epoch": 1965, "lr": 2.816705259967094e-05}
{"train_loss": 0.06107291579246521, "global_step": 174948, "epoch": 1965, "lr": 2.816653105436511e-05}
{"train_loss": 0.09735874831676483, "global_step": 174949, "epoch": 1965, "lr": 2.8166009511994495e-05}
{"train_loss": 0.12659770250320435, "global_step": 174950, "epoch": 1965, "lr": 2.8165487972559223e-05}
{"train_loss": 0.0762140154838562, "global_step": 174951, "epoch": 1965, "lr": 2.8164966436059313e-05}
{"train_loss": 0.09113869071006775, "global_step": 174952, "epoch": 1965, "lr": 2.8164444902494874e-05}
{"train_loss": 0.11341425031423569, "global_step": 174953, "epoch": 1965, "lr": 2.8163923371865942e-05}
{"train_loss": 0.08651530742645264, "global_step": 174954, "epoch": 1965, "lr": 2.816340184417262e-05}
{"train_loss": 0.12170455604791641, "global_step": 174955, "epoch": 1965, "lr": 2.816288031941495e-05}
{"train_loss": 0.12782996892929077, "global_step": 174956, "epoch": 1965, "lr": 2.8162358797593025e-05}
{"train_loss": 0.10429880023002625, "global_step": 174957, "epoch": 1965, "lr": 2.8161837278706894e-05}
{"train_loss": 0.05823523923754692, "global_step": 174958, "epoch": 1965, "lr": 2.816131576275665e-05}
{"train_loss": 0.07243014872074127, "global_step": 174959, "epoch": 1965, "lr": 2.816079424974234e-05}
{"train_loss": 0.05147199332714081, "global_step": 174960, "epoch": 1965, "lr": 2.8160272739664056e-05}
{"train_loss": 0.09757692366838455, "global_step": 174961, "epoch": 1965, "lr": 2.815975123252184e-05}
{"train_loss": 0.10835660994052887, "global_step": 174962, "epoch": 1965, "lr": 2.81592297283158e-05}
{"train_loss": 0.13411009311676025, "global_step": 174963, "epoch": 1965, "lr": 2.8158708227045966e-05}
{"train_loss": 0.12781789898872375, "global_step": 174964, "epoch": 1965, "lr": 2.8158186728712425e-05}
{"train_loss": 0.07475288957357407, "global_step": 174965, "epoch": 1965, "lr": 2.8157665233315277e-05}
{"train_loss": 0.05298006162047386, "global_step": 174966, "epoch": 1965, "lr": 2.8157143740854534e-05}
{"train_loss": 0.06601672619581223, "global_step": 174967, "epoch": 1965, "lr": 2.8156622251330323e-05}
{"train_loss": 0.11415906250476837, "global_step": 174968, "epoch": 1965, "lr": 2.8156100764742666e-05}
{"train_loss": 0.04261896759271622, "global_step": 174969, "epoch": 1965, "lr": 2.8155579281091683e-05}
{"train_loss": 0.04272438958287239, "global_step": 174970, "epoch": 1965, "lr": 2.8155057800377383e-05}
{"train_loss": 0.10378937423229218, "global_step": 174971, "epoch": 1965, "lr": 2.8154536322599896e-05}
{"train_loss": 0.06427416205406189, "global_step": 174972, "epoch": 1965, "lr": 2.8154014847759248e-05}
{"train_loss": 0.08562274930182467, "global_step": 174973, "epoch": 1965, "lr": 2.8153493375855537e-05, "val_loss": 7.1818108558654785, "train_action_mse_error": 18.440649032592773}
{"train_loss": 0.11659468710422516, "global_step": 174974, "epoch": 1966, "lr": 2.8152971906888815e-05}
{"train_loss": 0.06671972572803497, "global_step": 174975, "epoch": 1966, "lr": 2.8152450440859167e-05}
{"train_loss": 0.04573925584554672, "global_step": 174976, "epoch": 1966, "lr": 2.8151928977766645e-05}
{"train_loss": 0.09075367450714111, "global_step": 174977, "epoch": 1966, "lr": 2.815140751761135e-05}
{"train_loss": 0.05989520624279976, "global_step": 174978, "epoch": 1966, "lr": 2.8150886060393307e-05}
{"train_loss": 0.09031280875205994, "global_step": 174979, "epoch": 1966, "lr": 2.8150364606112632e-05}
{"train_loss": 0.11110642552375793, "global_step": 174980, "epoch": 1966, "lr": 2.8149843154769352e-05}
{"train_loss": 0.11182968318462372, "global_step": 174981, "epoch": 1966, "lr": 2.8149321706363586e-05}
{"train_loss": 0.10121866315603256, "global_step": 174982, "epoch": 1966, "lr": 2.8148800260895354e-05}
{"train_loss": 0.03596282750368118, "global_step": 174983, "epoch": 1966, "lr": 2.814827881836477e-05}
{"train_loss": 0.060873936861753464, "global_step": 174984, "epoch": 1966, "lr": 2.814775737877186e-05}
{"train_loss": 0.07123561948537827, "global_step": 174985, "epoch": 1966, "lr": 2.814723594211674e-05}
{"train_loss": 0.1226358488202095, "global_step": 174986, "epoch": 1966, "lr": 2.814671450839946e-05}
{"train_loss": 0.09414126724004745, "global_step": 174987, "epoch": 1966, "lr": 2.8146193077620066e-05}
{"train_loss": 0.14022615551948547, "global_step": 174988, "epoch": 1966, "lr": 2.814567164977867e-05}
{"train_loss": 0.08762861043214798, "global_step": 174989, "epoch": 1966, "lr": 2.81451502248753e-05}
{"train_loss": 0.10982076823711395, "global_step": 174990, "epoch": 1966, "lr": 2.8144628802910067e-05}
{"train_loss": 0.09297139942646027, "global_step": 174991, "epoch": 1966, "lr": 2.8144107383883012e-05}
{"train_loss": 0.06751181930303574, "global_step": 174992, "epoch": 1966, "lr": 2.814358596779423e-05}
{"train_loss": 0.14547280967235565, "global_step": 174993, "epoch": 1966, "lr": 2.814306455464375e-05}
{"train_loss": 0.034462422132492065, "global_step": 174994, "epoch": 1966, "lr": 2.81425431444317e-05}
{"train_loss": 0.15483956038951874, "global_step": 174995, "epoch": 1966, "lr": 2.8142021737158086e-05}
{"train_loss": 0.12097181379795074, "global_step": 174996, "epoch": 1966, "lr": 2.814150033282304e-05}
{"train_loss": 0.028952300548553467, "global_step": 174997, "epoch": 1966, "lr": 2.8140978931426577e-05}
{"train_loss": 0.12137170881032944, "global_step": 174998, "epoch": 1966, "lr": 2.8140457532968823e-05}
{"train_loss": 0.057129841297864914, "global_step": 174999, "epoch": 1966, "lr": 2.8139936137449784e-05}
{"train_loss": 0.10521085560321808, "global_step": 175000, "epoch": 1966, "lr": 2.8139414744869575e-05}
{"train_loss": 0.07084010541439056, "global_step": 175001, "epoch": 1966, "lr": 2.8138893355228268e-05}
{"train_loss": 0.06997490674257278, "global_step": 175002, "epoch": 1966, "lr": 2.8138371968525907e-05}
{"train_loss": 0.08446628600358963, "global_step": 175003, "epoch": 1966, "lr": 2.8137850584762592e-05}
{"train_loss": 0.18848472833633423, "global_step": 175004, "epoch": 1966, "lr": 2.8137329203938356e-05}
{"train_loss": 0.07316620647907257, "global_step": 175005, "epoch": 1966, "lr": 2.8136807826053314e-05}
{"train_loss": 0.07180019468069077, "global_step": 175006, "epoch": 1966, "lr": 2.8136286451107486e-05}
{"train_loss": 0.03317167982459068, "global_step": 175007, "epoch": 1966, "lr": 2.813576507910099e-05}
{"train_loss": 0.08029098808765411, "global_step": 175008, "epoch": 1966, "lr": 2.8135243710033854e-05}
{"train_loss": 0.04378098249435425, "global_step": 175009, "epoch": 1966, "lr": 2.813472234390619e-05}
{"train_loss": 0.08049184083938599, "global_step": 175010, "epoch": 1966, "lr": 2.8134200980718022e-05}
{"train_loss": 0.07860321551561356, "global_step": 175011, "epoch": 1966, "lr": 2.8133679620469466e-05}
{"train_loss": 0.10451977699995041, "global_step": 175012, "epoch": 1966, "lr": 2.8133158263160552e-05}
{"train_loss": 0.04293396323919296, "global_step": 175013, "epoch": 1966, "lr": 2.8132636908791387e-05}
{"train_loss": 0.05716143921017647, "global_step": 175014, "epoch": 1966, "lr": 2.8132115557362003e-05}
{"train_loss": 0.07491263002157211, "global_step": 175015, "epoch": 1966, "lr": 2.8131594208872508e-05}
{"train_loss": 0.03290955349802971, "global_step": 175016, "epoch": 1966, "lr": 2.813107286332294e-05}
{"train_loss": 0.10722340643405914, "global_step": 175017, "epoch": 1966, "lr": 2.813055152071339e-05}
{"train_loss": 0.07077488303184509, "global_step": 175018, "epoch": 1966, "lr": 2.813003018104391e-05}
{"train_loss": 0.0876031294465065, "global_step": 175019, "epoch": 1966, "lr": 2.81295088443146e-05}
{"train_loss": 0.10829273611307144, "global_step": 175020, "epoch": 1966, "lr": 2.8128987510525482e-05}
{"train_loss": 0.060462865978479385, "global_step": 175021, "epoch": 1966, "lr": 2.8128466179676688e-05}
{"train_loss": 0.09289559721946716, "global_step": 175022, "epoch": 1966, "lr": 2.812794485176823e-05}
{"train_loss": 0.09406432509422302, "global_step": 175023, "epoch": 1966, "lr": 2.8127423526800223e-05}
{"train_loss": 0.08243607729673386, "global_step": 175024, "epoch": 1966, "lr": 2.812690220477272e-05}
{"train_loss": 0.08701107650995255, "global_step": 175025, "epoch": 1966, "lr": 2.8126380885685765e-05}
{"train_loss": 0.05740411952137947, "global_step": 175026, "epoch": 1966, "lr": 2.812585956953947e-05}
{"train_loss": 0.10356654971837997, "global_step": 175027, "epoch": 1966, "lr": 2.812533825633387e-05}
{"train_loss": 0.06696920096874237, "global_step": 175028, "epoch": 1966, "lr": 2.8124816946069072e-05}
{"train_loss": 0.07219933718442917, "global_step": 175029, "epoch": 1966, "lr": 2.8124295638745103e-05}
{"train_loss": 0.11778496205806732, "global_step": 175030, "epoch": 1966, "lr": 2.812377433436208e-05}
{"train_loss": 0.07692486047744751, "global_step": 175031, "epoch": 1966, "lr": 2.8123253032920027e-05}
{"train_loss": 0.06189547851681709, "global_step": 175032, "epoch": 1966, "lr": 2.812273173441905e-05}
{"train_loss": 0.05402718856930733, "global_step": 175033, "epoch": 1966, "lr": 2.8122210438859192e-05}
{"train_loss": 0.08061889559030533, "global_step": 175034, "epoch": 1966, "lr": 2.812168914624056e-05}
{"train_loss": 0.1175500676035881, "global_step": 175035, "epoch": 1966, "lr": 2.812116785656317e-05}
{"train_loss": 0.07034317404031754, "global_step": 175036, "epoch": 1966, "lr": 2.8120646569827147e-05}
{"train_loss": 0.09018081426620483, "global_step": 175037, "epoch": 1966, "lr": 2.8120125286032517e-05}
{"train_loss": 0.10716809332370758, "global_step": 175038, "epoch": 1966, "lr": 2.8119604005179368e-05}
{"train_loss": 0.060625653713941574, "global_step": 175039, "epoch": 1966, "lr": 2.811908272726779e-05}
{"train_loss": 0.11674683541059494, "global_step": 175040, "epoch": 1966, "lr": 2.8118561452297822e-05}
{"train_loss": 0.061332814395427704, "global_step": 175041, "epoch": 1966, "lr": 2.8118040180269557e-05}
{"train_loss": 0.0512094683945179, "global_step": 175042, "epoch": 1966, "lr": 2.8117518911183037e-05}
{"train_loss": 0.11009608954191208, "global_step": 175043, "epoch": 1966, "lr": 2.811699764503838e-05}
{"train_loss": 0.10136988759040833, "global_step": 175044, "epoch": 1966, "lr": 2.81164763818356e-05}
{"train_loss": 0.06468641012907028, "global_step": 175045, "epoch": 1966, "lr": 2.8115955121574812e-05}
{"train_loss": 0.08162371814250946, "global_step": 175046, "epoch": 1966, "lr": 2.811543386425604e-05}
{"train_loss": 0.04949342459440231, "global_step": 175047, "epoch": 1966, "lr": 2.8114912609879406e-05}
{"train_loss": 0.06101539731025696, "global_step": 175048, "epoch": 1966, "lr": 2.8114391358444934e-05}
{"train_loss": 0.04067616909742355, "global_step": 175049, "epoch": 1966, "lr": 2.8113870109952744e-05}
{"train_loss": 0.06982029229402542, "global_step": 175050, "epoch": 1966, "lr": 2.8113348864402854e-05}
{"train_loss": 0.049793634563684464, "global_step": 175051, "epoch": 1966, "lr": 2.8112827621795374e-05}
{"train_loss": 0.10676001012325287, "global_step": 175052, "epoch": 1966, "lr": 2.8112306382130337e-05}
{"train_loss": 0.063052237033844, "global_step": 175053, "epoch": 1966, "lr": 2.8111785145407855e-05}
{"train_loss": 0.04613572731614113, "global_step": 175054, "epoch": 1966, "lr": 2.811126391162796e-05}
{"train_loss": 0.10656727105379105, "global_step": 175055, "epoch": 1966, "lr": 2.8110742680790737e-05}
{"train_loss": 0.1267605423927307, "global_step": 175056, "epoch": 1966, "lr": 2.8110221452896274e-05}
{"train_loss": 0.06400725990533829, "global_step": 175057, "epoch": 1966, "lr": 2.8109700227944623e-05}
{"train_loss": 0.09093587845563889, "global_step": 175058, "epoch": 1966, "lr": 2.810917900593585e-05}
{"train_loss": 0.11488428711891174, "global_step": 175059, "epoch": 1966, "lr": 2.8108657786870037e-05}
{"train_loss": 0.11984536796808243, "global_step": 175060, "epoch": 1966, "lr": 2.8108136570747234e-05}
{"train_loss": 0.10904711484909058, "global_step": 175061, "epoch": 1966, "lr": 2.810761535756754e-05}
{"train_loss": 0.08362323983331744, "global_step": 175062, "epoch": 1966, "lr": 2.8107094147331014e-05, "val_loss": 7.275794506072998}
{"train_loss": 0.08247868716716766, "global_step": 175063, "epoch": 1967, "lr": 2.8106572940037707e-05}
{"train_loss": 0.048266928642988205, "global_step": 175064, "epoch": 1967, "lr": 2.8106051735687723e-05}
{"train_loss": 0.1123291477560997, "global_step": 175065, "epoch": 1967, "lr": 2.8105530534281088e-05}
{"train_loss": 0.10305442661046982, "global_step": 175066, "epoch": 1967, "lr": 2.8105009335817915e-05}
{"train_loss": 0.07649222016334534, "global_step": 175067, "epoch": 1967, "lr": 2.8104488140298247e-05}
{"train_loss": 0.09061917662620544, "global_step": 175068, "epoch": 1967, "lr": 2.8103966947722177e-05}
{"train_loss": 0.17987635731697083, "global_step": 175069, "epoch": 1967, "lr": 2.810344575808974e-05}
{"train_loss": 0.2154846489429474, "global_step": 175070, "epoch": 1967, "lr": 2.8102924571401058e-05}
{"train_loss": 0.09693234413862228, "global_step": 175071, "epoch": 1967, "lr": 2.810240338765614e-05}
{"train_loss": 0.12506337463855743, "global_step": 175072, "epoch": 1967, "lr": 2.8101882206855112e-05}
{"train_loss": 0.12525899708271027, "global_step": 175073, "epoch": 1967, "lr": 2.8101361028997995e-05}
{"train_loss": 0.09186139702796936, "global_step": 175074, "epoch": 1967, "lr": 2.8100839854084887e-05}
{"train_loss": 0.17503607273101807, "global_step": 175075, "epoch": 1967, "lr": 2.810031868211587e-05}
{"train_loss": 0.07878702133893967, "global_step": 175076, "epoch": 1967, "lr": 2.8099797513090987e-05}
{"train_loss": 0.07429835200309753, "global_step": 175077, "epoch": 1967, "lr": 2.8099276347010328e-05}
{"train_loss": 0.085447758436203, "global_step": 175078, "epoch": 1967, "lr": 2.809875518387395e-05}
{"train_loss": 0.06905807554721832, "global_step": 175079, "epoch": 1967, "lr": 2.8098234023681934e-05}
{"train_loss": 0.06559095531702042, "global_step": 175080, "epoch": 1967, "lr": 2.809771286643432e-05}
{"train_loss": 0.05905911326408386, "global_step": 175081, "epoch": 1967, "lr": 2.8097191712131232e-05}
{"train_loss": 0.11108499765396118, "global_step": 175082, "epoch": 1967, "lr": 2.8096670560772687e-05}
{"train_loss": 0.10443044453859329, "global_step": 175083, "epoch": 1967, "lr": 2.8096149412358798e-05}
{"train_loss": 0.10705654323101044, "global_step": 175084, "epoch": 1967, "lr": 2.8095628266889594e-05}
{"train_loss": 0.0862388089299202, "global_step": 175085, "epoch": 1967, "lr": 2.8095107124365183e-05}
{"train_loss": 0.19452346861362457, "global_step": 175086, "epoch": 1967, "lr": 2.8094585984785603e-05}
{"train_loss": 0.06918404996395111, "global_step": 175087, "epoch": 1967, "lr": 2.8094064848150957e-05}
{"train_loss": 0.08142706751823425, "global_step": 175088, "epoch": 1967, "lr": 2.8093543714461275e-05}
{"train_loss": 0.06557998806238174, "global_step": 175089, "epoch": 1967, "lr": 2.8093022583716666e-05}
{"train_loss": 0.16926082968711853, "global_step": 175090, "epoch": 1967, "lr": 2.8092501455917176e-05}
{"train_loss": 0.08701848983764648, "global_step": 175091, "epoch": 1967, "lr": 2.809198033106287e-05}
{"train_loss": 0.06304612755775452, "global_step": 175092, "epoch": 1967, "lr": 2.8091459209153853e-05}
{"train_loss": 0.09554444253444672, "global_step": 175093, "epoch": 1967, "lr": 2.8090938090190157e-05}
{"train_loss": 0.06240234896540642, "global_step": 175094, "epoch": 1967, "lr": 2.809041697417188e-05}
{"train_loss": 0.06363217532634735, "global_step": 175095, "epoch": 1967, "lr": 2.808989586109908e-05}
{"train_loss": 0.07414770871400833, "global_step": 175096, "epoch": 1967, "lr": 2.808937475097181e-05}
{"train_loss": 0.07368198037147522, "global_step": 175097, "epoch": 1967, "lr": 2.808885364379018e-05}
{"train_loss": 0.04753107577562332, "global_step": 175098, "epoch": 1967, "lr": 2.8088332539554208e-05}
{"train_loss": 0.03507393226027489, "global_step": 175099, "epoch": 1967, "lr": 2.808781143826402e-05}
{"train_loss": 0.07878608256578445, "global_step": 175100, "epoch": 1967, "lr": 2.8087290339919635e-05}
{"train_loss": 0.11504960060119629, "global_step": 175101, "epoch": 1967, "lr": 2.8086769244521165e-05}
{"train_loss": 0.02239106595516205, "global_step": 175102, "epoch": 1967, "lr": 2.8086248152068662e-05}
{"train_loss": 0.14681805670261383, "global_step": 175103, "epoch": 1967, "lr": 2.8085727062562184e-05}
{"train_loss": 0.07703888416290283, "global_step": 175104, "epoch": 1967, "lr": 2.808520597600183e-05}
{"train_loss": 0.11164964735507965, "global_step": 175105, "epoch": 1967, "lr": 2.8084684892387626e-05}
{"train_loss": 0.09307830780744553, "global_step": 175106, "epoch": 1967, "lr": 2.808416381171969e-05}
{"train_loss": 0.10879777371883392, "global_step": 175107, "epoch": 1967, "lr": 2.8083642733998055e-05}
{"train_loss": 0.08468779176473618, "global_step": 175108, "epoch": 1967, "lr": 2.8083121659222822e-05}
{"train_loss": 0.1130463182926178, "global_step": 175109, "epoch": 1967, "lr": 2.8082600587394036e-05}
{"train_loss": 0.10384038835763931, "global_step": 175110, "epoch": 1967, "lr": 2.8082079518511774e-05}
{"train_loss": 0.08777287602424622, "global_step": 175111, "epoch": 1967, "lr": 2.8081558452576125e-05}
{"train_loss": 0.04380828142166138, "global_step": 175112, "epoch": 1967, "lr": 2.808103738958712e-05}
{"train_loss": 0.11196780204772949, "global_step": 175113, "epoch": 1967, "lr": 2.808051632954488e-05}
{"train_loss": 0.08722709864377975, "global_step": 175114, "epoch": 1967, "lr": 2.8079995272449423e-05}
{"train_loss": 0.11107248812913895, "global_step": 175115, "epoch": 1967, "lr": 2.8079474218300865e-05}
{"train_loss": 0.059039849787950516, "global_step": 175116, "epoch": 1967, "lr": 2.8078953167099236e-05}
{"train_loss": 0.06335663050413132, "global_step": 175117, "epoch": 1967, "lr": 2.8078432118844645e-05}
{"train_loss": 0.09992056339979172, "global_step": 175118, "epoch": 1967, "lr": 2.807791107353712e-05}
{"train_loss": 0.0627141147851944, "global_step": 175119, "epoch": 1967, "lr": 2.807739003117677e-05}
{"train_loss": 0.09417301416397095, "global_step": 175120, "epoch": 1967, "lr": 2.8076868991763632e-05}
{"train_loss": 0.06562850624322891, "global_step": 175121, "epoch": 1967, "lr": 2.807634795529781e-05}
{"train_loss": 0.07344542443752289, "global_step": 175122, "epoch": 1967, "lr": 2.8075826921779337e-05}
{"train_loss": 0.0787777304649353, "global_step": 175123, "epoch": 1967, "lr": 2.807530589120832e-05}
{"train_loss": 0.08745574206113815, "global_step": 175124, "epoch": 1967, "lr": 2.807478486358479e-05}
{"train_loss": 0.08064065128564835, "global_step": 175125, "epoch": 1967, "lr": 2.807426383890886e-05}
{"train_loss": 0.06892991065979004, "global_step": 175126, "epoch": 1967, "lr": 2.807374281718056e-05}
{"train_loss": 0.075645312666893, "global_step": 175127, "epoch": 1967, "lr": 2.8073221798399973e-05}
{"train_loss": 0.08687690645456314, "global_step": 175128, "epoch": 1967, "lr": 2.8072700782567197e-05}
{"train_loss": 0.03984879329800606, "global_step": 175129, "epoch": 1967, "lr": 2.8072179769682262e-05}
{"train_loss": 0.10039114207029343, "global_step": 175130, "epoch": 1967, "lr": 2.8071658759745266e-05}
{"train_loss": 0.13414183259010315, "global_step": 175131, "epoch": 1967, "lr": 2.807113775275626e-05}
{"train_loss": 0.06837976723909378, "global_step": 175132, "epoch": 1967, "lr": 2.8070616748715335e-05}
{"train_loss": 0.06969230622053146, "global_step": 175133, "epoch": 1967, "lr": 2.807009574762253e-05}
{"train_loss": 0.060367438942193985, "global_step": 175134, "epoch": 1967, "lr": 2.806957474947795e-05}
{"train_loss": 0.11079266667366028, "global_step": 175135, "epoch": 1967, "lr": 2.8069053754281654e-05}
{"train_loss": 0.07879727333784103, "global_step": 175136, "epoch": 1967, "lr": 2.8068532762033684e-05}
{"train_loss": 0.07250595837831497, "global_step": 175137, "epoch": 1967, "lr": 2.8068011772734148e-05}
{"train_loss": 0.042608875781297684, "global_step": 175138, "epoch": 1967, "lr": 2.806749078638309e-05}
{"train_loss": 0.033966172486543655, "global_step": 175139, "epoch": 1967, "lr": 2.8066969802980603e-05}
{"train_loss": 0.0665496289730072, "global_step": 175140, "epoch": 1967, "lr": 2.8066448822526747e-05}
{"train_loss": 0.13681556284427643, "global_step": 175141, "epoch": 1967, "lr": 2.8065927845021568e-05}
{"train_loss": 0.0550839938223362, "global_step": 175142, "epoch": 1967, "lr": 2.8065406870465182e-05}
{"train_loss": 0.07260838150978088, "global_step": 175143, "epoch": 1967, "lr": 2.8064885898857613e-05}
{"train_loss": 0.08650215715169907, "global_step": 175144, "epoch": 1967, "lr": 2.8064364930198972e-05}
{"train_loss": 0.09699497371912003, "global_step": 175145, "epoch": 1967, "lr": 2.806384396448929e-05}
{"train_loss": 0.07146111875772476, "global_step": 175146, "epoch": 1967, "lr": 2.8063323001728654e-05}
{"train_loss": 0.03609134629368782, "global_step": 175147, "epoch": 1967, "lr": 2.8062802041917157e-05}
{"train_loss": 0.031725794076919556, "global_step": 175148, "epoch": 1967, "lr": 2.8062281085054832e-05}
{"train_loss": 0.0692124143242836, "global_step": 175149, "epoch": 1967, "lr": 2.806176013114179e-05}
{"train_loss": 0.07300487905740738, "global_step": 175150, "epoch": 1967, "lr": 2.8061239180178045e-05}
{"train_loss": 0.0875237299485153, "global_step": 175151, "epoch": 1967, "lr": 2.806071823216373e-05, "val_loss": 7.2546162605285645}
{"train_loss": 0.0583806186914444, "global_step": 175152, "epoch": 1968, "lr": 2.8060197287098854e-05}
{"train_loss": 0.05827122926712036, "global_step": 175153, "epoch": 1968, "lr": 2.805967634498355e-05}
{"train_loss": 0.11756999790668488, "global_step": 175154, "epoch": 1968, "lr": 2.8059155405817826e-05}
{"train_loss": 0.05637047812342644, "global_step": 175155, "epoch": 1968, "lr": 2.80586344696018e-05}
{"train_loss": 0.07263276726007462, "global_step": 175156, "epoch": 1968, "lr": 2.8058113536335506e-05}
{"train_loss": 0.04800530523061752, "global_step": 175157, "epoch": 1968, "lr": 2.8057592606019055e-05}
{"train_loss": 0.11680879443883896, "global_step": 175158, "epoch": 1968, "lr": 2.805707167865247e-05}
{"train_loss": 0.06658042222261429, "global_step": 175159, "epoch": 1968, "lr": 2.8056550754235865e-05}
{"train_loss": 0.09635813534259796, "global_step": 175160, "epoch": 1968, "lr": 2.805602983276927e-05}
{"train_loss": 0.08631963282823563, "global_step": 175161, "epoch": 1968, "lr": 2.8055508914252792e-05}
{"train_loss": 0.0887916311621666, "global_step": 175162, "epoch": 1968, "lr": 2.8054987998686467e-05}
{"train_loss": 0.14437396824359894, "global_step": 175163, "epoch": 1968, "lr": 2.805446708607039e-05}
{"train_loss": 0.033141423016786575, "global_step": 175164, "epoch": 1968, "lr": 2.805394617640461e-05}
{"train_loss": 0.12065356969833374, "global_step": 175165, "epoch": 1968, "lr": 2.8053425269689215e-05}
{"train_loss": 0.10310317575931549, "global_step": 175166, "epoch": 1968, "lr": 2.8052904365924282e-05}
{"train_loss": 0.10584878921508789, "global_step": 175167, "epoch": 1968, "lr": 2.8052383465109857e-05}
{"train_loss": 0.10428960621356964, "global_step": 175168, "epoch": 1968, "lr": 2.8051862567246036e-05}
{"train_loss": 0.062011487782001495, "global_step": 175169, "epoch": 1968, "lr": 2.8051341672332855e-05}
{"train_loss": 0.1048518568277359, "global_step": 175170, "epoch": 1968, "lr": 2.8050820780370425e-05}
{"train_loss": 0.10521508008241653, "global_step": 175171, "epoch": 1968, "lr": 2.8050299891358766e-05}
{"train_loss": 0.06275705993175507, "global_step": 175172, "epoch": 1968, "lr": 2.8049779005298004e-05}
{"train_loss": 0.05210902541875839, "global_step": 175173, "epoch": 1968, "lr": 2.8049258122188183e-05}
{"train_loss": 0.062031060457229614, "global_step": 175174, "epoch": 1968, "lr": 2.804873724202935e-05}
{"train_loss": 0.07409914582967758, "global_step": 175175, "epoch": 1968, "lr": 2.804821636482161e-05}
{"train_loss": 0.08126891404390335, "global_step": 175176, "epoch": 1968, "lr": 2.8047695490565012e-05}
{"train_loss": 0.07055340707302094, "global_step": 175177, "epoch": 1968, "lr": 2.804717461925965e-05}
{"train_loss": 0.12903882563114166, "global_step": 175178, "epoch": 1968, "lr": 2.8046653750905572e-05}
{"train_loss": 0.06398800760507584, "global_step": 175179, "epoch": 1968, "lr": 2.804613288550284e-05}
{"train_loss": 0.0999271422624588, "global_step": 175180, "epoch": 1968, "lr": 2.8045612023051555e-05}
{"train_loss": 0.038881853222846985, "global_step": 175181, "epoch": 1968, "lr": 2.804509116355175e-05}
{"train_loss": 0.0747700035572052, "global_step": 175182, "epoch": 1968, "lr": 2.8044570307003514e-05}
{"train_loss": 0.038978636264801025, "global_step": 175183, "epoch": 1968, "lr": 2.804404945340694e-05}
{"train_loss": 0.10295065492391586, "global_step": 175184, "epoch": 1968, "lr": 2.8043528602762052e-05}
{"train_loss": 0.09648281335830688, "global_step": 175185, "epoch": 1968, "lr": 2.8043007755068972e-05}
{"train_loss": 0.059498973190784454, "global_step": 175186, "epoch": 1968, "lr": 2.8042486910327714e-05}
{"train_loss": 0.05256202816963196, "global_step": 175187, "epoch": 1968, "lr": 2.80419660685384e-05}
{"train_loss": 0.08029291033744812, "global_step": 175188, "epoch": 1968, "lr": 2.8041445229701056e-05}
{"train_loss": 0.11322493851184845, "global_step": 175189, "epoch": 1968, "lr": 2.8040924393815792e-05}
{"train_loss": 0.08312146365642548, "global_step": 175190, "epoch": 1968, "lr": 2.8040403560882634e-05}
{"train_loss": 0.07886793464422226, "global_step": 175191, "epoch": 1968, "lr": 2.80398827309017e-05}
{"train_loss": 0.14111481606960297, "global_step": 175192, "epoch": 1968, "lr": 2.803936190387302e-05}
{"train_loss": 0.06911975890398026, "global_step": 175193, "epoch": 1968, "lr": 2.803884107979669e-05}
{"train_loss": 0.08904106914997101, "global_step": 175194, "epoch": 1968, "lr": 2.8038320258672756e-05}
{"train_loss": 0.09067978709936142, "global_step": 175195, "epoch": 1968, "lr": 2.8037799440501322e-05}
{"train_loss": 0.11283185333013535, "global_step": 175196, "epoch": 1968, "lr": 2.8037278625282422e-05}
{"train_loss": 0.029203159734606743, "global_step": 175197, "epoch": 1968, "lr": 2.8036757813016156e-05}
{"train_loss": 0.04837632179260254, "global_step": 175198, "epoch": 1968, "lr": 2.803623700370256e-05}
{"train_loss": 0.071694515645504, "global_step": 175199, "epoch": 1968, "lr": 2.803571619734175e-05}
{"train_loss": 0.0682961493730545, "global_step": 175200, "epoch": 1968, "lr": 2.8035195393933745e-05}
{"train_loss": 0.08334770798683167, "global_step": 175201, "epoch": 1968, "lr": 2.8034674593478642e-05}
{"train_loss": 0.09611476212739944, "global_step": 175202, "epoch": 1968, "lr": 2.803415379597653e-05}
{"train_loss": 0.11040414869785309, "global_step": 175203, "epoch": 1968, "lr": 2.803363300142744e-05}
{"train_loss": 0.09149185568094254, "global_step": 175204, "epoch": 1968, "lr": 2.803311220983148e-05}
{"train_loss": 0.0495750717818737, "global_step": 175205, "epoch": 1968, "lr": 2.8032591421188677e-05}
{"train_loss": 0.1027199849486351, "global_step": 175206, "epoch": 1968, "lr": 2.803207063549914e-05}
{"train_loss": 0.04752976447343826, "global_step": 175207, "epoch": 1968, "lr": 2.8031549852762913e-05}
{"train_loss": 0.06479614973068237, "global_step": 175208, "epoch": 1968, "lr": 2.803102907298009e-05}
{"train_loss": 0.0982765182852745, "global_step": 175209, "epoch": 1968, "lr": 2.8030508296150715e-05}
{"train_loss": 0.08865846693515778, "global_step": 175210, "epoch": 1968, "lr": 2.802998752227488e-05}
{"train_loss": 0.0882478579878807, "global_step": 175211, "epoch": 1968, "lr": 2.8029466751352634e-05}
{"train_loss": 0.1595628559589386, "global_step": 175212, "epoch": 1968, "lr": 2.8028945983384074e-05}
{"train_loss": 0.0817757248878479, "global_step": 175213, "epoch": 1968, "lr": 2.8028425218369257e-05}
{"train_loss": 0.09972228854894638, "global_step": 175214, "epoch": 1968, "lr": 2.802790445630823e-05}
{"train_loss": 0.055734556168317795, "global_step": 175215, "epoch": 1968, "lr": 2.8027383697201105e-05}
{"train_loss": 0.08584319055080414, "global_step": 175216, "epoch": 1968, "lr": 2.8026862941047906e-05}
{"train_loss": 0.05001210793852806, "global_step": 175217, "epoch": 1968, "lr": 2.8026342187848748e-05}
{"train_loss": 0.056332867592573166, "global_step": 175218, "epoch": 1968, "lr": 2.802582143760366e-05}
{"train_loss": 0.05801510065793991, "global_step": 175219, "epoch": 1968, "lr": 2.802530069031275e-05}
{"train_loss": 0.060357291251420975, "global_step": 175220, "epoch": 1968, "lr": 2.802477994597606e-05}
{"train_loss": 0.07679197937250137, "global_step": 175221, "epoch": 1968, "lr": 2.8024259204593683e-05}
{"train_loss": 0.0836518257856369, "global_step": 175222, "epoch": 1968, "lr": 2.802373846616566e-05}
{"train_loss": 0.05545414984226227, "global_step": 175223, "epoch": 1968, "lr": 2.8023217730692097e-05}
{"train_loss": 0.04487829655408859, "global_step": 175224, "epoch": 1968, "lr": 2.8022696998173025e-05}
{"train_loss": 0.09994091093540192, "global_step": 175225, "epoch": 1968, "lr": 2.8022176268608546e-05}
{"train_loss": 0.1410309225320816, "global_step": 175226, "epoch": 1968, "lr": 2.802165554199871e-05}
{"train_loss": 0.15002362430095673, "global_step": 175227, "epoch": 1968, "lr": 2.8021134818343607e-05}
{"train_loss": 0.07762288302183151, "global_step": 175228, "epoch": 1968, "lr": 2.802061409764327e-05}
{"train_loss": 0.04030942916870117, "global_step": 175229, "epoch": 1968, "lr": 2.8020093379897828e-05}
{"train_loss": 0.08117501437664032, "global_step": 175230, "epoch": 1968, "lr": 2.8019572665107284e-05}
{"train_loss": 0.08999401330947876, "global_step": 175231, "epoch": 1968, "lr": 2.801905195327177e-05}
{"train_loss": 0.08672269433736801, "global_step": 175232, "epoch": 1968, "lr": 2.80185312443913e-05}
{"train_loss": 0.07609900087118149, "global_step": 175233, "epoch": 1968, "lr": 2.801801053846599e-05}
{"train_loss": 0.14283499121665955, "global_step": 175234, "epoch": 1968, "lr": 2.8017489835495868e-05}
{"train_loss": 0.08916453272104263, "global_step": 175235, "epoch": 1968, "lr": 2.801696913548105e-05}
{"train_loss": 0.10143319517374039, "global_step": 175236, "epoch": 1968, "lr": 2.8016448438421566e-05}
{"train_loss": 0.14348025619983673, "global_step": 175237, "epoch": 1968, "lr": 2.8015927744317498e-05}
{"train_loss": 0.11195742338895798, "global_step": 175238, "epoch": 1968, "lr": 2.801540705316894e-05}
{"train_loss": 0.06031410023570061, "global_step": 175239, "epoch": 1968, "lr": 2.801488636497593e-05}
{"train_loss": 0.08364510182500555, "global_step": 175240, "epoch": 1968, "lr": 2.8014365679738565e-05, "val_loss": 7.3224968910217285}
{"train_loss": 0.09679414331912994, "global_step": 175241, "epoch": 1969, "lr": 2.8013844997456884e-05}
{"train_loss": 0.061320967972278595, "global_step": 175242, "epoch": 1969, "lr": 2.8013324318130983e-05}
{"train_loss": 0.05786124989390373, "global_step": 175243, "epoch": 1969, "lr": 2.8012803641760915e-05}
{"train_loss": 0.08100762218236923, "global_step": 175244, "epoch": 1969, "lr": 2.801228296834677e-05}
{"train_loss": 0.06255493313074112, "global_step": 175245, "epoch": 1969, "lr": 2.801176229788859e-05}
{"train_loss": 0.06867434829473495, "global_step": 175246, "epoch": 1969, "lr": 2.8011241630386483e-05}
{"train_loss": 0.15353508293628693, "global_step": 175247, "epoch": 1969, "lr": 2.801072096584047e-05}
{"train_loss": 0.08224478363990784, "global_step": 175248, "epoch": 1969, "lr": 2.8010200304250665e-05}
{"train_loss": 0.12480909377336502, "global_step": 175249, "epoch": 1969, "lr": 2.800967964561711e-05}
{"train_loss": 0.08068230748176575, "global_step": 175250, "epoch": 1969, "lr": 2.80091589899399e-05}
{"train_loss": 0.06152203306555748, "global_step": 175251, "epoch": 1969, "lr": 2.8008638337219096e-05}
{"train_loss": 0.1081622838973999, "global_step": 175252, "epoch": 1969, "lr": 2.800811768745474e-05}
{"train_loss": 0.06069684401154518, "global_step": 175253, "epoch": 1969, "lr": 2.8007597040646937e-05}
{"train_loss": 0.12815409898757935, "global_step": 175254, "epoch": 1969, "lr": 2.8007076396795728e-05}
{"train_loss": 0.0848209336400032, "global_step": 175255, "epoch": 1969, "lr": 2.8006555755901225e-05}
{"train_loss": 0.05691414698958397, "global_step": 175256, "epoch": 1969, "lr": 2.8006035117963448e-05}
{"train_loss": 0.03900852054357529, "global_step": 175257, "epoch": 1969, "lr": 2.8005514482982513e-05}
{"train_loss": 0.09932460635900497, "global_step": 175258, "epoch": 1969, "lr": 2.8004993850958448e-05}
{"train_loss": 0.1266905963420868, "global_step": 175259, "epoch": 1969, "lr": 2.800447322189137e-05}
{"train_loss": 0.044777024537324905, "global_step": 175260, "epoch": 1969, "lr": 2.800395259578129e-05}
{"train_loss": 0.11422793567180634, "global_step": 175261, "epoch": 1969, "lr": 2.8003431972628335e-05}
{"train_loss": 0.12941429018974304, "global_step": 175262, "epoch": 1969, "lr": 2.8002911352432536e-05}
{"train_loss": 0.059266600757837296, "global_step": 175263, "epoch": 1969, "lr": 2.8002390735193994e-05}
{"train_loss": 0.1318991780281067, "global_step": 175264, "epoch": 1969, "lr": 2.8001870120912742e-05}
{"train_loss": 0.13117046654224396, "global_step": 175265, "epoch": 1969, "lr": 2.8001349509588892e-05}
{"train_loss": 0.06591234356164932, "global_step": 175266, "epoch": 1969, "lr": 2.800082890122247e-05}
{"train_loss": 0.08053586632013321, "global_step": 175267, "epoch": 1969, "lr": 2.8000308295813594e-05}
{"train_loss": 0.08068747073411942, "global_step": 175268, "epoch": 1969, "lr": 2.799978769336228e-05}
{"train_loss": 0.13045738637447357, "global_step": 175269, "epoch": 1969, "lr": 2.7999267093868653e-05}
{"train_loss": 0.09639621526002884, "global_step": 175270, "epoch": 1969, "lr": 2.7998746497332728e-05}
{"train_loss": 0.05997065082192421, "global_step": 175271, "epoch": 1969, "lr": 2.799822590375463e-05}
{"train_loss": 0.056264135986566544, "global_step": 175272, "epoch": 1969, "lr": 2.799770531313438e-05}
{"train_loss": 0.16219763457775116, "global_step": 175273, "epoch": 1969, "lr": 2.7997184725472074e-05}
{"train_loss": 0.137601837515831, "global_step": 175274, "epoch": 1969, "lr": 2.7996664140767796e-05}
{"train_loss": 0.10874280333518982, "global_step": 175275, "epoch": 1969, "lr": 2.799614355902158e-05}
{"train_loss": 0.05660763010382652, "global_step": 175276, "epoch": 1969, "lr": 2.7995622980233534e-05}
{"train_loss": 0.10575763136148453, "global_step": 175277, "epoch": 1969, "lr": 2.7995102404403685e-05}
{"train_loss": 0.09581553190946579, "global_step": 175278, "epoch": 1969, "lr": 2.7994581831532152e-05}
{"train_loss": 0.07685445249080658, "global_step": 175279, "epoch": 1969, "lr": 2.7994061261618953e-05}
{"train_loss": 0.06607894599437714, "global_step": 175280, "epoch": 1969, "lr": 2.7993540694664207e-05}
{"train_loss": 0.0856478288769722, "global_step": 175281, "epoch": 1969, "lr": 2.799302013066794e-05}
{"train_loss": 0.1335475593805313, "global_step": 175282, "epoch": 1969, "lr": 2.7992499569630264e-05}
{"train_loss": 0.10339436680078506, "global_step": 175283, "epoch": 1969, "lr": 2.7991979011551207e-05}
{"train_loss": 0.11345890909433365, "global_step": 175284, "epoch": 1969, "lr": 2.7991458456430886e-05}
{"train_loss": 0.07307328283786774, "global_step": 175285, "epoch": 1969, "lr": 2.799093790426932e-05}
{"train_loss": 0.041192058473825455, "global_step": 175286, "epoch": 1969, "lr": 2.7990417355066622e-05}
{"train_loss": 0.09342338144779205, "global_step": 175287, "epoch": 1969, "lr": 2.798989680882283e-05}
{"train_loss": 0.16326141357421875, "global_step": 175288, "epoch": 1969, "lr": 2.7989376265538047e-05}
{"train_loss": 0.08375655114650726, "global_step": 175289, "epoch": 1969, "lr": 2.798885572521232e-05}
{"train_loss": 0.08113200962543488, "global_step": 175290, "epoch": 1969, "lr": 2.7988335187845704e-05}
{"train_loss": 0.11151804774999619, "global_step": 175291, "epoch": 1969, "lr": 2.798781465343831e-05}
{"train_loss": 0.08299942314624786, "global_step": 175292, "epoch": 1969, "lr": 2.7987294121990165e-05}
{"train_loss": 0.1040726900100708, "global_step": 175293, "epoch": 1969, "lr": 2.798677359350138e-05}
{"train_loss": 0.07626877725124359, "global_step": 175294, "epoch": 1969, "lr": 2.798625306797199e-05}
{"train_loss": 0.058657705783843994, "global_step": 175295, "epoch": 1969, "lr": 2.7985732545402095e-05}
{"train_loss": 0.07270271331071854, "global_step": 175296, "epoch": 1969, "lr": 2.798521202579173e-05}
{"train_loss": 0.042809028178453445, "global_step": 175297, "epoch": 1969, "lr": 2.7984691509141003e-05}
{"train_loss": 0.10939610004425049, "global_step": 175298, "epoch": 1969, "lr": 2.798417099544995e-05}
{"train_loss": 0.0749501958489418, "global_step": 175299, "epoch": 1969, "lr": 2.798365048471867e-05}
{"train_loss": 0.0587216354906559, "global_step": 175300, "epoch": 1969, "lr": 2.7983129976947207e-05}
{"train_loss": 0.11649182438850403, "global_step": 175301, "epoch": 1969, "lr": 2.7982609472135663e-05}
{"train_loss": 0.05989318713545799, "global_step": 175302, "epoch": 1969, "lr": 2.798208897028407e-05}
{"train_loss": 0.04680461063981056, "global_step": 175303, "epoch": 1969, "lr": 2.7981568471392533e-05}
{"train_loss": 0.0949079617857933, "global_step": 175304, "epoch": 1969, "lr": 2.7981047975461083e-05}
{"train_loss": 0.08939949423074722, "global_step": 175305, "epoch": 1969, "lr": 2.7980527482489836e-05}
{"train_loss": 0.16458843648433685, "global_step": 175306, "epoch": 1969, "lr": 2.7980006992478824e-05}
{"train_loss": 0.11078157275915146, "global_step": 175307, "epoch": 1969, "lr": 2.797948650542814e-05}
{"train_loss": 0.09862814098596573, "global_step": 175308, "epoch": 1969, "lr": 2.797896602133784e-05}
{"train_loss": 0.04472308233380318, "global_step": 175309, "epoch": 1969, "lr": 2.7978445540207994e-05}
{"train_loss": 0.10152529925107956, "global_step": 175310, "epoch": 1969, "lr": 2.797792506203869e-05}
{"train_loss": 0.0650370866060257, "global_step": 175311, "epoch": 1969, "lr": 2.7977404586829974e-05}
{"train_loss": 0.07418148964643478, "global_step": 175312, "epoch": 1969, "lr": 2.7976884114581948e-05}
{"train_loss": 0.09531782567501068, "global_step": 175313, "epoch": 1969, "lr": 2.7976363645294633e-05}
{"train_loss": 0.0984838455915451, "global_step": 175314, "epoch": 1969, "lr": 2.797584317896816e-05}
{"train_loss": 0.06166598200798035, "global_step": 175315, "epoch": 1969, "lr": 2.7975322715602538e-05}
{"train_loss": 0.13907818496227264, "global_step": 175316, "epoch": 1969, "lr": 2.7974802255197885e-05}
{"train_loss": 0.05185265094041824, "global_step": 175317, "epoch": 1969, "lr": 2.7974281797754233e-05}
{"train_loss": 0.09511403739452362, "global_step": 175318, "epoch": 1969, "lr": 2.7973761343271692e-05}
{"train_loss": 0.060698140412569046, "global_step": 175319, "epoch": 1969, "lr": 2.7973240891750284e-05}
{"train_loss": 0.11140910536050797, "global_step": 175320, "epoch": 1969, "lr": 2.7972720443190137e-05}
{"train_loss": 0.07068560272455215, "global_step": 175321, "epoch": 1969, "lr": 2.7972199997591265e-05}
{"train_loss": 0.06099613383412361, "global_step": 175322, "epoch": 1969, "lr": 2.797167955495378e-05}
{"train_loss": 0.10508447885513306, "global_step": 175323, "epoch": 1969, "lr": 2.7971159115277713e-05}
{"train_loss": 0.059857796877622604, "global_step": 175324, "epoch": 1969, "lr": 2.7970638678563176e-05}
{"train_loss": 0.10625120997428894, "global_step": 175325, "epoch": 1969, "lr": 2.7970118244810195e-05}
{"train_loss": 0.09930025041103363, "global_step": 175326, "epoch": 1969, "lr": 2.796959781401889e-05}
{"train_loss": 0.06308184564113617, "global_step": 175327, "epoch": 1969, "lr": 2.796907738618928e-05}
{"train_loss": 0.09818179905414581, "global_step": 175328, "epoch": 1969, "lr": 2.7968556961321478e-05}
{"train_loss": 0.08897747409142805, "global_step": 175329, "epoch": 1969, "lr": 2.7968036539415533e-05, "val_loss": 7.2373127937316895}
{"train_loss": 0.10524646937847137, "global_step": 175330, "epoch": 1970, "lr": 2.7967516120471504e-05}
{"train_loss": 0.07087661325931549, "global_step": 175331, "epoch": 1970, "lr": 2.7966995704489484e-05}
{"train_loss": 0.11575724929571152, "global_step": 175332, "epoch": 1970, "lr": 2.7966475291469525e-05}
{"train_loss": 0.1422426998615265, "global_step": 175333, "epoch": 1970, "lr": 2.7965954881411714e-05}
{"train_loss": 0.12881670892238617, "global_step": 175334, "epoch": 1970, "lr": 2.79654344743161e-05}
{"train_loss": 0.16163024306297302, "global_step": 175335, "epoch": 1970, "lr": 2.796491407018278e-05}
{"train_loss": 0.06854595243930817, "global_step": 175336, "epoch": 1970, "lr": 2.796439366901179e-05}
{"train_loss": 0.06921286880970001, "global_step": 175337, "epoch": 1970, "lr": 2.796387327080324e-05}
{"train_loss": 0.09372206777334213, "global_step": 175338, "epoch": 1970, "lr": 2.7963352875557157e-05}
{"train_loss": 0.0709255188703537, "global_step": 175339, "epoch": 1970, "lr": 2.796283248327366e-05}
{"train_loss": 0.08027809113264084, "global_step": 175340, "epoch": 1970, "lr": 2.7962312093952757e-05}
{"train_loss": 0.07834985852241516, "global_step": 175341, "epoch": 1970, "lr": 2.7961791707594586e-05}
{"train_loss": 0.08222991973161697, "global_step": 175342, "epoch": 1970, "lr": 2.7961271324199157e-05}
{"train_loss": 0.08674722164869308, "global_step": 175343, "epoch": 1970, "lr": 2.7960750943766588e-05}
{"train_loss": 0.07547972351312637, "global_step": 175344, "epoch": 1970, "lr": 2.7960230566296907e-05}
{"train_loss": 0.02854856289923191, "global_step": 175345, "epoch": 1970, "lr": 2.79597101917902e-05}
{"train_loss": 0.051598697900772095, "global_step": 175346, "epoch": 1970, "lr": 2.795918982024657e-05}
{"train_loss": 0.056692007929086685, "global_step": 175347, "epoch": 1970, "lr": 2.795866945166603e-05}
{"train_loss": 0.09812832623720169, "global_step": 175348, "epoch": 1970, "lr": 2.79581490860487e-05}
{"train_loss": 0.08728084713220596, "global_step": 175349, "epoch": 1970, "lr": 2.7957628723394612e-05}
{"train_loss": 0.07703959196805954, "global_step": 175350, "epoch": 1970, "lr": 2.7957108363703878e-05}
{"train_loss": 0.08616748452186584, "global_step": 175351, "epoch": 1970, "lr": 2.7956588006976516e-05}
{"train_loss": 0.06745443493127823, "global_step": 175352, "epoch": 1970, "lr": 2.7956067653212637e-05}
{"train_loss": 0.06621365249156952, "global_step": 175353, "epoch": 1970, "lr": 2.7955547302412276e-05}
{"train_loss": 0.06844466924667358, "global_step": 175354, "epoch": 1970, "lr": 2.7955026954575547e-05}
{"train_loss": 0.059358976781368256, "global_step": 175355, "epoch": 1970, "lr": 2.795450660970248e-05}
{"train_loss": 0.04340158775448799, "global_step": 175356, "epoch": 1970, "lr": 2.7953986267793175e-05}
{"train_loss": 0.11347457766532898, "global_step": 175357, "epoch": 1970, "lr": 2.7953465928847667e-05}
{"train_loss": 0.08852989226579666, "global_step": 175358, "epoch": 1970, "lr": 2.795294559286607e-05}
{"train_loss": 0.05953587591648102, "global_step": 175359, "epoch": 1970, "lr": 2.7952425259848412e-05}
{"train_loss": 0.07142694294452667, "global_step": 175360, "epoch": 1970, "lr": 2.79519049297948e-05}
{"train_loss": 0.06902677565813065, "global_step": 175361, "epoch": 1970, "lr": 2.7951384602705265e-05}
{"train_loss": 0.0933893546462059, "global_step": 175362, "epoch": 1970, "lr": 2.7950864278579912e-05}
{"train_loss": 0.06662057340145111, "global_step": 175363, "epoch": 1970, "lr": 2.795034395741879e-05}
{"train_loss": 0.06336592137813568, "global_step": 175364, "epoch": 1970, "lr": 2.7949823639221983e-05}
{"train_loss": 0.06987769156694412, "global_step": 175365, "epoch": 1970, "lr": 2.7949303323989538e-05}
{"train_loss": 0.09558120369911194, "global_step": 175366, "epoch": 1970, "lr": 2.7948783011721563e-05}
{"train_loss": 0.04281384125351906, "global_step": 175367, "epoch": 1970, "lr": 2.7948262702418105e-05}
{"train_loss": 0.06697158515453339, "global_step": 175368, "epoch": 1970, "lr": 2.794774239607921e-05}
{"train_loss": 0.07708369195461273, "global_step": 175369, "epoch": 1970, "lr": 2.7947222092704994e-05}
{"train_loss": 0.11410816758871078, "global_step": 175370, "epoch": 1970, "lr": 2.7946701792295483e-05}
{"train_loss": 0.10074802488088608, "global_step": 175371, "epoch": 1970, "lr": 2.7946181494850798e-05}
{"train_loss": 0.06631387025117874, "global_step": 175372, "epoch": 1970, "lr": 2.7945661200370955e-05}
{"train_loss": 0.07650536298751831, "global_step": 175373, "epoch": 1970, "lr": 2.7945140908856067e-05}
{"train_loss": 0.06270980834960938, "global_step": 175374, "epoch": 1970, "lr": 2.7944620620306163e-05}
{"train_loss": 0.09683800488710403, "global_step": 175375, "epoch": 1970, "lr": 2.794410033472136e-05}
{"train_loss": 0.10260380059480667, "global_step": 175376, "epoch": 1970, "lr": 2.7943580052101688e-05}
{"train_loss": 0.0724087804555893, "global_step": 175377, "epoch": 1970, "lr": 2.7943059772447246e-05}
{"train_loss": 0.05426102504134178, "global_step": 175378, "epoch": 1970, "lr": 2.7942539495758073e-05}
{"train_loss": 0.11608461290597916, "global_step": 175379, "epoch": 1970, "lr": 2.7942019222034278e-05}
{"train_loss": 0.0617966465651989, "global_step": 175380, "epoch": 1970, "lr": 2.794149895127589e-05}
{"train_loss": 0.04559847339987755, "global_step": 175381, "epoch": 1970, "lr": 2.7940978683483e-05}
{"train_loss": 0.07563842087984085, "global_step": 175382, "epoch": 1970, "lr": 2.794045841865569e-05}
{"train_loss": 0.09054000675678253, "global_step": 175383, "epoch": 1970, "lr": 2.7939938156794e-05}
{"train_loss": 0.04519890248775482, "global_step": 175384, "epoch": 1970, "lr": 2.7939417897898033e-05}
{"train_loss": 0.10009844601154327, "global_step": 175385, "epoch": 1970, "lr": 2.7938897641967832e-05}
{"train_loss": 0.05400842800736427, "global_step": 175386, "epoch": 1970, "lr": 2.793837738900349e-05}
{"train_loss": 0.05542929470539093, "global_step": 175387, "epoch": 1970, "lr": 2.7937857139005047e-05}
{"train_loss": 0.08470740914344788, "global_step": 175388, "epoch": 1970, "lr": 2.793733689197261e-05}
{"train_loss": 0.08676719665527344, "global_step": 175389, "epoch": 1970, "lr": 2.7936816647906205e-05}
{"train_loss": 0.05454209819436073, "global_step": 175390, "epoch": 1970, "lr": 2.793629640680595e-05}
{"train_loss": 0.0729028731584549, "global_step": 175391, "epoch": 1970, "lr": 2.7935776168671872e-05}
{"train_loss": 0.08172048628330231, "global_step": 175392, "epoch": 1970, "lr": 2.7935255933504078e-05}
{"train_loss": 0.09369563311338425, "global_step": 175393, "epoch": 1970, "lr": 2.7934735701302607e-05}
{"train_loss": 0.10658598691225052, "global_step": 175394, "epoch": 1970, "lr": 2.7934215472067547e-05}
{"train_loss": 0.09244215488433838, "global_step": 175395, "epoch": 1970, "lr": 2.7933695245798952e-05}
{"train_loss": 0.06129670888185501, "global_step": 175396, "epoch": 1970, "lr": 2.7933175022496928e-05}
{"train_loss": 0.1148756593465805, "global_step": 175397, "epoch": 1970, "lr": 2.793265480216149e-05}
{"train_loss": 0.11753840744495392, "global_step": 175398, "epoch": 1970, "lr": 2.7932134584792763e-05}
{"train_loss": 0.1089572086930275, "global_step": 175399, "epoch": 1970, "lr": 2.7931614370390767e-05}
{"train_loss": 0.052207715809345245, "global_step": 175400, "epoch": 1970, "lr": 2.7931094158955617e-05}
{"train_loss": 0.03271416574716568, "global_step": 175401, "epoch": 1970, "lr": 2.793057395048735e-05}
{"train_loss": 0.09645643085241318, "global_step": 175402, "epoch": 1970, "lr": 2.7930053744986064e-05}
{"train_loss": 0.10224252194166183, "global_step": 175403, "epoch": 1970, "lr": 2.7929533542451793e-05}
{"train_loss": 0.05344494804739952, "global_step": 175404, "epoch": 1970, "lr": 2.792901334288464e-05}
{"train_loss": 0.04559055343270302, "global_step": 175405, "epoch": 1970, "lr": 2.7928493146284673e-05}
{"train_loss": 0.04519451782107353, "global_step": 175406, "epoch": 1970, "lr": 2.7927972952651926e-05}
{"train_loss": 0.05736901983618736, "global_step": 175407, "epoch": 1970, "lr": 2.792745276198651e-05}
{"train_loss": 0.09187569469213486, "global_step": 175408, "epoch": 1970, "lr": 2.7926932574288466e-05}
{"train_loss": 0.09919087588787079, "global_step": 175409, "epoch": 1970, "lr": 2.7926412389557893e-05}
{"train_loss": 0.041350193321704865, "global_step": 175410, "epoch": 1970, "lr": 2.7925892207794825e-05}
{"train_loss": 0.07223882526159286, "global_step": 175411, "epoch": 1970, "lr": 2.7925372028999374e-05}
{"train_loss": 0.11709437519311905, "global_step": 175412, "epoch": 1970, "lr": 2.7924851853171568e-05}
{"train_loss": 0.06441444158554077, "global_step": 175413, "epoch": 1970, "lr": 2.7924331680311518e-05}
{"train_loss": 0.11710869520902634, "global_step": 175414, "epoch": 1970, "lr": 2.792381151041925e-05}
{"train_loss": 0.06413163989782333, "global_step": 175415, "epoch": 1970, "lr": 2.7923291343494872e-05}
{"train_loss": 0.03452908992767334, "global_step": 175416, "epoch": 1970, "lr": 2.7922771179538422e-05}
{"train_loss": 0.10315394401550293, "global_step": 175417, "epoch": 1970, "lr": 2.7922251018550006e-05}
{"train_loss": 0.07859232575957024, "global_step": 175418, "epoch": 1970, "lr": 2.792173086052966e-05, "val_loss": 7.106244087219238, "train_action_mse_error": 14.553640365600586}
{"train_loss": 0.0689844861626625, "global_step": 175419, "epoch": 1971, "lr": 2.7921210705477462e-05}
{"train_loss": 0.07887767255306244, "global_step": 175420, "epoch": 1971, "lr": 2.7920690553393507e-05}
{"train_loss": 0.08473078906536102, "global_step": 175421, "epoch": 1971, "lr": 2.7920170404277833e-05}
{"train_loss": 0.1534125804901123, "global_step": 175422, "epoch": 1971, "lr": 2.7919650258130536e-05}
{"train_loss": 0.1394917368888855, "global_step": 175423, "epoch": 1971, "lr": 2.791913011495165e-05}
{"train_loss": 0.08232881873846054, "global_step": 175424, "epoch": 1971, "lr": 2.79186099747413e-05}
{"train_loss": 0.07426310330629349, "global_step": 175425, "epoch": 1971, "lr": 2.7918089837499496e-05}
{"train_loss": 0.10486534982919693, "global_step": 175426, "epoch": 1971, "lr": 2.7917569703226354e-05}
{"train_loss": 0.07695124298334122, "global_step": 175427, "epoch": 1971, "lr": 2.791704957192191e-05}
{"train_loss": 0.15977701544761658, "global_step": 175428, "epoch": 1971, "lr": 2.791652944358627e-05}
{"train_loss": 0.07264476269483566, "global_step": 175429, "epoch": 1971, "lr": 2.791600931821946e-05}
{"train_loss": 0.05086385831236839, "global_step": 175430, "epoch": 1971, "lr": 2.7915489195821597e-05}
{"train_loss": 0.06053634732961655, "global_step": 175431, "epoch": 1971, "lr": 2.7914969076392704e-05}
{"train_loss": 0.07406466454267502, "global_step": 175432, "epoch": 1971, "lr": 2.7914448959932894e-05}
{"train_loss": 0.09483025223016739, "global_step": 175433, "epoch": 1971, "lr": 2.79139288464422e-05}
{"train_loss": 0.07386478036642075, "global_step": 175434, "epoch": 1971, "lr": 2.7913408735920725e-05}
{"train_loss": 0.09146131575107574, "global_step": 175435, "epoch": 1971, "lr": 2.7912888628368504e-05}
{"train_loss": 0.09301643818616867, "global_step": 175436, "epoch": 1971, "lr": 2.7912368523785625e-05}
{"train_loss": 0.07202743738889694, "global_step": 175437, "epoch": 1971, "lr": 2.7911848422172182e-05}
{"train_loss": 0.10581020265817642, "global_step": 175438, "epoch": 1971, "lr": 2.7911328323528224e-05}
{"train_loss": 0.059389229863882065, "global_step": 175439, "epoch": 1971, "lr": 2.7910808227853796e-05}
{"train_loss": 0.07026302069425583, "global_step": 175440, "epoch": 1971, "lr": 2.791028813514901e-05}
{"train_loss": 0.11513913422822952, "global_step": 175441, "epoch": 1971, "lr": 2.79097680454139e-05}
{"train_loss": 0.06020289286971092, "global_step": 175442, "epoch": 1971, "lr": 2.7909247958648565e-05}
{"train_loss": 0.08418354392051697, "global_step": 175443, "epoch": 1971, "lr": 2.7908727874853046e-05}
{"train_loss": 0.07328084856271744, "global_step": 175444, "epoch": 1971, "lr": 2.7908207794027452e-05}
{"train_loss": 0.05787079036235809, "global_step": 175445, "epoch": 1971, "lr": 2.7907687716171827e-05}
{"train_loss": 0.06094092130661011, "global_step": 175446, "epoch": 1971, "lr": 2.7907167641286224e-05}
{"train_loss": 0.09118828922510147, "global_step": 175447, "epoch": 1971, "lr": 2.790664756937076e-05}
{"train_loss": 0.07502568513154984, "global_step": 175448, "epoch": 1971, "lr": 2.790612750042545e-05}
{"train_loss": 0.06374569982290268, "global_step": 175449, "epoch": 1971, "lr": 2.790560743445041e-05}
{"train_loss": 0.06460551917552948, "global_step": 175450, "epoch": 1971, "lr": 2.790508737144568e-05}
{"train_loss": 0.05337174981832504, "global_step": 175451, "epoch": 1971, "lr": 2.7904567311411355e-05}
{"train_loss": 0.09219484031200409, "global_step": 175452, "epoch": 1971, "lr": 2.7904047254347475e-05}
{"train_loss": 0.044661831110715866, "global_step": 175453, "epoch": 1971, "lr": 2.7903527200254142e-05}
{"train_loss": 0.07445438206195831, "global_step": 175454, "epoch": 1971, "lr": 2.79030071491314e-05}
{"train_loss": 0.11398792266845703, "global_step": 175455, "epoch": 1971, "lr": 2.7902487100979325e-05}
{"train_loss": 0.07641668617725372, "global_step": 175456, "epoch": 1971, "lr": 2.7901967055798005e-05}
{"train_loss": 0.08328883349895477, "global_step": 175457, "epoch": 1971, "lr": 2.7901447013587485e-05}
{"train_loss": 0.09765659272670746, "global_step": 175458, "epoch": 1971, "lr": 2.7900926974347864e-05}
{"train_loss": 0.10394886136054993, "global_step": 175459, "epoch": 1971, "lr": 2.790040693807917e-05}
{"train_loss": 0.1127023994922638, "global_step": 175460, "epoch": 1971, "lr": 2.7899886904781514e-05}
{"train_loss": 0.08720757812261581, "global_step": 175461, "epoch": 1971, "lr": 2.7899366874454934e-05}
{"train_loss": 0.0418548509478569, "global_step": 175462, "epoch": 1971, "lr": 2.7898846847099535e-05}
{"train_loss": 0.1006353348493576, "global_step": 175463, "epoch": 1971, "lr": 2.7898326822715348e-05}
{"train_loss": 0.04821542650461197, "global_step": 175464, "epoch": 1971, "lr": 2.7897806801302484e-05}
{"train_loss": 0.05217345058917999, "global_step": 175465, "epoch": 1971, "lr": 2.7897286782860965e-05}
{"train_loss": 0.101164810359478, "global_step": 175466, "epoch": 1971, "lr": 2.7896766767390902e-05}
{"train_loss": 0.0907798707485199, "global_step": 175467, "epoch": 1971, "lr": 2.7896246754892342e-05}
{"train_loss": 0.10664516687393188, "global_step": 175468, "epoch": 1971, "lr": 2.789572674536537e-05}
{"train_loss": 0.05340181291103363, "global_step": 175469, "epoch": 1971, "lr": 2.7895206738810038e-05}
{"train_loss": 0.10477592796087265, "global_step": 175470, "epoch": 1971, "lr": 2.789468673522644e-05}
{"train_loss": 0.09530078619718552, "global_step": 175471, "epoch": 1971, "lr": 2.7894166734614614e-05}
{"train_loss": 0.0717552900314331, "global_step": 175472, "epoch": 1971, "lr": 2.7893646736974642e-05}
{"train_loss": 0.061540231108665466, "global_step": 175473, "epoch": 1971, "lr": 2.789312674230663e-05}
{"train_loss": 0.11246360838413239, "global_step": 175474, "epoch": 1971, "lr": 2.7892606750610594e-05}
{"train_loss": 0.07483335584402084, "global_step": 175475, "epoch": 1971, "lr": 2.7892086761886643e-05}
{"train_loss": 0.0998162105679512, "global_step": 175476, "epoch": 1971, "lr": 2.7891566776134815e-05}
{"train_loss": 0.12118330597877502, "global_step": 175477, "epoch": 1971, "lr": 2.7891046793355213e-05}
{"train_loss": 0.06136400252580643, "global_step": 175478, "epoch": 1971, "lr": 2.7890526813547896e-05}
{"train_loss": 0.07620401680469513, "global_step": 175479, "epoch": 1971, "lr": 2.7890006836712907e-05}
{"train_loss": 0.06413311511278152, "global_step": 175480, "epoch": 1971, "lr": 2.7889486862850355e-05}
{"train_loss": 0.06374800950288773, "global_step": 175481, "epoch": 1971, "lr": 2.7888966891960273e-05}
{"train_loss": 0.05514722317457199, "global_step": 175482, "epoch": 1971, "lr": 2.7888446924042765e-05}
{"train_loss": 0.06785035878419876, "global_step": 175483, "epoch": 1971, "lr": 2.78879269590979e-05}
{"train_loss": 0.08641611784696579, "global_step": 175484, "epoch": 1971, "lr": 2.7887406997125698e-05}
{"train_loss": 0.10904347151517868, "global_step": 175485, "epoch": 1971, "lr": 2.788688703812629e-05}
{"train_loss": 0.06266359984874725, "global_step": 175486, "epoch": 1971, "lr": 2.7886367082099706e-05}
{"train_loss": 0.09521739929914474, "global_step": 175487, "epoch": 1971, "lr": 2.7885847129046037e-05}
{"train_loss": 0.12202230840921402, "global_step": 175488, "epoch": 1971, "lr": 2.788532717896534e-05}
{"train_loss": 0.04603300243616104, "global_step": 175489, "epoch": 1971, "lr": 2.7884807231857705e-05}
{"train_loss": 0.12397211045026779, "global_step": 175490, "epoch": 1971, "lr": 2.788428728772317e-05}
{"train_loss": 0.03465709090232849, "global_step": 175491, "epoch": 1971, "lr": 2.7883767346561825e-05}
{"train_loss": 0.107515349984169, "global_step": 175492, "epoch": 1971, "lr": 2.7883247408373753e-05}
{"train_loss": 0.11299882084131241, "global_step": 175493, "epoch": 1971, "lr": 2.788272747315899e-05}
{"train_loss": 0.08980867266654968, "global_step": 175494, "epoch": 1971, "lr": 2.7882207540917647e-05}
{"train_loss": 0.09383877366781235, "global_step": 175495, "epoch": 1971, "lr": 2.7881687611649754e-05}
{"train_loss": 0.10021965205669403, "global_step": 175496, "epoch": 1971, "lr": 2.7881167685355415e-05}
{"train_loss": 0.09625215828418732, "global_step": 175497, "epoch": 1971, "lr": 2.7880647762034662e-05}
{"train_loss": 0.07414012402296066, "global_step": 175498, "epoch": 1971, "lr": 2.788012784168761e-05}
{"train_loss": 0.10069672018289566, "global_step": 175499, "epoch": 1971, "lr": 2.7879607924314277e-05}
{"train_loss": 0.04478367418050766, "global_step": 175500, "epoch": 1971, "lr": 2.7879088009914788e-05}
{"train_loss": 0.055417828261852264, "global_step": 175501, "epoch": 1971, "lr": 2.7878568098489166e-05}
{"train_loss": 0.04622047767043114, "global_step": 175502, "epoch": 1971, "lr": 2.787804819003752e-05}
{"train_loss": 0.09556086361408234, "global_step": 175503, "epoch": 1971, "lr": 2.7877528284559884e-05}
{"train_loss": 0.10052889585494995, "global_step": 175504, "epoch": 1971, "lr": 2.787700838205636e-05}
{"train_loss": 0.06425075978040695, "global_step": 175505, "epoch": 1971, "lr": 2.7876488482526986e-05}
{"train_loss": 0.08836144208908081, "global_step": 175506, "epoch": 1971, "lr": 2.787596858597186e-05}
{"train_loss": 0.08282838029305586, "global_step": 175507, "epoch": 1971, "lr": 2.7875448692391025e-05, "val_loss": 7.443979740142822}
{"train_loss": 0.07017021626234055, "global_step": 175508, "epoch": 1972, "lr": 2.7874928801784572e-05}
{"train_loss": 0.05383319407701492, "global_step": 175509, "epoch": 1972, "lr": 2.7874408914152584e-05}
{"train_loss": 0.06133468076586723, "global_step": 175510, "epoch": 1972, "lr": 2.7873889029495087e-05}
{"train_loss": 0.06749214977025986, "global_step": 175511, "epoch": 1972, "lr": 2.78733691478122e-05}
{"train_loss": 0.09317746758460999, "global_step": 175512, "epoch": 1972, "lr": 2.7872849269103945e-05}
{"train_loss": 0.06909159570932388, "global_step": 175513, "epoch": 1972, "lr": 2.787232939337044e-05}
{"train_loss": 0.08255022019147873, "global_step": 175514, "epoch": 1972, "lr": 2.7871809520611702e-05}
{"train_loss": 0.05469411984086037, "global_step": 175515, "epoch": 1972, "lr": 2.787128965082786e-05}
{"train_loss": 0.12611064314842224, "global_step": 175516, "epoch": 1972, "lr": 2.7870769784018947e-05}
{"train_loss": 0.1711830496788025, "global_step": 175517, "epoch": 1972, "lr": 2.787024992018502e-05}
{"train_loss": 0.10057366639375687, "global_step": 175518, "epoch": 1972, "lr": 2.7869730059326195e-05}
{"train_loss": 0.07843636721372604, "global_step": 175519, "epoch": 1972, "lr": 2.7869210201442485e-05}
{"train_loss": 0.0810624212026596, "global_step": 175520, "epoch": 1972, "lr": 2.7868690346534022e-05}
{"train_loss": 0.08513825386762619, "global_step": 175521, "epoch": 1972, "lr": 2.786817049460083e-05}
{"train_loss": 0.09201384335756302, "global_step": 175522, "epoch": 1972, "lr": 2.7867650645642984e-05}
{"train_loss": 0.10090965032577515, "global_step": 175523, "epoch": 1972, "lr": 2.7867130799660578e-05}
{"train_loss": 0.07094158977270126, "global_step": 175524, "epoch": 1972, "lr": 2.7866610956653643e-05}
{"train_loss": 0.10489978641271591, "global_step": 175525, "epoch": 1972, "lr": 2.7866091116622294e-05}
{"train_loss": 0.09994151443243027, "global_step": 175526, "epoch": 1972, "lr": 2.7865571279566562e-05}
{"train_loss": 0.08247847855091095, "global_step": 175527, "epoch": 1972, "lr": 2.7865051445486534e-05}
{"train_loss": 0.07364651560783386, "global_step": 175528, "epoch": 1972, "lr": 2.78645316143823e-05}
{"train_loss": 0.12023310363292694, "global_step": 175529, "epoch": 1972, "lr": 2.786401178625389e-05}
{"train_loss": 0.07533781975507736, "global_step": 175530, "epoch": 1972, "lr": 2.786349196110141e-05}
{"train_loss": 0.09054189920425415, "global_step": 175531, "epoch": 1972, "lr": 2.7862972138924903e-05}
{"train_loss": 0.10724782943725586, "global_step": 175532, "epoch": 1972, "lr": 2.7862452319724463e-05}
{"train_loss": 0.1377708613872528, "global_step": 175533, "epoch": 1972, "lr": 2.786193250350012e-05}
{"train_loss": 0.06080104038119316, "global_step": 175534, "epoch": 1972, "lr": 2.7861412690251997e-05}
{"train_loss": 0.08434893190860748, "global_step": 175535, "epoch": 1972, "lr": 2.786089287998012e-05}
{"train_loss": 0.07853290438652039, "global_step": 175536, "epoch": 1972, "lr": 2.7860373072684596e-05}
{"train_loss": 0.04936195909976959, "global_step": 175537, "epoch": 1972, "lr": 2.7859853268365454e-05}
{"train_loss": 0.08701612055301666, "global_step": 175538, "epoch": 1972, "lr": 2.7859333467022796e-05}
{"train_loss": 0.07769433408975601, "global_step": 175539, "epoch": 1972, "lr": 2.785881366865667e-05}
{"train_loss": 0.08344865590333939, "global_step": 175540, "epoch": 1972, "lr": 2.785829387326718e-05}
{"train_loss": 0.05477527901530266, "global_step": 175541, "epoch": 1972, "lr": 2.7857774080854348e-05}
{"train_loss": 0.14320531487464905, "global_step": 175542, "epoch": 1972, "lr": 2.7857254291418283e-05}
{"train_loss": 0.08580733090639114, "global_step": 175543, "epoch": 1972, "lr": 2.7856734504959027e-05}
{"train_loss": 0.13054579496383667, "global_step": 175544, "epoch": 1972, "lr": 2.7856214721476682e-05}
{"train_loss": 0.05542251467704773, "global_step": 175545, "epoch": 1972, "lr": 2.7855694940971278e-05}
{"train_loss": 0.09624329209327698, "global_step": 175546, "epoch": 1972, "lr": 2.7855175163442904e-05}
{"train_loss": 0.06755074113607407, "global_step": 175547, "epoch": 1972, "lr": 2.7854655388891654e-05}
{"train_loss": 0.10751977562904358, "global_step": 175548, "epoch": 1972, "lr": 2.7854135617317555e-05}
{"train_loss": 0.12987835705280304, "global_step": 175549, "epoch": 1972, "lr": 2.7853615848720716e-05}
{"train_loss": 0.1170140877366066, "global_step": 175550, "epoch": 1972, "lr": 2.785309608310117e-05}
{"train_loss": 0.04963218420743942, "global_step": 175551, "epoch": 1972, "lr": 2.7852576320459028e-05}
{"train_loss": 0.12485366314649582, "global_step": 175552, "epoch": 1972, "lr": 2.7852056560794314e-05}
{"train_loss": 0.03691110387444496, "global_step": 175553, "epoch": 1972, "lr": 2.7851536804107148e-05}
{"train_loss": 0.09459701180458069, "global_step": 175554, "epoch": 1972, "lr": 2.7851017050397544e-05}
{"train_loss": 0.06767494231462479, "global_step": 175555, "epoch": 1972, "lr": 2.7850497299665622e-05}
{"train_loss": 0.028783857822418213, "global_step": 175556, "epoch": 1972, "lr": 2.7849977551911432e-05}
{"train_loss": 0.119057297706604, "global_step": 175557, "epoch": 1972, "lr": 2.7849457807135026e-05}
{"train_loss": 0.08612557500600815, "global_step": 175558, "epoch": 1972, "lr": 2.7848938065336504e-05}
{"train_loss": 0.06167994439601898, "global_step": 175559, "epoch": 1972, "lr": 2.7848418326515906e-05}
{"train_loss": 0.11743830889463425, "global_step": 175560, "epoch": 1972, "lr": 2.784789859067334e-05}
{"train_loss": 0.050774406641721725, "global_step": 175561, "epoch": 1972, "lr": 2.7847378857808846e-05}
{"train_loss": 0.05020485445857048, "global_step": 175562, "epoch": 1972, "lr": 2.784685912792249e-05}
{"train_loss": 0.10690140724182129, "global_step": 175563, "epoch": 1972, "lr": 2.7846339401014354e-05}
{"train_loss": 0.04181811213493347, "global_step": 175564, "epoch": 1972, "lr": 2.7845819677084527e-05}
{"train_loss": 0.0819694921374321, "global_step": 175565, "epoch": 1972, "lr": 2.7845299956133037e-05}
{"train_loss": 0.0526844784617424, "global_step": 175566, "epoch": 1972, "lr": 2.7844780238159996e-05}
{"train_loss": 0.06763894855976105, "global_step": 175567, "epoch": 1972, "lr": 2.7844260523165432e-05}
{"train_loss": 0.0884736031293869, "global_step": 175568, "epoch": 1972, "lr": 2.7843740811149466e-05}
{"train_loss": 0.07439584285020828, "global_step": 175569, "epoch": 1972, "lr": 2.784322110211211e-05}
{"train_loss": 0.07596974819898605, "global_step": 175570, "epoch": 1972, "lr": 2.784270139605349e-05}
{"train_loss": 0.11426378786563873, "global_step": 175571, "epoch": 1972, "lr": 2.7842181692973623e-05}
{"train_loss": 0.08270206302404404, "global_step": 175572, "epoch": 1972, "lr": 2.7841661992872625e-05}
{"train_loss": 0.10462476313114166, "global_step": 175573, "epoch": 1972, "lr": 2.7841142295750534e-05}
{"train_loss": 0.0668412372469902, "global_step": 175574, "epoch": 1972, "lr": 2.7840622601607437e-05}
{"train_loss": 0.08603009581565857, "global_step": 175575, "epoch": 1972, "lr": 2.784010291044339e-05}
{"train_loss": 0.0672152116894722, "global_step": 175576, "epoch": 1972, "lr": 2.7839583222258488e-05}
{"train_loss": 0.06985367089509964, "global_step": 175577, "epoch": 1972, "lr": 2.7839063537052758e-05}
{"train_loss": 0.0983208566904068, "global_step": 175578, "epoch": 1972, "lr": 2.7838543854826327e-05}
{"train_loss": 0.04151901230216026, "global_step": 175579, "epoch": 1972, "lr": 2.7838024175579202e-05}
{"train_loss": 0.06775546073913574, "global_step": 175580, "epoch": 1972, "lr": 2.7837504499311513e-05}
{"train_loss": 0.08866581320762634, "global_step": 175581, "epoch": 1972, "lr": 2.783698482602328e-05}
{"train_loss": 0.08191481232643127, "global_step": 175582, "epoch": 1972, "lr": 2.7836465155714588e-05}
{"train_loss": 0.04085664823651314, "global_step": 175583, "epoch": 1972, "lr": 2.783594548838554e-05}
{"train_loss": 0.051820844411849976, "global_step": 175584, "epoch": 1972, "lr": 2.7835425824036155e-05}
{"train_loss": 0.05633704364299774, "global_step": 175585, "epoch": 1972, "lr": 2.7834906162666545e-05}
{"train_loss": 0.09561260044574738, "global_step": 175586, "epoch": 1972, "lr": 2.7834386504276743e-05}
{"train_loss": 0.07481447607278824, "global_step": 175587, "epoch": 1972, "lr": 2.7833866848866863e-05}
{"train_loss": 0.04284778609871864, "global_step": 175588, "epoch": 1972, "lr": 2.7833347196436927e-05}
{"train_loss": 0.17773611843585968, "global_step": 175589, "epoch": 1972, "lr": 2.783282754698705e-05}
{"train_loss": 0.09382330626249313, "global_step": 175590, "epoch": 1972, "lr": 2.783230790051725e-05}
{"train_loss": 0.14165844023227692, "global_step": 175591, "epoch": 1972, "lr": 2.783178825702766e-05}
{"train_loss": 0.08477011322975159, "global_step": 175592, "epoch": 1972, "lr": 2.783126861651828e-05}
{"train_loss": 0.06063966825604439, "global_step": 175593, "epoch": 1972, "lr": 2.7830748978989244e-05}
{"train_loss": 0.0776362195611, "global_step": 175594, "epoch": 1972, "lr": 2.7830229344440594e-05}
{"train_loss": 0.05953052639961243, "global_step": 175595, "epoch": 1972, "lr": 2.7829709712872375e-05}
{"train_loss": 0.08343495035104537, "global_step": 175596, "epoch": 1972, "lr": 2.7829190084284705e-05, "val_loss": 7.3923659324646}
{"train_loss": 0.11369871348142624, "global_step": 175597, "epoch": 1973, "lr": 2.782867045867761e-05}
{"train_loss": 0.12902463972568512, "global_step": 175598, "epoch": 1973, "lr": 2.7828150836051205e-05}
{"train_loss": 0.04669777676463127, "global_step": 175599, "epoch": 1973, "lr": 2.7827631216405503e-05}
{"train_loss": 0.10150045156478882, "global_step": 175600, "epoch": 1973, "lr": 2.7827111599740635e-05}
{"train_loss": 0.08523967862129211, "global_step": 175601, "epoch": 1973, "lr": 2.7826591986056615e-05}
{"train_loss": 0.056653980165719986, "global_step": 175602, "epoch": 1973, "lr": 2.7826072375353568e-05}
{"train_loss": 0.13487163186073303, "global_step": 175603, "epoch": 1973, "lr": 2.782555276763151e-05}
{"train_loss": 0.0875461995601654, "global_step": 175604, "epoch": 1973, "lr": 2.7825033162890557e-05}
{"train_loss": 0.08607006072998047, "global_step": 175605, "epoch": 1973, "lr": 2.782451356113074e-05}
{"train_loss": 0.10904714465141296, "global_step": 175606, "epoch": 1973, "lr": 2.7823993962352164e-05}
{"train_loss": 0.07152164727449417, "global_step": 175607, "epoch": 1973, "lr": 2.7823474366554857e-05}
{"train_loss": 0.11066709458827972, "global_step": 175608, "epoch": 1973, "lr": 2.782295477373894e-05}
{"train_loss": 0.10523394495248795, "global_step": 175609, "epoch": 1973, "lr": 2.7822435183904438e-05}
{"train_loss": 0.09133238345384598, "global_step": 175610, "epoch": 1973, "lr": 2.782191559705146e-05}
{"train_loss": 0.08757659047842026, "global_step": 175611, "epoch": 1973, "lr": 2.782139601318003e-05}
{"train_loss": 0.09220991283655167, "global_step": 175612, "epoch": 1973, "lr": 2.7820876432290265e-05}
{"train_loss": 0.0465291365981102, "global_step": 175613, "epoch": 1973, "lr": 2.782035685438219e-05}
{"train_loss": 0.04576025903224945, "global_step": 175614, "epoch": 1973, "lr": 2.7819837279455924e-05}
{"train_loss": 0.08185990154743195, "global_step": 175615, "epoch": 1973, "lr": 2.7819317707511482e-05}
{"train_loss": 0.1392400562763214, "global_step": 175616, "epoch": 1973, "lr": 2.781879813854899e-05}
{"train_loss": 0.09687342494726181, "global_step": 175617, "epoch": 1973, "lr": 2.781827857256847e-05}
{"train_loss": 0.07681635767221451, "global_step": 175618, "epoch": 1973, "lr": 2.781775900957001e-05}
{"train_loss": 0.061034660786390305, "global_step": 175619, "epoch": 1973, "lr": 2.78172394495537e-05}
{"train_loss": 0.028267666697502136, "global_step": 175620, "epoch": 1973, "lr": 2.781671989251957e-05}
{"train_loss": 0.08317182213068008, "global_step": 175621, "epoch": 1973, "lr": 2.7816200338467735e-05}
{"train_loss": 0.06970828771591187, "global_step": 175622, "epoch": 1973, "lr": 2.7815680787398223e-05}
{"train_loss": 0.11112614721059799, "global_step": 175623, "epoch": 1973, "lr": 2.7815161239311145e-05}
{"train_loss": 0.10018692165613174, "global_step": 175624, "epoch": 1973, "lr": 2.781464169420652e-05}
{"train_loss": 0.04197089001536369, "global_step": 175625, "epoch": 1973, "lr": 2.7814122152084476e-05}
{"train_loss": 0.028399139642715454, "global_step": 175626, "epoch": 1973, "lr": 2.781360261294502e-05}
{"train_loss": 0.13366848230361938, "global_step": 175627, "epoch": 1973, "lr": 2.7813083076788283e-05}
{"train_loss": 0.15551982820034027, "global_step": 175628, "epoch": 1973, "lr": 2.781256354361429e-05}
{"train_loss": 0.10158126801252365, "global_step": 175629, "epoch": 1973, "lr": 2.781204401342314e-05}
{"train_loss": 0.06453191488981247, "global_step": 175630, "epoch": 1973, "lr": 2.7811524486214874e-05}
{"train_loss": 0.15719060599803925, "global_step": 175631, "epoch": 1973, "lr": 2.7811004961989595e-05}
{"train_loss": 0.11564742773771286, "global_step": 175632, "epoch": 1973, "lr": 2.781048544074736e-05}
{"train_loss": 0.1425691395998001, "global_step": 175633, "epoch": 1973, "lr": 2.780996592248821e-05}
{"train_loss": 0.0513368621468544, "global_step": 175634, "epoch": 1973, "lr": 2.7809446407212263e-05}
{"train_loss": 0.05291486904025078, "global_step": 175635, "epoch": 1973, "lr": 2.7808926894919547e-05}
{"train_loss": 0.11351344734430313, "global_step": 175636, "epoch": 1973, "lr": 2.780840738561017e-05}
{"train_loss": 0.10540729761123657, "global_step": 175637, "epoch": 1973, "lr": 2.7807887879284157e-05}
{"train_loss": 0.10104353725910187, "global_step": 175638, "epoch": 1973, "lr": 2.780736837594162e-05}
{"train_loss": 0.11978352069854736, "global_step": 175639, "epoch": 1973, "lr": 2.7806848875582603e-05}
{"train_loss": 0.06256181001663208, "global_step": 175640, "epoch": 1973, "lr": 2.7806329378207195e-05}
{"train_loss": 0.08040519803762436, "global_step": 175641, "epoch": 1973, "lr": 2.7805809883815446e-05}
{"train_loss": 0.08583489060401917, "global_step": 175642, "epoch": 1973, "lr": 2.7805290392407446e-05}
{"train_loss": 0.12046544998884201, "global_step": 175643, "epoch": 1973, "lr": 2.780477090398324e-05}
{"train_loss": 0.09749054908752441, "global_step": 175644, "epoch": 1973, "lr": 2.780425141854293e-05}
{"train_loss": 0.09174320101737976, "global_step": 175645, "epoch": 1973, "lr": 2.7803731936086553e-05}
{"train_loss": 0.053877901285886765, "global_step": 175646, "epoch": 1973, "lr": 2.7803212456614203e-05}
{"train_loss": 0.08281569182872772, "global_step": 175647, "epoch": 1973, "lr": 2.780269298012593e-05}
{"train_loss": 0.06401996314525604, "global_step": 175648, "epoch": 1973, "lr": 2.7802173506621826e-05}
{"train_loss": 0.08401510119438171, "global_step": 175649, "epoch": 1973, "lr": 2.7801654036101933e-05}
{"train_loss": 0.0613718144595623, "global_step": 175650, "epoch": 1973, "lr": 2.7801134568566356e-05}
{"train_loss": 0.17352628707885742, "global_step": 175651, "epoch": 1973, "lr": 2.7800615104015128e-05}
{"train_loss": 0.06353696435689926, "global_step": 175652, "epoch": 1973, "lr": 2.780009564244836e-05}
{"train_loss": 0.10471388697624207, "global_step": 175653, "epoch": 1973, "lr": 2.779957618386607e-05}
{"train_loss": 0.10090932995080948, "global_step": 175654, "epoch": 1973, "lr": 2.7799056728268363e-05}
{"train_loss": 0.06975524127483368, "global_step": 175655, "epoch": 1973, "lr": 2.7798537275655322e-05}
{"train_loss": 0.1097528487443924, "global_step": 175656, "epoch": 1973, "lr": 2.7798017826026978e-05}
{"train_loss": 0.09784624725580215, "global_step": 175657, "epoch": 1973, "lr": 2.7797498379383435e-05}
{"train_loss": 0.08667616546154022, "global_step": 175658, "epoch": 1973, "lr": 2.7796978935724725e-05}
{"train_loss": 0.09819334745407104, "global_step": 175659, "epoch": 1973, "lr": 2.779645949505097e-05}
{"train_loss": 0.05925453081727028, "global_step": 175660, "epoch": 1973, "lr": 2.7795940057362185e-05}
{"train_loss": 0.06372450292110443, "global_step": 175661, "epoch": 1973, "lr": 2.7795420622658486e-05}
{"train_loss": 0.09070485085248947, "global_step": 175662, "epoch": 1973, "lr": 2.7794901190939905e-05}
{"train_loss": 0.0851437970995903, "global_step": 175663, "epoch": 1973, "lr": 2.7794381762206546e-05}
{"train_loss": 0.07545655965805054, "global_step": 175664, "epoch": 1973, "lr": 2.7793862336458442e-05}
{"train_loss": 0.05213717743754387, "global_step": 175665, "epoch": 1973, "lr": 2.7793342913695703e-05}
{"train_loss": 0.09750135987997055, "global_step": 175666, "epoch": 1973, "lr": 2.7792823493918358e-05}
{"train_loss": 0.0948415994644165, "global_step": 175667, "epoch": 1973, "lr": 2.779230407712652e-05}
{"train_loss": 0.10058266669511795, "global_step": 175668, "epoch": 1973, "lr": 2.779178466332021e-05}
{"train_loss": 0.07997205853462219, "global_step": 175669, "epoch": 1973, "lr": 2.7791265252499543e-05}
{"train_loss": 0.07838473469018936, "global_step": 175670, "epoch": 1973, "lr": 2.7790745844664556e-05}
{"train_loss": 0.10413753986358643, "global_step": 175671, "epoch": 1973, "lr": 2.7790226439815347e-05}
{"train_loss": 0.10555140674114227, "global_step": 175672, "epoch": 1973, "lr": 2.7789707037951977e-05}
{"train_loss": 0.10126436501741409, "global_step": 175673, "epoch": 1973, "lr": 2.7789187639074486e-05}
{"train_loss": 0.05894675850868225, "global_step": 175674, "epoch": 1973, "lr": 2.7788668243182993e-05}
{"train_loss": 0.05562618374824524, "global_step": 175675, "epoch": 1973, "lr": 2.7788148850277518e-05}
{"train_loss": 0.06613879650831223, "global_step": 175676, "epoch": 1973, "lr": 2.778762946035818e-05}
{"train_loss": 0.14893141388893127, "global_step": 175677, "epoch": 1973, "lr": 2.7787110073425005e-05}
{"train_loss": 0.059950090944767, "global_step": 175678, "epoch": 1973, "lr": 2.77865906894781e-05}
{"train_loss": 0.06711927056312561, "global_step": 175679, "epoch": 1973, "lr": 2.7786071308517493e-05}
{"train_loss": 0.11911796778440475, "global_step": 175680, "epoch": 1973, "lr": 2.7785551930543297e-05}
{"train_loss": 0.05735606327652931, "global_step": 175681, "epoch": 1973, "lr": 2.7785032555555547e-05}
{"train_loss": 0.0616757906973362, "global_step": 175682, "epoch": 1973, "lr": 2.778451318355435e-05}
{"train_loss": 0.1155642718076706, "global_step": 175683, "epoch": 1973, "lr": 2.778399381453973e-05}
{"train_loss": 0.1353251188993454, "global_step": 175684, "epoch": 1973, "lr": 2.7783474448511804e-05}
{"train_loss": 0.08873244722405177, "global_step": 175685, "epoch": 1973, "lr": 2.7782955085470603e-05, "val_loss": 7.097951412200928}
{"train_loss": 0.03824487701058388, "global_step": 175686, "epoch": 1974, "lr": 2.7782435725416227e-05}
{"train_loss": 0.06540226191282272, "global_step": 175687, "epoch": 1974, "lr": 2.7781916368348716e-05}
{"train_loss": 0.07158695161342621, "global_step": 175688, "epoch": 1974, "lr": 2.7781397014268175e-05}
{"train_loss": 0.06802448630332947, "global_step": 175689, "epoch": 1974, "lr": 2.778087766317463e-05}
{"train_loss": 0.04967093467712402, "global_step": 175690, "epoch": 1974, "lr": 2.7780358315068177e-05}
{"train_loss": 0.06937571614980698, "global_step": 175691, "epoch": 1974, "lr": 2.7779838969948905e-05}
{"train_loss": 0.06257365643978119, "global_step": 175692, "epoch": 1974, "lr": 2.7779319627816845e-05}
{"train_loss": 0.1123824268579483, "global_step": 175693, "epoch": 1974, "lr": 2.7778800288672104e-05}
{"train_loss": 0.11305953562259674, "global_step": 175694, "epoch": 1974, "lr": 2.7778280952514712e-05}
{"train_loss": 0.051660653203725815, "global_step": 175695, "epoch": 1974, "lr": 2.777776161934478e-05}
{"train_loss": 0.11041467636823654, "global_step": 175696, "epoch": 1974, "lr": 2.777724228916233e-05}
{"train_loss": 0.08920735865831375, "global_step": 175697, "epoch": 1974, "lr": 2.7776722961967483e-05}
{"train_loss": 0.11226452142000198, "global_step": 175698, "epoch": 1974, "lr": 2.7776203637760266e-05}
{"train_loss": 0.07851768285036087, "global_step": 175699, "epoch": 1974, "lr": 2.777568431654079e-05}
{"train_loss": 0.08986461162567139, "global_step": 175700, "epoch": 1974, "lr": 2.7775164998309077e-05}
{"train_loss": 0.07608362287282944, "global_step": 175701, "epoch": 1974, "lr": 2.7774645683065248e-05}
{"train_loss": 0.09322722256183624, "global_step": 175702, "epoch": 1974, "lr": 2.7774126370809327e-05}
{"train_loss": 0.11484652757644653, "global_step": 175703, "epoch": 1974, "lr": 2.7773607061541416e-05}
{"train_loss": 0.04113459587097168, "global_step": 175704, "epoch": 1974, "lr": 2.7773087755261564e-05}
{"train_loss": 0.08237660676240921, "global_step": 175705, "epoch": 1974, "lr": 2.777256845196986e-05}
{"train_loss": 0.08504550158977509, "global_step": 175706, "epoch": 1974, "lr": 2.7772049151666347e-05}
{"train_loss": 0.07842514663934708, "global_step": 175707, "epoch": 1974, "lr": 2.777152985435113e-05}
{"train_loss": 0.10524718463420868, "global_step": 175708, "epoch": 1974, "lr": 2.7771010560024235e-05}
{"train_loss": 0.05940544605255127, "global_step": 175709, "epoch": 1974, "lr": 2.7770491268685783e-05}
{"train_loss": 0.08813674002885818, "global_step": 175710, "epoch": 1974, "lr": 2.776997198033582e-05}
{"train_loss": 0.06744862347841263, "global_step": 175711, "epoch": 1974, "lr": 2.776945269497439e-05}
{"train_loss": 0.04880688712000847, "global_step": 175712, "epoch": 1974, "lr": 2.7768933412601605e-05}
{"train_loss": 0.08694210648536682, "global_step": 175713, "epoch": 1974, "lr": 2.7768414133217495e-05}
{"train_loss": 0.0713692456483841, "global_step": 175714, "epoch": 1974, "lr": 2.7767894856822174e-05}
{"train_loss": 0.11997241526842117, "global_step": 175715, "epoch": 1974, "lr": 2.7767375583415667e-05}
{"train_loss": 0.06973130255937576, "global_step": 175716, "epoch": 1974, "lr": 2.7766856312998084e-05}
{"train_loss": 0.055123474448919296, "global_step": 175717, "epoch": 1974, "lr": 2.7766337045569458e-05}
{"train_loss": 0.07781696319580078, "global_step": 175718, "epoch": 1974, "lr": 2.77658177811299e-05}
{"train_loss": 0.08790221810340881, "global_step": 175719, "epoch": 1974, "lr": 2.7765298519679434e-05}
{"train_loss": 0.0659557655453682, "global_step": 175720, "epoch": 1974, "lr": 2.7764779261218173e-05}
{"train_loss": 0.05832487344741821, "global_step": 175721, "epoch": 1974, "lr": 2.7764260005746144e-05}
{"train_loss": 0.1131490096449852, "global_step": 175722, "epoch": 1974, "lr": 2.7763740753263455e-05}
{"train_loss": 0.10785806179046631, "global_step": 175723, "epoch": 1974, "lr": 2.7763221503770155e-05}
{"train_loss": 0.07599585503339767, "global_step": 175724, "epoch": 1974, "lr": 2.7762702257266325e-05}
{"train_loss": 0.10038453340530396, "global_step": 175725, "epoch": 1974, "lr": 2.7762183013752014e-05}
{"train_loss": 0.10132304579019547, "global_step": 175726, "epoch": 1974, "lr": 2.776166377322731e-05}
{"train_loss": 0.08172649145126343, "global_step": 175727, "epoch": 1974, "lr": 2.776114453569229e-05}
{"train_loss": 0.06214801222085953, "global_step": 175728, "epoch": 1974, "lr": 2.7760625301146997e-05}
{"train_loss": 0.09261210262775421, "global_step": 175729, "epoch": 1974, "lr": 2.776010606959154e-05}
{"train_loss": 0.05101015791296959, "global_step": 175730, "epoch": 1974, "lr": 2.775958684102594e-05}
{"train_loss": 0.11492221057415009, "global_step": 175731, "epoch": 1974, "lr": 2.7759067615450317e-05}
{"train_loss": 0.074469655752182, "global_step": 175732, "epoch": 1974, "lr": 2.7758548392864696e-05}
{"train_loss": 0.13036221265792847, "global_step": 175733, "epoch": 1974, "lr": 2.7758029173269185e-05}
{"train_loss": 0.0776573196053505, "global_step": 175734, "epoch": 1974, "lr": 2.775750995666382e-05}
{"train_loss": 0.057167068123817444, "global_step": 175735, "epoch": 1974, "lr": 2.7756990743048704e-05}
{"train_loss": 0.061158593744039536, "global_step": 175736, "epoch": 1974, "lr": 2.7756471532423865e-05}
{"train_loss": 0.025410121306777, "global_step": 175737, "epoch": 1974, "lr": 2.7755952324789426e-05}
{"train_loss": 0.08906231820583344, "global_step": 175738, "epoch": 1974, "lr": 2.7755433120145402e-05}
{"train_loss": 0.08908020704984665, "global_step": 175739, "epoch": 1974, "lr": 2.775491391849191e-05}
{"train_loss": 0.10727265477180481, "global_step": 175740, "epoch": 1974, "lr": 2.775439471982898e-05}
{"train_loss": 0.04539806395769119, "global_step": 175741, "epoch": 1974, "lr": 2.7753875524156724e-05}
{"train_loss": 0.1177639290690422, "global_step": 175742, "epoch": 1974, "lr": 2.7753356331475167e-05}
{"train_loss": 0.0924747958779335, "global_step": 175743, "epoch": 1974, "lr": 2.7752837141784416e-05}
{"train_loss": 0.0722021609544754, "global_step": 175744, "epoch": 1974, "lr": 2.7752317955084507e-05}
{"train_loss": 0.06029370054602623, "global_step": 175745, "epoch": 1974, "lr": 2.7751798771375543e-05}
{"train_loss": 0.09505292773246765, "global_step": 175746, "epoch": 1974, "lr": 2.7751279590657564e-05}
{"train_loss": 0.09332045912742615, "global_step": 175747, "epoch": 1974, "lr": 2.7750760412930672e-05}
{"train_loss": 0.09355003386735916, "global_step": 175748, "epoch": 1974, "lr": 2.7750241238194917e-05}
{"train_loss": 0.1108027845621109, "global_step": 175749, "epoch": 1974, "lr": 2.7749722066450358e-05}
{"train_loss": 0.07404951751232147, "global_step": 175750, "epoch": 1974, "lr": 2.7749202897697092e-05}
{"train_loss": 0.1422233283519745, "global_step": 175751, "epoch": 1974, "lr": 2.7748683731935154e-05}
{"train_loss": 0.0926801785826683, "global_step": 175752, "epoch": 1974, "lr": 2.7748164569164658e-05}
{"train_loss": 0.10726433247327805, "global_step": 175753, "epoch": 1974, "lr": 2.7747645409385624e-05}
{"train_loss": 0.05973779410123825, "global_step": 175754, "epoch": 1974, "lr": 2.7747126252598176e-05}
{"train_loss": 0.05332322046160698, "global_step": 175755, "epoch": 1974, "lr": 2.7746607098802336e-05}
{"train_loss": 0.15871122479438782, "global_step": 175756, "epoch": 1974, "lr": 2.77460879479982e-05}
{"train_loss": 0.08555702120065689, "global_step": 175757, "epoch": 1974, "lr": 2.7745568800185828e-05}
{"train_loss": 0.10631447285413742, "global_step": 175758, "epoch": 1974, "lr": 2.7745049655365302e-05}
{"train_loss": 0.08350212126970291, "global_step": 175759, "epoch": 1974, "lr": 2.7744530513536666e-05}
{"train_loss": 0.13568325340747833, "global_step": 175760, "epoch": 1974, "lr": 2.774401137470003e-05}
{"train_loss": 0.11631274968385696, "global_step": 175761, "epoch": 1974, "lr": 2.7743492238855418e-05}
{"train_loss": 0.06297530978918076, "global_step": 175762, "epoch": 1974, "lr": 2.774297310600292e-05}
{"train_loss": 0.05232400447130203, "global_step": 175763, "epoch": 1974, "lr": 2.7742453976142624e-05}
{"train_loss": 0.07141689956188202, "global_step": 175764, "epoch": 1974, "lr": 2.7741934849274575e-05}
{"train_loss": 0.07185955345630646, "global_step": 175765, "epoch": 1974, "lr": 2.7741415725398866e-05}
{"train_loss": 0.07109496742486954, "global_step": 175766, "epoch": 1974, "lr": 2.7740896604515532e-05}
{"train_loss": 0.05109867453575134, "global_step": 175767, "epoch": 1974, "lr": 2.774037748662468e-05}
{"train_loss": 0.09208463132381439, "global_step": 175768, "epoch": 1974, "lr": 2.7739858371726345e-05}
{"train_loss": 0.07517066597938538, "global_step": 175769, "epoch": 1974, "lr": 2.7739339259820642e-05}
{"train_loss": 0.062083229422569275, "global_step": 175770, "epoch": 1974, "lr": 2.7738820150907584e-05}
{"train_loss": 0.0838467925786972, "global_step": 175771, "epoch": 1974, "lr": 2.7738301044987293e-05}
{"train_loss": 0.06651747226715088, "global_step": 175772, "epoch": 1974, "lr": 2.7737781942059793e-05}
{"train_loss": 0.07739861309528351, "global_step": 175773, "epoch": 1974, "lr": 2.7737262842125207e-05}
{"train_loss": 0.08235934809938575, "global_step": 175774, "epoch": 1974, "lr": 2.7736743745183547e-05, "val_loss": 7.029170036315918}
{"train_loss": 0.07467176765203476, "global_step": 175775, "epoch": 1975, "lr": 2.773622465123493e-05}
{"train_loss": 0.1153944656252861, "global_step": 175776, "epoch": 1975, "lr": 2.773570556027939e-05}
{"train_loss": 0.06600989401340485, "global_step": 175777, "epoch": 1975, "lr": 2.7735186472317033e-05}
{"train_loss": 0.05784250795841217, "global_step": 175778, "epoch": 1975, "lr": 2.7734667387347886e-05}
{"train_loss": 0.062238145619630814, "global_step": 175779, "epoch": 1975, "lr": 2.773414830537206e-05}
{"train_loss": 0.07588033378124237, "global_step": 175780, "epoch": 1975, "lr": 2.773362922638959e-05}
{"train_loss": 0.143274188041687, "global_step": 175781, "epoch": 1975, "lr": 2.7733110150400577e-05}
{"train_loss": 0.07066506892442703, "global_step": 175782, "epoch": 1975, "lr": 2.7732591077405064e-05}
{"train_loss": 0.038039032369852066, "global_step": 175783, "epoch": 1975, "lr": 2.7732072007403148e-05}
{"train_loss": 0.051356442272663116, "global_step": 175784, "epoch": 1975, "lr": 2.7731552940394866e-05}
{"train_loss": 0.08378173410892487, "global_step": 175785, "epoch": 1975, "lr": 2.7731033876380318e-05}
{"train_loss": 0.06686437875032425, "global_step": 175786, "epoch": 1975, "lr": 2.7730514815359553e-05}
{"train_loss": 0.05767063796520233, "global_step": 175787, "epoch": 1975, "lr": 2.772999575733266e-05}
{"train_loss": 0.06909636408090591, "global_step": 175788, "epoch": 1975, "lr": 2.77294767022997e-05}
{"train_loss": 0.04007313400506973, "global_step": 175789, "epoch": 1975, "lr": 2.7728957650260723e-05}
{"train_loss": 0.10683005303144455, "global_step": 175790, "epoch": 1975, "lr": 2.7728438601215835e-05}
{"train_loss": 0.025762075558304787, "global_step": 175791, "epoch": 1975, "lr": 2.772791955516506e-05}
{"train_loss": 0.07038561254739761, "global_step": 175792, "epoch": 1975, "lr": 2.7727400512108525e-05}
{"train_loss": 0.08107057213783264, "global_step": 175793, "epoch": 1975, "lr": 2.772688147204624e-05}
{"train_loss": 0.08845552057027817, "global_step": 175794, "epoch": 1975, "lr": 2.7726362434978332e-05}
{"train_loss": 0.1266842633485794, "global_step": 175795, "epoch": 1975, "lr": 2.7725843400904816e-05}
{"train_loss": 0.07313484698534012, "global_step": 175796, "epoch": 1975, "lr": 2.7725324369825816e-05}
{"train_loss": 0.04240121692419052, "global_step": 175797, "epoch": 1975, "lr": 2.7724805341741354e-05}
{"train_loss": 0.037591684609651566, "global_step": 175798, "epoch": 1975, "lr": 2.7724286316651537e-05}
{"train_loss": 0.05054175853729248, "global_step": 175799, "epoch": 1975, "lr": 2.7723767294556402e-05}
{"train_loss": 0.11305207014083862, "global_step": 175800, "epoch": 1975, "lr": 2.7723248275456033e-05}
{"train_loss": 0.042269568890333176, "global_step": 175801, "epoch": 1975, "lr": 2.7722729259350522e-05}
{"train_loss": 0.11175128817558289, "global_step": 175802, "epoch": 1975, "lr": 2.7722210246239892e-05}
{"train_loss": 0.10205256193876266, "global_step": 175803, "epoch": 1975, "lr": 2.772169123612427e-05}
{"train_loss": 0.08626269549131393, "global_step": 175804, "epoch": 1975, "lr": 2.7721172229003667e-05}
{"train_loss": 0.07689952850341797, "global_step": 175805, "epoch": 1975, "lr": 2.7720653224878203e-05}
{"train_loss": 0.06684810668230057, "global_step": 175806, "epoch": 1975, "lr": 2.7720134223747906e-05}
{"train_loss": 0.0726502388715744, "global_step": 175807, "epoch": 1975, "lr": 2.771961522561289e-05}
{"train_loss": 0.08195184916257858, "global_step": 175808, "epoch": 1975, "lr": 2.7719096230473175e-05}
{"train_loss": 0.13316240906715393, "global_step": 175809, "epoch": 1975, "lr": 2.771857723832888e-05}
{"train_loss": 0.02782427705824375, "global_step": 175810, "epoch": 1975, "lr": 2.7718058249180024e-05}
{"train_loss": 0.09543640166521072, "global_step": 175811, "epoch": 1975, "lr": 2.7717539263026726e-05}
{"train_loss": 0.10747271776199341, "global_step": 175812, "epoch": 1975, "lr": 2.7717020279869017e-05}
{"train_loss": 0.06838290393352509, "global_step": 175813, "epoch": 1975, "lr": 2.7716501299707e-05}
{"train_loss": 0.07164335250854492, "global_step": 175814, "epoch": 1975, "lr": 2.771598232254071e-05}
{"train_loss": 0.04735439270734787, "global_step": 175815, "epoch": 1975, "lr": 2.7715463348370253e-05}
{"train_loss": 0.07084701210260391, "global_step": 175816, "epoch": 1975, "lr": 2.7714944377195662e-05}
{"train_loss": 0.09309995919466019, "global_step": 175817, "epoch": 1975, "lr": 2.771442540901702e-05}
{"train_loss": 0.08560479432344437, "global_step": 175818, "epoch": 1975, "lr": 2.7713906443834425e-05}
{"train_loss": 0.10026663541793823, "global_step": 175819, "epoch": 1975, "lr": 2.7713387481647907e-05}
{"train_loss": 0.06829249858856201, "global_step": 175820, "epoch": 1975, "lr": 2.7712868522457568e-05}
{"train_loss": 0.06195377558469772, "global_step": 175821, "epoch": 1975, "lr": 2.7712349566263463e-05}
{"train_loss": 0.07189366966485977, "global_step": 175822, "epoch": 1975, "lr": 2.7711830613065642e-05}
{"train_loss": 0.06363143771886826, "global_step": 175823, "epoch": 1975, "lr": 2.7711311662864214e-05}
{"train_loss": 0.05536482110619545, "global_step": 175824, "epoch": 1975, "lr": 2.7710792715659205e-05}
{"train_loss": 0.09721995145082474, "global_step": 175825, "epoch": 1975, "lr": 2.7710273771450734e-05}
{"train_loss": 0.0398479625582695, "global_step": 175826, "epoch": 1975, "lr": 2.7709754830238843e-05}
{"train_loss": 0.06611853837966919, "global_step": 175827, "epoch": 1975, "lr": 2.770923589202358e-05}
{"train_loss": 0.06372971832752228, "global_step": 175828, "epoch": 1975, "lr": 2.7708716956805065e-05}
{"train_loss": 0.05384569615125656, "global_step": 175829, "epoch": 1975, "lr": 2.7708198024583317e-05}
{"train_loss": 0.0620981864631176, "global_step": 175830, "epoch": 1975, "lr": 2.7707679095358452e-05}
{"train_loss": 0.09004522114992142, "global_step": 175831, "epoch": 1975, "lr": 2.7707160169130496e-05}
{"train_loss": 0.046164944767951965, "global_step": 175832, "epoch": 1975, "lr": 2.7706641245899555e-05}
{"train_loss": 0.10676052421331406, "global_step": 175833, "epoch": 1975, "lr": 2.7706122325665675e-05}
{"train_loss": 0.056423574686050415, "global_step": 175834, "epoch": 1975, "lr": 2.770560340842895e-05}
{"train_loss": 0.05923769623041153, "global_step": 175835, "epoch": 1975, "lr": 2.7705084494189416e-05}
{"train_loss": 0.09549618512392044, "global_step": 175836, "epoch": 1975, "lr": 2.7704565582947163e-05}
{"train_loss": 0.034091632813215256, "global_step": 175837, "epoch": 1975, "lr": 2.770404667470228e-05}
{"train_loss": 0.06071867421269417, "global_step": 175838, "epoch": 1975, "lr": 2.770352776945479e-05}
{"train_loss": 0.05197776108980179, "global_step": 175839, "epoch": 1975, "lr": 2.7703008867204817e-05}
{"train_loss": 0.07377894222736359, "global_step": 175840, "epoch": 1975, "lr": 2.7702489967952373e-05}
{"train_loss": 0.11888552457094193, "global_step": 175841, "epoch": 1975, "lr": 2.7701971071697586e-05}
{"train_loss": 0.15249431133270264, "global_step": 175842, "epoch": 1975, "lr": 2.7701452178440475e-05}
{"train_loss": 0.07087711244821548, "global_step": 175843, "epoch": 1975, "lr": 2.7700933288181152e-05}
{"train_loss": 0.07120165228843689, "global_step": 175844, "epoch": 1975, "lr": 2.770041440091964e-05}
{"train_loss": 0.08981850743293762, "global_step": 175845, "epoch": 1975, "lr": 2.769989551665606e-05}
{"train_loss": 0.0681469738483429, "global_step": 175846, "epoch": 1975, "lr": 2.769937663539044e-05}
{"train_loss": 0.10837649554014206, "global_step": 175847, "epoch": 1975, "lr": 2.7698857757122886e-05}
{"train_loss": 0.09495284408330917, "global_step": 175848, "epoch": 1975, "lr": 2.769833888185343e-05}
{"train_loss": 0.08087675273418427, "global_step": 175849, "epoch": 1975, "lr": 2.7697820009582174e-05}
{"train_loss": 0.11963234096765518, "global_step": 175850, "epoch": 1975, "lr": 2.769730114030916e-05}
{"train_loss": 0.0585649311542511, "global_step": 175851, "epoch": 1975, "lr": 2.7696782274034494e-05}
{"train_loss": 0.05101794749498367, "global_step": 175852, "epoch": 1975, "lr": 2.7696263410758206e-05}
{"train_loss": 0.09993138164281845, "global_step": 175853, "epoch": 1975, "lr": 2.7695744550480373e-05}
{"train_loss": 0.05181824788451195, "global_step": 175854, "epoch": 1975, "lr": 2.7695225693201105e-05}
{"train_loss": 0.07220722734928131, "global_step": 175855, "epoch": 1975, "lr": 2.7694706838920415e-05}
{"train_loss": 0.027221817523241043, "global_step": 175856, "epoch": 1975, "lr": 2.7694187987638422e-05}
{"train_loss": 0.061970170587301254, "global_step": 175857, "epoch": 1975, "lr": 2.7693669139355155e-05}
{"train_loss": 0.06653208285570145, "global_step": 175858, "epoch": 1975, "lr": 2.7693150294070724e-05}
{"train_loss": 0.061448875814676285, "global_step": 175859, "epoch": 1975, "lr": 2.7692631451785173e-05}
{"train_loss": 0.054518960416316986, "global_step": 175860, "epoch": 1975, "lr": 2.7692112612498555e-05}
{"train_loss": 0.08661447465419769, "global_step": 175861, "epoch": 1975, "lr": 2.769159377621099e-05}
{"train_loss": 0.10391891747713089, "global_step": 175862, "epoch": 1975, "lr": 2.769107494292249e-05}
{"train_loss": 0.07483686695105574, "global_step": 175863, "epoch": 1975, "lr": 2.769055611263317e-05, "val_loss": 7.265720844268799, "train_action_mse_error": 32.77300262451172}
{"train_loss": 0.07752324640750885, "global_step": 175864, "epoch": 1976, "lr": 2.7690037285343084e-05}
{"train_loss": 0.06621354073286057, "global_step": 175865, "epoch": 1976, "lr": 2.7689518461052287e-05}
{"train_loss": 0.04897679015994072, "global_step": 175866, "epoch": 1976, "lr": 2.768899963976088e-05}
{"train_loss": 0.08356405049562454, "global_step": 175867, "epoch": 1976, "lr": 2.7688480821468893e-05}
{"train_loss": 0.059925638139247894, "global_step": 175868, "epoch": 1976, "lr": 2.7687962006176438e-05}
{"train_loss": 0.03989718481898308, "global_step": 175869, "epoch": 1976, "lr": 2.7687443193883544e-05}
{"train_loss": 0.05398004502058029, "global_step": 175870, "epoch": 1976, "lr": 2.768692438459033e-05}
{"train_loss": 0.05343855544924736, "global_step": 175871, "epoch": 1976, "lr": 2.7686405578296803e-05}
{"train_loss": 0.11926184594631195, "global_step": 175872, "epoch": 1976, "lr": 2.768588677500308e-05}
{"train_loss": 0.08207570761442184, "global_step": 175873, "epoch": 1976, "lr": 2.768536797470923e-05}
{"train_loss": 0.09408572316169739, "global_step": 175874, "epoch": 1976, "lr": 2.768484917741529e-05}
{"train_loss": 0.052996933460235596, "global_step": 175875, "epoch": 1976, "lr": 2.768433038312137e-05}
{"train_loss": 0.05711488053202629, "global_step": 175876, "epoch": 1976, "lr": 2.76838115918275e-05}
{"train_loss": 0.07670978456735611, "global_step": 175877, "epoch": 1976, "lr": 2.7683292803533795e-05}
{"train_loss": 0.12554539740085602, "global_step": 175878, "epoch": 1976, "lr": 2.7682774018240276e-05}
{"train_loss": 0.06411994248628616, "global_step": 175879, "epoch": 1976, "lr": 2.7682255235947053e-05}
{"train_loss": 0.044605888426303864, "global_step": 175880, "epoch": 1976, "lr": 2.7681736456654162e-05}
{"train_loss": 0.11741238832473755, "global_step": 175881, "epoch": 1976, "lr": 2.7681217680361703e-05}
{"train_loss": 0.07674567401409149, "global_step": 175882, "epoch": 1976, "lr": 2.7680698907069723e-05}
{"train_loss": 0.0649094507098198, "global_step": 175883, "epoch": 1976, "lr": 2.768018013677831e-05}
{"train_loss": 0.04909910261631012, "global_step": 175884, "epoch": 1976, "lr": 2.7679661369487513e-05}
{"train_loss": 0.04597989097237587, "global_step": 175885, "epoch": 1976, "lr": 2.7679142605197433e-05}
{"train_loss": 0.07823553681373596, "global_step": 175886, "epoch": 1976, "lr": 2.7678623843908103e-05}
{"train_loss": 0.04963928833603859, "global_step": 175887, "epoch": 1976, "lr": 2.767810508561962e-05}
{"train_loss": 0.09646857529878616, "global_step": 175888, "epoch": 1976, "lr": 2.7677586330332027e-05}
{"train_loss": 0.07289161533117294, "global_step": 175889, "epoch": 1976, "lr": 2.7677067578045413e-05}
{"train_loss": 0.10487722605466843, "global_step": 175890, "epoch": 1976, "lr": 2.7676548828759865e-05}
{"train_loss": 0.04991363734006882, "global_step": 175891, "epoch": 1976, "lr": 2.767603008247541e-05}
{"train_loss": 0.04409507289528847, "global_step": 175892, "epoch": 1976, "lr": 2.7675511339192168e-05}
{"train_loss": 0.044028639793395996, "global_step": 175893, "epoch": 1976, "lr": 2.7674992598910153e-05}
{"train_loss": 0.0934402272105217, "global_step": 175894, "epoch": 1976, "lr": 2.767447386162949e-05}
{"train_loss": 0.05747177451848984, "global_step": 175895, "epoch": 1976, "lr": 2.7673955127350197e-05}
{"train_loss": 0.08636092394590378, "global_step": 175896, "epoch": 1976, "lr": 2.7673436396072393e-05}
{"train_loss": 0.0900217741727829, "global_step": 175897, "epoch": 1976, "lr": 2.7672917667796117e-05}
{"train_loss": 0.06723684072494507, "global_step": 175898, "epoch": 1976, "lr": 2.767239894252143e-05}
{"train_loss": 0.10429838299751282, "global_step": 175899, "epoch": 1976, "lr": 2.7671880220248435e-05}
{"train_loss": 0.029487954452633858, "global_step": 175900, "epoch": 1976, "lr": 2.7671361500977166e-05}
{"train_loss": 0.11937233805656433, "global_step": 175901, "epoch": 1976, "lr": 2.767084278470773e-05}
{"train_loss": 0.09477749466896057, "global_step": 175902, "epoch": 1976, "lr": 2.7670324071440152e-05}
{"train_loss": 0.20305904746055603, "global_step": 175903, "epoch": 1976, "lr": 2.766980536117455e-05}
{"train_loss": 0.0673382580280304, "global_step": 175904, "epoch": 1976, "lr": 2.7669286653910968e-05}
{"train_loss": 0.12816525995731354, "global_step": 175905, "epoch": 1976, "lr": 2.7668767949649466e-05}
{"train_loss": 0.06619345396757126, "global_step": 175906, "epoch": 1976, "lr": 2.766824924839014e-05}
{"train_loss": 0.034825801849365234, "global_step": 175907, "epoch": 1976, "lr": 2.766773055013302e-05}
{"train_loss": 0.044202517718076706, "global_step": 175908, "epoch": 1976, "lr": 2.7667211854878212e-05}
{"train_loss": 0.03498382493853569, "global_step": 175909, "epoch": 1976, "lr": 2.766669316262579e-05}
{"train_loss": 0.03495388105511665, "global_step": 175910, "epoch": 1976, "lr": 2.7666174473375796e-05}
{"train_loss": 0.056780196726322174, "global_step": 175911, "epoch": 1976, "lr": 2.766565578712832e-05}
{"train_loss": 0.10539982467889786, "global_step": 175912, "epoch": 1976, "lr": 2.766513710388341e-05}
{"train_loss": 0.07048114389181137, "global_step": 175913, "epoch": 1976, "lr": 2.7664618423641176e-05}
{"train_loss": 0.062174245715141296, "global_step": 175914, "epoch": 1976, "lr": 2.7664099746401634e-05}
{"train_loss": 0.08651173114776611, "global_step": 175915, "epoch": 1976, "lr": 2.7663581072164902e-05}
{"train_loss": 0.043175313621759415, "global_step": 175916, "epoch": 1976, "lr": 2.7663062400931017e-05}
{"train_loss": 0.12904903292655945, "global_step": 175917, "epoch": 1976, "lr": 2.7662543732700073e-05}
{"train_loss": 0.10141745954751968, "global_step": 175918, "epoch": 1976, "lr": 2.766202506747211e-05}
{"train_loss": 0.1448034793138504, "global_step": 175919, "epoch": 1976, "lr": 2.766150640524724e-05}
{"train_loss": 0.1284293532371521, "global_step": 175920, "epoch": 1976, "lr": 2.766098774602548e-05}
{"train_loss": 0.0666108950972557, "global_step": 175921, "epoch": 1976, "lr": 2.7660469089806952e-05}
{"train_loss": 0.08415412157773972, "global_step": 175922, "epoch": 1976, "lr": 2.765995043659168e-05}
{"train_loss": 0.09860409796237946, "global_step": 175923, "epoch": 1976, "lr": 2.765943178637978e-05}
{"train_loss": 0.05024106055498123, "global_step": 175924, "epoch": 1976, "lr": 2.7658913139171273e-05}
{"train_loss": 0.03209059312939644, "global_step": 175925, "epoch": 1976, "lr": 2.7658394494966277e-05}
{"train_loss": 0.11939846724271774, "global_step": 175926, "epoch": 1976, "lr": 2.7657875853764813e-05}
{"train_loss": 0.12094099819660187, "global_step": 175927, "epoch": 1976, "lr": 2.7657357215566974e-05}
{"train_loss": 0.09302463382482529, "global_step": 175928, "epoch": 1976, "lr": 2.7656838580372857e-05}
{"train_loss": 0.11149933934211731, "global_step": 175929, "epoch": 1976, "lr": 2.765631994818248e-05}
{"train_loss": 0.07931973040103912, "global_step": 175930, "epoch": 1976, "lr": 2.7655801318995962e-05}
{"train_loss": 0.09136621654033661, "global_step": 175931, "epoch": 1976, "lr": 2.7655282692813327e-05}
{"train_loss": 0.06242441013455391, "global_step": 175932, "epoch": 1976, "lr": 2.7654764069634687e-05}
{"train_loss": 0.07570719718933105, "global_step": 175933, "epoch": 1976, "lr": 2.7654245449460076e-05}
{"train_loss": 0.12429778277873993, "global_step": 175934, "epoch": 1976, "lr": 2.7653726832289594e-05}
{"train_loss": 0.09652144461870193, "global_step": 175935, "epoch": 1976, "lr": 2.7653208218123276e-05}
{"train_loss": 0.11236684024333954, "global_step": 175936, "epoch": 1976, "lr": 2.7652689606961235e-05}
{"train_loss": 0.07877786457538605, "global_step": 175937, "epoch": 1976, "lr": 2.765217099880352e-05}
{"train_loss": 0.0902642160654068, "global_step": 175938, "epoch": 1976, "lr": 2.7651652393650173e-05}
{"train_loss": 0.11808605492115021, "global_step": 175939, "epoch": 1976, "lr": 2.7651133791501312e-05}
{"train_loss": 0.05442190170288086, "global_step": 175940, "epoch": 1976, "lr": 2.765061519235696e-05}
{"train_loss": 0.10589738190174103, "global_step": 175941, "epoch": 1976, "lr": 2.7650096596217236e-05}
{"train_loss": 0.11512964963912964, "global_step": 175942, "epoch": 1976, "lr": 2.7649578003082177e-05}
{"train_loss": 0.0979972630739212, "global_step": 175943, "epoch": 1976, "lr": 2.7649059412951845e-05}
{"train_loss": 0.07759986072778702, "global_step": 175944, "epoch": 1976, "lr": 2.764854082582632e-05}
{"train_loss": 0.10348925739526749, "global_step": 175945, "epoch": 1976, "lr": 2.76480222417057e-05}
{"train_loss": 0.16199305653572083, "global_step": 175946, "epoch": 1976, "lr": 2.7647503660590012e-05}
{"train_loss": 0.15517374873161316, "global_step": 175947, "epoch": 1976, "lr": 2.764698508247936e-05}
{"train_loss": 0.05613381788134575, "global_step": 175948, "epoch": 1976, "lr": 2.7646466507373786e-05}
{"train_loss": 0.036276381462812424, "global_step": 175949, "epoch": 1976, "lr": 2.7645947935273385e-05}
{"train_loss": 0.0638563334941864, "global_step": 175950, "epoch": 1976, "lr": 2.76454293661782e-05}
{"train_loss": 0.09638161957263947, "global_step": 175951, "epoch": 1976, "lr": 2.764491080008833e-05}
{"train_loss": 0.08156664600449332, "global_step": 175952, "epoch": 1976, "lr": 2.7644392237003807e-05, "val_loss": 7.290443420410156}
{"train_loss": 0.07951097935438156, "global_step": 175953, "epoch": 1977, "lr": 2.7643873676924748e-05}
{"train_loss": 0.054893068969249725, "global_step": 175954, "epoch": 1977, "lr": 2.7643355119851178e-05}
{"train_loss": 0.03424624353647232, "global_step": 175955, "epoch": 1977, "lr": 2.7642836565783203e-05}
{"train_loss": 0.08465006947517395, "global_step": 175956, "epoch": 1977, "lr": 2.7642318014720857e-05}
{"train_loss": 0.05395101383328438, "global_step": 175957, "epoch": 1977, "lr": 2.7641799466664255e-05}
{"train_loss": 0.032972678542137146, "global_step": 175958, "epoch": 1977, "lr": 2.764128092161341e-05}
{"train_loss": 0.10309156775474548, "global_step": 175959, "epoch": 1977, "lr": 2.7640762379568448e-05}
{"train_loss": 0.07528062909841537, "global_step": 175960, "epoch": 1977, "lr": 2.764024384052939e-05}
{"train_loss": 0.07571786642074585, "global_step": 175961, "epoch": 1977, "lr": 2.763972530449635e-05}
{"train_loss": 0.12054265290498734, "global_step": 175962, "epoch": 1977, "lr": 2.7639206771469363e-05}
{"train_loss": 0.1195865347981453, "global_step": 175963, "epoch": 1977, "lr": 2.76386882414485e-05}
{"train_loss": 0.05326636880636215, "global_step": 175964, "epoch": 1977, "lr": 2.763816971443387e-05}
{"train_loss": 0.056607432663440704, "global_step": 175965, "epoch": 1977, "lr": 2.7637651190425496e-05}
{"train_loss": 0.10679832100868225, "global_step": 175966, "epoch": 1977, "lr": 2.763713266942348e-05}
{"train_loss": 0.0645456612110138, "global_step": 175967, "epoch": 1977, "lr": 2.763661415142787e-05}
{"train_loss": 0.12619970738887787, "global_step": 175968, "epoch": 1977, "lr": 2.7636095636438765e-05}
{"train_loss": 0.06517776101827621, "global_step": 175969, "epoch": 1977, "lr": 2.763557712445618e-05}
{"train_loss": 0.10384766012430191, "global_step": 175970, "epoch": 1977, "lr": 2.7635058615480253e-05}
{"train_loss": 0.10841871798038483, "global_step": 175971, "epoch": 1977, "lr": 2.7634540109510993e-05}
{"train_loss": 0.08549219369888306, "global_step": 175972, "epoch": 1977, "lr": 2.763402160654852e-05}
{"train_loss": 0.13947223126888275, "global_step": 175973, "epoch": 1977, "lr": 2.7633503106592856e-05}
{"train_loss": 0.048253174871206284, "global_step": 175974, "epoch": 1977, "lr": 2.7632984609644118e-05}
{"train_loss": 0.10508470982313156, "global_step": 175975, "epoch": 1977, "lr": 2.763246611570235e-05}
{"train_loss": 0.1146261990070343, "global_step": 175976, "epoch": 1977, "lr": 2.7631947624767606e-05}
{"train_loss": 0.05860817804932594, "global_step": 175977, "epoch": 1977, "lr": 2.7631429136839992e-05}
{"train_loss": 0.1140551045536995, "global_step": 175978, "epoch": 1977, "lr": 2.763091065191955e-05}
{"train_loss": 0.06913942098617554, "global_step": 175979, "epoch": 1977, "lr": 2.763039217000637e-05}
{"train_loss": 0.10365234315395355, "global_step": 175980, "epoch": 1977, "lr": 2.7629873691100494e-05}
{"train_loss": 0.07159820199012756, "global_step": 175981, "epoch": 1977, "lr": 2.7629355215202022e-05}
{"train_loss": 0.0722670778632164, "global_step": 175982, "epoch": 1977, "lr": 2.7628836742311e-05}
{"train_loss": 0.06011047959327698, "global_step": 175983, "epoch": 1977, "lr": 2.762831827242752e-05}
{"train_loss": 0.10737914592027664, "global_step": 175984, "epoch": 1977, "lr": 2.7627799805551623e-05}
{"train_loss": 0.06393878906965256, "global_step": 175985, "epoch": 1977, "lr": 2.7627281341683418e-05}
{"train_loss": 0.09973577409982681, "global_step": 175986, "epoch": 1977, "lr": 2.7626762880822932e-05}
{"train_loss": 0.05625326931476593, "global_step": 175987, "epoch": 1977, "lr": 2.762624442297027e-05}
{"train_loss": 0.08096735924482346, "global_step": 175988, "epoch": 1977, "lr": 2.7625725968125472e-05}
{"train_loss": 0.12498462200164795, "global_step": 175989, "epoch": 1977, "lr": 2.762520751628864e-05}
{"train_loss": 0.13264422118663788, "global_step": 175990, "epoch": 1977, "lr": 2.7624689067459808e-05}
{"train_loss": 0.11695052683353424, "global_step": 175991, "epoch": 1977, "lr": 2.762417062163909e-05}
{"train_loss": 0.07401109486818314, "global_step": 175992, "epoch": 1977, "lr": 2.7623652178826497e-05}
{"train_loss": 0.10122358798980713, "global_step": 175993, "epoch": 1977, "lr": 2.7623133739022155e-05}
{"train_loss": 0.10934741050004959, "global_step": 175994, "epoch": 1977, "lr": 2.7622615302226095e-05}
{"train_loss": 0.06316158175468445, "global_step": 175995, "epoch": 1977, "lr": 2.762209686843842e-05}
{"train_loss": 0.04142861068248749, "global_step": 175996, "epoch": 1977, "lr": 2.7621578437659156e-05}
{"train_loss": 0.07785569131374359, "global_step": 175997, "epoch": 1977, "lr": 2.762106000988842e-05}
{"train_loss": 0.060048907995224, "global_step": 175998, "epoch": 1977, "lr": 2.7620541585126247e-05}
{"train_loss": 0.11629658192396164, "global_step": 175999, "epoch": 1977, "lr": 2.762002316337271e-05}
{"train_loss": 0.0643150731921196, "global_step": 176000, "epoch": 1977, "lr": 2.7619504744627917e-05}
{"train_loss": 0.07165396958589554, "global_step": 176001, "epoch": 1977, "lr": 2.761898632889188e-05}
{"train_loss": 0.05638770014047623, "global_step": 176002, "epoch": 1977, "lr": 2.7618467916164727e-05}
{"train_loss": 0.09709758311510086, "global_step": 176003, "epoch": 1977, "lr": 2.7617949506446474e-05}
{"train_loss": 0.04949410259723663, "global_step": 176004, "epoch": 1977, "lr": 2.7617431099737227e-05}
{"train_loss": 0.08951277285814285, "global_step": 176005, "epoch": 1977, "lr": 2.7616912696037035e-05}
{"train_loss": 0.08413150906562805, "global_step": 176006, "epoch": 1977, "lr": 2.7616394295345988e-05}
{"train_loss": 0.06289861351251602, "global_step": 176007, "epoch": 1977, "lr": 2.7615875897664127e-05}
{"train_loss": 0.0570993535220623, "global_step": 176008, "epoch": 1977, "lr": 2.7615357502991558e-05}
{"train_loss": 0.07248471677303314, "global_step": 176009, "epoch": 1977, "lr": 2.7614839111328317e-05}
{"train_loss": 0.11553855240345001, "global_step": 176010, "epoch": 1977, "lr": 2.7614320722674503e-05}
{"train_loss": 0.04221353679895401, "global_step": 176011, "epoch": 1977, "lr": 2.761380233703015e-05}
{"train_loss": 0.04990354925394058, "global_step": 176012, "epoch": 1977, "lr": 2.7613283954395368e-05}
{"train_loss": 0.0528317466378212, "global_step": 176013, "epoch": 1977, "lr": 2.761276557477019e-05}
{"train_loss": 0.04262244328856468, "global_step": 176014, "epoch": 1977, "lr": 2.7612247198154716e-05}
{"train_loss": 0.07264743000268936, "global_step": 176015, "epoch": 1977, "lr": 2.7611728824549012e-05}
{"train_loss": 0.11269534379243851, "global_step": 176016, "epoch": 1977, "lr": 2.7611210453953117e-05}
{"train_loss": 0.10417086631059647, "global_step": 176017, "epoch": 1977, "lr": 2.761069208636714e-05}
{"train_loss": 0.07228577136993408, "global_step": 176018, "epoch": 1977, "lr": 2.761017372179111e-05}
{"train_loss": 0.04977628216147423, "global_step": 176019, "epoch": 1977, "lr": 2.760965536022514e-05}
{"train_loss": 0.04194440692663193, "global_step": 176020, "epoch": 1977, "lr": 2.7609137001669262e-05}
{"train_loss": 0.04120872914791107, "global_step": 176021, "epoch": 1977, "lr": 2.7608618646123574e-05}
{"train_loss": 0.05288581922650337, "global_step": 176022, "epoch": 1977, "lr": 2.7608100293588126e-05}
{"train_loss": 0.06191543862223625, "global_step": 176023, "epoch": 1977, "lr": 2.760758194406301e-05}
{"train_loss": 0.08474094420671463, "global_step": 176024, "epoch": 1977, "lr": 2.760706359754826e-05}
{"train_loss": 0.03168698027729988, "global_step": 176025, "epoch": 1977, "lr": 2.7606545254043992e-05}
{"train_loss": 0.08031462132930756, "global_step": 176026, "epoch": 1977, "lr": 2.7606026913550232e-05}
{"train_loss": 0.06270557641983032, "global_step": 176027, "epoch": 1977, "lr": 2.7605508576067084e-05}
{"train_loss": 0.11708638072013855, "global_step": 176028, "epoch": 1977, "lr": 2.760499024159458e-05}
{"train_loss": 0.14368276298046112, "global_step": 176029, "epoch": 1977, "lr": 2.7604471910132833e-05}
{"train_loss": 0.07092893123626709, "global_step": 176030, "epoch": 1977, "lr": 2.760395358168188e-05}
{"train_loss": 0.059027496725320816, "global_step": 176031, "epoch": 1977, "lr": 2.7603435256241815e-05}
{"train_loss": 0.07490924000740051, "global_step": 176032, "epoch": 1977, "lr": 2.760291693381268e-05}
{"train_loss": 0.07289525866508484, "global_step": 176033, "epoch": 1977, "lr": 2.7602398614394576e-05}
{"train_loss": 0.08277913182973862, "global_step": 176034, "epoch": 1977, "lr": 2.7601880297987535e-05}
{"train_loss": 0.11271935701370239, "global_step": 176035, "epoch": 1977, "lr": 2.7601361984591657e-05}
{"train_loss": 0.09457919746637344, "global_step": 176036, "epoch": 1977, "lr": 2.760084367420701e-05}
{"train_loss": 0.09110450744628906, "global_step": 176037, "epoch": 1977, "lr": 2.7600325366833644e-05}
{"train_loss": 0.1371344029903412, "global_step": 176038, "epoch": 1977, "lr": 2.759980706247166e-05}
{"train_loss": 0.08466597646474838, "global_step": 176039, "epoch": 1977, "lr": 2.7599288761121095e-05}
{"train_loss": 0.11797258257865906, "global_step": 176040, "epoch": 1977, "lr": 2.7598770462782047e-05}
{"train_loss": 0.08074193925000309, "global_step": 176041, "epoch": 1977, "lr": 2.7598252167454553e-05, "val_loss": 7.376204013824463}
{"train_loss": 0.07952352613210678, "global_step": 176042, "epoch": 1978, "lr": 2.759773387513872e-05}
{"train_loss": 0.07713254541158676, "global_step": 176043, "epoch": 1978, "lr": 2.7597215585834578e-05}
{"train_loss": 0.11536281555891037, "global_step": 176044, "epoch": 1978, "lr": 2.7596697299542236e-05}
{"train_loss": 0.06298346072435379, "global_step": 176045, "epoch": 1978, "lr": 2.7596179016261726e-05}
{"train_loss": 0.06704786419868469, "global_step": 176046, "epoch": 1978, "lr": 2.7595660735993157e-05}
{"train_loss": 0.08844778686761856, "global_step": 176047, "epoch": 1978, "lr": 2.7595142458736555e-05}
{"train_loss": 0.07423345744609833, "global_step": 176048, "epoch": 1978, "lr": 2.759462418449204e-05}
{"train_loss": 0.08002269268035889, "global_step": 176049, "epoch": 1978, "lr": 2.7594105913259625e-05}
{"train_loss": 0.07568787038326263, "global_step": 176050, "epoch": 1978, "lr": 2.7593587645039442e-05}
{"train_loss": 0.039895378053188324, "global_step": 176051, "epoch": 1978, "lr": 2.7593069379831505e-05}
{"train_loss": 0.056279122829437256, "global_step": 176052, "epoch": 1978, "lr": 2.759255111763592e-05}
{"train_loss": 0.08223176747560501, "global_step": 176053, "epoch": 1978, "lr": 2.7592032858452744e-05}
{"train_loss": 0.15163883566856384, "global_step": 176054, "epoch": 1978, "lr": 2.7591514602282032e-05}
{"train_loss": 0.08053120970726013, "global_step": 176055, "epoch": 1978, "lr": 2.7590996349123884e-05}
{"train_loss": 0.08781088143587112, "global_step": 176056, "epoch": 1978, "lr": 2.7590478098978334e-05}
{"train_loss": 0.06598559767007828, "global_step": 176057, "epoch": 1978, "lr": 2.7589959851845493e-05}
{"train_loss": 0.07375146448612213, "global_step": 176058, "epoch": 1978, "lr": 2.7589441607725387e-05}
{"train_loss": 0.07971745729446411, "global_step": 176059, "epoch": 1978, "lr": 2.7588923366618125e-05}
{"train_loss": 0.08429906517267227, "global_step": 176060, "epoch": 1978, "lr": 2.7588405128523743e-05}
{"train_loss": 0.11656872183084488, "global_step": 176061, "epoch": 1978, "lr": 2.758788689344234e-05}
{"train_loss": 0.07089423388242722, "global_step": 176062, "epoch": 1978, "lr": 2.758736866137396e-05}
{"train_loss": 0.11445365846157074, "global_step": 176063, "epoch": 1978, "lr": 2.75868504323187e-05}
{"train_loss": 0.046540919691324234, "global_step": 176064, "epoch": 1978, "lr": 2.75863322062766e-05}
{"train_loss": 0.07494183629751205, "global_step": 176065, "epoch": 1978, "lr": 2.758581398324776e-05}
{"train_loss": 0.12359582632780075, "global_step": 176066, "epoch": 1978, "lr": 2.7585295763232215e-05}
{"train_loss": 0.1064005196094513, "global_step": 176067, "epoch": 1978, "lr": 2.7584777546230078e-05}
{"train_loss": 0.09503071010112762, "global_step": 176068, "epoch": 1978, "lr": 2.758425933224137e-05}
{"train_loss": 0.08877100050449371, "global_step": 176069, "epoch": 1978, "lr": 2.7583741121266204e-05}
{"train_loss": 0.064339280128479, "global_step": 176070, "epoch": 1978, "lr": 2.758322291330462e-05}
{"train_loss": 0.08400247246026993, "global_step": 176071, "epoch": 1978, "lr": 2.758270470835669e-05}
{"train_loss": 0.095216304063797, "global_step": 176072, "epoch": 1978, "lr": 2.7582186506422514e-05}
{"train_loss": 0.14725664258003235, "global_step": 176073, "epoch": 1978, "lr": 2.7581668307502118e-05}
{"train_loss": 0.06442420929670334, "global_step": 176074, "epoch": 1978, "lr": 2.758115011159561e-05}
{"train_loss": 0.057085175067186356, "global_step": 176075, "epoch": 1978, "lr": 2.758063191870303e-05}
{"train_loss": 0.059601981192827225, "global_step": 176076, "epoch": 1978, "lr": 2.758011372882448e-05}
{"train_loss": 0.0302165187895298, "global_step": 176077, "epoch": 1978, "lr": 2.757959554195999e-05}
{"train_loss": 0.06086277961730957, "global_step": 176078, "epoch": 1978, "lr": 2.757907735810967e-05}
{"train_loss": 0.10969923436641693, "global_step": 176079, "epoch": 1978, "lr": 2.7578559177273545e-05}
{"train_loss": 0.09996646642684937, "global_step": 176080, "epoch": 1978, "lr": 2.7578040999451737e-05}
{"train_loss": 0.07420635968446732, "global_step": 176081, "epoch": 1978, "lr": 2.7577522824644266e-05}
{"train_loss": 0.08397568762302399, "global_step": 176082, "epoch": 1978, "lr": 2.7577004652851242e-05}
{"train_loss": 0.1112806499004364, "global_step": 176083, "epoch": 1978, "lr": 2.7576486484072693e-05}
{"train_loss": 0.04824144393205643, "global_step": 176084, "epoch": 1978, "lr": 2.757596831830874e-05}
{"train_loss": 0.07295376807451248, "global_step": 176085, "epoch": 1978, "lr": 2.75754501555594e-05}
{"train_loss": 0.07511591166257858, "global_step": 176086, "epoch": 1978, "lr": 2.757493199582479e-05}
{"train_loss": 0.08551234006881714, "global_step": 176087, "epoch": 1978, "lr": 2.757441383910493e-05}
{"train_loss": 0.03212333843111992, "global_step": 176088, "epoch": 1978, "lr": 2.7573895685399942e-05}
{"train_loss": 0.06178628280758858, "global_step": 176089, "epoch": 1978, "lr": 2.7573377534709853e-05}
{"train_loss": 0.05140852555632591, "global_step": 176090, "epoch": 1978, "lr": 2.7572859387034767e-05}
{"train_loss": 0.04012966528534889, "global_step": 176091, "epoch": 1978, "lr": 2.757234124237474e-05}
{"train_loss": 0.05795155093073845, "global_step": 176092, "epoch": 1978, "lr": 2.757182310072981e-05}
{"train_loss": 0.06283046305179596, "global_step": 176093, "epoch": 1978, "lr": 2.7571304962100102e-05}
{"train_loss": 0.08559680730104446, "global_step": 176094, "epoch": 1978, "lr": 2.757078682648564e-05}
{"train_loss": 0.08140970021486282, "global_step": 176095, "epoch": 1978, "lr": 2.7570268693886526e-05}
{"train_loss": 0.07747800648212433, "global_step": 176096, "epoch": 1978, "lr": 2.7569750564302797e-05}
{"train_loss": 0.05465295910835266, "global_step": 176097, "epoch": 1978, "lr": 2.756923243773457e-05}
{"train_loss": 0.1150858998298645, "global_step": 176098, "epoch": 1978, "lr": 2.7568714314181858e-05}
{"train_loss": 0.05888799950480461, "global_step": 176099, "epoch": 1978, "lr": 2.756819619364478e-05}
{"train_loss": 0.11228073388338089, "global_step": 176100, "epoch": 1978, "lr": 2.756767807612337e-05}
{"train_loss": 0.05173765495419502, "global_step": 176101, "epoch": 1978, "lr": 2.756715996161773e-05}
{"train_loss": 0.04913729801774025, "global_step": 176102, "epoch": 1978, "lr": 2.7566641850127884e-05}
{"train_loss": 0.12355569005012512, "global_step": 176103, "epoch": 1978, "lr": 2.756612374165396e-05}
{"train_loss": 0.07483362406492233, "global_step": 176104, "epoch": 1978, "lr": 2.7565605636195967e-05}
{"train_loss": 0.03881519287824631, "global_step": 176105, "epoch": 1978, "lr": 2.7565087533754036e-05}
{"train_loss": 0.04878627508878708, "global_step": 176106, "epoch": 1978, "lr": 2.756456943432818e-05}
{"train_loss": 0.09447916597127914, "global_step": 176107, "epoch": 1978, "lr": 2.7564051337918494e-05}
{"train_loss": 0.08578513562679291, "global_step": 176108, "epoch": 1978, "lr": 2.756353324452507e-05}
{"train_loss": 0.0873788520693779, "global_step": 176109, "epoch": 1978, "lr": 2.7563015154147933e-05}
{"train_loss": 0.08301004022359848, "global_step": 176110, "epoch": 1978, "lr": 2.7562497066787197e-05}
{"train_loss": 0.08087202906608582, "global_step": 176111, "epoch": 1978, "lr": 2.7561978982442888e-05}
{"train_loss": 0.060132261365652084, "global_step": 176112, "epoch": 1978, "lr": 2.756146090111512e-05}
{"train_loss": 0.12275991588830948, "global_step": 176113, "epoch": 1978, "lr": 2.7560942822803916e-05}
{"train_loss": 0.1304379254579544, "global_step": 176114, "epoch": 1978, "lr": 2.7560424747509395e-05}
{"train_loss": 0.0953565239906311, "global_step": 176115, "epoch": 1978, "lr": 2.755990667523158e-05}
{"train_loss": 0.057505276054143906, "global_step": 176116, "epoch": 1978, "lr": 2.7559388605970575e-05}
{"train_loss": 0.0797141045331955, "global_step": 176117, "epoch": 1978, "lr": 2.7558870539726422e-05}
{"train_loss": 0.10276548564434052, "global_step": 176118, "epoch": 1978, "lr": 2.7558352476499227e-05}
{"train_loss": 0.07626653462648392, "global_step": 176119, "epoch": 1978, "lr": 2.7557834416289018e-05}
{"train_loss": 0.07111844420433044, "global_step": 176120, "epoch": 1978, "lr": 2.7557316359095903e-05}
{"train_loss": 0.059560663998126984, "global_step": 176121, "epoch": 1978, "lr": 2.7556798304919916e-05}
{"train_loss": 0.10120422393083572, "global_step": 176122, "epoch": 1978, "lr": 2.755628025376116e-05}
{"train_loss": 0.05710890144109726, "global_step": 176123, "epoch": 1978, "lr": 2.7555762205619675e-05}
{"train_loss": 0.08146515488624573, "global_step": 176124, "epoch": 1978, "lr": 2.7555244160495553e-05}
{"train_loss": 0.12035390734672546, "global_step": 176125, "epoch": 1978, "lr": 2.755472611838884e-05}
{"train_loss": 0.06042621284723282, "global_step": 176126, "epoch": 1978, "lr": 2.755420807929964e-05}
{"train_loss": 0.09514963626861572, "global_step": 176127, "epoch": 1978, "lr": 2.7553690043227987e-05}
{"train_loss": 0.13473550975322723, "global_step": 176128, "epoch": 1978, "lr": 2.755317201017399e-05}
{"train_loss": 0.07926282286643982, "global_step": 176129, "epoch": 1978, "lr": 2.7552653980137666e-05}
{"train_loss": 0.080017661822311, "global_step": 176130, "epoch": 1978, "lr": 2.755213595311914e-05, "val_loss": 7.499447345733643}
{"train_loss": 0.05988045036792755, "global_step": 176131, "epoch": 1979, "lr": 2.755161792911845e-05}
{"train_loss": 0.1394556164741516, "global_step": 176132, "epoch": 1979, "lr": 2.755109990813566e-05}
{"train_loss": 0.06764963269233704, "global_step": 176133, "epoch": 1979, "lr": 2.7550581890170863e-05}
{"train_loss": 0.08705226331949234, "global_step": 176134, "epoch": 1979, "lr": 2.7550063875224095e-05}
{"train_loss": 0.11247226595878601, "global_step": 176135, "epoch": 1979, "lr": 2.754954586329548e-05}
{"train_loss": 0.03248858079314232, "global_step": 176136, "epoch": 1979, "lr": 2.7549027854385024e-05}
{"train_loss": 0.07186926901340485, "global_step": 176137, "epoch": 1979, "lr": 2.7548509848492853e-05}
{"train_loss": 0.05100850388407707, "global_step": 176138, "epoch": 1979, "lr": 2.7547991845618988e-05}
{"train_loss": 0.12208415567874908, "global_step": 176139, "epoch": 1979, "lr": 2.754747384576354e-05}
{"train_loss": 0.07359185069799423, "global_step": 176140, "epoch": 1979, "lr": 2.7546955848926546e-05}
{"train_loss": 0.12315978854894638, "global_step": 176141, "epoch": 1979, "lr": 2.7546437855108104e-05}
{"train_loss": 0.1357421576976776, "global_step": 176142, "epoch": 1979, "lr": 2.7545919864308257e-05}
{"train_loss": 0.08335632085800171, "global_step": 176143, "epoch": 1979, "lr": 2.7545401876527077e-05}
{"train_loss": 0.0912843570113182, "global_step": 176144, "epoch": 1979, "lr": 2.7544883891764674e-05}
{"train_loss": 0.08869475871324539, "global_step": 176145, "epoch": 1979, "lr": 2.7544365910021063e-05}
{"train_loss": 0.06506797671318054, "global_step": 176146, "epoch": 1979, "lr": 2.754384793129636e-05}
{"train_loss": 0.05862218141555786, "global_step": 176147, "epoch": 1979, "lr": 2.754332995559059e-05}
{"train_loss": 0.06820425391197205, "global_step": 176148, "epoch": 1979, "lr": 2.7542811982903873e-05}
{"train_loss": 0.11015070229768753, "global_step": 176149, "epoch": 1979, "lr": 2.7542294013236224e-05}
{"train_loss": 0.07825984060764313, "global_step": 176150, "epoch": 1979, "lr": 2.7541776046587763e-05}
{"train_loss": 0.051019974052906036, "global_step": 176151, "epoch": 1979, "lr": 2.7541258082958514e-05}
{"train_loss": 0.04977697506546974, "global_step": 176152, "epoch": 1979, "lr": 2.754074012234859e-05}
{"train_loss": 0.060210518538951874, "global_step": 176153, "epoch": 1979, "lr": 2.7540222164758023e-05}
{"train_loss": 0.17367680370807648, "global_step": 176154, "epoch": 1979, "lr": 2.753970421018691e-05}
{"train_loss": 0.06819897890090942, "global_step": 176155, "epoch": 1979, "lr": 2.75391862586353e-05}
{"train_loss": 0.11901610344648361, "global_step": 176156, "epoch": 1979, "lr": 2.753866831010329e-05}
{"train_loss": 0.060121022164821625, "global_step": 176157, "epoch": 1979, "lr": 2.753815036459091e-05}
{"train_loss": 0.12366753816604614, "global_step": 176158, "epoch": 1979, "lr": 2.7537632422098276e-05}
{"train_loss": 0.14276129007339478, "global_step": 176159, "epoch": 1979, "lr": 2.7537114482625414e-05}
{"train_loss": 0.11864527314901352, "global_step": 176160, "epoch": 1979, "lr": 2.7536596546172427e-05}
{"train_loss": 0.07724426686763763, "global_step": 176161, "epoch": 1979, "lr": 2.7536078612739357e-05}
{"train_loss": 0.08783268183469772, "global_step": 176162, "epoch": 1979, "lr": 2.7535560682326284e-05}
{"train_loss": 0.07950563728809357, "global_step": 176163, "epoch": 1979, "lr": 2.7535042754933298e-05}
{"train_loss": 0.0743381604552269, "global_step": 176164, "epoch": 1979, "lr": 2.753452483056046e-05}
{"train_loss": 0.10453970730304718, "global_step": 176165, "epoch": 1979, "lr": 2.7534006909207806e-05}
{"train_loss": 0.02495589479804039, "global_step": 176166, "epoch": 1979, "lr": 2.753348899087545e-05}
{"train_loss": 0.12037478387355804, "global_step": 176167, "epoch": 1979, "lr": 2.7532971075563422e-05}
{"train_loss": 0.07112735509872437, "global_step": 176168, "epoch": 1979, "lr": 2.7532453163271837e-05}
{"train_loss": 0.0731833204627037, "global_step": 176169, "epoch": 1979, "lr": 2.753193525400074e-05}
{"train_loss": 0.07955005019903183, "global_step": 176170, "epoch": 1979, "lr": 2.7531417347750178e-05}
{"train_loss": 0.08643972128629684, "global_step": 176171, "epoch": 1979, "lr": 2.753089944452026e-05}
{"train_loss": 0.10416314750909805, "global_step": 176172, "epoch": 1979, "lr": 2.7530381544311024e-05}
{"train_loss": 0.03496989980340004, "global_step": 176173, "epoch": 1979, "lr": 2.7529863647122568e-05}
{"train_loss": 0.06871920824050903, "global_step": 176174, "epoch": 1979, "lr": 2.7529345752954933e-05}
{"train_loss": 0.08444742113351822, "global_step": 176175, "epoch": 1979, "lr": 2.752882786180822e-05}
{"train_loss": 0.06424763798713684, "global_step": 176176, "epoch": 1979, "lr": 2.7528309973682465e-05}
{"train_loss": 0.048328641802072525, "global_step": 176177, "epoch": 1979, "lr": 2.7527792088577775e-05}
{"train_loss": 0.06973093748092651, "global_step": 176178, "epoch": 1979, "lr": 2.752727420649418e-05}
{"train_loss": 0.0936708077788353, "global_step": 176179, "epoch": 1979, "lr": 2.7526756327431785e-05}
{"train_loss": 0.04691431671380997, "global_step": 176180, "epoch": 1979, "lr": 2.752623845139063e-05}
{"train_loss": 0.06658689677715302, "global_step": 176181, "epoch": 1979, "lr": 2.7525720578370794e-05}
{"train_loss": 0.07635806500911713, "global_step": 176182, "epoch": 1979, "lr": 2.752520270837237e-05}
{"train_loss": 0.06063569337129593, "global_step": 176183, "epoch": 1979, "lr": 2.7524684841395388e-05}
{"train_loss": 0.03976786136627197, "global_step": 176184, "epoch": 1979, "lr": 2.7524166977439965e-05}
{"train_loss": 0.029923301190137863, "global_step": 176185, "epoch": 1979, "lr": 2.752364911650612e-05}
{"train_loss": 0.13821768760681152, "global_step": 176186, "epoch": 1979, "lr": 2.7523131258593964e-05}
{"train_loss": 0.11104108393192291, "global_step": 176187, "epoch": 1979, "lr": 2.7522613403703535e-05}
{"train_loss": 0.12002316862344742, "global_step": 176188, "epoch": 1979, "lr": 2.7522095551834926e-05}
{"train_loss": 0.09926634281873703, "global_step": 176189, "epoch": 1979, "lr": 2.752157770298819e-05}
{"train_loss": 0.09317665547132492, "global_step": 176190, "epoch": 1979, "lr": 2.752105985716341e-05}
{"train_loss": 0.12902024388313293, "global_step": 176191, "epoch": 1979, "lr": 2.752054201436064e-05}
{"train_loss": 0.12056241184473038, "global_step": 176192, "epoch": 1979, "lr": 2.752002417457998e-05}
{"train_loss": 0.09804831445217133, "global_step": 176193, "epoch": 1979, "lr": 2.7519506337821455e-05}
{"train_loss": 0.04012816399335861, "global_step": 176194, "epoch": 1979, "lr": 2.7518988504085176e-05}
{"train_loss": 0.10200203210115433, "global_step": 176195, "epoch": 1979, "lr": 2.7518470673371178e-05}
{"train_loss": 0.0678618922829628, "global_step": 176196, "epoch": 1979, "lr": 2.7517952845679567e-05}
{"train_loss": 0.12679807841777802, "global_step": 176197, "epoch": 1979, "lr": 2.7517435021010375e-05}
{"train_loss": 0.11113649606704712, "global_step": 176198, "epoch": 1979, "lr": 2.751691719936369e-05}
{"train_loss": 0.05486270785331726, "global_step": 176199, "epoch": 1979, "lr": 2.75163993807396e-05}
{"train_loss": 0.051172543317079544, "global_step": 176200, "epoch": 1979, "lr": 2.7515881565138136e-05}
{"train_loss": 0.12892599403858185, "global_step": 176201, "epoch": 1979, "lr": 2.75153637525594e-05}
{"train_loss": 0.08507205545902252, "global_step": 176202, "epoch": 1979, "lr": 2.751484594300346e-05}
{"train_loss": 0.07825025916099548, "global_step": 176203, "epoch": 1979, "lr": 2.7514328136470352e-05}
{"train_loss": 0.06642546504735947, "global_step": 176204, "epoch": 1979, "lr": 2.751381033296019e-05}
{"train_loss": 0.03701159358024597, "global_step": 176205, "epoch": 1979, "lr": 2.7513292532473e-05}
{"train_loss": 0.02415129914879799, "global_step": 176206, "epoch": 1979, "lr": 2.7512774735008894e-05}
{"train_loss": 0.049042973667383194, "global_step": 176207, "epoch": 1979, "lr": 2.7512256940567916e-05}
{"train_loss": 0.09045664966106415, "global_step": 176208, "epoch": 1979, "lr": 2.7511739149150128e-05}
{"train_loss": 0.06962349265813828, "global_step": 176209, "epoch": 1979, "lr": 2.7511221360755634e-05}
{"train_loss": 0.06778847426176071, "global_step": 176210, "epoch": 1979, "lr": 2.7510703575384455e-05}
{"train_loss": 0.0585976280272007, "global_step": 176211, "epoch": 1979, "lr": 2.7510185793036713e-05}
{"train_loss": 0.05461893230676651, "global_step": 176212, "epoch": 1979, "lr": 2.750966801371243e-05}
{"train_loss": 0.0908636823296547, "global_step": 176213, "epoch": 1979, "lr": 2.7509150237411718e-05}
{"train_loss": 0.03458357974886894, "global_step": 176214, "epoch": 1979, "lr": 2.750863246413461e-05}
{"train_loss": 0.06652740389108658, "global_step": 176215, "epoch": 1979, "lr": 2.7508114693881215e-05}
{"train_loss": 0.12480072677135468, "global_step": 176216, "epoch": 1979, "lr": 2.7507596926651554e-05}
{"train_loss": 0.13433779776096344, "global_step": 176217, "epoch": 1979, "lr": 2.7507079162445726e-05}
{"train_loss": 0.05080334097146988, "global_step": 176218, "epoch": 1979, "lr": 2.750656140126381e-05}
{"train_loss": 0.08222243943241205, "global_step": 176219, "epoch": 1979, "lr": 2.750604364310585e-05, "val_loss": 7.3798699378967285}
{"train_loss": 0.07116252928972244, "global_step": 176220, "epoch": 1980, "lr": 2.7505525887971938e-05}
{"train_loss": 0.10476814955472946, "global_step": 176221, "epoch": 1980, "lr": 2.7505008135862127e-05}
{"train_loss": 0.06195341795682907, "global_step": 176222, "epoch": 1980, "lr": 2.7504490386776504e-05}
{"train_loss": 0.0889367014169693, "global_step": 176223, "epoch": 1980, "lr": 2.7503972640715115e-05}
{"train_loss": 0.04474075511097908, "global_step": 176224, "epoch": 1980, "lr": 2.750345489767806e-05}
{"train_loss": 0.07778166979551315, "global_step": 176225, "epoch": 1980, "lr": 2.750293715766537e-05}
{"train_loss": 0.057986240833997726, "global_step": 176226, "epoch": 1980, "lr": 2.7502419420677156e-05}
{"train_loss": 0.11364959925413132, "global_step": 176227, "epoch": 1980, "lr": 2.750190168671345e-05}
{"train_loss": 0.07247661799192429, "global_step": 176228, "epoch": 1980, "lr": 2.7501383955774357e-05}
{"train_loss": 0.08539433777332306, "global_step": 176229, "epoch": 1980, "lr": 2.7500866227859913e-05}
{"train_loss": 0.05808340385556221, "global_step": 176230, "epoch": 1980, "lr": 2.750034850297022e-05}
{"train_loss": 0.06936785578727722, "global_step": 176231, "epoch": 1980, "lr": 2.7499830781105306e-05}
{"train_loss": 0.09269692748785019, "global_step": 176232, "epoch": 1980, "lr": 2.7499313062265285e-05}
{"train_loss": 0.059492167085409164, "global_step": 176233, "epoch": 1980, "lr": 2.7498795346450195e-05}
{"train_loss": 0.07095099240541458, "global_step": 176234, "epoch": 1980, "lr": 2.7498277633660118e-05}
{"train_loss": 0.04872288927435875, "global_step": 176235, "epoch": 1980, "lr": 2.749775992389514e-05}
{"train_loss": 0.08431877195835114, "global_step": 176236, "epoch": 1980, "lr": 2.7497242217155293e-05}
{"train_loss": 0.0781649574637413, "global_step": 176237, "epoch": 1980, "lr": 2.7496724513440686e-05}
{"train_loss": 0.054543811827898026, "global_step": 176238, "epoch": 1980, "lr": 2.7496206812751358e-05}
{"train_loss": 0.09570910781621933, "global_step": 176239, "epoch": 1980, "lr": 2.7495689115087398e-05}
{"train_loss": 0.02931496687233448, "global_step": 176240, "epoch": 1980, "lr": 2.7495171420448877e-05}
{"train_loss": 0.08525253087282181, "global_step": 176241, "epoch": 1980, "lr": 2.749465372883584e-05}
{"train_loss": 0.04906569793820381, "global_step": 176242, "epoch": 1980, "lr": 2.7494136040248387e-05}
{"train_loss": 0.09762278944253922, "global_step": 176243, "epoch": 1980, "lr": 2.7493618354686546e-05}
{"train_loss": 0.052238110452890396, "global_step": 176244, "epoch": 1980, "lr": 2.749310067215044e-05}
{"train_loss": 0.07886437326669693, "global_step": 176245, "epoch": 1980, "lr": 2.74925829926401e-05}
{"train_loss": 0.048230212181806564, "global_step": 176246, "epoch": 1980, "lr": 2.7492065316155618e-05}
{"train_loss": 0.11740012466907501, "global_step": 176247, "epoch": 1980, "lr": 2.7491547642697058e-05}
{"train_loss": 0.0605960488319397, "global_step": 176248, "epoch": 1980, "lr": 2.749102997226446e-05}
{"train_loss": 0.04434244707226753, "global_step": 176249, "epoch": 1980, "lr": 2.7490512304857942e-05}
{"train_loss": 0.07537428289651871, "global_step": 176250, "epoch": 1980, "lr": 2.7489994640477535e-05}
{"train_loss": 0.08641143143177032, "global_step": 176251, "epoch": 1980, "lr": 2.7489476979123337e-05}
{"train_loss": 0.08979038894176483, "global_step": 176252, "epoch": 1980, "lr": 2.7488959320795384e-05}
{"train_loss": 0.05099543184041977, "global_step": 176253, "epoch": 1980, "lr": 2.7488441665493768e-05}
{"train_loss": 0.11535482108592987, "global_step": 176254, "epoch": 1980, "lr": 2.748792401321858e-05}
{"train_loss": 0.1058787852525711, "global_step": 176255, "epoch": 1980, "lr": 2.7487406363969848e-05}
{"train_loss": 0.1689683049917221, "global_step": 176256, "epoch": 1980, "lr": 2.7486888717747672e-05}
{"train_loss": 0.06011701375246048, "global_step": 176257, "epoch": 1980, "lr": 2.748637107455209e-05}
{"train_loss": 0.11650411039590836, "global_step": 176258, "epoch": 1980, "lr": 2.7485853434383214e-05}
{"train_loss": 0.08950591832399368, "global_step": 176259, "epoch": 1980, "lr": 2.7485335797241073e-05}
{"train_loss": 0.06173919141292572, "global_step": 176260, "epoch": 1980, "lr": 2.748481816312577e-05}
{"train_loss": 0.05737592652440071, "global_step": 176261, "epoch": 1980, "lr": 2.7484300532037343e-05}
{"train_loss": 0.08309762179851532, "global_step": 176262, "epoch": 1980, "lr": 2.7483782903975896e-05}
{"train_loss": 0.06270068138837814, "global_step": 176263, "epoch": 1980, "lr": 2.7483265278941457e-05}
{"train_loss": 0.16205444931983948, "global_step": 176264, "epoch": 1980, "lr": 2.7482747656934138e-05}
{"train_loss": 0.10009102523326874, "global_step": 176265, "epoch": 1980, "lr": 2.7482230037953972e-05}
{"train_loss": 0.04480354115366936, "global_step": 176266, "epoch": 1980, "lr": 2.7481712422001065e-05}
{"train_loss": 0.0986701101064682, "global_step": 176267, "epoch": 1980, "lr": 2.748119480907545e-05}
{"train_loss": 0.08523830771446228, "global_step": 176268, "epoch": 1980, "lr": 2.748067719917723e-05}
{"train_loss": 0.09696311503648758, "global_step": 176269, "epoch": 1980, "lr": 2.7480159592306443e-05}
{"train_loss": 0.09688474237918854, "global_step": 176270, "epoch": 1980, "lr": 2.7479641988463166e-05}
{"train_loss": 0.12146840244531631, "global_step": 176271, "epoch": 1980, "lr": 2.7479124387647503e-05}
{"train_loss": 0.05959181860089302, "global_step": 176272, "epoch": 1980, "lr": 2.7478606789859475e-05}
{"train_loss": 0.08753854781389236, "global_step": 176273, "epoch": 1980, "lr": 2.74780891950992e-05}
{"train_loss": 0.12823878228664398, "global_step": 176274, "epoch": 1980, "lr": 2.747757160336669e-05}
{"train_loss": 0.12288139015436172, "global_step": 176275, "epoch": 1980, "lr": 2.7477054014662075e-05}
{"train_loss": 0.10136684775352478, "global_step": 176276, "epoch": 1980, "lr": 2.747653642898537e-05}
{"train_loss": 0.06158769130706787, "global_step": 176277, "epoch": 1980, "lr": 2.747601884633669e-05}
{"train_loss": 0.08967409282922745, "global_step": 176278, "epoch": 1980, "lr": 2.7475501266716076e-05}
{"train_loss": 0.0677778571844101, "global_step": 176279, "epoch": 1980, "lr": 2.747498369012361e-05}
{"train_loss": 0.10142937302589417, "global_step": 176280, "epoch": 1980, "lr": 2.7474466116559373e-05}
{"train_loss": 0.03317510709166527, "global_step": 176281, "epoch": 1980, "lr": 2.7473948546023387e-05}
{"train_loss": 0.045650262385606766, "global_step": 176282, "epoch": 1980, "lr": 2.747343097851578e-05}
{"train_loss": 0.061527661979198456, "global_step": 176283, "epoch": 1980, "lr": 2.747291341403657e-05}
{"train_loss": 0.02162741683423519, "global_step": 176284, "epoch": 1980, "lr": 2.7472395852585886e-05}
{"train_loss": 0.09201361238956451, "global_step": 176285, "epoch": 1980, "lr": 2.7471878294163755e-05}
{"train_loss": 0.06676068156957626, "global_step": 176286, "epoch": 1980, "lr": 2.7471360738770234e-05}
{"train_loss": 0.12477505952119827, "global_step": 176287, "epoch": 1980, "lr": 2.747084318640544e-05}
{"train_loss": 0.029363712295889854, "global_step": 176288, "epoch": 1980, "lr": 2.7470325637069382e-05}
{"train_loss": 0.07163391262292862, "global_step": 176289, "epoch": 1980, "lr": 2.7469808090762177e-05}
{"train_loss": 0.08334092050790787, "global_step": 176290, "epoch": 1980, "lr": 2.7469290547483893e-05}
{"train_loss": 0.06102897226810455, "global_step": 176291, "epoch": 1980, "lr": 2.7468773007234583e-05}
{"train_loss": 0.07745672017335892, "global_step": 176292, "epoch": 1980, "lr": 2.7468255470014326e-05}
{"train_loss": 0.0880725234746933, "global_step": 176293, "epoch": 1980, "lr": 2.746773793582318e-05}
{"train_loss": 0.09790323674678802, "global_step": 176294, "epoch": 1980, "lr": 2.746722040466123e-05}
{"train_loss": 0.08598758280277252, "global_step": 176295, "epoch": 1980, "lr": 2.7466702876528526e-05}
{"train_loss": 0.12673017382621765, "global_step": 176296, "epoch": 1980, "lr": 2.7466185351425168e-05}
{"train_loss": 0.07782874256372452, "global_step": 176297, "epoch": 1980, "lr": 2.7465667829351182e-05}
{"train_loss": 0.059595316648483276, "global_step": 176298, "epoch": 1980, "lr": 2.7465150310306687e-05}
{"train_loss": 0.07024314254522324, "global_step": 176299, "epoch": 1980, "lr": 2.746463279429171e-05}
{"train_loss": 0.0871863141655922, "global_step": 176300, "epoch": 1980, "lr": 2.7464115281306358e-05}
{"train_loss": 0.09983023256063461, "global_step": 176301, "epoch": 1980, "lr": 2.7463597771350656e-05}
{"train_loss": 0.11463774740695953, "global_step": 176302, "epoch": 1980, "lr": 2.7463080264424723e-05}
{"train_loss": 0.0889902114868164, "global_step": 176303, "epoch": 1980, "lr": 2.7462562760528576e-05}
{"train_loss": 0.0491546131670475, "global_step": 176304, "epoch": 1980, "lr": 2.746204525966234e-05}
{"train_loss": 0.0764327123761177, "global_step": 176305, "epoch": 1980, "lr": 2.746152776182604e-05}
{"train_loss": 0.0576518289744854, "global_step": 176306, "epoch": 1980, "lr": 2.7461010267019775e-05}
{"train_loss": 0.0939975455403328, "global_step": 176307, "epoch": 1980, "lr": 2.7460492775243592e-05}
{"train_loss": 0.08028487352591553, "global_step": 176308, "epoch": 1980, "lr": 2.7459975286497564e-05, "val_loss": 7.347859859466553, "train_action_mse_error": 8.542762756347656}
{"train_loss": 0.10479546338319778, "global_step": 176309, "epoch": 1981, "lr": 2.7459457800781786e-05}
{"train_loss": 0.08943302184343338, "global_step": 176310, "epoch": 1981, "lr": 2.7458940318096293e-05}
{"train_loss": 0.13479049503803253, "global_step": 176311, "epoch": 1981, "lr": 2.745842283844119e-05}
{"train_loss": 0.083924800157547, "global_step": 176312, "epoch": 1981, "lr": 2.7457905361816504e-05}
{"train_loss": 0.08805767446756363, "global_step": 176313, "epoch": 1981, "lr": 2.7457387888222354e-05}
{"train_loss": 0.03066130541265011, "global_step": 176314, "epoch": 1981, "lr": 2.7456870417658766e-05}
{"train_loss": 0.0534113310277462, "global_step": 176315, "epoch": 1981, "lr": 2.745635295012584e-05}
{"train_loss": 0.09058084338903427, "global_step": 176316, "epoch": 1981, "lr": 2.7455835485623614e-05}
{"train_loss": 0.12302937358617783, "global_step": 176317, "epoch": 1981, "lr": 2.7455318024152193e-05}
{"train_loss": 0.10673060268163681, "global_step": 176318, "epoch": 1981, "lr": 2.7454800565711635e-05}
{"train_loss": 0.04400329664349556, "global_step": 176319, "epoch": 1981, "lr": 2.745428311030198e-05}
{"train_loss": 0.04403194785118103, "global_step": 176320, "epoch": 1981, "lr": 2.7453765657923347e-05}
{"train_loss": 0.05543263629078865, "global_step": 176321, "epoch": 1981, "lr": 2.7453248208575756e-05}
{"train_loss": 0.09609870612621307, "global_step": 176322, "epoch": 1981, "lr": 2.7452730762259326e-05}
{"train_loss": 0.07841284573078156, "global_step": 176323, "epoch": 1981, "lr": 2.7452213318974095e-05}
{"train_loss": 0.10910946875810623, "global_step": 176324, "epoch": 1981, "lr": 2.7451695878720124e-05}
{"train_loss": 0.1041635274887085, "global_step": 176325, "epoch": 1981, "lr": 2.7451178441497495e-05}
{"train_loss": 0.06792452186346054, "global_step": 176326, "epoch": 1981, "lr": 2.74506610073063e-05}
{"train_loss": 0.058737993240356445, "global_step": 176327, "epoch": 1981, "lr": 2.745014357614657e-05}
{"train_loss": 0.09857858717441559, "global_step": 176328, "epoch": 1981, "lr": 2.7449626148018415e-05}
{"train_loss": 0.056945350021123886, "global_step": 176329, "epoch": 1981, "lr": 2.7449108722921858e-05}
{"train_loss": 0.0446920171380043, "global_step": 176330, "epoch": 1981, "lr": 2.7448591300857018e-05}
{"train_loss": 0.04647757112979889, "global_step": 176331, "epoch": 1981, "lr": 2.7448073881823923e-05}
{"train_loss": 0.061341963708400726, "global_step": 176332, "epoch": 1981, "lr": 2.744755646582267e-05}
{"train_loss": 0.09243949502706528, "global_step": 176333, "epoch": 1981, "lr": 2.7447039052853308e-05}
{"train_loss": 0.045337412506341934, "global_step": 176334, "epoch": 1981, "lr": 2.744652164291593e-05}
{"train_loss": 0.10184559226036072, "global_step": 176335, "epoch": 1981, "lr": 2.744600423601057e-05}
{"train_loss": 0.05671811103820801, "global_step": 176336, "epoch": 1981, "lr": 2.7445486832137346e-05}
{"train_loss": 0.058484550565481186, "global_step": 176337, "epoch": 1981, "lr": 2.7444969431296286e-05}
{"train_loss": 0.061082303524017334, "global_step": 176338, "epoch": 1981, "lr": 2.7444452033487485e-05}
{"train_loss": 0.04561469703912735, "global_step": 176339, "epoch": 1981, "lr": 2.7443934638710987e-05}
{"train_loss": 0.10381802916526794, "global_step": 176340, "epoch": 1981, "lr": 2.7443417246966895e-05}
{"train_loss": 0.1243559792637825, "global_step": 176341, "epoch": 1981, "lr": 2.7442899858255244e-05}
{"train_loss": 0.04207978397607803, "global_step": 176342, "epoch": 1981, "lr": 2.7442382472576135e-05}
{"train_loss": 0.11996565014123917, "global_step": 176343, "epoch": 1981, "lr": 2.7441865089929606e-05}
{"train_loss": 0.10882829874753952, "global_step": 176344, "epoch": 1981, "lr": 2.7441347710315744e-05}
{"train_loss": 0.04990975931286812, "global_step": 176345, "epoch": 1981, "lr": 2.7440830333734636e-05}
{"train_loss": 0.04206303507089615, "global_step": 176346, "epoch": 1981, "lr": 2.7440312960186315e-05}
{"train_loss": 0.07551462203264236, "global_step": 176347, "epoch": 1981, "lr": 2.743979558967089e-05}
{"train_loss": 0.0947418063879013, "global_step": 176348, "epoch": 1981, "lr": 2.7439278222188387e-05}
{"train_loss": 0.07614705711603165, "global_step": 176349, "epoch": 1981, "lr": 2.743876085773892e-05}
{"train_loss": 0.12207584828138351, "global_step": 176350, "epoch": 1981, "lr": 2.743824349632251e-05}
{"train_loss": 0.044483933597803116, "global_step": 176351, "epoch": 1981, "lr": 2.743772613793928e-05}
{"train_loss": 0.07347822934389114, "global_step": 176352, "epoch": 1981, "lr": 2.743720878258925e-05}
{"train_loss": 0.04559645801782608, "global_step": 176353, "epoch": 1981, "lr": 2.7436691430272534e-05}
{"train_loss": 0.07287462800741196, "global_step": 176354, "epoch": 1981, "lr": 2.7436174080989163e-05}
{"train_loss": 0.11126089841127396, "global_step": 176355, "epoch": 1981, "lr": 2.7435656734739233e-05}
{"train_loss": 0.13416807353496552, "global_step": 176356, "epoch": 1981, "lr": 2.7435139391522797e-05}
{"train_loss": 0.09442251175642014, "global_step": 176357, "epoch": 1981, "lr": 2.743462205133994e-05}
{"train_loss": 0.045262660831213, "global_step": 176358, "epoch": 1981, "lr": 2.7434104714190734e-05}
{"train_loss": 0.09680119156837463, "global_step": 176359, "epoch": 1981, "lr": 2.743358738007521e-05}
{"train_loss": 0.05059855431318283, "global_step": 176360, "epoch": 1981, "lr": 2.7433070048993488e-05}
{"train_loss": 0.047427840530872345, "global_step": 176361, "epoch": 1981, "lr": 2.7432552720945592e-05}
{"train_loss": 0.0718640461564064, "global_step": 176362, "epoch": 1981, "lr": 2.743203539593164e-05}
{"train_loss": 0.12183179706335068, "global_step": 176363, "epoch": 1981, "lr": 2.7431518073951644e-05}
{"train_loss": 0.11991618573665619, "global_step": 176364, "epoch": 1981, "lr": 2.7431000755005737e-05}
{"train_loss": 0.07870274037122726, "global_step": 176365, "epoch": 1981, "lr": 2.7430483439093936e-05}
{"train_loss": 0.07695827633142471, "global_step": 176366, "epoch": 1981, "lr": 2.7429966126216346e-05}
{"train_loss": 0.0906306803226471, "global_step": 176367, "epoch": 1981, "lr": 2.7429448816373004e-05}
{"train_loss": 0.12738095223903656, "global_step": 176368, "epoch": 1981, "lr": 2.7428931509564015e-05}
{"train_loss": 0.08153875172138214, "global_step": 176369, "epoch": 1981, "lr": 2.7428414205789414e-05}
{"train_loss": 0.08293324708938599, "global_step": 176370, "epoch": 1981, "lr": 2.742789690504931e-05}
{"train_loss": 0.06514476984739304, "global_step": 176371, "epoch": 1981, "lr": 2.7427379607343728e-05}
{"train_loss": 0.10018102824687958, "global_step": 176372, "epoch": 1981, "lr": 2.742686231267278e-05}
{"train_loss": 0.1599048525094986, "global_step": 176373, "epoch": 1981, "lr": 2.7426345021036492e-05}
{"train_loss": 0.05919777601957321, "global_step": 176374, "epoch": 1981, "lr": 2.7425827732434985e-05}
{"train_loss": 0.08787520229816437, "global_step": 176375, "epoch": 1981, "lr": 2.7425310446868268e-05}
{"train_loss": 0.059295061975717545, "global_step": 176376, "epoch": 1981, "lr": 2.742479316433647e-05}
{"train_loss": 0.07477398216724396, "global_step": 176377, "epoch": 1981, "lr": 2.742427588483961e-05}
{"train_loss": 0.11971430480480194, "global_step": 176378, "epoch": 1981, "lr": 2.7423758608377808e-05}
{"train_loss": 0.08572255074977875, "global_step": 176379, "epoch": 1981, "lr": 2.7423241334951082e-05}
{"train_loss": 0.07703615725040436, "global_step": 176380, "epoch": 1981, "lr": 2.7422724064559523e-05}
{"train_loss": 0.058752622455358505, "global_step": 176381, "epoch": 1981, "lr": 2.7422206797203232e-05}
{"train_loss": 0.05937822908163071, "global_step": 176382, "epoch": 1981, "lr": 2.7421689532882222e-05}
{"train_loss": 0.08390122652053833, "global_step": 176383, "epoch": 1981, "lr": 2.742117227159662e-05}
{"train_loss": 0.07681475579738617, "global_step": 176384, "epoch": 1981, "lr": 2.7420655013346437e-05}
{"train_loss": 0.0909622460603714, "global_step": 176385, "epoch": 1981, "lr": 2.742013775813179e-05}
{"train_loss": 0.034936677664518356, "global_step": 176386, "epoch": 1981, "lr": 2.7419620505952714e-05}
{"train_loss": 0.08988213539123535, "global_step": 176387, "epoch": 1981, "lr": 2.741910325680932e-05}
{"train_loss": 0.03720594942569733, "global_step": 176388, "epoch": 1981, "lr": 2.741858601070163e-05}
{"train_loss": 0.04951239749789238, "global_step": 176389, "epoch": 1981, "lr": 2.741806876762975e-05}
{"train_loss": 0.08112387359142303, "global_step": 176390, "epoch": 1981, "lr": 2.7417551527593722e-05}
{"train_loss": 0.038705311715602875, "global_step": 176391, "epoch": 1981, "lr": 2.7417034290593645e-05}
{"train_loss": 0.03049052506685257, "global_step": 176392, "epoch": 1981, "lr": 2.7416517056629554e-05}
{"train_loss": 0.10944787412881851, "global_step": 176393, "epoch": 1981, "lr": 2.7415999825701554e-05}
{"train_loss": 0.08546219766139984, "global_step": 176394, "epoch": 1981, "lr": 2.7415482597809684e-05}
{"train_loss": 0.09998351335525513, "global_step": 176395, "epoch": 1981, "lr": 2.7414965372954036e-05}
{"train_loss": 0.10716316848993301, "global_step": 176396, "epoch": 1981, "lr": 2.7414448151134687e-05}
{"train_loss": 0.07914966815726811, "global_step": 176397, "epoch": 1981, "lr": 2.7413930932351663e-05, "val_loss": 7.379276752471924}
{"train_loss": 0.06619168817996979, "global_step": 176398, "epoch": 1982, "lr": 2.7413413716605076e-05}
{"train_loss": 0.031603824347257614, "global_step": 176399, "epoch": 1982, "lr": 2.741289650389496e-05}
{"train_loss": 0.07632087171077728, "global_step": 176400, "epoch": 1982, "lr": 2.741237929422143e-05}
{"train_loss": 0.0536021813750267, "global_step": 176401, "epoch": 1982, "lr": 2.7411862087584505e-05}
{"train_loss": 0.11486034095287323, "global_step": 176402, "epoch": 1982, "lr": 2.7411344883984302e-05}
{"train_loss": 0.16358384490013123, "global_step": 176403, "epoch": 1982, "lr": 2.741082768342085e-05}
{"train_loss": 0.1501842737197876, "global_step": 176404, "epoch": 1982, "lr": 2.7410310485894258e-05}
{"train_loss": 0.06441409885883331, "global_step": 176405, "epoch": 1982, "lr": 2.7409793291404552e-05}
{"train_loss": 0.07532824575901031, "global_step": 176406, "epoch": 1982, "lr": 2.7409276099951846e-05}
{"train_loss": 0.08863231539726257, "global_step": 176407, "epoch": 1982, "lr": 2.740875891153617e-05}
{"train_loss": 0.09323600679636002, "global_step": 176408, "epoch": 1982, "lr": 2.7408241726157624e-05}
{"train_loss": 0.072898268699646, "global_step": 176409, "epoch": 1982, "lr": 2.7407724543816247e-05}
{"train_loss": 0.04144623503088951, "global_step": 176410, "epoch": 1982, "lr": 2.7407207364512145e-05}
{"train_loss": 0.11122152209281921, "global_step": 176411, "epoch": 1982, "lr": 2.7406690188245353e-05}
{"train_loss": 0.05001569539308548, "global_step": 176412, "epoch": 1982, "lr": 2.7406173015015975e-05}
{"train_loss": 0.08462076634168625, "global_step": 176413, "epoch": 1982, "lr": 2.7405655844824042e-05}
{"train_loss": 0.06157238036394119, "global_step": 176414, "epoch": 1982, "lr": 2.7405138677669662e-05}
{"train_loss": 0.11622824519872665, "global_step": 176415, "epoch": 1982, "lr": 2.7404621513552863e-05}
{"train_loss": 0.03184036910533905, "global_step": 176416, "epoch": 1982, "lr": 2.7404104352473735e-05}
{"train_loss": 0.08392118662595749, "global_step": 176417, "epoch": 1982, "lr": 2.740358719443238e-05}
{"train_loss": 0.13312551379203796, "global_step": 176418, "epoch": 1982, "lr": 2.7403070039428808e-05}
{"train_loss": 0.07232017815113068, "global_step": 176419, "epoch": 1982, "lr": 2.740255288746314e-05}
{"train_loss": 0.09719077497720718, "global_step": 176420, "epoch": 1982, "lr": 2.7402035738535405e-05}
{"train_loss": 0.05818549543619156, "global_step": 176421, "epoch": 1982, "lr": 2.7401518592645715e-05}
{"train_loss": 0.10533429682254791, "global_step": 176422, "epoch": 1982, "lr": 2.7401001449794085e-05}
{"train_loss": 0.1055588573217392, "global_step": 176423, "epoch": 1982, "lr": 2.7400484309980644e-05}
{"train_loss": 0.0733254998922348, "global_step": 176424, "epoch": 1982, "lr": 2.739996717320541e-05}
{"train_loss": 0.08693249523639679, "global_step": 176425, "epoch": 1982, "lr": 2.7399450039468487e-05}
{"train_loss": 0.02779504843056202, "global_step": 176426, "epoch": 1982, "lr": 2.7398932908769927e-05}
{"train_loss": 0.0772659182548523, "global_step": 176427, "epoch": 1982, "lr": 2.7398415781109816e-05}
{"train_loss": 0.08653968572616577, "global_step": 176428, "epoch": 1982, "lr": 2.7397898656488195e-05}
{"train_loss": 0.0557558573782444, "global_step": 176429, "epoch": 1982, "lr": 2.7397381534905163e-05}
{"train_loss": 0.10431928187608719, "global_step": 176430, "epoch": 1982, "lr": 2.7396864416360766e-05}
{"train_loss": 0.06786347180604935, "global_step": 176431, "epoch": 1982, "lr": 2.7396347300855108e-05}
{"train_loss": 0.11342833936214447, "global_step": 176432, "epoch": 1982, "lr": 2.739583018838821e-05}
{"train_loss": 0.08414125442504883, "global_step": 176433, "epoch": 1982, "lr": 2.739531307896019e-05}
{"train_loss": 0.045103926211595535, "global_step": 176434, "epoch": 1982, "lr": 2.7394795972571098e-05}
{"train_loss": 0.0654049664735794, "global_step": 176435, "epoch": 1982, "lr": 2.7394278869220968e-05}
{"train_loss": 0.10855024307966232, "global_step": 176436, "epoch": 1982, "lr": 2.739376176890993e-05}
{"train_loss": 0.056500062346458435, "global_step": 176437, "epoch": 1982, "lr": 2.739324467163801e-05}
{"train_loss": 0.10558882355690002, "global_step": 176438, "epoch": 1982, "lr": 2.7392727577405304e-05}
{"train_loss": 0.02196221426129341, "global_step": 176439, "epoch": 1982, "lr": 2.7392210486211856e-05}
{"train_loss": 0.08269628137350082, "global_step": 176440, "epoch": 1982, "lr": 2.739169339805776e-05}
{"train_loss": 0.07976410537958145, "global_step": 176441, "epoch": 1982, "lr": 2.739117631294307e-05}
{"train_loss": 0.04994310438632965, "global_step": 176442, "epoch": 1982, "lr": 2.7390659230867866e-05}
{"train_loss": 0.14290009438991547, "global_step": 176443, "epoch": 1982, "lr": 2.73901421518322e-05}
{"train_loss": 0.0745750367641449, "global_step": 176444, "epoch": 1982, "lr": 2.7389625075836172e-05}
{"train_loss": 0.06847072392702103, "global_step": 176445, "epoch": 1982, "lr": 2.7389108002879814e-05}
{"train_loss": 0.08355104178190231, "global_step": 176446, "epoch": 1982, "lr": 2.7388590932963237e-05}
{"train_loss": 0.08901273459196091, "global_step": 176447, "epoch": 1982, "lr": 2.7388073866086468e-05}
{"train_loss": 0.06518977135419846, "global_step": 176448, "epoch": 1982, "lr": 2.7387556802249615e-05}
{"train_loss": 0.06953893601894379, "global_step": 176449, "epoch": 1982, "lr": 2.7387039741452713e-05}
{"train_loss": 0.07454981654882431, "global_step": 176450, "epoch": 1982, "lr": 2.738652268369586e-05}
{"train_loss": 0.06089287996292114, "global_step": 176451, "epoch": 1982, "lr": 2.73860056289791e-05}
{"train_loss": 0.0851377472281456, "global_step": 176452, "epoch": 1982, "lr": 2.7385488577302516e-05}
{"train_loss": 0.0603906586766243, "global_step": 176453, "epoch": 1982, "lr": 2.7384971528666192e-05}
{"train_loss": 0.05852365866303444, "global_step": 176454, "epoch": 1982, "lr": 2.7384454483070167e-05}
{"train_loss": 0.09706304222345352, "global_step": 176455, "epoch": 1982, "lr": 2.7383937440514546e-05}
{"train_loss": 0.0815163105726242, "global_step": 176456, "epoch": 1982, "lr": 2.738342040099936e-05}
{"train_loss": 0.06824621558189392, "global_step": 176457, "epoch": 1982, "lr": 2.7382903364524715e-05}
{"train_loss": 0.120670847594738, "global_step": 176458, "epoch": 1982, "lr": 2.738238633109065e-05}
{"train_loss": 0.08800730854272842, "global_step": 176459, "epoch": 1982, "lr": 2.738186930069726e-05}
{"train_loss": 0.10023605078458786, "global_step": 176460, "epoch": 1982, "lr": 2.738135227334458e-05}
{"train_loss": 0.07887325435876846, "global_step": 176461, "epoch": 1982, "lr": 2.7380835249032728e-05}
{"train_loss": 0.12749166786670685, "global_step": 176462, "epoch": 1982, "lr": 2.7380318227761727e-05}
{"train_loss": 0.037762463092803955, "global_step": 176463, "epoch": 1982, "lr": 2.737980120953168e-05}
{"train_loss": 0.06021079048514366, "global_step": 176464, "epoch": 1982, "lr": 2.7379284194342626e-05}
{"train_loss": 0.08231750875711441, "global_step": 176465, "epoch": 1982, "lr": 2.7378767182194675e-05}
{"train_loss": 0.07416827231645584, "global_step": 176466, "epoch": 1982, "lr": 2.737825017308785e-05}
{"train_loss": 0.0850776806473732, "global_step": 176467, "epoch": 1982, "lr": 2.7377733167022264e-05}
{"train_loss": 0.11864230036735535, "global_step": 176468, "epoch": 1982, "lr": 2.7377216163997942e-05}
{"train_loss": 0.056321900337934494, "global_step": 176469, "epoch": 1982, "lr": 2.7376699164015007e-05}
{"train_loss": 0.0812000036239624, "global_step": 176470, "epoch": 1982, "lr": 2.7376182167073472e-05}
{"train_loss": 0.11266902089118958, "global_step": 176471, "epoch": 1982, "lr": 2.7375665173173454e-05}
{"train_loss": 0.08668716251850128, "global_step": 176472, "epoch": 1982, "lr": 2.737514818231498e-05}
{"train_loss": 0.10577798634767532, "global_step": 176473, "epoch": 1982, "lr": 2.7374631194498167e-05}
{"train_loss": 0.05899188295006752, "global_step": 176474, "epoch": 1982, "lr": 2.7374114209723055e-05}
{"train_loss": 0.2081739902496338, "global_step": 176475, "epoch": 1982, "lr": 2.73735972279897e-05}
{"train_loss": 0.04906352981925011, "global_step": 176476, "epoch": 1982, "lr": 2.737308024929821e-05}
{"train_loss": 0.08655822277069092, "global_step": 176477, "epoch": 1982, "lr": 2.737256327364861e-05}
{"train_loss": 0.0671011358499527, "global_step": 176478, "epoch": 1982, "lr": 2.7372046301041023e-05}
{"train_loss": 0.12443789094686508, "global_step": 176479, "epoch": 1982, "lr": 2.7371529331475454e-05}
{"train_loss": 0.060028139501810074, "global_step": 176480, "epoch": 1982, "lr": 2.7371012364952032e-05}
{"train_loss": 0.09650292247533798, "global_step": 176481, "epoch": 1982, "lr": 2.7370495401470786e-05}
{"train_loss": 0.07343139499425888, "global_step": 176482, "epoch": 1982, "lr": 2.7369978441031818e-05}
{"train_loss": 0.06155932694673538, "global_step": 176483, "epoch": 1982, "lr": 2.736946148363516e-05}
{"train_loss": 0.06905922293663025, "global_step": 176484, "epoch": 1982, "lr": 2.7368944529280926e-05}
{"train_loss": 0.1270686835050583, "global_step": 176485, "epoch": 1982, "lr": 2.7368427577969134e-05}
{"train_loss": 0.08245990146997939, "global_step": 176486, "epoch": 1982, "lr": 2.7367910629699912e-05, "val_loss": 7.3630852699279785}
{"train_loss": 0.11390512436628342, "global_step": 176487, "epoch": 1983, "lr": 2.736739368447327e-05}
{"train_loss": 0.0770903080701828, "global_step": 176488, "epoch": 1983, "lr": 2.7366876742289305e-05}
{"train_loss": 0.09504790604114532, "global_step": 176489, "epoch": 1983, "lr": 2.7366359803148113e-05}
{"train_loss": 0.08048610389232635, "global_step": 176490, "epoch": 1983, "lr": 2.7365842867049718e-05}
{"train_loss": 0.040830932557582855, "global_step": 176491, "epoch": 1983, "lr": 2.7365325933994224e-05}
{"train_loss": 0.07350365817546844, "global_step": 176492, "epoch": 1983, "lr": 2.7364809003981672e-05}
{"train_loss": 0.12075482308864594, "global_step": 176493, "epoch": 1983, "lr": 2.7364292077012167e-05}
{"train_loss": 0.10370049625635147, "global_step": 176494, "epoch": 1983, "lr": 2.736377515308573e-05}
{"train_loss": 0.06298892199993134, "global_step": 176495, "epoch": 1983, "lr": 2.7363258232202483e-05}
{"train_loss": 0.06946215033531189, "global_step": 176496, "epoch": 1983, "lr": 2.7362741314362445e-05}
{"train_loss": 0.04427967220544815, "global_step": 176497, "epoch": 1983, "lr": 2.7362224399565732e-05}
{"train_loss": 0.08220788836479187, "global_step": 176498, "epoch": 1983, "lr": 2.736170748781237e-05}
{"train_loss": 0.08233446627855301, "global_step": 176499, "epoch": 1983, "lr": 2.736119057910248e-05}
{"train_loss": 0.04074579104781151, "global_step": 176500, "epoch": 1983, "lr": 2.7360673673436076e-05}
{"train_loss": 0.09270883351564407, "global_step": 176501, "epoch": 1983, "lr": 2.7360156770813273e-05}
{"train_loss": 0.05991878733038902, "global_step": 176502, "epoch": 1983, "lr": 2.73596398712341e-05}
{"train_loss": 0.0853586494922638, "global_step": 176503, "epoch": 1983, "lr": 2.735912297469867e-05}
{"train_loss": 0.05989895388484001, "global_step": 176504, "epoch": 1983, "lr": 2.7358606081207006e-05}
{"train_loss": 0.048359327018260956, "global_step": 176505, "epoch": 1983, "lr": 2.7358089190759228e-05}
{"train_loss": 0.04558669775724411, "global_step": 176506, "epoch": 1983, "lr": 2.735757230335535e-05}
{"train_loss": 0.05590192601084709, "global_step": 176507, "epoch": 1983, "lr": 2.7357055418995503e-05}
{"train_loss": 0.03711001202464104, "global_step": 176508, "epoch": 1983, "lr": 2.7356538537679695e-05}
{"train_loss": 0.07828931510448456, "global_step": 176509, "epoch": 1983, "lr": 2.7356021659408042e-05}
{"train_loss": 0.04296932369470596, "global_step": 176510, "epoch": 1983, "lr": 2.735550478418058e-05}
{"train_loss": 0.07392663508653641, "global_step": 176511, "epoch": 1983, "lr": 2.7354987911997416e-05}
{"train_loss": 0.08988126367330551, "global_step": 176512, "epoch": 1983, "lr": 2.7354471042858598e-05}
{"train_loss": 0.08097442984580994, "global_step": 176513, "epoch": 1983, "lr": 2.7353954176764175e-05}
{"train_loss": 0.06592293828725815, "global_step": 176514, "epoch": 1983, "lr": 2.7353437313714247e-05}
{"train_loss": 0.04736229032278061, "global_step": 176515, "epoch": 1983, "lr": 2.7352920453708863e-05}
{"train_loss": 0.09159571677446365, "global_step": 176516, "epoch": 1983, "lr": 2.7352403596748122e-05}
{"train_loss": 0.08135991543531418, "global_step": 176517, "epoch": 1983, "lr": 2.7351886742832055e-05}
{"train_loss": 0.10206003487110138, "global_step": 176518, "epoch": 1983, "lr": 2.735136989196077e-05}
{"train_loss": 0.08060634136199951, "global_step": 176519, "epoch": 1983, "lr": 2.7350853044134293e-05}
{"train_loss": 0.10660909861326218, "global_step": 176520, "epoch": 1983, "lr": 2.7350336199352745e-05}
{"train_loss": 0.13492819666862488, "global_step": 176521, "epoch": 1983, "lr": 2.7349819357616147e-05}
{"train_loss": 0.056912653148174286, "global_step": 176522, "epoch": 1983, "lr": 2.7349302518924603e-05}
{"train_loss": 0.04738396406173706, "global_step": 176523, "epoch": 1983, "lr": 2.7348785683278155e-05}
{"train_loss": 0.05289863795042038, "global_step": 176524, "epoch": 1983, "lr": 2.7348268850676882e-05}
{"train_loss": 0.03631526976823807, "global_step": 176525, "epoch": 1983, "lr": 2.7347752021120886e-05}
{"train_loss": 0.059050895273685455, "global_step": 176526, "epoch": 1983, "lr": 2.7347235194610187e-05}
{"train_loss": 0.04400627315044403, "global_step": 176527, "epoch": 1983, "lr": 2.734671837114489e-05}
{"train_loss": 0.041030921041965485, "global_step": 176528, "epoch": 1983, "lr": 2.7346201550725036e-05}
{"train_loss": 0.08001216500997543, "global_step": 176529, "epoch": 1983, "lr": 2.734568473335073e-05}
{"train_loss": 0.05768342316150665, "global_step": 176530, "epoch": 1983, "lr": 2.7345167919021996e-05}
{"train_loss": 0.05648104473948479, "global_step": 176531, "epoch": 1983, "lr": 2.734465110773895e-05}
{"train_loss": 0.05509505420923233, "global_step": 176532, "epoch": 1983, "lr": 2.734413429950162e-05}
{"train_loss": 0.060652513056993484, "global_step": 176533, "epoch": 1983, "lr": 2.7343617494310115e-05}
{"train_loss": 0.1319243460893631, "global_step": 176534, "epoch": 1983, "lr": 2.7343100692164458e-05}
{"train_loss": 0.07656668871641159, "global_step": 176535, "epoch": 1983, "lr": 2.734258389306477e-05}
{"train_loss": 0.08779887109994888, "global_step": 176536, "epoch": 1983, "lr": 2.734206709701108e-05}
{"train_loss": 0.076830193400383, "global_step": 176537, "epoch": 1983, "lr": 2.734155030400348e-05}
{"train_loss": 0.12503103911876678, "global_step": 176538, "epoch": 1983, "lr": 2.7341033514042024e-05}
{"train_loss": 0.09856731444597244, "global_step": 176539, "epoch": 1983, "lr": 2.7340516727126808e-05}
{"train_loss": 0.08288721740245819, "global_step": 176540, "epoch": 1983, "lr": 2.7339999943257865e-05}
{"train_loss": 0.08682601898908615, "global_step": 176541, "epoch": 1983, "lr": 2.7339483162435296e-05}
{"train_loss": 0.06498389691114426, "global_step": 176542, "epoch": 1983, "lr": 2.733896638465915e-05}
{"train_loss": 0.09637171030044556, "global_step": 176543, "epoch": 1983, "lr": 2.7338449609929496e-05}
{"train_loss": 0.04266529157757759, "global_step": 176544, "epoch": 1983, "lr": 2.7337932838246426e-05}
{"train_loss": 0.17303913831710815, "global_step": 176545, "epoch": 1983, "lr": 2.7337416069610002e-05}
{"train_loss": 0.09761961549520493, "global_step": 176546, "epoch": 1983, "lr": 2.7336899304020257e-05}
{"train_loss": 0.06866288185119629, "global_step": 176547, "epoch": 1983, "lr": 2.7336382541477323e-05}
{"train_loss": 0.06061115115880966, "global_step": 176548, "epoch": 1983, "lr": 2.7335865781981208e-05}
{"train_loss": 0.08170080184936523, "global_step": 176549, "epoch": 1983, "lr": 2.7335349025532033e-05}
{"train_loss": 0.09084612131118774, "global_step": 176550, "epoch": 1983, "lr": 2.733483227212984e-05}
{"train_loss": 0.08111315965652466, "global_step": 176551, "epoch": 1983, "lr": 2.7334315521774685e-05}
{"train_loss": 0.12306439131498337, "global_step": 176552, "epoch": 1983, "lr": 2.7333798774466673e-05}
{"train_loss": 0.06802409142255783, "global_step": 176553, "epoch": 1983, "lr": 2.733328203020584e-05}
{"train_loss": 0.07870891690254211, "global_step": 176554, "epoch": 1983, "lr": 2.733276528899229e-05}
{"train_loss": 0.06192034110426903, "global_step": 176555, "epoch": 1983, "lr": 2.7332248550826044e-05}
{"train_loss": 0.07659028470516205, "global_step": 176556, "epoch": 1983, "lr": 2.733173181570723e-05}
{"train_loss": 0.03037596121430397, "global_step": 176557, "epoch": 1983, "lr": 2.733121508363587e-05}
{"train_loss": 0.05289547145366669, "global_step": 176558, "epoch": 1983, "lr": 2.733069835461207e-05}
{"train_loss": 0.09712757170200348, "global_step": 176559, "epoch": 1983, "lr": 2.733018162863586e-05}
{"train_loss": 0.05579619109630585, "global_step": 176560, "epoch": 1983, "lr": 2.732966490570735e-05}
{"train_loss": 0.06818839907646179, "global_step": 176561, "epoch": 1983, "lr": 2.732914818582657e-05}
{"train_loss": 0.10332974046468735, "global_step": 176562, "epoch": 1983, "lr": 2.732863146899361e-05}
{"train_loss": 0.033075470477342606, "global_step": 176563, "epoch": 1983, "lr": 2.732811475520856e-05}
{"train_loss": 0.0857468768954277, "global_step": 176564, "epoch": 1983, "lr": 2.7327598044471446e-05}
{"train_loss": 0.06834013015031815, "global_step": 176565, "epoch": 1983, "lr": 2.732708133678239e-05}
{"train_loss": 0.1041659340262413, "global_step": 176566, "epoch": 1983, "lr": 2.73265646321414e-05}
{"train_loss": 0.05653023719787598, "global_step": 176567, "epoch": 1983, "lr": 2.7326047930548604e-05}
{"train_loss": 0.07612988352775574, "global_step": 176568, "epoch": 1983, "lr": 2.732553123200402e-05}
{"train_loss": 0.04163917899131775, "global_step": 176569, "epoch": 1983, "lr": 2.7325014536507764e-05}
{"train_loss": 0.1233106255531311, "global_step": 176570, "epoch": 1983, "lr": 2.732449784405986e-05}
{"train_loss": 0.05576709657907486, "global_step": 176571, "epoch": 1983, "lr": 2.7323981154660426e-05}
{"train_loss": 0.1110241487622261, "global_step": 176572, "epoch": 1983, "lr": 2.7323464468309485e-05}
{"train_loss": 0.10360511392354965, "global_step": 176573, "epoch": 1983, "lr": 2.7322947785007148e-05}
{"train_loss": 0.05536048486828804, "global_step": 176574, "epoch": 1983, "lr": 2.7322431104753445e-05}
{"train_loss": 0.07598831538069115, "global_step": 176575, "epoch": 1983, "lr": 2.7321914427548478e-05, "val_loss": 7.403538703918457}
{"train_loss": 0.12199769169092178, "global_step": 176576, "epoch": 1984, "lr": 2.7321397753392287e-05}
{"train_loss": 0.08895320445299149, "global_step": 176577, "epoch": 1984, "lr": 2.732088108228498e-05}
{"train_loss": 0.053707003593444824, "global_step": 176578, "epoch": 1984, "lr": 2.7320364414226584e-05}
{"train_loss": 0.05461942404508591, "global_step": 176579, "epoch": 1984, "lr": 2.7319847749217187e-05}
{"train_loss": 0.1355699747800827, "global_step": 176580, "epoch": 1984, "lr": 2.731933108725688e-05}
{"train_loss": 0.05376386642456055, "global_step": 176581, "epoch": 1984, "lr": 2.7318814428345695e-05}
{"train_loss": 0.05741802975535393, "global_step": 176582, "epoch": 1984, "lr": 2.7318297772483735e-05}
{"train_loss": 0.09818152338266373, "global_step": 176583, "epoch": 1984, "lr": 2.731778111967106e-05}
{"train_loss": 0.024801425635814667, "global_step": 176584, "epoch": 1984, "lr": 2.731726446990771e-05}
{"train_loss": 0.053273871541023254, "global_step": 176585, "epoch": 1984, "lr": 2.7316747823193795e-05}
{"train_loss": 0.06277970224618912, "global_step": 176586, "epoch": 1984, "lr": 2.7316231179529344e-05}
{"train_loss": 0.09179709851741791, "global_step": 176587, "epoch": 1984, "lr": 2.7315714538914472e-05}
{"train_loss": 0.03773888200521469, "global_step": 176588, "epoch": 1984, "lr": 2.731519790134921e-05}
{"train_loss": 0.07959417998790741, "global_step": 176589, "epoch": 1984, "lr": 2.7314681266833657e-05}
{"train_loss": 0.12321873754262924, "global_step": 176590, "epoch": 1984, "lr": 2.7314164635367872e-05}
{"train_loss": 0.061483416706323624, "global_step": 176591, "epoch": 1984, "lr": 2.7313648006951904e-05}
{"train_loss": 0.06534746289253235, "global_step": 176592, "epoch": 1984, "lr": 2.7313131381585855e-05}
{"train_loss": 0.16702429950237274, "global_step": 176593, "epoch": 1984, "lr": 2.731261475926976e-05}
{"train_loss": 0.07947058975696564, "global_step": 176594, "epoch": 1984, "lr": 2.7312098140003728e-05}
{"train_loss": 0.19528326392173767, "global_step": 176595, "epoch": 1984, "lr": 2.731158152378779e-05}
{"train_loss": 0.10410868376493454, "global_step": 176596, "epoch": 1984, "lr": 2.7311064910622043e-05}
{"train_loss": 0.08695544302463531, "global_step": 176597, "epoch": 1984, "lr": 2.7310548300506534e-05}
{"train_loss": 0.046856898814439774, "global_step": 176598, "epoch": 1984, "lr": 2.731003169344135e-05}
{"train_loss": 0.0895734652876854, "global_step": 176599, "epoch": 1984, "lr": 2.730951508942657e-05}
{"train_loss": 0.06621252000331879, "global_step": 176600, "epoch": 1984, "lr": 2.7308998488462234e-05}
{"train_loss": 0.06286578625440598, "global_step": 176601, "epoch": 1984, "lr": 2.7308481890548444e-05}
{"train_loss": 0.05539548024535179, "global_step": 176602, "epoch": 1984, "lr": 2.7307965295685224e-05}
{"train_loss": 0.17990167438983917, "global_step": 176603, "epoch": 1984, "lr": 2.73074487038727e-05}
{"train_loss": 0.0900694876909256, "global_step": 176604, "epoch": 1984, "lr": 2.7306932115110895e-05}
{"train_loss": 0.12102766335010529, "global_step": 176605, "epoch": 1984, "lr": 2.730641552939991e-05}
{"train_loss": 0.06649864464998245, "global_step": 176606, "epoch": 1984, "lr": 2.7305898946739784e-05}
{"train_loss": 0.07112208008766174, "global_step": 176607, "epoch": 1984, "lr": 2.730538236713062e-05}
{"train_loss": 0.05867372825741768, "global_step": 176608, "epoch": 1984, "lr": 2.7304865790572453e-05}
{"train_loss": 0.06139602139592171, "global_step": 176609, "epoch": 1984, "lr": 2.7304349217065396e-05}
{"train_loss": 0.06940679997205734, "global_step": 176610, "epoch": 1984, "lr": 2.7303832646609463e-05}
{"train_loss": 0.09871627390384674, "global_step": 176611, "epoch": 1984, "lr": 2.7303316079204777e-05}
{"train_loss": 0.06067696586251259, "global_step": 176612, "epoch": 1984, "lr": 2.730279951485136e-05}
{"train_loss": 0.06999050825834274, "global_step": 176613, "epoch": 1984, "lr": 2.7302282953549335e-05}
{"train_loss": 0.05318242311477661, "global_step": 176614, "epoch": 1984, "lr": 2.7301766395298717e-05}
{"train_loss": 0.0507829450070858, "global_step": 176615, "epoch": 1984, "lr": 2.7301249840099596e-05}
{"train_loss": 0.069344662129879, "global_step": 176616, "epoch": 1984, "lr": 2.7300733287952067e-05}
{"train_loss": 0.06329181790351868, "global_step": 176617, "epoch": 1984, "lr": 2.730021673885616e-05}
{"train_loss": 0.05693987011909485, "global_step": 176618, "epoch": 1984, "lr": 2.7299700192811982e-05}
{"train_loss": 0.07948309183120728, "global_step": 176619, "epoch": 1984, "lr": 2.729918364981956e-05}
{"train_loss": 0.10907012969255447, "global_step": 176620, "epoch": 1984, "lr": 2.7298667109879004e-05}
{"train_loss": 0.13572604954242706, "global_step": 176621, "epoch": 1984, "lr": 2.7298150572990354e-05}
{"train_loss": 0.07513007521629333, "global_step": 176622, "epoch": 1984, "lr": 2.7297634039153702e-05}
{"train_loss": 0.06990017741918564, "global_step": 176623, "epoch": 1984, "lr": 2.7297117508369107e-05}
{"train_loss": 0.14677050709724426, "global_step": 176624, "epoch": 1984, "lr": 2.7296600980636622e-05}
{"train_loss": 0.04808061942458153, "global_step": 176625, "epoch": 1984, "lr": 2.7296084455956356e-05}
{"train_loss": 0.07312452793121338, "global_step": 176626, "epoch": 1984, "lr": 2.7295567934328325e-05}
{"train_loss": 0.10019780695438385, "global_step": 176627, "epoch": 1984, "lr": 2.729505141575266e-05}
{"train_loss": 0.07354220002889633, "global_step": 176628, "epoch": 1984, "lr": 2.7294534900229385e-05}
{"train_loss": 0.06019698083400726, "global_step": 176629, "epoch": 1984, "lr": 2.7294018387758574e-05}
{"train_loss": 0.09222014248371124, "global_step": 176630, "epoch": 1984, "lr": 2.7293501878340323e-05}
{"train_loss": 0.06354298442602158, "global_step": 176631, "epoch": 1984, "lr": 2.729298537197466e-05}
{"train_loss": 0.12251827120780945, "global_step": 176632, "epoch": 1984, "lr": 2.7292468868661702e-05}
{"train_loss": 0.07046554982662201, "global_step": 176633, "epoch": 1984, "lr": 2.7291952368401474e-05}
{"train_loss": 0.06698352098464966, "global_step": 176634, "epoch": 1984, "lr": 2.729143587119406e-05}
{"train_loss": 0.09057645499706268, "global_step": 176635, "epoch": 1984, "lr": 2.7290919377039564e-05}
{"train_loss": 0.06145187467336655, "global_step": 176636, "epoch": 1984, "lr": 2.7290402885938005e-05}
{"train_loss": 0.05651433393359184, "global_step": 176637, "epoch": 1984, "lr": 2.7289886397889498e-05}
{"train_loss": 0.050389476120471954, "global_step": 176638, "epoch": 1984, "lr": 2.7289369912894057e-05}
{"train_loss": 0.0569785051047802, "global_step": 176639, "epoch": 1984, "lr": 2.728885343095181e-05}
{"train_loss": 0.030404651537537575, "global_step": 176640, "epoch": 1984, "lr": 2.7288336952062786e-05}
{"train_loss": 0.0629621297121048, "global_step": 176641, "epoch": 1984, "lr": 2.7287820476227084e-05}
{"train_loss": 0.08089099824428558, "global_step": 176642, "epoch": 1984, "lr": 2.7287304003444737e-05}
{"train_loss": 0.11665424704551697, "global_step": 176643, "epoch": 1984, "lr": 2.7286787533715852e-05}
{"train_loss": 0.0989532545208931, "global_step": 176644, "epoch": 1984, "lr": 2.728627106704047e-05}
{"train_loss": 0.07115929573774338, "global_step": 176645, "epoch": 1984, "lr": 2.7285754603418685e-05}
{"train_loss": 0.1545257270336151, "global_step": 176646, "epoch": 1984, "lr": 2.7285238142850533e-05}
{"train_loss": 0.1065325140953064, "global_step": 176647, "epoch": 1984, "lr": 2.7284721685336123e-05}
{"train_loss": 0.04445596784353256, "global_step": 176648, "epoch": 1984, "lr": 2.7284205230875492e-05}
{"train_loss": 0.04657639563083649, "global_step": 176649, "epoch": 1984, "lr": 2.7283688779468742e-05}
{"train_loss": 0.07387042045593262, "global_step": 176650, "epoch": 1984, "lr": 2.7283172331115896e-05}
{"train_loss": 0.07564292848110199, "global_step": 176651, "epoch": 1984, "lr": 2.7282655885817054e-05}
{"train_loss": 0.08220959454774857, "global_step": 176652, "epoch": 1984, "lr": 2.7282139443572307e-05}
{"train_loss": 0.05384325608611107, "global_step": 176653, "epoch": 1984, "lr": 2.7281623004381674e-05}
{"train_loss": 0.08368615806102753, "global_step": 176654, "epoch": 1984, "lr": 2.7281106568245275e-05}
{"train_loss": 0.05553498491644859, "global_step": 176655, "epoch": 1984, "lr": 2.728059013516313e-05}
{"train_loss": 0.056688372045755386, "global_step": 176656, "epoch": 1984, "lr": 2.728007370513535e-05}
{"train_loss": 0.11753170937299728, "global_step": 176657, "epoch": 1984, "lr": 2.7279557278161972e-05}
{"train_loss": 0.06983153522014618, "global_step": 176658, "epoch": 1984, "lr": 2.72790408542431e-05}
{"train_loss": 0.09688825905323029, "global_step": 176659, "epoch": 1984, "lr": 2.727852443337876e-05}
{"train_loss": 0.047956373542547226, "global_step": 176660, "epoch": 1984, "lr": 2.727800801556907e-05}
{"train_loss": 0.03619850054383278, "global_step": 176661, "epoch": 1984, "lr": 2.7277491600814076e-05}
{"train_loss": 0.0706024095416069, "global_step": 176662, "epoch": 1984, "lr": 2.727697518911382e-05}
{"train_loss": 0.10209006816148758, "global_step": 176663, "epoch": 1984, "lr": 2.7276458780468423e-05}
{"train_loss": 0.0806407481641247, "global_step": 176664, "epoch": 1984, "lr": 2.7275942374877906e-05, "val_loss": 7.407698154449463}
{"train_loss": 0.0647362470626831, "global_step": 176665, "epoch": 1985, "lr": 2.7275425972342383e-05}
{"train_loss": 0.0911065861582756, "global_step": 176666, "epoch": 1985, "lr": 2.7274909572861908e-05}
{"train_loss": 0.08269202709197998, "global_step": 176667, "epoch": 1985, "lr": 2.7274393176436518e-05}
{"train_loss": 0.07118730247020721, "global_step": 176668, "epoch": 1985, "lr": 2.7273876783066328e-05}
{"train_loss": 0.07213912904262543, "global_step": 176669, "epoch": 1985, "lr": 2.727336039275137e-05}
{"train_loss": 0.0554189532995224, "global_step": 176670, "epoch": 1985, "lr": 2.7272844005491726e-05}
{"train_loss": 0.07149431854486465, "global_step": 176671, "epoch": 1985, "lr": 2.7272327621287503e-05}
{"train_loss": 0.05686016380786896, "global_step": 176672, "epoch": 1985, "lr": 2.727181124013871e-05}
{"train_loss": 0.06889469176530838, "global_step": 176673, "epoch": 1985, "lr": 2.7271294862045472e-05}
{"train_loss": 0.08990176022052765, "global_step": 176674, "epoch": 1985, "lr": 2.7270778487007803e-05}
{"train_loss": 0.09850427508354187, "global_step": 176675, "epoch": 1985, "lr": 2.7270262115025826e-05}
{"train_loss": 0.1073029637336731, "global_step": 176676, "epoch": 1985, "lr": 2.7269745746099563e-05}
{"train_loss": 0.10974318534135818, "global_step": 176677, "epoch": 1985, "lr": 2.726922938022912e-05}
{"train_loss": 0.07789948582649231, "global_step": 176678, "epoch": 1985, "lr": 2.7268713017414544e-05}
{"train_loss": 0.08011070638895035, "global_step": 176679, "epoch": 1985, "lr": 2.7268196657655926e-05}
{"train_loss": 0.09043613076210022, "global_step": 176680, "epoch": 1985, "lr": 2.72676803009533e-05}
{"train_loss": 0.12719978392124176, "global_step": 176681, "epoch": 1985, "lr": 2.7267163947306785e-05}
{"train_loss": 0.04980352520942688, "global_step": 176682, "epoch": 1985, "lr": 2.7266647596716394e-05}
{"train_loss": 0.09590362012386322, "global_step": 176683, "epoch": 1985, "lr": 2.7266131249182247e-05}
{"train_loss": 0.07251303642988205, "global_step": 176684, "epoch": 1985, "lr": 2.726561490470437e-05}
{"train_loss": 0.037921056151390076, "global_step": 176685, "epoch": 1985, "lr": 2.726509856328289e-05}
{"train_loss": 0.06222222372889519, "global_step": 176686, "epoch": 1985, "lr": 2.7264582224917802e-05}
{"train_loss": 0.0924869105219841, "global_step": 176687, "epoch": 1985, "lr": 2.7264065889609238e-05}
{"train_loss": 0.0741085484623909, "global_step": 176688, "epoch": 1985, "lr": 2.7263549557357225e-05}
{"train_loss": 0.054894961416721344, "global_step": 176689, "epoch": 1985, "lr": 2.726303322816185e-05}
{"train_loss": 0.06755086034536362, "global_step": 176690, "epoch": 1985, "lr": 2.7262516902023206e-05}
{"train_loss": 0.09054529666900635, "global_step": 176691, "epoch": 1985, "lr": 2.7262000578941315e-05}
{"train_loss": 0.1273259073495865, "global_step": 176692, "epoch": 1985, "lr": 2.726148425891629e-05}
{"train_loss": 0.024464048445224762, "global_step": 176693, "epoch": 1985, "lr": 2.7260967941948168e-05}
{"train_loss": 0.07014883309602737, "global_step": 176694, "epoch": 1985, "lr": 2.7260451628037048e-05}
{"train_loss": 0.06513097137212753, "global_step": 176695, "epoch": 1985, "lr": 2.725993531718296e-05}
{"train_loss": 0.08437172323465347, "global_step": 176696, "epoch": 1985, "lr": 2.7259419009386023e-05}
{"train_loss": 0.05417965352535248, "global_step": 176697, "epoch": 1985, "lr": 2.7258902704646254e-05}
{"train_loss": 0.1003328189253807, "global_step": 176698, "epoch": 1985, "lr": 2.7258386402963776e-05}
{"train_loss": 0.13407938182353973, "global_step": 176699, "epoch": 1985, "lr": 2.7257870104338623e-05}
{"train_loss": 0.10428524762392044, "global_step": 176700, "epoch": 1985, "lr": 2.725735380877086e-05}
{"train_loss": 0.10860396921634674, "global_step": 176701, "epoch": 1985, "lr": 2.7256837516260582e-05}
{"train_loss": 0.13405194878578186, "global_step": 176702, "epoch": 1985, "lr": 2.7256321226807823e-05}
{"train_loss": 0.13994669914245605, "global_step": 176703, "epoch": 1985, "lr": 2.72558049404127e-05}
{"train_loss": 0.060735560953617096, "global_step": 176704, "epoch": 1985, "lr": 2.7255288657075228e-05}
{"train_loss": 0.11675664037466049, "global_step": 176705, "epoch": 1985, "lr": 2.7254772376795534e-05}
{"train_loss": 0.08674875646829605, "global_step": 176706, "epoch": 1985, "lr": 2.725425609957364e-05}
{"train_loss": 0.08507311344146729, "global_step": 176707, "epoch": 1985, "lr": 2.7253739825409653e-05}
{"train_loss": 0.11472246795892715, "global_step": 176708, "epoch": 1985, "lr": 2.7253223554303597e-05}
{"train_loss": 0.06941495835781097, "global_step": 176709, "epoch": 1985, "lr": 2.725270728625559e-05}
{"train_loss": 0.07405547052621841, "global_step": 176710, "epoch": 1985, "lr": 2.7252191021265662e-05}
{"train_loss": 0.10023852437734604, "global_step": 176711, "epoch": 1985, "lr": 2.725167475933392e-05}
{"train_loss": 0.07312391698360443, "global_step": 176712, "epoch": 1985, "lr": 2.7251158500460394e-05}
{"train_loss": 0.07220776379108429, "global_step": 176713, "epoch": 1985, "lr": 2.7250642244645187e-05}
{"train_loss": 0.05938911810517311, "global_step": 176714, "epoch": 1985, "lr": 2.7250125991888338e-05}
{"train_loss": 0.0970768854022026, "global_step": 176715, "epoch": 1985, "lr": 2.7249609742189952e-05}
{"train_loss": 0.06574179232120514, "global_step": 176716, "epoch": 1985, "lr": 2.724909349555006e-05}
{"train_loss": 0.09353427588939667, "global_step": 176717, "epoch": 1985, "lr": 2.7248577251968767e-05}
{"train_loss": 0.05574360862374306, "global_step": 176718, "epoch": 1985, "lr": 2.7248061011446103e-05}
{"train_loss": 0.12178478389978409, "global_step": 176719, "epoch": 1985, "lr": 2.724754477398218e-05}
{"train_loss": 0.12786328792572021, "global_step": 176720, "epoch": 1985, "lr": 2.7247028539577035e-05}
{"train_loss": 0.10421504080295563, "global_step": 176721, "epoch": 1985, "lr": 2.7246512308230767e-05}
{"train_loss": 0.12164568901062012, "global_step": 176722, "epoch": 1985, "lr": 2.7245996079943402e-05}
{"train_loss": 0.047737959772348404, "global_step": 176723, "epoch": 1985, "lr": 2.7245479854715063e-05}
{"train_loss": 0.08401267230510712, "global_step": 176724, "epoch": 1985, "lr": 2.7244963632545773e-05}
{"train_loss": 0.10931230336427689, "global_step": 176725, "epoch": 1985, "lr": 2.7244447413435614e-05}
{"train_loss": 0.11869150400161743, "global_step": 176726, "epoch": 1985, "lr": 2.7243931197384687e-05}
{"train_loss": 0.08364704996347427, "global_step": 176727, "epoch": 1985, "lr": 2.7243414984393013e-05}
{"train_loss": 0.10826858878135681, "global_step": 176728, "epoch": 1985, "lr": 2.72428987744607e-05}
{"train_loss": 0.08412738144397736, "global_step": 176729, "epoch": 1985, "lr": 2.724238256758779e-05}
{"train_loss": 0.12252029776573181, "global_step": 176730, "epoch": 1985, "lr": 2.724186636377438e-05}
{"train_loss": 0.12508238852024078, "global_step": 176731, "epoch": 1985, "lr": 2.7241350163020503e-05}
{"train_loss": 0.10094136744737625, "global_step": 176732, "epoch": 1985, "lr": 2.7240833965326274e-05}
{"train_loss": 0.07069339603185654, "global_step": 176733, "epoch": 1985, "lr": 2.7240317770691708e-05}
{"train_loss": 0.09141863137483597, "global_step": 176734, "epoch": 1985, "lr": 2.723980157911693e-05}
{"train_loss": 0.12389398366212845, "global_step": 176735, "epoch": 1985, "lr": 2.7239285390601967e-05}
{"train_loss": 0.07609868049621582, "global_step": 176736, "epoch": 1985, "lr": 2.723876920514692e-05}
{"train_loss": 0.07172242552042007, "global_step": 176737, "epoch": 1985, "lr": 2.7238253022751825e-05}
{"train_loss": 0.11329527199268341, "global_step": 176738, "epoch": 1985, "lr": 2.7237736843416784e-05}
{"train_loss": 0.07498934864997864, "global_step": 176739, "epoch": 1985, "lr": 2.723722066714186e-05}
{"train_loss": 0.08305208384990692, "global_step": 176740, "epoch": 1985, "lr": 2.723670449392709e-05}
{"train_loss": 0.12407862395048141, "global_step": 176741, "epoch": 1985, "lr": 2.7236188323772587e-05}
{"train_loss": 0.09611358493566513, "global_step": 176742, "epoch": 1985, "lr": 2.7235672156678382e-05}
{"train_loss": 0.07804425805807114, "global_step": 176743, "epoch": 1985, "lr": 2.7235155992644585e-05}
{"train_loss": 0.08760984987020493, "global_step": 176744, "epoch": 1985, "lr": 2.723463983167122e-05}
{"train_loss": 0.08979801833629608, "global_step": 176745, "epoch": 1985, "lr": 2.7234123673758406e-05}
{"train_loss": 0.07493177801370621, "global_step": 176746, "epoch": 1985, "lr": 2.723360751890616e-05}
{"train_loss": 0.08731641620397568, "global_step": 176747, "epoch": 1985, "lr": 2.7233091367114603e-05}
{"train_loss": 0.07917047291994095, "global_step": 176748, "epoch": 1985, "lr": 2.7232575218383755e-05}
{"train_loss": 0.07457756996154785, "global_step": 176749, "epoch": 1985, "lr": 2.7232059072713734e-05}
{"train_loss": 0.08371945470571518, "global_step": 176750, "epoch": 1985, "lr": 2.723154293010456e-05}
{"train_loss": 0.07303638756275177, "global_step": 176751, "epoch": 1985, "lr": 2.7231026790556353e-05}
{"train_loss": 0.08574514836072922, "global_step": 176752, "epoch": 1985, "lr": 2.7230510654069135e-05}
{"train_loss": 0.08774740667490477, "global_step": 176753, "epoch": 1985, "lr": 2.722999452064302e-05, "val_loss": 7.5494608879089355, "train_action_mse_error": 10.184557914733887}
{"train_loss": 0.08788279443979263, "global_step": 176754, "epoch": 1986, "lr": 2.722947839027803e-05}
{"train_loss": 0.043690212070941925, "global_step": 176755, "epoch": 1986, "lr": 2.722896226297428e-05}
{"train_loss": 0.05794427916407585, "global_step": 176756, "epoch": 1986, "lr": 2.72284461387318e-05}
{"train_loss": 0.08026348799467087, "global_step": 176757, "epoch": 1986, "lr": 2.7227930017550695e-05}
{"train_loss": 0.08953773975372314, "global_step": 176758, "epoch": 1986, "lr": 2.7227413899430997e-05}
{"train_loss": 0.08259105682373047, "global_step": 176759, "epoch": 1986, "lr": 2.7226897784372825e-05}
{"train_loss": 0.15308420360088348, "global_step": 176760, "epoch": 1986, "lr": 2.722638167237619e-05}
{"train_loss": 0.08904960006475449, "global_step": 176761, "epoch": 1986, "lr": 2.7225865563441194e-05}
{"train_loss": 0.07201585173606873, "global_step": 176762, "epoch": 1986, "lr": 2.7225349457567915e-05}
{"train_loss": 0.07562168687582016, "global_step": 176763, "epoch": 1986, "lr": 2.7224833354756403e-05}
{"train_loss": 0.08196023851633072, "global_step": 176764, "epoch": 1986, "lr": 2.722431725500675e-05}
{"train_loss": 0.06126290187239647, "global_step": 176765, "epoch": 1986, "lr": 2.7223801158318984e-05}
{"train_loss": 0.09999633580446243, "global_step": 176766, "epoch": 1986, "lr": 2.722328506469323e-05}
{"train_loss": 0.09037073701620102, "global_step": 176767, "epoch": 1986, "lr": 2.7222768974129498e-05}
{"train_loss": 0.07541453838348389, "global_step": 176768, "epoch": 1986, "lr": 2.722225288662791e-05}
{"train_loss": 0.053442612290382385, "global_step": 176769, "epoch": 1986, "lr": 2.7221736802188498e-05}
{"train_loss": 0.05792117118835449, "global_step": 176770, "epoch": 1986, "lr": 2.7221220720811358e-05}
{"train_loss": 0.09312403947114944, "global_step": 176771, "epoch": 1986, "lr": 2.722070464249653e-05}
{"train_loss": 0.08446317911148071, "global_step": 176772, "epoch": 1986, "lr": 2.722018856724412e-05}
{"train_loss": 0.0774037167429924, "global_step": 176773, "epoch": 1986, "lr": 2.721967249505416e-05}
{"train_loss": 0.10512237250804901, "global_step": 176774, "epoch": 1986, "lr": 2.721915642592676e-05}
{"train_loss": 0.07622860372066498, "global_step": 176775, "epoch": 1986, "lr": 2.7218640359861948e-05}
{"train_loss": 0.07351116836071014, "global_step": 176776, "epoch": 1986, "lr": 2.7218124296859825e-05}
{"train_loss": 0.06639695167541504, "global_step": 176777, "epoch": 1986, "lr": 2.721760823692045e-05}
{"train_loss": 0.07325447350740433, "global_step": 176778, "epoch": 1986, "lr": 2.721709218004387e-05}
{"train_loss": 0.140068918466568, "global_step": 176779, "epoch": 1986, "lr": 2.7216576126230194e-05}
{"train_loss": 0.13279379904270172, "global_step": 176780, "epoch": 1986, "lr": 2.7216060075479454e-05}
{"train_loss": 0.1339733600616455, "global_step": 176781, "epoch": 1986, "lr": 2.721554402779175e-05}
{"train_loss": 0.09774678200483322, "global_step": 176782, "epoch": 1986, "lr": 2.721502798316713e-05}
{"train_loss": 0.20391613245010376, "global_step": 176783, "epoch": 1986, "lr": 2.7214511941605682e-05}
{"train_loss": 0.10326690971851349, "global_step": 176784, "epoch": 1986, "lr": 2.7213995903107447e-05}
{"train_loss": 0.12575072050094604, "global_step": 176785, "epoch": 1986, "lr": 2.721347986767253e-05}
{"train_loss": 0.03269733116030693, "global_step": 176786, "epoch": 1986, "lr": 2.7212963835300976e-05}
{"train_loss": 0.1013999953866005, "global_step": 176787, "epoch": 1986, "lr": 2.721244780599287e-05}
{"train_loss": 0.06279077380895615, "global_step": 176788, "epoch": 1986, "lr": 2.7211931779748256e-05}
{"train_loss": 0.03299696743488312, "global_step": 176789, "epoch": 1986, "lr": 2.721141575656724e-05}
{"train_loss": 0.13649654388427734, "global_step": 176790, "epoch": 1986, "lr": 2.7210899736449846e-05}
{"train_loss": 0.05086139217019081, "global_step": 176791, "epoch": 1986, "lr": 2.721038371939619e-05}
{"train_loss": 0.0717884823679924, "global_step": 176792, "epoch": 1986, "lr": 2.7209867705406306e-05}
{"train_loss": 0.07573992013931274, "global_step": 176793, "epoch": 1986, "lr": 2.7209351694480294e-05}
{"train_loss": 0.11758729815483093, "global_step": 176794, "epoch": 1986, "lr": 2.720883568661819e-05}
{"train_loss": 0.057962387800216675, "global_step": 176795, "epoch": 1986, "lr": 2.7208319681820093e-05}
{"train_loss": 0.0740407183766365, "global_step": 176796, "epoch": 1986, "lr": 2.7207803680086045e-05}
{"train_loss": 0.07616288959980011, "global_step": 176797, "epoch": 1986, "lr": 2.7207287681416133e-05}
{"train_loss": 0.15765343606472015, "global_step": 176798, "epoch": 1986, "lr": 2.720677168581044e-05}
{"train_loss": 0.0997423604130745, "global_step": 176799, "epoch": 1986, "lr": 2.7206255693269e-05}
{"train_loss": 0.058196112513542175, "global_step": 176800, "epoch": 1986, "lr": 2.720573970379192e-05}
{"train_loss": 0.047664422541856766, "global_step": 176801, "epoch": 1986, "lr": 2.7205223717379225e-05}
{"train_loss": 0.08926483243703842, "global_step": 176802, "epoch": 1986, "lr": 2.7204707734031037e-05}
{"train_loss": 0.10752683132886887, "global_step": 176803, "epoch": 1986, "lr": 2.7204191753747377e-05}
{"train_loss": 0.1108732521533966, "global_step": 176804, "epoch": 1986, "lr": 2.7203675776528354e-05}
{"train_loss": 0.05010020360350609, "global_step": 176805, "epoch": 1986, "lr": 2.7203159802374e-05}
{"train_loss": 0.07496752589941025, "global_step": 176806, "epoch": 1986, "lr": 2.7202643831284423e-05}
{"train_loss": 0.06762136518955231, "global_step": 176807, "epoch": 1986, "lr": 2.720212786325965e-05}
{"train_loss": 0.09087088704109192, "global_step": 176808, "epoch": 1986, "lr": 2.7201611898299793e-05}
{"train_loss": 0.06213236600160599, "global_step": 176809, "epoch": 1986, "lr": 2.720109593640489e-05}
{"train_loss": 0.10248284041881561, "global_step": 176810, "epoch": 1986, "lr": 2.7200579977575037e-05}
{"train_loss": 0.060382504016160965, "global_step": 176811, "epoch": 1986, "lr": 2.7200064021810267e-05}
{"train_loss": 0.06122394651174545, "global_step": 176812, "epoch": 1986, "lr": 2.7199548069110693e-05}
{"train_loss": 0.09031490236520767, "global_step": 176813, "epoch": 1986, "lr": 2.7199032119476338e-05}
{"train_loss": 0.05040004104375839, "global_step": 176814, "epoch": 1986, "lr": 2.7198516172907317e-05}
{"train_loss": 0.05808515101671219, "global_step": 176815, "epoch": 1986, "lr": 2.7198000229403654e-05}
{"train_loss": 0.07591372728347778, "global_step": 176816, "epoch": 1986, "lr": 2.7197484288965467e-05}
{"train_loss": 0.07771372050046921, "global_step": 176817, "epoch": 1986, "lr": 2.71969683515928e-05}
{"train_loss": 0.10708541423082352, "global_step": 176818, "epoch": 1986, "lr": 2.71964524172857e-05}
{"train_loss": 0.0932687371969223, "global_step": 176819, "epoch": 1986, "lr": 2.7195936486044283e-05}
{"train_loss": 0.11383218318223953, "global_step": 176820, "epoch": 1986, "lr": 2.7195420557868566e-05}
{"train_loss": 0.14717774093151093, "global_step": 176821, "epoch": 1986, "lr": 2.7194904632758677e-05}
{"train_loss": 0.04898630455136299, "global_step": 176822, "epoch": 1986, "lr": 2.7194388710714625e-05}
{"train_loss": 0.08199199289083481, "global_step": 176823, "epoch": 1986, "lr": 2.719387279173654e-05}
{"train_loss": 0.10656040161848068, "global_step": 176824, "epoch": 1986, "lr": 2.7193356875824433e-05}
{"train_loss": 0.04077340289950371, "global_step": 176825, "epoch": 1986, "lr": 2.7192840962978426e-05}
{"train_loss": 0.08541649580001831, "global_step": 176826, "epoch": 1986, "lr": 2.7192325053198543e-05}
{"train_loss": 0.049237195402383804, "global_step": 176827, "epoch": 1986, "lr": 2.7191809146484893e-05}
{"train_loss": 0.06113984435796738, "global_step": 176828, "epoch": 1986, "lr": 2.71912932428375e-05}
{"train_loss": 0.0861971378326416, "global_step": 176829, "epoch": 1986, "lr": 2.719077734225649e-05}
{"train_loss": 0.07061433047056198, "global_step": 176830, "epoch": 1986, "lr": 2.7190261444741872e-05}
{"train_loss": 0.06780008971691132, "global_step": 176831, "epoch": 1986, "lr": 2.7189745550293777e-05}
{"train_loss": 0.03230994939804077, "global_step": 176832, "epoch": 1986, "lr": 2.718922965891221e-05}
{"train_loss": 0.07629969716072083, "global_step": 176833, "epoch": 1986, "lr": 2.7188713770597285e-05}
{"train_loss": 0.07473099231719971, "global_step": 176834, "epoch": 1986, "lr": 2.718819788534907e-05}
{"train_loss": 0.05681225657463074, "global_step": 176835, "epoch": 1986, "lr": 2.718768200316761e-05}
{"train_loss": 0.04696134105324745, "global_step": 176836, "epoch": 1986, "lr": 2.7187166124053003e-05}
{"train_loss": 0.08345846831798553, "global_step": 176837, "epoch": 1986, "lr": 2.7186650248005286e-05}
{"train_loss": 0.10716512799263, "global_step": 176838, "epoch": 1986, "lr": 2.7186134375024563e-05}
{"train_loss": 0.08231262117624283, "global_step": 176839, "epoch": 1986, "lr": 2.7185618505110878e-05}
{"train_loss": 0.09030716121196747, "global_step": 176840, "epoch": 1986, "lr": 2.7185102638264314e-05}
{"train_loss": 0.09167221933603287, "global_step": 176841, "epoch": 1986, "lr": 2.718458677448492e-05}
{"train_loss": 0.08287887802619613, "global_step": 176842, "epoch": 1986, "lr": 2.71840709137728e-05, "val_loss": 7.29660701751709}
{"train_loss": 0.13430246710777283, "global_step": 176843, "epoch": 1987, "lr": 2.7183555056127986e-05}
{"train_loss": 0.04205704107880592, "global_step": 176844, "epoch": 1987, "lr": 2.7183039201550586e-05}
{"train_loss": 0.09082905948162079, "global_step": 176845, "epoch": 1987, "lr": 2.7182523350040623e-05}
{"train_loss": 0.0611538290977478, "global_step": 176846, "epoch": 1987, "lr": 2.7182007501598218e-05}
{"train_loss": 0.08462647348642349, "global_step": 176847, "epoch": 1987, "lr": 2.718149165622339e-05}
{"train_loss": 0.09201698005199432, "global_step": 176848, "epoch": 1987, "lr": 2.7180975813916247e-05}
{"train_loss": 0.0627206489443779, "global_step": 176849, "epoch": 1987, "lr": 2.718045997467683e-05}
{"train_loss": 0.053684160113334656, "global_step": 176850, "epoch": 1987, "lr": 2.7179944138505243e-05}
{"train_loss": 0.06160103157162666, "global_step": 176851, "epoch": 1987, "lr": 2.717942830540151e-05}
{"train_loss": 0.06106654182076454, "global_step": 176852, "epoch": 1987, "lr": 2.7178912475365747e-05}
{"train_loss": 0.11272550374269485, "global_step": 176853, "epoch": 1987, "lr": 2.7178396648397985e-05}
{"train_loss": 0.060006607323884964, "global_step": 176854, "epoch": 1987, "lr": 2.7177880824498325e-05}
{"train_loss": 0.059433214366436005, "global_step": 176855, "epoch": 1987, "lr": 2.7177365003666822e-05}
{"train_loss": 0.12251223623752594, "global_step": 176856, "epoch": 1987, "lr": 2.717684918590352e-05}
{"train_loss": 0.07936543971300125, "global_step": 176857, "epoch": 1987, "lr": 2.717633337120854e-05}
{"train_loss": 0.0948297306895256, "global_step": 176858, "epoch": 1987, "lr": 2.7175817559581896e-05}
{"train_loss": 0.06810680031776428, "global_step": 176859, "epoch": 1987, "lr": 2.7175301751023707e-05}
{"train_loss": 0.07335484772920609, "global_step": 176860, "epoch": 1987, "lr": 2.717478594553401e-05}
{"train_loss": 0.055825937539339066, "global_step": 176861, "epoch": 1987, "lr": 2.717427014311289e-05}
{"train_loss": 0.10644083470106125, "global_step": 176862, "epoch": 1987, "lr": 2.7173754343760405e-05}
{"train_loss": 0.07245554029941559, "global_step": 176863, "epoch": 1987, "lr": 2.7173238547476642e-05}
{"train_loss": 0.0889960303902626, "global_step": 176864, "epoch": 1987, "lr": 2.717272275426164e-05}
{"train_loss": 0.07276695966720581, "global_step": 176865, "epoch": 1987, "lr": 2.717220696411551e-05}
{"train_loss": 0.07242894172668457, "global_step": 176866, "epoch": 1987, "lr": 2.7171691177038273e-05}
{"train_loss": 0.07076049596071243, "global_step": 176867, "epoch": 1987, "lr": 2.7171175393030047e-05}
{"train_loss": 0.060798488557338715, "global_step": 176868, "epoch": 1987, "lr": 2.7170659612090865e-05}
{"train_loss": 0.07914458215236664, "global_step": 176869, "epoch": 1987, "lr": 2.7170143834220795e-05}
{"train_loss": 0.07309161871671677, "global_step": 176870, "epoch": 1987, "lr": 2.7169628059419956e-05}
{"train_loss": 0.09717202186584473, "global_step": 176871, "epoch": 1987, "lr": 2.7169112287688354e-05}
{"train_loss": 0.09087201952934265, "global_step": 176872, "epoch": 1987, "lr": 2.7168596519026108e-05}
{"train_loss": 0.07963099330663681, "global_step": 176873, "epoch": 1987, "lr": 2.7168080753433245e-05}
{"train_loss": 0.10710357874631882, "global_step": 176874, "epoch": 1987, "lr": 2.7167564990909878e-05}
{"train_loss": 0.06555421650409698, "global_step": 176875, "epoch": 1987, "lr": 2.7167049231456032e-05}
{"train_loss": 0.021189244464039803, "global_step": 176876, "epoch": 1987, "lr": 2.7166533475071813e-05}
{"train_loss": 0.07765272259712219, "global_step": 176877, "epoch": 1987, "lr": 2.7166017721757266e-05}
{"train_loss": 0.07154886424541473, "global_step": 176878, "epoch": 1987, "lr": 2.716550197151248e-05}
{"train_loss": 0.056663479655981064, "global_step": 176879, "epoch": 1987, "lr": 2.7164986224337496e-05}
{"train_loss": 0.024301497265696526, "global_step": 176880, "epoch": 1987, "lr": 2.716447048023243e-05}
{"train_loss": 0.036202047020196915, "global_step": 176881, "epoch": 1987, "lr": 2.7163954739197288e-05}
{"train_loss": 0.12713612616062164, "global_step": 176882, "epoch": 1987, "lr": 2.7163439001232204e-05}
{"train_loss": 0.09260985255241394, "global_step": 176883, "epoch": 1987, "lr": 2.716292326633719e-05}
{"train_loss": 0.08255033195018768, "global_step": 176884, "epoch": 1987, "lr": 2.7162407534512367e-05}
{"train_loss": 0.10219355672597885, "global_step": 176885, "epoch": 1987, "lr": 2.7161891805757767e-05}
{"train_loss": 0.03316575661301613, "global_step": 176886, "epoch": 1987, "lr": 2.7161376080073485e-05}
{"train_loss": 0.13945139944553375, "global_step": 176887, "epoch": 1987, "lr": 2.7160860357459562e-05}
{"train_loss": 0.09807136654853821, "global_step": 176888, "epoch": 1987, "lr": 2.716034463791609e-05}
{"train_loss": 0.08052228391170502, "global_step": 176889, "epoch": 1987, "lr": 2.7159828921443125e-05}
{"train_loss": 0.06590359658002853, "global_step": 176890, "epoch": 1987, "lr": 2.715931320804076e-05}
{"train_loss": 0.06306831538677216, "global_step": 176891, "epoch": 1987, "lr": 2.715879749770902e-05}
{"train_loss": 0.08700650185346603, "global_step": 176892, "epoch": 1987, "lr": 2.7158281790448024e-05}
{"train_loss": 0.11371215432882309, "global_step": 176893, "epoch": 1987, "lr": 2.715776608625783e-05}
{"train_loss": 0.05078665912151337, "global_step": 176894, "epoch": 1987, "lr": 2.7157250385138466e-05}
{"train_loss": 0.08360258489847183, "global_step": 176895, "epoch": 1987, "lr": 2.7156734687090047e-05}
{"train_loss": 0.039601460099220276, "global_step": 176896, "epoch": 1987, "lr": 2.7156218992112613e-05}
{"train_loss": 0.103089340031147, "global_step": 176897, "epoch": 1987, "lr": 2.715570330020627e-05}
{"train_loss": 0.09034083783626556, "global_step": 176898, "epoch": 1987, "lr": 2.7155187611371036e-05}
{"train_loss": 0.0706411749124527, "global_step": 176899, "epoch": 1987, "lr": 2.7154671925607034e-05}
{"train_loss": 0.1131313294172287, "global_step": 176900, "epoch": 1987, "lr": 2.7154156242914285e-05}
{"train_loss": 0.06335817277431488, "global_step": 176901, "epoch": 1987, "lr": 2.7153640563292905e-05}
{"train_loss": 0.13996246457099915, "global_step": 176902, "epoch": 1987, "lr": 2.7153124886742914e-05}
{"train_loss": 0.056424692273139954, "global_step": 176903, "epoch": 1987, "lr": 2.715260921326443e-05}
{"train_loss": 0.06618008017539978, "global_step": 176904, "epoch": 1987, "lr": 2.715209354285748e-05}
{"train_loss": 0.08798697590827942, "global_step": 176905, "epoch": 1987, "lr": 2.7151577875522148e-05}
{"train_loss": 0.08970058709383011, "global_step": 176906, "epoch": 1987, "lr": 2.715106221125854e-05}
{"train_loss": 0.06769349426031113, "global_step": 176907, "epoch": 1987, "lr": 2.7150546550066658e-05}
{"train_loss": 0.11897898465394974, "global_step": 176908, "epoch": 1987, "lr": 2.7150030891946637e-05}
{"train_loss": 0.10761010646820068, "global_step": 176909, "epoch": 1987, "lr": 2.7149515236898492e-05}
{"train_loss": 0.11592496186494827, "global_step": 176910, "epoch": 1987, "lr": 2.7148999584922336e-05}
{"train_loss": 0.1152106374502182, "global_step": 176911, "epoch": 1987, "lr": 2.71484839360182e-05}
{"train_loss": 0.06765225529670715, "global_step": 176912, "epoch": 1987, "lr": 2.7147968290186187e-05}
{"train_loss": 0.09778477996587753, "global_step": 176913, "epoch": 1987, "lr": 2.7147452647426343e-05}
{"train_loss": 0.07771007716655731, "global_step": 176914, "epoch": 1987, "lr": 2.7146937007738748e-05}
{"train_loss": 0.09707619249820709, "global_step": 176915, "epoch": 1987, "lr": 2.7146421371123466e-05}
{"train_loss": 0.049728766083717346, "global_step": 176916, "epoch": 1987, "lr": 2.714590573758058e-05}
{"train_loss": 0.10873498022556305, "global_step": 176917, "epoch": 1987, "lr": 2.7145390107110124e-05}
{"train_loss": 0.06846869736909866, "global_step": 176918, "epoch": 1987, "lr": 2.7144874479712224e-05}
{"train_loss": 0.07757965475320816, "global_step": 176919, "epoch": 1987, "lr": 2.7144358855386887e-05}
{"train_loss": 0.08066879957914352, "global_step": 176920, "epoch": 1987, "lr": 2.7143843234134235e-05}
{"train_loss": 0.09038647264242172, "global_step": 176921, "epoch": 1987, "lr": 2.7143327615954294e-05}
{"train_loss": 0.08598388731479645, "global_step": 176922, "epoch": 1987, "lr": 2.714281200084718e-05}
{"train_loss": 0.11254990845918655, "global_step": 176923, "epoch": 1987, "lr": 2.714229638881291e-05}
{"train_loss": 0.052810586988925934, "global_step": 176924, "epoch": 1987, "lr": 2.7141780779851576e-05}
{"train_loss": 0.09649308025836945, "global_step": 176925, "epoch": 1987, "lr": 2.7141265173963283e-05}
{"train_loss": 0.08865755796432495, "global_step": 176926, "epoch": 1987, "lr": 2.7140749571148062e-05}
{"train_loss": 0.1072680801153183, "global_step": 176927, "epoch": 1987, "lr": 2.714023397140597e-05}
{"train_loss": 0.05655607208609581, "global_step": 176928, "epoch": 1987, "lr": 2.713971837473711e-05}
{"train_loss": 0.0624673068523407, "global_step": 176929, "epoch": 1987, "lr": 2.713920278114152e-05}
{"train_loss": 0.1299039125442505, "global_step": 176930, "epoch": 1987, "lr": 2.7138687190619306e-05}
{"train_loss": 0.08157738327477755, "global_step": 176931, "epoch": 1987, "lr": 2.7138171603170493e-05, "val_loss": 7.519832611083984}
{"train_loss": 0.09419208765029907, "global_step": 176932, "epoch": 1988, "lr": 2.71376560187952e-05}
{"train_loss": 0.11590351164340973, "global_step": 176933, "epoch": 1988, "lr": 2.713714043749347e-05}
{"train_loss": 0.1069696843624115, "global_step": 176934, "epoch": 1988, "lr": 2.7136624859265346e-05}
{"train_loss": 0.14780205488204956, "global_step": 176935, "epoch": 1988, "lr": 2.713610928411095e-05}
{"train_loss": 0.1006191223859787, "global_step": 176936, "epoch": 1988, "lr": 2.7135593712030304e-05}
{"train_loss": 0.13951221108436584, "global_step": 176937, "epoch": 1988, "lr": 2.713507814302352e-05}
{"train_loss": 0.043386660516262054, "global_step": 176938, "epoch": 1988, "lr": 2.7134562577090627e-05}
{"train_loss": 0.10282809287309647, "global_step": 176939, "epoch": 1988, "lr": 2.713404701423173e-05}
{"train_loss": 0.10299792885780334, "global_step": 176940, "epoch": 1988, "lr": 2.713353145444686e-05}
{"train_loss": 0.13489338755607605, "global_step": 176941, "epoch": 1988, "lr": 2.7133015897736136e-05}
{"train_loss": 0.13322952389717102, "global_step": 176942, "epoch": 1988, "lr": 2.713250034409957e-05}
{"train_loss": 0.0655221939086914, "global_step": 176943, "epoch": 1988, "lr": 2.7131984793537267e-05}
{"train_loss": 0.10746369510889053, "global_step": 176944, "epoch": 1988, "lr": 2.7131469246049306e-05}
{"train_loss": 0.06702963262796402, "global_step": 176945, "epoch": 1988, "lr": 2.713095370163572e-05}
{"train_loss": 0.08350816369056702, "global_step": 176946, "epoch": 1988, "lr": 2.7130438160296622e-05}
{"train_loss": 0.11331694573163986, "global_step": 176947, "epoch": 1988, "lr": 2.7129922622032034e-05}
{"train_loss": 0.07478763908147812, "global_step": 176948, "epoch": 1988, "lr": 2.7129407086842074e-05}
{"train_loss": 0.044883668422698975, "global_step": 176949, "epoch": 1988, "lr": 2.7128891554726765e-05}
{"train_loss": 0.05947238579392433, "global_step": 176950, "epoch": 1988, "lr": 2.712837602568622e-05}
{"train_loss": 0.0871225893497467, "global_step": 176951, "epoch": 1988, "lr": 2.712786049972046e-05}
{"train_loss": 0.10693392902612686, "global_step": 176952, "epoch": 1988, "lr": 2.7127344976829604e-05}
{"train_loss": 0.10522132366895676, "global_step": 176953, "epoch": 1988, "lr": 2.7126829457013674e-05}
{"train_loss": 0.09492169320583344, "global_step": 176954, "epoch": 1988, "lr": 2.7126313940272786e-05}
{"train_loss": 0.06504618376493454, "global_step": 176955, "epoch": 1988, "lr": 2.7125798426606963e-05}
{"train_loss": 0.08456503599882126, "global_step": 176956, "epoch": 1988, "lr": 2.7125282916016327e-05}
{"train_loss": 0.11320221424102783, "global_step": 176957, "epoch": 1988, "lr": 2.7124767408500888e-05}
{"train_loss": 0.05573773384094238, "global_step": 176958, "epoch": 1988, "lr": 2.7124251904060772e-05}
{"train_loss": 0.12469760328531265, "global_step": 176959, "epoch": 1988, "lr": 2.7123736402695998e-05}
{"train_loss": 0.10105766355991364, "global_step": 176960, "epoch": 1988, "lr": 2.7123220904406656e-05}
{"train_loss": 0.051820848137140274, "global_step": 176961, "epoch": 1988, "lr": 2.7122705409192846e-05}
{"train_loss": 0.14090195298194885, "global_step": 176962, "epoch": 1988, "lr": 2.7122189917054585e-05}
{"train_loss": 0.056475136429071426, "global_step": 176963, "epoch": 1988, "lr": 2.7121674427991983e-05}
{"train_loss": 0.08634979277849197, "global_step": 176964, "epoch": 1988, "lr": 2.7121158942005075e-05}
{"train_loss": 0.10540126264095306, "global_step": 176965, "epoch": 1988, "lr": 2.7120643459093975e-05}
{"train_loss": 0.16421200335025787, "global_step": 176966, "epoch": 1988, "lr": 2.7120127979258712e-05}
{"train_loss": 0.1297505795955658, "global_step": 176967, "epoch": 1988, "lr": 2.7119612502499365e-05}
{"train_loss": 0.09936277568340302, "global_step": 176968, "epoch": 1988, "lr": 2.711909702881602e-05}
{"train_loss": 0.08083304017782211, "global_step": 176969, "epoch": 1988, "lr": 2.711858155820871e-05}
{"train_loss": 0.1099228709936142, "global_step": 176970, "epoch": 1988, "lr": 2.711806609067755e-05}
{"train_loss": 0.09576313197612762, "global_step": 176971, "epoch": 1988, "lr": 2.7117550626222587e-05}
{"train_loss": 0.0417962446808815, "global_step": 176972, "epoch": 1988, "lr": 2.7117035164843874e-05}
{"train_loss": 0.053125474601984024, "global_step": 176973, "epoch": 1988, "lr": 2.7116519706541517e-05}
{"train_loss": 0.0704847127199173, "global_step": 176974, "epoch": 1988, "lr": 2.711600425131554e-05}
{"train_loss": 0.07532820105552673, "global_step": 176975, "epoch": 1988, "lr": 2.7115488799166057e-05}
{"train_loss": 0.06263965368270874, "global_step": 176976, "epoch": 1988, "lr": 2.7114973350093103e-05}
{"train_loss": 0.05718913674354553, "global_step": 176977, "epoch": 1988, "lr": 2.711445790409678e-05}
{"train_loss": 0.05271196737885475, "global_step": 176978, "epoch": 1988, "lr": 2.7113942461177117e-05}
{"train_loss": 0.108172707259655, "global_step": 176979, "epoch": 1988, "lr": 2.7113427021334203e-05}
{"train_loss": 0.09743361175060272, "global_step": 176980, "epoch": 1988, "lr": 2.7112911584568136e-05}
{"train_loss": 0.08604913204908371, "global_step": 176981, "epoch": 1988, "lr": 2.7112396150878937e-05}
{"train_loss": 0.0536235012114048, "global_step": 176982, "epoch": 1988, "lr": 2.711188072026672e-05}
{"train_loss": 0.041563648730516434, "global_step": 176983, "epoch": 1988, "lr": 2.7111365292731507e-05}
{"train_loss": 0.10707932710647583, "global_step": 176984, "epoch": 1988, "lr": 2.7110849868273415e-05}
{"train_loss": 0.09706519544124603, "global_step": 176985, "epoch": 1988, "lr": 2.7110334446892472e-05}
{"train_loss": 0.11762217432260513, "global_step": 176986, "epoch": 1988, "lr": 2.710981902858878e-05}
{"train_loss": 0.05619824305176735, "global_step": 176987, "epoch": 1988, "lr": 2.7109303613362374e-05}
{"train_loss": 0.10152561217546463, "global_step": 176988, "epoch": 1988, "lr": 2.7108788201213365e-05}
{"train_loss": 0.027309155091643333, "global_step": 176989, "epoch": 1988, "lr": 2.7108272792141783e-05}
{"train_loss": 0.11623750627040863, "global_step": 176990, "epoch": 1988, "lr": 2.7107757386147732e-05}
{"train_loss": 0.11266491562128067, "global_step": 176991, "epoch": 1988, "lr": 2.7107241983231247e-05}
{"train_loss": 0.05663846433162689, "global_step": 176992, "epoch": 1988, "lr": 2.7106726583392428e-05}
{"train_loss": 0.06757775694131851, "global_step": 176993, "epoch": 1988, "lr": 2.7106211186631314e-05}
{"train_loss": 0.12106345593929291, "global_step": 176994, "epoch": 1988, "lr": 2.7105695792948017e-05}
{"train_loss": 0.12116134911775589, "global_step": 176995, "epoch": 1988, "lr": 2.7105180402342552e-05}
{"train_loss": 0.0598430335521698, "global_step": 176996, "epoch": 1988, "lr": 2.710466501481502e-05}
{"train_loss": 0.16018325090408325, "global_step": 176997, "epoch": 1988, "lr": 2.7104149630365505e-05}
{"train_loss": 0.0677240788936615, "global_step": 176998, "epoch": 1988, "lr": 2.7103634248994038e-05}
{"train_loss": 0.054525405168533325, "global_step": 176999, "epoch": 1988, "lr": 2.710311887070073e-05}
{"train_loss": 0.14577828347682953, "global_step": 177000, "epoch": 1988, "lr": 2.7102603495485602e-05}
{"train_loss": 0.08243542164564133, "global_step": 177001, "epoch": 1988, "lr": 2.710208812334878e-05}
{"train_loss": 0.05112787336111069, "global_step": 177002, "epoch": 1988, "lr": 2.7101572754290278e-05}
{"train_loss": 0.0918617695569992, "global_step": 177003, "epoch": 1988, "lr": 2.7101057388310213e-05}
{"train_loss": 0.0741487368941307, "global_step": 177004, "epoch": 1988, "lr": 2.7100542025408628e-05}
{"train_loss": 0.05849634110927582, "global_step": 177005, "epoch": 1988, "lr": 2.7100026665585577e-05}
{"train_loss": 0.12013127654790878, "global_step": 177006, "epoch": 1988, "lr": 2.709951130884117e-05}
{"train_loss": 0.09720897674560547, "global_step": 177007, "epoch": 1988, "lr": 2.709899595517543e-05}
{"train_loss": 0.04464936628937721, "global_step": 177008, "epoch": 1988, "lr": 2.709848060458848e-05}
{"train_loss": 0.051459405571222305, "global_step": 177009, "epoch": 1988, "lr": 2.709796525708035e-05}
{"train_loss": 0.08842654526233673, "global_step": 177010, "epoch": 1988, "lr": 2.709744991265111e-05}
{"train_loss": 0.10866980999708176, "global_step": 177011, "epoch": 1988, "lr": 2.7096934571300847e-05}
{"train_loss": 0.05102822929620743, "global_step": 177012, "epoch": 1988, "lr": 2.709641923302961e-05}
{"train_loss": 0.06198977679014206, "global_step": 177013, "epoch": 1988, "lr": 2.70959038978375e-05}
{"train_loss": 0.10484550893306732, "global_step": 177014, "epoch": 1988, "lr": 2.7095388565724546e-05}
{"train_loss": 0.059439852833747864, "global_step": 177015, "epoch": 1988, "lr": 2.709487323669083e-05}
{"train_loss": 0.08928045630455017, "global_step": 177016, "epoch": 1988, "lr": 2.7094357910736457e-05}
{"train_loss": 0.05268646404147148, "global_step": 177017, "epoch": 1988, "lr": 2.709384258786145e-05}
{"train_loss": 0.13566593825817108, "global_step": 177018, "epoch": 1988, "lr": 2.7093327268065915e-05}
{"train_loss": 0.09485897421836853, "global_step": 177019, "epoch": 1988, "lr": 2.7092811951349882e-05}
{"train_loss": 0.08804856569328334, "global_step": 177020, "epoch": 1988, "lr": 2.7092296637713454e-05, "val_loss": 7.36386251449585}
{"train_loss": 0.02838847227394581, "global_step": 177021, "epoch": 1989, "lr": 2.709178132715668e-05}
{"train_loss": 0.04652656614780426, "global_step": 177022, "epoch": 1989, "lr": 2.709126601967965e-05}
{"train_loss": 0.11742483824491501, "global_step": 177023, "epoch": 1989, "lr": 2.7090750715282398e-05}
{"train_loss": 0.09124596416950226, "global_step": 177024, "epoch": 1989, "lr": 2.709023541396504e-05}
{"train_loss": 0.11527380347251892, "global_step": 177025, "epoch": 1989, "lr": 2.70897201157276e-05}
{"train_loss": 0.1213548555970192, "global_step": 177026, "epoch": 1989, "lr": 2.708920482057019e-05}
{"train_loss": 0.10641226917505264, "global_step": 177027, "epoch": 1989, "lr": 2.7088689528492835e-05}
{"train_loss": 0.08799871057271957, "global_step": 177028, "epoch": 1989, "lr": 2.7088174239495643e-05}
{"train_loss": 0.0629129633307457, "global_step": 177029, "epoch": 1989, "lr": 2.7087658953578653e-05}
{"train_loss": 0.04988734424114227, "global_step": 177030, "epoch": 1989, "lr": 2.7087143670741967e-05}
{"train_loss": 0.062458403408527374, "global_step": 177031, "epoch": 1989, "lr": 2.7086628390985607e-05}
{"train_loss": 0.11249540746212006, "global_step": 177032, "epoch": 1989, "lr": 2.7086113114309675e-05}
{"train_loss": 0.07752515375614166, "global_step": 177033, "epoch": 1989, "lr": 2.708559784071426e-05}
{"train_loss": 0.0938376635313034, "global_step": 177034, "epoch": 1989, "lr": 2.708508257019938e-05}
{"train_loss": 0.08842718601226807, "global_step": 177035, "epoch": 1989, "lr": 2.7084567302765162e-05}
{"train_loss": 0.09954128414392471, "global_step": 177036, "epoch": 1989, "lr": 2.7084052038411612e-05}
{"train_loss": 0.04969821870326996, "global_step": 177037, "epoch": 1989, "lr": 2.7083536777138862e-05}
{"train_loss": 0.045624990016222, "global_step": 177038, "epoch": 1989, "lr": 2.7083021518946928e-05}
{"train_loss": 0.02680819481611252, "global_step": 177039, "epoch": 1989, "lr": 2.7082506263835916e-05}
{"train_loss": 0.06905119121074677, "global_step": 177040, "epoch": 1989, "lr": 2.7081991011805875e-05}
{"train_loss": 0.10671164095401764, "global_step": 177041, "epoch": 1989, "lr": 2.708147576285689e-05}
{"train_loss": 0.07751990109682083, "global_step": 177042, "epoch": 1989, "lr": 2.7080960516989024e-05}
{"train_loss": 0.08954513072967529, "global_step": 177043, "epoch": 1989, "lr": 2.7080445274202328e-05}
{"train_loss": 0.12860797345638275, "global_step": 177044, "epoch": 1989, "lr": 2.7079930034496904e-05}
{"train_loss": 0.06530755758285522, "global_step": 177045, "epoch": 1989, "lr": 2.7079414797872783e-05}
{"train_loss": 0.13145802915096283, "global_step": 177046, "epoch": 1989, "lr": 2.7078899564330073e-05}
{"train_loss": 0.10409853607416153, "global_step": 177047, "epoch": 1989, "lr": 2.7078384333868805e-05}
{"train_loss": 0.04823862016201019, "global_step": 177048, "epoch": 1989, "lr": 2.7077869106489096e-05}
{"train_loss": 0.09646401554346085, "global_step": 177049, "epoch": 1989, "lr": 2.7077353882190982e-05}
{"train_loss": 0.06668649613857269, "global_step": 177050, "epoch": 1989, "lr": 2.707683866097452e-05}
{"train_loss": 0.11135012656450272, "global_step": 177051, "epoch": 1989, "lr": 2.7076323442839795e-05}
{"train_loss": 0.06481967121362686, "global_step": 177052, "epoch": 1989, "lr": 2.7075808227786903e-05}
{"train_loss": 0.07247497141361237, "global_step": 177053, "epoch": 1989, "lr": 2.7075293015815866e-05}
{"train_loss": 0.06313470005989075, "global_step": 177054, "epoch": 1989, "lr": 2.70747778069268e-05}
{"train_loss": 0.07264299690723419, "global_step": 177055, "epoch": 1989, "lr": 2.7074262601119725e-05}
{"train_loss": 0.040781740099191666, "global_step": 177056, "epoch": 1989, "lr": 2.7073747398394767e-05}
{"train_loss": 0.10254170745611191, "global_step": 177057, "epoch": 1989, "lr": 2.7073232198751935e-05}
{"train_loss": 0.10489678382873535, "global_step": 177058, "epoch": 1989, "lr": 2.707271700219135e-05}
{"train_loss": 0.04636000841856003, "global_step": 177059, "epoch": 1989, "lr": 2.707220180871304e-05}
{"train_loss": 0.06273937225341797, "global_step": 177060, "epoch": 1989, "lr": 2.707168661831711e-05}
{"train_loss": 0.07154381275177002, "global_step": 177061, "epoch": 1989, "lr": 2.7071171431003594e-05}
{"train_loss": 0.09914321452379227, "global_step": 177062, "epoch": 1989, "lr": 2.7070656246772598e-05}
{"train_loss": 0.033746544271707535, "global_step": 177063, "epoch": 1989, "lr": 2.7070141065624154e-05}
{"train_loss": 0.08123036473989487, "global_step": 177064, "epoch": 1989, "lr": 2.706962588755837e-05}
{"train_loss": 0.1354372799396515, "global_step": 177065, "epoch": 1989, "lr": 2.7069110712575275e-05}
{"train_loss": 0.059675075113773346, "global_step": 177066, "epoch": 1989, "lr": 2.7068595540674978e-05}
{"train_loss": 0.04648829624056816, "global_step": 177067, "epoch": 1989, "lr": 2.7068080371857514e-05}
{"train_loss": 0.05394914001226425, "global_step": 177068, "epoch": 1989, "lr": 2.706756520612298e-05}
{"train_loss": 0.12077904492616653, "global_step": 177069, "epoch": 1989, "lr": 2.7067050043471422e-05}
{"train_loss": 0.08964904397726059, "global_step": 177070, "epoch": 1989, "lr": 2.7066534883902904e-05}
{"train_loss": 0.08844904601573944, "global_step": 177071, "epoch": 1989, "lr": 2.7066019727417546e-05}
{"train_loss": 0.07294150441884995, "global_step": 177072, "epoch": 1989, "lr": 2.7065504574015354e-05}
{"train_loss": 0.08944962918758392, "global_step": 177073, "epoch": 1989, "lr": 2.7064989423696447e-05}
{"train_loss": 0.06666514277458191, "global_step": 177074, "epoch": 1989, "lr": 2.7064474276460848e-05}
{"train_loss": 0.09597691893577576, "global_step": 177075, "epoch": 1989, "lr": 2.706395913230868e-05}
{"train_loss": 0.10419628024101257, "global_step": 177076, "epoch": 1989, "lr": 2.7063443991239958e-05}
{"train_loss": 0.04043243080377579, "global_step": 177077, "epoch": 1989, "lr": 2.70629288532548e-05}
{"train_loss": 0.14880822598934174, "global_step": 177078, "epoch": 1989, "lr": 2.706241371835323e-05}
{"train_loss": 0.0528700165450573, "global_step": 177079, "epoch": 1989, "lr": 2.7061898586535354e-05}
{"train_loss": 0.10706906020641327, "global_step": 177080, "epoch": 1989, "lr": 2.7061383457801215e-05}
{"train_loss": 0.09966466575860977, "global_step": 177081, "epoch": 1989, "lr": 2.706086833215091e-05}
{"train_loss": 0.11427143961191177, "global_step": 177082, "epoch": 1989, "lr": 2.7060353209584487e-05}
{"train_loss": 0.07911252230405807, "global_step": 177083, "epoch": 1989, "lr": 2.7059838090102004e-05}
{"train_loss": 0.0694463923573494, "global_step": 177084, "epoch": 1989, "lr": 2.705932297370356e-05}
{"train_loss": 0.06535769253969193, "global_step": 177085, "epoch": 1989, "lr": 2.70588078603892e-05}
{"train_loss": 0.06002838537096977, "global_step": 177086, "epoch": 1989, "lr": 2.705829275015902e-05}
{"train_loss": 0.06598510593175888, "global_step": 177087, "epoch": 1989, "lr": 2.705777764301305e-05}
{"train_loss": 0.07558287680149078, "global_step": 177088, "epoch": 1989, "lr": 2.7057262538951406e-05}
{"train_loss": 0.11580676585435867, "global_step": 177089, "epoch": 1989, "lr": 2.7056747437974113e-05}
{"train_loss": 0.08008168637752533, "global_step": 177090, "epoch": 1989, "lr": 2.7056232340081276e-05}
{"train_loss": 0.1206214502453804, "global_step": 177091, "epoch": 1989, "lr": 2.7055717245272938e-05}
{"train_loss": 0.08873521536588669, "global_step": 177092, "epoch": 1989, "lr": 2.7055202153549188e-05}
{"train_loss": 0.08390805125236511, "global_step": 177093, "epoch": 1989, "lr": 2.7054687064910078e-05}
{"train_loss": 0.08092299848794937, "global_step": 177094, "epoch": 1989, "lr": 2.7054171979355696e-05}
{"train_loss": 0.057366352528333664, "global_step": 177095, "epoch": 1989, "lr": 2.7053656896886082e-05}
{"train_loss": 0.05774008110165596, "global_step": 177096, "epoch": 1989, "lr": 2.705314181750135e-05}
{"train_loss": 0.0856221616268158, "global_step": 177097, "epoch": 1989, "lr": 2.705262674120152e-05}
{"train_loss": 0.06800493597984314, "global_step": 177098, "epoch": 1989, "lr": 2.70521116679867e-05}
{"train_loss": 0.10478290170431137, "global_step": 177099, "epoch": 1989, "lr": 2.705159659785693e-05}
{"train_loss": 0.15946230292320251, "global_step": 177100, "epoch": 1989, "lr": 2.7051081530812316e-05}
{"train_loss": 0.15314920246601105, "global_step": 177101, "epoch": 1989, "lr": 2.7050566466852877e-05}
{"train_loss": 0.09982576221227646, "global_step": 177102, "epoch": 1989, "lr": 2.7050051405978737e-05}
{"train_loss": 0.030675889924168587, "global_step": 177103, "epoch": 1989, "lr": 2.7049536348189913e-05}
{"train_loss": 0.10365954041481018, "global_step": 177104, "epoch": 1989, "lr": 2.704902129348652e-05}
{"train_loss": 0.07117584347724915, "global_step": 177105, "epoch": 1989, "lr": 2.7048506241868588e-05}
{"train_loss": 0.06796394288539886, "global_step": 177106, "epoch": 1989, "lr": 2.7047991193336196e-05}
{"train_loss": 0.06983377784490585, "global_step": 177107, "epoch": 1989, "lr": 2.704747614788945e-05}
{"train_loss": 0.1134396567940712, "global_step": 177108, "epoch": 1989, "lr": 2.7046961105528373e-05}
{"train_loss": 0.08286827815215239, "global_step": 177109, "epoch": 1989, "lr": 2.704644606625307e-05, "val_loss": 7.108856678009033}
{"train_loss": 0.11546554416418076, "global_step": 177110, "epoch": 1990, "lr": 2.704593103006357e-05}
{"train_loss": 0.07730764150619507, "global_step": 177111, "epoch": 1990, "lr": 2.704541599695999e-05}
{"train_loss": 0.13461804389953613, "global_step": 177112, "epoch": 1990, "lr": 2.7044900966942342e-05}
{"train_loss": 0.11904703825712204, "global_step": 177113, "epoch": 1990, "lr": 2.7044385940010765e-05}
{"train_loss": 0.09318381547927856, "global_step": 177114, "epoch": 1990, "lr": 2.704387091616526e-05}
{"train_loss": 0.09940337389707565, "global_step": 177115, "epoch": 1990, "lr": 2.704335589540594e-05}
{"train_loss": 0.16473212838172913, "global_step": 177116, "epoch": 1990, "lr": 2.7042840877732854e-05}
{"train_loss": 0.06380360573530197, "global_step": 177117, "epoch": 1990, "lr": 2.704232586314609e-05}
{"train_loss": 0.06672520935535431, "global_step": 177118, "epoch": 1990, "lr": 2.704181085164569e-05}
{"train_loss": 0.06528398394584656, "global_step": 177119, "epoch": 1990, "lr": 2.7041295843231763e-05}
{"train_loss": 0.1003647968173027, "global_step": 177120, "epoch": 1990, "lr": 2.7040780837904343e-05}
{"train_loss": 0.11375872790813446, "global_step": 177121, "epoch": 1990, "lr": 2.70402658356635e-05}
{"train_loss": 0.09394580125808716, "global_step": 177122, "epoch": 1990, "lr": 2.7039750836509327e-05}
{"train_loss": 0.08234051614999771, "global_step": 177123, "epoch": 1990, "lr": 2.7039235840441858e-05}
{"train_loss": 0.09492049366235733, "global_step": 177124, "epoch": 1990, "lr": 2.7038720847461208e-05}
{"train_loss": 0.05055660381913185, "global_step": 177125, "epoch": 1990, "lr": 2.70382058575674e-05}
{"train_loss": 0.0630292072892189, "global_step": 177126, "epoch": 1990, "lr": 2.7037690870760552e-05}
{"train_loss": 0.07052989304065704, "global_step": 177127, "epoch": 1990, "lr": 2.703717588704068e-05}
{"train_loss": 0.08967258781194687, "global_step": 177128, "epoch": 1990, "lr": 2.7036660906407902e-05}
{"train_loss": 0.06938191503286362, "global_step": 177129, "epoch": 1990, "lr": 2.7036145928862246e-05}
{"train_loss": 0.04676057770848274, "global_step": 177130, "epoch": 1990, "lr": 2.7035630954403817e-05}
{"train_loss": 0.12656155228614807, "global_step": 177131, "epoch": 1990, "lr": 2.7035115983032648e-05}
{"train_loss": 0.1025475561618805, "global_step": 177132, "epoch": 1990, "lr": 2.703460101474885e-05}
{"train_loss": 0.07415555417537689, "global_step": 177133, "epoch": 1990, "lr": 2.7034086049552448e-05}
{"train_loss": 0.07919259369373322, "global_step": 177134, "epoch": 1990, "lr": 2.7033571087443554e-05}
{"train_loss": 0.08283805847167969, "global_step": 177135, "epoch": 1990, "lr": 2.7033056128422196e-05}
{"train_loss": 0.0737251564860344, "global_step": 177136, "epoch": 1990, "lr": 2.7032541172488486e-05}
{"train_loss": 0.0974712148308754, "global_step": 177137, "epoch": 1990, "lr": 2.703202621964245e-05}
{"train_loss": 0.046703074127435684, "global_step": 177138, "epoch": 1990, "lr": 2.7031511269884196e-05}
{"train_loss": 0.06626810133457184, "global_step": 177139, "epoch": 1990, "lr": 2.7030996323213758e-05}
{"train_loss": 0.08343756198883057, "global_step": 177140, "epoch": 1990, "lr": 2.703048137963124e-05}
{"train_loss": 0.05835232883691788, "global_step": 177141, "epoch": 1990, "lr": 2.7029966439136677e-05}
{"train_loss": 0.1379912942647934, "global_step": 177142, "epoch": 1990, "lr": 2.7029451501730153e-05}
{"train_loss": 0.12449882924556732, "global_step": 177143, "epoch": 1990, "lr": 2.7028936567411755e-05}
{"train_loss": 0.0557829849421978, "global_step": 177144, "epoch": 1990, "lr": 2.702842163618152e-05}
{"train_loss": 0.043260712176561356, "global_step": 177145, "epoch": 1990, "lr": 2.7027906708039563e-05}
{"train_loss": 0.10633254051208496, "global_step": 177146, "epoch": 1990, "lr": 2.702739178298589e-05}
{"train_loss": 0.11056073009967804, "global_step": 177147, "epoch": 1990, "lr": 2.7026876861020634e-05}
{"train_loss": 0.08688120543956757, "global_step": 177148, "epoch": 1990, "lr": 2.7026361942143806e-05}
{"train_loss": 0.07542510330677032, "global_step": 177149, "epoch": 1990, "lr": 2.7025847026355532e-05}
{"train_loss": 0.051335323601961136, "global_step": 177150, "epoch": 1990, "lr": 2.7025332113655827e-05}
{"train_loss": 0.045319125056266785, "global_step": 177151, "epoch": 1990, "lr": 2.7024817204044805e-05}
{"train_loss": 0.08147512376308441, "global_step": 177152, "epoch": 1990, "lr": 2.70243022975225e-05}
{"train_loss": 0.041563455015420914, "global_step": 177153, "epoch": 1990, "lr": 2.7023787394089016e-05}
{"train_loss": 0.09531042724847794, "global_step": 177154, "epoch": 1990, "lr": 2.7023272493744388e-05}
{"train_loss": 0.09968198835849762, "global_step": 177155, "epoch": 1990, "lr": 2.7022757596488713e-05}
{"train_loss": 0.05960206314921379, "global_step": 177156, "epoch": 1990, "lr": 2.7022242702322032e-05}
{"train_loss": 0.051084525883197784, "global_step": 177157, "epoch": 1990, "lr": 2.7021727811244456e-05}
{"train_loss": 0.0952090471982956, "global_step": 177158, "epoch": 1990, "lr": 2.7021212923256e-05}
{"train_loss": 0.05127391591668129, "global_step": 177159, "epoch": 1990, "lr": 2.7020698038356784e-05}
{"train_loss": 0.08814278990030289, "global_step": 177160, "epoch": 1990, "lr": 2.7020183156546863e-05}
{"train_loss": 0.12321516126394272, "global_step": 177161, "epoch": 1990, "lr": 2.7019668277826272e-05}
{"train_loss": 0.11089161038398743, "global_step": 177162, "epoch": 1990, "lr": 2.7019153402195118e-05}
{"train_loss": 0.03384749963879585, "global_step": 177163, "epoch": 1990, "lr": 2.701863852965345e-05}
{"train_loss": 0.06929057091474533, "global_step": 177164, "epoch": 1990, "lr": 2.7018123660201366e-05}
{"train_loss": 0.1117255762219429, "global_step": 177165, "epoch": 1990, "lr": 2.7017608793838887e-05}
{"train_loss": 0.07538941502571106, "global_step": 177166, "epoch": 1990, "lr": 2.701709393056614e-05}
{"train_loss": 0.061962470412254333, "global_step": 177167, "epoch": 1990, "lr": 2.701657907038314e-05}
{"train_loss": 0.11655121296644211, "global_step": 177168, "epoch": 1990, "lr": 2.701606421329e-05}
{"train_loss": 0.07107454538345337, "global_step": 177169, "epoch": 1990, "lr": 2.701554935928675e-05}
{"train_loss": 0.06470058113336563, "global_step": 177170, "epoch": 1990, "lr": 2.7015034508373506e-05}
{"train_loss": 0.06481889635324478, "global_step": 177171, "epoch": 1990, "lr": 2.701451966055028e-05}
{"train_loss": 0.049672555178403854, "global_step": 177172, "epoch": 1990, "lr": 2.7014004815817195e-05}
{"train_loss": 0.13922330737113953, "global_step": 177173, "epoch": 1990, "lr": 2.7013489974174278e-05}
{"train_loss": 0.06815337389707565, "global_step": 177174, "epoch": 1990, "lr": 2.701297513562164e-05}
{"train_loss": 0.08674091100692749, "global_step": 177175, "epoch": 1990, "lr": 2.7012460300159305e-05}
{"train_loss": 0.10677958279848099, "global_step": 177176, "epoch": 1990, "lr": 2.701194546778738e-05}
{"train_loss": 0.06433028727769852, "global_step": 177177, "epoch": 1990, "lr": 2.7011430638505906e-05}
{"train_loss": 0.11763828992843628, "global_step": 177178, "epoch": 1990, "lr": 2.7010915812314964e-05}
{"train_loss": 0.10979542136192322, "global_step": 177179, "epoch": 1990, "lr": 2.701040098921464e-05}
{"train_loss": 0.08992310613393784, "global_step": 177180, "epoch": 1990, "lr": 2.7009886169204967e-05}
{"train_loss": 0.08065096288919449, "global_step": 177181, "epoch": 1990, "lr": 2.700937135228606e-05}
{"train_loss": 0.05330454558134079, "global_step": 177182, "epoch": 1990, "lr": 2.7008856538457938e-05}
{"train_loss": 0.17054259777069092, "global_step": 177183, "epoch": 1990, "lr": 2.7008341727720722e-05}
{"train_loss": 0.1196482852101326, "global_step": 177184, "epoch": 1990, "lr": 2.7007826920074425e-05}
{"train_loss": 0.06572555750608444, "global_step": 177185, "epoch": 1990, "lr": 2.7007312115519167e-05}
{"train_loss": 0.12208070605993271, "global_step": 177186, "epoch": 1990, "lr": 2.7006797314054977e-05}
{"train_loss": 0.1523083597421646, "global_step": 177187, "epoch": 1990, "lr": 2.700628251568197e-05}
{"train_loss": 0.06006179749965668, "global_step": 177188, "epoch": 1990, "lr": 2.700576772040016e-05}
{"train_loss": 0.09533502161502838, "global_step": 177189, "epoch": 1990, "lr": 2.7005252928209667e-05}
{"train_loss": 0.07549890130758286, "global_step": 177190, "epoch": 1990, "lr": 2.7004738139110518e-05}
{"train_loss": 0.06337670981884003, "global_step": 177191, "epoch": 1990, "lr": 2.700422335310282e-05}
{"train_loss": 0.1213466003537178, "global_step": 177192, "epoch": 1990, "lr": 2.7003708570186604e-05}
{"train_loss": 0.06811061501502991, "global_step": 177193, "epoch": 1990, "lr": 2.700319379036198e-05}
{"train_loss": 0.08316991478204727, "global_step": 177194, "epoch": 1990, "lr": 2.7002679013628974e-05}
{"train_loss": 0.12058830261230469, "global_step": 177195, "epoch": 1990, "lr": 2.70021642399877e-05}
{"train_loss": 0.1003529354929924, "global_step": 177196, "epoch": 1990, "lr": 2.7001649469438185e-05}
{"train_loss": 0.015364941209554672, "global_step": 177197, "epoch": 1990, "lr": 2.7001134701980535e-05}
{"train_loss": 0.08551319475086887, "global_step": 177198, "epoch": 1990, "lr": 2.7000619937614802e-05, "val_loss": 7.288672924041748, "train_action_mse_error": 14.039264678955078}
{"train_loss": 0.07559239119291306, "global_step": 177199, "epoch": 1991, "lr": 2.700010517634104e-05}
{"train_loss": 0.0644637942314148, "global_step": 177200, "epoch": 1991, "lr": 2.6999590418159347e-05}
{"train_loss": 0.0990019217133522, "global_step": 177201, "epoch": 1991, "lr": 2.6999075663069763e-05}
{"train_loss": 0.12451986968517303, "global_step": 177202, "epoch": 1991, "lr": 2.6998560911072395e-05}
{"train_loss": 0.04782770574092865, "global_step": 177203, "epoch": 1991, "lr": 2.6998046162167267e-05}
{"train_loss": 0.10815661400556564, "global_step": 177204, "epoch": 1991, "lr": 2.6997531416354493e-05}
{"train_loss": 0.053278714418411255, "global_step": 177205, "epoch": 1991, "lr": 2.69970166736341e-05}
{"train_loss": 0.07156287133693695, "global_step": 177206, "epoch": 1991, "lr": 2.6996501934006196e-05}
{"train_loss": 0.06384836882352829, "global_step": 177207, "epoch": 1991, "lr": 2.699598719747082e-05}
{"train_loss": 0.07466598600149155, "global_step": 177208, "epoch": 1991, "lr": 2.699547246402806e-05}
{"train_loss": 0.05091312155127525, "global_step": 177209, "epoch": 1991, "lr": 2.699495773367796e-05}
{"train_loss": 0.0426829494535923, "global_step": 177210, "epoch": 1991, "lr": 2.6994443006420633e-05}
{"train_loss": 0.13758555054664612, "global_step": 177211, "epoch": 1991, "lr": 2.699392828225611e-05}
{"train_loss": 0.09174111485481262, "global_step": 177212, "epoch": 1991, "lr": 2.6993413561184478e-05}
{"train_loss": 0.07477234303951263, "global_step": 177213, "epoch": 1991, "lr": 2.6992898843205793e-05}
{"train_loss": 0.11183279007673264, "global_step": 177214, "epoch": 1991, "lr": 2.6992384128320124e-05}
{"train_loss": 0.07280664145946503, "global_step": 177215, "epoch": 1991, "lr": 2.699186941652757e-05}
{"train_loss": 0.13771188259124756, "global_step": 177216, "epoch": 1991, "lr": 2.6991354707828166e-05}
{"train_loss": 0.061262764036655426, "global_step": 177217, "epoch": 1991, "lr": 2.6990840002222007e-05}
{"train_loss": 0.09004110097885132, "global_step": 177218, "epoch": 1991, "lr": 2.699032529970913e-05}
{"train_loss": 0.040000878274440765, "global_step": 177219, "epoch": 1991, "lr": 2.6989810600289645e-05}
{"train_loss": 0.04446052014827728, "global_step": 177220, "epoch": 1991, "lr": 2.698929590396358e-05}
{"train_loss": 0.056825343519449234, "global_step": 177221, "epoch": 1991, "lr": 2.698878121073104e-05}
{"train_loss": 0.08105089515447617, "global_step": 177222, "epoch": 1991, "lr": 2.6988266520592064e-05}
{"train_loss": 0.0986204445362091, "global_step": 177223, "epoch": 1991, "lr": 2.6987751833546748e-05}
{"train_loss": 0.14233334362506866, "global_step": 177224, "epoch": 1991, "lr": 2.6987237149595135e-05}
{"train_loss": 0.10896894335746765, "global_step": 177225, "epoch": 1991, "lr": 2.6986722468737324e-05}
{"train_loss": 0.07775843143463135, "global_step": 177226, "epoch": 1991, "lr": 2.6986207790973344e-05}
{"train_loss": 0.057198140770196915, "global_step": 177227, "epoch": 1991, "lr": 2.698569311630331e-05}
{"train_loss": 0.05520091950893402, "global_step": 177228, "epoch": 1991, "lr": 2.6985178444727256e-05}
{"train_loss": 0.11539115756750107, "global_step": 177229, "epoch": 1991, "lr": 2.6984663776245277e-05}
{"train_loss": 0.10297399759292603, "global_step": 177230, "epoch": 1991, "lr": 2.698414911085741e-05}
{"train_loss": 0.13220269978046417, "global_step": 177231, "epoch": 1991, "lr": 2.6983634448563766e-05}
{"train_loss": 0.07028309255838394, "global_step": 177232, "epoch": 1991, "lr": 2.698311978936437e-05}
{"train_loss": 0.054106585681438446, "global_step": 177233, "epoch": 1991, "lr": 2.698260513325933e-05}
{"train_loss": 0.09971960633993149, "global_step": 177234, "epoch": 1991, "lr": 2.698209048024869e-05}
{"train_loss": 0.06739674508571625, "global_step": 177235, "epoch": 1991, "lr": 2.6981575830332535e-05}
{"train_loss": 0.04910070821642876, "global_step": 177236, "epoch": 1991, "lr": 2.698106118351093e-05}
{"train_loss": 0.08061739057302475, "global_step": 177237, "epoch": 1991, "lr": 2.6980546539783925e-05}
{"train_loss": 0.09453233331441879, "global_step": 177238, "epoch": 1991, "lr": 2.698003189915162e-05}
{"train_loss": 0.061592403799295425, "global_step": 177239, "epoch": 1991, "lr": 2.6979517261614052e-05}
{"train_loss": 0.07212917506694794, "global_step": 177240, "epoch": 1991, "lr": 2.6979002627171325e-05}
{"train_loss": 0.07358171045780182, "global_step": 177241, "epoch": 1991, "lr": 2.697848799582347e-05}
{"train_loss": 0.09746681898832321, "global_step": 177242, "epoch": 1991, "lr": 2.6977973367570596e-05}
{"train_loss": 0.054630205035209656, "global_step": 177243, "epoch": 1991, "lr": 2.6977458742412736e-05}
{"train_loss": 0.05481168255209923, "global_step": 177244, "epoch": 1991, "lr": 2.697694412035e-05}
{"train_loss": 0.10856177657842636, "global_step": 177245, "epoch": 1991, "lr": 2.6976429501382404e-05}
{"train_loss": 0.08321379870176315, "global_step": 177246, "epoch": 1991, "lr": 2.697591488551007e-05}
{"train_loss": 0.08808454126119614, "global_step": 177247, "epoch": 1991, "lr": 2.6975400272733024e-05}
{"train_loss": 0.07313818484544754, "global_step": 177248, "epoch": 1991, "lr": 2.6974885663051374e-05}
{"train_loss": 0.0456852950155735, "global_step": 177249, "epoch": 1991, "lr": 2.6974371056465152e-05}
{"train_loss": 0.048441849648952484, "global_step": 177250, "epoch": 1991, "lr": 2.6973856452974446e-05}
{"train_loss": 0.027238942682743073, "global_step": 177251, "epoch": 1991, "lr": 2.6973341852579346e-05}
{"train_loss": 0.04880573973059654, "global_step": 177252, "epoch": 1991, "lr": 2.6972827255279875e-05}
{"train_loss": 0.0397866815328598, "global_step": 177253, "epoch": 1991, "lr": 2.697231266107615e-05}
{"train_loss": 0.08378686010837555, "global_step": 177254, "epoch": 1991, "lr": 2.697179806996819e-05}
{"train_loss": 0.09429041296243668, "global_step": 177255, "epoch": 1991, "lr": 2.6971283481956123e-05}
{"train_loss": 0.0835423693060875, "global_step": 177256, "epoch": 1991, "lr": 2.6970768897039955e-05}
{"train_loss": 0.08266042172908783, "global_step": 177257, "epoch": 1991, "lr": 2.697025431521981e-05}
{"train_loss": 0.07134595513343811, "global_step": 177258, "epoch": 1991, "lr": 2.6969739736495714e-05}
{"train_loss": 0.13019594550132751, "global_step": 177259, "epoch": 1991, "lr": 2.6969225160867777e-05}
{"train_loss": 0.10445507615804672, "global_step": 177260, "epoch": 1991, "lr": 2.6968710588336026e-05}
{"train_loss": 0.05752837285399437, "global_step": 177261, "epoch": 1991, "lr": 2.696819601890057e-05}
{"train_loss": 0.07976657152175903, "global_step": 177262, "epoch": 1991, "lr": 2.6967681452561444e-05}
{"train_loss": 0.0650554820895195, "global_step": 177263, "epoch": 1991, "lr": 2.696716688931875e-05}
{"train_loss": 0.13242633640766144, "global_step": 177264, "epoch": 1991, "lr": 2.6966652329172513e-05}
{"train_loss": 0.08567604422569275, "global_step": 177265, "epoch": 1991, "lr": 2.6966137772122858e-05}
{"train_loss": 0.04667368531227112, "global_step": 177266, "epoch": 1991, "lr": 2.69656232181698e-05}
{"train_loss": 0.1089363545179367, "global_step": 177267, "epoch": 1991, "lr": 2.6965108667313456e-05}
{"train_loss": 0.03209391236305237, "global_step": 177268, "epoch": 1991, "lr": 2.6964594119553855e-05}
{"train_loss": 0.1200910210609436, "global_step": 177269, "epoch": 1991, "lr": 2.6964079574891098e-05}
{"train_loss": 0.046451594680547714, "global_step": 177270, "epoch": 1991, "lr": 2.6963565033325222e-05}
{"train_loss": 0.08163666725158691, "global_step": 177271, "epoch": 1991, "lr": 2.6963050494856336e-05}
{"train_loss": 0.07294009625911713, "global_step": 177272, "epoch": 1991, "lr": 2.696253595948446e-05}
{"train_loss": 0.060065679252147675, "global_step": 177273, "epoch": 1991, "lr": 2.6962021427209715e-05}
{"train_loss": 0.04514973983168602, "global_step": 177274, "epoch": 1991, "lr": 2.696150689803213e-05}
{"train_loss": 0.0674462616443634, "global_step": 177275, "epoch": 1991, "lr": 2.6960992371951798e-05}
{"train_loss": 0.07416775822639465, "global_step": 177276, "epoch": 1991, "lr": 2.696047784896879e-05}
{"train_loss": 0.10066143423318863, "global_step": 177277, "epoch": 1991, "lr": 2.6959963329083137e-05}
{"train_loss": 0.07207649201154709, "global_step": 177278, "epoch": 1991, "lr": 2.6959448812294963e-05}
{"train_loss": 0.04620277136564255, "global_step": 177279, "epoch": 1991, "lr": 2.6958934298604288e-05}
{"train_loss": 0.03776361420750618, "global_step": 177280, "epoch": 1991, "lr": 2.6958419788011225e-05}
{"train_loss": 0.08623089641332626, "global_step": 177281, "epoch": 1991, "lr": 2.69579052805158e-05}
{"train_loss": 0.08228133618831635, "global_step": 177282, "epoch": 1991, "lr": 2.6957390776118126e-05}
{"train_loss": 0.10143072158098221, "global_step": 177283, "epoch": 1991, "lr": 2.6956876274818227e-05}
{"train_loss": 0.09698332846164703, "global_step": 177284, "epoch": 1991, "lr": 2.6956361776616213e-05}
{"train_loss": 0.047576624900102615, "global_step": 177285, "epoch": 1991, "lr": 2.695584728151212e-05}
{"train_loss": 0.07255411148071289, "global_step": 177286, "epoch": 1991, "lr": 2.6955332789506027e-05}
{"train_loss": 0.07787114172504189, "global_step": 177287, "epoch": 1991, "lr": 2.6954818300598038e-05, "val_loss": 7.418025016784668}
{"train_loss": 0.06574591249227524, "global_step": 177288, "epoch": 1992, "lr": 2.6954303814788163e-05}
{"train_loss": 0.10612726956605911, "global_step": 177289, "epoch": 1992, "lr": 2.695378933207653e-05}
{"train_loss": 0.1341976672410965, "global_step": 177290, "epoch": 1992, "lr": 2.695327485246315e-05}
{"train_loss": 0.04673156142234802, "global_step": 177291, "epoch": 1992, "lr": 2.6952760375948148e-05}
{"train_loss": 0.06072912737727165, "global_step": 177292, "epoch": 1992, "lr": 2.695224590253154e-05}
{"train_loss": 0.0488317534327507, "global_step": 177293, "epoch": 1992, "lr": 2.6951731432213445e-05}
{"train_loss": 0.09234116226434708, "global_step": 177294, "epoch": 1992, "lr": 2.695121696499389e-05}
{"train_loss": 0.09646067768335342, "global_step": 177295, "epoch": 1992, "lr": 2.6950702500872982e-05}
{"train_loss": 0.12402446568012238, "global_step": 177296, "epoch": 1992, "lr": 2.6950188039850754e-05}
{"train_loss": 0.05852910876274109, "global_step": 177297, "epoch": 1992, "lr": 2.6949673581927308e-05}
{"train_loss": 0.059601135551929474, "global_step": 177298, "epoch": 1992, "lr": 2.694915912710268e-05}
{"train_loss": 0.0737534686923027, "global_step": 177299, "epoch": 1992, "lr": 2.6948644675376976e-05}
{"train_loss": 0.07815248519182205, "global_step": 177300, "epoch": 1992, "lr": 2.6948130226750224e-05}
{"train_loss": 0.049235664308071136, "global_step": 177301, "epoch": 1992, "lr": 2.6947615781222536e-05}
{"train_loss": 0.10519109666347504, "global_step": 177302, "epoch": 1992, "lr": 2.6947101338793945e-05}
{"train_loss": 0.0762052834033966, "global_step": 177303, "epoch": 1992, "lr": 2.694658689946455e-05}
{"train_loss": 0.056254103779792786, "global_step": 177304, "epoch": 1992, "lr": 2.6946072463234383e-05}
{"train_loss": 0.05553753674030304, "global_step": 177305, "epoch": 1992, "lr": 2.694555803010354e-05}
{"train_loss": 0.08571325987577438, "global_step": 177306, "epoch": 1992, "lr": 2.6945043600072107e-05}
{"train_loss": 0.04731905087828636, "global_step": 177307, "epoch": 1992, "lr": 2.694452917314011e-05}
{"train_loss": 0.0583701953291893, "global_step": 177308, "epoch": 1992, "lr": 2.6944014749307656e-05}
{"train_loss": 0.06106363609433174, "global_step": 177309, "epoch": 1992, "lr": 2.69435003285748e-05}
{"train_loss": 0.06903253495693207, "global_step": 177310, "epoch": 1992, "lr": 2.694298591094159e-05}
{"train_loss": 0.06296595931053162, "global_step": 177311, "epoch": 1992, "lr": 2.694247149640814e-05}
{"train_loss": 0.08446301519870758, "global_step": 177312, "epoch": 1992, "lr": 2.6941957084974467e-05}
{"train_loss": 0.051140688359737396, "global_step": 177313, "epoch": 1992, "lr": 2.6941442676640693e-05}
{"train_loss": 0.1035814881324768, "global_step": 177314, "epoch": 1992, "lr": 2.694092827140685e-05}
{"train_loss": 0.04841405898332596, "global_step": 177315, "epoch": 1992, "lr": 2.6940413869273006e-05}
{"train_loss": 0.11629863828420639, "global_step": 177316, "epoch": 1992, "lr": 2.6939899470239265e-05}
{"train_loss": 0.07353202998638153, "global_step": 177317, "epoch": 1992, "lr": 2.693938507430565e-05}
{"train_loss": 0.040585264563560486, "global_step": 177318, "epoch": 1992, "lr": 2.6938870681472272e-05}
{"train_loss": 0.07042558491230011, "global_step": 177319, "epoch": 1992, "lr": 2.6938356291739163e-05}
{"train_loss": 0.055342480540275574, "global_step": 177320, "epoch": 1992, "lr": 2.693784190510643e-05}
{"train_loss": 0.09253115952014923, "global_step": 177321, "epoch": 1992, "lr": 2.6937327521574108e-05}
{"train_loss": 0.057621490210294724, "global_step": 177322, "epoch": 1992, "lr": 2.6936813141142298e-05}
{"train_loss": 0.10742989927530289, "global_step": 177323, "epoch": 1992, "lr": 2.693629876381103e-05}
{"train_loss": 0.08743412047624588, "global_step": 177324, "epoch": 1992, "lr": 2.6935784389580397e-05}
{"train_loss": 0.0701296329498291, "global_step": 177325, "epoch": 1992, "lr": 2.6935270018450486e-05}
{"train_loss": 0.12057795375585556, "global_step": 177326, "epoch": 1992, "lr": 2.693475565042133e-05}
{"train_loss": 0.050241973251104355, "global_step": 177327, "epoch": 1992, "lr": 2.6934241285493033e-05}
{"train_loss": 0.06450164318084717, "global_step": 177328, "epoch": 1992, "lr": 2.6933726923665624e-05}
{"train_loss": 0.07282818853855133, "global_step": 177329, "epoch": 1992, "lr": 2.6933212564939213e-05}
{"train_loss": 0.06718545407056808, "global_step": 177330, "epoch": 1992, "lr": 2.693269820931384e-05}
{"train_loss": 0.07008090615272522, "global_step": 177331, "epoch": 1992, "lr": 2.6932183856789594e-05}
{"train_loss": 0.09220841526985168, "global_step": 177332, "epoch": 1992, "lr": 2.693166950736652e-05}
{"train_loss": 0.10955692082643509, "global_step": 177333, "epoch": 1992, "lr": 2.6931155161044718e-05}
{"train_loss": 0.14144203066825867, "global_step": 177334, "epoch": 1992, "lr": 2.6930640817824228e-05}
{"train_loss": 0.07351931929588318, "global_step": 177335, "epoch": 1992, "lr": 2.6930126477705148e-05}
{"train_loss": 0.10828433930873871, "global_step": 177336, "epoch": 1992, "lr": 2.692961214068751e-05}
{"train_loss": 0.08277063816785812, "global_step": 177337, "epoch": 1992, "lr": 2.692909780677143e-05}
{"train_loss": 0.047954365611076355, "global_step": 177338, "epoch": 1992, "lr": 2.6928583475956926e-05}
{"train_loss": 0.06327858567237854, "global_step": 177339, "epoch": 1992, "lr": 2.692806914824411e-05}
{"train_loss": 0.039620786905288696, "global_step": 177340, "epoch": 1992, "lr": 2.6927554823633022e-05}
{"train_loss": 0.10407896339893341, "global_step": 177341, "epoch": 1992, "lr": 2.6927040502123734e-05}
{"train_loss": 0.08410804718732834, "global_step": 177342, "epoch": 1992, "lr": 2.692652618371635e-05}
{"train_loss": 0.09658744186162949, "global_step": 177343, "epoch": 1992, "lr": 2.6926011868410893e-05}
{"train_loss": 0.09113424271345139, "global_step": 177344, "epoch": 1992, "lr": 2.6925497556207475e-05}
{"train_loss": 0.04332159087061882, "global_step": 177345, "epoch": 1992, "lr": 2.692498324710612e-05}
{"train_loss": 0.13599523901939392, "global_step": 177346, "epoch": 1992, "lr": 2.692446894110694e-05}
{"train_loss": 0.08380033820867538, "global_step": 177347, "epoch": 1992, "lr": 2.692395463820998e-05}
{"train_loss": 0.08911126852035522, "global_step": 177348, "epoch": 1992, "lr": 2.6923440338415297e-05}
{"train_loss": 0.036280520260334015, "global_step": 177349, "epoch": 1992, "lr": 2.6922926041722997e-05}
{"train_loss": 0.07684104144573212, "global_step": 177350, "epoch": 1992, "lr": 2.6922411748133108e-05}
{"train_loss": 0.0766049325466156, "global_step": 177351, "epoch": 1992, "lr": 2.692189745764574e-05}
{"train_loss": 0.035769470036029816, "global_step": 177352, "epoch": 1992, "lr": 2.692138317026094e-05}
{"train_loss": 0.1399642378091812, "global_step": 177353, "epoch": 1992, "lr": 2.6920868885978757e-05}
{"train_loss": 0.05931556224822998, "global_step": 177354, "epoch": 1992, "lr": 2.6920354604799303e-05}
{"train_loss": 0.12078562378883362, "global_step": 177355, "epoch": 1992, "lr": 2.6919840326722606e-05}
{"train_loss": 0.06411363929510117, "global_step": 177356, "epoch": 1992, "lr": 2.6919326051748773e-05}
{"train_loss": 0.05438445508480072, "global_step": 177357, "epoch": 1992, "lr": 2.6918811779877838e-05}
{"train_loss": 0.09639359265565872, "global_step": 177358, "epoch": 1992, "lr": 2.6918297511109914e-05}
{"train_loss": 0.06600704789161682, "global_step": 177359, "epoch": 1992, "lr": 2.6917783245445017e-05}
{"train_loss": 0.05264681205153465, "global_step": 177360, "epoch": 1992, "lr": 2.691726898288324e-05}
{"train_loss": 0.0470816045999527, "global_step": 177361, "epoch": 1992, "lr": 2.6916754723424674e-05}
{"train_loss": 0.07868625223636627, "global_step": 177362, "epoch": 1992, "lr": 2.6916240467069354e-05}
{"train_loss": 0.08345123380422592, "global_step": 177363, "epoch": 1992, "lr": 2.6915726213817382e-05}
{"train_loss": 0.14073680341243744, "global_step": 177364, "epoch": 1992, "lr": 2.6915211963668797e-05}
{"train_loss": 0.10544625669717789, "global_step": 177365, "epoch": 1992, "lr": 2.6914697716623687e-05}
{"train_loss": 0.07361679524183273, "global_step": 177366, "epoch": 1992, "lr": 2.6914183472682102e-05}
{"train_loss": 0.08969029784202576, "global_step": 177367, "epoch": 1992, "lr": 2.691366923184414e-05}
{"train_loss": 0.0944879874587059, "global_step": 177368, "epoch": 1992, "lr": 2.691315499410984e-05}
{"train_loss": 0.10218207538127899, "global_step": 177369, "epoch": 1992, "lr": 2.6912640759479302e-05}
{"train_loss": 0.09591501206159592, "global_step": 177370, "epoch": 1992, "lr": 2.6912126527952553e-05}
{"train_loss": 0.08216933906078339, "global_step": 177371, "epoch": 1992, "lr": 2.6911612299529716e-05}
{"train_loss": 0.05985356122255325, "global_step": 177372, "epoch": 1992, "lr": 2.6911098074210805e-05}
{"train_loss": 0.11250431090593338, "global_step": 177373, "epoch": 1992, "lr": 2.691058385199594e-05}
{"train_loss": 0.08698958903551102, "global_step": 177374, "epoch": 1992, "lr": 2.691006963288514e-05}
{"train_loss": 0.061848290264606476, "global_step": 177375, "epoch": 1992, "lr": 2.6909555416878525e-05}
{"train_loss": 0.07938333822602636, "global_step": 177376, "epoch": 1992, "lr": 2.6909041203976116e-05, "val_loss": 7.256900310516357}
{"train_loss": 0.051324184983968735, "global_step": 177377, "epoch": 1993, "lr": 2.6908526994178006e-05}
{"train_loss": 0.04536362737417221, "global_step": 177378, "epoch": 1993, "lr": 2.690801278748428e-05}
{"train_loss": 0.10306298732757568, "global_step": 177379, "epoch": 1993, "lr": 2.6907498583894973e-05}
{"train_loss": 0.049785830080509186, "global_step": 177380, "epoch": 1993, "lr": 2.6906984383410193e-05}
{"train_loss": 0.028075972571969032, "global_step": 177381, "epoch": 1993, "lr": 2.6906470186029965e-05}
{"train_loss": 0.11404296010732651, "global_step": 177382, "epoch": 1993, "lr": 2.69059559917544e-05}
{"train_loss": 0.10519880801439285, "global_step": 177383, "epoch": 1993, "lr": 2.690544180058353e-05}
{"train_loss": 0.10180018097162247, "global_step": 177384, "epoch": 1993, "lr": 2.690492761251746e-05}
{"train_loss": 0.06332138925790787, "global_step": 177385, "epoch": 1993, "lr": 2.690441342755624e-05}
{"train_loss": 0.07746051251888275, "global_step": 177386, "epoch": 1993, "lr": 2.6903899245699927e-05}
{"train_loss": 0.0794459730386734, "global_step": 177387, "epoch": 1993, "lr": 2.6903385066948616e-05}
{"train_loss": 0.06838065385818481, "global_step": 177388, "epoch": 1993, "lr": 2.6902870891302346e-05}
{"train_loss": 0.09323610365390778, "global_step": 177389, "epoch": 1993, "lr": 2.6902356718761223e-05}
{"train_loss": 0.10038116574287415, "global_step": 177390, "epoch": 1993, "lr": 2.6901842549325273e-05}
{"train_loss": 0.0783214271068573, "global_step": 177391, "epoch": 1993, "lr": 2.6901328382994617e-05}
{"train_loss": 0.10502278804779053, "global_step": 177392, "epoch": 1993, "lr": 2.6900814219769288e-05}
{"train_loss": 0.06623418629169464, "global_step": 177393, "epoch": 1993, "lr": 2.6900300059649348e-05}
{"train_loss": 0.05727937072515488, "global_step": 177394, "epoch": 1993, "lr": 2.68997859026349e-05}
{"train_loss": 0.08628189563751221, "global_step": 177395, "epoch": 1993, "lr": 2.6899271748725974e-05}
{"train_loss": 0.08343666046857834, "global_step": 177396, "epoch": 1993, "lr": 2.689875759792265e-05}
{"train_loss": 0.08205555379390717, "global_step": 177397, "epoch": 1993, "lr": 2.6898243450225042e-05}
{"train_loss": 0.05175801366567612, "global_step": 177398, "epoch": 1993, "lr": 2.6897729305633158e-05}
{"train_loss": 0.13152073323726654, "global_step": 177399, "epoch": 1993, "lr": 2.689721516414711e-05}
{"train_loss": 0.07615865021944046, "global_step": 177400, "epoch": 1993, "lr": 2.6896701025766934e-05}
{"train_loss": 0.0708833858370781, "global_step": 177401, "epoch": 1993, "lr": 2.689618689049273e-05}
{"train_loss": 0.06562100350856781, "global_step": 177402, "epoch": 1993, "lr": 2.6895672758324536e-05}
{"train_loss": 0.07850218564271927, "global_step": 177403, "epoch": 1993, "lr": 2.689515862926245e-05}
{"train_loss": 0.12523309886455536, "global_step": 177404, "epoch": 1993, "lr": 2.6894644503306522e-05}
{"train_loss": 0.0760086178779602, "global_step": 177405, "epoch": 1993, "lr": 2.6894130380456838e-05}
{"train_loss": 0.051210492849349976, "global_step": 177406, "epoch": 1993, "lr": 2.6893616260713438e-05}
{"train_loss": 0.07092834264039993, "global_step": 177407, "epoch": 1993, "lr": 2.6893102144076425e-05}
{"train_loss": 0.06744378805160522, "global_step": 177408, "epoch": 1993, "lr": 2.689258803054584e-05}
{"train_loss": 0.1051509827375412, "global_step": 177409, "epoch": 1993, "lr": 2.6892073920121786e-05}
{"train_loss": 0.0740199014544487, "global_step": 177410, "epoch": 1993, "lr": 2.6891559812804285e-05}
{"train_loss": 0.0843837782740593, "global_step": 177411, "epoch": 1993, "lr": 2.6891045708593456e-05}
{"train_loss": 0.05466669425368309, "global_step": 177412, "epoch": 1993, "lr": 2.6890531607489324e-05}
{"train_loss": 0.06340941041707993, "global_step": 177413, "epoch": 1993, "lr": 2.6890017509491973e-05}
{"train_loss": 0.04634556174278259, "global_step": 177414, "epoch": 1993, "lr": 2.68895034146015e-05}
{"train_loss": 0.060117024928331375, "global_step": 177415, "epoch": 1993, "lr": 2.6888989322817937e-05}
{"train_loss": 0.11707327514886856, "global_step": 177416, "epoch": 1993, "lr": 2.6888475234141387e-05}
{"train_loss": 0.1166190579533577, "global_step": 177417, "epoch": 1993, "lr": 2.6887961148571873e-05}
{"train_loss": 0.059982433915138245, "global_step": 177418, "epoch": 1993, "lr": 2.6887447066109517e-05}
{"train_loss": 0.12849651277065277, "global_step": 177419, "epoch": 1993, "lr": 2.6886932986754338e-05}
{"train_loss": 0.1082136332988739, "global_step": 177420, "epoch": 1993, "lr": 2.688641891050645e-05}
{"train_loss": 0.07334952056407928, "global_step": 177421, "epoch": 1993, "lr": 2.6885904837365883e-05}
{"train_loss": 0.075773686170578, "global_step": 177422, "epoch": 1993, "lr": 2.6885390767332742e-05}
{"train_loss": 0.04666009172797203, "global_step": 177423, "epoch": 1993, "lr": 2.6884876700407057e-05}
{"train_loss": 0.06354938447475433, "global_step": 177424, "epoch": 1993, "lr": 2.6884362636588945e-05}
{"train_loss": 0.1053377240896225, "global_step": 177425, "epoch": 1993, "lr": 2.6883848575878446e-05}
{"train_loss": 0.06600870937108994, "global_step": 177426, "epoch": 1993, "lr": 2.688333451827561e-05}
{"train_loss": 0.10932368040084839, "global_step": 177427, "epoch": 1993, "lr": 2.6882820463780545e-05}
{"train_loss": 0.06475452333688736, "global_step": 177428, "epoch": 1993, "lr": 2.688230641239329e-05}
{"train_loss": 0.05945413187146187, "global_step": 177429, "epoch": 1993, "lr": 2.6881792364113943e-05}
{"train_loss": 0.13270129263401031, "global_step": 177430, "epoch": 1993, "lr": 2.688127831894256e-05}
{"train_loss": 0.11080651730298996, "global_step": 177431, "epoch": 1993, "lr": 2.688076427687919e-05}
{"train_loss": 0.04220884293317795, "global_step": 177432, "epoch": 1993, "lr": 2.688025023792391e-05}
{"train_loss": 0.08865105360746384, "global_step": 177433, "epoch": 1993, "lr": 2.687973620207682e-05}
{"train_loss": 0.09293599426746368, "global_step": 177434, "epoch": 1993, "lr": 2.6879222169337958e-05}
{"train_loss": 0.09215851128101349, "global_step": 177435, "epoch": 1993, "lr": 2.6878708139707415e-05}
{"train_loss": 0.13704366981983185, "global_step": 177436, "epoch": 1993, "lr": 2.6878194113185228e-05}
{"train_loss": 0.08719369024038315, "global_step": 177437, "epoch": 1993, "lr": 2.6877680089771507e-05}
{"train_loss": 0.0399695485830307, "global_step": 177438, "epoch": 1993, "lr": 2.687716606946628e-05}
{"train_loss": 0.054986897855997086, "global_step": 177439, "epoch": 1993, "lr": 2.6876652052269656e-05}
{"train_loss": 0.10850390791893005, "global_step": 177440, "epoch": 1993, "lr": 2.6876138038181662e-05}
{"train_loss": 0.09883318841457367, "global_step": 177441, "epoch": 1993, "lr": 2.687562402720242e-05}
{"train_loss": 0.12172091752290726, "global_step": 177442, "epoch": 1993, "lr": 2.6875110019331938e-05}
{"train_loss": 0.13637028634548187, "global_step": 177443, "epoch": 1993, "lr": 2.6874596014570336e-05}
{"train_loss": 0.10866675525903702, "global_step": 177444, "epoch": 1993, "lr": 2.6874082012917644e-05}
{"train_loss": 0.08740949630737305, "global_step": 177445, "epoch": 1993, "lr": 2.6873568014373972e-05}
{"train_loss": 0.0510285347700119, "global_step": 177446, "epoch": 1993, "lr": 2.6873054018939348e-05}
{"train_loss": 0.09543953835964203, "global_step": 177447, "epoch": 1993, "lr": 2.687254002661388e-05}
{"train_loss": 0.02742830663919449, "global_step": 177448, "epoch": 1993, "lr": 2.687202603739759e-05}
{"train_loss": 0.040675535798072815, "global_step": 177449, "epoch": 1993, "lr": 2.6871512051290597e-05}
{"train_loss": 0.04130730777978897, "global_step": 177450, "epoch": 1993, "lr": 2.687099806829293e-05}
{"train_loss": 0.08132067322731018, "global_step": 177451, "epoch": 1993, "lr": 2.687048408840468e-05}
{"train_loss": 0.07719357311725616, "global_step": 177452, "epoch": 1993, "lr": 2.6869970111625918e-05}
{"train_loss": 0.02699965052306652, "global_step": 177453, "epoch": 1993, "lr": 2.68694561379567e-05}
{"train_loss": 0.12964481115341187, "global_step": 177454, "epoch": 1993, "lr": 2.6868942167397117e-05}
{"train_loss": 0.0556366927921772, "global_step": 177455, "epoch": 1993, "lr": 2.6868428199947204e-05}
{"train_loss": 0.10071347653865814, "global_step": 177456, "epoch": 1993, "lr": 2.686791423560706e-05}
{"train_loss": 0.06583824753761292, "global_step": 177457, "epoch": 1993, "lr": 2.686740027437673e-05}
{"train_loss": 0.04618970304727554, "global_step": 177458, "epoch": 1993, "lr": 2.6866886316256322e-05}
{"train_loss": 0.07944335043430328, "global_step": 177459, "epoch": 1993, "lr": 2.686637236124585e-05}
{"train_loss": 0.08756611496210098, "global_step": 177460, "epoch": 1993, "lr": 2.6865858409345434e-05}
{"train_loss": 0.08393990993499756, "global_step": 177461, "epoch": 1993, "lr": 2.686534446055511e-05}
{"train_loss": 0.07256898283958435, "global_step": 177462, "epoch": 1993, "lr": 2.6864830514874973e-05}
{"train_loss": 0.10129743069410324, "global_step": 177463, "epoch": 1993, "lr": 2.686431657230507e-05}
{"train_loss": 0.10466203093528748, "global_step": 177464, "epoch": 1993, "lr": 2.6863802632845465e-05}
{"train_loss": 0.0810945699975062, "global_step": 177465, "epoch": 1993, "lr": 2.6863288696496258e-05, "val_loss": 7.475539684295654}
{"train_loss": 0.07846736907958984, "global_step": 177466, "epoch": 1994, "lr": 2.686277476325748e-05}
{"train_loss": 0.14232930541038513, "global_step": 177467, "epoch": 1994, "lr": 2.686226083312924e-05}
{"train_loss": 0.08641239255666733, "global_step": 177468, "epoch": 1994, "lr": 2.6861746906111567e-05}
{"train_loss": 0.1305808275938034, "global_step": 177469, "epoch": 1994, "lr": 2.686123298220457e-05}
{"train_loss": 0.08137772232294083, "global_step": 177470, "epoch": 1994, "lr": 2.686071906140828e-05}
{"train_loss": 0.06699401140213013, "global_step": 177471, "epoch": 1994, "lr": 2.6860205143722807e-05}
{"train_loss": 0.15658703446388245, "global_step": 177472, "epoch": 1994, "lr": 2.685969122914817e-05}
{"train_loss": 0.03035578690469265, "global_step": 177473, "epoch": 1994, "lr": 2.6859177317684493e-05}
{"train_loss": 0.12617067992687225, "global_step": 177474, "epoch": 1994, "lr": 2.6858663409331785e-05}
{"train_loss": 0.018576718866825104, "global_step": 177475, "epoch": 1994, "lr": 2.685814950409018e-05}
{"train_loss": 0.06696345657110214, "global_step": 177476, "epoch": 1994, "lr": 2.685763560195969e-05}
{"train_loss": 0.07720887660980225, "global_step": 177477, "epoch": 1994, "lr": 2.685712170294043e-05}
{"train_loss": 0.07629645615816116, "global_step": 177478, "epoch": 1994, "lr": 2.6856607807032432e-05}
{"train_loss": 0.09930378198623657, "global_step": 177479, "epoch": 1994, "lr": 2.6856093914235792e-05}
{"train_loss": 0.0877191573381424, "global_step": 177480, "epoch": 1994, "lr": 2.685558002455055e-05}
{"train_loss": 0.07824364304542542, "global_step": 177481, "epoch": 1994, "lr": 2.6855066137976815e-05}
{"train_loss": 0.08482305705547333, "global_step": 177482, "epoch": 1994, "lr": 2.685455225451462e-05}
{"train_loss": 0.07985008507966995, "global_step": 177483, "epoch": 1994, "lr": 2.685403837416406e-05}
{"train_loss": 0.048247240483760834, "global_step": 177484, "epoch": 1994, "lr": 2.6853524496925176e-05}
{"train_loss": 0.07449449598789215, "global_step": 177485, "epoch": 1994, "lr": 2.6853010622798074e-05}
{"train_loss": 0.13244900107383728, "global_step": 177486, "epoch": 1994, "lr": 2.6852496751782784e-05}
{"train_loss": 0.08871140331029892, "global_step": 177487, "epoch": 1994, "lr": 2.685198288387939e-05}
{"train_loss": 0.07827598601579666, "global_step": 177488, "epoch": 1994, "lr": 2.6851469019087983e-05}
{"train_loss": 0.12042839825153351, "global_step": 177489, "epoch": 1994, "lr": 2.6850955157408596e-05}
{"train_loss": 0.06657657772302628, "global_step": 177490, "epoch": 1994, "lr": 2.6850441298841344e-05}
{"train_loss": 0.06430856883525848, "global_step": 177491, "epoch": 1994, "lr": 2.6849927443386236e-05}
{"train_loss": 0.04939837381243706, "global_step": 177492, "epoch": 1994, "lr": 2.6849413591043405e-05}
{"train_loss": 0.07194703817367554, "global_step": 177493, "epoch": 1994, "lr": 2.684889974181286e-05}
{"train_loss": 0.09115611761808395, "global_step": 177494, "epoch": 1994, "lr": 2.6848385895694717e-05}
{"train_loss": 0.08275457471609116, "global_step": 177495, "epoch": 1994, "lr": 2.6847872052689004e-05}
{"train_loss": 0.1913101077079773, "global_step": 177496, "epoch": 1994, "lr": 2.6847358212795844e-05}
{"train_loss": 0.04375506192445755, "global_step": 177497, "epoch": 1994, "lr": 2.6846844376015244e-05}
{"train_loss": 0.051403410732746124, "global_step": 177498, "epoch": 1994, "lr": 2.6846330542347332e-05}
{"train_loss": 0.12055710703134537, "global_step": 177499, "epoch": 1994, "lr": 2.684581671179212e-05}
{"train_loss": 0.07944239675998688, "global_step": 177500, "epoch": 1994, "lr": 2.6845302884349733e-05}
{"train_loss": 0.09480338543653488, "global_step": 177501, "epoch": 1994, "lr": 2.6844789060020203e-05}
{"train_loss": 0.14777982234954834, "global_step": 177502, "epoch": 1994, "lr": 2.68442752388036e-05}
{"train_loss": 0.06700872629880905, "global_step": 177503, "epoch": 1994, "lr": 2.684376142070002e-05}
{"train_loss": 0.06592939794063568, "global_step": 177504, "epoch": 1994, "lr": 2.684324760570949e-05}
{"train_loss": 0.09035798907279968, "global_step": 177505, "epoch": 1994, "lr": 2.684273379383212e-05}
{"train_loss": 0.06287512183189392, "global_step": 177506, "epoch": 1994, "lr": 2.684221998506795e-05}
{"train_loss": 0.09917909651994705, "global_step": 177507, "epoch": 1994, "lr": 2.684170617941708e-05}
{"train_loss": 0.06511667370796204, "global_step": 177508, "epoch": 1994, "lr": 2.6841192376879535e-05}
{"train_loss": 0.09431453049182892, "global_step": 177509, "epoch": 1994, "lr": 2.6840678577455437e-05}
{"train_loss": 0.13803699612617493, "global_step": 177510, "epoch": 1994, "lr": 2.6840164781144805e-05}
{"train_loss": 0.08159853518009186, "global_step": 177511, "epoch": 1994, "lr": 2.6839650987947747e-05}
{"train_loss": 0.05475691333413124, "global_step": 177512, "epoch": 1994, "lr": 2.6839137197864294e-05}
{"train_loss": 0.0470760352909565, "global_step": 177513, "epoch": 1994, "lr": 2.6838623410894558e-05}
{"train_loss": 0.12546803057193756, "global_step": 177514, "epoch": 1994, "lr": 2.6838109627038576e-05}
{"train_loss": 0.10711884498596191, "global_step": 177515, "epoch": 1994, "lr": 2.6837595846296437e-05}
{"train_loss": 0.1169060543179512, "global_step": 177516, "epoch": 1994, "lr": 2.6837082068668183e-05}
{"train_loss": 0.1255430281162262, "global_step": 177517, "epoch": 1994, "lr": 2.6836568294153928e-05}
{"train_loss": 0.07627682387828827, "global_step": 177518, "epoch": 1994, "lr": 2.6836054522753684e-05}
{"train_loss": 0.12034870684146881, "global_step": 177519, "epoch": 1994, "lr": 2.683554075446758e-05}
{"train_loss": 0.10151369869709015, "global_step": 177520, "epoch": 1994, "lr": 2.683502698929563e-05}
{"train_loss": 0.12447067350149155, "global_step": 177521, "epoch": 1994, "lr": 2.6834513227237944e-05}
{"train_loss": 0.07489225268363953, "global_step": 177522, "epoch": 1994, "lr": 2.6833999468294564e-05}
{"train_loss": 0.09835852682590485, "global_step": 177523, "epoch": 1994, "lr": 2.6833485712465567e-05}
{"train_loss": 0.1001853421330452, "global_step": 177524, "epoch": 1994, "lr": 2.683297195975104e-05}
{"train_loss": 0.09594175964593887, "global_step": 177525, "epoch": 1994, "lr": 2.6832458210151025e-05}
{"train_loss": 0.0644725114107132, "global_step": 177526, "epoch": 1994, "lr": 2.6831944463665622e-05}
{"train_loss": 0.10145384073257446, "global_step": 177527, "epoch": 1994, "lr": 2.683143072029486e-05}
{"train_loss": 0.05197211727499962, "global_step": 177528, "epoch": 1994, "lr": 2.683091698003885e-05}
{"train_loss": 0.0785459652543068, "global_step": 177529, "epoch": 1994, "lr": 2.6830403242897622e-05}
{"train_loss": 0.05137859284877777, "global_step": 177530, "epoch": 1994, "lr": 2.6829889508871288e-05}
{"train_loss": 0.13027462363243103, "global_step": 177531, "epoch": 1994, "lr": 2.6829375777959863e-05}
{"train_loss": 0.09261712431907654, "global_step": 177532, "epoch": 1994, "lr": 2.6828862050163472e-05}
{"train_loss": 0.09479735046625137, "global_step": 177533, "epoch": 1994, "lr": 2.6828348325482134e-05}
{"train_loss": 0.10856162011623383, "global_step": 177534, "epoch": 1994, "lr": 2.682783460391597e-05}
{"train_loss": 0.054272957146167755, "global_step": 177535, "epoch": 1994, "lr": 2.682732088546499e-05}
{"train_loss": 0.10698279738426208, "global_step": 177536, "epoch": 1994, "lr": 2.6826807170129327e-05}
{"train_loss": 0.05237719789147377, "global_step": 177537, "epoch": 1994, "lr": 2.682629345790899e-05}
{"train_loss": 0.1136377677321434, "global_step": 177538, "epoch": 1994, "lr": 2.682577974880409e-05}
{"train_loss": 0.06694069504737854, "global_step": 177539, "epoch": 1994, "lr": 2.6825266042814672e-05}
{"train_loss": 0.09351980686187744, "global_step": 177540, "epoch": 1994, "lr": 2.6824752339940827e-05}
{"train_loss": 0.17635278403759003, "global_step": 177541, "epoch": 1994, "lr": 2.682423864018262e-05}
{"train_loss": 0.0777817964553833, "global_step": 177542, "epoch": 1994, "lr": 2.6823724943540084e-05}
{"train_loss": 0.04672729969024658, "global_step": 177543, "epoch": 1994, "lr": 2.6823211250013343e-05}
{"train_loss": 0.10932531952857971, "global_step": 177544, "epoch": 1994, "lr": 2.6822697559602415e-05}
{"train_loss": 0.10486555844545364, "global_step": 177545, "epoch": 1994, "lr": 2.682218387230741e-05}
{"train_loss": 0.1269228160381317, "global_step": 177546, "epoch": 1994, "lr": 2.682167018812836e-05}
{"train_loss": 0.09115608781576157, "global_step": 177547, "epoch": 1994, "lr": 2.682115650706538e-05}
{"train_loss": 0.06642370671033859, "global_step": 177548, "epoch": 1994, "lr": 2.6820642829118485e-05}
{"train_loss": 0.08976192027330399, "global_step": 177549, "epoch": 1994, "lr": 2.6820129154287797e-05}
{"train_loss": 0.07272585481405258, "global_step": 177550, "epoch": 1994, "lr": 2.681961548257334e-05}
{"train_loss": 0.042322564870119095, "global_step": 177551, "epoch": 1994, "lr": 2.681910181397522e-05}
{"train_loss": 0.08211909234523773, "global_step": 177552, "epoch": 1994, "lr": 2.6818588148493473e-05}
{"train_loss": 0.08726099878549576, "global_step": 177553, "epoch": 1994, "lr": 2.6818074486128203e-05}
{"train_loss": 0.0886446645043874, "global_step": 177554, "epoch": 1994, "lr": 2.681756082687944e-05, "val_loss": 7.165249347686768}
{"train_loss": 0.035271309316158295, "global_step": 177555, "epoch": 1995, "lr": 2.681704717074729e-05}
{"train_loss": 0.07943575084209442, "global_step": 177556, "epoch": 1995, "lr": 2.681653351773179e-05}
{"train_loss": 0.1491609662771225, "global_step": 177557, "epoch": 1995, "lr": 2.681601986783304e-05}
{"train_loss": 0.08684872835874557, "global_step": 177558, "epoch": 1995, "lr": 2.681550622105108e-05}
{"train_loss": 0.06980906426906586, "global_step": 177559, "epoch": 1995, "lr": 2.681499257738599e-05}
{"train_loss": 0.08660992980003357, "global_step": 177560, "epoch": 1995, "lr": 2.6814478936837862e-05}
{"train_loss": 0.06264741718769073, "global_step": 177561, "epoch": 1995, "lr": 2.6813965299406728e-05}
{"train_loss": 0.07246754318475723, "global_step": 177562, "epoch": 1995, "lr": 2.681345166509269e-05}
{"train_loss": 0.06735391169786453, "global_step": 177563, "epoch": 1995, "lr": 2.6812938033895774e-05}
{"train_loss": 0.16266027092933655, "global_step": 177564, "epoch": 1995, "lr": 2.6812424405816104e-05}
{"train_loss": 0.05509650707244873, "global_step": 177565, "epoch": 1995, "lr": 2.68119107808537e-05}
{"train_loss": 0.051332809031009674, "global_step": 177566, "epoch": 1995, "lr": 2.681139715900867e-05}
{"train_loss": 0.11660242080688477, "global_step": 177567, "epoch": 1995, "lr": 2.681088354028105e-05}
{"train_loss": 0.08917608857154846, "global_step": 177568, "epoch": 1995, "lr": 2.6810369924670942e-05}
{"train_loss": 0.057429809123277664, "global_step": 177569, "epoch": 1995, "lr": 2.6809856312178372e-05}
{"train_loss": 0.1346624493598938, "global_step": 177570, "epoch": 1995, "lr": 2.680934270280346e-05}
{"train_loss": 0.10658231377601624, "global_step": 177571, "epoch": 1995, "lr": 2.6808829096546227e-05}
{"train_loss": 0.07064281404018402, "global_step": 177572, "epoch": 1995, "lr": 2.680831549340678e-05}
{"train_loss": 0.1437780112028122, "global_step": 177573, "epoch": 1995, "lr": 2.680780189338516e-05}
{"train_loss": 0.049662839621305466, "global_step": 177574, "epoch": 1995, "lr": 2.6807288296481464e-05}
{"train_loss": 0.04700852930545807, "global_step": 177575, "epoch": 1995, "lr": 2.6806774702695725e-05}
{"train_loss": 0.07360805571079254, "global_step": 177576, "epoch": 1995, "lr": 2.6806261112028054e-05}
{"train_loss": 0.05274452269077301, "global_step": 177577, "epoch": 1995, "lr": 2.6805747524478476e-05}
{"train_loss": 0.0629197359085083, "global_step": 177578, "epoch": 1995, "lr": 2.6805233940047108e-05}
{"train_loss": 0.05158848688006401, "global_step": 177579, "epoch": 1995, "lr": 2.6804720358733994e-05}
{"train_loss": 0.059715379029512405, "global_step": 177580, "epoch": 1995, "lr": 2.6804206780539177e-05}
{"train_loss": 0.05298342928290367, "global_step": 177581, "epoch": 1995, "lr": 2.6803693205462773e-05}
{"train_loss": 0.04998597875237465, "global_step": 177582, "epoch": 1995, "lr": 2.6803179633504805e-05}
{"train_loss": 0.0840013325214386, "global_step": 177583, "epoch": 1995, "lr": 2.6802666064665392e-05}
{"train_loss": 0.061847664415836334, "global_step": 177584, "epoch": 1995, "lr": 2.6802152498944562e-05}
{"train_loss": 0.09400034695863724, "global_step": 177585, "epoch": 1995, "lr": 2.6801638936342416e-05}
{"train_loss": 0.07072410732507706, "global_step": 177586, "epoch": 1995, "lr": 2.6801125376858987e-05}
{"train_loss": 0.0836912989616394, "global_step": 177587, "epoch": 1995, "lr": 2.680061182049438e-05}
{"train_loss": 0.05031179264187813, "global_step": 177588, "epoch": 1995, "lr": 2.6800098267248635e-05}
{"train_loss": 0.05977434292435646, "global_step": 177589, "epoch": 1995, "lr": 2.679958471712185e-05}
{"train_loss": 0.16730435192584991, "global_step": 177590, "epoch": 1995, "lr": 2.6799071170114058e-05}
{"train_loss": 0.06799256801605225, "global_step": 177591, "epoch": 1995, "lr": 2.6798557626225373e-05}
{"train_loss": 0.09294089674949646, "global_step": 177592, "epoch": 1995, "lr": 2.679804408545581e-05}
{"train_loss": 0.10001951456069946, "global_step": 177593, "epoch": 1995, "lr": 2.6797530547805495e-05}
{"train_loss": 0.059147439897060394, "global_step": 177594, "epoch": 1995, "lr": 2.6797017013274444e-05}
{"train_loss": 0.07768764346837997, "global_step": 177595, "epoch": 1995, "lr": 2.6796503481862758e-05}
{"train_loss": 0.07262497395277023, "global_step": 177596, "epoch": 1995, "lr": 2.679598995357051e-05}
{"train_loss": 0.059541165828704834, "global_step": 177597, "epoch": 1995, "lr": 2.6795476428397747e-05}
{"train_loss": 0.0886785089969635, "global_step": 177598, "epoch": 1995, "lr": 2.679496290634457e-05}
{"train_loss": 0.08953458815813065, "global_step": 177599, "epoch": 1995, "lr": 2.6794449387410996e-05}
{"train_loss": 0.08038549870252609, "global_step": 177600, "epoch": 1995, "lr": 2.6793935871597154e-05}
{"train_loss": 0.0804310292005539, "global_step": 177601, "epoch": 1995, "lr": 2.6793422358903066e-05}
{"train_loss": 0.10583023726940155, "global_step": 177602, "epoch": 1995, "lr": 2.6792908849328834e-05}
{"train_loss": 0.040187202394008636, "global_step": 177603, "epoch": 1995, "lr": 2.6792395342874493e-05}
{"train_loss": 0.061033256351947784, "global_step": 177604, "epoch": 1995, "lr": 2.6791881839540156e-05}
{"train_loss": 0.08377710729837418, "global_step": 177605, "epoch": 1995, "lr": 2.679136833932584e-05}
{"train_loss": 0.06752339005470276, "global_step": 177606, "epoch": 1995, "lr": 2.6790854842231672e-05}
{"train_loss": 0.08671528100967407, "global_step": 177607, "epoch": 1995, "lr": 2.6790341348257663e-05}
{"train_loss": 0.08816448599100113, "global_step": 177608, "epoch": 1995, "lr": 2.6789827857403933e-05}
{"train_loss": 0.13678795099258423, "global_step": 177609, "epoch": 1995, "lr": 2.6789314369670504e-05}
{"train_loss": 0.06634441763162613, "global_step": 177610, "epoch": 1995, "lr": 2.6788800885057496e-05}
{"train_loss": 0.04794630780816078, "global_step": 177611, "epoch": 1995, "lr": 2.6788287403564928e-05}
{"train_loss": 0.09379179775714874, "global_step": 177612, "epoch": 1995, "lr": 2.6787773925192906e-05}
{"train_loss": 0.10339242219924927, "global_step": 177613, "epoch": 1995, "lr": 2.678726044994147e-05}
{"train_loss": 0.05666821822524071, "global_step": 177614, "epoch": 1995, "lr": 2.6786746977810723e-05}
{"train_loss": 0.1046927198767662, "global_step": 177615, "epoch": 1995, "lr": 2.6786233508800695e-05}
{"train_loss": 0.0585615336894989, "global_step": 177616, "epoch": 1995, "lr": 2.67857200429115e-05}
{"train_loss": 0.07146000117063522, "global_step": 177617, "epoch": 1995, "lr": 2.678520658014315e-05}
{"train_loss": 0.08133971691131592, "global_step": 177618, "epoch": 1995, "lr": 2.6784693120495775e-05}
{"train_loss": 0.11218731105327606, "global_step": 177619, "epoch": 1995, "lr": 2.678417966396941e-05}
{"train_loss": 0.0873711034655571, "global_step": 177620, "epoch": 1995, "lr": 2.6783666210564116e-05}
{"train_loss": 0.07804334908723831, "global_step": 177621, "epoch": 1995, "lr": 2.6783152760279995e-05}
{"train_loss": 0.13096365332603455, "global_step": 177622, "epoch": 1995, "lr": 2.678263931311707e-05}
{"train_loss": 0.03231308236718178, "global_step": 177623, "epoch": 1995, "lr": 2.678212586907546e-05}
{"train_loss": 0.09178166836500168, "global_step": 177624, "epoch": 1995, "lr": 2.678161242815519e-05}
{"train_loss": 0.08487267047166824, "global_step": 177625, "epoch": 1995, "lr": 2.6781098990356368e-05}
{"train_loss": 0.1132165938615799, "global_step": 177626, "epoch": 1995, "lr": 2.6780585555679022e-05}
{"train_loss": 0.06757526099681854, "global_step": 177627, "epoch": 1995, "lr": 2.6780072124123268e-05}
{"train_loss": 0.09847847372293472, "global_step": 177628, "epoch": 1995, "lr": 2.677955869568912e-05}
{"train_loss": 0.08803839981555939, "global_step": 177629, "epoch": 1995, "lr": 2.6779045270376714e-05}
{"train_loss": 0.051716677844524384, "global_step": 177630, "epoch": 1995, "lr": 2.677853184818605e-05}
{"train_loss": 0.06590878218412399, "global_step": 177631, "epoch": 1995, "lr": 2.6778018429117234e-05}
{"train_loss": 0.12523071467876434, "global_step": 177632, "epoch": 1995, "lr": 2.6777505013170345e-05}
{"train_loss": 0.09063264727592468, "global_step": 177633, "epoch": 1995, "lr": 2.6776991600345423e-05}
{"train_loss": 0.0875130295753479, "global_step": 177634, "epoch": 1995, "lr": 2.6776478190642573e-05}
{"train_loss": 0.08529765158891678, "global_step": 177635, "epoch": 1995, "lr": 2.6775964784061818e-05}
{"train_loss": 0.09801341593265533, "global_step": 177636, "epoch": 1995, "lr": 2.6775451380603278e-05}
{"train_loss": 0.04032251983880997, "global_step": 177637, "epoch": 1995, "lr": 2.677493798026697e-05}
{"train_loss": 0.060644105076789856, "global_step": 177638, "epoch": 1995, "lr": 2.6774424583053003e-05}
{"train_loss": 0.05180593952536583, "global_step": 177639, "epoch": 1995, "lr": 2.677391118896142e-05}
{"train_loss": 0.15133017301559448, "global_step": 177640, "epoch": 1995, "lr": 2.6773397797992317e-05}
{"train_loss": 0.15208786725997925, "global_step": 177641, "epoch": 1995, "lr": 2.6772884410145732e-05}
{"train_loss": 0.09904403984546661, "global_step": 177642, "epoch": 1995, "lr": 2.677237102542177e-05}
{"train_loss": 0.0821512569202466, "global_step": 177643, "epoch": 1995, "lr": 2.6771857643820452e-05, "val_loss": 7.320967197418213, "train_action_mse_error": 18.959701538085938}
{"train_loss": 0.10251893103122711, "global_step": 177644, "epoch": 1996, "lr": 2.6771344265341903e-05}
{"train_loss": 0.0699099525809288, "global_step": 177645, "epoch": 1996, "lr": 2.6770830889986144e-05}
{"train_loss": 0.06294233351945877, "global_step": 177646, "epoch": 1996, "lr": 2.6770317517753273e-05}
{"train_loss": 0.08291422575712204, "global_step": 177647, "epoch": 1996, "lr": 2.6769804148643342e-05}
{"train_loss": 0.07008025050163269, "global_step": 177648, "epoch": 1996, "lr": 2.676929078265644e-05}
{"train_loss": 0.03906937316060066, "global_step": 177649, "epoch": 1996, "lr": 2.6768777419792606e-05}
{"train_loss": 0.04670245200395584, "global_step": 177650, "epoch": 1996, "lr": 2.676826406005193e-05}
{"train_loss": 0.07693765312433243, "global_step": 177651, "epoch": 1996, "lr": 2.676775070343449e-05}
{"train_loss": 0.059416599571704865, "global_step": 177652, "epoch": 1996, "lr": 2.676723734994035e-05}
{"train_loss": 0.16316184401512146, "global_step": 177653, "epoch": 1996, "lr": 2.676672399956955e-05}
{"train_loss": 0.11954144388437271, "global_step": 177654, "epoch": 1996, "lr": 2.67662106523222e-05}
{"train_loss": 0.14265118539333344, "global_step": 177655, "epoch": 1996, "lr": 2.6765697308198327e-05}
{"train_loss": 0.07527748495340347, "global_step": 177656, "epoch": 1996, "lr": 2.6765183967198048e-05}
{"train_loss": 0.07742907106876373, "global_step": 177657, "epoch": 1996, "lr": 2.6764670629321396e-05}
{"train_loss": 0.1176726371049881, "global_step": 177658, "epoch": 1996, "lr": 2.676415729456845e-05}
{"train_loss": 0.07968585938215256, "global_step": 177659, "epoch": 1996, "lr": 2.6763643962939282e-05}
{"train_loss": 0.07710534334182739, "global_step": 177660, "epoch": 1996, "lr": 2.6763130634433946e-05}
{"train_loss": 0.08450046926736832, "global_step": 177661, "epoch": 1996, "lr": 2.6762617309052544e-05}
{"train_loss": 0.09350494295358658, "global_step": 177662, "epoch": 1996, "lr": 2.6762103986795105e-05}
{"train_loss": 0.10529080033302307, "global_step": 177663, "epoch": 1996, "lr": 2.676159066766174e-05}
{"train_loss": 0.09807369112968445, "global_step": 177664, "epoch": 1996, "lr": 2.6761077351652475e-05}
{"train_loss": 0.05919944867491722, "global_step": 177665, "epoch": 1996, "lr": 2.676056403876741e-05}
{"train_loss": 0.05391346663236618, "global_step": 177666, "epoch": 1996, "lr": 2.67600507290066e-05}
{"train_loss": 0.07672179490327835, "global_step": 177667, "epoch": 1996, "lr": 2.6759537422370107e-05}
{"train_loss": 0.06538116186857224, "global_step": 177668, "epoch": 1996, "lr": 2.6759024118858033e-05}
{"train_loss": 0.07629352807998657, "global_step": 177669, "epoch": 1996, "lr": 2.6758510818470405e-05}
{"train_loss": 0.08952127397060394, "global_step": 177670, "epoch": 1996, "lr": 2.6757997521207333e-05}
{"train_loss": 0.06113772466778755, "global_step": 177671, "epoch": 1996, "lr": 2.6757484227068847e-05}
{"train_loss": 0.034599341452121735, "global_step": 177672, "epoch": 1996, "lr": 2.675697093605505e-05}
{"train_loss": 0.10630263388156891, "global_step": 177673, "epoch": 1996, "lr": 2.675645764816598e-05}
{"train_loss": 0.0718693807721138, "global_step": 177674, "epoch": 1996, "lr": 2.6755944363401735e-05}
{"train_loss": 0.07914453744888306, "global_step": 177675, "epoch": 1996, "lr": 2.675543108176235e-05}
{"train_loss": 0.06280189007520676, "global_step": 177676, "epoch": 1996, "lr": 2.6754917803247937e-05}
{"train_loss": 0.07330432534217834, "global_step": 177677, "epoch": 1996, "lr": 2.6754404527858522e-05}
{"train_loss": 0.050571758300065994, "global_step": 177678, "epoch": 1996, "lr": 2.6753891255594215e-05}
{"train_loss": 0.05649194121360779, "global_step": 177679, "epoch": 1996, "lr": 2.6753377986455042e-05}
{"train_loss": 0.03513091430068016, "global_step": 177680, "epoch": 1996, "lr": 2.6752864720441116e-05}
{"train_loss": 0.07826553285121918, "global_step": 177681, "epoch": 1996, "lr": 2.675235145755246e-05}
{"train_loss": 0.07909383624792099, "global_step": 177682, "epoch": 1996, "lr": 2.675183819778919e-05}
{"train_loss": 0.07341144233942032, "global_step": 177683, "epoch": 1996, "lr": 2.6751324941151333e-05}
{"train_loss": 0.05864909663796425, "global_step": 177684, "epoch": 1996, "lr": 2.6750811687639e-05}
{"train_loss": 0.07381884753704071, "global_step": 177685, "epoch": 1996, "lr": 2.675029843725221e-05}
{"train_loss": 0.08275613933801651, "global_step": 177686, "epoch": 1996, "lr": 2.674978518999106e-05}
{"train_loss": 0.10277664661407471, "global_step": 177687, "epoch": 1996, "lr": 2.6749271945855643e-05}
{"train_loss": 0.06590525805950165, "global_step": 177688, "epoch": 1996, "lr": 2.6748758704845977e-05}
{"train_loss": 0.07721533626317978, "global_step": 177689, "epoch": 1996, "lr": 2.6748245466962178e-05}
{"train_loss": 0.04056534171104431, "global_step": 177690, "epoch": 1996, "lr": 2.6747732232204296e-05}
{"train_loss": 0.057528164237737656, "global_step": 177691, "epoch": 1996, "lr": 2.6747219000572376e-05}
{"train_loss": 0.056822214275598526, "global_step": 177692, "epoch": 1996, "lr": 2.6746705772066533e-05}
{"train_loss": 0.059492576867341995, "global_step": 177693, "epoch": 1996, "lr": 2.6746192546686787e-05}
{"train_loss": 0.055883802473545074, "global_step": 177694, "epoch": 1996, "lr": 2.674567932443326e-05}
{"train_loss": 0.049706052988767624, "global_step": 177695, "epoch": 1996, "lr": 2.6745166105305985e-05}
{"train_loss": 0.06459436565637589, "global_step": 177696, "epoch": 1996, "lr": 2.674465288930502e-05}
{"train_loss": 0.03965752199292183, "global_step": 177697, "epoch": 1996, "lr": 2.674413967643048e-05}
{"train_loss": 0.05155003070831299, "global_step": 177698, "epoch": 1996, "lr": 2.674362646668238e-05}
{"train_loss": 0.07008346915245056, "global_step": 177699, "epoch": 1996, "lr": 2.6743113260060836e-05}
{"train_loss": 0.05521000549197197, "global_step": 177700, "epoch": 1996, "lr": 2.674260005656588e-05}
{"train_loss": 0.05130451172590256, "global_step": 177701, "epoch": 1996, "lr": 2.674208685619761e-05}
{"train_loss": 0.04897894710302353, "global_step": 177702, "epoch": 1996, "lr": 2.6741573658956076e-05}
{"train_loss": 0.054932378232479095, "global_step": 177703, "epoch": 1996, "lr": 2.6741060464841362e-05}
{"train_loss": 0.09695368260145187, "global_step": 177704, "epoch": 1996, "lr": 2.6740547273853515e-05}
{"train_loss": 0.05754629895091057, "global_step": 177705, "epoch": 1996, "lr": 2.6740034085992616e-05}
{"train_loss": 0.07452517002820969, "global_step": 177706, "epoch": 1996, "lr": 2.673952090125876e-05}
{"train_loss": 0.07194037735462189, "global_step": 177707, "epoch": 1996, "lr": 2.6739007719651964e-05}
{"train_loss": 0.10397308319807053, "global_step": 177708, "epoch": 1996, "lr": 2.6738494541172354e-05}
{"train_loss": 0.09275727719068527, "global_step": 177709, "epoch": 1996, "lr": 2.673798136581995e-05}
{"train_loss": 0.10177557170391083, "global_step": 177710, "epoch": 1996, "lr": 2.673746819359485e-05}
{"train_loss": 0.03436499461531639, "global_step": 177711, "epoch": 1996, "lr": 2.6736955024497103e-05}
{"train_loss": 0.053544096648693085, "global_step": 177712, "epoch": 1996, "lr": 2.6736441858526807e-05}
{"train_loss": 0.07642915099859238, "global_step": 177713, "epoch": 1996, "lr": 2.673592869568399e-05}
{"train_loss": 0.09279122203588486, "global_step": 177714, "epoch": 1996, "lr": 2.6735415535968767e-05}
{"train_loss": 0.0687766894698143, "global_step": 177715, "epoch": 1996, "lr": 2.6734902379381167e-05}
{"train_loss": 0.04589167237281799, "global_step": 177716, "epoch": 1996, "lr": 2.6734389225921284e-05}
{"train_loss": 0.10005105286836624, "global_step": 177717, "epoch": 1996, "lr": 2.6733876075589172e-05}
{"train_loss": 0.07095964252948761, "global_step": 177718, "epoch": 1996, "lr": 2.673336292838492e-05}
{"train_loss": 0.07369431853294373, "global_step": 177719, "epoch": 1996, "lr": 2.673284978430857e-05}
{"train_loss": 0.05505004897713661, "global_step": 177720, "epoch": 1996, "lr": 2.6732336643360213e-05}
{"train_loss": 0.07130763679742813, "global_step": 177721, "epoch": 1996, "lr": 2.67318235055399e-05}
{"train_loss": 0.030367206782102585, "global_step": 177722, "epoch": 1996, "lr": 2.6731310370847708e-05}
{"train_loss": 0.053504686802625656, "global_step": 177723, "epoch": 1996, "lr": 2.673079723928372e-05}
{"train_loss": 0.08109679818153381, "global_step": 177724, "epoch": 1996, "lr": 2.673028411084798e-05}
{"train_loss": 0.12754014134407043, "global_step": 177725, "epoch": 1996, "lr": 2.6729770985540592e-05}
{"train_loss": 0.03310009837150574, "global_step": 177726, "epoch": 1996, "lr": 2.6729257863361578e-05}
{"train_loss": 0.07807572931051254, "global_step": 177727, "epoch": 1996, "lr": 2.672874474431105e-05}
{"train_loss": 0.0910782739520073, "global_step": 177728, "epoch": 1996, "lr": 2.6728231628389065e-05}
{"train_loss": 0.08512943983078003, "global_step": 177729, "epoch": 1996, "lr": 2.6727718515595658e-05}
{"train_loss": 0.07087400555610657, "global_step": 177730, "epoch": 1996, "lr": 2.6727205405930956e-05}
{"train_loss": 0.07493838667869568, "global_step": 177731, "epoch": 1996, "lr": 2.6726692299394967e-05}
{"train_loss": 0.07299533917495374, "global_step": 177732, "epoch": 1996, "lr": 2.672617919598781e-05, "val_loss": 7.445485591888428}
{"train_loss": 0.06370153278112411, "global_step": 177733, "epoch": 1997, "lr": 2.6725666095709523e-05}
{"train_loss": 0.06388596445322037, "global_step": 177734, "epoch": 1997, "lr": 2.6725152998560197e-05}
{"train_loss": 0.09769877791404724, "global_step": 177735, "epoch": 1997, "lr": 2.6724639904539906e-05}
{"train_loss": 0.07406801730394363, "global_step": 177736, "epoch": 1997, "lr": 2.672412681364867e-05}
{"train_loss": 0.0642361268401146, "global_step": 177737, "epoch": 1997, "lr": 2.6723613725886614e-05}
{"train_loss": 0.05753916874527931, "global_step": 177738, "epoch": 1997, "lr": 2.672310064125377e-05}
{"train_loss": 0.05587898939847946, "global_step": 177739, "epoch": 1997, "lr": 2.6722587559750232e-05}
{"train_loss": 0.03987078741192818, "global_step": 177740, "epoch": 1997, "lr": 2.6722074481376046e-05}
{"train_loss": 0.06787736713886261, "global_step": 177741, "epoch": 1997, "lr": 2.672156140613129e-05}
{"train_loss": 0.10423959791660309, "global_step": 177742, "epoch": 1997, "lr": 2.6721048334016057e-05}
{"train_loss": 0.05749345198273659, "global_step": 177743, "epoch": 1997, "lr": 2.6720535265030377e-05}
{"train_loss": 0.0741264745593071, "global_step": 177744, "epoch": 1997, "lr": 2.6720022199174355e-05}
{"train_loss": 0.050530798733234406, "global_step": 177745, "epoch": 1997, "lr": 2.6719509136448024e-05}
{"train_loss": 0.07310579717159271, "global_step": 177746, "epoch": 1997, "lr": 2.6718996076851488e-05}
{"train_loss": 0.1017180010676384, "global_step": 177747, "epoch": 1997, "lr": 2.671848302038478e-05}
{"train_loss": 0.06924150884151459, "global_step": 177748, "epoch": 1997, "lr": 2.671796996704801e-05}
{"train_loss": 0.0887145847082138, "global_step": 177749, "epoch": 1997, "lr": 2.671745691684121e-05}
{"train_loss": 0.05201876908540726, "global_step": 177750, "epoch": 1997, "lr": 2.671694386976448e-05}
{"train_loss": 0.09967093914747238, "global_step": 177751, "epoch": 1997, "lr": 2.671643082581785e-05}
{"train_loss": 0.04814811423420906, "global_step": 177752, "epoch": 1997, "lr": 2.671591778500143e-05}
{"train_loss": 0.10212485492229462, "global_step": 177753, "epoch": 1997, "lr": 2.6715404747315254e-05}
{"train_loss": 0.09422747045755386, "global_step": 177754, "epoch": 1997, "lr": 2.6714891712759434e-05}
{"train_loss": 0.0556885190308094, "global_step": 177755, "epoch": 1997, "lr": 2.6714378681333983e-05}
{"train_loss": 0.13176126778125763, "global_step": 177756, "epoch": 1997, "lr": 2.6713865653039027e-05}
{"train_loss": 0.07927199453115463, "global_step": 177757, "epoch": 1997, "lr": 2.671335262787459e-05}
{"train_loss": 0.06596410274505615, "global_step": 177758, "epoch": 1997, "lr": 2.6712839605840755e-05}
{"train_loss": 0.08117891103029251, "global_step": 177759, "epoch": 1997, "lr": 2.6712326586937613e-05}
{"train_loss": 0.05608735606074333, "global_step": 177760, "epoch": 1997, "lr": 2.6711813571165202e-05}
{"train_loss": 0.08307655155658722, "global_step": 177761, "epoch": 1997, "lr": 2.671130055852361e-05}
{"train_loss": 0.08096299320459366, "global_step": 177762, "epoch": 1997, "lr": 2.6710787549012895e-05}
{"train_loss": 0.10303772240877151, "global_step": 177763, "epoch": 1997, "lr": 2.671027454263314e-05}
{"train_loss": 0.06376916170120239, "global_step": 177764, "epoch": 1997, "lr": 2.6709761539384394e-05}
{"train_loss": 0.038357656449079514, "global_step": 177765, "epoch": 1997, "lr": 2.670924853926675e-05}
{"train_loss": 0.04906582459807396, "global_step": 177766, "epoch": 1997, "lr": 2.6708735542280244e-05}
{"train_loss": 0.10050202161073685, "global_step": 177767, "epoch": 1997, "lr": 2.6708222548424986e-05}
{"train_loss": 0.05819088965654373, "global_step": 177768, "epoch": 1997, "lr": 2.670770955770102e-05}
{"train_loss": 0.12648767232894897, "global_step": 177769, "epoch": 1997, "lr": 2.6707196570108407e-05}
{"train_loss": 0.05386006087064743, "global_step": 177770, "epoch": 1997, "lr": 2.670668358564724e-05}
{"train_loss": 0.11771872639656067, "global_step": 177771, "epoch": 1997, "lr": 2.670617060431756e-05}
{"train_loss": 0.039445821195840836, "global_step": 177772, "epoch": 1997, "lr": 2.6705657626119467e-05}
{"train_loss": 0.056674323976039886, "global_step": 177773, "epoch": 1997, "lr": 2.6705144651053016e-05}
{"train_loss": 0.08343890309333801, "global_step": 177774, "epoch": 1997, "lr": 2.6704631679118254e-05}
{"train_loss": 0.08535007387399673, "global_step": 177775, "epoch": 1997, "lr": 2.6704118710315297e-05}
{"train_loss": 0.1628209501504898, "global_step": 177776, "epoch": 1997, "lr": 2.6703605744644157e-05}
{"train_loss": 0.13674411177635193, "global_step": 177777, "epoch": 1997, "lr": 2.6703092782104944e-05}
{"train_loss": 0.04749304801225662, "global_step": 177778, "epoch": 1997, "lr": 2.6702579822697728e-05}
{"train_loss": 0.09534665197134018, "global_step": 177779, "epoch": 1997, "lr": 2.6702066866422548e-05}
{"train_loss": 0.06877122819423676, "global_step": 177780, "epoch": 1997, "lr": 2.670155391327951e-05}
{"train_loss": 0.04339221492409706, "global_step": 177781, "epoch": 1997, "lr": 2.6701040963268646e-05}
{"train_loss": 0.037068139761686325, "global_step": 177782, "epoch": 1997, "lr": 2.6700528016390063e-05}
{"train_loss": 0.05988466739654541, "global_step": 177783, "epoch": 1997, "lr": 2.670001507264379e-05}
{"train_loss": 0.1322912871837616, "global_step": 177784, "epoch": 1997, "lr": 2.6699502132029942e-05}
{"train_loss": 0.07044871151447296, "global_step": 177785, "epoch": 1997, "lr": 2.669898919454853e-05}
{"train_loss": 0.0677940770983696, "global_step": 177786, "epoch": 1997, "lr": 2.6698476260199678e-05}
{"train_loss": 0.03417091444134712, "global_step": 177787, "epoch": 1997, "lr": 2.6697963328983422e-05}
{"train_loss": 0.059646669775247574, "global_step": 177788, "epoch": 1997, "lr": 2.669745040089985e-05}
{"train_loss": 0.07748278975486755, "global_step": 177789, "epoch": 1997, "lr": 2.6696937475949002e-05}
{"train_loss": 0.04233497008681297, "global_step": 177790, "epoch": 1997, "lr": 2.6696424554130993e-05}
{"train_loss": 0.049750443547964096, "global_step": 177791, "epoch": 1997, "lr": 2.6695911635445848e-05}
{"train_loss": 0.05627379193902016, "global_step": 177792, "epoch": 1997, "lr": 2.669539871989366e-05}
{"train_loss": 0.0712079405784607, "global_step": 177793, "epoch": 1997, "lr": 2.6694885807474486e-05}
{"train_loss": 0.10975488275289536, "global_step": 177794, "epoch": 1997, "lr": 2.669437289818839e-05}
{"train_loss": 0.07426857948303223, "global_step": 177795, "epoch": 1997, "lr": 2.6693859992035474e-05}
{"train_loss": 0.05150645226240158, "global_step": 177796, "epoch": 1997, "lr": 2.6693347089015776e-05}
{"train_loss": 0.13798730075359344, "global_step": 177797, "epoch": 1997, "lr": 2.6692834189129378e-05}
{"train_loss": 0.06463231891393661, "global_step": 177798, "epoch": 1997, "lr": 2.669232129237633e-05}
{"train_loss": 0.044987235218286514, "global_step": 177799, "epoch": 1997, "lr": 2.6691808398756734e-05}
{"train_loss": 0.10628876090049744, "global_step": 177800, "epoch": 1997, "lr": 2.6691295508270624e-05}
{"train_loss": 0.09412004798650742, "global_step": 177801, "epoch": 1997, "lr": 2.6690782620918104e-05}
{"train_loss": 0.04251468926668167, "global_step": 177802, "epoch": 1997, "lr": 2.6690269736699204e-05}
{"train_loss": 0.04887460544705391, "global_step": 177803, "epoch": 1997, "lr": 2.6689756855614034e-05}
{"train_loss": 0.03845689445734024, "global_step": 177804, "epoch": 1997, "lr": 2.6689243977662615e-05}
{"train_loss": 0.05392519757151604, "global_step": 177805, "epoch": 1997, "lr": 2.6688731102845072e-05}
{"train_loss": 0.09704933315515518, "global_step": 177806, "epoch": 1997, "lr": 2.6688218231161443e-05}
{"train_loss": 0.10070245712995529, "global_step": 177807, "epoch": 1997, "lr": 2.668770536261177e-05}
{"train_loss": 0.10569300502538681, "global_step": 177808, "epoch": 1997, "lr": 2.668719249719618e-05}
{"train_loss": 0.09078918397426605, "global_step": 177809, "epoch": 1997, "lr": 2.668667963491469e-05}
{"train_loss": 0.12378031015396118, "global_step": 177810, "epoch": 1997, "lr": 2.668616677576742e-05}
{"train_loss": 0.0729023739695549, "global_step": 177811, "epoch": 1997, "lr": 2.66856539197544e-05}
{"train_loss": 0.09643366932868958, "global_step": 177812, "epoch": 1997, "lr": 2.6685141066875696e-05}
{"train_loss": 0.09039350599050522, "global_step": 177813, "epoch": 1997, "lr": 2.6684628217131385e-05}
{"train_loss": 0.10383550822734833, "global_step": 177814, "epoch": 1997, "lr": 2.6684115370521568e-05}
{"train_loss": 0.04338647425174713, "global_step": 177815, "epoch": 1997, "lr": 2.668360252704626e-05}
{"train_loss": 0.10914258658885956, "global_step": 177816, "epoch": 1997, "lr": 2.668308968670558e-05}
{"train_loss": 0.07908754795789719, "global_step": 177817, "epoch": 1997, "lr": 2.6682576849499557e-05}
{"train_loss": 0.10268077254295349, "global_step": 177818, "epoch": 1997, "lr": 2.66820640154283e-05}
{"train_loss": 0.08347262442111969, "global_step": 177819, "epoch": 1997, "lr": 2.6681551184491828e-05}
{"train_loss": 0.051689472049474716, "global_step": 177820, "epoch": 1997, "lr": 2.6681038356690258e-05}
{"train_loss": 0.07646392846710226, "global_step": 177821, "epoch": 1997, "lr": 2.6680525532023625e-05, "val_loss": 7.492178916931152}
{"train_loss": 0.08254381269216537, "global_step": 177822, "epoch": 1998, "lr": 2.6680012710492024e-05}
{"train_loss": 0.07401401549577713, "global_step": 177823, "epoch": 1998, "lr": 2.6679499892095495e-05}
{"train_loss": 0.05195815488696098, "global_step": 177824, "epoch": 1998, "lr": 2.667898707683414e-05}
{"train_loss": 0.05808213725686073, "global_step": 177825, "epoch": 1998, "lr": 2.6678474264708e-05}
{"train_loss": 0.08890783786773682, "global_step": 177826, "epoch": 1998, "lr": 2.667796145571716e-05}
{"train_loss": 0.06953279674053192, "global_step": 177827, "epoch": 1998, "lr": 2.6677448649861676e-05}
{"train_loss": 0.08648459613323212, "global_step": 177828, "epoch": 1998, "lr": 2.6676935847141636e-05}
{"train_loss": 0.1135934591293335, "global_step": 177829, "epoch": 1998, "lr": 2.6676423047557085e-05}
{"train_loss": 0.12663014233112335, "global_step": 177830, "epoch": 1998, "lr": 2.6675910251108122e-05}
{"train_loss": 0.04552565887570381, "global_step": 177831, "epoch": 1998, "lr": 2.6675397457794783e-05}
{"train_loss": 0.029575016349554062, "global_step": 177832, "epoch": 1998, "lr": 2.6674884667617146e-05}
{"train_loss": 0.0790817067027092, "global_step": 177833, "epoch": 1998, "lr": 2.6674371880575304e-05}
{"train_loss": 0.054191455245018005, "global_step": 177834, "epoch": 1998, "lr": 2.6673859096669295e-05}
{"train_loss": 0.10940036922693253, "global_step": 177835, "epoch": 1998, "lr": 2.6673346315899217e-05}
{"train_loss": 0.07196109741926193, "global_step": 177836, "epoch": 1998, "lr": 2.6672833538265108e-05}
{"train_loss": 0.06974238902330399, "global_step": 177837, "epoch": 1998, "lr": 2.6672320763767068e-05}
{"train_loss": 0.035429697483778, "global_step": 177838, "epoch": 1998, "lr": 2.667180799240513e-05}
{"train_loss": 0.06305396556854248, "global_step": 177839, "epoch": 1998, "lr": 2.6671295224179405e-05}
{"train_loss": 0.06021132692694664, "global_step": 177840, "epoch": 1998, "lr": 2.6670782459089922e-05}
{"train_loss": 0.06245029345154762, "global_step": 177841, "epoch": 1998, "lr": 2.6670269697136786e-05}
{"train_loss": 0.04541334509849548, "global_step": 177842, "epoch": 1998, "lr": 2.666975693832003e-05}
{"train_loss": 0.07671009004116058, "global_step": 177843, "epoch": 1998, "lr": 2.6669244182639754e-05}
{"train_loss": 0.07404007017612457, "global_step": 177844, "epoch": 1998, "lr": 2.6668731430096018e-05}
{"train_loss": 0.03778674826025963, "global_step": 177845, "epoch": 1998, "lr": 2.6668218680688872e-05}
{"train_loss": 0.07471508532762527, "global_step": 177846, "epoch": 1998, "lr": 2.6667705934418408e-05}
{"train_loss": 0.08294536918401718, "global_step": 177847, "epoch": 1998, "lr": 2.666719319128468e-05}
{"train_loss": 0.10103774070739746, "global_step": 177848, "epoch": 1998, "lr": 2.6666680451287772e-05}
{"train_loss": 0.08114156126976013, "global_step": 177849, "epoch": 1998, "lr": 2.6666167714427732e-05}
{"train_loss": 0.09907272458076477, "global_step": 177850, "epoch": 1998, "lr": 2.666565498070466e-05}
{"train_loss": 0.12682604789733887, "global_step": 177851, "epoch": 1998, "lr": 2.6665142250118578e-05}
{"train_loss": 0.11778595298528671, "global_step": 177852, "epoch": 1998, "lr": 2.6664629522669617e-05}
{"train_loss": 0.0637030303478241, "global_step": 177853, "epoch": 1998, "lr": 2.6664116798357787e-05}
{"train_loss": 0.10434239357709885, "global_step": 177854, "epoch": 1998, "lr": 2.666360407718319e-05}
{"train_loss": 0.06617729365825653, "global_step": 177855, "epoch": 1998, "lr": 2.666309135914588e-05}
{"train_loss": 0.03333905711770058, "global_step": 177856, "epoch": 1998, "lr": 2.666257864424595e-05}
{"train_loss": 0.09771350771188736, "global_step": 177857, "epoch": 1998, "lr": 2.6662065932483438e-05}
{"train_loss": 0.1728796511888504, "global_step": 177858, "epoch": 1998, "lr": 2.666155322385844e-05}
{"train_loss": 0.06931378692388535, "global_step": 177859, "epoch": 1998, "lr": 2.6661040518370988e-05}
{"train_loss": 0.07470410317182541, "global_step": 177860, "epoch": 1998, "lr": 2.6660527816021198e-05}
{"train_loss": 0.10864998400211334, "global_step": 177861, "epoch": 1998, "lr": 2.66600151168091e-05}
{"train_loss": 0.12095905840396881, "global_step": 177862, "epoch": 1998, "lr": 2.6659502420734794e-05}
{"train_loss": 0.07288358360528946, "global_step": 177863, "epoch": 1998, "lr": 2.665898972779831e-05}
{"train_loss": 0.04710555449128151, "global_step": 177864, "epoch": 1998, "lr": 2.6658477037999768e-05}
{"train_loss": 0.049784667789936066, "global_step": 177865, "epoch": 1998, "lr": 2.6657964351339186e-05}
{"train_loss": 0.03812369704246521, "global_step": 177866, "epoch": 1998, "lr": 2.665745166781668e-05}
{"train_loss": 0.07121188193559647, "global_step": 177867, "epoch": 1998, "lr": 2.6656938987432266e-05}
{"train_loss": 0.06939885765314102, "global_step": 177868, "epoch": 1998, "lr": 2.665642631018605e-05}
{"train_loss": 0.047336045652627945, "global_step": 177869, "epoch": 1998, "lr": 2.665591363607811e-05}
{"train_loss": 0.09046655893325806, "global_step": 177870, "epoch": 1998, "lr": 2.665540096510848e-05}
{"train_loss": 0.04997442290186882, "global_step": 177871, "epoch": 1998, "lr": 2.6654888297277263e-05}
{"train_loss": 0.08928891271352768, "global_step": 177872, "epoch": 1998, "lr": 2.6654375632584495e-05}
{"train_loss": 0.13179367780685425, "global_step": 177873, "epoch": 1998, "lr": 2.665386297103028e-05}
{"train_loss": 0.057124994695186615, "global_step": 177874, "epoch": 1998, "lr": 2.665335031261465e-05}
{"train_loss": 0.05692118778824806, "global_step": 177875, "epoch": 1998, "lr": 2.6652837657337714e-05}
{"train_loss": 0.06827939301729202, "global_step": 177876, "epoch": 1998, "lr": 2.6652325005199495e-05}
{"train_loss": 0.15196433663368225, "global_step": 177877, "epoch": 1998, "lr": 2.6651812356200106e-05}
{"train_loss": 0.0644216388463974, "global_step": 177878, "epoch": 1998, "lr": 2.6651299710339584e-05}
{"train_loss": 0.044695623219013214, "global_step": 177879, "epoch": 1998, "lr": 2.665078706761802e-05}
{"train_loss": 0.0777697041630745, "global_step": 177880, "epoch": 1998, "lr": 2.665027442803546e-05}
{"train_loss": 0.09755287319421768, "global_step": 177881, "epoch": 1998, "lr": 2.6649761791591998e-05}
{"train_loss": 0.09136457741260529, "global_step": 177882, "epoch": 1998, "lr": 2.6649249158287682e-05}
{"train_loss": 0.05225493013858795, "global_step": 177883, "epoch": 1998, "lr": 2.6648736528122603e-05}
{"train_loss": 0.0853823646903038, "global_step": 177884, "epoch": 1998, "lr": 2.664822390109682e-05}
{"train_loss": 0.09482239931821823, "global_step": 177885, "epoch": 1998, "lr": 2.6647711277210376e-05}
{"train_loss": 0.07855043560266495, "global_step": 177886, "epoch": 1998, "lr": 2.6647198656463384e-05}
{"train_loss": 0.08804120123386383, "global_step": 177887, "epoch": 1998, "lr": 2.6646686038855873e-05}
{"train_loss": 0.14228500425815582, "global_step": 177888, "epoch": 1998, "lr": 2.6646173424387944e-05}
{"train_loss": 0.11190866678953171, "global_step": 177889, "epoch": 1998, "lr": 2.6645660813059637e-05}
{"train_loss": 0.14360667765140533, "global_step": 177890, "epoch": 1998, "lr": 2.664514820487106e-05}
{"train_loss": 0.08397708088159561, "global_step": 177891, "epoch": 1998, "lr": 2.6644635599822233e-05}
{"train_loss": 0.02607746794819832, "global_step": 177892, "epoch": 1998, "lr": 2.6644122997913267e-05}
{"train_loss": 0.0951540470123291, "global_step": 177893, "epoch": 1998, "lr": 2.6643610399144203e-05}
{"train_loss": 0.08436881005764008, "global_step": 177894, "epoch": 1998, "lr": 2.6643097803515136e-05}
{"train_loss": 0.07656624913215637, "global_step": 177895, "epoch": 1998, "lr": 2.66425852110261e-05}
{"train_loss": 0.092320017516613, "global_step": 177896, "epoch": 1998, "lr": 2.6642072621677206e-05}
{"train_loss": 0.07001356780529022, "global_step": 177897, "epoch": 1998, "lr": 2.6641560035468478e-05}
{"train_loss": 0.027262702584266663, "global_step": 177898, "epoch": 1998, "lr": 2.6641047452400026e-05}
{"train_loss": 0.058717742562294006, "global_step": 177899, "epoch": 1998, "lr": 2.6640534872471884e-05}
{"train_loss": 0.07484780251979828, "global_step": 177900, "epoch": 1998, "lr": 2.6640022295684154e-05}
{"train_loss": 0.08409763127565384, "global_step": 177901, "epoch": 1998, "lr": 2.663950972203687e-05}
{"train_loss": 0.08992837369441986, "global_step": 177902, "epoch": 1998, "lr": 2.6638997151530144e-05}
{"train_loss": 0.048710569739341736, "global_step": 177903, "epoch": 1998, "lr": 2.6638484584163992e-05}
{"train_loss": 0.05319473519921303, "global_step": 177904, "epoch": 1998, "lr": 2.6637972019938517e-05}
{"train_loss": 0.08791796863079071, "global_step": 177905, "epoch": 1998, "lr": 2.66374594588538e-05}
{"train_loss": 0.11907491087913513, "global_step": 177906, "epoch": 1998, "lr": 2.6636946900909872e-05}
{"train_loss": 0.09309802949428558, "global_step": 177907, "epoch": 1998, "lr": 2.6636434346106838e-05}
{"train_loss": 0.21144968271255493, "global_step": 177908, "epoch": 1998, "lr": 2.6635921794444735e-05}
{"train_loss": 0.0776115357875824, "global_step": 177909, "epoch": 1998, "lr": 2.6635409245923664e-05}
{"train_loss": 0.07979485055620081, "global_step": 177910, "epoch": 1998, "lr": 2.663489670054366e-05, "val_loss": 7.642165660858154}
{"train_loss": 0.06222149729728699, "global_step": 177911, "epoch": 1999, "lr": 2.6634384158304827e-05}
{"train_loss": 0.12536557018756866, "global_step": 177912, "epoch": 1999, "lr": 2.66338716192072e-05}
{"train_loss": 0.09106209874153137, "global_step": 177913, "epoch": 1999, "lr": 2.6633359083250887e-05}
{"train_loss": 0.03891716152429581, "global_step": 177914, "epoch": 1999, "lr": 2.6632846550435907e-05}
{"train_loss": 0.10348504781723022, "global_step": 177915, "epoch": 1999, "lr": 2.663233402076238e-05}
{"train_loss": 0.11447018384933472, "global_step": 177916, "epoch": 1999, "lr": 2.6631821494230323e-05}
{"train_loss": 0.06077185273170471, "global_step": 177917, "epoch": 1999, "lr": 2.6631308970839862e-05}
{"train_loss": 0.08117129653692245, "global_step": 177918, "epoch": 1999, "lr": 2.6630796450591013e-05}
{"train_loss": 0.09520462900400162, "global_step": 177919, "epoch": 1999, "lr": 2.663028393348389e-05}
{"train_loss": 0.10797014832496643, "global_step": 177920, "epoch": 1999, "lr": 2.6629771419518523e-05}
{"train_loss": 0.04365279898047447, "global_step": 177921, "epoch": 1999, "lr": 2.662925890869501e-05}
{"train_loss": 0.04241928085684776, "global_step": 177922, "epoch": 1999, "lr": 2.6628746401013414e-05}
{"train_loss": 0.058476001024246216, "global_step": 177923, "epoch": 1999, "lr": 2.662823389647378e-05}
{"train_loss": 0.044055692851543427, "global_step": 177924, "epoch": 1999, "lr": 2.662772139507621e-05}
{"train_loss": 0.057100068777799606, "global_step": 177925, "epoch": 1999, "lr": 2.6627208896820744e-05}
{"train_loss": 0.09694864600896835, "global_step": 177926, "epoch": 1999, "lr": 2.6626696401707485e-05}
{"train_loss": 0.13501977920532227, "global_step": 177927, "epoch": 1999, "lr": 2.6626183909736458e-05}
{"train_loss": 0.13191424310207367, "global_step": 177928, "epoch": 1999, "lr": 2.662567142090777e-05}
{"train_loss": 0.06508288532495499, "global_step": 177929, "epoch": 1999, "lr": 2.6625158935221463e-05}
{"train_loss": 0.09925667941570282, "global_step": 177930, "epoch": 1999, "lr": 2.6624646452677638e-05}
{"train_loss": 0.043025292456150055, "global_step": 177931, "epoch": 1999, "lr": 2.6624133973276322e-05}
{"train_loss": 0.03555234894156456, "global_step": 177932, "epoch": 1999, "lr": 2.662362149701763e-05}
{"train_loss": 0.0728512853384018, "global_step": 177933, "epoch": 1999, "lr": 2.6623109023901583e-05}
{"train_loss": 0.05935436859726906, "global_step": 177934, "epoch": 1999, "lr": 2.6622596553928303e-05}
{"train_loss": 0.09472440928220749, "global_step": 177935, "epoch": 1999, "lr": 2.66220840870978e-05}
{"train_loss": 0.10044442862272263, "global_step": 177936, "epoch": 1999, "lr": 2.6621571623410198e-05}
{"train_loss": 0.060956284403800964, "global_step": 177937, "epoch": 1999, "lr": 2.6621059162865513e-05}
{"train_loss": 0.03548017516732216, "global_step": 177938, "epoch": 1999, "lr": 2.6620546705463868e-05}
{"train_loss": 0.08754245191812515, "global_step": 177939, "epoch": 1999, "lr": 2.6620034251205285e-05}
{"train_loss": 0.11726804077625275, "global_step": 177940, "epoch": 1999, "lr": 2.6619521800089852e-05}
{"train_loss": 0.06797967106103897, "global_step": 177941, "epoch": 1999, "lr": 2.6619009352117663e-05}
{"train_loss": 0.10817760974168777, "global_step": 177942, "epoch": 1999, "lr": 2.6618496907288742e-05}
{"train_loss": 0.041860442608594894, "global_step": 177943, "epoch": 1999, "lr": 2.6617984465603197e-05}
{"train_loss": 0.09493539482355118, "global_step": 177944, "epoch": 1999, "lr": 2.6617472027061063e-05}
{"train_loss": 0.07955502718687057, "global_step": 177945, "epoch": 1999, "lr": 2.6616959591662438e-05}
{"train_loss": 0.17333261668682098, "global_step": 177946, "epoch": 1999, "lr": 2.6616447159407366e-05}
{"train_loss": 0.1024325042963028, "global_step": 177947, "epoch": 1999, "lr": 2.6615934730295944e-05}
{"train_loss": 0.10472767055034637, "global_step": 177948, "epoch": 1999, "lr": 2.6615422304328204e-05}
{"train_loss": 0.047923021018505096, "global_step": 177949, "epoch": 1999, "lr": 2.661490988150425e-05}
{"train_loss": 0.07698599994182587, "global_step": 177950, "epoch": 1999, "lr": 2.6614397461824124e-05}
{"train_loss": 0.08347269892692566, "global_step": 177951, "epoch": 1999, "lr": 2.6613885045287923e-05}
{"train_loss": 0.07002543658018112, "global_step": 177952, "epoch": 1999, "lr": 2.6613372631895684e-05}
{"train_loss": 0.059997472912073135, "global_step": 177953, "epoch": 1999, "lr": 2.6612860221647514e-05}
{"train_loss": 0.039803650230169296, "global_step": 177954, "epoch": 1999, "lr": 2.6612347814543438e-05}
{"train_loss": 0.061085984110832214, "global_step": 177955, "epoch": 1999, "lr": 2.6611835410583562e-05}
{"train_loss": 0.13706006109714508, "global_step": 177956, "epoch": 1999, "lr": 2.661132300976792e-05}
{"train_loss": 0.08198194205760956, "global_step": 177957, "epoch": 1999, "lr": 2.6610810612096626e-05}
{"train_loss": 0.09910576790571213, "global_step": 177958, "epoch": 1999, "lr": 2.6610298217569695e-05}
{"train_loss": 0.06407809257507324, "global_step": 177959, "epoch": 1999, "lr": 2.660978582618725e-05}
{"train_loss": 0.07903306186199188, "global_step": 177960, "epoch": 1999, "lr": 2.6609273437949318e-05}
{"train_loss": 0.07666902989149094, "global_step": 177961, "epoch": 1999, "lr": 2.6608761052856003e-05}
{"train_loss": 0.08210105448961258, "global_step": 177962, "epoch": 1999, "lr": 2.6608248670907342e-05}
{"train_loss": 0.10803697258234024, "global_step": 177963, "epoch": 1999, "lr": 2.6607736292103413e-05}
{"train_loss": 0.04019923135638237, "global_step": 177964, "epoch": 1999, "lr": 2.66072239164443e-05}
{"train_loss": 0.13418106734752655, "global_step": 177965, "epoch": 1999, "lr": 2.660671154393004e-05}
{"train_loss": 0.06506612151861191, "global_step": 177966, "epoch": 1999, "lr": 2.6606199174560755e-05}
{"train_loss": 0.04943789169192314, "global_step": 177967, "epoch": 1999, "lr": 2.660568680833645e-05}
{"train_loss": 0.10908782482147217, "global_step": 177968, "epoch": 1999, "lr": 2.660517444525724e-05}
{"train_loss": 0.06633780896663666, "global_step": 177969, "epoch": 1999, "lr": 2.6604662085323173e-05}
{"train_loss": 0.08086366206407547, "global_step": 177970, "epoch": 1999, "lr": 2.6604149728534335e-05}
{"train_loss": 0.05126093327999115, "global_step": 177971, "epoch": 1999, "lr": 2.6603637374890767e-05}
{"train_loss": 0.050789669156074524, "global_step": 177972, "epoch": 1999, "lr": 2.660312502439257e-05}
{"train_loss": 0.08678241819143295, "global_step": 177973, "epoch": 1999, "lr": 2.6602612677039785e-05}
{"train_loss": 0.05409763380885124, "global_step": 177974, "epoch": 1999, "lr": 2.6602100332832514e-05}
{"train_loss": 0.09874241799116135, "global_step": 177975, "epoch": 1999, "lr": 2.6601587991770784e-05}
{"train_loss": 0.08327201008796692, "global_step": 177976, "epoch": 1999, "lr": 2.660107565385468e-05}
{"train_loss": 0.06433100998401642, "global_step": 177977, "epoch": 1999, "lr": 2.6600563319084297e-05}
{"train_loss": 0.07708768546581268, "global_step": 177978, "epoch": 1999, "lr": 2.660005098745967e-05}
{"train_loss": 0.05705808475613594, "global_step": 177979, "epoch": 1999, "lr": 2.6599538658980895e-05}
{"train_loss": 0.05200866237282753, "global_step": 177980, "epoch": 1999, "lr": 2.6599026333648008e-05}
{"train_loss": 0.08125977963209152, "global_step": 177981, "epoch": 1999, "lr": 2.6598514011461116e-05}
{"train_loss": 0.10322488844394684, "global_step": 177982, "epoch": 1999, "lr": 2.659800169242025e-05}
{"train_loss": 0.06926438957452774, "global_step": 177983, "epoch": 1999, "lr": 2.659748937652552e-05}
{"train_loss": 0.06932968646287918, "global_step": 177984, "epoch": 1999, "lr": 2.6596977063776946e-05}
{"train_loss": 0.12986932694911957, "global_step": 177985, "epoch": 1999, "lr": 2.6596464754174643e-05}
{"train_loss": 0.09634444117546082, "global_step": 177986, "epoch": 1999, "lr": 2.6595952447718647e-05}
{"train_loss": 0.08374810218811035, "global_step": 177987, "epoch": 1999, "lr": 2.6595440144409057e-05}
{"train_loss": 0.12150617688894272, "global_step": 177988, "epoch": 1999, "lr": 2.6594927844245903e-05}
{"train_loss": 0.05726176127791405, "global_step": 177989, "epoch": 1999, "lr": 2.65944155472293e-05}
{"train_loss": 0.05595950037240982, "global_step": 177990, "epoch": 1999, "lr": 2.6593903253359265e-05}
{"train_loss": 0.027499299496412277, "global_step": 177991, "epoch": 1999, "lr": 2.6593390962635924e-05}
{"train_loss": 0.049391716718673706, "global_step": 177992, "epoch": 1999, "lr": 2.659287867505929e-05}
{"train_loss": 0.048285819590091705, "global_step": 177993, "epoch": 1999, "lr": 2.6592366390629474e-05}
{"train_loss": 0.11870162934064865, "global_step": 177994, "epoch": 1999, "lr": 2.6591854109346522e-05}
{"train_loss": 0.10757266730070114, "global_step": 177995, "epoch": 1999, "lr": 2.659134183121052e-05}
{"train_loss": 0.11713041365146637, "global_step": 177996, "epoch": 1999, "lr": 2.6590829556221507e-05}
{"train_loss": 0.09288962185382843, "global_step": 177997, "epoch": 1999, "lr": 2.6590317284379594e-05}
{"train_loss": 0.08268287032842636, "global_step": 177998, "epoch": 1999, "lr": 2.658980501568481e-05}
{"train_loss": 0.0800514803006408, "global_step": 177999, "epoch": 1999, "lr": 2.6589292750137258e-05, "val_loss": 7.721046447753906}
{"train_loss": 0.08426690101623535, "global_step": 178000, "epoch": 2000, "lr": 2.6588780487736988e-05}
{"train_loss": 0.1537809818983078, "global_step": 178001, "epoch": 2000, "lr": 2.6588268228484058e-05}
{"train_loss": 0.07727047055959702, "global_step": 178002, "epoch": 2000, "lr": 2.6587755972378563e-05}
{"train_loss": 0.04675159975886345, "global_step": 178003, "epoch": 2000, "lr": 2.6587243719420542e-05}
{"train_loss": 0.0970478281378746, "global_step": 178004, "epoch": 2000, "lr": 2.6586731469610105e-05}
{"train_loss": 0.07981424778699875, "global_step": 178005, "epoch": 2000, "lr": 2.6586219222947273e-05}
{"train_loss": 0.08474577963352203, "global_step": 178006, "epoch": 2000, "lr": 2.6585706979432157e-05}
{"train_loss": 0.05253409594297409, "global_step": 178007, "epoch": 2000, "lr": 2.6585194739064782e-05}
{"train_loss": 0.09278205037117004, "global_step": 178008, "epoch": 2000, "lr": 2.658468250184527e-05}
{"train_loss": 0.14526723325252533, "global_step": 178009, "epoch": 2000, "lr": 2.6584170267773644e-05}
{"train_loss": 0.08039484173059464, "global_step": 178010, "epoch": 2000, "lr": 2.6583658036850002e-05}
{"train_loss": 0.07942335307598114, "global_step": 178011, "epoch": 2000, "lr": 2.6583145809074388e-05}
{"train_loss": 0.09953658282756805, "global_step": 178012, "epoch": 2000, "lr": 2.6582633584446888e-05}
{"train_loss": 0.05153508111834526, "global_step": 178013, "epoch": 2000, "lr": 2.6582121362967582e-05}
{"train_loss": 0.05415009707212448, "global_step": 178014, "epoch": 2000, "lr": 2.6581609144636504e-05}
{"train_loss": 0.09120821952819824, "global_step": 178015, "epoch": 2000, "lr": 2.6581096929453764e-05}
{"train_loss": 0.05971909686923027, "global_step": 178016, "epoch": 2000, "lr": 2.6580584717419387e-05}
{"train_loss": 0.08422990143299103, "global_step": 178017, "epoch": 2000, "lr": 2.6580072508533492e-05}
{"train_loss": 0.0763983428478241, "global_step": 178018, "epoch": 2000, "lr": 2.6579560302796097e-05}
{"train_loss": 0.08541280776262283, "global_step": 178019, "epoch": 2000, "lr": 2.6579048100207317e-05}
{"train_loss": 0.0578036904335022, "global_step": 178020, "epoch": 2000, "lr": 2.6578535900767178e-05}
{"train_loss": 0.06597020477056503, "global_step": 178021, "epoch": 2000, "lr": 2.6578023704475786e-05}
{"train_loss": 0.06859174370765686, "global_step": 178022, "epoch": 2000, "lr": 2.6577511511333174e-05}
{"train_loss": 0.11098621785640717, "global_step": 178023, "epoch": 2000, "lr": 2.6576999321339452e-05}
{"train_loss": 0.0513143427670002, "global_step": 178024, "epoch": 2000, "lr": 2.6576487134494642e-05}
{"train_loss": 0.1107461079955101, "global_step": 178025, "epoch": 2000, "lr": 2.6575974950798864e-05}
{"train_loss": 0.03422928601503372, "global_step": 178026, "epoch": 2000, "lr": 2.657546277025214e-05}
{"train_loss": 0.09384141117334366, "global_step": 178027, "epoch": 2000, "lr": 2.657495059285458e-05}
{"train_loss": 0.0405735969543457, "global_step": 178028, "epoch": 2000, "lr": 2.6574438418606208e-05}
{"train_loss": 0.08537571132183075, "global_step": 178029, "epoch": 2000, "lr": 2.657392624750714e-05}
{"train_loss": 0.05311470478773117, "global_step": 178030, "epoch": 2000, "lr": 2.65734140795574e-05}
{"train_loss": 0.06919602304697037, "global_step": 178031, "epoch": 2000, "lr": 2.6572901914757087e-05}
{"train_loss": 0.12084852159023285, "global_step": 178032, "epoch": 2000, "lr": 2.6572389753106273e-05}
{"train_loss": 0.0755496621131897, "global_step": 178033, "epoch": 2000, "lr": 2.6571877594605022e-05}
{"train_loss": 0.10594769567251205, "global_step": 178034, "epoch": 2000, "lr": 2.6571365439253372e-05}
{"train_loss": 0.08542433381080627, "global_step": 178035, "epoch": 2000, "lr": 2.6570853287051434e-05}
{"train_loss": 0.047453660517930984, "global_step": 178036, "epoch": 2000, "lr": 2.6570341137999246e-05}
{"train_loss": 0.08439008891582489, "global_step": 178037, "epoch": 2000, "lr": 2.65698289920969e-05}
{"train_loss": 0.07000592350959778, "global_step": 178038, "epoch": 2000, "lr": 2.6569316849344467e-05}
{"train_loss": 0.0622355192899704, "global_step": 178039, "epoch": 2000, "lr": 2.6568804709741978e-05}
{"train_loss": 0.07178575545549393, "global_step": 178040, "epoch": 2000, "lr": 2.6568292573289543e-05}
{"train_loss": 0.10299356281757355, "global_step": 178041, "epoch": 2000, "lr": 2.6567780439987204e-05}
{"train_loss": 0.09034422039985657, "global_step": 178042, "epoch": 2000, "lr": 2.6567268309835058e-05}
{"train_loss": 0.055564530193805695, "global_step": 178043, "epoch": 2000, "lr": 2.656675618283314e-05}
{"train_loss": 0.07676954567432404, "global_step": 178044, "epoch": 2000, "lr": 2.6566244058981547e-05}
{"train_loss": 0.08547255396842957, "global_step": 178045, "epoch": 2000, "lr": 2.6565731938280324e-05}
{"train_loss": 0.0851266086101532, "global_step": 178046, "epoch": 2000, "lr": 2.656521982072957e-05}
{"train_loss": 0.03538381680846214, "global_step": 178047, "epoch": 2000, "lr": 2.6564707706329322e-05}
{"train_loss": 0.031098971143364906, "global_step": 178048, "epoch": 2000, "lr": 2.6564195595079655e-05}
{"train_loss": 0.04496312886476517, "global_step": 178049, "epoch": 2000, "lr": 2.6563683486980673e-05}
{"train_loss": 0.13329391181468964, "global_step": 178050, "epoch": 2000, "lr": 2.65631713820324e-05}
{"train_loss": 0.06957534700632095, "global_step": 178051, "epoch": 2000, "lr": 2.6562659280234935e-05}
{"train_loss": 0.07263124734163284, "global_step": 178052, "epoch": 2000, "lr": 2.6562147181588322e-05}
{"train_loss": 0.11236083507537842, "global_step": 178053, "epoch": 2000, "lr": 2.6561635086092658e-05}
{"train_loss": 0.04235534369945526, "global_step": 178054, "epoch": 2000, "lr": 2.6561122993747977e-05}
{"train_loss": 0.07727281749248505, "global_step": 178055, "epoch": 2000, "lr": 2.6560610904554384e-05}
{"train_loss": 0.07836370915174484, "global_step": 178056, "epoch": 2000, "lr": 2.656009881851192e-05}
{"train_loss": 0.05609694495797157, "global_step": 178057, "epoch": 2000, "lr": 2.6559586735620678e-05}
{"train_loss": 0.07215067744255066, "global_step": 178058, "epoch": 2000, "lr": 2.65590746558807e-05}
{"train_loss": 0.08007699251174927, "global_step": 178059, "epoch": 2000, "lr": 2.655856257929208e-05}
{"train_loss": 0.06502911448478699, "global_step": 178060, "epoch": 2000, "lr": 2.6558050505854858e-05}
{"train_loss": 0.056590672582387924, "global_step": 178061, "epoch": 2000, "lr": 2.6557538435569144e-05}
{"train_loss": 0.07142193615436554, "global_step": 178062, "epoch": 2000, "lr": 2.6557026368434966e-05}
{"train_loss": 0.09826137870550156, "global_step": 178063, "epoch": 2000, "lr": 2.6556514304452427e-05}
{"train_loss": 0.05887326970696449, "global_step": 178064, "epoch": 2000, "lr": 2.655600224362156e-05}
{"train_loss": 0.034070517867803574, "global_step": 178065, "epoch": 2000, "lr": 2.6555490185942468e-05}
{"train_loss": 0.04846079647541046, "global_step": 178066, "epoch": 2000, "lr": 2.6554978131415185e-05}
{"train_loss": 0.11407960951328278, "global_step": 178067, "epoch": 2000, "lr": 2.6554466080039804e-05}
{"train_loss": 0.08795047551393509, "global_step": 178068, "epoch": 2000, "lr": 2.655395403181641e-05}
{"train_loss": 0.08312439173460007, "global_step": 178069, "epoch": 2000, "lr": 2.655344198674502e-05}
{"train_loss": 0.059889938682317734, "global_step": 178070, "epoch": 2000, "lr": 2.6552929944825768e-05}
{"train_loss": 0.09081563353538513, "global_step": 178071, "epoch": 2000, "lr": 2.655241790605868e-05}
{"train_loss": 0.07420021295547485, "global_step": 178072, "epoch": 2000, "lr": 2.6551905870443816e-05}
{"train_loss": 0.1312488317489624, "global_step": 178073, "epoch": 2000, "lr": 2.6551393837981288e-05}
{"train_loss": 0.04814307019114494, "global_step": 178074, "epoch": 2000, "lr": 2.6550881808671115e-05}
{"train_loss": 0.04482418671250343, "global_step": 178075, "epoch": 2000, "lr": 2.6550369782513405e-05}
{"train_loss": 0.14368151128292084, "global_step": 178076, "epoch": 2000, "lr": 2.6549857759508196e-05}
{"train_loss": 0.12054053694009781, "global_step": 178077, "epoch": 2000, "lr": 2.6549345739655597e-05}
{"train_loss": 0.04351452365517616, "global_step": 178078, "epoch": 2000, "lr": 2.6548833722955647e-05}
{"train_loss": 0.03547929972410202, "global_step": 178079, "epoch": 2000, "lr": 2.6548321709408398e-05}
{"train_loss": 0.08500552177429199, "global_step": 178080, "epoch": 2000, "lr": 2.654780969901397e-05}
{"train_loss": 0.05917022004723549, "global_step": 178081, "epoch": 2000, "lr": 2.6547297691772378e-05}
{"train_loss": 0.09143432974815369, "global_step": 178082, "epoch": 2000, "lr": 2.6546785687683733e-05}
{"train_loss": 0.08335035294294357, "global_step": 178083, "epoch": 2000, "lr": 2.6546273686748076e-05}
{"train_loss": 0.11066728830337524, "global_step": 178084, "epoch": 2000, "lr": 2.65457616889655e-05}
{"train_loss": 0.05326123908162117, "global_step": 178085, "epoch": 2000, "lr": 2.654524969433604e-05}
{"train_loss": 0.06732653081417084, "global_step": 178086, "epoch": 2000, "lr": 2.654473770285979e-05}
{"train_loss": 0.16299767792224884, "global_step": 178087, "epoch": 2000, "lr": 2.654422571453683e-05}
{"train_loss": 0.07839102886114897, "global_step": 178088, "epoch": 2000, "lr": 2.6543713729367193e-05, "train/sim_max_reward_0": 0.5370321160174071, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 0.4762335591718945, "train/sim_max_reward_3": 0.3916908394438463, "train/sim_max_reward_4": 0.9678599172451964, "train/sim_max_reward_5": 0.4347366030454537, "test/sim_max_reward_4500000": 0.372147255118777, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.954022346339751, "test/sim_max_reward_4500003": 0.6022700030899111, "test/sim_max_reward_4500004": 0.21771652286146567, "test/sim_max_reward_4500005": 0.3247592854974067, "test/sim_max_reward_4500006": 0.20116261138494715, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.196780295696233, "test/sim_max_reward_4500009": 0.05063797954430559, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9866419225168532, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.6477994378337542, "test/sim_max_reward_4500014": 0.9793245149457831, "test/sim_max_reward_4500015": 0.8498560109854885, "test/sim_max_reward_4500016": 0.6662726504243258, "test/sim_max_reward_4500017": 0.6638085100875897, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.6099265097903724, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.14861511220147822, "test/sim_max_reward_4500022": 0.9799621769953747, "test/sim_max_reward_4500023": 0.6841647909934208, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9848570150208742, "test/sim_max_reward_4500026": 0.17605462642272754, "test/sim_max_reward_4500027": 0.9941165866860009, "test/sim_max_reward_4500028": 0.7116312955934893, "test/sim_max_reward_4500029": 0.6276077562611992, "test/sim_max_reward_4500030": 0.949502417867273, "test/sim_max_reward_4500031": 0.6971887525664825, "test/sim_max_reward_4500032": 0.9493913545694896, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.9421997112403566, "test/sim_max_reward_4500035": 0.9344666662206859, "test/sim_max_reward_4500036": 0.09383960885440319, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.9819376365585999, "test/sim_max_reward_4500039": 0.970566110946652, "test/sim_max_reward_4500040": 0.9334734181075959, "test/sim_max_reward_4500041": 0.6173176291011025, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.3639440363893624, "test/sim_max_reward_4500044": 0.0012345343609664083, "test/sim_max_reward_4500045": 0.5786183130972189, "test/sim_max_reward_4500046": 0.41504652552404914, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.7992514058152859, "test/sim_max_reward_4500049": 0.9476341525443868, "train/mean_score": 0.6345921724872997, "test/mean_score": 0.5961149498011088, "val_loss": 7.406713008880615, "train_action_mse_error": 7.964087963104248}
{"train_loss": 0.05694112181663513, "global_step": 178089, "epoch": 2001, "lr": 2.654320174735099e-05}
{"train_loss": 0.09249003976583481, "global_step": 178090, "epoch": 2001, "lr": 2.6542689768488245e-05}
{"train_loss": 0.08658445626497269, "global_step": 178091, "epoch": 2001, "lr": 2.6542177792779073e-05}
{"train_loss": 0.13146533071994781, "global_step": 178092, "epoch": 2001, "lr": 2.6541665820223495e-05}
{"train_loss": 0.08957317471504211, "global_step": 178093, "epoch": 2001, "lr": 2.6541153850821625e-05}
{"train_loss": 0.03521603345870972, "global_step": 178094, "epoch": 2001, "lr": 2.654064188457349e-05}
{"train_loss": 0.06954875588417053, "global_step": 178095, "epoch": 2001, "lr": 2.6540129921479208e-05}
{"train_loss": 0.08834609389305115, "global_step": 178096, "epoch": 2001, "lr": 2.6539617961538787e-05}
{"train_loss": 0.041973013430833817, "global_step": 178097, "epoch": 2001, "lr": 2.6539106004752357e-05}
{"train_loss": 0.07586012035608292, "global_step": 178098, "epoch": 2001, "lr": 2.6538594051119937e-05}
{"train_loss": 0.06033199653029442, "global_step": 178099, "epoch": 2001, "lr": 2.6538082100641636e-05}
{"train_loss": 0.0735679417848587, "global_step": 178100, "epoch": 2001, "lr": 2.6537570153317486e-05}
{"train_loss": 0.08331672847270966, "global_step": 178101, "epoch": 2001, "lr": 2.6537058209147592e-05}
{"train_loss": 0.08390100300312042, "global_step": 178102, "epoch": 2001, "lr": 2.6536546268131984e-05}
{"train_loss": 0.016564663499593735, "global_step": 178103, "epoch": 2001, "lr": 2.6536034330270754e-05}
{"train_loss": 0.07743184268474579, "global_step": 178104, "epoch": 2001, "lr": 2.6535522395563984e-05}
{"train_loss": 0.07636885344982147, "global_step": 178105, "epoch": 2001, "lr": 2.6535010464011713e-05}
{"train_loss": 0.07011492550373077, "global_step": 178106, "epoch": 2001, "lr": 2.653449853561404e-05}
{"train_loss": 0.06941143423318863, "global_step": 178107, "epoch": 2001, "lr": 2.6533986610371e-05}
{"train_loss": 0.06437895447015762, "global_step": 178108, "epoch": 2001, "lr": 2.6533474688282688e-05}
{"train_loss": 0.09842439740896225, "global_step": 178109, "epoch": 2001, "lr": 2.6532962769349158e-05}
{"train_loss": 0.04080309346318245, "global_step": 178110, "epoch": 2001, "lr": 2.65324508535705e-05}
{"train_loss": 0.11212386935949326, "global_step": 178111, "epoch": 2001, "lr": 2.6531938940946766e-05}
{"train_loss": 0.05642944946885109, "global_step": 178112, "epoch": 2001, "lr": 2.6531427031478013e-05}
{"train_loss": 0.03513583168387413, "global_step": 178113, "epoch": 2001, "lr": 2.6530915125164335e-05}
{"train_loss": 0.0826595202088356, "global_step": 178114, "epoch": 2001, "lr": 2.6530403222005778e-05}
{"train_loss": 0.06699533760547638, "global_step": 178115, "epoch": 2001, "lr": 2.6529891322002438e-05}
{"train_loss": 0.11959582567214966, "global_step": 178116, "epoch": 2001, "lr": 2.6529379425154367e-05}
{"train_loss": 0.07220187038183212, "global_step": 178117, "epoch": 2001, "lr": 2.652886753146162e-05}
{"train_loss": 0.044778745621442795, "global_step": 178118, "epoch": 2001, "lr": 2.6528355640924297e-05}
{"train_loss": 0.10045681893825531, "global_step": 178119, "epoch": 2001, "lr": 2.6527843753542426e-05}
{"train_loss": 0.09342539310455322, "global_step": 178120, "epoch": 2001, "lr": 2.652733186931613e-05}
{"train_loss": 0.08171112835407257, "global_step": 178121, "epoch": 2001, "lr": 2.6526819988245422e-05}
{"train_loss": 0.052710387855768204, "global_step": 178122, "epoch": 2001, "lr": 2.65263081103304e-05}
{"train_loss": 0.0959060937166214, "global_step": 178123, "epoch": 2001, "lr": 2.6525796235571144e-05}
{"train_loss": 0.04968341067433357, "global_step": 178124, "epoch": 2001, "lr": 2.6525284363967694e-05}
{"train_loss": 0.08814521878957748, "global_step": 178125, "epoch": 2001, "lr": 2.652477249552015e-05}
{"train_loss": 0.07834074646234512, "global_step": 178126, "epoch": 2001, "lr": 2.652426063022855e-05}
{"train_loss": 0.07062271982431412, "global_step": 178127, "epoch": 2001, "lr": 2.6523748768092993e-05}
{"train_loss": 0.0707881897687912, "global_step": 178128, "epoch": 2001, "lr": 2.6523236909113512e-05}
{"train_loss": 0.03170269727706909, "global_step": 178129, "epoch": 2001, "lr": 2.6522725053290208e-05}
{"train_loss": 0.07207471877336502, "global_step": 178130, "epoch": 2001, "lr": 2.6522213200623132e-05}
{"train_loss": 0.033657778054475784, "global_step": 178131, "epoch": 2001, "lr": 2.652170135111237e-05}
{"train_loss": 0.07768616825342178, "global_step": 178132, "epoch": 2001, "lr": 2.652118950475796e-05}
{"train_loss": 0.11362477391958237, "global_step": 178133, "epoch": 2001, "lr": 2.6520677661560012e-05}
{"train_loss": 0.05957595631480217, "global_step": 178134, "epoch": 2001, "lr": 2.652016582151855e-05}
{"train_loss": 0.06643431633710861, "global_step": 178135, "epoch": 2001, "lr": 2.6519653984633684e-05}
{"train_loss": 0.059934649616479874, "global_step": 178136, "epoch": 2001, "lr": 2.651914215090545e-05}
{"train_loss": 0.041548557579517365, "global_step": 178137, "epoch": 2001, "lr": 2.6518630320333948e-05}
{"train_loss": 0.11349627375602722, "global_step": 178138, "epoch": 2001, "lr": 2.6518118492919208e-05}
{"train_loss": 0.0407898984849453, "global_step": 178139, "epoch": 2001, "lr": 2.6517606668661322e-05}
{"train_loss": 0.08696585148572922, "global_step": 178140, "epoch": 2001, "lr": 2.6517094847560382e-05}
{"train_loss": 0.09803378582000732, "global_step": 178141, "epoch": 2001, "lr": 2.6516583029616405e-05}
{"train_loss": 0.052438803017139435, "global_step": 178142, "epoch": 2001, "lr": 2.6516071214829513e-05}
{"train_loss": 0.08725515753030777, "global_step": 178143, "epoch": 2001, "lr": 2.651555940319973e-05}
{"train_loss": 0.05438246950507164, "global_step": 178144, "epoch": 2001, "lr": 2.651504759472716e-05}
{"train_loss": 0.08571910113096237, "global_step": 178145, "epoch": 2001, "lr": 2.651453578941183e-05}
{"train_loss": 0.11153285205364227, "global_step": 178146, "epoch": 2001, "lr": 2.6514023987253868e-05}
{"train_loss": 0.06882943958044052, "global_step": 178147, "epoch": 2001, "lr": 2.6513512188253275e-05}
{"train_loss": 0.047046124935150146, "global_step": 178148, "epoch": 2001, "lr": 2.651300039241018e-05}
{"train_loss": 0.0779038518667221, "global_step": 178149, "epoch": 2001, "lr": 2.6512488599724634e-05}
{"train_loss": 0.08769307285547256, "global_step": 178150, "epoch": 2001, "lr": 2.651197681019667e-05}
{"train_loss": 0.08150045573711395, "global_step": 178151, "epoch": 2001, "lr": 2.6511465023826403e-05}
{"train_loss": 0.11356733739376068, "global_step": 178152, "epoch": 2001, "lr": 2.6510953240613867e-05}
{"train_loss": 0.06875748187303543, "global_step": 178153, "epoch": 2001, "lr": 2.6510441460559164e-05}
{"train_loss": 0.05989738181233406, "global_step": 178154, "epoch": 2001, "lr": 2.650992968366235e-05}
{"train_loss": 0.06181967630982399, "global_step": 178155, "epoch": 2001, "lr": 2.650941790992346e-05}
{"train_loss": 0.0509335957467556, "global_step": 178156, "epoch": 2001, "lr": 2.650890613934262e-05}
{"train_loss": 0.10917339473962784, "global_step": 178157, "epoch": 2001, "lr": 2.6508394371919853e-05}
{"train_loss": 0.06691355258226395, "global_step": 178158, "epoch": 2001, "lr": 2.6507882607655244e-05}
{"train_loss": 0.07631572335958481, "global_step": 178159, "epoch": 2001, "lr": 2.6507370846548886e-05}
{"train_loss": 0.10381712019443512, "global_step": 178160, "epoch": 2001, "lr": 2.6506859088600798e-05}
{"train_loss": 0.06190970540046692, "global_step": 178161, "epoch": 2001, "lr": 2.65063473338111e-05}
{"train_loss": 0.08333256840705872, "global_step": 178162, "epoch": 2001, "lr": 2.650583558217982e-05}
{"train_loss": 0.09837077558040619, "global_step": 178163, "epoch": 2001, "lr": 2.6505323833707064e-05}
{"train_loss": 0.08228838443756104, "global_step": 178164, "epoch": 2001, "lr": 2.6504812088392862e-05}
{"train_loss": 0.08691343665122986, "global_step": 178165, "epoch": 2001, "lr": 2.650430034623732e-05}
{"train_loss": 0.09381858259439468, "global_step": 178166, "epoch": 2001, "lr": 2.6503788607240463e-05}
{"train_loss": 0.1345030963420868, "global_step": 178167, "epoch": 2001, "lr": 2.650327687140241e-05}
{"train_loss": 0.046239469200372696, "global_step": 178168, "epoch": 2001, "lr": 2.6502765138723184e-05}
{"train_loss": 0.10728171467781067, "global_step": 178169, "epoch": 2001, "lr": 2.6502253409202898e-05}
{"train_loss": 0.058824434876441956, "global_step": 178170, "epoch": 2001, "lr": 2.6501741682841573e-05}
{"train_loss": 0.11399617791175842, "global_step": 178171, "epoch": 2001, "lr": 2.650122995963933e-05}
{"train_loss": 0.06550412625074387, "global_step": 178172, "epoch": 2001, "lr": 2.6500718239596182e-05}
{"train_loss": 0.07976680248975754, "global_step": 178173, "epoch": 2001, "lr": 2.6500206522712252e-05}
{"train_loss": 0.09538306295871735, "global_step": 178174, "epoch": 2001, "lr": 2.6499694808987558e-05}
{"train_loss": 0.05582059547305107, "global_step": 178175, "epoch": 2001, "lr": 2.6499183098422198e-05}
{"train_loss": 0.12673908472061157, "global_step": 178176, "epoch": 2001, "lr": 2.649867139101625e-05}
{"train_loss": 0.07629915649134121, "global_step": 178177, "epoch": 2001, "lr": 2.649815968676975e-05, "val_loss": 7.3841705322265625}
{"train_loss": 0.07605881243944168, "global_step": 178178, "epoch": 2002, "lr": 2.6497647985682815e-05}
{"train_loss": 0.07852298766374588, "global_step": 178179, "epoch": 2002, "lr": 2.649713628775546e-05}
{"train_loss": 0.08379611372947693, "global_step": 178180, "epoch": 2002, "lr": 2.6496624592987795e-05}
{"train_loss": 0.10925058275461197, "global_step": 178181, "epoch": 2002, "lr": 2.6496112901379853e-05}
{"train_loss": 0.12693044543266296, "global_step": 178182, "epoch": 2002, "lr": 2.6495601212931744e-05}
{"train_loss": 0.09249305725097656, "global_step": 178183, "epoch": 2002, "lr": 2.6495089527643497e-05}
{"train_loss": 0.09248813986778259, "global_step": 178184, "epoch": 2002, "lr": 2.6494577845515216e-05}
{"train_loss": 0.09417508542537689, "global_step": 178185, "epoch": 2002, "lr": 2.6494066166546933e-05}
{"train_loss": 0.026924557983875275, "global_step": 178186, "epoch": 2002, "lr": 2.6493554490738757e-05}
{"train_loss": 0.06690805405378342, "global_step": 178187, "epoch": 2002, "lr": 2.6493042818090736e-05}
{"train_loss": 0.11612381041049957, "global_step": 178188, "epoch": 2002, "lr": 2.6492531148602917e-05}
{"train_loss": 0.09548192471265793, "global_step": 178189, "epoch": 2002, "lr": 2.649201948227541e-05}
{"train_loss": 0.11543960869312286, "global_step": 178190, "epoch": 2002, "lr": 2.649150781910825e-05}
{"train_loss": 0.10963382571935654, "global_step": 178191, "epoch": 2002, "lr": 2.6490996159101532e-05}
{"train_loss": 0.0738702267408371, "global_step": 178192, "epoch": 2002, "lr": 2.6490484502255304e-05}
{"train_loss": 0.0891505777835846, "global_step": 178193, "epoch": 2002, "lr": 2.6489972848569654e-05}
{"train_loss": 0.09135410934686661, "global_step": 178194, "epoch": 2002, "lr": 2.6489461198044625e-05}
{"train_loss": 0.06245250254869461, "global_step": 178195, "epoch": 2002, "lr": 2.648894955068032e-05}
{"train_loss": 0.06331025063991547, "global_step": 178196, "epoch": 2002, "lr": 2.648843790647677e-05}
{"train_loss": 0.0625762864947319, "global_step": 178197, "epoch": 2002, "lr": 2.6487926265434083e-05}
{"train_loss": 0.08359421044588089, "global_step": 178198, "epoch": 2002, "lr": 2.648741462755229e-05}
{"train_loss": 0.09854131937026978, "global_step": 178199, "epoch": 2002, "lr": 2.6486902992831493e-05}
{"train_loss": 0.1090770810842514, "global_step": 178200, "epoch": 2002, "lr": 2.648639136127173e-05}
{"train_loss": 0.11946847289800644, "global_step": 178201, "epoch": 2002, "lr": 2.64858797328731e-05}
{"train_loss": 0.13553433120250702, "global_step": 178202, "epoch": 2002, "lr": 2.648536810763564e-05}
{"train_loss": 0.10233648866415024, "global_step": 178203, "epoch": 2002, "lr": 2.648485648555945e-05}
{"train_loss": 0.09241314232349396, "global_step": 178204, "epoch": 2002, "lr": 2.6484344866644572e-05}
{"train_loss": 0.04212717339396477, "global_step": 178205, "epoch": 2002, "lr": 2.6483833250891104e-05}
{"train_loss": 0.08165369927883148, "global_step": 178206, "epoch": 2002, "lr": 2.648332163829908e-05}
{"train_loss": 0.05380713567137718, "global_step": 178207, "epoch": 2002, "lr": 2.6482810028868605e-05}
{"train_loss": 0.07074737548828125, "global_step": 178208, "epoch": 2002, "lr": 2.648229842259971e-05}
{"train_loss": 0.07763316482305527, "global_step": 178209, "epoch": 2002, "lr": 2.6481786819492504e-05}
{"train_loss": 0.11068443953990936, "global_step": 178210, "epoch": 2002, "lr": 2.6481275219547015e-05}
{"train_loss": 0.0589170940220356, "global_step": 178211, "epoch": 2002, "lr": 2.648076362276335e-05}
{"train_loss": 0.08339916169643402, "global_step": 178212, "epoch": 2002, "lr": 2.6480252029141538e-05}
{"train_loss": 0.07630730420351028, "global_step": 178213, "epoch": 2002, "lr": 2.6479740438681676e-05}
{"train_loss": 0.05858032777905464, "global_step": 178214, "epoch": 2002, "lr": 2.6479228851383843e-05}
{"train_loss": 0.09071770310401917, "global_step": 178215, "epoch": 2002, "lr": 2.647871726724807e-05}
{"train_loss": 0.05883011594414711, "global_step": 178216, "epoch": 2002, "lr": 2.647820568627447e-05}
{"train_loss": 0.024055851623415947, "global_step": 178217, "epoch": 2002, "lr": 2.6477694108463063e-05}
{"train_loss": 0.06141365319490433, "global_step": 178218, "epoch": 2002, "lr": 2.6477182533813965e-05}
{"train_loss": 0.044053614139556885, "global_step": 178219, "epoch": 2002, "lr": 2.647667096232721e-05}
{"train_loss": 0.07175594568252563, "global_step": 178220, "epoch": 2002, "lr": 2.6476159394002885e-05}
{"train_loss": 0.07368921488523483, "global_step": 178221, "epoch": 2002, "lr": 2.647564782884105e-05}
{"train_loss": 0.061303842812776566, "global_step": 178222, "epoch": 2002, "lr": 2.6475136266841787e-05}
{"train_loss": 0.05443767085671425, "global_step": 178223, "epoch": 2002, "lr": 2.647462470800514e-05}
{"train_loss": 0.042199138551950455, "global_step": 178224, "epoch": 2002, "lr": 2.6474113152331215e-05}
{"train_loss": 0.07052940130233765, "global_step": 178225, "epoch": 2002, "lr": 2.6473601599820032e-05}
{"train_loss": 0.08807776123285294, "global_step": 178226, "epoch": 2002, "lr": 2.647309005047171e-05}
{"train_loss": 0.0654139444231987, "global_step": 178227, "epoch": 2002, "lr": 2.6472578504286294e-05}
{"train_loss": 0.019921619445085526, "global_step": 178228, "epoch": 2002, "lr": 2.647206696126383e-05}
{"train_loss": 0.05550543591380119, "global_step": 178229, "epoch": 2002, "lr": 2.6471555421404437e-05}
{"train_loss": 0.08926192671060562, "global_step": 178230, "epoch": 2002, "lr": 2.647104388470813e-05}
{"train_loss": 0.12077465653419495, "global_step": 178231, "epoch": 2002, "lr": 2.6470532351175026e-05}
{"train_loss": 0.0788501501083374, "global_step": 178232, "epoch": 2002, "lr": 2.6470020820805152e-05}
{"train_loss": 0.055036671459674835, "global_step": 178233, "epoch": 2002, "lr": 2.646950929359862e-05}
{"train_loss": 0.04660782590508461, "global_step": 178234, "epoch": 2002, "lr": 2.6468997769555452e-05}
{"train_loss": 0.09531550109386444, "global_step": 178235, "epoch": 2002, "lr": 2.6468486248675762e-05}
{"train_loss": 0.04059417545795441, "global_step": 178236, "epoch": 2002, "lr": 2.6467974730959577e-05}
{"train_loss": 0.05895175784826279, "global_step": 178237, "epoch": 2002, "lr": 2.6467463216407007e-05}
{"train_loss": 0.09690727293491364, "global_step": 178238, "epoch": 2002, "lr": 2.6466951705018084e-05}
{"train_loss": 0.098004549741745, "global_step": 178239, "epoch": 2002, "lr": 2.6466440196792907e-05}
{"train_loss": 0.09728076308965683, "global_step": 178240, "epoch": 2002, "lr": 2.6465928691731513e-05}
{"train_loss": 0.03541288152337074, "global_step": 178241, "epoch": 2002, "lr": 2.6465417189834e-05}
{"train_loss": 0.07775439321994781, "global_step": 178242, "epoch": 2002, "lr": 2.6464905691100418e-05}
{"train_loss": 0.027216190472245216, "global_step": 178243, "epoch": 2002, "lr": 2.646439419553085e-05}
{"train_loss": 0.09438484907150269, "global_step": 178244, "epoch": 2002, "lr": 2.6463882703125343e-05}
{"train_loss": 0.08174449950456619, "global_step": 178245, "epoch": 2002, "lr": 2.6463371213884004e-05}
{"train_loss": 0.06259529292583466, "global_step": 178246, "epoch": 2002, "lr": 2.6462859727806848e-05}
{"train_loss": 0.13281622529029846, "global_step": 178247, "epoch": 2002, "lr": 2.6462348244893997e-05}
{"train_loss": 0.05529586598277092, "global_step": 178248, "epoch": 2002, "lr": 2.646183676514548e-05}
{"train_loss": 0.07113993912935257, "global_step": 178249, "epoch": 2002, "lr": 2.6461325288561383e-05}
{"train_loss": 0.07031145691871643, "global_step": 178250, "epoch": 2002, "lr": 2.6460813815141792e-05}
{"train_loss": 0.08621491491794586, "global_step": 178251, "epoch": 2002, "lr": 2.6460302344886734e-05}
{"train_loss": 0.07601207494735718, "global_step": 178252, "epoch": 2002, "lr": 2.645979087779633e-05}
{"train_loss": 0.05152834951877594, "global_step": 178253, "epoch": 2002, "lr": 2.6459279413870596e-05}
{"train_loss": 0.08030254393815994, "global_step": 178254, "epoch": 2002, "lr": 2.6458767953109642e-05}
{"train_loss": 0.06456299126148224, "global_step": 178255, "epoch": 2002, "lr": 2.6458256495513502e-05}
{"train_loss": 0.07555046677589417, "global_step": 178256, "epoch": 2002, "lr": 2.6457745041082283e-05}
{"train_loss": 0.06532279402017593, "global_step": 178257, "epoch": 2002, "lr": 2.645723358981601e-05}
{"train_loss": 0.0579027459025383, "global_step": 178258, "epoch": 2002, "lr": 2.6456722141714796e-05}
{"train_loss": 0.08203572779893875, "global_step": 178259, "epoch": 2002, "lr": 2.645621069677867e-05}
{"train_loss": 0.1241297796368599, "global_step": 178260, "epoch": 2002, "lr": 2.6455699255007738e-05}
{"train_loss": 0.04781505838036537, "global_step": 178261, "epoch": 2002, "lr": 2.645518781640204e-05}
{"train_loss": 0.08577439188957214, "global_step": 178262, "epoch": 2002, "lr": 2.6454676380961664e-05}
{"train_loss": 0.08011875301599503, "global_step": 178263, "epoch": 2002, "lr": 2.6454164948686655e-05}
{"train_loss": 0.05904702842235565, "global_step": 178264, "epoch": 2002, "lr": 2.645365351957711e-05}
{"train_loss": 0.09397073090076447, "global_step": 178265, "epoch": 2002, "lr": 2.6453142093633086e-05}
{"train_loss": 0.0775761872195126, "global_step": 178266, "epoch": 2002, "lr": 2.6452630670854638e-05, "val_loss": 7.402279376983643}
{"train_loss": 0.08498997986316681, "global_step": 178267, "epoch": 2003, "lr": 2.6452119251241857e-05}
{"train_loss": 0.04246482253074646, "global_step": 178268, "epoch": 2003, "lr": 2.645160783479479e-05}
{"train_loss": 0.04978622496128082, "global_step": 178269, "epoch": 2003, "lr": 2.6451096421513533e-05}
{"train_loss": 0.044330425560474396, "global_step": 178270, "epoch": 2003, "lr": 2.6450585011398122e-05}
{"train_loss": 0.019324883818626404, "global_step": 178271, "epoch": 2003, "lr": 2.6450073604448656e-05}
{"train_loss": 0.06894413381814957, "global_step": 178272, "epoch": 2003, "lr": 2.644956220066518e-05}
{"train_loss": 0.04880305379629135, "global_step": 178273, "epoch": 2003, "lr": 2.644905080004778e-05}
{"train_loss": 0.07054276764392853, "global_step": 178274, "epoch": 2003, "lr": 2.6448539402596507e-05}
{"train_loss": 0.0747106745839119, "global_step": 178275, "epoch": 2003, "lr": 2.6448028008311465e-05}
{"train_loss": 0.20209451019763947, "global_step": 178276, "epoch": 2003, "lr": 2.644751661719267e-05}
{"train_loss": 0.060801319777965546, "global_step": 178277, "epoch": 2003, "lr": 2.644700522924024e-05}
{"train_loss": 0.060327403247356415, "global_step": 178278, "epoch": 2003, "lr": 2.6446493844454212e-05}
{"train_loss": 0.08505260199308395, "global_step": 178279, "epoch": 2003, "lr": 2.6445982462834677e-05}
{"train_loss": 0.1382884383201599, "global_step": 178280, "epoch": 2003, "lr": 2.644547108438168e-05}
{"train_loss": 0.08879385888576508, "global_step": 178281, "epoch": 2003, "lr": 2.6444959709095312e-05}
{"train_loss": 0.05745001137256622, "global_step": 178282, "epoch": 2003, "lr": 2.644444833697562e-05}
{"train_loss": 0.06772958487272263, "global_step": 178283, "epoch": 2003, "lr": 2.6443936968022704e-05}
{"train_loss": 0.10002477467060089, "global_step": 178284, "epoch": 2003, "lr": 2.6443425602236594e-05}
{"train_loss": 0.052768707275390625, "global_step": 178285, "epoch": 2003, "lr": 2.6442914239617373e-05}
{"train_loss": 0.04533888399600983, "global_step": 178286, "epoch": 2003, "lr": 2.644240288016514e-05}
{"train_loss": 0.05096925422549248, "global_step": 178287, "epoch": 2003, "lr": 2.6441891523879914e-05}
{"train_loss": 0.0704970434308052, "global_step": 178288, "epoch": 2003, "lr": 2.6441380170761814e-05}
{"train_loss": 0.10858824849128723, "global_step": 178289, "epoch": 2003, "lr": 2.644086882081086e-05}
{"train_loss": 0.11259011179208755, "global_step": 178290, "epoch": 2003, "lr": 2.6440357474027167e-05}
{"train_loss": 0.062192291021347046, "global_step": 178291, "epoch": 2003, "lr": 2.643984613041076e-05}
{"train_loss": 0.0897056981921196, "global_step": 178292, "epoch": 2003, "lr": 2.6439334789961743e-05}
{"train_loss": 0.044196754693984985, "global_step": 178293, "epoch": 2003, "lr": 2.643882345268016e-05}
{"train_loss": 0.09784567356109619, "global_step": 178294, "epoch": 2003, "lr": 2.64383121185661e-05}
{"train_loss": 0.15037792921066284, "global_step": 178295, "epoch": 2003, "lr": 2.643780078761961e-05}
{"train_loss": 0.07304274290800095, "global_step": 178296, "epoch": 2003, "lr": 2.6437289459840785e-05}
{"train_loss": 0.08547237515449524, "global_step": 178297, "epoch": 2003, "lr": 2.643677813522966e-05}
{"train_loss": 0.06895344704389572, "global_step": 178298, "epoch": 2003, "lr": 2.643626681378635e-05}
{"train_loss": 0.12517888844013214, "global_step": 178299, "epoch": 2003, "lr": 2.643575549551087e-05}
{"train_loss": 0.07273942232131958, "global_step": 178300, "epoch": 2003, "lr": 2.6435244180403335e-05}
{"train_loss": 0.07679352909326553, "global_step": 178301, "epoch": 2003, "lr": 2.6434732868463776e-05}
{"train_loss": 0.09597253054380417, "global_step": 178302, "epoch": 2003, "lr": 2.6434221559692296e-05}
{"train_loss": 0.08098931610584259, "global_step": 178303, "epoch": 2003, "lr": 2.6433710254088933e-05}
{"train_loss": 0.10478190332651138, "global_step": 178304, "epoch": 2003, "lr": 2.6433198951653792e-05}
{"train_loss": 0.08568276464939117, "global_step": 178305, "epoch": 2003, "lr": 2.643268765238691e-05}
{"train_loss": 0.06055764853954315, "global_step": 178306, "epoch": 2003, "lr": 2.6432176356288358e-05}
{"train_loss": 0.05944661796092987, "global_step": 178307, "epoch": 2003, "lr": 2.6431665063358224e-05}
{"train_loss": 0.0918843224644661, "global_step": 178308, "epoch": 2003, "lr": 2.643115377359655e-05}
{"train_loss": 0.051465824246406555, "global_step": 178309, "epoch": 2003, "lr": 2.6430642487003442e-05}
{"train_loss": 0.07224203646183014, "global_step": 178310, "epoch": 2003, "lr": 2.6430131203578922e-05}
{"train_loss": 0.0675562247633934, "global_step": 178311, "epoch": 2003, "lr": 2.6429619923323108e-05}
{"train_loss": 0.0641138032078743, "global_step": 178312, "epoch": 2003, "lr": 2.6429108646236024e-05}
{"train_loss": 0.08278221637010574, "global_step": 178313, "epoch": 2003, "lr": 2.6428597372317775e-05}
{"train_loss": 0.08450952172279358, "global_step": 178314, "epoch": 2003, "lr": 2.6428086101568395e-05}
{"train_loss": 0.12872730195522308, "global_step": 178315, "epoch": 2003, "lr": 2.642757483398799e-05}
{"train_loss": 0.09052567929029465, "global_step": 178316, "epoch": 2003, "lr": 2.6427063569576594e-05}
{"train_loss": 0.07164542376995087, "global_step": 178317, "epoch": 2003, "lr": 2.642655230833431e-05}
{"train_loss": 0.06575450301170349, "global_step": 178318, "epoch": 2003, "lr": 2.642604105026117e-05}
{"train_loss": 0.07673317939043045, "global_step": 178319, "epoch": 2003, "lr": 2.642552979535728e-05}
{"train_loss": 0.06524583697319031, "global_step": 178320, "epoch": 2003, "lr": 2.642501854362267e-05}
{"train_loss": 0.08886165171861649, "global_step": 178321, "epoch": 2003, "lr": 2.6424507295057432e-05}
{"train_loss": 0.07831890136003494, "global_step": 178322, "epoch": 2003, "lr": 2.642399604966165e-05}
{"train_loss": 0.06277719885110855, "global_step": 178323, "epoch": 2003, "lr": 2.642348480743535e-05}
{"train_loss": 0.07596676796674728, "global_step": 178324, "epoch": 2003, "lr": 2.642297356837865e-05}
{"train_loss": 0.0791100412607193, "global_step": 178325, "epoch": 2003, "lr": 2.642246233249157e-05}
{"train_loss": 0.07354608178138733, "global_step": 178326, "epoch": 2003, "lr": 2.642195109977422e-05}
{"train_loss": 0.11163473129272461, "global_step": 178327, "epoch": 2003, "lr": 2.6421439870226645e-05}
{"train_loss": 0.11664963513612747, "global_step": 178328, "epoch": 2003, "lr": 2.642092864384893e-05}
{"train_loss": 0.053229812532663345, "global_step": 178329, "epoch": 2003, "lr": 2.6420417420641114e-05}
{"train_loss": 0.04443392530083656, "global_step": 178330, "epoch": 2003, "lr": 2.6419906200603305e-05}
{"train_loss": 0.1294100135564804, "global_step": 178331, "epoch": 2003, "lr": 2.6419394983735534e-05}
{"train_loss": 0.13018706440925598, "global_step": 178332, "epoch": 2003, "lr": 2.6418883770037912e-05}
{"train_loss": 0.07865028083324432, "global_step": 178333, "epoch": 2003, "lr": 2.6418372559510457e-05}
{"train_loss": 0.06330025941133499, "global_step": 178334, "epoch": 2003, "lr": 2.6417861352153284e-05}
{"train_loss": 0.07308690249919891, "global_step": 178335, "epoch": 2003, "lr": 2.641735014796643e-05}
{"train_loss": 0.02621590904891491, "global_step": 178336, "epoch": 2003, "lr": 2.641683894694999e-05}
{"train_loss": 0.052550043910741806, "global_step": 178337, "epoch": 2003, "lr": 2.6416327749103996e-05}
{"train_loss": 0.060085479170084, "global_step": 178338, "epoch": 2003, "lr": 2.6415816554428562e-05}
{"train_loss": 0.06901460140943527, "global_step": 178339, "epoch": 2003, "lr": 2.6415305362923716e-05}
{"train_loss": 0.07359787821769714, "global_step": 178340, "epoch": 2003, "lr": 2.6414794174589564e-05}
{"train_loss": 0.028727097436785698, "global_step": 178341, "epoch": 2003, "lr": 2.6414282989426138e-05}
{"train_loss": 0.07662408798933029, "global_step": 178342, "epoch": 2003, "lr": 2.6413771807433542e-05}
{"train_loss": 0.05809807777404785, "global_step": 178343, "epoch": 2003, "lr": 2.6413260628611825e-05}
{"train_loss": 0.1402052640914917, "global_step": 178344, "epoch": 2003, "lr": 2.641274945296104e-05}
{"train_loss": 0.0671406015753746, "global_step": 178345, "epoch": 2003, "lr": 2.6412238280481288e-05}
{"train_loss": 0.07110031694173813, "global_step": 178346, "epoch": 2003, "lr": 2.6411727111172614e-05}
{"train_loss": 0.030974142253398895, "global_step": 178347, "epoch": 2003, "lr": 2.6411215945035106e-05}
{"train_loss": 0.04653659090399742, "global_step": 178348, "epoch": 2003, "lr": 2.6410704782068805e-05}
{"train_loss": 0.10732194036245346, "global_step": 178349, "epoch": 2003, "lr": 2.6410193622273816e-05}
{"train_loss": 0.056514859199523926, "global_step": 178350, "epoch": 2003, "lr": 2.6409682465650172e-05}
{"train_loss": 0.060898616909980774, "global_step": 178351, "epoch": 2003, "lr": 2.640917131219798e-05}
{"train_loss": 0.09682182967662811, "global_step": 178352, "epoch": 2003, "lr": 2.640866016191726e-05}
{"train_loss": 0.05849354714155197, "global_step": 178353, "epoch": 2003, "lr": 2.640814901480813e-05}
{"train_loss": 0.06943739950656891, "global_step": 178354, "epoch": 2003, "lr": 2.6407637870870622e-05}
{"train_loss": 0.07677829922752434, "global_step": 178355, "epoch": 2003, "lr": 2.6407126730104837e-05, "val_loss": 7.5132927894592285}
{"train_loss": 0.09985312819480896, "global_step": 178356, "epoch": 2004, "lr": 2.640661559251081e-05}
{"train_loss": 0.08097660541534424, "global_step": 178357, "epoch": 2004, "lr": 2.640610445808862e-05}
{"train_loss": 0.10061881691217422, "global_step": 178358, "epoch": 2004, "lr": 2.6405593326838363e-05}
{"train_loss": 0.09509982913732529, "global_step": 178359, "epoch": 2004, "lr": 2.640508219876007e-05}
{"train_loss": 0.08050501346588135, "global_step": 178360, "epoch": 2004, "lr": 2.640457107385384e-05}
{"train_loss": 0.07666373252868652, "global_step": 178361, "epoch": 2004, "lr": 2.640405995211971e-05}
{"train_loss": 0.06995847076177597, "global_step": 178362, "epoch": 2004, "lr": 2.6403548833557788e-05}
{"train_loss": 0.058069415390491486, "global_step": 178363, "epoch": 2004, "lr": 2.64030377181681e-05}
{"train_loss": 0.06787148863077164, "global_step": 178364, "epoch": 2004, "lr": 2.6402526605950763e-05}
{"train_loss": 0.07604273408651352, "global_step": 178365, "epoch": 2004, "lr": 2.640201549690579e-05}
{"train_loss": 0.07872655987739563, "global_step": 178366, "epoch": 2004, "lr": 2.6401504391033304e-05}
{"train_loss": 0.0716865286231041, "global_step": 178367, "epoch": 2004, "lr": 2.640099328833332e-05}
{"train_loss": 0.0337948352098465, "global_step": 178368, "epoch": 2004, "lr": 2.6400482188805965e-05}
{"train_loss": 0.0949726551771164, "global_step": 178369, "epoch": 2004, "lr": 2.639997109245125e-05}
{"train_loss": 0.09904297441244125, "global_step": 178370, "epoch": 2004, "lr": 2.6399459999269293e-05}
{"train_loss": 0.0691830962896347, "global_step": 178371, "epoch": 2004, "lr": 2.6398948909260125e-05}
{"train_loss": 0.07862065732479095, "global_step": 178372, "epoch": 2004, "lr": 2.639843782242385e-05}
{"train_loss": 0.035959966480731964, "global_step": 178373, "epoch": 2004, "lr": 2.639792673876049e-05}
{"train_loss": 0.08463574945926666, "global_step": 178374, "epoch": 2004, "lr": 2.6397415658270164e-05}
{"train_loss": 0.11933184415102005, "global_step": 178375, "epoch": 2004, "lr": 2.6396904580952904e-05}
{"train_loss": 0.06420695036649704, "global_step": 178376, "epoch": 2004, "lr": 2.6396393506808802e-05}
{"train_loss": 0.0335230752825737, "global_step": 178377, "epoch": 2004, "lr": 2.6395882435837903e-05}
{"train_loss": 0.09983650594949722, "global_step": 178378, "epoch": 2004, "lr": 2.639537136804031e-05}
{"train_loss": 0.09017916768789291, "global_step": 178379, "epoch": 2004, "lr": 2.639486030341606e-05}
{"train_loss": 0.038088079541921616, "global_step": 178380, "epoch": 2004, "lr": 2.6394349241965244e-05}
{"train_loss": 0.09070080518722534, "global_step": 178381, "epoch": 2004, "lr": 2.639383818368792e-05}
{"train_loss": 0.07822085916996002, "global_step": 178382, "epoch": 2004, "lr": 2.639332712858414e-05}
{"train_loss": 0.08901060372591019, "global_step": 178383, "epoch": 2004, "lr": 2.6392816076654005e-05}
{"train_loss": 0.06593339890241623, "global_step": 178384, "epoch": 2004, "lr": 2.639230502789755e-05}
{"train_loss": 0.03570253774523735, "global_step": 178385, "epoch": 2004, "lr": 2.6391793982314884e-05}
{"train_loss": 0.0883522480726242, "global_step": 178386, "epoch": 2004, "lr": 2.639128293990603e-05}
{"train_loss": 0.046712055802345276, "global_step": 178387, "epoch": 2004, "lr": 2.6390771900671108e-05}
{"train_loss": 0.0488940067589283, "global_step": 178388, "epoch": 2004, "lr": 2.639026086461013e-05}
{"train_loss": 0.05351153761148453, "global_step": 178389, "epoch": 2004, "lr": 2.6389749831723216e-05}
{"train_loss": 0.04010357707738876, "global_step": 178390, "epoch": 2004, "lr": 2.6389238802010397e-05}
{"train_loss": 0.07922413945198059, "global_step": 178391, "epoch": 2004, "lr": 2.6388727775471766e-05}
{"train_loss": 0.11496109515428543, "global_step": 178392, "epoch": 2004, "lr": 2.6388216752107374e-05}
{"train_loss": 0.08242800086736679, "global_step": 178393, "epoch": 2004, "lr": 2.6387705731917296e-05}
{"train_loss": 0.10222647339105606, "global_step": 178394, "epoch": 2004, "lr": 2.638719471490162e-05}
{"train_loss": 0.06385420262813568, "global_step": 178395, "epoch": 2004, "lr": 2.6386683701060378e-05}
{"train_loss": 0.1113477423787117, "global_step": 178396, "epoch": 2004, "lr": 2.6386172690393684e-05}
{"train_loss": 0.10211586952209473, "global_step": 178397, "epoch": 2004, "lr": 2.638566168290155e-05}
{"train_loss": 0.07030659168958664, "global_step": 178398, "epoch": 2004, "lr": 2.6385150678584107e-05}
{"train_loss": 0.047225821763277054, "global_step": 178399, "epoch": 2004, "lr": 2.638463967744137e-05}
{"train_loss": 0.07327119261026382, "global_step": 178400, "epoch": 2004, "lr": 2.6384128679473452e-05}
{"train_loss": 0.1291969120502472, "global_step": 178401, "epoch": 2004, "lr": 2.6383617684680373e-05}
{"train_loss": 0.05542530491948128, "global_step": 178402, "epoch": 2004, "lr": 2.6383106693062254e-05}
{"train_loss": 0.13343021273612976, "global_step": 178403, "epoch": 2004, "lr": 2.6382595704619116e-05}
{"train_loss": 0.12198891490697861, "global_step": 178404, "epoch": 2004, "lr": 2.6382084719351074e-05}
{"train_loss": 0.10322380065917969, "global_step": 178405, "epoch": 2004, "lr": 2.6381573737258153e-05}
{"train_loss": 0.06150010973215103, "global_step": 178406, "epoch": 2004, "lr": 2.638106275834046e-05}
{"train_loss": 0.07462753355503082, "global_step": 178407, "epoch": 2004, "lr": 2.6380551782598028e-05}
{"train_loss": 0.06946597993373871, "global_step": 178408, "epoch": 2004, "lr": 2.6380040810030958e-05}
{"train_loss": 0.09811083227396011, "global_step": 178409, "epoch": 2004, "lr": 2.6379529840639285e-05}
{"train_loss": 0.077094666659832, "global_step": 178410, "epoch": 2004, "lr": 2.637901887442311e-05}
{"train_loss": 0.039205264300107956, "global_step": 178411, "epoch": 2004, "lr": 2.6378507911382478e-05}
{"train_loss": 0.04321174696087837, "global_step": 178412, "epoch": 2004, "lr": 2.6377996951517463e-05}
{"train_loss": 0.07933587580919266, "global_step": 178413, "epoch": 2004, "lr": 2.637748599482816e-05}
{"train_loss": 0.07478032261133194, "global_step": 178414, "epoch": 2004, "lr": 2.6376975041314618e-05}
{"train_loss": 0.06266950815916061, "global_step": 178415, "epoch": 2004, "lr": 2.6376464090976884e-05}
{"train_loss": 0.05860254541039467, "global_step": 178416, "epoch": 2004, "lr": 2.637595314381506e-05}
{"train_loss": 0.11645783483982086, "global_step": 178417, "epoch": 2004, "lr": 2.6375442199829187e-05}
{"train_loss": 0.03983432427048683, "global_step": 178418, "epoch": 2004, "lr": 2.6374931259019364e-05}
{"train_loss": 0.16679732501506805, "global_step": 178419, "epoch": 2004, "lr": 2.637442032138563e-05}
{"train_loss": 0.1337578296661377, "global_step": 178420, "epoch": 2004, "lr": 2.6373909386928075e-05}
{"train_loss": 0.08978722244501114, "global_step": 178421, "epoch": 2004, "lr": 2.637339845564677e-05}
{"train_loss": 0.09047850221395493, "global_step": 178422, "epoch": 2004, "lr": 2.6372887527541756e-05}
{"train_loss": 0.07711951434612274, "global_step": 178423, "epoch": 2004, "lr": 2.637237660261313e-05}
{"train_loss": 0.11097873002290726, "global_step": 178424, "epoch": 2004, "lr": 2.6371865680860934e-05}
{"train_loss": 0.06290274858474731, "global_step": 178425, "epoch": 2004, "lr": 2.6371354762285272e-05}
{"train_loss": 0.1496533751487732, "global_step": 178426, "epoch": 2004, "lr": 2.637084384688618e-05}
{"train_loss": 0.09528210759162903, "global_step": 178427, "epoch": 2004, "lr": 2.6370332934663755e-05}
{"train_loss": 0.07315730303525925, "global_step": 178428, "epoch": 2004, "lr": 2.636982202561803e-05}
{"train_loss": 0.06931211054325104, "global_step": 178429, "epoch": 2004, "lr": 2.6369311119749112e-05}
{"train_loss": 0.08109453320503235, "global_step": 178430, "epoch": 2004, "lr": 2.636880021705703e-05}
{"train_loss": 0.050599586218595505, "global_step": 178431, "epoch": 2004, "lr": 2.6368289317541884e-05}
{"train_loss": 0.153720885515213, "global_step": 178432, "epoch": 2004, "lr": 2.6367778421203747e-05}
{"train_loss": 0.06094873696565628, "global_step": 178433, "epoch": 2004, "lr": 2.636726752804266e-05}
{"train_loss": 0.048202089965343475, "global_step": 178434, "epoch": 2004, "lr": 2.636675663805872e-05}
{"train_loss": 0.07335454970598221, "global_step": 178435, "epoch": 2004, "lr": 2.636624575125196e-05}
{"train_loss": 0.04195106774568558, "global_step": 178436, "epoch": 2004, "lr": 2.63657348676225e-05}
{"train_loss": 0.09780578315258026, "global_step": 178437, "epoch": 2004, "lr": 2.636522398717035e-05}
{"train_loss": 0.08974593877792358, "global_step": 178438, "epoch": 2004, "lr": 2.6364713109895634e-05}
{"train_loss": 0.13253207504749298, "global_step": 178439, "epoch": 2004, "lr": 2.636420223579837e-05}
{"train_loss": 0.11523360013961792, "global_step": 178440, "epoch": 2004, "lr": 2.6363691364878675e-05}
{"train_loss": 0.13414828479290009, "global_step": 178441, "epoch": 2004, "lr": 2.6363180497136563e-05}
{"train_loss": 0.04117424786090851, "global_step": 178442, "epoch": 2004, "lr": 2.6362669632572163e-05}
{"train_loss": 0.1147664338350296, "global_step": 178443, "epoch": 2004, "lr": 2.6362158771185492e-05}
{"train_loss": 0.08131839410307702, "global_step": 178444, "epoch": 2004, "lr": 2.636164791297666e-05, "val_loss": 7.425971031188965}
{"train_loss": 0.06937862187623978, "global_step": 178445, "epoch": 2005, "lr": 2.6361137057945695e-05}
{"train_loss": 0.10806678235530853, "global_step": 178446, "epoch": 2005, "lr": 2.636062620609271e-05}
{"train_loss": 0.15812736749649048, "global_step": 178447, "epoch": 2005, "lr": 2.6360115357417725e-05}
{"train_loss": 0.053805336356163025, "global_step": 178448, "epoch": 2005, "lr": 2.6359604511920843e-05}
{"train_loss": 0.0824107751250267, "global_step": 178449, "epoch": 2005, "lr": 2.6359093669602137e-05}
{"train_loss": 0.08378491550683975, "global_step": 178450, "epoch": 2005, "lr": 2.635858283046164e-05}
{"train_loss": 0.060580842196941376, "global_step": 178451, "epoch": 2005, "lr": 2.6358071994499474e-05}
{"train_loss": 0.061591461300849915, "global_step": 178452, "epoch": 2005, "lr": 2.635756116171565e-05}
{"train_loss": 0.07561040669679642, "global_step": 178453, "epoch": 2005, "lr": 2.635705033211029e-05}
{"train_loss": 0.05443129688501358, "global_step": 178454, "epoch": 2005, "lr": 2.635653950568343e-05}
{"train_loss": 0.06671830266714096, "global_step": 178455, "epoch": 2005, "lr": 2.6356028682435123e-05}
{"train_loss": 0.09546682238578796, "global_step": 178456, "epoch": 2005, "lr": 2.6355517862365486e-05}
{"train_loss": 0.04235481470823288, "global_step": 178457, "epoch": 2005, "lr": 2.635500704547454e-05}
{"train_loss": 0.1001618430018425, "global_step": 178458, "epoch": 2005, "lr": 2.6354496231762393e-05}
{"train_loss": 0.04102890193462372, "global_step": 178459, "epoch": 2005, "lr": 2.63539854212291e-05}
{"train_loss": 0.06288830190896988, "global_step": 178460, "epoch": 2005, "lr": 2.6353474613874697e-05}
{"train_loss": 0.04054680094122887, "global_step": 178461, "epoch": 2005, "lr": 2.6352963809699305e-05}
{"train_loss": 0.0953848734498024, "global_step": 178462, "epoch": 2005, "lr": 2.635245300870295e-05}
{"train_loss": 0.05043709650635719, "global_step": 178463, "epoch": 2005, "lr": 2.6351942210885738e-05}
{"train_loss": 0.04769675433635712, "global_step": 178464, "epoch": 2005, "lr": 2.6351431416247706e-05}
{"train_loss": 0.039549924433231354, "global_step": 178465, "epoch": 2005, "lr": 2.635092062478895e-05}
{"train_loss": 0.05760961398482323, "global_step": 178466, "epoch": 2005, "lr": 2.6350409836509504e-05}
{"train_loss": 0.1202300488948822, "global_step": 178467, "epoch": 2005, "lr": 2.634989905140946e-05}
{"train_loss": 0.07089076191186905, "global_step": 178468, "epoch": 2005, "lr": 2.63493882694889e-05}
{"train_loss": 0.07789897173643112, "global_step": 178469, "epoch": 2005, "lr": 2.634887749074786e-05}
{"train_loss": 0.07771716266870499, "global_step": 178470, "epoch": 2005, "lr": 2.6348366715186445e-05}
{"train_loss": 0.039115384221076965, "global_step": 178471, "epoch": 2005, "lr": 2.6347855942804678e-05}
{"train_loss": 0.0860547125339508, "global_step": 178472, "epoch": 2005, "lr": 2.6347345173602676e-05}
{"train_loss": 0.09059597551822662, "global_step": 178473, "epoch": 2005, "lr": 2.6346834407580473e-05}
{"train_loss": 0.08882037550210953, "global_step": 178474, "epoch": 2005, "lr": 2.6346323644738164e-05}
{"train_loss": 0.1277560591697693, "global_step": 178475, "epoch": 2005, "lr": 2.6345812885075783e-05}
{"train_loss": 0.07233352959156036, "global_step": 178476, "epoch": 2005, "lr": 2.634530212859344e-05}
{"train_loss": 0.07872354984283447, "global_step": 178477, "epoch": 2005, "lr": 2.6344791375291165e-05}
{"train_loss": 0.07154407352209091, "global_step": 178478, "epoch": 2005, "lr": 2.6344280625169058e-05}
{"train_loss": 0.05610119551420212, "global_step": 178479, "epoch": 2005, "lr": 2.634376987822716e-05}
{"train_loss": 0.10751514881849289, "global_step": 178480, "epoch": 2005, "lr": 2.634325913446557e-05}
{"train_loss": 0.05168038234114647, "global_step": 178481, "epoch": 2005, "lr": 2.6342748393884326e-05}
{"train_loss": 0.0789603739976883, "global_step": 178482, "epoch": 2005, "lr": 2.6342237656483526e-05}
{"train_loss": 0.05921947956085205, "global_step": 178483, "epoch": 2005, "lr": 2.6341726922263205e-05}
{"train_loss": 0.09783617407083511, "global_step": 178484, "epoch": 2005, "lr": 2.634121619122345e-05}
{"train_loss": 0.1463354229927063, "global_step": 178485, "epoch": 2005, "lr": 2.634070546336435e-05}
{"train_loss": 0.05915101617574692, "global_step": 178486, "epoch": 2005, "lr": 2.6340194738685937e-05}
{"train_loss": 0.05684764310717583, "global_step": 178487, "epoch": 2005, "lr": 2.6339684017188315e-05}
{"train_loss": 0.023167502135038376, "global_step": 178488, "epoch": 2005, "lr": 2.6339173298871513e-05}
{"train_loss": 0.033784184604883194, "global_step": 178489, "epoch": 2005, "lr": 2.633866258373564e-05}
{"train_loss": 0.04386153072118759, "global_step": 178490, "epoch": 2005, "lr": 2.6338151871780726e-05}
{"train_loss": 0.04932187870144844, "global_step": 178491, "epoch": 2005, "lr": 2.633764116300688e-05}
{"train_loss": 0.044922295957803726, "global_step": 178492, "epoch": 2005, "lr": 2.633713045741415e-05}
{"train_loss": 0.03664112091064453, "global_step": 178493, "epoch": 2005, "lr": 2.6336619755002584e-05}
{"train_loss": 0.14336329698562622, "global_step": 178494, "epoch": 2005, "lr": 2.633610905577229e-05}
{"train_loss": 0.07855338603258133, "global_step": 178495, "epoch": 2005, "lr": 2.6335598359723302e-05}
{"train_loss": 0.14218534529209137, "global_step": 178496, "epoch": 2005, "lr": 2.6335087666855716e-05}
{"train_loss": 0.07627490907907486, "global_step": 178497, "epoch": 2005, "lr": 2.6334576977169596e-05}
{"train_loss": 0.10015026479959488, "global_step": 178498, "epoch": 2005, "lr": 2.6334066290664983e-05}
{"train_loss": 0.09603229910135269, "global_step": 178499, "epoch": 2005, "lr": 2.6333555607341988e-05}
{"train_loss": 0.06828632950782776, "global_step": 178500, "epoch": 2005, "lr": 2.6333044927200635e-05}
{"train_loss": 0.04658966511487961, "global_step": 178501, "epoch": 2005, "lr": 2.633253425024103e-05}
{"train_loss": 0.058464888483285904, "global_step": 178502, "epoch": 2005, "lr": 2.6332023576463215e-05}
{"train_loss": 0.09551464766263962, "global_step": 178503, "epoch": 2005, "lr": 2.6331512905867266e-05}
{"train_loss": 0.05109593644738197, "global_step": 178504, "epoch": 2005, "lr": 2.6331002238453284e-05}
{"train_loss": 0.1107368916273117, "global_step": 178505, "epoch": 2005, "lr": 2.6330491574221284e-05}
{"train_loss": 0.058025289326906204, "global_step": 178506, "epoch": 2005, "lr": 2.6329980913171383e-05}
{"train_loss": 0.11636167019605637, "global_step": 178507, "epoch": 2005, "lr": 2.6329470255303608e-05}
{"train_loss": 0.060780446976423264, "global_step": 178508, "epoch": 2005, "lr": 2.6328959600618065e-05}
{"train_loss": 0.14393633604049683, "global_step": 178509, "epoch": 2005, "lr": 2.632844894911478e-05}
{"train_loss": 0.01805827207863331, "global_step": 178510, "epoch": 2005, "lr": 2.6327938300793874e-05}
{"train_loss": 0.07481877505779266, "global_step": 178511, "epoch": 2005, "lr": 2.632742765565537e-05}
{"train_loss": 0.05669692903757095, "global_step": 178512, "epoch": 2005, "lr": 2.632691701369937e-05}
{"train_loss": 0.044946592301130295, "global_step": 178513, "epoch": 2005, "lr": 2.632640637492591e-05}
{"train_loss": 0.057640157639980316, "global_step": 178514, "epoch": 2005, "lr": 2.6325895739335094e-05}
{"train_loss": 0.03561839461326599, "global_step": 178515, "epoch": 2005, "lr": 2.632538510692696e-05}
{"train_loss": 0.03477878496050835, "global_step": 178516, "epoch": 2005, "lr": 2.63248744777016e-05}
{"train_loss": 0.0666821077466011, "global_step": 178517, "epoch": 2005, "lr": 2.6324363851659055e-05}
{"train_loss": 0.06469929218292236, "global_step": 178518, "epoch": 2005, "lr": 2.6323853228799434e-05}
{"train_loss": 0.069601871073246, "global_step": 178519, "epoch": 2005, "lr": 2.632334260912276e-05}
{"train_loss": 0.05842597410082817, "global_step": 178520, "epoch": 2005, "lr": 2.6322831992629125e-05}
{"train_loss": 0.05851142480969429, "global_step": 178521, "epoch": 2005, "lr": 2.6322321379318614e-05}
{"train_loss": 0.03494437411427498, "global_step": 178522, "epoch": 2005, "lr": 2.6321810769191256e-05}
{"train_loss": 0.0800577700138092, "global_step": 178523, "epoch": 2005, "lr": 2.6321300162247164e-05}
{"train_loss": 0.08039301633834839, "global_step": 178524, "epoch": 2005, "lr": 2.632078955848637e-05}
{"train_loss": 0.1723879724740982, "global_step": 178525, "epoch": 2005, "lr": 2.6320278957908973e-05}
{"train_loss": 0.16052968800067902, "global_step": 178526, "epoch": 2005, "lr": 2.6319768360515007e-05}
{"train_loss": 0.13284331560134888, "global_step": 178527, "epoch": 2005, "lr": 2.6319257766304578e-05}
{"train_loss": 0.11046479642391205, "global_step": 178528, "epoch": 2005, "lr": 2.6318747175277718e-05}
{"train_loss": 0.042657095938920975, "global_step": 178529, "epoch": 2005, "lr": 2.631823658743453e-05}
{"train_loss": 0.04598985239863396, "global_step": 178530, "epoch": 2005, "lr": 2.631772600277507e-05}
{"train_loss": 0.12701676785945892, "global_step": 178531, "epoch": 2005, "lr": 2.6317215421299386e-05}
{"train_loss": 0.04958067461848259, "global_step": 178532, "epoch": 2005, "lr": 2.6316704843007578e-05}
{"train_loss": 0.07529539604451549, "global_step": 178533, "epoch": 2005, "lr": 2.6316194267899686e-05, "val_loss": 7.511528968811035, "train_action_mse_error": 25.38892936706543}
{"train_loss": 0.08685774356126785, "global_step": 178534, "epoch": 2006, "lr": 2.631568369597581e-05}
{"train_loss": 0.09697719663381577, "global_step": 178535, "epoch": 2006, "lr": 2.631517312723598e-05}
{"train_loss": 0.05066670477390289, "global_step": 178536, "epoch": 2006, "lr": 2.6314662561680304e-05}
{"train_loss": 0.05628495663404465, "global_step": 178537, "epoch": 2006, "lr": 2.6314151999308833e-05}
{"train_loss": 0.0815110057592392, "global_step": 178538, "epoch": 2006, "lr": 2.6313641440121623e-05}
{"train_loss": 0.041808344423770905, "global_step": 178539, "epoch": 2006, "lr": 2.6313130884118753e-05}
{"train_loss": 0.0840902030467987, "global_step": 178540, "epoch": 2006, "lr": 2.631262033130031e-05}
{"train_loss": 0.08086763322353363, "global_step": 178541, "epoch": 2006, "lr": 2.6312109781666327e-05}
{"train_loss": 0.09624692052602768, "global_step": 178542, "epoch": 2006, "lr": 2.6311599235216917e-05}
{"train_loss": 0.05851378291845322, "global_step": 178543, "epoch": 2006, "lr": 2.63110886919521e-05}
{"train_loss": 0.06568238139152527, "global_step": 178544, "epoch": 2006, "lr": 2.631057815187199e-05}
{"train_loss": 0.057332511991262436, "global_step": 178545, "epoch": 2006, "lr": 2.6310067614976606e-05}
{"train_loss": 0.07507240772247314, "global_step": 178546, "epoch": 2006, "lr": 2.6309557081266078e-05}
{"train_loss": 0.09984336793422699, "global_step": 178547, "epoch": 2006, "lr": 2.6309046550740413e-05}
{"train_loss": 0.04884437099099159, "global_step": 178548, "epoch": 2006, "lr": 2.630853602339973e-05}
{"train_loss": 0.11772674322128296, "global_step": 178549, "epoch": 2006, "lr": 2.6308025499244053e-05}
{"train_loss": 0.076631560921669, "global_step": 178550, "epoch": 2006, "lr": 2.6307514978273496e-05}
{"train_loss": 0.06544340401887894, "global_step": 178551, "epoch": 2006, "lr": 2.630700446048808e-05}
{"train_loss": 0.12680405378341675, "global_step": 178552, "epoch": 2006, "lr": 2.6306493945887927e-05}
{"train_loss": 0.059942055493593216, "global_step": 178553, "epoch": 2006, "lr": 2.6305983434473057e-05}
{"train_loss": 0.05159768834710121, "global_step": 178554, "epoch": 2006, "lr": 2.630547292624357e-05}
{"train_loss": 0.04820183292031288, "global_step": 178555, "epoch": 2006, "lr": 2.630496242119951e-05}
{"train_loss": 0.05573507025837898, "global_step": 178556, "epoch": 2006, "lr": 2.6304451919340977e-05}
{"train_loss": 0.10958976298570633, "global_step": 178557, "epoch": 2006, "lr": 2.6303941420668e-05}
{"train_loss": 0.05660003423690796, "global_step": 178558, "epoch": 2006, "lr": 2.6303430925180673e-05}
{"train_loss": 0.08230803906917572, "global_step": 178559, "epoch": 2006, "lr": 2.630292043287908e-05}
{"train_loss": 0.0708903893828392, "global_step": 178560, "epoch": 2006, "lr": 2.630240994376325e-05}
{"train_loss": 0.15155060589313507, "global_step": 178561, "epoch": 2006, "lr": 2.630189945783329e-05}
{"train_loss": 0.07441680133342743, "global_step": 178562, "epoch": 2006, "lr": 2.6301388975089237e-05}
{"train_loss": 0.0769425705075264, "global_step": 178563, "epoch": 2006, "lr": 2.6300878495531185e-05}
{"train_loss": 0.039877746254205704, "global_step": 178564, "epoch": 2006, "lr": 2.6300368019159173e-05}
{"train_loss": 0.07279994338750839, "global_step": 178565, "epoch": 2006, "lr": 2.6299857545973312e-05}
{"train_loss": 0.05804504454135895, "global_step": 178566, "epoch": 2006, "lr": 2.6299347075973623e-05}
{"train_loss": 0.07799293100833893, "global_step": 178567, "epoch": 2006, "lr": 2.6298836609160215e-05}
{"train_loss": 0.09347710013389587, "global_step": 178568, "epoch": 2006, "lr": 2.6298326145533125e-05}
{"train_loss": 0.07015553116798401, "global_step": 178569, "epoch": 2006, "lr": 2.6297815685092454e-05}
{"train_loss": 0.08145596086978912, "global_step": 178570, "epoch": 2006, "lr": 2.6297305227838253e-05}
{"train_loss": 0.03474883735179901, "global_step": 178571, "epoch": 2006, "lr": 2.6296794773770573e-05}
{"train_loss": 0.10634397715330124, "global_step": 178572, "epoch": 2006, "lr": 2.6296284322889512e-05}
{"train_loss": 0.11185739189386368, "global_step": 178573, "epoch": 2006, "lr": 2.629577387519512e-05}
{"train_loss": 0.13974539935588837, "global_step": 178574, "epoch": 2006, "lr": 2.629526343068748e-05}
{"train_loss": 0.10480132699012756, "global_step": 178575, "epoch": 2006, "lr": 2.629475298936664e-05}
{"train_loss": 0.15168121457099915, "global_step": 178576, "epoch": 2006, "lr": 2.6294242551232696e-05}
{"train_loss": 0.08042021095752716, "global_step": 178577, "epoch": 2006, "lr": 2.6293732116285687e-05}
{"train_loss": 0.07900841534137726, "global_step": 178578, "epoch": 2006, "lr": 2.6293221684525714e-05}
{"train_loss": 0.1185104250907898, "global_step": 178579, "epoch": 2006, "lr": 2.6292711255952808e-05}
{"train_loss": 0.07356484979391098, "global_step": 178580, "epoch": 2006, "lr": 2.6292200830567083e-05}
{"train_loss": 0.09135948866605759, "global_step": 178581, "epoch": 2006, "lr": 2.6291690408368553e-05}
{"train_loss": 0.07140808552503586, "global_step": 178582, "epoch": 2006, "lr": 2.6291179989357345e-05}
{"train_loss": 0.07236278057098389, "global_step": 178583, "epoch": 2006, "lr": 2.6290669573533477e-05}
{"train_loss": 0.10112815350294113, "global_step": 178584, "epoch": 2006, "lr": 2.6290159160897055e-05}
{"train_loss": 0.04930364340543747, "global_step": 178585, "epoch": 2006, "lr": 2.6289648751448116e-05}
{"train_loss": 0.08901100605726242, "global_step": 178586, "epoch": 2006, "lr": 2.628913834518677e-05}
{"train_loss": 0.08462046086788177, "global_step": 178587, "epoch": 2006, "lr": 2.628862794211303e-05}
{"train_loss": 0.0726468563079834, "global_step": 178588, "epoch": 2006, "lr": 2.628811754222702e-05}
{"train_loss": 0.07608620822429657, "global_step": 178589, "epoch": 2006, "lr": 2.6287607145528766e-05}
{"train_loss": 0.06334806978702545, "global_step": 178590, "epoch": 2006, "lr": 2.6287096752018368e-05}
{"train_loss": 0.07186193764209747, "global_step": 178591, "epoch": 2006, "lr": 2.6286586361695864e-05}
{"train_loss": 0.1128431111574173, "global_step": 178592, "epoch": 2006, "lr": 2.6286075974561364e-05}
{"train_loss": 0.15559524297714233, "global_step": 178593, "epoch": 2006, "lr": 2.6285565590614887e-05}
{"train_loss": 0.04174376651644707, "global_step": 178594, "epoch": 2006, "lr": 2.6285055209856523e-05}
{"train_loss": 0.04875633120536804, "global_step": 178595, "epoch": 2006, "lr": 2.6284544832286374e-05}
{"train_loss": 0.10266004502773285, "global_step": 178596, "epoch": 2006, "lr": 2.628403445790445e-05}
{"train_loss": 0.042809274047613144, "global_step": 178597, "epoch": 2006, "lr": 2.6283524086710877e-05}
{"train_loss": 0.07468930631875992, "global_step": 178598, "epoch": 2006, "lr": 2.6283013718705668e-05}
{"train_loss": 0.07430121302604675, "global_step": 178599, "epoch": 2006, "lr": 2.6282503353888942e-05}
{"train_loss": 0.051232848316431046, "global_step": 178600, "epoch": 2006, "lr": 2.6281992992260722e-05}
{"train_loss": 0.15024349093437195, "global_step": 178601, "epoch": 2006, "lr": 2.628148263382112e-05}
{"train_loss": 0.10590279847383499, "global_step": 178602, "epoch": 2006, "lr": 2.6280972278570164e-05}
{"train_loss": 0.09349897503852844, "global_step": 178603, "epoch": 2006, "lr": 2.628046192650796e-05}
{"train_loss": 0.05706428736448288, "global_step": 178604, "epoch": 2006, "lr": 2.6279951577634544e-05}
{"train_loss": 0.045844193547964096, "global_step": 178605, "epoch": 2006, "lr": 2.6279441231950018e-05}
{"train_loss": 0.06017976626753807, "global_step": 178606, "epoch": 2006, "lr": 2.6278930889454413e-05}
{"train_loss": 0.08904226124286652, "global_step": 178607, "epoch": 2006, "lr": 2.6278420550147832e-05}
{"train_loss": 0.05042491853237152, "global_step": 178608, "epoch": 2006, "lr": 2.6277910214030327e-05}
{"train_loss": 0.0547318160533905, "global_step": 178609, "epoch": 2006, "lr": 2.6277399881101948e-05}
{"train_loss": 0.06692636013031006, "global_step": 178610, "epoch": 2006, "lr": 2.627688955136281e-05}
{"train_loss": 0.07727189362049103, "global_step": 178611, "epoch": 2006, "lr": 2.627637922481293e-05}
{"train_loss": 0.11022195965051651, "global_step": 178612, "epoch": 2006, "lr": 2.6275868901452428e-05}
{"train_loss": 0.08576129376888275, "global_step": 178613, "epoch": 2006, "lr": 2.6275358581281318e-05}
{"train_loss": 0.11157649755477905, "global_step": 178614, "epoch": 2006, "lr": 2.6274848264299724e-05}
{"train_loss": 0.14391523599624634, "global_step": 178615, "epoch": 2006, "lr": 2.627433795050766e-05}
{"train_loss": 0.04606917127966881, "global_step": 178616, "epoch": 2006, "lr": 2.6273827639905246e-05}
{"train_loss": 0.08802716434001923, "global_step": 178617, "epoch": 2006, "lr": 2.627331733249251e-05}
{"train_loss": 0.08943244814872742, "global_step": 178618, "epoch": 2006, "lr": 2.627280702826955e-05}
{"train_loss": 0.09259537607431412, "global_step": 178619, "epoch": 2006, "lr": 2.6272296727236405e-05}
{"train_loss": 0.1111961305141449, "global_step": 178620, "epoch": 2006, "lr": 2.6271786429393185e-05}
{"train_loss": 0.07670346647500992, "global_step": 178621, "epoch": 2006, "lr": 2.6271276134739907e-05}
{"train_loss": 0.08095854177568736, "global_step": 178622, "epoch": 2006, "lr": 2.6270765843276692e-05, "val_loss": 7.387064456939697}
{"train_loss": 0.08875681459903717, "global_step": 178623, "epoch": 2007, "lr": 2.6270255555003565e-05}
{"train_loss": 0.08844597637653351, "global_step": 178624, "epoch": 2007, "lr": 2.6269745269920625e-05}
{"train_loss": 0.06025514379143715, "global_step": 178625, "epoch": 2007, "lr": 2.6269234988027913e-05}
{"train_loss": 0.0897950530052185, "global_step": 178626, "epoch": 2007, "lr": 2.6268724709325533e-05}
{"train_loss": 0.18322664499282837, "global_step": 178627, "epoch": 2007, "lr": 2.6268214433813516e-05}
{"train_loss": 0.026437262073159218, "global_step": 178628, "epoch": 2007, "lr": 2.626770416149196e-05}
{"train_loss": 0.07183235138654709, "global_step": 178629, "epoch": 2007, "lr": 2.6267193892360908e-05}
{"train_loss": 0.1135750561952591, "global_step": 178630, "epoch": 2007, "lr": 2.6266683626420445e-05}
{"train_loss": 0.07437901943922043, "global_step": 178631, "epoch": 2007, "lr": 2.6266173363670654e-05}
{"train_loss": 0.044786110520362854, "global_step": 178632, "epoch": 2007, "lr": 2.626566310411157e-05}
{"train_loss": 0.08698072284460068, "global_step": 178633, "epoch": 2007, "lr": 2.6265152847743295e-05}
{"train_loss": 0.16321314871311188, "global_step": 178634, "epoch": 2007, "lr": 2.6264642594565857e-05}
{"train_loss": 0.11440134793519974, "global_step": 178635, "epoch": 2007, "lr": 2.626413234457938e-05}
{"train_loss": 0.058969803154468536, "global_step": 178636, "epoch": 2007, "lr": 2.626362209778387e-05}
{"train_loss": 0.12641163170337677, "global_step": 178637, "epoch": 2007, "lr": 2.6263111854179455e-05}
{"train_loss": 0.05188867822289467, "global_step": 178638, "epoch": 2007, "lr": 2.6262601613766146e-05}
{"train_loss": 0.04523215442895889, "global_step": 178639, "epoch": 2007, "lr": 2.6262091376544068e-05}
{"train_loss": 0.08189170062541962, "global_step": 178640, "epoch": 2007, "lr": 2.6261581142513243e-05}
{"train_loss": 0.08204969018697739, "global_step": 178641, "epoch": 2007, "lr": 2.6261070911673775e-05}
{"train_loss": 0.07065661996603012, "global_step": 178642, "epoch": 2007, "lr": 2.6260560684025696e-05}
{"train_loss": 0.05563925951719284, "global_step": 178643, "epoch": 2007, "lr": 2.626005045956912e-05}
{"train_loss": 0.06979669630527496, "global_step": 178644, "epoch": 2007, "lr": 2.625954023830407e-05}
{"train_loss": 0.07892628759145737, "global_step": 178645, "epoch": 2007, "lr": 2.6259030020230653e-05}
{"train_loss": 0.118306964635849, "global_step": 178646, "epoch": 2007, "lr": 2.625851980534892e-05}
{"train_loss": 0.03176301717758179, "global_step": 178647, "epoch": 2007, "lr": 2.625800959365892e-05}
{"train_loss": 0.026501024141907692, "global_step": 178648, "epoch": 2007, "lr": 2.6257499385160766e-05}
{"train_loss": 0.09609092772006989, "global_step": 178649, "epoch": 2007, "lr": 2.6256989179854473e-05}
{"train_loss": 0.046368956565856934, "global_step": 178650, "epoch": 2007, "lr": 2.625647897774016e-05}
{"train_loss": 0.08229571580886841, "global_step": 178651, "epoch": 2007, "lr": 2.6255968778817864e-05}
{"train_loss": 0.12799130380153656, "global_step": 178652, "epoch": 2007, "lr": 2.6255458583087677e-05}
{"train_loss": 0.06767488270998001, "global_step": 178653, "epoch": 2007, "lr": 2.625494839054963e-05}
{"train_loss": 0.12745459377765656, "global_step": 178654, "epoch": 2007, "lr": 2.625443820120384e-05}
{"train_loss": 0.04324117302894592, "global_step": 178655, "epoch": 2007, "lr": 2.625392801505033e-05}
{"train_loss": 0.08227858692407608, "global_step": 178656, "epoch": 2007, "lr": 2.625341783208921e-05}
{"train_loss": 0.07409203052520752, "global_step": 178657, "epoch": 2007, "lr": 2.6252907652320513e-05}
{"train_loss": 0.05552650988101959, "global_step": 178658, "epoch": 2007, "lr": 2.625239747574434e-05}
{"train_loss": 0.0992254912853241, "global_step": 178659, "epoch": 2007, "lr": 2.6251887302360723e-05}
{"train_loss": 0.059431303292512894, "global_step": 178660, "epoch": 2007, "lr": 2.625137713216977e-05}
{"train_loss": 0.08184942603111267, "global_step": 178661, "epoch": 2007, "lr": 2.6250866965171506e-05}
{"train_loss": 0.03208477050065994, "global_step": 178662, "epoch": 2007, "lr": 2.625035680136605e-05}
{"train_loss": 0.08572801202535629, "global_step": 178663, "epoch": 2007, "lr": 2.6249846640753418e-05}
{"train_loss": 0.09131571650505066, "global_step": 178664, "epoch": 2007, "lr": 2.624933648333373e-05}
{"train_loss": 0.0621604397892952, "global_step": 178665, "epoch": 2007, "lr": 2.6248826329107008e-05}
{"train_loss": 0.04752685874700546, "global_step": 178666, "epoch": 2007, "lr": 2.624831617807334e-05}
{"train_loss": 0.053358498960733414, "global_step": 178667, "epoch": 2007, "lr": 2.6247806030232814e-05}
{"train_loss": 0.06609690189361572, "global_step": 178668, "epoch": 2007, "lr": 2.6247295885585467e-05}
{"train_loss": 0.13139691948890686, "global_step": 178669, "epoch": 2007, "lr": 2.6246785744131397e-05}
{"train_loss": 0.057375580072402954, "global_step": 178670, "epoch": 2007, "lr": 2.624627560587064e-05}
{"train_loss": 0.06897729635238647, "global_step": 178671, "epoch": 2007, "lr": 2.62457654708033e-05}
{"train_loss": 0.08999928086996078, "global_step": 178672, "epoch": 2007, "lr": 2.624525533892941e-05}
{"train_loss": 0.06094735115766525, "global_step": 178673, "epoch": 2007, "lr": 2.624474521024907e-05}
{"train_loss": 0.11475475877523422, "global_step": 178674, "epoch": 2007, "lr": 2.624423508476232e-05}
{"train_loss": 0.03441260755062103, "global_step": 178675, "epoch": 2007, "lr": 2.6243724962469264e-05}
{"train_loss": 0.0602559857070446, "global_step": 178676, "epoch": 2007, "lr": 2.6243214843369924e-05}
{"train_loss": 0.04055022448301315, "global_step": 178677, "epoch": 2007, "lr": 2.6242704727464417e-05}
{"train_loss": 0.0525536984205246, "global_step": 178678, "epoch": 2007, "lr": 2.624219461475277e-05}
{"train_loss": 0.08732203394174576, "global_step": 178679, "epoch": 2007, "lr": 2.6241684505235088e-05}
{"train_loss": 0.08559782058000565, "global_step": 178680, "epoch": 2007, "lr": 2.6241174398911404e-05}
{"train_loss": 0.046211566776037216, "global_step": 178681, "epoch": 2007, "lr": 2.6240664295781824e-05}
{"train_loss": 0.043303657323122025, "global_step": 178682, "epoch": 2007, "lr": 2.6240154195846378e-05}
{"train_loss": 0.0534362718462944, "global_step": 178683, "epoch": 2007, "lr": 2.623964409910517e-05}
{"train_loss": 0.03368472680449486, "global_step": 178684, "epoch": 2007, "lr": 2.6239134005558234e-05}
{"train_loss": 0.08702205121517181, "global_step": 178685, "epoch": 2007, "lr": 2.6238623915205685e-05}
{"train_loss": 0.08341861516237259, "global_step": 178686, "epoch": 2007, "lr": 2.6238113828047552e-05}
{"train_loss": 0.037946559488773346, "global_step": 178687, "epoch": 2007, "lr": 2.6237603744083898e-05}
{"train_loss": 0.10595230013132095, "global_step": 178688, "epoch": 2007, "lr": 2.6237093663314826e-05}
{"train_loss": 0.10760548710823059, "global_step": 178689, "epoch": 2007, "lr": 2.6236583585740372e-05}
{"train_loss": 0.08182170242071152, "global_step": 178690, "epoch": 2007, "lr": 2.6236073511360638e-05}
{"train_loss": 0.07053771615028381, "global_step": 178691, "epoch": 2007, "lr": 2.623556344017566e-05}
{"train_loss": 0.07963061332702637, "global_step": 178692, "epoch": 2007, "lr": 2.6235053372185535e-05}
{"train_loss": 0.07596385478973389, "global_step": 178693, "epoch": 2007, "lr": 2.6234543307390298e-05}
{"train_loss": 0.05764451250433922, "global_step": 178694, "epoch": 2007, "lr": 2.6234033245790056e-05}
{"train_loss": 0.12342316657304764, "global_step": 178695, "epoch": 2007, "lr": 2.623352318738484e-05}
{"train_loss": 0.07441962510347366, "global_step": 178696, "epoch": 2007, "lr": 2.6233013132174755e-05}
{"train_loss": 0.0655568391084671, "global_step": 178697, "epoch": 2007, "lr": 2.6232503080159832e-05}
{"train_loss": 0.063896045088768, "global_step": 178698, "epoch": 2007, "lr": 2.6231993031340185e-05}
{"train_loss": 0.06395641714334488, "global_step": 178699, "epoch": 2007, "lr": 2.6231482985715827e-05}
{"train_loss": 0.12623879313468933, "global_step": 178700, "epoch": 2007, "lr": 2.6230972943286885e-05}
{"train_loss": 0.10687834769487381, "global_step": 178701, "epoch": 2007, "lr": 2.6230462904053377e-05}
{"train_loss": 0.060744285583496094, "global_step": 178702, "epoch": 2007, "lr": 2.6229952868015397e-05}
{"train_loss": 0.057036157697439194, "global_step": 178703, "epoch": 2007, "lr": 2.6229442835173017e-05}
{"train_loss": 0.04112495854496956, "global_step": 178704, "epoch": 2007, "lr": 2.622893280552629e-05}
{"train_loss": 0.10679689794778824, "global_step": 178705, "epoch": 2007, "lr": 2.6228422779075312e-05}
{"train_loss": 0.0715356320142746, "global_step": 178706, "epoch": 2007, "lr": 2.6227912755820112e-05}
{"train_loss": 0.05204996466636658, "global_step": 178707, "epoch": 2007, "lr": 2.62274027357608e-05}
{"train_loss": 0.06149189546704292, "global_step": 178708, "epoch": 2007, "lr": 2.62268927188974e-05}
{"train_loss": 0.06883668899536133, "global_step": 178709, "epoch": 2007, "lr": 2.6226382705230024e-05}
{"train_loss": 0.09885846078395844, "global_step": 178710, "epoch": 2007, "lr": 2.6225872694758706e-05}
{"train_loss": 0.07584168381068145, "global_step": 178711, "epoch": 2007, "lr": 2.6225362687483545e-05, "val_loss": 7.500081539154053}
{"train_loss": 0.06365993618965149, "global_step": 178712, "epoch": 2008, "lr": 2.622485268340458e-05}
{"train_loss": 0.05904949828982353, "global_step": 178713, "epoch": 2008, "lr": 2.6224342682521907e-05}
{"train_loss": 0.06387300789356232, "global_step": 178714, "epoch": 2008, "lr": 2.622383268483557e-05}
{"train_loss": 0.10329475998878479, "global_step": 178715, "epoch": 2008, "lr": 2.6223322690345666e-05}
{"train_loss": 0.08871211856603622, "global_step": 178716, "epoch": 2008, "lr": 2.6222812699052224e-05}
{"train_loss": 0.08408757299184799, "global_step": 178717, "epoch": 2008, "lr": 2.6222302710955353e-05}
{"train_loss": 0.10223423689603806, "global_step": 178718, "epoch": 2008, "lr": 2.6221792726055082e-05}
{"train_loss": 0.03034527413547039, "global_step": 178719, "epoch": 2008, "lr": 2.6221282744351527e-05}
{"train_loss": 0.14036163687705994, "global_step": 178720, "epoch": 2008, "lr": 2.6220772765844715e-05}
{"train_loss": 0.0903494730591774, "global_step": 178721, "epoch": 2008, "lr": 2.6220262790534745e-05}
{"train_loss": 0.06983163207769394, "global_step": 178722, "epoch": 2008, "lr": 2.621975281842165e-05}
{"train_loss": 0.048464205116033554, "global_step": 178723, "epoch": 2008, "lr": 2.6219242849505538e-05}
{"train_loss": 0.025615263730287552, "global_step": 178724, "epoch": 2008, "lr": 2.6218732883786457e-05}
{"train_loss": 0.05632761865854263, "global_step": 178725, "epoch": 2008, "lr": 2.621822292126446e-05}
{"train_loss": 0.08204952627420425, "global_step": 178726, "epoch": 2008, "lr": 2.6217712961939656e-05}
{"train_loss": 0.08396083116531372, "global_step": 178727, "epoch": 2008, "lr": 2.6217203005812068e-05}
{"train_loss": 0.06474007666110992, "global_step": 178728, "epoch": 2008, "lr": 2.6216693052881812e-05}
{"train_loss": 0.1027049720287323, "global_step": 178729, "epoch": 2008, "lr": 2.6216183103148906e-05}
{"train_loss": 0.037297431379556656, "global_step": 178730, "epoch": 2008, "lr": 2.6215673156613464e-05}
{"train_loss": 0.10031767189502716, "global_step": 178731, "epoch": 2008, "lr": 2.6215163213275518e-05}
{"train_loss": 0.1452343761920929, "global_step": 178732, "epoch": 2008, "lr": 2.621465327313517e-05}
{"train_loss": 0.063398577272892, "global_step": 178733, "epoch": 2008, "lr": 2.6214143336192453e-05}
{"train_loss": 0.07322247326374054, "global_step": 178734, "epoch": 2008, "lr": 2.621363340244748e-05}
{"train_loss": 0.04817443713545799, "global_step": 178735, "epoch": 2008, "lr": 2.621312347190027e-05}
{"train_loss": 0.06933661550283432, "global_step": 178736, "epoch": 2008, "lr": 2.6212613544550935e-05}
{"train_loss": 0.09901255369186401, "global_step": 178737, "epoch": 2008, "lr": 2.621210362039951e-05}
{"train_loss": 0.09547768533229828, "global_step": 178738, "epoch": 2008, "lr": 2.6211593699446068e-05}
{"train_loss": 0.08050120621919632, "global_step": 178739, "epoch": 2008, "lr": 2.621108378169071e-05}
{"train_loss": 0.07188133150339127, "global_step": 178740, "epoch": 2008, "lr": 2.6210573867133465e-05}
{"train_loss": 0.09657087922096252, "global_step": 178741, "epoch": 2008, "lr": 2.6210063955774434e-05}
{"train_loss": 0.06024256348609924, "global_step": 178742, "epoch": 2008, "lr": 2.6209554047613655e-05}
{"train_loss": 0.06882323324680328, "global_step": 178743, "epoch": 2008, "lr": 2.620904414265123e-05}
{"train_loss": 0.030399475246667862, "global_step": 178744, "epoch": 2008, "lr": 2.6208534240887183e-05}
{"train_loss": 0.08659660816192627, "global_step": 178745, "epoch": 2008, "lr": 2.6208024342321636e-05}
{"train_loss": 0.03961401805281639, "global_step": 178746, "epoch": 2008, "lr": 2.6207514446954605e-05}
{"train_loss": 0.03828931599855423, "global_step": 178747, "epoch": 2008, "lr": 2.6207004554786214e-05}
{"train_loss": 0.09794547408819199, "global_step": 178748, "epoch": 2008, "lr": 2.6206494665816473e-05}
{"train_loss": 0.1040901243686676, "global_step": 178749, "epoch": 2008, "lr": 2.62059847800455e-05}
{"train_loss": 0.08980502188205719, "global_step": 178750, "epoch": 2008, "lr": 2.620547489747332e-05}
{"train_loss": 0.1008659228682518, "global_step": 178751, "epoch": 2008, "lr": 2.620496501810005e-05}
{"train_loss": 0.08958674967288971, "global_step": 178752, "epoch": 2008, "lr": 2.6204455141925703e-05}
{"train_loss": 0.1075410321354866, "global_step": 178753, "epoch": 2008, "lr": 2.6203945268950404e-05}
{"train_loss": 0.06632544100284576, "global_step": 178754, "epoch": 2008, "lr": 2.620343539917417e-05}
{"train_loss": 0.08603673428297043, "global_step": 178755, "epoch": 2008, "lr": 2.6202925532597124e-05}
{"train_loss": 0.08265238255262375, "global_step": 178756, "epoch": 2008, "lr": 2.620241566921927e-05}
{"train_loss": 0.06868789345026016, "global_step": 178757, "epoch": 2008, "lr": 2.6201905809040746e-05}
{"train_loss": 0.08281648904085159, "global_step": 178758, "epoch": 2008, "lr": 2.620139595206156e-05}
{"train_loss": 0.1215527355670929, "global_step": 178759, "epoch": 2008, "lr": 2.620088609828182e-05}
{"train_loss": 0.09522814303636551, "global_step": 178760, "epoch": 2008, "lr": 2.6200376247701563e-05}
{"train_loss": 0.08588765561580658, "global_step": 178761, "epoch": 2008, "lr": 2.6199866400320903e-05}
{"train_loss": 0.043085817247629166, "global_step": 178762, "epoch": 2008, "lr": 2.6199356556139853e-05}
{"train_loss": 0.08633732795715332, "global_step": 178763, "epoch": 2008, "lr": 2.6198846715158533e-05}
{"train_loss": 0.09333831071853638, "global_step": 178764, "epoch": 2008, "lr": 2.6198336877376982e-05}
{"train_loss": 0.08055146783590317, "global_step": 178765, "epoch": 2008, "lr": 2.6197827042795265e-05}
{"train_loss": 0.08871994912624359, "global_step": 178766, "epoch": 2008, "lr": 2.6197317211413475e-05}
{"train_loss": 0.07069261372089386, "global_step": 178767, "epoch": 2008, "lr": 2.6196807383231646e-05}
{"train_loss": 0.06428604573011398, "global_step": 178768, "epoch": 2008, "lr": 2.619629755824988e-05}
{"train_loss": 0.0554683580994606, "global_step": 178769, "epoch": 2008, "lr": 2.6195787736468224e-05}
{"train_loss": 0.0659032091498375, "global_step": 178770, "epoch": 2008, "lr": 2.6195277917886762e-05}
{"train_loss": 0.07880214601755142, "global_step": 178771, "epoch": 2008, "lr": 2.619476810250554e-05}
{"train_loss": 0.07950249314308167, "global_step": 178772, "epoch": 2008, "lr": 2.6194258290324662e-05}
{"train_loss": 0.06960839033126831, "global_step": 178773, "epoch": 2008, "lr": 2.619374848134416e-05}
{"train_loss": 0.08566591143608093, "global_step": 178774, "epoch": 2008, "lr": 2.6193238675564113e-05}
{"train_loss": 0.0823311135172844, "global_step": 178775, "epoch": 2008, "lr": 2.619272887298461e-05}
{"train_loss": 0.15074746310710907, "global_step": 178776, "epoch": 2008, "lr": 2.6192219073605694e-05}
{"train_loss": 0.06008774787187576, "global_step": 178777, "epoch": 2008, "lr": 2.6191709277427458e-05}
{"train_loss": 0.08194057643413544, "global_step": 178778, "epoch": 2008, "lr": 2.6191199484449935e-05}
{"train_loss": 0.10366816073656082, "global_step": 178779, "epoch": 2008, "lr": 2.619068969467324e-05}
{"train_loss": 0.13810831308364868, "global_step": 178780, "epoch": 2008, "lr": 2.61901799080974e-05}
{"train_loss": 0.09320160746574402, "global_step": 178781, "epoch": 2008, "lr": 2.618967012472251e-05}
{"train_loss": 0.09768274426460266, "global_step": 178782, "epoch": 2008, "lr": 2.6189160344548612e-05}
{"train_loss": 0.026064584031701088, "global_step": 178783, "epoch": 2008, "lr": 2.6188650567575818e-05}
{"train_loss": 0.08692487329244614, "global_step": 178784, "epoch": 2008, "lr": 2.6188140793804144e-05}
{"train_loss": 0.07243337482213974, "global_step": 178785, "epoch": 2008, "lr": 2.6187631023233705e-05}
{"train_loss": 0.07717664539813995, "global_step": 178786, "epoch": 2008, "lr": 2.6187121255864526e-05}
{"train_loss": 0.06724728643894196, "global_step": 178787, "epoch": 2008, "lr": 2.618661149169672e-05}
{"train_loss": 0.04510830342769623, "global_step": 178788, "epoch": 2008, "lr": 2.618610173073032e-05}
{"train_loss": 0.03096134029328823, "global_step": 178789, "epoch": 2008, "lr": 2.618559197296542e-05}
{"train_loss": 0.09977952390909195, "global_step": 178790, "epoch": 2008, "lr": 2.6185082218402057e-05}
{"train_loss": 0.12134275585412979, "global_step": 178791, "epoch": 2008, "lr": 2.6184572467040347e-05}
{"train_loss": 0.09827534854412079, "global_step": 178792, "epoch": 2008, "lr": 2.618406271888031e-05}
{"train_loss": 0.11334645748138428, "global_step": 178793, "epoch": 2008, "lr": 2.6183552973922027e-05}
{"train_loss": 0.07527086138725281, "global_step": 178794, "epoch": 2008, "lr": 2.61830432321656e-05}
{"train_loss": 0.06033460795879364, "global_step": 178795, "epoch": 2008, "lr": 2.618253349361105e-05}
{"train_loss": 0.08984602987766266, "global_step": 178796, "epoch": 2008, "lr": 2.618202375825849e-05}
{"train_loss": 0.019060766324400902, "global_step": 178797, "epoch": 2008, "lr": 2.6181514026107966e-05}
{"train_loss": 0.1166396364569664, "global_step": 178798, "epoch": 2008, "lr": 2.6181004297159534e-05}
{"train_loss": 0.11372504383325577, "global_step": 178799, "epoch": 2008, "lr": 2.618049457141329e-05}
{"train_loss": 0.07939373957139723, "global_step": 178800, "epoch": 2008, "lr": 2.6179984848869267e-05, "val_loss": 7.297505855560303}
{"train_loss": 0.07208789139986038, "global_step": 178801, "epoch": 2009, "lr": 2.6179475129527574e-05}
{"train_loss": 0.11577649414539337, "global_step": 178802, "epoch": 2009, "lr": 2.6178965413388262e-05}
{"train_loss": 0.04078920558094978, "global_step": 178803, "epoch": 2009, "lr": 2.6178455700451375e-05}
{"train_loss": 0.10873189568519592, "global_step": 178804, "epoch": 2009, "lr": 2.6177945990717035e-05}
{"train_loss": 0.08367656916379929, "global_step": 178805, "epoch": 2009, "lr": 2.617743628418525e-05}
{"train_loss": 0.08696837723255157, "global_step": 178806, "epoch": 2009, "lr": 2.6176926580856138e-05}
{"train_loss": 0.05543234571814537, "global_step": 178807, "epoch": 2009, "lr": 2.6176416880729733e-05}
{"train_loss": 0.09342119842767715, "global_step": 178808, "epoch": 2009, "lr": 2.6175907183806136e-05}
{"train_loss": 0.08507056534290314, "global_step": 178809, "epoch": 2009, "lr": 2.6175397490085374e-05}
{"train_loss": 0.07135333865880966, "global_step": 178810, "epoch": 2009, "lr": 2.617488779956756e-05}
{"train_loss": 0.10492290556430817, "global_step": 178811, "epoch": 2009, "lr": 2.6174378112252727e-05}
{"train_loss": 0.09408100694417953, "global_step": 178812, "epoch": 2009, "lr": 2.6173868428140958e-05}
{"train_loss": 0.0881720706820488, "global_step": 178813, "epoch": 2009, "lr": 2.617335874723234e-05}
{"train_loss": 0.023702122271060944, "global_step": 178814, "epoch": 2009, "lr": 2.6172849069526906e-05}
{"train_loss": 0.1381196230649948, "global_step": 178815, "epoch": 2009, "lr": 2.617233939502476e-05}
{"train_loss": 0.0711471363902092, "global_step": 178816, "epoch": 2009, "lr": 2.6171829723725928e-05}
{"train_loss": 0.05963042378425598, "global_step": 178817, "epoch": 2009, "lr": 2.617132005563053e-05}
{"train_loss": 0.09923848509788513, "global_step": 178818, "epoch": 2009, "lr": 2.6170810390738586e-05}
{"train_loss": 0.099288709461689, "global_step": 178819, "epoch": 2009, "lr": 2.61703007290502e-05}
{"train_loss": 0.04111624136567116, "global_step": 178820, "epoch": 2009, "lr": 2.6169791070565414e-05}
{"train_loss": 0.11912097781896591, "global_step": 178821, "epoch": 2009, "lr": 2.616928141528433e-05}
{"train_loss": 0.04387832432985306, "global_step": 178822, "epoch": 2009, "lr": 2.616877176320697e-05}
{"train_loss": 0.025644347071647644, "global_step": 178823, "epoch": 2009, "lr": 2.6168262114333453e-05}
{"train_loss": 0.08255772292613983, "global_step": 178824, "epoch": 2009, "lr": 2.6167752468663796e-05}
{"train_loss": 0.1102200448513031, "global_step": 178825, "epoch": 2009, "lr": 2.6167242826198123e-05}
{"train_loss": 0.07009315490722656, "global_step": 178826, "epoch": 2009, "lr": 2.616673318693645e-05}
{"train_loss": 0.0848117247223854, "global_step": 178827, "epoch": 2009, "lr": 2.616622355087889e-05}
{"train_loss": 0.055790361016988754, "global_step": 178828, "epoch": 2009, "lr": 2.6165713918025463e-05}
{"train_loss": 0.1415344476699829, "global_step": 178829, "epoch": 2009, "lr": 2.6165204288376276e-05}
{"train_loss": 0.1547919660806656, "global_step": 178830, "epoch": 2009, "lr": 2.61646946619314e-05}
{"train_loss": 0.10556290298700333, "global_step": 178831, "epoch": 2009, "lr": 2.616418503869087e-05}
{"train_loss": 0.03796619176864624, "global_step": 178832, "epoch": 2009, "lr": 2.61636754186548e-05}
{"train_loss": 0.03630660101771355, "global_step": 178833, "epoch": 2009, "lr": 2.616316580182321e-05}
{"train_loss": 0.1075742170214653, "global_step": 178834, "epoch": 2009, "lr": 2.6162656188196212e-05}
{"train_loss": 0.04634831100702286, "global_step": 178835, "epoch": 2009, "lr": 2.6162146577773856e-05}
{"train_loss": 0.07971332967281342, "global_step": 178836, "epoch": 2009, "lr": 2.616163697055619e-05}
{"train_loss": 0.07907705008983612, "global_step": 178837, "epoch": 2009, "lr": 2.6161127366543313e-05}
{"train_loss": 0.04395129904150963, "global_step": 178838, "epoch": 2009, "lr": 2.6160617765735272e-05}
{"train_loss": 0.05115485563874245, "global_step": 178839, "epoch": 2009, "lr": 2.616010816813216e-05}
{"train_loss": 0.07899460196495056, "global_step": 178840, "epoch": 2009, "lr": 2.6159598573734034e-05}
{"train_loss": 0.08274239301681519, "global_step": 178841, "epoch": 2009, "lr": 2.6159088982540935e-05}
{"train_loss": 0.1020275205373764, "global_step": 178842, "epoch": 2009, "lr": 2.6158579394552985e-05}
{"train_loss": 0.06530013680458069, "global_step": 178843, "epoch": 2009, "lr": 2.615806980977019e-05}
{"train_loss": 0.1297454833984375, "global_step": 178844, "epoch": 2009, "lr": 2.6157560228192677e-05}
{"train_loss": 0.05956953391432762, "global_step": 178845, "epoch": 2009, "lr": 2.615705064982048e-05}
{"train_loss": 0.1017410010099411, "global_step": 178846, "epoch": 2009, "lr": 2.6156541074653684e-05}
{"train_loss": 0.10662249475717545, "global_step": 178847, "epoch": 2009, "lr": 2.6156031502692336e-05}
{"train_loss": 0.08867501467466354, "global_step": 178848, "epoch": 2009, "lr": 2.6155521933936516e-05}
{"train_loss": 0.04098128899931908, "global_step": 178849, "epoch": 2009, "lr": 2.6155012368386317e-05}
{"train_loss": 0.061813950538635254, "global_step": 178850, "epoch": 2009, "lr": 2.615450280604177e-05}
{"train_loss": 0.1074848473072052, "global_step": 178851, "epoch": 2009, "lr": 2.6153993246902974e-05}
{"train_loss": 0.10747456550598145, "global_step": 178852, "epoch": 2009, "lr": 2.615348369096996e-05}
{"train_loss": 0.034186460077762604, "global_step": 178853, "epoch": 2009, "lr": 2.615297413824285e-05}
{"train_loss": 0.1193295419216156, "global_step": 178854, "epoch": 2009, "lr": 2.6152464588721653e-05}
{"train_loss": 0.052580010145902634, "global_step": 178855, "epoch": 2009, "lr": 2.615195504240649e-05}
{"train_loss": 0.06902031600475311, "global_step": 178856, "epoch": 2009, "lr": 2.615144549929739e-05}
{"train_loss": 0.08896910399198532, "global_step": 178857, "epoch": 2009, "lr": 2.6150935959394453e-05}
{"train_loss": 0.08579280227422714, "global_step": 178858, "epoch": 2009, "lr": 2.6150426422697705e-05}
{"train_loss": 0.05158700421452522, "global_step": 178859, "epoch": 2009, "lr": 2.6149916889207272e-05}
{"train_loss": 0.11484497040510178, "global_step": 178860, "epoch": 2009, "lr": 2.6149407358923167e-05}
{"train_loss": 0.08134154975414276, "global_step": 178861, "epoch": 2009, "lr": 2.6148897831845502e-05}
{"train_loss": 0.08972001820802689, "global_step": 178862, "epoch": 2009, "lr": 2.6148388307974314e-05}
{"train_loss": 0.0838623121380806, "global_step": 178863, "epoch": 2009, "lr": 2.61478787873097e-05}
{"train_loss": 0.05933135002851486, "global_step": 178864, "epoch": 2009, "lr": 2.6147369269851695e-05}
{"train_loss": 0.037149347364902496, "global_step": 178865, "epoch": 2009, "lr": 2.6146859755600382e-05}
{"train_loss": 0.058600980788469315, "global_step": 178866, "epoch": 2009, "lr": 2.614635024455585e-05}
{"train_loss": 0.06878092139959335, "global_step": 178867, "epoch": 2009, "lr": 2.614584073671814e-05}
{"train_loss": 0.09508691728115082, "global_step": 178868, "epoch": 2009, "lr": 2.6145331232087346e-05}
{"train_loss": 0.06198412552475929, "global_step": 178869, "epoch": 2009, "lr": 2.6144821730663504e-05}
{"train_loss": 0.0788022130727768, "global_step": 178870, "epoch": 2009, "lr": 2.614431223244671e-05}
{"train_loss": 0.10382326692342758, "global_step": 178871, "epoch": 2009, "lr": 2.614380273743702e-05}
{"train_loss": 0.050258081406354904, "global_step": 178872, "epoch": 2009, "lr": 2.6143293245634508e-05}
{"train_loss": 0.03911035507917404, "global_step": 178873, "epoch": 2009, "lr": 2.6142783757039247e-05}
{"train_loss": 0.10677983611822128, "global_step": 178874, "epoch": 2009, "lr": 2.614227427165128e-05}
{"train_loss": 0.07633057981729507, "global_step": 178875, "epoch": 2009, "lr": 2.6141764789470708e-05}
{"train_loss": 0.05912495404481888, "global_step": 178876, "epoch": 2009, "lr": 2.614125531049757e-05}
{"train_loss": 0.069340281188488, "global_step": 178877, "epoch": 2009, "lr": 2.614074583473196e-05}
{"train_loss": 0.07389168441295624, "global_step": 178878, "epoch": 2009, "lr": 2.614023636217393e-05}
{"train_loss": 0.08026103675365448, "global_step": 178879, "epoch": 2009, "lr": 2.613972689282357e-05}
{"train_loss": 0.08311813324689865, "global_step": 178880, "epoch": 2009, "lr": 2.613921742668093e-05}
{"train_loss": 0.06604229658842087, "global_step": 178881, "epoch": 2009, "lr": 2.6138707963746058e-05}
{"train_loss": 0.04043526202440262, "global_step": 178882, "epoch": 2009, "lr": 2.6138198504019072e-05}
{"train_loss": 0.05538489669561386, "global_step": 178883, "epoch": 2009, "lr": 2.613768904749999e-05}
{"train_loss": 0.0704212635755539, "global_step": 178884, "epoch": 2009, "lr": 2.613717959418891e-05}
{"train_loss": 0.04742059484124184, "global_step": 178885, "epoch": 2009, "lr": 2.6136670144085906e-05}
{"train_loss": 0.06514888256788254, "global_step": 178886, "epoch": 2009, "lr": 2.6136160697191026e-05}
{"train_loss": 0.07503167539834976, "global_step": 178887, "epoch": 2009, "lr": 2.6135651253504368e-05}
{"train_loss": 0.04172702506184578, "global_step": 178888, "epoch": 2009, "lr": 2.6135141813025955e-05}
{"train_loss": 0.07750971858086211, "global_step": 178889, "epoch": 2009, "lr": 2.61346323757559e-05, "val_loss": 7.4735283851623535}
{"train_loss": 0.06603178381919861, "global_step": 178890, "epoch": 2010, "lr": 2.613412294169424e-05}
{"train_loss": 0.09789226949214935, "global_step": 178891, "epoch": 2010, "lr": 2.613361351084107e-05}
{"train_loss": 0.12650315463542938, "global_step": 178892, "epoch": 2010, "lr": 2.6133104083196426e-05}
{"train_loss": 0.056058794260025024, "global_step": 178893, "epoch": 2010, "lr": 2.613259465876042e-05}
{"train_loss": 0.06241053715348244, "global_step": 178894, "epoch": 2010, "lr": 2.6132085237533072e-05}
{"train_loss": 0.0566890724003315, "global_step": 178895, "epoch": 2010, "lr": 2.6131575819514497e-05}
{"train_loss": 0.04780568182468414, "global_step": 178896, "epoch": 2010, "lr": 2.6131066404704717e-05}
{"train_loss": 0.09932549297809601, "global_step": 178897, "epoch": 2010, "lr": 2.6130556993103843e-05}
{"train_loss": 0.10161857306957245, "global_step": 178898, "epoch": 2010, "lr": 2.613004758471191e-05}
{"train_loss": 0.09179795533418655, "global_step": 178899, "epoch": 2010, "lr": 2.6129538179529018e-05}
{"train_loss": 0.046118710190057755, "global_step": 178900, "epoch": 2010, "lr": 2.61290287775552e-05}
{"train_loss": 0.08130471408367157, "global_step": 178901, "epoch": 2010, "lr": 2.612851937879054e-05}
{"train_loss": 0.12694428861141205, "global_step": 178902, "epoch": 2010, "lr": 2.6128009983235136e-05}
{"train_loss": 0.06249936670064926, "global_step": 178903, "epoch": 2010, "lr": 2.6127500590889005e-05}
{"train_loss": 0.09869732707738876, "global_step": 178904, "epoch": 2010, "lr": 2.612699120175226e-05}
{"train_loss": 0.07683227956295013, "global_step": 178905, "epoch": 2010, "lr": 2.6126481815824932e-05}
{"train_loss": 0.0838572308421135, "global_step": 178906, "epoch": 2010, "lr": 2.6125972433107126e-05}
{"train_loss": 0.054629094898700714, "global_step": 178907, "epoch": 2010, "lr": 2.6125463053598874e-05}
{"train_loss": 0.09442286193370819, "global_step": 178908, "epoch": 2010, "lr": 2.612495367730028e-05}
{"train_loss": 0.0940576121211052, "global_step": 178909, "epoch": 2010, "lr": 2.6124444304211382e-05}
{"train_loss": 0.07756040245294571, "global_step": 178910, "epoch": 2010, "lr": 2.612393493433227e-05}
{"train_loss": 0.1083318218588829, "global_step": 178911, "epoch": 2010, "lr": 2.6123425567662995e-05}
{"train_loss": 0.08569134026765823, "global_step": 178912, "epoch": 2010, "lr": 2.6122916204203645e-05}
{"train_loss": 0.03589855507016182, "global_step": 178913, "epoch": 2010, "lr": 2.6122406843954283e-05}
{"train_loss": 0.11344590783119202, "global_step": 178914, "epoch": 2010, "lr": 2.6121897486914955e-05}
{"train_loss": 0.029736028984189034, "global_step": 178915, "epoch": 2010, "lr": 2.612138813308576e-05}
{"train_loss": 0.060006145387887955, "global_step": 178916, "epoch": 2010, "lr": 2.612087878246674e-05}
{"train_loss": 0.08751804381608963, "global_step": 178917, "epoch": 2010, "lr": 2.612036943505799e-05}
{"train_loss": 0.09218832850456238, "global_step": 178918, "epoch": 2010, "lr": 2.6119860090859572e-05}
{"train_loss": 0.1341429501771927, "global_step": 178919, "epoch": 2010, "lr": 2.6119350749871525e-05}
{"train_loss": 0.11619864404201508, "global_step": 178920, "epoch": 2010, "lr": 2.6118841412093943e-05}
{"train_loss": 0.04329158365726471, "global_step": 178921, "epoch": 2010, "lr": 2.6118332077526908e-05}
{"train_loss": 0.08583461493253708, "global_step": 178922, "epoch": 2010, "lr": 2.6117822746170452e-05}
{"train_loss": 0.06698925048112869, "global_step": 178923, "epoch": 2010, "lr": 2.611731341802468e-05}
{"train_loss": 0.03782926872372627, "global_step": 178924, "epoch": 2010, "lr": 2.611680409308963e-05}
{"train_loss": 0.08319053053855896, "global_step": 178925, "epoch": 2010, "lr": 2.6116294771365403e-05}
{"train_loss": 0.08530602604150772, "global_step": 178926, "epoch": 2010, "lr": 2.6115785452852027e-05}
{"train_loss": 0.0427217073738575, "global_step": 178927, "epoch": 2010, "lr": 2.6115276137549617e-05}
{"train_loss": 0.06401122361421585, "global_step": 178928, "epoch": 2010, "lr": 2.611476682545819e-05}
{"train_loss": 0.12160266190767288, "global_step": 178929, "epoch": 2010, "lr": 2.6114257516577873e-05}
{"train_loss": 0.061231765896081924, "global_step": 178930, "epoch": 2010, "lr": 2.611374821090867e-05}
{"train_loss": 0.07066531479358673, "global_step": 178931, "epoch": 2010, "lr": 2.611323890845071e-05}
{"train_loss": 0.07810325920581818, "global_step": 178932, "epoch": 2010, "lr": 2.6112729609204016e-05}
{"train_loss": 0.1594679355621338, "global_step": 178933, "epoch": 2010, "lr": 2.611222031316869e-05}
{"train_loss": 0.09256888180971146, "global_step": 178934, "epoch": 2010, "lr": 2.6111711020344765e-05}
{"train_loss": 0.08154188096523285, "global_step": 178935, "epoch": 2010, "lr": 2.611120173073235e-05}
{"train_loss": 0.07925306260585785, "global_step": 178936, "epoch": 2010, "lr": 2.6110692444331473e-05}
{"train_loss": 0.0770469382405281, "global_step": 178937, "epoch": 2010, "lr": 2.6110183161142244e-05}
{"train_loss": 0.0683627650141716, "global_step": 178938, "epoch": 2010, "lr": 2.6109673881164687e-05}
{"train_loss": 0.042117029428482056, "global_step": 178939, "epoch": 2010, "lr": 2.6109164604398895e-05}
{"train_loss": 0.0959990993142128, "global_step": 178940, "epoch": 2010, "lr": 2.6108655330844957e-05}
{"train_loss": 0.029972536489367485, "global_step": 178941, "epoch": 2010, "lr": 2.610814606050289e-05}
{"train_loss": 0.0715876966714859, "global_step": 178942, "epoch": 2010, "lr": 2.610763679337282e-05}
{"train_loss": 0.0672386884689331, "global_step": 178943, "epoch": 2010, "lr": 2.6107127529454766e-05}
{"train_loss": 0.11394570022821426, "global_step": 178944, "epoch": 2010, "lr": 2.6106618268748834e-05}
{"train_loss": 0.05754637345671654, "global_step": 178945, "epoch": 2010, "lr": 2.6106109011255057e-05}
{"train_loss": 0.09402582049369812, "global_step": 178946, "epoch": 2010, "lr": 2.610559975697354e-05}
{"train_loss": 0.07871559262275696, "global_step": 178947, "epoch": 2010, "lr": 2.6105090505904317e-05}
{"train_loss": 0.10654742270708084, "global_step": 178948, "epoch": 2010, "lr": 2.61045812580475e-05}
{"train_loss": 0.048124149441719055, "global_step": 178949, "epoch": 2010, "lr": 2.6104072013403103e-05}
{"train_loss": 0.08546438068151474, "global_step": 178950, "epoch": 2010, "lr": 2.610356277197124e-05}
{"train_loss": 0.08432482182979584, "global_step": 178951, "epoch": 2010, "lr": 2.6103053533751963e-05}
{"train_loss": 0.08966405689716339, "global_step": 178952, "epoch": 2010, "lr": 2.6102544298745325e-05}
{"train_loss": 0.06497959047555923, "global_step": 178953, "epoch": 2010, "lr": 2.6102035066951423e-05}
{"train_loss": 0.10745954513549805, "global_step": 178954, "epoch": 2010, "lr": 2.6101525838370293e-05}
{"train_loss": 0.09576350450515747, "global_step": 178955, "epoch": 2010, "lr": 2.6101016613002044e-05}
{"train_loss": 0.07116016745567322, "global_step": 178956, "epoch": 2010, "lr": 2.6100507390846695e-05}
{"train_loss": 0.1293560117483139, "global_step": 178957, "epoch": 2010, "lr": 2.6099998171904373e-05}
{"train_loss": 0.08527716249227524, "global_step": 178958, "epoch": 2010, "lr": 2.6099488956175088e-05}
{"train_loss": 0.053345613181591034, "global_step": 178959, "epoch": 2010, "lr": 2.6098979743658954e-05}
{"train_loss": 0.09596280753612518, "global_step": 178960, "epoch": 2010, "lr": 2.6098470534356002e-05}
{"train_loss": 0.11000658571720123, "global_step": 178961, "epoch": 2010, "lr": 2.609796132826634e-05}
{"train_loss": 0.06837344914674759, "global_step": 178962, "epoch": 2010, "lr": 2.6097452125389998e-05}
{"train_loss": 0.09363386780023575, "global_step": 178963, "epoch": 2010, "lr": 2.6096942925727074e-05}
{"train_loss": 0.0721440538764, "global_step": 178964, "epoch": 2010, "lr": 2.6096433729277614e-05}
{"train_loss": 0.09294969588518143, "global_step": 178965, "epoch": 2010, "lr": 2.6095924536041715e-05}
{"train_loss": 0.06256350129842758, "global_step": 178966, "epoch": 2010, "lr": 2.6095415346019402e-05}
{"train_loss": 0.05622369796037674, "global_step": 178967, "epoch": 2010, "lr": 2.6094906159210796e-05}
{"train_loss": 0.07806304097175598, "global_step": 178968, "epoch": 2010, "lr": 2.6094396975615914e-05}
{"train_loss": 0.0885300263762474, "global_step": 178969, "epoch": 2010, "lr": 2.6093887795234872e-05}
{"train_loss": 0.06760798394680023, "global_step": 178970, "epoch": 2010, "lr": 2.6093378618067693e-05}
{"train_loss": 0.06898362934589386, "global_step": 178971, "epoch": 2010, "lr": 2.6092869444114486e-05}
{"train_loss": 0.073881596326828, "global_step": 178972, "epoch": 2010, "lr": 2.6092360273375288e-05}
{"train_loss": 0.15447606146335602, "global_step": 178973, "epoch": 2010, "lr": 2.6091851105850197e-05}
{"train_loss": 0.04516800120472908, "global_step": 178974, "epoch": 2010, "lr": 2.6091341941539243e-05}
{"train_loss": 0.06295386701822281, "global_step": 178975, "epoch": 2010, "lr": 2.609083278044252e-05}
{"train_loss": 0.07365331798791885, "global_step": 178976, "epoch": 2010, "lr": 2.609032362256011e-05}
{"train_loss": 0.06987877190113068, "global_step": 178977, "epoch": 2010, "lr": 2.608981446789205e-05}
{"train_loss": 0.08029114287555887, "global_step": 178978, "epoch": 2010, "lr": 2.608930531643844e-05, "val_loss": 7.5375261306762695, "train_action_mse_error": 30.055946350097656}
{"train_loss": 0.08375688642263412, "global_step": 178979, "epoch": 2011, "lr": 2.608879616819931e-05}
{"train_loss": 0.14605014026165009, "global_step": 178980, "epoch": 2011, "lr": 2.6088287023174774e-05}
{"train_loss": 0.06757361441850662, "global_step": 178981, "epoch": 2011, "lr": 2.608777788136485e-05}
{"train_loss": 0.0730288028717041, "global_step": 178982, "epoch": 2011, "lr": 2.6087268742769655e-05}
{"train_loss": 0.08676701039075851, "global_step": 178983, "epoch": 2011, "lr": 2.6086759607389217e-05}
{"train_loss": 0.10995209217071533, "global_step": 178984, "epoch": 2011, "lr": 2.608625047522364e-05}
{"train_loss": 0.05359089747071266, "global_step": 178985, "epoch": 2011, "lr": 2.6085741346272963e-05}
{"train_loss": 0.07720768451690674, "global_step": 178986, "epoch": 2011, "lr": 2.6085232220537282e-05}
{"train_loss": 0.039540547877550125, "global_step": 178987, "epoch": 2011, "lr": 2.6084723098016627e-05}
{"train_loss": 0.08507703244686127, "global_step": 178988, "epoch": 2011, "lr": 2.608421397871111e-05}
{"train_loss": 0.07211881130933762, "global_step": 178989, "epoch": 2011, "lr": 2.608370486262078e-05}
{"train_loss": 0.03713134303689003, "global_step": 178990, "epoch": 2011, "lr": 2.608319574974568e-05}
{"train_loss": 0.039588697254657745, "global_step": 178991, "epoch": 2011, "lr": 2.608268664008593e-05}
{"train_loss": 0.06690215319395065, "global_step": 178992, "epoch": 2011, "lr": 2.608217753364155e-05}
{"train_loss": 0.1267344206571579, "global_step": 178993, "epoch": 2011, "lr": 2.6081668430412647e-05}
{"train_loss": 0.12800544500350952, "global_step": 178994, "epoch": 2011, "lr": 2.6081159330399253e-05}
{"train_loss": 0.06019236892461777, "global_step": 178995, "epoch": 2011, "lr": 2.6080650233601477e-05}
{"train_loss": 0.09055175632238388, "global_step": 178996, "epoch": 2011, "lr": 2.6080141140019347e-05}
{"train_loss": 0.1326332688331604, "global_step": 178997, "epoch": 2011, "lr": 2.6079632049652967e-05}
{"train_loss": 0.09100136905908585, "global_step": 178998, "epoch": 2011, "lr": 2.6079122962502372e-05}
{"train_loss": 0.035006117075681686, "global_step": 178999, "epoch": 2011, "lr": 2.607861387856767e-05}
{"train_loss": 0.06427235156297684, "global_step": 179000, "epoch": 2011, "lr": 2.607810479784888e-05}
{"train_loss": 0.06964264810085297, "global_step": 179001, "epoch": 2011, "lr": 2.607759572034612e-05}
{"train_loss": 0.08038396388292313, "global_step": 179002, "epoch": 2011, "lr": 2.607708664605942e-05}
{"train_loss": 0.0930599644780159, "global_step": 179003, "epoch": 2011, "lr": 2.6076577574988876e-05}
{"train_loss": 0.07949327677488327, "global_step": 179004, "epoch": 2011, "lr": 2.607606850713453e-05}
{"train_loss": 0.07053832709789276, "global_step": 179005, "epoch": 2011, "lr": 2.6075559442496482e-05}
{"train_loss": 0.06785997748374939, "global_step": 179006, "epoch": 2011, "lr": 2.607505038107476e-05}
{"train_loss": 0.04657578095793724, "global_step": 179007, "epoch": 2011, "lr": 2.6074541322869482e-05}
{"train_loss": 0.1202457845211029, "global_step": 179008, "epoch": 2011, "lr": 2.6074032267880677e-05}
{"train_loss": 0.06932754814624786, "global_step": 179009, "epoch": 2011, "lr": 2.6073523216108435e-05}
{"train_loss": 0.07105670869350433, "global_step": 179010, "epoch": 2011, "lr": 2.60730141675528e-05}
{"train_loss": 0.0505048893392086, "global_step": 179011, "epoch": 2011, "lr": 2.607250512221386e-05}
{"train_loss": 0.04269697889685631, "global_step": 179012, "epoch": 2011, "lr": 2.6071996080091694e-05}
{"train_loss": 0.15884894132614136, "global_step": 179013, "epoch": 2011, "lr": 2.6071487041186337e-05}
{"train_loss": 0.07328661531209946, "global_step": 179014, "epoch": 2011, "lr": 2.6070978005497903e-05}
{"train_loss": 0.1134033277630806, "global_step": 179015, "epoch": 2011, "lr": 2.6070468973026408e-05}
{"train_loss": 0.13021905720233917, "global_step": 179016, "epoch": 2011, "lr": 2.606995994377197e-05}
{"train_loss": 0.07618681341409683, "global_step": 179017, "epoch": 2011, "lr": 2.6069450917734616e-05}
{"train_loss": 0.0726814940571785, "global_step": 179018, "epoch": 2011, "lr": 2.6068941894914445e-05}
{"train_loss": 0.0930589884519577, "global_step": 179019, "epoch": 2011, "lr": 2.6068432875311498e-05}
{"train_loss": 0.08604839444160461, "global_step": 179020, "epoch": 2011, "lr": 2.606792385892588e-05}
{"train_loss": 0.0560668408870697, "global_step": 179021, "epoch": 2011, "lr": 2.606741484575762e-05}
{"train_loss": 0.10688260197639465, "global_step": 179022, "epoch": 2011, "lr": 2.606690583580682e-05}
{"train_loss": 0.1149207279086113, "global_step": 179023, "epoch": 2011, "lr": 2.6066396829073515e-05}
{"train_loss": 0.0657133087515831, "global_step": 179024, "epoch": 2011, "lr": 2.606588782555781e-05}
{"train_loss": 0.07302597165107727, "global_step": 179025, "epoch": 2011, "lr": 2.6065378825259734e-05}
{"train_loss": 0.07960209995508194, "global_step": 179026, "epoch": 2011, "lr": 2.6064869828179393e-05}
{"train_loss": 0.052488479763269424, "global_step": 179027, "epoch": 2011, "lr": 2.6064360834316826e-05}
{"train_loss": 0.06311967223882675, "global_step": 179028, "epoch": 2011, "lr": 2.6063851843672126e-05}
{"train_loss": 0.09642206132411957, "global_step": 179029, "epoch": 2011, "lr": 2.6063342856245355e-05}
{"train_loss": 0.05868353322148323, "global_step": 179030, "epoch": 2011, "lr": 2.6062833872036558e-05}
{"train_loss": 0.09003420919179916, "global_step": 179031, "epoch": 2011, "lr": 2.6062324891045835e-05}
{"train_loss": 0.06098763272166252, "global_step": 179032, "epoch": 2011, "lr": 2.6061815913273224e-05}
{"train_loss": 0.06201371178030968, "global_step": 179033, "epoch": 2011, "lr": 2.6061306938718833e-05}
{"train_loss": 0.054486583918333054, "global_step": 179034, "epoch": 2011, "lr": 2.6060797967382683e-05}
{"train_loss": 0.10175321251153946, "global_step": 179035, "epoch": 2011, "lr": 2.6060288999264886e-05}
{"train_loss": 0.08968204259872437, "global_step": 179036, "epoch": 2011, "lr": 2.6059780034365478e-05}
{"train_loss": 0.04040253907442093, "global_step": 179037, "epoch": 2011, "lr": 2.6059271072684555e-05}
{"train_loss": 0.13553385436534882, "global_step": 179038, "epoch": 2011, "lr": 2.6058762114222147e-05}
{"train_loss": 0.15178631246089935, "global_step": 179039, "epoch": 2011, "lr": 2.6058253158978373e-05}
{"train_loss": 0.08172506839036942, "global_step": 179040, "epoch": 2011, "lr": 2.6057744206953256e-05}
{"train_loss": 0.11877545714378357, "global_step": 179041, "epoch": 2011, "lr": 2.6057235258146902e-05}
{"train_loss": 0.0628972128033638, "global_step": 179042, "epoch": 2011, "lr": 2.6056726312559344e-05}
{"train_loss": 0.05964914336800575, "global_step": 179043, "epoch": 2011, "lr": 2.6056217370190684e-05}
{"train_loss": 0.0724458321928978, "global_step": 179044, "epoch": 2011, "lr": 2.605570843104095e-05}
{"train_loss": 0.05034228414297104, "global_step": 179045, "epoch": 2011, "lr": 2.6055199495110254e-05}
{"train_loss": 0.04558519273996353, "global_step": 179046, "epoch": 2011, "lr": 2.605469056239863e-05}
{"train_loss": 0.08098962903022766, "global_step": 179047, "epoch": 2011, "lr": 2.605418163290616e-05}
{"train_loss": 0.07565987855195999, "global_step": 179048, "epoch": 2011, "lr": 2.605367270663293e-05}
{"train_loss": 0.0810326412320137, "global_step": 179049, "epoch": 2011, "lr": 2.6053163783578977e-05}
{"train_loss": 0.04540620371699333, "global_step": 179050, "epoch": 2011, "lr": 2.6052654863744402e-05}
{"train_loss": 0.04040179401636124, "global_step": 179051, "epoch": 2011, "lr": 2.6052145947129237e-05}
{"train_loss": 0.0750831589102745, "global_step": 179052, "epoch": 2011, "lr": 2.605163703373359e-05}
{"train_loss": 0.10503759980201721, "global_step": 179053, "epoch": 2011, "lr": 2.6051128123557477e-05}
{"train_loss": 0.06747656315565109, "global_step": 179054, "epoch": 2011, "lr": 2.6050619216601035e-05}
{"train_loss": 0.09077375382184982, "global_step": 179055, "epoch": 2011, "lr": 2.605011031286426e-05}
{"train_loss": 0.10299203544855118, "global_step": 179056, "epoch": 2011, "lr": 2.6049601412347285e-05}
{"train_loss": 0.041865792125463486, "global_step": 179057, "epoch": 2011, "lr": 2.6049092515050133e-05}
{"train_loss": 0.09205132722854614, "global_step": 179058, "epoch": 2011, "lr": 2.6048583620972898e-05}
{"train_loss": 0.09991763532161713, "global_step": 179059, "epoch": 2011, "lr": 2.6048074730115625e-05}
{"train_loss": 0.052551738917827606, "global_step": 179060, "epoch": 2011, "lr": 2.6047565842478416e-05}
{"train_loss": 0.036225128918886185, "global_step": 179061, "epoch": 2011, "lr": 2.60470569580613e-05}
{"train_loss": 0.11471860110759735, "global_step": 179062, "epoch": 2011, "lr": 2.6046548076864384e-05}
{"train_loss": 0.07456209510564804, "global_step": 179063, "epoch": 2011, "lr": 2.6046039198887694e-05}
{"train_loss": 0.10033776611089706, "global_step": 179064, "epoch": 2011, "lr": 2.6045530324131352e-05}
{"train_loss": 0.03006068617105484, "global_step": 179065, "epoch": 2011, "lr": 2.6045021452595365e-05}
{"train_loss": 0.04803471639752388, "global_step": 179066, "epoch": 2011, "lr": 2.6044512584279858e-05}
{"train_loss": 0.07897721996970392, "global_step": 179067, "epoch": 2011, "lr": 2.6044003719184872e-05, "val_loss": 7.416159629821777}
{"train_loss": 0.1357884705066681, "global_step": 179068, "epoch": 2012, "lr": 2.6043494857310468e-05}
{"train_loss": 0.08272261917591095, "global_step": 179069, "epoch": 2012, "lr": 2.6042985998656734e-05}
{"train_loss": 0.0899970754981041, "global_step": 179070, "epoch": 2012, "lr": 2.604247714322371e-05}
{"train_loss": 0.039420466870069504, "global_step": 179071, "epoch": 2012, "lr": 2.6041968291011503e-05}
{"train_loss": 0.0727376788854599, "global_step": 179072, "epoch": 2012, "lr": 2.6041459442020145e-05}
{"train_loss": 0.1441289484500885, "global_step": 179073, "epoch": 2012, "lr": 2.6040950596249734e-05}
{"train_loss": 0.09855855256319046, "global_step": 179074, "epoch": 2012, "lr": 2.6040441753700312e-05}
{"train_loss": 0.15205949544906616, "global_step": 179075, "epoch": 2012, "lr": 2.6039932914371977e-05}
{"train_loss": 0.05394880101084709, "global_step": 179076, "epoch": 2012, "lr": 2.6039424078264763e-05}
{"train_loss": 0.045868758112192154, "global_step": 179077, "epoch": 2012, "lr": 2.6038915245378777e-05}
{"train_loss": 0.08589065819978714, "global_step": 179078, "epoch": 2012, "lr": 2.6038406415714044e-05}
{"train_loss": 0.07434869557619095, "global_step": 179079, "epoch": 2012, "lr": 2.6037897589270676e-05}
{"train_loss": 0.08927401900291443, "global_step": 179080, "epoch": 2012, "lr": 2.60373887660487e-05}
{"train_loss": 0.09774305671453476, "global_step": 179081, "epoch": 2012, "lr": 2.6036879946048226e-05}
{"train_loss": 0.08425376564264297, "global_step": 179082, "epoch": 2012, "lr": 2.6036371129269284e-05}
{"train_loss": 0.09670192003250122, "global_step": 179083, "epoch": 2012, "lr": 2.6035862315711956e-05}
{"train_loss": 0.08223473280668259, "global_step": 179084, "epoch": 2012, "lr": 2.6035353505376336e-05}
{"train_loss": 0.08110108226537704, "global_step": 179085, "epoch": 2012, "lr": 2.603484469826245e-05}
{"train_loss": 0.06267422437667847, "global_step": 179086, "epoch": 2012, "lr": 2.6034335894370408e-05}
{"train_loss": 0.05571536719799042, "global_step": 179087, "epoch": 2012, "lr": 2.6033827093700237e-05}
{"train_loss": 0.09508345276117325, "global_step": 179088, "epoch": 2012, "lr": 2.603331829625204e-05}
{"train_loss": 0.06181008741259575, "global_step": 179089, "epoch": 2012, "lr": 2.6032809502025857e-05}
{"train_loss": 0.06677138805389404, "global_step": 179090, "epoch": 2012, "lr": 2.6032300711021795e-05}
{"train_loss": 0.0888308584690094, "global_step": 179091, "epoch": 2012, "lr": 2.6031791923239867e-05}
{"train_loss": 0.06829941272735596, "global_step": 179092, "epoch": 2012, "lr": 2.60312831386802e-05}
{"train_loss": 0.10605518519878387, "global_step": 179093, "epoch": 2012, "lr": 2.6030774357342814e-05}
{"train_loss": 0.05864189565181732, "global_step": 179094, "epoch": 2012, "lr": 2.603026557922782e-05}
{"train_loss": 0.11385796964168549, "global_step": 179095, "epoch": 2012, "lr": 2.6029756804335242e-05}
{"train_loss": 0.1011786237359047, "global_step": 179096, "epoch": 2012, "lr": 2.6029248032665188e-05}
{"train_loss": 0.0680367648601532, "global_step": 179097, "epoch": 2012, "lr": 2.6028739264217695e-05}
{"train_loss": 0.09823232889175415, "global_step": 179098, "epoch": 2012, "lr": 2.6028230498992867e-05}
{"train_loss": 0.1267385184764862, "global_step": 179099, "epoch": 2012, "lr": 2.602772173699073e-05}
{"train_loss": 0.04110102728009224, "global_step": 179100, "epoch": 2012, "lr": 2.602721297821139e-05}
{"train_loss": 0.033559489995241165, "global_step": 179101, "epoch": 2012, "lr": 2.602670422265488e-05}
{"train_loss": 0.08353465050458908, "global_step": 179102, "epoch": 2012, "lr": 2.6026195470321307e-05}
{"train_loss": 0.05147324502468109, "global_step": 179103, "epoch": 2012, "lr": 2.60256867212107e-05}
{"train_loss": 0.07851973176002502, "global_step": 179104, "epoch": 2012, "lr": 2.6025177975323168e-05}
{"train_loss": 0.24436995387077332, "global_step": 179105, "epoch": 2012, "lr": 2.6024669232658738e-05}
{"train_loss": 0.09053540974855423, "global_step": 179106, "epoch": 2012, "lr": 2.602416049321752e-05}
{"train_loss": 0.0996900424361229, "global_step": 179107, "epoch": 2012, "lr": 2.602365175699956e-05}
{"train_loss": 0.07884038239717484, "global_step": 179108, "epoch": 2012, "lr": 2.6023143024004913e-05}
{"train_loss": 0.08875398337841034, "global_step": 179109, "epoch": 2012, "lr": 2.602263429423368e-05}
{"train_loss": 0.04849948734045029, "global_step": 179110, "epoch": 2012, "lr": 2.6022125567685884e-05}
{"train_loss": 0.10415373742580414, "global_step": 179111, "epoch": 2012, "lr": 2.602161684436165e-05}
{"train_loss": 0.066463902592659, "global_step": 179112, "epoch": 2012, "lr": 2.6021108124260996e-05}
{"train_loss": 0.08539120107889175, "global_step": 179113, "epoch": 2012, "lr": 2.6020599407384027e-05}
{"train_loss": 0.07444466650485992, "global_step": 179114, "epoch": 2012, "lr": 2.602009069373078e-05}
{"train_loss": 0.031104546040296555, "global_step": 179115, "epoch": 2012, "lr": 2.6019581983301357e-05}
{"train_loss": 0.06657180190086365, "global_step": 179116, "epoch": 2012, "lr": 2.601907327609579e-05}
{"train_loss": 0.060677871108055115, "global_step": 179117, "epoch": 2012, "lr": 2.601856457211419e-05}
{"train_loss": 0.07890568673610687, "global_step": 179118, "epoch": 2012, "lr": 2.6018055871356573e-05}
{"train_loss": 0.1482151448726654, "global_step": 179119, "epoch": 2012, "lr": 2.601754717382304e-05}
{"train_loss": 0.08535102009773254, "global_step": 179120, "epoch": 2012, "lr": 2.601703847951368e-05}
{"train_loss": 0.14998072385787964, "global_step": 179121, "epoch": 2012, "lr": 2.6016529788428513e-05}
{"train_loss": 0.10111398994922638, "global_step": 179122, "epoch": 2012, "lr": 2.6016021100567656e-05}
{"train_loss": 0.05169715732336044, "global_step": 179123, "epoch": 2012, "lr": 2.6015512415931126e-05}
{"train_loss": 0.09482929110527039, "global_step": 179124, "epoch": 2012, "lr": 2.601500373451904e-05}
{"train_loss": 0.14144380390644073, "global_step": 179125, "epoch": 2012, "lr": 2.6014495056331424e-05}
{"train_loss": 0.04254278913140297, "global_step": 179126, "epoch": 2012, "lr": 2.6013986381368392e-05}
{"train_loss": 0.09493793547153473, "global_step": 179127, "epoch": 2012, "lr": 2.6013477709629964e-05}
{"train_loss": 0.05985000729560852, "global_step": 179128, "epoch": 2012, "lr": 2.6012969041116248e-05}
{"train_loss": 0.059767432510852814, "global_step": 179129, "epoch": 2012, "lr": 2.601246037582728e-05}
{"train_loss": 0.041987184435129166, "global_step": 179130, "epoch": 2012, "lr": 2.6011951713763162e-05}
{"train_loss": 0.07761316001415253, "global_step": 179131, "epoch": 2012, "lr": 2.6011443054923924e-05}
{"train_loss": 0.06558530032634735, "global_step": 179132, "epoch": 2012, "lr": 2.6010934399309676e-05}
{"train_loss": 0.11970536410808563, "global_step": 179133, "epoch": 2012, "lr": 2.6010425746920452e-05}
{"train_loss": 0.059547364711761475, "global_step": 179134, "epoch": 2012, "lr": 2.6009917097756343e-05}
{"train_loss": 0.1281803995370865, "global_step": 179135, "epoch": 2012, "lr": 2.6009408451817397e-05}
{"train_loss": 0.09522544592618942, "global_step": 179136, "epoch": 2012, "lr": 2.6008899809103705e-05}
{"train_loss": 0.06495507061481476, "global_step": 179137, "epoch": 2012, "lr": 2.600839116961531e-05}
{"train_loss": 0.08899616450071335, "global_step": 179138, "epoch": 2012, "lr": 2.6007882533352296e-05}
{"train_loss": 0.05706948786973953, "global_step": 179139, "epoch": 2012, "lr": 2.6007373900314746e-05}
{"train_loss": 0.14318591356277466, "global_step": 179140, "epoch": 2012, "lr": 2.600686527050271e-05}
{"train_loss": 0.06264281272888184, "global_step": 179141, "epoch": 2012, "lr": 2.6006356643916248e-05}
{"train_loss": 0.13852640986442566, "global_step": 179142, "epoch": 2012, "lr": 2.6005848020555452e-05}
{"train_loss": 0.10792617499828339, "global_step": 179143, "epoch": 2012, "lr": 2.600533940042036e-05}
{"train_loss": 0.07711394131183624, "global_step": 179144, "epoch": 2012, "lr": 2.600483078351107e-05}
{"train_loss": 0.10555306822061539, "global_step": 179145, "epoch": 2012, "lr": 2.6004322169827644e-05}
{"train_loss": 0.06460169702768326, "global_step": 179146, "epoch": 2012, "lr": 2.600381355937013e-05}
{"train_loss": 0.13836944103240967, "global_step": 179147, "epoch": 2012, "lr": 2.600330495213862e-05}
{"train_loss": 0.14223191142082214, "global_step": 179148, "epoch": 2012, "lr": 2.6002796348133164e-05}
{"train_loss": 0.07095947116613388, "global_step": 179149, "epoch": 2012, "lr": 2.6002287747353847e-05}
{"train_loss": 0.14548590779304504, "global_step": 179150, "epoch": 2012, "lr": 2.600177914980072e-05}
{"train_loss": 0.03464488685131073, "global_step": 179151, "epoch": 2012, "lr": 2.600127055547388e-05}
{"train_loss": 0.12435347586870193, "global_step": 179152, "epoch": 2012, "lr": 2.6000761964373353e-05}
{"train_loss": 0.07914087176322937, "global_step": 179153, "epoch": 2012, "lr": 2.600025337649925e-05}
{"train_loss": 0.10170856863260269, "global_step": 179154, "epoch": 2012, "lr": 2.5999744791851605e-05}
{"train_loss": 0.12322473526000977, "global_step": 179155, "epoch": 2012, "lr": 2.59992362104305e-05}
{"train_loss": 0.08808151603247343, "global_step": 179156, "epoch": 2012, "lr": 2.599872763223602e-05, "val_loss": 7.54909610748291}
{"train_loss": 0.046688470989465714, "global_step": 179157, "epoch": 2013, "lr": 2.599821905726821e-05}
{"train_loss": 0.0657910630106926, "global_step": 179158, "epoch": 2013, "lr": 2.599771048552716e-05}
{"train_loss": 0.1395379900932312, "global_step": 179159, "epoch": 2013, "lr": 2.5997201917012904e-05}
{"train_loss": 0.04644593223929405, "global_step": 179160, "epoch": 2013, "lr": 2.5996693351725555e-05}
{"train_loss": 0.09307938069105148, "global_step": 179161, "epoch": 2013, "lr": 2.5996184789665133e-05}
{"train_loss": 0.0910133644938469, "global_step": 179162, "epoch": 2013, "lr": 2.5995676230831755e-05}
{"train_loss": 0.10098057240247726, "global_step": 179163, "epoch": 2013, "lr": 2.5995167675225447e-05}
{"train_loss": 0.05851864069700241, "global_step": 179164, "epoch": 2013, "lr": 2.599465912284631e-05}
{"train_loss": 0.048343513160943985, "global_step": 179165, "epoch": 2013, "lr": 2.5994150573694388e-05}
{"train_loss": 0.05362351983785629, "global_step": 179166, "epoch": 2013, "lr": 2.599364202776977e-05}
{"train_loss": 0.05577782541513443, "global_step": 179167, "epoch": 2013, "lr": 2.5993133485072506e-05}
{"train_loss": 0.04799420014023781, "global_step": 179168, "epoch": 2013, "lr": 2.5992624945602684e-05}
{"train_loss": 0.08900074660778046, "global_step": 179169, "epoch": 2013, "lr": 2.5992116409360346e-05}
{"train_loss": 0.07536758482456207, "global_step": 179170, "epoch": 2013, "lr": 2.5991607876345592e-05}
{"train_loss": 0.07474415004253387, "global_step": 179171, "epoch": 2013, "lr": 2.599109934655845e-05}
{"train_loss": 0.07432719320058823, "global_step": 179172, "epoch": 2013, "lr": 2.5990590819999038e-05}
{"train_loss": 0.07861078530550003, "global_step": 179173, "epoch": 2013, "lr": 2.5990082296667372e-05}
{"train_loss": 0.13201436400413513, "global_step": 179174, "epoch": 2013, "lr": 2.5989573776563552e-05}
{"train_loss": 0.05513671785593033, "global_step": 179175, "epoch": 2013, "lr": 2.5989065259687663e-05}
{"train_loss": 0.06824801862239838, "global_step": 179176, "epoch": 2013, "lr": 2.598855674603973e-05}
{"train_loss": 0.07952965050935745, "global_step": 179177, "epoch": 2013, "lr": 2.598804823561986e-05}
{"train_loss": 0.08525502681732178, "global_step": 179178, "epoch": 2013, "lr": 2.5987539728428106e-05}
{"train_loss": 0.11122250556945801, "global_step": 179179, "epoch": 2013, "lr": 2.5987031224464516e-05}
{"train_loss": 0.12559211254119873, "global_step": 179180, "epoch": 2013, "lr": 2.5986522723729188e-05}
{"train_loss": 0.14687713980674744, "global_step": 179181, "epoch": 2013, "lr": 2.598601422622217e-05}
{"train_loss": 0.08770684152841568, "global_step": 179182, "epoch": 2013, "lr": 2.598550573194356e-05}
{"train_loss": 0.055189672857522964, "global_step": 179183, "epoch": 2013, "lr": 2.5984997240893407e-05}
{"train_loss": 0.09901955723762512, "global_step": 179184, "epoch": 2013, "lr": 2.5984488753071744e-05}
{"train_loss": 0.11609717458486557, "global_step": 179185, "epoch": 2013, "lr": 2.598398026847871e-05}
{"train_loss": 0.05887546390295029, "global_step": 179186, "epoch": 2013, "lr": 2.598347178711431e-05}
{"train_loss": 0.06997659802436829, "global_step": 179187, "epoch": 2013, "lr": 2.598296330897866e-05}
{"train_loss": 0.09337931126356125, "global_step": 179188, "epoch": 2013, "lr": 2.598245483407179e-05}
{"train_loss": 0.06369608640670776, "global_step": 179189, "epoch": 2013, "lr": 2.5981946362393806e-05}
{"train_loss": 0.12105298787355423, "global_step": 179190, "epoch": 2013, "lr": 2.598143789394473e-05}
{"train_loss": 0.05299827829003334, "global_step": 179191, "epoch": 2013, "lr": 2.5980929428724686e-05}
{"train_loss": 0.07106640934944153, "global_step": 179192, "epoch": 2013, "lr": 2.5980420966733686e-05}
{"train_loss": 0.12218315154314041, "global_step": 179193, "epoch": 2013, "lr": 2.5979912507971826e-05}
{"train_loss": 0.03485212102532387, "global_step": 179194, "epoch": 2013, "lr": 2.5979404052439205e-05}
{"train_loss": 0.04555578529834747, "global_step": 179195, "epoch": 2013, "lr": 2.597889560013583e-05}
{"train_loss": 0.13107825815677643, "global_step": 179196, "epoch": 2013, "lr": 2.5978387151061824e-05}
{"train_loss": 0.10443361103534698, "global_step": 179197, "epoch": 2013, "lr": 2.5977878705217206e-05}
{"train_loss": 0.05690433084964752, "global_step": 179198, "epoch": 2013, "lr": 2.5977370262602092e-05}
{"train_loss": 0.07202655076980591, "global_step": 179199, "epoch": 2013, "lr": 2.5976861823216507e-05}
{"train_loss": 0.0792631134390831, "global_step": 179200, "epoch": 2013, "lr": 2.597635338706056e-05}
{"train_loss": 0.06261581182479858, "global_step": 179201, "epoch": 2013, "lr": 2.5975844954134287e-05}
{"train_loss": 0.027631442993879318, "global_step": 179202, "epoch": 2013, "lr": 2.5975336524437776e-05}
{"train_loss": 0.10284222662448883, "global_step": 179203, "epoch": 2013, "lr": 2.5974828097971084e-05}
{"train_loss": 0.1150069311261177, "global_step": 179204, "epoch": 2013, "lr": 2.5974319674734294e-05}
{"train_loss": 0.10220354050397873, "global_step": 179205, "epoch": 2013, "lr": 2.597381125472744e-05}
{"train_loss": 0.05489824339747429, "global_step": 179206, "epoch": 2013, "lr": 2.5973302837950643e-05}
{"train_loss": 0.050773706287145615, "global_step": 179207, "epoch": 2013, "lr": 2.5972794424403913e-05}
{"train_loss": 0.12835901975631714, "global_step": 179208, "epoch": 2013, "lr": 2.5972286014087378e-05}
{"train_loss": 0.10202225297689438, "global_step": 179209, "epoch": 2013, "lr": 2.5971777607001047e-05}
{"train_loss": 0.09446846693754196, "global_step": 179210, "epoch": 2013, "lr": 2.597126920314502e-05}
{"train_loss": 0.05857948958873749, "global_step": 179211, "epoch": 2013, "lr": 2.597076080251939e-05}
{"train_loss": 0.08166802674531937, "global_step": 179212, "epoch": 2013, "lr": 2.5970252405124173e-05}
{"train_loss": 0.13072185218334198, "global_step": 179213, "epoch": 2013, "lr": 2.5969744010959478e-05}
{"train_loss": 0.09000230580568314, "global_step": 179214, "epoch": 2013, "lr": 2.5969235620025344e-05}
{"train_loss": 0.06211932748556137, "global_step": 179215, "epoch": 2013, "lr": 2.5968727232321877e-05}
{"train_loss": 0.05114927887916565, "global_step": 179216, "epoch": 2013, "lr": 2.5968218847849113e-05}
{"train_loss": 0.07571352273225784, "global_step": 179217, "epoch": 2013, "lr": 2.5967710466607116e-05}
{"train_loss": 0.04680579900741577, "global_step": 179218, "epoch": 2013, "lr": 2.5967202088595984e-05}
{"train_loss": 0.11844674497842789, "global_step": 179219, "epoch": 2013, "lr": 2.596669371381575e-05}
{"train_loss": 0.051098763942718506, "global_step": 179220, "epoch": 2013, "lr": 2.596618534226652e-05}
{"train_loss": 0.062036119401454926, "global_step": 179221, "epoch": 2013, "lr": 2.596567697394832e-05}
{"train_loss": 0.06588689237833023, "global_step": 179222, "epoch": 2013, "lr": 2.5965168608861268e-05}
{"train_loss": 0.06154797226190567, "global_step": 179223, "epoch": 2013, "lr": 2.59646602470054e-05}
{"train_loss": 0.10946781933307648, "global_step": 179224, "epoch": 2013, "lr": 2.5964151888380776e-05}
{"train_loss": 0.07881366461515427, "global_step": 179225, "epoch": 2013, "lr": 2.5963643532987504e-05}
{"train_loss": 0.06434821337461472, "global_step": 179226, "epoch": 2013, "lr": 2.5963135180825594e-05}
{"train_loss": 0.10915809869766235, "global_step": 179227, "epoch": 2013, "lr": 2.5962626831895176e-05}
{"train_loss": 0.09445269405841827, "global_step": 179228, "epoch": 2013, "lr": 2.5962118486196267e-05}
{"train_loss": 0.06612613052129745, "global_step": 179229, "epoch": 2013, "lr": 2.5961610143728958e-05}
{"train_loss": 0.09365154057741165, "global_step": 179230, "epoch": 2013, "lr": 2.5961101804493337e-05}
{"train_loss": 0.10533797740936279, "global_step": 179231, "epoch": 2013, "lr": 2.5960593468489435e-05}
{"train_loss": 0.044874563813209534, "global_step": 179232, "epoch": 2013, "lr": 2.596008513571736e-05}
{"train_loss": 0.07679847627878189, "global_step": 179233, "epoch": 2013, "lr": 2.5959576806177132e-05}
{"train_loss": 0.07800343632698059, "global_step": 179234, "epoch": 2013, "lr": 2.5959068479868877e-05}
{"train_loss": 0.08508644253015518, "global_step": 179235, "epoch": 2013, "lr": 2.5958560156792598e-05}
{"train_loss": 0.0823405310511589, "global_step": 179236, "epoch": 2013, "lr": 2.5958051836948428e-05}
{"train_loss": 0.05376988649368286, "global_step": 179237, "epoch": 2013, "lr": 2.595754352033638e-05}
{"train_loss": 0.09970651566982269, "global_step": 179238, "epoch": 2013, "lr": 2.595703520695657e-05}
{"train_loss": 0.02447192743420601, "global_step": 179239, "epoch": 2013, "lr": 2.5956526896809024e-05}
{"train_loss": 0.06392321735620499, "global_step": 179240, "epoch": 2013, "lr": 2.595601858989385e-05}
{"train_loss": 0.06421392410993576, "global_step": 179241, "epoch": 2013, "lr": 2.595551028621107e-05}
{"train_loss": 0.06012998893857002, "global_step": 179242, "epoch": 2013, "lr": 2.5955001985760807e-05}
{"train_loss": 0.0842609852552414, "global_step": 179243, "epoch": 2013, "lr": 2.595449368854308e-05}
{"train_loss": 0.09285573661327362, "global_step": 179244, "epoch": 2013, "lr": 2.595398539455799e-05}
{"train_loss": 0.07951789751146617, "global_step": 179245, "epoch": 2013, "lr": 2.595347710380558e-05, "val_loss": 7.483673572540283}
{"train_loss": 0.08283452689647675, "global_step": 179246, "epoch": 2014, "lr": 2.5952968816285932e-05}
{"train_loss": 0.03437625244259834, "global_step": 179247, "epoch": 2014, "lr": 2.5952460531999134e-05}
{"train_loss": 0.074822798371315, "global_step": 179248, "epoch": 2014, "lr": 2.595195225094521e-05}
{"train_loss": 0.11400970816612244, "global_step": 179249, "epoch": 2014, "lr": 2.595144397312428e-05}
{"train_loss": 0.11978679150342941, "global_step": 179250, "epoch": 2014, "lr": 2.595093569853636e-05}
{"train_loss": 0.04683838039636612, "global_step": 179251, "epoch": 2014, "lr": 2.5950427427181567e-05}
{"train_loss": 0.0927717387676239, "global_step": 179252, "epoch": 2014, "lr": 2.5949919159059922e-05}
{"train_loss": 0.07829870283603668, "global_step": 179253, "epoch": 2014, "lr": 2.5949410894171544e-05}
{"train_loss": 0.08048059791326523, "global_step": 179254, "epoch": 2014, "lr": 2.5948902632516446e-05}
{"train_loss": 0.08691627532243729, "global_step": 179255, "epoch": 2014, "lr": 2.5948394374094755e-05}
{"train_loss": 0.0845787525177002, "global_step": 179256, "epoch": 2014, "lr": 2.59478861189065e-05}
{"train_loss": 0.08075875043869019, "global_step": 179257, "epoch": 2014, "lr": 2.594737786695174e-05}
{"train_loss": 0.12452170252799988, "global_step": 179258, "epoch": 2014, "lr": 2.5946869618230584e-05}
{"train_loss": 0.05287470668554306, "global_step": 179259, "epoch": 2014, "lr": 2.5946361372743055e-05}
{"train_loss": 0.03416738659143448, "global_step": 179260, "epoch": 2014, "lr": 2.5945853130489274e-05}
{"train_loss": 0.06864359229803085, "global_step": 179261, "epoch": 2014, "lr": 2.594534489146927e-05}
{"train_loss": 0.05830642953515053, "global_step": 179262, "epoch": 2014, "lr": 2.5944836655683103e-05}
{"train_loss": 0.08555778115987778, "global_step": 179263, "epoch": 2014, "lr": 2.5944328423130882e-05}
{"train_loss": 0.05852166563272476, "global_step": 179264, "epoch": 2014, "lr": 2.594382019381263e-05}
{"train_loss": 0.10887552797794342, "global_step": 179265, "epoch": 2014, "lr": 2.5943311967728447e-05}
{"train_loss": 0.07960501313209534, "global_step": 179266, "epoch": 2014, "lr": 2.5942803744878402e-05}
{"train_loss": 0.10788692533969879, "global_step": 179267, "epoch": 2014, "lr": 2.594229552526254e-05}
{"train_loss": 0.09381633996963501, "global_step": 179268, "epoch": 2014, "lr": 2.594178730888096e-05}
{"train_loss": 0.14206010103225708, "global_step": 179269, "epoch": 2014, "lr": 2.5941279095733694e-05}
{"train_loss": 0.07060394436120987, "global_step": 179270, "epoch": 2014, "lr": 2.5940770885820848e-05}
{"train_loss": 0.14167487621307373, "global_step": 179271, "epoch": 2014, "lr": 2.5940262679142458e-05}
{"train_loss": 0.07715439796447754, "global_step": 179272, "epoch": 2014, "lr": 2.593975447569862e-05}
{"train_loss": 0.07823285460472107, "global_step": 179273, "epoch": 2014, "lr": 2.593924627548937e-05}
{"train_loss": 0.07307294011116028, "global_step": 179274, "epoch": 2014, "lr": 2.593873807851481e-05}
{"train_loss": 0.04759489372372627, "global_step": 179275, "epoch": 2014, "lr": 2.5938229884774985e-05}
{"train_loss": 0.13073578476905823, "global_step": 179276, "epoch": 2014, "lr": 2.593772169426999e-05}
{"train_loss": 0.05099726468324661, "global_step": 179277, "epoch": 2014, "lr": 2.5937213506999846e-05}
{"train_loss": 0.06838294118642807, "global_step": 179278, "epoch": 2014, "lr": 2.5936705322964676e-05}
{"train_loss": 0.11722911894321442, "global_step": 179279, "epoch": 2014, "lr": 2.5936197142164505e-05}
{"train_loss": 0.09245043992996216, "global_step": 179280, "epoch": 2014, "lr": 2.5935688964599437e-05}
{"train_loss": 0.09227524697780609, "global_step": 179281, "epoch": 2014, "lr": 2.59351807902695e-05}
{"train_loss": 0.08843973278999329, "global_step": 179282, "epoch": 2014, "lr": 2.5934672619174783e-05}
{"train_loss": 0.084522545337677, "global_step": 179283, "epoch": 2014, "lr": 2.5934164451315378e-05}
{"train_loss": 0.10979608446359634, "global_step": 179284, "epoch": 2014, "lr": 2.5933656286691317e-05}
{"train_loss": 0.05228670686483383, "global_step": 179285, "epoch": 2014, "lr": 2.5933148125302697e-05}
{"train_loss": 0.1381787806749344, "global_step": 179286, "epoch": 2014, "lr": 2.593263996714955e-05}
{"train_loss": 0.06413167715072632, "global_step": 179287, "epoch": 2014, "lr": 2.5932131812231996e-05}
{"train_loss": 0.08003897964954376, "global_step": 179288, "epoch": 2014, "lr": 2.5931623660550042e-05}
{"train_loss": 0.12937241792678833, "global_step": 179289, "epoch": 2014, "lr": 2.5931115512103815e-05}
{"train_loss": 0.06649517267942429, "global_step": 179290, "epoch": 2014, "lr": 2.5930607366893332e-05}
{"train_loss": 0.03364185988903046, "global_step": 179291, "epoch": 2014, "lr": 2.5930099224918708e-05}
{"train_loss": 0.13521213829517365, "global_step": 179292, "epoch": 2014, "lr": 2.5929591086179973e-05}
{"train_loss": 0.11987850815057755, "global_step": 179293, "epoch": 2014, "lr": 2.592908295067722e-05}
{"train_loss": 0.10534095019102097, "global_step": 179294, "epoch": 2014, "lr": 2.5928574818410518e-05}
{"train_loss": 0.03287307545542717, "global_step": 179295, "epoch": 2014, "lr": 2.5928066689379902e-05}
{"train_loss": 0.09217178821563721, "global_step": 179296, "epoch": 2014, "lr": 2.5927558563585485e-05}
{"train_loss": 0.04888606816530228, "global_step": 179297, "epoch": 2014, "lr": 2.592705044102729e-05}
{"train_loss": 0.07577034831047058, "global_step": 179298, "epoch": 2014, "lr": 2.592654232170544e-05}
{"train_loss": 0.06620781868696213, "global_step": 179299, "epoch": 2014, "lr": 2.5926034205619964e-05}
{"train_loss": 0.07417336851358414, "global_step": 179300, "epoch": 2014, "lr": 2.592552609277092e-05}
{"train_loss": 0.10139568895101547, "global_step": 179301, "epoch": 2014, "lr": 2.5925017983158396e-05}
{"train_loss": 0.0652397871017456, "global_step": 179302, "epoch": 2014, "lr": 2.5924509876782476e-05}
{"train_loss": 0.08700765669345856, "global_step": 179303, "epoch": 2014, "lr": 2.5924001773643204e-05}
{"train_loss": 0.055043529719114304, "global_step": 179304, "epoch": 2014, "lr": 2.5923493673740666e-05}
{"train_loss": 0.04219982028007507, "global_step": 179305, "epoch": 2014, "lr": 2.592298557707491e-05}
{"train_loss": 0.0496031790971756, "global_step": 179306, "epoch": 2014, "lr": 2.5922477483646023e-05}
{"train_loss": 0.05766076594591141, "global_step": 179307, "epoch": 2014, "lr": 2.592196939345406e-05}
{"train_loss": 0.07835810631513596, "global_step": 179308, "epoch": 2014, "lr": 2.5921461306499105e-05}
{"train_loss": 0.05833040550351143, "global_step": 179309, "epoch": 2014, "lr": 2.5920953222781198e-05}
{"train_loss": 0.07650923728942871, "global_step": 179310, "epoch": 2014, "lr": 2.5920445142300444e-05}
{"train_loss": 0.07826537638902664, "global_step": 179311, "epoch": 2014, "lr": 2.5919937065056877e-05}
{"train_loss": 0.062357865273952484, "global_step": 179312, "epoch": 2014, "lr": 2.5919428991050593e-05}
{"train_loss": 0.03644849359989166, "global_step": 179313, "epoch": 2014, "lr": 2.591892092028163e-05}
{"train_loss": 0.08597679436206818, "global_step": 179314, "epoch": 2014, "lr": 2.59184128527501e-05}
{"train_loss": 0.06817559897899628, "global_step": 179315, "epoch": 2014, "lr": 2.5917904788456027e-05}
{"train_loss": 0.08038731664419174, "global_step": 179316, "epoch": 2014, "lr": 2.591739672739951e-05}
{"train_loss": 0.07667815685272217, "global_step": 179317, "epoch": 2014, "lr": 2.5916888669580598e-05}
{"train_loss": 0.09049014747142792, "global_step": 179318, "epoch": 2014, "lr": 2.5916380614999374e-05}
{"train_loss": 0.06979560852050781, "global_step": 179319, "epoch": 2014, "lr": 2.5915872563655886e-05}
{"train_loss": 0.05289440229535103, "global_step": 179320, "epoch": 2014, "lr": 2.5915364515550212e-05}
{"train_loss": 0.05364421010017395, "global_step": 179321, "epoch": 2014, "lr": 2.5914856470682447e-05}
{"train_loss": 0.08790095150470734, "global_step": 179322, "epoch": 2014, "lr": 2.5914348429052615e-05}
{"train_loss": 0.06478878855705261, "global_step": 179323, "epoch": 2014, "lr": 2.5913840390660826e-05}
{"train_loss": 0.06414273381233215, "global_step": 179324, "epoch": 2014, "lr": 2.5913332355507103e-05}
{"train_loss": 0.04644005745649338, "global_step": 179325, "epoch": 2014, "lr": 2.5912824323591566e-05}
{"train_loss": 0.14548738300800323, "global_step": 179326, "epoch": 2014, "lr": 2.591231629491423e-05}
{"train_loss": 0.047832995653152466, "global_step": 179327, "epoch": 2014, "lr": 2.5911808269475214e-05}
{"train_loss": 0.039666444063186646, "global_step": 179328, "epoch": 2014, "lr": 2.591130024727454e-05}
{"train_loss": 0.0756475031375885, "global_step": 179329, "epoch": 2014, "lr": 2.5910792228312314e-05}
{"train_loss": 0.09793082624673843, "global_step": 179330, "epoch": 2014, "lr": 2.5910284212588574e-05}
{"train_loss": 0.09847301244735718, "global_step": 179331, "epoch": 2014, "lr": 2.590977620010342e-05}
{"train_loss": 0.04732777923345566, "global_step": 179332, "epoch": 2014, "lr": 2.5909268190856905e-05}
{"train_loss": 0.1303299218416214, "global_step": 179333, "epoch": 2014, "lr": 2.590876018484907e-05}
{"train_loss": 0.07926998492539598, "global_step": 179334, "epoch": 2014, "lr": 2.5908252182080035e-05, "val_loss": 7.568572521209717}
{"train_loss": 0.0785931870341301, "global_step": 179335, "epoch": 2015, "lr": 2.590774418254982e-05}
{"train_loss": 0.07005686312913895, "global_step": 179336, "epoch": 2015, "lr": 2.5907236186258534e-05}
{"train_loss": 0.07303116470575333, "global_step": 179337, "epoch": 2015, "lr": 2.590672819320621e-05}
{"train_loss": 0.0796826109290123, "global_step": 179338, "epoch": 2015, "lr": 2.5906220203392944e-05}
{"train_loss": 0.0631241723895073, "global_step": 179339, "epoch": 2015, "lr": 2.590571221681878e-05}
{"train_loss": 0.08119127154350281, "global_step": 179340, "epoch": 2015, "lr": 2.5905204233483814e-05}
{"train_loss": 0.11302328109741211, "global_step": 179341, "epoch": 2015, "lr": 2.590469625338809e-05}
{"train_loss": 0.054955631494522095, "global_step": 179342, "epoch": 2015, "lr": 2.5904188276531692e-05}
{"train_loss": 0.061846110969781876, "global_step": 179343, "epoch": 2015, "lr": 2.5903680302914675e-05}
{"train_loss": 0.1283157616853714, "global_step": 179344, "epoch": 2015, "lr": 2.590317233253713e-05}
{"train_loss": 0.06468875706195831, "global_step": 179345, "epoch": 2015, "lr": 2.5902664365399087e-05}
{"train_loss": 0.125822514295578, "global_step": 179346, "epoch": 2015, "lr": 2.590215640150066e-05}
{"train_loss": 0.10938399285078049, "global_step": 179347, "epoch": 2015, "lr": 2.5901648440841876e-05}
{"train_loss": 0.08535882085561752, "global_step": 179348, "epoch": 2015, "lr": 2.5901140483422836e-05}
{"train_loss": 0.101702481508255, "global_step": 179349, "epoch": 2015, "lr": 2.5900632529243578e-05}
{"train_loss": 0.07764386385679245, "global_step": 179350, "epoch": 2015, "lr": 2.5900124578304208e-05}
{"train_loss": 0.09464030712842941, "global_step": 179351, "epoch": 2015, "lr": 2.589961663060475e-05}
{"train_loss": 0.05713888257741928, "global_step": 179352, "epoch": 2015, "lr": 2.5899108686145313e-05}
{"train_loss": 0.056768715381622314, "global_step": 179353, "epoch": 2015, "lr": 2.5898600744925927e-05}
{"train_loss": 0.08230478316545486, "global_step": 179354, "epoch": 2015, "lr": 2.5898092806946707e-05}
{"train_loss": 0.08971678465604782, "global_step": 179355, "epoch": 2015, "lr": 2.5897584872207665e-05}
{"train_loss": 0.053010281175374985, "global_step": 179356, "epoch": 2015, "lr": 2.5897076940708903e-05}
{"train_loss": 0.06274328380823135, "global_step": 179357, "epoch": 2015, "lr": 2.5896569012450504e-05}
{"train_loss": 0.06691573560237885, "global_step": 179358, "epoch": 2015, "lr": 2.58960610874325e-05}
{"train_loss": 0.06894255429506302, "global_step": 179359, "epoch": 2015, "lr": 2.5895553165654994e-05}
{"train_loss": 0.08441942930221558, "global_step": 179360, "epoch": 2015, "lr": 2.5895045247118022e-05}
{"train_loss": 0.05059725046157837, "global_step": 179361, "epoch": 2015, "lr": 2.589453733182169e-05}
{"train_loss": 0.044756073504686356, "global_step": 179362, "epoch": 2015, "lr": 2.5894029419766014e-05}
{"train_loss": 0.12527774274349213, "global_step": 179363, "epoch": 2015, "lr": 2.5893521510951114e-05}
{"train_loss": 0.08685243129730225, "global_step": 179364, "epoch": 2015, "lr": 2.589301360537702e-05}
{"train_loss": 0.12239956855773926, "global_step": 179365, "epoch": 2015, "lr": 2.5892505703043835e-05}
{"train_loss": 0.11545103788375854, "global_step": 179366, "epoch": 2015, "lr": 2.589199780395159e-05}
{"train_loss": 0.05526963993906975, "global_step": 179367, "epoch": 2015, "lr": 2.589148990810039e-05}
{"train_loss": 0.053492553532123566, "global_step": 179368, "epoch": 2015, "lr": 2.5890982015490268e-05}
{"train_loss": 0.10389795899391174, "global_step": 179369, "epoch": 2015, "lr": 2.5890474126121323e-05}
{"train_loss": 0.09132799506187439, "global_step": 179370, "epoch": 2015, "lr": 2.58899662399936e-05}
{"train_loss": 0.06063520163297653, "global_step": 179371, "epoch": 2015, "lr": 2.5889458357107188e-05}
{"train_loss": 0.08720405399799347, "global_step": 179372, "epoch": 2015, "lr": 2.5888950477462148e-05}
{"train_loss": 0.07645556330680847, "global_step": 179373, "epoch": 2015, "lr": 2.5888442601058526e-05}
{"train_loss": 0.07762379944324493, "global_step": 179374, "epoch": 2015, "lr": 2.5887934727896428e-05}
{"train_loss": 0.06004077196121216, "global_step": 179375, "epoch": 2015, "lr": 2.588742685797588e-05}
{"train_loss": 0.11147934943437576, "global_step": 179376, "epoch": 2015, "lr": 2.5886918991297004e-05}
{"train_loss": 0.11431552469730377, "global_step": 179377, "epoch": 2015, "lr": 2.588641112785981e-05}
{"train_loss": 0.09732204675674438, "global_step": 179378, "epoch": 2015, "lr": 2.5885903267664414e-05}
{"train_loss": 0.05894601717591286, "global_step": 179379, "epoch": 2015, "lr": 2.5885395410710844e-05}
{"train_loss": 0.10606265813112259, "global_step": 179380, "epoch": 2015, "lr": 2.588488755699921e-05}
{"train_loss": 0.11893081665039062, "global_step": 179381, "epoch": 2015, "lr": 2.588437970652954e-05}
{"train_loss": 0.10220053791999817, "global_step": 179382, "epoch": 2015, "lr": 2.588387185930194e-05}
{"train_loss": 0.07064621895551682, "global_step": 179383, "epoch": 2015, "lr": 2.5883364015316434e-05}
{"train_loss": 0.06755760312080383, "global_step": 179384, "epoch": 2015, "lr": 2.5882856174573144e-05}
{"train_loss": 0.08202888816595078, "global_step": 179385, "epoch": 2015, "lr": 2.588234833707209e-05}
{"train_loss": 0.08272448927164078, "global_step": 179386, "epoch": 2015, "lr": 2.5881840502813377e-05}
{"train_loss": 0.09220078587532043, "global_step": 179387, "epoch": 2015, "lr": 2.5881332671797036e-05}
{"train_loss": 0.10810305923223495, "global_step": 179388, "epoch": 2015, "lr": 2.588082484402318e-05}
{"train_loss": 0.04331796243786812, "global_step": 179389, "epoch": 2015, "lr": 2.5880317019491827e-05}
{"train_loss": 0.05832603946328163, "global_step": 179390, "epoch": 2015, "lr": 2.5879809198203098e-05}
{"train_loss": 0.14933602511882782, "global_step": 179391, "epoch": 2015, "lr": 2.5879301380157007e-05}
{"train_loss": 0.0558374859392643, "global_step": 179392, "epoch": 2015, "lr": 2.5878793565353653e-05}
{"train_loss": 0.11054952442646027, "global_step": 179393, "epoch": 2015, "lr": 2.587828575379313e-05}
{"train_loss": 0.06672517955303192, "global_step": 179394, "epoch": 2015, "lr": 2.5877777945475447e-05}
{"train_loss": 0.06685902923345566, "global_step": 179395, "epoch": 2015, "lr": 2.5877270140400724e-05}
{"train_loss": 0.098036989569664, "global_step": 179396, "epoch": 2015, "lr": 2.5876762338568995e-05}
{"train_loss": 0.07865343987941742, "global_step": 179397, "epoch": 2015, "lr": 2.587625453998035e-05}
{"train_loss": 0.08916100114583969, "global_step": 179398, "epoch": 2015, "lr": 2.587574674463484e-05}
{"train_loss": 0.07334175705909729, "global_step": 179399, "epoch": 2015, "lr": 2.5875238952532554e-05}
{"train_loss": 0.08246401697397232, "global_step": 179400, "epoch": 2015, "lr": 2.5874731163673528e-05}
{"train_loss": 0.03912913054227829, "global_step": 179401, "epoch": 2015, "lr": 2.5874223378057872e-05}
{"train_loss": 0.08535075187683105, "global_step": 179402, "epoch": 2015, "lr": 2.5873715595685617e-05}
{"train_loss": 0.0741732195019722, "global_step": 179403, "epoch": 2015, "lr": 2.5873207816556865e-05}
{"train_loss": 0.0700438842177391, "global_step": 179404, "epoch": 2015, "lr": 2.5872700040671643e-05}
{"train_loss": 0.08495690673589706, "global_step": 179405, "epoch": 2015, "lr": 2.587219226803006e-05}
{"train_loss": 0.09956757724285126, "global_step": 179406, "epoch": 2015, "lr": 2.5871684498632153e-05}
{"train_loss": 0.09356650710105896, "global_step": 179407, "epoch": 2015, "lr": 2.587117673247802e-05}
{"train_loss": 0.08004464954137802, "global_step": 179408, "epoch": 2015, "lr": 2.5870668969567697e-05}
{"train_loss": 0.04519763961434364, "global_step": 179409, "epoch": 2015, "lr": 2.5870161209901288e-05}
{"train_loss": 0.06106112524867058, "global_step": 179410, "epoch": 2015, "lr": 2.5869653453478836e-05}
{"train_loss": 0.10568909347057343, "global_step": 179411, "epoch": 2015, "lr": 2.5869145700300402e-05}
{"train_loss": 0.05019040033221245, "global_step": 179412, "epoch": 2015, "lr": 2.586863795036608e-05}
{"train_loss": 0.10318388789892197, "global_step": 179413, "epoch": 2015, "lr": 2.5868130203675912e-05}
{"train_loss": 0.08537398278713226, "global_step": 179414, "epoch": 2015, "lr": 2.586762246023e-05}
{"train_loss": 0.06657668203115463, "global_step": 179415, "epoch": 2015, "lr": 2.586711472002837e-05}
{"train_loss": 0.05616738647222519, "global_step": 179416, "epoch": 2015, "lr": 2.5866606983071128e-05}
{"train_loss": 0.058210570365190506, "global_step": 179417, "epoch": 2015, "lr": 2.586609924935831e-05}
{"train_loss": 0.030030183494091034, "global_step": 179418, "epoch": 2015, "lr": 2.586559151889002e-05}
{"train_loss": 0.10802750289440155, "global_step": 179419, "epoch": 2015, "lr": 2.586508379166629e-05}
{"train_loss": 0.07231976836919785, "global_step": 179420, "epoch": 2015, "lr": 2.5864576067687223e-05}
{"train_loss": 0.06324200332164764, "global_step": 179421, "epoch": 2015, "lr": 2.586406834695285e-05}
{"train_loss": 0.045826271176338196, "global_step": 179422, "epoch": 2015, "lr": 2.586356062946328e-05}
{"train_loss": 0.08071004406789715, "global_step": 179423, "epoch": 2015, "lr": 2.586305291521854e-05, "val_loss": 7.451312065124512, "train_action_mse_error": 20.15826988220215}
{"train_loss": 0.07067982107400894, "global_step": 179424, "epoch": 2016, "lr": 2.5862545204218737e-05}
{"train_loss": 0.056554730981588364, "global_step": 179425, "epoch": 2016, "lr": 2.5862037496463898e-05}
{"train_loss": 0.10428483039140701, "global_step": 179426, "epoch": 2016, "lr": 2.5861529791954138e-05}
{"train_loss": 0.06351712346076965, "global_step": 179427, "epoch": 2016, "lr": 2.586102209068948e-05}
{"train_loss": 0.10948014259338379, "global_step": 179428, "epoch": 2016, "lr": 2.586051439267001e-05}
{"train_loss": 0.10940930992364883, "global_step": 179429, "epoch": 2016, "lr": 2.5860006697895822e-05}
{"train_loss": 0.043569549918174744, "global_step": 179430, "epoch": 2016, "lr": 2.585949900636695e-05}
{"train_loss": 0.057323213666677475, "global_step": 179431, "epoch": 2016, "lr": 2.585899131808348e-05}
{"train_loss": 0.10094284266233444, "global_step": 179432, "epoch": 2016, "lr": 2.5858483633045465e-05}
{"train_loss": 0.04854290187358856, "global_step": 179433, "epoch": 2016, "lr": 2.5857975951253e-05}
{"train_loss": 0.0767342671751976, "global_step": 179434, "epoch": 2016, "lr": 2.5857468272706113e-05}
{"train_loss": 0.05072818323969841, "global_step": 179435, "epoch": 2016, "lr": 2.5856960597404918e-05}
{"train_loss": 0.10437461733818054, "global_step": 179436, "epoch": 2016, "lr": 2.5856452925349438e-05}
{"train_loss": 0.10190080106258392, "global_step": 179437, "epoch": 2016, "lr": 2.5855945256539782e-05}
{"train_loss": 0.06152999401092529, "global_step": 179438, "epoch": 2016, "lr": 2.5855437590975983e-05}
{"train_loss": 0.045105572789907455, "global_step": 179439, "epoch": 2016, "lr": 2.5854929928658146e-05}
{"train_loss": 0.07203496992588043, "global_step": 179440, "epoch": 2016, "lr": 2.5854422269586302e-05}
{"train_loss": 0.07581574469804764, "global_step": 179441, "epoch": 2016, "lr": 2.5853914613760556e-05}
{"train_loss": 0.05801912769675255, "global_step": 179442, "epoch": 2016, "lr": 2.585340696118094e-05}
{"train_loss": 0.0524241179227829, "global_step": 179443, "epoch": 2016, "lr": 2.5852899311847545e-05}
{"train_loss": 0.09003794938325882, "global_step": 179444, "epoch": 2016, "lr": 2.585239166576042e-05}
{"train_loss": 0.08253394067287445, "global_step": 179445, "epoch": 2016, "lr": 2.585188402291967e-05}
{"train_loss": 0.092849500477314, "global_step": 179446, "epoch": 2016, "lr": 2.585137638332532e-05}
{"train_loss": 0.10983522981405258, "global_step": 179447, "epoch": 2016, "lr": 2.5850868746977474e-05}
{"train_loss": 0.09174223989248276, "global_step": 179448, "epoch": 2016, "lr": 2.5850361113876193e-05}
{"train_loss": 0.09336636960506439, "global_step": 179449, "epoch": 2016, "lr": 2.5849853484021514e-05}
{"train_loss": 0.09115397185087204, "global_step": 179450, "epoch": 2016, "lr": 2.5849345857413542e-05}
{"train_loss": 0.07226588577032089, "global_step": 179451, "epoch": 2016, "lr": 2.584883823405231e-05}
{"train_loss": 0.0567479133605957, "global_step": 179452, "epoch": 2016, "lr": 2.5848330613937933e-05}
{"train_loss": 0.07705895602703094, "global_step": 179453, "epoch": 2016, "lr": 2.584782299707043e-05}
{"train_loss": 0.031032972037792206, "global_step": 179454, "epoch": 2016, "lr": 2.5847315383449915e-05}
{"train_loss": 0.06955277174711227, "global_step": 179455, "epoch": 2016, "lr": 2.5846807773076414e-05}
{"train_loss": 0.07834803313016891, "global_step": 179456, "epoch": 2016, "lr": 2.584630016595003e-05}
{"train_loss": 0.09013563394546509, "global_step": 179457, "epoch": 2016, "lr": 2.584579256207081e-05}
{"train_loss": 0.17246997356414795, "global_step": 179458, "epoch": 2016, "lr": 2.5845284961438836e-05}
{"train_loss": 0.09965101629495621, "global_step": 179459, "epoch": 2016, "lr": 2.584477736405416e-05}
{"train_loss": 0.1576348841190338, "global_step": 179460, "epoch": 2016, "lr": 2.5844269769916867e-05}
{"train_loss": 0.05477512627840042, "global_step": 179461, "epoch": 2016, "lr": 2.5843762179027004e-05}
{"train_loss": 0.09446676820516586, "global_step": 179462, "epoch": 2016, "lr": 2.584325459138467e-05}
{"train_loss": 0.052904631942510605, "global_step": 179463, "epoch": 2016, "lr": 2.5842747006989896e-05}
{"train_loss": 0.12690478563308716, "global_step": 179464, "epoch": 2016, "lr": 2.5842239425842773e-05}
{"train_loss": 0.07861431688070297, "global_step": 179465, "epoch": 2016, "lr": 2.5841731847943384e-05}
{"train_loss": 0.07558861374855042, "global_step": 179466, "epoch": 2016, "lr": 2.5841224273291763e-05}
{"train_loss": 0.039455436170101166, "global_step": 179467, "epoch": 2016, "lr": 2.5840716701888013e-05}
{"train_loss": 0.0740237757563591, "global_step": 179468, "epoch": 2016, "lr": 2.584020913373217e-05}
{"train_loss": 0.06543786078691483, "global_step": 179469, "epoch": 2016, "lr": 2.5839701568824327e-05}
{"train_loss": 0.06581616401672363, "global_step": 179470, "epoch": 2016, "lr": 2.5839194007164525e-05}
{"train_loss": 0.12670326232910156, "global_step": 179471, "epoch": 2016, "lr": 2.583868644875287e-05}
{"train_loss": 0.07200871407985687, "global_step": 179472, "epoch": 2016, "lr": 2.5838178893589394e-05}
{"train_loss": 0.13776209950447083, "global_step": 179473, "epoch": 2016, "lr": 2.583767134167419e-05}
{"train_loss": 0.06338921189308167, "global_step": 179474, "epoch": 2016, "lr": 2.58371637930073e-05}
{"train_loss": 0.07769137620925903, "global_step": 179475, "epoch": 2016, "lr": 2.583665624758883e-05}
{"train_loss": 0.08535383641719818, "global_step": 179476, "epoch": 2016, "lr": 2.5836148705418806e-05}
{"train_loss": 0.031512703746557236, "global_step": 179477, "epoch": 2016, "lr": 2.583564116649734e-05}
{"train_loss": 0.09170708060264587, "global_step": 179478, "epoch": 2016, "lr": 2.583513363082446e-05}
{"train_loss": 0.06871958822011948, "global_step": 179479, "epoch": 2016, "lr": 2.5834626098400268e-05}
{"train_loss": 0.04385364055633545, "global_step": 179480, "epoch": 2016, "lr": 2.5834118569224795e-05}
{"train_loss": 0.09037353098392487, "global_step": 179481, "epoch": 2016, "lr": 2.5833611043298152e-05}
{"train_loss": 0.03848772123456001, "global_step": 179482, "epoch": 2016, "lr": 2.583310352062036e-05}
{"train_loss": 0.05620448291301727, "global_step": 179483, "epoch": 2016, "lr": 2.5832596001191538e-05}
{"train_loss": 0.04500480368733406, "global_step": 179484, "epoch": 2016, "lr": 2.5832088485011706e-05}
{"train_loss": 0.061551935970783234, "global_step": 179485, "epoch": 2016, "lr": 2.583158097208098e-05}
{"train_loss": 0.056521765887737274, "global_step": 179486, "epoch": 2016, "lr": 2.583107346239938e-05}
{"train_loss": 0.06141354888677597, "global_step": 179487, "epoch": 2016, "lr": 2.583056595596702e-05}
{"train_loss": 0.0981065183877945, "global_step": 179488, "epoch": 2016, "lr": 2.5830058452783946e-05}
{"train_loss": 0.09342871606349945, "global_step": 179489, "epoch": 2016, "lr": 2.5829550952850202e-05}
{"train_loss": 0.07919791340827942, "global_step": 179490, "epoch": 2016, "lr": 2.58290434561659e-05}
{"train_loss": 0.07798119634389877, "global_step": 179491, "epoch": 2016, "lr": 2.5828535962731072e-05}
{"train_loss": 0.08945360034704208, "global_step": 179492, "epoch": 2016, "lr": 2.5828028472545818e-05}
{"train_loss": 0.09910363703966141, "global_step": 179493, "epoch": 2016, "lr": 2.5827520985610178e-05}
{"train_loss": 0.06529179215431213, "global_step": 179494, "epoch": 2016, "lr": 2.5827013501924248e-05}
{"train_loss": 0.06849928200244904, "global_step": 179495, "epoch": 2016, "lr": 2.582650602148806e-05}
{"train_loss": 0.10170703381299973, "global_step": 179496, "epoch": 2016, "lr": 2.5825998544301733e-05}
{"train_loss": 0.09587466716766357, "global_step": 179497, "epoch": 2016, "lr": 2.582549107036528e-05}
{"train_loss": 0.09888754785060883, "global_step": 179498, "epoch": 2016, "lr": 2.582498359967881e-05}
{"train_loss": 0.0396481528878212, "global_step": 179499, "epoch": 2016, "lr": 2.5824476132242366e-05}
{"train_loss": 0.12400327622890472, "global_step": 179500, "epoch": 2016, "lr": 2.582396866805602e-05}
{"train_loss": 0.09744011610746384, "global_step": 179501, "epoch": 2016, "lr": 2.582346120711987e-05}
{"train_loss": 0.07784025371074677, "global_step": 179502, "epoch": 2016, "lr": 2.5822953749433936e-05}
{"train_loss": 0.09219226986169815, "global_step": 179503, "epoch": 2016, "lr": 2.5822446294998337e-05}
{"train_loss": 0.05752090364694595, "global_step": 179504, "epoch": 2016, "lr": 2.5821938843813092e-05}
{"train_loss": 0.0885276049375534, "global_step": 179505, "epoch": 2016, "lr": 2.5821431395878315e-05}
{"train_loss": 0.05736830458045006, "global_step": 179506, "epoch": 2016, "lr": 2.582092395119403e-05}
{"train_loss": 0.10766148567199707, "global_step": 179507, "epoch": 2016, "lr": 2.5820416509760347e-05}
{"train_loss": 0.05558362230658531, "global_step": 179508, "epoch": 2016, "lr": 2.5819909071577297e-05}
{"train_loss": 0.08318792283535004, "global_step": 179509, "epoch": 2016, "lr": 2.581940163664498e-05}
{"train_loss": 0.12846039235591888, "global_step": 179510, "epoch": 2016, "lr": 2.5818894204963428e-05}
{"train_loss": 0.07806761562824249, "global_step": 179511, "epoch": 2016, "lr": 2.581838677653276e-05}
{"train_loss": 0.07976311357336098, "global_step": 179512, "epoch": 2016, "lr": 2.581787935135299e-05, "val_loss": 7.472245216369629}
{"train_loss": 0.04688924551010132, "global_step": 179513, "epoch": 2017, "lr": 2.5817371929424234e-05}
{"train_loss": 0.060454804450273514, "global_step": 179514, "epoch": 2017, "lr": 2.581686451074652e-05}
{"train_loss": 0.13425110280513763, "global_step": 179515, "epoch": 2017, "lr": 2.5816357095319943e-05}
{"train_loss": 0.04853251203894615, "global_step": 179516, "epoch": 2017, "lr": 2.5815849683144556e-05}
{"train_loss": 0.08974424749612808, "global_step": 179517, "epoch": 2017, "lr": 2.5815342274220444e-05}
{"train_loss": 0.09381693601608276, "global_step": 179518, "epoch": 2017, "lr": 2.5814834868547644e-05}
{"train_loss": 0.0497371181845665, "global_step": 179519, "epoch": 2017, "lr": 2.5814327466126248e-05}
{"train_loss": 0.03197290003299713, "global_step": 179520, "epoch": 2017, "lr": 2.581382006695634e-05}
{"train_loss": 0.09625190496444702, "global_step": 179521, "epoch": 2017, "lr": 2.581331267103797e-05}
{"train_loss": 0.06949080526828766, "global_step": 179522, "epoch": 2017, "lr": 2.5812805278371183e-05}
{"train_loss": 0.07369101047515869, "global_step": 179523, "epoch": 2017, "lr": 2.5812297888956093e-05}
{"train_loss": 0.10111754387617111, "global_step": 179524, "epoch": 2017, "lr": 2.581179050279272e-05}
{"train_loss": 0.05804416537284851, "global_step": 179525, "epoch": 2017, "lr": 2.581128311988118e-05}
{"train_loss": 0.11043297499418259, "global_step": 179526, "epoch": 2017, "lr": 2.5810775740221515e-05}
{"train_loss": 0.061738330870866776, "global_step": 179527, "epoch": 2017, "lr": 2.581026836381378e-05}
{"train_loss": 0.0968700721859932, "global_step": 179528, "epoch": 2017, "lr": 2.5809760990658082e-05}
{"train_loss": 0.039540719240903854, "global_step": 179529, "epoch": 2017, "lr": 2.5809253620754447e-05}
{"train_loss": 0.08436901122331619, "global_step": 179530, "epoch": 2017, "lr": 2.5808746254102977e-05}
{"train_loss": 0.06837638467550278, "global_step": 179531, "epoch": 2017, "lr": 2.5808238890703708e-05}
{"train_loss": 0.09048417210578918, "global_step": 179532, "epoch": 2017, "lr": 2.5807731530556744e-05}
{"train_loss": 0.0719938725233078, "global_step": 179533, "epoch": 2017, "lr": 2.5807224173662125e-05}
{"train_loss": 0.11285156011581421, "global_step": 179534, "epoch": 2017, "lr": 2.5806716820019937e-05}
{"train_loss": 0.074593186378479, "global_step": 179535, "epoch": 2017, "lr": 2.5806209469630233e-05}
{"train_loss": 0.11324477940797806, "global_step": 179536, "epoch": 2017, "lr": 2.580570212249308e-05}
{"train_loss": 0.050713829696178436, "global_step": 179537, "epoch": 2017, "lr": 2.5805194778608584e-05}
{"train_loss": 0.1134433001279831, "global_step": 179538, "epoch": 2017, "lr": 2.580468743797676e-05}
{"train_loss": 0.1042192131280899, "global_step": 179539, "epoch": 2017, "lr": 2.580418010059772e-05}
{"train_loss": 0.12261838465929031, "global_step": 179540, "epoch": 2017, "lr": 2.58036727664715e-05}
{"train_loss": 0.04878903552889824, "global_step": 179541, "epoch": 2017, "lr": 2.5803165435598187e-05}
{"train_loss": 0.04994660243391991, "global_step": 179542, "epoch": 2017, "lr": 2.580265810797784e-05}
{"train_loss": 0.08689470589160919, "global_step": 179543, "epoch": 2017, "lr": 2.5802150783610534e-05}
{"train_loss": 0.08634141832590103, "global_step": 179544, "epoch": 2017, "lr": 2.5801643462496328e-05}
{"train_loss": 0.14718227088451385, "global_step": 179545, "epoch": 2017, "lr": 2.580113614463531e-05}
{"train_loss": 0.054654791951179504, "global_step": 179546, "epoch": 2017, "lr": 2.580062883002752e-05}
{"train_loss": 0.05510083585977554, "global_step": 179547, "epoch": 2017, "lr": 2.5800121518673047e-05}
{"train_loss": 0.10649870336055756, "global_step": 179548, "epoch": 2017, "lr": 2.5799614210571945e-05}
{"train_loss": 0.08263418078422546, "global_step": 179549, "epoch": 2017, "lr": 2.5799106905724302e-05}
{"train_loss": 0.08332222700119019, "global_step": 179550, "epoch": 2017, "lr": 2.5798599604130157e-05}
{"train_loss": 0.06632882356643677, "global_step": 179551, "epoch": 2017, "lr": 2.5798092305789622e-05}
{"train_loss": 0.05819610133767128, "global_step": 179552, "epoch": 2017, "lr": 2.5797585010702713e-05}
{"train_loss": 0.09970053285360336, "global_step": 179553, "epoch": 2017, "lr": 2.5797077718869545e-05}
{"train_loss": 0.09365727752447128, "global_step": 179554, "epoch": 2017, "lr": 2.579657043029014e-05}
{"train_loss": 0.07796841114759445, "global_step": 179555, "epoch": 2017, "lr": 2.57960631449646e-05}
{"train_loss": 0.05999341979622841, "global_step": 179556, "epoch": 2017, "lr": 2.5795555862892997e-05}
{"train_loss": 0.05904367193579674, "global_step": 179557, "epoch": 2017, "lr": 2.5795048584075364e-05}
{"train_loss": 0.09336385875940323, "global_step": 179558, "epoch": 2017, "lr": 2.5794541308511823e-05}
{"train_loss": 0.05590565502643585, "global_step": 179559, "epoch": 2017, "lr": 2.5794034036202395e-05}
{"train_loss": 0.04976929351687431, "global_step": 179560, "epoch": 2017, "lr": 2.5793526767147157e-05}
{"train_loss": 0.09465664625167847, "global_step": 179561, "epoch": 2017, "lr": 2.5793019501346205e-05}
{"train_loss": 0.07961377501487732, "global_step": 179562, "epoch": 2017, "lr": 2.579251223879956e-05}
{"train_loss": 0.09551876038312912, "global_step": 179563, "epoch": 2017, "lr": 2.5792004979507332e-05}
{"train_loss": 0.0708845779299736, "global_step": 179564, "epoch": 2017, "lr": 2.5791497723469567e-05}
{"train_loss": 0.03922119364142418, "global_step": 179565, "epoch": 2017, "lr": 2.579099047068635e-05}
{"train_loss": 0.06081642210483551, "global_step": 179566, "epoch": 2017, "lr": 2.5790483221157748e-05}
{"train_loss": 0.13144966959953308, "global_step": 179567, "epoch": 2017, "lr": 2.5789975974883796e-05}
{"train_loss": 0.07958924025297165, "global_step": 179568, "epoch": 2017, "lr": 2.57894687318646e-05}
{"train_loss": 0.19438223540782928, "global_step": 179569, "epoch": 2017, "lr": 2.5788961492100206e-05}
{"train_loss": 0.03284255787730217, "global_step": 179570, "epoch": 2017, "lr": 2.5788454255590704e-05}
{"train_loss": 0.052276916801929474, "global_step": 179571, "epoch": 2017, "lr": 2.5787947022336133e-05}
{"train_loss": 0.07684575021266937, "global_step": 179572, "epoch": 2017, "lr": 2.5787439792336594e-05}
{"train_loss": 0.05091124027967453, "global_step": 179573, "epoch": 2017, "lr": 2.578693256559212e-05}
{"train_loss": 0.05010388791561127, "global_step": 179574, "epoch": 2017, "lr": 2.57864253421028e-05}
{"train_loss": 0.0778656154870987, "global_step": 179575, "epoch": 2017, "lr": 2.5785918121868714e-05}
{"train_loss": 0.07477028667926788, "global_step": 179576, "epoch": 2017, "lr": 2.5785410904889905e-05}
{"train_loss": 0.05593843013048172, "global_step": 179577, "epoch": 2017, "lr": 2.5784903691166463e-05}
{"train_loss": 0.0750722661614418, "global_step": 179578, "epoch": 2017, "lr": 2.578439648069843e-05}
{"train_loss": 0.07285596430301666, "global_step": 179579, "epoch": 2017, "lr": 2.578388927348591e-05}
{"train_loss": 0.1427442878484726, "global_step": 179580, "epoch": 2017, "lr": 2.5783382069528932e-05}
{"train_loss": 0.11729709059000015, "global_step": 179581, "epoch": 2017, "lr": 2.57828748688276e-05}
{"train_loss": 0.11848024278879166, "global_step": 179582, "epoch": 2017, "lr": 2.5782367671381945e-05}
{"train_loss": 0.06278003007173538, "global_step": 179583, "epoch": 2017, "lr": 2.5781860477192083e-05}
{"train_loss": 0.09208326786756516, "global_step": 179584, "epoch": 2017, "lr": 2.5781353286258025e-05}
{"train_loss": 0.09868848323822021, "global_step": 179585, "epoch": 2017, "lr": 2.5780846098579898e-05}
{"train_loss": 0.09617090225219727, "global_step": 179586, "epoch": 2017, "lr": 2.5780338914157715e-05}
{"train_loss": 0.03446875140070915, "global_step": 179587, "epoch": 2017, "lr": 2.5779831732991587e-05}
{"train_loss": 0.10600723326206207, "global_step": 179588, "epoch": 2017, "lr": 2.5779324555081552e-05}
{"train_loss": 0.05095010995864868, "global_step": 179589, "epoch": 2017, "lr": 2.5778817380427712e-05}
{"train_loss": 0.1013636663556099, "global_step": 179590, "epoch": 2017, "lr": 2.5778310209030094e-05}
{"train_loss": 0.12461996078491211, "global_step": 179591, "epoch": 2017, "lr": 2.577780304088878e-05}
{"train_loss": 0.03369709476828575, "global_step": 179592, "epoch": 2017, "lr": 2.577729587600387e-05}
{"train_loss": 0.11393866688013077, "global_step": 179593, "epoch": 2017, "lr": 2.5776788714375388e-05}
{"train_loss": 0.08770810067653656, "global_step": 179594, "epoch": 2017, "lr": 2.5776281556003445e-05}
{"train_loss": 0.07660134881734848, "global_step": 179595, "epoch": 2017, "lr": 2.5775774400888063e-05}
{"train_loss": 0.11026246100664139, "global_step": 179596, "epoch": 2017, "lr": 2.577526724902935e-05}
{"train_loss": 0.10392755270004272, "global_step": 179597, "epoch": 2017, "lr": 2.5774760100427343e-05}
{"train_loss": 0.06405071169137955, "global_step": 179598, "epoch": 2017, "lr": 2.5774252955082134e-05}
{"train_loss": 0.04562768340110779, "global_step": 179599, "epoch": 2017, "lr": 2.5773745812993786e-05}
{"train_loss": 0.0645042210817337, "global_step": 179600, "epoch": 2017, "lr": 2.577323867416235e-05}
{"train_loss": 0.08027682976608866, "global_step": 179601, "epoch": 2017, "lr": 2.577273153858792e-05, "val_loss": 7.503899574279785}
{"train_loss": 0.12223771214485168, "global_step": 179602, "epoch": 2018, "lr": 2.577222440627053e-05}
{"train_loss": 0.044728536158800125, "global_step": 179603, "epoch": 2018, "lr": 2.5771717277210295e-05}
{"train_loss": 0.09246306121349335, "global_step": 179604, "epoch": 2018, "lr": 2.5771210151407253e-05}
{"train_loss": 0.05135287716984749, "global_step": 179605, "epoch": 2018, "lr": 2.5770703028861454e-05}
{"train_loss": 0.10174424201250076, "global_step": 179606, "epoch": 2018, "lr": 2.5770195909573014e-05}
{"train_loss": 0.11733343452215195, "global_step": 179607, "epoch": 2018, "lr": 2.576968879354195e-05}
{"train_loss": 0.08514349162578583, "global_step": 179608, "epoch": 2018, "lr": 2.5769181680768384e-05}
{"train_loss": 0.07224103808403015, "global_step": 179609, "epoch": 2018, "lr": 2.5768674571252327e-05}
{"train_loss": 0.08085063099861145, "global_step": 179610, "epoch": 2018, "lr": 2.5768167464993886e-05}
{"train_loss": 0.1125185638666153, "global_step": 179611, "epoch": 2018, "lr": 2.5767660361993128e-05}
{"train_loss": 0.07651032507419586, "global_step": 179612, "epoch": 2018, "lr": 2.57671532622501e-05}
{"train_loss": 0.09235373139381409, "global_step": 179613, "epoch": 2018, "lr": 2.5766646165764908e-05}
{"train_loss": 0.08653205633163452, "global_step": 179614, "epoch": 2018, "lr": 2.5766139072537564e-05}
{"train_loss": 0.08460056781768799, "global_step": 179615, "epoch": 2018, "lr": 2.5765631982568194e-05}
{"train_loss": 0.10273437947034836, "global_step": 179616, "epoch": 2018, "lr": 2.5765124895856818e-05}
{"train_loss": 0.06648165732622147, "global_step": 179617, "epoch": 2018, "lr": 2.5764617812403547e-05}
{"train_loss": 0.05532490834593773, "global_step": 179618, "epoch": 2018, "lr": 2.5764110732208402e-05}
{"train_loss": 0.06191118434071541, "global_step": 179619, "epoch": 2018, "lr": 2.5763603655271505e-05}
{"train_loss": 0.05338601395487785, "global_step": 179620, "epoch": 2018, "lr": 2.5763096581592872e-05}
{"train_loss": 0.07633113861083984, "global_step": 179621, "epoch": 2018, "lr": 2.5762589511172613e-05}
{"train_loss": 0.08491921424865723, "global_step": 179622, "epoch": 2018, "lr": 2.576208244401076e-05}
{"train_loss": 0.07664801925420761, "global_step": 179623, "epoch": 2018, "lr": 2.5761575380107423e-05}
{"train_loss": 0.085389643907547, "global_step": 179624, "epoch": 2018, "lr": 2.5761068319462622e-05}
{"train_loss": 0.15172220766544342, "global_step": 179625, "epoch": 2018, "lr": 2.576056126207648e-05}
{"train_loss": 0.07034438103437424, "global_step": 179626, "epoch": 2018, "lr": 2.5760054207949003e-05}
{"train_loss": 0.05422380939126015, "global_step": 179627, "epoch": 2018, "lr": 2.5759547157080293e-05}
{"train_loss": 0.10057123005390167, "global_step": 179628, "epoch": 2018, "lr": 2.5759040109470438e-05}
{"train_loss": 0.052810151129961014, "global_step": 179629, "epoch": 2018, "lr": 2.5758533065119467e-05}
{"train_loss": 0.11194945126771927, "global_step": 179630, "epoch": 2018, "lr": 2.5758026024027483e-05}
{"train_loss": 0.0553385354578495, "global_step": 179631, "epoch": 2018, "lr": 2.575751898619452e-05}
{"train_loss": 0.06546191871166229, "global_step": 179632, "epoch": 2018, "lr": 2.575701195162068e-05}
{"train_loss": 0.0632147416472435, "global_step": 179633, "epoch": 2018, "lr": 2.5756504920305997e-05}
{"train_loss": 0.1356123983860016, "global_step": 179634, "epoch": 2018, "lr": 2.5755997892250572e-05}
{"train_loss": 0.09505752474069595, "global_step": 179635, "epoch": 2018, "lr": 2.5755490867454444e-05}
{"train_loss": 0.0602077953517437, "global_step": 179636, "epoch": 2018, "lr": 2.5754983845917713e-05}
{"train_loss": 0.06053157523274422, "global_step": 179637, "epoch": 2018, "lr": 2.575447682764043e-05}
{"train_loss": 0.13278518617153168, "global_step": 179638, "epoch": 2018, "lr": 2.5753969812622637e-05}
{"train_loss": 0.10954449325799942, "global_step": 179639, "epoch": 2018, "lr": 2.5753462800864447e-05}
{"train_loss": 0.08859498798847198, "global_step": 179640, "epoch": 2018, "lr": 2.575295579236589e-05}
{"train_loss": 0.14315640926361084, "global_step": 179641, "epoch": 2018, "lr": 2.575244878712707e-05}
{"train_loss": 0.08882179856300354, "global_step": 179642, "epoch": 2018, "lr": 2.5751941785148048e-05}
{"train_loss": 0.09828569740056992, "global_step": 179643, "epoch": 2018, "lr": 2.5751434786428853e-05}
{"train_loss": 0.08756557106971741, "global_step": 179644, "epoch": 2018, "lr": 2.5750927790969603e-05}
{"train_loss": 0.08759532868862152, "global_step": 179645, "epoch": 2018, "lr": 2.5750420798770326e-05}
{"train_loss": 0.12387459725141525, "global_step": 179646, "epoch": 2018, "lr": 2.5749913809831112e-05}
{"train_loss": 0.06846332550048828, "global_step": 179647, "epoch": 2018, "lr": 2.5749406824152044e-05}
{"train_loss": 0.05138011649250984, "global_step": 179648, "epoch": 2018, "lr": 2.5748899841733154e-05}
{"train_loss": 0.05381675809621811, "global_step": 179649, "epoch": 2018, "lr": 2.5748392862574545e-05}
{"train_loss": 0.07946161925792694, "global_step": 179650, "epoch": 2018, "lr": 2.5747885886676248e-05}
{"train_loss": 0.11392181366682053, "global_step": 179651, "epoch": 2018, "lr": 2.574737891403838e-05}
{"train_loss": 0.06889961659908295, "global_step": 179652, "epoch": 2018, "lr": 2.5746871944660955e-05}
{"train_loss": 0.07217121124267578, "global_step": 179653, "epoch": 2018, "lr": 2.5746364978544092e-05}
{"train_loss": 0.07812700420618057, "global_step": 179654, "epoch": 2018, "lr": 2.574585801568781e-05}
{"train_loss": 0.09650090336799622, "global_step": 179655, "epoch": 2018, "lr": 2.5745351056092228e-05}
{"train_loss": 0.06271247565746307, "global_step": 179656, "epoch": 2018, "lr": 2.574484409975736e-05}
{"train_loss": 0.12351550161838531, "global_step": 179657, "epoch": 2018, "lr": 2.574433714668333e-05}
{"train_loss": 0.0770639106631279, "global_step": 179658, "epoch": 2018, "lr": 2.5743830196870157e-05}
{"train_loss": 0.041277725249528885, "global_step": 179659, "epoch": 2018, "lr": 2.5743323250317942e-05}
{"train_loss": 0.10476868599653244, "global_step": 179660, "epoch": 2018, "lr": 2.574281630702673e-05}
{"train_loss": 0.10168618708848953, "global_step": 179661, "epoch": 2018, "lr": 2.574230936699662e-05}
{"train_loss": 0.03843875229358673, "global_step": 179662, "epoch": 2018, "lr": 2.5741802430227634e-05}
{"train_loss": 0.04587843641638756, "global_step": 179663, "epoch": 2018, "lr": 2.5741295496719876e-05}
{"train_loss": 0.10692425072193146, "global_step": 179664, "epoch": 2018, "lr": 2.5740788566473416e-05}
{"train_loss": 0.10112572461366653, "global_step": 179665, "epoch": 2018, "lr": 2.5740281639488296e-05}
{"train_loss": 0.06530305743217468, "global_step": 179666, "epoch": 2018, "lr": 2.573977471576462e-05}
{"train_loss": 0.08247755467891693, "global_step": 179667, "epoch": 2018, "lr": 2.5739267795302418e-05}
{"train_loss": 0.12148533016443253, "global_step": 179668, "epoch": 2018, "lr": 2.5738760878101793e-05}
{"train_loss": 0.08607200533151627, "global_step": 179669, "epoch": 2018, "lr": 2.5738253964162773e-05}
{"train_loss": 0.11368145793676376, "global_step": 179670, "epoch": 2018, "lr": 2.5737747053485473e-05}
{"train_loss": 0.04827040433883667, "global_step": 179671, "epoch": 2018, "lr": 2.5737240146069918e-05}
{"train_loss": 0.07126742601394653, "global_step": 179672, "epoch": 2018, "lr": 2.5736733241916212e-05}
{"train_loss": 0.11213900148868561, "global_step": 179673, "epoch": 2018, "lr": 2.573622634102439e-05}
{"train_loss": 0.08935115486383438, "global_step": 179674, "epoch": 2018, "lr": 2.573571944339456e-05}
{"train_loss": 0.08585494756698608, "global_step": 179675, "epoch": 2018, "lr": 2.5735212549026756e-05}
{"train_loss": 0.059859152883291245, "global_step": 179676, "epoch": 2018, "lr": 2.5734705657921044e-05}
{"train_loss": 0.08840804547071457, "global_step": 179677, "epoch": 2018, "lr": 2.5734198770077523e-05}
{"train_loss": 0.04245610535144806, "global_step": 179678, "epoch": 2018, "lr": 2.5733691885496225e-05}
{"train_loss": 0.04692968726158142, "global_step": 179679, "epoch": 2018, "lr": 2.573318500417725e-05}
{"train_loss": 0.09486768394708633, "global_step": 179680, "epoch": 2018, "lr": 2.573267812612064e-05}
{"train_loss": 0.0791003555059433, "global_step": 179681, "epoch": 2018, "lr": 2.573217125132649e-05}
{"train_loss": 0.03473476320505142, "global_step": 179682, "epoch": 2018, "lr": 2.573166437979484e-05}
{"train_loss": 0.056164566427469254, "global_step": 179683, "epoch": 2018, "lr": 2.573115751152579e-05}
{"train_loss": 0.09787286818027496, "global_step": 179684, "epoch": 2018, "lr": 2.5730650646519365e-05}
{"train_loss": 0.10648784041404724, "global_step": 179685, "epoch": 2018, "lr": 2.5730143784775684e-05}
{"train_loss": 0.08378319442272186, "global_step": 179686, "epoch": 2018, "lr": 2.5729636926294764e-05}
{"train_loss": 0.097832091152668, "global_step": 179687, "epoch": 2018, "lr": 2.5729130071076723e-05}
{"train_loss": 0.12087124586105347, "global_step": 179688, "epoch": 2018, "lr": 2.5728623219121585e-05}
{"train_loss": 0.04759208485484123, "global_step": 179689, "epoch": 2018, "lr": 2.5728116370429455e-05}
{"train_loss": 0.08365581586454691, "global_step": 179690, "epoch": 2018, "lr": 2.5727609525000367e-05, "val_loss": 7.364720821380615}
{"train_loss": 0.03597148135304451, "global_step": 179691, "epoch": 2019, "lr": 2.5727102682834426e-05}
{"train_loss": 0.06751461327075958, "global_step": 179692, "epoch": 2019, "lr": 2.5726595843931656e-05}
{"train_loss": 0.07476157695055008, "global_step": 179693, "epoch": 2019, "lr": 2.5726089008292176e-05}
{"train_loss": 0.08705401420593262, "global_step": 179694, "epoch": 2019, "lr": 2.5725582175916e-05}
{"train_loss": 0.08139587193727493, "global_step": 179695, "epoch": 2019, "lr": 2.572507534680325e-05}
{"train_loss": 0.07690717279911041, "global_step": 179696, "epoch": 2019, "lr": 2.5724568520953947e-05}
{"train_loss": 0.04648275673389435, "global_step": 179697, "epoch": 2019, "lr": 2.5724061698368196e-05}
{"train_loss": 0.055249009281396866, "global_step": 179698, "epoch": 2019, "lr": 2.572355487904603e-05}
{"train_loss": 0.09884267300367355, "global_step": 179699, "epoch": 2019, "lr": 2.5723048062987554e-05}
{"train_loss": 0.06593949347734451, "global_step": 179700, "epoch": 2019, "lr": 2.57225412501928e-05}
{"train_loss": 0.11044518649578094, "global_step": 179701, "epoch": 2019, "lr": 2.572203444066186e-05}
{"train_loss": 0.04929475858807564, "global_step": 179702, "epoch": 2019, "lr": 2.5721527634394814e-05}
{"train_loss": 0.12423693388700485, "global_step": 179703, "epoch": 2019, "lr": 2.572102083139169e-05}
{"train_loss": 0.12381259351968765, "global_step": 179704, "epoch": 2019, "lr": 2.57205140316526e-05}
{"train_loss": 0.12035368382930756, "global_step": 179705, "epoch": 2019, "lr": 2.5720007235177566e-05}
{"train_loss": 0.08241404592990875, "global_step": 179706, "epoch": 2019, "lr": 2.571950044196671e-05}
{"train_loss": 0.06207221746444702, "global_step": 179707, "epoch": 2019, "lr": 2.5718993652020047e-05}
{"train_loss": 0.07511413097381592, "global_step": 179708, "epoch": 2019, "lr": 2.5718486865337688e-05}
{"train_loss": 0.06675763428211212, "global_step": 179709, "epoch": 2019, "lr": 2.571798008191967e-05}
{"train_loss": 0.083223357796669, "global_step": 179710, "epoch": 2019, "lr": 2.571747330176608e-05}
{"train_loss": 0.11349336057901382, "global_step": 179711, "epoch": 2019, "lr": 2.5716966524876968e-05}
{"train_loss": 0.09762498736381531, "global_step": 179712, "epoch": 2019, "lr": 2.571645975125243e-05}
{"train_loss": 0.062352631241083145, "global_step": 179713, "epoch": 2019, "lr": 2.5715952980892506e-05}
{"train_loss": 0.09999991208314896, "global_step": 179714, "epoch": 2019, "lr": 2.571544621379729e-05}
{"train_loss": 0.06324797868728638, "global_step": 179715, "epoch": 2019, "lr": 2.5714939449966836e-05}
{"train_loss": 0.02848171442747116, "global_step": 179716, "epoch": 2019, "lr": 2.571443268940119e-05}
{"train_loss": 0.12496120482683182, "global_step": 179717, "epoch": 2019, "lr": 2.571392593210047e-05}
{"train_loss": 0.12214764952659607, "global_step": 179718, "epoch": 2019, "lr": 2.5713419178064695e-05}
{"train_loss": 0.09645976126194, "global_step": 179719, "epoch": 2019, "lr": 2.571291242729397e-05}
{"train_loss": 0.10524266213178635, "global_step": 179720, "epoch": 2019, "lr": 2.5712405679788333e-05}
{"train_loss": 0.1142822802066803, "global_step": 179721, "epoch": 2019, "lr": 2.5711898935547885e-05}
{"train_loss": 0.052791085094213486, "global_step": 179722, "epoch": 2019, "lr": 2.5711392194572657e-05}
{"train_loss": 0.09179434180259705, "global_step": 179723, "epoch": 2019, "lr": 2.5710885456862755e-05}
{"train_loss": 0.05785830318927765, "global_step": 179724, "epoch": 2019, "lr": 2.5710378722418206e-05}
{"train_loss": 0.06954597681760788, "global_step": 179725, "epoch": 2019, "lr": 2.5709871991239127e-05}
{"train_loss": 0.04753832519054413, "global_step": 179726, "epoch": 2019, "lr": 2.5709365263325534e-05}
{"train_loss": 0.11786067485809326, "global_step": 179727, "epoch": 2019, "lr": 2.5708858538677537e-05}
{"train_loss": 0.0502953864634037, "global_step": 179728, "epoch": 2019, "lr": 2.5708351817295178e-05}
{"train_loss": 0.04919400066137314, "global_step": 179729, "epoch": 2019, "lr": 2.570784509917855e-05}
{"train_loss": 0.04366741701960564, "global_step": 179730, "epoch": 2019, "lr": 2.570733838432769e-05}
{"train_loss": 0.10262490808963776, "global_step": 179731, "epoch": 2019, "lr": 2.57068316727427e-05}
{"train_loss": 0.057782869786024094, "global_step": 179732, "epoch": 2019, "lr": 2.5706324964423606e-05}
{"train_loss": 0.059186991304159164, "global_step": 179733, "epoch": 2019, "lr": 2.5705818259370524e-05}
{"train_loss": 0.07787581533193588, "global_step": 179734, "epoch": 2019, "lr": 2.570531155758348e-05}
{"train_loss": 0.05924089998006821, "global_step": 179735, "epoch": 2019, "lr": 2.570480485906258e-05}
{"train_loss": 0.1241956502199173, "global_step": 179736, "epoch": 2019, "lr": 2.570429816380785e-05}
{"train_loss": 0.0652599185705185, "global_step": 179737, "epoch": 2019, "lr": 2.5703791471819383e-05}
{"train_loss": 0.09136369079351425, "global_step": 179738, "epoch": 2019, "lr": 2.5703284783097265e-05}
{"train_loss": 0.04780144244432449, "global_step": 179739, "epoch": 2019, "lr": 2.5702778097641524e-05}
{"train_loss": 0.05651045963168144, "global_step": 179740, "epoch": 2019, "lr": 2.5702271415452273e-05}
{"train_loss": 0.08516210317611694, "global_step": 179741, "epoch": 2019, "lr": 2.5701764736529534e-05}
{"train_loss": 0.07257963716983795, "global_step": 179742, "epoch": 2019, "lr": 2.570125806087341e-05}
{"train_loss": 0.0774112120270729, "global_step": 179743, "epoch": 2019, "lr": 2.5700751388483935e-05}
{"train_loss": 0.09896626323461533, "global_step": 179744, "epoch": 2019, "lr": 2.570024471936122e-05}
{"train_loss": 0.07536664605140686, "global_step": 179745, "epoch": 2019, "lr": 2.5699738053505297e-05}
{"train_loss": 0.025298304855823517, "global_step": 179746, "epoch": 2019, "lr": 2.5699231390916267e-05}
{"train_loss": 0.07436731457710266, "global_step": 179747, "epoch": 2019, "lr": 2.5698724731594158e-05}
{"train_loss": 0.0801483690738678, "global_step": 179748, "epoch": 2019, "lr": 2.5698218075539072e-05}
{"train_loss": 0.06383277475833893, "global_step": 179749, "epoch": 2019, "lr": 2.5697711422751054e-05}
{"train_loss": 0.10622738301753998, "global_step": 179750, "epoch": 2019, "lr": 2.56972047732302e-05}
{"train_loss": 0.06332515180110931, "global_step": 179751, "epoch": 2019, "lr": 2.5696698126976536e-05}
{"train_loss": 0.08467185497283936, "global_step": 179752, "epoch": 2019, "lr": 2.569619148399018e-05}
{"train_loss": 0.05842852592468262, "global_step": 179753, "epoch": 2019, "lr": 2.5695684844271174e-05}
{"train_loss": 0.06660419702529907, "global_step": 179754, "epoch": 2019, "lr": 2.5695178207819558e-05}
{"train_loss": 0.05205567553639412, "global_step": 179755, "epoch": 2019, "lr": 2.5694671574635466e-05}
{"train_loss": 0.06308845430612564, "global_step": 179756, "epoch": 2019, "lr": 2.5694164944718896e-05}
{"train_loss": 0.12172498553991318, "global_step": 179757, "epoch": 2019, "lr": 2.5693658318069968e-05}
{"train_loss": 0.07092569768428802, "global_step": 179758, "epoch": 2019, "lr": 2.569315169468872e-05}
{"train_loss": 0.06451878696680069, "global_step": 179759, "epoch": 2019, "lr": 2.569264507457525e-05}
{"train_loss": 0.055623237043619156, "global_step": 179760, "epoch": 2019, "lr": 2.5692138457729586e-05}
{"train_loss": 0.053326454013586044, "global_step": 179761, "epoch": 2019, "lr": 2.5691631844151837e-05}
{"train_loss": 0.03515741601586342, "global_step": 179762, "epoch": 2019, "lr": 2.569112523384203e-05}
{"train_loss": 0.10542695224285126, "global_step": 179763, "epoch": 2019, "lr": 2.5690618626800278e-05}
{"train_loss": 0.07574081420898438, "global_step": 179764, "epoch": 2019, "lr": 2.5690112023026604e-05}
{"train_loss": 0.05072493106126785, "global_step": 179765, "epoch": 2019, "lr": 2.5689605422521124e-05}
{"train_loss": 0.07279680669307709, "global_step": 179766, "epoch": 2019, "lr": 2.5689098825283854e-05}
{"train_loss": 0.16577929258346558, "global_step": 179767, "epoch": 2019, "lr": 2.568859223131491e-05}
{"train_loss": 0.09993031620979309, "global_step": 179768, "epoch": 2019, "lr": 2.568808564061432e-05}
{"train_loss": 0.059492576867341995, "global_step": 179769, "epoch": 2019, "lr": 2.5687579053182187e-05}
{"train_loss": 0.10875269770622253, "global_step": 179770, "epoch": 2019, "lr": 2.568707246901855e-05}
{"train_loss": 0.03772468492388725, "global_step": 179771, "epoch": 2019, "lr": 2.5686565888123498e-05}
{"train_loss": 0.032636094838380814, "global_step": 179772, "epoch": 2019, "lr": 2.568605931049708e-05}
{"train_loss": 0.08738131821155548, "global_step": 179773, "epoch": 2019, "lr": 2.5685552736139372e-05}
{"train_loss": 0.050844497978687286, "global_step": 179774, "epoch": 2019, "lr": 2.5685046165050463e-05}
{"train_loss": 0.027046171948313713, "global_step": 179775, "epoch": 2019, "lr": 2.5684539597230383e-05}
{"train_loss": 0.09000616520643234, "global_step": 179776, "epoch": 2019, "lr": 2.5684033032679243e-05}
{"train_loss": 0.1063125804066658, "global_step": 179777, "epoch": 2019, "lr": 2.5683526471397068e-05}
{"train_loss": 0.08104778081178665, "global_step": 179778, "epoch": 2019, "lr": 2.5683019913383965e-05}
{"train_loss": 0.07697378435915106, "global_step": 179779, "epoch": 2019, "lr": 2.568251335863997e-05, "val_loss": 7.4654717445373535}
{"train_loss": 0.09503699094057083, "global_step": 179780, "epoch": 2020, "lr": 2.5682006807165182e-05}
{"train_loss": 0.09981401264667511, "global_step": 179781, "epoch": 2020, "lr": 2.5681500258959624e-05}
{"train_loss": 0.09407593309879303, "global_step": 179782, "epoch": 2020, "lr": 2.5680993714023422e-05}
{"train_loss": 0.05231829360127449, "global_step": 179783, "epoch": 2020, "lr": 2.568048717235659e-05}
{"train_loss": 0.06352362781763077, "global_step": 179784, "epoch": 2020, "lr": 2.5679980633959238e-05}
{"train_loss": 0.060577474534511566, "global_step": 179785, "epoch": 2020, "lr": 2.56794740988314e-05}
{"train_loss": 0.04637790471315384, "global_step": 179786, "epoch": 2020, "lr": 2.5678967566973178e-05}
{"train_loss": 0.0886848196387291, "global_step": 179787, "epoch": 2020, "lr": 2.5678461038384604e-05}
{"train_loss": 0.055302128195762634, "global_step": 179788, "epoch": 2020, "lr": 2.5677954513065784e-05}
{"train_loss": 0.11551713943481445, "global_step": 179789, "epoch": 2020, "lr": 2.5677447991016745e-05}
{"train_loss": 0.0743764340877533, "global_step": 179790, "epoch": 2020, "lr": 2.5676941472237592e-05}
{"train_loss": 0.03906626254320145, "global_step": 179791, "epoch": 2020, "lr": 2.5676434956728386e-05}
{"train_loss": 0.0775260254740715, "global_step": 179792, "epoch": 2020, "lr": 2.567592844448916e-05}
{"train_loss": 0.18400682508945465, "global_step": 179793, "epoch": 2020, "lr": 2.567542193552004e-05}
{"train_loss": 0.04475323483347893, "global_step": 179794, "epoch": 2020, "lr": 2.5674915429821033e-05}
{"train_loss": 0.07635904848575592, "global_step": 179795, "epoch": 2020, "lr": 2.567440892739226e-05}
{"train_loss": 0.07523061335086823, "global_step": 179796, "epoch": 2020, "lr": 2.5673902428233743e-05}
{"train_loss": 0.05901026725769043, "global_step": 179797, "epoch": 2020, "lr": 2.56733959323456e-05}
{"train_loss": 0.09082458168268204, "global_step": 179798, "epoch": 2020, "lr": 2.5672889439727843e-05}
{"train_loss": 0.06795382499694824, "global_step": 179799, "epoch": 2020, "lr": 2.5672382950380602e-05}
{"train_loss": 0.06419356912374496, "global_step": 179800, "epoch": 2020, "lr": 2.5671876464303883e-05}
{"train_loss": 0.05930342897772789, "global_step": 179801, "epoch": 2020, "lr": 2.5671369981497806e-05}
{"train_loss": 0.05282892659306526, "global_step": 179802, "epoch": 2020, "lr": 2.56708635019624e-05}
{"train_loss": 0.061403218656778336, "global_step": 179803, "epoch": 2020, "lr": 2.5670357025697766e-05}
{"train_loss": 0.06791500002145767, "global_step": 179804, "epoch": 2020, "lr": 2.566985055270394e-05}
{"train_loss": 0.07752519845962524, "global_step": 179805, "epoch": 2020, "lr": 2.5669344082981018e-05}
{"train_loss": 0.10606350749731064, "global_step": 179806, "epoch": 2020, "lr": 2.5668837616529046e-05}
{"train_loss": 0.10608163475990295, "global_step": 179807, "epoch": 2020, "lr": 2.5668331153348113e-05}
{"train_loss": 0.10546047240495682, "global_step": 179808, "epoch": 2020, "lr": 2.5667824693438265e-05}
{"train_loss": 0.08490313589572906, "global_step": 179809, "epoch": 2020, "lr": 2.5667318236799577e-05}
{"train_loss": 0.11991624534130096, "global_step": 179810, "epoch": 2020, "lr": 2.5666811783432143e-05}
{"train_loss": 0.07389819622039795, "global_step": 179811, "epoch": 2020, "lr": 2.5666305333335988e-05}
{"train_loss": 0.05998062714934349, "global_step": 179812, "epoch": 2020, "lr": 2.566579888651122e-05}
{"train_loss": 0.09575630724430084, "global_step": 179813, "epoch": 2020, "lr": 2.5665292442957877e-05}
{"train_loss": 0.08331045508384705, "global_step": 179814, "epoch": 2020, "lr": 2.566478600267605e-05}
{"train_loss": 0.05067899450659752, "global_step": 179815, "epoch": 2020, "lr": 2.5664279565665777e-05}
{"train_loss": 0.060441937297582626, "global_step": 179816, "epoch": 2020, "lr": 2.566377313192717e-05}
{"train_loss": 0.07438522577285767, "global_step": 179817, "epoch": 2020, "lr": 2.5663266701460248e-05}
{"train_loss": 0.04755185917019844, "global_step": 179818, "epoch": 2020, "lr": 2.5662760274265123e-05}
{"train_loss": 0.0652187168598175, "global_step": 179819, "epoch": 2020, "lr": 2.5662253850341822e-05}
{"train_loss": 0.06323204934597015, "global_step": 179820, "epoch": 2020, "lr": 2.5661747429690457e-05}
{"train_loss": 0.0701979398727417, "global_step": 179821, "epoch": 2020, "lr": 2.5661241012311054e-05}
{"train_loss": 0.15197211503982544, "global_step": 179822, "epoch": 2020, "lr": 2.5660734598203717e-05}
{"train_loss": 0.060323361307382584, "global_step": 179823, "epoch": 2020, "lr": 2.5660228187368484e-05}
{"train_loss": 0.057777151465415955, "global_step": 179824, "epoch": 2020, "lr": 2.5659721779805452e-05}
{"train_loss": 0.06397740542888641, "global_step": 179825, "epoch": 2020, "lr": 2.5659215375514657e-05}
{"train_loss": 0.08053679019212723, "global_step": 179826, "epoch": 2020, "lr": 2.5658708974496197e-05}
{"train_loss": 0.05928203463554382, "global_step": 179827, "epoch": 2020, "lr": 2.565820257675011e-05}
{"train_loss": 0.07974188029766083, "global_step": 179828, "epoch": 2020, "lr": 2.5657696182276502e-05}
{"train_loss": 0.1042250469326973, "global_step": 179829, "epoch": 2020, "lr": 2.5657189791075402e-05}
{"train_loss": 0.09948066622018814, "global_step": 179830, "epoch": 2020, "lr": 2.5656683403146918e-05}
{"train_loss": 0.08079684525728226, "global_step": 179831, "epoch": 2020, "lr": 2.565617701849109e-05}
{"train_loss": 0.06101442500948906, "global_step": 179832, "epoch": 2020, "lr": 2.5655670637107975e-05}
{"train_loss": 0.06978650391101837, "global_step": 179833, "epoch": 2020, "lr": 2.5655164258997677e-05}
{"train_loss": 0.02926764450967312, "global_step": 179834, "epoch": 2020, "lr": 2.5654657884160227e-05}
{"train_loss": 0.06355837732553482, "global_step": 179835, "epoch": 2020, "lr": 2.5654151512595732e-05}
{"train_loss": 0.08593406528234482, "global_step": 179836, "epoch": 2020, "lr": 2.5653645144304217e-05}
{"train_loss": 0.07228250056505203, "global_step": 179837, "epoch": 2020, "lr": 2.5653138779285796e-05}
{"train_loss": 0.10524556785821915, "global_step": 179838, "epoch": 2020, "lr": 2.565263241754049e-05}
{"train_loss": 0.0889996811747551, "global_step": 179839, "epoch": 2020, "lr": 2.5652126059068414e-05}
{"train_loss": 0.13720040023326874, "global_step": 179840, "epoch": 2020, "lr": 2.5651619703869596e-05}
{"train_loss": 0.09031623601913452, "global_step": 179841, "epoch": 2020, "lr": 2.5651113351944132e-05}
{"train_loss": 0.0989733412861824, "global_step": 179842, "epoch": 2020, "lr": 2.5650607003292072e-05}
{"train_loss": 0.10482703894376755, "global_step": 179843, "epoch": 2020, "lr": 2.56501006579135e-05}
{"train_loss": 0.03636596351861954, "global_step": 179844, "epoch": 2020, "lr": 2.5649594315808455e-05}
{"train_loss": 0.11006785929203033, "global_step": 179845, "epoch": 2020, "lr": 2.5649087976977026e-05}
{"train_loss": 0.05474495142698288, "global_step": 179846, "epoch": 2020, "lr": 2.5648581641419306e-05}
{"train_loss": 0.059012506157159805, "global_step": 179847, "epoch": 2020, "lr": 2.5648075309135317e-05}
{"train_loss": 0.05660892650485039, "global_step": 179848, "epoch": 2020, "lr": 2.5647568980125165e-05}
{"train_loss": 0.06818056851625443, "global_step": 179849, "epoch": 2020, "lr": 2.5647062654388876e-05}
{"train_loss": 0.05515357851982117, "global_step": 179850, "epoch": 2020, "lr": 2.564655633192657e-05}
{"train_loss": 0.07105458527803421, "global_step": 179851, "epoch": 2020, "lr": 2.5646050012738266e-05}
{"train_loss": 0.10804976522922516, "global_step": 179852, "epoch": 2020, "lr": 2.564554369682407e-05}
{"train_loss": 0.03941559046506882, "global_step": 179853, "epoch": 2020, "lr": 2.564503738418401e-05}
{"train_loss": 0.07059256732463837, "global_step": 179854, "epoch": 2020, "lr": 2.5644531074818213e-05}
{"train_loss": 0.13552211225032806, "global_step": 179855, "epoch": 2020, "lr": 2.5644024768726682e-05}
{"train_loss": 0.0871439054608345, "global_step": 179856, "epoch": 2020, "lr": 2.5643518465909533e-05}
{"train_loss": 0.06980334222316742, "global_step": 179857, "epoch": 2020, "lr": 2.56430121663668e-05}
{"train_loss": 0.07741570472717285, "global_step": 179858, "epoch": 2020, "lr": 2.5642505870098587e-05}
{"train_loss": 0.06931225955486298, "global_step": 179859, "epoch": 2020, "lr": 2.564199957710492e-05}
{"train_loss": 0.07329850643873215, "global_step": 179860, "epoch": 2020, "lr": 2.5641493287385908e-05}
{"train_loss": 0.06982588022947311, "global_step": 179861, "epoch": 2020, "lr": 2.5640987000941584e-05}
{"train_loss": 0.05839696153998375, "global_step": 179862, "epoch": 2020, "lr": 2.5640480717772052e-05}
{"train_loss": 0.03802717104554176, "global_step": 179863, "epoch": 2020, "lr": 2.563997443787734e-05}
{"train_loss": 0.039347611367702484, "global_step": 179864, "epoch": 2020, "lr": 2.5639468161257556e-05}
{"train_loss": 0.05768473818898201, "global_step": 179865, "epoch": 2020, "lr": 2.5638961887912725e-05}
{"train_loss": 0.0686473548412323, "global_step": 179866, "epoch": 2020, "lr": 2.563845561784297e-05}
{"train_loss": 0.10581446439027786, "global_step": 179867, "epoch": 2020, "lr": 2.5637949351048297e-05}
{"train_loss": 0.07643796493079555, "global_step": 179868, "epoch": 2020, "lr": 2.5637443087528828e-05, "val_loss": 7.331246376037598, "train_action_mse_error": 9.337900161743164}
{"train_loss": 0.09352434426546097, "global_step": 179869, "epoch": 2021, "lr": 2.5636936827284608e-05}
{"train_loss": 0.10979285836219788, "global_step": 179870, "epoch": 2021, "lr": 2.5636430570315685e-05}
{"train_loss": 0.07124459743499756, "global_step": 179871, "epoch": 2021, "lr": 2.5635924316622168e-05}
{"train_loss": 0.0653512105345726, "global_step": 179872, "epoch": 2021, "lr": 2.563541806620408e-05}
{"train_loss": 0.06199269741773605, "global_step": 179873, "epoch": 2021, "lr": 2.5634911819061535e-05}
{"train_loss": 0.09736388921737671, "global_step": 179874, "epoch": 2021, "lr": 2.5634405575194565e-05}
{"train_loss": 0.05376795679330826, "global_step": 179875, "epoch": 2021, "lr": 2.5633899334603262e-05}
{"train_loss": 0.0554702989757061, "global_step": 179876, "epoch": 2021, "lr": 2.5633393097287672e-05}
{"train_loss": 0.10561169683933258, "global_step": 179877, "epoch": 2021, "lr": 2.5632886863247897e-05}
{"train_loss": 0.04496288299560547, "global_step": 179878, "epoch": 2021, "lr": 2.5632380632483954e-05}
{"train_loss": 0.12918409705162048, "global_step": 179879, "epoch": 2021, "lr": 2.563187440499597e-05}
{"train_loss": 0.059522923082113266, "global_step": 179880, "epoch": 2021, "lr": 2.5631368180783955e-05}
{"train_loss": 0.09171666949987411, "global_step": 179881, "epoch": 2021, "lr": 2.5630861959848007e-05}
{"train_loss": 0.05494283139705658, "global_step": 179882, "epoch": 2021, "lr": 2.5630355742188217e-05}
{"train_loss": 0.0646032765507698, "global_step": 179883, "epoch": 2021, "lr": 2.5629849527804607e-05}
{"train_loss": 0.053857672959566116, "global_step": 179884, "epoch": 2021, "lr": 2.5629343316697284e-05}
{"train_loss": 0.09423334896564484, "global_step": 179885, "epoch": 2021, "lr": 2.5628837108866287e-05}
{"train_loss": 0.05807080864906311, "global_step": 179886, "epoch": 2021, "lr": 2.562833090431171e-05}
{"train_loss": 0.05022800341248512, "global_step": 179887, "epoch": 2021, "lr": 2.5627824703033587e-05}
{"train_loss": 0.07787066698074341, "global_step": 179888, "epoch": 2021, "lr": 2.5627318505032028e-05}
{"train_loss": 0.12205750495195389, "global_step": 179889, "epoch": 2021, "lr": 2.5626812310307054e-05}
{"train_loss": 0.08092100918292999, "global_step": 179890, "epoch": 2021, "lr": 2.562630611885878e-05}
{"train_loss": 0.07781172543764114, "global_step": 179891, "epoch": 2021, "lr": 2.5625799930687234e-05}
{"train_loss": 0.0894918367266655, "global_step": 179892, "epoch": 2021, "lr": 2.562529374579252e-05}
{"train_loss": 0.07839062064886093, "global_step": 179893, "epoch": 2021, "lr": 2.5624787564174667e-05}
{"train_loss": 0.053745657205581665, "global_step": 179894, "epoch": 2021, "lr": 2.5624281385833793e-05}
{"train_loss": 0.08489589393138885, "global_step": 179895, "epoch": 2021, "lr": 2.5623775210769907e-05}
{"train_loss": 0.07477137446403503, "global_step": 179896, "epoch": 2021, "lr": 2.5623269038983134e-05}
{"train_loss": 0.03922300040721893, "global_step": 179897, "epoch": 2021, "lr": 2.5622762870473487e-05}
{"train_loss": 0.05895944684743881, "global_step": 179898, "epoch": 2021, "lr": 2.562225670524109e-05}
{"train_loss": 0.0714530274271965, "global_step": 179899, "epoch": 2021, "lr": 2.562175054328596e-05}
{"train_loss": 0.05127355083823204, "global_step": 179900, "epoch": 2021, "lr": 2.562124438460819e-05}
{"train_loss": 0.06022518500685692, "global_step": 179901, "epoch": 2021, "lr": 2.5620738229207865e-05}
{"train_loss": 0.07176582515239716, "global_step": 179902, "epoch": 2021, "lr": 2.5620232077085036e-05}
{"train_loss": 0.11146940290927887, "global_step": 179903, "epoch": 2021, "lr": 2.561972592823975e-05}
{"train_loss": 0.02632187306880951, "global_step": 179904, "epoch": 2021, "lr": 2.5619219782672117e-05}
{"train_loss": 0.08768216520547867, "global_step": 179905, "epoch": 2021, "lr": 2.5618713640382158e-05}
{"train_loss": 0.050472091883420944, "global_step": 179906, "epoch": 2021, "lr": 2.561820750136999e-05}
{"train_loss": 0.09336334466934204, "global_step": 179907, "epoch": 2021, "lr": 2.5617701365635632e-05}
{"train_loss": 0.08949314802885056, "global_step": 179908, "epoch": 2021, "lr": 2.5617195233179204e-05}
{"train_loss": 0.056988585740327835, "global_step": 179909, "epoch": 2021, "lr": 2.5616689104000736e-05}
{"train_loss": 0.05252298712730408, "global_step": 179910, "epoch": 2021, "lr": 2.5616182978100296e-05}
{"train_loss": 0.06909888982772827, "global_step": 179911, "epoch": 2021, "lr": 2.5615676855477978e-05}
{"train_loss": 0.0671771839261055, "global_step": 179912, "epoch": 2021, "lr": 2.561517073613382e-05}
{"train_loss": 0.07886601984500885, "global_step": 179913, "epoch": 2021, "lr": 2.561466462006792e-05}
{"train_loss": 0.058147937059402466, "global_step": 179914, "epoch": 2021, "lr": 2.561415850728032e-05}
{"train_loss": 0.0710783302783966, "global_step": 179915, "epoch": 2021, "lr": 2.561365239777111e-05}
{"train_loss": 0.05687547102570534, "global_step": 179916, "epoch": 2021, "lr": 2.561314629154033e-05}
{"train_loss": 0.12921138107776642, "global_step": 179917, "epoch": 2021, "lr": 2.5612640188588066e-05}
{"train_loss": 0.057269759476184845, "global_step": 179918, "epoch": 2021, "lr": 2.56121340889144e-05}
{"train_loss": 0.06280988454818726, "global_step": 179919, "epoch": 2021, "lr": 2.561162799251937e-05}
{"train_loss": 0.06254150718450546, "global_step": 179920, "epoch": 2021, "lr": 2.561112189940308e-05}
{"train_loss": 0.07846888899803162, "global_step": 179921, "epoch": 2021, "lr": 2.561061580956556e-05}
{"train_loss": 0.09339655935764313, "global_step": 179922, "epoch": 2021, "lr": 2.5610109723006916e-05}
{"train_loss": 0.056098129600286484, "global_step": 179923, "epoch": 2021, "lr": 2.5609603639727164e-05}
{"train_loss": 0.07449192553758621, "global_step": 179924, "epoch": 2021, "lr": 2.5609097559726436e-05}
{"train_loss": 0.06521529704332352, "global_step": 179925, "epoch": 2021, "lr": 2.560859148300474e-05}
{"train_loss": 0.08940494805574417, "global_step": 179926, "epoch": 2021, "lr": 2.5608085409562194e-05}
{"train_loss": 0.06954102218151093, "global_step": 179927, "epoch": 2021, "lr": 2.5607579339398825e-05}
{"train_loss": 0.09437064081430435, "global_step": 179928, "epoch": 2021, "lr": 2.5607073272514738e-05}
{"train_loss": 0.12959709763526917, "global_step": 179929, "epoch": 2021, "lr": 2.560656720890996e-05}
{"train_loss": 0.04664010554552078, "global_step": 179930, "epoch": 2021, "lr": 2.5606061148584604e-05}
{"train_loss": 0.050504207611083984, "global_step": 179931, "epoch": 2021, "lr": 2.5605555091538703e-05}
{"train_loss": 0.07303120195865631, "global_step": 179932, "epoch": 2021, "lr": 2.560504903777235e-05}
{"train_loss": 0.05421581491827965, "global_step": 179933, "epoch": 2021, "lr": 2.5604542987285584e-05}
{"train_loss": 0.10370282083749771, "global_step": 179934, "epoch": 2021, "lr": 2.5604036940078503e-05}
{"train_loss": 0.074875108897686, "global_step": 179935, "epoch": 2021, "lr": 2.560353089615115e-05}
{"train_loss": 0.060532934963703156, "global_step": 179936, "epoch": 2021, "lr": 2.56030248555036e-05}
{"train_loss": 0.0580558143556118, "global_step": 179937, "epoch": 2021, "lr": 2.5602518818135944e-05}
{"train_loss": 0.03977128490805626, "global_step": 179938, "epoch": 2021, "lr": 2.5602012784048224e-05}
{"train_loss": 0.10196100920438766, "global_step": 179939, "epoch": 2021, "lr": 2.5601506753240523e-05}
{"train_loss": 0.06530727446079254, "global_step": 179940, "epoch": 2021, "lr": 2.5601000725712887e-05}
{"train_loss": 0.09647700935602188, "global_step": 179941, "epoch": 2021, "lr": 2.5600494701465416e-05}
{"train_loss": 0.10260903090238571, "global_step": 179942, "epoch": 2021, "lr": 2.5599988680498165e-05}
{"train_loss": 0.054640281945466995, "global_step": 179943, "epoch": 2021, "lr": 2.559948266281118e-05}
{"train_loss": 0.11771613359451294, "global_step": 179944, "epoch": 2021, "lr": 2.5598976648404566e-05}
{"train_loss": 0.06652176380157471, "global_step": 179945, "epoch": 2021, "lr": 2.559847063727835e-05}
{"train_loss": 0.1133575439453125, "global_step": 179946, "epoch": 2021, "lr": 2.5597964629432642e-05}
{"train_loss": 0.05330009385943413, "global_step": 179947, "epoch": 2021, "lr": 2.5597458624867486e-05}
{"train_loss": 0.06797988712787628, "global_step": 179948, "epoch": 2021, "lr": 2.559695262358295e-05}
{"train_loss": 0.08656694740056992, "global_step": 179949, "epoch": 2021, "lr": 2.5596446625579107e-05}
{"train_loss": 0.044173434376716614, "global_step": 179950, "epoch": 2021, "lr": 2.559594063085602e-05}
{"train_loss": 0.0536847859621048, "global_step": 179951, "epoch": 2021, "lr": 2.5595434639413773e-05}
{"train_loss": 0.10618475824594498, "global_step": 179952, "epoch": 2021, "lr": 2.5594928651252402e-05}
{"train_loss": 0.04951298609375954, "global_step": 179953, "epoch": 2021, "lr": 2.5594422666372014e-05}
{"train_loss": 0.07568473368883133, "global_step": 179954, "epoch": 2021, "lr": 2.5593916684772645e-05}
{"train_loss": 0.06704146414995193, "global_step": 179955, "epoch": 2021, "lr": 2.5593410706454367e-05}
{"train_loss": 0.14216437935829163, "global_step": 179956, "epoch": 2021, "lr": 2.5592904731417288e-05}
{"train_loss": 0.07484317813696487, "global_step": 179957, "epoch": 2021, "lr": 2.5592398759661418e-05, "val_loss": 7.6484150886535645}
{"train_loss": 0.10019729286432266, "global_step": 179958, "epoch": 2022, "lr": 2.559189279118688e-05}
{"train_loss": 0.08889822661876678, "global_step": 179959, "epoch": 2022, "lr": 2.5591386825993684e-05}
{"train_loss": 0.061150502413511276, "global_step": 179960, "epoch": 2022, "lr": 2.5590880864081956e-05}
{"train_loss": 0.08777043223381042, "global_step": 179961, "epoch": 2022, "lr": 2.559037490545172e-05}
{"train_loss": 0.07993350178003311, "global_step": 179962, "epoch": 2022, "lr": 2.558986895010307e-05}
{"train_loss": 0.07682617008686066, "global_step": 179963, "epoch": 2022, "lr": 2.5589362998036047e-05}
{"train_loss": 0.06511524319648743, "global_step": 179964, "epoch": 2022, "lr": 2.5588857049250765e-05}
{"train_loss": 0.08032882958650589, "global_step": 179965, "epoch": 2022, "lr": 2.5588351103747238e-05}
{"train_loss": 0.04634133726358414, "global_step": 179966, "epoch": 2022, "lr": 2.558784516152558e-05}
{"train_loss": 0.06270691752433777, "global_step": 179967, "epoch": 2022, "lr": 2.5587339222585816e-05}
{"train_loss": 0.08862589299678802, "global_step": 179968, "epoch": 2022, "lr": 2.5586833286928058e-05}
{"train_loss": 0.09307482093572617, "global_step": 179969, "epoch": 2022, "lr": 2.5586327354552335e-05}
{"train_loss": 0.07630046457052231, "global_step": 179970, "epoch": 2022, "lr": 2.5585821425458756e-05}
{"train_loss": 0.0777081772685051, "global_step": 179971, "epoch": 2022, "lr": 2.558531549964734e-05}
{"train_loss": 0.05730637535452843, "global_step": 179972, "epoch": 2022, "lr": 2.5584809577118185e-05}
{"train_loss": 0.0468476265668869, "global_step": 179973, "epoch": 2022, "lr": 2.5584303657871367e-05}
{"train_loss": 0.03856337070465088, "global_step": 179974, "epoch": 2022, "lr": 2.5583797741906928e-05}
{"train_loss": 0.06720060110092163, "global_step": 179975, "epoch": 2022, "lr": 2.5583291829224965e-05}
{"train_loss": 0.06871048361063004, "global_step": 179976, "epoch": 2022, "lr": 2.558278591982552e-05}
{"train_loss": 0.041943419724702835, "global_step": 179977, "epoch": 2022, "lr": 2.5582280013708682e-05}
{"train_loss": 0.08793967962265015, "global_step": 179978, "epoch": 2022, "lr": 2.5581774110874497e-05}
{"train_loss": 0.06886684149503708, "global_step": 179979, "epoch": 2022, "lr": 2.5581268211323063e-05}
{"train_loss": 0.03280727565288544, "global_step": 179980, "epoch": 2022, "lr": 2.5580762315054423e-05}
{"train_loss": 0.10682830214500427, "global_step": 179981, "epoch": 2022, "lr": 2.558025642206864e-05}
{"train_loss": 0.038050711154937744, "global_step": 179982, "epoch": 2022, "lr": 2.557975053236581e-05}
{"train_loss": 0.06940250098705292, "global_step": 179983, "epoch": 2022, "lr": 2.5579244645945965e-05}
{"train_loss": 0.05375349894165993, "global_step": 179984, "epoch": 2022, "lr": 2.5578738762809212e-05}
{"train_loss": 0.16688783466815948, "global_step": 179985, "epoch": 2022, "lr": 2.55782328829556e-05}
{"train_loss": 0.06748306006193161, "global_step": 179986, "epoch": 2022, "lr": 2.5577727006385176e-05}
{"train_loss": 0.08704406768083572, "global_step": 179987, "epoch": 2022, "lr": 2.557722113309805e-05}
{"train_loss": 0.08729570358991623, "global_step": 179988, "epoch": 2022, "lr": 2.5576715263094254e-05}
{"train_loss": 0.042382631450891495, "global_step": 179989, "epoch": 2022, "lr": 2.5576209396373884e-05}
{"train_loss": 0.07055536657571793, "global_step": 179990, "epoch": 2022, "lr": 2.5575703532936977e-05}
{"train_loss": 0.0973234623670578, "global_step": 179991, "epoch": 2022, "lr": 2.5575197672783624e-05}
{"train_loss": 0.07255807518959045, "global_step": 179992, "epoch": 2022, "lr": 2.5574691815913905e-05}
{"train_loss": 0.0998353660106659, "global_step": 179993, "epoch": 2022, "lr": 2.5574185962327852e-05}
{"train_loss": 0.09519684314727783, "global_step": 179994, "epoch": 2022, "lr": 2.5573680112025566e-05}
{"train_loss": 0.04019554331898689, "global_step": 179995, "epoch": 2022, "lr": 2.557317426500709e-05}
{"train_loss": 0.08805520087480545, "global_step": 179996, "epoch": 2022, "lr": 2.5572668421272516e-05}
{"train_loss": 0.08826740831136703, "global_step": 179997, "epoch": 2022, "lr": 2.5572162580821878e-05}
{"train_loss": 0.0481056310236454, "global_step": 179998, "epoch": 2022, "lr": 2.557165674365529e-05}
{"train_loss": 0.06721422076225281, "global_step": 179999, "epoch": 2022, "lr": 2.5571150909772772e-05}
{"train_loss": 0.12130571156740189, "global_step": 180000, "epoch": 2022, "lr": 2.5570645079174438e-05}
{"train_loss": 0.11401454359292984, "global_step": 180001, "epoch": 2022, "lr": 2.5570139251860308e-05}
{"train_loss": 0.0736059695482254, "global_step": 180002, "epoch": 2022, "lr": 2.5569633427830497e-05}
{"train_loss": 0.11992402374744415, "global_step": 180003, "epoch": 2022, "lr": 2.5569127607085037e-05}
{"train_loss": 0.05100838840007782, "global_step": 180004, "epoch": 2022, "lr": 2.5568621789624024e-05}
{"train_loss": 0.09420318901538849, "global_step": 180005, "epoch": 2022, "lr": 2.5568115975447494e-05}
{"train_loss": 0.1042560562491417, "global_step": 180006, "epoch": 2022, "lr": 2.556761016455555e-05}
{"train_loss": 0.1104203388094902, "global_step": 180007, "epoch": 2022, "lr": 2.5567104356948224e-05}
{"train_loss": 0.0434294119477272, "global_step": 180008, "epoch": 2022, "lr": 2.5566598552625607e-05}
{"train_loss": 0.1561710238456726, "global_step": 180009, "epoch": 2022, "lr": 2.5566092751587778e-05}
{"train_loss": 0.07025205343961716, "global_step": 180010, "epoch": 2022, "lr": 2.5565586953834776e-05}
{"train_loss": 0.06961251050233841, "global_step": 180011, "epoch": 2022, "lr": 2.55650811593667e-05}
{"train_loss": 0.09655662626028061, "global_step": 180012, "epoch": 2022, "lr": 2.5564575368183573e-05}
{"train_loss": 0.08721160888671875, "global_step": 180013, "epoch": 2022, "lr": 2.556406958028552e-05}
{"train_loss": 0.11870510876178741, "global_step": 180014, "epoch": 2022, "lr": 2.556356379567256e-05}
{"train_loss": 0.06645835191011429, "global_step": 180015, "epoch": 2022, "lr": 2.55630580143448e-05}
{"train_loss": 0.14157909154891968, "global_step": 180016, "epoch": 2022, "lr": 2.5562552236302267e-05}
{"train_loss": 0.15618373453617096, "global_step": 180017, "epoch": 2022, "lr": 2.5562046461545074e-05}
{"train_loss": 0.09612801671028137, "global_step": 180018, "epoch": 2022, "lr": 2.556154069007326e-05}
{"train_loss": 0.07084721326828003, "global_step": 180019, "epoch": 2022, "lr": 2.5561034921886883e-05}
{"train_loss": 0.06999529153108597, "global_step": 180020, "epoch": 2022, "lr": 2.556052915698605e-05}
{"train_loss": 0.08876360952854156, "global_step": 180021, "epoch": 2022, "lr": 2.556002339537078e-05}
{"train_loss": 0.04966384917497635, "global_step": 180022, "epoch": 2022, "lr": 2.5559517637041196e-05}
{"train_loss": 0.12234295904636383, "global_step": 180023, "epoch": 2022, "lr": 2.555901188199731e-05}
{"train_loss": 0.04725756496191025, "global_step": 180024, "epoch": 2022, "lr": 2.5558506130239236e-05}
{"train_loss": 0.16448834538459778, "global_step": 180025, "epoch": 2022, "lr": 2.555800038176702e-05}
{"train_loss": 0.07922423630952835, "global_step": 180026, "epoch": 2022, "lr": 2.5557494636580724e-05}
{"train_loss": 0.08046504110097885, "global_step": 180027, "epoch": 2022, "lr": 2.5556988894680424e-05}
{"train_loss": 0.09140247851610184, "global_step": 180028, "epoch": 2022, "lr": 2.5556483156066202e-05}
{"train_loss": 0.08922041952610016, "global_step": 180029, "epoch": 2022, "lr": 2.555597742073809e-05}
{"train_loss": 0.12150195986032486, "global_step": 180030, "epoch": 2022, "lr": 2.5555471688696208e-05}
{"train_loss": 0.0644158348441124, "global_step": 180031, "epoch": 2022, "lr": 2.5554965959940568e-05}
{"train_loss": 0.07895272970199585, "global_step": 180032, "epoch": 2022, "lr": 2.5554460234471288e-05}
{"train_loss": 0.034348923712968826, "global_step": 180033, "epoch": 2022, "lr": 2.5553954512288393e-05}
{"train_loss": 0.08092658966779709, "global_step": 180034, "epoch": 2022, "lr": 2.5553448793391987e-05}
{"train_loss": 0.05718711391091347, "global_step": 180035, "epoch": 2022, "lr": 2.555294307778211e-05}
{"train_loss": 0.0643601343035698, "global_step": 180036, "epoch": 2022, "lr": 2.5552437365458854e-05}
{"train_loss": 0.11843176931142807, "global_step": 180037, "epoch": 2022, "lr": 2.5551931656422257e-05}
{"train_loss": 0.09662581980228424, "global_step": 180038, "epoch": 2022, "lr": 2.5551425950672425e-05}
{"train_loss": 0.07912016659975052, "global_step": 180039, "epoch": 2022, "lr": 2.5550920248209386e-05}
{"train_loss": 0.0351775698363781, "global_step": 180040, "epoch": 2022, "lr": 2.555041454903325e-05}
{"train_loss": 0.06263408809900284, "global_step": 180041, "epoch": 2022, "lr": 2.5549908853144045e-05}
{"train_loss": 0.07976150512695312, "global_step": 180042, "epoch": 2022, "lr": 2.554940316054187e-05}
{"train_loss": 0.08150063455104828, "global_step": 180043, "epoch": 2022, "lr": 2.5548897471226764e-05}
{"train_loss": 0.07407399266958237, "global_step": 180044, "epoch": 2022, "lr": 2.5548391785198806e-05}
{"train_loss": 0.025203026831150055, "global_step": 180045, "epoch": 2022, "lr": 2.5547886102458096e-05}
{"train_loss": 0.08017910560697653, "global_step": 180046, "epoch": 2022, "lr": 2.5547380423004646e-05, "val_loss": 7.5413713455200195}
{"train_loss": 0.11172425001859665, "global_step": 180047, "epoch": 2023, "lr": 2.554687474683858e-05}
{"train_loss": 0.056918926537036896, "global_step": 180048, "epoch": 2023, "lr": 2.5546369073959913e-05}
{"train_loss": 0.0834958478808403, "global_step": 180049, "epoch": 2023, "lr": 2.5545863404368763e-05}
{"train_loss": 0.0656084343791008, "global_step": 180050, "epoch": 2023, "lr": 2.554535773806515e-05}
{"train_loss": 0.11824902147054672, "global_step": 180051, "epoch": 2023, "lr": 2.5544852075049185e-05}
{"train_loss": 0.11418894678354263, "global_step": 180052, "epoch": 2023, "lr": 2.5544346415320903e-05}
{"train_loss": 0.11470310389995575, "global_step": 180053, "epoch": 2023, "lr": 2.55438407588804e-05}
{"train_loss": 0.034082479774951935, "global_step": 180054, "epoch": 2023, "lr": 2.554333510572772e-05}
{"train_loss": 0.06926016509532928, "global_step": 180055, "epoch": 2023, "lr": 2.5542829455862942e-05}
{"train_loss": 0.06082240119576454, "global_step": 180056, "epoch": 2023, "lr": 2.5542323809286127e-05}
{"train_loss": 0.06014051288366318, "global_step": 180057, "epoch": 2023, "lr": 2.5541818165997362e-05}
{"train_loss": 0.07579758018255234, "global_step": 180058, "epoch": 2023, "lr": 2.55413125259967e-05}
{"train_loss": 0.0711013674736023, "global_step": 180059, "epoch": 2023, "lr": 2.55408068892842e-05}
{"train_loss": 0.06924096494913101, "global_step": 180060, "epoch": 2023, "lr": 2.554030125585995e-05}
{"train_loss": 0.1110239177942276, "global_step": 180061, "epoch": 2023, "lr": 2.5539795625723996e-05}
{"train_loss": 0.09214668720960617, "global_step": 180062, "epoch": 2023, "lr": 2.553928999887643e-05}
{"train_loss": 0.07128249853849411, "global_step": 180063, "epoch": 2023, "lr": 2.5538784375317293e-05}
{"train_loss": 0.08568644523620605, "global_step": 180064, "epoch": 2023, "lr": 2.553827875504669e-05}
{"train_loss": 0.05071817710995674, "global_step": 180065, "epoch": 2023, "lr": 2.5537773138064647e-05}
{"train_loss": 0.067160964012146, "global_step": 180066, "epoch": 2023, "lr": 2.5537267524371267e-05}
{"train_loss": 0.06132880970835686, "global_step": 180067, "epoch": 2023, "lr": 2.5536761913966588e-05}
{"train_loss": 0.10358452051877975, "global_step": 180068, "epoch": 2023, "lr": 2.5536256306850714e-05}
{"train_loss": 0.09596551954746246, "global_step": 180069, "epoch": 2023, "lr": 2.553575070302367e-05}
{"train_loss": 0.06499674916267395, "global_step": 180070, "epoch": 2023, "lr": 2.5535245102485568e-05}
{"train_loss": 0.13094115257263184, "global_step": 180071, "epoch": 2023, "lr": 2.553473950523643e-05}
{"train_loss": 0.06233099102973938, "global_step": 180072, "epoch": 2023, "lr": 2.5534233911276372e-05}
{"train_loss": 0.08765562623739243, "global_step": 180073, "epoch": 2023, "lr": 2.5533728320605415e-05}
{"train_loss": 0.10941340774297714, "global_step": 180074, "epoch": 2023, "lr": 2.553322273322367e-05}
{"train_loss": 0.05160773918032646, "global_step": 180075, "epoch": 2023, "lr": 2.553271714913117e-05}
{"train_loss": 0.05936579778790474, "global_step": 180076, "epoch": 2023, "lr": 2.5532211568328014e-05}
{"train_loss": 0.08684513717889786, "global_step": 180077, "epoch": 2023, "lr": 2.5531705990814237e-05}
{"train_loss": 0.05944501981139183, "global_step": 180078, "epoch": 2023, "lr": 2.553120041658994e-05}
{"train_loss": 0.0726558119058609, "global_step": 180079, "epoch": 2023, "lr": 2.5530694845655156e-05}
{"train_loss": 0.05676233768463135, "global_step": 180080, "epoch": 2023, "lr": 2.5530189278009998e-05}
{"train_loss": 0.07164216786623001, "global_step": 180081, "epoch": 2023, "lr": 2.5529683713654478e-05}
{"train_loss": 0.0977894589304924, "global_step": 180082, "epoch": 2023, "lr": 2.5529178152588695e-05}
{"train_loss": 0.10950729250907898, "global_step": 180083, "epoch": 2023, "lr": 2.552867259481274e-05}
{"train_loss": 0.0736001655459404, "global_step": 180084, "epoch": 2023, "lr": 2.5528167040326633e-05}
{"train_loss": 0.0865771546959877, "global_step": 180085, "epoch": 2023, "lr": 2.5527661489130495e-05}
{"train_loss": 0.05530661344528198, "global_step": 180086, "epoch": 2023, "lr": 2.5527155941224334e-05}
{"train_loss": 0.09388997405767441, "global_step": 180087, "epoch": 2023, "lr": 2.5526650396608277e-05}
{"train_loss": 0.03796469420194626, "global_step": 180088, "epoch": 2023, "lr": 2.5526144855282336e-05}
{"train_loss": 0.06490684300661087, "global_step": 180089, "epoch": 2023, "lr": 2.552563931724663e-05}
{"train_loss": 0.13570907711982727, "global_step": 180090, "epoch": 2023, "lr": 2.552513378250118e-05}
{"train_loss": 0.04105345159769058, "global_step": 180091, "epoch": 2023, "lr": 2.55246282510461e-05}
{"train_loss": 0.054679516702890396, "global_step": 180092, "epoch": 2023, "lr": 2.5524122722881416e-05}
{"train_loss": 0.05069279298186302, "global_step": 180093, "epoch": 2023, "lr": 2.5523617198007233e-05}
{"train_loss": 0.08132413029670715, "global_step": 180094, "epoch": 2023, "lr": 2.5523111676423584e-05}
{"train_loss": 0.04548332467675209, "global_step": 180095, "epoch": 2023, "lr": 2.552260615813058e-05}
{"train_loss": 0.08342020213603973, "global_step": 180096, "epoch": 2023, "lr": 2.5522100643128255e-05}
{"train_loss": 0.057443153113126755, "global_step": 180097, "epoch": 2023, "lr": 2.5521595131416663e-05}
{"train_loss": 0.0716099739074707, "global_step": 180098, "epoch": 2023, "lr": 2.552108962299592e-05}
{"train_loss": 0.06954833120107651, "global_step": 180099, "epoch": 2023, "lr": 2.5520584117866048e-05}
{"train_loss": 0.08558780699968338, "global_step": 180100, "epoch": 2023, "lr": 2.5520078616027154e-05}
{"train_loss": 0.09725886583328247, "global_step": 180101, "epoch": 2023, "lr": 2.5519573117479267e-05}
{"train_loss": 0.043872881680727005, "global_step": 180102, "epoch": 2023, "lr": 2.5519067622222493e-05}
{"train_loss": 0.05284794792532921, "global_step": 180103, "epoch": 2023, "lr": 2.551856213025686e-05}
{"train_loss": 0.04056559503078461, "global_step": 180104, "epoch": 2023, "lr": 2.551805664158249e-05}
{"train_loss": 0.08880340307950974, "global_step": 180105, "epoch": 2023, "lr": 2.5517551156199388e-05}
{"train_loss": 0.06136128678917885, "global_step": 180106, "epoch": 2023, "lr": 2.5517045674107674e-05}
{"train_loss": 0.054963141679763794, "global_step": 180107, "epoch": 2023, "lr": 2.551654019530738e-05}
{"train_loss": 0.07720428705215454, "global_step": 180108, "epoch": 2023, "lr": 2.5516034719798597e-05}
{"train_loss": 0.04195086658000946, "global_step": 180109, "epoch": 2023, "lr": 2.5515529247581384e-05}
{"train_loss": 0.13093650341033936, "global_step": 180110, "epoch": 2023, "lr": 2.551502377865581e-05}
{"train_loss": 0.07462861388921738, "global_step": 180111, "epoch": 2023, "lr": 2.5514518313021934e-05}
{"train_loss": 0.0447642058134079, "global_step": 180112, "epoch": 2023, "lr": 2.5514012850679847e-05}
{"train_loss": 0.10945910960435867, "global_step": 180113, "epoch": 2023, "lr": 2.5513507391629588e-05}
{"train_loss": 0.05290338397026062, "global_step": 180114, "epoch": 2023, "lr": 2.5513001935871254e-05}
{"train_loss": 0.07515048235654831, "global_step": 180115, "epoch": 2023, "lr": 2.551249648340488e-05}
{"train_loss": 0.07037219405174255, "global_step": 180116, "epoch": 2023, "lr": 2.5511991034230574e-05}
{"train_loss": 0.0576932318508625, "global_step": 180117, "epoch": 2023, "lr": 2.551148558834836e-05}
{"train_loss": 0.08012755215167999, "global_step": 180118, "epoch": 2023, "lr": 2.5510980145758333e-05}
{"train_loss": 0.0636998862028122, "global_step": 180119, "epoch": 2023, "lr": 2.5510474706460574e-05}
{"train_loss": 0.09312421828508377, "global_step": 180120, "epoch": 2023, "lr": 2.5509969270455113e-05}
{"train_loss": 0.10928889364004135, "global_step": 180121, "epoch": 2023, "lr": 2.5509463837742052e-05}
{"train_loss": 0.06895332783460617, "global_step": 180122, "epoch": 2023, "lr": 2.550895840832144e-05}
{"train_loss": 0.04995538294315338, "global_step": 180123, "epoch": 2023, "lr": 2.5508452982193354e-05}
{"train_loss": 0.1303313970565796, "global_step": 180124, "epoch": 2023, "lr": 2.5507947559357846e-05}
{"train_loss": 0.04238056764006615, "global_step": 180125, "epoch": 2023, "lr": 2.550744213981502e-05}
{"train_loss": 0.16642743349075317, "global_step": 180126, "epoch": 2023, "lr": 2.5506936723564888e-05}
{"train_loss": 0.07000653445720673, "global_step": 180127, "epoch": 2023, "lr": 2.550643131060758e-05}
{"train_loss": 0.09458012878894806, "global_step": 180128, "epoch": 2023, "lr": 2.5505925900943106e-05}
{"train_loss": 0.05755077302455902, "global_step": 180129, "epoch": 2023, "lr": 2.5505420494571587e-05}
{"train_loss": 0.11775106936693192, "global_step": 180130, "epoch": 2023, "lr": 2.5504915091493048e-05}
{"train_loss": 0.12202844768762589, "global_step": 180131, "epoch": 2023, "lr": 2.5504409691707587e-05}
{"train_loss": 0.1595349907875061, "global_step": 180132, "epoch": 2023, "lr": 2.5503904295215247e-05}
{"train_loss": 0.06346477568149567, "global_step": 180133, "epoch": 2023, "lr": 2.5503398902016128e-05}
{"train_loss": 0.08205603063106537, "global_step": 180134, "epoch": 2023, "lr": 2.5502893512110272e-05}
{"train_loss": 0.07816652777824509, "global_step": 180135, "epoch": 2023, "lr": 2.550238812549774e-05, "val_loss": 7.31679105758667}
{"train_loss": 0.08990536630153656, "global_step": 180136, "epoch": 2024, "lr": 2.5501882742178628e-05}
{"train_loss": 0.09520608186721802, "global_step": 180137, "epoch": 2024, "lr": 2.5501377362152972e-05}
{"train_loss": 0.07629343867301941, "global_step": 180138, "epoch": 2024, "lr": 2.5500871985420878e-05}
{"train_loss": 0.12273931503295898, "global_step": 180139, "epoch": 2024, "lr": 2.5500366611982373e-05}
{"train_loss": 0.09622068703174591, "global_step": 180140, "epoch": 2024, "lr": 2.5499861241837562e-05}
{"train_loss": 0.09727989137172699, "global_step": 180141, "epoch": 2024, "lr": 2.5499355874986473e-05}
{"train_loss": 0.020928654819726944, "global_step": 180142, "epoch": 2024, "lr": 2.5498850511429228e-05}
{"train_loss": 0.05775607377290726, "global_step": 180143, "epoch": 2024, "lr": 2.5498345151165836e-05}
{"train_loss": 0.06563836336135864, "global_step": 180144, "epoch": 2024, "lr": 2.5497839794196416e-05}
{"train_loss": 0.04406816512346268, "global_step": 180145, "epoch": 2024, "lr": 2.5497334440520992e-05}
{"train_loss": 0.06038530170917511, "global_step": 180146, "epoch": 2024, "lr": 2.5496829090139673e-05}
{"train_loss": 0.0864102691411972, "global_step": 180147, "epoch": 2024, "lr": 2.5496323743052485e-05}
{"train_loss": 0.04952947422862053, "global_step": 180148, "epoch": 2024, "lr": 2.549581839925954e-05}
{"train_loss": 0.07572557032108307, "global_step": 180149, "epoch": 2024, "lr": 2.5495313058760863e-05}
{"train_loss": 0.10595367103815079, "global_step": 180150, "epoch": 2024, "lr": 2.5494807721556557e-05}
{"train_loss": 0.05346459150314331, "global_step": 180151, "epoch": 2024, "lr": 2.549430238764666e-05}
{"train_loss": 0.10439428687095642, "global_step": 180152, "epoch": 2024, "lr": 2.5493797057031275e-05}
{"train_loss": 0.06753937155008316, "global_step": 180153, "epoch": 2024, "lr": 2.5493291729710435e-05}
{"train_loss": 0.07224304974079132, "global_step": 180154, "epoch": 2024, "lr": 2.549278640568422e-05}
{"train_loss": 0.07126697897911072, "global_step": 180155, "epoch": 2024, "lr": 2.5492281084952717e-05}
{"train_loss": 0.05098497495055199, "global_step": 180156, "epoch": 2024, "lr": 2.5491775767515968e-05}
{"train_loss": 0.09974416345357895, "global_step": 180157, "epoch": 2024, "lr": 2.5491270453374062e-05}
{"train_loss": 0.12073914706707001, "global_step": 180158, "epoch": 2024, "lr": 2.549076514252704e-05}
{"train_loss": 0.0387972928583622, "global_step": 180159, "epoch": 2024, "lr": 2.5490259834975006e-05}
{"train_loss": 0.03198958560824394, "global_step": 180160, "epoch": 2024, "lr": 2.548975453071798e-05}
{"train_loss": 0.10496759414672852, "global_step": 180161, "epoch": 2024, "lr": 2.5489249229756085e-05}
{"train_loss": 0.11402174830436707, "global_step": 180162, "epoch": 2024, "lr": 2.5488743932089343e-05}
{"train_loss": 0.07284949719905853, "global_step": 180163, "epoch": 2024, "lr": 2.5488238637717853e-05}
{"train_loss": 0.06049942597746849, "global_step": 180164, "epoch": 2024, "lr": 2.548773334664166e-05}
{"train_loss": 0.1212611049413681, "global_step": 180165, "epoch": 2024, "lr": 2.5487228058860857e-05}
{"train_loss": 0.10122658312320709, "global_step": 180166, "epoch": 2024, "lr": 2.5486722774375482e-05}
{"train_loss": 0.06635016947984695, "global_step": 180167, "epoch": 2024, "lr": 2.548621749318563e-05}
{"train_loss": 0.08807273954153061, "global_step": 180168, "epoch": 2024, "lr": 2.5485712215291345e-05}
{"train_loss": 0.14840567111968994, "global_step": 180169, "epoch": 2024, "lr": 2.5485206940692725e-05}
{"train_loss": 0.06799983233213425, "global_step": 180170, "epoch": 2024, "lr": 2.54847016693898e-05}
{"train_loss": 0.1670861393213272, "global_step": 180171, "epoch": 2024, "lr": 2.5484196401382677e-05}
{"train_loss": 0.057924967259168625, "global_step": 180172, "epoch": 2024, "lr": 2.5483691136671385e-05}
{"train_loss": 0.06039426848292351, "global_step": 180173, "epoch": 2024, "lr": 2.5483185875256033e-05}
{"train_loss": 0.10728956758975983, "global_step": 180174, "epoch": 2024, "lr": 2.5482680617136667e-05}
{"train_loss": 0.09585527330636978, "global_step": 180175, "epoch": 2024, "lr": 2.548217536231334e-05}
{"train_loss": 0.08505918830633163, "global_step": 180176, "epoch": 2024, "lr": 2.5481670110786148e-05}
{"train_loss": 0.13800375163555145, "global_step": 180177, "epoch": 2024, "lr": 2.548116486255513e-05}
{"train_loss": 0.12267496436834335, "global_step": 180178, "epoch": 2024, "lr": 2.5480659617620395e-05}
{"train_loss": 0.11173181980848312, "global_step": 180179, "epoch": 2024, "lr": 2.5480154375981958e-05}
{"train_loss": 0.11136344820261002, "global_step": 180180, "epoch": 2024, "lr": 2.5479649137639937e-05}
{"train_loss": 0.07175960391759872, "global_step": 180181, "epoch": 2024, "lr": 2.5479143902594367e-05}
{"train_loss": 0.07124491780996323, "global_step": 180182, "epoch": 2024, "lr": 2.5478638670845338e-05}
{"train_loss": 0.1017642617225647, "global_step": 180183, "epoch": 2024, "lr": 2.5478133442392894e-05}
{"train_loss": 0.03368163853883743, "global_step": 180184, "epoch": 2024, "lr": 2.5477628217237132e-05}
{"train_loss": 0.10083747655153275, "global_step": 180185, "epoch": 2024, "lr": 2.5477122995378083e-05}
{"train_loss": 0.060349829494953156, "global_step": 180186, "epoch": 2024, "lr": 2.547661777681586e-05}
{"train_loss": 0.08550065755844116, "global_step": 180187, "epoch": 2024, "lr": 2.547611256155048e-05}
{"train_loss": 0.06975312530994415, "global_step": 180188, "epoch": 2024, "lr": 2.5475607349582064e-05}
{"train_loss": 0.1300082504749298, "global_step": 180189, "epoch": 2024, "lr": 2.5475102140910634e-05}
{"train_loss": 0.09657507389783859, "global_step": 180190, "epoch": 2024, "lr": 2.5474596935536276e-05}
{"train_loss": 0.08517593890428543, "global_step": 180191, "epoch": 2024, "lr": 2.5474091733459083e-05}
{"train_loss": 0.025441277772188187, "global_step": 180192, "epoch": 2024, "lr": 2.5473586534679073e-05}
{"train_loss": 0.1461368203163147, "global_step": 180193, "epoch": 2024, "lr": 2.547308133919636e-05}
{"train_loss": 0.035601917654275894, "global_step": 180194, "epoch": 2024, "lr": 2.5472576147010972e-05}
{"train_loss": 0.10206632316112518, "global_step": 180195, "epoch": 2024, "lr": 2.5472070958123023e-05}
{"train_loss": 0.08954600244760513, "global_step": 180196, "epoch": 2024, "lr": 2.5471565772532537e-05}
{"train_loss": 0.04884365573525429, "global_step": 180197, "epoch": 2024, "lr": 2.5471060590239613e-05}
{"train_loss": 0.046992767602205276, "global_step": 180198, "epoch": 2024, "lr": 2.547055541124429e-05}
{"train_loss": 0.0897691547870636, "global_step": 180199, "epoch": 2024, "lr": 2.5470050235546668e-05}
{"train_loss": 0.11379878222942352, "global_step": 180200, "epoch": 2024, "lr": 2.546954506314678e-05}
{"train_loss": 0.047572456300258636, "global_step": 180201, "epoch": 2024, "lr": 2.5469039894044732e-05}
{"train_loss": 0.0804285928606987, "global_step": 180202, "epoch": 2024, "lr": 2.5468534728240556e-05}
{"train_loss": 0.05938557907938957, "global_step": 180203, "epoch": 2024, "lr": 2.5468029565734353e-05}
{"train_loss": 0.08274997025728226, "global_step": 180204, "epoch": 2024, "lr": 2.5467524406526157e-05}
{"train_loss": 0.059068724513053894, "global_step": 180205, "epoch": 2024, "lr": 2.5467019250616074e-05}
{"train_loss": 0.09545686841011047, "global_step": 180206, "epoch": 2024, "lr": 2.546651409800413e-05}
{"train_loss": 0.04958602041006088, "global_step": 180207, "epoch": 2024, "lr": 2.5466008948690435e-05}
{"train_loss": 0.07854274660348892, "global_step": 180208, "epoch": 2024, "lr": 2.5465503802675018e-05}
{"train_loss": 0.08188936114311218, "global_step": 180209, "epoch": 2024, "lr": 2.5464998659957978e-05}
{"train_loss": 0.08848876506090164, "global_step": 180210, "epoch": 2024, "lr": 2.5464493520539355e-05}
{"train_loss": 0.07697588205337524, "global_step": 180211, "epoch": 2024, "lr": 2.5463988384419245e-05}
{"train_loss": 0.0631045326590538, "global_step": 180212, "epoch": 2024, "lr": 2.5463483251597708e-05}
{"train_loss": 0.11358901858329773, "global_step": 180213, "epoch": 2024, "lr": 2.5462978122074788e-05}
{"train_loss": 0.12837517261505127, "global_step": 180214, "epoch": 2024, "lr": 2.5462472995850594e-05}
{"train_loss": 0.03329455479979515, "global_step": 180215, "epoch": 2024, "lr": 2.5461967872925153e-05}
{"train_loss": 0.05832034349441528, "global_step": 180216, "epoch": 2024, "lr": 2.546146275329856e-05}
{"train_loss": 0.16330203413963318, "global_step": 180217, "epoch": 2024, "lr": 2.546095763697086e-05}
{"train_loss": 0.0551876574754715, "global_step": 180218, "epoch": 2024, "lr": 2.5460452523942157e-05}
{"train_loss": 0.10539371520280838, "global_step": 180219, "epoch": 2024, "lr": 2.5459947414212476e-05}
{"train_loss": 0.06975677609443665, "global_step": 180220, "epoch": 2024, "lr": 2.545944230778192e-05}
{"train_loss": 0.0672672763466835, "global_step": 180221, "epoch": 2024, "lr": 2.5458937204650535e-05}
{"train_loss": 0.18471169471740723, "global_step": 180222, "epoch": 2024, "lr": 2.5458432104818407e-05}
{"train_loss": 0.05123927816748619, "global_step": 180223, "epoch": 2024, "lr": 2.545792700828558e-05}
{"train_loss": 0.08391666441653552, "global_step": 180224, "epoch": 2024, "lr": 2.5457421915052155e-05, "val_loss": 7.421138763427734}
{"train_loss": 0.1272149384021759, "global_step": 180225, "epoch": 2025, "lr": 2.545691682511816e-05}
{"train_loss": 0.10939644277095795, "global_step": 180226, "epoch": 2025, "lr": 2.5456411738483675e-05}
{"train_loss": 0.08559796959161758, "global_step": 180227, "epoch": 2025, "lr": 2.5455906655148808e-05}
{"train_loss": 0.03561598062515259, "global_step": 180228, "epoch": 2025, "lr": 2.5455401575113568e-05}
{"train_loss": 0.15130627155303955, "global_step": 180229, "epoch": 2025, "lr": 2.5454896498378074e-05}
{"train_loss": 0.08258960396051407, "global_step": 180230, "epoch": 2025, "lr": 2.545439142494235e-05}
{"train_loss": 0.06841254234313965, "global_step": 180231, "epoch": 2025, "lr": 2.54538863548065e-05}
{"train_loss": 0.05501091852784157, "global_step": 180232, "epoch": 2025, "lr": 2.5453381287970555e-05}
{"train_loss": 0.07095925509929657, "global_step": 180233, "epoch": 2025, "lr": 2.545287622443463e-05}
{"train_loss": 0.08248867839574814, "global_step": 180234, "epoch": 2025, "lr": 2.545237116419874e-05}
{"train_loss": 0.056269336491823196, "global_step": 180235, "epoch": 2025, "lr": 2.5451866107263e-05}
{"train_loss": 0.05972316861152649, "global_step": 180236, "epoch": 2025, "lr": 2.5451361053627443e-05}
{"train_loss": 0.09737128764390945, "global_step": 180237, "epoch": 2025, "lr": 2.5450856003292167e-05}
{"train_loss": 0.07993916422128677, "global_step": 180238, "epoch": 2025, "lr": 2.5450350956257208e-05}
{"train_loss": 0.08724336326122284, "global_step": 180239, "epoch": 2025, "lr": 2.544984591252267e-05}
{"train_loss": 0.09028779715299606, "global_step": 180240, "epoch": 2025, "lr": 2.5449340872088578e-05}
{"train_loss": 0.08626759052276611, "global_step": 180241, "epoch": 2025, "lr": 2.5448835834955042e-05}
{"train_loss": 0.04919911548495293, "global_step": 180242, "epoch": 2025, "lr": 2.5448330801122088e-05}
{"train_loss": 0.052971065044403076, "global_step": 180243, "epoch": 2025, "lr": 2.5447825770589835e-05}
{"train_loss": 0.04620738327503204, "global_step": 180244, "epoch": 2025, "lr": 2.54473207433583e-05}
{"train_loss": 0.0738748088479042, "global_step": 180245, "epoch": 2025, "lr": 2.5446815719427592e-05}
{"train_loss": 0.10364628583192825, "global_step": 180246, "epoch": 2025, "lr": 2.5446310698797738e-05}
{"train_loss": 0.07437047362327576, "global_step": 180247, "epoch": 2025, "lr": 2.5445805681468848e-05}
{"train_loss": 0.09612444043159485, "global_step": 180248, "epoch": 2025, "lr": 2.5445300667440958e-05}
{"train_loss": 0.08467449247837067, "global_step": 180249, "epoch": 2025, "lr": 2.5444795656714158e-05}
{"train_loss": 0.08526348322629929, "global_step": 180250, "epoch": 2025, "lr": 2.5444290649288493e-05}
{"train_loss": 0.11519525945186615, "global_step": 180251, "epoch": 2025, "lr": 2.5443785645164065e-05}
{"train_loss": 0.09952595084905624, "global_step": 180252, "epoch": 2025, "lr": 2.5443280644340917e-05}
{"train_loss": 0.06862448900938034, "global_step": 180253, "epoch": 2025, "lr": 2.54427756468191e-05}
{"train_loss": 0.07637254148721695, "global_step": 180254, "epoch": 2025, "lr": 2.544227065259872e-05}
{"train_loss": 0.06862626224756241, "global_step": 180255, "epoch": 2025, "lr": 2.5441765661679816e-05}
{"train_loss": 0.10707993060350418, "global_step": 180256, "epoch": 2025, "lr": 2.5441260674062484e-05}
{"train_loss": 0.12048490345478058, "global_step": 180257, "epoch": 2025, "lr": 2.5440755689746748e-05}
{"train_loss": 0.1650158315896988, "global_step": 180258, "epoch": 2025, "lr": 2.5440250708732726e-05}
{"train_loss": 0.06777143478393555, "global_step": 180259, "epoch": 2025, "lr": 2.5439745731020448e-05}
{"train_loss": 0.09735812991857529, "global_step": 180260, "epoch": 2025, "lr": 2.5439240756610017e-05}
{"train_loss": 0.0559050627052784, "global_step": 180261, "epoch": 2025, "lr": 2.5438735785501455e-05}
{"train_loss": 0.14651373028755188, "global_step": 180262, "epoch": 2025, "lr": 2.5438230817694865e-05}
{"train_loss": 0.09856105595827103, "global_step": 180263, "epoch": 2025, "lr": 2.5437725853190315e-05}
{"train_loss": 0.06800905615091324, "global_step": 180264, "epoch": 2025, "lr": 2.5437220891987844e-05}
{"train_loss": 0.06799214333295822, "global_step": 180265, "epoch": 2025, "lr": 2.5436715934087562e-05}
{"train_loss": 0.027165288105607033, "global_step": 180266, "epoch": 2025, "lr": 2.5436210979489496e-05}
{"train_loss": 0.09993726015090942, "global_step": 180267, "epoch": 2025, "lr": 2.543570602819375e-05}
{"train_loss": 0.07904677093029022, "global_step": 180268, "epoch": 2025, "lr": 2.5435201080200355e-05}
{"train_loss": 0.0347621887922287, "global_step": 180269, "epoch": 2025, "lr": 2.5434696135509416e-05}
{"train_loss": 0.061042528599500656, "global_step": 180270, "epoch": 2025, "lr": 2.543419119412097e-05}
{"train_loss": 0.033421989530324936, "global_step": 180271, "epoch": 2025, "lr": 2.543368625603511e-05}
{"train_loss": 0.09747635573148727, "global_step": 180272, "epoch": 2025, "lr": 2.5433181321251876e-05}
{"train_loss": 0.08362548798322678, "global_step": 180273, "epoch": 2025, "lr": 2.543267638977137e-05}
{"train_loss": 0.04902570694684982, "global_step": 180274, "epoch": 2025, "lr": 2.5432171461593618e-05}
{"train_loss": 0.08159782737493515, "global_step": 180275, "epoch": 2025, "lr": 2.5431666536718733e-05}
{"train_loss": 0.026119176298379898, "global_step": 180276, "epoch": 2025, "lr": 2.5431161615146736e-05}
{"train_loss": 0.044558774679899216, "global_step": 180277, "epoch": 2025, "lr": 2.5430656696877746e-05}
{"train_loss": 0.07088389992713928, "global_step": 180278, "epoch": 2025, "lr": 2.543015178191179e-05}
{"train_loss": 0.0936707854270935, "global_step": 180279, "epoch": 2025, "lr": 2.542964687024896e-05}
{"train_loss": 0.12711700797080994, "global_step": 180280, "epoch": 2025, "lr": 2.5429141961889303e-05}
{"train_loss": 0.051551900804042816, "global_step": 180281, "epoch": 2025, "lr": 2.5428637056832892e-05}
{"train_loss": 0.10628864914178848, "global_step": 180282, "epoch": 2025, "lr": 2.542813215507982e-05}
{"train_loss": 0.0937713161110878, "global_step": 180283, "epoch": 2025, "lr": 2.542762725663012e-05}
{"train_loss": 0.09063954651355743, "global_step": 180284, "epoch": 2025, "lr": 2.5427122361483897e-05}
{"train_loss": 0.06268440932035446, "global_step": 180285, "epoch": 2025, "lr": 2.54266174696412e-05}
{"train_loss": 0.08669064193964005, "global_step": 180286, "epoch": 2025, "lr": 2.5426112581102068e-05}
{"train_loss": 0.0928194671869278, "global_step": 180287, "epoch": 2025, "lr": 2.5425607695866614e-05}
{"train_loss": 0.11060991138219833, "global_step": 180288, "epoch": 2025, "lr": 2.5425102813934874e-05}
{"train_loss": 0.08067641407251358, "global_step": 180289, "epoch": 2025, "lr": 2.542459793530695e-05}
{"train_loss": 0.08693830668926239, "global_step": 180290, "epoch": 2025, "lr": 2.542409305998289e-05}
{"train_loss": 0.09616696089506149, "global_step": 180291, "epoch": 2025, "lr": 2.5423588187962732e-05}
{"train_loss": 0.08773098140954971, "global_step": 180292, "epoch": 2025, "lr": 2.5423083319246603e-05}
{"train_loss": 0.12604795396327972, "global_step": 180293, "epoch": 2025, "lr": 2.5422578453834516e-05}
{"train_loss": 0.05177801102399826, "global_step": 180294, "epoch": 2025, "lr": 2.5422073591726586e-05}
{"train_loss": 0.06738722324371338, "global_step": 180295, "epoch": 2025, "lr": 2.542156873292284e-05}
{"train_loss": 0.08624744415283203, "global_step": 180296, "epoch": 2025, "lr": 2.542106387742338e-05}
{"train_loss": 0.06313294172286987, "global_step": 180297, "epoch": 2025, "lr": 2.5420559025228242e-05}
{"train_loss": 0.03364305943250656, "global_step": 180298, "epoch": 2025, "lr": 2.5420054176337506e-05}
{"train_loss": 0.06473629176616669, "global_step": 180299, "epoch": 2025, "lr": 2.541954933075127e-05}
{"train_loss": 0.051113445311784744, "global_step": 180300, "epoch": 2025, "lr": 2.541904448846955e-05}
{"train_loss": 0.11320513486862183, "global_step": 180301, "epoch": 2025, "lr": 2.5418539649492467e-05}
{"train_loss": 0.10238415747880936, "global_step": 180302, "epoch": 2025, "lr": 2.5418034813820036e-05}
{"train_loss": 0.0921628400683403, "global_step": 180303, "epoch": 2025, "lr": 2.541752998145237e-05}
{"train_loss": 0.07313471287488937, "global_step": 180304, "epoch": 2025, "lr": 2.54170251523895e-05}
{"train_loss": 0.106197290122509, "global_step": 180305, "epoch": 2025, "lr": 2.541652032663153e-05}
{"train_loss": 0.04438432306051254, "global_step": 180306, "epoch": 2025, "lr": 2.5416015504178492e-05}
{"train_loss": 0.043977539986371994, "global_step": 180307, "epoch": 2025, "lr": 2.5415510685030484e-05}
{"train_loss": 0.11955719441175461, "global_step": 180308, "epoch": 2025, "lr": 2.541500586918755e-05}
{"train_loss": 0.11541002243757248, "global_step": 180309, "epoch": 2025, "lr": 2.5414501056649785e-05}
{"train_loss": 0.06313742697238922, "global_step": 180310, "epoch": 2025, "lr": 2.5413996247417223e-05}
{"train_loss": 0.06299727410078049, "global_step": 180311, "epoch": 2025, "lr": 2.5413491441489966e-05}
{"train_loss": 0.14334650337696075, "global_step": 180312, "epoch": 2025, "lr": 2.541298663886805e-05}
{"train_loss": 0.0816695573164171, "global_step": 180313, "epoch": 2025, "lr": 2.5412481839551572e-05, "val_loss": 7.34029483795166, "train_action_mse_error": 19.793176651000977}
{"train_loss": 0.0574997141957283, "global_step": 180314, "epoch": 2026, "lr": 2.541197704354057e-05}
{"train_loss": 0.1099725067615509, "global_step": 180315, "epoch": 2026, "lr": 2.5411472250835145e-05}
{"train_loss": 0.041558887809515, "global_step": 180316, "epoch": 2026, "lr": 2.541096746143533e-05}
{"train_loss": 0.05357549712061882, "global_step": 180317, "epoch": 2026, "lr": 2.541046267534121e-05}
{"train_loss": 0.08639667183160782, "global_step": 180318, "epoch": 2026, "lr": 2.5409957892552875e-05}
{"train_loss": 0.08397051692008972, "global_step": 180319, "epoch": 2026, "lr": 2.5409453113070353e-05}
{"train_loss": 0.09687551856040955, "global_step": 180320, "epoch": 2026, "lr": 2.5408948336893746e-05}
{"train_loss": 0.12130006402730942, "global_step": 180321, "epoch": 2026, "lr": 2.5408443564023088e-05}
{"train_loss": 0.10702041536569595, "global_step": 180322, "epoch": 2026, "lr": 2.5407938794458485e-05}
{"train_loss": 0.06084245815873146, "global_step": 180323, "epoch": 2026, "lr": 2.5407434028199983e-05}
{"train_loss": 0.051125869154930115, "global_step": 180324, "epoch": 2026, "lr": 2.5406929265247637e-05}
{"train_loss": 0.06889335811138153, "global_step": 180325, "epoch": 2026, "lr": 2.5406424505601546e-05}
{"train_loss": 0.11227671802043915, "global_step": 180326, "epoch": 2026, "lr": 2.5405919749261742e-05}
{"train_loss": 0.0684344619512558, "global_step": 180327, "epoch": 2026, "lr": 2.5405414996228332e-05}
{"train_loss": 0.06441755592823029, "global_step": 180328, "epoch": 2026, "lr": 2.540491024650137e-05}
{"train_loss": 0.09318199753761292, "global_step": 180329, "epoch": 2026, "lr": 2.5404405500080887e-05}
{"train_loss": 0.07990092039108276, "global_step": 180330, "epoch": 2026, "lr": 2.5403900756967e-05}
{"train_loss": 0.1405528336763382, "global_step": 180331, "epoch": 2026, "lr": 2.5403396017159753e-05}
{"train_loss": 0.05726365000009537, "global_step": 180332, "epoch": 2026, "lr": 2.5402891280659236e-05}
{"train_loss": 0.1023961752653122, "global_step": 180333, "epoch": 2026, "lr": 2.5402386547465472e-05}
{"train_loss": 0.05696086212992668, "global_step": 180334, "epoch": 2026, "lr": 2.540188181757859e-05}
{"train_loss": 0.08730518072843552, "global_step": 180335, "epoch": 2026, "lr": 2.540137709099859e-05}
{"train_loss": 0.07648223638534546, "global_step": 180336, "epoch": 2026, "lr": 2.5400872367725587e-05}
{"train_loss": 0.04949298873543739, "global_step": 180337, "epoch": 2026, "lr": 2.540036764775965e-05}
{"train_loss": 0.05878603458404541, "global_step": 180338, "epoch": 2026, "lr": 2.5399862931100815e-05}
{"train_loss": 0.06981471925973892, "global_step": 180339, "epoch": 2026, "lr": 2.539935821774919e-05}
{"train_loss": 0.05729624256491661, "global_step": 180340, "epoch": 2026, "lr": 2.53988535077048e-05}
{"train_loss": 0.05099461227655411, "global_step": 180341, "epoch": 2026, "lr": 2.5398348800967757e-05}
{"train_loss": 0.07221106439828873, "global_step": 180342, "epoch": 2026, "lr": 2.5397844097538087e-05}
{"train_loss": 0.07173887640237808, "global_step": 180343, "epoch": 2026, "lr": 2.5397339397415897e-05}
{"train_loss": 0.07050260156393051, "global_step": 180344, "epoch": 2026, "lr": 2.5396834700601212e-05}
{"train_loss": 0.09972739219665527, "global_step": 180345, "epoch": 2026, "lr": 2.5396330007094143e-05}
{"train_loss": 0.06145337224006653, "global_step": 180346, "epoch": 2026, "lr": 2.5395825316894717e-05}
{"train_loss": 0.08783761411905289, "global_step": 180347, "epoch": 2026, "lr": 2.5395320630003044e-05}
{"train_loss": 0.06224006041884422, "global_step": 180348, "epoch": 2026, "lr": 2.539481594641915e-05}
{"train_loss": 0.07535713165998459, "global_step": 180349, "epoch": 2026, "lr": 2.5394311266143144e-05}
{"train_loss": 0.06551116704940796, "global_step": 180350, "epoch": 2026, "lr": 2.539380658917505e-05}
{"train_loss": 0.051920875906944275, "global_step": 180351, "epoch": 2026, "lr": 2.5393301915514977e-05}
{"train_loss": 0.04072709381580353, "global_step": 180352, "epoch": 2026, "lr": 2.5392797245162957e-05}
{"train_loss": 0.06652408093214035, "global_step": 180353, "epoch": 2026, "lr": 2.539229257811908e-05}
{"train_loss": 0.07661062479019165, "global_step": 180354, "epoch": 2026, "lr": 2.5391787914383424e-05}
{"train_loss": 0.02313876897096634, "global_step": 180355, "epoch": 2026, "lr": 2.5391283253956028e-05}
{"train_loss": 0.02156016044318676, "global_step": 180356, "epoch": 2026, "lr": 2.539077859683699e-05}
{"train_loss": 0.0630737915635109, "global_step": 180357, "epoch": 2026, "lr": 2.5390273943026343e-05}
{"train_loss": 0.06865642964839935, "global_step": 180358, "epoch": 2026, "lr": 2.538976929252419e-05}
{"train_loss": 0.08867292106151581, "global_step": 180359, "epoch": 2026, "lr": 2.538926464533057e-05}
{"train_loss": 0.12035859376192093, "global_step": 180360, "epoch": 2026, "lr": 2.5388760001445577e-05}
{"train_loss": 0.059295859187841415, "global_step": 180361, "epoch": 2026, "lr": 2.5388255360869262e-05}
{"train_loss": 0.08522874116897583, "global_step": 180362, "epoch": 2026, "lr": 2.5387750723601688e-05}
{"train_loss": 0.10048805177211761, "global_step": 180363, "epoch": 2026, "lr": 2.5387246089642942e-05}
{"train_loss": 0.14914453029632568, "global_step": 180364, "epoch": 2026, "lr": 2.5386741458993057e-05}
{"train_loss": 0.05005640909075737, "global_step": 180365, "epoch": 2026, "lr": 2.5386236831652153e-05}
{"train_loss": 0.0965019166469574, "global_step": 180366, "epoch": 2026, "lr": 2.538573220762025e-05}
{"train_loss": 0.07034409791231155, "global_step": 180367, "epoch": 2026, "lr": 2.5385227586897453e-05}
{"train_loss": 0.08766675740480423, "global_step": 180368, "epoch": 2026, "lr": 2.538472296948381e-05}
{"train_loss": 0.07558634132146835, "global_step": 180369, "epoch": 2026, "lr": 2.5384218355379374e-05}
{"train_loss": 0.03917858004570007, "global_step": 180370, "epoch": 2026, "lr": 2.538371374458425e-05}
{"train_loss": 0.024978425353765488, "global_step": 180371, "epoch": 2026, "lr": 2.538320913709846e-05}
{"train_loss": 0.06428961455821991, "global_step": 180372, "epoch": 2026, "lr": 2.5382704532922096e-05}
{"train_loss": 0.0562763512134552, "global_step": 180373, "epoch": 2026, "lr": 2.5382199932055255e-05}
{"train_loss": 0.08484482765197754, "global_step": 180374, "epoch": 2026, "lr": 2.5381695334497957e-05}
{"train_loss": 0.06052049621939659, "global_step": 180375, "epoch": 2026, "lr": 2.5381190740250305e-05}
{"train_loss": 0.084775909781456, "global_step": 180376, "epoch": 2026, "lr": 2.538068614931234e-05}
{"train_loss": 0.06829182058572769, "global_step": 180377, "epoch": 2026, "lr": 2.5380181561684152e-05}
{"train_loss": 0.08527187258005142, "global_step": 180378, "epoch": 2026, "lr": 2.5379676977365786e-05}
{"train_loss": 0.06910523027181625, "global_step": 180379, "epoch": 2026, "lr": 2.5379172396357337e-05}
{"train_loss": 0.09541063755750656, "global_step": 180380, "epoch": 2026, "lr": 2.5378667818658842e-05}
{"train_loss": 0.07048051059246063, "global_step": 180381, "epoch": 2026, "lr": 2.5378163244270403e-05}
{"train_loss": 0.041736479848623276, "global_step": 180382, "epoch": 2026, "lr": 2.537765867319205e-05}
{"train_loss": 0.040810465812683105, "global_step": 180383, "epoch": 2026, "lr": 2.5377154105423888e-05}
{"train_loss": 0.03463771566748619, "global_step": 180384, "epoch": 2026, "lr": 2.537664954096595e-05}
{"train_loss": 0.07726593315601349, "global_step": 180385, "epoch": 2026, "lr": 2.5376144979818345e-05}
{"train_loss": 0.049497369676828384, "global_step": 180386, "epoch": 2026, "lr": 2.5375640421981095e-05}
{"train_loss": 0.06843499094247818, "global_step": 180387, "epoch": 2026, "lr": 2.5375135867454308e-05}
{"train_loss": 0.1224108636379242, "global_step": 180388, "epoch": 2026, "lr": 2.5374631316238017e-05}
{"train_loss": 0.06741242110729218, "global_step": 180389, "epoch": 2026, "lr": 2.53741267683323e-05}
{"train_loss": 0.06813523173332214, "global_step": 180390, "epoch": 2026, "lr": 2.537362222373726e-05}
{"train_loss": 0.10959013551473618, "global_step": 180391, "epoch": 2026, "lr": 2.5373117682452907e-05}
{"train_loss": 0.028386395424604416, "global_step": 180392, "epoch": 2026, "lr": 2.5372613144479362e-05}
{"train_loss": 0.08785515278577805, "global_step": 180393, "epoch": 2026, "lr": 2.5372108609816657e-05}
{"train_loss": 0.024811750277876854, "global_step": 180394, "epoch": 2026, "lr": 2.5371604078464884e-05}
{"train_loss": 0.07116037607192993, "global_step": 180395, "epoch": 2026, "lr": 2.5371099550424082e-05}
{"train_loss": 0.05811881273984909, "global_step": 180396, "epoch": 2026, "lr": 2.5370595025694355e-05}
{"train_loss": 0.08666177093982697, "global_step": 180397, "epoch": 2026, "lr": 2.5370090504275724e-05}
{"train_loss": 0.06244739517569542, "global_step": 180398, "epoch": 2026, "lr": 2.5369585986168308e-05}
{"train_loss": 0.09007784724235535, "global_step": 180399, "epoch": 2026, "lr": 2.5369081471372136e-05}
{"train_loss": 0.10203618556261063, "global_step": 180400, "epoch": 2026, "lr": 2.5368576959887308e-05}
{"train_loss": 0.07714948058128357, "global_step": 180401, "epoch": 2026, "lr": 2.5368072451713877e-05}
{"train_loss": 0.07352950302570053, "global_step": 180402, "epoch": 2026, "lr": 2.5367567946851884e-05, "val_loss": 7.543066501617432}
{"train_loss": 0.08953731507062912, "global_step": 180403, "epoch": 2027, "lr": 2.536706344530144e-05}
{"train_loss": 0.0804663747549057, "global_step": 180404, "epoch": 2027, "lr": 2.536655894706258e-05}
{"train_loss": 0.03880785405635834, "global_step": 180405, "epoch": 2027, "lr": 2.53660544521354e-05}
{"train_loss": 0.09204652905464172, "global_step": 180406, "epoch": 2027, "lr": 2.5365549960519957e-05}
{"train_loss": 0.0848521962761879, "global_step": 180407, "epoch": 2027, "lr": 2.53650454722163e-05}
{"train_loss": 0.12027732282876968, "global_step": 180408, "epoch": 2027, "lr": 2.5364540987224507e-05}
{"train_loss": 0.12814916670322418, "global_step": 180409, "epoch": 2027, "lr": 2.536403650554467e-05}
{"train_loss": 0.13137750327587128, "global_step": 180410, "epoch": 2027, "lr": 2.5363532027176818e-05}
{"train_loss": 0.1213827133178711, "global_step": 180411, "epoch": 2027, "lr": 2.536302755212106e-05}
{"train_loss": 0.08720716089010239, "global_step": 180412, "epoch": 2027, "lr": 2.5362523080377426e-05}
{"train_loss": 0.09864004701375961, "global_step": 180413, "epoch": 2027, "lr": 2.5362018611946015e-05}
{"train_loss": 0.04502563551068306, "global_step": 180414, "epoch": 2027, "lr": 2.5361514146826866e-05}
{"train_loss": 0.09149874001741409, "global_step": 180415, "epoch": 2027, "lr": 2.5361009685020076e-05}
{"train_loss": 0.06025451421737671, "global_step": 180416, "epoch": 2027, "lr": 2.5360505226525683e-05}
{"train_loss": 0.038456231355667114, "global_step": 180417, "epoch": 2027, "lr": 2.5360000771343785e-05}
{"train_loss": 0.06405819952487946, "global_step": 180418, "epoch": 2027, "lr": 2.535949631947442e-05}
{"train_loss": 0.10907517373561859, "global_step": 180419, "epoch": 2027, "lr": 2.535899187091768e-05}
{"train_loss": 0.07019998133182526, "global_step": 180420, "epoch": 2027, "lr": 2.5358487425673615e-05}
{"train_loss": 0.0895381048321724, "global_step": 180421, "epoch": 2027, "lr": 2.535798298374231e-05}
{"train_loss": 0.05193352699279785, "global_step": 180422, "epoch": 2027, "lr": 2.5357478545123814e-05}
{"train_loss": 0.05360295623540878, "global_step": 180423, "epoch": 2027, "lr": 2.535697410981822e-05}
{"train_loss": 0.08643702417612076, "global_step": 180424, "epoch": 2027, "lr": 2.535646967782557e-05}
{"train_loss": 0.062385253608226776, "global_step": 180425, "epoch": 2027, "lr": 2.535596524914593e-05}
{"train_loss": 0.11320288479328156, "global_step": 180426, "epoch": 2027, "lr": 2.5355460823779398e-05}
{"train_loss": 0.05962373688817024, "global_step": 180427, "epoch": 2027, "lr": 2.5354956401726016e-05}
{"train_loss": 0.0649222657084465, "global_step": 180428, "epoch": 2027, "lr": 2.535445198298587e-05}
{"train_loss": 0.13919833302497864, "global_step": 180429, "epoch": 2027, "lr": 2.5353947567559e-05}
{"train_loss": 0.12460910528898239, "global_step": 180430, "epoch": 2027, "lr": 2.5353443155445512e-05}
{"train_loss": 0.07184509187936783, "global_step": 180431, "epoch": 2027, "lr": 2.535293874664543e-05}
{"train_loss": 0.07274443656206131, "global_step": 180432, "epoch": 2027, "lr": 2.5352434341158872e-05}
{"train_loss": 0.09248629957437515, "global_step": 180433, "epoch": 2027, "lr": 2.535192993898585e-05}
{"train_loss": 0.06375780701637268, "global_step": 180434, "epoch": 2027, "lr": 2.5351425540126487e-05}
{"train_loss": 0.061001840978860855, "global_step": 180435, "epoch": 2027, "lr": 2.5350921144580808e-05}
{"train_loss": 0.03883131593465805, "global_step": 180436, "epoch": 2027, "lr": 2.535041675234891e-05}
{"train_loss": 0.09927302598953247, "global_step": 180437, "epoch": 2027, "lr": 2.5349912363430828e-05}
{"train_loss": 0.07774465531110764, "global_step": 180438, "epoch": 2027, "lr": 2.5349407977826677e-05}
{"train_loss": 0.05302330479025841, "global_step": 180439, "epoch": 2027, "lr": 2.5348903595536487e-05}
{"train_loss": 0.03041091375052929, "global_step": 180440, "epoch": 2027, "lr": 2.5348399216560324e-05}
{"train_loss": 0.1332836002111435, "global_step": 180441, "epoch": 2027, "lr": 2.534789484089829e-05}
{"train_loss": 0.0844971239566803, "global_step": 180442, "epoch": 2027, "lr": 2.5347390468550403e-05}
{"train_loss": 0.03746424615383148, "global_step": 180443, "epoch": 2027, "lr": 2.5346886099516788e-05}
{"train_loss": 0.10004079341888428, "global_step": 180444, "epoch": 2027, "lr": 2.534638173379746e-05}
{"train_loss": 0.07056049257516861, "global_step": 180445, "epoch": 2027, "lr": 2.5345877371392524e-05}
{"train_loss": 0.13852958381175995, "global_step": 180446, "epoch": 2027, "lr": 2.534537301230202e-05}
{"train_loss": 0.05089012533426285, "global_step": 180447, "epoch": 2027, "lr": 2.5344868656526057e-05}
{"train_loss": 0.037919070571660995, "global_step": 180448, "epoch": 2027, "lr": 2.5344364304064644e-05}
{"train_loss": 0.09982271492481232, "global_step": 180449, "epoch": 2027, "lr": 2.5343859954917905e-05}
{"train_loss": 0.09017080813646317, "global_step": 180450, "epoch": 2027, "lr": 2.5343355609085867e-05}
{"train_loss": 0.1274445801973343, "global_step": 180451, "epoch": 2027, "lr": 2.5342851266568628e-05}
{"train_loss": 0.0586189329624176, "global_step": 180452, "epoch": 2027, "lr": 2.534234692736623e-05}
{"train_loss": 0.07261943817138672, "global_step": 180453, "epoch": 2027, "lr": 2.5341842591478764e-05}
{"train_loss": 0.0755317285656929, "global_step": 180454, "epoch": 2027, "lr": 2.5341338258906277e-05}
{"train_loss": 0.09638017416000366, "global_step": 180455, "epoch": 2027, "lr": 2.5340833929648855e-05}
{"train_loss": 0.06020433455705643, "global_step": 180456, "epoch": 2027, "lr": 2.5340329603706546e-05}
{"train_loss": 0.03311890363693237, "global_step": 180457, "epoch": 2027, "lr": 2.5339825281079443e-05}
{"train_loss": 0.10898347198963165, "global_step": 180458, "epoch": 2027, "lr": 2.5339320961767586e-05}
{"train_loss": 0.10910706967115402, "global_step": 180459, "epoch": 2027, "lr": 2.5338816645771074e-05}
{"train_loss": 0.04776909574866295, "global_step": 180460, "epoch": 2027, "lr": 2.5338312333089937e-05}
{"train_loss": 0.05365607514977455, "global_step": 180461, "epoch": 2027, "lr": 2.5337808023724287e-05}
{"train_loss": 0.07873239368200302, "global_step": 180462, "epoch": 2027, "lr": 2.5337303717674144e-05}
{"train_loss": 0.0900924876332283, "global_step": 180463, "epoch": 2027, "lr": 2.53367994149396e-05}
{"train_loss": 0.0466601699590683, "global_step": 180464, "epoch": 2027, "lr": 2.533629511552074e-05}
{"train_loss": 0.07119455933570862, "global_step": 180465, "epoch": 2027, "lr": 2.5335790819417605e-05}
{"train_loss": 0.12113337218761444, "global_step": 180466, "epoch": 2027, "lr": 2.533528652663028e-05}
{"train_loss": 0.02009836956858635, "global_step": 180467, "epoch": 2027, "lr": 2.533478223715881e-05}
{"train_loss": 0.06326813995838165, "global_step": 180468, "epoch": 2027, "lr": 2.53342779510033e-05}
{"train_loss": 0.07082716375589371, "global_step": 180469, "epoch": 2027, "lr": 2.5333773668163774e-05}
{"train_loss": 0.07994982600212097, "global_step": 180470, "epoch": 2027, "lr": 2.533326938864034e-05}
{"train_loss": 0.06966079771518707, "global_step": 180471, "epoch": 2027, "lr": 2.5332765112433032e-05}
{"train_loss": 0.10298717767000198, "global_step": 180472, "epoch": 2027, "lr": 2.5332260839541942e-05}
{"train_loss": 0.0959324985742569, "global_step": 180473, "epoch": 2027, "lr": 2.5331756569967125e-05}
{"train_loss": 0.040673960000276566, "global_step": 180474, "epoch": 2027, "lr": 2.5331252303708662e-05}
{"train_loss": 0.025186626240611076, "global_step": 180475, "epoch": 2027, "lr": 2.5330748040766594e-05}
{"train_loss": 0.08060676604509354, "global_step": 180476, "epoch": 2027, "lr": 2.533024378114103e-05}
{"train_loss": 0.04528970643877983, "global_step": 180477, "epoch": 2027, "lr": 2.5329739524832008e-05}
{"train_loss": 0.054060742259025574, "global_step": 180478, "epoch": 2027, "lr": 2.532923527183958e-05}
{"train_loss": 0.09592635929584503, "global_step": 180479, "epoch": 2027, "lr": 2.532873102216386e-05}
{"train_loss": 0.07945775985717773, "global_step": 180480, "epoch": 2027, "lr": 2.5328226775804876e-05}
{"train_loss": 0.09137257188558578, "global_step": 180481, "epoch": 2027, "lr": 2.5327722532762722e-05}
{"train_loss": 0.14206203818321228, "global_step": 180482, "epoch": 2027, "lr": 2.5327218293037434e-05}
{"train_loss": 0.09721891582012177, "global_step": 180483, "epoch": 2027, "lr": 2.532671405662913e-05}
{"train_loss": 0.0581417940557003, "global_step": 180484, "epoch": 2027, "lr": 2.5326209823537828e-05}
{"train_loss": 0.09908291697502136, "global_step": 180485, "epoch": 2027, "lr": 2.5325705593763626e-05}
{"train_loss": 0.012737590819597244, "global_step": 180486, "epoch": 2027, "lr": 2.5325201367306573e-05}
{"train_loss": 0.0832391083240509, "global_step": 180487, "epoch": 2027, "lr": 2.5324697144166766e-05}
{"train_loss": 0.060829758644104004, "global_step": 180488, "epoch": 2027, "lr": 2.5324192924344225e-05}
{"train_loss": 0.07279479503631592, "global_step": 180489, "epoch": 2027, "lr": 2.5323688707839068e-05}
{"train_loss": 0.08937536180019379, "global_step": 180490, "epoch": 2027, "lr": 2.5323184494651318e-05}
{"train_loss": 0.07832418528667996, "global_step": 180491, "epoch": 2027, "lr": 2.53226802847811e-05, "val_loss": 7.586510181427002}
{"train_loss": 0.06055750697851181, "global_step": 180492, "epoch": 2028, "lr": 2.532217607822841e-05}
{"train_loss": 0.05977838858962059, "global_step": 180493, "epoch": 2028, "lr": 2.532167187499338e-05}
{"train_loss": 0.07717740535736084, "global_step": 180494, "epoch": 2028, "lr": 2.532116767507603e-05}
{"train_loss": 0.06722727417945862, "global_step": 180495, "epoch": 2028, "lr": 2.532066347847646e-05}
{"train_loss": 0.042195893824100494, "global_step": 180496, "epoch": 2028, "lr": 2.5320159285194716e-05}
{"train_loss": 0.08340767025947571, "global_step": 180497, "epoch": 2028, "lr": 2.531965509523089e-05}
{"train_loss": 0.07261299341917038, "global_step": 180498, "epoch": 2028, "lr": 2.5319150908585022e-05}
{"train_loss": 0.0763348862528801, "global_step": 180499, "epoch": 2028, "lr": 2.531864672525719e-05}
{"train_loss": 0.07483074069023132, "global_step": 180500, "epoch": 2028, "lr": 2.531814254524748e-05}
{"train_loss": 0.11909832805395126, "global_step": 180501, "epoch": 2028, "lr": 2.5317638368555928e-05}
{"train_loss": 0.022093480452895164, "global_step": 180502, "epoch": 2028, "lr": 2.5317134195182647e-05}
{"train_loss": 0.06041647493839264, "global_step": 180503, "epoch": 2028, "lr": 2.5316630025127645e-05}
{"train_loss": 0.048141296952962875, "global_step": 180504, "epoch": 2028, "lr": 2.5316125858391047e-05}
{"train_loss": 0.04666493833065033, "global_step": 180505, "epoch": 2028, "lr": 2.5315621694972874e-05}
{"train_loss": 0.09648596495389938, "global_step": 180506, "epoch": 2028, "lr": 2.5315117534873233e-05}
{"train_loss": 0.11878056079149246, "global_step": 180507, "epoch": 2028, "lr": 2.5314613378092155e-05}
{"train_loss": 0.07902810722589493, "global_step": 180508, "epoch": 2028, "lr": 2.5314109224629746e-05}
{"train_loss": 0.09741265326738358, "global_step": 180509, "epoch": 2028, "lr": 2.5313605074486036e-05}
{"train_loss": 0.07208968698978424, "global_step": 180510, "epoch": 2028, "lr": 2.5313100927661133e-05}
{"train_loss": 0.0484481081366539, "global_step": 180511, "epoch": 2028, "lr": 2.5312596784155057e-05}
{"train_loss": 0.09965124726295471, "global_step": 180512, "epoch": 2028, "lr": 2.531209264396792e-05}
{"train_loss": 0.059821031987667084, "global_step": 180513, "epoch": 2028, "lr": 2.531158850709976e-05}
{"train_loss": 0.08317184448242188, "global_step": 180514, "epoch": 2028, "lr": 2.531108437355067e-05}
{"train_loss": 0.06862558424472809, "global_step": 180515, "epoch": 2028, "lr": 2.5310580243320687e-05}
{"train_loss": 0.09360744804143906, "global_step": 180516, "epoch": 2028, "lr": 2.5310076116409908e-05}
{"train_loss": 0.08318091183900833, "global_step": 180517, "epoch": 2028, "lr": 2.5309571992818393e-05}
{"train_loss": 0.1411631852388382, "global_step": 180518, "epoch": 2028, "lr": 2.5309067872546187e-05}
{"train_loss": 0.052598267793655396, "global_step": 180519, "epoch": 2028, "lr": 2.5308563755593395e-05}
{"train_loss": 0.06427457183599472, "global_step": 180520, "epoch": 2028, "lr": 2.5308059641960047e-05}
{"train_loss": 0.0736972913146019, "global_step": 180521, "epoch": 2028, "lr": 2.530755553164625e-05}
{"train_loss": 0.041008807718753815, "global_step": 180522, "epoch": 2028, "lr": 2.530705142465203e-05}
{"train_loss": 0.048221904784440994, "global_step": 180523, "epoch": 2028, "lr": 2.530654732097749e-05}
{"train_loss": 0.07081783562898636, "global_step": 180524, "epoch": 2028, "lr": 2.5306043220622673e-05}
{"train_loss": 0.07608785480260849, "global_step": 180525, "epoch": 2028, "lr": 2.5305539123587674e-05}
{"train_loss": 0.0665341466665268, "global_step": 180526, "epoch": 2028, "lr": 2.530503502987252e-05}
{"train_loss": 0.08581613749265671, "global_step": 180527, "epoch": 2028, "lr": 2.5304530939477327e-05}
{"train_loss": 0.08074156194925308, "global_step": 180528, "epoch": 2028, "lr": 2.5304026852402118e-05}
{"train_loss": 0.08390490710735321, "global_step": 180529, "epoch": 2028, "lr": 2.5303522768647003e-05}
{"train_loss": 0.06753307580947876, "global_step": 180530, "epoch": 2028, "lr": 2.530301868821201e-05}
{"train_loss": 0.06801164895296097, "global_step": 180531, "epoch": 2028, "lr": 2.5302514611097238e-05}
{"train_loss": 0.08207109570503235, "global_step": 180532, "epoch": 2028, "lr": 2.530201053730273e-05}
{"train_loss": 0.08502759784460068, "global_step": 180533, "epoch": 2028, "lr": 2.5301506466828583e-05}
{"train_loss": 0.06854802370071411, "global_step": 180534, "epoch": 2028, "lr": 2.5301002399674823e-05}
{"train_loss": 0.07705561816692352, "global_step": 180535, "epoch": 2028, "lr": 2.530049833584155e-05}
{"train_loss": 0.12542903423309326, "global_step": 180536, "epoch": 2028, "lr": 2.5299994275328837e-05}
{"train_loss": 0.09891017526388168, "global_step": 180537, "epoch": 2028, "lr": 2.5299490218136723e-05}
{"train_loss": 0.11373819410800934, "global_step": 180538, "epoch": 2028, "lr": 2.5298986164265304e-05}
{"train_loss": 0.044952392578125, "global_step": 180539, "epoch": 2028, "lr": 2.5298482113714626e-05}
{"train_loss": 0.05272788181900978, "global_step": 180540, "epoch": 2028, "lr": 2.529797806648478e-05}
{"train_loss": 0.11986126005649567, "global_step": 180541, "epoch": 2028, "lr": 2.52974740225758e-05}
{"train_loss": 0.048800766468048096, "global_step": 180542, "epoch": 2028, "lr": 2.5296969981987793e-05}
{"train_loss": 0.07262682914733887, "global_step": 180543, "epoch": 2028, "lr": 2.5296465944720795e-05}
{"train_loss": 0.05691325664520264, "global_step": 180544, "epoch": 2028, "lr": 2.5295961910774897e-05}
{"train_loss": 0.12295480817556381, "global_step": 180545, "epoch": 2028, "lr": 2.5295457880150137e-05}
{"train_loss": 0.08906977623701096, "global_step": 180546, "epoch": 2028, "lr": 2.5294953852846625e-05}
{"train_loss": 0.06608396023511887, "global_step": 180547, "epoch": 2028, "lr": 2.529444982886438e-05}
{"train_loss": 0.08553197979927063, "global_step": 180548, "epoch": 2028, "lr": 2.5293945808203522e-05}
{"train_loss": 0.05062933638691902, "global_step": 180549, "epoch": 2028, "lr": 2.529344179086407e-05}
{"train_loss": 0.1184106171131134, "global_step": 180550, "epoch": 2028, "lr": 2.5292937776846128e-05}
{"train_loss": 0.08065427094697952, "global_step": 180551, "epoch": 2028, "lr": 2.5292433766149736e-05}
{"train_loss": 0.13911280035972595, "global_step": 180552, "epoch": 2028, "lr": 2.529192975877499e-05}
{"train_loss": 0.11019536107778549, "global_step": 180553, "epoch": 2028, "lr": 2.5291425754721926e-05}
{"train_loss": 0.04082392156124115, "global_step": 180554, "epoch": 2028, "lr": 2.529092175399065e-05}
{"train_loss": 0.06926924735307693, "global_step": 180555, "epoch": 2028, "lr": 2.5290417756581203e-05}
{"train_loss": 0.07283692061901093, "global_step": 180556, "epoch": 2028, "lr": 2.5289913762493644e-05}
{"train_loss": 0.08756119012832642, "global_step": 180557, "epoch": 2028, "lr": 2.5289409771728066e-05}
{"train_loss": 0.13780727982521057, "global_step": 180558, "epoch": 2028, "lr": 2.5288905784284518e-05}
{"train_loss": 0.08032125979661942, "global_step": 180559, "epoch": 2028, "lr": 2.5288401800163087e-05}
{"train_loss": 0.08365621417760849, "global_step": 180560, "epoch": 2028, "lr": 2.5287897819363814e-05}
{"train_loss": 0.11325259506702423, "global_step": 180561, "epoch": 2028, "lr": 2.5287393841886797e-05}
{"train_loss": 0.06662879139184952, "global_step": 180562, "epoch": 2028, "lr": 2.5286889867732078e-05}
{"train_loss": 0.0842907577753067, "global_step": 180563, "epoch": 2028, "lr": 2.5286385896899743e-05}
{"train_loss": 0.09850763529539108, "global_step": 180564, "epoch": 2028, "lr": 2.528588192938984e-05}
{"train_loss": 0.06161456182599068, "global_step": 180565, "epoch": 2028, "lr": 2.5285377965202467e-05}
{"train_loss": 0.07577767223119736, "global_step": 180566, "epoch": 2028, "lr": 2.5284874004337654e-05}
{"train_loss": 0.07108642905950546, "global_step": 180567, "epoch": 2028, "lr": 2.5284370046795507e-05}
{"train_loss": 0.08483275771141052, "global_step": 180568, "epoch": 2028, "lr": 2.528386609257606e-05}
{"train_loss": 0.06598182767629623, "global_step": 180569, "epoch": 2028, "lr": 2.5283362141679412e-05}
{"train_loss": 0.06474192440509796, "global_step": 180570, "epoch": 2028, "lr": 2.5282858194105596e-05}
{"train_loss": 0.07511460036039352, "global_step": 180571, "epoch": 2028, "lr": 2.5282354249854693e-05}
{"train_loss": 0.08529149740934372, "global_step": 180572, "epoch": 2028, "lr": 2.5281850308926803e-05}
{"train_loss": 0.07674220204353333, "global_step": 180573, "epoch": 2028, "lr": 2.5281346371321945e-05}
{"train_loss": 0.07358517497777939, "global_step": 180574, "epoch": 2028, "lr": 2.528084243704023e-05}
{"train_loss": 0.0773773118853569, "global_step": 180575, "epoch": 2028, "lr": 2.528033850608168e-05}
{"train_loss": 0.11812450736761093, "global_step": 180576, "epoch": 2028, "lr": 2.5279834578446414e-05}
{"train_loss": 0.08891768008470535, "global_step": 180577, "epoch": 2028, "lr": 2.5279330654134447e-05}
{"train_loss": 0.04287772998213768, "global_step": 180578, "epoch": 2028, "lr": 2.5278826733145895e-05}
{"train_loss": 0.07813642919063568, "global_step": 180579, "epoch": 2028, "lr": 2.5278322815480782e-05}
{"train_loss": 0.07865412217261417, "global_step": 180580, "epoch": 2028, "lr": 2.5277818901139216e-05, "val_loss": 7.573599338531494}
{"train_loss": 0.05909208580851555, "global_step": 180581, "epoch": 2029, "lr": 2.527731499012123e-05}
{"train_loss": 0.12199917435646057, "global_step": 180582, "epoch": 2029, "lr": 2.5276811082426926e-05}
{"train_loss": 0.11092222481966019, "global_step": 180583, "epoch": 2029, "lr": 2.5276307178056334e-05}
{"train_loss": 0.0527801513671875, "global_step": 180584, "epoch": 2029, "lr": 2.5275803277009556e-05}
{"train_loss": 0.07181964814662933, "global_step": 180585, "epoch": 2029, "lr": 2.527529937928663e-05}
{"train_loss": 0.12146896123886108, "global_step": 180586, "epoch": 2029, "lr": 2.5274795484887663e-05}
{"train_loss": 0.10439430177211761, "global_step": 180587, "epoch": 2029, "lr": 2.527429159381267e-05}
{"train_loss": 0.05135917291045189, "global_step": 180588, "epoch": 2029, "lr": 2.5273787706061775e-05}
{"train_loss": 0.08104404807090759, "global_step": 180589, "epoch": 2029, "lr": 2.5273283821634997e-05}
{"train_loss": 0.11118867993354797, "global_step": 180590, "epoch": 2029, "lr": 2.5272779940532437e-05}
{"train_loss": 0.06534445285797119, "global_step": 180591, "epoch": 2029, "lr": 2.5272276062754136e-05}
{"train_loss": 0.04749518632888794, "global_step": 180592, "epoch": 2029, "lr": 2.5271772188300192e-05}
{"train_loss": 0.11693815141916275, "global_step": 180593, "epoch": 2029, "lr": 2.527126831717066e-05}
{"train_loss": 0.06865779310464859, "global_step": 180594, "epoch": 2029, "lr": 2.527076444936558e-05}
{"train_loss": 0.07543594390153885, "global_step": 180595, "epoch": 2029, "lr": 2.5270260584885074e-05}
{"train_loss": 0.053811054676771164, "global_step": 180596, "epoch": 2029, "lr": 2.5269756723729154e-05}
{"train_loss": 0.16159261763095856, "global_step": 180597, "epoch": 2029, "lr": 2.526925286589793e-05}
{"train_loss": 0.05655190721154213, "global_step": 180598, "epoch": 2029, "lr": 2.526874901139144e-05}
{"train_loss": 0.0863012820482254, "global_step": 180599, "epoch": 2029, "lr": 2.526824516020978e-05}
{"train_loss": 0.06949585676193237, "global_step": 180600, "epoch": 2029, "lr": 2.526774131235299e-05}
{"train_loss": 0.08298461139202118, "global_step": 180601, "epoch": 2029, "lr": 2.526723746782117e-05}
{"train_loss": 0.08241461217403412, "global_step": 180602, "epoch": 2029, "lr": 2.526673362661434e-05}
{"train_loss": 0.08413460850715637, "global_step": 180603, "epoch": 2029, "lr": 2.526622978873262e-05}
{"train_loss": 0.07326405495405197, "global_step": 180604, "epoch": 2029, "lr": 2.5265725954176035e-05}
{"train_loss": 0.10938724875450134, "global_step": 180605, "epoch": 2029, "lr": 2.5265222122944698e-05}
{"train_loss": 0.07624337822198868, "global_step": 180606, "epoch": 2029, "lr": 2.5264718295038626e-05}
{"train_loss": 0.06022348999977112, "global_step": 180607, "epoch": 2029, "lr": 2.5264214470457904e-05}
{"train_loss": 0.06628147512674332, "global_step": 180608, "epoch": 2029, "lr": 2.5263710649202633e-05}
{"train_loss": 0.07552685588598251, "global_step": 180609, "epoch": 2029, "lr": 2.5263206831272834e-05}
{"train_loss": 0.09434787184000015, "global_step": 180610, "epoch": 2029, "lr": 2.5262703016668615e-05}
{"train_loss": 0.13960443437099457, "global_step": 180611, "epoch": 2029, "lr": 2.5262199205390003e-05}
{"train_loss": 0.040386710315942764, "global_step": 180612, "epoch": 2029, "lr": 2.5261695397437114e-05}
{"train_loss": 0.07678096741437912, "global_step": 180613, "epoch": 2029, "lr": 2.5261191592809964e-05}
{"train_loss": 0.05166991800069809, "global_step": 180614, "epoch": 2029, "lr": 2.5260687791508658e-05}
{"train_loss": 0.07949139922857285, "global_step": 180615, "epoch": 2029, "lr": 2.5260183993533238e-05}
{"train_loss": 0.061527807265520096, "global_step": 180616, "epoch": 2029, "lr": 2.5259680198883807e-05}
{"train_loss": 0.1083824560046196, "global_step": 180617, "epoch": 2029, "lr": 2.525917640756038e-05}
{"train_loss": 0.0809541642665863, "global_step": 180618, "epoch": 2029, "lr": 2.5258672619563085e-05}
{"train_loss": 0.06226855143904686, "global_step": 180619, "epoch": 2029, "lr": 2.5258168834891937e-05}
{"train_loss": 0.05982407554984093, "global_step": 180620, "epoch": 2029, "lr": 2.5257665053547046e-05}
{"train_loss": 0.1296653300523758, "global_step": 180621, "epoch": 2029, "lr": 2.525716127552844e-05}
{"train_loss": 0.062327880412340164, "global_step": 180622, "epoch": 2029, "lr": 2.5256657500836234e-05}
{"train_loss": 0.11635730415582657, "global_step": 180623, "epoch": 2029, "lr": 2.5256153729470445e-05}
{"train_loss": 0.06764727085828781, "global_step": 180624, "epoch": 2029, "lr": 2.5255649961431184e-05}
{"train_loss": 0.03633025661110878, "global_step": 180625, "epoch": 2029, "lr": 2.525514619671847e-05}
{"train_loss": 0.08117100596427917, "global_step": 180626, "epoch": 2029, "lr": 2.525464243533241e-05}
{"train_loss": 0.07588592171669006, "global_step": 180627, "epoch": 2029, "lr": 2.525413867727308e-05}
{"train_loss": 0.08681046217679977, "global_step": 180628, "epoch": 2029, "lr": 2.5253634922540532e-05}
{"train_loss": 0.0914783850312233, "global_step": 180629, "epoch": 2029, "lr": 2.5253131171134803e-05}
{"train_loss": 0.03286832943558693, "global_step": 180630, "epoch": 2029, "lr": 2.525262742305602e-05}
{"train_loss": 0.12274570763111115, "global_step": 180631, "epoch": 2029, "lr": 2.525212367830419e-05}
{"train_loss": 0.05845780298113823, "global_step": 180632, "epoch": 2029, "lr": 2.5251619936879432e-05}
{"train_loss": 0.1239682212471962, "global_step": 180633, "epoch": 2029, "lr": 2.52511161987818e-05}
{"train_loss": 0.06375649571418762, "global_step": 180634, "epoch": 2029, "lr": 2.5250612464011324e-05}
{"train_loss": 0.1279466599225998, "global_step": 180635, "epoch": 2029, "lr": 2.5250108732568123e-05}
{"train_loss": 0.07952408492565155, "global_step": 180636, "epoch": 2029, "lr": 2.5249605004452232e-05}
{"train_loss": 0.08051832020282745, "global_step": 180637, "epoch": 2029, "lr": 2.5249101279663745e-05}
{"train_loss": 0.08043421059846878, "global_step": 180638, "epoch": 2029, "lr": 2.524859755820269e-05}
{"train_loss": 0.07337351888418198, "global_step": 180639, "epoch": 2029, "lr": 2.524809384006918e-05}
{"train_loss": 0.05459621548652649, "global_step": 180640, "epoch": 2029, "lr": 2.524759012526325e-05}
{"train_loss": 0.09088681638240814, "global_step": 180641, "epoch": 2029, "lr": 2.5247086413784993e-05}
{"train_loss": 0.05022461339831352, "global_step": 180642, "epoch": 2029, "lr": 2.5246582705634448e-05}
{"train_loss": 0.04965399578213692, "global_step": 180643, "epoch": 2029, "lr": 2.5246079000811696e-05}
{"train_loss": 0.08185112476348877, "global_step": 180644, "epoch": 2029, "lr": 2.5245575299316825e-05}
{"train_loss": 0.055939964950084686, "global_step": 180645, "epoch": 2029, "lr": 2.524507160114986e-05}
{"train_loss": 0.09035980701446533, "global_step": 180646, "epoch": 2029, "lr": 2.5244567906310924e-05}
{"train_loss": 0.08633832633495331, "global_step": 180647, "epoch": 2029, "lr": 2.524406421480003e-05}
{"train_loss": 0.09764096885919571, "global_step": 180648, "epoch": 2029, "lr": 2.5243560526617287e-05}
{"train_loss": 0.03725168853998184, "global_step": 180649, "epoch": 2029, "lr": 2.524305684176273e-05}
{"train_loss": 0.08848553895950317, "global_step": 180650, "epoch": 2029, "lr": 2.524255316023646e-05}
{"train_loss": 0.09561417251825333, "global_step": 180651, "epoch": 2029, "lr": 2.524204948203851e-05}
{"train_loss": 0.04886944964528084, "global_step": 180652, "epoch": 2029, "lr": 2.5241545807168977e-05}
{"train_loss": 0.1014712005853653, "global_step": 180653, "epoch": 2029, "lr": 2.52410421356279e-05}
{"train_loss": 0.025534749031066895, "global_step": 180654, "epoch": 2029, "lr": 2.5240538467415387e-05}
{"train_loss": 0.05812256410717964, "global_step": 180655, "epoch": 2029, "lr": 2.5240034802531458e-05}
{"train_loss": 0.038380131125450134, "global_step": 180656, "epoch": 2029, "lr": 2.5239531140976226e-05}
{"train_loss": 0.047916196286678314, "global_step": 180657, "epoch": 2029, "lr": 2.5239027482749722e-05}
{"train_loss": 0.07060867547988892, "global_step": 180658, "epoch": 2029, "lr": 2.523852382785204e-05}
{"train_loss": 0.09002706408500671, "global_step": 180659, "epoch": 2029, "lr": 2.5238020176283222e-05}
{"train_loss": 0.053387854248285294, "global_step": 180660, "epoch": 2029, "lr": 2.523751652804337e-05}
{"train_loss": 0.11245058476924896, "global_step": 180661, "epoch": 2029, "lr": 2.523701288313251e-05}
{"train_loss": 0.08912943303585052, "global_step": 180662, "epoch": 2029, "lr": 2.5236509241550733e-05}
{"train_loss": 0.06684534251689911, "global_step": 180663, "epoch": 2029, "lr": 2.523600560329813e-05}
{"train_loss": 0.09137263149023056, "global_step": 180664, "epoch": 2029, "lr": 2.523550196837472e-05}
{"train_loss": 0.04445875063538551, "global_step": 180665, "epoch": 2029, "lr": 2.5234998336780614e-05}
{"train_loss": 0.07598415017127991, "global_step": 180666, "epoch": 2029, "lr": 2.5234494708515854e-05}
{"train_loss": 0.11365704238414764, "global_step": 180667, "epoch": 2029, "lr": 2.5233991083580506e-05}
{"train_loss": 0.10449281334877014, "global_step": 180668, "epoch": 2029, "lr": 2.523348746197466e-05}
{"train_loss": 0.07863660085569607, "global_step": 180669, "epoch": 2029, "lr": 2.5232983843698355e-05, "val_loss": 7.579472064971924}
{"train_loss": 0.049112170934677124, "global_step": 180670, "epoch": 2030, "lr": 2.5232480228751688e-05}
{"train_loss": 0.13864745199680328, "global_step": 180671, "epoch": 2030, "lr": 2.5231976617134713e-05}
{"train_loss": 0.08613815158605576, "global_step": 180672, "epoch": 2030, "lr": 2.523147300884748e-05}
{"train_loss": 0.03195727616548538, "global_step": 180673, "epoch": 2030, "lr": 2.5230969403890088e-05}
{"train_loss": 0.12291714549064636, "global_step": 180674, "epoch": 2030, "lr": 2.523046580226257e-05}
{"train_loss": 0.06517307460308075, "global_step": 180675, "epoch": 2030, "lr": 2.522996220396504e-05}
{"train_loss": 0.034680530428886414, "global_step": 180676, "epoch": 2030, "lr": 2.522945860899752e-05}
{"train_loss": 0.08281558752059937, "global_step": 180677, "epoch": 2030, "lr": 2.5228955017360112e-05}
{"train_loss": 0.0344545803964138, "global_step": 180678, "epoch": 2030, "lr": 2.5228451429052857e-05}
{"train_loss": 0.05411330983042717, "global_step": 180679, "epoch": 2030, "lr": 2.5227947844075828e-05}
{"train_loss": 0.04587572440505028, "global_step": 180680, "epoch": 2030, "lr": 2.5227444262429122e-05}
{"train_loss": 0.08878762274980545, "global_step": 180681, "epoch": 2030, "lr": 2.522694068411276e-05}
{"train_loss": 0.10405617952346802, "global_step": 180682, "epoch": 2030, "lr": 2.5226437109126854e-05}
{"train_loss": 0.09023384749889374, "global_step": 180683, "epoch": 2030, "lr": 2.5225933537471437e-05}
{"train_loss": 0.12058699876070023, "global_step": 180684, "epoch": 2030, "lr": 2.5225429969146608e-05}
{"train_loss": 0.15275928378105164, "global_step": 180685, "epoch": 2030, "lr": 2.5224926404152403e-05}
{"train_loss": 0.10002836585044861, "global_step": 180686, "epoch": 2030, "lr": 2.522442284248892e-05}
{"train_loss": 0.09273400157690048, "global_step": 180687, "epoch": 2030, "lr": 2.5223919284156196e-05}
{"train_loss": 0.023936627432703972, "global_step": 180688, "epoch": 2030, "lr": 2.5223415729154327e-05}
{"train_loss": 0.08804072439670563, "global_step": 180689, "epoch": 2030, "lr": 2.5222912177483348e-05}
{"train_loss": 0.17034943401813507, "global_step": 180690, "epoch": 2030, "lr": 2.5222408629143367e-05}
{"train_loss": 0.05928238108754158, "global_step": 180691, "epoch": 2030, "lr": 2.5221905084134417e-05}
{"train_loss": 0.07432974129915237, "global_step": 180692, "epoch": 2030, "lr": 2.5221401542456598e-05}
{"train_loss": 0.054033223539590836, "global_step": 180693, "epoch": 2030, "lr": 2.5220898004109938e-05}
{"train_loss": 0.038989074528217316, "global_step": 180694, "epoch": 2030, "lr": 2.5220394469094544e-05}
{"train_loss": 0.03542742505669594, "global_step": 180695, "epoch": 2030, "lr": 2.521989093741045e-05}
{"train_loss": 0.09550893306732178, "global_step": 180696, "epoch": 2030, "lr": 2.521938740905776e-05}
{"train_loss": 0.06267623603343964, "global_step": 180697, "epoch": 2030, "lr": 2.5218883884036493e-05}
{"train_loss": 0.123113252222538, "global_step": 180698, "epoch": 2030, "lr": 2.5218380362346754e-05}
{"train_loss": 0.06910198926925659, "global_step": 180699, "epoch": 2030, "lr": 2.5217876843988618e-05}
{"train_loss": 0.05054222047328949, "global_step": 180700, "epoch": 2030, "lr": 2.5217373328962118e-05}
{"train_loss": 0.11738548427820206, "global_step": 180701, "epoch": 2030, "lr": 2.5216869817267363e-05}
{"train_loss": 0.0808490514755249, "global_step": 180702, "epoch": 2030, "lr": 2.5216366308904373e-05}
{"train_loss": 0.07108188420534134, "global_step": 180703, "epoch": 2030, "lr": 2.5215862803873257e-05}
{"train_loss": 0.13711346685886383, "global_step": 180704, "epoch": 2030, "lr": 2.5215359302174075e-05}
{"train_loss": 0.18211863934993744, "global_step": 180705, "epoch": 2030, "lr": 2.521485580380686e-05}
{"train_loss": 0.0501299686729908, "global_step": 180706, "epoch": 2030, "lr": 2.5214352308771728e-05}
{"train_loss": 0.05702392756938934, "global_step": 180707, "epoch": 2030, "lr": 2.5213848817068703e-05}
{"train_loss": 0.08585847169160843, "global_step": 180708, "epoch": 2030, "lr": 2.521334532869789e-05}
{"train_loss": 0.08444827049970627, "global_step": 180709, "epoch": 2030, "lr": 2.521284184365933e-05}
{"train_loss": 0.09026739001274109, "global_step": 180710, "epoch": 2030, "lr": 2.5212338361953113e-05}
{"train_loss": 0.059334538877010345, "global_step": 180711, "epoch": 2030, "lr": 2.52118348835793e-05}
{"train_loss": 0.07970749586820602, "global_step": 180712, "epoch": 2030, "lr": 2.521133140853793e-05}
{"train_loss": 0.07031477242708206, "global_step": 180713, "epoch": 2030, "lr": 2.521082793682912e-05}
{"train_loss": 0.07620540261268616, "global_step": 180714, "epoch": 2030, "lr": 2.5210324468452883e-05}
{"train_loss": 0.07727251946926117, "global_step": 180715, "epoch": 2030, "lr": 2.5209821003409346e-05}
{"train_loss": 0.12447474896907806, "global_step": 180716, "epoch": 2030, "lr": 2.5209317541698518e-05}
{"train_loss": 0.025592608377337456, "global_step": 180717, "epoch": 2030, "lr": 2.5208814083320496e-05}
{"train_loss": 0.06120971962809563, "global_step": 180718, "epoch": 2030, "lr": 2.5208310628275377e-05}
{"train_loss": 0.07077614963054657, "global_step": 180719, "epoch": 2030, "lr": 2.520780717656317e-05}
{"train_loss": 0.08605993539094925, "global_step": 180720, "epoch": 2030, "lr": 2.520730372818399e-05}
{"train_loss": 0.1606069654226303, "global_step": 180721, "epoch": 2030, "lr": 2.5206800283137867e-05}
{"train_loss": 0.05027580261230469, "global_step": 180722, "epoch": 2030, "lr": 2.520629684142491e-05}
{"train_loss": 0.08598556369543076, "global_step": 180723, "epoch": 2030, "lr": 2.520579340304514e-05}
{"train_loss": 0.09709200263023376, "global_step": 180724, "epoch": 2030, "lr": 2.520528996799867e-05}
{"train_loss": 0.08840761333703995, "global_step": 180725, "epoch": 2030, "lr": 2.520478653628553e-05}
{"train_loss": 0.08943834155797958, "global_step": 180726, "epoch": 2030, "lr": 2.5204283107905823e-05}
{"train_loss": 0.11281240731477737, "global_step": 180727, "epoch": 2030, "lr": 2.5203779682859575e-05}
{"train_loss": 0.09564340114593506, "global_step": 180728, "epoch": 2030, "lr": 2.5203276261146903e-05}
{"train_loss": 0.12625378370285034, "global_step": 180729, "epoch": 2030, "lr": 2.520277284276782e-05}
{"train_loss": 0.061475712805986404, "global_step": 180730, "epoch": 2030, "lr": 2.5202269427722446e-05}
{"train_loss": 0.07468195259571075, "global_step": 180731, "epoch": 2030, "lr": 2.5201766016010807e-05}
{"train_loss": 0.12717321515083313, "global_step": 180732, "epoch": 2030, "lr": 2.5201262607633e-05}
{"train_loss": 0.06560166925191879, "global_step": 180733, "epoch": 2030, "lr": 2.5200759202589074e-05}
{"train_loss": 0.09004736691713333, "global_step": 180734, "epoch": 2030, "lr": 2.520025580087909e-05}
{"train_loss": 0.10864683985710144, "global_step": 180735, "epoch": 2030, "lr": 2.5199752402503163e-05}
{"train_loss": 0.06216692924499512, "global_step": 180736, "epoch": 2030, "lr": 2.51992490074613e-05}
{"train_loss": 0.12797516584396362, "global_step": 180737, "epoch": 2030, "lr": 2.5198745615753616e-05}
{"train_loss": 0.11586907505989075, "global_step": 180738, "epoch": 2030, "lr": 2.5198242227380137e-05}
{"train_loss": 0.1260300874710083, "global_step": 180739, "epoch": 2030, "lr": 2.519773884234097e-05}
{"train_loss": 0.055748939514160156, "global_step": 180740, "epoch": 2030, "lr": 2.519723546063615e-05}
{"train_loss": 0.06278438866138458, "global_step": 180741, "epoch": 2030, "lr": 2.5196732082265774e-05}
{"train_loss": 0.03996715322136879, "global_step": 180742, "epoch": 2030, "lr": 2.519622870722988e-05}
{"train_loss": 0.044855453073978424, "global_step": 180743, "epoch": 2030, "lr": 2.519572533552857e-05}
{"train_loss": 0.19002509117126465, "global_step": 180744, "epoch": 2030, "lr": 2.5195221967161887e-05}
{"train_loss": 0.04778723046183586, "global_step": 180745, "epoch": 2030, "lr": 2.519471860212989e-05}
{"train_loss": 0.088618203997612, "global_step": 180746, "epoch": 2030, "lr": 2.5194215240432685e-05}
{"train_loss": 0.06390415132045746, "global_step": 180747, "epoch": 2030, "lr": 2.5193711882070288e-05}
{"train_loss": 0.09495057910680771, "global_step": 180748, "epoch": 2030, "lr": 2.5193208527042812e-05}
{"train_loss": 0.04388478770852089, "global_step": 180749, "epoch": 2030, "lr": 2.519270517535031e-05}
{"train_loss": 0.12468206882476807, "global_step": 180750, "epoch": 2030, "lr": 2.5192201826992827e-05}
{"train_loss": 0.05021032691001892, "global_step": 180751, "epoch": 2030, "lr": 2.519169848197047e-05}
{"train_loss": 0.10270734131336212, "global_step": 180752, "epoch": 2030, "lr": 2.519119514028327e-05}
{"train_loss": 0.13851463794708252, "global_step": 180753, "epoch": 2030, "lr": 2.519069180193131e-05}
{"train_loss": 0.045482080429792404, "global_step": 180754, "epoch": 2030, "lr": 2.5190188466914676e-05}
{"train_loss": 0.10328630357980728, "global_step": 180755, "epoch": 2030, "lr": 2.5189685135233397e-05}
{"train_loss": 0.09550056606531143, "global_step": 180756, "epoch": 2030, "lr": 2.5189181806887584e-05}
{"train_loss": 0.1072104200720787, "global_step": 180757, "epoch": 2030, "lr": 2.518867848187727e-05}
{"train_loss": 0.08482277104526423, "global_step": 180758, "epoch": 2030, "lr": 2.5188175160202542e-05, "val_loss": 7.447945594787598, "train_action_mse_error": 10.93588638305664}
{"train_loss": 0.08284749835729599, "global_step": 180759, "epoch": 2031, "lr": 2.5187671841863458e-05}
{"train_loss": 0.04961176961660385, "global_step": 180760, "epoch": 2031, "lr": 2.5187168526860095e-05}
{"train_loss": 0.0743120089173317, "global_step": 180761, "epoch": 2031, "lr": 2.5186665215192506e-05}
{"train_loss": 0.016526658087968826, "global_step": 180762, "epoch": 2031, "lr": 2.5186161906860778e-05}
{"train_loss": 0.06505695730447769, "global_step": 180763, "epoch": 2031, "lr": 2.5185658601864953e-05}
{"train_loss": 0.10254556685686111, "global_step": 180764, "epoch": 2031, "lr": 2.5185155300205134e-05}
{"train_loss": 0.0386030450463295, "global_step": 180765, "epoch": 2031, "lr": 2.5184652001881347e-05}
{"train_loss": 0.06914971768856049, "global_step": 180766, "epoch": 2031, "lr": 2.51841487068937e-05}
{"train_loss": 0.10179047286510468, "global_step": 180767, "epoch": 2031, "lr": 2.5183645415242222e-05}
{"train_loss": 0.10462816059589386, "global_step": 180768, "epoch": 2031, "lr": 2.5183142126927024e-05}
{"train_loss": 0.06310328841209412, "global_step": 180769, "epoch": 2031, "lr": 2.5182638841948126e-05}
{"train_loss": 0.042226724326610565, "global_step": 180770, "epoch": 2031, "lr": 2.518213556030562e-05}
{"train_loss": 0.06910116225481033, "global_step": 180771, "epoch": 2031, "lr": 2.51816322819996e-05}
{"train_loss": 0.08938940614461899, "global_step": 180772, "epoch": 2031, "lr": 2.518112900703008e-05}
{"train_loss": 0.06846644729375839, "global_step": 180773, "epoch": 2031, "lr": 2.518062573539718e-05}
{"train_loss": 0.04714185371994972, "global_step": 180774, "epoch": 2031, "lr": 2.5180122467100913e-05}
{"train_loss": 0.13179156184196472, "global_step": 180775, "epoch": 2031, "lr": 2.51796192021414e-05}
{"train_loss": 0.07827231287956238, "global_step": 180776, "epoch": 2031, "lr": 2.517911594051867e-05}
{"train_loss": 0.09063445031642914, "global_step": 180777, "epoch": 2031, "lr": 2.517861268223282e-05}
{"train_loss": 0.0909208357334137, "global_step": 180778, "epoch": 2031, "lr": 2.517810942728389e-05}
{"train_loss": 0.10621387511491776, "global_step": 180779, "epoch": 2031, "lr": 2.5177606175671974e-05}
{"train_loss": 0.06363733857870102, "global_step": 180780, "epoch": 2031, "lr": 2.5177102927397104e-05}
{"train_loss": 0.043091025203466415, "global_step": 180781, "epoch": 2031, "lr": 2.5176599682459395e-05}
{"train_loss": 0.0693720281124115, "global_step": 180782, "epoch": 2031, "lr": 2.5176096440858887e-05}
{"train_loss": 0.07729179412126541, "global_step": 180783, "epoch": 2031, "lr": 2.517559320259564e-05}
{"train_loss": 0.1426144391298294, "global_step": 180784, "epoch": 2031, "lr": 2.517508996766974e-05}
{"train_loss": 0.07943800836801529, "global_step": 180785, "epoch": 2031, "lr": 2.517458673608123e-05}
{"train_loss": 0.09399136900901794, "global_step": 180786, "epoch": 2031, "lr": 2.5174083507830215e-05}
{"train_loss": 0.0657188668847084, "global_step": 180787, "epoch": 2031, "lr": 2.5173580282916744e-05}
{"train_loss": 0.08322039991617203, "global_step": 180788, "epoch": 2031, "lr": 2.5173077061340854e-05}
{"train_loss": 0.06881151348352432, "global_step": 180789, "epoch": 2031, "lr": 2.517257384310265e-05}
{"train_loss": 0.06418199092149734, "global_step": 180790, "epoch": 2031, "lr": 2.5172070628202206e-05}
{"train_loss": 0.08086603879928589, "global_step": 180791, "epoch": 2031, "lr": 2.5171567416639557e-05}
{"train_loss": 0.06426656991243362, "global_step": 180792, "epoch": 2031, "lr": 2.5171064208414806e-05}
{"train_loss": 0.05597047135233879, "global_step": 180793, "epoch": 2031, "lr": 2.5170561003527983e-05}
{"train_loss": 0.10642709583044052, "global_step": 180794, "epoch": 2031, "lr": 2.5170057801979198e-05}
{"train_loss": 0.07189760357141495, "global_step": 180795, "epoch": 2031, "lr": 2.5169554603768474e-05}
{"train_loss": 0.08067180961370468, "global_step": 180796, "epoch": 2031, "lr": 2.5169051408895928e-05}
{"train_loss": 0.10261495411396027, "global_step": 180797, "epoch": 2031, "lr": 2.516854821736157e-05}
{"train_loss": 0.05947288125753403, "global_step": 180798, "epoch": 2031, "lr": 2.516804502916552e-05}
{"train_loss": 0.07608543336391449, "global_step": 180799, "epoch": 2031, "lr": 2.5167541844307806e-05}
{"train_loss": 0.08505652844905853, "global_step": 180800, "epoch": 2031, "lr": 2.5167038662788535e-05}
{"train_loss": 0.06314440816640854, "global_step": 180801, "epoch": 2031, "lr": 2.5166535484607735e-05}
{"train_loss": 0.047942906618118286, "global_step": 180802, "epoch": 2031, "lr": 2.5166032309765502e-05}
{"train_loss": 0.04365022853016853, "global_step": 180803, "epoch": 2031, "lr": 2.516552913826188e-05}
{"train_loss": 0.08593684434890747, "global_step": 180804, "epoch": 2031, "lr": 2.5165025970096967e-05}
{"train_loss": 0.04417479410767555, "global_step": 180805, "epoch": 2031, "lr": 2.5164522805270797e-05}
{"train_loss": 0.06770959496498108, "global_step": 180806, "epoch": 2031, "lr": 2.5164019643783455e-05}
{"train_loss": 0.10709050297737122, "global_step": 180807, "epoch": 2031, "lr": 2.5163516485635025e-05}
{"train_loss": 0.05861888825893402, "global_step": 180808, "epoch": 2031, "lr": 2.5163013330825535e-05}
{"train_loss": 0.047610439360141754, "global_step": 180809, "epoch": 2031, "lr": 2.5162510179355102e-05}
{"train_loss": 0.040838874876499176, "global_step": 180810, "epoch": 2031, "lr": 2.5162007031223745e-05}
{"train_loss": 0.08383036404848099, "global_step": 180811, "epoch": 2031, "lr": 2.516150388643157e-05}
{"train_loss": 0.027164096012711525, "global_step": 180812, "epoch": 2031, "lr": 2.5161000744978608e-05}
{"train_loss": 0.08092810958623886, "global_step": 180813, "epoch": 2031, "lr": 2.5160497606864964e-05}
{"train_loss": 0.08763951063156128, "global_step": 180814, "epoch": 2031, "lr": 2.5159994472090676e-05}
{"train_loss": 0.05982159450650215, "global_step": 180815, "epoch": 2031, "lr": 2.515949134065584e-05}
{"train_loss": 0.09691818803548813, "global_step": 180816, "epoch": 2031, "lr": 2.5158988212560487e-05}
{"train_loss": 0.0760498195886612, "global_step": 180817, "epoch": 2031, "lr": 2.5158485087804728e-05}
{"train_loss": 0.1017884612083435, "global_step": 180818, "epoch": 2031, "lr": 2.5157981966388587e-05}
{"train_loss": 0.0762212872505188, "global_step": 180819, "epoch": 2031, "lr": 2.5157478848312164e-05}
{"train_loss": 0.07127401977777481, "global_step": 180820, "epoch": 2031, "lr": 2.5156975733575528e-05}
{"train_loss": 0.08636625856161118, "global_step": 180821, "epoch": 2031, "lr": 2.5156472622178706e-05}
{"train_loss": 0.06795116513967514, "global_step": 180822, "epoch": 2031, "lr": 2.515596951412182e-05}
{"train_loss": 0.07079143077135086, "global_step": 180823, "epoch": 2031, "lr": 2.5155466409404888e-05}
{"train_loss": 0.05257626622915268, "global_step": 180824, "epoch": 2031, "lr": 2.5154963308028018e-05}
{"train_loss": 0.05998048558831215, "global_step": 180825, "epoch": 2031, "lr": 2.5154460209991243e-05}
{"train_loss": 0.09617189317941666, "global_step": 180826, "epoch": 2031, "lr": 2.5153957115294668e-05}
{"train_loss": 0.0658603087067604, "global_step": 180827, "epoch": 2031, "lr": 2.5153454023938316e-05}
{"train_loss": 0.08147840201854706, "global_step": 180828, "epoch": 2031, "lr": 2.5152950935922304e-05}
{"train_loss": 0.031511541455984116, "global_step": 180829, "epoch": 2031, "lr": 2.5152447851246657e-05}
{"train_loss": 0.11248332262039185, "global_step": 180830, "epoch": 2031, "lr": 2.5151944769911478e-05}
{"train_loss": 0.08657345175743103, "global_step": 180831, "epoch": 2031, "lr": 2.51514416919168e-05}
{"train_loss": 0.12275943160057068, "global_step": 180832, "epoch": 2031, "lr": 2.515093861726272e-05}
{"train_loss": 0.04965579882264137, "global_step": 180833, "epoch": 2031, "lr": 2.515043554594928e-05}
{"train_loss": 0.06707438826560974, "global_step": 180834, "epoch": 2031, "lr": 2.5149932477976578e-05}
{"train_loss": 0.0660536140203476, "global_step": 180835, "epoch": 2031, "lr": 2.514942941334465e-05}
{"train_loss": 0.11050577461719513, "global_step": 180836, "epoch": 2031, "lr": 2.514892635205359e-05}
{"train_loss": 0.10211339592933655, "global_step": 180837, "epoch": 2031, "lr": 2.514842329410344e-05}
{"train_loss": 0.10263580083847046, "global_step": 180838, "epoch": 2031, "lr": 2.51479202394943e-05}
{"train_loss": 0.07541051506996155, "global_step": 180839, "epoch": 2031, "lr": 2.5147417188226197e-05}
{"train_loss": 0.06797652691602707, "global_step": 180840, "epoch": 2031, "lr": 2.5146914140299244e-05}
{"train_loss": 0.15167558193206787, "global_step": 180841, "epoch": 2031, "lr": 2.5146411095713463e-05}
{"train_loss": 0.04849681258201599, "global_step": 180842, "epoch": 2031, "lr": 2.514590805446897e-05}
{"train_loss": 0.04472963139414787, "global_step": 180843, "epoch": 2031, "lr": 2.514540501656578e-05}
{"train_loss": 0.09609227627515793, "global_step": 180844, "epoch": 2031, "lr": 2.5144901982003988e-05}
{"train_loss": 0.03949647396802902, "global_step": 180845, "epoch": 2031, "lr": 2.5144398950783686e-05}
{"train_loss": 0.10054581612348557, "global_step": 180846, "epoch": 2031, "lr": 2.514389592290489e-05}
{"train_loss": 0.07578097591490558, "global_step": 180847, "epoch": 2031, "lr": 2.5143392898367716e-05, "val_loss": 7.449930667877197}
{"train_loss": 0.06616166234016418, "global_step": 180848, "epoch": 2032, "lr": 2.5142889877172195e-05}
{"train_loss": 0.06296262890100479, "global_step": 180849, "epoch": 2032, "lr": 2.514238685931842e-05}
{"train_loss": 0.09525265544652939, "global_step": 180850, "epoch": 2032, "lr": 2.5141883844806434e-05}
{"train_loss": 0.12808352708816528, "global_step": 180851, "epoch": 2032, "lr": 2.514138083363634e-05}
{"train_loss": 0.0661432147026062, "global_step": 180852, "epoch": 2032, "lr": 2.5140877825808164e-05}
{"train_loss": 0.06803227961063385, "global_step": 180853, "epoch": 2032, "lr": 2.5140374821322017e-05}
{"train_loss": 0.03768845647573471, "global_step": 180854, "epoch": 2032, "lr": 2.5139871820177917e-05}
{"train_loss": 0.0938461422920227, "global_step": 180855, "epoch": 2032, "lr": 2.5139368822375982e-05}
{"train_loss": 0.10487799346446991, "global_step": 180856, "epoch": 2032, "lr": 2.5138865827916235e-05}
{"train_loss": 0.0787719190120697, "global_step": 180857, "epoch": 2032, "lr": 2.513836283679878e-05}
{"train_loss": 0.06356489658355713, "global_step": 180858, "epoch": 2032, "lr": 2.513785984902366e-05}
{"train_loss": 0.09234555065631866, "global_step": 180859, "epoch": 2032, "lr": 2.513735686459096e-05}
{"train_loss": 0.0759865865111351, "global_step": 180860, "epoch": 2032, "lr": 2.5136853883500743e-05}
{"train_loss": 0.09707071632146835, "global_step": 180861, "epoch": 2032, "lr": 2.5136350905753058e-05}
{"train_loss": 0.07740329205989838, "global_step": 180862, "epoch": 2032, "lr": 2.5135847931348e-05}
{"train_loss": 0.04875617101788521, "global_step": 180863, "epoch": 2032, "lr": 2.5135344960285613e-05}
{"train_loss": 0.10992581397294998, "global_step": 180864, "epoch": 2032, "lr": 2.513484199256599e-05}
{"train_loss": 0.09404582530260086, "global_step": 180865, "epoch": 2032, "lr": 2.5134339028189163e-05}
{"train_loss": 0.015069265849888325, "global_step": 180866, "epoch": 2032, "lr": 2.5133836067155244e-05}
{"train_loss": 0.07679805904626846, "global_step": 180867, "epoch": 2032, "lr": 2.5133333109464252e-05}
{"train_loss": 0.06627985835075378, "global_step": 180868, "epoch": 2032, "lr": 2.513283015511631e-05}
{"train_loss": 0.10529676079750061, "global_step": 180869, "epoch": 2032, "lr": 2.5132327204111427e-05}
{"train_loss": 0.09869539737701416, "global_step": 180870, "epoch": 2032, "lr": 2.5131824256449726e-05}
{"train_loss": 0.041986603289842606, "global_step": 180871, "epoch": 2032, "lr": 2.513132131213123e-05}
{"train_loss": 0.11814309656620026, "global_step": 180872, "epoch": 2032, "lr": 2.513081837115604e-05}
{"train_loss": 0.0618947371840477, "global_step": 180873, "epoch": 2032, "lr": 2.5130315433524188e-05}
{"train_loss": 0.0815926343202591, "global_step": 180874, "epoch": 2032, "lr": 2.5129812499235782e-05}
{"train_loss": 0.05855714902281761, "global_step": 180875, "epoch": 2032, "lr": 2.512930956829085e-05}
{"train_loss": 0.03631982579827309, "global_step": 180876, "epoch": 2032, "lr": 2.5128806640689496e-05}
{"train_loss": 0.04863744229078293, "global_step": 180877, "epoch": 2032, "lr": 2.5128303716431756e-05}
{"train_loss": 0.06541815400123596, "global_step": 180878, "epoch": 2032, "lr": 2.5127800795517732e-05}
{"train_loss": 0.09981933236122131, "global_step": 180879, "epoch": 2032, "lr": 2.5127297877947452e-05}
{"train_loss": 0.07799345254898071, "global_step": 180880, "epoch": 2032, "lr": 2.5126794963721e-05}
{"train_loss": 0.027803009375929832, "global_step": 180881, "epoch": 2032, "lr": 2.5126292052838473e-05}
{"train_loss": 0.0579567588865757, "global_step": 180882, "epoch": 2032, "lr": 2.512578914529989e-05}
{"train_loss": 0.12348924577236176, "global_step": 180883, "epoch": 2032, "lr": 2.512528624110536e-05}
{"train_loss": 0.07743117958307266, "global_step": 180884, "epoch": 2032, "lr": 2.512478334025491e-05}
{"train_loss": 0.11852674931287766, "global_step": 180885, "epoch": 2032, "lr": 2.5124280442748653e-05}
{"train_loss": 0.04451176896691322, "global_step": 180886, "epoch": 2032, "lr": 2.5123777548586613e-05}
{"train_loss": 0.09335223585367203, "global_step": 180887, "epoch": 2032, "lr": 2.5123274657768903e-05}
{"train_loss": 0.09404053539037704, "global_step": 180888, "epoch": 2032, "lr": 2.512277177029554e-05}
{"train_loss": 0.07982996851205826, "global_step": 180889, "epoch": 2032, "lr": 2.5122268886166633e-05}
{"train_loss": 0.1016620397567749, "global_step": 180890, "epoch": 2032, "lr": 2.5121766005382218e-05}
{"train_loss": 0.1268412321805954, "global_step": 180891, "epoch": 2032, "lr": 2.51212631279424e-05}
{"train_loss": 0.09682750701904297, "global_step": 180892, "epoch": 2032, "lr": 2.5120760253847208e-05}
{"train_loss": 0.06857834756374359, "global_step": 180893, "epoch": 2032, "lr": 2.512025738309674e-05}
{"train_loss": 0.05371113494038582, "global_step": 180894, "epoch": 2032, "lr": 2.511975451569103e-05}
{"train_loss": 0.12611036002635956, "global_step": 180895, "epoch": 2032, "lr": 2.5119251651630193e-05}
{"train_loss": 0.05005314573645592, "global_step": 180896, "epoch": 2032, "lr": 2.511874879091424e-05}
{"train_loss": 0.11987099796533585, "global_step": 180897, "epoch": 2032, "lr": 2.5118245933543295e-05}
{"train_loss": 0.0987161248922348, "global_step": 180898, "epoch": 2032, "lr": 2.5117743079517396e-05}
{"train_loss": 0.10579142719507217, "global_step": 180899, "epoch": 2032, "lr": 2.5117240228836593e-05}
{"train_loss": 0.07467725872993469, "global_step": 180900, "epoch": 2032, "lr": 2.5116737381500998e-05}
{"train_loss": 0.10616157948970795, "global_step": 180901, "epoch": 2032, "lr": 2.511623453751063e-05}
{"train_loss": 0.10665164142847061, "global_step": 180902, "epoch": 2032, "lr": 2.5115731696865595e-05}
{"train_loss": 0.06979107111692429, "global_step": 180903, "epoch": 2032, "lr": 2.5115228859565938e-05}
{"train_loss": 0.09934428334236145, "global_step": 180904, "epoch": 2032, "lr": 2.5114726025611745e-05}
{"train_loss": 0.06059429794549942, "global_step": 180905, "epoch": 2032, "lr": 2.511422319500306e-05}
{"train_loss": 0.09300822019577026, "global_step": 180906, "epoch": 2032, "lr": 2.5113720367739985e-05}
{"train_loss": 0.0757114440202713, "global_step": 180907, "epoch": 2032, "lr": 2.5113217543822544e-05}
{"train_loss": 0.08430178463459015, "global_step": 180908, "epoch": 2032, "lr": 2.511271472325085e-05}
{"train_loss": 0.11524622142314911, "global_step": 180909, "epoch": 2032, "lr": 2.511221190602493e-05}
{"train_loss": 0.0968085378408432, "global_step": 180910, "epoch": 2032, "lr": 2.5111709092144886e-05}
{"train_loss": 0.09040937572717667, "global_step": 180911, "epoch": 2032, "lr": 2.511120628161075e-05}
{"train_loss": 0.053020209074020386, "global_step": 180912, "epoch": 2032, "lr": 2.5110703474422627e-05}
{"train_loss": 0.07295919209718704, "global_step": 180913, "epoch": 2032, "lr": 2.5110200670580554e-05}
{"train_loss": 0.04263017699122429, "global_step": 180914, "epoch": 2032, "lr": 2.5109697870084627e-05}
{"train_loss": 0.07198767364025116, "global_step": 180915, "epoch": 2032, "lr": 2.5109195072934873e-05}
{"train_loss": 0.12213288992643356, "global_step": 180916, "epoch": 2032, "lr": 2.5108692279131384e-05}
{"train_loss": 0.017639677971601486, "global_step": 180917, "epoch": 2032, "lr": 2.5108189488674254e-05}
{"train_loss": 0.08745163679122925, "global_step": 180918, "epoch": 2032, "lr": 2.5107686701563504e-05}
{"train_loss": 0.09721167385578156, "global_step": 180919, "epoch": 2032, "lr": 2.510718391779924e-05}
{"train_loss": 0.060121115297079086, "global_step": 180920, "epoch": 2032, "lr": 2.5106681137381493e-05}
{"train_loss": 0.07781960815191269, "global_step": 180921, "epoch": 2032, "lr": 2.5106178360310367e-05}
{"train_loss": 0.059676021337509155, "global_step": 180922, "epoch": 2032, "lr": 2.51056755865859e-05}
{"train_loss": 0.08996552228927612, "global_step": 180923, "epoch": 2032, "lr": 2.5105172816208176e-05}
{"train_loss": 0.09295815229415894, "global_step": 180924, "epoch": 2032, "lr": 2.5104670049177247e-05}
{"train_loss": 0.0570341981947422, "global_step": 180925, "epoch": 2032, "lr": 2.510416728549322e-05}
{"train_loss": 0.14614391326904297, "global_step": 180926, "epoch": 2032, "lr": 2.51036645251561e-05}
{"train_loss": 0.09315531700849533, "global_step": 180927, "epoch": 2032, "lr": 2.510316176816602e-05}
{"train_loss": 0.0869063213467598, "global_step": 180928, "epoch": 2032, "lr": 2.510265901452299e-05}
{"train_loss": 0.11675351113080978, "global_step": 180929, "epoch": 2032, "lr": 2.510215626422712e-05}
{"train_loss": 0.07510755211114883, "global_step": 180930, "epoch": 2032, "lr": 2.510165351727845e-05}
{"train_loss": 0.041625604033470154, "global_step": 180931, "epoch": 2032, "lr": 2.5101150773677074e-05}
{"train_loss": 0.10460231453180313, "global_step": 180932, "epoch": 2032, "lr": 2.510064803342303e-05}
{"train_loss": 0.1024436503648758, "global_step": 180933, "epoch": 2032, "lr": 2.510014529651642e-05}
{"train_loss": 0.09666374325752258, "global_step": 180934, "epoch": 2032, "lr": 2.5099642562957264e-05}
{"train_loss": 0.12002413719892502, "global_step": 180935, "epoch": 2032, "lr": 2.509913983274569e-05}
{"train_loss": 0.08247385110203805, "global_step": 180936, "epoch": 2032, "lr": 2.5098637105881716e-05, "val_loss": 7.575572490692139}
{"train_loss": 0.12926526367664337, "global_step": 180937, "epoch": 2033, "lr": 2.5098134382365423e-05}
{"train_loss": 0.050011299550533295, "global_step": 180938, "epoch": 2033, "lr": 2.5097631662196897e-05}
{"train_loss": 0.0632253959774971, "global_step": 180939, "epoch": 2033, "lr": 2.509712894537617e-05}
{"train_loss": 0.0593348853290081, "global_step": 180940, "epoch": 2033, "lr": 2.5096626231903347e-05}
{"train_loss": 0.09728679060935974, "global_step": 180941, "epoch": 2033, "lr": 2.5096123521778463e-05}
{"train_loss": 0.06741106510162354, "global_step": 180942, "epoch": 2033, "lr": 2.509562081500162e-05}
{"train_loss": 0.06972569972276688, "global_step": 180943, "epoch": 2033, "lr": 2.5095118111572852e-05}
{"train_loss": 0.04972553253173828, "global_step": 180944, "epoch": 2033, "lr": 2.5094615411492254e-05}
{"train_loss": 0.08725802600383759, "global_step": 180945, "epoch": 2033, "lr": 2.5094112714759866e-05}
{"train_loss": 0.04520858824253082, "global_step": 180946, "epoch": 2033, "lr": 2.509361002137579e-05}
{"train_loss": 0.06592495739459991, "global_step": 180947, "epoch": 2033, "lr": 2.509310733134006e-05}
{"train_loss": 0.08364640176296234, "global_step": 180948, "epoch": 2033, "lr": 2.5092604644652773e-05}
{"train_loss": 0.09117531776428223, "global_step": 180949, "epoch": 2033, "lr": 2.5092101961313964e-05}
{"train_loss": 0.05195554345846176, "global_step": 180950, "epoch": 2033, "lr": 2.509159928132373e-05}
{"train_loss": 0.10736154764890671, "global_step": 180951, "epoch": 2033, "lr": 2.5091096604682117e-05}
{"train_loss": 0.10058706998825073, "global_step": 180952, "epoch": 2033, "lr": 2.50905939313892e-05}
{"train_loss": 0.08280249685049057, "global_step": 180953, "epoch": 2033, "lr": 2.509009126144507e-05}
{"train_loss": 0.09374108910560608, "global_step": 180954, "epoch": 2033, "lr": 2.508958859484975e-05}
{"train_loss": 0.05885612964630127, "global_step": 180955, "epoch": 2033, "lr": 2.508908593160335e-05}
{"train_loss": 0.05992762744426727, "global_step": 180956, "epoch": 2033, "lr": 2.5088583271705902e-05}
{"train_loss": 0.09834805876016617, "global_step": 180957, "epoch": 2033, "lr": 2.5088080615157512e-05}
{"train_loss": 0.07479146122932434, "global_step": 180958, "epoch": 2033, "lr": 2.5087577961958202e-05}
{"train_loss": 0.0958368331193924, "global_step": 180959, "epoch": 2033, "lr": 2.5087075312108087e-05}
{"train_loss": 0.09374170005321503, "global_step": 180960, "epoch": 2033, "lr": 2.508657266560719e-05}
{"train_loss": 0.04203448444604874, "global_step": 180961, "epoch": 2033, "lr": 2.508607002245562e-05}
{"train_loss": 0.04889386147260666, "global_step": 180962, "epoch": 2033, "lr": 2.50855673826534e-05}
{"train_loss": 0.05689125880599022, "global_step": 180963, "epoch": 2033, "lr": 2.5085064746200647e-05}
{"train_loss": 0.03884253650903702, "global_step": 180964, "epoch": 2033, "lr": 2.5084562113097377e-05}
{"train_loss": 0.06380421668291092, "global_step": 180965, "epoch": 2033, "lr": 2.5084059483343712e-05}
{"train_loss": 0.07320461422204971, "global_step": 180966, "epoch": 2033, "lr": 2.5083556856939673e-05}
{"train_loss": 0.08587346225976944, "global_step": 180967, "epoch": 2033, "lr": 2.5083054233885354e-05}
{"train_loss": 0.06392602622509003, "global_step": 180968, "epoch": 2033, "lr": 2.5082551614180804e-05}
{"train_loss": 0.04532307758927345, "global_step": 180969, "epoch": 2033, "lr": 2.508204899782612e-05}
{"train_loss": 0.07160308212041855, "global_step": 180970, "epoch": 2033, "lr": 2.508154638482133e-05}
{"train_loss": 0.056534770876169205, "global_step": 180971, "epoch": 2033, "lr": 2.5081043775166542e-05}
{"train_loss": 0.10084494203329086, "global_step": 180972, "epoch": 2033, "lr": 2.508054116886178e-05}
{"train_loss": 0.03486056253314018, "global_step": 180973, "epoch": 2033, "lr": 2.5080038565907162e-05}
{"train_loss": 0.10592103004455566, "global_step": 180974, "epoch": 2033, "lr": 2.5079535966302703e-05}
{"train_loss": 0.06590293347835541, "global_step": 180975, "epoch": 2033, "lr": 2.507903337004852e-05}
{"train_loss": 0.10524050146341324, "global_step": 180976, "epoch": 2033, "lr": 2.507853077714466e-05}
{"train_loss": 0.06611667573451996, "global_step": 180977, "epoch": 2033, "lr": 2.5078028187591163e-05}
{"train_loss": 0.07330457866191864, "global_step": 180978, "epoch": 2033, "lr": 2.5077525601388148e-05}
{"train_loss": 0.08387921750545502, "global_step": 180979, "epoch": 2033, "lr": 2.507702301853563e-05}
{"train_loss": 0.049162786453962326, "global_step": 180980, "epoch": 2033, "lr": 2.5076520439033725e-05}
{"train_loss": 0.094106025993824, "global_step": 180981, "epoch": 2033, "lr": 2.5076017862882457e-05}
{"train_loss": 0.05840997397899628, "global_step": 180982, "epoch": 2033, "lr": 2.507551529008193e-05}
{"train_loss": 0.0341574028134346, "global_step": 180983, "epoch": 2033, "lr": 2.5075012720632185e-05}
{"train_loss": 0.08682902902364731, "global_step": 180984, "epoch": 2033, "lr": 2.5074510154533314e-05}
{"train_loss": 0.09418398886919022, "global_step": 180985, "epoch": 2033, "lr": 2.5074007591785352e-05}
{"train_loss": 0.030815288424491882, "global_step": 180986, "epoch": 2033, "lr": 2.5073505032388412e-05}
{"train_loss": 0.07275249063968658, "global_step": 180987, "epoch": 2033, "lr": 2.507300247634251e-05}
{"train_loss": 0.13334298133850098, "global_step": 180988, "epoch": 2033, "lr": 2.507249992364774e-05}
{"train_loss": 0.09256454557180405, "global_step": 180989, "epoch": 2033, "lr": 2.507199737430419e-05}
{"train_loss": 0.06862318515777588, "global_step": 180990, "epoch": 2033, "lr": 2.507149482831188e-05}
{"train_loss": 0.08323034644126892, "global_step": 180991, "epoch": 2033, "lr": 2.5070992285670934e-05}
{"train_loss": 0.10630875080823898, "global_step": 180992, "epoch": 2033, "lr": 2.5070489746381364e-05}
{"train_loss": 0.08214358985424042, "global_step": 180993, "epoch": 2033, "lr": 2.506998721044328e-05}
{"train_loss": 0.04321737587451935, "global_step": 180994, "epoch": 2033, "lr": 2.5069484677856714e-05}
{"train_loss": 0.08493007719516754, "global_step": 180995, "epoch": 2033, "lr": 2.5068982148621777e-05}
{"train_loss": 0.06837891042232513, "global_step": 180996, "epoch": 2033, "lr": 2.5068479622738482e-05}
{"train_loss": 0.09475668519735336, "global_step": 180997, "epoch": 2033, "lr": 2.506797710020695e-05}
{"train_loss": 0.06839336454868317, "global_step": 180998, "epoch": 2033, "lr": 2.5067474581027213e-05}
{"train_loss": 0.07738850265741348, "global_step": 180999, "epoch": 2033, "lr": 2.506697206519936e-05}
{"train_loss": 0.08957380056381226, "global_step": 181000, "epoch": 2033, "lr": 2.5066469552723438e-05}
{"train_loss": 0.06308320164680481, "global_step": 181001, "epoch": 2033, "lr": 2.5065967043599537e-05}
{"train_loss": 0.08507784456014633, "global_step": 181002, "epoch": 2033, "lr": 2.5065464537827694e-05}
{"train_loss": 0.12111987173557281, "global_step": 181003, "epoch": 2033, "lr": 2.5064962035408014e-05}
{"train_loss": 0.07443511486053467, "global_step": 181004, "epoch": 2033, "lr": 2.506445953634053e-05}
{"train_loss": 0.04017622768878937, "global_step": 181005, "epoch": 2033, "lr": 2.506395704062534e-05}
{"train_loss": 0.06613558530807495, "global_step": 181006, "epoch": 2033, "lr": 2.5063454548262488e-05}
{"train_loss": 0.04225290194153786, "global_step": 181007, "epoch": 2033, "lr": 2.5062952059252044e-05}
{"train_loss": 0.10050101578235626, "global_step": 181008, "epoch": 2033, "lr": 2.50624495735941e-05}
{"train_loss": 0.04672650620341301, "global_step": 181009, "epoch": 2033, "lr": 2.5061947091288708e-05}
{"train_loss": 0.10485945641994476, "global_step": 181010, "epoch": 2033, "lr": 2.506144461233591e-05}
{"train_loss": 0.07766009867191315, "global_step": 181011, "epoch": 2033, "lr": 2.506094213673582e-05}
{"train_loss": 0.10201700776815414, "global_step": 181012, "epoch": 2033, "lr": 2.506043966448846e-05}
{"train_loss": 0.07800235599279404, "global_step": 181013, "epoch": 2033, "lr": 2.505993719559394e-05}
{"train_loss": 0.07833731174468994, "global_step": 181014, "epoch": 2033, "lr": 2.5059434730052312e-05}
{"train_loss": 0.08575807511806488, "global_step": 181015, "epoch": 2033, "lr": 2.5058932267863615e-05}
{"train_loss": 0.10268670320510864, "global_step": 181016, "epoch": 2033, "lr": 2.505842980902796e-05}
{"train_loss": 0.04893156513571739, "global_step": 181017, "epoch": 2033, "lr": 2.505792735354538e-05}
{"train_loss": 0.07072771340608597, "global_step": 181018, "epoch": 2033, "lr": 2.505742490141597e-05}
{"train_loss": 0.044496700167655945, "global_step": 181019, "epoch": 2033, "lr": 2.5056922452639774e-05}
{"train_loss": 0.046420853585004807, "global_step": 181020, "epoch": 2033, "lr": 2.5056420007216884e-05}
{"train_loss": 0.07954058796167374, "global_step": 181021, "epoch": 2033, "lr": 2.5055917565147336e-05}
{"train_loss": 0.05854224041104317, "global_step": 181022, "epoch": 2033, "lr": 2.5055415126431236e-05}
{"train_loss": 0.09498744457960129, "global_step": 181023, "epoch": 2033, "lr": 2.5054912691068617e-05}
{"train_loss": 0.0724867731332779, "global_step": 181024, "epoch": 2033, "lr": 2.5054410259059558e-05}
{"train_loss": 0.07419638323147645, "global_step": 181025, "epoch": 2033, "lr": 2.5053907830404145e-05, "val_loss": 7.413669109344482}
{"train_loss": 0.13464824855327606, "global_step": 181026, "epoch": 2034, "lr": 2.5053405405102414e-05}
{"train_loss": 0.09287051111459732, "global_step": 181027, "epoch": 2034, "lr": 2.505290298315447e-05}
{"train_loss": 0.07682370394468307, "global_step": 181028, "epoch": 2034, "lr": 2.505240056456033e-05}
{"train_loss": 0.07856295257806778, "global_step": 181029, "epoch": 2034, "lr": 2.505189814932012e-05}
{"train_loss": 0.06791801750659943, "global_step": 181030, "epoch": 2034, "lr": 2.505139573743386e-05}
{"train_loss": 0.05761304497718811, "global_step": 181031, "epoch": 2034, "lr": 2.505089332890165e-05}
{"train_loss": 0.06366387009620667, "global_step": 181032, "epoch": 2034, "lr": 2.505039092372352e-05}
{"train_loss": 0.1075485348701477, "global_step": 181033, "epoch": 2034, "lr": 2.5049888521899584e-05}
{"train_loss": 0.11175547540187836, "global_step": 181034, "epoch": 2034, "lr": 2.5049386123429868e-05}
{"train_loss": 0.11172805726528168, "global_step": 181035, "epoch": 2034, "lr": 2.5048883728314476e-05}
{"train_loss": 0.060014013200998306, "global_step": 181036, "epoch": 2034, "lr": 2.5048381336553438e-05}
{"train_loss": 0.06182970479130745, "global_step": 181037, "epoch": 2034, "lr": 2.5047878948146863e-05}
{"train_loss": 0.05553652346134186, "global_step": 181038, "epoch": 2034, "lr": 2.5047376563094775e-05}
{"train_loss": 0.12956485152244568, "global_step": 181039, "epoch": 2034, "lr": 2.5046874181397284e-05}
{"train_loss": 0.08816523104906082, "global_step": 181040, "epoch": 2034, "lr": 2.5046371803054412e-05}
{"train_loss": 0.06040497124195099, "global_step": 181041, "epoch": 2034, "lr": 2.5045869428066283e-05}
{"train_loss": 0.08290289342403412, "global_step": 181042, "epoch": 2034, "lr": 2.50453670564329e-05}
{"train_loss": 0.046711623668670654, "global_step": 181043, "epoch": 2034, "lr": 2.5044864688154367e-05}
{"train_loss": 0.16287556290626526, "global_step": 181044, "epoch": 2034, "lr": 2.5044362323230774e-05}
{"train_loss": 0.04425524175167084, "global_step": 181045, "epoch": 2034, "lr": 2.5043859961662137e-05}
{"train_loss": 0.024761436507105827, "global_step": 181046, "epoch": 2034, "lr": 2.504335760344857e-05}
{"train_loss": 0.072594553232193, "global_step": 181047, "epoch": 2034, "lr": 2.5042855248590113e-05}
{"train_loss": 0.06961940228939056, "global_step": 181048, "epoch": 2034, "lr": 2.504235289708683e-05}
{"train_loss": 0.12524491548538208, "global_step": 181049, "epoch": 2034, "lr": 2.5041850548938815e-05}
{"train_loss": 0.0660846009850502, "global_step": 181050, "epoch": 2034, "lr": 2.50413482041461e-05}
{"train_loss": 0.09660852700471878, "global_step": 181051, "epoch": 2034, "lr": 2.5040845862708785e-05}
{"train_loss": 0.07912248373031616, "global_step": 181052, "epoch": 2034, "lr": 2.5040343524626907e-05}
{"train_loss": 0.09693574905395508, "global_step": 181053, "epoch": 2034, "lr": 2.5039841189900575e-05}
{"train_loss": 0.10915985703468323, "global_step": 181054, "epoch": 2034, "lr": 2.5039338858529827e-05}
{"train_loss": 0.10936112701892853, "global_step": 181055, "epoch": 2034, "lr": 2.503883653051472e-05}
{"train_loss": 0.04982921853661537, "global_step": 181056, "epoch": 2034, "lr": 2.5038334205855352e-05}
{"train_loss": 0.09016116708517075, "global_step": 181057, "epoch": 2034, "lr": 2.5037831884551764e-05}
{"train_loss": 0.030631357803940773, "global_step": 181058, "epoch": 2034, "lr": 2.503732956660405e-05}
{"train_loss": 0.03249774128198624, "global_step": 181059, "epoch": 2034, "lr": 2.5036827252012242e-05}
{"train_loss": 0.06967342644929886, "global_step": 181060, "epoch": 2034, "lr": 2.503632494077643e-05}
{"train_loss": 0.05412651225924492, "global_step": 181061, "epoch": 2034, "lr": 2.5035822632896698e-05}
{"train_loss": 0.08034338057041168, "global_step": 181062, "epoch": 2034, "lr": 2.503532032837308e-05}
{"train_loss": 0.0788135826587677, "global_step": 181063, "epoch": 2034, "lr": 2.5034818027205674e-05}
{"train_loss": 0.05720634385943413, "global_step": 181064, "epoch": 2034, "lr": 2.5034315729394514e-05}
{"train_loss": 0.04331611469388008, "global_step": 181065, "epoch": 2034, "lr": 2.5033813434939702e-05}
{"train_loss": 0.07690293341875076, "global_step": 181066, "epoch": 2034, "lr": 2.5033311143841275e-05}
{"train_loss": 0.105250284075737, "global_step": 181067, "epoch": 2034, "lr": 2.503280885609933e-05}
{"train_loss": 0.07991103827953339, "global_step": 181068, "epoch": 2034, "lr": 2.5032306571713904e-05}
{"train_loss": 0.06728551536798477, "global_step": 181069, "epoch": 2034, "lr": 2.5031804290685103e-05}
{"train_loss": 0.04660864174365997, "global_step": 181070, "epoch": 2034, "lr": 2.5031302013012948e-05}
{"train_loss": 0.10595612227916718, "global_step": 181071, "epoch": 2034, "lr": 2.5030799738697547e-05}
{"train_loss": 0.06389990448951721, "global_step": 181072, "epoch": 2034, "lr": 2.5030297467738934e-05}
{"train_loss": 0.07667019218206406, "global_step": 181073, "epoch": 2034, "lr": 2.5029795200137214e-05}
{"train_loss": 0.07210249453783035, "global_step": 181074, "epoch": 2034, "lr": 2.502929293589241e-05}
{"train_loss": 0.09815064817667007, "global_step": 181075, "epoch": 2034, "lr": 2.5028790675004638e-05}
{"train_loss": 0.09976060688495636, "global_step": 181076, "epoch": 2034, "lr": 2.5028288417473916e-05}
{"train_loss": 0.04721277579665184, "global_step": 181077, "epoch": 2034, "lr": 2.502778616330036e-05}
{"train_loss": 0.09374688565731049, "global_step": 181078, "epoch": 2034, "lr": 2.502728391248399e-05}
{"train_loss": 0.04859742894768715, "global_step": 181079, "epoch": 2034, "lr": 2.50267816650249e-05}
{"train_loss": 0.03338014334440231, "global_step": 181080, "epoch": 2034, "lr": 2.502627942092317e-05}
{"train_loss": 0.060478903353214264, "global_step": 181081, "epoch": 2034, "lr": 2.5025777180178843e-05}
{"train_loss": 0.08916281163692474, "global_step": 181082, "epoch": 2034, "lr": 2.502527494279201e-05}
{"train_loss": 0.09698700904846191, "global_step": 181083, "epoch": 2034, "lr": 2.50247727087627e-05}
{"train_loss": 0.07831921428442001, "global_step": 181084, "epoch": 2034, "lr": 2.5024270478091028e-05}
{"train_loss": 0.04792255535721779, "global_step": 181085, "epoch": 2034, "lr": 2.5023768250777025e-05}
{"train_loss": 0.08867637813091278, "global_step": 181086, "epoch": 2034, "lr": 2.5023266026820778e-05}
{"train_loss": 0.09495024383068085, "global_step": 181087, "epoch": 2034, "lr": 2.502276380622236e-05}
{"train_loss": 0.05344349145889282, "global_step": 181088, "epoch": 2034, "lr": 2.5022261588981804e-05}
{"train_loss": 0.09051103889942169, "global_step": 181089, "epoch": 2034, "lr": 2.502175937509922e-05}
{"train_loss": 0.08793061226606369, "global_step": 181090, "epoch": 2034, "lr": 2.5021257164574642e-05}
{"train_loss": 0.04116383567452431, "global_step": 181091, "epoch": 2034, "lr": 2.502075495740816e-05}
{"train_loss": 0.10185661166906357, "global_step": 181092, "epoch": 2034, "lr": 2.5020252753599837e-05}
{"train_loss": 0.1099766343832016, "global_step": 181093, "epoch": 2034, "lr": 2.5019750553149723e-05}
{"train_loss": 0.11615914106369019, "global_step": 181094, "epoch": 2034, "lr": 2.5019248356057913e-05}
{"train_loss": 0.056447386741638184, "global_step": 181095, "epoch": 2034, "lr": 2.5018746162324443e-05}
{"train_loss": 0.03320414945483208, "global_step": 181096, "epoch": 2034, "lr": 2.5018243971949417e-05}
{"train_loss": 0.07750522345304489, "global_step": 181097, "epoch": 2034, "lr": 2.5017741784932857e-05}
{"train_loss": 0.033005982637405396, "global_step": 181098, "epoch": 2034, "lr": 2.5017239601274867e-05}
{"train_loss": 0.08325217664241791, "global_step": 181099, "epoch": 2034, "lr": 2.5016737420975522e-05}
{"train_loss": 0.1179429441690445, "global_step": 181100, "epoch": 2034, "lr": 2.5016235244034853e-05}
{"train_loss": 0.05617956072092056, "global_step": 181101, "epoch": 2034, "lr": 2.5015733070452964e-05}
{"train_loss": 0.08480606228113174, "global_step": 181102, "epoch": 2034, "lr": 2.5015230900229885e-05}
{"train_loss": 0.06767790764570236, "global_step": 181103, "epoch": 2034, "lr": 2.5014728733365726e-05}
{"train_loss": 0.08888594061136246, "global_step": 181104, "epoch": 2034, "lr": 2.501422656986051e-05}
{"train_loss": 0.04021500423550606, "global_step": 181105, "epoch": 2034, "lr": 2.501372440971435e-05}
{"train_loss": 0.06114475801587105, "global_step": 181106, "epoch": 2034, "lr": 2.501322225292727e-05}
{"train_loss": 0.041824325919151306, "global_step": 181107, "epoch": 2034, "lr": 2.5012720099499375e-05}
{"train_loss": 0.08177568763494492, "global_step": 181108, "epoch": 2034, "lr": 2.5012217949430693e-05}
{"train_loss": 0.09726566821336746, "global_step": 181109, "epoch": 2034, "lr": 2.5011715802721338e-05}
{"train_loss": 0.06277360022068024, "global_step": 181110, "epoch": 2034, "lr": 2.5011213659371335e-05}
{"train_loss": 0.07129190862178802, "global_step": 181111, "epoch": 2034, "lr": 2.501071151938078e-05}
{"train_loss": 0.052989114075899124, "global_step": 181112, "epoch": 2034, "lr": 2.501020938274972e-05}
{"train_loss": 0.04973715916275978, "global_step": 181113, "epoch": 2034, "lr": 2.5009707249478255e-05}
{"train_loss": 0.076699704475952, "global_step": 181114, "epoch": 2034, "lr": 2.5009205119566403e-05, "val_loss": 7.4611358642578125}
{"train_loss": 0.0975569561123848, "global_step": 181115, "epoch": 2035, "lr": 2.500870299301426e-05}
{"train_loss": 0.06844835728406906, "global_step": 181116, "epoch": 2035, "lr": 2.5008200869821914e-05}
{"train_loss": 0.13557349145412445, "global_step": 181117, "epoch": 2035, "lr": 2.5007698749989388e-05}
{"train_loss": 0.06916528195142746, "global_step": 181118, "epoch": 2035, "lr": 2.5007196633516794e-05}
{"train_loss": 0.07483828067779541, "global_step": 181119, "epoch": 2035, "lr": 2.500669452040416e-05}
{"train_loss": 0.10367165505886078, "global_step": 181120, "epoch": 2035, "lr": 2.5006192410651584e-05}
{"train_loss": 0.09582064300775528, "global_step": 181121, "epoch": 2035, "lr": 2.500569030425911e-05}
{"train_loss": 0.08080512285232544, "global_step": 181122, "epoch": 2035, "lr": 2.5005188201226835e-05}
{"train_loss": 0.1275552213191986, "global_step": 181123, "epoch": 2035, "lr": 2.5004686101554785e-05}
{"train_loss": 0.12042056769132614, "global_step": 181124, "epoch": 2035, "lr": 2.500418400524307e-05}
{"train_loss": 0.14124615490436554, "global_step": 181125, "epoch": 2035, "lr": 2.5003681912291737e-05}
{"train_loss": 0.06301933526992798, "global_step": 181126, "epoch": 2035, "lr": 2.500317982270084e-05}
{"train_loss": 0.0685717761516571, "global_step": 181127, "epoch": 2035, "lr": 2.500267773647047e-05}
{"train_loss": 0.13876457512378693, "global_step": 181128, "epoch": 2035, "lr": 2.5002175653600672e-05}
{"train_loss": 0.0938233733177185, "global_step": 181129, "epoch": 2035, "lr": 2.5001673574091543e-05}
{"train_loss": 0.048469576984643936, "global_step": 181130, "epoch": 2035, "lr": 2.5001171497943142e-05}
{"train_loss": 0.06889704614877701, "global_step": 181131, "epoch": 2035, "lr": 2.500066942515551e-05}
{"train_loss": 0.14711707830429077, "global_step": 181132, "epoch": 2035, "lr": 2.5000167355728743e-05}
{"train_loss": 0.08628590404987335, "global_step": 181133, "epoch": 2035, "lr": 2.4999665289662884e-05}
{"train_loss": 0.0822216272354126, "global_step": 181134, "epoch": 2035, "lr": 2.4999163226958018e-05}
{"train_loss": 0.06403092294931412, "global_step": 181135, "epoch": 2035, "lr": 2.4998661167614222e-05}
{"train_loss": 0.10061623901128769, "global_step": 181136, "epoch": 2035, "lr": 2.4998159111631534e-05}
{"train_loss": 0.13916198909282684, "global_step": 181137, "epoch": 2035, "lr": 2.4997657059010067e-05}
{"train_loss": 0.11340518295764923, "global_step": 181138, "epoch": 2035, "lr": 2.4997155009749836e-05}
{"train_loss": 0.07725688815116882, "global_step": 181139, "epoch": 2035, "lr": 2.4996652963850947e-05}
{"train_loss": 0.0675291195511818, "global_step": 181140, "epoch": 2035, "lr": 2.4996150921313437e-05}
{"train_loss": 0.1077108085155487, "global_step": 181141, "epoch": 2035, "lr": 2.499564888213741e-05}
{"train_loss": 0.05062687024474144, "global_step": 181142, "epoch": 2035, "lr": 2.4995146846322893e-05}
{"train_loss": 0.041648589074611664, "global_step": 181143, "epoch": 2035, "lr": 2.4994644813869995e-05}
{"train_loss": 0.07793833315372467, "global_step": 181144, "epoch": 2035, "lr": 2.499414278477874e-05}
{"train_loss": 0.06331251561641693, "global_step": 181145, "epoch": 2035, "lr": 2.4993640759049243e-05}
{"train_loss": 0.05549364537000656, "global_step": 181146, "epoch": 2035, "lr": 2.4993138736681522e-05}
{"train_loss": 0.04639164358377457, "global_step": 181147, "epoch": 2035, "lr": 2.4992636717675682e-05}
{"train_loss": 0.07114069163799286, "global_step": 181148, "epoch": 2035, "lr": 2.4992134702031766e-05}
{"train_loss": 0.09791196882724762, "global_step": 181149, "epoch": 2035, "lr": 2.4991632689749873e-05}
{"train_loss": 0.06737560033798218, "global_step": 181150, "epoch": 2035, "lr": 2.4991130680830027e-05}
{"train_loss": 0.09060973674058914, "global_step": 181151, "epoch": 2035, "lr": 2.4990628675272322e-05}
{"train_loss": 0.09888773411512375, "global_step": 181152, "epoch": 2035, "lr": 2.4990126673076836e-05}
{"train_loss": 0.04750658944249153, "global_step": 181153, "epoch": 2035, "lr": 2.4989624674243607e-05}
{"train_loss": 0.09607149660587311, "global_step": 181154, "epoch": 2035, "lr": 2.4989122678772737e-05}
{"train_loss": 0.07023793458938599, "global_step": 181155, "epoch": 2035, "lr": 2.4988620686664255e-05}
{"train_loss": 0.03255269303917885, "global_step": 181156, "epoch": 2035, "lr": 2.4988118697918267e-05}
{"train_loss": 0.10458025336265564, "global_step": 181157, "epoch": 2035, "lr": 2.4987616712534806e-05}
{"train_loss": 0.0709400400519371, "global_step": 181158, "epoch": 2035, "lr": 2.4987114730513972e-05}
{"train_loss": 0.058188408613204956, "global_step": 181159, "epoch": 2035, "lr": 2.4986612751855794e-05}
{"train_loss": 0.04622165113687515, "global_step": 181160, "epoch": 2035, "lr": 2.4986110776560384e-05}
{"train_loss": 0.047547824680805206, "global_step": 181161, "epoch": 2035, "lr": 2.4985608804627763e-05}
{"train_loss": 0.1352599859237671, "global_step": 181162, "epoch": 2035, "lr": 2.4985106836058042e-05}
{"train_loss": 0.1127161979675293, "global_step": 181163, "epoch": 2035, "lr": 2.4984604870851275e-05}
{"train_loss": 0.06594432145357132, "global_step": 181164, "epoch": 2035, "lr": 2.4984102909007494e-05}
{"train_loss": 0.11822472512722015, "global_step": 181165, "epoch": 2035, "lr": 2.498360095052682e-05}
{"train_loss": 0.0701039656996727, "global_step": 181166, "epoch": 2035, "lr": 2.4983098995409277e-05}
{"train_loss": 0.10152248293161392, "global_step": 181167, "epoch": 2035, "lr": 2.498259704365497e-05}
{"train_loss": 0.10695501416921616, "global_step": 181168, "epoch": 2035, "lr": 2.498209509526393e-05}
{"train_loss": 0.09283441305160522, "global_step": 181169, "epoch": 2035, "lr": 2.498159315023626e-05}
{"train_loss": 0.054378632456064224, "global_step": 181170, "epoch": 2035, "lr": 2.498109120857199e-05}
{"train_loss": 0.09972526133060455, "global_step": 181171, "epoch": 2035, "lr": 2.4980589270271228e-05}
{"train_loss": 0.06910189241170883, "global_step": 181172, "epoch": 2035, "lr": 2.4980087335334e-05}
{"train_loss": 0.06728754937648773, "global_step": 181173, "epoch": 2035, "lr": 2.4979585403760412e-05}
{"train_loss": 0.09640051424503326, "global_step": 181174, "epoch": 2035, "lr": 2.4979083475550503e-05}
{"train_loss": 0.06870397925376892, "global_step": 181175, "epoch": 2035, "lr": 2.4978581550704362e-05}
{"train_loss": 0.06961636245250702, "global_step": 181176, "epoch": 2035, "lr": 2.497807962922203e-05}
{"train_loss": 0.03564409166574478, "global_step": 181177, "epoch": 2035, "lr": 2.4977577711103607e-05}
{"train_loss": 0.07633300125598907, "global_step": 181178, "epoch": 2035, "lr": 2.4977075796349125e-05}
{"train_loss": 0.0540679506957531, "global_step": 181179, "epoch": 2035, "lr": 2.4976573884958692e-05}
{"train_loss": 0.11378316581249237, "global_step": 181180, "epoch": 2035, "lr": 2.4976071976932335e-05}
{"train_loss": 0.0943451002240181, "global_step": 181181, "epoch": 2035, "lr": 2.4975570072270155e-05}
{"train_loss": 0.06847569346427917, "global_step": 181182, "epoch": 2035, "lr": 2.4975068170972187e-05}
{"train_loss": 0.11633996665477753, "global_step": 181183, "epoch": 2035, "lr": 2.4974566273038536e-05}
{"train_loss": 0.11308930069208145, "global_step": 181184, "epoch": 2035, "lr": 2.4974064378469232e-05}
{"train_loss": 0.07228253036737442, "global_step": 181185, "epoch": 2035, "lr": 2.497356248726438e-05}
{"train_loss": 0.11456947028636932, "global_step": 181186, "epoch": 2035, "lr": 2.4973060599424004e-05}
{"train_loss": 0.04769408702850342, "global_step": 181187, "epoch": 2035, "lr": 2.4972558714948195e-05}
{"train_loss": 0.05209760367870331, "global_step": 181188, "epoch": 2035, "lr": 2.4972056833837042e-05}
{"train_loss": 0.06811506301164627, "global_step": 181189, "epoch": 2035, "lr": 2.497155495609057e-05}
{"train_loss": 0.1379571259021759, "global_step": 181190, "epoch": 2035, "lr": 2.4971053081708888e-05}
{"train_loss": 0.10348783433437347, "global_step": 181191, "epoch": 2035, "lr": 2.4970551210692023e-05}
{"train_loss": 0.07932721823453903, "global_step": 181192, "epoch": 2035, "lr": 2.497004934304008e-05}
{"train_loss": 0.035472407937049866, "global_step": 181193, "epoch": 2035, "lr": 2.4969547478753092e-05}
{"train_loss": 0.06205310299992561, "global_step": 181194, "epoch": 2035, "lr": 2.4969045617831165e-05}
{"train_loss": 0.11953802406787872, "global_step": 181195, "epoch": 2035, "lr": 2.496854376027432e-05}
{"train_loss": 0.05159715563058853, "global_step": 181196, "epoch": 2035, "lr": 2.4968041906082674e-05}
{"train_loss": 0.10342832654714584, "global_step": 181197, "epoch": 2035, "lr": 2.4967540055256246e-05}
{"train_loss": 0.07585369050502777, "global_step": 181198, "epoch": 2035, "lr": 2.4967038207795146e-05}
{"train_loss": 0.06952618062496185, "global_step": 181199, "epoch": 2035, "lr": 2.4966536363699404e-05}
{"train_loss": 0.12360818684101105, "global_step": 181200, "epoch": 2035, "lr": 2.4966034522969127e-05}
{"train_loss": 0.06348524987697601, "global_step": 181201, "epoch": 2035, "lr": 2.4965532685604342e-05}
{"train_loss": 0.1356581598520279, "global_step": 181202, "epoch": 2035, "lr": 2.4965030851605154e-05}
{"train_loss": 0.08433103917187519, "global_step": 181203, "epoch": 2035, "lr": 2.4964529020971618e-05, "val_loss": 7.48236083984375, "train_action_mse_error": 9.17979621887207}
{"train_loss": 0.07461951673030853, "global_step": 181204, "epoch": 2036, "lr": 2.496402719370377e-05}
{"train_loss": 0.07637759298086166, "global_step": 181205, "epoch": 2036, "lr": 2.496352536980172e-05}
{"train_loss": 0.053796831518411636, "global_step": 181206, "epoch": 2036, "lr": 2.496302354926551e-05}
{"train_loss": 0.05705622583627701, "global_step": 181207, "epoch": 2036, "lr": 2.496252173209523e-05}
{"train_loss": 0.07622582465410233, "global_step": 181208, "epoch": 2036, "lr": 2.4962019918290913e-05}
{"train_loss": 0.0615575946867466, "global_step": 181209, "epoch": 2036, "lr": 2.4961518107852665e-05}
{"train_loss": 0.11409380286931992, "global_step": 181210, "epoch": 2036, "lr": 2.496101630078052e-05}
{"train_loss": 0.025869395583868027, "global_step": 181211, "epoch": 2036, "lr": 2.4960514497074584e-05}
{"train_loss": 0.11605358123779297, "global_step": 181212, "epoch": 2036, "lr": 2.4960012696734873e-05}
{"train_loss": 0.07861588895320892, "global_step": 181213, "epoch": 2036, "lr": 2.495951089976151e-05}
{"train_loss": 0.03802412003278732, "global_step": 181214, "epoch": 2036, "lr": 2.4959009106154513e-05}
{"train_loss": 0.06859515607357025, "global_step": 181215, "epoch": 2036, "lr": 2.4958507315913993e-05}
{"train_loss": 0.06707970798015594, "global_step": 181216, "epoch": 2036, "lr": 2.4958005529039975e-05}
{"train_loss": 0.08096688985824585, "global_step": 181217, "epoch": 2036, "lr": 2.4957503745532567e-05}
{"train_loss": 0.0720461830496788, "global_step": 181218, "epoch": 2036, "lr": 2.4957001965391803e-05}
{"train_loss": 0.08329618722200394, "global_step": 181219, "epoch": 2036, "lr": 2.4956500188617782e-05}
{"train_loss": 0.08010761439800262, "global_step": 181220, "epoch": 2036, "lr": 2.495599841521053e-05}
{"train_loss": 0.04762006178498268, "global_step": 181221, "epoch": 2036, "lr": 2.4955496645170164e-05}
{"train_loss": 0.08997288346290588, "global_step": 181222, "epoch": 2036, "lr": 2.4954994878496702e-05}
{"train_loss": 0.05473870411515236, "global_step": 181223, "epoch": 2036, "lr": 2.4954493115190257e-05}
{"train_loss": 0.06748844683170319, "global_step": 181224, "epoch": 2036, "lr": 2.495399135525086e-05}
{"train_loss": 0.05128970369696617, "global_step": 181225, "epoch": 2036, "lr": 2.4953489598678592e-05}
{"train_loss": 0.07044993340969086, "global_step": 181226, "epoch": 2036, "lr": 2.4952987845473537e-05}
{"train_loss": 0.09459178894758224, "global_step": 181227, "epoch": 2036, "lr": 2.495248609563573e-05}
{"train_loss": 0.05745136737823486, "global_step": 181228, "epoch": 2036, "lr": 2.4951984349165274e-05}
{"train_loss": 0.05755392462015152, "global_step": 181229, "epoch": 2036, "lr": 2.4951482606062205e-05}
{"train_loss": 0.05270254239439964, "global_step": 181230, "epoch": 2036, "lr": 2.4950980866326617e-05}
{"train_loss": 0.10158692300319672, "global_step": 181231, "epoch": 2036, "lr": 2.4950479129958548e-05}
{"train_loss": 0.06976020336151123, "global_step": 181232, "epoch": 2036, "lr": 2.49499773969581e-05}
{"train_loss": 0.08955105394124985, "global_step": 181233, "epoch": 2036, "lr": 2.4949475667325306e-05}
{"train_loss": 0.06711888313293457, "global_step": 181234, "epoch": 2036, "lr": 2.4948973941060266e-05}
{"train_loss": 0.0787450447678566, "global_step": 181235, "epoch": 2036, "lr": 2.494847221816301e-05}
{"train_loss": 0.08268439769744873, "global_step": 181236, "epoch": 2036, "lr": 2.494797049863365e-05}
{"train_loss": 0.06818103790283203, "global_step": 181237, "epoch": 2036, "lr": 2.4947468782472205e-05}
{"train_loss": 0.07078958302736282, "global_step": 181238, "epoch": 2036, "lr": 2.4946967069678794e-05}
{"train_loss": 0.046080976724624634, "global_step": 181239, "epoch": 2036, "lr": 2.4946465360253436e-05}
{"train_loss": 0.10762859880924225, "global_step": 181240, "epoch": 2036, "lr": 2.4945963654196237e-05}
{"train_loss": 0.06071788817644119, "global_step": 181241, "epoch": 2036, "lr": 2.4945461951507248e-05}
{"train_loss": 0.06757866591215134, "global_step": 181242, "epoch": 2036, "lr": 2.4944960252186516e-05}
{"train_loss": 0.05896621197462082, "global_step": 181243, "epoch": 2036, "lr": 2.4944458556234153e-05}
{"train_loss": 0.07148462533950806, "global_step": 181244, "epoch": 2036, "lr": 2.4943956863650182e-05}
{"train_loss": 0.11573786288499832, "global_step": 181245, "epoch": 2036, "lr": 2.494345517443471e-05}
{"train_loss": 0.05546358972787857, "global_step": 181246, "epoch": 2036, "lr": 2.494295348858776e-05}
{"train_loss": 0.03835580125451088, "global_step": 181247, "epoch": 2036, "lr": 2.494245180610945e-05}
{"train_loss": 0.07920639216899872, "global_step": 181248, "epoch": 2036, "lr": 2.4941950126999803e-05}
{"train_loss": 0.06589598208665848, "global_step": 181249, "epoch": 2036, "lr": 2.494144845125892e-05}
{"train_loss": 0.0812935084104538, "global_step": 181250, "epoch": 2036, "lr": 2.4940946778886835e-05}
{"train_loss": 0.06611675769090652, "global_step": 181251, "epoch": 2036, "lr": 2.494044510988366e-05}
{"train_loss": 0.05279974639415741, "global_step": 181252, "epoch": 2036, "lr": 2.493994344424941e-05}
{"train_loss": 0.03326498344540596, "global_step": 181253, "epoch": 2036, "lr": 2.49394417819842e-05}
{"train_loss": 0.07226185500621796, "global_step": 181254, "epoch": 2036, "lr": 2.4938940123088062e-05}
{"train_loss": 0.08275141566991806, "global_step": 181255, "epoch": 2036, "lr": 2.493843846756109e-05}
{"train_loss": 0.06064256280660629, "global_step": 181256, "epoch": 2036, "lr": 2.4937936815403323e-05}
{"train_loss": 0.07304874062538147, "global_step": 181257, "epoch": 2036, "lr": 2.493743516661487e-05}
{"train_loss": 0.06598032265901566, "global_step": 181258, "epoch": 2036, "lr": 2.493693352119575e-05}
{"train_loss": 0.03989666700363159, "global_step": 181259, "epoch": 2036, "lr": 2.4936431879146072e-05}
{"train_loss": 0.04949074983596802, "global_step": 181260, "epoch": 2036, "lr": 2.4935930240465866e-05}
{"train_loss": 0.05560529977083206, "global_step": 181261, "epoch": 2036, "lr": 2.4935428605155213e-05}
{"train_loss": 0.08072406053543091, "global_step": 181262, "epoch": 2036, "lr": 2.4934926973214217e-05}
{"train_loss": 0.09268258512020111, "global_step": 181263, "epoch": 2036, "lr": 2.4934425344642888e-05}
{"train_loss": 0.12061114609241486, "global_step": 181264, "epoch": 2036, "lr": 2.4933923719441344e-05}
{"train_loss": 0.10210394859313965, "global_step": 181265, "epoch": 2036, "lr": 2.4933422097609605e-05}
{"train_loss": 0.04457690939307213, "global_step": 181266, "epoch": 2036, "lr": 2.4932920479147783e-05}
{"train_loss": 0.04539753124117851, "global_step": 181267, "epoch": 2036, "lr": 2.4932418864055902e-05}
{"train_loss": 0.1100865975022316, "global_step": 181268, "epoch": 2036, "lr": 2.4931917252334074e-05}
{"train_loss": 0.09872762858867645, "global_step": 181269, "epoch": 2036, "lr": 2.4931415643982325e-05}
{"train_loss": 0.08824541419744492, "global_step": 181270, "epoch": 2036, "lr": 2.493091403900076e-05}
{"train_loss": 0.047075435519218445, "global_step": 181271, "epoch": 2036, "lr": 2.493041243738941e-05}
{"train_loss": 0.07270392030477524, "global_step": 181272, "epoch": 2036, "lr": 2.492991083914838e-05}
{"train_loss": 0.06435497105121613, "global_step": 181273, "epoch": 2036, "lr": 2.4929409244277695e-05}
{"train_loss": 0.06440044194459915, "global_step": 181274, "epoch": 2036, "lr": 2.4928907652777468e-05}
{"train_loss": 0.07129695266485214, "global_step": 181275, "epoch": 2036, "lr": 2.4928406064647726e-05}
{"train_loss": 0.13307861983776093, "global_step": 181276, "epoch": 2036, "lr": 2.492790447988857e-05}
{"train_loss": 0.07552195340394974, "global_step": 181277, "epoch": 2036, "lr": 2.492740289850003e-05}
{"train_loss": 0.1124335378408432, "global_step": 181278, "epoch": 2036, "lr": 2.492690132048222e-05}
{"train_loss": 0.05159550532698631, "global_step": 181279, "epoch": 2036, "lr": 2.4926399745835176e-05}
{"train_loss": 0.1355297565460205, "global_step": 181280, "epoch": 2036, "lr": 2.4925898174558958e-05}
{"train_loss": 0.14437048137187958, "global_step": 181281, "epoch": 2036, "lr": 2.4925396606653666e-05}
{"train_loss": 0.07870065420866013, "global_step": 181282, "epoch": 2036, "lr": 2.4924895042119336e-05}
{"train_loss": 0.0837857574224472, "global_step": 181283, "epoch": 2036, "lr": 2.492439348095606e-05}
{"train_loss": 0.01827983930706978, "global_step": 181284, "epoch": 2036, "lr": 2.492389192316387e-05}
{"train_loss": 0.05718710646033287, "global_step": 181285, "epoch": 2036, "lr": 2.4923390368742883e-05}
{"train_loss": 0.04189383238554001, "global_step": 181286, "epoch": 2036, "lr": 2.4922888817693117e-05}
{"train_loss": 0.05919848755002022, "global_step": 181287, "epoch": 2036, "lr": 2.492238727001469e-05}
{"train_loss": 0.12803402543067932, "global_step": 181288, "epoch": 2036, "lr": 2.4921885725707616e-05}
{"train_loss": 0.14480262994766235, "global_step": 181289, "epoch": 2036, "lr": 2.4921384184772012e-05}
{"train_loss": 0.05036415532231331, "global_step": 181290, "epoch": 2036, "lr": 2.4920882647207898e-05}
{"train_loss": 0.10124865174293518, "global_step": 181291, "epoch": 2036, "lr": 2.492038111301539e-05}
{"train_loss": 0.07348779842173786, "global_step": 181292, "epoch": 2036, "lr": 2.491987958219451e-05, "val_loss": 7.580573081970215}
{"train_loss": 0.08570469915866852, "global_step": 181293, "epoch": 2037, "lr": 2.491937805474537e-05}
{"train_loss": 0.15809839963912964, "global_step": 181294, "epoch": 2037, "lr": 2.4918876530667985e-05}
{"train_loss": 0.057810842990875244, "global_step": 181295, "epoch": 2037, "lr": 2.4918375009962473e-05}
{"train_loss": 0.06736789643764496, "global_step": 181296, "epoch": 2037, "lr": 2.4917873492628867e-05}
{"train_loss": 0.09059898555278778, "global_step": 181297, "epoch": 2037, "lr": 2.491737197866724e-05}
{"train_loss": 0.053298499435186386, "global_step": 181298, "epoch": 2037, "lr": 2.4916870468077698e-05}
{"train_loss": 0.04833918809890747, "global_step": 181299, "epoch": 2037, "lr": 2.4916368960860247e-05}
{"train_loss": 0.09889592975378036, "global_step": 181300, "epoch": 2037, "lr": 2.4915867457015007e-05}
{"train_loss": 0.06230272725224495, "global_step": 181301, "epoch": 2037, "lr": 2.4915365956542e-05}
{"train_loss": 0.09127327799797058, "global_step": 181302, "epoch": 2037, "lr": 2.491486445944134e-05}
{"train_loss": 0.08665058761835098, "global_step": 181303, "epoch": 2037, "lr": 2.491436296571305e-05}
{"train_loss": 0.07044664770364761, "global_step": 181304, "epoch": 2037, "lr": 2.4913861475357237e-05}
{"train_loss": 0.0632094070315361, "global_step": 181305, "epoch": 2037, "lr": 2.491335998837393e-05}
{"train_loss": 0.09093809127807617, "global_step": 181306, "epoch": 2037, "lr": 2.4912858504763235e-05}
{"train_loss": 0.11508588492870331, "global_step": 181307, "epoch": 2037, "lr": 2.491235702452518e-05}
{"train_loss": 0.05281178280711174, "global_step": 181308, "epoch": 2037, "lr": 2.4911855547659873e-05}
{"train_loss": 0.06104517728090286, "global_step": 181309, "epoch": 2037, "lr": 2.4911354074167348e-05}
{"train_loss": 0.0678861141204834, "global_step": 181310, "epoch": 2037, "lr": 2.4910852604047697e-05}
{"train_loss": 0.04540728032588959, "global_step": 181311, "epoch": 2037, "lr": 2.4910351137300958e-05}
{"train_loss": 0.1476532518863678, "global_step": 181312, "epoch": 2037, "lr": 2.4909849673927237e-05}
{"train_loss": 0.07717045396566391, "global_step": 181313, "epoch": 2037, "lr": 2.4909348213926563e-05}
{"train_loss": 0.05392802134156227, "global_step": 181314, "epoch": 2037, "lr": 2.4908846757299043e-05}
{"train_loss": 0.06719102710485458, "global_step": 181315, "epoch": 2037, "lr": 2.49083453040447e-05}
{"train_loss": 0.08595508337020874, "global_step": 181316, "epoch": 2037, "lr": 2.4907843854163643e-05}
{"train_loss": 0.04341130703687668, "global_step": 181317, "epoch": 2037, "lr": 2.49073424076559e-05}
{"train_loss": 0.08242331445217133, "global_step": 181318, "epoch": 2037, "lr": 2.4906840964521583e-05}
{"train_loss": 0.03839888423681259, "global_step": 181319, "epoch": 2037, "lr": 2.490633952476073e-05}
{"train_loss": 0.07187557220458984, "global_step": 181320, "epoch": 2037, "lr": 2.4905838088373396e-05}
{"train_loss": 0.035761285573244095, "global_step": 181321, "epoch": 2037, "lr": 2.490533665535969e-05}
{"train_loss": 0.10766351222991943, "global_step": 181322, "epoch": 2037, "lr": 2.4904835225719632e-05}
{"train_loss": 0.1140594482421875, "global_step": 181323, "epoch": 2037, "lr": 2.4904333799453333e-05}
{"train_loss": 0.08466881513595581, "global_step": 181324, "epoch": 2037, "lr": 2.490383237656082e-05}
{"train_loss": 0.0598088875412941, "global_step": 181325, "epoch": 2037, "lr": 2.4903330957042203e-05}
{"train_loss": 0.0996309444308281, "global_step": 181326, "epoch": 2037, "lr": 2.4902829540897504e-05}
{"train_loss": 0.1112440824508667, "global_step": 181327, "epoch": 2037, "lr": 2.4902328128126836e-05}
{"train_loss": 0.05533446744084358, "global_step": 181328, "epoch": 2037, "lr": 2.4901826718730224e-05}
{"train_loss": 0.09427505731582642, "global_step": 181329, "epoch": 2037, "lr": 2.4901325312707773e-05}
{"train_loss": 0.03563600033521652, "global_step": 181330, "epoch": 2037, "lr": 2.4900823910059517e-05}
{"train_loss": 0.06833330541849136, "global_step": 181331, "epoch": 2037, "lr": 2.4900322510785556e-05}
{"train_loss": 0.05705821514129639, "global_step": 181332, "epoch": 2037, "lr": 2.4899821114885924e-05}
{"train_loss": 0.041228726506233215, "global_step": 181333, "epoch": 2037, "lr": 2.4899319722360705e-05}
{"train_loss": 0.057723335921764374, "global_step": 181334, "epoch": 2037, "lr": 2.4898818333209984e-05}
{"train_loss": 0.05022710561752319, "global_step": 181335, "epoch": 2037, "lr": 2.489831694743379e-05}
{"train_loss": 0.0380873829126358, "global_step": 181336, "epoch": 2037, "lr": 2.4897815565032235e-05}
{"train_loss": 0.030762627720832825, "global_step": 181337, "epoch": 2037, "lr": 2.4897314186005343e-05}
{"train_loss": 0.13655467331409454, "global_step": 181338, "epoch": 2037, "lr": 2.4896812810353213e-05}
{"train_loss": 0.05839262530207634, "global_step": 181339, "epoch": 2037, "lr": 2.4896311438075893e-05}
{"train_loss": 0.020721834152936935, "global_step": 181340, "epoch": 2037, "lr": 2.489581006917347e-05}
{"train_loss": 0.05843252316117287, "global_step": 181341, "epoch": 2037, "lr": 2.489530870364598e-05}
{"train_loss": 0.07249005138874054, "global_step": 181342, "epoch": 2037, "lr": 2.489480734149353e-05}
{"train_loss": 0.08999800682067871, "global_step": 181343, "epoch": 2037, "lr": 2.489430598271615e-05}
{"train_loss": 0.05515546351671219, "global_step": 181344, "epoch": 2037, "lr": 2.4893804627313942e-05}
{"train_loss": 0.059029266238212585, "global_step": 181345, "epoch": 2037, "lr": 2.489330327528694e-05}
{"train_loss": 0.07268953323364258, "global_step": 181346, "epoch": 2037, "lr": 2.489280192663524e-05}
{"train_loss": 0.07788383960723877, "global_step": 181347, "epoch": 2037, "lr": 2.489230058135888e-05}
{"train_loss": 0.06928417086601257, "global_step": 181348, "epoch": 2037, "lr": 2.4891799239457964e-05}
{"train_loss": 0.08603063225746155, "global_step": 181349, "epoch": 2037, "lr": 2.489129790093252e-05}
{"train_loss": 0.08426009863615036, "global_step": 181350, "epoch": 2037, "lr": 2.4890796565782655e-05}
{"train_loss": 0.0907074511051178, "global_step": 181351, "epoch": 2037, "lr": 2.4890295234008397e-05}
{"train_loss": 0.04505990073084831, "global_step": 181352, "epoch": 2037, "lr": 2.4889793905609854e-05}
{"train_loss": 0.05505715310573578, "global_step": 181353, "epoch": 2037, "lr": 2.488929258058705e-05}
{"train_loss": 0.032965052872896194, "global_step": 181354, "epoch": 2037, "lr": 2.4888791258940096e-05}
{"train_loss": 0.0645684227347374, "global_step": 181355, "epoch": 2037, "lr": 2.4888289940669018e-05}
{"train_loss": 0.08918019384145737, "global_step": 181356, "epoch": 2037, "lr": 2.488778862577392e-05}
{"train_loss": 0.13044990599155426, "global_step": 181357, "epoch": 2037, "lr": 2.4887287314254854e-05}
{"train_loss": 0.09535306692123413, "global_step": 181358, "epoch": 2037, "lr": 2.4886786006111867e-05}
{"train_loss": 0.05377769470214844, "global_step": 181359, "epoch": 2037, "lr": 2.4886284701345068e-05}
{"train_loss": 0.07165911793708801, "global_step": 181360, "epoch": 2037, "lr": 2.4885783399954475e-05}
{"train_loss": 0.07524771988391876, "global_step": 181361, "epoch": 2037, "lr": 2.488528210194021e-05}
{"train_loss": 0.03405661880970001, "global_step": 181362, "epoch": 2037, "lr": 2.4884780807302292e-05}
{"train_loss": 0.09485705941915512, "global_step": 181363, "epoch": 2037, "lr": 2.4884279516040828e-05}
{"train_loss": 0.05185262858867645, "global_step": 181364, "epoch": 2037, "lr": 2.488377822815584e-05}
{"train_loss": 0.13857266306877136, "global_step": 181365, "epoch": 2037, "lr": 2.488327694364745e-05}
{"train_loss": 0.07665114849805832, "global_step": 181366, "epoch": 2037, "lr": 2.4882775662515673e-05}
{"train_loss": 0.08728905767202377, "global_step": 181367, "epoch": 2037, "lr": 2.4882274384760623e-05}
{"train_loss": 0.08296709507703781, "global_step": 181368, "epoch": 2037, "lr": 2.4881773110382327e-05}
{"train_loss": 0.07431236654520035, "global_step": 181369, "epoch": 2037, "lr": 2.4881271839380864e-05}
{"train_loss": 0.08288391679525375, "global_step": 181370, "epoch": 2037, "lr": 2.4880770571756335e-05}
{"train_loss": 0.049020905047655106, "global_step": 181371, "epoch": 2037, "lr": 2.4880269307508757e-05}
{"train_loss": 0.1263512223958969, "global_step": 181372, "epoch": 2037, "lr": 2.487976804663824e-05}
{"train_loss": 0.08057175576686859, "global_step": 181373, "epoch": 2037, "lr": 2.4879266789144813e-05}
{"train_loss": 0.05404479801654816, "global_step": 181374, "epoch": 2037, "lr": 2.4878765535028582e-05}
{"train_loss": 0.1090863049030304, "global_step": 181375, "epoch": 2037, "lr": 2.487826428428957e-05}
{"train_loss": 0.04696561396121979, "global_step": 181376, "epoch": 2037, "lr": 2.48777630369279e-05}
{"train_loss": 0.07010473310947418, "global_step": 181377, "epoch": 2037, "lr": 2.4877261792943578e-05}
{"train_loss": 0.05490827560424805, "global_step": 181378, "epoch": 2037, "lr": 2.4876760552336727e-05}
{"train_loss": 0.05405588820576668, "global_step": 181379, "epoch": 2037, "lr": 2.4876259315107368e-05}
{"train_loss": 0.1254553496837616, "global_step": 181380, "epoch": 2037, "lr": 2.487575808125561e-05}
{"train_loss": 0.07365684349382862, "global_step": 181381, "epoch": 2037, "lr": 2.487525685078148e-05, "val_loss": 7.432164669036865}
{"train_loss": 0.07190226763486862, "global_step": 181382, "epoch": 2038, "lr": 2.4874755623685085e-05}
{"train_loss": 0.06101129204034805, "global_step": 181383, "epoch": 2038, "lr": 2.4874254399966458e-05}
{"train_loss": 0.09510144591331482, "global_step": 181384, "epoch": 2038, "lr": 2.4873753179625697e-05}
{"train_loss": 0.11069151014089584, "global_step": 181385, "epoch": 2038, "lr": 2.4873251962662835e-05}
{"train_loss": 0.05261247605085373, "global_step": 181386, "epoch": 2038, "lr": 2.487275074907797e-05}
{"train_loss": 0.1493362933397293, "global_step": 181387, "epoch": 2038, "lr": 2.487224953887115e-05}
{"train_loss": 0.07440826296806335, "global_step": 181388, "epoch": 2038, "lr": 2.4871748332042444e-05}
{"train_loss": 0.05496606603264809, "global_step": 181389, "epoch": 2038, "lr": 2.4871247128591946e-05}
{"train_loss": 0.1519497036933899, "global_step": 181390, "epoch": 2038, "lr": 2.48707459285197e-05}
{"train_loss": 0.10310975462198257, "global_step": 181391, "epoch": 2038, "lr": 2.4870244731825758e-05}
{"train_loss": 0.06778864562511444, "global_step": 181392, "epoch": 2038, "lr": 2.4869743538510227e-05}
{"train_loss": 0.12441221624612808, "global_step": 181393, "epoch": 2038, "lr": 2.486924234857313e-05}
{"train_loss": 0.17809849977493286, "global_step": 181394, "epoch": 2038, "lr": 2.4868741162014574e-05}
{"train_loss": 0.065253347158432, "global_step": 181395, "epoch": 2038, "lr": 2.486823997883461e-05}
{"train_loss": 0.08967740088701248, "global_step": 181396, "epoch": 2038, "lr": 2.486773879903329e-05}
{"train_loss": 0.052027758210897446, "global_step": 181397, "epoch": 2038, "lr": 2.4867237622610717e-05}
{"train_loss": 0.0736008808016777, "global_step": 181398, "epoch": 2038, "lr": 2.4866736449566913e-05}
{"train_loss": 0.07353013753890991, "global_step": 181399, "epoch": 2038, "lr": 2.4866235279901995e-05}
{"train_loss": 0.07206511497497559, "global_step": 181400, "epoch": 2038, "lr": 2.486573411361598e-05}
{"train_loss": 0.048036254942417145, "global_step": 181401, "epoch": 2038, "lr": 2.486523295070898e-05}
{"train_loss": 0.16466526687145233, "global_step": 181402, "epoch": 2038, "lr": 2.486473179118103e-05}
{"train_loss": 0.09166606515645981, "global_step": 181403, "epoch": 2038, "lr": 2.4864230635032226e-05}
{"train_loss": 0.08851777017116547, "global_step": 181404, "epoch": 2038, "lr": 2.48637294822626e-05}
{"train_loss": 0.10679498314857483, "global_step": 181405, "epoch": 2038, "lr": 2.486322833287224e-05}
{"train_loss": 0.058273158967494965, "global_step": 181406, "epoch": 2038, "lr": 2.486272718686123e-05}
{"train_loss": 0.05176107585430145, "global_step": 181407, "epoch": 2038, "lr": 2.4862226044229602e-05}
{"train_loss": 0.062370169907808304, "global_step": 181408, "epoch": 2038, "lr": 2.4861724904977463e-05}
{"train_loss": 0.05305781215429306, "global_step": 181409, "epoch": 2038, "lr": 2.4861223769104837e-05}
{"train_loss": 0.07693993300199509, "global_step": 181410, "epoch": 2038, "lr": 2.4860722636611834e-05}
{"train_loss": 0.09457872807979584, "global_step": 181411, "epoch": 2038, "lr": 2.4860221507498476e-05}
{"train_loss": 0.0929916724562645, "global_step": 181412, "epoch": 2038, "lr": 2.4859720381764878e-05}
{"train_loss": 0.07139766216278076, "global_step": 181413, "epoch": 2038, "lr": 2.4859219259411066e-05}
{"train_loss": 0.043432120233774185, "global_step": 181414, "epoch": 2038, "lr": 2.4858718140437143e-05}
{"train_loss": 0.1018567755818367, "global_step": 181415, "epoch": 2038, "lr": 2.485821702484314e-05}
{"train_loss": 0.05101616308093071, "global_step": 181416, "epoch": 2038, "lr": 2.485771591262916e-05}
{"train_loss": 0.09799985587596893, "global_step": 181417, "epoch": 2038, "lr": 2.4857214803795237e-05}
{"train_loss": 0.05564701557159424, "global_step": 181418, "epoch": 2038, "lr": 2.4856713698341478e-05}
{"train_loss": 0.09985212236642838, "global_step": 181419, "epoch": 2038, "lr": 2.4856212596267902e-05}
{"train_loss": 0.08679445087909698, "global_step": 181420, "epoch": 2038, "lr": 2.4855711497574624e-05}
{"train_loss": 0.07348251342773438, "global_step": 181421, "epoch": 2038, "lr": 2.485521040226167e-05}
{"train_loss": 0.10223225504159927, "global_step": 181422, "epoch": 2038, "lr": 2.4854709310329144e-05}
{"train_loss": 0.12979857623577118, "global_step": 181423, "epoch": 2038, "lr": 2.485420822177708e-05}
{"train_loss": 0.07368214428424835, "global_step": 181424, "epoch": 2038, "lr": 2.485370713660556e-05}
{"train_loss": 0.06938084959983826, "global_step": 181425, "epoch": 2038, "lr": 2.485320605481467e-05}
{"train_loss": 0.05261319875717163, "global_step": 181426, "epoch": 2038, "lr": 2.485270497640444e-05}
{"train_loss": 0.10812287032604218, "global_step": 181427, "epoch": 2038, "lr": 2.4852203901374977e-05}
{"train_loss": 0.0442470908164978, "global_step": 181428, "epoch": 2038, "lr": 2.485170282972631e-05}
{"train_loss": 0.07867711782455444, "global_step": 181429, "epoch": 2038, "lr": 2.4851201761458546e-05}
{"train_loss": 0.09863825887441635, "global_step": 181430, "epoch": 2038, "lr": 2.4850700696571733e-05}
{"train_loss": 0.06234218180179596, "global_step": 181431, "epoch": 2038, "lr": 2.4850199635065912e-05}
{"train_loss": 0.0873551070690155, "global_step": 181432, "epoch": 2038, "lr": 2.48496985769412e-05}
{"train_loss": 0.057253461331129074, "global_step": 181433, "epoch": 2038, "lr": 2.4849197522197615e-05}
{"train_loss": 0.046836815774440765, "global_step": 181434, "epoch": 2038, "lr": 2.484869647083527e-05}
{"train_loss": 0.04615890979766846, "global_step": 181435, "epoch": 2038, "lr": 2.484819542285421e-05}
{"train_loss": 0.04633358493447304, "global_step": 181436, "epoch": 2038, "lr": 2.484769437825449e-05}
{"train_loss": 0.03743071109056473, "global_step": 181437, "epoch": 2038, "lr": 2.48471933370362e-05}
{"train_loss": 0.07724618166685104, "global_step": 181438, "epoch": 2038, "lr": 2.4846692299199388e-05}
{"train_loss": 0.10495194792747498, "global_step": 181439, "epoch": 2038, "lr": 2.4846191264744144e-05}
{"train_loss": 0.0960962101817131, "global_step": 181440, "epoch": 2038, "lr": 2.484569023367051e-05}
{"train_loss": 0.08976668119430542, "global_step": 181441, "epoch": 2038, "lr": 2.4845189205978566e-05}
{"train_loss": 0.09667718410491943, "global_step": 181442, "epoch": 2038, "lr": 2.4844688181668398e-05}
{"train_loss": 0.05570800602436066, "global_step": 181443, "epoch": 2038, "lr": 2.4844187160740036e-05}
{"train_loss": 0.07946138828992844, "global_step": 181444, "epoch": 2038, "lr": 2.4843686143193585e-05}
{"train_loss": 0.035834986716508865, "global_step": 181445, "epoch": 2038, "lr": 2.4843185129029073e-05}
{"train_loss": 0.09749424457550049, "global_step": 181446, "epoch": 2038, "lr": 2.4842684118246607e-05}
{"train_loss": 0.06318876147270203, "global_step": 181447, "epoch": 2038, "lr": 2.484218311084622e-05}
{"train_loss": 0.0649583712220192, "global_step": 181448, "epoch": 2038, "lr": 2.4841682106828006e-05}
{"train_loss": 0.06910621374845505, "global_step": 181449, "epoch": 2038, "lr": 2.4841181106192012e-05}
{"train_loss": 0.12261677533388138, "global_step": 181450, "epoch": 2038, "lr": 2.484068010893833e-05}
{"train_loss": 0.06432162970304489, "global_step": 181451, "epoch": 2038, "lr": 2.4840179115066992e-05}
{"train_loss": 0.0818827748298645, "global_step": 181452, "epoch": 2038, "lr": 2.4839678124578104e-05}
{"train_loss": 0.10198754072189331, "global_step": 181453, "epoch": 2038, "lr": 2.4839177137471697e-05}
{"train_loss": 0.0709780752658844, "global_step": 181454, "epoch": 2038, "lr": 2.4838676153747875e-05}
{"train_loss": 0.058871857821941376, "global_step": 181455, "epoch": 2038, "lr": 2.4838175173406662e-05}
{"train_loss": 0.1480722278356552, "global_step": 181456, "epoch": 2038, "lr": 2.4837674196448174e-05}
{"train_loss": 0.0402790904045105, "global_step": 181457, "epoch": 2038, "lr": 2.4837173222872433e-05}
{"train_loss": 0.07488831132650375, "global_step": 181458, "epoch": 2038, "lr": 2.483667225267955e-05}
{"train_loss": 0.10162409394979477, "global_step": 181459, "epoch": 2038, "lr": 2.4836171285869546e-05}
{"train_loss": 0.08366110175848007, "global_step": 181460, "epoch": 2038, "lr": 2.4835670322442517e-05}
{"train_loss": 0.036848828196525574, "global_step": 181461, "epoch": 2038, "lr": 2.483516936239854e-05}
{"train_loss": 0.10937606543302536, "global_step": 181462, "epoch": 2038, "lr": 2.4834668405737656e-05}
{"train_loss": 0.07024437189102173, "global_step": 181463, "epoch": 2038, "lr": 2.4834167452459955e-05}
{"train_loss": 0.10626786947250366, "global_step": 181464, "epoch": 2038, "lr": 2.4833666502565477e-05}
{"train_loss": 0.06626635789871216, "global_step": 181465, "epoch": 2038, "lr": 2.4833165556054327e-05}
{"train_loss": 0.109300896525383, "global_step": 181466, "epoch": 2038, "lr": 2.4832664612926527e-05}
{"train_loss": 0.07506690919399261, "global_step": 181467, "epoch": 2038, "lr": 2.483216367318219e-05}
{"train_loss": 0.051213108003139496, "global_step": 181468, "epoch": 2038, "lr": 2.4831662736821365e-05}
{"train_loss": 0.062003668397665024, "global_step": 181469, "epoch": 2038, "lr": 2.4831161803844093e-05}
{"train_loss": 0.08033461977591676, "global_step": 181470, "epoch": 2038, "lr": 2.4830660874250487e-05, "val_loss": 7.647876262664795}
{"train_loss": 0.05687428265810013, "global_step": 181471, "epoch": 2039, "lr": 2.4830159948040572e-05}
{"train_loss": 0.06574033200740814, "global_step": 181472, "epoch": 2039, "lr": 2.482965902521445e-05}
{"train_loss": 0.10803969204425812, "global_step": 181473, "epoch": 2039, "lr": 2.4829158105772177e-05}
{"train_loss": 0.09442012757062912, "global_step": 181474, "epoch": 2039, "lr": 2.48286571897138e-05}
{"train_loss": 0.07268834859132767, "global_step": 181475, "epoch": 2039, "lr": 2.4828156277039422e-05}
{"train_loss": 0.03525857627391815, "global_step": 181476, "epoch": 2039, "lr": 2.482765536774907e-05}
{"train_loss": 0.109307661652565, "global_step": 181477, "epoch": 2039, "lr": 2.482715446184285e-05}
{"train_loss": 0.101824551820755, "global_step": 181478, "epoch": 2039, "lr": 2.4826653559320794e-05}
{"train_loss": 0.10262072831392288, "global_step": 181479, "epoch": 2039, "lr": 2.482615266018299e-05}
{"train_loss": 0.08817685395479202, "global_step": 181480, "epoch": 2039, "lr": 2.4825651764429526e-05}
{"train_loss": 0.09207148849964142, "global_step": 181481, "epoch": 2039, "lr": 2.482515087206042e-05}
{"train_loss": 0.061220888048410416, "global_step": 181482, "epoch": 2039, "lr": 2.4824649983075787e-05}
{"train_loss": 0.1264224350452423, "global_step": 181483, "epoch": 2039, "lr": 2.4824149097475653e-05}
{"train_loss": 0.07912138849496841, "global_step": 181484, "epoch": 2039, "lr": 2.4823648215260127e-05}
{"train_loss": 0.05415768176317215, "global_step": 181485, "epoch": 2039, "lr": 2.482314733642923e-05}
{"train_loss": 0.06964188814163208, "global_step": 181486, "epoch": 2039, "lr": 2.482264646098308e-05}
{"train_loss": 0.12221762537956238, "global_step": 181487, "epoch": 2039, "lr": 2.4822145588921697e-05}
{"train_loss": 0.05952320620417595, "global_step": 181488, "epoch": 2039, "lr": 2.4821644720245186e-05}
{"train_loss": 0.05942808836698532, "global_step": 181489, "epoch": 2039, "lr": 2.4821143854953582e-05}
{"train_loss": 0.11051126569509506, "global_step": 181490, "epoch": 2039, "lr": 2.482064299304699e-05}
{"train_loss": 0.07469428330659866, "global_step": 181491, "epoch": 2039, "lr": 2.4820142134525432e-05}
{"train_loss": 0.06186017394065857, "global_step": 181492, "epoch": 2039, "lr": 2.4819641279389018e-05}
{"train_loss": 0.052457451820373535, "global_step": 181493, "epoch": 2039, "lr": 2.4819140427637777e-05}
{"train_loss": 0.07879675924777985, "global_step": 181494, "epoch": 2039, "lr": 2.481863957927182e-05}
{"train_loss": 0.05808418244123459, "global_step": 181495, "epoch": 2039, "lr": 2.4818138734291164e-05}
{"train_loss": 0.08761699497699738, "global_step": 181496, "epoch": 2039, "lr": 2.4817637892695906e-05}
{"train_loss": 0.043489206582307816, "global_step": 181497, "epoch": 2039, "lr": 2.4817137054486127e-05}
{"train_loss": 0.05750161409378052, "global_step": 181498, "epoch": 2039, "lr": 2.4816636219661858e-05}
{"train_loss": 0.06445620208978653, "global_step": 181499, "epoch": 2039, "lr": 2.4816135388223206e-05}
{"train_loss": 0.08293936401605606, "global_step": 181500, "epoch": 2039, "lr": 2.48156345601702e-05}
{"train_loss": 0.05496150627732277, "global_step": 181501, "epoch": 2039, "lr": 2.481513373550294e-05}
{"train_loss": 0.023467812687158585, "global_step": 181502, "epoch": 2039, "lr": 2.481463291422146e-05}
{"train_loss": 0.08585923910140991, "global_step": 181503, "epoch": 2039, "lr": 2.4814132096325865e-05}
{"train_loss": 0.08060386031866074, "global_step": 181504, "epoch": 2039, "lr": 2.4813631281816187e-05}
{"train_loss": 0.051623690873384476, "global_step": 181505, "epoch": 2039, "lr": 2.4813130470692526e-05}
{"train_loss": 0.06561174243688583, "global_step": 181506, "epoch": 2039, "lr": 2.4812629662954934e-05}
{"train_loss": 0.07242852449417114, "global_step": 181507, "epoch": 2039, "lr": 2.4812128858603457e-05}
{"train_loss": 0.0636695921421051, "global_step": 181508, "epoch": 2039, "lr": 2.4811628057638198e-05}
{"train_loss": 0.10347390919923782, "global_step": 181509, "epoch": 2039, "lr": 2.481112726005919e-05}
{"train_loss": 0.08088304847478867, "global_step": 181510, "epoch": 2039, "lr": 2.4810626465866543e-05}
{"train_loss": 0.06438162177801132, "global_step": 181511, "epoch": 2039, "lr": 2.481012567506028e-05}
{"train_loss": 0.07334237545728683, "global_step": 181512, "epoch": 2039, "lr": 2.4809624887640502e-05}
{"train_loss": 0.04964244365692139, "global_step": 181513, "epoch": 2039, "lr": 2.4809124103607273e-05}
{"train_loss": 0.06641406565904617, "global_step": 181514, "epoch": 2039, "lr": 2.4808623322960623e-05}
{"train_loss": 0.04925919696688652, "global_step": 181515, "epoch": 2039, "lr": 2.4808122545700652e-05}
{"train_loss": 0.12209728360176086, "global_step": 181516, "epoch": 2039, "lr": 2.480762177182744e-05}
{"train_loss": 0.10693860799074173, "global_step": 181517, "epoch": 2039, "lr": 2.480712100134101e-05}
{"train_loss": 0.04840531572699547, "global_step": 181518, "epoch": 2039, "lr": 2.4806620234241478e-05}
{"train_loss": 0.02745957300066948, "global_step": 181519, "epoch": 2039, "lr": 2.480611947052887e-05}
{"train_loss": 0.09530586749315262, "global_step": 181520, "epoch": 2039, "lr": 2.48056187102033e-05}
{"train_loss": 0.1056458055973053, "global_step": 181521, "epoch": 2039, "lr": 2.4805117953264778e-05}
{"train_loss": 0.07344617694616318, "global_step": 181522, "epoch": 2039, "lr": 2.480461719971342e-05}
{"train_loss": 0.05023477226495743, "global_step": 181523, "epoch": 2039, "lr": 2.4804116449549258e-05}
{"train_loss": 0.06692387908697128, "global_step": 181524, "epoch": 2039, "lr": 2.48036157027724e-05}
{"train_loss": 0.09536052495241165, "global_step": 181525, "epoch": 2039, "lr": 2.480311495938286e-05}
{"train_loss": 0.04636034369468689, "global_step": 181526, "epoch": 2039, "lr": 2.480261421938076e-05}
{"train_loss": 0.049531783908605576, "global_step": 181527, "epoch": 2039, "lr": 2.480211348276612e-05}
{"train_loss": 0.1002730205655098, "global_step": 181528, "epoch": 2039, "lr": 2.4801612749539044e-05}
{"train_loss": 0.06302943080663681, "global_step": 181529, "epoch": 2039, "lr": 2.480111201969957e-05}
{"train_loss": 0.0802065059542656, "global_step": 181530, "epoch": 2039, "lr": 2.4800611293247798e-05}
{"train_loss": 0.03302426263689995, "global_step": 181531, "epoch": 2039, "lr": 2.4800110570183753e-05}
{"train_loss": 0.06957025080919266, "global_step": 181532, "epoch": 2039, "lr": 2.4799609850507534e-05}
{"train_loss": 0.08904298394918442, "global_step": 181533, "epoch": 2039, "lr": 2.479910913421921e-05}
{"train_loss": 0.1569896787405014, "global_step": 181534, "epoch": 2039, "lr": 2.4798608421318825e-05}
{"train_loss": 0.07322037220001221, "global_step": 181535, "epoch": 2039, "lr": 2.4798107711806483e-05}
{"train_loss": 0.0836145430803299, "global_step": 181536, "epoch": 2039, "lr": 2.4797607005682206e-05}
{"train_loss": 0.11619497090578079, "global_step": 181537, "epoch": 2039, "lr": 2.47971063029461e-05}
{"train_loss": 0.06763393431901932, "global_step": 181538, "epoch": 2039, "lr": 2.4796605603598195e-05}
{"train_loss": 0.06318921595811844, "global_step": 181539, "epoch": 2039, "lr": 2.47961049076386e-05}
{"train_loss": 0.05655306205153465, "global_step": 181540, "epoch": 2039, "lr": 2.479560421506735e-05}
{"train_loss": 0.12005780637264252, "global_step": 181541, "epoch": 2039, "lr": 2.4795103525884533e-05}
{"train_loss": 0.08677476644515991, "global_step": 181542, "epoch": 2039, "lr": 2.479460284009019e-05}
{"train_loss": 0.09229421615600586, "global_step": 181543, "epoch": 2039, "lr": 2.479410215768443e-05}
{"train_loss": 0.06931743025779724, "global_step": 181544, "epoch": 2039, "lr": 2.4793601478667272e-05}
{"train_loss": 0.04311146214604378, "global_step": 181545, "epoch": 2039, "lr": 2.4793100803038827e-05}
{"train_loss": 0.04025232046842575, "global_step": 181546, "epoch": 2039, "lr": 2.4792600130799143e-05}
{"train_loss": 0.0508267842233181, "global_step": 181547, "epoch": 2039, "lr": 2.4792099461948266e-05}
{"train_loss": 0.10336963832378387, "global_step": 181548, "epoch": 2039, "lr": 2.4791598796486303e-05}
{"train_loss": 0.07195429503917694, "global_step": 181549, "epoch": 2039, "lr": 2.479109813441329e-05}
{"train_loss": 0.07183339446783066, "global_step": 181550, "epoch": 2039, "lr": 2.4790597475729317e-05}
{"train_loss": 0.08215555548667908, "global_step": 181551, "epoch": 2039, "lr": 2.4790096820434423e-05}
{"train_loss": 0.06476536393165588, "global_step": 181552, "epoch": 2039, "lr": 2.478959616852871e-05}
{"train_loss": 0.0820067822933197, "global_step": 181553, "epoch": 2039, "lr": 2.4789095520012218e-05}
{"train_loss": 0.12174846231937408, "global_step": 181554, "epoch": 2039, "lr": 2.478859487488504e-05}
{"train_loss": 0.05600125715136528, "global_step": 181555, "epoch": 2039, "lr": 2.478809423314721e-05}
{"train_loss": 0.05798454210162163, "global_step": 181556, "epoch": 2039, "lr": 2.478759359479883e-05}
{"train_loss": 0.08143698424100876, "global_step": 181557, "epoch": 2039, "lr": 2.478709295983993e-05}
{"train_loss": 0.0699993297457695, "global_step": 181558, "epoch": 2039, "lr": 2.478659232827062e-05}
{"train_loss": 0.07505647838115692, "global_step": 181559, "epoch": 2039, "lr": 2.4786091700090924e-05, "val_loss": 7.720571041107178}
{"train_loss": 0.035598818212747574, "global_step": 181560, "epoch": 2040, "lr": 2.4785591075300955e-05}
{"train_loss": 0.08814581483602524, "global_step": 181561, "epoch": 2040, "lr": 2.4785090453900735e-05}
{"train_loss": 0.05684150010347366, "global_step": 181562, "epoch": 2040, "lr": 2.478458983589037e-05}
{"train_loss": 0.061244431883096695, "global_step": 181563, "epoch": 2040, "lr": 2.478408922126989e-05}
{"train_loss": 0.09855950623750687, "global_step": 181564, "epoch": 2040, "lr": 2.4783588610039398e-05}
{"train_loss": 0.06157863885164261, "global_step": 181565, "epoch": 2040, "lr": 2.4783088002198935e-05}
{"train_loss": 0.06915751844644547, "global_step": 181566, "epoch": 2040, "lr": 2.478258739774859e-05}
{"train_loss": 0.04949236288666725, "global_step": 181567, "epoch": 2040, "lr": 2.4782086796688402e-05}
{"train_loss": 0.05695808678865433, "global_step": 181568, "epoch": 2040, "lr": 2.4781586199018453e-05}
{"train_loss": 0.05190388485789299, "global_step": 181569, "epoch": 2040, "lr": 2.4781085604738835e-05}
{"train_loss": 0.059340331703424454, "global_step": 181570, "epoch": 2040, "lr": 2.478058501384957e-05}
{"train_loss": 0.08184527605772018, "global_step": 181571, "epoch": 2040, "lr": 2.4780084426350768e-05}
{"train_loss": 0.1712021827697754, "global_step": 181572, "epoch": 2040, "lr": 2.477958384224246e-05}
{"train_loss": 0.1002776101231575, "global_step": 181573, "epoch": 2040, "lr": 2.4779083261524748e-05}
{"train_loss": 0.05246730521321297, "global_step": 181574, "epoch": 2040, "lr": 2.477858268419766e-05}
{"train_loss": 0.11347154527902603, "global_step": 181575, "epoch": 2040, "lr": 2.47780821102613e-05}
{"train_loss": 0.10631400346755981, "global_step": 181576, "epoch": 2040, "lr": 2.4777581539715705e-05}
{"train_loss": 0.0976346880197525, "global_step": 181577, "epoch": 2040, "lr": 2.477708097256098e-05}
{"train_loss": 0.09134067595005035, "global_step": 181578, "epoch": 2040, "lr": 2.477658040879714e-05}
{"train_loss": 0.062276691198349, "global_step": 181579, "epoch": 2040, "lr": 2.4776079848424305e-05}
{"train_loss": 0.0839305967092514, "global_step": 181580, "epoch": 2040, "lr": 2.4775579291442502e-05}
{"train_loss": 0.0687154233455658, "global_step": 181581, "epoch": 2040, "lr": 2.4775078737851833e-05}
{"train_loss": 0.05785037577152252, "global_step": 181582, "epoch": 2040, "lr": 2.4774578187652324e-05}
{"train_loss": 0.05945068597793579, "global_step": 181583, "epoch": 2040, "lr": 2.477407764084409e-05}
{"train_loss": 0.11573004722595215, "global_step": 181584, "epoch": 2040, "lr": 2.477357709742717e-05}
{"train_loss": 0.021667180582880974, "global_step": 181585, "epoch": 2040, "lr": 2.477307655740162e-05}
{"train_loss": 0.058404821902513504, "global_step": 181586, "epoch": 2040, "lr": 2.4772576020767535e-05}
{"train_loss": 0.05061408132314682, "global_step": 181587, "epoch": 2040, "lr": 2.4772075487524956e-05}
{"train_loss": 0.05063104256987572, "global_step": 181588, "epoch": 2040, "lr": 2.477157495767398e-05}
{"train_loss": 0.09663856774568558, "global_step": 181589, "epoch": 2040, "lr": 2.477107443121464e-05}
{"train_loss": 0.08837088942527771, "global_step": 181590, "epoch": 2040, "lr": 2.4770573908147044e-05}
{"train_loss": 0.036079391837120056, "global_step": 181591, "epoch": 2040, "lr": 2.4770073388471215e-05}
{"train_loss": 0.0786295011639595, "global_step": 181592, "epoch": 2040, "lr": 2.4769572872187262e-05}
{"train_loss": 0.06486852467060089, "global_step": 181593, "epoch": 2040, "lr": 2.4769072359295215e-05}
{"train_loss": 0.06740058213472366, "global_step": 181594, "epoch": 2040, "lr": 2.4768571849795175e-05}
{"train_loss": 0.0732497125864029, "global_step": 181595, "epoch": 2040, "lr": 2.476807134368717e-05}
{"train_loss": 0.044641878455877304, "global_step": 181596, "epoch": 2040, "lr": 2.4767570840971316e-05}
{"train_loss": 0.04919520765542984, "global_step": 181597, "epoch": 2040, "lr": 2.476707034164763e-05}
{"train_loss": 0.09430898725986481, "global_step": 181598, "epoch": 2040, "lr": 2.476656984571623e-05}
{"train_loss": 0.04597357660531998, "global_step": 181599, "epoch": 2040, "lr": 2.4766069353177135e-05}
{"train_loss": 0.10933486372232437, "global_step": 181600, "epoch": 2040, "lr": 2.4765568864030454e-05}
{"train_loss": 0.14562049508094788, "global_step": 181601, "epoch": 2040, "lr": 2.4765068378276214e-05}
{"train_loss": 0.06480959802865982, "global_step": 181602, "epoch": 2040, "lr": 2.476456789591452e-05}
{"train_loss": 0.0759454295039177, "global_step": 181603, "epoch": 2040, "lr": 2.47640674169454e-05}
{"train_loss": 0.049751825630664825, "global_step": 181604, "epoch": 2040, "lr": 2.476356694136897e-05}
{"train_loss": 0.08117195963859558, "global_step": 181605, "epoch": 2040, "lr": 2.476306646918525e-05}
{"train_loss": 0.023838181048631668, "global_step": 181606, "epoch": 2040, "lr": 2.4762566000394334e-05}
{"train_loss": 0.09481088817119598, "global_step": 181607, "epoch": 2040, "lr": 2.476206553499629e-05}
{"train_loss": 0.03889359161257744, "global_step": 181608, "epoch": 2040, "lr": 2.4761565072991165e-05}
{"train_loss": 0.1467820256948471, "global_step": 181609, "epoch": 2040, "lr": 2.4761064614379066e-05}
{"train_loss": 0.13176599144935608, "global_step": 181610, "epoch": 2040, "lr": 2.4760564159160004e-05}
{"train_loss": 0.07114087045192719, "global_step": 181611, "epoch": 2040, "lr": 2.4760063707334103e-05}
{"train_loss": 0.02618967927992344, "global_step": 181612, "epoch": 2040, "lr": 2.4759563258901385e-05}
{"train_loss": 0.06227625161409378, "global_step": 181613, "epoch": 2040, "lr": 2.475906281386195e-05}
{"train_loss": 0.07331594079732895, "global_step": 181614, "epoch": 2040, "lr": 2.4758562372215838e-05}
{"train_loss": 0.04485996440052986, "global_step": 181615, "epoch": 2040, "lr": 2.4758061933963145e-05}
{"train_loss": 0.06385812163352966, "global_step": 181616, "epoch": 2040, "lr": 2.4757561499103903e-05}
{"train_loss": 0.14196762442588806, "global_step": 181617, "epoch": 2040, "lr": 2.475706106763822e-05}
{"train_loss": 0.05793080851435661, "global_step": 181618, "epoch": 2040, "lr": 2.475656063956612e-05}
{"train_loss": 0.061915792524814606, "global_step": 181619, "epoch": 2040, "lr": 2.4756060214887718e-05}
{"train_loss": 0.03899410367012024, "global_step": 181620, "epoch": 2040, "lr": 2.4755559793603032e-05}
{"train_loss": 0.08453501015901566, "global_step": 181621, "epoch": 2040, "lr": 2.4755059375712174e-05}
{"train_loss": 0.10465816408395767, "global_step": 181622, "epoch": 2040, "lr": 2.4754558961215198e-05}
{"train_loss": 0.08831793069839478, "global_step": 181623, "epoch": 2040, "lr": 2.4754058550112137e-05}
{"train_loss": 0.057628314942121506, "global_step": 181624, "epoch": 2040, "lr": 2.47535581424031e-05}
{"train_loss": 0.11285630613565445, "global_step": 181625, "epoch": 2040, "lr": 2.4753057738088126e-05}
{"train_loss": 0.07045134902000427, "global_step": 181626, "epoch": 2040, "lr": 2.4752557337167316e-05}
{"train_loss": 0.08794430643320084, "global_step": 181627, "epoch": 2040, "lr": 2.4752056939640695e-05}
{"train_loss": 0.07728340476751328, "global_step": 181628, "epoch": 2040, "lr": 2.475155654550837e-05}
{"train_loss": 0.06524539738893509, "global_step": 181629, "epoch": 2040, "lr": 2.4751056154770376e-05}
{"train_loss": 0.09037008136510849, "global_step": 181630, "epoch": 2040, "lr": 2.4750555767426816e-05}
{"train_loss": 0.07463768869638443, "global_step": 181631, "epoch": 2040, "lr": 2.4750055383477706e-05}
{"train_loss": 0.085558220744133, "global_step": 181632, "epoch": 2040, "lr": 2.474955500292317e-05}
{"train_loss": 0.09663163870573044, "global_step": 181633, "epoch": 2040, "lr": 2.474905462576323e-05}
{"train_loss": 0.0824756920337677, "global_step": 181634, "epoch": 2040, "lr": 2.4748554251997987e-05}
{"train_loss": 0.04685298353433609, "global_step": 181635, "epoch": 2040, "lr": 2.4748053881627476e-05}
{"train_loss": 0.05040820688009262, "global_step": 181636, "epoch": 2040, "lr": 2.47475535146518e-05}
{"train_loss": 0.048358816653490067, "global_step": 181637, "epoch": 2040, "lr": 2.4747053151070993e-05}
{"train_loss": 0.05019116401672363, "global_step": 181638, "epoch": 2040, "lr": 2.4746552790885152e-05}
{"train_loss": 0.06369628757238388, "global_step": 181639, "epoch": 2040, "lr": 2.4746052434094308e-05}
{"train_loss": 0.1098087951540947, "global_step": 181640, "epoch": 2040, "lr": 2.4745552080698568e-05}
{"train_loss": 0.12069829553365707, "global_step": 181641, "epoch": 2040, "lr": 2.4745051730697967e-05}
{"train_loss": 0.03673291951417923, "global_step": 181642, "epoch": 2040, "lr": 2.4744551384092585e-05}
{"train_loss": 0.046636227518320084, "global_step": 181643, "epoch": 2040, "lr": 2.4744051040882504e-05}
{"train_loss": 0.09376338124275208, "global_step": 181644, "epoch": 2040, "lr": 2.4743550701067762e-05}
{"train_loss": 0.034071244299411774, "global_step": 181645, "epoch": 2040, "lr": 2.4743050364648463e-05}
{"train_loss": 0.07709795981645584, "global_step": 181646, "epoch": 2040, "lr": 2.4742550031624635e-05}
{"train_loss": 0.08819765597581863, "global_step": 181647, "epoch": 2040, "lr": 2.474204970199638e-05}
{"train_loss": 0.07457110829902498, "global_step": 181648, "epoch": 2040, "lr": 2.474154937576373e-05, "val_loss": 7.5161356925964355, "train_action_mse_error": 12.194974899291992}
{"train_loss": 0.13725240528583527, "global_step": 181649, "epoch": 2041, "lr": 2.4741049052926786e-05}
{"train_loss": 0.10424744337797165, "global_step": 181650, "epoch": 2041, "lr": 2.4740548733485585e-05}
{"train_loss": 0.07122118026018143, "global_step": 181651, "epoch": 2041, "lr": 2.4740048417440227e-05}
{"train_loss": 0.04431699588894844, "global_step": 181652, "epoch": 2041, "lr": 2.473954810479074e-05}
{"train_loss": 0.06884424388408661, "global_step": 181653, "epoch": 2041, "lr": 2.4739047795537234e-05}
{"train_loss": 0.09176812320947647, "global_step": 181654, "epoch": 2041, "lr": 2.4738547489679737e-05}
{"train_loss": 0.08817414939403534, "global_step": 181655, "epoch": 2041, "lr": 2.4738047187218355e-05}
{"train_loss": 0.09646491706371307, "global_step": 181656, "epoch": 2041, "lr": 2.473754688815311e-05}
{"train_loss": 0.07341054081916809, "global_step": 181657, "epoch": 2041, "lr": 2.473704659248412e-05}
{"train_loss": 0.043190985918045044, "global_step": 181658, "epoch": 2041, "lr": 2.4736546300211406e-05}
{"train_loss": 0.05807707458734512, "global_step": 181659, "epoch": 2041, "lr": 2.4736046011335074e-05}
{"train_loss": 0.04325544834136963, "global_step": 181660, "epoch": 2041, "lr": 2.4735545725855147e-05}
{"train_loss": 0.06260643899440765, "global_step": 181661, "epoch": 2041, "lr": 2.473504544377174e-05}
{"train_loss": 0.04740467667579651, "global_step": 181662, "epoch": 2041, "lr": 2.4734545165084904e-05}
{"train_loss": 0.0956692025065422, "global_step": 181663, "epoch": 2041, "lr": 2.4734044889794678e-05}
{"train_loss": 0.0957670584321022, "global_step": 181664, "epoch": 2041, "lr": 2.4733544617901174e-05}
{"train_loss": 0.059918541461229324, "global_step": 181665, "epoch": 2041, "lr": 2.473304434940441e-05}
{"train_loss": 0.05805181339383125, "global_step": 181666, "epoch": 2041, "lr": 2.4732544084304503e-05}
{"train_loss": 0.0521511472761631, "global_step": 181667, "epoch": 2041, "lr": 2.4732043822601482e-05}
{"train_loss": 0.09661377966403961, "global_step": 181668, "epoch": 2041, "lr": 2.4731543564295446e-05}
{"train_loss": 0.09204617887735367, "global_step": 181669, "epoch": 2041, "lr": 2.4731043309386426e-05}
{"train_loss": 0.06380832195281982, "global_step": 181670, "epoch": 2041, "lr": 2.4730543057874526e-05}
{"train_loss": 0.08280423283576965, "global_step": 181671, "epoch": 2041, "lr": 2.4730042809759784e-05}
{"train_loss": 0.09825218468904495, "global_step": 181672, "epoch": 2041, "lr": 2.4729542565042295e-05}
{"train_loss": 0.04594167694449425, "global_step": 181673, "epoch": 2041, "lr": 2.472904232372209e-05}
{"train_loss": 0.08639658987522125, "global_step": 181674, "epoch": 2041, "lr": 2.4728542085799278e-05}
{"train_loss": 0.07483266294002533, "global_step": 181675, "epoch": 2041, "lr": 2.4728041851273885e-05}
{"train_loss": 0.040768545120954514, "global_step": 181676, "epoch": 2041, "lr": 2.4727541620146023e-05}
{"train_loss": 0.05131451413035393, "global_step": 181677, "epoch": 2041, "lr": 2.4727041392415712e-05}
{"train_loss": 0.06040198355913162, "global_step": 181678, "epoch": 2041, "lr": 2.4726541168083046e-05}
{"train_loss": 0.05215629190206528, "global_step": 181679, "epoch": 2041, "lr": 2.4726040947148104e-05}
{"train_loss": 0.06678729504346848, "global_step": 181680, "epoch": 2041, "lr": 2.472554072961092e-05}
{"train_loss": 0.06603380292654037, "global_step": 181681, "epoch": 2041, "lr": 2.4725040515471598e-05}
{"train_loss": 0.11080124229192734, "global_step": 181682, "epoch": 2041, "lr": 2.4724540304730165e-05}
{"train_loss": 0.03018922731280327, "global_step": 181683, "epoch": 2041, "lr": 2.472404009738673e-05}
{"train_loss": 0.11324179917573929, "global_step": 181684, "epoch": 2041, "lr": 2.472353989344132e-05}
{"train_loss": 0.07528598606586456, "global_step": 181685, "epoch": 2041, "lr": 2.472303969289404e-05}
{"train_loss": 0.07491132616996765, "global_step": 181686, "epoch": 2041, "lr": 2.4722539495744924e-05}
{"train_loss": 0.04648633301258087, "global_step": 181687, "epoch": 2041, "lr": 2.472203930199407e-05}
{"train_loss": 0.06116366386413574, "global_step": 181688, "epoch": 2041, "lr": 2.472153911164151e-05}
{"train_loss": 0.04604954272508621, "global_step": 181689, "epoch": 2041, "lr": 2.4721038924687356e-05}
{"train_loss": 0.04364641010761261, "global_step": 181690, "epoch": 2041, "lr": 2.4720538741131625e-05}
{"train_loss": 0.08361929655075073, "global_step": 181691, "epoch": 2041, "lr": 2.4720038560974435e-05}
{"train_loss": 0.05777854472398758, "global_step": 181692, "epoch": 2041, "lr": 2.47195383842158e-05}
{"train_loss": 0.04907383397221565, "global_step": 181693, "epoch": 2041, "lr": 2.4719038210855837e-05}
{"train_loss": 0.055373188108205795, "global_step": 181694, "epoch": 2041, "lr": 2.471853804089458e-05}
{"train_loss": 0.10958380997180939, "global_step": 181695, "epoch": 2041, "lr": 2.471803787433212e-05}
{"train_loss": 0.0592050775885582, "global_step": 181696, "epoch": 2041, "lr": 2.4717537711168493e-05}
{"train_loss": 0.06702741235494614, "global_step": 181697, "epoch": 2041, "lr": 2.47170375514038e-05}
{"train_loss": 0.04805539920926094, "global_step": 181698, "epoch": 2041, "lr": 2.471653739503808e-05}
{"train_loss": 0.0893126055598259, "global_step": 181699, "epoch": 2041, "lr": 2.4716037242071428e-05}
{"train_loss": 0.07843827456235886, "global_step": 181700, "epoch": 2041, "lr": 2.4715537092503898e-05}
{"train_loss": 0.09996918588876724, "global_step": 181701, "epoch": 2041, "lr": 2.471503694633554e-05}
{"train_loss": 0.09332409501075745, "global_step": 181702, "epoch": 2041, "lr": 2.4714536803566453e-05}
{"train_loss": 0.06581857055425644, "global_step": 181703, "epoch": 2041, "lr": 2.471403666419667e-05}
{"train_loss": 0.11927640438079834, "global_step": 181704, "epoch": 2041, "lr": 2.4713536528226295e-05}
{"train_loss": 0.052343856543302536, "global_step": 181705, "epoch": 2041, "lr": 2.4713036395655358e-05}
{"train_loss": 0.039664480835199356, "global_step": 181706, "epoch": 2041, "lr": 2.4712536266483964e-05}
{"train_loss": 0.108770452439785, "global_step": 181707, "epoch": 2041, "lr": 2.4712036140712147e-05}
{"train_loss": 0.09395559132099152, "global_step": 181708, "epoch": 2041, "lr": 2.471153601834e-05}
{"train_loss": 0.05005628243088722, "global_step": 181709, "epoch": 2041, "lr": 2.4711035899367562e-05}
{"train_loss": 0.07625039666891098, "global_step": 181710, "epoch": 2041, "lr": 2.471053578379494e-05}
{"train_loss": 0.09771318733692169, "global_step": 181711, "epoch": 2041, "lr": 2.4710035671622156e-05}
{"train_loss": 0.04064381122589111, "global_step": 181712, "epoch": 2041, "lr": 2.4709535562849318e-05}
{"train_loss": 0.0697454884648323, "global_step": 181713, "epoch": 2041, "lr": 2.4709035457476453e-05}
{"train_loss": 0.08518190681934357, "global_step": 181714, "epoch": 2041, "lr": 2.470853535550366e-05}
{"train_loss": 0.06964395940303802, "global_step": 181715, "epoch": 2041, "lr": 2.4708035256931006e-05}
{"train_loss": 0.10526766628026962, "global_step": 181716, "epoch": 2041, "lr": 2.4707535161758532e-05}
{"train_loss": 0.09667053073644638, "global_step": 181717, "epoch": 2041, "lr": 2.470703506998634e-05}
{"train_loss": 0.0642656609416008, "global_step": 181718, "epoch": 2041, "lr": 2.470653498161446e-05}
{"train_loss": 0.09772079437971115, "global_step": 181719, "epoch": 2041, "lr": 2.4706034896643e-05}
{"train_loss": 0.058929286897182465, "global_step": 181720, "epoch": 2041, "lr": 2.470553481507198e-05}
{"train_loss": 0.11803598701953888, "global_step": 181721, "epoch": 2041, "lr": 2.4705034736901518e-05}
{"train_loss": 0.08891009539365768, "global_step": 181722, "epoch": 2041, "lr": 2.4704534662131635e-05}
{"train_loss": 0.04332546517252922, "global_step": 181723, "epoch": 2041, "lr": 2.4704034590762437e-05}
{"train_loss": 0.062167804688215256, "global_step": 181724, "epoch": 2041, "lr": 2.4703534522793954e-05}
{"train_loss": 0.0543813481926918, "global_step": 181725, "epoch": 2041, "lr": 2.470303445822629e-05}
{"train_loss": 0.06424962729215622, "global_step": 181726, "epoch": 2041, "lr": 2.470253439705948e-05}
{"train_loss": 0.10789578408002853, "global_step": 181727, "epoch": 2041, "lr": 2.4702034339293624e-05}
{"train_loss": 0.06426147371530533, "global_step": 181728, "epoch": 2041, "lr": 2.4701534284928745e-05}
{"train_loss": 0.10954593867063522, "global_step": 181729, "epoch": 2041, "lr": 2.4701034233964963e-05}
{"train_loss": 0.11625483632087708, "global_step": 181730, "epoch": 2041, "lr": 2.4700534186402297e-05}
{"train_loss": 0.06263213604688644, "global_step": 181731, "epoch": 2041, "lr": 2.4700034142240857e-05}
{"train_loss": 0.05467790737748146, "global_step": 181732, "epoch": 2041, "lr": 2.469953410148067e-05}
{"train_loss": 0.058322738856077194, "global_step": 181733, "epoch": 2041, "lr": 2.469903406412184e-05}
{"train_loss": 0.039738621562719345, "global_step": 181734, "epoch": 2041, "lr": 2.4698534030164395e-05}
{"train_loss": 0.10006308555603027, "global_step": 181735, "epoch": 2041, "lr": 2.4698033999608443e-05}
{"train_loss": 0.04309482499957085, "global_step": 181736, "epoch": 2041, "lr": 2.469753397245402e-05}
{"train_loss": 0.07324699454762962, "global_step": 181737, "epoch": 2041, "lr": 2.4697033948701215e-05, "val_loss": 7.632653713226318}
{"train_loss": 0.0658833384513855, "global_step": 181738, "epoch": 2042, "lr": 2.4696533928350092e-05}
{"train_loss": 0.1572248488664627, "global_step": 181739, "epoch": 2042, "lr": 2.469603391140069e-05}
{"train_loss": 0.07521651685237885, "global_step": 181740, "epoch": 2042, "lr": 2.4695533897853113e-05}
{"train_loss": 0.08742211014032364, "global_step": 181741, "epoch": 2042, "lr": 2.46950338877074e-05}
{"train_loss": 0.15088693797588348, "global_step": 181742, "epoch": 2042, "lr": 2.469453388096365e-05}
{"train_loss": 0.08864959329366684, "global_step": 181743, "epoch": 2042, "lr": 2.4694033877621887e-05}
{"train_loss": 0.059213023632764816, "global_step": 181744, "epoch": 2042, "lr": 2.4693533877682228e-05}
{"train_loss": 0.1311708241701126, "global_step": 181745, "epoch": 2042, "lr": 2.4693033881144695e-05}
{"train_loss": 0.0833258107304573, "global_step": 181746, "epoch": 2042, "lr": 2.469253388800939e-05}
{"train_loss": 0.08152614533901215, "global_step": 181747, "epoch": 2042, "lr": 2.469203389827635e-05}
{"train_loss": 0.047931134700775146, "global_step": 181748, "epoch": 2042, "lr": 2.4691533911945676e-05}
{"train_loss": 0.07818092405796051, "global_step": 181749, "epoch": 2042, "lr": 2.4691033929017393e-05}
{"train_loss": 0.05460500344634056, "global_step": 181750, "epoch": 2042, "lr": 2.46905339494916e-05}
{"train_loss": 0.08327402919530869, "global_step": 181751, "epoch": 2042, "lr": 2.469003397336837e-05}
{"train_loss": 0.022902697324752808, "global_step": 181752, "epoch": 2042, "lr": 2.468953400064774e-05}
{"train_loss": 0.07126867771148682, "global_step": 181753, "epoch": 2042, "lr": 2.468903403132981e-05}
{"train_loss": 0.072892926633358, "global_step": 181754, "epoch": 2042, "lr": 2.4688534065414615e-05}
{"train_loss": 0.13172021508216858, "global_step": 181755, "epoch": 2042, "lr": 2.4688034102902253e-05}
{"train_loss": 0.06554888933897018, "global_step": 181756, "epoch": 2042, "lr": 2.468753414379276e-05}
{"train_loss": 0.11288797855377197, "global_step": 181757, "epoch": 2042, "lr": 2.468703418808624e-05}
{"train_loss": 0.02430856227874756, "global_step": 181758, "epoch": 2042, "lr": 2.468653423578272e-05}
{"train_loss": 0.05322159826755524, "global_step": 181759, "epoch": 2042, "lr": 2.4686034286882304e-05}
{"train_loss": 0.0236093457788229, "global_step": 181760, "epoch": 2042, "lr": 2.4685534341385024e-05}
{"train_loss": 0.0576377771794796, "global_step": 181761, "epoch": 2042, "lr": 2.4685034399290984e-05}
{"train_loss": 0.15453973412513733, "global_step": 181762, "epoch": 2042, "lr": 2.468453446060021e-05}
{"train_loss": 0.13254448771476746, "global_step": 181763, "epoch": 2042, "lr": 2.468403452531282e-05}
{"train_loss": 0.1277901828289032, "global_step": 181764, "epoch": 2042, "lr": 2.468353459342883e-05}
{"train_loss": 0.07789844274520874, "global_step": 181765, "epoch": 2042, "lr": 2.4683034664948345e-05}
{"train_loss": 0.05124788358807564, "global_step": 181766, "epoch": 2042, "lr": 2.4682534739871403e-05}
{"train_loss": 0.10927025973796844, "global_step": 181767, "epoch": 2042, "lr": 2.46820348181981e-05}
{"train_loss": 0.11458319425582886, "global_step": 181768, "epoch": 2042, "lr": 2.4681534899928478e-05}
{"train_loss": 0.17666396498680115, "global_step": 181769, "epoch": 2042, "lr": 2.4681034985062607e-05}
{"train_loss": 0.06351802498102188, "global_step": 181770, "epoch": 2042, "lr": 2.4680535073600584e-05}
{"train_loss": 0.07600738853216171, "global_step": 181771, "epoch": 2042, "lr": 2.4680035165542435e-05}
{"train_loss": 0.059827305376529694, "global_step": 181772, "epoch": 2042, "lr": 2.467953526088827e-05}
{"train_loss": 0.062303923070430756, "global_step": 181773, "epoch": 2042, "lr": 2.4679035359638125e-05}
{"train_loss": 0.09656587988138199, "global_step": 181774, "epoch": 2042, "lr": 2.4678535461792063e-05}
{"train_loss": 0.12376239895820618, "global_step": 181775, "epoch": 2042, "lr": 2.467803556735018e-05}
{"train_loss": 0.1503482311964035, "global_step": 181776, "epoch": 2042, "lr": 2.467753567631251e-05}
{"train_loss": 0.09800097346305847, "global_step": 181777, "epoch": 2042, "lr": 2.4677035788679147e-05}
{"train_loss": 0.09208714216947556, "global_step": 181778, "epoch": 2042, "lr": 2.4676535904450155e-05}
{"train_loss": 0.07242389023303986, "global_step": 181779, "epoch": 2042, "lr": 2.4676036023625576e-05}
{"train_loss": 0.05830804631114006, "global_step": 181780, "epoch": 2042, "lr": 2.4675536146205513e-05}
{"train_loss": 0.0681983083486557, "global_step": 181781, "epoch": 2042, "lr": 2.4675036272189994e-05}
{"train_loss": 0.03246701508760452, "global_step": 181782, "epoch": 2042, "lr": 2.4674536401579127e-05}
{"train_loss": 0.11978460103273392, "global_step": 181783, "epoch": 2042, "lr": 2.4674036534372947e-05}
{"train_loss": 0.03561929985880852, "global_step": 181784, "epoch": 2042, "lr": 2.4673536670571544e-05}
{"train_loss": 0.09986910223960876, "global_step": 181785, "epoch": 2042, "lr": 2.467303681017496e-05}
{"train_loss": 0.07446326315402985, "global_step": 181786, "epoch": 2042, "lr": 2.467253695318328e-05}
{"train_loss": 0.09808360040187836, "global_step": 181787, "epoch": 2042, "lr": 2.4672037099596585e-05}
{"train_loss": 0.06012091413140297, "global_step": 181788, "epoch": 2042, "lr": 2.467153724941491e-05}
{"train_loss": 0.0807504653930664, "global_step": 181789, "epoch": 2042, "lr": 2.4671037402638354e-05}
{"train_loss": 0.06568840891122818, "global_step": 181790, "epoch": 2042, "lr": 2.467053755926695e-05}
{"train_loss": 0.11249274760484695, "global_step": 181791, "epoch": 2042, "lr": 2.4670037719300803e-05}
{"train_loss": 0.13080482184886932, "global_step": 181792, "epoch": 2042, "lr": 2.466953788273994e-05}
{"train_loss": 0.0933510884642601, "global_step": 181793, "epoch": 2042, "lr": 2.466903804958447e-05}
{"train_loss": 0.05030359327793121, "global_step": 181794, "epoch": 2042, "lr": 2.4668538219834418e-05}
{"train_loss": 0.06922104209661484, "global_step": 181795, "epoch": 2042, "lr": 2.4668038393489897e-05}
{"train_loss": 0.06606921553611755, "global_step": 181796, "epoch": 2042, "lr": 2.4667538570550925e-05}
{"train_loss": 0.06898070126771927, "global_step": 181797, "epoch": 2042, "lr": 2.4667038751017612e-05}
{"train_loss": 0.07984957844018936, "global_step": 181798, "epoch": 2042, "lr": 2.466653893488999e-05}
{"train_loss": 0.06362899392843246, "global_step": 181799, "epoch": 2042, "lr": 2.4666039122168162e-05}
{"train_loss": 0.2051214724779129, "global_step": 181800, "epoch": 2042, "lr": 2.4665539312852154e-05}
{"train_loss": 0.12710583209991455, "global_step": 181801, "epoch": 2042, "lr": 2.466503950694208e-05}
{"train_loss": 0.08254572004079819, "global_step": 181802, "epoch": 2042, "lr": 2.466453970443796e-05}
{"train_loss": 0.10509084165096283, "global_step": 181803, "epoch": 2042, "lr": 2.4664039905339903e-05}
{"train_loss": 0.10286782681941986, "global_step": 181804, "epoch": 2042, "lr": 2.4663540109647942e-05}
{"train_loss": 0.10633499175310135, "global_step": 181805, "epoch": 2042, "lr": 2.4663040317362158e-05}
{"train_loss": 0.06924216449260712, "global_step": 181806, "epoch": 2042, "lr": 2.4662540528482635e-05}
{"train_loss": 0.07102419435977936, "global_step": 181807, "epoch": 2042, "lr": 2.466204074300941e-05}
{"train_loss": 0.09482329338788986, "global_step": 181808, "epoch": 2042, "lr": 2.4661540960942575e-05}
{"train_loss": 0.07313374429941177, "global_step": 181809, "epoch": 2042, "lr": 2.4661041182282173e-05}
{"train_loss": 0.09644939750432968, "global_step": 181810, "epoch": 2042, "lr": 2.466054140702831e-05}
{"train_loss": 0.049192652106285095, "global_step": 181811, "epoch": 2042, "lr": 2.4660041635181025e-05}
{"train_loss": 0.10975325852632523, "global_step": 181812, "epoch": 2042, "lr": 2.4659541866740367e-05}
{"train_loss": 0.06413432955741882, "global_step": 181813, "epoch": 2042, "lr": 2.465904210170644e-05}
{"train_loss": 0.12308963388204575, "global_step": 181814, "epoch": 2042, "lr": 2.4658542340079282e-05}
{"train_loss": 0.07026325911283493, "global_step": 181815, "epoch": 2042, "lr": 2.4658042581859e-05}
{"train_loss": 0.090524822473526, "global_step": 181816, "epoch": 2042, "lr": 2.4657542827045627e-05}
{"train_loss": 0.06342659145593643, "global_step": 181817, "epoch": 2042, "lr": 2.4657043075639218e-05}
{"train_loss": 0.056286297738552094, "global_step": 181818, "epoch": 2042, "lr": 2.465654332763988e-05}
{"train_loss": 0.09198439866304398, "global_step": 181819, "epoch": 2042, "lr": 2.4656043583047645e-05}
{"train_loss": 0.05980135500431061, "global_step": 181820, "epoch": 2042, "lr": 2.4655543841862616e-05}
{"train_loss": 0.05634504184126854, "global_step": 181821, "epoch": 2042, "lr": 2.465504410408482e-05}
{"train_loss": 0.05949145928025246, "global_step": 181822, "epoch": 2042, "lr": 2.4654544369714343e-05}
{"train_loss": 0.11959273368120193, "global_step": 181823, "epoch": 2042, "lr": 2.465404463875127e-05}
{"train_loss": 0.07862377911806107, "global_step": 181824, "epoch": 2042, "lr": 2.4653544911195643e-05}
{"train_loss": 0.07585365325212479, "global_step": 181825, "epoch": 2042, "lr": 2.4653045187047547e-05}
{"train_loss": 0.08555010825479298, "global_step": 181826, "epoch": 2042, "lr": 2.4652545466307024e-05, "val_loss": 7.426219940185547}
{"train_loss": 0.06914595514535904, "global_step": 181827, "epoch": 2043, "lr": 2.465204574897418e-05}
{"train_loss": 0.08806996792554855, "global_step": 181828, "epoch": 2043, "lr": 2.4651546035049033e-05}
{"train_loss": 0.12764447927474976, "global_step": 181829, "epoch": 2043, "lr": 2.4651046324531703e-05}
{"train_loss": 0.06853518635034561, "global_step": 181830, "epoch": 2043, "lr": 2.4650546617422203e-05}
{"train_loss": 0.09025154262781143, "global_step": 181831, "epoch": 2043, "lr": 2.465004691372066e-05}
{"train_loss": 0.08660749346017838, "global_step": 181832, "epoch": 2043, "lr": 2.4649547213427082e-05}
{"train_loss": 0.09379561990499496, "global_step": 181833, "epoch": 2043, "lr": 2.4649047516541578e-05}
{"train_loss": 0.0607423335313797, "global_step": 181834, "epoch": 2043, "lr": 2.464854782306419e-05}
{"train_loss": 0.06865333765745163, "global_step": 181835, "epoch": 2043, "lr": 2.464804813299501e-05}
{"train_loss": 0.05496960133314133, "global_step": 181836, "epoch": 2043, "lr": 2.4647548446334073e-05}
{"train_loss": 0.0657448023557663, "global_step": 181837, "epoch": 2043, "lr": 2.4647048763081487e-05}
{"train_loss": 0.06895311921834946, "global_step": 181838, "epoch": 2043, "lr": 2.4646549083237273e-05}
{"train_loss": 0.06585835665464401, "global_step": 181839, "epoch": 2043, "lr": 2.464604940680154e-05}
{"train_loss": 0.09943214803934097, "global_step": 181840, "epoch": 2043, "lr": 2.464554973377432e-05}
{"train_loss": 0.08515045791864395, "global_step": 181841, "epoch": 2043, "lr": 2.4645050064155694e-05}
{"train_loss": 0.09994478523731232, "global_step": 181842, "epoch": 2043, "lr": 2.4644550397945753e-05}
{"train_loss": 0.11269094794988632, "global_step": 181843, "epoch": 2043, "lr": 2.4644050735144526e-05}
{"train_loss": 0.08121609687805176, "global_step": 181844, "epoch": 2043, "lr": 2.4643551075752113e-05}
{"train_loss": 0.08682861924171448, "global_step": 181845, "epoch": 2043, "lr": 2.4643051419768547e-05}
{"train_loss": 0.02493310533463955, "global_step": 181846, "epoch": 2043, "lr": 2.464255176719394e-05}
{"train_loss": 0.12387698143720627, "global_step": 181847, "epoch": 2043, "lr": 2.4642052118028303e-05}
{"train_loss": 0.10382749140262604, "global_step": 181848, "epoch": 2043, "lr": 2.4641552472271754e-05}
{"train_loss": 0.07552109658718109, "global_step": 181849, "epoch": 2043, "lr": 2.4641052829924344e-05}
{"train_loss": 0.13005430996418, "global_step": 181850, "epoch": 2043, "lr": 2.4640553190986116e-05}
{"train_loss": 0.08946412056684494, "global_step": 181851, "epoch": 2043, "lr": 2.4640053555457172e-05}
{"train_loss": 0.0680248960852623, "global_step": 181852, "epoch": 2043, "lr": 2.463955392333755e-05}
{"train_loss": 0.11808498203754425, "global_step": 181853, "epoch": 2043, "lr": 2.4639054294627346e-05}
{"train_loss": 0.05774553865194321, "global_step": 181854, "epoch": 2043, "lr": 2.4638554669326593e-05}
{"train_loss": 0.06298350542783737, "global_step": 181855, "epoch": 2043, "lr": 2.4638055047435394e-05}
{"train_loss": 0.08529859036207199, "global_step": 181856, "epoch": 2043, "lr": 2.46375554289538e-05}
{"train_loss": 0.07811082899570465, "global_step": 181857, "epoch": 2043, "lr": 2.463705581388186e-05}
{"train_loss": 0.1500837355852127, "global_step": 181858, "epoch": 2043, "lr": 2.463655620221968e-05}
{"train_loss": 0.07663698494434357, "global_step": 181859, "epoch": 2043, "lr": 2.4636056593967284e-05}
{"train_loss": 0.04782358184456825, "global_step": 181860, "epoch": 2043, "lr": 2.4635556989124765e-05}
{"train_loss": 0.08647359162569046, "global_step": 181861, "epoch": 2043, "lr": 2.46350573876922e-05}
{"train_loss": 0.05463806912302971, "global_step": 181862, "epoch": 2043, "lr": 2.463455778966963e-05}
{"train_loss": 0.08822030574083328, "global_step": 181863, "epoch": 2043, "lr": 2.4634058195057146e-05}
{"train_loss": 0.08810198307037354, "global_step": 181864, "epoch": 2043, "lr": 2.4633558603854785e-05}
{"train_loss": 0.05335336923599243, "global_step": 181865, "epoch": 2043, "lr": 2.4633059016062653e-05}
{"train_loss": 0.07130960375070572, "global_step": 181866, "epoch": 2043, "lr": 2.4632559431680783e-05}
{"train_loss": 0.1262693703174591, "global_step": 181867, "epoch": 2043, "lr": 2.4632059850709272e-05}
{"train_loss": 0.131209135055542, "global_step": 181868, "epoch": 2043, "lr": 2.4631560273148153e-05}
{"train_loss": 0.10470058023929596, "global_step": 181869, "epoch": 2043, "lr": 2.4631060698997527e-05}
{"train_loss": 0.060556575655937195, "global_step": 181870, "epoch": 2043, "lr": 2.4630561128257428e-05}
{"train_loss": 0.08988353610038757, "global_step": 181871, "epoch": 2043, "lr": 2.4630061560927965e-05}
{"train_loss": 0.06514433771371841, "global_step": 181872, "epoch": 2043, "lr": 2.462956199700916e-05}
{"train_loss": 0.026360291987657547, "global_step": 181873, "epoch": 2043, "lr": 2.4629062436501116e-05}
{"train_loss": 0.09944036602973938, "global_step": 181874, "epoch": 2043, "lr": 2.4628562879403873e-05}
{"train_loss": 0.0755041241645813, "global_step": 181875, "epoch": 2043, "lr": 2.4628063325717522e-05}
{"train_loss": 0.07786519825458527, "global_step": 181876, "epoch": 2043, "lr": 2.4627563775442102e-05}
{"train_loss": 0.06659989058971405, "global_step": 181877, "epoch": 2043, "lr": 2.46270642285777e-05}
{"train_loss": 0.07811000943183899, "global_step": 181878, "epoch": 2043, "lr": 2.46265646851244e-05}
{"train_loss": 0.11332619190216064, "global_step": 181879, "epoch": 2043, "lr": 2.462606514508222e-05}
{"train_loss": 0.0575350746512413, "global_step": 181880, "epoch": 2043, "lr": 2.4625565608451286e-05}
{"train_loss": 0.05445887893438339, "global_step": 181881, "epoch": 2043, "lr": 2.4625066075231612e-05}
{"train_loss": 0.04887694865465164, "global_step": 181882, "epoch": 2043, "lr": 2.4624566545423307e-05}
{"train_loss": 0.07321888208389282, "global_step": 181883, "epoch": 2043, "lr": 2.4624067019026403e-05}
{"train_loss": 0.035199958831071854, "global_step": 181884, "epoch": 2043, "lr": 2.4623567496041e-05}
{"train_loss": 0.10629038512706757, "global_step": 181885, "epoch": 2043, "lr": 2.4623067976467135e-05}
{"train_loss": 0.12633712589740753, "global_step": 181886, "epoch": 2043, "lr": 2.46225684603049e-05}
{"train_loss": 0.0654451996088028, "global_step": 181887, "epoch": 2043, "lr": 2.462206894755434e-05}
{"train_loss": 0.05010262504220009, "global_step": 181888, "epoch": 2043, "lr": 2.462156943821555e-05}
{"train_loss": 0.05925627797842026, "global_step": 181889, "epoch": 2043, "lr": 2.462106993228858e-05}
{"train_loss": 0.0861130952835083, "global_step": 181890, "epoch": 2043, "lr": 2.4620570429773483e-05}
{"train_loss": 0.06808934360742569, "global_step": 181891, "epoch": 2043, "lr": 2.4620070930670352e-05}
{"train_loss": 0.11946076899766922, "global_step": 181892, "epoch": 2043, "lr": 2.4619571434979233e-05}
{"train_loss": 0.0978107824921608, "global_step": 181893, "epoch": 2043, "lr": 2.4619071942700216e-05}
{"train_loss": 0.058505527675151825, "global_step": 181894, "epoch": 2043, "lr": 2.4618572453833357e-05}
{"train_loss": 0.08404237776994705, "global_step": 181895, "epoch": 2043, "lr": 2.46180729683787e-05}
{"train_loss": 0.05250653252005577, "global_step": 181896, "epoch": 2043, "lr": 2.4617573486336336e-05}
{"train_loss": 0.06589643657207489, "global_step": 181897, "epoch": 2043, "lr": 2.4617074007706354e-05}
{"train_loss": 0.058309875428676605, "global_step": 181898, "epoch": 2043, "lr": 2.4616574532488773e-05}
{"train_loss": 0.029992343857884407, "global_step": 181899, "epoch": 2043, "lr": 2.46160750606837e-05}
{"train_loss": 0.11713328957557678, "global_step": 181900, "epoch": 2043, "lr": 2.4615575592291173e-05}
{"train_loss": 0.07944726198911667, "global_step": 181901, "epoch": 2043, "lr": 2.4615076127311287e-05}
{"train_loss": 0.14053285121917725, "global_step": 181902, "epoch": 2043, "lr": 2.4614576665744082e-05}
{"train_loss": 0.10236511379480362, "global_step": 181903, "epoch": 2043, "lr": 2.4614077207589655e-05}
{"train_loss": 0.15292930603027344, "global_step": 181904, "epoch": 2043, "lr": 2.461357775284803e-05}
{"train_loss": 0.044402722269296646, "global_step": 181905, "epoch": 2043, "lr": 2.4613078301519328e-05}
{"train_loss": 0.072422094643116, "global_step": 181906, "epoch": 2043, "lr": 2.4612578853603563e-05}
{"train_loss": 0.06986206024885178, "global_step": 181907, "epoch": 2043, "lr": 2.4612079409100847e-05}
{"train_loss": 0.08534969389438629, "global_step": 181908, "epoch": 2043, "lr": 2.461157996801121e-05}
{"train_loss": 0.10368716716766357, "global_step": 181909, "epoch": 2043, "lr": 2.4611080530334756e-05}
{"train_loss": 0.06788072735071182, "global_step": 181910, "epoch": 2043, "lr": 2.461058109607151e-05}
{"train_loss": 0.08264631778001785, "global_step": 181911, "epoch": 2043, "lr": 2.4610081665221585e-05}
{"train_loss": 0.07068649679422379, "global_step": 181912, "epoch": 2043, "lr": 2.4609582237785004e-05}
{"train_loss": 0.04846912622451782, "global_step": 181913, "epoch": 2043, "lr": 2.4609082813761858e-05}
{"train_loss": 0.044012416154146194, "global_step": 181914, "epoch": 2043, "lr": 2.460858339315223e-05}
{"train_loss": 0.08099040154660686, "global_step": 181915, "epoch": 2043, "lr": 2.460808397595615e-05, "val_loss": 7.460171222686768}
{"train_loss": 0.08056914061307907, "global_step": 181916, "epoch": 2044, "lr": 2.4607584562173718e-05}
{"train_loss": 0.055388350039720535, "global_step": 181917, "epoch": 2044, "lr": 2.460708515180497e-05}
{"train_loss": 0.08096993714570999, "global_step": 181918, "epoch": 2044, "lr": 2.4606585744850008e-05}
{"train_loss": 0.04236549511551857, "global_step": 181919, "epoch": 2044, "lr": 2.4606086341308866e-05}
{"train_loss": 0.07541340589523315, "global_step": 181920, "epoch": 2044, "lr": 2.460558694118164e-05}
{"train_loss": 0.1213078424334526, "global_step": 181921, "epoch": 2044, "lr": 2.460508754446837e-05}
{"train_loss": 0.09355169534683228, "global_step": 181922, "epoch": 2044, "lr": 2.460458815116915e-05}
{"train_loss": 0.07033565640449524, "global_step": 181923, "epoch": 2044, "lr": 2.4604088761284022e-05}
{"train_loss": 0.074863500893116, "global_step": 181924, "epoch": 2044, "lr": 2.4603589374813075e-05}
{"train_loss": 0.05763280391693115, "global_step": 181925, "epoch": 2044, "lr": 2.460308999175635e-05}
{"train_loss": 0.037601351737976074, "global_step": 181926, "epoch": 2044, "lr": 2.4602590612113953e-05}
{"train_loss": 0.11814653873443604, "global_step": 181927, "epoch": 2044, "lr": 2.460209123588592e-05}
{"train_loss": 0.05449261516332626, "global_step": 181928, "epoch": 2044, "lr": 2.4601591863072316e-05}
{"train_loss": 0.09205704927444458, "global_step": 181929, "epoch": 2044, "lr": 2.4601092493673233e-05}
{"train_loss": 0.08285670727491379, "global_step": 181930, "epoch": 2044, "lr": 2.4600593127688703e-05}
{"train_loss": 0.071341373026371, "global_step": 181931, "epoch": 2044, "lr": 2.4600093765118836e-05}
{"train_loss": 0.04916108027100563, "global_step": 181932, "epoch": 2044, "lr": 2.4599594405963655e-05}
{"train_loss": 0.05998146906495094, "global_step": 181933, "epoch": 2044, "lr": 2.459909505022327e-05}
{"train_loss": 0.06344917416572571, "global_step": 181934, "epoch": 2044, "lr": 2.4598595697897708e-05}
{"train_loss": 0.03469875082373619, "global_step": 181935, "epoch": 2044, "lr": 2.4598096348987078e-05}
{"train_loss": 0.08508266508579254, "global_step": 181936, "epoch": 2044, "lr": 2.4597597003491397e-05}
{"train_loss": 0.11291257292032242, "global_step": 181937, "epoch": 2044, "lr": 2.4597097661410783e-05}
{"train_loss": 0.057746656239032745, "global_step": 181938, "epoch": 2044, "lr": 2.459659832274526e-05}
{"train_loss": 0.1494789570569992, "global_step": 181939, "epoch": 2044, "lr": 2.4596098987494938e-05}
{"train_loss": 0.08206629753112793, "global_step": 181940, "epoch": 2044, "lr": 2.4595599655659835e-05}
{"train_loss": 0.0690775141119957, "global_step": 181941, "epoch": 2044, "lr": 2.4595100327240067e-05}
{"train_loss": 0.07836748659610748, "global_step": 181942, "epoch": 2044, "lr": 2.4594601002235658e-05}
{"train_loss": 0.053688161075115204, "global_step": 181943, "epoch": 2044, "lr": 2.4594101680646713e-05}
{"train_loss": 0.04477570205926895, "global_step": 181944, "epoch": 2044, "lr": 2.4593602362473262e-05}
{"train_loss": 0.024567293003201485, "global_step": 181945, "epoch": 2044, "lr": 2.4593103047715414e-05}
{"train_loss": 0.0707528367638588, "global_step": 181946, "epoch": 2044, "lr": 2.4592603736373192e-05}
{"train_loss": 0.04807339981198311, "global_step": 181947, "epoch": 2044, "lr": 2.4592104428446706e-05}
{"train_loss": 0.03924741595983505, "global_step": 181948, "epoch": 2044, "lr": 2.4591605123935978e-05}
{"train_loss": 0.06881853938102722, "global_step": 181949, "epoch": 2044, "lr": 2.4591105822841103e-05}
{"train_loss": 0.04623667150735855, "global_step": 181950, "epoch": 2044, "lr": 2.4590606525162163e-05}
{"train_loss": 0.07887928932905197, "global_step": 181951, "epoch": 2044, "lr": 2.459010723089919e-05}
{"train_loss": 0.05561249703168869, "global_step": 181952, "epoch": 2044, "lr": 2.4589607940052278e-05}
{"train_loss": 0.053418051451444626, "global_step": 181953, "epoch": 2044, "lr": 2.4589108652621474e-05}
{"train_loss": 0.0417657233774662, "global_step": 181954, "epoch": 2044, "lr": 2.458860936860687e-05}
{"train_loss": 0.09171660244464874, "global_step": 181955, "epoch": 2044, "lr": 2.45881100880085e-05}
{"train_loss": 0.06859368085861206, "global_step": 181956, "epoch": 2044, "lr": 2.4587610810826462e-05}
{"train_loss": 0.04158305376768112, "global_step": 181957, "epoch": 2044, "lr": 2.4587111537060794e-05}
{"train_loss": 0.04424142837524414, "global_step": 181958, "epoch": 2044, "lr": 2.4586612266711602e-05}
{"train_loss": 0.0892290472984314, "global_step": 181959, "epoch": 2044, "lr": 2.4586112999778905e-05}
{"train_loss": 0.043157219886779785, "global_step": 181960, "epoch": 2044, "lr": 2.458561373626282e-05}
{"train_loss": 0.05999937281012535, "global_step": 181961, "epoch": 2044, "lr": 2.4585114476163367e-05}
{"train_loss": 0.06321332603693008, "global_step": 181962, "epoch": 2044, "lr": 2.4584615219480656e-05}
{"train_loss": 0.05608990043401718, "global_step": 181963, "epoch": 2044, "lr": 2.458411596621471e-05}
{"train_loss": 0.04976445436477661, "global_step": 181964, "epoch": 2044, "lr": 2.4583616716365647e-05}
{"train_loss": 0.09123975038528442, "global_step": 181965, "epoch": 2044, "lr": 2.4583117469933505e-05}
{"train_loss": 0.07605361938476562, "global_step": 181966, "epoch": 2044, "lr": 2.458261822691833e-05}
{"train_loss": 0.11784537136554718, "global_step": 181967, "epoch": 2044, "lr": 2.458211898732023e-05}
{"train_loss": 0.0885983482003212, "global_step": 181968, "epoch": 2044, "lr": 2.458161975113924e-05}
{"train_loss": 0.09976896643638611, "global_step": 181969, "epoch": 2044, "lr": 2.4581120518375462e-05}
{"train_loss": 0.05851353332400322, "global_step": 181970, "epoch": 2044, "lr": 2.458062128902892e-05}
{"train_loss": 0.06028422340750694, "global_step": 181971, "epoch": 2044, "lr": 2.4580122063099718e-05}
{"train_loss": 0.05653662234544754, "global_step": 181972, "epoch": 2044, "lr": 2.4579622840587902e-05}
{"train_loss": 0.06328950077295303, "global_step": 181973, "epoch": 2044, "lr": 2.4579123621493555e-05}
{"train_loss": 0.09106072783470154, "global_step": 181974, "epoch": 2044, "lr": 2.457862440581672e-05}
{"train_loss": 0.06422818452119827, "global_step": 181975, "epoch": 2044, "lr": 2.4578125193557494e-05}
{"train_loss": 0.08046874403953552, "global_step": 181976, "epoch": 2044, "lr": 2.457762598471591e-05}
{"train_loss": 0.04592888802289963, "global_step": 181977, "epoch": 2044, "lr": 2.457712677929208e-05}
{"train_loss": 0.09564384073019028, "global_step": 181978, "epoch": 2044, "lr": 2.4576627577286027e-05}
{"train_loss": 0.07707936316728592, "global_step": 181979, "epoch": 2044, "lr": 2.457612837869785e-05}
{"train_loss": 0.06057076156139374, "global_step": 181980, "epoch": 2044, "lr": 2.4575629183527582e-05}
{"train_loss": 0.0746678039431572, "global_step": 181981, "epoch": 2044, "lr": 2.4575129991775337e-05}
{"train_loss": 0.04221029579639435, "global_step": 181982, "epoch": 2044, "lr": 2.4574630803441127e-05}
{"train_loss": 0.05117711424827576, "global_step": 181983, "epoch": 2044, "lr": 2.4574131618525075e-05}
{"train_loss": 0.09324173629283905, "global_step": 181984, "epoch": 2044, "lr": 2.45736324370272e-05}
{"train_loss": 0.04755457490682602, "global_step": 181985, "epoch": 2044, "lr": 2.4573133258947612e-05}
{"train_loss": 0.08933647722005844, "global_step": 181986, "epoch": 2044, "lr": 2.4572634084286333e-05}
{"train_loss": 0.07719678431749344, "global_step": 181987, "epoch": 2044, "lr": 2.4572134913043456e-05}
{"train_loss": 0.0908973217010498, "global_step": 181988, "epoch": 2044, "lr": 2.457163574521906e-05}
{"train_loss": 0.11073926836252213, "global_step": 181989, "epoch": 2044, "lr": 2.4571136580813187e-05}
{"train_loss": 0.05195975303649902, "global_step": 181990, "epoch": 2044, "lr": 2.4570637419825927e-05}
{"train_loss": 0.08031709492206573, "global_step": 181991, "epoch": 2044, "lr": 2.457013826225732e-05}
{"train_loss": 0.07304872572422028, "global_step": 181992, "epoch": 2044, "lr": 2.456963910810746e-05}
{"train_loss": 0.09311196953058243, "global_step": 181993, "epoch": 2044, "lr": 2.4569139957376392e-05}
{"train_loss": 0.15318726003170013, "global_step": 181994, "epoch": 2044, "lr": 2.4568640810064204e-05}
{"train_loss": 0.07594470679759979, "global_step": 181995, "epoch": 2044, "lr": 2.456814166617094e-05}
{"train_loss": 0.04869873449206352, "global_step": 181996, "epoch": 2044, "lr": 2.4567642525696694e-05}
{"train_loss": 0.043985530734062195, "global_step": 181997, "epoch": 2044, "lr": 2.45671433886415e-05}
{"train_loss": 0.040623098611831665, "global_step": 181998, "epoch": 2044, "lr": 2.4566644255005466e-05}
{"train_loss": 0.12113955616950989, "global_step": 181999, "epoch": 2044, "lr": 2.4566145124788616e-05}
{"train_loss": 0.09155090898275375, "global_step": 182000, "epoch": 2044, "lr": 2.4565645997991055e-05}
{"train_loss": 0.05833721160888672, "global_step": 182001, "epoch": 2044, "lr": 2.4565146874612817e-05}
{"train_loss": 0.13646367192268372, "global_step": 182002, "epoch": 2044, "lr": 2.4564647754654003e-05}
{"train_loss": 0.046996571123600006, "global_step": 182003, "epoch": 2044, "lr": 2.4564148638114638e-05}
{"train_loss": 0.07154982195895039, "global_step": 182004, "epoch": 2044, "lr": 2.4563649524994836e-05, "val_loss": 7.384334564208984}
{"train_loss": 0.10582456737756729, "global_step": 182005, "epoch": 2045, "lr": 2.4563150415294645e-05}
{"train_loss": 0.07616983354091644, "global_step": 182006, "epoch": 2045, "lr": 2.4562651309014106e-05}
{"train_loss": 0.1216307133436203, "global_step": 182007, "epoch": 2045, "lr": 2.4562152206153327e-05}
{"train_loss": 0.05793915316462517, "global_step": 182008, "epoch": 2045, "lr": 2.4561653106712335e-05}
{"train_loss": 0.10650131851434708, "global_step": 182009, "epoch": 2045, "lr": 2.456115401069124e-05}
{"train_loss": 0.04832189157605171, "global_step": 182010, "epoch": 2045, "lr": 2.456065491809007e-05}
{"train_loss": 0.07012317329645157, "global_step": 182011, "epoch": 2045, "lr": 2.4560155828908932e-05}
{"train_loss": 0.06190146878361702, "global_step": 182012, "epoch": 2045, "lr": 2.4559656743147847e-05}
{"train_loss": 0.08577477186918259, "global_step": 182013, "epoch": 2045, "lr": 2.4559157660806926e-05}
{"train_loss": 0.08236876130104065, "global_step": 182014, "epoch": 2045, "lr": 2.4558658581886194e-05}
{"train_loss": 0.07948978990316391, "global_step": 182015, "epoch": 2045, "lr": 2.455815950638577e-05}
{"train_loss": 0.09138083457946777, "global_step": 182016, "epoch": 2045, "lr": 2.4557660434305662e-05}
{"train_loss": 0.07527744024991989, "global_step": 182017, "epoch": 2045, "lr": 2.4557161365645986e-05}
{"train_loss": 0.07345753908157349, "global_step": 182018, "epoch": 2045, "lr": 2.4556662300406775e-05}
{"train_loss": 0.03399338573217392, "global_step": 182019, "epoch": 2045, "lr": 2.455616323858813e-05}
{"train_loss": 0.08583725243806839, "global_step": 182020, "epoch": 2045, "lr": 2.4555664180190075e-05}
{"train_loss": 0.04571348428726196, "global_step": 182021, "epoch": 2045, "lr": 2.4555165125212726e-05}
{"train_loss": 0.11679276823997498, "global_step": 182022, "epoch": 2045, "lr": 2.4554666073656102e-05}
{"train_loss": 0.05635937303304672, "global_step": 182023, "epoch": 2045, "lr": 2.4554167025520292e-05}
{"train_loss": 0.07137982547283173, "global_step": 182024, "epoch": 2045, "lr": 2.4553667980805384e-05}
{"train_loss": 0.0730980932712555, "global_step": 182025, "epoch": 2045, "lr": 2.4553168939511405e-05}
{"train_loss": 0.10831747949123383, "global_step": 182026, "epoch": 2045, "lr": 2.4552669901638466e-05}
{"train_loss": 0.11098440736532211, "global_step": 182027, "epoch": 2045, "lr": 2.4552170867186585e-05}
{"train_loss": 0.11199983209371567, "global_step": 182028, "epoch": 2045, "lr": 2.455167183615588e-05}
{"train_loss": 0.05092205852270126, "global_step": 182029, "epoch": 2045, "lr": 2.455117280854637e-05}
{"train_loss": 0.11658106744289398, "global_step": 182030, "epoch": 2045, "lr": 2.4550673784358162e-05}
{"train_loss": 0.08135071396827698, "global_step": 182031, "epoch": 2045, "lr": 2.4550174763591294e-05}
{"train_loss": 0.058085694909095764, "global_step": 182032, "epoch": 2045, "lr": 2.4549675746245854e-05}
{"train_loss": 0.04010109230875969, "global_step": 182033, "epoch": 2045, "lr": 2.4549176732321887e-05}
{"train_loss": 0.06952887028455734, "global_step": 182034, "epoch": 2045, "lr": 2.454867772181949e-05}
{"train_loss": 0.046874258667230606, "global_step": 182035, "epoch": 2045, "lr": 2.4548178714738696e-05}
{"train_loss": 0.06027001142501831, "global_step": 182036, "epoch": 2045, "lr": 2.4547679711079607e-05}
{"train_loss": 0.05042345076799393, "global_step": 182037, "epoch": 2045, "lr": 2.4547180710842253e-05}
{"train_loss": 0.06212252005934715, "global_step": 182038, "epoch": 2045, "lr": 2.4546681714026736e-05}
{"train_loss": 0.07420029491186142, "global_step": 182039, "epoch": 2045, "lr": 2.4546182720633093e-05}
{"train_loss": 0.08028066903352737, "global_step": 182040, "epoch": 2045, "lr": 2.4545683730661423e-05}
{"train_loss": 0.07532011717557907, "global_step": 182041, "epoch": 2045, "lr": 2.454518474411176e-05}
{"train_loss": 0.10169646143913269, "global_step": 182042, "epoch": 2045, "lr": 2.45446857609842e-05}
{"train_loss": 0.1292605698108673, "global_step": 182043, "epoch": 2045, "lr": 2.45441867812788e-05}
{"train_loss": 0.057693034410476685, "global_step": 182044, "epoch": 2045, "lr": 2.45436878049956e-05}
{"train_loss": 0.12656262516975403, "global_step": 182045, "epoch": 2045, "lr": 2.4543188832134718e-05}
{"train_loss": 0.04107996076345444, "global_step": 182046, "epoch": 2045, "lr": 2.4542689862696173e-05}
{"train_loss": 0.08917383849620819, "global_step": 182047, "epoch": 2045, "lr": 2.454219089668007e-05}
{"train_loss": 0.05500193312764168, "global_step": 182048, "epoch": 2045, "lr": 2.4541691934086434e-05}
{"train_loss": 0.04548582434654236, "global_step": 182049, "epoch": 2045, "lr": 2.4541192974915382e-05}
{"train_loss": 0.06475681066513062, "global_step": 182050, "epoch": 2045, "lr": 2.4540694019166943e-05}
{"train_loss": 0.06061534211039543, "global_step": 182051, "epoch": 2045, "lr": 2.454019506684121e-05}
{"train_loss": 0.04488373547792435, "global_step": 182052, "epoch": 2045, "lr": 2.4539696117938222e-05}
{"train_loss": 0.08333524316549301, "global_step": 182053, "epoch": 2045, "lr": 2.4539197172458072e-05}
{"train_loss": 0.07517016679048538, "global_step": 182054, "epoch": 2045, "lr": 2.453869823040081e-05}
{"train_loss": 0.12057586014270782, "global_step": 182055, "epoch": 2045, "lr": 2.4538199291766518e-05}
{"train_loss": 0.1026601567864418, "global_step": 182056, "epoch": 2045, "lr": 2.4537700356555243e-05}
{"train_loss": 0.08411513268947601, "global_step": 182057, "epoch": 2045, "lr": 2.4537201424767076e-05}
{"train_loss": 0.0889139398932457, "global_step": 182058, "epoch": 2045, "lr": 2.453670249640206e-05}
{"train_loss": 0.09292938560247421, "global_step": 182059, "epoch": 2045, "lr": 2.453620357146027e-05}
{"train_loss": 0.0933573991060257, "global_step": 182060, "epoch": 2045, "lr": 2.4535704649941798e-05}
{"train_loss": 0.05501682683825493, "global_step": 182061, "epoch": 2045, "lr": 2.4535205731846667e-05}
{"train_loss": 0.10095742344856262, "global_step": 182062, "epoch": 2045, "lr": 2.453470681717499e-05}
{"train_loss": 0.055884040892124176, "global_step": 182063, "epoch": 2045, "lr": 2.453420790592679e-05}
{"train_loss": 0.17016227543354034, "global_step": 182064, "epoch": 2045, "lr": 2.453370899810218e-05}
{"train_loss": 0.04547469690442085, "global_step": 182065, "epoch": 2045, "lr": 2.4533210093701175e-05}
{"train_loss": 0.10463954508304596, "global_step": 182066, "epoch": 2045, "lr": 2.4532711192723895e-05}
{"train_loss": 0.06321051716804504, "global_step": 182067, "epoch": 2045, "lr": 2.453221229517036e-05}
{"train_loss": 0.048372991383075714, "global_step": 182068, "epoch": 2045, "lr": 2.4531713401040678e-05}
{"train_loss": 0.09163425117731094, "global_step": 182069, "epoch": 2045, "lr": 2.4531214510334876e-05}
{"train_loss": 0.08071736991405487, "global_step": 182070, "epoch": 2045, "lr": 2.453071562305307e-05}
{"train_loss": 0.057684361934661865, "global_step": 182071, "epoch": 2045, "lr": 2.4530216739195267e-05}
{"train_loss": 0.06628675013780594, "global_step": 182072, "epoch": 2045, "lr": 2.4529717858761592e-05}
{"train_loss": 0.06505585461854935, "global_step": 182073, "epoch": 2045, "lr": 2.4529218981752065e-05}
{"train_loss": 0.08321961015462875, "global_step": 182074, "epoch": 2045, "lr": 2.452872010816679e-05}
{"train_loss": 0.050550676882267, "global_step": 182075, "epoch": 2045, "lr": 2.4528221238005807e-05}
{"train_loss": 0.040580373257398605, "global_step": 182076, "epoch": 2045, "lr": 2.4527722371269207e-05}
{"train_loss": 0.05336639657616615, "global_step": 182077, "epoch": 2045, "lr": 2.4527223507957026e-05}
{"train_loss": 0.08892741799354553, "global_step": 182078, "epoch": 2045, "lr": 2.4526724648069373e-05}
{"train_loss": 0.09974204748868942, "global_step": 182079, "epoch": 2045, "lr": 2.4526225791606266e-05}
{"train_loss": 0.0646115317940712, "global_step": 182080, "epoch": 2045, "lr": 2.4525726938567823e-05}
{"train_loss": 0.11877111345529556, "global_step": 182081, "epoch": 2045, "lr": 2.452522808895408e-05}
{"train_loss": 0.06722094863653183, "global_step": 182082, "epoch": 2045, "lr": 2.4524729242765093e-05}
{"train_loss": 0.05186641588807106, "global_step": 182083, "epoch": 2045, "lr": 2.4524230400000962e-05}
{"train_loss": 0.040446944534778595, "global_step": 182084, "epoch": 2045, "lr": 2.452373156066172e-05}
{"train_loss": 0.08123360574245453, "global_step": 182085, "epoch": 2045, "lr": 2.4523232724747475e-05}
{"train_loss": 0.02788100577890873, "global_step": 182086, "epoch": 2045, "lr": 2.4522733892258247e-05}
{"train_loss": 0.07625001668930054, "global_step": 182087, "epoch": 2045, "lr": 2.452223506319415e-05}
{"train_loss": 0.10218405723571777, "global_step": 182088, "epoch": 2045, "lr": 2.4521736237555202e-05}
{"train_loss": 0.05838507413864136, "global_step": 182089, "epoch": 2045, "lr": 2.4521237415341518e-05}
{"train_loss": 0.06956429779529572, "global_step": 182090, "epoch": 2045, "lr": 2.452073859655313e-05}
{"train_loss": 0.12170106172561646, "global_step": 182091, "epoch": 2045, "lr": 2.4520239781190128e-05}
{"train_loss": 0.07000173628330231, "global_step": 182092, "epoch": 2045, "lr": 2.451974096925256e-05}
{"train_loss": 0.07654436851401678, "global_step": 182093, "epoch": 2045, "lr": 2.4519242160740513e-05, "val_loss": 7.4994120597839355, "train_action_mse_error": 23.869369506835938}
{"train_loss": 0.04430657625198364, "global_step": 182094, "epoch": 2046, "lr": 2.4518743355654027e-05}
{"train_loss": 0.1092052161693573, "global_step": 182095, "epoch": 2046, "lr": 2.4518244553993187e-05}
{"train_loss": 0.0708000436425209, "global_step": 182096, "epoch": 2046, "lr": 2.4517745755758077e-05}
{"train_loss": 0.05706014856696129, "global_step": 182097, "epoch": 2046, "lr": 2.4517246960948727e-05}
{"train_loss": 0.05826687067747116, "global_step": 182098, "epoch": 2046, "lr": 2.4516748169565235e-05}
{"train_loss": 0.038025662302970886, "global_step": 182099, "epoch": 2046, "lr": 2.4516249381607643e-05}
{"train_loss": 0.04572243243455887, "global_step": 182100, "epoch": 2046, "lr": 2.451575059707605e-05}
{"train_loss": 0.1103375256061554, "global_step": 182101, "epoch": 2046, "lr": 2.451525181597048e-05}
{"train_loss": 0.10412900149822235, "global_step": 182102, "epoch": 2046, "lr": 2.451475303829105e-05}
{"train_loss": 0.08789754658937454, "global_step": 182103, "epoch": 2046, "lr": 2.4514254264037772e-05}
{"train_loss": 0.058958761394023895, "global_step": 182104, "epoch": 2046, "lr": 2.451375549321077e-05}
{"train_loss": 0.033109150826931, "global_step": 182105, "epoch": 2046, "lr": 2.4513256725810058e-05}
{"train_loss": 0.040848467499017715, "global_step": 182106, "epoch": 2046, "lr": 2.4512757961835753e-05}
{"train_loss": 0.07695949077606201, "global_step": 182107, "epoch": 2046, "lr": 2.4512259201287873e-05}
{"train_loss": 0.07379709184169769, "global_step": 182108, "epoch": 2046, "lr": 2.4511760444166532e-05}
{"train_loss": 0.1385922133922577, "global_step": 182109, "epoch": 2046, "lr": 2.4511261690471748e-05}
{"train_loss": 0.036260075867176056, "global_step": 182110, "epoch": 2046, "lr": 2.451076294020364e-05}
{"train_loss": 0.06686106324195862, "global_step": 182111, "epoch": 2046, "lr": 2.4510264193362233e-05}
{"train_loss": 0.05269595608115196, "global_step": 182112, "epoch": 2046, "lr": 2.450976544994762e-05}
{"train_loss": 0.03209458291530609, "global_step": 182113, "epoch": 2046, "lr": 2.4509266709959845e-05}
{"train_loss": 0.05034623295068741, "global_step": 182114, "epoch": 2046, "lr": 2.450876797339899e-05}
{"train_loss": 0.05187678337097168, "global_step": 182115, "epoch": 2046, "lr": 2.4508269240265136e-05}
{"train_loss": 0.059113550931215286, "global_step": 182116, "epoch": 2046, "lr": 2.4507770510558332e-05}
{"train_loss": 0.06535155326128006, "global_step": 182117, "epoch": 2046, "lr": 2.4507271784278633e-05}
{"train_loss": 0.06719473004341125, "global_step": 182118, "epoch": 2046, "lr": 2.450677306142614e-05}
{"train_loss": 0.06203209608793259, "global_step": 182119, "epoch": 2046, "lr": 2.4506274342000877e-05}
{"train_loss": 0.10890790820121765, "global_step": 182120, "epoch": 2046, "lr": 2.450577562600296e-05}
{"train_loss": 0.08980018645524979, "global_step": 182121, "epoch": 2046, "lr": 2.4505276913432428e-05}
{"train_loss": 0.07283749431371689, "global_step": 182122, "epoch": 2046, "lr": 2.450477820428933e-05}
{"train_loss": 0.05182122066617012, "global_step": 182123, "epoch": 2046, "lr": 2.4504279498573772e-05}
{"train_loss": 0.07301101833581924, "global_step": 182124, "epoch": 2046, "lr": 2.4503780796285787e-05}
{"train_loss": 0.09934485703706741, "global_step": 182125, "epoch": 2046, "lr": 2.450328209742548e-05}
{"train_loss": 0.10061249881982803, "global_step": 182126, "epoch": 2046, "lr": 2.4502783401992867e-05}
{"train_loss": 0.08764231204986572, "global_step": 182127, "epoch": 2046, "lr": 2.4502284709988072e-05}
{"train_loss": 0.0850384458899498, "global_step": 182128, "epoch": 2046, "lr": 2.450178602141111e-05}
{"train_loss": 0.13547056913375854, "global_step": 182129, "epoch": 2046, "lr": 2.4501287336262095e-05}
{"train_loss": 0.11064115166664124, "global_step": 182130, "epoch": 2046, "lr": 2.4500788654541052e-05}
{"train_loss": 0.017757359892129898, "global_step": 182131, "epoch": 2046, "lr": 2.4500289976248064e-05}
{"train_loss": 0.1463184654712677, "global_step": 182132, "epoch": 2046, "lr": 2.4499791301383217e-05}
{"train_loss": 0.0804537832736969, "global_step": 182133, "epoch": 2046, "lr": 2.449929262994655e-05}
{"train_loss": 0.08788491785526276, "global_step": 182134, "epoch": 2046, "lr": 2.449879396193816e-05}
{"train_loss": 0.08939068019390106, "global_step": 182135, "epoch": 2046, "lr": 2.4498295297358077e-05}
{"train_loss": 0.11039147526025772, "global_step": 182136, "epoch": 2046, "lr": 2.4497796636206406e-05}
{"train_loss": 0.044909462332725525, "global_step": 182137, "epoch": 2046, "lr": 2.4497297978483176e-05}
{"train_loss": 0.07092221826314926, "global_step": 182138, "epoch": 2046, "lr": 2.449679932418849e-05}
{"train_loss": 0.06281419843435287, "global_step": 182139, "epoch": 2046, "lr": 2.4496300673322388e-05}
{"train_loss": 0.09657622873783112, "global_step": 182140, "epoch": 2046, "lr": 2.4495802025884957e-05}
{"train_loss": 0.08116480708122253, "global_step": 182141, "epoch": 2046, "lr": 2.449530338187624e-05}
{"train_loss": 0.06561902165412903, "global_step": 182142, "epoch": 2046, "lr": 2.4494804741296336e-05}
{"train_loss": 0.08085871487855911, "global_step": 182143, "epoch": 2046, "lr": 2.4494306104145276e-05}
{"train_loss": 0.08359123766422272, "global_step": 182144, "epoch": 2046, "lr": 2.4493807470423168e-05}
{"train_loss": 0.09728573262691498, "global_step": 182145, "epoch": 2046, "lr": 2.4493308840130035e-05}
{"train_loss": 0.12967351078987122, "global_step": 182146, "epoch": 2046, "lr": 2.4492810213265984e-05}
{"train_loss": 0.1320902407169342, "global_step": 182147, "epoch": 2046, "lr": 2.4492311589831047e-05}
{"train_loss": 0.02767515741288662, "global_step": 182148, "epoch": 2046, "lr": 2.4491812969825322e-05}
{"train_loss": 0.07587037235498428, "global_step": 182149, "epoch": 2046, "lr": 2.4491314353248845e-05}
{"train_loss": 0.0737704262137413, "global_step": 182150, "epoch": 2046, "lr": 2.4490815740101703e-05}
{"train_loss": 0.026752350851893425, "global_step": 182151, "epoch": 2046, "lr": 2.4490317130383976e-05}
{"train_loss": 0.11665132641792297, "global_step": 182152, "epoch": 2046, "lr": 2.4489818524095697e-05}
{"train_loss": 0.09488740563392639, "global_step": 182153, "epoch": 2046, "lr": 2.448931992123697e-05}
{"train_loss": 0.07159342616796494, "global_step": 182154, "epoch": 2046, "lr": 2.4488821321807843e-05}
{"train_loss": 0.05145307630300522, "global_step": 182155, "epoch": 2046, "lr": 2.4488322725808364e-05}
{"train_loss": 0.06111426651477814, "global_step": 182156, "epoch": 2046, "lr": 2.448782413323864e-05}
{"train_loss": 0.05897730961441994, "global_step": 182157, "epoch": 2046, "lr": 2.4487325544098695e-05}
{"train_loss": 0.08279754966497421, "global_step": 182158, "epoch": 2046, "lr": 2.4486826958388637e-05}
{"train_loss": 0.07839709520339966, "global_step": 182159, "epoch": 2046, "lr": 2.4486328376108513e-05}
{"train_loss": 0.09013356268405914, "global_step": 182160, "epoch": 2046, "lr": 2.4485829797258375e-05}
{"train_loss": 0.048859499394893646, "global_step": 182161, "epoch": 2046, "lr": 2.4485331221838324e-05}
{"train_loss": 0.038714535534381866, "global_step": 182162, "epoch": 2046, "lr": 2.4484832649848394e-05}
{"train_loss": 0.09802303463220596, "global_step": 182163, "epoch": 2046, "lr": 2.4484334081288683e-05}
{"train_loss": 0.06493505090475082, "global_step": 182164, "epoch": 2046, "lr": 2.448383551615922e-05}
{"train_loss": 0.05759806931018829, "global_step": 182165, "epoch": 2046, "lr": 2.448333695446012e-05}
{"train_loss": 0.05885499343276024, "global_step": 182166, "epoch": 2046, "lr": 2.44828383961914e-05}
{"train_loss": 0.11437001824378967, "global_step": 182167, "epoch": 2046, "lr": 2.4482339841353152e-05}
{"train_loss": 0.07857158035039902, "global_step": 182168, "epoch": 2046, "lr": 2.448184128994547e-05}
{"train_loss": 0.06360355764627457, "global_step": 182169, "epoch": 2046, "lr": 2.448134274196836e-05}
{"train_loss": 0.1246025413274765, "global_step": 182170, "epoch": 2046, "lr": 2.4480844197421955e-05}
{"train_loss": 0.09317740052938461, "global_step": 182171, "epoch": 2046, "lr": 2.4480345656306257e-05}
{"train_loss": 0.11257953941822052, "global_step": 182172, "epoch": 2046, "lr": 2.4479847118621395e-05}
{"train_loss": 0.10726677626371384, "global_step": 182173, "epoch": 2046, "lr": 2.4479348584367386e-05}
{"train_loss": 0.05572943389415741, "global_step": 182174, "epoch": 2046, "lr": 2.4478850053544334e-05}
{"train_loss": 0.07329802215099335, "global_step": 182175, "epoch": 2046, "lr": 2.447835152615227e-05}
{"train_loss": 0.057482920587062836, "global_step": 182176, "epoch": 2046, "lr": 2.4477853002191304e-05}
{"train_loss": 0.0708310529589653, "global_step": 182177, "epoch": 2046, "lr": 2.4477354481661458e-05}
{"train_loss": 0.05270175635814667, "global_step": 182178, "epoch": 2046, "lr": 2.4476855964562834e-05}
{"train_loss": 0.12148477137088776, "global_step": 182179, "epoch": 2046, "lr": 2.4476357450895476e-05}
{"train_loss": 0.11938537657260895, "global_step": 182180, "epoch": 2046, "lr": 2.4475858940659476e-05}
{"train_loss": 0.06202000752091408, "global_step": 182181, "epoch": 2046, "lr": 2.447536043385486e-05}
{"train_loss": 0.0773409019025524, "global_step": 182182, "epoch": 2046, "lr": 2.4474861930481747e-05, "val_loss": 7.497487545013428}
{"train_loss": 0.06486275047063828, "global_step": 182183, "epoch": 2047, "lr": 2.447436343054016e-05}
{"train_loss": 0.04015548899769783, "global_step": 182184, "epoch": 2047, "lr": 2.4473864934030198e-05}
{"train_loss": 0.03076218254864216, "global_step": 182185, "epoch": 2047, "lr": 2.44733664409519e-05}
{"train_loss": 0.07913568615913391, "global_step": 182186, "epoch": 2047, "lr": 2.447286795130534e-05}
{"train_loss": 0.06398624926805496, "global_step": 182187, "epoch": 2047, "lr": 2.4472369465090617e-05}
{"train_loss": 0.05068551003932953, "global_step": 182188, "epoch": 2047, "lr": 2.4471870982307748e-05}
{"train_loss": 0.06792182475328445, "global_step": 182189, "epoch": 2047, "lr": 2.4471372502956845e-05}
{"train_loss": 0.0735805407166481, "global_step": 182190, "epoch": 2047, "lr": 2.4470874027037944e-05}
{"train_loss": 0.05390506982803345, "global_step": 182191, "epoch": 2047, "lr": 2.4470375554551133e-05}
{"train_loss": 0.06950980424880981, "global_step": 182192, "epoch": 2047, "lr": 2.4469877085496475e-05}
{"train_loss": 0.034806977957487106, "global_step": 182193, "epoch": 2047, "lr": 2.4469378619874005e-05}
{"train_loss": 0.08591718971729279, "global_step": 182194, "epoch": 2047, "lr": 2.4468880157683837e-05}
{"train_loss": 0.0337943471968174, "global_step": 182195, "epoch": 2047, "lr": 2.4468381698926003e-05}
{"train_loss": 0.05784811079502106, "global_step": 182196, "epoch": 2047, "lr": 2.4467883243600597e-05}
{"train_loss": 0.0750940814614296, "global_step": 182197, "epoch": 2047, "lr": 2.4467384791707678e-05}
{"train_loss": 0.11178132146596909, "global_step": 182198, "epoch": 2047, "lr": 2.446688634324728e-05}
{"train_loss": 0.06095946952700615, "global_step": 182199, "epoch": 2047, "lr": 2.4466387898219522e-05}
{"train_loss": 0.1343059539794922, "global_step": 182200, "epoch": 2047, "lr": 2.446588945662443e-05}
{"train_loss": 0.05574122816324234, "global_step": 182201, "epoch": 2047, "lr": 2.446539101846211e-05}
{"train_loss": 0.0913916528224945, "global_step": 182202, "epoch": 2047, "lr": 2.4464892583732578e-05}
{"train_loss": 0.11563009768724442, "global_step": 182203, "epoch": 2047, "lr": 2.4464394152435938e-05}
{"train_loss": 0.05917412415146828, "global_step": 182204, "epoch": 2047, "lr": 2.4463895724572265e-05}
{"train_loss": 0.07202830910682678, "global_step": 182205, "epoch": 2047, "lr": 2.4463397300141594e-05}
{"train_loss": 0.10548414289951324, "global_step": 182206, "epoch": 2047, "lr": 2.446289887914402e-05}
{"train_loss": 0.11694269627332687, "global_step": 182207, "epoch": 2047, "lr": 2.4462400461579588e-05}
{"train_loss": 0.04929900914430618, "global_step": 182208, "epoch": 2047, "lr": 2.4461902047448386e-05}
{"train_loss": 0.05994151905179024, "global_step": 182209, "epoch": 2047, "lr": 2.4461403636750456e-05}
{"train_loss": 0.054058752954006195, "global_step": 182210, "epoch": 2047, "lr": 2.4460905229485896e-05}
{"train_loss": 0.05000294744968414, "global_step": 182211, "epoch": 2047, "lr": 2.4460406825654737e-05}
{"train_loss": 0.09165237098932266, "global_step": 182212, "epoch": 2047, "lr": 2.4459908425257084e-05}
{"train_loss": 0.11089416593313217, "global_step": 182213, "epoch": 2047, "lr": 2.4459410028292966e-05}
{"train_loss": 0.10018955171108246, "global_step": 182214, "epoch": 2047, "lr": 2.4458911634762487e-05}
{"train_loss": 0.04165224730968475, "global_step": 182215, "epoch": 2047, "lr": 2.445841324466568e-05}
{"train_loss": 0.08683837205171585, "global_step": 182216, "epoch": 2047, "lr": 2.4457914858002644e-05}
{"train_loss": 0.04132138192653656, "global_step": 182217, "epoch": 2047, "lr": 2.4457416474773413e-05}
{"train_loss": 0.05599309131503105, "global_step": 182218, "epoch": 2047, "lr": 2.4456918094978095e-05}
{"train_loss": 0.09655819088220596, "global_step": 182219, "epoch": 2047, "lr": 2.445641971861671e-05}
{"train_loss": 0.10582675039768219, "global_step": 182220, "epoch": 2047, "lr": 2.445592134568936e-05}
{"train_loss": 0.08487455546855927, "global_step": 182221, "epoch": 2047, "lr": 2.445542297619609e-05}
{"train_loss": 0.07307140529155731, "global_step": 182222, "epoch": 2047, "lr": 2.445492461013697e-05}
{"train_loss": 0.1407562494277954, "global_step": 182223, "epoch": 2047, "lr": 2.4454426247512096e-05}
{"train_loss": 0.06567291915416718, "global_step": 182224, "epoch": 2047, "lr": 2.4453927888321497e-05}
{"train_loss": 0.08057738095521927, "global_step": 182225, "epoch": 2047, "lr": 2.4453429532565275e-05}
{"train_loss": 0.05807947367429733, "global_step": 182226, "epoch": 2047, "lr": 2.4452931180243454e-05}
{"train_loss": 0.041080743074417114, "global_step": 182227, "epoch": 2047, "lr": 2.4452432831356147e-05}
{"train_loss": 0.042467713356018066, "global_step": 182228, "epoch": 2047, "lr": 2.4451934485903383e-05}
{"train_loss": 0.08480333536863327, "global_step": 182229, "epoch": 2047, "lr": 2.445143614388526e-05}
{"train_loss": 0.08888337016105652, "global_step": 182230, "epoch": 2047, "lr": 2.445093780530181e-05}
{"train_loss": 0.08367027342319489, "global_step": 182231, "epoch": 2047, "lr": 2.445043947015314e-05}
{"train_loss": 0.10098669677972794, "global_step": 182232, "epoch": 2047, "lr": 2.4449941138439302e-05}
{"train_loss": 0.07429995387792587, "global_step": 182233, "epoch": 2047, "lr": 2.4449442810160334e-05}
{"train_loss": 0.060267750173807144, "global_step": 182234, "epoch": 2047, "lr": 2.4448944485316342e-05}
{"train_loss": 0.06626341491937637, "global_step": 182235, "epoch": 2047, "lr": 2.4448446163907367e-05}
{"train_loss": 0.13498815894126892, "global_step": 182236, "epoch": 2047, "lr": 2.4447947845933504e-05}
{"train_loss": 0.06634753197431564, "global_step": 182237, "epoch": 2047, "lr": 2.44474495313948e-05}
{"train_loss": 0.10023587942123413, "global_step": 182238, "epoch": 2047, "lr": 2.444695122029131e-05}
{"train_loss": 0.11251093447208405, "global_step": 182239, "epoch": 2047, "lr": 2.4446452912623136e-05}
{"train_loss": 0.06890901923179626, "global_step": 182240, "epoch": 2047, "lr": 2.4445954608390298e-05}
{"train_loss": 0.03477894887328148, "global_step": 182241, "epoch": 2047, "lr": 2.4445456307592894e-05}
{"train_loss": 0.06258121877908707, "global_step": 182242, "epoch": 2047, "lr": 2.4444958010231012e-05}
{"train_loss": 0.0771736204624176, "global_step": 182243, "epoch": 2047, "lr": 2.4444459716304668e-05}
{"train_loss": 0.06521538645029068, "global_step": 182244, "epoch": 2047, "lr": 2.4443961425813978e-05}
{"train_loss": 0.05185515433549881, "global_step": 182245, "epoch": 2047, "lr": 2.4443463138758958e-05}
{"train_loss": 0.026975084096193314, "global_step": 182246, "epoch": 2047, "lr": 2.444296485513973e-05}
{"train_loss": 0.0621703639626503, "global_step": 182247, "epoch": 2047, "lr": 2.4442466574956314e-05}
{"train_loss": 0.06969799846410751, "global_step": 182248, "epoch": 2047, "lr": 2.4441968298208818e-05}
{"train_loss": 0.0565938837826252, "global_step": 182249, "epoch": 2047, "lr": 2.444147002489726e-05}
{"train_loss": 0.09358882158994675, "global_step": 182250, "epoch": 2047, "lr": 2.4440971755021763e-05}
{"train_loss": 0.07358520478010178, "global_step": 182251, "epoch": 2047, "lr": 2.4440473488582343e-05}
{"train_loss": 0.08075864613056183, "global_step": 182252, "epoch": 2047, "lr": 2.443997522557911e-05}
{"train_loss": 0.08472713083028793, "global_step": 182253, "epoch": 2047, "lr": 2.4439476966012094e-05}
{"train_loss": 0.08973206579685211, "global_step": 182254, "epoch": 2047, "lr": 2.4438978709881394e-05}
{"train_loss": 0.06288015097379684, "global_step": 182255, "epoch": 2047, "lr": 2.443848045718704e-05}
{"train_loss": 0.07046090066432953, "global_step": 182256, "epoch": 2047, "lr": 2.4437982207929143e-05}
{"train_loss": 0.051260825246572495, "global_step": 182257, "epoch": 2047, "lr": 2.443748396210773e-05}
{"train_loss": 0.09246833622455597, "global_step": 182258, "epoch": 2047, "lr": 2.4436985719722888e-05}
{"train_loss": 0.09382346272468567, "global_step": 182259, "epoch": 2047, "lr": 2.4436487480774695e-05}
{"train_loss": 0.06163995340466499, "global_step": 182260, "epoch": 2047, "lr": 2.4435989245263192e-05}
{"train_loss": 0.101960189640522, "global_step": 182261, "epoch": 2047, "lr": 2.443549101318847e-05}
{"train_loss": 0.06833616644144058, "global_step": 182262, "epoch": 2047, "lr": 2.443499278455057e-05}
{"train_loss": 0.10000713914632797, "global_step": 182263, "epoch": 2047, "lr": 2.4434494559349596e-05}
{"train_loss": 0.0899479016661644, "global_step": 182264, "epoch": 2047, "lr": 2.443399633758557e-05}
{"train_loss": 0.09036287665367126, "global_step": 182265, "epoch": 2047, "lr": 2.4433498119258597e-05}
{"train_loss": 0.057082850486040115, "global_step": 182266, "epoch": 2047, "lr": 2.4432999904368714e-05}
{"train_loss": 0.07306089252233505, "global_step": 182267, "epoch": 2047, "lr": 2.443250169291602e-05}
{"train_loss": 0.058899104595184326, "global_step": 182268, "epoch": 2047, "lr": 2.4432003484900545e-05}
{"train_loss": 0.050706375390291214, "global_step": 182269, "epoch": 2047, "lr": 2.4431505280322398e-05}
{"train_loss": 0.04188193008303642, "global_step": 182270, "epoch": 2047, "lr": 2.4431007079181622e-05}
{"train_loss": 0.07379911289539899, "global_step": 182271, "epoch": 2047, "lr": 2.4430508881478265e-05, "val_loss": 7.484377384185791}
{"train_loss": 0.06529854238033295, "global_step": 182272, "epoch": 2048, "lr": 2.4430010687212435e-05}
{"train_loss": 0.1132858544588089, "global_step": 182273, "epoch": 2048, "lr": 2.4429512496384153e-05}
{"train_loss": 0.058791738003492355, "global_step": 182274, "epoch": 2048, "lr": 2.442901430899354e-05}
{"train_loss": 0.08414451032876968, "global_step": 182275, "epoch": 2048, "lr": 2.442851612504063e-05}
{"train_loss": 0.06700330227613449, "global_step": 182276, "epoch": 2048, "lr": 2.4428017944525478e-05}
{"train_loss": 0.0790109857916832, "global_step": 182277, "epoch": 2048, "lr": 2.4427519767448166e-05}
{"train_loss": 0.04732619225978851, "global_step": 182278, "epoch": 2048, "lr": 2.4427021593808774e-05}
{"train_loss": 0.07063058018684387, "global_step": 182279, "epoch": 2048, "lr": 2.4426523423607344e-05}
{"train_loss": 0.14306971430778503, "global_step": 182280, "epoch": 2048, "lr": 2.4426025256843976e-05}
{"train_loss": 0.09566782414913177, "global_step": 182281, "epoch": 2048, "lr": 2.44255270935187e-05}
{"train_loss": 0.06967515498399734, "global_step": 182282, "epoch": 2048, "lr": 2.4425028933631618e-05}
{"train_loss": 0.07721743732690811, "global_step": 182283, "epoch": 2048, "lr": 2.4424530777182757e-05}
{"train_loss": 0.08101806789636612, "global_step": 182284, "epoch": 2048, "lr": 2.442403262417223e-05}
{"train_loss": 0.08646967262029648, "global_step": 182285, "epoch": 2048, "lr": 2.442353447460006e-05}
{"train_loss": 0.10399263352155685, "global_step": 182286, "epoch": 2048, "lr": 2.4423036328466348e-05}
{"train_loss": 0.08861598372459412, "global_step": 182287, "epoch": 2048, "lr": 2.4422538185771133e-05}
{"train_loss": 0.0728098601102829, "global_step": 182288, "epoch": 2048, "lr": 2.442204004651451e-05}
{"train_loss": 0.11005592346191406, "global_step": 182289, "epoch": 2048, "lr": 2.4421541910696515e-05}
{"train_loss": 0.06704441457986832, "global_step": 182290, "epoch": 2048, "lr": 2.4421043778317253e-05}
{"train_loss": 0.052605342119932175, "global_step": 182291, "epoch": 2048, "lr": 2.4420545649376753e-05}
{"train_loss": 0.05325809866189957, "global_step": 182292, "epoch": 2048, "lr": 2.4420047523875118e-05}
{"train_loss": 0.06365301460027695, "global_step": 182293, "epoch": 2048, "lr": 2.441954940181237e-05}
{"train_loss": 0.09098561108112335, "global_step": 182294, "epoch": 2048, "lr": 2.4419051283188604e-05}
{"train_loss": 0.06638415157794952, "global_step": 182295, "epoch": 2048, "lr": 2.441855316800391e-05}
{"train_loss": 0.10184157639741898, "global_step": 182296, "epoch": 2048, "lr": 2.4418055056258305e-05}
{"train_loss": 0.12449860572814941, "global_step": 182297, "epoch": 2048, "lr": 2.4417556947951896e-05}
{"train_loss": 0.07276152819395065, "global_step": 182298, "epoch": 2048, "lr": 2.4417058843084722e-05}
{"train_loss": 0.06560564786195755, "global_step": 182299, "epoch": 2048, "lr": 2.4416560741656878e-05}
{"train_loss": 0.03989506885409355, "global_step": 182300, "epoch": 2048, "lr": 2.44160626436684e-05}
{"train_loss": 0.05595889687538147, "global_step": 182301, "epoch": 2048, "lr": 2.441556454911939e-05}
{"train_loss": 0.07070506364107132, "global_step": 182302, "epoch": 2048, "lr": 2.4415066458009878e-05}
{"train_loss": 0.11699219793081284, "global_step": 182303, "epoch": 2048, "lr": 2.4414568370339963e-05}
{"train_loss": 0.058828335255384445, "global_step": 182304, "epoch": 2048, "lr": 2.441407028610968e-05}
{"train_loss": 0.06271049380302429, "global_step": 182305, "epoch": 2048, "lr": 2.441357220531913e-05}
{"train_loss": 0.11800577491521835, "global_step": 182306, "epoch": 2048, "lr": 2.441307412796835e-05}
{"train_loss": 0.07056628912687302, "global_step": 182307, "epoch": 2048, "lr": 2.4412576054057434e-05}
{"train_loss": 0.12335818260908127, "global_step": 182308, "epoch": 2048, "lr": 2.4412077983586436e-05}
{"train_loss": 0.06088905408978462, "global_step": 182309, "epoch": 2048, "lr": 2.4411579916555403e-05}
{"train_loss": 0.08672166615724564, "global_step": 182310, "epoch": 2048, "lr": 2.4411081852964447e-05}
{"train_loss": 0.07177889347076416, "global_step": 182311, "epoch": 2048, "lr": 2.441058379281358e-05}
{"train_loss": 0.08933953195810318, "global_step": 182312, "epoch": 2048, "lr": 2.441008573610292e-05}
{"train_loss": 0.08505292981863022, "global_step": 182313, "epoch": 2048, "lr": 2.4409587682832495e-05}
{"train_loss": 0.08394629508256912, "global_step": 182314, "epoch": 2048, "lr": 2.4409089633002412e-05}
{"train_loss": 0.07367351651191711, "global_step": 182315, "epoch": 2048, "lr": 2.440859158661269e-05}
{"train_loss": 0.06964091211557388, "global_step": 182316, "epoch": 2048, "lr": 2.440809354366344e-05}
{"train_loss": 0.041088469326496124, "global_step": 182317, "epoch": 2048, "lr": 2.4407595504154694e-05}
{"train_loss": 0.14183491468429565, "global_step": 182318, "epoch": 2048, "lr": 2.4407097468086555e-05}
{"train_loss": 0.13822339475154877, "global_step": 182319, "epoch": 2048, "lr": 2.440659943545905e-05}
{"train_loss": 0.03755376860499382, "global_step": 182320, "epoch": 2048, "lr": 2.440610140627228e-05}
{"train_loss": 0.07562408596277237, "global_step": 182321, "epoch": 2048, "lr": 2.4405603380526282e-05}
{"train_loss": 0.056757811456918716, "global_step": 182322, "epoch": 2048, "lr": 2.4405105358221158e-05}
{"train_loss": 0.04176357761025429, "global_step": 182323, "epoch": 2048, "lr": 2.440460733935694e-05}
{"train_loss": 0.10944759100675583, "global_step": 182324, "epoch": 2048, "lr": 2.440410932393372e-05}
{"train_loss": 0.08846814930438995, "global_step": 182325, "epoch": 2048, "lr": 2.4403611311951546e-05}
{"train_loss": 0.13115300238132477, "global_step": 182326, "epoch": 2048, "lr": 2.4403113303410508e-05}
{"train_loss": 0.07389836758375168, "global_step": 182327, "epoch": 2048, "lr": 2.4402615298310645e-05}
{"train_loss": 0.09899742156267166, "global_step": 182328, "epoch": 2048, "lr": 2.440211729665205e-05}
{"train_loss": 0.058334410190582275, "global_step": 182329, "epoch": 2048, "lr": 2.4401619298434765e-05}
{"train_loss": 0.085523322224617, "global_step": 182330, "epoch": 2048, "lr": 2.440112130365887e-05}
{"train_loss": 0.09370346367359161, "global_step": 182331, "epoch": 2048, "lr": 2.440062331232445e-05}
{"train_loss": 0.06463261693716049, "global_step": 182332, "epoch": 2048, "lr": 2.4400125324431537e-05}
{"train_loss": 0.10873579978942871, "global_step": 182333, "epoch": 2048, "lr": 2.439962733998023e-05}
{"train_loss": 0.043045587837696075, "global_step": 182334, "epoch": 2048, "lr": 2.439912935897057e-05}
{"train_loss": 0.054459623992443085, "global_step": 182335, "epoch": 2048, "lr": 2.4398631381402648e-05}
{"train_loss": 0.14251920580863953, "global_step": 182336, "epoch": 2048, "lr": 2.4398133407276497e-05}
{"train_loss": 0.1119631752371788, "global_step": 182337, "epoch": 2048, "lr": 2.4397635436592227e-05}
{"train_loss": 0.08826804906129837, "global_step": 182338, "epoch": 2048, "lr": 2.4397137469349862e-05}
{"train_loss": 0.0780084878206253, "global_step": 182339, "epoch": 2048, "lr": 2.4396639505549513e-05}
{"train_loss": 0.02369944006204605, "global_step": 182340, "epoch": 2048, "lr": 2.4396141545191202e-05}
{"train_loss": 0.0897630900144577, "global_step": 182341, "epoch": 2048, "lr": 2.4395643588275036e-05}
{"train_loss": 0.059822406619787216, "global_step": 182342, "epoch": 2048, "lr": 2.439514563480105e-05}
{"train_loss": 0.05715300515294075, "global_step": 182343, "epoch": 2048, "lr": 2.4394647684769333e-05}
{"train_loss": 0.051346905529499054, "global_step": 182344, "epoch": 2048, "lr": 2.4394149738179935e-05}
{"train_loss": 0.1145055890083313, "global_step": 182345, "epoch": 2048, "lr": 2.4393651795032947e-05}
{"train_loss": 0.13443320989608765, "global_step": 182346, "epoch": 2048, "lr": 2.43931538553284e-05}
{"train_loss": 0.059207215905189514, "global_step": 182347, "epoch": 2048, "lr": 2.43926559190664e-05}
{"train_loss": 0.14832599461078644, "global_step": 182348, "epoch": 2048, "lr": 2.4392157986246996e-05}
{"train_loss": 0.09791915863752365, "global_step": 182349, "epoch": 2048, "lr": 2.4391660056870235e-05}
{"train_loss": 0.05706210806965828, "global_step": 182350, "epoch": 2048, "lr": 2.4391162130936224e-05}
{"train_loss": 0.06945352256298065, "global_step": 182351, "epoch": 2048, "lr": 2.4390664208444984e-05}
{"train_loss": 0.04954645782709122, "global_step": 182352, "epoch": 2048, "lr": 2.4390166289396632e-05}
{"train_loss": 0.08916127681732178, "global_step": 182353, "epoch": 2048, "lr": 2.438966837379119e-05}
{"train_loss": 0.07824961841106415, "global_step": 182354, "epoch": 2048, "lr": 2.438917046162877e-05}
{"train_loss": 0.06534720212221146, "global_step": 182355, "epoch": 2048, "lr": 2.4388672552909385e-05}
{"train_loss": 0.14683222770690918, "global_step": 182356, "epoch": 2048, "lr": 2.4388174647633155e-05}
{"train_loss": 0.07385789602994919, "global_step": 182357, "epoch": 2048, "lr": 2.4387676745800097e-05}
{"train_loss": 0.08184211701154709, "global_step": 182358, "epoch": 2048, "lr": 2.438717884741033e-05}
{"train_loss": 0.07065219432115555, "global_step": 182359, "epoch": 2048, "lr": 2.4386680952463874e-05}
{"train_loss": 0.08243961879209186, "global_step": 182360, "epoch": 2048, "lr": 2.4386183060960836e-05, "val_loss": 7.734959602355957}
{"train_loss": 0.07096394896507263, "global_step": 182361, "epoch": 2049, "lr": 2.4385685172901246e-05}
{"train_loss": 0.11364786326885223, "global_step": 182362, "epoch": 2049, "lr": 2.4385187288285207e-05}
{"train_loss": 0.07524034380912781, "global_step": 182363, "epoch": 2049, "lr": 2.4384689407112742e-05}
{"train_loss": 0.09717854857444763, "global_step": 182364, "epoch": 2049, "lr": 2.4384191529383965e-05}
{"train_loss": 0.05836757645010948, "global_step": 182365, "epoch": 2049, "lr": 2.4383693655098906e-05}
{"train_loss": 0.05713040009140968, "global_step": 182366, "epoch": 2049, "lr": 2.4383195784257662e-05}
{"train_loss": 0.06324268132448196, "global_step": 182367, "epoch": 2049, "lr": 2.438269791686027e-05}
{"train_loss": 0.05829046666622162, "global_step": 182368, "epoch": 2049, "lr": 2.438220005290681e-05}
{"train_loss": 0.04791424423456192, "global_step": 182369, "epoch": 2049, "lr": 2.438170219239737e-05}
{"train_loss": 0.05921535566449165, "global_step": 182370, "epoch": 2049, "lr": 2.4381204335331982e-05}
{"train_loss": 0.06050161272287369, "global_step": 182371, "epoch": 2049, "lr": 2.4380706481710742e-05}
{"train_loss": 0.05615047365427017, "global_step": 182372, "epoch": 2049, "lr": 2.438020863153369e-05}
{"train_loss": 0.07813990116119385, "global_step": 182373, "epoch": 2049, "lr": 2.4379710784800925e-05}
{"train_loss": 0.1080755814909935, "global_step": 182374, "epoch": 2049, "lr": 2.4379212941512474e-05}
{"train_loss": 0.10500451177358627, "global_step": 182375, "epoch": 2049, "lr": 2.437871510166844e-05}
{"train_loss": 0.07572677731513977, "global_step": 182376, "epoch": 2049, "lr": 2.4378217265268865e-05}
{"train_loss": 0.13878273963928223, "global_step": 182377, "epoch": 2049, "lr": 2.4377719432313843e-05}
{"train_loss": 0.07183206081390381, "global_step": 182378, "epoch": 2049, "lr": 2.4377221602803406e-05}
{"train_loss": 0.09101524949073792, "global_step": 182379, "epoch": 2049, "lr": 2.4376723776737652e-05}
{"train_loss": 0.0808735340833664, "global_step": 182380, "epoch": 2049, "lr": 2.437622595411663e-05}
{"train_loss": 0.0888131782412529, "global_step": 182381, "epoch": 2049, "lr": 2.4375728134940417e-05}
{"train_loss": 0.026447800919413567, "global_step": 182382, "epoch": 2049, "lr": 2.4375230319209064e-05}
{"train_loss": 0.08968240767717361, "global_step": 182383, "epoch": 2049, "lr": 2.4374732506922666e-05}
{"train_loss": 0.07461234927177429, "global_step": 182384, "epoch": 2049, "lr": 2.437423469808125e-05}
{"train_loss": 0.06057222560048103, "global_step": 182385, "epoch": 2049, "lr": 2.4373736892684928e-05}
{"train_loss": 0.06273223459720612, "global_step": 182386, "epoch": 2049, "lr": 2.4373239090733745e-05}
{"train_loss": 0.13662375509738922, "global_step": 182387, "epoch": 2049, "lr": 2.4372741292227746e-05}
{"train_loss": 0.05224572494626045, "global_step": 182388, "epoch": 2049, "lr": 2.4372243497167047e-05}
{"train_loss": 0.11602912843227386, "global_step": 182389, "epoch": 2049, "lr": 2.4371745705551657e-05}
{"train_loss": 0.08216355741024017, "global_step": 182390, "epoch": 2049, "lr": 2.43712479173817e-05}
{"train_loss": 0.07280724495649338, "global_step": 182391, "epoch": 2049, "lr": 2.4370750132657195e-05}
{"train_loss": 0.0821804329752922, "global_step": 182392, "epoch": 2049, "lr": 2.4370252351378247e-05}
{"train_loss": 0.05991493910551071, "global_step": 182393, "epoch": 2049, "lr": 2.4369754573544897e-05}
{"train_loss": 0.04255897179245949, "global_step": 182394, "epoch": 2049, "lr": 2.436925679915723e-05}
{"train_loss": 0.12760518491268158, "global_step": 182395, "epoch": 2049, "lr": 2.4368759028215288e-05}
{"train_loss": 0.06062939763069153, "global_step": 182396, "epoch": 2049, "lr": 2.4368261260719172e-05}
{"train_loss": 0.05748101323843002, "global_step": 182397, "epoch": 2049, "lr": 2.436776349666891e-05}
{"train_loss": 0.09690853953361511, "global_step": 182398, "epoch": 2049, "lr": 2.4367265736064615e-05}
{"train_loss": 0.060333721339702606, "global_step": 182399, "epoch": 2049, "lr": 2.436676797890631e-05}
{"train_loss": 0.08119598031044006, "global_step": 182400, "epoch": 2049, "lr": 2.4366270225194092e-05}
{"train_loss": 0.12357644736766815, "global_step": 182401, "epoch": 2049, "lr": 2.4365772474928e-05}
{"train_loss": 0.08627770841121674, "global_step": 182402, "epoch": 2049, "lr": 2.436527472810814e-05}
{"train_loss": 0.06304040551185608, "global_step": 182403, "epoch": 2049, "lr": 2.4364776984734533e-05}
{"train_loss": 0.17106936872005463, "global_step": 182404, "epoch": 2049, "lr": 2.436427924480727e-05}
{"train_loss": 0.10538457334041595, "global_step": 182405, "epoch": 2049, "lr": 2.4363781508326437e-05}
{"train_loss": 0.10284753888845444, "global_step": 182406, "epoch": 2049, "lr": 2.4363283775292067e-05}
{"train_loss": 0.06710364669561386, "global_step": 182407, "epoch": 2049, "lr": 2.4362786045704256e-05}
{"train_loss": 0.11078231781721115, "global_step": 182408, "epoch": 2049, "lr": 2.4362288319563037e-05}
{"train_loss": 0.06361876428127289, "global_step": 182409, "epoch": 2049, "lr": 2.4361790596868512e-05}
{"train_loss": 0.049542807042598724, "global_step": 182410, "epoch": 2049, "lr": 2.4361292877620718e-05}
{"train_loss": 0.06099623069167137, "global_step": 182411, "epoch": 2049, "lr": 2.4360795161819748e-05}
{"train_loss": 0.14616012573242188, "global_step": 182412, "epoch": 2049, "lr": 2.4360297449465643e-05}
{"train_loss": 0.03879069536924362, "global_step": 182413, "epoch": 2049, "lr": 2.4359799740558504e-05}
{"train_loss": 0.07785534858703613, "global_step": 182414, "epoch": 2049, "lr": 2.4359302035098353e-05}
{"train_loss": 0.03621545061469078, "global_step": 182415, "epoch": 2049, "lr": 2.43588043330853e-05}
{"train_loss": 0.0753021314740181, "global_step": 182416, "epoch": 2049, "lr": 2.4358306634519377e-05}
{"train_loss": 0.0628049224615097, "global_step": 182417, "epoch": 2049, "lr": 2.435780893940069e-05}
{"train_loss": 0.05859865993261337, "global_step": 182418, "epoch": 2049, "lr": 2.4357311247729258e-05}
{"train_loss": 0.12319199740886688, "global_step": 182419, "epoch": 2049, "lr": 2.4356813559505193e-05}
{"train_loss": 0.11518062651157379, "global_step": 182420, "epoch": 2049, "lr": 2.4356315874728526e-05}
{"train_loss": 0.054989252239465714, "global_step": 182421, "epoch": 2049, "lr": 2.435581819339936e-05}
{"train_loss": 0.10256512463092804, "global_step": 182422, "epoch": 2049, "lr": 2.4355320515517715e-05}
{"train_loss": 0.07430823892354965, "global_step": 182423, "epoch": 2049, "lr": 2.435482284108371e-05}
{"train_loss": 0.05281662940979004, "global_step": 182424, "epoch": 2049, "lr": 2.4354325170097385e-05}
{"train_loss": 0.05646209791302681, "global_step": 182425, "epoch": 2049, "lr": 2.435382750255879e-05}
{"train_loss": 0.06815502047538757, "global_step": 182426, "epoch": 2049, "lr": 2.4353329838468027e-05}
{"train_loss": 0.07536840438842773, "global_step": 182427, "epoch": 2049, "lr": 2.435283217782513e-05}
{"train_loss": 0.06562245637178421, "global_step": 182428, "epoch": 2049, "lr": 2.43523345206302e-05}
{"train_loss": 0.0933266133069992, "global_step": 182429, "epoch": 2049, "lr": 2.4351836866883264e-05}
{"train_loss": 0.09610165655612946, "global_step": 182430, "epoch": 2049, "lr": 2.4351339216584436e-05}
{"train_loss": 0.11586713790893555, "global_step": 182431, "epoch": 2049, "lr": 2.4350841569733733e-05}
{"train_loss": 0.06207766383886337, "global_step": 182432, "epoch": 2049, "lr": 2.435034392633127e-05}
{"train_loss": 0.12376274913549423, "global_step": 182433, "epoch": 2049, "lr": 2.434984628637707e-05}
{"train_loss": 0.10277017205953598, "global_step": 182434, "epoch": 2049, "lr": 2.434934864987124e-05}
{"train_loss": 0.06612105667591095, "global_step": 182435, "epoch": 2049, "lr": 2.4348851016813807e-05}
{"train_loss": 0.12548255920410156, "global_step": 182436, "epoch": 2049, "lr": 2.4348353387204876e-05}
{"train_loss": 0.06427796185016632, "global_step": 182437, "epoch": 2049, "lr": 2.434785576104448e-05}
{"train_loss": 0.12242477387189865, "global_step": 182438, "epoch": 2049, "lr": 2.434735813833272e-05}
{"train_loss": 0.06303689628839493, "global_step": 182439, "epoch": 2049, "lr": 2.4346860519069626e-05}
{"train_loss": 0.06420315057039261, "global_step": 182440, "epoch": 2049, "lr": 2.4346362903255288e-05}
{"train_loss": 0.07501876354217529, "global_step": 182441, "epoch": 2049, "lr": 2.434586529088978e-05}
{"train_loss": 0.054786503314971924, "global_step": 182442, "epoch": 2049, "lr": 2.4345367681973147e-05}
{"train_loss": 0.06875230371952057, "global_step": 182443, "epoch": 2049, "lr": 2.4344870076505482e-05}
{"train_loss": 0.057960011065006256, "global_step": 182444, "epoch": 2049, "lr": 2.4344372474486814e-05}
{"train_loss": 0.09831308573484421, "global_step": 182445, "epoch": 2049, "lr": 2.4343874875917257e-05}
{"train_loss": 0.09080516546964645, "global_step": 182446, "epoch": 2049, "lr": 2.434337728079683e-05}
{"train_loss": 0.06333476305007935, "global_step": 182447, "epoch": 2049, "lr": 2.4342879689125646e-05}
{"train_loss": 0.06610700488090515, "global_step": 182448, "epoch": 2049, "lr": 2.4342382100903728e-05}
{"train_loss": 0.07989384863902343, "global_step": 182449, "epoch": 2049, "lr": 2.4341884516131186e-05, "val_loss": 7.5584235191345215}
{"train_loss": 0.09937689453363419, "global_step": 182450, "epoch": 2050, "lr": 2.4341386934808043e-05}
{"train_loss": 0.07844610512256622, "global_step": 182451, "epoch": 2050, "lr": 2.4340889356934404e-05}
{"train_loss": 0.0827007070183754, "global_step": 182452, "epoch": 2050, "lr": 2.43403917825103e-05}
{"train_loss": 0.04979664459824562, "global_step": 182453, "epoch": 2050, "lr": 2.4339894211535847e-05}
{"train_loss": 0.06666579842567444, "global_step": 182454, "epoch": 2050, "lr": 2.4339396644011054e-05}
{"train_loss": 0.1335502564907074, "global_step": 182455, "epoch": 2050, "lr": 2.433889907993604e-05}
{"train_loss": 0.12348013371229172, "global_step": 182456, "epoch": 2050, "lr": 2.4338401519310828e-05}
{"train_loss": 0.07385873049497604, "global_step": 182457, "epoch": 2050, "lr": 2.43379039621355e-05}
{"train_loss": 0.07766850292682648, "global_step": 182458, "epoch": 2050, "lr": 2.433740640841015e-05}
{"train_loss": 0.06903967261314392, "global_step": 182459, "epoch": 2050, "lr": 2.4336908858134828e-05}
{"train_loss": 0.07047475874423981, "global_step": 182460, "epoch": 2050, "lr": 2.4336411311309575e-05}
{"train_loss": 0.051126282662153244, "global_step": 182461, "epoch": 2050, "lr": 2.4335913767934494e-05}
{"train_loss": 0.07606221735477448, "global_step": 182462, "epoch": 2050, "lr": 2.4335416228009622e-05}
{"train_loss": 0.09131667017936707, "global_step": 182463, "epoch": 2050, "lr": 2.433491869153505e-05}
{"train_loss": 0.03868497908115387, "global_step": 182464, "epoch": 2050, "lr": 2.4334421158510845e-05}
{"train_loss": 0.051774390041828156, "global_step": 182465, "epoch": 2050, "lr": 2.4333923628937044e-05}
{"train_loss": 0.02850339747965336, "global_step": 182466, "epoch": 2050, "lr": 2.4333426102813755e-05}
{"train_loss": 0.11670473217964172, "global_step": 182467, "epoch": 2050, "lr": 2.4332928580141003e-05}
{"train_loss": 0.09167220443487167, "global_step": 182468, "epoch": 2050, "lr": 2.4332431060918893e-05}
{"train_loss": 0.059447865933179855, "global_step": 182469, "epoch": 2050, "lr": 2.4331933545147457e-05}
{"train_loss": 0.055049579590559006, "global_step": 182470, "epoch": 2050, "lr": 2.4331436032826793e-05}
{"train_loss": 0.06395892798900604, "global_step": 182471, "epoch": 2050, "lr": 2.433093852395694e-05}
{"train_loss": 0.04215685650706291, "global_step": 182472, "epoch": 2050, "lr": 2.4330441018538003e-05}
{"train_loss": 0.04181564971804619, "global_step": 182473, "epoch": 2050, "lr": 2.4329943516570002e-05}
{"train_loss": 0.07451990991830826, "global_step": 182474, "epoch": 2050, "lr": 2.4329446018053047e-05}
{"train_loss": 0.07676467299461365, "global_step": 182475, "epoch": 2050, "lr": 2.4328948522987166e-05}
{"train_loss": 0.09419537335634232, "global_step": 182476, "epoch": 2050, "lr": 2.4328451031372445e-05}
{"train_loss": 0.10022971034049988, "global_step": 182477, "epoch": 2050, "lr": 2.432795354320897e-05}
{"train_loss": 0.06937770545482635, "global_step": 182478, "epoch": 2050, "lr": 2.4327456058496766e-05}
{"train_loss": 0.08635400235652924, "global_step": 182479, "epoch": 2050, "lr": 2.4326958577235946e-05}
{"train_loss": 0.09064806997776031, "global_step": 182480, "epoch": 2050, "lr": 2.4326461099426535e-05}
{"train_loss": 0.08831442147493362, "global_step": 182481, "epoch": 2050, "lr": 2.4325963625068637e-05}
{"train_loss": 0.10903163254261017, "global_step": 182482, "epoch": 2050, "lr": 2.432546615416228e-05}
{"train_loss": 0.08377705514431, "global_step": 182483, "epoch": 2050, "lr": 2.432496868670757e-05}
{"train_loss": 0.09145410358905792, "global_step": 182484, "epoch": 2050, "lr": 2.4324471222704536e-05}
{"train_loss": 0.058102864772081375, "global_step": 182485, "epoch": 2050, "lr": 2.4323973762153286e-05}
{"train_loss": 0.0673779845237732, "global_step": 182486, "epoch": 2050, "lr": 2.4323476305053843e-05}
{"train_loss": 0.05167091637849808, "global_step": 182487, "epoch": 2050, "lr": 2.4322978851406314e-05}
{"train_loss": 0.08341099321842194, "global_step": 182488, "epoch": 2050, "lr": 2.432248140121073e-05}
{"train_loss": 0.06185774505138397, "global_step": 182489, "epoch": 2050, "lr": 2.4321983954467193e-05}
{"train_loss": 0.10314913094043732, "global_step": 182490, "epoch": 2050, "lr": 2.4321486511175733e-05}
{"train_loss": 0.05395686626434326, "global_step": 182491, "epoch": 2050, "lr": 2.4320989071336457e-05}
{"train_loss": 0.0850600078701973, "global_step": 182492, "epoch": 2050, "lr": 2.4320491634949393e-05}
{"train_loss": 0.07706335186958313, "global_step": 182493, "epoch": 2050, "lr": 2.431999420201464e-05}
{"train_loss": 0.09732213616371155, "global_step": 182494, "epoch": 2050, "lr": 2.4319496772532234e-05}
{"train_loss": 0.09455133229494095, "global_step": 182495, "epoch": 2050, "lr": 2.4318999346502262e-05}
{"train_loss": 0.1080595999956131, "global_step": 182496, "epoch": 2050, "lr": 2.4318501923924803e-05}
{"train_loss": 0.1028698980808258, "global_step": 182497, "epoch": 2050, "lr": 2.4318004504799914e-05}
{"train_loss": 0.07661449164152145, "global_step": 182498, "epoch": 2050, "lr": 2.4317507089127627e-05}
{"train_loss": 0.037580300122499466, "global_step": 182499, "epoch": 2050, "lr": 2.4317009676908066e-05}
{"train_loss": 0.05336170643568039, "global_step": 182500, "epoch": 2050, "lr": 2.4316512268141252e-05}
{"train_loss": 0.10532938688993454, "global_step": 182501, "epoch": 2050, "lr": 2.431601486282728e-05}
{"train_loss": 0.044669702649116516, "global_step": 182502, "epoch": 2050, "lr": 2.4315517460966214e-05}
{"train_loss": 0.09277438372373581, "global_step": 182503, "epoch": 2050, "lr": 2.431502006255809e-05}
{"train_loss": 0.0755322128534317, "global_step": 182504, "epoch": 2050, "lr": 2.4314522667603016e-05}
{"train_loss": 0.11591633409261703, "global_step": 182505, "epoch": 2050, "lr": 2.431402527610103e-05}
{"train_loss": 0.08344424515962601, "global_step": 182506, "epoch": 2050, "lr": 2.4313527888052223e-05}
{"train_loss": 0.08380020409822464, "global_step": 182507, "epoch": 2050, "lr": 2.4313030503456636e-05}
{"train_loss": 0.1019194945693016, "global_step": 182508, "epoch": 2050, "lr": 2.431253312231436e-05}
{"train_loss": 0.12027213722467422, "global_step": 182509, "epoch": 2050, "lr": 2.431203574462544e-05}
{"train_loss": 0.08797954767942429, "global_step": 182510, "epoch": 2050, "lr": 2.431153837038997e-05}
{"train_loss": 0.061589565128088, "global_step": 182511, "epoch": 2050, "lr": 2.431104099960798e-05}
{"train_loss": 0.13097967207431793, "global_step": 182512, "epoch": 2050, "lr": 2.431054363227956e-05}
{"train_loss": 0.13088901340961456, "global_step": 182513, "epoch": 2050, "lr": 2.431004626840479e-05}
{"train_loss": 0.04262983053922653, "global_step": 182514, "epoch": 2050, "lr": 2.4309548907983703e-05}
{"train_loss": 0.08564737439155579, "global_step": 182515, "epoch": 2050, "lr": 2.4309051551016403e-05}
{"train_loss": 0.06446560472249985, "global_step": 182516, "epoch": 2050, "lr": 2.430855419750292e-05}
{"train_loss": 0.06431546062231064, "global_step": 182517, "epoch": 2050, "lr": 2.4308056847443357e-05}
{"train_loss": 0.0592275895178318, "global_step": 182518, "epoch": 2050, "lr": 2.4307559500837746e-05}
{"train_loss": 0.09419794380664825, "global_step": 182519, "epoch": 2050, "lr": 2.4307062157686184e-05}
{"train_loss": 0.036507830023765564, "global_step": 182520, "epoch": 2050, "lr": 2.430656481798871e-05}
{"train_loss": 0.06110960990190506, "global_step": 182521, "epoch": 2050, "lr": 2.430606748174542e-05}
{"train_loss": 0.0670960545539856, "global_step": 182522, "epoch": 2050, "lr": 2.4305570148956352e-05}
{"train_loss": 0.08177868276834488, "global_step": 182523, "epoch": 2050, "lr": 2.4305072819621605e-05}
{"train_loss": 0.05360474810004234, "global_step": 182524, "epoch": 2050, "lr": 2.4304575493741206e-05}
{"train_loss": 0.11308456212282181, "global_step": 182525, "epoch": 2050, "lr": 2.4304078171315263e-05}
{"train_loss": 0.09905144572257996, "global_step": 182526, "epoch": 2050, "lr": 2.4303580852343804e-05}
{"train_loss": 0.08656873553991318, "global_step": 182527, "epoch": 2050, "lr": 2.430308353682694e-05}
{"train_loss": 0.07694140821695328, "global_step": 182528, "epoch": 2050, "lr": 2.430258622476469e-05}
{"train_loss": 0.10108058154582977, "global_step": 182529, "epoch": 2050, "lr": 2.430208891615716e-05}
{"train_loss": 0.06979706138372421, "global_step": 182530, "epoch": 2050, "lr": 2.4301591611004387e-05}
{"train_loss": 0.07769303768873215, "global_step": 182531, "epoch": 2050, "lr": 2.4301094309306448e-05}
{"train_loss": 0.10174616426229477, "global_step": 182532, "epoch": 2050, "lr": 2.4300597011063437e-05}
{"train_loss": 0.08362489193677902, "global_step": 182533, "epoch": 2050, "lr": 2.4300099716275375e-05}
{"train_loss": 0.07209791243076324, "global_step": 182534, "epoch": 2050, "lr": 2.4299602424942375e-05}
{"train_loss": 0.0760839432477951, "global_step": 182535, "epoch": 2050, "lr": 2.4299105137064476e-05}
{"train_loss": 0.0783141702413559, "global_step": 182536, "epoch": 2050, "lr": 2.429860785264173e-05}
{"train_loss": 0.08600004762411118, "global_step": 182537, "epoch": 2050, "lr": 2.4298110571674243e-05}
{"train_loss": 0.07919551873726122, "global_step": 182538, "epoch": 2050, "lr": 2.4297613294162042e-05, "train/sim_max_reward_0": 0.6508298684193763, "train/sim_max_reward_1": 0.9469841851233062, "train/sim_max_reward_2": 0.3779566629299147, "train/sim_max_reward_3": 1.0, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.4805925682928884, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.5107716802997784, "test/sim_max_reward_4500004": 0.25040345570814543, "test/sim_max_reward_4500005": 0.5747861223275983, "test/sim_max_reward_4500006": 0.9702216194839706, "test/sim_max_reward_4500007": 0.8106745137282422, "test/sim_max_reward_4500008": 0.9982274539445523, "test/sim_max_reward_4500009": 0.04948105118961793, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5287907255316422, "test/sim_max_reward_4500014": 0.6191648767763258, "test/sim_max_reward_4500015": 0.9844962820148856, "test/sim_max_reward_4500016": 0.9929103157849165, "test/sim_max_reward_4500017": 0.9888330013453995, "test/sim_max_reward_4500018": 0.9237533760963552, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9984497659746848, "test/sim_max_reward_4500023": 0.608493111487377, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.31268901185981507, "test/sim_max_reward_4500026": 0.9098268418344164, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.9209883689962562, "test/sim_max_reward_4500029": 0.9823063172045569, "test/sim_max_reward_4500030": 0.9226810291492273, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.9680794506953356, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.9801291924482061, "test/sim_max_reward_4500035": 1.0, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.9994008616259561, "test/sim_max_reward_4500038": 0.1623496196960387, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9381577190274472, "test/sim_max_reward_4500041": 0.4807986599472945, "test/sim_max_reward_4500042": 2.5590191870492765e-05, "test/sim_max_reward_4500043": 0.9980186503696558, "test/sim_max_reward_4500044": 0.7252473586535824, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.6370822465988579, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9796600545801253, "test/sim_max_reward_4500049": 0.9852358674318333, "train/mean_score": 0.8292951194120995, "test/mean_score": 0.7038545352059371, "val_loss": 7.52870512008667, "train_action_mse_error": 18.505088806152344}
{"train_loss": 0.11799086630344391, "global_step": 182539, "epoch": 2051, "lr": 2.429711602010523e-05}
{"train_loss": 0.08433621376752853, "global_step": 182540, "epoch": 2051, "lr": 2.4296618749503862e-05}
{"train_loss": 0.05106291547417641, "global_step": 182541, "epoch": 2051, "lr": 2.4296121482357982e-05}
{"train_loss": 0.0635698214173317, "global_step": 182542, "epoch": 2051, "lr": 2.429562421866769e-05}
{"train_loss": 0.03523227572441101, "global_step": 182543, "epoch": 2051, "lr": 2.4295126958433025e-05}
{"train_loss": 0.1265977919101715, "global_step": 182544, "epoch": 2051, "lr": 2.4294629701654083e-05}
{"train_loss": 0.123143270611763, "global_step": 182545, "epoch": 2051, "lr": 2.4294132448330893e-05}
{"train_loss": 0.046887967735528946, "global_step": 182546, "epoch": 2051, "lr": 2.429363519846356e-05}
{"train_loss": 0.05531531199812889, "global_step": 182547, "epoch": 2051, "lr": 2.4293137952052118e-05}
{"train_loss": 0.07759446650743484, "global_step": 182548, "epoch": 2051, "lr": 2.4292640709096652e-05}
{"train_loss": 0.09371444582939148, "global_step": 182549, "epoch": 2051, "lr": 2.429214346959725e-05}
{"train_loss": 0.05207034945487976, "global_step": 182550, "epoch": 2051, "lr": 2.429164623355393e-05}
{"train_loss": 0.0849081426858902, "global_step": 182551, "epoch": 2051, "lr": 2.4291149000966813e-05}
{"train_loss": 0.1197393387556076, "global_step": 182552, "epoch": 2051, "lr": 2.429065177183591e-05}
{"train_loss": 0.03266028314828873, "global_step": 182553, "epoch": 2051, "lr": 2.4290154546161336e-05}
{"train_loss": 0.07118062674999237, "global_step": 182554, "epoch": 2051, "lr": 2.428965732394312e-05}
{"train_loss": 0.06878196448087692, "global_step": 182555, "epoch": 2051, "lr": 2.4289160105181365e-05}
{"train_loss": 0.06732481718063354, "global_step": 182556, "epoch": 2051, "lr": 2.42886628898761e-05}
{"train_loss": 0.07877930998802185, "global_step": 182557, "epoch": 2051, "lr": 2.428816567802743e-05}
{"train_loss": 0.060251835733652115, "global_step": 182558, "epoch": 2051, "lr": 2.4287668469635382e-05}
{"train_loss": 0.04719613119959831, "global_step": 182559, "epoch": 2051, "lr": 2.4287171264700064e-05}
{"train_loss": 0.06998154520988464, "global_step": 182560, "epoch": 2051, "lr": 2.4286674063221504e-05}
{"train_loss": 0.07417399436235428, "global_step": 182561, "epoch": 2051, "lr": 2.4286176865199804e-05}
{"train_loss": 0.08870749920606613, "global_step": 182562, "epoch": 2051, "lr": 2.4285679670634998e-05}
{"train_loss": 0.05532200634479523, "global_step": 182563, "epoch": 2051, "lr": 2.4285182479527185e-05}
{"train_loss": 0.1185382679104805, "global_step": 182564, "epoch": 2051, "lr": 2.4284685291876404e-05}
{"train_loss": 0.05375487729907036, "global_step": 182565, "epoch": 2051, "lr": 2.4284188107682744e-05}
{"train_loss": 0.09031883627176285, "global_step": 182566, "epoch": 2051, "lr": 2.4283690926946246e-05}
{"train_loss": 0.08612160384654999, "global_step": 182567, "epoch": 2051, "lr": 2.4283193749666994e-05}
{"train_loss": 0.12673288583755493, "global_step": 182568, "epoch": 2051, "lr": 2.4282696575845066e-05}
{"train_loss": 0.09255865961313248, "global_step": 182569, "epoch": 2051, "lr": 2.4282199405480506e-05}
{"train_loss": 0.044276848435401917, "global_step": 182570, "epoch": 2051, "lr": 2.42817022385734e-05}
{"train_loss": 0.06603587418794632, "global_step": 182571, "epoch": 2051, "lr": 2.4281205075123793e-05}
{"train_loss": 0.06053483113646507, "global_step": 182572, "epoch": 2051, "lr": 2.4280707915131783e-05}
{"train_loss": 0.07895531505346298, "global_step": 182573, "epoch": 2051, "lr": 2.42802107585974e-05}
{"train_loss": 0.07017426192760468, "global_step": 182574, "epoch": 2051, "lr": 2.4279713605520743e-05}
{"train_loss": 0.04701267555356026, "global_step": 182575, "epoch": 2051, "lr": 2.427921645590187e-05}
{"train_loss": 0.0247790589928627, "global_step": 182576, "epoch": 2051, "lr": 2.4278719309740823e-05}
{"train_loss": 0.03810380399227142, "global_step": 182577, "epoch": 2051, "lr": 2.4278222167037707e-05}
{"train_loss": 0.08282918483018875, "global_step": 182578, "epoch": 2051, "lr": 2.4277725027792552e-05}
{"train_loss": 0.04402271285653114, "global_step": 182579, "epoch": 2051, "lr": 2.427722789200546e-05}
{"train_loss": 0.04868237301707268, "global_step": 182580, "epoch": 2051, "lr": 2.4276730759676487e-05}
{"train_loss": 0.05062361806631088, "global_step": 182581, "epoch": 2051, "lr": 2.4276233630805666e-05}
{"train_loss": 0.07401837408542633, "global_step": 182582, "epoch": 2051, "lr": 2.427573650539312e-05}
{"train_loss": 0.05266030505299568, "global_step": 182583, "epoch": 2051, "lr": 2.4275239383438863e-05}
{"train_loss": 0.09524435549974442, "global_step": 182584, "epoch": 2051, "lr": 2.427474226494299e-05}
{"train_loss": 0.04499834030866623, "global_step": 182585, "epoch": 2051, "lr": 2.4274245149905582e-05}
{"train_loss": 0.08364084362983704, "global_step": 182586, "epoch": 2051, "lr": 2.427374803832667e-05}
{"train_loss": 0.06144283339381218, "global_step": 182587, "epoch": 2051, "lr": 2.4273250930206355e-05}
{"train_loss": 0.07321711629629135, "global_step": 182588, "epoch": 2051, "lr": 2.427275382554467e-05}
{"train_loss": 0.0341862291097641, "global_step": 182589, "epoch": 2051, "lr": 2.427225672434172e-05}
{"train_loss": 0.06936506927013397, "global_step": 182590, "epoch": 2051, "lr": 2.4271759626597534e-05}
{"train_loss": 0.054431334137916565, "global_step": 182591, "epoch": 2051, "lr": 2.427126253231221e-05}
{"train_loss": 0.04458723962306976, "global_step": 182592, "epoch": 2051, "lr": 2.427076544148579e-05}
{"train_loss": 0.07436636835336685, "global_step": 182593, "epoch": 2051, "lr": 2.4270268354118362e-05}
{"train_loss": 0.05675140768289566, "global_step": 182594, "epoch": 2051, "lr": 2.4269771270209975e-05}
{"train_loss": 0.12781645357608795, "global_step": 182595, "epoch": 2051, "lr": 2.426927418976071e-05}
{"train_loss": 0.06457091867923737, "global_step": 182596, "epoch": 2051, "lr": 2.426877711277062e-05}
{"train_loss": 0.04643823206424713, "global_step": 182597, "epoch": 2051, "lr": 2.426828003923979e-05}
{"train_loss": 0.04848982393741608, "global_step": 182598, "epoch": 2051, "lr": 2.4267782969168257e-05}
{"train_loss": 0.048357464373111725, "global_step": 182599, "epoch": 2051, "lr": 2.426728590255613e-05}
{"train_loss": 0.12905630469322205, "global_step": 182600, "epoch": 2051, "lr": 2.426678883940343e-05}
{"train_loss": 0.07498564571142197, "global_step": 182601, "epoch": 2051, "lr": 2.426629177971027e-05}
{"train_loss": 0.08646605163812637, "global_step": 182602, "epoch": 2051, "lr": 2.4265794723476676e-05}
{"train_loss": 0.07024693489074707, "global_step": 182603, "epoch": 2051, "lr": 2.4265297670702725e-05}
{"train_loss": 0.0535893440246582, "global_step": 182604, "epoch": 2051, "lr": 2.426480062138852e-05}
{"train_loss": 0.03587789461016655, "global_step": 182605, "epoch": 2051, "lr": 2.426430357553407e-05}
{"train_loss": 0.06912095099687576, "global_step": 182606, "epoch": 2051, "lr": 2.4263806533139494e-05}
{"train_loss": 0.08849577605724335, "global_step": 182607, "epoch": 2051, "lr": 2.4263309494204815e-05}
{"train_loss": 0.033759526908397675, "global_step": 182608, "epoch": 2051, "lr": 2.4262812458730137e-05}
{"train_loss": 0.05129548907279968, "global_step": 182609, "epoch": 2051, "lr": 2.4262315426715498e-05}
{"train_loss": 0.12367706000804901, "global_step": 182610, "epoch": 2051, "lr": 2.426181839816099e-05}
{"train_loss": 0.04042532294988632, "global_step": 182611, "epoch": 2051, "lr": 2.4261321373066654e-05}
{"train_loss": 0.059555891901254654, "global_step": 182612, "epoch": 2051, "lr": 2.4260824351432583e-05}
{"train_loss": 0.11504064500331879, "global_step": 182613, "epoch": 2051, "lr": 2.426032733325883e-05}
{"train_loss": 0.028436385095119476, "global_step": 182614, "epoch": 2051, "lr": 2.4259830318545452e-05}
{"train_loss": 0.039229247719049454, "global_step": 182615, "epoch": 2051, "lr": 2.4259333307292535e-05}
{"train_loss": 0.06516947597265244, "global_step": 182616, "epoch": 2051, "lr": 2.425883629950012e-05}
{"train_loss": 0.0805353969335556, "global_step": 182617, "epoch": 2051, "lr": 2.4258339295168313e-05}
{"train_loss": 0.04130122810602188, "global_step": 182618, "epoch": 2051, "lr": 2.4257842294297157e-05}
{"train_loss": 0.08474915474653244, "global_step": 182619, "epoch": 2051, "lr": 2.4257345296886704e-05}
{"train_loss": 0.05396019294857979, "global_step": 182620, "epoch": 2051, "lr": 2.4256848302937054e-05}
{"train_loss": 0.02739539183676243, "global_step": 182621, "epoch": 2051, "lr": 2.4256351312448235e-05}
{"train_loss": 0.0327083021402359, "global_step": 182622, "epoch": 2051, "lr": 2.4255854325420335e-05}
{"train_loss": 0.039729516953229904, "global_step": 182623, "epoch": 2051, "lr": 2.4255357341853445e-05}
{"train_loss": 0.1051800474524498, "global_step": 182624, "epoch": 2051, "lr": 2.425486036174759e-05}
{"train_loss": 0.06872598081827164, "global_step": 182625, "epoch": 2051, "lr": 2.4254363385102874e-05}
{"train_loss": 0.07452508062124252, "global_step": 182626, "epoch": 2051, "lr": 2.4253866411919324e-05}
{"train_loss": 0.06830383274243788, "global_step": 182627, "epoch": 2051, "lr": 2.425336944219705e-05, "val_loss": 7.705561637878418}
{"train_loss": 0.02560257539153099, "global_step": 182628, "epoch": 2052, "lr": 2.4252872475936073e-05}
{"train_loss": 0.06261539459228516, "global_step": 182629, "epoch": 2052, "lr": 2.4252375513136503e-05}
{"train_loss": 0.05147900804877281, "global_step": 182630, "epoch": 2052, "lr": 2.4251878553798368e-05}
{"train_loss": 0.07747966796159744, "global_step": 182631, "epoch": 2052, "lr": 2.4251381597921778e-05}
{"train_loss": 0.09579899162054062, "global_step": 182632, "epoch": 2052, "lr": 2.4250884645506755e-05}
{"train_loss": 0.05247179791331291, "global_step": 182633, "epoch": 2052, "lr": 2.42503876965534e-05}
{"train_loss": 0.10763920098543167, "global_step": 182634, "epoch": 2052, "lr": 2.4249890751061755e-05}
{"train_loss": 0.03865756466984749, "global_step": 182635, "epoch": 2052, "lr": 2.424939380903191e-05}
{"train_loss": 0.0580289363861084, "global_step": 182636, "epoch": 2052, "lr": 2.424889687046391e-05}
{"train_loss": 0.11005106568336487, "global_step": 182637, "epoch": 2052, "lr": 2.4248399935357847e-05}
{"train_loss": 0.06498237699270248, "global_step": 182638, "epoch": 2052, "lr": 2.4247903003713752e-05}
{"train_loss": 0.050272464752197266, "global_step": 182639, "epoch": 2052, "lr": 2.4247406075531714e-05}
{"train_loss": 0.05245353281497955, "global_step": 182640, "epoch": 2052, "lr": 2.424690915081182e-05}
{"train_loss": 0.14051097631454468, "global_step": 182641, "epoch": 2052, "lr": 2.424641222955409e-05}
{"train_loss": 0.014912783168256283, "global_step": 182642, "epoch": 2052, "lr": 2.4245915311758637e-05}
{"train_loss": 0.06787453591823578, "global_step": 182643, "epoch": 2052, "lr": 2.424541839742549e-05}
{"train_loss": 0.04189487174153328, "global_step": 182644, "epoch": 2052, "lr": 2.424492148655475e-05}
{"train_loss": 0.08446778357028961, "global_step": 182645, "epoch": 2052, "lr": 2.4244424579146452e-05}
{"train_loss": 0.06995871663093567, "global_step": 182646, "epoch": 2052, "lr": 2.424392767520069e-05}
{"train_loss": 0.13566561043262482, "global_step": 182647, "epoch": 2052, "lr": 2.42434307747175e-05}
{"train_loss": 0.0804811492562294, "global_step": 182648, "epoch": 2052, "lr": 2.4242933877696988e-05}
{"train_loss": 0.06713682413101196, "global_step": 182649, "epoch": 2052, "lr": 2.4242436984139184e-05}
{"train_loss": 0.0914258360862732, "global_step": 182650, "epoch": 2052, "lr": 2.4241940094044186e-05}
{"train_loss": 0.03507096692919731, "global_step": 182651, "epoch": 2052, "lr": 2.424144320741204e-05}
{"train_loss": 0.12721560895442963, "global_step": 182652, "epoch": 2052, "lr": 2.4240946324242802e-05}
{"train_loss": 0.05721598118543625, "global_step": 182653, "epoch": 2052, "lr": 2.424044944453658e-05}
{"train_loss": 0.0418572798371315, "global_step": 182654, "epoch": 2052, "lr": 2.423995256829339e-05}
{"train_loss": 0.058397404849529266, "global_step": 182655, "epoch": 2052, "lr": 2.4239455695513347e-05}
{"train_loss": 0.10837069153785706, "global_step": 182656, "epoch": 2052, "lr": 2.4238958826196473e-05}
{"train_loss": 0.09302525967359543, "global_step": 182657, "epoch": 2052, "lr": 2.4238461960342877e-05}
{"train_loss": 0.09139960259199142, "global_step": 182658, "epoch": 2052, "lr": 2.4237965097952586e-05}
{"train_loss": 0.048853542655706406, "global_step": 182659, "epoch": 2052, "lr": 2.4237468239025707e-05}
{"train_loss": 0.059012316167354584, "global_step": 182660, "epoch": 2052, "lr": 2.4236971383562268e-05}
{"train_loss": 0.07972205430269241, "global_step": 182661, "epoch": 2052, "lr": 2.4236474531562377e-05}
{"train_loss": 0.05930249020457268, "global_step": 182662, "epoch": 2052, "lr": 2.423597768302605e-05}
{"train_loss": 0.1028512716293335, "global_step": 182663, "epoch": 2052, "lr": 2.4235480837953405e-05}
{"train_loss": 0.08717257529497147, "global_step": 182664, "epoch": 2052, "lr": 2.4234983996344467e-05}
{"train_loss": 0.07765887677669525, "global_step": 182665, "epoch": 2052, "lr": 2.423448715819934e-05}
{"train_loss": 0.10927777737379074, "global_step": 182666, "epoch": 2052, "lr": 2.4233990323518052e-05}
{"train_loss": 0.10659726709127426, "global_step": 182667, "epoch": 2052, "lr": 2.4233493492300714e-05}
{"train_loss": 0.06812715530395508, "global_step": 182668, "epoch": 2052, "lr": 2.423299666454734e-05}
{"train_loss": 0.0562361478805542, "global_step": 182669, "epoch": 2052, "lr": 2.4232499840258055e-05}
{"train_loss": 0.07216975092887878, "global_step": 182670, "epoch": 2052, "lr": 2.423200301943287e-05}
{"train_loss": 0.037700846791267395, "global_step": 182671, "epoch": 2052, "lr": 2.4231506202071903e-05}
{"train_loss": 0.05138825625181198, "global_step": 182672, "epoch": 2052, "lr": 2.4231009388175173e-05}
{"train_loss": 0.05020928755402565, "global_step": 182673, "epoch": 2052, "lr": 2.4230512577742793e-05}
{"train_loss": 0.11543263494968414, "global_step": 182674, "epoch": 2052, "lr": 2.4230015770774782e-05}
{"train_loss": 0.05696023628115654, "global_step": 182675, "epoch": 2052, "lr": 2.422951896727123e-05}
{"train_loss": 0.06544086337089539, "global_step": 182676, "epoch": 2052, "lr": 2.422902216723223e-05}
{"train_loss": 0.06507570296525955, "global_step": 182677, "epoch": 2052, "lr": 2.42285253706578e-05}
{"train_loss": 0.0488838367164135, "global_step": 182678, "epoch": 2052, "lr": 2.4228028577548055e-05}
{"train_loss": 0.07415597885847092, "global_step": 182679, "epoch": 2052, "lr": 2.4227531787903013e-05}
{"train_loss": 0.08287401497364044, "global_step": 182680, "epoch": 2052, "lr": 2.4227035001722786e-05}
{"train_loss": 0.046990182250738144, "global_step": 182681, "epoch": 2052, "lr": 2.4226538219007405e-05}
{"train_loss": 0.04638640582561493, "global_step": 182682, "epoch": 2052, "lr": 2.422604143975697e-05}
{"train_loss": 0.12503737211227417, "global_step": 182683, "epoch": 2052, "lr": 2.4225544663971506e-05}
{"train_loss": 0.04341144859790802, "global_step": 182684, "epoch": 2052, "lr": 2.4225047891651127e-05}
{"train_loss": 0.0432913638651371, "global_step": 182685, "epoch": 2052, "lr": 2.422455112279585e-05}
{"train_loss": 0.1010468453168869, "global_step": 182686, "epoch": 2052, "lr": 2.422405435740579e-05}
{"train_loss": 0.04480988532304764, "global_step": 182687, "epoch": 2052, "lr": 2.422355759548098e-05}
{"train_loss": 0.04911961406469345, "global_step": 182688, "epoch": 2052, "lr": 2.422306083702151e-05}
{"train_loss": 0.040483806282281876, "global_step": 182689, "epoch": 2052, "lr": 2.4222564082027415e-05}
{"train_loss": 0.07742777466773987, "global_step": 182690, "epoch": 2052, "lr": 2.4222067330498805e-05}
{"train_loss": 0.04735822230577469, "global_step": 182691, "epoch": 2052, "lr": 2.4221570582435728e-05}
{"train_loss": 0.11029276251792908, "global_step": 182692, "epoch": 2052, "lr": 2.4221073837838222e-05}
{"train_loss": 0.07578542083501816, "global_step": 182693, "epoch": 2052, "lr": 2.4220577096706392e-05}
{"train_loss": 0.10394404083490372, "global_step": 182694, "epoch": 2052, "lr": 2.4220080359040282e-05}
{"train_loss": 0.04482165351510048, "global_step": 182695, "epoch": 2052, "lr": 2.421958362483998e-05}
{"train_loss": 0.12489472329616547, "global_step": 182696, "epoch": 2052, "lr": 2.4219086894105524e-05}
{"train_loss": 0.09980566799640656, "global_step": 182697, "epoch": 2052, "lr": 2.4218590166837018e-05}
{"train_loss": 0.09609724581241608, "global_step": 182698, "epoch": 2052, "lr": 2.421809344303449e-05}
{"train_loss": 0.05756537988781929, "global_step": 182699, "epoch": 2052, "lr": 2.4217596722698044e-05}
{"train_loss": 0.062134530395269394, "global_step": 182700, "epoch": 2052, "lr": 2.4217100005827708e-05}
{"train_loss": 0.0830056294798851, "global_step": 182701, "epoch": 2052, "lr": 2.421660329242359e-05}
{"train_loss": 0.0885181650519371, "global_step": 182702, "epoch": 2052, "lr": 2.4216106582485716e-05}
{"train_loss": 0.0675380751490593, "global_step": 182703, "epoch": 2052, "lr": 2.4215609876014183e-05}
{"train_loss": 0.05295826122164726, "global_step": 182704, "epoch": 2052, "lr": 2.4215113173009034e-05}
{"train_loss": 0.14240308105945587, "global_step": 182705, "epoch": 2052, "lr": 2.421461647347037e-05}
{"train_loss": 0.041735485196113586, "global_step": 182706, "epoch": 2052, "lr": 2.4214119777398214e-05}
{"train_loss": 0.13366207480430603, "global_step": 182707, "epoch": 2052, "lr": 2.421362308479268e-05}
{"train_loss": 0.07957903295755386, "global_step": 182708, "epoch": 2052, "lr": 2.4213126395653784e-05}
{"train_loss": 0.08214817196130753, "global_step": 182709, "epoch": 2052, "lr": 2.4212629709981643e-05}
{"train_loss": 0.0642385482788086, "global_step": 182710, "epoch": 2052, "lr": 2.421213302777628e-05}
{"train_loss": 0.0524006225168705, "global_step": 182711, "epoch": 2052, "lr": 2.4211636349037773e-05}
{"train_loss": 0.07887087762355804, "global_step": 182712, "epoch": 2052, "lr": 2.4211139673766225e-05}
{"train_loss": 0.04651380702853203, "global_step": 182713, "epoch": 2052, "lr": 2.421064300196166e-05}
{"train_loss": 0.09163229167461395, "global_step": 182714, "epoch": 2052, "lr": 2.4210146333624167e-05}
{"train_loss": 0.08208663016557693, "global_step": 182715, "epoch": 2052, "lr": 2.4209649668753793e-05}
{"train_loss": 0.0739247484886077, "global_step": 182716, "epoch": 2052, "lr": 2.4209153007350637e-05, "val_loss": 7.5860819816589355}
{"train_loss": 0.03903814032673836, "global_step": 182717, "epoch": 2053, "lr": 2.4208656349414725e-05}
{"train_loss": 0.054772038012742996, "global_step": 182718, "epoch": 2053, "lr": 2.4208159694946163e-05}
{"train_loss": 0.07032699137926102, "global_step": 182719, "epoch": 2053, "lr": 2.4207663043944985e-05}
{"train_loss": 0.048304878175258636, "global_step": 182720, "epoch": 2053, "lr": 2.4207166396411285e-05}
{"train_loss": 0.09139373153448105, "global_step": 182721, "epoch": 2053, "lr": 2.42066697523451e-05}
{"train_loss": 0.043360330164432526, "global_step": 182722, "epoch": 2053, "lr": 2.4206173111746534e-05}
{"train_loss": 0.06535553187131882, "global_step": 182723, "epoch": 2053, "lr": 2.420567647461562e-05}
{"train_loss": 0.04435613751411438, "global_step": 182724, "epoch": 2053, "lr": 2.4205179840952446e-05}
{"train_loss": 0.03721275553107262, "global_step": 182725, "epoch": 2053, "lr": 2.4204683210757063e-05}
{"train_loss": 0.1121424213051796, "global_step": 182726, "epoch": 2053, "lr": 2.420418658402956e-05}
{"train_loss": 0.07468284666538239, "global_step": 182727, "epoch": 2053, "lr": 2.420368996076997e-05}
{"train_loss": 0.07600010931491852, "global_step": 182728, "epoch": 2053, "lr": 2.4203193340978398e-05}
{"train_loss": 0.11172429472208023, "global_step": 182729, "epoch": 2053, "lr": 2.4202696724654894e-05}
{"train_loss": 0.11603273451328278, "global_step": 182730, "epoch": 2053, "lr": 2.4202200111799506e-05}
{"train_loss": 0.08512689918279648, "global_step": 182731, "epoch": 2053, "lr": 2.420170350241233e-05}
{"train_loss": 0.08663538843393326, "global_step": 182732, "epoch": 2053, "lr": 2.4201206896493404e-05}
{"train_loss": 0.0746799185872078, "global_step": 182733, "epoch": 2053, "lr": 2.420071029404284e-05}
{"train_loss": 0.08591669797897339, "global_step": 182734, "epoch": 2053, "lr": 2.420021369506065e-05}
{"train_loss": 0.08067500591278076, "global_step": 182735, "epoch": 2053, "lr": 2.4199717099546942e-05}
{"train_loss": 0.07382522523403168, "global_step": 182736, "epoch": 2053, "lr": 2.419922050750175e-05}
{"train_loss": 0.12856446206569672, "global_step": 182737, "epoch": 2053, "lr": 2.4198723918925183e-05}
{"train_loss": 0.07297838479280472, "global_step": 182738, "epoch": 2053, "lr": 2.419822733381726e-05}
{"train_loss": 0.052811894565820694, "global_step": 182739, "epoch": 2053, "lr": 2.4197730752178087e-05}
{"train_loss": 0.08255238085985184, "global_step": 182740, "epoch": 2053, "lr": 2.41972341740077e-05}
{"train_loss": 0.046261996030807495, "global_step": 182741, "epoch": 2053, "lr": 2.4196737599306198e-05}
{"train_loss": 0.02818957157433033, "global_step": 182742, "epoch": 2053, "lr": 2.419624102807361e-05}
{"train_loss": 0.07304781675338745, "global_step": 182743, "epoch": 2053, "lr": 2.4195744460310043e-05}
{"train_loss": 0.055891551077365875, "global_step": 182744, "epoch": 2053, "lr": 2.419524789601552e-05}
{"train_loss": 0.0781761109828949, "global_step": 182745, "epoch": 2053, "lr": 2.4194751335190158e-05}
{"train_loss": 0.06275065988302231, "global_step": 182746, "epoch": 2053, "lr": 2.4194254777833975e-05}
{"train_loss": 0.06471291929483414, "global_step": 182747, "epoch": 2053, "lr": 2.4193758223947082e-05}
{"train_loss": 0.08992242068052292, "global_step": 182748, "epoch": 2053, "lr": 2.41932616735295e-05}
{"train_loss": 0.06387975811958313, "global_step": 182749, "epoch": 2053, "lr": 2.4192765126581323e-05}
{"train_loss": 0.07405515760183334, "global_step": 182750, "epoch": 2053, "lr": 2.419226858310263e-05}
{"train_loss": 0.03504950925707817, "global_step": 182751, "epoch": 2053, "lr": 2.4191772043093453e-05}
{"train_loss": 0.03571150079369545, "global_step": 182752, "epoch": 2053, "lr": 2.41912755065539e-05}
{"train_loss": 0.07192113250494003, "global_step": 182753, "epoch": 2053, "lr": 2.419077897348399e-05}
{"train_loss": 0.08593806624412537, "global_step": 182754, "epoch": 2053, "lr": 2.4190282443883844e-05}
{"train_loss": 0.10653319954872131, "global_step": 182755, "epoch": 2053, "lr": 2.4189785917753472e-05}
{"train_loss": 0.09853152185678482, "global_step": 182756, "epoch": 2053, "lr": 2.4189289395092994e-05}
{"train_loss": 0.06209462136030197, "global_step": 182757, "epoch": 2053, "lr": 2.4188792875902428e-05}
{"train_loss": 0.10607679188251495, "global_step": 182758, "epoch": 2053, "lr": 2.4188296360181885e-05}
{"train_loss": 0.05307013913989067, "global_step": 182759, "epoch": 2053, "lr": 2.418779984793139e-05}
{"train_loss": 0.06374938040971756, "global_step": 182760, "epoch": 2053, "lr": 2.4187303339151053e-05}
{"train_loss": 0.08769838511943817, "global_step": 182761, "epoch": 2053, "lr": 2.4186806833840897e-05}
{"train_loss": 0.07050921022891998, "global_step": 182762, "epoch": 2053, "lr": 2.418631033200103e-05}
{"train_loss": 0.10992205888032913, "global_step": 182763, "epoch": 2053, "lr": 2.4185813833631484e-05}
{"train_loss": 0.13583964109420776, "global_step": 182764, "epoch": 2053, "lr": 2.4185317338732355e-05}
{"train_loss": 0.09349571913480759, "global_step": 182765, "epoch": 2053, "lr": 2.418482084730368e-05}
{"train_loss": 0.04952120780944824, "global_step": 182766, "epoch": 2053, "lr": 2.4184324359345556e-05}
{"train_loss": 0.07151919603347778, "global_step": 182767, "epoch": 2053, "lr": 2.4183827874858035e-05}
{"train_loss": 0.07412949204444885, "global_step": 182768, "epoch": 2053, "lr": 2.418333139384117e-05}
{"train_loss": 0.06616844236850739, "global_step": 182769, "epoch": 2053, "lr": 2.4182834916295056e-05}
{"train_loss": 0.074571892619133, "global_step": 182770, "epoch": 2053, "lr": 2.4182338442219728e-05}
{"train_loss": 0.14525246620178223, "global_step": 182771, "epoch": 2053, "lr": 2.4181841971615283e-05}
{"train_loss": 0.06949523836374283, "global_step": 182772, "epoch": 2053, "lr": 2.4181345504481763e-05}
{"train_loss": 0.03799410164356232, "global_step": 182773, "epoch": 2053, "lr": 2.4180849040819258e-05}
{"train_loss": 0.06466979533433914, "global_step": 182774, "epoch": 2053, "lr": 2.4180352580627812e-05}
{"train_loss": 0.04587700590491295, "global_step": 182775, "epoch": 2053, "lr": 2.417985612390752e-05}
{"train_loss": 0.13869257271289825, "global_step": 182776, "epoch": 2053, "lr": 2.417935967065841e-05}
{"train_loss": 0.03228376805782318, "global_step": 182777, "epoch": 2053, "lr": 2.4178863220880594e-05}
{"train_loss": 0.102004773914814, "global_step": 182778, "epoch": 2053, "lr": 2.417836677457409e-05}
{"train_loss": 0.043282292783260345, "global_step": 182779, "epoch": 2053, "lr": 2.417787033173901e-05}
{"train_loss": 0.09829071164131165, "global_step": 182780, "epoch": 2053, "lr": 2.417737389237539e-05}
{"train_loss": 0.03826332092285156, "global_step": 182781, "epoch": 2053, "lr": 2.4176877456483317e-05}
{"train_loss": 0.07571793347597122, "global_step": 182782, "epoch": 2053, "lr": 2.4176381024062826e-05}
{"train_loss": 0.07949231564998627, "global_step": 182783, "epoch": 2053, "lr": 2.4175884595114035e-05}
{"train_loss": 0.07348600775003433, "global_step": 182784, "epoch": 2053, "lr": 2.4175388169636958e-05}
{"train_loss": 0.06563518941402435, "global_step": 182785, "epoch": 2053, "lr": 2.4174891747631683e-05}
{"train_loss": 0.10667236149311066, "global_step": 182786, "epoch": 2053, "lr": 2.4174395329098304e-05}
{"train_loss": 0.12157178670167923, "global_step": 182787, "epoch": 2053, "lr": 2.417389891403684e-05}
{"train_loss": 0.04981500655412674, "global_step": 182788, "epoch": 2053, "lr": 2.41734025024474e-05}
{"train_loss": 0.06412544846534729, "global_step": 182789, "epoch": 2053, "lr": 2.417290609433001e-05}
{"train_loss": 0.12496308237314224, "global_step": 182790, "epoch": 2053, "lr": 2.4172409689684784e-05}
{"train_loss": 0.10544677078723907, "global_step": 182791, "epoch": 2053, "lr": 2.4171913288511738e-05}
{"train_loss": 0.08479789644479752, "global_step": 182792, "epoch": 2053, "lr": 2.4171416890810982e-05}
{"train_loss": 0.07335123419761658, "global_step": 182793, "epoch": 2053, "lr": 2.4170920496582543e-05}
{"train_loss": 0.09905743598937988, "global_step": 182794, "epoch": 2053, "lr": 2.4170424105826535e-05}
{"train_loss": 0.043363917618989944, "global_step": 182795, "epoch": 2053, "lr": 2.416992771854298e-05}
{"train_loss": 0.1127762645483017, "global_step": 182796, "epoch": 2053, "lr": 2.4169431334731972e-05}
{"train_loss": 0.0732184499502182, "global_step": 182797, "epoch": 2053, "lr": 2.4168934954393558e-05}
{"train_loss": 0.06139114871621132, "global_step": 182798, "epoch": 2053, "lr": 2.4168438577527834e-05}
{"train_loss": 0.09054896980524063, "global_step": 182799, "epoch": 2053, "lr": 2.416794220413483e-05}
{"train_loss": 0.06735460460186005, "global_step": 182800, "epoch": 2053, "lr": 2.4167445834214647e-05}
{"train_loss": 0.05538281798362732, "global_step": 182801, "epoch": 2053, "lr": 2.416694946776732e-05}
{"train_loss": 0.05264955013990402, "global_step": 182802, "epoch": 2053, "lr": 2.4166453104792952e-05}
{"train_loss": 0.09766960144042969, "global_step": 182803, "epoch": 2053, "lr": 2.416595674529157e-05}
{"train_loss": 0.10234721750020981, "global_step": 182804, "epoch": 2053, "lr": 2.416546038926328e-05}
{"train_loss": 0.07557070445729776, "global_step": 182805, "epoch": 2053, "lr": 2.4164964036708105e-05, "val_loss": 7.61749792098999}
{"train_loss": 0.07107280939817429, "global_step": 182806, "epoch": 2054, "lr": 2.4164467687626156e-05}
{"train_loss": 0.06543946266174316, "global_step": 182807, "epoch": 2054, "lr": 2.4163971342017483e-05}
{"train_loss": 0.027529025450348854, "global_step": 182808, "epoch": 2054, "lr": 2.416347499988213e-05}
{"train_loss": 0.0811644196510315, "global_step": 182809, "epoch": 2054, "lr": 2.41629786612202e-05}
{"train_loss": 0.03378161042928696, "global_step": 182810, "epoch": 2054, "lr": 2.416248232603172e-05}
{"train_loss": 0.1097988709807396, "global_step": 182811, "epoch": 2054, "lr": 2.4161985994316804e-05}
{"train_loss": 0.019367210566997528, "global_step": 182812, "epoch": 2054, "lr": 2.416148966607547e-05}
{"train_loss": 0.0732058510184288, "global_step": 182813, "epoch": 2054, "lr": 2.4160993341307835e-05}
{"train_loss": 0.08849121630191803, "global_step": 182814, "epoch": 2054, "lr": 2.4160497020013917e-05}
{"train_loss": 0.050819482654333115, "global_step": 182815, "epoch": 2054, "lr": 2.4160000702193826e-05}
{"train_loss": 0.05133279412984848, "global_step": 182816, "epoch": 2054, "lr": 2.415950438784758e-05}
{"train_loss": 0.09154205024242401, "global_step": 182817, "epoch": 2054, "lr": 2.4159008076975304e-05}
{"train_loss": 0.07477415353059769, "global_step": 182818, "epoch": 2054, "lr": 2.4158511769577008e-05}
{"train_loss": 0.07922296226024628, "global_step": 182819, "epoch": 2054, "lr": 2.4158015465652804e-05}
{"train_loss": 0.037055786699056625, "global_step": 182820, "epoch": 2054, "lr": 2.415751916520273e-05}
{"train_loss": 0.06857682764530182, "global_step": 182821, "epoch": 2054, "lr": 2.4157022868226853e-05}
{"train_loss": 0.043856579810380936, "global_step": 182822, "epoch": 2054, "lr": 2.4156526574725274e-05}
{"train_loss": 0.0963912382721901, "global_step": 182823, "epoch": 2054, "lr": 2.415603028469801e-05}
{"train_loss": 0.048389725387096405, "global_step": 182824, "epoch": 2054, "lr": 2.415553399814518e-05}
{"train_loss": 0.07284420728683472, "global_step": 182825, "epoch": 2054, "lr": 2.4155037715066796e-05}
{"train_loss": 0.046878632158041, "global_step": 182826, "epoch": 2054, "lr": 2.4154541435462975e-05}
{"train_loss": 0.07050754129886627, "global_step": 182827, "epoch": 2054, "lr": 2.4154045159333744e-05}
{"train_loss": 0.11316043138504028, "global_step": 182828, "epoch": 2054, "lr": 2.41535488866792e-05}
{"train_loss": 0.09975775331258774, "global_step": 182829, "epoch": 2054, "lr": 2.4153052617499385e-05}
{"train_loss": 0.09430034458637238, "global_step": 182830, "epoch": 2054, "lr": 2.4152556351794392e-05}
{"train_loss": 0.07607779651880264, "global_step": 182831, "epoch": 2054, "lr": 2.4152060089564253e-05}
{"train_loss": 0.04396665468811989, "global_step": 182832, "epoch": 2054, "lr": 2.415156383080907e-05}
{"train_loss": 0.048396997153759, "global_step": 182833, "epoch": 2054, "lr": 2.4151067575528885e-05}
{"train_loss": 0.09376141428947449, "global_step": 182834, "epoch": 2054, "lr": 2.4150571323723788e-05}
{"train_loss": 0.11206592619419098, "global_step": 182835, "epoch": 2054, "lr": 2.415007507539381e-05}
{"train_loss": 0.043853290379047394, "global_step": 182836, "epoch": 2054, "lr": 2.414957883053906e-05}
{"train_loss": 0.047195158898830414, "global_step": 182837, "epoch": 2054, "lr": 2.4149082589159567e-05}
{"train_loss": 0.0734761506319046, "global_step": 182838, "epoch": 2054, "lr": 2.4148586351255408e-05}
{"train_loss": 0.05100039765238762, "global_step": 182839, "epoch": 2054, "lr": 2.4148090116826683e-05}
{"train_loss": 0.09739921241998672, "global_step": 182840, "epoch": 2054, "lr": 2.4147593885873426e-05}
{"train_loss": 0.07461582869291306, "global_step": 182841, "epoch": 2054, "lr": 2.4147097658395694e-05}
{"train_loss": 0.05744754523038864, "global_step": 182842, "epoch": 2054, "lr": 2.4146601434393596e-05}
{"train_loss": 0.06759018450975418, "global_step": 182843, "epoch": 2054, "lr": 2.414610521386714e-05}
{"train_loss": 0.11639717221260071, "global_step": 182844, "epoch": 2054, "lr": 2.4145608996816453e-05}
{"train_loss": 0.09762568771839142, "global_step": 182845, "epoch": 2054, "lr": 2.4145112783241574e-05}
{"train_loss": 0.06265740841627121, "global_step": 182846, "epoch": 2054, "lr": 2.4144616573142546e-05}
{"train_loss": 0.1292884349822998, "global_step": 182847, "epoch": 2054, "lr": 2.414412036651948e-05}
{"train_loss": 0.08613942563533783, "global_step": 182848, "epoch": 2054, "lr": 2.4143624163372404e-05}
{"train_loss": 0.09860201179981232, "global_step": 182849, "epoch": 2054, "lr": 2.414312796370142e-05}
{"train_loss": 0.07621538639068604, "global_step": 182850, "epoch": 2054, "lr": 2.414263176750656e-05}
{"train_loss": 0.08333060890436172, "global_step": 182851, "epoch": 2054, "lr": 2.414213557478792e-05}
{"train_loss": 0.041805338114500046, "global_step": 182852, "epoch": 2054, "lr": 2.4141639385545545e-05}
{"train_loss": 0.07350005209445953, "global_step": 182853, "epoch": 2054, "lr": 2.4141143199779527e-05}
{"train_loss": 0.0535353384912014, "global_step": 182854, "epoch": 2054, "lr": 2.4140647017489898e-05}
{"train_loss": 0.05752665549516678, "global_step": 182855, "epoch": 2054, "lr": 2.414015083867676e-05}
{"train_loss": 0.04990767687559128, "global_step": 182856, "epoch": 2054, "lr": 2.413965466334015e-05}
{"train_loss": 0.1080758273601532, "global_step": 182857, "epoch": 2054, "lr": 2.4139158491480147e-05}
{"train_loss": 0.08746381103992462, "global_step": 182858, "epoch": 2054, "lr": 2.4138662323096834e-05}
{"train_loss": 0.09684805572032928, "global_step": 182859, "epoch": 2054, "lr": 2.4138166158190246e-05}
{"train_loss": 0.08673368394374847, "global_step": 182860, "epoch": 2054, "lr": 2.4137669996760488e-05}
{"train_loss": 0.06729884445667267, "global_step": 182861, "epoch": 2054, "lr": 2.413717383880758e-05}
{"train_loss": 0.16655442118644714, "global_step": 182862, "epoch": 2054, "lr": 2.413667768433164e-05}
{"train_loss": 0.0826038271188736, "global_step": 182863, "epoch": 2054, "lr": 2.4136181533332685e-05}
{"train_loss": 0.06596380472183228, "global_step": 182864, "epoch": 2054, "lr": 2.4135685385810825e-05}
{"train_loss": 0.04281521588563919, "global_step": 182865, "epoch": 2054, "lr": 2.413518924176609e-05}
{"train_loss": 0.04656830057501793, "global_step": 182866, "epoch": 2054, "lr": 2.4134693101198576e-05}
{"train_loss": 0.06512076407670975, "global_step": 182867, "epoch": 2054, "lr": 2.4134196964108326e-05}
{"train_loss": 0.08104795962572098, "global_step": 182868, "epoch": 2054, "lr": 2.413370083049543e-05}
{"train_loss": 0.027633730322122574, "global_step": 182869, "epoch": 2054, "lr": 2.4133204700359928e-05}
{"train_loss": 0.08512462675571442, "global_step": 182870, "epoch": 2054, "lr": 2.413270857370192e-05}
{"train_loss": 0.02723628655076027, "global_step": 182871, "epoch": 2054, "lr": 2.4132212450521434e-05}
{"train_loss": 0.03845416381955147, "global_step": 182872, "epoch": 2054, "lr": 2.4131716330818575e-05}
{"train_loss": 0.08251642435789108, "global_step": 182873, "epoch": 2054, "lr": 2.4131220214593376e-05}
{"train_loss": 0.043718304485082626, "global_step": 182874, "epoch": 2054, "lr": 2.4130724101845935e-05}
{"train_loss": 0.05203375220298767, "global_step": 182875, "epoch": 2054, "lr": 2.4130227992576286e-05}
{"train_loss": 0.09004926681518555, "global_step": 182876, "epoch": 2054, "lr": 2.4129731886784514e-05}
{"train_loss": 0.0738857090473175, "global_step": 182877, "epoch": 2054, "lr": 2.4129235784470692e-05}
{"train_loss": 0.06187813729047775, "global_step": 182878, "epoch": 2054, "lr": 2.4128739685634893e-05}
{"train_loss": 0.1099284216761589, "global_step": 182879, "epoch": 2054, "lr": 2.4128243590277144e-05}
{"train_loss": 0.04331740736961365, "global_step": 182880, "epoch": 2054, "lr": 2.412774749839756e-05}
{"train_loss": 0.08315424621105194, "global_step": 182881, "epoch": 2054, "lr": 2.412725140999616e-05}
{"train_loss": 0.0667264461517334, "global_step": 182882, "epoch": 2054, "lr": 2.412675532507306e-05}
{"train_loss": 0.08373311161994934, "global_step": 182883, "epoch": 2054, "lr": 2.4126259243628303e-05}
{"train_loss": 0.057517390698194504, "global_step": 182884, "epoch": 2054, "lr": 2.412576316566193e-05}
{"train_loss": 0.06782118231058121, "global_step": 182885, "epoch": 2054, "lr": 2.4125267091174058e-05}
{"train_loss": 0.11783547699451447, "global_step": 182886, "epoch": 2054, "lr": 2.4124771020164704e-05}
{"train_loss": 0.06738080829381943, "global_step": 182887, "epoch": 2054, "lr": 2.412427495263398e-05}
{"train_loss": 0.11013802140951157, "global_step": 182888, "epoch": 2054, "lr": 2.4123778888581917e-05}
{"train_loss": 0.052321288734674454, "global_step": 182889, "epoch": 2054, "lr": 2.4123282828008608e-05}
{"train_loss": 0.05637241527438164, "global_step": 182890, "epoch": 2054, "lr": 2.4122786770914095e-05}
{"train_loss": 0.17358916997909546, "global_step": 182891, "epoch": 2054, "lr": 2.4122290717298478e-05}
{"train_loss": 0.05924142897129059, "global_step": 182892, "epoch": 2054, "lr": 2.4121794667161778e-05}
{"train_loss": 0.0782112404704094, "global_step": 182893, "epoch": 2054, "lr": 2.4121298620504096e-05}
{"train_loss": 0.0730963031436955, "global_step": 182894, "epoch": 2054, "lr": 2.4120802577325504e-05, "val_loss": 7.754778861999512}
{"train_loss": 0.07460076361894608, "global_step": 182895, "epoch": 2055, "lr": 2.4120306537626035e-05}
{"train_loss": 0.06110716238617897, "global_step": 182896, "epoch": 2055, "lr": 2.4119810501405798e-05}
{"train_loss": 0.08841075748205185, "global_step": 182897, "epoch": 2055, "lr": 2.411931446866481e-05}
{"train_loss": 0.08016368746757507, "global_step": 182898, "epoch": 2055, "lr": 2.411881843940319e-05}
{"train_loss": 0.06192367523908615, "global_step": 182899, "epoch": 2055, "lr": 2.4118322413620963e-05}
{"train_loss": 0.10905832052230835, "global_step": 182900, "epoch": 2055, "lr": 2.4117826391318226e-05}
{"train_loss": 0.15136601030826569, "global_step": 182901, "epoch": 2055, "lr": 2.4117330372495016e-05}
{"train_loss": 0.07829087972640991, "global_step": 182902, "epoch": 2055, "lr": 2.4116834357151434e-05}
{"train_loss": 0.04504771903157234, "global_step": 182903, "epoch": 2055, "lr": 2.4116338345287508e-05}
{"train_loss": 0.043807223439216614, "global_step": 182904, "epoch": 2055, "lr": 2.4115842336903345e-05}
{"train_loss": 0.13274826109409332, "global_step": 182905, "epoch": 2055, "lr": 2.4115346331998967e-05}
{"train_loss": 0.0616125762462616, "global_step": 182906, "epoch": 2055, "lr": 2.411485033057449e-05}
{"train_loss": 0.04038892313838005, "global_step": 182907, "epoch": 2055, "lr": 2.4114354332629935e-05}
{"train_loss": 0.02696625329554081, "global_step": 182908, "epoch": 2055, "lr": 2.4113858338165414e-05}
{"train_loss": 0.10076989978551865, "global_step": 182909, "epoch": 2055, "lr": 2.411336234718094e-05}
{"train_loss": 0.12420905381441116, "global_step": 182910, "epoch": 2055, "lr": 2.4112866359676635e-05}
{"train_loss": 0.025711270049214363, "global_step": 182911, "epoch": 2055, "lr": 2.411237037565252e-05}
{"train_loss": 0.10028333961963654, "global_step": 182912, "epoch": 2055, "lr": 2.411187439510868e-05}
{"train_loss": 0.08576525747776031, "global_step": 182913, "epoch": 2055, "lr": 2.41113784180452e-05}
{"train_loss": 0.0904831811785698, "global_step": 182914, "epoch": 2055, "lr": 2.411088244446211e-05}
{"train_loss": 0.04855082929134369, "global_step": 182915, "epoch": 2055, "lr": 2.4110386474359525e-05}
{"train_loss": 0.07995955646038055, "global_step": 182916, "epoch": 2055, "lr": 2.4109890507737455e-05}
{"train_loss": 0.030961455777287483, "global_step": 182917, "epoch": 2055, "lr": 2.4109394544596015e-05}
{"train_loss": 0.05529460310935974, "global_step": 182918, "epoch": 2055, "lr": 2.410889858493525e-05}
{"train_loss": 0.10935121029615402, "global_step": 182919, "epoch": 2055, "lr": 2.410840262875521e-05}
{"train_loss": 0.061914991587400436, "global_step": 182920, "epoch": 2055, "lr": 2.4107906676056003e-05}
{"train_loss": 0.0728236585855484, "global_step": 182921, "epoch": 2055, "lr": 2.4107410726837647e-05}
{"train_loss": 0.11944221705198288, "global_step": 182922, "epoch": 2055, "lr": 2.4106914781100258e-05}
{"train_loss": 0.10067544132471085, "global_step": 182923, "epoch": 2055, "lr": 2.410641883884388e-05}
{"train_loss": 0.042582862079143524, "global_step": 182924, "epoch": 2055, "lr": 2.4105922900068556e-05}
{"train_loss": 0.1809728443622589, "global_step": 182925, "epoch": 2055, "lr": 2.4105426964774392e-05}
{"train_loss": 0.1123277097940445, "global_step": 182926, "epoch": 2055, "lr": 2.4104931032961418e-05}
{"train_loss": 0.09007901698350906, "global_step": 182927, "epoch": 2055, "lr": 2.4104435104629746e-05}
{"train_loss": 0.04572254419326782, "global_step": 182928, "epoch": 2055, "lr": 2.4103939179779394e-05}
{"train_loss": 0.11072321236133575, "global_step": 182929, "epoch": 2055, "lr": 2.410344325841045e-05}
{"train_loss": 0.09819895774126053, "global_step": 182930, "epoch": 2055, "lr": 2.4102947340523008e-05}
{"train_loss": 0.10557249933481216, "global_step": 182931, "epoch": 2055, "lr": 2.4102451426117083e-05}
{"train_loss": 0.10391552746295929, "global_step": 182932, "epoch": 2055, "lr": 2.4101955515192792e-05}
{"train_loss": 0.10185844451189041, "global_step": 182933, "epoch": 2055, "lr": 2.4101459607750155e-05}
{"train_loss": 0.11509157717227936, "global_step": 182934, "epoch": 2055, "lr": 2.4100963703789282e-05}
{"train_loss": 0.0827639028429985, "global_step": 182935, "epoch": 2055, "lr": 2.41004678033102e-05}
{"train_loss": 0.052221961319446564, "global_step": 182936, "epoch": 2055, "lr": 2.4099971906313013e-05}
{"train_loss": 0.07747672498226166, "global_step": 182937, "epoch": 2055, "lr": 2.4099476012797746e-05}
{"train_loss": 0.0872495099902153, "global_step": 182938, "epoch": 2055, "lr": 2.4098980122764513e-05}
{"train_loss": 0.12145872414112091, "global_step": 182939, "epoch": 2055, "lr": 2.4098484236213338e-05}
{"train_loss": 0.048976097255945206, "global_step": 182940, "epoch": 2055, "lr": 2.4097988353144318e-05}
{"train_loss": 0.06526187062263489, "global_step": 182941, "epoch": 2055, "lr": 2.40974924735575e-05}
{"train_loss": 0.0998886227607727, "global_step": 182942, "epoch": 2055, "lr": 2.4096996597452964e-05}
{"train_loss": 0.04931487515568733, "global_step": 182943, "epoch": 2055, "lr": 2.409650072483076e-05}
{"train_loss": 0.036910563707351685, "global_step": 182944, "epoch": 2055, "lr": 2.4096004855690985e-05}
{"train_loss": 0.08613678812980652, "global_step": 182945, "epoch": 2055, "lr": 2.4095508990033666e-05}
{"train_loss": 0.026801500469446182, "global_step": 182946, "epoch": 2055, "lr": 2.4095013127858913e-05}
{"train_loss": 0.09183365106582642, "global_step": 182947, "epoch": 2055, "lr": 2.4094517269166743e-05}
{"train_loss": 0.05396794155240059, "global_step": 182948, "epoch": 2055, "lr": 2.409402141395725e-05}
{"train_loss": 0.07530380040407181, "global_step": 182949, "epoch": 2055, "lr": 2.409352556223053e-05}
{"train_loss": 0.09156307578086853, "global_step": 182950, "epoch": 2055, "lr": 2.4093029713986593e-05}
{"train_loss": 0.06718975305557251, "global_step": 182951, "epoch": 2055, "lr": 2.4092533869225546e-05}
{"train_loss": 0.06164534389972687, "global_step": 182952, "epoch": 2055, "lr": 2.4092038027947435e-05}
{"train_loss": 0.06646709144115448, "global_step": 182953, "epoch": 2055, "lr": 2.4091542190152344e-05}
{"train_loss": 0.046492211520671844, "global_step": 182954, "epoch": 2055, "lr": 2.4091046355840313e-05}
{"train_loss": 0.06557118892669678, "global_step": 182955, "epoch": 2055, "lr": 2.409055052501144e-05}
{"train_loss": 0.06222790852189064, "global_step": 182956, "epoch": 2055, "lr": 2.409005469766578e-05}
{"train_loss": 0.06824733316898346, "global_step": 182957, "epoch": 2055, "lr": 2.4089558873803382e-05}
{"train_loss": 0.11730977892875671, "global_step": 182958, "epoch": 2055, "lr": 2.4089063053424338e-05}
{"train_loss": 0.07048377394676208, "global_step": 182959, "epoch": 2055, "lr": 2.408856723652869e-05}
{"train_loss": 0.0548979789018631, "global_step": 182960, "epoch": 2055, "lr": 2.4088071423116537e-05}
{"train_loss": 0.04753049463033676, "global_step": 182961, "epoch": 2055, "lr": 2.4087575613187924e-05}
{"train_loss": 0.05093859136104584, "global_step": 182962, "epoch": 2055, "lr": 2.4087079806742908e-05}
{"train_loss": 0.11611782014369965, "global_step": 182963, "epoch": 2055, "lr": 2.4086584003781582e-05}
{"train_loss": 0.07131106406450272, "global_step": 182964, "epoch": 2055, "lr": 2.408608820430398e-05}
{"train_loss": 0.10780060291290283, "global_step": 182965, "epoch": 2055, "lr": 2.4085592408310186e-05}
{"train_loss": 0.10956795513629913, "global_step": 182966, "epoch": 2055, "lr": 2.408509661580029e-05}
{"train_loss": 0.07637234032154083, "global_step": 182967, "epoch": 2055, "lr": 2.4084600826774318e-05}
{"train_loss": 0.10966415703296661, "global_step": 182968, "epoch": 2055, "lr": 2.4084105041232374e-05}
{"train_loss": 0.09485527127981186, "global_step": 182969, "epoch": 2055, "lr": 2.4083609259174483e-05}
{"train_loss": 0.07728550583124161, "global_step": 182970, "epoch": 2055, "lr": 2.408311348060076e-05}
{"train_loss": 0.03787047788500786, "global_step": 182971, "epoch": 2055, "lr": 2.408261770551123e-05}
{"train_loss": 0.10227847099304199, "global_step": 182972, "epoch": 2055, "lr": 2.4082121933905986e-05}
{"train_loss": 0.05845651030540466, "global_step": 182973, "epoch": 2055, "lr": 2.408162616578507e-05}
{"train_loss": 0.09446825832128525, "global_step": 182974, "epoch": 2055, "lr": 2.4081130401148584e-05}
{"train_loss": 0.08626923710107803, "global_step": 182975, "epoch": 2055, "lr": 2.408063463999655e-05}
{"train_loss": 0.07734592258930206, "global_step": 182976, "epoch": 2055, "lr": 2.4080138882329085e-05}
{"train_loss": 0.09105472266674042, "global_step": 182977, "epoch": 2055, "lr": 2.4079643128146205e-05}
{"train_loss": 0.04504290223121643, "global_step": 182978, "epoch": 2055, "lr": 2.4079147377448025e-05}
{"train_loss": 0.05282288044691086, "global_step": 182979, "epoch": 2055, "lr": 2.4078651630234566e-05}
{"train_loss": 0.11410601437091827, "global_step": 182980, "epoch": 2055, "lr": 2.407815588650594e-05}
{"train_loss": 0.07171313464641571, "global_step": 182981, "epoch": 2055, "lr": 2.4077660146262165e-05}
{"train_loss": 0.015458246693015099, "global_step": 182982, "epoch": 2055, "lr": 2.407716440950335e-05}
{"train_loss": 0.07909314187892368, "global_step": 182983, "epoch": 2055, "lr": 2.407666867622953e-05, "val_loss": 7.770115375518799, "train_action_mse_error": 27.638652801513672}
{"train_loss": 0.08367978781461716, "global_step": 182984, "epoch": 2056, "lr": 2.4076172946440783e-05}
{"train_loss": 0.09113609790802002, "global_step": 182985, "epoch": 2056, "lr": 2.40756772201372e-05}
{"train_loss": 0.07803011685609818, "global_step": 182986, "epoch": 2056, "lr": 2.4075181497318804e-05}
{"train_loss": 0.11814412474632263, "global_step": 182987, "epoch": 2056, "lr": 2.4074685777985705e-05}
{"train_loss": 0.03046605922281742, "global_step": 182988, "epoch": 2056, "lr": 2.407419006213793e-05}
{"train_loss": 0.03853710740804672, "global_step": 182989, "epoch": 2056, "lr": 2.4073694349775584e-05}
{"train_loss": 0.08887907862663269, "global_step": 182990, "epoch": 2056, "lr": 2.407319864089869e-05}
{"train_loss": 0.15962931513786316, "global_step": 182991, "epoch": 2056, "lr": 2.4072702935507364e-05}
{"train_loss": 0.05890867859125137, "global_step": 182992, "epoch": 2056, "lr": 2.4072207233601623e-05}
{"train_loss": 0.08745209127664566, "global_step": 182993, "epoch": 2056, "lr": 2.4071711535181578e-05}
{"train_loss": 0.09777796268463135, "global_step": 182994, "epoch": 2056, "lr": 2.407121584024728e-05}
{"train_loss": 0.06273920089006424, "global_step": 182995, "epoch": 2056, "lr": 2.4070720148798764e-05}
{"train_loss": 0.10370851308107376, "global_step": 182996, "epoch": 2056, "lr": 2.407022446083615e-05}
{"train_loss": 0.07321204245090485, "global_step": 182997, "epoch": 2056, "lr": 2.406972877635945e-05}
{"train_loss": 0.0470900684595108, "global_step": 182998, "epoch": 2056, "lr": 2.4069233095368788e-05}
{"train_loss": 0.018335416913032532, "global_step": 182999, "epoch": 2056, "lr": 2.406873741786418e-05}
{"train_loss": 0.07014722377061844, "global_step": 183000, "epoch": 2056, "lr": 2.4068241743845733e-05}
{"train_loss": 0.08319354057312012, "global_step": 183001, "epoch": 2056, "lr": 2.406774607331349e-05}
{"train_loss": 0.1008978858590126, "global_step": 183002, "epoch": 2056, "lr": 2.4067250406267504e-05}
{"train_loss": 0.10839865356683731, "global_step": 183003, "epoch": 2056, "lr": 2.4066754742707865e-05}
{"train_loss": 0.09091749042272568, "global_step": 183004, "epoch": 2056, "lr": 2.4066259082634652e-05}
{"train_loss": 0.0503697395324707, "global_step": 183005, "epoch": 2056, "lr": 2.4065763426047894e-05}
{"train_loss": 0.09021425992250443, "global_step": 183006, "epoch": 2056, "lr": 2.4065267772947696e-05}
{"train_loss": 0.08001499623060226, "global_step": 183007, "epoch": 2056, "lr": 2.4064772123334096e-05}
{"train_loss": 0.06245523318648338, "global_step": 183008, "epoch": 2056, "lr": 2.4064276477207182e-05}
{"train_loss": 0.05530012398958206, "global_step": 183009, "epoch": 2056, "lr": 2.406378083456699e-05}
{"train_loss": 0.09895505756139755, "global_step": 183010, "epoch": 2056, "lr": 2.406328519541363e-05}
{"train_loss": 0.09999234974384308, "global_step": 183011, "epoch": 2056, "lr": 2.4062789559747123e-05}
{"train_loss": 0.09042399376630783, "global_step": 183012, "epoch": 2056, "lr": 2.4062293927567582e-05}
{"train_loss": 0.09733514487743378, "global_step": 183013, "epoch": 2056, "lr": 2.406179829887503e-05}
{"train_loss": 0.10376040637493134, "global_step": 183014, "epoch": 2056, "lr": 2.4061302673669567e-05}
{"train_loss": 0.0827869325876236, "global_step": 183015, "epoch": 2056, "lr": 2.406080705195123e-05}
{"train_loss": 0.07121558487415314, "global_step": 183016, "epoch": 2056, "lr": 2.4060311433720118e-05}
{"train_loss": 0.08335480839014053, "global_step": 183017, "epoch": 2056, "lr": 2.4059815818976262e-05}
{"train_loss": 0.027668574824929237, "global_step": 183018, "epoch": 2056, "lr": 2.4059320207719767e-05}
{"train_loss": 0.043866854161024094, "global_step": 183019, "epoch": 2056, "lr": 2.4058824599950665e-05}
{"train_loss": 0.06294403970241547, "global_step": 183020, "epoch": 2056, "lr": 2.405832899566903e-05}
{"train_loss": 0.09927449375391006, "global_step": 183021, "epoch": 2056, "lr": 2.405783339487496e-05}
{"train_loss": 0.03755098208785057, "global_step": 183022, "epoch": 2056, "lr": 2.405733779756848e-05}
{"train_loss": 0.12573394179344177, "global_step": 183023, "epoch": 2056, "lr": 2.405684220374969e-05}
{"train_loss": 0.09751233458518982, "global_step": 183024, "epoch": 2056, "lr": 2.405634661341863e-05}
{"train_loss": 0.06311328709125519, "global_step": 183025, "epoch": 2056, "lr": 2.4055851026575395e-05}
{"train_loss": 0.05725358799099922, "global_step": 183026, "epoch": 2056, "lr": 2.4055355443220005e-05}
{"train_loss": 0.05419261381030083, "global_step": 183027, "epoch": 2056, "lr": 2.4054859863352586e-05}
{"train_loss": 0.060587335377931595, "global_step": 183028, "epoch": 2056, "lr": 2.4054364286973153e-05}
{"train_loss": 0.1200026422739029, "global_step": 183029, "epoch": 2056, "lr": 2.405386871408182e-05}
{"train_loss": 0.055972322821617126, "global_step": 183030, "epoch": 2056, "lr": 2.40533731446786e-05}
{"train_loss": 0.08498954772949219, "global_step": 183031, "epoch": 2056, "lr": 2.4052877578763605e-05}
{"train_loss": 0.03775342181324959, "global_step": 183032, "epoch": 2056, "lr": 2.405238201633687e-05}
{"train_loss": 0.06781599670648575, "global_step": 183033, "epoch": 2056, "lr": 2.4051886457398497e-05}
{"train_loss": 0.052966441959142685, "global_step": 183034, "epoch": 2056, "lr": 2.4051390901948527e-05}
{"train_loss": 0.07847144454717636, "global_step": 183035, "epoch": 2056, "lr": 2.4050895349987017e-05}
{"train_loss": 0.06875389069318771, "global_step": 183036, "epoch": 2056, "lr": 2.405039980151406e-05}
{"train_loss": 0.09323354065418243, "global_step": 183037, "epoch": 2056, "lr": 2.4049904256529697e-05}
{"train_loss": 0.025781450793147087, "global_step": 183038, "epoch": 2056, "lr": 2.404940871503403e-05}
{"train_loss": 0.14059919118881226, "global_step": 183039, "epoch": 2056, "lr": 2.4048913177027076e-05}
{"train_loss": 0.03379293531179428, "global_step": 183040, "epoch": 2056, "lr": 2.404841764250896e-05}
{"train_loss": 0.07365331053733826, "global_step": 183041, "epoch": 2056, "lr": 2.4047922111479687e-05}
{"train_loss": 0.0647212490439415, "global_step": 183042, "epoch": 2056, "lr": 2.404742658393938e-05}
{"train_loss": 0.09742250293493271, "global_step": 183043, "epoch": 2056, "lr": 2.4046931059888062e-05}
{"train_loss": 0.05072269216179848, "global_step": 183044, "epoch": 2056, "lr": 2.4046435539325833e-05}
{"train_loss": 0.03374960646033287, "global_step": 183045, "epoch": 2056, "lr": 2.4045940022252727e-05}
{"train_loss": 0.10388320684432983, "global_step": 183046, "epoch": 2056, "lr": 2.4045444508668847e-05}
{"train_loss": 0.08984819799661636, "global_step": 183047, "epoch": 2056, "lr": 2.4044948998574218e-05}
{"train_loss": 0.05972456932067871, "global_step": 183048, "epoch": 2056, "lr": 2.4044453491968955e-05}
{"train_loss": 0.07111440598964691, "global_step": 183049, "epoch": 2056, "lr": 2.4043957988853078e-05}
{"train_loss": 0.11578770726919174, "global_step": 183050, "epoch": 2056, "lr": 2.404346248922669e-05}
{"train_loss": 0.06674788147211075, "global_step": 183051, "epoch": 2056, "lr": 2.4042966993089827e-05}
{"train_loss": 0.08965957164764404, "global_step": 183052, "epoch": 2056, "lr": 2.4042471500442586e-05}
{"train_loss": 0.07235201448202133, "global_step": 183053, "epoch": 2056, "lr": 2.4041976011285e-05}
{"train_loss": 0.07665704935789108, "global_step": 183054, "epoch": 2056, "lr": 2.4041480525617177e-05}
{"train_loss": 0.0660833865404129, "global_step": 183055, "epoch": 2056, "lr": 2.404098504343914e-05}
{"train_loss": 0.06802897900342941, "global_step": 183056, "epoch": 2056, "lr": 2.4040489564750972e-05}
{"train_loss": 0.07244057953357697, "global_step": 183057, "epoch": 2056, "lr": 2.4039994089552766e-05}
{"train_loss": 0.21986037492752075, "global_step": 183058, "epoch": 2056, "lr": 2.4039498617844548e-05}
{"train_loss": 0.11468136310577393, "global_step": 183059, "epoch": 2056, "lr": 2.403900314962642e-05}
{"train_loss": 0.06123242899775505, "global_step": 183060, "epoch": 2056, "lr": 2.4038507684898416e-05}
{"train_loss": 0.0843668207526207, "global_step": 183061, "epoch": 2056, "lr": 2.4038012223660634e-05}
{"train_loss": 0.08636565506458282, "global_step": 183062, "epoch": 2056, "lr": 2.403751676591311e-05}
{"train_loss": 0.055286556482315063, "global_step": 183063, "epoch": 2056, "lr": 2.403702131165594e-05}
{"train_loss": 0.1558281034231186, "global_step": 183064, "epoch": 2056, "lr": 2.4036525860889157e-05}
{"train_loss": 0.05783946067094803, "global_step": 183065, "epoch": 2056, "lr": 2.4036030413612863e-05}
{"train_loss": 0.05227416753768921, "global_step": 183066, "epoch": 2056, "lr": 2.4035534969827094e-05}
{"train_loss": 0.06230240315198898, "global_step": 183067, "epoch": 2056, "lr": 2.403503952953195e-05}
{"train_loss": 0.07873199880123138, "global_step": 183068, "epoch": 2056, "lr": 2.4034544092727464e-05}
{"train_loss": 0.043813977390527725, "global_step": 183069, "epoch": 2056, "lr": 2.403404865941373e-05}
{"train_loss": 0.10795935243368149, "global_step": 183070, "epoch": 2056, "lr": 2.403355322959078e-05}
{"train_loss": 0.05730151757597923, "global_step": 183071, "epoch": 2056, "lr": 2.4033057803258723e-05}
{"train_loss": 0.07774069321373206, "global_step": 183072, "epoch": 2056, "lr": 2.4032562380417605e-05, "val_loss": 7.720470905303955}
{"train_loss": 0.09526960551738739, "global_step": 183073, "epoch": 2057, "lr": 2.403206696106748e-05}
{"train_loss": 0.05395927652716637, "global_step": 183074, "epoch": 2057, "lr": 2.4031571545208436e-05}
{"train_loss": 0.05410923436284065, "global_step": 183075, "epoch": 2057, "lr": 2.4031076132840518e-05}
{"train_loss": 0.04370992258191109, "global_step": 183076, "epoch": 2057, "lr": 2.4030580723963826e-05}
{"train_loss": 0.06804163008928299, "global_step": 183077, "epoch": 2057, "lr": 2.4030085318578383e-05}
{"train_loss": 0.06761012226343155, "global_step": 183078, "epoch": 2057, "lr": 2.4029589916684298e-05}
{"train_loss": 0.10839879512786865, "global_step": 183079, "epoch": 2057, "lr": 2.40290945182816e-05}
{"train_loss": 0.11318562924861908, "global_step": 183080, "epoch": 2057, "lr": 2.4028599123370392e-05}
{"train_loss": 0.056068308651447296, "global_step": 183081, "epoch": 2057, "lr": 2.4028103731950703e-05}
{"train_loss": 0.06759282946586609, "global_step": 183082, "epoch": 2057, "lr": 2.402760834402264e-05}
{"train_loss": 0.08475068211555481, "global_step": 183083, "epoch": 2057, "lr": 2.402711295958623e-05}
{"train_loss": 0.0964226946234703, "global_step": 183084, "epoch": 2057, "lr": 2.402661757864157e-05}
{"train_loss": 0.08562836050987244, "global_step": 183085, "epoch": 2057, "lr": 2.40261222011887e-05}
{"train_loss": 0.028100714087486267, "global_step": 183086, "epoch": 2057, "lr": 2.402562682722772e-05}
{"train_loss": 0.10621725022792816, "global_step": 183087, "epoch": 2057, "lr": 2.4025131456758664e-05}
{"train_loss": 0.1193104162812233, "global_step": 183088, "epoch": 2057, "lr": 2.4024636089781625e-05}
{"train_loss": 0.11289142817258835, "global_step": 183089, "epoch": 2057, "lr": 2.402414072629664e-05}
{"train_loss": 0.07990448921918869, "global_step": 183090, "epoch": 2057, "lr": 2.4023645366303814e-05}
{"train_loss": 0.10767126828432083, "global_step": 183091, "epoch": 2057, "lr": 2.402315000980317e-05}
{"train_loss": 0.09550247341394424, "global_step": 183092, "epoch": 2057, "lr": 2.4022654656794803e-05}
{"train_loss": 0.08726736158132553, "global_step": 183093, "epoch": 2057, "lr": 2.4022159307278786e-05}
{"train_loss": 0.04139736667275429, "global_step": 183094, "epoch": 2057, "lr": 2.4021663961255153e-05}
{"train_loss": 0.08605130761861801, "global_step": 183095, "epoch": 2057, "lr": 2.4021168618724017e-05}
{"train_loss": 0.13218960165977478, "global_step": 183096, "epoch": 2057, "lr": 2.402067327968539e-05}
{"train_loss": 0.09414032101631165, "global_step": 183097, "epoch": 2057, "lr": 2.4020177944139394e-05}
{"train_loss": 0.05283236503601074, "global_step": 183098, "epoch": 2057, "lr": 2.4019682612086042e-05}
{"train_loss": 0.060316599905490875, "global_step": 183099, "epoch": 2057, "lr": 2.401918728352545e-05}
{"train_loss": 0.0659244954586029, "global_step": 183100, "epoch": 2057, "lr": 2.4018691958457644e-05}
{"train_loss": 0.05414694547653198, "global_step": 183101, "epoch": 2057, "lr": 2.4018196636882727e-05}
{"train_loss": 0.07603536546230316, "global_step": 183102, "epoch": 2057, "lr": 2.4017701318800723e-05}
{"train_loss": 0.06685104221105576, "global_step": 183103, "epoch": 2057, "lr": 2.4017206004211747e-05}
{"train_loss": 0.1389785259962082, "global_step": 183104, "epoch": 2057, "lr": 2.401671069311582e-05}
{"train_loss": 0.05540865287184715, "global_step": 183105, "epoch": 2057, "lr": 2.4016215385513046e-05}
{"train_loss": 0.11453317850828171, "global_step": 183106, "epoch": 2057, "lr": 2.401572008140346e-05}
{"train_loss": 0.05882958322763443, "global_step": 183107, "epoch": 2057, "lr": 2.401522478078716e-05}
{"train_loss": 0.0525507777929306, "global_step": 183108, "epoch": 2057, "lr": 2.4014729483664183e-05}
{"train_loss": 0.029346268624067307, "global_step": 183109, "epoch": 2057, "lr": 2.4014234190034624e-05}
{"train_loss": 0.059653766453266144, "global_step": 183110, "epoch": 2057, "lr": 2.401373889989853e-05}
{"train_loss": 0.046273186802864075, "global_step": 183111, "epoch": 2057, "lr": 2.401324361325596e-05}
{"train_loss": 0.03092668019235134, "global_step": 183112, "epoch": 2057, "lr": 2.401274833010701e-05}
{"train_loss": 0.07455506920814514, "global_step": 183113, "epoch": 2057, "lr": 2.4012253050451707e-05}
{"train_loss": 0.14242179691791534, "global_step": 183114, "epoch": 2057, "lr": 2.4011757774290163e-05}
{"train_loss": 0.08260447531938553, "global_step": 183115, "epoch": 2057, "lr": 2.40112625016224e-05}
{"train_loss": 0.01492999866604805, "global_step": 183116, "epoch": 2057, "lr": 2.401076723244852e-05}
{"train_loss": 0.058395665138959885, "global_step": 183117, "epoch": 2057, "lr": 2.4010271966768566e-05}
{"train_loss": 0.08126205205917358, "global_step": 183118, "epoch": 2057, "lr": 2.4009776704582625e-05}
{"train_loss": 0.09530077129602432, "global_step": 183119, "epoch": 2057, "lr": 2.4009281445890743e-05}
{"train_loss": 0.19022946059703827, "global_step": 183120, "epoch": 2057, "lr": 2.4008786190693006e-05}
{"train_loss": 0.08371584117412567, "global_step": 183121, "epoch": 2057, "lr": 2.4008290938989457e-05}
{"train_loss": 0.08267582952976227, "global_step": 183122, "epoch": 2057, "lr": 2.4007795690780194e-05}
{"train_loss": 0.04397917911410332, "global_step": 183123, "epoch": 2057, "lr": 2.400730044606525e-05}
{"train_loss": 0.08712088316679001, "global_step": 183124, "epoch": 2057, "lr": 2.4006805204844722e-05}
{"train_loss": 0.0668209046125412, "global_step": 183125, "epoch": 2057, "lr": 2.4006309967118647e-05}
{"train_loss": 0.06071678176522255, "global_step": 183126, "epoch": 2057, "lr": 2.400581473288712e-05}
{"train_loss": 0.05671601742506027, "global_step": 183127, "epoch": 2057, "lr": 2.400531950215018e-05}
{"train_loss": 0.10879304260015488, "global_step": 183128, "epoch": 2057, "lr": 2.4004824274907928e-05}
{"train_loss": 0.08584713190793991, "global_step": 183129, "epoch": 2057, "lr": 2.400432905116039e-05}
{"train_loss": 0.11309713125228882, "global_step": 183130, "epoch": 2057, "lr": 2.4003833830907658e-05}
{"train_loss": 0.10829003155231476, "global_step": 183131, "epoch": 2057, "lr": 2.4003338614149805e-05}
{"train_loss": 0.16001500189304352, "global_step": 183132, "epoch": 2057, "lr": 2.400284340088687e-05}
{"train_loss": 0.0572744719684124, "global_step": 183133, "epoch": 2057, "lr": 2.4002348191118958e-05}
{"train_loss": 0.06073124706745148, "global_step": 183134, "epoch": 2057, "lr": 2.4001852984846092e-05}
{"train_loss": 0.13704250752925873, "global_step": 183135, "epoch": 2057, "lr": 2.4001357782068372e-05}
{"train_loss": 0.07707271724939346, "global_step": 183136, "epoch": 2057, "lr": 2.4000862582785842e-05}
{"train_loss": 0.058569058775901794, "global_step": 183137, "epoch": 2057, "lr": 2.4000367386998594e-05}
{"train_loss": 0.1150493249297142, "global_step": 183138, "epoch": 2057, "lr": 2.3999872194706668e-05}
{"train_loss": 0.04368548095226288, "global_step": 183139, "epoch": 2057, "lr": 2.3999377005910155e-05}
{"train_loss": 0.07840131223201752, "global_step": 183140, "epoch": 2057, "lr": 2.3998881820609093e-05}
{"train_loss": 0.062259819358587265, "global_step": 183141, "epoch": 2057, "lr": 2.3998386638803577e-05}
{"train_loss": 0.06663788855075836, "global_step": 183142, "epoch": 2057, "lr": 2.3997891460493648e-05}
{"train_loss": 0.12993022799491882, "global_step": 183143, "epoch": 2057, "lr": 2.3997396285679403e-05}
{"train_loss": 0.06234578415751457, "global_step": 183144, "epoch": 2057, "lr": 2.399690111436087e-05}
{"train_loss": 0.05724545568227768, "global_step": 183145, "epoch": 2057, "lr": 2.399640594653816e-05}
{"train_loss": 0.08478816598653793, "global_step": 183146, "epoch": 2057, "lr": 2.3995910782211294e-05}
{"train_loss": 0.05876518413424492, "global_step": 183147, "epoch": 2057, "lr": 2.3995415621380384e-05}
{"train_loss": 0.06750471144914627, "global_step": 183148, "epoch": 2057, "lr": 2.399492046404545e-05}
{"train_loss": 0.06587553769350052, "global_step": 183149, "epoch": 2057, "lr": 2.3994425310206603e-05}
{"train_loss": 0.06801826506853104, "global_step": 183150, "epoch": 2057, "lr": 2.3993930159863885e-05}
{"train_loss": 0.045609038323163986, "global_step": 183151, "epoch": 2057, "lr": 2.399343501301735e-05}
{"train_loss": 0.09652306884527206, "global_step": 183152, "epoch": 2057, "lr": 2.3992939869667103e-05}
{"train_loss": 0.08418422192335129, "global_step": 183153, "epoch": 2057, "lr": 2.3992444729813163e-05}
{"train_loss": 0.0615929514169693, "global_step": 183154, "epoch": 2057, "lr": 2.3991949593455647e-05}
{"train_loss": 0.0674857422709465, "global_step": 183155, "epoch": 2057, "lr": 2.3991454460594575e-05}
{"train_loss": 0.0926295593380928, "global_step": 183156, "epoch": 2057, "lr": 2.399095933123005e-05}
{"train_loss": 0.05184193328022957, "global_step": 183157, "epoch": 2057, "lr": 2.3990464205362106e-05}
{"train_loss": 0.07862064987421036, "global_step": 183158, "epoch": 2057, "lr": 2.398996908299085e-05}
{"train_loss": 0.09399569034576416, "global_step": 183159, "epoch": 2057, "lr": 2.3989473964116303e-05}
{"train_loss": 0.06870267540216446, "global_step": 183160, "epoch": 2057, "lr": 2.3988978848738573e-05}
{"train_loss": 0.07824476992481211, "global_step": 183161, "epoch": 2057, "lr": 2.398848373685769e-05, "val_loss": 7.642692565917969}
{"train_loss": 0.06692282110452652, "global_step": 183162, "epoch": 2058, "lr": 2.3987988628473763e-05}
{"train_loss": 0.05428772792220116, "global_step": 183163, "epoch": 2058, "lr": 2.3987493523586807e-05}
{"train_loss": 0.08209215849637985, "global_step": 183164, "epoch": 2058, "lr": 2.398699842219694e-05}
{"train_loss": 0.0679335743188858, "global_step": 183165, "epoch": 2058, "lr": 2.3986503324304182e-05}
{"train_loss": 0.06779544055461884, "global_step": 183166, "epoch": 2058, "lr": 2.3986008229908617e-05}
{"train_loss": 0.06455443799495697, "global_step": 183167, "epoch": 2058, "lr": 2.398551313901034e-05}
{"train_loss": 0.09267622232437134, "global_step": 183168, "epoch": 2058, "lr": 2.3985018051609376e-05}
{"train_loss": 0.09021863341331482, "global_step": 183169, "epoch": 2058, "lr": 2.398452296770583e-05}
{"train_loss": 0.10254263132810593, "global_step": 183170, "epoch": 2058, "lr": 2.3984027887299725e-05}
{"train_loss": 0.07700292766094208, "global_step": 183171, "epoch": 2058, "lr": 2.398353281039117e-05}
{"train_loss": 0.08920487761497498, "global_step": 183172, "epoch": 2058, "lr": 2.398303773698019e-05}
{"train_loss": 0.14063498377799988, "global_step": 183173, "epoch": 2058, "lr": 2.3982542667066897e-05}
{"train_loss": 0.10422424226999283, "global_step": 183174, "epoch": 2058, "lr": 2.3982047600651314e-05}
{"train_loss": 0.05504604056477547, "global_step": 183175, "epoch": 2058, "lr": 2.3981552537733543e-05}
{"train_loss": 0.06278754025697708, "global_step": 183176, "epoch": 2058, "lr": 2.398105747831362e-05}
{"train_loss": 0.09751112014055252, "global_step": 183177, "epoch": 2058, "lr": 2.3980562422391646e-05}
{"train_loss": 0.07279541343450546, "global_step": 183178, "epoch": 2058, "lr": 2.3980067369967644e-05}
{"train_loss": 0.06777285039424896, "global_step": 183179, "epoch": 2058, "lr": 2.3979572321041733e-05}
{"train_loss": 0.08612434566020966, "global_step": 183180, "epoch": 2058, "lr": 2.3979077275613926e-05}
{"train_loss": 0.066644087433815, "global_step": 183181, "epoch": 2058, "lr": 2.3978582233684334e-05}
{"train_loss": 0.056197162717580795, "global_step": 183182, "epoch": 2058, "lr": 2.3978087195252984e-05}
{"train_loss": 0.14833500981330872, "global_step": 183183, "epoch": 2058, "lr": 2.3977592160319983e-05}
{"train_loss": 0.03695482015609741, "global_step": 183184, "epoch": 2058, "lr": 2.3977097128885363e-05}
{"train_loss": 0.04933498427271843, "global_step": 183185, "epoch": 2058, "lr": 2.397660210094922e-05}
{"train_loss": 0.0843365266919136, "global_step": 183186, "epoch": 2058, "lr": 2.397610707651159e-05}
{"train_loss": 0.057984232902526855, "global_step": 183187, "epoch": 2058, "lr": 2.3975612055572572e-05}
{"train_loss": 0.09471544623374939, "global_step": 183188, "epoch": 2058, "lr": 2.3975117038132217e-05}
{"train_loss": 0.09904308617115021, "global_step": 183189, "epoch": 2058, "lr": 2.397462202419057e-05}
{"train_loss": 0.14257864654064178, "global_step": 183190, "epoch": 2058, "lr": 2.397412701374774e-05}
{"train_loss": 0.09007322788238525, "global_step": 183191, "epoch": 2058, "lr": 2.3973632006803743e-05}
{"train_loss": 0.07581702619791031, "global_step": 183192, "epoch": 2058, "lr": 2.3973137003358703e-05}
{"train_loss": 0.08956920355558395, "global_step": 183193, "epoch": 2058, "lr": 2.397264200341263e-05}
{"train_loss": 0.03440367057919502, "global_step": 183194, "epoch": 2058, "lr": 2.3972147006965646e-05}
{"train_loss": 0.11143362522125244, "global_step": 183195, "epoch": 2058, "lr": 2.3971652014017763e-05}
{"train_loss": 0.07407648116350174, "global_step": 183196, "epoch": 2058, "lr": 2.3971157024569096e-05}
{"train_loss": 0.06128918379545212, "global_step": 183197, "epoch": 2058, "lr": 2.3970662038619675e-05}
{"train_loss": 0.07822085916996002, "global_step": 183198, "epoch": 2058, "lr": 2.3970167056169597e-05}
{"train_loss": 0.05940967798233032, "global_step": 183199, "epoch": 2058, "lr": 2.396967207721889e-05}
{"train_loss": 0.061729706823825836, "global_step": 183200, "epoch": 2058, "lr": 2.396917710176767e-05}
{"train_loss": 0.029842017218470573, "global_step": 183201, "epoch": 2058, "lr": 2.396868212981595e-05}
{"train_loss": 0.04909148067235947, "global_step": 183202, "epoch": 2058, "lr": 2.3968187161363832e-05}
{"train_loss": 0.06756585091352463, "global_step": 183203, "epoch": 2058, "lr": 2.3967692196411385e-05}
{"train_loss": 0.04890309274196625, "global_step": 183204, "epoch": 2058, "lr": 2.396719723495865e-05}
{"train_loss": 0.08106409013271332, "global_step": 183205, "epoch": 2058, "lr": 2.3966702277005724e-05}
{"train_loss": 0.04411183297634125, "global_step": 183206, "epoch": 2058, "lr": 2.396620732255264e-05}
{"train_loss": 0.04995240643620491, "global_step": 183207, "epoch": 2058, "lr": 2.3965712371599503e-05}
{"train_loss": 0.05353357642889023, "global_step": 183208, "epoch": 2058, "lr": 2.3965217424146336e-05}
{"train_loss": 0.06886348128318787, "global_step": 183209, "epoch": 2058, "lr": 2.3964722480193245e-05}
{"train_loss": 0.0308090727776289, "global_step": 183210, "epoch": 2058, "lr": 2.396422753974027e-05}
{"train_loss": 0.09086816757917404, "global_step": 183211, "epoch": 2058, "lr": 2.3963732602787497e-05}
{"train_loss": 0.07996205985546112, "global_step": 183212, "epoch": 2058, "lr": 2.396323766933497e-05}
{"train_loss": 0.1048634946346283, "global_step": 183213, "epoch": 2058, "lr": 2.3962742739382786e-05}
{"train_loss": 0.06715260446071625, "global_step": 183214, "epoch": 2058, "lr": 2.396224781293097e-05}
{"train_loss": 0.13872164487838745, "global_step": 183215, "epoch": 2058, "lr": 2.396175288997964e-05}
{"train_loss": 0.07885994017124176, "global_step": 183216, "epoch": 2058, "lr": 2.3961257970528812e-05}
{"train_loss": 0.056072548031806946, "global_step": 183217, "epoch": 2058, "lr": 2.3960763054578593e-05}
{"train_loss": 0.032109107822179794, "global_step": 183218, "epoch": 2058, "lr": 2.3960268142129012e-05}
{"train_loss": 0.06198102980852127, "global_step": 183219, "epoch": 2058, "lr": 2.395977323318016e-05}
{"train_loss": 0.0977521687746048, "global_step": 183220, "epoch": 2058, "lr": 2.3959278327732117e-05}
{"train_loss": 0.05147682502865791, "global_step": 183221, "epoch": 2058, "lr": 2.3958783425784925e-05}
{"train_loss": 0.0945960134267807, "global_step": 183222, "epoch": 2058, "lr": 2.3958288527338645e-05}
{"train_loss": 0.033498913049697876, "global_step": 183223, "epoch": 2058, "lr": 2.3957793632393372e-05}
{"train_loss": 0.09306881576776505, "global_step": 183224, "epoch": 2058, "lr": 2.395729874094914e-05}
{"train_loss": 0.07054072618484497, "global_step": 183225, "epoch": 2058, "lr": 2.395680385300605e-05}
{"train_loss": 0.08118834346532822, "global_step": 183226, "epoch": 2058, "lr": 2.3956308968564152e-05}
{"train_loss": 0.08425355702638626, "global_step": 183227, "epoch": 2058, "lr": 2.3955814087623485e-05}
{"train_loss": 0.09550262242555618, "global_step": 183228, "epoch": 2058, "lr": 2.3955319210184168e-05}
{"train_loss": 0.07610363513231277, "global_step": 183229, "epoch": 2058, "lr": 2.3954824336246217e-05}
{"train_loss": 0.021608833223581314, "global_step": 183230, "epoch": 2058, "lr": 2.3954329465809745e-05}
{"train_loss": 0.08722761273384094, "global_step": 183231, "epoch": 2058, "lr": 2.3953834598874775e-05}
{"train_loss": 0.08345744013786316, "global_step": 183232, "epoch": 2058, "lr": 2.395333973544141e-05}
{"train_loss": 0.0818902850151062, "global_step": 183233, "epoch": 2058, "lr": 2.3952844875509682e-05}
{"train_loss": 0.07929562777280807, "global_step": 183234, "epoch": 2058, "lr": 2.3952350019079706e-05}
{"train_loss": 0.1209920346736908, "global_step": 183235, "epoch": 2058, "lr": 2.3951855166151486e-05}
{"train_loss": 0.05098410323262215, "global_step": 183236, "epoch": 2058, "lr": 2.395136031672515e-05}
{"train_loss": 0.046486176550388336, "global_step": 183237, "epoch": 2058, "lr": 2.3950865470800715e-05}
{"train_loss": 0.10148152709007263, "global_step": 183238, "epoch": 2058, "lr": 2.395037062837827e-05}
{"train_loss": 0.0711413100361824, "global_step": 183239, "epoch": 2058, "lr": 2.39498757894579e-05}
{"train_loss": 0.07722265273332596, "global_step": 183240, "epoch": 2058, "lr": 2.394938095403963e-05}
{"train_loss": 0.12487535178661346, "global_step": 183241, "epoch": 2058, "lr": 2.3948886122123566e-05}
{"train_loss": 0.10924192517995834, "global_step": 183242, "epoch": 2058, "lr": 2.3948391293709736e-05}
{"train_loss": 0.09426481276750565, "global_step": 183243, "epoch": 2058, "lr": 2.3947896468798253e-05}
{"train_loss": 0.03928282856941223, "global_step": 183244, "epoch": 2058, "lr": 2.3947401647389135e-05}
{"train_loss": 0.05237927287817001, "global_step": 183245, "epoch": 2058, "lr": 2.394690682948249e-05}
{"train_loss": 0.07797802239656448, "global_step": 183246, "epoch": 2058, "lr": 2.3946412015078346e-05}
{"train_loss": 0.13921524584293365, "global_step": 183247, "epoch": 2058, "lr": 2.394591720417681e-05}
{"train_loss": 0.07641637325286865, "global_step": 183248, "epoch": 2058, "lr": 2.394542239677791e-05}
{"train_loss": 0.0891990214586258, "global_step": 183249, "epoch": 2058, "lr": 2.394492759288175e-05}
{"train_loss": 0.07665309297402254, "global_step": 183250, "epoch": 2058, "lr": 2.3944432792488354e-05, "val_loss": 7.672057151794434}
{"train_loss": 0.056815482676029205, "global_step": 183251, "epoch": 2059, "lr": 2.394393799559783e-05}
{"train_loss": 0.09372138977050781, "global_step": 183252, "epoch": 2059, "lr": 2.394344320221021e-05}
{"train_loss": 0.061391036957502365, "global_step": 183253, "epoch": 2059, "lr": 2.3942948412325595e-05}
{"train_loss": 0.054983992129564285, "global_step": 183254, "epoch": 2059, "lr": 2.3942453625944017e-05}
{"train_loss": 0.08592712134122849, "global_step": 183255, "epoch": 2059, "lr": 2.394195884306557e-05}
{"train_loss": 0.03953807055950165, "global_step": 183256, "epoch": 2059, "lr": 2.3941464063690294e-05}
{"train_loss": 0.07969546318054199, "global_step": 183257, "epoch": 2059, "lr": 2.394096928781827e-05}
{"train_loss": 0.05517285689711571, "global_step": 183258, "epoch": 2059, "lr": 2.394047451544959e-05}
{"train_loss": 0.08165580034255981, "global_step": 183259, "epoch": 2059, "lr": 2.393997974658427e-05}
{"train_loss": 0.0762423649430275, "global_step": 183260, "epoch": 2059, "lr": 2.393948498122242e-05}
{"train_loss": 0.07353510707616806, "global_step": 183261, "epoch": 2059, "lr": 2.3938990219364093e-05}
{"train_loss": 0.046297863125801086, "global_step": 183262, "epoch": 2059, "lr": 2.393849546100933e-05}
{"train_loss": 0.10238218307495117, "global_step": 183263, "epoch": 2059, "lr": 2.3938000706158236e-05}
{"train_loss": 0.10748285055160522, "global_step": 183264, "epoch": 2059, "lr": 2.3937505954810845e-05}
{"train_loss": 0.12672704458236694, "global_step": 183265, "epoch": 2059, "lr": 2.3937011206967257e-05}
{"train_loss": 0.06211807206273079, "global_step": 183266, "epoch": 2059, "lr": 2.3936516462627516e-05}
{"train_loss": 0.08221712708473206, "global_step": 183267, "epoch": 2059, "lr": 2.393602172179168e-05}
{"train_loss": 0.07493055611848831, "global_step": 183268, "epoch": 2059, "lr": 2.3935526984459844e-05}
{"train_loss": 0.05100740119814873, "global_step": 183269, "epoch": 2059, "lr": 2.393503225063204e-05}
{"train_loss": 0.04726400598883629, "global_step": 183270, "epoch": 2059, "lr": 2.393453752030837e-05}
{"train_loss": 0.06030628830194473, "global_step": 183271, "epoch": 2059, "lr": 2.3934042793488864e-05}
{"train_loss": 0.08463507145643234, "global_step": 183272, "epoch": 2059, "lr": 2.3933548070173632e-05}
{"train_loss": 0.036186810582876205, "global_step": 183273, "epoch": 2059, "lr": 2.3933053350362694e-05}
{"train_loss": 0.10242793709039688, "global_step": 183274, "epoch": 2059, "lr": 2.3932558634056145e-05}
{"train_loss": 0.08784331381320953, "global_step": 183275, "epoch": 2059, "lr": 2.393206392125406e-05}
{"train_loss": 0.06517505645751953, "global_step": 183276, "epoch": 2059, "lr": 2.393156921195647e-05}
{"train_loss": 0.10178858786821365, "global_step": 183277, "epoch": 2059, "lr": 2.393107450616349e-05}
{"train_loss": 0.08130739629268646, "global_step": 183278, "epoch": 2059, "lr": 2.393057980387513e-05}
{"train_loss": 0.07036254554986954, "global_step": 183279, "epoch": 2059, "lr": 2.393008510509151e-05}
{"train_loss": 0.07524564117193222, "global_step": 183280, "epoch": 2059, "lr": 2.392959040981265e-05}
{"train_loss": 0.05739295855164528, "global_step": 183281, "epoch": 2059, "lr": 2.3929095718038663e-05}
{"train_loss": 0.10688022524118423, "global_step": 183282, "epoch": 2059, "lr": 2.3928601029769572e-05}
{"train_loss": 0.06560315936803818, "global_step": 183283, "epoch": 2059, "lr": 2.392810634500548e-05}
{"train_loss": 0.10705660283565521, "global_step": 183284, "epoch": 2059, "lr": 2.392761166374642e-05}
{"train_loss": 0.08770574629306793, "global_step": 183285, "epoch": 2059, "lr": 2.3927116985992487e-05}
{"train_loss": 0.07402990013360977, "global_step": 183286, "epoch": 2059, "lr": 2.3926622311743723e-05}
{"train_loss": 0.04175228998064995, "global_step": 183287, "epoch": 2059, "lr": 2.3926127641000225e-05}
{"train_loss": 0.03939104825258255, "global_step": 183288, "epoch": 2059, "lr": 2.392563297376202e-05}
{"train_loss": 0.06118185818195343, "global_step": 183289, "epoch": 2059, "lr": 2.3925138310029217e-05}
{"train_loss": 0.07479104399681091, "global_step": 183290, "epoch": 2059, "lr": 2.3924643649801836e-05}
{"train_loss": 0.08507711440324783, "global_step": 183291, "epoch": 2059, "lr": 2.392414899308e-05}
{"train_loss": 0.1419646441936493, "global_step": 183292, "epoch": 2059, "lr": 2.3923654339863717e-05}
{"train_loss": 0.0966416448354721, "global_step": 183293, "epoch": 2059, "lr": 2.392315969015308e-05}
{"train_loss": 0.07105576992034912, "global_step": 183294, "epoch": 2059, "lr": 2.3922665043948177e-05}
{"train_loss": 0.061923690140247345, "global_step": 183295, "epoch": 2059, "lr": 2.3922170401249032e-05}
{"train_loss": 0.04903490096330643, "global_step": 183296, "epoch": 2059, "lr": 2.3921675762055756e-05}
{"train_loss": 0.06752049922943115, "global_step": 183297, "epoch": 2059, "lr": 2.3921181126368376e-05}
{"train_loss": 0.07418905943632126, "global_step": 183298, "epoch": 2059, "lr": 2.392068649418699e-05}
{"train_loss": 0.12347477674484253, "global_step": 183299, "epoch": 2059, "lr": 2.3920191865511653e-05}
{"train_loss": 0.04948863759636879, "global_step": 183300, "epoch": 2059, "lr": 2.391969724034241e-05}
{"train_loss": 0.062149301171302795, "global_step": 183301, "epoch": 2059, "lr": 2.3919202618679358e-05}
{"train_loss": 0.04551812261343002, "global_step": 183302, "epoch": 2059, "lr": 2.391870800052254e-05}
{"train_loss": 0.08702276647090912, "global_step": 183303, "epoch": 2059, "lr": 2.3918213385872052e-05}
{"train_loss": 0.057867370545864105, "global_step": 183304, "epoch": 2059, "lr": 2.3917718774727937e-05}
{"train_loss": 0.0925796777009964, "global_step": 183305, "epoch": 2059, "lr": 2.3917224167090253e-05}
{"train_loss": 0.06864159554243088, "global_step": 183306, "epoch": 2059, "lr": 2.3916729562959093e-05}
{"train_loss": 0.14719340205192566, "global_step": 183307, "epoch": 2059, "lr": 2.39162349623345e-05}
{"train_loss": 0.08990306407213211, "global_step": 183308, "epoch": 2059, "lr": 2.3915740365216566e-05}
{"train_loss": 0.08236867189407349, "global_step": 183309, "epoch": 2059, "lr": 2.391524577160532e-05}
{"train_loss": 0.03506381809711456, "global_step": 183310, "epoch": 2059, "lr": 2.391475118150086e-05}
{"train_loss": 0.03696846216917038, "global_step": 183311, "epoch": 2059, "lr": 2.3914256594903255e-05}
{"train_loss": 0.1364540308713913, "global_step": 183312, "epoch": 2059, "lr": 2.3913762011812546e-05}
{"train_loss": 0.03490786999464035, "global_step": 183313, "epoch": 2059, "lr": 2.391326743222883e-05}
{"train_loss": 0.057082150131464005, "global_step": 183314, "epoch": 2059, "lr": 2.3912772856152144e-05}
{"train_loss": 0.09733261168003082, "global_step": 183315, "epoch": 2059, "lr": 2.391227828358258e-05}
{"train_loss": 0.0414394848048687, "global_step": 183316, "epoch": 2059, "lr": 2.3911783714520176e-05}
{"train_loss": 0.06990350037813187, "global_step": 183317, "epoch": 2059, "lr": 2.3911289148965034e-05}
{"train_loss": 0.08946429938077927, "global_step": 183318, "epoch": 2059, "lr": 2.3910794586917178e-05}
{"train_loss": 0.08285363018512726, "global_step": 183319, "epoch": 2059, "lr": 2.391030002837672e-05}
{"train_loss": 0.12588584423065186, "global_step": 183320, "epoch": 2059, "lr": 2.3909805473343688e-05}
{"train_loss": 0.08084897696971893, "global_step": 183321, "epoch": 2059, "lr": 2.390931092181818e-05}
{"train_loss": 0.06208905577659607, "global_step": 183322, "epoch": 2059, "lr": 2.390881637380023e-05}
{"train_loss": 0.03669285401701927, "global_step": 183323, "epoch": 2059, "lr": 2.3908321829289938e-05}
{"train_loss": 0.0823335126042366, "global_step": 183324, "epoch": 2059, "lr": 2.3907827288287337e-05}
{"train_loss": 0.04918971657752991, "global_step": 183325, "epoch": 2059, "lr": 2.3907332750792527e-05}
{"train_loss": 0.0969451367855072, "global_step": 183326, "epoch": 2059, "lr": 2.3906838216805544e-05}
{"train_loss": 0.13584037125110626, "global_step": 183327, "epoch": 2059, "lr": 2.3906343686326483e-05}
{"train_loss": 0.07072826474905014, "global_step": 183328, "epoch": 2059, "lr": 2.3905849159355383e-05}
{"train_loss": 0.06790536642074585, "global_step": 183329, "epoch": 2059, "lr": 2.390535463589232e-05}
{"train_loss": 0.06630124896764755, "global_step": 183330, "epoch": 2059, "lr": 2.3904860115937377e-05}
{"train_loss": 0.07521405816078186, "global_step": 183331, "epoch": 2059, "lr": 2.3904365599490596e-05}
{"train_loss": 0.14000147581100464, "global_step": 183332, "epoch": 2059, "lr": 2.390387108655207e-05}
{"train_loss": 0.07001706212759018, "global_step": 183333, "epoch": 2059, "lr": 2.390337657712184e-05}
{"train_loss": 0.05771876871585846, "global_step": 183334, "epoch": 2059, "lr": 2.3902882071199996e-05}
{"train_loss": 0.08151277154684067, "global_step": 183335, "epoch": 2059, "lr": 2.3902387568786566e-05}
{"train_loss": 0.060459088534116745, "global_step": 183336, "epoch": 2059, "lr": 2.390189306988167e-05}
{"train_loss": 0.08754803240299225, "global_step": 183337, "epoch": 2059, "lr": 2.390139857448534e-05}
{"train_loss": 0.08452147990465164, "global_step": 183338, "epoch": 2059, "lr": 2.3900904082597636e-05}
{"train_loss": 0.07599491651138562, "global_step": 183339, "epoch": 2059, "lr": 2.3900409594218655e-05, "val_loss": 7.700247764587402}
{"train_loss": 0.06749545782804489, "global_step": 183340, "epoch": 2060, "lr": 2.3899915109348425e-05}
{"train_loss": 0.08127494901418686, "global_step": 183341, "epoch": 2060, "lr": 2.389942062798705e-05}
{"train_loss": 0.10515390336513519, "global_step": 183342, "epoch": 2060, "lr": 2.389892615013458e-05}
{"train_loss": 0.08036968111991882, "global_step": 183343, "epoch": 2060, "lr": 2.3898431675791067e-05}
{"train_loss": 0.020624205470085144, "global_step": 183344, "epoch": 2060, "lr": 2.389793720495661e-05}
{"train_loss": 0.10837629437446594, "global_step": 183345, "epoch": 2060, "lr": 2.3897442737631236e-05}
{"train_loss": 0.11829707026481628, "global_step": 183346, "epoch": 2060, "lr": 2.3896948273815035e-05}
{"train_loss": 0.12457121163606644, "global_step": 183347, "epoch": 2060, "lr": 2.3896453813508087e-05}
{"train_loss": 0.0888959988951683, "global_step": 183348, "epoch": 2060, "lr": 2.3895959356710423e-05}
{"train_loss": 0.06368018686771393, "global_step": 183349, "epoch": 2060, "lr": 2.389546490342215e-05}
{"train_loss": 0.02848956733942032, "global_step": 183350, "epoch": 2060, "lr": 2.3894970453643296e-05}
{"train_loss": 0.031431492418050766, "global_step": 183351, "epoch": 2060, "lr": 2.389447600737396e-05}
{"train_loss": 0.09081728011369705, "global_step": 183352, "epoch": 2060, "lr": 2.3893981564614182e-05}
{"train_loss": 0.07841365784406662, "global_step": 183353, "epoch": 2060, "lr": 2.389348712536405e-05}
{"train_loss": 0.09855621308088303, "global_step": 183354, "epoch": 2060, "lr": 2.3892992689623605e-05}
{"train_loss": 0.0820944756269455, "global_step": 183355, "epoch": 2060, "lr": 2.3892498257392953e-05}
{"train_loss": 0.04323365539312363, "global_step": 183356, "epoch": 2060, "lr": 2.389200382867211e-05}
{"train_loss": 0.046117279678583145, "global_step": 183357, "epoch": 2060, "lr": 2.3891509403461194e-05}
{"train_loss": 0.06747914105653763, "global_step": 183358, "epoch": 2060, "lr": 2.3891014981760223e-05}
{"train_loss": 0.0685645192861557, "global_step": 183359, "epoch": 2060, "lr": 2.3890520563569308e-05}
{"train_loss": 0.05068087577819824, "global_step": 183360, "epoch": 2060, "lr": 2.389002614888847e-05}
{"train_loss": 0.06722940504550934, "global_step": 183361, "epoch": 2060, "lr": 2.3889531737717823e-05}
{"train_loss": 0.10526034981012344, "global_step": 183362, "epoch": 2060, "lr": 2.3889037330057395e-05}
{"train_loss": 0.04865357652306557, "global_step": 183363, "epoch": 2060, "lr": 2.388854292590728e-05}
{"train_loss": 0.05951005965471268, "global_step": 183364, "epoch": 2060, "lr": 2.388804852526752e-05}
{"train_loss": 0.11049485206604004, "global_step": 183365, "epoch": 2060, "lr": 2.388755412813819e-05}
{"train_loss": 0.07047458738088608, "global_step": 183366, "epoch": 2060, "lr": 2.3887059734519374e-05}
{"train_loss": 0.08205962181091309, "global_step": 183367, "epoch": 2060, "lr": 2.388656534441111e-05}
{"train_loss": 0.06852594017982483, "global_step": 183368, "epoch": 2060, "lr": 2.38860709578135e-05}
{"train_loss": 0.1276814043521881, "global_step": 183369, "epoch": 2060, "lr": 2.388557657472657e-05}
{"train_loss": 0.0742773786187172, "global_step": 183370, "epoch": 2060, "lr": 2.388508219515042e-05}
{"train_loss": 0.07095279544591904, "global_step": 183371, "epoch": 2060, "lr": 2.388458781908509e-05}
{"train_loss": 0.08129921555519104, "global_step": 183372, "epoch": 2060, "lr": 2.3884093446530675e-05}
{"train_loss": 0.04355333372950554, "global_step": 183373, "epoch": 2060, "lr": 2.3883599077487208e-05}
{"train_loss": 0.09073443710803986, "global_step": 183374, "epoch": 2060, "lr": 2.388310471195479e-05}
{"train_loss": 0.04506579414010048, "global_step": 183375, "epoch": 2060, "lr": 2.388261034993345e-05}
{"train_loss": 0.06975492835044861, "global_step": 183376, "epoch": 2060, "lr": 2.3882115991423297e-05}
{"train_loss": 0.049172308295965195, "global_step": 183377, "epoch": 2060, "lr": 2.3881621636424374e-05}
{"train_loss": 0.08419689536094666, "global_step": 183378, "epoch": 2060, "lr": 2.3881127284936728e-05}
{"train_loss": 0.08975326269865036, "global_step": 183379, "epoch": 2060, "lr": 2.3880632936960467e-05}
{"train_loss": 0.06217752397060394, "global_step": 183380, "epoch": 2060, "lr": 2.3880138592495616e-05}
{"train_loss": 0.06321606785058975, "global_step": 183381, "epoch": 2060, "lr": 2.387964425154228e-05}
{"train_loss": 0.072600357234478, "global_step": 183382, "epoch": 2060, "lr": 2.3879149914100508e-05}
{"train_loss": 0.06542636454105377, "global_step": 183383, "epoch": 2060, "lr": 2.387865558017035e-05}
{"train_loss": 0.033722300082445145, "global_step": 183384, "epoch": 2060, "lr": 2.3878161249751884e-05}
{"train_loss": 0.07591553777456284, "global_step": 183385, "epoch": 2060, "lr": 2.3877666922845205e-05}
{"train_loss": 0.12060803174972534, "global_step": 183386, "epoch": 2060, "lr": 2.3877172599450332e-05}
{"train_loss": 0.09748610854148865, "global_step": 183387, "epoch": 2060, "lr": 2.3876678279567376e-05}
{"train_loss": 0.0445554256439209, "global_step": 183388, "epoch": 2060, "lr": 2.3876183963196364e-05}
{"train_loss": 0.0809609442949295, "global_step": 183389, "epoch": 2060, "lr": 2.3875689650337395e-05}
{"train_loss": 0.14243462681770325, "global_step": 183390, "epoch": 2060, "lr": 2.387519534099051e-05}
{"train_loss": 0.041330330073833466, "global_step": 183391, "epoch": 2060, "lr": 2.38747010351558e-05}
{"train_loss": 0.043001752346754074, "global_step": 183392, "epoch": 2060, "lr": 2.3874206732833298e-05}
{"train_loss": 0.06105189025402069, "global_step": 183393, "epoch": 2060, "lr": 2.387371243402311e-05}
{"train_loss": 0.08858076483011246, "global_step": 183394, "epoch": 2060, "lr": 2.3873218138725267e-05}
{"train_loss": 0.09789656102657318, "global_step": 183395, "epoch": 2060, "lr": 2.3872723846939864e-05}
{"train_loss": 0.08767588436603546, "global_step": 183396, "epoch": 2060, "lr": 2.387222955866694e-05}
{"train_loss": 0.057785600423812866, "global_step": 183397, "epoch": 2060, "lr": 2.38717352739066e-05}
{"train_loss": 0.0608837865293026, "global_step": 183398, "epoch": 2060, "lr": 2.3871240992658862e-05}
{"train_loss": 0.07755762338638306, "global_step": 183399, "epoch": 2060, "lr": 2.3870746714923835e-05}
{"train_loss": 0.08378394693136215, "global_step": 183400, "epoch": 2060, "lr": 2.387025244070156e-05}
{"train_loss": 0.06436218321323395, "global_step": 183401, "epoch": 2060, "lr": 2.3869758169992097e-05}
{"train_loss": 0.05256330221891403, "global_step": 183402, "epoch": 2060, "lr": 2.3869263902795553e-05}
{"train_loss": 0.114723339676857, "global_step": 183403, "epoch": 2060, "lr": 2.3868769639111942e-05}
{"train_loss": 0.09380560368299484, "global_step": 183404, "epoch": 2060, "lr": 2.3868275378941387e-05}
{"train_loss": 0.09590038657188416, "global_step": 183405, "epoch": 2060, "lr": 2.3867781122283894e-05}
{"train_loss": 0.0490606427192688, "global_step": 183406, "epoch": 2060, "lr": 2.386728686913958e-05}
{"train_loss": 0.1059960201382637, "global_step": 183407, "epoch": 2060, "lr": 2.3866792619508478e-05}
{"train_loss": 0.10622026026248932, "global_step": 183408, "epoch": 2060, "lr": 2.3866298373390683e-05}
{"train_loss": 0.018214404582977295, "global_step": 183409, "epoch": 2060, "lr": 2.3865804130786223e-05}
{"train_loss": 0.11202720552682877, "global_step": 183410, "epoch": 2060, "lr": 2.386530989169521e-05}
{"train_loss": 0.03504885360598564, "global_step": 183411, "epoch": 2060, "lr": 2.3864815656117667e-05}
{"train_loss": 0.06573989987373352, "global_step": 183412, "epoch": 2060, "lr": 2.3864321424053705e-05}
{"train_loss": 0.05543564260005951, "global_step": 183413, "epoch": 2060, "lr": 2.386382719550334e-05}
{"train_loss": 0.10718287527561188, "global_step": 183414, "epoch": 2060, "lr": 2.386333297046669e-05}
{"train_loss": 0.047781188040971756, "global_step": 183415, "epoch": 2060, "lr": 2.386283874894379e-05}
{"train_loss": 0.04049587994813919, "global_step": 183416, "epoch": 2060, "lr": 2.38623445309347e-05}
{"train_loss": 0.061220668256282806, "global_step": 183417, "epoch": 2060, "lr": 2.3861850316439514e-05}
{"train_loss": 0.1116190180182457, "global_step": 183418, "epoch": 2060, "lr": 2.3861356105458266e-05}
{"train_loss": 0.08287009596824646, "global_step": 183419, "epoch": 2060, "lr": 2.386086189799106e-05}
{"train_loss": 0.035428717732429504, "global_step": 183420, "epoch": 2060, "lr": 2.3860367694037922e-05}
{"train_loss": 0.08627356588840485, "global_step": 183421, "epoch": 2060, "lr": 2.3859873493598954e-05}
{"train_loss": 0.042483340948820114, "global_step": 183422, "epoch": 2060, "lr": 2.3859379296674194e-05}
{"train_loss": 0.0674881562590599, "global_step": 183423, "epoch": 2060, "lr": 2.3858885103263743e-05}
{"train_loss": 0.12115049362182617, "global_step": 183424, "epoch": 2060, "lr": 2.3858390913367622e-05}
{"train_loss": 0.08402663469314575, "global_step": 183425, "epoch": 2060, "lr": 2.3857896726985946e-05}
{"train_loss": 0.07521042227745056, "global_step": 183426, "epoch": 2060, "lr": 2.3857402544118734e-05}
{"train_loss": 0.07753127068281174, "global_step": 183427, "epoch": 2060, "lr": 2.3856908364766096e-05}
{"train_loss": 0.07508450026592511, "global_step": 183428, "epoch": 2060, "lr": 2.3856414188928055e-05, "val_loss": 7.7476911544799805, "train_action_mse_error": 12.438825607299805}
{"train_loss": 0.06878473609685898, "global_step": 183429, "epoch": 2061, "lr": 2.3855920016604727e-05}
{"train_loss": 0.06542040407657623, "global_step": 183430, "epoch": 2061, "lr": 2.3855425847796124e-05}
{"train_loss": 0.0474579818546772, "global_step": 183431, "epoch": 2061, "lr": 2.385493168250237e-05}
{"train_loss": 0.06575792282819748, "global_step": 183432, "epoch": 2061, "lr": 2.3854437520723477e-05}
{"train_loss": 0.08573857694864273, "global_step": 183433, "epoch": 2061, "lr": 2.3853943362459553e-05}
{"train_loss": 0.09448746591806412, "global_step": 183434, "epoch": 2061, "lr": 2.385344920771063e-05}
{"train_loss": 0.06446024775505066, "global_step": 183435, "epoch": 2061, "lr": 2.3852955056476817e-05}
{"train_loss": 0.059811849147081375, "global_step": 183436, "epoch": 2061, "lr": 2.385246090875813e-05}
{"train_loss": 0.06109381839632988, "global_step": 183437, "epoch": 2061, "lr": 2.385196676455466e-05}
{"train_loss": 0.07959511131048203, "global_step": 183438, "epoch": 2061, "lr": 2.3851472623866495e-05}
{"train_loss": 0.07150724530220032, "global_step": 183439, "epoch": 2061, "lr": 2.3850978486693665e-05}
{"train_loss": 0.0884329155087471, "global_step": 183440, "epoch": 2061, "lr": 2.385048435303627e-05}
{"train_loss": 0.041358619928359985, "global_step": 183441, "epoch": 2061, "lr": 2.3849990222894334e-05}
{"train_loss": 0.12519875168800354, "global_step": 183442, "epoch": 2061, "lr": 2.3849496096267976e-05}
{"train_loss": 0.05146153271198273, "global_step": 183443, "epoch": 2061, "lr": 2.3849001973157208e-05}
{"train_loss": 0.07694992423057556, "global_step": 183444, "epoch": 2061, "lr": 2.3848507853562148e-05}
{"train_loss": 0.07366461306810379, "global_step": 183445, "epoch": 2061, "lr": 2.3848013737482817e-05}
{"train_loss": 0.06619541347026825, "global_step": 183446, "epoch": 2061, "lr": 2.3847519624919316e-05}
{"train_loss": 0.05742524564266205, "global_step": 183447, "epoch": 2061, "lr": 2.384702551587168e-05}
{"train_loss": 0.09436887502670288, "global_step": 183448, "epoch": 2061, "lr": 2.3846531410340016e-05}
{"train_loss": 0.06539840251207352, "global_step": 183449, "epoch": 2061, "lr": 2.3846037308324343e-05}
{"train_loss": 0.098895363509655, "global_step": 183450, "epoch": 2061, "lr": 2.384554320982477e-05}
{"train_loss": 0.04307050630450249, "global_step": 183451, "epoch": 2061, "lr": 2.3845049114841334e-05}
{"train_loss": 0.11017961800098419, "global_step": 183452, "epoch": 2061, "lr": 2.3844555023374127e-05}
{"train_loss": 0.1057811751961708, "global_step": 183453, "epoch": 2061, "lr": 2.3844060935423195e-05}
{"train_loss": 0.08064697682857513, "global_step": 183454, "epoch": 2061, "lr": 2.3843566850988598e-05}
{"train_loss": 0.05636384338140488, "global_step": 183455, "epoch": 2061, "lr": 2.3843072770070435e-05}
{"train_loss": 0.10831530392169952, "global_step": 183456, "epoch": 2061, "lr": 2.3842578692668725e-05}
{"train_loss": 0.06091543287038803, "global_step": 183457, "epoch": 2061, "lr": 2.384208461878359e-05}
{"train_loss": 0.03135932236909866, "global_step": 183458, "epoch": 2061, "lr": 2.3841590548415044e-05}
{"train_loss": 0.08479244261980057, "global_step": 183459, "epoch": 2061, "lr": 2.3841096481563198e-05}
{"train_loss": 0.08733154833316803, "global_step": 183460, "epoch": 2061, "lr": 2.3840602418228076e-05}
{"train_loss": 0.04233970493078232, "global_step": 183461, "epoch": 2061, "lr": 2.3840108358409785e-05}
{"train_loss": 0.07516169548034668, "global_step": 183462, "epoch": 2061, "lr": 2.3839614302108354e-05}
{"train_loss": 0.07543054968118668, "global_step": 183463, "epoch": 2061, "lr": 2.383912024932389e-05}
{"train_loss": 0.061089955270290375, "global_step": 183464, "epoch": 2061, "lr": 2.3838626200056412e-05}
{"train_loss": 0.08924704045057297, "global_step": 183465, "epoch": 2061, "lr": 2.3838132154306037e-05}
{"train_loss": 0.07139855623245239, "global_step": 183466, "epoch": 2061, "lr": 2.3837638112072787e-05}
{"train_loss": 0.0784863531589508, "global_step": 183467, "epoch": 2061, "lr": 2.3837144073356766e-05}
{"train_loss": 0.09943791478872299, "global_step": 183468, "epoch": 2061, "lr": 2.3836650038158e-05}
{"train_loss": 0.07362502068281174, "global_step": 183469, "epoch": 2061, "lr": 2.38361560064766e-05}
{"train_loss": 0.04714541882276535, "global_step": 183470, "epoch": 2061, "lr": 2.383566197831259e-05}
{"train_loss": 0.1286439150571823, "global_step": 183471, "epoch": 2061, "lr": 2.3835167953666072e-05}
{"train_loss": 0.034849442541599274, "global_step": 183472, "epoch": 2061, "lr": 2.383467393253708e-05}
{"train_loss": 0.053638819605112076, "global_step": 183473, "epoch": 2061, "lr": 2.3834179914925703e-05}
{"train_loss": 0.08223037421703339, "global_step": 183474, "epoch": 2061, "lr": 2.3833685900832015e-05}
{"train_loss": 0.057439617812633514, "global_step": 183475, "epoch": 2061, "lr": 2.3833191890256052e-05}
{"train_loss": 0.09628020972013474, "global_step": 183476, "epoch": 2061, "lr": 2.383269788319791e-05}
{"train_loss": 0.05239007994532585, "global_step": 183477, "epoch": 2061, "lr": 2.3832203879657634e-05}
{"train_loss": 0.06080312281847, "global_step": 183478, "epoch": 2061, "lr": 2.3831709879635312e-05}
{"train_loss": 0.0634833499789238, "global_step": 183479, "epoch": 2061, "lr": 2.3831215883130975e-05}
{"train_loss": 0.08172237128019333, "global_step": 183480, "epoch": 2061, "lr": 2.3830721890144735e-05}
{"train_loss": 0.045931898057460785, "global_step": 183481, "epoch": 2061, "lr": 2.3830227900676623e-05}
{"train_loss": 0.0633620172739029, "global_step": 183482, "epoch": 2061, "lr": 2.3829733914726726e-05}
{"train_loss": 0.11218944936990738, "global_step": 183483, "epoch": 2061, "lr": 2.382923993229509e-05}
{"train_loss": 0.10201907902956009, "global_step": 183484, "epoch": 2061, "lr": 2.3828745953381814e-05}
{"train_loss": 0.03122352436184883, "global_step": 183485, "epoch": 2061, "lr": 2.382825197798692e-05}
{"train_loss": 0.04587916284799576, "global_step": 183486, "epoch": 2061, "lr": 2.3827758006110524e-05}
{"train_loss": 0.08068419992923737, "global_step": 183487, "epoch": 2061, "lr": 2.3827264037752637e-05}
{"train_loss": 0.052226606756448746, "global_step": 183488, "epoch": 2061, "lr": 2.3826770072913384e-05}
{"train_loss": 0.10255949944257736, "global_step": 183489, "epoch": 2061, "lr": 2.3826276111592783e-05}
{"train_loss": 0.12698081135749817, "global_step": 183490, "epoch": 2061, "lr": 2.3825782153790936e-05}
{"train_loss": 0.08296164870262146, "global_step": 183491, "epoch": 2061, "lr": 2.3825288199507878e-05}
{"train_loss": 0.10359478741884232, "global_step": 183492, "epoch": 2061, "lr": 2.3824794248743703e-05}
{"train_loss": 0.07758662104606628, "global_step": 183493, "epoch": 2061, "lr": 2.382430030149847e-05}
{"train_loss": 0.04463433474302292, "global_step": 183494, "epoch": 2061, "lr": 2.3823806357772226e-05}
{"train_loss": 0.08876015245914459, "global_step": 183495, "epoch": 2061, "lr": 2.3823312417565063e-05}
{"train_loss": 0.06087389215826988, "global_step": 183496, "epoch": 2061, "lr": 2.3822818480877024e-05}
{"train_loss": 0.07380692660808563, "global_step": 183497, "epoch": 2061, "lr": 2.3822324547708203e-05}
{"train_loss": 0.05645643547177315, "global_step": 183498, "epoch": 2061, "lr": 2.382183061805864e-05}
{"train_loss": 0.0746447890996933, "global_step": 183499, "epoch": 2061, "lr": 2.382133669192842e-05}
{"train_loss": 0.03765329718589783, "global_step": 183500, "epoch": 2061, "lr": 2.3820842769317592e-05}
{"train_loss": 0.08304949104785919, "global_step": 183501, "epoch": 2061, "lr": 2.382034885022625e-05}
{"train_loss": 0.10054422914981842, "global_step": 183502, "epoch": 2061, "lr": 2.381985493465442e-05}
{"train_loss": 0.04408123344182968, "global_step": 183503, "epoch": 2061, "lr": 2.381936102260221e-05}
{"train_loss": 0.06408326327800751, "global_step": 183504, "epoch": 2061, "lr": 2.3818867114069656e-05}
{"train_loss": 0.08027704805135727, "global_step": 183505, "epoch": 2061, "lr": 2.381837320905685e-05}
{"train_loss": 0.03949202969670296, "global_step": 183506, "epoch": 2061, "lr": 2.381787930756383e-05}
{"train_loss": 0.051486000418663025, "global_step": 183507, "epoch": 2061, "lr": 2.3817385409590685e-05}
{"train_loss": 0.09064947068691254, "global_step": 183508, "epoch": 2061, "lr": 2.3816891515137467e-05}
{"train_loss": 0.08340759575366974, "global_step": 183509, "epoch": 2061, "lr": 2.381639762420426e-05}
{"train_loss": 0.08362113684415817, "global_step": 183510, "epoch": 2061, "lr": 2.3815903736791106e-05}
{"train_loss": 0.10573829710483551, "global_step": 183511, "epoch": 2061, "lr": 2.3815409852898078e-05}
{"train_loss": 0.10612600296735764, "global_step": 183512, "epoch": 2061, "lr": 2.3814915972525275e-05}
{"train_loss": 0.08585434406995773, "global_step": 183513, "epoch": 2061, "lr": 2.3814422095672717e-05}
{"train_loss": 0.05892748758196831, "global_step": 183514, "epoch": 2061, "lr": 2.3813928222340503e-05}
{"train_loss": 0.12094275653362274, "global_step": 183515, "epoch": 2061, "lr": 2.3813434352528673e-05}
{"train_loss": 0.06893710047006607, "global_step": 183516, "epoch": 2061, "lr": 2.381294048623733e-05}
{"train_loss": 0.07454203671953652, "global_step": 183517, "epoch": 2061, "lr": 2.3812446623466494e-05, "val_loss": 7.646690845489502}
{"train_loss": 0.07314913719892502, "global_step": 183518, "epoch": 2062, "lr": 2.3811952764216277e-05}
{"train_loss": 0.080703005194664, "global_step": 183519, "epoch": 2062, "lr": 2.3811458908486705e-05}
{"train_loss": 0.07286350429058075, "global_step": 183520, "epoch": 2062, "lr": 2.381096505627788e-05}
{"train_loss": 0.10566553473472595, "global_step": 183521, "epoch": 2062, "lr": 2.3810471207589834e-05}
{"train_loss": 0.10579079389572144, "global_step": 183522, "epoch": 2062, "lr": 2.380997736242267e-05}
{"train_loss": 0.08269796520471573, "global_step": 183523, "epoch": 2062, "lr": 2.3809483520776415e-05}
{"train_loss": 0.09993923455476761, "global_step": 183524, "epoch": 2062, "lr": 2.380898968265118e-05}
{"train_loss": 0.04021678492426872, "global_step": 183525, "epoch": 2062, "lr": 2.380849584804698e-05}
{"train_loss": 0.04106887802481651, "global_step": 183526, "epoch": 2062, "lr": 2.380800201696393e-05}
{"train_loss": 0.10449821501970291, "global_step": 183527, "epoch": 2062, "lr": 2.3807508189402056e-05}
{"train_loss": 0.07759816199541092, "global_step": 183528, "epoch": 2062, "lr": 2.3807014365361464e-05}
{"train_loss": 0.056485239416360855, "global_step": 183529, "epoch": 2062, "lr": 2.380652054484218e-05}
{"train_loss": 0.09310421347618103, "global_step": 183530, "epoch": 2062, "lr": 2.3806026727844312e-05}
{"train_loss": 0.09414523094892502, "global_step": 183531, "epoch": 2062, "lr": 2.38055329143679e-05}
{"train_loss": 0.06129099056124687, "global_step": 183532, "epoch": 2062, "lr": 2.3805039104412996e-05}
{"train_loss": 0.06516373157501221, "global_step": 183533, "epoch": 2062, "lr": 2.3804545297979708e-05}
{"train_loss": 0.07198213785886765, "global_step": 183534, "epoch": 2062, "lr": 2.3804051495068058e-05}
{"train_loss": 0.07710018754005432, "global_step": 183535, "epoch": 2062, "lr": 2.3803557695678154e-05}
{"train_loss": 0.029201431199908257, "global_step": 183536, "epoch": 2062, "lr": 2.380306389981002e-05}
{"train_loss": 0.08072013407945633, "global_step": 183537, "epoch": 2062, "lr": 2.3802570107463767e-05}
{"train_loss": 0.07894450426101685, "global_step": 183538, "epoch": 2062, "lr": 2.380207631863942e-05}
{"train_loss": 0.053265366703271866, "global_step": 183539, "epoch": 2062, "lr": 2.380158253333708e-05}
{"train_loss": 0.095160573720932, "global_step": 183540, "epoch": 2062, "lr": 2.380108875155678e-05}
{"train_loss": 0.0912998840212822, "global_step": 183541, "epoch": 2062, "lr": 2.3800594973298628e-05}
{"train_loss": 0.039689354598522186, "global_step": 183542, "epoch": 2062, "lr": 2.3800101198562648e-05}
{"train_loss": 0.06475784629583359, "global_step": 183543, "epoch": 2062, "lr": 2.3799607427348943e-05}
{"train_loss": 0.08917538821697235, "global_step": 183544, "epoch": 2062, "lr": 2.3799113659657536e-05}
{"train_loss": 0.06327357143163681, "global_step": 183545, "epoch": 2062, "lr": 2.3798619895488546e-05}
{"train_loss": 0.11434956640005112, "global_step": 183546, "epoch": 2062, "lr": 2.3798126134841987e-05}
{"train_loss": 0.07388114184141159, "global_step": 183547, "epoch": 2062, "lr": 2.379763237771795e-05}
{"train_loss": 0.0811154693365097, "global_step": 183548, "epoch": 2062, "lr": 2.379713862411653e-05}
{"train_loss": 0.10407020151615143, "global_step": 183549, "epoch": 2062, "lr": 2.379664487403774e-05}
{"train_loss": 0.07441574335098267, "global_step": 183550, "epoch": 2062, "lr": 2.3796151127481688e-05}
{"train_loss": 0.07762116938829422, "global_step": 183551, "epoch": 2062, "lr": 2.3795657384448406e-05}
{"train_loss": 0.0379672534763813, "global_step": 183552, "epoch": 2062, "lr": 2.3795163644938002e-05}
{"train_loss": 0.0737590566277504, "global_step": 183553, "epoch": 2062, "lr": 2.3794669908950496e-05}
{"train_loss": 0.05941891670227051, "global_step": 183554, "epoch": 2062, "lr": 2.3794176176486e-05}
{"train_loss": 0.08161794394254684, "global_step": 183555, "epoch": 2062, "lr": 2.3793682447544535e-05}
{"train_loss": 0.06648648530244827, "global_step": 183556, "epoch": 2062, "lr": 2.3793188722126208e-05}
{"train_loss": 0.046696290373802185, "global_step": 183557, "epoch": 2062, "lr": 2.3792695000231053e-05}
{"train_loss": 0.07164992392063141, "global_step": 183558, "epoch": 2062, "lr": 2.3792201281859166e-05}
{"train_loss": 0.07343712449073792, "global_step": 183559, "epoch": 2062, "lr": 2.379170756701058e-05}
{"train_loss": 0.058156419545412064, "global_step": 183560, "epoch": 2062, "lr": 2.37912138556854e-05}
{"train_loss": 0.10229310393333435, "global_step": 183561, "epoch": 2062, "lr": 2.379072014788365e-05}
{"train_loss": 0.039636287838220596, "global_step": 183562, "epoch": 2062, "lr": 2.379022644360544e-05}
{"train_loss": 0.08123306185007095, "global_step": 183563, "epoch": 2062, "lr": 2.3789732742850796e-05}
{"train_loss": 0.06481222808361053, "global_step": 183564, "epoch": 2062, "lr": 2.378923904561982e-05}
{"train_loss": 0.0949568822979927, "global_step": 183565, "epoch": 2062, "lr": 2.3788745351912545e-05}
{"train_loss": 0.07342611998319626, "global_step": 183566, "epoch": 2062, "lr": 2.3788251661729067e-05}
{"train_loss": 0.06746654957532883, "global_step": 183567, "epoch": 2062, "lr": 2.3787757975069426e-05}
{"train_loss": 0.11782460659742355, "global_step": 183568, "epoch": 2062, "lr": 2.378726429193372e-05}
{"train_loss": 0.07626603543758392, "global_step": 183569, "epoch": 2062, "lr": 2.378677061232199e-05}
{"train_loss": 0.0804637148976326, "global_step": 183570, "epoch": 2062, "lr": 2.3786276936234296e-05}
{"train_loss": 0.05856053903698921, "global_step": 183571, "epoch": 2062, "lr": 2.3785783263670737e-05}
{"train_loss": 0.06755571812391281, "global_step": 183572, "epoch": 2062, "lr": 2.378528959463134e-05}
{"train_loss": 0.06909873336553574, "global_step": 183573, "epoch": 2062, "lr": 2.378479592911621e-05}
{"train_loss": 0.07744332402944565, "global_step": 183574, "epoch": 2062, "lr": 2.3784302267125373e-05}
{"train_loss": 0.046740490943193436, "global_step": 183575, "epoch": 2062, "lr": 2.3783808608658942e-05}
{"train_loss": 0.039230622351169586, "global_step": 183576, "epoch": 2062, "lr": 2.3783314953716935e-05}
{"train_loss": 0.026144489645957947, "global_step": 183577, "epoch": 2062, "lr": 2.378282130229946e-05}
{"train_loss": 0.025852080434560776, "global_step": 183578, "epoch": 2062, "lr": 2.378232765440655e-05}
{"train_loss": 0.0823659598827362, "global_step": 183579, "epoch": 2062, "lr": 2.3781834010038306e-05}
{"train_loss": 0.10779181867837906, "global_step": 183580, "epoch": 2062, "lr": 2.3781340369194748e-05}
{"train_loss": 0.08326465636491776, "global_step": 183581, "epoch": 2062, "lr": 2.3780846731876e-05}
{"train_loss": 0.05415624380111694, "global_step": 183582, "epoch": 2062, "lr": 2.378035309808207e-05}
{"train_loss": 0.09518016129732132, "global_step": 183583, "epoch": 2062, "lr": 2.3779859467813055e-05}
{"train_loss": 0.13130715489387512, "global_step": 183584, "epoch": 2062, "lr": 2.3779365841069035e-05}
{"train_loss": 0.036362700164318085, "global_step": 183585, "epoch": 2062, "lr": 2.377887221785004e-05}
{"train_loss": 0.11324655264616013, "global_step": 183586, "epoch": 2062, "lr": 2.377837859815618e-05}
{"train_loss": 0.18897397816181183, "global_step": 183587, "epoch": 2062, "lr": 2.3777884981987476e-05}
{"train_loss": 0.10917074978351593, "global_step": 183588, "epoch": 2062, "lr": 2.3777391369344033e-05}
{"train_loss": 0.047901809215545654, "global_step": 183589, "epoch": 2062, "lr": 2.377689776022588e-05}
{"train_loss": 0.07744939625263214, "global_step": 183590, "epoch": 2062, "lr": 2.3776404154633124e-05}
{"train_loss": 0.1366022676229477, "global_step": 183591, "epoch": 2062, "lr": 2.3775910552565793e-05}
{"train_loss": 0.06706562638282776, "global_step": 183592, "epoch": 2062, "lr": 2.3775416954023993e-05}
{"train_loss": 0.0357947014272213, "global_step": 183593, "epoch": 2062, "lr": 2.3774923359007743e-05}
{"train_loss": 0.08454523980617523, "global_step": 183594, "epoch": 2062, "lr": 2.377442976751716e-05}
{"train_loss": 0.07194431871175766, "global_step": 183595, "epoch": 2062, "lr": 2.3773936179552263e-05}
{"train_loss": 0.0644460991024971, "global_step": 183596, "epoch": 2062, "lr": 2.3773442595113165e-05}
{"train_loss": 0.05692658945918083, "global_step": 183597, "epoch": 2062, "lr": 2.377294901419988e-05}
{"train_loss": 0.08105216175317764, "global_step": 183598, "epoch": 2062, "lr": 2.377245543681253e-05}
{"train_loss": 0.09389714896678925, "global_step": 183599, "epoch": 2062, "lr": 2.377196186295113e-05}
{"train_loss": 0.12709230184555054, "global_step": 183600, "epoch": 2062, "lr": 2.377146829261577e-05}
{"train_loss": 0.04214101657271385, "global_step": 183601, "epoch": 2062, "lr": 2.3770974725806533e-05}
{"train_loss": 0.033381786197423935, "global_step": 183602, "epoch": 2062, "lr": 2.3770481162523454e-05}
{"train_loss": 0.057588815689086914, "global_step": 183603, "epoch": 2062, "lr": 2.3769987602766632e-05}
{"train_loss": 0.10296764224767685, "global_step": 183604, "epoch": 2062, "lr": 2.3769494046536116e-05}
{"train_loss": 0.09303280711174011, "global_step": 183605, "epoch": 2062, "lr": 2.376900049383195e-05}
{"train_loss": 0.07552889765899502, "global_step": 183606, "epoch": 2062, "lr": 2.3768506944654244e-05, "val_loss": 7.637453556060791}
{"train_loss": 0.02534174546599388, "global_step": 183607, "epoch": 2063, "lr": 2.3768013399003024e-05}
{"train_loss": 0.10132471472024918, "global_step": 183608, "epoch": 2063, "lr": 2.376751985687839e-05}
{"train_loss": 0.06546692550182343, "global_step": 183609, "epoch": 2063, "lr": 2.37670263182804e-05}
{"train_loss": 0.08568819612264633, "global_step": 183610, "epoch": 2063, "lr": 2.376653278320909e-05}
{"train_loss": 0.053189996629953384, "global_step": 183611, "epoch": 2063, "lr": 2.3766039251664567e-05}
{"train_loss": 0.040120214223861694, "global_step": 183612, "epoch": 2063, "lr": 2.3765545723646865e-05}
{"train_loss": 0.04414541274309158, "global_step": 183613, "epoch": 2063, "lr": 2.3765052199156083e-05}
{"train_loss": 0.09684040397405624, "global_step": 183614, "epoch": 2063, "lr": 2.376455867819225e-05}
{"train_loss": 0.08173736929893494, "global_step": 183615, "epoch": 2063, "lr": 2.3764065160755472e-05}
{"train_loss": 0.10535936057567596, "global_step": 183616, "epoch": 2063, "lr": 2.3763571646845778e-05}
{"train_loss": 0.12254748493432999, "global_step": 183617, "epoch": 2063, "lr": 2.376307813646327e-05}
{"train_loss": 0.11386220902204514, "global_step": 183618, "epoch": 2063, "lr": 2.3762584629607977e-05}
{"train_loss": 0.11073669791221619, "global_step": 183619, "epoch": 2063, "lr": 2.3762091126279984e-05}
{"train_loss": 0.10685237497091293, "global_step": 183620, "epoch": 2063, "lr": 2.3761597626479383e-05}
{"train_loss": 0.1710224449634552, "global_step": 183621, "epoch": 2063, "lr": 2.3761104130206192e-05}
{"train_loss": 0.07690230011940002, "global_step": 183622, "epoch": 2063, "lr": 2.3760610637460518e-05}
{"train_loss": 0.05006296560168266, "global_step": 183623, "epoch": 2063, "lr": 2.376011714824239e-05}
{"train_loss": 0.08644995093345642, "global_step": 183624, "epoch": 2063, "lr": 2.3759623662551917e-05}
{"train_loss": 0.06357162445783615, "global_step": 183625, "epoch": 2063, "lr": 2.3759130180389123e-05}
{"train_loss": 0.0899706482887268, "global_step": 183626, "epoch": 2063, "lr": 2.375863670175411e-05}
{"train_loss": 0.08720976859331131, "global_step": 183627, "epoch": 2063, "lr": 2.3758143226646918e-05}
{"train_loss": 0.10458584129810333, "global_step": 183628, "epoch": 2063, "lr": 2.3757649755067634e-05}
{"train_loss": 0.07018314301967621, "global_step": 183629, "epoch": 2063, "lr": 2.3757156287016297e-05}
{"train_loss": 0.08992470800876617, "global_step": 183630, "epoch": 2063, "lr": 2.3756662822493015e-05}
{"train_loss": 0.05013594776391983, "global_step": 183631, "epoch": 2063, "lr": 2.3756169361497805e-05}
{"train_loss": 0.03935740143060684, "global_step": 183632, "epoch": 2063, "lr": 2.375567590403078e-05}
{"train_loss": 0.11427869647741318, "global_step": 183633, "epoch": 2063, "lr": 2.3755182450091963e-05}
{"train_loss": 0.04426873102784157, "global_step": 183634, "epoch": 2063, "lr": 2.3754688999681463e-05}
{"train_loss": 0.11562870442867279, "global_step": 183635, "epoch": 2063, "lr": 2.3754195552799302e-05}
{"train_loss": 0.09878221154212952, "global_step": 183636, "epoch": 2063, "lr": 2.3753702109445593e-05}
{"train_loss": 0.0874597579240799, "global_step": 183637, "epoch": 2063, "lr": 2.375320866962036e-05}
{"train_loss": 0.14688216149806976, "global_step": 183638, "epoch": 2063, "lr": 2.3752715233323687e-05}
{"train_loss": 0.10198821872472763, "global_step": 183639, "epoch": 2063, "lr": 2.3752221800555657e-05}
{"train_loss": 0.11631741374731064, "global_step": 183640, "epoch": 2063, "lr": 2.3751728371316308e-05}
{"train_loss": 0.09623056650161743, "global_step": 183641, "epoch": 2063, "lr": 2.3751234945605732e-05}
{"train_loss": 0.07068350911140442, "global_step": 183642, "epoch": 2063, "lr": 2.375074152342399e-05}
{"train_loss": 0.08766105771064758, "global_step": 183643, "epoch": 2063, "lr": 2.3750248104771118e-05}
{"train_loss": 0.05549587309360504, "global_step": 183644, "epoch": 2063, "lr": 2.3749754689647224e-05}
{"train_loss": 0.0638301819562912, "global_step": 183645, "epoch": 2063, "lr": 2.3749261278052337e-05}
{"train_loss": 0.07357928156852722, "global_step": 183646, "epoch": 2063, "lr": 2.374876786998656e-05}
{"train_loss": 0.09807774424552917, "global_step": 183647, "epoch": 2063, "lr": 2.3748274465449937e-05}
{"train_loss": 0.08802630007266998, "global_step": 183648, "epoch": 2063, "lr": 2.3747781064442525e-05}
{"train_loss": 0.06780803203582764, "global_step": 183649, "epoch": 2063, "lr": 2.3747287666964425e-05}
{"train_loss": 0.08416024595499039, "global_step": 183650, "epoch": 2063, "lr": 2.374679427301566e-05}
{"train_loss": 0.11293748021125793, "global_step": 183651, "epoch": 2063, "lr": 2.374630088259634e-05}
{"train_loss": 0.06771110743284225, "global_step": 183652, "epoch": 2063, "lr": 2.3745807495706484e-05}
{"train_loss": 0.14635397493839264, "global_step": 183653, "epoch": 2063, "lr": 2.374531411234621e-05}
{"train_loss": 0.1266207993030548, "global_step": 183654, "epoch": 2063, "lr": 2.3744820732515538e-05}
{"train_loss": 0.06779205799102783, "global_step": 183655, "epoch": 2063, "lr": 2.3744327356214552e-05}
{"train_loss": 0.08831575512886047, "global_step": 183656, "epoch": 2063, "lr": 2.3743833983443343e-05}
{"train_loss": 0.10730588436126709, "global_step": 183657, "epoch": 2063, "lr": 2.3743340614201936e-05}
{"train_loss": 0.06681489944458008, "global_step": 183658, "epoch": 2063, "lr": 2.3742847248490435e-05}
{"train_loss": 0.07365760207176208, "global_step": 183659, "epoch": 2063, "lr": 2.3742353886308872e-05}
{"train_loss": 0.059138886630535126, "global_step": 183660, "epoch": 2063, "lr": 2.3741860527657346e-05}
{"train_loss": 0.08104173839092255, "global_step": 183661, "epoch": 2063, "lr": 2.3741367172535887e-05}
{"train_loss": 0.10749582201242447, "global_step": 183662, "epoch": 2063, "lr": 2.3740873820944605e-05}
{"train_loss": 0.041369713842868805, "global_step": 183663, "epoch": 2063, "lr": 2.3740380472883518e-05}
{"train_loss": 0.07061076164245605, "global_step": 183664, "epoch": 2063, "lr": 2.373988712835274e-05}
{"train_loss": 0.050526347011327744, "global_step": 183665, "epoch": 2063, "lr": 2.373939378735229e-05}
{"train_loss": 0.10273321717977524, "global_step": 183666, "epoch": 2063, "lr": 2.373890044988228e-05}
{"train_loss": 0.08616384118795395, "global_step": 183667, "epoch": 2063, "lr": 2.3738407115942736e-05}
{"train_loss": 0.06163458898663521, "global_step": 183668, "epoch": 2063, "lr": 2.3737913785533762e-05}
{"train_loss": 0.08281619101762772, "global_step": 183669, "epoch": 2063, "lr": 2.3737420458655384e-05}
{"train_loss": 0.04596554487943649, "global_step": 183670, "epoch": 2063, "lr": 2.3736927135307713e-05}
{"train_loss": 0.07543357461690903, "global_step": 183671, "epoch": 2063, "lr": 2.373643381549077e-05}
{"train_loss": 0.07892540842294693, "global_step": 183672, "epoch": 2063, "lr": 2.373594049920466e-05}
{"train_loss": 0.0734783262014389, "global_step": 183673, "epoch": 2063, "lr": 2.3735447186449423e-05}
{"train_loss": 0.08142909407615662, "global_step": 183674, "epoch": 2063, "lr": 2.3734953877225125e-05}
{"train_loss": 0.10659441351890564, "global_step": 183675, "epoch": 2063, "lr": 2.3734460571531868e-05}
{"train_loss": 0.05603119358420372, "global_step": 183676, "epoch": 2063, "lr": 2.3733967269369674e-05}
{"train_loss": 0.0628269836306572, "global_step": 183677, "epoch": 2063, "lr": 2.373347397073864e-05}
{"train_loss": 0.041941747069358826, "global_step": 183678, "epoch": 2063, "lr": 2.3732980675638805e-05}
{"train_loss": 0.11898276209831238, "global_step": 183679, "epoch": 2063, "lr": 2.3732487384070273e-05}
{"train_loss": 0.0652163103222847, "global_step": 183680, "epoch": 2063, "lr": 2.3731994096033085e-05}
{"train_loss": 0.14094431698322296, "global_step": 183681, "epoch": 2063, "lr": 2.373150081152729e-05}
{"train_loss": 0.10271847993135452, "global_step": 183682, "epoch": 2063, "lr": 2.3731007530552994e-05}
{"train_loss": 0.07762967795133591, "global_step": 183683, "epoch": 2063, "lr": 2.3730514253110227e-05}
{"train_loss": 0.04867023602128029, "global_step": 183684, "epoch": 2063, "lr": 2.3730020979199086e-05}
{"train_loss": 0.14553587138652802, "global_step": 183685, "epoch": 2063, "lr": 2.3729527708819628e-05}
{"train_loss": 0.09499876201152802, "global_step": 183686, "epoch": 2063, "lr": 2.3729034441971898e-05}
{"train_loss": 0.07428599894046783, "global_step": 183687, "epoch": 2063, "lr": 2.3728541178655993e-05}
{"train_loss": 0.07825068384408951, "global_step": 183688, "epoch": 2063, "lr": 2.372804791887195e-05}
{"train_loss": 0.05584387853741646, "global_step": 183689, "epoch": 2063, "lr": 2.3727554662619865e-05}
{"train_loss": 0.08864966034889221, "global_step": 183690, "epoch": 2063, "lr": 2.3727061409899777e-05}
{"train_loss": 0.04413627088069916, "global_step": 183691, "epoch": 2063, "lr": 2.372656816071176e-05}
{"train_loss": 0.11885767430067062, "global_step": 183692, "epoch": 2063, "lr": 2.3726074915055905e-05}
{"train_loss": 0.08847808837890625, "global_step": 183693, "epoch": 2063, "lr": 2.3725581672932247e-05}
{"train_loss": 0.07690516114234924, "global_step": 183694, "epoch": 2063, "lr": 2.3725088434340875e-05}
{"train_loss": 0.08397898124007697, "global_step": 183695, "epoch": 2063, "lr": 2.3724595199281834e-05, "val_loss": 7.394002437591553}
{"train_loss": 0.03530965372920036, "global_step": 183696, "epoch": 2064, "lr": 2.3724101967755215e-05}
{"train_loss": 0.10917456448078156, "global_step": 183697, "epoch": 2064, "lr": 2.372360873976105e-05}
{"train_loss": 0.046443842351436615, "global_step": 183698, "epoch": 2064, "lr": 2.3723115515299443e-05}
{"train_loss": 0.14428488910198212, "global_step": 183699, "epoch": 2064, "lr": 2.3722622294370426e-05}
{"train_loss": 0.04876535013318062, "global_step": 183700, "epoch": 2064, "lr": 2.3722129076974103e-05}
{"train_loss": 0.11142434179782867, "global_step": 183701, "epoch": 2064, "lr": 2.3721635863110502e-05}
{"train_loss": 0.07085417956113815, "global_step": 183702, "epoch": 2064, "lr": 2.372114265277972e-05}
{"train_loss": 0.12425264716148376, "global_step": 183703, "epoch": 2064, "lr": 2.3720649445981795e-05}
{"train_loss": 0.13607452809810638, "global_step": 183704, "epoch": 2064, "lr": 2.372015624271683e-05}
{"train_loss": 0.09645921736955643, "global_step": 183705, "epoch": 2064, "lr": 2.3719663042984847e-05}
{"train_loss": 0.0964067131280899, "global_step": 183706, "epoch": 2064, "lr": 2.3719169846785956e-05}
{"train_loss": 0.03514695167541504, "global_step": 183707, "epoch": 2064, "lr": 2.3718676654120182e-05}
{"train_loss": 0.08241236209869385, "global_step": 183708, "epoch": 2064, "lr": 2.3718183464987626e-05}
{"train_loss": 0.08140866458415985, "global_step": 183709, "epoch": 2064, "lr": 2.3717690279388326e-05}
{"train_loss": 0.10501064360141754, "global_step": 183710, "epoch": 2064, "lr": 2.3717197097322363e-05}
{"train_loss": 0.03936275467276573, "global_step": 183711, "epoch": 2064, "lr": 2.371670391878982e-05}
{"train_loss": 0.03613927587866783, "global_step": 183712, "epoch": 2064, "lr": 2.371621074379073e-05}
{"train_loss": 0.0768328309059143, "global_step": 183713, "epoch": 2064, "lr": 2.3715717572325184e-05}
{"train_loss": 0.07984308153390884, "global_step": 183714, "epoch": 2064, "lr": 2.3715224404393233e-05}
{"train_loss": 0.07541592419147491, "global_step": 183715, "epoch": 2064, "lr": 2.3714731239994958e-05}
{"train_loss": 0.07090961933135986, "global_step": 183716, "epoch": 2064, "lr": 2.371423807913041e-05}
{"train_loss": 0.08485911786556244, "global_step": 183717, "epoch": 2064, "lr": 2.3713744921799668e-05}
{"train_loss": 0.1397576481103897, "global_step": 183718, "epoch": 2064, "lr": 2.3713251768002782e-05}
{"train_loss": 0.048948537558317184, "global_step": 183719, "epoch": 2064, "lr": 2.3712758617739844e-05}
{"train_loss": 0.053416453301906586, "global_step": 183720, "epoch": 2064, "lr": 2.371226547101091e-05}
{"train_loss": 0.061312705278396606, "global_step": 183721, "epoch": 2064, "lr": 2.3711772327816017e-05}
{"train_loss": 0.045984070748090744, "global_step": 183722, "epoch": 2064, "lr": 2.3711279188155278e-05}
{"train_loss": 0.05251696705818176, "global_step": 183723, "epoch": 2064, "lr": 2.3710786052028716e-05}
{"train_loss": 0.08920209854841232, "global_step": 183724, "epoch": 2064, "lr": 2.371029291943644e-05}
{"train_loss": 0.07605317234992981, "global_step": 183725, "epoch": 2064, "lr": 2.3709799790378497e-05}
{"train_loss": 0.10445782542228699, "global_step": 183726, "epoch": 2064, "lr": 2.3709306664854924e-05}
{"train_loss": 0.07926832884550095, "global_step": 183727, "epoch": 2064, "lr": 2.370881354286582e-05}
{"train_loss": 0.07812433689832687, "global_step": 183728, "epoch": 2064, "lr": 2.370832042441126e-05}
{"train_loss": 0.052028216421604156, "global_step": 183729, "epoch": 2064, "lr": 2.370782730949128e-05}
{"train_loss": 0.07129041105508804, "global_step": 183730, "epoch": 2064, "lr": 2.3707334198105986e-05}
{"train_loss": 0.046840179711580276, "global_step": 183731, "epoch": 2064, "lr": 2.3706841090255393e-05}
{"train_loss": 0.09408068656921387, "global_step": 183732, "epoch": 2064, "lr": 2.370634798593962e-05}
{"train_loss": 0.046308159828186035, "global_step": 183733, "epoch": 2064, "lr": 2.3705854885158683e-05}
{"train_loss": 0.06539468467235565, "global_step": 183734, "epoch": 2064, "lr": 2.37053617879127e-05}
{"train_loss": 0.03189195320010185, "global_step": 183735, "epoch": 2064, "lr": 2.3704868694201688e-05}
{"train_loss": 0.08650176972150803, "global_step": 183736, "epoch": 2064, "lr": 2.3704375604025752e-05}
{"train_loss": 0.06569625437259674, "global_step": 183737, "epoch": 2064, "lr": 2.370388251738493e-05}
{"train_loss": 0.03816593438386917, "global_step": 183738, "epoch": 2064, "lr": 2.3703389434279312e-05}
{"train_loss": 0.09061684459447861, "global_step": 183739, "epoch": 2064, "lr": 2.3702896354708937e-05}
{"train_loss": 0.05531798303127289, "global_step": 183740, "epoch": 2064, "lr": 2.3702403278673908e-05}
{"train_loss": 0.11149374395608902, "global_step": 183741, "epoch": 2064, "lr": 2.3701910206174248e-05}
{"train_loss": 0.047107014805078506, "global_step": 183742, "epoch": 2064, "lr": 2.3701417137210064e-05}
{"train_loss": 0.02853621356189251, "global_step": 183743, "epoch": 2064, "lr": 2.3700924071781388e-05}
{"train_loss": 0.07151303440332413, "global_step": 183744, "epoch": 2064, "lr": 2.370043100988832e-05}
{"train_loss": 0.0473727211356163, "global_step": 183745, "epoch": 2064, "lr": 2.3699937951530888e-05}
{"train_loss": 0.032591573894023895, "global_step": 183746, "epoch": 2064, "lr": 2.3699444896709182e-05}
{"train_loss": 0.06592255085706711, "global_step": 183747, "epoch": 2064, "lr": 2.3698951845423284e-05}
{"train_loss": 0.06990215927362442, "global_step": 183748, "epoch": 2064, "lr": 2.369845879767322e-05}
{"train_loss": 0.06270869076251984, "global_step": 183749, "epoch": 2064, "lr": 2.3697965753459093e-05}
{"train_loss": 0.06652960926294327, "global_step": 183750, "epoch": 2064, "lr": 2.3697472712780944e-05}
{"train_loss": 0.07761436700820923, "global_step": 183751, "epoch": 2064, "lr": 2.3696979675638866e-05}
{"train_loss": 0.09097842127084732, "global_step": 183752, "epoch": 2064, "lr": 2.3696486642032882e-05}
{"train_loss": 0.08139078319072723, "global_step": 183753, "epoch": 2064, "lr": 2.3695993611963112e-05}
{"train_loss": 0.06492233276367188, "global_step": 183754, "epoch": 2064, "lr": 2.369550058542957e-05}
{"train_loss": 0.05689040943980217, "global_step": 183755, "epoch": 2064, "lr": 2.369500756243237e-05}
{"train_loss": 0.11138098686933517, "global_step": 183756, "epoch": 2064, "lr": 2.3694514542971534e-05}
{"train_loss": 0.08585518598556519, "global_step": 183757, "epoch": 2064, "lr": 2.369402152704717e-05}
{"train_loss": 0.08253113180398941, "global_step": 183758, "epoch": 2064, "lr": 2.369352851465933e-05}
{"train_loss": 0.042804788798093796, "global_step": 183759, "epoch": 2064, "lr": 2.3693035505808046e-05}
{"train_loss": 0.0780397579073906, "global_step": 183760, "epoch": 2064, "lr": 2.369254250049344e-05}
{"train_loss": 0.09039107710123062, "global_step": 183761, "epoch": 2064, "lr": 2.369204949871553e-05}
{"train_loss": 0.20381683111190796, "global_step": 183762, "epoch": 2064, "lr": 2.3691556500474416e-05}
{"train_loss": 0.13523037731647491, "global_step": 183763, "epoch": 2064, "lr": 2.3691063505770156e-05}
{"train_loss": 0.08035945892333984, "global_step": 183764, "epoch": 2064, "lr": 2.3690570514602793e-05}
{"train_loss": 0.054685790091753006, "global_step": 183765, "epoch": 2064, "lr": 2.3690077526972416e-05}
{"train_loss": 0.10317815095186234, "global_step": 183766, "epoch": 2064, "lr": 2.36895845428791e-05}
{"train_loss": 0.05207240954041481, "global_step": 183767, "epoch": 2064, "lr": 2.3689091562322885e-05}
{"train_loss": 0.15254178643226624, "global_step": 183768, "epoch": 2064, "lr": 2.3688598585303866e-05}
{"train_loss": 0.09943997859954834, "global_step": 183769, "epoch": 2064, "lr": 2.3688105611822077e-05}
{"train_loss": 0.07159030437469482, "global_step": 183770, "epoch": 2064, "lr": 2.3687612641877616e-05}
{"train_loss": 0.11791110783815384, "global_step": 183771, "epoch": 2064, "lr": 2.368711967547052e-05}
{"train_loss": 0.11558036506175995, "global_step": 183772, "epoch": 2064, "lr": 2.368662671260089e-05}
{"train_loss": 0.026689639315009117, "global_step": 183773, "epoch": 2064, "lr": 2.3686133753268752e-05}
{"train_loss": 0.07804275304079056, "global_step": 183774, "epoch": 2064, "lr": 2.3685640797474213e-05}
{"train_loss": 0.09657180309295654, "global_step": 183775, "epoch": 2064, "lr": 2.3685147845217298e-05}
{"train_loss": 0.10154125094413757, "global_step": 183776, "epoch": 2064, "lr": 2.3684654896498114e-05}
{"train_loss": 0.06424935907125473, "global_step": 183777, "epoch": 2064, "lr": 2.3684161951316685e-05}
{"train_loss": 0.07435064762830734, "global_step": 183778, "epoch": 2064, "lr": 2.3683669009673127e-05}
{"train_loss": 0.04311329871416092, "global_step": 183779, "epoch": 2064, "lr": 2.3683176071567452e-05}
{"train_loss": 0.08219724148511887, "global_step": 183780, "epoch": 2064, "lr": 2.3682683136999773e-05}
{"train_loss": 0.07564030587673187, "global_step": 183781, "epoch": 2064, "lr": 2.3682190205970123e-05}
{"train_loss": 0.09209366887807846, "global_step": 183782, "epoch": 2064, "lr": 2.3681697278478577e-05}
{"train_loss": 0.0958368107676506, "global_step": 183783, "epoch": 2064, "lr": 2.3681204354525227e-05}
{"train_loss": 0.07760786373963517, "global_step": 183784, "epoch": 2064, "lr": 2.3680711434110096e-05, "val_loss": 7.582284450531006}
{"train_loss": 0.10959552973508835, "global_step": 183785, "epoch": 2065, "lr": 2.3680218517233293e-05}
{"train_loss": 0.041592199355363846, "global_step": 183786, "epoch": 2065, "lr": 2.3679725603894848e-05}
{"train_loss": 0.08440759032964706, "global_step": 183787, "epoch": 2065, "lr": 2.367923269409486e-05}
{"train_loss": 0.062441907823085785, "global_step": 183788, "epoch": 2065, "lr": 2.3678739787833358e-05}
{"train_loss": 0.08636573702096939, "global_step": 183789, "epoch": 2065, "lr": 2.3678246885110448e-05}
{"train_loss": 0.05600270628929138, "global_step": 183790, "epoch": 2065, "lr": 2.367775398592616e-05}
{"train_loss": 0.06037669628858566, "global_step": 183791, "epoch": 2065, "lr": 2.3677261090280594e-05}
{"train_loss": 0.0643058493733406, "global_step": 183792, "epoch": 2065, "lr": 2.3676768198173782e-05}
{"train_loss": 0.09960849583148956, "global_step": 183793, "epoch": 2065, "lr": 2.3676275309605827e-05}
{"train_loss": 0.028079841285943985, "global_step": 183794, "epoch": 2065, "lr": 2.3675782424576754e-05}
{"train_loss": 0.057611625641584396, "global_step": 183795, "epoch": 2065, "lr": 2.367528954308667e-05}
{"train_loss": 0.08273773640394211, "global_step": 183796, "epoch": 2065, "lr": 2.367479666513563e-05}
{"train_loss": 0.06231090426445007, "global_step": 183797, "epoch": 2065, "lr": 2.3674303790723667e-05}
{"train_loss": 0.03833013027906418, "global_step": 183798, "epoch": 2065, "lr": 2.367381091985089e-05}
{"train_loss": 0.0684146136045456, "global_step": 183799, "epoch": 2065, "lr": 2.3673318052517335e-05}
{"train_loss": 0.13770484924316406, "global_step": 183800, "epoch": 2065, "lr": 2.3672825188723098e-05}
{"train_loss": 0.062135227024555206, "global_step": 183801, "epoch": 2065, "lr": 2.367233232846821e-05}
{"train_loss": 0.0732356607913971, "global_step": 183802, "epoch": 2065, "lr": 2.3671839471752777e-05}
{"train_loss": 0.07931153476238251, "global_step": 183803, "epoch": 2065, "lr": 2.3671346618576822e-05}
{"train_loss": 0.10685960948467255, "global_step": 183804, "epoch": 2065, "lr": 2.3670853768940455e-05}
{"train_loss": 0.08653126657009125, "global_step": 183805, "epoch": 2065, "lr": 2.3670360922843697e-05}
{"train_loss": 0.04644755646586418, "global_step": 183806, "epoch": 2065, "lr": 2.3669868080286663e-05}
{"train_loss": 0.07849813997745514, "global_step": 183807, "epoch": 2065, "lr": 2.3669375241269372e-05}
{"train_loss": 0.13270092010498047, "global_step": 183808, "epoch": 2065, "lr": 2.366888240579193e-05}
{"train_loss": 0.11750668287277222, "global_step": 183809, "epoch": 2065, "lr": 2.366838957385437e-05}
{"train_loss": 0.08005817234516144, "global_step": 183810, "epoch": 2065, "lr": 2.3667896745456797e-05}
{"train_loss": 0.06402862817049026, "global_step": 183811, "epoch": 2065, "lr": 2.366740392059923e-05}
{"train_loss": 0.05621649697422981, "global_step": 183812, "epoch": 2065, "lr": 2.366691109928178e-05}
{"train_loss": 0.10145318508148193, "global_step": 183813, "epoch": 2065, "lr": 2.366641828150447e-05}
{"train_loss": 0.10107345879077911, "global_step": 183814, "epoch": 2065, "lr": 2.3665925467267414e-05}
{"train_loss": 0.06225336343050003, "global_step": 183815, "epoch": 2065, "lr": 2.366543265657063e-05}
{"train_loss": 0.03276372328400612, "global_step": 183816, "epoch": 2065, "lr": 2.366493984941423e-05}
{"train_loss": 0.046377409249544144, "global_step": 183817, "epoch": 2065, "lr": 2.3664447045798233e-05}
{"train_loss": 0.06280437111854553, "global_step": 183818, "epoch": 2065, "lr": 2.3663954245722737e-05}
{"train_loss": 0.12673041224479675, "global_step": 183819, "epoch": 2065, "lr": 2.3663461449187818e-05}
{"train_loss": 0.09303903579711914, "global_step": 183820, "epoch": 2065, "lr": 2.3662968656193506e-05}
{"train_loss": 0.08281821012496948, "global_step": 183821, "epoch": 2065, "lr": 2.36624758667399e-05}
{"train_loss": 0.09436903148889542, "global_step": 183822, "epoch": 2065, "lr": 2.3661983080827043e-05}
{"train_loss": 0.10693347454071045, "global_step": 183823, "epoch": 2065, "lr": 2.366149029845503e-05}
{"train_loss": 0.09994838386774063, "global_step": 183824, "epoch": 2065, "lr": 2.3660997519623886e-05}
{"train_loss": 0.0889311134815216, "global_step": 183825, "epoch": 2065, "lr": 2.3660504744333722e-05}
{"train_loss": 0.10068777203559875, "global_step": 183826, "epoch": 2065, "lr": 2.3660011972584562e-05}
{"train_loss": 0.06901183724403381, "global_step": 183827, "epoch": 2065, "lr": 2.3659519204376505e-05}
{"train_loss": 0.07786350697278976, "global_step": 183828, "epoch": 2065, "lr": 2.36590264397096e-05}
{"train_loss": 0.09962566196918488, "global_step": 183829, "epoch": 2065, "lr": 2.3658533678583923e-05}
{"train_loss": 0.10606418550014496, "global_step": 183830, "epoch": 2065, "lr": 2.365804092099952e-05}
{"train_loss": 0.09948654472827911, "global_step": 183831, "epoch": 2065, "lr": 2.3657548166956496e-05}
{"train_loss": 0.11348506063222885, "global_step": 183832, "epoch": 2065, "lr": 2.3657055416454866e-05}
{"train_loss": 0.12455496191978455, "global_step": 183833, "epoch": 2065, "lr": 2.3656562669494746e-05}
{"train_loss": 0.15219104290008545, "global_step": 183834, "epoch": 2065, "lr": 2.365606992607617e-05}
{"train_loss": 0.05927245318889618, "global_step": 183835, "epoch": 2065, "lr": 2.3655577186199222e-05}
{"train_loss": 0.10237565636634827, "global_step": 183836, "epoch": 2065, "lr": 2.365508444986396e-05}
{"train_loss": 0.1237010806798935, "global_step": 183837, "epoch": 2065, "lr": 2.365459171707044e-05}
{"train_loss": 0.08146224170923233, "global_step": 183838, "epoch": 2065, "lr": 2.3654098987818757e-05}
{"train_loss": 0.05468594655394554, "global_step": 183839, "epoch": 2065, "lr": 2.3653606262108934e-05}
{"train_loss": 0.05209822207689285, "global_step": 183840, "epoch": 2065, "lr": 2.3653113539941086e-05}
{"train_loss": 0.05844338610768318, "global_step": 183841, "epoch": 2065, "lr": 2.3652620821315237e-05}
{"train_loss": 0.12812440097332, "global_step": 183842, "epoch": 2065, "lr": 2.365212810623149e-05}
{"train_loss": 0.07209561765193939, "global_step": 183843, "epoch": 2065, "lr": 2.3651635394689874e-05}
{"train_loss": 0.06729379296302795, "global_step": 183844, "epoch": 2065, "lr": 2.3651142686690498e-05}
{"train_loss": 0.07070157676935196, "global_step": 183845, "epoch": 2065, "lr": 2.365064998223338e-05}
{"train_loss": 0.05521097779273987, "global_step": 183846, "epoch": 2065, "lr": 2.365015728131863e-05}
{"train_loss": 0.09861347824335098, "global_step": 183847, "epoch": 2065, "lr": 2.364966458394628e-05}
{"train_loss": 0.08366631716489792, "global_step": 183848, "epoch": 2065, "lr": 2.3649171890116422e-05}
{"train_loss": 0.04312114045023918, "global_step": 183849, "epoch": 2065, "lr": 2.3648679199829105e-05}
{"train_loss": 0.06308134645223618, "global_step": 183850, "epoch": 2065, "lr": 2.3648186513084414e-05}
{"train_loss": 0.03463149443268776, "global_step": 183851, "epoch": 2065, "lr": 2.3647693829882385e-05}
{"train_loss": 0.059762828052043915, "global_step": 183852, "epoch": 2065, "lr": 2.3647201150223118e-05}
{"train_loss": 0.03915269672870636, "global_step": 183853, "epoch": 2065, "lr": 2.364670847410665e-05}
{"train_loss": 0.1098519116640091, "global_step": 183854, "epoch": 2065, "lr": 2.3646215801533056e-05}
{"train_loss": 0.06782984733581543, "global_step": 183855, "epoch": 2065, "lr": 2.3645723132502427e-05}
{"train_loss": 0.08177816867828369, "global_step": 183856, "epoch": 2065, "lr": 2.3645230467014794e-05}
{"train_loss": 0.12098728120326996, "global_step": 183857, "epoch": 2065, "lr": 2.3644737805070256e-05}
{"train_loss": 0.11299274861812592, "global_step": 183858, "epoch": 2065, "lr": 2.3644245146668843e-05}
{"train_loss": 0.10689359158277512, "global_step": 183859, "epoch": 2065, "lr": 2.3643752491810654e-05}
{"train_loss": 0.05491483956575394, "global_step": 183860, "epoch": 2065, "lr": 2.364325984049573e-05}
{"train_loss": 0.06060957908630371, "global_step": 183861, "epoch": 2065, "lr": 2.3642767192724163e-05}
{"train_loss": 0.0636090412735939, "global_step": 183862, "epoch": 2065, "lr": 2.364227454849599e-05}
{"train_loss": 0.04470179229974747, "global_step": 183863, "epoch": 2065, "lr": 2.3641781907811304e-05}
{"train_loss": 0.13131499290466309, "global_step": 183864, "epoch": 2065, "lr": 2.3641289270670147e-05}
{"train_loss": 0.08292372524738312, "global_step": 183865, "epoch": 2065, "lr": 2.364079663707261e-05}
{"train_loss": 0.06197673827409744, "global_step": 183866, "epoch": 2065, "lr": 2.364030400701873e-05}
{"train_loss": 0.10229642689228058, "global_step": 183867, "epoch": 2065, "lr": 2.3639811380508613e-05}
{"train_loss": 0.049144282937049866, "global_step": 183868, "epoch": 2065, "lr": 2.3639318757542278e-05}
{"train_loss": 0.07490212470293045, "global_step": 183869, "epoch": 2065, "lr": 2.3638826138119834e-05}
{"train_loss": 0.07518575340509415, "global_step": 183870, "epoch": 2065, "lr": 2.3638333522241312e-05}
{"train_loss": 0.1179492324590683, "global_step": 183871, "epoch": 2065, "lr": 2.3637840909906816e-05}
{"train_loss": 0.03837601840496063, "global_step": 183872, "epoch": 2065, "lr": 2.363734830111637e-05}
{"train_loss": 0.0808537035259638, "global_step": 183873, "epoch": 2065, "lr": 2.363685569587008e-05, "val_loss": 7.477013111114502, "train_action_mse_error": 8.991259574890137}
{"train_loss": 0.050191450864076614, "global_step": 183874, "epoch": 2066, "lr": 2.3636363094167992e-05}
{"train_loss": 0.08840849995613098, "global_step": 183875, "epoch": 2066, "lr": 2.3635870496010154e-05}
{"train_loss": 0.08433561027050018, "global_step": 183876, "epoch": 2066, "lr": 2.363537790139667e-05}
{"train_loss": 0.058375537395477295, "global_step": 183877, "epoch": 2066, "lr": 2.3634885310327575e-05}
{"train_loss": 0.08443118631839752, "global_step": 183878, "epoch": 2066, "lr": 2.363439272280297e-05}
{"train_loss": 0.08509425073862076, "global_step": 183879, "epoch": 2066, "lr": 2.3633900138822867e-05}
{"train_loss": 0.06104563549160957, "global_step": 183880, "epoch": 2066, "lr": 2.3633407558387394e-05}
{"train_loss": 0.07689297199249268, "global_step": 183881, "epoch": 2066, "lr": 2.363291498149657e-05}
{"train_loss": 0.10834113508462906, "global_step": 183882, "epoch": 2066, "lr": 2.363242240815049e-05}
{"train_loss": 0.07617741078138351, "global_step": 183883, "epoch": 2066, "lr": 2.3631929838349192e-05}
{"train_loss": 0.05378793179988861, "global_step": 183884, "epoch": 2066, "lr": 2.3631437272092783e-05}
{"train_loss": 0.03804977983236313, "global_step": 183885, "epoch": 2066, "lr": 2.3630944709381286e-05}
{"train_loss": 0.08284775912761688, "global_step": 183886, "epoch": 2066, "lr": 2.3630452150214804e-05}
{"train_loss": 0.12327143549919128, "global_step": 183887, "epoch": 2066, "lr": 2.362995959459337e-05}
{"train_loss": 0.07872780412435532, "global_step": 183888, "epoch": 2066, "lr": 2.3629467042517084e-05}
{"train_loss": 0.1068856418132782, "global_step": 183889, "epoch": 2066, "lr": 2.362897449398598e-05}
{"train_loss": 0.05875923112034798, "global_step": 183890, "epoch": 2066, "lr": 2.3628481949000153e-05}
{"train_loss": 0.09141180664300919, "global_step": 183891, "epoch": 2066, "lr": 2.3627989407559638e-05}
{"train_loss": 0.03577737882733345, "global_step": 183892, "epoch": 2066, "lr": 2.362749686966452e-05}
{"train_loss": 0.08444662392139435, "global_step": 183893, "epoch": 2066, "lr": 2.3627004335314872e-05}
{"train_loss": 0.09272774308919907, "global_step": 183894, "epoch": 2066, "lr": 2.3626511804510743e-05}
{"train_loss": 0.09394672513008118, "global_step": 183895, "epoch": 2066, "lr": 2.3626019277252226e-05}
{"train_loss": 0.03698064014315605, "global_step": 183896, "epoch": 2066, "lr": 2.3625526753539352e-05}
{"train_loss": 0.05998625233769417, "global_step": 183897, "epoch": 2066, "lr": 2.362503423337222e-05}
{"train_loss": 0.10234063863754272, "global_step": 183898, "epoch": 2066, "lr": 2.362454171675086e-05}
{"train_loss": 0.02439408004283905, "global_step": 183899, "epoch": 2066, "lr": 2.362404920367538e-05}
{"train_loss": 0.059687960892915726, "global_step": 183900, "epoch": 2066, "lr": 2.362355669414581e-05}
{"train_loss": 0.06601833552122116, "global_step": 183901, "epoch": 2066, "lr": 2.3623064188162243e-05}
{"train_loss": 0.04614127427339554, "global_step": 183902, "epoch": 2066, "lr": 2.362257168572472e-05}
{"train_loss": 0.11454381048679352, "global_step": 183903, "epoch": 2066, "lr": 2.362207918683334e-05}
{"train_loss": 0.10934726148843765, "global_step": 183904, "epoch": 2066, "lr": 2.362158669148813e-05}
{"train_loss": 0.15453198552131653, "global_step": 183905, "epoch": 2066, "lr": 2.3621094199689193e-05}
{"train_loss": 0.09884104132652283, "global_step": 183906, "epoch": 2066, "lr": 2.3620601711436557e-05}
{"train_loss": 0.14350678026676178, "global_step": 183907, "epoch": 2066, "lr": 2.3620109226730336e-05}
{"train_loss": 0.15523892641067505, "global_step": 183908, "epoch": 2066, "lr": 2.3619616745570544e-05}
{"train_loss": 0.061151113361120224, "global_step": 183909, "epoch": 2066, "lr": 2.3619124267957292e-05}
{"train_loss": 0.039506711065769196, "global_step": 183910, "epoch": 2066, "lr": 2.3618631793890618e-05}
{"train_loss": 0.021163638681173325, "global_step": 183911, "epoch": 2066, "lr": 2.361813932337061e-05}
{"train_loss": 0.06581966578960419, "global_step": 183912, "epoch": 2066, "lr": 2.3617646856397318e-05}
{"train_loss": 0.07680355757474899, "global_step": 183913, "epoch": 2066, "lr": 2.3617154392970798e-05}
{"train_loss": 0.11020753532648087, "global_step": 183914, "epoch": 2066, "lr": 2.3616661933091143e-05}
{"train_loss": 0.06473436206579208, "global_step": 183915, "epoch": 2066, "lr": 2.3616169476758394e-05}
{"train_loss": 0.062298733741045, "global_step": 183916, "epoch": 2066, "lr": 2.3615677023972644e-05}
{"train_loss": 0.07522661238908768, "global_step": 183917, "epoch": 2066, "lr": 2.361518457473393e-05}
{"train_loss": 0.06977735459804535, "global_step": 183918, "epoch": 2066, "lr": 2.361469212904235e-05}
{"train_loss": 0.08893817663192749, "global_step": 183919, "epoch": 2066, "lr": 2.3614199686897932e-05}
{"train_loss": 0.09416873753070831, "global_step": 183920, "epoch": 2066, "lr": 2.3613707248300783e-05}
{"train_loss": 0.13242506980895996, "global_step": 183921, "epoch": 2066, "lr": 2.3613214813250926e-05}
{"train_loss": 0.2009274661540985, "global_step": 183922, "epoch": 2066, "lr": 2.3612722381748475e-05}
{"train_loss": 0.08592554181814194, "global_step": 183923, "epoch": 2066, "lr": 2.3612229953793458e-05}
{"train_loss": 0.06110439822077751, "global_step": 183924, "epoch": 2066, "lr": 2.3611737529385964e-05}
{"train_loss": 0.03562389686703682, "global_step": 183925, "epoch": 2066, "lr": 2.3611245108526037e-05}
{"train_loss": 0.0717141404747963, "global_step": 183926, "epoch": 2066, "lr": 2.3610752691213772e-05}
{"train_loss": 0.044328901916742325, "global_step": 183927, "epoch": 2066, "lr": 2.3610260277449203e-05}
{"train_loss": 0.1109263151884079, "global_step": 183928, "epoch": 2066, "lr": 2.360976786723241e-05}
{"train_loss": 0.05779081955552101, "global_step": 183929, "epoch": 2066, "lr": 2.3609275460563492e-05}
{"train_loss": 0.051729995757341385, "global_step": 183930, "epoch": 2066, "lr": 2.3608783057442458e-05}
{"train_loss": 0.10083124786615372, "global_step": 183931, "epoch": 2066, "lr": 2.360829065786942e-05}
{"train_loss": 0.06059523671865463, "global_step": 183932, "epoch": 2066, "lr": 2.3607798261844405e-05}
{"train_loss": 0.09364401549100876, "global_step": 183933, "epoch": 2066, "lr": 2.360730586936752e-05}
{"train_loss": 0.08480037748813629, "global_step": 183934, "epoch": 2066, "lr": 2.3606813480438793e-05}
{"train_loss": 0.06493020057678223, "global_step": 183935, "epoch": 2066, "lr": 2.3606321095058325e-05}
{"train_loss": 0.06716213375329971, "global_step": 183936, "epoch": 2066, "lr": 2.3605828713226153e-05}
{"train_loss": 0.13202765583992004, "global_step": 183937, "epoch": 2066, "lr": 2.3605336334942368e-05}
{"train_loss": 0.09362199902534485, "global_step": 183938, "epoch": 2066, "lr": 2.3604843960207007e-05}
{"train_loss": 0.06364574283361435, "global_step": 183939, "epoch": 2066, "lr": 2.3604351589020172e-05}
{"train_loss": 0.12758027017116547, "global_step": 183940, "epoch": 2066, "lr": 2.360385922138189e-05}
{"train_loss": 0.11289884895086288, "global_step": 183941, "epoch": 2066, "lr": 2.360336685729227e-05}
{"train_loss": 0.038787681609392166, "global_step": 183942, "epoch": 2066, "lr": 2.3602874496751338e-05}
{"train_loss": 0.0907212421298027, "global_step": 183943, "epoch": 2066, "lr": 2.360238213975919e-05}
{"train_loss": 0.04386211931705475, "global_step": 183944, "epoch": 2066, "lr": 2.360188978631587e-05}
{"train_loss": 0.09268291294574738, "global_step": 183945, "epoch": 2066, "lr": 2.360139743642144e-05}
{"train_loss": 0.06004951894283295, "global_step": 183946, "epoch": 2066, "lr": 2.3600905090076008e-05}
{"train_loss": 0.08788371831178665, "global_step": 183947, "epoch": 2066, "lr": 2.3600412747279617e-05}
{"train_loss": 0.09897363185882568, "global_step": 183948, "epoch": 2066, "lr": 2.3599920408032307e-05}
{"train_loss": 0.08299924433231354, "global_step": 183949, "epoch": 2066, "lr": 2.3599428072334177e-05}
{"train_loss": 0.12425711750984192, "global_step": 183950, "epoch": 2066, "lr": 2.3598935740185268e-05}
{"train_loss": 0.08505842089653015, "global_step": 183951, "epoch": 2066, "lr": 2.3598443411585684e-05}
{"train_loss": 0.10604478418827057, "global_step": 183952, "epoch": 2066, "lr": 2.3597951086535464e-05}
{"train_loss": 0.084694042801857, "global_step": 183953, "epoch": 2066, "lr": 2.3597458765034656e-05}
{"train_loss": 0.06054207682609558, "global_step": 183954, "epoch": 2066, "lr": 2.3596966447083368e-05}
{"train_loss": 0.03127952665090561, "global_step": 183955, "epoch": 2066, "lr": 2.3596474132681635e-05}
{"train_loss": 0.05896871164441109, "global_step": 183956, "epoch": 2066, "lr": 2.3595981821829543e-05}
{"train_loss": 0.06665383279323578, "global_step": 183957, "epoch": 2066, "lr": 2.3595489514527136e-05}
{"train_loss": 0.10075734555721283, "global_step": 183958, "epoch": 2066, "lr": 2.3594997210774512e-05}
{"train_loss": 0.046761415898799896, "global_step": 183959, "epoch": 2066, "lr": 2.3594504910571698e-05}
{"train_loss": 0.1148606687784195, "global_step": 183960, "epoch": 2066, "lr": 2.35940126139188e-05}
{"train_loss": 0.06685122847557068, "global_step": 183961, "epoch": 2066, "lr": 2.359352032081584e-05}
{"train_loss": 0.08117055386472284, "global_step": 183962, "epoch": 2066, "lr": 2.359302803126294e-05, "val_loss": 7.364935398101807}
{"train_loss": 0.034025389701128006, "global_step": 183963, "epoch": 2067, "lr": 2.3592535745260113e-05}
{"train_loss": 0.14960214495658875, "global_step": 183964, "epoch": 2067, "lr": 2.3592043462807443e-05}
{"train_loss": 0.09249915927648544, "global_step": 183965, "epoch": 2067, "lr": 2.3591551183905024e-05}
{"train_loss": 0.04506317153573036, "global_step": 183966, "epoch": 2067, "lr": 2.3591058908552883e-05}
{"train_loss": 0.07313068211078644, "global_step": 183967, "epoch": 2067, "lr": 2.3590566636751117e-05}
{"train_loss": 0.08978281915187836, "global_step": 183968, "epoch": 2067, "lr": 2.3590074368499754e-05}
{"train_loss": 0.09536111354827881, "global_step": 183969, "epoch": 2067, "lr": 2.3589582103798906e-05}
{"train_loss": 0.12498413771390915, "global_step": 183970, "epoch": 2067, "lr": 2.3589089842648604e-05}
{"train_loss": 0.06940388679504395, "global_step": 183971, "epoch": 2067, "lr": 2.3588597585048938e-05}
{"train_loss": 0.07366471737623215, "global_step": 183972, "epoch": 2067, "lr": 2.358810533099995e-05}
{"train_loss": 0.11077065765857697, "global_step": 183973, "epoch": 2067, "lr": 2.3587613080501737e-05}
{"train_loss": 0.0725015252828598, "global_step": 183974, "epoch": 2067, "lr": 2.3587120833554327e-05}
{"train_loss": 0.06947401911020279, "global_step": 183975, "epoch": 2067, "lr": 2.358662859015782e-05}
{"train_loss": 0.037192124873399734, "global_step": 183976, "epoch": 2067, "lr": 2.358613635031226e-05}
{"train_loss": 0.03987226262688637, "global_step": 183977, "epoch": 2067, "lr": 2.3585644114017737e-05}
{"train_loss": 0.049698151648044586, "global_step": 183978, "epoch": 2067, "lr": 2.358515188127428e-05}
{"train_loss": 0.09425180405378342, "global_step": 183979, "epoch": 2067, "lr": 2.3584659652082e-05}
{"train_loss": 0.09034814685583115, "global_step": 183980, "epoch": 2067, "lr": 2.3584167426440923e-05}
{"train_loss": 0.07005234062671661, "global_step": 183981, "epoch": 2067, "lr": 2.358367520435113e-05}
{"train_loss": 0.051316820085048676, "global_step": 183982, "epoch": 2067, "lr": 2.3583182985812708e-05}
{"train_loss": 0.08436300605535507, "global_step": 183983, "epoch": 2067, "lr": 2.3582690770825687e-05}
{"train_loss": 0.1333359181880951, "global_step": 183984, "epoch": 2067, "lr": 2.3582198559390174e-05}
{"train_loss": 0.07626654952764511, "global_step": 183985, "epoch": 2067, "lr": 2.358170635150621e-05}
{"train_loss": 0.09881848096847534, "global_step": 183986, "epoch": 2067, "lr": 2.358121414717384e-05}
{"train_loss": 0.06594666093587875, "global_step": 183987, "epoch": 2067, "lr": 2.358072194639318e-05}
{"train_loss": 0.09078605473041534, "global_step": 183988, "epoch": 2067, "lr": 2.3580229749164244e-05}
{"train_loss": 0.051433462649583817, "global_step": 183989, "epoch": 2067, "lr": 2.3579737555487146e-05}
{"train_loss": 0.10798977315425873, "global_step": 183990, "epoch": 2067, "lr": 2.3579245365361935e-05}
{"train_loss": 0.08257446438074112, "global_step": 183991, "epoch": 2067, "lr": 2.357875317878865e-05}
{"train_loss": 0.12148873507976532, "global_step": 183992, "epoch": 2067, "lr": 2.3578260995767398e-05}
{"train_loss": 0.10354647040367126, "global_step": 183993, "epoch": 2067, "lr": 2.3577768816298207e-05}
{"train_loss": 0.09200054407119751, "global_step": 183994, "epoch": 2067, "lr": 2.357727664038118e-05}
{"train_loss": 0.07134480029344559, "global_step": 183995, "epoch": 2067, "lr": 2.357678446801635e-05}
{"train_loss": 0.15386798977851868, "global_step": 183996, "epoch": 2067, "lr": 2.3576292299203816e-05}
{"train_loss": 0.07172611355781555, "global_step": 183997, "epoch": 2067, "lr": 2.3575800133943616e-05}
{"train_loss": 0.0428144633769989, "global_step": 183998, "epoch": 2067, "lr": 2.357530797223584e-05}
{"train_loss": 0.1035618931055069, "global_step": 183999, "epoch": 2067, "lr": 2.357481581408052e-05}
{"train_loss": 0.02084978111088276, "global_step": 184000, "epoch": 2067, "lr": 2.3574323659477747e-05}
{"train_loss": 0.06730693578720093, "global_step": 184001, "epoch": 2067, "lr": 2.3573831508427606e-05}
{"train_loss": 0.11733750253915787, "global_step": 184002, "epoch": 2067, "lr": 2.3573339360930117e-05}
{"train_loss": 0.06634862720966339, "global_step": 184003, "epoch": 2067, "lr": 2.3572847216985393e-05}
{"train_loss": 0.025884069502353668, "global_step": 184004, "epoch": 2067, "lr": 2.357235507659346e-05}
{"train_loss": 0.06577015668153763, "global_step": 184005, "epoch": 2067, "lr": 2.3571862939754418e-05}
{"train_loss": 0.04716070368885994, "global_step": 184006, "epoch": 2067, "lr": 2.3571370806468303e-05}
{"train_loss": 0.09613194316625595, "global_step": 184007, "epoch": 2067, "lr": 2.357087867673521e-05}
{"train_loss": 0.06393338739871979, "global_step": 184008, "epoch": 2067, "lr": 2.357038655055517e-05}
{"train_loss": 0.06044243276119232, "global_step": 184009, "epoch": 2067, "lr": 2.3569894427928292e-05}
{"train_loss": 0.06813949346542358, "global_step": 184010, "epoch": 2067, "lr": 2.3569402308854593e-05}
{"train_loss": 0.10634593665599823, "global_step": 184011, "epoch": 2067, "lr": 2.3568910193334194e-05}
{"train_loss": 0.10206398367881775, "global_step": 184012, "epoch": 2067, "lr": 2.3568418081367105e-05}
{"train_loss": 0.04819155111908913, "global_step": 184013, "epoch": 2067, "lr": 2.3567925972953448e-05}
{"train_loss": 0.08998246490955353, "global_step": 184014, "epoch": 2067, "lr": 2.3567433868093237e-05}
{"train_loss": 0.058758147060871124, "global_step": 184015, "epoch": 2067, "lr": 2.3566941766786582e-05}
{"train_loss": 0.11706292629241943, "global_step": 184016, "epoch": 2067, "lr": 2.356644966903351e-05}
{"train_loss": 0.07401133328676224, "global_step": 184017, "epoch": 2067, "lr": 2.3565957574834126e-05}
{"train_loss": 0.0656052827835083, "global_step": 184018, "epoch": 2067, "lr": 2.356546548418846e-05}
{"train_loss": 0.07163362205028534, "global_step": 184019, "epoch": 2067, "lr": 2.356497339709659e-05}
{"train_loss": 0.06209322065114975, "global_step": 184020, "epoch": 2067, "lr": 2.3564481313558613e-05}
{"train_loss": 0.048076946288347244, "global_step": 184021, "epoch": 2067, "lr": 2.3563989233574546e-05}
{"train_loss": 0.07681162655353546, "global_step": 184022, "epoch": 2067, "lr": 2.3563497157144497e-05}
{"train_loss": 0.057608604431152344, "global_step": 184023, "epoch": 2067, "lr": 2.3563005084268514e-05}
{"train_loss": 0.055332355201244354, "global_step": 184024, "epoch": 2067, "lr": 2.3562513014946648e-05}
{"train_loss": 0.022482167929410934, "global_step": 184025, "epoch": 2067, "lr": 2.3562020949178992e-05}
{"train_loss": 0.09664204716682434, "global_step": 184026, "epoch": 2067, "lr": 2.3561528886965585e-05}
{"train_loss": 0.09732691943645477, "global_step": 184027, "epoch": 2067, "lr": 2.3561036828306525e-05}
{"train_loss": 0.11090351641178131, "global_step": 184028, "epoch": 2067, "lr": 2.3560544773201865e-05}
{"train_loss": 0.06175743415951729, "global_step": 184029, "epoch": 2067, "lr": 2.3560052721651644e-05}
{"train_loss": 0.052404239773750305, "global_step": 184030, "epoch": 2067, "lr": 2.3559560673655973e-05}
{"train_loss": 0.1378709226846695, "global_step": 184031, "epoch": 2067, "lr": 2.355906862921488e-05}
{"train_loss": 0.129930779337883, "global_step": 184032, "epoch": 2067, "lr": 2.355857658832846e-05}
{"train_loss": 0.04399139806628227, "global_step": 184033, "epoch": 2067, "lr": 2.3558084550996755e-05}
{"train_loss": 0.05221174284815788, "global_step": 184034, "epoch": 2067, "lr": 2.355759251721986e-05}
{"train_loss": 0.058383915573358536, "global_step": 184035, "epoch": 2067, "lr": 2.3557100486997807e-05}
{"train_loss": 0.08518185466527939, "global_step": 184036, "epoch": 2067, "lr": 2.3556608460330683e-05}
{"train_loss": 0.06067003309726715, "global_step": 184037, "epoch": 2067, "lr": 2.3556116437218563e-05}
{"train_loss": 0.11211193352937698, "global_step": 184038, "epoch": 2067, "lr": 2.355562441766148e-05}
{"train_loss": 0.04104885086417198, "global_step": 184039, "epoch": 2067, "lr": 2.3555132401659546e-05}
{"train_loss": 0.0590771809220314, "global_step": 184040, "epoch": 2067, "lr": 2.355464038921278e-05}
{"train_loss": 0.07030180096626282, "global_step": 184041, "epoch": 2067, "lr": 2.355414838032129e-05}
{"train_loss": 0.09723198413848877, "global_step": 184042, "epoch": 2067, "lr": 2.3553656374985105e-05}
{"train_loss": 0.06918463855981827, "global_step": 184043, "epoch": 2067, "lr": 2.3553164373204324e-05}
{"train_loss": 0.03706217184662819, "global_step": 184044, "epoch": 2067, "lr": 2.3552672374978978e-05}
{"train_loss": 0.06398224085569382, "global_step": 184045, "epoch": 2067, "lr": 2.355218038030918e-05}
{"train_loss": 0.06325022876262665, "global_step": 184046, "epoch": 2067, "lr": 2.3551688389194943e-05}
{"train_loss": 0.06526274979114532, "global_step": 184047, "epoch": 2067, "lr": 2.3551196401636377e-05}
{"train_loss": 0.05556609109044075, "global_step": 184048, "epoch": 2067, "lr": 2.3550704417633517e-05}
{"train_loss": 0.13382484018802643, "global_step": 184049, "epoch": 2067, "lr": 2.3550212437186457e-05}
{"train_loss": 0.11296917498111725, "global_step": 184050, "epoch": 2067, "lr": 2.354972046029523e-05}
{"train_loss": 0.07695864683038063, "global_step": 184051, "epoch": 2067, "lr": 2.354922848695994e-05, "val_loss": 7.586029529571533}
{"train_loss": 0.08265716582536697, "global_step": 184052, "epoch": 2068, "lr": 2.354873651718062e-05}
{"train_loss": 0.05823128670454025, "global_step": 184053, "epoch": 2068, "lr": 2.354824455095736e-05}
{"train_loss": 0.07642759382724762, "global_step": 184054, "epoch": 2068, "lr": 2.3547752588290207e-05}
{"train_loss": 0.07584752142429352, "global_step": 184055, "epoch": 2068, "lr": 2.354726062917923e-05}
{"train_loss": 0.0689806416630745, "global_step": 184056, "epoch": 2068, "lr": 2.354676867362452e-05}
{"train_loss": 0.05812613666057587, "global_step": 184057, "epoch": 2068, "lr": 2.3546276721626104e-05}
{"train_loss": 0.08475197851657867, "global_step": 184058, "epoch": 2068, "lr": 2.3545784773184092e-05}
{"train_loss": 0.042917195707559586, "global_step": 184059, "epoch": 2068, "lr": 2.3545292828298504e-05}
{"train_loss": 0.10431646555662155, "global_step": 184060, "epoch": 2068, "lr": 2.3544800886969455e-05}
{"train_loss": 0.052354052662849426, "global_step": 184061, "epoch": 2068, "lr": 2.3544308949196958e-05}
{"train_loss": 0.05439865589141846, "global_step": 184062, "epoch": 2068, "lr": 2.354381701498113e-05}
{"train_loss": 0.054519809782505035, "global_step": 184063, "epoch": 2068, "lr": 2.354332508432201e-05}
{"train_loss": 0.11253122985363007, "global_step": 184064, "epoch": 2068, "lr": 2.354283315721965e-05}
{"train_loss": 0.05784812569618225, "global_step": 184065, "epoch": 2068, "lr": 2.3542341233674153e-05}
{"train_loss": 0.10949082672595978, "global_step": 184066, "epoch": 2068, "lr": 2.354184931368555e-05}
{"train_loss": 0.06456539034843445, "global_step": 184067, "epoch": 2068, "lr": 2.354135739725394e-05}
{"train_loss": 0.0742785781621933, "global_step": 184068, "epoch": 2068, "lr": 2.3540865484379367e-05}
{"train_loss": 0.08225315809249878, "global_step": 184069, "epoch": 2068, "lr": 2.3540373575061885e-05}
{"train_loss": 0.1146887019276619, "global_step": 184070, "epoch": 2068, "lr": 2.35398816693016e-05}
{"train_loss": 0.0795162171125412, "global_step": 184071, "epoch": 2068, "lr": 2.3539389767098536e-05}
{"train_loss": 0.05265921354293823, "global_step": 184072, "epoch": 2068, "lr": 2.3538897868452775e-05}
{"train_loss": 0.024600984528660774, "global_step": 184073, "epoch": 2068, "lr": 2.3538405973364406e-05}
{"train_loss": 0.06031167879700661, "global_step": 184074, "epoch": 2068, "lr": 2.353791408183346e-05}
{"train_loss": 0.06957808881998062, "global_step": 184075, "epoch": 2068, "lr": 2.3537422193860037e-05}
{"train_loss": 0.07211022824048996, "global_step": 184076, "epoch": 2068, "lr": 2.3536930309444165e-05}
{"train_loss": 0.06931036710739136, "global_step": 184077, "epoch": 2068, "lr": 2.353643842858595e-05}
{"train_loss": 0.1046774610877037, "global_step": 184078, "epoch": 2068, "lr": 2.353594655128542e-05}
{"train_loss": 0.06772029399871826, "global_step": 184079, "epoch": 2068, "lr": 2.3535454677542674e-05}
{"train_loss": 0.0643749088048935, "global_step": 184080, "epoch": 2068, "lr": 2.3534962807357748e-05}
{"train_loss": 0.0837273970246315, "global_step": 184081, "epoch": 2068, "lr": 2.3534470940730747e-05}
{"train_loss": 0.07348806411027908, "global_step": 184082, "epoch": 2068, "lr": 2.3533979077661687e-05}
{"train_loss": 0.08847077935934067, "global_step": 184083, "epoch": 2068, "lr": 2.3533487218150685e-05}
{"train_loss": 0.0453825443983078, "global_step": 184084, "epoch": 2068, "lr": 2.3532995362197762e-05}
{"train_loss": 0.09743107110261917, "global_step": 184085, "epoch": 2068, "lr": 2.3532503509803022e-05}
{"train_loss": 0.08968311548233032, "global_step": 184086, "epoch": 2068, "lr": 2.3532011660966495e-05}
{"train_loss": 0.06902605295181274, "global_step": 184087, "epoch": 2068, "lr": 2.353151981568829e-05}
{"train_loss": 0.1293969601392746, "global_step": 184088, "epoch": 2068, "lr": 2.3531027973968433e-05}
{"train_loss": 0.0827983096241951, "global_step": 184089, "epoch": 2068, "lr": 2.3530536135807018e-05}
{"train_loss": 0.06515292823314667, "global_step": 184090, "epoch": 2068, "lr": 2.3530044301204086e-05}
{"train_loss": 0.07486261427402496, "global_step": 184091, "epoch": 2068, "lr": 2.3529552470159716e-05}
{"train_loss": 0.08749520033597946, "global_step": 184092, "epoch": 2068, "lr": 2.3529060642673996e-05}
{"train_loss": 0.10828623920679092, "global_step": 184093, "epoch": 2068, "lr": 2.3528568818746942e-05}
{"train_loss": 0.07801728695631027, "global_step": 184094, "epoch": 2068, "lr": 2.3528076998378677e-05}
{"train_loss": 0.12242584675550461, "global_step": 184095, "epoch": 2068, "lr": 2.3527585181569224e-05}
{"train_loss": 0.06914404034614563, "global_step": 184096, "epoch": 2068, "lr": 2.3527093368318675e-05}
{"train_loss": 0.07289332896471024, "global_step": 184097, "epoch": 2068, "lr": 2.352660155862707e-05}
{"train_loss": 0.06887909024953842, "global_step": 184098, "epoch": 2068, "lr": 2.3526109752494512e-05}
{"train_loss": 0.03883274644613266, "global_step": 184099, "epoch": 2068, "lr": 2.3525617949921023e-05}
{"train_loss": 0.06339745968580246, "global_step": 184100, "epoch": 2068, "lr": 2.3525126150906713e-05}
{"train_loss": 0.050798140466213226, "global_step": 184101, "epoch": 2068, "lr": 2.352463435545163e-05}
{"train_loss": 0.08465297520160675, "global_step": 184102, "epoch": 2068, "lr": 2.3524142563555813e-05}
{"train_loss": 0.058951299637556076, "global_step": 184103, "epoch": 2068, "lr": 2.3523650775219376e-05}
{"train_loss": 0.04715697094798088, "global_step": 184104, "epoch": 2068, "lr": 2.352315899044234e-05}
{"train_loss": 0.07470850646495819, "global_step": 184105, "epoch": 2068, "lr": 2.352266720922481e-05}
{"train_loss": 0.08585777133703232, "global_step": 184106, "epoch": 2068, "lr": 2.3522175431566836e-05}
{"train_loss": 0.055015914142131805, "global_step": 184107, "epoch": 2068, "lr": 2.3521683657468463e-05}
{"train_loss": 0.08956233412027359, "global_step": 184108, "epoch": 2068, "lr": 2.3521191886929778e-05}
{"train_loss": 0.04645407944917679, "global_step": 184109, "epoch": 2068, "lr": 2.3520700119950867e-05}
{"train_loss": 0.12512460350990295, "global_step": 184110, "epoch": 2068, "lr": 2.3520208356531754e-05}
{"train_loss": 0.06365819275379181, "global_step": 184111, "epoch": 2068, "lr": 2.351971659667254e-05}
{"train_loss": 0.06703665852546692, "global_step": 184112, "epoch": 2068, "lr": 2.3519224840373265e-05}
{"train_loss": 0.05301036685705185, "global_step": 184113, "epoch": 2068, "lr": 2.3518733087634025e-05}
{"train_loss": 0.06661035120487213, "global_step": 184114, "epoch": 2068, "lr": 2.3518241338454854e-05}
{"train_loss": 0.07295837998390198, "global_step": 184115, "epoch": 2068, "lr": 2.3517749592835847e-05}
{"train_loss": 0.07293897122144699, "global_step": 184116, "epoch": 2068, "lr": 2.3517257850777034e-05}
{"train_loss": 0.0968681201338768, "global_step": 184117, "epoch": 2068, "lr": 2.3516766112278525e-05}
{"train_loss": 0.11659395694732666, "global_step": 184118, "epoch": 2068, "lr": 2.3516274377340346e-05}
{"train_loss": 0.05360758677124977, "global_step": 184119, "epoch": 2068, "lr": 2.3515782645962603e-05}
{"train_loss": 0.05553100258111954, "global_step": 184120, "epoch": 2068, "lr": 2.3515290918145315e-05}
{"train_loss": 0.06156076490879059, "global_step": 184121, "epoch": 2068, "lr": 2.3514799193888594e-05}
{"train_loss": 0.1015399694442749, "global_step": 184122, "epoch": 2068, "lr": 2.351430747319247e-05}
{"train_loss": 0.03658926486968994, "global_step": 184123, "epoch": 2068, "lr": 2.351381575605704e-05}
{"train_loss": 0.03962763026356697, "global_step": 184124, "epoch": 2068, "lr": 2.3513324042482337e-05}
{"train_loss": 0.06904112547636032, "global_step": 184125, "epoch": 2068, "lr": 2.351283233246846e-05}
{"train_loss": 0.0660891979932785, "global_step": 184126, "epoch": 2068, "lr": 2.3512340626015446e-05}
{"train_loss": 0.04900510236620903, "global_step": 184127, "epoch": 2068, "lr": 2.3511848923123376e-05}
{"train_loss": 0.07299254089593887, "global_step": 184128, "epoch": 2068, "lr": 2.351135722379233e-05}
{"train_loss": 0.09367787837982178, "global_step": 184129, "epoch": 2068, "lr": 2.351086552802234e-05}
{"train_loss": 0.10443703830242157, "global_step": 184130, "epoch": 2068, "lr": 2.351037383581351e-05}
{"train_loss": 0.057143647223711014, "global_step": 184131, "epoch": 2068, "lr": 2.3509882147165867e-05}
{"train_loss": 0.06906337291002274, "global_step": 184132, "epoch": 2068, "lr": 2.3509390462079524e-05}
{"train_loss": 0.05897356942296028, "global_step": 184133, "epoch": 2068, "lr": 2.3508898780554494e-05}
{"train_loss": 0.0358334556221962, "global_step": 184134, "epoch": 2068, "lr": 2.3508407102590895e-05}
{"train_loss": 0.10846118628978729, "global_step": 184135, "epoch": 2068, "lr": 2.350791542818874e-05}
{"train_loss": 0.07511308789253235, "global_step": 184136, "epoch": 2068, "lr": 2.350742375734815e-05}
{"train_loss": 0.05058012902736664, "global_step": 184137, "epoch": 2068, "lr": 2.3506932090069144e-05}
{"train_loss": 0.10784374922513962, "global_step": 184138, "epoch": 2068, "lr": 2.3506440426351823e-05}
{"train_loss": 0.056795671582221985, "global_step": 184139, "epoch": 2068, "lr": 2.350594876619624e-05}
{"train_loss": 0.0733075165974625, "global_step": 184140, "epoch": 2068, "lr": 2.350545710960244e-05, "val_loss": 7.648174285888672}
{"train_loss": 0.035560980439186096, "global_step": 184141, "epoch": 2069, "lr": 2.3504965456570532e-05}
{"train_loss": 0.10894211381673813, "global_step": 184142, "epoch": 2069, "lr": 2.3504473807100536e-05}
{"train_loss": 0.11260362714529037, "global_step": 184143, "epoch": 2069, "lr": 2.3503982161192557e-05}
{"train_loss": 0.07044333964586258, "global_step": 184144, "epoch": 2069, "lr": 2.350349051884665e-05}
{"train_loss": 0.028521891683340073, "global_step": 184145, "epoch": 2069, "lr": 2.3502998880062848e-05}
{"train_loss": 0.07566378265619278, "global_step": 184146, "epoch": 2069, "lr": 2.3502507244841253e-05}
{"train_loss": 0.07539951056241989, "global_step": 184147, "epoch": 2069, "lr": 2.3502015613181937e-05}
{"train_loss": 0.08108387142419815, "global_step": 184148, "epoch": 2069, "lr": 2.350152398508494e-05}
{"train_loss": 0.08737823367118835, "global_step": 184149, "epoch": 2069, "lr": 2.350103236055035e-05}
{"train_loss": 0.13001219928264618, "global_step": 184150, "epoch": 2069, "lr": 2.3500540739578215e-05}
{"train_loss": 0.06536711007356644, "global_step": 184151, "epoch": 2069, "lr": 2.3500049122168622e-05}
{"train_loss": 0.1022423803806305, "global_step": 184152, "epoch": 2069, "lr": 2.3499557508321602e-05}
{"train_loss": 0.0893549844622612, "global_step": 184153, "epoch": 2069, "lr": 2.3499065898037264e-05}
{"train_loss": 0.053399331867694855, "global_step": 184154, "epoch": 2069, "lr": 2.3498574291315638e-05}
{"train_loss": 0.08448504656553268, "global_step": 184155, "epoch": 2069, "lr": 2.3498082688156826e-05}
{"train_loss": 0.09751415252685547, "global_step": 184156, "epoch": 2069, "lr": 2.349759108856085e-05}
{"train_loss": 0.04979582875967026, "global_step": 184157, "epoch": 2069, "lr": 2.349709949252782e-05}
{"train_loss": 0.08201872557401657, "global_step": 184158, "epoch": 2069, "lr": 2.3496607900057766e-05}
{"train_loss": 0.029577311128377914, "global_step": 184159, "epoch": 2069, "lr": 2.3496116311150784e-05}
{"train_loss": 0.06737636774778366, "global_step": 184160, "epoch": 2069, "lr": 2.3495624725806915e-05}
{"train_loss": 0.07069645076990128, "global_step": 184161, "epoch": 2069, "lr": 2.349513314402625e-05}
{"train_loss": 0.05819173902273178, "global_step": 184162, "epoch": 2069, "lr": 2.3494641565808816e-05}
{"train_loss": 0.15945760905742645, "global_step": 184163, "epoch": 2069, "lr": 2.3494149991154712e-05}
{"train_loss": 0.06646835058927536, "global_step": 184164, "epoch": 2069, "lr": 2.349365842006402e-05}
{"train_loss": 0.08574165403842926, "global_step": 184165, "epoch": 2069, "lr": 2.349316685253675e-05}
{"train_loss": 0.03362049162387848, "global_step": 184166, "epoch": 2069, "lr": 2.3492675288573034e-05}
{"train_loss": 0.07425978779792786, "global_step": 184167, "epoch": 2069, "lr": 2.3492183728172878e-05}
{"train_loss": 0.09686878323554993, "global_step": 184168, "epoch": 2069, "lr": 2.3491692171336394e-05}
{"train_loss": 0.04893555864691734, "global_step": 184169, "epoch": 2069, "lr": 2.3491200618063614e-05}
{"train_loss": 0.09430458396673203, "global_step": 184170, "epoch": 2069, "lr": 2.3490709068354634e-05}
{"train_loss": 0.053563158959150314, "global_step": 184171, "epoch": 2069, "lr": 2.3490217522209484e-05}
{"train_loss": 0.06814086437225342, "global_step": 184172, "epoch": 2069, "lr": 2.3489725979628276e-05}
{"train_loss": 0.04688660427927971, "global_step": 184173, "epoch": 2069, "lr": 2.348923444061103e-05}
{"train_loss": 0.05832095444202423, "global_step": 184174, "epoch": 2069, "lr": 2.348874290515785e-05}
{"train_loss": 0.05216896906495094, "global_step": 184175, "epoch": 2069, "lr": 2.3488251373268768e-05}
{"train_loss": 0.06665593385696411, "global_step": 184176, "epoch": 2069, "lr": 2.3487759844943886e-05}
{"train_loss": 0.08618270605802536, "global_step": 184177, "epoch": 2069, "lr": 2.3487268320183232e-05}
{"train_loss": 0.0796910896897316, "global_step": 184178, "epoch": 2069, "lr": 2.3486776798986915e-05}
{"train_loss": 0.05819731205701828, "global_step": 184179, "epoch": 2069, "lr": 2.3486285281354975e-05}
{"train_loss": 0.08403607457876205, "global_step": 184180, "epoch": 2069, "lr": 2.348579376728746e-05}
{"train_loss": 0.03373267501592636, "global_step": 184181, "epoch": 2069, "lr": 2.3485302256784474e-05}
{"train_loss": 0.04161813482642174, "global_step": 184182, "epoch": 2069, "lr": 2.3484810749846054e-05}
{"train_loss": 0.05069532245397568, "global_step": 184183, "epoch": 2069, "lr": 2.3484319246472293e-05}
{"train_loss": 0.11991468816995621, "global_step": 184184, "epoch": 2069, "lr": 2.3483827746663223e-05}
{"train_loss": 0.1027800589799881, "global_step": 184185, "epoch": 2069, "lr": 2.3483336250418946e-05}
{"train_loss": 0.06416205316781998, "global_step": 184186, "epoch": 2069, "lr": 2.3482844757739498e-05}
{"train_loss": 0.08361345529556274, "global_step": 184187, "epoch": 2069, "lr": 2.3482353268624975e-05}
{"train_loss": 0.06467076390981674, "global_step": 184188, "epoch": 2069, "lr": 2.3481861783075403e-05}
{"train_loss": 0.05730196461081505, "global_step": 184189, "epoch": 2069, "lr": 2.3481370301090895e-05}
{"train_loss": 0.09972373396158218, "global_step": 184190, "epoch": 2069, "lr": 2.348087882267147e-05}
{"train_loss": 0.04742264747619629, "global_step": 184191, "epoch": 2069, "lr": 2.3480387347817238e-05}
{"train_loss": 0.16900353133678436, "global_step": 184192, "epoch": 2069, "lr": 2.347989587652823e-05}
{"train_loss": 0.04040464386343956, "global_step": 184193, "epoch": 2069, "lr": 2.3479404408804547e-05}
{"train_loss": 0.1586059033870697, "global_step": 184194, "epoch": 2069, "lr": 2.347891294464621e-05}
{"train_loss": 0.16984915733337402, "global_step": 184195, "epoch": 2069, "lr": 2.3478421484053332e-05}
{"train_loss": 0.12045304477214813, "global_step": 184196, "epoch": 2069, "lr": 2.3477930027025934e-05}
{"train_loss": 0.05088098719716072, "global_step": 184197, "epoch": 2069, "lr": 2.3477438573564132e-05}
{"train_loss": 0.05307392403483391, "global_step": 184198, "epoch": 2069, "lr": 2.3476947123667943e-05}
{"train_loss": 0.16812093555927277, "global_step": 184199, "epoch": 2069, "lr": 2.347645567733745e-05}
{"train_loss": 0.06811937689781189, "global_step": 184200, "epoch": 2069, "lr": 2.3475964234572746e-05}
{"train_loss": 0.06927285343408585, "global_step": 184201, "epoch": 2069, "lr": 2.347547279537386e-05}
{"train_loss": 0.04204040393233299, "global_step": 184202, "epoch": 2069, "lr": 2.347498135974089e-05}
{"train_loss": 0.07225698977708817, "global_step": 184203, "epoch": 2069, "lr": 2.3474489927673865e-05}
{"train_loss": 0.05778531730175018, "global_step": 184204, "epoch": 2069, "lr": 2.3473998499172888e-05}
{"train_loss": 0.054359398782253265, "global_step": 184205, "epoch": 2069, "lr": 2.3473507074237993e-05}
{"train_loss": 0.05197610706090927, "global_step": 184206, "epoch": 2069, "lr": 2.3473015652869284e-05}
{"train_loss": 0.09169377386569977, "global_step": 184207, "epoch": 2069, "lr": 2.347252423506678e-05}
{"train_loss": 0.10657647997140884, "global_step": 184208, "epoch": 2069, "lr": 2.3472032820830592e-05}
{"train_loss": 0.125584214925766, "global_step": 184209, "epoch": 2069, "lr": 2.3471541410160752e-05}
{"train_loss": 0.07586167752742767, "global_step": 184210, "epoch": 2069, "lr": 2.3471050003057355e-05}
{"train_loss": 0.05902240052819252, "global_step": 184211, "epoch": 2069, "lr": 2.347055859952043e-05}
{"train_loss": 0.07878319919109344, "global_step": 184212, "epoch": 2069, "lr": 2.3470067199550087e-05}
{"train_loss": 0.05662831291556358, "global_step": 184213, "epoch": 2069, "lr": 2.3469575803146356e-05}
{"train_loss": 0.05712977051734924, "global_step": 184214, "epoch": 2069, "lr": 2.3469084410309327e-05}
{"train_loss": 0.08089949935674667, "global_step": 184215, "epoch": 2069, "lr": 2.3468593021039043e-05}
{"train_loss": 0.043722666800022125, "global_step": 184216, "epoch": 2069, "lr": 2.3468101635335598e-05}
{"train_loss": 0.07731281965970993, "global_step": 184217, "epoch": 2069, "lr": 2.3467610253199045e-05}
{"train_loss": 0.0849393904209137, "global_step": 184218, "epoch": 2069, "lr": 2.3467118874629433e-05}
{"train_loss": 0.11043716967105865, "global_step": 184219, "epoch": 2069, "lr": 2.346662749962686e-05}
{"train_loss": 0.09015925228595734, "global_step": 184220, "epoch": 2069, "lr": 2.3466136128191356e-05}
{"train_loss": 0.1348094642162323, "global_step": 184221, "epoch": 2069, "lr": 2.346564476032303e-05}
{"train_loss": 0.07901110500097275, "global_step": 184222, "epoch": 2069, "lr": 2.34651533960219e-05}
{"train_loss": 0.06937967240810394, "global_step": 184223, "epoch": 2069, "lr": 2.3464662035288076e-05}
{"train_loss": 0.0777125135064125, "global_step": 184224, "epoch": 2069, "lr": 2.346417067812159e-05}
{"train_loss": 0.06730727106332779, "global_step": 184225, "epoch": 2069, "lr": 2.3463679324522537e-05}
{"train_loss": 0.05142778530716896, "global_step": 184226, "epoch": 2069, "lr": 2.3463187974490952e-05}
{"train_loss": 0.09711182117462158, "global_step": 184227, "epoch": 2069, "lr": 2.3462696628026938e-05}
{"train_loss": 0.09425609558820724, "global_step": 184228, "epoch": 2069, "lr": 2.3462205285130522e-05}
{"train_loss": 0.07825705136978224, "global_step": 184229, "epoch": 2069, "lr": 2.3461713945801805e-05, "val_loss": 7.522961139678955}
{"train_loss": 0.07368988543748856, "global_step": 184230, "epoch": 2070, "lr": 2.346122261004082e-05}
{"train_loss": 0.04397205263376236, "global_step": 184231, "epoch": 2070, "lr": 2.346073127784767e-05}
{"train_loss": 0.10035894066095352, "global_step": 184232, "epoch": 2070, "lr": 2.346023994922238e-05}
{"train_loss": 0.09719254821538925, "global_step": 184233, "epoch": 2070, "lr": 2.3459748624165063e-05}
{"train_loss": 0.06096632778644562, "global_step": 184234, "epoch": 2070, "lr": 2.345925730267573e-05}
{"train_loss": 0.0995575338602066, "global_step": 184235, "epoch": 2070, "lr": 2.345876598475448e-05}
{"train_loss": 0.07694174349308014, "global_step": 184236, "epoch": 2070, "lr": 2.3458274670401398e-05}
{"train_loss": 0.03188025951385498, "global_step": 184237, "epoch": 2070, "lr": 2.345778335961651e-05}
{"train_loss": 0.07866134494543076, "global_step": 184238, "epoch": 2070, "lr": 2.3457292052399916e-05}
{"train_loss": 0.026823746040463448, "global_step": 184239, "epoch": 2070, "lr": 2.345680074875164e-05}
{"train_loss": 0.05586053431034088, "global_step": 184240, "epoch": 2070, "lr": 2.34563094486718e-05}
{"train_loss": 0.08944366127252579, "global_step": 184241, "epoch": 2070, "lr": 2.3455818152160415e-05}
{"train_loss": 0.09132041782140732, "global_step": 184242, "epoch": 2070, "lr": 2.345532685921759e-05}
{"train_loss": 0.049084752798080444, "global_step": 184243, "epoch": 2070, "lr": 2.3454835569843353e-05}
{"train_loss": 0.06438867747783661, "global_step": 184244, "epoch": 2070, "lr": 2.3454344284037812e-05}
{"train_loss": 0.05954433232545853, "global_step": 184245, "epoch": 2070, "lr": 2.345385300180099e-05}
{"train_loss": 0.0893053188920021, "global_step": 184246, "epoch": 2070, "lr": 2.3453361723132994e-05}
{"train_loss": 0.08403832465410233, "global_step": 184247, "epoch": 2070, "lr": 2.3452870448033847e-05}
{"train_loss": 0.08151514083147049, "global_step": 184248, "epoch": 2070, "lr": 2.3452379176503657e-05}
{"train_loss": 0.04887958616018295, "global_step": 184249, "epoch": 2070, "lr": 2.3451887908542456e-05}
{"train_loss": 0.053014203906059265, "global_step": 184250, "epoch": 2070, "lr": 2.3451396644150348e-05}
{"train_loss": 0.099495530128479, "global_step": 184251, "epoch": 2070, "lr": 2.345090538332735e-05}
{"train_loss": 0.0718170553445816, "global_step": 184252, "epoch": 2070, "lr": 2.3450414126073576e-05}
{"train_loss": 0.08339464664459229, "global_step": 184253, "epoch": 2070, "lr": 2.3449922872389052e-05}
{"train_loss": 0.06897861510515213, "global_step": 184254, "epoch": 2070, "lr": 2.3449431622273877e-05}
{"train_loss": 0.08171454071998596, "global_step": 184255, "epoch": 2070, "lr": 2.3448940375728106e-05}
{"train_loss": 0.06706906855106354, "global_step": 184256, "epoch": 2070, "lr": 2.3448449132751782e-05}
{"train_loss": 0.04798128828406334, "global_step": 184257, "epoch": 2070, "lr": 2.3447957893345008e-05}
{"train_loss": 0.09245073795318604, "global_step": 184258, "epoch": 2070, "lr": 2.3447466657507817e-05}
{"train_loss": 0.05613791197538376, "global_step": 184259, "epoch": 2070, "lr": 2.3446975425240304e-05}
{"train_loss": 0.13022169470787048, "global_step": 184260, "epoch": 2070, "lr": 2.3446484196542506e-05}
{"train_loss": 0.08448240160942078, "global_step": 184261, "epoch": 2070, "lr": 2.344599297141452e-05}
{"train_loss": 0.07347093522548676, "global_step": 184262, "epoch": 2070, "lr": 2.344550174985638e-05}
{"train_loss": 0.08121621608734131, "global_step": 184263, "epoch": 2070, "lr": 2.344501053186819e-05}
{"train_loss": 0.07562632113695145, "global_step": 184264, "epoch": 2070, "lr": 2.3444519317449975e-05}
{"train_loss": 0.06450390815734863, "global_step": 184265, "epoch": 2070, "lr": 2.344402810660184e-05}
{"train_loss": 0.0643174946308136, "global_step": 184266, "epoch": 2070, "lr": 2.344353689932381e-05}
{"train_loss": 0.14863194525241852, "global_step": 184267, "epoch": 2070, "lr": 2.3443045695615988e-05}
{"train_loss": 0.06982686370611191, "global_step": 184268, "epoch": 2070, "lr": 2.3442554495478413e-05}
{"train_loss": 0.07358725368976593, "global_step": 184269, "epoch": 2070, "lr": 2.344206329891118e-05}
{"train_loss": 0.14534428715705872, "global_step": 184270, "epoch": 2070, "lr": 2.3441572105914324e-05}
{"train_loss": 0.0659772977232933, "global_step": 184271, "epoch": 2070, "lr": 2.3441080916487935e-05}
{"train_loss": 0.07137410342693329, "global_step": 184272, "epoch": 2070, "lr": 2.344058973063205e-05}
{"train_loss": 0.15013526380062103, "global_step": 184273, "epoch": 2070, "lr": 2.3440098548346762e-05}
{"train_loss": 0.06186166778206825, "global_step": 184274, "epoch": 2070, "lr": 2.3439607369632145e-05}
{"train_loss": 0.09707746654748917, "global_step": 184275, "epoch": 2070, "lr": 2.3439116194488225e-05}
{"train_loss": 0.0480203852057457, "global_step": 184276, "epoch": 2070, "lr": 2.3438625022915116e-05}
{"train_loss": 0.05400087311863899, "global_step": 184277, "epoch": 2070, "lr": 2.3438133854912843e-05}
{"train_loss": 0.06676655262708664, "global_step": 184278, "epoch": 2070, "lr": 2.3437642690481498e-05}
{"train_loss": 0.061078719794750214, "global_step": 184279, "epoch": 2070, "lr": 2.3437151529621126e-05}
{"train_loss": 0.07790753990411758, "global_step": 184280, "epoch": 2070, "lr": 2.3436660372331827e-05}
{"train_loss": 0.10061782598495483, "global_step": 184281, "epoch": 2070, "lr": 2.3436169218613623e-05}
{"train_loss": 0.07019468396902084, "global_step": 184282, "epoch": 2070, "lr": 2.343567806846662e-05}
{"train_loss": 0.06275952607393265, "global_step": 184283, "epoch": 2070, "lr": 2.343518692189085e-05}
{"train_loss": 0.09624390304088593, "global_step": 184284, "epoch": 2070, "lr": 2.3434695778886417e-05}
{"train_loss": 0.1376505047082901, "global_step": 184285, "epoch": 2070, "lr": 2.343420463945334e-05}
{"train_loss": 0.030648943036794662, "global_step": 184286, "epoch": 2070, "lr": 2.343371350359173e-05}
{"train_loss": 0.07953246682882309, "global_step": 184287, "epoch": 2070, "lr": 2.343322237130162e-05}
{"train_loss": 0.030773259699344635, "global_step": 184288, "epoch": 2070, "lr": 2.3432731242583106e-05}
{"train_loss": 0.07075875252485275, "global_step": 184289, "epoch": 2070, "lr": 2.3432240117436217e-05}
{"train_loss": 0.08784949779510498, "global_step": 184290, "epoch": 2070, "lr": 2.3431748995861057e-05}
{"train_loss": 0.025121144950389862, "global_step": 184291, "epoch": 2070, "lr": 2.3431257877857658e-05}
{"train_loss": 0.08078747242689133, "global_step": 184292, "epoch": 2070, "lr": 2.343076676342612e-05}
{"train_loss": 0.07108515501022339, "global_step": 184293, "epoch": 2070, "lr": 2.3430275652566475e-05}
{"train_loss": 0.07407363504171371, "global_step": 184294, "epoch": 2070, "lr": 2.3429784545278825e-05}
{"train_loss": 0.07810065895318985, "global_step": 184295, "epoch": 2070, "lr": 2.3429293441563216e-05}
{"train_loss": 0.05090204253792763, "global_step": 184296, "epoch": 2070, "lr": 2.3428802341419693e-05}
{"train_loss": 0.07335233688354492, "global_step": 184297, "epoch": 2070, "lr": 2.3428311244848366e-05}
{"train_loss": 0.07976795732975006, "global_step": 184298, "epoch": 2070, "lr": 2.3427820151849254e-05}
{"train_loss": 0.09197662770748138, "global_step": 184299, "epoch": 2070, "lr": 2.3427329062422477e-05}
{"train_loss": 0.11567697674036026, "global_step": 184300, "epoch": 2070, "lr": 2.3426837976568044e-05}
{"train_loss": 0.11667757481336594, "global_step": 184301, "epoch": 2070, "lr": 2.342634689428607e-05}
{"train_loss": 0.08054781705141068, "global_step": 184302, "epoch": 2070, "lr": 2.3425855815576582e-05}
{"train_loss": 0.08336985111236572, "global_step": 184303, "epoch": 2070, "lr": 2.342536474043968e-05}
{"train_loss": 0.08102890849113464, "global_step": 184304, "epoch": 2070, "lr": 2.3424873668875396e-05}
{"train_loss": 0.08368192613124847, "global_step": 184305, "epoch": 2070, "lr": 2.342438260088383e-05}
{"train_loss": 0.07982071489095688, "global_step": 184306, "epoch": 2070, "lr": 2.342389153646502e-05}
{"train_loss": 0.05514504387974739, "global_step": 184307, "epoch": 2070, "lr": 2.3423400475619063e-05}
{"train_loss": 0.13355964422225952, "global_step": 184308, "epoch": 2070, "lr": 2.3422909418345978e-05}
{"train_loss": 0.06626192480325699, "global_step": 184309, "epoch": 2070, "lr": 2.3422418364645865e-05}
{"train_loss": 0.0930594727396965, "global_step": 184310, "epoch": 2070, "lr": 2.3421927314518804e-05}
{"train_loss": 0.12455140054225922, "global_step": 184311, "epoch": 2070, "lr": 2.342143626796482e-05}
{"train_loss": 0.06042724847793579, "global_step": 184312, "epoch": 2070, "lr": 2.3420945224984014e-05}
{"train_loss": 0.0711495652794838, "global_step": 184313, "epoch": 2070, "lr": 2.3420454185576423e-05}
{"train_loss": 0.08629390597343445, "global_step": 184314, "epoch": 2070, "lr": 2.3419963149742152e-05}
{"train_loss": 0.0686577707529068, "global_step": 184315, "epoch": 2070, "lr": 2.341947211748122e-05}
{"train_loss": 0.07060693949460983, "global_step": 184316, "epoch": 2070, "lr": 2.3418981088793734e-05}
{"train_loss": 0.06975274533033371, "global_step": 184317, "epoch": 2070, "lr": 2.3418490063679722e-05}
{"train_loss": 0.07804440047717497, "global_step": 184318, "epoch": 2070, "lr": 2.3417999042139295e-05, "val_loss": 7.731980323791504, "train_action_mse_error": 24.768470764160156}
{"train_loss": 0.12257158011198044, "global_step": 184319, "epoch": 2071, "lr": 2.341750802417247e-05}
{"train_loss": 0.04823542758822441, "global_step": 184320, "epoch": 2071, "lr": 2.341701700977935e-05}
{"train_loss": 0.05412580445408821, "global_step": 184321, "epoch": 2071, "lr": 2.3416525998959982e-05}
{"train_loss": 0.07533129304647446, "global_step": 184322, "epoch": 2071, "lr": 2.341603499171445e-05}
{"train_loss": 0.09816283732652664, "global_step": 184323, "epoch": 2071, "lr": 2.3415543988042788e-05}
{"train_loss": 0.07149223238229752, "global_step": 184324, "epoch": 2071, "lr": 2.3415052987945102e-05}
{"train_loss": 0.051316313445568085, "global_step": 184325, "epoch": 2071, "lr": 2.3414561991421424e-05}
{"train_loss": 0.05596749484539032, "global_step": 184326, "epoch": 2071, "lr": 2.3414070998471828e-05}
{"train_loss": 0.0541747510433197, "global_step": 184327, "epoch": 2071, "lr": 2.341358000909641e-05}
{"train_loss": 0.06939934194087982, "global_step": 184328, "epoch": 2071, "lr": 2.341308902329521e-05}
{"train_loss": 0.08326593041419983, "global_step": 184329, "epoch": 2071, "lr": 2.3412598041068272e-05}
{"train_loss": 0.09869305789470673, "global_step": 184330, "epoch": 2071, "lr": 2.341210706241571e-05}
{"train_loss": 0.07311813533306122, "global_step": 184331, "epoch": 2071, "lr": 2.3411616087337545e-05}
{"train_loss": 0.05644749850034714, "global_step": 184332, "epoch": 2071, "lr": 2.341112511583389e-05}
{"train_loss": 0.10057379305362701, "global_step": 184333, "epoch": 2071, "lr": 2.3410634147904782e-05}
{"train_loss": 0.09435102343559265, "global_step": 184334, "epoch": 2071, "lr": 2.3410143183550263e-05}
{"train_loss": 0.07360755652189255, "global_step": 184335, "epoch": 2071, "lr": 2.3409652222770457e-05}
{"train_loss": 0.09542406350374222, "global_step": 184336, "epoch": 2071, "lr": 2.340916126556537e-05}
{"train_loss": 0.06774444878101349, "global_step": 184337, "epoch": 2071, "lr": 2.3408670311935122e-05}
{"train_loss": 0.04786500334739685, "global_step": 184338, "epoch": 2071, "lr": 2.3408179361879732e-05}
{"train_loss": 0.04672221839427948, "global_step": 184339, "epoch": 2071, "lr": 2.3407688415399316e-05}
{"train_loss": 0.05043952167034149, "global_step": 184340, "epoch": 2071, "lr": 2.340719747249388e-05}
{"train_loss": 0.0913507491350174, "global_step": 184341, "epoch": 2071, "lr": 2.3406706533163554e-05}
{"train_loss": 0.03425056114792824, "global_step": 184342, "epoch": 2071, "lr": 2.3406215597408344e-05}
{"train_loss": 0.06270204484462738, "global_step": 184343, "epoch": 2071, "lr": 2.340572466522836e-05}
{"train_loss": 0.08843013644218445, "global_step": 184344, "epoch": 2071, "lr": 2.3405233736623644e-05}
{"train_loss": 0.09385740011930466, "global_step": 184345, "epoch": 2071, "lr": 2.3404742811594266e-05}
{"train_loss": 0.05165763944387436, "global_step": 184346, "epoch": 2071, "lr": 2.3404251890140316e-05}
{"train_loss": 0.08753609657287598, "global_step": 184347, "epoch": 2071, "lr": 2.3403760972261817e-05}
{"train_loss": 0.05814613029360771, "global_step": 184348, "epoch": 2071, "lr": 2.340327005795888e-05}
{"train_loss": 0.04028443247079849, "global_step": 184349, "epoch": 2071, "lr": 2.340277914723153e-05}
{"train_loss": 0.06709586083889008, "global_step": 184350, "epoch": 2071, "lr": 2.3402288240079873e-05}
{"train_loss": 0.06209017336368561, "global_step": 184351, "epoch": 2071, "lr": 2.3401797336503932e-05}
{"train_loss": 0.07590509951114655, "global_step": 184352, "epoch": 2071, "lr": 2.340130643650382e-05}
{"train_loss": 0.0796249732375145, "global_step": 184353, "epoch": 2071, "lr": 2.3400815540079553e-05}
{"train_loss": 0.05808645486831665, "global_step": 184354, "epoch": 2071, "lr": 2.340032464723124e-05}
{"train_loss": 0.05798336863517761, "global_step": 184355, "epoch": 2071, "lr": 2.339983375795891e-05}
{"train_loss": 0.06653110682964325, "global_step": 184356, "epoch": 2071, "lr": 2.3399342872262674e-05}
{"train_loss": 0.0755540058016777, "global_step": 184357, "epoch": 2071, "lr": 2.339885199014255e-05}
{"train_loss": 0.02522611990571022, "global_step": 184358, "epoch": 2071, "lr": 2.3398361111598643e-05}
{"train_loss": 0.08092445880174637, "global_step": 184359, "epoch": 2071, "lr": 2.3397870236630986e-05}
{"train_loss": 0.05853639915585518, "global_step": 184360, "epoch": 2071, "lr": 2.3397379365239675e-05}
{"train_loss": 0.09175102412700653, "global_step": 184361, "epoch": 2071, "lr": 2.339688849742475e-05}
{"train_loss": 0.07904919236898422, "global_step": 184362, "epoch": 2071, "lr": 2.3396397633186283e-05}
{"train_loss": 0.10486521571874619, "global_step": 184363, "epoch": 2071, "lr": 2.339590677252437e-05}
{"train_loss": 0.08367666602134705, "global_step": 184364, "epoch": 2071, "lr": 2.3395415915439034e-05}
{"train_loss": 0.09714356064796448, "global_step": 184365, "epoch": 2071, "lr": 2.3394925061930374e-05}
{"train_loss": 0.08400480449199677, "global_step": 184366, "epoch": 2071, "lr": 2.3394434211998446e-05}
{"train_loss": 0.07099219411611557, "global_step": 184367, "epoch": 2071, "lr": 2.339394336564329e-05}
{"train_loss": 0.10548342019319534, "global_step": 184368, "epoch": 2071, "lr": 2.3393452522865016e-05}
{"train_loss": 0.02804897539317608, "global_step": 184369, "epoch": 2071, "lr": 2.339296168366365e-05}
{"train_loss": 0.10323972254991531, "global_step": 184370, "epoch": 2071, "lr": 2.3392470848039295e-05}
{"train_loss": 0.077975794672966, "global_step": 184371, "epoch": 2071, "lr": 2.3391980015991993e-05}
{"train_loss": 0.06339240074157715, "global_step": 184372, "epoch": 2071, "lr": 2.3391489187521796e-05}
{"train_loss": 0.08518841117620468, "global_step": 184373, "epoch": 2071, "lr": 2.3390998362628813e-05}
{"train_loss": 0.08243513852357864, "global_step": 184374, "epoch": 2071, "lr": 2.3390507541313066e-05}
{"train_loss": 0.09011723846197128, "global_step": 184375, "epoch": 2071, "lr": 2.3390016723574658e-05}
{"train_loss": 0.09441535174846649, "global_step": 184376, "epoch": 2071, "lr": 2.3389525909413622e-05}
{"train_loss": 0.06916316598653793, "global_step": 184377, "epoch": 2071, "lr": 2.3389035098830058e-05}
{"train_loss": 0.09929560869932175, "global_step": 184378, "epoch": 2071, "lr": 2.3388544291823994e-05}
{"train_loss": 0.06624103337526321, "global_step": 184379, "epoch": 2071, "lr": 2.3388053488395533e-05}
{"train_loss": 0.10959769785404205, "global_step": 184380, "epoch": 2071, "lr": 2.3387562688544706e-05}
{"train_loss": 0.06622051447629929, "global_step": 184381, "epoch": 2071, "lr": 2.33870718922716e-05}
{"train_loss": 0.10205898433923721, "global_step": 184382, "epoch": 2071, "lr": 2.3386581099576294e-05}
{"train_loss": 0.04141910374164581, "global_step": 184383, "epoch": 2071, "lr": 2.3386090310458818e-05}
{"train_loss": 0.05790157988667488, "global_step": 184384, "epoch": 2071, "lr": 2.3385599524919276e-05}
{"train_loss": 0.08663703501224518, "global_step": 184385, "epoch": 2071, "lr": 2.33851087429577e-05}
{"train_loss": 0.0425095371901989, "global_step": 184386, "epoch": 2071, "lr": 2.338461796457419e-05}
{"train_loss": 0.07302603870630264, "global_step": 184387, "epoch": 2071, "lr": 2.3384127189768773e-05}
{"train_loss": 0.0699630081653595, "global_step": 184388, "epoch": 2071, "lr": 2.3383636418541554e-05}
{"train_loss": 0.05891427397727966, "global_step": 184389, "epoch": 2071, "lr": 2.3383145650892564e-05}
{"train_loss": 0.07890494912862778, "global_step": 184390, "epoch": 2071, "lr": 2.3382654886821903e-05}
{"train_loss": 0.09135094285011292, "global_step": 184391, "epoch": 2071, "lr": 2.3382164126329603e-05}
{"train_loss": 0.06388185918331146, "global_step": 184392, "epoch": 2071, "lr": 2.3381673369415764e-05}
{"train_loss": 0.11340203881263733, "global_step": 184393, "epoch": 2071, "lr": 2.3381182616080415e-05}
{"train_loss": 0.0633515864610672, "global_step": 184394, "epoch": 2071, "lr": 2.3380691866323662e-05}
{"train_loss": 0.07812164723873138, "global_step": 184395, "epoch": 2071, "lr": 2.3380201120145528e-05}
{"train_loss": 0.07818502187728882, "global_step": 184396, "epoch": 2071, "lr": 2.3379710377546125e-05}
{"train_loss": 0.06763160228729248, "global_step": 184397, "epoch": 2071, "lr": 2.3379219638525474e-05}
{"train_loss": 0.03439649939537048, "global_step": 184398, "epoch": 2071, "lr": 2.3378728903083686e-05}
{"train_loss": 0.07745391875505447, "global_step": 184399, "epoch": 2071, "lr": 2.3378238171220778e-05}
{"train_loss": 0.09360823035240173, "global_step": 184400, "epoch": 2071, "lr": 2.3377747442936842e-05}
{"train_loss": 0.08383607119321823, "global_step": 184401, "epoch": 2071, "lr": 2.3377256718231966e-05}
{"train_loss": 0.1045919880270958, "global_step": 184402, "epoch": 2071, "lr": 2.3376765997106176e-05}
{"train_loss": 0.06550184637308121, "global_step": 184403, "epoch": 2071, "lr": 2.337627527955957e-05}
{"train_loss": 0.1657082438468933, "global_step": 184404, "epoch": 2071, "lr": 2.337578456559218e-05}
{"train_loss": 0.08905697613954544, "global_step": 184405, "epoch": 2071, "lr": 2.3375293855204112e-05}
{"train_loss": 0.0916176438331604, "global_step": 184406, "epoch": 2071, "lr": 2.3374803148395412e-05}
{"train_loss": 0.07528449475681515, "global_step": 184407, "epoch": 2071, "lr": 2.3374312445166124e-05, "val_loss": 7.7088799476623535}
{"train_loss": 0.04789426550269127, "global_step": 184408, "epoch": 2072, "lr": 2.3373821745516354e-05}
{"train_loss": 0.0811072364449501, "global_step": 184409, "epoch": 2072, "lr": 2.3373331049446135e-05}
{"train_loss": 0.052533891052007675, "global_step": 184410, "epoch": 2072, "lr": 2.337284035695556e-05}
{"train_loss": 0.07935045659542084, "global_step": 184411, "epoch": 2072, "lr": 2.3372349668044684e-05}
{"train_loss": 0.11196894198656082, "global_step": 184412, "epoch": 2072, "lr": 2.337185898271356e-05}
{"train_loss": 0.08676953613758087, "global_step": 184413, "epoch": 2072, "lr": 2.3371368300962272e-05}
{"train_loss": 0.08589120209217072, "global_step": 184414, "epoch": 2072, "lr": 2.3370877622790865e-05}
{"train_loss": 0.046927716583013535, "global_step": 184415, "epoch": 2072, "lr": 2.3370386948199436e-05}
{"train_loss": 0.10684490203857422, "global_step": 184416, "epoch": 2072, "lr": 2.336989627718802e-05}
{"train_loss": 0.04501966014504433, "global_step": 184417, "epoch": 2072, "lr": 2.3369405609756688e-05}
{"train_loss": 0.08717722445726395, "global_step": 184418, "epoch": 2072, "lr": 2.336891494590554e-05}
{"train_loss": 0.07387018948793411, "global_step": 184419, "epoch": 2072, "lr": 2.3368424285634598e-05}
{"train_loss": 0.057799823582172394, "global_step": 184420, "epoch": 2072, "lr": 2.3367933628943966e-05}
{"train_loss": 0.06264912337064743, "global_step": 184421, "epoch": 2072, "lr": 2.336744297583367e-05}
{"train_loss": 0.12336687743663788, "global_step": 184422, "epoch": 2072, "lr": 2.3366952326303814e-05}
{"train_loss": 0.13280853629112244, "global_step": 184423, "epoch": 2072, "lr": 2.3366461680354435e-05}
{"train_loss": 0.03928029164671898, "global_step": 184424, "epoch": 2072, "lr": 2.3365971037985623e-05}
{"train_loss": 0.040330592542886734, "global_step": 184425, "epoch": 2072, "lr": 2.3365480399197416e-05}
{"train_loss": 0.06454141438007355, "global_step": 184426, "epoch": 2072, "lr": 2.336498976398991e-05}
{"train_loss": 0.060930877923965454, "global_step": 184427, "epoch": 2072, "lr": 2.336449913236314e-05}
{"train_loss": 0.11117291450500488, "global_step": 184428, "epoch": 2072, "lr": 2.336400850431721e-05}
{"train_loss": 0.07519698888063431, "global_step": 184429, "epoch": 2072, "lr": 2.3363517879852137e-05}
{"train_loss": 0.05865424498915672, "global_step": 184430, "epoch": 2072, "lr": 2.336302725896804e-05}
{"train_loss": 0.07571025937795639, "global_step": 184431, "epoch": 2072, "lr": 2.336253664166494e-05}
{"train_loss": 0.0685572624206543, "global_step": 184432, "epoch": 2072, "lr": 2.336204602794294e-05}
{"train_loss": 0.06270699203014374, "global_step": 184433, "epoch": 2072, "lr": 2.336155541780207e-05}
{"train_loss": 0.0966687873005867, "global_step": 184434, "epoch": 2072, "lr": 2.3361064811242435e-05}
{"train_loss": 0.06628081947565079, "global_step": 184435, "epoch": 2072, "lr": 2.3360574208264057e-05}
{"train_loss": 0.05735456570982933, "global_step": 184436, "epoch": 2072, "lr": 2.3360083608867032e-05}
{"train_loss": 0.07208360731601715, "global_step": 184437, "epoch": 2072, "lr": 2.3359593013051433e-05}
{"train_loss": 0.058332137763500214, "global_step": 184438, "epoch": 2072, "lr": 2.335910242081729e-05}
{"train_loss": 0.10512588173151016, "global_step": 184439, "epoch": 2072, "lr": 2.3358611832164717e-05}
{"train_loss": 0.08448199927806854, "global_step": 184440, "epoch": 2072, "lr": 2.3358121247093734e-05}
{"train_loss": 0.0748237669467926, "global_step": 184441, "epoch": 2072, "lr": 2.335763066560444e-05}
{"train_loss": 0.041539378464221954, "global_step": 184442, "epoch": 2072, "lr": 2.335714008769687e-05}
{"train_loss": 0.11084851622581482, "global_step": 184443, "epoch": 2072, "lr": 2.335664951337113e-05}
{"train_loss": 0.08592119067907333, "global_step": 184444, "epoch": 2072, "lr": 2.3356158942627266e-05}
{"train_loss": 0.05339802801609039, "global_step": 184445, "epoch": 2072, "lr": 2.3355668375465318e-05}
{"train_loss": 0.09958186745643616, "global_step": 184446, "epoch": 2072, "lr": 2.3355177811885396e-05}
{"train_loss": 0.08929522335529327, "global_step": 184447, "epoch": 2072, "lr": 2.3354687251887523e-05}
{"train_loss": 0.1219571977853775, "global_step": 184448, "epoch": 2072, "lr": 2.3354196695471814e-05}
{"train_loss": 0.052893806248903275, "global_step": 184449, "epoch": 2072, "lr": 2.3353706142638305e-05}
{"train_loss": 0.0830695778131485, "global_step": 184450, "epoch": 2072, "lr": 2.3353215593387047e-05}
{"train_loss": 0.07958649098873138, "global_step": 184451, "epoch": 2072, "lr": 2.3352725047718142e-05}
{"train_loss": 0.027621006593108177, "global_step": 184452, "epoch": 2072, "lr": 2.335223450563162e-05}
{"train_loss": 0.09786663949489594, "global_step": 184453, "epoch": 2072, "lr": 2.3351743967127565e-05}
{"train_loss": 0.07052730768918991, "global_step": 184454, "epoch": 2072, "lr": 2.3351253432206062e-05}
{"train_loss": 0.09760745614767075, "global_step": 184455, "epoch": 2072, "lr": 2.335076290086714e-05}
{"train_loss": 0.09834590554237366, "global_step": 184456, "epoch": 2072, "lr": 2.3350272373110904e-05}
{"train_loss": 0.06301611661911011, "global_step": 184457, "epoch": 2072, "lr": 2.3349781848937375e-05}
{"train_loss": 0.045960161834955215, "global_step": 184458, "epoch": 2072, "lr": 2.3349291328346666e-05}
{"train_loss": 0.058848921209573746, "global_step": 184459, "epoch": 2072, "lr": 2.3348800811338802e-05}
{"train_loss": 0.07365147024393082, "global_step": 184460, "epoch": 2072, "lr": 2.3348310297913877e-05}
{"train_loss": 0.06380610167980194, "global_step": 184461, "epoch": 2072, "lr": 2.3347819788071938e-05}
{"train_loss": 0.11638396233320236, "global_step": 184462, "epoch": 2072, "lr": 2.3347329281813074e-05}
{"train_loss": 0.07236446440219879, "global_step": 184463, "epoch": 2072, "lr": 2.334683877913732e-05}
{"train_loss": 0.06666693091392517, "global_step": 184464, "epoch": 2072, "lr": 2.3346348280044778e-05}
{"train_loss": 0.06587854027748108, "global_step": 184465, "epoch": 2072, "lr": 2.3345857784535474e-05}
{"train_loss": 0.07210395485162735, "global_step": 184466, "epoch": 2072, "lr": 2.3345367292609515e-05}
{"train_loss": 0.0797024518251419, "global_step": 184467, "epoch": 2072, "lr": 2.3344876804266923e-05}
{"train_loss": 0.09987960010766983, "global_step": 184468, "epoch": 2072, "lr": 2.3344386319507812e-05}
{"train_loss": 0.07958895713090897, "global_step": 184469, "epoch": 2072, "lr": 2.3343895838332203e-05}
{"train_loss": 0.07382076978683472, "global_step": 184470, "epoch": 2072, "lr": 2.33434053607402e-05}
{"train_loss": 0.10581430792808533, "global_step": 184471, "epoch": 2072, "lr": 2.334291488673183e-05}
{"train_loss": 0.07731939107179642, "global_step": 184472, "epoch": 2072, "lr": 2.3342424416307184e-05}
{"train_loss": 0.1236974224448204, "global_step": 184473, "epoch": 2072, "lr": 2.3341933949466344e-05}
{"train_loss": 0.06745151430368423, "global_step": 184474, "epoch": 2072, "lr": 2.3341443486209337e-05}
{"train_loss": 0.11280383169651031, "global_step": 184475, "epoch": 2072, "lr": 2.3340953026536262e-05}
{"train_loss": 0.1224600151181221, "global_step": 184476, "epoch": 2072, "lr": 2.3340462570447153e-05}
{"train_loss": 0.04271749407052994, "global_step": 184477, "epoch": 2072, "lr": 2.3339972117942115e-05}
{"train_loss": 0.08558697253465652, "global_step": 184478, "epoch": 2072, "lr": 2.3339481669021174e-05}
{"train_loss": 0.15729066729545593, "global_step": 184479, "epoch": 2072, "lr": 2.3338991223684437e-05}
{"train_loss": 0.05867588147521019, "global_step": 184480, "epoch": 2072, "lr": 2.333850078193192e-05}
{"train_loss": 0.05198400095105171, "global_step": 184481, "epoch": 2072, "lr": 2.3338010343763744e-05}
{"train_loss": 0.10156400501728058, "global_step": 184482, "epoch": 2072, "lr": 2.3337519909179944e-05}
{"train_loss": 0.08963654935359955, "global_step": 184483, "epoch": 2072, "lr": 2.333702947818057e-05}
{"train_loss": 0.06101561337709427, "global_step": 184484, "epoch": 2072, "lr": 2.3336539050765726e-05}
{"train_loss": 0.0823163166642189, "global_step": 184485, "epoch": 2072, "lr": 2.333604862693544e-05}
{"train_loss": 0.043679557740688324, "global_step": 184486, "epoch": 2072, "lr": 2.333555820668982e-05}
{"train_loss": 0.07716315984725952, "global_step": 184487, "epoch": 2072, "lr": 2.333506779002891e-05}
{"train_loss": 0.06206922233104706, "global_step": 184488, "epoch": 2072, "lr": 2.3334577376952747e-05}
{"train_loss": 0.07967865467071533, "global_step": 184489, "epoch": 2072, "lr": 2.333408696746143e-05}
{"train_loss": 0.10307911783456802, "global_step": 184490, "epoch": 2072, "lr": 2.3333596561555043e-05}
{"train_loss": 0.05998053401708603, "global_step": 184491, "epoch": 2072, "lr": 2.333310615923361e-05}
{"train_loss": 0.11330835521221161, "global_step": 184492, "epoch": 2072, "lr": 2.3332615760497227e-05}
{"train_loss": 0.03342954069375992, "global_step": 184493, "epoch": 2072, "lr": 2.3332125365345935e-05}
{"train_loss": 0.08389940112829208, "global_step": 184494, "epoch": 2072, "lr": 2.3331634973779835e-05}
{"train_loss": 0.13736297190189362, "global_step": 184495, "epoch": 2072, "lr": 2.333114458579895e-05}
{"train_loss": 0.0788050265668818, "global_step": 184496, "epoch": 2072, "lr": 2.333065420140339e-05, "val_loss": 7.558578014373779}
{"train_loss": 0.04811340570449829, "global_step": 184497, "epoch": 2073, "lr": 2.3330163820593177e-05}
{"train_loss": 0.07524381577968597, "global_step": 184498, "epoch": 2073, "lr": 2.332967344336841e-05}
{"train_loss": 0.08383296430110931, "global_step": 184499, "epoch": 2073, "lr": 2.332918306972913e-05}
{"train_loss": 0.07034764438867569, "global_step": 184500, "epoch": 2073, "lr": 2.3328692699675443e-05}
{"train_loss": 0.10364782065153122, "global_step": 184501, "epoch": 2073, "lr": 2.332820233320736e-05}
{"train_loss": 0.044501662254333496, "global_step": 184502, "epoch": 2073, "lr": 2.3327711970324993e-05}
{"train_loss": 0.05447295308113098, "global_step": 184503, "epoch": 2073, "lr": 2.3327221611028375e-05}
{"train_loss": 0.05874430388212204, "global_step": 184504, "epoch": 2073, "lr": 2.3326731255317607e-05}
{"train_loss": 0.05017156898975372, "global_step": 184505, "epoch": 2073, "lr": 2.3326240903192715e-05}
{"train_loss": 0.13504010438919067, "global_step": 184506, "epoch": 2073, "lr": 2.3325750554653802e-05}
{"train_loss": 0.08671049028635025, "global_step": 184507, "epoch": 2073, "lr": 2.33252602097009e-05}
{"train_loss": 0.0474909245967865, "global_step": 184508, "epoch": 2073, "lr": 2.332476986833409e-05}
{"train_loss": 0.0745726227760315, "global_step": 184509, "epoch": 2073, "lr": 2.332427953055346e-05}
{"train_loss": 0.046675339341163635, "global_step": 184510, "epoch": 2073, "lr": 2.3323789196359037e-05}
{"train_loss": 0.11519510298967361, "global_step": 184511, "epoch": 2073, "lr": 2.332329886575092e-05}
{"train_loss": 0.062429942190647125, "global_step": 184512, "epoch": 2073, "lr": 2.3322808538729145e-05}
{"train_loss": 0.11701210588216782, "global_step": 184513, "epoch": 2073, "lr": 2.3322318215293808e-05}
{"train_loss": 0.08183897286653519, "global_step": 184514, "epoch": 2073, "lr": 2.3321827895444947e-05}
{"train_loss": 0.12439828366041183, "global_step": 184515, "epoch": 2073, "lr": 2.332133757918266e-05}
{"train_loss": 0.08851165324449539, "global_step": 184516, "epoch": 2073, "lr": 2.3320847266506974e-05}
{"train_loss": 0.03908166661858559, "global_step": 184517, "epoch": 2073, "lr": 2.3320356957417987e-05}
{"train_loss": 0.05604949966073036, "global_step": 184518, "epoch": 2073, "lr": 2.3319866651915744e-05}
{"train_loss": 0.09109386056661606, "global_step": 184519, "epoch": 2073, "lr": 2.3319376350000333e-05}
{"train_loss": 0.07342950254678726, "global_step": 184520, "epoch": 2073, "lr": 2.331888605167179e-05}
{"train_loss": 0.052020274102687836, "global_step": 184521, "epoch": 2073, "lr": 2.3318395756930217e-05}
{"train_loss": 0.10209352523088455, "global_step": 184522, "epoch": 2073, "lr": 2.331790546577566e-05}
{"train_loss": 0.04238158464431763, "global_step": 184523, "epoch": 2073, "lr": 2.3317415178208162e-05}
{"train_loss": 0.08168566226959229, "global_step": 184524, "epoch": 2073, "lr": 2.3316924894227833e-05}
{"train_loss": 0.06462686508893967, "global_step": 184525, "epoch": 2073, "lr": 2.3316434613834703e-05}
{"train_loss": 0.04394451156258583, "global_step": 184526, "epoch": 2073, "lr": 2.3315944337028865e-05}
{"train_loss": 0.12122682482004166, "global_step": 184527, "epoch": 2073, "lr": 2.3315454063810365e-05}
{"train_loss": 0.04804691672325134, "global_step": 184528, "epoch": 2073, "lr": 2.331496379417929e-05}
{"train_loss": 0.09327071905136108, "global_step": 184529, "epoch": 2073, "lr": 2.3314473528135672e-05}
{"train_loss": 0.08398954570293427, "global_step": 184530, "epoch": 2073, "lr": 2.3313983265679617e-05}
{"train_loss": 0.09790553152561188, "global_step": 184531, "epoch": 2073, "lr": 2.3313493006811155e-05}
{"train_loss": 0.09437410533428192, "global_step": 184532, "epoch": 2073, "lr": 2.331300275153039e-05}
{"train_loss": 0.08753027766942978, "global_step": 184533, "epoch": 2073, "lr": 2.331251249983734e-05}
{"train_loss": 0.12685823440551758, "global_step": 184534, "epoch": 2073, "lr": 2.3312022251732124e-05}
{"train_loss": 0.044463105499744415, "global_step": 184535, "epoch": 2073, "lr": 2.331153200721476e-05}
{"train_loss": 0.06889288872480392, "global_step": 184536, "epoch": 2073, "lr": 2.331104176628535e-05}
{"train_loss": 0.04111523553729057, "global_step": 184537, "epoch": 2073, "lr": 2.331055152894393e-05}
{"train_loss": 0.0889548808336258, "global_step": 184538, "epoch": 2073, "lr": 2.3310061295190598e-05}
{"train_loss": 0.07856301218271255, "global_step": 184539, "epoch": 2073, "lr": 2.3309571065025386e-05}
{"train_loss": 0.052462950348854065, "global_step": 184540, "epoch": 2073, "lr": 2.33090808384484e-05}
{"train_loss": 0.07176634669303894, "global_step": 184541, "epoch": 2073, "lr": 2.3308590615459657e-05}
{"train_loss": 0.05443233251571655, "global_step": 184542, "epoch": 2073, "lr": 2.3308100396059275e-05}
{"train_loss": 0.050509922206401825, "global_step": 184543, "epoch": 2073, "lr": 2.3307610180247264e-05}
{"train_loss": 0.05764731764793396, "global_step": 184544, "epoch": 2073, "lr": 2.3307119968023728e-05}
{"train_loss": 0.03901468217372894, "global_step": 184545, "epoch": 2073, "lr": 2.3306629759388738e-05}
{"train_loss": 0.059273216873407364, "global_step": 184546, "epoch": 2073, "lr": 2.3306139554342332e-05}
{"train_loss": 0.05912575125694275, "global_step": 184547, "epoch": 2073, "lr": 2.3305649352884605e-05}
{"train_loss": 0.09731198847293854, "global_step": 184548, "epoch": 2073, "lr": 2.3305159155015594e-05}
{"train_loss": 0.09027788043022156, "global_step": 184549, "epoch": 2073, "lr": 2.3304668960735394e-05}
{"train_loss": 0.05056780204176903, "global_step": 184550, "epoch": 2073, "lr": 2.330417877004404e-05}
{"train_loss": 0.019297152757644653, "global_step": 184551, "epoch": 2073, "lr": 2.3303688582941622e-05}
{"train_loss": 0.05358601734042168, "global_step": 184552, "epoch": 2073, "lr": 2.3303198399428194e-05}
{"train_loss": 0.11855313181877136, "global_step": 184553, "epoch": 2073, "lr": 2.330270821950384e-05}
{"train_loss": 0.06323754787445068, "global_step": 184554, "epoch": 2073, "lr": 2.3302218043168585e-05}
{"train_loss": 0.0765262097120285, "global_step": 184555, "epoch": 2073, "lr": 2.3301727870422547e-05}
{"train_loss": 0.06559882313013077, "global_step": 184556, "epoch": 2073, "lr": 2.3301237701265744e-05}
{"train_loss": 0.07652317732572556, "global_step": 184557, "epoch": 2073, "lr": 2.3300747535698287e-05}
{"train_loss": 0.06374134868383408, "global_step": 184558, "epoch": 2073, "lr": 2.3300257373720196e-05}
{"train_loss": 0.0887540802359581, "global_step": 184559, "epoch": 2073, "lr": 2.329976721533158e-05}
{"train_loss": 0.04488181322813034, "global_step": 184560, "epoch": 2073, "lr": 2.3299277060532487e-05}
{"train_loss": 0.13005000352859497, "global_step": 184561, "epoch": 2073, "lr": 2.329878690932296e-05}
{"train_loss": 0.08654162287712097, "global_step": 184562, "epoch": 2073, "lr": 2.3298296761703108e-05}
{"train_loss": 0.06657280772924423, "global_step": 184563, "epoch": 2073, "lr": 2.329780661767295e-05}
{"train_loss": 0.0712565928697586, "global_step": 184564, "epoch": 2073, "lr": 2.32973164772326e-05}
{"train_loss": 0.06478848308324814, "global_step": 184565, "epoch": 2073, "lr": 2.3296826340382082e-05}
{"train_loss": 0.0393575057387352, "global_step": 184566, "epoch": 2073, "lr": 2.3296336207121495e-05}
{"train_loss": 0.10109108686447144, "global_step": 184567, "epoch": 2073, "lr": 2.3295846077450874e-05}
{"train_loss": 0.08791738748550415, "global_step": 184568, "epoch": 2073, "lr": 2.3295355951370317e-05}
{"train_loss": 0.1263834834098816, "global_step": 184569, "epoch": 2073, "lr": 2.329486582887986e-05}
{"train_loss": 0.041264504194259644, "global_step": 184570, "epoch": 2073, "lr": 2.3294375709979598e-05}
{"train_loss": 0.12545746564865112, "global_step": 184571, "epoch": 2073, "lr": 2.3293885594669567e-05}
{"train_loss": 0.09486157447099686, "global_step": 184572, "epoch": 2073, "lr": 2.3293395482949858e-05}
{"train_loss": 0.06239668279886246, "global_step": 184573, "epoch": 2073, "lr": 2.3292905374820518e-05}
{"train_loss": 0.06203748285770416, "global_step": 184574, "epoch": 2073, "lr": 2.3292415270281632e-05}
{"train_loss": 0.047288812696933746, "global_step": 184575, "epoch": 2073, "lr": 2.329192516933324e-05}
{"train_loss": 0.05785781890153885, "global_step": 184576, "epoch": 2073, "lr": 2.3291435071975443e-05}
{"train_loss": 0.07841850817203522, "global_step": 184577, "epoch": 2073, "lr": 2.3290944978208262e-05}
{"train_loss": 0.09210404753684998, "global_step": 184578, "epoch": 2073, "lr": 2.3290454888031816e-05}
{"train_loss": 0.0690181627869606, "global_step": 184579, "epoch": 2073, "lr": 2.328996480144612e-05}
{"train_loss": 0.09369109570980072, "global_step": 184580, "epoch": 2073, "lr": 2.328947471845126e-05}
{"train_loss": 0.057335276156663895, "global_step": 184581, "epoch": 2073, "lr": 2.3288984639047324e-05}
{"train_loss": 0.06909879297018051, "global_step": 184582, "epoch": 2073, "lr": 2.328849456323434e-05}
{"train_loss": 0.14536252617835999, "global_step": 184583, "epoch": 2073, "lr": 2.3288004491012415e-05}
{"train_loss": 0.09485305100679398, "global_step": 184584, "epoch": 2073, "lr": 2.3287514422381567e-05}
{"train_loss": 0.07509151628513015, "global_step": 184585, "epoch": 2073, "lr": 2.3287024357341914e-05, "val_loss": 7.753893852233887}
{"train_loss": 0.09290748834609985, "global_step": 184586, "epoch": 2074, "lr": 2.3286534295893468e-05}
{"train_loss": 0.05924458056688309, "global_step": 184587, "epoch": 2074, "lr": 2.3286044238036347e-05}
{"train_loss": 0.06477604806423187, "global_step": 184588, "epoch": 2074, "lr": 2.3285554183770568e-05}
{"train_loss": 0.06271608918905258, "global_step": 184589, "epoch": 2074, "lr": 2.3285064133096242e-05}
{"train_loss": 0.06833160668611526, "global_step": 184590, "epoch": 2074, "lr": 2.3284574086013393e-05}
{"train_loss": 0.08011668920516968, "global_step": 184591, "epoch": 2074, "lr": 2.328408404252213e-05}
{"train_loss": 0.07557547837495804, "global_step": 184592, "epoch": 2074, "lr": 2.328359400262247e-05}
{"train_loss": 0.1214243471622467, "global_step": 184593, "epoch": 2074, "lr": 2.3283103966314535e-05}
{"train_loss": 0.049830324947834015, "global_step": 184594, "epoch": 2074, "lr": 2.3282613933598334e-05}
{"train_loss": 0.09041354805231094, "global_step": 184595, "epoch": 2074, "lr": 2.328212390447398e-05}
{"train_loss": 0.08142548054456711, "global_step": 184596, "epoch": 2074, "lr": 2.32816338789415e-05}
{"train_loss": 0.06999175250530243, "global_step": 184597, "epoch": 2074, "lr": 2.3281143857001004e-05}
{"train_loss": 0.027327870950102806, "global_step": 184598, "epoch": 2074, "lr": 2.3280653838652522e-05}
{"train_loss": 0.034543029963970184, "global_step": 184599, "epoch": 2074, "lr": 2.3280163823896117e-05}
{"train_loss": 0.0329294353723526, "global_step": 184600, "epoch": 2074, "lr": 2.327967381273189e-05}
{"train_loss": 0.09008283913135529, "global_step": 184601, "epoch": 2074, "lr": 2.327918380515986e-05}
{"train_loss": 0.04111456871032715, "global_step": 184602, "epoch": 2074, "lr": 2.327869380118014e-05}
{"train_loss": 0.07729345560073853, "global_step": 184603, "epoch": 2074, "lr": 2.327820380079276e-05}
{"train_loss": 0.09216289967298508, "global_step": 184604, "epoch": 2074, "lr": 2.3277713803997807e-05}
{"train_loss": 0.065381720662117, "global_step": 184605, "epoch": 2074, "lr": 2.327722381079533e-05}
{"train_loss": 0.09104686975479126, "global_step": 184606, "epoch": 2074, "lr": 2.327673382118542e-05}
{"train_loss": 0.03586967661976814, "global_step": 184607, "epoch": 2074, "lr": 2.327624383516811e-05}
{"train_loss": 0.04848190024495125, "global_step": 184608, "epoch": 2074, "lr": 2.3275753852743504e-05}
{"train_loss": 0.04730894789099693, "global_step": 184609, "epoch": 2074, "lr": 2.3275263873911623e-05}
{"train_loss": 0.09487475454807281, "global_step": 184610, "epoch": 2074, "lr": 2.327477389867258e-05}
{"train_loss": 0.03771831840276718, "global_step": 184611, "epoch": 2074, "lr": 2.3274283927026395e-05}
{"train_loss": 0.06780939549207687, "global_step": 184612, "epoch": 2074, "lr": 2.3273793958973184e-05}
{"train_loss": 0.1104152649641037, "global_step": 184613, "epoch": 2074, "lr": 2.327330399451296e-05}
{"train_loss": 0.07133583724498749, "global_step": 184614, "epoch": 2074, "lr": 2.3272814033645836e-05}
{"train_loss": 0.11663313955068588, "global_step": 184615, "epoch": 2074, "lr": 2.327232407637183e-05}
{"train_loss": 0.06033923476934433, "global_step": 184616, "epoch": 2074, "lr": 2.3271834122691045e-05}
{"train_loss": 0.06569124013185501, "global_step": 184617, "epoch": 2074, "lr": 2.3271344172603548e-05}
{"train_loss": 0.07692591100931168, "global_step": 184618, "epoch": 2074, "lr": 2.327085422610938e-05}
{"train_loss": 0.10677280277013779, "global_step": 184619, "epoch": 2074, "lr": 2.3270364283208635e-05}
{"train_loss": 0.11346293240785599, "global_step": 184620, "epoch": 2074, "lr": 2.3269874343901344e-05}
{"train_loss": 0.12911193072795868, "global_step": 184621, "epoch": 2074, "lr": 2.326938440818761e-05}
{"train_loss": 0.08767294883728027, "global_step": 184622, "epoch": 2074, "lr": 2.3268894476067472e-05}
{"train_loss": 0.0737251564860344, "global_step": 184623, "epoch": 2074, "lr": 2.3268404547541016e-05}
{"train_loss": 0.10420549660921097, "global_step": 184624, "epoch": 2074, "lr": 2.3267914622608278e-05}
{"train_loss": 0.04871075600385666, "global_step": 184625, "epoch": 2074, "lr": 2.3267424701269364e-05}
{"train_loss": 0.11078694462776184, "global_step": 184626, "epoch": 2074, "lr": 2.3266934783524303e-05}
{"train_loss": 0.07272657006978989, "global_step": 184627, "epoch": 2074, "lr": 2.326644486937319e-05}
{"train_loss": 0.10197804868221283, "global_step": 184628, "epoch": 2074, "lr": 2.326595495881606e-05}
{"train_loss": 0.044269487261772156, "global_step": 184629, "epoch": 2074, "lr": 2.326546505185302e-05}
{"train_loss": 0.07953868806362152, "global_step": 184630, "epoch": 2074, "lr": 2.3264975148484086e-05}
{"train_loss": 0.04031609371304512, "global_step": 184631, "epoch": 2074, "lr": 2.3264485248709378e-05}
{"train_loss": 0.10596923530101776, "global_step": 184632, "epoch": 2074, "lr": 2.3263995352528905e-05}
{"train_loss": 0.1046920120716095, "global_step": 184633, "epoch": 2074, "lr": 2.3263505459942792e-05}
{"train_loss": 0.058123182505369186, "global_step": 184634, "epoch": 2074, "lr": 2.3263015570951047e-05}
{"train_loss": 0.04733150079846382, "global_step": 184635, "epoch": 2074, "lr": 2.3262525685553788e-05}
{"train_loss": 0.06545984745025635, "global_step": 184636, "epoch": 2074, "lr": 2.3262035803751035e-05}
{"train_loss": 0.05043770745396614, "global_step": 184637, "epoch": 2074, "lr": 2.3261545925542894e-05}
{"train_loss": 0.04553057625889778, "global_step": 184638, "epoch": 2074, "lr": 2.3261056050929414e-05}
{"train_loss": 0.0998152494430542, "global_step": 184639, "epoch": 2074, "lr": 2.3260566179910637e-05}
{"train_loss": 0.10523530840873718, "global_step": 184640, "epoch": 2074, "lr": 2.3260076312486672e-05}
{"train_loss": 0.08026258647441864, "global_step": 184641, "epoch": 2074, "lr": 2.3259586448657545e-05}
{"train_loss": 0.103561170399189, "global_step": 184642, "epoch": 2074, "lr": 2.325909658842336e-05}
{"train_loss": 0.08002879470586777, "global_step": 184643, "epoch": 2074, "lr": 2.3258606731784144e-05}
{"train_loss": 0.07524999976158142, "global_step": 184644, "epoch": 2074, "lr": 2.3258116878740007e-05}
{"train_loss": 0.07358863949775696, "global_step": 184645, "epoch": 2074, "lr": 2.325762702929096e-05}
{"train_loss": 0.052095551043748856, "global_step": 184646, "epoch": 2074, "lr": 2.3257137183437127e-05}
{"train_loss": 0.04588539898395538, "global_step": 184647, "epoch": 2074, "lr": 2.3256647341178524e-05}
{"train_loss": 0.019602324813604355, "global_step": 184648, "epoch": 2074, "lr": 2.3256157502515257e-05}
{"train_loss": 0.06444968283176422, "global_step": 184649, "epoch": 2074, "lr": 2.3255667667447358e-05}
{"train_loss": 0.10947288572788239, "global_step": 184650, "epoch": 2074, "lr": 2.3255177835974923e-05}
{"train_loss": 0.07606719434261322, "global_step": 184651, "epoch": 2074, "lr": 2.3254688008097987e-05}
{"train_loss": 0.0496755950152874, "global_step": 184652, "epoch": 2074, "lr": 2.325419818381665e-05}
{"train_loss": 0.047321952879428864, "global_step": 184653, "epoch": 2074, "lr": 2.325370836313094e-05}
{"train_loss": 0.07079684734344482, "global_step": 184654, "epoch": 2074, "lr": 2.3253218546040944e-05}
{"train_loss": 0.06574838608503342, "global_step": 184655, "epoch": 2074, "lr": 2.3252728732546746e-05}
{"train_loss": 0.06466316431760788, "global_step": 184656, "epoch": 2074, "lr": 2.3252238922648373e-05}
{"train_loss": 0.12982787191867828, "global_step": 184657, "epoch": 2074, "lr": 2.3251749116345932e-05}
{"train_loss": 0.160877987742424, "global_step": 184658, "epoch": 2074, "lr": 2.3251259313639444e-05}
{"train_loss": 0.07871228456497192, "global_step": 184659, "epoch": 2074, "lr": 2.3250769514529025e-05}
{"train_loss": 0.06897906959056854, "global_step": 184660, "epoch": 2074, "lr": 2.325027971901469e-05}
{"train_loss": 0.06330907344818115, "global_step": 184661, "epoch": 2074, "lr": 2.3249789927096543e-05}
{"train_loss": 0.09790464490652084, "global_step": 184662, "epoch": 2074, "lr": 2.324930013877462e-05}
{"train_loss": 0.05014361813664436, "global_step": 184663, "epoch": 2074, "lr": 2.324881035404903e-05}
{"train_loss": 0.06365585327148438, "global_step": 184664, "epoch": 2074, "lr": 2.324832057291978e-05}
{"train_loss": 0.06396566331386566, "global_step": 184665, "epoch": 2074, "lr": 2.3247830795386994e-05}
{"train_loss": 0.07572153210639954, "global_step": 184666, "epoch": 2074, "lr": 2.324734102145069e-05}
{"train_loss": 0.09399651736021042, "global_step": 184667, "epoch": 2074, "lr": 2.3246851251110973e-05}
{"train_loss": 0.07760410755872726, "global_step": 184668, "epoch": 2074, "lr": 2.3246361484367874e-05}
{"train_loss": 0.0663272887468338, "global_step": 184669, "epoch": 2074, "lr": 2.324587172122149e-05}
{"train_loss": 0.04165301471948624, "global_step": 184670, "epoch": 2074, "lr": 2.3245381961671858e-05}
{"train_loss": 0.07859284430742264, "global_step": 184671, "epoch": 2074, "lr": 2.3244892205719078e-05}
{"train_loss": 0.16560855507850647, "global_step": 184672, "epoch": 2074, "lr": 2.3244402453363173e-05}
{"train_loss": 0.049647074192762375, "global_step": 184673, "epoch": 2074, "lr": 2.3243912704604255e-05}
{"train_loss": 0.0755365128639374, "global_step": 184674, "epoch": 2074, "lr": 2.324342295944234e-05, "val_loss": 7.404628276824951}
{"train_loss": 0.05871144309639931, "global_step": 184675, "epoch": 2075, "lr": 2.324293321787755e-05}
{"train_loss": 0.11270023882389069, "global_step": 184676, "epoch": 2075, "lr": 2.3242443479909915e-05}
{"train_loss": 0.04862673580646515, "global_step": 184677, "epoch": 2075, "lr": 2.324195374553949e-05}
{"train_loss": 0.12613965570926666, "global_step": 184678, "epoch": 2075, "lr": 2.324146401476638e-05}
{"train_loss": 0.04062033072113991, "global_step": 184679, "epoch": 2075, "lr": 2.3240974287590605e-05}
{"train_loss": 0.13415926694869995, "global_step": 184680, "epoch": 2075, "lr": 2.324048456401228e-05}
{"train_loss": 0.04263509064912796, "global_step": 184681, "epoch": 2075, "lr": 2.323999484403142e-05}
{"train_loss": 0.1293492615222931, "global_step": 184682, "epoch": 2075, "lr": 2.3239505127648137e-05}
{"train_loss": 0.049968261271715164, "global_step": 184683, "epoch": 2075, "lr": 2.323901541486246e-05}
{"train_loss": 0.09971804916858673, "global_step": 184684, "epoch": 2075, "lr": 2.323852570567449e-05}
{"train_loss": 0.06990364193916321, "global_step": 184685, "epoch": 2075, "lr": 2.323803600008425e-05}
{"train_loss": 0.1364806592464447, "global_step": 184686, "epoch": 2075, "lr": 2.3237546298091855e-05}
{"train_loss": 0.11624382436275482, "global_step": 184687, "epoch": 2075, "lr": 2.3237056599697325e-05}
{"train_loss": 0.06914113461971283, "global_step": 184688, "epoch": 2075, "lr": 2.3236566904900765e-05}
{"train_loss": 0.11369454860687256, "global_step": 184689, "epoch": 2075, "lr": 2.3236077213702208e-05}
{"train_loss": 0.04933036491274834, "global_step": 184690, "epoch": 2075, "lr": 2.3235587526101728e-05}
{"train_loss": 0.13776886463165283, "global_step": 184691, "epoch": 2075, "lr": 2.323509784209942e-05}
{"train_loss": 0.07982978224754333, "global_step": 184692, "epoch": 2075, "lr": 2.3234608161695304e-05}
{"train_loss": 0.09992638230323792, "global_step": 184693, "epoch": 2075, "lr": 2.323411848488949e-05}
{"train_loss": 0.07157552242279053, "global_step": 184694, "epoch": 2075, "lr": 2.3233628811682e-05}
{"train_loss": 0.05607233941555023, "global_step": 184695, "epoch": 2075, "lr": 2.3233139142072948e-05}
{"train_loss": 0.07431194186210632, "global_step": 184696, "epoch": 2075, "lr": 2.323264947606235e-05}
{"train_loss": 0.048460472375154495, "global_step": 184697, "epoch": 2075, "lr": 2.3232159813650317e-05}
{"train_loss": 0.0636954978108406, "global_step": 184698, "epoch": 2075, "lr": 2.3231670154836872e-05}
{"train_loss": 0.1043727695941925, "global_step": 184699, "epoch": 2075, "lr": 2.3231180499622124e-05}
{"train_loss": 0.08105407655239105, "global_step": 184700, "epoch": 2075, "lr": 2.3230690848006097e-05}
{"train_loss": 0.06754359602928162, "global_step": 184701, "epoch": 2075, "lr": 2.32302011999889e-05}
{"train_loss": 0.09250274300575256, "global_step": 184702, "epoch": 2075, "lr": 2.3229711555570556e-05}
{"train_loss": 0.05344368517398834, "global_step": 184703, "epoch": 2075, "lr": 2.322922191475117e-05}
{"train_loss": 0.08412881195545197, "global_step": 184704, "epoch": 2075, "lr": 2.322873227753077e-05}
{"train_loss": 0.07651880383491516, "global_step": 184705, "epoch": 2075, "lr": 2.322824264390946e-05}
{"train_loss": 0.09896720945835114, "global_step": 184706, "epoch": 2075, "lr": 2.3227753013887267e-05}
{"train_loss": 0.04164842888712883, "global_step": 184707, "epoch": 2075, "lr": 2.322726338746427e-05}
{"train_loss": 0.063927561044693, "global_step": 184708, "epoch": 2075, "lr": 2.322677376464057e-05}
{"train_loss": 0.07897026836872101, "global_step": 184709, "epoch": 2075, "lr": 2.32262841454162e-05}
{"train_loss": 0.06980486214160919, "global_step": 184710, "epoch": 2075, "lr": 2.3225794529791205e-05}
{"train_loss": 0.05215486139059067, "global_step": 184711, "epoch": 2075, "lr": 2.32253049177657e-05}
{"train_loss": 0.07399925589561462, "global_step": 184712, "epoch": 2075, "lr": 2.3224815309339708e-05}
{"train_loss": 0.07512842118740082, "global_step": 184713, "epoch": 2075, "lr": 2.3224325704513334e-05}
{"train_loss": 0.09424543380737305, "global_step": 184714, "epoch": 2075, "lr": 2.3223836103286623e-05}
{"train_loss": 0.07201575487852097, "global_step": 184715, "epoch": 2075, "lr": 2.3223346505659616e-05}
{"train_loss": 0.048330988734960556, "global_step": 184716, "epoch": 2075, "lr": 2.322285691163243e-05}
{"train_loss": 0.10019368678331375, "global_step": 184717, "epoch": 2075, "lr": 2.322236732120508e-05}
{"train_loss": 0.09713634848594666, "global_step": 184718, "epoch": 2075, "lr": 2.3221877734377673e-05}
{"train_loss": 0.1120363175868988, "global_step": 184719, "epoch": 2075, "lr": 2.3221388151150246e-05}
{"train_loss": 0.05321840941905975, "global_step": 184720, "epoch": 2075, "lr": 2.3220898571522892e-05}
{"train_loss": 0.06967853754758835, "global_step": 184721, "epoch": 2075, "lr": 2.3220408995495645e-05}
{"train_loss": 0.05972665920853615, "global_step": 184722, "epoch": 2075, "lr": 2.3219919423068603e-05}
{"train_loss": 0.05894805118441582, "global_step": 184723, "epoch": 2075, "lr": 2.3219429854241803e-05}
{"train_loss": 0.05711803212761879, "global_step": 184724, "epoch": 2075, "lr": 2.321894028901534e-05}
{"train_loss": 0.06826715171337128, "global_step": 184725, "epoch": 2075, "lr": 2.3218450727389247e-05}
{"train_loss": 0.0976439118385315, "global_step": 184726, "epoch": 2075, "lr": 2.3217961169363605e-05}
{"train_loss": 0.09518767893314362, "global_step": 184727, "epoch": 2075, "lr": 2.3217471614938503e-05}
{"train_loss": 0.0585579015314579, "global_step": 184728, "epoch": 2075, "lr": 2.3216982064113962e-05}
{"train_loss": 0.08004644513130188, "global_step": 184729, "epoch": 2075, "lr": 2.321649251689009e-05}
{"train_loss": 0.05603226274251938, "global_step": 184730, "epoch": 2075, "lr": 2.321600297326692e-05}
{"train_loss": 0.08710178732872009, "global_step": 184731, "epoch": 2075, "lr": 2.3215513433244546e-05}
{"train_loss": 0.12093677371740341, "global_step": 184732, "epoch": 2075, "lr": 2.3215023896823008e-05}
{"train_loss": 0.05365852639079094, "global_step": 184733, "epoch": 2075, "lr": 2.3214534364002393e-05}
{"train_loss": 0.06140916794538498, "global_step": 184734, "epoch": 2075, "lr": 2.3214044834782745e-05}
{"train_loss": 0.0507889948785305, "global_step": 184735, "epoch": 2075, "lr": 2.321355530916416e-05}
{"train_loss": 0.0510125569999218, "global_step": 184736, "epoch": 2075, "lr": 2.321306578714667e-05}
{"train_loss": 0.08879086375236511, "global_step": 184737, "epoch": 2075, "lr": 2.3212576268730367e-05}
{"train_loss": 0.0894228145480156, "global_step": 184738, "epoch": 2075, "lr": 2.3212086753915298e-05}
{"train_loss": 0.07843034714460373, "global_step": 184739, "epoch": 2075, "lr": 2.3211597242701548e-05}
{"train_loss": 0.06714743375778198, "global_step": 184740, "epoch": 2075, "lr": 2.321110773508916e-05}
{"train_loss": 0.07067764550447464, "global_step": 184741, "epoch": 2075, "lr": 2.321061823107823e-05}
{"train_loss": 0.08437863737344742, "global_step": 184742, "epoch": 2075, "lr": 2.3210128730668784e-05}
{"train_loss": 0.10745054483413696, "global_step": 184743, "epoch": 2075, "lr": 2.3209639233860907e-05}
{"train_loss": 0.09271321445703506, "global_step": 184744, "epoch": 2075, "lr": 2.320914974065469e-05}
{"train_loss": 0.07474575191736221, "global_step": 184745, "epoch": 2075, "lr": 2.3208660251050158e-05}
{"train_loss": 0.05571078136563301, "global_step": 184746, "epoch": 2075, "lr": 2.3208170765047416e-05}
{"train_loss": 0.059471067041158676, "global_step": 184747, "epoch": 2075, "lr": 2.3207681282646488e-05}
{"train_loss": 0.04243699461221695, "global_step": 184748, "epoch": 2075, "lr": 2.3207191803847476e-05}
{"train_loss": 0.06538670510053635, "global_step": 184749, "epoch": 2075, "lr": 2.320670232865043e-05}
{"train_loss": 0.0770585909485817, "global_step": 184750, "epoch": 2075, "lr": 2.3206212857055405e-05}
{"train_loss": 0.09906110912561417, "global_step": 184751, "epoch": 2075, "lr": 2.320572338906249e-05}
{"train_loss": 0.06993546336889267, "global_step": 184752, "epoch": 2075, "lr": 2.3205233924671726e-05}
{"train_loss": 0.024126382544636726, "global_step": 184753, "epoch": 2075, "lr": 2.3204744463883203e-05}
{"train_loss": 0.056663818657398224, "global_step": 184754, "epoch": 2075, "lr": 2.3204255006696986e-05}
{"train_loss": 0.13265542685985565, "global_step": 184755, "epoch": 2075, "lr": 2.3203765553113105e-05}
{"train_loss": 0.07975401729345322, "global_step": 184756, "epoch": 2075, "lr": 2.3203276103131665e-05}
{"train_loss": 0.08355803787708282, "global_step": 184757, "epoch": 2075, "lr": 2.3202786656752707e-05}
{"train_loss": 0.028020357713103294, "global_step": 184758, "epoch": 2075, "lr": 2.3202297213976326e-05}
{"train_loss": 0.07405737042427063, "global_step": 184759, "epoch": 2075, "lr": 2.3201807774802546e-05}
{"train_loss": 0.07757847011089325, "global_step": 184760, "epoch": 2075, "lr": 2.3201318339231474e-05}
{"train_loss": 0.05533425509929657, "global_step": 184761, "epoch": 2075, "lr": 2.320082890726315e-05}
{"train_loss": 0.06866704672574997, "global_step": 184762, "epoch": 2075, "lr": 2.3200339478897638e-05}
{"train_loss": 0.07627537021978517, "global_step": 184763, "epoch": 2075, "lr": 2.3199850054135036e-05, "val_loss": 7.463967323303223, "train_action_mse_error": 17.24740982055664}
{"train_loss": 0.09226004779338837, "global_step": 184764, "epoch": 2076, "lr": 2.319936063297537e-05}
{"train_loss": 0.05902569741010666, "global_step": 184765, "epoch": 2076, "lr": 2.3198871215418737e-05}
{"train_loss": 0.07942413538694382, "global_step": 184766, "epoch": 2076, "lr": 2.3198381801465173e-05}
{"train_loss": 0.07997975498437881, "global_step": 184767, "epoch": 2076, "lr": 2.3197892391114785e-05}
{"train_loss": 0.03781330585479736, "global_step": 184768, "epoch": 2076, "lr": 2.3197402984367588e-05}
{"train_loss": 0.05482718348503113, "global_step": 184769, "epoch": 2076, "lr": 2.319691358122369e-05}
{"train_loss": 0.07036038488149643, "global_step": 184770, "epoch": 2076, "lr": 2.319642418168313e-05}
{"train_loss": 0.08411248028278351, "global_step": 184771, "epoch": 2076, "lr": 2.3195934785745992e-05}
{"train_loss": 0.04255284368991852, "global_step": 184772, "epoch": 2076, "lr": 2.3195445393412324e-05}
{"train_loss": 0.04520965740084648, "global_step": 184773, "epoch": 2076, "lr": 2.3194956004682215e-05}
{"train_loss": 0.03056352585554123, "global_step": 184774, "epoch": 2076, "lr": 2.3194466619555703e-05}
{"train_loss": 0.06659380346536636, "global_step": 184775, "epoch": 2076, "lr": 2.319397723803289e-05}
{"train_loss": 0.06456177681684494, "global_step": 184776, "epoch": 2076, "lr": 2.3193487860113794e-05}
{"train_loss": 0.11919249594211578, "global_step": 184777, "epoch": 2076, "lr": 2.319299848579853e-05}
{"train_loss": 0.0494062677025795, "global_step": 184778, "epoch": 2076, "lr": 2.3192509115087112e-05}
{"train_loss": 0.04216018691658974, "global_step": 184779, "epoch": 2076, "lr": 2.3192019747979664e-05}
{"train_loss": 0.07814151793718338, "global_step": 184780, "epoch": 2076, "lr": 2.3191530384476202e-05}
{"train_loss": 0.08900357037782669, "global_step": 184781, "epoch": 2076, "lr": 2.319104102457681e-05}
{"train_loss": 0.05742615833878517, "global_step": 184782, "epoch": 2076, "lr": 2.3190551668281572e-05}
{"train_loss": 0.029002700001001358, "global_step": 184783, "epoch": 2076, "lr": 2.3190062315590527e-05}
{"train_loss": 0.07337402552366257, "global_step": 184784, "epoch": 2076, "lr": 2.3189572966503765e-05}
{"train_loss": 0.07672993093729019, "global_step": 184785, "epoch": 2076, "lr": 2.318908362102132e-05}
{"train_loss": 0.08353013545274734, "global_step": 184786, "epoch": 2076, "lr": 2.318859427914329e-05}
{"train_loss": 0.07753060013055801, "global_step": 184787, "epoch": 2076, "lr": 2.318810494086973e-05}
{"train_loss": 0.061025168746709824, "global_step": 184788, "epoch": 2076, "lr": 2.3187615606200686e-05}
{"train_loss": 0.06035545840859413, "global_step": 184789, "epoch": 2076, "lr": 2.3187126275136255e-05}
{"train_loss": 0.05642906576395035, "global_step": 184790, "epoch": 2076, "lr": 2.3186636947676478e-05}
{"train_loss": 0.04923887923359871, "global_step": 184791, "epoch": 2076, "lr": 2.3186147623821443e-05}
{"train_loss": 0.11015615612268448, "global_step": 184792, "epoch": 2076, "lr": 2.31856583035712e-05}
{"train_loss": 0.04344945028424263, "global_step": 184793, "epoch": 2076, "lr": 2.3185168986925804e-05}
{"train_loss": 0.06397360563278198, "global_step": 184794, "epoch": 2076, "lr": 2.3184679673885357e-05}
{"train_loss": 0.09960964322090149, "global_step": 184795, "epoch": 2076, "lr": 2.318419036444988e-05}
{"train_loss": 0.04724012687802315, "global_step": 184796, "epoch": 2076, "lr": 2.3183701058619477e-05}
{"train_loss": 0.05013277381658554, "global_step": 184797, "epoch": 2076, "lr": 2.3183211756394186e-05}
{"train_loss": 0.057003315538167953, "global_step": 184798, "epoch": 2076, "lr": 2.3182722457774085e-05}
{"train_loss": 0.10110475867986679, "global_step": 184799, "epoch": 2076, "lr": 2.3182233162759258e-05}
{"train_loss": 0.06224405765533447, "global_step": 184800, "epoch": 2076, "lr": 2.318174387134973e-05}
{"train_loss": 0.09817473590373993, "global_step": 184801, "epoch": 2076, "lr": 2.3181254583545614e-05}
{"train_loss": 0.05824314057826996, "global_step": 184802, "epoch": 2076, "lr": 2.318076529934693e-05}
{"train_loss": 0.09269227832555771, "global_step": 184803, "epoch": 2076, "lr": 2.3180276018753787e-05}
{"train_loss": 0.01555620040744543, "global_step": 184804, "epoch": 2076, "lr": 2.3179786741766206e-05}
{"train_loss": 0.060044124722480774, "global_step": 184805, "epoch": 2076, "lr": 2.31792974683843e-05}
{"train_loss": 0.08379223942756653, "global_step": 184806, "epoch": 2076, "lr": 2.317880819860809e-05}
{"train_loss": 0.06508629769086838, "global_step": 184807, "epoch": 2076, "lr": 2.3178318932437677e-05}
{"train_loss": 0.08353162556886673, "global_step": 184808, "epoch": 2076, "lr": 2.3177829669873092e-05}
{"train_loss": 0.045766059309244156, "global_step": 184809, "epoch": 2076, "lr": 2.3177340410914445e-05}
{"train_loss": 0.090754434466362, "global_step": 184810, "epoch": 2076, "lr": 2.317685115556176e-05}
{"train_loss": 0.07106398046016693, "global_step": 184811, "epoch": 2076, "lr": 2.3176361903815136e-05}
{"train_loss": 0.0442122146487236, "global_step": 184812, "epoch": 2076, "lr": 2.31758726556746e-05}
{"train_loss": 0.11312741786241531, "global_step": 184813, "epoch": 2076, "lr": 2.3175383411140267e-05}
{"train_loss": 0.12778513133525848, "global_step": 184814, "epoch": 2076, "lr": 2.317489417021216e-05}
{"train_loss": 0.07808626443147659, "global_step": 184815, "epoch": 2076, "lr": 2.317440493289037e-05}
{"train_loss": 0.07124998420476913, "global_step": 184816, "epoch": 2076, "lr": 2.3173915699174943e-05}
{"train_loss": 0.06828497350215912, "global_step": 184817, "epoch": 2076, "lr": 2.3173426469065957e-05}
{"train_loss": 0.05442262068390846, "global_step": 184818, "epoch": 2076, "lr": 2.317293724256349e-05}
{"train_loss": 0.061059579253196716, "global_step": 184819, "epoch": 2076, "lr": 2.317244801966758e-05}
{"train_loss": 0.06882211565971375, "global_step": 184820, "epoch": 2076, "lr": 2.3171958800378325e-05}
{"train_loss": 0.0854148268699646, "global_step": 184821, "epoch": 2076, "lr": 2.3171469584695753e-05}
{"train_loss": 0.034397345036268234, "global_step": 184822, "epoch": 2076, "lr": 2.3170980372619972e-05}
{"train_loss": 0.06934823095798492, "global_step": 184823, "epoch": 2076, "lr": 2.3170491164151004e-05}
{"train_loss": 0.046692319214344025, "global_step": 184824, "epoch": 2076, "lr": 2.3170001959288956e-05}
{"train_loss": 0.034098174422979355, "global_step": 184825, "epoch": 2076, "lr": 2.3169512758033873e-05}
{"train_loss": 0.08226045221090317, "global_step": 184826, "epoch": 2076, "lr": 2.3169023560385804e-05}
{"train_loss": 0.06562241911888123, "global_step": 184827, "epoch": 2076, "lr": 2.3168534366344856e-05}
{"train_loss": 0.16187360882759094, "global_step": 184828, "epoch": 2076, "lr": 2.3168045175911046e-05}
{"train_loss": 0.08536534756422043, "global_step": 184829, "epoch": 2076, "lr": 2.3167555989084488e-05}
{"train_loss": 0.0859832689166069, "global_step": 184830, "epoch": 2076, "lr": 2.316706680586523e-05}
{"train_loss": 0.0750628113746643, "global_step": 184831, "epoch": 2076, "lr": 2.3166577626253305e-05}
{"train_loss": 0.05883908271789551, "global_step": 184832, "epoch": 2076, "lr": 2.316608845024883e-05}
{"train_loss": 0.08122118562459946, "global_step": 184833, "epoch": 2076, "lr": 2.3165599277851822e-05}
{"train_loss": 0.10000793635845184, "global_step": 184834, "epoch": 2076, "lr": 2.3165110109062384e-05}
{"train_loss": 0.08590725809335709, "global_step": 184835, "epoch": 2076, "lr": 2.3164620943880577e-05}
{"train_loss": 0.032988131046295166, "global_step": 184836, "epoch": 2076, "lr": 2.316413178230645e-05}
{"train_loss": 0.1800786554813385, "global_step": 184837, "epoch": 2076, "lr": 2.316364262434009e-05}
{"train_loss": 0.045967474579811096, "global_step": 184838, "epoch": 2076, "lr": 2.3163153469981537e-05}
{"train_loss": 0.04835304245352745, "global_step": 184839, "epoch": 2076, "lr": 2.3162664319230888e-05}
{"train_loss": 0.060831036418676376, "global_step": 184840, "epoch": 2076, "lr": 2.316217517208817e-05}
{"train_loss": 0.056044913828372955, "global_step": 184841, "epoch": 2076, "lr": 2.316168602855349e-05}
{"train_loss": 0.11007428169250488, "global_step": 184842, "epoch": 2076, "lr": 2.316119688862688e-05}
{"train_loss": 0.0690576359629631, "global_step": 184843, "epoch": 2076, "lr": 2.3160707752308435e-05}
{"train_loss": 0.09021959453821182, "global_step": 184844, "epoch": 2076, "lr": 2.316021861959819e-05}
{"train_loss": 0.061857957392930984, "global_step": 184845, "epoch": 2076, "lr": 2.3159729490496245e-05}
{"train_loss": 0.08122635632753372, "global_step": 184846, "epoch": 2076, "lr": 2.3159240365002627e-05}
{"train_loss": 0.05673186108469963, "global_step": 184847, "epoch": 2076, "lr": 2.3158751243117438e-05}
{"train_loss": 0.07724868506193161, "global_step": 184848, "epoch": 2076, "lr": 2.3158262124840714e-05}
{"train_loss": 0.07995247840881348, "global_step": 184849, "epoch": 2076, "lr": 2.315777301017255e-05}
{"train_loss": 0.08173881471157074, "global_step": 184850, "epoch": 2076, "lr": 2.315728389911298e-05}
{"train_loss": 0.17417392134666443, "global_step": 184851, "epoch": 2076, "lr": 2.31567947916621e-05}
{"train_loss": 0.07199608767928367, "global_step": 184852, "epoch": 2076, "lr": 2.3156305687819946e-05, "val_loss": 7.572845458984375}
{"train_loss": 0.050789155066013336, "global_step": 184853, "epoch": 2077, "lr": 2.3155816587586603e-05}
{"train_loss": 0.06880505383014679, "global_step": 184854, "epoch": 2077, "lr": 2.315532749096215e-05}
{"train_loss": 0.1110420748591423, "global_step": 184855, "epoch": 2077, "lr": 2.3154838397946617e-05}
{"train_loss": 0.07158070802688599, "global_step": 184856, "epoch": 2077, "lr": 2.3154349308540103e-05}
{"train_loss": 0.0634637400507927, "global_step": 184857, "epoch": 2077, "lr": 2.3153860222742647e-05}
{"train_loss": 0.053466372191905975, "global_step": 184858, "epoch": 2077, "lr": 2.315337114055434e-05}
{"train_loss": 0.07219532132148743, "global_step": 184859, "epoch": 2077, "lr": 2.3152882061975223e-05}
{"train_loss": 0.06392088532447815, "global_step": 184860, "epoch": 2077, "lr": 2.315239298700539e-05}
{"train_loss": 0.055209655314683914, "global_step": 184861, "epoch": 2077, "lr": 2.315190391564487e-05}
{"train_loss": 0.05736381188035011, "global_step": 184862, "epoch": 2077, "lr": 2.3151414847893765e-05}
{"train_loss": 0.07811745256185532, "global_step": 184863, "epoch": 2077, "lr": 2.3150925783752116e-05}
{"train_loss": 0.04097996652126312, "global_step": 184864, "epoch": 2077, "lr": 2.315043672322001e-05}
{"train_loss": 0.07872316241264343, "global_step": 184865, "epoch": 2077, "lr": 2.31499476662975e-05}
{"train_loss": 0.0722515732049942, "global_step": 184866, "epoch": 2077, "lr": 2.3149458612984633e-05}
{"train_loss": 0.12602844834327698, "global_step": 184867, "epoch": 2077, "lr": 2.3148969563281514e-05}
{"train_loss": 0.09568306058645248, "global_step": 184868, "epoch": 2077, "lr": 2.314848051718817e-05}
{"train_loss": 0.10512596368789673, "global_step": 184869, "epoch": 2077, "lr": 2.3147991474704705e-05}
{"train_loss": 0.06879576295614243, "global_step": 184870, "epoch": 2077, "lr": 2.3147502435831147e-05}
{"train_loss": 0.04093905910849571, "global_step": 184871, "epoch": 2077, "lr": 2.3147013400567603e-05}
{"train_loss": 0.042661990970373154, "global_step": 184872, "epoch": 2077, "lr": 2.3146524368914092e-05}
{"train_loss": 0.09172376990318298, "global_step": 184873, "epoch": 2077, "lr": 2.314603534087072e-05}
{"train_loss": 0.09462903439998627, "global_step": 184874, "epoch": 2077, "lr": 2.3145546316437523e-05}
{"train_loss": 0.08289685845375061, "global_step": 184875, "epoch": 2077, "lr": 2.3145057295614592e-05}
{"train_loss": 0.08708291500806808, "global_step": 184876, "epoch": 2077, "lr": 2.3144568278401967e-05}
{"train_loss": 0.08451802283525467, "global_step": 184877, "epoch": 2077, "lr": 2.3144079264799745e-05}
{"train_loss": 0.08565996587276459, "global_step": 184878, "epoch": 2077, "lr": 2.3143590254807957e-05}
{"train_loss": 0.05752451717853546, "global_step": 184879, "epoch": 2077, "lr": 2.31431012484267e-05}
{"train_loss": 0.08380333334207535, "global_step": 184880, "epoch": 2077, "lr": 2.3142612245656008e-05}
{"train_loss": 0.09151652455329895, "global_step": 184881, "epoch": 2077, "lr": 2.314212324649599e-05}
{"train_loss": 0.05765673518180847, "global_step": 184882, "epoch": 2077, "lr": 2.314163425094666e-05}
{"train_loss": 0.03889011964201927, "global_step": 184883, "epoch": 2077, "lr": 2.3141145259008134e-05}
{"train_loss": 0.03498386591672897, "global_step": 184884, "epoch": 2077, "lr": 2.3140656270680434e-05}
{"train_loss": 0.07083304971456528, "global_step": 184885, "epoch": 2077, "lr": 2.3140167285963658e-05}
{"train_loss": 0.06560368090867996, "global_step": 184886, "epoch": 2077, "lr": 2.313967830485785e-05}
{"train_loss": 0.13499462604522705, "global_step": 184887, "epoch": 2077, "lr": 2.3139189327363093e-05}
{"train_loss": 0.07416384667158127, "global_step": 184888, "epoch": 2077, "lr": 2.313870035347943e-05}
{"train_loss": 0.0782238021492958, "global_step": 184889, "epoch": 2077, "lr": 2.3138211383206946e-05}
{"train_loss": 0.04357810690999031, "global_step": 184890, "epoch": 2077, "lr": 2.3137722416545715e-05}
{"train_loss": 0.03334717079997063, "global_step": 184891, "epoch": 2077, "lr": 2.3137233453495778e-05}
{"train_loss": 0.05556129664182663, "global_step": 184892, "epoch": 2077, "lr": 2.3136744494057224e-05}
{"train_loss": 0.055350977927446365, "global_step": 184893, "epoch": 2077, "lr": 2.313625553823009e-05}
{"train_loss": 0.09545159339904785, "global_step": 184894, "epoch": 2077, "lr": 2.313576658601448e-05}
{"train_loss": 0.06165273115038872, "global_step": 184895, "epoch": 2077, "lr": 2.3135277637410418e-05}
{"train_loss": 0.0789107158780098, "global_step": 184896, "epoch": 2077, "lr": 2.3134788692418014e-05}
{"train_loss": 0.060652781277894974, "global_step": 184897, "epoch": 2077, "lr": 2.3134299751037287e-05}
{"train_loss": 0.06634214520454407, "global_step": 184898, "epoch": 2077, "lr": 2.3133810813268346e-05}
{"train_loss": 0.07538475096225739, "global_step": 184899, "epoch": 2077, "lr": 2.313332187911122e-05}
{"train_loss": 0.03815615922212601, "global_step": 184900, "epoch": 2077, "lr": 2.3132832948566e-05}
{"train_loss": 0.1021331176161766, "global_step": 184901, "epoch": 2077, "lr": 2.3132344021632736e-05}
{"train_loss": 0.08112490177154541, "global_step": 184902, "epoch": 2077, "lr": 2.3131855098311517e-05}
{"train_loss": 0.05942092090845108, "global_step": 184903, "epoch": 2077, "lr": 2.3131366178602393e-05}
{"train_loss": 0.07670863717794418, "global_step": 184904, "epoch": 2077, "lr": 2.3130877262505407e-05}
{"train_loss": 0.09348420798778534, "global_step": 184905, "epoch": 2077, "lr": 2.3130388350020667e-05}
{"train_loss": 0.05178753659129143, "global_step": 184906, "epoch": 2077, "lr": 2.31298994411482e-05}
{"train_loss": 0.10809677839279175, "global_step": 184907, "epoch": 2077, "lr": 2.3129410535888113e-05}
{"train_loss": 0.06602678447961807, "global_step": 184908, "epoch": 2077, "lr": 2.3128921634240424e-05}
{"train_loss": 0.07292930036783218, "global_step": 184909, "epoch": 2077, "lr": 2.3128432736205247e-05}
{"train_loss": 0.10246259719133377, "global_step": 184910, "epoch": 2077, "lr": 2.3127943841782607e-05}
{"train_loss": 0.05839208886027336, "global_step": 184911, "epoch": 2077, "lr": 2.3127454950972603e-05}
{"train_loss": 0.09176766872406006, "global_step": 184912, "epoch": 2077, "lr": 2.312696606377527e-05}
{"train_loss": 0.17937277257442474, "global_step": 184913, "epoch": 2077, "lr": 2.3126477180190705e-05}
{"train_loss": 0.03278103843331337, "global_step": 184914, "epoch": 2077, "lr": 2.3125988300218938e-05}
{"train_loss": 0.06063510850071907, "global_step": 184915, "epoch": 2077, "lr": 2.3125499423860076e-05}
{"train_loss": 0.04800458624958992, "global_step": 184916, "epoch": 2077, "lr": 2.312501055111414e-05}
{"train_loss": 0.03348636254668236, "global_step": 184917, "epoch": 2077, "lr": 2.312452168198124e-05}
{"train_loss": 0.15326763689517975, "global_step": 184918, "epoch": 2077, "lr": 2.31240328164614e-05}
{"train_loss": 0.1566418558359146, "global_step": 184919, "epoch": 2077, "lr": 2.3123543954554728e-05}
{"train_loss": 0.05083771049976349, "global_step": 184920, "epoch": 2077, "lr": 2.3123055096261248e-05}
{"train_loss": 0.04701400175690651, "global_step": 184921, "epoch": 2077, "lr": 2.312256624158107e-05}
{"train_loss": 0.0517231710255146, "global_step": 184922, "epoch": 2077, "lr": 2.3122077390514207e-05}
{"train_loss": 0.08730807900428772, "global_step": 184923, "epoch": 2077, "lr": 2.3121588543060774e-05}
{"train_loss": 0.06806027144193649, "global_step": 184924, "epoch": 2077, "lr": 2.31210996992208e-05}
{"train_loss": 0.07620812207460403, "global_step": 184925, "epoch": 2077, "lr": 2.3120610858994368e-05}
{"train_loss": 0.07505950331687927, "global_step": 184926, "epoch": 2077, "lr": 2.3120122022381557e-05}
{"train_loss": 0.1587122082710266, "global_step": 184927, "epoch": 2077, "lr": 2.31196331893824e-05}
{"train_loss": 0.08195152878761292, "global_step": 184928, "epoch": 2077, "lr": 2.3119144359997003e-05}
{"train_loss": 0.1200418472290039, "global_step": 184929, "epoch": 2077, "lr": 2.3118655534225387e-05}
{"train_loss": 0.030559809878468513, "global_step": 184930, "epoch": 2077, "lr": 2.3118166712067658e-05}
{"train_loss": 0.040558841079473495, "global_step": 184931, "epoch": 2077, "lr": 2.311767789352385e-05}
{"train_loss": 0.10062351077795029, "global_step": 184932, "epoch": 2077, "lr": 2.3117189078594052e-05}
{"train_loss": 0.04819633066654205, "global_step": 184933, "epoch": 2077, "lr": 2.3116700267278307e-05}
{"train_loss": 0.046642400324344635, "global_step": 184934, "epoch": 2077, "lr": 2.3116211459576713e-05}
{"train_loss": 0.06776965409517288, "global_step": 184935, "epoch": 2077, "lr": 2.3115722655489304e-05}
{"train_loss": 0.09534121304750443, "global_step": 184936, "epoch": 2077, "lr": 2.311523385501617e-05}
{"train_loss": 0.09242526441812515, "global_step": 184937, "epoch": 2077, "lr": 2.3114745058157344e-05}
{"train_loss": 0.08201342076063156, "global_step": 184938, "epoch": 2077, "lr": 2.3114256264912935e-05}
{"train_loss": 0.05895635485649109, "global_step": 184939, "epoch": 2077, "lr": 2.3113767475282972e-05}
{"train_loss": 0.07780762016773224, "global_step": 184940, "epoch": 2077, "lr": 2.311327868926755e-05}
{"train_loss": 0.07494817270238079, "global_step": 184941, "epoch": 2077, "lr": 2.311278990686672e-05, "val_loss": 7.601022243499756}
{"train_loss": 0.06374062597751617, "global_step": 184942, "epoch": 2078, "lr": 2.3112301128080528e-05}
{"train_loss": 0.07576951384544373, "global_step": 184943, "epoch": 2078, "lr": 2.311181235290908e-05}
{"train_loss": 0.06461746245622635, "global_step": 184944, "epoch": 2078, "lr": 2.31113235813524e-05}
{"train_loss": 0.07055911421775818, "global_step": 184945, "epoch": 2078, "lr": 2.3110834813410597e-05}
{"train_loss": 0.1442909687757492, "global_step": 184946, "epoch": 2078, "lr": 2.3110346049083693e-05}
{"train_loss": 0.08900018036365509, "global_step": 184947, "epoch": 2078, "lr": 2.3109857288371794e-05}
{"train_loss": 0.09967663884162903, "global_step": 184948, "epoch": 2078, "lr": 2.3109368531274933e-05}
{"train_loss": 0.06062738224864006, "global_step": 184949, "epoch": 2078, "lr": 2.3108879777793202e-05}
{"train_loss": 0.0576217919588089, "global_step": 184950, "epoch": 2078, "lr": 2.310839102792664e-05}
{"train_loss": 0.06822220981121063, "global_step": 184951, "epoch": 2078, "lr": 2.3107902281675343e-05}
{"train_loss": 0.07731817662715912, "global_step": 184952, "epoch": 2078, "lr": 2.3107413539039347e-05}
{"train_loss": 0.0682745948433876, "global_step": 184953, "epoch": 2078, "lr": 2.3106924800018742e-05}
{"train_loss": 0.11347207427024841, "global_step": 184954, "epoch": 2078, "lr": 2.310643606461357e-05}
{"train_loss": 0.08978339284658432, "global_step": 184955, "epoch": 2078, "lr": 2.3105947332823928e-05}
{"train_loss": 0.10841268301010132, "global_step": 184956, "epoch": 2078, "lr": 2.3105458604649844e-05}
{"train_loss": 0.0710081085562706, "global_step": 184957, "epoch": 2078, "lr": 2.3104969880091422e-05}
{"train_loss": 0.09074416011571884, "global_step": 184958, "epoch": 2078, "lr": 2.310448115914869e-05}
{"train_loss": 0.07159488648176193, "global_step": 184959, "epoch": 2078, "lr": 2.3103992441821753e-05}
{"train_loss": 0.08532638102769852, "global_step": 184960, "epoch": 2078, "lr": 2.3103503728110638e-05}
{"train_loss": 0.07292490452528, "global_step": 184961, "epoch": 2078, "lr": 2.310301501801543e-05}
{"train_loss": 0.05427232012152672, "global_step": 184962, "epoch": 2078, "lr": 2.3102526311536203e-05}
{"train_loss": 0.062134940177202225, "global_step": 184963, "epoch": 2078, "lr": 2.3102037608673005e-05}
{"train_loss": 0.07160282135009766, "global_step": 184964, "epoch": 2078, "lr": 2.3101548909425923e-05}
{"train_loss": 0.12744027376174927, "global_step": 184965, "epoch": 2078, "lr": 2.3101060213794994e-05}
{"train_loss": 0.11322301626205444, "global_step": 184966, "epoch": 2078, "lr": 2.3100571521780323e-05}
{"train_loss": 0.086402527987957, "global_step": 184967, "epoch": 2078, "lr": 2.3100082833381924e-05}
{"train_loss": 0.1085192933678627, "global_step": 184968, "epoch": 2078, "lr": 2.3099594148599918e-05}
{"train_loss": 0.1130523607134819, "global_step": 184969, "epoch": 2078, "lr": 2.3099105467434323e-05}
{"train_loss": 0.08127599209547043, "global_step": 184970, "epoch": 2078, "lr": 2.309861678988524e-05}
{"train_loss": 0.12816183269023895, "global_step": 184971, "epoch": 2078, "lr": 2.3098128115952707e-05}
{"train_loss": 0.09294699132442474, "global_step": 184972, "epoch": 2078, "lr": 2.3097639445636814e-05}
{"train_loss": 0.10552731901407242, "global_step": 184973, "epoch": 2078, "lr": 2.3097150778937603e-05}
{"train_loss": 0.08948806673288345, "global_step": 184974, "epoch": 2078, "lr": 2.3096662115855172e-05}
{"train_loss": 0.06438253074884415, "global_step": 184975, "epoch": 2078, "lr": 2.309617345638954e-05}
{"train_loss": 0.1020662933588028, "global_step": 184976, "epoch": 2078, "lr": 2.309568480054083e-05}
{"train_loss": 0.16134724020957947, "global_step": 184977, "epoch": 2078, "lr": 2.309519614830905e-05}
{"train_loss": 0.13592314720153809, "global_step": 184978, "epoch": 2078, "lr": 2.3094707499694314e-05}
{"train_loss": 0.0601479634642601, "global_step": 184979, "epoch": 2078, "lr": 2.3094218854696646e-05}
{"train_loss": 0.056951556354761124, "global_step": 184980, "epoch": 2078, "lr": 2.3093730213316155e-05}
{"train_loss": 0.04789276793599129, "global_step": 184981, "epoch": 2078, "lr": 2.309324157555288e-05}
{"train_loss": 0.08174889534711838, "global_step": 184982, "epoch": 2078, "lr": 2.3092752941406877e-05}
{"train_loss": 0.08848229795694351, "global_step": 184983, "epoch": 2078, "lr": 2.3092264310878243e-05}
{"train_loss": 0.08303163200616837, "global_step": 184984, "epoch": 2078, "lr": 2.3091775683967006e-05}
{"train_loss": 0.07198241353034973, "global_step": 184985, "epoch": 2078, "lr": 2.3091287060673266e-05}
{"train_loss": 0.10131952166557312, "global_step": 184986, "epoch": 2078, "lr": 2.3090798440997062e-05}
{"train_loss": 0.08916491270065308, "global_step": 184987, "epoch": 2078, "lr": 2.309030982493849e-05}
{"train_loss": 0.06463415920734406, "global_step": 184988, "epoch": 2078, "lr": 2.3089821212497576e-05}
{"train_loss": 0.07055284827947617, "global_step": 184989, "epoch": 2078, "lr": 2.308933260367443e-05}
{"train_loss": 0.05683664605021477, "global_step": 184990, "epoch": 2078, "lr": 2.3088843998469075e-05}
{"train_loss": 0.06603644788265228, "global_step": 184991, "epoch": 2078, "lr": 2.308835539688161e-05}
{"train_loss": 0.09763611853122711, "global_step": 184992, "epoch": 2078, "lr": 2.3087866798912077e-05}
{"train_loss": 0.1297437995672226, "global_step": 184993, "epoch": 2078, "lr": 2.3087378204560566e-05}
{"train_loss": 0.07881317287683487, "global_step": 184994, "epoch": 2078, "lr": 2.3086889613827112e-05}
{"train_loss": 0.076966293156147, "global_step": 184995, "epoch": 2078, "lr": 2.3086401026711817e-05}
{"train_loss": 0.11182790994644165, "global_step": 184996, "epoch": 2078, "lr": 2.3085912443214714e-05}
{"train_loss": 0.0937802642583847, "global_step": 184997, "epoch": 2078, "lr": 2.3085423863335876e-05}
{"train_loss": 0.08402954041957855, "global_step": 184998, "epoch": 2078, "lr": 2.3084935287075392e-05}
{"train_loss": 0.0992451161146164, "global_step": 184999, "epoch": 2078, "lr": 2.3084446714433295e-05}
{"train_loss": 0.0515628345310688, "global_step": 185000, "epoch": 2078, "lr": 2.3083958145409678e-05}
{"train_loss": 0.0799640342593193, "global_step": 185001, "epoch": 2078, "lr": 2.3083469580004586e-05}
{"train_loss": 0.03295927494764328, "global_step": 185002, "epoch": 2078, "lr": 2.3082981018218107e-05}
{"train_loss": 0.05130202695727348, "global_step": 185003, "epoch": 2078, "lr": 2.3082492460050275e-05}
{"train_loss": 0.06823109835386276, "global_step": 185004, "epoch": 2078, "lr": 2.3082003905501194e-05}
{"train_loss": 0.08118000626564026, "global_step": 185005, "epoch": 2078, "lr": 2.308151535457089e-05}
{"train_loss": 0.041471999138593674, "global_step": 185006, "epoch": 2078, "lr": 2.3081026807259464e-05}
{"train_loss": 0.05188382789492607, "global_step": 185007, "epoch": 2078, "lr": 2.3080538263566947e-05}
{"train_loss": 0.05006378889083862, "global_step": 185008, "epoch": 2078, "lr": 2.3080049723493446e-05}
{"train_loss": 0.06455816328525543, "global_step": 185009, "epoch": 2078, "lr": 2.3079561187038985e-05}
{"train_loss": 0.13046737015247345, "global_step": 185010, "epoch": 2078, "lr": 2.3079072654203665e-05}
{"train_loss": 0.0669686496257782, "global_step": 185011, "epoch": 2078, "lr": 2.3078584124987517e-05}
{"train_loss": 0.033442422747612, "global_step": 185012, "epoch": 2078, "lr": 2.307809559939065e-05}
{"train_loss": 0.07476170361042023, "global_step": 185013, "epoch": 2078, "lr": 2.3077607077413082e-05}
{"train_loss": 0.060768142342567444, "global_step": 185014, "epoch": 2078, "lr": 2.3077118559054917e-05}
{"train_loss": 0.05324290320277214, "global_step": 185015, "epoch": 2078, "lr": 2.307663004431619e-05}
{"train_loss": 0.05788722261786461, "global_step": 185016, "epoch": 2078, "lr": 2.3076141533196998e-05}
{"train_loss": 0.07414747774600983, "global_step": 185017, "epoch": 2078, "lr": 2.3075653025697375e-05}
{"train_loss": 0.051987625658512115, "global_step": 185018, "epoch": 2078, "lr": 2.3075164521817422e-05}
{"train_loss": 0.054551903158426285, "global_step": 185019, "epoch": 2078, "lr": 2.3074676021557184e-05}
{"train_loss": 0.09112689644098282, "global_step": 185020, "epoch": 2078, "lr": 2.3074187524916707e-05}
{"train_loss": 0.05776269733905792, "global_step": 185021, "epoch": 2078, "lr": 2.3073699031896096e-05}
{"train_loss": 0.03839895501732826, "global_step": 185022, "epoch": 2078, "lr": 2.3073210542495382e-05}
{"train_loss": 0.04041726514697075, "global_step": 185023, "epoch": 2078, "lr": 2.3072722056714663e-05}
{"train_loss": 0.035390984266996384, "global_step": 185024, "epoch": 2078, "lr": 2.307223357455397e-05}
{"train_loss": 0.037187494337558746, "global_step": 185025, "epoch": 2078, "lr": 2.307174509601341e-05}
{"train_loss": 0.04764487221837044, "global_step": 185026, "epoch": 2078, "lr": 2.3071256621093e-05}
{"train_loss": 0.08730453252792358, "global_step": 185027, "epoch": 2078, "lr": 2.3070768149792856e-05}
{"train_loss": 0.1187666654586792, "global_step": 185028, "epoch": 2078, "lr": 2.3070279682112994e-05}
{"train_loss": 0.06091393902897835, "global_step": 185029, "epoch": 2078, "lr": 2.3069791218053527e-05}
{"train_loss": 0.07900070482760332, "global_step": 185030, "epoch": 2078, "lr": 2.306930275761448e-05, "val_loss": 7.71511173248291}
{"train_loss": 0.10145650058984756, "global_step": 185031, "epoch": 2079, "lr": 2.3068814300795955e-05}
{"train_loss": 0.056898169219493866, "global_step": 185032, "epoch": 2079, "lr": 2.3068325847597978e-05}
{"train_loss": 0.0873330682516098, "global_step": 185033, "epoch": 2079, "lr": 2.3067837398020657e-05}
{"train_loss": 0.06602049618959427, "global_step": 185034, "epoch": 2079, "lr": 2.3067348952064023e-05}
{"train_loss": 0.0471254400908947, "global_step": 185035, "epoch": 2079, "lr": 2.306686050972815e-05}
{"train_loss": 0.07818777114152908, "global_step": 185036, "epoch": 2079, "lr": 2.306637207101313e-05}
{"train_loss": 0.09358260780572891, "global_step": 185037, "epoch": 2079, "lr": 2.306588363591899e-05}
{"train_loss": 0.06702858954668045, "global_step": 185038, "epoch": 2079, "lr": 2.3065395204445833e-05}
{"train_loss": 0.06511076539754868, "global_step": 185039, "epoch": 2079, "lr": 2.306490677659368e-05}
{"train_loss": 0.05701734125614166, "global_step": 185040, "epoch": 2079, "lr": 2.3064418352362648e-05}
{"train_loss": 0.09537512063980103, "global_step": 185041, "epoch": 2079, "lr": 2.306392993175276e-05}
{"train_loss": 0.034679897129535675, "global_step": 185042, "epoch": 2079, "lr": 2.306344151476411e-05}
{"train_loss": 0.04774104803800583, "global_step": 185043, "epoch": 2079, "lr": 2.306295310139674e-05}
{"train_loss": 0.07921183854341507, "global_step": 185044, "epoch": 2079, "lr": 2.3062464691650736e-05}
{"train_loss": 0.0983826220035553, "global_step": 185045, "epoch": 2079, "lr": 2.3061976285526148e-05}
{"train_loss": 0.044770497828722, "global_step": 185046, "epoch": 2079, "lr": 2.306148788302306e-05}
{"train_loss": 0.08071941137313843, "global_step": 185047, "epoch": 2079, "lr": 2.306099948414151e-05}
{"train_loss": 0.03958941996097565, "global_step": 185048, "epoch": 2079, "lr": 2.3060511088881604e-05}
{"train_loss": 0.09072476625442505, "global_step": 185049, "epoch": 2079, "lr": 2.306002269724336e-05}
{"train_loss": 0.03490530699491501, "global_step": 185050, "epoch": 2079, "lr": 2.305953430922689e-05}
{"train_loss": 0.11852727830410004, "global_step": 185051, "epoch": 2079, "lr": 2.3059045924832212e-05}
{"train_loss": 0.084299735724926, "global_step": 185052, "epoch": 2079, "lr": 2.305855754405944e-05}
{"train_loss": 0.052719514816999435, "global_step": 185053, "epoch": 2079, "lr": 2.30580691669086e-05}
{"train_loss": 0.02962086908519268, "global_step": 185054, "epoch": 2079, "lr": 2.3057580793379786e-05}
{"train_loss": 0.10087377578020096, "global_step": 185055, "epoch": 2079, "lr": 2.3057092423473036e-05}
{"train_loss": 0.07824371010065079, "global_step": 185056, "epoch": 2079, "lr": 2.3056604057188452e-05}
{"train_loss": 0.12558309733867645, "global_step": 185057, "epoch": 2079, "lr": 2.305611569452608e-05}
{"train_loss": 0.043202728033065796, "global_step": 185058, "epoch": 2079, "lr": 2.3055627335485963e-05}
{"train_loss": 0.06986675411462784, "global_step": 185059, "epoch": 2079, "lr": 2.3055138980068212e-05}
{"train_loss": 0.05547284334897995, "global_step": 185060, "epoch": 2079, "lr": 2.3054650628272846e-05}
{"train_loss": 0.06996171176433563, "global_step": 185061, "epoch": 2079, "lr": 2.305416228009998e-05}
{"train_loss": 0.08669161051511765, "global_step": 185062, "epoch": 2079, "lr": 2.3053673935549624e-05}
{"train_loss": 0.05381381884217262, "global_step": 185063, "epoch": 2079, "lr": 2.3053185594621896e-05}
{"train_loss": 0.054888077080249786, "global_step": 185064, "epoch": 2079, "lr": 2.3052697257316824e-05}
{"train_loss": 0.08394729346036911, "global_step": 185065, "epoch": 2079, "lr": 2.30522089236345e-05}
{"train_loss": 0.10224194079637527, "global_step": 185066, "epoch": 2079, "lr": 2.3051720593574966e-05}
{"train_loss": 0.09888871759176254, "global_step": 185067, "epoch": 2079, "lr": 2.3051232267138313e-05}
{"train_loss": 0.07046380639076233, "global_step": 185068, "epoch": 2079, "lr": 2.3050743944324575e-05}
{"train_loss": 0.06182579696178436, "global_step": 185069, "epoch": 2079, "lr": 2.305025562513386e-05}
{"train_loss": 0.0588274747133255, "global_step": 185070, "epoch": 2079, "lr": 2.304976730956619e-05}
{"train_loss": 0.07391124963760376, "global_step": 185071, "epoch": 2079, "lr": 2.304927899762165e-05}
{"train_loss": 0.06630919128656387, "global_step": 185072, "epoch": 2079, "lr": 2.304879068930032e-05}
{"train_loss": 0.07582960277795792, "global_step": 185073, "epoch": 2079, "lr": 2.3048302384602238e-05}
{"train_loss": 0.056754790246486664, "global_step": 185074, "epoch": 2079, "lr": 2.3047814083527496e-05}
{"train_loss": 0.08369041979312897, "global_step": 185075, "epoch": 2079, "lr": 2.3047325786076134e-05}
{"train_loss": 0.05312538146972656, "global_step": 185076, "epoch": 2079, "lr": 2.3046837492248246e-05}
{"train_loss": 0.1722266972064972, "global_step": 185077, "epoch": 2079, "lr": 2.3046349202043864e-05}
{"train_loss": 0.05927269160747528, "global_step": 185078, "epoch": 2079, "lr": 2.3045860915463092e-05}
{"train_loss": 0.10742352157831192, "global_step": 185079, "epoch": 2079, "lr": 2.3045372632505958e-05}
{"train_loss": 0.040664203464984894, "global_step": 185080, "epoch": 2079, "lr": 2.304488435317256e-05}
{"train_loss": 0.07611878216266632, "global_step": 185081, "epoch": 2079, "lr": 2.3044396077462933e-05}
{"train_loss": 0.056308966130018234, "global_step": 185082, "epoch": 2079, "lr": 2.3043907805377175e-05}
{"train_loss": 0.028454696759581566, "global_step": 185083, "epoch": 2079, "lr": 2.3043419536915316e-05}
{"train_loss": 0.10505050420761108, "global_step": 185084, "epoch": 2079, "lr": 2.3042931272077463e-05}
{"train_loss": 0.07505062967538834, "global_step": 185085, "epoch": 2079, "lr": 2.304244301086364e-05}
{"train_loss": 0.07751329243183136, "global_step": 185086, "epoch": 2079, "lr": 2.3041954753273947e-05}
{"train_loss": 0.056082647293806076, "global_step": 185087, "epoch": 2079, "lr": 2.3041466499308416e-05}
{"train_loss": 0.016626477241516113, "global_step": 185088, "epoch": 2079, "lr": 2.3040978248967132e-05}
{"train_loss": 0.039969220757484436, "global_step": 185089, "epoch": 2079, "lr": 2.304049000225018e-05}
{"train_loss": 0.06565165519714355, "global_step": 185090, "epoch": 2079, "lr": 2.3040001759157588e-05}
{"train_loss": 0.0522347092628479, "global_step": 185091, "epoch": 2079, "lr": 2.303951351968946e-05}
{"train_loss": 0.03541157394647598, "global_step": 185092, "epoch": 2079, "lr": 2.3039025283845832e-05}
{"train_loss": 0.06542810797691345, "global_step": 185093, "epoch": 2079, "lr": 2.3038537051626768e-05}
{"train_loss": 0.06784188002347946, "global_step": 185094, "epoch": 2079, "lr": 2.303804882303236e-05}
{"train_loss": 0.06376296281814575, "global_step": 185095, "epoch": 2079, "lr": 2.303756059806264e-05}
{"train_loss": 0.12491191178560257, "global_step": 185096, "epoch": 2079, "lr": 2.3037072376717707e-05}
{"train_loss": 0.07690457254648209, "global_step": 185097, "epoch": 2079, "lr": 2.3036584158997614e-05}
{"train_loss": 0.11890026181936264, "global_step": 185098, "epoch": 2079, "lr": 2.3036095944902404e-05}
{"train_loss": 0.0751488208770752, "global_step": 185099, "epoch": 2079, "lr": 2.303560773443218e-05}
{"train_loss": 0.06658695638179779, "global_step": 185100, "epoch": 2079, "lr": 2.3035119527586974e-05}
{"train_loss": 0.1277453899383545, "global_step": 185101, "epoch": 2079, "lr": 2.3034631324366886e-05}
{"train_loss": 0.09129202365875244, "global_step": 185102, "epoch": 2079, "lr": 2.303414312477194e-05}
{"train_loss": 0.13169698417186737, "global_step": 185103, "epoch": 2079, "lr": 2.3033654928802245e-05}
{"train_loss": 0.06904471665620804, "global_step": 185104, "epoch": 2079, "lr": 2.303316673645783e-05}
{"train_loss": 0.05420096591114998, "global_step": 185105, "epoch": 2079, "lr": 2.3032678547738796e-05}
{"train_loss": 0.1063164621591568, "global_step": 185106, "epoch": 2079, "lr": 2.3032190362645168e-05}
{"train_loss": 0.09324592351913452, "global_step": 185107, "epoch": 2079, "lr": 2.3031702181177034e-05}
{"train_loss": 0.08080312609672546, "global_step": 185108, "epoch": 2079, "lr": 2.303121400333448e-05}
{"train_loss": 0.04633686691522598, "global_step": 185109, "epoch": 2079, "lr": 2.3030725829117528e-05}
{"train_loss": 0.07556985318660736, "global_step": 185110, "epoch": 2079, "lr": 2.3030237658526286e-05}
{"train_loss": 0.029568172991275787, "global_step": 185111, "epoch": 2079, "lr": 2.302974949156078e-05}
{"train_loss": 0.09437563270330429, "global_step": 185112, "epoch": 2079, "lr": 2.302926132822112e-05}
{"train_loss": 0.07178765535354614, "global_step": 185113, "epoch": 2079, "lr": 2.3028773168507327e-05}
{"train_loss": 0.06757520139217377, "global_step": 185114, "epoch": 2079, "lr": 2.3028285012419505e-05}
{"train_loss": 0.0802941843867302, "global_step": 185115, "epoch": 2079, "lr": 2.302779685995768e-05}
{"train_loss": 0.05270454287528992, "global_step": 185116, "epoch": 2079, "lr": 2.3027308711121958e-05}
{"train_loss": 0.057720012962818146, "global_step": 185117, "epoch": 2079, "lr": 2.3026820565912367e-05}
{"train_loss": 0.09625179320573807, "global_step": 185118, "epoch": 2079, "lr": 2.302633242432901e-05}
{"train_loss": 0.07285549709301316, "global_step": 185119, "epoch": 2079, "lr": 2.302584428637192e-05, "val_loss": 7.533387660980225}
{"train_loss": 0.05212094634771347, "global_step": 185120, "epoch": 2080, "lr": 2.302535615204119e-05}
{"train_loss": 0.08070127665996552, "global_step": 185121, "epoch": 2080, "lr": 2.3024868021336848e-05}
{"train_loss": 0.03820934146642685, "global_step": 185122, "epoch": 2080, "lr": 2.3024379894259014e-05}
{"train_loss": 0.13291621208190918, "global_step": 185123, "epoch": 2080, "lr": 2.3023891770807697e-05}
{"train_loss": 0.09394946694374084, "global_step": 185124, "epoch": 2080, "lr": 2.302340365098299e-05}
{"train_loss": 0.07439261674880981, "global_step": 185125, "epoch": 2080, "lr": 2.302291553478498e-05}
{"train_loss": 0.07858153432607651, "global_step": 185126, "epoch": 2080, "lr": 2.302242742221369e-05}
{"train_loss": 0.09932821989059448, "global_step": 185127, "epoch": 2080, "lr": 2.302193931326922e-05}
{"train_loss": 0.07959447801113129, "global_step": 185128, "epoch": 2080, "lr": 2.3021451207951605e-05}
{"train_loss": 0.11311103403568268, "global_step": 185129, "epoch": 2080, "lr": 2.302096310626095e-05}
{"train_loss": 0.03449244797229767, "global_step": 185130, "epoch": 2080, "lr": 2.3020475008197296e-05}
{"train_loss": 0.10855426639318466, "global_step": 185131, "epoch": 2080, "lr": 2.301998691376069e-05}
{"train_loss": 0.07474619895219803, "global_step": 185132, "epoch": 2080, "lr": 2.3019498822951235e-05}
{"train_loss": 0.06803858280181885, "global_step": 185133, "epoch": 2080, "lr": 2.3019010735768964e-05}
{"train_loss": 0.05494226887822151, "global_step": 185134, "epoch": 2080, "lr": 2.3018522652213974e-05}
{"train_loss": 0.05308595299720764, "global_step": 185135, "epoch": 2080, "lr": 2.3018034572286317e-05}
{"train_loss": 0.10783524066209793, "global_step": 185136, "epoch": 2080, "lr": 2.3017546495986038e-05}
{"train_loss": 0.08763124793767929, "global_step": 185137, "epoch": 2080, "lr": 2.301705842331324e-05}
{"train_loss": 0.10256042331457138, "global_step": 185138, "epoch": 2080, "lr": 2.3016570354267946e-05}
{"train_loss": 0.05651042237877846, "global_step": 185139, "epoch": 2080, "lr": 2.3016082288850272e-05}
{"train_loss": 0.043796658515930176, "global_step": 185140, "epoch": 2080, "lr": 2.3015594227060233e-05}
{"train_loss": 0.07827441394329071, "global_step": 185141, "epoch": 2080, "lr": 2.301510616889794e-05}
{"train_loss": 0.05564473196864128, "global_step": 185142, "epoch": 2080, "lr": 2.301461811436341e-05}
{"train_loss": 0.04123834893107414, "global_step": 185143, "epoch": 2080, "lr": 2.3014130063456746e-05}
{"train_loss": 0.04265867918729782, "global_step": 185144, "epoch": 2080, "lr": 2.3013642016178015e-05}
{"train_loss": 0.051683273166418076, "global_step": 185145, "epoch": 2080, "lr": 2.3013153972527258e-05}
{"train_loss": 0.07877064496278763, "global_step": 185146, "epoch": 2080, "lr": 2.3012665932504567e-05}
{"train_loss": 0.07689367979764938, "global_step": 185147, "epoch": 2080, "lr": 2.3012177896109975e-05}
{"train_loss": 0.08354909718036652, "global_step": 185148, "epoch": 2080, "lr": 2.3011689863343595e-05}
{"train_loss": 0.06010827049612999, "global_step": 185149, "epoch": 2080, "lr": 2.3011201834205437e-05}
{"train_loss": 0.07100086659193039, "global_step": 185150, "epoch": 2080, "lr": 2.301071380869561e-05}
{"train_loss": 0.0920664519071579, "global_step": 185151, "epoch": 2080, "lr": 2.3010225786814153e-05}
{"train_loss": 0.06440120190382004, "global_step": 185152, "epoch": 2080, "lr": 2.3009737768561158e-05}
{"train_loss": 0.042718466371297836, "global_step": 185153, "epoch": 2080, "lr": 2.3009249753936658e-05}
{"train_loss": 0.07356879115104675, "global_step": 185154, "epoch": 2080, "lr": 2.300876174294075e-05}
{"train_loss": 0.10797359049320221, "global_step": 185155, "epoch": 2080, "lr": 2.3008273735573465e-05}
{"train_loss": 0.06873157620429993, "global_step": 185156, "epoch": 2080, "lr": 2.3007785731834914e-05}
{"train_loss": 0.08748574554920197, "global_step": 185157, "epoch": 2080, "lr": 2.3007297731725113e-05}
{"train_loss": 0.05819379910826683, "global_step": 185158, "epoch": 2080, "lr": 2.3006809735244174e-05}
{"train_loss": 0.05652106925845146, "global_step": 185159, "epoch": 2080, "lr": 2.3006321742392118e-05}
{"train_loss": 0.1291305124759674, "global_step": 185160, "epoch": 2080, "lr": 2.3005833753169055e-05}
{"train_loss": 0.060037434101104736, "global_step": 185161, "epoch": 2080, "lr": 2.300534576757501e-05}
{"train_loss": 0.05064890533685684, "global_step": 185162, "epoch": 2080, "lr": 2.300485778561006e-05}
{"train_loss": 0.09566579759120941, "global_step": 185163, "epoch": 2080, "lr": 2.3004369807274306e-05}
{"train_loss": 0.09013817459344864, "global_step": 185164, "epoch": 2080, "lr": 2.3003881832567766e-05}
{"train_loss": 0.060247618705034256, "global_step": 185165, "epoch": 2080, "lr": 2.3003393861490534e-05}
{"train_loss": 0.09416370093822479, "global_step": 185166, "epoch": 2080, "lr": 2.300290589404266e-05}
{"train_loss": 0.147109717130661, "global_step": 185167, "epoch": 2080, "lr": 2.300241793022423e-05}
{"train_loss": 0.03718653321266174, "global_step": 185168, "epoch": 2080, "lr": 2.3001929970035298e-05}
{"train_loss": 0.06422121822834015, "global_step": 185169, "epoch": 2080, "lr": 2.3001442013475906e-05}
{"train_loss": 0.06342365592718124, "global_step": 185170, "epoch": 2080, "lr": 2.3000954060546155e-05}
{"train_loss": 0.07403811067342758, "global_step": 185171, "epoch": 2080, "lr": 2.3000466111246083e-05}
{"train_loss": 0.09016667306423187, "global_step": 185172, "epoch": 2080, "lr": 2.299997816557579e-05}
{"train_loss": 0.09059296548366547, "global_step": 185173, "epoch": 2080, "lr": 2.299949022353532e-05}
{"train_loss": 0.14498144388198853, "global_step": 185174, "epoch": 2080, "lr": 2.299900228512472e-05}
{"train_loss": 0.07032687216997147, "global_step": 185175, "epoch": 2080, "lr": 2.299851435034409e-05}
{"train_loss": 0.1411592811346054, "global_step": 185176, "epoch": 2080, "lr": 2.299802641919347e-05}
{"train_loss": 0.023917725309729576, "global_step": 185177, "epoch": 2080, "lr": 2.2997538491672948e-05}
{"train_loss": 0.05915592610836029, "global_step": 185178, "epoch": 2080, "lr": 2.2997050567782563e-05}
{"train_loss": 0.036895181983709335, "global_step": 185179, "epoch": 2080, "lr": 2.2996562647522392e-05}
{"train_loss": 0.08826343715190887, "global_step": 185180, "epoch": 2080, "lr": 2.2996074730892527e-05}
{"train_loss": 0.08726987987756729, "global_step": 185181, "epoch": 2080, "lr": 2.299558681789299e-05}
{"train_loss": 0.033919867128133774, "global_step": 185182, "epoch": 2080, "lr": 2.299509890852388e-05}
{"train_loss": 0.0434378981590271, "global_step": 185183, "epoch": 2080, "lr": 2.299461100278524e-05}
{"train_loss": 0.09467718750238419, "global_step": 185184, "epoch": 2080, "lr": 2.299412310067716e-05}
{"train_loss": 0.053637560456991196, "global_step": 185185, "epoch": 2080, "lr": 2.2993635202199675e-05}
{"train_loss": 0.029379894956946373, "global_step": 185186, "epoch": 2080, "lr": 2.299314730735288e-05}
{"train_loss": 0.0990835577249527, "global_step": 185187, "epoch": 2080, "lr": 2.2992659416136813e-05}
{"train_loss": 0.06351363658905029, "global_step": 185188, "epoch": 2080, "lr": 2.299217152855157e-05}
{"train_loss": 0.06576411426067352, "global_step": 185189, "epoch": 2080, "lr": 2.2991683644597184e-05}
{"train_loss": 0.12069658190011978, "global_step": 185190, "epoch": 2080, "lr": 2.2991195764273753e-05}
{"train_loss": 0.12043297290802002, "global_step": 185191, "epoch": 2080, "lr": 2.2990707887581308e-05}
{"train_loss": 0.0877208337187767, "global_step": 185192, "epoch": 2080, "lr": 2.2990220014519958e-05}
{"train_loss": 0.06643465906381607, "global_step": 185193, "epoch": 2080, "lr": 2.2989732145089722e-05}
{"train_loss": 0.09594781696796417, "global_step": 185194, "epoch": 2080, "lr": 2.29892442792907e-05}
{"train_loss": 0.06828442215919495, "global_step": 185195, "epoch": 2080, "lr": 2.2988756417122932e-05}
{"train_loss": 0.09276364743709564, "global_step": 185196, "epoch": 2080, "lr": 2.2988268558586513e-05}
{"train_loss": 0.03138671815395355, "global_step": 185197, "epoch": 2080, "lr": 2.2987780703681473e-05}
{"train_loss": 0.02709553763270378, "global_step": 185198, "epoch": 2080, "lr": 2.29872928524079e-05}
{"train_loss": 0.06595230102539062, "global_step": 185199, "epoch": 2080, "lr": 2.298680500476587e-05}
{"train_loss": 0.07455477863550186, "global_step": 185200, "epoch": 2080, "lr": 2.2986317160755422e-05}
{"train_loss": 0.06286592036485672, "global_step": 185201, "epoch": 2080, "lr": 2.2985829320376645e-05}
{"train_loss": 0.06862019747495651, "global_step": 185202, "epoch": 2080, "lr": 2.298534148362958e-05}
{"train_loss": 0.05029473453760147, "global_step": 185203, "epoch": 2080, "lr": 2.2984853650514326e-05}
{"train_loss": 0.06608730554580688, "global_step": 185204, "epoch": 2080, "lr": 2.2984365821030913e-05}
{"train_loss": 0.04735444113612175, "global_step": 185205, "epoch": 2080, "lr": 2.2983877995179436e-05}
{"train_loss": 0.06392789632081985, "global_step": 185206, "epoch": 2080, "lr": 2.2983390172959934e-05}
{"train_loss": 0.050111543387174606, "global_step": 185207, "epoch": 2080, "lr": 2.29829023543725e-05}
{"train_loss": 0.07351189604803418, "global_step": 185208, "epoch": 2080, "lr": 2.2982414539417185e-05, "val_loss": 7.583392143249512, "train_action_mse_error": 15.885141372680664}
{"train_loss": 0.027986785396933556, "global_step": 185209, "epoch": 2081, "lr": 2.298192672809404e-05}
{"train_loss": 0.04311157017946243, "global_step": 185210, "epoch": 2081, "lr": 2.2981438920403165e-05}
{"train_loss": 0.046866729855537415, "global_step": 185211, "epoch": 2081, "lr": 2.2980951116344585e-05}
{"train_loss": 0.109612837433815, "global_step": 185212, "epoch": 2081, "lr": 2.2980463315918404e-05}
{"train_loss": 0.0440528467297554, "global_step": 185213, "epoch": 2081, "lr": 2.297997551912467e-05}
{"train_loss": 0.04004804417490959, "global_step": 185214, "epoch": 2081, "lr": 2.2979487725963435e-05}
{"train_loss": 0.10258152335882187, "global_step": 185215, "epoch": 2081, "lr": 2.2978999936434776e-05}
{"train_loss": 0.04494275525212288, "global_step": 185216, "epoch": 2081, "lr": 2.2978512150538778e-05}
{"train_loss": 0.050827398896217346, "global_step": 185217, "epoch": 2081, "lr": 2.2978024368275473e-05}
{"train_loss": 0.08253112435340881, "global_step": 185218, "epoch": 2081, "lr": 2.297753658964496e-05}
{"train_loss": 0.05073525756597519, "global_step": 185219, "epoch": 2081, "lr": 2.2977048814647274e-05}
{"train_loss": 0.11385434120893478, "global_step": 185220, "epoch": 2081, "lr": 2.297656104328251e-05}
{"train_loss": 0.07377965748310089, "global_step": 185221, "epoch": 2081, "lr": 2.29760732755507e-05}
{"train_loss": 0.03624047338962555, "global_step": 185222, "epoch": 2081, "lr": 2.297558551145194e-05}
{"train_loss": 0.09728574007749557, "global_step": 185223, "epoch": 2081, "lr": 2.2975097750986273e-05}
{"train_loss": 0.12927967309951782, "global_step": 185224, "epoch": 2081, "lr": 2.2974609994153788e-05}
{"train_loss": 0.09049966186285019, "global_step": 185225, "epoch": 2081, "lr": 2.297412224095452e-05}
{"train_loss": 0.07252035290002823, "global_step": 185226, "epoch": 2081, "lr": 2.2973634491388573e-05}
{"train_loss": 0.08479718118906021, "global_step": 185227, "epoch": 2081, "lr": 2.2973146745455975e-05}
{"train_loss": 0.09825063496828079, "global_step": 185228, "epoch": 2081, "lr": 2.297265900315682e-05}
{"train_loss": 0.05034813657402992, "global_step": 185229, "epoch": 2081, "lr": 2.2972171264491148e-05}
{"train_loss": 0.049586597830057144, "global_step": 185230, "epoch": 2081, "lr": 2.2971683529459054e-05}
{"train_loss": 0.0806453600525856, "global_step": 185231, "epoch": 2081, "lr": 2.297119579806057e-05}
{"train_loss": 0.05758262053132057, "global_step": 185232, "epoch": 2081, "lr": 2.2970708070295798e-05}
{"train_loss": 0.04558141902089119, "global_step": 185233, "epoch": 2081, "lr": 2.2970220346164768e-05}
{"train_loss": 0.046998437494039536, "global_step": 185234, "epoch": 2081, "lr": 2.2969732625667557e-05}
{"train_loss": 0.0757613554596901, "global_step": 185235, "epoch": 2081, "lr": 2.2969244908804262e-05}
{"train_loss": 0.06547225266695023, "global_step": 185236, "epoch": 2081, "lr": 2.2968757195574898e-05}
{"train_loss": 0.0639055073261261, "global_step": 185237, "epoch": 2081, "lr": 2.2968269485979577e-05}
{"train_loss": 0.0743609145283699, "global_step": 185238, "epoch": 2081, "lr": 2.296778178001832e-05}
{"train_loss": 0.11486787348985672, "global_step": 185239, "epoch": 2081, "lr": 2.2967294077691237e-05}
{"train_loss": 0.08233616501092911, "global_step": 185240, "epoch": 2081, "lr": 2.296680637899835e-05}
{"train_loss": 0.06475933641195297, "global_step": 185241, "epoch": 2081, "lr": 2.296631868393977e-05}
{"train_loss": 0.050423625856637955, "global_step": 185242, "epoch": 2081, "lr": 2.296583099251552e-05}
{"train_loss": 0.03494706004858017, "global_step": 185243, "epoch": 2081, "lr": 2.29653433047257e-05}
{"train_loss": 0.06113729625940323, "global_step": 185244, "epoch": 2081, "lr": 2.296485562057034e-05}
{"train_loss": 0.04985853284597397, "global_step": 185245, "epoch": 2081, "lr": 2.2964367940049548e-05}
{"train_loss": 0.09061266481876373, "global_step": 185246, "epoch": 2081, "lr": 2.2963880263163366e-05}
{"train_loss": 0.05736533924937248, "global_step": 185247, "epoch": 2081, "lr": 2.2963392589911843e-05}
{"train_loss": 0.11196810752153397, "global_step": 185248, "epoch": 2081, "lr": 2.296290492029508e-05}
{"train_loss": 0.0911172404885292, "global_step": 185249, "epoch": 2081, "lr": 2.2962417254313106e-05}
{"train_loss": 0.03157644346356392, "global_step": 185250, "epoch": 2081, "lr": 2.2961929591966025e-05}
{"train_loss": 0.10715632885694504, "global_step": 185251, "epoch": 2081, "lr": 2.296144193325387e-05}
{"train_loss": 0.10152048617601395, "global_step": 185252, "epoch": 2081, "lr": 2.2960954278176726e-05}
{"train_loss": 0.056744903326034546, "global_step": 185253, "epoch": 2081, "lr": 2.2960466626734645e-05}
{"train_loss": 0.08537725359201431, "global_step": 185254, "epoch": 2081, "lr": 2.295997897892771e-05}
{"train_loss": 0.0962672084569931, "global_step": 185255, "epoch": 2081, "lr": 2.295949133475596e-05}
{"train_loss": 0.10061898827552795, "global_step": 185256, "epoch": 2081, "lr": 2.29590036942195e-05}
{"train_loss": 0.03829066455364227, "global_step": 185257, "epoch": 2081, "lr": 2.2958516057318348e-05}
{"train_loss": 0.15708807110786438, "global_step": 185258, "epoch": 2081, "lr": 2.2958028424052614e-05}
{"train_loss": 0.05913623422384262, "global_step": 185259, "epoch": 2081, "lr": 2.2957540794422326e-05}
{"train_loss": 0.058780934661626816, "global_step": 185260, "epoch": 2081, "lr": 2.2957053168427584e-05}
{"train_loss": 0.07170219719409943, "global_step": 185261, "epoch": 2081, "lr": 2.295656554606842e-05}
{"train_loss": 0.08497190475463867, "global_step": 185262, "epoch": 2081, "lr": 2.295607792734493e-05}
{"train_loss": 0.07334382086992264, "global_step": 185263, "epoch": 2081, "lr": 2.295559031225715e-05}
{"train_loss": 0.11914825439453125, "global_step": 185264, "epoch": 2081, "lr": 2.2955102700805186e-05}
{"train_loss": 0.05300547182559967, "global_step": 185265, "epoch": 2081, "lr": 2.2954615092989052e-05}
{"train_loss": 0.03834948316216469, "global_step": 185266, "epoch": 2081, "lr": 2.2954127488808865e-05}
{"train_loss": 0.05411564186215401, "global_step": 185267, "epoch": 2081, "lr": 2.2953639888264644e-05}
{"train_loss": 0.12991608679294586, "global_step": 185268, "epoch": 2081, "lr": 2.2953152291356496e-05}
{"train_loss": 0.09594326466321945, "global_step": 185269, "epoch": 2081, "lr": 2.2952664698084446e-05}
{"train_loss": 0.09027465432882309, "global_step": 185270, "epoch": 2081, "lr": 2.295217710844858e-05}
{"train_loss": 0.12490464746952057, "global_step": 185271, "epoch": 2081, "lr": 2.295168952244899e-05}
{"train_loss": 0.09555447846651077, "global_step": 185272, "epoch": 2081, "lr": 2.2951201940085687e-05}
{"train_loss": 0.032678138464689255, "global_step": 185273, "epoch": 2081, "lr": 2.2950714361358792e-05}
{"train_loss": 0.05743573233485222, "global_step": 185274, "epoch": 2081, "lr": 2.295022678626832e-05}
{"train_loss": 0.05429824814200401, "global_step": 185275, "epoch": 2081, "lr": 2.294973921481438e-05}
{"train_loss": 0.04434739425778389, "global_step": 185276, "epoch": 2081, "lr": 2.2949251646997006e-05}
{"train_loss": 0.06852744519710541, "global_step": 185277, "epoch": 2081, "lr": 2.2948764082816286e-05}
{"train_loss": 0.13812869787216187, "global_step": 185278, "epoch": 2081, "lr": 2.2948276522272255e-05}
{"train_loss": 0.058103058487176895, "global_step": 185279, "epoch": 2081, "lr": 2.2947788965365023e-05}
{"train_loss": 0.044804781675338745, "global_step": 185280, "epoch": 2081, "lr": 2.2947301412094613e-05}
{"train_loss": 0.05848384648561478, "global_step": 185281, "epoch": 2081, "lr": 2.2946813862461118e-05}
{"train_loss": 0.10711264610290527, "global_step": 185282, "epoch": 2081, "lr": 2.2946326316464583e-05}
{"train_loss": 0.0494283027946949, "global_step": 185283, "epoch": 2081, "lr": 2.2945838774105105e-05}
{"train_loss": 0.07359044253826141, "global_step": 185284, "epoch": 2081, "lr": 2.294535123538273e-05}
{"train_loss": 0.08075543493032455, "global_step": 185285, "epoch": 2081, "lr": 2.2944863700297496e-05}
{"train_loss": 0.05352715775370598, "global_step": 185286, "epoch": 2081, "lr": 2.2944376168849517e-05}
{"train_loss": 0.0393156073987484, "global_step": 185287, "epoch": 2081, "lr": 2.2943888641038823e-05}
{"train_loss": 0.08407828956842422, "global_step": 185288, "epoch": 2081, "lr": 2.294340111686551e-05}
{"train_loss": 0.04556119814515114, "global_step": 185289, "epoch": 2081, "lr": 2.294291359632961e-05}
{"train_loss": 0.10608506947755814, "global_step": 185290, "epoch": 2081, "lr": 2.2942426079431216e-05}
{"train_loss": 0.053122617304325104, "global_step": 185291, "epoch": 2081, "lr": 2.2941938566170373e-05}
{"train_loss": 0.07981403917074203, "global_step": 185292, "epoch": 2081, "lr": 2.294145105654717e-05}
{"train_loss": 0.07624997198581696, "global_step": 185293, "epoch": 2081, "lr": 2.294096355056164e-05}
{"train_loss": 0.06886467337608337, "global_step": 185294, "epoch": 2081, "lr": 2.2940476048213887e-05}
{"train_loss": 0.08315527439117432, "global_step": 185295, "epoch": 2081, "lr": 2.2939988549503942e-05}
{"train_loss": 0.01965232379734516, "global_step": 185296, "epoch": 2081, "lr": 2.2939501054431905e-05}
{"train_loss": 0.0719480783249555, "global_step": 185297, "epoch": 2081, "lr": 2.2939013562997803e-05, "val_loss": 7.529236316680908}
{"train_loss": 0.049317892640829086, "global_step": 185298, "epoch": 2082, "lr": 2.293852607520173e-05}
{"train_loss": 0.06803084909915924, "global_step": 185299, "epoch": 2082, "lr": 2.2938038591043738e-05}
{"train_loss": 0.06351573020219803, "global_step": 185300, "epoch": 2082, "lr": 2.293755111052391e-05}
{"train_loss": 0.08658188581466675, "global_step": 185301, "epoch": 2082, "lr": 2.2937063633642274e-05}
{"train_loss": 0.035784851759672165, "global_step": 185302, "epoch": 2082, "lr": 2.2936576160398944e-05}
{"train_loss": 0.08286748826503754, "global_step": 185303, "epoch": 2082, "lr": 2.2936088690793945e-05}
{"train_loss": 0.05637272074818611, "global_step": 185304, "epoch": 2082, "lr": 2.2935601224827373e-05}
{"train_loss": 0.09944374859333038, "global_step": 185305, "epoch": 2082, "lr": 2.293511376249926e-05}
{"train_loss": 0.08005059510469437, "global_step": 185306, "epoch": 2082, "lr": 2.2934626303809698e-05}
{"train_loss": 0.040071096271276474, "global_step": 185307, "epoch": 2082, "lr": 2.293413884875876e-05}
{"train_loss": 0.09554547816514969, "global_step": 185308, "epoch": 2082, "lr": 2.293365139734648e-05}
{"train_loss": 0.1111283004283905, "global_step": 185309, "epoch": 2082, "lr": 2.2933163949572955e-05}
{"train_loss": 0.06472623348236084, "global_step": 185310, "epoch": 2082, "lr": 2.2932676505438216e-05}
{"train_loss": 0.10696465522050858, "global_step": 185311, "epoch": 2082, "lr": 2.2932189064942372e-05}
{"train_loss": 0.043725091964006424, "global_step": 185312, "epoch": 2082, "lr": 2.293170162808545e-05}
{"train_loss": 0.07339015603065491, "global_step": 185313, "epoch": 2082, "lr": 2.2931214194867546e-05}
{"train_loss": 0.10423989593982697, "global_step": 185314, "epoch": 2082, "lr": 2.293072676528869e-05}
{"train_loss": 0.07733780890703201, "global_step": 185315, "epoch": 2082, "lr": 2.2930239339348984e-05}
{"train_loss": 0.13456949591636658, "global_step": 185316, "epoch": 2082, "lr": 2.2929751917048465e-05}
{"train_loss": 0.09768371284008026, "global_step": 185317, "epoch": 2082, "lr": 2.2929264498387233e-05}
{"train_loss": 0.07024429738521576, "global_step": 185318, "epoch": 2082, "lr": 2.2928777083365305e-05}
{"train_loss": 0.09583057463169098, "global_step": 185319, "epoch": 2082, "lr": 2.2928289671982796e-05}
{"train_loss": 0.06492280960083008, "global_step": 185320, "epoch": 2082, "lr": 2.2927802264239724e-05}
{"train_loss": 0.06627070158720016, "global_step": 185321, "epoch": 2082, "lr": 2.29273148601362e-05}
{"train_loss": 0.09917283058166504, "global_step": 185322, "epoch": 2082, "lr": 2.292682745967225e-05}
{"train_loss": 0.09870681166648865, "global_step": 185323, "epoch": 2082, "lr": 2.2926340062847978e-05}
{"train_loss": 0.05692236125469208, "global_step": 185324, "epoch": 2082, "lr": 2.2925852669663427e-05}
{"train_loss": 0.07542819529771805, "global_step": 185325, "epoch": 2082, "lr": 2.2925365280118654e-05}
{"train_loss": 0.14316681027412415, "global_step": 185326, "epoch": 2082, "lr": 2.2924877894213748e-05}
{"train_loss": 0.11808951944112778, "global_step": 185327, "epoch": 2082, "lr": 2.2924390511948746e-05}
{"train_loss": 0.09468401968479156, "global_step": 185328, "epoch": 2082, "lr": 2.2923903133323742e-05}
{"train_loss": 0.06704919785261154, "global_step": 185329, "epoch": 2082, "lr": 2.2923415758338778e-05}
{"train_loss": 0.092990443110466, "global_step": 185330, "epoch": 2082, "lr": 2.2922928386993948e-05}
{"train_loss": 0.046533215790987015, "global_step": 185331, "epoch": 2082, "lr": 2.2922441019289283e-05}
{"train_loss": 0.04499159753322601, "global_step": 185332, "epoch": 2082, "lr": 2.292195365522488e-05}
{"train_loss": 0.12239541113376617, "global_step": 185333, "epoch": 2082, "lr": 2.2921466294800775e-05}
{"train_loss": 0.06397753208875656, "global_step": 185334, "epoch": 2082, "lr": 2.2920978938017062e-05}
{"train_loss": 0.04080777242779732, "global_step": 185335, "epoch": 2082, "lr": 2.292049158487378e-05}
{"train_loss": 0.10580205917358398, "global_step": 185336, "epoch": 2082, "lr": 2.292000423537103e-05}
{"train_loss": 0.10704939067363739, "global_step": 185337, "epoch": 2082, "lr": 2.291951688950883e-05}
{"train_loss": 0.06203121691942215, "global_step": 185338, "epoch": 2082, "lr": 2.291902954728729e-05}
{"train_loss": 0.07663993537425995, "global_step": 185339, "epoch": 2082, "lr": 2.291854220870644e-05}
{"train_loss": 0.06375160068273544, "global_step": 185340, "epoch": 2082, "lr": 2.2918054873766382e-05}
{"train_loss": 0.06861262023448944, "global_step": 185341, "epoch": 2082, "lr": 2.291756754246714e-05}
{"train_loss": 0.06108511611819267, "global_step": 185342, "epoch": 2082, "lr": 2.2917080214808802e-05}
{"train_loss": 0.06612438708543777, "global_step": 185343, "epoch": 2082, "lr": 2.291659289079145e-05}
{"train_loss": 0.038004692643880844, "global_step": 185344, "epoch": 2082, "lr": 2.2916105570415113e-05}
{"train_loss": 0.0508306585252285, "global_step": 185345, "epoch": 2082, "lr": 2.2915618253679894e-05}
{"train_loss": 0.06365539878606796, "global_step": 185346, "epoch": 2082, "lr": 2.291513094058582e-05}
{"train_loss": 0.0992961972951889, "global_step": 185347, "epoch": 2082, "lr": 2.2914643631132997e-05}
{"train_loss": 0.08422692120075226, "global_step": 185348, "epoch": 2082, "lr": 2.2914156325321456e-05}
{"train_loss": 0.05622659623622894, "global_step": 185349, "epoch": 2082, "lr": 2.291366902315129e-05}
{"train_loss": 0.06123419478535652, "global_step": 185350, "epoch": 2082, "lr": 2.2913181724622535e-05}
{"train_loss": 0.052709296345710754, "global_step": 185351, "epoch": 2082, "lr": 2.2912694429735286e-05}
{"train_loss": 0.02974003180861473, "global_step": 185352, "epoch": 2082, "lr": 2.2912207138489582e-05}
{"train_loss": 0.05472757667303085, "global_step": 185353, "epoch": 2082, "lr": 2.2911719850885522e-05}
{"train_loss": 0.08901330828666687, "global_step": 185354, "epoch": 2082, "lr": 2.291123256692313e-05}
{"train_loss": 0.07141172885894775, "global_step": 185355, "epoch": 2082, "lr": 2.291074528660251e-05}
{"train_loss": 0.05624515935778618, "global_step": 185356, "epoch": 2082, "lr": 2.291025800992369e-05}
{"train_loss": 0.07059033215045929, "global_step": 185357, "epoch": 2082, "lr": 2.2909770736886775e-05}
{"train_loss": 0.11354292929172516, "global_step": 185358, "epoch": 2082, "lr": 2.2909283467491793e-05}
{"train_loss": 0.10457351058721542, "global_step": 185359, "epoch": 2082, "lr": 2.2908796201738846e-05}
{"train_loss": 0.057661473751068115, "global_step": 185360, "epoch": 2082, "lr": 2.2908308939627964e-05}
{"train_loss": 0.07554976642131805, "global_step": 185361, "epoch": 2082, "lr": 2.2907821681159246e-05}
{"train_loss": 0.05336261913180351, "global_step": 185362, "epoch": 2082, "lr": 2.290733442633275e-05}
{"train_loss": 0.10364159941673279, "global_step": 185363, "epoch": 2082, "lr": 2.2906847175148505e-05}
{"train_loss": 0.04392329603433609, "global_step": 185364, "epoch": 2082, "lr": 2.2906359927606624e-05}
{"train_loss": 0.06325631588697433, "global_step": 185365, "epoch": 2082, "lr": 2.2905872683707137e-05}
{"train_loss": 0.05838710069656372, "global_step": 185366, "epoch": 2082, "lr": 2.290538544345014e-05}
{"train_loss": 0.10133258998394012, "global_step": 185367, "epoch": 2082, "lr": 2.2904898206835667e-05}
{"train_loss": 0.09783788025379181, "global_step": 185368, "epoch": 2082, "lr": 2.2904410973863816e-05}
{"train_loss": 0.05314456298947334, "global_step": 185369, "epoch": 2082, "lr": 2.290392374453462e-05}
{"train_loss": 0.05262408033013344, "global_step": 185370, "epoch": 2082, "lr": 2.2903436518848182e-05}
{"train_loss": 0.13090312480926514, "global_step": 185371, "epoch": 2082, "lr": 2.2902949296804527e-05}
{"train_loss": 0.08281850814819336, "global_step": 185372, "epoch": 2082, "lr": 2.2902462078403757e-05}
{"train_loss": 0.05567016080021858, "global_step": 185373, "epoch": 2082, "lr": 2.2901974863645904e-05}
{"train_loss": 0.10275658220052719, "global_step": 185374, "epoch": 2082, "lr": 2.2901487652531067e-05}
{"train_loss": 0.1554260104894638, "global_step": 185375, "epoch": 2082, "lr": 2.290100044505928e-05}
{"train_loss": 0.030895570293068886, "global_step": 185376, "epoch": 2082, "lr": 2.2900513241230637e-05}
{"train_loss": 0.07349635660648346, "global_step": 185377, "epoch": 2082, "lr": 2.2900026041045176e-05}
{"train_loss": 0.03985482081770897, "global_step": 185378, "epoch": 2082, "lr": 2.2899538844502972e-05}
{"train_loss": 0.04913811758160591, "global_step": 185379, "epoch": 2082, "lr": 2.2899051651604115e-05}
{"train_loss": 0.05797894299030304, "global_step": 185380, "epoch": 2082, "lr": 2.289856446234863e-05}
{"train_loss": 0.04687846824526787, "global_step": 185381, "epoch": 2082, "lr": 2.2898077276736625e-05}
{"train_loss": 0.07015284150838852, "global_step": 185382, "epoch": 2082, "lr": 2.2897590094768118e-05}
{"train_loss": 0.08745615184307098, "global_step": 185383, "epoch": 2082, "lr": 2.2897102916443225e-05}
{"train_loss": 0.04312993586063385, "global_step": 185384, "epoch": 2082, "lr": 2.2896615741761963e-05}
{"train_loss": 0.10011018812656403, "global_step": 185385, "epoch": 2082, "lr": 2.2896128570724446e-05}
{"train_loss": 0.07477688655424654, "global_step": 185386, "epoch": 2082, "lr": 2.2895641403330686e-05, "val_loss": 7.702111721038818}
{"train_loss": 0.07756483554840088, "global_step": 185387, "epoch": 2083, "lr": 2.28951542395808e-05}
{"train_loss": 0.05949794873595238, "global_step": 185388, "epoch": 2083, "lr": 2.2894667079474814e-05}
{"train_loss": 0.09995071589946747, "global_step": 185389, "epoch": 2083, "lr": 2.2894179923012827e-05}
{"train_loss": 0.08470027148723602, "global_step": 185390, "epoch": 2083, "lr": 2.2893692770194864e-05}
{"train_loss": 0.07470373809337616, "global_step": 185391, "epoch": 2083, "lr": 2.2893205621021036e-05}
{"train_loss": 0.06142698600888252, "global_step": 185392, "epoch": 2083, "lr": 2.2892718475491372e-05}
{"train_loss": 0.1277582049369812, "global_step": 185393, "epoch": 2083, "lr": 2.289223133360596e-05}
{"train_loss": 0.02797636203467846, "global_step": 185394, "epoch": 2083, "lr": 2.2891744195364846e-05}
{"train_loss": 0.0733719989657402, "global_step": 185395, "epoch": 2083, "lr": 2.2891257060768124e-05}
{"train_loss": 0.11972489953041077, "global_step": 185396, "epoch": 2083, "lr": 2.2890769929815818e-05}
{"train_loss": 0.13082581758499146, "global_step": 185397, "epoch": 2083, "lr": 2.2890282802508045e-05}
{"train_loss": 0.06634402275085449, "global_step": 185398, "epoch": 2083, "lr": 2.2889795678844817e-05}
{"train_loss": 0.06856440007686615, "global_step": 185399, "epoch": 2083, "lr": 2.288930855882625e-05}
{"train_loss": 0.07922529429197311, "global_step": 185400, "epoch": 2083, "lr": 2.288882144245238e-05}
{"train_loss": 0.05511632561683655, "global_step": 185401, "epoch": 2083, "lr": 2.2888334329723264e-05}
{"train_loss": 0.0716480240225792, "global_step": 185402, "epoch": 2083, "lr": 2.2887847220638998e-05}
{"train_loss": 0.1394725888967514, "global_step": 185403, "epoch": 2083, "lr": 2.288736011519961e-05}
{"train_loss": 0.1078023836016655, "global_step": 185404, "epoch": 2083, "lr": 2.2886873013405203e-05}
{"train_loss": 0.09894762188196182, "global_step": 185405, "epoch": 2083, "lr": 2.288638591525581e-05}
{"train_loss": 0.04935974255204201, "global_step": 185406, "epoch": 2083, "lr": 2.288589882075153e-05}
{"train_loss": 0.047633469104766846, "global_step": 185407, "epoch": 2083, "lr": 2.2885411729892393e-05}
{"train_loss": 0.0635477751493454, "global_step": 185408, "epoch": 2083, "lr": 2.2884924642678497e-05}
{"train_loss": 0.05827496945858002, "global_step": 185409, "epoch": 2083, "lr": 2.288443755910988e-05}
{"train_loss": 0.07447022944688797, "global_step": 185410, "epoch": 2083, "lr": 2.2883950479186633e-05}
{"train_loss": 0.08397272974252701, "global_step": 185411, "epoch": 2083, "lr": 2.288346340290879e-05}
{"train_loss": 0.10803373157978058, "global_step": 185412, "epoch": 2083, "lr": 2.288297633027645e-05}
{"train_loss": 0.04780269414186478, "global_step": 185413, "epoch": 2083, "lr": 2.2882489261289654e-05}
{"train_loss": 0.055285461246967316, "global_step": 185414, "epoch": 2083, "lr": 2.2882002195948486e-05}
{"train_loss": 0.08627955615520477, "global_step": 185415, "epoch": 2083, "lr": 2.288151513425299e-05}
{"train_loss": 0.05000949651002884, "global_step": 185416, "epoch": 2083, "lr": 2.288102807620324e-05}
{"train_loss": 0.04203074425458908, "global_step": 185417, "epoch": 2083, "lr": 2.288054102179933e-05}
{"train_loss": 0.08260635286569595, "global_step": 185418, "epoch": 2083, "lr": 2.2880053971041277e-05}
{"train_loss": 0.05781026929616928, "global_step": 185419, "epoch": 2083, "lr": 2.287956692392919e-05}
{"train_loss": 0.08465749025344849, "global_step": 185420, "epoch": 2083, "lr": 2.2879079880463093e-05}
{"train_loss": 0.10213730484247208, "global_step": 185421, "epoch": 2083, "lr": 2.28785928406431e-05}
{"train_loss": 0.06863719969987869, "global_step": 185422, "epoch": 2083, "lr": 2.2878105804469224e-05}
{"train_loss": 0.10889574885368347, "global_step": 185423, "epoch": 2083, "lr": 2.2877618771941578e-05}
{"train_loss": 0.08304654061794281, "global_step": 185424, "epoch": 2083, "lr": 2.287713174306018e-05}
{"train_loss": 0.05590431019663811, "global_step": 185425, "epoch": 2083, "lr": 2.2876644717825153e-05}
{"train_loss": 0.05653147026896477, "global_step": 185426, "epoch": 2083, "lr": 2.28761576962365e-05}
{"train_loss": 0.05711480975151062, "global_step": 185427, "epoch": 2083, "lr": 2.287567067829434e-05}
{"train_loss": 0.0944475457072258, "global_step": 185428, "epoch": 2083, "lr": 2.2875183663998696e-05}
{"train_loss": 0.08076704293489456, "global_step": 185429, "epoch": 2083, "lr": 2.287469665334967e-05}
{"train_loss": 0.02791641093790531, "global_step": 185430, "epoch": 2083, "lr": 2.2874209646347294e-05}
{"train_loss": 0.04874029755592346, "global_step": 185431, "epoch": 2083, "lr": 2.287372264299167e-05}
{"train_loss": 0.09086829423904419, "global_step": 185432, "epoch": 2083, "lr": 2.2873235643282826e-05}
{"train_loss": 0.08836332708597183, "global_step": 185433, "epoch": 2083, "lr": 2.2872748647220836e-05}
{"train_loss": 0.039286982268095016, "global_step": 185434, "epoch": 2083, "lr": 2.2872261654805804e-05}
{"train_loss": 0.10815497487783432, "global_step": 185435, "epoch": 2083, "lr": 2.287177466603776e-05}
{"train_loss": 0.08679564297199249, "global_step": 185436, "epoch": 2083, "lr": 2.287128768091676e-05}
{"train_loss": 0.1285947859287262, "global_step": 185437, "epoch": 2083, "lr": 2.2870800699442895e-05}
{"train_loss": 0.08395448327064514, "global_step": 185438, "epoch": 2083, "lr": 2.287031372161621e-05}
{"train_loss": 0.03524632006883621, "global_step": 185439, "epoch": 2083, "lr": 2.2869826747436794e-05}
{"train_loss": 0.08331811428070068, "global_step": 185440, "epoch": 2083, "lr": 2.2869339776904702e-05}
{"train_loss": 0.1179874837398529, "global_step": 185441, "epoch": 2083, "lr": 2.2868852810019982e-05}
{"train_loss": 0.07420212775468826, "global_step": 185442, "epoch": 2083, "lr": 2.2868365846782724e-05}
{"train_loss": 0.09663267433643341, "global_step": 185443, "epoch": 2083, "lr": 2.2867878887192973e-05}
{"train_loss": 0.05940292403101921, "global_step": 185444, "epoch": 2083, "lr": 2.2867391931250814e-05}
{"train_loss": 0.026741711422801018, "global_step": 185445, "epoch": 2083, "lr": 2.2866904978956293e-05}
{"train_loss": 0.05536062270402908, "global_step": 185446, "epoch": 2083, "lr": 2.2866418030309506e-05}
{"train_loss": 0.09229905158281326, "global_step": 185447, "epoch": 2083, "lr": 2.286593108531047e-05}
{"train_loss": 0.049881599843502045, "global_step": 185448, "epoch": 2083, "lr": 2.2865444143959307e-05}
{"train_loss": 0.07101308554410934, "global_step": 185449, "epoch": 2083, "lr": 2.2864957206256034e-05}
{"train_loss": 0.08758450299501419, "global_step": 185450, "epoch": 2083, "lr": 2.286447027220075e-05}
{"train_loss": 0.10991369187831879, "global_step": 185451, "epoch": 2083, "lr": 2.2863983341793495e-05}
{"train_loss": 0.03802479803562164, "global_step": 185452, "epoch": 2083, "lr": 2.286349641503434e-05}
{"train_loss": 0.043366435915231705, "global_step": 185453, "epoch": 2083, "lr": 2.2863009491923382e-05}
{"train_loss": 0.04314431548118591, "global_step": 185454, "epoch": 2083, "lr": 2.2862522572460638e-05}
{"train_loss": 0.058036308735609055, "global_step": 185455, "epoch": 2083, "lr": 2.286203565664622e-05}
{"train_loss": 0.07593461871147156, "global_step": 185456, "epoch": 2083, "lr": 2.286154874448015e-05}
{"train_loss": 0.07486879825592041, "global_step": 185457, "epoch": 2083, "lr": 2.2861061835962532e-05}
{"train_loss": 0.08279774338006973, "global_step": 185458, "epoch": 2083, "lr": 2.28605749310934e-05}
{"train_loss": 0.1004120334982872, "global_step": 185459, "epoch": 2083, "lr": 2.2860088029872845e-05}
{"train_loss": 0.05280633270740509, "global_step": 185460, "epoch": 2083, "lr": 2.28596011323009e-05}
{"train_loss": 0.10362625867128372, "global_step": 185461, "epoch": 2083, "lr": 2.2859114238377678e-05}
{"train_loss": 0.09153902530670166, "global_step": 185462, "epoch": 2083, "lr": 2.285862734810319e-05}
{"train_loss": 0.05158337578177452, "global_step": 185463, "epoch": 2083, "lr": 2.285814046147755e-05}
{"train_loss": 0.06969612836837769, "global_step": 185464, "epoch": 2083, "lr": 2.2857653578500787e-05}
{"train_loss": 0.11918094754219055, "global_step": 185465, "epoch": 2083, "lr": 2.2857166699172998e-05}
{"train_loss": 0.06742183119058609, "global_step": 185466, "epoch": 2083, "lr": 2.2856679823494213e-05}
{"train_loss": 0.03637578338384628, "global_step": 185467, "epoch": 2083, "lr": 2.285619295146454e-05}
{"train_loss": 0.08641587942838669, "global_step": 185468, "epoch": 2083, "lr": 2.2855706083083993e-05}
{"train_loss": 0.10358153283596039, "global_step": 185469, "epoch": 2083, "lr": 2.2855219218352676e-05}
{"train_loss": 0.039479248225688934, "global_step": 185470, "epoch": 2083, "lr": 2.2854732357270658e-05}
{"train_loss": 0.03759998828172684, "global_step": 185471, "epoch": 2083, "lr": 2.2854245499837966e-05}
{"train_loss": 0.0530734620988369, "global_step": 185472, "epoch": 2083, "lr": 2.285375864605472e-05}
{"train_loss": 0.04215952008962631, "global_step": 185473, "epoch": 2083, "lr": 2.2853271795920944e-05}
{"train_loss": 0.08807218819856644, "global_step": 185474, "epoch": 2083, "lr": 2.2852784949436702e-05}
{"train_loss": 0.07442069474314705, "global_step": 185475, "epoch": 2083, "lr": 2.2852298106602087e-05, "val_loss": 7.497012138366699}
{"train_loss": 0.08946351706981659, "global_step": 185476, "epoch": 2084, "lr": 2.2851811267417138e-05}
{"train_loss": 0.05738753825426102, "global_step": 185477, "epoch": 2084, "lr": 2.2851324431881942e-05}
{"train_loss": 0.11391236633062363, "global_step": 185478, "epoch": 2084, "lr": 2.285083759999656e-05}
{"train_loss": 0.1298232525587082, "global_step": 185479, "epoch": 2084, "lr": 2.2850350771761036e-05}
{"train_loss": 0.05564231425523758, "global_step": 185480, "epoch": 2084, "lr": 2.2849863947175466e-05}
{"train_loss": 0.10371147096157074, "global_step": 185481, "epoch": 2084, "lr": 2.2849377126239885e-05}
{"train_loss": 0.14775687456130981, "global_step": 185482, "epoch": 2084, "lr": 2.284889030895439e-05}
{"train_loss": 0.10889096558094025, "global_step": 185483, "epoch": 2084, "lr": 2.284840349531901e-05}
{"train_loss": 0.05538228526711464, "global_step": 185484, "epoch": 2084, "lr": 2.2847916685333852e-05}
{"train_loss": 0.07704009115695953, "global_step": 185485, "epoch": 2084, "lr": 2.2847429878998944e-05}
{"train_loss": 0.06917068362236023, "global_step": 185486, "epoch": 2084, "lr": 2.2846943076314386e-05}
{"train_loss": 0.03525957092642784, "global_step": 185487, "epoch": 2084, "lr": 2.2846456277280205e-05}
{"train_loss": 0.07614652812480927, "global_step": 185488, "epoch": 2084, "lr": 2.284596948189649e-05}
{"train_loss": 0.13266786932945251, "global_step": 185489, "epoch": 2084, "lr": 2.2845482690163317e-05}
{"train_loss": 0.07520488649606705, "global_step": 185490, "epoch": 2084, "lr": 2.2844995902080723e-05}
{"train_loss": 0.05377385392785072, "global_step": 185491, "epoch": 2084, "lr": 2.284450911764881e-05}
{"train_loss": 0.08565816283226013, "global_step": 185492, "epoch": 2084, "lr": 2.28440223368676e-05}
{"train_loss": 0.09050649404525757, "global_step": 185493, "epoch": 2084, "lr": 2.2843535559737195e-05}
{"train_loss": 0.05545133724808693, "global_step": 185494, "epoch": 2084, "lr": 2.284304878625763e-05}
{"train_loss": 0.05506254360079765, "global_step": 185495, "epoch": 2084, "lr": 2.284256201642901e-05}
{"train_loss": 0.10409257560968399, "global_step": 185496, "epoch": 2084, "lr": 2.284207525025135e-05}
{"train_loss": 0.056726519018411636, "global_step": 185497, "epoch": 2084, "lr": 2.2841588487724765e-05}
{"train_loss": 0.07009971141815186, "global_step": 185498, "epoch": 2084, "lr": 2.284110172884928e-05}
{"train_loss": 0.1053786352276802, "global_step": 185499, "epoch": 2084, "lr": 2.284061497362499e-05}
{"train_loss": 0.09588728100061417, "global_step": 185500, "epoch": 2084, "lr": 2.284012822205194e-05}
{"train_loss": 0.0859103798866272, "global_step": 185501, "epoch": 2084, "lr": 2.2839641474130212e-05}
{"train_loss": 0.09042789041996002, "global_step": 185502, "epoch": 2084, "lr": 2.2839154729859848e-05}
{"train_loss": 0.07319460064172745, "global_step": 185503, "epoch": 2084, "lr": 2.2838667989240948e-05}
{"train_loss": 0.07247860729694366, "global_step": 185504, "epoch": 2084, "lr": 2.283818125227354e-05}
{"train_loss": 0.0930994376540184, "global_step": 185505, "epoch": 2084, "lr": 2.2837694518957707e-05}
{"train_loss": 0.04951445758342743, "global_step": 185506, "epoch": 2084, "lr": 2.2837207789293535e-05}
{"train_loss": 0.07092341780662537, "global_step": 185507, "epoch": 2084, "lr": 2.2836721063281047e-05}
{"train_loss": 0.054630097001791, "global_step": 185508, "epoch": 2084, "lr": 2.2836234340920353e-05}
{"train_loss": 0.06194857507944107, "global_step": 185509, "epoch": 2084, "lr": 2.2835747622211474e-05}
{"train_loss": 0.07369497418403625, "global_step": 185510, "epoch": 2084, "lr": 2.283526090715452e-05}
{"train_loss": 0.1078755110502243, "global_step": 185511, "epoch": 2084, "lr": 2.283477419574953e-05}
{"train_loss": 0.08794275671243668, "global_step": 185512, "epoch": 2084, "lr": 2.283428748799656e-05}
{"train_loss": 0.12303580343723297, "global_step": 185513, "epoch": 2084, "lr": 2.2833800783895702e-05}
{"train_loss": 0.04694894701242447, "global_step": 185514, "epoch": 2084, "lr": 2.2833314083446994e-05}
{"train_loss": 0.060946810990571976, "global_step": 185515, "epoch": 2084, "lr": 2.2832827386650533e-05}
{"train_loss": 0.08419219404459, "global_step": 185516, "epoch": 2084, "lr": 2.2832340693506366e-05}
{"train_loss": 0.065996915102005, "global_step": 185517, "epoch": 2084, "lr": 2.283185400401454e-05}
{"train_loss": 0.0989592894911766, "global_step": 185518, "epoch": 2084, "lr": 2.2831367318175166e-05}
{"train_loss": 0.11709889769554138, "global_step": 185519, "epoch": 2084, "lr": 2.283088063598826e-05}
{"train_loss": 0.059131015092134476, "global_step": 185520, "epoch": 2084, "lr": 2.2830393957453926e-05}
{"train_loss": 0.0762290507555008, "global_step": 185521, "epoch": 2084, "lr": 2.28299072825722e-05}
{"train_loss": 0.08194319903850555, "global_step": 185522, "epoch": 2084, "lr": 2.2829420611343184e-05}
{"train_loss": 0.06999120116233826, "global_step": 185523, "epoch": 2084, "lr": 2.2828933943766896e-05}
{"train_loss": 0.10862690210342407, "global_step": 185524, "epoch": 2084, "lr": 2.2828447279843428e-05}
{"train_loss": 0.09260142594575882, "global_step": 185525, "epoch": 2084, "lr": 2.2827960619572864e-05}
{"train_loss": 0.049406807869672775, "global_step": 185526, "epoch": 2084, "lr": 2.282747396295523e-05}
{"train_loss": 0.06331516057252884, "global_step": 185527, "epoch": 2084, "lr": 2.2826987309990634e-05}
{"train_loss": 0.10597667098045349, "global_step": 185528, "epoch": 2084, "lr": 2.2826500660679094e-05}
{"train_loss": 0.06336670368909836, "global_step": 185529, "epoch": 2084, "lr": 2.282601401502072e-05}
{"train_loss": 0.04900739714503288, "global_step": 185530, "epoch": 2084, "lr": 2.2825527373015537e-05}
{"train_loss": 0.05597253143787384, "global_step": 185531, "epoch": 2084, "lr": 2.282504073466365e-05}
{"train_loss": 0.0790923535823822, "global_step": 185532, "epoch": 2084, "lr": 2.282455409996509e-05}
{"train_loss": 0.08425025641918182, "global_step": 185533, "epoch": 2084, "lr": 2.282406746891995e-05}
{"train_loss": 0.0768389031291008, "global_step": 185534, "epoch": 2084, "lr": 2.2823580841528268e-05}
{"train_loss": 0.08440820127725601, "global_step": 185535, "epoch": 2084, "lr": 2.2823094217790147e-05}
{"train_loss": 0.11080325394868851, "global_step": 185536, "epoch": 2084, "lr": 2.2822607597705607e-05}
{"train_loss": 0.08750172704458237, "global_step": 185537, "epoch": 2084, "lr": 2.282212098127475e-05}
{"train_loss": 0.12615878880023956, "global_step": 185538, "epoch": 2084, "lr": 2.282163436849762e-05}
{"train_loss": 0.04651283845305443, "global_step": 185539, "epoch": 2084, "lr": 2.28211477593743e-05}
{"train_loss": 0.0660446286201477, "global_step": 185540, "epoch": 2084, "lr": 2.2820661153904833e-05}
{"train_loss": 0.061449818313121796, "global_step": 185541, "epoch": 2084, "lr": 2.2820174552089307e-05}
{"train_loss": 0.1338607370853424, "global_step": 185542, "epoch": 2084, "lr": 2.2819687953927765e-05}
{"train_loss": 0.06709064543247223, "global_step": 185543, "epoch": 2084, "lr": 2.2819201359420284e-05}
{"train_loss": 0.059815891087055206, "global_step": 185544, "epoch": 2084, "lr": 2.2818714768566946e-05}
{"train_loss": 0.11154359579086304, "global_step": 185545, "epoch": 2084, "lr": 2.2818228181367783e-05}
{"train_loss": 0.10961192101240158, "global_step": 185546, "epoch": 2084, "lr": 2.28177415978229e-05}
{"train_loss": 0.08765274286270142, "global_step": 185547, "epoch": 2084, "lr": 2.2817255017932314e-05}
{"train_loss": 0.07666950672864914, "global_step": 185548, "epoch": 2084, "lr": 2.281676844169614e-05}
{"train_loss": 0.020677585154771805, "global_step": 185549, "epoch": 2084, "lr": 2.2816281869114403e-05}
{"train_loss": 0.0726458728313446, "global_step": 185550, "epoch": 2084, "lr": 2.2815795300187205e-05}
{"train_loss": 0.05576806887984276, "global_step": 185551, "epoch": 2084, "lr": 2.2815308734914588e-05}
{"train_loss": 0.062125954777002335, "global_step": 185552, "epoch": 2084, "lr": 2.2814822173296606e-05}
{"train_loss": 0.05578888952732086, "global_step": 185553, "epoch": 2084, "lr": 2.2814335615333354e-05}
{"train_loss": 0.1434781551361084, "global_step": 185554, "epoch": 2084, "lr": 2.2813849061024874e-05}
{"train_loss": 0.05697563290596008, "global_step": 185555, "epoch": 2084, "lr": 2.281336251037125e-05}
{"train_loss": 0.06381476670503616, "global_step": 185556, "epoch": 2084, "lr": 2.2812875963372544e-05}
{"train_loss": 0.09268578886985779, "global_step": 185557, "epoch": 2084, "lr": 2.281238942002879e-05}
{"train_loss": 0.09920969605445862, "global_step": 185558, "epoch": 2084, "lr": 2.2811902880340102e-05}
{"train_loss": 0.0768318921327591, "global_step": 185559, "epoch": 2084, "lr": 2.2811416344306503e-05}
{"train_loss": 0.09219972789287567, "global_step": 185560, "epoch": 2084, "lr": 2.2810929811928078e-05}
{"train_loss": 0.05903714895248413, "global_step": 185561, "epoch": 2084, "lr": 2.2810443283204912e-05}
{"train_loss": 0.05582135170698166, "global_step": 185562, "epoch": 2084, "lr": 2.280995675813703e-05}
{"train_loss": 0.051894836127758026, "global_step": 185563, "epoch": 2084, "lr": 2.2809470236724533e-05}
{"train_loss": 0.07989911898300889, "global_step": 185564, "epoch": 2084, "lr": 2.280898371896746e-05, "val_loss": 7.650380611419678}
{"train_loss": 0.07576452940702438, "global_step": 185565, "epoch": 2085, "lr": 2.28084972048659e-05}
{"train_loss": 0.07870887219905853, "global_step": 185566, "epoch": 2085, "lr": 2.280801069441989e-05}
{"train_loss": 0.05954065918922424, "global_step": 185567, "epoch": 2085, "lr": 2.280752418762953e-05}
{"train_loss": 0.06805211305618286, "global_step": 185568, "epoch": 2085, "lr": 2.2807037684494855e-05}
{"train_loss": 0.1157180592417717, "global_step": 185569, "epoch": 2085, "lr": 2.2806551185015952e-05}
{"train_loss": 0.07426486909389496, "global_step": 185570, "epoch": 2085, "lr": 2.2806064689192864e-05}
{"train_loss": 0.12873105704784393, "global_step": 185571, "epoch": 2085, "lr": 2.2805578197025685e-05}
{"train_loss": 0.07772763818502426, "global_step": 185572, "epoch": 2085, "lr": 2.2805091708514442e-05}
{"train_loss": 0.12220214307308197, "global_step": 185573, "epoch": 2085, "lr": 2.280460522365925e-05}
{"train_loss": 0.037406668066978455, "global_step": 185574, "epoch": 2085, "lr": 2.2804118742460116e-05}
{"train_loss": 0.10403545200824738, "global_step": 185575, "epoch": 2085, "lr": 2.2803632264917168e-05}
{"train_loss": 0.08895523101091385, "global_step": 185576, "epoch": 2085, "lr": 2.2803145791030417e-05}
{"train_loss": 0.07290532439947128, "global_step": 185577, "epoch": 2085, "lr": 2.280265932079997e-05}
{"train_loss": 0.05877692997455597, "global_step": 185578, "epoch": 2085, "lr": 2.2802172854225856e-05}
{"train_loss": 0.05656340718269348, "global_step": 185579, "epoch": 2085, "lr": 2.2801686391308158e-05}
{"train_loss": 0.059437960386276245, "global_step": 185580, "epoch": 2085, "lr": 2.2801199932046956e-05}
{"train_loss": 0.057446934282779694, "global_step": 185581, "epoch": 2085, "lr": 2.2800713476442293e-05}
{"train_loss": 0.10251312702894211, "global_step": 185582, "epoch": 2085, "lr": 2.2800227024494254e-05}
{"train_loss": 0.07788309454917908, "global_step": 185583, "epoch": 2085, "lr": 2.279974057620287e-05}
{"train_loss": 0.062054190784692764, "global_step": 185584, "epoch": 2085, "lr": 2.2799254131568253e-05}
{"train_loss": 0.13331352174282074, "global_step": 185585, "epoch": 2085, "lr": 2.2798767690590426e-05}
{"train_loss": 0.10107208788394928, "global_step": 185586, "epoch": 2085, "lr": 2.27982812532695e-05}
{"train_loss": 0.0818251520395279, "global_step": 185587, "epoch": 2085, "lr": 2.2797794819605482e-05}
{"train_loss": 0.05708261951804161, "global_step": 185588, "epoch": 2085, "lr": 2.2797308389598494e-05}
{"train_loss": 0.0810946524143219, "global_step": 185589, "epoch": 2085, "lr": 2.2796821963248577e-05}
{"train_loss": 0.023047516122460365, "global_step": 185590, "epoch": 2085, "lr": 2.2796335540555776e-05}
{"train_loss": 0.05041033402085304, "global_step": 185591, "epoch": 2085, "lr": 2.2795849121520192e-05}
{"train_loss": 0.1204468384385109, "global_step": 185592, "epoch": 2085, "lr": 2.279536270614186e-05}
{"train_loss": 0.06537869572639465, "global_step": 185593, "epoch": 2085, "lr": 2.2794876294420875e-05}
{"train_loss": 0.07411341369152069, "global_step": 185594, "epoch": 2085, "lr": 2.279438988635729e-05}
{"train_loss": 0.06582795083522797, "global_step": 185595, "epoch": 2085, "lr": 2.2793903481951146e-05}
{"train_loss": 0.0612800195813179, "global_step": 185596, "epoch": 2085, "lr": 2.279341708120253e-05}
{"train_loss": 0.05293550714850426, "global_step": 185597, "epoch": 2085, "lr": 2.279293068411153e-05}
{"train_loss": 0.0656757727265358, "global_step": 185598, "epoch": 2085, "lr": 2.2792444290678166e-05}
{"train_loss": 0.09589855372905731, "global_step": 185599, "epoch": 2085, "lr": 2.279195790090255e-05}
{"train_loss": 0.1045917421579361, "global_step": 185600, "epoch": 2085, "lr": 2.27914715147847e-05}
{"train_loss": 0.057491205632686615, "global_step": 185601, "epoch": 2085, "lr": 2.279098513232473e-05}
{"train_loss": 0.13101708889007568, "global_step": 185602, "epoch": 2085, "lr": 2.2790498753522654e-05}
{"train_loss": 0.08573673665523529, "global_step": 185603, "epoch": 2085, "lr": 2.2790012378378582e-05}
{"train_loss": 0.16515348851680756, "global_step": 185604, "epoch": 2085, "lr": 2.2789526006892548e-05}
{"train_loss": 0.09371571987867355, "global_step": 185605, "epoch": 2085, "lr": 2.2789039639064646e-05}
{"train_loss": 0.10293319076299667, "global_step": 185606, "epoch": 2085, "lr": 2.278855327489491e-05}
{"train_loss": 0.06487647444009781, "global_step": 185607, "epoch": 2085, "lr": 2.2788066914383437e-05}
{"train_loss": 0.1395154893398285, "global_step": 185608, "epoch": 2085, "lr": 2.2787580557530263e-05}
{"train_loss": 0.09207673370838165, "global_step": 185609, "epoch": 2085, "lr": 2.2787094204335485e-05}
{"train_loss": 0.03881317377090454, "global_step": 185610, "epoch": 2085, "lr": 2.2786607854799123e-05}
{"train_loss": 0.04513035714626312, "global_step": 185611, "epoch": 2085, "lr": 2.2786121508921297e-05}
{"train_loss": 0.05994929000735283, "global_step": 185612, "epoch": 2085, "lr": 2.278563516670203e-05}
{"train_loss": 0.08898620307445526, "global_step": 185613, "epoch": 2085, "lr": 2.2785148828141413e-05}
{"train_loss": 0.06784478574991226, "global_step": 185614, "epoch": 2085, "lr": 2.2784662493239484e-05}
{"train_loss": 0.05036327987909317, "global_step": 185615, "epoch": 2085, "lr": 2.2784176161996327e-05}
{"train_loss": 0.05472269654273987, "global_step": 185616, "epoch": 2085, "lr": 2.2783689834412024e-05}
{"train_loss": 0.12190436571836472, "global_step": 185617, "epoch": 2085, "lr": 2.2783203510486605e-05}
{"train_loss": 0.10853517800569534, "global_step": 185618, "epoch": 2085, "lr": 2.2782717190220172e-05}
{"train_loss": 0.03224661946296692, "global_step": 185619, "epoch": 2085, "lr": 2.278223087361275e-05}
{"train_loss": 0.06948141008615494, "global_step": 185620, "epoch": 2085, "lr": 2.2781744560664443e-05}
{"train_loss": 0.07167068868875504, "global_step": 185621, "epoch": 2085, "lr": 2.2781258251375282e-05}
{"train_loss": 0.09066478908061981, "global_step": 185622, "epoch": 2085, "lr": 2.2780771945745365e-05}
{"train_loss": 0.08179118484258652, "global_step": 185623, "epoch": 2085, "lr": 2.278028564377473e-05}
{"train_loss": 0.08565641194581985, "global_step": 185624, "epoch": 2085, "lr": 2.2779799345463464e-05}
{"train_loss": 0.03557513281702995, "global_step": 185625, "epoch": 2085, "lr": 2.277931305081161e-05}
{"train_loss": 0.08918496966362, "global_step": 185626, "epoch": 2085, "lr": 2.277882675981926e-05}
{"train_loss": 0.10953927040100098, "global_step": 185627, "epoch": 2085, "lr": 2.2778340472486465e-05}
{"train_loss": 0.1171659380197525, "global_step": 185628, "epoch": 2085, "lr": 2.2777854188813275e-05}
{"train_loss": 0.059278447180986404, "global_step": 185629, "epoch": 2085, "lr": 2.2777367908799785e-05}
{"train_loss": 0.05292677506804466, "global_step": 185630, "epoch": 2085, "lr": 2.277688163244603e-05}
{"train_loss": 0.09163536876440048, "global_step": 185631, "epoch": 2085, "lr": 2.2776395359752112e-05}
{"train_loss": 0.09285075217485428, "global_step": 185632, "epoch": 2085, "lr": 2.277590909071805e-05}
{"train_loss": 0.07924982905387878, "global_step": 185633, "epoch": 2085, "lr": 2.2775422825343963e-05}
{"train_loss": 0.054484348744153976, "global_step": 185634, "epoch": 2085, "lr": 2.277493656362986e-05}
{"train_loss": 0.05036094784736633, "global_step": 185635, "epoch": 2085, "lr": 2.277445030557586e-05}
{"train_loss": 0.06381143629550934, "global_step": 185636, "epoch": 2085, "lr": 2.277396405118199e-05}
{"train_loss": 0.09111806750297546, "global_step": 185637, "epoch": 2085, "lr": 2.277347780044834e-05}
{"train_loss": 0.04526585713028908, "global_step": 185638, "epoch": 2085, "lr": 2.2772991553374945e-05}
{"train_loss": 0.06946218013763428, "global_step": 185639, "epoch": 2085, "lr": 2.277250530996191e-05}
{"train_loss": 0.07552560418844223, "global_step": 185640, "epoch": 2085, "lr": 2.277201907020926e-05}
{"train_loss": 0.061526041477918625, "global_step": 185641, "epoch": 2085, "lr": 2.2771532834117106e-05}
{"train_loss": 0.07974924892187119, "global_step": 185642, "epoch": 2085, "lr": 2.2771046601685458e-05}
{"train_loss": 0.06360696256160736, "global_step": 185643, "epoch": 2085, "lr": 2.2770560372914433e-05}
{"train_loss": 0.1292443722486496, "global_step": 185644, "epoch": 2085, "lr": 2.277007414780406e-05}
{"train_loss": 0.05212757736444473, "global_step": 185645, "epoch": 2085, "lr": 2.2769587926354436e-05}
{"train_loss": 0.09376832097768784, "global_step": 185646, "epoch": 2085, "lr": 2.276910170856559e-05}
{"train_loss": 0.03685282543301582, "global_step": 185647, "epoch": 2085, "lr": 2.2768615494437616e-05}
{"train_loss": 0.05909736454486847, "global_step": 185648, "epoch": 2085, "lr": 2.2768129283970563e-05}
{"train_loss": 0.052878379821777344, "global_step": 185649, "epoch": 2085, "lr": 2.2767643077164518e-05}
{"train_loss": 0.09982335567474365, "global_step": 185650, "epoch": 2085, "lr": 2.2767156874019513e-05}
{"train_loss": 0.06835544109344482, "global_step": 185651, "epoch": 2085, "lr": 2.2766670674535627e-05}
{"train_loss": 0.14410659670829773, "global_step": 185652, "epoch": 2085, "lr": 2.2766184478712954e-05}
{"train_loss": 0.07910067472983612, "global_step": 185653, "epoch": 2085, "lr": 2.2765698286551516e-05, "val_loss": 7.589707851409912, "train_action_mse_error": 14.430827140808105}
{"train_loss": 0.0992974042892456, "global_step": 185654, "epoch": 2086, "lr": 2.276521209805142e-05}
{"train_loss": 0.08022993057966232, "global_step": 185655, "epoch": 2086, "lr": 2.2764725913212686e-05}
{"train_loss": 0.04347125440835953, "global_step": 185656, "epoch": 2086, "lr": 2.276423973203542e-05}
{"train_loss": 0.09513883292675018, "global_step": 185657, "epoch": 2086, "lr": 2.2763753554519655e-05}
{"train_loss": 0.03869393840432167, "global_step": 185658, "epoch": 2086, "lr": 2.2763267380665486e-05}
{"train_loss": 0.03957220911979675, "global_step": 185659, "epoch": 2086, "lr": 2.276278121047295e-05}
{"train_loss": 0.07715754210948944, "global_step": 185660, "epoch": 2086, "lr": 2.2762295043942145e-05}
{"train_loss": 0.040277235209941864, "global_step": 185661, "epoch": 2086, "lr": 2.2761808881073092e-05}
{"train_loss": 0.07723403722047806, "global_step": 185662, "epoch": 2086, "lr": 2.2761322721865908e-05}
{"train_loss": 0.06098870933055878, "global_step": 185663, "epoch": 2086, "lr": 2.276083656632061e-05}
{"train_loss": 0.06400232017040253, "global_step": 185664, "epoch": 2086, "lr": 2.2760350414437303e-05}
{"train_loss": 0.07815003395080566, "global_step": 185665, "epoch": 2086, "lr": 2.2759864266216018e-05}
{"train_loss": 0.059049881994724274, "global_step": 185666, "epoch": 2086, "lr": 2.2759378121656856e-05}
{"train_loss": 0.06355015188455582, "global_step": 185667, "epoch": 2086, "lr": 2.2758891980759868e-05}
{"train_loss": 0.06334131211042404, "global_step": 185668, "epoch": 2086, "lr": 2.2758405843525088e-05}
{"train_loss": 0.10236436128616333, "global_step": 185669, "epoch": 2086, "lr": 2.275791970995263e-05}
{"train_loss": 0.07777181267738342, "global_step": 185670, "epoch": 2086, "lr": 2.275743358004252e-05}
{"train_loss": 0.024888359010219574, "global_step": 185671, "epoch": 2086, "lr": 2.2756947453794856e-05}
{"train_loss": 0.09076879173517227, "global_step": 185672, "epoch": 2086, "lr": 2.2756461331209677e-05}
{"train_loss": 0.06528227776288986, "global_step": 185673, "epoch": 2086, "lr": 2.2755975212287072e-05}
{"train_loss": 0.09907156974077225, "global_step": 185674, "epoch": 2086, "lr": 2.2755489097027078e-05}
{"train_loss": 0.04750771075487137, "global_step": 185675, "epoch": 2086, "lr": 2.2755002985429796e-05}
{"train_loss": 0.07549959421157837, "global_step": 185676, "epoch": 2086, "lr": 2.275451687749525e-05}
{"train_loss": 0.07811892032623291, "global_step": 185677, "epoch": 2086, "lr": 2.275403077322355e-05}
{"train_loss": 0.0722709596157074, "global_step": 185678, "epoch": 2086, "lr": 2.275354467261472e-05}
{"train_loss": 0.11191932111978531, "global_step": 185679, "epoch": 2086, "lr": 2.2753058575668855e-05}
{"train_loss": 0.09143765270709991, "global_step": 185680, "epoch": 2086, "lr": 2.2752572482386e-05}
{"train_loss": 0.05949961394071579, "global_step": 185681, "epoch": 2086, "lr": 2.275208639276624e-05}
{"train_loss": 0.04545542970299721, "global_step": 185682, "epoch": 2086, "lr": 2.275160030680962e-05}
{"train_loss": 0.07941046357154846, "global_step": 185683, "epoch": 2086, "lr": 2.2751114224516225e-05}
{"train_loss": 0.0784967765212059, "global_step": 185684, "epoch": 2086, "lr": 2.2750628145886092e-05}
{"train_loss": 0.044244877994060516, "global_step": 185685, "epoch": 2086, "lr": 2.2750142070919334e-05}
{"train_loss": 0.04010011628270149, "global_step": 185686, "epoch": 2086, "lr": 2.274965599961596e-05}
{"train_loss": 0.10737039148807526, "global_step": 185687, "epoch": 2086, "lr": 2.2749169931976065e-05}
{"train_loss": 0.059132788330316544, "global_step": 185688, "epoch": 2086, "lr": 2.274868386799973e-05}
{"train_loss": 0.10556317120790482, "global_step": 185689, "epoch": 2086, "lr": 2.274819780768698e-05}
{"train_loss": 0.07340352982282639, "global_step": 185690, "epoch": 2086, "lr": 2.2747711751037927e-05}
{"train_loss": 0.10144033282995224, "global_step": 185691, "epoch": 2086, "lr": 2.2747225698052594e-05}
{"train_loss": 0.0727248415350914, "global_step": 185692, "epoch": 2086, "lr": 2.274673964873108e-05}
{"train_loss": 0.08897765725851059, "global_step": 185693, "epoch": 2086, "lr": 2.2746253603073415e-05}
{"train_loss": 0.063518226146698, "global_step": 185694, "epoch": 2086, "lr": 2.2745767561079704e-05}
{"train_loss": 0.0866643637418747, "global_step": 185695, "epoch": 2086, "lr": 2.2745281522749973e-05}
{"train_loss": 0.08076097816228867, "global_step": 185696, "epoch": 2086, "lr": 2.274479548808433e-05}
{"train_loss": 0.07682958245277405, "global_step": 185697, "epoch": 2086, "lr": 2.274430945708279e-05}
{"train_loss": 0.07633041590452194, "global_step": 185698, "epoch": 2086, "lr": 2.274382342974547e-05}
{"train_loss": 0.052443794906139374, "global_step": 185699, "epoch": 2086, "lr": 2.274333740607239e-05}
{"train_loss": 0.08369211852550507, "global_step": 185700, "epoch": 2086, "lr": 2.2742851386063656e-05}
{"train_loss": 0.10772722959518433, "global_step": 185701, "epoch": 2086, "lr": 2.27423653697193e-05}
{"train_loss": 0.11955269426107407, "global_step": 185702, "epoch": 2086, "lr": 2.274187935703941e-05}
{"train_loss": 0.06895793229341507, "global_step": 185703, "epoch": 2086, "lr": 2.2741393348024025e-05}
{"train_loss": 0.10115210711956024, "global_step": 185704, "epoch": 2086, "lr": 2.274090734267325e-05}
{"train_loss": 0.04763466492295265, "global_step": 185705, "epoch": 2086, "lr": 2.2740421340987127e-05}
{"train_loss": 0.1008017286658287, "global_step": 185706, "epoch": 2086, "lr": 2.27399353429657e-05}
{"train_loss": 0.08048050105571747, "global_step": 185707, "epoch": 2086, "lr": 2.2739449348609076e-05}
{"train_loss": 0.08805545419454575, "global_step": 185708, "epoch": 2086, "lr": 2.2738963357917288e-05}
{"train_loss": 0.08943358808755875, "global_step": 185709, "epoch": 2086, "lr": 2.2738477370890428e-05}
{"train_loss": 0.07817448675632477, "global_step": 185710, "epoch": 2086, "lr": 2.2737991387528527e-05}
{"train_loss": 0.047763969749212265, "global_step": 185711, "epoch": 2086, "lr": 2.2737505407831694e-05}
{"train_loss": 0.0528227798640728, "global_step": 185712, "epoch": 2086, "lr": 2.273701943179995e-05}
{"train_loss": 0.03183287754654884, "global_step": 185713, "epoch": 2086, "lr": 2.27365334594334e-05}
{"train_loss": 0.11444825679063797, "global_step": 185714, "epoch": 2086, "lr": 2.273604749073207e-05}
{"train_loss": 0.03484367951750755, "global_step": 185715, "epoch": 2086, "lr": 2.273556152569607e-05}
{"train_loss": 0.04402780160307884, "global_step": 185716, "epoch": 2086, "lr": 2.273507556432542e-05}
{"train_loss": 0.09003143757581711, "global_step": 185717, "epoch": 2086, "lr": 2.273458960662022e-05}
{"train_loss": 0.09429416805505753, "global_step": 185718, "epoch": 2086, "lr": 2.2734103652580513e-05}
{"train_loss": 0.10550399124622345, "global_step": 185719, "epoch": 2086, "lr": 2.273361770220639e-05}
{"train_loss": 0.174355149269104, "global_step": 185720, "epoch": 2086, "lr": 2.2733131755497877e-05}
{"train_loss": 0.09735146164894104, "global_step": 185721, "epoch": 2086, "lr": 2.2732645812455083e-05}
{"train_loss": 0.057952623814344406, "global_step": 185722, "epoch": 2086, "lr": 2.2732159873078035e-05}
{"train_loss": 0.08683431148529053, "global_step": 185723, "epoch": 2086, "lr": 2.273167393736681e-05}
{"train_loss": 0.11427481472492218, "global_step": 185724, "epoch": 2086, "lr": 2.2731188005321502e-05}
{"train_loss": 0.0502069890499115, "global_step": 185725, "epoch": 2086, "lr": 2.2730702076942133e-05}
{"train_loss": 0.08863768726587296, "global_step": 185726, "epoch": 2086, "lr": 2.273021615222881e-05}
{"train_loss": 0.051441460847854614, "global_step": 185727, "epoch": 2086, "lr": 2.2729730231181562e-05}
{"train_loss": 0.06077170372009277, "global_step": 185728, "epoch": 2086, "lr": 2.272924431380048e-05}
{"train_loss": 0.03760450705885887, "global_step": 185729, "epoch": 2086, "lr": 2.2728758400085604e-05}
{"train_loss": 0.0554727241396904, "global_step": 185730, "epoch": 2086, "lr": 2.2728272490037034e-05}
{"train_loss": 0.08877023309469223, "global_step": 185731, "epoch": 2086, "lr": 2.27277865836548e-05}
{"train_loss": 0.04700738564133644, "global_step": 185732, "epoch": 2086, "lr": 2.272730068093899e-05}
{"train_loss": 0.050419509410858154, "global_step": 185733, "epoch": 2086, "lr": 2.2726814781889648e-05}
{"train_loss": 0.10244103521108627, "global_step": 185734, "epoch": 2086, "lr": 2.2726328886506874e-05}
{"train_loss": 0.035513635724782944, "global_step": 185735, "epoch": 2086, "lr": 2.2725842994790696e-05}
{"train_loss": 0.07551789283752441, "global_step": 185736, "epoch": 2086, "lr": 2.272535710674121e-05}
{"train_loss": 0.045253004878759384, "global_step": 185737, "epoch": 2086, "lr": 2.272487122235846e-05}
{"train_loss": 0.051353078335523605, "global_step": 185738, "epoch": 2086, "lr": 2.2724385341642524e-05}
{"train_loss": 0.0797661542892456, "global_step": 185739, "epoch": 2086, "lr": 2.2723899464593447e-05}
{"train_loss": 0.09278221428394318, "global_step": 185740, "epoch": 2086, "lr": 2.272341359121133e-05}
{"train_loss": 0.08743540197610855, "global_step": 185741, "epoch": 2086, "lr": 2.2722927721496196e-05}
{"train_loss": 0.07380385939659698, "global_step": 185742, "epoch": 2086, "lr": 2.2722441855448152e-05, "val_loss": 7.769047260284424}
{"train_loss": 0.06613526493310928, "global_step": 185743, "epoch": 2087, "lr": 2.2721955993067245e-05}
{"train_loss": 0.04058430716395378, "global_step": 185744, "epoch": 2087, "lr": 2.2721470134353517e-05}
{"train_loss": 0.026350583881139755, "global_step": 185745, "epoch": 2087, "lr": 2.2720984279307072e-05}
{"train_loss": 0.06063558906316757, "global_step": 185746, "epoch": 2087, "lr": 2.2720498427927945e-05}
{"train_loss": 0.02630016952753067, "global_step": 185747, "epoch": 2087, "lr": 2.2720012580216233e-05}
{"train_loss": 0.1299431174993515, "global_step": 185748, "epoch": 2087, "lr": 2.271952673617196e-05}
{"train_loss": 0.07555848360061646, "global_step": 185749, "epoch": 2087, "lr": 2.2719040895795235e-05}
{"train_loss": 0.07029590755701065, "global_step": 185750, "epoch": 2087, "lr": 2.2718555059086084e-05}
{"train_loss": 0.039173513650894165, "global_step": 185751, "epoch": 2087, "lr": 2.2718069226044603e-05}
{"train_loss": 0.10241317003965378, "global_step": 185752, "epoch": 2087, "lr": 2.2717583396670833e-05}
{"train_loss": 0.09086064994335175, "global_step": 185753, "epoch": 2087, "lr": 2.2717097570964867e-05}
{"train_loss": 0.07824889570474625, "global_step": 185754, "epoch": 2087, "lr": 2.271661174892674e-05}
{"train_loss": 0.05343609303236008, "global_step": 185755, "epoch": 2087, "lr": 2.2716125930556548e-05}
{"train_loss": 0.06508023291826248, "global_step": 185756, "epoch": 2087, "lr": 2.2715640115854327e-05}
{"train_loss": 0.035872284322977066, "global_step": 185757, "epoch": 2087, "lr": 2.2715154304820168e-05}
{"train_loss": 0.05571846663951874, "global_step": 185758, "epoch": 2087, "lr": 2.2714668497454105e-05}
{"train_loss": 0.040906451642513275, "global_step": 185759, "epoch": 2087, "lr": 2.2714182693756224e-05}
{"train_loss": 0.12654632329940796, "global_step": 185760, "epoch": 2087, "lr": 2.271369689372661e-05}
{"train_loss": 0.06679050624370575, "global_step": 185761, "epoch": 2087, "lr": 2.2713211097365283e-05}
{"train_loss": 0.05274643748998642, "global_step": 185762, "epoch": 2087, "lr": 2.2712725304672356e-05}
{"train_loss": 0.09445591270923615, "global_step": 185763, "epoch": 2087, "lr": 2.271223951564785e-05}
{"train_loss": 0.052316173911094666, "global_step": 185764, "epoch": 2087, "lr": 2.2711753730291868e-05}
{"train_loss": 0.07418584078550339, "global_step": 185765, "epoch": 2087, "lr": 2.2711267948604436e-05}
{"train_loss": 0.05141085386276245, "global_step": 185766, "epoch": 2087, "lr": 2.2710782170585666e-05}
{"train_loss": 0.07837013900279999, "global_step": 185767, "epoch": 2087, "lr": 2.271029639623558e-05}
{"train_loss": 0.08537910878658295, "global_step": 185768, "epoch": 2087, "lr": 2.270981062555428e-05}
{"train_loss": 0.020141450688242912, "global_step": 185769, "epoch": 2087, "lr": 2.2709324858541793e-05}
{"train_loss": 0.0591248944401741, "global_step": 185770, "epoch": 2087, "lr": 2.2708839095198225e-05}
{"train_loss": 0.07363932579755783, "global_step": 185771, "epoch": 2087, "lr": 2.2708353335523604e-05}
{"train_loss": 0.07373739033937454, "global_step": 185772, "epoch": 2087, "lr": 2.270786757951802e-05}
{"train_loss": 0.08452728390693665, "global_step": 185773, "epoch": 2087, "lr": 2.2707381827181524e-05}
{"train_loss": 0.13757620751857758, "global_step": 185774, "epoch": 2087, "lr": 2.2706896078514205e-05}
{"train_loss": 0.08822773396968842, "global_step": 185775, "epoch": 2087, "lr": 2.2706410333516086e-05}
{"train_loss": 0.11649619042873383, "global_step": 185776, "epoch": 2087, "lr": 2.2705924592187284e-05}
{"train_loss": 0.06953271478414536, "global_step": 185777, "epoch": 2087, "lr": 2.2705438854527815e-05}
{"train_loss": 0.07454262673854828, "global_step": 185778, "epoch": 2087, "lr": 2.2704953120537788e-05}
{"train_loss": 0.04477521777153015, "global_step": 185779, "epoch": 2087, "lr": 2.2704467390217226e-05}
{"train_loss": 0.04295116290450096, "global_step": 185780, "epoch": 2087, "lr": 2.270398166356623e-05}
{"train_loss": 0.06313877552747726, "global_step": 185781, "epoch": 2087, "lr": 2.270349594058484e-05}
{"train_loss": 0.05140376463532448, "global_step": 185782, "epoch": 2087, "lr": 2.270301022127315e-05}
{"train_loss": 0.07898755371570587, "global_step": 185783, "epoch": 2087, "lr": 2.2702524505631206e-05}
{"train_loss": 0.03739924728870392, "global_step": 185784, "epoch": 2087, "lr": 2.2702038793659058e-05}
{"train_loss": 0.07671482861042023, "global_step": 185785, "epoch": 2087, "lr": 2.2701553085356802e-05}
{"train_loss": 0.052931006997823715, "global_step": 185786, "epoch": 2087, "lr": 2.270106738072448e-05}
{"train_loss": 0.08784478902816772, "global_step": 185787, "epoch": 2087, "lr": 2.2700581679762174e-05}
{"train_loss": 0.11637510359287262, "global_step": 185788, "epoch": 2087, "lr": 2.2700095982469927e-05}
{"train_loss": 0.04047944024205208, "global_step": 185789, "epoch": 2087, "lr": 2.2699610288847843e-05}
{"train_loss": 0.08967509120702744, "global_step": 185790, "epoch": 2087, "lr": 2.2699124598895943e-05}
{"train_loss": 0.06837587058544159, "global_step": 185791, "epoch": 2087, "lr": 2.2698638912614328e-05}
{"train_loss": 0.07670964300632477, "global_step": 185792, "epoch": 2087, "lr": 2.2698153230003032e-05}
{"train_loss": 0.03325851634144783, "global_step": 185793, "epoch": 2087, "lr": 2.2697667551062157e-05}
{"train_loss": 0.12041249871253967, "global_step": 185794, "epoch": 2087, "lr": 2.2697181875791723e-05}
{"train_loss": 0.06891218572854996, "global_step": 185795, "epoch": 2087, "lr": 2.269669620419185e-05}
{"train_loss": 0.10775035619735718, "global_step": 185796, "epoch": 2087, "lr": 2.269621053626254e-05}
{"train_loss": 0.08379318565130234, "global_step": 185797, "epoch": 2087, "lr": 2.2695724872003905e-05}
{"train_loss": 0.06636874377727509, "global_step": 185798, "epoch": 2087, "lr": 2.2695239211416008e-05}
{"train_loss": 0.06804066896438599, "global_step": 185799, "epoch": 2087, "lr": 2.2694753554498887e-05}
{"train_loss": 0.08609387278556824, "global_step": 185800, "epoch": 2087, "lr": 2.269426790125264e-05}
{"train_loss": 0.10829725116491318, "global_step": 185801, "epoch": 2087, "lr": 2.2693782251677304e-05}
{"train_loss": 0.08091630786657333, "global_step": 185802, "epoch": 2087, "lr": 2.2693296605772967e-05}
{"train_loss": 0.051601771265268326, "global_step": 185803, "epoch": 2087, "lr": 2.2692810963539667e-05}
{"train_loss": 0.09159114211797714, "global_step": 185804, "epoch": 2087, "lr": 2.269232532497751e-05}
{"train_loss": 0.04443119466304779, "global_step": 185805, "epoch": 2087, "lr": 2.269183969008651e-05}
{"train_loss": 0.09539251029491425, "global_step": 185806, "epoch": 2087, "lr": 2.2691354058866782e-05}
{"train_loss": 0.10202496498823166, "global_step": 185807, "epoch": 2087, "lr": 2.269086843131835e-05}
{"train_loss": 0.05807747319340706, "global_step": 185808, "epoch": 2087, "lr": 2.2690382807441312e-05}
{"train_loss": 0.06003400683403015, "global_step": 185809, "epoch": 2087, "lr": 2.2689897187235708e-05}
{"train_loss": 0.11823300272226334, "global_step": 185810, "epoch": 2087, "lr": 2.2689411570701628e-05}
{"train_loss": 0.10192850977182388, "global_step": 185811, "epoch": 2087, "lr": 2.2688925957839108e-05}
{"train_loss": 0.141806498169899, "global_step": 185812, "epoch": 2087, "lr": 2.268844034864825e-05}
{"train_loss": 0.08924072980880737, "global_step": 185813, "epoch": 2087, "lr": 2.2687954743129074e-05}
{"train_loss": 0.11700472980737686, "global_step": 185814, "epoch": 2087, "lr": 2.268746914128167e-05}
{"train_loss": 0.08379318565130234, "global_step": 185815, "epoch": 2087, "lr": 2.2686983543106123e-05}
{"train_loss": 0.07733431458473206, "global_step": 185816, "epoch": 2087, "lr": 2.268649794860248e-05}
{"train_loss": 0.04780268669128418, "global_step": 185817, "epoch": 2087, "lr": 2.2686012357770785e-05}
{"train_loss": 0.06026599556207657, "global_step": 185818, "epoch": 2087, "lr": 2.2685526770611137e-05}
{"train_loss": 0.03931853175163269, "global_step": 185819, "epoch": 2087, "lr": 2.2685041187123574e-05}
{"train_loss": 0.060426145792007446, "global_step": 185820, "epoch": 2087, "lr": 2.2684555607308193e-05}
{"train_loss": 0.08990826457738876, "global_step": 185821, "epoch": 2087, "lr": 2.268407003116504e-05}
{"train_loss": 0.10305143892765045, "global_step": 185822, "epoch": 2087, "lr": 2.2683584458694155e-05}
{"train_loss": 0.06424655765295029, "global_step": 185823, "epoch": 2087, "lr": 2.2683098889895655e-05}
{"train_loss": 0.05419555678963661, "global_step": 185824, "epoch": 2087, "lr": 2.268261332476956e-05}
{"train_loss": 0.05609065666794777, "global_step": 185825, "epoch": 2087, "lr": 2.2682127763315973e-05}
{"train_loss": 0.04444674029946327, "global_step": 185826, "epoch": 2087, "lr": 2.268164220553492e-05}
{"train_loss": 0.10442718118429184, "global_step": 185827, "epoch": 2087, "lr": 2.268115665142651e-05}
{"train_loss": 0.11861947178840637, "global_step": 185828, "epoch": 2087, "lr": 2.2680671100990764e-05}
{"train_loss": 0.05388028547167778, "global_step": 185829, "epoch": 2087, "lr": 2.268018555422779e-05}
{"train_loss": 0.04309752583503723, "global_step": 185830, "epoch": 2087, "lr": 2.267970001113761e-05}
{"train_loss": 0.07345274950932251, "global_step": 185831, "epoch": 2087, "lr": 2.2679214471720332e-05, "val_loss": 7.687046051025391}
{"train_loss": 0.105864517390728, "global_step": 185832, "epoch": 2088, "lr": 2.2678728935975977e-05}
{"train_loss": 0.1016692966222763, "global_step": 185833, "epoch": 2088, "lr": 2.2678243403904642e-05}
{"train_loss": 0.11090117692947388, "global_step": 185834, "epoch": 2088, "lr": 2.26777578755064e-05}
{"train_loss": 0.053508393466472626, "global_step": 185835, "epoch": 2088, "lr": 2.267727235078128e-05}
{"train_loss": 0.05553991347551346, "global_step": 185836, "epoch": 2088, "lr": 2.267678682972939e-05}
{"train_loss": 0.07514034956693649, "global_step": 185837, "epoch": 2088, "lr": 2.267630131235075e-05}
{"train_loss": 0.08020051568746567, "global_step": 185838, "epoch": 2088, "lr": 2.2675815798645467e-05}
{"train_loss": 0.0966276153922081, "global_step": 185839, "epoch": 2088, "lr": 2.2675330288613573e-05}
{"train_loss": 0.04454323649406433, "global_step": 185840, "epoch": 2088, "lr": 2.2674844782255163e-05}
{"train_loss": 0.06446550786495209, "global_step": 185841, "epoch": 2088, "lr": 2.2674359279570267e-05}
{"train_loss": 0.05791214108467102, "global_step": 185842, "epoch": 2088, "lr": 2.2673873780558997e-05}
{"train_loss": 0.06141265481710434, "global_step": 185843, "epoch": 2088, "lr": 2.2673388285221363e-05}
{"train_loss": 0.08445057272911072, "global_step": 185844, "epoch": 2088, "lr": 2.2672902793557484e-05}
{"train_loss": 0.11708924919366837, "global_step": 185845, "epoch": 2088, "lr": 2.267241730556738e-05}
{"train_loss": 0.10123871266841888, "global_step": 185846, "epoch": 2088, "lr": 2.2671931821251152e-05}
{"train_loss": 0.04639824479818344, "global_step": 185847, "epoch": 2088, "lr": 2.267144634060883e-05}
{"train_loss": 0.12833233177661896, "global_step": 185848, "epoch": 2088, "lr": 2.2670960863640523e-05}
{"train_loss": 0.045922599732875824, "global_step": 185849, "epoch": 2088, "lr": 2.267047539034625e-05}
{"train_loss": 0.11221323162317276, "global_step": 185850, "epoch": 2088, "lr": 2.2669989920726103e-05}
{"train_loss": 0.0619174987077713, "global_step": 185851, "epoch": 2088, "lr": 2.266950445478016e-05}
{"train_loss": 0.11414147168397903, "global_step": 185852, "epoch": 2088, "lr": 2.2669018992508452e-05}
{"train_loss": 0.08684059232473373, "global_step": 185853, "epoch": 2088, "lr": 2.2668533533911072e-05}
{"train_loss": 0.08449500799179077, "global_step": 185854, "epoch": 2088, "lr": 2.266804807898808e-05}
{"train_loss": 0.06146315857768059, "global_step": 185855, "epoch": 2088, "lr": 2.2667562627739518e-05}
{"train_loss": 0.12629055976867676, "global_step": 185856, "epoch": 2088, "lr": 2.2667077180165486e-05}
{"train_loss": 0.05386239290237427, "global_step": 185857, "epoch": 2088, "lr": 2.2666591736266015e-05}
{"train_loss": 0.08310577273368835, "global_step": 185858, "epoch": 2088, "lr": 2.26661062960412e-05}
{"train_loss": 0.0719551146030426, "global_step": 185859, "epoch": 2088, "lr": 2.2665620859491104e-05}
{"train_loss": 0.11718874424695969, "global_step": 185860, "epoch": 2088, "lr": 2.2665135426615757e-05}
{"train_loss": 0.09151414781808853, "global_step": 185861, "epoch": 2088, "lr": 2.2664649997415272e-05}
{"train_loss": 0.03986024111509323, "global_step": 185862, "epoch": 2088, "lr": 2.2664164571889667e-05}
{"train_loss": 0.07419189810752869, "global_step": 185863, "epoch": 2088, "lr": 2.2663679150039053e-05}
{"train_loss": 0.10788649320602417, "global_step": 185864, "epoch": 2088, "lr": 2.2663193731863457e-05}
{"train_loss": 0.12021642178297043, "global_step": 185865, "epoch": 2088, "lr": 2.266270831736298e-05}
{"train_loss": 0.025740673765540123, "global_step": 185866, "epoch": 2088, "lr": 2.266222290653765e-05}
{"train_loss": 0.0695914477109909, "global_step": 185867, "epoch": 2088, "lr": 2.266173749938757e-05}
{"train_loss": 0.06590528041124344, "global_step": 185868, "epoch": 2088, "lr": 2.2661252095912768e-05}
{"train_loss": 0.062043026089668274, "global_step": 185869, "epoch": 2088, "lr": 2.266076669611332e-05}
{"train_loss": 0.047845929861068726, "global_step": 185870, "epoch": 2088, "lr": 2.2660281299989327e-05}
{"train_loss": 0.12847653031349182, "global_step": 185871, "epoch": 2088, "lr": 2.2659795907540803e-05}
{"train_loss": 0.07589482516050339, "global_step": 185872, "epoch": 2088, "lr": 2.2659310518767855e-05}
{"train_loss": 0.022938072681427002, "global_step": 185873, "epoch": 2088, "lr": 2.2658825133670512e-05}
{"train_loss": 0.062434326857328415, "global_step": 185874, "epoch": 2088, "lr": 2.2658339752248876e-05}
{"train_loss": 0.07372254133224487, "global_step": 185875, "epoch": 2088, "lr": 2.2657854374502973e-05}
{"train_loss": 0.05986281856894493, "global_step": 185876, "epoch": 2088, "lr": 2.265736900043291e-05}
{"train_loss": 0.16363398730754852, "global_step": 185877, "epoch": 2088, "lr": 2.2656883630038705e-05}
{"train_loss": 0.06958679109811783, "global_step": 185878, "epoch": 2088, "lr": 2.2656398263320473e-05}
{"train_loss": 0.08892285078763962, "global_step": 185879, "epoch": 2088, "lr": 2.2655912900278235e-05}
{"train_loss": 0.05395649001002312, "global_step": 185880, "epoch": 2088, "lr": 2.2655427540912093e-05}
{"train_loss": 0.0832754522562027, "global_step": 185881, "epoch": 2088, "lr": 2.2654942185222077e-05}
{"train_loss": 0.08506085723638535, "global_step": 185882, "epoch": 2088, "lr": 2.2654456833208293e-05}
{"train_loss": 0.07862850278615952, "global_step": 185883, "epoch": 2088, "lr": 2.265397148487076e-05}
{"train_loss": 0.11756179481744766, "global_step": 185884, "epoch": 2088, "lr": 2.2653486140209588e-05}
{"train_loss": 0.054552581161260605, "global_step": 185885, "epoch": 2088, "lr": 2.2653000799224807e-05}
{"train_loss": 0.13399243354797363, "global_step": 185886, "epoch": 2088, "lr": 2.2652515461916497e-05}
{"train_loss": 0.07099007815122604, "global_step": 185887, "epoch": 2088, "lr": 2.2652030128284734e-05}
{"train_loss": 0.10498910397291183, "global_step": 185888, "epoch": 2088, "lr": 2.265154479832956e-05}
{"train_loss": 0.13562990725040436, "global_step": 185889, "epoch": 2088, "lr": 2.2651059472051067e-05}
{"train_loss": 0.07392473518848419, "global_step": 185890, "epoch": 2088, "lr": 2.265057414944929e-05}
{"train_loss": 0.07953650504350662, "global_step": 185891, "epoch": 2088, "lr": 2.265008883052433e-05}
{"train_loss": 0.0950019583106041, "global_step": 185892, "epoch": 2088, "lr": 2.264960351527621e-05}
{"train_loss": 0.019672559574246407, "global_step": 185893, "epoch": 2088, "lr": 2.2649118203705033e-05}
{"train_loss": 0.049013521522283554, "global_step": 185894, "epoch": 2088, "lr": 2.2648632895810857e-05}
{"train_loss": 0.049387477338314056, "global_step": 185895, "epoch": 2088, "lr": 2.2648147591593715e-05}
{"train_loss": 0.06429213285446167, "global_step": 185896, "epoch": 2088, "lr": 2.2647662291053717e-05}
{"train_loss": 0.06576166301965714, "global_step": 185897, "epoch": 2088, "lr": 2.2647176994190888e-05}
{"train_loss": 0.09075478464365005, "global_step": 185898, "epoch": 2088, "lr": 2.2646691701005328e-05}
{"train_loss": 0.09124013781547546, "global_step": 185899, "epoch": 2088, "lr": 2.264620641149709e-05}
{"train_loss": 0.09403325617313385, "global_step": 185900, "epoch": 2088, "lr": 2.2645721125666218e-05}
{"train_loss": 0.08202122896909714, "global_step": 185901, "epoch": 2088, "lr": 2.2645235843512814e-05}
{"train_loss": 0.07384853810071945, "global_step": 185902, "epoch": 2088, "lr": 2.2644750565036902e-05}
{"train_loss": 0.03188447654247284, "global_step": 185903, "epoch": 2088, "lr": 2.2644265290238598e-05}
{"train_loss": 0.047622743993997574, "global_step": 185904, "epoch": 2088, "lr": 2.264378001911791e-05}
{"train_loss": 0.047303423285484314, "global_step": 185905, "epoch": 2088, "lr": 2.2643294751674938e-05}
{"train_loss": 0.07649614661931992, "global_step": 185906, "epoch": 2088, "lr": 2.2642809487909762e-05}
{"train_loss": 0.07373176515102386, "global_step": 185907, "epoch": 2088, "lr": 2.2642324227822404e-05}
{"train_loss": 0.042182981967926025, "global_step": 185908, "epoch": 2088, "lr": 2.264183897141297e-05}
{"train_loss": 0.12416453659534454, "global_step": 185909, "epoch": 2088, "lr": 2.264135371868149e-05}
{"train_loss": 0.06689731031656265, "global_step": 185910, "epoch": 2088, "lr": 2.2640868469628068e-05}
{"train_loss": 0.09015584737062454, "global_step": 185911, "epoch": 2088, "lr": 2.2640383224252726e-05}
{"train_loss": 0.11186230182647705, "global_step": 185912, "epoch": 2088, "lr": 2.2639897982555568e-05}
{"train_loss": 0.06787118315696716, "global_step": 185913, "epoch": 2088, "lr": 2.2639412744536633e-05}
{"train_loss": 0.08850382268428802, "global_step": 185914, "epoch": 2088, "lr": 2.2638927510196e-05}
{"train_loss": 0.0824969932436943, "global_step": 185915, "epoch": 2088, "lr": 2.2638442279533722e-05}
{"train_loss": 0.049007248133420944, "global_step": 185916, "epoch": 2088, "lr": 2.2637957052549885e-05}
{"train_loss": 0.05635065212845802, "global_step": 185917, "epoch": 2088, "lr": 2.263747182924453e-05}
{"train_loss": 0.051866985857486725, "global_step": 185918, "epoch": 2088, "lr": 2.2636986609617746e-05}
{"train_loss": 0.09035217761993408, "global_step": 185919, "epoch": 2088, "lr": 2.2636501393669562e-05}
{"train_loss": 0.0786350214163239, "global_step": 185920, "epoch": 2088, "lr": 2.2636016181400093e-05, "val_loss": 7.6567511558532715}
{"train_loss": 0.0642974004149437, "global_step": 185921, "epoch": 2089, "lr": 2.263553097280936e-05}
{"train_loss": 0.0898602157831192, "global_step": 185922, "epoch": 2089, "lr": 2.263504576789746e-05}
{"train_loss": 0.048433151096105576, "global_step": 185923, "epoch": 2089, "lr": 2.263456056666443e-05}
{"train_loss": 0.037704262882471085, "global_step": 185924, "epoch": 2089, "lr": 2.2634075369110347e-05}
{"train_loss": 0.0789928212761879, "global_step": 185925, "epoch": 2089, "lr": 2.2633590175235296e-05}
{"train_loss": 0.07942428439855576, "global_step": 185926, "epoch": 2089, "lr": 2.263310498503931e-05}
{"train_loss": 0.051899056881666183, "global_step": 185927, "epoch": 2089, "lr": 2.2632619798522492e-05}
{"train_loss": 0.08339949697256088, "global_step": 185928, "epoch": 2089, "lr": 2.2632134615684857e-05}
{"train_loss": 0.06942492723464966, "global_step": 185929, "epoch": 2089, "lr": 2.2631649436526526e-05}
{"train_loss": 0.07546856999397278, "global_step": 185930, "epoch": 2089, "lr": 2.263116426104751e-05}
{"train_loss": 0.09407313168048859, "global_step": 185931, "epoch": 2089, "lr": 2.2630679089247924e-05}
{"train_loss": 0.031183280050754547, "global_step": 185932, "epoch": 2089, "lr": 2.2630193921127808e-05}
{"train_loss": 0.04050975292921066, "global_step": 185933, "epoch": 2089, "lr": 2.2629708756687207e-05}
{"train_loss": 0.062097467482089996, "global_step": 185934, "epoch": 2089, "lr": 2.262922359592623e-05}
{"train_loss": 0.12953874468803406, "global_step": 185935, "epoch": 2089, "lr": 2.26287384388449e-05}
{"train_loss": 0.09593330323696136, "global_step": 185936, "epoch": 2089, "lr": 2.2628253285443323e-05}
{"train_loss": 0.08998587727546692, "global_step": 185937, "epoch": 2089, "lr": 2.2627768135721546e-05}
{"train_loss": 0.02520793303847313, "global_step": 185938, "epoch": 2089, "lr": 2.2627282989679604e-05}
{"train_loss": 0.1280224323272705, "global_step": 185939, "epoch": 2089, "lr": 2.2626797847317615e-05}
{"train_loss": 0.08978457748889923, "global_step": 185940, "epoch": 2089, "lr": 2.2626312708635604e-05}
{"train_loss": 0.05512223765254021, "global_step": 185941, "epoch": 2089, "lr": 2.2625827573633646e-05}
{"train_loss": 0.09552866220474243, "global_step": 185942, "epoch": 2089, "lr": 2.2625342442311836e-05}
{"train_loss": 0.05626366659998894, "global_step": 185943, "epoch": 2089, "lr": 2.262485731467019e-05}
{"train_loss": 0.07589980959892273, "global_step": 185944, "epoch": 2089, "lr": 2.262437219070882e-05}
{"train_loss": 0.03292807936668396, "global_step": 185945, "epoch": 2089, "lr": 2.262388707042775e-05}
{"train_loss": 0.05320439115166664, "global_step": 185946, "epoch": 2089, "lr": 2.262340195382709e-05}
{"train_loss": 0.08267197012901306, "global_step": 185947, "epoch": 2089, "lr": 2.2622916840906856e-05}
{"train_loss": 0.07779716700315475, "global_step": 185948, "epoch": 2089, "lr": 2.2622431731667155e-05}
{"train_loss": 0.10401985049247742, "global_step": 185949, "epoch": 2089, "lr": 2.2621946626108015e-05}
{"train_loss": 0.056878529489040375, "global_step": 185950, "epoch": 2089, "lr": 2.2621461524229538e-05}
{"train_loss": 0.07271646708250046, "global_step": 185951, "epoch": 2089, "lr": 2.2620976426031755e-05}
{"train_loss": 0.05762035772204399, "global_step": 185952, "epoch": 2089, "lr": 2.262049133151477e-05}
{"train_loss": 0.044547587633132935, "global_step": 185953, "epoch": 2089, "lr": 2.2620006240678605e-05}
{"train_loss": 0.1113133653998375, "global_step": 185954, "epoch": 2089, "lr": 2.2619521153523366e-05}
{"train_loss": 0.029319198802113533, "global_step": 185955, "epoch": 2089, "lr": 2.2619036070049076e-05}
{"train_loss": 0.08059602975845337, "global_step": 185956, "epoch": 2089, "lr": 2.2618550990255845e-05}
{"train_loss": 0.05420500412583351, "global_step": 185957, "epoch": 2089, "lr": 2.2618065914143693e-05}
{"train_loss": 0.09756682068109512, "global_step": 185958, "epoch": 2089, "lr": 2.2617580841712734e-05}
{"train_loss": 0.07245943695306778, "global_step": 185959, "epoch": 2089, "lr": 2.2617095772962986e-05}
{"train_loss": 0.038927752524614334, "global_step": 185960, "epoch": 2089, "lr": 2.2616610707894535e-05}
{"train_loss": 0.07801418006420135, "global_step": 185961, "epoch": 2089, "lr": 2.2616125646507462e-05}
{"train_loss": 0.07238588482141495, "global_step": 185962, "epoch": 2089, "lr": 2.2615640588801808e-05}
{"train_loss": 0.04924644157290459, "global_step": 185963, "epoch": 2089, "lr": 2.261515553477766e-05}
{"train_loss": 0.07097108662128448, "global_step": 185964, "epoch": 2089, "lr": 2.2614670484435053e-05}
{"train_loss": 0.06753414869308472, "global_step": 185965, "epoch": 2089, "lr": 2.2614185437774088e-05}
{"train_loss": 0.0922485888004303, "global_step": 185966, "epoch": 2089, "lr": 2.261370039479479e-05}
{"train_loss": 0.1238386258482933, "global_step": 185967, "epoch": 2089, "lr": 2.2613215355497268e-05}
{"train_loss": 0.08850998431444168, "global_step": 185968, "epoch": 2089, "lr": 2.261273031988155e-05}
{"train_loss": 0.07173041999340057, "global_step": 185969, "epoch": 2089, "lr": 2.2612245287947732e-05}
{"train_loss": 0.07938113063573837, "global_step": 185970, "epoch": 2089, "lr": 2.2611760259695863e-05}
{"train_loss": 0.14450562000274658, "global_step": 185971, "epoch": 2089, "lr": 2.261127523512599e-05}
{"train_loss": 0.07950834929943085, "global_step": 185972, "epoch": 2089, "lr": 2.2610790214238213e-05}
{"train_loss": 0.11660855263471603, "global_step": 185973, "epoch": 2089, "lr": 2.2610305197032572e-05}
{"train_loss": 0.09009014815092087, "global_step": 185974, "epoch": 2089, "lr": 2.2609820183509156e-05}
{"train_loss": 0.043405525386333466, "global_step": 185975, "epoch": 2089, "lr": 2.2609335173668012e-05}
{"train_loss": 0.03802625462412834, "global_step": 185976, "epoch": 2089, "lr": 2.2608850167509195e-05}
{"train_loss": 0.07476706057786942, "global_step": 185977, "epoch": 2089, "lr": 2.260836516503278e-05}
{"train_loss": 0.05601881816983223, "global_step": 185978, "epoch": 2089, "lr": 2.260788016623886e-05}
{"train_loss": 0.16062621772289276, "global_step": 185979, "epoch": 2089, "lr": 2.2607395171127455e-05}
{"train_loss": 0.0605124905705452, "global_step": 185980, "epoch": 2089, "lr": 2.260691017969867e-05}
{"train_loss": 0.06509482115507126, "global_step": 185981, "epoch": 2089, "lr": 2.2606425191952536e-05}
{"train_loss": 0.10568639636039734, "global_step": 185982, "epoch": 2089, "lr": 2.260594020788915e-05}
{"train_loss": 0.07240702956914902, "global_step": 185983, "epoch": 2089, "lr": 2.2605455227508543e-05}
{"train_loss": 0.05308942496776581, "global_step": 185984, "epoch": 2089, "lr": 2.260497025081082e-05}
{"train_loss": 0.05086684599518776, "global_step": 185985, "epoch": 2089, "lr": 2.2604485277796004e-05}
{"train_loss": 0.10640225559473038, "global_step": 185986, "epoch": 2089, "lr": 2.26040003084642e-05}
{"train_loss": 0.08329036831855774, "global_step": 185987, "epoch": 2089, "lr": 2.260351534281544e-05}
{"train_loss": 0.1019439846277237, "global_step": 185988, "epoch": 2089, "lr": 2.2603030380849817e-05}
{"train_loss": 0.0814637765288353, "global_step": 185989, "epoch": 2089, "lr": 2.2602545422567362e-05}
{"train_loss": 0.06922537833452225, "global_step": 185990, "epoch": 2089, "lr": 2.2602060467968182e-05}
{"train_loss": 0.11875524371862411, "global_step": 185991, "epoch": 2089, "lr": 2.2601575517052305e-05}
{"train_loss": 0.08329891413450241, "global_step": 185992, "epoch": 2089, "lr": 2.2601090569819828e-05}
{"train_loss": 0.07904300093650818, "global_step": 185993, "epoch": 2089, "lr": 2.2600605626270783e-05}
{"train_loss": 0.0946183055639267, "global_step": 185994, "epoch": 2089, "lr": 2.2600120686405274e-05}
{"train_loss": 0.07454214245080948, "global_step": 185995, "epoch": 2089, "lr": 2.2599635750223318e-05}
{"train_loss": 0.08289548754692078, "global_step": 185996, "epoch": 2089, "lr": 2.2599150817725016e-05}
{"train_loss": 0.07462204992771149, "global_step": 185997, "epoch": 2089, "lr": 2.2598665888910438e-05}
{"train_loss": 0.09981974214315414, "global_step": 185998, "epoch": 2089, "lr": 2.259818096377962e-05}
{"train_loss": 0.06148094683885574, "global_step": 185999, "epoch": 2089, "lr": 2.2597696042332655e-05}
{"train_loss": 0.050571173429489136, "global_step": 186000, "epoch": 2089, "lr": 2.2597211124569583e-05}
{"train_loss": 0.11970718204975128, "global_step": 186001, "epoch": 2089, "lr": 2.2596726210490497e-05}
{"train_loss": 0.08567187190055847, "global_step": 186002, "epoch": 2089, "lr": 2.2596241300095433e-05}
{"train_loss": 0.09237388521432877, "global_step": 186003, "epoch": 2089, "lr": 2.2595756393384487e-05}
{"train_loss": 0.13037699460983276, "global_step": 186004, "epoch": 2089, "lr": 2.2595271490357688e-05}
{"train_loss": 0.04400401934981346, "global_step": 186005, "epoch": 2089, "lr": 2.259478659101514e-05}
{"train_loss": 0.048546064645051956, "global_step": 186006, "epoch": 2089, "lr": 2.259430169535687e-05}
{"train_loss": 0.030707811936736107, "global_step": 186007, "epoch": 2089, "lr": 2.259381680338298e-05}
{"train_loss": 0.04834112897515297, "global_step": 186008, "epoch": 2089, "lr": 2.2593331915093495e-05}
{"train_loss": 0.07623386705357037, "global_step": 186009, "epoch": 2089, "lr": 2.2592847030488528e-05, "val_loss": 7.687490940093994}
{"train_loss": 0.1471884399652481, "global_step": 186010, "epoch": 2090, "lr": 2.2592362149568112e-05}
{"train_loss": 0.0734972134232521, "global_step": 186011, "epoch": 2090, "lr": 2.2591877272332307e-05}
{"train_loss": 0.10717491060495377, "global_step": 186012, "epoch": 2090, "lr": 2.2591392398781207e-05}
{"train_loss": 0.025113843381404877, "global_step": 186013, "epoch": 2090, "lr": 2.259090752891484e-05}
{"train_loss": 0.06998910754919052, "global_step": 186014, "epoch": 2090, "lr": 2.2590422662733313e-05}
{"train_loss": 0.11511608958244324, "global_step": 186015, "epoch": 2090, "lr": 2.2589937800236643e-05}
{"train_loss": 0.07758130878210068, "global_step": 186016, "epoch": 2090, "lr": 2.258945294142495e-05}
{"train_loss": 0.028404388576745987, "global_step": 186017, "epoch": 2090, "lr": 2.2588968086298247e-05}
{"train_loss": 0.058998122811317444, "global_step": 186018, "epoch": 2090, "lr": 2.2588483234856645e-05}
{"train_loss": 0.08115410059690475, "global_step": 186019, "epoch": 2090, "lr": 2.2587998387100168e-05}
{"train_loss": 0.08828271180391312, "global_step": 186020, "epoch": 2090, "lr": 2.2587513543028915e-05}
{"train_loss": 0.09796357154846191, "global_step": 186021, "epoch": 2090, "lr": 2.258702870264292e-05}
{"train_loss": 0.05714971199631691, "global_step": 186022, "epoch": 2090, "lr": 2.2586543865942283e-05}
{"train_loss": 0.07453245669603348, "global_step": 186023, "epoch": 2090, "lr": 2.258605903292704e-05}
{"train_loss": 0.13313323259353638, "global_step": 186024, "epoch": 2090, "lr": 2.2585574203597276e-05}
{"train_loss": 0.09546803683042526, "global_step": 186025, "epoch": 2090, "lr": 2.2585089377953035e-05}
{"train_loss": 0.06708761304616928, "global_step": 186026, "epoch": 2090, "lr": 2.258460455599441e-05}
{"train_loss": 0.031285881996154785, "global_step": 186027, "epoch": 2090, "lr": 2.2584119737721433e-05}
{"train_loss": 0.058960143476724625, "global_step": 186028, "epoch": 2090, "lr": 2.2583634923134207e-05}
{"train_loss": 0.07304789870977402, "global_step": 186029, "epoch": 2090, "lr": 2.2583150112232754e-05}
{"train_loss": 0.056886397302150726, "global_step": 186030, "epoch": 2090, "lr": 2.258266530501718e-05}
{"train_loss": 0.054195474833250046, "global_step": 186031, "epoch": 2090, "lr": 2.258218050148752e-05}
{"train_loss": 0.04199550673365593, "global_step": 186032, "epoch": 2090, "lr": 2.2581695701643852e-05}
{"train_loss": 0.04378192499279976, "global_step": 186033, "epoch": 2090, "lr": 2.2581210905486254e-05}
{"train_loss": 0.08205714821815491, "global_step": 186034, "epoch": 2090, "lr": 2.258072611301476e-05}
{"train_loss": 0.06639768183231354, "global_step": 186035, "epoch": 2090, "lr": 2.2580241324229477e-05}
{"train_loss": 0.08913922309875488, "global_step": 186036, "epoch": 2090, "lr": 2.2579756539130426e-05}
{"train_loss": 0.0892377495765686, "global_step": 186037, "epoch": 2090, "lr": 2.2579271757717707e-05}
{"train_loss": 0.06222522631287575, "global_step": 186038, "epoch": 2090, "lr": 2.2578786979991356e-05}
{"train_loss": 0.04736305773258209, "global_step": 186039, "epoch": 2090, "lr": 2.2578302205951473e-05}
{"train_loss": 0.0963032990694046, "global_step": 186040, "epoch": 2090, "lr": 2.257781743559808e-05}
{"train_loss": 0.07716868817806244, "global_step": 186041, "epoch": 2090, "lr": 2.2577332668931282e-05}
{"train_loss": 0.05738220363855362, "global_step": 186042, "epoch": 2090, "lr": 2.2576847905951115e-05}
{"train_loss": 0.07493939250707626, "global_step": 186043, "epoch": 2090, "lr": 2.2576363146657675e-05}
{"train_loss": 0.05253124237060547, "global_step": 186044, "epoch": 2090, "lr": 2.2575878391050987e-05}
{"train_loss": 0.04462624713778496, "global_step": 186045, "epoch": 2090, "lr": 2.2575393639131158e-05}
{"train_loss": 0.1237260103225708, "global_step": 186046, "epoch": 2090, "lr": 2.257490889089821e-05}
{"train_loss": 0.09179019182920456, "global_step": 186047, "epoch": 2090, "lr": 2.2574424146352258e-05}
{"train_loss": 0.09428819268941879, "global_step": 186048, "epoch": 2090, "lr": 2.2573939405493335e-05}
{"train_loss": 0.07454398274421692, "global_step": 186049, "epoch": 2090, "lr": 2.257345466832149e-05}
{"train_loss": 0.12334812432527542, "global_step": 186050, "epoch": 2090, "lr": 2.2572969934836834e-05}
{"train_loss": 0.042168691754341125, "global_step": 186051, "epoch": 2090, "lr": 2.257248520503939e-05}
{"train_loss": 0.04232092946767807, "global_step": 186052, "epoch": 2090, "lr": 2.2572000478929257e-05}
{"train_loss": 0.05998198688030243, "global_step": 186053, "epoch": 2090, "lr": 2.2571515756506467e-05}
{"train_loss": 0.05463111028075218, "global_step": 186054, "epoch": 2090, "lr": 2.2571031037771116e-05}
{"train_loss": 0.0669282004237175, "global_step": 186055, "epoch": 2090, "lr": 2.2570546322723242e-05}
{"train_loss": 0.13020409643650055, "global_step": 186056, "epoch": 2090, "lr": 2.2570061611362942e-05}
{"train_loss": 0.06348567456007004, "global_step": 186057, "epoch": 2090, "lr": 2.256957690369024e-05}
{"train_loss": 0.06517217308282852, "global_step": 186058, "epoch": 2090, "lr": 2.2569092199705248e-05}
{"train_loss": 0.08284319192171097, "global_step": 186059, "epoch": 2090, "lr": 2.2568607499407988e-05}
{"train_loss": 0.054230399429798126, "global_step": 186060, "epoch": 2090, "lr": 2.2568122802798564e-05}
{"train_loss": 0.07255235314369202, "global_step": 186061, "epoch": 2090, "lr": 2.2567638109877e-05}
{"train_loss": 0.10825038701295853, "global_step": 186062, "epoch": 2090, "lr": 2.25671534206434e-05}
{"train_loss": 0.08426769077777863, "global_step": 186063, "epoch": 2090, "lr": 2.2566668735097795e-05}
{"train_loss": 0.08556883037090302, "global_step": 186064, "epoch": 2090, "lr": 2.2566184053240287e-05}
{"train_loss": 0.07929304242134094, "global_step": 186065, "epoch": 2090, "lr": 2.2565699375070902e-05}
{"train_loss": 0.13188980519771576, "global_step": 186066, "epoch": 2090, "lr": 2.2565214700589743e-05}
{"train_loss": 0.07072661817073822, "global_step": 186067, "epoch": 2090, "lr": 2.2564730029796837e-05}
{"train_loss": 0.08508028090000153, "global_step": 186068, "epoch": 2090, "lr": 2.2564245362692267e-05}
{"train_loss": 0.08889787644147873, "global_step": 186069, "epoch": 2090, "lr": 2.256376069927612e-05}
{"train_loss": 0.06953316181898117, "global_step": 186070, "epoch": 2090, "lr": 2.2563276039548425e-05}
{"train_loss": 0.05220133438706398, "global_step": 186071, "epoch": 2090, "lr": 2.256279138350928e-05}
{"train_loss": 0.14901858568191528, "global_step": 186072, "epoch": 2090, "lr": 2.256230673115871e-05}
{"train_loss": 0.14122852683067322, "global_step": 186073, "epoch": 2090, "lr": 2.256182208249683e-05}
{"train_loss": 0.06504333764314651, "global_step": 186074, "epoch": 2090, "lr": 2.2561337437523656e-05}
{"train_loss": 0.10314644873142242, "global_step": 186075, "epoch": 2090, "lr": 2.2560852796239297e-05}
{"train_loss": 0.02976790815591812, "global_step": 186076, "epoch": 2090, "lr": 2.2560368158643773e-05}
{"train_loss": 0.043615035712718964, "global_step": 186077, "epoch": 2090, "lr": 2.255988352473719e-05}
{"train_loss": 0.0431007519364357, "global_step": 186078, "epoch": 2090, "lr": 2.2559398894519584e-05}
{"train_loss": 0.06516745686531067, "global_step": 186079, "epoch": 2090, "lr": 2.255891426799105e-05}
{"train_loss": 0.06163822114467621, "global_step": 186080, "epoch": 2090, "lr": 2.255842964515162e-05}
{"train_loss": 0.05623156577348709, "global_step": 186081, "epoch": 2090, "lr": 2.2557945026001387e-05}
{"train_loss": 0.08082346618175507, "global_step": 186082, "epoch": 2090, "lr": 2.255746041054039e-05}
{"train_loss": 0.10035005956888199, "global_step": 186083, "epoch": 2090, "lr": 2.255697579876872e-05}
{"train_loss": 0.1693573296070099, "global_step": 186084, "epoch": 2090, "lr": 2.2556491190686417e-05}
{"train_loss": 0.062137171626091, "global_step": 186085, "epoch": 2090, "lr": 2.2556006586293575e-05}
{"train_loss": 0.07404562085866928, "global_step": 186086, "epoch": 2090, "lr": 2.2555521985590243e-05}
{"train_loss": 0.07698872685432434, "global_step": 186087, "epoch": 2090, "lr": 2.255503738857647e-05}
{"train_loss": 0.049185965210199356, "global_step": 186088, "epoch": 2090, "lr": 2.255455279525236e-05}
{"train_loss": 0.0976395383477211, "global_step": 186089, "epoch": 2090, "lr": 2.2554068205617933e-05}
{"train_loss": 0.06596413999795914, "global_step": 186090, "epoch": 2090, "lr": 2.2553583619673296e-05}
{"train_loss": 0.07660115510225296, "global_step": 186091, "epoch": 2090, "lr": 2.2553099037418473e-05}
{"train_loss": 0.055920399725437164, "global_step": 186092, "epoch": 2090, "lr": 2.2552614458853577e-05}
{"train_loss": 0.08228453248739243, "global_step": 186093, "epoch": 2090, "lr": 2.2552129883978623e-05}
{"train_loss": 0.03963828831911087, "global_step": 186094, "epoch": 2090, "lr": 2.2551645312793722e-05}
{"train_loss": 0.0882011279463768, "global_step": 186095, "epoch": 2090, "lr": 2.25511607452989e-05}
{"train_loss": 0.07821360975503922, "global_step": 186096, "epoch": 2090, "lr": 2.2550676181494256e-05}
{"train_loss": 0.09226624667644501, "global_step": 186097, "epoch": 2090, "lr": 2.2550191621379824e-05}
{"train_loss": 0.07607174706592988, "global_step": 186098, "epoch": 2090, "lr": 2.25497070649557e-05, "val_loss": 7.585647106170654, "train_action_mse_error": 45.720306396484375}
{"train_loss": 0.08835209906101227, "global_step": 186099, "epoch": 2091, "lr": 2.2549222512221912e-05}
{"train_loss": 0.05976640805602074, "global_step": 186100, "epoch": 2091, "lr": 2.2548737963178567e-05}
{"train_loss": 0.05705486238002777, "global_step": 186101, "epoch": 2091, "lr": 2.2548253417825693e-05}
{"train_loss": 0.056914038956165314, "global_step": 186102, "epoch": 2091, "lr": 2.254776887616339e-05}
{"train_loss": 0.08689229190349579, "global_step": 186103, "epoch": 2091, "lr": 2.2547284338191687e-05}
{"train_loss": 0.07072979211807251, "global_step": 186104, "epoch": 2091, "lr": 2.254679980391067e-05}
{"train_loss": 0.048388995230197906, "global_step": 186105, "epoch": 2091, "lr": 2.254631527332041e-05}
{"train_loss": 0.1969284564256668, "global_step": 186106, "epoch": 2091, "lr": 2.2545830746420953e-05}
{"train_loss": 0.0492505244910717, "global_step": 186107, "epoch": 2091, "lr": 2.2545346223212387e-05}
{"train_loss": 0.044459305703639984, "global_step": 186108, "epoch": 2091, "lr": 2.2544861703694754e-05}
{"train_loss": 0.08528149127960205, "global_step": 186109, "epoch": 2091, "lr": 2.2544377187868138e-05}
{"train_loss": 0.05250217393040657, "global_step": 186110, "epoch": 2091, "lr": 2.2543892675732586e-05}
{"train_loss": 0.0899784117937088, "global_step": 186111, "epoch": 2091, "lr": 2.2543408167288187e-05}
{"train_loss": 0.06800909340381622, "global_step": 186112, "epoch": 2091, "lr": 2.2542923662534975e-05}
{"train_loss": 0.10383102297782898, "global_step": 186113, "epoch": 2091, "lr": 2.254243916147305e-05}
{"train_loss": 0.12409713119268417, "global_step": 186114, "epoch": 2091, "lr": 2.2541954664102443e-05}
{"train_loss": 0.08479593694210052, "global_step": 186115, "epoch": 2091, "lr": 2.254147017042325e-05}
{"train_loss": 0.08717919141054153, "global_step": 186116, "epoch": 2091, "lr": 2.2540985680435504e-05}
{"train_loss": 0.07255005836486816, "global_step": 186117, "epoch": 2091, "lr": 2.2540501194139313e-05}
{"train_loss": 0.06889083981513977, "global_step": 186118, "epoch": 2091, "lr": 2.2540016711534694e-05}
{"train_loss": 0.06246679648756981, "global_step": 186119, "epoch": 2091, "lr": 2.253953223262175e-05}
{"train_loss": 0.05132758989930153, "global_step": 186120, "epoch": 2091, "lr": 2.2539047757400518e-05}
{"train_loss": 0.11245473474264145, "global_step": 186121, "epoch": 2091, "lr": 2.253856328587109e-05}
{"train_loss": 0.08022280037403107, "global_step": 186122, "epoch": 2091, "lr": 2.2538078818033504e-05}
{"train_loss": 0.08222602307796478, "global_step": 186123, "epoch": 2091, "lr": 2.2537594353887854e-05}
{"train_loss": 0.08143424242734909, "global_step": 186124, "epoch": 2091, "lr": 2.253710989343417e-05}
{"train_loss": 0.05904684215784073, "global_step": 186125, "epoch": 2091, "lr": 2.253662543667256e-05}
{"train_loss": 0.043185822665691376, "global_step": 186126, "epoch": 2091, "lr": 2.253614098360306e-05}
{"train_loss": 0.12276887893676758, "global_step": 186127, "epoch": 2091, "lr": 2.253565653422573e-05}
{"train_loss": 0.07676181942224503, "global_step": 186128, "epoch": 2091, "lr": 2.2535172088540656e-05}
{"train_loss": 0.07888898998498917, "global_step": 186129, "epoch": 2091, "lr": 2.253468764654788e-05}
{"train_loss": 0.08325762301683426, "global_step": 186130, "epoch": 2091, "lr": 2.2534203208247502e-05}
{"train_loss": 0.05776194483041763, "global_step": 186131, "epoch": 2091, "lr": 2.2533718773639538e-05}
{"train_loss": 0.05792386457324028, "global_step": 186132, "epoch": 2091, "lr": 2.253323434272411e-05}
{"train_loss": 0.06100156158208847, "global_step": 186133, "epoch": 2091, "lr": 2.253274991550123e-05}
{"train_loss": 0.05897346884012222, "global_step": 186134, "epoch": 2091, "lr": 2.2532265491971005e-05}
{"train_loss": 0.11598221212625504, "global_step": 186135, "epoch": 2091, "lr": 2.2531781072133463e-05}
{"train_loss": 0.07436580210924149, "global_step": 186136, "epoch": 2091, "lr": 2.253129665598871e-05}
{"train_loss": 0.061434339731931686, "global_step": 186137, "epoch": 2091, "lr": 2.253081224353677e-05}
{"train_loss": 0.03769635036587715, "global_step": 186138, "epoch": 2091, "lr": 2.2530327834777743e-05}
{"train_loss": 0.045647069811820984, "global_step": 186139, "epoch": 2091, "lr": 2.2529843429711662e-05}
{"train_loss": 0.04127217084169388, "global_step": 186140, "epoch": 2091, "lr": 2.2529359028338608e-05}
{"train_loss": 0.10381826013326645, "global_step": 186141, "epoch": 2091, "lr": 2.2528874630658665e-05}
{"train_loss": 0.04977094754576683, "global_step": 186142, "epoch": 2091, "lr": 2.2528390236671863e-05}
{"train_loss": 0.08874078840017319, "global_step": 186143, "epoch": 2091, "lr": 2.25279058463783e-05}
{"train_loss": 0.06789667159318924, "global_step": 186144, "epoch": 2091, "lr": 2.252742145977801e-05}
{"train_loss": 0.0745396539568901, "global_step": 186145, "epoch": 2091, "lr": 2.252693707687109e-05}
{"train_loss": 0.04871082678437233, "global_step": 186146, "epoch": 2091, "lr": 2.2526452697657573e-05}
{"train_loss": 0.10055918991565704, "global_step": 186147, "epoch": 2091, "lr": 2.2525968322137553e-05}
{"train_loss": 0.0787566676735878, "global_step": 186148, "epoch": 2091, "lr": 2.2525483950311065e-05}
{"train_loss": 0.0924689918756485, "global_step": 186149, "epoch": 2091, "lr": 2.2524999582178212e-05}
{"train_loss": 0.1012297198176384, "global_step": 186150, "epoch": 2091, "lr": 2.2524515217739016e-05}
{"train_loss": 0.035425640642642975, "global_step": 186151, "epoch": 2091, "lr": 2.252403085699359e-05}
{"train_loss": 0.04495507851243019, "global_step": 186152, "epoch": 2091, "lr": 2.2523546499941944e-05}
{"train_loss": 0.07917279005050659, "global_step": 186153, "epoch": 2091, "lr": 2.2523062146584195e-05}
{"train_loss": 0.07340610772371292, "global_step": 186154, "epoch": 2091, "lr": 2.2522577796920368e-05}
{"train_loss": 0.1404602974653244, "global_step": 186155, "epoch": 2091, "lr": 2.2522093450950566e-05}
{"train_loss": 0.04316449910402298, "global_step": 186156, "epoch": 2091, "lr": 2.2521609108674814e-05}
{"train_loss": 0.05005927383899689, "global_step": 186157, "epoch": 2091, "lr": 2.252112477009321e-05}
{"train_loss": 0.08666698634624481, "global_step": 186158, "epoch": 2091, "lr": 2.252064043520579e-05}
{"train_loss": 0.032217785716056824, "global_step": 186159, "epoch": 2091, "lr": 2.252015610401266e-05}
{"train_loss": 0.08864349871873856, "global_step": 186160, "epoch": 2091, "lr": 2.2519671776513834e-05}
{"train_loss": 0.0868171975016594, "global_step": 186161, "epoch": 2091, "lr": 2.2519187452709424e-05}
{"train_loss": 0.07974762469530106, "global_step": 186162, "epoch": 2091, "lr": 2.2518703132599457e-05}
{"train_loss": 0.04321436583995819, "global_step": 186163, "epoch": 2091, "lr": 2.2518218816184034e-05}
{"train_loss": 0.03661352023482323, "global_step": 186164, "epoch": 2091, "lr": 2.25177345034632e-05}
{"train_loss": 0.061582762748003006, "global_step": 186165, "epoch": 2091, "lr": 2.2517250194437e-05}
{"train_loss": 0.07295461744070053, "global_step": 186166, "epoch": 2091, "lr": 2.2516765889105547e-05}
{"train_loss": 0.06055586785078049, "global_step": 186167, "epoch": 2091, "lr": 2.2516281587468857e-05}
{"train_loss": 0.06980552524328232, "global_step": 186168, "epoch": 2091, "lr": 2.2515797289527036e-05}
{"train_loss": 0.0730886310338974, "global_step": 186169, "epoch": 2091, "lr": 2.2515312995280112e-05}
{"train_loss": 0.1224438026547432, "global_step": 186170, "epoch": 2091, "lr": 2.2514828704728187e-05}
{"train_loss": 0.0376661941409111, "global_step": 186171, "epoch": 2091, "lr": 2.2514344417871293e-05}
{"train_loss": 0.08206532150506973, "global_step": 186172, "epoch": 2091, "lr": 2.2513860134709523e-05}
{"train_loss": 0.07054217159748077, "global_step": 186173, "epoch": 2091, "lr": 2.2513375855242918e-05}
{"train_loss": 0.07446444779634476, "global_step": 186174, "epoch": 2091, "lr": 2.2512891579471575e-05}
{"train_loss": 0.0773690864443779, "global_step": 186175, "epoch": 2091, "lr": 2.251240730739551e-05}
{"train_loss": 0.05851589888334274, "global_step": 186176, "epoch": 2091, "lr": 2.2511923039014842e-05}
{"train_loss": 0.099593386054039, "global_step": 186177, "epoch": 2091, "lr": 2.2511438774329592e-05}
{"train_loss": 0.03837006166577339, "global_step": 186178, "epoch": 2091, "lr": 2.2510954513339845e-05}
{"train_loss": 0.07685588300228119, "global_step": 186179, "epoch": 2091, "lr": 2.251047025604568e-05}
{"train_loss": 0.10208070278167725, "global_step": 186180, "epoch": 2091, "lr": 2.250998600244713e-05}
{"train_loss": 0.07599514722824097, "global_step": 186181, "epoch": 2091, "lr": 2.25095017525443e-05}
{"train_loss": 0.06546427309513092, "global_step": 186182, "epoch": 2091, "lr": 2.250901750633721e-05}
{"train_loss": 0.10276880860328674, "global_step": 186183, "epoch": 2091, "lr": 2.2508533263825965e-05}
{"train_loss": 0.06853246688842773, "global_step": 186184, "epoch": 2091, "lr": 2.2508049025010595e-05}
{"train_loss": 0.05294390022754669, "global_step": 186185, "epoch": 2091, "lr": 2.2507564789891206e-05}
{"train_loss": 0.06129617616534233, "global_step": 186186, "epoch": 2091, "lr": 2.2507080558467815e-05}
{"train_loss": 0.07362244140063778, "global_step": 186187, "epoch": 2091, "lr": 2.2506596330740533e-05, "val_loss": 7.619299411773682}
{"train_loss": 0.13999034464359283, "global_step": 186188, "epoch": 2092, "lr": 2.2506112106709388e-05}
{"train_loss": 0.06942180544137955, "global_step": 186189, "epoch": 2092, "lr": 2.2505627886374474e-05}
{"train_loss": 0.08229055255651474, "global_step": 186190, "epoch": 2092, "lr": 2.250514366973583e-05}
{"train_loss": 0.05921579897403717, "global_step": 186191, "epoch": 2092, "lr": 2.2504659456793548e-05}
{"train_loss": 0.048795029520988464, "global_step": 186192, "epoch": 2092, "lr": 2.250417524754766e-05}
{"train_loss": 0.05303570628166199, "global_step": 186193, "epoch": 2092, "lr": 2.2503691041998277e-05}
{"train_loss": 0.06145748496055603, "global_step": 186194, "epoch": 2092, "lr": 2.2503206840145412e-05}
{"train_loss": 0.053943902254104614, "global_step": 186195, "epoch": 2092, "lr": 2.250272264198916e-05}
{"train_loss": 0.0345606654882431, "global_step": 186196, "epoch": 2092, "lr": 2.2502238447529593e-05}
{"train_loss": 0.12503603100776672, "global_step": 186197, "epoch": 2092, "lr": 2.2501754256766772e-05}
{"train_loss": 0.05338846147060394, "global_step": 186198, "epoch": 2092, "lr": 2.2501270069700737e-05}
{"train_loss": 0.05857524648308754, "global_step": 186199, "epoch": 2092, "lr": 2.250078588633159e-05}
{"train_loss": 0.023841500282287598, "global_step": 186200, "epoch": 2092, "lr": 2.2500301706659355e-05}
{"train_loss": 0.09850272536277771, "global_step": 186201, "epoch": 2092, "lr": 2.2499817530684137e-05}
{"train_loss": 0.0765397846698761, "global_step": 186202, "epoch": 2092, "lr": 2.2499333358405983e-05}
{"train_loss": 0.06808988749980927, "global_step": 186203, "epoch": 2092, "lr": 2.2498849189824945e-05}
{"train_loss": 0.06519682705402374, "global_step": 186204, "epoch": 2092, "lr": 2.2498365024941114e-05}
{"train_loss": 0.09469448029994965, "global_step": 186205, "epoch": 2092, "lr": 2.249788086375453e-05}
{"train_loss": 0.09446081519126892, "global_step": 186206, "epoch": 2092, "lr": 2.2497396706265282e-05}
{"train_loss": 0.0999627634882927, "global_step": 186207, "epoch": 2092, "lr": 2.2496912552473416e-05}
{"train_loss": 0.07279141247272491, "global_step": 186208, "epoch": 2092, "lr": 2.2496428402379017e-05}
{"train_loss": 0.07579519599676132, "global_step": 186209, "epoch": 2092, "lr": 2.2495944255982115e-05}
{"train_loss": 0.05816301703453064, "global_step": 186210, "epoch": 2092, "lr": 2.2495460113282822e-05}
{"train_loss": 0.08143579214811325, "global_step": 186211, "epoch": 2092, "lr": 2.2494975974281164e-05}
{"train_loss": 0.1237497329711914, "global_step": 186212, "epoch": 2092, "lr": 2.2494491838977232e-05}
{"train_loss": 0.1043132022023201, "global_step": 186213, "epoch": 2092, "lr": 2.2494007707371066e-05}
{"train_loss": 0.07041366398334503, "global_step": 186214, "epoch": 2092, "lr": 2.2493523579462744e-05}
{"train_loss": 0.06396929919719696, "global_step": 186215, "epoch": 2092, "lr": 2.249303945525235e-05}
{"train_loss": 0.03805277869105339, "global_step": 186216, "epoch": 2092, "lr": 2.249255533473992e-05}
{"train_loss": 0.04955687001347542, "global_step": 186217, "epoch": 2092, "lr": 2.2492071217925543e-05}
{"train_loss": 0.04454308748245239, "global_step": 186218, "epoch": 2092, "lr": 2.2491587104809252e-05}
{"train_loss": 0.06945963203907013, "global_step": 186219, "epoch": 2092, "lr": 2.2491102995391155e-05}
{"train_loss": 0.05741151422262192, "global_step": 186220, "epoch": 2092, "lr": 2.249061888967127e-05}
{"train_loss": 0.1687900424003601, "global_step": 186221, "epoch": 2092, "lr": 2.2490134787649707e-05}
{"train_loss": 0.04527357220649719, "global_step": 186222, "epoch": 2092, "lr": 2.2489650689326497e-05}
{"train_loss": 0.0593392439186573, "global_step": 186223, "epoch": 2092, "lr": 2.2489166594701726e-05}
{"train_loss": 0.10877741128206253, "global_step": 186224, "epoch": 2092, "lr": 2.2488682503775442e-05}
{"train_loss": 0.019533321261405945, "global_step": 186225, "epoch": 2092, "lr": 2.2488198416547734e-05}
{"train_loss": 0.039090149104595184, "global_step": 186226, "epoch": 2092, "lr": 2.2487714333018633e-05}
{"train_loss": 0.031272903084754944, "global_step": 186227, "epoch": 2092, "lr": 2.2487230253188247e-05}
{"train_loss": 0.10412469506263733, "global_step": 186228, "epoch": 2092, "lr": 2.2486746177056593e-05}
{"train_loss": 0.06822295486927032, "global_step": 186229, "epoch": 2092, "lr": 2.248626210462378e-05}
{"train_loss": 0.0687679797410965, "global_step": 186230, "epoch": 2092, "lr": 2.2485778035889837e-05}
{"train_loss": 0.08219973742961884, "global_step": 186231, "epoch": 2092, "lr": 2.248529397085485e-05}
{"train_loss": 0.07382501661777496, "global_step": 186232, "epoch": 2092, "lr": 2.248480990951889e-05}
{"train_loss": 0.052493948489427567, "global_step": 186233, "epoch": 2092, "lr": 2.2484325851882e-05}
{"train_loss": 0.14545078575611115, "global_step": 186234, "epoch": 2092, "lr": 2.2483841797944273e-05}
{"train_loss": 0.08325323462486267, "global_step": 186235, "epoch": 2092, "lr": 2.2483357747705735e-05}
{"train_loss": 0.09218473732471466, "global_step": 186236, "epoch": 2092, "lr": 2.2482873701166495e-05}
{"train_loss": 0.044723547995090485, "global_step": 186237, "epoch": 2092, "lr": 2.2482389658326604e-05}
{"train_loss": 0.037078771740198135, "global_step": 186238, "epoch": 2092, "lr": 2.2481905619186095e-05}
{"train_loss": 0.029097499325871468, "global_step": 186239, "epoch": 2092, "lr": 2.2481421583745077e-05}
{"train_loss": 0.07956460118293762, "global_step": 186240, "epoch": 2092, "lr": 2.2480937552003584e-05}
{"train_loss": 0.08697180449962616, "global_step": 186241, "epoch": 2092, "lr": 2.2480453523961707e-05}
{"train_loss": 0.09644942730665207, "global_step": 186242, "epoch": 2092, "lr": 2.2479969499619497e-05}
{"train_loss": 0.042585887014865875, "global_step": 186243, "epoch": 2092, "lr": 2.2479485478977002e-05}
{"train_loss": 0.06180308759212494, "global_step": 186244, "epoch": 2092, "lr": 2.2479001462034327e-05}
{"train_loss": 0.06520852446556091, "global_step": 186245, "epoch": 2092, "lr": 2.247851744879149e-05}
{"train_loss": 0.10777169466018677, "global_step": 186246, "epoch": 2092, "lr": 2.2478033439248604e-05}
{"train_loss": 0.09379690140485764, "global_step": 186247, "epoch": 2092, "lr": 2.2477549433405692e-05}
{"train_loss": 0.11120457947254181, "global_step": 186248, "epoch": 2092, "lr": 2.2477065431262857e-05}
{"train_loss": 0.07649488747119904, "global_step": 186249, "epoch": 2092, "lr": 2.2476581432820127e-05}
{"train_loss": 0.0672498568892479, "global_step": 186250, "epoch": 2092, "lr": 2.247609743807758e-05}
{"train_loss": 0.05431525036692619, "global_step": 186251, "epoch": 2092, "lr": 2.247561344703531e-05}
{"train_loss": 0.07440923154354095, "global_step": 186252, "epoch": 2092, "lr": 2.247512945969334e-05}
{"train_loss": 0.04425547271966934, "global_step": 186253, "epoch": 2092, "lr": 2.2474645476051764e-05}
{"train_loss": 0.10255119949579239, "global_step": 186254, "epoch": 2092, "lr": 2.2474161496110625e-05}
{"train_loss": 0.07590024918317795, "global_step": 186255, "epoch": 2092, "lr": 2.2473677519870017e-05}
{"train_loss": 0.07046803086996078, "global_step": 186256, "epoch": 2092, "lr": 2.2473193547329967e-05}
{"train_loss": 0.05949658527970314, "global_step": 186257, "epoch": 2092, "lr": 2.2472709578490588e-05}
{"train_loss": 0.06998931616544724, "global_step": 186258, "epoch": 2092, "lr": 2.2472225613351888e-05}
{"train_loss": 0.09569723904132843, "global_step": 186259, "epoch": 2092, "lr": 2.2471741651913984e-05}
{"train_loss": 0.06426866352558136, "global_step": 186260, "epoch": 2092, "lr": 2.24712576941769e-05}
{"train_loss": 0.054492153227329254, "global_step": 186261, "epoch": 2092, "lr": 2.2470773740140738e-05}
{"train_loss": 0.04342469945549965, "global_step": 186262, "epoch": 2092, "lr": 2.247028978980553e-05}
{"train_loss": 0.033653438091278076, "global_step": 186263, "epoch": 2092, "lr": 2.246980584317137e-05}
{"train_loss": 0.1660982370376587, "global_step": 186264, "epoch": 2092, "lr": 2.2469321900238294e-05}
{"train_loss": 0.06436346471309662, "global_step": 186265, "epoch": 2092, "lr": 2.24688379610064e-05}
{"train_loss": 0.0785684660077095, "global_step": 186266, "epoch": 2092, "lr": 2.246835402547572e-05}
{"train_loss": 0.06633172929286957, "global_step": 186267, "epoch": 2092, "lr": 2.2467870093646326e-05}
{"train_loss": 0.09680517017841339, "global_step": 186268, "epoch": 2092, "lr": 2.246738616551831e-05}
{"train_loss": 0.0750754326581955, "global_step": 186269, "epoch": 2092, "lr": 2.2466902241091708e-05}
{"train_loss": 0.03836461901664734, "global_step": 186270, "epoch": 2092, "lr": 2.2466418320366612e-05}
{"train_loss": 0.059070322662591934, "global_step": 186271, "epoch": 2092, "lr": 2.2465934403343047e-05}
{"train_loss": 0.1048102155327797, "global_step": 186272, "epoch": 2092, "lr": 2.2465450490021123e-05}
{"train_loss": 0.054243359714746475, "global_step": 186273, "epoch": 2092, "lr": 2.2464966580400866e-05}
{"train_loss": 0.08450380712747574, "global_step": 186274, "epoch": 2092, "lr": 2.2464482674482374e-05}
{"train_loss": 0.09285195916891098, "global_step": 186275, "epoch": 2092, "lr": 2.24639987722657e-05}
{"train_loss": 0.07292048672862937, "global_step": 186276, "epoch": 2092, "lr": 2.2463514873750885e-05, "val_loss": 7.610358238220215}
{"train_loss": 0.06448784470558167, "global_step": 186277, "epoch": 2093, "lr": 2.246303097893803e-05}
{"train_loss": 0.05542544648051262, "global_step": 186278, "epoch": 2093, "lr": 2.2462547087827175e-05}
{"train_loss": 0.06114593893289566, "global_step": 186279, "epoch": 2093, "lr": 2.246206320041841e-05}
{"train_loss": 0.1131317988038063, "global_step": 186280, "epoch": 2093, "lr": 2.2461579316711783e-05}
{"train_loss": 0.04472164809703827, "global_step": 186281, "epoch": 2093, "lr": 2.2461095436707342e-05}
{"train_loss": 0.07342979311943054, "global_step": 186282, "epoch": 2093, "lr": 2.2460611560405194e-05}
{"train_loss": 0.06447188556194305, "global_step": 186283, "epoch": 2093, "lr": 2.2460127687805355e-05}
{"train_loss": 0.08218812942504883, "global_step": 186284, "epoch": 2093, "lr": 2.2459643818907945e-05}
{"train_loss": 0.054215919226408005, "global_step": 186285, "epoch": 2093, "lr": 2.245915995371297e-05}
{"train_loss": 0.05570032820105553, "global_step": 186286, "epoch": 2093, "lr": 2.2458676092220533e-05}
{"train_loss": 0.05709225684404373, "global_step": 186287, "epoch": 2093, "lr": 2.245819223443071e-05}
{"train_loss": 0.07993490993976593, "global_step": 186288, "epoch": 2093, "lr": 2.245770838034352e-05}
{"train_loss": 0.08344680815935135, "global_step": 186289, "epoch": 2093, "lr": 2.2457224529959087e-05}
{"train_loss": 0.09070144593715668, "global_step": 186290, "epoch": 2093, "lr": 2.2456740683277417e-05}
{"train_loss": 0.07061424106359482, "global_step": 186291, "epoch": 2093, "lr": 2.245625684029863e-05}
{"train_loss": 0.06256071478128433, "global_step": 186292, "epoch": 2093, "lr": 2.2455773001022734e-05}
{"train_loss": 0.034713394939899445, "global_step": 186293, "epoch": 2093, "lr": 2.245528916544985e-05}
{"train_loss": 0.13449670374393463, "global_step": 186294, "epoch": 2093, "lr": 2.2454805333579993e-05}
{"train_loss": 0.09586929529905319, "global_step": 186295, "epoch": 2093, "lr": 2.2454321505413274e-05}
{"train_loss": 0.023207714781165123, "global_step": 186296, "epoch": 2093, "lr": 2.245383768094971e-05}
{"train_loss": 0.1575089991092682, "global_step": 186297, "epoch": 2093, "lr": 2.2453353860189418e-05}
{"train_loss": 0.07518064975738525, "global_step": 186298, "epoch": 2093, "lr": 2.2452870043132416e-05}
{"train_loss": 0.08786700665950775, "global_step": 186299, "epoch": 2093, "lr": 2.245238622977881e-05}
{"train_loss": 0.05199044942855835, "global_step": 186300, "epoch": 2093, "lr": 2.2451902420128623e-05}
{"train_loss": 0.04025861993432045, "global_step": 186301, "epoch": 2093, "lr": 2.2451418614181963e-05}
{"train_loss": 0.09820473194122314, "global_step": 186302, "epoch": 2093, "lr": 2.2450934811938856e-05}
{"train_loss": 0.09788283705711365, "global_step": 186303, "epoch": 2093, "lr": 2.2450451013399405e-05}
{"train_loss": 0.05820667743682861, "global_step": 186304, "epoch": 2093, "lr": 2.2449967218563635e-05}
{"train_loss": 0.08324847370386124, "global_step": 186305, "epoch": 2093, "lr": 2.244948342743163e-05}
{"train_loss": 0.06029766798019409, "global_step": 186306, "epoch": 2093, "lr": 2.244899964000347e-05}
{"train_loss": 0.05618108808994293, "global_step": 186307, "epoch": 2093, "lr": 2.2448515856279194e-05}
{"train_loss": 0.10177768766880035, "global_step": 186308, "epoch": 2093, "lr": 2.2448032076258895e-05}
{"train_loss": 0.0339675098657608, "global_step": 186309, "epoch": 2093, "lr": 2.2447548299942606e-05}
{"train_loss": 0.050583455711603165, "global_step": 186310, "epoch": 2093, "lr": 2.2447064527330426e-05}
{"train_loss": 0.08869940787553787, "global_step": 186311, "epoch": 2093, "lr": 2.2446580758422386e-05}
{"train_loss": 0.07079831510782242, "global_step": 186312, "epoch": 2093, "lr": 2.244609699321858e-05}
{"train_loss": 0.1219598576426506, "global_step": 186313, "epoch": 2093, "lr": 2.2445613231719066e-05}
{"train_loss": 0.06407427042722702, "global_step": 186314, "epoch": 2093, "lr": 2.244512947392388e-05}
{"train_loss": 0.1260969638824463, "global_step": 186315, "epoch": 2093, "lr": 2.2444645719833134e-05}
{"train_loss": 0.052463188767433167, "global_step": 186316, "epoch": 2093, "lr": 2.2444161969446848e-05}
{"train_loss": 0.0664057731628418, "global_step": 186317, "epoch": 2093, "lr": 2.244367822276513e-05}
{"train_loss": 0.01856148988008499, "global_step": 186318, "epoch": 2093, "lr": 2.2443194479788022e-05}
{"train_loss": 0.09620910882949829, "global_step": 186319, "epoch": 2093, "lr": 2.244271074051557e-05}
{"train_loss": 0.04121807590126991, "global_step": 186320, "epoch": 2093, "lr": 2.244222700494788e-05}
{"train_loss": 0.034183401614427567, "global_step": 186321, "epoch": 2093, "lr": 2.244174327308498e-05}
{"train_loss": 0.06003441661596298, "global_step": 186322, "epoch": 2093, "lr": 2.2441259544926947e-05}
{"train_loss": 0.11038094758987427, "global_step": 186323, "epoch": 2093, "lr": 2.2440775820473875e-05}
{"train_loss": 0.08017276972532272, "global_step": 186324, "epoch": 2093, "lr": 2.244029209972578e-05}
{"train_loss": 0.07916011661291122, "global_step": 186325, "epoch": 2093, "lr": 2.2439808382682774e-05}
{"train_loss": 0.11277278512716293, "global_step": 186326, "epoch": 2093, "lr": 2.243932466934488e-05}
{"train_loss": 0.03853520378470421, "global_step": 186327, "epoch": 2093, "lr": 2.2438840959712204e-05}
{"train_loss": 0.09195093810558319, "global_step": 186328, "epoch": 2093, "lr": 2.243835725378477e-05}
{"train_loss": 0.09156560897827148, "global_step": 186329, "epoch": 2093, "lr": 2.243787355156268e-05}
{"train_loss": 0.060055166482925415, "global_step": 186330, "epoch": 2093, "lr": 2.2437389853045966e-05}
{"train_loss": 0.028360316529870033, "global_step": 186331, "epoch": 2093, "lr": 2.2436906158234728e-05}
{"train_loss": 0.08622685819864273, "global_step": 186332, "epoch": 2093, "lr": 2.243642246712899e-05}
{"train_loss": 0.07416810095310211, "global_step": 186333, "epoch": 2093, "lr": 2.243593877972886e-05}
{"train_loss": 0.0811227411031723, "global_step": 186334, "epoch": 2093, "lr": 2.2435455096034362e-05}
{"train_loss": 0.13804319500923157, "global_step": 186335, "epoch": 2093, "lr": 2.2434971416045604e-05}
{"train_loss": 0.06165648624300957, "global_step": 186336, "epoch": 2093, "lr": 2.2434487739762606e-05}
{"train_loss": 0.04436855763196945, "global_step": 186337, "epoch": 2093, "lr": 2.243400406718547e-05}
{"train_loss": 0.04753981530666351, "global_step": 186338, "epoch": 2093, "lr": 2.2433520398314234e-05}
{"train_loss": 0.0570867583155632, "global_step": 186339, "epoch": 2093, "lr": 2.2433036733148987e-05}
{"train_loss": 0.09628640860319138, "global_step": 186340, "epoch": 2093, "lr": 2.2432553071689773e-05}
{"train_loss": 0.03533966466784477, "global_step": 186341, "epoch": 2093, "lr": 2.2432069413936656e-05}
{"train_loss": 0.06233217194676399, "global_step": 186342, "epoch": 2093, "lr": 2.243158575988974e-05}
{"train_loss": 0.1226031482219696, "global_step": 186343, "epoch": 2093, "lr": 2.2431102109549036e-05}
{"train_loss": 0.06929142028093338, "global_step": 186344, "epoch": 2093, "lr": 2.2430618462914654e-05}
{"train_loss": 0.07222091406583786, "global_step": 186345, "epoch": 2093, "lr": 2.243013481998662e-05}
{"train_loss": 0.08147643506526947, "global_step": 186346, "epoch": 2093, "lr": 2.242965118076504e-05}
{"train_loss": 0.04757377877831459, "global_step": 186347, "epoch": 2093, "lr": 2.2429167545249936e-05}
{"train_loss": 0.05233941227197647, "global_step": 186348, "epoch": 2093, "lr": 2.2428683913441412e-05}
{"train_loss": 0.1168069988489151, "global_step": 186349, "epoch": 2093, "lr": 2.24282002853395e-05}
{"train_loss": 0.0631430447101593, "global_step": 186350, "epoch": 2093, "lr": 2.2427716660944297e-05}
{"train_loss": 0.0877315029501915, "global_step": 186351, "epoch": 2093, "lr": 2.2427233040255835e-05}
{"train_loss": 0.11543698608875275, "global_step": 186352, "epoch": 2093, "lr": 2.2426749423274213e-05}
{"train_loss": 0.0529354065656662, "global_step": 186353, "epoch": 2093, "lr": 2.2426265809999474e-05}
{"train_loss": 0.07221746444702148, "global_step": 186354, "epoch": 2093, "lr": 2.2425782200431677e-05}
{"train_loss": 0.039510577917099, "global_step": 186355, "epoch": 2093, "lr": 2.242529859457091e-05}
{"train_loss": 0.05629991739988327, "global_step": 186356, "epoch": 2093, "lr": 2.242481499241721e-05}
{"train_loss": 0.048765793442726135, "global_step": 186357, "epoch": 2093, "lr": 2.2424331393970676e-05}
{"train_loss": 0.07351983338594437, "global_step": 186358, "epoch": 2093, "lr": 2.2423847799231333e-05}
{"train_loss": 0.09886693954467773, "global_step": 186359, "epoch": 2093, "lr": 2.2423364208199287e-05}
{"train_loss": 0.07297927141189575, "global_step": 186360, "epoch": 2093, "lr": 2.242288062087457e-05}
{"train_loss": 0.10979148745536804, "global_step": 186361, "epoch": 2093, "lr": 2.2422397037257275e-05}
{"train_loss": 0.06014721095561981, "global_step": 186362, "epoch": 2093, "lr": 2.2421913457347432e-05}
{"train_loss": 0.1541537344455719, "global_step": 186363, "epoch": 2093, "lr": 2.242142988114515e-05}
{"train_loss": 0.10285525768995285, "global_step": 186364, "epoch": 2093, "lr": 2.2420946308650447e-05}
{"train_loss": 0.0744513003678804, "global_step": 186365, "epoch": 2093, "lr": 2.2420462739863434e-05, "val_loss": 7.544370174407959}
{"train_loss": 0.09404949843883514, "global_step": 186366, "epoch": 2094, "lr": 2.2419979174784135e-05}
{"train_loss": 0.0916585624217987, "global_step": 186367, "epoch": 2094, "lr": 2.241949561341265e-05}
{"train_loss": 0.05095569044351578, "global_step": 186368, "epoch": 2094, "lr": 2.241901205574901e-05}
{"train_loss": 0.08958176523447037, "global_step": 186369, "epoch": 2094, "lr": 2.2418528501793318e-05}
{"train_loss": 0.049532532691955566, "global_step": 186370, "epoch": 2094, "lr": 2.2418044951545596e-05}
{"train_loss": 0.08446241170167923, "global_step": 186371, "epoch": 2094, "lr": 2.2417561405005955e-05}
{"train_loss": 0.03504477068781853, "global_step": 186372, "epoch": 2094, "lr": 2.2417077862174413e-05}
{"train_loss": 0.11953817307949066, "global_step": 186373, "epoch": 2094, "lr": 2.2416594323051083e-05}
{"train_loss": 0.07500740140676498, "global_step": 186374, "epoch": 2094, "lr": 2.2416110787635986e-05}
{"train_loss": 0.08496666699647903, "global_step": 186375, "epoch": 2094, "lr": 2.2415627255929223e-05}
{"train_loss": 0.06685328483581543, "global_step": 186376, "epoch": 2094, "lr": 2.241514372793082e-05}
{"train_loss": 0.08015763014554977, "global_step": 186377, "epoch": 2094, "lr": 2.241466020364087e-05}
{"train_loss": 0.07536774128675461, "global_step": 186378, "epoch": 2094, "lr": 2.241417668305945e-05}
{"train_loss": 0.13653339445590973, "global_step": 186379, "epoch": 2094, "lr": 2.2413693166186593e-05}
{"train_loss": 0.07937003672122955, "global_step": 186380, "epoch": 2094, "lr": 2.2413209653022394e-05}
{"train_loss": 0.06303001195192337, "global_step": 186381, "epoch": 2094, "lr": 2.241272614356688e-05}
{"train_loss": 0.02629190869629383, "global_step": 186382, "epoch": 2094, "lr": 2.2412242637820163e-05}
{"train_loss": 0.15217961370944977, "global_step": 186383, "epoch": 2094, "lr": 2.2411759135782267e-05}
{"train_loss": 0.05289090424776077, "global_step": 186384, "epoch": 2094, "lr": 2.2411275637453288e-05}
{"train_loss": 0.08559607714414597, "global_step": 186385, "epoch": 2094, "lr": 2.241079214283326e-05}
{"train_loss": 0.0770169049501419, "global_step": 186386, "epoch": 2094, "lr": 2.241030865192228e-05}
{"train_loss": 0.08988329023122787, "global_step": 186387, "epoch": 2094, "lr": 2.240982516472038e-05}
{"train_loss": 0.0597120001912117, "global_step": 186388, "epoch": 2094, "lr": 2.2409341681227662e-05}
{"train_loss": 0.04837920144200325, "global_step": 186389, "epoch": 2094, "lr": 2.240885820144416e-05}
{"train_loss": 0.05291955545544624, "global_step": 186390, "epoch": 2094, "lr": 2.2408374725369962e-05}
{"train_loss": 0.11053087562322617, "global_step": 186391, "epoch": 2094, "lr": 2.240789125300512e-05}
{"train_loss": 0.08969438821077347, "global_step": 186392, "epoch": 2094, "lr": 2.2407407784349688e-05}
{"train_loss": 0.09008267521858215, "global_step": 186393, "epoch": 2094, "lr": 2.240692431940376e-05}
{"train_loss": 0.04055352881550789, "global_step": 186394, "epoch": 2094, "lr": 2.2406440858167367e-05}
{"train_loss": 0.06954873353242874, "global_step": 186395, "epoch": 2094, "lr": 2.240595740064061e-05}
{"train_loss": 0.05355700105428696, "global_step": 186396, "epoch": 2094, "lr": 2.2405473946823512e-05}
{"train_loss": 0.1108536571264267, "global_step": 186397, "epoch": 2094, "lr": 2.240499049671619e-05}
{"train_loss": 0.10916738957166672, "global_step": 186398, "epoch": 2094, "lr": 2.240450705031865e-05}
{"train_loss": 0.06098277494311333, "global_step": 186399, "epoch": 2094, "lr": 2.2404023607631015e-05}
{"train_loss": 0.06391429156064987, "global_step": 186400, "epoch": 2094, "lr": 2.2403540168653302e-05}
{"train_loss": 0.0616726279258728, "global_step": 186401, "epoch": 2094, "lr": 2.2403056733385614e-05}
{"train_loss": 0.09908338636159897, "global_step": 186402, "epoch": 2094, "lr": 2.2402573301827984e-05}
{"train_loss": 0.02418587915599346, "global_step": 186403, "epoch": 2094, "lr": 2.2402089873980504e-05}
{"train_loss": 0.12725301086902618, "global_step": 186404, "epoch": 2094, "lr": 2.2401606449843215e-05}
{"train_loss": 0.1103907972574234, "global_step": 186405, "epoch": 2094, "lr": 2.2401123029416205e-05}
{"train_loss": 0.07391829043626785, "global_step": 186406, "epoch": 2094, "lr": 2.2400639612699514e-05}
{"train_loss": 0.05786969140172005, "global_step": 186407, "epoch": 2094, "lr": 2.240015619969324e-05}
{"train_loss": 0.11979348957538605, "global_step": 186408, "epoch": 2094, "lr": 2.239967279039741e-05}
{"train_loss": 0.04539274424314499, "global_step": 186409, "epoch": 2094, "lr": 2.239918938481212e-05}
{"train_loss": 0.04380034655332565, "global_step": 186410, "epoch": 2094, "lr": 2.2398705982937408e-05}
{"train_loss": 0.03043137863278389, "global_step": 186411, "epoch": 2094, "lr": 2.239822258477337e-05}
{"train_loss": 0.07676000148057938, "global_step": 186412, "epoch": 2094, "lr": 2.2397739190320037e-05}
{"train_loss": 0.057577792555093765, "global_step": 186413, "epoch": 2094, "lr": 2.23972557995775e-05}
{"train_loss": 0.06934463977813721, "global_step": 186414, "epoch": 2094, "lr": 2.2396772412545824e-05}
{"train_loss": 0.09362363815307617, "global_step": 186415, "epoch": 2094, "lr": 2.2396289029225047e-05}
{"train_loss": 0.07337605953216553, "global_step": 186416, "epoch": 2094, "lr": 2.2395805649615277e-05}
{"train_loss": 0.05422710254788399, "global_step": 186417, "epoch": 2094, "lr": 2.2395322273716534e-05}
{"train_loss": 0.08066470175981522, "global_step": 186418, "epoch": 2094, "lr": 2.239483890152892e-05}
{"train_loss": 0.10067480057477951, "global_step": 186419, "epoch": 2094, "lr": 2.2394355533052464e-05}
{"train_loss": 0.09677702933549881, "global_step": 186420, "epoch": 2094, "lr": 2.2393872168287272e-05}
{"train_loss": 0.058182891458272934, "global_step": 186421, "epoch": 2094, "lr": 2.2393388807233368e-05}
{"train_loss": 0.09000690281391144, "global_step": 186422, "epoch": 2094, "lr": 2.2392905449890854e-05}
{"train_loss": 0.07331692427396774, "global_step": 186423, "epoch": 2094, "lr": 2.2392422096259758e-05}
{"train_loss": 0.09299005568027496, "global_step": 186424, "epoch": 2094, "lr": 2.2391938746340187e-05}
{"train_loss": 0.06813429296016693, "global_step": 186425, "epoch": 2094, "lr": 2.239145540013216e-05}
{"train_loss": 0.07940568029880524, "global_step": 186426, "epoch": 2094, "lr": 2.2390972057635784e-05}
{"train_loss": 0.08037340641021729, "global_step": 186427, "epoch": 2094, "lr": 2.2390488718851095e-05}
{"train_loss": 0.0756516382098198, "global_step": 186428, "epoch": 2094, "lr": 2.2390005383778178e-05}
{"train_loss": 0.036623116582632065, "global_step": 186429, "epoch": 2094, "lr": 2.238952205241709e-05}
{"train_loss": 0.06045741215348244, "global_step": 186430, "epoch": 2094, "lr": 2.2389038724767874e-05}
{"train_loss": 0.11705775558948517, "global_step": 186431, "epoch": 2094, "lr": 2.2388555400830642e-05}
{"train_loss": 0.08935225009918213, "global_step": 186432, "epoch": 2094, "lr": 2.2388072080605404e-05}
{"train_loss": 0.05887415632605553, "global_step": 186433, "epoch": 2094, "lr": 2.2387588764092277e-05}
{"train_loss": 0.08616305887699127, "global_step": 186434, "epoch": 2094, "lr": 2.238710545129128e-05}
{"train_loss": 0.1054496318101883, "global_step": 186435, "epoch": 2094, "lr": 2.2386622142202524e-05}
{"train_loss": 0.05836605280637741, "global_step": 186436, "epoch": 2094, "lr": 2.2386138836826027e-05}
{"train_loss": 0.03905845060944557, "global_step": 186437, "epoch": 2094, "lr": 2.23856555351619e-05}
{"train_loss": 0.06620263308286667, "global_step": 186438, "epoch": 2094, "lr": 2.238517223721016e-05}
{"train_loss": 0.0676368996500969, "global_step": 186439, "epoch": 2094, "lr": 2.2384688942970918e-05}
{"train_loss": 0.05742486193776131, "global_step": 186440, "epoch": 2094, "lr": 2.2384205652444202e-05}
{"train_loss": 0.04294849559664726, "global_step": 186441, "epoch": 2094, "lr": 2.238372236563011e-05}
{"train_loss": 0.06372767686843872, "global_step": 186442, "epoch": 2094, "lr": 2.2383239082528673e-05}
{"train_loss": 0.08536396920681, "global_step": 186443, "epoch": 2094, "lr": 2.2382755803139987e-05}
{"train_loss": 0.04943820834159851, "global_step": 186444, "epoch": 2094, "lr": 2.2382272527464087e-05}
{"train_loss": 0.05987941101193428, "global_step": 186445, "epoch": 2094, "lr": 2.2381789255501074e-05}
{"train_loss": 0.07413321733474731, "global_step": 186446, "epoch": 2094, "lr": 2.2381305987250973e-05}
{"train_loss": 0.08352169394493103, "global_step": 186447, "epoch": 2094, "lr": 2.2380822722713885e-05}
{"train_loss": 0.09441353380680084, "global_step": 186448, "epoch": 2094, "lr": 2.2380339461889844e-05}
{"train_loss": 0.053031545132398605, "global_step": 186449, "epoch": 2094, "lr": 2.2379856204778928e-05}
{"train_loss": 0.08591333031654358, "global_step": 186450, "epoch": 2094, "lr": 2.237937295138122e-05}
{"train_loss": 0.11083325743675232, "global_step": 186451, "epoch": 2094, "lr": 2.2378889701696754e-05}
{"train_loss": 0.08788663893938065, "global_step": 186452, "epoch": 2094, "lr": 2.2378406455725625e-05}
{"train_loss": 0.09723444283008575, "global_step": 186453, "epoch": 2094, "lr": 2.2377923213467865e-05}
{"train_loss": 0.07560351181231188, "global_step": 186454, "epoch": 2094, "lr": 2.2377439974923582e-05, "val_loss": 7.512120723724365}
{"train_loss": 0.05980845168232918, "global_step": 186455, "epoch": 2095, "lr": 2.2376956740092787e-05}
{"train_loss": 0.058746062219142914, "global_step": 186456, "epoch": 2095, "lr": 2.2376473508975597e-05}
{"train_loss": 0.04439646750688553, "global_step": 186457, "epoch": 2095, "lr": 2.2375990281572034e-05}
{"train_loss": 0.08494652807712555, "global_step": 186458, "epoch": 2095, "lr": 2.2375507057882202e-05}
{"train_loss": 0.041558925062417984, "global_step": 186459, "epoch": 2095, "lr": 2.237502383790612e-05}
{"train_loss": 0.06925269961357117, "global_step": 186460, "epoch": 2095, "lr": 2.2374540621643907e-05}
{"train_loss": 0.07799039781093597, "global_step": 186461, "epoch": 2095, "lr": 2.2374057409095577e-05}
{"train_loss": 0.09344534575939178, "global_step": 186462, "epoch": 2095, "lr": 2.2373574200261243e-05}
{"train_loss": 0.07724311947822571, "global_step": 186463, "epoch": 2095, "lr": 2.237309099514092e-05}
{"train_loss": 0.09010898321866989, "global_step": 186464, "epoch": 2095, "lr": 2.2372607793734723e-05}
{"train_loss": 0.06072649732232094, "global_step": 186465, "epoch": 2095, "lr": 2.237212459604267e-05}
{"train_loss": 0.05170423164963722, "global_step": 186466, "epoch": 2095, "lr": 2.2371641402064863e-05}
{"train_loss": 0.12854517996311188, "global_step": 186467, "epoch": 2095, "lr": 2.237115821180134e-05}
{"train_loss": 0.16612555086612701, "global_step": 186468, "epoch": 2095, "lr": 2.237067502525219e-05}
{"train_loss": 0.06922673434019089, "global_step": 186469, "epoch": 2095, "lr": 2.237019184241747e-05}
{"train_loss": 0.04931463673710823, "global_step": 186470, "epoch": 2095, "lr": 2.2369708663297222e-05}
{"train_loss": 0.10722032189369202, "global_step": 186471, "epoch": 2095, "lr": 2.2369225487891548e-05}
{"train_loss": 0.12144970893859863, "global_step": 186472, "epoch": 2095, "lr": 2.2368742316200482e-05}
{"train_loss": 0.0798940435051918, "global_step": 186473, "epoch": 2095, "lr": 2.2368259148224113e-05}
{"train_loss": 0.09558592736721039, "global_step": 186474, "epoch": 2095, "lr": 2.236777598396248e-05}
{"train_loss": 0.1048167422413826, "global_step": 186475, "epoch": 2095, "lr": 2.236729282341568e-05}
{"train_loss": 0.0791686400771141, "global_step": 186476, "epoch": 2095, "lr": 2.2366809666583737e-05}
{"train_loss": 0.05641665682196617, "global_step": 186477, "epoch": 2095, "lr": 2.236632651346677e-05}
{"train_loss": 0.031362760812044144, "global_step": 186478, "epoch": 2095, "lr": 2.2365843364064786e-05}
{"train_loss": 0.0853838324546814, "global_step": 186479, "epoch": 2095, "lr": 2.2365360218377896e-05}
{"train_loss": 0.14608365297317505, "global_step": 186480, "epoch": 2095, "lr": 2.2364877076406133e-05}
{"train_loss": 0.10159347206354141, "global_step": 186481, "epoch": 2095, "lr": 2.2364393938149592e-05}
{"train_loss": 0.06439686566591263, "global_step": 186482, "epoch": 2095, "lr": 2.23639108036083e-05}
{"train_loss": 0.08831178396940231, "global_step": 186483, "epoch": 2095, "lr": 2.236342767278236e-05}
{"train_loss": 0.040499284863471985, "global_step": 186484, "epoch": 2095, "lr": 2.236294454567181e-05}
{"train_loss": 0.06985945254564285, "global_step": 186485, "epoch": 2095, "lr": 2.236246142227672e-05}
{"train_loss": 0.0988226905465126, "global_step": 186486, "epoch": 2095, "lr": 2.236197830259718e-05}
{"train_loss": 0.05012037977576256, "global_step": 186487, "epoch": 2095, "lr": 2.2361495186633212e-05}
{"train_loss": 0.06702842563390732, "global_step": 186488, "epoch": 2095, "lr": 2.236101207438493e-05}
{"train_loss": 0.07911749929189682, "global_step": 186489, "epoch": 2095, "lr": 2.236052896585235e-05}
{"train_loss": 0.1114293783903122, "global_step": 186490, "epoch": 2095, "lr": 2.2360045861035583e-05}
{"train_loss": 0.048204876482486725, "global_step": 186491, "epoch": 2095, "lr": 2.2359562759934648e-05}
{"train_loss": 0.0928427055478096, "global_step": 186492, "epoch": 2095, "lr": 2.2359079662549652e-05}
{"train_loss": 0.08622955530881882, "global_step": 186493, "epoch": 2095, "lr": 2.2358596568880625e-05}
{"train_loss": 0.04010346159338951, "global_step": 186494, "epoch": 2095, "lr": 2.2358113478927666e-05}
{"train_loss": 0.055300042033195496, "global_step": 186495, "epoch": 2095, "lr": 2.23576303926908e-05}
{"train_loss": 0.0786743015050888, "global_step": 186496, "epoch": 2095, "lr": 2.2357147310170135e-05}
{"train_loss": 0.12386870384216309, "global_step": 186497, "epoch": 2095, "lr": 2.235666423136569e-05}
{"train_loss": 0.04582181200385094, "global_step": 186498, "epoch": 2095, "lr": 2.2356181156277585e-05}
{"train_loss": 0.12251244485378265, "global_step": 186499, "epoch": 2095, "lr": 2.2355698084905825e-05}
{"train_loss": 0.09210971742868423, "global_step": 186500, "epoch": 2095, "lr": 2.235521501725053e-05}
{"train_loss": 0.08473507314920425, "global_step": 186501, "epoch": 2095, "lr": 2.2354731953311716e-05}
{"train_loss": 0.08053411543369293, "global_step": 186502, "epoch": 2095, "lr": 2.235424889308949e-05}
{"train_loss": 0.07257702201604843, "global_step": 186503, "epoch": 2095, "lr": 2.2353765836583885e-05}
{"train_loss": 0.06537798047065735, "global_step": 186504, "epoch": 2095, "lr": 2.2353282783794994e-05}
{"train_loss": 0.047573983669281006, "global_step": 186505, "epoch": 2095, "lr": 2.2352799734722844e-05}
{"train_loss": 0.07708001136779785, "global_step": 186506, "epoch": 2095, "lr": 2.2352316689367552e-05}
{"train_loss": 0.057810597121715546, "global_step": 186507, "epoch": 2095, "lr": 2.235183364772915e-05}
{"train_loss": 0.06567218899726868, "global_step": 186508, "epoch": 2095, "lr": 2.2351350609807686e-05}
{"train_loss": 0.04197699576616287, "global_step": 186509, "epoch": 2095, "lr": 2.235086757560326e-05}
{"train_loss": 0.08958633244037628, "global_step": 186510, "epoch": 2095, "lr": 2.2350384545115915e-05}
{"train_loss": 0.04599032551050186, "global_step": 186511, "epoch": 2095, "lr": 2.2349901518345732e-05}
{"train_loss": 0.07781582325696945, "global_step": 186512, "epoch": 2095, "lr": 2.2349418495292757e-05}
{"train_loss": 0.06200459972023964, "global_step": 186513, "epoch": 2095, "lr": 2.2348935475957084e-05}
{"train_loss": 0.10396116226911545, "global_step": 186514, "epoch": 2095, "lr": 2.2348452460338737e-05}
{"train_loss": 0.07246324419975281, "global_step": 186515, "epoch": 2095, "lr": 2.234796944843782e-05}
{"train_loss": 0.10760116577148438, "global_step": 186516, "epoch": 2095, "lr": 2.2347486440254374e-05}
{"train_loss": 0.05841289088129997, "global_step": 186517, "epoch": 2095, "lr": 2.2347003435788476e-05}
{"train_loss": 0.10399210453033447, "global_step": 186518, "epoch": 2095, "lr": 2.234652043504018e-05}
{"train_loss": 0.08030401915311813, "global_step": 186519, "epoch": 2095, "lr": 2.234603743800957e-05}
{"train_loss": 0.08124536275863647, "global_step": 186520, "epoch": 2095, "lr": 2.2345554444696677e-05}
{"train_loss": 0.08196975290775299, "global_step": 186521, "epoch": 2095, "lr": 2.234507145510159e-05}
{"train_loss": 0.11797706037759781, "global_step": 186522, "epoch": 2095, "lr": 2.234458846922439e-05}
{"train_loss": 0.07576718926429749, "global_step": 186523, "epoch": 2095, "lr": 2.2344105487065102e-05}
{"train_loss": 0.07653568685054779, "global_step": 186524, "epoch": 2095, "lr": 2.2343622508623834e-05}
{"train_loss": 0.08143437653779984, "global_step": 186525, "epoch": 2095, "lr": 2.2343139533900608e-05}
{"train_loss": 0.054876867681741714, "global_step": 186526, "epoch": 2095, "lr": 2.234265656289553e-05}
{"train_loss": 0.09275515377521515, "global_step": 186527, "epoch": 2095, "lr": 2.234217359560863e-05}
{"train_loss": 0.06958166509866714, "global_step": 186528, "epoch": 2095, "lr": 2.2341690632039996e-05}
{"train_loss": 0.0756223127245903, "global_step": 186529, "epoch": 2095, "lr": 2.2341207672189675e-05}
{"train_loss": 0.08707472681999207, "global_step": 186530, "epoch": 2095, "lr": 2.234072471605776e-05}
{"train_loss": 0.05331858620047569, "global_step": 186531, "epoch": 2095, "lr": 2.234024176364427e-05}
{"train_loss": 0.0548282265663147, "global_step": 186532, "epoch": 2095, "lr": 2.2339758814949326e-05}
{"train_loss": 0.08257484436035156, "global_step": 186533, "epoch": 2095, "lr": 2.2339275869972944e-05}
{"train_loss": 0.0847172662615776, "global_step": 186534, "epoch": 2095, "lr": 2.233879292871522e-05}
{"train_loss": 0.06519737839698792, "global_step": 186535, "epoch": 2095, "lr": 2.2338309991176193e-05}
{"train_loss": 0.06582257151603699, "global_step": 186536, "epoch": 2095, "lr": 2.2337827057355964e-05}
{"train_loss": 0.1534854620695114, "global_step": 186537, "epoch": 2095, "lr": 2.233734412725456e-05}
{"train_loss": 0.11702746897935867, "global_step": 186538, "epoch": 2095, "lr": 2.2336861200872077e-05}
{"train_loss": 0.11231325566768646, "global_step": 186539, "epoch": 2095, "lr": 2.2336378278208553e-05}
{"train_loss": 0.058254681527614594, "global_step": 186540, "epoch": 2095, "lr": 2.233589535926408e-05}
{"train_loss": 0.07105264067649841, "global_step": 186541, "epoch": 2095, "lr": 2.233541244403869e-05}
{"train_loss": 0.06646527349948883, "global_step": 186542, "epoch": 2095, "lr": 2.2334929532532485e-05}
{"train_loss": 0.07882382038436579, "global_step": 186543, "epoch": 2095, "lr": 2.2334446624745498e-05, "val_loss": 7.67337703704834, "train_action_mse_error": 26.409608840942383}
{"train_loss": 0.08655137568712234, "global_step": 186544, "epoch": 2096, "lr": 2.233396372067782e-05}
{"train_loss": 0.0874810591340065, "global_step": 186545, "epoch": 2096, "lr": 2.2333480820329506e-05}
{"train_loss": 0.07842344790697098, "global_step": 186546, "epoch": 2096, "lr": 2.23329979237006e-05}
{"train_loss": 0.11167441308498383, "global_step": 186547, "epoch": 2096, "lr": 2.23325150307912e-05}
{"train_loss": 0.06903322786092758, "global_step": 186548, "epoch": 2096, "lr": 2.2332032141601344e-05}
{"train_loss": 0.08400791883468628, "global_step": 186549, "epoch": 2096, "lr": 2.2331549256131124e-05}
{"train_loss": 0.06109778955578804, "global_step": 186550, "epoch": 2096, "lr": 2.233106637438057e-05}
{"train_loss": 0.026331502944231033, "global_step": 186551, "epoch": 2096, "lr": 2.2330583496349787e-05}
{"train_loss": 0.03935969993472099, "global_step": 186552, "epoch": 2096, "lr": 2.2330100622038802e-05}
{"train_loss": 0.02845590002834797, "global_step": 186553, "epoch": 2096, "lr": 2.232961775144772e-05}
{"train_loss": 0.06280036270618439, "global_step": 186554, "epoch": 2096, "lr": 2.2329134884576557e-05}
{"train_loss": 0.0643908753991127, "global_step": 186555, "epoch": 2096, "lr": 2.2328652021425428e-05}
{"train_loss": 0.038770824670791626, "global_step": 186556, "epoch": 2096, "lr": 2.232816916199435e-05}
{"train_loss": 0.07896087318658829, "global_step": 186557, "epoch": 2096, "lr": 2.232768630628344e-05}
{"train_loss": 0.07951578497886658, "global_step": 186558, "epoch": 2096, "lr": 2.232720345429271e-05}
{"train_loss": 0.054657258093357086, "global_step": 186559, "epoch": 2096, "lr": 2.2326720606022257e-05}
{"train_loss": 0.08674793690443039, "global_step": 186560, "epoch": 2096, "lr": 2.232623776147215e-05}
{"train_loss": 0.0667303279042244, "global_step": 186561, "epoch": 2096, "lr": 2.2325754920642428e-05}
{"train_loss": 0.09641897678375244, "global_step": 186562, "epoch": 2096, "lr": 2.2325272083533188e-05}
{"train_loss": 0.08202171325683594, "global_step": 186563, "epoch": 2096, "lr": 2.2324789250144462e-05}
{"train_loss": 0.08103984594345093, "global_step": 186564, "epoch": 2096, "lr": 2.232430642047635e-05}
{"train_loss": 0.05298984795808792, "global_step": 186565, "epoch": 2096, "lr": 2.2323823594528876e-05}
{"train_loss": 0.10823536664247513, "global_step": 186566, "epoch": 2096, "lr": 2.2323340772302144e-05}
{"train_loss": 0.0623885914683342, "global_step": 186567, "epoch": 2096, "lr": 2.2322857953796183e-05}
{"train_loss": 0.07530078291893005, "global_step": 186568, "epoch": 2096, "lr": 2.2322375139011097e-05}
{"train_loss": 0.07812954485416412, "global_step": 186569, "epoch": 2096, "lr": 2.2321892327946914e-05}
{"train_loss": 0.03736061230301857, "global_step": 186570, "epoch": 2096, "lr": 2.232140952060373e-05}
{"train_loss": 0.09135431796312332, "global_step": 186571, "epoch": 2096, "lr": 2.232092671698158e-05}
{"train_loss": 0.08710893243551254, "global_step": 186572, "epoch": 2096, "lr": 2.2320443917080564e-05}
{"train_loss": 0.10143597424030304, "global_step": 186573, "epoch": 2096, "lr": 2.2319961120900702e-05}
{"train_loss": 0.07829979807138443, "global_step": 186574, "epoch": 2096, "lr": 2.2319478328442106e-05}
{"train_loss": 0.07625719904899597, "global_step": 186575, "epoch": 2096, "lr": 2.23189955397048e-05}
{"train_loss": 0.10631673783063889, "global_step": 186576, "epoch": 2096, "lr": 2.231851275468887e-05}
{"train_loss": 0.06482192873954773, "global_step": 186577, "epoch": 2096, "lr": 2.2318029973394394e-05}
{"train_loss": 0.0816732719540596, "global_step": 186578, "epoch": 2096, "lr": 2.2317547195821403e-05}
{"train_loss": 0.07868409156799316, "global_step": 186579, "epoch": 2096, "lr": 2.2317064421969996e-05}
{"train_loss": 0.06226414069533348, "global_step": 186580, "epoch": 2096, "lr": 2.231658165184023e-05}
{"train_loss": 0.08573964983224869, "global_step": 186581, "epoch": 2096, "lr": 2.2316098885432135e-05}
{"train_loss": 0.04327185079455376, "global_step": 186582, "epoch": 2096, "lr": 2.2315616122745832e-05}
{"train_loss": 0.05796213075518608, "global_step": 186583, "epoch": 2096, "lr": 2.231513336378133e-05}
{"train_loss": 0.07654375582933426, "global_step": 186584, "epoch": 2096, "lr": 2.231465060853874e-05}
{"train_loss": 0.07595597207546234, "global_step": 186585, "epoch": 2096, "lr": 2.2314167857018107e-05}
{"train_loss": 0.05136309564113617, "global_step": 186586, "epoch": 2096, "lr": 2.2313685109219483e-05}
{"train_loss": 0.05985952913761139, "global_step": 186587, "epoch": 2096, "lr": 2.2313202365142965e-05}
{"train_loss": 0.06054104492068291, "global_step": 186588, "epoch": 2096, "lr": 2.231271962478858e-05}
{"train_loss": 0.14657576382160187, "global_step": 186589, "epoch": 2096, "lr": 2.231223688815643e-05}
{"train_loss": 0.03661669045686722, "global_step": 186590, "epoch": 2096, "lr": 2.2311754155246546e-05}
{"train_loss": 0.05938142538070679, "global_step": 186591, "epoch": 2096, "lr": 2.231127142605902e-05}
{"train_loss": 0.03916313126683235, "global_step": 186592, "epoch": 2096, "lr": 2.2310788700593894e-05}
{"train_loss": 0.08289247751235962, "global_step": 186593, "epoch": 2096, "lr": 2.2310305978851266e-05}
{"train_loss": 0.05858449637889862, "global_step": 186594, "epoch": 2096, "lr": 2.2309823260831152e-05}
{"train_loss": 0.07105322927236557, "global_step": 186595, "epoch": 2096, "lr": 2.2309340546533653e-05}
{"train_loss": 0.08802998065948486, "global_step": 186596, "epoch": 2096, "lr": 2.230885783595884e-05}
{"train_loss": 0.061176612973213196, "global_step": 186597, "epoch": 2096, "lr": 2.230837512910675e-05}
{"train_loss": 0.0980134978890419, "global_step": 186598, "epoch": 2096, "lr": 2.2307892425977472e-05}
{"train_loss": 0.09103840589523315, "global_step": 186599, "epoch": 2096, "lr": 2.230740972657105e-05}
{"train_loss": 0.0817803144454956, "global_step": 186600, "epoch": 2096, "lr": 2.2306927030887577e-05}
{"train_loss": 0.0483260378241539, "global_step": 186601, "epoch": 2096, "lr": 2.2306444338927075e-05}
{"train_loss": 0.10477281361818314, "global_step": 186602, "epoch": 2096, "lr": 2.2305961650689656e-05}
{"train_loss": 0.12211515754461288, "global_step": 186603, "epoch": 2096, "lr": 2.2305478966175346e-05}
{"train_loss": 0.08523473143577576, "global_step": 186604, "epoch": 2096, "lr": 2.2304996285384245e-05}
{"train_loss": 0.06372657418251038, "global_step": 186605, "epoch": 2096, "lr": 2.230451360831638e-05}
{"train_loss": 0.12494964897632599, "global_step": 186606, "epoch": 2096, "lr": 2.2304030934971854e-05}
{"train_loss": 0.08844608813524246, "global_step": 186607, "epoch": 2096, "lr": 2.2303548265350695e-05}
{"train_loss": 0.03781202435493469, "global_step": 186608, "epoch": 2096, "lr": 2.2303065599453004e-05}
{"train_loss": 0.09349048137664795, "global_step": 186609, "epoch": 2096, "lr": 2.230258293727881e-05}
{"train_loss": 0.08962371200323105, "global_step": 186610, "epoch": 2096, "lr": 2.230210027882822e-05}
{"train_loss": 0.13513346016407013, "global_step": 186611, "epoch": 2096, "lr": 2.230161762410125e-05}
{"train_loss": 0.05228846147656441, "global_step": 186612, "epoch": 2096, "lr": 2.230113497309799e-05}
{"train_loss": 0.12542030215263367, "global_step": 186613, "epoch": 2096, "lr": 2.2300652325818528e-05}
{"train_loss": 0.02310158684849739, "global_step": 186614, "epoch": 2096, "lr": 2.2300169682262883e-05}
{"train_loss": 0.07282311469316483, "global_step": 186615, "epoch": 2096, "lr": 2.229968704243116e-05}
{"train_loss": 0.07595334947109222, "global_step": 186616, "epoch": 2096, "lr": 2.229920440632339e-05}
{"train_loss": 0.046683210879564285, "global_step": 186617, "epoch": 2096, "lr": 2.229872177393968e-05}
{"train_loss": 0.10358282178640366, "global_step": 186618, "epoch": 2096, "lr": 2.2298239145280058e-05}
{"train_loss": 0.052575740963220596, "global_step": 186619, "epoch": 2096, "lr": 2.2297756520344586e-05}
{"train_loss": 0.06052946671843529, "global_step": 186620, "epoch": 2096, "lr": 2.2297273899133365e-05}
{"train_loss": 0.03940655663609505, "global_step": 186621, "epoch": 2096, "lr": 2.229679128164642e-05}
{"train_loss": 0.07299192249774933, "global_step": 186622, "epoch": 2096, "lr": 2.2296308667883843e-05}
{"train_loss": 0.0813378393650055, "global_step": 186623, "epoch": 2096, "lr": 2.2295826057845698e-05}
{"train_loss": 0.09178002178668976, "global_step": 186624, "epoch": 2096, "lr": 2.2295343451532018e-05}
{"train_loss": 0.08960595726966858, "global_step": 186625, "epoch": 2096, "lr": 2.229486084894291e-05}
{"train_loss": 0.0787176713347435, "global_step": 186626, "epoch": 2096, "lr": 2.2294378250078408e-05}
{"train_loss": 0.09449481964111328, "global_step": 186627, "epoch": 2096, "lr": 2.22938956549386e-05}
{"train_loss": 0.019683077931404114, "global_step": 186628, "epoch": 2096, "lr": 2.229341306352353e-05}
{"train_loss": 0.06792107969522476, "global_step": 186629, "epoch": 2096, "lr": 2.2292930475833284e-05}
{"train_loss": 0.0456017442047596, "global_step": 186630, "epoch": 2096, "lr": 2.2292447891867897e-05}
{"train_loss": 0.04066786170005798, "global_step": 186631, "epoch": 2096, "lr": 2.2291965311627457e-05}
{"train_loss": 0.07392377860509278, "global_step": 186632, "epoch": 2096, "lr": 2.229148273511204e-05, "val_loss": 7.449168682098389}
{"train_loss": 0.06012905761599541, "global_step": 186633, "epoch": 2097, "lr": 2.2291000162321678e-05}
{"train_loss": 0.048213567584753036, "global_step": 186634, "epoch": 2097, "lr": 2.229051759325647e-05}
{"train_loss": 0.08756467700004578, "global_step": 186635, "epoch": 2097, "lr": 2.229003502791645e-05}
{"train_loss": 0.0627279281616211, "global_step": 186636, "epoch": 2097, "lr": 2.2289552466301717e-05}
{"train_loss": 0.03230816870927811, "global_step": 186637, "epoch": 2097, "lr": 2.228906990841229e-05}
{"train_loss": 0.06327884644269943, "global_step": 186638, "epoch": 2097, "lr": 2.228858735424828e-05}
{"train_loss": 0.07033589482307434, "global_step": 186639, "epoch": 2097, "lr": 2.228810480380972e-05}
{"train_loss": 0.06483975797891617, "global_step": 186640, "epoch": 2097, "lr": 2.2287622257096695e-05}
{"train_loss": 0.09893083572387695, "global_step": 186641, "epoch": 2097, "lr": 2.228713971410925e-05}
{"train_loss": 0.05344367027282715, "global_step": 186642, "epoch": 2097, "lr": 2.2286657174847482e-05}
{"train_loss": 0.05387478321790695, "global_step": 186643, "epoch": 2097, "lr": 2.2286174639311408e-05}
{"train_loss": 0.06125437095761299, "global_step": 186644, "epoch": 2097, "lr": 2.2285692107501143e-05}
{"train_loss": 0.07783941179513931, "global_step": 186645, "epoch": 2097, "lr": 2.2285209579416712e-05}
{"train_loss": 0.05560341104865074, "global_step": 186646, "epoch": 2097, "lr": 2.2284727055058213e-05}
{"train_loss": 0.06525819003582001, "global_step": 186647, "epoch": 2097, "lr": 2.228424453442568e-05}
{"train_loss": 0.0960584506392479, "global_step": 186648, "epoch": 2097, "lr": 2.2283762017519187e-05}
{"train_loss": 0.07599295675754547, "global_step": 186649, "epoch": 2097, "lr": 2.228327950433883e-05}
{"train_loss": 0.07552801072597504, "global_step": 186650, "epoch": 2097, "lr": 2.228279699488462e-05}
{"train_loss": 0.10852228105068207, "global_step": 186651, "epoch": 2097, "lr": 2.2282314489156675e-05}
{"train_loss": 0.14077730476856232, "global_step": 186652, "epoch": 2097, "lr": 2.2281831987155018e-05}
{"train_loss": 0.10361368209123611, "global_step": 186653, "epoch": 2097, "lr": 2.2281349488879743e-05}
{"train_loss": 0.09024930745363235, "global_step": 186654, "epoch": 2097, "lr": 2.228086699433089e-05}
{"train_loss": 0.04169480502605438, "global_step": 186655, "epoch": 2097, "lr": 2.228038450350855e-05}
{"train_loss": 0.09710489958524704, "global_step": 186656, "epoch": 2097, "lr": 2.227990201641278e-05}
{"train_loss": 0.09492550045251846, "global_step": 186657, "epoch": 2097, "lr": 2.2279419533043615e-05}
{"train_loss": 0.04514765739440918, "global_step": 186658, "epoch": 2097, "lr": 2.2278937053401167e-05}
{"train_loss": 0.062018021941185, "global_step": 186659, "epoch": 2097, "lr": 2.227845457748546e-05}
{"train_loss": 0.073923259973526, "global_step": 186660, "epoch": 2097, "lr": 2.2277972105296596e-05}
{"train_loss": 0.04039962589740753, "global_step": 186661, "epoch": 2097, "lr": 2.2277489636834614e-05}
{"train_loss": 0.08964652568101883, "global_step": 186662, "epoch": 2097, "lr": 2.2277007172099574e-05}
{"train_loss": 0.053853459656238556, "global_step": 186663, "epoch": 2097, "lr": 2.2276524711091567e-05}
{"train_loss": 0.05404989793896675, "global_step": 186664, "epoch": 2097, "lr": 2.227604225381063e-05}
{"train_loss": 0.11070597171783447, "global_step": 186665, "epoch": 2097, "lr": 2.2275559800256856e-05}
{"train_loss": 0.03025844879448414, "global_step": 186666, "epoch": 2097, "lr": 2.2275077350430273e-05}
{"train_loss": 0.027698704972863197, "global_step": 186667, "epoch": 2097, "lr": 2.2274594904330974e-05}
{"train_loss": 0.08793733268976212, "global_step": 186668, "epoch": 2097, "lr": 2.2274112461959034e-05}
{"train_loss": 0.07626935839653015, "global_step": 186669, "epoch": 2097, "lr": 2.2273630023314485e-05}
{"train_loss": 0.06335566937923431, "global_step": 186670, "epoch": 2097, "lr": 2.2273147588397424e-05}
{"train_loss": 0.036730773746967316, "global_step": 186671, "epoch": 2097, "lr": 2.2272665157207884e-05}
{"train_loss": 0.049235839396715164, "global_step": 186672, "epoch": 2097, "lr": 2.227218272974596e-05}
{"train_loss": 0.10402964800596237, "global_step": 186673, "epoch": 2097, "lr": 2.2271700306011688e-05}
{"train_loss": 0.13719363510608673, "global_step": 186674, "epoch": 2097, "lr": 2.2271217886005168e-05}
{"train_loss": 0.08658254891633987, "global_step": 186675, "epoch": 2097, "lr": 2.227073546972643e-05}
{"train_loss": 0.05483251065015793, "global_step": 186676, "epoch": 2097, "lr": 2.2270253057175565e-05}
{"train_loss": 0.06496798992156982, "global_step": 186677, "epoch": 2097, "lr": 2.226977064835261e-05}
{"train_loss": 0.061941906809806824, "global_step": 186678, "epoch": 2097, "lr": 2.226928824325766e-05}
{"train_loss": 0.10183525830507278, "global_step": 186679, "epoch": 2097, "lr": 2.2268805841890756e-05}
{"train_loss": 0.057300958782434464, "global_step": 186680, "epoch": 2097, "lr": 2.2268323444251987e-05}
{"train_loss": 0.11869446188211441, "global_step": 186681, "epoch": 2097, "lr": 2.2267841050341387e-05}
{"train_loss": 0.052817102521657944, "global_step": 186682, "epoch": 2097, "lr": 2.226735866015906e-05}
{"train_loss": 0.059134602546691895, "global_step": 186683, "epoch": 2097, "lr": 2.2266876273705022e-05}
{"train_loss": 0.05443843826651573, "global_step": 186684, "epoch": 2097, "lr": 2.2266393890979387e-05}
{"train_loss": 0.08458353579044342, "global_step": 186685, "epoch": 2097, "lr": 2.226591151198218e-05}
{"train_loss": 0.0588214211165905, "global_step": 186686, "epoch": 2097, "lr": 2.2265429136713484e-05}
{"train_loss": 0.1146775335073471, "global_step": 186687, "epoch": 2097, "lr": 2.226494676517338e-05}
{"train_loss": 0.0910290777683258, "global_step": 186688, "epoch": 2097, "lr": 2.2264464397361894e-05}
{"train_loss": 0.07196927070617676, "global_step": 186689, "epoch": 2097, "lr": 2.2263982033279135e-05}
{"train_loss": 0.0746842697262764, "global_step": 186690, "epoch": 2097, "lr": 2.2263499672925127e-05}
{"train_loss": 0.11332842707633972, "global_step": 186691, "epoch": 2097, "lr": 2.226301731629997e-05}
{"train_loss": 0.09772294014692307, "global_step": 186692, "epoch": 2097, "lr": 2.2262534963403696e-05}
{"train_loss": 0.09080840647220612, "global_step": 186693, "epoch": 2097, "lr": 2.2262052614236404e-05}
{"train_loss": 0.11298497021198273, "global_step": 186694, "epoch": 2097, "lr": 2.226157026879812e-05}
{"train_loss": 0.05348693206906319, "global_step": 186695, "epoch": 2097, "lr": 2.2261087927088952e-05}
{"train_loss": 0.04546970874071121, "global_step": 186696, "epoch": 2097, "lr": 2.226060558910894e-05}
{"train_loss": 0.09634074568748474, "global_step": 186697, "epoch": 2097, "lr": 2.2260123254858133e-05}
{"train_loss": 0.040455516427755356, "global_step": 186698, "epoch": 2097, "lr": 2.2259640924336632e-05}
{"train_loss": 0.05663024261593819, "global_step": 186699, "epoch": 2097, "lr": 2.2259158597544472e-05}
{"train_loss": 0.09451533854007721, "global_step": 186700, "epoch": 2097, "lr": 2.2258676274481743e-05}
{"train_loss": 0.11681900918483734, "global_step": 186701, "epoch": 2097, "lr": 2.2258193955148498e-05}
{"train_loss": 0.11309709399938583, "global_step": 186702, "epoch": 2097, "lr": 2.2257711639544782e-05}
{"train_loss": 0.04638671875, "global_step": 186703, "epoch": 2097, "lr": 2.2257229327670674e-05}
{"train_loss": 0.03641902282834053, "global_step": 186704, "epoch": 2097, "lr": 2.225674701952627e-05}
{"train_loss": 0.09304749965667725, "global_step": 186705, "epoch": 2097, "lr": 2.2256264715111585e-05}
{"train_loss": 0.08732590824365616, "global_step": 186706, "epoch": 2097, "lr": 2.2255782414426728e-05}
{"train_loss": 0.07568538188934326, "global_step": 186707, "epoch": 2097, "lr": 2.2255300117471723e-05}
{"train_loss": 0.0907655879855156, "global_step": 186708, "epoch": 2097, "lr": 2.2254817824246672e-05}
{"train_loss": 0.07255318015813828, "global_step": 186709, "epoch": 2097, "lr": 2.225433553475161e-05}
{"train_loss": 0.05824130028486252, "global_step": 186710, "epoch": 2097, "lr": 2.225385324898662e-05}
{"train_loss": 0.08935552090406418, "global_step": 186711, "epoch": 2097, "lr": 2.2253370966951753e-05}
{"train_loss": 0.06609633564949036, "global_step": 186712, "epoch": 2097, "lr": 2.22528886886471e-05}
{"train_loss": 0.07010860741138458, "global_step": 186713, "epoch": 2097, "lr": 2.225240641407269e-05}
{"train_loss": 0.04862501844763756, "global_step": 186714, "epoch": 2097, "lr": 2.2251924143228618e-05}
{"train_loss": 0.09339623153209686, "global_step": 186715, "epoch": 2097, "lr": 2.2251441876114926e-05}
{"train_loss": 0.03329657390713692, "global_step": 186716, "epoch": 2097, "lr": 2.2250959612731703e-05}
{"train_loss": 0.043198589235544205, "global_step": 186717, "epoch": 2097, "lr": 2.2250477353078986e-05}
{"train_loss": 0.056148711591959, "global_step": 186718, "epoch": 2097, "lr": 2.2249995097156866e-05}
{"train_loss": 0.06418872624635696, "global_step": 186719, "epoch": 2097, "lr": 2.224951284496538e-05}
{"train_loss": 0.0933125764131546, "global_step": 186720, "epoch": 2097, "lr": 2.2249030596504632e-05}
{"train_loss": 0.07388219258256173, "global_step": 186721, "epoch": 2097, "lr": 2.224854835177464e-05, "val_loss": 7.639362812042236}
{"train_loss": 0.0655164048075676, "global_step": 186722, "epoch": 2098, "lr": 2.2248066110775496e-05}
{"train_loss": 0.06834559887647629, "global_step": 186723, "epoch": 2098, "lr": 2.2247583873507278e-05}
{"train_loss": 0.08692575246095657, "global_step": 186724, "epoch": 2098, "lr": 2.2247101639970013e-05}
{"train_loss": 0.11082538962364197, "global_step": 186725, "epoch": 2098, "lr": 2.224661941016381e-05}
{"train_loss": 0.10562806576490402, "global_step": 186726, "epoch": 2098, "lr": 2.2246137184088687e-05}
{"train_loss": 0.08287698030471802, "global_step": 186727, "epoch": 2098, "lr": 2.2245654961744756e-05}
{"train_loss": 0.04148232936859131, "global_step": 186728, "epoch": 2098, "lr": 2.2245172743132035e-05}
{"train_loss": 0.057230014353990555, "global_step": 186729, "epoch": 2098, "lr": 2.224469052825064e-05}
{"train_loss": 0.05317972972989082, "global_step": 186730, "epoch": 2098, "lr": 2.224420831710058e-05}
{"train_loss": 0.07627306133508682, "global_step": 186731, "epoch": 2098, "lr": 2.2243726109681967e-05}
{"train_loss": 0.05108539015054703, "global_step": 186732, "epoch": 2098, "lr": 2.224324390599483e-05}
{"train_loss": 0.08196941018104553, "global_step": 186733, "epoch": 2098, "lr": 2.224276170603927e-05}
{"train_loss": 0.06164456158876419, "global_step": 186734, "epoch": 2098, "lr": 2.2242279509815332e-05}
{"train_loss": 0.10350102186203003, "global_step": 186735, "epoch": 2098, "lr": 2.2241797317323065e-05}
{"train_loss": 0.07363194972276688, "global_step": 186736, "epoch": 2098, "lr": 2.2241315128562567e-05}
{"train_loss": 0.09108351171016693, "global_step": 186737, "epoch": 2098, "lr": 2.224083294353387e-05}
{"train_loss": 0.08493193238973618, "global_step": 186738, "epoch": 2098, "lr": 2.2240350762237066e-05}
{"train_loss": 0.0696941465139389, "global_step": 186739, "epoch": 2098, "lr": 2.2239868584672196e-05}
{"train_loss": 0.10622820258140564, "global_step": 186740, "epoch": 2098, "lr": 2.2239386410839353e-05}
{"train_loss": 0.09487591683864594, "global_step": 186741, "epoch": 2098, "lr": 2.2238904240738567e-05}
{"train_loss": 0.08617045730352402, "global_step": 186742, "epoch": 2098, "lr": 2.2238422074369947e-05}
{"train_loss": 0.04251544550061226, "global_step": 186743, "epoch": 2098, "lr": 2.2237939911733507e-05}
{"train_loss": 0.07815883308649063, "global_step": 186744, "epoch": 2098, "lr": 2.223745775282936e-05}
{"train_loss": 0.04562089592218399, "global_step": 186745, "epoch": 2098, "lr": 2.223697559765753e-05}
{"train_loss": 0.0816599503159523, "global_step": 186746, "epoch": 2098, "lr": 2.2236493446218116e-05}
{"train_loss": 0.05370021611452103, "global_step": 186747, "epoch": 2098, "lr": 2.2236011298511143e-05}
{"train_loss": 0.029839882627129555, "global_step": 186748, "epoch": 2098, "lr": 2.2235529154536726e-05}
{"train_loss": 0.11261647194623947, "global_step": 186749, "epoch": 2098, "lr": 2.2235047014294885e-05}
{"train_loss": 0.0625620037317276, "global_step": 186750, "epoch": 2098, "lr": 2.2234564877785726e-05}
{"train_loss": 0.03640094771981239, "global_step": 186751, "epoch": 2098, "lr": 2.2234082745009267e-05}
{"train_loss": 0.09538156539201736, "global_step": 186752, "epoch": 2098, "lr": 2.2233600615965615e-05}
{"train_loss": 0.047547850757837296, "global_step": 186753, "epoch": 2098, "lr": 2.2233118490654797e-05}
{"train_loss": 0.0735301524400711, "global_step": 186754, "epoch": 2098, "lr": 2.223263636907692e-05}
{"train_loss": 0.07245421409606934, "global_step": 186755, "epoch": 2098, "lr": 2.2232154251232013e-05}
{"train_loss": 0.05367373675107956, "global_step": 186756, "epoch": 2098, "lr": 2.2231672137120163e-05}
{"train_loss": 0.06362003087997437, "global_step": 186757, "epoch": 2098, "lr": 2.2231190026741417e-05}
{"train_loss": 0.039295971393585205, "global_step": 186758, "epoch": 2098, "lr": 2.223070792009584e-05}
{"train_loss": 0.03369175270199776, "global_step": 186759, "epoch": 2098, "lr": 2.223022581718353e-05}
{"train_loss": 0.08257892727851868, "global_step": 186760, "epoch": 2098, "lr": 2.222974371800451e-05}
{"train_loss": 0.0866512805223465, "global_step": 186761, "epoch": 2098, "lr": 2.2229261622558876e-05}
{"train_loss": 0.06660333275794983, "global_step": 186762, "epoch": 2098, "lr": 2.2228779530846667e-05}
{"train_loss": 0.09797786921262741, "global_step": 186763, "epoch": 2098, "lr": 2.2228297442867975e-05}
{"train_loss": 0.07172630727291107, "global_step": 186764, "epoch": 2098, "lr": 2.222781535862283e-05}
{"train_loss": 0.0919099897146225, "global_step": 186765, "epoch": 2098, "lr": 2.2227333278111335e-05}
{"train_loss": 0.08471474051475525, "global_step": 186766, "epoch": 2098, "lr": 2.222685120133352e-05}
{"train_loss": 0.044012658298015594, "global_step": 186767, "epoch": 2098, "lr": 2.2226369128289486e-05}
{"train_loss": 0.07990759611129761, "global_step": 186768, "epoch": 2098, "lr": 2.2225887058979254e-05}
{"train_loss": 0.11353936046361923, "global_step": 186769, "epoch": 2098, "lr": 2.2225404993402937e-05}
{"train_loss": 0.043660156428813934, "global_step": 186770, "epoch": 2098, "lr": 2.222492293156056e-05}
{"train_loss": 0.08257976174354553, "global_step": 186771, "epoch": 2098, "lr": 2.222444087345222e-05}
{"train_loss": 0.042987290769815445, "global_step": 186772, "epoch": 2098, "lr": 2.222395881907796e-05}
{"train_loss": 0.05387696251273155, "global_step": 186773, "epoch": 2098, "lr": 2.2223476768437833e-05}
{"train_loss": 0.06867814809083939, "global_step": 186774, "epoch": 2098, "lr": 2.2222994721531942e-05}
{"train_loss": 0.06581193208694458, "global_step": 186775, "epoch": 2098, "lr": 2.2222512678360314e-05}
{"train_loss": 0.07179784774780273, "global_step": 186776, "epoch": 2098, "lr": 2.2222030638923048e-05}
{"train_loss": 0.04955272004008293, "global_step": 186777, "epoch": 2098, "lr": 2.222154860322017e-05}
{"train_loss": 0.0969412550330162, "global_step": 186778, "epoch": 2098, "lr": 2.2221066571251786e-05}
{"train_loss": 0.05414489656686783, "global_step": 186779, "epoch": 2098, "lr": 2.2220584543017926e-05}
{"train_loss": 0.06332289427518845, "global_step": 186780, "epoch": 2098, "lr": 2.2220102518518686e-05}
{"train_loss": 0.14518332481384277, "global_step": 186781, "epoch": 2098, "lr": 2.2219620497754095e-05}
{"train_loss": 0.06101461499929428, "global_step": 186782, "epoch": 2098, "lr": 2.2219138480724255e-05}
{"train_loss": 0.07605204731225967, "global_step": 186783, "epoch": 2098, "lr": 2.22186564674292e-05}
{"train_loss": 0.08698239177465439, "global_step": 186784, "epoch": 2098, "lr": 2.2218174457869022e-05}
{"train_loss": 0.03798849135637283, "global_step": 186785, "epoch": 2098, "lr": 2.2217692452043754e-05}
{"train_loss": 0.08474119007587433, "global_step": 186786, "epoch": 2098, "lr": 2.2217210449953495e-05}
{"train_loss": 0.06712935864925385, "global_step": 186787, "epoch": 2098, "lr": 2.221672845159828e-05}
{"train_loss": 0.06883303076028824, "global_step": 186788, "epoch": 2098, "lr": 2.22162464569782e-05}
{"train_loss": 0.08991578221321106, "global_step": 186789, "epoch": 2098, "lr": 2.2215764466093293e-05}
{"train_loss": 0.06702148169279099, "global_step": 186790, "epoch": 2098, "lr": 2.2215282478943655e-05}
{"train_loss": 0.07579834014177322, "global_step": 186791, "epoch": 2098, "lr": 2.2214800495529314e-05}
{"train_loss": 0.06969013810157776, "global_step": 186792, "epoch": 2098, "lr": 2.2214318515850374e-05}
{"train_loss": 0.0712096095085144, "global_step": 186793, "epoch": 2098, "lr": 2.2213836539906863e-05}
{"train_loss": 0.08624473959207535, "global_step": 186794, "epoch": 2098, "lr": 2.2213354567698863e-05}
{"train_loss": 0.04802444577217102, "global_step": 186795, "epoch": 2098, "lr": 2.2212872599226457e-05}
{"train_loss": 0.11112063378095627, "global_step": 186796, "epoch": 2098, "lr": 2.221239063448967e-05}
{"train_loss": 0.04657212272286415, "global_step": 186797, "epoch": 2098, "lr": 2.2211908673488614e-05}
{"train_loss": 0.09553153067827225, "global_step": 186798, "epoch": 2098, "lr": 2.2211426716223298e-05}
{"train_loss": 0.07307956367731094, "global_step": 186799, "epoch": 2098, "lr": 2.2210944762693848e-05}
{"train_loss": 0.10527738183736801, "global_step": 186800, "epoch": 2098, "lr": 2.2210462812900274e-05}
{"train_loss": 0.08383255451917648, "global_step": 186801, "epoch": 2098, "lr": 2.2209980866842677e-05}
{"train_loss": 0.09455880522727966, "global_step": 186802, "epoch": 2098, "lr": 2.22094989245211e-05}
{"train_loss": 0.07029350101947784, "global_step": 186803, "epoch": 2098, "lr": 2.220901698593563e-05}
{"train_loss": 0.07229475677013397, "global_step": 186804, "epoch": 2098, "lr": 2.2208535051086304e-05}
{"train_loss": 0.07019983232021332, "global_step": 186805, "epoch": 2098, "lr": 2.2208053119973215e-05}
{"train_loss": 0.07143867760896683, "global_step": 186806, "epoch": 2098, "lr": 2.2207571192596405e-05}
{"train_loss": 0.04952577129006386, "global_step": 186807, "epoch": 2098, "lr": 2.2207089268955962e-05}
{"train_loss": 0.08574455231428146, "global_step": 186808, "epoch": 2098, "lr": 2.2206607349051918e-05}
{"train_loss": 0.06347595900297165, "global_step": 186809, "epoch": 2098, "lr": 2.2206125432884373e-05}
{"train_loss": 0.0725525151519628, "global_step": 186810, "epoch": 2098, "lr": 2.220564352045336e-05, "val_loss": 7.970549583435059}
{"train_loss": 0.09230373054742813, "global_step": 186811, "epoch": 2099, "lr": 2.2205161611758977e-05}
{"train_loss": 0.0656806156039238, "global_step": 186812, "epoch": 2099, "lr": 2.2204679706801273e-05}
{"train_loss": 0.12862004339694977, "global_step": 186813, "epoch": 2099, "lr": 2.220419780558029e-05}
{"train_loss": 0.06711167842149734, "global_step": 186814, "epoch": 2099, "lr": 2.2203715908096135e-05}
{"train_loss": 0.08495429158210754, "global_step": 186815, "epoch": 2099, "lr": 2.2203234014348834e-05}
{"train_loss": 0.050742149353027344, "global_step": 186816, "epoch": 2099, "lr": 2.2202752124338483e-05}
{"train_loss": 0.1080513745546341, "global_step": 186817, "epoch": 2099, "lr": 2.220227023806512e-05}
{"train_loss": 0.09568558633327484, "global_step": 186818, "epoch": 2099, "lr": 2.2201788355528834e-05}
{"train_loss": 0.06181858479976654, "global_step": 186819, "epoch": 2099, "lr": 2.2201306476729665e-05}
{"train_loss": 0.06876007467508316, "global_step": 186820, "epoch": 2099, "lr": 2.2200824601667714e-05}
{"train_loss": 0.06467002630233765, "global_step": 186821, "epoch": 2099, "lr": 2.2200342730342998e-05}
{"train_loss": 0.06004882603883743, "global_step": 186822, "epoch": 2099, "lr": 2.2199860862755635e-05}
{"train_loss": 0.06308162957429886, "global_step": 186823, "epoch": 2099, "lr": 2.2199378998905636e-05}
{"train_loss": 0.07394170761108398, "global_step": 186824, "epoch": 2099, "lr": 2.2198897138793108e-05}
{"train_loss": 0.08020655065774918, "global_step": 186825, "epoch": 2099, "lr": 2.2198415282418083e-05}
{"train_loss": 0.0921010822057724, "global_step": 186826, "epoch": 2099, "lr": 2.2197933429780665e-05}
{"train_loss": 0.0798736959695816, "global_step": 186827, "epoch": 2099, "lr": 2.2197451580880875e-05}
{"train_loss": 0.04617776349186897, "global_step": 186828, "epoch": 2099, "lr": 2.2196969735718814e-05}
{"train_loss": 0.06500218063592911, "global_step": 186829, "epoch": 2099, "lr": 2.2196487894294514e-05}
{"train_loss": 0.09170880168676376, "global_step": 186830, "epoch": 2099, "lr": 2.219600605660806e-05}
{"train_loss": 0.05001423507928848, "global_step": 186831, "epoch": 2099, "lr": 2.219552422265953e-05}
{"train_loss": 0.059483565390110016, "global_step": 186832, "epoch": 2099, "lr": 2.2195042392448957e-05}
{"train_loss": 0.0647764727473259, "global_step": 186833, "epoch": 2099, "lr": 2.219456056597644e-05}
{"train_loss": 0.07037677615880966, "global_step": 186834, "epoch": 2099, "lr": 2.2194078743242007e-05}
{"train_loss": 0.06497456133365631, "global_step": 186835, "epoch": 2099, "lr": 2.2193596924245756e-05}
{"train_loss": 0.05347348749637604, "global_step": 186836, "epoch": 2099, "lr": 2.2193115108987723e-05}
{"train_loss": 0.10568219423294067, "global_step": 186837, "epoch": 2099, "lr": 2.2192633297468002e-05}
{"train_loss": 0.06221149489283562, "global_step": 186838, "epoch": 2099, "lr": 2.2192151489686623e-05}
{"train_loss": 0.08889210969209671, "global_step": 186839, "epoch": 2099, "lr": 2.2191669685643696e-05}
{"train_loss": 0.07495361566543579, "global_step": 186840, "epoch": 2099, "lr": 2.2191187885339236e-05}
{"train_loss": 0.04890913888812065, "global_step": 186841, "epoch": 2099, "lr": 2.219070608877335e-05}
{"train_loss": 0.04576073959469795, "global_step": 186842, "epoch": 2099, "lr": 2.2190224295946067e-05}
{"train_loss": 0.09164776653051376, "global_step": 186843, "epoch": 2099, "lr": 2.2189742506857492e-05}
{"train_loss": 0.07060542702674866, "global_step": 186844, "epoch": 2099, "lr": 2.2189260721507643e-05}
{"train_loss": 0.10077282786369324, "global_step": 186845, "epoch": 2099, "lr": 2.2188778939896625e-05}
{"train_loss": 0.06647331267595291, "global_step": 186846, "epoch": 2099, "lr": 2.2188297162024474e-05}
{"train_loss": 0.06178145110607147, "global_step": 186847, "epoch": 2099, "lr": 2.2187815387891287e-05}
{"train_loss": 0.11819589138031006, "global_step": 186848, "epoch": 2099, "lr": 2.2187333617497085e-05}
{"train_loss": 0.10988672822713852, "global_step": 186849, "epoch": 2099, "lr": 2.2186851850841982e-05}
{"train_loss": 0.052417367696762085, "global_step": 186850, "epoch": 2099, "lr": 2.218637008792601e-05}
{"train_loss": 0.04258072376251221, "global_step": 186851, "epoch": 2099, "lr": 2.2185888328749232e-05}
{"train_loss": 0.09944173693656921, "global_step": 186852, "epoch": 2099, "lr": 2.2185406573311735e-05}
{"train_loss": 0.05590156465768814, "global_step": 186853, "epoch": 2099, "lr": 2.2184924821613557e-05}
{"train_loss": 0.07602705806493759, "global_step": 186854, "epoch": 2099, "lr": 2.2184443073654788e-05}
{"train_loss": 0.09054984152317047, "global_step": 186855, "epoch": 2099, "lr": 2.218396132943547e-05}
{"train_loss": 0.06989926099777222, "global_step": 186856, "epoch": 2099, "lr": 2.218347958895569e-05}
{"train_loss": 0.06095534935593605, "global_step": 186857, "epoch": 2099, "lr": 2.2182997852215494e-05}
{"train_loss": 0.03388968110084534, "global_step": 186858, "epoch": 2099, "lr": 2.2182516119214968e-05}
{"train_loss": 0.05980140343308449, "global_step": 186859, "epoch": 2099, "lr": 2.2182034389954143e-05}
{"train_loss": 0.0855545848608017, "global_step": 186860, "epoch": 2099, "lr": 2.2181552664433125e-05}
{"train_loss": 0.10929840058088303, "global_step": 186861, "epoch": 2099, "lr": 2.2181070942651943e-05}
{"train_loss": 0.17020215094089508, "global_step": 186862, "epoch": 2099, "lr": 2.2180589224610694e-05}
{"train_loss": 0.10802736133337021, "global_step": 186863, "epoch": 2099, "lr": 2.21801075103094e-05}
{"train_loss": 0.08101274073123932, "global_step": 186864, "epoch": 2099, "lr": 2.217962579974818e-05}
{"train_loss": 0.08677845448255539, "global_step": 186865, "epoch": 2099, "lr": 2.2179144092927046e-05}
{"train_loss": 0.12256816029548645, "global_step": 186866, "epoch": 2099, "lr": 2.2178662389846088e-05}
{"train_loss": 0.0763891339302063, "global_step": 186867, "epoch": 2099, "lr": 2.217818069050539e-05}
{"train_loss": 0.056608594954013824, "global_step": 186868, "epoch": 2099, "lr": 2.2177698994904978e-05}
{"train_loss": 0.06388973444700241, "global_step": 186869, "epoch": 2099, "lr": 2.217721730304495e-05}
{"train_loss": 0.06929534673690796, "global_step": 186870, "epoch": 2099, "lr": 2.2176735614925336e-05}
{"train_loss": 0.035002127289772034, "global_step": 186871, "epoch": 2099, "lr": 2.2176253930546247e-05}
{"train_loss": 0.041312698274850845, "global_step": 186872, "epoch": 2099, "lr": 2.21757722499077e-05}
{"train_loss": 0.04752250015735626, "global_step": 186873, "epoch": 2099, "lr": 2.21752905730098e-05}
{"train_loss": 0.0660448744893074, "global_step": 186874, "epoch": 2099, "lr": 2.2174808899852572e-05}
{"train_loss": 0.06977753341197968, "global_step": 186875, "epoch": 2099, "lr": 2.217432723043612e-05}
{"train_loss": 0.07486363500356674, "global_step": 186876, "epoch": 2099, "lr": 2.2173845564760475e-05}
{"train_loss": 0.11912495642900467, "global_step": 186877, "epoch": 2099, "lr": 2.2173363902825732e-05}
{"train_loss": 0.05205825716257095, "global_step": 186878, "epoch": 2099, "lr": 2.2172882244631928e-05}
{"train_loss": 0.0767369344830513, "global_step": 186879, "epoch": 2099, "lr": 2.2172400590179155e-05}
{"train_loss": 0.07084260135889053, "global_step": 186880, "epoch": 2099, "lr": 2.2171918939467447e-05}
{"train_loss": 0.08113820105791092, "global_step": 186881, "epoch": 2099, "lr": 2.2171437292496906e-05}
{"train_loss": 0.08302702009677887, "global_step": 186882, "epoch": 2099, "lr": 2.2170955649267554e-05}
{"train_loss": 0.0835811197757721, "global_step": 186883, "epoch": 2099, "lr": 2.2170474009779496e-05}
{"train_loss": 0.08988038450479507, "global_step": 186884, "epoch": 2099, "lr": 2.2169992374032762e-05}
{"train_loss": 0.0966801866889, "global_step": 186885, "epoch": 2099, "lr": 2.216951074202745e-05}
{"train_loss": 0.1014951840043068, "global_step": 186886, "epoch": 2099, "lr": 2.216902911376359e-05}
{"train_loss": 0.04199377819895744, "global_step": 186887, "epoch": 2099, "lr": 2.216854748924129e-05}
{"train_loss": 0.09736916422843933, "global_step": 186888, "epoch": 2099, "lr": 2.216806586846058e-05}
{"train_loss": 0.04879673197865486, "global_step": 186889, "epoch": 2099, "lr": 2.216758425142152e-05}
{"train_loss": 0.04579213634133339, "global_step": 186890, "epoch": 2099, "lr": 2.2167102638124203e-05}
{"train_loss": 0.04379740357398987, "global_step": 186891, "epoch": 2099, "lr": 2.216662102856867e-05}
{"train_loss": 0.07094763964414597, "global_step": 186892, "epoch": 2099, "lr": 2.2166139422755012e-05}
{"train_loss": 0.06563301384449005, "global_step": 186893, "epoch": 2099, "lr": 2.2165657820683256e-05}
{"train_loss": 0.11579713225364685, "global_step": 186894, "epoch": 2099, "lr": 2.2165176222353508e-05}
{"train_loss": 0.06501007825136185, "global_step": 186895, "epoch": 2099, "lr": 2.2164694627765792e-05}
{"train_loss": 0.08997707813978195, "global_step": 186896, "epoch": 2099, "lr": 2.216421303692021e-05}
{"train_loss": 0.07798150926828384, "global_step": 186897, "epoch": 2099, "lr": 2.2163731449816793e-05}
{"train_loss": 0.07694641500711441, "global_step": 186898, "epoch": 2099, "lr": 2.2163249866455642e-05}
{"train_loss": 0.07545588441779104, "global_step": 186899, "epoch": 2099, "lr": 2.2162768286836787e-05, "val_loss": 7.764528751373291}
{"train_loss": 0.13518042862415314, "global_step": 186900, "epoch": 2100, "lr": 2.2162286710960328e-05}
{"train_loss": 0.08340173959732056, "global_step": 186901, "epoch": 2100, "lr": 2.216180513882629e-05}
{"train_loss": 0.12415177375078201, "global_step": 186902, "epoch": 2100, "lr": 2.2161323570434755e-05}
{"train_loss": 0.06336087733507156, "global_step": 186903, "epoch": 2100, "lr": 2.2160842005785816e-05}
{"train_loss": 0.049191854894161224, "global_step": 186904, "epoch": 2100, "lr": 2.2160360444879484e-05}
{"train_loss": 0.07378575950860977, "global_step": 186905, "epoch": 2100, "lr": 2.215987888771588e-05}
{"train_loss": 0.04458630084991455, "global_step": 186906, "epoch": 2100, "lr": 2.215939733429502e-05}
{"train_loss": 0.031086409464478493, "global_step": 186907, "epoch": 2100, "lr": 2.2158915784617007e-05}
{"train_loss": 0.08098416775465012, "global_step": 186908, "epoch": 2100, "lr": 2.2158434238681868e-05}
{"train_loss": 0.05475940927863121, "global_step": 186909, "epoch": 2100, "lr": 2.215795269648971e-05}
{"train_loss": 0.0918012261390686, "global_step": 186910, "epoch": 2100, "lr": 2.2157471158040556e-05}
{"train_loss": 0.09062069654464722, "global_step": 186911, "epoch": 2100, "lr": 2.2156989623334512e-05}
{"train_loss": 0.06060033291578293, "global_step": 186912, "epoch": 2100, "lr": 2.2156508092371598e-05}
{"train_loss": 0.05495959520339966, "global_step": 186913, "epoch": 2100, "lr": 2.2156026565151923e-05}
{"train_loss": 0.10159996151924133, "global_step": 186914, "epoch": 2100, "lr": 2.2155545041675513e-05}
{"train_loss": 0.07401353120803833, "global_step": 186915, "epoch": 2100, "lr": 2.2155063521942464e-05}
{"train_loss": 0.09861499816179276, "global_step": 186916, "epoch": 2100, "lr": 2.215458200595282e-05}
{"train_loss": 0.15409418940544128, "global_step": 186917, "epoch": 2100, "lr": 2.215410049370666e-05}
{"train_loss": 0.04104779660701752, "global_step": 186918, "epoch": 2100, "lr": 2.2153618985204028e-05}
{"train_loss": 0.053392574191093445, "global_step": 186919, "epoch": 2100, "lr": 2.2153137480445024e-05}
{"train_loss": 0.08210767060518265, "global_step": 186920, "epoch": 2100, "lr": 2.215265597942967e-05}
{"train_loss": 0.06995321065187454, "global_step": 186921, "epoch": 2100, "lr": 2.215217448215805e-05}
{"train_loss": 0.09429224580526352, "global_step": 186922, "epoch": 2100, "lr": 2.215169298863025e-05}
{"train_loss": 0.0893232598900795, "global_step": 186923, "epoch": 2100, "lr": 2.2151211498846318e-05}
{"train_loss": 0.11264467984437943, "global_step": 186924, "epoch": 2100, "lr": 2.2150730012806298e-05}
{"train_loss": 0.05460139364004135, "global_step": 186925, "epoch": 2100, "lr": 2.2150248530510285e-05}
{"train_loss": 0.04165478050708771, "global_step": 186926, "epoch": 2100, "lr": 2.214976705195832e-05}
{"train_loss": 0.06223528832197189, "global_step": 186927, "epoch": 2100, "lr": 2.2149285577150497e-05}
{"train_loss": 0.08370620012283325, "global_step": 186928, "epoch": 2100, "lr": 2.2148804106086863e-05}
{"train_loss": 0.02992640808224678, "global_step": 186929, "epoch": 2100, "lr": 2.214832263876746e-05}
{"train_loss": 0.09022095054388046, "global_step": 186930, "epoch": 2100, "lr": 2.2147841175192398e-05}
{"train_loss": 0.14421752095222473, "global_step": 186931, "epoch": 2100, "lr": 2.2147359715361698e-05}
{"train_loss": 0.07962089031934738, "global_step": 186932, "epoch": 2100, "lr": 2.2146878259275467e-05}
{"train_loss": 0.05987134203314781, "global_step": 186933, "epoch": 2100, "lr": 2.2146396806933728e-05}
{"train_loss": 0.091457299888134, "global_step": 186934, "epoch": 2100, "lr": 2.214591535833659e-05}
{"train_loss": 0.07442965358495712, "global_step": 186935, "epoch": 2100, "lr": 2.214543391348407e-05}
{"train_loss": 0.08666259050369263, "global_step": 186936, "epoch": 2100, "lr": 2.214495247237628e-05}
{"train_loss": 0.048156291246414185, "global_step": 186937, "epoch": 2100, "lr": 2.2144471035013237e-05}
{"train_loss": 0.07071975618600845, "global_step": 186938, "epoch": 2100, "lr": 2.2143989601395053e-05}
{"train_loss": 0.06324931234121323, "global_step": 186939, "epoch": 2100, "lr": 2.2143508171521754e-05}
{"train_loss": 0.05497194826602936, "global_step": 186940, "epoch": 2100, "lr": 2.214302674539342e-05}
{"train_loss": 0.07340133935213089, "global_step": 186941, "epoch": 2100, "lr": 2.214254532301013e-05}
{"train_loss": 0.03885013610124588, "global_step": 186942, "epoch": 2100, "lr": 2.214206390437192e-05}
{"train_loss": 0.08108693361282349, "global_step": 186943, "epoch": 2100, "lr": 2.2141582489478886e-05}
{"train_loss": 0.06136246770620346, "global_step": 186944, "epoch": 2100, "lr": 2.2141101078331062e-05}
{"train_loss": 0.054276514798402786, "global_step": 186945, "epoch": 2100, "lr": 2.214061967092854e-05}
{"train_loss": 0.060416534543037415, "global_step": 186946, "epoch": 2100, "lr": 2.214013826727136e-05}
{"train_loss": 0.09282118827104568, "global_step": 186947, "epoch": 2100, "lr": 2.2139656867359614e-05}
{"train_loss": 0.09217025339603424, "global_step": 186948, "epoch": 2100, "lr": 2.2139175471193334e-05}
{"train_loss": 0.0832473561167717, "global_step": 186949, "epoch": 2100, "lr": 2.2138694078772626e-05}
{"train_loss": 0.055366888642311096, "global_step": 186950, "epoch": 2100, "lr": 2.213821269009751e-05}
{"train_loss": 0.081329345703125, "global_step": 186951, "epoch": 2100, "lr": 2.2137731305168086e-05}
{"train_loss": 0.078641377389431, "global_step": 186952, "epoch": 2100, "lr": 2.213724992398439e-05}
{"train_loss": 0.05298827588558197, "global_step": 186953, "epoch": 2100, "lr": 2.2136768546546515e-05}
{"train_loss": 0.10922904312610626, "global_step": 186954, "epoch": 2100, "lr": 2.2136287172854496e-05}
{"train_loss": 0.07980429381132126, "global_step": 186955, "epoch": 2100, "lr": 2.2135805802908432e-05}
{"train_loss": 0.07357257604598999, "global_step": 186956, "epoch": 2100, "lr": 2.2135324436708348e-05}
{"train_loss": 0.07748836278915405, "global_step": 186957, "epoch": 2100, "lr": 2.2134843074254335e-05}
{"train_loss": 0.06657243520021439, "global_step": 186958, "epoch": 2100, "lr": 2.213436171554647e-05}
{"train_loss": 0.07392754405736923, "global_step": 186959, "epoch": 2100, "lr": 2.213388036058478e-05}
{"train_loss": 0.07405571639537811, "global_step": 186960, "epoch": 2100, "lr": 2.2133399009369365e-05}
{"train_loss": 0.06647671014070511, "global_step": 186961, "epoch": 2100, "lr": 2.2132917661900282e-05}
{"train_loss": 0.020780811086297035, "global_step": 186962, "epoch": 2100, "lr": 2.213243631817757e-05}
{"train_loss": 0.1075308546423912, "global_step": 186963, "epoch": 2100, "lr": 2.2131954978201324e-05}
{"train_loss": 0.05324281379580498, "global_step": 186964, "epoch": 2100, "lr": 2.2131473641971583e-05}
{"train_loss": 0.11168411374092102, "global_step": 186965, "epoch": 2100, "lr": 2.2130992309488448e-05}
{"train_loss": 0.05150701850652695, "global_step": 186966, "epoch": 2100, "lr": 2.2130510980751957e-05}
{"train_loss": 0.08528245240449905, "global_step": 186967, "epoch": 2100, "lr": 2.2130029655762156e-05}
{"train_loss": 0.06554572284221649, "global_step": 186968, "epoch": 2100, "lr": 2.2129548334519156e-05}
{"train_loss": 0.10755506902933121, "global_step": 186969, "epoch": 2100, "lr": 2.212906701702298e-05}
{"train_loss": 0.09751690924167633, "global_step": 186970, "epoch": 2100, "lr": 2.2128585703273735e-05}
{"train_loss": 0.06539476662874222, "global_step": 186971, "epoch": 2100, "lr": 2.2128104393271438e-05}
{"train_loss": 0.05118321627378464, "global_step": 186972, "epoch": 2100, "lr": 2.2127623087016197e-05}
{"train_loss": 0.06524765491485596, "global_step": 186973, "epoch": 2100, "lr": 2.2127141784508037e-05}
{"train_loss": 0.09913406521081924, "global_step": 186974, "epoch": 2100, "lr": 2.2126660485747063e-05}
{"train_loss": 0.08671263605356216, "global_step": 186975, "epoch": 2100, "lr": 2.2126179190733304e-05}
{"train_loss": 0.07471520453691483, "global_step": 186976, "epoch": 2100, "lr": 2.2125697899466834e-05}
{"train_loss": 0.0929039865732193, "global_step": 186977, "epoch": 2100, "lr": 2.2125216611947748e-05}
{"train_loss": 0.11518274247646332, "global_step": 186978, "epoch": 2100, "lr": 2.2124735328176065e-05}
{"train_loss": 0.10867618769407272, "global_step": 186979, "epoch": 2100, "lr": 2.2124254048151894e-05}
{"train_loss": 0.09297443926334381, "global_step": 186980, "epoch": 2100, "lr": 2.2123772771875256e-05}
{"train_loss": 0.07442285120487213, "global_step": 186981, "epoch": 2100, "lr": 2.2123291499346255e-05}
{"train_loss": 0.08141391724348068, "global_step": 186982, "epoch": 2100, "lr": 2.212281023056492e-05}
{"train_loss": 0.06854911893606186, "global_step": 186983, "epoch": 2100, "lr": 2.2122328965531352e-05}
{"train_loss": 0.061096008867025375, "global_step": 186984, "epoch": 2100, "lr": 2.212184770424558e-05}
{"train_loss": 0.0758427157998085, "global_step": 186985, "epoch": 2100, "lr": 2.2121366446707702e-05}
{"train_loss": 0.08735205233097076, "global_step": 186986, "epoch": 2100, "lr": 2.212088519291775e-05}
{"train_loss": 0.056387245655059814, "global_step": 186987, "epoch": 2100, "lr": 2.2120403942875824e-05}
{"train_loss": 0.07593631951577878, "global_step": 186988, "epoch": 2100, "lr": 2.211992269658195e-05, "train/sim_max_reward_0": 0.4231146722024454, "train/sim_max_reward_1": 0.013469553543733338, "train/sim_max_reward_2": 1.0, "train/sim_max_reward_3": 0.8007871610480297, "train/sim_max_reward_4": 0.14056479104619082, "train/sim_max_reward_5": 0.5693075841546764, "test/sim_max_reward_4500000": 0.9992861500821725, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9796489667468528, "test/sim_max_reward_4500003": 0.560785204684805, "test/sim_max_reward_4500004": 0.5836886533983009, "test/sim_max_reward_4500005": 0.9600733138728507, "test/sim_max_reward_4500006": 0.8572529125937318, "test/sim_max_reward_4500007": 0.9401689685125924, "test/sim_max_reward_4500008": 0.3421293690667855, "test/sim_max_reward_4500009": 0.05004358612352402, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0072783052635949025, "test/sim_max_reward_4500013": 0.6813103595673824, "test/sim_max_reward_4500014": 0.9401547279044573, "test/sim_max_reward_4500015": 0.9457546611973151, "test/sim_max_reward_4500016": 0.9328491846068843, "test/sim_max_reward_4500017": 0.07561272544748421, "test/sim_max_reward_4500018": 0.9588728782540699, "test/sim_max_reward_4500019": 0.9437810235520971, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9538207067205382, "test/sim_max_reward_4500022": 0.9986175249614015, "test/sim_max_reward_4500023": 0.9110986438630126, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9124774296101891, "test/sim_max_reward_4500026": 0.218412342313675, "test/sim_max_reward_4500027": 0.9455613144477237, "test/sim_max_reward_4500028": 0.6506391538677667, "test/sim_max_reward_4500029": 0.2919904434585916, "test/sim_max_reward_4500030": 0.9639839831340734, "test/sim_max_reward_4500031": 0.34620403057886867, "test/sim_max_reward_4500032": 0.26665151132095305, "test/sim_max_reward_4500033": 0.9546797158222734, "test/sim_max_reward_4500034": 0.41281387335082154, "test/sim_max_reward_4500035": 0.5625935352556122, "test/sim_max_reward_4500036": 0.9023100207481809, "test/sim_max_reward_4500037": 0.5325124130208219, "test/sim_max_reward_4500038": 0.9391978619238125, "test/sim_max_reward_4500039": 0.9500062589684763, "test/sim_max_reward_4500040": 0.7653812550254844, "test/sim_max_reward_4500041": 0.5964365723530692, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9020992232729141, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.43113632384689404, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.23744763017394224, "test/sim_max_reward_4500049": 0.9502034572664686, "train/mean_score": 0.491207293665846, "test/mean_score": 0.6070993243236094, "val_loss": 7.777390480041504, "train_action_mse_error": 22.218786239624023}
{"train_loss": 0.046209413558244705, "global_step": 186989, "epoch": 2101, "lr": 2.2119441454036232e-05}
{"train_loss": 0.06362274289131165, "global_step": 186990, "epoch": 2101, "lr": 2.2118960215238692e-05}
{"train_loss": 0.0855850875377655, "global_step": 186991, "epoch": 2101, "lr": 2.2118478980189445e-05}
{"train_loss": 0.0922364741563797, "global_step": 186992, "epoch": 2101, "lr": 2.2117997748888503e-05}
{"train_loss": 0.03384502977132797, "global_step": 186993, "epoch": 2101, "lr": 2.2117516521335962e-05}
{"train_loss": 0.0682418942451477, "global_step": 186994, "epoch": 2101, "lr": 2.2117035297531898e-05}
{"train_loss": 0.06285742670297623, "global_step": 186995, "epoch": 2101, "lr": 2.2116554077476337e-05}
{"train_loss": 0.09463340789079666, "global_step": 186996, "epoch": 2101, "lr": 2.2116072861169385e-05}
{"train_loss": 0.09804318845272064, "global_step": 186997, "epoch": 2101, "lr": 2.2115591648611066e-05}
{"train_loss": 0.10498613864183426, "global_step": 186998, "epoch": 2101, "lr": 2.211511043980149e-05}
{"train_loss": 0.07493562251329422, "global_step": 186999, "epoch": 2101, "lr": 2.2114629234740696e-05}
{"train_loss": 0.09463304281234741, "global_step": 187000, "epoch": 2101, "lr": 2.2114148033428735e-05}
{"train_loss": 0.10070037096738815, "global_step": 187001, "epoch": 2101, "lr": 2.2113666835865697e-05}
{"train_loss": 0.04961397498846054, "global_step": 187002, "epoch": 2101, "lr": 2.2113185642051627e-05}
{"train_loss": 0.10467824339866638, "global_step": 187003, "epoch": 2101, "lr": 2.211270445198661e-05}
{"train_loss": 0.05373481288552284, "global_step": 187004, "epoch": 2101, "lr": 2.2112223265670706e-05}
{"train_loss": 0.09869145601987839, "global_step": 187005, "epoch": 2101, "lr": 2.211174208310395e-05}
{"train_loss": 0.10531297326087952, "global_step": 187006, "epoch": 2101, "lr": 2.2111260904286452e-05}
{"train_loss": 0.08124129474163055, "global_step": 187007, "epoch": 2101, "lr": 2.2110779729218234e-05}
{"train_loss": 0.0917648896574974, "global_step": 187008, "epoch": 2101, "lr": 2.2110298557899405e-05}
{"train_loss": 0.029579797759652138, "global_step": 187009, "epoch": 2101, "lr": 2.2109817390329983e-05}
{"train_loss": 0.061191439628601074, "global_step": 187010, "epoch": 2101, "lr": 2.210933622651008e-05}
{"train_loss": 0.05701844394207001, "global_step": 187011, "epoch": 2101, "lr": 2.2108855066439714e-05}
{"train_loss": 0.10604818910360336, "global_step": 187012, "epoch": 2101, "lr": 2.2108373910118975e-05}
{"train_loss": 0.08490500599145889, "global_step": 187013, "epoch": 2101, "lr": 2.210789275754794e-05}
{"train_loss": 0.1279282569885254, "global_step": 187014, "epoch": 2101, "lr": 2.2107411608726643e-05}
{"train_loss": 0.07945183664560318, "global_step": 187015, "epoch": 2101, "lr": 2.2106930463655185e-05}
{"train_loss": 0.08531482517719269, "global_step": 187016, "epoch": 2101, "lr": 2.210644932233359e-05}
{"train_loss": 0.11631820350885391, "global_step": 187017, "epoch": 2101, "lr": 2.210596818476196e-05}
{"train_loss": 0.03954021632671356, "global_step": 187018, "epoch": 2101, "lr": 2.2105487050940322e-05}
{"train_loss": 0.07550467550754547, "global_step": 187019, "epoch": 2101, "lr": 2.2105005920868782e-05}
{"train_loss": 0.03562871739268303, "global_step": 187020, "epoch": 2101, "lr": 2.2104524794547364e-05}
{"train_loss": 0.09659066051244736, "global_step": 187021, "epoch": 2101, "lr": 2.2104043671976175e-05}
{"train_loss": 0.050841815769672394, "global_step": 187022, "epoch": 2101, "lr": 2.2103562553155238e-05}
{"train_loss": 0.0770399272441864, "global_step": 187023, "epoch": 2101, "lr": 2.2103081438084654e-05}
{"train_loss": 0.12670911848545074, "global_step": 187024, "epoch": 2101, "lr": 2.2102600326764454e-05}
{"train_loss": 0.08161241561174393, "global_step": 187025, "epoch": 2101, "lr": 2.210211921919474e-05}
{"train_loss": 0.0578971765935421, "global_step": 187026, "epoch": 2101, "lr": 2.2101638115375533e-05}
{"train_loss": 0.10191550105810165, "global_step": 187027, "epoch": 2101, "lr": 2.210115701530694e-05}
{"train_loss": 0.05389126390218735, "global_step": 187028, "epoch": 2101, "lr": 2.2100675918988987e-05}
{"train_loss": 0.08068565279245377, "global_step": 187029, "epoch": 2101, "lr": 2.210019482642176e-05}
{"train_loss": 0.07178162038326263, "global_step": 187030, "epoch": 2101, "lr": 2.2099713737605337e-05}
{"train_loss": 0.13492831587791443, "global_step": 187031, "epoch": 2101, "lr": 2.2099232652539753e-05}
{"train_loss": 0.07466226071119308, "global_step": 187032, "epoch": 2101, "lr": 2.2098751571225106e-05}
{"train_loss": 0.08593130856752396, "global_step": 187033, "epoch": 2101, "lr": 2.2098270493661423e-05}
{"train_loss": 0.04607820138335228, "global_step": 187034, "epoch": 2101, "lr": 2.20977894198488e-05}
{"train_loss": 0.07372201979160309, "global_step": 187035, "epoch": 2101, "lr": 2.2097308349787278e-05}
{"train_loss": 0.0626511201262474, "global_step": 187036, "epoch": 2101, "lr": 2.2096827283476946e-05}
{"train_loss": 0.07047230005264282, "global_step": 187037, "epoch": 2101, "lr": 2.209634622091784e-05}
{"train_loss": 0.050410784780979156, "global_step": 187038, "epoch": 2101, "lr": 2.209586516211006e-05}
{"train_loss": 0.08574528992176056, "global_step": 187039, "epoch": 2101, "lr": 2.2095384107053646e-05}
{"train_loss": 0.04955257847905159, "global_step": 187040, "epoch": 2101, "lr": 2.2094903055748655e-05}
{"train_loss": 0.08301147818565369, "global_step": 187041, "epoch": 2101, "lr": 2.2094422008195176e-05}
{"train_loss": 0.09762414544820786, "global_step": 187042, "epoch": 2101, "lr": 2.209394096439325e-05}
{"train_loss": 0.07859676331281662, "global_step": 187043, "epoch": 2101, "lr": 2.2093459924342968e-05}
{"train_loss": 0.07471875846385956, "global_step": 187044, "epoch": 2101, "lr": 2.209297888804438e-05}
{"train_loss": 0.09878010302782059, "global_step": 187045, "epoch": 2101, "lr": 2.2092497855497535e-05}
{"train_loss": 0.09884551912546158, "global_step": 187046, "epoch": 2101, "lr": 2.209201682670253e-05}
{"train_loss": 0.10110814869403839, "global_step": 187047, "epoch": 2101, "lr": 2.2091535801659397e-05}
{"train_loss": 0.078291155397892, "global_step": 187048, "epoch": 2101, "lr": 2.2091054780368215e-05}
{"train_loss": 0.06758497655391693, "global_step": 187049, "epoch": 2101, "lr": 2.2090573762829064e-05}
{"train_loss": 0.06906149536371231, "global_step": 187050, "epoch": 2101, "lr": 2.2090092749041984e-05}
{"train_loss": 0.07168934494256973, "global_step": 187051, "epoch": 2101, "lr": 2.2089611739007064e-05}
{"train_loss": 0.10738932341337204, "global_step": 187052, "epoch": 2101, "lr": 2.2089130732724344e-05}
{"train_loss": 0.07970046997070312, "global_step": 187053, "epoch": 2101, "lr": 2.2088649730193915e-05}
{"train_loss": 0.03562352806329727, "global_step": 187054, "epoch": 2101, "lr": 2.208816873141581e-05}
{"train_loss": 0.08272262662649155, "global_step": 187055, "epoch": 2101, "lr": 2.2087687736390122e-05}
{"train_loss": 0.07436920702457428, "global_step": 187056, "epoch": 2101, "lr": 2.2087206745116897e-05}
{"train_loss": 0.09742894023656845, "global_step": 187057, "epoch": 2101, "lr": 2.2086725757596217e-05}
{"train_loss": 0.0963071957230568, "global_step": 187058, "epoch": 2101, "lr": 2.2086244773828117e-05}
{"train_loss": 0.0887560024857521, "global_step": 187059, "epoch": 2101, "lr": 2.20857637938127e-05}
{"train_loss": 0.0411897674202919, "global_step": 187060, "epoch": 2101, "lr": 2.208528281755e-05}
{"train_loss": 0.0866907611489296, "global_step": 187061, "epoch": 2101, "lr": 2.208480184504011e-05}
{"train_loss": 0.06799948215484619, "global_step": 187062, "epoch": 2101, "lr": 2.2084320876283054e-05}
{"train_loss": 0.1050214096903801, "global_step": 187063, "epoch": 2101, "lr": 2.2083839911278946e-05}
{"train_loss": 0.10740786045789719, "global_step": 187064, "epoch": 2101, "lr": 2.2083358950027804e-05}
{"train_loss": 0.06229456886649132, "global_step": 187065, "epoch": 2101, "lr": 2.2082877992529733e-05}
{"train_loss": 0.07304907590150833, "global_step": 187066, "epoch": 2101, "lr": 2.2082397038784757e-05}
{"train_loss": 0.0982956662774086, "global_step": 187067, "epoch": 2101, "lr": 2.208191608879297e-05}
{"train_loss": 0.07597822695970535, "global_step": 187068, "epoch": 2101, "lr": 2.208143514255444e-05}
{"train_loss": 0.07512886077165604, "global_step": 187069, "epoch": 2101, "lr": 2.2080954200069204e-05}
{"train_loss": 0.08002754300832748, "global_step": 187070, "epoch": 2101, "lr": 2.2080473261337365e-05}
{"train_loss": 0.045127592980861664, "global_step": 187071, "epoch": 2101, "lr": 2.2079992326358945e-05}
{"train_loss": 0.07686001807451248, "global_step": 187072, "epoch": 2101, "lr": 2.207951139513405e-05}
{"train_loss": 0.07554842531681061, "global_step": 187073, "epoch": 2101, "lr": 2.2079030467662705e-05}
{"train_loss": 0.05766145884990692, "global_step": 187074, "epoch": 2101, "lr": 2.2078549543945015e-05}
{"train_loss": 0.05212792381644249, "global_step": 187075, "epoch": 2101, "lr": 2.2078068623981e-05}
{"train_loss": 0.0739232525229454, "global_step": 187076, "epoch": 2101, "lr": 2.207758770777077e-05}
{"train_loss": 0.0778367556840851, "global_step": 187077, "epoch": 2101, "lr": 2.2077106795314368e-05, "val_loss": 7.741305351257324}
{"train_loss": 0.07158183306455612, "global_step": 187078, "epoch": 2102, "lr": 2.207662588661184e-05}
{"train_loss": 0.04539742320775986, "global_step": 187079, "epoch": 2102, "lr": 2.207614498166329e-05}
{"train_loss": 0.1121424064040184, "global_step": 187080, "epoch": 2102, "lr": 2.2075664080468737e-05}
{"train_loss": 0.06713055819272995, "global_step": 187081, "epoch": 2102, "lr": 2.207518318302829e-05}
{"train_loss": 0.08102060854434967, "global_step": 187082, "epoch": 2102, "lr": 2.2074702289342002e-05}
{"train_loss": 0.12279674410820007, "global_step": 187083, "epoch": 2102, "lr": 2.2074221399409906e-05}
{"train_loss": 0.032189011573791504, "global_step": 187084, "epoch": 2102, "lr": 2.2073740513232094e-05}
{"train_loss": 0.07435501366853714, "global_step": 187085, "epoch": 2102, "lr": 2.207325963080864e-05}
{"train_loss": 0.06387697905302048, "global_step": 187086, "epoch": 2102, "lr": 2.2072778752139585e-05}
{"train_loss": 0.13962309062480927, "global_step": 187087, "epoch": 2102, "lr": 2.2072297877225014e-05}
{"train_loss": 0.03692632541060448, "global_step": 187088, "epoch": 2102, "lr": 2.207181700606497e-05}
{"train_loss": 0.13634350895881653, "global_step": 187089, "epoch": 2102, "lr": 2.2071336138659543e-05}
{"train_loss": 0.1256428062915802, "global_step": 187090, "epoch": 2102, "lr": 2.207085527500877e-05}
{"train_loss": 0.10780449956655502, "global_step": 187091, "epoch": 2102, "lr": 2.2070374415112748e-05}
{"train_loss": 0.06329942494630814, "global_step": 187092, "epoch": 2102, "lr": 2.2069893558971504e-05}
{"train_loss": 0.11401854455471039, "global_step": 187093, "epoch": 2102, "lr": 2.2069412706585146e-05}
{"train_loss": 0.061703529208898544, "global_step": 187094, "epoch": 2102, "lr": 2.206893185795369e-05}
{"train_loss": 0.08028377592563629, "global_step": 187095, "epoch": 2102, "lr": 2.2068451013077247e-05}
{"train_loss": 0.05693235620856285, "global_step": 187096, "epoch": 2102, "lr": 2.2067970171955844e-05}
{"train_loss": 0.07746128737926483, "global_step": 187097, "epoch": 2102, "lr": 2.2067489334589576e-05}
{"train_loss": 0.15441635251045227, "global_step": 187098, "epoch": 2102, "lr": 2.2067008500978476e-05}
{"train_loss": 0.08684203028678894, "global_step": 187099, "epoch": 2102, "lr": 2.206652767112265e-05}
{"train_loss": 0.08460057526826859, "global_step": 187100, "epoch": 2102, "lr": 2.2066046845022116e-05}
{"train_loss": 0.05590609833598137, "global_step": 187101, "epoch": 2102, "lr": 2.206556602267698e-05}
{"train_loss": 0.09605447202920914, "global_step": 187102, "epoch": 2102, "lr": 2.206508520408727e-05}
{"train_loss": 0.04209558665752411, "global_step": 187103, "epoch": 2102, "lr": 2.206460438925307e-05}
{"train_loss": 0.12955117225646973, "global_step": 187104, "epoch": 2102, "lr": 2.2064123578174455e-05}
{"train_loss": 0.08908085525035858, "global_step": 187105, "epoch": 2102, "lr": 2.2063642770851466e-05}
{"train_loss": 0.07323615998029709, "global_step": 187106, "epoch": 2102, "lr": 2.20631619672842e-05}
{"train_loss": 0.08675221353769302, "global_step": 187107, "epoch": 2102, "lr": 2.206268116747268e-05}
{"train_loss": 0.10450249910354614, "global_step": 187108, "epoch": 2102, "lr": 2.2062200371417012e-05}
{"train_loss": 0.03818522393703461, "global_step": 187109, "epoch": 2102, "lr": 2.2061719579117223e-05}
{"train_loss": 0.05183390527963638, "global_step": 187110, "epoch": 2102, "lr": 2.2061238790573406e-05}
{"train_loss": 0.07803595066070557, "global_step": 187111, "epoch": 2102, "lr": 2.206075800578561e-05}
{"train_loss": 0.1034255251288414, "global_step": 187112, "epoch": 2102, "lr": 2.206027722475391e-05}
{"train_loss": 0.07033531367778778, "global_step": 187113, "epoch": 2102, "lr": 2.2059796447478352e-05}
{"train_loss": 0.095722496509552, "global_step": 187114, "epoch": 2102, "lr": 2.2059315673959036e-05}
{"train_loss": 0.07451952248811722, "global_step": 187115, "epoch": 2102, "lr": 2.2058834904196002e-05}
{"train_loss": 0.04363200441002846, "global_step": 187116, "epoch": 2102, "lr": 2.2058354138189298e-05}
{"train_loss": 0.027978667989373207, "global_step": 187117, "epoch": 2102, "lr": 2.205787337593902e-05}
{"train_loss": 0.08376534283161163, "global_step": 187118, "epoch": 2102, "lr": 2.205739261744521e-05}
{"train_loss": 0.07964815944433212, "global_step": 187119, "epoch": 2102, "lr": 2.205691186270796e-05}
{"train_loss": 0.05342918634414673, "global_step": 187120, "epoch": 2102, "lr": 2.20564311117273e-05}
{"train_loss": 0.14148588478565216, "global_step": 187121, "epoch": 2102, "lr": 2.2055950364503326e-05}
{"train_loss": 0.05607878044247627, "global_step": 187122, "epoch": 2102, "lr": 2.2055469621036074e-05}
{"train_loss": 0.09574472904205322, "global_step": 187123, "epoch": 2102, "lr": 2.205498888132564e-05}
{"train_loss": 0.0801672637462616, "global_step": 187124, "epoch": 2102, "lr": 2.2054508145372056e-05}
{"train_loss": 0.07706528902053833, "global_step": 187125, "epoch": 2102, "lr": 2.205402741317542e-05}
{"train_loss": 0.08385376632213593, "global_step": 187126, "epoch": 2102, "lr": 2.205354668473576e-05}
{"train_loss": 0.08844483643770218, "global_step": 187127, "epoch": 2102, "lr": 2.2053065960053182e-05}
{"train_loss": 0.04166613146662712, "global_step": 187128, "epoch": 2102, "lr": 2.2052585239127703e-05}
{"train_loss": 0.04988609999418259, "global_step": 187129, "epoch": 2102, "lr": 2.205210452195944e-05}
{"train_loss": 0.10651137679815292, "global_step": 187130, "epoch": 2102, "lr": 2.2051623808548412e-05}
{"train_loss": 0.13339734077453613, "global_step": 187131, "epoch": 2102, "lr": 2.2051143098894716e-05}
{"train_loss": 0.05424419790506363, "global_step": 187132, "epoch": 2102, "lr": 2.205066239299839e-05}
{"train_loss": 0.06738421320915222, "global_step": 187133, "epoch": 2102, "lr": 2.2050181690859522e-05}
{"train_loss": 0.09754562377929688, "global_step": 187134, "epoch": 2102, "lr": 2.2049700992478157e-05}
{"train_loss": 0.06201819330453873, "global_step": 187135, "epoch": 2102, "lr": 2.204922029785438e-05}
{"train_loss": 0.05212313309311867, "global_step": 187136, "epoch": 2102, "lr": 2.204873960698823e-05}
{"train_loss": 0.07385578751564026, "global_step": 187137, "epoch": 2102, "lr": 2.2048258919879806e-05}
{"train_loss": 0.08358999341726303, "global_step": 187138, "epoch": 2102, "lr": 2.204777823652913e-05}
{"train_loss": 0.05796978995203972, "global_step": 187139, "epoch": 2102, "lr": 2.2047297556936292e-05}
{"train_loss": 0.07735107094049454, "global_step": 187140, "epoch": 2102, "lr": 2.2046816881101372e-05}
{"train_loss": 0.13888655602931976, "global_step": 187141, "epoch": 2102, "lr": 2.2046336209024405e-05}
{"train_loss": 0.07409189641475677, "global_step": 187142, "epoch": 2102, "lr": 2.204585554070548e-05}
{"train_loss": 0.046689022332429886, "global_step": 187143, "epoch": 2102, "lr": 2.2045374876144626e-05}
{"train_loss": 0.05863751471042633, "global_step": 187144, "epoch": 2102, "lr": 2.2044894215341955e-05}
{"train_loss": 0.07966078072786331, "global_step": 187145, "epoch": 2102, "lr": 2.204441355829749e-05}
{"train_loss": 0.06956019997596741, "global_step": 187146, "epoch": 2102, "lr": 2.2043932905011328e-05}
{"train_loss": 0.08480658382177353, "global_step": 187147, "epoch": 2102, "lr": 2.20434522554835e-05}
{"train_loss": 0.07909759879112244, "global_step": 187148, "epoch": 2102, "lr": 2.204297160971411e-05}
{"train_loss": 0.060298822820186615, "global_step": 187149, "epoch": 2102, "lr": 2.2042490967703182e-05}
{"train_loss": 0.08002082258462906, "global_step": 187150, "epoch": 2102, "lr": 2.204201032945082e-05}
{"train_loss": 0.03944114223122597, "global_step": 187151, "epoch": 2102, "lr": 2.2041529694957047e-05}
{"train_loss": 0.09474819153547287, "global_step": 187152, "epoch": 2102, "lr": 2.204104906422197e-05}
{"train_loss": 0.08983071893453598, "global_step": 187153, "epoch": 2102, "lr": 2.2040568437245613e-05}
{"train_loss": 0.04669564589858055, "global_step": 187154, "epoch": 2102, "lr": 2.2040087814028082e-05}
{"train_loss": 0.08854719251394272, "global_step": 187155, "epoch": 2102, "lr": 2.2039607194569418e-05}
{"train_loss": 0.03280851989984512, "global_step": 187156, "epoch": 2102, "lr": 2.2039126578869673e-05}
{"train_loss": 0.0236174575984478, "global_step": 187157, "epoch": 2102, "lr": 2.2038645966928946e-05}
{"train_loss": 0.05410826578736305, "global_step": 187158, "epoch": 2102, "lr": 2.203816535874726e-05}
{"train_loss": 0.03630226105451584, "global_step": 187159, "epoch": 2102, "lr": 2.2037684754324717e-05}
{"train_loss": 0.06982150673866272, "global_step": 187160, "epoch": 2102, "lr": 2.2037204153661356e-05}
{"train_loss": 0.05167152360081673, "global_step": 187161, "epoch": 2102, "lr": 2.2036723556757265e-05}
{"train_loss": 0.09593037515878677, "global_step": 187162, "epoch": 2102, "lr": 2.203624296361248e-05}
{"train_loss": 0.09613703191280365, "global_step": 187163, "epoch": 2102, "lr": 2.2035762374227097e-05}
{"train_loss": 0.05947123467922211, "global_step": 187164, "epoch": 2102, "lr": 2.203528178860115e-05}
{"train_loss": 0.0711980015039444, "global_step": 187165, "epoch": 2102, "lr": 2.2034801206734736e-05}
{"train_loss": 0.0769603592082021, "global_step": 187166, "epoch": 2102, "lr": 2.2034320628627882e-05, "val_loss": 7.915962219238281}
{"train_loss": 0.07898581773042679, "global_step": 187167, "epoch": 2103, "lr": 2.2033840054280692e-05}
{"train_loss": 0.09084156155586243, "global_step": 187168, "epoch": 2103, "lr": 2.203335948369319e-05}
{"train_loss": 0.15521353483200073, "global_step": 187169, "epoch": 2103, "lr": 2.2032878916865484e-05}
{"train_loss": 0.05857609957456589, "global_step": 187170, "epoch": 2103, "lr": 2.2032398353797595e-05}
{"train_loss": 0.044039078056812286, "global_step": 187171, "epoch": 2103, "lr": 2.203191779448963e-05}
{"train_loss": 0.04934355989098549, "global_step": 187172, "epoch": 2103, "lr": 2.2031437238941616e-05}
{"train_loss": 0.0600174218416214, "global_step": 187173, "epoch": 2103, "lr": 2.2030956687153655e-05}
{"train_loss": 0.06525755673646927, "global_step": 187174, "epoch": 2103, "lr": 2.2030476139125768e-05}
{"train_loss": 0.09323835372924805, "global_step": 187175, "epoch": 2103, "lr": 2.2029995594858043e-05}
{"train_loss": 0.19693633913993835, "global_step": 187176, "epoch": 2103, "lr": 2.2029515054350564e-05}
{"train_loss": 0.08475824445486069, "global_step": 187177, "epoch": 2103, "lr": 2.2029034517603354e-05}
{"train_loss": 0.06182502210140228, "global_step": 187178, "epoch": 2103, "lr": 2.2028553984616518e-05}
{"train_loss": 0.08622744679450989, "global_step": 187179, "epoch": 2103, "lr": 2.2028073455390086e-05}
{"train_loss": 0.06785362958908081, "global_step": 187180, "epoch": 2103, "lr": 2.2027592929924158e-05}
{"train_loss": 0.06378781795501709, "global_step": 187181, "epoch": 2103, "lr": 2.2027112408218758e-05}
{"train_loss": 0.10094180703163147, "global_step": 187182, "epoch": 2103, "lr": 2.202663189027399e-05}
{"train_loss": 0.04272707924246788, "global_step": 187183, "epoch": 2103, "lr": 2.2026151376089888e-05}
{"train_loss": 0.09361354261636734, "global_step": 187184, "epoch": 2103, "lr": 2.2025670865666543e-05}
{"train_loss": 0.04663247615098953, "global_step": 187185, "epoch": 2103, "lr": 2.2025190359003988e-05}
{"train_loss": 0.07475720345973969, "global_step": 187186, "epoch": 2103, "lr": 2.2024709856102323e-05}
{"train_loss": 0.06776425242424011, "global_step": 187187, "epoch": 2103, "lr": 2.2024229356961572e-05}
{"train_loss": 0.04885225370526314, "global_step": 187188, "epoch": 2103, "lr": 2.202374886158185e-05}
{"train_loss": 0.09555710107088089, "global_step": 187189, "epoch": 2103, "lr": 2.2023268369963167e-05}
{"train_loss": 0.07154061645269394, "global_step": 187190, "epoch": 2103, "lr": 2.202278788210564e-05}
{"train_loss": 0.0458902083337307, "global_step": 187191, "epoch": 2103, "lr": 2.2022307398009283e-05}
{"train_loss": 0.04570813477039337, "global_step": 187192, "epoch": 2103, "lr": 2.202182691767421e-05}
{"train_loss": 0.07756871730089188, "global_step": 187193, "epoch": 2103, "lr": 2.202134644110046e-05}
{"train_loss": 0.10548592358827591, "global_step": 187194, "epoch": 2103, "lr": 2.202086596828808e-05}
{"train_loss": 0.05898508056998253, "global_step": 187195, "epoch": 2103, "lr": 2.2020385499237166e-05}
{"train_loss": 0.06375777721405029, "global_step": 187196, "epoch": 2103, "lr": 2.201990503394776e-05}
{"train_loss": 0.04476332291960716, "global_step": 187197, "epoch": 2103, "lr": 2.201942457241995e-05}
{"train_loss": 0.0865369662642479, "global_step": 187198, "epoch": 2103, "lr": 2.2018944114653774e-05}
{"train_loss": 0.04772500693798065, "global_step": 187199, "epoch": 2103, "lr": 2.2018463660649325e-05}
{"train_loss": 0.04958447813987732, "global_step": 187200, "epoch": 2103, "lr": 2.2017983210406635e-05}
{"train_loss": 0.04462239518761635, "global_step": 187201, "epoch": 2103, "lr": 2.20175027639258e-05}
{"train_loss": 0.09251005947589874, "global_step": 187202, "epoch": 2103, "lr": 2.201702232120686e-05}
{"train_loss": 0.051482200622558594, "global_step": 187203, "epoch": 2103, "lr": 2.20165418822499e-05}
{"train_loss": 0.05490577965974808, "global_step": 187204, "epoch": 2103, "lr": 2.2016061447054964e-05}
{"train_loss": 0.0535256490111351, "global_step": 187205, "epoch": 2103, "lr": 2.2015581015622145e-05}
{"train_loss": 0.06680506467819214, "global_step": 187206, "epoch": 2103, "lr": 2.201510058795147e-05}
{"train_loss": 0.04426904022693634, "global_step": 187207, "epoch": 2103, "lr": 2.2014620164043044e-05}
{"train_loss": 0.0712970644235611, "global_step": 187208, "epoch": 2103, "lr": 2.2014139743896887e-05}
{"train_loss": 0.0722338929772377, "global_step": 187209, "epoch": 2103, "lr": 2.2013659327513116e-05}
{"train_loss": 0.03842891380190849, "global_step": 187210, "epoch": 2103, "lr": 2.2013178914891743e-05}
{"train_loss": 0.09690538048744202, "global_step": 187211, "epoch": 2103, "lr": 2.2012698506032853e-05}
{"train_loss": 0.06300608813762665, "global_step": 187212, "epoch": 2103, "lr": 2.2012218100936537e-05}
{"train_loss": 0.05159313604235649, "global_step": 187213, "epoch": 2103, "lr": 2.201173769960282e-05}
{"train_loss": 0.09466015547513962, "global_step": 187214, "epoch": 2103, "lr": 2.2011257302031803e-05}
{"train_loss": 0.17941361665725708, "global_step": 187215, "epoch": 2103, "lr": 2.2010776908223512e-05}
{"train_loss": 0.07213246077299118, "global_step": 187216, "epoch": 2103, "lr": 2.2010296518178047e-05}
{"train_loss": 0.08112575113773346, "global_step": 187217, "epoch": 2103, "lr": 2.2009816131895438e-05}
{"train_loss": 0.09061741083860397, "global_step": 187218, "epoch": 2103, "lr": 2.2009335749375786e-05}
{"train_loss": 0.03691805526614189, "global_step": 187219, "epoch": 2103, "lr": 2.2008855370619126e-05}
{"train_loss": 0.059056200087070465, "global_step": 187220, "epoch": 2103, "lr": 2.200837499562555e-05}
{"train_loss": 0.08337568491697311, "global_step": 187221, "epoch": 2103, "lr": 2.200789462439508e-05}
{"train_loss": 0.052021339535713196, "global_step": 187222, "epoch": 2103, "lr": 2.2007414256927834e-05}
{"train_loss": 0.1211463138461113, "global_step": 187223, "epoch": 2103, "lr": 2.200693389322383e-05}
{"train_loss": 0.07625015079975128, "global_step": 187224, "epoch": 2103, "lr": 2.2006453533283167e-05}
{"train_loss": 0.037840764969587326, "global_step": 187225, "epoch": 2103, "lr": 2.200597317710588e-05}
{"train_loss": 0.03930458053946495, "global_step": 187226, "epoch": 2103, "lr": 2.200549282469207e-05}
{"train_loss": 0.05177071690559387, "global_step": 187227, "epoch": 2103, "lr": 2.200501247604176e-05}
{"train_loss": 0.03832282871007919, "global_step": 187228, "epoch": 2103, "lr": 2.2004532131155053e-05}
{"train_loss": 0.09505581855773926, "global_step": 187229, "epoch": 2103, "lr": 2.2004051790031978e-05}
{"train_loss": 0.07213381677865982, "global_step": 187230, "epoch": 2103, "lr": 2.2003571452672633e-05}
{"train_loss": 0.09051037579774857, "global_step": 187231, "epoch": 2103, "lr": 2.2003091119077062e-05}
{"train_loss": 0.058294810354709625, "global_step": 187232, "epoch": 2103, "lr": 2.2002610789245326e-05}
{"train_loss": 0.08095463365316391, "global_step": 187233, "epoch": 2103, "lr": 2.200213046317751e-05}
{"train_loss": 0.03059859201312065, "global_step": 187234, "epoch": 2103, "lr": 2.2001650140873647e-05}
{"train_loss": 0.056306142359972, "global_step": 187235, "epoch": 2103, "lr": 2.2001169822333843e-05}
{"train_loss": 0.09166708588600159, "global_step": 187236, "epoch": 2103, "lr": 2.200068950755812e-05}
{"train_loss": 0.057158462703228, "global_step": 187237, "epoch": 2103, "lr": 2.2000209196546583e-05}
{"train_loss": 0.10687854886054993, "global_step": 187238, "epoch": 2103, "lr": 2.199972888929926e-05}
{"train_loss": 0.06522365659475327, "global_step": 187239, "epoch": 2103, "lr": 2.1999248585816246e-05}
{"train_loss": 0.07136823982000351, "global_step": 187240, "epoch": 2103, "lr": 2.1998768286097578e-05}
{"train_loss": 0.09090090543031693, "global_step": 187241, "epoch": 2103, "lr": 2.1998287990143345e-05}
{"train_loss": 0.08238894492387772, "global_step": 187242, "epoch": 2103, "lr": 2.1997807697953587e-05}
{"train_loss": 0.0475764125585556, "global_step": 187243, "epoch": 2103, "lr": 2.19973274095284e-05}
{"train_loss": 0.07339399307966232, "global_step": 187244, "epoch": 2103, "lr": 2.1996847124867813e-05}
{"train_loss": 0.033083315938711166, "global_step": 187245, "epoch": 2103, "lr": 2.199636684397193e-05}
{"train_loss": 0.11328663676977158, "global_step": 187246, "epoch": 2103, "lr": 2.1995886566840773e-05}
{"train_loss": 0.14652679860591888, "global_step": 187247, "epoch": 2103, "lr": 2.199540629347443e-05}
{"train_loss": 0.08979877829551697, "global_step": 187248, "epoch": 2103, "lr": 2.199492602387298e-05}
{"train_loss": 0.06774721294641495, "global_step": 187249, "epoch": 2103, "lr": 2.1994445758036453e-05}
{"train_loss": 0.06194489076733589, "global_step": 187250, "epoch": 2103, "lr": 2.1993965495964947e-05}
{"train_loss": 0.043842196464538574, "global_step": 187251, "epoch": 2103, "lr": 2.1993485237658496e-05}
{"train_loss": 0.08340104669332504, "global_step": 187252, "epoch": 2103, "lr": 2.1993004983117198e-05}
{"train_loss": 0.07515329122543335, "global_step": 187253, "epoch": 2103, "lr": 2.199252473234108e-05}
{"train_loss": 0.07029689103364944, "global_step": 187254, "epoch": 2103, "lr": 2.1992044485330244e-05}
{"train_loss": 0.07302191711208794, "global_step": 187255, "epoch": 2103, "lr": 2.1991564242084716e-05, "val_loss": 7.745131969451904}
{"train_loss": 0.10270782560110092, "global_step": 187256, "epoch": 2104, "lr": 2.19910840026046e-05}
{"train_loss": 0.03638382628560066, "global_step": 187257, "epoch": 2104, "lr": 2.1990603766889926e-05}
{"train_loss": 0.06071389466524124, "global_step": 187258, "epoch": 2104, "lr": 2.199012353494079e-05}
{"train_loss": 0.06280270218849182, "global_step": 187259, "epoch": 2104, "lr": 2.1989643306757225e-05}
{"train_loss": 0.060279179364442825, "global_step": 187260, "epoch": 2104, "lr": 2.1989163082339326e-05}
{"train_loss": 0.06839988380670547, "global_step": 187261, "epoch": 2104, "lr": 2.198868286168712e-05}
{"train_loss": 0.05903305485844612, "global_step": 187262, "epoch": 2104, "lr": 2.198820264480072e-05}
{"train_loss": 0.10753370821475983, "global_step": 187263, "epoch": 2104, "lr": 2.198772243168014e-05}
{"train_loss": 0.056623682379722595, "global_step": 187264, "epoch": 2104, "lr": 2.1987242222325495e-05}
{"train_loss": 0.07396604865789413, "global_step": 187265, "epoch": 2104, "lr": 2.19867620167368e-05}
{"train_loss": 0.09057237952947617, "global_step": 187266, "epoch": 2104, "lr": 2.1986281814914167e-05}
{"train_loss": 0.035667017102241516, "global_step": 187267, "epoch": 2104, "lr": 2.1985801616857614e-05}
{"train_loss": 0.07919663935899734, "global_step": 187268, "epoch": 2104, "lr": 2.1985321422567246e-05}
{"train_loss": 0.0588577538728714, "global_step": 187269, "epoch": 2104, "lr": 2.1984841232043096e-05}
{"train_loss": 0.044074732810258865, "global_step": 187270, "epoch": 2104, "lr": 2.1984361045285257e-05}
{"train_loss": 0.08445584774017334, "global_step": 187271, "epoch": 2104, "lr": 2.198388086229378e-05}
{"train_loss": 0.1399696171283722, "global_step": 187272, "epoch": 2104, "lr": 2.1983400683068718e-05}
{"train_loss": 0.051967788487672806, "global_step": 187273, "epoch": 2104, "lr": 2.1982920507610156e-05}
{"train_loss": 0.10289333015680313, "global_step": 187274, "epoch": 2104, "lr": 2.198244033591813e-05}
{"train_loss": 0.058246638625860214, "global_step": 187275, "epoch": 2104, "lr": 2.1981960167992744e-05}
{"train_loss": 0.05643131583929062, "global_step": 187276, "epoch": 2104, "lr": 2.1981480003834025e-05}
{"train_loss": 0.04638618603348732, "global_step": 187277, "epoch": 2104, "lr": 2.1980999843442075e-05}
{"train_loss": 0.09764175862073898, "global_step": 187278, "epoch": 2104, "lr": 2.1980519686816915e-05}
{"train_loss": 0.07952196151018143, "global_step": 187279, "epoch": 2104, "lr": 2.1980039533958653e-05}
{"train_loss": 0.0910002663731575, "global_step": 187280, "epoch": 2104, "lr": 2.1979559384867316e-05}
{"train_loss": 0.050719067454338074, "global_step": 187281, "epoch": 2104, "lr": 2.1979079239542998e-05}
{"train_loss": 0.08370571583509445, "global_step": 187282, "epoch": 2104, "lr": 2.197859909798574e-05}
{"train_loss": 0.05891342833638191, "global_step": 187283, "epoch": 2104, "lr": 2.1978118960195615e-05}
{"train_loss": 0.0625964030623436, "global_step": 187284, "epoch": 2104, "lr": 2.197763882617271e-05}
{"train_loss": 0.023512033745646477, "global_step": 187285, "epoch": 2104, "lr": 2.197715869591705e-05}
{"train_loss": 0.0876961201429367, "global_step": 187286, "epoch": 2104, "lr": 2.1976678569428736e-05}
{"train_loss": 0.10280635952949524, "global_step": 187287, "epoch": 2104, "lr": 2.19761984467078e-05}
{"train_loss": 0.059425920248031616, "global_step": 187288, "epoch": 2104, "lr": 2.197571832775434e-05}
{"train_loss": 0.021021157503128052, "global_step": 187289, "epoch": 2104, "lr": 2.197523821256839e-05}
{"train_loss": 0.04674208536744118, "global_step": 187290, "epoch": 2104, "lr": 2.197475810115004e-05}
{"train_loss": 0.06524240225553513, "global_step": 187291, "epoch": 2104, "lr": 2.1974277993499328e-05}
{"train_loss": 0.08120111376047134, "global_step": 187292, "epoch": 2104, "lr": 2.1973797889616344e-05}
{"train_loss": 0.053045108914375305, "global_step": 187293, "epoch": 2104, "lr": 2.197331778950113e-05}
{"train_loss": 0.08114589750766754, "global_step": 187294, "epoch": 2104, "lr": 2.1972837693153775e-05}
{"train_loss": 0.038992688059806824, "global_step": 187295, "epoch": 2104, "lr": 2.1972357600574317e-05}
{"train_loss": 0.05871298536658287, "global_step": 187296, "epoch": 2104, "lr": 2.1971877511762852e-05}
{"train_loss": 0.07728848606348038, "global_step": 187297, "epoch": 2104, "lr": 2.1971397426719403e-05}
{"train_loss": 0.06910722702741623, "global_step": 187298, "epoch": 2104, "lr": 2.197091734544408e-05}
{"train_loss": 0.06809286773204803, "global_step": 187299, "epoch": 2104, "lr": 2.1970437267936904e-05}
{"train_loss": 0.038004323840141296, "global_step": 187300, "epoch": 2104, "lr": 2.196995719419798e-05}
{"train_loss": 0.10020829737186432, "global_step": 187301, "epoch": 2104, "lr": 2.196947712422734e-05}
{"train_loss": 0.06449376046657562, "global_step": 187302, "epoch": 2104, "lr": 2.1968997058025058e-05}
{"train_loss": 0.03785565122961998, "global_step": 187303, "epoch": 2104, "lr": 2.1968516995591225e-05}
{"train_loss": 0.053922053426504135, "global_step": 187304, "epoch": 2104, "lr": 2.196803693692588e-05}
{"train_loss": 0.03294377774000168, "global_step": 187305, "epoch": 2104, "lr": 2.196755688202907e-05}
{"train_loss": 0.05757157504558563, "global_step": 187306, "epoch": 2104, "lr": 2.1967076830900906e-05}
{"train_loss": 0.1217842549085617, "global_step": 187307, "epoch": 2104, "lr": 2.1966596783541404e-05}
{"train_loss": 0.07501299679279327, "global_step": 187308, "epoch": 2104, "lr": 2.196611673995067e-05}
{"train_loss": 0.1271500438451767, "global_step": 187309, "epoch": 2104, "lr": 2.196563670012875e-05}
{"train_loss": 0.04077723249793053, "global_step": 187310, "epoch": 2104, "lr": 2.196515666407569e-05}
{"train_loss": 0.07885818183422089, "global_step": 187311, "epoch": 2104, "lr": 2.1964676631791593e-05}
{"train_loss": 0.13795435428619385, "global_step": 187312, "epoch": 2104, "lr": 2.1964196603276484e-05}
{"train_loss": 0.035389527678489685, "global_step": 187313, "epoch": 2104, "lr": 2.1963716578530464e-05}
{"train_loss": 0.040281783789396286, "global_step": 187314, "epoch": 2104, "lr": 2.196323655755356e-05}
{"train_loss": 0.09599867463111877, "global_step": 187315, "epoch": 2104, "lr": 2.196275654034588e-05}
{"train_loss": 0.04111591354012489, "global_step": 187316, "epoch": 2104, "lr": 2.196227652690745e-05}
{"train_loss": 0.060049161314964294, "global_step": 187317, "epoch": 2104, "lr": 2.1961796517238365e-05}
{"train_loss": 0.05477878451347351, "global_step": 187318, "epoch": 2104, "lr": 2.1961316511338652e-05}
{"train_loss": 0.12429758161306381, "global_step": 187319, "epoch": 2104, "lr": 2.1960836509208422e-05}
{"train_loss": 0.08715353161096573, "global_step": 187320, "epoch": 2104, "lr": 2.1960356510847695e-05}
{"train_loss": 0.07959648966789246, "global_step": 187321, "epoch": 2104, "lr": 2.1959876516256556e-05}
{"train_loss": 0.08855019509792328, "global_step": 187322, "epoch": 2104, "lr": 2.1959396525435087e-05}
{"train_loss": 0.07928859442472458, "global_step": 187323, "epoch": 2104, "lr": 2.1958916538383324e-05}
{"train_loss": 0.06329170614480972, "global_step": 187324, "epoch": 2104, "lr": 2.195843655510136e-05}
{"train_loss": 0.04080552980303764, "global_step": 187325, "epoch": 2104, "lr": 2.1957956575589217e-05}
{"train_loss": 0.04579632729291916, "global_step": 187326, "epoch": 2104, "lr": 2.1957476599847004e-05}
{"train_loss": 0.12146828323602676, "global_step": 187327, "epoch": 2104, "lr": 2.1956996627874755e-05}
{"train_loss": 0.07410787791013718, "global_step": 187328, "epoch": 2104, "lr": 2.195651665967256e-05}
{"train_loss": 0.06535869836807251, "global_step": 187329, "epoch": 2104, "lr": 2.195603669524045e-05}
{"train_loss": 0.06980277597904205, "global_step": 187330, "epoch": 2104, "lr": 2.1955556734578532e-05}
{"train_loss": 0.08711268752813339, "global_step": 187331, "epoch": 2104, "lr": 2.1955076777686824e-05}
{"train_loss": 0.06264511495828629, "global_step": 187332, "epoch": 2104, "lr": 2.1954596824565436e-05}
{"train_loss": 0.06627175956964493, "global_step": 187333, "epoch": 2104, "lr": 2.195411687521439e-05}
{"train_loss": 0.08756569027900696, "global_step": 187334, "epoch": 2104, "lr": 2.1953636929633792e-05}
{"train_loss": 0.07483318448066711, "global_step": 187335, "epoch": 2104, "lr": 2.1953156987823665e-05}
{"train_loss": 0.09127980470657349, "global_step": 187336, "epoch": 2104, "lr": 2.1952677049784115e-05}
{"train_loss": 0.07756146788597107, "global_step": 187337, "epoch": 2104, "lr": 2.1952197115515167e-05}
{"train_loss": 0.07751672714948654, "global_step": 187338, "epoch": 2104, "lr": 2.19517171850169e-05}
{"train_loss": 0.06488946825265884, "global_step": 187339, "epoch": 2104, "lr": 2.1951237258289404e-05}
{"train_loss": 0.04931275546550751, "global_step": 187340, "epoch": 2104, "lr": 2.1950757335332706e-05}
{"train_loss": 0.05586963891983032, "global_step": 187341, "epoch": 2104, "lr": 2.1950277416146903e-05}
{"train_loss": 0.07502083480358124, "global_step": 187342, "epoch": 2104, "lr": 2.1949797500732044e-05}
{"train_loss": 0.06510990113019943, "global_step": 187343, "epoch": 2104, "lr": 2.1949317589088176e-05}
{"train_loss": 0.07006884908324547, "global_step": 187344, "epoch": 2104, "lr": 2.1948837681215396e-05, "val_loss": 7.687312602996826}
{"train_loss": 0.07063072919845581, "global_step": 187345, "epoch": 2105, "lr": 2.1948357777113732e-05}
{"train_loss": 0.06839833408594131, "global_step": 187346, "epoch": 2105, "lr": 2.194787787678329e-05}
{"train_loss": 0.043850455433130264, "global_step": 187347, "epoch": 2105, "lr": 2.1947397980224116e-05}
{"train_loss": 0.05552568659186363, "global_step": 187348, "epoch": 2105, "lr": 2.1946918087436258e-05}
{"train_loss": 0.08352264016866684, "global_step": 187349, "epoch": 2105, "lr": 2.1946438198419807e-05}
{"train_loss": 0.09462060034275055, "global_step": 187350, "epoch": 2105, "lr": 2.1945958313174797e-05}
{"train_loss": 0.08917252719402313, "global_step": 187351, "epoch": 2105, "lr": 2.1945478431701334e-05}
{"train_loss": 0.09452573955059052, "global_step": 187352, "epoch": 2105, "lr": 2.1944998553999435e-05}
{"train_loss": 0.09807274490594864, "global_step": 187353, "epoch": 2105, "lr": 2.194451868006921e-05}
{"train_loss": 0.07924740016460419, "global_step": 187354, "epoch": 2105, "lr": 2.1944038809910684e-05}
{"train_loss": 0.07273206859827042, "global_step": 187355, "epoch": 2105, "lr": 2.1943558943523962e-05}
{"train_loss": 0.1363144814968109, "global_step": 187356, "epoch": 2105, "lr": 2.1943079080909064e-05}
{"train_loss": 0.07994837313890457, "global_step": 187357, "epoch": 2105, "lr": 2.194259922206608e-05}
{"train_loss": 0.057068757712841034, "global_step": 187358, "epoch": 2105, "lr": 2.1942119366995083e-05}
{"train_loss": 0.10359509289264679, "global_step": 187359, "epoch": 2105, "lr": 2.1941639515696115e-05}
{"train_loss": 0.04851832613348961, "global_step": 187360, "epoch": 2105, "lr": 2.194115966816927e-05}
{"train_loss": 0.06882680207490921, "global_step": 187361, "epoch": 2105, "lr": 2.1940679824414573e-05}
{"train_loss": 0.07347916811704636, "global_step": 187362, "epoch": 2105, "lr": 2.1940199984432124e-05}
{"train_loss": 0.08761069178581238, "global_step": 187363, "epoch": 2105, "lr": 2.193972014822196e-05}
{"train_loss": 0.1495412141084671, "global_step": 187364, "epoch": 2105, "lr": 2.1939240315784177e-05}
{"train_loss": 0.04584604129195213, "global_step": 187365, "epoch": 2105, "lr": 2.1938760487118803e-05}
{"train_loss": 0.0538947619497776, "global_step": 187366, "epoch": 2105, "lr": 2.1938280662225935e-05}
{"train_loss": 0.10895949602127075, "global_step": 187367, "epoch": 2105, "lr": 2.193780084110561e-05}
{"train_loss": 0.04657403379678726, "global_step": 187368, "epoch": 2105, "lr": 2.1937321023757924e-05}
{"train_loss": 0.060075536370277405, "global_step": 187369, "epoch": 2105, "lr": 2.19368412101829e-05}
{"train_loss": 0.07753628492355347, "global_step": 187370, "epoch": 2105, "lr": 2.1936361400380645e-05}
{"train_loss": 0.07621096074581146, "global_step": 187371, "epoch": 2105, "lr": 2.1935881594351187e-05}
{"train_loss": 0.04815578833222389, "global_step": 187372, "epoch": 2105, "lr": 2.193540179209463e-05}
{"train_loss": 0.04838493838906288, "global_step": 187373, "epoch": 2105, "lr": 2.1934921993610992e-05}
{"train_loss": 0.05698590353131294, "global_step": 187374, "epoch": 2105, "lr": 2.193444219890036e-05}
{"train_loss": 0.05656718090176582, "global_step": 187375, "epoch": 2105, "lr": 2.1933962407962828e-05}
{"train_loss": 0.07391151040792465, "global_step": 187376, "epoch": 2105, "lr": 2.1933482620798406e-05}
{"train_loss": 0.07674175500869751, "global_step": 187377, "epoch": 2105, "lr": 2.1933002837407208e-05}
{"train_loss": 0.07195506989955902, "global_step": 187378, "epoch": 2105, "lr": 2.1932523057789257e-05}
{"train_loss": 0.08602778613567352, "global_step": 187379, "epoch": 2105, "lr": 2.1932043281944654e-05}
{"train_loss": 0.10750356316566467, "global_step": 187380, "epoch": 2105, "lr": 2.193156350987342e-05}
{"train_loss": 0.06385243684053421, "global_step": 187381, "epoch": 2105, "lr": 2.1931083741575676e-05}
{"train_loss": 0.07185640186071396, "global_step": 187382, "epoch": 2105, "lr": 2.1930603977051445e-05}
{"train_loss": 0.06132165715098381, "global_step": 187383, "epoch": 2105, "lr": 2.193012421630079e-05}
{"train_loss": 0.10399455577135086, "global_step": 187384, "epoch": 2105, "lr": 2.1929644459323807e-05}
{"train_loss": 0.052709922194480896, "global_step": 187385, "epoch": 2105, "lr": 2.192916470612052e-05}
{"train_loss": 0.16298595070838928, "global_step": 187386, "epoch": 2105, "lr": 2.192868495669103e-05}
{"train_loss": 0.04152855649590492, "global_step": 187387, "epoch": 2105, "lr": 2.192820521103539e-05}
{"train_loss": 0.1497732251882553, "global_step": 187388, "epoch": 2105, "lr": 2.1927725469153642e-05}
{"train_loss": 0.10442392528057098, "global_step": 187389, "epoch": 2105, "lr": 2.1927245731045885e-05}
{"train_loss": 0.10877333581447601, "global_step": 187390, "epoch": 2105, "lr": 2.1926765996712155e-05}
{"train_loss": 0.054487090557813644, "global_step": 187391, "epoch": 2105, "lr": 2.1926286266152547e-05}
{"train_loss": 0.10314647853374481, "global_step": 187392, "epoch": 2105, "lr": 2.1925806539367083e-05}
{"train_loss": 0.09038915485143661, "global_step": 187393, "epoch": 2105, "lr": 2.1925326816355856e-05}
{"train_loss": 0.07241792231798172, "global_step": 187394, "epoch": 2105, "lr": 2.1924847097118944e-05}
{"train_loss": 0.11846142262220383, "global_step": 187395, "epoch": 2105, "lr": 2.1924367381656376e-05}
{"train_loss": 0.06561759859323502, "global_step": 187396, "epoch": 2105, "lr": 2.192388766996825e-05}
{"train_loss": 0.09787794202566147, "global_step": 187397, "epoch": 2105, "lr": 2.1923407962054604e-05}
{"train_loss": 0.07929690182209015, "global_step": 187398, "epoch": 2105, "lr": 2.1922928257915525e-05}
{"train_loss": 0.09480255097150803, "global_step": 187399, "epoch": 2105, "lr": 2.1922448557551052e-05}
{"train_loss": 0.0753495842218399, "global_step": 187400, "epoch": 2105, "lr": 2.1921968860961283e-05}
{"train_loss": 0.02652757428586483, "global_step": 187401, "epoch": 2105, "lr": 2.1921489168146242e-05}
{"train_loss": 0.06817561388015747, "global_step": 187402, "epoch": 2105, "lr": 2.192100947910603e-05}
{"train_loss": 0.0713132992386818, "global_step": 187403, "epoch": 2105, "lr": 2.192052979384068e-05}
{"train_loss": 0.059837352484464645, "global_step": 187404, "epoch": 2105, "lr": 2.1920050112350295e-05}
{"train_loss": 0.0719820111989975, "global_step": 187405, "epoch": 2105, "lr": 2.1919570434634896e-05}
{"train_loss": 0.14284302294254303, "global_step": 187406, "epoch": 2105, "lr": 2.1919090760694582e-05}
{"train_loss": 0.06968080252408981, "global_step": 187407, "epoch": 2105, "lr": 2.1918611090529395e-05}
{"train_loss": 0.05025627464056015, "global_step": 187408, "epoch": 2105, "lr": 2.191813142413942e-05}
{"train_loss": 0.09314204752445221, "global_step": 187409, "epoch": 2105, "lr": 2.1917651761524695e-05}
{"train_loss": 0.09551665931940079, "global_step": 187410, "epoch": 2105, "lr": 2.1917172102685297e-05}
{"train_loss": 0.10489270836114883, "global_step": 187411, "epoch": 2105, "lr": 2.191669244762131e-05}
{"train_loss": 0.08584032207727432, "global_step": 187412, "epoch": 2105, "lr": 2.1916212796332765e-05}
{"train_loss": 0.06294567137956619, "global_step": 187413, "epoch": 2105, "lr": 2.191573314881976e-05}
{"train_loss": 0.05101771280169487, "global_step": 187414, "epoch": 2105, "lr": 2.191525350508233e-05}
{"train_loss": 0.06829562783241272, "global_step": 187415, "epoch": 2105, "lr": 2.1914773865120565e-05}
{"train_loss": 0.09716314822435379, "global_step": 187416, "epoch": 2105, "lr": 2.1914294228934494e-05}
{"train_loss": 0.06638133525848389, "global_step": 187417, "epoch": 2105, "lr": 2.191381459652423e-05}
{"train_loss": 0.05679580941796303, "global_step": 187418, "epoch": 2105, "lr": 2.1913334967889785e-05}
{"train_loss": 0.06361131370067596, "global_step": 187419, "epoch": 2105, "lr": 2.1912855343031276e-05}
{"train_loss": 0.0848412960767746, "global_step": 187420, "epoch": 2105, "lr": 2.191237572194874e-05}
{"train_loss": 0.07115815579891205, "global_step": 187421, "epoch": 2105, "lr": 2.1911896104642222e-05}
{"train_loss": 0.05629865452647209, "global_step": 187422, "epoch": 2105, "lr": 2.1911416491111826e-05}
{"train_loss": 0.062051527202129364, "global_step": 187423, "epoch": 2105, "lr": 2.191093688135758e-05}
{"train_loss": 0.08770444244146347, "global_step": 187424, "epoch": 2105, "lr": 2.191045727537958e-05}
{"train_loss": 0.05437838286161423, "global_step": 187425, "epoch": 2105, "lr": 2.1909977673177885e-05}
{"train_loss": 0.06597545742988586, "global_step": 187426, "epoch": 2105, "lr": 2.1909498074752527e-05}
{"train_loss": 0.051898133009672165, "global_step": 187427, "epoch": 2105, "lr": 2.1909018480103617e-05}
{"train_loss": 0.060405269265174866, "global_step": 187428, "epoch": 2105, "lr": 2.1908538889231173e-05}
{"train_loss": 0.09509553760290146, "global_step": 187429, "epoch": 2105, "lr": 2.1908059302135286e-05}
{"train_loss": 0.06917059421539307, "global_step": 187430, "epoch": 2105, "lr": 2.190757971881604e-05}
{"train_loss": 0.04567408561706543, "global_step": 187431, "epoch": 2105, "lr": 2.1907100139273456e-05}
{"train_loss": 0.08209511637687683, "global_step": 187432, "epoch": 2105, "lr": 2.1906620563507634e-05}
{"train_loss": 0.07724985457287076, "global_step": 187433, "epoch": 2105, "lr": 2.1906140991518614e-05, "val_loss": 7.761457920074463, "train_action_mse_error": 31.339033126831055}
{"train_loss": 0.09655917435884476, "global_step": 187434, "epoch": 2106, "lr": 2.1905661423306483e-05}
{"train_loss": 0.051008693873882294, "global_step": 187435, "epoch": 2106, "lr": 2.1905181858871283e-05}
{"train_loss": 0.08349837362766266, "global_step": 187436, "epoch": 2106, "lr": 2.1904702298213097e-05}
{"train_loss": 0.12441114336252213, "global_step": 187437, "epoch": 2106, "lr": 2.190422274133197e-05}
{"train_loss": 0.07369715720415115, "global_step": 187438, "epoch": 2106, "lr": 2.1903743188227993e-05}
{"train_loss": 0.12062321603298187, "global_step": 187439, "epoch": 2106, "lr": 2.19032636389012e-05}
{"train_loss": 0.07793208211660385, "global_step": 187440, "epoch": 2106, "lr": 2.1902784093351687e-05}
{"train_loss": 0.08014114946126938, "global_step": 187441, "epoch": 2106, "lr": 2.1902304551579483e-05}
{"train_loss": 0.05672669783234596, "global_step": 187442, "epoch": 2106, "lr": 2.1901825013584693e-05}
{"train_loss": 0.0999821349978447, "global_step": 187443, "epoch": 2106, "lr": 2.1901345479367342e-05}
{"train_loss": 0.07229438424110413, "global_step": 187444, "epoch": 2106, "lr": 2.1900865948927528e-05}
{"train_loss": 0.0899888426065445, "global_step": 187445, "epoch": 2106, "lr": 2.1900386422265284e-05}
{"train_loss": 0.0498485304415226, "global_step": 187446, "epoch": 2106, "lr": 2.1899906899380708e-05}
{"train_loss": 0.05259627848863602, "global_step": 187447, "epoch": 2106, "lr": 2.189942738027383e-05}
{"train_loss": 0.08715400844812393, "global_step": 187448, "epoch": 2106, "lr": 2.1898947864944726e-05}
{"train_loss": 0.0645369440317154, "global_step": 187449, "epoch": 2106, "lr": 2.1898468353393485e-05}
{"train_loss": 0.09836388379335403, "global_step": 187450, "epoch": 2106, "lr": 2.1897988845620144e-05}
{"train_loss": 0.07843280583620071, "global_step": 187451, "epoch": 2106, "lr": 2.1897509341624782e-05}
{"train_loss": 0.09753171354532242, "global_step": 187452, "epoch": 2106, "lr": 2.1897029841407446e-05}
{"train_loss": 0.04340650513768196, "global_step": 187453, "epoch": 2106, "lr": 2.189655034496823e-05}
{"train_loss": 0.03760436177253723, "global_step": 187454, "epoch": 2106, "lr": 2.189607085230716e-05}
{"train_loss": 0.06335736066102982, "global_step": 187455, "epoch": 2106, "lr": 2.189559136342434e-05}
{"train_loss": 0.11184294521808624, "global_step": 187456, "epoch": 2106, "lr": 2.1895111878319796e-05}
{"train_loss": 0.07310997694730759, "global_step": 187457, "epoch": 2106, "lr": 2.1894632396993632e-05}
{"train_loss": 0.057137079536914825, "global_step": 187458, "epoch": 2106, "lr": 2.189415291944589e-05}
{"train_loss": 0.07872903347015381, "global_step": 187459, "epoch": 2106, "lr": 2.1893673445676617e-05}
{"train_loss": 0.0818834975361824, "global_step": 187460, "epoch": 2106, "lr": 2.1893193975685915e-05}
{"train_loss": 0.060443174093961716, "global_step": 187461, "epoch": 2106, "lr": 2.1892714509473817e-05}
{"train_loss": 0.10761047899723053, "global_step": 187462, "epoch": 2106, "lr": 2.1892235047040414e-05}
{"train_loss": 0.09220758080482483, "global_step": 187463, "epoch": 2106, "lr": 2.1891755588385765e-05}
{"train_loss": 0.055023517459630966, "global_step": 187464, "epoch": 2106, "lr": 2.18912761335099e-05}
{"train_loss": 0.08746831864118576, "global_step": 187465, "epoch": 2106, "lr": 2.1890796682412917e-05}
{"train_loss": 0.059991367161273956, "global_step": 187466, "epoch": 2106, "lr": 2.1890317235094888e-05}
{"train_loss": 0.056209124624729156, "global_step": 187467, "epoch": 2106, "lr": 2.1889837791555846e-05}
{"train_loss": 0.06455129384994507, "global_step": 187468, "epoch": 2106, "lr": 2.1889358351795893e-05}
{"train_loss": 0.07194440811872482, "global_step": 187469, "epoch": 2106, "lr": 2.1888878915815055e-05}
{"train_loss": 0.05305440351366997, "global_step": 187470, "epoch": 2106, "lr": 2.1888399483613427e-05}
{"train_loss": 0.1116454154253006, "global_step": 187471, "epoch": 2106, "lr": 2.1887920055191046e-05}
{"train_loss": 0.09110146015882492, "global_step": 187472, "epoch": 2106, "lr": 2.1887440630548012e-05}
{"train_loss": 0.06975313276052475, "global_step": 187473, "epoch": 2106, "lr": 2.1886961209684347e-05}
{"train_loss": 0.05776664987206459, "global_step": 187474, "epoch": 2106, "lr": 2.188648179260016e-05}
{"train_loss": 0.04956405609846115, "global_step": 187475, "epoch": 2106, "lr": 2.1886002379295474e-05}
{"train_loss": 0.10846668481826782, "global_step": 187476, "epoch": 2106, "lr": 2.1885522969770382e-05}
{"train_loss": 0.09680372476577759, "global_step": 187477, "epoch": 2106, "lr": 2.188504356402493e-05}
{"train_loss": 0.07595310360193253, "global_step": 187478, "epoch": 2106, "lr": 2.1884564162059207e-05}
{"train_loss": 0.10416015982627869, "global_step": 187479, "epoch": 2106, "lr": 2.1884084763873247e-05}
{"train_loss": 0.12300591170787811, "global_step": 187480, "epoch": 2106, "lr": 2.188360536946714e-05}
{"train_loss": 0.05681616812944412, "global_step": 187481, "epoch": 2106, "lr": 2.1883125978840922e-05}
{"train_loss": 0.0541531965136528, "global_step": 187482, "epoch": 2106, "lr": 2.18826465919947e-05}
{"train_loss": 0.05056415870785713, "global_step": 187483, "epoch": 2106, "lr": 2.1882167208928488e-05}
{"train_loss": 0.11715923994779587, "global_step": 187484, "epoch": 2106, "lr": 2.1881687829642384e-05}
{"train_loss": 0.07615957409143448, "global_step": 187485, "epoch": 2106, "lr": 2.188120845413646e-05}
{"train_loss": 0.04911017417907715, "global_step": 187486, "epoch": 2106, "lr": 2.188072908241074e-05}
{"train_loss": 0.09403188526630402, "global_step": 187487, "epoch": 2106, "lr": 2.188024971446534e-05}
{"train_loss": 0.09174016118049622, "global_step": 187488, "epoch": 2106, "lr": 2.1879770350300272e-05}
{"train_loss": 0.12538635730743408, "global_step": 187489, "epoch": 2106, "lr": 2.1879290989915647e-05}
{"train_loss": 0.16002970933914185, "global_step": 187490, "epoch": 2106, "lr": 2.1878811633311493e-05}
{"train_loss": 0.0666905865073204, "global_step": 187491, "epoch": 2106, "lr": 2.1878332280487908e-05}
{"train_loss": 0.07009009271860123, "global_step": 187492, "epoch": 2106, "lr": 2.1877852931444915e-05}
{"train_loss": 0.06156441941857338, "global_step": 187493, "epoch": 2106, "lr": 2.1877373586182627e-05}
{"train_loss": 0.09202886372804642, "global_step": 187494, "epoch": 2106, "lr": 2.1876894244701063e-05}
{"train_loss": 0.031032178550958633, "global_step": 187495, "epoch": 2106, "lr": 2.187641490700032e-05}
{"train_loss": 0.10039642453193665, "global_step": 187496, "epoch": 2106, "lr": 2.1875935573080443e-05}
{"train_loss": 0.10735023021697998, "global_step": 187497, "epoch": 2106, "lr": 2.187545624294151e-05}
{"train_loss": 0.07533709704875946, "global_step": 187498, "epoch": 2106, "lr": 2.187497691658359e-05}
{"train_loss": 0.06308478862047195, "global_step": 187499, "epoch": 2106, "lr": 2.187449759400671e-05}
{"train_loss": 0.10124478489160538, "global_step": 187500, "epoch": 2106, "lr": 2.1874018275210983e-05}
{"train_loss": 0.08226703852415085, "global_step": 187501, "epoch": 2106, "lr": 2.187353896019643e-05}
{"train_loss": 0.05612111836671829, "global_step": 187502, "epoch": 2106, "lr": 2.1873059648963163e-05}
{"train_loss": 0.10558716952800751, "global_step": 187503, "epoch": 2106, "lr": 2.1872580341511196e-05}
{"train_loss": 0.10083988308906555, "global_step": 187504, "epoch": 2106, "lr": 2.1872101037840637e-05}
{"train_loss": 0.056258346885442734, "global_step": 187505, "epoch": 2106, "lr": 2.1871621737951516e-05}
{"train_loss": 0.12031012773513794, "global_step": 187506, "epoch": 2106, "lr": 2.187114244184392e-05}
{"train_loss": 0.05073659121990204, "global_step": 187507, "epoch": 2106, "lr": 2.18706631495179e-05}
{"train_loss": 0.1087557002902031, "global_step": 187508, "epoch": 2106, "lr": 2.187018386097354e-05}
{"train_loss": 0.06583799421787262, "global_step": 187509, "epoch": 2106, "lr": 2.186970457621087e-05}
{"train_loss": 0.08573942631483078, "global_step": 187510, "epoch": 2106, "lr": 2.186922529523e-05}
{"train_loss": 0.09019006788730621, "global_step": 187511, "epoch": 2106, "lr": 2.1868746018030945e-05}
{"train_loss": 0.03830403462052345, "global_step": 187512, "epoch": 2106, "lr": 2.186826674461382e-05}
{"train_loss": 0.09641379863023758, "global_step": 187513, "epoch": 2106, "lr": 2.186778747497864e-05}
{"train_loss": 0.05904725566506386, "global_step": 187514, "epoch": 2106, "lr": 2.1867308209125508e-05}
{"train_loss": 0.08148650079965591, "global_step": 187515, "epoch": 2106, "lr": 2.1866828947054462e-05}
{"train_loss": 0.13940474390983582, "global_step": 187516, "epoch": 2106, "lr": 2.18663496887656e-05}
{"train_loss": 0.08336075395345688, "global_step": 187517, "epoch": 2106, "lr": 2.1865870434258933e-05}
{"train_loss": 0.08719035983085632, "global_step": 187518, "epoch": 2106, "lr": 2.1865391183534585e-05}
{"train_loss": 0.04923484846949577, "global_step": 187519, "epoch": 2106, "lr": 2.1864911936592568e-05}
{"train_loss": 0.07418898493051529, "global_step": 187520, "epoch": 2106, "lr": 2.1864432693432973e-05}
{"train_loss": 0.07533491402864456, "global_step": 187521, "epoch": 2106, "lr": 2.1863953454055885e-05}
{"train_loss": 0.08113967669144105, "global_step": 187522, "epoch": 2106, "lr": 2.186347421846132e-05, "val_loss": 7.8099846839904785}
{"train_loss": 0.04395056515932083, "global_step": 187523, "epoch": 2107, "lr": 2.1862994986649392e-05}
{"train_loss": 0.08331815898418427, "global_step": 187524, "epoch": 2107, "lr": 2.1862515758620116e-05}
{"train_loss": 0.025338122621178627, "global_step": 187525, "epoch": 2107, "lr": 2.1862036534373608e-05}
{"train_loss": 0.09760453552007675, "global_step": 187526, "epoch": 2107, "lr": 2.1861557313909885e-05}
{"train_loss": 0.02799489162862301, "global_step": 187527, "epoch": 2107, "lr": 2.1861078097229048e-05}
{"train_loss": 0.06120593845844269, "global_step": 187528, "epoch": 2107, "lr": 2.1860598884331135e-05}
{"train_loss": 0.05485348403453827, "global_step": 187529, "epoch": 2107, "lr": 2.1860119675216234e-05}
{"train_loss": 0.08875381201505661, "global_step": 187530, "epoch": 2107, "lr": 2.1859640469884378e-05}
{"train_loss": 0.07710179686546326, "global_step": 187531, "epoch": 2107, "lr": 2.185916126833567e-05}
{"train_loss": 0.07949750870466232, "global_step": 187532, "epoch": 2107, "lr": 2.1858682070570135e-05}
{"train_loss": 0.07179275155067444, "global_step": 187533, "epoch": 2107, "lr": 2.185820287658788e-05}
{"train_loss": 0.0789107233285904, "global_step": 187534, "epoch": 2107, "lr": 2.1857723686388926e-05}
{"train_loss": 0.04504557326436043, "global_step": 187535, "epoch": 2107, "lr": 2.185724449997338e-05}
{"train_loss": 0.025477983057498932, "global_step": 187536, "epoch": 2107, "lr": 2.1856765317341278e-05}
{"train_loss": 0.11778520047664642, "global_step": 187537, "epoch": 2107, "lr": 2.1856286138492676e-05}
{"train_loss": 0.05931733548641205, "global_step": 187538, "epoch": 2107, "lr": 2.1855806963427673e-05}
{"train_loss": 0.07086946070194244, "global_step": 187539, "epoch": 2107, "lr": 2.185532779214629e-05}
{"train_loss": 0.05091945827007294, "global_step": 187540, "epoch": 2107, "lr": 2.185484862464864e-05}
{"train_loss": 0.027880392968654633, "global_step": 187541, "epoch": 2107, "lr": 2.1854369460934737e-05}
{"train_loss": 0.06502395868301392, "global_step": 187542, "epoch": 2107, "lr": 2.185389030100469e-05}
{"train_loss": 0.06495829671621323, "global_step": 187543, "epoch": 2107, "lr": 2.185341114485853e-05}
{"train_loss": 0.04455660283565521, "global_step": 187544, "epoch": 2107, "lr": 2.1852931992496355e-05}
{"train_loss": 0.0709514319896698, "global_step": 187545, "epoch": 2107, "lr": 2.1852452843918194e-05}
{"train_loss": 0.06171899661421776, "global_step": 187546, "epoch": 2107, "lr": 2.185197369912414e-05}
{"train_loss": 0.04703100025653839, "global_step": 187547, "epoch": 2107, "lr": 2.1851494558114226e-05}
{"train_loss": 0.07244156301021576, "global_step": 187548, "epoch": 2107, "lr": 2.185101542088856e-05}
{"train_loss": 0.0758095309138298, "global_step": 187549, "epoch": 2107, "lr": 2.1850536287447158e-05}
{"train_loss": 0.09828343242406845, "global_step": 187550, "epoch": 2107, "lr": 2.1850057157790134e-05}
{"train_loss": 0.09487081319093704, "global_step": 187551, "epoch": 2107, "lr": 2.18495780319175e-05}
{"train_loss": 0.056640803813934326, "global_step": 187552, "epoch": 2107, "lr": 2.184909890982937e-05}
{"train_loss": 0.08420401066541672, "global_step": 187553, "epoch": 2107, "lr": 2.1848619791525765e-05}
{"train_loss": 0.06576073169708252, "global_step": 187554, "epoch": 2107, "lr": 2.1848140677006788e-05}
{"train_loss": 0.044958554208278656, "global_step": 187555, "epoch": 2107, "lr": 2.184766156627247e-05}
{"train_loss": 0.1257750540971756, "global_step": 187556, "epoch": 2107, "lr": 2.184718245932289e-05}
{"train_loss": 0.04205161705613136, "global_step": 187557, "epoch": 2107, "lr": 2.1846703356158134e-05}
{"train_loss": 0.05662579834461212, "global_step": 187558, "epoch": 2107, "lr": 2.184622425677822e-05}
{"train_loss": 0.016698220744729042, "global_step": 187559, "epoch": 2107, "lr": 2.184574516118326e-05}
{"train_loss": 0.13289226591587067, "global_step": 187560, "epoch": 2107, "lr": 2.184526606937328e-05}
{"train_loss": 0.0736512839794159, "global_step": 187561, "epoch": 2107, "lr": 2.1844786981348374e-05}
{"train_loss": 0.05465085059404373, "global_step": 187562, "epoch": 2107, "lr": 2.184430789710858e-05}
{"train_loss": 0.03727877140045166, "global_step": 187563, "epoch": 2107, "lr": 2.1843828816653994e-05}
{"train_loss": 0.042583782225847244, "global_step": 187564, "epoch": 2107, "lr": 2.184334973998464e-05}
{"train_loss": 0.06788834184408188, "global_step": 187565, "epoch": 2107, "lr": 2.1842870667100623e-05}
{"train_loss": 0.06934226304292679, "global_step": 187566, "epoch": 2107, "lr": 2.1842391598001976e-05}
{"train_loss": 0.12172786891460419, "global_step": 187567, "epoch": 2107, "lr": 2.184191253268879e-05}
{"train_loss": 0.058478303253650665, "global_step": 187568, "epoch": 2107, "lr": 2.18414334711611e-05}
{"train_loss": 0.08112290501594543, "global_step": 187569, "epoch": 2107, "lr": 2.1840954413419e-05}
{"train_loss": 0.04344012588262558, "global_step": 187570, "epoch": 2107, "lr": 2.1840475359462524e-05}
{"train_loss": 0.032513123005628586, "global_step": 187571, "epoch": 2107, "lr": 2.183999630929177e-05}
{"train_loss": 0.03366009145975113, "global_step": 187572, "epoch": 2107, "lr": 2.183951726290677e-05}
{"train_loss": 0.060623858124017715, "global_step": 187573, "epoch": 2107, "lr": 2.1839038220307624e-05}
{"train_loss": 0.09474105387926102, "global_step": 187574, "epoch": 2107, "lr": 2.1838559181494372e-05}
{"train_loss": 0.05704405531287193, "global_step": 187575, "epoch": 2107, "lr": 2.1838080146467066e-05}
{"train_loss": 0.03939918801188469, "global_step": 187576, "epoch": 2107, "lr": 2.1837601115225804e-05}
{"train_loss": 0.05018387734889984, "global_step": 187577, "epoch": 2107, "lr": 2.1837122087770616e-05}
{"train_loss": 0.15034320950508118, "global_step": 187578, "epoch": 2107, "lr": 2.18366430641016e-05}
{"train_loss": 0.05159891024231911, "global_step": 187579, "epoch": 2107, "lr": 2.1836164044218787e-05}
{"train_loss": 0.03834482282400131, "global_step": 187580, "epoch": 2107, "lr": 2.1835685028122277e-05}
{"train_loss": 0.06491734832525253, "global_step": 187581, "epoch": 2107, "lr": 2.18352060158121e-05}
{"train_loss": 0.06391709297895432, "global_step": 187582, "epoch": 2107, "lr": 2.1834727007288353e-05}
{"train_loss": 0.08607359975576401, "global_step": 187583, "epoch": 2107, "lr": 2.183424800255106e-05}
{"train_loss": 0.060670144855976105, "global_step": 187584, "epoch": 2107, "lr": 2.1833769001600336e-05}
{"train_loss": 0.045003972947597504, "global_step": 187585, "epoch": 2107, "lr": 2.1833290004436195e-05}
{"train_loss": 0.03770802170038223, "global_step": 187586, "epoch": 2107, "lr": 2.183281101105874e-05}
{"train_loss": 0.042467840015888214, "global_step": 187587, "epoch": 2107, "lr": 2.183233202146801e-05}
{"train_loss": 0.06821445375680923, "global_step": 187588, "epoch": 2107, "lr": 2.183185303566409e-05}
{"train_loss": 0.018787294626235962, "global_step": 187589, "epoch": 2107, "lr": 2.1831374053647026e-05}
{"train_loss": 0.10034660249948502, "global_step": 187590, "epoch": 2107, "lr": 2.18308950754169e-05}
{"train_loss": 0.07204676419496536, "global_step": 187591, "epoch": 2107, "lr": 2.1830416100973748e-05}
{"train_loss": 0.0784582570195198, "global_step": 187592, "epoch": 2107, "lr": 2.1829937130317656e-05}
{"train_loss": 0.06100873276591301, "global_step": 187593, "epoch": 2107, "lr": 2.182945816344871e-05}
{"train_loss": 0.022330671548843384, "global_step": 187594, "epoch": 2107, "lr": 2.182897920036692e-05}
{"train_loss": 0.08905043452978134, "global_step": 187595, "epoch": 2107, "lr": 2.1828500241072407e-05}
{"train_loss": 0.05400421842932701, "global_step": 187596, "epoch": 2107, "lr": 2.1828021285565187e-05}
{"train_loss": 0.055179379880428314, "global_step": 187597, "epoch": 2107, "lr": 2.182754233384537e-05}
{"train_loss": 0.04953795671463013, "global_step": 187598, "epoch": 2107, "lr": 2.1827063385912973e-05}
{"train_loss": 0.07325564324855804, "global_step": 187599, "epoch": 2107, "lr": 2.18265844417681e-05}
{"train_loss": 0.11044207215309143, "global_step": 187600, "epoch": 2107, "lr": 2.182610550141078e-05}
{"train_loss": 0.07138540595769882, "global_step": 187601, "epoch": 2107, "lr": 2.1825626564841123e-05}
{"train_loss": 0.06598939001560211, "global_step": 187602, "epoch": 2107, "lr": 2.1825147632059144e-05}
{"train_loss": 0.06564152985811234, "global_step": 187603, "epoch": 2107, "lr": 2.182466870306495e-05}
{"train_loss": 0.05470917746424675, "global_step": 187604, "epoch": 2107, "lr": 2.182418977785857e-05}
{"train_loss": 0.04394214227795601, "global_step": 187605, "epoch": 2107, "lr": 2.18237108564401e-05}
{"train_loss": 0.07021057605743408, "global_step": 187606, "epoch": 2107, "lr": 2.182323193880957e-05}
{"train_loss": 0.13081137835979462, "global_step": 187607, "epoch": 2107, "lr": 2.182275302496708e-05}
{"train_loss": 0.04788415879011154, "global_step": 187608, "epoch": 2107, "lr": 2.1822274114912662e-05}
{"train_loss": 0.07639109343290329, "global_step": 187609, "epoch": 2107, "lr": 2.1821795208646412e-05}
{"train_loss": 0.08832480758428574, "global_step": 187610, "epoch": 2107, "lr": 2.182131630616836e-05}
{"train_loss": 0.06495563432657986, "global_step": 187611, "epoch": 2107, "lr": 2.1820837407478607e-05, "val_loss": 8.023870468139648}
{"train_loss": 0.11279454827308655, "global_step": 187612, "epoch": 2108, "lr": 2.1820358512577184e-05}
{"train_loss": 0.08217665553092957, "global_step": 187613, "epoch": 2108, "lr": 2.1819879621464184e-05}
{"train_loss": 0.061049867421388626, "global_step": 187614, "epoch": 2108, "lr": 2.1819400734139657e-05}
{"train_loss": 0.0996968224644661, "global_step": 187615, "epoch": 2108, "lr": 2.181892185060365e-05}
{"train_loss": 0.06761495023965836, "global_step": 187616, "epoch": 2108, "lr": 2.1818442970856268e-05}
{"train_loss": 0.06530120968818665, "global_step": 187617, "epoch": 2108, "lr": 2.181796409489753e-05}
{"train_loss": 0.062214720994234085, "global_step": 187618, "epoch": 2108, "lr": 2.1817485222727545e-05}
{"train_loss": 0.12545713782310486, "global_step": 187619, "epoch": 2108, "lr": 2.1817006354346338e-05}
{"train_loss": 0.0942453145980835, "global_step": 187620, "epoch": 2108, "lr": 2.1816527489754007e-05}
{"train_loss": 0.04503495991230011, "global_step": 187621, "epoch": 2108, "lr": 2.1816048628950587e-05}
{"train_loss": 0.03989306837320328, "global_step": 187622, "epoch": 2108, "lr": 2.181556977193617e-05}
{"train_loss": 0.0784635916352272, "global_step": 187623, "epoch": 2108, "lr": 2.1815090918710786e-05}
{"train_loss": 0.037233248353004456, "global_step": 187624, "epoch": 2108, "lr": 2.1814612069274538e-05}
{"train_loss": 0.09163349121809006, "global_step": 187625, "epoch": 2108, "lr": 2.1814133223627454e-05}
{"train_loss": 0.12161882221698761, "global_step": 187626, "epoch": 2108, "lr": 2.1813654381769637e-05}
{"train_loss": 0.06287403404712677, "global_step": 187627, "epoch": 2108, "lr": 2.181317554370111e-05}
{"train_loss": 0.03194905072450638, "global_step": 187628, "epoch": 2108, "lr": 2.181269670942196e-05}
{"train_loss": 0.0626584067940712, "global_step": 187629, "epoch": 2108, "lr": 2.1812217878932266e-05}
{"train_loss": 0.05216637998819351, "global_step": 187630, "epoch": 2108, "lr": 2.181173905223206e-05}
{"train_loss": 0.05271315574645996, "global_step": 187631, "epoch": 2108, "lr": 2.181126022932144e-05}
{"train_loss": 0.14268183708190918, "global_step": 187632, "epoch": 2108, "lr": 2.181078141020043e-05}
{"train_loss": 0.054581526666879654, "global_step": 187633, "epoch": 2108, "lr": 2.1810302594869137e-05}
{"train_loss": 0.07959788292646408, "global_step": 187634, "epoch": 2108, "lr": 2.1809823783327588e-05}
{"train_loss": 0.11100781708955765, "global_step": 187635, "epoch": 2108, "lr": 2.1809344975575884e-05}
{"train_loss": 0.08138260990381241, "global_step": 187636, "epoch": 2108, "lr": 2.180886617161405e-05}
{"train_loss": 0.050704970955848694, "global_step": 187637, "epoch": 2108, "lr": 2.180838737144219e-05}
{"train_loss": 0.09424825012683868, "global_step": 187638, "epoch": 2108, "lr": 2.1807908575060328e-05}
{"train_loss": 0.09021555632352829, "global_step": 187639, "epoch": 2108, "lr": 2.180742978246857e-05}
{"train_loss": 0.05812734365463257, "global_step": 187640, "epoch": 2108, "lr": 2.1806950993666937e-05}
{"train_loss": 0.03329256549477577, "global_step": 187641, "epoch": 2108, "lr": 2.1806472208655537e-05}
{"train_loss": 0.05141959711909294, "global_step": 187642, "epoch": 2108, "lr": 2.1805993427434396e-05}
{"train_loss": 0.048509255051612854, "global_step": 187643, "epoch": 2108, "lr": 2.1805514650003612e-05}
{"train_loss": 0.05795251578092575, "global_step": 187644, "epoch": 2108, "lr": 2.1805035876363216e-05}
{"train_loss": 0.06646659970283508, "global_step": 187645, "epoch": 2108, "lr": 2.1804557106513306e-05}
{"train_loss": 0.041114162653684616, "global_step": 187646, "epoch": 2108, "lr": 2.1804078340453916e-05}
{"train_loss": 0.041056230664253235, "global_step": 187647, "epoch": 2108, "lr": 2.1803599578185137e-05}
{"train_loss": 0.05129100754857063, "global_step": 187648, "epoch": 2108, "lr": 2.1803120819707003e-05}
{"train_loss": 0.07060860842466354, "global_step": 187649, "epoch": 2108, "lr": 2.1802642065019613e-05}
{"train_loss": 0.0554361529648304, "global_step": 187650, "epoch": 2108, "lr": 2.1802163314122997e-05}
{"train_loss": 0.05961424484848976, "global_step": 187651, "epoch": 2108, "lr": 2.1801684567017256e-05}
{"train_loss": 0.05595598742365837, "global_step": 187652, "epoch": 2108, "lr": 2.1801205823702435e-05}
{"train_loss": 0.1147083044052124, "global_step": 187653, "epoch": 2108, "lr": 2.1800727084178578e-05}
{"train_loss": 0.06063993647694588, "global_step": 187654, "epoch": 2108, "lr": 2.180024834844579e-05}
{"train_loss": 0.05917445570230484, "global_step": 187655, "epoch": 2108, "lr": 2.1799769616504094e-05}
{"train_loss": 0.04511597752571106, "global_step": 187656, "epoch": 2108, "lr": 2.1799290888353595e-05}
{"train_loss": 0.09554988890886307, "global_step": 187657, "epoch": 2108, "lr": 2.179881216399432e-05}
{"train_loss": 0.07047931849956512, "global_step": 187658, "epoch": 2108, "lr": 2.1798333443426367e-05}
{"train_loss": 0.044965166598558426, "global_step": 187659, "epoch": 2108, "lr": 2.179785472664977e-05}
{"train_loss": 0.10667581111192703, "global_step": 187660, "epoch": 2108, "lr": 2.1797376013664624e-05}
{"train_loss": 0.03721765801310539, "global_step": 187661, "epoch": 2108, "lr": 2.179689730447096e-05}
{"train_loss": 0.08347461372613907, "global_step": 187662, "epoch": 2108, "lr": 2.1796418599068873e-05}
{"train_loss": 0.13718870282173157, "global_step": 187663, "epoch": 2108, "lr": 2.17959398974584e-05}
{"train_loss": 0.04983554780483246, "global_step": 187664, "epoch": 2108, "lr": 2.1795461199639616e-05}
{"train_loss": 0.03742339089512825, "global_step": 187665, "epoch": 2108, "lr": 2.179498250561261e-05}
{"train_loss": 0.12166371941566467, "global_step": 187666, "epoch": 2108, "lr": 2.1794503815377403e-05}
{"train_loss": 0.09537284821271896, "global_step": 187667, "epoch": 2108, "lr": 2.1794025128934105e-05}
{"train_loss": 0.0893935114145279, "global_step": 187668, "epoch": 2108, "lr": 2.179354644628273e-05}
{"train_loss": 0.09691251069307327, "global_step": 187669, "epoch": 2108, "lr": 2.1793067767423393e-05}
{"train_loss": 0.09156955778598785, "global_step": 187670, "epoch": 2108, "lr": 2.1792589092356114e-05}
{"train_loss": 0.08218923956155777, "global_step": 187671, "epoch": 2108, "lr": 2.1792110421080998e-05}
{"train_loss": 0.03483162447810173, "global_step": 187672, "epoch": 2108, "lr": 2.179163175359807e-05}
{"train_loss": 0.12843641638755798, "global_step": 187673, "epoch": 2108, "lr": 2.179115308990743e-05}
{"train_loss": 0.0414779856801033, "global_step": 187674, "epoch": 2108, "lr": 2.179067443000911e-05}
{"train_loss": 0.025809599086642265, "global_step": 187675, "epoch": 2108, "lr": 2.1790195773903206e-05}
{"train_loss": 0.07795647531747818, "global_step": 187676, "epoch": 2108, "lr": 2.178971712158975e-05}
{"train_loss": 0.09143225848674774, "global_step": 187677, "epoch": 2108, "lr": 2.178923847306884e-05}
{"train_loss": 0.05537600442767143, "global_step": 187678, "epoch": 2108, "lr": 2.1788759828340504e-05}
{"train_loss": 0.07868137210607529, "global_step": 187679, "epoch": 2108, "lr": 2.178828118740484e-05}
{"train_loss": 0.09484149515628815, "global_step": 187680, "epoch": 2108, "lr": 2.1787802550261887e-05}
{"train_loss": 0.02731993794441223, "global_step": 187681, "epoch": 2108, "lr": 2.1787323916911732e-05}
{"train_loss": 0.05397995933890343, "global_step": 187682, "epoch": 2108, "lr": 2.1786845287354412e-05}
{"train_loss": 0.0940825417637825, "global_step": 187683, "epoch": 2108, "lr": 2.1786366661590007e-05}
{"train_loss": 0.08007378876209259, "global_step": 187684, "epoch": 2108, "lr": 2.1785888039618602e-05}
{"train_loss": 0.12527699768543243, "global_step": 187685, "epoch": 2108, "lr": 2.1785409421440238e-05}
{"train_loss": 0.05019296705722809, "global_step": 187686, "epoch": 2108, "lr": 2.1784930807054958e-05}
{"train_loss": 0.06545278429985046, "global_step": 187687, "epoch": 2108, "lr": 2.1784452196462878e-05}
{"train_loss": 0.06679104268550873, "global_step": 187688, "epoch": 2108, "lr": 2.1783973589664015e-05}
{"train_loss": 0.07726054638624191, "global_step": 187689, "epoch": 2108, "lr": 2.1783494986658464e-05}
{"train_loss": 0.07815659791231155, "global_step": 187690, "epoch": 2108, "lr": 2.1783016387446282e-05}
{"train_loss": 0.11536354571580887, "global_step": 187691, "epoch": 2108, "lr": 2.1782537792027513e-05}
{"train_loss": 0.031124848872423172, "global_step": 187692, "epoch": 2108, "lr": 2.1782059200402256e-05}
{"train_loss": 0.09099648147821426, "global_step": 187693, "epoch": 2108, "lr": 2.1781580612570535e-05}
{"train_loss": 0.0598330944776535, "global_step": 187694, "epoch": 2108, "lr": 2.1781102028532457e-05}
{"train_loss": 0.13404406607151031, "global_step": 187695, "epoch": 2108, "lr": 2.178062344828805e-05}
{"train_loss": 0.06875728815793991, "global_step": 187696, "epoch": 2108, "lr": 2.178014487183741e-05}
{"train_loss": 0.03069336712360382, "global_step": 187697, "epoch": 2108, "lr": 2.1779666299180566e-05}
{"train_loss": 0.20433548092842102, "global_step": 187698, "epoch": 2108, "lr": 2.1779187730317625e-05}
{"train_loss": 0.05870504304766655, "global_step": 187699, "epoch": 2108, "lr": 2.1778709165248606e-05}
{"train_loss": 0.07421790341731538, "global_step": 187700, "epoch": 2108, "lr": 2.1778230603973614e-05, "val_loss": 7.8077592849731445}
{"train_loss": 0.055980224162340164, "global_step": 187701, "epoch": 2109, "lr": 2.177775204649268e-05}
{"train_loss": 0.06307832151651382, "global_step": 187702, "epoch": 2109, "lr": 2.1777273492805878e-05}
{"train_loss": 0.07843908667564392, "global_step": 187703, "epoch": 2109, "lr": 2.1776794942913297e-05}
{"train_loss": 0.05819127336144447, "global_step": 187704, "epoch": 2109, "lr": 2.1776316396814967e-05}
{"train_loss": 0.06896838545799255, "global_step": 187705, "epoch": 2109, "lr": 2.1775837854510985e-05}
{"train_loss": 0.046006105840206146, "global_step": 187706, "epoch": 2109, "lr": 2.1775359316001375e-05}
{"train_loss": 0.044070057570934296, "global_step": 187707, "epoch": 2109, "lr": 2.177488078128625e-05}
{"train_loss": 0.07021089643239975, "global_step": 187708, "epoch": 2109, "lr": 2.177440225036562e-05}
{"train_loss": 0.07699299603700638, "global_step": 187709, "epoch": 2109, "lr": 2.17739237232396e-05}
{"train_loss": 0.06350182741880417, "global_step": 187710, "epoch": 2109, "lr": 2.1773445199908223e-05}
{"train_loss": 0.07965736091136932, "global_step": 187711, "epoch": 2109, "lr": 2.177296668037157e-05}
{"train_loss": 0.06568644940853119, "global_step": 187712, "epoch": 2109, "lr": 2.177248816462968e-05}
{"train_loss": 0.048914216458797455, "global_step": 187713, "epoch": 2109, "lr": 2.1772009652682656e-05}
{"train_loss": 0.06927620619535446, "global_step": 187714, "epoch": 2109, "lr": 2.177153114453052e-05}
{"train_loss": 0.08734125643968582, "global_step": 187715, "epoch": 2109, "lr": 2.1771052640173377e-05}
{"train_loss": 0.06797235459089279, "global_step": 187716, "epoch": 2109, "lr": 2.1770574139611254e-05}
{"train_loss": 0.12003470957279205, "global_step": 187717, "epoch": 2109, "lr": 2.177009564284425e-05}
{"train_loss": 0.0722866803407669, "global_step": 187718, "epoch": 2109, "lr": 2.1769617149872403e-05}
{"train_loss": 0.0240744911134243, "global_step": 187719, "epoch": 2109, "lr": 2.1769138660695777e-05}
{"train_loss": 0.045020535588264465, "global_step": 187720, "epoch": 2109, "lr": 2.1768660175314466e-05}
{"train_loss": 0.06380896270275116, "global_step": 187721, "epoch": 2109, "lr": 2.1768181693728502e-05}
{"train_loss": 0.06717903912067413, "global_step": 187722, "epoch": 2109, "lr": 2.1767703215937975e-05}
{"train_loss": 0.02963133715093136, "global_step": 187723, "epoch": 2109, "lr": 2.1767224741942914e-05}
{"train_loss": 0.05140913277864456, "global_step": 187724, "epoch": 2109, "lr": 2.176674627174343e-05}
{"train_loss": 0.04864608496427536, "global_step": 187725, "epoch": 2109, "lr": 2.176626780533956e-05}
{"train_loss": 0.10096687078475952, "global_step": 187726, "epoch": 2109, "lr": 2.1765789342731358e-05}
{"train_loss": 0.04634397476911545, "global_step": 187727, "epoch": 2109, "lr": 2.1765310883918915e-05}
{"train_loss": 0.09192679822444916, "global_step": 187728, "epoch": 2109, "lr": 2.1764832428902264e-05}
{"train_loss": 0.10438212752342224, "global_step": 187729, "epoch": 2109, "lr": 2.1764353977681507e-05}
{"train_loss": 0.05286180227994919, "global_step": 187730, "epoch": 2109, "lr": 2.1763875530256687e-05}
{"train_loss": 0.10211163014173508, "global_step": 187731, "epoch": 2109, "lr": 2.1763397086627853e-05}
{"train_loss": 0.08346903324127197, "global_step": 187732, "epoch": 2109, "lr": 2.1762918646795104e-05}
{"train_loss": 0.05499335750937462, "global_step": 187733, "epoch": 2109, "lr": 2.1762440210758467e-05}
{"train_loss": 0.13859593868255615, "global_step": 187734, "epoch": 2109, "lr": 2.1761961778518047e-05}
{"train_loss": 0.0711287409067154, "global_step": 187735, "epoch": 2109, "lr": 2.1761483350073864e-05}
{"train_loss": 0.08510106801986694, "global_step": 187736, "epoch": 2109, "lr": 2.1761004925426028e-05}
{"train_loss": 0.05435798689723015, "global_step": 187737, "epoch": 2109, "lr": 2.176052650457456e-05}
{"train_loss": 0.0617537721991539, "global_step": 187738, "epoch": 2109, "lr": 2.1760048087519547e-05}
{"train_loss": 0.054553769528865814, "global_step": 187739, "epoch": 2109, "lr": 2.1759569674261065e-05}
{"train_loss": 0.06909816712141037, "global_step": 187740, "epoch": 2109, "lr": 2.1759091264799154e-05}
{"train_loss": 0.0898548886179924, "global_step": 187741, "epoch": 2109, "lr": 2.17586128591339e-05}
{"train_loss": 0.15531328320503235, "global_step": 187742, "epoch": 2109, "lr": 2.175813445726534e-05}
{"train_loss": 0.053959738463163376, "global_step": 187743, "epoch": 2109, "lr": 2.1757656059193576e-05}
{"train_loss": 0.08341943472623825, "global_step": 187744, "epoch": 2109, "lr": 2.1757177664918627e-05}
{"train_loss": 0.0930628851056099, "global_step": 187745, "epoch": 2109, "lr": 2.1756699274440605e-05}
{"train_loss": 0.05322539433836937, "global_step": 187746, "epoch": 2109, "lr": 2.1756220887759526e-05}
{"train_loss": 0.0785786360502243, "global_step": 187747, "epoch": 2109, "lr": 2.1755742504875503e-05}
{"train_loss": 0.04624248296022415, "global_step": 187748, "epoch": 2109, "lr": 2.1755264125788556e-05}
{"train_loss": 0.06143774092197418, "global_step": 187749, "epoch": 2109, "lr": 2.175478575049879e-05}
{"train_loss": 0.029446028172969818, "global_step": 187750, "epoch": 2109, "lr": 2.175430737900623e-05}
{"train_loss": 0.11830618977546692, "global_step": 187751, "epoch": 2109, "lr": 2.175382901131097e-05}
{"train_loss": 0.17392250895500183, "global_step": 187752, "epoch": 2109, "lr": 2.1753350647413056e-05}
{"train_loss": 0.06518996506929398, "global_step": 187753, "epoch": 2109, "lr": 2.175287228731257e-05}
{"train_loss": 0.07818704843521118, "global_step": 187754, "epoch": 2109, "lr": 2.175239393100955e-05}
{"train_loss": 0.07700123637914658, "global_step": 187755, "epoch": 2109, "lr": 2.1751915578504077e-05}
{"train_loss": 0.0841788500547409, "global_step": 187756, "epoch": 2109, "lr": 2.1751437229796235e-05}
{"train_loss": 0.11473991721868515, "global_step": 187757, "epoch": 2109, "lr": 2.1750958884886047e-05}
{"train_loss": 0.029106590896844864, "global_step": 187758, "epoch": 2109, "lr": 2.1750480543773617e-05}
{"train_loss": 0.0860922709107399, "global_step": 187759, "epoch": 2109, "lr": 2.1750002206458974e-05}
{"train_loss": 0.08878976106643677, "global_step": 187760, "epoch": 2109, "lr": 2.1749523872942217e-05}
{"train_loss": 0.03313260152935982, "global_step": 187761, "epoch": 2109, "lr": 2.1749045543223372e-05}
{"train_loss": 0.024932414293289185, "global_step": 187762, "epoch": 2109, "lr": 2.174856721730254e-05}
{"train_loss": 0.07643263041973114, "global_step": 187763, "epoch": 2109, "lr": 2.1748088895179774e-05}
{"train_loss": 0.12865495681762695, "global_step": 187764, "epoch": 2109, "lr": 2.174761057685511e-05}
{"train_loss": 0.06936225295066833, "global_step": 187765, "epoch": 2109, "lr": 2.1747132262328656e-05}
{"train_loss": 0.07659627497196198, "global_step": 187766, "epoch": 2109, "lr": 2.1746653951600436e-05}
{"train_loss": 0.03787820786237717, "global_step": 187767, "epoch": 2109, "lr": 2.1746175644670552e-05}
{"train_loss": 0.12250903993844986, "global_step": 187768, "epoch": 2109, "lr": 2.1745697341539056e-05}
{"train_loss": 0.13602328300476074, "global_step": 187769, "epoch": 2109, "lr": 2.1745219042205983e-05}
{"train_loss": 0.08780420571565628, "global_step": 187770, "epoch": 2109, "lr": 2.1744740746671437e-05}
{"train_loss": 0.05009174719452858, "global_step": 187771, "epoch": 2109, "lr": 2.1744262454935454e-05}
{"train_loss": 0.06284140795469284, "global_step": 187772, "epoch": 2109, "lr": 2.1743784166998125e-05}
{"train_loss": 0.02010367624461651, "global_step": 187773, "epoch": 2109, "lr": 2.1743305882859482e-05}
{"train_loss": 0.07283420115709305, "global_step": 187774, "epoch": 2109, "lr": 2.174282760251961e-05}
{"train_loss": 0.10266965627670288, "global_step": 187775, "epoch": 2109, "lr": 2.1742349325978584e-05}
{"train_loss": 0.059712208807468414, "global_step": 187776, "epoch": 2109, "lr": 2.174187105323644e-05}
{"train_loss": 0.0855623334646225, "global_step": 187777, "epoch": 2109, "lr": 2.1741392784293274e-05}
{"train_loss": 0.08760663866996765, "global_step": 187778, "epoch": 2109, "lr": 2.1740914519149115e-05}
{"train_loss": 0.05934618040919304, "global_step": 187779, "epoch": 2109, "lr": 2.1740436257804064e-05}
{"train_loss": 0.09846006333827972, "global_step": 187780, "epoch": 2109, "lr": 2.173995800025815e-05}
{"train_loss": 0.05216514691710472, "global_step": 187781, "epoch": 2109, "lr": 2.173947974651147e-05}
{"train_loss": 0.05700451135635376, "global_step": 187782, "epoch": 2109, "lr": 2.1739001496564055e-05}
{"train_loss": 0.04781613498926163, "global_step": 187783, "epoch": 2109, "lr": 2.1738523250416006e-05}
{"train_loss": 0.06577855348587036, "global_step": 187784, "epoch": 2109, "lr": 2.173804500806735e-05}
{"train_loss": 0.09245980530977249, "global_step": 187785, "epoch": 2109, "lr": 2.1737566769518188e-05}
{"train_loss": 0.0640304759144783, "global_step": 187786, "epoch": 2109, "lr": 2.1737088534768545e-05}
{"train_loss": 0.0791754499077797, "global_step": 187787, "epoch": 2109, "lr": 2.1736610303818526e-05}
{"train_loss": 0.0529290996491909, "global_step": 187788, "epoch": 2109, "lr": 2.1736132076668158e-05}
{"train_loss": 0.07286803533187074, "global_step": 187789, "epoch": 2109, "lr": 2.173565385331754e-05, "val_loss": 7.685083866119385}
{"train_loss": 0.11600758135318756, "global_step": 187790, "epoch": 2110, "lr": 2.1735175633766697e-05}
{"train_loss": 0.13035377860069275, "global_step": 187791, "epoch": 2110, "lr": 2.1734697418015722e-05}
{"train_loss": 0.06853196024894714, "global_step": 187792, "epoch": 2110, "lr": 2.1734219206064683e-05}
{"train_loss": 0.09633217006921768, "global_step": 187793, "epoch": 2110, "lr": 2.173374099791362e-05}
{"train_loss": 0.03688720613718033, "global_step": 187794, "epoch": 2110, "lr": 2.173326279356263e-05}
{"train_loss": 0.025435401126742363, "global_step": 187795, "epoch": 2110, "lr": 2.1732784593011733e-05}
{"train_loss": 0.09334130585193634, "global_step": 187796, "epoch": 2110, "lr": 2.173230639626105e-05}
{"train_loss": 0.11804956942796707, "global_step": 187797, "epoch": 2110, "lr": 2.1731828203310585e-05}
{"train_loss": 0.04703940823674202, "global_step": 187798, "epoch": 2110, "lr": 2.173135001416045e-05}
{"train_loss": 0.12737490236759186, "global_step": 187799, "epoch": 2110, "lr": 2.173087182881068e-05}
{"train_loss": 0.08727726340293884, "global_step": 187800, "epoch": 2110, "lr": 2.173039364726136e-05}
{"train_loss": 0.03421073779463768, "global_step": 187801, "epoch": 2110, "lr": 2.1729915469512546e-05}
{"train_loss": 0.062251877039670944, "global_step": 187802, "epoch": 2110, "lr": 2.172943729556428e-05}
{"train_loss": 0.05810892581939697, "global_step": 187803, "epoch": 2110, "lr": 2.172895912541667e-05}
{"train_loss": 0.085706926882267, "global_step": 187804, "epoch": 2110, "lr": 2.1728480959069737e-05}
{"train_loss": 0.08977297693490982, "global_step": 187805, "epoch": 2110, "lr": 2.1728002796523585e-05}
{"train_loss": 0.0353979729115963, "global_step": 187806, "epoch": 2110, "lr": 2.1727524637778252e-05}
{"train_loss": 0.11250104755163193, "global_step": 187807, "epoch": 2110, "lr": 2.1727046482833798e-05}
{"train_loss": 0.06423122435808182, "global_step": 187808, "epoch": 2110, "lr": 2.172656833169031e-05}
{"train_loss": 0.050684019923210144, "global_step": 187809, "epoch": 2110, "lr": 2.1726090184347825e-05}
{"train_loss": 0.05567796900868416, "global_step": 187810, "epoch": 2110, "lr": 2.172561204080642e-05}
{"train_loss": 0.05333642289042473, "global_step": 187811, "epoch": 2110, "lr": 2.1725133901066185e-05}
{"train_loss": 0.04106077179312706, "global_step": 187812, "epoch": 2110, "lr": 2.1724655765127137e-05}
{"train_loss": 0.07403451949357986, "global_step": 187813, "epoch": 2110, "lr": 2.172417763298939e-05}
{"train_loss": 0.13863113522529602, "global_step": 187814, "epoch": 2110, "lr": 2.172369950465296e-05}
{"train_loss": 0.097516268491745, "global_step": 187815, "epoch": 2110, "lr": 2.1723221380117954e-05}
{"train_loss": 0.03298748657107353, "global_step": 187816, "epoch": 2110, "lr": 2.172274325938439e-05}
{"train_loss": 0.0587424635887146, "global_step": 187817, "epoch": 2110, "lr": 2.1722265142452386e-05}
{"train_loss": 0.12053631991147995, "global_step": 187818, "epoch": 2110, "lr": 2.172178702932196e-05}
{"train_loss": 0.10013649612665176, "global_step": 187819, "epoch": 2110, "lr": 2.172130891999321e-05}
{"train_loss": 0.07829833030700684, "global_step": 187820, "epoch": 2110, "lr": 2.172083081446617e-05}
{"train_loss": 0.06647788733243942, "global_step": 187821, "epoch": 2110, "lr": 2.1720352712740934e-05}
{"train_loss": 0.15512971580028534, "global_step": 187822, "epoch": 2110, "lr": 2.1719874614817537e-05}
{"train_loss": 0.08814234286546707, "global_step": 187823, "epoch": 2110, "lr": 2.1719396520696072e-05}
{"train_loss": 0.12004861980676651, "global_step": 187824, "epoch": 2110, "lr": 2.171891843037658e-05}
{"train_loss": 0.05333058163523674, "global_step": 187825, "epoch": 2110, "lr": 2.1718440343859142e-05}
{"train_loss": 0.08167856931686401, "global_step": 187826, "epoch": 2110, "lr": 2.1717962261143805e-05}
{"train_loss": 0.053271934390068054, "global_step": 187827, "epoch": 2110, "lr": 2.1717484182230658e-05}
{"train_loss": 0.047229740768671036, "global_step": 187828, "epoch": 2110, "lr": 2.171700610711973e-05}
{"train_loss": 0.11480966210365295, "global_step": 187829, "epoch": 2110, "lr": 2.171652803581111e-05}
{"train_loss": 0.07730145752429962, "global_step": 187830, "epoch": 2110, "lr": 2.1716049968304874e-05}
{"train_loss": 0.044487759470939636, "global_step": 187831, "epoch": 2110, "lr": 2.1715571904601057e-05}
{"train_loss": 0.06050741299986839, "global_step": 187832, "epoch": 2110, "lr": 2.1715093844699753e-05}
{"train_loss": 0.08499468117952347, "global_step": 187833, "epoch": 2110, "lr": 2.1714615788600984e-05}
{"train_loss": 0.041968733072280884, "global_step": 187834, "epoch": 2110, "lr": 2.1714137736304867e-05}
{"train_loss": 0.0788361206650734, "global_step": 187835, "epoch": 2110, "lr": 2.1713659687811417e-05}
{"train_loss": 0.10335621237754822, "global_step": 187836, "epoch": 2110, "lr": 2.1713181643120738e-05}
{"train_loss": 0.09223766624927521, "global_step": 187837, "epoch": 2110, "lr": 2.1712703602232864e-05}
{"train_loss": 0.10791485756635666, "global_step": 187838, "epoch": 2110, "lr": 2.1712225565147887e-05}
{"train_loss": 0.08167441189289093, "global_step": 187839, "epoch": 2110, "lr": 2.1711747531865838e-05}
{"train_loss": 0.0743393674492836, "global_step": 187840, "epoch": 2110, "lr": 2.171126950238682e-05}
{"train_loss": 0.07233549654483795, "global_step": 187841, "epoch": 2110, "lr": 2.1710791476710872e-05}
{"train_loss": 0.061967410147190094, "global_step": 187842, "epoch": 2110, "lr": 2.1710313454838048e-05}
{"train_loss": 0.03040088154375553, "global_step": 187843, "epoch": 2110, "lr": 2.170983543676845e-05}
{"train_loss": 0.05002451688051224, "global_step": 187844, "epoch": 2110, "lr": 2.1709357422502096e-05}
{"train_loss": 0.07424542307853699, "global_step": 187845, "epoch": 2110, "lr": 2.1708879412039097e-05}
{"train_loss": 0.09019341319799423, "global_step": 187846, "epoch": 2110, "lr": 2.1708401405379475e-05}
{"train_loss": 0.04721682518720627, "global_step": 187847, "epoch": 2110, "lr": 2.1707923402523327e-05}
{"train_loss": 0.06826627254486084, "global_step": 187848, "epoch": 2110, "lr": 2.1707445403470688e-05}
{"train_loss": 0.08396834880113602, "global_step": 187849, "epoch": 2110, "lr": 2.170696740822166e-05}
{"train_loss": 0.1020422950387001, "global_step": 187850, "epoch": 2110, "lr": 2.1706489416776266e-05}
{"train_loss": 0.06980118155479431, "global_step": 187851, "epoch": 2110, "lr": 2.1706011429134605e-05}
{"train_loss": 0.045577552169561386, "global_step": 187852, "epoch": 2110, "lr": 2.170553344529671e-05}
{"train_loss": 0.08618541061878204, "global_step": 187853, "epoch": 2110, "lr": 2.1705055465262676e-05}
{"train_loss": 0.09231114387512207, "global_step": 187854, "epoch": 2110, "lr": 2.170457748903254e-05}
{"train_loss": 0.0497383177280426, "global_step": 187855, "epoch": 2110, "lr": 2.1704099516606392e-05}
{"train_loss": 0.0614396333694458, "global_step": 187856, "epoch": 2110, "lr": 2.1703621547984266e-05}
{"train_loss": 0.06506876647472382, "global_step": 187857, "epoch": 2110, "lr": 2.1703143583166263e-05}
{"train_loss": 0.02504478208720684, "global_step": 187858, "epoch": 2110, "lr": 2.1702665622152407e-05}
{"train_loss": 0.12524117529392242, "global_step": 187859, "epoch": 2110, "lr": 2.1702187664942803e-05}
{"train_loss": 0.057745933532714844, "global_step": 187860, "epoch": 2110, "lr": 2.170170971153747e-05}
{"train_loss": 0.07307574898004532, "global_step": 187861, "epoch": 2110, "lr": 2.1701231761936524e-05}
{"train_loss": 0.0635191947221756, "global_step": 187862, "epoch": 2110, "lr": 2.170075381613998e-05}
{"train_loss": 0.04970481991767883, "global_step": 187863, "epoch": 2110, "lr": 2.1700275874147947e-05}
{"train_loss": 0.08499209582805634, "global_step": 187864, "epoch": 2110, "lr": 2.1699797935960446e-05}
{"train_loss": 0.09477580338716507, "global_step": 187865, "epoch": 2110, "lr": 2.169932000157756e-05}
{"train_loss": 0.08154557645320892, "global_step": 187866, "epoch": 2110, "lr": 2.1698842070999377e-05}
{"train_loss": 0.1140621080994606, "global_step": 187867, "epoch": 2110, "lr": 2.169836414422592e-05}
{"train_loss": 0.039063841104507446, "global_step": 187868, "epoch": 2110, "lr": 2.169788622125729e-05}
{"train_loss": 0.10364332050085068, "global_step": 187869, "epoch": 2110, "lr": 2.169740830209352e-05}
{"train_loss": 0.0402955487370491, "global_step": 187870, "epoch": 2110, "lr": 2.1696930386734702e-05}
{"train_loss": 0.03961462527513504, "global_step": 187871, "epoch": 2110, "lr": 2.1696452475180872e-05}
{"train_loss": 0.09585640579462051, "global_step": 187872, "epoch": 2110, "lr": 2.1695974567432124e-05}
{"train_loss": 0.07114498317241669, "global_step": 187873, "epoch": 2110, "lr": 2.1695496663488495e-05}
{"train_loss": 0.09101060032844543, "global_step": 187874, "epoch": 2110, "lr": 2.1695018763350077e-05}
{"train_loss": 0.10058745741844177, "global_step": 187875, "epoch": 2110, "lr": 2.16945408670169e-05}
{"train_loss": 0.08399590849876404, "global_step": 187876, "epoch": 2110, "lr": 2.169406297448906e-05}
{"train_loss": 0.08676610141992569, "global_step": 187877, "epoch": 2110, "lr": 2.16935850857666e-05}
{"train_loss": 0.07628129889372359, "global_step": 187878, "epoch": 2110, "lr": 2.1693107200849604e-05, "val_loss": 7.624831199645996, "train_action_mse_error": 26.439523696899414}
{"train_loss": 0.04614034295082092, "global_step": 187879, "epoch": 2111, "lr": 2.169262931973812e-05}
{"train_loss": 0.10258034616708755, "global_step": 187880, "epoch": 2111, "lr": 2.1692151442432208e-05}
{"train_loss": 0.05948016047477722, "global_step": 187881, "epoch": 2111, "lr": 2.169167356893196e-05}
{"train_loss": 0.10952184349298477, "global_step": 187882, "epoch": 2111, "lr": 2.16911956992374e-05}
{"train_loss": 0.054113853722810745, "global_step": 187883, "epoch": 2111, "lr": 2.1690717833348627e-05}
{"train_loss": 0.07051641494035721, "global_step": 187884, "epoch": 2111, "lr": 2.1690239971265682e-05}
{"train_loss": 0.06541552394628525, "global_step": 187885, "epoch": 2111, "lr": 2.168976211298865e-05}
{"train_loss": 0.0625799298286438, "global_step": 187886, "epoch": 2111, "lr": 2.168928425851757e-05}
{"train_loss": 0.09065845608711243, "global_step": 187887, "epoch": 2111, "lr": 2.1688806407852542e-05}
{"train_loss": 0.06026472896337509, "global_step": 187888, "epoch": 2111, "lr": 2.1688328560993587e-05}
{"train_loss": 0.06574562191963196, "global_step": 187889, "epoch": 2111, "lr": 2.1687850717940806e-05}
{"train_loss": 0.052423473447561264, "global_step": 187890, "epoch": 2111, "lr": 2.1687372878694234e-05}
{"train_loss": 0.042625341564416885, "global_step": 187891, "epoch": 2111, "lr": 2.1686895043253968e-05}
{"train_loss": 0.0741419717669487, "global_step": 187892, "epoch": 2111, "lr": 2.1686417211620034e-05}
{"train_loss": 0.06866328418254852, "global_step": 187893, "epoch": 2111, "lr": 2.1685939383792538e-05}
{"train_loss": 0.047127384692430496, "global_step": 187894, "epoch": 2111, "lr": 2.1685461559771502e-05}
{"train_loss": 0.08839528262615204, "global_step": 187895, "epoch": 2111, "lr": 2.168498373955703e-05}
{"train_loss": 0.04110112413764, "global_step": 187896, "epoch": 2111, "lr": 2.1684505923149146e-05}
{"train_loss": 0.07337680459022522, "global_step": 187897, "epoch": 2111, "lr": 2.1684028110547955e-05}
{"train_loss": 0.05298561975359917, "global_step": 187898, "epoch": 2111, "lr": 2.1683550301753482e-05}
{"train_loss": 0.09427918493747711, "global_step": 187899, "epoch": 2111, "lr": 2.1683072496765826e-05}
{"train_loss": 0.08713003247976303, "global_step": 187900, "epoch": 2111, "lr": 2.1682594695585023e-05}
{"train_loss": 0.09470271319150925, "global_step": 187901, "epoch": 2111, "lr": 2.1682116898211146e-05}
{"train_loss": 0.055636338889598846, "global_step": 187902, "epoch": 2111, "lr": 2.1681639104644286e-05}
{"train_loss": 0.06423253566026688, "global_step": 187903, "epoch": 2111, "lr": 2.1681161314884463e-05}
{"train_loss": 0.11707524210214615, "global_step": 187904, "epoch": 2111, "lr": 2.1680683528931785e-05}
{"train_loss": 0.06395525485277176, "global_step": 187905, "epoch": 2111, "lr": 2.168020574678627e-05}
{"train_loss": 0.10117431730031967, "global_step": 187906, "epoch": 2111, "lr": 2.1679727968448028e-05}
{"train_loss": 0.05582974851131439, "global_step": 187907, "epoch": 2111, "lr": 2.1679250193917078e-05}
{"train_loss": 0.07156597822904587, "global_step": 187908, "epoch": 2111, "lr": 2.1678772423193534e-05}
{"train_loss": 0.06630124896764755, "global_step": 187909, "epoch": 2111, "lr": 2.167829465627741e-05}
{"train_loss": 0.10440094023942947, "global_step": 187910, "epoch": 2111, "lr": 2.167781689316881e-05}
{"train_loss": 0.07278523594141006, "global_step": 187911, "epoch": 2111, "lr": 2.167733913386777e-05}
{"train_loss": 0.07502420991659164, "global_step": 187912, "epoch": 2111, "lr": 2.167686137837438e-05}
{"train_loss": 0.06539050489664078, "global_step": 187913, "epoch": 2111, "lr": 2.1676383626688678e-05}
{"train_loss": 0.03861973062157631, "global_step": 187914, "epoch": 2111, "lr": 2.1675905878810754e-05}
{"train_loss": 0.06866364926099777, "global_step": 187915, "epoch": 2111, "lr": 2.1675428134740644e-05}
{"train_loss": 0.16731414198875427, "global_step": 187916, "epoch": 2111, "lr": 2.1674950394478444e-05}
{"train_loss": 0.07505370676517487, "global_step": 187917, "epoch": 2111, "lr": 2.1674472658024203e-05}
{"train_loss": 0.041653018444776535, "global_step": 187918, "epoch": 2111, "lr": 2.1673994925377965e-05}
{"train_loss": 0.09461365640163422, "global_step": 187919, "epoch": 2111, "lr": 2.1673517196539835e-05}
{"train_loss": 0.07663391530513763, "global_step": 187920, "epoch": 2111, "lr": 2.1673039471509832e-05}
{"train_loss": 0.05970952287316322, "global_step": 187921, "epoch": 2111, "lr": 2.1672561750288063e-05}
{"train_loss": 0.08171039819717407, "global_step": 187922, "epoch": 2111, "lr": 2.1672084032874557e-05}
{"train_loss": 0.07607880979776382, "global_step": 187923, "epoch": 2111, "lr": 2.1671606319269412e-05}
{"train_loss": 0.06641294062137604, "global_step": 187924, "epoch": 2111, "lr": 2.167112860947265e-05}
{"train_loss": 0.07945893704891205, "global_step": 187925, "epoch": 2111, "lr": 2.1670650903484385e-05}
{"train_loss": 0.0560324601829052, "global_step": 187926, "epoch": 2111, "lr": 2.1670173201304635e-05}
{"train_loss": 0.0989791601896286, "global_step": 187927, "epoch": 2111, "lr": 2.16696955029335e-05}
{"train_loss": 0.10583579540252686, "global_step": 187928, "epoch": 2111, "lr": 2.166921780837102e-05}
{"train_loss": 0.09023751318454742, "global_step": 187929, "epoch": 2111, "lr": 2.1668740117617275e-05}
{"train_loss": 0.05161839723587036, "global_step": 187930, "epoch": 2111, "lr": 2.166826243067231e-05}
{"train_loss": 0.042625170201063156, "global_step": 187931, "epoch": 2111, "lr": 2.1667784747536218e-05}
{"train_loss": 0.08210987597703934, "global_step": 187932, "epoch": 2111, "lr": 2.166730706820903e-05}
{"train_loss": 0.040849894285202026, "global_step": 187933, "epoch": 2111, "lr": 2.166682939269085e-05}
{"train_loss": 0.10209348797798157, "global_step": 187934, "epoch": 2111, "lr": 2.1666351720981698e-05}
{"train_loss": 0.057329874485731125, "global_step": 187935, "epoch": 2111, "lr": 2.1665874053081675e-05}
{"train_loss": 0.08464578539133072, "global_step": 187936, "epoch": 2111, "lr": 2.1665396388990818e-05}
{"train_loss": 0.09055766463279724, "global_step": 187937, "epoch": 2111, "lr": 2.1664918728709193e-05}
{"train_loss": 0.0573047436773777, "global_step": 187938, "epoch": 2111, "lr": 2.1664441072236902e-05}
{"train_loss": 0.04289208725094795, "global_step": 187939, "epoch": 2111, "lr": 2.166396341957396e-05}
{"train_loss": 0.050203800201416016, "global_step": 187940, "epoch": 2111, "lr": 2.1663485770720472e-05}
{"train_loss": 0.08093245327472687, "global_step": 187941, "epoch": 2111, "lr": 2.1663008125676458e-05}
{"train_loss": 0.02862239070236683, "global_step": 187942, "epoch": 2111, "lr": 2.1662530484442032e-05}
{"train_loss": 0.07603705674409866, "global_step": 187943, "epoch": 2111, "lr": 2.1662052847017215e-05}
{"train_loss": 0.05597921833395958, "global_step": 187944, "epoch": 2111, "lr": 2.166157521340211e-05}
{"train_loss": 0.04493514075875282, "global_step": 187945, "epoch": 2111, "lr": 2.1661097583596736e-05}
{"train_loss": 0.0631636381149292, "global_step": 187946, "epoch": 2111, "lr": 2.1660619957601202e-05}
{"train_loss": 0.051664188504219055, "global_step": 187947, "epoch": 2111, "lr": 2.166014233541554e-05}
{"train_loss": 0.034109652042388916, "global_step": 187948, "epoch": 2111, "lr": 2.165966471703984e-05}
{"train_loss": 0.07055222243070602, "global_step": 187949, "epoch": 2111, "lr": 2.1659187102474133e-05}
{"train_loss": 0.0869993269443512, "global_step": 187950, "epoch": 2111, "lr": 2.165870949171852e-05}
{"train_loss": 0.04541580751538277, "global_step": 187951, "epoch": 2111, "lr": 2.1658231884773032e-05}
{"train_loss": 0.07982499897480011, "global_step": 187952, "epoch": 2111, "lr": 2.1657754281637773e-05}
{"train_loss": 0.06023649871349335, "global_step": 187953, "epoch": 2111, "lr": 2.1657276682312762e-05}
{"train_loss": 0.16338133811950684, "global_step": 187954, "epoch": 2111, "lr": 2.1656799086798095e-05}
{"train_loss": 0.021193886175751686, "global_step": 187955, "epoch": 2111, "lr": 2.1656321495093816e-05}
{"train_loss": 0.032396577298641205, "global_step": 187956, "epoch": 2111, "lr": 2.165584390720002e-05}
{"train_loss": 0.05831402540206909, "global_step": 187957, "epoch": 2111, "lr": 2.1655366323116743e-05}
{"train_loss": 0.05433889105916023, "global_step": 187958, "epoch": 2111, "lr": 2.165488874284404e-05}
{"train_loss": 0.08540968596935272, "global_step": 187959, "epoch": 2111, "lr": 2.165441116638201e-05}
{"train_loss": 0.07282097637653351, "global_step": 187960, "epoch": 2111, "lr": 2.1653933593730685e-05}
{"train_loss": 0.09016051888465881, "global_step": 187961, "epoch": 2111, "lr": 2.1653456024890156e-05}
{"train_loss": 0.0492352694272995, "global_step": 187962, "epoch": 2111, "lr": 2.1652978459860456e-05}
{"train_loss": 0.08855833113193512, "global_step": 187963, "epoch": 2111, "lr": 2.1652500898641692e-05}
{"train_loss": 0.054795317351818085, "global_step": 187964, "epoch": 2111, "lr": 2.165202334123388e-05}
{"train_loss": 0.09150176495313644, "global_step": 187965, "epoch": 2111, "lr": 2.1651545787637127e-05}
{"train_loss": 0.07435036450624466, "global_step": 187966, "epoch": 2111, "lr": 2.1651068237851462e-05}
{"train_loss": 0.07063989705416594, "global_step": 187967, "epoch": 2111, "lr": 2.165059069187698e-05, "val_loss": 7.727272987365723}
{"train_loss": 0.06099196523427963, "global_step": 187968, "epoch": 2112, "lr": 2.1650113149713717e-05}
{"train_loss": 0.059137795120477676, "global_step": 187969, "epoch": 2112, "lr": 2.1649635611361767e-05}
{"train_loss": 0.07222574204206467, "global_step": 187970, "epoch": 2112, "lr": 2.1649158076821156e-05}
{"train_loss": 0.041598349809646606, "global_step": 187971, "epoch": 2112, "lr": 2.1648680546091994e-05}
{"train_loss": 0.01678476482629776, "global_step": 187972, "epoch": 2112, "lr": 2.16482030191743e-05}
{"train_loss": 0.0778937041759491, "global_step": 187973, "epoch": 2112, "lr": 2.164772549606816e-05}
{"train_loss": 0.043680716305971146, "global_step": 187974, "epoch": 2112, "lr": 2.164724797677366e-05}
{"train_loss": 0.05059605464339256, "global_step": 187975, "epoch": 2112, "lr": 2.1646770461290815e-05}
{"train_loss": 0.037211403250694275, "global_step": 187976, "epoch": 2112, "lr": 2.164629294961974e-05}
{"train_loss": 0.07772587239742279, "global_step": 187977, "epoch": 2112, "lr": 2.164581544176046e-05}
{"train_loss": 0.045903127640485764, "global_step": 187978, "epoch": 2112, "lr": 2.1645337937713068e-05}
{"train_loss": 0.04315816983580589, "global_step": 187979, "epoch": 2112, "lr": 2.16448604374776e-05}
{"train_loss": 0.09700877964496613, "global_step": 187980, "epoch": 2112, "lr": 2.1644382941054146e-05}
{"train_loss": 0.07067687809467316, "global_step": 187981, "epoch": 2112, "lr": 2.1643905448442747e-05}
{"train_loss": 0.06986090540885925, "global_step": 187982, "epoch": 2112, "lr": 2.1643427959643498e-05}
{"train_loss": 0.11503004282712936, "global_step": 187983, "epoch": 2112, "lr": 2.164295047465642e-05}
{"train_loss": 0.069807268679142, "global_step": 187984, "epoch": 2112, "lr": 2.164247299348163e-05}
{"train_loss": 0.07765332609415054, "global_step": 187985, "epoch": 2112, "lr": 2.164199551611914e-05}
{"train_loss": 0.09316504746675491, "global_step": 187986, "epoch": 2112, "lr": 2.1641518042569052e-05}
{"train_loss": 0.05212930589914322, "global_step": 187987, "epoch": 2112, "lr": 2.1641040572831405e-05}
{"train_loss": 0.08245939016342163, "global_step": 187988, "epoch": 2112, "lr": 2.164056310690629e-05}
{"train_loss": 0.038338132202625275, "global_step": 187989, "epoch": 2112, "lr": 2.1640085644793733e-05}
{"train_loss": 0.061026982963085175, "global_step": 187990, "epoch": 2112, "lr": 2.1639608186493847e-05}
{"train_loss": 0.07420875132083893, "global_step": 187991, "epoch": 2112, "lr": 2.1639130732006645e-05}
{"train_loss": 0.03737610578536987, "global_step": 187992, "epoch": 2112, "lr": 2.1638653281332237e-05}
{"train_loss": 0.08824653923511505, "global_step": 187993, "epoch": 2112, "lr": 2.1638175834470653e-05}
{"train_loss": 0.0500195138156414, "global_step": 187994, "epoch": 2112, "lr": 2.1637698391421984e-05}
{"train_loss": 0.071821428835392, "global_step": 187995, "epoch": 2112, "lr": 2.1637220952186282e-05}
{"train_loss": 0.0798560380935669, "global_step": 187996, "epoch": 2112, "lr": 2.163674351676359e-05}
{"train_loss": 0.059691622853279114, "global_step": 187997, "epoch": 2112, "lr": 2.163626608515401e-05}
{"train_loss": 0.06992095708847046, "global_step": 187998, "epoch": 2112, "lr": 2.1635788657357575e-05}
{"train_loss": 0.06712613999843597, "global_step": 187999, "epoch": 2112, "lr": 2.1635311233374378e-05}
{"train_loss": 0.045191213488578796, "global_step": 188000, "epoch": 2112, "lr": 2.1634833813204445e-05}
{"train_loss": 0.1072535291314125, "global_step": 188001, "epoch": 2112, "lr": 2.163435639684789e-05}
{"train_loss": 0.05263904854655266, "global_step": 188002, "epoch": 2112, "lr": 2.1633878984304722e-05}
{"train_loss": 0.07667382806539536, "global_step": 188003, "epoch": 2112, "lr": 2.1633401575575057e-05}
{"train_loss": 0.1095375269651413, "global_step": 188004, "epoch": 2112, "lr": 2.163292417065892e-05}
{"train_loss": 0.03940075635910034, "global_step": 188005, "epoch": 2112, "lr": 2.16324467695564e-05}
{"train_loss": 0.051580894738435745, "global_step": 188006, "epoch": 2112, "lr": 2.1631969372267542e-05}
{"train_loss": 0.06308723241090775, "global_step": 188007, "epoch": 2112, "lr": 2.1631491978792436e-05}
{"train_loss": 0.08284679055213928, "global_step": 188008, "epoch": 2112, "lr": 2.163101458913111e-05}
{"train_loss": 0.08162161707878113, "global_step": 188009, "epoch": 2112, "lr": 2.1630537203283645e-05}
{"train_loss": 0.0782119408249855, "global_step": 188010, "epoch": 2112, "lr": 2.1630059821250138e-05}
{"train_loss": 0.123543381690979, "global_step": 188011, "epoch": 2112, "lr": 2.1629582443030593e-05}
{"train_loss": 0.04373716190457344, "global_step": 188012, "epoch": 2112, "lr": 2.1629105068625132e-05}
{"train_loss": 0.050279926508665085, "global_step": 188013, "epoch": 2112, "lr": 2.1628627698033772e-05}
{"train_loss": 0.04094496741890907, "global_step": 188014, "epoch": 2112, "lr": 2.162815033125661e-05}
{"train_loss": 0.055083490908145905, "global_step": 188015, "epoch": 2112, "lr": 2.1627672968293682e-05}
{"train_loss": 0.041174665093421936, "global_step": 188016, "epoch": 2112, "lr": 2.162719560914509e-05}
{"train_loss": 0.06668221205472946, "global_step": 188017, "epoch": 2112, "lr": 2.162671825381085e-05}
{"train_loss": 0.13028451800346375, "global_step": 188018, "epoch": 2112, "lr": 2.162624090229108e-05}
{"train_loss": 0.11167408525943756, "global_step": 188019, "epoch": 2112, "lr": 2.1625763554585787e-05}
{"train_loss": 0.10320492088794708, "global_step": 188020, "epoch": 2112, "lr": 2.1625286210695088e-05}
{"train_loss": 0.1406630128622055, "global_step": 188021, "epoch": 2112, "lr": 2.1624808870619005e-05}
{"train_loss": 0.06804582476615906, "global_step": 188022, "epoch": 2112, "lr": 2.162433153435764e-05}
{"train_loss": 0.03915996477007866, "global_step": 188023, "epoch": 2112, "lr": 2.1623854201911014e-05}
{"train_loss": 0.07902020215988159, "global_step": 188024, "epoch": 2112, "lr": 2.1623376873279237e-05}
{"train_loss": 0.1254945993423462, "global_step": 188025, "epoch": 2112, "lr": 2.1622899548462337e-05}
{"train_loss": 0.130318745970726, "global_step": 188026, "epoch": 2112, "lr": 2.162242222746041e-05}
{"train_loss": 0.07494594901800156, "global_step": 188027, "epoch": 2112, "lr": 2.1621944910273478e-05}
{"train_loss": 0.10062761604785919, "global_step": 188028, "epoch": 2112, "lr": 2.162146759690165e-05}
{"train_loss": 0.04177962616086006, "global_step": 188029, "epoch": 2112, "lr": 2.1620990287344954e-05}
{"train_loss": 0.09285467863082886, "global_step": 188030, "epoch": 2112, "lr": 2.1620512981603484e-05}
{"train_loss": 0.06203918904066086, "global_step": 188031, "epoch": 2112, "lr": 2.1620035679677274e-05}
{"train_loss": 0.09412959963083267, "global_step": 188032, "epoch": 2112, "lr": 2.1619558381566422e-05}
{"train_loss": 0.05120537057518959, "global_step": 188033, "epoch": 2112, "lr": 2.1619081087270976e-05}
{"train_loss": 0.1480899602174759, "global_step": 188034, "epoch": 2112, "lr": 2.1618603796790983e-05}
{"train_loss": 0.06084918975830078, "global_step": 188035, "epoch": 2112, "lr": 2.1618126510126534e-05}
{"train_loss": 0.1090737134218216, "global_step": 188036, "epoch": 2112, "lr": 2.161764922727767e-05}
{"train_loss": 0.1006212830543518, "global_step": 188037, "epoch": 2112, "lr": 2.1617171948244485e-05}
{"train_loss": 0.10147438198328018, "global_step": 188038, "epoch": 2112, "lr": 2.1616694673027006e-05}
{"train_loss": 0.06257516145706177, "global_step": 188039, "epoch": 2112, "lr": 2.1616217401625332e-05}
{"train_loss": 0.07463038712739944, "global_step": 188040, "epoch": 2112, "lr": 2.1615740134039498e-05}
{"train_loss": 0.05289420112967491, "global_step": 188041, "epoch": 2112, "lr": 2.161526287026959e-05}
{"train_loss": 0.12127791345119476, "global_step": 188042, "epoch": 2112, "lr": 2.161478561031566e-05}
{"train_loss": 0.05547012761235237, "global_step": 188043, "epoch": 2112, "lr": 2.1614308354177787e-05}
{"train_loss": 0.02918991632759571, "global_step": 188044, "epoch": 2112, "lr": 2.1613831101856e-05}
{"train_loss": 0.0366348996758461, "global_step": 188045, "epoch": 2112, "lr": 2.16133538533504e-05}
{"train_loss": 0.08395075052976608, "global_step": 188046, "epoch": 2112, "lr": 2.161287660866105e-05}
{"train_loss": 0.07673540711402893, "global_step": 188047, "epoch": 2112, "lr": 2.1612399367787984e-05}
{"train_loss": 0.12254641950130463, "global_step": 188048, "epoch": 2112, "lr": 2.1611922130731305e-05}
{"train_loss": 0.0566156767308712, "global_step": 188049, "epoch": 2112, "lr": 2.1611444897491033e-05}
{"train_loss": 0.0707322508096695, "global_step": 188050, "epoch": 2112, "lr": 2.1610967668067283e-05}
{"train_loss": 0.06310903280973434, "global_step": 188051, "epoch": 2112, "lr": 2.161049044246007e-05}
{"train_loss": 0.10555684566497803, "global_step": 188052, "epoch": 2112, "lr": 2.16100132206695e-05}
{"train_loss": 0.07798434048891068, "global_step": 188053, "epoch": 2112, "lr": 2.1609536002695596e-05}
{"train_loss": 0.056440237909555435, "global_step": 188054, "epoch": 2112, "lr": 2.1609058788538466e-05}
{"train_loss": 0.060102429240942, "global_step": 188055, "epoch": 2112, "lr": 2.1608581578198134e-05}
{"train_loss": 0.07295158557761251, "global_step": 188056, "epoch": 2112, "lr": 2.1608104371674698e-05, "val_loss": 7.73219633102417}
{"train_loss": 0.06231724098324776, "global_step": 188057, "epoch": 2113, "lr": 2.1607627168968192e-05}
{"train_loss": 0.11171749979257584, "global_step": 188058, "epoch": 2113, "lr": 2.1607149970078712e-05}
{"train_loss": 0.06375644356012344, "global_step": 188059, "epoch": 2113, "lr": 2.160667277500628e-05}
{"train_loss": 0.07015201449394226, "global_step": 188060, "epoch": 2113, "lr": 2.160619558375101e-05}
{"train_loss": 0.048455145210027695, "global_step": 188061, "epoch": 2113, "lr": 2.1605718396312918e-05}
{"train_loss": 0.07750040292739868, "global_step": 188062, "epoch": 2113, "lr": 2.160524121269211e-05}
{"train_loss": 0.13194838166236877, "global_step": 188063, "epoch": 2113, "lr": 2.1604764032888614e-05}
{"train_loss": 0.05653468519449234, "global_step": 188064, "epoch": 2113, "lr": 2.160428685690251e-05}
{"train_loss": 0.0699801966547966, "global_step": 188065, "epoch": 2113, "lr": 2.160380968473389e-05}
{"train_loss": 0.054629869759082794, "global_step": 188066, "epoch": 2113, "lr": 2.160333251638276e-05}
{"train_loss": 0.047371696680784225, "global_step": 188067, "epoch": 2113, "lr": 2.1602855351849243e-05}
{"train_loss": 0.061389658600091934, "global_step": 188068, "epoch": 2113, "lr": 2.1602378191133376e-05}
{"train_loss": 0.079501673579216, "global_step": 188069, "epoch": 2113, "lr": 2.16019010342352e-05}
{"train_loss": 0.04399862885475159, "global_step": 188070, "epoch": 2113, "lr": 2.1601423881154825e-05}
{"train_loss": 0.04085412248969078, "global_step": 188071, "epoch": 2113, "lr": 2.1600946731892273e-05}
{"train_loss": 0.06824871152639389, "global_step": 188072, "epoch": 2113, "lr": 2.1600469586447648e-05}
{"train_loss": 0.039898697286844254, "global_step": 188073, "epoch": 2113, "lr": 2.1599992444820987e-05}
{"train_loss": 0.09165647625923157, "global_step": 188074, "epoch": 2113, "lr": 2.1599515307012347e-05}
{"train_loss": 0.1208631619811058, "global_step": 188075, "epoch": 2113, "lr": 2.1599038173021823e-05}
{"train_loss": 0.12446233630180359, "global_step": 188076, "epoch": 2113, "lr": 2.1598561042849445e-05}
{"train_loss": 0.05813341587781906, "global_step": 188077, "epoch": 2113, "lr": 2.1598083916495317e-05}
{"train_loss": 0.07918469607830048, "global_step": 188078, "epoch": 2113, "lr": 2.1597606793959456e-05}
{"train_loss": 0.05838451161980629, "global_step": 188079, "epoch": 2113, "lr": 2.1597129675241973e-05}
{"train_loss": 0.034883610904216766, "global_step": 188080, "epoch": 2113, "lr": 2.1596652560342884e-05}
{"train_loss": 0.09337066113948822, "global_step": 188081, "epoch": 2113, "lr": 2.1596175449262308e-05}
{"train_loss": 0.08195897191762924, "global_step": 188082, "epoch": 2113, "lr": 2.1595698342000254e-05}
{"train_loss": 0.0523722805082798, "global_step": 188083, "epoch": 2113, "lr": 2.159522123855681e-05}
{"train_loss": 0.06141268461942673, "global_step": 188084, "epoch": 2113, "lr": 2.1594744138932067e-05}
{"train_loss": 0.10738454014062881, "global_step": 188085, "epoch": 2113, "lr": 2.1594267043126043e-05}
{"train_loss": 0.029269587248563766, "global_step": 188086, "epoch": 2113, "lr": 2.1593789951138843e-05}
{"train_loss": 0.06960197538137436, "global_step": 188087, "epoch": 2113, "lr": 2.1593312862970494e-05}
{"train_loss": 0.07389083504676819, "global_step": 188088, "epoch": 2113, "lr": 2.1592835778621095e-05}
{"train_loss": 0.12155210971832275, "global_step": 188089, "epoch": 2113, "lr": 2.1592358698090676e-05}
{"train_loss": 0.10538829863071442, "global_step": 188090, "epoch": 2113, "lr": 2.1591881621379335e-05}
{"train_loss": 0.07847241312265396, "global_step": 188091, "epoch": 2113, "lr": 2.1591404548487097e-05}
{"train_loss": 0.10650515556335449, "global_step": 188092, "epoch": 2113, "lr": 2.1590927479414076e-05}
{"train_loss": 0.03791043907403946, "global_step": 188093, "epoch": 2113, "lr": 2.1590450414160286e-05}
{"train_loss": 0.06352768838405609, "global_step": 188094, "epoch": 2113, "lr": 2.158997335272583e-05}
{"train_loss": 0.06177407503128052, "global_step": 188095, "epoch": 2113, "lr": 2.158949629511074e-05}
{"train_loss": 0.08586985617876053, "global_step": 188096, "epoch": 2113, "lr": 2.158901924131511e-05}
{"train_loss": 0.13466185331344604, "global_step": 188097, "epoch": 2113, "lr": 2.1588542191338973e-05}
{"train_loss": 0.1349351704120636, "global_step": 188098, "epoch": 2113, "lr": 2.1588065145182435e-05}
{"train_loss": 0.09928331524133682, "global_step": 188099, "epoch": 2113, "lr": 2.1587588102845507e-05}
{"train_loss": 0.08989062160253525, "global_step": 188100, "epoch": 2113, "lr": 2.1587111064328292e-05}
{"train_loss": 0.0545230396091938, "global_step": 188101, "epoch": 2113, "lr": 2.1586634029630853e-05}
{"train_loss": 0.05417124554514885, "global_step": 188102, "epoch": 2113, "lr": 2.158615699875323e-05}
{"train_loss": 0.049829792231321335, "global_step": 188103, "epoch": 2113, "lr": 2.158567997169552e-05}
{"train_loss": 0.05260614678263664, "global_step": 188104, "epoch": 2113, "lr": 2.1585202948457757e-05}
{"train_loss": 0.08069609105587006, "global_step": 188105, "epoch": 2113, "lr": 2.1584725929040023e-05}
{"train_loss": 0.08498144149780273, "global_step": 188106, "epoch": 2113, "lr": 2.1584248913442384e-05}
{"train_loss": 0.051176875829696655, "global_step": 188107, "epoch": 2113, "lr": 2.1583771901664877e-05}
{"train_loss": 0.08694219589233398, "global_step": 188108, "epoch": 2113, "lr": 2.1583294893707608e-05}
{"train_loss": 0.07820397615432739, "global_step": 188109, "epoch": 2113, "lr": 2.1582817889570596e-05}
{"train_loss": 0.08531174063682556, "global_step": 188110, "epoch": 2113, "lr": 2.1582340889253943e-05}
{"train_loss": 0.06440578401088715, "global_step": 188111, "epoch": 2113, "lr": 2.1581863892757703e-05}
{"train_loss": 0.08715526759624481, "global_step": 188112, "epoch": 2113, "lr": 2.1581386900081914e-05}
{"train_loss": 0.045985147356987, "global_step": 188113, "epoch": 2113, "lr": 2.1580909911226677e-05}
{"train_loss": 0.06625738739967346, "global_step": 188114, "epoch": 2113, "lr": 2.158043292619203e-05}
{"train_loss": 0.1078554019331932, "global_step": 188115, "epoch": 2113, "lr": 2.1579955944978057e-05}
{"train_loss": 0.03792574256658554, "global_step": 188116, "epoch": 2113, "lr": 2.15794789675848e-05}
{"train_loss": 0.1236875057220459, "global_step": 188117, "epoch": 2113, "lr": 2.1579001994012348e-05}
{"train_loss": 0.05481844022870064, "global_step": 188118, "epoch": 2113, "lr": 2.1578525024260743e-05}
{"train_loss": 0.06121718883514404, "global_step": 188119, "epoch": 2113, "lr": 2.1578048058330054e-05}
{"train_loss": 0.09609150141477585, "global_step": 188120, "epoch": 2113, "lr": 2.1577571096220367e-05}
{"train_loss": 0.11979600042104721, "global_step": 188121, "epoch": 2113, "lr": 2.1577094137931713e-05}
{"train_loss": 0.10468973964452744, "global_step": 188122, "epoch": 2113, "lr": 2.157661718346419e-05}
{"train_loss": 0.06903313100337982, "global_step": 188123, "epoch": 2113, "lr": 2.1576140232817825e-05}
{"train_loss": 0.0787028968334198, "global_step": 188124, "epoch": 2113, "lr": 2.1575663285992726e-05}
{"train_loss": 0.08429772406816483, "global_step": 188125, "epoch": 2113, "lr": 2.1575186342988907e-05}
{"train_loss": 0.04500335827469826, "global_step": 188126, "epoch": 2113, "lr": 2.1574709403806474e-05}
{"train_loss": 0.05673349276185036, "global_step": 188127, "epoch": 2113, "lr": 2.1574232468445464e-05}
{"train_loss": 0.1565479040145874, "global_step": 188128, "epoch": 2113, "lr": 2.157375553690597e-05}
{"train_loss": 0.08526692539453506, "global_step": 188129, "epoch": 2113, "lr": 2.1573278609188012e-05}
{"train_loss": 0.13997408747673035, "global_step": 188130, "epoch": 2113, "lr": 2.1572801685291705e-05}
{"train_loss": 0.06465677171945572, "global_step": 188131, "epoch": 2113, "lr": 2.1572324765217072e-05}
{"train_loss": 0.052197832614183426, "global_step": 188132, "epoch": 2113, "lr": 2.1571847848964204e-05}
{"train_loss": 0.06684385240077972, "global_step": 188133, "epoch": 2113, "lr": 2.157137093653314e-05}
{"train_loss": 0.06387661397457123, "global_step": 188134, "epoch": 2113, "lr": 2.1570894027923983e-05}
{"train_loss": 0.11102528125047684, "global_step": 188135, "epoch": 2113, "lr": 2.157041712313675e-05}
{"train_loss": 0.03367123752832413, "global_step": 188136, "epoch": 2113, "lr": 2.1569940222171525e-05}
{"train_loss": 0.06439542025327682, "global_step": 188137, "epoch": 2113, "lr": 2.1569463325028395e-05}
{"train_loss": 0.06180911883711815, "global_step": 188138, "epoch": 2113, "lr": 2.156898643170739e-05}
{"train_loss": 0.07947415858507156, "global_step": 188139, "epoch": 2113, "lr": 2.15685095422086e-05}
{"train_loss": 0.07614953070878983, "global_step": 188140, "epoch": 2113, "lr": 2.156803265653206e-05}
{"train_loss": 0.09424839913845062, "global_step": 188141, "epoch": 2113, "lr": 2.1567555774677872e-05}
{"train_loss": 0.06939741224050522, "global_step": 188142, "epoch": 2113, "lr": 2.1567078896646058e-05}
{"train_loss": 0.10338735580444336, "global_step": 188143, "epoch": 2113, "lr": 2.1566602022436727e-05}
{"train_loss": 0.05102712661027908, "global_step": 188144, "epoch": 2113, "lr": 2.156612515204992e-05}
{"train_loss": 0.076899449477035, "global_step": 188145, "epoch": 2113, "lr": 2.156564828548568e-05, "val_loss": 7.685614585876465}
{"train_loss": 0.10107403993606567, "global_step": 188146, "epoch": 2114, "lr": 2.1565171422744114e-05}
{"train_loss": 0.09589079022407532, "global_step": 188147, "epoch": 2114, "lr": 2.156469456382524e-05}
{"train_loss": 0.06861859560012817, "global_step": 188148, "epoch": 2114, "lr": 2.1564217708729166e-05}
{"train_loss": 0.09617089480161667, "global_step": 188149, "epoch": 2114, "lr": 2.156374085745594e-05}
{"train_loss": 0.05442149192094803, "global_step": 188150, "epoch": 2114, "lr": 2.1563264010005596e-05}
{"train_loss": 0.06151888146996498, "global_step": 188151, "epoch": 2114, "lr": 2.156278716637825e-05}
{"train_loss": 0.12896732985973358, "global_step": 188152, "epoch": 2114, "lr": 2.156231032657392e-05}
{"train_loss": 0.035890549421310425, "global_step": 188153, "epoch": 2114, "lr": 2.156183349059271e-05}
{"train_loss": 0.07912790030241013, "global_step": 188154, "epoch": 2114, "lr": 2.1561356658434644e-05}
{"train_loss": 0.02852184697985649, "global_step": 188155, "epoch": 2114, "lr": 2.156087983009981e-05}
{"train_loss": 0.08629801124334335, "global_step": 188156, "epoch": 2114, "lr": 2.1560403005588292e-05}
{"train_loss": 0.06565506011247635, "global_step": 188157, "epoch": 2114, "lr": 2.1559926184900104e-05}
{"train_loss": 0.09291215986013412, "global_step": 188158, "epoch": 2114, "lr": 2.155944936803536e-05}
{"train_loss": 0.07538987696170807, "global_step": 188159, "epoch": 2114, "lr": 2.1558972554994085e-05}
{"train_loss": 0.07949456572532654, "global_step": 188160, "epoch": 2114, "lr": 2.1558495745776376e-05}
{"train_loss": 0.06447622179985046, "global_step": 188161, "epoch": 2114, "lr": 2.1558018940382258e-05}
{"train_loss": 0.08826971054077148, "global_step": 188162, "epoch": 2114, "lr": 2.1557542138811837e-05}
{"train_loss": 0.07959042489528656, "global_step": 188163, "epoch": 2114, "lr": 2.1557065341065146e-05}
{"train_loss": 0.06277596205472946, "global_step": 188164, "epoch": 2114, "lr": 2.1556588547142274e-05}
{"train_loss": 0.046732641756534576, "global_step": 188165, "epoch": 2114, "lr": 2.155611175704325e-05}
{"train_loss": 0.06592452526092529, "global_step": 188166, "epoch": 2114, "lr": 2.1555634970768185e-05}
{"train_loss": 0.038146939128637314, "global_step": 188167, "epoch": 2114, "lr": 2.15551581883171e-05}
{"train_loss": 0.10806266218423843, "global_step": 188168, "epoch": 2114, "lr": 2.155468140969009e-05}
{"train_loss": 0.055735036730766296, "global_step": 188169, "epoch": 2114, "lr": 2.155420463488719e-05}
{"train_loss": 0.09224351495504379, "global_step": 188170, "epoch": 2114, "lr": 2.1553727863908507e-05}
{"train_loss": 0.07393509149551392, "global_step": 188171, "epoch": 2114, "lr": 2.155325109675405e-05}
{"train_loss": 0.049292951822280884, "global_step": 188172, "epoch": 2114, "lr": 2.155277433342392e-05}
{"train_loss": 0.07549726217985153, "global_step": 188173, "epoch": 2114, "lr": 2.1552297573918184e-05}
{"train_loss": 0.06805513799190521, "global_step": 188174, "epoch": 2114, "lr": 2.1551820818236877e-05}
{"train_loss": 0.043767452239990234, "global_step": 188175, "epoch": 2114, "lr": 2.1551344066380104e-05}
{"train_loss": 0.04853537678718567, "global_step": 188176, "epoch": 2114, "lr": 2.1550867318347884e-05}
{"train_loss": 0.12890566885471344, "global_step": 188177, "epoch": 2114, "lr": 2.1550390574140323e-05}
{"train_loss": 0.08078458160161972, "global_step": 188178, "epoch": 2114, "lr": 2.1549913833757445e-05}
{"train_loss": 0.058590035885572433, "global_step": 188179, "epoch": 2114, "lr": 2.1549437097199354e-05}
{"train_loss": 0.0772940069437027, "global_step": 188180, "epoch": 2114, "lr": 2.154896036446608e-05}
{"train_loss": 0.10908618569374084, "global_step": 188181, "epoch": 2114, "lr": 2.1548483635557713e-05}
{"train_loss": 0.0688949003815651, "global_step": 188182, "epoch": 2114, "lr": 2.1548006910474293e-05}
{"train_loss": 0.08164768666028976, "global_step": 188183, "epoch": 2114, "lr": 2.1547530189215907e-05}
{"train_loss": 0.10393211245536804, "global_step": 188184, "epoch": 2114, "lr": 2.154705347178262e-05}
{"train_loss": 0.0639086440205574, "global_step": 188185, "epoch": 2114, "lr": 2.1546576758174457e-05}
{"train_loss": 0.07384912669658661, "global_step": 188186, "epoch": 2114, "lr": 2.154610004839153e-05}
{"train_loss": 0.07050813734531403, "global_step": 188187, "epoch": 2114, "lr": 2.1545623342433867e-05}
{"train_loss": 0.06432077288627625, "global_step": 188188, "epoch": 2114, "lr": 2.1545146640301567e-05}
{"train_loss": 0.1448441445827484, "global_step": 188189, "epoch": 2114, "lr": 2.1544669941994677e-05}
{"train_loss": 0.11417683213949203, "global_step": 188190, "epoch": 2114, "lr": 2.154419324751324e-05}
{"train_loss": 0.05966440960764885, "global_step": 188191, "epoch": 2114, "lr": 2.1543716556857334e-05}
{"train_loss": 0.08705020695924759, "global_step": 188192, "epoch": 2114, "lr": 2.1543239870027053e-05}
{"train_loss": 0.035556159913539886, "global_step": 188193, "epoch": 2114, "lr": 2.1542763187022414e-05}
{"train_loss": 0.05866714194417, "global_step": 188194, "epoch": 2114, "lr": 2.1542286507843524e-05}
{"train_loss": 0.0298245120793581, "global_step": 188195, "epoch": 2114, "lr": 2.1541809832490407e-05}
{"train_loss": 0.07159730046987534, "global_step": 188196, "epoch": 2114, "lr": 2.154133316096317e-05}
{"train_loss": 0.053255967795848846, "global_step": 188197, "epoch": 2114, "lr": 2.154085649326183e-05}
{"train_loss": 0.08379457890987396, "global_step": 188198, "epoch": 2114, "lr": 2.1540379829386494e-05}
{"train_loss": 0.06047534570097923, "global_step": 188199, "epoch": 2114, "lr": 2.153990316933719e-05}
{"train_loss": 0.06732573360204697, "global_step": 188200, "epoch": 2114, "lr": 2.153942651311402e-05}
{"train_loss": 0.0721181184053421, "global_step": 188201, "epoch": 2114, "lr": 2.1538949860717007e-05}
{"train_loss": 0.045569244772195816, "global_step": 188202, "epoch": 2114, "lr": 2.153847321214625e-05}
{"train_loss": 0.11508000642061234, "global_step": 188203, "epoch": 2114, "lr": 2.153799656740178e-05}
{"train_loss": 0.05607821047306061, "global_step": 188204, "epoch": 2114, "lr": 2.1537519926483703e-05}
{"train_loss": 0.05764000117778778, "global_step": 188205, "epoch": 2114, "lr": 2.153704328939204e-05}
{"train_loss": 0.09010646492242813, "global_step": 188206, "epoch": 2114, "lr": 2.153656665612689e-05}
{"train_loss": 0.04362041503190994, "global_step": 188207, "epoch": 2114, "lr": 2.153609002668829e-05}
{"train_loss": 0.1014857068657875, "global_step": 188208, "epoch": 2114, "lr": 2.1535613401076328e-05}
{"train_loss": 0.06471626460552216, "global_step": 188209, "epoch": 2114, "lr": 2.153513677929104e-05}
{"train_loss": 0.0672992616891861, "global_step": 188210, "epoch": 2114, "lr": 2.1534660161332498e-05}
{"train_loss": 0.09169431775808334, "global_step": 188211, "epoch": 2114, "lr": 2.1534183547200802e-05}
{"train_loss": 0.07793167978525162, "global_step": 188212, "epoch": 2114, "lr": 2.1533706936895963e-05}
{"train_loss": 0.07321272790431976, "global_step": 188213, "epoch": 2114, "lr": 2.1533230330418098e-05}
{"train_loss": 0.08371530473232269, "global_step": 188214, "epoch": 2114, "lr": 2.1532753727767218e-05}
{"train_loss": 0.07283809036016464, "global_step": 188215, "epoch": 2114, "lr": 2.153227712894343e-05}
{"train_loss": 0.09265345335006714, "global_step": 188216, "epoch": 2114, "lr": 2.1531800533946767e-05}
{"train_loss": 0.06275215744972229, "global_step": 188217, "epoch": 2114, "lr": 2.153132394277732e-05}
{"train_loss": 0.06921501457691193, "global_step": 188218, "epoch": 2114, "lr": 2.153084735543512e-05}
{"train_loss": 0.06864938884973526, "global_step": 188219, "epoch": 2114, "lr": 2.1530370771920272e-05}
{"train_loss": 0.05960964038968086, "global_step": 188220, "epoch": 2114, "lr": 2.15298941922328e-05}
{"train_loss": 0.0657924935221672, "global_step": 188221, "epoch": 2114, "lr": 2.1529417616372805e-05}
{"train_loss": 0.06332380324602127, "global_step": 188222, "epoch": 2114, "lr": 2.1528941044340334e-05}
{"train_loss": 0.06959637999534607, "global_step": 188223, "epoch": 2114, "lr": 2.1528464476135428e-05}
{"train_loss": 0.05665966868400574, "global_step": 188224, "epoch": 2114, "lr": 2.1527987911758196e-05}
{"train_loss": 0.07750402390956879, "global_step": 188225, "epoch": 2114, "lr": 2.1527511351208656e-05}
{"train_loss": 0.06867574900388718, "global_step": 188226, "epoch": 2114, "lr": 2.1527034794486918e-05}
{"train_loss": 0.08549730479717255, "global_step": 188227, "epoch": 2114, "lr": 2.1526558241593004e-05}
{"train_loss": 0.1113375872373581, "global_step": 188228, "epoch": 2114, "lr": 2.152608169252701e-05}
{"train_loss": 0.1353769302368164, "global_step": 188229, "epoch": 2114, "lr": 2.1525605147288973e-05}
{"train_loss": 0.11802920699119568, "global_step": 188230, "epoch": 2114, "lr": 2.1525128605878993e-05}
{"train_loss": 0.05308622866868973, "global_step": 188231, "epoch": 2114, "lr": 2.152465206829709e-05}
{"train_loss": 0.11073381453752518, "global_step": 188232, "epoch": 2114, "lr": 2.152417553454337e-05}
{"train_loss": 0.11502823978662491, "global_step": 188233, "epoch": 2114, "lr": 2.152369900461786e-05}
{"train_loss": 0.07575385136467018, "global_step": 188234, "epoch": 2114, "lr": 2.1523222478520656e-05, "val_loss": 7.620145320892334}
{"train_loss": 0.09796677529811859, "global_step": 188235, "epoch": 2115, "lr": 2.1522745956251794e-05}
{"train_loss": 0.05210540071129799, "global_step": 188236, "epoch": 2115, "lr": 2.1522269437811365e-05}
{"train_loss": 0.0823117271065712, "global_step": 188237, "epoch": 2115, "lr": 2.1521792923199406e-05}
{"train_loss": 0.12329025566577911, "global_step": 188238, "epoch": 2115, "lr": 2.152131641241601e-05}
{"train_loss": 0.08621495962142944, "global_step": 188239, "epoch": 2115, "lr": 2.152083990546121e-05}
{"train_loss": 0.09219758212566376, "global_step": 188240, "epoch": 2115, "lr": 2.1520363402335104e-05}
{"train_loss": 0.07796993851661682, "global_step": 188241, "epoch": 2115, "lr": 2.1519886903037718e-05}
{"train_loss": 0.1293676793575287, "global_step": 188242, "epoch": 2115, "lr": 2.1519410407569157e-05}
{"train_loss": 0.06823724508285522, "global_step": 188243, "epoch": 2115, "lr": 2.1518933915929447e-05}
{"train_loss": 0.07009348273277283, "global_step": 188244, "epoch": 2115, "lr": 2.151845742811868e-05}
{"train_loss": 0.1274222582578659, "global_step": 188245, "epoch": 2115, "lr": 2.1517980944136894e-05}
{"train_loss": 0.07323997467756271, "global_step": 188246, "epoch": 2115, "lr": 2.1517504463984174e-05}
{"train_loss": 0.10871628671884537, "global_step": 188247, "epoch": 2115, "lr": 2.1517027987660594e-05}
{"train_loss": 0.051959168165922165, "global_step": 188248, "epoch": 2115, "lr": 2.1516551515166183e-05}
{"train_loss": 0.060271333903074265, "global_step": 188249, "epoch": 2115, "lr": 2.151607504650104e-05}
{"train_loss": 0.04840148985385895, "global_step": 188250, "epoch": 2115, "lr": 2.1515598581665198e-05}
{"train_loss": 0.0915534645318985, "global_step": 188251, "epoch": 2115, "lr": 2.1515122120658753e-05}
{"train_loss": 0.05841951444745064, "global_step": 188252, "epoch": 2115, "lr": 2.1514645663481743e-05}
{"train_loss": 0.09625426679849625, "global_step": 188253, "epoch": 2115, "lr": 2.151416921013425e-05}
{"train_loss": 0.08152993768453598, "global_step": 188254, "epoch": 2115, "lr": 2.1513692760616317e-05}
{"train_loss": 0.04924318939447403, "global_step": 188255, "epoch": 2115, "lr": 2.1513216314928035e-05}
{"train_loss": 0.019206732511520386, "global_step": 188256, "epoch": 2115, "lr": 2.1512739873069437e-05}
{"train_loss": 0.13297806680202484, "global_step": 188257, "epoch": 2115, "lr": 2.151226343504062e-05}
{"train_loss": 0.07380814105272293, "global_step": 188258, "epoch": 2115, "lr": 2.1511787000841616e-05}
{"train_loss": 0.12304413318634033, "global_step": 188259, "epoch": 2115, "lr": 2.1511310570472525e-05}
{"train_loss": 0.0670391172170639, "global_step": 188260, "epoch": 2115, "lr": 2.151083414393339e-05}
{"train_loss": 0.08270058780908585, "global_step": 188261, "epoch": 2115, "lr": 2.1510357721224257e-05}
{"train_loss": 0.07696493715047836, "global_step": 188262, "epoch": 2115, "lr": 2.150988130234523e-05}
{"train_loss": 0.08636125922203064, "global_step": 188263, "epoch": 2115, "lr": 2.1509404887296332e-05}
{"train_loss": 0.07918756455183029, "global_step": 188264, "epoch": 2115, "lr": 2.1508928476077667e-05}
{"train_loss": 0.06462696194648743, "global_step": 188265, "epoch": 2115, "lr": 2.1508452068689254e-05}
{"train_loss": 0.0900932177901268, "global_step": 188266, "epoch": 2115, "lr": 2.150797566513121e-05}
{"train_loss": 0.09137611091136932, "global_step": 188267, "epoch": 2115, "lr": 2.1507499265403548e-05}
{"train_loss": 0.051036179065704346, "global_step": 188268, "epoch": 2115, "lr": 2.1507022869506377e-05}
{"train_loss": 0.03495365008711815, "global_step": 188269, "epoch": 2115, "lr": 2.1506546477439717e-05}
{"train_loss": 0.10489597916603088, "global_step": 188270, "epoch": 2115, "lr": 2.1506070089203674e-05}
{"train_loss": 0.04207059368491173, "global_step": 188271, "epoch": 2115, "lr": 2.1505593704798277e-05}
{"train_loss": 0.10507144033908844, "global_step": 188272, "epoch": 2115, "lr": 2.1505117324223624e-05}
{"train_loss": 0.06605106592178345, "global_step": 188273, "epoch": 2115, "lr": 2.150464094747974e-05}
{"train_loss": 0.1477849781513214, "global_step": 188274, "epoch": 2115, "lr": 2.1504164574566726e-05}
{"train_loss": 0.05923549830913544, "global_step": 188275, "epoch": 2115, "lr": 2.1503688205484612e-05}
{"train_loss": 0.03384847566485405, "global_step": 188276, "epoch": 2115, "lr": 2.15032118402335e-05}
{"train_loss": 0.09678269922733307, "global_step": 188277, "epoch": 2115, "lr": 2.150273547881342e-05}
{"train_loss": 0.09651122242212296, "global_step": 188278, "epoch": 2115, "lr": 2.1502259121224462e-05}
{"train_loss": 0.03352547064423561, "global_step": 188279, "epoch": 2115, "lr": 2.1501782767466662e-05}
{"train_loss": 0.06568580120801926, "global_step": 188280, "epoch": 2115, "lr": 2.1501306417540118e-05}
{"train_loss": 0.07967910170555115, "global_step": 188281, "epoch": 2115, "lr": 2.1500830071444855e-05}
{"train_loss": 0.06670234352350235, "global_step": 188282, "epoch": 2115, "lr": 2.150035372918096e-05}
{"train_loss": 0.06879813969135284, "global_step": 188283, "epoch": 2115, "lr": 2.1499877390748514e-05}
{"train_loss": 0.07593848556280136, "global_step": 188284, "epoch": 2115, "lr": 2.1499401056147543e-05}
{"train_loss": 0.01890031434595585, "global_step": 188285, "epoch": 2115, "lr": 2.149892472537815e-05}
{"train_loss": 0.07151495665311813, "global_step": 188286, "epoch": 2115, "lr": 2.1498448398440363e-05}
{"train_loss": 0.07552310824394226, "global_step": 188287, "epoch": 2115, "lr": 2.1497972075334277e-05}
{"train_loss": 0.13430748879909515, "global_step": 188288, "epoch": 2115, "lr": 2.149749575605992e-05}
{"train_loss": 0.031468335539102554, "global_step": 188289, "epoch": 2115, "lr": 2.14970194406174e-05}
{"train_loss": 0.06901778280735016, "global_step": 188290, "epoch": 2115, "lr": 2.149654312900674e-05}
{"train_loss": 0.06820131093263626, "global_step": 188291, "epoch": 2115, "lr": 2.149606682122804e-05}
{"train_loss": 0.06912315636873245, "global_step": 188292, "epoch": 2115, "lr": 2.1495590517281327e-05}
{"train_loss": 0.04366428777575493, "global_step": 188293, "epoch": 2115, "lr": 2.1495114217166707e-05}
{"train_loss": 0.06258193403482437, "global_step": 188294, "epoch": 2115, "lr": 2.14946379208842e-05}
{"train_loss": 0.03643253445625305, "global_step": 188295, "epoch": 2115, "lr": 2.1494161628433908e-05}
{"train_loss": 0.05613814666867256, "global_step": 188296, "epoch": 2115, "lr": 2.1493685339815862e-05}
{"train_loss": 0.07313382625579834, "global_step": 188297, "epoch": 2115, "lr": 2.1493209055030163e-05}
{"train_loss": 0.07677707076072693, "global_step": 188298, "epoch": 2115, "lr": 2.149273277407683e-05}
{"train_loss": 0.023049714043736458, "global_step": 188299, "epoch": 2115, "lr": 2.1492256496955976e-05}
{"train_loss": 0.03972340747714043, "global_step": 188300, "epoch": 2115, "lr": 2.1491780223667634e-05}
{"train_loss": 0.03825180232524872, "global_step": 188301, "epoch": 2115, "lr": 2.149130395421186e-05}
{"train_loss": 0.04749532788991928, "global_step": 188302, "epoch": 2115, "lr": 2.1490827688588755e-05}
{"train_loss": 0.07133915275335312, "global_step": 188303, "epoch": 2115, "lr": 2.149035142679834e-05}
{"train_loss": 0.06823955476284027, "global_step": 188304, "epoch": 2115, "lr": 2.1489875168840718e-05}
{"train_loss": 0.027849305421113968, "global_step": 188305, "epoch": 2115, "lr": 2.1489398914715914e-05}
{"train_loss": 0.1292812079191208, "global_step": 188306, "epoch": 2115, "lr": 2.1488922664424032e-05}
{"train_loss": 0.06421541422605515, "global_step": 188307, "epoch": 2115, "lr": 2.1488446417965103e-05}
{"train_loss": 0.04872415214776993, "global_step": 188308, "epoch": 2115, "lr": 2.1487970175339218e-05}
{"train_loss": 0.09319542348384857, "global_step": 188309, "epoch": 2115, "lr": 2.148749393654641e-05}
{"train_loss": 0.02811611257493496, "global_step": 188310, "epoch": 2115, "lr": 2.148701770158678e-05}
{"train_loss": 0.020079080015420914, "global_step": 188311, "epoch": 2115, "lr": 2.1486541470460358e-05}
{"train_loss": 0.1129598319530487, "global_step": 188312, "epoch": 2115, "lr": 2.1486065243167237e-05}
{"train_loss": 0.056844428181648254, "global_step": 188313, "epoch": 2115, "lr": 2.148558901970745e-05}
{"train_loss": 0.05717535689473152, "global_step": 188314, "epoch": 2115, "lr": 2.1485112800081096e-05}
{"train_loss": 0.07510899752378464, "global_step": 188315, "epoch": 2115, "lr": 2.1484636584288202e-05}
{"train_loss": 0.030933139845728874, "global_step": 188316, "epoch": 2115, "lr": 2.148416037232887e-05}
{"train_loss": 0.05283364653587341, "global_step": 188317, "epoch": 2115, "lr": 2.1483684164203132e-05}
{"train_loss": 0.11977846175432205, "global_step": 188318, "epoch": 2115, "lr": 2.1483207959911062e-05}
{"train_loss": 0.10808011889457703, "global_step": 188319, "epoch": 2115, "lr": 2.148273175945274e-05}
{"train_loss": 0.0677296444773674, "global_step": 188320, "epoch": 2115, "lr": 2.1482255562828202e-05}
{"train_loss": 0.05236604064702988, "global_step": 188321, "epoch": 2115, "lr": 2.148177937003755e-05}
{"train_loss": 0.05751829221844673, "global_step": 188322, "epoch": 2115, "lr": 2.1481303181080803e-05}
{"train_loss": 0.07171964092870776, "global_step": 188323, "epoch": 2115, "lr": 2.1480826995958064e-05, "val_loss": 7.880461692810059, "train_action_mse_error": 17.62080955505371}
{"train_loss": 0.05983826145529747, "global_step": 188324, "epoch": 2116, "lr": 2.1480350814669365e-05}
{"train_loss": 0.041730307042598724, "global_step": 188325, "epoch": 2116, "lr": 2.14798746372148e-05}
{"train_loss": 0.05512547120451927, "global_step": 188326, "epoch": 2116, "lr": 2.14793984635944e-05}
{"train_loss": 0.10298430919647217, "global_step": 188327, "epoch": 2116, "lr": 2.1478922293808267e-05}
{"train_loss": 0.07820197939872742, "global_step": 188328, "epoch": 2116, "lr": 2.147844612785643e-05}
{"train_loss": 0.06072331219911575, "global_step": 188329, "epoch": 2116, "lr": 2.1477969965738982e-05}
{"train_loss": 0.04112925007939339, "global_step": 188330, "epoch": 2116, "lr": 2.1477493807455956e-05}
{"train_loss": 0.05213819071650505, "global_step": 188331, "epoch": 2116, "lr": 2.147701765300745e-05}
{"train_loss": 0.09864800423383713, "global_step": 188332, "epoch": 2116, "lr": 2.14765415023935e-05}
{"train_loss": 0.030868250876665115, "global_step": 188333, "epoch": 2116, "lr": 2.1476065355614193e-05}
{"train_loss": 0.04267973452806473, "global_step": 188334, "epoch": 2116, "lr": 2.1475589212669568e-05}
{"train_loss": 0.05125392973423004, "global_step": 188335, "epoch": 2116, "lr": 2.1475113073559715e-05}
{"train_loss": 0.05107259750366211, "global_step": 188336, "epoch": 2116, "lr": 2.1474636938284675e-05}
{"train_loss": 0.0742429792881012, "global_step": 188337, "epoch": 2116, "lr": 2.1474160806844533e-05}
{"train_loss": 0.07628019899129868, "global_step": 188338, "epoch": 2116, "lr": 2.1473684679239346e-05}
{"train_loss": 0.06730856746435165, "global_step": 188339, "epoch": 2116, "lr": 2.1473208555469155e-05}
{"train_loss": 0.05140456184744835, "global_step": 188340, "epoch": 2116, "lr": 2.1472732435534065e-05}
{"train_loss": 0.040278658270835876, "global_step": 188341, "epoch": 2116, "lr": 2.14722563194341e-05}
{"train_loss": 0.025747884064912796, "global_step": 188342, "epoch": 2116, "lr": 2.147178020716935e-05}
{"train_loss": 0.12358936667442322, "global_step": 188343, "epoch": 2116, "lr": 2.1471304098739864e-05}
{"train_loss": 0.05198762193322182, "global_step": 188344, "epoch": 2116, "lr": 2.1470827994145733e-05}
{"train_loss": 0.06458255648612976, "global_step": 188345, "epoch": 2116, "lr": 2.1470351893386974e-05}
{"train_loss": 0.10520567744970322, "global_step": 188346, "epoch": 2116, "lr": 2.1469875796463702e-05}
{"train_loss": 0.1321285218000412, "global_step": 188347, "epoch": 2116, "lr": 2.146939970337594e-05}
{"train_loss": 0.0678408220410347, "global_step": 188348, "epoch": 2116, "lr": 2.1468923614123786e-05}
{"train_loss": 0.0941443219780922, "global_step": 188349, "epoch": 2116, "lr": 2.146844752870727e-05}
{"train_loss": 0.08078701794147491, "global_step": 188350, "epoch": 2116, "lr": 2.146797144712649e-05}
{"train_loss": 0.06165868788957596, "global_step": 188351, "epoch": 2116, "lr": 2.146749536938148e-05}
{"train_loss": 0.10030292719602585, "global_step": 188352, "epoch": 2116, "lr": 2.146701929547233e-05}
{"train_loss": 0.07944297045469284, "global_step": 188353, "epoch": 2116, "lr": 2.146654322539908e-05}
{"train_loss": 0.05124887451529503, "global_step": 188354, "epoch": 2116, "lr": 2.1466067159161794e-05}
{"train_loss": 0.11194955557584763, "global_step": 188355, "epoch": 2116, "lr": 2.146559109676058e-05}
{"train_loss": 0.10231930762529373, "global_step": 188356, "epoch": 2116, "lr": 2.1465115038195443e-05}
{"train_loss": 0.044651735574007034, "global_step": 188357, "epoch": 2116, "lr": 2.146463898346649e-05}
{"train_loss": 0.09457992762327194, "global_step": 188358, "epoch": 2116, "lr": 2.1464162932573757e-05}
{"train_loss": 0.03382169455289841, "global_step": 188359, "epoch": 2116, "lr": 2.146368688551733e-05}
{"train_loss": 0.12198076397180557, "global_step": 188360, "epoch": 2116, "lr": 2.146321084229725e-05}
{"train_loss": 0.047338422387838364, "global_step": 188361, "epoch": 2116, "lr": 2.146273480291361e-05}
{"train_loss": 0.06249957159161568, "global_step": 188362, "epoch": 2116, "lr": 2.146225876736644e-05}
{"train_loss": 0.06730537116527557, "global_step": 188363, "epoch": 2116, "lr": 2.146178273565584e-05}
{"train_loss": 0.09559132903814316, "global_step": 188364, "epoch": 2116, "lr": 2.1461306707781836e-05}
{"train_loss": 0.0679640918970108, "global_step": 188365, "epoch": 2116, "lr": 2.1460830683744526e-05}
{"train_loss": 0.05749683082103729, "global_step": 188366, "epoch": 2116, "lr": 2.1460354663543947e-05}
{"train_loss": 0.05811842530965805, "global_step": 188367, "epoch": 2116, "lr": 2.1459878647180193e-05}
{"train_loss": 0.08642162382602692, "global_step": 188368, "epoch": 2116, "lr": 2.1459402634653292e-05}
{"train_loss": 0.05254543200135231, "global_step": 188369, "epoch": 2116, "lr": 2.1458926625963345e-05}
{"train_loss": 0.029430583119392395, "global_step": 188370, "epoch": 2116, "lr": 2.1458450621110376e-05}
{"train_loss": 0.05869531258940697, "global_step": 188371, "epoch": 2116, "lr": 2.145797462009449e-05}
{"train_loss": 0.07383611798286438, "global_step": 188372, "epoch": 2116, "lr": 2.145749862291571e-05}
{"train_loss": 0.06226997822523117, "global_step": 188373, "epoch": 2116, "lr": 2.1457022629574146e-05}
{"train_loss": 0.06795988976955414, "global_step": 188374, "epoch": 2116, "lr": 2.1456546640069818e-05}
{"train_loss": 0.027067672461271286, "global_step": 188375, "epoch": 2116, "lr": 2.1456070654402826e-05}
{"train_loss": 0.047541871666908264, "global_step": 188376, "epoch": 2116, "lr": 2.1455594672573215e-05}
{"train_loss": 0.04827199876308441, "global_step": 188377, "epoch": 2116, "lr": 2.145511869458104e-05}
{"train_loss": 0.0515156053006649, "global_step": 188378, "epoch": 2116, "lr": 2.1454642720426388e-05}
{"train_loss": 0.05784185603260994, "global_step": 188379, "epoch": 2116, "lr": 2.1454166750109294e-05}
{"train_loss": 0.05917206406593323, "global_step": 188380, "epoch": 2116, "lr": 2.1453690783629858e-05}
{"train_loss": 0.052307769656181335, "global_step": 188381, "epoch": 2116, "lr": 2.145321482098811e-05}
{"train_loss": 0.07530932873487473, "global_step": 188382, "epoch": 2116, "lr": 2.1452738862184147e-05}
{"train_loss": 0.049039460718631744, "global_step": 188383, "epoch": 2116, "lr": 2.1452262907217996e-05}
{"train_loss": 0.08727964013814926, "global_step": 188384, "epoch": 2116, "lr": 2.145178695608976e-05}
{"train_loss": 0.07127571851015091, "global_step": 188385, "epoch": 2116, "lr": 2.1451311008799462e-05}
{"train_loss": 0.08011537790298462, "global_step": 188386, "epoch": 2116, "lr": 2.1450835065347207e-05}
{"train_loss": 0.0734756663441658, "global_step": 188387, "epoch": 2116, "lr": 2.145035912573302e-05}
{"train_loss": 0.07925616949796677, "global_step": 188388, "epoch": 2116, "lr": 2.1449883189957004e-05}
{"train_loss": 0.0535246804356575, "global_step": 188389, "epoch": 2116, "lr": 2.1449407258019183e-05}
{"train_loss": 0.04958547651767731, "global_step": 188390, "epoch": 2116, "lr": 2.1448931329919647e-05}
{"train_loss": 0.09735334664583206, "global_step": 188391, "epoch": 2116, "lr": 2.1448455405658468e-05}
{"train_loss": 0.09735333919525146, "global_step": 188392, "epoch": 2116, "lr": 2.1447979485235676e-05}
{"train_loss": 0.08100225776433945, "global_step": 188393, "epoch": 2116, "lr": 2.1447503568651377e-05}
{"train_loss": 0.05557344853878021, "global_step": 188394, "epoch": 2116, "lr": 2.1447027655905594e-05}
{"train_loss": 0.07410626113414764, "global_step": 188395, "epoch": 2116, "lr": 2.144655174699843e-05}
{"train_loss": 0.05215883255004883, "global_step": 188396, "epoch": 2116, "lr": 2.1446075841929902e-05}
{"train_loss": 0.0515722930431366, "global_step": 188397, "epoch": 2116, "lr": 2.1445599940700128e-05}
{"train_loss": 0.04805257171392441, "global_step": 188398, "epoch": 2116, "lr": 2.1445124043309128e-05}
{"train_loss": 0.0948687419295311, "global_step": 188399, "epoch": 2116, "lr": 2.1444648149756996e-05}
{"train_loss": 0.14833888411521912, "global_step": 188400, "epoch": 2116, "lr": 2.1444172260043767e-05}
{"train_loss": 0.11630793660879135, "global_step": 188401, "epoch": 2116, "lr": 2.1443696374169535e-05}
{"train_loss": 0.05350469425320625, "global_step": 188402, "epoch": 2116, "lr": 2.144322049213434e-05}
{"train_loss": 0.07838647067546844, "global_step": 188403, "epoch": 2116, "lr": 2.144274461393827e-05}
{"train_loss": 0.04663414508104324, "global_step": 188404, "epoch": 2116, "lr": 2.1442268739581355e-05}
{"train_loss": 0.11070447415113449, "global_step": 188405, "epoch": 2116, "lr": 2.1441792869063698e-05}
{"train_loss": 0.08758246898651123, "global_step": 188406, "epoch": 2116, "lr": 2.144131700238533e-05}
{"train_loss": 0.04326467961072922, "global_step": 188407, "epoch": 2116, "lr": 2.1440841139546342e-05}
{"train_loss": 0.0417431965470314, "global_step": 188408, "epoch": 2116, "lr": 2.144036528054677e-05}
{"train_loss": 0.06607135385274887, "global_step": 188409, "epoch": 2116, "lr": 2.143988942538669e-05}
{"train_loss": 0.12831686437129974, "global_step": 188410, "epoch": 2116, "lr": 2.1439413574066188e-05}
{"train_loss": 0.07008615136146545, "global_step": 188411, "epoch": 2116, "lr": 2.1438937726585306e-05}
{"train_loss": 0.06926379721151309, "global_step": 188412, "epoch": 2116, "lr": 2.1438461882944093e-05, "val_loss": 7.882483959197998}
{"train_loss": 0.078913114964962, "global_step": 188413, "epoch": 2117, "lr": 2.1437986043142656e-05}
{"train_loss": 0.04405660554766655, "global_step": 188414, "epoch": 2117, "lr": 2.143751020718101e-05}
{"train_loss": 0.08132154494524002, "global_step": 188415, "epoch": 2117, "lr": 2.143703437505926e-05}
{"train_loss": 0.04010649025440216, "global_step": 188416, "epoch": 2117, "lr": 2.1436558546777453e-05}
{"train_loss": 0.08065090328454971, "global_step": 188417, "epoch": 2117, "lr": 2.1436082722335637e-05}
{"train_loss": 0.03369076922535896, "global_step": 188418, "epoch": 2117, "lr": 2.1435606901733906e-05}
{"train_loss": 0.08757235109806061, "global_step": 188419, "epoch": 2117, "lr": 2.1435131084972294e-05}
{"train_loss": 0.03631912171840668, "global_step": 188420, "epoch": 2117, "lr": 2.1434655272050896e-05}
{"train_loss": 0.0675622820854187, "global_step": 188421, "epoch": 2117, "lr": 2.1434179462969745e-05}
{"train_loss": 0.04756883531808853, "global_step": 188422, "epoch": 2117, "lr": 2.1433703657728937e-05}
{"train_loss": 0.08008890599012375, "global_step": 188423, "epoch": 2117, "lr": 2.1433227856328504e-05}
{"train_loss": 0.056051310151815414, "global_step": 188424, "epoch": 2117, "lr": 2.143275205876854e-05}
{"train_loss": 0.053621627390384674, "global_step": 188425, "epoch": 2117, "lr": 2.1432276265049077e-05}
{"train_loss": 0.07318543642759323, "global_step": 188426, "epoch": 2117, "lr": 2.14318004751702e-05}
{"train_loss": 0.06363166868686676, "global_step": 188427, "epoch": 2117, "lr": 2.143132468913198e-05}
{"train_loss": 0.06497325003147125, "global_step": 188428, "epoch": 2117, "lr": 2.143084890693446e-05}
{"train_loss": 0.053893834352493286, "global_step": 188429, "epoch": 2117, "lr": 2.143037312857773e-05}
{"train_loss": 0.055539317429065704, "global_step": 188430, "epoch": 2117, "lr": 2.1429897354061816e-05}
{"train_loss": 0.09380055963993073, "global_step": 188431, "epoch": 2117, "lr": 2.1429421583386827e-05}
{"train_loss": 0.04054398462176323, "global_step": 188432, "epoch": 2117, "lr": 2.1428945816552782e-05}
{"train_loss": 0.06829796731472015, "global_step": 188433, "epoch": 2117, "lr": 2.1428470053559792e-05}
{"train_loss": 0.06265042722225189, "global_step": 188434, "epoch": 2117, "lr": 2.1427994294407865e-05}
{"train_loss": 0.06941960752010345, "global_step": 188435, "epoch": 2117, "lr": 2.1427518539097125e-05}
{"train_loss": 0.08452676981687546, "global_step": 188436, "epoch": 2117, "lr": 2.1427042787627587e-05}
{"train_loss": 0.02974880114197731, "global_step": 188437, "epoch": 2117, "lr": 2.142656703999935e-05}
{"train_loss": 0.07477051764726639, "global_step": 188438, "epoch": 2117, "lr": 2.1426091296212446e-05}
{"train_loss": 0.06319503486156464, "global_step": 188439, "epoch": 2117, "lr": 2.1425615556266975e-05}
{"train_loss": 0.035056039690971375, "global_step": 188440, "epoch": 2117, "lr": 2.142513982016296e-05}
{"train_loss": 0.07360212504863739, "global_step": 188441, "epoch": 2117, "lr": 2.142466408790051e-05}
{"train_loss": 0.1320132613182068, "global_step": 188442, "epoch": 2117, "lr": 2.1424188359479646e-05}
{"train_loss": 0.09220297634601593, "global_step": 188443, "epoch": 2117, "lr": 2.1423712634900467e-05}
{"train_loss": 0.05372947081923485, "global_step": 188444, "epoch": 2117, "lr": 2.1423236914163002e-05}
{"train_loss": 0.0687650814652443, "global_step": 188445, "epoch": 2117, "lr": 2.1422761197267337e-05}
{"train_loss": 0.024928776547312737, "global_step": 188446, "epoch": 2117, "lr": 2.1422285484213556e-05}
{"train_loss": 0.06859489530324936, "global_step": 188447, "epoch": 2117, "lr": 2.142180977500168e-05}
{"train_loss": 0.04721369594335556, "global_step": 188448, "epoch": 2117, "lr": 2.142133406963181e-05}
{"train_loss": 0.07535642385482788, "global_step": 188449, "epoch": 2117, "lr": 2.142085836810399e-05}
{"train_loss": 0.03205990791320801, "global_step": 188450, "epoch": 2117, "lr": 2.142038267041827e-05}
{"train_loss": 0.055797792971134186, "global_step": 188451, "epoch": 2117, "lr": 2.1419906976574755e-05}
{"train_loss": 0.12599992752075195, "global_step": 188452, "epoch": 2117, "lr": 2.1419431286573467e-05}
{"train_loss": 0.09121786057949066, "global_step": 188453, "epoch": 2117, "lr": 2.1418955600414497e-05}
{"train_loss": 0.047298818826675415, "global_step": 188454, "epoch": 2117, "lr": 2.141847991809791e-05}
{"train_loss": 0.1105775386095047, "global_step": 188455, "epoch": 2117, "lr": 2.1418004239623735e-05}
{"train_loss": 0.05440346524119377, "global_step": 188456, "epoch": 2117, "lr": 2.1417528564992083e-05}
{"train_loss": 0.053035836666822433, "global_step": 188457, "epoch": 2117, "lr": 2.1417052894202976e-05}
{"train_loss": 0.06863671541213989, "global_step": 188458, "epoch": 2117, "lr": 2.141657722725652e-05}
{"train_loss": 0.06676169484853745, "global_step": 188459, "epoch": 2117, "lr": 2.1416101564152734e-05}
{"train_loss": 0.06737347692251205, "global_step": 188460, "epoch": 2117, "lr": 2.1415625904891722e-05}
{"train_loss": 0.02952607162296772, "global_step": 188461, "epoch": 2117, "lr": 2.1415150249473515e-05}
{"train_loss": 0.0776059553027153, "global_step": 188462, "epoch": 2117, "lr": 2.141467459789821e-05}
{"train_loss": 0.08906827121973038, "global_step": 188463, "epoch": 2117, "lr": 2.1414198950165836e-05}
{"train_loss": 0.06323710083961487, "global_step": 188464, "epoch": 2117, "lr": 2.141372330627648e-05}
{"train_loss": 0.04624936729669571, "global_step": 188465, "epoch": 2117, "lr": 2.1413247666230213e-05}
{"train_loss": 0.08719216287136078, "global_step": 188466, "epoch": 2117, "lr": 2.1412772030027067e-05}
{"train_loss": 0.08556398749351501, "global_step": 188467, "epoch": 2117, "lr": 2.1412296397667143e-05}
{"train_loss": 0.07866378128528595, "global_step": 188468, "epoch": 2117, "lr": 2.141182076915047e-05}
{"train_loss": 0.05353856831789017, "global_step": 188469, "epoch": 2117, "lr": 2.1411345144477152e-05}
{"train_loss": 0.07005812972784042, "global_step": 188470, "epoch": 2117, "lr": 2.141086952364721e-05}
{"train_loss": 0.039657268673181534, "global_step": 188471, "epoch": 2117, "lr": 2.141039390666074e-05}
{"train_loss": 0.04999789595603943, "global_step": 188472, "epoch": 2117, "lr": 2.1409918293517783e-05}
{"train_loss": 0.08249486237764359, "global_step": 188473, "epoch": 2117, "lr": 2.140944268421843e-05}
{"train_loss": 0.05716068297624588, "global_step": 188474, "epoch": 2117, "lr": 2.1408967078762708e-05}
{"train_loss": 0.07968799769878387, "global_step": 188475, "epoch": 2117, "lr": 2.1408491477150727e-05}
{"train_loss": 0.06665665656328201, "global_step": 188476, "epoch": 2117, "lr": 2.1408015879382504e-05}
{"train_loss": 0.07387854158878326, "global_step": 188477, "epoch": 2117, "lr": 2.1407540285458144e-05}
{"train_loss": 0.08601129055023193, "global_step": 188478, "epoch": 2117, "lr": 2.140706469537767e-05}
{"train_loss": 0.05539456754922867, "global_step": 188479, "epoch": 2117, "lr": 2.140658910914119e-05}
{"train_loss": 0.09881601482629776, "global_step": 188480, "epoch": 2117, "lr": 2.1406113526748723e-05}
{"train_loss": 0.05531713366508484, "global_step": 188481, "epoch": 2117, "lr": 2.140563794820036e-05}
{"train_loss": 0.11357630789279938, "global_step": 188482, "epoch": 2117, "lr": 2.1405162373496175e-05}
{"train_loss": 0.0663987472653389, "global_step": 188483, "epoch": 2117, "lr": 2.1404686802636197e-05}
{"train_loss": 0.05841986835002899, "global_step": 188484, "epoch": 2117, "lr": 2.1404211235620535e-05}
{"train_loss": 0.09269550442695618, "global_step": 188485, "epoch": 2117, "lr": 2.140373567244921e-05}
{"train_loss": 0.05069669336080551, "global_step": 188486, "epoch": 2117, "lr": 2.140326011312232e-05}
{"train_loss": 0.08564883470535278, "global_step": 188487, "epoch": 2117, "lr": 2.140278455763991e-05}
{"train_loss": 0.1155601218342781, "global_step": 188488, "epoch": 2117, "lr": 2.140230900600203e-05}
{"train_loss": 0.05886295810341835, "global_step": 188489, "epoch": 2117, "lr": 2.140183345820878e-05}
{"train_loss": 0.05397530272603035, "global_step": 188490, "epoch": 2117, "lr": 2.140135791426019e-05}
{"train_loss": 0.051446497440338135, "global_step": 188491, "epoch": 2117, "lr": 2.1400882374156355e-05}
{"train_loss": 0.09270770102739334, "global_step": 188492, "epoch": 2117, "lr": 2.1400406837897323e-05}
{"train_loss": 0.05136485770344734, "global_step": 188493, "epoch": 2117, "lr": 2.139993130548314e-05}
{"train_loss": 0.07540109008550644, "global_step": 188494, "epoch": 2117, "lr": 2.13994557769139e-05}
{"train_loss": 0.05823010951280594, "global_step": 188495, "epoch": 2117, "lr": 2.1398980252189642e-05}
{"train_loss": 0.08391419053077698, "global_step": 188496, "epoch": 2117, "lr": 2.1398504731310465e-05}
{"train_loss": 0.0350894033908844, "global_step": 188497, "epoch": 2117, "lr": 2.139802921427638e-05}
{"train_loss": 0.019951000809669495, "global_step": 188498, "epoch": 2117, "lr": 2.1397553701087515e-05}
{"train_loss": 0.04107864201068878, "global_step": 188499, "epoch": 2117, "lr": 2.139707819174387e-05}
{"train_loss": 0.04578334093093872, "global_step": 188500, "epoch": 2117, "lr": 2.139660268624554e-05}
{"train_loss": 0.06665154872985368, "global_step": 188501, "epoch": 2117, "lr": 2.1396127184592613e-05, "val_loss": 7.740738391876221}
{"train_loss": 0.08005636930465698, "global_step": 188502, "epoch": 2118, "lr": 2.1395651686785107e-05}
{"train_loss": 0.0647251158952713, "global_step": 188503, "epoch": 2118, "lr": 2.1395176192823123e-05}
{"train_loss": 0.05822954326868057, "global_step": 188504, "epoch": 2118, "lr": 2.1394700702706695e-05}
{"train_loss": 0.026576941832900047, "global_step": 188505, "epoch": 2118, "lr": 2.1394225216435916e-05}
{"train_loss": 0.10132241249084473, "global_step": 188506, "epoch": 2118, "lr": 2.1393749734010816e-05}
{"train_loss": 0.05524076521396637, "global_step": 188507, "epoch": 2118, "lr": 2.1393274255431496e-05}
{"train_loss": 0.06616691499948502, "global_step": 188508, "epoch": 2118, "lr": 2.1392798780697983e-05}
{"train_loss": 0.12416331470012665, "global_step": 188509, "epoch": 2118, "lr": 2.139232330981038e-05}
{"train_loss": 0.07904060184955597, "global_step": 188510, "epoch": 2118, "lr": 2.1391847842768708e-05}
{"train_loss": 0.07404239475727081, "global_step": 188511, "epoch": 2118, "lr": 2.139137237957308e-05}
{"train_loss": 0.06566428393125534, "global_step": 188512, "epoch": 2118, "lr": 2.1390896920223507e-05}
{"train_loss": 0.09562196582555771, "global_step": 188513, "epoch": 2118, "lr": 2.1390421464720102e-05}
{"train_loss": 0.045524291694164276, "global_step": 188514, "epoch": 2118, "lr": 2.1389946013062884e-05}
{"train_loss": 0.07982321083545685, "global_step": 188515, "epoch": 2118, "lr": 2.138947056525196e-05}
{"train_loss": 0.10113541036844254, "global_step": 188516, "epoch": 2118, "lr": 2.1388995121287353e-05}
{"train_loss": 0.08435630053281784, "global_step": 188517, "epoch": 2118, "lr": 2.138851968116915e-05}
{"train_loss": 0.0705893263220787, "global_step": 188518, "epoch": 2118, "lr": 2.1388044244897425e-05}
{"train_loss": 0.0446481816470623, "global_step": 188519, "epoch": 2118, "lr": 2.1387568812472215e-05}
{"train_loss": 0.07144864648580551, "global_step": 188520, "epoch": 2118, "lr": 2.1387093383893616e-05}
{"train_loss": 0.02184588648378849, "global_step": 188521, "epoch": 2118, "lr": 2.1386617959161654e-05}
{"train_loss": 0.08340180665254593, "global_step": 188522, "epoch": 2118, "lr": 2.1386142538276428e-05}
{"train_loss": 0.10082768648862839, "global_step": 188523, "epoch": 2118, "lr": 2.1385667121237974e-05}
{"train_loss": 0.03540549799799919, "global_step": 188524, "epoch": 2118, "lr": 2.1385191708046382e-05}
{"train_loss": 0.10587064921855927, "global_step": 188525, "epoch": 2118, "lr": 2.1384716298701686e-05}
{"train_loss": 0.09259548038244247, "global_step": 188526, "epoch": 2118, "lr": 2.1384240893203982e-05}
{"train_loss": 0.09716949611902237, "global_step": 188527, "epoch": 2118, "lr": 2.138376549155332e-05}
{"train_loss": 0.048160962760448456, "global_step": 188528, "epoch": 2118, "lr": 2.138329009374974e-05}
{"train_loss": 0.06726789474487305, "global_step": 188529, "epoch": 2118, "lr": 2.1382814699793352e-05}
{"train_loss": 0.042240604758262634, "global_step": 188530, "epoch": 2118, "lr": 2.1382339309684174e-05}
{"train_loss": 0.03155219927430153, "global_step": 188531, "epoch": 2118, "lr": 2.1381863923422314e-05}
{"train_loss": 0.059976402670145035, "global_step": 188532, "epoch": 2118, "lr": 2.1381388541007806e-05}
{"train_loss": 0.042287591844797134, "global_step": 188533, "epoch": 2118, "lr": 2.138091316244071e-05}
{"train_loss": 0.0909615010023117, "global_step": 188534, "epoch": 2118, "lr": 2.1380437787721112e-05}
{"train_loss": 0.062017008662223816, "global_step": 188535, "epoch": 2118, "lr": 2.1379962416849052e-05}
{"train_loss": 0.06176518276333809, "global_step": 188536, "epoch": 2118, "lr": 2.1379487049824608e-05}
{"train_loss": 0.1116599515080452, "global_step": 188537, "epoch": 2118, "lr": 2.1379011686647865e-05}
{"train_loss": 0.07343901693820953, "global_step": 188538, "epoch": 2118, "lr": 2.1378536327318844e-05}
{"train_loss": 0.09237389266490936, "global_step": 188539, "epoch": 2118, "lr": 2.1378060971837643e-05}
{"train_loss": 0.041518036276102066, "global_step": 188540, "epoch": 2118, "lr": 2.13775856202043e-05}
{"train_loss": 0.0376618430018425, "global_step": 188541, "epoch": 2118, "lr": 2.137711027241891e-05}
{"train_loss": 0.0800887942314148, "global_step": 188542, "epoch": 2118, "lr": 2.13766349284815e-05}
{"train_loss": 0.0514460951089859, "global_step": 188543, "epoch": 2118, "lr": 2.1376159588392163e-05}
{"train_loss": 0.11851730197668076, "global_step": 188544, "epoch": 2118, "lr": 2.1375684252150947e-05}
{"train_loss": 0.06164056435227394, "global_step": 188545, "epoch": 2118, "lr": 2.1375208919757932e-05}
{"train_loss": 0.09281136840581894, "global_step": 188546, "epoch": 2118, "lr": 2.137473359121315e-05}
{"train_loss": 0.0751878172159195, "global_step": 188547, "epoch": 2118, "lr": 2.1374258266516713e-05}
{"train_loss": 0.07714524120092392, "global_step": 188548, "epoch": 2118, "lr": 2.1373782945668636e-05}
{"train_loss": 0.07677874714136124, "global_step": 188549, "epoch": 2118, "lr": 2.137330762866902e-05}
{"train_loss": 0.09217830747365952, "global_step": 188550, "epoch": 2118, "lr": 2.13728323155179e-05}
{"train_loss": 0.06993695348501205, "global_step": 188551, "epoch": 2118, "lr": 2.137235700621537e-05}
{"train_loss": 0.048238713294267654, "global_step": 188552, "epoch": 2118, "lr": 2.137188170076146e-05}
{"train_loss": 0.05074091628193855, "global_step": 188553, "epoch": 2118, "lr": 2.1371406399156253e-05}
{"train_loss": 0.07089769095182419, "global_step": 188554, "epoch": 2118, "lr": 2.1370931101399826e-05}
{"train_loss": 0.09758463501930237, "global_step": 188555, "epoch": 2118, "lr": 2.1370455807492218e-05}
{"train_loss": 0.03740182891488075, "global_step": 188556, "epoch": 2118, "lr": 2.1369980517433514e-05}
{"train_loss": 0.08237955719232559, "global_step": 188557, "epoch": 2118, "lr": 2.136950523122375e-05}
{"train_loss": 0.06206516548991203, "global_step": 188558, "epoch": 2118, "lr": 2.136902994886303e-05}
{"train_loss": 0.09696384519338608, "global_step": 188559, "epoch": 2118, "lr": 2.136855467035137e-05}
{"train_loss": 0.10176440328359604, "global_step": 188560, "epoch": 2118, "lr": 2.136807939568888e-05}
{"train_loss": 0.060775674879550934, "global_step": 188561, "epoch": 2118, "lr": 2.1367604124875583e-05}
{"train_loss": 0.06909175217151642, "global_step": 188562, "epoch": 2118, "lr": 2.1367128857911585e-05}
{"train_loss": 0.16264058649539948, "global_step": 188563, "epoch": 2118, "lr": 2.1366653594796905e-05}
{"train_loss": 0.04174989461898804, "global_step": 188564, "epoch": 2118, "lr": 2.136617833553165e-05}
{"train_loss": 0.07165096700191498, "global_step": 188565, "epoch": 2118, "lr": 2.1365703080115862e-05}
{"train_loss": 0.1686687022447586, "global_step": 188566, "epoch": 2118, "lr": 2.1365227828549588e-05}
{"train_loss": 0.06765686720609665, "global_step": 188567, "epoch": 2118, "lr": 2.136475258083293e-05}
{"train_loss": 0.12414669990539551, "global_step": 188568, "epoch": 2118, "lr": 2.1364277336965917e-05}
{"train_loss": 0.0934847742319107, "global_step": 188569, "epoch": 2118, "lr": 2.136380209694864e-05}
{"train_loss": 0.052606720477342606, "global_step": 188570, "epoch": 2118, "lr": 2.1363326860781153e-05}
{"train_loss": 0.04280118644237518, "global_step": 188571, "epoch": 2118, "lr": 2.13628516284635e-05}
{"train_loss": 0.04777968302369118, "global_step": 188572, "epoch": 2118, "lr": 2.136237639999576e-05}
{"train_loss": 0.11159830540418625, "global_step": 188573, "epoch": 2118, "lr": 2.1361901175378018e-05}
{"train_loss": 0.04167471081018448, "global_step": 188574, "epoch": 2118, "lr": 2.1361425954610303e-05}
{"train_loss": 0.044146228581666946, "global_step": 188575, "epoch": 2118, "lr": 2.1360950737692714e-05}
{"train_loss": 0.04430025815963745, "global_step": 188576, "epoch": 2118, "lr": 2.136047552462528e-05}
{"train_loss": 0.08165700733661652, "global_step": 188577, "epoch": 2118, "lr": 2.1360000315408092e-05}
{"train_loss": 0.08226024359464645, "global_step": 188578, "epoch": 2118, "lr": 2.135952511004119e-05}
{"train_loss": 0.1039973646402359, "global_step": 188579, "epoch": 2118, "lr": 2.1359049908524663e-05}
{"train_loss": 0.08715437352657318, "global_step": 188580, "epoch": 2118, "lr": 2.135857471085855e-05}
{"train_loss": 0.05407886579632759, "global_step": 188581, "epoch": 2118, "lr": 2.1358099517042947e-05}
{"train_loss": 0.04217948764562607, "global_step": 188582, "epoch": 2118, "lr": 2.135762432707788e-05}
{"train_loss": 0.06227337568998337, "global_step": 188583, "epoch": 2118, "lr": 2.1357149140963447e-05}
{"train_loss": 0.03019510954618454, "global_step": 188584, "epoch": 2118, "lr": 2.135667395869968e-05}
{"train_loss": 0.04229864478111267, "global_step": 188585, "epoch": 2118, "lr": 2.1356198780286675e-05}
{"train_loss": 0.08201583474874496, "global_step": 188586, "epoch": 2118, "lr": 2.1355723605724464e-05}
{"train_loss": 0.08974923193454742, "global_step": 188587, "epoch": 2118, "lr": 2.1355248435013143e-05}
{"train_loss": 0.08901823312044144, "global_step": 188588, "epoch": 2118, "lr": 2.135477326815274e-05}
{"train_loss": 0.07532049715518951, "global_step": 188589, "epoch": 2118, "lr": 2.135429810514336e-05}
{"train_loss": 0.07298519880918974, "global_step": 188590, "epoch": 2118, "lr": 2.1353822945985026e-05, "val_loss": 7.796994209289551}
{"train_loss": 0.12167201936244965, "global_step": 188591, "epoch": 2119, "lr": 2.1353347790677826e-05}
{"train_loss": 0.10424631834030151, "global_step": 188592, "epoch": 2119, "lr": 2.1352872639221827e-05}
{"train_loss": 0.05928162485361099, "global_step": 188593, "epoch": 2119, "lr": 2.1352397491617076e-05}
{"train_loss": 0.1369158774614334, "global_step": 188594, "epoch": 2119, "lr": 2.1351922347863663e-05}
{"train_loss": 0.08008425682783127, "global_step": 188595, "epoch": 2119, "lr": 2.135144720796161e-05}
{"train_loss": 0.06558196991682053, "global_step": 188596, "epoch": 2119, "lr": 2.135097207191103e-05}
{"train_loss": 0.059366293251514435, "global_step": 188597, "epoch": 2119, "lr": 2.1350496939711946e-05}
{"train_loss": 0.09214094281196594, "global_step": 188598, "epoch": 2119, "lr": 2.1350021811364452e-05}
{"train_loss": 0.08165572583675385, "global_step": 188599, "epoch": 2119, "lr": 2.1349546686868584e-05}
{"train_loss": 0.06077485904097557, "global_step": 188600, "epoch": 2119, "lr": 2.1349071566224438e-05}
{"train_loss": 0.04500171169638634, "global_step": 188601, "epoch": 2119, "lr": 2.134859644943204e-05}
{"train_loss": 0.08844906836748123, "global_step": 188602, "epoch": 2119, "lr": 2.134812133649149e-05}
{"train_loss": 0.0824725478887558, "global_step": 188603, "epoch": 2119, "lr": 2.134764622740284e-05}
{"train_loss": 0.05660405382514, "global_step": 188604, "epoch": 2119, "lr": 2.134717112216613e-05}
{"train_loss": 0.0301323551684618, "global_step": 188605, "epoch": 2119, "lr": 2.1346696020781458e-05}
{"train_loss": 0.06211305782198906, "global_step": 188606, "epoch": 2119, "lr": 2.134622092324886e-05}
{"train_loss": 0.02709299325942993, "global_step": 188607, "epoch": 2119, "lr": 2.1345745829568437e-05}
{"train_loss": 0.07980277389287949, "global_step": 188608, "epoch": 2119, "lr": 2.1345270739740204e-05}
{"train_loss": 0.05867193266749382, "global_step": 188609, "epoch": 2119, "lr": 2.1344795653764266e-05}
{"train_loss": 0.08743879199028015, "global_step": 188610, "epoch": 2119, "lr": 2.134432057164066e-05}
{"train_loss": 0.09025005251169205, "global_step": 188611, "epoch": 2119, "lr": 2.1343845493369475e-05}
{"train_loss": 0.08318396657705307, "global_step": 188612, "epoch": 2119, "lr": 2.1343370418950737e-05}
{"train_loss": 0.05159442499279976, "global_step": 188613, "epoch": 2119, "lr": 2.134289534838456e-05}
{"train_loss": 0.04569078981876373, "global_step": 188614, "epoch": 2119, "lr": 2.134242028167096e-05}
{"train_loss": 0.06723426282405853, "global_step": 188615, "epoch": 2119, "lr": 2.134194521881004e-05}
{"train_loss": 0.11043702065944672, "global_step": 188616, "epoch": 2119, "lr": 2.134147015980183e-05}
{"train_loss": 0.05320937931537628, "global_step": 188617, "epoch": 2119, "lr": 2.134099510464643e-05}
{"train_loss": 0.06963581591844559, "global_step": 188618, "epoch": 2119, "lr": 2.134052005334386e-05}
{"train_loss": 0.0558832585811615, "global_step": 188619, "epoch": 2119, "lr": 2.1340045005894227e-05}
{"train_loss": 0.06134391203522682, "global_step": 188620, "epoch": 2119, "lr": 2.1339569962297563e-05}
{"train_loss": 0.07315108925104141, "global_step": 188621, "epoch": 2119, "lr": 2.1339094922553955e-05}
{"train_loss": 0.05305975675582886, "global_step": 188622, "epoch": 2119, "lr": 2.133861988666344e-05}
{"train_loss": 0.049316585063934326, "global_step": 188623, "epoch": 2119, "lr": 2.1338144854626125e-05}
{"train_loss": 0.023222895339131355, "global_step": 188624, "epoch": 2119, "lr": 2.133766982644202e-05}
{"train_loss": 0.09177722036838531, "global_step": 188625, "epoch": 2119, "lr": 2.133719480211124e-05}
{"train_loss": 0.0760711207985878, "global_step": 188626, "epoch": 2119, "lr": 2.13367197816338e-05}
{"train_loss": 0.04533830285072327, "global_step": 188627, "epoch": 2119, "lr": 2.1336244765009788e-05}
{"train_loss": 0.07504413276910782, "global_step": 188628, "epoch": 2119, "lr": 2.1335769752239293e-05}
{"train_loss": 0.04556095972657204, "global_step": 188629, "epoch": 2119, "lr": 2.133529474332233e-05}
{"train_loss": 0.03796740621328354, "global_step": 188630, "epoch": 2119, "lr": 2.1334819738259016e-05}
{"train_loss": 0.07931707799434662, "global_step": 188631, "epoch": 2119, "lr": 2.133434473704936e-05}
{"train_loss": 0.07257897406816483, "global_step": 188632, "epoch": 2119, "lr": 2.1333869739693467e-05}
{"train_loss": 0.09529775381088257, "global_step": 188633, "epoch": 2119, "lr": 2.1333394746191376e-05}
{"train_loss": 0.05491815507411957, "global_step": 188634, "epoch": 2119, "lr": 2.1332919756543173e-05}
{"train_loss": 0.0974796861410141, "global_step": 188635, "epoch": 2119, "lr": 2.1332444770748895e-05}
{"train_loss": 0.03146308660507202, "global_step": 188636, "epoch": 2119, "lr": 2.1331969788808638e-05}
{"train_loss": 0.10560790449380875, "global_step": 188637, "epoch": 2119, "lr": 2.1331494810722424e-05}
{"train_loss": 0.05285610258579254, "global_step": 188638, "epoch": 2119, "lr": 2.1331019836490367e-05}
{"train_loss": 0.12591445446014404, "global_step": 188639, "epoch": 2119, "lr": 2.1330544866112485e-05}
{"train_loss": 0.09936562925577164, "global_step": 188640, "epoch": 2119, "lr": 2.1330069899588878e-05}
{"train_loss": 0.06937094777822495, "global_step": 188641, "epoch": 2119, "lr": 2.1329594936919572e-05}
{"train_loss": 0.03755361959338188, "global_step": 188642, "epoch": 2119, "lr": 2.132911997810468e-05}
{"train_loss": 0.05610177665948868, "global_step": 188643, "epoch": 2119, "lr": 2.1328645023144234e-05}
{"train_loss": 0.114262156188488, "global_step": 188644, "epoch": 2119, "lr": 2.1328170072038277e-05}
{"train_loss": 0.07518884539604187, "global_step": 188645, "epoch": 2119, "lr": 2.132769512478693e-05}
{"train_loss": 0.05851057544350624, "global_step": 188646, "epoch": 2119, "lr": 2.1327220181390194e-05}
{"train_loss": 0.11975477635860443, "global_step": 188647, "epoch": 2119, "lr": 2.1326745241848188e-05}
{"train_loss": 0.050362780690193176, "global_step": 188648, "epoch": 2119, "lr": 2.1326270306160934e-05}
{"train_loss": 0.11382965743541718, "global_step": 188649, "epoch": 2119, "lr": 2.1325795374328533e-05}
{"train_loss": 0.13505274057388306, "global_step": 188650, "epoch": 2119, "lr": 2.1325320446351003e-05}
{"train_loss": 0.10275474190711975, "global_step": 188651, "epoch": 2119, "lr": 2.132484552222846e-05}
{"train_loss": 0.08012453466653824, "global_step": 188652, "epoch": 2119, "lr": 2.1324370601960925e-05}
{"train_loss": 0.10311019420623779, "global_step": 188653, "epoch": 2119, "lr": 2.1323895685548488e-05}
{"train_loss": 0.07383712381124496, "global_step": 188654, "epoch": 2119, "lr": 2.1323420772991193e-05}
{"train_loss": 0.08921466022729874, "global_step": 188655, "epoch": 2119, "lr": 2.1322945864289135e-05}
{"train_loss": 0.057743366807699203, "global_step": 188656, "epoch": 2119, "lr": 2.132247095944233e-05}
{"train_loss": 0.02348141372203827, "global_step": 188657, "epoch": 2119, "lr": 2.1321996058450893e-05}
{"train_loss": 0.05410691723227501, "global_step": 188658, "epoch": 2119, "lr": 2.132152116131485e-05}
{"train_loss": 0.0819629356265068, "global_step": 188659, "epoch": 2119, "lr": 2.132104626803429e-05}
{"train_loss": 0.041342075914144516, "global_step": 188660, "epoch": 2119, "lr": 2.132057137860925e-05}
{"train_loss": 0.08257424831390381, "global_step": 188661, "epoch": 2119, "lr": 2.1320096493039832e-05}
{"train_loss": 0.10233629494905472, "global_step": 188662, "epoch": 2119, "lr": 2.1319621611326052e-05}
{"train_loss": 0.08825189620256424, "global_step": 188663, "epoch": 2119, "lr": 2.1319146733468004e-05}
{"train_loss": 0.03797353431582451, "global_step": 188664, "epoch": 2119, "lr": 2.1318671859465765e-05}
{"train_loss": 0.058931902050971985, "global_step": 188665, "epoch": 2119, "lr": 2.1318196989319362e-05}
{"train_loss": 0.09297014027833939, "global_step": 188666, "epoch": 2119, "lr": 2.1317722123028894e-05}
{"train_loss": 0.07204370945692062, "global_step": 188667, "epoch": 2119, "lr": 2.131724726059439e-05}
{"train_loss": 0.05356880649924278, "global_step": 188668, "epoch": 2119, "lr": 2.1316772402015954e-05}
{"train_loss": 0.07822877913713455, "global_step": 188669, "epoch": 2119, "lr": 2.131629754729361e-05}
{"train_loss": 0.07110118120908737, "global_step": 188670, "epoch": 2119, "lr": 2.131582269642746e-05}
{"train_loss": 0.06604158133268356, "global_step": 188671, "epoch": 2119, "lr": 2.131534784941753e-05}
{"train_loss": 0.03969693183898926, "global_step": 188672, "epoch": 2119, "lr": 2.1314873006263915e-05}
{"train_loss": 0.03524838015437126, "global_step": 188673, "epoch": 2119, "lr": 2.131439816696665e-05}
{"train_loss": 0.07726646214723587, "global_step": 188674, "epoch": 2119, "lr": 2.131392333152583e-05}
{"train_loss": 0.07836907356977463, "global_step": 188675, "epoch": 2119, "lr": 2.131344849994149e-05}
{"train_loss": 0.025808122009038925, "global_step": 188676, "epoch": 2119, "lr": 2.1312973672213726e-05}
{"train_loss": 0.06341227144002914, "global_step": 188677, "epoch": 2119, "lr": 2.131249884834256e-05}
{"train_loss": 0.09462092071771622, "global_step": 188678, "epoch": 2119, "lr": 2.13120240283281e-05}
{"train_loss": 0.07201669758625245, "global_step": 188679, "epoch": 2119, "lr": 2.1311549212170367e-05, "val_loss": 7.758688926696777}
{"train_loss": 0.06206822767853737, "global_step": 188680, "epoch": 2120, "lr": 2.1311074399869467e-05}
{"train_loss": 0.07762832194566727, "global_step": 188681, "epoch": 2120, "lr": 2.1310599591425444e-05}
{"train_loss": 0.07664036750793457, "global_step": 188682, "epoch": 2120, "lr": 2.1310124786838348e-05}
{"train_loss": 0.049550846219062805, "global_step": 188683, "epoch": 2120, "lr": 2.1309649986108267e-05}
{"train_loss": 0.05633355677127838, "global_step": 188684, "epoch": 2120, "lr": 2.1309175189235235e-05}
{"train_loss": 0.0721404105424881, "global_step": 188685, "epoch": 2120, "lr": 2.130870039621936e-05}
{"train_loss": 0.09047674387693405, "global_step": 188686, "epoch": 2120, "lr": 2.1308225607060655e-05}
{"train_loss": 0.05955612659454346, "global_step": 188687, "epoch": 2120, "lr": 2.1307750821759232e-05}
{"train_loss": 0.04927471652626991, "global_step": 188688, "epoch": 2120, "lr": 2.1307276040315115e-05}
{"train_loss": 0.042433008551597595, "global_step": 188689, "epoch": 2120, "lr": 2.1306801262728397e-05}
{"train_loss": 0.04507996141910553, "global_step": 188690, "epoch": 2120, "lr": 2.1306326488999118e-05}
{"train_loss": 0.06598462909460068, "global_step": 188691, "epoch": 2120, "lr": 2.130585171912737e-05}
{"train_loss": 0.13765966892242432, "global_step": 188692, "epoch": 2120, "lr": 2.130537695311318e-05}
{"train_loss": 0.08417383581399918, "global_step": 188693, "epoch": 2120, "lr": 2.1304902190956655e-05}
{"train_loss": 0.14375437796115875, "global_step": 188694, "epoch": 2120, "lr": 2.1304427432657815e-05}
{"train_loss": 0.07780672609806061, "global_step": 188695, "epoch": 2120, "lr": 2.130395267821676e-05}
{"train_loss": 0.05580390989780426, "global_step": 188696, "epoch": 2120, "lr": 2.1303477927633523e-05}
{"train_loss": 0.0407254733145237, "global_step": 188697, "epoch": 2120, "lr": 2.1303003180908203e-05}
{"train_loss": 0.03341541439294815, "global_step": 188698, "epoch": 2120, "lr": 2.130252843804083e-05}
{"train_loss": 0.09179385751485825, "global_step": 188699, "epoch": 2120, "lr": 2.1302053699031475e-05}
{"train_loss": 0.08145369589328766, "global_step": 188700, "epoch": 2120, "lr": 2.130157896388023e-05}
{"train_loss": 0.07057639956474304, "global_step": 188701, "epoch": 2120, "lr": 2.1301104232587122e-05}
{"train_loss": 0.039722785353660583, "global_step": 188702, "epoch": 2120, "lr": 2.130062950515225e-05}
{"train_loss": 0.06253528594970703, "global_step": 188703, "epoch": 2120, "lr": 2.1300154781575638e-05}
{"train_loss": 0.0661226212978363, "global_step": 188704, "epoch": 2120, "lr": 2.1299680061857387e-05}
{"train_loss": 0.06917086988687515, "global_step": 188705, "epoch": 2120, "lr": 2.1299205345997524e-05}
{"train_loss": 0.07293443381786346, "global_step": 188706, "epoch": 2120, "lr": 2.1298730633996157e-05}
{"train_loss": 0.07657580822706223, "global_step": 188707, "epoch": 2120, "lr": 2.1298255925853306e-05}
{"train_loss": 0.06307508796453476, "global_step": 188708, "epoch": 2120, "lr": 2.1297781221569073e-05}
{"train_loss": 0.0236454326659441, "global_step": 188709, "epoch": 2120, "lr": 2.1297306521143485e-05}
{"train_loss": 0.062055956572294235, "global_step": 188710, "epoch": 2120, "lr": 2.129683182457664e-05}
{"train_loss": 0.06636498123407364, "global_step": 188711, "epoch": 2120, "lr": 2.1296357131868573e-05}
{"train_loss": 0.03562362492084503, "global_step": 188712, "epoch": 2120, "lr": 2.129588244301937e-05}
{"train_loss": 0.0762031078338623, "global_step": 188713, "epoch": 2120, "lr": 2.1295407758029075e-05}
{"train_loss": 0.08842874318361282, "global_step": 188714, "epoch": 2120, "lr": 2.1294933076897783e-05}
{"train_loss": 0.04791996255517006, "global_step": 188715, "epoch": 2120, "lr": 2.1294458399625512e-05}
{"train_loss": 0.10583841055631638, "global_step": 188716, "epoch": 2120, "lr": 2.1293983726212374e-05}
{"train_loss": 0.09812605381011963, "global_step": 188717, "epoch": 2120, "lr": 2.129350905665839e-05}
{"train_loss": 0.05601505562663078, "global_step": 188718, "epoch": 2120, "lr": 2.1293034390963663e-05}
{"train_loss": 0.06859621405601501, "global_step": 188719, "epoch": 2120, "lr": 2.1292559729128236e-05}
{"train_loss": 0.05504631996154785, "global_step": 188720, "epoch": 2120, "lr": 2.1292085071152158e-05}
{"train_loss": 0.07615236937999725, "global_step": 188721, "epoch": 2120, "lr": 2.129161041703553e-05}
{"train_loss": 0.10235139727592468, "global_step": 188722, "epoch": 2120, "lr": 2.129113576677837e-05}
{"train_loss": 0.06731373071670532, "global_step": 188723, "epoch": 2120, "lr": 2.129066112038079e-05}
{"train_loss": 0.05927155166864395, "global_step": 188724, "epoch": 2120, "lr": 2.1290186477842812e-05}
{"train_loss": 0.043670278042554855, "global_step": 188725, "epoch": 2120, "lr": 2.1289711839164534e-05}
{"train_loss": 0.06324408948421478, "global_step": 188726, "epoch": 2120, "lr": 2.128923720434599e-05}
{"train_loss": 0.07630876451730728, "global_step": 188727, "epoch": 2120, "lr": 2.1288762573387273e-05}
{"train_loss": 0.02752963826060295, "global_step": 188728, "epoch": 2120, "lr": 2.1288287946288416e-05}
{"train_loss": 0.07136911898851395, "global_step": 188729, "epoch": 2120, "lr": 2.1287813323049515e-05}
{"train_loss": 0.048015009611845016, "global_step": 188730, "epoch": 2120, "lr": 2.12873387036706e-05}
{"train_loss": 0.04686475545167923, "global_step": 188731, "epoch": 2120, "lr": 2.1286864088151775e-05}
{"train_loss": 0.03760797530412674, "global_step": 188732, "epoch": 2120, "lr": 2.1286389476493058e-05}
{"train_loss": 0.0735229104757309, "global_step": 188733, "epoch": 2120, "lr": 2.1285914868694555e-05}
{"train_loss": 0.07799455523490906, "global_step": 188734, "epoch": 2120, "lr": 2.128544026475629e-05}
{"train_loss": 0.07243508845567703, "global_step": 188735, "epoch": 2120, "lr": 2.1284965664678353e-05}
{"train_loss": 0.04028704762458801, "global_step": 188736, "epoch": 2120, "lr": 2.1284491068460816e-05}
{"train_loss": 0.049975086003541946, "global_step": 188737, "epoch": 2120, "lr": 2.1284016476103713e-05}
{"train_loss": 0.09911756962537766, "global_step": 188738, "epoch": 2120, "lr": 2.1283541887607146e-05}
{"train_loss": 0.17891252040863037, "global_step": 188739, "epoch": 2120, "lr": 2.128306730297113e-05}
{"train_loss": 0.06418050080537796, "global_step": 188740, "epoch": 2120, "lr": 2.128259272219578e-05}
{"train_loss": 0.07452016323804855, "global_step": 188741, "epoch": 2120, "lr": 2.1282118145281116e-05}
{"train_loss": 0.12065424025058746, "global_step": 188742, "epoch": 2120, "lr": 2.128164357222724e-05}
{"train_loss": 0.047114770859479904, "global_step": 188743, "epoch": 2120, "lr": 2.1281169003034178e-05}
{"train_loss": 0.06151844933629036, "global_step": 188744, "epoch": 2120, "lr": 2.128069443770203e-05}
{"train_loss": 0.0494726225733757, "global_step": 188745, "epoch": 2120, "lr": 2.128021987623083e-05}
{"train_loss": 0.05578669533133507, "global_step": 188746, "epoch": 2120, "lr": 2.1279745318620664e-05}
{"train_loss": 0.08598125725984573, "global_step": 188747, "epoch": 2120, "lr": 2.1279270764871574e-05}
{"train_loss": 0.0742383822798729, "global_step": 188748, "epoch": 2120, "lr": 2.127879621498365e-05}
{"train_loss": 0.08521047979593277, "global_step": 188749, "epoch": 2120, "lr": 2.1278321668956925e-05}
{"train_loss": 0.07810091227293015, "global_step": 188750, "epoch": 2120, "lr": 2.12778471267915e-05}
{"train_loss": 0.10121390223503113, "global_step": 188751, "epoch": 2120, "lr": 2.12773725884874e-05}
{"train_loss": 0.05704555660486221, "global_step": 188752, "epoch": 2120, "lr": 2.1276898054044718e-05}
{"train_loss": 0.08411248028278351, "global_step": 188753, "epoch": 2120, "lr": 2.12764235234635e-05}
{"train_loss": 0.0504167303442955, "global_step": 188754, "epoch": 2120, "lr": 2.1275948996743827e-05}
{"train_loss": 0.05464518070220947, "global_step": 188755, "epoch": 2120, "lr": 2.1275474473885737e-05}
{"train_loss": 0.06801942735910416, "global_step": 188756, "epoch": 2120, "lr": 2.127499995488933e-05}
{"train_loss": 0.10166238993406296, "global_step": 188757, "epoch": 2120, "lr": 2.1274525439754627e-05}
{"train_loss": 0.08411381393671036, "global_step": 188758, "epoch": 2120, "lr": 2.1274050928481738e-05}
{"train_loss": 0.08794587105512619, "global_step": 188759, "epoch": 2120, "lr": 2.1273576421070697e-05}
{"train_loss": 0.0749685987830162, "global_step": 188760, "epoch": 2120, "lr": 2.1273101917521553e-05}
{"train_loss": 0.052464548498392105, "global_step": 188761, "epoch": 2120, "lr": 2.127262741783441e-05}
{"train_loss": 0.0828704982995987, "global_step": 188762, "epoch": 2120, "lr": 2.1272152922009297e-05}
{"train_loss": 0.06660451740026474, "global_step": 188763, "epoch": 2120, "lr": 2.1271678430046316e-05}
{"train_loss": 0.057678889483213425, "global_step": 188764, "epoch": 2120, "lr": 2.1271203941945485e-05}
{"train_loss": 0.09176081418991089, "global_step": 188765, "epoch": 2120, "lr": 2.1270729457706907e-05}
{"train_loss": 0.10295990109443665, "global_step": 188766, "epoch": 2120, "lr": 2.1270254977330605e-05}
{"train_loss": 0.08536283671855927, "global_step": 188767, "epoch": 2120, "lr": 2.1269780500816693e-05}
{"train_loss": 0.07002977428309033, "global_step": 188768, "epoch": 2120, "lr": 2.1269306028165194e-05, "val_loss": 7.733412742614746, "train_action_mse_error": 27.80117416381836}
{"train_loss": 0.12114217877388, "global_step": 188769, "epoch": 2121, "lr": 2.1268831559376195e-05}
{"train_loss": 0.08023130893707275, "global_step": 188770, "epoch": 2121, "lr": 2.1268357094449744e-05}
{"train_loss": 0.07218626886606216, "global_step": 188771, "epoch": 2121, "lr": 2.1267882633385906e-05}
{"train_loss": 0.06430956721305847, "global_step": 188772, "epoch": 2121, "lr": 2.1267408176184765e-05}
{"train_loss": 0.11765975505113602, "global_step": 188773, "epoch": 2121, "lr": 2.1266933722846362e-05}
{"train_loss": 0.040558941662311554, "global_step": 188774, "epoch": 2121, "lr": 2.126645927337078e-05}
{"train_loss": 0.1087741106748581, "global_step": 188775, "epoch": 2121, "lr": 2.1265984827758058e-05}
{"train_loss": 0.06867679953575134, "global_step": 188776, "epoch": 2121, "lr": 2.1265510386008292e-05}
{"train_loss": 0.06465159356594086, "global_step": 188777, "epoch": 2121, "lr": 2.1265035948121508e-05}
{"train_loss": 0.06463257223367691, "global_step": 188778, "epoch": 2121, "lr": 2.1264561514097807e-05}
{"train_loss": 0.05192504823207855, "global_step": 188779, "epoch": 2121, "lr": 2.1264087083937224e-05}
{"train_loss": 0.07016487419605255, "global_step": 188780, "epoch": 2121, "lr": 2.1263612657639847e-05}
{"train_loss": 0.061355262994766235, "global_step": 188781, "epoch": 2121, "lr": 2.126313823520571e-05}
{"train_loss": 0.043406110256910324, "global_step": 188782, "epoch": 2121, "lr": 2.1262663816634908e-05}
{"train_loss": 0.12692046165466309, "global_step": 188783, "epoch": 2121, "lr": 2.126218940192748e-05}
{"train_loss": 0.03142847865819931, "global_step": 188784, "epoch": 2121, "lr": 2.1261714991083516e-05}
{"train_loss": 0.09849509596824646, "global_step": 188785, "epoch": 2121, "lr": 2.1261240584103043e-05}
{"train_loss": 0.07771965116262436, "global_step": 188786, "epoch": 2121, "lr": 2.1260766180986165e-05}
{"train_loss": 0.11301583796739578, "global_step": 188787, "epoch": 2121, "lr": 2.126029178173291e-05}
{"train_loss": 0.11610393971204758, "global_step": 188788, "epoch": 2121, "lr": 2.1259817386343378e-05}
{"train_loss": 0.07033977657556534, "global_step": 188789, "epoch": 2121, "lr": 2.1259342994817587e-05}
{"train_loss": 0.10127884894609451, "global_step": 188790, "epoch": 2121, "lr": 2.1258868607155636e-05}
{"train_loss": 0.03550175204873085, "global_step": 188791, "epoch": 2121, "lr": 2.1258394223357597e-05}
{"train_loss": 0.07185013592243195, "global_step": 188792, "epoch": 2121, "lr": 2.1257919843423513e-05}
{"train_loss": 0.07720058411359787, "global_step": 188793, "epoch": 2121, "lr": 2.1257445467353438e-05}
{"train_loss": 0.08691247552633286, "global_step": 188794, "epoch": 2121, "lr": 2.125697109514746e-05}
{"train_loss": 0.07935944944620132, "global_step": 188795, "epoch": 2121, "lr": 2.125649672680562e-05}
{"train_loss": 0.1992722600698471, "global_step": 188796, "epoch": 2121, "lr": 2.125602236232801e-05}
{"train_loss": 0.09006395936012268, "global_step": 188797, "epoch": 2121, "lr": 2.1255548001714677e-05}
{"train_loss": 0.03643334284424782, "global_step": 188798, "epoch": 2121, "lr": 2.1255073644965667e-05}
{"train_loss": 0.06502911448478699, "global_step": 188799, "epoch": 2121, "lr": 2.125459929208108e-05}
{"train_loss": 0.08749014139175415, "global_step": 188800, "epoch": 2121, "lr": 2.1254124943060943e-05}
{"train_loss": 0.06644870340824127, "global_step": 188801, "epoch": 2121, "lr": 2.125365059790536e-05}
{"train_loss": 0.07466153800487518, "global_step": 188802, "epoch": 2121, "lr": 2.1253176256614353e-05}
{"train_loss": 0.06206206977367401, "global_step": 188803, "epoch": 2121, "lr": 2.125270191918802e-05}
{"train_loss": 0.03205893933773041, "global_step": 188804, "epoch": 2121, "lr": 2.12522275856264e-05}
{"train_loss": 0.03884558752179146, "global_step": 188805, "epoch": 2121, "lr": 2.125175325592958e-05}
{"train_loss": 0.11343877762556076, "global_step": 188806, "epoch": 2121, "lr": 2.1251278930097597e-05}
{"train_loss": 0.03772709146142006, "global_step": 188807, "epoch": 2121, "lr": 2.125080460813053e-05}
{"train_loss": 0.037226371467113495, "global_step": 188808, "epoch": 2121, "lr": 2.125033029002846e-05}
{"train_loss": 0.0472903698682785, "global_step": 188809, "epoch": 2121, "lr": 2.124985597579141e-05}
{"train_loss": 0.045320507138967514, "global_step": 188810, "epoch": 2121, "lr": 2.1249381665419486e-05}
{"train_loss": 0.09666202962398529, "global_step": 188811, "epoch": 2121, "lr": 2.1248907358912717e-05}
{"train_loss": 0.043725673109292984, "global_step": 188812, "epoch": 2121, "lr": 2.12484330562712e-05}
{"train_loss": 0.08438625186681747, "global_step": 188813, "epoch": 2121, "lr": 2.1247958757494957e-05}
{"train_loss": 0.05608120933175087, "global_step": 188814, "epoch": 2121, "lr": 2.12474844625841e-05}
{"train_loss": 0.05921453982591629, "global_step": 188815, "epoch": 2121, "lr": 2.1247010171538645e-05}
{"train_loss": 0.0858953595161438, "global_step": 188816, "epoch": 2121, "lr": 2.1246535884358704e-05}
{"train_loss": 0.11297294497489929, "global_step": 188817, "epoch": 2121, "lr": 2.1246061601044293e-05}
{"train_loss": 0.07295334339141846, "global_step": 188818, "epoch": 2121, "lr": 2.1245587321595516e-05}
{"train_loss": 0.04886305332183838, "global_step": 188819, "epoch": 2121, "lr": 2.1245113046012398e-05}
{"train_loss": 0.055012088268995285, "global_step": 188820, "epoch": 2121, "lr": 2.124463877429505e-05}
{"train_loss": 0.08229482918977737, "global_step": 188821, "epoch": 2121, "lr": 2.1244164506443487e-05}
{"train_loss": 0.04151046648621559, "global_step": 188822, "epoch": 2121, "lr": 2.1243690242457815e-05}
{"train_loss": 0.09889595210552216, "global_step": 188823, "epoch": 2121, "lr": 2.124321598233806e-05}
{"train_loss": 0.06059931218624115, "global_step": 188824, "epoch": 2121, "lr": 2.1242741726084325e-05}
{"train_loss": 0.03139427676796913, "global_step": 188825, "epoch": 2121, "lr": 2.124226747369663e-05}
{"train_loss": 0.11486638337373734, "global_step": 188826, "epoch": 2121, "lr": 2.1241793225175067e-05}
{"train_loss": 0.09606954455375671, "global_step": 188827, "epoch": 2121, "lr": 2.1241318980519707e-05}
{"train_loss": 0.062329113483428955, "global_step": 188828, "epoch": 2121, "lr": 2.1240844739730585e-05}
{"train_loss": 0.04834467172622681, "global_step": 188829, "epoch": 2121, "lr": 2.12403705028078e-05}
{"train_loss": 0.03718743100762367, "global_step": 188830, "epoch": 2121, "lr": 2.1239896269751397e-05}
{"train_loss": 0.1335577368736267, "global_step": 188831, "epoch": 2121, "lr": 2.123942204056142e-05}
{"train_loss": 0.06415043026208878, "global_step": 188832, "epoch": 2121, "lr": 2.123894781523797e-05}
{"train_loss": 0.05074451118707657, "global_step": 188833, "epoch": 2121, "lr": 2.1238473593781076e-05}
{"train_loss": 0.07161583006381989, "global_step": 188834, "epoch": 2121, "lr": 2.1237999376190838e-05}
{"train_loss": 0.01973136141896248, "global_step": 188835, "epoch": 2121, "lr": 2.12375251624673e-05}
{"train_loss": 0.07671325653791428, "global_step": 188836, "epoch": 2121, "lr": 2.1237050952610505e-05}
{"train_loss": 0.06381820142269135, "global_step": 188837, "epoch": 2121, "lr": 2.1236576746620556e-05}
{"train_loss": 0.07087008655071259, "global_step": 188838, "epoch": 2121, "lr": 2.123610254449748e-05}
{"train_loss": 0.025184692814946175, "global_step": 188839, "epoch": 2121, "lr": 2.123562834624138e-05}
{"train_loss": 0.07059027254581451, "global_step": 188840, "epoch": 2121, "lr": 2.1235154151852283e-05}
{"train_loss": 0.0578281432390213, "global_step": 188841, "epoch": 2121, "lr": 2.1234679961330283e-05}
{"train_loss": 0.04975057393312454, "global_step": 188842, "epoch": 2121, "lr": 2.123420577467542e-05}
{"train_loss": 0.06712822616100311, "global_step": 188843, "epoch": 2121, "lr": 2.123373159188778e-05}
{"train_loss": 0.03932061791419983, "global_step": 188844, "epoch": 2121, "lr": 2.1233257412967388e-05}
{"train_loss": 0.059204231947660446, "global_step": 188845, "epoch": 2121, "lr": 2.1232783237914344e-05}
{"train_loss": 0.05937298759818077, "global_step": 188846, "epoch": 2121, "lr": 2.1232309066728718e-05}
{"train_loss": 0.0492313988506794, "global_step": 188847, "epoch": 2121, "lr": 2.1231834899410536e-05}
{"train_loss": 0.051841091364622116, "global_step": 188848, "epoch": 2121, "lr": 2.1231360735959903e-05}
{"train_loss": 0.12310590595006943, "global_step": 188849, "epoch": 2121, "lr": 2.1230886576376846e-05}
{"train_loss": 0.05475308373570442, "global_step": 188850, "epoch": 2121, "lr": 2.1230412420661466e-05}
{"train_loss": 0.04297908768057823, "global_step": 188851, "epoch": 2121, "lr": 2.1229938268813782e-05}
{"train_loss": 0.05891573429107666, "global_step": 188852, "epoch": 2121, "lr": 2.1229464120833904e-05}
{"train_loss": 0.08179173618555069, "global_step": 188853, "epoch": 2121, "lr": 2.122898997672186e-05}
{"train_loss": 0.05676161125302315, "global_step": 188854, "epoch": 2121, "lr": 2.1228515836477737e-05}
{"train_loss": 0.06615433096885681, "global_step": 188855, "epoch": 2121, "lr": 2.1228041700101574e-05}
{"train_loss": 0.09375115483999252, "global_step": 188856, "epoch": 2121, "lr": 2.122756756759347e-05}
{"train_loss": 0.07085473293417625, "global_step": 188857, "epoch": 2121, "lr": 2.122709343895345e-05, "val_loss": 7.798086643218994}
{"train_loss": 0.06544230133295059, "global_step": 188858, "epoch": 2122, "lr": 2.1226619314181616e-05}
{"train_loss": 0.1220988854765892, "global_step": 188859, "epoch": 2122, "lr": 2.1226145193277997e-05}
{"train_loss": 0.05149996653199196, "global_step": 188860, "epoch": 2122, "lr": 2.1225671076242682e-05}
{"train_loss": 0.060154978185892105, "global_step": 188861, "epoch": 2122, "lr": 2.122519696307571e-05}
{"train_loss": 0.10636024177074432, "global_step": 188862, "epoch": 2122, "lr": 2.1224722853777163e-05}
{"train_loss": 0.04030875489115715, "global_step": 188863, "epoch": 2122, "lr": 2.122424874834712e-05}
{"train_loss": 0.08677195757627487, "global_step": 188864, "epoch": 2122, "lr": 2.1223774646785604e-05}
{"train_loss": 0.06689976900815964, "global_step": 188865, "epoch": 2122, "lr": 2.1223300549092717e-05}
{"train_loss": 0.10691236704587936, "global_step": 188866, "epoch": 2122, "lr": 2.122282645526849e-05}
{"train_loss": 0.06938409805297852, "global_step": 188867, "epoch": 2122, "lr": 2.1222352365313025e-05}
{"train_loss": 0.12083584815263748, "global_step": 188868, "epoch": 2122, "lr": 2.1221878279226343e-05}
{"train_loss": 0.08197120577096939, "global_step": 188869, "epoch": 2122, "lr": 2.1221404197008548e-05}
{"train_loss": 0.05930610001087189, "global_step": 188870, "epoch": 2122, "lr": 2.122093011865968e-05}
{"train_loss": 0.07832148671150208, "global_step": 188871, "epoch": 2122, "lr": 2.12204560441798e-05}
{"train_loss": 0.06946638971567154, "global_step": 188872, "epoch": 2122, "lr": 2.1219981973568985e-05}
{"train_loss": 0.08874277025461197, "global_step": 188873, "epoch": 2122, "lr": 2.1219507906827278e-05}
{"train_loss": 0.025963975116610527, "global_step": 188874, "epoch": 2122, "lr": 2.1219033843954777e-05}
{"train_loss": 0.10319395363330841, "global_step": 188875, "epoch": 2122, "lr": 2.121855978495153e-05}
{"train_loss": 0.12723982334136963, "global_step": 188876, "epoch": 2122, "lr": 2.1218085729817572e-05}
{"train_loss": 0.04009711742401123, "global_step": 188877, "epoch": 2122, "lr": 2.1217611678553014e-05}
{"train_loss": 0.11042638123035431, "global_step": 188878, "epoch": 2122, "lr": 2.1217137631157878e-05}
{"train_loss": 0.08643556386232376, "global_step": 188879, "epoch": 2122, "lr": 2.1216663587632263e-05}
{"train_loss": 0.06609831005334854, "global_step": 188880, "epoch": 2122, "lr": 2.12161895479762e-05}
{"train_loss": 0.05380044877529144, "global_step": 188881, "epoch": 2122, "lr": 2.121571551218977e-05}
{"train_loss": 0.034151919186115265, "global_step": 188882, "epoch": 2122, "lr": 2.1215241480273057e-05}
{"train_loss": 0.05424991250038147, "global_step": 188883, "epoch": 2122, "lr": 2.121476745222608e-05}
{"train_loss": 0.041326284408569336, "global_step": 188884, "epoch": 2122, "lr": 2.121429342804895e-05}
{"train_loss": 0.076179638504982, "global_step": 188885, "epoch": 2122, "lr": 2.1213819407741686e-05}
{"train_loss": 0.0560162253677845, "global_step": 188886, "epoch": 2122, "lr": 2.1213345391304396e-05}
{"train_loss": 0.07717594504356384, "global_step": 188887, "epoch": 2122, "lr": 2.1212871378737096e-05}
{"train_loss": 0.0909745991230011, "global_step": 188888, "epoch": 2122, "lr": 2.1212397370039894e-05}
{"train_loss": 0.10745026916265488, "global_step": 188889, "epoch": 2122, "lr": 2.121192336521282e-05}
{"train_loss": 0.12170356512069702, "global_step": 188890, "epoch": 2122, "lr": 2.121144936425597e-05}
{"train_loss": 0.03792208060622215, "global_step": 188891, "epoch": 2122, "lr": 2.1210975367169368e-05}
{"train_loss": 0.08942723274230957, "global_step": 188892, "epoch": 2122, "lr": 2.121050137395312e-05}
{"train_loss": 0.04465051740407944, "global_step": 188893, "epoch": 2122, "lr": 2.1210027384607252e-05}
{"train_loss": 0.08478806912899017, "global_step": 188894, "epoch": 2122, "lr": 2.120955339913186e-05}
{"train_loss": 0.10964812338352203, "global_step": 188895, "epoch": 2122, "lr": 2.1209079417526974e-05}
{"train_loss": 0.10428618639707565, "global_step": 188896, "epoch": 2122, "lr": 2.1208605439792695e-05}
{"train_loss": 0.08610042929649353, "global_step": 188897, "epoch": 2122, "lr": 2.1208131465929056e-05}
{"train_loss": 0.029908962547779083, "global_step": 188898, "epoch": 2122, "lr": 2.1207657495936123e-05}
{"train_loss": 0.08646320551633835, "global_step": 188899, "epoch": 2122, "lr": 2.1207183529813995e-05}
{"train_loss": 0.04707080125808716, "global_step": 188900, "epoch": 2122, "lr": 2.120670956756269e-05}
{"train_loss": 0.08544385433197021, "global_step": 188901, "epoch": 2122, "lr": 2.120623560918231e-05}
{"train_loss": 0.0715089812874794, "global_step": 188902, "epoch": 2122, "lr": 2.1205761654672886e-05}
{"train_loss": 0.12946951389312744, "global_step": 188903, "epoch": 2122, "lr": 2.120528770403451e-05}
{"train_loss": 0.07232510298490524, "global_step": 188904, "epoch": 2122, "lr": 2.120481375726721e-05}
{"train_loss": 0.06867608428001404, "global_step": 188905, "epoch": 2122, "lr": 2.12043398143711e-05}
{"train_loss": 0.0987793356180191, "global_step": 188906, "epoch": 2122, "lr": 2.1203865875346192e-05}
{"train_loss": 0.060446593910455704, "global_step": 188907, "epoch": 2122, "lr": 2.1203391940192592e-05}
{"train_loss": 0.11920124292373657, "global_step": 188908, "epoch": 2122, "lr": 2.1202918008910345e-05}
{"train_loss": 0.0990695133805275, "global_step": 188909, "epoch": 2122, "lr": 2.1202444081499497e-05}
{"train_loss": 0.06525322794914246, "global_step": 188910, "epoch": 2122, "lr": 2.1201970157960145e-05}
{"train_loss": 0.06928861141204834, "global_step": 188911, "epoch": 2122, "lr": 2.1201496238292322e-05}
{"train_loss": 0.03151701018214226, "global_step": 188912, "epoch": 2122, "lr": 2.1201022322496122e-05}
{"train_loss": 0.09828025102615356, "global_step": 188913, "epoch": 2122, "lr": 2.1200548410571597e-05}
{"train_loss": 0.07544070482254028, "global_step": 188914, "epoch": 2122, "lr": 2.1200074502518785e-05}
{"train_loss": 0.05841536447405815, "global_step": 188915, "epoch": 2122, "lr": 2.1199600598337792e-05}
{"train_loss": 0.05113270506262779, "global_step": 188916, "epoch": 2122, "lr": 2.119912669802865e-05}
{"train_loss": 0.07937072962522507, "global_step": 188917, "epoch": 2122, "lr": 2.1198652801591425e-05}
{"train_loss": 0.03674889728426933, "global_step": 188918, "epoch": 2122, "lr": 2.119817890902621e-05}
{"train_loss": 0.03916580229997635, "global_step": 188919, "epoch": 2122, "lr": 2.1197705020333035e-05}
{"train_loss": 0.06644151359796524, "global_step": 188920, "epoch": 2122, "lr": 2.1197231135511993e-05}
{"train_loss": 0.06903425604104996, "global_step": 188921, "epoch": 2122, "lr": 2.1196757254563117e-05}
{"train_loss": 0.09944287687540054, "global_step": 188922, "epoch": 2122, "lr": 2.1196283377486497e-05}
{"train_loss": 0.10466839373111725, "global_step": 188923, "epoch": 2122, "lr": 2.1195809504282172e-05}
{"train_loss": 0.10823629796504974, "global_step": 188924, "epoch": 2122, "lr": 2.1195335634950235e-05}
{"train_loss": 0.024196529760956764, "global_step": 188925, "epoch": 2122, "lr": 2.119486176949072e-05}
{"train_loss": 0.03524281084537506, "global_step": 188926, "epoch": 2122, "lr": 2.1194387907903722e-05}
{"train_loss": 0.06038206070661545, "global_step": 188927, "epoch": 2122, "lr": 2.119391405018927e-05}
{"train_loss": 0.06769838184118271, "global_step": 188928, "epoch": 2122, "lr": 2.1193440196347457e-05}
{"train_loss": 0.08182355016469955, "global_step": 188929, "epoch": 2122, "lr": 2.1192966346378324e-05}
{"train_loss": 0.09798797965049744, "global_step": 188930, "epoch": 2122, "lr": 2.1192492500281957e-05}
{"train_loss": 0.04327299818396568, "global_step": 188931, "epoch": 2122, "lr": 2.1192018658058398e-05}
{"train_loss": 0.09543078392744064, "global_step": 188932, "epoch": 2122, "lr": 2.119154481970773e-05}
{"train_loss": 0.0734088271856308, "global_step": 188933, "epoch": 2122, "lr": 2.119107098523e-05}
{"train_loss": 0.07358451187610626, "global_step": 188934, "epoch": 2122, "lr": 2.1190597154625273e-05}
{"train_loss": 0.0547039695084095, "global_step": 188935, "epoch": 2122, "lr": 2.1190123327893642e-05}
{"train_loss": 0.059863246977329254, "global_step": 188936, "epoch": 2122, "lr": 2.1189649505035126e-05}
{"train_loss": 0.07295774668455124, "global_step": 188937, "epoch": 2122, "lr": 2.118917568604982e-05}
{"train_loss": 0.0879194512963295, "global_step": 188938, "epoch": 2122, "lr": 2.118870187093777e-05}
{"train_loss": 0.031383175402879715, "global_step": 188939, "epoch": 2122, "lr": 2.1188228059699067e-05}
{"train_loss": 0.07574982941150665, "global_step": 188940, "epoch": 2122, "lr": 2.1187754252333736e-05}
{"train_loss": 0.057930923998355865, "global_step": 188941, "epoch": 2122, "lr": 2.118728044884188e-05}
{"train_loss": 0.07542214542627335, "global_step": 188942, "epoch": 2122, "lr": 2.1186806649223522e-05}
{"train_loss": 0.06392170488834381, "global_step": 188943, "epoch": 2122, "lr": 2.1186332853478765e-05}
{"train_loss": 0.08768545091152191, "global_step": 188944, "epoch": 2122, "lr": 2.118585906160763e-05}
{"train_loss": 0.07295418530702591, "global_step": 188945, "epoch": 2122, "lr": 2.1185385273610237e-05}
{"train_loss": 0.07411698607748814, "global_step": 188946, "epoch": 2122, "lr": 2.118491148948661e-05, "val_loss": 7.695166110992432}
{"train_loss": 0.05106059089303017, "global_step": 188947, "epoch": 2123, "lr": 2.11844377092368e-05}
{"train_loss": 0.07170181721448898, "global_step": 188948, "epoch": 2123, "lr": 2.1183963932860913e-05}
{"train_loss": 0.09467557072639465, "global_step": 188949, "epoch": 2123, "lr": 2.118349016035897e-05}
{"train_loss": 0.07847726345062256, "global_step": 188950, "epoch": 2123, "lr": 2.1183016391731075e-05}
{"train_loss": 0.025941401720046997, "global_step": 188951, "epoch": 2123, "lr": 2.1182542626977275e-05}
{"train_loss": 0.10407934337854385, "global_step": 188952, "epoch": 2123, "lr": 2.118206886609761e-05}
{"train_loss": 0.051023442298173904, "global_step": 188953, "epoch": 2123, "lr": 2.1181595109092162e-05}
{"train_loss": 0.0506022647023201, "global_step": 188954, "epoch": 2123, "lr": 2.118112135596102e-05}
{"train_loss": 0.11326508224010468, "global_step": 188955, "epoch": 2123, "lr": 2.1180647606704208e-05}
{"train_loss": 0.0729251354932785, "global_step": 188956, "epoch": 2123, "lr": 2.1180173861321817e-05}
{"train_loss": 0.06673223525285721, "global_step": 188957, "epoch": 2123, "lr": 2.1179700119813888e-05}
{"train_loss": 0.05694561079144478, "global_step": 188958, "epoch": 2123, "lr": 2.1179226382180517e-05}
{"train_loss": 0.11986597627401352, "global_step": 188959, "epoch": 2123, "lr": 2.1178752648421722e-05}
{"train_loss": 0.07157481461763382, "global_step": 188960, "epoch": 2123, "lr": 2.117827891853762e-05}
{"train_loss": 0.07219251990318298, "global_step": 188961, "epoch": 2123, "lr": 2.117780519252822e-05}
{"train_loss": 0.08369600027799606, "global_step": 188962, "epoch": 2123, "lr": 2.1177331470393635e-05}
{"train_loss": 0.09579092264175415, "global_step": 188963, "epoch": 2123, "lr": 2.1176857752133893e-05}
{"train_loss": 0.08066804707050323, "global_step": 188964, "epoch": 2123, "lr": 2.1176384037749086e-05}
{"train_loss": 0.10498782247304916, "global_step": 188965, "epoch": 2123, "lr": 2.117591032723924e-05}
{"train_loss": 0.0817485824227333, "global_step": 188966, "epoch": 2123, "lr": 2.1175436620604465e-05}
{"train_loss": 0.024050146341323853, "global_step": 188967, "epoch": 2123, "lr": 2.1174962917844782e-05}
{"train_loss": 0.08148764073848724, "global_step": 188968, "epoch": 2123, "lr": 2.1174489218960292e-05}
{"train_loss": 0.06834471225738525, "global_step": 188969, "epoch": 2123, "lr": 2.117401552395102e-05}
{"train_loss": 0.024069776758551598, "global_step": 188970, "epoch": 2123, "lr": 2.1173541832817073e-05}
{"train_loss": 0.061482496559619904, "global_step": 188971, "epoch": 2123, "lr": 2.1173068145558473e-05}
{"train_loss": 0.09178578853607178, "global_step": 188972, "epoch": 2123, "lr": 2.11725944621753e-05}
{"train_loss": 0.10789664834737778, "global_step": 188973, "epoch": 2123, "lr": 2.1172120782667637e-05}
{"train_loss": 0.11153284460306168, "global_step": 188974, "epoch": 2123, "lr": 2.1171647107035512e-05}
{"train_loss": 0.0747838243842125, "global_step": 188975, "epoch": 2123, "lr": 2.1171173435279035e-05}
{"train_loss": 0.10626029223203659, "global_step": 188976, "epoch": 2123, "lr": 2.1170699767398217e-05}
{"train_loss": 0.04825858771800995, "global_step": 188977, "epoch": 2123, "lr": 2.1170226103393165e-05}
{"train_loss": 0.06955292820930481, "global_step": 188978, "epoch": 2123, "lr": 2.1169752443263906e-05}
{"train_loss": 0.12408111244440079, "global_step": 188979, "epoch": 2123, "lr": 2.1169278787010543e-05}
{"train_loss": 0.030881205573678017, "global_step": 188980, "epoch": 2123, "lr": 2.1168805134633102e-05}
{"train_loss": 0.06169069558382034, "global_step": 188981, "epoch": 2123, "lr": 2.116833148613168e-05}
{"train_loss": 0.09211897850036621, "global_step": 188982, "epoch": 2123, "lr": 2.1167857841506305e-05}
{"train_loss": 0.0820678174495697, "global_step": 188983, "epoch": 2123, "lr": 2.1167384200757077e-05}
{"train_loss": 0.04022180289030075, "global_step": 188984, "epoch": 2123, "lr": 2.116691056388403e-05}
{"train_loss": 0.04864245280623436, "global_step": 188985, "epoch": 2123, "lr": 2.1166436930887257e-05}
{"train_loss": 0.06283067911863327, "global_step": 188986, "epoch": 2123, "lr": 2.1165963301766806e-05}
{"train_loss": 0.061433661729097366, "global_step": 188987, "epoch": 2123, "lr": 2.116548967652272e-05}
{"train_loss": 0.12125974893569946, "global_step": 188988, "epoch": 2123, "lr": 2.11650160551551e-05}
{"train_loss": 0.05291870981454849, "global_step": 188989, "epoch": 2123, "lr": 2.1164542437663976e-05}
{"train_loss": 0.05725677311420441, "global_step": 188990, "epoch": 2123, "lr": 2.1164068824049442e-05}
{"train_loss": 0.06335785984992981, "global_step": 188991, "epoch": 2123, "lr": 2.116359521431153e-05}
{"train_loss": 0.07504665851593018, "global_step": 188992, "epoch": 2123, "lr": 2.1163121608450344e-05}
{"train_loss": 0.07619065791368484, "global_step": 188993, "epoch": 2123, "lr": 2.11626480064659e-05}
{"train_loss": 0.0422021858394146, "global_step": 188994, "epoch": 2123, "lr": 2.116217440835831e-05}
{"train_loss": 0.03745066747069359, "global_step": 188995, "epoch": 2123, "lr": 2.1161700814127595e-05}
{"train_loss": 0.039225753396749496, "global_step": 188996, "epoch": 2123, "lr": 2.1161227223773854e-05}
{"train_loss": 0.08914738148450851, "global_step": 188997, "epoch": 2123, "lr": 2.116075363729712e-05}
{"train_loss": 0.09043397009372711, "global_step": 188998, "epoch": 2123, "lr": 2.1160280054697485e-05}
{"train_loss": 0.0381510891020298, "global_step": 188999, "epoch": 2123, "lr": 2.1159806475974986e-05}
{"train_loss": 0.033949293196201324, "global_step": 189000, "epoch": 2123, "lr": 2.1159332901129713e-05}
{"train_loss": 0.0903187170624733, "global_step": 189001, "epoch": 2123, "lr": 2.1158859330161702e-05}
{"train_loss": 0.06726458668708801, "global_step": 189002, "epoch": 2123, "lr": 2.1158385763071043e-05}
{"train_loss": 0.056604836136102676, "global_step": 189003, "epoch": 2123, "lr": 2.1157912199857772e-05}
{"train_loss": 0.04546579346060753, "global_step": 189004, "epoch": 2123, "lr": 2.1157438640521987e-05}
{"train_loss": 0.046765752136707306, "global_step": 189005, "epoch": 2123, "lr": 2.115696508506372e-05}
{"train_loss": 0.08709254115819931, "global_step": 189006, "epoch": 2123, "lr": 2.115649153348306e-05}
{"train_loss": 0.09506762772798538, "global_step": 189007, "epoch": 2123, "lr": 2.1156017985780036e-05}
{"train_loss": 0.0505794882774353, "global_step": 189008, "epoch": 2123, "lr": 2.115554444195474e-05}
{"train_loss": 0.06999456882476807, "global_step": 189009, "epoch": 2123, "lr": 2.1155070902007247e-05}
{"train_loss": 0.13170084357261658, "global_step": 189010, "epoch": 2123, "lr": 2.1154597365937584e-05}
{"train_loss": 0.07962910085916519, "global_step": 189011, "epoch": 2123, "lr": 2.1154123833745847e-05}
{"train_loss": 0.09469263255596161, "global_step": 189012, "epoch": 2123, "lr": 2.115365030543208e-05}
{"train_loss": 0.03185903653502464, "global_step": 189013, "epoch": 2123, "lr": 2.115317678099636e-05}
{"train_loss": 0.08901114016771317, "global_step": 189014, "epoch": 2123, "lr": 2.1152703260438735e-05}
{"train_loss": 0.07836207002401352, "global_step": 189015, "epoch": 2123, "lr": 2.115222974375929e-05}
{"train_loss": 0.028763547539711, "global_step": 189016, "epoch": 2123, "lr": 2.1151756230958054e-05}
{"train_loss": 0.07582572847604752, "global_step": 189017, "epoch": 2123, "lr": 2.1151282722035136e-05}
{"train_loss": 0.03974713012576103, "global_step": 189018, "epoch": 2123, "lr": 2.1150809216990558e-05}
{"train_loss": 0.02558555267751217, "global_step": 189019, "epoch": 2123, "lr": 2.1150335715824422e-05}
{"train_loss": 0.06388380378484726, "global_step": 189020, "epoch": 2123, "lr": 2.114986221853675e-05}
{"train_loss": 0.11030298471450806, "global_step": 189021, "epoch": 2123, "lr": 2.1149388725127645e-05}
{"train_loss": 0.054047368466854095, "global_step": 189022, "epoch": 2123, "lr": 2.114891523559714e-05}
{"train_loss": 0.07295534759759903, "global_step": 189023, "epoch": 2123, "lr": 2.1148441749945324e-05}
{"train_loss": 0.06386033445596695, "global_step": 189024, "epoch": 2123, "lr": 2.114796826817225e-05}
{"train_loss": 0.057130612432956696, "global_step": 189025, "epoch": 2123, "lr": 2.1147494790277966e-05}
{"train_loss": 0.07203701138496399, "global_step": 189026, "epoch": 2123, "lr": 2.1147021316262567e-05}
{"train_loss": 0.05295810103416443, "global_step": 189027, "epoch": 2123, "lr": 2.114654784612608e-05}
{"train_loss": 0.024255624040961266, "global_step": 189028, "epoch": 2123, "lr": 2.11460743798686e-05}
{"train_loss": 0.08751837909221649, "global_step": 189029, "epoch": 2123, "lr": 2.1145600917490167e-05}
{"train_loss": 0.05219554156064987, "global_step": 189030, "epoch": 2123, "lr": 2.1145127458990877e-05}
{"train_loss": 0.046790607273578644, "global_step": 189031, "epoch": 2123, "lr": 2.114465400437075e-05}
{"train_loss": 0.08110540360212326, "global_step": 189032, "epoch": 2123, "lr": 2.1144180553629888e-05}
{"train_loss": 0.059915754944086075, "global_step": 189033, "epoch": 2123, "lr": 2.1143707106768328e-05}
{"train_loss": 0.041418638080358505, "global_step": 189034, "epoch": 2123, "lr": 2.114323366378616e-05}
{"train_loss": 0.06863277221328756, "global_step": 189035, "epoch": 2123, "lr": 2.114276022468341e-05, "val_loss": 7.847527503967285}
{"train_loss": 0.025563061237335205, "global_step": 189036, "epoch": 2124, "lr": 2.1142286789460186e-05}
{"train_loss": 0.043431758880615234, "global_step": 189037, "epoch": 2124, "lr": 2.1141813358116515e-05}
{"train_loss": 0.07270677387714386, "global_step": 189038, "epoch": 2124, "lr": 2.1141339930652494e-05}
{"train_loss": 0.09775947779417038, "global_step": 189039, "epoch": 2124, "lr": 2.1140866507068142e-05}
{"train_loss": 0.033718694001436234, "global_step": 189040, "epoch": 2124, "lr": 2.1140393087363574e-05}
{"train_loss": 0.07180828601121902, "global_step": 189041, "epoch": 2124, "lr": 2.1139919671538806e-05}
{"train_loss": 0.12809284031391144, "global_step": 189042, "epoch": 2124, "lr": 2.1139446259593944e-05}
{"train_loss": 0.05224349722266197, "global_step": 189043, "epoch": 2124, "lr": 2.1138972851529014e-05}
{"train_loss": 0.0459650419652462, "global_step": 189044, "epoch": 2124, "lr": 2.1138499447344095e-05}
{"train_loss": 0.06908423453569412, "global_step": 189045, "epoch": 2124, "lr": 2.113802604703927e-05}
{"train_loss": 0.0341070331633091, "global_step": 189046, "epoch": 2124, "lr": 2.1137552650614573e-05}
{"train_loss": 0.045485369861125946, "global_step": 189047, "epoch": 2124, "lr": 2.1137079258070096e-05}
{"train_loss": 0.055950455367565155, "global_step": 189048, "epoch": 2124, "lr": 2.1136605869405873e-05}
{"train_loss": 0.06271715462207794, "global_step": 189049, "epoch": 2124, "lr": 2.1136132484621996e-05}
{"train_loss": 0.11378657817840576, "global_step": 189050, "epoch": 2124, "lr": 2.1135659103718493e-05}
{"train_loss": 0.060174666345119476, "global_step": 189051, "epoch": 2124, "lr": 2.1135185726695467e-05}
{"train_loss": 0.02951303869485855, "global_step": 189052, "epoch": 2124, "lr": 2.113471235355295e-05}
{"train_loss": 0.04135064780712128, "global_step": 189053, "epoch": 2124, "lr": 2.1134238984291034e-05}
{"train_loss": 0.1079339012503624, "global_step": 189054, "epoch": 2124, "lr": 2.113376561890975e-05}
{"train_loss": 0.03898469731211662, "global_step": 189055, "epoch": 2124, "lr": 2.1133292257409197e-05}
{"train_loss": 0.05949384719133377, "global_step": 189056, "epoch": 2124, "lr": 2.113281889978941e-05}
{"train_loss": 0.09937229752540588, "global_step": 189057, "epoch": 2124, "lr": 2.113234554605047e-05}
{"train_loss": 0.1284969598054886, "global_step": 189058, "epoch": 2124, "lr": 2.113187219619242e-05}
{"train_loss": 0.06462368369102478, "global_step": 189059, "epoch": 2124, "lr": 2.1131398850215362e-05}
{"train_loss": 0.08306176960468292, "global_step": 189060, "epoch": 2124, "lr": 2.1130925508119313e-05}
{"train_loss": 0.04103792458772659, "global_step": 189061, "epoch": 2124, "lr": 2.1130452169904373e-05}
{"train_loss": 0.03891535475850105, "global_step": 189062, "epoch": 2124, "lr": 2.1129978835570602e-05}
{"train_loss": 0.08346912264823914, "global_step": 189063, "epoch": 2124, "lr": 2.112950550511803e-05}
{"train_loss": 0.09022019803524017, "global_step": 189064, "epoch": 2124, "lr": 2.112903217854676e-05}
{"train_loss": 0.1177629604935646, "global_step": 189065, "epoch": 2124, "lr": 2.1128558855856823e-05}
{"train_loss": 0.07418029010295868, "global_step": 189066, "epoch": 2124, "lr": 2.112808553704832e-05}
{"train_loss": 0.049034830182790756, "global_step": 189067, "epoch": 2124, "lr": 2.1127612222121275e-05}
{"train_loss": 0.08839219808578491, "global_step": 189068, "epoch": 2124, "lr": 2.1127138911075787e-05}
{"train_loss": 0.06910701841115952, "global_step": 189069, "epoch": 2124, "lr": 2.1126665603911882e-05}
{"train_loss": 0.12365267425775528, "global_step": 189070, "epoch": 2124, "lr": 2.112619230062967e-05}
{"train_loss": 0.07622881233692169, "global_step": 189071, "epoch": 2124, "lr": 2.1125719001229165e-05}
{"train_loss": 0.053740669041872025, "global_step": 189072, "epoch": 2124, "lr": 2.112524570571048e-05}
{"train_loss": 0.03955421224236488, "global_step": 189073, "epoch": 2124, "lr": 2.1124772414073633e-05}
{"train_loss": 0.06735514104366302, "global_step": 189074, "epoch": 2124, "lr": 2.1124299126318726e-05}
{"train_loss": 0.06304580718278885, "global_step": 189075, "epoch": 2124, "lr": 2.1123825842445787e-05}
{"train_loss": 0.06592970341444016, "global_step": 189076, "epoch": 2124, "lr": 2.1123352562454912e-05}
{"train_loss": 0.07742493599653244, "global_step": 189077, "epoch": 2124, "lr": 2.1122879286346135e-05}
{"train_loss": 0.059292349964380264, "global_step": 189078, "epoch": 2124, "lr": 2.1122406014119554e-05}
{"train_loss": 0.057666704058647156, "global_step": 189079, "epoch": 2124, "lr": 2.1121932745775196e-05}
{"train_loss": 0.08355468511581421, "global_step": 189080, "epoch": 2124, "lr": 2.1121459481313146e-05}
{"train_loss": 0.07524298131465912, "global_step": 189081, "epoch": 2124, "lr": 2.1120986220733474e-05}
{"train_loss": 0.11611367017030716, "global_step": 189082, "epoch": 2124, "lr": 2.1120512964036222e-05}
{"train_loss": 0.03857491537928581, "global_step": 189083, "epoch": 2124, "lr": 2.1120039711221478e-05}
{"train_loss": 0.10081511735916138, "global_step": 189084, "epoch": 2124, "lr": 2.1119566462289275e-05}
{"train_loss": 0.13728885352611542, "global_step": 189085, "epoch": 2124, "lr": 2.1119093217239716e-05}
{"train_loss": 0.08632951229810715, "global_step": 189086, "epoch": 2124, "lr": 2.1118619976072816e-05}
{"train_loss": 0.06979090720415115, "global_step": 189087, "epoch": 2124, "lr": 2.1118146738788695e-05}
{"train_loss": 0.05007858946919441, "global_step": 189088, "epoch": 2124, "lr": 2.1117673505387364e-05}
{"train_loss": 0.09591825306415558, "global_step": 189089, "epoch": 2124, "lr": 2.1117200275868926e-05}
{"train_loss": 0.05358993634581566, "global_step": 189090, "epoch": 2124, "lr": 2.1116727050233415e-05}
{"train_loss": 0.04701903834939003, "global_step": 189091, "epoch": 2124, "lr": 2.111625382848092e-05}
{"train_loss": 0.05137950927019119, "global_step": 189092, "epoch": 2124, "lr": 2.111578061061148e-05}
{"train_loss": 0.10869253426790237, "global_step": 189093, "epoch": 2124, "lr": 2.111530739662519e-05}
{"train_loss": 0.06947540491819382, "global_step": 189094, "epoch": 2124, "lr": 2.111483418652207e-05}
{"train_loss": 0.10038993507623672, "global_step": 189095, "epoch": 2124, "lr": 2.1114360980302234e-05}
{"train_loss": 0.11167547106742859, "global_step": 189096, "epoch": 2124, "lr": 2.1113887777965697e-05}
{"train_loss": 0.08519706130027771, "global_step": 189097, "epoch": 2124, "lr": 2.1113414579512565e-05}
{"train_loss": 0.030863773077726364, "global_step": 189098, "epoch": 2124, "lr": 2.1112941384942862e-05}
{"train_loss": 0.112962506711483, "global_step": 189099, "epoch": 2124, "lr": 2.1112468194256697e-05}
{"train_loss": 0.05646573379635811, "global_step": 189100, "epoch": 2124, "lr": 2.1111995007454083e-05}
{"train_loss": 0.10676074773073196, "global_step": 189101, "epoch": 2124, "lr": 2.111152182453513e-05}
{"train_loss": 0.03691915050148964, "global_step": 189102, "epoch": 2124, "lr": 2.1111048645499882e-05}
{"train_loss": 0.06626524031162262, "global_step": 189103, "epoch": 2124, "lr": 2.1110575470348382e-05}
{"train_loss": 0.06835431605577469, "global_step": 189104, "epoch": 2124, "lr": 2.1110102299080735e-05}
{"train_loss": 0.06647320091724396, "global_step": 189105, "epoch": 2124, "lr": 2.1109629131696962e-05}
{"train_loss": 0.10765799134969711, "global_step": 189106, "epoch": 2124, "lr": 2.1109155968197163e-05}
{"train_loss": 0.09929995238780975, "global_step": 189107, "epoch": 2124, "lr": 2.110868280858137e-05}
{"train_loss": 0.0597255676984787, "global_step": 189108, "epoch": 2124, "lr": 2.1108209652849682e-05}
{"train_loss": 0.05473439395427704, "global_step": 189109, "epoch": 2124, "lr": 2.110773650100212e-05}
{"train_loss": 0.09461832046508789, "global_step": 189110, "epoch": 2124, "lr": 2.110726335303879e-05}
{"train_loss": 0.09113222360610962, "global_step": 189111, "epoch": 2124, "lr": 2.110679020895972e-05}
{"train_loss": 0.07686161249876022, "global_step": 189112, "epoch": 2124, "lr": 2.110631706876501e-05}
{"train_loss": 0.09158230572938919, "global_step": 189113, "epoch": 2124, "lr": 2.1105843932454683e-05}
{"train_loss": 0.04105936735868454, "global_step": 189114, "epoch": 2124, "lr": 2.110537080002884e-05}
{"train_loss": 0.06251604110002518, "global_step": 189115, "epoch": 2124, "lr": 2.110489767148751e-05}
{"train_loss": 0.0979522094130516, "global_step": 189116, "epoch": 2124, "lr": 2.1104424546830776e-05}
{"train_loss": 0.0494118332862854, "global_step": 189117, "epoch": 2124, "lr": 2.1103951426058717e-05}
{"train_loss": 0.04997701942920685, "global_step": 189118, "epoch": 2124, "lr": 2.1103478309171365e-05}
{"train_loss": 0.042021509259939194, "global_step": 189119, "epoch": 2124, "lr": 2.1103005196168814e-05}
{"train_loss": 0.06035049259662628, "global_step": 189120, "epoch": 2124, "lr": 2.110253208705109e-05}
{"train_loss": 0.07670212537050247, "global_step": 189121, "epoch": 2124, "lr": 2.1102058981818303e-05}
{"train_loss": 0.050260938704013824, "global_step": 189122, "epoch": 2124, "lr": 2.1101585880470464e-05}
{"train_loss": 0.07392691820859909, "global_step": 189123, "epoch": 2124, "lr": 2.110111278300769e-05}
{"train_loss": 0.07158637147271231, "global_step": 189124, "epoch": 2124, "lr": 2.1100639689430002e-05, "val_loss": 7.939034461975098}
{"train_loss": 0.059174228459596634, "global_step": 189125, "epoch": 2125, "lr": 2.1100166599737493e-05}
{"train_loss": 0.08286748081445694, "global_step": 189126, "epoch": 2125, "lr": 2.10996935139302e-05}
{"train_loss": 0.057431910187006, "global_step": 189127, "epoch": 2125, "lr": 2.109922043200822e-05}
{"train_loss": 0.08623805642127991, "global_step": 189128, "epoch": 2125, "lr": 2.1098747353971576e-05}
{"train_loss": 0.09359949827194214, "global_step": 189129, "epoch": 2125, "lr": 2.1098274279820373e-05}
{"train_loss": 0.05185861513018608, "global_step": 189130, "epoch": 2125, "lr": 2.109780120955464e-05}
{"train_loss": 0.08588118851184845, "global_step": 189131, "epoch": 2125, "lr": 2.1097328143174472e-05}
{"train_loss": 0.03894274681806564, "global_step": 189132, "epoch": 2125, "lr": 2.1096855080679896e-05}
{"train_loss": 0.05615174025297165, "global_step": 189133, "epoch": 2125, "lr": 2.1096382022071016e-05}
{"train_loss": 0.06112023815512657, "global_step": 189134, "epoch": 2125, "lr": 2.1095908967347856e-05}
{"train_loss": 0.06260237097740173, "global_step": 189135, "epoch": 2125, "lr": 2.1095435916510513e-05}
{"train_loss": 0.04864048585295677, "global_step": 189136, "epoch": 2125, "lr": 2.109496286955902e-05}
{"train_loss": 0.044640712440013885, "global_step": 189137, "epoch": 2125, "lr": 2.109448982649348e-05}
{"train_loss": 0.10126635432243347, "global_step": 189138, "epoch": 2125, "lr": 2.1094016787313908e-05}
{"train_loss": 0.043499160557985306, "global_step": 189139, "epoch": 2125, "lr": 2.1093543752020416e-05}
{"train_loss": 0.027951952069997787, "global_step": 189140, "epoch": 2125, "lr": 2.1093070720613044e-05}
{"train_loss": 0.04838911443948746, "global_step": 189141, "epoch": 2125, "lr": 2.1092597693091838e-05}
{"train_loss": 0.058393798768520355, "global_step": 189142, "epoch": 2125, "lr": 2.1092124669456898e-05}
{"train_loss": 0.08398096263408661, "global_step": 189143, "epoch": 2125, "lr": 2.1091651649708256e-05}
{"train_loss": 0.03309730067849159, "global_step": 189144, "epoch": 2125, "lr": 2.1091178633846e-05}
{"train_loss": 0.05257563292980194, "global_step": 189145, "epoch": 2125, "lr": 2.1090705621870165e-05}
{"train_loss": 0.08729638159275055, "global_step": 189146, "epoch": 2125, "lr": 2.1090232613780854e-05}
{"train_loss": 0.05134300887584686, "global_step": 189147, "epoch": 2125, "lr": 2.1089759609578093e-05}
{"train_loss": 0.09027642756700516, "global_step": 189148, "epoch": 2125, "lr": 2.1089286609261973e-05}
{"train_loss": 0.07154965400695801, "global_step": 189149, "epoch": 2125, "lr": 2.108881361283253e-05}
{"train_loss": 0.10909055173397064, "global_step": 189150, "epoch": 2125, "lr": 2.1088340620289866e-05}
{"train_loss": 0.06561784446239471, "global_step": 189151, "epoch": 2125, "lr": 2.1087867631633995e-05}
{"train_loss": 0.04941873252391815, "global_step": 189152, "epoch": 2125, "lr": 2.1087394646865016e-05}
{"train_loss": 0.10584532469511032, "global_step": 189153, "epoch": 2125, "lr": 2.1086921665983e-05}
{"train_loss": 0.04636165499687195, "global_step": 189154, "epoch": 2125, "lr": 2.1086448688987977e-05}
{"train_loss": 0.10896992683410645, "global_step": 189155, "epoch": 2125, "lr": 2.1085975715880046e-05}
{"train_loss": 0.04764220118522644, "global_step": 189156, "epoch": 2125, "lr": 2.1085502746659237e-05}
{"train_loss": 0.11505766957998276, "global_step": 189157, "epoch": 2125, "lr": 2.1085029781325645e-05}
{"train_loss": 0.066122867166996, "global_step": 189158, "epoch": 2125, "lr": 2.1084556819879305e-05}
{"train_loss": 0.08389894664287567, "global_step": 189159, "epoch": 2125, "lr": 2.108408386232031e-05}
{"train_loss": 0.056029513478279114, "global_step": 189160, "epoch": 2125, "lr": 2.1083610908648682e-05}
{"train_loss": 0.046509187668561935, "global_step": 189161, "epoch": 2125, "lr": 2.1083137958864535e-05}
{"train_loss": 0.13062676787376404, "global_step": 189162, "epoch": 2125, "lr": 2.108266501296789e-05}
{"train_loss": 0.10281466692686081, "global_step": 189163, "epoch": 2125, "lr": 2.1082192070958844e-05}
{"train_loss": 0.0797620415687561, "global_step": 189164, "epoch": 2125, "lr": 2.108171913283743e-05}
{"train_loss": 0.08089061826467514, "global_step": 189165, "epoch": 2125, "lr": 2.1081246198603744e-05}
{"train_loss": 0.10353430360555649, "global_step": 189166, "epoch": 2125, "lr": 2.1080773268257807e-05}
{"train_loss": 0.05065521225333214, "global_step": 189167, "epoch": 2125, "lr": 2.1080300341799734e-05}
{"train_loss": 0.05527608096599579, "global_step": 189168, "epoch": 2125, "lr": 2.1079827419229543e-05}
{"train_loss": 0.07931255549192429, "global_step": 189169, "epoch": 2125, "lr": 2.107935450054733e-05}
{"train_loss": 0.07737258821725845, "global_step": 189170, "epoch": 2125, "lr": 2.107888158575313e-05}
{"train_loss": 0.09589112550020218, "global_step": 189171, "epoch": 2125, "lr": 2.1078408674847022e-05}
{"train_loss": 0.05688612163066864, "global_step": 189172, "epoch": 2125, "lr": 2.1077935767829088e-05}
{"train_loss": 0.07214546948671341, "global_step": 189173, "epoch": 2125, "lr": 2.107746286469937e-05}
{"train_loss": 0.06525012105703354, "global_step": 189174, "epoch": 2125, "lr": 2.1076989965457923e-05}
{"train_loss": 0.04436413198709488, "global_step": 189175, "epoch": 2125, "lr": 2.1076517070104835e-05}
{"train_loss": 0.054058805108070374, "global_step": 189176, "epoch": 2125, "lr": 2.1076044178640136e-05}
{"train_loss": 0.09244450181722641, "global_step": 189177, "epoch": 2125, "lr": 2.1075571291063932e-05}
{"train_loss": 0.039201974868774414, "global_step": 189178, "epoch": 2125, "lr": 2.107509840737627e-05}
{"train_loss": 0.08218832314014435, "global_step": 189179, "epoch": 2125, "lr": 2.1074625527577187e-05}
{"train_loss": 0.09268618375062943, "global_step": 189180, "epoch": 2125, "lr": 2.1074152651666778e-05}
{"train_loss": 0.07081176340579987, "global_step": 189181, "epoch": 2125, "lr": 2.107367977964509e-05}
{"train_loss": 0.10214546322822571, "global_step": 189182, "epoch": 2125, "lr": 2.1073206911512205e-05}
{"train_loss": 0.06844218075275421, "global_step": 189183, "epoch": 2125, "lr": 2.1072734047268154e-05}
{"train_loss": 0.05572587996721268, "global_step": 189184, "epoch": 2125, "lr": 2.1072261186913044e-05}
{"train_loss": 0.0771985724568367, "global_step": 189185, "epoch": 2125, "lr": 2.10717883304469e-05}
{"train_loss": 0.06819117814302444, "global_step": 189186, "epoch": 2125, "lr": 2.107131547786982e-05}
{"train_loss": 0.09095881879329681, "global_step": 189187, "epoch": 2125, "lr": 2.1070842629181826e-05}
{"train_loss": 0.1098155677318573, "global_step": 189188, "epoch": 2125, "lr": 2.1070369784383005e-05}
{"train_loss": 0.12815789878368378, "global_step": 189189, "epoch": 2125, "lr": 2.1069896943473437e-05}
{"train_loss": 0.05420494079589844, "global_step": 189190, "epoch": 2125, "lr": 2.1069424106453157e-05}
{"train_loss": 0.051815930753946304, "global_step": 189191, "epoch": 2125, "lr": 2.106895127332226e-05}
{"train_loss": 0.044957611709833145, "global_step": 189192, "epoch": 2125, "lr": 2.1068478444080763e-05}
{"train_loss": 0.08376549184322357, "global_step": 189193, "epoch": 2125, "lr": 2.1068005618728775e-05}
{"train_loss": 0.0625704973936081, "global_step": 189194, "epoch": 2125, "lr": 2.1067532797266325e-05}
{"train_loss": 0.07951591908931732, "global_step": 189195, "epoch": 2125, "lr": 2.1067059979693514e-05}
{"train_loss": 0.05487899109721184, "global_step": 189196, "epoch": 2125, "lr": 2.1066587166010364e-05}
{"train_loss": 0.0801967903971672, "global_step": 189197, "epoch": 2125, "lr": 2.1066114356216975e-05}
{"train_loss": 0.05279526114463806, "global_step": 189198, "epoch": 2125, "lr": 2.106564155031338e-05}
{"train_loss": 0.06590068340301514, "global_step": 189199, "epoch": 2125, "lr": 2.1065168748299675e-05}
{"train_loss": 0.05947898328304291, "global_step": 189200, "epoch": 2125, "lr": 2.1064695950175883e-05}
{"train_loss": 0.039238542318344116, "global_step": 189201, "epoch": 2125, "lr": 2.106422315594211e-05}
{"train_loss": 0.06656406074762344, "global_step": 189202, "epoch": 2125, "lr": 2.1063750365598383e-05}
{"train_loss": 0.020728938281536102, "global_step": 189203, "epoch": 2125, "lr": 2.1063277579144797e-05}
{"train_loss": 0.055310722440481186, "global_step": 189204, "epoch": 2125, "lr": 2.106280479658138e-05}
{"train_loss": 0.08178862929344177, "global_step": 189205, "epoch": 2125, "lr": 2.1062332017908242e-05}
{"train_loss": 0.04559943452477455, "global_step": 189206, "epoch": 2125, "lr": 2.10618592431254e-05}
{"train_loss": 0.05936328321695328, "global_step": 189207, "epoch": 2125, "lr": 2.1061386472232936e-05}
{"train_loss": 0.04075310006737709, "global_step": 189208, "epoch": 2125, "lr": 2.1060913705230934e-05}
{"train_loss": 0.05630113556981087, "global_step": 189209, "epoch": 2125, "lr": 2.1060440942119425e-05}
{"train_loss": 0.1028977781534195, "global_step": 189210, "epoch": 2125, "lr": 2.1059968182898504e-05}
{"train_loss": 0.09485369175672531, "global_step": 189211, "epoch": 2125, "lr": 2.10594954275682e-05}
{"train_loss": 0.10929913818836212, "global_step": 189212, "epoch": 2125, "lr": 2.105902267612861e-05}
{"train_loss": 0.0700453036370572, "global_step": 189213, "epoch": 2125, "lr": 2.105854992857978e-05, "val_loss": 7.727846622467041, "train_action_mse_error": 14.229426383972168}
{"train_loss": 0.05552590265870094, "global_step": 189214, "epoch": 2126, "lr": 2.105807718492176e-05}
{"train_loss": 0.08352929353713989, "global_step": 189215, "epoch": 2126, "lr": 2.1057604445154643e-05}
{"train_loss": 0.06391498446464539, "global_step": 189216, "epoch": 2126, "lr": 2.1057131709278465e-05}
{"train_loss": 0.04956863075494766, "global_step": 189217, "epoch": 2126, "lr": 2.105665897729332e-05}
{"train_loss": 0.03921414166688919, "global_step": 189218, "epoch": 2126, "lr": 2.1056186249199257e-05}
{"train_loss": 0.052288781851530075, "global_step": 189219, "epoch": 2126, "lr": 2.105571352499631e-05}
{"train_loss": 0.05675819516181946, "global_step": 189220, "epoch": 2126, "lr": 2.1055240804684594e-05}
{"train_loss": 0.02068358100950718, "global_step": 189221, "epoch": 2126, "lr": 2.1054768088264133e-05}
{"train_loss": 0.07431759685277939, "global_step": 189222, "epoch": 2126, "lr": 2.1054295375735015e-05}
{"train_loss": 0.05451704561710358, "global_step": 189223, "epoch": 2126, "lr": 2.1053822667097283e-05}
{"train_loss": 0.04893629997968674, "global_step": 189224, "epoch": 2126, "lr": 2.1053349962351033e-05}
{"train_loss": 0.12404854595661163, "global_step": 189225, "epoch": 2126, "lr": 2.105287726149628e-05}
{"train_loss": 0.08485598862171173, "global_step": 189226, "epoch": 2126, "lr": 2.1052404564533117e-05}
{"train_loss": 0.05724204704165459, "global_step": 189227, "epoch": 2126, "lr": 2.1051931871461627e-05}
{"train_loss": 0.030556276440620422, "global_step": 189228, "epoch": 2126, "lr": 2.1051459182281834e-05}
{"train_loss": 0.09886138141155243, "global_step": 189229, "epoch": 2126, "lr": 2.1050986496993835e-05}
{"train_loss": 0.0915173888206482, "global_step": 189230, "epoch": 2126, "lr": 2.105051381559766e-05}
{"train_loss": 0.050788041204214096, "global_step": 189231, "epoch": 2126, "lr": 2.1050041138093413e-05}
{"train_loss": 0.05013106018304825, "global_step": 189232, "epoch": 2126, "lr": 2.1049568464481113e-05}
{"train_loss": 0.061282526701688766, "global_step": 189233, "epoch": 2126, "lr": 2.1049095794760865e-05}
{"train_loss": 0.038316428661346436, "global_step": 189234, "epoch": 2126, "lr": 2.1048623128932692e-05}
{"train_loss": 0.0669817253947258, "global_step": 189235, "epoch": 2126, "lr": 2.1048150466996697e-05}
{"train_loss": 0.06384047120809555, "global_step": 189236, "epoch": 2126, "lr": 2.104767780895291e-05}
{"train_loss": 0.06493302434682846, "global_step": 189237, "epoch": 2126, "lr": 2.1047205154801432e-05}
{"train_loss": 0.05373278632760048, "global_step": 189238, "epoch": 2126, "lr": 2.1046732504542277e-05}
{"train_loss": 0.03835301846265793, "global_step": 189239, "epoch": 2126, "lr": 2.104625985817556e-05}
{"train_loss": 0.1388956606388092, "global_step": 189240, "epoch": 2126, "lr": 2.1045787215701303e-05}
{"train_loss": 0.08753959089517593, "global_step": 189241, "epoch": 2126, "lr": 2.10453145771196e-05}
{"train_loss": 0.05747076869010925, "global_step": 189242, "epoch": 2126, "lr": 2.1044841942430488e-05}
{"train_loss": 0.06115332990884781, "global_step": 189243, "epoch": 2126, "lr": 2.1044369311634043e-05}
{"train_loss": 0.07019373029470444, "global_step": 189244, "epoch": 2126, "lr": 2.1043896684730348e-05}
{"train_loss": 0.08491964638233185, "global_step": 189245, "epoch": 2126, "lr": 2.104342406171943e-05}
{"train_loss": 0.11488427221775055, "global_step": 189246, "epoch": 2126, "lr": 2.1042951442601387e-05}
{"train_loss": 0.05915936455130577, "global_step": 189247, "epoch": 2126, "lr": 2.104247882737625e-05}
{"train_loss": 0.08329293131828308, "global_step": 189248, "epoch": 2126, "lr": 2.1042006216044114e-05}
{"train_loss": 0.06632620096206665, "global_step": 189249, "epoch": 2126, "lr": 2.1041533608605013e-05}
{"train_loss": 0.09312336146831512, "global_step": 189250, "epoch": 2126, "lr": 2.1041061005059044e-05}
{"train_loss": 0.16402758657932281, "global_step": 189251, "epoch": 2126, "lr": 2.104058840540625e-05}
{"train_loss": 0.06302616000175476, "global_step": 189252, "epoch": 2126, "lr": 2.104011580964667e-05}
{"train_loss": 0.02105831541121006, "global_step": 189253, "epoch": 2126, "lr": 2.103964321778042e-05}
{"train_loss": 0.07282678037881851, "global_step": 189254, "epoch": 2126, "lr": 2.103917062980752e-05}
{"train_loss": 0.05065702274441719, "global_step": 189255, "epoch": 2126, "lr": 2.103869804572806e-05}
{"train_loss": 0.05764633044600487, "global_step": 189256, "epoch": 2126, "lr": 2.1038225465542104e-05}
{"train_loss": 0.06615331023931503, "global_step": 189257, "epoch": 2126, "lr": 2.1037752889249685e-05}
{"train_loss": 0.05076569691300392, "global_step": 189258, "epoch": 2126, "lr": 2.10372803168509e-05}
{"train_loss": 0.05696478858590126, "global_step": 189259, "epoch": 2126, "lr": 2.103680774834579e-05}
{"train_loss": 0.07319649308919907, "global_step": 189260, "epoch": 2126, "lr": 2.103633518373444e-05}
{"train_loss": 0.02648613601922989, "global_step": 189261, "epoch": 2126, "lr": 2.1035862623016882e-05}
{"train_loss": 0.07472561299800873, "global_step": 189262, "epoch": 2126, "lr": 2.1035390066193207e-05}
{"train_loss": 0.05769040808081627, "global_step": 189263, "epoch": 2126, "lr": 2.103491751326348e-05}
{"train_loss": 0.061172619462013245, "global_step": 189264, "epoch": 2126, "lr": 2.1034444964227738e-05}
{"train_loss": 0.035414472222328186, "global_step": 189265, "epoch": 2126, "lr": 2.103397241908609e-05}
{"train_loss": 0.06490938365459442, "global_step": 189266, "epoch": 2126, "lr": 2.103349987783854e-05}
{"train_loss": 0.06960425525903702, "global_step": 189267, "epoch": 2126, "lr": 2.1033027340485212e-05}
{"train_loss": 0.025531761348247528, "global_step": 189268, "epoch": 2126, "lr": 2.1032554807026117e-05}
{"train_loss": 0.07305487990379333, "global_step": 189269, "epoch": 2126, "lr": 2.103208227746136e-05}
{"train_loss": 0.04283539205789566, "global_step": 189270, "epoch": 2126, "lr": 2.1031609751790972e-05}
{"train_loss": 0.03988945484161377, "global_step": 189271, "epoch": 2126, "lr": 2.1031137230015047e-05}
{"train_loss": 0.05312393605709076, "global_step": 189272, "epoch": 2126, "lr": 2.1030664712133613e-05}
{"train_loss": 0.07846471667289734, "global_step": 189273, "epoch": 2126, "lr": 2.103019219814677e-05}
{"train_loss": 0.047809116542339325, "global_step": 189274, "epoch": 2126, "lr": 2.102971968805455e-05}
{"train_loss": 0.1096293106675148, "global_step": 189275, "epoch": 2126, "lr": 2.102924718185705e-05}
{"train_loss": 0.04478989541530609, "global_step": 189276, "epoch": 2126, "lr": 2.1028774679554293e-05}
{"train_loss": 0.05613337829709053, "global_step": 189277, "epoch": 2126, "lr": 2.102830218114638e-05}
{"train_loss": 0.06540360301733017, "global_step": 189278, "epoch": 2126, "lr": 2.1027829686633345e-05}
{"train_loss": 0.06460181623697281, "global_step": 189279, "epoch": 2126, "lr": 2.1027357196015262e-05}
{"train_loss": 0.07369548827409744, "global_step": 189280, "epoch": 2126, "lr": 2.1026884709292217e-05}
{"train_loss": 0.057885754853487015, "global_step": 189281, "epoch": 2126, "lr": 2.102641222646424e-05}
{"train_loss": 0.10468365252017975, "global_step": 189282, "epoch": 2126, "lr": 2.102593974753142e-05}
{"train_loss": 0.061565760523080826, "global_step": 189283, "epoch": 2126, "lr": 2.10254672724938e-05}
{"train_loss": 0.028124826028943062, "global_step": 189284, "epoch": 2126, "lr": 2.102499480135146e-05}
{"train_loss": 0.11896715313196182, "global_step": 189285, "epoch": 2126, "lr": 2.1024522334104447e-05}
{"train_loss": 0.10180523991584778, "global_step": 189286, "epoch": 2126, "lr": 2.102404987075285e-05}
{"train_loss": 0.06307022273540497, "global_step": 189287, "epoch": 2126, "lr": 2.1023577411296695e-05}
{"train_loss": 0.034012727439403534, "global_step": 189288, "epoch": 2126, "lr": 2.102310495573609e-05}
{"train_loss": 0.11661729216575623, "global_step": 189289, "epoch": 2126, "lr": 2.102263250407107e-05}
{"train_loss": 0.07226362824440002, "global_step": 189290, "epoch": 2126, "lr": 2.1022160056301688e-05}
{"train_loss": 0.0510999970138073, "global_step": 189291, "epoch": 2126, "lr": 2.1021687612428043e-05}
{"train_loss": 0.04612383991479874, "global_step": 189292, "epoch": 2126, "lr": 2.102121517245016e-05}
{"train_loss": 0.1039493978023529, "global_step": 189293, "epoch": 2126, "lr": 2.102074273636814e-05}
{"train_loss": 0.08188670128583908, "global_step": 189294, "epoch": 2126, "lr": 2.1020270304182027e-05}
{"train_loss": 0.1288955956697464, "global_step": 189295, "epoch": 2126, "lr": 2.1019797875891868e-05}
{"train_loss": 0.05623042955994606, "global_step": 189296, "epoch": 2126, "lr": 2.1019325451497755e-05}
{"train_loss": 0.05667079985141754, "global_step": 189297, "epoch": 2126, "lr": 2.101885303099973e-05}
{"train_loss": 0.13021863996982574, "global_step": 189298, "epoch": 2126, "lr": 2.101838061439787e-05}
{"train_loss": 0.09530416876077652, "global_step": 189299, "epoch": 2126, "lr": 2.101790820169225e-05}
{"train_loss": 0.0700354352593422, "global_step": 189300, "epoch": 2126, "lr": 2.10174357928829e-05}
{"train_loss": 0.10720966011285782, "global_step": 189301, "epoch": 2126, "lr": 2.1016963387969917e-05}
{"train_loss": 0.06890574794555648, "global_step": 189302, "epoch": 2126, "lr": 2.1016490986953334e-05, "val_loss": 7.856113433837891}
{"train_loss": 0.08236870169639587, "global_step": 189303, "epoch": 2127, "lr": 2.1016018589833252e-05}
{"train_loss": 0.08632440119981766, "global_step": 189304, "epoch": 2127, "lr": 2.101554619660969e-05}
{"train_loss": 0.08115794509649277, "global_step": 189305, "epoch": 2127, "lr": 2.101507380728276e-05}
{"train_loss": 0.09576112031936646, "global_step": 189306, "epoch": 2127, "lr": 2.101460142185248e-05}
{"train_loss": 0.07164972275495529, "global_step": 189307, "epoch": 2127, "lr": 2.1014129040318946e-05}
{"train_loss": 0.05501171573996544, "global_step": 189308, "epoch": 2127, "lr": 2.1013656662682198e-05}
{"train_loss": 0.09508028626441956, "global_step": 189309, "epoch": 2127, "lr": 2.1013184288942322e-05}
{"train_loss": 0.057343386113643646, "global_step": 189310, "epoch": 2127, "lr": 2.101271191909936e-05}
{"train_loss": 0.06341663002967834, "global_step": 189311, "epoch": 2127, "lr": 2.1012239553153394e-05}
{"train_loss": 0.06111466884613037, "global_step": 189312, "epoch": 2127, "lr": 2.101176719110447e-05}
{"train_loss": 0.07818668335676193, "global_step": 189313, "epoch": 2127, "lr": 2.101129483295267e-05}
{"train_loss": 0.066409170627594, "global_step": 189314, "epoch": 2127, "lr": 2.101082247869804e-05}
{"train_loss": 0.05659164860844612, "global_step": 189315, "epoch": 2127, "lr": 2.1010350128340644e-05}
{"train_loss": 0.08088061213493347, "global_step": 189316, "epoch": 2127, "lr": 2.1009877781880573e-05}
{"train_loss": 0.08356127142906189, "global_step": 189317, "epoch": 2127, "lr": 2.1009405439317852e-05}
{"train_loss": 0.09017788618803024, "global_step": 189318, "epoch": 2127, "lr": 2.1008933100652585e-05}
{"train_loss": 0.07199299335479736, "global_step": 189319, "epoch": 2127, "lr": 2.100846076588479e-05}
{"train_loss": 0.09649636596441269, "global_step": 189320, "epoch": 2127, "lr": 2.1007988435014577e-05}
{"train_loss": 0.04664240777492523, "global_step": 189321, "epoch": 2127, "lr": 2.1007516108041965e-05}
{"train_loss": 0.09471184015274048, "global_step": 189322, "epoch": 2127, "lr": 2.1007043784967057e-05}
{"train_loss": 0.0772063210606575, "global_step": 189323, "epoch": 2127, "lr": 2.1006571465789883e-05}
{"train_loss": 0.05893559008836746, "global_step": 189324, "epoch": 2127, "lr": 2.1006099150510537e-05}
{"train_loss": 0.09033404290676117, "global_step": 189325, "epoch": 2127, "lr": 2.100562683912905e-05}
{"train_loss": 0.12372863292694092, "global_step": 189326, "epoch": 2127, "lr": 2.1005154531645523e-05}
{"train_loss": 0.05235804617404938, "global_step": 189327, "epoch": 2127, "lr": 2.100468222805998e-05}
{"train_loss": 0.06311778724193573, "global_step": 189328, "epoch": 2127, "lr": 2.1004209928372524e-05}
{"train_loss": 0.0793934166431427, "global_step": 189329, "epoch": 2127, "lr": 2.1003737632583197e-05}
{"train_loss": 0.08965972810983658, "global_step": 189330, "epoch": 2127, "lr": 2.1003265340692047e-05}
{"train_loss": 0.05069369077682495, "global_step": 189331, "epoch": 2127, "lr": 2.1002793052699177e-05}
{"train_loss": 0.07401608675718307, "global_step": 189332, "epoch": 2127, "lr": 2.1002320768604607e-05}
{"train_loss": 0.0690271332859993, "global_step": 189333, "epoch": 2127, "lr": 2.100184848840843e-05}
{"train_loss": 0.0828595906496048, "global_step": 189334, "epoch": 2127, "lr": 2.1001376212110697e-05}
{"train_loss": 0.08176808059215546, "global_step": 189335, "epoch": 2127, "lr": 2.1000903939711486e-05}
{"train_loss": 0.06995419412851334, "global_step": 189336, "epoch": 2127, "lr": 2.1000431671210837e-05}
{"train_loss": 0.09188605844974518, "global_step": 189337, "epoch": 2127, "lr": 2.099995940660884e-05}
{"train_loss": 0.09199734777212143, "global_step": 189338, "epoch": 2127, "lr": 2.0999487145905527e-05}
{"train_loss": 0.05506501719355583, "global_step": 189339, "epoch": 2127, "lr": 2.0999014889101e-05}
{"train_loss": 0.07553711533546448, "global_step": 189340, "epoch": 2127, "lr": 2.099854263619528e-05}
{"train_loss": 0.11967802792787552, "global_step": 189341, "epoch": 2127, "lr": 2.0998070387188473e-05}
{"train_loss": 0.09104274213314056, "global_step": 189342, "epoch": 2127, "lr": 2.0997598142080603e-05}
{"train_loss": 0.04380377382040024, "global_step": 189343, "epoch": 2127, "lr": 2.0997125900871772e-05}
{"train_loss": 0.06372201442718506, "global_step": 189344, "epoch": 2127, "lr": 2.0996653663562006e-05}
{"train_loss": 0.10367683321237564, "global_step": 189345, "epoch": 2127, "lr": 2.09961814301514e-05}
{"train_loss": 0.06389901041984558, "global_step": 189346, "epoch": 2127, "lr": 2.0995709200639992e-05}
{"train_loss": 0.05196917802095413, "global_step": 189347, "epoch": 2127, "lr": 2.0995236975027873e-05}
{"train_loss": 0.08829798549413681, "global_step": 189348, "epoch": 2127, "lr": 2.0994764753315073e-05}
{"train_loss": 0.040427304804325104, "global_step": 189349, "epoch": 2127, "lr": 2.099429253550169e-05}
{"train_loss": 0.03419080376625061, "global_step": 189350, "epoch": 2127, "lr": 2.0993820321587754e-05}
{"train_loss": 0.08990340679883957, "global_step": 189351, "epoch": 2127, "lr": 2.0993348111573364e-05}
{"train_loss": 0.10106217116117477, "global_step": 189352, "epoch": 2127, "lr": 2.0992875905458548e-05}
{"train_loss": 0.04413877800107002, "global_step": 189353, "epoch": 2127, "lr": 2.0992403703243386e-05}
{"train_loss": 0.059008702635765076, "global_step": 189354, "epoch": 2127, "lr": 2.0991931504927954e-05}
{"train_loss": 0.039064813405275345, "global_step": 189355, "epoch": 2127, "lr": 2.0991459310512292e-05}
{"train_loss": 0.046609677374362946, "global_step": 189356, "epoch": 2127, "lr": 2.0990987119996492e-05}
{"train_loss": 0.1381659209728241, "global_step": 189357, "epoch": 2127, "lr": 2.0990514933380577e-05}
{"train_loss": 0.06146729737520218, "global_step": 189358, "epoch": 2127, "lr": 2.0990042750664656e-05}
{"train_loss": 0.06283527612686157, "global_step": 189359, "epoch": 2127, "lr": 2.098957057184875e-05}
{"train_loss": 0.03427410125732422, "global_step": 189360, "epoch": 2127, "lr": 2.0989098396932964e-05}
{"train_loss": 0.10095889866352081, "global_step": 189361, "epoch": 2127, "lr": 2.098862622591732e-05}
{"train_loss": 0.06410772353410721, "global_step": 189362, "epoch": 2127, "lr": 2.098815405880192e-05}
{"train_loss": 0.056014545261859894, "global_step": 189363, "epoch": 2127, "lr": 2.098768189558679e-05}
{"train_loss": 0.06126999109983444, "global_step": 189364, "epoch": 2127, "lr": 2.0987209736272028e-05}
{"train_loss": 0.11781921982765198, "global_step": 189365, "epoch": 2127, "lr": 2.098673758085767e-05}
{"train_loss": 0.06549108773469925, "global_step": 189366, "epoch": 2127, "lr": 2.098626542934381e-05}
{"train_loss": 0.085331492125988, "global_step": 189367, "epoch": 2127, "lr": 2.0985793281730486e-05}
{"train_loss": 0.07184195518493652, "global_step": 189368, "epoch": 2127, "lr": 2.0985321138017755e-05}
{"train_loss": 0.044313449412584305, "global_step": 189369, "epoch": 2127, "lr": 2.0984848998205713e-05}
{"train_loss": 0.08558674901723862, "global_step": 189370, "epoch": 2127, "lr": 2.0984376862294387e-05}
{"train_loss": 0.08953007310628891, "global_step": 189371, "epoch": 2127, "lr": 2.098390473028387e-05}
{"train_loss": 0.09335684031248093, "global_step": 189372, "epoch": 2127, "lr": 2.0983432602174198e-05}
{"train_loss": 0.05943784862756729, "global_step": 189373, "epoch": 2127, "lr": 2.098296047796547e-05}
{"train_loss": 0.11047281324863434, "global_step": 189374, "epoch": 2127, "lr": 2.098248835765771e-05}
{"train_loss": 0.04426775872707367, "global_step": 189375, "epoch": 2127, "lr": 2.0982016241251016e-05}
{"train_loss": 0.042987603694200516, "global_step": 189376, "epoch": 2127, "lr": 2.0981544128745423e-05}
{"train_loss": 0.01781921088695526, "global_step": 189377, "epoch": 2127, "lr": 2.0981072020141025e-05}
{"train_loss": 0.04427347332239151, "global_step": 189378, "epoch": 2127, "lr": 2.098059991543785e-05}
{"train_loss": 0.07355020195245743, "global_step": 189379, "epoch": 2127, "lr": 2.0980127814635997e-05}
{"train_loss": 0.08039705455303192, "global_step": 189380, "epoch": 2127, "lr": 2.097965571773549e-05}
{"train_loss": 0.11986162513494492, "global_step": 189381, "epoch": 2127, "lr": 2.0979183624736436e-05}
{"train_loss": 0.048338670283555984, "global_step": 189382, "epoch": 2127, "lr": 2.097871153563886e-05}
{"train_loss": 0.06099861487746239, "global_step": 189383, "epoch": 2127, "lr": 2.097823945044286e-05}
{"train_loss": 0.03917304798960686, "global_step": 189384, "epoch": 2127, "lr": 2.0977767369148465e-05}
{"train_loss": 0.07091809809207916, "global_step": 189385, "epoch": 2127, "lr": 2.0977295291755772e-05}
{"train_loss": 0.08690416812896729, "global_step": 189386, "epoch": 2127, "lr": 2.0976823218264814e-05}
{"train_loss": 0.05934963747859001, "global_step": 189387, "epoch": 2127, "lr": 2.0976351148675675e-05}
{"train_loss": 0.09733057767152786, "global_step": 189388, "epoch": 2127, "lr": 2.0975879082988405e-05}
{"train_loss": 0.04220930486917496, "global_step": 189389, "epoch": 2127, "lr": 2.097540702120307e-05}
{"train_loss": 0.12494538724422455, "global_step": 189390, "epoch": 2127, "lr": 2.097493496331976e-05}
{"train_loss": 0.07270070676053508, "global_step": 189391, "epoch": 2127, "lr": 2.0974462909338495e-05, "val_loss": 7.740290641784668}
{"train_loss": 0.08518281579017639, "global_step": 189392, "epoch": 2128, "lr": 2.0973990859259375e-05}
{"train_loss": 0.07634469121694565, "global_step": 189393, "epoch": 2128, "lr": 2.097351881308243e-05}
{"train_loss": 0.07082611322402954, "global_step": 189394, "epoch": 2128, "lr": 2.0973046770807764e-05}
{"train_loss": 0.05033815652132034, "global_step": 189395, "epoch": 2128, "lr": 2.0972574732435395e-05}
{"train_loss": 0.07415593415498734, "global_step": 189396, "epoch": 2128, "lr": 2.0972102697965436e-05}
{"train_loss": 0.1006522923707962, "global_step": 189397, "epoch": 2128, "lr": 2.0971630667397896e-05}
{"train_loss": 0.05093079060316086, "global_step": 189398, "epoch": 2128, "lr": 2.097115864073289e-05}
{"train_loss": 0.06497105211019516, "global_step": 189399, "epoch": 2128, "lr": 2.097068661797044e-05}
{"train_loss": 0.08840515464544296, "global_step": 189400, "epoch": 2128, "lr": 2.0970214599110645e-05}
{"train_loss": 0.07738498598337173, "global_step": 189401, "epoch": 2128, "lr": 2.0969742584153533e-05}
{"train_loss": 0.14758965373039246, "global_step": 189402, "epoch": 2128, "lr": 2.0969270573099203e-05}
{"train_loss": 0.06945797801017761, "global_step": 189403, "epoch": 2128, "lr": 2.0968798565947682e-05}
{"train_loss": 0.09865717589855194, "global_step": 189404, "epoch": 2128, "lr": 2.096832656269907e-05}
{"train_loss": 0.08469001203775406, "global_step": 189405, "epoch": 2128, "lr": 2.096785456335341e-05}
{"train_loss": 0.07184811681509018, "global_step": 189406, "epoch": 2128, "lr": 2.0967382567910755e-05}
{"train_loss": 0.08845768868923187, "global_step": 189407, "epoch": 2128, "lr": 2.0966910576371197e-05}
{"train_loss": 0.09019136428833008, "global_step": 189408, "epoch": 2128, "lr": 2.0966438588734767e-05}
{"train_loss": 0.05988955870270729, "global_step": 189409, "epoch": 2128, "lr": 2.0965966605001562e-05}
{"train_loss": 0.02025974914431572, "global_step": 189410, "epoch": 2128, "lr": 2.0965494625171618e-05}
{"train_loss": 0.05526111647486687, "global_step": 189411, "epoch": 2128, "lr": 2.0965022649245018e-05}
{"train_loss": 0.1811894327402115, "global_step": 189412, "epoch": 2128, "lr": 2.09645506772218e-05}
{"train_loss": 0.05753679201006889, "global_step": 189413, "epoch": 2128, "lr": 2.096407870910207e-05}
{"train_loss": 0.09002971649169922, "global_step": 189414, "epoch": 2128, "lr": 2.096360674488584e-05}
{"train_loss": 0.0823693796992302, "global_step": 189415, "epoch": 2128, "lr": 2.0963134784573218e-05}
{"train_loss": 0.037699781358242035, "global_step": 189416, "epoch": 2128, "lr": 2.0962662828164232e-05}
{"train_loss": 0.08893433213233948, "global_step": 189417, "epoch": 2128, "lr": 2.096219087565898e-05}
{"train_loss": 0.03313731402158737, "global_step": 189418, "epoch": 2128, "lr": 2.0961718927057488e-05}
{"train_loss": 0.07034614682197571, "global_step": 189419, "epoch": 2128, "lr": 2.096124698235986e-05}
{"train_loss": 0.06411854177713394, "global_step": 189420, "epoch": 2128, "lr": 2.0960775041566122e-05}
{"train_loss": 0.07574182748794556, "global_step": 189421, "epoch": 2128, "lr": 2.0960303104676367e-05}
{"train_loss": 0.045895665884017944, "global_step": 189422, "epoch": 2128, "lr": 2.095983117169063e-05}
{"train_loss": 0.07221627980470657, "global_step": 189423, "epoch": 2128, "lr": 2.0959359242609005e-05}
{"train_loss": 0.14983434975147247, "global_step": 189424, "epoch": 2128, "lr": 2.0958887317431524e-05}
{"train_loss": 0.05150824040174484, "global_step": 189425, "epoch": 2128, "lr": 2.0958415396158272e-05}
{"train_loss": 0.048390015959739685, "global_step": 189426, "epoch": 2128, "lr": 2.0957943478789316e-05}
{"train_loss": 0.08591346442699432, "global_step": 189427, "epoch": 2128, "lr": 2.0957471565324698e-05}
{"train_loss": 0.047725215554237366, "global_step": 189428, "epoch": 2128, "lr": 2.095699965576451e-05}
{"train_loss": 0.03782106935977936, "global_step": 189429, "epoch": 2128, "lr": 2.095652775010878e-05}
{"train_loss": 0.1024559810757637, "global_step": 189430, "epoch": 2128, "lr": 2.0956055848357613e-05}
{"train_loss": 0.070587657392025, "global_step": 189431, "epoch": 2128, "lr": 2.0955583950511025e-05}
{"train_loss": 0.06243695318698883, "global_step": 189432, "epoch": 2128, "lr": 2.095511205656913e-05}
{"train_loss": 0.056040264666080475, "global_step": 189433, "epoch": 2128, "lr": 2.0954640166531946e-05}
{"train_loss": 0.06386644393205643, "global_step": 189434, "epoch": 2128, "lr": 2.0954168280399577e-05}
{"train_loss": 0.07846880704164505, "global_step": 189435, "epoch": 2128, "lr": 2.0953696398172046e-05}
{"train_loss": 0.09926498681306839, "global_step": 189436, "epoch": 2128, "lr": 2.0953224519849453e-05}
{"train_loss": 0.13250941038131714, "global_step": 189437, "epoch": 2128, "lr": 2.0952752645431827e-05}
{"train_loss": 0.05538417771458626, "global_step": 189438, "epoch": 2128, "lr": 2.095228077491927e-05}
{"train_loss": 0.04707147181034088, "global_step": 189439, "epoch": 2128, "lr": 2.0951808908311803e-05}
{"train_loss": 0.09297185391187668, "global_step": 189440, "epoch": 2128, "lr": 2.095133704560953e-05}
{"train_loss": 0.11192137748003006, "global_step": 189441, "epoch": 2128, "lr": 2.095086518681248e-05}
{"train_loss": 0.06093868613243103, "global_step": 189442, "epoch": 2128, "lr": 2.0950393331920753e-05}
{"train_loss": 0.04360105097293854, "global_step": 189443, "epoch": 2128, "lr": 2.0949921480934365e-05}
{"train_loss": 0.08531150966882706, "global_step": 189444, "epoch": 2128, "lr": 2.094944963385343e-05}
{"train_loss": 0.05843237414956093, "global_step": 189445, "epoch": 2128, "lr": 2.0948977790677988e-05}
{"train_loss": 0.06582881510257721, "global_step": 189446, "epoch": 2128, "lr": 2.0948505951408082e-05}
{"train_loss": 0.052911024540662766, "global_step": 189447, "epoch": 2128, "lr": 2.094803411604381e-05}
{"train_loss": 0.0323895625770092, "global_step": 189448, "epoch": 2128, "lr": 2.094756228458521e-05}
{"train_loss": 0.09077653288841248, "global_step": 189449, "epoch": 2128, "lr": 2.094709045703237e-05}
{"train_loss": 0.0748688355088234, "global_step": 189450, "epoch": 2128, "lr": 2.0946618633385322e-05}
{"train_loss": 0.08880340307950974, "global_step": 189451, "epoch": 2128, "lr": 2.0946146813644164e-05}
{"train_loss": 0.064597487449646, "global_step": 189452, "epoch": 2128, "lr": 2.0945674997808922e-05}
{"train_loss": 0.10450350493192673, "global_step": 189453, "epoch": 2128, "lr": 2.0945203185879702e-05}
{"train_loss": 0.05384787544608116, "global_step": 189454, "epoch": 2128, "lr": 2.094473137785653e-05}
{"train_loss": 0.10389239341020584, "global_step": 189455, "epoch": 2128, "lr": 2.09442595737395e-05}
{"train_loss": 0.09024898707866669, "global_step": 189456, "epoch": 2128, "lr": 2.094378777352864e-05}
{"train_loss": 0.0451883040368557, "global_step": 189457, "epoch": 2128, "lr": 2.0943315977224053e-05}
{"train_loss": 0.09499247372150421, "global_step": 189458, "epoch": 2128, "lr": 2.094284418482576e-05}
{"train_loss": 0.10788601636886597, "global_step": 189459, "epoch": 2128, "lr": 2.0942372396333877e-05}
{"train_loss": 0.08483649790287018, "global_step": 189460, "epoch": 2128, "lr": 2.094190061174841e-05}
{"train_loss": 0.025434566661715508, "global_step": 189461, "epoch": 2128, "lr": 2.0941428831069455e-05}
{"train_loss": 0.05926283448934555, "global_step": 189462, "epoch": 2128, "lr": 2.0940957054297083e-05}
{"train_loss": 0.08983686566352844, "global_step": 189463, "epoch": 2128, "lr": 2.094048528143133e-05}
{"train_loss": 0.09800254553556442, "global_step": 189464, "epoch": 2128, "lr": 2.0940013512472296e-05}
{"train_loss": 0.03850654512643814, "global_step": 189465, "epoch": 2128, "lr": 2.0939541747420006e-05}
{"train_loss": 0.04471985623240471, "global_step": 189466, "epoch": 2128, "lr": 2.093906998627455e-05}
{"train_loss": 0.07313892245292664, "global_step": 189467, "epoch": 2128, "lr": 2.0938598229035972e-05}
{"train_loss": 0.043542519211769104, "global_step": 189468, "epoch": 2128, "lr": 2.0938126475704362e-05}
{"train_loss": 0.06260906159877777, "global_step": 189469, "epoch": 2128, "lr": 2.093765472627975e-05}
{"train_loss": 0.13154058158397675, "global_step": 189470, "epoch": 2128, "lr": 2.0937182980762225e-05}
{"train_loss": 0.06067486107349396, "global_step": 189471, "epoch": 2128, "lr": 2.093671123915183e-05}
{"train_loss": 0.07362490892410278, "global_step": 189472, "epoch": 2128, "lr": 2.0936239501448663e-05}
{"train_loss": 0.09203182905912399, "global_step": 189473, "epoch": 2128, "lr": 2.0935767767652736e-05}
{"train_loss": 0.052171796560287476, "global_step": 189474, "epoch": 2128, "lr": 2.093529603776417e-05}
{"train_loss": 0.057955071330070496, "global_step": 189475, "epoch": 2128, "lr": 2.093482431178297e-05}
{"train_loss": 0.04884523153305054, "global_step": 189476, "epoch": 2128, "lr": 2.093435258970925e-05}
{"train_loss": 0.05173519626259804, "global_step": 189477, "epoch": 2128, "lr": 2.0933880871543036e-05}
{"train_loss": 0.0357513427734375, "global_step": 189478, "epoch": 2128, "lr": 2.093340915728443e-05}
{"train_loss": 0.05902047082781792, "global_step": 189479, "epoch": 2128, "lr": 2.0932937446933447e-05}
{"train_loss": 0.07265189463754049, "global_step": 189480, "epoch": 2128, "lr": 2.09324657404902e-05, "val_loss": 7.751628875732422}
{"train_loss": 0.07293473184108734, "global_step": 189481, "epoch": 2129, "lr": 2.0931994037954705e-05}
{"train_loss": 0.07649235427379608, "global_step": 189482, "epoch": 2129, "lr": 2.0931522339327068e-05}
{"train_loss": 0.07942432910203934, "global_step": 189483, "epoch": 2129, "lr": 2.0931050644607337e-05}
{"train_loss": 0.09510117769241333, "global_step": 189484, "epoch": 2129, "lr": 2.0930578953795554e-05}
{"train_loss": 0.05408243462443352, "global_step": 189485, "epoch": 2129, "lr": 2.093010726689182e-05}
{"train_loss": 0.08272559940814972, "global_step": 189486, "epoch": 2129, "lr": 2.092963558389615e-05}
{"train_loss": 0.05480019748210907, "global_step": 189487, "epoch": 2129, "lr": 2.0929163904808663e-05}
{"train_loss": 0.04428883269429207, "global_step": 189488, "epoch": 2129, "lr": 2.0928692229629375e-05}
{"train_loss": 0.06325870007276535, "global_step": 189489, "epoch": 2129, "lr": 2.092822055835839e-05}
{"train_loss": 0.07185098528862, "global_step": 189490, "epoch": 2129, "lr": 2.0927748890995737e-05}
{"train_loss": 0.061244651675224304, "global_step": 189491, "epoch": 2129, "lr": 2.092727722754151e-05}
{"train_loss": 0.08939298987388611, "global_step": 189492, "epoch": 2129, "lr": 2.0926805567995732e-05}
{"train_loss": 0.09113239496946335, "global_step": 189493, "epoch": 2129, "lr": 2.0926333912358514e-05}
{"train_loss": 0.06470275670289993, "global_step": 189494, "epoch": 2129, "lr": 2.0925862260629875e-05}
{"train_loss": 0.06198587268590927, "global_step": 189495, "epoch": 2129, "lr": 2.092539061280992e-05}
{"train_loss": 0.08416210860013962, "global_step": 189496, "epoch": 2129, "lr": 2.092491896889867e-05}
{"train_loss": 0.07941164821386337, "global_step": 189497, "epoch": 2129, "lr": 2.0924447328896213e-05}
{"train_loss": 0.064116932451725, "global_step": 189498, "epoch": 2129, "lr": 2.0923975692802633e-05}
{"train_loss": 0.04656761512160301, "global_step": 189499, "epoch": 2129, "lr": 2.0923504060617945e-05}
{"train_loss": 0.05897822603583336, "global_step": 189500, "epoch": 2129, "lr": 2.0923032432342256e-05}
{"train_loss": 0.04317541420459747, "global_step": 189501, "epoch": 2129, "lr": 2.0922560807975593e-05}
{"train_loss": 0.031103091314435005, "global_step": 189502, "epoch": 2129, "lr": 2.0922089187518055e-05}
{"train_loss": 0.04922766983509064, "global_step": 189503, "epoch": 2129, "lr": 2.0921617570969676e-05}
{"train_loss": 0.05867467075586319, "global_step": 189504, "epoch": 2129, "lr": 2.0921145958330547e-05}
{"train_loss": 0.11370617896318436, "global_step": 189505, "epoch": 2129, "lr": 2.092067434960069e-05}
{"train_loss": 0.0627683773636818, "global_step": 189506, "epoch": 2129, "lr": 2.0920202744780216e-05}
{"train_loss": 0.03504899516701698, "global_step": 189507, "epoch": 2129, "lr": 2.0919731143869153e-05}
{"train_loss": 0.1040940135717392, "global_step": 189508, "epoch": 2129, "lr": 2.091925954686759e-05}
{"train_loss": 0.039619214832782745, "global_step": 189509, "epoch": 2129, "lr": 2.091878795377556e-05}
{"train_loss": 0.06984776258468628, "global_step": 189510, "epoch": 2129, "lr": 2.0918316364593165e-05}
{"train_loss": 0.03805321455001831, "global_step": 189511, "epoch": 2129, "lr": 2.0917844779320427e-05}
{"train_loss": 0.0646720603108406, "global_step": 189512, "epoch": 2129, "lr": 2.091737319795745e-05}
{"train_loss": 0.08661037683486938, "global_step": 189513, "epoch": 2129, "lr": 2.0916901620504264e-05}
{"train_loss": 0.043088648468256, "global_step": 189514, "epoch": 2129, "lr": 2.091643004696096e-05}
{"train_loss": 0.049385763704776764, "global_step": 189515, "epoch": 2129, "lr": 2.0915958477327568e-05}
{"train_loss": 0.0768524557352066, "global_step": 189516, "epoch": 2129, "lr": 2.091548691160419e-05}
{"train_loss": 0.1369103044271469, "global_step": 189517, "epoch": 2129, "lr": 2.0915015349790858e-05}
{"train_loss": 0.038608308881521225, "global_step": 189518, "epoch": 2129, "lr": 2.091454379188766e-05}
{"train_loss": 0.06227748095989227, "global_step": 189519, "epoch": 2129, "lr": 2.0914072237894628e-05}
{"train_loss": 0.0692019984126091, "global_step": 189520, "epoch": 2129, "lr": 2.0913600687811867e-05}
{"train_loss": 0.064882792532444, "global_step": 189521, "epoch": 2129, "lr": 2.0913129141639414e-05}
{"train_loss": 0.01621343195438385, "global_step": 189522, "epoch": 2129, "lr": 2.091265759937732e-05}
{"train_loss": 0.06976161897182465, "global_step": 189523, "epoch": 2129, "lr": 2.091218606102568e-05}
{"train_loss": 0.06529605388641357, "global_step": 189524, "epoch": 2129, "lr": 2.091171452658453e-05}
{"train_loss": 0.05769430100917816, "global_step": 189525, "epoch": 2129, "lr": 2.091124299605396e-05}
{"train_loss": 0.11530445516109467, "global_step": 189526, "epoch": 2129, "lr": 2.0910771469433998e-05}
{"train_loss": 0.04164208844304085, "global_step": 189527, "epoch": 2129, "lr": 2.091029994672475e-05}
{"train_loss": 0.05951298773288727, "global_step": 189528, "epoch": 2129, "lr": 2.0909828427926236e-05}
{"train_loss": 0.03987891599535942, "global_step": 189529, "epoch": 2129, "lr": 2.0909356913038562e-05}
{"train_loss": 0.04068965092301369, "global_step": 189530, "epoch": 2129, "lr": 2.090888540206175e-05}
{"train_loss": 0.053889986127614975, "global_step": 189531, "epoch": 2129, "lr": 2.0908413894995903e-05}
{"train_loss": 0.09899202734231949, "global_step": 189532, "epoch": 2129, "lr": 2.090794239184105e-05}
{"train_loss": 0.062090914696455, "global_step": 189533, "epoch": 2129, "lr": 2.090747089259727e-05}
{"train_loss": 0.10985168814659119, "global_step": 189534, "epoch": 2129, "lr": 2.0906999397264638e-05}
{"train_loss": 0.06528107821941376, "global_step": 189535, "epoch": 2129, "lr": 2.090652790584319e-05}
{"train_loss": 0.050160881131887436, "global_step": 189536, "epoch": 2129, "lr": 2.0906056418333024e-05}
{"train_loss": 0.0432526059448719, "global_step": 189537, "epoch": 2129, "lr": 2.0905584934734167e-05}
{"train_loss": 0.08279619365930557, "global_step": 189538, "epoch": 2129, "lr": 2.0905113455046716e-05}
{"train_loss": 0.0899716392159462, "global_step": 189539, "epoch": 2129, "lr": 2.09046419792707e-05}
{"train_loss": 0.06919099390506744, "global_step": 189540, "epoch": 2129, "lr": 2.0904170507406217e-05}
{"train_loss": 0.0762329027056694, "global_step": 189541, "epoch": 2129, "lr": 2.0903699039453295e-05}
{"train_loss": 0.052048686891794205, "global_step": 189542, "epoch": 2129, "lr": 2.0903227575412036e-05}
{"train_loss": 0.06385013461112976, "global_step": 189543, "epoch": 2129, "lr": 2.0902756115282467e-05}
{"train_loss": 0.06448671221733093, "global_step": 189544, "epoch": 2129, "lr": 2.0902284659064685e-05}
{"train_loss": 0.10789903253316879, "global_step": 189545, "epoch": 2129, "lr": 2.0901813206758712e-05}
{"train_loss": 0.12249574065208435, "global_step": 189546, "epoch": 2129, "lr": 2.0901341758364662e-05}
{"train_loss": 0.0781053826212883, "global_step": 189547, "epoch": 2129, "lr": 2.0900870313882553e-05}
{"train_loss": 0.09465836733579636, "global_step": 189548, "epoch": 2129, "lr": 2.0900398873312482e-05}
{"train_loss": 0.0936334878206253, "global_step": 189549, "epoch": 2129, "lr": 2.0899927436654483e-05}
{"train_loss": 0.12062135338783264, "global_step": 189550, "epoch": 2129, "lr": 2.0899456003908645e-05}
{"train_loss": 0.0776258111000061, "global_step": 189551, "epoch": 2129, "lr": 2.089898457507501e-05}
{"train_loss": 0.03176948055624962, "global_step": 189552, "epoch": 2129, "lr": 2.0898513150153648e-05}
{"train_loss": 0.07944270223379135, "global_step": 189553, "epoch": 2129, "lr": 2.0898041729144648e-05}
{"train_loss": 0.055884554982185364, "global_step": 189554, "epoch": 2129, "lr": 2.0897570312048025e-05}
{"train_loss": 0.041968926787376404, "global_step": 189555, "epoch": 2129, "lr": 2.0897098898863894e-05}
{"train_loss": 0.038802240043878555, "global_step": 189556, "epoch": 2129, "lr": 2.089662748959229e-05}
{"train_loss": 0.07746509462594986, "global_step": 189557, "epoch": 2129, "lr": 2.0896156084233264e-05}
{"train_loss": 0.05470641329884529, "global_step": 189558, "epoch": 2129, "lr": 2.0895684682786916e-05}
{"train_loss": 0.11747856438159943, "global_step": 189559, "epoch": 2129, "lr": 2.0895213285253268e-05}
{"train_loss": 0.09772799909114838, "global_step": 189560, "epoch": 2129, "lr": 2.0894741891632417e-05}
{"train_loss": 0.10112855583429337, "global_step": 189561, "epoch": 2129, "lr": 2.0894270501924422e-05}
{"train_loss": 0.09294385462999344, "global_step": 189562, "epoch": 2129, "lr": 2.0893799116129315e-05}
{"train_loss": 0.036476537585258484, "global_step": 189563, "epoch": 2129, "lr": 2.08933277342472e-05}
{"train_loss": 0.08019629120826721, "global_step": 189564, "epoch": 2129, "lr": 2.08928563562781e-05}
{"train_loss": 0.06986716389656067, "global_step": 189565, "epoch": 2129, "lr": 2.0892384982222123e-05}
{"train_loss": 0.08252143859863281, "global_step": 189566, "epoch": 2129, "lr": 2.0891913612079294e-05}
{"train_loss": 0.07512950152158737, "global_step": 189567, "epoch": 2129, "lr": 2.0891442245849706e-05}
{"train_loss": 0.045774783939123154, "global_step": 189568, "epoch": 2129, "lr": 2.0890970883533394e-05}
{"train_loss": 0.06929687006671108, "global_step": 189569, "epoch": 2129, "lr": 2.0890499525130435e-05, "val_loss": 7.620517730712891}
{"train_loss": 0.040409304201602936, "global_step": 189570, "epoch": 2130, "lr": 2.0890028170640914e-05}
{"train_loss": 0.10750062018632889, "global_step": 189571, "epoch": 2130, "lr": 2.0889556820064855e-05}
{"train_loss": 0.06627737730741501, "global_step": 189572, "epoch": 2130, "lr": 2.088908547340236e-05}
{"train_loss": 0.09147301316261292, "global_step": 189573, "epoch": 2130, "lr": 2.088861413065345e-05}
{"train_loss": 0.060810819268226624, "global_step": 189574, "epoch": 2130, "lr": 2.0888142791818233e-05}
{"train_loss": 0.1108061671257019, "global_step": 189575, "epoch": 2130, "lr": 2.0887671456896728e-05}
{"train_loss": 0.06732656806707382, "global_step": 189576, "epoch": 2130, "lr": 2.088720012588904e-05}
{"train_loss": 0.06608978658914566, "global_step": 189577, "epoch": 2130, "lr": 2.0886728798795198e-05}
{"train_loss": 0.05031004548072815, "global_step": 189578, "epoch": 2130, "lr": 2.0886257475615294e-05}
{"train_loss": 0.07310555875301361, "global_step": 189579, "epoch": 2130, "lr": 2.0885786156349364e-05}
{"train_loss": 0.049134090542793274, "global_step": 189580, "epoch": 2130, "lr": 2.08853148409975e-05}
{"train_loss": 0.11576586216688156, "global_step": 189581, "epoch": 2130, "lr": 2.0884843529559735e-05}
{"train_loss": 0.07723277062177658, "global_step": 189582, "epoch": 2130, "lr": 2.0884372222036165e-05}
{"train_loss": 0.06630389392375946, "global_step": 189583, "epoch": 2130, "lr": 2.088390091842682e-05}
{"train_loss": 0.054417479783296585, "global_step": 189584, "epoch": 2130, "lr": 2.088342961873179e-05}
{"train_loss": 0.04743706434965134, "global_step": 189585, "epoch": 2130, "lr": 2.088295832295112e-05}
{"train_loss": 0.12029790878295898, "global_step": 189586, "epoch": 2130, "lr": 2.0882487031084896e-05}
{"train_loss": 0.10925475507974625, "global_step": 189587, "epoch": 2130, "lr": 2.088201574313315e-05}
{"train_loss": 0.08255275338888168, "global_step": 189588, "epoch": 2130, "lr": 2.0881544459095957e-05}
{"train_loss": 0.05945281311869621, "global_step": 189589, "epoch": 2130, "lr": 2.088107317897341e-05}
{"train_loss": 0.1334538459777832, "global_step": 189590, "epoch": 2130, "lr": 2.0880601902765524e-05}
{"train_loss": 0.04730093106627464, "global_step": 189591, "epoch": 2130, "lr": 2.0880130630472405e-05}
{"train_loss": 0.052189555019140244, "global_step": 189592, "epoch": 2130, "lr": 2.0879659362094083e-05}
{"train_loss": 0.05523114651441574, "global_step": 189593, "epoch": 2130, "lr": 2.0879188097630653e-05}
{"train_loss": 0.05833156779408455, "global_step": 189594, "epoch": 2130, "lr": 2.0878716837082158e-05}
{"train_loss": 0.05868450179696083, "global_step": 189595, "epoch": 2130, "lr": 2.0878245580448645e-05}
{"train_loss": 0.1255701184272766, "global_step": 189596, "epoch": 2130, "lr": 2.087777432773022e-05}
{"train_loss": 0.04184505715966225, "global_step": 189597, "epoch": 2130, "lr": 2.087730307892691e-05}
{"train_loss": 0.08664163947105408, "global_step": 189598, "epoch": 2130, "lr": 2.0876831834038802e-05}
{"train_loss": 0.07238119840621948, "global_step": 189599, "epoch": 2130, "lr": 2.0876360593065947e-05}
{"train_loss": 0.08135790377855301, "global_step": 189600, "epoch": 2130, "lr": 2.08758893560084e-05}
{"train_loss": 0.08988180756568909, "global_step": 189601, "epoch": 2130, "lr": 2.0875418122866248e-05}
{"train_loss": 0.061151545494794846, "global_step": 189602, "epoch": 2130, "lr": 2.0874946893639524e-05}
{"train_loss": 0.0736822858452797, "global_step": 189603, "epoch": 2130, "lr": 2.0874475668328324e-05}
{"train_loss": 0.07340866327285767, "global_step": 189604, "epoch": 2130, "lr": 2.0874004446932684e-05}
{"train_loss": 0.05828088894486427, "global_step": 189605, "epoch": 2130, "lr": 2.087353322945269e-05}
{"train_loss": 0.11186094582080841, "global_step": 189606, "epoch": 2130, "lr": 2.0873062015888378e-05}
{"train_loss": 0.06169862672686577, "global_step": 189607, "epoch": 2130, "lr": 2.0872590806239827e-05}
{"train_loss": 0.0761147066950798, "global_step": 189608, "epoch": 2130, "lr": 2.0872119600507123e-05}
{"train_loss": 0.06376306712627411, "global_step": 189609, "epoch": 2130, "lr": 2.087164839869029e-05}
{"train_loss": 0.045614104717969894, "global_step": 189610, "epoch": 2130, "lr": 2.0871177200789426e-05}
{"train_loss": 0.06320185214281082, "global_step": 189611, "epoch": 2130, "lr": 2.0870706006804558e-05}
{"train_loss": 0.04245743900537491, "global_step": 189612, "epoch": 2130, "lr": 2.0870234816735783e-05}
{"train_loss": 0.09251180291175842, "global_step": 189613, "epoch": 2130, "lr": 2.0869763630583137e-05}
{"train_loss": 0.03338518738746643, "global_step": 189614, "epoch": 2130, "lr": 2.086929244834671e-05}
{"train_loss": 0.06546948105096817, "global_step": 189615, "epoch": 2130, "lr": 2.086882127002654e-05}
{"train_loss": 0.0695597380399704, "global_step": 189616, "epoch": 2130, "lr": 2.0868350095622713e-05}
{"train_loss": 0.04101762920618057, "global_step": 189617, "epoch": 2130, "lr": 2.086787892513527e-05}
{"train_loss": 0.06676839292049408, "global_step": 189618, "epoch": 2130, "lr": 2.08674077585643e-05}
{"train_loss": 0.09033053368330002, "global_step": 189619, "epoch": 2130, "lr": 2.086693659590983e-05}
{"train_loss": 0.06085261330008507, "global_step": 189620, "epoch": 2130, "lr": 2.086646543717197e-05}
{"train_loss": 0.04875312000513077, "global_step": 189621, "epoch": 2130, "lr": 2.086599428235074e-05}
{"train_loss": 0.06747393310070038, "global_step": 189622, "epoch": 2130, "lr": 2.0865523131446234e-05}
{"train_loss": 0.06974831968545914, "global_step": 189623, "epoch": 2130, "lr": 2.086505198445849e-05}
{"train_loss": 0.04273300990462303, "global_step": 189624, "epoch": 2130, "lr": 2.0864580841387583e-05}
{"train_loss": 0.040532056242227554, "global_step": 189625, "epoch": 2130, "lr": 2.0864109702233597e-05}
{"train_loss": 0.08053120225667953, "global_step": 189626, "epoch": 2130, "lr": 2.086363856699656e-05}
{"train_loss": 0.05116446316242218, "global_step": 189627, "epoch": 2130, "lr": 2.0863167435676567e-05}
{"train_loss": 0.05171116814017296, "global_step": 189628, "epoch": 2130, "lr": 2.0862696308273648e-05}
{"train_loss": 0.06494162231683731, "global_step": 189629, "epoch": 2130, "lr": 2.0862225184787905e-05}
{"train_loss": 0.06529144942760468, "global_step": 189630, "epoch": 2130, "lr": 2.0861754065219357e-05}
{"train_loss": 0.0466378815472126, "global_step": 189631, "epoch": 2130, "lr": 2.086128294956812e-05}
{"train_loss": 0.10192565619945526, "global_step": 189632, "epoch": 2130, "lr": 2.0860811837834216e-05}
{"train_loss": 0.08242415636777878, "global_step": 189633, "epoch": 2130, "lr": 2.0860340730017707e-05}
{"train_loss": 0.05992081016302109, "global_step": 189634, "epoch": 2130, "lr": 2.085986962611869e-05}
{"train_loss": 0.06488413363695145, "global_step": 189635, "epoch": 2130, "lr": 2.0859398526137192e-05}
{"train_loss": 0.07363168150186539, "global_step": 189636, "epoch": 2130, "lr": 2.0858927430073307e-05}
{"train_loss": 0.11788837611675262, "global_step": 189637, "epoch": 2130, "lr": 2.0858456337927084e-05}
{"train_loss": 0.06031817942857742, "global_step": 189638, "epoch": 2130, "lr": 2.085798524969857e-05}
{"train_loss": 0.10687793791294098, "global_step": 189639, "epoch": 2130, "lr": 2.0857514165387865e-05}
{"train_loss": 0.07983708381652832, "global_step": 189640, "epoch": 2130, "lr": 2.0857043084994987e-05}
{"train_loss": 0.07160921394824982, "global_step": 189641, "epoch": 2130, "lr": 2.0856572008520046e-05}
{"train_loss": 0.06088904291391373, "global_step": 189642, "epoch": 2130, "lr": 2.085610093596307e-05}
{"train_loss": 0.05525350570678711, "global_step": 189643, "epoch": 2130, "lr": 2.0855629867324126e-05}
{"train_loss": 0.054347798228263855, "global_step": 189644, "epoch": 2130, "lr": 2.0855158802603313e-05}
{"train_loss": 0.08292178064584732, "global_step": 189645, "epoch": 2130, "lr": 2.085468774180065e-05}
{"train_loss": 0.11292462050914764, "global_step": 189646, "epoch": 2130, "lr": 2.085421668491624e-05}
{"train_loss": 0.07430040836334229, "global_step": 189647, "epoch": 2130, "lr": 2.08537456319501e-05}
{"train_loss": 0.05509349703788757, "global_step": 189648, "epoch": 2130, "lr": 2.0853274582902343e-05}
{"train_loss": 0.0813799723982811, "global_step": 189649, "epoch": 2130, "lr": 2.0852803537772985e-05}
{"train_loss": 0.11187000572681427, "global_step": 189650, "epoch": 2130, "lr": 2.0852332496562133e-05}
{"train_loss": 0.1234879195690155, "global_step": 189651, "epoch": 2130, "lr": 2.085186145926981e-05}
{"train_loss": 0.10076278448104858, "global_step": 189652, "epoch": 2130, "lr": 2.0851390425896115e-05}
{"train_loss": 0.05894758179783821, "global_step": 189653, "epoch": 2130, "lr": 2.085091939644108e-05}
{"train_loss": 0.1276422142982483, "global_step": 189654, "epoch": 2130, "lr": 2.085044837090481e-05}
{"train_loss": 0.06366679817438126, "global_step": 189655, "epoch": 2130, "lr": 2.0849977349287313e-05}
{"train_loss": 0.11178602278232574, "global_step": 189656, "epoch": 2130, "lr": 2.0849506331588703e-05}
{"train_loss": 0.08147326111793518, "global_step": 189657, "epoch": 2130, "lr": 2.0849035317809e-05}
{"train_loss": 0.07321679843275734, "global_step": 189658, "epoch": 2130, "lr": 2.084856430794831e-05, "val_loss": 7.719524383544922, "train_action_mse_error": 16.665502548217773}
{"train_loss": 0.12809211015701294, "global_step": 189659, "epoch": 2131, "lr": 2.0848093302006662e-05}
{"train_loss": 0.01961730234324932, "global_step": 189660, "epoch": 2131, "lr": 2.084762229998412e-05}
{"train_loss": 0.14167830348014832, "global_step": 189661, "epoch": 2131, "lr": 2.084715130188079e-05}
{"train_loss": 0.054370224475860596, "global_step": 189662, "epoch": 2131, "lr": 2.084668030769668e-05}
{"train_loss": 0.09012619405984879, "global_step": 189663, "epoch": 2131, "lr": 2.0846209317431897e-05}
{"train_loss": 0.04624726623296738, "global_step": 189664, "epoch": 2131, "lr": 2.0845738331086473e-05}
{"train_loss": 0.06283709406852722, "global_step": 189665, "epoch": 2131, "lr": 2.0845267348660498e-05}
{"train_loss": 0.053611963987350464, "global_step": 189666, "epoch": 2131, "lr": 2.0844796370154008e-05}
{"train_loss": 0.05781865492463112, "global_step": 189667, "epoch": 2131, "lr": 2.0844325395567095e-05}
{"train_loss": 0.050436049699783325, "global_step": 189668, "epoch": 2131, "lr": 2.0843854424899783e-05}
{"train_loss": 0.09889326244592667, "global_step": 189669, "epoch": 2131, "lr": 2.0843383458152184e-05}
{"train_loss": 0.04421057924628258, "global_step": 189670, "epoch": 2131, "lr": 2.0842912495324318e-05}
{"train_loss": 0.09445439279079437, "global_step": 189671, "epoch": 2131, "lr": 2.0842441536416287e-05}
{"train_loss": 0.07352111488580704, "global_step": 189672, "epoch": 2131, "lr": 2.0841970581428127e-05}
{"train_loss": 0.06575999408960342, "global_step": 189673, "epoch": 2131, "lr": 2.0841499630359894e-05}
{"train_loss": 0.053790025413036346, "global_step": 189674, "epoch": 2131, "lr": 2.084102868321169e-05}
{"train_loss": 0.06788700073957443, "global_step": 189675, "epoch": 2131, "lr": 2.0840557739983528e-05}
{"train_loss": 0.07024461776018143, "global_step": 189676, "epoch": 2131, "lr": 2.0840086800675524e-05}
{"train_loss": 0.0947931557893753, "global_step": 189677, "epoch": 2131, "lr": 2.083961586528771e-05}
{"train_loss": 0.0864897295832634, "global_step": 189678, "epoch": 2131, "lr": 2.083914493382013e-05}
{"train_loss": 0.050898484885692596, "global_step": 189679, "epoch": 2131, "lr": 2.0838674006272885e-05}
{"train_loss": 0.081484854221344, "global_step": 189680, "epoch": 2131, "lr": 2.0838203082646036e-05}
{"train_loss": 0.08757338672876358, "global_step": 189681, "epoch": 2131, "lr": 2.0837732162939614e-05}
{"train_loss": 0.05129938945174217, "global_step": 189682, "epoch": 2131, "lr": 2.0837261247153722e-05}
{"train_loss": 0.05053582042455673, "global_step": 189683, "epoch": 2131, "lr": 2.0836790335288393e-05}
{"train_loss": 0.07916681468486786, "global_step": 189684, "epoch": 2131, "lr": 2.0836319427343715e-05}
{"train_loss": 0.06465791165828705, "global_step": 189685, "epoch": 2131, "lr": 2.0835848523319722e-05}
{"train_loss": 0.050085484981536865, "global_step": 189686, "epoch": 2131, "lr": 2.0835377623216513e-05}
{"train_loss": 0.06718574464321136, "global_step": 189687, "epoch": 2131, "lr": 2.083490672703411e-05}
{"train_loss": 0.07537579536437988, "global_step": 189688, "epoch": 2131, "lr": 2.083443583477262e-05}
{"train_loss": 0.12025179713964462, "global_step": 189689, "epoch": 2131, "lr": 2.0833964946432066e-05}
{"train_loss": 0.020424729213118553, "global_step": 189690, "epoch": 2131, "lr": 2.083349406201255e-05}
{"train_loss": 0.06526977568864822, "global_step": 189691, "epoch": 2131, "lr": 2.083302318151409e-05}
{"train_loss": 0.11884641647338867, "global_step": 189692, "epoch": 2131, "lr": 2.0832552304936797e-05}
{"train_loss": 0.06429912894964218, "global_step": 189693, "epoch": 2131, "lr": 2.083208143228069e-05}
{"train_loss": 0.05940151959657669, "global_step": 189694, "epoch": 2131, "lr": 2.0831610563545883e-05}
{"train_loss": 0.04915288835763931, "global_step": 189695, "epoch": 2131, "lr": 2.0831139698732384e-05}
{"train_loss": 0.10378655791282654, "global_step": 189696, "epoch": 2131, "lr": 2.0830668837840284e-05}
{"train_loss": 0.007813380099833012, "global_step": 189697, "epoch": 2131, "lr": 2.0830197980869663e-05}
{"train_loss": 0.08798135071992874, "global_step": 189698, "epoch": 2131, "lr": 2.0829727127820547e-05}
{"train_loss": 0.06059667095541954, "global_step": 189699, "epoch": 2131, "lr": 2.0829256278693037e-05}
{"train_loss": 0.03538668900728226, "global_step": 189700, "epoch": 2131, "lr": 2.0828785433487165e-05}
{"train_loss": 0.07064221799373627, "global_step": 189701, "epoch": 2131, "lr": 2.0828314592203017e-05}
{"train_loss": 0.1104993149638176, "global_step": 189702, "epoch": 2131, "lr": 2.0827843754840636e-05}
{"train_loss": 0.07975656539201736, "global_step": 189703, "epoch": 2131, "lr": 2.0827372921400112e-05}
{"train_loss": 0.05263343080878258, "global_step": 189704, "epoch": 2131, "lr": 2.0826902091881473e-05}
{"train_loss": 0.0463305301964283, "global_step": 189705, "epoch": 2131, "lr": 2.082643126628482e-05}
{"train_loss": 0.09276021271944046, "global_step": 189706, "epoch": 2131, "lr": 2.0825960444610183e-05}
{"train_loss": 0.041517022997140884, "global_step": 189707, "epoch": 2131, "lr": 2.082548962685765e-05}
{"train_loss": 0.05958142504096031, "global_step": 189708, "epoch": 2131, "lr": 2.0825018813027265e-05}
{"train_loss": 0.07789723575115204, "global_step": 189709, "epoch": 2131, "lr": 2.082454800311911e-05}
{"train_loss": 0.06212145462632179, "global_step": 189710, "epoch": 2131, "lr": 2.0824077197133247e-05}
{"train_loss": 0.12376981228590012, "global_step": 189711, "epoch": 2131, "lr": 2.0823606395069707e-05}
{"train_loss": 0.06539255380630493, "global_step": 189712, "epoch": 2131, "lr": 2.0823135596928593e-05}
{"train_loss": 0.0670076236128807, "global_step": 189713, "epoch": 2131, "lr": 2.0822664802709947e-05}
{"train_loss": 0.07773482799530029, "global_step": 189714, "epoch": 2131, "lr": 2.082219401241385e-05}
{"train_loss": 0.13633489608764648, "global_step": 189715, "epoch": 2131, "lr": 2.082172322604033e-05}
{"train_loss": 0.15600742399692535, "global_step": 189716, "epoch": 2131, "lr": 2.0821252443589496e-05}
{"train_loss": 0.07175573706626892, "global_step": 189717, "epoch": 2131, "lr": 2.0820781665061373e-05}
{"train_loss": 0.13069717586040497, "global_step": 189718, "epoch": 2131, "lr": 2.0820310890456053e-05}
{"train_loss": 0.047004178166389465, "global_step": 189719, "epoch": 2131, "lr": 2.0819840119773572e-05}
{"train_loss": 0.14184615015983582, "global_step": 189720, "epoch": 2131, "lr": 2.0819369353014023e-05}
{"train_loss": 0.0962957963347435, "global_step": 189721, "epoch": 2131, "lr": 2.0818898590177442e-05}
{"train_loss": 0.030874324962496758, "global_step": 189722, "epoch": 2131, "lr": 2.081842783126392e-05}
{"train_loss": 0.11758633702993393, "global_step": 189723, "epoch": 2131, "lr": 2.0817957076273486e-05}
{"train_loss": 0.10701990127563477, "global_step": 189724, "epoch": 2131, "lr": 2.081748632520624e-05}
{"train_loss": 0.059824440628290176, "global_step": 189725, "epoch": 2131, "lr": 2.0817015578062204e-05}
{"train_loss": 0.09586318582296371, "global_step": 189726, "epoch": 2131, "lr": 2.081654483484149e-05}
{"train_loss": 0.07779138535261154, "global_step": 189727, "epoch": 2131, "lr": 2.0816074095544115e-05}
{"train_loss": 0.06351877748966217, "global_step": 189728, "epoch": 2131, "lr": 2.0815603360170182e-05}
{"train_loss": 0.10936769843101501, "global_step": 189729, "epoch": 2131, "lr": 2.081513262871972e-05}
{"train_loss": 0.07296313345432281, "global_step": 189730, "epoch": 2131, "lr": 2.0814661901192823e-05}
{"train_loss": 0.12550899386405945, "global_step": 189731, "epoch": 2131, "lr": 2.081419117758952e-05}
{"train_loss": 0.06385073065757751, "global_step": 189732, "epoch": 2131, "lr": 2.081372045790991e-05}
{"train_loss": 0.06306268274784088, "global_step": 189733, "epoch": 2131, "lr": 2.0813249742154022e-05}
{"train_loss": 0.08937667310237885, "global_step": 189734, "epoch": 2131, "lr": 2.0812779030321937e-05}
{"train_loss": 0.089337557554245, "global_step": 189735, "epoch": 2131, "lr": 2.0812308322413743e-05}
{"train_loss": 0.06436242908239365, "global_step": 189736, "epoch": 2131, "lr": 2.0811837618429452e-05}
{"train_loss": 0.08567768335342407, "global_step": 189737, "epoch": 2131, "lr": 2.0811366918369173e-05}
{"train_loss": 0.0570584274828434, "global_step": 189738, "epoch": 2131, "lr": 2.0810896222232934e-05}
{"train_loss": 0.08028888702392578, "global_step": 189739, "epoch": 2131, "lr": 2.0810425530020833e-05}
{"train_loss": 0.03022938407957554, "global_step": 189740, "epoch": 2131, "lr": 2.08099548417329e-05}
{"train_loss": 0.07037852704524994, "global_step": 189741, "epoch": 2131, "lr": 2.080948415736922e-05}
{"train_loss": 0.06439360231161118, "global_step": 189742, "epoch": 2131, "lr": 2.080901347692984e-05}
{"train_loss": 0.05906456336379051, "global_step": 189743, "epoch": 2131, "lr": 2.080854280041485e-05}
{"train_loss": 0.06447505950927734, "global_step": 189744, "epoch": 2131, "lr": 2.0808072127824273e-05}
{"train_loss": 0.040071822702884674, "global_step": 189745, "epoch": 2131, "lr": 2.0807601459158217e-05}
{"train_loss": 0.053276341408491135, "global_step": 189746, "epoch": 2131, "lr": 2.0807130794416702e-05}
{"train_loss": 0.07449090105136123, "global_step": 189747, "epoch": 2131, "lr": 2.0806660133599833e-05, "val_loss": 7.768547534942627}
{"train_loss": 0.0643419474363327, "global_step": 189748, "epoch": 2132, "lr": 2.080618947670765e-05}
{"train_loss": 0.07180914282798767, "global_step": 189749, "epoch": 2132, "lr": 2.08057188237402e-05}
{"train_loss": 0.05196956917643547, "global_step": 189750, "epoch": 2132, "lr": 2.0805248174697584e-05}
{"train_loss": 0.07283910363912582, "global_step": 189751, "epoch": 2132, "lr": 2.0804777529579833e-05}
{"train_loss": 0.11082474142313004, "global_step": 189752, "epoch": 2132, "lr": 2.0804306888387037e-05}
{"train_loss": 0.11153268814086914, "global_step": 189753, "epoch": 2132, "lr": 2.080383625111923e-05}
{"train_loss": 0.09786425530910492, "global_step": 189754, "epoch": 2132, "lr": 2.080336561777651e-05}
{"train_loss": 0.029925614595413208, "global_step": 189755, "epoch": 2132, "lr": 2.08028949883589e-05}
{"train_loss": 0.08281248062849045, "global_step": 189756, "epoch": 2132, "lr": 2.0802424362866507e-05}
{"train_loss": 0.0632595494389534, "global_step": 189757, "epoch": 2132, "lr": 2.0801953741299356e-05}
{"train_loss": 0.07872892171144485, "global_step": 189758, "epoch": 2132, "lr": 2.0801483123657538e-05}
{"train_loss": 0.06803622841835022, "global_step": 189759, "epoch": 2132, "lr": 2.080101250994109e-05}
{"train_loss": 0.06635389477014542, "global_step": 189760, "epoch": 2132, "lr": 2.080054190015011e-05}
{"train_loss": 0.09500172734260559, "global_step": 189761, "epoch": 2132, "lr": 2.0800071294284625e-05}
{"train_loss": 0.032342735677957535, "global_step": 189762, "epoch": 2132, "lr": 2.0799600692344727e-05}
{"train_loss": 0.07241767644882202, "global_step": 189763, "epoch": 2132, "lr": 2.079913009433045e-05}
{"train_loss": 0.06894946098327637, "global_step": 189764, "epoch": 2132, "lr": 2.0798659500241896e-05}
{"train_loss": 0.05832068994641304, "global_step": 189765, "epoch": 2132, "lr": 2.0798188910079087e-05}
{"train_loss": 0.0550139918923378, "global_step": 189766, "epoch": 2132, "lr": 2.079771832384212e-05}
{"train_loss": 0.07911714166402817, "global_step": 189767, "epoch": 2132, "lr": 2.079724774153103e-05}
{"train_loss": 0.10604441165924072, "global_step": 189768, "epoch": 2132, "lr": 2.079677716314591e-05}
{"train_loss": 0.05189354345202446, "global_step": 189769, "epoch": 2132, "lr": 2.0796306588686786e-05}
{"train_loss": 0.05982106178998947, "global_step": 189770, "epoch": 2132, "lr": 2.0795836018153746e-05}
{"train_loss": 0.08268888294696808, "global_step": 189771, "epoch": 2132, "lr": 2.0795365451546866e-05}
{"train_loss": 0.07088952511548996, "global_step": 189772, "epoch": 2132, "lr": 2.079489488886618e-05}
{"train_loss": 0.07751068472862244, "global_step": 189773, "epoch": 2132, "lr": 2.0794424330111776e-05}
{"train_loss": 0.07376854866743088, "global_step": 189774, "epoch": 2132, "lr": 2.0793953775283693e-05}
{"train_loss": 0.07312086969614029, "global_step": 189775, "epoch": 2132, "lr": 2.079348322438202e-05}
{"train_loss": 0.02120383456349373, "global_step": 189776, "epoch": 2132, "lr": 2.0793012677406792e-05}
{"train_loss": 0.05363697558641434, "global_step": 189777, "epoch": 2132, "lr": 2.0792542134358107e-05}
{"train_loss": 0.054367899894714355, "global_step": 189778, "epoch": 2132, "lr": 2.079207159523599e-05}
{"train_loss": 0.032716572284698486, "global_step": 189779, "epoch": 2132, "lr": 2.079160106004054e-05}
{"train_loss": 0.0670866072177887, "global_step": 189780, "epoch": 2132, "lr": 2.0791130528771786e-05}
{"train_loss": 0.03982812166213989, "global_step": 189781, "epoch": 2132, "lr": 2.0790660001429823e-05}
{"train_loss": 0.10555949062108994, "global_step": 189782, "epoch": 2132, "lr": 2.0790189478014688e-05}
{"train_loss": 0.06077691540122032, "global_step": 189783, "epoch": 2132, "lr": 2.0789718958526472e-05}
{"train_loss": 0.07771406322717667, "global_step": 189784, "epoch": 2132, "lr": 2.0789248442965203e-05}
{"train_loss": 0.0706896111369133, "global_step": 189785, "epoch": 2132, "lr": 2.0788777931330978e-05}
{"train_loss": 0.03186071291565895, "global_step": 189786, "epoch": 2132, "lr": 2.078830742362383e-05}
{"train_loss": 0.0756884515285492, "global_step": 189787, "epoch": 2132, "lr": 2.078783691984386e-05}
{"train_loss": 0.08879759162664413, "global_step": 189788, "epoch": 2132, "lr": 2.0787366419991106e-05}
{"train_loss": 0.0633336752653122, "global_step": 189789, "epoch": 2132, "lr": 2.078689592406562e-05}
{"train_loss": 0.0557992123067379, "global_step": 189790, "epoch": 2132, "lr": 2.078642543206749e-05}
{"train_loss": 0.05559905245900154, "global_step": 189791, "epoch": 2132, "lr": 2.078595494399676e-05}
{"train_loss": 0.04295285418629646, "global_step": 189792, "epoch": 2132, "lr": 2.0785484459853515e-05}
{"train_loss": 0.066294364631176, "global_step": 189793, "epoch": 2132, "lr": 2.0785013979637792e-05}
{"train_loss": 0.062321439385414124, "global_step": 189794, "epoch": 2132, "lr": 2.078454350334968e-05}
{"train_loss": 0.08209042996168137, "global_step": 189795, "epoch": 2132, "lr": 2.0784073030989215e-05}
{"train_loss": 0.037453778088092804, "global_step": 189796, "epoch": 2132, "lr": 2.0783602562556492e-05}
{"train_loss": 0.09629929065704346, "global_step": 189797, "epoch": 2132, "lr": 2.078313209805154e-05}
{"train_loss": 0.0414196252822876, "global_step": 189798, "epoch": 2132, "lr": 2.078266163747446e-05}
{"train_loss": 0.065558522939682, "global_step": 189799, "epoch": 2132, "lr": 2.0782191180825272e-05}
{"train_loss": 0.08224612474441528, "global_step": 189800, "epoch": 2132, "lr": 2.0781720728104086e-05}
{"train_loss": 0.036573465913534164, "global_step": 189801, "epoch": 2132, "lr": 2.078125027931092e-05}
{"train_loss": 0.055452313274145126, "global_step": 189802, "epoch": 2132, "lr": 2.078077983444588e-05}
{"train_loss": 0.10881171375513077, "global_step": 189803, "epoch": 2132, "lr": 2.0780309393508986e-05}
{"train_loss": 0.04222358018159866, "global_step": 189804, "epoch": 2132, "lr": 2.0779838956500342e-05}
{"train_loss": 0.05240580439567566, "global_step": 189805, "epoch": 2132, "lr": 2.077936852341998e-05}
{"train_loss": 0.08431612700223923, "global_step": 189806, "epoch": 2132, "lr": 2.0778898094267967e-05}
{"train_loss": 0.07517965137958527, "global_step": 189807, "epoch": 2132, "lr": 2.0778427669044398e-05}
{"train_loss": 0.07901949435472488, "global_step": 189808, "epoch": 2132, "lr": 2.0777957247749296e-05}
{"train_loss": 0.06398320943117142, "global_step": 189809, "epoch": 2132, "lr": 2.077748683038276e-05}
{"train_loss": 0.03970003500580788, "global_step": 189810, "epoch": 2132, "lr": 2.077701641694481e-05}
{"train_loss": 0.05769653990864754, "global_step": 189811, "epoch": 2132, "lr": 2.077654600743556e-05}
{"train_loss": 0.05632023140788078, "global_step": 189812, "epoch": 2132, "lr": 2.0776075601855026e-05}
{"train_loss": 0.06675716489553452, "global_step": 189813, "epoch": 2132, "lr": 2.0775605200203307e-05}
{"train_loss": 0.05290237441658974, "global_step": 189814, "epoch": 2132, "lr": 2.077513480248044e-05}
{"train_loss": 0.055514857172966, "global_step": 189815, "epoch": 2132, "lr": 2.0774664408686517e-05}
{"train_loss": 0.06317797303199768, "global_step": 189816, "epoch": 2132, "lr": 2.0774194018821562e-05}
{"train_loss": 0.060249947011470795, "global_step": 189817, "epoch": 2132, "lr": 2.0773723632885685e-05}
{"train_loss": 0.035282351076602936, "global_step": 189818, "epoch": 2132, "lr": 2.07732532508789e-05}
{"train_loss": 0.08015015721321106, "global_step": 189819, "epoch": 2132, "lr": 2.0772782872801312e-05}
{"train_loss": 0.09801220148801804, "global_step": 189820, "epoch": 2132, "lr": 2.0772312498652952e-05}
{"train_loss": 0.05805732309818268, "global_step": 189821, "epoch": 2132, "lr": 2.0771842128433917e-05}
{"train_loss": 0.06125789135694504, "global_step": 189822, "epoch": 2132, "lr": 2.077137176214423e-05}
{"train_loss": 0.04914651811122894, "global_step": 189823, "epoch": 2132, "lr": 2.0770901399784004e-05}
{"train_loss": 0.06606105715036392, "global_step": 189824, "epoch": 2132, "lr": 2.0770431041353245e-05}
{"train_loss": 0.0481279231607914, "global_step": 189825, "epoch": 2132, "lr": 2.0769960686852068e-05}
{"train_loss": 0.0831366628408432, "global_step": 189826, "epoch": 2132, "lr": 2.076949033628051e-05}
{"train_loss": 0.07020657509565353, "global_step": 189827, "epoch": 2132, "lr": 2.0769019989638627e-05}
{"train_loss": 0.08522067964076996, "global_step": 189828, "epoch": 2132, "lr": 2.0768549646926506e-05}
{"train_loss": 0.05481943488121033, "global_step": 189829, "epoch": 2132, "lr": 2.0768079308144177e-05}
{"train_loss": 0.07645275443792343, "global_step": 189830, "epoch": 2132, "lr": 2.0767608973291746e-05}
{"train_loss": 0.05132078751921654, "global_step": 189831, "epoch": 2132, "lr": 2.0767138642369232e-05}
{"train_loss": 0.040379367768764496, "global_step": 189832, "epoch": 2132, "lr": 2.076666831537674e-05}
{"train_loss": 0.06754134595394135, "global_step": 189833, "epoch": 2132, "lr": 2.0766197992314296e-05}
{"train_loss": 0.04936768859624863, "global_step": 189834, "epoch": 2132, "lr": 2.0765727673182e-05}
{"train_loss": 0.10139404237270355, "global_step": 189835, "epoch": 2132, "lr": 2.0765257357979873e-05}
{"train_loss": 0.06581304130259524, "global_step": 189836, "epoch": 2132, "lr": 2.0764787046708022e-05, "val_loss": 7.840641021728516}
{"train_loss": 0.040149059146642685, "global_step": 189837, "epoch": 2133, "lr": 2.0764316739366468e-05}
{"train_loss": 0.07510092109441757, "global_step": 189838, "epoch": 2133, "lr": 2.076384643595532e-05}
{"train_loss": 0.07012736052274704, "global_step": 189839, "epoch": 2133, "lr": 2.0763376136474588e-05}
{"train_loss": 0.08020943403244019, "global_step": 189840, "epoch": 2133, "lr": 2.076290584092439e-05}
{"train_loss": 0.05821213871240616, "global_step": 189841, "epoch": 2133, "lr": 2.0762435549304744e-05}
{"train_loss": 0.014255238696932793, "global_step": 189842, "epoch": 2133, "lr": 2.0761965261615727e-05}
{"train_loss": 0.06280697137117386, "global_step": 189843, "epoch": 2133, "lr": 2.076149497785743e-05}
{"train_loss": 0.06947948783636093, "global_step": 189844, "epoch": 2133, "lr": 2.0761024698029873e-05}
{"train_loss": 0.04676934331655502, "global_step": 189845, "epoch": 2133, "lr": 2.076055442213316e-05}
{"train_loss": 0.09275811165571213, "global_step": 189846, "epoch": 2133, "lr": 2.0760084150167318e-05}
{"train_loss": 0.035496585071086884, "global_step": 189847, "epoch": 2133, "lr": 2.075961388213244e-05}
{"train_loss": 0.10062636435031891, "global_step": 189848, "epoch": 2133, "lr": 2.075914361802856e-05}
{"train_loss": 0.029628552496433258, "global_step": 189849, "epoch": 2133, "lr": 2.0758673357855775e-05}
{"train_loss": 0.047674261033535004, "global_step": 189850, "epoch": 2133, "lr": 2.075820310161411e-05}
{"train_loss": 0.055317141115665436, "global_step": 189851, "epoch": 2133, "lr": 2.0757732849303667e-05}
{"train_loss": 0.034551165997982025, "global_step": 189852, "epoch": 2133, "lr": 2.075726260092447e-05}
{"train_loss": 0.11606785655021667, "global_step": 189853, "epoch": 2133, "lr": 2.0756792356476627e-05}
{"train_loss": 0.054171666502952576, "global_step": 189854, "epoch": 2133, "lr": 2.0756322115960152e-05}
{"train_loss": 0.11929386854171753, "global_step": 189855, "epoch": 2133, "lr": 2.0755851879375154e-05}
{"train_loss": 0.08878911286592484, "global_step": 189856, "epoch": 2133, "lr": 2.0755381646721662e-05}
{"train_loss": 0.0880466178059578, "global_step": 189857, "epoch": 2133, "lr": 2.0754911417999762e-05}
{"train_loss": 0.06588339805603027, "global_step": 189858, "epoch": 2133, "lr": 2.0754441193209494e-05}
{"train_loss": 0.0383559949696064, "global_step": 189859, "epoch": 2133, "lr": 2.0753970972350956e-05}
{"train_loss": 0.056402988731861115, "global_step": 189860, "epoch": 2133, "lr": 2.0753500755424166e-05}
{"train_loss": 0.09541533142328262, "global_step": 189861, "epoch": 2133, "lr": 2.0753030542429235e-05}
{"train_loss": 0.06962839514017105, "global_step": 189862, "epoch": 2133, "lr": 2.0752560333366177e-05}
{"train_loss": 0.15119792520999908, "global_step": 189863, "epoch": 2133, "lr": 2.0752090128235106e-05}
{"train_loss": 0.0756339505314827, "global_step": 189864, "epoch": 2133, "lr": 2.0751619927036058e-05}
{"train_loss": 0.055539000779390335, "global_step": 189865, "epoch": 2133, "lr": 2.0751149729769082e-05}
{"train_loss": 0.09602773934602737, "global_step": 189866, "epoch": 2133, "lr": 2.075067953643427e-05}
{"train_loss": 0.10257711261510849, "global_step": 189867, "epoch": 2133, "lr": 2.0750209347031656e-05}
{"train_loss": 0.0940055176615715, "global_step": 189868, "epoch": 2133, "lr": 2.0749739161561342e-05}
{"train_loss": 0.040402285754680634, "global_step": 189869, "epoch": 2133, "lr": 2.074926898002335e-05}
{"train_loss": 0.09417140483856201, "global_step": 189870, "epoch": 2133, "lr": 2.074879880241778e-05}
{"train_loss": 0.03816855326294899, "global_step": 189871, "epoch": 2133, "lr": 2.0748328628744658e-05}
{"train_loss": 0.043517906218767166, "global_step": 189872, "epoch": 2133, "lr": 2.074785845900408e-05}
{"train_loss": 0.06968630850315094, "global_step": 189873, "epoch": 2133, "lr": 2.0747388293196085e-05}
{"train_loss": 0.05485367029905319, "global_step": 189874, "epoch": 2133, "lr": 2.0746918131320765e-05}
{"train_loss": 0.10621590167284012, "global_step": 189875, "epoch": 2133, "lr": 2.074644797337814e-05}
{"train_loss": 0.13958360254764557, "global_step": 189876, "epoch": 2133, "lr": 2.0745977819368327e-05}
{"train_loss": 0.10275214910507202, "global_step": 189877, "epoch": 2133, "lr": 2.0745507669291336e-05}
{"train_loss": 0.054949354380369186, "global_step": 189878, "epoch": 2133, "lr": 2.074503752314726e-05}
{"train_loss": 0.07544177025556564, "global_step": 189879, "epoch": 2133, "lr": 2.074456738093617e-05}
{"train_loss": 0.026085656136274338, "global_step": 189880, "epoch": 2133, "lr": 2.07440972426581e-05}
{"train_loss": 0.08891254663467407, "global_step": 189881, "epoch": 2133, "lr": 2.0743627108313145e-05}
{"train_loss": 0.04407982900738716, "global_step": 189882, "epoch": 2133, "lr": 2.074315697790134e-05}
{"train_loss": 0.05782885104417801, "global_step": 189883, "epoch": 2133, "lr": 2.0742686851422776e-05}
{"train_loss": 0.052220556885004044, "global_step": 189884, "epoch": 2133, "lr": 2.0742216728877485e-05}
{"train_loss": 0.020655570551753044, "global_step": 189885, "epoch": 2133, "lr": 2.0741746610265562e-05}
{"train_loss": 0.060331810265779495, "global_step": 189886, "epoch": 2133, "lr": 2.0741276495587036e-05}
{"train_loss": 0.09689042717218399, "global_step": 189887, "epoch": 2133, "lr": 2.074080638484201e-05}
{"train_loss": 0.0789286270737648, "global_step": 189888, "epoch": 2133, "lr": 2.07403362780305e-05}
{"train_loss": 0.10372445732355118, "global_step": 189889, "epoch": 2133, "lr": 2.0739866175152622e-05}
{"train_loss": 0.09556938707828522, "global_step": 189890, "epoch": 2133, "lr": 2.0739396076208394e-05}
{"train_loss": 0.024837981909513474, "global_step": 189891, "epoch": 2133, "lr": 2.0738925981197908e-05}
{"train_loss": 0.0771670714020729, "global_step": 189892, "epoch": 2133, "lr": 2.0738455890121207e-05}
{"train_loss": 0.07412664592266083, "global_step": 189893, "epoch": 2133, "lr": 2.0737985802978377e-05}
{"train_loss": 0.0642842948436737, "global_step": 189894, "epoch": 2133, "lr": 2.0737515719769456e-05}
{"train_loss": 0.08874986320734024, "global_step": 189895, "epoch": 2133, "lr": 2.0737045640494535e-05}
{"train_loss": 0.03919633850455284, "global_step": 189896, "epoch": 2133, "lr": 2.073657556515364e-05}
{"train_loss": 0.0975629910826683, "global_step": 189897, "epoch": 2133, "lr": 2.0736105493746855e-05}
{"train_loss": 0.04857335612177849, "global_step": 189898, "epoch": 2133, "lr": 2.0735635426274265e-05}
{"train_loss": 0.05341333895921707, "global_step": 189899, "epoch": 2133, "lr": 2.0735165362735915e-05}
{"train_loss": 0.06488333642482758, "global_step": 189900, "epoch": 2133, "lr": 2.073469530313184e-05}
{"train_loss": 0.10686897486448288, "global_step": 189901, "epoch": 2133, "lr": 2.0734225247462153e-05}
{"train_loss": 0.03680029883980751, "global_step": 189902, "epoch": 2133, "lr": 2.0733755195726873e-05}
{"train_loss": 0.07610973715782166, "global_step": 189903, "epoch": 2133, "lr": 2.07332851479261e-05}
{"train_loss": 0.03241800144314766, "global_step": 189904, "epoch": 2133, "lr": 2.0732815104059878e-05}
{"train_loss": 0.07169217616319656, "global_step": 189905, "epoch": 2133, "lr": 2.0732345064128256e-05}
{"train_loss": 0.07872848957777023, "global_step": 189906, "epoch": 2133, "lr": 2.0731875028131335e-05}
{"train_loss": 0.08128152787685394, "global_step": 189907, "epoch": 2133, "lr": 2.0731404996069137e-05}
{"train_loss": 0.02843492664396763, "global_step": 189908, "epoch": 2133, "lr": 2.0730934967941762e-05}
{"train_loss": 0.04088055342435837, "global_step": 189909, "epoch": 2133, "lr": 2.0730464943749235e-05}
{"train_loss": 0.030744286254048347, "global_step": 189910, "epoch": 2133, "lr": 2.0729994923491664e-05}
{"train_loss": 0.06690509617328644, "global_step": 189911, "epoch": 2133, "lr": 2.0729524907169062e-05}
{"train_loss": 0.05749416723847389, "global_step": 189912, "epoch": 2133, "lr": 2.0729054894781545e-05}
{"train_loss": 0.05637522041797638, "global_step": 189913, "epoch": 2133, "lr": 2.0728584886329126e-05}
{"train_loss": 0.06675566732883453, "global_step": 189914, "epoch": 2133, "lr": 2.0728114881811894e-05}
{"train_loss": 0.10157286375761032, "global_step": 189915, "epoch": 2133, "lr": 2.072764488122993e-05}
{"train_loss": 0.06214073672890663, "global_step": 189916, "epoch": 2133, "lr": 2.0727174884583262e-05}
{"train_loss": 0.07996977120637894, "global_step": 189917, "epoch": 2133, "lr": 2.0726704891871978e-05}
{"train_loss": 0.04280458390712738, "global_step": 189918, "epoch": 2133, "lr": 2.072623490309612e-05}
{"train_loss": 0.08346715569496155, "global_step": 189919, "epoch": 2133, "lr": 2.072576491825578e-05}
{"train_loss": 0.09124691784381866, "global_step": 189920, "epoch": 2133, "lr": 2.0725294937350986e-05}
{"train_loss": 0.03482959419488907, "global_step": 189921, "epoch": 2133, "lr": 2.072482496038184e-05}
{"train_loss": 0.08821380138397217, "global_step": 189922, "epoch": 2133, "lr": 2.072435498734836e-05}
{"train_loss": 0.05161042883992195, "global_step": 189923, "epoch": 2133, "lr": 2.0723885018250655e-05}
{"train_loss": 0.04491624981164932, "global_step": 189924, "epoch": 2133, "lr": 2.0723415053088752e-05}
{"train_loss": 0.06775513253687473, "global_step": 189925, "epoch": 2133, "lr": 2.0722945091862743e-05, "val_loss": 7.839823246002197}
{"train_loss": 0.08313727378845215, "global_step": 189926, "epoch": 2134, "lr": 2.072247513457266e-05}
{"train_loss": 0.11520200222730637, "global_step": 189927, "epoch": 2134, "lr": 2.0722005181218602e-05}
{"train_loss": 0.06840933859348297, "global_step": 189928, "epoch": 2134, "lr": 2.0721535231800592e-05}
{"train_loss": 0.03970909118652344, "global_step": 189929, "epoch": 2134, "lr": 2.0721065286318737e-05}
{"train_loss": 0.11890596896409988, "global_step": 189930, "epoch": 2134, "lr": 2.0720595344773058e-05}
{"train_loss": 0.04142674058675766, "global_step": 189931, "epoch": 2134, "lr": 2.072012540716366e-05}
{"train_loss": 0.06834337115287781, "global_step": 189932, "epoch": 2134, "lr": 2.071965547349056e-05}
{"train_loss": 0.09998498857021332, "global_step": 189933, "epoch": 2134, "lr": 2.071918554375385e-05}
{"train_loss": 0.09794074296951294, "global_step": 189934, "epoch": 2134, "lr": 2.0718715617953605e-05}
{"train_loss": 0.05716459080576897, "global_step": 189935, "epoch": 2134, "lr": 2.0718245696089854e-05}
{"train_loss": 0.0227975957095623, "global_step": 189936, "epoch": 2134, "lr": 2.0717775778162696e-05}
{"train_loss": 0.07493467628955841, "global_step": 189937, "epoch": 2134, "lr": 2.0717305864172177e-05}
{"train_loss": 0.08040996640920639, "global_step": 189938, "epoch": 2134, "lr": 2.0716835954118343e-05}
{"train_loss": 0.08940037339925766, "global_step": 189939, "epoch": 2134, "lr": 2.0716366048001283e-05}
{"train_loss": 0.08322035521268845, "global_step": 189940, "epoch": 2134, "lr": 2.071589614582104e-05}
{"train_loss": 0.08650767803192139, "global_step": 189941, "epoch": 2134, "lr": 2.07154262475777e-05}
{"train_loss": 0.08121580630540848, "global_step": 189942, "epoch": 2134, "lr": 2.071495635327132e-05}
{"train_loss": 0.12564285099506378, "global_step": 189943, "epoch": 2134, "lr": 2.071448646290194e-05}
{"train_loss": 0.04967167228460312, "global_step": 189944, "epoch": 2134, "lr": 2.0714016576469657e-05}
{"train_loss": 0.07656390964984894, "global_step": 189945, "epoch": 2134, "lr": 2.0713546693974493e-05}
{"train_loss": 0.07204948365688324, "global_step": 189946, "epoch": 2134, "lr": 2.0713076815416566e-05}
{"train_loss": 0.06127024441957474, "global_step": 189947, "epoch": 2134, "lr": 2.071260694079588e-05}
{"train_loss": 0.07380479574203491, "global_step": 189948, "epoch": 2134, "lr": 2.071213707011255e-05}
{"train_loss": 0.11510472744703293, "global_step": 189949, "epoch": 2134, "lr": 2.0711667203366597e-05}
{"train_loss": 0.05656632408499718, "global_step": 189950, "epoch": 2134, "lr": 2.07111973405581e-05}
{"train_loss": 0.08779238909482956, "global_step": 189951, "epoch": 2134, "lr": 2.0710727481687154e-05}
{"train_loss": 0.08991676568984985, "global_step": 189952, "epoch": 2134, "lr": 2.0710257626753765e-05}
{"train_loss": 0.05367794260382652, "global_step": 189953, "epoch": 2134, "lr": 2.070978777575805e-05}
{"train_loss": 0.054344192147254944, "global_step": 189954, "epoch": 2134, "lr": 2.0709317928700023e-05}
{"train_loss": 0.0762905552983284, "global_step": 189955, "epoch": 2134, "lr": 2.070884808557979e-05}
{"train_loss": 0.04276253655552864, "global_step": 189956, "epoch": 2134, "lr": 2.070837824639738e-05}
{"train_loss": 0.06146671250462532, "global_step": 189957, "epoch": 2134, "lr": 2.0707908411152887e-05}
{"train_loss": 0.04147186875343323, "global_step": 189958, "epoch": 2134, "lr": 2.0707438579846343e-05}
{"train_loss": 0.12722148001194, "global_step": 189959, "epoch": 2134, "lr": 2.0706968752477845e-05}
{"train_loss": 0.03953945264220238, "global_step": 189960, "epoch": 2134, "lr": 2.0706498929047417e-05}
{"train_loss": 0.13536401093006134, "global_step": 189961, "epoch": 2134, "lr": 2.070602910955517e-05}
{"train_loss": 0.12787003815174103, "global_step": 189962, "epoch": 2134, "lr": 2.0705559294001116e-05}
{"train_loss": 0.036546021699905396, "global_step": 189963, "epoch": 2134, "lr": 2.0705089482385358e-05}
{"train_loss": 0.09215576946735382, "global_step": 189964, "epoch": 2134, "lr": 2.0704619674707926e-05}
{"train_loss": 0.07517927139997482, "global_step": 189965, "epoch": 2134, "lr": 2.0704149870968924e-05}
{"train_loss": 0.08326108008623123, "global_step": 189966, "epoch": 2134, "lr": 2.0703680071168367e-05}
{"train_loss": 0.03236853703856468, "global_step": 189967, "epoch": 2134, "lr": 2.0703210275306368e-05}
{"train_loss": 0.020252056419849396, "global_step": 189968, "epoch": 2134, "lr": 2.0702740483382943e-05}
{"train_loss": 0.011027313768863678, "global_step": 189969, "epoch": 2134, "lr": 2.070227069539818e-05}
{"train_loss": 0.06802386790513992, "global_step": 189970, "epoch": 2134, "lr": 2.0701800911352152e-05}
{"train_loss": 0.04166584089398384, "global_step": 189971, "epoch": 2134, "lr": 2.0701331131244898e-05}
{"train_loss": 0.03706316649913788, "global_step": 189972, "epoch": 2134, "lr": 2.0700861355076513e-05}
{"train_loss": 0.06360849738121033, "global_step": 189973, "epoch": 2134, "lr": 2.0700391582847018e-05}
{"train_loss": 0.07770141214132309, "global_step": 189974, "epoch": 2134, "lr": 2.0699921814556517e-05}
{"train_loss": 0.05214957892894745, "global_step": 189975, "epoch": 2134, "lr": 2.0699452050205055e-05}
{"train_loss": 0.06802457571029663, "global_step": 189976, "epoch": 2134, "lr": 2.0698982289792672e-05}
{"train_loss": 0.037013158202171326, "global_step": 189977, "epoch": 2134, "lr": 2.0698512533319477e-05}
{"train_loss": 0.07132997363805771, "global_step": 189978, "epoch": 2134, "lr": 2.0698042780785486e-05}
{"train_loss": 0.09583133459091187, "global_step": 189979, "epoch": 2134, "lr": 2.069757303219081e-05}
{"train_loss": 0.05094734579324722, "global_step": 189980, "epoch": 2134, "lr": 2.0697103287535485e-05}
{"train_loss": 0.07875695079565048, "global_step": 189981, "epoch": 2134, "lr": 2.069663354681956e-05}
{"train_loss": 0.06396737694740295, "global_step": 189982, "epoch": 2134, "lr": 2.0696163810043136e-05}
{"train_loss": 0.03447400778532028, "global_step": 189983, "epoch": 2134, "lr": 2.0695694077206236e-05}
{"train_loss": 0.08603017032146454, "global_step": 189984, "epoch": 2134, "lr": 2.069522434830896e-05}
{"train_loss": 0.08542238920927048, "global_step": 189985, "epoch": 2134, "lr": 2.0694754623351337e-05}
{"train_loss": 0.05128586292266846, "global_step": 189986, "epoch": 2134, "lr": 2.069428490233346e-05}
{"train_loss": 0.040457770228385925, "global_step": 189987, "epoch": 2134, "lr": 2.0693815185255365e-05}
{"train_loss": 0.05745929479598999, "global_step": 189988, "epoch": 2134, "lr": 2.069334547211713e-05}
{"train_loss": 0.119239442050457, "global_step": 189989, "epoch": 2134, "lr": 2.069287576291884e-05}
{"train_loss": 0.03513725847005844, "global_step": 189990, "epoch": 2134, "lr": 2.0692406057660513e-05}
{"train_loss": 0.04629731550812721, "global_step": 189991, "epoch": 2134, "lr": 2.069193635634225e-05}
{"train_loss": 0.07812165468931198, "global_step": 189992, "epoch": 2134, "lr": 2.0691466658964088e-05}
{"train_loss": 0.08426953852176666, "global_step": 189993, "epoch": 2134, "lr": 2.069099696552611e-05}
{"train_loss": 0.0631430596113205, "global_step": 189994, "epoch": 2134, "lr": 2.069052727602836e-05}
{"train_loss": 0.04558304324746132, "global_step": 189995, "epoch": 2134, "lr": 2.0690057590470922e-05}
{"train_loss": 0.08685123175382614, "global_step": 189996, "epoch": 2134, "lr": 2.0689587908853837e-05}
{"train_loss": 0.08082296699285507, "global_step": 189997, "epoch": 2134, "lr": 2.0689118231177196e-05}
{"train_loss": 0.04557649418711662, "global_step": 189998, "epoch": 2134, "lr": 2.0688648557441027e-05}
{"train_loss": 0.037141259759664536, "global_step": 189999, "epoch": 2134, "lr": 2.0688178887645432e-05}
{"train_loss": 0.06159406155347824, "global_step": 190000, "epoch": 2134, "lr": 2.0687709221790437e-05}
{"train_loss": 0.16225184500217438, "global_step": 190001, "epoch": 2134, "lr": 2.068723955987613e-05}
{"train_loss": 0.09935009479522705, "global_step": 190002, "epoch": 2134, "lr": 2.0686769901902557e-05}
{"train_loss": 0.08634484559297562, "global_step": 190003, "epoch": 2134, "lr": 2.068630024786981e-05}
{"train_loss": 0.07002265006303787, "global_step": 190004, "epoch": 2134, "lr": 2.068583059777791e-05}
{"train_loss": 0.09719731658697128, "global_step": 190005, "epoch": 2134, "lr": 2.0685360951626948e-05}
{"train_loss": 0.043704453855752945, "global_step": 190006, "epoch": 2134, "lr": 2.0684891309417e-05}
{"train_loss": 0.10096190869808197, "global_step": 190007, "epoch": 2134, "lr": 2.0684421671148086e-05}
{"train_loss": 0.07197007536888123, "global_step": 190008, "epoch": 2134, "lr": 2.0683952036820315e-05}
{"train_loss": 0.024404732510447502, "global_step": 190009, "epoch": 2134, "lr": 2.0683482406433713e-05}
{"train_loss": 0.1541987806558609, "global_step": 190010, "epoch": 2134, "lr": 2.0683012779988376e-05}
{"train_loss": 0.03492553159594536, "global_step": 190011, "epoch": 2134, "lr": 2.0682543157484335e-05}
{"train_loss": 0.10793456435203552, "global_step": 190012, "epoch": 2134, "lr": 2.0682073538921686e-05}
{"train_loss": 0.09175825119018555, "global_step": 190013, "epoch": 2134, "lr": 2.068160392430045e-05}
{"train_loss": 0.07161572803690862, "global_step": 190014, "epoch": 2134, "lr": 2.0681134313620743e-05, "val_loss": 7.893154144287109}
{"train_loss": 0.04256003722548485, "global_step": 190015, "epoch": 2135, "lr": 2.0680664706882596e-05}
{"train_loss": 0.0603693462908268, "global_step": 190016, "epoch": 2135, "lr": 2.068019510408606e-05}
{"train_loss": 0.09246966242790222, "global_step": 190017, "epoch": 2135, "lr": 2.067972550523123e-05}
{"train_loss": 0.0757816955447197, "global_step": 190018, "epoch": 2135, "lr": 2.0679255910318134e-05}
{"train_loss": 0.04069044813513756, "global_step": 190019, "epoch": 2135, "lr": 2.0678786319346875e-05}
{"train_loss": 0.06673633307218552, "global_step": 190020, "epoch": 2135, "lr": 2.06783167323175e-05}
{"train_loss": 0.051514700055122375, "global_step": 190021, "epoch": 2135, "lr": 2.0677847149230047e-05}
{"train_loss": 0.06128471717238426, "global_step": 190022, "epoch": 2135, "lr": 2.0677377570084615e-05}
{"train_loss": 0.07059969753026962, "global_step": 190023, "epoch": 2135, "lr": 2.0676907994881235e-05}
{"train_loss": 0.06730201095342636, "global_step": 190024, "epoch": 2135, "lr": 2.0676438423619993e-05}
{"train_loss": 0.04702359810471535, "global_step": 190025, "epoch": 2135, "lr": 2.067596885630096e-05}
{"train_loss": 0.03947342559695244, "global_step": 190026, "epoch": 2135, "lr": 2.0675499292924166e-05}
{"train_loss": 0.06915668398141861, "global_step": 190027, "epoch": 2135, "lr": 2.067502973348971e-05}
{"train_loss": 0.06415870785713196, "global_step": 190028, "epoch": 2135, "lr": 2.0674560177997625e-05}
{"train_loss": 0.037749290466308594, "global_step": 190029, "epoch": 2135, "lr": 2.0674090626448006e-05}
{"train_loss": 0.0693981721997261, "global_step": 190030, "epoch": 2135, "lr": 2.0673621078840877e-05}
{"train_loss": 0.027026426047086716, "global_step": 190031, "epoch": 2135, "lr": 2.0673151535176344e-05}
{"train_loss": 0.08951780200004578, "global_step": 190032, "epoch": 2135, "lr": 2.067268199545443e-05}
{"train_loss": 0.07210339605808258, "global_step": 190033, "epoch": 2135, "lr": 2.0672212459675234e-05}
{"train_loss": 0.0934310257434845, "global_step": 190034, "epoch": 2135, "lr": 2.0671742927838785e-05}
{"train_loss": 0.039871327579021454, "global_step": 190035, "epoch": 2135, "lr": 2.0671273399945172e-05}
{"train_loss": 0.0683048665523529, "global_step": 190036, "epoch": 2135, "lr": 2.0670803875994437e-05}
{"train_loss": 0.09431633353233337, "global_step": 190037, "epoch": 2135, "lr": 2.0670334355986674e-05}
{"train_loss": 0.04032491147518158, "global_step": 190038, "epoch": 2135, "lr": 2.066986483992191e-05}
{"train_loss": 0.07490932941436768, "global_step": 190039, "epoch": 2135, "lr": 2.0669395327800238e-05}
{"train_loss": 0.11760357022285461, "global_step": 190040, "epoch": 2135, "lr": 2.0668925819621697e-05}
{"train_loss": 0.04881497099995613, "global_step": 190041, "epoch": 2135, "lr": 2.066845631538636e-05}
{"train_loss": 0.029524175450205803, "global_step": 190042, "epoch": 2135, "lr": 2.0667986815094304e-05}
{"train_loss": 0.02476540021598339, "global_step": 190043, "epoch": 2135, "lr": 2.0667517318745567e-05}
{"train_loss": 0.08608721941709518, "global_step": 190044, "epoch": 2135, "lr": 2.066704782634024e-05}
{"train_loss": 0.04976871982216835, "global_step": 190045, "epoch": 2135, "lr": 2.0666578337878356e-05}
{"train_loss": 0.08287856727838516, "global_step": 190046, "epoch": 2135, "lr": 2.0666108853360006e-05}
{"train_loss": 0.04664238914847374, "global_step": 190047, "epoch": 2135, "lr": 2.066563937278523e-05}
{"train_loss": 0.05284102261066437, "global_step": 190048, "epoch": 2135, "lr": 2.0665169896154114e-05}
{"train_loss": 0.0679657831788063, "global_step": 190049, "epoch": 2135, "lr": 2.0664700423466687e-05}
{"train_loss": 0.03024446777999401, "global_step": 190050, "epoch": 2135, "lr": 2.0664230954723056e-05}
{"train_loss": 0.08731313794851303, "global_step": 190051, "epoch": 2135, "lr": 2.0663761489923244e-05}
{"train_loss": 0.06473160535097122, "global_step": 190052, "epoch": 2135, "lr": 2.066329202906735e-05}
{"train_loss": 0.06689012795686722, "global_step": 190053, "epoch": 2135, "lr": 2.0662822572155416e-05}
{"train_loss": 0.09007548540830612, "global_step": 190054, "epoch": 2135, "lr": 2.066235311918749e-05}
{"train_loss": 0.02196592092514038, "global_step": 190055, "epoch": 2135, "lr": 2.0661883670163672e-05}
{"train_loss": 0.05568431690335274, "global_step": 190056, "epoch": 2135, "lr": 2.0661414225083993e-05}
{"train_loss": 0.0733812004327774, "global_step": 190057, "epoch": 2135, "lr": 2.0660944783948536e-05}
{"train_loss": 0.09351304918527603, "global_step": 190058, "epoch": 2135, "lr": 2.066047534675737e-05}
{"train_loss": 0.05965263396501541, "global_step": 190059, "epoch": 2135, "lr": 2.066000591351052e-05}
{"train_loss": 0.09515786170959473, "global_step": 190060, "epoch": 2135, "lr": 2.0659536484208074e-05}
{"train_loss": 0.031278178095817566, "global_step": 190061, "epoch": 2135, "lr": 2.0659067058850118e-05}
{"train_loss": 0.06279964745044708, "global_step": 190062, "epoch": 2135, "lr": 2.065859763743667e-05}
{"train_loss": 0.06235384941101074, "global_step": 190063, "epoch": 2135, "lr": 2.0658128219967837e-05}
{"train_loss": 0.06813772022724152, "global_step": 190064, "epoch": 2135, "lr": 2.0657658806443646e-05}
{"train_loss": 0.07208514213562012, "global_step": 190065, "epoch": 2135, "lr": 2.065718939686419e-05}
{"train_loss": 0.042632464319467545, "global_step": 190066, "epoch": 2135, "lr": 2.0656719991229502e-05}
{"train_loss": 0.08199124038219452, "global_step": 190067, "epoch": 2135, "lr": 2.0656250589539672e-05}
{"train_loss": 0.13786067068576813, "global_step": 190068, "epoch": 2135, "lr": 2.065578119179474e-05}
{"train_loss": 0.08898638188838959, "global_step": 190069, "epoch": 2135, "lr": 2.0655311797994793e-05}
{"train_loss": 0.06918863207101822, "global_step": 190070, "epoch": 2135, "lr": 2.0654842408139867e-05}
{"train_loss": 0.1338610053062439, "global_step": 190071, "epoch": 2135, "lr": 2.0654373022230057e-05}
{"train_loss": 0.0621102936565876, "global_step": 190072, "epoch": 2135, "lr": 2.065390364026539e-05}
{"train_loss": 0.04653621464967728, "global_step": 190073, "epoch": 2135, "lr": 2.065343426224597e-05}
{"train_loss": 0.0896657407283783, "global_step": 190074, "epoch": 2135, "lr": 2.065296488817181e-05}
{"train_loss": 0.07126890867948532, "global_step": 190075, "epoch": 2135, "lr": 2.065249551804303e-05}
{"train_loss": 0.054090291261672974, "global_step": 190076, "epoch": 2135, "lr": 2.0652026151859644e-05}
{"train_loss": 0.06431771069765091, "global_step": 190077, "epoch": 2135, "lr": 2.0651556789621735e-05}
{"train_loss": 0.03800510615110397, "global_step": 190078, "epoch": 2135, "lr": 2.065108743132938e-05}
{"train_loss": 0.08397470414638519, "global_step": 190079, "epoch": 2135, "lr": 2.0650618076982614e-05}
{"train_loss": 0.08966346830129623, "global_step": 190080, "epoch": 2135, "lr": 2.065014872658153e-05}
{"train_loss": 0.1538621038198471, "global_step": 190081, "epoch": 2135, "lr": 2.0649679380126163e-05}
{"train_loss": 0.03164391964673996, "global_step": 190082, "epoch": 2135, "lr": 2.06492100376166e-05}
{"train_loss": 0.1120157390832901, "global_step": 190083, "epoch": 2135, "lr": 2.0648740699052883e-05}
{"train_loss": 0.13272902369499207, "global_step": 190084, "epoch": 2135, "lr": 2.06482713644351e-05}
{"train_loss": 0.05980350449681282, "global_step": 190085, "epoch": 2135, "lr": 2.064780203376328e-05}
{"train_loss": 0.04679180681705475, "global_step": 190086, "epoch": 2135, "lr": 2.0647332707037526e-05}
{"train_loss": 0.049308646470308304, "global_step": 190087, "epoch": 2135, "lr": 2.0646863384257858e-05}
{"train_loss": 0.13470885157585144, "global_step": 190088, "epoch": 2135, "lr": 2.064639406542438e-05}
{"train_loss": 0.07633081823587418, "global_step": 190089, "epoch": 2135, "lr": 2.064592475053712e-05}
{"train_loss": 0.08268903940916061, "global_step": 190090, "epoch": 2135, "lr": 2.064545543959618e-05}
{"train_loss": 0.1321273297071457, "global_step": 190091, "epoch": 2135, "lr": 2.064498613260159e-05}
{"train_loss": 0.08491625636816025, "global_step": 190092, "epoch": 2135, "lr": 2.0644516829553412e-05}
{"train_loss": 0.0559317022562027, "global_step": 190093, "epoch": 2135, "lr": 2.0644047530451743e-05}
{"train_loss": 0.04976828396320343, "global_step": 190094, "epoch": 2135, "lr": 2.06435782352966e-05}
{"train_loss": 0.060734011232852936, "global_step": 190095, "epoch": 2135, "lr": 2.064310894408809e-05}
{"train_loss": 0.05866698548197746, "global_step": 190096, "epoch": 2135, "lr": 2.0642639656826242e-05}
{"train_loss": 0.10973270237445831, "global_step": 190097, "epoch": 2135, "lr": 2.0642170373511143e-05}
{"train_loss": 0.038883890956640244, "global_step": 190098, "epoch": 2135, "lr": 2.0641701094142835e-05}
{"train_loss": 0.049683310091495514, "global_step": 190099, "epoch": 2135, "lr": 2.0641231818721406e-05}
{"train_loss": 0.13600628077983856, "global_step": 190100, "epoch": 2135, "lr": 2.0640762547246896e-05}
{"train_loss": 0.07379729300737381, "global_step": 190101, "epoch": 2135, "lr": 2.0640293279719387e-05}
{"train_loss": 0.08245157450437546, "global_step": 190102, "epoch": 2135, "lr": 2.0639824016138916e-05}
{"train_loss": 0.0702582582454668, "global_step": 190103, "epoch": 2135, "lr": 2.0639354756505585e-05, "val_loss": 7.8221635818481445, "train_action_mse_error": 17.207124710083008}
{"train_loss": 0.07132123410701752, "global_step": 190104, "epoch": 2136, "lr": 2.063888550081941e-05}
{"train_loss": 0.08084496855735779, "global_step": 190105, "epoch": 2136, "lr": 2.0638416249080507e-05}
{"train_loss": 0.09268558025360107, "global_step": 190106, "epoch": 2136, "lr": 2.0637947001288882e-05}
{"train_loss": 0.13244135677814484, "global_step": 190107, "epoch": 2136, "lr": 2.063747775744465e-05}
{"train_loss": 0.06983939558267593, "global_step": 190108, "epoch": 2136, "lr": 2.0637008517547835e-05}
{"train_loss": 0.08163092285394669, "global_step": 190109, "epoch": 2136, "lr": 2.0636539281598534e-05}
{"train_loss": 0.08002059906721115, "global_step": 190110, "epoch": 2136, "lr": 2.0636070049596773e-05}
{"train_loss": 0.12815773487091064, "global_step": 190111, "epoch": 2136, "lr": 2.063560082154265e-05}
{"train_loss": 0.09539288282394409, "global_step": 190112, "epoch": 2136, "lr": 2.0635131597436204e-05}
{"train_loss": 0.031098969280719757, "global_step": 190113, "epoch": 2136, "lr": 2.0634662377277514e-05}
{"train_loss": 0.08432123810052872, "global_step": 190114, "epoch": 2136, "lr": 2.0634193161066622e-05}
{"train_loss": 0.05832923948764801, "global_step": 190115, "epoch": 2136, "lr": 2.063372394880361e-05}
{"train_loss": 0.06459495425224304, "global_step": 190116, "epoch": 2136, "lr": 2.0633254740488544e-05}
{"train_loss": 0.04642836004495621, "global_step": 190117, "epoch": 2136, "lr": 2.0632785536121463e-05}
{"train_loss": 0.04056974872946739, "global_step": 190118, "epoch": 2136, "lr": 2.0632316335702467e-05}
{"train_loss": 0.13462966680526733, "global_step": 190119, "epoch": 2136, "lr": 2.063184713923158e-05}
{"train_loss": 0.03830332309007645, "global_step": 190120, "epoch": 2136, "lr": 2.0631377946708902e-05}
{"train_loss": 0.12373131513595581, "global_step": 190121, "epoch": 2136, "lr": 2.063090875813446e-05}
{"train_loss": 0.06460070610046387, "global_step": 190122, "epoch": 2136, "lr": 2.0630439573508348e-05}
{"train_loss": 0.027517497539520264, "global_step": 190123, "epoch": 2136, "lr": 2.06299703928306e-05}
{"train_loss": 0.0462174154818058, "global_step": 190124, "epoch": 2136, "lr": 2.062950121610131e-05}
{"train_loss": 0.03999392315745354, "global_step": 190125, "epoch": 2136, "lr": 2.0629032043320514e-05}
{"train_loss": 0.0728989839553833, "global_step": 190126, "epoch": 2136, "lr": 2.06285628744883e-05}
{"train_loss": 0.06942128390073776, "global_step": 190127, "epoch": 2136, "lr": 2.0628093709604702e-05}
{"train_loss": 0.05761948227882385, "global_step": 190128, "epoch": 2136, "lr": 2.062762454866981e-05}
{"train_loss": 0.040261074900627136, "global_step": 190129, "epoch": 2136, "lr": 2.0627155391683666e-05}
{"train_loss": 0.11101645231246948, "global_step": 190130, "epoch": 2136, "lr": 2.0626686238646358e-05}
{"train_loss": 0.10593628138303757, "global_step": 190131, "epoch": 2136, "lr": 2.062621708955793e-05}
{"train_loss": 0.07116016745567322, "global_step": 190132, "epoch": 2136, "lr": 2.062574794441844e-05}
{"train_loss": 0.07635676860809326, "global_step": 190133, "epoch": 2136, "lr": 2.062527880322797e-05}
{"train_loss": 0.09811776876449585, "global_step": 190134, "epoch": 2136, "lr": 2.062480966598656e-05}
{"train_loss": 0.04437095671892166, "global_step": 190135, "epoch": 2136, "lr": 2.0624340532694302e-05}
{"train_loss": 0.032399773597717285, "global_step": 190136, "epoch": 2136, "lr": 2.0623871403351226e-05}
{"train_loss": 0.03648724406957626, "global_step": 190137, "epoch": 2136, "lr": 2.062340227795743e-05}
{"train_loss": 0.10128823667764664, "global_step": 190138, "epoch": 2136, "lr": 2.062293315651294e-05}
{"train_loss": 0.09293944388628006, "global_step": 190139, "epoch": 2136, "lr": 2.0622464039017864e-05}
{"train_loss": 0.06904254853725433, "global_step": 190140, "epoch": 2136, "lr": 2.0621994925472217e-05}
{"train_loss": 0.0711708664894104, "global_step": 190141, "epoch": 2136, "lr": 2.06215258158761e-05}
{"train_loss": 0.046782948076725006, "global_step": 190142, "epoch": 2136, "lr": 2.0621056710229548e-05}
{"train_loss": 0.05729418247938156, "global_step": 190143, "epoch": 2136, "lr": 2.062058760853265e-05}
{"train_loss": 0.1039571613073349, "global_step": 190144, "epoch": 2136, "lr": 2.0620118510785436e-05}
{"train_loss": 0.06036591902375221, "global_step": 190145, "epoch": 2136, "lr": 2.0619649416988014e-05}
{"train_loss": 0.06213648244738579, "global_step": 190146, "epoch": 2136, "lr": 2.0619180327140397e-05}
{"train_loss": 0.0641830712556839, "global_step": 190147, "epoch": 2136, "lr": 2.06187112412427e-05}
{"train_loss": 0.08776716142892838, "global_step": 190148, "epoch": 2136, "lr": 2.061824215929493e-05}
{"train_loss": 0.02626919187605381, "global_step": 190149, "epoch": 2136, "lr": 2.061777308129721e-05}
{"train_loss": 0.04739808663725853, "global_step": 190150, "epoch": 2136, "lr": 2.0617304007249542e-05}
{"train_loss": 0.06018081307411194, "global_step": 190151, "epoch": 2136, "lr": 2.0616834937152023e-05}
{"train_loss": 0.1151750311255455, "global_step": 190152, "epoch": 2136, "lr": 2.0616365871004734e-05}
{"train_loss": 0.0957866758108139, "global_step": 190153, "epoch": 2136, "lr": 2.0615896808807698e-05}
{"train_loss": 0.06551503390073776, "global_step": 190154, "epoch": 2136, "lr": 2.061542775056101e-05}
{"train_loss": 0.05667322501540184, "global_step": 190155, "epoch": 2136, "lr": 2.0614958696264708e-05}
{"train_loss": 0.046233080327510834, "global_step": 190156, "epoch": 2136, "lr": 2.0614489645918883e-05}
{"train_loss": 0.052175771445035934, "global_step": 190157, "epoch": 2136, "lr": 2.0614020599523558e-05}
{"train_loss": 0.023059777915477753, "global_step": 190158, "epoch": 2136, "lr": 2.0613551557078846e-05}
{"train_loss": 0.09792576730251312, "global_step": 190159, "epoch": 2136, "lr": 2.061308251858476e-05}
{"train_loss": 0.03573682904243469, "global_step": 190160, "epoch": 2136, "lr": 2.061261348404141e-05}
{"train_loss": 0.04408083111047745, "global_step": 190161, "epoch": 2136, "lr": 2.061214445344881e-05}
{"train_loss": 0.038054030388593674, "global_step": 190162, "epoch": 2136, "lr": 2.0611675426807074e-05}
{"train_loss": 0.08031841367483139, "global_step": 190163, "epoch": 2136, "lr": 2.0611206404116217e-05}
{"train_loss": 0.05607396364212036, "global_step": 190164, "epoch": 2136, "lr": 2.061073738537635e-05}
{"train_loss": 0.06502307951450348, "global_step": 190165, "epoch": 2136, "lr": 2.0610268370587486e-05}
{"train_loss": 0.08119203895330429, "global_step": 190166, "epoch": 2136, "lr": 2.0609799359749737e-05}
{"train_loss": 0.10399255156517029, "global_step": 190167, "epoch": 2136, "lr": 2.0609330352863123e-05}
{"train_loss": 0.0627327635884285, "global_step": 190168, "epoch": 2136, "lr": 2.0608861349927737e-05}
{"train_loss": 0.10255308449268341, "global_step": 190169, "epoch": 2136, "lr": 2.0608392350943638e-05}
{"train_loss": 0.06292037665843964, "global_step": 190170, "epoch": 2136, "lr": 2.0607923355910863e-05}
{"train_loss": 0.03682371601462364, "global_step": 190171, "epoch": 2136, "lr": 2.0607454364829514e-05}
{"train_loss": 0.13115625083446503, "global_step": 190172, "epoch": 2136, "lr": 2.0606985377699618e-05}
{"train_loss": 0.0676661878824234, "global_step": 190173, "epoch": 2136, "lr": 2.060651639452127e-05}
{"train_loss": 0.05826976150274277, "global_step": 190174, "epoch": 2136, "lr": 2.06060474152945e-05}
{"train_loss": 0.06780589371919632, "global_step": 190175, "epoch": 2136, "lr": 2.06055784400194e-05}
{"train_loss": 0.07832351326942444, "global_step": 190176, "epoch": 2136, "lr": 2.0605109468696017e-05}
{"train_loss": 0.1202336773276329, "global_step": 190177, "epoch": 2136, "lr": 2.0604640501324424e-05}
{"train_loss": 0.07660562545061111, "global_step": 190178, "epoch": 2136, "lr": 2.060417153790467e-05}
{"train_loss": 0.0859827846288681, "global_step": 190179, "epoch": 2136, "lr": 2.060370257843684e-05}
{"train_loss": 0.07042660564184189, "global_step": 190180, "epoch": 2136, "lr": 2.0603233622920965e-05}
{"train_loss": 0.08707946538925171, "global_step": 190181, "epoch": 2136, "lr": 2.060276467135715e-05}
{"train_loss": 0.058828774839639664, "global_step": 190182, "epoch": 2136, "lr": 2.0602295723745412e-05}
{"train_loss": 0.05526752769947052, "global_step": 190183, "epoch": 2136, "lr": 2.060182678008586e-05}
{"train_loss": 0.08343017101287842, "global_step": 190184, "epoch": 2136, "lr": 2.060135784037851e-05}
{"train_loss": 0.10294543951749802, "global_step": 190185, "epoch": 2136, "lr": 2.060088890462347e-05}
{"train_loss": 0.09388979524374008, "global_step": 190186, "epoch": 2136, "lr": 2.0600419972820767e-05}
{"train_loss": 0.11433335393667221, "global_step": 190187, "epoch": 2136, "lr": 2.059995104497048e-05}
{"train_loss": 0.06847888976335526, "global_step": 190188, "epoch": 2136, "lr": 2.0599482121072683e-05}
{"train_loss": 0.12417066097259521, "global_step": 190189, "epoch": 2136, "lr": 2.059901320112741e-05}
{"train_loss": 0.0718693807721138, "global_step": 190190, "epoch": 2136, "lr": 2.0598544285134763e-05}
{"train_loss": 0.06133335456252098, "global_step": 190191, "epoch": 2136, "lr": 2.0598075373094767e-05}
{"train_loss": 0.07249577090227871, "global_step": 190192, "epoch": 2136, "lr": 2.0597606465007512e-05, "val_loss": 7.788307189941406}
{"train_loss": 0.06542187929153442, "global_step": 190193, "epoch": 2137, "lr": 2.0597137560873038e-05}
{"train_loss": 0.09926077723503113, "global_step": 190194, "epoch": 2137, "lr": 2.0596668660691436e-05}
{"train_loss": 0.11740174889564514, "global_step": 190195, "epoch": 2137, "lr": 2.0596199764462737e-05}
{"train_loss": 0.0449703074991703, "global_step": 190196, "epoch": 2137, "lr": 2.0595730872187043e-05}
{"train_loss": 0.06220187246799469, "global_step": 190197, "epoch": 2137, "lr": 2.0595261983864373e-05}
{"train_loss": 0.09702762216329575, "global_step": 190198, "epoch": 2137, "lr": 2.0594793099494823e-05}
{"train_loss": 0.07226070016622543, "global_step": 190199, "epoch": 2137, "lr": 2.0594324219078437e-05}
{"train_loss": 0.05814541503787041, "global_step": 190200, "epoch": 2137, "lr": 2.05938553426153e-05}
{"train_loss": 0.026447676122188568, "global_step": 190201, "epoch": 2137, "lr": 2.059338647010544e-05}
{"train_loss": 0.06157774478197098, "global_step": 190202, "epoch": 2137, "lr": 2.0592917601548962e-05}
{"train_loss": 0.0724581852555275, "global_step": 190203, "epoch": 2137, "lr": 2.0592448736945885e-05}
{"train_loss": 0.09882107377052307, "global_step": 190204, "epoch": 2137, "lr": 2.0591979876296317e-05}
{"train_loss": 0.057803675532341, "global_step": 190205, "epoch": 2137, "lr": 2.059151101960028e-05}
{"train_loss": 0.09063178300857544, "global_step": 190206, "epoch": 2137, "lr": 2.0591042166857872e-05}
{"train_loss": 0.08185563236474991, "global_step": 190207, "epoch": 2137, "lr": 2.0590573318069144e-05}
{"train_loss": 0.04459911212325096, "global_step": 190208, "epoch": 2137, "lr": 2.0590104473234135e-05}
{"train_loss": 0.06911782175302505, "global_step": 190209, "epoch": 2137, "lr": 2.0589635632352944e-05}
{"train_loss": 0.10141725838184357, "global_step": 190210, "epoch": 2137, "lr": 2.05891667954256e-05}
{"train_loss": 0.037506990134716034, "global_step": 190211, "epoch": 2137, "lr": 2.0588697962452208e-05}
{"train_loss": 0.06561645865440369, "global_step": 190212, "epoch": 2137, "lr": 2.0588229133432785e-05}
{"train_loss": 0.08457903563976288, "global_step": 190213, "epoch": 2137, "lr": 2.0587760308367434e-05}
{"train_loss": 0.10360445082187653, "global_step": 190214, "epoch": 2137, "lr": 2.058729148725618e-05}
{"train_loss": 0.10847611725330353, "global_step": 190215, "epoch": 2137, "lr": 2.0586822670099126e-05}
{"train_loss": 0.0902000144124031, "global_step": 190216, "epoch": 2137, "lr": 2.0586353856896302e-05}
{"train_loss": 0.06104985997080803, "global_step": 190217, "epoch": 2137, "lr": 2.0585885047647797e-05}
{"train_loss": 0.09882713854312897, "global_step": 190218, "epoch": 2137, "lr": 2.0585416242353644e-05}
{"train_loss": 0.07048925757408142, "global_step": 190219, "epoch": 2137, "lr": 2.058494744101394e-05}
{"train_loss": 0.04015999287366867, "global_step": 190220, "epoch": 2137, "lr": 2.0584478643628714e-05}
{"train_loss": 0.08289148658514023, "global_step": 190221, "epoch": 2137, "lr": 2.0584009850198065e-05}
{"train_loss": 0.06673964858055115, "global_step": 190222, "epoch": 2137, "lr": 2.0583541060722023e-05}
{"train_loss": 0.07903813570737839, "global_step": 190223, "epoch": 2137, "lr": 2.058307227520066e-05}
{"train_loss": 0.14201101660728455, "global_step": 190224, "epoch": 2137, "lr": 2.058260349363406e-05}
{"train_loss": 0.11806844919919968, "global_step": 190225, "epoch": 2137, "lr": 2.058213471602226e-05}
{"train_loss": 0.11635972559452057, "global_step": 190226, "epoch": 2137, "lr": 2.0581665942365347e-05}
{"train_loss": 0.061465661972761154, "global_step": 190227, "epoch": 2137, "lr": 2.0581197172663354e-05}
{"train_loss": 0.1143319383263588, "global_step": 190228, "epoch": 2137, "lr": 2.0580728406916377e-05}
{"train_loss": 0.13584266602993011, "global_step": 190229, "epoch": 2137, "lr": 2.0580259645124445e-05}
{"train_loss": 0.10989857465028763, "global_step": 190230, "epoch": 2137, "lr": 2.0579790887287658e-05}
{"train_loss": 0.03217696398496628, "global_step": 190231, "epoch": 2137, "lr": 2.0579322133406044e-05}
{"train_loss": 0.07263641804456711, "global_step": 190232, "epoch": 2137, "lr": 2.057885338347969e-05}
{"train_loss": 0.05703529715538025, "global_step": 190233, "epoch": 2137, "lr": 2.057838463750864e-05}
{"train_loss": 0.07691361010074615, "global_step": 190234, "epoch": 2137, "lr": 2.0577915895492982e-05}
{"train_loss": 0.02472430281341076, "global_step": 190235, "epoch": 2137, "lr": 2.057744715743275e-05}
{"train_loss": 0.07043585181236267, "global_step": 190236, "epoch": 2137, "lr": 2.0576978423328042e-05}
{"train_loss": 0.03898051008582115, "global_step": 190237, "epoch": 2137, "lr": 2.0576509693178875e-05}
{"train_loss": 0.03352092579007149, "global_step": 190238, "epoch": 2137, "lr": 2.057604096698536e-05}
{"train_loss": 0.07771380245685577, "global_step": 190239, "epoch": 2137, "lr": 2.057557224474752e-05}
{"train_loss": 0.07314860820770264, "global_step": 190240, "epoch": 2137, "lr": 2.0575103526465455e-05}
{"train_loss": 0.061726685613393784, "global_step": 190241, "epoch": 2137, "lr": 2.057463481213919e-05}
{"train_loss": 0.08342915773391724, "global_step": 190242, "epoch": 2137, "lr": 2.057416610176882e-05}
{"train_loss": 0.07961202412843704, "global_step": 190243, "epoch": 2137, "lr": 2.057369739535438e-05}
{"train_loss": 0.051949311047792435, "global_step": 190244, "epoch": 2137, "lr": 2.0573228692895968e-05}
{"train_loss": 0.04998140409588814, "global_step": 190245, "epoch": 2137, "lr": 2.0572759994393602e-05}
{"train_loss": 0.07953585684299469, "global_step": 190246, "epoch": 2137, "lr": 2.0572291299847395e-05}
{"train_loss": 0.047380320727825165, "global_step": 190247, "epoch": 2137, "lr": 2.0571822609257384e-05}
{"train_loss": 0.10265035927295685, "global_step": 190248, "epoch": 2137, "lr": 2.057135392262361e-05}
{"train_loss": 0.09432092308998108, "global_step": 190249, "epoch": 2137, "lr": 2.057088523994618e-05}
{"train_loss": 0.08392561972141266, "global_step": 190250, "epoch": 2137, "lr": 2.057041656122512e-05}
{"train_loss": 0.06751957535743713, "global_step": 190251, "epoch": 2137, "lr": 2.056994788646052e-05}
{"train_loss": 0.0404181145131588, "global_step": 190252, "epoch": 2137, "lr": 2.056947921565242e-05}
{"train_loss": 0.11065762490034103, "global_step": 190253, "epoch": 2137, "lr": 2.056901054880091e-05}
{"train_loss": 0.12687057256698608, "global_step": 190254, "epoch": 2137, "lr": 2.0568541885906024e-05}
{"train_loss": 0.03247358277440071, "global_step": 190255, "epoch": 2137, "lr": 2.0568073226967855e-05}
{"train_loss": 0.12919583916664124, "global_step": 190256, "epoch": 2137, "lr": 2.0567604571986425e-05}
{"train_loss": 0.06309450417757034, "global_step": 190257, "epoch": 2137, "lr": 2.0567135920961848e-05}
{"train_loss": 0.08934437483549118, "global_step": 190258, "epoch": 2137, "lr": 2.056666727389414e-05}
{"train_loss": 0.10059113055467606, "global_step": 190259, "epoch": 2137, "lr": 2.0566198630783384e-05}
{"train_loss": 0.10542666912078857, "global_step": 190260, "epoch": 2137, "lr": 2.0565729991629662e-05}
{"train_loss": 0.0609419122338295, "global_step": 190261, "epoch": 2137, "lr": 2.0565261356433004e-05}
{"train_loss": 0.08552175015211105, "global_step": 190262, "epoch": 2137, "lr": 2.0564792725193504e-05}
{"train_loss": 0.06746694445610046, "global_step": 190263, "epoch": 2137, "lr": 2.0564324097911186e-05}
{"train_loss": 0.09419314563274384, "global_step": 190264, "epoch": 2137, "lr": 2.056385547458616e-05}
{"train_loss": 0.07772615551948547, "global_step": 190265, "epoch": 2137, "lr": 2.056338685521845e-05}
{"train_loss": 0.06714379042387009, "global_step": 190266, "epoch": 2137, "lr": 2.0562918239808145e-05}
{"train_loss": 0.04121634364128113, "global_step": 190267, "epoch": 2137, "lr": 2.056244962835528e-05}
{"train_loss": 0.033743858337402344, "global_step": 190268, "epoch": 2137, "lr": 2.056198102085995e-05}
{"train_loss": 0.048430029302835464, "global_step": 190269, "epoch": 2137, "lr": 2.056151241732219e-05}
{"train_loss": 0.04413966089487076, "global_step": 190270, "epoch": 2137, "lr": 2.0561043817742093e-05}
{"train_loss": 0.07359064370393753, "global_step": 190271, "epoch": 2137, "lr": 2.0560575222119682e-05}
{"train_loss": 0.07051537185907364, "global_step": 190272, "epoch": 2137, "lr": 2.0560106630455065e-05}
{"train_loss": 0.14490412175655365, "global_step": 190273, "epoch": 2137, "lr": 2.0559638042748268e-05}
{"train_loss": 0.12074680626392365, "global_step": 190274, "epoch": 2137, "lr": 2.0559169458999385e-05}
{"train_loss": 0.05469485744833946, "global_step": 190275, "epoch": 2137, "lr": 2.0558700879208443e-05}
{"train_loss": 0.07769792526960373, "global_step": 190276, "epoch": 2137, "lr": 2.055823230337554e-05}
{"train_loss": 0.04882311820983887, "global_step": 190277, "epoch": 2137, "lr": 2.0557763731500713e-05}
{"train_loss": 0.04644937813282013, "global_step": 190278, "epoch": 2137, "lr": 2.0557295163584028e-05}
{"train_loss": 0.07466231286525726, "global_step": 190279, "epoch": 2137, "lr": 2.0556826599625578e-05}
{"train_loss": 0.08740417659282684, "global_step": 190280, "epoch": 2137, "lr": 2.0556358039625408e-05}
{"train_loss": 0.07571507653409845, "global_step": 190281, "epoch": 2137, "lr": 2.055588948358355e-05, "val_loss": 7.838240146636963}
{"train_loss": 0.05197957903146744, "global_step": 190282, "epoch": 2138, "lr": 2.0555420931500117e-05}
{"train_loss": 0.050630588084459305, "global_step": 190283, "epoch": 2138, "lr": 2.0554952383375133e-05}
{"train_loss": 0.09311748296022415, "global_step": 190284, "epoch": 2138, "lr": 2.055448383920869e-05}
{"train_loss": 0.06083230674266815, "global_step": 190285, "epoch": 2138, "lr": 2.055401529900084e-05}
{"train_loss": 0.043903838843107224, "global_step": 190286, "epoch": 2138, "lr": 2.0553546762751626e-05}
{"train_loss": 0.04776239022612572, "global_step": 190287, "epoch": 2138, "lr": 2.0553078230461142e-05}
{"train_loss": 0.0571889691054821, "global_step": 190288, "epoch": 2138, "lr": 2.0552609702129427e-05}
{"train_loss": 0.08678776025772095, "global_step": 190289, "epoch": 2138, "lr": 2.0552141177756567e-05}
{"train_loss": 0.05832979828119278, "global_step": 190290, "epoch": 2138, "lr": 2.0551672657342597e-05}
{"train_loss": 0.1142842173576355, "global_step": 190291, "epoch": 2138, "lr": 2.0551204140887615e-05}
{"train_loss": 0.07953250408172607, "global_step": 190292, "epoch": 2138, "lr": 2.055073562839164e-05}
{"train_loss": 0.05722114071249962, "global_step": 190293, "epoch": 2138, "lr": 2.0550267119854788e-05}
{"train_loss": 0.08787661045789719, "global_step": 190294, "epoch": 2138, "lr": 2.054979861527707e-05}
{"train_loss": 0.05953061208128929, "global_step": 190295, "epoch": 2138, "lr": 2.0549330114658573e-05}
{"train_loss": 0.08864350616931915, "global_step": 190296, "epoch": 2138, "lr": 2.0548861617999378e-05}
{"train_loss": 0.05213319510221481, "global_step": 190297, "epoch": 2138, "lr": 2.054839312529951e-05}
{"train_loss": 0.1145661473274231, "global_step": 190298, "epoch": 2138, "lr": 2.0547924636559072e-05}
{"train_loss": 0.0701797753572464, "global_step": 190299, "epoch": 2138, "lr": 2.0547456151778088e-05}
{"train_loss": 0.0725116953253746, "global_step": 190300, "epoch": 2138, "lr": 2.0546987670956657e-05}
{"train_loss": 0.09553831815719604, "global_step": 190301, "epoch": 2138, "lr": 2.0546519194094804e-05}
{"train_loss": 0.06669335812330246, "global_step": 190302, "epoch": 2138, "lr": 2.0546050721192633e-05}
{"train_loss": 0.056870296597480774, "global_step": 190303, "epoch": 2138, "lr": 2.0545582252250172e-05}
{"train_loss": 0.0911610797047615, "global_step": 190304, "epoch": 2138, "lr": 2.0545113787267512e-05}
{"train_loss": 0.034782763570547104, "global_step": 190305, "epoch": 2138, "lr": 2.0544645326244687e-05}
{"train_loss": 0.07219412177801132, "global_step": 190306, "epoch": 2138, "lr": 2.054417686918179e-05}
{"train_loss": 0.04506045952439308, "global_step": 190307, "epoch": 2138, "lr": 2.0543708416078857e-05}
{"train_loss": 0.0782330185174942, "global_step": 190308, "epoch": 2138, "lr": 2.0543239966935974e-05}
{"train_loss": 0.1011502742767334, "global_step": 190309, "epoch": 2138, "lr": 2.054277152175318e-05}
{"train_loss": 0.18049147725105286, "global_step": 190310, "epoch": 2138, "lr": 2.0542303080530566e-05}
{"train_loss": 0.055390115827322006, "global_step": 190311, "epoch": 2138, "lr": 2.054183464326817e-05}
{"train_loss": 0.09451699256896973, "global_step": 190312, "epoch": 2138, "lr": 2.0541366209966078e-05}
{"train_loss": 0.10047357529401779, "global_step": 190313, "epoch": 2138, "lr": 2.0540897780624318e-05}
{"train_loss": 0.10243849456310272, "global_step": 190314, "epoch": 2138, "lr": 2.0540429355242986e-05}
{"train_loss": 0.032910484820604324, "global_step": 190315, "epoch": 2138, "lr": 2.0539960933822146e-05}
{"train_loss": 0.040195681154727936, "global_step": 190316, "epoch": 2138, "lr": 2.0539492516361825e-05}
{"train_loss": 0.05379367247223854, "global_step": 190317, "epoch": 2138, "lr": 2.0539024102862136e-05}
{"train_loss": 0.04849043861031532, "global_step": 190318, "epoch": 2138, "lr": 2.0538555693323115e-05}
{"train_loss": 0.04547123238444328, "global_step": 190319, "epoch": 2138, "lr": 2.0538087287744807e-05}
{"train_loss": 0.0700492262840271, "global_step": 190320, "epoch": 2138, "lr": 2.0537618886127313e-05}
{"train_loss": 0.06889215856790543, "global_step": 190321, "epoch": 2138, "lr": 2.0537150488470657e-05}
{"train_loss": 0.11438295245170593, "global_step": 190322, "epoch": 2138, "lr": 2.0536682094774944e-05}
{"train_loss": 0.05952446907758713, "global_step": 190323, "epoch": 2138, "lr": 2.0536213705040207e-05}
{"train_loss": 0.052671805024147034, "global_step": 190324, "epoch": 2138, "lr": 2.0535745319266508e-05}
{"train_loss": 0.027518000453710556, "global_step": 190325, "epoch": 2138, "lr": 2.0535276937453935e-05}
{"train_loss": 0.08976762741804123, "global_step": 190326, "epoch": 2138, "lr": 2.0534808559602515e-05}
{"train_loss": 0.10469618439674377, "global_step": 190327, "epoch": 2138, "lr": 2.0534340185712347e-05}
{"train_loss": 0.15514475107192993, "global_step": 190328, "epoch": 2138, "lr": 2.0533871815783463e-05}
{"train_loss": 0.07642005383968353, "global_step": 190329, "epoch": 2138, "lr": 2.053340344981596e-05}
{"train_loss": 0.057965319603681564, "global_step": 190330, "epoch": 2138, "lr": 2.0532935087809856e-05}
{"train_loss": 0.06805924326181412, "global_step": 190331, "epoch": 2138, "lr": 2.0532466729765248e-05}
{"train_loss": 0.08703097701072693, "global_step": 190332, "epoch": 2138, "lr": 2.0531998375682203e-05}
{"train_loss": 0.07701639831066132, "global_step": 190333, "epoch": 2138, "lr": 2.0531530025560758e-05}
{"train_loss": 0.061617154628038406, "global_step": 190334, "epoch": 2138, "lr": 2.0531061679401008e-05}
{"train_loss": 0.07867799699306488, "global_step": 190335, "epoch": 2138, "lr": 2.053059333720298e-05}
{"train_loss": 0.05197804048657417, "global_step": 190336, "epoch": 2138, "lr": 2.0530124998966772e-05}
{"train_loss": 0.027975784614682198, "global_step": 190337, "epoch": 2138, "lr": 2.0529656664692402e-05}
{"train_loss": 0.025329265743494034, "global_step": 190338, "epoch": 2138, "lr": 2.052918833437999e-05}
{"train_loss": 0.03711400181055069, "global_step": 190339, "epoch": 2138, "lr": 2.052872000802955e-05}
{"train_loss": 0.06477955728769302, "global_step": 190340, "epoch": 2138, "lr": 2.052825168564118e-05}
{"train_loss": 0.09026458114385605, "global_step": 190341, "epoch": 2138, "lr": 2.052778336721491e-05}
{"train_loss": 0.05807255581021309, "global_step": 190342, "epoch": 2138, "lr": 2.0527315052750834e-05}
{"train_loss": 0.07462363690137863, "global_step": 190343, "epoch": 2138, "lr": 2.052684674224899e-05}
{"train_loss": 0.08067326247692108, "global_step": 190344, "epoch": 2138, "lr": 2.0526378435709466e-05}
{"train_loss": 0.04585864022374153, "global_step": 190345, "epoch": 2138, "lr": 2.0525910133132292e-05}
{"train_loss": 0.060100603848695755, "global_step": 190346, "epoch": 2138, "lr": 2.0525441834517572e-05}
{"train_loss": 0.14185839891433716, "global_step": 190347, "epoch": 2138, "lr": 2.0524973539865326e-05}
{"train_loss": 0.08842392265796661, "global_step": 190348, "epoch": 2138, "lr": 2.0524505249175663e-05}
{"train_loss": 0.10017107427120209, "global_step": 190349, "epoch": 2138, "lr": 2.0524036962448597e-05}
{"train_loss": 0.08384449779987335, "global_step": 190350, "epoch": 2138, "lr": 2.0523568679684214e-05}
{"train_loss": 0.06097765266895294, "global_step": 190351, "epoch": 2138, "lr": 2.05231004008826e-05}
{"train_loss": 0.045468516647815704, "global_step": 190352, "epoch": 2138, "lr": 2.052263212604377e-05}
{"train_loss": 0.06181255728006363, "global_step": 190353, "epoch": 2138, "lr": 2.0522163855167842e-05}
{"train_loss": 0.050418343394994736, "global_step": 190354, "epoch": 2138, "lr": 2.0521695588254824e-05}
{"train_loss": 0.08010930567979813, "global_step": 190355, "epoch": 2138, "lr": 2.0521227325304826e-05}
{"train_loss": 0.07025402784347534, "global_step": 190356, "epoch": 2138, "lr": 2.0520759066317868e-05}
{"train_loss": 0.06969380378723145, "global_step": 190357, "epoch": 2138, "lr": 2.0520290811294056e-05}
{"train_loss": 0.09135846048593521, "global_step": 190358, "epoch": 2138, "lr": 2.0519822560233432e-05}
{"train_loss": 0.08922850340604782, "global_step": 190359, "epoch": 2138, "lr": 2.0519354313136042e-05}
{"train_loss": 0.07722575217485428, "global_step": 190360, "epoch": 2138, "lr": 2.051888607000198e-05}
{"train_loss": 0.05577278137207031, "global_step": 190361, "epoch": 2138, "lr": 2.0518417830831277e-05}
{"train_loss": 0.07100889086723328, "global_step": 190362, "epoch": 2138, "lr": 2.051794959562403e-05}
{"train_loss": 0.07380329072475433, "global_step": 190363, "epoch": 2138, "lr": 2.0517481364380285e-05}
{"train_loss": 0.0382208414375782, "global_step": 190364, "epoch": 2138, "lr": 2.0517013137100088e-05}
{"train_loss": 0.05216722562909126, "global_step": 190365, "epoch": 2138, "lr": 2.051654491378354e-05}
{"train_loss": 0.06032490357756615, "global_step": 190366, "epoch": 2138, "lr": 2.0516076694430664e-05}
{"train_loss": 0.055507782846689224, "global_step": 190367, "epoch": 2138, "lr": 2.051560847904156e-05}
{"train_loss": 0.04835829883813858, "global_step": 190368, "epoch": 2138, "lr": 2.0515140267616257e-05}
{"train_loss": 0.050157222896814346, "global_step": 190369, "epoch": 2138, "lr": 2.0514672060154832e-05}
{"train_loss": 0.07069106197005577, "global_step": 190370, "epoch": 2138, "lr": 2.0514203856657365e-05, "val_loss": 7.935770511627197}
{"train_loss": 0.08693332970142365, "global_step": 190371, "epoch": 2139, "lr": 2.051373565712389e-05}
{"train_loss": 0.08699629455804825, "global_step": 190372, "epoch": 2139, "lr": 2.0513267461554503e-05}
{"train_loss": 0.06662826985120773, "global_step": 190373, "epoch": 2139, "lr": 2.051279926994923e-05}
{"train_loss": 0.06216741353273392, "global_step": 190374, "epoch": 2139, "lr": 2.051233108230816e-05}
{"train_loss": 0.060747433453798294, "global_step": 190375, "epoch": 2139, "lr": 2.051186289863134e-05}
{"train_loss": 0.0979829728603363, "global_step": 190376, "epoch": 2139, "lr": 2.0511394718918857e-05}
{"train_loss": 0.04642624780535698, "global_step": 190377, "epoch": 2139, "lr": 2.0510926543170738e-05}
{"train_loss": 0.09145013242959976, "global_step": 190378, "epoch": 2139, "lr": 2.0510458371387082e-05}
{"train_loss": 0.06608650833368301, "global_step": 190379, "epoch": 2139, "lr": 2.050999020356792e-05}
{"train_loss": 0.13169622421264648, "global_step": 190380, "epoch": 2139, "lr": 2.0509522039713347e-05}
{"train_loss": 0.07820052653551102, "global_step": 190381, "epoch": 2139, "lr": 2.0509053879823393e-05}
{"train_loss": 0.09360986202955246, "global_step": 190382, "epoch": 2139, "lr": 2.0508585723898156e-05}
{"train_loss": 0.09653099626302719, "global_step": 190383, "epoch": 2139, "lr": 2.0508117571937657e-05}
{"train_loss": 0.1062101200222969, "global_step": 190384, "epoch": 2139, "lr": 2.0507649423942003e-05}
{"train_loss": 0.06045631319284439, "global_step": 190385, "epoch": 2139, "lr": 2.050718127991122e-05}
{"train_loss": 0.03837420046329498, "global_step": 190386, "epoch": 2139, "lr": 2.0506713139845384e-05}
{"train_loss": 0.053532954305410385, "global_step": 190387, "epoch": 2139, "lr": 2.050624500374458e-05}
{"train_loss": 0.10372360795736313, "global_step": 190388, "epoch": 2139, "lr": 2.0505776871608833e-05}
{"train_loss": 0.10070358961820602, "global_step": 190389, "epoch": 2139, "lr": 2.0505308743438246e-05}
{"train_loss": 0.07747367769479752, "global_step": 190390, "epoch": 2139, "lr": 2.050484061923284e-05}
{"train_loss": 0.060053933411836624, "global_step": 190391, "epoch": 2139, "lr": 2.050437249899272e-05}
{"train_loss": 0.05267301946878433, "global_step": 190392, "epoch": 2139, "lr": 2.0503904382717905e-05}
{"train_loss": 0.048885807394981384, "global_step": 190393, "epoch": 2139, "lr": 2.0503436270408498e-05}
{"train_loss": 0.05966907739639282, "global_step": 190394, "epoch": 2139, "lr": 2.0502968162064534e-05}
{"train_loss": 0.0799594596028328, "global_step": 190395, "epoch": 2139, "lr": 2.0502500057686097e-05}
{"train_loss": 0.07348114252090454, "global_step": 190396, "epoch": 2139, "lr": 2.0502031957273245e-05}
{"train_loss": 0.04994936287403107, "global_step": 190397, "epoch": 2139, "lr": 2.050156386082601e-05}
{"train_loss": 0.10915477573871613, "global_step": 190398, "epoch": 2139, "lr": 2.0501095768344504e-05}
{"train_loss": 0.07630997896194458, "global_step": 190399, "epoch": 2139, "lr": 2.0500627679828747e-05}
{"train_loss": 0.07810347527265549, "global_step": 190400, "epoch": 2139, "lr": 2.050015959527884e-05}
{"train_loss": 0.0756840929389, "global_step": 190401, "epoch": 2139, "lr": 2.0499691514694825e-05}
{"train_loss": 0.11305426061153412, "global_step": 190402, "epoch": 2139, "lr": 2.0499223438076748e-05}
{"train_loss": 0.10508143901824951, "global_step": 190403, "epoch": 2139, "lr": 2.049875536542471e-05}
{"train_loss": 0.07035831362009048, "global_step": 190404, "epoch": 2139, "lr": 2.049828729673874e-05}
{"train_loss": 0.04394582659006119, "global_step": 190405, "epoch": 2139, "lr": 2.049781923201891e-05}
{"train_loss": 0.06662517040967941, "global_step": 190406, "epoch": 2139, "lr": 2.0497351171265312e-05}
{"train_loss": 0.06753235310316086, "global_step": 190407, "epoch": 2139, "lr": 2.0496883114477967e-05}
{"train_loss": 0.07343535125255585, "global_step": 190408, "epoch": 2139, "lr": 2.0496415061656972e-05}
{"train_loss": 0.1302875429391861, "global_step": 190409, "epoch": 2139, "lr": 2.049594701280235e-05}
{"train_loss": 0.15009166300296783, "global_step": 190410, "epoch": 2139, "lr": 2.0495478967914216e-05}
{"train_loss": 0.11765125393867493, "global_step": 190411, "epoch": 2139, "lr": 2.0495010926992587e-05}
{"train_loss": 0.022708354517817497, "global_step": 190412, "epoch": 2139, "lr": 2.049454289003756e-05}
{"train_loss": 0.048823583871126175, "global_step": 190413, "epoch": 2139, "lr": 2.0494074857049162e-05}
{"train_loss": 0.10259535908699036, "global_step": 190414, "epoch": 2139, "lr": 2.0493606828027496e-05}
{"train_loss": 0.028045788407325745, "global_step": 190415, "epoch": 2139, "lr": 2.0493138802972583e-05}
{"train_loss": 0.03239457309246063, "global_step": 190416, "epoch": 2139, "lr": 2.0492670781884532e-05}
{"train_loss": 0.08230847865343094, "global_step": 190417, "epoch": 2139, "lr": 2.0492202764763363e-05}
{"train_loss": 0.06650380790233612, "global_step": 190418, "epoch": 2139, "lr": 2.0491734751609172e-05}
{"train_loss": 0.06816906481981277, "global_step": 190419, "epoch": 2139, "lr": 2.0491266742421995e-05}
{"train_loss": 0.10181818157434464, "global_step": 190420, "epoch": 2139, "lr": 2.049079873720192e-05}
{"train_loss": 0.04592321813106537, "global_step": 190421, "epoch": 2139, "lr": 2.0490330735948982e-05}
{"train_loss": 0.027890950441360474, "global_step": 190422, "epoch": 2139, "lr": 2.048986273866326e-05}
{"train_loss": 0.04353118687868118, "global_step": 190423, "epoch": 2139, "lr": 2.048939474534483e-05}
{"train_loss": 0.08095593005418777, "global_step": 190424, "epoch": 2139, "lr": 2.0488926755993725e-05}
{"train_loss": 0.12506641447544098, "global_step": 190425, "epoch": 2139, "lr": 2.0488458770610047e-05}
{"train_loss": 0.05285194143652916, "global_step": 190426, "epoch": 2139, "lr": 2.048799078919381e-05}
{"train_loss": 0.06039370223879814, "global_step": 190427, "epoch": 2139, "lr": 2.048752281174513e-05}
{"train_loss": 0.07610639929771423, "global_step": 190428, "epoch": 2139, "lr": 2.0487054838264018e-05}
{"train_loss": 0.05035044252872467, "global_step": 190429, "epoch": 2139, "lr": 2.0486586868750584e-05}
{"train_loss": 0.06858061254024506, "global_step": 190430, "epoch": 2139, "lr": 2.0486118903204848e-05}
{"train_loss": 0.09655626863241196, "global_step": 190431, "epoch": 2139, "lr": 2.048565094162691e-05}
{"train_loss": 0.07943624258041382, "global_step": 190432, "epoch": 2139, "lr": 2.0485182984016805e-05}
{"train_loss": 0.10528996586799622, "global_step": 190433, "epoch": 2139, "lr": 2.0484715030374617e-05}
{"train_loss": 0.08291111886501312, "global_step": 190434, "epoch": 2139, "lr": 2.04842470807004e-05}
{"train_loss": 0.04552963003516197, "global_step": 190435, "epoch": 2139, "lr": 2.0483779134994207e-05}
{"train_loss": 0.039697617292404175, "global_step": 190436, "epoch": 2139, "lr": 2.048331119325612e-05}
{"train_loss": 0.05210965499281883, "global_step": 190437, "epoch": 2139, "lr": 2.0482843255486178e-05}
{"train_loss": 0.06931628286838531, "global_step": 190438, "epoch": 2139, "lr": 2.0482375321684476e-05}
{"train_loss": 0.06386195868253708, "global_step": 190439, "epoch": 2139, "lr": 2.048190739185106e-05}
{"train_loss": 0.08348764479160309, "global_step": 190440, "epoch": 2139, "lr": 2.0481439465985967e-05}
{"train_loss": 0.0959228053689003, "global_step": 190441, "epoch": 2139, "lr": 2.0480971544089296e-05}
{"train_loss": 0.07622046768665314, "global_step": 190442, "epoch": 2139, "lr": 2.0480503626161107e-05}
{"train_loss": 0.08514861017465591, "global_step": 190443, "epoch": 2139, "lr": 2.048003571220144e-05}
{"train_loss": 0.05461452901363373, "global_step": 190444, "epoch": 2139, "lr": 2.0479567802210385e-05}
{"train_loss": 0.1193632259964943, "global_step": 190445, "epoch": 2139, "lr": 2.0479099896187983e-05}
{"train_loss": 0.1066485270857811, "global_step": 190446, "epoch": 2139, "lr": 2.047863199413432e-05}
{"train_loss": 0.08700673282146454, "global_step": 190447, "epoch": 2139, "lr": 2.0478164096049428e-05}
{"train_loss": 0.1139533519744873, "global_step": 190448, "epoch": 2139, "lr": 2.0477696201933405e-05}
{"train_loss": 0.024207446724176407, "global_step": 190449, "epoch": 2139, "lr": 2.0477228311786273e-05}
{"train_loss": 0.0686490386724472, "global_step": 190450, "epoch": 2139, "lr": 2.0476760425608144e-05}
{"train_loss": 0.04622402414679527, "global_step": 190451, "epoch": 2139, "lr": 2.0476292543399033e-05}
{"train_loss": 0.06295789778232574, "global_step": 190452, "epoch": 2139, "lr": 2.047582466515904e-05}
{"train_loss": 0.058598604053258896, "global_step": 190453, "epoch": 2139, "lr": 2.0475356790888195e-05}
{"train_loss": 0.049924276769161224, "global_step": 190454, "epoch": 2139, "lr": 2.0474888920586598e-05}
{"train_loss": 0.06436562538146973, "global_step": 190455, "epoch": 2139, "lr": 2.047442105425428e-05}
{"train_loss": 0.08164961636066437, "global_step": 190456, "epoch": 2139, "lr": 2.0473953191891325e-05}
{"train_loss": 0.04828278720378876, "global_step": 190457, "epoch": 2139, "lr": 2.0473485333497778e-05}
{"train_loss": 0.06721948087215424, "global_step": 190458, "epoch": 2139, "lr": 2.0473017479073702e-05}
{"train_loss": 0.07460715845645813, "global_step": 190459, "epoch": 2139, "lr": 2.047254962861919e-05, "val_loss": 7.989475727081299}
{"train_loss": 0.07096225023269653, "global_step": 190460, "epoch": 2140, "lr": 2.0472081782134267e-05}
{"train_loss": 0.06759169697761536, "global_step": 190461, "epoch": 2140, "lr": 2.0471613939619033e-05}
{"train_loss": 0.06355466693639755, "global_step": 190462, "epoch": 2140, "lr": 2.0471146101073508e-05}
{"train_loss": 0.06268008798360825, "global_step": 190463, "epoch": 2140, "lr": 2.04706782664978e-05}
{"train_loss": 0.03653286397457123, "global_step": 190464, "epoch": 2140, "lr": 2.047021043589193e-05}
{"train_loss": 0.051473114639520645, "global_step": 190465, "epoch": 2140, "lr": 2.0469742609255993e-05}
{"train_loss": 0.11417534947395325, "global_step": 190466, "epoch": 2140, "lr": 2.0469274786590024e-05}
{"train_loss": 0.1085580587387085, "global_step": 190467, "epoch": 2140, "lr": 2.0468806967894123e-05}
{"train_loss": 0.03364269435405731, "global_step": 190468, "epoch": 2140, "lr": 2.0468339153168316e-05}
{"train_loss": 0.07936128973960876, "global_step": 190469, "epoch": 2140, "lr": 2.0467871342412688e-05}
{"train_loss": 0.04765816032886505, "global_step": 190470, "epoch": 2140, "lr": 2.0467403535627283e-05}
{"train_loss": 0.08933600038290024, "global_step": 190471, "epoch": 2140, "lr": 2.0466935732812192e-05}
{"train_loss": 0.06762154400348663, "global_step": 190472, "epoch": 2140, "lr": 2.046646793396745e-05}
{"train_loss": 0.10579865425825119, "global_step": 190473, "epoch": 2140, "lr": 2.0466000139093137e-05}
{"train_loss": 0.0393773689866066, "global_step": 190474, "epoch": 2140, "lr": 2.046553234818932e-05}
{"train_loss": 0.06221810355782509, "global_step": 190475, "epoch": 2140, "lr": 2.046506456125603e-05}
{"train_loss": 0.07728073745965958, "global_step": 190476, "epoch": 2140, "lr": 2.046459677829337e-05}
{"train_loss": 0.05527682229876518, "global_step": 190477, "epoch": 2140, "lr": 2.046412899930137e-05}
{"train_loss": 0.12459942698478699, "global_step": 190478, "epoch": 2140, "lr": 2.0463661224280123e-05}
{"train_loss": 0.06503227353096008, "global_step": 190479, "epoch": 2140, "lr": 2.0463193453229663e-05}
{"train_loss": 0.041170116513967514, "global_step": 190480, "epoch": 2140, "lr": 2.046272568615008e-05}
{"train_loss": 0.045298654586076736, "global_step": 190481, "epoch": 2140, "lr": 2.0462257923041406e-05}
{"train_loss": 0.06505879759788513, "global_step": 190482, "epoch": 2140, "lr": 2.0461790163903742e-05}
{"train_loss": 0.08673969656229019, "global_step": 190483, "epoch": 2140, "lr": 2.046132240873711e-05}
{"train_loss": 0.08488299697637558, "global_step": 190484, "epoch": 2140, "lr": 2.046085465754161e-05}
{"train_loss": 0.07590527087450027, "global_step": 190485, "epoch": 2140, "lr": 2.0460386910317276e-05}
{"train_loss": 0.03048277087509632, "global_step": 190486, "epoch": 2140, "lr": 2.0459919167064197e-05}
{"train_loss": 0.05271846801042557, "global_step": 190487, "epoch": 2140, "lr": 2.04594514277824e-05}
{"train_loss": 0.03667961061000824, "global_step": 190488, "epoch": 2140, "lr": 2.0458983692471995e-05}
{"train_loss": 0.15766392648220062, "global_step": 190489, "epoch": 2140, "lr": 2.0458515961132997e-05}
{"train_loss": 0.05485422536730766, "global_step": 190490, "epoch": 2140, "lr": 2.045804823376551e-05}
{"train_loss": 0.07984600961208344, "global_step": 190491, "epoch": 2140, "lr": 2.0457580510369562e-05}
{"train_loss": 0.028580261394381523, "global_step": 190492, "epoch": 2140, "lr": 2.045711279094525e-05}
{"train_loss": 0.05998527631163597, "global_step": 190493, "epoch": 2140, "lr": 2.0456645075492597e-05}
{"train_loss": 0.05098748207092285, "global_step": 190494, "epoch": 2140, "lr": 2.0456177364011707e-05}
{"train_loss": 0.0474301353096962, "global_step": 190495, "epoch": 2140, "lr": 2.0455709656502612e-05}
{"train_loss": 0.11892402917146683, "global_step": 190496, "epoch": 2140, "lr": 2.045524195296538e-05}
{"train_loss": 0.10306419432163239, "global_step": 190497, "epoch": 2140, "lr": 2.0454774253400106e-05}
{"train_loss": 0.10281313955783844, "global_step": 190498, "epoch": 2140, "lr": 2.0454306557806803e-05}
{"train_loss": 0.05343485251069069, "global_step": 190499, "epoch": 2140, "lr": 2.0453838866185576e-05}
{"train_loss": 0.10033094882965088, "global_step": 190500, "epoch": 2140, "lr": 2.0453371178536456e-05}
{"train_loss": 0.06004629284143448, "global_step": 190501, "epoch": 2140, "lr": 2.0452903494859538e-05}
{"train_loss": 0.023946821689605713, "global_step": 190502, "epoch": 2140, "lr": 2.0452435815154845e-05}
{"train_loss": 0.09389603137969971, "global_step": 190503, "epoch": 2140, "lr": 2.0451968139422483e-05}
{"train_loss": 0.0841858759522438, "global_step": 190504, "epoch": 2140, "lr": 2.0451500467662478e-05}
{"train_loss": 0.0834387019276619, "global_step": 190505, "epoch": 2140, "lr": 2.0451032799874925e-05}
{"train_loss": 0.07060263305902481, "global_step": 190506, "epoch": 2140, "lr": 2.0450565136059852e-05}
{"train_loss": 0.0326162613928318, "global_step": 190507, "epoch": 2140, "lr": 2.0450097476217357e-05}
{"train_loss": 0.09748747944831848, "global_step": 190508, "epoch": 2140, "lr": 2.0449629820347466e-05}
{"train_loss": 0.0672377496957779, "global_step": 190509, "epoch": 2140, "lr": 2.0449162168450286e-05}
{"train_loss": 0.0862618014216423, "global_step": 190510, "epoch": 2140, "lr": 2.0448694520525836e-05}
{"train_loss": 0.054706718772649765, "global_step": 190511, "epoch": 2140, "lr": 2.044822687657421e-05}
{"train_loss": 0.09438912570476532, "global_step": 190512, "epoch": 2140, "lr": 2.044775923659547e-05}
{"train_loss": 0.035702336579561234, "global_step": 190513, "epoch": 2140, "lr": 2.0447291600589645e-05}
{"train_loss": 0.07468432933092117, "global_step": 190514, "epoch": 2140, "lr": 2.0446823968556837e-05}
{"train_loss": 0.06671465933322906, "global_step": 190515, "epoch": 2140, "lr": 2.044635634049708e-05}
{"train_loss": 0.08095219731330872, "global_step": 190516, "epoch": 2140, "lr": 2.0445888716410473e-05}
{"train_loss": 0.03644104301929474, "global_step": 190517, "epoch": 2140, "lr": 2.044542109629703e-05}
{"train_loss": 0.06373032182455063, "global_step": 190518, "epoch": 2140, "lr": 2.044495348015686e-05}
{"train_loss": 0.0805303081870079, "global_step": 190519, "epoch": 2140, "lr": 2.044448586798999e-05}
{"train_loss": 0.06839893758296967, "global_step": 190520, "epoch": 2140, "lr": 2.044401825979651e-05}
{"train_loss": 0.0939163938164711, "global_step": 190521, "epoch": 2140, "lr": 2.044355065557646e-05}
{"train_loss": 0.0731988251209259, "global_step": 190522, "epoch": 2140, "lr": 2.044308305532993e-05}
{"train_loss": 0.06845518946647644, "global_step": 190523, "epoch": 2140, "lr": 2.044261545905695e-05}
{"train_loss": 0.11973647028207779, "global_step": 190524, "epoch": 2140, "lr": 2.0442147866757618e-05}
{"train_loss": 0.1000688225030899, "global_step": 190525, "epoch": 2140, "lr": 2.0441680278431964e-05}
{"train_loss": 0.09303679317235947, "global_step": 190526, "epoch": 2140, "lr": 2.0441212694080074e-05}
{"train_loss": 0.08392621576786041, "global_step": 190527, "epoch": 2140, "lr": 2.044074511370199e-05}
{"train_loss": 0.09461317211389542, "global_step": 190528, "epoch": 2140, "lr": 2.0440277537297815e-05}
{"train_loss": 0.05899054557085037, "global_step": 190529, "epoch": 2140, "lr": 2.0439809964867558e-05}
{"train_loss": 0.057154860347509384, "global_step": 190530, "epoch": 2140, "lr": 2.0439342396411322e-05}
{"train_loss": 0.12125545740127563, "global_step": 190531, "epoch": 2140, "lr": 2.0438874831929144e-05}
{"train_loss": 0.061329908668994904, "global_step": 190532, "epoch": 2140, "lr": 2.0438407271421096e-05}
{"train_loss": 0.04201928526163101, "global_step": 190533, "epoch": 2140, "lr": 2.0437939714887266e-05}
{"train_loss": 0.08459071815013885, "global_step": 190534, "epoch": 2140, "lr": 2.043747216232767e-05}
{"train_loss": 0.09252458065748215, "global_step": 190535, "epoch": 2140, "lr": 2.0437004613742423e-05}
{"train_loss": 0.05893179774284363, "global_step": 190536, "epoch": 2140, "lr": 2.0436537069131534e-05}
{"train_loss": 0.04748702421784401, "global_step": 190537, "epoch": 2140, "lr": 2.0436069528495112e-05}
{"train_loss": 0.04549746960401535, "global_step": 190538, "epoch": 2140, "lr": 2.0435601991833186e-05}
{"train_loss": 0.05511372908949852, "global_step": 190539, "epoch": 2140, "lr": 2.043513445914585e-05}
{"train_loss": 0.06890568137168884, "global_step": 190540, "epoch": 2140, "lr": 2.043466693043313e-05}
{"train_loss": 0.0918617993593216, "global_step": 190541, "epoch": 2140, "lr": 2.0434199405695126e-05}
{"train_loss": 0.05562705174088478, "global_step": 190542, "epoch": 2140, "lr": 2.0433731884931866e-05}
{"train_loss": 0.042873285710811615, "global_step": 190543, "epoch": 2140, "lr": 2.0433264368143444e-05}
{"train_loss": 0.06869041919708252, "global_step": 190544, "epoch": 2140, "lr": 2.04327968553299e-05}
{"train_loss": 0.08983344584703445, "global_step": 190545, "epoch": 2140, "lr": 2.043232934649132e-05}
{"train_loss": 0.03807757794857025, "global_step": 190546, "epoch": 2140, "lr": 2.0431861841627735e-05}
{"train_loss": 0.11117877066135406, "global_step": 190547, "epoch": 2140, "lr": 2.0431394340739246e-05}
{"train_loss": 0.07124490822466571, "global_step": 190548, "epoch": 2140, "lr": 2.043092684382587e-05, "val_loss": 7.924128532409668, "train_action_mse_error": 23.58074951171875}
{"train_loss": 0.05982782691717148, "global_step": 190549, "epoch": 2141, "lr": 2.043045935088772e-05}
{"train_loss": 0.041091129183769226, "global_step": 190550, "epoch": 2141, "lr": 2.0429991861924835e-05}
{"train_loss": 0.15778842568397522, "global_step": 190551, "epoch": 2141, "lr": 2.0429524376937258e-05}
{"train_loss": 0.08053379505872726, "global_step": 190552, "epoch": 2141, "lr": 2.0429056895925085e-05}
{"train_loss": 0.027702802792191505, "global_step": 190553, "epoch": 2141, "lr": 2.0428589418888356e-05}
{"train_loss": 0.0749090313911438, "global_step": 190554, "epoch": 2141, "lr": 2.042812194582715e-05}
{"train_loss": 0.09112513810396194, "global_step": 190555, "epoch": 2141, "lr": 2.0427654476741515e-05}
{"train_loss": 0.06835799664258957, "global_step": 190556, "epoch": 2141, "lr": 2.0427187011631532e-05}
{"train_loss": 0.11471527814865112, "global_step": 190557, "epoch": 2141, "lr": 2.0426719550497237e-05}
{"train_loss": 0.09945034235715866, "global_step": 190558, "epoch": 2141, "lr": 2.0426252093338728e-05}
{"train_loss": 0.06544245034456253, "global_step": 190559, "epoch": 2141, "lr": 2.0425784640156032e-05}
{"train_loss": 0.055801860988140106, "global_step": 190560, "epoch": 2141, "lr": 2.0425317190949245e-05}
{"train_loss": 0.026658717542886734, "global_step": 190561, "epoch": 2141, "lr": 2.0424849745718393e-05}
{"train_loss": 0.04410996288061142, "global_step": 190562, "epoch": 2141, "lr": 2.042438230446358e-05}
{"train_loss": 0.05987703055143356, "global_step": 190563, "epoch": 2141, "lr": 2.0423914867184835e-05}
{"train_loss": 0.11668851226568222, "global_step": 190564, "epoch": 2141, "lr": 2.0423447433882243e-05}
{"train_loss": 0.061708442866802216, "global_step": 190565, "epoch": 2141, "lr": 2.0422980004555847e-05}
{"train_loss": 0.03980022296309471, "global_step": 190566, "epoch": 2141, "lr": 2.042251257920574e-05}
{"train_loss": 0.09865804016590118, "global_step": 190567, "epoch": 2141, "lr": 2.0422045157831944e-05}
{"train_loss": 0.04529692977666855, "global_step": 190568, "epoch": 2141, "lr": 2.042157774043454e-05}
{"train_loss": 0.07329387962818146, "global_step": 190569, "epoch": 2141, "lr": 2.0421110327013614e-05}
{"train_loss": 0.06943745166063309, "global_step": 190570, "epoch": 2141, "lr": 2.04206429175692e-05}
{"train_loss": 0.04168263077735901, "global_step": 190571, "epoch": 2141, "lr": 2.042017551210138e-05}
{"train_loss": 0.09934742003679276, "global_step": 190572, "epoch": 2141, "lr": 2.0419708110610186e-05}
{"train_loss": 0.05688006430864334, "global_step": 190573, "epoch": 2141, "lr": 2.041924071309573e-05}
{"train_loss": 0.10121899098157883, "global_step": 190574, "epoch": 2141, "lr": 2.041877331955802e-05}
{"train_loss": 0.049492597579956055, "global_step": 190575, "epoch": 2141, "lr": 2.0418305929997166e-05}
{"train_loss": 0.08050782233476639, "global_step": 190576, "epoch": 2141, "lr": 2.041783854441319e-05}
{"train_loss": 0.05754108726978302, "global_step": 190577, "epoch": 2141, "lr": 2.04173711628062e-05}
{"train_loss": 0.04902172461152077, "global_step": 190578, "epoch": 2141, "lr": 2.0416903785176207e-05}
{"train_loss": 0.05655772611498833, "global_step": 190579, "epoch": 2141, "lr": 2.0416436411523326e-05}
{"train_loss": 0.028739774599671364, "global_step": 190580, "epoch": 2141, "lr": 2.0415969041847572e-05}
{"train_loss": 0.08284439891576767, "global_step": 190581, "epoch": 2141, "lr": 2.0415501676149052e-05}
{"train_loss": 0.04811700060963631, "global_step": 190582, "epoch": 2141, "lr": 2.0415034314427785e-05}
{"train_loss": 0.08182011544704437, "global_step": 190583, "epoch": 2141, "lr": 2.041456695668388e-05}
{"train_loss": 0.06130485236644745, "global_step": 190584, "epoch": 2141, "lr": 2.0414099602917353e-05}
{"train_loss": 0.03207888454198837, "global_step": 190585, "epoch": 2141, "lr": 2.0413632253128307e-05}
{"train_loss": 0.07825919985771179, "global_step": 190586, "epoch": 2141, "lr": 2.0413164907316768e-05}
{"train_loss": 0.09010975062847137, "global_step": 190587, "epoch": 2141, "lr": 2.041269756548284e-05}
{"train_loss": 0.13188901543617249, "global_step": 190588, "epoch": 2141, "lr": 2.0412230227626548e-05}
{"train_loss": 0.09093252569437027, "global_step": 190589, "epoch": 2141, "lr": 2.041176289374799e-05}
{"train_loss": 0.08231989294290543, "global_step": 190590, "epoch": 2141, "lr": 2.0411295563847206e-05}
{"train_loss": 0.05704563856124878, "global_step": 190591, "epoch": 2141, "lr": 2.041082823792424e-05}
{"train_loss": 0.06768900901079178, "global_step": 190592, "epoch": 2141, "lr": 2.0410360915979204e-05}
{"train_loss": 0.06299228966236115, "global_step": 190593, "epoch": 2141, "lr": 2.0409893598012115e-05}
{"train_loss": 0.08203664422035217, "global_step": 190594, "epoch": 2141, "lr": 2.040942628402307e-05}
{"train_loss": 0.028552867472171783, "global_step": 190595, "epoch": 2141, "lr": 2.04089589740121e-05}
{"train_loss": 0.0915098562836647, "global_step": 190596, "epoch": 2141, "lr": 2.04084916679793e-05}
{"train_loss": 0.05597323179244995, "global_step": 190597, "epoch": 2141, "lr": 2.0408024365924705e-05}
{"train_loss": 0.07595096528530121, "global_step": 190598, "epoch": 2141, "lr": 2.0407557067848403e-05}
{"train_loss": 0.04775819182395935, "global_step": 190599, "epoch": 2141, "lr": 2.040708977375043e-05}
{"train_loss": 0.07823609560728073, "global_step": 190600, "epoch": 2141, "lr": 2.0406622483630882e-05}
{"train_loss": 0.05966559052467346, "global_step": 190601, "epoch": 2141, "lr": 2.0406155197489785e-05}
{"train_loss": 0.07372363656759262, "global_step": 190602, "epoch": 2141, "lr": 2.0405687915327236e-05}
{"train_loss": 0.0587286539375782, "global_step": 190603, "epoch": 2141, "lr": 2.0405220637143262e-05}
{"train_loss": 0.08236264437437057, "global_step": 190604, "epoch": 2141, "lr": 2.040475336293795e-05}
{"train_loss": 0.07847287505865097, "global_step": 190605, "epoch": 2141, "lr": 2.0404286092711373e-05}
{"train_loss": 0.07902035117149353, "global_step": 190606, "epoch": 2141, "lr": 2.040381882646356e-05}
{"train_loss": 0.03116193599998951, "global_step": 190607, "epoch": 2141, "lr": 2.040335156419461e-05}
{"train_loss": 0.0645608976483345, "global_step": 190608, "epoch": 2141, "lr": 2.0402884305904557e-05}
{"train_loss": 0.12332304567098618, "global_step": 190609, "epoch": 2141, "lr": 2.0402417051593488e-05}
{"train_loss": 0.08845008164644241, "global_step": 190610, "epoch": 2141, "lr": 2.0401949801261434e-05}
{"train_loss": 0.051773570477962494, "global_step": 190611, "epoch": 2141, "lr": 2.0401482554908497e-05}
{"train_loss": 0.08098982274532318, "global_step": 190612, "epoch": 2141, "lr": 2.0401015312534704e-05}
{"train_loss": 0.05278346687555313, "global_step": 190613, "epoch": 2141, "lr": 2.0400548074140154e-05}
{"train_loss": 0.05172296240925789, "global_step": 190614, "epoch": 2141, "lr": 2.0400080839724866e-05}
{"train_loss": 0.07787229865789413, "global_step": 190615, "epoch": 2141, "lr": 2.039961360928894e-05}
{"train_loss": 0.04167568311095238, "global_step": 190616, "epoch": 2141, "lr": 2.039914638283242e-05}
{"train_loss": 0.032879848033189774, "global_step": 190617, "epoch": 2141, "lr": 2.0398679160355383e-05}
{"train_loss": 0.08704010397195816, "global_step": 190618, "epoch": 2141, "lr": 2.039821194185787e-05}
{"train_loss": 0.07166143506765366, "global_step": 190619, "epoch": 2141, "lr": 2.0397744727339968e-05}
{"train_loss": 0.08920454978942871, "global_step": 190620, "epoch": 2141, "lr": 2.0397277516801716e-05}
{"train_loss": 0.07111728936433792, "global_step": 190621, "epoch": 2141, "lr": 2.0396810310243203e-05}
{"train_loss": 0.03920046240091324, "global_step": 190622, "epoch": 2141, "lr": 2.0396343107664468e-05}
{"train_loss": 0.03351220861077309, "global_step": 190623, "epoch": 2141, "lr": 2.0395875909065592e-05}
{"train_loss": 0.09601275622844696, "global_step": 190624, "epoch": 2141, "lr": 2.039540871444662e-05}
{"train_loss": 0.034187767654657364, "global_step": 190625, "epoch": 2141, "lr": 2.0394941523807637e-05}
{"train_loss": 0.04107879102230072, "global_step": 190626, "epoch": 2141, "lr": 2.039447433714868e-05}
{"train_loss": 0.0964687392115593, "global_step": 190627, "epoch": 2141, "lr": 2.0394007154469835e-05}
{"train_loss": 0.06907179206609726, "global_step": 190628, "epoch": 2141, "lr": 2.0393539975771158e-05}
{"train_loss": 0.05515817180275917, "global_step": 190629, "epoch": 2141, "lr": 2.0393072801052692e-05}
{"train_loss": 0.07175027579069138, "global_step": 190630, "epoch": 2141, "lr": 2.039260563031454e-05}
{"train_loss": 0.0603313185274601, "global_step": 190631, "epoch": 2141, "lr": 2.0392138463556715e-05}
{"train_loss": 0.10286915302276611, "global_step": 190632, "epoch": 2141, "lr": 2.0391671300779323e-05}
{"train_loss": 0.04889257624745369, "global_step": 190633, "epoch": 2141, "lr": 2.03912041419824e-05}
{"train_loss": 0.09281104803085327, "global_step": 190634, "epoch": 2141, "lr": 2.039073698716603e-05}
{"train_loss": 0.04685027897357941, "global_step": 190635, "epoch": 2141, "lr": 2.0390269836330254e-05}
{"train_loss": 0.07625514268875122, "global_step": 190636, "epoch": 2141, "lr": 2.0389802689475157e-05}
{"train_loss": 0.06857672005138371, "global_step": 190637, "epoch": 2141, "lr": 2.0389335546600773e-05, "val_loss": 7.972393035888672}
{"train_loss": 0.076123908162117, "global_step": 190638, "epoch": 2142, "lr": 2.03888684077072e-05}
{"train_loss": 0.13119065761566162, "global_step": 190639, "epoch": 2142, "lr": 2.0388401272794468e-05}
{"train_loss": 0.06834990531206131, "global_step": 190640, "epoch": 2142, "lr": 2.038793414186265e-05}
{"train_loss": 0.039166409522295, "global_step": 190641, "epoch": 2142, "lr": 2.0387467014911836e-05}
{"train_loss": 0.08849304169416428, "global_step": 190642, "epoch": 2142, "lr": 2.0386999891942045e-05}
{"train_loss": 0.05235876515507698, "global_step": 190643, "epoch": 2142, "lr": 2.0386532772953383e-05}
{"train_loss": 0.0749274492263794, "global_step": 190644, "epoch": 2142, "lr": 2.0386065657945868e-05}
{"train_loss": 0.10712351649999619, "global_step": 190645, "epoch": 2142, "lr": 2.0385598546919604e-05}
{"train_loss": 0.0524444505572319, "global_step": 190646, "epoch": 2142, "lr": 2.0385131439874615e-05}
{"train_loss": 0.04878365620970726, "global_step": 190647, "epoch": 2142, "lr": 2.038466433681101e-05}
{"train_loss": 0.04963691160082817, "global_step": 190648, "epoch": 2142, "lr": 2.03841972377288e-05}
{"train_loss": 0.04304513335227966, "global_step": 190649, "epoch": 2142, "lr": 2.0383730142628095e-05}
{"train_loss": 0.046284522861242294, "global_step": 190650, "epoch": 2142, "lr": 2.038326305150892e-05}
{"train_loss": 0.06267677247524261, "global_step": 190651, "epoch": 2142, "lr": 2.038279596437137e-05}
{"train_loss": 0.05770708993077278, "global_step": 190652, "epoch": 2142, "lr": 2.0382328881215473e-05}
{"train_loss": 0.08002495765686035, "global_step": 190653, "epoch": 2142, "lr": 2.038186180204133e-05}
{"train_loss": 0.09645047038793564, "global_step": 190654, "epoch": 2142, "lr": 2.038139472684897e-05}
{"train_loss": 0.10396986454725266, "global_step": 190655, "epoch": 2142, "lr": 2.0380927655638488e-05}
{"train_loss": 0.03549715876579285, "global_step": 190656, "epoch": 2142, "lr": 2.0380460588409905e-05}
{"train_loss": 0.036013636738061905, "global_step": 190657, "epoch": 2142, "lr": 2.0379993525163333e-05}
{"train_loss": 0.08373601734638214, "global_step": 190658, "epoch": 2142, "lr": 2.0379526465898792e-05}
{"train_loss": 0.09089194238185883, "global_step": 190659, "epoch": 2142, "lr": 2.037905941061636e-05}
{"train_loss": 0.1467384397983551, "global_step": 190660, "epoch": 2142, "lr": 2.0378592359316116e-05}
{"train_loss": 0.06539104133844376, "global_step": 190661, "epoch": 2142, "lr": 2.0378125311998114e-05}
{"train_loss": 0.09777837991714478, "global_step": 190662, "epoch": 2142, "lr": 2.0377658268662393e-05}
{"train_loss": 0.04864764213562012, "global_step": 190663, "epoch": 2142, "lr": 2.037719122930905e-05}
{"train_loss": 0.033250704407691956, "global_step": 190664, "epoch": 2142, "lr": 2.0376724193938117e-05}
{"train_loss": 0.07447779178619385, "global_step": 190665, "epoch": 2142, "lr": 2.0376257162549694e-05}
{"train_loss": 0.04719860851764679, "global_step": 190666, "epoch": 2142, "lr": 2.0375790135143817e-05}
{"train_loss": 0.05971691757440567, "global_step": 190667, "epoch": 2142, "lr": 2.037532311172054e-05}
{"train_loss": 0.09964015334844589, "global_step": 190668, "epoch": 2142, "lr": 2.037485609227996e-05}
{"train_loss": 0.05539832264184952, "global_step": 190669, "epoch": 2142, "lr": 2.0374389076822093e-05}
{"train_loss": 0.050390686839818954, "global_step": 190670, "epoch": 2142, "lr": 2.0373922065347052e-05}
{"train_loss": 0.10777108371257782, "global_step": 190671, "epoch": 2142, "lr": 2.0373455057854855e-05}
{"train_loss": 0.06158654764294624, "global_step": 190672, "epoch": 2142, "lr": 2.037298805434561e-05}
{"train_loss": 0.1095457524061203, "global_step": 190673, "epoch": 2142, "lr": 2.037252105481933e-05}
{"train_loss": 0.09643159061670303, "global_step": 190674, "epoch": 2142, "lr": 2.0372054059276123e-05}
{"train_loss": 0.11725933849811554, "global_step": 190675, "epoch": 2142, "lr": 2.037158706771602e-05}
{"train_loss": 0.09560245275497437, "global_step": 190676, "epoch": 2142, "lr": 2.037112008013909e-05}
{"train_loss": 0.09056682884693146, "global_step": 190677, "epoch": 2142, "lr": 2.037065309654542e-05}
{"train_loss": 0.04159093648195267, "global_step": 190678, "epoch": 2142, "lr": 2.037018611693504e-05}
{"train_loss": 0.05481136590242386, "global_step": 190679, "epoch": 2142, "lr": 2.0369719141308048e-05}
{"train_loss": 0.053396254777908325, "global_step": 190680, "epoch": 2142, "lr": 2.036925216966446e-05}
{"train_loss": 0.06564699858427048, "global_step": 190681, "epoch": 2142, "lr": 2.0368785202004388e-05}
{"train_loss": 0.061465151607990265, "global_step": 190682, "epoch": 2142, "lr": 2.0368318238327853e-05}
{"train_loss": 0.06828895956277847, "global_step": 190683, "epoch": 2142, "lr": 2.0367851278634952e-05}
{"train_loss": 0.04926760122179985, "global_step": 190684, "epoch": 2142, "lr": 2.036738432292572e-05}
{"train_loss": 0.06720389425754547, "global_step": 190685, "epoch": 2142, "lr": 2.0366917371200246e-05}
{"train_loss": 0.0685308575630188, "global_step": 190686, "epoch": 2142, "lr": 2.036645042345856e-05}
{"train_loss": 0.14274513721466064, "global_step": 190687, "epoch": 2142, "lr": 2.0365983479700763e-05}
{"train_loss": 0.06787687540054321, "global_step": 190688, "epoch": 2142, "lr": 2.0365516539926875e-05}
{"train_loss": 0.05317971110343933, "global_step": 190689, "epoch": 2142, "lr": 2.036504960413701e-05}
{"train_loss": 0.05719014257192612, "global_step": 190690, "epoch": 2142, "lr": 2.0364582672331178e-05}
{"train_loss": 0.04672853276133537, "global_step": 190691, "epoch": 2142, "lr": 2.0364115744509487e-05}
{"train_loss": 0.05538492649793625, "global_step": 190692, "epoch": 2142, "lr": 2.036364882067196e-05}
{"train_loss": 0.06889193505048752, "global_step": 190693, "epoch": 2142, "lr": 2.0363181900818696e-05}
{"train_loss": 0.06878366321325302, "global_step": 190694, "epoch": 2142, "lr": 2.0362714984949727e-05}
{"train_loss": 0.07558830827474594, "global_step": 190695, "epoch": 2142, "lr": 2.0362248073065127e-05}
{"train_loss": 0.0671602413058281, "global_step": 190696, "epoch": 2142, "lr": 2.0361781165164977e-05}
{"train_loss": 0.1160946935415268, "global_step": 190697, "epoch": 2142, "lr": 2.0361314261249314e-05}
{"train_loss": 0.05309327319264412, "global_step": 190698, "epoch": 2142, "lr": 2.0360847361318226e-05}
{"train_loss": 0.04648396372795105, "global_step": 190699, "epoch": 2142, "lr": 2.0360380465371744e-05}
{"train_loss": 0.05999079719185829, "global_step": 190700, "epoch": 2142, "lr": 2.0359913573409962e-05}
{"train_loss": 0.07820945233106613, "global_step": 190701, "epoch": 2142, "lr": 2.0359446685432925e-05}
{"train_loss": 0.0966123715043068, "global_step": 190702, "epoch": 2142, "lr": 2.0358979801440686e-05}
{"train_loss": 0.06765551120042801, "global_step": 190703, "epoch": 2142, "lr": 2.0358512921433338e-05}
{"train_loss": 0.07634900510311127, "global_step": 190704, "epoch": 2142, "lr": 2.035804604541091e-05}
{"train_loss": 0.021253179758787155, "global_step": 190705, "epoch": 2142, "lr": 2.0357579173373498e-05}
{"train_loss": 0.08606831729412079, "global_step": 190706, "epoch": 2142, "lr": 2.0357112305321145e-05}
{"train_loss": 0.08250070363283157, "global_step": 190707, "epoch": 2142, "lr": 2.0356645441253907e-05}
{"train_loss": 0.05558660998940468, "global_step": 190708, "epoch": 2142, "lr": 2.035617858117187e-05}
{"train_loss": 0.07647114247083664, "global_step": 190709, "epoch": 2142, "lr": 2.0355711725075066e-05}
{"train_loss": 0.05779130011796951, "global_step": 190710, "epoch": 2142, "lr": 2.0355244872963592e-05}
{"train_loss": 0.08229716867208481, "global_step": 190711, "epoch": 2142, "lr": 2.035477802483748e-05}
{"train_loss": 0.05054716393351555, "global_step": 190712, "epoch": 2142, "lr": 2.03543111806968e-05}
{"train_loss": 0.05559604987502098, "global_step": 190713, "epoch": 2142, "lr": 2.0353844340541643e-05}
{"train_loss": 0.08045372366905212, "global_step": 190714, "epoch": 2142, "lr": 2.0353377504372034e-05}
{"train_loss": 0.10596221685409546, "global_step": 190715, "epoch": 2142, "lr": 2.0352910672188068e-05}
{"train_loss": 0.05060352385044098, "global_step": 190716, "epoch": 2142, "lr": 2.0352443843989778e-05}
{"train_loss": 0.11646363139152527, "global_step": 190717, "epoch": 2142, "lr": 2.0351977019777252e-05}
{"train_loss": 0.0584939606487751, "global_step": 190718, "epoch": 2142, "lr": 2.0351510199550528e-05}
{"train_loss": 0.11602253466844559, "global_step": 190719, "epoch": 2142, "lr": 2.0351043383309696e-05}
{"train_loss": 0.06427088379859924, "global_step": 190720, "epoch": 2142, "lr": 2.0350576571054792e-05}
{"train_loss": 0.06371066719293594, "global_step": 190721, "epoch": 2142, "lr": 2.0350109762785907e-05}
{"train_loss": 0.05100591480731964, "global_step": 190722, "epoch": 2142, "lr": 2.0349642958503067e-05}
{"train_loss": 0.10077878087759018, "global_step": 190723, "epoch": 2142, "lr": 2.034917615820638e-05}
{"train_loss": 0.10786918550729752, "global_step": 190724, "epoch": 2142, "lr": 2.0348709361895862e-05}
{"train_loss": 0.029594140127301216, "global_step": 190725, "epoch": 2142, "lr": 2.0348242569571623e-05}
{"train_loss": 0.07242618118193042, "global_step": 190726, "epoch": 2142, "lr": 2.034777578123368e-05, "val_loss": 7.921964168548584}
{"train_loss": 0.04392705485224724, "global_step": 190727, "epoch": 2143, "lr": 2.034730899688213e-05}
{"train_loss": 0.12073993682861328, "global_step": 190728, "epoch": 2143, "lr": 2.0346842216517015e-05}
{"train_loss": 0.05752257630228996, "global_step": 190729, "epoch": 2143, "lr": 2.0346375440138415e-05}
{"train_loss": 0.09606343507766724, "global_step": 190730, "epoch": 2143, "lr": 2.0345908667746365e-05}
{"train_loss": 0.05085252597928047, "global_step": 190731, "epoch": 2143, "lr": 2.0345441899340954e-05}
{"train_loss": 0.03396983444690704, "global_step": 190732, "epoch": 2143, "lr": 2.034497513492225e-05}
{"train_loss": 0.119549959897995, "global_step": 190733, "epoch": 2143, "lr": 2.0344508374490285e-05}
{"train_loss": 0.0647040456533432, "global_step": 190734, "epoch": 2143, "lr": 2.0344041618045155e-05}
{"train_loss": 0.049329403787851334, "global_step": 190735, "epoch": 2143, "lr": 2.0343574865586894e-05}
{"train_loss": 0.10322076082229614, "global_step": 190736, "epoch": 2143, "lr": 2.0343108117115594e-05}
{"train_loss": 0.04205334559082985, "global_step": 190737, "epoch": 2143, "lr": 2.0342641372631284e-05}
{"train_loss": 0.03870965540409088, "global_step": 190738, "epoch": 2143, "lr": 2.034217463213406e-05}
{"train_loss": 0.0824543759226799, "global_step": 190739, "epoch": 2143, "lr": 2.0341707895623973e-05}
{"train_loss": 0.07369297742843628, "global_step": 190740, "epoch": 2143, "lr": 2.034124116310106e-05}
{"train_loss": 0.07150967419147491, "global_step": 190741, "epoch": 2143, "lr": 2.0340774434565424e-05}
{"train_loss": 0.04186319187283516, "global_step": 190742, "epoch": 2143, "lr": 2.0340307710017092e-05}
{"train_loss": 0.05034413933753967, "global_step": 190743, "epoch": 2143, "lr": 2.0339840989456164e-05}
{"train_loss": 0.07009846717119217, "global_step": 190744, "epoch": 2143, "lr": 2.033937427288268e-05}
{"train_loss": 0.05611180141568184, "global_step": 190745, "epoch": 2143, "lr": 2.0338907560296694e-05}
{"train_loss": 0.08807166665792465, "global_step": 190746, "epoch": 2143, "lr": 2.0338440851698294e-05}
{"train_loss": 0.07635623961687088, "global_step": 190747, "epoch": 2143, "lr": 2.0337974147087508e-05}
{"train_loss": 0.04217790067195892, "global_step": 190748, "epoch": 2143, "lr": 2.0337507446464438e-05}
{"train_loss": 0.08189962059259415, "global_step": 190749, "epoch": 2143, "lr": 2.0337040749829118e-05}
{"train_loss": 0.07936734706163406, "global_step": 190750, "epoch": 2143, "lr": 2.0336574057181612e-05}
{"train_loss": 0.12272784858942032, "global_step": 190751, "epoch": 2143, "lr": 2.033610736852202e-05}
{"train_loss": 0.07207003235816956, "global_step": 190752, "epoch": 2143, "lr": 2.0335640683850347e-05}
{"train_loss": 0.05780956894159317, "global_step": 190753, "epoch": 2143, "lr": 2.0335174003166714e-05}
{"train_loss": 0.06539005041122437, "global_step": 190754, "epoch": 2143, "lr": 2.0334707326471132e-05}
{"train_loss": 0.046036962419748306, "global_step": 190755, "epoch": 2143, "lr": 2.033424065376371e-05}
{"train_loss": 0.10354612022638321, "global_step": 190756, "epoch": 2143, "lr": 2.033377398504446e-05}
{"train_loss": 0.07611596584320068, "global_step": 190757, "epoch": 2143, "lr": 2.0333307320313498e-05}
{"train_loss": 0.11495980620384216, "global_step": 190758, "epoch": 2143, "lr": 2.033284065957084e-05}
{"train_loss": 0.05807507410645485, "global_step": 190759, "epoch": 2143, "lr": 2.033237400281659e-05}
{"train_loss": 0.11568322032690048, "global_step": 190760, "epoch": 2143, "lr": 2.033190735005078e-05}
{"train_loss": 0.10030992329120636, "global_step": 190761, "epoch": 2143, "lr": 2.033144070127349e-05}
{"train_loss": 0.1149345263838768, "global_step": 190762, "epoch": 2143, "lr": 2.033097405648476e-05}
{"train_loss": 0.14221370220184326, "global_step": 190763, "epoch": 2143, "lr": 2.0330507415684692e-05}
{"train_loss": 0.08388687670230865, "global_step": 190764, "epoch": 2143, "lr": 2.0330040778873303e-05}
{"train_loss": 0.079254150390625, "global_step": 190765, "epoch": 2143, "lr": 2.0329574146050694e-05}
{"train_loss": 0.03618406504392624, "global_step": 190766, "epoch": 2143, "lr": 2.03291075172169e-05}
{"train_loss": 0.1304638832807541, "global_step": 190767, "epoch": 2143, "lr": 2.0328640892372e-05}
{"train_loss": 0.07366638630628586, "global_step": 190768, "epoch": 2143, "lr": 2.032817427151606e-05}
{"train_loss": 0.03749620169401169, "global_step": 190769, "epoch": 2143, "lr": 2.0327707654649126e-05}
{"train_loss": 0.06862969696521759, "global_step": 190770, "epoch": 2143, "lr": 2.0327241041771284e-05}
{"train_loss": 0.09467621147632599, "global_step": 190771, "epoch": 2143, "lr": 2.0326774432882563e-05}
{"train_loss": 0.10082481056451797, "global_step": 190772, "epoch": 2143, "lr": 2.0326307827983066e-05}
{"train_loss": 0.08680487424135208, "global_step": 190773, "epoch": 2143, "lr": 2.0325841227072813e-05}
{"train_loss": 0.1219262033700943, "global_step": 190774, "epoch": 2143, "lr": 2.0325374630151912e-05}
{"train_loss": 0.048763666301965714, "global_step": 190775, "epoch": 2143, "lr": 2.032490803722038e-05}
{"train_loss": 0.04891559109091759, "global_step": 190776, "epoch": 2143, "lr": 2.0324441448278327e-05}
{"train_loss": 0.044384051114320755, "global_step": 190777, "epoch": 2143, "lr": 2.032397486332579e-05}
{"train_loss": 0.060638681054115295, "global_step": 190778, "epoch": 2143, "lr": 2.032350828236281e-05}
{"train_loss": 0.036638036370277405, "global_step": 190779, "epoch": 2143, "lr": 2.0323041705389496e-05}
{"train_loss": 0.07379265129566193, "global_step": 190780, "epoch": 2143, "lr": 2.0322575132405862e-05}
{"train_loss": 0.10961863398551941, "global_step": 190781, "epoch": 2143, "lr": 2.0322108563412024e-05}
{"train_loss": 0.10765431821346283, "global_step": 190782, "epoch": 2143, "lr": 2.032164199840801e-05}
{"train_loss": 0.11767580360174179, "global_step": 190783, "epoch": 2143, "lr": 2.0321175437393868e-05}
{"train_loss": 0.06347307562828064, "global_step": 190784, "epoch": 2143, "lr": 2.032070888036971e-05}
{"train_loss": 0.10536722838878632, "global_step": 190785, "epoch": 2143, "lr": 2.0320242327335548e-05}
{"train_loss": 0.058714885264635086, "global_step": 190786, "epoch": 2143, "lr": 2.0319775778291466e-05}
{"train_loss": 0.08246807008981705, "global_step": 190787, "epoch": 2143, "lr": 2.0319309233237547e-05}
{"train_loss": 0.08867810666561127, "global_step": 190788, "epoch": 2143, "lr": 2.0318842692173818e-05}
{"train_loss": 0.08418715745210648, "global_step": 190789, "epoch": 2143, "lr": 2.031837615510038e-05}
{"train_loss": 0.10732376575469971, "global_step": 190790, "epoch": 2143, "lr": 2.031790962201725e-05}
{"train_loss": 0.045972730964422226, "global_step": 190791, "epoch": 2143, "lr": 2.0317443092924536e-05}
{"train_loss": 0.056401483714580536, "global_step": 190792, "epoch": 2143, "lr": 2.0316976567822265e-05}
{"train_loss": 0.09433592110872269, "global_step": 190793, "epoch": 2143, "lr": 2.0316510046710535e-05}
{"train_loss": 0.06087639555335045, "global_step": 190794, "epoch": 2143, "lr": 2.0316043529589364e-05}
{"train_loss": 0.07463005185127258, "global_step": 190795, "epoch": 2143, "lr": 2.031557701645886e-05}
{"train_loss": 0.09033934026956558, "global_step": 190796, "epoch": 2143, "lr": 2.0315110507319046e-05}
{"train_loss": 0.06127331405878067, "global_step": 190797, "epoch": 2143, "lr": 2.0314644002170018e-05}
{"train_loss": 0.12689578533172607, "global_step": 190798, "epoch": 2143, "lr": 2.031417750101181e-05}
{"train_loss": 0.10442338138818741, "global_step": 190799, "epoch": 2143, "lr": 2.0313711003844522e-05}
{"train_loss": 0.05826129764318466, "global_step": 190800, "epoch": 2143, "lr": 2.0313244510668165e-05}
{"train_loss": 0.06705427169799805, "global_step": 190801, "epoch": 2143, "lr": 2.031277802148286e-05}
{"train_loss": 0.11186262965202332, "global_step": 190802, "epoch": 2143, "lr": 2.0312311536288613e-05}
{"train_loss": 0.07765594124794006, "global_step": 190803, "epoch": 2143, "lr": 2.031184505508552e-05}
{"train_loss": 0.07380030304193497, "global_step": 190804, "epoch": 2143, "lr": 2.0311378577873652e-05}
{"train_loss": 0.06032169610261917, "global_step": 190805, "epoch": 2143, "lr": 2.031091210465304e-05}
{"train_loss": 0.11793692409992218, "global_step": 190806, "epoch": 2143, "lr": 2.031044563542378e-05}
{"train_loss": 0.12386807799339294, "global_step": 190807, "epoch": 2143, "lr": 2.0309979170185907e-05}
{"train_loss": 0.034225836396217346, "global_step": 190808, "epoch": 2143, "lr": 2.0309512708939504e-05}
{"train_loss": 0.06519024074077606, "global_step": 190809, "epoch": 2143, "lr": 2.0309046251684616e-05}
{"train_loss": 0.06773581355810165, "global_step": 190810, "epoch": 2143, "lr": 2.030857979842133e-05}
{"train_loss": 0.08184020221233368, "global_step": 190811, "epoch": 2143, "lr": 2.030811334914967e-05}
{"train_loss": 0.0973508283495903, "global_step": 190812, "epoch": 2143, "lr": 2.0307646903869747e-05}
{"train_loss": 0.05861847475171089, "global_step": 190813, "epoch": 2143, "lr": 2.0307180462581582e-05}
{"train_loss": 0.0995829850435257, "global_step": 190814, "epoch": 2143, "lr": 2.0306714025285267e-05}
{"train_loss": 0.07828422317679008, "global_step": 190815, "epoch": 2143, "lr": 2.0306247591980838e-05, "val_loss": 7.824185848236084}
{"train_loss": 0.08241745084524155, "global_step": 190816, "epoch": 2144, "lr": 2.030578116266839e-05}
{"train_loss": 0.10311292111873627, "global_step": 190817, "epoch": 2144, "lr": 2.030531473734797e-05}
{"train_loss": 0.07925047725439072, "global_step": 190818, "epoch": 2144, "lr": 2.0304848316019614e-05}
{"train_loss": 0.06648533046245575, "global_step": 190819, "epoch": 2144, "lr": 2.0304381898683434e-05}
{"train_loss": 0.06151118129491806, "global_step": 190820, "epoch": 2144, "lr": 2.030391548533944e-05}
{"train_loss": 0.05375032499432564, "global_step": 190821, "epoch": 2144, "lr": 2.0303449075987753e-05}
{"train_loss": 0.07080303877592087, "global_step": 190822, "epoch": 2144, "lr": 2.0302982670628378e-05}
{"train_loss": 0.028607457876205444, "global_step": 190823, "epoch": 2144, "lr": 2.0302516269261428e-05}
{"train_loss": 0.09235460311174393, "global_step": 190824, "epoch": 2144, "lr": 2.0302049871886924e-05}
{"train_loss": 0.06475389748811722, "global_step": 190825, "epoch": 2144, "lr": 2.0301583478504964e-05}
{"train_loss": 0.0957103744149208, "global_step": 190826, "epoch": 2144, "lr": 2.030111708911558e-05}
{"train_loss": 0.07439485937356949, "global_step": 190827, "epoch": 2144, "lr": 2.030065070371886e-05}
{"train_loss": 0.041800353676080704, "global_step": 190828, "epoch": 2144, "lr": 2.0300184322314843e-05}
{"train_loss": 0.07133342325687408, "global_step": 190829, "epoch": 2144, "lr": 2.0299717944903618e-05}
{"train_loss": 0.07866054773330688, "global_step": 190830, "epoch": 2144, "lr": 2.029925157148522e-05}
{"train_loss": 0.08586277067661285, "global_step": 190831, "epoch": 2144, "lr": 2.0298785202059745e-05}
{"train_loss": 0.050028763711452484, "global_step": 190832, "epoch": 2144, "lr": 2.0298318836627213e-05}
{"train_loss": 0.11974752694368362, "global_step": 190833, "epoch": 2144, "lr": 2.029785247518773e-05}
{"train_loss": 0.08381994068622589, "global_step": 190834, "epoch": 2144, "lr": 2.0297386117741325e-05}
{"train_loss": 0.06246582418680191, "global_step": 190835, "epoch": 2144, "lr": 2.0296919764288085e-05}
{"train_loss": 0.08614921569824219, "global_step": 190836, "epoch": 2144, "lr": 2.029645341482805e-05}
{"train_loss": 0.036308303475379944, "global_step": 190837, "epoch": 2144, "lr": 2.029598706936131e-05}
{"train_loss": 0.0628761425614357, "global_step": 190838, "epoch": 2144, "lr": 2.0295520727887894e-05}
{"train_loss": 0.0631466656923294, "global_step": 190839, "epoch": 2144, "lr": 2.0295054390407886e-05}
{"train_loss": 0.08263124525547028, "global_step": 190840, "epoch": 2144, "lr": 2.029458805692136e-05}
{"train_loss": 0.061731014400720596, "global_step": 190841, "epoch": 2144, "lr": 2.029412172742835e-05}
{"train_loss": 0.08932473510503769, "global_step": 190842, "epoch": 2144, "lr": 2.0293655401928947e-05}
{"train_loss": 0.09212499111890793, "global_step": 190843, "epoch": 2144, "lr": 2.0293189080423187e-05}
{"train_loss": 0.08200059831142426, "global_step": 190844, "epoch": 2144, "lr": 2.029272276291116e-05}
{"train_loss": 0.08832957595586777, "global_step": 190845, "epoch": 2144, "lr": 2.02922564493929e-05}
{"train_loss": 0.07143151015043259, "global_step": 190846, "epoch": 2144, "lr": 2.02917901398685e-05}
{"train_loss": 0.06073053553700447, "global_step": 190847, "epoch": 2144, "lr": 2.029132383433799e-05}
{"train_loss": 0.10414237529039383, "global_step": 190848, "epoch": 2144, "lr": 2.029085753280146e-05}
{"train_loss": 0.09025207161903381, "global_step": 190849, "epoch": 2144, "lr": 2.0290391235258953e-05}
{"train_loss": 0.054882556200027466, "global_step": 190850, "epoch": 2144, "lr": 2.028992494171056e-05}
{"train_loss": 0.04279961809515953, "global_step": 190851, "epoch": 2144, "lr": 2.0289458652156295e-05}
{"train_loss": 0.03809667006134987, "global_step": 190852, "epoch": 2144, "lr": 2.0288992366596276e-05}
{"train_loss": 0.07795094698667526, "global_step": 190853, "epoch": 2144, "lr": 2.0288526085030522e-05}
{"train_loss": 0.05756644904613495, "global_step": 190854, "epoch": 2144, "lr": 2.0288059807459132e-05}
{"train_loss": 0.05735769867897034, "global_step": 190855, "epoch": 2144, "lr": 2.028759353388215e-05}
{"train_loss": 0.05857662484049797, "global_step": 190856, "epoch": 2144, "lr": 2.028712726429962e-05}
{"train_loss": 0.08017370849847794, "global_step": 190857, "epoch": 2144, "lr": 2.0286660998711643e-05}
{"train_loss": 0.047071341425180435, "global_step": 190858, "epoch": 2144, "lr": 2.028619473711824e-05}
{"train_loss": 0.03375636041164398, "global_step": 190859, "epoch": 2144, "lr": 2.028572847951952e-05}
{"train_loss": 0.10669486969709396, "global_step": 190860, "epoch": 2144, "lr": 2.0285262225915498e-05}
{"train_loss": 0.06719089299440384, "global_step": 190861, "epoch": 2144, "lr": 2.028479597630628e-05}
{"train_loss": 0.07838308811187744, "global_step": 190862, "epoch": 2144, "lr": 2.0284329730691892e-05}
{"train_loss": 0.06874234974384308, "global_step": 190863, "epoch": 2144, "lr": 2.0283863489072436e-05}
{"train_loss": 0.04881945624947548, "global_step": 190864, "epoch": 2144, "lr": 2.028339725144793e-05}
{"train_loss": 0.04195563122630119, "global_step": 190865, "epoch": 2144, "lr": 2.0282931017818473e-05}
{"train_loss": 0.06964769959449768, "global_step": 190866, "epoch": 2144, "lr": 2.0282464788184103e-05}
{"train_loss": 0.07337360829114914, "global_step": 190867, "epoch": 2144, "lr": 2.028199856254491e-05}
{"train_loss": 0.1101190522313118, "global_step": 190868, "epoch": 2144, "lr": 2.0281532340900915e-05}
{"train_loss": 0.0562313087284565, "global_step": 190869, "epoch": 2144, "lr": 2.028106612325223e-05}
{"train_loss": 0.12961095571517944, "global_step": 190870, "epoch": 2144, "lr": 2.0280599909598874e-05}
{"train_loss": 0.05874432250857353, "global_step": 190871, "epoch": 2144, "lr": 2.0280133699940946e-05}
{"train_loss": 0.05298040062189102, "global_step": 190872, "epoch": 2144, "lr": 2.0279667494278475e-05}
{"train_loss": 0.1715335100889206, "global_step": 190873, "epoch": 2144, "lr": 2.027920129261156e-05}
{"train_loss": 0.10021282732486725, "global_step": 190874, "epoch": 2144, "lr": 2.0278735094940225e-05}
{"train_loss": 0.023420142009854317, "global_step": 190875, "epoch": 2144, "lr": 2.0278268901264568e-05}
{"train_loss": 0.08255379647016525, "global_step": 190876, "epoch": 2144, "lr": 2.027780271158462e-05}
{"train_loss": 0.1137193813920021, "global_step": 190877, "epoch": 2144, "lr": 2.0277336525900458e-05}
{"train_loss": 0.09213827550411224, "global_step": 190878, "epoch": 2144, "lr": 2.0276870344212167e-05}
{"train_loss": 0.03974679484963417, "global_step": 190879, "epoch": 2144, "lr": 2.0276404166519768e-05}
{"train_loss": 0.04081626981496811, "global_step": 190880, "epoch": 2144, "lr": 2.027593799282336e-05}
{"train_loss": 0.055948611348867416, "global_step": 190881, "epoch": 2144, "lr": 2.0275471823122976e-05}
{"train_loss": 0.11657692492008209, "global_step": 190882, "epoch": 2144, "lr": 2.0275005657418705e-05}
{"train_loss": 0.04889684170484543, "global_step": 190883, "epoch": 2144, "lr": 2.0274539495710583e-05}
{"train_loss": 0.07665272057056427, "global_step": 190884, "epoch": 2144, "lr": 2.0274073337998705e-05}
{"train_loss": 0.13046598434448242, "global_step": 190885, "epoch": 2144, "lr": 2.0273607184283094e-05}
{"train_loss": 0.08231592178344727, "global_step": 190886, "epoch": 2144, "lr": 2.027314103456386e-05}
{"train_loss": 0.10893545299768448, "global_step": 190887, "epoch": 2144, "lr": 2.0272674888841015e-05}
{"train_loss": 0.050142426043748856, "global_step": 190888, "epoch": 2144, "lr": 2.0272208747114668e-05}
{"train_loss": 0.07082973420619965, "global_step": 190889, "epoch": 2144, "lr": 2.0271742609384836e-05}
{"train_loss": 0.05140604451298714, "global_step": 190890, "epoch": 2144, "lr": 2.027127647565163e-05}
{"train_loss": 0.09379184246063232, "global_step": 190891, "epoch": 2144, "lr": 2.027081034591507e-05}
{"train_loss": 0.07779920101165771, "global_step": 190892, "epoch": 2144, "lr": 2.0270344220175252e-05}
{"train_loss": 0.049712054431438446, "global_step": 190893, "epoch": 2144, "lr": 2.0269878098432225e-05}
{"train_loss": 0.10043700784444809, "global_step": 190894, "epoch": 2144, "lr": 2.026941198068604e-05}
{"train_loss": 0.029507789760828018, "global_step": 190895, "epoch": 2144, "lr": 2.026894586693678e-05}
{"train_loss": 0.05853548273444176, "global_step": 190896, "epoch": 2144, "lr": 2.0268479757184483e-05}
{"train_loss": 0.09510482102632523, "global_step": 190897, "epoch": 2144, "lr": 2.026801365142924e-05}
{"train_loss": 0.0313628613948822, "global_step": 190898, "epoch": 2144, "lr": 2.0267547549671085e-05}
{"train_loss": 0.09016798436641693, "global_step": 190899, "epoch": 2144, "lr": 2.0267081451910115e-05}
{"train_loss": 0.08058757334947586, "global_step": 190900, "epoch": 2144, "lr": 2.0266615358146358e-05}
{"train_loss": 0.0511389896273613, "global_step": 190901, "epoch": 2144, "lr": 2.0266149268379902e-05}
{"train_loss": 0.06053658574819565, "global_step": 190902, "epoch": 2144, "lr": 2.026568318261079e-05}
{"train_loss": 0.0656789019703865, "global_step": 190903, "epoch": 2144, "lr": 2.0265217100839107e-05}
{"train_loss": 0.0723412103657977, "global_step": 190904, "epoch": 2144, "lr": 2.026475102306489e-05, "val_loss": 7.779003143310547}
{"train_loss": 0.05499935522675514, "global_step": 190905, "epoch": 2145, "lr": 2.0264284949288226e-05}
{"train_loss": 0.10975519567728043, "global_step": 190906, "epoch": 2145, "lr": 2.0263818879509156e-05}
{"train_loss": 0.03507540002465248, "global_step": 190907, "epoch": 2145, "lr": 2.0263352813727766e-05}
{"train_loss": 0.059339359402656555, "global_step": 190908, "epoch": 2145, "lr": 2.0262886751944087e-05}
{"train_loss": 0.03925585374236107, "global_step": 190909, "epoch": 2145, "lr": 2.0262420694158224e-05}
{"train_loss": 0.08245903998613358, "global_step": 190910, "epoch": 2145, "lr": 2.0261954640370194e-05}
{"train_loss": 0.045679397881031036, "global_step": 190911, "epoch": 2145, "lr": 2.0261488590580102e-05}
{"train_loss": 0.08992819488048553, "global_step": 190912, "epoch": 2145, "lr": 2.026102254478797e-05}
{"train_loss": 0.09921323508024216, "global_step": 190913, "epoch": 2145, "lr": 2.0260556502993883e-05}
{"train_loss": 0.06786949932575226, "global_step": 190914, "epoch": 2145, "lr": 2.0260090465197917e-05}
{"train_loss": 0.03155723959207535, "global_step": 190915, "epoch": 2145, "lr": 2.0259624431400105e-05}
{"train_loss": 0.08569350838661194, "global_step": 190916, "epoch": 2145, "lr": 2.0259158401600542e-05}
{"train_loss": 0.09085353463888168, "global_step": 190917, "epoch": 2145, "lr": 2.025869237579925e-05}
{"train_loss": 0.047850027680397034, "global_step": 190918, "epoch": 2145, "lr": 2.0258226353996336e-05}
{"train_loss": 0.09391468018293381, "global_step": 190919, "epoch": 2145, "lr": 2.0257760336191824e-05}
{"train_loss": 0.07777024805545807, "global_step": 190920, "epoch": 2145, "lr": 2.0257294322385812e-05}
{"train_loss": 0.05274166539311409, "global_step": 190921, "epoch": 2145, "lr": 2.0256828312578318e-05}
{"train_loss": 0.03175444155931473, "global_step": 190922, "epoch": 2145, "lr": 2.0256362306769456e-05}
{"train_loss": 0.04872498661279678, "global_step": 190923, "epoch": 2145, "lr": 2.025589630495925e-05}
{"train_loss": 0.07132065296173096, "global_step": 190924, "epoch": 2145, "lr": 2.0255430307147782e-05}
{"train_loss": 0.05041421949863434, "global_step": 190925, "epoch": 2145, "lr": 2.0254964313335102e-05}
{"train_loss": 0.0772971585392952, "global_step": 190926, "epoch": 2145, "lr": 2.0254498323521292e-05}
{"train_loss": 0.09805630892515182, "global_step": 190927, "epoch": 2145, "lr": 2.0254032337706385e-05}
{"train_loss": 0.09455841779708862, "global_step": 190928, "epoch": 2145, "lr": 2.025356635589048e-05}
{"train_loss": 0.06933276355266571, "global_step": 190929, "epoch": 2145, "lr": 2.0253100378073604e-05}
{"train_loss": 0.08035381883382797, "global_step": 190930, "epoch": 2145, "lr": 2.025263440425585e-05}
{"train_loss": 0.0379168763756752, "global_step": 190931, "epoch": 2145, "lr": 2.0252168434437247e-05}
{"train_loss": 0.041764914989471436, "global_step": 190932, "epoch": 2145, "lr": 2.0251702468617902e-05}
{"train_loss": 0.044030264019966125, "global_step": 190933, "epoch": 2145, "lr": 2.025123650679785e-05}
{"train_loss": 0.05315614864230156, "global_step": 190934, "epoch": 2145, "lr": 2.0250770548977138e-05}
{"train_loss": 0.06423261016607285, "global_step": 190935, "epoch": 2145, "lr": 2.0250304595155868e-05}
{"train_loss": 0.04876982048153877, "global_step": 190936, "epoch": 2145, "lr": 2.0249838645334058e-05}
{"train_loss": 0.09131421148777008, "global_step": 190937, "epoch": 2145, "lr": 2.0249372699511815e-05}
{"train_loss": 0.04414597526192665, "global_step": 190938, "epoch": 2145, "lr": 2.0248906757689167e-05}
{"train_loss": 0.1064591333270073, "global_step": 190939, "epoch": 2145, "lr": 2.0248440819866205e-05}
{"train_loss": 0.05200624093413353, "global_step": 190940, "epoch": 2145, "lr": 2.0247974886042964e-05}
{"train_loss": 0.12887416779994965, "global_step": 190941, "epoch": 2145, "lr": 2.024750895621953e-05}
{"train_loss": 0.05197139084339142, "global_step": 190942, "epoch": 2145, "lr": 2.0247043030395945e-05}
{"train_loss": 0.05487794429063797, "global_step": 190943, "epoch": 2145, "lr": 2.02465771085723e-05}
{"train_loss": 0.04900529608130455, "global_step": 190944, "epoch": 2145, "lr": 2.0246111190748617e-05}
{"train_loss": 0.05847733095288277, "global_step": 190945, "epoch": 2145, "lr": 2.0245645276925006e-05}
{"train_loss": 0.04939448833465576, "global_step": 190946, "epoch": 2145, "lr": 2.0245179367101487e-05}
{"train_loss": 0.06631754338741302, "global_step": 190947, "epoch": 2145, "lr": 2.0244713461278153e-05}
{"train_loss": 0.08067382872104645, "global_step": 190948, "epoch": 2145, "lr": 2.0244247559455037e-05}
{"train_loss": 0.0643502026796341, "global_step": 190949, "epoch": 2145, "lr": 2.0243781661632227e-05}
{"train_loss": 0.05895751714706421, "global_step": 190950, "epoch": 2145, "lr": 2.024331576780979e-05}
{"train_loss": 0.051034171134233475, "global_step": 190951, "epoch": 2145, "lr": 2.024284987798776e-05}
{"train_loss": 0.052574921399354935, "global_step": 190952, "epoch": 2145, "lr": 2.024238399216623e-05}
{"train_loss": 0.034396007657051086, "global_step": 190953, "epoch": 2145, "lr": 2.0241918110345238e-05}
{"train_loss": 0.135477215051651, "global_step": 190954, "epoch": 2145, "lr": 2.024145223252487e-05}
{"train_loss": 0.0961054265499115, "global_step": 190955, "epoch": 2145, "lr": 2.024098635870516e-05}
{"train_loss": 0.10427386313676834, "global_step": 190956, "epoch": 2145, "lr": 2.0240520488886206e-05}
{"train_loss": 0.05148463696241379, "global_step": 190957, "epoch": 2145, "lr": 2.0240054623068034e-05}
{"train_loss": 0.07463579624891281, "global_step": 190958, "epoch": 2145, "lr": 2.0239588761250738e-05}
{"train_loss": 0.06624536216259003, "global_step": 190959, "epoch": 2145, "lr": 2.0239122903434348e-05}
{"train_loss": 0.07873012125492096, "global_step": 190960, "epoch": 2145, "lr": 2.023865704961897e-05}
{"train_loss": 0.0533609502017498, "global_step": 190961, "epoch": 2145, "lr": 2.0238191199804617e-05}
{"train_loss": 0.09950082749128342, "global_step": 190962, "epoch": 2145, "lr": 2.02377253539914e-05}
{"train_loss": 0.0849839523434639, "global_step": 190963, "epoch": 2145, "lr": 2.0237259512179334e-05}
{"train_loss": 0.0570446252822876, "global_step": 190964, "epoch": 2145, "lr": 2.0236793674368527e-05}
{"train_loss": 0.0508401058614254, "global_step": 190965, "epoch": 2145, "lr": 2.0236327840559e-05}
{"train_loss": 0.033549632877111435, "global_step": 190966, "epoch": 2145, "lr": 2.0235862010750857e-05}
{"train_loss": 0.08955438435077667, "global_step": 190967, "epoch": 2145, "lr": 2.0235396184944123e-05}
{"train_loss": 0.03777756169438362, "global_step": 190968, "epoch": 2145, "lr": 2.0234930363138888e-05}
{"train_loss": 0.033086713403463364, "global_step": 190969, "epoch": 2145, "lr": 2.0234464545335193e-05}
{"train_loss": 0.04589364305138588, "global_step": 190970, "epoch": 2145, "lr": 2.0233998731533128e-05}
{"train_loss": 0.08001784235239029, "global_step": 190971, "epoch": 2145, "lr": 2.0233532921732735e-05}
{"train_loss": 0.08214876055717468, "global_step": 190972, "epoch": 2145, "lr": 2.0233067115934068e-05}
{"train_loss": 0.07745059579610825, "global_step": 190973, "epoch": 2145, "lr": 2.0232601314137218e-05}
{"train_loss": 0.08986782282590866, "global_step": 190974, "epoch": 2145, "lr": 2.0232135516342216e-05}
{"train_loss": 0.07426241785287857, "global_step": 190975, "epoch": 2145, "lr": 2.0231669722549157e-05}
{"train_loss": 0.054271891713142395, "global_step": 190976, "epoch": 2145, "lr": 2.0231203932758064e-05}
{"train_loss": 0.10017047822475433, "global_step": 190977, "epoch": 2145, "lr": 2.0230738146969046e-05}
{"train_loss": 0.07273083180189133, "global_step": 190978, "epoch": 2145, "lr": 2.0230272365182123e-05}
{"train_loss": 0.11434297263622284, "global_step": 190979, "epoch": 2145, "lr": 2.02298065873974e-05}
{"train_loss": 0.0920322984457016, "global_step": 190980, "epoch": 2145, "lr": 2.0229340813614894e-05}
{"train_loss": 0.050892867147922516, "global_step": 190981, "epoch": 2145, "lr": 2.022887504383471e-05}
{"train_loss": 0.06596118956804276, "global_step": 190982, "epoch": 2145, "lr": 2.022840927805687e-05}
{"train_loss": 0.10035281628370285, "global_step": 190983, "epoch": 2145, "lr": 2.0227943516281476e-05}
{"train_loss": 0.07878243178129196, "global_step": 190984, "epoch": 2145, "lr": 2.0227477758508558e-05}
{"train_loss": 0.11788284778594971, "global_step": 190985, "epoch": 2145, "lr": 2.0227012004738188e-05}
{"train_loss": 0.07568458467721939, "global_step": 190986, "epoch": 2145, "lr": 2.022654625497046e-05}
{"train_loss": 0.15258818864822388, "global_step": 190987, "epoch": 2145, "lr": 2.022608050920538e-05}
{"train_loss": 0.05134337767958641, "global_step": 190988, "epoch": 2145, "lr": 2.022561476744306e-05}
{"train_loss": 0.07291741669178009, "global_step": 190989, "epoch": 2145, "lr": 2.0225149029683532e-05}
{"train_loss": 0.06726161390542984, "global_step": 190990, "epoch": 2145, "lr": 2.0224683295926888e-05}
{"train_loss": 0.07363550364971161, "global_step": 190991, "epoch": 2145, "lr": 2.0224217566173148e-05}
{"train_loss": 0.037227142602205276, "global_step": 190992, "epoch": 2145, "lr": 2.022375184042242e-05}
{"train_loss": 0.06977770812390896, "global_step": 190993, "epoch": 2145, "lr": 2.0223286118674732e-05, "val_loss": 7.887473106384277, "train_action_mse_error": 7.580522060394287}
{"train_loss": 0.06473852694034576, "global_step": 190994, "epoch": 2146, "lr": 2.0222820400930174e-05}
{"train_loss": 0.09223940223455429, "global_step": 190995, "epoch": 2146, "lr": 2.0222354687188776e-05}
{"train_loss": 0.06323709338903427, "global_step": 190996, "epoch": 2146, "lr": 2.022188897745064e-05}
{"train_loss": 0.0971313938498497, "global_step": 190997, "epoch": 2146, "lr": 2.0221423271715788e-05}
{"train_loss": 0.14846812188625336, "global_step": 190998, "epoch": 2146, "lr": 2.0220957569984324e-05}
{"train_loss": 0.05540435016155243, "global_step": 190999, "epoch": 2146, "lr": 2.022049187225627e-05}
{"train_loss": 0.058791279792785645, "global_step": 191000, "epoch": 2146, "lr": 2.022002617853172e-05}
{"train_loss": 0.11699499189853668, "global_step": 191001, "epoch": 2146, "lr": 2.0219560488810712e-05}
{"train_loss": 0.09837900102138519, "global_step": 191002, "epoch": 2146, "lr": 2.021909480309334e-05}
{"train_loss": 0.06915025413036346, "global_step": 191003, "epoch": 2146, "lr": 2.0218629121379627e-05}
{"train_loss": 0.078806571662426, "global_step": 191004, "epoch": 2146, "lr": 2.0218163443669676e-05}
{"train_loss": 0.07548853009939194, "global_step": 191005, "epoch": 2146, "lr": 2.0217697769963513e-05}
{"train_loss": 0.039254385977983475, "global_step": 191006, "epoch": 2146, "lr": 2.0217232100261234e-05}
{"train_loss": 0.05373910069465637, "global_step": 191007, "epoch": 2146, "lr": 2.0216766434562867e-05}
{"train_loss": 0.057860687375068665, "global_step": 191008, "epoch": 2146, "lr": 2.0216300772868513e-05}
{"train_loss": 0.08203457295894623, "global_step": 191009, "epoch": 2146, "lr": 2.021583511517821e-05}
{"train_loss": 0.0823962539434433, "global_step": 191010, "epoch": 2146, "lr": 2.0215369461492005e-05}
{"train_loss": 0.06629392504692078, "global_step": 191011, "epoch": 2146, "lr": 2.0214903811810004e-05}
{"train_loss": 0.09841839969158173, "global_step": 191012, "epoch": 2146, "lr": 2.0214438166132226e-05}
{"train_loss": 0.04756520316004753, "global_step": 191013, "epoch": 2146, "lr": 2.0213972524458776e-05}
{"train_loss": 0.07466035336256027, "global_step": 191014, "epoch": 2146, "lr": 2.0213506886789674e-05}
{"train_loss": 0.06740473210811615, "global_step": 191015, "epoch": 2146, "lr": 2.021304125312502e-05}
{"train_loss": 0.11246687173843384, "global_step": 191016, "epoch": 2146, "lr": 2.021257562346484e-05}
{"train_loss": 0.07899889349937439, "global_step": 191017, "epoch": 2146, "lr": 2.0212109997809237e-05}
{"train_loss": 0.07471296936273575, "global_step": 191018, "epoch": 2146, "lr": 2.0211644376158233e-05}
{"train_loss": 0.07918606698513031, "global_step": 191019, "epoch": 2146, "lr": 2.021117875851193e-05}
{"train_loss": 0.07040172070264816, "global_step": 191020, "epoch": 2146, "lr": 2.0210713144870347e-05}
{"train_loss": 0.08843083679676056, "global_step": 191021, "epoch": 2146, "lr": 2.0210247535233577e-05}
{"train_loss": 0.07739631831645966, "global_step": 191022, "epoch": 2146, "lr": 2.0209781929601694e-05}
{"train_loss": 0.07511944323778152, "global_step": 191023, "epoch": 2146, "lr": 2.020931632797472e-05}
{"train_loss": 0.04021184518933296, "global_step": 191024, "epoch": 2146, "lr": 2.020885073035276e-05}
{"train_loss": 0.09287766367197037, "global_step": 191025, "epoch": 2146, "lr": 2.020838513673584e-05}
{"train_loss": 0.030936460942029953, "global_step": 191026, "epoch": 2146, "lr": 2.020791954712406e-05}
{"train_loss": 0.07149390131235123, "global_step": 191027, "epoch": 2146, "lr": 2.0207453961517443e-05}
{"train_loss": 0.053472381085157394, "global_step": 191028, "epoch": 2146, "lr": 2.0206988379916085e-05}
{"train_loss": 0.10071519017219543, "global_step": 191029, "epoch": 2146, "lr": 2.020652280232002e-05}
{"train_loss": 0.06536982953548431, "global_step": 191030, "epoch": 2146, "lr": 2.020605722872934e-05}
{"train_loss": 0.08316835761070251, "global_step": 191031, "epoch": 2146, "lr": 2.020559165914408e-05}
{"train_loss": 0.07377123087644577, "global_step": 191032, "epoch": 2146, "lr": 2.0205126093564332e-05}
{"train_loss": 0.047966886311769485, "global_step": 191033, "epoch": 2146, "lr": 2.0204660531990115e-05}
{"train_loss": 0.08773256093263626, "global_step": 191034, "epoch": 2146, "lr": 2.0204194974421548e-05}
{"train_loss": 0.10254014283418655, "global_step": 191035, "epoch": 2146, "lr": 2.0203729420858637e-05}
{"train_loss": 0.04417809844017029, "global_step": 191036, "epoch": 2146, "lr": 2.020326387130149e-05}
{"train_loss": 0.04732298105955124, "global_step": 191037, "epoch": 2146, "lr": 2.020279832575014e-05}
{"train_loss": 0.06971639394760132, "global_step": 191038, "epoch": 2146, "lr": 2.020233278420467e-05}
{"train_loss": 0.058751024305820465, "global_step": 191039, "epoch": 2146, "lr": 2.0201867246665122e-05}
{"train_loss": 0.09210082143545151, "global_step": 191040, "epoch": 2146, "lr": 2.020140171313157e-05}
{"train_loss": 0.11646468192338943, "global_step": 191041, "epoch": 2146, "lr": 2.0200936183604085e-05}
{"train_loss": 0.10649017244577408, "global_step": 191042, "epoch": 2146, "lr": 2.020047065808273e-05}
{"train_loss": 0.09673520177602768, "global_step": 191043, "epoch": 2146, "lr": 2.0200005136567536e-05}
{"train_loss": 0.044409047812223434, "global_step": 191044, "epoch": 2146, "lr": 2.019953961905861e-05}
{"train_loss": 0.06558512151241302, "global_step": 191045, "epoch": 2146, "lr": 2.0199074105555972e-05}
{"train_loss": 0.10782203823328018, "global_step": 191046, "epoch": 2146, "lr": 2.0198608596059726e-05}
{"train_loss": 0.1294744908809662, "global_step": 191047, "epoch": 2146, "lr": 2.0198143090569892e-05}
{"train_loss": 0.07224273681640625, "global_step": 191048, "epoch": 2146, "lr": 2.019767758908657e-05}
{"train_loss": 0.03927680850028992, "global_step": 191049, "epoch": 2146, "lr": 2.019721209160981e-05}
{"train_loss": 0.0257068183273077, "global_step": 191050, "epoch": 2146, "lr": 2.019674659813966e-05}
{"train_loss": 0.07928857952356339, "global_step": 191051, "epoch": 2146, "lr": 2.0196281108676207e-05}
{"train_loss": 0.0768677294254303, "global_step": 191052, "epoch": 2146, "lr": 2.0195815623219482e-05}
{"train_loss": 0.11119915544986725, "global_step": 191053, "epoch": 2146, "lr": 2.019535014176958e-05}
{"train_loss": 0.08405047655105591, "global_step": 191054, "epoch": 2146, "lr": 2.019488466432654e-05}
{"train_loss": 0.04395116865634918, "global_step": 191055, "epoch": 2146, "lr": 2.019441919089045e-05}
{"train_loss": 0.05769231542944908, "global_step": 191056, "epoch": 2146, "lr": 2.019395372146134e-05}
{"train_loss": 0.02668015845119953, "global_step": 191057, "epoch": 2146, "lr": 2.019348825603929e-05}
{"train_loss": 0.049234263598918915, "global_step": 191058, "epoch": 2146, "lr": 2.019302279462437e-05}
{"train_loss": 0.05071675032377243, "global_step": 191059, "epoch": 2146, "lr": 2.0192557337216626e-05}
{"train_loss": 0.08185527473688126, "global_step": 191060, "epoch": 2146, "lr": 2.019209188381615e-05}
{"train_loss": 0.07502393424510956, "global_step": 191061, "epoch": 2146, "lr": 2.019162643442296e-05}
{"train_loss": 0.08385638892650604, "global_step": 191062, "epoch": 2146, "lr": 2.0191160989037165e-05}
{"train_loss": 0.09602341800928116, "global_step": 191063, "epoch": 2146, "lr": 2.0190695547658784e-05}
{"train_loss": 0.04704359918832779, "global_step": 191064, "epoch": 2146, "lr": 2.0190230110287922e-05}
{"train_loss": 0.09596440196037292, "global_step": 191065, "epoch": 2146, "lr": 2.01897646769246e-05}
{"train_loss": 0.0989195927977562, "global_step": 191066, "epoch": 2146, "lr": 2.0189299247568917e-05}
{"train_loss": 0.07701987773180008, "global_step": 191067, "epoch": 2146, "lr": 2.0188833822220905e-05}
{"train_loss": 0.07448561489582062, "global_step": 191068, "epoch": 2146, "lr": 2.0188368400880654e-05}
{"train_loss": 0.06691537797451019, "global_step": 191069, "epoch": 2146, "lr": 2.0187902983548203e-05}
{"train_loss": 0.10855452716350555, "global_step": 191070, "epoch": 2146, "lr": 2.0187437570223634e-05}
{"train_loss": 0.06686798483133316, "global_step": 191071, "epoch": 2146, "lr": 2.0186972160906985e-05}
{"train_loss": 0.04004008695483208, "global_step": 191072, "epoch": 2146, "lr": 2.0186506755598357e-05}
{"train_loss": 0.07509353756904602, "global_step": 191073, "epoch": 2146, "lr": 2.018604135429777e-05}
{"train_loss": 0.06838568300008774, "global_step": 191074, "epoch": 2146, "lr": 2.018557595700532e-05}
{"train_loss": 0.10440319776535034, "global_step": 191075, "epoch": 2146, "lr": 2.0185110563721045e-05}
{"train_loss": 0.08650827407836914, "global_step": 191076, "epoch": 2146, "lr": 2.0184645174445015e-05}
{"train_loss": 0.061402615159749985, "global_step": 191077, "epoch": 2146, "lr": 2.0184179789177315e-05}
{"train_loss": 0.12120524048805237, "global_step": 191078, "epoch": 2146, "lr": 2.018371440791797e-05}
{"train_loss": 0.06012953817844391, "global_step": 191079, "epoch": 2146, "lr": 2.018324903066708e-05}
{"train_loss": 0.04633722081780434, "global_step": 191080, "epoch": 2146, "lr": 2.0182783657424665e-05}
{"train_loss": 0.06420602649450302, "global_step": 191081, "epoch": 2146, "lr": 2.018231828819083e-05}
{"train_loss": 0.07531478737344903, "global_step": 191082, "epoch": 2146, "lr": 2.018185292296563e-05, "val_loss": 7.651195526123047}
{"train_loss": 0.08315479010343552, "global_step": 191083, "epoch": 2147, "lr": 2.0181387561749087e-05}
{"train_loss": 0.05187111347913742, "global_step": 191084, "epoch": 2147, "lr": 2.0180922204541318e-05}
{"train_loss": 0.10120481997728348, "global_step": 191085, "epoch": 2147, "lr": 2.018045685134234e-05}
{"train_loss": 0.043520595878362656, "global_step": 191086, "epoch": 2147, "lr": 2.017999150215225e-05}
{"train_loss": 0.07168838381767273, "global_step": 191087, "epoch": 2147, "lr": 2.0179526156971102e-05}
{"train_loss": 0.08269937336444855, "global_step": 191088, "epoch": 2147, "lr": 2.0179060815798935e-05}
{"train_loss": 0.04034974053502083, "global_step": 191089, "epoch": 2147, "lr": 2.0178595478635848e-05}
{"train_loss": 0.024936547502875328, "global_step": 191090, "epoch": 2147, "lr": 2.0178130145481867e-05}
{"train_loss": 0.09649349004030228, "global_step": 191091, "epoch": 2147, "lr": 2.017766481633709e-05}
{"train_loss": 0.06061379611492157, "global_step": 191092, "epoch": 2147, "lr": 2.0177199491201543e-05}
{"train_loss": 0.0758160650730133, "global_step": 191093, "epoch": 2147, "lr": 2.0176734170075308e-05}
{"train_loss": 0.06821880489587784, "global_step": 191094, "epoch": 2147, "lr": 2.017626885295847e-05}
{"train_loss": 0.0994836613535881, "global_step": 191095, "epoch": 2147, "lr": 2.0175803539851047e-05}
{"train_loss": 0.0603630430996418, "global_step": 191096, "epoch": 2147, "lr": 2.0175338230753144e-05}
{"train_loss": 0.07881534844636917, "global_step": 191097, "epoch": 2147, "lr": 2.0174872925664783e-05}
{"train_loss": 0.032303307205438614, "global_step": 191098, "epoch": 2147, "lr": 2.0174407624586066e-05}
{"train_loss": 0.055494315922260284, "global_step": 191099, "epoch": 2147, "lr": 2.0173942327517014e-05}
{"train_loss": 0.0874188020825386, "global_step": 191100, "epoch": 2147, "lr": 2.0173477034457737e-05}
{"train_loss": 0.11772200465202332, "global_step": 191101, "epoch": 2147, "lr": 2.0173011745408253e-05}
{"train_loss": 0.115504190325737, "global_step": 191102, "epoch": 2147, "lr": 2.0172546460368657e-05}
{"train_loss": 0.12498656660318375, "global_step": 191103, "epoch": 2147, "lr": 2.0172081179338987e-05}
{"train_loss": 0.06713905930519104, "global_step": 191104, "epoch": 2147, "lr": 2.0171615902319336e-05}
{"train_loss": 0.05302426591515541, "global_step": 191105, "epoch": 2147, "lr": 2.0171150629309722e-05}
{"train_loss": 0.07251621782779694, "global_step": 191106, "epoch": 2147, "lr": 2.0170685360310254e-05}
{"train_loss": 0.11549407988786697, "global_step": 191107, "epoch": 2147, "lr": 2.0170220095320958e-05}
{"train_loss": 0.12479770928621292, "global_step": 191108, "epoch": 2147, "lr": 2.016975483434193e-05}
{"train_loss": 0.056318871676921844, "global_step": 191109, "epoch": 2147, "lr": 2.0169289577373195e-05}
{"train_loss": 0.07774192839860916, "global_step": 191110, "epoch": 2147, "lr": 2.016882432441486e-05}
{"train_loss": 0.013129069469869137, "global_step": 191111, "epoch": 2147, "lr": 2.0168359075466935e-05}
{"train_loss": 0.09855247288942337, "global_step": 191112, "epoch": 2147, "lr": 2.0167893830529515e-05}
{"train_loss": 0.07973030209541321, "global_step": 191113, "epoch": 2147, "lr": 2.0167428589602676e-05}
{"train_loss": 0.11227206140756607, "global_step": 191114, "epoch": 2147, "lr": 2.0166963352686448e-05}
{"train_loss": 0.10587441176176071, "global_step": 191115, "epoch": 2147, "lr": 2.0166498119780923e-05}
{"train_loss": 0.09895685315132141, "global_step": 191116, "epoch": 2147, "lr": 2.0166032890886134e-05}
{"train_loss": 0.04551350325345993, "global_step": 191117, "epoch": 2147, "lr": 2.016556766600217e-05}
{"train_loss": 0.09601125866174698, "global_step": 191118, "epoch": 2147, "lr": 2.0165102445129068e-05}
{"train_loss": 0.0997244343161583, "global_step": 191119, "epoch": 2147, "lr": 2.016463722826692e-05}
{"train_loss": 0.06823816150426865, "global_step": 191120, "epoch": 2147, "lr": 2.016417201541578e-05}
{"train_loss": 0.0709233507514, "global_step": 191121, "epoch": 2147, "lr": 2.0163706806575677e-05}
{"train_loss": 0.0662386491894722, "global_step": 191122, "epoch": 2147, "lr": 2.016324160174672e-05}
{"train_loss": 0.06707199662923813, "global_step": 191123, "epoch": 2147, "lr": 2.0162776400928935e-05}
{"train_loss": 0.05774011090397835, "global_step": 191124, "epoch": 2147, "lr": 2.0162311204122413e-05}
{"train_loss": 0.08270970731973648, "global_step": 191125, "epoch": 2147, "lr": 2.0161846011327212e-05}
{"train_loss": 0.0180441252887249, "global_step": 191126, "epoch": 2147, "lr": 2.016138082254337e-05}
{"train_loss": 0.05056328698992729, "global_step": 191127, "epoch": 2147, "lr": 2.0160915637770977e-05}
{"train_loss": 0.14936354756355286, "global_step": 191128, "epoch": 2147, "lr": 2.0160450457010072e-05}
{"train_loss": 0.0776500329375267, "global_step": 191129, "epoch": 2147, "lr": 2.015998528026075e-05}
{"train_loss": 0.056956108659505844, "global_step": 191130, "epoch": 2147, "lr": 2.0159520107523034e-05}
{"train_loss": 0.07893280684947968, "global_step": 191131, "epoch": 2147, "lr": 2.0159054938797006e-05}
{"train_loss": 0.04835061356425285, "global_step": 191132, "epoch": 2147, "lr": 2.0158589774082753e-05}
{"train_loss": 0.06488966196775436, "global_step": 191133, "epoch": 2147, "lr": 2.0158124613380293e-05}
{"train_loss": 0.09050878882408142, "global_step": 191134, "epoch": 2147, "lr": 2.015765945668972e-05}
{"train_loss": 0.08868569135665894, "global_step": 191135, "epoch": 2147, "lr": 2.015719430401108e-05}
{"train_loss": 0.01229744590818882, "global_step": 191136, "epoch": 2147, "lr": 2.015672915534445e-05}
{"train_loss": 0.06944747269153595, "global_step": 191137, "epoch": 2147, "lr": 2.0156264010689868e-05}
{"train_loss": 0.10043350607156754, "global_step": 191138, "epoch": 2147, "lr": 2.0155798870047438e-05}
{"train_loss": 0.059278298169374466, "global_step": 191139, "epoch": 2147, "lr": 2.015533373341717e-05}
{"train_loss": 0.08600613474845886, "global_step": 191140, "epoch": 2147, "lr": 2.0154868600799177e-05}
{"train_loss": 0.07389391958713531, "global_step": 191141, "epoch": 2147, "lr": 2.0154403472193478e-05}
{"train_loss": 0.06058865040540695, "global_step": 191142, "epoch": 2147, "lr": 2.015393834760017e-05}
{"train_loss": 0.062053270637989044, "global_step": 191143, "epoch": 2147, "lr": 2.0153473227019286e-05}
{"train_loss": 0.07406151294708252, "global_step": 191144, "epoch": 2147, "lr": 2.0153008110450928e-05}
{"train_loss": 0.02822374366223812, "global_step": 191145, "epoch": 2147, "lr": 2.0152542997895108e-05}
{"train_loss": 0.05535252019762993, "global_step": 191146, "epoch": 2147, "lr": 2.015207788935194e-05}
{"train_loss": 0.06010913476347923, "global_step": 191147, "epoch": 2147, "lr": 2.0151612784821438e-05}
{"train_loss": 0.04713961482048035, "global_step": 191148, "epoch": 2147, "lr": 2.015114768430369e-05}
{"train_loss": 0.11213575303554535, "global_step": 191149, "epoch": 2147, "lr": 2.015068258779877e-05}
{"train_loss": 0.07849125564098358, "global_step": 191150, "epoch": 2147, "lr": 2.0150217495306713e-05}
{"train_loss": 0.06255973130464554, "global_step": 191151, "epoch": 2147, "lr": 2.0149752406827614e-05}
{"train_loss": 0.058674633502960205, "global_step": 191152, "epoch": 2147, "lr": 2.01492873223615e-05}
{"train_loss": 0.07493390887975693, "global_step": 191153, "epoch": 2147, "lr": 2.0148822241908462e-05}
{"train_loss": 0.15825772285461426, "global_step": 191154, "epoch": 2147, "lr": 2.014835716546854e-05}
{"train_loss": 0.0718260407447815, "global_step": 191155, "epoch": 2147, "lr": 2.014789209304182e-05}
{"train_loss": 0.0478590726852417, "global_step": 191156, "epoch": 2147, "lr": 2.014742702462834e-05}
{"train_loss": 0.07745946198701859, "global_step": 191157, "epoch": 2147, "lr": 2.014696196022819e-05}
{"train_loss": 0.16718140244483948, "global_step": 191158, "epoch": 2147, "lr": 2.0146496899841393e-05}
{"train_loss": 0.09217850863933563, "global_step": 191159, "epoch": 2147, "lr": 2.014603184346806e-05}
{"train_loss": 0.07081890851259232, "global_step": 191160, "epoch": 2147, "lr": 2.0145566791108226e-05}
{"train_loss": 0.08510783314704895, "global_step": 191161, "epoch": 2147, "lr": 2.0145101742761945e-05}
{"train_loss": 0.06271833181381226, "global_step": 191162, "epoch": 2147, "lr": 2.0144636698429303e-05}
{"train_loss": 0.036110833287239075, "global_step": 191163, "epoch": 2147, "lr": 2.014417165811034e-05}
{"train_loss": 0.049274373799562454, "global_step": 191164, "epoch": 2147, "lr": 2.0143706621805143e-05}
{"train_loss": 0.08089220523834229, "global_step": 191165, "epoch": 2147, "lr": 2.0143241589513755e-05}
{"train_loss": 0.05420105904340744, "global_step": 191166, "epoch": 2147, "lr": 2.0142776561236232e-05}
{"train_loss": 0.06984511017799377, "global_step": 191167, "epoch": 2147, "lr": 2.0142311536972646e-05}
{"train_loss": 0.06808998435735703, "global_step": 191168, "epoch": 2147, "lr": 2.0141846516723085e-05}
{"train_loss": 0.03244568035006523, "global_step": 191169, "epoch": 2147, "lr": 2.0141381500487565e-05}
{"train_loss": 0.06646370142698288, "global_step": 191170, "epoch": 2147, "lr": 2.01409164882662e-05}
{"train_loss": 0.07357755053286137, "global_step": 191171, "epoch": 2147, "lr": 2.0140451480058997e-05, "val_loss": 7.663883686065674}
{"train_loss": 0.1130988821387291, "global_step": 191172, "epoch": 2148, "lr": 2.013998647586607e-05}
{"train_loss": 0.10273297131061554, "global_step": 191173, "epoch": 2148, "lr": 2.0139521475687435e-05}
{"train_loss": 0.052699219435453415, "global_step": 191174, "epoch": 2148, "lr": 2.0139056479523198e-05}
{"train_loss": 0.07831675559282303, "global_step": 191175, "epoch": 2148, "lr": 2.0138591487373382e-05}
{"train_loss": 0.11177212744951248, "global_step": 191176, "epoch": 2148, "lr": 2.0138126499238086e-05}
{"train_loss": 0.05198284983634949, "global_step": 191177, "epoch": 2148, "lr": 2.013766151511734e-05}
{"train_loss": 0.06270933151245117, "global_step": 191178, "epoch": 2148, "lr": 2.0137196535011237e-05}
{"train_loss": 0.08077472448348999, "global_step": 191179, "epoch": 2148, "lr": 2.0136731558919808e-05}
{"train_loss": 0.09371501207351685, "global_step": 191180, "epoch": 2148, "lr": 2.0136266586843145e-05}
{"train_loss": 0.07384076714515686, "global_step": 191181, "epoch": 2148, "lr": 2.0135801618781286e-05}
{"train_loss": 0.028050802648067474, "global_step": 191182, "epoch": 2148, "lr": 2.0135336654734315e-05}
{"train_loss": 0.0402156338095665, "global_step": 191183, "epoch": 2148, "lr": 2.0134871694702274e-05}
{"train_loss": 0.08200670033693314, "global_step": 191184, "epoch": 2148, "lr": 2.013440673868523e-05}
{"train_loss": 0.06852424889802933, "global_step": 191185, "epoch": 2148, "lr": 2.013394178668327e-05}
{"train_loss": 0.06137852743268013, "global_step": 191186, "epoch": 2148, "lr": 2.013347683869642e-05}
{"train_loss": 0.02928958274424076, "global_step": 191187, "epoch": 2148, "lr": 2.0133011894724773e-05}
{"train_loss": 0.05386844277381897, "global_step": 191188, "epoch": 2148, "lr": 2.0132546954768365e-05}
{"train_loss": 0.056619592010974884, "global_step": 191189, "epoch": 2148, "lr": 2.0132082018827293e-05}
{"train_loss": 0.14980167150497437, "global_step": 191190, "epoch": 2148, "lr": 2.0131617086901578e-05}
{"train_loss": 0.0796843022108078, "global_step": 191191, "epoch": 2148, "lr": 2.0131152158991316e-05}
{"train_loss": 0.06918366253376007, "global_step": 191192, "epoch": 2148, "lr": 2.0130687235096547e-05}
{"train_loss": 0.052316658198833466, "global_step": 191193, "epoch": 2148, "lr": 2.0130222315217358e-05}
{"train_loss": 0.0436706505715847, "global_step": 191194, "epoch": 2148, "lr": 2.0129757399353775e-05}
{"train_loss": 0.06677836179733276, "global_step": 191195, "epoch": 2148, "lr": 2.01292924875059e-05}
{"train_loss": 0.06644323468208313, "global_step": 191196, "epoch": 2148, "lr": 2.0128827579673764e-05}
{"train_loss": 0.03889765962958336, "global_step": 191197, "epoch": 2148, "lr": 2.012836267585746e-05}
{"train_loss": 0.09255976229906082, "global_step": 191198, "epoch": 2148, "lr": 2.012789777605703e-05}
{"train_loss": 0.06829749047756195, "global_step": 191199, "epoch": 2148, "lr": 2.012743288027253e-05}
{"train_loss": 0.0702536553144455, "global_step": 191200, "epoch": 2148, "lr": 2.0126967988504043e-05}
{"train_loss": 0.09797438234090805, "global_step": 191201, "epoch": 2148, "lr": 2.012650310075161e-05}
{"train_loss": 0.03353632986545563, "global_step": 191202, "epoch": 2148, "lr": 2.0126038217015314e-05}
{"train_loss": 0.054860059171915054, "global_step": 191203, "epoch": 2148, "lr": 2.0125573337295196e-05}
{"train_loss": 0.07447115331888199, "global_step": 191204, "epoch": 2148, "lr": 2.0125108461591347e-05}
{"train_loss": 0.11066686362028122, "global_step": 191205, "epoch": 2148, "lr": 2.0124643589903792e-05}
{"train_loss": 0.028506644070148468, "global_step": 191206, "epoch": 2148, "lr": 2.0124178722232633e-05}
{"train_loss": 0.05700192227959633, "global_step": 191207, "epoch": 2148, "lr": 2.0123713858577902e-05}
{"train_loss": 0.04634740203619003, "global_step": 191208, "epoch": 2148, "lr": 2.0123248998939686e-05}
{"train_loss": 0.04414210468530655, "global_step": 191209, "epoch": 2148, "lr": 2.0122784143318018e-05}
{"train_loss": 0.03297564387321472, "global_step": 191210, "epoch": 2148, "lr": 2.0122319291712998e-05}
{"train_loss": 0.02023777924478054, "global_step": 191211, "epoch": 2148, "lr": 2.012185444412465e-05}
{"train_loss": 0.08989419043064117, "global_step": 191212, "epoch": 2148, "lr": 2.012138960055307e-05}
{"train_loss": 0.06719601154327393, "global_step": 191213, "epoch": 2148, "lr": 2.0120924760998288e-05}
{"train_loss": 0.04463374987244606, "global_step": 191214, "epoch": 2148, "lr": 2.01204599254604e-05}
{"train_loss": 0.10767398774623871, "global_step": 191215, "epoch": 2148, "lr": 2.0119995093939437e-05}
{"train_loss": 0.09187464416027069, "global_step": 191216, "epoch": 2148, "lr": 2.0119530266435492e-05}
{"train_loss": 0.04295116662979126, "global_step": 191217, "epoch": 2148, "lr": 2.0119065442948598e-05}
{"train_loss": 0.06440968066453934, "global_step": 191218, "epoch": 2148, "lr": 2.0118600623478846e-05}
{"train_loss": 0.0556054450571537, "global_step": 191219, "epoch": 2148, "lr": 2.0118135808026262e-05}
{"train_loss": 0.061002083122730255, "global_step": 191220, "epoch": 2148, "lr": 2.0117670996590936e-05}
{"train_loss": 0.0656522735953331, "global_step": 191221, "epoch": 2148, "lr": 2.0117206189172944e-05}
{"train_loss": 0.05921844765543938, "global_step": 191222, "epoch": 2148, "lr": 2.0116741385772308e-05}
{"train_loss": 0.05475898087024689, "global_step": 191223, "epoch": 2148, "lr": 2.0116276586389132e-05}
{"train_loss": 0.08599454164505005, "global_step": 191224, "epoch": 2148, "lr": 2.0115811791023437e-05}
{"train_loss": 0.04217485710978508, "global_step": 191225, "epoch": 2148, "lr": 2.0115346999675328e-05}
{"train_loss": 0.08421234786510468, "global_step": 191226, "epoch": 2148, "lr": 2.0114882212344828e-05}
{"train_loss": 0.05485378950834274, "global_step": 191227, "epoch": 2148, "lr": 2.0114417429032035e-05}
{"train_loss": 0.07335648685693741, "global_step": 191228, "epoch": 2148, "lr": 2.0113952649736978e-05}
{"train_loss": 0.1294453740119934, "global_step": 191229, "epoch": 2148, "lr": 2.0113487874459753e-05}
{"train_loss": 0.03769839555025101, "global_step": 191230, "epoch": 2148, "lr": 2.0113023103200385e-05}
{"train_loss": 0.04388977959752083, "global_step": 191231, "epoch": 2148, "lr": 2.0112558335958976e-05}
{"train_loss": 0.12437689304351807, "global_step": 191232, "epoch": 2148, "lr": 2.011209357273555e-05}
{"train_loss": 0.07839921116828918, "global_step": 191233, "epoch": 2148, "lr": 2.0111628813530208e-05}
{"train_loss": 0.0705617293715477, "global_step": 191234, "epoch": 2148, "lr": 2.011116405834298e-05}
{"train_loss": 0.027547411620616913, "global_step": 191235, "epoch": 2148, "lr": 2.011069930717395e-05}
{"train_loss": 0.048317696899175644, "global_step": 191236, "epoch": 2148, "lr": 2.0110234560023178e-05}
{"train_loss": 0.11211948841810226, "global_step": 191237, "epoch": 2148, "lr": 2.0109769816890704e-05}
{"train_loss": 0.062094900757074356, "global_step": 191238, "epoch": 2148, "lr": 2.0109305077776623e-05}
{"train_loss": 0.04297412186861038, "global_step": 191239, "epoch": 2148, "lr": 2.0108840342680968e-05}
{"train_loss": 0.05266167223453522, "global_step": 191240, "epoch": 2148, "lr": 2.0108375611603826e-05}
{"train_loss": 0.048155538737773895, "global_step": 191241, "epoch": 2148, "lr": 2.0107910884545233e-05}
{"train_loss": 0.0653705969452858, "global_step": 191242, "epoch": 2148, "lr": 2.010744616150529e-05}
{"train_loss": 0.09814615547657013, "global_step": 191243, "epoch": 2148, "lr": 2.0106981442484012e-05}
{"train_loss": 0.05221322178840637, "global_step": 191244, "epoch": 2148, "lr": 2.0106516727481506e-05}
{"train_loss": 0.097569540143013, "global_step": 191245, "epoch": 2148, "lr": 2.0106052016497795e-05}
{"train_loss": 0.052739329636096954, "global_step": 191246, "epoch": 2148, "lr": 2.0105587309532982e-05}
{"train_loss": 0.07536925375461578, "global_step": 191247, "epoch": 2148, "lr": 2.010512260658709e-05}
{"train_loss": 0.056774672120809555, "global_step": 191248, "epoch": 2148, "lr": 2.0104657907660218e-05}
{"train_loss": 0.06907802075147629, "global_step": 191249, "epoch": 2148, "lr": 2.0104193212752386e-05}
{"train_loss": 0.0688440129160881, "global_step": 191250, "epoch": 2148, "lr": 2.0103728521863706e-05}
{"train_loss": 0.07837484031915665, "global_step": 191251, "epoch": 2148, "lr": 2.0103263834994197e-05}
{"train_loss": 0.04711589589715004, "global_step": 191252, "epoch": 2148, "lr": 2.0102799152143957e-05}
{"train_loss": 0.05135494843125343, "global_step": 191253, "epoch": 2148, "lr": 2.010233447331301e-05}
{"train_loss": 0.05049651116132736, "global_step": 191254, "epoch": 2148, "lr": 2.010186979850146e-05}
{"train_loss": 0.0731356292963028, "global_step": 191255, "epoch": 2148, "lr": 2.0101405127709333e-05}
{"train_loss": 0.10420643538236618, "global_step": 191256, "epoch": 2148, "lr": 2.0100940460936724e-05}
{"train_loss": 0.048827461898326874, "global_step": 191257, "epoch": 2148, "lr": 2.010047579818366e-05}
{"train_loss": 0.07804065942764282, "global_step": 191258, "epoch": 2148, "lr": 2.0100011139450226e-05}
{"train_loss": 0.11944011598825455, "global_step": 191259, "epoch": 2148, "lr": 2.0099546484736497e-05}
{"train_loss": 0.06702301219063864, "global_step": 191260, "epoch": 2148, "lr": 2.0099081834042498e-05, "val_loss": 7.914224147796631}
{"train_loss": 0.08629579097032547, "global_step": 191261, "epoch": 2149, "lr": 2.009861718736834e-05}
{"train_loss": 0.05224566534161568, "global_step": 191262, "epoch": 2149, "lr": 2.0098152544714033e-05}
{"train_loss": 0.13632138073444366, "global_step": 191263, "epoch": 2149, "lr": 2.0097687906079682e-05}
{"train_loss": 0.06753227114677429, "global_step": 191264, "epoch": 2149, "lr": 2.0097223271465322e-05}
{"train_loss": 0.09766016900539398, "global_step": 191265, "epoch": 2149, "lr": 2.009675864087104e-05}
{"train_loss": 0.07532460987567902, "global_step": 191266, "epoch": 2149, "lr": 2.0096294014296862e-05}
{"train_loss": 0.07637728750705719, "global_step": 191267, "epoch": 2149, "lr": 2.0095829391742893e-05}
{"train_loss": 0.06680388003587723, "global_step": 191268, "epoch": 2149, "lr": 2.0095364773209163e-05}
{"train_loss": 0.05085419863462448, "global_step": 191269, "epoch": 2149, "lr": 2.0094900158695756e-05}
{"train_loss": 0.08393196761608124, "global_step": 191270, "epoch": 2149, "lr": 2.0094435548202713e-05}
{"train_loss": 0.02083062008023262, "global_step": 191271, "epoch": 2149, "lr": 2.009397094173013e-05}
{"train_loss": 0.0930001363158226, "global_step": 191272, "epoch": 2149, "lr": 2.0093506339278024e-05}
{"train_loss": 0.038685835897922516, "global_step": 191273, "epoch": 2149, "lr": 2.0093041740846496e-05}
{"train_loss": 0.05265101045370102, "global_step": 191274, "epoch": 2149, "lr": 2.0092577146435583e-05}
{"train_loss": 0.04126877337694168, "global_step": 191275, "epoch": 2149, "lr": 2.0092112556045373e-05}
{"train_loss": 0.053461313247680664, "global_step": 191276, "epoch": 2149, "lr": 2.0091647969675913e-05}
{"train_loss": 0.06802424788475037, "global_step": 191277, "epoch": 2149, "lr": 2.0091183387327257e-05}
{"train_loss": 0.10346025228500366, "global_step": 191278, "epoch": 2149, "lr": 2.0090718808999487e-05}
{"train_loss": 0.04275713488459587, "global_step": 191279, "epoch": 2149, "lr": 2.0090254234692634e-05}
{"train_loss": 0.07411324232816696, "global_step": 191280, "epoch": 2149, "lr": 2.0089789664406805e-05}
{"train_loss": 0.04240654036402702, "global_step": 191281, "epoch": 2149, "lr": 2.0089325098142024e-05}
{"train_loss": 0.05918550491333008, "global_step": 191282, "epoch": 2149, "lr": 2.0088860535898387e-05}
{"train_loss": 0.11470010876655579, "global_step": 191283, "epoch": 2149, "lr": 2.0088395977675913e-05}
{"train_loss": 0.11415968090295792, "global_step": 191284, "epoch": 2149, "lr": 2.008793142347471e-05}
{"train_loss": 0.05434230715036392, "global_step": 191285, "epoch": 2149, "lr": 2.00874668732948e-05}
{"train_loss": 0.03361029922962189, "global_step": 191286, "epoch": 2149, "lr": 2.008700232713629e-05}
{"train_loss": 0.11654475331306458, "global_step": 191287, "epoch": 2149, "lr": 2.0086537784999198e-05}
{"train_loss": 0.09489376842975616, "global_step": 191288, "epoch": 2149, "lr": 2.0086073246883626e-05}
{"train_loss": 0.08026100695133209, "global_step": 191289, "epoch": 2149, "lr": 2.0085608712789594e-05}
{"train_loss": 0.03997804597020149, "global_step": 191290, "epoch": 2149, "lr": 2.0085144182717203e-05}
{"train_loss": 0.06803571432828903, "global_step": 191291, "epoch": 2149, "lr": 2.0084679656666487e-05}
{"train_loss": 0.0645560696721077, "global_step": 191292, "epoch": 2149, "lr": 2.008421513463754e-05}
{"train_loss": 0.066135935485363, "global_step": 191293, "epoch": 2149, "lr": 2.0083750616630386e-05}
{"train_loss": 0.03571514040231705, "global_step": 191294, "epoch": 2149, "lr": 2.0083286102645106e-05}
{"train_loss": 0.11144012957811356, "global_step": 191295, "epoch": 2149, "lr": 2.008282159268178e-05}
{"train_loss": 0.056764617562294006, "global_step": 191296, "epoch": 2149, "lr": 2.0082357086740445e-05}
{"train_loss": 0.044437430799007416, "global_step": 191297, "epoch": 2149, "lr": 2.0081892584821183e-05}
{"train_loss": 0.08567176014184952, "global_step": 191298, "epoch": 2149, "lr": 2.0081428086924026e-05}
{"train_loss": 0.060510024428367615, "global_step": 191299, "epoch": 2149, "lr": 2.008096359304908e-05}
{"train_loss": 0.068036288022995, "global_step": 191300, "epoch": 2149, "lr": 2.008049910319636e-05}
{"train_loss": 0.06473278254270554, "global_step": 191301, "epoch": 2149, "lr": 2.008003461736597e-05}
{"train_loss": 0.10345516353845596, "global_step": 191302, "epoch": 2149, "lr": 2.0079570135557945e-05}
{"train_loss": 0.036356016993522644, "global_step": 191303, "epoch": 2149, "lr": 2.007910565777237e-05}
{"train_loss": 0.08937466144561768, "global_step": 191304, "epoch": 2149, "lr": 2.007864118400927e-05}
{"train_loss": 0.03563353419303894, "global_step": 191305, "epoch": 2149, "lr": 2.007817671426876e-05}
{"train_loss": 0.06840101629495621, "global_step": 191306, "epoch": 2149, "lr": 2.0077712248550856e-05}
{"train_loss": 0.052389685064554214, "global_step": 191307, "epoch": 2149, "lr": 2.007724778685565e-05}
{"train_loss": 0.04477834329009056, "global_step": 191308, "epoch": 2149, "lr": 2.007678332918318e-05}
{"train_loss": 0.08067233115434647, "global_step": 191309, "epoch": 2149, "lr": 2.007631887553354e-05}
{"train_loss": 0.1488046646118164, "global_step": 191310, "epoch": 2149, "lr": 2.0075854425906753e-05}
{"train_loss": 0.04901493340730667, "global_step": 191311, "epoch": 2149, "lr": 2.0075389980302927e-05}
{"train_loss": 0.053314268589019775, "global_step": 191312, "epoch": 2149, "lr": 2.0074925538722077e-05}
{"train_loss": 0.04104447364807129, "global_step": 191313, "epoch": 2149, "lr": 2.0074461101164305e-05}
{"train_loss": 0.06984003633260727, "global_step": 191314, "epoch": 2149, "lr": 2.0073996667629662e-05}
{"train_loss": 0.061848193407058716, "global_step": 191315, "epoch": 2149, "lr": 2.0073532238118188e-05}
{"train_loss": 0.06085115671157837, "global_step": 191316, "epoch": 2149, "lr": 2.0073067812629975e-05}
{"train_loss": 0.05810480937361717, "global_step": 191317, "epoch": 2149, "lr": 2.0072603391165056e-05}
{"train_loss": 0.11763370782136917, "global_step": 191318, "epoch": 2149, "lr": 2.007213897372353e-05}
{"train_loss": 0.06375382095575333, "global_step": 191319, "epoch": 2149, "lr": 2.0071674560305425e-05}
{"train_loss": 0.030953308567404747, "global_step": 191320, "epoch": 2149, "lr": 2.0071210150910835e-05}
{"train_loss": 0.047808799892663956, "global_step": 191321, "epoch": 2149, "lr": 2.0070745745539786e-05}
{"train_loss": 0.07761768996715546, "global_step": 191322, "epoch": 2149, "lr": 2.007028134419238e-05}
{"train_loss": 0.04067714884877205, "global_step": 191323, "epoch": 2149, "lr": 2.006981694686864e-05}
{"train_loss": 0.05966636538505554, "global_step": 191324, "epoch": 2149, "lr": 2.0069352553568672e-05}
{"train_loss": 0.10450860857963562, "global_step": 191325, "epoch": 2149, "lr": 2.0068888164292486e-05}
{"train_loss": 0.06892713904380798, "global_step": 191326, "epoch": 2149, "lr": 2.0068423779040202e-05}
{"train_loss": 0.06467913836240768, "global_step": 191327, "epoch": 2149, "lr": 2.0067959397811832e-05}
{"train_loss": 0.04687418416142464, "global_step": 191328, "epoch": 2149, "lr": 2.006749502060748e-05}
{"train_loss": 0.1433062106370926, "global_step": 191329, "epoch": 2149, "lr": 2.0067030647427165e-05}
{"train_loss": 0.044068798422813416, "global_step": 191330, "epoch": 2149, "lr": 2.0066566278270976e-05}
{"train_loss": 0.06857863813638687, "global_step": 191331, "epoch": 2149, "lr": 2.006610191313899e-05}
{"train_loss": 0.06982898712158203, "global_step": 191332, "epoch": 2149, "lr": 2.0065637552031234e-05}
{"train_loss": 0.07582210749387741, "global_step": 191333, "epoch": 2149, "lr": 2.0065173194947806e-05}
{"train_loss": 0.06881450116634369, "global_step": 191334, "epoch": 2149, "lr": 2.0064708841888736e-05}
{"train_loss": 0.04972248524427414, "global_step": 191335, "epoch": 2149, "lr": 2.0064244492854113e-05}
{"train_loss": 0.06420227885246277, "global_step": 191336, "epoch": 2149, "lr": 2.0063780147843974e-05}
{"train_loss": 0.0401938371360302, "global_step": 191337, "epoch": 2149, "lr": 2.006331580685841e-05}
{"train_loss": 0.09044227749109268, "global_step": 191338, "epoch": 2149, "lr": 2.0062851469897453e-05}
{"train_loss": 0.042260438203811646, "global_step": 191339, "epoch": 2149, "lr": 2.0062387136961196e-05}
{"train_loss": 0.06047791242599487, "global_step": 191340, "epoch": 2149, "lr": 2.006192280804967e-05}
{"train_loss": 0.09773905575275421, "global_step": 191341, "epoch": 2149, "lr": 2.006145848316297e-05}
{"train_loss": 0.033998072147369385, "global_step": 191342, "epoch": 2149, "lr": 2.006099416230113e-05}
{"train_loss": 0.05610832944512367, "global_step": 191343, "epoch": 2149, "lr": 2.0060529845464234e-05}
{"train_loss": 0.06089283153414726, "global_step": 191344, "epoch": 2149, "lr": 2.0060065532652317e-05}
{"train_loss": 0.06893113255500793, "global_step": 191345, "epoch": 2149, "lr": 2.005960122386548e-05}
{"train_loss": 0.04098169505596161, "global_step": 191346, "epoch": 2149, "lr": 2.005913691910375e-05}
{"train_loss": 0.034366633743047714, "global_step": 191347, "epoch": 2149, "lr": 2.005867261836722e-05}
{"train_loss": 0.06796504557132721, "global_step": 191348, "epoch": 2149, "lr": 2.005820832165592e-05}
{"train_loss": 0.06713362617857671, "global_step": 191349, "epoch": 2149, "lr": 2.0057744028969943e-05, "val_loss": 7.827841758728027}
{"train_loss": 0.10391096770763397, "global_step": 191350, "epoch": 2150, "lr": 2.005727974030932e-05}
{"train_loss": 0.1317715048789978, "global_step": 191351, "epoch": 2150, "lr": 2.005681545567415e-05}
{"train_loss": 0.07920468598604202, "global_step": 191352, "epoch": 2150, "lr": 2.0056351175064474e-05}
{"train_loss": 0.048067912459373474, "global_step": 191353, "epoch": 2150, "lr": 2.0055886898480343e-05}
{"train_loss": 0.06757118552923203, "global_step": 191354, "epoch": 2150, "lr": 2.005542262592185e-05}
{"train_loss": 0.05333830788731575, "global_step": 191355, "epoch": 2150, "lr": 2.005495835738902e-05}
{"train_loss": 0.13626278936862946, "global_step": 191356, "epoch": 2150, "lr": 2.0054494092881953e-05}
{"train_loss": 0.09646137058734894, "global_step": 191357, "epoch": 2150, "lr": 2.0054029832400684e-05}
{"train_loss": 0.06032959371805191, "global_step": 191358, "epoch": 2150, "lr": 2.0053565575945292e-05}
{"train_loss": 0.05827408656477928, "global_step": 191359, "epoch": 2150, "lr": 2.0053101323515827e-05}
{"train_loss": 0.06431438028812408, "global_step": 191360, "epoch": 2150, "lr": 2.0052637075112367e-05}
{"train_loss": 0.05724882334470749, "global_step": 191361, "epoch": 2150, "lr": 2.005217283073495e-05}
{"train_loss": 0.02351820468902588, "global_step": 191362, "epoch": 2150, "lr": 2.0051708590383674e-05}
{"train_loss": 0.115885891020298, "global_step": 191363, "epoch": 2150, "lr": 2.0051244354058558e-05}
{"train_loss": 0.08196398615837097, "global_step": 191364, "epoch": 2150, "lr": 2.0050780121759705e-05}
{"train_loss": 0.065728560090065, "global_step": 191365, "epoch": 2150, "lr": 2.0050315893487142e-05}
{"train_loss": 0.0840999037027359, "global_step": 191366, "epoch": 2150, "lr": 2.0049851669240954e-05}
{"train_loss": 0.05027490854263306, "global_step": 191367, "epoch": 2150, "lr": 2.004938744902121e-05}
{"train_loss": 0.07045971602201462, "global_step": 191368, "epoch": 2150, "lr": 2.0048923232827948e-05}
{"train_loss": 0.04981997609138489, "global_step": 191369, "epoch": 2150, "lr": 2.004845902066126e-05}
{"train_loss": 0.03126924857497215, "global_step": 191370, "epoch": 2150, "lr": 2.004799481252117e-05}
{"train_loss": 0.08618961274623871, "global_step": 191371, "epoch": 2150, "lr": 2.0047530608407784e-05}
{"train_loss": 0.07342291623353958, "global_step": 191372, "epoch": 2150, "lr": 2.0047066408321126e-05}
{"train_loss": 0.08150915056467056, "global_step": 191373, "epoch": 2150, "lr": 2.004660221226129e-05}
{"train_loss": 0.0911504328250885, "global_step": 191374, "epoch": 2150, "lr": 2.0046138020228307e-05}
{"train_loss": 0.057327087968587875, "global_step": 191375, "epoch": 2150, "lr": 2.0045673832222268e-05}
{"train_loss": 0.05724501982331276, "global_step": 191376, "epoch": 2150, "lr": 2.004520964824321e-05}
{"train_loss": 0.0623634047806263, "global_step": 191377, "epoch": 2150, "lr": 2.0044745468291226e-05}
{"train_loss": 0.033952344208955765, "global_step": 191378, "epoch": 2150, "lr": 2.004428129236634e-05}
{"train_loss": 0.06054835021495819, "global_step": 191379, "epoch": 2150, "lr": 2.0043817120468654e-05}
{"train_loss": 0.056058626621961594, "global_step": 191380, "epoch": 2150, "lr": 2.00433529525982e-05}
{"train_loss": 0.08856148272752762, "global_step": 191381, "epoch": 2150, "lr": 2.0042888788755066e-05}
{"train_loss": 0.07304980605840683, "global_step": 191382, "epoch": 2150, "lr": 2.0042424628939275e-05}
{"train_loss": 0.05544107407331467, "global_step": 191383, "epoch": 2150, "lr": 2.0041960473150944e-05}
{"train_loss": 0.10750571638345718, "global_step": 191384, "epoch": 2150, "lr": 2.0041496321390084e-05}
{"train_loss": 0.0517997182905674, "global_step": 191385, "epoch": 2150, "lr": 2.00410321736568e-05}
{"train_loss": 0.055899735540151596, "global_step": 191386, "epoch": 2150, "lr": 2.0040568029951114e-05}
{"train_loss": 0.04914805665612221, "global_step": 191387, "epoch": 2150, "lr": 2.004010389027312e-05}
{"train_loss": 0.10449840128421783, "global_step": 191388, "epoch": 2150, "lr": 2.003963975462286e-05}
{"train_loss": 0.04446489363908768, "global_step": 191389, "epoch": 2150, "lr": 2.003917562300042e-05}
{"train_loss": 0.08655077964067459, "global_step": 191390, "epoch": 2150, "lr": 2.0038711495405832e-05}
{"train_loss": 0.08325808495283127, "global_step": 191391, "epoch": 2150, "lr": 2.0038247371839186e-05}
{"train_loss": 0.10045677423477173, "global_step": 191392, "epoch": 2150, "lr": 2.003778325230054e-05}
{"train_loss": 0.0520910918712616, "global_step": 191393, "epoch": 2150, "lr": 2.003731913678993e-05}
{"train_loss": 0.0330851748585701, "global_step": 191394, "epoch": 2150, "lr": 2.003685502530745e-05}
{"train_loss": 0.04409515857696533, "global_step": 191395, "epoch": 2150, "lr": 2.003639091785313e-05}
{"train_loss": 0.04725015535950661, "global_step": 191396, "epoch": 2150, "lr": 2.0035926814427075e-05}
{"train_loss": 0.041772447526454926, "global_step": 191397, "epoch": 2150, "lr": 2.0035462715029308e-05}
{"train_loss": 0.0784793570637703, "global_step": 191398, "epoch": 2150, "lr": 2.0034998619659924e-05}
{"train_loss": 0.050233885645866394, "global_step": 191399, "epoch": 2150, "lr": 2.0034534528318948e-05}
{"train_loss": 0.04868939518928528, "global_step": 191400, "epoch": 2150, "lr": 2.0034070441006487e-05}
{"train_loss": 0.0614277645945549, "global_step": 191401, "epoch": 2150, "lr": 2.0033606357722556e-05}
{"train_loss": 0.08802119642496109, "global_step": 191402, "epoch": 2150, "lr": 2.003314227846725e-05}
{"train_loss": 0.08206547051668167, "global_step": 191403, "epoch": 2150, "lr": 2.0032678203240635e-05}
{"train_loss": 0.02141229622066021, "global_step": 191404, "epoch": 2150, "lr": 2.0032214132042742e-05}
{"train_loss": 0.12569035589694977, "global_step": 191405, "epoch": 2150, "lr": 2.003175006487367e-05}
{"train_loss": 0.08926449716091156, "global_step": 191406, "epoch": 2150, "lr": 2.003128600173345e-05}
{"train_loss": 0.0696818083524704, "global_step": 191407, "epoch": 2150, "lr": 2.0030821942622173e-05}
{"train_loss": 0.10609658807516098, "global_step": 191408, "epoch": 2150, "lr": 2.0030357887539875e-05}
{"train_loss": 0.09229784458875656, "global_step": 191409, "epoch": 2150, "lr": 2.0029893836486642e-05}
{"train_loss": 0.05980326607823372, "global_step": 191410, "epoch": 2150, "lr": 2.0029429789462508e-05}
{"train_loss": 0.0545942485332489, "global_step": 191411, "epoch": 2150, "lr": 2.0028965746467566e-05}
{"train_loss": 0.08396846801042557, "global_step": 191412, "epoch": 2150, "lr": 2.0028501707501853e-05}
{"train_loss": 0.09931378066539764, "global_step": 191413, "epoch": 2150, "lr": 2.0028037672565458e-05}
{"train_loss": 0.06726200878620148, "global_step": 191414, "epoch": 2150, "lr": 2.002757364165841e-05}
{"train_loss": 0.058630213141441345, "global_step": 191415, "epoch": 2150, "lr": 2.0027109614780804e-05}
{"train_loss": 0.052941303700208664, "global_step": 191416, "epoch": 2150, "lr": 2.0026645591932674e-05}
{"train_loss": 0.08091054111719131, "global_step": 191417, "epoch": 2150, "lr": 2.002618157311411e-05}
{"train_loss": 0.10353723913431168, "global_step": 191418, "epoch": 2150, "lr": 2.0025717558325145e-05}
{"train_loss": 0.07278095185756683, "global_step": 191419, "epoch": 2150, "lr": 2.002525354756588e-05}
{"train_loss": 0.051958758383989334, "global_step": 191420, "epoch": 2150, "lr": 2.0024789540836332e-05}
{"train_loss": 0.09175833314657211, "global_step": 191421, "epoch": 2150, "lr": 2.0024325538136585e-05}
{"train_loss": 0.06542034447193146, "global_step": 191422, "epoch": 2150, "lr": 2.0023861539466722e-05}
{"train_loss": 0.05250854045152664, "global_step": 191423, "epoch": 2150, "lr": 2.0023397544826768e-05}
{"train_loss": 0.08634798228740692, "global_step": 191424, "epoch": 2150, "lr": 2.0022933554216816e-05}
{"train_loss": 0.08022087067365646, "global_step": 191425, "epoch": 2150, "lr": 2.0022469567636916e-05}
{"train_loss": 0.08071538805961609, "global_step": 191426, "epoch": 2150, "lr": 2.0022005585087117e-05}
{"train_loss": 0.11188993602991104, "global_step": 191427, "epoch": 2150, "lr": 2.0021541606567507e-05}
{"train_loss": 0.08954101800918579, "global_step": 191428, "epoch": 2150, "lr": 2.0021077632078117e-05}
{"train_loss": 0.04297109320759773, "global_step": 191429, "epoch": 2150, "lr": 2.0020613661619046e-05}
{"train_loss": 0.05101960152387619, "global_step": 191430, "epoch": 2150, "lr": 2.0020149695190344e-05}
{"train_loss": 0.10480708628892899, "global_step": 191431, "epoch": 2150, "lr": 2.001968573279204e-05}
{"train_loss": 0.04124150052666664, "global_step": 191432, "epoch": 2150, "lr": 2.001922177442425e-05}
{"train_loss": 0.08471870422363281, "global_step": 191433, "epoch": 2150, "lr": 2.001875782008699e-05}
{"train_loss": 0.05089348182082176, "global_step": 191434, "epoch": 2150, "lr": 2.0018293869780358e-05}
{"train_loss": 0.10650549829006195, "global_step": 191435, "epoch": 2150, "lr": 2.001782992350439e-05}
{"train_loss": 0.02990025468170643, "global_step": 191436, "epoch": 2150, "lr": 2.001736598125917e-05}
{"train_loss": 0.06412018090486526, "global_step": 191437, "epoch": 2150, "lr": 2.0016902043044734e-05}
{"train_loss": 0.07086808016795791, "global_step": 191438, "epoch": 2150, "lr": 2.0016438108861156e-05, "train/sim_max_reward_0": 0.542711348818192, "train/sim_max_reward_1": 0.48849219434750735, "train/sim_max_reward_2": 0.5306447443130395, "train/sim_max_reward_3": 0.9162220566950136, "train/sim_max_reward_4": 0.5394890272235177, "train/sim_max_reward_5": 0.9773766317349215, "test/sim_max_reward_4500000": 0.9941326663960474, "test/sim_max_reward_4500001": 0.9920837208444033, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.7466855302506864, "test/sim_max_reward_4500004": 0.9909259591545211, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9849036935666845, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.04843398156077542, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5381913564853379, "test/sim_max_reward_4500014": 0.368925170083764, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 0.9643253093369906, "test/sim_max_reward_4500017": 1.0, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9682750219648926, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9966811122806388, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.9532692888294714, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.7274737713061707, "test/sim_max_reward_4500026": 0.03760329356664699, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.609707623590459, "test/sim_max_reward_4500029": 1.0, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.5039476610968938, "test/sim_max_reward_4500032": 0.6797006149658638, "test/sim_max_reward_4500033": 0.8451677236544031, "test/sim_max_reward_4500034": 0.8916289185475796, "test/sim_max_reward_4500035": 0.7188416404899041, "test/sim_max_reward_4500036": 0.46485549318057995, "test/sim_max_reward_4500037": 0.9354667136934627, "test/sim_max_reward_4500038": 0.1656018025255626, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9089652501473079, "test/sim_max_reward_4500041": 0.5205694591024422, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.19230404260800502, "test/sim_max_reward_4500045": 0.9826453119598384, "test/sim_max_reward_4500046": 0.65949624648164, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9546430027469184, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.6658226671886985, "test/mean_score": 0.6869090276083578, "val_loss": 7.85083532333374, "train_action_mse_error": 16.73014259338379}
{"train_loss": 0.07852014899253845, "global_step": 191439, "epoch": 2151, "lr": 2.001597417870853e-05}
{"train_loss": 0.09890194982290268, "global_step": 191440, "epoch": 2151, "lr": 2.0015510252586867e-05}
{"train_loss": 0.050251469016075134, "global_step": 191441, "epoch": 2151, "lr": 2.0015046330496267e-05}
{"train_loss": 0.06763061881065369, "global_step": 191442, "epoch": 2151, "lr": 2.0014582412436765e-05}
{"train_loss": 0.08866724371910095, "global_step": 191443, "epoch": 2151, "lr": 2.0014118498408458e-05}
{"train_loss": 0.08683241903781891, "global_step": 191444, "epoch": 2151, "lr": 2.0013654588411368e-05}
{"train_loss": 0.11314483731985092, "global_step": 191445, "epoch": 2151, "lr": 2.0013190682445592e-05}
{"train_loss": 0.08215328305959702, "global_step": 191446, "epoch": 2151, "lr": 2.001272678051116e-05}
{"train_loss": 0.09118849784135818, "global_step": 191447, "epoch": 2151, "lr": 2.001226288260817e-05}
{"train_loss": 0.0703401193022728, "global_step": 191448, "epoch": 2151, "lr": 2.0011798988736645e-05}
{"train_loss": 0.08964661508798599, "global_step": 191449, "epoch": 2151, "lr": 2.001133509889669e-05}
{"train_loss": 0.06775647401809692, "global_step": 191450, "epoch": 2151, "lr": 2.001087121308832e-05}
{"train_loss": 0.0589798279106617, "global_step": 191451, "epoch": 2151, "lr": 2.0010407331311648e-05}
{"train_loss": 0.04774248227477074, "global_step": 191452, "epoch": 2151, "lr": 2.0009943453566693e-05}
{"train_loss": 0.09718057513237, "global_step": 191453, "epoch": 2151, "lr": 2.0009479579853553e-05}
{"train_loss": 0.1048518717288971, "global_step": 191454, "epoch": 2151, "lr": 2.000901571017226e-05}
{"train_loss": 0.07764697074890137, "global_step": 191455, "epoch": 2151, "lr": 2.00085518445229e-05}
{"train_loss": 0.056716956198215485, "global_step": 191456, "epoch": 2151, "lr": 2.0008087982905506e-05}
{"train_loss": 0.08310346305370331, "global_step": 191457, "epoch": 2151, "lr": 2.000762412532016e-05}
{"train_loss": 0.07376343011856079, "global_step": 191458, "epoch": 2151, "lr": 2.0007160271766944e-05}
{"train_loss": 0.050209902226924896, "global_step": 191459, "epoch": 2151, "lr": 2.000669642224588e-05}
{"train_loss": 0.04804214462637901, "global_step": 191460, "epoch": 2151, "lr": 2.0006232576757066e-05}
{"train_loss": 0.07539873570203781, "global_step": 191461, "epoch": 2151, "lr": 2.0005768735300533e-05}
{"train_loss": 0.045701757073402405, "global_step": 191462, "epoch": 2151, "lr": 2.0005304897876375e-05}
{"train_loss": 0.05973365530371666, "global_step": 191463, "epoch": 2151, "lr": 2.0004841064484643e-05}
{"train_loss": 0.05253990367054939, "global_step": 191464, "epoch": 2151, "lr": 2.000437723512537e-05}
{"train_loss": 0.06445872038602829, "global_step": 191465, "epoch": 2151, "lr": 2.000391340979867e-05}
{"train_loss": 0.05810856074094772, "global_step": 191466, "epoch": 2151, "lr": 2.0003449588504553e-05}
{"train_loss": 0.08138073235750198, "global_step": 191467, "epoch": 2151, "lr": 2.0002985771243126e-05}
{"train_loss": 0.07972824573516846, "global_step": 191468, "epoch": 2151, "lr": 2.000252195801443e-05}
{"train_loss": 0.06331239640712738, "global_step": 191469, "epoch": 2151, "lr": 2.0002058148818513e-05}
{"train_loss": 0.06902477145195007, "global_step": 191470, "epoch": 2151, "lr": 2.0001594343655472e-05}
{"train_loss": 0.07298235595226288, "global_step": 191471, "epoch": 2151, "lr": 2.0001130542525337e-05}
{"train_loss": 0.07416252046823502, "global_step": 191472, "epoch": 2151, "lr": 2.0000666745428198e-05}
{"train_loss": 0.07406114786863327, "global_step": 191473, "epoch": 2151, "lr": 2.0000202952364084e-05}
{"train_loss": 0.10700812190771103, "global_step": 191474, "epoch": 2151, "lr": 1.999973916333308e-05}
{"train_loss": 0.06452225148677826, "global_step": 191475, "epoch": 2151, "lr": 1.9999275378335264e-05}
{"train_loss": 0.04521317407488823, "global_step": 191476, "epoch": 2151, "lr": 1.999881159737066e-05}
{"train_loss": 0.029199542477726936, "global_step": 191477, "epoch": 2151, "lr": 1.999834782043937e-05}
{"train_loss": 0.03943134844303131, "global_step": 191478, "epoch": 2151, "lr": 1.9997884047541415e-05}
{"train_loss": 0.053096577525138855, "global_step": 191479, "epoch": 2151, "lr": 1.99974202786769e-05}
{"train_loss": 0.05499331280589104, "global_step": 191480, "epoch": 2151, "lr": 1.9996956513845844e-05}
{"train_loss": 0.08112670481204987, "global_step": 191481, "epoch": 2151, "lr": 1.9996492753048352e-05}
{"train_loss": 0.05130352824926376, "global_step": 191482, "epoch": 2151, "lr": 1.9996028996284448e-05}
{"train_loss": 0.06324541568756104, "global_step": 191483, "epoch": 2151, "lr": 1.999556524355423e-05}
{"train_loss": 0.0534285306930542, "global_step": 191484, "epoch": 2151, "lr": 1.9995101494857726e-05}
{"train_loss": 0.11035983264446259, "global_step": 191485, "epoch": 2151, "lr": 1.9994637750195033e-05}
{"train_loss": 0.06774049252271652, "global_step": 191486, "epoch": 2151, "lr": 1.999417400956618e-05}
{"train_loss": 0.05321013927459717, "global_step": 191487, "epoch": 2151, "lr": 1.9993710272971257e-05}
{"train_loss": 0.08827044069766998, "global_step": 191488, "epoch": 2151, "lr": 1.99932465404103e-05}
{"train_loss": 0.06122960150241852, "global_step": 191489, "epoch": 2151, "lr": 1.9992782811883398e-05}
{"train_loss": 0.0603606216609478, "global_step": 191490, "epoch": 2151, "lr": 1.999231908739059e-05}
{"train_loss": 0.08177404850721359, "global_step": 191491, "epoch": 2151, "lr": 1.9991855366931957e-05}
{"train_loss": 0.047876857221126556, "global_step": 191492, "epoch": 2151, "lr": 1.999139165050754e-05}
{"train_loss": 0.048581503331661224, "global_step": 191493, "epoch": 2151, "lr": 1.9990927938117415e-05}
{"train_loss": 0.05594509467482567, "global_step": 191494, "epoch": 2151, "lr": 1.9990464229761664e-05}
{"train_loss": 0.09400442987680435, "global_step": 191495, "epoch": 2151, "lr": 1.999000052544031e-05}
{"train_loss": 0.054804570972919464, "global_step": 191496, "epoch": 2151, "lr": 1.9989536825153453e-05}
{"train_loss": 0.0579444095492363, "global_step": 191497, "epoch": 2151, "lr": 1.998907312890112e-05}
{"train_loss": 0.034797973930835724, "global_step": 191498, "epoch": 2151, "lr": 1.9988609436683403e-05}
{"train_loss": 0.05898202210664749, "global_step": 191499, "epoch": 2151, "lr": 1.998814574850034e-05}
{"train_loss": 0.10035840421915054, "global_step": 191500, "epoch": 2151, "lr": 1.998768206435202e-05}
{"train_loss": 0.1289442777633667, "global_step": 191501, "epoch": 2151, "lr": 1.998721838423847e-05}
{"train_loss": 0.06606675684452057, "global_step": 191502, "epoch": 2151, "lr": 1.99867547081598e-05}
{"train_loss": 0.073808453977108, "global_step": 191503, "epoch": 2151, "lr": 1.9986291036116032e-05}
{"train_loss": 0.06810775399208069, "global_step": 191504, "epoch": 2151, "lr": 1.9985827368107236e-05}
{"train_loss": 0.10748155415058136, "global_step": 191505, "epoch": 2151, "lr": 1.9985363704133487e-05}
{"train_loss": 0.08025982975959778, "global_step": 191506, "epoch": 2151, "lr": 1.9984900044194827e-05}
{"train_loss": 0.07609367370605469, "global_step": 191507, "epoch": 2151, "lr": 1.998443638829135e-05}
{"train_loss": 0.04299166798591614, "global_step": 191508, "epoch": 2151, "lr": 1.99839727364231e-05}
{"train_loss": 0.08780534565448761, "global_step": 191509, "epoch": 2151, "lr": 1.998350908859012e-05}
{"train_loss": 0.09418068826198578, "global_step": 191510, "epoch": 2151, "lr": 1.998304544479251e-05}
{"train_loss": 0.059226635843515396, "global_step": 191511, "epoch": 2151, "lr": 1.9982581805030293e-05}
{"train_loss": 0.0498121976852417, "global_step": 191512, "epoch": 2151, "lr": 1.9982118169303555e-05}
{"train_loss": 0.10158807784318924, "global_step": 191513, "epoch": 2151, "lr": 1.998165453761237e-05}
{"train_loss": 0.08639872819185257, "global_step": 191514, "epoch": 2151, "lr": 1.9981190909956772e-05}
{"train_loss": 0.09776660054922104, "global_step": 191515, "epoch": 2151, "lr": 1.9980727286336852e-05}
{"train_loss": 0.05913708731532097, "global_step": 191516, "epoch": 2151, "lr": 1.998026366675264e-05}
{"train_loss": 0.038704391568899155, "global_step": 191517, "epoch": 2151, "lr": 1.9979800051204234e-05}
{"train_loss": 0.10569118708372116, "global_step": 191518, "epoch": 2151, "lr": 1.997933643969166e-05}
{"train_loss": 0.07845450192689896, "global_step": 191519, "epoch": 2151, "lr": 1.9978872832215013e-05}
{"train_loss": 0.060235992074012756, "global_step": 191520, "epoch": 2151, "lr": 1.9978409228774325e-05}
{"train_loss": 0.05011134222149849, "global_step": 191521, "epoch": 2151, "lr": 1.9977945629369692e-05}
{"train_loss": 0.06853540986776352, "global_step": 191522, "epoch": 2151, "lr": 1.9977482034001143e-05}
{"train_loss": 0.08346167206764221, "global_step": 191523, "epoch": 2151, "lr": 1.997701844266877e-05}
{"train_loss": 0.04315374791622162, "global_step": 191524, "epoch": 2151, "lr": 1.997655485537261e-05}
{"train_loss": 0.058087386190891266, "global_step": 191525, "epoch": 2151, "lr": 1.9976091272112745e-05}
{"train_loss": 0.090727798640728, "global_step": 191526, "epoch": 2151, "lr": 1.9975627692889213e-05}
{"train_loss": 0.07106714182941432, "global_step": 191527, "epoch": 2151, "lr": 1.9975164117702112e-05, "val_loss": 7.912839412689209}
{"train_loss": 0.05160841345787048, "global_step": 191528, "epoch": 2152, "lr": 1.9974700546551465e-05}
{"train_loss": 0.0754443034529686, "global_step": 191529, "epoch": 2152, "lr": 1.9974236979437354e-05}
{"train_loss": 0.1260705143213272, "global_step": 191530, "epoch": 2152, "lr": 1.997377341635986e-05}
{"train_loss": 0.08008240163326263, "global_step": 191531, "epoch": 2152, "lr": 1.9973309857319006e-05}
{"train_loss": 0.062108784914016724, "global_step": 191532, "epoch": 2152, "lr": 1.9972846302314897e-05}
{"train_loss": 0.1245674341917038, "global_step": 191533, "epoch": 2152, "lr": 1.997238275134755e-05}
{"train_loss": 0.057635996490716934, "global_step": 191534, "epoch": 2152, "lr": 1.9971919204417073e-05}
{"train_loss": 0.0962890163064003, "global_step": 191535, "epoch": 2152, "lr": 1.997145566152348e-05}
{"train_loss": 0.05944453552365303, "global_step": 191536, "epoch": 2152, "lr": 1.997099212266688e-05}
{"train_loss": 0.1400926113128662, "global_step": 191537, "epoch": 2152, "lr": 1.9970528587847297e-05}
{"train_loss": 0.04434458166360855, "global_step": 191538, "epoch": 2152, "lr": 1.9970065057064828e-05}
{"train_loss": 0.05171891301870346, "global_step": 191539, "epoch": 2152, "lr": 1.9969601530319504e-05}
{"train_loss": 0.09603888541460037, "global_step": 191540, "epoch": 2152, "lr": 1.9969138007611417e-05}
{"train_loss": 0.0627085492014885, "global_step": 191541, "epoch": 2152, "lr": 1.9968674488940607e-05}
{"train_loss": 0.06302390247583389, "global_step": 191542, "epoch": 2152, "lr": 1.9968210974307132e-05}
{"train_loss": 0.05168714374303818, "global_step": 191543, "epoch": 2152, "lr": 1.9967747463711078e-05}
{"train_loss": 0.07666444778442383, "global_step": 191544, "epoch": 2152, "lr": 1.9967283957152476e-05}
{"train_loss": 0.05498750880360603, "global_step": 191545, "epoch": 2152, "lr": 1.996682045463143e-05}
{"train_loss": 0.060760099440813065, "global_step": 191546, "epoch": 2152, "lr": 1.9966356956147974e-05}
{"train_loss": 0.04118387773633003, "global_step": 191547, "epoch": 2152, "lr": 1.996589346170216e-05}
{"train_loss": 0.0716266930103302, "global_step": 191548, "epoch": 2152, "lr": 1.996542997129407e-05}
{"train_loss": 0.04595045745372772, "global_step": 191549, "epoch": 2152, "lr": 1.9964966484923774e-05}
{"train_loss": 0.0845460444688797, "global_step": 191550, "epoch": 2152, "lr": 1.9964503002591307e-05}
{"train_loss": 0.06768518686294556, "global_step": 191551, "epoch": 2152, "lr": 1.9964039524296756e-05}
{"train_loss": 0.043087977916002274, "global_step": 191552, "epoch": 2152, "lr": 1.9963576050040166e-05}
{"train_loss": 0.037104036659002304, "global_step": 191553, "epoch": 2152, "lr": 1.996311257982162e-05}
{"train_loss": 0.09025102108716965, "global_step": 191554, "epoch": 2152, "lr": 1.996264911364115e-05}
{"train_loss": 0.11429665982723236, "global_step": 191555, "epoch": 2152, "lr": 1.9962185651498856e-05}
{"train_loss": 0.05623957887291908, "global_step": 191556, "epoch": 2152, "lr": 1.996172219339476e-05}
{"train_loss": 0.06947124004364014, "global_step": 191557, "epoch": 2152, "lr": 1.9961258739328964e-05}
{"train_loss": 0.06195449084043503, "global_step": 191558, "epoch": 2152, "lr": 1.9960795289301494e-05}
{"train_loss": 0.12275063246488571, "global_step": 191559, "epoch": 2152, "lr": 1.996033184331244e-05}
{"train_loss": 0.08276079595088959, "global_step": 191560, "epoch": 2152, "lr": 1.9959868401361843e-05}
{"train_loss": 0.07840476930141449, "global_step": 191561, "epoch": 2152, "lr": 1.995940496344979e-05}
{"train_loss": 0.038205135613679886, "global_step": 191562, "epoch": 2152, "lr": 1.9958941529576307e-05}
{"train_loss": 0.08503789454698563, "global_step": 191563, "epoch": 2152, "lr": 1.9958478099741503e-05}
{"train_loss": 0.09624774754047394, "global_step": 191564, "epoch": 2152, "lr": 1.9958014673945396e-05}
{"train_loss": 0.06369577348232269, "global_step": 191565, "epoch": 2152, "lr": 1.9957551252188067e-05}
{"train_loss": 0.04228208214044571, "global_step": 191566, "epoch": 2152, "lr": 1.9957087834469596e-05}
{"train_loss": 0.058305222541093826, "global_step": 191567, "epoch": 2152, "lr": 1.995662442079001e-05}
{"train_loss": 0.08979745209217072, "global_step": 191568, "epoch": 2152, "lr": 1.995616101114941e-05}
{"train_loss": 0.06650005280971527, "global_step": 191569, "epoch": 2152, "lr": 1.995569760554782e-05}
{"train_loss": 0.06482458114624023, "global_step": 191570, "epoch": 2152, "lr": 1.9955234203985335e-05}
{"train_loss": 0.09399738162755966, "global_step": 191571, "epoch": 2152, "lr": 1.9954770806461985e-05}
{"train_loss": 0.08522618561983109, "global_step": 191572, "epoch": 2152, "lr": 1.9954307412977868e-05}
{"train_loss": 0.07178449630737305, "global_step": 191573, "epoch": 2152, "lr": 1.9953844023533014e-05}
{"train_loss": 0.07222015410661697, "global_step": 191574, "epoch": 2152, "lr": 1.9953380638127517e-05}
{"train_loss": 0.06789687275886536, "global_step": 191575, "epoch": 2152, "lr": 1.99529172567614e-05}
{"train_loss": 0.05477851629257202, "global_step": 191576, "epoch": 2152, "lr": 1.9952453879434767e-05}
{"train_loss": 0.06680098176002502, "global_step": 191577, "epoch": 2152, "lr": 1.995199050614764e-05}
{"train_loss": 0.0629706159234047, "global_step": 191578, "epoch": 2152, "lr": 1.9951527136900122e-05}
{"train_loss": 0.05195540934801102, "global_step": 191579, "epoch": 2152, "lr": 1.9951063771692258e-05}
{"train_loss": 0.08963324874639511, "global_step": 191580, "epoch": 2152, "lr": 1.9950600410524084e-05}
{"train_loss": 0.12616536021232605, "global_step": 191581, "epoch": 2152, "lr": 1.9950137053395705e-05}
{"train_loss": 0.08481986075639725, "global_step": 191582, "epoch": 2152, "lr": 1.9949673700307148e-05}
{"train_loss": 0.08476811647415161, "global_step": 191583, "epoch": 2152, "lr": 1.9949210351258508e-05}
{"train_loss": 0.05925200879573822, "global_step": 191584, "epoch": 2152, "lr": 1.994874700624981e-05}
{"train_loss": 0.07509056478738785, "global_step": 191585, "epoch": 2152, "lr": 1.9948283665281154e-05}
{"train_loss": 0.05832362174987793, "global_step": 191586, "epoch": 2152, "lr": 1.9947820328352573e-05}
{"train_loss": 0.07194086909294128, "global_step": 191587, "epoch": 2152, "lr": 1.9947356995464156e-05}
{"train_loss": 0.07308900356292725, "global_step": 191588, "epoch": 2152, "lr": 1.994689366661593e-05}
{"train_loss": 0.09260348975658417, "global_step": 191589, "epoch": 2152, "lr": 1.9946430341807997e-05}
{"train_loss": 0.045607369393110275, "global_step": 191590, "epoch": 2152, "lr": 1.9945967021040385e-05}
{"train_loss": 0.06461377441883087, "global_step": 191591, "epoch": 2152, "lr": 1.994550370431319e-05}
{"train_loss": 0.0488828681409359, "global_step": 191592, "epoch": 2152, "lr": 1.9945040391626434e-05}
{"train_loss": 0.06275677680969238, "global_step": 191593, "epoch": 2152, "lr": 1.994457708298022e-05}
{"train_loss": 0.1270483285188675, "global_step": 191594, "epoch": 2152, "lr": 1.9944113778374574e-05}
{"train_loss": 0.05206397548317909, "global_step": 191595, "epoch": 2152, "lr": 1.9943650477809595e-05}
{"train_loss": 0.04593484103679657, "global_step": 191596, "epoch": 2152, "lr": 1.9943187181285305e-05}
{"train_loss": 0.08243504911661148, "global_step": 191597, "epoch": 2152, "lr": 1.9942723888801806e-05}
{"train_loss": 0.05914868414402008, "global_step": 191598, "epoch": 2152, "lr": 1.9942260600359126e-05}
{"train_loss": 0.08178658783435822, "global_step": 191599, "epoch": 2152, "lr": 1.9941797315957355e-05}
{"train_loss": 0.05381237342953682, "global_step": 191600, "epoch": 2152, "lr": 1.9941334035596527e-05}
{"train_loss": 0.02363821677863598, "global_step": 191601, "epoch": 2152, "lr": 1.9940870759276724e-05}
{"train_loss": 0.06505141407251358, "global_step": 191602, "epoch": 2152, "lr": 1.994040748699802e-05}
{"train_loss": 0.07435115426778793, "global_step": 191603, "epoch": 2152, "lr": 1.9939944218760444e-05}
{"train_loss": 0.09071327745914459, "global_step": 191604, "epoch": 2152, "lr": 1.9939480954564095e-05}
{"train_loss": 0.059533603489398956, "global_step": 191605, "epoch": 2152, "lr": 1.9939017694408994e-05}
{"train_loss": 0.07322835922241211, "global_step": 191606, "epoch": 2152, "lr": 1.9938554438295247e-05}
{"train_loss": 0.050089385360479355, "global_step": 191607, "epoch": 2152, "lr": 1.993809118622288e-05}
{"train_loss": 0.07824572175741196, "global_step": 191608, "epoch": 2152, "lr": 1.9937627938191982e-05}
{"train_loss": 0.05627182498574257, "global_step": 191609, "epoch": 2152, "lr": 1.9937164694202594e-05}
{"train_loss": 0.037102989852428436, "global_step": 191610, "epoch": 2152, "lr": 1.9936701454254796e-05}
{"train_loss": 0.06823521107435226, "global_step": 191611, "epoch": 2152, "lr": 1.993623821834863e-05}
{"train_loss": 0.08792304247617722, "global_step": 191612, "epoch": 2152, "lr": 1.993577498648419e-05}
{"train_loss": 0.10490743815898895, "global_step": 191613, "epoch": 2152, "lr": 1.9935311758661497e-05}
{"train_loss": 0.03827035799622536, "global_step": 191614, "epoch": 2152, "lr": 1.993484853488065e-05}
{"train_loss": 0.061390187591314316, "global_step": 191615, "epoch": 2152, "lr": 1.9934385315141683e-05}
{"train_loss": 0.07089988433159469, "global_step": 191616, "epoch": 2152, "lr": 1.993392209944469e-05, "val_loss": 7.975566864013672}
{"train_loss": 0.03603864461183548, "global_step": 191617, "epoch": 2153, "lr": 1.9933458887789696e-05}
{"train_loss": 0.11691649258136749, "global_step": 191618, "epoch": 2153, "lr": 1.9932995680176798e-05}
{"train_loss": 0.04861875995993614, "global_step": 191619, "epoch": 2153, "lr": 1.9932532476606043e-05}
{"train_loss": 0.09077351540327072, "global_step": 191620, "epoch": 2153, "lr": 1.9932069277077474e-05}
{"train_loss": 0.031045585870742798, "global_step": 191621, "epoch": 2153, "lr": 1.993160608159119e-05}
{"train_loss": 0.07143453508615494, "global_step": 191622, "epoch": 2153, "lr": 1.993114289014722e-05}
{"train_loss": 0.08226552605628967, "global_step": 191623, "epoch": 2153, "lr": 1.9930679702745657e-05}
{"train_loss": 0.1038798913359642, "global_step": 191624, "epoch": 2153, "lr": 1.993021651938653e-05}
{"train_loss": 0.08382296562194824, "global_step": 191625, "epoch": 2153, "lr": 1.9929753340069933e-05}
{"train_loss": 0.08364204317331314, "global_step": 191626, "epoch": 2153, "lr": 1.9929290164795906e-05}
{"train_loss": 0.0625808984041214, "global_step": 191627, "epoch": 2153, "lr": 1.9928826993564527e-05}
{"train_loss": 0.15894058346748352, "global_step": 191628, "epoch": 2153, "lr": 1.992836382637584e-05}
{"train_loss": 0.07480044662952423, "global_step": 191629, "epoch": 2153, "lr": 1.992790066322993e-05}
{"train_loss": 0.06336446106433868, "global_step": 191630, "epoch": 2153, "lr": 1.9927437504126834e-05}
{"train_loss": 0.047554176300764084, "global_step": 191631, "epoch": 2153, "lr": 1.992697434906664e-05}
{"train_loss": 0.05905907601118088, "global_step": 191632, "epoch": 2153, "lr": 1.9926511198049384e-05}
{"train_loss": 0.07347218692302704, "global_step": 191633, "epoch": 2153, "lr": 1.9926048051075165e-05}
{"train_loss": 0.0861203595995903, "global_step": 191634, "epoch": 2153, "lr": 1.9925584908143997e-05}
{"train_loss": 0.0938844382762909, "global_step": 191635, "epoch": 2153, "lr": 1.9925121769255984e-05}
{"train_loss": 0.08387608826160431, "global_step": 191636, "epoch": 2153, "lr": 1.9924658634411158e-05}
{"train_loss": 0.07041057199239731, "global_step": 191637, "epoch": 2153, "lr": 1.992419550360961e-05}
{"train_loss": 0.05018464848399162, "global_step": 191638, "epoch": 2153, "lr": 1.9923732376851373e-05}
{"train_loss": 0.0586923211812973, "global_step": 191639, "epoch": 2153, "lr": 1.9923269254136523e-05}
{"train_loss": 0.05957426130771637, "global_step": 191640, "epoch": 2153, "lr": 1.9922806135465137e-05}
{"train_loss": 0.10133245587348938, "global_step": 191641, "epoch": 2153, "lr": 1.9922343020837253e-05}
{"train_loss": 0.1135171577334404, "global_step": 191642, "epoch": 2153, "lr": 1.9921879910252955e-05}
{"train_loss": 0.08877934515476227, "global_step": 191643, "epoch": 2153, "lr": 1.992141680371228e-05}
{"train_loss": 0.05789633467793465, "global_step": 191644, "epoch": 2153, "lr": 1.9920953701215317e-05}
{"train_loss": 0.06177714467048645, "global_step": 191645, "epoch": 2153, "lr": 1.99204906027621e-05}
{"train_loss": 0.053944557905197144, "global_step": 191646, "epoch": 2153, "lr": 1.9920027508352718e-05}
{"train_loss": 0.045311760157346725, "global_step": 191647, "epoch": 2153, "lr": 1.991956441798721e-05}
{"train_loss": 0.07982784509658813, "global_step": 191648, "epoch": 2153, "lr": 1.9919101331665667e-05}
{"train_loss": 0.03995600342750549, "global_step": 191649, "epoch": 2153, "lr": 1.9918638249388116e-05}
{"train_loss": 0.07020942121744156, "global_step": 191650, "epoch": 2153, "lr": 1.9918175171154655e-05}
{"train_loss": 0.09109900146722794, "global_step": 191651, "epoch": 2153, "lr": 1.9917712096965313e-05}
{"train_loss": 0.035816192626953125, "global_step": 191652, "epoch": 2153, "lr": 1.9917249026820184e-05}
{"train_loss": 0.06909047812223434, "global_step": 191653, "epoch": 2153, "lr": 1.9916785960719298e-05}
{"train_loss": 0.052846506237983704, "global_step": 191654, "epoch": 2153, "lr": 1.9916322898662747e-05}
{"train_loss": 0.06500936299562454, "global_step": 191655, "epoch": 2153, "lr": 1.991585984065057e-05}
{"train_loss": 0.05722849816083908, "global_step": 191656, "epoch": 2153, "lr": 1.991539678668285e-05}
{"train_loss": 0.04166550189256668, "global_step": 191657, "epoch": 2153, "lr": 1.991493373675964e-05}
{"train_loss": 0.10954996198415756, "global_step": 191658, "epoch": 2153, "lr": 1.9914470690880982e-05}
{"train_loss": 0.09322018176317215, "global_step": 191659, "epoch": 2153, "lr": 1.9914007649046977e-05}
{"train_loss": 0.05510908365249634, "global_step": 191660, "epoch": 2153, "lr": 1.9913544611257646e-05}
{"train_loss": 0.07465628534555435, "global_step": 191661, "epoch": 2153, "lr": 1.9913081577513093e-05}
{"train_loss": 0.041200507432222366, "global_step": 191662, "epoch": 2153, "lr": 1.991261854781334e-05}
{"train_loss": 0.057484161108732224, "global_step": 191663, "epoch": 2153, "lr": 1.991215552215848e-05}
{"train_loss": 0.05684977024793625, "global_step": 191664, "epoch": 2153, "lr": 1.9911692500548556e-05}
{"train_loss": 0.08939454704523087, "global_step": 191665, "epoch": 2153, "lr": 1.9911229482983658e-05}
{"train_loss": 0.0849725529551506, "global_step": 191666, "epoch": 2153, "lr": 1.9910766469463802e-05}
{"train_loss": 0.11575151234865189, "global_step": 191667, "epoch": 2153, "lr": 1.99103034599891e-05}
{"train_loss": 0.08857499808073044, "global_step": 191668, "epoch": 2153, "lr": 1.990984045455957e-05}
{"train_loss": 0.0942944660782814, "global_step": 191669, "epoch": 2153, "lr": 1.9909377453175315e-05}
{"train_loss": 0.05293271318078041, "global_step": 191670, "epoch": 2153, "lr": 1.9908914455836357e-05}
{"train_loss": 0.04876586049795151, "global_step": 191671, "epoch": 2153, "lr": 1.99084514625428e-05}
{"train_loss": 0.09463166445493698, "global_step": 191672, "epoch": 2153, "lr": 1.9907988473294666e-05}
{"train_loss": 0.10797411948442459, "global_step": 191673, "epoch": 2153, "lr": 1.990752548809205e-05}
{"train_loss": 0.06575015187263489, "global_step": 191674, "epoch": 2153, "lr": 1.9907062506934986e-05}
{"train_loss": 0.0409083254635334, "global_step": 191675, "epoch": 2153, "lr": 1.990659952982355e-05}
{"train_loss": 0.042621348053216934, "global_step": 191676, "epoch": 2153, "lr": 1.9906136556757827e-05}
{"train_loss": 0.07932517677545547, "global_step": 191677, "epoch": 2153, "lr": 1.9905673587737833e-05}
{"train_loss": 0.0625104233622551, "global_step": 191678, "epoch": 2153, "lr": 1.9905210622763676e-05}
{"train_loss": 0.08970694243907928, "global_step": 191679, "epoch": 2153, "lr": 1.9904747661835378e-05}
{"train_loss": 0.08710683882236481, "global_step": 191680, "epoch": 2153, "lr": 1.9904284704953035e-05}
{"train_loss": 0.06476561725139618, "global_step": 191681, "epoch": 2153, "lr": 1.990382175211668e-05}
{"train_loss": 0.05042550712823868, "global_step": 191682, "epoch": 2153, "lr": 1.9903358803326405e-05}
{"train_loss": 0.059437621384859085, "global_step": 191683, "epoch": 2153, "lr": 1.9902895858582237e-05}
{"train_loss": 0.056396760046482086, "global_step": 191684, "epoch": 2153, "lr": 1.9902432917884277e-05}
{"train_loss": 0.04920870438218117, "global_step": 191685, "epoch": 2153, "lr": 1.9901969981232553e-05}
{"train_loss": 0.059162434190511703, "global_step": 191686, "epoch": 2153, "lr": 1.990150704862716e-05}
{"train_loss": 0.08974777162075043, "global_step": 191687, "epoch": 2153, "lr": 1.9901044120068125e-05}
{"train_loss": 0.06806578487157822, "global_step": 191688, "epoch": 2153, "lr": 1.990058119555554e-05}
{"train_loss": 0.1183227151632309, "global_step": 191689, "epoch": 2153, "lr": 1.9900118275089442e-05}
{"train_loss": 0.10579900443553925, "global_step": 191690, "epoch": 2153, "lr": 1.989965535866992e-05}
{"train_loss": 0.06443221867084503, "global_step": 191691, "epoch": 2153, "lr": 1.989919244629701e-05}
{"train_loss": 0.08806725591421127, "global_step": 191692, "epoch": 2153, "lr": 1.9898729537970807e-05}
{"train_loss": 0.034605883061885834, "global_step": 191693, "epoch": 2153, "lr": 1.9898266633691327e-05}
{"train_loss": 0.08823037147521973, "global_step": 191694, "epoch": 2153, "lr": 1.989780373345868e-05}
{"train_loss": 0.07105031609535217, "global_step": 191695, "epoch": 2153, "lr": 1.9897340837272905e-05}
{"train_loss": 0.06194361671805382, "global_step": 191696, "epoch": 2153, "lr": 1.9896877945134053e-05}
{"train_loss": 0.04180644080042839, "global_step": 191697, "epoch": 2153, "lr": 1.9896415057042206e-05}
{"train_loss": 0.0896153673529625, "global_step": 191698, "epoch": 2153, "lr": 1.9895952172997412e-05}
{"train_loss": 0.0912877768278122, "global_step": 191699, "epoch": 2153, "lr": 1.989548929299975e-05}
{"train_loss": 0.0689350888133049, "global_step": 191700, "epoch": 2153, "lr": 1.989502641704926e-05}
{"train_loss": 0.05582672730088234, "global_step": 191701, "epoch": 2153, "lr": 1.9894563545146034e-05}
{"train_loss": 0.1007719412446022, "global_step": 191702, "epoch": 2153, "lr": 1.9894100677290096e-05}
{"train_loss": 0.031074784696102142, "global_step": 191703, "epoch": 2153, "lr": 1.989363781348155e-05}
{"train_loss": 0.0752599909901619, "global_step": 191704, "epoch": 2153, "lr": 1.9893174953720416e-05}
{"train_loss": 0.07166693116841691, "global_step": 191705, "epoch": 2153, "lr": 1.98927120980068e-05, "val_loss": 7.830090045928955}
{"train_loss": 0.059510488063097, "global_step": 191706, "epoch": 2154, "lr": 1.989224924634072e-05}
{"train_loss": 0.0482337586581707, "global_step": 191707, "epoch": 2154, "lr": 1.9891786398722284e-05}
{"train_loss": 0.09410271793603897, "global_step": 191708, "epoch": 2154, "lr": 1.989132355515151e-05}
{"train_loss": 0.06841710954904556, "global_step": 191709, "epoch": 2154, "lr": 1.9890860715628495e-05}
{"train_loss": 0.07026601582765579, "global_step": 191710, "epoch": 2154, "lr": 1.989039788015327e-05}
{"train_loss": 0.06249644234776497, "global_step": 191711, "epoch": 2154, "lr": 1.9889935048725916e-05}
{"train_loss": 0.04214029759168625, "global_step": 191712, "epoch": 2154, "lr": 1.988947222134651e-05}
{"train_loss": 0.17943121492862701, "global_step": 191713, "epoch": 2154, "lr": 1.988900939801508e-05}
{"train_loss": 0.09311147779226303, "global_step": 191714, "epoch": 2154, "lr": 1.9888546578731728e-05}
{"train_loss": 0.0746801570057869, "global_step": 191715, "epoch": 2154, "lr": 1.9888083763496468e-05}
{"train_loss": 0.15738312900066376, "global_step": 191716, "epoch": 2154, "lr": 1.988762095230941e-05}
{"train_loss": 0.10172251611948013, "global_step": 191717, "epoch": 2154, "lr": 1.9887158145170575e-05}
{"train_loss": 0.045806635171175, "global_step": 191718, "epoch": 2154, "lr": 1.9886695342080063e-05}
{"train_loss": 0.044199321419000626, "global_step": 191719, "epoch": 2154, "lr": 1.98862325430379e-05}
{"train_loss": 0.03864927962422371, "global_step": 191720, "epoch": 2154, "lr": 1.9885769748044185e-05}
{"train_loss": 0.050987359136343, "global_step": 191721, "epoch": 2154, "lr": 1.9885306957098943e-05}
{"train_loss": 0.08522854745388031, "global_step": 191722, "epoch": 2154, "lr": 1.9884844170202272e-05}
{"train_loss": 0.0664980486035347, "global_step": 191723, "epoch": 2154, "lr": 1.98843813873542e-05}
{"train_loss": 0.03004997968673706, "global_step": 191724, "epoch": 2154, "lr": 1.988391860855482e-05}
{"train_loss": 0.09118854254484177, "global_step": 191725, "epoch": 2154, "lr": 1.9883455833804167e-05}
{"train_loss": 0.07470647245645523, "global_step": 191726, "epoch": 2154, "lr": 1.988299306310233e-05}
{"train_loss": 0.08300872147083282, "global_step": 191727, "epoch": 2154, "lr": 1.988253029644934e-05}
{"train_loss": 0.08646383881568909, "global_step": 191728, "epoch": 2154, "lr": 1.9882067533845294e-05}
{"train_loss": 0.10873742401599884, "global_step": 191729, "epoch": 2154, "lr": 1.9881604775290226e-05}
{"train_loss": 0.05916697531938553, "global_step": 191730, "epoch": 2154, "lr": 1.988114202078422e-05}
{"train_loss": 0.052652765065431595, "global_step": 191731, "epoch": 2154, "lr": 1.9880679270327313e-05}
{"train_loss": 0.05411142855882645, "global_step": 191732, "epoch": 2154, "lr": 1.98802165239196e-05}
{"train_loss": 0.04369725286960602, "global_step": 191733, "epoch": 2154, "lr": 1.98797537815611e-05}
{"train_loss": 0.043532274663448334, "global_step": 191734, "epoch": 2154, "lr": 1.9879291043251925e-05}
{"train_loss": 0.07077512890100479, "global_step": 191735, "epoch": 2154, "lr": 1.987882830899211e-05}
{"train_loss": 0.04130004718899727, "global_step": 191736, "epoch": 2154, "lr": 1.98783655787817e-05}
{"train_loss": 0.048485226929187775, "global_step": 191737, "epoch": 2154, "lr": 1.9877902852620793e-05}
{"train_loss": 0.04809650778770447, "global_step": 191738, "epoch": 2154, "lr": 1.9877440130509424e-05}
{"train_loss": 0.09090384840965271, "global_step": 191739, "epoch": 2154, "lr": 1.9876977412447677e-05}
{"train_loss": 0.08548647910356522, "global_step": 191740, "epoch": 2154, "lr": 1.987651469843559e-05}
{"train_loss": 0.06510905921459198, "global_step": 191741, "epoch": 2154, "lr": 1.9876051988473253e-05}
{"train_loss": 0.06441861391067505, "global_step": 191742, "epoch": 2154, "lr": 1.9875589282560697e-05}
{"train_loss": 0.05101203918457031, "global_step": 191743, "epoch": 2154, "lr": 1.9875126580698023e-05}
{"train_loss": 0.06917241960763931, "global_step": 191744, "epoch": 2154, "lr": 1.9874663882885247e-05}
{"train_loss": 0.08220039308071136, "global_step": 191745, "epoch": 2154, "lr": 1.9874201189122477e-05}
{"train_loss": 0.09556379914283752, "global_step": 191746, "epoch": 2154, "lr": 1.9873738499409734e-05}
{"train_loss": 0.04680529981851578, "global_step": 191747, "epoch": 2154, "lr": 1.98732758137471e-05}
{"train_loss": 0.08892340213060379, "global_step": 191748, "epoch": 2154, "lr": 1.987281313213465e-05}
{"train_loss": 0.03414014354348183, "global_step": 191749, "epoch": 2154, "lr": 1.9872350454572418e-05}
{"train_loss": 0.06211482360959053, "global_step": 191750, "epoch": 2154, "lr": 1.9871887781060505e-05}
{"train_loss": 0.026849428191781044, "global_step": 191751, "epoch": 2154, "lr": 1.9871425111598924e-05}
{"train_loss": 0.0743744894862175, "global_step": 191752, "epoch": 2154, "lr": 1.9870962446187784e-05}
{"train_loss": 0.07490099221467972, "global_step": 191753, "epoch": 2154, "lr": 1.9870499784827108e-05}
{"train_loss": 0.06586422026157379, "global_step": 191754, "epoch": 2154, "lr": 1.9870037127516994e-05}
{"train_loss": 0.049491167068481445, "global_step": 191755, "epoch": 2154, "lr": 1.986957447425747e-05}
{"train_loss": 0.05067228898406029, "global_step": 191756, "epoch": 2154, "lr": 1.986911182504863e-05}
{"train_loss": 0.05739274248480797, "global_step": 191757, "epoch": 2154, "lr": 1.98686491798905e-05}
{"train_loss": 0.052993036806583405, "global_step": 191758, "epoch": 2154, "lr": 1.9868186538783186e-05}
{"train_loss": 0.08390824496746063, "global_step": 191759, "epoch": 2154, "lr": 1.986772390172671e-05}
{"train_loss": 0.06495314836502075, "global_step": 191760, "epoch": 2154, "lr": 1.9867261268721167e-05}
{"train_loss": 0.060430098325014114, "global_step": 191761, "epoch": 2154, "lr": 1.9866798639766583e-05}
{"train_loss": 0.05280200019478798, "global_step": 191762, "epoch": 2154, "lr": 1.986633601486306e-05}
{"train_loss": 0.09033630043268204, "global_step": 191763, "epoch": 2154, "lr": 1.9865873394010625e-05}
{"train_loss": 0.0605514794588089, "global_step": 191764, "epoch": 2154, "lr": 1.986541077720937e-05}
{"train_loss": 0.08790674060583115, "global_step": 191765, "epoch": 2154, "lr": 1.986494816445933e-05}
{"train_loss": 0.06257624924182892, "global_step": 191766, "epoch": 2154, "lr": 1.9864485555760575e-05}
{"train_loss": 0.06915692985057831, "global_step": 191767, "epoch": 2154, "lr": 1.9864022951113196e-05}
{"train_loss": 0.1148921474814415, "global_step": 191768, "epoch": 2154, "lr": 1.986356035051723e-05}
{"train_loss": 0.05690615624189377, "global_step": 191769, "epoch": 2154, "lr": 1.9863097753972722e-05}
{"train_loss": 0.04210603982210159, "global_step": 191770, "epoch": 2154, "lr": 1.986263516147977e-05}
{"train_loss": 0.07851041108369827, "global_step": 191771, "epoch": 2154, "lr": 1.9862172573038407e-05}
{"train_loss": 0.07537592202425003, "global_step": 191772, "epoch": 2154, "lr": 1.986170998864872e-05}
{"train_loss": 0.053878817707300186, "global_step": 191773, "epoch": 2154, "lr": 1.9861247408310756e-05}
{"train_loss": 0.06488918513059616, "global_step": 191774, "epoch": 2154, "lr": 1.986078483202457e-05}
{"train_loss": 0.07982206344604492, "global_step": 191775, "epoch": 2154, "lr": 1.9860322259790247e-05}
{"train_loss": 0.07437781989574432, "global_step": 191776, "epoch": 2154, "lr": 1.985985969160782e-05}
{"train_loss": 0.05226846784353256, "global_step": 191777, "epoch": 2154, "lr": 1.9859397127477385e-05}
{"train_loss": 0.06875711679458618, "global_step": 191778, "epoch": 2154, "lr": 1.985893456739897e-05}
{"train_loss": 0.12473312020301819, "global_step": 191779, "epoch": 2154, "lr": 1.985847201137267e-05}
{"train_loss": 0.0825498178601265, "global_step": 191780, "epoch": 2154, "lr": 1.9858009459398514e-05}
{"train_loss": 0.047863949090242386, "global_step": 191781, "epoch": 2154, "lr": 1.98575469114766e-05}
{"train_loss": 0.04815800115466118, "global_step": 191782, "epoch": 2154, "lr": 1.9857084367606956e-05}
{"train_loss": 0.07319959253072739, "global_step": 191783, "epoch": 2154, "lr": 1.9856621827789653e-05}
{"train_loss": 0.054229944944381714, "global_step": 191784, "epoch": 2154, "lr": 1.985615929202478e-05}
{"train_loss": 0.05338773503899574, "global_step": 191785, "epoch": 2154, "lr": 1.9855696760312365e-05}
{"train_loss": 0.042145468294620514, "global_step": 191786, "epoch": 2154, "lr": 1.98552342326525e-05}
{"train_loss": 0.10845585912466049, "global_step": 191787, "epoch": 2154, "lr": 1.985477170904521e-05}
{"train_loss": 0.06563498824834824, "global_step": 191788, "epoch": 2154, "lr": 1.98543091894906e-05}
{"train_loss": 0.05690901726484299, "global_step": 191789, "epoch": 2154, "lr": 1.9853846673988692e-05}
{"train_loss": 0.045396532863378525, "global_step": 191790, "epoch": 2154, "lr": 1.9853384162539583e-05}
{"train_loss": 0.0806725025177002, "global_step": 191791, "epoch": 2154, "lr": 1.9852921655143303e-05}
{"train_loss": 0.08194409310817719, "global_step": 191792, "epoch": 2154, "lr": 1.9852459151799947e-05}
{"train_loss": 0.11232546716928482, "global_step": 191793, "epoch": 2154, "lr": 1.9851996652509545e-05}
{"train_loss": 0.06875084376150972, "global_step": 191794, "epoch": 2154, "lr": 1.985153415727219e-05, "val_loss": 7.9895405769348145}
{"train_loss": 0.04954146593809128, "global_step": 191795, "epoch": 2155, "lr": 1.985107166608791e-05}
{"train_loss": 0.11618568003177643, "global_step": 191796, "epoch": 2155, "lr": 1.985060917895681e-05}
{"train_loss": 0.06834366917610168, "global_step": 191797, "epoch": 2155, "lr": 1.9850146695878902e-05}
{"train_loss": 0.12422984838485718, "global_step": 191798, "epoch": 2155, "lr": 1.98496842168543e-05}
{"train_loss": 0.059875283390283585, "global_step": 191799, "epoch": 2155, "lr": 1.9849221741883016e-05}
{"train_loss": 0.08096212893724442, "global_step": 191800, "epoch": 2155, "lr": 1.9848759270965155e-05}
{"train_loss": 0.07315241545438766, "global_step": 191801, "epoch": 2155, "lr": 1.9848296804100747e-05}
{"train_loss": 0.11028438061475754, "global_step": 191802, "epoch": 2155, "lr": 1.9847834341289866e-05}
{"train_loss": 0.05969921499490738, "global_step": 191803, "epoch": 2155, "lr": 1.9847371882532596e-05}
{"train_loss": 0.09354139864444733, "global_step": 191804, "epoch": 2155, "lr": 1.9846909427828958e-05}
{"train_loss": 0.060929425060749054, "global_step": 191805, "epoch": 2155, "lr": 1.9846446977179058e-05}
{"train_loss": 0.08004314452409744, "global_step": 191806, "epoch": 2155, "lr": 1.9845984530582928e-05}
{"train_loss": 0.04172647371888161, "global_step": 191807, "epoch": 2155, "lr": 1.9845522088040623e-05}
{"train_loss": 0.05027257278561592, "global_step": 191808, "epoch": 2155, "lr": 1.9845059649552246e-05}
{"train_loss": 0.07534967362880707, "global_step": 191809, "epoch": 2155, "lr": 1.9844597215117805e-05}
{"train_loss": 0.0620703287422657, "global_step": 191810, "epoch": 2155, "lr": 1.9844134784737413e-05}
{"train_loss": 0.07198493927717209, "global_step": 191811, "epoch": 2155, "lr": 1.9843672358411107e-05}
{"train_loss": 0.14633165299892426, "global_step": 191812, "epoch": 2155, "lr": 1.984320993613894e-05}
{"train_loss": 0.10333789139986038, "global_step": 191813, "epoch": 2155, "lr": 1.9842747517920997e-05}
{"train_loss": 0.03979992866516113, "global_step": 191814, "epoch": 2155, "lr": 1.9842285103757315e-05}
{"train_loss": 0.04204969108104706, "global_step": 191815, "epoch": 2155, "lr": 1.9841822693647983e-05}
{"train_loss": 0.04332831874489784, "global_step": 191816, "epoch": 2155, "lr": 1.9841360287593036e-05}
{"train_loss": 0.0644548311829567, "global_step": 191817, "epoch": 2155, "lr": 1.9840897885592567e-05}
{"train_loss": 0.022717352956533432, "global_step": 191818, "epoch": 2155, "lr": 1.984043548764661e-05}
{"train_loss": 0.07080338895320892, "global_step": 191819, "epoch": 2155, "lr": 1.9839973093755234e-05}
{"train_loss": 0.059506144374608994, "global_step": 191820, "epoch": 2155, "lr": 1.9839510703918524e-05}
{"train_loss": 0.05221002921462059, "global_step": 191821, "epoch": 2155, "lr": 1.9839048318136504e-05}
{"train_loss": 0.0618898868560791, "global_step": 191822, "epoch": 2155, "lr": 1.9838585936409278e-05}
{"train_loss": 0.05706581845879555, "global_step": 191823, "epoch": 2155, "lr": 1.9838123558736866e-05}
{"train_loss": 0.07164771854877472, "global_step": 191824, "epoch": 2155, "lr": 1.983766118511937e-05}
{"train_loss": 0.051013365387916565, "global_step": 191825, "epoch": 2155, "lr": 1.983719881555681e-05}
{"train_loss": 0.05459350347518921, "global_step": 191826, "epoch": 2155, "lr": 1.9836736450049297e-05}
{"train_loss": 0.07404907047748566, "global_step": 191827, "epoch": 2155, "lr": 1.9836274088596845e-05}
{"train_loss": 0.0628063976764679, "global_step": 191828, "epoch": 2155, "lr": 1.983581173119956e-05}
{"train_loss": 0.09089343249797821, "global_step": 191829, "epoch": 2155, "lr": 1.983534937785746e-05}
{"train_loss": 0.06399467587471008, "global_step": 191830, "epoch": 2155, "lr": 1.9834887028570648e-05}
{"train_loss": 0.12063106149435043, "global_step": 191831, "epoch": 2155, "lr": 1.9834424683339155e-05}
{"train_loss": 0.015034563839435577, "global_step": 191832, "epoch": 2155, "lr": 1.9833962342163065e-05}
{"train_loss": 0.059025127440690994, "global_step": 191833, "epoch": 2155, "lr": 1.983350000504242e-05}
{"train_loss": 0.02110617235302925, "global_step": 191834, "epoch": 2155, "lr": 1.9833037671977305e-05}
{"train_loss": 0.062497202306985855, "global_step": 191835, "epoch": 2155, "lr": 1.983257534296776e-05}
{"train_loss": 0.09712854772806168, "global_step": 191836, "epoch": 2155, "lr": 1.983211301801387e-05}
{"train_loss": 0.09185461699962616, "global_step": 191837, "epoch": 2155, "lr": 1.9831650697115673e-05}
{"train_loss": 0.04546321928501129, "global_step": 191838, "epoch": 2155, "lr": 1.9831188380273237e-05}
{"train_loss": 0.07224006950855255, "global_step": 191839, "epoch": 2155, "lr": 1.983072606748665e-05}
{"train_loss": 0.055908046662807465, "global_step": 191840, "epoch": 2155, "lr": 1.9830263758755933e-05}
{"train_loss": 0.09927333891391754, "global_step": 191841, "epoch": 2155, "lr": 1.9829801454081194e-05}
{"train_loss": 0.04698696732521057, "global_step": 191842, "epoch": 2155, "lr": 1.9829339153462444e-05}
{"train_loss": 0.038885556161403656, "global_step": 191843, "epoch": 2155, "lr": 1.9828876856899797e-05}
{"train_loss": 0.06450004130601883, "global_step": 191844, "epoch": 2155, "lr": 1.982841456439329e-05}
{"train_loss": 0.08158645778894424, "global_step": 191845, "epoch": 2155, "lr": 1.9827952275942964e-05}
{"train_loss": 0.06574837118387222, "global_step": 191846, "epoch": 2155, "lr": 1.9827489991548914e-05}
{"train_loss": 0.0441712960600853, "global_step": 191847, "epoch": 2155, "lr": 1.9827027711211183e-05}
{"train_loss": 0.07219318300485611, "global_step": 191848, "epoch": 2155, "lr": 1.982656543492985e-05}
{"train_loss": 0.06149721145629883, "global_step": 191849, "epoch": 2155, "lr": 1.9826103162704958e-05}
{"train_loss": 0.03965162858366966, "global_step": 191850, "epoch": 2155, "lr": 1.982564089453659e-05}
{"train_loss": 0.06135812774300575, "global_step": 191851, "epoch": 2155, "lr": 1.9825178630424794e-05}
{"train_loss": 0.051423244178295135, "global_step": 191852, "epoch": 2155, "lr": 1.9824716370369622e-05}
{"train_loss": 0.01643986627459526, "global_step": 191853, "epoch": 2155, "lr": 1.9824254114371166e-05}
{"train_loss": 0.09437067061662674, "global_step": 191854, "epoch": 2155, "lr": 1.982379186242946e-05}
{"train_loss": 0.10350356251001358, "global_step": 191855, "epoch": 2155, "lr": 1.9823329614544568e-05}
{"train_loss": 0.10181710869073868, "global_step": 191856, "epoch": 2155, "lr": 1.982286737071658e-05}
{"train_loss": 0.1185615211725235, "global_step": 191857, "epoch": 2155, "lr": 1.9822405130945526e-05}
{"train_loss": 0.04064227640628815, "global_step": 191858, "epoch": 2155, "lr": 1.98219428952315e-05}
{"train_loss": 0.07470311224460602, "global_step": 191859, "epoch": 2155, "lr": 1.9821480663574527e-05}
{"train_loss": 0.07955113053321838, "global_step": 191860, "epoch": 2155, "lr": 1.9821018435974703e-05}
{"train_loss": 0.052741508930921555, "global_step": 191861, "epoch": 2155, "lr": 1.9820556212432063e-05}
{"train_loss": 0.051171209663152695, "global_step": 191862, "epoch": 2155, "lr": 1.9820093992946693e-05}
{"train_loss": 0.039921898394823074, "global_step": 191863, "epoch": 2155, "lr": 1.9819631777518628e-05}
{"train_loss": 0.043443020433187485, "global_step": 191864, "epoch": 2155, "lr": 1.9819169566147965e-05}
{"train_loss": 0.066144198179245, "global_step": 191865, "epoch": 2155, "lr": 1.9818707358834727e-05}
{"train_loss": 0.049980346113443375, "global_step": 191866, "epoch": 2155, "lr": 1.9818245155579013e-05}
{"train_loss": 0.062241941690444946, "global_step": 191867, "epoch": 2155, "lr": 1.9817782956380855e-05}
{"train_loss": 0.15984615683555603, "global_step": 191868, "epoch": 2155, "lr": 1.9817320761240343e-05}
{"train_loss": 0.09597375243902206, "global_step": 191869, "epoch": 2155, "lr": 1.9816858570157503e-05}
{"train_loss": 0.10919136554002762, "global_step": 191870, "epoch": 2155, "lr": 1.981639638313244e-05}
{"train_loss": 0.05068563297390938, "global_step": 191871, "epoch": 2155, "lr": 1.9815934200165182e-05}
{"train_loss": 0.08497007936239243, "global_step": 191872, "epoch": 2155, "lr": 1.9815472021255814e-05}
{"train_loss": 0.08173149824142456, "global_step": 191873, "epoch": 2155, "lr": 1.981500984640437e-05}
{"train_loss": 0.052027665078639984, "global_step": 191874, "epoch": 2155, "lr": 1.9814547675610934e-05}
{"train_loss": 0.07815844565629959, "global_step": 191875, "epoch": 2155, "lr": 1.9814085508875578e-05}
{"train_loss": 0.0578702948987484, "global_step": 191876, "epoch": 2155, "lr": 1.9813623346198333e-05}
{"train_loss": 0.05373319983482361, "global_step": 191877, "epoch": 2155, "lr": 1.9813161187579304e-05}
{"train_loss": 0.08257367461919785, "global_step": 191878, "epoch": 2155, "lr": 1.98126990330185e-05}
{"train_loss": 0.07770944386720657, "global_step": 191879, "epoch": 2155, "lr": 1.9812236882516033e-05}
{"train_loss": 0.05710979923605919, "global_step": 191880, "epoch": 2155, "lr": 1.9811774736071924e-05}
{"train_loss": 0.04626816138625145, "global_step": 191881, "epoch": 2155, "lr": 1.981131259368627e-05}
{"train_loss": 0.031050512567162514, "global_step": 191882, "epoch": 2155, "lr": 1.9810850455359102e-05}
{"train_loss": 0.06760526969610306, "global_step": 191883, "epoch": 2155, "lr": 1.981038832109052e-05, "val_loss": 7.9642486572265625, "train_action_mse_error": 20.22623634338379}
{"train_loss": 0.04835042729973793, "global_step": 191884, "epoch": 2156, "lr": 1.9809926190880558e-05}
{"train_loss": 0.03027217835187912, "global_step": 191885, "epoch": 2156, "lr": 1.980946406472926e-05}
{"train_loss": 0.050044070929288864, "global_step": 191886, "epoch": 2156, "lr": 1.9809001942636735e-05}
{"train_loss": 0.05117907375097275, "global_step": 191887, "epoch": 2156, "lr": 1.9808539824603005e-05}
{"train_loss": 0.07511073350906372, "global_step": 191888, "epoch": 2156, "lr": 1.9808077710628166e-05}
{"train_loss": 0.04713581129908562, "global_step": 191889, "epoch": 2156, "lr": 1.980761560071226e-05}
{"train_loss": 0.18809737265110016, "global_step": 191890, "epoch": 2156, "lr": 1.980715349485534e-05}
{"train_loss": 0.12748369574546814, "global_step": 191891, "epoch": 2156, "lr": 1.980669139305749e-05}
{"train_loss": 0.030267061665654182, "global_step": 191892, "epoch": 2156, "lr": 1.980622929531875e-05}
{"train_loss": 0.05210509151220322, "global_step": 191893, "epoch": 2156, "lr": 1.9805767201639197e-05}
{"train_loss": 0.03536494821310043, "global_step": 191894, "epoch": 2156, "lr": 1.9805305112018903e-05}
{"train_loss": 0.06042579561471939, "global_step": 191895, "epoch": 2156, "lr": 1.9804843026457902e-05}
{"train_loss": 0.080869659781456, "global_step": 191896, "epoch": 2156, "lr": 1.980438094495629e-05}
{"train_loss": 0.057558659464120865, "global_step": 191897, "epoch": 2156, "lr": 1.9803918867514093e-05}
{"train_loss": 0.061874061822891235, "global_step": 191898, "epoch": 2156, "lr": 1.9803456794131407e-05}
{"train_loss": 0.0936729684472084, "global_step": 191899, "epoch": 2156, "lr": 1.980299472480826e-05}
{"train_loss": 0.05077488347887993, "global_step": 191900, "epoch": 2156, "lr": 1.9802532659544752e-05}
{"train_loss": 0.09427819401025772, "global_step": 191901, "epoch": 2156, "lr": 1.9802070598340912e-05}
{"train_loss": 0.07425214350223541, "global_step": 191902, "epoch": 2156, "lr": 1.9801608541196827e-05}
{"train_loss": 0.06057919189333916, "global_step": 191903, "epoch": 2156, "lr": 1.9801146488112533e-05}
{"train_loss": 0.06731398403644562, "global_step": 191904, "epoch": 2156, "lr": 1.980068443908812e-05}
{"train_loss": 0.030876366421580315, "global_step": 191905, "epoch": 2156, "lr": 1.9800222394123625e-05}
{"train_loss": 0.07034100592136383, "global_step": 191906, "epoch": 2156, "lr": 1.9799760353219136e-05}
{"train_loss": 0.047392066568136215, "global_step": 191907, "epoch": 2156, "lr": 1.9799298316374683e-05}
{"train_loss": 0.09636092931032181, "global_step": 191908, "epoch": 2156, "lr": 1.9798836283590367e-05}
{"train_loss": 0.08321459591388702, "global_step": 191909, "epoch": 2156, "lr": 1.979837425486621e-05}
{"train_loss": 0.04773212969303131, "global_step": 191910, "epoch": 2156, "lr": 1.9797912230202293e-05}
{"train_loss": 0.025888821110129356, "global_step": 191911, "epoch": 2156, "lr": 1.97974502095987e-05}
{"train_loss": 0.06995449960231781, "global_step": 191912, "epoch": 2156, "lr": 1.979698819305545e-05}
{"train_loss": 0.08597347885370255, "global_step": 191913, "epoch": 2156, "lr": 1.9796526180572646e-05}
{"train_loss": 0.08361229300498962, "global_step": 191914, "epoch": 2156, "lr": 1.979606417215031e-05}
{"train_loss": 0.050734490156173706, "global_step": 191915, "epoch": 2156, "lr": 1.9795602167788546e-05}
{"train_loss": 0.07379937916994095, "global_step": 191916, "epoch": 2156, "lr": 1.9795140167487374e-05}
{"train_loss": 0.06187163293361664, "global_step": 191917, "epoch": 2156, "lr": 1.97946781712469e-05}
{"train_loss": 0.060931622982025146, "global_step": 191918, "epoch": 2156, "lr": 1.9794216179067144e-05}
{"train_loss": 0.05720573663711548, "global_step": 191919, "epoch": 2156, "lr": 1.9793754190948204e-05}
{"train_loss": 0.08683229982852936, "global_step": 191920, "epoch": 2156, "lr": 1.979329220689011e-05}
{"train_loss": 0.04542821645736694, "global_step": 191921, "epoch": 2156, "lr": 1.9792830226892957e-05}
{"train_loss": 0.060868680477142334, "global_step": 191922, "epoch": 2156, "lr": 1.979236825095679e-05}
{"train_loss": 0.0909532755613327, "global_step": 191923, "epoch": 2156, "lr": 1.979190627908165e-05}
{"train_loss": 0.09759153425693512, "global_step": 191924, "epoch": 2156, "lr": 1.979144431126764e-05}
{"train_loss": 0.04667094722390175, "global_step": 191925, "epoch": 2156, "lr": 1.9790982347514787e-05}
{"train_loss": 0.04912155121564865, "global_step": 191926, "epoch": 2156, "lr": 1.979052038782318e-05}
{"train_loss": 0.03356226533651352, "global_step": 191927, "epoch": 2156, "lr": 1.979005843219287e-05}
{"train_loss": 0.07410746812820435, "global_step": 191928, "epoch": 2156, "lr": 1.9789596480623906e-05}
{"train_loss": 0.06464677304029465, "global_step": 191929, "epoch": 2156, "lr": 1.978913453311636e-05}
{"train_loss": 0.04300137236714363, "global_step": 191930, "epoch": 2156, "lr": 1.9788672589670313e-05}
{"train_loss": 0.0629739835858345, "global_step": 191931, "epoch": 2156, "lr": 1.9788210650285793e-05}
{"train_loss": 0.054146431386470795, "global_step": 191932, "epoch": 2156, "lr": 1.97877487149629e-05}
{"train_loss": 0.06670588254928589, "global_step": 191933, "epoch": 2156, "lr": 1.9787286783701657e-05}
{"train_loss": 0.0581093430519104, "global_step": 191934, "epoch": 2156, "lr": 1.9786824856502163e-05}
{"train_loss": 0.13893480598926544, "global_step": 191935, "epoch": 2156, "lr": 1.9786362933364443e-05}
{"train_loss": 0.0550004318356514, "global_step": 191936, "epoch": 2156, "lr": 1.9785901014288598e-05}
{"train_loss": 0.04871160909533501, "global_step": 191937, "epoch": 2156, "lr": 1.978543909927465e-05}
{"train_loss": 0.08076104521751404, "global_step": 191938, "epoch": 2156, "lr": 1.97849771883227e-05}
{"train_loss": 0.033369071781635284, "global_step": 191939, "epoch": 2156, "lr": 1.9784515281432776e-05}
{"train_loss": 0.08158767223358154, "global_step": 191940, "epoch": 2156, "lr": 1.978405337860497e-05}
{"train_loss": 0.07184266299009323, "global_step": 191941, "epoch": 2156, "lr": 1.9783591479839313e-05}
{"train_loss": 0.07228802144527435, "global_step": 191942, "epoch": 2156, "lr": 1.9783129585135906e-05}
{"train_loss": 0.1276809573173523, "global_step": 191943, "epoch": 2156, "lr": 1.9782667694494767e-05}
{"train_loss": 0.10578345507383347, "global_step": 191944, "epoch": 2156, "lr": 1.9782205807916e-05}
{"train_loss": 0.05797639116644859, "global_step": 191945, "epoch": 2156, "lr": 1.9781743925399627e-05}
{"train_loss": 0.08905432373285294, "global_step": 191946, "epoch": 2156, "lr": 1.978128204694573e-05}
{"train_loss": 0.07264503091573715, "global_step": 191947, "epoch": 2156, "lr": 1.9780820172554394e-05}
{"train_loss": 0.08265077322721481, "global_step": 191948, "epoch": 2156, "lr": 1.978035830222563e-05}
{"train_loss": 0.05749335139989853, "global_step": 191949, "epoch": 2156, "lr": 1.977989643595956e-05}
{"train_loss": 0.04663606733083725, "global_step": 191950, "epoch": 2156, "lr": 1.977943457375619e-05}
{"train_loss": 0.062009576708078384, "global_step": 191951, "epoch": 2156, "lr": 1.9778972715615624e-05}
{"train_loss": 0.0942130759358406, "global_step": 191952, "epoch": 2156, "lr": 1.9778510861537886e-05}
{"train_loss": 0.09651681035757065, "global_step": 191953, "epoch": 2156, "lr": 1.977804901152308e-05}
{"train_loss": 0.05988074094057083, "global_step": 191954, "epoch": 2156, "lr": 1.977758716557123e-05}
{"train_loss": 0.026074914261698723, "global_step": 191955, "epoch": 2156, "lr": 1.9777125323682432e-05}
{"train_loss": 0.0609922930598259, "global_step": 191956, "epoch": 2156, "lr": 1.9776663485856718e-05}
{"train_loss": 0.052949048578739166, "global_step": 191957, "epoch": 2156, "lr": 1.9776201652094174e-05}
{"train_loss": 0.047008030116558075, "global_step": 191958, "epoch": 2156, "lr": 1.9775739822394833e-05}
{"train_loss": 0.08737647533416748, "global_step": 191959, "epoch": 2156, "lr": 1.97752779967588e-05}
{"train_loss": 0.03142388537526131, "global_step": 191960, "epoch": 2156, "lr": 1.9774816175186088e-05}
{"train_loss": 0.11230592429637909, "global_step": 191961, "epoch": 2156, "lr": 1.9774354357676805e-05}
{"train_loss": 0.07206770032644272, "global_step": 191962, "epoch": 2156, "lr": 1.9773892544230993e-05}
{"train_loss": 0.05321644991636276, "global_step": 191963, "epoch": 2156, "lr": 1.977343073484869e-05}
{"train_loss": 0.07956249266862869, "global_step": 191964, "epoch": 2156, "lr": 1.977296892953e-05}
{"train_loss": 0.09527945518493652, "global_step": 191965, "epoch": 2156, "lr": 1.9772507128274954e-05}
{"train_loss": 0.09774554520845413, "global_step": 191966, "epoch": 2156, "lr": 1.9772045331083637e-05}
{"train_loss": 0.08799787610769272, "global_step": 191967, "epoch": 2156, "lr": 1.9771583537956086e-05}
{"train_loss": 0.09616200625896454, "global_step": 191968, "epoch": 2156, "lr": 1.977112174889239e-05}
{"train_loss": 0.08002809435129166, "global_step": 191969, "epoch": 2156, "lr": 1.9770659963892586e-05}
{"train_loss": 0.07695930451154709, "global_step": 191970, "epoch": 2156, "lr": 1.9770198182956766e-05}
{"train_loss": 0.11643404513597488, "global_step": 191971, "epoch": 2156, "lr": 1.9769736406084956e-05}
{"train_loss": 0.06931899295429166, "global_step": 191972, "epoch": 2156, "lr": 1.976927463327725e-05, "val_loss": 7.732491493225098}
{"train_loss": 0.05677089840173721, "global_step": 191973, "epoch": 2157, "lr": 1.976881286453368e-05}
{"train_loss": 0.053671613335609436, "global_step": 191974, "epoch": 2157, "lr": 1.9768351099854347e-05}
{"train_loss": 0.080229751765728, "global_step": 191975, "epoch": 2157, "lr": 1.976788933923927e-05}
{"train_loss": 0.09698373824357986, "global_step": 191976, "epoch": 2157, "lr": 1.9767427582688554e-05}
{"train_loss": 0.03260306268930435, "global_step": 191977, "epoch": 2157, "lr": 1.976696583020221e-05}
{"train_loss": 0.07669543474912643, "global_step": 191978, "epoch": 2157, "lr": 1.9766504081780356e-05}
{"train_loss": 0.05112273991107941, "global_step": 191979, "epoch": 2157, "lr": 1.9766042337423006e-05}
{"train_loss": 0.08632869273424149, "global_step": 191980, "epoch": 2157, "lr": 1.9765580597130263e-05}
{"train_loss": 0.09285977482795715, "global_step": 191981, "epoch": 2157, "lr": 1.976511886090215e-05}
{"train_loss": 0.07781060039997101, "global_step": 191982, "epoch": 2157, "lr": 1.9764657128738752e-05}
{"train_loss": 0.08784586936235428, "global_step": 191983, "epoch": 2157, "lr": 1.976419540064014e-05}
{"train_loss": 0.04864486679434776, "global_step": 191984, "epoch": 2157, "lr": 1.9763733676606343e-05}
{"train_loss": 0.06523127108812332, "global_step": 191985, "epoch": 2157, "lr": 1.976327195663747e-05}
{"train_loss": 0.1245531439781189, "global_step": 191986, "epoch": 2157, "lr": 1.976281024073353e-05}
{"train_loss": 0.09305194765329361, "global_step": 191987, "epoch": 2157, "lr": 1.9762348528894637e-05}
{"train_loss": 0.08187834173440933, "global_step": 191988, "epoch": 2157, "lr": 1.9761886821120806e-05}
{"train_loss": 0.09166460484266281, "global_step": 191989, "epoch": 2157, "lr": 1.9761425117412135e-05}
{"train_loss": 0.08950037509202957, "global_step": 191990, "epoch": 2157, "lr": 1.9760963417768662e-05}
{"train_loss": 0.07757546007633209, "global_step": 191991, "epoch": 2157, "lr": 1.9760501722190473e-05}
{"train_loss": 0.09767462313175201, "global_step": 191992, "epoch": 2157, "lr": 1.9760040030677595e-05}
{"train_loss": 0.08365710079669952, "global_step": 191993, "epoch": 2157, "lr": 1.9759578343230133e-05}
{"train_loss": 0.04186156764626503, "global_step": 191994, "epoch": 2157, "lr": 1.9759116659848108e-05}
{"train_loss": 0.030351022258400917, "global_step": 191995, "epoch": 2157, "lr": 1.975865498053162e-05}
{"train_loss": 0.06630253791809082, "global_step": 191996, "epoch": 2157, "lr": 1.9758193305280694e-05}
{"train_loss": 0.06526704877614975, "global_step": 191997, "epoch": 2157, "lr": 1.9757731634095428e-05}
{"train_loss": 0.09497340768575668, "global_step": 191998, "epoch": 2157, "lr": 1.9757269966975848e-05}
{"train_loss": 0.14249387383460999, "global_step": 191999, "epoch": 2157, "lr": 1.9756808303922054e-05}
{"train_loss": 0.11620248854160309, "global_step": 192000, "epoch": 2157, "lr": 1.9756346644934086e-05}
{"train_loss": 0.04747490957379341, "global_step": 192001, "epoch": 2157, "lr": 1.975588499001199e-05}
{"train_loss": 0.06190059334039688, "global_step": 192002, "epoch": 2157, "lr": 1.9755423339155865e-05}
{"train_loss": 0.05539168417453766, "global_step": 192003, "epoch": 2157, "lr": 1.975496169236574e-05}
{"train_loss": 0.09526719152927399, "global_step": 192004, "epoch": 2157, "lr": 1.9754500049641706e-05}
{"train_loss": 0.1417384296655655, "global_step": 192005, "epoch": 2157, "lr": 1.9754038410983793e-05}
{"train_loss": 0.10052407532930374, "global_step": 192006, "epoch": 2157, "lr": 1.97535767763921e-05}
{"train_loss": 0.07191383093595505, "global_step": 192007, "epoch": 2157, "lr": 1.9753115145866656e-05}
{"train_loss": 0.10473962128162384, "global_step": 192008, "epoch": 2157, "lr": 1.975265351940755e-05}
{"train_loss": 0.06637448072433472, "global_step": 192009, "epoch": 2157, "lr": 1.9752191897014816e-05}
{"train_loss": 0.04424368962645531, "global_step": 192010, "epoch": 2157, "lr": 1.9751730278688546e-05}
{"train_loss": 0.080311119556427, "global_step": 192011, "epoch": 2157, "lr": 1.9751268664428772e-05}
{"train_loss": 0.12063650786876678, "global_step": 192012, "epoch": 2157, "lr": 1.9750807054235587e-05}
{"train_loss": 0.04703371226787567, "global_step": 192013, "epoch": 2157, "lr": 1.975034544810902e-05}
{"train_loss": 0.06833747774362564, "global_step": 192014, "epoch": 2157, "lr": 1.974988384604917e-05}
{"train_loss": 0.06799877434968948, "global_step": 192015, "epoch": 2157, "lr": 1.974942224805606e-05}
{"train_loss": 0.09047171473503113, "global_step": 192016, "epoch": 2157, "lr": 1.974896065412979e-05}
{"train_loss": 0.043911464512348175, "global_step": 192017, "epoch": 2157, "lr": 1.9748499064270386e-05}
{"train_loss": 0.05554097890853882, "global_step": 192018, "epoch": 2157, "lr": 1.9748037478477944e-05}
{"train_loss": 0.07846453785896301, "global_step": 192019, "epoch": 2157, "lr": 1.9747575896752496e-05}
{"train_loss": 0.027834322303533554, "global_step": 192020, "epoch": 2157, "lr": 1.974711431909411e-05}
{"train_loss": 0.11544261127710342, "global_step": 192021, "epoch": 2157, "lr": 1.974665274550288e-05}
{"train_loss": 0.06757641583681107, "global_step": 192022, "epoch": 2157, "lr": 1.9746191175978823e-05}
{"train_loss": 0.047884587198495865, "global_step": 192023, "epoch": 2157, "lr": 1.9745729610522042e-05}
{"train_loss": 0.11334964632987976, "global_step": 192024, "epoch": 2157, "lr": 1.974526804913256e-05}
{"train_loss": 0.09412267804145813, "global_step": 192025, "epoch": 2157, "lr": 1.9744806491810474e-05}
{"train_loss": 0.06714710593223572, "global_step": 192026, "epoch": 2157, "lr": 1.9744344938555813e-05}
{"train_loss": 0.056694529950618744, "global_step": 192027, "epoch": 2157, "lr": 1.974388338936867e-05}
{"train_loss": 0.055841024965047836, "global_step": 192028, "epoch": 2157, "lr": 1.9743421844249084e-05}
{"train_loss": 0.03814040496945381, "global_step": 192029, "epoch": 2157, "lr": 1.974296030319714e-05}
{"train_loss": 0.13010790944099426, "global_step": 192030, "epoch": 2157, "lr": 1.974249876621287e-05}
{"train_loss": 0.12105036526918411, "global_step": 192031, "epoch": 2157, "lr": 1.9742037233296367e-05}
{"train_loss": 0.06609370559453964, "global_step": 192032, "epoch": 2157, "lr": 1.9741575704447662e-05}
{"train_loss": 0.0889018252491951, "global_step": 192033, "epoch": 2157, "lr": 1.9741114179666847e-05}
{"train_loss": 0.06780724227428436, "global_step": 192034, "epoch": 2157, "lr": 1.9740652658953958e-05}
{"train_loss": 0.12421263754367828, "global_step": 192035, "epoch": 2157, "lr": 1.9740191142309084e-05}
{"train_loss": 0.059604715555906296, "global_step": 192036, "epoch": 2157, "lr": 1.9739729629732258e-05}
{"train_loss": 0.04141883924603462, "global_step": 192037, "epoch": 2157, "lr": 1.9739268121223576e-05}
{"train_loss": 0.07920350134372711, "global_step": 192038, "epoch": 2157, "lr": 1.9738806616783074e-05}
{"train_loss": 0.0581117607653141, "global_step": 192039, "epoch": 2157, "lr": 1.9738345116410804e-05}
{"train_loss": 0.05423463135957718, "global_step": 192040, "epoch": 2157, "lr": 1.9737883620106867e-05}
{"train_loss": 0.14684060215950012, "global_step": 192041, "epoch": 2157, "lr": 1.973742212787128e-05}
{"train_loss": 0.13492707908153534, "global_step": 192042, "epoch": 2157, "lr": 1.973696063970415e-05}
{"train_loss": 0.03190422058105469, "global_step": 192043, "epoch": 2157, "lr": 1.97364991556055e-05}
{"train_loss": 0.04330616816878319, "global_step": 192044, "epoch": 2157, "lr": 1.973603767557542e-05}
{"train_loss": 0.058913614600896835, "global_step": 192045, "epoch": 2157, "lr": 1.9735576199613948e-05}
{"train_loss": 0.07919658720493317, "global_step": 192046, "epoch": 2157, "lr": 1.9735114727721172e-05}
{"train_loss": 0.03990452364087105, "global_step": 192047, "epoch": 2157, "lr": 1.973465325989713e-05}
{"train_loss": 0.11496773362159729, "global_step": 192048, "epoch": 2157, "lr": 1.973419179614191e-05}
{"train_loss": 0.0518498532474041, "global_step": 192049, "epoch": 2157, "lr": 1.9733730336455536e-05}
{"train_loss": 0.11014661937952042, "global_step": 192050, "epoch": 2157, "lr": 1.973326888083812e-05}
{"train_loss": 0.08356954902410507, "global_step": 192051, "epoch": 2157, "lr": 1.9732807429289675e-05}
{"train_loss": 0.047913484275341034, "global_step": 192052, "epoch": 2157, "lr": 1.9732345981810303e-05}
{"train_loss": 0.04913785308599472, "global_step": 192053, "epoch": 2157, "lr": 1.9731884538400026e-05}
{"train_loss": 0.08989742398262024, "global_step": 192054, "epoch": 2157, "lr": 1.973142309905895e-05}
{"train_loss": 0.07390189170837402, "global_step": 192055, "epoch": 2157, "lr": 1.9730961663787094e-05}
{"train_loss": 0.08104345202445984, "global_step": 192056, "epoch": 2157, "lr": 1.9730500232584548e-05}
{"train_loss": 0.06730315834283829, "global_step": 192057, "epoch": 2157, "lr": 1.9730038805451385e-05}
{"train_loss": 0.07403960824012756, "global_step": 192058, "epoch": 2157, "lr": 1.9729577382387625e-05}
{"train_loss": 0.05389220267534256, "global_step": 192059, "epoch": 2157, "lr": 1.9729115963393375e-05}
{"train_loss": 0.09421142935752869, "global_step": 192060, "epoch": 2157, "lr": 1.972865454846866e-05}
{"train_loss": 0.07668603946235072, "global_step": 192061, "epoch": 2157, "lr": 1.9728193137613577e-05, "val_loss": 7.651825904846191}
{"train_loss": 0.05495375394821167, "global_step": 192062, "epoch": 2158, "lr": 1.9727731730828147e-05}
{"train_loss": 0.06896769255399704, "global_step": 192063, "epoch": 2158, "lr": 1.9727270328112474e-05}
{"train_loss": 0.11853326112031937, "global_step": 192064, "epoch": 2158, "lr": 1.9726808929466584e-05}
{"train_loss": 0.09634680300951004, "global_step": 192065, "epoch": 2158, "lr": 1.9726347534890576e-05}
{"train_loss": 0.06827481091022491, "global_step": 192066, "epoch": 2158, "lr": 1.972588614438447e-05}
{"train_loss": 0.06038045138120651, "global_step": 192067, "epoch": 2158, "lr": 1.972542475794837e-05}
{"train_loss": 0.09333565831184387, "global_step": 192068, "epoch": 2158, "lr": 1.97249633755823e-05}
{"train_loss": 0.052636757493019104, "global_step": 192069, "epoch": 2158, "lr": 1.9724501997286353e-05}
{"train_loss": 0.038300078362226486, "global_step": 192070, "epoch": 2158, "lr": 1.9724040623060562e-05}
{"train_loss": 0.08113813400268555, "global_step": 192071, "epoch": 2158, "lr": 1.9723579252905022e-05}
{"train_loss": 0.11501562595367432, "global_step": 192072, "epoch": 2158, "lr": 1.9723117886819763e-05}
{"train_loss": 0.09349419176578522, "global_step": 192073, "epoch": 2158, "lr": 1.9722656524804876e-05}
{"train_loss": 0.045071668922901154, "global_step": 192074, "epoch": 2158, "lr": 1.9722195166860397e-05}
{"train_loss": 0.04118289053440094, "global_step": 192075, "epoch": 2158, "lr": 1.9721733812986415e-05}
{"train_loss": 0.048868704587221146, "global_step": 192076, "epoch": 2158, "lr": 1.9721272463182956e-05}
{"train_loss": 0.04707152768969536, "global_step": 192077, "epoch": 2158, "lr": 1.9720811117450123e-05}
{"train_loss": 0.07587914913892746, "global_step": 192078, "epoch": 2158, "lr": 1.9720349775787955e-05}
{"train_loss": 0.053712133318185806, "global_step": 192079, "epoch": 2158, "lr": 1.9719888438196498e-05}
{"train_loss": 0.08316735923290253, "global_step": 192080, "epoch": 2158, "lr": 1.9719427104675857e-05}
{"train_loss": 0.05670614540576935, "global_step": 192081, "epoch": 2158, "lr": 1.971896577522605e-05}
{"train_loss": 0.08217775821685791, "global_step": 192082, "epoch": 2158, "lr": 1.9718504449847176e-05}
{"train_loss": 0.07576757669448853, "global_step": 192083, "epoch": 2158, "lr": 1.971804312853926e-05}
{"train_loss": 0.06173945963382721, "global_step": 192084, "epoch": 2158, "lr": 1.9717581811302404e-05}
{"train_loss": 0.0820908173918724, "global_step": 192085, "epoch": 2158, "lr": 1.9717120498136633e-05}
{"train_loss": 0.11331082135438919, "global_step": 192086, "epoch": 2158, "lr": 1.971665918904204e-05}
{"train_loss": 0.027583789080381393, "global_step": 192087, "epoch": 2158, "lr": 1.9716197884018662e-05}
{"train_loss": 0.0799279659986496, "global_step": 192088, "epoch": 2158, "lr": 1.9715736583066585e-05}
{"train_loss": 0.047749113291502, "global_step": 192089, "epoch": 2158, "lr": 1.9715275286185842e-05}
{"train_loss": 0.10567323118448257, "global_step": 192090, "epoch": 2158, "lr": 1.971481399337653e-05}
{"train_loss": 0.07832580804824829, "global_step": 192091, "epoch": 2158, "lr": 1.9714352704638673e-05}
{"train_loss": 0.092323437333107, "global_step": 192092, "epoch": 2158, "lr": 1.9713891419972352e-05}
{"train_loss": 0.0681743174791336, "global_step": 192093, "epoch": 2158, "lr": 1.971343013937764e-05}
{"train_loss": 0.08985915780067444, "global_step": 192094, "epoch": 2158, "lr": 1.971296886285458e-05}
{"train_loss": 0.0688360333442688, "global_step": 192095, "epoch": 2158, "lr": 1.9712507590403263e-05}
{"train_loss": 0.10271045565605164, "global_step": 192096, "epoch": 2158, "lr": 1.97120463220237e-05}
{"train_loss": 0.12096477299928665, "global_step": 192097, "epoch": 2158, "lr": 1.971158505771601e-05}
{"train_loss": 0.154817596077919, "global_step": 192098, "epoch": 2158, "lr": 1.9711123797480207e-05}
{"train_loss": 0.07811802625656128, "global_step": 192099, "epoch": 2158, "lr": 1.9710662541316395e-05}
{"train_loss": 0.05305102840065956, "global_step": 192100, "epoch": 2158, "lr": 1.9710201289224593e-05}
{"train_loss": 0.0878216028213501, "global_step": 192101, "epoch": 2158, "lr": 1.9709740041204905e-05}
{"train_loss": 0.14529581367969513, "global_step": 192102, "epoch": 2158, "lr": 1.970927879725736e-05}
{"train_loss": 0.10042480379343033, "global_step": 192103, "epoch": 2158, "lr": 1.970881755738205e-05}
{"train_loss": 0.0775238573551178, "global_step": 192104, "epoch": 2158, "lr": 1.9708356321578998e-05}
{"train_loss": 0.03790930286049843, "global_step": 192105, "epoch": 2158, "lr": 1.9707895089848306e-05}
{"train_loss": 0.10276535898447037, "global_step": 192106, "epoch": 2158, "lr": 1.970743386219e-05}
{"train_loss": 0.11727208644151688, "global_step": 192107, "epoch": 2158, "lr": 1.9706972638604183e-05}
{"train_loss": 0.05282982066273689, "global_step": 192108, "epoch": 2158, "lr": 1.9706511419090874e-05}
{"train_loss": 0.04609757661819458, "global_step": 192109, "epoch": 2158, "lr": 1.970605020365015e-05}
{"train_loss": 0.07317955791950226, "global_step": 192110, "epoch": 2158, "lr": 1.9705588992282103e-05}
{"train_loss": 0.09348516166210175, "global_step": 192111, "epoch": 2158, "lr": 1.970512778498677e-05}
{"train_loss": 0.11660992354154587, "global_step": 192112, "epoch": 2158, "lr": 1.970466658176419e-05}
{"train_loss": 0.09979042410850525, "global_step": 192113, "epoch": 2158, "lr": 1.9704205382614473e-05}
{"train_loss": 0.08521395176649094, "global_step": 192114, "epoch": 2158, "lr": 1.9703744187537637e-05}
{"train_loss": 0.06603734940290451, "global_step": 192115, "epoch": 2158, "lr": 1.9703282996533777e-05}
{"train_loss": 0.045025765895843506, "global_step": 192116, "epoch": 2158, "lr": 1.9702821809602945e-05}
{"train_loss": 0.045249808579683304, "global_step": 192117, "epoch": 2158, "lr": 1.970236062674518e-05}
{"train_loss": 0.06714656949043274, "global_step": 192118, "epoch": 2158, "lr": 1.9701899447960587e-05}
{"train_loss": 0.078278549015522, "global_step": 192119, "epoch": 2158, "lr": 1.9701438273249178e-05}
{"train_loss": 0.0378708653151989, "global_step": 192120, "epoch": 2158, "lr": 1.9700977102611062e-05}
{"train_loss": 0.05537223070859909, "global_step": 192121, "epoch": 2158, "lr": 1.9700515936046264e-05}
{"train_loss": 0.06854136288166046, "global_step": 192122, "epoch": 2158, "lr": 1.9700054773554878e-05}
{"train_loss": 0.06552857160568237, "global_step": 192123, "epoch": 2158, "lr": 1.9699593615136937e-05}
{"train_loss": 0.05278048291802406, "global_step": 192124, "epoch": 2158, "lr": 1.969913246079253e-05}
{"train_loss": 0.10016565024852753, "global_step": 192125, "epoch": 2158, "lr": 1.9698671310521688e-05}
{"train_loss": 0.05061640962958336, "global_step": 192126, "epoch": 2158, "lr": 1.9698210164324508e-05}
{"train_loss": 0.030523091554641724, "global_step": 192127, "epoch": 2158, "lr": 1.969774902220102e-05}
{"train_loss": 0.041634175926446915, "global_step": 192128, "epoch": 2158, "lr": 1.9697287884151293e-05}
{"train_loss": 0.0512869767844677, "global_step": 192129, "epoch": 2158, "lr": 1.9696826750175416e-05}
{"train_loss": 0.06237729266285896, "global_step": 192130, "epoch": 2158, "lr": 1.9696365620273422e-05}
{"train_loss": 0.06927188485860825, "global_step": 192131, "epoch": 2158, "lr": 1.9695904494445393e-05}
{"train_loss": 0.06514662504196167, "global_step": 192132, "epoch": 2158, "lr": 1.9695443372691365e-05}
{"train_loss": 0.07837802916765213, "global_step": 192133, "epoch": 2158, "lr": 1.969498225501143e-05}
{"train_loss": 0.11356061697006226, "global_step": 192134, "epoch": 2158, "lr": 1.9694521141405618e-05}
{"train_loss": 0.049606721848249435, "global_step": 192135, "epoch": 2158, "lr": 1.969406003187403e-05}
{"train_loss": 0.03879431635141373, "global_step": 192136, "epoch": 2158, "lr": 1.9693598926416684e-05}
{"train_loss": 0.04811802878975868, "global_step": 192137, "epoch": 2158, "lr": 1.969313782503368e-05}
{"train_loss": 0.08283840119838715, "global_step": 192138, "epoch": 2158, "lr": 1.9692676727725053e-05}
{"train_loss": 0.05391908437013626, "global_step": 192139, "epoch": 2158, "lr": 1.9692215634490892e-05}
{"train_loss": 0.08901961147785187, "global_step": 192140, "epoch": 2158, "lr": 1.969175454533122e-05}
{"train_loss": 0.04541977867484093, "global_step": 192141, "epoch": 2158, "lr": 1.9691293460246146e-05}
{"train_loss": 0.07963579893112183, "global_step": 192142, "epoch": 2158, "lr": 1.969083237923569e-05}
{"train_loss": 0.054880429059267044, "global_step": 192143, "epoch": 2158, "lr": 1.9690371302299948e-05}
{"train_loss": 0.03082451969385147, "global_step": 192144, "epoch": 2158, "lr": 1.9689910229438952e-05}
{"train_loss": 0.06746864318847656, "global_step": 192145, "epoch": 2158, "lr": 1.968944916065279e-05}
{"train_loss": 0.06189679726958275, "global_step": 192146, "epoch": 2158, "lr": 1.9688988095941495e-05}
{"train_loss": 0.07919224351644516, "global_step": 192147, "epoch": 2158, "lr": 1.9688527035305153e-05}
{"train_loss": 0.0639108270406723, "global_step": 192148, "epoch": 2158, "lr": 1.968806597874383e-05}
{"train_loss": 0.0746338814496994, "global_step": 192149, "epoch": 2158, "lr": 1.968760492625758e-05}
{"train_loss": 0.07218493470984898, "global_step": 192150, "epoch": 2158, "lr": 1.968714387784644e-05, "val_loss": 7.792986869812012}
{"train_loss": 0.059385258704423904, "global_step": 192151, "epoch": 2159, "lr": 1.9686682833510523e-05}
{"train_loss": 0.09418600052595139, "global_step": 192152, "epoch": 2159, "lr": 1.968622179324983e-05}
{"train_loss": 0.061033789068460464, "global_step": 192153, "epoch": 2159, "lr": 1.9685760757064487e-05}
{"train_loss": 0.05398350954055786, "global_step": 192154, "epoch": 2159, "lr": 1.9685299724954513e-05}
{"train_loss": 0.07315578311681747, "global_step": 192155, "epoch": 2159, "lr": 1.9684838696919965e-05}
{"train_loss": 0.04696805030107498, "global_step": 192156, "epoch": 2159, "lr": 1.9684377672960942e-05}
{"train_loss": 0.019561907276511192, "global_step": 192157, "epoch": 2159, "lr": 1.9683916653077466e-05}
{"train_loss": 0.04249905049800873, "global_step": 192158, "epoch": 2159, "lr": 1.9683455637269638e-05}
{"train_loss": 0.04431464523077011, "global_step": 192159, "epoch": 2159, "lr": 1.9682994625537475e-05}
{"train_loss": 0.03733358159661293, "global_step": 192160, "epoch": 2159, "lr": 1.968253361788109e-05}
{"train_loss": 0.10120423138141632, "global_step": 192161, "epoch": 2159, "lr": 1.9682072614300497e-05}
{"train_loss": 0.0889703631401062, "global_step": 192162, "epoch": 2159, "lr": 1.9681611614795796e-05}
{"train_loss": 0.032062482088804245, "global_step": 192163, "epoch": 2159, "lr": 1.9681150619367015e-05}
{"train_loss": 0.05658711865544319, "global_step": 192164, "epoch": 2159, "lr": 1.9680689628014238e-05}
{"train_loss": 0.029294542968273163, "global_step": 192165, "epoch": 2159, "lr": 1.9680228640737536e-05}
{"train_loss": 0.057531509548425674, "global_step": 192166, "epoch": 2159, "lr": 1.9679767657536946e-05}
{"train_loss": 0.05707470700144768, "global_step": 192167, "epoch": 2159, "lr": 1.967930667841255e-05}
{"train_loss": 0.03615296259522438, "global_step": 192168, "epoch": 2159, "lr": 1.967884570336439e-05}
{"train_loss": 0.08794242143630981, "global_step": 192169, "epoch": 2159, "lr": 1.9678384732392558e-05}
{"train_loss": 0.09191267937421799, "global_step": 192170, "epoch": 2159, "lr": 1.967792376549708e-05}
{"train_loss": 0.06660572439432144, "global_step": 192171, "epoch": 2159, "lr": 1.9677462802678055e-05}
{"train_loss": 0.07610096037387848, "global_step": 192172, "epoch": 2159, "lr": 1.96770018439355e-05}
{"train_loss": 0.07969126105308533, "global_step": 192173, "epoch": 2159, "lr": 1.967654088926953e-05}
{"train_loss": 0.04286966845393181, "global_step": 192174, "epoch": 2159, "lr": 1.9676079938680152e-05}
{"train_loss": 0.0723312571644783, "global_step": 192175, "epoch": 2159, "lr": 1.9675618992167477e-05}
{"train_loss": 0.05220673978328705, "global_step": 192176, "epoch": 2159, "lr": 1.967515804973153e-05}
{"train_loss": 0.06508580595254898, "global_step": 192177, "epoch": 2159, "lr": 1.9674697111372403e-05}
{"train_loss": 0.11360733956098557, "global_step": 192178, "epoch": 2159, "lr": 1.9674236177090134e-05}
{"train_loss": 0.043616198003292084, "global_step": 192179, "epoch": 2159, "lr": 1.96737752468848e-05}
{"train_loss": 0.07661113888025284, "global_step": 192180, "epoch": 2159, "lr": 1.9673314320756447e-05}
{"train_loss": 0.06864189356565475, "global_step": 192181, "epoch": 2159, "lr": 1.967285339870516e-05}
{"train_loss": 0.05240578576922417, "global_step": 192182, "epoch": 2159, "lr": 1.9672392480730977e-05}
{"train_loss": 0.06624820828437805, "global_step": 192183, "epoch": 2159, "lr": 1.967193156683397e-05}
{"train_loss": 0.08101049065589905, "global_step": 192184, "epoch": 2159, "lr": 1.9671470657014212e-05}
{"train_loss": 0.11859464645385742, "global_step": 192185, "epoch": 2159, "lr": 1.9671009751271746e-05}
{"train_loss": 0.054607145488262177, "global_step": 192186, "epoch": 2159, "lr": 1.9670548849606656e-05}
{"train_loss": 0.03299044817686081, "global_step": 192187, "epoch": 2159, "lr": 1.9670087952018996e-05}
{"train_loss": 0.06905900686979294, "global_step": 192188, "epoch": 2159, "lr": 1.96696270585088e-05}
{"train_loss": 0.10993625223636627, "global_step": 192189, "epoch": 2159, "lr": 1.966916616907617e-05}
{"train_loss": 0.04042002186179161, "global_step": 192190, "epoch": 2159, "lr": 1.966870528372114e-05}
{"train_loss": 0.08791975677013397, "global_step": 192191, "epoch": 2159, "lr": 1.9668244402443798e-05}
{"train_loss": 0.06575462222099304, "global_step": 192192, "epoch": 2159, "lr": 1.966778352524417e-05}
{"train_loss": 0.0774168074131012, "global_step": 192193, "epoch": 2159, "lr": 1.9667322652122356e-05}
{"train_loss": 0.04782514646649361, "global_step": 192194, "epoch": 2159, "lr": 1.9666861783078405e-05}
{"train_loss": 0.05816107988357544, "global_step": 192195, "epoch": 2159, "lr": 1.9666400918112356e-05}
{"train_loss": 0.042007964104413986, "global_step": 192196, "epoch": 2159, "lr": 1.9665940057224303e-05}
{"train_loss": 0.09153541922569275, "global_step": 192197, "epoch": 2159, "lr": 1.9665479200414283e-05}
{"train_loss": 0.09180893748998642, "global_step": 192198, "epoch": 2159, "lr": 1.9665018347682384e-05}
{"train_loss": 0.08084478229284286, "global_step": 192199, "epoch": 2159, "lr": 1.9664557499028635e-05}
{"train_loss": 0.11427933722734451, "global_step": 192200, "epoch": 2159, "lr": 1.9664096654453118e-05}
{"train_loss": 0.07870251685380936, "global_step": 192201, "epoch": 2159, "lr": 1.966363581395591e-05}
{"train_loss": 0.05220580846071243, "global_step": 192202, "epoch": 2159, "lr": 1.9663174977537042e-05}
{"train_loss": 0.06257382035255432, "global_step": 192203, "epoch": 2159, "lr": 1.9662714145196605e-05}
{"train_loss": 0.07210514694452286, "global_step": 192204, "epoch": 2159, "lr": 1.966225331693463e-05}
{"train_loss": 0.05002284049987793, "global_step": 192205, "epoch": 2159, "lr": 1.9661792492751212e-05}
{"train_loss": 0.040472351014614105, "global_step": 192206, "epoch": 2159, "lr": 1.9661331672646376e-05}
{"train_loss": 0.07490437477827072, "global_step": 192207, "epoch": 2159, "lr": 1.9660870856620227e-05}
{"train_loss": 0.04781803488731384, "global_step": 192208, "epoch": 2159, "lr": 1.9660410044672785e-05}
{"train_loss": 0.04484819248318672, "global_step": 192209, "epoch": 2159, "lr": 1.965994923680415e-05}
{"train_loss": 0.0714016929268837, "global_step": 192210, "epoch": 2159, "lr": 1.9659488433014343e-05}
{"train_loss": 0.05699102580547333, "global_step": 192211, "epoch": 2159, "lr": 1.9659027633303468e-05}
{"train_loss": 0.06364721804857254, "global_step": 192212, "epoch": 2159, "lr": 1.9658566837671548e-05}
{"train_loss": 0.0757017731666565, "global_step": 192213, "epoch": 2159, "lr": 1.9658106046118685e-05}
{"train_loss": 0.07710538059473038, "global_step": 192214, "epoch": 2159, "lr": 1.96576452586449e-05}
{"train_loss": 0.04990852624177933, "global_step": 192215, "epoch": 2159, "lr": 1.9657184475250297e-05}
{"train_loss": 0.18778319656848907, "global_step": 192216, "epoch": 2159, "lr": 1.965672369593489e-05}
{"train_loss": 0.06074504181742668, "global_step": 192217, "epoch": 2159, "lr": 1.9656262920698786e-05}
{"train_loss": 0.05323980003595352, "global_step": 192218, "epoch": 2159, "lr": 1.9655802149542015e-05}
{"train_loss": 0.03424931317567825, "global_step": 192219, "epoch": 2159, "lr": 1.965534138246465e-05}
{"train_loss": 0.07474441081285477, "global_step": 192220, "epoch": 2159, "lr": 1.965488061946677e-05}
{"train_loss": 0.08722978085279465, "global_step": 192221, "epoch": 2159, "lr": 1.96544198605484e-05}
{"train_loss": 0.07498469948768616, "global_step": 192222, "epoch": 2159, "lr": 1.9653959105709647e-05}
{"train_loss": 0.03666841983795166, "global_step": 192223, "epoch": 2159, "lr": 1.9653498354950534e-05}
{"train_loss": 0.06998653709888458, "global_step": 192224, "epoch": 2159, "lr": 1.9653037608271145e-05}
{"train_loss": 0.09048499166965485, "global_step": 192225, "epoch": 2159, "lr": 1.9652576865671528e-05}
{"train_loss": 0.06401602178812027, "global_step": 192226, "epoch": 2159, "lr": 1.9652116127151766e-05}
{"train_loss": 0.04484814777970314, "global_step": 192227, "epoch": 2159, "lr": 1.965165539271191e-05}
{"train_loss": 0.07231909036636353, "global_step": 192228, "epoch": 2159, "lr": 1.9651194662351995e-05}
{"train_loss": 0.06625855714082718, "global_step": 192229, "epoch": 2159, "lr": 1.965073393607213e-05}
{"train_loss": 0.046716686338186264, "global_step": 192230, "epoch": 2159, "lr": 1.9650273213872333e-05}
{"train_loss": 0.07714100182056427, "global_step": 192231, "epoch": 2159, "lr": 1.9649812495752706e-05}
{"train_loss": 0.05375091731548309, "global_step": 192232, "epoch": 2159, "lr": 1.9649351781713295e-05}
{"train_loss": 0.14940792322158813, "global_step": 192233, "epoch": 2159, "lr": 1.964889107175414e-05}
{"train_loss": 0.014122311025857925, "global_step": 192234, "epoch": 2159, "lr": 1.9648430365875338e-05}
{"train_loss": 0.0666804313659668, "global_step": 192235, "epoch": 2159, "lr": 1.9647969664076915e-05}
{"train_loss": 0.04451897367835045, "global_step": 192236, "epoch": 2159, "lr": 1.964750896635895e-05}
{"train_loss": 0.049889110028743744, "global_step": 192237, "epoch": 2159, "lr": 1.964704827272153e-05}
{"train_loss": 0.07179596275091171, "global_step": 192238, "epoch": 2159, "lr": 1.9646587583164678e-05}
{"train_loss": 0.06580341019238649, "global_step": 192239, "epoch": 2159, "lr": 1.9646126897688488e-05, "val_loss": 8.046575546264648}
{"train_loss": 0.059451375156641006, "global_step": 192240, "epoch": 2160, "lr": 1.964566621629299e-05}
{"train_loss": 0.056033845990896225, "global_step": 192241, "epoch": 2160, "lr": 1.964520553897828e-05}
{"train_loss": 0.04643847048282623, "global_step": 192242, "epoch": 2160, "lr": 1.964474486574438e-05}
{"train_loss": 0.02711333893239498, "global_step": 192243, "epoch": 2160, "lr": 1.9644284196591394e-05}
{"train_loss": 0.08314752578735352, "global_step": 192244, "epoch": 2160, "lr": 1.9643823531519344e-05}
{"train_loss": 0.14431358873844147, "global_step": 192245, "epoch": 2160, "lr": 1.9643362870528335e-05}
{"train_loss": 0.09863732755184174, "global_step": 192246, "epoch": 2160, "lr": 1.9642902213618386e-05}
{"train_loss": 0.05463211238384247, "global_step": 192247, "epoch": 2160, "lr": 1.96424415607896e-05}
{"train_loss": 0.05651126056909561, "global_step": 192248, "epoch": 2160, "lr": 1.9641980912041992e-05}
{"train_loss": 0.09705700725317001, "global_step": 192249, "epoch": 2160, "lr": 1.9641520267375673e-05}
{"train_loss": 0.027044346556067467, "global_step": 192250, "epoch": 2160, "lr": 1.9641059626790658e-05}
{"train_loss": 0.036479584872722626, "global_step": 192251, "epoch": 2160, "lr": 1.9640598990287056e-05}
{"train_loss": 0.06217040494084358, "global_step": 192252, "epoch": 2160, "lr": 1.964013835786489e-05}
{"train_loss": 0.036463722586631775, "global_step": 192253, "epoch": 2160, "lr": 1.9639677729524248e-05}
{"train_loss": 0.04247065633535385, "global_step": 192254, "epoch": 2160, "lr": 1.963921710526517e-05}
{"train_loss": 0.07976005971431732, "global_step": 192255, "epoch": 2160, "lr": 1.9638756485087728e-05}
{"train_loss": 0.07938764244318008, "global_step": 192256, "epoch": 2160, "lr": 1.9638295868991996e-05}
{"train_loss": 0.06330609321594238, "global_step": 192257, "epoch": 2160, "lr": 1.9637835256978015e-05}
{"train_loss": 0.06437889486551285, "global_step": 192258, "epoch": 2160, "lr": 1.963737464904587e-05}
{"train_loss": 0.059159111231565475, "global_step": 192259, "epoch": 2160, "lr": 1.9636914045195598e-05}
{"train_loss": 0.06987182050943375, "global_step": 192260, "epoch": 2160, "lr": 1.9636453445427284e-05}
{"train_loss": 0.0392165333032608, "global_step": 192261, "epoch": 2160, "lr": 1.9635992849740964e-05}
{"train_loss": 0.040914155542850494, "global_step": 192262, "epoch": 2160, "lr": 1.9635532258136736e-05}
{"train_loss": 0.108435720205307, "global_step": 192263, "epoch": 2160, "lr": 1.9635071670614617e-05}
{"train_loss": 0.09095398336648941, "global_step": 192264, "epoch": 2160, "lr": 1.9634611087174713e-05}
{"train_loss": 0.05642121285200119, "global_step": 192265, "epoch": 2160, "lr": 1.963415050781706e-05}
{"train_loss": 0.059869080781936646, "global_step": 192266, "epoch": 2160, "lr": 1.9633689932541714e-05}
{"train_loss": 0.04969248175621033, "global_step": 192267, "epoch": 2160, "lr": 1.9633229361348765e-05}
{"train_loss": 0.07572898268699646, "global_step": 192268, "epoch": 2160, "lr": 1.9632768794238243e-05}
{"train_loss": 0.057529598474502563, "global_step": 192269, "epoch": 2160, "lr": 1.963230823121024e-05}
{"train_loss": 0.0628710389137268, "global_step": 192270, "epoch": 2160, "lr": 1.96318476722648e-05}
{"train_loss": 0.11930499225854874, "global_step": 192271, "epoch": 2160, "lr": 1.9631387117401973e-05}
{"train_loss": 0.049482014030218124, "global_step": 192272, "epoch": 2160, "lr": 1.9630926566621853e-05}
{"train_loss": 0.02442845329642296, "global_step": 192273, "epoch": 2160, "lr": 1.963046601992447e-05}
{"train_loss": 0.06295701861381531, "global_step": 192274, "epoch": 2160, "lr": 1.96300054773099e-05}
{"train_loss": 0.056593332439661026, "global_step": 192275, "epoch": 2160, "lr": 1.9629544938778223e-05}
{"train_loss": 0.049688518047332764, "global_step": 192276, "epoch": 2160, "lr": 1.962908440432946e-05}
{"train_loss": 0.05194477364420891, "global_step": 192277, "epoch": 2160, "lr": 1.962862387396372e-05}
{"train_loss": 0.10147259384393692, "global_step": 192278, "epoch": 2160, "lr": 1.9628163347681023e-05}
{"train_loss": 0.061176858842372894, "global_step": 192279, "epoch": 2160, "lr": 1.9627702825481465e-05}
{"train_loss": 0.1464119553565979, "global_step": 192280, "epoch": 2160, "lr": 1.962724230736508e-05}
{"train_loss": 0.05732366815209389, "global_step": 192281, "epoch": 2160, "lr": 1.962678179333195e-05}
{"train_loss": 0.10835014283657074, "global_step": 192282, "epoch": 2160, "lr": 1.962632128338212e-05}
{"train_loss": 0.09810620546340942, "global_step": 192283, "epoch": 2160, "lr": 1.9625860777515674e-05}
{"train_loss": 0.12708771228790283, "global_step": 192284, "epoch": 2160, "lr": 1.9625400275732647e-05}
{"train_loss": 0.06778370589017868, "global_step": 192285, "epoch": 2160, "lr": 1.962493977803313e-05}
{"train_loss": 0.057345788925886154, "global_step": 192286, "epoch": 2160, "lr": 1.962447928441715e-05}
{"train_loss": 0.050490014255046844, "global_step": 192287, "epoch": 2160, "lr": 1.9624018794884808e-05}
{"train_loss": 0.03030185028910637, "global_step": 192288, "epoch": 2160, "lr": 1.962355830943613e-05}
{"train_loss": 0.0726822093129158, "global_step": 192289, "epoch": 2160, "lr": 1.9623097828071207e-05}
{"train_loss": 0.051795125007629395, "global_step": 192290, "epoch": 2160, "lr": 1.962263735079008e-05}
{"train_loss": 0.12998852133750916, "global_step": 192291, "epoch": 2160, "lr": 1.9622176877592813e-05}
{"train_loss": 0.07058002799749374, "global_step": 192292, "epoch": 2160, "lr": 1.962171640847949e-05}
{"train_loss": 0.06393956393003464, "global_step": 192293, "epoch": 2160, "lr": 1.962125594345014e-05}
{"train_loss": 0.04027196019887924, "global_step": 192294, "epoch": 2160, "lr": 1.962079548250486e-05}
{"train_loss": 0.0755450576543808, "global_step": 192295, "epoch": 2160, "lr": 1.9620335025643684e-05}
{"train_loss": 0.06047024205327034, "global_step": 192296, "epoch": 2160, "lr": 1.9619874572866692e-05}
{"train_loss": 0.09273356944322586, "global_step": 192297, "epoch": 2160, "lr": 1.9619414124173924e-05}
{"train_loss": 0.06883955746889114, "global_step": 192298, "epoch": 2160, "lr": 1.961895367956547e-05}
{"train_loss": 0.08181099593639374, "global_step": 192299, "epoch": 2160, "lr": 1.9618493239041357e-05}
{"train_loss": 0.05001156032085419, "global_step": 192300, "epoch": 2160, "lr": 1.9618032802601693e-05}
{"train_loss": 0.055379752069711685, "global_step": 192301, "epoch": 2160, "lr": 1.9617572370246494e-05}
{"train_loss": 0.05778365954756737, "global_step": 192302, "epoch": 2160, "lr": 1.9617111941975856e-05}
{"train_loss": 0.05254639685153961, "global_step": 192303, "epoch": 2160, "lr": 1.9616651517789815e-05}
{"train_loss": 0.0713537260890007, "global_step": 192304, "epoch": 2160, "lr": 1.9616191097688458e-05}
{"train_loss": 0.07058215141296387, "global_step": 192305, "epoch": 2160, "lr": 1.9615730681671836e-05}
{"train_loss": 0.07480154186487198, "global_step": 192306, "epoch": 2160, "lr": 1.961527026973999e-05}
{"train_loss": 0.08722247183322906, "global_step": 192307, "epoch": 2160, "lr": 1.9614809861893017e-05}
{"train_loss": 0.08758089691400528, "global_step": 192308, "epoch": 2160, "lr": 1.961434945813095e-05}
{"train_loss": 0.04437970370054245, "global_step": 192309, "epoch": 2160, "lr": 1.9613889058453878e-05}
{"train_loss": 0.06659212708473206, "global_step": 192310, "epoch": 2160, "lr": 1.9613428662861826e-05}
{"train_loss": 0.056799497455358505, "global_step": 192311, "epoch": 2160, "lr": 1.9612968271354898e-05}
{"train_loss": 0.06578358262777328, "global_step": 192312, "epoch": 2160, "lr": 1.9612507883933123e-05}
{"train_loss": 0.06032814085483551, "global_step": 192313, "epoch": 2160, "lr": 1.9612047500596594e-05}
{"train_loss": 0.05953073129057884, "global_step": 192314, "epoch": 2160, "lr": 1.961158712134533e-05}
{"train_loss": 0.07059068977832794, "global_step": 192315, "epoch": 2160, "lr": 1.9611126746179442e-05}
{"train_loss": 0.12661848962306976, "global_step": 192316, "epoch": 2160, "lr": 1.9610666375098948e-05}
{"train_loss": 0.07105684280395508, "global_step": 192317, "epoch": 2160, "lr": 1.9610206008103944e-05}
{"train_loss": 0.05489899218082428, "global_step": 192318, "epoch": 2160, "lr": 1.9609745645194462e-05}
{"train_loss": 0.06688235700130463, "global_step": 192319, "epoch": 2160, "lr": 1.96092852863706e-05}
{"train_loss": 0.0623558834195137, "global_step": 192320, "epoch": 2160, "lr": 1.9608824931632375e-05}
{"train_loss": 0.016355931758880615, "global_step": 192321, "epoch": 2160, "lr": 1.9608364580979893e-05}
{"train_loss": 0.09451992809772491, "global_step": 192322, "epoch": 2160, "lr": 1.9607904234413183e-05}
{"train_loss": 0.06302336603403091, "global_step": 192323, "epoch": 2160, "lr": 1.960744389193233e-05}
{"train_loss": 0.06733608990907669, "global_step": 192324, "epoch": 2160, "lr": 1.9606983553537377e-05}
{"train_loss": 0.07578468322753906, "global_step": 192325, "epoch": 2160, "lr": 1.9606523219228406e-05}
{"train_loss": 0.055059272795915604, "global_step": 192326, "epoch": 2160, "lr": 1.960606288900545e-05}
{"train_loss": 0.12287256866693497, "global_step": 192327, "epoch": 2160, "lr": 1.9605602562868592e-05}
{"train_loss": 0.06859985965021541, "global_step": 192328, "epoch": 2160, "lr": 1.9605142240817902e-05, "val_loss": 7.972339153289795, "train_action_mse_error": 28.181373596191406}
{"train_loss": 0.025165610015392303, "global_step": 192329, "epoch": 2161, "lr": 1.9604681922853418e-05}
{"train_loss": 0.06709296256303787, "global_step": 192330, "epoch": 2161, "lr": 1.960422160897523e-05}
{"train_loss": 0.07287608087062836, "global_step": 192331, "epoch": 2161, "lr": 1.9603761299183364e-05}
{"train_loss": 0.05696688964962959, "global_step": 192332, "epoch": 2161, "lr": 1.9603300993477923e-05}
{"train_loss": 0.04724334180355072, "global_step": 192333, "epoch": 2161, "lr": 1.960284069185893e-05}
{"train_loss": 0.055782344192266464, "global_step": 192334, "epoch": 2161, "lr": 1.960238039432648e-05}
{"train_loss": 0.03545374795794487, "global_step": 192335, "epoch": 2161, "lr": 1.9601920100880605e-05}
{"train_loss": 0.053434811532497406, "global_step": 192336, "epoch": 2161, "lr": 1.9601459811521396e-05}
{"train_loss": 0.02716323919594288, "global_step": 192337, "epoch": 2161, "lr": 1.9600999526248886e-05}
{"train_loss": 0.039259012788534164, "global_step": 192338, "epoch": 2161, "lr": 1.9600539245063166e-05}
{"train_loss": 0.043548546731472015, "global_step": 192339, "epoch": 2161, "lr": 1.9600078967964263e-05}
{"train_loss": 0.02230945974588394, "global_step": 192340, "epoch": 2161, "lr": 1.959961869495228e-05}
{"train_loss": 0.11240542680025101, "global_step": 192341, "epoch": 2161, "lr": 1.9599158426027237e-05}
{"train_loss": 0.04573674499988556, "global_step": 192342, "epoch": 2161, "lr": 1.959869816118924e-05}
{"train_loss": 0.08618266880512238, "global_step": 192343, "epoch": 2161, "lr": 1.959823790043832e-05}
{"train_loss": 0.05454127863049507, "global_step": 192344, "epoch": 2161, "lr": 1.9597777643774535e-05}
{"train_loss": 0.022401347756385803, "global_step": 192345, "epoch": 2161, "lr": 1.959731739119797e-05}
{"train_loss": 0.11447323858737946, "global_step": 192346, "epoch": 2161, "lr": 1.9596857142708664e-05}
{"train_loss": 0.0350460410118103, "global_step": 192347, "epoch": 2161, "lr": 1.9596396898306708e-05}
{"train_loss": 0.10428333282470703, "global_step": 192348, "epoch": 2161, "lr": 1.959593665799212e-05}
{"train_loss": 0.139291912317276, "global_step": 192349, "epoch": 2161, "lr": 1.9595476421765013e-05}
{"train_loss": 0.04070953652262688, "global_step": 192350, "epoch": 2161, "lr": 1.9595016189625403e-05}
{"train_loss": 0.05110563337802887, "global_step": 192351, "epoch": 2161, "lr": 1.959455596157339e-05}
{"train_loss": 0.05691174417734146, "global_step": 192352, "epoch": 2161, "lr": 1.9594095737609e-05}
{"train_loss": 0.05847184360027313, "global_step": 192353, "epoch": 2161, "lr": 1.9593635517732335e-05}
{"train_loss": 0.07629691064357758, "global_step": 192354, "epoch": 2161, "lr": 1.9593175301943413e-05}
{"train_loss": 0.05159873887896538, "global_step": 192355, "epoch": 2161, "lr": 1.9592715090242335e-05}
{"train_loss": 0.03534068539738655, "global_step": 192356, "epoch": 2161, "lr": 1.959225488262913e-05}
{"train_loss": 0.05279025435447693, "global_step": 192357, "epoch": 2161, "lr": 1.9591794679103893e-05}
{"train_loss": 0.09076902270317078, "global_step": 192358, "epoch": 2161, "lr": 1.959133447966665e-05}
{"train_loss": 0.04431844875216484, "global_step": 192359, "epoch": 2161, "lr": 1.95908742843175e-05}
{"train_loss": 0.10121956467628479, "global_step": 192360, "epoch": 2161, "lr": 1.959041409305647e-05}
{"train_loss": 0.05828095227479935, "global_step": 192361, "epoch": 2161, "lr": 1.9589953905883657e-05}
{"train_loss": 0.0646347776055336, "global_step": 192362, "epoch": 2161, "lr": 1.9589493722799086e-05}
{"train_loss": 0.029742609709501266, "global_step": 192363, "epoch": 2161, "lr": 1.958903354380283e-05}
{"train_loss": 0.060320161283016205, "global_step": 192364, "epoch": 2161, "lr": 1.9588573368894985e-05}
{"train_loss": 0.0894928053021431, "global_step": 192365, "epoch": 2161, "lr": 1.9588113198075563e-05}
{"train_loss": 0.08099615573883057, "global_step": 192366, "epoch": 2161, "lr": 1.9587653031344665e-05}
{"train_loss": 0.10324511677026749, "global_step": 192367, "epoch": 2161, "lr": 1.9587192868702326e-05}
{"train_loss": 0.04598001390695572, "global_step": 192368, "epoch": 2161, "lr": 1.9586732710148626e-05}
{"train_loss": 0.05926021188497543, "global_step": 192369, "epoch": 2161, "lr": 1.9586272555683604e-05}
{"train_loss": 0.11118824034929276, "global_step": 192370, "epoch": 2161, "lr": 1.9585812405307357e-05}
{"train_loss": 0.05558464676141739, "global_step": 192371, "epoch": 2161, "lr": 1.9585352259019913e-05}
{"train_loss": 0.07566468417644501, "global_step": 192372, "epoch": 2161, "lr": 1.958489211682136e-05}
{"train_loss": 0.06340359896421432, "global_step": 192373, "epoch": 2161, "lr": 1.958443197871173e-05}
{"train_loss": 0.06600195169448853, "global_step": 192374, "epoch": 2161, "lr": 1.9583971844691123e-05}
{"train_loss": 0.05589236691594124, "global_step": 192375, "epoch": 2161, "lr": 1.958351171475956e-05}
{"train_loss": 0.1778312772512436, "global_step": 192376, "epoch": 2161, "lr": 1.958305158891714e-05}
{"train_loss": 0.047978539019823074, "global_step": 192377, "epoch": 2161, "lr": 1.958259146716389e-05}
{"train_loss": 0.0776640996336937, "global_step": 192378, "epoch": 2161, "lr": 1.958213134949991e-05}
{"train_loss": 0.07207044959068298, "global_step": 192379, "epoch": 2161, "lr": 1.9581671235925225e-05}
{"train_loss": 0.05359521880745888, "global_step": 192380, "epoch": 2161, "lr": 1.958121112643993e-05}
{"train_loss": 0.06596959382295609, "global_step": 192381, "epoch": 2161, "lr": 1.958075102104407e-05}
{"train_loss": 0.07850831747055054, "global_step": 192382, "epoch": 2161, "lr": 1.9580290919737687e-05}
{"train_loss": 0.1006515696644783, "global_step": 192383, "epoch": 2161, "lr": 1.9579830822520885e-05}
{"train_loss": 0.10647168755531311, "global_step": 192384, "epoch": 2161, "lr": 1.9579370729393687e-05}
{"train_loss": 0.04882391542196274, "global_step": 192385, "epoch": 2161, "lr": 1.9578910640356184e-05}
{"train_loss": 0.0489267073571682, "global_step": 192386, "epoch": 2161, "lr": 1.9578450555408413e-05}
{"train_loss": 0.11881866306066513, "global_step": 192387, "epoch": 2161, "lr": 1.957799047455046e-05}
{"train_loss": 0.052041828632354736, "global_step": 192388, "epoch": 2161, "lr": 1.9577530397782368e-05}
{"train_loss": 0.050091736018657684, "global_step": 192389, "epoch": 2161, "lr": 1.957707032510422e-05}
{"train_loss": 0.07406630367040634, "global_step": 192390, "epoch": 2161, "lr": 1.9576610256516043e-05}
{"train_loss": 0.056801341474056244, "global_step": 192391, "epoch": 2161, "lr": 1.9576150192017938e-05}
{"train_loss": 0.05296942964196205, "global_step": 192392, "epoch": 2161, "lr": 1.9575690131609932e-05}
{"train_loss": 0.05595822259783745, "global_step": 192393, "epoch": 2161, "lr": 1.9575230075292122e-05}
{"train_loss": 0.053118184208869934, "global_step": 192394, "epoch": 2161, "lr": 1.957477002306453e-05}
{"train_loss": 0.05428232252597809, "global_step": 192395, "epoch": 2161, "lr": 1.9574309974927263e-05}
{"train_loss": 0.023403719067573547, "global_step": 192396, "epoch": 2161, "lr": 1.957384993088034e-05}
{"train_loss": 0.027004245668649673, "global_step": 192397, "epoch": 2161, "lr": 1.9573389890923855e-05}
{"train_loss": 0.05790967866778374, "global_step": 192398, "epoch": 2161, "lr": 1.9572929855057842e-05}
{"train_loss": 0.08141092956066132, "global_step": 192399, "epoch": 2161, "lr": 1.9572469823282395e-05}
{"train_loss": 0.041760869324207306, "global_step": 192400, "epoch": 2161, "lr": 1.9572009795597545e-05}
{"train_loss": 0.06749452650547028, "global_step": 192401, "epoch": 2161, "lr": 1.9571549772003366e-05}
{"train_loss": 0.053207751363515854, "global_step": 192402, "epoch": 2161, "lr": 1.957108975249994e-05}
{"train_loss": 0.047489799559116364, "global_step": 192403, "epoch": 2161, "lr": 1.9570629737087286e-05}
{"train_loss": 0.08758159726858139, "global_step": 192404, "epoch": 2161, "lr": 1.9570169725765513e-05}
{"train_loss": 0.08308407664299011, "global_step": 192405, "epoch": 2161, "lr": 1.956970971853464e-05}
{"train_loss": 0.0405459851026535, "global_step": 192406, "epoch": 2161, "lr": 1.9569249715394767e-05}
{"train_loss": 0.03337161988019943, "global_step": 192407, "epoch": 2161, "lr": 1.956878971634592e-05}
{"train_loss": 0.042204733937978745, "global_step": 192408, "epoch": 2161, "lr": 1.9568329721388196e-05}
{"train_loss": 0.07350274920463562, "global_step": 192409, "epoch": 2161, "lr": 1.9567869730521622e-05}
{"train_loss": 0.05487198755145073, "global_step": 192410, "epoch": 2161, "lr": 1.956740974374629e-05}
{"train_loss": 0.049866173416376114, "global_step": 192411, "epoch": 2161, "lr": 1.9566949761062235e-05}
{"train_loss": 0.0502912737429142, "global_step": 192412, "epoch": 2161, "lr": 1.9566489782469548e-05}
{"train_loss": 0.06923224031925201, "global_step": 192413, "epoch": 2161, "lr": 1.956602980796826e-05}
{"train_loss": 0.03932901844382286, "global_step": 192414, "epoch": 2161, "lr": 1.9565569837558467e-05}
{"train_loss": 0.08900734037160873, "global_step": 192415, "epoch": 2161, "lr": 1.9565109871240195e-05}
{"train_loss": 0.0400400385260582, "global_step": 192416, "epoch": 2161, "lr": 1.9564649909013543e-05}
{"train_loss": 0.06293637135976486, "global_step": 192417, "epoch": 2161, "lr": 1.956418995087853e-05, "val_loss": 8.015677452087402}
{"train_loss": 0.09729896485805511, "global_step": 192418, "epoch": 2162, "lr": 1.956372999683526e-05}
{"train_loss": 0.04248251020908356, "global_step": 192419, "epoch": 2162, "lr": 1.956327004688376e-05}
{"train_loss": 0.04705917835235596, "global_step": 192420, "epoch": 2162, "lr": 1.956281010102412e-05}
{"train_loss": 0.06740862876176834, "global_step": 192421, "epoch": 2162, "lr": 1.9562350159256395e-05}
{"train_loss": 0.06952162086963654, "global_step": 192422, "epoch": 2162, "lr": 1.956189022158062e-05}
{"train_loss": 0.12677055597305298, "global_step": 192423, "epoch": 2162, "lr": 1.9561430287996895e-05}
{"train_loss": 0.0816647931933403, "global_step": 192424, "epoch": 2162, "lr": 1.9560970358505247e-05}
{"train_loss": 0.04989180341362953, "global_step": 192425, "epoch": 2162, "lr": 1.9560510433105772e-05}
{"train_loss": 0.037602748721838, "global_step": 192426, "epoch": 2162, "lr": 1.95600505117985e-05}
{"train_loss": 0.025837110355496407, "global_step": 192427, "epoch": 2162, "lr": 1.9559590594583526e-05}
{"train_loss": 0.06174156442284584, "global_step": 192428, "epoch": 2162, "lr": 1.955913068146088e-05}
{"train_loss": 0.046621616929769516, "global_step": 192429, "epoch": 2162, "lr": 1.955867077243065e-05}
{"train_loss": 0.07636524736881256, "global_step": 192430, "epoch": 2162, "lr": 1.955821086749287e-05}
{"train_loss": 0.043811969459056854, "global_step": 192431, "epoch": 2162, "lr": 1.9557750966647638e-05}
{"train_loss": 0.07302828878164291, "global_step": 192432, "epoch": 2162, "lr": 1.9557291069894972e-05}
{"train_loss": 0.06490661948919296, "global_step": 192433, "epoch": 2162, "lr": 1.9556831177234973e-05}
{"train_loss": 0.05755416676402092, "global_step": 192434, "epoch": 2162, "lr": 1.955637128866768e-05}
{"train_loss": 0.059644512832164764, "global_step": 192435, "epoch": 2162, "lr": 1.9555911404193172e-05}
{"train_loss": 0.10516554117202759, "global_step": 192436, "epoch": 2162, "lr": 1.9555451523811485e-05}
{"train_loss": 0.10101370513439178, "global_step": 192437, "epoch": 2162, "lr": 1.9554991647522692e-05}
{"train_loss": 0.05819625407457352, "global_step": 192438, "epoch": 2162, "lr": 1.9554531775326884e-05}
{"train_loss": 0.043214354664087296, "global_step": 192439, "epoch": 2162, "lr": 1.9554071907224074e-05}
{"train_loss": 0.0756712555885315, "global_step": 192440, "epoch": 2162, "lr": 1.955361204321437e-05}
{"train_loss": 0.0866314023733139, "global_step": 192441, "epoch": 2162, "lr": 1.955315218329779e-05}
{"train_loss": 0.08724326640367508, "global_step": 192442, "epoch": 2162, "lr": 1.955269232747444e-05}
{"train_loss": 0.05875649303197861, "global_step": 192443, "epoch": 2162, "lr": 1.9552232475744342e-05}
{"train_loss": 0.031971972435712814, "global_step": 192444, "epoch": 2162, "lr": 1.955177262810759e-05}
{"train_loss": 0.13429559767246246, "global_step": 192445, "epoch": 2162, "lr": 1.9551312784564213e-05}
{"train_loss": 0.11353592574596405, "global_step": 192446, "epoch": 2162, "lr": 1.9550852945114312e-05}
{"train_loss": 0.11946962028741837, "global_step": 192447, "epoch": 2162, "lr": 1.955039310975791e-05}
{"train_loss": 0.06087687984108925, "global_step": 192448, "epoch": 2162, "lr": 1.9549933278495096e-05}
{"train_loss": 0.056793462485075, "global_step": 192449, "epoch": 2162, "lr": 1.9549473451325913e-05}
{"train_loss": 0.050201382488012314, "global_step": 192450, "epoch": 2162, "lr": 1.9549013628250447e-05}
{"train_loss": 0.1019468680024147, "global_step": 192451, "epoch": 2162, "lr": 1.9548553809268728e-05}
{"train_loss": 0.0688503310084343, "global_step": 192452, "epoch": 2162, "lr": 1.954809399438085e-05}
{"train_loss": 0.06617191433906555, "global_step": 192453, "epoch": 2162, "lr": 1.9547634183586845e-05}
{"train_loss": 0.05008048564195633, "global_step": 192454, "epoch": 2162, "lr": 1.9547174376886805e-05}
{"train_loss": 0.02501668408513069, "global_step": 192455, "epoch": 2162, "lr": 1.954671457428076e-05}
{"train_loss": 0.07812465727329254, "global_step": 192456, "epoch": 2162, "lr": 1.9546254775768806e-05}
{"train_loss": 0.1093582808971405, "global_step": 192457, "epoch": 2162, "lr": 1.954579498135097e-05}
{"train_loss": 0.057678356766700745, "global_step": 192458, "epoch": 2162, "lr": 1.9545335191027346e-05}
{"train_loss": 0.08224885910749435, "global_step": 192459, "epoch": 2162, "lr": 1.9544875404797985e-05}
{"train_loss": 0.07117648422718048, "global_step": 192460, "epoch": 2162, "lr": 1.954441562266292e-05}
{"train_loss": 0.13265946507453918, "global_step": 192461, "epoch": 2162, "lr": 1.954395584462226e-05}
{"train_loss": 0.06954535841941833, "global_step": 192462, "epoch": 2162, "lr": 1.9543496070676022e-05}
{"train_loss": 0.031675200909376144, "global_step": 192463, "epoch": 2162, "lr": 1.9543036300824315e-05}
{"train_loss": 0.10966162383556366, "global_step": 192464, "epoch": 2162, "lr": 1.9542576535067148e-05}
{"train_loss": 0.07109571248292923, "global_step": 192465, "epoch": 2162, "lr": 1.9542116773404633e-05}
{"train_loss": 0.09380140900611877, "global_step": 192466, "epoch": 2162, "lr": 1.954165701583679e-05}
{"train_loss": 0.05418527126312256, "global_step": 192467, "epoch": 2162, "lr": 1.954119726236372e-05}
{"train_loss": 0.09075400978326797, "global_step": 192468, "epoch": 2162, "lr": 1.9540737512985445e-05}
{"train_loss": 0.04726852849125862, "global_step": 192469, "epoch": 2162, "lr": 1.9540277767702064e-05}
{"train_loss": 0.0637778788805008, "global_step": 192470, "epoch": 2162, "lr": 1.9539818026513607e-05}
{"train_loss": 0.042014773935079575, "global_step": 192471, "epoch": 2162, "lr": 1.9539358289420163e-05}
{"train_loss": 0.07169757783412933, "global_step": 192472, "epoch": 2162, "lr": 1.953889855642177e-05}
{"train_loss": 0.10972745716571808, "global_step": 192473, "epoch": 2162, "lr": 1.9538438827518497e-05}
{"train_loss": 0.03843708336353302, "global_step": 192474, "epoch": 2162, "lr": 1.9537979102710423e-05}
{"train_loss": 0.04582780972123146, "global_step": 192475, "epoch": 2162, "lr": 1.953751938199758e-05}
{"train_loss": 0.05107961222529411, "global_step": 192476, "epoch": 2162, "lr": 1.9537059665380068e-05}
{"train_loss": 0.06770681589841843, "global_step": 192477, "epoch": 2162, "lr": 1.953659995285791e-05}
{"train_loss": 0.04333417862653732, "global_step": 192478, "epoch": 2162, "lr": 1.95361402444312e-05}
{"train_loss": 0.11053211987018585, "global_step": 192479, "epoch": 2162, "lr": 1.953568054009997e-05}
{"train_loss": 0.06030949577689171, "global_step": 192480, "epoch": 2162, "lr": 1.953522083986431e-05}
{"train_loss": 0.08713159710168839, "global_step": 192481, "epoch": 2162, "lr": 1.9534761143724257e-05}
{"train_loss": 0.0752476379275322, "global_step": 192482, "epoch": 2162, "lr": 1.9534301451679892e-05}
{"train_loss": 0.06910688430070877, "global_step": 192483, "epoch": 2162, "lr": 1.953384176373126e-05}
{"train_loss": 0.06805107742547989, "global_step": 192484, "epoch": 2162, "lr": 1.9533382079878442e-05}
{"train_loss": 0.16750425100326538, "global_step": 192485, "epoch": 2162, "lr": 1.953292240012148e-05}
{"train_loss": 0.049562208354473114, "global_step": 192486, "epoch": 2162, "lr": 1.9532462724460455e-05}
{"train_loss": 0.07566235959529877, "global_step": 192487, "epoch": 2162, "lr": 1.9532003052895402e-05}
{"train_loss": 0.05360661447048187, "global_step": 192488, "epoch": 2162, "lr": 1.953154338542642e-05}
{"train_loss": 0.14245383441448212, "global_step": 192489, "epoch": 2162, "lr": 1.9531083722053534e-05}
{"train_loss": 0.04604491591453552, "global_step": 192490, "epoch": 2162, "lr": 1.9530624062776825e-05}
{"train_loss": 0.06671959161758423, "global_step": 192491, "epoch": 2162, "lr": 1.9530164407596364e-05}
{"train_loss": 0.05281961336731911, "global_step": 192492, "epoch": 2162, "lr": 1.95297047565122e-05}
{"train_loss": 0.07992104440927505, "global_step": 192493, "epoch": 2162, "lr": 1.9529245109524385e-05}
{"train_loss": 0.031878724694252014, "global_step": 192494, "epoch": 2162, "lr": 1.952878546663301e-05}
{"train_loss": 0.07682114839553833, "global_step": 192495, "epoch": 2162, "lr": 1.9528325827838096e-05}
{"train_loss": 0.033530816435813904, "global_step": 192496, "epoch": 2162, "lr": 1.9527866193139745e-05}
{"train_loss": 0.12353655695915222, "global_step": 192497, "epoch": 2162, "lr": 1.9527406562537998e-05}
{"train_loss": 0.06665878742933273, "global_step": 192498, "epoch": 2162, "lr": 1.9526946936032908e-05}
{"train_loss": 0.021366314962506294, "global_step": 192499, "epoch": 2162, "lr": 1.9526487313624565e-05}
{"train_loss": 0.02717246301472187, "global_step": 192500, "epoch": 2162, "lr": 1.952602769531299e-05}
{"train_loss": 0.12778480350971222, "global_step": 192501, "epoch": 2162, "lr": 1.9525568081098293e-05}
{"train_loss": 0.05888736993074417, "global_step": 192502, "epoch": 2162, "lr": 1.9525108470980496e-05}
{"train_loss": 0.07959839701652527, "global_step": 192503, "epoch": 2162, "lr": 1.9524648864959692e-05}
{"train_loss": 0.13237006962299347, "global_step": 192504, "epoch": 2162, "lr": 1.952418926303591e-05}
{"train_loss": 0.09188375622034073, "global_step": 192505, "epoch": 2162, "lr": 1.9523729665209245e-05}
{"train_loss": 0.07197167789249607, "global_step": 192506, "epoch": 2162, "lr": 1.9523270071479726e-05, "val_loss": 8.029667854309082}
{"train_loss": 0.06152042746543884, "global_step": 192507, "epoch": 2163, "lr": 1.9522810481847455e-05}
{"train_loss": 0.09110889583826065, "global_step": 192508, "epoch": 2163, "lr": 1.9522350896312448e-05}
{"train_loss": 0.0608358159661293, "global_step": 192509, "epoch": 2163, "lr": 1.952189131487479e-05}
{"train_loss": 0.08351267129182816, "global_step": 192510, "epoch": 2163, "lr": 1.9521431737534556e-05}
{"train_loss": 0.023046009242534637, "global_step": 192511, "epoch": 2163, "lr": 1.9520972164291783e-05}
{"train_loss": 0.0827479287981987, "global_step": 192512, "epoch": 2163, "lr": 1.9520512595146562e-05}
{"train_loss": 0.045103639364242554, "global_step": 192513, "epoch": 2163, "lr": 1.952005303009891e-05}
{"train_loss": 0.09437062591314316, "global_step": 192514, "epoch": 2163, "lr": 1.9519593469148943e-05}
{"train_loss": 0.035462636500597, "global_step": 192515, "epoch": 2163, "lr": 1.951913391229667e-05}
{"train_loss": 0.04795324429869652, "global_step": 192516, "epoch": 2163, "lr": 1.95186743595422e-05}
{"train_loss": 0.08256719261407852, "global_step": 192517, "epoch": 2163, "lr": 1.9518214810885556e-05}
{"train_loss": 0.0761319026350975, "global_step": 192518, "epoch": 2163, "lr": 1.9517755266326832e-05}
{"train_loss": 0.09269571304321289, "global_step": 192519, "epoch": 2163, "lr": 1.951729572586606e-05}
{"train_loss": 0.04492095485329628, "global_step": 192520, "epoch": 2163, "lr": 1.9516836189503328e-05}
{"train_loss": 0.03485146537423134, "global_step": 192521, "epoch": 2163, "lr": 1.951637665723867e-05}
{"train_loss": 0.05841792747378349, "global_step": 192522, "epoch": 2163, "lr": 1.9515917129072182e-05}
{"train_loss": 0.06071152910590172, "global_step": 192523, "epoch": 2163, "lr": 1.951545760500389e-05}
{"train_loss": 0.03913329169154167, "global_step": 192524, "epoch": 2163, "lr": 1.951499808503389e-05}
{"train_loss": 0.06694214791059494, "global_step": 192525, "epoch": 2163, "lr": 1.951453856916221e-05}
{"train_loss": 0.0774727389216423, "global_step": 192526, "epoch": 2163, "lr": 1.9514079057388945e-05}
{"train_loss": 0.03241594508290291, "global_step": 192527, "epoch": 2163, "lr": 1.9513619549714124e-05}
{"train_loss": 0.08633820712566376, "global_step": 192528, "epoch": 2163, "lr": 1.9513160046137822e-05}
{"train_loss": 0.060243863612413406, "global_step": 192529, "epoch": 2163, "lr": 1.951270054666013e-05}
{"train_loss": 0.03733043745160103, "global_step": 192530, "epoch": 2163, "lr": 1.9512241051281077e-05}
{"train_loss": 0.03767223656177521, "global_step": 192531, "epoch": 2163, "lr": 1.951178156000072e-05}
{"train_loss": 0.08678361773490906, "global_step": 192532, "epoch": 2163, "lr": 1.9511322072819142e-05}
{"train_loss": 0.05491616576910019, "global_step": 192533, "epoch": 2163, "lr": 1.9510862589736385e-05}
{"train_loss": 0.0656849592924118, "global_step": 192534, "epoch": 2163, "lr": 1.9510403110752533e-05}
{"train_loss": 0.037669990211725235, "global_step": 192535, "epoch": 2163, "lr": 1.9509943635867615e-05}
{"train_loss": 0.05579141899943352, "global_step": 192536, "epoch": 2163, "lr": 1.9509484165081737e-05}
{"train_loss": 0.05242021754384041, "global_step": 192537, "epoch": 2163, "lr": 1.950902469839494e-05}
{"train_loss": 0.0428684800863266, "global_step": 192538, "epoch": 2163, "lr": 1.9508565235807256e-05}
{"train_loss": 0.0387842021882534, "global_step": 192539, "epoch": 2163, "lr": 1.9508105777318803e-05}
{"train_loss": 0.07386287301778793, "global_step": 192540, "epoch": 2163, "lr": 1.9507646322929592e-05}
{"train_loss": 0.06252741813659668, "global_step": 192541, "epoch": 2163, "lr": 1.9507186872639716e-05}
{"train_loss": 0.09262008965015411, "global_step": 192542, "epoch": 2163, "lr": 1.9506727426449218e-05}
{"train_loss": 0.12227761000394821, "global_step": 192543, "epoch": 2163, "lr": 1.9506267984358184e-05}
{"train_loss": 0.0626639872789383, "global_step": 192544, "epoch": 2163, "lr": 1.9505808546366643e-05}
{"train_loss": 0.055194295942783356, "global_step": 192545, "epoch": 2163, "lr": 1.9505349112474674e-05}
{"train_loss": 0.05615997314453125, "global_step": 192546, "epoch": 2163, "lr": 1.9504889682682352e-05}
{"train_loss": 0.05449822172522545, "global_step": 192547, "epoch": 2163, "lr": 1.9504430256989713e-05}
{"train_loss": 0.10614459216594696, "global_step": 192548, "epoch": 2163, "lr": 1.9503970835396852e-05}
{"train_loss": 0.04502136632800102, "global_step": 192549, "epoch": 2163, "lr": 1.950351141790379e-05}
{"train_loss": 0.0807708352804184, "global_step": 192550, "epoch": 2163, "lr": 1.9503052004510632e-05}
{"train_loss": 0.0786057859659195, "global_step": 192551, "epoch": 2163, "lr": 1.950259259521739e-05}
{"train_loss": 0.09639817476272583, "global_step": 192552, "epoch": 2163, "lr": 1.9502133190024176e-05}
{"train_loss": 0.07195494323968887, "global_step": 192553, "epoch": 2163, "lr": 1.9501673788931006e-05}
{"train_loss": 0.07307800650596619, "global_step": 192554, "epoch": 2163, "lr": 1.9501214391937988e-05}
{"train_loss": 0.06601661443710327, "global_step": 192555, "epoch": 2163, "lr": 1.950075499904514e-05}
{"train_loss": 0.050131261348724365, "global_step": 192556, "epoch": 2163, "lr": 1.950029561025256e-05}
{"train_loss": 0.09296712279319763, "global_step": 192557, "epoch": 2163, "lr": 1.949983622556028e-05}
{"train_loss": 0.05309596657752991, "global_step": 192558, "epoch": 2163, "lr": 1.9499376844968385e-05}
{"train_loss": 0.059209421277046204, "global_step": 192559, "epoch": 2163, "lr": 1.949891746847692e-05}
{"train_loss": 0.05619831755757332, "global_step": 192560, "epoch": 2163, "lr": 1.9498458096085966e-05}
{"train_loss": 0.0427771732211113, "global_step": 192561, "epoch": 2163, "lr": 1.9497998727795553e-05}
{"train_loss": 0.06966204941272736, "global_step": 192562, "epoch": 2163, "lr": 1.9497539363605783e-05}
{"train_loss": 0.04672873392701149, "global_step": 192563, "epoch": 2163, "lr": 1.9497080003516676e-05}
{"train_loss": 0.027880584821105003, "global_step": 192564, "epoch": 2163, "lr": 1.949662064752832e-05}
{"train_loss": 0.06928838044404984, "global_step": 192565, "epoch": 2163, "lr": 1.9496161295640782e-05}
{"train_loss": 0.08222626149654388, "global_step": 192566, "epoch": 2163, "lr": 1.94957019478541e-05}
{"train_loss": 0.06350316107273102, "global_step": 192567, "epoch": 2163, "lr": 1.9495242604168368e-05}
{"train_loss": 0.05447092652320862, "global_step": 192568, "epoch": 2163, "lr": 1.949478326458361e-05}
{"train_loss": 0.06152193993330002, "global_step": 192569, "epoch": 2163, "lr": 1.9494323929099923e-05}
{"train_loss": 0.0733036994934082, "global_step": 192570, "epoch": 2163, "lr": 1.9493864597717356e-05}
{"train_loss": 0.035639308393001556, "global_step": 192571, "epoch": 2163, "lr": 1.9493405270435945e-05}
{"train_loss": 0.13288843631744385, "global_step": 192572, "epoch": 2163, "lr": 1.949294594725579e-05}
{"train_loss": 0.08955809473991394, "global_step": 192573, "epoch": 2163, "lr": 1.9492486628176925e-05}
{"train_loss": 0.06669091433286667, "global_step": 192574, "epoch": 2163, "lr": 1.9492027313199434e-05}
{"train_loss": 0.07901880145072937, "global_step": 192575, "epoch": 2163, "lr": 1.9491568002323375e-05}
{"train_loss": 0.10907427966594696, "global_step": 192576, "epoch": 2163, "lr": 1.949110869554878e-05}
{"train_loss": 0.060774046927690506, "global_step": 192577, "epoch": 2163, "lr": 1.9490649392875755e-05}
{"train_loss": 0.029824333265423775, "global_step": 192578, "epoch": 2163, "lr": 1.9490190094304323e-05}
{"train_loss": 0.06712540239095688, "global_step": 192579, "epoch": 2163, "lr": 1.9489730799834577e-05}
{"train_loss": 0.04912630096077919, "global_step": 192580, "epoch": 2163, "lr": 1.948927150946655e-05}
{"train_loss": 0.11479371041059494, "global_step": 192581, "epoch": 2163, "lr": 1.948881222320032e-05}
{"train_loss": 0.0707613080739975, "global_step": 192582, "epoch": 2163, "lr": 1.948835294103596e-05}
{"train_loss": 0.10242094099521637, "global_step": 192583, "epoch": 2163, "lr": 1.9487893662973505e-05}
{"train_loss": 0.02815866656601429, "global_step": 192584, "epoch": 2163, "lr": 1.948743438901305e-05}
{"train_loss": 0.04232827201485634, "global_step": 192585, "epoch": 2163, "lr": 1.9486975119154615e-05}
{"train_loss": 0.028347086161375046, "global_step": 192586, "epoch": 2163, "lr": 1.9486515853398303e-05}
{"train_loss": 0.049555759876966476, "global_step": 192587, "epoch": 2163, "lr": 1.9486056591744138e-05}
{"train_loss": 0.07791444659233093, "global_step": 192588, "epoch": 2163, "lr": 1.9485597334192225e-05}
{"train_loss": 0.07405318319797516, "global_step": 192589, "epoch": 2163, "lr": 1.9485138080742584e-05}
{"train_loss": 0.09603535383939743, "global_step": 192590, "epoch": 2163, "lr": 1.9484678831395303e-05}
{"train_loss": 0.09990106523036957, "global_step": 192591, "epoch": 2163, "lr": 1.9484219586150422e-05}
{"train_loss": 0.08957767486572266, "global_step": 192592, "epoch": 2163, "lr": 1.9483760345008034e-05}
{"train_loss": 0.07985945791006088, "global_step": 192593, "epoch": 2163, "lr": 1.9483301107968166e-05}
{"train_loss": 0.0691654160618782, "global_step": 192594, "epoch": 2163, "lr": 1.948284187503091e-05}
{"train_loss": 0.0656116580318534, "global_step": 192595, "epoch": 2163, "lr": 1.94823826461963e-05, "val_loss": 8.023561477661133}
{"train_loss": 0.047480225563049316, "global_step": 192596, "epoch": 2164, "lr": 1.9481923421464428e-05}
{"train_loss": 0.05733715370297432, "global_step": 192597, "epoch": 2164, "lr": 1.9481464200835325e-05}
{"train_loss": 0.04118335247039795, "global_step": 192598, "epoch": 2164, "lr": 1.9481004984309083e-05}
{"train_loss": 0.08193095028400421, "global_step": 192599, "epoch": 2164, "lr": 1.9480545771885726e-05}
{"train_loss": 0.06504873931407928, "global_step": 192600, "epoch": 2164, "lr": 1.9480086563565343e-05}
{"train_loss": 0.05902588739991188, "global_step": 192601, "epoch": 2164, "lr": 1.9479627359348e-05}
{"train_loss": 0.022580772638320923, "global_step": 192602, "epoch": 2164, "lr": 1.947916815923374e-05}
{"train_loss": 0.052944667637348175, "global_step": 192603, "epoch": 2164, "lr": 1.9478708963222648e-05}
{"train_loss": 0.03767445683479309, "global_step": 192604, "epoch": 2164, "lr": 1.9478249771314754e-05}
{"train_loss": 0.015172554180026054, "global_step": 192605, "epoch": 2164, "lr": 1.947779058351016e-05}
{"train_loss": 0.05675096809864044, "global_step": 192606, "epoch": 2164, "lr": 1.9477331399808877e-05}
{"train_loss": 0.06349464505910873, "global_step": 192607, "epoch": 2164, "lr": 1.947687222021102e-05}
{"train_loss": 0.09199555218219757, "global_step": 192608, "epoch": 2164, "lr": 1.9476413044716625e-05}
{"train_loss": 0.08140549808740616, "global_step": 192609, "epoch": 2164, "lr": 1.9475953873325737e-05}
{"train_loss": 0.10290096700191498, "global_step": 192610, "epoch": 2164, "lr": 1.947549470603845e-05}
{"train_loss": 0.05708567798137665, "global_step": 192611, "epoch": 2164, "lr": 1.947503554285479e-05}
{"train_loss": 0.053116943687200546, "global_step": 192612, "epoch": 2164, "lr": 1.9474576383774863e-05}
{"train_loss": 0.05237668380141258, "global_step": 192613, "epoch": 2164, "lr": 1.9474117228798704e-05}
{"train_loss": 0.03706667199730873, "global_step": 192614, "epoch": 2164, "lr": 1.9473658077926365e-05}
{"train_loss": 0.057300884276628494, "global_step": 192615, "epoch": 2164, "lr": 1.947319893115793e-05}
{"train_loss": 0.07627780735492706, "global_step": 192616, "epoch": 2164, "lr": 1.9472739788493444e-05}
{"train_loss": 0.029216378927230835, "global_step": 192617, "epoch": 2164, "lr": 1.9472280649932966e-05}
{"train_loss": 0.07664605230093002, "global_step": 192618, "epoch": 2164, "lr": 1.947182151547659e-05}
{"train_loss": 0.08343919366598129, "global_step": 192619, "epoch": 2164, "lr": 1.9471362385124343e-05}
{"train_loss": 0.043969858437776566, "global_step": 192620, "epoch": 2164, "lr": 1.947090325887631e-05}
{"train_loss": 0.05497726798057556, "global_step": 192621, "epoch": 2164, "lr": 1.9470444136732523e-05}
{"train_loss": 0.09351681917905807, "global_step": 192622, "epoch": 2164, "lr": 1.9469985018693083e-05}
{"train_loss": 0.07564511150121689, "global_step": 192623, "epoch": 2164, "lr": 1.9469525904758018e-05}
{"train_loss": 0.024042842909693718, "global_step": 192624, "epoch": 2164, "lr": 1.9469066794927414e-05}
{"train_loss": 0.07441168278455734, "global_step": 192625, "epoch": 2164, "lr": 1.9468607689201308e-05}
{"train_loss": 0.07261195033788681, "global_step": 192626, "epoch": 2164, "lr": 1.946814858757979e-05}
{"train_loss": 0.03387581184506416, "global_step": 192627, "epoch": 2164, "lr": 1.9467689490062895e-05}
{"train_loss": 0.08119402080774307, "global_step": 192628, "epoch": 2164, "lr": 1.946723039665071e-05}
{"train_loss": 0.026258762925863266, "global_step": 192629, "epoch": 2164, "lr": 1.9466771307343268e-05}
{"train_loss": 0.0544663667678833, "global_step": 192630, "epoch": 2164, "lr": 1.9466312222140665e-05}
{"train_loss": 0.05207308009266853, "global_step": 192631, "epoch": 2164, "lr": 1.9465853141042923e-05}
{"train_loss": 0.04836087301373482, "global_step": 192632, "epoch": 2164, "lr": 1.9465394064050147e-05}
{"train_loss": 0.0633017048239708, "global_step": 192633, "epoch": 2164, "lr": 1.9464934991162353e-05}
{"train_loss": 0.08079749345779419, "global_step": 192634, "epoch": 2164, "lr": 1.946447592237965e-05}
{"train_loss": 0.05651036649942398, "global_step": 192635, "epoch": 2164, "lr": 1.9464016857702055e-05}
{"train_loss": 0.07931939512491226, "global_step": 192636, "epoch": 2164, "lr": 1.9463557797129656e-05}
{"train_loss": 0.06228100135922432, "global_step": 192637, "epoch": 2164, "lr": 1.946309874066252e-05}
{"train_loss": 0.04817608743906021, "global_step": 192638, "epoch": 2164, "lr": 1.9462639688300683e-05}
{"train_loss": 0.031803105026483536, "global_step": 192639, "epoch": 2164, "lr": 1.946218064004424e-05}
{"train_loss": 0.07164620608091354, "global_step": 192640, "epoch": 2164, "lr": 1.946172159589322e-05}
{"train_loss": 0.06804753839969635, "global_step": 192641, "epoch": 2164, "lr": 1.9461262555847715e-05}
{"train_loss": 0.049779508262872696, "global_step": 192642, "epoch": 2164, "lr": 1.946080351990775e-05}
{"train_loss": 0.06754564493894577, "global_step": 192643, "epoch": 2164, "lr": 1.946034448807343e-05}
{"train_loss": 0.04068021476268768, "global_step": 192644, "epoch": 2164, "lr": 1.9459885460344774e-05}
{"train_loss": 0.09835929423570633, "global_step": 192645, "epoch": 2164, "lr": 1.945942643672188e-05}
{"train_loss": 0.04845094308257103, "global_step": 192646, "epoch": 2164, "lr": 1.9458967417204778e-05}
{"train_loss": 0.0610203854739666, "global_step": 192647, "epoch": 2164, "lr": 1.9458508401793564e-05}
{"train_loss": 0.043157294392585754, "global_step": 192648, "epoch": 2164, "lr": 1.945804939048828e-05}
{"train_loss": 0.09421908855438232, "global_step": 192649, "epoch": 2164, "lr": 1.9457590383288976e-05}
{"train_loss": 0.09868544340133667, "global_step": 192650, "epoch": 2164, "lr": 1.9457131380195738e-05}
{"train_loss": 0.06284842640161514, "global_step": 192651, "epoch": 2164, "lr": 1.94566723812086e-05}
{"train_loss": 0.08502314984798431, "global_step": 192652, "epoch": 2164, "lr": 1.945621338632766e-05}
{"train_loss": 0.050461381673812866, "global_step": 192653, "epoch": 2164, "lr": 1.945575439555296e-05}
{"train_loss": 0.10707683861255646, "global_step": 192654, "epoch": 2164, "lr": 1.9455295408884545e-05}
{"train_loss": 0.07066941261291504, "global_step": 192655, "epoch": 2164, "lr": 1.945483642632249e-05}
{"train_loss": 0.019673379138112068, "global_step": 192656, "epoch": 2164, "lr": 1.9454377447866877e-05}
{"train_loss": 0.12190273404121399, "global_step": 192657, "epoch": 2164, "lr": 1.945391847351774e-05}
{"train_loss": 0.07742854207754135, "global_step": 192658, "epoch": 2164, "lr": 1.9453459503275163e-05}
{"train_loss": 0.051492415368556976, "global_step": 192659, "epoch": 2164, "lr": 1.9453000537139176e-05}
{"train_loss": 0.055113449692726135, "global_step": 192660, "epoch": 2164, "lr": 1.9452541575109885e-05}
{"train_loss": 0.0591653548181057, "global_step": 192661, "epoch": 2164, "lr": 1.9452082617187305e-05}
{"train_loss": 0.057967670261859894, "global_step": 192662, "epoch": 2164, "lr": 1.9451623663371544e-05}
{"train_loss": 0.08036497980356216, "global_step": 192663, "epoch": 2164, "lr": 1.945116471366261e-05}
{"train_loss": 0.05266018584370613, "global_step": 192664, "epoch": 2164, "lr": 1.9450705768060624e-05}
{"train_loss": 0.04830985516309738, "global_step": 192665, "epoch": 2164, "lr": 1.94502468265656e-05}
{"train_loss": 0.042211469262838364, "global_step": 192666, "epoch": 2164, "lr": 1.944978788917763e-05}
{"train_loss": 0.13060349225997925, "global_step": 192667, "epoch": 2164, "lr": 1.9449328955896746e-05}
{"train_loss": 0.07343722879886627, "global_step": 192668, "epoch": 2164, "lr": 1.944887002672305e-05}
{"train_loss": 0.11771684139966965, "global_step": 192669, "epoch": 2164, "lr": 1.944841110165656e-05}
{"train_loss": 0.0771958976984024, "global_step": 192670, "epoch": 2164, "lr": 1.9447952180697376e-05}
{"train_loss": 0.061795782297849655, "global_step": 192671, "epoch": 2164, "lr": 1.9447493263845522e-05}
{"train_loss": 0.12165793031454086, "global_step": 192672, "epoch": 2164, "lr": 1.944703435110109e-05}
{"train_loss": 0.09123606979846954, "global_step": 192673, "epoch": 2164, "lr": 1.9446575442464143e-05}
{"train_loss": 0.06714177876710892, "global_step": 192674, "epoch": 2164, "lr": 1.944611653793471e-05}
{"train_loss": 0.06326672434806824, "global_step": 192675, "epoch": 2164, "lr": 1.94456576375129e-05}
{"train_loss": 0.12054294347763062, "global_step": 192676, "epoch": 2164, "lr": 1.9445198741198728e-05}
{"train_loss": 0.10304027795791626, "global_step": 192677, "epoch": 2164, "lr": 1.9444739848992294e-05}
{"train_loss": 0.055581267923116684, "global_step": 192678, "epoch": 2164, "lr": 1.9444280960893625e-05}
{"train_loss": 0.08494442701339722, "global_step": 192679, "epoch": 2164, "lr": 1.944382207690281e-05}
{"train_loss": 0.11263129115104675, "global_step": 192680, "epoch": 2164, "lr": 1.9443363197019894e-05}
{"train_loss": 0.04909393936395645, "global_step": 192681, "epoch": 2164, "lr": 1.9442904321244957e-05}
{"train_loss": 0.08436324447393417, "global_step": 192682, "epoch": 2164, "lr": 1.9442445449578033e-05}
{"train_loss": 0.07811074703931808, "global_step": 192683, "epoch": 2164, "lr": 1.944198658201922e-05}
{"train_loss": 0.0665724699626143, "global_step": 192684, "epoch": 2164, "lr": 1.9441527718568543e-05, "val_loss": 8.080488204956055}
{"train_loss": 0.035665471106767654, "global_step": 192685, "epoch": 2165, "lr": 1.944106885922609e-05}
{"train_loss": 0.0641571655869484, "global_step": 192686, "epoch": 2165, "lr": 1.9440610003991923e-05}
{"train_loss": 0.15432319045066833, "global_step": 192687, "epoch": 2165, "lr": 1.9440151152866065e-05}
{"train_loss": 0.07345361262559891, "global_step": 192688, "epoch": 2165, "lr": 1.943969230584863e-05}
{"train_loss": 0.08221031725406647, "global_step": 192689, "epoch": 2165, "lr": 1.943923346293964e-05}
{"train_loss": 0.07805225998163223, "global_step": 192690, "epoch": 2165, "lr": 1.9438774624139185e-05}
{"train_loss": 0.0777030661702156, "global_step": 192691, "epoch": 2165, "lr": 1.9438315789447298e-05}
{"train_loss": 0.08132188022136688, "global_step": 192692, "epoch": 2165, "lr": 1.9437856958864077e-05}
{"train_loss": 0.07811927795410156, "global_step": 192693, "epoch": 2165, "lr": 1.9437398132389545e-05}
{"train_loss": 0.05331556871533394, "global_step": 192694, "epoch": 2165, "lr": 1.94369393100238e-05}
{"train_loss": 0.09683041274547577, "global_step": 192695, "epoch": 2165, "lr": 1.9436480491766866e-05}
{"train_loss": 0.05980992317199707, "global_step": 192696, "epoch": 2165, "lr": 1.943602167761885e-05}
{"train_loss": 0.07516218721866608, "global_step": 192697, "epoch": 2165, "lr": 1.943556286757976e-05}
{"train_loss": 0.08981753140687943, "global_step": 192698, "epoch": 2165, "lr": 1.943510406164971e-05}
{"train_loss": 0.023576734587550163, "global_step": 192699, "epoch": 2165, "lr": 1.943464525982872e-05}
{"train_loss": 0.09969459474086761, "global_step": 192700, "epoch": 2165, "lr": 1.9434186462116883e-05}
{"train_loss": 0.06265880912542343, "global_step": 192701, "epoch": 2165, "lr": 1.9433727668514235e-05}
{"train_loss": 0.07766104489564896, "global_step": 192702, "epoch": 2165, "lr": 1.9433268879020867e-05}
{"train_loss": 0.04535142332315445, "global_step": 192703, "epoch": 2165, "lr": 1.94328100936368e-05}
{"train_loss": 0.049731750041246414, "global_step": 192704, "epoch": 2165, "lr": 1.943235131236214e-05}
{"train_loss": 0.09835152328014374, "global_step": 192705, "epoch": 2165, "lr": 1.943189253519691e-05}
{"train_loss": 0.0642312541604042, "global_step": 192706, "epoch": 2165, "lr": 1.943143376214121e-05}
{"train_loss": 0.05958116427063942, "global_step": 192707, "epoch": 2165, "lr": 1.9430974993195062e-05}
{"train_loss": 0.015075384639203548, "global_step": 192708, "epoch": 2165, "lr": 1.943051622835854e-05}
{"train_loss": 0.03642804175615311, "global_step": 192709, "epoch": 2165, "lr": 1.943005746763174e-05}
{"train_loss": 0.07290744036436081, "global_step": 192710, "epoch": 2165, "lr": 1.9429598711014674e-05}
{"train_loss": 0.10125307738780975, "global_step": 192711, "epoch": 2165, "lr": 1.9429139958507447e-05}
{"train_loss": 0.07963965833187103, "global_step": 192712, "epoch": 2165, "lr": 1.9428681210110077e-05}
{"train_loss": 0.0754234790802002, "global_step": 192713, "epoch": 2165, "lr": 1.942822246582267e-05}
{"train_loss": 0.15754175186157227, "global_step": 192714, "epoch": 2165, "lr": 1.942776372564524e-05}
{"train_loss": 0.10043127089738846, "global_step": 192715, "epoch": 2165, "lr": 1.9427304989577903e-05}
{"train_loss": 0.024797413498163223, "global_step": 192716, "epoch": 2165, "lr": 1.9426846257620667e-05}
{"train_loss": 0.08248976618051529, "global_step": 192717, "epoch": 2165, "lr": 1.9426387529773642e-05}
{"train_loss": 0.09448462724685669, "global_step": 192718, "epoch": 2165, "lr": 1.9425928806036847e-05}
{"train_loss": 0.1081000491976738, "global_step": 192719, "epoch": 2165, "lr": 1.942547008641038e-05}
{"train_loss": 0.03066849894821644, "global_step": 192720, "epoch": 2165, "lr": 1.942501137089427e-05}
{"train_loss": 0.07057377696037292, "global_step": 192721, "epoch": 2165, "lr": 1.942455265948861e-05}
{"train_loss": 0.15138059854507446, "global_step": 192722, "epoch": 2165, "lr": 1.942409395219343e-05}
{"train_loss": 0.07147625088691711, "global_step": 192723, "epoch": 2165, "lr": 1.942363524900882e-05}
{"train_loss": 0.10100659728050232, "global_step": 192724, "epoch": 2165, "lr": 1.9423176549934835e-05}
{"train_loss": 0.05989697575569153, "global_step": 192725, "epoch": 2165, "lr": 1.9422717854971516e-05}
{"train_loss": 0.05464189872145653, "global_step": 192726, "epoch": 2165, "lr": 1.9422259164118956e-05}
{"train_loss": 0.079674132168293, "global_step": 192727, "epoch": 2165, "lr": 1.942180047737718e-05}
{"train_loss": 0.033662497997283936, "global_step": 192728, "epoch": 2165, "lr": 1.942134179474629e-05}
{"train_loss": 0.0676666870713234, "global_step": 192729, "epoch": 2165, "lr": 1.9420883116226306e-05}
{"train_loss": 0.07459347695112228, "global_step": 192730, "epoch": 2165, "lr": 1.9420424441817332e-05}
{"train_loss": 0.12019292265176773, "global_step": 192731, "epoch": 2165, "lr": 1.9419965771519396e-05}
{"train_loss": 0.03898152709007263, "global_step": 192732, "epoch": 2165, "lr": 1.9419507105332584e-05}
{"train_loss": 0.08593939989805222, "global_step": 192733, "epoch": 2165, "lr": 1.9419048443256932e-05}
{"train_loss": 0.06872659921646118, "global_step": 192734, "epoch": 2165, "lr": 1.9418589785292533e-05}
{"train_loss": 0.08217823505401611, "global_step": 192735, "epoch": 2165, "lr": 1.9418131131439416e-05}
{"train_loss": 0.04940643906593323, "global_step": 192736, "epoch": 2165, "lr": 1.9417672481697675e-05}
{"train_loss": 0.018176132813096046, "global_step": 192737, "epoch": 2165, "lr": 1.941721383606734e-05}
{"train_loss": 0.06022518128156662, "global_step": 192738, "epoch": 2165, "lr": 1.9416755194548503e-05}
{"train_loss": 0.057909246534109116, "global_step": 192739, "epoch": 2165, "lr": 1.9416296557141188e-05}
{"train_loss": 0.11563108116388321, "global_step": 192740, "epoch": 2165, "lr": 1.9415837923845502e-05}
{"train_loss": 0.06175106018781662, "global_step": 192741, "epoch": 2165, "lr": 1.9415379294661468e-05}
{"train_loss": 0.03840283304452896, "global_step": 192742, "epoch": 2165, "lr": 1.9414920669589175e-05}
{"train_loss": 0.037776749581098557, "global_step": 192743, "epoch": 2165, "lr": 1.941446204862866e-05}
{"train_loss": 0.06857612729072571, "global_step": 192744, "epoch": 2165, "lr": 1.9414003431779997e-05}
{"train_loss": 0.15053685009479523, "global_step": 192745, "epoch": 2165, "lr": 1.9413544819043272e-05}
{"train_loss": 0.09835992008447647, "global_step": 192746, "epoch": 2165, "lr": 1.9413086210418496e-05}
{"train_loss": 0.0926501452922821, "global_step": 192747, "epoch": 2165, "lr": 1.9412627605905783e-05}
{"train_loss": 0.04556936025619507, "global_step": 192748, "epoch": 2165, "lr": 1.9412169005505144e-05}
{"train_loss": 0.04270782321691513, "global_step": 192749, "epoch": 2165, "lr": 1.9411710409216687e-05}
{"train_loss": 0.0685342475771904, "global_step": 192750, "epoch": 2165, "lr": 1.9411251817040436e-05}
{"train_loss": 0.0945533961057663, "global_step": 192751, "epoch": 2165, "lr": 1.9410793228976487e-05}
{"train_loss": 0.1405969262123108, "global_step": 192752, "epoch": 2165, "lr": 1.941033464502487e-05}
{"train_loss": 0.034817930310964584, "global_step": 192753, "epoch": 2165, "lr": 1.9409876065185673e-05}
{"train_loss": 0.04214885085821152, "global_step": 192754, "epoch": 2165, "lr": 1.940941748945893e-05}
{"train_loss": 0.05319448933005333, "global_step": 192755, "epoch": 2165, "lr": 1.9408958917844734e-05}
{"train_loss": 0.047349438071250916, "global_step": 192756, "epoch": 2165, "lr": 1.9408500350343112e-05}
{"train_loss": 0.060732051730155945, "global_step": 192757, "epoch": 2165, "lr": 1.940804178695416e-05}
{"train_loss": 0.08868200331926346, "global_step": 192758, "epoch": 2165, "lr": 1.9407583227677917e-05}
{"train_loss": 0.05777139216661453, "global_step": 192759, "epoch": 2165, "lr": 1.9407124672514458e-05}
{"train_loss": 0.07013843953609467, "global_step": 192760, "epoch": 2165, "lr": 1.940666612146383e-05}
{"train_loss": 0.0685541108250618, "global_step": 192761, "epoch": 2165, "lr": 1.9406207574526114e-05}
{"train_loss": 0.07617542147636414, "global_step": 192762, "epoch": 2165, "lr": 1.9405749031701343e-05}
{"train_loss": 0.13842293620109558, "global_step": 192763, "epoch": 2165, "lr": 1.9405290492989612e-05}
{"train_loss": 0.08647343516349792, "global_step": 192764, "epoch": 2165, "lr": 1.9404831958390972e-05}
{"train_loss": 0.044163770973682404, "global_step": 192765, "epoch": 2165, "lr": 1.940437342790546e-05}
{"train_loss": 0.0735919401049614, "global_step": 192766, "epoch": 2165, "lr": 1.9403914901533172e-05}
{"train_loss": 0.05450756475329399, "global_step": 192767, "epoch": 2165, "lr": 1.9403456379274138e-05}
{"train_loss": 0.06073245778679848, "global_step": 192768, "epoch": 2165, "lr": 1.940299786112846e-05}
{"train_loss": 0.06981387734413147, "global_step": 192769, "epoch": 2165, "lr": 1.940253934709615e-05}
{"train_loss": 0.07874031364917755, "global_step": 192770, "epoch": 2165, "lr": 1.9402080837177316e-05}
{"train_loss": 0.13925759494304657, "global_step": 192771, "epoch": 2165, "lr": 1.9401622331371982e-05}
{"train_loss": 0.08306536078453064, "global_step": 192772, "epoch": 2165, "lr": 1.9401163829680246e-05}
{"train_loss": 0.07320718547810665, "global_step": 192773, "epoch": 2165, "lr": 1.9400705332102133e-05, "val_loss": 7.809433460235596, "train_action_mse_error": 27.277433395385742}
{"train_loss": 0.06655975431203842, "global_step": 192774, "epoch": 2166, "lr": 1.9400246838637743e-05}
{"train_loss": 0.04405658692121506, "global_step": 192775, "epoch": 2166, "lr": 1.939978834928709e-05}
{"train_loss": 0.0566241592168808, "global_step": 192776, "epoch": 2166, "lr": 1.9399329864050285e-05}
{"train_loss": 0.07316602021455765, "global_step": 192777, "epoch": 2166, "lr": 1.939887138292735e-05}
{"train_loss": 0.020927822217345238, "global_step": 192778, "epoch": 2166, "lr": 1.9398412905918383e-05}
{"train_loss": 0.04941234737634659, "global_step": 192779, "epoch": 2166, "lr": 1.9397954433023406e-05}
{"train_loss": 0.05651718005537987, "global_step": 192780, "epoch": 2166, "lr": 1.939749596424252e-05}
{"train_loss": 0.03435184061527252, "global_step": 192781, "epoch": 2166, "lr": 1.939703749957575e-05}
{"train_loss": 0.09392218291759491, "global_step": 192782, "epoch": 2166, "lr": 1.9396579039023176e-05}
{"train_loss": 0.08966042101383209, "global_step": 192783, "epoch": 2166, "lr": 1.9396120582584877e-05}
{"train_loss": 0.03719170019030571, "global_step": 192784, "epoch": 2166, "lr": 1.9395662130260878e-05}
{"train_loss": 0.06804138422012329, "global_step": 192785, "epoch": 2166, "lr": 1.9395203682051273e-05}
{"train_loss": 0.035074908286333084, "global_step": 192786, "epoch": 2166, "lr": 1.9394745237956097e-05}
{"train_loss": 0.05112659931182861, "global_step": 192787, "epoch": 2166, "lr": 1.939428679797544e-05}
{"train_loss": 0.037959203124046326, "global_step": 192788, "epoch": 2166, "lr": 1.939382836210933e-05}
{"train_loss": 0.09973589330911636, "global_step": 192789, "epoch": 2166, "lr": 1.939336993035787e-05}
{"train_loss": 0.08529835939407349, "global_step": 192790, "epoch": 2166, "lr": 1.9392911502721073e-05}
{"train_loss": 0.08298531174659729, "global_step": 192791, "epoch": 2166, "lr": 1.9392453079199047e-05}
{"train_loss": 0.08978456258773804, "global_step": 192792, "epoch": 2166, "lr": 1.9391994659791812e-05}
{"train_loss": 0.08179374784231186, "global_step": 192793, "epoch": 2166, "lr": 1.9391536244499476e-05}
{"train_loss": 0.05912770703434944, "global_step": 192794, "epoch": 2166, "lr": 1.939107783332205e-05}
{"train_loss": 0.04419687017798424, "global_step": 192795, "epoch": 2166, "lr": 1.9390619426259636e-05}
{"train_loss": 0.09670353680849075, "global_step": 192796, "epoch": 2166, "lr": 1.939016102331227e-05}
{"train_loss": 0.05534602329134941, "global_step": 192797, "epoch": 2166, "lr": 1.938970262448004e-05}
{"train_loss": 0.05087391287088394, "global_step": 192798, "epoch": 2166, "lr": 1.938924422976297e-05}
{"train_loss": 0.03923237696290016, "global_step": 192799, "epoch": 2166, "lr": 1.938878583916116e-05}
{"train_loss": 0.06473381817340851, "global_step": 192800, "epoch": 2166, "lr": 1.9388327452674644e-05}
{"train_loss": 0.06603766977787018, "global_step": 192801, "epoch": 2166, "lr": 1.9387869070303504e-05}
{"train_loss": 0.025846799835562706, "global_step": 192802, "epoch": 2166, "lr": 1.9387410692047796e-05}
{"train_loss": 0.08559209853410721, "global_step": 192803, "epoch": 2166, "lr": 1.938695231790756e-05}
{"train_loss": 0.06633853167295456, "global_step": 192804, "epoch": 2166, "lr": 1.9386493947882895e-05}
{"train_loss": 0.07793910801410675, "global_step": 192805, "epoch": 2166, "lr": 1.938603558197382e-05}
{"train_loss": 0.11642671376466751, "global_step": 192806, "epoch": 2166, "lr": 1.938557722018044e-05}
{"train_loss": 0.04151710495352745, "global_step": 192807, "epoch": 2166, "lr": 1.9385118862502775e-05}
{"train_loss": 0.07735008746385574, "global_step": 192808, "epoch": 2166, "lr": 1.9384660508940927e-05}
{"train_loss": 0.048666104674339294, "global_step": 192809, "epoch": 2166, "lr": 1.9384202159494924e-05}
{"train_loss": 0.036312002688646317, "global_step": 192810, "epoch": 2166, "lr": 1.9383743814164857e-05}
{"train_loss": 0.08219302445650101, "global_step": 192811, "epoch": 2166, "lr": 1.9383285472950753e-05}
{"train_loss": 0.05678350850939751, "global_step": 192812, "epoch": 2166, "lr": 1.9382827135852715e-05}
{"train_loss": 0.08928746730089188, "global_step": 192813, "epoch": 2166, "lr": 1.938236880287076e-05}
{"train_loss": 0.12392772734165192, "global_step": 192814, "epoch": 2166, "lr": 1.9381910474004993e-05}
{"train_loss": 0.03287152945995331, "global_step": 192815, "epoch": 2166, "lr": 1.9381452149255434e-05}
{"train_loss": 0.0790693610906601, "global_step": 192816, "epoch": 2166, "lr": 1.9380993828622192e-05}
{"train_loss": 0.06642857939004898, "global_step": 192817, "epoch": 2166, "lr": 1.9380535512105275e-05}
{"train_loss": 0.11274360120296478, "global_step": 192818, "epoch": 2166, "lr": 1.9380077199704776e-05}
{"train_loss": 0.04291120544075966, "global_step": 192819, "epoch": 2166, "lr": 1.937961889142077e-05}
{"train_loss": 0.019525961950421333, "global_step": 192820, "epoch": 2166, "lr": 1.937916058725328e-05}
{"train_loss": 0.02737068012356758, "global_step": 192821, "epoch": 2166, "lr": 1.937870228720241e-05}
{"train_loss": 0.1372336447238922, "global_step": 192822, "epoch": 2166, "lr": 1.9378243991268185e-05}
{"train_loss": 0.07101422548294067, "global_step": 192823, "epoch": 2166, "lr": 1.9377785699450695e-05}
{"train_loss": 0.03984691575169563, "global_step": 192824, "epoch": 2166, "lr": 1.9377327411749975e-05}
{"train_loss": 0.10225872695446014, "global_step": 192825, "epoch": 2166, "lr": 1.9376869128166113e-05}
{"train_loss": 0.06947951018810272, "global_step": 192826, "epoch": 2166, "lr": 1.937641084869914e-05}
{"train_loss": 0.07257602363824844, "global_step": 192827, "epoch": 2166, "lr": 1.9375952573349156e-05}
{"train_loss": 0.062246885150671005, "global_step": 192828, "epoch": 2166, "lr": 1.9375494302116183e-05}
{"train_loss": 0.07676416635513306, "global_step": 192829, "epoch": 2166, "lr": 1.937503603500032e-05}
{"train_loss": 0.06550989300012589, "global_step": 192830, "epoch": 2166, "lr": 1.937457777200159e-05}
{"train_loss": 0.05920715257525444, "global_step": 192831, "epoch": 2166, "lr": 1.937411951312009e-05}
{"train_loss": 0.05978977307677269, "global_step": 192832, "epoch": 2166, "lr": 1.9373661258355858e-05}
{"train_loss": 0.06797905266284943, "global_step": 192833, "epoch": 2166, "lr": 1.9373203007708973e-05}
{"train_loss": 0.03482885658740997, "global_step": 192834, "epoch": 2166, "lr": 1.937274476117948e-05}
{"train_loss": 0.08248461782932281, "global_step": 192835, "epoch": 2166, "lr": 1.937228651876746e-05}
{"train_loss": 0.04238951578736305, "global_step": 192836, "epoch": 2166, "lr": 1.937182828047294e-05}
{"train_loss": 0.025821257382631302, "global_step": 192837, "epoch": 2166, "lr": 1.937137004629603e-05}
{"train_loss": 0.08490580320358276, "global_step": 192838, "epoch": 2166, "lr": 1.937091181623674e-05}
{"train_loss": 0.06287705153226852, "global_step": 192839, "epoch": 2166, "lr": 1.937045359029518e-05}
{"train_loss": 0.05142813175916672, "global_step": 192840, "epoch": 2166, "lr": 1.936999536847139e-05}
{"train_loss": 0.07283176481723785, "global_step": 192841, "epoch": 2166, "lr": 1.936953715076541e-05}
{"train_loss": 0.045002784579992294, "global_step": 192842, "epoch": 2166, "lr": 1.936907893717734e-05}
{"train_loss": 0.050820235162973404, "global_step": 192843, "epoch": 2166, "lr": 1.9368620727707216e-05}
{"train_loss": 0.16029293835163116, "global_step": 192844, "epoch": 2166, "lr": 1.9368162522355117e-05}
{"train_loss": 0.06820231676101685, "global_step": 192845, "epoch": 2166, "lr": 1.9367704321121078e-05}
{"train_loss": 0.062126558274030685, "global_step": 192846, "epoch": 2166, "lr": 1.9367246124005196e-05}
{"train_loss": 0.06330890953540802, "global_step": 192847, "epoch": 2166, "lr": 1.9366787931007496e-05}
{"train_loss": 0.14392855763435364, "global_step": 192848, "epoch": 2166, "lr": 1.9366329742128077e-05}
{"train_loss": 0.056712061166763306, "global_step": 192849, "epoch": 2166, "lr": 1.936587155736696e-05}
{"train_loss": 0.13302862644195557, "global_step": 192850, "epoch": 2166, "lr": 1.936541337672425e-05}
{"train_loss": 0.06003129482269287, "global_step": 192851, "epoch": 2166, "lr": 1.936495520019997e-05}
{"train_loss": 0.042157746851444244, "global_step": 192852, "epoch": 2166, "lr": 1.9364497027794216e-05}
{"train_loss": 0.04899057000875473, "global_step": 192853, "epoch": 2166, "lr": 1.9364038859507012e-05}
{"train_loss": 0.05252091586589813, "global_step": 192854, "epoch": 2166, "lr": 1.936358069533844e-05}
{"train_loss": 0.034143414348363876, "global_step": 192855, "epoch": 2166, "lr": 1.936312253528858e-05}
{"train_loss": 0.0588577538728714, "global_step": 192856, "epoch": 2166, "lr": 1.9362664379357453e-05}
{"train_loss": 0.08718851208686829, "global_step": 192857, "epoch": 2166, "lr": 1.9362206227545165e-05}
{"train_loss": 0.12426932156085968, "global_step": 192858, "epoch": 2166, "lr": 1.9361748079851733e-05}
{"train_loss": 0.07717936486005783, "global_step": 192859, "epoch": 2166, "lr": 1.936128993627726e-05}
{"train_loss": 0.07767569273710251, "global_step": 192860, "epoch": 2166, "lr": 1.936083179682177e-05}
{"train_loss": 0.030886119231581688, "global_step": 192861, "epoch": 2166, "lr": 1.9360373661485364e-05}
{"train_loss": 0.0666094608604908, "global_step": 192862, "epoch": 2166, "lr": 1.9359915530268065e-05, "val_loss": 7.91897439956665}
{"train_loss": 0.0994509756565094, "global_step": 192863, "epoch": 2167, "lr": 1.935945740316997e-05}
{"train_loss": 0.04797825217247009, "global_step": 192864, "epoch": 2167, "lr": 1.93589992801911e-05}
{"train_loss": 0.015961939468979836, "global_step": 192865, "epoch": 2167, "lr": 1.9358541161331557e-05}
{"train_loss": 0.08418521285057068, "global_step": 192866, "epoch": 2167, "lr": 1.9358083046591363e-05}
{"train_loss": 0.09293927252292633, "global_step": 192867, "epoch": 2167, "lr": 1.9357624935970626e-05}
{"train_loss": 0.05980915203690529, "global_step": 192868, "epoch": 2167, "lr": 1.9357166829469365e-05}
{"train_loss": 0.09284307062625885, "global_step": 192869, "epoch": 2167, "lr": 1.9356708727087674e-05}
{"train_loss": 0.07179973274469376, "global_step": 192870, "epoch": 2167, "lr": 1.9356250628825584e-05}
{"train_loss": 0.07534094899892807, "global_step": 192871, "epoch": 2167, "lr": 1.9355792534683175e-05}
{"train_loss": 0.06910670548677444, "global_step": 192872, "epoch": 2167, "lr": 1.9355334444660524e-05}
{"train_loss": 0.05155875161290169, "global_step": 192873, "epoch": 2167, "lr": 1.935487635875767e-05}
{"train_loss": 0.05228679999709129, "global_step": 192874, "epoch": 2167, "lr": 1.9354418276974663e-05}
{"train_loss": 0.0443449541926384, "global_step": 192875, "epoch": 2167, "lr": 1.9353960199311598e-05}
{"train_loss": 0.07615555077791214, "global_step": 192876, "epoch": 2167, "lr": 1.93535021257685e-05}
{"train_loss": 0.030832991003990173, "global_step": 192877, "epoch": 2167, "lr": 1.9353044056345475e-05}
{"train_loss": 0.1037868857383728, "global_step": 192878, "epoch": 2167, "lr": 1.935258599104254e-05}
{"train_loss": 0.04304531589150429, "global_step": 192879, "epoch": 2167, "lr": 1.9352127929859787e-05}
{"train_loss": 0.02381863258779049, "global_step": 192880, "epoch": 2167, "lr": 1.9351669872797272e-05}
{"train_loss": 0.06726888567209244, "global_step": 192881, "epoch": 2167, "lr": 1.9351211819855037e-05}
{"train_loss": 0.09751655906438828, "global_step": 192882, "epoch": 2167, "lr": 1.935075377103317e-05}
{"train_loss": 0.059405501931905746, "global_step": 192883, "epoch": 2167, "lr": 1.9350295726331707e-05}
{"train_loss": 0.06510686874389648, "global_step": 192884, "epoch": 2167, "lr": 1.934983768575074e-05}
{"train_loss": 0.05617779865860939, "global_step": 192885, "epoch": 2167, "lr": 1.9349379649290294e-05}
{"train_loss": 0.04455666244029999, "global_step": 192886, "epoch": 2167, "lr": 1.9348921616950472e-05}
{"train_loss": 0.0776711255311966, "global_step": 192887, "epoch": 2167, "lr": 1.9348463588731292e-05}
{"train_loss": 0.07716231793165207, "global_step": 192888, "epoch": 2167, "lr": 1.9348005564632853e-05}
{"train_loss": 0.05376088619232178, "global_step": 192889, "epoch": 2167, "lr": 1.9347547544655188e-05}
{"train_loss": 0.08793513476848602, "global_step": 192890, "epoch": 2167, "lr": 1.9347089528798373e-05}
{"train_loss": 0.017986638471484184, "global_step": 192891, "epoch": 2167, "lr": 1.934663151706248e-05}
{"train_loss": 0.07137823849916458, "global_step": 192892, "epoch": 2167, "lr": 1.9346173509447547e-05}
{"train_loss": 0.08834195882081985, "global_step": 192893, "epoch": 2167, "lr": 1.9345715505953658e-05}
{"train_loss": 0.10494314134120941, "global_step": 192894, "epoch": 2167, "lr": 1.9345257506580856e-05}
{"train_loss": 0.03945199400186539, "global_step": 192895, "epoch": 2167, "lr": 1.934479951132922e-05}
{"train_loss": 0.05771489068865776, "global_step": 192896, "epoch": 2167, "lr": 1.934434152019879e-05}
{"train_loss": 0.02242662012577057, "global_step": 192897, "epoch": 2167, "lr": 1.9343883533189656e-05}
{"train_loss": 0.08448009938001633, "global_step": 192898, "epoch": 2167, "lr": 1.9343425550301843e-05}
{"train_loss": 0.10242871195077896, "global_step": 192899, "epoch": 2167, "lr": 1.934296757153545e-05}
{"train_loss": 0.034206315875053406, "global_step": 192900, "epoch": 2167, "lr": 1.9342509596890506e-05}
{"train_loss": 0.035926368087530136, "global_step": 192901, "epoch": 2167, "lr": 1.9342051626367107e-05}
{"train_loss": 0.055324871093034744, "global_step": 192902, "epoch": 2167, "lr": 1.9341593659965272e-05}
{"train_loss": 0.0582243986427784, "global_step": 192903, "epoch": 2167, "lr": 1.934113569768511e-05}
{"train_loss": 0.04775416851043701, "global_step": 192904, "epoch": 2167, "lr": 1.934067773952664e-05}
{"train_loss": 0.06400419771671295, "global_step": 192905, "epoch": 2167, "lr": 1.9340219785489956e-05}
{"train_loss": 0.06416402757167816, "global_step": 192906, "epoch": 2167, "lr": 1.9339761835575093e-05}
{"train_loss": 0.09299977123737335, "global_step": 192907, "epoch": 2167, "lr": 1.933930388978214e-05}
{"train_loss": 0.06921051442623138, "global_step": 192908, "epoch": 2167, "lr": 1.933884594811113e-05}
{"train_loss": 0.07736771553754807, "global_step": 192909, "epoch": 2167, "lr": 1.933838801056213e-05}
{"train_loss": 0.03603949397802353, "global_step": 192910, "epoch": 2167, "lr": 1.9337930077135237e-05}
{"train_loss": 0.05122491344809532, "global_step": 192911, "epoch": 2167, "lr": 1.933747214783046e-05}
{"train_loss": 0.07652824372053146, "global_step": 192912, "epoch": 2167, "lr": 1.933701422264791e-05}
{"train_loss": 0.06678182631731033, "global_step": 192913, "epoch": 2167, "lr": 1.9336556301587623e-05}
{"train_loss": 0.09096905589103699, "global_step": 192914, "epoch": 2167, "lr": 1.9336098384649642e-05}
{"train_loss": 0.07388176769018173, "global_step": 192915, "epoch": 2167, "lr": 1.9335640471834067e-05}
{"train_loss": 0.07836561650037766, "global_step": 192916, "epoch": 2167, "lr": 1.9335182563140925e-05}
{"train_loss": 0.06543954461812973, "global_step": 192917, "epoch": 2167, "lr": 1.9334724658570314e-05}
{"train_loss": 0.05588570237159729, "global_step": 192918, "epoch": 2167, "lr": 1.9334266758122273e-05}
{"train_loss": 0.0594085231423378, "global_step": 192919, "epoch": 2167, "lr": 1.933380886179685e-05}
{"train_loss": 0.0604863204061985, "global_step": 192920, "epoch": 2167, "lr": 1.933335096959414e-05}
{"train_loss": 0.06643479317426682, "global_step": 192921, "epoch": 2167, "lr": 1.9332893081514165e-05}
{"train_loss": 0.09503812342882156, "global_step": 192922, "epoch": 2167, "lr": 1.9332435197557032e-05}
{"train_loss": 0.0671411082148552, "global_step": 192923, "epoch": 2167, "lr": 1.9331977317722754e-05}
{"train_loss": 0.09322863072156906, "global_step": 192924, "epoch": 2167, "lr": 1.9331519442011443e-05}
{"train_loss": 0.0844353660941124, "global_step": 192925, "epoch": 2167, "lr": 1.9331061570423115e-05}
{"train_loss": 0.031298864632844925, "global_step": 192926, "epoch": 2167, "lr": 1.9330603702957856e-05}
{"train_loss": 0.05714850872755051, "global_step": 192927, "epoch": 2167, "lr": 1.9330145839615737e-05}
{"train_loss": 0.06302367150783539, "global_step": 192928, "epoch": 2167, "lr": 1.9329687980396787e-05}
{"train_loss": 0.12593600153923035, "global_step": 192929, "epoch": 2167, "lr": 1.9329230125301106e-05}
{"train_loss": 0.051441166549921036, "global_step": 192930, "epoch": 2167, "lr": 1.932877227432872e-05}
{"train_loss": 0.05740264803171158, "global_step": 192931, "epoch": 2167, "lr": 1.9328314427479728e-05}
{"train_loss": 0.07846738398075104, "global_step": 192932, "epoch": 2167, "lr": 1.9327856584754143e-05}
{"train_loss": 0.07893868535757065, "global_step": 192933, "epoch": 2167, "lr": 1.932739874615208e-05}
{"train_loss": 0.02461427077651024, "global_step": 192934, "epoch": 2167, "lr": 1.932694091167355e-05}
{"train_loss": 0.09380611777305603, "global_step": 192935, "epoch": 2167, "lr": 1.9326483081318665e-05}
{"train_loss": 0.06450594961643219, "global_step": 192936, "epoch": 2167, "lr": 1.9326025255087437e-05}
{"train_loss": 0.07439583539962769, "global_step": 192937, "epoch": 2167, "lr": 1.932556743297997e-05}
{"train_loss": 0.07934414595365524, "global_step": 192938, "epoch": 2167, "lr": 1.9325109614996285e-05}
{"train_loss": 0.02486012503504753, "global_step": 192939, "epoch": 2167, "lr": 1.9324651801136488e-05}
{"train_loss": 0.04147278890013695, "global_step": 192940, "epoch": 2167, "lr": 1.9324193991400597e-05}
{"train_loss": 0.12248179316520691, "global_step": 192941, "epoch": 2167, "lr": 1.9323736185788716e-05}
{"train_loss": 0.04282868653535843, "global_step": 192942, "epoch": 2167, "lr": 1.9323278384300862e-05}
{"train_loss": 0.07187718898057938, "global_step": 192943, "epoch": 2167, "lr": 1.932282058693714e-05}
{"train_loss": 0.12457459419965744, "global_step": 192944, "epoch": 2167, "lr": 1.9322362793697572e-05}
{"train_loss": 0.0687805786728859, "global_step": 192945, "epoch": 2167, "lr": 1.932190500458224e-05}
{"train_loss": 0.06035827845335007, "global_step": 192946, "epoch": 2167, "lr": 1.932144721959122e-05}
{"train_loss": 0.07085511833429337, "global_step": 192947, "epoch": 2167, "lr": 1.932098943872454e-05}
{"train_loss": 0.07744374126195908, "global_step": 192948, "epoch": 2167, "lr": 1.93205316619823e-05}
{"train_loss": 0.04186653718352318, "global_step": 192949, "epoch": 2167, "lr": 1.9320073889364514e-05}
{"train_loss": 0.10633546113967896, "global_step": 192950, "epoch": 2167, "lr": 1.9319616120871298e-05}
{"train_loss": 0.06670710986501045, "global_step": 192951, "epoch": 2167, "lr": 1.931915835650268e-05, "val_loss": 7.809055805206299}
{"train_loss": 0.09627453237771988, "global_step": 192952, "epoch": 2168, "lr": 1.931870059625871e-05}
{"train_loss": 0.06971543282270432, "global_step": 192953, "epoch": 2168, "lr": 1.9318242840139487e-05}
{"train_loss": 0.10573134571313858, "global_step": 192954, "epoch": 2168, "lr": 1.9317785088145036e-05}
{"train_loss": 0.049115173518657684, "global_step": 192955, "epoch": 2168, "lr": 1.931732734027545e-05}
{"train_loss": 0.05499378591775894, "global_step": 192956, "epoch": 2168, "lr": 1.9316869596530773e-05}
{"train_loss": 0.058051466941833496, "global_step": 192957, "epoch": 2168, "lr": 1.9316411856911053e-05}
{"train_loss": 0.05127827078104019, "global_step": 192958, "epoch": 2168, "lr": 1.9315954121416386e-05}
{"train_loss": 0.06840824335813522, "global_step": 192959, "epoch": 2168, "lr": 1.9315496390046794e-05}
{"train_loss": 0.03687714412808418, "global_step": 192960, "epoch": 2168, "lr": 1.931503866280238e-05}
{"train_loss": 0.11472566425800323, "global_step": 192961, "epoch": 2168, "lr": 1.9314580939683168e-05}
{"train_loss": 0.08294795453548431, "global_step": 192962, "epoch": 2168, "lr": 1.9314123220689235e-05}
{"train_loss": 0.13884632289409637, "global_step": 192963, "epoch": 2168, "lr": 1.9313665505820665e-05}
{"train_loss": 0.05607634410262108, "global_step": 192964, "epoch": 2168, "lr": 1.9313207795077477e-05}
{"train_loss": 0.06524119526147842, "global_step": 192965, "epoch": 2168, "lr": 1.9312750088459773e-05}
{"train_loss": 0.08415482938289642, "global_step": 192966, "epoch": 2168, "lr": 1.9312292385967577e-05}
{"train_loss": 0.05591548606753349, "global_step": 192967, "epoch": 2168, "lr": 1.9311834687600987e-05}
{"train_loss": 0.043795302510261536, "global_step": 192968, "epoch": 2168, "lr": 1.9311376993360037e-05}
{"train_loss": 0.06949464976787567, "global_step": 192969, "epoch": 2168, "lr": 1.9310919303244806e-05}
{"train_loss": 0.01773325726389885, "global_step": 192970, "epoch": 2168, "lr": 1.9310461617255332e-05}
{"train_loss": 0.10959891974925995, "global_step": 192971, "epoch": 2168, "lr": 1.9310003935391712e-05}
{"train_loss": 0.048998698592185974, "global_step": 192972, "epoch": 2168, "lr": 1.930954625765397e-05}
{"train_loss": 0.06308504194021225, "global_step": 192973, "epoch": 2168, "lr": 1.9309088584042206e-05}
{"train_loss": 0.08142991364002228, "global_step": 192974, "epoch": 2168, "lr": 1.930863091455644e-05}
{"train_loss": 0.034404169768095016, "global_step": 192975, "epoch": 2168, "lr": 1.930817324919677e-05}
{"train_loss": 0.09572602808475494, "global_step": 192976, "epoch": 2168, "lr": 1.9307715587963227e-05}
{"train_loss": 0.08713621646165848, "global_step": 192977, "epoch": 2168, "lr": 1.930725793085591e-05}
{"train_loss": 0.024807758629322052, "global_step": 192978, "epoch": 2168, "lr": 1.9306800277874833e-05}
{"train_loss": 0.11577805876731873, "global_step": 192979, "epoch": 2168, "lr": 1.93063426290201e-05}
{"train_loss": 0.0590248629450798, "global_step": 192980, "epoch": 2168, "lr": 1.9305884984291743e-05}
{"train_loss": 0.08458089083433151, "global_step": 192981, "epoch": 2168, "lr": 1.930542734368983e-05}
{"train_loss": 0.15632890164852142, "global_step": 192982, "epoch": 2168, "lr": 1.9304969707214453e-05}
{"train_loss": 0.062400996685028076, "global_step": 192983, "epoch": 2168, "lr": 1.9304512074865628e-05}
{"train_loss": 0.06581409275531769, "global_step": 192984, "epoch": 2168, "lr": 1.930405444664345e-05}
{"train_loss": 0.048709411174058914, "global_step": 192985, "epoch": 2168, "lr": 1.9303596822547954e-05}
{"train_loss": 0.1289929747581482, "global_step": 192986, "epoch": 2168, "lr": 1.9303139202579234e-05}
{"train_loss": 0.06525228917598724, "global_step": 192987, "epoch": 2168, "lr": 1.9302681586737316e-05}
{"train_loss": 0.06131911650300026, "global_step": 192988, "epoch": 2168, "lr": 1.9302223975022295e-05}
{"train_loss": 0.10927677154541016, "global_step": 192989, "epoch": 2168, "lr": 1.930176636743421e-05}
{"train_loss": 0.04119105264544487, "global_step": 192990, "epoch": 2168, "lr": 1.9301308763973113e-05}
{"train_loss": 0.10779345780611038, "global_step": 192991, "epoch": 2168, "lr": 1.9300851164639105e-05}
{"train_loss": 0.09428514540195465, "global_step": 192992, "epoch": 2168, "lr": 1.93003935694322e-05}
{"train_loss": 0.05689051002264023, "global_step": 192993, "epoch": 2168, "lr": 1.92999359783525e-05}
{"train_loss": 0.055013060569763184, "global_step": 192994, "epoch": 2168, "lr": 1.9299478391400037e-05}
{"train_loss": 0.06307218968868256, "global_step": 192995, "epoch": 2168, "lr": 1.92990208085749e-05}
{"train_loss": 0.03809961676597595, "global_step": 192996, "epoch": 2168, "lr": 1.9298563229877136e-05}
{"train_loss": 0.10656473785638809, "global_step": 192997, "epoch": 2168, "lr": 1.9298105655306787e-05}
{"train_loss": 0.05925518274307251, "global_step": 192998, "epoch": 2168, "lr": 1.9297648084863934e-05}
{"train_loss": 0.07247725129127502, "global_step": 192999, "epoch": 2168, "lr": 1.9297190518548652e-05}
{"train_loss": 0.07758881151676178, "global_step": 193000, "epoch": 2168, "lr": 1.9296732956360976e-05}
{"train_loss": 0.12791801989078522, "global_step": 193001, "epoch": 2168, "lr": 1.9296275398300996e-05}
{"train_loss": 0.059044647961854935, "global_step": 193002, "epoch": 2168, "lr": 1.929581784436874e-05}
{"train_loss": 0.059429142624139786, "global_step": 193003, "epoch": 2168, "lr": 1.929536029456431e-05}
{"train_loss": 0.10939183831214905, "global_step": 193004, "epoch": 2168, "lr": 1.9294902748887717e-05}
{"train_loss": 0.11481387913227081, "global_step": 193005, "epoch": 2168, "lr": 1.9294445207339072e-05}
{"train_loss": 0.07358024269342422, "global_step": 193006, "epoch": 2168, "lr": 1.9293987669918395e-05}
{"train_loss": 0.07985550910234451, "global_step": 193007, "epoch": 2168, "lr": 1.9293530136625787e-05}
{"train_loss": 0.07646048069000244, "global_step": 193008, "epoch": 2168, "lr": 1.9293072607461276e-05}
{"train_loss": 0.14158058166503906, "global_step": 193009, "epoch": 2168, "lr": 1.929261508242495e-05}
{"train_loss": 0.13135641813278198, "global_step": 193010, "epoch": 2168, "lr": 1.9292157561516837e-05}
{"train_loss": 0.028107138350605965, "global_step": 193011, "epoch": 2168, "lr": 1.9291700044737045e-05}
{"train_loss": 0.08018148690462112, "global_step": 193012, "epoch": 2168, "lr": 1.9291242532085587e-05}
{"train_loss": 0.10530470311641693, "global_step": 193013, "epoch": 2168, "lr": 1.9290785023562563e-05}
{"train_loss": 0.10341902822256088, "global_step": 193014, "epoch": 2168, "lr": 1.9290327519168006e-05}
{"train_loss": 0.06694892793893814, "global_step": 193015, "epoch": 2168, "lr": 1.9289870018902005e-05}
{"train_loss": 0.08527164906263351, "global_step": 193016, "epoch": 2168, "lr": 1.928941252276459e-05}
{"train_loss": 0.07047776877880096, "global_step": 193017, "epoch": 2168, "lr": 1.9288955030755835e-05}
{"train_loss": 0.07130613923072815, "global_step": 193018, "epoch": 2168, "lr": 1.928849754287583e-05}
{"train_loss": 0.12297720462083817, "global_step": 193019, "epoch": 2168, "lr": 1.928804005912459e-05}
{"train_loss": 0.08608803153038025, "global_step": 193020, "epoch": 2168, "lr": 1.9287582579502218e-05}
{"train_loss": 0.03399059921503067, "global_step": 193021, "epoch": 2168, "lr": 1.9287125104008736e-05}
{"train_loss": 0.060836296528577805, "global_step": 193022, "epoch": 2168, "lr": 1.9286667632644245e-05}
{"train_loss": 0.036754392087459564, "global_step": 193023, "epoch": 2168, "lr": 1.928621016540877e-05}
{"train_loss": 0.03512990102171898, "global_step": 193024, "epoch": 2168, "lr": 1.9285752702302406e-05}
{"train_loss": 0.04753732308745384, "global_step": 193025, "epoch": 2168, "lr": 1.9285295243325186e-05}
{"train_loss": 0.06612879782915115, "global_step": 193026, "epoch": 2168, "lr": 1.9284837788477194e-05}
{"train_loss": 0.05183282494544983, "global_step": 193027, "epoch": 2168, "lr": 1.9284380337758467e-05}
{"train_loss": 0.06786300241947174, "global_step": 193028, "epoch": 2168, "lr": 1.9283922891169097e-05}
{"train_loss": 0.0675462931394577, "global_step": 193029, "epoch": 2168, "lr": 1.928346544870913e-05}
{"train_loss": 0.07944004982709885, "global_step": 193030, "epoch": 2168, "lr": 1.928300801037861e-05}
{"train_loss": 0.06053014472126961, "global_step": 193031, "epoch": 2168, "lr": 1.9282550576177632e-05}
{"train_loss": 0.0440489836037159, "global_step": 193032, "epoch": 2168, "lr": 1.9282093146106228e-05}
{"train_loss": 0.06905770301818848, "global_step": 193033, "epoch": 2168, "lr": 1.9281635720164482e-05}
{"train_loss": 0.03981202095746994, "global_step": 193034, "epoch": 2168, "lr": 1.928117829835245e-05}
{"train_loss": 0.06240427866578102, "global_step": 193035, "epoch": 2168, "lr": 1.9280720880670172e-05}
{"train_loss": 0.06167558953166008, "global_step": 193036, "epoch": 2168, "lr": 1.928026346711772e-05}
{"train_loss": 0.10111332684755325, "global_step": 193037, "epoch": 2168, "lr": 1.9279806057695186e-05}
{"train_loss": 0.07154004275798798, "global_step": 193038, "epoch": 2168, "lr": 1.927934865240259e-05}
{"train_loss": 0.05389158055186272, "global_step": 193039, "epoch": 2168, "lr": 1.9278891251240026e-05}
{"train_loss": 0.07494209641904644, "global_step": 193040, "epoch": 2168, "lr": 1.9278433854207528e-05, "val_loss": 7.852725505828857}
{"train_loss": 0.07638825476169586, "global_step": 193041, "epoch": 2169, "lr": 1.927797646130518e-05}
{"train_loss": 0.08885956555604935, "global_step": 193042, "epoch": 2169, "lr": 1.9277519072533022e-05}
{"train_loss": 0.06948274374008179, "global_step": 193043, "epoch": 2169, "lr": 1.9277061687891146e-05}
{"train_loss": 0.08185764402151108, "global_step": 193044, "epoch": 2169, "lr": 1.927660430737957e-05}
{"train_loss": 0.09197023510932922, "global_step": 193045, "epoch": 2169, "lr": 1.92761469309984e-05}
{"train_loss": 0.06977396458387375, "global_step": 193046, "epoch": 2169, "lr": 1.9275689558747668e-05}
{"train_loss": 0.06865596026182175, "global_step": 193047, "epoch": 2169, "lr": 1.927523219062745e-05}
{"train_loss": 0.05175619199872017, "global_step": 193048, "epoch": 2169, "lr": 1.9274774826637793e-05}
{"train_loss": 0.10210880637168884, "global_step": 193049, "epoch": 2169, "lr": 1.9274317466778786e-05}
{"train_loss": 0.0900830328464508, "global_step": 193050, "epoch": 2169, "lr": 1.927386011105045e-05}
{"train_loss": 0.055620912462472916, "global_step": 193051, "epoch": 2169, "lr": 1.927340275945289e-05}
{"train_loss": 0.08485621213912964, "global_step": 193052, "epoch": 2169, "lr": 1.9272945411986125e-05}
{"train_loss": 0.04310625419020653, "global_step": 193053, "epoch": 2169, "lr": 1.9272488068650248e-05}
{"train_loss": 0.06972965598106384, "global_step": 193054, "epoch": 2169, "lr": 1.9272030729445316e-05}
{"train_loss": 0.0795530453324318, "global_step": 193055, "epoch": 2169, "lr": 1.9271573394371374e-05}
{"train_loss": 0.09277724474668503, "global_step": 193056, "epoch": 2169, "lr": 1.9271116063428513e-05}
{"train_loss": 0.06169898808002472, "global_step": 193057, "epoch": 2169, "lr": 1.9270658736616755e-05}
{"train_loss": 0.20858734846115112, "global_step": 193058, "epoch": 2169, "lr": 1.9270201413936202e-05}
{"train_loss": 0.11438379436731339, "global_step": 193059, "epoch": 2169, "lr": 1.9269744095386877e-05}
{"train_loss": 0.04670845717191696, "global_step": 193060, "epoch": 2169, "lr": 1.9269286780968877e-05}
{"train_loss": 0.07457096129655838, "global_step": 193061, "epoch": 2169, "lr": 1.9268829470682233e-05}
{"train_loss": 0.07956434041261673, "global_step": 193062, "epoch": 2169, "lr": 1.9268372164527033e-05}
{"train_loss": 0.10001358389854431, "global_step": 193063, "epoch": 2169, "lr": 1.926791486250331e-05}
{"train_loss": 0.09860706329345703, "global_step": 193064, "epoch": 2169, "lr": 1.9267457564611157e-05}
{"train_loss": 0.03313877806067467, "global_step": 193065, "epoch": 2169, "lr": 1.926700027085061e-05}
{"train_loss": 0.12174008786678314, "global_step": 193066, "epoch": 2169, "lr": 1.926654298122175e-05}
{"train_loss": 0.05031558871269226, "global_step": 193067, "epoch": 2169, "lr": 1.9266085695724634e-05}
{"train_loss": 0.04585467651486397, "global_step": 193068, "epoch": 2169, "lr": 1.9265628414359293e-05}
{"train_loss": 0.05257368087768555, "global_step": 193069, "epoch": 2169, "lr": 1.9265171137125836e-05}
{"train_loss": 0.04678169637918472, "global_step": 193070, "epoch": 2169, "lr": 1.9264713864024287e-05}
{"train_loss": 0.059073306620121, "global_step": 193071, "epoch": 2169, "lr": 1.9264256595054738e-05}
{"train_loss": 0.05254619941115379, "global_step": 193072, "epoch": 2169, "lr": 1.9263799330217218e-05}
{"train_loss": 0.02873123809695244, "global_step": 193073, "epoch": 2169, "lr": 1.926334206951182e-05}
{"train_loss": 0.05965816229581833, "global_step": 193074, "epoch": 2169, "lr": 1.9262884812938575e-05}
{"train_loss": 0.10507849603891373, "global_step": 193075, "epoch": 2169, "lr": 1.9262427560497582e-05}
{"train_loss": 0.07671599090099335, "global_step": 193076, "epoch": 2169, "lr": 1.926197031218886e-05}
{"train_loss": 0.038902293890714645, "global_step": 193077, "epoch": 2169, "lr": 1.926151306801251e-05}
{"train_loss": 0.034938495606184006, "global_step": 193078, "epoch": 2169, "lr": 1.9261055827968554e-05}
{"train_loss": 0.08422734588384628, "global_step": 193079, "epoch": 2169, "lr": 1.9260598592057095e-05}
{"train_loss": 0.06008297577500343, "global_step": 193080, "epoch": 2169, "lr": 1.9260141360278157e-05}
{"train_loss": 0.02524331957101822, "global_step": 193081, "epoch": 2169, "lr": 1.9259684132631835e-05}
{"train_loss": 0.03971567377448082, "global_step": 193082, "epoch": 2169, "lr": 1.9259226909118156e-05}
{"train_loss": 0.07222903519868851, "global_step": 193083, "epoch": 2169, "lr": 1.9258769689737216e-05}
{"train_loss": 0.08175200968980789, "global_step": 193084, "epoch": 2169, "lr": 1.9258312474489044e-05}
{"train_loss": 0.0474427193403244, "global_step": 193085, "epoch": 2169, "lr": 1.9257855263373732e-05}
{"train_loss": 0.053706005215644836, "global_step": 193086, "epoch": 2169, "lr": 1.9257398056391314e-05}
{"train_loss": 0.01578754372894764, "global_step": 193087, "epoch": 2169, "lr": 1.925694085354188e-05}
{"train_loss": 0.029429560527205467, "global_step": 193088, "epoch": 2169, "lr": 1.9256483654825452e-05}
{"train_loss": 0.04164019599556923, "global_step": 193089, "epoch": 2169, "lr": 1.9256026460242122e-05}
{"train_loss": 0.10492384433746338, "global_step": 193090, "epoch": 2169, "lr": 1.9255569269791958e-05}
{"train_loss": 0.0885099470615387, "global_step": 193091, "epoch": 2169, "lr": 1.925511208347499e-05}
{"train_loss": 0.05029980465769768, "global_step": 193092, "epoch": 2169, "lr": 1.9254654901291318e-05}
{"train_loss": 0.06468154489994049, "global_step": 193093, "epoch": 2169, "lr": 1.9254197723240967e-05}
{"train_loss": 0.049330443143844604, "global_step": 193094, "epoch": 2169, "lr": 1.925374054932403e-05}
{"train_loss": 0.07366481423377991, "global_step": 193095, "epoch": 2169, "lr": 1.9253283379540533e-05}
{"train_loss": 0.08890040218830109, "global_step": 193096, "epoch": 2169, "lr": 1.925282621389058e-05}
{"train_loss": 0.12160803377628326, "global_step": 193097, "epoch": 2169, "lr": 1.9252369052374186e-05}
{"train_loss": 0.10325369983911514, "global_step": 193098, "epoch": 2169, "lr": 1.925191189499146e-05}
{"train_loss": 0.06241247057914734, "global_step": 193099, "epoch": 2169, "lr": 1.9251454741742413e-05}
{"train_loss": 0.031750328838825226, "global_step": 193100, "epoch": 2169, "lr": 1.9250997592627158e-05}
{"train_loss": 0.07910730689764023, "global_step": 193101, "epoch": 2169, "lr": 1.9250540447645716e-05}
{"train_loss": 0.029220061376690865, "global_step": 193102, "epoch": 2169, "lr": 1.9250083306798177e-05}
{"train_loss": 0.07711166888475418, "global_step": 193103, "epoch": 2169, "lr": 1.9249626170084572e-05}
{"train_loss": 0.13968110084533691, "global_step": 193104, "epoch": 2169, "lr": 1.9249169037505e-05}
{"train_loss": 0.04225768521428108, "global_step": 193105, "epoch": 2169, "lr": 1.924871190905948e-05}
{"train_loss": 0.051026295870542526, "global_step": 193106, "epoch": 2169, "lr": 1.9248254784748122e-05}
{"train_loss": 0.04902723431587219, "global_step": 193107, "epoch": 2169, "lr": 1.9247797664570955e-05}
{"train_loss": 0.04638788849115372, "global_step": 193108, "epoch": 2169, "lr": 1.9247340548528027e-05}
{"train_loss": 0.06781771034002304, "global_step": 193109, "epoch": 2169, "lr": 1.924688343661944e-05}
{"train_loss": 0.04164184629917145, "global_step": 193110, "epoch": 2169, "lr": 1.9246426328845218e-05}
{"train_loss": 0.05881888046860695, "global_step": 193111, "epoch": 2169, "lr": 1.9245969225205458e-05}
{"train_loss": 0.07511100172996521, "global_step": 193112, "epoch": 2169, "lr": 1.924551212570018e-05}
{"train_loss": 0.04923796281218529, "global_step": 193113, "epoch": 2169, "lr": 1.924505503032949e-05}
{"train_loss": 0.04990202933549881, "global_step": 193114, "epoch": 2169, "lr": 1.924459793909341e-05}
{"train_loss": 0.09204406291246414, "global_step": 193115, "epoch": 2169, "lr": 1.924414085199203e-05}
{"train_loss": 0.04998699948191643, "global_step": 193116, "epoch": 2169, "lr": 1.9243683769025394e-05}
{"train_loss": 0.0789979100227356, "global_step": 193117, "epoch": 2169, "lr": 1.924322669019358e-05}
{"train_loss": 0.07205178588628769, "global_step": 193118, "epoch": 2169, "lr": 1.9242769615496627e-05}
{"train_loss": 0.06355510652065277, "global_step": 193119, "epoch": 2169, "lr": 1.924231254493462e-05}
{"train_loss": 0.031363535672426224, "global_step": 193120, "epoch": 2169, "lr": 1.9241855478507594e-05}
{"train_loss": 0.058561377227306366, "global_step": 193121, "epoch": 2169, "lr": 1.9241398416215646e-05}
{"train_loss": 0.048435430973768234, "global_step": 193122, "epoch": 2169, "lr": 1.9240941358058794e-05}
{"train_loss": 0.0851278081536293, "global_step": 193123, "epoch": 2169, "lr": 1.9240484304037143e-05}
{"train_loss": 0.023456981405615807, "global_step": 193124, "epoch": 2169, "lr": 1.9240027254150717e-05}
{"train_loss": 0.07390783727169037, "global_step": 193125, "epoch": 2169, "lr": 1.9239570208399594e-05}
{"train_loss": 0.09732850641012192, "global_step": 193126, "epoch": 2169, "lr": 1.923911316678385e-05}
{"train_loss": 0.07780272513628006, "global_step": 193127, "epoch": 2169, "lr": 1.923865612930352e-05}
{"train_loss": 0.09276030212640762, "global_step": 193128, "epoch": 2169, "lr": 1.923819909595869e-05}
{"train_loss": 0.06846507320578178, "global_step": 193129, "epoch": 2169, "lr": 1.9237742066749397e-05, "val_loss": 7.829089164733887}
{"train_loss": 0.08581434190273285, "global_step": 193130, "epoch": 2170, "lr": 1.9237285041675727e-05}
{"train_loss": 0.08847454190254211, "global_step": 193131, "epoch": 2170, "lr": 1.923682802073772e-05}
{"train_loss": 0.06023038923740387, "global_step": 193132, "epoch": 2170, "lr": 1.9236371003935456e-05}
{"train_loss": 0.0773775652050972, "global_step": 193133, "epoch": 2170, "lr": 1.9235913991268972e-05}
{"train_loss": 0.11617151647806168, "global_step": 193134, "epoch": 2170, "lr": 1.9235456982738364e-05}
{"train_loss": 0.06770136207342148, "global_step": 193135, "epoch": 2170, "lr": 1.9234999978343652e-05}
{"train_loss": 0.05531786009669304, "global_step": 193136, "epoch": 2170, "lr": 1.923454297808494e-05}
{"train_loss": 0.11452095210552216, "global_step": 193137, "epoch": 2170, "lr": 1.923408598196225e-05}
{"train_loss": 0.047890182584524155, "global_step": 193138, "epoch": 2170, "lr": 1.923362898997568e-05}
{"train_loss": 0.07651402801275253, "global_step": 193139, "epoch": 2170, "lr": 1.9233172002125254e-05}
{"train_loss": 0.07328244298696518, "global_step": 193140, "epoch": 2170, "lr": 1.9232715018411075e-05}
{"train_loss": 0.041089363396167755, "global_step": 193141, "epoch": 2170, "lr": 1.923225803883316e-05}
{"train_loss": 0.05308590084314346, "global_step": 193142, "epoch": 2170, "lr": 1.9231801063391614e-05}
{"train_loss": 0.05633435770869255, "global_step": 193143, "epoch": 2170, "lr": 1.923134409208646e-05}
{"train_loss": 0.08869870752096176, "global_step": 193144, "epoch": 2170, "lr": 1.9230887124917797e-05}
{"train_loss": 0.054747894406318665, "global_step": 193145, "epoch": 2170, "lr": 1.923043016188566e-05}
{"train_loss": 0.07001209259033203, "global_step": 193146, "epoch": 2170, "lr": 1.9229973202990104e-05}
{"train_loss": 0.06761088222265244, "global_step": 193147, "epoch": 2170, "lr": 1.9229516248231218e-05}
{"train_loss": 0.09308646619319916, "global_step": 193148, "epoch": 2170, "lr": 1.9229059297609027e-05}
{"train_loss": 0.07815171778202057, "global_step": 193149, "epoch": 2170, "lr": 1.9228602351123637e-05}
{"train_loss": 0.084624744951725, "global_step": 193150, "epoch": 2170, "lr": 1.9228145408775067e-05}
{"train_loss": 0.0428851954638958, "global_step": 193151, "epoch": 2170, "lr": 1.9227688470563414e-05}
{"train_loss": 0.055481620132923126, "global_step": 193152, "epoch": 2170, "lr": 1.922723153648871e-05}
{"train_loss": 0.038958046585321426, "global_step": 193153, "epoch": 2170, "lr": 1.9226774606551046e-05}
{"train_loss": 0.05869890749454498, "global_step": 193154, "epoch": 2170, "lr": 1.9226317680750445e-05}
{"train_loss": 0.07172518223524094, "global_step": 193155, "epoch": 2170, "lr": 1.922586075908701e-05}
{"train_loss": 0.0592912919819355, "global_step": 193156, "epoch": 2170, "lr": 1.9225403841560768e-05}
{"train_loss": 0.04610337316989899, "global_step": 193157, "epoch": 2170, "lr": 1.922494692817181e-05}
{"train_loss": 0.09077253192663193, "global_step": 193158, "epoch": 2170, "lr": 1.9224490018920166e-05}
{"train_loss": 0.03631603345274925, "global_step": 193159, "epoch": 2170, "lr": 1.9224033113805935e-05}
{"train_loss": 0.058089692145586014, "global_step": 193160, "epoch": 2170, "lr": 1.9223576212829135e-05}
{"train_loss": 0.14571703970432281, "global_step": 193161, "epoch": 2170, "lr": 1.9223119315989874e-05}
{"train_loss": 0.0404512882232666, "global_step": 193162, "epoch": 2170, "lr": 1.922266242328817e-05}
{"train_loss": 0.04774262383580208, "global_step": 193163, "epoch": 2170, "lr": 1.9222205534724097e-05}
{"train_loss": 0.04361608624458313, "global_step": 193164, "epoch": 2170, "lr": 1.922174865029775e-05}
{"train_loss": 0.03287477418780327, "global_step": 193165, "epoch": 2170, "lr": 1.9221291770009135e-05}
{"train_loss": 0.05636608973145485, "global_step": 193166, "epoch": 2170, "lr": 1.922083489385837e-05}
{"train_loss": 0.05701703578233719, "global_step": 193167, "epoch": 2170, "lr": 1.922037802184547e-05}
{"train_loss": 0.08714596927165985, "global_step": 193168, "epoch": 2170, "lr": 1.9219921153970528e-05}
{"train_loss": 0.08559796214103699, "global_step": 193169, "epoch": 2170, "lr": 1.921946429023358e-05}
{"train_loss": 0.08383981138467789, "global_step": 193170, "epoch": 2170, "lr": 1.921900743063471e-05}
{"train_loss": 0.07148642838001251, "global_step": 193171, "epoch": 2170, "lr": 1.921855057517396e-05}
{"train_loss": 0.0578237846493721, "global_step": 193172, "epoch": 2170, "lr": 1.9218093723851415e-05}
{"train_loss": 0.03878525272011757, "global_step": 193173, "epoch": 2170, "lr": 1.9217636876667105e-05}
{"train_loss": 0.060366854071617126, "global_step": 193174, "epoch": 2170, "lr": 1.9217180033621124e-05}
{"train_loss": 0.05459984391927719, "global_step": 193175, "epoch": 2170, "lr": 1.9216723194713503e-05}
{"train_loss": 0.08284398913383484, "global_step": 193176, "epoch": 2170, "lr": 1.9216266359944335e-05}
{"train_loss": 0.12806013226509094, "global_step": 193177, "epoch": 2170, "lr": 1.9215809529313644e-05}
{"train_loss": 0.08975503593683243, "global_step": 193178, "epoch": 2170, "lr": 1.9215352702821533e-05}
{"train_loss": 0.09388156980276108, "global_step": 193179, "epoch": 2170, "lr": 1.9214895880468025e-05}
{"train_loss": 0.054288093000650406, "global_step": 193180, "epoch": 2170, "lr": 1.921443906225322e-05}
{"train_loss": 0.07764586806297302, "global_step": 193181, "epoch": 2170, "lr": 1.9213982248177137e-05}
{"train_loss": 0.09599749743938446, "global_step": 193182, "epoch": 2170, "lr": 1.9213525438239878e-05}
{"train_loss": 0.044017717242240906, "global_step": 193183, "epoch": 2170, "lr": 1.9213068632441477e-05}
{"train_loss": 0.031554680317640305, "global_step": 193184, "epoch": 2170, "lr": 1.9212611830781996e-05}
{"train_loss": 0.04845793917775154, "global_step": 193185, "epoch": 2170, "lr": 1.921215503326152e-05}
{"train_loss": 0.07058367133140564, "global_step": 193186, "epoch": 2170, "lr": 1.921169823988008e-05}
{"train_loss": 0.03418200463056564, "global_step": 193187, "epoch": 2170, "lr": 1.9211241450637763e-05}
{"train_loss": 0.06376421451568604, "global_step": 193188, "epoch": 2170, "lr": 1.9210784665534602e-05}
{"train_loss": 0.04084504768252373, "global_step": 193189, "epoch": 2170, "lr": 1.9210327884570695e-05}
{"train_loss": 0.11511262506246567, "global_step": 193190, "epoch": 2170, "lr": 1.9209871107746065e-05}
{"train_loss": 0.08393803983926773, "global_step": 193191, "epoch": 2170, "lr": 1.920941433506081e-05}
{"train_loss": 0.03939461708068848, "global_step": 193192, "epoch": 2170, "lr": 1.9208957566514962e-05}
{"train_loss": 0.0689273402094841, "global_step": 193193, "epoch": 2170, "lr": 1.9208500802108608e-05}
{"train_loss": 0.07008746266365051, "global_step": 193194, "epoch": 2170, "lr": 1.9208044041841778e-05}
{"train_loss": 0.06200054660439491, "global_step": 193195, "epoch": 2170, "lr": 1.9207587285714567e-05}
{"train_loss": 0.05375562235713005, "global_step": 193196, "epoch": 2170, "lr": 1.9207130533727007e-05}
{"train_loss": 0.06766236573457718, "global_step": 193197, "epoch": 2170, "lr": 1.9206673785879188e-05}
{"train_loss": 0.1155647337436676, "global_step": 193198, "epoch": 2170, "lr": 1.9206217042171144e-05}
{"train_loss": 0.07017074525356293, "global_step": 193199, "epoch": 2170, "lr": 1.9205760302602943e-05}
{"train_loss": 0.08648384362459183, "global_step": 193200, "epoch": 2170, "lr": 1.920530356717467e-05}
{"train_loss": 0.06301703304052353, "global_step": 193201, "epoch": 2170, "lr": 1.920484683588635e-05}
{"train_loss": 0.04553782567381859, "global_step": 193202, "epoch": 2170, "lr": 1.9204390108738086e-05}
{"train_loss": 0.06306397914886475, "global_step": 193203, "epoch": 2170, "lr": 1.9203933385729895e-05}
{"train_loss": 0.08355404436588287, "global_step": 193204, "epoch": 2170, "lr": 1.9203476666861875e-05}
{"train_loss": 0.0924578532576561, "global_step": 193205, "epoch": 2170, "lr": 1.920301995213406e-05}
{"train_loss": 0.06412365287542343, "global_step": 193206, "epoch": 2170, "lr": 1.920256324154654e-05}
{"train_loss": 0.1294994056224823, "global_step": 193207, "epoch": 2170, "lr": 1.920210653509934e-05}
{"train_loss": 0.1399405151605606, "global_step": 193208, "epoch": 2170, "lr": 1.9201649832792558e-05}
{"train_loss": 0.07433268427848816, "global_step": 193209, "epoch": 2170, "lr": 1.9201193134626223e-05}
{"train_loss": 0.06152725592255592, "global_step": 193210, "epoch": 2170, "lr": 1.920073644060043e-05}
{"train_loss": 0.10333116352558136, "global_step": 193211, "epoch": 2170, "lr": 1.9200279750715207e-05}
{"train_loss": 0.0788135901093483, "global_step": 193212, "epoch": 2170, "lr": 1.9199823064970644e-05}
{"train_loss": 0.07812632620334625, "global_step": 193213, "epoch": 2170, "lr": 1.9199366383366778e-05}
{"train_loss": 0.057367682456970215, "global_step": 193214, "epoch": 2170, "lr": 1.9198909705903694e-05}
{"train_loss": 0.08673799782991409, "global_step": 193215, "epoch": 2170, "lr": 1.9198453032581425e-05}
{"train_loss": 0.06866692006587982, "global_step": 193216, "epoch": 2170, "lr": 1.9197996363400066e-05}
{"train_loss": 0.06997157633304596, "global_step": 193217, "epoch": 2170, "lr": 1.9197539698359644e-05}
{"train_loss": 0.07049950823355257, "global_step": 193218, "epoch": 2170, "lr": 1.9197083037460255e-05, "val_loss": 7.923704624176025, "train_action_mse_error": 8.387246131896973}
{"train_loss": 0.025289883837103844, "global_step": 193219, "epoch": 2171, "lr": 1.9196626380701926e-05}
{"train_loss": 0.0385727696120739, "global_step": 193220, "epoch": 2171, "lr": 1.919616972808475e-05}
{"train_loss": 0.06446247547864914, "global_step": 193221, "epoch": 2171, "lr": 1.9195713079608763e-05}
{"train_loss": 0.04415784031152725, "global_step": 193222, "epoch": 2171, "lr": 1.9195256435274045e-05}
{"train_loss": 0.0883711650967598, "global_step": 193223, "epoch": 2171, "lr": 1.9194799795080654e-05}
{"train_loss": 0.11710003763437271, "global_step": 193224, "epoch": 2171, "lr": 1.919434315902863e-05}
{"train_loss": 0.08430670201778412, "global_step": 193225, "epoch": 2171, "lr": 1.9193886527118067e-05}
{"train_loss": 0.05252905562520027, "global_step": 193226, "epoch": 2171, "lr": 1.9193429899348993e-05}
{"train_loss": 0.04881998151540756, "global_step": 193227, "epoch": 2171, "lr": 1.9192973275721503e-05}
{"train_loss": 0.03370993956923485, "global_step": 193228, "epoch": 2171, "lr": 1.9192516656235632e-05}
{"train_loss": 0.06279953569173813, "global_step": 193229, "epoch": 2171, "lr": 1.9192060040891464e-05}
{"train_loss": 0.06650238484144211, "global_step": 193230, "epoch": 2171, "lr": 1.9191603429689026e-05}
{"train_loss": 0.0702112466096878, "global_step": 193231, "epoch": 2171, "lr": 1.9191146822628426e-05}
{"train_loss": 0.06038103997707367, "global_step": 193232, "epoch": 2171, "lr": 1.919069021970968e-05}
{"train_loss": 0.023953044787049294, "global_step": 193233, "epoch": 2171, "lr": 1.9190233620932886e-05}
{"train_loss": 0.07176391780376434, "global_step": 193234, "epoch": 2171, "lr": 1.9189777026298073e-05}
{"train_loss": 0.10770120471715927, "global_step": 193235, "epoch": 2171, "lr": 1.918932043580532e-05}
{"train_loss": 0.06320829689502716, "global_step": 193236, "epoch": 2171, "lr": 1.9188863849454707e-05}
{"train_loss": 0.05805985629558563, "global_step": 193237, "epoch": 2171, "lr": 1.9188407267246257e-05}
{"train_loss": 0.12591645121574402, "global_step": 193238, "epoch": 2171, "lr": 1.9187950689180067e-05}
{"train_loss": 0.04940323159098625, "global_step": 193239, "epoch": 2171, "lr": 1.9187494115256162e-05}
{"train_loss": 0.02642274834215641, "global_step": 193240, "epoch": 2171, "lr": 1.918703754547464e-05}
{"train_loss": 0.04759099707007408, "global_step": 193241, "epoch": 2171, "lr": 1.9186580979835532e-05}
{"train_loss": 0.07430937886238098, "global_step": 193242, "epoch": 2171, "lr": 1.918612441833893e-05}
{"train_loss": 0.04408269375562668, "global_step": 193243, "epoch": 2171, "lr": 1.918566786098486e-05}
{"train_loss": 0.07782105356454849, "global_step": 193244, "epoch": 2171, "lr": 1.9185211307773417e-05}
{"train_loss": 0.0942557230591774, "global_step": 193245, "epoch": 2171, "lr": 1.9184754758704626e-05}
{"train_loss": 0.027915358543395996, "global_step": 193246, "epoch": 2171, "lr": 1.918429821377859e-05}
{"train_loss": 0.03815769404172897, "global_step": 193247, "epoch": 2171, "lr": 1.9183841672995334e-05}
{"train_loss": 0.08124452829360962, "global_step": 193248, "epoch": 2171, "lr": 1.9183385136354946e-05}
{"train_loss": 0.07343751192092896, "global_step": 193249, "epoch": 2171, "lr": 1.918292860385747e-05}
{"train_loss": 0.04743674769997597, "global_step": 193250, "epoch": 2171, "lr": 1.9182472075502984e-05}
{"train_loss": 0.09695693850517273, "global_step": 193251, "epoch": 2171, "lr": 1.918201555129152e-05}
{"train_loss": 0.04192332178354263, "global_step": 193252, "epoch": 2171, "lr": 1.9181559031223163e-05}
{"train_loss": 0.09614694863557816, "global_step": 193253, "epoch": 2171, "lr": 1.9181102515297983e-05}
{"train_loss": 0.08581489324569702, "global_step": 193254, "epoch": 2171, "lr": 1.9180646003516013e-05}
{"train_loss": 0.09223758429288864, "global_step": 193255, "epoch": 2171, "lr": 1.9180189495877343e-05}
{"train_loss": 0.036679696291685104, "global_step": 193256, "epoch": 2171, "lr": 1.9179732992382025e-05}
{"train_loss": 0.05593319609761238, "global_step": 193257, "epoch": 2171, "lr": 1.91792764930301e-05}
{"train_loss": 0.07285516709089279, "global_step": 193258, "epoch": 2171, "lr": 1.9178819997821663e-05}
{"train_loss": 0.09258664399385452, "global_step": 193259, "epoch": 2171, "lr": 1.9178363506756737e-05}
{"train_loss": 0.07128550112247467, "global_step": 193260, "epoch": 2171, "lr": 1.9177907019835422e-05}
{"train_loss": 0.07727805525064468, "global_step": 193261, "epoch": 2171, "lr": 1.917745053705776e-05}
{"train_loss": 0.07057678699493408, "global_step": 193262, "epoch": 2171, "lr": 1.9176994058423803e-05}
{"train_loss": 0.0727255716919899, "global_step": 193263, "epoch": 2171, "lr": 1.917653758393364e-05}
{"train_loss": 0.03132695332169533, "global_step": 193264, "epoch": 2171, "lr": 1.9176081113587297e-05}
{"train_loss": 0.04601510241627693, "global_step": 193265, "epoch": 2171, "lr": 1.917562464738487e-05}
{"train_loss": 0.09464826434850693, "global_step": 193266, "epoch": 2171, "lr": 1.9175168185326385e-05}
{"train_loss": 0.10324118286371231, "global_step": 193267, "epoch": 2171, "lr": 1.9174711727411944e-05}
{"train_loss": 0.06722275912761688, "global_step": 193268, "epoch": 2171, "lr": 1.9174255273641573e-05}
{"train_loss": 0.04707726836204529, "global_step": 193269, "epoch": 2171, "lr": 1.917379882401536e-05}
{"train_loss": 0.07595153898000717, "global_step": 193270, "epoch": 2171, "lr": 1.917334237853334e-05}
{"train_loss": 0.0865955576300621, "global_step": 193271, "epoch": 2171, "lr": 1.9172885937195582e-05}
{"train_loss": 0.09439358860254288, "global_step": 193272, "epoch": 2171, "lr": 1.9172429500002175e-05}
{"train_loss": 0.04712904989719391, "global_step": 193273, "epoch": 2171, "lr": 1.917197306695314e-05}
{"train_loss": 0.05781693756580353, "global_step": 193274, "epoch": 2171, "lr": 1.9171516638048582e-05}
{"train_loss": 0.07754074037075043, "global_step": 193275, "epoch": 2171, "lr": 1.9171060213288518e-05}
{"train_loss": 0.0917903482913971, "global_step": 193276, "epoch": 2171, "lr": 1.9170603792673043e-05}
{"train_loss": 0.03941726312041283, "global_step": 193277, "epoch": 2171, "lr": 1.9170147376202185e-05}
{"train_loss": 0.05740692839026451, "global_step": 193278, "epoch": 2171, "lr": 1.9169690963876046e-05}
{"train_loss": 0.05774218961596489, "global_step": 193279, "epoch": 2171, "lr": 1.916923455569465e-05}
{"train_loss": 0.062054261565208435, "global_step": 193280, "epoch": 2171, "lr": 1.916877815165809e-05}
{"train_loss": 0.12266115844249725, "global_step": 193281, "epoch": 2171, "lr": 1.9168321751766394e-05}
{"train_loss": 0.08559461683034897, "global_step": 193282, "epoch": 2171, "lr": 1.916786535601966e-05}
{"train_loss": 0.0752817913889885, "global_step": 193283, "epoch": 2171, "lr": 1.9167408964417915e-05}
{"train_loss": 0.0653298944234848, "global_step": 193284, "epoch": 2171, "lr": 1.916695257696125e-05}
{"train_loss": 0.11837138235569, "global_step": 193285, "epoch": 2171, "lr": 1.9166496193649698e-05}
{"train_loss": 0.03623215854167938, "global_step": 193286, "epoch": 2171, "lr": 1.916603981448335e-05}
{"train_loss": 0.05498971417546272, "global_step": 193287, "epoch": 2171, "lr": 1.9165583439462237e-05}
{"train_loss": 0.051819294691085815, "global_step": 193288, "epoch": 2171, "lr": 1.9165127068586448e-05}
{"train_loss": 0.0499502532184124, "global_step": 193289, "epoch": 2171, "lr": 1.9164670701856017e-05}
{"train_loss": 0.05150958523154259, "global_step": 193290, "epoch": 2171, "lr": 1.9164214339271024e-05}
{"train_loss": 0.06682474911212921, "global_step": 193291, "epoch": 2171, "lr": 1.9163757980831542e-05}
{"train_loss": 0.033607397228479385, "global_step": 193292, "epoch": 2171, "lr": 1.91633016265376e-05}
{"train_loss": 0.0412331186234951, "global_step": 193293, "epoch": 2171, "lr": 1.9162845276389292e-05}
{"train_loss": 0.037649888545274734, "global_step": 193294, "epoch": 2171, "lr": 1.916238893038666e-05}
{"train_loss": 0.07234129309654236, "global_step": 193295, "epoch": 2171, "lr": 1.916193258852976e-05}
{"train_loss": 0.042596977204084396, "global_step": 193296, "epoch": 2171, "lr": 1.916147625081867e-05}
{"train_loss": 0.02639004774391651, "global_step": 193297, "epoch": 2171, "lr": 1.9161019917253432e-05}
{"train_loss": 0.05238831043243408, "global_step": 193298, "epoch": 2171, "lr": 1.9160563587834136e-05}
{"train_loss": 0.04923096299171448, "global_step": 193299, "epoch": 2171, "lr": 1.916010726256083e-05}
{"train_loss": 0.06644826382398605, "global_step": 193300, "epoch": 2171, "lr": 1.9159650941433548e-05}
{"train_loss": 0.04508417844772339, "global_step": 193301, "epoch": 2171, "lr": 1.9159194624452392e-05}
{"train_loss": 0.04347796365618706, "global_step": 193302, "epoch": 2171, "lr": 1.915873831161739e-05}
{"train_loss": 0.09585621953010559, "global_step": 193303, "epoch": 2171, "lr": 1.915828200292864e-05}
{"train_loss": 0.09605872631072998, "global_step": 193304, "epoch": 2171, "lr": 1.9157825698386166e-05}
{"train_loss": 0.06866108626127243, "global_step": 193305, "epoch": 2171, "lr": 1.9157369397990068e-05}
{"train_loss": 0.10158760100603104, "global_step": 193306, "epoch": 2171, "lr": 1.9156913101740358e-05}
{"train_loss": 0.06541157028313434, "global_step": 193307, "epoch": 2171, "lr": 1.915645680963713e-05, "val_loss": 7.982058048248291}
{"train_loss": 0.053266510367393494, "global_step": 193308, "epoch": 2172, "lr": 1.9156000521680458e-05}
{"train_loss": 0.0574636273086071, "global_step": 193309, "epoch": 2172, "lr": 1.9155544237870365e-05}
{"train_loss": 0.038481809198856354, "global_step": 193310, "epoch": 2172, "lr": 1.9155087958206957e-05}
{"train_loss": 0.08335462212562561, "global_step": 193311, "epoch": 2172, "lr": 1.915463168269025e-05}
{"train_loss": 0.10918629914522171, "global_step": 193312, "epoch": 2172, "lr": 1.9154175411320346e-05}
{"train_loss": 0.10802647471427917, "global_step": 193313, "epoch": 2172, "lr": 1.9153719144097267e-05}
{"train_loss": 0.07540713995695114, "global_step": 193314, "epoch": 2172, "lr": 1.9153262881021118e-05}
{"train_loss": 0.11252816021442413, "global_step": 193315, "epoch": 2172, "lr": 1.9152806622091914e-05}
{"train_loss": 0.03732156381011009, "global_step": 193316, "epoch": 2172, "lr": 1.9152350367309756e-05}
{"train_loss": 0.07331493496894836, "global_step": 193317, "epoch": 2172, "lr": 1.9151894116674674e-05}
{"train_loss": 0.08503618836402893, "global_step": 193318, "epoch": 2172, "lr": 1.9151437870186754e-05}
{"train_loss": 0.06142493337392807, "global_step": 193319, "epoch": 2172, "lr": 1.9150981627846036e-05}
{"train_loss": 0.04512723907828331, "global_step": 193320, "epoch": 2172, "lr": 1.9150525389652612e-05}
{"train_loss": 0.07631948590278625, "global_step": 193321, "epoch": 2172, "lr": 1.9150069155606505e-05}
{"train_loss": 0.03792255371809006, "global_step": 193322, "epoch": 2172, "lr": 1.914961292570781e-05}
{"train_loss": 0.09735581278800964, "global_step": 193323, "epoch": 2172, "lr": 1.914915669995656e-05}
{"train_loss": 0.09127864986658096, "global_step": 193324, "epoch": 2172, "lr": 1.9148700478352843e-05}
{"train_loss": 0.09094541519880295, "global_step": 193325, "epoch": 2172, "lr": 1.9148244260896693e-05}
{"train_loss": 0.06440358608961105, "global_step": 193326, "epoch": 2172, "lr": 1.914778804758818e-05}
{"train_loss": 0.04419242590665817, "global_step": 193327, "epoch": 2172, "lr": 1.9147331838427394e-05}
{"train_loss": 0.07531385868787766, "global_step": 193328, "epoch": 2172, "lr": 1.9146875633414358e-05}
{"train_loss": 0.08494334667921066, "global_step": 193329, "epoch": 2172, "lr": 1.914641943254916e-05}
{"train_loss": 0.03125838562846184, "global_step": 193330, "epoch": 2172, "lr": 1.9145963235831836e-05}
{"train_loss": 0.08601116389036179, "global_step": 193331, "epoch": 2172, "lr": 1.914550704326248e-05}
{"train_loss": 0.032493218779563904, "global_step": 193332, "epoch": 2172, "lr": 1.9145050854841128e-05}
{"train_loss": 0.08063052594661713, "global_step": 193333, "epoch": 2172, "lr": 1.914459467056784e-05}
{"train_loss": 0.048473965376615524, "global_step": 193334, "epoch": 2172, "lr": 1.9144138490442694e-05}
{"train_loss": 0.07716865092515945, "global_step": 193335, "epoch": 2172, "lr": 1.9143682314465727e-05}
{"train_loss": 0.031621482223272324, "global_step": 193336, "epoch": 2172, "lr": 1.914322614263704e-05}
{"train_loss": 0.06118262931704521, "global_step": 193337, "epoch": 2172, "lr": 1.914276997495664e-05}
{"train_loss": 0.04441000148653984, "global_step": 193338, "epoch": 2172, "lr": 1.9142313811424645e-05}
{"train_loss": 0.05469169467687607, "global_step": 193339, "epoch": 2172, "lr": 1.914185765204109e-05}
{"train_loss": 0.05146603658795357, "global_step": 193340, "epoch": 2172, "lr": 1.9141401496806015e-05}
{"train_loss": 0.05254937708377838, "global_step": 193341, "epoch": 2172, "lr": 1.9140945345719524e-05}
{"train_loss": 0.09694058448076248, "global_step": 193342, "epoch": 2172, "lr": 1.9140489198781635e-05}
{"train_loss": 0.05430499464273453, "global_step": 193343, "epoch": 2172, "lr": 1.9140033055992435e-05}
{"train_loss": 0.07541366666555405, "global_step": 193344, "epoch": 2172, "lr": 1.9139576917351993e-05}
{"train_loss": 0.03635626658797264, "global_step": 193345, "epoch": 2172, "lr": 1.9139120782860348e-05}
{"train_loss": 0.05431528389453888, "global_step": 193346, "epoch": 2172, "lr": 1.9138664652517586e-05}
{"train_loss": 0.07023683190345764, "global_step": 193347, "epoch": 2172, "lr": 1.9138208526323737e-05}
{"train_loss": 0.065029077231884, "global_step": 193348, "epoch": 2172, "lr": 1.9137752404278895e-05}
{"train_loss": 0.04250511899590492, "global_step": 193349, "epoch": 2172, "lr": 1.9137296286383093e-05}
{"train_loss": 0.0643678531050682, "global_step": 193350, "epoch": 2172, "lr": 1.9136840172636423e-05}
{"train_loss": 0.08833208680152893, "global_step": 193351, "epoch": 2172, "lr": 1.913638406303891e-05}
{"train_loss": 0.08380857855081558, "global_step": 193352, "epoch": 2172, "lr": 1.913592795759065e-05}
{"train_loss": 0.039951223880052567, "global_step": 193353, "epoch": 2172, "lr": 1.913547185629167e-05}
{"train_loss": 0.07142049819231033, "global_step": 193354, "epoch": 2172, "lr": 1.9135015759142068e-05}
{"train_loss": 0.031058579683303833, "global_step": 193355, "epoch": 2172, "lr": 1.913455966614187e-05}
{"train_loss": 0.11867959797382355, "global_step": 193356, "epoch": 2172, "lr": 1.9134103577291174e-05}
{"train_loss": 0.040502969175577164, "global_step": 193357, "epoch": 2172, "lr": 1.9133647492590002e-05}
{"train_loss": 0.07239805161952972, "global_step": 193358, "epoch": 2172, "lr": 1.913319141203845e-05}
{"train_loss": 0.10670723021030426, "global_step": 193359, "epoch": 2172, "lr": 1.9132735335636548e-05}
{"train_loss": 0.05781707540154457, "global_step": 193360, "epoch": 2172, "lr": 1.913227926338439e-05}
{"train_loss": 0.12253626435995102, "global_step": 193361, "epoch": 2172, "lr": 1.9131823195282004e-05}
{"train_loss": 0.08635316789150238, "global_step": 193362, "epoch": 2172, "lr": 1.9131367131329474e-05}
{"train_loss": 0.08459053933620453, "global_step": 193363, "epoch": 2172, "lr": 1.913091107152687e-05}
{"train_loss": 0.11758061498403549, "global_step": 193364, "epoch": 2172, "lr": 1.9130455015874217e-05}
{"train_loss": 0.03942471370100975, "global_step": 193365, "epoch": 2172, "lr": 1.912999896437162e-05}
{"train_loss": 0.07750626653432846, "global_step": 193366, "epoch": 2172, "lr": 1.91295429170191e-05}
{"train_loss": 0.049169108271598816, "global_step": 193367, "epoch": 2172, "lr": 1.9129086873816747e-05}
{"train_loss": 0.048071373254060745, "global_step": 193368, "epoch": 2172, "lr": 1.9128630834764604e-05}
{"train_loss": 0.08902832120656967, "global_step": 193369, "epoch": 2172, "lr": 1.9128174799862753e-05}
{"train_loss": 0.11471454799175262, "global_step": 193370, "epoch": 2172, "lr": 1.912771876911123e-05}
{"train_loss": 0.06707197427749634, "global_step": 193371, "epoch": 2172, "lr": 1.912726274251012e-05}
{"train_loss": 0.11972244083881378, "global_step": 193372, "epoch": 2172, "lr": 1.9126806720059477e-05}
{"train_loss": 0.08202091604471207, "global_step": 193373, "epoch": 2172, "lr": 1.9126350701759343e-05}
{"train_loss": 0.02956680953502655, "global_step": 193374, "epoch": 2172, "lr": 1.9125894687609806e-05}
{"train_loss": 0.037077050656080246, "global_step": 193375, "epoch": 2172, "lr": 1.9125438677610907e-05}
{"train_loss": 0.06912949681282043, "global_step": 193376, "epoch": 2172, "lr": 1.9124982671762726e-05}
{"train_loss": 0.06368529796600342, "global_step": 193377, "epoch": 2172, "lr": 1.912452667006532e-05}
{"train_loss": 0.0648796558380127, "global_step": 193378, "epoch": 2172, "lr": 1.9124070672518722e-05}
{"train_loss": 0.03199400752782822, "global_step": 193379, "epoch": 2172, "lr": 1.9123614679123024e-05}
{"train_loss": 0.04408857598900795, "global_step": 193380, "epoch": 2172, "lr": 1.9123158689878295e-05}
{"train_loss": 0.0475793294608593, "global_step": 193381, "epoch": 2172, "lr": 1.9122702704784563e-05}
{"train_loss": 0.050754107534885406, "global_step": 193382, "epoch": 2172, "lr": 1.912224672384192e-05}
{"train_loss": 0.09851588308811188, "global_step": 193383, "epoch": 2172, "lr": 1.91217907470504e-05}
{"train_loss": 0.08225632458925247, "global_step": 193384, "epoch": 2172, "lr": 1.91213347744101e-05}
{"train_loss": 0.07974336296319962, "global_step": 193385, "epoch": 2172, "lr": 1.912087880592104e-05}
{"train_loss": 0.07935729622840881, "global_step": 193386, "epoch": 2172, "lr": 1.9120422841583318e-05}
{"train_loss": 0.07586625218391418, "global_step": 193387, "epoch": 2172, "lr": 1.9119966881396962e-05}
{"train_loss": 0.05995996296405792, "global_step": 193388, "epoch": 2172, "lr": 1.9119510925362068e-05}
{"train_loss": 0.08132954686880112, "global_step": 193389, "epoch": 2172, "lr": 1.911905497347866e-05}
{"train_loss": 0.0571492463350296, "global_step": 193390, "epoch": 2172, "lr": 1.9118599025746838e-05}
{"train_loss": 0.10883815586566925, "global_step": 193391, "epoch": 2172, "lr": 1.9118143082166624e-05}
{"train_loss": 0.049809180200099945, "global_step": 193392, "epoch": 2172, "lr": 1.9117687142738122e-05}
{"train_loss": 0.10063894093036652, "global_step": 193393, "epoch": 2172, "lr": 1.911723120746135e-05}
{"train_loss": 0.0515715591609478, "global_step": 193394, "epoch": 2172, "lr": 1.9116775276336406e-05}
{"train_loss": 0.10121931880712509, "global_step": 193395, "epoch": 2172, "lr": 1.911631934936332e-05}
{"train_loss": 0.06960591213421875, "global_step": 193396, "epoch": 2172, "lr": 1.9115863426542187e-05, "val_loss": 7.947938442230225}
{"train_loss": 0.07295098155736923, "global_step": 193397, "epoch": 2173, "lr": 1.9115407507873023e-05}
{"train_loss": 0.06886232644319534, "global_step": 193398, "epoch": 2173, "lr": 1.911495159335593e-05}
{"train_loss": 0.07683710753917694, "global_step": 193399, "epoch": 2173, "lr": 1.9114495682990964e-05}
{"train_loss": 0.05754163861274719, "global_step": 193400, "epoch": 2173, "lr": 1.9114039776778158e-05}
{"train_loss": 0.08748599886894226, "global_step": 193401, "epoch": 2173, "lr": 1.9113583874717616e-05}
{"train_loss": 0.04256552830338478, "global_step": 193402, "epoch": 2173, "lr": 1.9113127976809358e-05}
{"train_loss": 0.12055232375860214, "global_step": 193403, "epoch": 2173, "lr": 1.911267208305348e-05}
{"train_loss": 0.045471467077732086, "global_step": 193404, "epoch": 2173, "lr": 1.911221619345001e-05}
{"train_loss": 0.08175644278526306, "global_step": 193405, "epoch": 2173, "lr": 1.9111760307999045e-05}
{"train_loss": 0.04962485283613205, "global_step": 193406, "epoch": 2173, "lr": 1.9111304426700605e-05}
{"train_loss": 0.11606168001890182, "global_step": 193407, "epoch": 2173, "lr": 1.9110848549554795e-05}
{"train_loss": 0.06031636893749237, "global_step": 193408, "epoch": 2173, "lr": 1.9110392676561635e-05}
{"train_loss": 0.06533744931221008, "global_step": 193409, "epoch": 2173, "lr": 1.9109936807721223e-05}
{"train_loss": 0.0703638419508934, "global_step": 193410, "epoch": 2173, "lr": 1.9109480943033607e-05}
{"train_loss": 0.0747271478176117, "global_step": 193411, "epoch": 2173, "lr": 1.910902508249883e-05}
{"train_loss": 0.04070984944701195, "global_step": 193412, "epoch": 2173, "lr": 1.910856922611698e-05}
{"train_loss": 0.041052572429180145, "global_step": 193413, "epoch": 2173, "lr": 1.9108113373888088e-05}
{"train_loss": 0.05066227912902832, "global_step": 193414, "epoch": 2173, "lr": 1.9107657525812255e-05}
{"train_loss": 0.07474386692047119, "global_step": 193415, "epoch": 2173, "lr": 1.910720168188952e-05}
{"train_loss": 0.07085947692394257, "global_step": 193416, "epoch": 2173, "lr": 1.9106745842119927e-05}
{"train_loss": 0.09169893711805344, "global_step": 193417, "epoch": 2173, "lr": 1.9106290006503553e-05}
{"train_loss": 0.029850410297513008, "global_step": 193418, "epoch": 2173, "lr": 1.9105834175040482e-05}
{"train_loss": 0.11240378767251968, "global_step": 193419, "epoch": 2173, "lr": 1.910537834773074e-05}
{"train_loss": 0.09001076966524124, "global_step": 193420, "epoch": 2173, "lr": 1.9104922524574413e-05}
{"train_loss": 0.07320697605609894, "global_step": 193421, "epoch": 2173, "lr": 1.9104466705571544e-05}
{"train_loss": 0.04516321420669556, "global_step": 193422, "epoch": 2173, "lr": 1.9104010890722218e-05}
{"train_loss": 0.056155793368816376, "global_step": 193423, "epoch": 2173, "lr": 1.9103555080026458e-05}
{"train_loss": 0.0794341117143631, "global_step": 193424, "epoch": 2173, "lr": 1.9103099273484374e-05}
{"train_loss": 0.047832295298576355, "global_step": 193425, "epoch": 2173, "lr": 1.910264347109598e-05}
{"train_loss": 0.07083359360694885, "global_step": 193426, "epoch": 2173, "lr": 1.9102187672861377e-05}
{"train_loss": 0.10147306323051453, "global_step": 193427, "epoch": 2173, "lr": 1.9101731878780594e-05}
{"train_loss": 0.13342030346393585, "global_step": 193428, "epoch": 2173, "lr": 1.910127608885372e-05}
{"train_loss": 0.054662711918354034, "global_step": 193429, "epoch": 2173, "lr": 1.9100820303080786e-05}
{"train_loss": 0.06915642321109772, "global_step": 193430, "epoch": 2173, "lr": 1.9100364521461884e-05}
{"train_loss": 0.07058092206716537, "global_step": 193431, "epoch": 2173, "lr": 1.9099908743997046e-05}
{"train_loss": 0.11243119835853577, "global_step": 193432, "epoch": 2173, "lr": 1.9099452970686372e-05}
{"train_loss": 0.08112876117229462, "global_step": 193433, "epoch": 2173, "lr": 1.9098997201529877e-05}
{"train_loss": 0.08553170412778854, "global_step": 193434, "epoch": 2173, "lr": 1.9098541436527644e-05}
{"train_loss": 0.09522576630115509, "global_step": 193435, "epoch": 2173, "lr": 1.9098085675679763e-05}
{"train_loss": 0.10009711980819702, "global_step": 193436, "epoch": 2173, "lr": 1.909762991898624e-05}
{"train_loss": 0.07323867082595825, "global_step": 193437, "epoch": 2173, "lr": 1.9097174166447185e-05}
{"train_loss": 0.06502362340688705, "global_step": 193438, "epoch": 2173, "lr": 1.9096718418062614e-05}
{"train_loss": 0.10633295029401779, "global_step": 193439, "epoch": 2173, "lr": 1.9096262673832638e-05}
{"train_loss": 0.11423882097005844, "global_step": 193440, "epoch": 2173, "lr": 1.9095806933757277e-05}
{"train_loss": 0.07017574459314346, "global_step": 193441, "epoch": 2173, "lr": 1.909535119783662e-05}
{"train_loss": 0.10155593603849411, "global_step": 193442, "epoch": 2173, "lr": 1.90948954660707e-05}
{"train_loss": 0.07085287570953369, "global_step": 193443, "epoch": 2173, "lr": 1.909443973845961e-05}
{"train_loss": 0.029098542407155037, "global_step": 193444, "epoch": 2173, "lr": 1.9093984015003386e-05}
{"train_loss": 0.125074103474617, "global_step": 193445, "epoch": 2173, "lr": 1.909352829570211e-05}
{"train_loss": 0.07323924452066422, "global_step": 193446, "epoch": 2173, "lr": 1.9093072580555815e-05}
{"train_loss": 0.07429948449134827, "global_step": 193447, "epoch": 2173, "lr": 1.90926168695646e-05}
{"train_loss": 0.0830613225698471, "global_step": 193448, "epoch": 2173, "lr": 1.909216116272849e-05}
{"train_loss": 0.07747843861579895, "global_step": 193449, "epoch": 2173, "lr": 1.9091705460047577e-05}
{"train_loss": 0.044467635452747345, "global_step": 193450, "epoch": 2173, "lr": 1.9091249761521907e-05}
{"train_loss": 0.07601917535066605, "global_step": 193451, "epoch": 2173, "lr": 1.9090794067151522e-05}
{"train_loss": 0.047952234745025635, "global_step": 193452, "epoch": 2173, "lr": 1.909033837693653e-05}
{"train_loss": 0.08136331290006638, "global_step": 193453, "epoch": 2173, "lr": 1.9089882690876937e-05}
{"train_loss": 0.04974263161420822, "global_step": 193454, "epoch": 2173, "lr": 1.9089427008972855e-05}
{"train_loss": 0.07108504325151443, "global_step": 193455, "epoch": 2173, "lr": 1.9088971331224303e-05}
{"train_loss": 0.055790454149246216, "global_step": 193456, "epoch": 2173, "lr": 1.9088515657631382e-05}
{"train_loss": 0.05406252294778824, "global_step": 193457, "epoch": 2173, "lr": 1.9088059988194114e-05}
{"train_loss": 0.06924740225076675, "global_step": 193458, "epoch": 2173, "lr": 1.90876043229126e-05}
{"train_loss": 0.09338374435901642, "global_step": 193459, "epoch": 2173, "lr": 1.908714866178686e-05}
{"train_loss": 0.05846185237169266, "global_step": 193460, "epoch": 2173, "lr": 1.9086693004816997e-05}
{"train_loss": 0.07178891450166702, "global_step": 193461, "epoch": 2173, "lr": 1.908623735200303e-05}
{"train_loss": 0.08052612841129303, "global_step": 193462, "epoch": 2173, "lr": 1.9085781703345064e-05}
{"train_loss": 0.06102978438138962, "global_step": 193463, "epoch": 2173, "lr": 1.908532605884311e-05}
{"train_loss": 0.07430193573236465, "global_step": 193464, "epoch": 2173, "lr": 1.9084870418497287e-05}
{"train_loss": 0.05584605783224106, "global_step": 193465, "epoch": 2173, "lr": 1.9084414782307604e-05}
{"train_loss": 0.0647030919790268, "global_step": 193466, "epoch": 2173, "lr": 1.908395915027416e-05}
{"train_loss": 0.1108848825097084, "global_step": 193467, "epoch": 2173, "lr": 1.9083503522396985e-05}
{"train_loss": 0.06935387849807739, "global_step": 193468, "epoch": 2173, "lr": 1.9083047898676175e-05}
{"train_loss": 0.0853785052895546, "global_step": 193469, "epoch": 2173, "lr": 1.9082592279111756e-05}
{"train_loss": 0.08485093712806702, "global_step": 193470, "epoch": 2173, "lr": 1.90821366637038e-05}
{"train_loss": 0.05399126559495926, "global_step": 193471, "epoch": 2173, "lr": 1.9081681052452395e-05}
{"train_loss": 0.0811874270439148, "global_step": 193472, "epoch": 2173, "lr": 1.9081225445357565e-05}
{"train_loss": 0.054951075464487076, "global_step": 193473, "epoch": 2173, "lr": 1.90807698424194e-05}
{"train_loss": 0.06957880407571793, "global_step": 193474, "epoch": 2173, "lr": 1.9080314243637937e-05}
{"train_loss": 0.050539642572402954, "global_step": 193475, "epoch": 2173, "lr": 1.907985864901326e-05}
{"train_loss": 0.025144703686237335, "global_step": 193476, "epoch": 2173, "lr": 1.907940305854541e-05}
{"train_loss": 0.04758988320827484, "global_step": 193477, "epoch": 2173, "lr": 1.907894747223447e-05}
{"train_loss": 0.07541221380233765, "global_step": 193478, "epoch": 2173, "lr": 1.907849189008047e-05}
{"train_loss": 0.06048380210995674, "global_step": 193479, "epoch": 2173, "lr": 1.9078036312083507e-05}
{"train_loss": 0.04705012962222099, "global_step": 193480, "epoch": 2173, "lr": 1.907758073824361e-05}
{"train_loss": 0.09379483759403229, "global_step": 193481, "epoch": 2173, "lr": 1.9077125168560874e-05}
{"train_loss": 0.04753204435110092, "global_step": 193482, "epoch": 2173, "lr": 1.9076669603035324e-05}
{"train_loss": 0.0706920251250267, "global_step": 193483, "epoch": 2173, "lr": 1.9076214041667052e-05}
{"train_loss": 0.04286830872297287, "global_step": 193484, "epoch": 2173, "lr": 1.9075758484456093e-05}
{"train_loss": 0.07169009209348914, "global_step": 193485, "epoch": 2173, "lr": 1.907530293140254e-05, "val_loss": 7.892561912536621}
{"train_loss": 0.0534411258995533, "global_step": 193486, "epoch": 2174, "lr": 1.9074847382506412e-05}
{"train_loss": 0.10170187801122665, "global_step": 193487, "epoch": 2174, "lr": 1.9074391837767813e-05}
{"train_loss": 0.05299697816371918, "global_step": 193488, "epoch": 2174, "lr": 1.907393629718679e-05}
{"train_loss": 0.046715907752513885, "global_step": 193489, "epoch": 2174, "lr": 1.907348076076338e-05}
{"train_loss": 0.04544370621442795, "global_step": 193490, "epoch": 2174, "lr": 1.9073025228497675e-05}
{"train_loss": 0.05348241701722145, "global_step": 193491, "epoch": 2174, "lr": 1.9072569700389714e-05}
{"train_loss": 0.026761868968605995, "global_step": 193492, "epoch": 2174, "lr": 1.9072114176439587e-05}
{"train_loss": 0.06212940812110901, "global_step": 193493, "epoch": 2174, "lr": 1.9071658656647318e-05}
{"train_loss": 0.09256036579608917, "global_step": 193494, "epoch": 2174, "lr": 1.9071203141013e-05}
{"train_loss": 0.03297942504286766, "global_step": 193495, "epoch": 2174, "lr": 1.9070747629536674e-05}
{"train_loss": 0.07502440363168716, "global_step": 193496, "epoch": 2174, "lr": 1.907029212221842e-05}
{"train_loss": 0.07597770541906357, "global_step": 193497, "epoch": 2174, "lr": 1.9069836619058268e-05}
{"train_loss": 0.05981219559907913, "global_step": 193498, "epoch": 2174, "lr": 1.9069381120056324e-05}
{"train_loss": 0.07762376219034195, "global_step": 193499, "epoch": 2174, "lr": 1.9068925625212604e-05}
{"train_loss": 0.11538061499595642, "global_step": 193500, "epoch": 2174, "lr": 1.9068470134527206e-05}
{"train_loss": 0.07246388494968414, "global_step": 193501, "epoch": 2174, "lr": 1.906801464800016e-05}
{"train_loss": 0.1585293710231781, "global_step": 193502, "epoch": 2174, "lr": 1.906755916563156e-05}
{"train_loss": 0.04170946031808853, "global_step": 193503, "epoch": 2174, "lr": 1.9067103687421427e-05}
{"train_loss": 0.0824400931596756, "global_step": 193504, "epoch": 2174, "lr": 1.906664821336987e-05}
{"train_loss": 0.02803143486380577, "global_step": 193505, "epoch": 2174, "lr": 1.90661927434769e-05}
{"train_loss": 0.0767437070608139, "global_step": 193506, "epoch": 2174, "lr": 1.9065737277742612e-05}
{"train_loss": 0.057475123554468155, "global_step": 193507, "epoch": 2174, "lr": 1.906528181616707e-05}
{"train_loss": 0.038253169506788254, "global_step": 193508, "epoch": 2174, "lr": 1.906482635875031e-05}
{"train_loss": 0.10441029816865921, "global_step": 193509, "epoch": 2174, "lr": 1.9064370905492425e-05}
{"train_loss": 0.10659899562597275, "global_step": 193510, "epoch": 2174, "lr": 1.9063915456393437e-05}
{"train_loss": 0.041838496923446655, "global_step": 193511, "epoch": 2174, "lr": 1.906346001145345e-05}
{"train_loss": 0.06294640153646469, "global_step": 193512, "epoch": 2174, "lr": 1.906300457067248e-05}
{"train_loss": 0.05763193964958191, "global_step": 193513, "epoch": 2174, "lr": 1.9062549134050635e-05}
{"train_loss": 0.09076671302318573, "global_step": 193514, "epoch": 2174, "lr": 1.906209370158793e-05}
{"train_loss": 0.05744144320487976, "global_step": 193515, "epoch": 2174, "lr": 1.9061638273284472e-05}
{"train_loss": 0.10463916510343552, "global_step": 193516, "epoch": 2174, "lr": 1.906118284914028e-05}
{"train_loss": 0.03870191425085068, "global_step": 193517, "epoch": 2174, "lr": 1.9060727429155456e-05}
{"train_loss": 0.05394672602415085, "global_step": 193518, "epoch": 2174, "lr": 1.9060272013330023e-05}
{"train_loss": 0.05772408843040466, "global_step": 193519, "epoch": 2174, "lr": 1.905981660166407e-05}
{"train_loss": 0.07035430520772934, "global_step": 193520, "epoch": 2174, "lr": 1.9059361194157633e-05}
{"train_loss": 0.057532791048288345, "global_step": 193521, "epoch": 2174, "lr": 1.9058905790810805e-05}
{"train_loss": 0.06900739669799805, "global_step": 193522, "epoch": 2174, "lr": 1.9058450391623617e-05}
{"train_loss": 0.06733965128660202, "global_step": 193523, "epoch": 2174, "lr": 1.9057994996596157e-05}
{"train_loss": 0.06940003484487534, "global_step": 193524, "epoch": 2174, "lr": 1.905753960572845e-05}
{"train_loss": 0.05331464484333992, "global_step": 193525, "epoch": 2174, "lr": 1.9057084219020603e-05}
{"train_loss": 0.06179727986454964, "global_step": 193526, "epoch": 2174, "lr": 1.9056628836472655e-05}
{"train_loss": 0.08725959807634354, "global_step": 193527, "epoch": 2174, "lr": 1.905617345808465e-05}
{"train_loss": 0.06003529950976372, "global_step": 193528, "epoch": 2174, "lr": 1.905571808385668e-05}
{"train_loss": 0.0513898991048336, "global_step": 193529, "epoch": 2174, "lr": 1.9055262713788772e-05}
{"train_loss": 0.04798165708780289, "global_step": 193530, "epoch": 2174, "lr": 1.9054807347881028e-05}
{"train_loss": 0.07114352285861969, "global_step": 193531, "epoch": 2174, "lr": 1.9054351986133468e-05}
{"train_loss": 0.07611199468374252, "global_step": 193532, "epoch": 2174, "lr": 1.9053896628546192e-05}
{"train_loss": 0.0857865959405899, "global_step": 193533, "epoch": 2174, "lr": 1.905344127511922e-05}
{"train_loss": 0.04743967205286026, "global_step": 193534, "epoch": 2174, "lr": 1.9052985925852657e-05}
{"train_loss": 0.03920364752411842, "global_step": 193535, "epoch": 2174, "lr": 1.9052530580746524e-05}
{"train_loss": 0.03453056514263153, "global_step": 193536, "epoch": 2174, "lr": 1.9052075239800924e-05}
{"train_loss": 0.08092793077230453, "global_step": 193537, "epoch": 2174, "lr": 1.9051619903015873e-05}
{"train_loss": 0.038463741540908813, "global_step": 193538, "epoch": 2174, "lr": 1.905116457039147e-05}
{"train_loss": 0.049313683062791824, "global_step": 193539, "epoch": 2174, "lr": 1.905070924192775e-05}
{"train_loss": 0.07171696424484253, "global_step": 193540, "epoch": 2174, "lr": 1.90502539176248e-05}
{"train_loss": 0.09592744708061218, "global_step": 193541, "epoch": 2174, "lr": 1.9049798597482644e-05}
{"train_loss": 0.03359461575746536, "global_step": 193542, "epoch": 2174, "lr": 1.9049343281501386e-05}
{"train_loss": 0.12883101403713226, "global_step": 193543, "epoch": 2174, "lr": 1.9048887969681052e-05}
{"train_loss": 0.1092061698436737, "global_step": 193544, "epoch": 2174, "lr": 1.9048432662021713e-05}
{"train_loss": 0.0683816522359848, "global_step": 193545, "epoch": 2174, "lr": 1.9047977358523454e-05}
{"train_loss": 0.11825994402170181, "global_step": 193546, "epoch": 2174, "lr": 1.9047522059186302e-05}
{"train_loss": 0.03528478741645813, "global_step": 193547, "epoch": 2174, "lr": 1.9047066764010345e-05}
{"train_loss": 0.037989143282175064, "global_step": 193548, "epoch": 2174, "lr": 1.9046611472995624e-05}
{"train_loss": 0.0937584787607193, "global_step": 193549, "epoch": 2174, "lr": 1.9046156186142223e-05}
{"train_loss": 0.028570296242833138, "global_step": 193550, "epoch": 2174, "lr": 1.9045700903450166e-05}
{"train_loss": 0.037345509976148605, "global_step": 193551, "epoch": 2174, "lr": 1.904524562491956e-05}
{"train_loss": 0.08476527035236359, "global_step": 193552, "epoch": 2174, "lr": 1.9044790350550423e-05}
{"train_loss": 0.08692758530378342, "global_step": 193553, "epoch": 2174, "lr": 1.9044335080342856e-05}
{"train_loss": 0.0868653878569603, "global_step": 193554, "epoch": 2174, "lr": 1.9043879814296884e-05}
{"train_loss": 0.07754532992839813, "global_step": 193555, "epoch": 2174, "lr": 1.9043424552412605e-05}
{"train_loss": 0.12574402987957, "global_step": 193556, "epoch": 2174, "lr": 1.904296929469004e-05}
{"train_loss": 0.06087316945195198, "global_step": 193557, "epoch": 2174, "lr": 1.9042514041129285e-05}
{"train_loss": 0.051024436950683594, "global_step": 193558, "epoch": 2174, "lr": 1.9042058791730373e-05}
{"train_loss": 0.10383173823356628, "global_step": 193559, "epoch": 2174, "lr": 1.904160354649339e-05}
{"train_loss": 0.05382842943072319, "global_step": 193560, "epoch": 2174, "lr": 1.9041148305418377e-05}
{"train_loss": 0.06744567304849625, "global_step": 193561, "epoch": 2174, "lr": 1.9040693068505416e-05}
{"train_loss": 0.0319925881922245, "global_step": 193562, "epoch": 2174, "lr": 1.9040237835754543e-05}
{"train_loss": 0.05420878529548645, "global_step": 193563, "epoch": 2174, "lr": 1.9039782607165845e-05}
{"train_loss": 0.07103033363819122, "global_step": 193564, "epoch": 2174, "lr": 1.903932738273936e-05}
{"train_loss": 0.047111429274082184, "global_step": 193565, "epoch": 2174, "lr": 1.9038872162475175e-05}
{"train_loss": 0.061445899307727814, "global_step": 193566, "epoch": 2174, "lr": 1.903841694637333e-05}
{"train_loss": 0.12856008112430573, "global_step": 193567, "epoch": 2174, "lr": 1.9037961734433886e-05}
{"train_loss": 0.03436533361673355, "global_step": 193568, "epoch": 2174, "lr": 1.9037506526656923e-05}
{"train_loss": 0.03546532243490219, "global_step": 193569, "epoch": 2174, "lr": 1.903705132304247e-05}
{"train_loss": 0.06442530453205109, "global_step": 193570, "epoch": 2174, "lr": 1.9036596123590627e-05}
{"train_loss": 0.07297863811254501, "global_step": 193571, "epoch": 2174, "lr": 1.903614092830142e-05}
{"train_loss": 0.05134309455752373, "global_step": 193572, "epoch": 2174, "lr": 1.9035685737174946e-05}
{"train_loss": 0.03254132345318794, "global_step": 193573, "epoch": 2174, "lr": 1.9035230550211226e-05}
{"train_loss": 0.06635213513555152, "global_step": 193574, "epoch": 2174, "lr": 1.9034775367410357e-05, "val_loss": 7.862903118133545}
{"train_loss": 0.03877045214176178, "global_step": 193575, "epoch": 2175, "lr": 1.903432018877237e-05}
{"train_loss": 0.04831258952617645, "global_step": 193576, "epoch": 2175, "lr": 1.9033865014297366e-05}
{"train_loss": 0.0464613139629364, "global_step": 193577, "epoch": 2175, "lr": 1.903340984398535e-05}
{"train_loss": 0.05457781255245209, "global_step": 193578, "epoch": 2175, "lr": 1.903295467783644e-05}
{"train_loss": 0.06625806540250778, "global_step": 193579, "epoch": 2175, "lr": 1.9032499515850656e-05}
{"train_loss": 0.06545271724462509, "global_step": 193580, "epoch": 2175, "lr": 1.9032044358028074e-05}
{"train_loss": 0.1234666034579277, "global_step": 193581, "epoch": 2175, "lr": 1.903158920436877e-05}
{"train_loss": 0.050189293920993805, "global_step": 193582, "epoch": 2175, "lr": 1.9031134054872778e-05}
{"train_loss": 0.15361812710762024, "global_step": 193583, "epoch": 2175, "lr": 1.9030678909540184e-05}
{"train_loss": 0.11397980898618698, "global_step": 193584, "epoch": 2175, "lr": 1.9030223768371026e-05}
{"train_loss": 0.06290140002965927, "global_step": 193585, "epoch": 2175, "lr": 1.9029768631365397e-05}
{"train_loss": 0.057388488203287125, "global_step": 193586, "epoch": 2175, "lr": 1.9029313498523314e-05}
{"train_loss": 0.0904831811785698, "global_step": 193587, "epoch": 2175, "lr": 1.9028858369844883e-05}
{"train_loss": 0.03847027197480202, "global_step": 193588, "epoch": 2175, "lr": 1.9028403245330124e-05}
{"train_loss": 0.04776325076818466, "global_step": 193589, "epoch": 2175, "lr": 1.9027948124979138e-05}
{"train_loss": 0.07179612666368484, "global_step": 193590, "epoch": 2175, "lr": 1.9027493008791952e-05}
{"train_loss": 0.0421643927693367, "global_step": 193591, "epoch": 2175, "lr": 1.9027037896768656e-05}
{"train_loss": 0.05374328792095184, "global_step": 193592, "epoch": 2175, "lr": 1.902658278890928e-05}
{"train_loss": 0.07223417609930038, "global_step": 193593, "epoch": 2175, "lr": 1.9026127685213924e-05}
{"train_loss": 0.06688141822814941, "global_step": 193594, "epoch": 2175, "lr": 1.902567258568261e-05}
{"train_loss": 0.031147392466664314, "global_step": 193595, "epoch": 2175, "lr": 1.902521749031543e-05}
{"train_loss": 0.06438101828098297, "global_step": 193596, "epoch": 2175, "lr": 1.9024762399112418e-05}
{"train_loss": 0.04761790111660957, "global_step": 193597, "epoch": 2175, "lr": 1.9024307312073642e-05}
{"train_loss": 0.0687703862786293, "global_step": 193598, "epoch": 2175, "lr": 1.9023852229199195e-05}
{"train_loss": 0.050871118903160095, "global_step": 193599, "epoch": 2175, "lr": 1.9023397150489115e-05}
{"train_loss": 0.03566901385784149, "global_step": 193600, "epoch": 2175, "lr": 1.902294207594344e-05}
{"train_loss": 0.06264948844909668, "global_step": 193601, "epoch": 2175, "lr": 1.9022487005562266e-05}
{"train_loss": 0.05903752148151398, "global_step": 193602, "epoch": 2175, "lr": 1.902203193934563e-05}
{"train_loss": 0.07457531243562698, "global_step": 193603, "epoch": 2175, "lr": 1.9021576877293625e-05}
{"train_loss": 0.08351614326238632, "global_step": 193604, "epoch": 2175, "lr": 1.9021121819406284e-05}
{"train_loss": 0.04440994933247566, "global_step": 193605, "epoch": 2175, "lr": 1.902066676568366e-05}
{"train_loss": 0.0684833899140358, "global_step": 193606, "epoch": 2175, "lr": 1.9020211716125846e-05}
{"train_loss": 0.056571245193481445, "global_step": 193607, "epoch": 2175, "lr": 1.9019756670732875e-05}
{"train_loss": 0.08819106221199036, "global_step": 193608, "epoch": 2175, "lr": 1.9019301629504833e-05}
{"train_loss": 0.09139502048492432, "global_step": 193609, "epoch": 2175, "lr": 1.9018846592441753e-05}
{"train_loss": 0.051730819046497345, "global_step": 193610, "epoch": 2175, "lr": 1.901839155954373e-05}
{"train_loss": 0.06851876527070999, "global_step": 193611, "epoch": 2175, "lr": 1.9017936530810786e-05}
{"train_loss": 0.09639310836791992, "global_step": 193612, "epoch": 2175, "lr": 1.9017481506243017e-05}
{"train_loss": 0.06604507565498352, "global_step": 193613, "epoch": 2175, "lr": 1.9017026485840462e-05}
{"train_loss": 0.056399840861558914, "global_step": 193614, "epoch": 2175, "lr": 1.9016571469603202e-05}
{"train_loss": 0.06376703828573227, "global_step": 193615, "epoch": 2175, "lr": 1.901611645753127e-05}
{"train_loss": 0.047434594482183456, "global_step": 193616, "epoch": 2175, "lr": 1.9015661449624745e-05}
{"train_loss": 0.05877244472503662, "global_step": 193617, "epoch": 2175, "lr": 1.90152064458837e-05}
{"train_loss": 0.06173147261142731, "global_step": 193618, "epoch": 2175, "lr": 1.9014751446308164e-05}
{"train_loss": 0.046908579766750336, "global_step": 193619, "epoch": 2175, "lr": 1.9014296450898238e-05}
{"train_loss": 0.08589272946119308, "global_step": 193620, "epoch": 2175, "lr": 1.9013841459653948e-05}
{"train_loss": 0.07751426845788956, "global_step": 193621, "epoch": 2175, "lr": 1.9013386472575384e-05}
{"train_loss": 0.06797303259372711, "global_step": 193622, "epoch": 2175, "lr": 1.901293148966257e-05}
{"train_loss": 0.053428955376148224, "global_step": 193623, "epoch": 2175, "lr": 1.9012476510915616e-05}
{"train_loss": 0.05360359326004982, "global_step": 193624, "epoch": 2175, "lr": 1.9012021536334534e-05}
{"train_loss": 0.07741760462522507, "global_step": 193625, "epoch": 2175, "lr": 1.9011566565919425e-05}
{"train_loss": 0.016813885420560837, "global_step": 193626, "epoch": 2175, "lr": 1.9011111599670316e-05}
{"train_loss": 0.1488177478313446, "global_step": 193627, "epoch": 2175, "lr": 1.901065663758731e-05}
{"train_loss": 0.054416004568338394, "global_step": 193628, "epoch": 2175, "lr": 1.901020167967042e-05}
{"train_loss": 0.09827939420938492, "global_step": 193629, "epoch": 2175, "lr": 1.9009746725919746e-05}
{"train_loss": 0.07329881191253662, "global_step": 193630, "epoch": 2175, "lr": 1.900929177633532e-05}
{"train_loss": 0.05229082703590393, "global_step": 193631, "epoch": 2175, "lr": 1.9008836830917232e-05}
{"train_loss": 0.14428871870040894, "global_step": 193632, "epoch": 2175, "lr": 1.9008381889665512e-05}
{"train_loss": 0.045595213770866394, "global_step": 193633, "epoch": 2175, "lr": 1.9007926952580235e-05}
{"train_loss": 0.042607277631759644, "global_step": 193634, "epoch": 2175, "lr": 1.900747201966149e-05}
{"train_loss": 0.0803757905960083, "global_step": 193635, "epoch": 2175, "lr": 1.9007017090909284e-05}
{"train_loss": 0.03875631466507912, "global_step": 193636, "epoch": 2175, "lr": 1.9006562166323733e-05}
{"train_loss": 0.1437608301639557, "global_step": 193637, "epoch": 2175, "lr": 1.9006107245904862e-05}
{"train_loss": 0.050777602940797806, "global_step": 193638, "epoch": 2175, "lr": 1.9005652329652733e-05}
{"train_loss": 0.03989534080028534, "global_step": 193639, "epoch": 2175, "lr": 1.9005197417567433e-05}
{"train_loss": 0.06797075271606445, "global_step": 193640, "epoch": 2175, "lr": 1.9004742509648987e-05}
{"train_loss": 0.07243962585926056, "global_step": 193641, "epoch": 2175, "lr": 1.90042876058975e-05}
{"train_loss": 0.1022760346531868, "global_step": 193642, "epoch": 2175, "lr": 1.9003832706312996e-05}
{"train_loss": 0.057496506720781326, "global_step": 193643, "epoch": 2175, "lr": 1.900337781089554e-05}
{"train_loss": 0.11693482100963593, "global_step": 193644, "epoch": 2175, "lr": 1.9002922919645217e-05}
{"train_loss": 0.04684530198574066, "global_step": 193645, "epoch": 2175, "lr": 1.900246803256206e-05}
{"train_loss": 0.03818751871585846, "global_step": 193646, "epoch": 2175, "lr": 1.9002013149646152e-05}
{"train_loss": 0.07734616100788116, "global_step": 193647, "epoch": 2175, "lr": 1.9001558270897535e-05}
{"train_loss": 0.05573359504342079, "global_step": 193648, "epoch": 2175, "lr": 1.9001103396316296e-05}
{"train_loss": 0.07583189010620117, "global_step": 193649, "epoch": 2175, "lr": 1.9000648525902466e-05}
{"train_loss": 0.054372671991586685, "global_step": 193650, "epoch": 2175, "lr": 1.9000193659656136e-05}
{"train_loss": 0.054087307304143906, "global_step": 193651, "epoch": 2175, "lr": 1.8999738797577334e-05}
{"train_loss": 0.06659764796495438, "global_step": 193652, "epoch": 2175, "lr": 1.899928393966614e-05}
{"train_loss": 0.08984506130218506, "global_step": 193653, "epoch": 2175, "lr": 1.899882908592263e-05}
{"train_loss": 0.06251241266727448, "global_step": 193654, "epoch": 2175, "lr": 1.8998374236346837e-05}
{"train_loss": 0.07085109502077103, "global_step": 193655, "epoch": 2175, "lr": 1.899791939093885e-05}
{"train_loss": 0.0796523168683052, "global_step": 193656, "epoch": 2175, "lr": 1.8997464549698694e-05}
{"train_loss": 0.09180701524019241, "global_step": 193657, "epoch": 2175, "lr": 1.8997009712626473e-05}
{"train_loss": 0.03980884701013565, "global_step": 193658, "epoch": 2175, "lr": 1.8996554879722206e-05}
{"train_loss": 0.056225746870040894, "global_step": 193659, "epoch": 2175, "lr": 1.8996100050985992e-05}
{"train_loss": 0.05699709802865982, "global_step": 193660, "epoch": 2175, "lr": 1.8995645226417858e-05}
{"train_loss": 0.03319944068789482, "global_step": 193661, "epoch": 2175, "lr": 1.8995190406017898e-05}
{"train_loss": 0.07298000901937485, "global_step": 193662, "epoch": 2175, "lr": 1.899473558978614e-05}
{"train_loss": 0.06710093838863829, "global_step": 193663, "epoch": 2175, "lr": 1.8994280777722678e-05, "val_loss": 7.991913795471191, "train_action_mse_error": 23.9545955657959}
{"train_loss": 0.09488216787576675, "global_step": 193664, "epoch": 2176, "lr": 1.8993825969827543e-05}
{"train_loss": 0.04658591374754906, "global_step": 193665, "epoch": 2176, "lr": 1.8993371166100827e-05}
{"train_loss": 0.08560346066951752, "global_step": 193666, "epoch": 2176, "lr": 1.8992916366542557e-05}
{"train_loss": 0.06595239043235779, "global_step": 193667, "epoch": 2176, "lr": 1.8992461571152824e-05}
{"train_loss": 0.08831676095724106, "global_step": 193668, "epoch": 2176, "lr": 1.8992006779931665e-05}
{"train_loss": 0.0562136173248291, "global_step": 193669, "epoch": 2176, "lr": 1.899155199287917e-05}
{"train_loss": 0.0741393119096756, "global_step": 193670, "epoch": 2176, "lr": 1.899109720999536e-05}
{"train_loss": 0.0851864442229271, "global_step": 193671, "epoch": 2176, "lr": 1.899064243128032e-05}
{"train_loss": 0.1046772301197052, "global_step": 193672, "epoch": 2176, "lr": 1.8990187656734132e-05}
{"train_loss": 0.09671711921691895, "global_step": 193673, "epoch": 2176, "lr": 1.8989732886356814e-05}
{"train_loss": 0.0947418063879013, "global_step": 193674, "epoch": 2176, "lr": 1.8989278120148467e-05}
{"train_loss": 0.03148391470313072, "global_step": 193675, "epoch": 2176, "lr": 1.8988823358109138e-05}
{"train_loss": 0.08247317373752594, "global_step": 193676, "epoch": 2176, "lr": 1.8988368600238858e-05}
{"train_loss": 0.04626483470201492, "global_step": 193677, "epoch": 2176, "lr": 1.8987913846537742e-05}
{"train_loss": 0.0499318391084671, "global_step": 193678, "epoch": 2176, "lr": 1.89874590970058e-05}
{"train_loss": 0.12171667069196701, "global_step": 193679, "epoch": 2176, "lr": 1.898700435164313e-05}
{"train_loss": 0.0458904467523098, "global_step": 193680, "epoch": 2176, "lr": 1.8986549610449766e-05}
{"train_loss": 0.08792120218276978, "global_step": 193681, "epoch": 2176, "lr": 1.8986094873425797e-05}
{"train_loss": 0.08430417627096176, "global_step": 193682, "epoch": 2176, "lr": 1.898564014057127e-05}
{"train_loss": 0.08329612761735916, "global_step": 193683, "epoch": 2176, "lr": 1.898518541188623e-05}
{"train_loss": 0.05659027770161629, "global_step": 193684, "epoch": 2176, "lr": 1.898473068737077e-05}
{"train_loss": 0.11137218028306961, "global_step": 193685, "epoch": 2176, "lr": 1.8984275967024924e-05}
{"train_loss": 0.020316923037171364, "global_step": 193686, "epoch": 2176, "lr": 1.898382125084877e-05}
{"train_loss": 0.073208749294281, "global_step": 193687, "epoch": 2176, "lr": 1.898336653884235e-05}
{"train_loss": 0.07796919345855713, "global_step": 193688, "epoch": 2176, "lr": 1.8982911831005746e-05}
{"train_loss": 0.08700956404209137, "global_step": 193689, "epoch": 2176, "lr": 1.898245712733902e-05}
{"train_loss": 0.1339871883392334, "global_step": 193690, "epoch": 2176, "lr": 1.898200242784221e-05}
{"train_loss": 0.04525349289178848, "global_step": 193691, "epoch": 2176, "lr": 1.898154773251541e-05}
{"train_loss": 0.07960059493780136, "global_step": 193692, "epoch": 2176, "lr": 1.8981093041358645e-05}
{"train_loss": 0.10328178107738495, "global_step": 193693, "epoch": 2176, "lr": 1.898063835437201e-05}
{"train_loss": 0.06827718764543533, "global_step": 193694, "epoch": 2176, "lr": 1.8980183671555536e-05}
{"train_loss": 0.05777723342180252, "global_step": 193695, "epoch": 2176, "lr": 1.8979728992909308e-05}
{"train_loss": 0.06688615679740906, "global_step": 193696, "epoch": 2176, "lr": 1.8979274318433365e-05}
{"train_loss": 0.0710025280714035, "global_step": 193697, "epoch": 2176, "lr": 1.89788196481278e-05}
{"train_loss": 0.10652666538953781, "global_step": 193698, "epoch": 2176, "lr": 1.8978364981992637e-05}
{"train_loss": 0.0933334231376648, "global_step": 193699, "epoch": 2176, "lr": 1.897791032002797e-05}
{"train_loss": 0.03540199249982834, "global_step": 193700, "epoch": 2176, "lr": 1.8977455662233824e-05}
{"train_loss": 0.06423453241586685, "global_step": 193701, "epoch": 2176, "lr": 1.89770010086103e-05}
{"train_loss": 0.09632395207881927, "global_step": 193702, "epoch": 2176, "lr": 1.8976546359157426e-05}
{"train_loss": 0.053418878465890884, "global_step": 193703, "epoch": 2176, "lr": 1.8976091713875294e-05}
{"train_loss": 0.049392156302928925, "global_step": 193704, "epoch": 2176, "lr": 1.897563707276393e-05}
{"train_loss": 0.11188636720180511, "global_step": 193705, "epoch": 2176, "lr": 1.8975182435823434e-05}
{"train_loss": 0.10122328996658325, "global_step": 193706, "epoch": 2176, "lr": 1.8974727803053828e-05}
{"train_loss": 0.06627406179904938, "global_step": 193707, "epoch": 2176, "lr": 1.897427317445519e-05}
{"train_loss": 0.052414149045944214, "global_step": 193708, "epoch": 2176, "lr": 1.89738185500276e-05}
{"train_loss": 0.03508104756474495, "global_step": 193709, "epoch": 2176, "lr": 1.897336392977108e-05}
{"train_loss": 0.06769485771656036, "global_step": 193710, "epoch": 2176, "lr": 1.8972909313685737e-05}
{"train_loss": 0.040441371500492096, "global_step": 193711, "epoch": 2176, "lr": 1.897245470177159e-05}
{"train_loss": 0.08340325951576233, "global_step": 193712, "epoch": 2176, "lr": 1.8972000094028735e-05}
{"train_loss": 0.189324289560318, "global_step": 193713, "epoch": 2176, "lr": 1.89715454904572e-05}
{"train_loss": 0.07725609838962555, "global_step": 193714, "epoch": 2176, "lr": 1.8971090891057076e-05}
{"train_loss": 0.06367102265357971, "global_step": 193715, "epoch": 2176, "lr": 1.8970636295828414e-05}
{"train_loss": 0.068331278860569, "global_step": 193716, "epoch": 2176, "lr": 1.8970181704771255e-05}
{"train_loss": 0.02447764202952385, "global_step": 193717, "epoch": 2176, "lr": 1.8969727117885695e-05}
{"train_loss": 0.03141723573207855, "global_step": 193718, "epoch": 2176, "lr": 1.8969272535171755e-05}
{"train_loss": 0.07298336178064346, "global_step": 193719, "epoch": 2176, "lr": 1.8968817956629538e-05}
{"train_loss": 0.0198338832706213, "global_step": 193720, "epoch": 2176, "lr": 1.8968363382259086e-05}
{"train_loss": 0.04774424433708191, "global_step": 193721, "epoch": 2176, "lr": 1.896790881206044e-05}
{"train_loss": 0.07582727074623108, "global_step": 193722, "epoch": 2176, "lr": 1.89674542460337e-05}
{"train_loss": 0.1528865247964859, "global_step": 193723, "epoch": 2176, "lr": 1.8966999684178887e-05}
{"train_loss": 0.11427780240774155, "global_step": 193724, "epoch": 2176, "lr": 1.896654512649609e-05}
{"train_loss": 0.0603005588054657, "global_step": 193725, "epoch": 2176, "lr": 1.8966090572985372e-05}
{"train_loss": 0.18455636501312256, "global_step": 193726, "epoch": 2176, "lr": 1.896563602364677e-05}
{"train_loss": 0.06770369410514832, "global_step": 193727, "epoch": 2176, "lr": 1.8965181478480374e-05}
{"train_loss": 0.08188345283269882, "global_step": 193728, "epoch": 2176, "lr": 1.8964726937486223e-05}
{"train_loss": 0.10284387320280075, "global_step": 193729, "epoch": 2176, "lr": 1.8964272400664396e-05}
{"train_loss": 0.0588931068778038, "global_step": 193730, "epoch": 2176, "lr": 1.896381786801493e-05}
{"train_loss": 0.05451201647520065, "global_step": 193731, "epoch": 2176, "lr": 1.8963363339537915e-05}
{"train_loss": 0.052889399230480194, "global_step": 193732, "epoch": 2176, "lr": 1.8962908815233382e-05}
{"train_loss": 0.1083609089255333, "global_step": 193733, "epoch": 2176, "lr": 1.8962454295101428e-05}
{"train_loss": 0.06144258379936218, "global_step": 193734, "epoch": 2176, "lr": 1.896199977914207e-05}
{"train_loss": 0.11326771229505539, "global_step": 193735, "epoch": 2176, "lr": 1.8961545267355414e-05}
{"train_loss": 0.07975033670663834, "global_step": 193736, "epoch": 2176, "lr": 1.8961090759741486e-05}
{"train_loss": 0.08898729830980301, "global_step": 193737, "epoch": 2176, "lr": 1.8960636256300374e-05}
{"train_loss": 0.07575346529483795, "global_step": 193738, "epoch": 2176, "lr": 1.896018175703211e-05}
{"train_loss": 0.0657404437661171, "global_step": 193739, "epoch": 2176, "lr": 1.895972726193678e-05}
{"train_loss": 0.0616728849709034, "global_step": 193740, "epoch": 2176, "lr": 1.895927277101443e-05}
{"train_loss": 0.07014166563749313, "global_step": 193741, "epoch": 2176, "lr": 1.8958818284265146e-05}
{"train_loss": 0.10377058386802673, "global_step": 193742, "epoch": 2176, "lr": 1.8958363801688943e-05}
{"train_loss": 0.1193227618932724, "global_step": 193743, "epoch": 2176, "lr": 1.8957909323285915e-05}
{"train_loss": 0.06047062203288078, "global_step": 193744, "epoch": 2176, "lr": 1.895745484905614e-05}
{"train_loss": 0.08121398091316223, "global_step": 193745, "epoch": 2176, "lr": 1.8957000378999635e-05}
{"train_loss": 0.08829908072948456, "global_step": 193746, "epoch": 2176, "lr": 1.8956545913116496e-05}
{"train_loss": 0.10594479739665985, "global_step": 193747, "epoch": 2176, "lr": 1.895609145140676e-05}
{"train_loss": 0.03833809867501259, "global_step": 193748, "epoch": 2176, "lr": 1.8955636993870517e-05}
{"train_loss": 0.04962631314992905, "global_step": 193749, "epoch": 2176, "lr": 1.895518254050779e-05}
{"train_loss": 0.12256774306297302, "global_step": 193750, "epoch": 2176, "lr": 1.895472809131868e-05}
{"train_loss": 0.046986885368824005, "global_step": 193751, "epoch": 2176, "lr": 1.8954273646303207e-05}
{"train_loss": 0.07672604566879487, "global_step": 193752, "epoch": 2176, "lr": 1.8953819205461472e-05, "val_loss": 7.865832805633545}
{"train_loss": 0.08000960946083069, "global_step": 193753, "epoch": 2177, "lr": 1.895336476879352e-05}
{"train_loss": 0.0441475510597229, "global_step": 193754, "epoch": 2177, "lr": 1.895291033629939e-05}
{"train_loss": 0.055582474917173386, "global_step": 193755, "epoch": 2177, "lr": 1.8952455907979182e-05}
{"train_loss": 0.06870289891958237, "global_step": 193756, "epoch": 2177, "lr": 1.895200148383292e-05}
{"train_loss": 0.05015940964221954, "global_step": 193757, "epoch": 2177, "lr": 1.89515470638607e-05}
{"train_loss": 0.028873475268483162, "global_step": 193758, "epoch": 2177, "lr": 1.8951092648062564e-05}
{"train_loss": 0.05129928141832352, "global_step": 193759, "epoch": 2177, "lr": 1.8950638236438563e-05}
{"train_loss": 0.035678163170814514, "global_step": 193760, "epoch": 2177, "lr": 1.8950183828988764e-05}
{"train_loss": 0.08627835661172867, "global_step": 193761, "epoch": 2177, "lr": 1.8949729425713252e-05}
{"train_loss": 0.10756780207157135, "global_step": 193762, "epoch": 2177, "lr": 1.8949275026612057e-05}
{"train_loss": 0.05032878369092941, "global_step": 193763, "epoch": 2177, "lr": 1.894882063168527e-05}
{"train_loss": 0.028710713610053062, "global_step": 193764, "epoch": 2177, "lr": 1.8948366240932918e-05}
{"train_loss": 0.03501731529831886, "global_step": 193765, "epoch": 2177, "lr": 1.89479118543551e-05}
{"train_loss": 0.03263617679476738, "global_step": 193766, "epoch": 2177, "lr": 1.8947457471951834e-05}
{"train_loss": 0.08170750737190247, "global_step": 193767, "epoch": 2177, "lr": 1.8947003093723225e-05}
{"train_loss": 0.06483547389507294, "global_step": 193768, "epoch": 2177, "lr": 1.8946548719669292e-05}
{"train_loss": 0.08403870463371277, "global_step": 193769, "epoch": 2177, "lr": 1.8946094349790138e-05}
{"train_loss": 0.16358743607997894, "global_step": 193770, "epoch": 2177, "lr": 1.8945639984085782e-05}
{"train_loss": 0.11374035477638245, "global_step": 193771, "epoch": 2177, "lr": 1.8945185622556326e-05}
{"train_loss": 0.08189310133457184, "global_step": 193772, "epoch": 2177, "lr": 1.8944731265201797e-05}
{"train_loss": 0.060450948774814606, "global_step": 193773, "epoch": 2177, "lr": 1.8944276912022284e-05}
{"train_loss": 0.07041150331497192, "global_step": 193774, "epoch": 2177, "lr": 1.8943822563017816e-05}
{"train_loss": 0.03285861387848854, "global_step": 193775, "epoch": 2177, "lr": 1.8943368218188496e-05}
{"train_loss": 0.05934366211295128, "global_step": 193776, "epoch": 2177, "lr": 1.8942913877534336e-05}
{"train_loss": 0.04522649943828583, "global_step": 193777, "epoch": 2177, "lr": 1.8942459541055446e-05}
{"train_loss": 0.02893727459013462, "global_step": 193778, "epoch": 2177, "lr": 1.8942005208751845e-05}
{"train_loss": 0.09146853536367416, "global_step": 193779, "epoch": 2177, "lr": 1.894155088062361e-05}
{"train_loss": 0.11395713686943054, "global_step": 193780, "epoch": 2177, "lr": 1.8941096556670822e-05}
{"train_loss": 0.05135265737771988, "global_step": 193781, "epoch": 2177, "lr": 1.8940642236893514e-05}
{"train_loss": 0.044933367520570755, "global_step": 193782, "epoch": 2177, "lr": 1.894018792129177e-05}
{"train_loss": 0.09143314510583878, "global_step": 193783, "epoch": 2177, "lr": 1.8939733609865624e-05}
{"train_loss": 0.10491416603326797, "global_step": 193784, "epoch": 2177, "lr": 1.893927930261517e-05}
{"train_loss": 0.12104272097349167, "global_step": 193785, "epoch": 2177, "lr": 1.8938824999540435e-05}
{"train_loss": 0.05733003094792366, "global_step": 193786, "epoch": 2177, "lr": 1.893837070064151e-05}
{"train_loss": 0.048960376530885696, "global_step": 193787, "epoch": 2177, "lr": 1.8937916405918426e-05}
{"train_loss": 0.0970599502325058, "global_step": 193788, "epoch": 2177, "lr": 1.893746211537128e-05}
{"train_loss": 0.04652733355760574, "global_step": 193789, "epoch": 2177, "lr": 1.893700782900009e-05}
{"train_loss": 0.0774589255452156, "global_step": 193790, "epoch": 2177, "lr": 1.893655354680497e-05}
{"train_loss": 0.09005901217460632, "global_step": 193791, "epoch": 2177, "lr": 1.893609926878594e-05}
{"train_loss": 0.06938564032316208, "global_step": 193792, "epoch": 2177, "lr": 1.8935644994943063e-05}
{"train_loss": 0.08112483471632004, "global_step": 193793, "epoch": 2177, "lr": 1.8935190725276424e-05}
{"train_loss": 0.019700683653354645, "global_step": 193794, "epoch": 2177, "lr": 1.893473645978605e-05}
{"train_loss": 0.03535519540309906, "global_step": 193795, "epoch": 2177, "lr": 1.8934282198472047e-05}
{"train_loss": 0.06692918390035629, "global_step": 193796, "epoch": 2177, "lr": 1.8933827941334425e-05}
{"train_loss": 0.07138458639383316, "global_step": 193797, "epoch": 2177, "lr": 1.8933373688373295e-05}
{"train_loss": 0.06255145370960236, "global_step": 193798, "epoch": 2177, "lr": 1.8932919439588675e-05}
{"train_loss": 0.09417436271905899, "global_step": 193799, "epoch": 2177, "lr": 1.8932465194980665e-05}
{"train_loss": 0.05204363539814949, "global_step": 193800, "epoch": 2177, "lr": 1.8932010954549284e-05}
{"train_loss": 0.0647498071193695, "global_step": 193801, "epoch": 2177, "lr": 1.8931556718294635e-05}
{"train_loss": 0.08277332037687302, "global_step": 193802, "epoch": 2177, "lr": 1.8931102486216744e-05}
{"train_loss": 0.0609847716987133, "global_step": 193803, "epoch": 2177, "lr": 1.8930648258315704e-05}
{"train_loss": 0.04440190643072128, "global_step": 193804, "epoch": 2177, "lr": 1.8930194034591542e-05}
{"train_loss": 0.09998518228530884, "global_step": 193805, "epoch": 2177, "lr": 1.892973981504435e-05}
{"train_loss": 0.08754870295524597, "global_step": 193806, "epoch": 2177, "lr": 1.8929285599674164e-05}
{"train_loss": 0.08583830296993256, "global_step": 193807, "epoch": 2177, "lr": 1.8928831388481072e-05}
{"train_loss": 0.08324345201253891, "global_step": 193808, "epoch": 2177, "lr": 1.8928377181465107e-05}
{"train_loss": 0.08900079131126404, "global_step": 193809, "epoch": 2177, "lr": 1.8927922978626355e-05}
{"train_loss": 0.05683991312980652, "global_step": 193810, "epoch": 2177, "lr": 1.8927468779964845e-05}
{"train_loss": 0.07982101291418076, "global_step": 193811, "epoch": 2177, "lr": 1.8927014585480685e-05}
{"train_loss": 0.07298580557107925, "global_step": 193812, "epoch": 2177, "lr": 1.8926560395173887e-05}
{"train_loss": 0.09065518528223038, "global_step": 193813, "epoch": 2177, "lr": 1.8926106209044548e-05}
{"train_loss": 0.05580335110425949, "global_step": 193814, "epoch": 2177, "lr": 1.89256520270927e-05}
{"train_loss": 0.06686278432607651, "global_step": 193815, "epoch": 2177, "lr": 1.892519784931842e-05}
{"train_loss": 0.06148885563015938, "global_step": 193816, "epoch": 2177, "lr": 1.8924743675721783e-05}
{"train_loss": 0.07330981642007828, "global_step": 193817, "epoch": 2177, "lr": 1.8924289506302822e-05}
{"train_loss": 0.08678852021694183, "global_step": 193818, "epoch": 2177, "lr": 1.892383534106163e-05}
{"train_loss": 0.08253715932369232, "global_step": 193819, "epoch": 2177, "lr": 1.892338117999823e-05}
{"train_loss": 0.04291371628642082, "global_step": 193820, "epoch": 2177, "lr": 1.892292702311272e-05}
{"train_loss": 0.05661964416503906, "global_step": 193821, "epoch": 2177, "lr": 1.8922472870405128e-05}
{"train_loss": 0.0913536325097084, "global_step": 193822, "epoch": 2177, "lr": 1.8922018721875544e-05}
{"train_loss": 0.05335130915045738, "global_step": 193823, "epoch": 2177, "lr": 1.8921564577524004e-05}
{"train_loss": 0.11700139194726944, "global_step": 193824, "epoch": 2177, "lr": 1.8921110437350593e-05}
{"train_loss": 0.0638655498623848, "global_step": 193825, "epoch": 2177, "lr": 1.8920656301355345e-05}
{"train_loss": 0.04008550941944122, "global_step": 193826, "epoch": 2177, "lr": 1.8920202169538355e-05}
{"train_loss": 0.04685655236244202, "global_step": 193827, "epoch": 2177, "lr": 1.8919748041899643e-05}
{"train_loss": 0.059434983879327774, "global_step": 193828, "epoch": 2177, "lr": 1.8919293918439313e-05}
{"train_loss": 0.05405702069401741, "global_step": 193829, "epoch": 2177, "lr": 1.891883979915739e-05}
{"train_loss": 0.06330973654985428, "global_step": 193830, "epoch": 2177, "lr": 1.8918385684053963e-05}
{"train_loss": 0.03817250579595566, "global_step": 193831, "epoch": 2177, "lr": 1.8917931573129084e-05}
{"train_loss": 0.10835392773151398, "global_step": 193832, "epoch": 2177, "lr": 1.891747746638279e-05}
{"train_loss": 0.10549410432577133, "global_step": 193833, "epoch": 2177, "lr": 1.8917023363815177e-05}
{"train_loss": 0.10250779986381531, "global_step": 193834, "epoch": 2177, "lr": 1.8916569265426274e-05}
{"train_loss": 0.07203241437673569, "global_step": 193835, "epoch": 2177, "lr": 1.891611517121618e-05}
{"train_loss": 0.05298147350549698, "global_step": 193836, "epoch": 2177, "lr": 1.8915661081184922e-05}
{"train_loss": 0.055542368441820145, "global_step": 193837, "epoch": 2177, "lr": 1.8915206995332585e-05}
{"train_loss": 0.11632725596427917, "global_step": 193838, "epoch": 2177, "lr": 1.8914752913659206e-05}
{"train_loss": 0.06337514519691467, "global_step": 193839, "epoch": 2177, "lr": 1.8914298836164878e-05}
{"train_loss": 0.07045856863260269, "global_step": 193840, "epoch": 2177, "lr": 1.8913844762849624e-05}
{"train_loss": 0.06911767753406187, "global_step": 193841, "epoch": 2177, "lr": 1.8913390693713547e-05, "val_loss": 7.772541522979736}
{"train_loss": 0.06410852819681168, "global_step": 193842, "epoch": 2178, "lr": 1.891293662875666e-05}
{"train_loss": 0.08265633136034012, "global_step": 193843, "epoch": 2178, "lr": 1.8912482567979067e-05}
{"train_loss": 0.07095661014318466, "global_step": 193844, "epoch": 2178, "lr": 1.89120285113808e-05}
{"train_loss": 0.059058357030153275, "global_step": 193845, "epoch": 2178, "lr": 1.891157445896195e-05}
{"train_loss": 0.0738871619105339, "global_step": 193846, "epoch": 2178, "lr": 1.891112041072254e-05}
{"train_loss": 0.03915562853217125, "global_step": 193847, "epoch": 2178, "lr": 1.8910666366662665e-05}
{"train_loss": 0.06915880739688873, "global_step": 193848, "epoch": 2178, "lr": 1.891021232678236e-05}
{"train_loss": 0.05778703838586807, "global_step": 193849, "epoch": 2178, "lr": 1.8909758291081715e-05}
{"train_loss": 0.06941080838441849, "global_step": 193850, "epoch": 2178, "lr": 1.8909304259560758e-05}
{"train_loss": 0.07864968478679657, "global_step": 193851, "epoch": 2178, "lr": 1.890885023221956e-05}
{"train_loss": 0.08912934362888336, "global_step": 193852, "epoch": 2178, "lr": 1.890839620905821e-05}
{"train_loss": 0.09144990146160126, "global_step": 193853, "epoch": 2178, "lr": 1.8907942190076728e-05}
{"train_loss": 0.11868744343519211, "global_step": 193854, "epoch": 2178, "lr": 1.8907488175275212e-05}
{"train_loss": 0.04172634333372116, "global_step": 193855, "epoch": 2178, "lr": 1.8907034164653688e-05}
{"train_loss": 0.12979155778884888, "global_step": 193856, "epoch": 2178, "lr": 1.890658015821225e-05}
{"train_loss": 0.023561587557196617, "global_step": 193857, "epoch": 2178, "lr": 1.8906126155950925e-05}
{"train_loss": 0.08522673696279526, "global_step": 193858, "epoch": 2178, "lr": 1.8905672157869815e-05}
{"train_loss": 0.0374862439930439, "global_step": 193859, "epoch": 2178, "lr": 1.8905218163968942e-05}
{"train_loss": 0.04001054912805557, "global_step": 193860, "epoch": 2178, "lr": 1.890476417424839e-05}
{"train_loss": 0.05264252796769142, "global_step": 193861, "epoch": 2178, "lr": 1.890431018870821e-05}
{"train_loss": 0.05470651015639305, "global_step": 193862, "epoch": 2178, "lr": 1.8903856207348475e-05}
{"train_loss": 0.04591790586709976, "global_step": 193863, "epoch": 2178, "lr": 1.890340223016922e-05}
{"train_loss": 0.07033322751522064, "global_step": 193864, "epoch": 2178, "lr": 1.8902948257170542e-05}
{"train_loss": 0.06731199473142624, "global_step": 193865, "epoch": 2178, "lr": 1.890249428835247e-05}
{"train_loss": 0.07628962397575378, "global_step": 193866, "epoch": 2178, "lr": 1.8902040323715093e-05}
{"train_loss": 0.07171674072742462, "global_step": 193867, "epoch": 2178, "lr": 1.890158636325844e-05}
{"train_loss": 0.03135877847671509, "global_step": 193868, "epoch": 2178, "lr": 1.890113240698261e-05}
{"train_loss": 0.0707290917634964, "global_step": 193869, "epoch": 2178, "lr": 1.8900678454887636e-05}
{"train_loss": 0.08262816816568375, "global_step": 193870, "epoch": 2178, "lr": 1.890022450697358e-05}
{"train_loss": 0.10221955180168152, "global_step": 193871, "epoch": 2178, "lr": 1.889977056324052e-05}
{"train_loss": 0.08891893923282623, "global_step": 193872, "epoch": 2178, "lr": 1.889931662368849e-05}
{"train_loss": 0.04334762692451477, "global_step": 193873, "epoch": 2178, "lr": 1.8898862688317586e-05}
{"train_loss": 0.05666773393750191, "global_step": 193874, "epoch": 2178, "lr": 1.8898408757127834e-05}
{"train_loss": 0.09751368314027786, "global_step": 193875, "epoch": 2178, "lr": 1.889795483011933e-05}
{"train_loss": 0.07137376070022583, "global_step": 193876, "epoch": 2178, "lr": 1.8897500907292098e-05}
{"train_loss": 0.10228948295116425, "global_step": 193877, "epoch": 2178, "lr": 1.889704698864624e-05}
{"train_loss": 0.09552145004272461, "global_step": 193878, "epoch": 2178, "lr": 1.8896593074181772e-05}
{"train_loss": 0.1086127758026123, "global_step": 193879, "epoch": 2178, "lr": 1.8896139163898795e-05}
{"train_loss": 0.05392434075474739, "global_step": 193880, "epoch": 2178, "lr": 1.8895685257797342e-05}
{"train_loss": 0.0406172052025795, "global_step": 193881, "epoch": 2178, "lr": 1.8895231355877498e-05}
{"train_loss": 0.08028078824281693, "global_step": 193882, "epoch": 2178, "lr": 1.889477745813929e-05}
{"train_loss": 0.057374756783246994, "global_step": 193883, "epoch": 2178, "lr": 1.889432356458282e-05}
{"train_loss": 0.07723461836576462, "global_step": 193884, "epoch": 2178, "lr": 1.8893869675208113e-05}
{"train_loss": 0.05546743795275688, "global_step": 193885, "epoch": 2178, "lr": 1.8893415790015262e-05}
{"train_loss": 0.07622860372066498, "global_step": 193886, "epoch": 2178, "lr": 1.8892961909004297e-05}
{"train_loss": 0.05696757510304451, "global_step": 193887, "epoch": 2178, "lr": 1.8892508032175293e-05}
{"train_loss": 0.062080174684524536, "global_step": 193888, "epoch": 2178, "lr": 1.8892054159528334e-05}
{"train_loss": 0.05518779158592224, "global_step": 193889, "epoch": 2178, "lr": 1.8891600291063432e-05}
{"train_loss": 0.05146784707903862, "global_step": 193890, "epoch": 2178, "lr": 1.8891146426780697e-05}
{"train_loss": 0.06122890114784241, "global_step": 193891, "epoch": 2178, "lr": 1.8890692566680153e-05}
{"train_loss": 0.06060221418738365, "global_step": 193892, "epoch": 2178, "lr": 1.8890238710761892e-05}
{"train_loss": 0.05391278862953186, "global_step": 193893, "epoch": 2178, "lr": 1.8889784859025944e-05}
{"train_loss": 0.08678648620843887, "global_step": 193894, "epoch": 2178, "lr": 1.88893310114724e-05}
{"train_loss": 0.06906510144472122, "global_step": 193895, "epoch": 2178, "lr": 1.8888877168101286e-05}
{"train_loss": 0.09171662479639053, "global_step": 193896, "epoch": 2178, "lr": 1.888842332891271e-05}
{"train_loss": 0.03869610279798508, "global_step": 193897, "epoch": 2178, "lr": 1.888796949390668e-05}
{"train_loss": 0.09750869125127792, "global_step": 193898, "epoch": 2178, "lr": 1.8887515663083306e-05}
{"train_loss": 0.08395050466060638, "global_step": 193899, "epoch": 2178, "lr": 1.8887061836442604e-05}
{"train_loss": 0.07343284785747528, "global_step": 193900, "epoch": 2178, "lr": 1.888660801398468e-05}
{"train_loss": 0.04169844090938568, "global_step": 193901, "epoch": 2178, "lr": 1.8886154195709554e-05}
{"train_loss": 0.1513921618461609, "global_step": 193902, "epoch": 2178, "lr": 1.8885700381617323e-05}
{"train_loss": 0.07303492724895477, "global_step": 193903, "epoch": 2178, "lr": 1.888524657170801e-05}
{"train_loss": 0.06563717871904373, "global_step": 193904, "epoch": 2178, "lr": 1.8884792765981714e-05}
{"train_loss": 0.06214812025427818, "global_step": 193905, "epoch": 2178, "lr": 1.8884338964438458e-05}
{"train_loss": 0.047741688787937164, "global_step": 193906, "epoch": 2178, "lr": 1.8883885167078346e-05}
{"train_loss": 0.08456503599882126, "global_step": 193907, "epoch": 2178, "lr": 1.88834313739014e-05}
{"train_loss": 0.04083767533302307, "global_step": 193908, "epoch": 2178, "lr": 1.888297758490771e-05}
{"train_loss": 0.06144954636693001, "global_step": 193909, "epoch": 2178, "lr": 1.888252380009733e-05}
{"train_loss": 0.05350454896688461, "global_step": 193910, "epoch": 2178, "lr": 1.8882070019470294e-05}
{"train_loss": 0.024783244356513023, "global_step": 193911, "epoch": 2178, "lr": 1.88816162430267e-05}
{"train_loss": 0.058026522397994995, "global_step": 193912, "epoch": 2178, "lr": 1.8881162470766578e-05}
{"train_loss": 0.09327654540538788, "global_step": 193913, "epoch": 2178, "lr": 1.8880708702690024e-05}
{"train_loss": 0.10019059479236603, "global_step": 193914, "epoch": 2178, "lr": 1.8880254938797058e-05}
{"train_loss": 0.046811606734991074, "global_step": 193915, "epoch": 2178, "lr": 1.8879801179087788e-05}
{"train_loss": 0.07642120122909546, "global_step": 193916, "epoch": 2178, "lr": 1.8879347423562226e-05}
{"train_loss": 0.03967510536313057, "global_step": 193917, "epoch": 2178, "lr": 1.8878893672220472e-05}
{"train_loss": 0.06783835589885712, "global_step": 193918, "epoch": 2178, "lr": 1.8878439925062556e-05}
{"train_loss": 0.03488108143210411, "global_step": 193919, "epoch": 2178, "lr": 1.8877986182088576e-05}
{"train_loss": 0.06271137297153473, "global_step": 193920, "epoch": 2178, "lr": 1.8877532443298546e-05}
{"train_loss": 0.06337025761604309, "global_step": 193921, "epoch": 2178, "lr": 1.8877078708692574e-05}
{"train_loss": 0.05406555160880089, "global_step": 193922, "epoch": 2178, "lr": 1.8876624978270684e-05}
{"train_loss": 0.0378592275083065, "global_step": 193923, "epoch": 2178, "lr": 1.8876171252032964e-05}
{"train_loss": 0.025007296353578568, "global_step": 193924, "epoch": 2178, "lr": 1.8875717529979455e-05}
{"train_loss": 0.08023668080568314, "global_step": 193925, "epoch": 2178, "lr": 1.8875263812110217e-05}
{"train_loss": 0.05448399484157562, "global_step": 193926, "epoch": 2178, "lr": 1.8874810098425345e-05}
{"train_loss": 0.1076052114367485, "global_step": 193927, "epoch": 2178, "lr": 1.887435638892485e-05}
{"train_loss": 0.07815355807542801, "global_step": 193928, "epoch": 2178, "lr": 1.8873902683608848e-05}
{"train_loss": 0.07041876018047333, "global_step": 193929, "epoch": 2178, "lr": 1.8873448982477343e-05}
{"train_loss": 0.06743154101324884, "global_step": 193930, "epoch": 2178, "lr": 1.8872995285530443e-05, "val_loss": 8.035806655883789}
{"train_loss": 0.10535498708486557, "global_step": 193931, "epoch": 2179, "lr": 1.8872541592768173e-05}
{"train_loss": 0.02685142122209072, "global_step": 193932, "epoch": 2179, "lr": 1.887208790419063e-05}
{"train_loss": 0.03677307814359665, "global_step": 193933, "epoch": 2179, "lr": 1.8871634219797833e-05}
{"train_loss": 0.052043572068214417, "global_step": 193934, "epoch": 2179, "lr": 1.8871180539589883e-05}
{"train_loss": 0.043261848390102386, "global_step": 193935, "epoch": 2179, "lr": 1.8870726863566813e-05}
{"train_loss": 0.053330324590206146, "global_step": 193936, "epoch": 2179, "lr": 1.8870273191728704e-05}
{"train_loss": 0.06846655905246735, "global_step": 193937, "epoch": 2179, "lr": 1.8869819524075594e-05}
{"train_loss": 0.05243341997265816, "global_step": 193938, "epoch": 2179, "lr": 1.8869365860607575e-05}
{"train_loss": 0.10336530953645706, "global_step": 193939, "epoch": 2179, "lr": 1.8868912201324672e-05}
{"train_loss": 0.07632957398891449, "global_step": 193940, "epoch": 2179, "lr": 1.886845854622698e-05}
{"train_loss": 0.04258851706981659, "global_step": 193941, "epoch": 2179, "lr": 1.8868004895314527e-05}
{"train_loss": 0.10314485430717468, "global_step": 193942, "epoch": 2179, "lr": 1.886755124858741e-05}
{"train_loss": 0.07795766741037369, "global_step": 193943, "epoch": 2179, "lr": 1.8867097606045653e-05}
{"train_loss": 0.07231815904378891, "global_step": 193944, "epoch": 2179, "lr": 1.8866643967689356e-05}
{"train_loss": 0.09639837592840195, "global_step": 193945, "epoch": 2179, "lr": 1.886619033351854e-05}
{"train_loss": 0.09850719571113586, "global_step": 193946, "epoch": 2179, "lr": 1.8865736703533297e-05}
{"train_loss": 0.04413745552301407, "global_step": 193947, "epoch": 2179, "lr": 1.8865283077733688e-05}
{"train_loss": 0.06576802581548691, "global_step": 193948, "epoch": 2179, "lr": 1.8864829456119736e-05}
{"train_loss": 0.04012254625558853, "global_step": 193949, "epoch": 2179, "lr": 1.8864375838691546e-05}
{"train_loss": 0.09496454894542694, "global_step": 193950, "epoch": 2179, "lr": 1.886392222544915e-05}
{"train_loss": 0.08121451735496521, "global_step": 193951, "epoch": 2179, "lr": 1.886346861639263e-05}
{"train_loss": 0.0485641248524189, "global_step": 193952, "epoch": 2179, "lr": 1.8863015011522023e-05}
{"train_loss": 0.0685042142868042, "global_step": 193953, "epoch": 2179, "lr": 1.886256141083742e-05}
{"train_loss": 0.11777269095182419, "global_step": 193954, "epoch": 2179, "lr": 1.8862107814338854e-05}
{"train_loss": 0.02898220159113407, "global_step": 193955, "epoch": 2179, "lr": 1.886165422202641e-05}
{"train_loss": 0.08182039856910706, "global_step": 193956, "epoch": 2179, "lr": 1.8861200633900117e-05}
{"train_loss": 0.0640355721116066, "global_step": 193957, "epoch": 2179, "lr": 1.8860747049960083e-05}
{"train_loss": 0.11627180129289627, "global_step": 193958, "epoch": 2179, "lr": 1.8860293470206313e-05}
{"train_loss": 0.03342081233859062, "global_step": 193959, "epoch": 2179, "lr": 1.8859839894638925e-05}
{"train_loss": 0.09941036999225616, "global_step": 193960, "epoch": 2179, "lr": 1.8859386323257926e-05}
{"train_loss": 0.05617310479283333, "global_step": 193961, "epoch": 2179, "lr": 1.885893275606341e-05}
{"train_loss": 0.05329059809446335, "global_step": 193962, "epoch": 2179, "lr": 1.8858479193055446e-05}
{"train_loss": 0.05683168023824692, "global_step": 193963, "epoch": 2179, "lr": 1.8858025634234062e-05}
{"train_loss": 0.055218521505594254, "global_step": 193964, "epoch": 2179, "lr": 1.8857572079599356e-05}
{"train_loss": 0.09641217440366745, "global_step": 193965, "epoch": 2179, "lr": 1.8857118529151347e-05}
{"train_loss": 0.045713599771261215, "global_step": 193966, "epoch": 2179, "lr": 1.8856664982890144e-05}
{"train_loss": 0.05348530411720276, "global_step": 193967, "epoch": 2179, "lr": 1.8856211440815757e-05}
{"train_loss": 0.052203673869371414, "global_step": 193968, "epoch": 2179, "lr": 1.8855757902928295e-05}
{"train_loss": 0.12217580527067184, "global_step": 193969, "epoch": 2179, "lr": 1.8855304369227777e-05}
{"train_loss": 0.07591090351343155, "global_step": 193970, "epoch": 2179, "lr": 1.8854850839714306e-05}
{"train_loss": 0.10432788729667664, "global_step": 193971, "epoch": 2179, "lr": 1.8854397314387902e-05}
{"train_loss": 0.08846073597669601, "global_step": 193972, "epoch": 2179, "lr": 1.8853943793248656e-05}
{"train_loss": 0.08106803148984909, "global_step": 193973, "epoch": 2179, "lr": 1.885349027629661e-05}
{"train_loss": 0.08889462053775787, "global_step": 193974, "epoch": 2179, "lr": 1.8853036763531845e-05}
{"train_loss": 0.07585611194372177, "global_step": 193975, "epoch": 2179, "lr": 1.8852583254954386e-05}
{"train_loss": 0.065195731818676, "global_step": 193976, "epoch": 2179, "lr": 1.8852129750564346e-05}
{"train_loss": 0.06366684287786484, "global_step": 193977, "epoch": 2179, "lr": 1.885167625036173e-05}
{"train_loss": 0.033649593591690063, "global_step": 193978, "epoch": 2179, "lr": 1.8851222754346638e-05}
{"train_loss": 0.03396360203623772, "global_step": 193979, "epoch": 2179, "lr": 1.8850769262519123e-05}
{"train_loss": 0.15095731616020203, "global_step": 193980, "epoch": 2179, "lr": 1.885031577487925e-05}
{"train_loss": 0.10884656012058258, "global_step": 193981, "epoch": 2179, "lr": 1.8849862291427056e-05}
{"train_loss": 0.055481813848018646, "global_step": 193982, "epoch": 2179, "lr": 1.884940881216264e-05}
{"train_loss": 0.04041196405887604, "global_step": 193983, "epoch": 2179, "lr": 1.8848955337086017e-05}
{"train_loss": 0.05548664554953575, "global_step": 193984, "epoch": 2179, "lr": 1.8848501866197288e-05}
{"train_loss": 0.07668765634298325, "global_step": 193985, "epoch": 2179, "lr": 1.8848048399496504e-05}
{"train_loss": 0.07750480622053146, "global_step": 193986, "epoch": 2179, "lr": 1.88475949369837e-05}
{"train_loss": 0.07250514626502991, "global_step": 193987, "epoch": 2179, "lr": 1.884714147865897e-05}
{"train_loss": 0.08607053756713867, "global_step": 193988, "epoch": 2179, "lr": 1.8846688024522353e-05}
{"train_loss": 0.062439169734716415, "global_step": 193989, "epoch": 2179, "lr": 1.884623457457393e-05}
{"train_loss": 0.07793928682804108, "global_step": 193990, "epoch": 2179, "lr": 1.8845781128813738e-05}
{"train_loss": 0.09287405014038086, "global_step": 193991, "epoch": 2179, "lr": 1.8845327687241863e-05}
{"train_loss": 0.07440900057554245, "global_step": 193992, "epoch": 2179, "lr": 1.8844874249858347e-05}
{"train_loss": 0.03087938390672207, "global_step": 193993, "epoch": 2179, "lr": 1.8844420816663265e-05}
{"train_loss": 0.023008933290839195, "global_step": 193994, "epoch": 2179, "lr": 1.8843967387656657e-05}
{"train_loss": 0.042502131313085556, "global_step": 193995, "epoch": 2179, "lr": 1.8843513962838617e-05}
{"train_loss": 0.08400028198957443, "global_step": 193996, "epoch": 2179, "lr": 1.8843060542209167e-05}
{"train_loss": 0.037798043340444565, "global_step": 193997, "epoch": 2179, "lr": 1.884260712576839e-05}
{"train_loss": 0.07172008603811264, "global_step": 193998, "epoch": 2179, "lr": 1.884215371351636e-05}
{"train_loss": 0.05859280005097389, "global_step": 193999, "epoch": 2179, "lr": 1.8841700305453108e-05}
{"train_loss": 0.03538387641310692, "global_step": 194000, "epoch": 2179, "lr": 1.8841246901578722e-05}
{"train_loss": 0.06634597480297089, "global_step": 194001, "epoch": 2179, "lr": 1.884079350189324e-05}
{"train_loss": 0.0732167512178421, "global_step": 194002, "epoch": 2179, "lr": 1.8840340106396743e-05}
{"train_loss": 0.0851055458188057, "global_step": 194003, "epoch": 2179, "lr": 1.883988671508927e-05}
{"train_loss": 0.07663612812757492, "global_step": 194004, "epoch": 2179, "lr": 1.883943332797091e-05}
{"train_loss": 0.05156967043876648, "global_step": 194005, "epoch": 2179, "lr": 1.8838979945041696e-05}
{"train_loss": 0.0826457142829895, "global_step": 194006, "epoch": 2179, "lr": 1.8838526566301712e-05}
{"train_loss": 0.10157924890518188, "global_step": 194007, "epoch": 2179, "lr": 1.8838073191750995e-05}
{"train_loss": 0.09215066581964493, "global_step": 194008, "epoch": 2179, "lr": 1.8837619821389634e-05}
{"train_loss": 0.03941096365451813, "global_step": 194009, "epoch": 2179, "lr": 1.8837166455217658e-05}
{"train_loss": 0.06572160124778748, "global_step": 194010, "epoch": 2179, "lr": 1.8836713093235164e-05}
{"train_loss": 0.04579683393239975, "global_step": 194011, "epoch": 2179, "lr": 1.883625973544218e-05}
{"train_loss": 0.05756863206624985, "global_step": 194012, "epoch": 2179, "lr": 1.8835806381838793e-05}
{"train_loss": 0.05494621396064758, "global_step": 194013, "epoch": 2179, "lr": 1.8835353032425035e-05}
{"train_loss": 0.05946177616715431, "global_step": 194014, "epoch": 2179, "lr": 1.8834899687200987e-05}
{"train_loss": 0.05711125209927559, "global_step": 194015, "epoch": 2179, "lr": 1.8834446346166722e-05}
{"train_loss": 0.09098789840936661, "global_step": 194016, "epoch": 2179, "lr": 1.883399300932227e-05}
{"train_loss": 0.05230952426791191, "global_step": 194017, "epoch": 2179, "lr": 1.883353967666773e-05}
{"train_loss": 0.08156410604715347, "global_step": 194018, "epoch": 2179, "lr": 1.8833086348203128e-05}
{"train_loss": 0.06825391395708148, "global_step": 194019, "epoch": 2179, "lr": 1.8832633023928532e-05, "val_loss": 8.064395904541016}
{"train_loss": 0.07985720783472061, "global_step": 194020, "epoch": 2180, "lr": 1.8832179703844015e-05}
{"train_loss": 0.03690645471215248, "global_step": 194021, "epoch": 2180, "lr": 1.8831726387949624e-05}
{"train_loss": 0.0781928151845932, "global_step": 194022, "epoch": 2180, "lr": 1.8831273076245444e-05}
{"train_loss": 0.08383987098932266, "global_step": 194023, "epoch": 2180, "lr": 1.8830819768731495e-05}
{"train_loss": 0.07802490890026093, "global_step": 194024, "epoch": 2180, "lr": 1.8830366465407883e-05}
{"train_loss": 0.08169001340866089, "global_step": 194025, "epoch": 2180, "lr": 1.8829913166274654e-05}
{"train_loss": 0.09223858267068863, "global_step": 194026, "epoch": 2180, "lr": 1.882945987133184e-05}
{"train_loss": 0.083967424929142, "global_step": 194027, "epoch": 2180, "lr": 1.8829006580579545e-05}
{"train_loss": 0.05695783719420433, "global_step": 194028, "epoch": 2180, "lr": 1.8828553294017793e-05}
{"train_loss": 0.10196222364902496, "global_step": 194029, "epoch": 2180, "lr": 1.8828100011646676e-05}
{"train_loss": 0.05861999839544296, "global_step": 194030, "epoch": 2180, "lr": 1.8827646733466224e-05}
{"train_loss": 0.04917382821440697, "global_step": 194031, "epoch": 2180, "lr": 1.8827193459476538e-05}
{"train_loss": 0.051239315420389175, "global_step": 194032, "epoch": 2180, "lr": 1.8826740189677633e-05}
{"train_loss": 0.05789021775126457, "global_step": 194033, "epoch": 2180, "lr": 1.8826286924069596e-05}
{"train_loss": 0.06033734604716301, "global_step": 194034, "epoch": 2180, "lr": 1.8825833662652498e-05}
{"train_loss": 0.09779274463653564, "global_step": 194035, "epoch": 2180, "lr": 1.8825380405426374e-05}
{"train_loss": 0.036620836704969406, "global_step": 194036, "epoch": 2180, "lr": 1.8824927152391314e-05}
{"train_loss": 0.10395058244466782, "global_step": 194037, "epoch": 2180, "lr": 1.882447390354734e-05}
{"train_loss": 0.08504889160394669, "global_step": 194038, "epoch": 2180, "lr": 1.882402065889456e-05}
{"train_loss": 0.13586705923080444, "global_step": 194039, "epoch": 2180, "lr": 1.8823567418432992e-05}
{"train_loss": 0.07795654237270355, "global_step": 194040, "epoch": 2180, "lr": 1.8823114182162727e-05}
{"train_loss": 0.09001213312149048, "global_step": 194041, "epoch": 2180, "lr": 1.8822660950083805e-05}
{"train_loss": 0.031285542994737625, "global_step": 194042, "epoch": 2180, "lr": 1.8822207722196306e-05}
{"train_loss": 0.09128163754940033, "global_step": 194043, "epoch": 2180, "lr": 1.8821754498500266e-05}
{"train_loss": 0.04047394543886185, "global_step": 194044, "epoch": 2180, "lr": 1.8821301278995785e-05}
{"train_loss": 0.07193674892187119, "global_step": 194045, "epoch": 2180, "lr": 1.8820848063682874e-05}
{"train_loss": 0.05845143646001816, "global_step": 194046, "epoch": 2180, "lr": 1.882039485256164e-05}
{"train_loss": 0.07171221822500229, "global_step": 194047, "epoch": 2180, "lr": 1.8819941645632112e-05}
{"train_loss": 0.10798874497413635, "global_step": 194048, "epoch": 2180, "lr": 1.8819488442894374e-05}
{"train_loss": 0.025409284979104996, "global_step": 194049, "epoch": 2180, "lr": 1.881903524434846e-05}
{"train_loss": 0.0598565973341465, "global_step": 194050, "epoch": 2180, "lr": 1.881858204999446e-05}
{"train_loss": 0.04231852665543556, "global_step": 194051, "epoch": 2180, "lr": 1.8818128859832413e-05}
{"train_loss": 0.060005124658346176, "global_step": 194052, "epoch": 2180, "lr": 1.881767567386238e-05}
{"train_loss": 0.043549954891204834, "global_step": 194053, "epoch": 2180, "lr": 1.881722249208445e-05}
{"train_loss": 0.04621446877717972, "global_step": 194054, "epoch": 2180, "lr": 1.881676931449865e-05}
{"train_loss": 0.09132227301597595, "global_step": 194055, "epoch": 2180, "lr": 1.8816316141105072e-05}
{"train_loss": 0.07336689531803131, "global_step": 194056, "epoch": 2180, "lr": 1.881586297190374e-05}
{"train_loss": 0.051779285073280334, "global_step": 194057, "epoch": 2180, "lr": 1.8815409806894757e-05}
{"train_loss": 0.061474449932575226, "global_step": 194058, "epoch": 2180, "lr": 1.8814956646078156e-05}
{"train_loss": 0.054875295609235764, "global_step": 194059, "epoch": 2180, "lr": 1.8814503489453993e-05}
{"train_loss": 0.1298152357339859, "global_step": 194060, "epoch": 2180, "lr": 1.8814050337022355e-05}
{"train_loss": 0.028708728030323982, "global_step": 194061, "epoch": 2180, "lr": 1.8813597188783265e-05}
{"train_loss": 0.0781414806842804, "global_step": 194062, "epoch": 2180, "lr": 1.881314404473683e-05}
{"train_loss": 0.07063723355531693, "global_step": 194063, "epoch": 2180, "lr": 1.881269090488309e-05}
{"train_loss": 0.05210354924201965, "global_step": 194064, "epoch": 2180, "lr": 1.881223776922208e-05}
{"train_loss": 0.07675161957740784, "global_step": 194065, "epoch": 2180, "lr": 1.8811784637753903e-05}
{"train_loss": 0.025721268728375435, "global_step": 194066, "epoch": 2180, "lr": 1.881133151047859e-05}
{"train_loss": 0.10204153507947922, "global_step": 194067, "epoch": 2180, "lr": 1.8810878387396224e-05}
{"train_loss": 0.09000076353549957, "global_step": 194068, "epoch": 2180, "lr": 1.8810425268506843e-05}
{"train_loss": 0.05278896540403366, "global_step": 194069, "epoch": 2180, "lr": 1.880997215381052e-05}
{"train_loss": 0.057080306112766266, "global_step": 194070, "epoch": 2180, "lr": 1.8809519043307326e-05}
{"train_loss": 0.10043168812990189, "global_step": 194071, "epoch": 2180, "lr": 1.88090659369973e-05}
{"train_loss": 0.017116045579314232, "global_step": 194072, "epoch": 2180, "lr": 1.8808612834880536e-05}
{"train_loss": 0.05319678410887718, "global_step": 194073, "epoch": 2180, "lr": 1.8808159736957047e-05}
{"train_loss": 0.04688755050301552, "global_step": 194074, "epoch": 2180, "lr": 1.8807706643226946e-05}
{"train_loss": 0.03732161223888397, "global_step": 194075, "epoch": 2180, "lr": 1.8807253553690252e-05}
{"train_loss": 0.10449505597352982, "global_step": 194076, "epoch": 2180, "lr": 1.8806800468347052e-05}
{"train_loss": 0.06868670135736465, "global_step": 194077, "epoch": 2180, "lr": 1.8806347387197387e-05}
{"train_loss": 0.08878197520971298, "global_step": 194078, "epoch": 2180, "lr": 1.8805894310241344e-05}
{"train_loss": 0.047637809067964554, "global_step": 194079, "epoch": 2180, "lr": 1.8805441237478948e-05}
{"train_loss": 0.03169155865907669, "global_step": 194080, "epoch": 2180, "lr": 1.8804988168910304e-05}
{"train_loss": 0.09540300071239471, "global_step": 194081, "epoch": 2180, "lr": 1.8804535104535424e-05}
{"train_loss": 0.07827279716730118, "global_step": 194082, "epoch": 2180, "lr": 1.8804082044354415e-05}
{"train_loss": 0.09374997019767761, "global_step": 194083, "epoch": 2180, "lr": 1.88036289883673e-05}
{"train_loss": 0.04631306976079941, "global_step": 194084, "epoch": 2180, "lr": 1.8803175936574174e-05}
{"train_loss": 0.07232048362493515, "global_step": 194085, "epoch": 2180, "lr": 1.8802722888975067e-05}
{"train_loss": 0.07132256031036377, "global_step": 194086, "epoch": 2180, "lr": 1.880226984557007e-05}
{"train_loss": 0.05656362324953079, "global_step": 194087, "epoch": 2180, "lr": 1.880181680635921e-05}
{"train_loss": 0.05348017066717148, "global_step": 194088, "epoch": 2180, "lr": 1.8801363771342555e-05}
{"train_loss": 0.06492684781551361, "global_step": 194089, "epoch": 2180, "lr": 1.8800910740520206e-05}
{"train_loss": 0.07512810826301575, "global_step": 194090, "epoch": 2180, "lr": 1.880045771389217e-05}
{"train_loss": 0.07293254882097244, "global_step": 194091, "epoch": 2180, "lr": 1.8800004691458555e-05}
{"train_loss": 0.048702459782361984, "global_step": 194092, "epoch": 2180, "lr": 1.879955167321938e-05}
{"train_loss": 0.029073935002088547, "global_step": 194093, "epoch": 2180, "lr": 1.879909865917474e-05}
{"train_loss": 0.04712020978331566, "global_step": 194094, "epoch": 2180, "lr": 1.8798645649324666e-05}
{"train_loss": 0.056142792105674744, "global_step": 194095, "epoch": 2180, "lr": 1.8798192643669248e-05}
{"train_loss": 0.0790979415178299, "global_step": 194096, "epoch": 2180, "lr": 1.8797739642208534e-05}
{"train_loss": 0.0418303981423378, "global_step": 194097, "epoch": 2180, "lr": 1.879728664494257e-05}
{"train_loss": 0.06145678460597992, "global_step": 194098, "epoch": 2180, "lr": 1.8796833651871443e-05}
{"train_loss": 0.03816661238670349, "global_step": 194099, "epoch": 2180, "lr": 1.879638066299519e-05}
{"train_loss": 0.09179134666919708, "global_step": 194100, "epoch": 2180, "lr": 1.8795927678313896e-05}
{"train_loss": 0.0512128621339798, "global_step": 194101, "epoch": 2180, "lr": 1.8795474697827608e-05}
{"train_loss": 0.0602564811706543, "global_step": 194102, "epoch": 2180, "lr": 1.8795021721536373e-05}
{"train_loss": 0.09831049293279648, "global_step": 194103, "epoch": 2180, "lr": 1.8794568749440285e-05}
{"train_loss": 0.12865719199180603, "global_step": 194104, "epoch": 2180, "lr": 1.8794115781539368e-05}
{"train_loss": 0.04631778225302696, "global_step": 194105, "epoch": 2180, "lr": 1.87936628178337e-05}
{"train_loss": 0.059667885303497314, "global_step": 194106, "epoch": 2180, "lr": 1.8793209858323367e-05}
{"train_loss": 0.06319587677717209, "global_step": 194107, "epoch": 2180, "lr": 1.8792756903008385e-05}
{"train_loss": 0.06714063474720114, "global_step": 194108, "epoch": 2180, "lr": 1.8792303951888858e-05, "val_loss": 8.053055763244629, "train_action_mse_error": 9.690773963928223}
{"train_loss": 0.04962732270359993, "global_step": 194109, "epoch": 2181, "lr": 1.87918510049648e-05}
{"train_loss": 0.08797312527894974, "global_step": 194110, "epoch": 2181, "lr": 1.879139806223632e-05}
{"train_loss": 0.03963456675410271, "global_step": 194111, "epoch": 2181, "lr": 1.8790945123703436e-05}
{"train_loss": 0.03638968616724014, "global_step": 194112, "epoch": 2181, "lr": 1.879049218936625e-05}
{"train_loss": 0.10324116051197052, "global_step": 194113, "epoch": 2181, "lr": 1.8790039259224785e-05}
{"train_loss": 0.05307690054178238, "global_step": 194114, "epoch": 2181, "lr": 1.878958633327913e-05}
{"train_loss": 0.041804101318120956, "global_step": 194115, "epoch": 2181, "lr": 1.8789133411529324e-05}
{"train_loss": 0.08002714067697525, "global_step": 194116, "epoch": 2181, "lr": 1.8788680493975453e-05}
{"train_loss": 0.05730648338794708, "global_step": 194117, "epoch": 2181, "lr": 1.8788227580617546e-05}
{"train_loss": 0.07808616757392883, "global_step": 194118, "epoch": 2181, "lr": 1.8787774671455698e-05}
{"train_loss": 0.0636925920844078, "global_step": 194119, "epoch": 2181, "lr": 1.878732176648994e-05}
{"train_loss": 0.039748720824718475, "global_step": 194120, "epoch": 2181, "lr": 1.8786868865720358e-05}
{"train_loss": 0.061637070029973984, "global_step": 194121, "epoch": 2181, "lr": 1.8786415969146988e-05}
{"train_loss": 0.0926547646522522, "global_step": 194122, "epoch": 2181, "lr": 1.8785963076769918e-05}
{"train_loss": 0.07639864087104797, "global_step": 194123, "epoch": 2181, "lr": 1.8785510188589178e-05}
{"train_loss": 0.0919281393289566, "global_step": 194124, "epoch": 2181, "lr": 1.878505730460485e-05}
{"train_loss": 0.09430408477783203, "global_step": 194125, "epoch": 2181, "lr": 1.8784604424817005e-05}
{"train_loss": 0.08514661341905594, "global_step": 194126, "epoch": 2181, "lr": 1.878415154922567e-05}
{"train_loss": 0.13315607607364655, "global_step": 194127, "epoch": 2181, "lr": 1.8783698677830946e-05}
{"train_loss": 0.051253411918878555, "global_step": 194128, "epoch": 2181, "lr": 1.8783245810632855e-05}
{"train_loss": 0.03719548508524895, "global_step": 194129, "epoch": 2181, "lr": 1.878279294763149e-05}
{"train_loss": 0.0779217779636383, "global_step": 194130, "epoch": 2181, "lr": 1.8782340088826882e-05}
{"train_loss": 0.10286001116037369, "global_step": 194131, "epoch": 2181, "lr": 1.878188723421913e-05}
{"train_loss": 0.047613903880119324, "global_step": 194132, "epoch": 2181, "lr": 1.878143438380825e-05}
{"train_loss": 0.10334298759698868, "global_step": 194133, "epoch": 2181, "lr": 1.878098153759435e-05}
{"train_loss": 0.04993902146816254, "global_step": 194134, "epoch": 2181, "lr": 1.878052869557746e-05}
{"train_loss": 0.05282929167151451, "global_step": 194135, "epoch": 2181, "lr": 1.878007585775763e-05}
{"train_loss": 0.06217074766755104, "global_step": 194136, "epoch": 2181, "lr": 1.8779623024134955e-05}
{"train_loss": 0.046175453811883926, "global_step": 194137, "epoch": 2181, "lr": 1.8779170194709465e-05}
{"train_loss": 0.05436145141720772, "global_step": 194138, "epoch": 2181, "lr": 1.8778717369481254e-05}
{"train_loss": 0.09610281139612198, "global_step": 194139, "epoch": 2181, "lr": 1.8778264548450348e-05}
{"train_loss": 0.06356637924909592, "global_step": 194140, "epoch": 2181, "lr": 1.8777811731616836e-05}
{"train_loss": 0.06480782479047775, "global_step": 194141, "epoch": 2181, "lr": 1.877735891898075e-05}
{"train_loss": 0.039729781448841095, "global_step": 194142, "epoch": 2181, "lr": 1.8776906110542187e-05}
{"train_loss": 0.10683540254831314, "global_step": 194143, "epoch": 2181, "lr": 1.8776453306301168e-05}
{"train_loss": 0.12555736303329468, "global_step": 194144, "epoch": 2181, "lr": 1.877600050625779e-05}
{"train_loss": 0.020036501809954643, "global_step": 194145, "epoch": 2181, "lr": 1.8775547710412088e-05}
{"train_loss": 0.08286293596029282, "global_step": 194146, "epoch": 2181, "lr": 1.8775094918764145e-05}
{"train_loss": 0.0952427014708519, "global_step": 194147, "epoch": 2181, "lr": 1.8774642131313996e-05}
{"train_loss": 0.057709962129592896, "global_step": 194148, "epoch": 2181, "lr": 1.8774189348061725e-05}
{"train_loss": 0.07570616155862808, "global_step": 194149, "epoch": 2181, "lr": 1.8773736569007377e-05}
{"train_loss": 0.059000786393880844, "global_step": 194150, "epoch": 2181, "lr": 1.8773283794151033e-05}
{"train_loss": 0.06080172583460808, "global_step": 194151, "epoch": 2181, "lr": 1.8772831023492722e-05}
{"train_loss": 0.06522388011217117, "global_step": 194152, "epoch": 2181, "lr": 1.877237825703254e-05}
{"train_loss": 0.059843599796295166, "global_step": 194153, "epoch": 2181, "lr": 1.8771925494770516e-05}
{"train_loss": 0.05880490690469742, "global_step": 194154, "epoch": 2181, "lr": 1.877147273670674e-05}
{"train_loss": 0.053079161792993546, "global_step": 194155, "epoch": 2181, "lr": 1.8771019982841242e-05}
{"train_loss": 0.11994615942239761, "global_step": 194156, "epoch": 2181, "lr": 1.8770567233174126e-05}
{"train_loss": 0.030328061431646347, "global_step": 194157, "epoch": 2181, "lr": 1.87701144877054e-05}
{"train_loss": 0.07086008787155151, "global_step": 194158, "epoch": 2181, "lr": 1.8769661746435168e-05}
{"train_loss": 0.08594439923763275, "global_step": 194159, "epoch": 2181, "lr": 1.876920900936346e-05}
{"train_loss": 0.06227308139204979, "global_step": 194160, "epoch": 2181, "lr": 1.876875627649035e-05}
{"train_loss": 0.04390683397650719, "global_step": 194161, "epoch": 2181, "lr": 1.876830354781592e-05}
{"train_loss": 0.08727052062749863, "global_step": 194162, "epoch": 2181, "lr": 1.8767850823340193e-05}
{"train_loss": 0.09681683778762817, "global_step": 194163, "epoch": 2181, "lr": 1.876739810306326e-05}
{"train_loss": 0.039378415793180466, "global_step": 194164, "epoch": 2181, "lr": 1.8766945386985157e-05}
{"train_loss": 0.03872537240386009, "global_step": 194165, "epoch": 2181, "lr": 1.876649267510598e-05}
{"train_loss": 0.07156135141849518, "global_step": 194166, "epoch": 2181, "lr": 1.876603996742574e-05}
{"train_loss": 0.06212674081325531, "global_step": 194167, "epoch": 2181, "lr": 1.876558726394455e-05}
{"train_loss": 0.0612814724445343, "global_step": 194168, "epoch": 2181, "lr": 1.8765134564662423e-05}
{"train_loss": 0.05119681358337402, "global_step": 194169, "epoch": 2181, "lr": 1.8764681869579465e-05}
{"train_loss": 0.0677294209599495, "global_step": 194170, "epoch": 2181, "lr": 1.8764229178695697e-05}
{"train_loss": 0.1095326617360115, "global_step": 194171, "epoch": 2181, "lr": 1.876377649201121e-05}
{"train_loss": 0.052490416914224625, "global_step": 194172, "epoch": 2181, "lr": 1.876332380952604e-05}
{"train_loss": 0.10303489118814468, "global_step": 194173, "epoch": 2181, "lr": 1.8762871131240278e-05}
{"train_loss": 0.06524470448493958, "global_step": 194174, "epoch": 2181, "lr": 1.8762418457153962e-05}
{"train_loss": 0.06801703572273254, "global_step": 194175, "epoch": 2181, "lr": 1.8761965787267145e-05}
{"train_loss": 0.04894394427537918, "global_step": 194176, "epoch": 2181, "lr": 1.876151312157992e-05}
{"train_loss": 0.06237886846065521, "global_step": 194177, "epoch": 2181, "lr": 1.8761060460092305e-05}
{"train_loss": 0.09019908308982849, "global_step": 194178, "epoch": 2181, "lr": 1.8760607802804407e-05}
{"train_loss": 0.051062971353530884, "global_step": 194179, "epoch": 2181, "lr": 1.8760155149716245e-05}
{"train_loss": 0.09318488836288452, "global_step": 194180, "epoch": 2181, "lr": 1.875970250082792e-05}
{"train_loss": 0.06952372193336487, "global_step": 194181, "epoch": 2181, "lr": 1.8759249856139455e-05}
{"train_loss": 0.1506287306547165, "global_step": 194182, "epoch": 2181, "lr": 1.8758797215650942e-05}
{"train_loss": 0.10659155994653702, "global_step": 194183, "epoch": 2181, "lr": 1.8758344579362415e-05}
{"train_loss": 0.020320851355791092, "global_step": 194184, "epoch": 2181, "lr": 1.8757891947273958e-05}
{"train_loss": 0.06148023530840874, "global_step": 194185, "epoch": 2181, "lr": 1.8757439319385612e-05}
{"train_loss": 0.061967551708221436, "global_step": 194186, "epoch": 2181, "lr": 1.875698669569746e-05}
{"train_loss": 0.05722909793257713, "global_step": 194187, "epoch": 2181, "lr": 1.8756534076209535e-05}
{"train_loss": 0.039796557277441025, "global_step": 194188, "epoch": 2181, "lr": 1.8756081460921933e-05}
{"train_loss": 0.032708894461393356, "global_step": 194189, "epoch": 2181, "lr": 1.8755628849834672e-05}
{"train_loss": 0.09274430572986603, "global_step": 194190, "epoch": 2181, "lr": 1.8755176242947854e-05}
{"train_loss": 0.07143176347017288, "global_step": 194191, "epoch": 2181, "lr": 1.8754723640261507e-05}
{"train_loss": 0.04304135590791702, "global_step": 194192, "epoch": 2181, "lr": 1.8754271041775724e-05}
{"train_loss": 0.06158042699098587, "global_step": 194193, "epoch": 2181, "lr": 1.8753818447490527e-05}
{"train_loss": 0.06861699372529984, "global_step": 194194, "epoch": 2181, "lr": 1.8753365857406018e-05}
{"train_loss": 0.041599664837121964, "global_step": 194195, "epoch": 2181, "lr": 1.8752913271522222e-05}
{"train_loss": 0.1280531883239746, "global_step": 194196, "epoch": 2181, "lr": 1.875246068983922e-05}
{"train_loss": 0.06885792409184943, "global_step": 194197, "epoch": 2181, "lr": 1.875200811235708e-05, "val_loss": 8.041224479675293}
{"train_loss": 0.06629179418087006, "global_step": 194198, "epoch": 2182, "lr": 1.8751555539075832e-05}
{"train_loss": 0.06380970031023026, "global_step": 194199, "epoch": 2182, "lr": 1.875110296999558e-05}
{"train_loss": 0.042905859649181366, "global_step": 194200, "epoch": 2182, "lr": 1.8750650405116343e-05}
{"train_loss": 0.05294516310095787, "global_step": 194201, "epoch": 2182, "lr": 1.8750197844438215e-05}
{"train_loss": 0.10849951207637787, "global_step": 194202, "epoch": 2182, "lr": 1.8749745287961228e-05}
{"train_loss": 0.05104483291506767, "global_step": 194203, "epoch": 2182, "lr": 1.874929273568547e-05}
{"train_loss": 0.12855654954910278, "global_step": 194204, "epoch": 2182, "lr": 1.8748840187610973e-05}
{"train_loss": 0.06102348491549492, "global_step": 194205, "epoch": 2182, "lr": 1.874838764373783e-05}
{"train_loss": 0.0645681619644165, "global_step": 194206, "epoch": 2182, "lr": 1.874793510406607e-05}
{"train_loss": 0.07866841554641724, "global_step": 194207, "epoch": 2182, "lr": 1.8747482568595786e-05}
{"train_loss": 0.04573648422956467, "global_step": 194208, "epoch": 2182, "lr": 1.8747030037327008e-05}
{"train_loss": 0.07570039480924606, "global_step": 194209, "epoch": 2182, "lr": 1.874657751025982e-05}
{"train_loss": 0.08688659965991974, "global_step": 194210, "epoch": 2182, "lr": 1.8746124987394264e-05}
{"train_loss": 0.05659351497888565, "global_step": 194211, "epoch": 2182, "lr": 1.8745672468730424e-05}
{"train_loss": 0.061799272894859314, "global_step": 194212, "epoch": 2182, "lr": 1.8745219954268352e-05}
{"train_loss": 0.046851541846990585, "global_step": 194213, "epoch": 2182, "lr": 1.874476744400808e-05}
{"train_loss": 0.06740229576826096, "global_step": 194214, "epoch": 2182, "lr": 1.8744314937949714e-05}
{"train_loss": 0.09799228608608246, "global_step": 194215, "epoch": 2182, "lr": 1.8743862436093278e-05}
{"train_loss": 0.08687296509742737, "global_step": 194216, "epoch": 2182, "lr": 1.8743409938438867e-05}
{"train_loss": 0.03052336908876896, "global_step": 194217, "epoch": 2182, "lr": 1.8742957444986505e-05}
{"train_loss": 0.06392831355333328, "global_step": 194218, "epoch": 2182, "lr": 1.8742504955736283e-05}
{"train_loss": 0.10533076524734497, "global_step": 194219, "epoch": 2182, "lr": 1.8742052470688237e-05}
{"train_loss": 0.053430430591106415, "global_step": 194220, "epoch": 2182, "lr": 1.8741599989842462e-05}
{"train_loss": 0.08490721881389618, "global_step": 194221, "epoch": 2182, "lr": 1.8741147513198975e-05}
{"train_loss": 0.08932934701442719, "global_step": 194222, "epoch": 2182, "lr": 1.874069504075788e-05}
{"train_loss": 0.09058725833892822, "global_step": 194223, "epoch": 2182, "lr": 1.87402425725192e-05}
{"train_loss": 0.06559278070926666, "global_step": 194224, "epoch": 2182, "lr": 1.873979010848303e-05}
{"train_loss": 0.07815052568912506, "global_step": 194225, "epoch": 2182, "lr": 1.8739337648649398e-05}
{"train_loss": 0.05841898173093796, "global_step": 194226, "epoch": 2182, "lr": 1.8738885193018395e-05}
{"train_loss": 0.030772794038057327, "global_step": 194227, "epoch": 2182, "lr": 1.8738432741590055e-05}
{"train_loss": 0.05227874591946602, "global_step": 194228, "epoch": 2182, "lr": 1.8737980294364465e-05}
{"train_loss": 0.07621564716100693, "global_step": 194229, "epoch": 2182, "lr": 1.8737527851341664e-05}
{"train_loss": 0.047426339238882065, "global_step": 194230, "epoch": 2182, "lr": 1.8737075412521727e-05}
{"train_loss": 0.07387647777795792, "global_step": 194231, "epoch": 2182, "lr": 1.87366229779047e-05}
{"train_loss": 0.08959613740444183, "global_step": 194232, "epoch": 2182, "lr": 1.8736170547490652e-05}
{"train_loss": 0.0714486688375473, "global_step": 194233, "epoch": 2182, "lr": 1.8735718121279655e-05}
{"train_loss": 0.04634321853518486, "global_step": 194234, "epoch": 2182, "lr": 1.8735265699271747e-05}
{"train_loss": 0.05778604373335838, "global_step": 194235, "epoch": 2182, "lr": 1.8734813281467023e-05}
{"train_loss": 0.030429353937506676, "global_step": 194236, "epoch": 2182, "lr": 1.8734360867865498e-05}
{"train_loss": 0.05139315500855446, "global_step": 194237, "epoch": 2182, "lr": 1.873390845846728e-05}
{"train_loss": 0.10550940781831741, "global_step": 194238, "epoch": 2182, "lr": 1.8733456053272385e-05}
{"train_loss": 0.07028992474079132, "global_step": 194239, "epoch": 2182, "lr": 1.8733003652280916e-05}
{"train_loss": 0.055475495755672455, "global_step": 194240, "epoch": 2182, "lr": 1.8732551255492896e-05}
{"train_loss": 0.10632947087287903, "global_step": 194241, "epoch": 2182, "lr": 1.873209886290842e-05}
{"train_loss": 0.054469525814056396, "global_step": 194242, "epoch": 2182, "lr": 1.8731646474527513e-05}
{"train_loss": 0.02619040384888649, "global_step": 194243, "epoch": 2182, "lr": 1.8731194090350275e-05}
{"train_loss": 0.07060135900974274, "global_step": 194244, "epoch": 2182, "lr": 1.873074171037673e-05}
{"train_loss": 0.05948169529438019, "global_step": 194245, "epoch": 2182, "lr": 1.873028933460697e-05}
{"train_loss": 0.04796437546610832, "global_step": 194246, "epoch": 2182, "lr": 1.8729836963041025e-05}
{"train_loss": 0.07340798527002335, "global_step": 194247, "epoch": 2182, "lr": 1.8729384595678993e-05}
{"train_loss": 0.07737241685390472, "global_step": 194248, "epoch": 2182, "lr": 1.8728932232520895e-05}
{"train_loss": 0.06200317665934563, "global_step": 194249, "epoch": 2182, "lr": 1.8728479873566828e-05}
{"train_loss": 0.027998486533761024, "global_step": 194250, "epoch": 2182, "lr": 1.8728027518816816e-05}
{"train_loss": 0.09388089925050735, "global_step": 194251, "epoch": 2182, "lr": 1.872757516827096e-05}
{"train_loss": 0.05736248195171356, "global_step": 194252, "epoch": 2182, "lr": 1.87271228219293e-05}
{"train_loss": 0.04349274933338165, "global_step": 194253, "epoch": 2182, "lr": 1.8726670479791882e-05}
{"train_loss": 0.07836799323558807, "global_step": 194254, "epoch": 2182, "lr": 1.872621814185879e-05}
{"train_loss": 0.09813609719276428, "global_step": 194255, "epoch": 2182, "lr": 1.872576580813007e-05}
{"train_loss": 0.07588239014148712, "global_step": 194256, "epoch": 2182, "lr": 1.8725313478605807e-05}
{"train_loss": 0.05756800249218941, "global_step": 194257, "epoch": 2182, "lr": 1.8724861153286015e-05}
{"train_loss": 0.04444470256567001, "global_step": 194258, "epoch": 2182, "lr": 1.8724408832170814e-05}
{"train_loss": 0.0582699328660965, "global_step": 194259, "epoch": 2182, "lr": 1.872395651526021e-05}
{"train_loss": 0.06825458258390427, "global_step": 194260, "epoch": 2182, "lr": 1.8723504202554308e-05}
{"train_loss": 0.06958874315023422, "global_step": 194261, "epoch": 2182, "lr": 1.872305189405314e-05}
{"train_loss": 0.09612992405891418, "global_step": 194262, "epoch": 2182, "lr": 1.8722599589756783e-05}
{"train_loss": 0.053987644612789154, "global_step": 194263, "epoch": 2182, "lr": 1.872214728966528e-05}
{"train_loss": 0.056141261011362076, "global_step": 194264, "epoch": 2182, "lr": 1.872169499377872e-05}
{"train_loss": 0.07081450521945953, "global_step": 194265, "epoch": 2182, "lr": 1.872124270209713e-05}
{"train_loss": 0.09653395414352417, "global_step": 194266, "epoch": 2182, "lr": 1.8720790414620603e-05}
{"train_loss": 0.06373459100723267, "global_step": 194267, "epoch": 2182, "lr": 1.8720338131349164e-05}
{"train_loss": 0.08368635177612305, "global_step": 194268, "epoch": 2182, "lr": 1.8719885852282897e-05}
{"train_loss": 0.05774272605776787, "global_step": 194269, "epoch": 2182, "lr": 1.871943357742188e-05}
{"train_loss": 0.07326239347457886, "global_step": 194270, "epoch": 2182, "lr": 1.8718981306766137e-05}
{"train_loss": 0.03592090308666229, "global_step": 194271, "epoch": 2182, "lr": 1.8718529040315756e-05}
{"train_loss": 0.1240767240524292, "global_step": 194272, "epoch": 2182, "lr": 1.871807677807078e-05}
{"train_loss": 0.10748632997274399, "global_step": 194273, "epoch": 2182, "lr": 1.871762452003129e-05}
{"train_loss": 0.09781207144260406, "global_step": 194274, "epoch": 2182, "lr": 1.8717172266197313e-05}
{"train_loss": 0.101723812520504, "global_step": 194275, "epoch": 2182, "lr": 1.8716720016568955e-05}
{"train_loss": 0.04838530346751213, "global_step": 194276, "epoch": 2182, "lr": 1.8716267771146227e-05}
{"train_loss": 0.03934451565146446, "global_step": 194277, "epoch": 2182, "lr": 1.8715815529929242e-05}
{"train_loss": 0.08996647596359253, "global_step": 194278, "epoch": 2182, "lr": 1.8715363292918016e-05}
{"train_loss": 0.11548751592636108, "global_step": 194279, "epoch": 2182, "lr": 1.871491106011264e-05}
{"train_loss": 0.07470086961984634, "global_step": 194280, "epoch": 2182, "lr": 1.871445883151315e-05}
{"train_loss": 0.030475666746497154, "global_step": 194281, "epoch": 2182, "lr": 1.8714006607119634e-05}
{"train_loss": 0.06906738877296448, "global_step": 194282, "epoch": 2182, "lr": 1.871355438693212e-05}
{"train_loss": 0.04601975902915001, "global_step": 194283, "epoch": 2182, "lr": 1.8713102170950708e-05}
{"train_loss": 0.06788855046033859, "global_step": 194284, "epoch": 2182, "lr": 1.8712649959175415e-05}
{"train_loss": 0.0708467960357666, "global_step": 194285, "epoch": 2182, "lr": 1.8712197751606346e-05}
{"train_loss": 0.06851210796766066, "global_step": 194286, "epoch": 2182, "lr": 1.8711745548243524e-05, "val_loss": 8.040499687194824}
{"train_loss": 0.041786786168813705, "global_step": 194287, "epoch": 2183, "lr": 1.871129334908705e-05}
{"train_loss": 0.08333148807287216, "global_step": 194288, "epoch": 2183, "lr": 1.8710841154136936e-05}
{"train_loss": 0.07490233331918716, "global_step": 194289, "epoch": 2183, "lr": 1.871038896339329e-05}
{"train_loss": 0.06318788230419159, "global_step": 194290, "epoch": 2183, "lr": 1.8709936776856146e-05}
{"train_loss": 0.03800734505057335, "global_step": 194291, "epoch": 2183, "lr": 1.8709484594525552e-05}
{"train_loss": 0.09479890763759613, "global_step": 194292, "epoch": 2183, "lr": 1.8709032416401608e-05}
{"train_loss": 0.07627806812524796, "global_step": 194293, "epoch": 2183, "lr": 1.870858024248433e-05}
{"train_loss": 0.05653934180736542, "global_step": 194294, "epoch": 2183, "lr": 1.870812807277383e-05}
{"train_loss": 0.04860179126262665, "global_step": 194295, "epoch": 2183, "lr": 1.870767590727011e-05}
{"train_loss": 0.07655281573534012, "global_step": 194296, "epoch": 2183, "lr": 1.870722374597329e-05}
{"train_loss": 0.07663404941558838, "global_step": 194297, "epoch": 2183, "lr": 1.8706771588883376e-05}
{"train_loss": 0.07861285656690598, "global_step": 194298, "epoch": 2183, "lr": 1.870631943600048e-05}
{"train_loss": 0.06256422400474548, "global_step": 194299, "epoch": 2183, "lr": 1.8705867287324617e-05}
{"train_loss": 0.06143859401345253, "global_step": 194300, "epoch": 2183, "lr": 1.8705415142855888e-05}
{"train_loss": 0.08984754979610443, "global_step": 194301, "epoch": 2183, "lr": 1.8704963002594316e-05}
{"train_loss": 0.04493327438831329, "global_step": 194302, "epoch": 2183, "lr": 1.870451086654e-05}
{"train_loss": 0.07640216499567032, "global_step": 194303, "epoch": 2183, "lr": 1.8704058734692957e-05}
{"train_loss": 0.0812491849064827, "global_step": 194304, "epoch": 2183, "lr": 1.8703606607053292e-05}
{"train_loss": 0.05782713741064072, "global_step": 194305, "epoch": 2183, "lr": 1.8703154483621032e-05}
{"train_loss": 0.12416151911020279, "global_step": 194306, "epoch": 2183, "lr": 1.8702702364396246e-05}
{"train_loss": 0.07326897233724594, "global_step": 194307, "epoch": 2183, "lr": 1.870225024937902e-05}
{"train_loss": 0.12151503562927246, "global_step": 194308, "epoch": 2183, "lr": 1.870179813856938e-05}
{"train_loss": 0.06855672597885132, "global_step": 194309, "epoch": 2183, "lr": 1.870134603196742e-05}
{"train_loss": 0.05061551183462143, "global_step": 194310, "epoch": 2183, "lr": 1.870089392957316e-05}
{"train_loss": 0.08889748901128769, "global_step": 194311, "epoch": 2183, "lr": 1.8700441831386707e-05}
{"train_loss": 0.07508986443281174, "global_step": 194312, "epoch": 2183, "lr": 1.8699989737408073e-05}
{"train_loss": 0.07785646617412567, "global_step": 194313, "epoch": 2183, "lr": 1.8699537647637367e-05}
{"train_loss": 0.11035268008708954, "global_step": 194314, "epoch": 2183, "lr": 1.869908556207461e-05}
{"train_loss": 0.07818011194467545, "global_step": 194315, "epoch": 2183, "lr": 1.869863348071989e-05}
{"train_loss": 0.0787172019481659, "global_step": 194316, "epoch": 2183, "lr": 1.8698181403573243e-05}
{"train_loss": 0.040555547922849655, "global_step": 194317, "epoch": 2183, "lr": 1.8697729330634766e-05}
{"train_loss": 0.05389048531651497, "global_step": 194318, "epoch": 2183, "lr": 1.8697277261904476e-05}
{"train_loss": 0.04116310551762581, "global_step": 194319, "epoch": 2183, "lr": 1.8696825197382477e-05}
{"train_loss": 0.08044911921024323, "global_step": 194320, "epoch": 2183, "lr": 1.8696373137068785e-05}
{"train_loss": 0.050074443221092224, "global_step": 194321, "epoch": 2183, "lr": 1.8695921080963508e-05}
{"train_loss": 0.04997844621539116, "global_step": 194322, "epoch": 2183, "lr": 1.869546902906666e-05}
{"train_loss": 0.053338535130023956, "global_step": 194323, "epoch": 2183, "lr": 1.8695016981378343e-05}
{"train_loss": 0.050257086753845215, "global_step": 194324, "epoch": 2183, "lr": 1.8694564937898583e-05}
{"train_loss": 0.039314914494752884, "global_step": 194325, "epoch": 2183, "lr": 1.8694112898627476e-05}
{"train_loss": 0.04372231289744377, "global_step": 194326, "epoch": 2183, "lr": 1.8693660863565042e-05}
{"train_loss": 0.048615291714668274, "global_step": 194327, "epoch": 2183, "lr": 1.8693208832711383e-05}
{"train_loss": 0.06686540693044662, "global_step": 194328, "epoch": 2183, "lr": 1.8692756806066546e-05}
{"train_loss": 0.07551173865795135, "global_step": 194329, "epoch": 2183, "lr": 1.8692304783630564e-05}
{"train_loss": 0.09672664850950241, "global_step": 194330, "epoch": 2183, "lr": 1.869185276540354e-05}
{"train_loss": 0.07242545485496521, "global_step": 194331, "epoch": 2183, "lr": 1.869140075138549e-05}
{"train_loss": 0.0674562081694603, "global_step": 194332, "epoch": 2183, "lr": 1.8690948741576524e-05}
{"train_loss": 0.057248324155807495, "global_step": 194333, "epoch": 2183, "lr": 1.8690496735976658e-05}
{"train_loss": 0.059743382036685944, "global_step": 194334, "epoch": 2183, "lr": 1.8690044734585992e-05}
{"train_loss": 0.056432001292705536, "global_step": 194335, "epoch": 2183, "lr": 1.8689592737404556e-05}
{"train_loss": 0.09522529691457748, "global_step": 194336, "epoch": 2183, "lr": 1.8689140744432432e-05}
{"train_loss": 0.05184021592140198, "global_step": 194337, "epoch": 2183, "lr": 1.8688688755669653e-05}
{"train_loss": 0.1088816374540329, "global_step": 194338, "epoch": 2183, "lr": 1.8688236771116325e-05}
{"train_loss": 0.0907924473285675, "global_step": 194339, "epoch": 2183, "lr": 1.8687784790772457e-05}
{"train_loss": 0.14035595953464508, "global_step": 194340, "epoch": 2183, "lr": 1.8687332814638153e-05}
{"train_loss": 0.09296028316020966, "global_step": 194341, "epoch": 2183, "lr": 1.868688084271344e-05}
{"train_loss": 0.0723135694861412, "global_step": 194342, "epoch": 2183, "lr": 1.868642887499839e-05}
{"train_loss": 0.06504521518945694, "global_step": 194343, "epoch": 2183, "lr": 1.8685976911493088e-05}
{"train_loss": 0.04717682674527168, "global_step": 194344, "epoch": 2183, "lr": 1.8685524952197558e-05}
{"train_loss": 0.06173034757375717, "global_step": 194345, "epoch": 2183, "lr": 1.8685072997111897e-05}
{"train_loss": 0.07434097677469254, "global_step": 194346, "epoch": 2183, "lr": 1.868462104623612e-05}
{"train_loss": 0.0670909658074379, "global_step": 194347, "epoch": 2183, "lr": 1.868416909957034e-05}
{"train_loss": 0.039759278297424316, "global_step": 194348, "epoch": 2183, "lr": 1.868371715711457e-05}
{"train_loss": 0.05768710747361183, "global_step": 194349, "epoch": 2183, "lr": 1.8683265218868918e-05}
{"train_loss": 0.10915663093328476, "global_step": 194350, "epoch": 2183, "lr": 1.8682813284833394e-05}
{"train_loss": 0.061355311423540115, "global_step": 194351, "epoch": 2183, "lr": 1.8682361355008106e-05}
{"train_loss": 0.043955232948064804, "global_step": 194352, "epoch": 2183, "lr": 1.868190942939307e-05}
{"train_loss": 0.09218249469995499, "global_step": 194353, "epoch": 2183, "lr": 1.8681457507988386e-05}
{"train_loss": 0.02213091216981411, "global_step": 194354, "epoch": 2183, "lr": 1.8681005590794086e-05}
{"train_loss": 0.06852253526449203, "global_step": 194355, "epoch": 2183, "lr": 1.868055367781026e-05}
{"train_loss": 0.03369361534714699, "global_step": 194356, "epoch": 2183, "lr": 1.8680101769036934e-05}
{"train_loss": 0.06659014523029327, "global_step": 194357, "epoch": 2183, "lr": 1.8679649864474203e-05}
{"train_loss": 0.09006410092115402, "global_step": 194358, "epoch": 2183, "lr": 1.8679197964122097e-05}
{"train_loss": 0.09916847199201584, "global_step": 194359, "epoch": 2183, "lr": 1.8678746067980686e-05}
{"train_loss": 0.06587597727775574, "global_step": 194360, "epoch": 2183, "lr": 1.8678294176050052e-05}
{"train_loss": 0.04113506153225899, "global_step": 194361, "epoch": 2183, "lr": 1.867784228833025e-05}
{"train_loss": 0.06839248538017273, "global_step": 194362, "epoch": 2183, "lr": 1.8677390404821306e-05}
{"train_loss": 0.14139077067375183, "global_step": 194363, "epoch": 2183, "lr": 1.8676938525523318e-05}
{"train_loss": 0.062213245779275894, "global_step": 194364, "epoch": 2183, "lr": 1.8676486650436324e-05}
{"train_loss": 0.08744590729475021, "global_step": 194365, "epoch": 2183, "lr": 1.867603477956041e-05}
{"train_loss": 0.12146487087011337, "global_step": 194366, "epoch": 2183, "lr": 1.8675582912895602e-05}
{"train_loss": 0.03928934037685394, "global_step": 194367, "epoch": 2183, "lr": 1.8675131050441996e-05}
{"train_loss": 0.06954962760210037, "global_step": 194368, "epoch": 2183, "lr": 1.867467919219964e-05}
{"train_loss": 0.10097309947013855, "global_step": 194369, "epoch": 2183, "lr": 1.8674227338168575e-05}
{"train_loss": 0.08733533322811127, "global_step": 194370, "epoch": 2183, "lr": 1.8673775488348893e-05}
{"train_loss": 0.06006130576133728, "global_step": 194371, "epoch": 2183, "lr": 1.8673323642740624e-05}
{"train_loss": 0.06821925938129425, "global_step": 194372, "epoch": 2183, "lr": 1.8672871801343867e-05}
{"train_loss": 0.03684317320585251, "global_step": 194373, "epoch": 2183, "lr": 1.8672419964158637e-05}
{"train_loss": 0.06597752124071121, "global_step": 194374, "epoch": 2183, "lr": 1.8671968131185037e-05}
{"train_loss": 0.07090560194063053, "global_step": 194375, "epoch": 2183, "lr": 1.867151630242309e-05, "val_loss": 8.00440788269043}
{"train_loss": 0.09063003212213516, "global_step": 194376, "epoch": 2184, "lr": 1.8671064477872895e-05}
{"train_loss": 0.027492769062519073, "global_step": 194377, "epoch": 2184, "lr": 1.8670612657534475e-05}
{"train_loss": 0.02990335039794445, "global_step": 194378, "epoch": 2184, "lr": 1.8670160841407913e-05}
{"train_loss": 0.03303573653101921, "global_step": 194379, "epoch": 2184, "lr": 1.866970902949328e-05}
{"train_loss": 0.07059761136770248, "global_step": 194380, "epoch": 2184, "lr": 1.8669257221790603e-05}
{"train_loss": 0.0425066240131855, "global_step": 194381, "epoch": 2184, "lr": 1.8668805418299984e-05}
{"train_loss": 0.03161589428782463, "global_step": 194382, "epoch": 2184, "lr": 1.8668353619021446e-05}
{"train_loss": 0.03499840945005417, "global_step": 194383, "epoch": 2184, "lr": 1.866790182395508e-05}
{"train_loss": 0.05442322790622711, "global_step": 194384, "epoch": 2184, "lr": 1.866745003310092e-05}
{"train_loss": 0.12578292191028595, "global_step": 194385, "epoch": 2184, "lr": 1.866699824645905e-05}
{"train_loss": 0.06483690440654755, "global_step": 194386, "epoch": 2184, "lr": 1.8666546464029505e-05}
{"train_loss": 0.05617111548781395, "global_step": 194387, "epoch": 2184, "lr": 1.866609468581238e-05}
{"train_loss": 0.057373326271772385, "global_step": 194388, "epoch": 2184, "lr": 1.8665642911807703e-05}
{"train_loss": 0.08141575753688812, "global_step": 194389, "epoch": 2184, "lr": 1.8665191142015557e-05}
{"train_loss": 0.10988670587539673, "global_step": 194390, "epoch": 2184, "lr": 1.8664739376435985e-05}
{"train_loss": 0.029555754736065865, "global_step": 194391, "epoch": 2184, "lr": 1.866428761506907e-05}
{"train_loss": 0.05953037738800049, "global_step": 194392, "epoch": 2184, "lr": 1.8663835857914842e-05}
{"train_loss": 0.08322296291589737, "global_step": 194393, "epoch": 2184, "lr": 1.8663384104973397e-05}
{"train_loss": 0.10977716743946075, "global_step": 194394, "epoch": 2184, "lr": 1.8662932356244763e-05}
{"train_loss": 0.06301144510507584, "global_step": 194395, "epoch": 2184, "lr": 1.8662480611729012e-05}
{"train_loss": 0.08202027529478073, "global_step": 194396, "epoch": 2184, "lr": 1.866202887142623e-05}
{"train_loss": 0.08274757117033005, "global_step": 194397, "epoch": 2184, "lr": 1.866157713533644e-05}
{"train_loss": 0.11011136323213577, "global_step": 194398, "epoch": 2184, "lr": 1.8661125403459728e-05}
{"train_loss": 0.04993695020675659, "global_step": 194399, "epoch": 2184, "lr": 1.8660673675796135e-05}
{"train_loss": 0.07812724262475967, "global_step": 194400, "epoch": 2184, "lr": 1.866022195234575e-05}
{"train_loss": 0.04766799509525299, "global_step": 194401, "epoch": 2184, "lr": 1.8659770233108615e-05}
{"train_loss": 0.04177005961537361, "global_step": 194402, "epoch": 2184, "lr": 1.8659318518084773e-05}
{"train_loss": 0.06782478094100952, "global_step": 194403, "epoch": 2184, "lr": 1.865886680727432e-05}
{"train_loss": 0.0486195832490921, "global_step": 194404, "epoch": 2184, "lr": 1.8658415100677284e-05}
{"train_loss": 0.11361616104841232, "global_step": 194405, "epoch": 2184, "lr": 1.8657963398293766e-05}
{"train_loss": 0.048657532781362534, "global_step": 194406, "epoch": 2184, "lr": 1.8657511700123792e-05}
{"train_loss": 0.0765732079744339, "global_step": 194407, "epoch": 2184, "lr": 1.8657060006167425e-05}
{"train_loss": 0.05164484307169914, "global_step": 194408, "epoch": 2184, "lr": 1.8656608316424745e-05}
{"train_loss": 0.04706273972988129, "global_step": 194409, "epoch": 2184, "lr": 1.865615663089579e-05}
{"train_loss": 0.08835238218307495, "global_step": 194410, "epoch": 2184, "lr": 1.865570494958065e-05}
{"train_loss": 0.08028166741132736, "global_step": 194411, "epoch": 2184, "lr": 1.8655253272479344e-05}
{"train_loss": 0.05718269571661949, "global_step": 194412, "epoch": 2184, "lr": 1.865480159959198e-05}
{"train_loss": 0.03849903494119644, "global_step": 194413, "epoch": 2184, "lr": 1.865434993091858e-05}
{"train_loss": 0.06303352117538452, "global_step": 194414, "epoch": 2184, "lr": 1.865389826645922e-05}
{"train_loss": 0.026770051568746567, "global_step": 194415, "epoch": 2184, "lr": 1.8653446606213975e-05}
{"train_loss": 0.061299797147512436, "global_step": 194416, "epoch": 2184, "lr": 1.8652994950182877e-05}
{"train_loss": 0.07508689910173416, "global_step": 194417, "epoch": 2184, "lr": 1.865254329836602e-05}
{"train_loss": 0.07593929767608643, "global_step": 194418, "epoch": 2184, "lr": 1.8652091650763425e-05}
{"train_loss": 0.08322563022375107, "global_step": 194419, "epoch": 2184, "lr": 1.86516400073752e-05}
{"train_loss": 0.04800587147474289, "global_step": 194420, "epoch": 2184, "lr": 1.8651188368201355e-05}
{"train_loss": 0.06885932385921478, "global_step": 194421, "epoch": 2184, "lr": 1.8650736733242e-05}
{"train_loss": 0.0760420709848404, "global_step": 194422, "epoch": 2184, "lr": 1.8650285102497145e-05}
{"train_loss": 0.05463763326406479, "global_step": 194423, "epoch": 2184, "lr": 1.8649833475966898e-05}
{"train_loss": 0.0776861384510994, "global_step": 194424, "epoch": 2184, "lr": 1.8649381853651283e-05}
{"train_loss": 0.06715608388185501, "global_step": 194425, "epoch": 2184, "lr": 1.8648930235550395e-05}
{"train_loss": 0.05987837538123131, "global_step": 194426, "epoch": 2184, "lr": 1.8648478621664256e-05}
{"train_loss": 0.07405257225036621, "global_step": 194427, "epoch": 2184, "lr": 1.8648027011992962e-05}
{"train_loss": 0.0479288250207901, "global_step": 194428, "epoch": 2184, "lr": 1.8647575406536545e-05}
{"train_loss": 0.035988811403512955, "global_step": 194429, "epoch": 2184, "lr": 1.8647123805295097e-05}
{"train_loss": 0.1063821092247963, "global_step": 194430, "epoch": 2184, "lr": 1.8646672208268646e-05}
{"train_loss": 0.07781747728586197, "global_step": 194431, "epoch": 2184, "lr": 1.8646220615457276e-05}
{"train_loss": 0.08519415557384491, "global_step": 194432, "epoch": 2184, "lr": 1.8645769026861033e-05}
{"train_loss": 0.07561209052801132, "global_step": 194433, "epoch": 2184, "lr": 1.8645317442479975e-05}
{"train_loss": 0.08403417468070984, "global_step": 194434, "epoch": 2184, "lr": 1.8644865862314197e-05}
{"train_loss": 0.03212781995534897, "global_step": 194435, "epoch": 2184, "lr": 1.8644414286363715e-05}
{"train_loss": 0.07548520714044571, "global_step": 194436, "epoch": 2184, "lr": 1.8643962714628622e-05}
{"train_loss": 0.020527366548776627, "global_step": 194437, "epoch": 2184, "lr": 1.8643511147108954e-05}
{"train_loss": 0.10528416931629181, "global_step": 194438, "epoch": 2184, "lr": 1.86430595838048e-05}
{"train_loss": 0.11105140298604965, "global_step": 194439, "epoch": 2184, "lr": 1.8642608024716203e-05}
{"train_loss": 0.0679403692483902, "global_step": 194440, "epoch": 2184, "lr": 1.8642156469843208e-05}
{"train_loss": 0.0821586400270462, "global_step": 194441, "epoch": 2184, "lr": 1.864170491918591e-05}
{"train_loss": 0.06626076996326447, "global_step": 194442, "epoch": 2184, "lr": 1.8641253372744338e-05}
{"train_loss": 0.04700319096446037, "global_step": 194443, "epoch": 2184, "lr": 1.8640801830518585e-05}
{"train_loss": 0.06871135532855988, "global_step": 194444, "epoch": 2184, "lr": 1.864035029250869e-05}
{"train_loss": 0.0973302498459816, "global_step": 194445, "epoch": 2184, "lr": 1.86398987587147e-05}
{"train_loss": 0.07032649964094162, "global_step": 194446, "epoch": 2184, "lr": 1.8639447229136715e-05}
{"train_loss": 0.06668557971715927, "global_step": 194447, "epoch": 2184, "lr": 1.8638995703774755e-05}
{"train_loss": 0.05710539594292641, "global_step": 194448, "epoch": 2184, "lr": 1.8638544182628914e-05}
{"train_loss": 0.06623540073633194, "global_step": 194449, "epoch": 2184, "lr": 1.8638092665699224e-05}
{"train_loss": 0.05584239587187767, "global_step": 194450, "epoch": 2184, "lr": 1.8637641152985756e-05}
{"train_loss": 0.10300946235656738, "global_step": 194451, "epoch": 2184, "lr": 1.86371896444886e-05}
{"train_loss": 0.08107125759124756, "global_step": 194452, "epoch": 2184, "lr": 1.8636738140207773e-05}
{"train_loss": 0.08157071471214294, "global_step": 194453, "epoch": 2184, "lr": 1.8636286640143362e-05}
{"train_loss": 0.06336149573326111, "global_step": 194454, "epoch": 2184, "lr": 1.8635835144295415e-05}
{"train_loss": 0.04519766941666603, "global_step": 194455, "epoch": 2184, "lr": 1.8635383652664007e-05}
{"train_loss": 0.11160796135663986, "global_step": 194456, "epoch": 2184, "lr": 1.8634932165249177e-05}
{"train_loss": 0.04991301894187927, "global_step": 194457, "epoch": 2184, "lr": 1.8634480682051015e-05}
{"train_loss": 0.04332079738378525, "global_step": 194458, "epoch": 2184, "lr": 1.8634029203069543e-05}
{"train_loss": 0.03334392234683037, "global_step": 194459, "epoch": 2184, "lr": 1.8633577728304862e-05}
{"train_loss": 0.04205678030848503, "global_step": 194460, "epoch": 2184, "lr": 1.8633126257757e-05}
{"train_loss": 0.056923504918813705, "global_step": 194461, "epoch": 2184, "lr": 1.863267479142605e-05}
{"train_loss": 0.06744962185621262, "global_step": 194462, "epoch": 2184, "lr": 1.863222332931203e-05}
{"train_loss": 0.05505474656820297, "global_step": 194463, "epoch": 2184, "lr": 1.863177187141505e-05}
{"train_loss": 0.06589725768465675, "global_step": 194464, "epoch": 2184, "lr": 1.8631320417735125e-05, "val_loss": 7.985016345977783}
{"train_loss": 0.06306319683790207, "global_step": 194465, "epoch": 2185, "lr": 1.863086896827235e-05}
{"train_loss": 0.04159555956721306, "global_step": 194466, "epoch": 2185, "lr": 1.8630417523026765e-05}
{"train_loss": 0.07785005867481232, "global_step": 194467, "epoch": 2185, "lr": 1.862996608199845e-05}
{"train_loss": 0.06712174415588379, "global_step": 194468, "epoch": 2185, "lr": 1.8629514645187435e-05}
{"train_loss": 0.03839178383350372, "global_step": 194469, "epoch": 2185, "lr": 1.86290632125938e-05}
{"train_loss": 0.0529579259455204, "global_step": 194470, "epoch": 2185, "lr": 1.8628611784217624e-05}
{"train_loss": 0.07057231664657593, "global_step": 194471, "epoch": 2185, "lr": 1.8628160360058933e-05}
{"train_loss": 0.06875278055667877, "global_step": 194472, "epoch": 2185, "lr": 1.8627708940117816e-05}
{"train_loss": 0.041748613119125366, "global_step": 194473, "epoch": 2185, "lr": 1.862725752439431e-05}
{"train_loss": 0.061188600957393646, "global_step": 194474, "epoch": 2185, "lr": 1.86268061128885e-05}
{"train_loss": 0.07330675423145294, "global_step": 194475, "epoch": 2185, "lr": 1.862635470560042e-05}
{"train_loss": 0.05428950861096382, "global_step": 194476, "epoch": 2185, "lr": 1.8625903302530156e-05}
{"train_loss": 0.03729523345828056, "global_step": 194477, "epoch": 2185, "lr": 1.862545190367776e-05}
{"train_loss": 0.04715581238269806, "global_step": 194478, "epoch": 2185, "lr": 1.8625000509043273e-05}
{"train_loss": 0.04198490083217621, "global_step": 194479, "epoch": 2185, "lr": 1.8624549118626783e-05}
{"train_loss": 0.06949059665203094, "global_step": 194480, "epoch": 2185, "lr": 1.862409773242833e-05}
{"train_loss": 0.07513783872127533, "global_step": 194481, "epoch": 2185, "lr": 1.8623646350447998e-05}
{"train_loss": 0.04009786993265152, "global_step": 194482, "epoch": 2185, "lr": 1.862319497268582e-05}
{"train_loss": 0.05892449989914894, "global_step": 194483, "epoch": 2185, "lr": 1.862274359914189e-05}
{"train_loss": 0.02917427010834217, "global_step": 194484, "epoch": 2185, "lr": 1.8622292229816245e-05}
{"train_loss": 0.09156601130962372, "global_step": 194485, "epoch": 2185, "lr": 1.8621840864708934e-05}
{"train_loss": 0.022799143567681313, "global_step": 194486, "epoch": 2185, "lr": 1.8621389503820036e-05}
{"train_loss": 0.06267059594392776, "global_step": 194487, "epoch": 2185, "lr": 1.8620938147149624e-05}
{"train_loss": 0.10723447799682617, "global_step": 194488, "epoch": 2185, "lr": 1.8620486794697726e-05}
{"train_loss": 0.053819090127944946, "global_step": 194489, "epoch": 2185, "lr": 1.8620035446464446e-05}
{"train_loss": 0.09691425412893295, "global_step": 194490, "epoch": 2185, "lr": 1.86195841024498e-05}
{"train_loss": 0.09831448644399643, "global_step": 194491, "epoch": 2185, "lr": 1.861913276265388e-05}
{"train_loss": 0.08089820295572281, "global_step": 194492, "epoch": 2185, "lr": 1.8618681427076722e-05}
{"train_loss": 0.09108423441648483, "global_step": 194493, "epoch": 2185, "lr": 1.8618230095718416e-05}
{"train_loss": 0.04633156955242157, "global_step": 194494, "epoch": 2185, "lr": 1.861777876857899e-05}
{"train_loss": 0.07487107068300247, "global_step": 194495, "epoch": 2185, "lr": 1.8617327445658538e-05}
{"train_loss": 0.10808979719877243, "global_step": 194496, "epoch": 2185, "lr": 1.861687612695709e-05}
{"train_loss": 0.041261933743953705, "global_step": 194497, "epoch": 2185, "lr": 1.861642481247473e-05}
{"train_loss": 0.061642907559871674, "global_step": 194498, "epoch": 2185, "lr": 1.8615973502211498e-05}
{"train_loss": 0.047155678272247314, "global_step": 194499, "epoch": 2185, "lr": 1.8615522196167483e-05}
{"train_loss": 0.06822490692138672, "global_step": 194500, "epoch": 2185, "lr": 1.8615070894342707e-05}
{"train_loss": 0.0636613667011261, "global_step": 194501, "epoch": 2185, "lr": 1.8614619596737274e-05}
{"train_loss": 0.06130347028374672, "global_step": 194502, "epoch": 2185, "lr": 1.8614168303351204e-05}
{"train_loss": 0.043439362198114395, "global_step": 194503, "epoch": 2185, "lr": 1.8613717014184594e-05}
{"train_loss": 0.05089086294174194, "global_step": 194504, "epoch": 2185, "lr": 1.8613265729237466e-05}
{"train_loss": 0.07728762179613113, "global_step": 194505, "epoch": 2185, "lr": 1.8612814448509906e-05}
{"train_loss": 0.047264326363801956, "global_step": 194506, "epoch": 2185, "lr": 1.861236317200199e-05}
{"train_loss": 0.09097777307033539, "global_step": 194507, "epoch": 2185, "lr": 1.8611911899713742e-05}
{"train_loss": 0.07393079251050949, "global_step": 194508, "epoch": 2185, "lr": 1.861146063164525e-05}
{"train_loss": 0.030777819454669952, "global_step": 194509, "epoch": 2185, "lr": 1.861100936779655e-05}
{"train_loss": 0.05970494821667671, "global_step": 194510, "epoch": 2185, "lr": 1.8610558108167736e-05}
{"train_loss": 0.07861781865358353, "global_step": 194511, "epoch": 2185, "lr": 1.8610106852758835e-05}
{"train_loss": 0.046765536069869995, "global_step": 194512, "epoch": 2185, "lr": 1.8609655601569937e-05}
{"train_loss": 0.031921837478876114, "global_step": 194513, "epoch": 2185, "lr": 1.8609204354601072e-05}
{"train_loss": 0.07232655584812164, "global_step": 194514, "epoch": 2185, "lr": 1.8608753111852332e-05}
{"train_loss": 0.08940643072128296, "global_step": 194515, "epoch": 2185, "lr": 1.8608301873323748e-05}
{"train_loss": 0.04437371715903282, "global_step": 194516, "epoch": 2185, "lr": 1.8607850639015413e-05}
{"train_loss": 0.08892040699720383, "global_step": 194517, "epoch": 2185, "lr": 1.8607399408927366e-05}
{"train_loss": 0.0586651973426342, "global_step": 194518, "epoch": 2185, "lr": 1.8606948183059657e-05}
{"train_loss": 0.032039884477853775, "global_step": 194519, "epoch": 2185, "lr": 1.8606496961412378e-05}
{"train_loss": 0.07168307155370712, "global_step": 194520, "epoch": 2185, "lr": 1.8606045743985556e-05}
{"train_loss": 0.052860572934150696, "global_step": 194521, "epoch": 2185, "lr": 1.8605594530779285e-05}
{"train_loss": 0.025998523458838463, "global_step": 194522, "epoch": 2185, "lr": 1.8605143321793594e-05}
{"train_loss": 0.03672059625387192, "global_step": 194523, "epoch": 2185, "lr": 1.8604692117028572e-05}
{"train_loss": 0.06175632402300835, "global_step": 194524, "epoch": 2185, "lr": 1.860424091648425e-05}
{"train_loss": 0.04468062147498131, "global_step": 194525, "epoch": 2185, "lr": 1.8603789720160723e-05}
{"train_loss": 0.0295685026794672, "global_step": 194526, "epoch": 2185, "lr": 1.8603338528058025e-05}
{"train_loss": 0.04599887877702713, "global_step": 194527, "epoch": 2185, "lr": 1.860288734017623e-05}
{"train_loss": 0.033656150102615356, "global_step": 194528, "epoch": 2185, "lr": 1.8602436156515383e-05}
{"train_loss": 0.04591388627886772, "global_step": 194529, "epoch": 2185, "lr": 1.860198497707557e-05}
{"train_loss": 0.05322229489684105, "global_step": 194530, "epoch": 2185, "lr": 1.8601533801856823e-05}
{"train_loss": 0.10786260664463043, "global_step": 194531, "epoch": 2185, "lr": 1.8601082630859234e-05}
{"train_loss": 0.040118198841810226, "global_step": 194532, "epoch": 2185, "lr": 1.860063146408283e-05}
{"train_loss": 0.09801845252513885, "global_step": 194533, "epoch": 2185, "lr": 1.8600180301527704e-05}
{"train_loss": 0.10517849773168564, "global_step": 194534, "epoch": 2185, "lr": 1.8599729143193884e-05}
{"train_loss": 0.03379899635910988, "global_step": 194535, "epoch": 2185, "lr": 1.8599277989081464e-05}
{"train_loss": 0.03319329768419266, "global_step": 194536, "epoch": 2185, "lr": 1.8598826839190476e-05}
{"train_loss": 0.12519916892051697, "global_step": 194537, "epoch": 2185, "lr": 1.859837569352101e-05}
{"train_loss": 0.11411548405885696, "global_step": 194538, "epoch": 2185, "lr": 1.8597924552073086e-05}
{"train_loss": 0.06927572935819626, "global_step": 194539, "epoch": 2185, "lr": 1.8597473414846806e-05}
{"train_loss": 0.06524960696697235, "global_step": 194540, "epoch": 2185, "lr": 1.8597022281842202e-05}
{"train_loss": 0.0983535572886467, "global_step": 194541, "epoch": 2185, "lr": 1.859657115305934e-05}
{"train_loss": 0.07368645817041397, "global_step": 194542, "epoch": 2185, "lr": 1.8596120028498303e-05}
{"train_loss": 0.05342205613851547, "global_step": 194543, "epoch": 2185, "lr": 1.8595668908159125e-05}
{"train_loss": 0.09246805310249329, "global_step": 194544, "epoch": 2185, "lr": 1.8595217792041885e-05}
{"train_loss": 0.053249236196279526, "global_step": 194545, "epoch": 2185, "lr": 1.8594766680146624e-05}
{"train_loss": 0.09691659361124039, "global_step": 194546, "epoch": 2185, "lr": 1.859431557247343e-05}
{"train_loss": 0.038711756467819214, "global_step": 194547, "epoch": 2185, "lr": 1.8593864469022326e-05}
{"train_loss": 0.03523032367229462, "global_step": 194548, "epoch": 2185, "lr": 1.8593413369793415e-05}
{"train_loss": 0.05396342650055885, "global_step": 194549, "epoch": 2185, "lr": 1.8592962274786723e-05}
{"train_loss": 0.08588559180498123, "global_step": 194550, "epoch": 2185, "lr": 1.8592511184002338e-05}
{"train_loss": 0.05779305100440979, "global_step": 194551, "epoch": 2185, "lr": 1.859206009744029e-05}
{"train_loss": 0.1308050900697708, "global_step": 194552, "epoch": 2185, "lr": 1.859160901510067e-05}
{"train_loss": 0.0631860744082526, "global_step": 194553, "epoch": 2185, "lr": 1.8591157936983518e-05, "val_loss": 7.995410919189453, "train_action_mse_error": 31.934293746948242}
{"train_loss": 0.05408729985356331, "global_step": 194554, "epoch": 2186, "lr": 1.859070686308891e-05}
{"train_loss": 0.06630861014127731, "global_step": 194555, "epoch": 2186, "lr": 1.85902557934169e-05}
{"train_loss": 0.06144912540912628, "global_step": 194556, "epoch": 2186, "lr": 1.858980472796753e-05}
{"train_loss": 0.06246327608823776, "global_step": 194557, "epoch": 2186, "lr": 1.85893536667409e-05}
{"train_loss": 0.03698702156543732, "global_step": 194558, "epoch": 2186, "lr": 1.858890260973703e-05}
{"train_loss": 0.11822839826345444, "global_step": 194559, "epoch": 2186, "lr": 1.858845155695601e-05}
{"train_loss": 0.04533742368221283, "global_step": 194560, "epoch": 2186, "lr": 1.8588000508397878e-05}
{"train_loss": 0.08946745842695236, "global_step": 194561, "epoch": 2186, "lr": 1.8587549464062726e-05}
{"train_loss": 0.06884241849184036, "global_step": 194562, "epoch": 2186, "lr": 1.8587098423950573e-05}
{"train_loss": 0.0567631870508194, "global_step": 194563, "epoch": 2186, "lr": 1.8586647388061522e-05}
{"train_loss": 0.0691002681851387, "global_step": 194564, "epoch": 2186, "lr": 1.858619635639559e-05}
{"train_loss": 0.05369563028216362, "global_step": 194565, "epoch": 2186, "lr": 1.8585745328952886e-05}
{"train_loss": 0.0691753551363945, "global_step": 194566, "epoch": 2186, "lr": 1.8585294305733425e-05}
{"train_loss": 0.03750329837203026, "global_step": 194567, "epoch": 2186, "lr": 1.858484328673731e-05}
{"train_loss": 0.07958655059337616, "global_step": 194568, "epoch": 2186, "lr": 1.8584392271964555e-05}
{"train_loss": 0.05685960128903389, "global_step": 194569, "epoch": 2186, "lr": 1.8583941261415268e-05}
{"train_loss": 0.08012303709983826, "global_step": 194570, "epoch": 2186, "lr": 1.858349025508947e-05}
{"train_loss": 0.08577387779951096, "global_step": 194571, "epoch": 2186, "lr": 1.8583039252987255e-05}
{"train_loss": 0.059241343289613724, "global_step": 194572, "epoch": 2186, "lr": 1.8582588255108646e-05}
{"train_loss": 0.04389229416847229, "global_step": 194573, "epoch": 2186, "lr": 1.858213726145375e-05}
{"train_loss": 0.032880764454603195, "global_step": 194574, "epoch": 2186, "lr": 1.858168627202258e-05}
{"train_loss": 0.07675071805715561, "global_step": 194575, "epoch": 2186, "lr": 1.858123528681524e-05}
{"train_loss": 0.05566108599305153, "global_step": 194576, "epoch": 2186, "lr": 1.858078430583175e-05}
{"train_loss": 0.07989596575498581, "global_step": 194577, "epoch": 2186, "lr": 1.858033332907219e-05}
{"train_loss": 0.0375189445912838, "global_step": 194578, "epoch": 2186, "lr": 1.8579882356536644e-05}
{"train_loss": 0.09889573603868484, "global_step": 194579, "epoch": 2186, "lr": 1.8579431388225126e-05}
{"train_loss": 0.025163063779473305, "global_step": 194580, "epoch": 2186, "lr": 1.8578980424137736e-05}
{"train_loss": 0.0746961161494255, "global_step": 194581, "epoch": 2186, "lr": 1.8578529464274508e-05}
{"train_loss": 0.11119087785482407, "global_step": 194582, "epoch": 2186, "lr": 1.8578078508635533e-05}
{"train_loss": 0.0715511292219162, "global_step": 194583, "epoch": 2186, "lr": 1.8577627557220828e-05}
{"train_loss": 0.06800160557031631, "global_step": 194584, "epoch": 2186, "lr": 1.85771766100305e-05}
{"train_loss": 0.07222416996955872, "global_step": 194585, "epoch": 2186, "lr": 1.857672566706457e-05}
{"train_loss": 0.06526706367731094, "global_step": 194586, "epoch": 2186, "lr": 1.8576274728323136e-05}
{"train_loss": 0.09883760660886765, "global_step": 194587, "epoch": 2186, "lr": 1.8575823793806212e-05}
{"train_loss": 0.08285962045192719, "global_step": 194588, "epoch": 2186, "lr": 1.8575372863513913e-05}
{"train_loss": 0.07021813094615936, "global_step": 194589, "epoch": 2186, "lr": 1.857492193744625e-05}
{"train_loss": 0.03929886221885681, "global_step": 194590, "epoch": 2186, "lr": 1.8574471015603323e-05}
{"train_loss": 0.09484528005123138, "global_step": 194591, "epoch": 2186, "lr": 1.8574020097985158e-05}
{"train_loss": 0.028817037120461464, "global_step": 194592, "epoch": 2186, "lr": 1.857356918459185e-05}
{"train_loss": 0.04542980343103409, "global_step": 194593, "epoch": 2186, "lr": 1.8573118275423428e-05}
{"train_loss": 0.05629270151257515, "global_step": 194594, "epoch": 2186, "lr": 1.857266737047998e-05}
{"train_loss": 0.0768708810210228, "global_step": 194595, "epoch": 2186, "lr": 1.8572216469761554e-05}
{"train_loss": 0.07605956494808197, "global_step": 194596, "epoch": 2186, "lr": 1.85717655732682e-05}
{"train_loss": 0.06661265343427658, "global_step": 194597, "epoch": 2186, "lr": 1.8571314681000003e-05}
{"train_loss": 0.04883139207959175, "global_step": 194598, "epoch": 2186, "lr": 1.857086379295699e-05}
{"train_loss": 0.05377841368317604, "global_step": 194599, "epoch": 2186, "lr": 1.8570412909139263e-05}
{"train_loss": 0.12407608330249786, "global_step": 194600, "epoch": 2186, "lr": 1.856996202954684e-05}
{"train_loss": 0.09622547030448914, "global_step": 194601, "epoch": 2186, "lr": 1.856951115417982e-05}
{"train_loss": 0.07942941784858704, "global_step": 194602, "epoch": 2186, "lr": 1.8569060283038226e-05}
{"train_loss": 0.053786564618349075, "global_step": 194603, "epoch": 2186, "lr": 1.8568609416122162e-05}
{"train_loss": 0.043353401124477386, "global_step": 194604, "epoch": 2186, "lr": 1.8568158553431647e-05}
{"train_loss": 0.0838766023516655, "global_step": 194605, "epoch": 2186, "lr": 1.8567707694966773e-05}
{"train_loss": 0.06502178311347961, "global_step": 194606, "epoch": 2186, "lr": 1.8567256840727577e-05}
{"train_loss": 0.0918601006269455, "global_step": 194607, "epoch": 2186, "lr": 1.8566805990714143e-05}
{"train_loss": 0.08354255557060242, "global_step": 194608, "epoch": 2186, "lr": 1.85663551449265e-05}
{"train_loss": 0.05904499813914299, "global_step": 194609, "epoch": 2186, "lr": 1.8565904303364744e-05}
{"train_loss": 0.05605228990316391, "global_step": 194610, "epoch": 2186, "lr": 1.856545346602891e-05}
{"train_loss": 0.0850633978843689, "global_step": 194611, "epoch": 2186, "lr": 1.8565002632919076e-05}
{"train_loss": 0.057754598557949066, "global_step": 194612, "epoch": 2186, "lr": 1.856455180403528e-05}
{"train_loss": 0.10484670847654343, "global_step": 194613, "epoch": 2186, "lr": 1.8564100979377597e-05}
{"train_loss": 0.08445926755666733, "global_step": 194614, "epoch": 2186, "lr": 1.8563650158946105e-05}
{"train_loss": 0.07627829909324646, "global_step": 194615, "epoch": 2186, "lr": 1.8563199342740833e-05}
{"train_loss": 0.03644319996237755, "global_step": 194616, "epoch": 2186, "lr": 1.856274853076187e-05}
{"train_loss": 0.07040101289749146, "global_step": 194617, "epoch": 2186, "lr": 1.8562297723009243e-05}
{"train_loss": 0.10253672301769257, "global_step": 194618, "epoch": 2186, "lr": 1.856184691948305e-05}
{"train_loss": 0.038490764796733856, "global_step": 194619, "epoch": 2186, "lr": 1.8561396120183317e-05}
{"train_loss": 0.09346884489059448, "global_step": 194620, "epoch": 2186, "lr": 1.856094532511014e-05}
{"train_loss": 0.08174122124910355, "global_step": 194621, "epoch": 2186, "lr": 1.856049453426354e-05}
{"train_loss": 0.09694435447454453, "global_step": 194622, "epoch": 2186, "lr": 1.856004374764362e-05}
{"train_loss": 0.07046402245759964, "global_step": 194623, "epoch": 2186, "lr": 1.85595929652504e-05}
{"train_loss": 0.05272883176803589, "global_step": 194624, "epoch": 2186, "lr": 1.8559142187083972e-05}
{"train_loss": 0.10593221336603165, "global_step": 194625, "epoch": 2186, "lr": 1.8558691413144373e-05}
{"train_loss": 0.05420524999499321, "global_step": 194626, "epoch": 2186, "lr": 1.855824064343169e-05}
{"train_loss": 0.0080754728987813, "global_step": 194627, "epoch": 2186, "lr": 1.8557789877945954e-05}
{"train_loss": 0.0662803128361702, "global_step": 194628, "epoch": 2186, "lr": 1.855733911668725e-05}
{"train_loss": 0.0805421844124794, "global_step": 194629, "epoch": 2186, "lr": 1.8556888359655616e-05}
{"train_loss": 0.09452959150075912, "global_step": 194630, "epoch": 2186, "lr": 1.8556437606851146e-05}
{"train_loss": 0.06450115144252777, "global_step": 194631, "epoch": 2186, "lr": 1.8555986858273855e-05}
{"train_loss": 0.04951869696378708, "global_step": 194632, "epoch": 2186, "lr": 1.855553611392385e-05}
{"train_loss": 0.05976087227463722, "global_step": 194633, "epoch": 2186, "lr": 1.855508537380117e-05}
{"train_loss": 0.13215413689613342, "global_step": 194634, "epoch": 2186, "lr": 1.8554634637905856e-05}
{"train_loss": 0.06840283423662186, "global_step": 194635, "epoch": 2186, "lr": 1.8554183906238005e-05}
{"train_loss": 0.07401345670223236, "global_step": 194636, "epoch": 2186, "lr": 1.8553733178797646e-05}
{"train_loss": 0.042374785989522934, "global_step": 194637, "epoch": 2186, "lr": 1.8553282455584868e-05}
{"train_loss": 0.08810621500015259, "global_step": 194638, "epoch": 2186, "lr": 1.8552831736599707e-05}
{"train_loss": 0.052038632333278656, "global_step": 194639, "epoch": 2186, "lr": 1.8552381021842246e-05}
{"train_loss": 0.06759534031152725, "global_step": 194640, "epoch": 2186, "lr": 1.855193031131252e-05}
{"train_loss": 0.09782465547323227, "global_step": 194641, "epoch": 2186, "lr": 1.855147960501062e-05}
{"train_loss": 0.06959405981966954, "global_step": 194642, "epoch": 2186, "lr": 1.855102890293657e-05, "val_loss": 8.106322288513184}
{"train_loss": 0.07555484026670456, "global_step": 194643, "epoch": 2187, "lr": 1.855057820509047e-05}
{"train_loss": 0.06338571012020111, "global_step": 194644, "epoch": 2187, "lr": 1.855012751147235e-05}
{"train_loss": 0.06436654180288315, "global_step": 194645, "epoch": 2187, "lr": 1.854967682208229e-05}
{"train_loss": 0.13482151925563812, "global_step": 194646, "epoch": 2187, "lr": 1.8549226136920328e-05}
{"train_loss": 0.07516742497682571, "global_step": 194647, "epoch": 2187, "lr": 1.8548775455986563e-05}
{"train_loss": 0.12128785997629166, "global_step": 194648, "epoch": 2187, "lr": 1.8548324779281007e-05}
{"train_loss": 0.06934379786252975, "global_step": 194649, "epoch": 2187, "lr": 1.8547874106803748e-05}
{"train_loss": 0.036259688436985016, "global_step": 194650, "epoch": 2187, "lr": 1.854742343855486e-05}
{"train_loss": 0.029889410361647606, "global_step": 194651, "epoch": 2187, "lr": 1.854697277453437e-05}
{"train_loss": 0.06254217773675919, "global_step": 194652, "epoch": 2187, "lr": 1.8546522114742376e-05}
{"train_loss": 0.04884243384003639, "global_step": 194653, "epoch": 2187, "lr": 1.8546071459178898e-05}
{"train_loss": 0.10438007116317749, "global_step": 194654, "epoch": 2187, "lr": 1.854562080784404e-05}
{"train_loss": 0.1431753933429718, "global_step": 194655, "epoch": 2187, "lr": 1.8545170160737818e-05}
{"train_loss": 0.027339037507772446, "global_step": 194656, "epoch": 2187, "lr": 1.8544719517860336e-05}
{"train_loss": 0.048881739377975464, "global_step": 194657, "epoch": 2187, "lr": 1.8544268879211614e-05}
{"train_loss": 0.06072768196463585, "global_step": 194658, "epoch": 2187, "lr": 1.8543818244791745e-05}
{"train_loss": 0.08934514969587326, "global_step": 194659, "epoch": 2187, "lr": 1.8543367614600766e-05}
{"train_loss": 0.12102877348661423, "global_step": 194660, "epoch": 2187, "lr": 1.854291698863876e-05}
{"train_loss": 0.042695920914411545, "global_step": 194661, "epoch": 2187, "lr": 1.854246636690576e-05}
{"train_loss": 0.12036243081092834, "global_step": 194662, "epoch": 2187, "lr": 1.854201574940186e-05}
{"train_loss": 0.038301847875118256, "global_step": 194663, "epoch": 2187, "lr": 1.854156513612708e-05}
{"train_loss": 0.04838135465979576, "global_step": 194664, "epoch": 2187, "lr": 1.8541114527081528e-05}
{"train_loss": 0.09787924587726593, "global_step": 194665, "epoch": 2187, "lr": 1.8540663922265217e-05}
{"train_loss": 0.058668140321969986, "global_step": 194666, "epoch": 2187, "lr": 1.8540213321678246e-05}
{"train_loss": 0.09047979861497879, "global_step": 194667, "epoch": 2187, "lr": 1.8539762725320646e-05}
{"train_loss": 0.06269416213035583, "global_step": 194668, "epoch": 2187, "lr": 1.853931213319251e-05}
{"train_loss": 0.070491261780262, "global_step": 194669, "epoch": 2187, "lr": 1.853886154529386e-05}
{"train_loss": 0.03359271585941315, "global_step": 194670, "epoch": 2187, "lr": 1.8538410961624798e-05}
{"train_loss": 0.07663719356060028, "global_step": 194671, "epoch": 2187, "lr": 1.853796038218536e-05}
{"train_loss": 0.031909435987472534, "global_step": 194672, "epoch": 2187, "lr": 1.8537509806975588e-05}
{"train_loss": 0.06248094514012337, "global_step": 194673, "epoch": 2187, "lr": 1.853705923599559e-05}
{"train_loss": 0.08887283504009247, "global_step": 194674, "epoch": 2187, "lr": 1.853660866924538e-05}
{"train_loss": 0.05528009310364723, "global_step": 194675, "epoch": 2187, "lr": 1.8536158106725056e-05}
{"train_loss": 0.025440139696002007, "global_step": 194676, "epoch": 2187, "lr": 1.8535707548434645e-05}
{"train_loss": 0.054123908281326294, "global_step": 194677, "epoch": 2187, "lr": 1.8535256994374245e-05}
{"train_loss": 0.06399474292993546, "global_step": 194678, "epoch": 2187, "lr": 1.8534806444543874e-05}
{"train_loss": 0.07437992841005325, "global_step": 194679, "epoch": 2187, "lr": 1.8534355898943633e-05}
{"train_loss": 0.09982229769229889, "global_step": 194680, "epoch": 2187, "lr": 1.853390535757355e-05}
{"train_loss": 0.0406353697180748, "global_step": 194681, "epoch": 2187, "lr": 1.8533454820433716e-05}
{"train_loss": 0.058481913059949875, "global_step": 194682, "epoch": 2187, "lr": 1.853300428752416e-05}
{"train_loss": 0.09597306698560715, "global_step": 194683, "epoch": 2187, "lr": 1.8532553758844973e-05}
{"train_loss": 0.09629573673009872, "global_step": 194684, "epoch": 2187, "lr": 1.853210323439618e-05}
{"train_loss": 0.059396542608737946, "global_step": 194685, "epoch": 2187, "lr": 1.853165271417789e-05}
{"train_loss": 0.04077906161546707, "global_step": 194686, "epoch": 2187, "lr": 1.8531202198190106e-05}
{"train_loss": 0.059942375868558884, "global_step": 194687, "epoch": 2187, "lr": 1.8530751686432924e-05}
{"train_loss": 0.04983983561396599, "global_step": 194688, "epoch": 2187, "lr": 1.8530301178906417e-05}
{"train_loss": 0.04813379794359207, "global_step": 194689, "epoch": 2187, "lr": 1.852985067561061e-05}
{"train_loss": 0.05203810706734657, "global_step": 194690, "epoch": 2187, "lr": 1.8529400176545596e-05}
{"train_loss": 0.05109134316444397, "global_step": 194691, "epoch": 2187, "lr": 1.8528949681711403e-05}
{"train_loss": 0.09723558276891708, "global_step": 194692, "epoch": 2187, "lr": 1.8528499191108134e-05}
{"train_loss": 0.0727057084441185, "global_step": 194693, "epoch": 2187, "lr": 1.85280487047358e-05}
{"train_loss": 0.08491884171962738, "global_step": 194694, "epoch": 2187, "lr": 1.8527598222594505e-05}
{"train_loss": 0.05013315752148628, "global_step": 194695, "epoch": 2187, "lr": 1.8527147744684275e-05}
{"train_loss": 0.052784692496061325, "global_step": 194696, "epoch": 2187, "lr": 1.85266972710052e-05}
{"train_loss": 0.05315060541033745, "global_step": 194697, "epoch": 2187, "lr": 1.852624680155731e-05}
{"train_loss": 0.047674305737018585, "global_step": 194698, "epoch": 2187, "lr": 1.8525796336340707e-05}
{"train_loss": 0.05871813744306564, "global_step": 194699, "epoch": 2187, "lr": 1.8525345875355405e-05}
{"train_loss": 0.04860023036599159, "global_step": 194700, "epoch": 2187, "lr": 1.85248954186015e-05}
{"train_loss": 0.027435794472694397, "global_step": 194701, "epoch": 2187, "lr": 1.8524444966079026e-05}
{"train_loss": 0.039521392434835434, "global_step": 194702, "epoch": 2187, "lr": 1.8523994517788074e-05}
{"train_loss": 0.06265132129192352, "global_step": 194703, "epoch": 2187, "lr": 1.8523544073728665e-05}
{"train_loss": 0.030516916885972023, "global_step": 194704, "epoch": 2187, "lr": 1.8523093633900907e-05}
{"train_loss": 0.04765800014138222, "global_step": 194705, "epoch": 2187, "lr": 1.852264319830481e-05}
{"train_loss": 0.07580269128084183, "global_step": 194706, "epoch": 2187, "lr": 1.852219276694049e-05}
{"train_loss": 0.0503043457865715, "global_step": 194707, "epoch": 2187, "lr": 1.8521742339807952e-05}
{"train_loss": 0.07167830318212509, "global_step": 194708, "epoch": 2187, "lr": 1.8521291916907298e-05}
{"train_loss": 0.11193985491991043, "global_step": 194709, "epoch": 2187, "lr": 1.8520841498238555e-05}
{"train_loss": 0.07178839296102524, "global_step": 194710, "epoch": 2187, "lr": 1.852039108380182e-05}
{"train_loss": 0.03918921574950218, "global_step": 194711, "epoch": 2187, "lr": 1.8519940673597135e-05}
{"train_loss": 0.0898161306977272, "global_step": 194712, "epoch": 2187, "lr": 1.8519490267624546e-05}
{"train_loss": 0.06797045469284058, "global_step": 194713, "epoch": 2187, "lr": 1.851903986588414e-05}
{"train_loss": 0.036666139960289, "global_step": 194714, "epoch": 2187, "lr": 1.851858946837595e-05}
{"train_loss": 0.050672318786382675, "global_step": 194715, "epoch": 2187, "lr": 1.851813907510007e-05}
{"train_loss": 0.09363040328025818, "global_step": 194716, "epoch": 2187, "lr": 1.8517688686056522e-05}
{"train_loss": 0.13689231872558594, "global_step": 194717, "epoch": 2187, "lr": 1.851723830124541e-05}
{"train_loss": 0.08892253786325455, "global_step": 194718, "epoch": 2187, "lr": 1.851678792066675e-05}
{"train_loss": 0.09136564284563065, "global_step": 194719, "epoch": 2187, "lr": 1.8516337544320643e-05}
{"train_loss": 0.0739826112985611, "global_step": 194720, "epoch": 2187, "lr": 1.8515887172207115e-05}
{"train_loss": 0.08529514819383621, "global_step": 194721, "epoch": 2187, "lr": 1.851543680432626e-05}
{"train_loss": 0.1010148823261261, "global_step": 194722, "epoch": 2187, "lr": 1.85149864406781e-05}
{"train_loss": 0.04932789504528046, "global_step": 194723, "epoch": 2187, "lr": 1.8514536081262717e-05}
{"train_loss": 0.07414703071117401, "global_step": 194724, "epoch": 2187, "lr": 1.8514085726080188e-05}
{"train_loss": 0.035113219171762466, "global_step": 194725, "epoch": 2187, "lr": 1.851363537513054e-05}
{"train_loss": 0.046358153223991394, "global_step": 194726, "epoch": 2187, "lr": 1.8513185028413866e-05}
{"train_loss": 0.0889340192079544, "global_step": 194727, "epoch": 2187, "lr": 1.8512734685930194e-05}
{"train_loss": 0.11778464168310165, "global_step": 194728, "epoch": 2187, "lr": 1.851228434767962e-05}
{"train_loss": 0.05701782926917076, "global_step": 194729, "epoch": 2187, "lr": 1.8511834013662165e-05}
{"train_loss": 0.12596409022808075, "global_step": 194730, "epoch": 2187, "lr": 1.8511383683877932e-05}
{"train_loss": 0.06849492615444606, "global_step": 194731, "epoch": 2187, "lr": 1.8510933358326937e-05, "val_loss": 7.988087177276611}
{"train_loss": 0.04593346640467644, "global_step": 194732, "epoch": 2188, "lr": 1.8510483037009284e-05}
{"train_loss": 0.04068373516201973, "global_step": 194733, "epoch": 2188, "lr": 1.8510032719924996e-05}
{"train_loss": 0.08905073255300522, "global_step": 194734, "epoch": 2188, "lr": 1.8509582407074165e-05}
{"train_loss": 0.021819885820150375, "global_step": 194735, "epoch": 2188, "lr": 1.8509132098456822e-05}
{"train_loss": 0.13280580937862396, "global_step": 194736, "epoch": 2188, "lr": 1.850868179407306e-05}
{"train_loss": 0.06929350644350052, "global_step": 194737, "epoch": 2188, "lr": 1.85082314939229e-05}
{"train_loss": 0.055667728185653687, "global_step": 194738, "epoch": 2188, "lr": 1.8507781198006447e-05}
{"train_loss": 0.07522854208946228, "global_step": 194739, "epoch": 2188, "lr": 1.8507330906323716e-05}
{"train_loss": 0.05623839423060417, "global_step": 194740, "epoch": 2188, "lr": 1.8506880618874793e-05}
{"train_loss": 0.03457720950245857, "global_step": 194741, "epoch": 2188, "lr": 1.8506430335659758e-05}
{"train_loss": 0.05073151737451553, "global_step": 194742, "epoch": 2188, "lr": 1.8505980056678628e-05}
{"train_loss": 0.04814999923110008, "global_step": 194743, "epoch": 2188, "lr": 1.8505529781931503e-05}
{"train_loss": 0.05661056935787201, "global_step": 194744, "epoch": 2188, "lr": 1.850507951141843e-05}
{"train_loss": 0.05970224365592003, "global_step": 194745, "epoch": 2188, "lr": 1.8504629245139438e-05}
{"train_loss": 0.05795300751924515, "global_step": 194746, "epoch": 2188, "lr": 1.850417898309464e-05}
{"train_loss": 0.024259353056550026, "global_step": 194747, "epoch": 2188, "lr": 1.8503728725284054e-05}
{"train_loss": 0.08769385516643524, "global_step": 194748, "epoch": 2188, "lr": 1.850327847170777e-05}
{"train_loss": 0.0807611495256424, "global_step": 194749, "epoch": 2188, "lr": 1.8502828222365835e-05}
{"train_loss": 0.04779771342873573, "global_step": 194750, "epoch": 2188, "lr": 1.85023779772583e-05}
{"train_loss": 0.0412408821284771, "global_step": 194751, "epoch": 2188, "lr": 1.8501927736385254e-05}
{"train_loss": 0.07611513882875443, "global_step": 194752, "epoch": 2188, "lr": 1.8501477499746718e-05}
{"train_loss": 0.09422551840543747, "global_step": 194753, "epoch": 2188, "lr": 1.8501027267342797e-05}
{"train_loss": 0.046990953385829926, "global_step": 194754, "epoch": 2188, "lr": 1.8500577039173506e-05}
{"train_loss": 0.06626122444868088, "global_step": 194755, "epoch": 2188, "lr": 1.850012681523895e-05}
{"train_loss": 0.041634701192379, "global_step": 194756, "epoch": 2188, "lr": 1.8499676595539144e-05}
{"train_loss": 0.11471468955278397, "global_step": 194757, "epoch": 2188, "lr": 1.84992263800742e-05}
{"train_loss": 0.03551587834954262, "global_step": 194758, "epoch": 2188, "lr": 1.849877616884413e-05}
{"train_loss": 0.06669546663761139, "global_step": 194759, "epoch": 2188, "lr": 1.8498325961849012e-05}
{"train_loss": 0.08794647455215454, "global_step": 194760, "epoch": 2188, "lr": 1.8497875759088928e-05}
{"train_loss": 0.05257762596011162, "global_step": 194761, "epoch": 2188, "lr": 1.8497425560563903e-05}
{"train_loss": 0.04801357537508011, "global_step": 194762, "epoch": 2188, "lr": 1.8496975366274032e-05}
{"train_loss": 0.07132961601018906, "global_step": 194763, "epoch": 2188, "lr": 1.8496525176219347e-05}
{"train_loss": 0.046737104654312134, "global_step": 194764, "epoch": 2188, "lr": 1.8496074990399938e-05}
{"train_loss": 0.03350282832980156, "global_step": 194765, "epoch": 2188, "lr": 1.849562480881582e-05}
{"train_loss": 0.05723261833190918, "global_step": 194766, "epoch": 2188, "lr": 1.8495174631467112e-05}
{"train_loss": 0.09019742906093597, "global_step": 194767, "epoch": 2188, "lr": 1.8494724458353817e-05}
{"train_loss": 0.06625071167945862, "global_step": 194768, "epoch": 2188, "lr": 1.8494274289476043e-05}
{"train_loss": 0.0346231535077095, "global_step": 194769, "epoch": 2188, "lr": 1.849382412483381e-05}
{"train_loss": 0.05649957433342934, "global_step": 194770, "epoch": 2188, "lr": 1.849337396442722e-05}
{"train_loss": 0.043924566358327866, "global_step": 194771, "epoch": 2188, "lr": 1.849292380825629e-05}
{"train_loss": 0.09317076951265335, "global_step": 194772, "epoch": 2188, "lr": 1.849247365632112e-05}
{"train_loss": 0.029080567881464958, "global_step": 194773, "epoch": 2188, "lr": 1.8492023508621738e-05}
{"train_loss": 0.08792909234762192, "global_step": 194774, "epoch": 2188, "lr": 1.849157336515824e-05}
{"train_loss": 0.04459112510085106, "global_step": 194775, "epoch": 2188, "lr": 1.849112322593064e-05}
{"train_loss": 0.05940777808427811, "global_step": 194776, "epoch": 2188, "lr": 1.849067309093903e-05}
{"train_loss": 0.09888337552547455, "global_step": 194777, "epoch": 2188, "lr": 1.849022296018348e-05}
{"train_loss": 0.058899957686662674, "global_step": 194778, "epoch": 2188, "lr": 1.848977283366402e-05}
{"train_loss": 0.12993480265140533, "global_step": 194779, "epoch": 2188, "lr": 1.8489322711380742e-05}
{"train_loss": 0.11604492366313934, "global_step": 194780, "epoch": 2188, "lr": 1.8488872593333673e-05}
{"train_loss": 0.08271953463554382, "global_step": 194781, "epoch": 2188, "lr": 1.848842247952291e-05}
{"train_loss": 0.057931289076805115, "global_step": 194782, "epoch": 2188, "lr": 1.848797236994849e-05}
{"train_loss": 0.05321597307920456, "global_step": 194783, "epoch": 2188, "lr": 1.848752226461047e-05}
{"train_loss": 0.08222693204879761, "global_step": 194784, "epoch": 2188, "lr": 1.8487072163508927e-05}
{"train_loss": 0.07017098367214203, "global_step": 194785, "epoch": 2188, "lr": 1.84866220666439e-05}
{"train_loss": 0.05755651742219925, "global_step": 194786, "epoch": 2188, "lr": 1.848617197401548e-05}
{"train_loss": 0.07965011149644852, "global_step": 194787, "epoch": 2188, "lr": 1.8485721885623706e-05}
{"train_loss": 0.0666629895567894, "global_step": 194788, "epoch": 2188, "lr": 1.8485271801468634e-05}
{"train_loss": 0.03308062627911568, "global_step": 194789, "epoch": 2188, "lr": 1.8484821721550343e-05}
{"train_loss": 0.09444878250360489, "global_step": 194790, "epoch": 2188, "lr": 1.848437164586887e-05}
{"train_loss": 0.0656033456325531, "global_step": 194791, "epoch": 2188, "lr": 1.8483921574424302e-05}
{"train_loss": 0.035953182727098465, "global_step": 194792, "epoch": 2188, "lr": 1.8483471507216672e-05}
{"train_loss": 0.04588877409696579, "global_step": 194793, "epoch": 2188, "lr": 1.8483021444246073e-05}
{"train_loss": 0.11785811185836792, "global_step": 194794, "epoch": 2188, "lr": 1.8482571385512527e-05}
{"train_loss": 0.04427952691912651, "global_step": 194795, "epoch": 2188, "lr": 1.8482121331016117e-05}
{"train_loss": 0.09214901179075241, "global_step": 194796, "epoch": 2188, "lr": 1.8481671280756924e-05}
{"train_loss": 0.08901242166757584, "global_step": 194797, "epoch": 2188, "lr": 1.8481221234734964e-05}
{"train_loss": 0.10538811981678009, "global_step": 194798, "epoch": 2188, "lr": 1.8480771192950336e-05}
{"train_loss": 0.036426521837711334, "global_step": 194799, "epoch": 2188, "lr": 1.848032115540307e-05}
{"train_loss": 0.06657887995243073, "global_step": 194800, "epoch": 2188, "lr": 1.8479871122093257e-05}
{"train_loss": 0.048129066824913025, "global_step": 194801, "epoch": 2188, "lr": 1.847942109302092e-05}
{"train_loss": 0.11006800830364227, "global_step": 194802, "epoch": 2188, "lr": 1.8478971068186164e-05}
{"train_loss": 0.1093362346291542, "global_step": 194803, "epoch": 2188, "lr": 1.8478521047589005e-05}
{"train_loss": 0.044802289456129074, "global_step": 194804, "epoch": 2188, "lr": 1.8478071031229538e-05}
{"train_loss": 0.09807580709457397, "global_step": 194805, "epoch": 2188, "lr": 1.8477621019107798e-05}
{"train_loss": 0.07666444033384323, "global_step": 194806, "epoch": 2188, "lr": 1.8477171011223875e-05}
{"train_loss": 0.08427540212869644, "global_step": 194807, "epoch": 2188, "lr": 1.8476721007577797e-05}
{"train_loss": 0.032828256487846375, "global_step": 194808, "epoch": 2188, "lr": 1.8476271008169655e-05}
{"train_loss": 0.0966116413474083, "global_step": 194809, "epoch": 2188, "lr": 1.8475821012999473e-05}
{"train_loss": 0.062004588544368744, "global_step": 194810, "epoch": 2188, "lr": 1.8475371022067356e-05}
{"train_loss": 0.07583841681480408, "global_step": 194811, "epoch": 2188, "lr": 1.847492103537332e-05}
{"train_loss": 0.07434865087270737, "global_step": 194812, "epoch": 2188, "lr": 1.8474471052917463e-05}
{"train_loss": 0.04094408079981804, "global_step": 194813, "epoch": 2188, "lr": 1.8474021074699816e-05}
{"train_loss": 0.07784213125705719, "global_step": 194814, "epoch": 2188, "lr": 1.847357110072045e-05}
{"train_loss": 0.04934101551771164, "global_step": 194815, "epoch": 2188, "lr": 1.8473121130979447e-05}
{"train_loss": 0.08166943490505219, "global_step": 194816, "epoch": 2188, "lr": 1.847267116547683e-05}
{"train_loss": 0.054671913385391235, "global_step": 194817, "epoch": 2188, "lr": 1.847222120421269e-05}
{"train_loss": 0.07957116514444351, "global_step": 194818, "epoch": 2188, "lr": 1.847177124718707e-05}
{"train_loss": 0.06916956603527069, "global_step": 194819, "epoch": 2188, "lr": 1.8471321294400044e-05}
{"train_loss": 0.06639944967091753, "global_step": 194820, "epoch": 2188, "lr": 1.847087134585167e-05, "val_loss": 8.17521858215332}
{"train_loss": 0.059712156653404236, "global_step": 194821, "epoch": 2189, "lr": 1.8470421401541982e-05}
{"train_loss": 0.11680705845355988, "global_step": 194822, "epoch": 2189, "lr": 1.846997146147108e-05}
{"train_loss": 0.15920697152614594, "global_step": 194823, "epoch": 2189, "lr": 1.8469521525638988e-05}
{"train_loss": 0.06591367721557617, "global_step": 194824, "epoch": 2189, "lr": 1.8469071594045805e-05}
{"train_loss": 0.03197357431054115, "global_step": 194825, "epoch": 2189, "lr": 1.8468621666691547e-05}
{"train_loss": 0.02337142638862133, "global_step": 194826, "epoch": 2189, "lr": 1.8468171743576324e-05}
{"train_loss": 0.08183170109987259, "global_step": 194827, "epoch": 2189, "lr": 1.8467721824700168e-05}
{"train_loss": 0.06774678081274033, "global_step": 194828, "epoch": 2189, "lr": 1.8467271910063128e-05}
{"train_loss": 0.03233756124973297, "global_step": 194829, "epoch": 2189, "lr": 1.846682199966529e-05}
{"train_loss": 0.07236464321613312, "global_step": 194830, "epoch": 2189, "lr": 1.846637209350669e-05}
{"train_loss": 0.09876153618097305, "global_step": 194831, "epoch": 2189, "lr": 1.84659221915874e-05}
{"train_loss": 0.05050421133637428, "global_step": 194832, "epoch": 2189, "lr": 1.8465472293907498e-05}
{"train_loss": 0.03242850303649902, "global_step": 194833, "epoch": 2189, "lr": 1.846502240046702e-05}
{"train_loss": 0.09736339747905731, "global_step": 194834, "epoch": 2189, "lr": 1.8464572511266044e-05}
{"train_loss": 0.08870282024145126, "global_step": 194835, "epoch": 2189, "lr": 1.846412262630461e-05}
{"train_loss": 0.052266690880060196, "global_step": 194836, "epoch": 2189, "lr": 1.84636727455828e-05}
{"train_loss": 0.1218111664056778, "global_step": 194837, "epoch": 2189, "lr": 1.8463222869100656e-05}
{"train_loss": 0.03997930884361267, "global_step": 194838, "epoch": 2189, "lr": 1.8462772996858258e-05}
{"train_loss": 0.05431205779314041, "global_step": 194839, "epoch": 2189, "lr": 1.8462323128855647e-05}
{"train_loss": 0.08615023642778397, "global_step": 194840, "epoch": 2189, "lr": 1.84618732650929e-05}
{"train_loss": 0.04175977408885956, "global_step": 194841, "epoch": 2189, "lr": 1.8461423405570057e-05}
{"train_loss": 0.020406166091561317, "global_step": 194842, "epoch": 2189, "lr": 1.8460973550287207e-05}
{"train_loss": 0.0694199949502945, "global_step": 194843, "epoch": 2189, "lr": 1.846052369924438e-05}
{"train_loss": 0.10141286998987198, "global_step": 194844, "epoch": 2189, "lr": 1.846007385244167e-05}
{"train_loss": 0.050737008452415466, "global_step": 194845, "epoch": 2189, "lr": 1.8459624009879095e-05}
{"train_loss": 0.08351081609725952, "global_step": 194846, "epoch": 2189, "lr": 1.845917417155676e-05}
{"train_loss": 0.06926161795854568, "global_step": 194847, "epoch": 2189, "lr": 1.8458724337474686e-05}
{"train_loss": 0.07307107746601105, "global_step": 194848, "epoch": 2189, "lr": 1.8458274507632972e-05}
{"train_loss": 0.08310539275407791, "global_step": 194849, "epoch": 2189, "lr": 1.8457824682031644e-05}
{"train_loss": 0.05924715846776962, "global_step": 194850, "epoch": 2189, "lr": 1.8457374860670767e-05}
{"train_loss": 0.09941839426755905, "global_step": 194851, "epoch": 2189, "lr": 1.8456925043550437e-05}
{"train_loss": 0.034493591636419296, "global_step": 194852, "epoch": 2189, "lr": 1.8456475230670673e-05}
{"train_loss": 0.10170230269432068, "global_step": 194853, "epoch": 2189, "lr": 1.8456025422031563e-05}
{"train_loss": 0.08793055266141891, "global_step": 194854, "epoch": 2189, "lr": 1.8455575617633143e-05}
{"train_loss": 0.05487705022096634, "global_step": 194855, "epoch": 2189, "lr": 1.84551258174755e-05}
{"train_loss": 0.0648273229598999, "global_step": 194856, "epoch": 2189, "lr": 1.8454676021558665e-05}
{"train_loss": 0.08726184070110321, "global_step": 194857, "epoch": 2189, "lr": 1.845422622988273e-05}
{"train_loss": 0.09258142113685608, "global_step": 194858, "epoch": 2189, "lr": 1.8453776442447725e-05}
{"train_loss": 0.040827587246894836, "global_step": 194859, "epoch": 2189, "lr": 1.8453326659253738e-05}
{"train_loss": 0.06286562234163284, "global_step": 194860, "epoch": 2189, "lr": 1.8452876880300822e-05}
{"train_loss": 0.09817828238010406, "global_step": 194861, "epoch": 2189, "lr": 1.8452427105589017e-05}
{"train_loss": 0.0581449493765831, "global_step": 194862, "epoch": 2189, "lr": 1.845197733511841e-05}
{"train_loss": 0.05447489768266678, "global_step": 194863, "epoch": 2189, "lr": 1.8451527568889033e-05}
{"train_loss": 0.04946267232298851, "global_step": 194864, "epoch": 2189, "lr": 1.8451077806900984e-05}
{"train_loss": 0.09339439123868942, "global_step": 194865, "epoch": 2189, "lr": 1.84506280491543e-05}
{"train_loss": 0.05833094194531441, "global_step": 194866, "epoch": 2189, "lr": 1.8450178295649035e-05}
{"train_loss": 0.13311095535755157, "global_step": 194867, "epoch": 2189, "lr": 1.8449728546385247e-05}
{"train_loss": 0.05015743151307106, "global_step": 194868, "epoch": 2189, "lr": 1.8449278801363034e-05}
{"train_loss": 0.05732530355453491, "global_step": 194869, "epoch": 2189, "lr": 1.8448829060582413e-05}
{"train_loss": 0.048203300684690475, "global_step": 194870, "epoch": 2189, "lr": 1.844837932404348e-05}
{"train_loss": 0.041054222732782364, "global_step": 194871, "epoch": 2189, "lr": 1.8447929591746256e-05}
{"train_loss": 0.06795593351125717, "global_step": 194872, "epoch": 2189, "lr": 1.844747986369084e-05}
{"train_loss": 0.09948650747537613, "global_step": 194873, "epoch": 2189, "lr": 1.8447030139877263e-05}
{"train_loss": 0.03746562451124191, "global_step": 194874, "epoch": 2189, "lr": 1.844658042030562e-05}
{"train_loss": 0.0991852879524231, "global_step": 194875, "epoch": 2189, "lr": 1.844613070497592e-05}
{"train_loss": 0.03640161082148552, "global_step": 194876, "epoch": 2189, "lr": 1.844568099388828e-05}
{"train_loss": 0.045498449355363846, "global_step": 194877, "epoch": 2189, "lr": 1.8445231287042714e-05}
{"train_loss": 0.06736080348491669, "global_step": 194878, "epoch": 2189, "lr": 1.8444781584439318e-05}
{"train_loss": 0.0516422763466835, "global_step": 194879, "epoch": 2189, "lr": 1.844433188607812e-05}
{"train_loss": 0.022945517674088478, "global_step": 194880, "epoch": 2189, "lr": 1.8443882191959217e-05}
{"train_loss": 0.03733687847852707, "global_step": 194881, "epoch": 2189, "lr": 1.844343250208263e-05}
{"train_loss": 0.05029810965061188, "global_step": 194882, "epoch": 2189, "lr": 1.8442982816448458e-05}
{"train_loss": 0.050511617213487625, "global_step": 194883, "epoch": 2189, "lr": 1.8442533135056715e-05}
{"train_loss": 0.047492992132902145, "global_step": 194884, "epoch": 2189, "lr": 1.8442083457907517e-05}
{"train_loss": 0.035101957619190216, "global_step": 194885, "epoch": 2189, "lr": 1.844163378500088e-05}
{"train_loss": 0.07332150638103485, "global_step": 194886, "epoch": 2189, "lr": 1.8441184116336874e-05}
{"train_loss": 0.09286317229270935, "global_step": 194887, "epoch": 2189, "lr": 1.8440734451915585e-05}
{"train_loss": 0.10379272699356079, "global_step": 194888, "epoch": 2189, "lr": 1.844028479173704e-05}
{"train_loss": 0.08123151957988739, "global_step": 194889, "epoch": 2189, "lr": 1.8439835135801325e-05}
{"train_loss": 0.06693590432405472, "global_step": 194890, "epoch": 2189, "lr": 1.8439385484108478e-05}
{"train_loss": 0.06595182418823242, "global_step": 194891, "epoch": 2189, "lr": 1.843893583665859e-05}
{"train_loss": 0.10180500149726868, "global_step": 194892, "epoch": 2189, "lr": 1.843848619345168e-05}
{"train_loss": 0.07154147326946259, "global_step": 194893, "epoch": 2189, "lr": 1.8438036554487853e-05}
{"train_loss": 0.04948185011744499, "global_step": 194894, "epoch": 2189, "lr": 1.8437586919767124e-05}
{"train_loss": 0.09361599385738373, "global_step": 194895, "epoch": 2189, "lr": 1.8437137289289596e-05}
{"train_loss": 0.05862019956111908, "global_step": 194896, "epoch": 2189, "lr": 1.8436687663055298e-05}
{"train_loss": 0.07497581839561462, "global_step": 194897, "epoch": 2189, "lr": 1.8436238041064312e-05}
{"train_loss": 0.047995343804359436, "global_step": 194898, "epoch": 2189, "lr": 1.8435788423316697e-05}
{"train_loss": 0.037031952291727066, "global_step": 194899, "epoch": 2189, "lr": 1.8435338809812487e-05}
{"train_loss": 0.03442232683300972, "global_step": 194900, "epoch": 2189, "lr": 1.843488920055177e-05}
{"train_loss": 0.03864873945713043, "global_step": 194901, "epoch": 2189, "lr": 1.8434439595534585e-05}
{"train_loss": 0.06639350950717926, "global_step": 194902, "epoch": 2189, "lr": 1.8433989994761026e-05}
{"train_loss": 0.092439204454422, "global_step": 194903, "epoch": 2189, "lr": 1.8433540398231115e-05}
{"train_loss": 0.12619352340698242, "global_step": 194904, "epoch": 2189, "lr": 1.843309080594494e-05}
{"train_loss": 0.03217872977256775, "global_step": 194905, "epoch": 2189, "lr": 1.8432641217902542e-05}
{"train_loss": 0.08866797387599945, "global_step": 194906, "epoch": 2189, "lr": 1.8432191634104e-05}
{"train_loss": 0.09114689379930496, "global_step": 194907, "epoch": 2189, "lr": 1.843174205454935e-05}
{"train_loss": 0.05610416457056999, "global_step": 194908, "epoch": 2189, "lr": 1.8431292479238687e-05}
{"train_loss": 0.06712916631544574, "global_step": 194909, "epoch": 2189, "lr": 1.8430842908172034e-05, "val_loss": 8.22193717956543}
{"train_loss": 0.05990719795227051, "global_step": 194910, "epoch": 2190, "lr": 1.843039334134949e-05}
{"train_loss": 0.0909433364868164, "global_step": 194911, "epoch": 2190, "lr": 1.8429943778771076e-05}
{"train_loss": 0.09080609679222107, "global_step": 194912, "epoch": 2190, "lr": 1.8429494220436883e-05}
{"train_loss": 0.10068752616643906, "global_step": 194913, "epoch": 2190, "lr": 1.8429044666346945e-05}
{"train_loss": 0.14560282230377197, "global_step": 194914, "epoch": 2190, "lr": 1.8428595116501357e-05}
{"train_loss": 0.052222780883312225, "global_step": 194915, "epoch": 2190, "lr": 1.842814557090014e-05}
{"train_loss": 0.0909859910607338, "global_step": 194916, "epoch": 2190, "lr": 1.8427696029543397e-05}
{"train_loss": 0.07611259818077087, "global_step": 194917, "epoch": 2190, "lr": 1.8427246492431143e-05}
{"train_loss": 0.06765688210725784, "global_step": 194918, "epoch": 2190, "lr": 1.8426796959563473e-05}
{"train_loss": 0.048569269478321075, "global_step": 194919, "epoch": 2190, "lr": 1.8426347430940428e-05}
{"train_loss": 0.07552696019411087, "global_step": 194920, "epoch": 2190, "lr": 1.842589790656209e-05}
{"train_loss": 0.07042013853788376, "global_step": 194921, "epoch": 2190, "lr": 1.8425448386428485e-05}
{"train_loss": 0.02871810831129551, "global_step": 194922, "epoch": 2190, "lr": 1.8424998870539696e-05}
{"train_loss": 0.10363442450761795, "global_step": 194923, "epoch": 2190, "lr": 1.84245493588958e-05}
{"train_loss": 0.07360778748989105, "global_step": 194924, "epoch": 2190, "lr": 1.8424099851496817e-05}
{"train_loss": 0.07162170112133026, "global_step": 194925, "epoch": 2190, "lr": 1.8423650348342846e-05}
{"train_loss": 0.059003353118896484, "global_step": 194926, "epoch": 2190, "lr": 1.842320084943392e-05}
{"train_loss": 0.11542351543903351, "global_step": 194927, "epoch": 2190, "lr": 1.842275135477012e-05}
{"train_loss": 0.055061887949705124, "global_step": 194928, "epoch": 2190, "lr": 1.8422301864351482e-05}
{"train_loss": 0.05229821801185608, "global_step": 194929, "epoch": 2190, "lr": 1.8421852378178096e-05}
{"train_loss": 0.09787986427545547, "global_step": 194930, "epoch": 2190, "lr": 1.8421402896249988e-05}
{"train_loss": 0.05215898156166077, "global_step": 194931, "epoch": 2190, "lr": 1.842095341856726e-05}
{"train_loss": 0.07458032667636871, "global_step": 194932, "epoch": 2190, "lr": 1.842050394512993e-05}
{"train_loss": 0.06506163626909256, "global_step": 194933, "epoch": 2190, "lr": 1.8420054475938097e-05}
{"train_loss": 0.06543585658073425, "global_step": 194934, "epoch": 2190, "lr": 1.841960501099178e-05}
{"train_loss": 0.0659475103020668, "global_step": 194935, "epoch": 2190, "lr": 1.841915555029109e-05}
{"train_loss": 0.0815972313284874, "global_step": 194936, "epoch": 2190, "lr": 1.841870609383605e-05}
{"train_loss": 0.027725759893655777, "global_step": 194937, "epoch": 2190, "lr": 1.8418256641626718e-05}
{"train_loss": 0.05676136538386345, "global_step": 194938, "epoch": 2190, "lr": 1.8417807193663185e-05}
{"train_loss": 0.010714642703533173, "global_step": 194939, "epoch": 2190, "lr": 1.841735774994547e-05}
{"train_loss": 0.06729695200920105, "global_step": 194940, "epoch": 2190, "lr": 1.8416908310473674e-05}
{"train_loss": 0.052132077515125275, "global_step": 194941, "epoch": 2190, "lr": 1.841645887524783e-05}
{"train_loss": 0.04825098067522049, "global_step": 194942, "epoch": 2190, "lr": 1.8416009444268018e-05}
{"train_loss": 0.052413634955883026, "global_step": 194943, "epoch": 2190, "lr": 1.841556001753427e-05}
{"train_loss": 0.07769453525543213, "global_step": 194944, "epoch": 2190, "lr": 1.8415110595046686e-05}
{"train_loss": 0.06701548397541046, "global_step": 194945, "epoch": 2190, "lr": 1.841466117680529e-05}
{"train_loss": 0.04954437538981438, "global_step": 194946, "epoch": 2190, "lr": 1.8414211762810168e-05}
{"train_loss": 0.045086752623319626, "global_step": 194947, "epoch": 2190, "lr": 1.841376235306136e-05}
{"train_loss": 0.06667955964803696, "global_step": 194948, "epoch": 2190, "lr": 1.841331294755895e-05}
{"train_loss": 0.04243260994553566, "global_step": 194949, "epoch": 2190, "lr": 1.8412863546302973e-05}
{"train_loss": 0.06853949278593063, "global_step": 194950, "epoch": 2190, "lr": 1.8412414149293516e-05}
{"train_loss": 0.0489475354552269, "global_step": 194951, "epoch": 2190, "lr": 1.8411964756530608e-05}
{"train_loss": 0.051133982837200165, "global_step": 194952, "epoch": 2190, "lr": 1.8411515368014344e-05}
{"train_loss": 0.07371458411216736, "global_step": 194953, "epoch": 2190, "lr": 1.8411065983744748e-05}
{"train_loss": 0.05974371358752251, "global_step": 194954, "epoch": 2190, "lr": 1.8410616603721913e-05}
{"train_loss": 0.07607944309711456, "global_step": 194955, "epoch": 2190, "lr": 1.8410167227945875e-05}
{"train_loss": 0.04809858277440071, "global_step": 194956, "epoch": 2190, "lr": 1.8409717856416714e-05}
{"train_loss": 0.07578083127737045, "global_step": 194957, "epoch": 2190, "lr": 1.8409268489134472e-05}
{"train_loss": 0.07010950893163681, "global_step": 194958, "epoch": 2190, "lr": 1.8408819126099212e-05}
{"train_loss": 0.11413395404815674, "global_step": 194959, "epoch": 2190, "lr": 1.840836976731103e-05}
{"train_loss": 0.0393826849758625, "global_step": 194960, "epoch": 2190, "lr": 1.840792041276993e-05}
{"train_loss": 0.028105448931455612, "global_step": 194961, "epoch": 2190, "lr": 1.840747106247602e-05}
{"train_loss": 0.04553423449397087, "global_step": 194962, "epoch": 2190, "lr": 1.8407021716429323e-05}
{"train_loss": 0.05559869855642319, "global_step": 194963, "epoch": 2190, "lr": 1.840657237462993e-05}
{"train_loss": 0.08825881034135818, "global_step": 194964, "epoch": 2190, "lr": 1.8406123037077884e-05}
{"train_loss": 0.06561652570962906, "global_step": 194965, "epoch": 2190, "lr": 1.8405673703773257e-05}
{"train_loss": 0.03643767535686493, "global_step": 194966, "epoch": 2190, "lr": 1.8405224374716083e-05}
{"train_loss": 0.05008680373430252, "global_step": 194967, "epoch": 2190, "lr": 1.8404775049906465e-05}
{"train_loss": 0.09858925640583038, "global_step": 194968, "epoch": 2190, "lr": 1.840432572934442e-05}
{"train_loss": 0.07561381161212921, "global_step": 194969, "epoch": 2190, "lr": 1.8403876413030043e-05}
{"train_loss": 0.09559807181358337, "global_step": 194970, "epoch": 2190, "lr": 1.8403427100963367e-05}
{"train_loss": 0.10678543895483017, "global_step": 194971, "epoch": 2190, "lr": 1.8402977793144478e-05}
{"train_loss": 0.11344735324382782, "global_step": 194972, "epoch": 2190, "lr": 1.8402528489573407e-05}
{"train_loss": 0.02541644498705864, "global_step": 194973, "epoch": 2190, "lr": 1.8402079190250248e-05}
{"train_loss": 0.03639611601829529, "global_step": 194974, "epoch": 2190, "lr": 1.8401629895175033e-05}
{"train_loss": 0.04191318526864052, "global_step": 194975, "epoch": 2190, "lr": 1.8401180604347845e-05}
{"train_loss": 0.09559450298547745, "global_step": 194976, "epoch": 2190, "lr": 1.8400731317768736e-05}
{"train_loss": 0.03972005471587181, "global_step": 194977, "epoch": 2190, "lr": 1.8400282035437742e-05}
{"train_loss": 0.030502378940582275, "global_step": 194978, "epoch": 2190, "lr": 1.8399832757354968e-05}
{"train_loss": 0.06389571726322174, "global_step": 194979, "epoch": 2190, "lr": 1.8399383483520432e-05}
{"train_loss": 0.05794965848326683, "global_step": 194980, "epoch": 2190, "lr": 1.8398934213934228e-05}
{"train_loss": 0.05536631867289543, "global_step": 194981, "epoch": 2190, "lr": 1.839848494859639e-05}
{"train_loss": 0.06824827939271927, "global_step": 194982, "epoch": 2190, "lr": 1.8398035687507005e-05}
{"train_loss": 0.098948635160923, "global_step": 194983, "epoch": 2190, "lr": 1.8397586430666102e-05}
{"train_loss": 0.06797032058238983, "global_step": 194984, "epoch": 2190, "lr": 1.8397137178073774e-05}
{"train_loss": 0.054494570940732956, "global_step": 194985, "epoch": 2190, "lr": 1.839668792973005e-05}
{"train_loss": 0.06709173321723938, "global_step": 194986, "epoch": 2190, "lr": 1.839623868563502e-05}
{"train_loss": 0.08237739652395248, "global_step": 194987, "epoch": 2190, "lr": 1.8395789445788718e-05}
{"train_loss": 0.06824339926242828, "global_step": 194988, "epoch": 2190, "lr": 1.8395340210191232e-05}
{"train_loss": 0.039985135197639465, "global_step": 194989, "epoch": 2190, "lr": 1.8394890978842595e-05}
{"train_loss": 0.048646118491888046, "global_step": 194990, "epoch": 2190, "lr": 1.8394441751742893e-05}
{"train_loss": 0.11845805495977402, "global_step": 194991, "epoch": 2190, "lr": 1.839399252889215e-05}
{"train_loss": 0.042724478989839554, "global_step": 194992, "epoch": 2190, "lr": 1.8393543310290473e-05}
{"train_loss": 0.07135225832462311, "global_step": 194993, "epoch": 2190, "lr": 1.839309409593788e-05}
{"train_loss": 0.09524232149124146, "global_step": 194994, "epoch": 2190, "lr": 1.8392644885834454e-05}
{"train_loss": 0.0857328400015831, "global_step": 194995, "epoch": 2190, "lr": 1.8392195679980262e-05}
{"train_loss": 0.02159333974123001, "global_step": 194996, "epoch": 2190, "lr": 1.8391746478375343e-05}
{"train_loss": 0.042434561997652054, "global_step": 194997, "epoch": 2190, "lr": 1.8391297281019782e-05}
{"train_loss": 0.06575156202058444, "global_step": 194998, "epoch": 2190, "lr": 1.839084808791361e-05, "val_loss": 8.11696720123291, "train_action_mse_error": 15.5457124710083}
{"train_loss": 0.060581520199775696, "global_step": 194999, "epoch": 2191, "lr": 1.8390398899056922e-05}
{"train_loss": 0.11247079074382782, "global_step": 195000, "epoch": 2191, "lr": 1.838994971444974e-05}
{"train_loss": 0.05060078203678131, "global_step": 195001, "epoch": 2191, "lr": 1.8389500534092163e-05}
{"train_loss": 0.04489828273653984, "global_step": 195002, "epoch": 2191, "lr": 1.8389051357984217e-05}
{"train_loss": 0.03696565702557564, "global_step": 195003, "epoch": 2191, "lr": 1.8388602186125987e-05}
{"train_loss": 0.07745542377233505, "global_step": 195004, "epoch": 2191, "lr": 1.8388153018517512e-05}
{"train_loss": 0.0807385966181755, "global_step": 195005, "epoch": 2191, "lr": 1.838770385515888e-05}
{"train_loss": 0.10078773647546768, "global_step": 195006, "epoch": 2191, "lr": 1.838725469605012e-05}
{"train_loss": 0.08043579757213593, "global_step": 195007, "epoch": 2191, "lr": 1.838680554119133e-05}
{"train_loss": 0.07791925221681595, "global_step": 195008, "epoch": 2191, "lr": 1.8386356390582525e-05}
{"train_loss": 0.038996621966362, "global_step": 195009, "epoch": 2191, "lr": 1.838590724422381e-05}
{"train_loss": 0.049328941851854324, "global_step": 195010, "epoch": 2191, "lr": 1.8385458102115206e-05}
{"train_loss": 0.06963028013706207, "global_step": 195011, "epoch": 2191, "lr": 1.838500896425681e-05}
{"train_loss": 0.08090943843126297, "global_step": 195012, "epoch": 2191, "lr": 1.8384559830648645e-05}
{"train_loss": 0.07621107250452042, "global_step": 195013, "epoch": 2191, "lr": 1.8384110701290808e-05}
{"train_loss": 0.042483795434236526, "global_step": 195014, "epoch": 2191, "lr": 1.838366157618335e-05}
{"train_loss": 0.07212181389331818, "global_step": 195015, "epoch": 2191, "lr": 1.8383212455326294e-05}
{"train_loss": 0.07135488837957382, "global_step": 195016, "epoch": 2191, "lr": 1.838276333871975e-05}
{"train_loss": 0.03920456022024155, "global_step": 195017, "epoch": 2191, "lr": 1.8382314226363745e-05}
{"train_loss": 0.09322676807641983, "global_step": 195018, "epoch": 2191, "lr": 1.8381865118258368e-05}
{"train_loss": 0.042496103793382645, "global_step": 195019, "epoch": 2191, "lr": 1.8381416014403647e-05}
{"train_loss": 0.03728385642170906, "global_step": 195020, "epoch": 2191, "lr": 1.8380966914799675e-05}
{"train_loss": 0.07318054884672165, "global_step": 195021, "epoch": 2191, "lr": 1.838051781944648e-05}
{"train_loss": 0.024260912090539932, "global_step": 195022, "epoch": 2191, "lr": 1.8380068728344156e-05}
{"train_loss": 0.0263977088034153, "global_step": 195023, "epoch": 2191, "lr": 1.837961964149273e-05}
{"train_loss": 0.04571513831615448, "global_step": 195024, "epoch": 2191, "lr": 1.8379170558892294e-05}
{"train_loss": 0.07558263093233109, "global_step": 195025, "epoch": 2191, "lr": 1.8378721480542877e-05}
{"train_loss": 0.07256260514259338, "global_step": 195026, "epoch": 2191, "lr": 1.8378272406444575e-05}
{"train_loss": 0.08657905459403992, "global_step": 195027, "epoch": 2191, "lr": 1.8377823336597403e-05}
{"train_loss": 0.12160255014896393, "global_step": 195028, "epoch": 2191, "lr": 1.8377374271001473e-05}
{"train_loss": 0.06275060027837753, "global_step": 195029, "epoch": 2191, "lr": 1.8376925209656804e-05}
{"train_loss": 0.04333185404539108, "global_step": 195030, "epoch": 2191, "lr": 1.8376476152563465e-05}
{"train_loss": 0.0717276930809021, "global_step": 195031, "epoch": 2191, "lr": 1.8376027099721543e-05}
{"train_loss": 0.09872565418481827, "global_step": 195032, "epoch": 2191, "lr": 1.8375578051131066e-05}
{"train_loss": 0.07529886811971664, "global_step": 195033, "epoch": 2191, "lr": 1.837512900679212e-05}
{"train_loss": 0.10986641049385071, "global_step": 195034, "epoch": 2191, "lr": 1.8374679966704735e-05}
{"train_loss": 0.06380651891231537, "global_step": 195035, "epoch": 2191, "lr": 1.837423093086901e-05}
{"train_loss": 0.050064537674188614, "global_step": 195036, "epoch": 2191, "lr": 1.8373781899284965e-05}
{"train_loss": 0.1164131909608841, "global_step": 195037, "epoch": 2191, "lr": 1.837333287195269e-05}
{"train_loss": 0.09228051453828812, "global_step": 195038, "epoch": 2191, "lr": 1.8372883848872223e-05}
{"train_loss": 0.05528128147125244, "global_step": 195039, "epoch": 2191, "lr": 1.8372434830043655e-05}
{"train_loss": 0.09080564975738525, "global_step": 195040, "epoch": 2191, "lr": 1.8371985815467013e-05}
{"train_loss": 0.10877189040184021, "global_step": 195041, "epoch": 2191, "lr": 1.8371536805142388e-05}
{"train_loss": 0.1319698691368103, "global_step": 195042, "epoch": 2191, "lr": 1.8371087799069804e-05}
{"train_loss": 0.07450959831476212, "global_step": 195043, "epoch": 2191, "lr": 1.8370638797249355e-05}
{"train_loss": 0.08121798932552338, "global_step": 195044, "epoch": 2191, "lr": 1.8370189799681077e-05}
{"train_loss": 0.03691576421260834, "global_step": 195045, "epoch": 2191, "lr": 1.836974080636506e-05}
{"train_loss": 0.062254808843135834, "global_step": 195046, "epoch": 2191, "lr": 1.836929181730132e-05}
{"train_loss": 0.04131614789366722, "global_step": 195047, "epoch": 2191, "lr": 1.836884283248997e-05}
{"train_loss": 0.12258565425872803, "global_step": 195048, "epoch": 2191, "lr": 1.836839385193102e-05}
{"train_loss": 0.06766141206026077, "global_step": 195049, "epoch": 2191, "lr": 1.8367944875624576e-05}
{"train_loss": 0.04294731095433235, "global_step": 195050, "epoch": 2191, "lr": 1.8367495903570658e-05}
{"train_loss": 0.05671664699912071, "global_step": 195051, "epoch": 2191, "lr": 1.836704693576936e-05}
{"train_loss": 0.08656289428472519, "global_step": 195052, "epoch": 2191, "lr": 1.836659797222071e-05}
{"train_loss": 0.07825244963169098, "global_step": 195053, "epoch": 2191, "lr": 1.8366149012924806e-05}
{"train_loss": 0.10001208633184433, "global_step": 195054, "epoch": 2191, "lr": 1.836570005788168e-05}
{"train_loss": 0.11575546115636826, "global_step": 195055, "epoch": 2191, "lr": 1.836525110709139e-05}
{"train_loss": 0.07634352147579193, "global_step": 195056, "epoch": 2191, "lr": 1.8364802160554017e-05}
{"train_loss": 0.04891277104616165, "global_step": 195057, "epoch": 2191, "lr": 1.8364353218269604e-05}
{"train_loss": 0.06250663101673126, "global_step": 195058, "epoch": 2191, "lr": 1.8363904280238225e-05}
{"train_loss": 0.07251322269439697, "global_step": 195059, "epoch": 2191, "lr": 1.8363455346459923e-05}
{"train_loss": 0.05397436022758484, "global_step": 195060, "epoch": 2191, "lr": 1.836300641693478e-05}
{"train_loss": 0.08707136660814285, "global_step": 195061, "epoch": 2191, "lr": 1.8362557491662834e-05}
{"train_loss": 0.06901531666517258, "global_step": 195062, "epoch": 2191, "lr": 1.8362108570644172e-05}
{"train_loss": 0.05576912313699722, "global_step": 195063, "epoch": 2191, "lr": 1.8361659653878822e-05}
{"train_loss": 0.0807049423456192, "global_step": 195064, "epoch": 2191, "lr": 1.8361210741366875e-05}
{"train_loss": 0.06725382059812546, "global_step": 195065, "epoch": 2191, "lr": 1.8360761833108365e-05}
{"train_loss": 0.05051617696881294, "global_step": 195066, "epoch": 2191, "lr": 1.836031292910338e-05}
{"train_loss": 0.08497042953968048, "global_step": 195067, "epoch": 2191, "lr": 1.8359864029351946e-05}
{"train_loss": 0.055179186165332794, "global_step": 195068, "epoch": 2191, "lr": 1.835941513385414e-05}
{"train_loss": 0.056320227682590485, "global_step": 195069, "epoch": 2191, "lr": 1.835896624261005e-05}
{"train_loss": 0.04738307371735573, "global_step": 195070, "epoch": 2191, "lr": 1.8358517355619692e-05}
{"train_loss": 0.052884042263031006, "global_step": 195071, "epoch": 2191, "lr": 1.8358068472883162e-05}
{"train_loss": 0.08931073546409607, "global_step": 195072, "epoch": 2191, "lr": 1.8357619594400482e-05}
{"train_loss": 0.09146658331155777, "global_step": 195073, "epoch": 2191, "lr": 1.8357170720171758e-05}
{"train_loss": 0.046466510742902756, "global_step": 195074, "epoch": 2191, "lr": 1.8356721850197012e-05}
{"train_loss": 0.04898020997643471, "global_step": 195075, "epoch": 2191, "lr": 1.835627298447633e-05}
{"train_loss": 0.08398116379976273, "global_step": 195076, "epoch": 2191, "lr": 1.8355824123009747e-05}
{"train_loss": 0.06488408893346786, "global_step": 195077, "epoch": 2191, "lr": 1.8355375265797352e-05}
{"train_loss": 0.0666000097990036, "global_step": 195078, "epoch": 2191, "lr": 1.8354926412839173e-05}
{"train_loss": 0.10632622987031937, "global_step": 195079, "epoch": 2191, "lr": 1.8354477564135312e-05}
{"train_loss": 0.08373790979385376, "global_step": 195080, "epoch": 2191, "lr": 1.8354028719685783e-05}
{"train_loss": 0.06677302718162537, "global_step": 195081, "epoch": 2191, "lr": 1.835357987949069e-05}
{"train_loss": 0.0924583300948143, "global_step": 195082, "epoch": 2191, "lr": 1.8353131043550054e-05}
{"train_loss": 0.033554237335920334, "global_step": 195083, "epoch": 2191, "lr": 1.835268221186397e-05}
{"train_loss": 0.09326113760471344, "global_step": 195084, "epoch": 2191, "lr": 1.8352233384432466e-05}
{"train_loss": 0.0754857286810875, "global_step": 195085, "epoch": 2191, "lr": 1.835178456125562e-05}
{"train_loss": 0.06518512964248657, "global_step": 195086, "epoch": 2191, "lr": 1.835133574233351e-05}
{"train_loss": 0.07082193273674237, "global_step": 195087, "epoch": 2191, "lr": 1.8350886927666173e-05, "val_loss": 7.945802211761475}
{"train_loss": 0.03803831338882446, "global_step": 195088, "epoch": 2192, "lr": 1.8350438117253653e-05}
{"train_loss": 0.0703047439455986, "global_step": 195089, "epoch": 2192, "lr": 1.834998931109605e-05}
{"train_loss": 0.06155621260404587, "global_step": 195090, "epoch": 2192, "lr": 1.8349540509193393e-05}
{"train_loss": 0.07817906141281128, "global_step": 195091, "epoch": 2192, "lr": 1.8349091711545767e-05}
{"train_loss": 0.03428880125284195, "global_step": 195092, "epoch": 2192, "lr": 1.834864291815322e-05}
{"train_loss": 0.12277194112539291, "global_step": 195093, "epoch": 2192, "lr": 1.83481941290158e-05}
{"train_loss": 0.08708884567022324, "global_step": 195094, "epoch": 2192, "lr": 1.8347745344133593e-05}
{"train_loss": 0.017967384308576584, "global_step": 195095, "epoch": 2192, "lr": 1.8347296563506627e-05}
{"train_loss": 0.05394558236002922, "global_step": 195096, "epoch": 2192, "lr": 1.8346847787135003e-05}
{"train_loss": 0.11488272249698639, "global_step": 195097, "epoch": 2192, "lr": 1.8346399015018733e-05}
{"train_loss": 0.04388294741511345, "global_step": 195098, "epoch": 2192, "lr": 1.834595024715793e-05}
{"train_loss": 0.05837564170360565, "global_step": 195099, "epoch": 2192, "lr": 1.834550148355261e-05}
{"train_loss": 0.10633675009012222, "global_step": 195100, "epoch": 2192, "lr": 1.8345052724202867e-05}
{"train_loss": 0.03492351248860359, "global_step": 195101, "epoch": 2192, "lr": 1.8344603969108727e-05}
{"train_loss": 0.07486963272094727, "global_step": 195102, "epoch": 2192, "lr": 1.8344155218270287e-05}
{"train_loss": 0.07821155339479446, "global_step": 195103, "epoch": 2192, "lr": 1.8343706471687578e-05}
{"train_loss": 0.07480322569608688, "global_step": 195104, "epoch": 2192, "lr": 1.8343257729360664e-05}
{"train_loss": 0.05382296442985535, "global_step": 195105, "epoch": 2192, "lr": 1.8342808991289633e-05}
{"train_loss": 0.03843372315168381, "global_step": 195106, "epoch": 2192, "lr": 1.8342360257474512e-05}
{"train_loss": 0.06379316747188568, "global_step": 195107, "epoch": 2192, "lr": 1.834191152791539e-05}
{"train_loss": 0.06332062929868698, "global_step": 195108, "epoch": 2192, "lr": 1.8341462802612293e-05}
{"train_loss": 0.03948516771197319, "global_step": 195109, "epoch": 2192, "lr": 1.8341014081565317e-05}
{"train_loss": 0.0441465750336647, "global_step": 195110, "epoch": 2192, "lr": 1.834056536477449e-05}
{"train_loss": 0.10006646811962128, "global_step": 195111, "epoch": 2192, "lr": 1.8340116652239907e-05}
{"train_loss": 0.11190230399370193, "global_step": 195112, "epoch": 2192, "lr": 1.833966794396159e-05}
{"train_loss": 0.06160465255379677, "global_step": 195113, "epoch": 2192, "lr": 1.8339219239939637e-05}
{"train_loss": 0.06533029675483704, "global_step": 195114, "epoch": 2192, "lr": 1.8338770540174072e-05}
{"train_loss": 0.10000846534967422, "global_step": 195115, "epoch": 2192, "lr": 1.833832184466499e-05}
{"train_loss": 0.09958431124687195, "global_step": 195116, "epoch": 2192, "lr": 1.833787315341242e-05}
{"train_loss": 0.06413611024618149, "global_step": 195117, "epoch": 2192, "lr": 1.8337424466416453e-05}
{"train_loss": 0.06265496462583542, "global_step": 195118, "epoch": 2192, "lr": 1.8336975783677118e-05}
{"train_loss": 0.06885853409767151, "global_step": 195119, "epoch": 2192, "lr": 1.8336527105194503e-05}
{"train_loss": 0.11074899882078171, "global_step": 195120, "epoch": 2192, "lr": 1.8336078430968644e-05}
{"train_loss": 0.06878543645143509, "global_step": 195121, "epoch": 2192, "lr": 1.833562976099961e-05}
{"train_loss": 0.05810915306210518, "global_step": 195122, "epoch": 2192, "lr": 1.8335181095287485e-05}
{"train_loss": 0.0894472599029541, "global_step": 195123, "epoch": 2192, "lr": 1.8334732433832285e-05}
{"train_loss": 0.07550916820764542, "global_step": 195124, "epoch": 2192, "lr": 1.833428377663412e-05}
{"train_loss": 0.08908193558454514, "global_step": 195125, "epoch": 2192, "lr": 1.833383512369302e-05}
{"train_loss": 0.05476618558168411, "global_step": 195126, "epoch": 2192, "lr": 1.8333386475009034e-05}
{"train_loss": 0.04816654697060585, "global_step": 195127, "epoch": 2192, "lr": 1.8332937830582253e-05}
{"train_loss": 0.07260773330926895, "global_step": 195128, "epoch": 2192, "lr": 1.833248919041271e-05}
{"train_loss": 0.05348443239927292, "global_step": 195129, "epoch": 2192, "lr": 1.833204055450049e-05}
{"train_loss": 0.056998539716005325, "global_step": 195130, "epoch": 2192, "lr": 1.8331591922845643e-05}
{"train_loss": 0.08955582976341248, "global_step": 195131, "epoch": 2192, "lr": 1.833114329544821e-05}
{"train_loss": 0.0912003368139267, "global_step": 195132, "epoch": 2192, "lr": 1.833069467230829e-05}
{"train_loss": 0.08232268691062927, "global_step": 195133, "epoch": 2192, "lr": 1.8330246053425897e-05}
{"train_loss": 0.07066028565168381, "global_step": 195134, "epoch": 2192, "lr": 1.832979743880114e-05}
{"train_loss": 0.0472370870411396, "global_step": 195135, "epoch": 2192, "lr": 1.832934882843404e-05}
{"train_loss": 0.06451799720525742, "global_step": 195136, "epoch": 2192, "lr": 1.8328900222324685e-05}
{"train_loss": 0.06860383599996567, "global_step": 195137, "epoch": 2192, "lr": 1.832845162047311e-05}
{"train_loss": 0.08058615773916245, "global_step": 195138, "epoch": 2192, "lr": 1.8328003022879403e-05}
{"train_loss": 0.0853552371263504, "global_step": 195139, "epoch": 2192, "lr": 1.8327554429543592e-05}
{"train_loss": 0.07384906709194183, "global_step": 195140, "epoch": 2192, "lr": 1.8327105840465758e-05}
{"train_loss": 0.038854122161865234, "global_step": 195141, "epoch": 2192, "lr": 1.8326657255645973e-05}
{"train_loss": 0.04590390995144844, "global_step": 195142, "epoch": 2192, "lr": 1.8326208675084272e-05}
{"train_loss": 0.0549316480755806, "global_step": 195143, "epoch": 2192, "lr": 1.8325760098780736e-05}
{"train_loss": 0.054960720241069794, "global_step": 195144, "epoch": 2192, "lr": 1.83253115267354e-05}
{"train_loss": 0.07243737578392029, "global_step": 195145, "epoch": 2192, "lr": 1.832486295894836e-05}
{"train_loss": 0.05035875737667084, "global_step": 195146, "epoch": 2192, "lr": 1.8324414395419637e-05}
{"train_loss": 0.03517520800232887, "global_step": 195147, "epoch": 2192, "lr": 1.8323965836149326e-05}
{"train_loss": 0.08121076226234436, "global_step": 195148, "epoch": 2192, "lr": 1.832351728113746e-05}
{"train_loss": 0.12288381159305573, "global_step": 195149, "epoch": 2192, "lr": 1.8323068730384123e-05}
{"train_loss": 0.10008031129837036, "global_step": 195150, "epoch": 2192, "lr": 1.832262018388935e-05}
{"train_loss": 0.03370371088385582, "global_step": 195151, "epoch": 2192, "lr": 1.832217164165323e-05}
{"train_loss": 0.059558384120464325, "global_step": 195152, "epoch": 2192, "lr": 1.8321723103675787e-05}
{"train_loss": 0.057267606258392334, "global_step": 195153, "epoch": 2192, "lr": 1.8321274569957126e-05}
{"train_loss": 0.045857541263103485, "global_step": 195154, "epoch": 2192, "lr": 1.832082604049726e-05}
{"train_loss": 0.04363630339503288, "global_step": 195155, "epoch": 2192, "lr": 1.8320377515296294e-05}
{"train_loss": 0.049495309591293335, "global_step": 195156, "epoch": 2192, "lr": 1.8319928994354247e-05}
{"train_loss": 0.0619550459086895, "global_step": 195157, "epoch": 2192, "lr": 1.83194804776712e-05}
{"train_loss": 0.03730940818786621, "global_step": 195158, "epoch": 2192, "lr": 1.8319031965247234e-05}
{"train_loss": 0.06848955899477005, "global_step": 195159, "epoch": 2192, "lr": 1.8318583457082368e-05}
{"train_loss": 0.028601109981536865, "global_step": 195160, "epoch": 2192, "lr": 1.8318134953176695e-05}
{"train_loss": 0.0717742070555687, "global_step": 195161, "epoch": 2192, "lr": 1.8317686453530253e-05}
{"train_loss": 0.07072862982749939, "global_step": 195162, "epoch": 2192, "lr": 1.8317237958143124e-05}
{"train_loss": 0.053081922233104706, "global_step": 195163, "epoch": 2192, "lr": 1.831678946701536e-05}
{"train_loss": 0.039564795792102814, "global_step": 195164, "epoch": 2192, "lr": 1.8316340980147e-05}
{"train_loss": 0.030179429799318314, "global_step": 195165, "epoch": 2192, "lr": 1.831589249753814e-05}
{"train_loss": 0.08674952387809753, "global_step": 195166, "epoch": 2192, "lr": 1.8315444019188797e-05}
{"train_loss": 0.08372882008552551, "global_step": 195167, "epoch": 2192, "lr": 1.8314995545099084e-05}
{"train_loss": 0.08302357792854309, "global_step": 195168, "epoch": 2192, "lr": 1.8314547075269012e-05}
{"train_loss": 0.05474499240517616, "global_step": 195169, "epoch": 2192, "lr": 1.8314098609698675e-05}
{"train_loss": 0.044381700456142426, "global_step": 195170, "epoch": 2192, "lr": 1.8313650148388128e-05}
{"train_loss": 0.08292635530233383, "global_step": 195171, "epoch": 2192, "lr": 1.8313201691337408e-05}
{"train_loss": 0.0782974436879158, "global_step": 195172, "epoch": 2192, "lr": 1.8312753238546602e-05}
{"train_loss": 0.10538921505212784, "global_step": 195173, "epoch": 2192, "lr": 1.831230479001575e-05}
{"train_loss": 0.0974215716123581, "global_step": 195174, "epoch": 2192, "lr": 1.8311856345744934e-05}
{"train_loss": 0.03915727883577347, "global_step": 195175, "epoch": 2192, "lr": 1.8311407905734192e-05}
{"train_loss": 0.0672167054471675, "global_step": 195176, "epoch": 2192, "lr": 1.8310959469983592e-05, "val_loss": 8.011649131774902}
{"train_loss": 0.061609696596860886, "global_step": 195177, "epoch": 2193, "lr": 1.831051103849321e-05}
{"train_loss": 0.09348348528146744, "global_step": 195178, "epoch": 2193, "lr": 1.831006261126308e-05}
{"train_loss": 0.06539542227983475, "global_step": 195179, "epoch": 2193, "lr": 1.8309614188293296e-05}
{"train_loss": 0.08767876774072647, "global_step": 195180, "epoch": 2193, "lr": 1.8309165769583875e-05}
{"train_loss": 0.0861787274479866, "global_step": 195181, "epoch": 2193, "lr": 1.830871735513492e-05}
{"train_loss": 0.038334812968969345, "global_step": 195182, "epoch": 2193, "lr": 1.8308268944946454e-05}
{"train_loss": 0.03470386937260628, "global_step": 195183, "epoch": 2193, "lr": 1.8307820539018577e-05}
{"train_loss": 0.0650225430727005, "global_step": 195184, "epoch": 2193, "lr": 1.8307372137351302e-05}
{"train_loss": 0.07576941698789597, "global_step": 195185, "epoch": 2193, "lr": 1.8306923739944735e-05}
{"train_loss": 0.07281018793582916, "global_step": 195186, "epoch": 2193, "lr": 1.83064753467989e-05}
{"train_loss": 0.03927219286561012, "global_step": 195187, "epoch": 2193, "lr": 1.8306026957913892e-05}
{"train_loss": 0.07350163906812668, "global_step": 195188, "epoch": 2193, "lr": 1.8305578573289733e-05}
{"train_loss": 0.050622470676898956, "global_step": 195189, "epoch": 2193, "lr": 1.8305130192926523e-05}
{"train_loss": 0.040008459240198135, "global_step": 195190, "epoch": 2193, "lr": 1.8304681816824282e-05}
{"train_loss": 0.049686603248119354, "global_step": 195191, "epoch": 2193, "lr": 1.8304233444983104e-05}
{"train_loss": 0.09910396486520767, "global_step": 195192, "epoch": 2193, "lr": 1.8303785077403024e-05}
{"train_loss": 0.062334492802619934, "global_step": 195193, "epoch": 2193, "lr": 1.8303336714084123e-05}
{"train_loss": 0.1025550365447998, "global_step": 195194, "epoch": 2193, "lr": 1.8302888355026444e-05}
{"train_loss": 0.04868806526064873, "global_step": 195195, "epoch": 2193, "lr": 1.8302440000230053e-05}
{"train_loss": 0.047615326941013336, "global_step": 195196, "epoch": 2193, "lr": 1.8301991649695027e-05}
{"train_loss": 0.1120990514755249, "global_step": 195197, "epoch": 2193, "lr": 1.8301543303421396e-05}
{"train_loss": 0.03992239013314247, "global_step": 195198, "epoch": 2193, "lr": 1.8301094961409253e-05}
{"train_loss": 0.08128084242343903, "global_step": 195199, "epoch": 2193, "lr": 1.8300646623658624e-05}
{"train_loss": 0.06967449933290482, "global_step": 195200, "epoch": 2193, "lr": 1.8300198290169607e-05}
{"train_loss": 0.11321157217025757, "global_step": 195201, "epoch": 2193, "lr": 1.829974996094222e-05}
{"train_loss": 0.051929302513599396, "global_step": 195202, "epoch": 2193, "lr": 1.8299301635976568e-05}
{"train_loss": 0.09454412013292313, "global_step": 195203, "epoch": 2193, "lr": 1.8298853315272683e-05}
{"train_loss": 0.08825614303350449, "global_step": 195204, "epoch": 2193, "lr": 1.8298404998830615e-05}
{"train_loss": 0.09629376977682114, "global_step": 195205, "epoch": 2193, "lr": 1.8297956686650457e-05}
{"train_loss": 0.0629795640707016, "global_step": 195206, "epoch": 2193, "lr": 1.8297508378732237e-05}
{"train_loss": 0.06051509827375412, "global_step": 195207, "epoch": 2193, "lr": 1.829706007507605e-05}
{"train_loss": 0.04037943109869957, "global_step": 195208, "epoch": 2193, "lr": 1.829661177568193e-05}
{"train_loss": 0.07150699943304062, "global_step": 195209, "epoch": 2193, "lr": 1.8296163480549934e-05}
{"train_loss": 0.06916353851556778, "global_step": 195210, "epoch": 2193, "lr": 1.8295715189680146e-05}
{"train_loss": 0.07504608482122421, "global_step": 195211, "epoch": 2193, "lr": 1.8295266903072594e-05}
{"train_loss": 0.07198209315538406, "global_step": 195212, "epoch": 2193, "lr": 1.8294818620727362e-05}
{"train_loss": 0.04950113222002983, "global_step": 195213, "epoch": 2193, "lr": 1.829437034264452e-05}
{"train_loss": 0.06383508443832397, "global_step": 195214, "epoch": 2193, "lr": 1.8293922068824095e-05}
{"train_loss": 0.028837181627750397, "global_step": 195215, "epoch": 2193, "lr": 1.829347379926618e-05}
{"train_loss": 0.06443844735622406, "global_step": 195216, "epoch": 2193, "lr": 1.8293025533970808e-05}
{"train_loss": 0.03041447512805462, "global_step": 195217, "epoch": 2193, "lr": 1.8292577272938073e-05}
{"train_loss": 0.10949289798736572, "global_step": 195218, "epoch": 2193, "lr": 1.8292129016167992e-05}
{"train_loss": 0.10450741648674011, "global_step": 195219, "epoch": 2193, "lr": 1.829168076366066e-05}
{"train_loss": 0.06688421219587326, "global_step": 195220, "epoch": 2193, "lr": 1.829123251541612e-05}
{"train_loss": 0.0726279616355896, "global_step": 195221, "epoch": 2193, "lr": 1.8290784271434446e-05}
{"train_loss": 0.10451510548591614, "global_step": 195222, "epoch": 2193, "lr": 1.8290336031715677e-05}
{"train_loss": 0.10365545004606247, "global_step": 195223, "epoch": 2193, "lr": 1.8289887796259895e-05}
{"train_loss": 0.05921429395675659, "global_step": 195224, "epoch": 2193, "lr": 1.8289439565067146e-05}
{"train_loss": 0.08098364621400833, "global_step": 195225, "epoch": 2193, "lr": 1.8288991338137507e-05}
{"train_loss": 0.130718395113945, "global_step": 195226, "epoch": 2193, "lr": 1.8288543115471008e-05}
{"train_loss": 0.028927695006132126, "global_step": 195227, "epoch": 2193, "lr": 1.828809489706774e-05}
{"train_loss": 0.06408783048391342, "global_step": 195228, "epoch": 2193, "lr": 1.8287646682927744e-05}
{"train_loss": 0.09133044630289078, "global_step": 195229, "epoch": 2193, "lr": 1.8287198473051097e-05}
{"train_loss": 0.12081240862607956, "global_step": 195230, "epoch": 2193, "lr": 1.8286750267437835e-05}
{"train_loss": 0.09507743269205093, "global_step": 195231, "epoch": 2193, "lr": 1.8286302066088035e-05}
{"train_loss": 0.07478810101747513, "global_step": 195232, "epoch": 2193, "lr": 1.8285853869001772e-05}
{"train_loss": 0.12872843444347382, "global_step": 195233, "epoch": 2193, "lr": 1.8285405676179073e-05}
{"train_loss": 0.06895142793655396, "global_step": 195234, "epoch": 2193, "lr": 1.828495748762003e-05}
{"train_loss": 0.07652084529399872, "global_step": 195235, "epoch": 2193, "lr": 1.828450930332467e-05}
{"train_loss": 0.13029952347278595, "global_step": 195236, "epoch": 2193, "lr": 1.828406112329309e-05}
{"train_loss": 0.07800675928592682, "global_step": 195237, "epoch": 2193, "lr": 1.8283612947525314e-05}
{"train_loss": 0.041556354612112045, "global_step": 195238, "epoch": 2193, "lr": 1.828316477602143e-05}
{"train_loss": 0.05650598928332329, "global_step": 195239, "epoch": 2193, "lr": 1.828271660878148e-05}
{"train_loss": 0.02847996912896633, "global_step": 195240, "epoch": 2193, "lr": 1.8282268445805544e-05}
{"train_loss": 0.07461711764335632, "global_step": 195241, "epoch": 2193, "lr": 1.8281820287093676e-05}
{"train_loss": 0.05571554973721504, "global_step": 195242, "epoch": 2193, "lr": 1.8281372132645913e-05}
{"train_loss": 0.08393268287181854, "global_step": 195243, "epoch": 2193, "lr": 1.8280923982462344e-05}
{"train_loss": 0.07437969744205475, "global_step": 195244, "epoch": 2193, "lr": 1.8280475836543004e-05}
{"train_loss": 0.11000309139490128, "global_step": 195245, "epoch": 2193, "lr": 1.8280027694887986e-05}
{"train_loss": 0.07253226637840271, "global_step": 195246, "epoch": 2193, "lr": 1.827957955749733e-05}
{"train_loss": 0.047718338668346405, "global_step": 195247, "epoch": 2193, "lr": 1.8279131424371083e-05}
{"train_loss": 0.046376533806324005, "global_step": 195248, "epoch": 2193, "lr": 1.827868329550932e-05}
{"train_loss": 0.035753004252910614, "global_step": 195249, "epoch": 2193, "lr": 1.8278235170912116e-05}
{"train_loss": 0.08702349662780762, "global_step": 195250, "epoch": 2193, "lr": 1.82777870505795e-05}
{"train_loss": 0.028648100793361664, "global_step": 195251, "epoch": 2193, "lr": 1.827733893451157e-05}
{"train_loss": 0.06980259716510773, "global_step": 195252, "epoch": 2193, "lr": 1.8276890822708343e-05}
{"train_loss": 0.053253546357154846, "global_step": 195253, "epoch": 2193, "lr": 1.8276442715169916e-05}
{"train_loss": 0.06207656115293503, "global_step": 195254, "epoch": 2193, "lr": 1.8275994611896323e-05}
{"train_loss": 0.059478700160980225, "global_step": 195255, "epoch": 2193, "lr": 1.8275546512887652e-05}
{"train_loss": 0.04124800115823746, "global_step": 195256, "epoch": 2193, "lr": 1.8275098418143925e-05}
{"train_loss": 0.06814490258693695, "global_step": 195257, "epoch": 2193, "lr": 1.827465032766525e-05}
{"train_loss": 0.046152081340551376, "global_step": 195258, "epoch": 2193, "lr": 1.8274202241451638e-05}
{"train_loss": 0.034355685114860535, "global_step": 195259, "epoch": 2193, "lr": 1.8273754159503194e-05}
{"train_loss": 0.07269500941038132, "global_step": 195260, "epoch": 2193, "lr": 1.827330608181994e-05}
{"train_loss": 0.10911950469017029, "global_step": 195261, "epoch": 2193, "lr": 1.8272858008401965e-05}
{"train_loss": 0.10463081300258636, "global_step": 195262, "epoch": 2193, "lr": 1.827240993924931e-05}
{"train_loss": 0.08518309146165848, "global_step": 195263, "epoch": 2193, "lr": 1.827196187436205e-05}
{"train_loss": 0.09507247805595398, "global_step": 195264, "epoch": 2193, "lr": 1.8271513813740228e-05}
{"train_loss": 0.07100032123454501, "global_step": 195265, "epoch": 2193, "lr": 1.8271065757383925e-05, "val_loss": 8.02717113494873}
{"train_loss": 0.05618777871131897, "global_step": 195266, "epoch": 2194, "lr": 1.8270617705293182e-05}
{"train_loss": 0.013633674941956997, "global_step": 195267, "epoch": 2194, "lr": 1.8270169657468056e-05}
{"train_loss": 0.058898985385894775, "global_step": 195268, "epoch": 2194, "lr": 1.8269721613908646e-05}
{"train_loss": 0.06971653550863266, "global_step": 195269, "epoch": 2194, "lr": 1.8269273574614965e-05}
{"train_loss": 0.1285228282213211, "global_step": 195270, "epoch": 2194, "lr": 1.8268825539587108e-05}
{"train_loss": 0.04996681213378906, "global_step": 195271, "epoch": 2194, "lr": 1.826837750882511e-05}
{"train_loss": 0.06227024644613266, "global_step": 195272, "epoch": 2194, "lr": 1.826792948232905e-05}
{"train_loss": 0.07203914225101471, "global_step": 195273, "epoch": 2194, "lr": 1.8267481460098968e-05}
{"train_loss": 0.06141829490661621, "global_step": 195274, "epoch": 2194, "lr": 1.8267033442134952e-05}
{"train_loss": 0.05528238043189049, "global_step": 195275, "epoch": 2194, "lr": 1.826658542843703e-05}
{"train_loss": 0.06585417687892914, "global_step": 195276, "epoch": 2194, "lr": 1.8266137419005297e-05}
{"train_loss": 0.06248731538653374, "global_step": 195277, "epoch": 2194, "lr": 1.8265689413839783e-05}
{"train_loss": 0.06374713033437729, "global_step": 195278, "epoch": 2194, "lr": 1.826524141294057e-05}
{"train_loss": 0.08806747198104858, "global_step": 195279, "epoch": 2194, "lr": 1.8264793416307707e-05}
{"train_loss": 0.04843258857727051, "global_step": 195280, "epoch": 2194, "lr": 1.8264345423941244e-05}
{"train_loss": 0.06956145912408829, "global_step": 195281, "epoch": 2194, "lr": 1.8263897435841265e-05}
{"train_loss": 0.053897980600595474, "global_step": 195282, "epoch": 2194, "lr": 1.826344945200781e-05}
{"train_loss": 0.029482921585440636, "global_step": 195283, "epoch": 2194, "lr": 1.8263001472440955e-05}
{"train_loss": 0.0973559319972992, "global_step": 195284, "epoch": 2194, "lr": 1.826255349714074e-05}
{"train_loss": 0.04161926731467247, "global_step": 195285, "epoch": 2194, "lr": 1.8262105526107254e-05}
{"train_loss": 0.06452246755361557, "global_step": 195286, "epoch": 2194, "lr": 1.826165755934052e-05}
{"train_loss": 0.06248970329761505, "global_step": 195287, "epoch": 2194, "lr": 1.8261209596840647e-05}
{"train_loss": 0.040408797562122345, "global_step": 195288, "epoch": 2194, "lr": 1.8260761638607642e-05}
{"train_loss": 0.0641150176525116, "global_step": 195289, "epoch": 2194, "lr": 1.826031368464161e-05}
{"train_loss": 0.051577433943748474, "global_step": 195290, "epoch": 2194, "lr": 1.8259865734942572e-05}
{"train_loss": 0.09127993136644363, "global_step": 195291, "epoch": 2194, "lr": 1.825941778951063e-05}
{"train_loss": 0.08658002316951752, "global_step": 195292, "epoch": 2194, "lr": 1.8258969848345802e-05}
{"train_loss": 0.04382176324725151, "global_step": 195293, "epoch": 2194, "lr": 1.825852191144819e-05}
{"train_loss": 0.07746530324220657, "global_step": 195294, "epoch": 2194, "lr": 1.825807397881781e-05}
{"train_loss": 0.018883736804127693, "global_step": 195295, "epoch": 2194, "lr": 1.825762605045476e-05}
{"train_loss": 0.07621067017316818, "global_step": 195296, "epoch": 2194, "lr": 1.8257178126359076e-05}
{"train_loss": 0.09765112400054932, "global_step": 195297, "epoch": 2194, "lr": 1.8256730206530844e-05}
{"train_loss": 0.04948126897215843, "global_step": 195298, "epoch": 2194, "lr": 1.825628229097009e-05}
{"train_loss": 0.143548846244812, "global_step": 195299, "epoch": 2194, "lr": 1.8255834379676906e-05}
{"train_loss": 0.12049545347690582, "global_step": 195300, "epoch": 2194, "lr": 1.825538647265132e-05}
{"train_loss": 0.06334326416254044, "global_step": 195301, "epoch": 2194, "lr": 1.825493856989343e-05}
{"train_loss": 0.10421015322208405, "global_step": 195302, "epoch": 2194, "lr": 1.8254490671403263e-05}
{"train_loss": 0.08794094622135162, "global_step": 195303, "epoch": 2194, "lr": 1.8254042777180884e-05}
{"train_loss": 0.019024558365345, "global_step": 195304, "epoch": 2194, "lr": 1.825359488722639e-05}
{"train_loss": 0.08437535166740417, "global_step": 195305, "epoch": 2194, "lr": 1.825314700153979e-05}
{"train_loss": 0.025081826373934746, "global_step": 195306, "epoch": 2194, "lr": 1.8252699120121187e-05}
{"train_loss": 0.09895806014537811, "global_step": 195307, "epoch": 2194, "lr": 1.82522512429706e-05}
{"train_loss": 0.053191352635622025, "global_step": 195308, "epoch": 2194, "lr": 1.8251803370088127e-05}
{"train_loss": 0.04478086531162262, "global_step": 195309, "epoch": 2194, "lr": 1.8251355501473798e-05}
{"train_loss": 0.08468690514564514, "global_step": 195310, "epoch": 2194, "lr": 1.8250907637127708e-05}
{"train_loss": 0.0642465278506279, "global_step": 195311, "epoch": 2194, "lr": 1.8250459777049878e-05}
{"train_loss": 0.10353466123342514, "global_step": 195312, "epoch": 2194, "lr": 1.8250011921240402e-05}
{"train_loss": 0.057491447776556015, "global_step": 195313, "epoch": 2194, "lr": 1.8249564069699306e-05}
{"train_loss": 0.044145919382572174, "global_step": 195314, "epoch": 2194, "lr": 1.8249116222426686e-05}
{"train_loss": 0.027924910187721252, "global_step": 195315, "epoch": 2194, "lr": 1.8248668379422574e-05}
{"train_loss": 0.09201114624738693, "global_step": 195316, "epoch": 2194, "lr": 1.8248220540687057e-05}
{"train_loss": 0.08968871086835861, "global_step": 195317, "epoch": 2194, "lr": 1.8247772706220156e-05}
{"train_loss": 0.03742065280675888, "global_step": 195318, "epoch": 2194, "lr": 1.8247324876021975e-05}
{"train_loss": 0.03699861094355583, "global_step": 195319, "epoch": 2194, "lr": 1.824687705009256e-05}
{"train_loss": 0.10618838667869568, "global_step": 195320, "epoch": 2194, "lr": 1.8246429228431943e-05}
{"train_loss": 0.06578608602285385, "global_step": 195321, "epoch": 2194, "lr": 1.8245981411040226e-05}
{"train_loss": 0.0656365230679512, "global_step": 195322, "epoch": 2194, "lr": 1.824553359791743e-05}
{"train_loss": 0.026707405224442482, "global_step": 195323, "epoch": 2194, "lr": 1.8245085789063654e-05}
{"train_loss": 0.040339116007089615, "global_step": 195324, "epoch": 2194, "lr": 1.8244637984478923e-05}
{"train_loss": 0.06680294126272202, "global_step": 195325, "epoch": 2194, "lr": 1.824419018416333e-05}
{"train_loss": 0.06014135479927063, "global_step": 195326, "epoch": 2194, "lr": 1.8243742388116898e-05}
{"train_loss": 0.06890954822301865, "global_step": 195327, "epoch": 2194, "lr": 1.824329459633973e-05}
{"train_loss": 0.14921841025352478, "global_step": 195328, "epoch": 2194, "lr": 1.8242846808831843e-05}
{"train_loss": 0.0744466781616211, "global_step": 195329, "epoch": 2194, "lr": 1.824239902559334e-05}
{"train_loss": 0.07733199000358582, "global_step": 195330, "epoch": 2194, "lr": 1.8241951246624232e-05}
{"train_loss": 0.028980350121855736, "global_step": 195331, "epoch": 2194, "lr": 1.8241503471924637e-05}
{"train_loss": 0.07126826047897339, "global_step": 195332, "epoch": 2194, "lr": 1.8241055701494558e-05}
{"train_loss": 0.06340048462152481, "global_step": 195333, "epoch": 2194, "lr": 1.8240607935334105e-05}
{"train_loss": 0.07696593552827835, "global_step": 195334, "epoch": 2194, "lr": 1.8240160173443294e-05}
{"train_loss": 0.10811911523342133, "global_step": 195335, "epoch": 2194, "lr": 1.8239712415822223e-05}
{"train_loss": 0.06991001218557358, "global_step": 195336, "epoch": 2194, "lr": 1.8239264662470924e-05}
{"train_loss": 0.06589705497026443, "global_step": 195337, "epoch": 2194, "lr": 1.8238816913389477e-05}
{"train_loss": 0.06818974763154984, "global_step": 195338, "epoch": 2194, "lr": 1.8238369168577924e-05}
{"train_loss": 0.06970570236444473, "global_step": 195339, "epoch": 2194, "lr": 1.8237921428036337e-05}
{"train_loss": 0.09644552320241928, "global_step": 195340, "epoch": 2194, "lr": 1.8237473691764783e-05}
{"train_loss": 0.04212598875164986, "global_step": 195341, "epoch": 2194, "lr": 1.8237025959763305e-05}
{"train_loss": 0.060511719435453415, "global_step": 195342, "epoch": 2194, "lr": 1.823657823203198e-05}
{"train_loss": 0.03599356859922409, "global_step": 195343, "epoch": 2194, "lr": 1.8236130508570852e-05}
{"train_loss": 0.041617706418037415, "global_step": 195344, "epoch": 2194, "lr": 1.823568278938e-05}
{"train_loss": 0.10107863694429398, "global_step": 195345, "epoch": 2194, "lr": 1.823523507445946e-05}
{"train_loss": 0.07701543718576431, "global_step": 195346, "epoch": 2194, "lr": 1.823478736380932e-05}
{"train_loss": 0.07397447526454926, "global_step": 195347, "epoch": 2194, "lr": 1.8234339657429612e-05}
{"train_loss": 0.09413456171751022, "global_step": 195348, "epoch": 2194, "lr": 1.823389195532042e-05}
{"train_loss": 0.06701762229204178, "global_step": 195349, "epoch": 2194, "lr": 1.823344425748178e-05}
{"train_loss": 0.04574405774474144, "global_step": 195350, "epoch": 2194, "lr": 1.8232996563913786e-05}
{"train_loss": 0.08309239894151688, "global_step": 195351, "epoch": 2194, "lr": 1.8232548874616463e-05}
{"train_loss": 0.08533240109682083, "global_step": 195352, "epoch": 2194, "lr": 1.82321011895899e-05}
{"train_loss": 0.10625062137842178, "global_step": 195353, "epoch": 2194, "lr": 1.8231653508834123e-05}
{"train_loss": 0.06837879641379198, "global_step": 195354, "epoch": 2194, "lr": 1.8231205832349234e-05, "val_loss": 8.030805587768555}
{"train_loss": 0.060748688876628876, "global_step": 195355, "epoch": 2195, "lr": 1.823075816013526e-05}
{"train_loss": 0.09061497449874878, "global_step": 195356, "epoch": 2195, "lr": 1.8230310492192283e-05}
{"train_loss": 0.05582486838102341, "global_step": 195357, "epoch": 2195, "lr": 1.8229862828520357e-05}
{"train_loss": 0.0585441030561924, "global_step": 195358, "epoch": 2195, "lr": 1.8229415169119523e-05}
{"train_loss": 0.04017210751771927, "global_step": 195359, "epoch": 2195, "lr": 1.822896751398987e-05}
{"train_loss": 0.05218065157532692, "global_step": 195360, "epoch": 2195, "lr": 1.8228519863131434e-05}
{"train_loss": 0.06338193267583847, "global_step": 195361, "epoch": 2195, "lr": 1.8228072216544294e-05}
{"train_loss": 0.0731593444943428, "global_step": 195362, "epoch": 2195, "lr": 1.822762457422849e-05}
{"train_loss": 0.059467919170856476, "global_step": 195363, "epoch": 2195, "lr": 1.8227176936184115e-05}
{"train_loss": 0.05377919226884842, "global_step": 195364, "epoch": 2195, "lr": 1.8226729302411188e-05}
{"train_loss": 0.02646651491522789, "global_step": 195365, "epoch": 2195, "lr": 1.822628167290981e-05}
{"train_loss": 0.09154604375362396, "global_step": 195366, "epoch": 2195, "lr": 1.822583404768e-05}
{"train_loss": 0.06807275861501694, "global_step": 195367, "epoch": 2195, "lr": 1.8225386426721865e-05}
{"train_loss": 0.08240965008735657, "global_step": 195368, "epoch": 2195, "lr": 1.8224938810035415e-05}
{"train_loss": 0.06363734602928162, "global_step": 195369, "epoch": 2195, "lr": 1.8224491197620752e-05}
{"train_loss": 0.08265470713376999, "global_step": 195370, "epoch": 2195, "lr": 1.82240435894779e-05}
{"train_loss": 0.09088878333568573, "global_step": 195371, "epoch": 2195, "lr": 1.822359598560696e-05}
{"train_loss": 0.051335420459508896, "global_step": 195372, "epoch": 2195, "lr": 1.822314838600795e-05}
{"train_loss": 0.07840850949287415, "global_step": 195373, "epoch": 2195, "lr": 1.8222700790680968e-05}
{"train_loss": 0.0659354105591774, "global_step": 195374, "epoch": 2195, "lr": 1.822225319962604e-05}
{"train_loss": 0.05218060314655304, "global_step": 195375, "epoch": 2195, "lr": 1.8221805612843244e-05}
{"train_loss": 0.054440852254629135, "global_step": 195376, "epoch": 2195, "lr": 1.8221358030332653e-05}
{"train_loss": 0.06568234413862228, "global_step": 195377, "epoch": 2195, "lr": 1.82209104520943e-05}
{"train_loss": 0.07715674489736557, "global_step": 195378, "epoch": 2195, "lr": 1.822046287812827e-05}
{"train_loss": 0.03477578982710838, "global_step": 195379, "epoch": 2195, "lr": 1.8220015308434603e-05}
{"train_loss": 0.08199694752693176, "global_step": 195380, "epoch": 2195, "lr": 1.821956774301338e-05}
{"train_loss": 0.056356143206357956, "global_step": 195381, "epoch": 2195, "lr": 1.8219120181864622e-05}
{"train_loss": 0.08762939274311066, "global_step": 195382, "epoch": 2195, "lr": 1.8218672624988446e-05}
{"train_loss": 0.08175497502088547, "global_step": 195383, "epoch": 2195, "lr": 1.821822507238486e-05}
{"train_loss": 0.05161983519792557, "global_step": 195384, "epoch": 2195, "lr": 1.8217777524053962e-05}
{"train_loss": 0.07959440350532532, "global_step": 195385, "epoch": 2195, "lr": 1.8217329979995783e-05}
{"train_loss": 0.06261246651411057, "global_step": 195386, "epoch": 2195, "lr": 1.8216882440210415e-05}
{"train_loss": 0.1016116738319397, "global_step": 195387, "epoch": 2195, "lr": 1.8216434904697877e-05}
{"train_loss": 0.08656744658946991, "global_step": 195388, "epoch": 2195, "lr": 1.821598737345827e-05}
{"train_loss": 0.0418335497379303, "global_step": 195389, "epoch": 2195, "lr": 1.8215539846491624e-05}
{"train_loss": 0.05268280953168869, "global_step": 195390, "epoch": 2195, "lr": 1.821509232379803e-05}
{"train_loss": 0.04383881390094757, "global_step": 195391, "epoch": 2195, "lr": 1.82146448053775e-05}
{"train_loss": 0.04683750122785568, "global_step": 195392, "epoch": 2195, "lr": 1.8214197291230152e-05}
{"train_loss": 0.09083394706249237, "global_step": 195393, "epoch": 2195, "lr": 1.8213749781356e-05}
{"train_loss": 0.057092245668172836, "global_step": 195394, "epoch": 2195, "lr": 1.8213302275755133e-05}
{"train_loss": 0.05177759751677513, "global_step": 195395, "epoch": 2195, "lr": 1.8212854774427584e-05}
{"train_loss": 0.05880017578601837, "global_step": 195396, "epoch": 2195, "lr": 1.821240727737345e-05}
{"train_loss": 0.048612721264362335, "global_step": 195397, "epoch": 2195, "lr": 1.821195978459277e-05}
{"train_loss": 0.06019449606537819, "global_step": 195398, "epoch": 2195, "lr": 1.8211512296085587e-05}
{"train_loss": 0.05242350697517395, "global_step": 195399, "epoch": 2195, "lr": 1.821106481185199e-05}
{"train_loss": 0.10813125222921371, "global_step": 195400, "epoch": 2195, "lr": 1.821061733189202e-05}
{"train_loss": 0.09019318222999573, "global_step": 195401, "epoch": 2195, "lr": 1.8210169856205754e-05}
{"train_loss": 0.05012064054608345, "global_step": 195402, "epoch": 2195, "lr": 1.820972238479323e-05}
{"train_loss": 0.03679453209042549, "global_step": 195403, "epoch": 2195, "lr": 1.820927491765454e-05}
{"train_loss": 0.06667623668909073, "global_step": 195404, "epoch": 2195, "lr": 1.8208827454789707e-05}
{"train_loss": 0.07650966942310333, "global_step": 195405, "epoch": 2195, "lr": 1.8208379996198826e-05}
{"train_loss": 0.10215292870998383, "global_step": 195406, "epoch": 2195, "lr": 1.8207932541881928e-05}
{"train_loss": 0.07424233853816986, "global_step": 195407, "epoch": 2195, "lr": 1.82074850918391e-05}
{"train_loss": 0.06887513399124146, "global_step": 195408, "epoch": 2195, "lr": 1.8207037646070367e-05}
{"train_loss": 0.03211606293916702, "global_step": 195409, "epoch": 2195, "lr": 1.8206590204575837e-05}
{"train_loss": 0.06185826286673546, "global_step": 195410, "epoch": 2195, "lr": 1.8206142767355516e-05}
{"train_loss": 0.0793447345495224, "global_step": 195411, "epoch": 2195, "lr": 1.8205695334409513e-05}
{"train_loss": 0.09912866353988647, "global_step": 195412, "epoch": 2195, "lr": 1.8205247905737848e-05}
{"train_loss": 0.05106677487492561, "global_step": 195413, "epoch": 2195, "lr": 1.82048004813406e-05}
{"train_loss": 0.04464918002486229, "global_step": 195414, "epoch": 2195, "lr": 1.820435306121785e-05}
{"train_loss": 0.08160768449306488, "global_step": 195415, "epoch": 2195, "lr": 1.820390564536962e-05}
{"train_loss": 0.027178281918168068, "global_step": 195416, "epoch": 2195, "lr": 1.8203458233796e-05}
{"train_loss": 0.07136000692844391, "global_step": 195417, "epoch": 2195, "lr": 1.8203010826497026e-05}
{"train_loss": 0.05761769786477089, "global_step": 195418, "epoch": 2195, "lr": 1.820256342347278e-05}
{"train_loss": 0.03860224783420563, "global_step": 195419, "epoch": 2195, "lr": 1.82021160247233e-05}
{"train_loss": 0.04104023799300194, "global_step": 195420, "epoch": 2195, "lr": 1.820166863024867e-05}
{"train_loss": 0.12123821675777435, "global_step": 195421, "epoch": 2195, "lr": 1.8201221240048926e-05}
{"train_loss": 0.07307973504066467, "global_step": 195422, "epoch": 2195, "lr": 1.8200773854124154e-05}
{"train_loss": 0.03745988756418228, "global_step": 195423, "epoch": 2195, "lr": 1.820032647247439e-05}
{"train_loss": 0.05385186895728111, "global_step": 195424, "epoch": 2195, "lr": 1.8199879095099714e-05}
{"train_loss": 0.04132472723722458, "global_step": 195425, "epoch": 2195, "lr": 1.819943172200016e-05}
{"train_loss": 0.08240503072738647, "global_step": 195426, "epoch": 2195, "lr": 1.8198984353175825e-05}
{"train_loss": 0.03701576590538025, "global_step": 195427, "epoch": 2195, "lr": 1.819853698862673e-05}
{"train_loss": 0.12399838864803314, "global_step": 195428, "epoch": 2195, "lr": 1.8198089628352972e-05}
{"train_loss": 0.11016449332237244, "global_step": 195429, "epoch": 2195, "lr": 1.8197642272354576e-05}
{"train_loss": 0.058835357427597046, "global_step": 195430, "epoch": 2195, "lr": 1.8197194920631633e-05}
{"train_loss": 0.06406380236148834, "global_step": 195431, "epoch": 2195, "lr": 1.819674757318418e-05}
{"train_loss": 0.10748057067394257, "global_step": 195432, "epoch": 2195, "lr": 1.81963002300123e-05}
{"train_loss": 0.06546135991811752, "global_step": 195433, "epoch": 2195, "lr": 1.8195852891116023e-05}
{"train_loss": 0.04612861946225166, "global_step": 195434, "epoch": 2195, "lr": 1.8195405556495438e-05}
{"train_loss": 0.06962595134973526, "global_step": 195435, "epoch": 2195, "lr": 1.81949582261506e-05}
{"train_loss": 0.07357250154018402, "global_step": 195436, "epoch": 2195, "lr": 1.819451090008154e-05}
{"train_loss": 0.07617361843585968, "global_step": 195437, "epoch": 2195, "lr": 1.819406357828836e-05}
{"train_loss": 0.059824854135513306, "global_step": 195438, "epoch": 2195, "lr": 1.8193616260771085e-05}
{"train_loss": 0.04528231918811798, "global_step": 195439, "epoch": 2195, "lr": 1.8193168947529805e-05}
{"train_loss": 0.13025127351284027, "global_step": 195440, "epoch": 2195, "lr": 1.8192721638564553e-05}
{"train_loss": 0.09915374219417572, "global_step": 195441, "epoch": 2195, "lr": 1.8192274333875415e-05}
{"train_loss": 0.09503438323736191, "global_step": 195442, "epoch": 2195, "lr": 1.819182703346242e-05}
{"train_loss": 0.0667847033087792, "global_step": 195443, "epoch": 2195, "lr": 1.8191379737325666e-05, "val_loss": 7.928725242614746, "train_action_mse_error": 4.498798370361328}
{"train_loss": 0.060750238597393036, "global_step": 195444, "epoch": 2196, "lr": 1.8190932445465174e-05}
{"train_loss": 0.07872936874628067, "global_step": 195445, "epoch": 2196, "lr": 1.8190485157881043e-05}
{"train_loss": 0.1146920844912529, "global_step": 195446, "epoch": 2196, "lr": 1.8190037874573297e-05}
{"train_loss": 0.07651287317276001, "global_step": 195447, "epoch": 2196, "lr": 1.8189590595542027e-05}
{"train_loss": 0.10736435651779175, "global_step": 195448, "epoch": 2196, "lr": 1.818914332078726e-05}
{"train_loss": 0.03991812467575073, "global_step": 195449, "epoch": 2196, "lr": 1.8188696050309078e-05}
{"train_loss": 0.10855657607316971, "global_step": 195450, "epoch": 2196, "lr": 1.8188248784107552e-05}
{"train_loss": 0.03821684792637825, "global_step": 195451, "epoch": 2196, "lr": 1.818780152218272e-05}
{"train_loss": 0.13270151615142822, "global_step": 195452, "epoch": 2196, "lr": 1.8187354264534663e-05}
{"train_loss": 0.104768306016922, "global_step": 195453, "epoch": 2196, "lr": 1.8186907011163406e-05}
{"train_loss": 0.05365586280822754, "global_step": 195454, "epoch": 2196, "lr": 1.818645976206905e-05}
{"train_loss": 0.04719645529985428, "global_step": 195455, "epoch": 2196, "lr": 1.8186012517251628e-05}
{"train_loss": 0.028970392420887947, "global_step": 195456, "epoch": 2196, "lr": 1.8185565276711215e-05}
{"train_loss": 0.0627472922205925, "global_step": 195457, "epoch": 2196, "lr": 1.818511804044786e-05}
{"train_loss": 0.023263603448867798, "global_step": 195458, "epoch": 2196, "lr": 1.8184670808461635e-05}
{"train_loss": 0.07056517899036407, "global_step": 195459, "epoch": 2196, "lr": 1.818422358075258e-05}
{"train_loss": 0.09043817967176437, "global_step": 195460, "epoch": 2196, "lr": 1.8183776357320782e-05}
{"train_loss": 0.019410792738199234, "global_step": 195461, "epoch": 2196, "lr": 1.818332913816627e-05}
{"train_loss": 0.14660297334194183, "global_step": 195462, "epoch": 2196, "lr": 1.8182881923289148e-05}
{"train_loss": 0.05564253032207489, "global_step": 195463, "epoch": 2196, "lr": 1.8182434712689422e-05}
{"train_loss": 0.05726921185851097, "global_step": 195464, "epoch": 2196, "lr": 1.8181987506367203e-05}
{"train_loss": 0.08407092839479446, "global_step": 195465, "epoch": 2196, "lr": 1.8181540304322507e-05}
{"train_loss": 0.03394573926925659, "global_step": 195466, "epoch": 2196, "lr": 1.818109310655542e-05}
{"train_loss": 0.07522398233413696, "global_step": 195467, "epoch": 2196, "lr": 1.8180645913066013e-05}
{"train_loss": 0.08540048450231552, "global_step": 195468, "epoch": 2196, "lr": 1.818019872385433e-05}
{"train_loss": 0.06766916066408157, "global_step": 195469, "epoch": 2196, "lr": 1.8179751538920407e-05}
{"train_loss": 0.08946658670902252, "global_step": 195470, "epoch": 2196, "lr": 1.8179304358264354e-05}
{"train_loss": 0.06382212787866592, "global_step": 195471, "epoch": 2196, "lr": 1.817885718188619e-05}
{"train_loss": 0.062060900032520294, "global_step": 195472, "epoch": 2196, "lr": 1.8178410009786e-05}
{"train_loss": 0.06273500621318817, "global_step": 195473, "epoch": 2196, "lr": 1.8177962841963842e-05}
{"train_loss": 0.06651180982589722, "global_step": 195474, "epoch": 2196, "lr": 1.8177515678419748e-05}
{"train_loss": 0.03983106091618538, "global_step": 195475, "epoch": 2196, "lr": 1.8177068519153824e-05}
{"train_loss": 0.06661923974752426, "global_step": 195476, "epoch": 2196, "lr": 1.817662136416608e-05}
{"train_loss": 0.09266161173582077, "global_step": 195477, "epoch": 2196, "lr": 1.8176174213456622e-05}
{"train_loss": 0.0730888769030571, "global_step": 195478, "epoch": 2196, "lr": 1.8175727067025466e-05}
{"train_loss": 0.061912473291158676, "global_step": 195479, "epoch": 2196, "lr": 1.8175279924872718e-05}
{"train_loss": 0.03620247170329094, "global_step": 195480, "epoch": 2196, "lr": 1.8174832786998398e-05}
{"train_loss": 0.03394246846437454, "global_step": 195481, "epoch": 2196, "lr": 1.8174385653402604e-05}
{"train_loss": 0.06865332275629044, "global_step": 195482, "epoch": 2196, "lr": 1.8173938524085356e-05}
{"train_loss": 0.056758712977170944, "global_step": 195483, "epoch": 2196, "lr": 1.817349139904675e-05}
{"train_loss": 0.06639575213193893, "global_step": 195484, "epoch": 2196, "lr": 1.817304427828681e-05}
{"train_loss": 0.06045738235116005, "global_step": 195485, "epoch": 2196, "lr": 1.8172597161805622e-05}
{"train_loss": 0.13604366779327393, "global_step": 195486, "epoch": 2196, "lr": 1.817215004960326e-05}
{"train_loss": 0.047890108078718185, "global_step": 195487, "epoch": 2196, "lr": 1.817170294167974e-05}
{"train_loss": 0.0339798666536808, "global_step": 195488, "epoch": 2196, "lr": 1.8171255838035167e-05}
{"train_loss": 0.06114199385046959, "global_step": 195489, "epoch": 2196, "lr": 1.8170808738669563e-05}
{"train_loss": 0.07319877296686172, "global_step": 195490, "epoch": 2196, "lr": 1.8170361643583024e-05}
{"train_loss": 0.07920137047767639, "global_step": 195491, "epoch": 2196, "lr": 1.8169914552775574e-05}
{"train_loss": 0.044959623366594315, "global_step": 195492, "epoch": 2196, "lr": 1.816946746624731e-05}
{"train_loss": 0.036297474056482315, "global_step": 195493, "epoch": 2196, "lr": 1.8169020383998253e-05}
{"train_loss": 0.06913471221923828, "global_step": 195494, "epoch": 2196, "lr": 1.81685733060285e-05}
{"train_loss": 0.08424196392297745, "global_step": 195495, "epoch": 2196, "lr": 1.8168126232338086e-05}
{"train_loss": 0.06574513763189316, "global_step": 195496, "epoch": 2196, "lr": 1.8167679162927087e-05}
{"train_loss": 0.09761488437652588, "global_step": 195497, "epoch": 2196, "lr": 1.816723209779554e-05}
{"train_loss": 0.04130198061466217, "global_step": 195498, "epoch": 2196, "lr": 1.816678503694354e-05}
{"train_loss": 0.06491335481405258, "global_step": 195499, "epoch": 2196, "lr": 1.8166337980371106e-05}
{"train_loss": 0.01495323982089758, "global_step": 195500, "epoch": 2196, "lr": 1.8165890928078345e-05}
{"train_loss": 0.08017027378082275, "global_step": 195501, "epoch": 2196, "lr": 1.816544388006527e-05}
{"train_loss": 0.029099546372890472, "global_step": 195502, "epoch": 2196, "lr": 1.816499683633196e-05}
{"train_loss": 0.057607218623161316, "global_step": 195503, "epoch": 2196, "lr": 1.81645497968785e-05}
{"train_loss": 0.05597805231809616, "global_step": 195504, "epoch": 2196, "lr": 1.816410276170491e-05}
{"train_loss": 0.07756298780441284, "global_step": 195505, "epoch": 2196, "lr": 1.8163655730811286e-05}
{"train_loss": 0.08334416896104813, "global_step": 195506, "epoch": 2196, "lr": 1.816320870419767e-05}
{"train_loss": 0.06262791901826859, "global_step": 195507, "epoch": 2196, "lr": 1.8162761681864106e-05}
{"train_loss": 0.09691586345434189, "global_step": 195508, "epoch": 2196, "lr": 1.8162314663810687e-05}
{"train_loss": 0.05354669690132141, "global_step": 195509, "epoch": 2196, "lr": 1.816186765003744e-05}
{"train_loss": 0.029642300680279732, "global_step": 195510, "epoch": 2196, "lr": 1.8161420640544458e-05}
{"train_loss": 0.07310841977596283, "global_step": 195511, "epoch": 2196, "lr": 1.8160973635331767e-05}
{"train_loss": 0.07774199545383453, "global_step": 195512, "epoch": 2196, "lr": 1.8160526634399466e-05}
{"train_loss": 0.1410713791847229, "global_step": 195513, "epoch": 2196, "lr": 1.816007963774759e-05}
{"train_loss": 0.09287066757678986, "global_step": 195514, "epoch": 2196, "lr": 1.8159632645376184e-05}
{"train_loss": 0.029086343944072723, "global_step": 195515, "epoch": 2196, "lr": 1.815918565728535e-05}
{"train_loss": 0.06580528616905212, "global_step": 195516, "epoch": 2196, "lr": 1.8158738673475107e-05}
{"train_loss": 0.08635949343442917, "global_step": 195517, "epoch": 2196, "lr": 1.815829169394555e-05}
{"train_loss": 0.06129425764083862, "global_step": 195518, "epoch": 2196, "lr": 1.8157844718696703e-05}
{"train_loss": 0.06479987502098083, "global_step": 195519, "epoch": 2196, "lr": 1.8157397747728666e-05}
{"train_loss": 0.10435820370912552, "global_step": 195520, "epoch": 2196, "lr": 1.8156950781041467e-05}
{"train_loss": 0.07679174095392227, "global_step": 195521, "epoch": 2196, "lr": 1.8156503818635168e-05}
{"train_loss": 0.054320573806762695, "global_step": 195522, "epoch": 2196, "lr": 1.815605686050986e-05}
{"train_loss": 0.018369989469647408, "global_step": 195523, "epoch": 2196, "lr": 1.8155609906665565e-05}
{"train_loss": 0.052643194794654846, "global_step": 195524, "epoch": 2196, "lr": 1.815516295710238e-05}
{"train_loss": 0.08715040981769562, "global_step": 195525, "epoch": 2196, "lr": 1.8154716011820322e-05}
{"train_loss": 0.08403150737285614, "global_step": 195526, "epoch": 2196, "lr": 1.8154269070819495e-05}
{"train_loss": 0.12149938941001892, "global_step": 195527, "epoch": 2196, "lr": 1.8153822134099924e-05}
{"train_loss": 0.05529677867889404, "global_step": 195528, "epoch": 2196, "lr": 1.8153375201661692e-05}
{"train_loss": 0.04591095447540283, "global_step": 195529, "epoch": 2196, "lr": 1.8152928273504845e-05}
{"train_loss": 0.11250517517328262, "global_step": 195530, "epoch": 2196, "lr": 1.815248134962946e-05}
{"train_loss": 0.04795365780591965, "global_step": 195531, "epoch": 2196, "lr": 1.8152034430035562e-05}
{"train_loss": 0.06828179630042797, "global_step": 195532, "epoch": 2196, "lr": 1.8151587514723262e-05, "val_loss": 8.012608528137207}
{"train_loss": 0.08946512639522552, "global_step": 195533, "epoch": 2197, "lr": 1.8151140603692568e-05}
{"train_loss": 0.05503925308585167, "global_step": 195534, "epoch": 2197, "lr": 1.8150693696943588e-05}
{"train_loss": 0.05669591203331947, "global_step": 195535, "epoch": 2197, "lr": 1.815024679447634e-05}
{"train_loss": 0.11923687160015106, "global_step": 195536, "epoch": 2197, "lr": 1.814979989629092e-05}
{"train_loss": 0.05444435775279999, "global_step": 195537, "epoch": 2197, "lr": 1.8149353002387353e-05}
{"train_loss": 0.08697664737701416, "global_step": 195538, "epoch": 2197, "lr": 1.8148906112765735e-05}
{"train_loss": 0.06679963320493698, "global_step": 195539, "epoch": 2197, "lr": 1.8148459227426095e-05}
{"train_loss": 0.06861542910337448, "global_step": 195540, "epoch": 2197, "lr": 1.8148012346368505e-05}
{"train_loss": 0.08068627864122391, "global_step": 195541, "epoch": 2197, "lr": 1.8147565469593036e-05}
{"train_loss": 0.06453920155763626, "global_step": 195542, "epoch": 2197, "lr": 1.8147118597099732e-05}
{"train_loss": 0.055950701236724854, "global_step": 195543, "epoch": 2197, "lr": 1.814667172888867e-05}
{"train_loss": 0.08309391140937805, "global_step": 195544, "epoch": 2197, "lr": 1.8146224864959886e-05}
{"train_loss": 0.04781893268227577, "global_step": 195545, "epoch": 2197, "lr": 1.814577800531347e-05}
{"train_loss": 0.07166527956724167, "global_step": 195546, "epoch": 2197, "lr": 1.8145331149949462e-05}
{"train_loss": 0.08795938640832901, "global_step": 195547, "epoch": 2197, "lr": 1.8144884298867914e-05}
{"train_loss": 0.03360230103135109, "global_step": 195548, "epoch": 2197, "lr": 1.8144437452068914e-05}
{"train_loss": 0.023977670818567276, "global_step": 195549, "epoch": 2197, "lr": 1.8143990609552492e-05}
{"train_loss": 0.08967658132314682, "global_step": 195550, "epoch": 2197, "lr": 1.8143543771318733e-05}
{"train_loss": 0.0851900577545166, "global_step": 195551, "epoch": 2197, "lr": 1.8143096937367688e-05}
{"train_loss": 0.08815456926822662, "global_step": 195552, "epoch": 2197, "lr": 1.81426501076994e-05}
{"train_loss": 0.041746921837329865, "global_step": 195553, "epoch": 2197, "lr": 1.8142203282313962e-05}
{"train_loss": 0.05196475237607956, "global_step": 195554, "epoch": 2197, "lr": 1.81417564612114e-05}
{"train_loss": 0.20897358655929565, "global_step": 195555, "epoch": 2197, "lr": 1.8141309644391807e-05}
{"train_loss": 0.06267236918210983, "global_step": 195556, "epoch": 2197, "lr": 1.8140862831855205e-05}
{"train_loss": 0.09221403300762177, "global_step": 195557, "epoch": 2197, "lr": 1.814041602360168e-05}
{"train_loss": 0.023686658591032028, "global_step": 195558, "epoch": 2197, "lr": 1.8139969219631304e-05}
{"train_loss": 0.03081119805574417, "global_step": 195559, "epoch": 2197, "lr": 1.8139522419944104e-05}
{"train_loss": 0.05413217097520828, "global_step": 195560, "epoch": 2197, "lr": 1.8139075624540176e-05}
{"train_loss": 0.04687563329935074, "global_step": 195561, "epoch": 2197, "lr": 1.813862883341954e-05}
{"train_loss": 0.04580570384860039, "global_step": 195562, "epoch": 2197, "lr": 1.81381820465823e-05}
{"train_loss": 0.0965893343091011, "global_step": 195563, "epoch": 2197, "lr": 1.8137735264028467e-05}
{"train_loss": 0.08760496228933334, "global_step": 195564, "epoch": 2197, "lr": 1.8137288485758146e-05}
{"train_loss": 0.07173701375722885, "global_step": 195565, "epoch": 2197, "lr": 1.8136841711771368e-05}
{"train_loss": 0.03886944055557251, "global_step": 195566, "epoch": 2197, "lr": 1.8136394942068213e-05}
{"train_loss": 0.09795855730772018, "global_step": 195567, "epoch": 2197, "lr": 1.813594817664872e-05}
{"train_loss": 0.09742364287376404, "global_step": 195568, "epoch": 2197, "lr": 1.8135501415512973e-05}
{"train_loss": 0.03799278661608696, "global_step": 195569, "epoch": 2197, "lr": 1.8135054658661005e-05}
{"train_loss": 0.06028019264340401, "global_step": 195570, "epoch": 2197, "lr": 1.813460790609291e-05}
{"train_loss": 0.04617883265018463, "global_step": 195571, "epoch": 2197, "lr": 1.813416115780871e-05}
{"train_loss": 0.06915980577468872, "global_step": 195572, "epoch": 2197, "lr": 1.81337144138085e-05}
{"train_loss": 0.0789942592382431, "global_step": 195573, "epoch": 2197, "lr": 1.8133267674092304e-05}
{"train_loss": 0.04307373985648155, "global_step": 195574, "epoch": 2197, "lr": 1.8132820938660223e-05}
{"train_loss": 0.04772306606173515, "global_step": 195575, "epoch": 2197, "lr": 1.8132374207512276e-05}
{"train_loss": 0.06839794665575027, "global_step": 195576, "epoch": 2197, "lr": 1.8131927480648543e-05}
{"train_loss": 0.049721527844667435, "global_step": 195577, "epoch": 2197, "lr": 1.8131480758069103e-05}
{"train_loss": 0.05713101476430893, "global_step": 195578, "epoch": 2197, "lr": 1.8131034039773976e-05}
{"train_loss": 0.10355596244335175, "global_step": 195579, "epoch": 2197, "lr": 1.813058732576326e-05}
{"train_loss": 0.09243828803300858, "global_step": 195580, "epoch": 2197, "lr": 1.8130140616036983e-05}
{"train_loss": 0.08311013877391815, "global_step": 195581, "epoch": 2197, "lr": 1.8129693910595243e-05}
{"train_loss": 0.04354582354426384, "global_step": 195582, "epoch": 2197, "lr": 1.812924720943805e-05}
{"train_loss": 0.0462338961660862, "global_step": 195583, "epoch": 2197, "lr": 1.8128800512565513e-05}
{"train_loss": 0.03158823773264885, "global_step": 195584, "epoch": 2197, "lr": 1.8128353819977674e-05}
{"train_loss": 0.03498179838061333, "global_step": 195585, "epoch": 2197, "lr": 1.8127907131674564e-05}
{"train_loss": 0.08675321936607361, "global_step": 195586, "epoch": 2197, "lr": 1.8127460447656292e-05}
{"train_loss": 0.0490979328751564, "global_step": 195587, "epoch": 2197, "lr": 1.812701376792288e-05}
{"train_loss": 0.07886981964111328, "global_step": 195588, "epoch": 2197, "lr": 1.8126567092474412e-05}
{"train_loss": 0.06781408190727234, "global_step": 195589, "epoch": 2197, "lr": 1.8126120421310943e-05}
{"train_loss": 0.047089315950870514, "global_step": 195590, "epoch": 2197, "lr": 1.8125673754432508e-05}
{"train_loss": 0.06416565924882889, "global_step": 195591, "epoch": 2197, "lr": 1.812522709183921e-05}
{"train_loss": 0.03452800586819649, "global_step": 195592, "epoch": 2197, "lr": 1.8124780433531068e-05}
{"train_loss": 0.053572896867990494, "global_step": 195593, "epoch": 2197, "lr": 1.8124333779508157e-05}
{"train_loss": 0.041004978120326996, "global_step": 195594, "epoch": 2197, "lr": 1.8123887129770562e-05}
{"train_loss": 0.030182618647813797, "global_step": 195595, "epoch": 2197, "lr": 1.8123440484318304e-05}
{"train_loss": 0.037547703832387924, "global_step": 195596, "epoch": 2197, "lr": 1.8122993843151476e-05}
{"train_loss": 0.08234544843435287, "global_step": 195597, "epoch": 2197, "lr": 1.8122547206270107e-05}
{"train_loss": 0.02776215597987175, "global_step": 195598, "epoch": 2197, "lr": 1.8122100573674287e-05}
{"train_loss": 0.047532323747873306, "global_step": 195599, "epoch": 2197, "lr": 1.8121653945364055e-05}
{"train_loss": 0.0844939574599266, "global_step": 195600, "epoch": 2197, "lr": 1.8121207321339483e-05}
{"train_loss": 0.06096000224351883, "global_step": 195601, "epoch": 2197, "lr": 1.8120760701600614e-05}
{"train_loss": 0.05265652388334274, "global_step": 195602, "epoch": 2197, "lr": 1.812031408614754e-05}
{"train_loss": 0.08035711944103241, "global_step": 195603, "epoch": 2197, "lr": 1.8119867474980283e-05}
{"train_loss": 0.05588790774345398, "global_step": 195604, "epoch": 2197, "lr": 1.8119420868098934e-05}
{"train_loss": 0.10563896596431732, "global_step": 195605, "epoch": 2197, "lr": 1.811897426550353e-05}
{"train_loss": 0.049405537545681, "global_step": 195606, "epoch": 2197, "lr": 1.811852766719415e-05}
{"train_loss": 0.06944915652275085, "global_step": 195607, "epoch": 2197, "lr": 1.8118081073170833e-05}
{"train_loss": 0.10289282351732254, "global_step": 195608, "epoch": 2197, "lr": 1.811763448343367e-05}
{"train_loss": 0.05255351960659027, "global_step": 195609, "epoch": 2197, "lr": 1.8117187897982685e-05}
{"train_loss": 0.04154520109295845, "global_step": 195610, "epoch": 2197, "lr": 1.811674131681797e-05}
{"train_loss": 0.131144180893898, "global_step": 195611, "epoch": 2197, "lr": 1.8116294739939555e-05}
{"train_loss": 0.083844855427742, "global_step": 195612, "epoch": 2197, "lr": 1.8115848167347515e-05}
{"train_loss": 0.07528822124004364, "global_step": 195613, "epoch": 2197, "lr": 1.811540159904193e-05}
{"train_loss": 0.05027295649051666, "global_step": 195614, "epoch": 2197, "lr": 1.8114955035022822e-05}
{"train_loss": 0.09655172377824783, "global_step": 195615, "epoch": 2197, "lr": 1.8114508475290288e-05}
{"train_loss": 0.08589750528335571, "global_step": 195616, "epoch": 2197, "lr": 1.811406191984435e-05}
{"train_loss": 0.07708702236413956, "global_step": 195617, "epoch": 2197, "lr": 1.8113615368685103e-05}
{"train_loss": 0.09963609278202057, "global_step": 195618, "epoch": 2197, "lr": 1.8113168821812582e-05}
{"train_loss": 0.059517864137887955, "global_step": 195619, "epoch": 2197, "lr": 1.8112722279226868e-05}
{"train_loss": 0.08609802275896072, "global_step": 195620, "epoch": 2197, "lr": 1.811227574092799e-05}
{"train_loss": 0.06670334085487248, "global_step": 195621, "epoch": 2197, "lr": 1.8111829206916052e-05, "val_loss": 8.058504104614258}
{"train_loss": 0.06951864063739777, "global_step": 195622, "epoch": 2198, "lr": 1.811138267719109e-05}
{"train_loss": 0.04254627227783203, "global_step": 195623, "epoch": 2198, "lr": 1.8110936151753144e-05}
{"train_loss": 0.04641476646065712, "global_step": 195624, "epoch": 2198, "lr": 1.8110489630602313e-05}
{"train_loss": 0.05979035049676895, "global_step": 195625, "epoch": 2198, "lr": 1.8110043113738617e-05}
{"train_loss": 0.04769608750939369, "global_step": 195626, "epoch": 2198, "lr": 1.8109596601162155e-05}
{"train_loss": 0.044252388179302216, "global_step": 195627, "epoch": 2198, "lr": 1.810915009287295e-05}
{"train_loss": 0.02873309887945652, "global_step": 195628, "epoch": 2198, "lr": 1.81087035888711e-05}
{"train_loss": 0.09292928874492645, "global_step": 195629, "epoch": 2198, "lr": 1.810825708915663e-05}
{"train_loss": 0.08871777355670929, "global_step": 195630, "epoch": 2198, "lr": 1.8107810593729635e-05}
{"train_loss": 0.07035738229751587, "global_step": 195631, "epoch": 2198, "lr": 1.810736410259014e-05}
{"train_loss": 0.05512396991252899, "global_step": 195632, "epoch": 2198, "lr": 1.8106917615738235e-05}
{"train_loss": 0.09885810315608978, "global_step": 195633, "epoch": 2198, "lr": 1.8106471133173953e-05}
{"train_loss": 0.051983144134283066, "global_step": 195634, "epoch": 2198, "lr": 1.810602465489738e-05}
{"train_loss": 0.052283044904470444, "global_step": 195635, "epoch": 2198, "lr": 1.8105578180908546e-05}
{"train_loss": 0.06300605088472366, "global_step": 195636, "epoch": 2198, "lr": 1.8105131711207547e-05}
{"train_loss": 0.11367516964673996, "global_step": 195637, "epoch": 2198, "lr": 1.8104685245794404e-05}
{"train_loss": 0.08376693725585938, "global_step": 195638, "epoch": 2198, "lr": 1.810423878466922e-05}
{"train_loss": 0.03276197984814644, "global_step": 195639, "epoch": 2198, "lr": 1.810379232783201e-05}
{"train_loss": 0.04180021584033966, "global_step": 195640, "epoch": 2198, "lr": 1.8103345875282878e-05}
{"train_loss": 0.0626121386885643, "global_step": 195641, "epoch": 2198, "lr": 1.810289942702184e-05}
{"train_loss": 0.06939379125833511, "global_step": 195642, "epoch": 2198, "lr": 1.8102452983049e-05}
{"train_loss": 0.09960198402404785, "global_step": 195643, "epoch": 2198, "lr": 1.8102006543364382e-05}
{"train_loss": 0.08366019278764725, "global_step": 195644, "epoch": 2198, "lr": 1.8101560107968068e-05}
{"train_loss": 0.060715049505233765, "global_step": 195645, "epoch": 2198, "lr": 1.8101113676860105e-05}
{"train_loss": 0.04811324179172516, "global_step": 195646, "epoch": 2198, "lr": 1.8100667250040566e-05}
{"train_loss": 0.10604003816843033, "global_step": 195647, "epoch": 2198, "lr": 1.810022082750949e-05}
{"train_loss": 0.04447568953037262, "global_step": 195648, "epoch": 2198, "lr": 1.809977440926695e-05}
{"train_loss": 0.06764398515224457, "global_step": 195649, "epoch": 2198, "lr": 1.8099327995313026e-05}
{"train_loss": 0.08101637661457062, "global_step": 195650, "epoch": 2198, "lr": 1.8098881585647736e-05}
{"train_loss": 0.055986437946558, "global_step": 195651, "epoch": 2198, "lr": 1.8098435180271185e-05}
{"train_loss": 0.0853019505739212, "global_step": 195652, "epoch": 2198, "lr": 1.8097988779183395e-05}
{"train_loss": 0.06229542940855026, "global_step": 195653, "epoch": 2198, "lr": 1.809754238238445e-05}
{"train_loss": 0.12478193640708923, "global_step": 195654, "epoch": 2198, "lr": 1.809709598987439e-05}
{"train_loss": 0.051758188754320145, "global_step": 195655, "epoch": 2198, "lr": 1.8096649601653302e-05}
{"train_loss": 0.07409387081861496, "global_step": 195656, "epoch": 2198, "lr": 1.809620321772122e-05}
{"train_loss": 0.05454619973897934, "global_step": 195657, "epoch": 2198, "lr": 1.8095756838078227e-05}
{"train_loss": 0.05606803297996521, "global_step": 195658, "epoch": 2198, "lr": 1.809531046272435e-05}
{"train_loss": 0.10002133250236511, "global_step": 195659, "epoch": 2198, "lr": 1.8094864091659692e-05}
{"train_loss": 0.04337136074900627, "global_step": 195660, "epoch": 2198, "lr": 1.809441772488427e-05}
{"train_loss": 0.049753908067941666, "global_step": 195661, "epoch": 2198, "lr": 1.8093971362398182e-05}
{"train_loss": 0.046097394078969955, "global_step": 195662, "epoch": 2198, "lr": 1.8093525004201474e-05}
{"train_loss": 0.07653799653053284, "global_step": 195663, "epoch": 2198, "lr": 1.809307865029418e-05}
{"train_loss": 0.09772530198097229, "global_step": 195664, "epoch": 2198, "lr": 1.809263230067641e-05}
{"train_loss": 0.03792712837457657, "global_step": 195665, "epoch": 2198, "lr": 1.8092185955348175e-05}
{"train_loss": 0.08039239048957825, "global_step": 195666, "epoch": 2198, "lr": 1.8091739614309566e-05}
{"train_loss": 0.06858988851308823, "global_step": 195667, "epoch": 2198, "lr": 1.8091293277560623e-05}
{"train_loss": 0.05618532374501228, "global_step": 195668, "epoch": 2198, "lr": 1.8090846945101435e-05}
{"train_loss": 0.035613011568784714, "global_step": 195669, "epoch": 2198, "lr": 1.8090400616932024e-05}
{"train_loss": 0.05966711416840553, "global_step": 195670, "epoch": 2198, "lr": 1.8089954293052487e-05}
{"train_loss": 0.07454860210418701, "global_step": 195671, "epoch": 2198, "lr": 1.808950797346285e-05}
{"train_loss": 0.06953831762075424, "global_step": 195672, "epoch": 2198, "lr": 1.808906165816321e-05}
{"train_loss": 0.05397928133606911, "global_step": 195673, "epoch": 2198, "lr": 1.8088615347153583e-05}
{"train_loss": 0.06648373603820801, "global_step": 195674, "epoch": 2198, "lr": 1.8088169040434072e-05}
{"train_loss": 0.08656563609838486, "global_step": 195675, "epoch": 2198, "lr": 1.8087722738004702e-05}
{"train_loss": 0.12813708186149597, "global_step": 195676, "epoch": 2198, "lr": 1.8087276439865568e-05}
{"train_loss": 0.030388403683900833, "global_step": 195677, "epoch": 2198, "lr": 1.8086830146016693e-05}
{"train_loss": 0.07012033462524414, "global_step": 195678, "epoch": 2198, "lr": 1.8086383856458168e-05}
{"train_loss": 0.062153078615665436, "global_step": 195679, "epoch": 2198, "lr": 1.8085937571190024e-05}
{"train_loss": 0.05806860327720642, "global_step": 195680, "epoch": 2198, "lr": 1.808549129021236e-05}
{"train_loss": 0.07215556502342224, "global_step": 195681, "epoch": 2198, "lr": 1.808504501352519e-05}
{"train_loss": 0.03429781645536423, "global_step": 195682, "epoch": 2198, "lr": 1.8084598741128612e-05}
{"train_loss": 0.054911058396101, "global_step": 195683, "epoch": 2198, "lr": 1.8084152473022653e-05}
{"train_loss": 0.0650077536702156, "global_step": 195684, "epoch": 2198, "lr": 1.8083706209207394e-05}
{"train_loss": 0.04902879521250725, "global_step": 195685, "epoch": 2198, "lr": 1.8083259949682913e-05}
{"train_loss": 0.10313599556684494, "global_step": 195686, "epoch": 2198, "lr": 1.8082813694449223e-05}
{"train_loss": 0.06414956599473953, "global_step": 195687, "epoch": 2198, "lr": 1.8082367443506432e-05}
{"train_loss": 0.026782477274537086, "global_step": 195688, "epoch": 2198, "lr": 1.8081921196854558e-05}
{"train_loss": 0.09422960132360458, "global_step": 195689, "epoch": 2198, "lr": 1.8081474954493698e-05}
{"train_loss": 0.06073521077632904, "global_step": 195690, "epoch": 2198, "lr": 1.808102871642388e-05}
{"train_loss": 0.08774121105670929, "global_step": 195691, "epoch": 2198, "lr": 1.8080582482645193e-05}
{"train_loss": 0.06348835676908493, "global_step": 195692, "epoch": 2198, "lr": 1.8080136253157672e-05}
{"train_loss": 0.0771360769867897, "global_step": 195693, "epoch": 2198, "lr": 1.8079690027961393e-05}
{"train_loss": 0.09758882969617844, "global_step": 195694, "epoch": 2198, "lr": 1.80792438070564e-05}
{"train_loss": 0.11204129457473755, "global_step": 195695, "epoch": 2198, "lr": 1.8078797590442787e-05}
{"train_loss": 0.12686683237552643, "global_step": 195696, "epoch": 2198, "lr": 1.807835137812056e-05}
{"train_loss": 0.0710110068321228, "global_step": 195697, "epoch": 2198, "lr": 1.8077905170089836e-05}
{"train_loss": 0.10191528499126434, "global_step": 195698, "epoch": 2198, "lr": 1.807745896635063e-05}
{"train_loss": 0.0701453760266304, "global_step": 195699, "epoch": 2198, "lr": 1.807701276690304e-05}
{"train_loss": 0.04686151072382927, "global_step": 195700, "epoch": 2198, "lr": 1.80765665717471e-05}
{"train_loss": 0.061714623123407364, "global_step": 195701, "epoch": 2198, "lr": 1.8076120380882865e-05}
{"train_loss": 0.07555081695318222, "global_step": 195702, "epoch": 2198, "lr": 1.807567419431042e-05}
{"train_loss": 0.03740227594971657, "global_step": 195703, "epoch": 2198, "lr": 1.80752280120298e-05}
{"train_loss": 0.07180584967136383, "global_step": 195704, "epoch": 2198, "lr": 1.8074781834041088e-05}
{"train_loss": 0.024064041674137115, "global_step": 195705, "epoch": 2198, "lr": 1.8074335660344317e-05}
{"train_loss": 0.0493813157081604, "global_step": 195706, "epoch": 2198, "lr": 1.8073889490939578e-05}
{"train_loss": 0.09160257875919342, "global_step": 195707, "epoch": 2198, "lr": 1.80734433258269e-05}
{"train_loss": 0.08016625791788101, "global_step": 195708, "epoch": 2198, "lr": 1.807299716500638e-05}
{"train_loss": 0.05069255456328392, "global_step": 195709, "epoch": 2198, "lr": 1.8072551008478034e-05}
{"train_loss": 0.06725557851657439, "global_step": 195710, "epoch": 2198, "lr": 1.807210485624196e-05, "val_loss": 8.111245155334473}
{"train_loss": 0.06025276705622673, "global_step": 195711, "epoch": 2199, "lr": 1.807165870829819e-05}
{"train_loss": 0.08254390954971313, "global_step": 195712, "epoch": 2199, "lr": 1.807121256464681e-05}
{"train_loss": 0.04883258789777756, "global_step": 195713, "epoch": 2199, "lr": 1.807076642528785e-05}
{"train_loss": 0.06951448321342468, "global_step": 195714, "epoch": 2199, "lr": 1.8070320290221403e-05}
{"train_loss": 0.03419455885887146, "global_step": 195715, "epoch": 2199, "lr": 1.8069874159447498e-05}
{"train_loss": 0.04429829865694046, "global_step": 195716, "epoch": 2199, "lr": 1.8069428032966224e-05}
{"train_loss": 0.06869800388813019, "global_step": 195717, "epoch": 2199, "lr": 1.8068981910777605e-05}
{"train_loss": 0.0485505573451519, "global_step": 195718, "epoch": 2199, "lr": 1.8068535792881745e-05}
{"train_loss": 0.05044855922460556, "global_step": 195719, "epoch": 2199, "lr": 1.806808967927866e-05}
{"train_loss": 0.05059277266263962, "global_step": 195720, "epoch": 2199, "lr": 1.8067643569968434e-05}
{"train_loss": 0.10094642639160156, "global_step": 195721, "epoch": 2199, "lr": 1.806719746495114e-05}
{"train_loss": 0.04387945309281349, "global_step": 195722, "epoch": 2199, "lr": 1.8066751364226807e-05}
{"train_loss": 0.08083122223615646, "global_step": 195723, "epoch": 2199, "lr": 1.806630526779553e-05}
{"train_loss": 0.11289119720458984, "global_step": 195724, "epoch": 2199, "lr": 1.8065859175657324e-05}
{"train_loss": 0.05333992838859558, "global_step": 195725, "epoch": 2199, "lr": 1.80654130878123e-05}
{"train_loss": 0.0634700134396553, "global_step": 195726, "epoch": 2199, "lr": 1.8064967004260468e-05}
{"train_loss": 0.06559160351753235, "global_step": 195727, "epoch": 2199, "lr": 1.806452092500193e-05}
{"train_loss": 0.048635002225637436, "global_step": 195728, "epoch": 2199, "lr": 1.806407485003671e-05}
{"train_loss": 0.0467023141682148, "global_step": 195729, "epoch": 2199, "lr": 1.8063628779364906e-05}
{"train_loss": 0.07651634514331818, "global_step": 195730, "epoch": 2199, "lr": 1.8063182712986542e-05}
{"train_loss": 0.08367332071065903, "global_step": 195731, "epoch": 2199, "lr": 1.8062736650901707e-05}
{"train_loss": 0.032418251037597656, "global_step": 195732, "epoch": 2199, "lr": 1.806229059311043e-05}
{"train_loss": 0.04406113922595978, "global_step": 195733, "epoch": 2199, "lr": 1.806184453961281e-05}
{"train_loss": 0.057343196123838425, "global_step": 195734, "epoch": 2199, "lr": 1.8061398490408865e-05}
{"train_loss": 0.08403194695711136, "global_step": 195735, "epoch": 2199, "lr": 1.8060952445498697e-05}
{"train_loss": 0.06930090487003326, "global_step": 195736, "epoch": 2199, "lr": 1.8060506404882325e-05}
{"train_loss": 0.06906411796808243, "global_step": 195737, "epoch": 2199, "lr": 1.8060060368559845e-05}
{"train_loss": 0.0547783263027668, "global_step": 195738, "epoch": 2199, "lr": 1.8059614336531307e-05}
{"train_loss": 0.07203662395477295, "global_step": 195739, "epoch": 2199, "lr": 1.8059168308796737e-05}
{"train_loss": 0.10691864043474197, "global_step": 195740, "epoch": 2199, "lr": 1.8058722285356246e-05}
{"train_loss": 0.050785843282938004, "global_step": 195741, "epoch": 2199, "lr": 1.8058276266209852e-05}
{"train_loss": 0.027705278247594833, "global_step": 195742, "epoch": 2199, "lr": 1.8057830251357648e-05}
{"train_loss": 0.03191041201353073, "global_step": 195743, "epoch": 2199, "lr": 1.805738424079967e-05}
{"train_loss": 0.0422348715364933, "global_step": 195744, "epoch": 2199, "lr": 1.8056938234535997e-05}
{"train_loss": 0.0732618048787117, "global_step": 195745, "epoch": 2199, "lr": 1.8056492232566662e-05}
{"train_loss": 0.05464489012956619, "global_step": 195746, "epoch": 2199, "lr": 1.805604623489176e-05}
{"train_loss": 0.09206971526145935, "global_step": 195747, "epoch": 2199, "lr": 1.8055600241511323e-05}
{"train_loss": 0.11202801018953323, "global_step": 195748, "epoch": 2199, "lr": 1.805515425242543e-05}
{"train_loss": 0.04703190550208092, "global_step": 195749, "epoch": 2199, "lr": 1.8054708267634117e-05}
{"train_loss": 0.05946972966194153, "global_step": 195750, "epoch": 2199, "lr": 1.805426228713748e-05}
{"train_loss": 0.05375015363097191, "global_step": 195751, "epoch": 2199, "lr": 1.805381631093554e-05}
{"train_loss": 0.11495760828256607, "global_step": 195752, "epoch": 2199, "lr": 1.8053370339028387e-05}
{"train_loss": 0.058596521615982056, "global_step": 195753, "epoch": 2199, "lr": 1.8052924371416057e-05}
{"train_loss": 0.05631681904196739, "global_step": 195754, "epoch": 2199, "lr": 1.8052478408098638e-05}
{"train_loss": 0.07463568449020386, "global_step": 195755, "epoch": 2199, "lr": 1.8052032449076157e-05}
{"train_loss": 0.06326163560152054, "global_step": 195756, "epoch": 2199, "lr": 1.805158649434869e-05}
{"train_loss": 0.1220436692237854, "global_step": 195757, "epoch": 2199, "lr": 1.8051140543916312e-05}
{"train_loss": 0.03240818530321121, "global_step": 195758, "epoch": 2199, "lr": 1.8050694597779056e-05}
{"train_loss": 0.06544245779514313, "global_step": 195759, "epoch": 2199, "lr": 1.8050248655937012e-05}
{"train_loss": 0.0783974677324295, "global_step": 195760, "epoch": 2199, "lr": 1.80498027183902e-05}
{"train_loss": 0.10179495811462402, "global_step": 195761, "epoch": 2199, "lr": 1.804935678513873e-05}
{"train_loss": 0.09234864264726639, "global_step": 195762, "epoch": 2199, "lr": 1.804891085618261e-05}
{"train_loss": 0.06953596323728561, "global_step": 195763, "epoch": 2199, "lr": 1.8048464931521943e-05}
{"train_loss": 0.05907362699508667, "global_step": 195764, "epoch": 2199, "lr": 1.804801901115675e-05}
{"train_loss": 0.03600389137864113, "global_step": 195765, "epoch": 2199, "lr": 1.8047573095087132e-05}
{"train_loss": 0.08910227566957474, "global_step": 195766, "epoch": 2199, "lr": 1.804712718331311e-05}
{"train_loss": 0.05735824257135391, "global_step": 195767, "epoch": 2199, "lr": 1.804668127583478e-05}
{"train_loss": 0.03574850782752037, "global_step": 195768, "epoch": 2199, "lr": 1.804623537265217e-05}
{"train_loss": 0.07474245876073837, "global_step": 195769, "epoch": 2199, "lr": 1.8045789473765367e-05}
{"train_loss": 0.033997148275375366, "global_step": 195770, "epoch": 2199, "lr": 1.8045343579174406e-05}
{"train_loss": 0.04970713332295418, "global_step": 195771, "epoch": 2199, "lr": 1.804489768887937e-05}
{"train_loss": 0.06380109488964081, "global_step": 195772, "epoch": 2199, "lr": 1.8044451802880297e-05}
{"train_loss": 0.07178381830453873, "global_step": 195773, "epoch": 2199, "lr": 1.804400592117727e-05}
{"train_loss": 0.06721225380897522, "global_step": 195774, "epoch": 2199, "lr": 1.8043560043770324e-05}
{"train_loss": 0.055783290416002274, "global_step": 195775, "epoch": 2199, "lr": 1.8043114170659546e-05}
{"train_loss": 0.10409484803676605, "global_step": 195776, "epoch": 2199, "lr": 1.8042668301844966e-05}
{"train_loss": 0.028331389650702477, "global_step": 195777, "epoch": 2199, "lr": 1.8042222437326677e-05}
{"train_loss": 0.05544969439506531, "global_step": 195778, "epoch": 2199, "lr": 1.8041776577104718e-05}
{"train_loss": 0.030336957424879074, "global_step": 195779, "epoch": 2199, "lr": 1.804133072117914e-05}
{"train_loss": 0.07354980707168579, "global_step": 195780, "epoch": 2199, "lr": 1.804088486955003e-05}
{"train_loss": 0.052999354898929596, "global_step": 195781, "epoch": 2199, "lr": 1.804043902221742e-05}
{"train_loss": 0.03441205248236656, "global_step": 195782, "epoch": 2199, "lr": 1.8039993179181398e-05}
{"train_loss": 0.0457431823015213, "global_step": 195783, "epoch": 2199, "lr": 1.8039547340441994e-05}
{"train_loss": 0.06210712715983391, "global_step": 195784, "epoch": 2199, "lr": 1.8039101505999296e-05}
{"train_loss": 0.07144194096326828, "global_step": 195785, "epoch": 2199, "lr": 1.803865567585334e-05}
{"train_loss": 0.09559153765439987, "global_step": 195786, "epoch": 2199, "lr": 1.8038209850004208e-05}
{"train_loss": 0.06212081015110016, "global_step": 195787, "epoch": 2199, "lr": 1.8037764028451935e-05}
{"train_loss": 0.049803752452135086, "global_step": 195788, "epoch": 2199, "lr": 1.8037318211196613e-05}
{"train_loss": 0.0653664767742157, "global_step": 195789, "epoch": 2199, "lr": 1.803687239823827e-05}
{"train_loss": 0.04438388720154762, "global_step": 195790, "epoch": 2199, "lr": 1.8036426589576988e-05}
{"train_loss": 0.08730262517929077, "global_step": 195791, "epoch": 2199, "lr": 1.8035980785212807e-05}
{"train_loss": 0.059323884546756744, "global_step": 195792, "epoch": 2199, "lr": 1.803553498514582e-05}
{"train_loss": 0.02866249904036522, "global_step": 195793, "epoch": 2199, "lr": 1.803508918937604e-05}
{"train_loss": 0.11365270614624023, "global_step": 195794, "epoch": 2199, "lr": 1.8034643397903562e-05}
{"train_loss": 0.049319662153720856, "global_step": 195795, "epoch": 2199, "lr": 1.8034197610728447e-05}
{"train_loss": 0.060034334659576416, "global_step": 195796, "epoch": 2199, "lr": 1.803375182785073e-05}
{"train_loss": 0.06444765627384186, "global_step": 195797, "epoch": 2199, "lr": 1.8033306049270503e-05}
{"train_loss": 0.05561312288045883, "global_step": 195798, "epoch": 2199, "lr": 1.8032860274987788e-05}
{"train_loss": 0.0631240021562978, "global_step": 195799, "epoch": 2199, "lr": 1.803241450500269e-05, "val_loss": 8.150259971618652}
{"train_loss": 0.02005426026880741, "global_step": 195800, "epoch": 2200, "lr": 1.803196873931522e-05}
{"train_loss": 0.04335717484354973, "global_step": 195801, "epoch": 2200, "lr": 1.803152297792548e-05}
{"train_loss": 0.04801657423377037, "global_step": 195802, "epoch": 2200, "lr": 1.8031077220833498e-05}
{"train_loss": 0.037307009100914, "global_step": 195803, "epoch": 2200, "lr": 1.803063146803936e-05}
{"train_loss": 0.05569036304950714, "global_step": 195804, "epoch": 2200, "lr": 1.8030185719543102e-05}
{"train_loss": 0.03840997815132141, "global_step": 195805, "epoch": 2200, "lr": 1.8029739975344813e-05}
{"train_loss": 0.061264313757419586, "global_step": 195806, "epoch": 2200, "lr": 1.802929423544451e-05}
{"train_loss": 0.0835680365562439, "global_step": 195807, "epoch": 2200, "lr": 1.8028848499842305e-05}
{"train_loss": 0.04834511876106262, "global_step": 195808, "epoch": 2200, "lr": 1.802840276853821e-05}
{"train_loss": 0.09983296692371368, "global_step": 195809, "epoch": 2200, "lr": 1.802795704153233e-05}
{"train_loss": 0.046874843537807465, "global_step": 195810, "epoch": 2200, "lr": 1.8027511318824682e-05}
{"train_loss": 0.04789130017161369, "global_step": 195811, "epoch": 2200, "lr": 1.8027065600415354e-05}
{"train_loss": 0.09795121103525162, "global_step": 195812, "epoch": 2200, "lr": 1.802661988630439e-05}
{"train_loss": 0.03797632455825806, "global_step": 195813, "epoch": 2200, "lr": 1.8026174176491874e-05}
{"train_loss": 0.03552233427762985, "global_step": 195814, "epoch": 2200, "lr": 1.8025728470977826e-05}
{"train_loss": 0.06557487696409225, "global_step": 195815, "epoch": 2200, "lr": 1.8025282769762352e-05}
{"train_loss": 0.052979644387960434, "global_step": 195816, "epoch": 2200, "lr": 1.8024837072845486e-05}
{"train_loss": 0.10311366617679596, "global_step": 195817, "epoch": 2200, "lr": 1.8024391380227274e-05}
{"train_loss": 0.07427635043859482, "global_step": 195818, "epoch": 2200, "lr": 1.8023945691907813e-05}
{"train_loss": 0.059406887739896774, "global_step": 195819, "epoch": 2200, "lr": 1.8023500007887118e-05}
{"train_loss": 0.06478661298751831, "global_step": 195820, "epoch": 2200, "lr": 1.8023054328165295e-05}
{"train_loss": 0.06484231352806091, "global_step": 195821, "epoch": 2200, "lr": 1.8022608652742367e-05}
{"train_loss": 0.06720543652772903, "global_step": 195822, "epoch": 2200, "lr": 1.802216298161843e-05}
{"train_loss": 0.10928229987621307, "global_step": 195823, "epoch": 2200, "lr": 1.80217173147935e-05}
{"train_loss": 0.07099752873182297, "global_step": 195824, "epoch": 2200, "lr": 1.802127165226768e-05}
{"train_loss": 0.07160402834415436, "global_step": 195825, "epoch": 2200, "lr": 1.802082599404099e-05}
{"train_loss": 0.04913373291492462, "global_step": 195826, "epoch": 2200, "lr": 1.8020380340113533e-05}
{"train_loss": 0.0777311772108078, "global_step": 195827, "epoch": 2200, "lr": 1.8019934690485324e-05}
{"train_loss": 0.038038548082113266, "global_step": 195828, "epoch": 2200, "lr": 1.8019489045156467e-05}
{"train_loss": 0.09829650819301605, "global_step": 195829, "epoch": 2200, "lr": 1.8019043404126977e-05}
{"train_loss": 0.0664239451289177, "global_step": 195830, "epoch": 2200, "lr": 1.8018597767396945e-05}
{"train_loss": 0.03349096700549126, "global_step": 195831, "epoch": 2200, "lr": 1.801815213496643e-05}
{"train_loss": 0.07188349217176437, "global_step": 195832, "epoch": 2200, "lr": 1.8017706506835474e-05}
{"train_loss": 0.05522125959396362, "global_step": 195833, "epoch": 2200, "lr": 1.801726088300416e-05}
{"train_loss": 0.07243134081363678, "global_step": 195834, "epoch": 2200, "lr": 1.8016815263472527e-05}
{"train_loss": 0.08617354929447174, "global_step": 195835, "epoch": 2200, "lr": 1.8016369648240648e-05}
{"train_loss": 0.049391042441129684, "global_step": 195836, "epoch": 2200, "lr": 1.8015924037308573e-05}
{"train_loss": 0.0592208169400692, "global_step": 195837, "epoch": 2200, "lr": 1.801547843067638e-05}
{"train_loss": 0.11444300413131714, "global_step": 195838, "epoch": 2200, "lr": 1.8015032828344097e-05}
{"train_loss": 0.048075709491968155, "global_step": 195839, "epoch": 2200, "lr": 1.801458723031182e-05}
{"train_loss": 0.06396446377038956, "global_step": 195840, "epoch": 2200, "lr": 1.8014141636579578e-05}
{"train_loss": 0.04909345507621765, "global_step": 195841, "epoch": 2200, "lr": 1.801369604714746e-05}
{"train_loss": 0.11084719747304916, "global_step": 195842, "epoch": 2200, "lr": 1.8013250462015492e-05}
{"train_loss": 0.07644451409578323, "global_step": 195843, "epoch": 2200, "lr": 1.801280488118377e-05}
{"train_loss": 0.05157613381743431, "global_step": 195844, "epoch": 2200, "lr": 1.801235930465232e-05}
{"train_loss": 0.14074736833572388, "global_step": 195845, "epoch": 2200, "lr": 1.8011913732421236e-05}
{"train_loss": 0.07779470086097717, "global_step": 195846, "epoch": 2200, "lr": 1.8011468164490547e-05}
{"train_loss": 0.07849984616041183, "global_step": 195847, "epoch": 2200, "lr": 1.8011022600860322e-05}
{"train_loss": 0.044786904007196426, "global_step": 195848, "epoch": 2200, "lr": 1.801057704153064e-05}
{"train_loss": 0.07089431583881378, "global_step": 195849, "epoch": 2200, "lr": 1.8010131486501553e-05}
{"train_loss": 0.0668022483587265, "global_step": 195850, "epoch": 2200, "lr": 1.800968593577309e-05}
{"train_loss": 0.03360302001237869, "global_step": 195851, "epoch": 2200, "lr": 1.8009240389345355e-05}
{"train_loss": 0.09423467516899109, "global_step": 195852, "epoch": 2200, "lr": 1.800879484721837e-05}
{"train_loss": 0.045405492186546326, "global_step": 195853, "epoch": 2200, "lr": 1.800834930939223e-05}
{"train_loss": 0.08457942306995392, "global_step": 195854, "epoch": 2200, "lr": 1.800790377586696e-05}
{"train_loss": 0.053934015333652496, "global_step": 195855, "epoch": 2200, "lr": 1.8007458246642657e-05}
{"train_loss": 0.08079113066196442, "global_step": 195856, "epoch": 2200, "lr": 1.8007012721719358e-05}
{"train_loss": 0.01700402982532978, "global_step": 195857, "epoch": 2200, "lr": 1.800656720109711e-05}
{"train_loss": 0.05326317995786667, "global_step": 195858, "epoch": 2200, "lr": 1.8006121684776005e-05}
{"train_loss": 0.1273369938135147, "global_step": 195859, "epoch": 2200, "lr": 1.8005676172756077e-05}
{"train_loss": 0.01422647014260292, "global_step": 195860, "epoch": 2200, "lr": 1.8005230665037402e-05}
{"train_loss": 0.058036670088768005, "global_step": 195861, "epoch": 2200, "lr": 1.8004785161620025e-05}
{"train_loss": 0.05410851910710335, "global_step": 195862, "epoch": 2200, "lr": 1.8004339662504028e-05}
{"train_loss": 0.04737285152077675, "global_step": 195863, "epoch": 2200, "lr": 1.800389416768944e-05}
{"train_loss": 0.05222544074058533, "global_step": 195864, "epoch": 2200, "lr": 1.8003448677176355e-05}
{"train_loss": 0.06622236967086792, "global_step": 195865, "epoch": 2200, "lr": 1.80030031909648e-05}
{"train_loss": 0.07500584423542023, "global_step": 195866, "epoch": 2200, "lr": 1.8002557709054856e-05}
{"train_loss": 0.06018870323896408, "global_step": 195867, "epoch": 2200, "lr": 1.8002112231446587e-05}
{"train_loss": 0.05413436517119408, "global_step": 195868, "epoch": 2200, "lr": 1.8001666758140036e-05}
{"train_loss": 0.02483634650707245, "global_step": 195869, "epoch": 2200, "lr": 1.8001221289135283e-05}
{"train_loss": 0.08912630379199982, "global_step": 195870, "epoch": 2200, "lr": 1.800077582443236e-05}
{"train_loss": 0.05793149396777153, "global_step": 195871, "epoch": 2200, "lr": 1.800033036403136e-05}
{"train_loss": 0.11816994845867157, "global_step": 195872, "epoch": 2200, "lr": 1.799988490793231e-05}
{"train_loss": 0.04387566074728966, "global_step": 195873, "epoch": 2200, "lr": 1.799943945613529e-05}
{"train_loss": 0.045985810458660126, "global_step": 195874, "epoch": 2200, "lr": 1.799899400864035e-05}
{"train_loss": 0.04204757139086723, "global_step": 195875, "epoch": 2200, "lr": 1.799854856544757e-05}
{"train_loss": 0.13361917436122894, "global_step": 195876, "epoch": 2200, "lr": 1.799810312655698e-05}
{"train_loss": 0.0665917918086052, "global_step": 195877, "epoch": 2200, "lr": 1.7997657691968673e-05}
{"train_loss": 0.04750106856226921, "global_step": 195878, "epoch": 2200, "lr": 1.799721226168267e-05}
{"train_loss": 0.08577708899974823, "global_step": 195879, "epoch": 2200, "lr": 1.799676683569907e-05}
{"train_loss": 0.03302837908267975, "global_step": 195880, "epoch": 2200, "lr": 1.7996321414017897e-05}
{"train_loss": 0.037191201001405716, "global_step": 195881, "epoch": 2200, "lr": 1.799587599663925e-05}
{"train_loss": 0.027404632419347763, "global_step": 195882, "epoch": 2200, "lr": 1.7995430583563145e-05}
{"train_loss": 0.10355349630117416, "global_step": 195883, "epoch": 2200, "lr": 1.7994985174789662e-05}
{"train_loss": 0.04820691794157028, "global_step": 195884, "epoch": 2200, "lr": 1.7994539770318886e-05}
{"train_loss": 0.11413147300481796, "global_step": 195885, "epoch": 2200, "lr": 1.7994094370150838e-05}
{"train_loss": 0.05827156826853752, "global_step": 195886, "epoch": 2200, "lr": 1.7993648974285605e-05}
{"train_loss": 0.14708155393600464, "global_step": 195887, "epoch": 2200, "lr": 1.7993203582723222e-05}
{"train_loss": 0.06566537037659227, "global_step": 195888, "epoch": 2200, "lr": 1.7992758195463783e-05, "train/sim_max_reward_0": 0.6956634898026027, "train/sim_max_reward_1": 0.48783224038882333, "train/sim_max_reward_2": 1.0, "train/sim_max_reward_3": 0.6875772224229227, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9050827451189237, "test/sim_max_reward_4500000": 0.4162465191696434, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9792645647488469, "test/sim_max_reward_4500003": 0.6774841972924788, "test/sim_max_reward_4500004": 0.9332353784330408, "test/sim_max_reward_4500005": 0.3926331832961536, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.901124922399256, "test/sim_max_reward_4500008": 0.9807078765254196, "test/sim_max_reward_4500009": 0.04925046575540526, "test/sim_max_reward_4500010": 0.4150291629895001, "test/sim_max_reward_4500011": 0.9911379791447296, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5469544127656806, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9613296390743086, "test/sim_max_reward_4500016": 0.9010392996268874, "test/sim_max_reward_4500017": 0.6540845032474135, "test/sim_max_reward_4500018": 0.9884544110175513, "test/sim_max_reward_4500019": 0.9692027862900664, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9960755849304155, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 0.6585542896351235, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.7317386163160633, "test/sim_max_reward_4500026": 0.32389631669609964, "test/sim_max_reward_4500027": 0.902344342793558, "test/sim_max_reward_4500028": 0.9952585512088101, "test/sim_max_reward_4500029": 0.044627909804164125, "test/sim_max_reward_4500030": 0.9461668798799803, "test/sim_max_reward_4500031": 0.5174287890942787, "test/sim_max_reward_4500032": 0.5544195064417565, "test/sim_max_reward_4500033": 0.9878027490390637, "test/sim_max_reward_4500034": 0.40154650890977117, "test/sim_max_reward_4500035": 0.40816864122035257, "test/sim_max_reward_4500036": 0.13216136728466063, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.1625169021945696, "test/sim_max_reward_4500039": 0.9730414103400249, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6094886364915427, "test/sim_max_reward_4500042": 0.04626333346952088, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.013838589474767338, "test/sim_max_reward_4500045": 0.9951394270290231, "test/sim_max_reward_4500046": 0.3960986517862417, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.996277966618993, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7960259496222122, "test/mean_score": 0.6510006854487034, "val_loss": 8.160477638244629, "train_action_mse_error": 22.95933723449707}
{"train_loss": 0.030897922813892365, "global_step": 195889, "epoch": 2201, "lr": 1.7992312812507322e-05}
{"train_loss": 0.08420243114233017, "global_step": 195890, "epoch": 2201, "lr": 1.7991867433853887e-05}
{"train_loss": 0.10591929405927658, "global_step": 195891, "epoch": 2201, "lr": 1.7991422059503575e-05}
{"train_loss": 0.06614409387111664, "global_step": 195892, "epoch": 2201, "lr": 1.799097668945641e-05}
{"train_loss": 0.10117800533771515, "global_step": 195893, "epoch": 2201, "lr": 1.7990531323712484e-05}
{"train_loss": 0.053844183683395386, "global_step": 195894, "epoch": 2201, "lr": 1.7990085962271842e-05}
{"train_loss": 0.056809645146131516, "global_step": 195895, "epoch": 2201, "lr": 1.798964060513452e-05}
{"train_loss": 0.07441660761833191, "global_step": 195896, "epoch": 2201, "lr": 1.798919525230063e-05}
{"train_loss": 0.1219215914607048, "global_step": 195897, "epoch": 2201, "lr": 1.7988749903770175e-05}
{"train_loss": 0.025025347247719765, "global_step": 195898, "epoch": 2201, "lr": 1.7988304559543263e-05}
{"train_loss": 0.050150029361248016, "global_step": 195899, "epoch": 2201, "lr": 1.798785921961991e-05}
{"train_loss": 0.03739795833826065, "global_step": 195900, "epoch": 2201, "lr": 1.7987413884000225e-05}
{"train_loss": 0.05411658436059952, "global_step": 195901, "epoch": 2201, "lr": 1.7986968552684226e-05}
{"train_loss": 0.03494966775178909, "global_step": 195902, "epoch": 2201, "lr": 1.798652322567198e-05}
{"train_loss": 0.012615105137228966, "global_step": 195903, "epoch": 2201, "lr": 1.7986077902963577e-05}
{"train_loss": 0.09191524982452393, "global_step": 195904, "epoch": 2201, "lr": 1.7985632584559043e-05}
{"train_loss": 0.03538918122649193, "global_step": 195905, "epoch": 2201, "lr": 1.798518727045846e-05}
{"train_loss": 0.06514444947242737, "global_step": 195906, "epoch": 2201, "lr": 1.7984741960661864e-05}
{"train_loss": 0.10456440597772598, "global_step": 195907, "epoch": 2201, "lr": 1.798429665516935e-05}
{"train_loss": 0.06656117737293243, "global_step": 195908, "epoch": 2201, "lr": 1.7983851353980935e-05}
{"train_loss": 0.07088425010442734, "global_step": 195909, "epoch": 2201, "lr": 1.798340605709672e-05}
{"train_loss": 0.0704081654548645, "global_step": 195910, "epoch": 2201, "lr": 1.7982960764516733e-05}
{"train_loss": 0.046044666320085526, "global_step": 195911, "epoch": 2201, "lr": 1.7982515476241056e-05}
{"train_loss": 0.11162900924682617, "global_step": 195912, "epoch": 2201, "lr": 1.798207019226973e-05}
{"train_loss": 0.06866146624088287, "global_step": 195913, "epoch": 2201, "lr": 1.7981624912602835e-05}
{"train_loss": 0.09654884040355682, "global_step": 195914, "epoch": 2201, "lr": 1.798117963724041e-05}
{"train_loss": 0.044358231127262115, "global_step": 195915, "epoch": 2201, "lr": 1.7980734366182533e-05}
{"train_loss": 0.05831364542245865, "global_step": 195916, "epoch": 2201, "lr": 1.798028909942924e-05}
{"train_loss": 0.053108058869838715, "global_step": 195917, "epoch": 2201, "lr": 1.7979843836980632e-05}
{"train_loss": 0.04772600904107094, "global_step": 195918, "epoch": 2201, "lr": 1.7979398578836726e-05}
{"train_loss": 0.05873589962720871, "global_step": 195919, "epoch": 2201, "lr": 1.797895332499761e-05}
{"train_loss": 0.07241447269916534, "global_step": 195920, "epoch": 2201, "lr": 1.7978508075463323e-05}
{"train_loss": 0.054015401750802994, "global_step": 195921, "epoch": 2201, "lr": 1.7978062830233932e-05}
{"train_loss": 0.11949414759874344, "global_step": 195922, "epoch": 2201, "lr": 1.7977617589309514e-05}
{"train_loss": 0.08299887180328369, "global_step": 195923, "epoch": 2201, "lr": 1.7977172352690104e-05}
{"train_loss": 0.030373185873031616, "global_step": 195924, "epoch": 2201, "lr": 1.7976727120375786e-05}
{"train_loss": 0.08651816844940186, "global_step": 195925, "epoch": 2201, "lr": 1.7976281892366592e-05}
{"train_loss": 0.03800409287214279, "global_step": 195926, "epoch": 2201, "lr": 1.797583666866261e-05}
{"train_loss": 0.04183606803417206, "global_step": 195927, "epoch": 2201, "lr": 1.797539144926389e-05}
{"train_loss": 0.061539482325315475, "global_step": 195928, "epoch": 2201, "lr": 1.797494623417047e-05}
{"train_loss": 0.04824519902467728, "global_step": 195929, "epoch": 2201, "lr": 1.7974501023382444e-05}
{"train_loss": 0.05720260366797447, "global_step": 195930, "epoch": 2201, "lr": 1.7974055816899843e-05}
{"train_loss": 0.026534315198659897, "global_step": 195931, "epoch": 2201, "lr": 1.797361061472275e-05}
{"train_loss": 0.06624497473239899, "global_step": 195932, "epoch": 2201, "lr": 1.7973165416851222e-05}
{"train_loss": 0.04450925439596176, "global_step": 195933, "epoch": 2201, "lr": 1.7972720223285288e-05}
{"train_loss": 0.056648045778274536, "global_step": 195934, "epoch": 2201, "lr": 1.7972275034025055e-05}
{"train_loss": 0.07183051109313965, "global_step": 195935, "epoch": 2201, "lr": 1.7971829849070538e-05}
{"train_loss": 0.10603279620409012, "global_step": 195936, "epoch": 2201, "lr": 1.7971384668421835e-05}
{"train_loss": 0.07093643397092819, "global_step": 195937, "epoch": 2201, "lr": 1.7970939492078974e-05}
{"train_loss": 0.0708952397108078, "global_step": 195938, "epoch": 2201, "lr": 1.797049432004203e-05}
{"train_loss": 0.10427938401699066, "global_step": 195939, "epoch": 2201, "lr": 1.797004915231108e-05}
{"train_loss": 0.029558144509792328, "global_step": 195940, "epoch": 2201, "lr": 1.7969603988886148e-05}
{"train_loss": 0.04587734118103981, "global_step": 195941, "epoch": 2201, "lr": 1.7969158829767328e-05}
{"train_loss": 0.06888648122549057, "global_step": 195942, "epoch": 2201, "lr": 1.7968713674954652e-05}
{"train_loss": 0.06360107660293579, "global_step": 195943, "epoch": 2201, "lr": 1.79682685244482e-05}
{"train_loss": 0.07354911416769028, "global_step": 195944, "epoch": 2201, "lr": 1.7967823378248018e-05}
{"train_loss": 0.08369331806898117, "global_step": 195945, "epoch": 2201, "lr": 1.7967378236354183e-05}
{"train_loss": 0.08364164084196091, "global_step": 195946, "epoch": 2201, "lr": 1.7966933098766724e-05}
{"train_loss": 0.06743890792131424, "global_step": 195947, "epoch": 2201, "lr": 1.796648796548574e-05}
{"train_loss": 0.08485327661037445, "global_step": 195948, "epoch": 2201, "lr": 1.7966042836511255e-05}
{"train_loss": 0.043578002601861954, "global_step": 195949, "epoch": 2201, "lr": 1.7965597711843363e-05}
{"train_loss": 0.06361573934555054, "global_step": 195950, "epoch": 2201, "lr": 1.7965152591482088e-05}
{"train_loss": 0.0490720272064209, "global_step": 195951, "epoch": 2201, "lr": 1.796470747542752e-05}
{"train_loss": 0.04238373041152954, "global_step": 195952, "epoch": 2201, "lr": 1.7964262363679697e-05}
{"train_loss": 0.0943434089422226, "global_step": 195953, "epoch": 2201, "lr": 1.79638172562387e-05}
{"train_loss": 0.05770585685968399, "global_step": 195954, "epoch": 2201, "lr": 1.7963372153104564e-05}
{"train_loss": 0.048403773456811905, "global_step": 195955, "epoch": 2201, "lr": 1.7962927054277378e-05}
{"train_loss": 0.04631805419921875, "global_step": 195956, "epoch": 2201, "lr": 1.796248195975717e-05}
{"train_loss": 0.045404594391584396, "global_step": 195957, "epoch": 2201, "lr": 1.7962036869544014e-05}
{"train_loss": 0.076771080493927, "global_step": 195958, "epoch": 2201, "lr": 1.796159178363799e-05}
{"train_loss": 0.08286663889884949, "global_step": 195959, "epoch": 2201, "lr": 1.7961146702039124e-05}
{"train_loss": 0.05515410751104355, "global_step": 195960, "epoch": 2201, "lr": 1.7960701624747506e-05}
{"train_loss": 0.04850901663303375, "global_step": 195961, "epoch": 2201, "lr": 1.7960256551763162e-05}
{"train_loss": 0.07528343796730042, "global_step": 195962, "epoch": 2201, "lr": 1.795981148308619e-05}
{"train_loss": 0.08807211369276047, "global_step": 195963, "epoch": 2201, "lr": 1.7959366418716616e-05}
{"train_loss": 0.08138873428106308, "global_step": 195964, "epoch": 2201, "lr": 1.795892135865453e-05}
{"train_loss": 0.07162138819694519, "global_step": 195965, "epoch": 2201, "lr": 1.7958476302899973e-05}
{"train_loss": 0.0612148754298687, "global_step": 195966, "epoch": 2201, "lr": 1.7958031251452994e-05}
{"train_loss": 0.09856616705656052, "global_step": 195967, "epoch": 2201, "lr": 1.7957586204313685e-05}
{"train_loss": 0.04936433210968971, "global_step": 195968, "epoch": 2201, "lr": 1.795714116148207e-05}
{"train_loss": 0.07351522892713547, "global_step": 195969, "epoch": 2201, "lr": 1.795669612295825e-05}
{"train_loss": 0.09514252841472626, "global_step": 195970, "epoch": 2201, "lr": 1.7956251088742238e-05}
{"train_loss": 0.06972471624612808, "global_step": 195971, "epoch": 2201, "lr": 1.7955806058834134e-05}
{"train_loss": 0.0451461486518383, "global_step": 195972, "epoch": 2201, "lr": 1.7955361033233986e-05}
{"train_loss": 0.04742075502872467, "global_step": 195973, "epoch": 2201, "lr": 1.795491601194183e-05}
{"train_loss": 0.08474951237440109, "global_step": 195974, "epoch": 2201, "lr": 1.7954470994957738e-05}
{"train_loss": 0.05391949042677879, "global_step": 195975, "epoch": 2201, "lr": 1.79540259822818e-05}
{"train_loss": 0.0748833417892456, "global_step": 195976, "epoch": 2201, "lr": 1.795358097391404e-05}
{"train_loss": 0.06506770388798767, "global_step": 195977, "epoch": 2201, "lr": 1.7953135969854546e-05, "val_loss": 8.143539428710938}
{"train_loss": 0.06544024497270584, "global_step": 195978, "epoch": 2202, "lr": 1.7952690970103336e-05}
{"train_loss": 0.045565731823444366, "global_step": 195979, "epoch": 2202, "lr": 1.795224597466052e-05}
{"train_loss": 0.042984526604413986, "global_step": 195980, "epoch": 2202, "lr": 1.7951800983526116e-05}
{"train_loss": 0.04831438511610031, "global_step": 195981, "epoch": 2202, "lr": 1.795135599670022e-05}
{"train_loss": 0.06309860199689865, "global_step": 195982, "epoch": 2202, "lr": 1.795091101418286e-05}
{"train_loss": 0.06604397296905518, "global_step": 195983, "epoch": 2202, "lr": 1.7950466035974117e-05}
{"train_loss": 0.020569920539855957, "global_step": 195984, "epoch": 2202, "lr": 1.7950021062074034e-05}
{"train_loss": 0.049829401075839996, "global_step": 195985, "epoch": 2202, "lr": 1.7949576092482695e-05}
{"train_loss": 0.1000421792268753, "global_step": 195986, "epoch": 2202, "lr": 1.7949131127200126e-05}
{"train_loss": 0.07695477455854416, "global_step": 195987, "epoch": 2202, "lr": 1.7948686166226426e-05}
{"train_loss": 0.08703147619962692, "global_step": 195988, "epoch": 2202, "lr": 1.794824120956161e-05}
{"train_loss": 0.04160391166806221, "global_step": 195989, "epoch": 2202, "lr": 1.794779625720579e-05}
{"train_loss": 0.041669249534606934, "global_step": 195990, "epoch": 2202, "lr": 1.7947351309158972e-05}
{"train_loss": 0.05132703483104706, "global_step": 195991, "epoch": 2202, "lr": 1.794690636542126e-05}
{"train_loss": 0.0421917587518692, "global_step": 195992, "epoch": 2202, "lr": 1.7946461425992688e-05}
{"train_loss": 0.08398228883743286, "global_step": 195993, "epoch": 2202, "lr": 1.7946016490873314e-05}
{"train_loss": 0.08530405163764954, "global_step": 195994, "epoch": 2202, "lr": 1.794557156006323e-05}
{"train_loss": 0.11070233583450317, "global_step": 195995, "epoch": 2202, "lr": 1.7945126633562453e-05}
{"train_loss": 0.08660757541656494, "global_step": 195996, "epoch": 2202, "lr": 1.7944681711371082e-05}
{"train_loss": 0.05411652848124504, "global_step": 195997, "epoch": 2202, "lr": 1.7944236793489137e-05}
{"train_loss": 0.04924357682466507, "global_step": 195998, "epoch": 2202, "lr": 1.7943791879916722e-05}
{"train_loss": 0.09740199148654938, "global_step": 195999, "epoch": 2202, "lr": 1.7943346970653852e-05}
{"train_loss": 0.04585089907050133, "global_step": 196000, "epoch": 2202, "lr": 1.794290206570062e-05}
{"train_loss": 0.08427409827709198, "global_step": 196001, "epoch": 2202, "lr": 1.7942457165057064e-05}
{"train_loss": 0.10663633793592453, "global_step": 196002, "epoch": 2202, "lr": 1.7942012268723274e-05}
{"train_loss": 0.0901251956820488, "global_step": 196003, "epoch": 2202, "lr": 1.794156737669927e-05}
{"train_loss": 0.10165511071681976, "global_step": 196004, "epoch": 2202, "lr": 1.7941122488985145e-05}
{"train_loss": 0.061215173453092575, "global_step": 196005, "epoch": 2202, "lr": 1.7940677605580947e-05}
{"train_loss": 0.0936642438173294, "global_step": 196006, "epoch": 2202, "lr": 1.7940232726486717e-05}
{"train_loss": 0.09040240198373795, "global_step": 196007, "epoch": 2202, "lr": 1.793978785170255e-05}
{"train_loss": 0.09182173758745193, "global_step": 196008, "epoch": 2202, "lr": 1.793934298122847e-05}
{"train_loss": 0.09756383299827576, "global_step": 196009, "epoch": 2202, "lr": 1.7938898115064574e-05}
{"train_loss": 0.12981250882148743, "global_step": 196010, "epoch": 2202, "lr": 1.793845325321088e-05}
{"train_loss": 0.08670002967119217, "global_step": 196011, "epoch": 2202, "lr": 1.7938008395667493e-05}
{"train_loss": 0.05747722461819649, "global_step": 196012, "epoch": 2202, "lr": 1.7937563542434432e-05}
{"train_loss": 0.0435895137488842, "global_step": 196013, "epoch": 2202, "lr": 1.793711869351179e-05}
{"train_loss": 0.04358197748661041, "global_step": 196014, "epoch": 2202, "lr": 1.7936673848899594e-05}
{"train_loss": 0.058075085282325745, "global_step": 196015, "epoch": 2202, "lr": 1.793622900859794e-05}
{"train_loss": 0.0604223906993866, "global_step": 196016, "epoch": 2202, "lr": 1.793578417260685e-05}
{"train_loss": 0.053480204194784164, "global_step": 196017, "epoch": 2202, "lr": 1.7935339340926422e-05}
{"train_loss": 0.0881526917219162, "global_step": 196018, "epoch": 2202, "lr": 1.7934894513556676e-05}
{"train_loss": 0.05646149814128876, "global_step": 196019, "epoch": 2202, "lr": 1.7934449690497718e-05}
{"train_loss": 0.03226931393146515, "global_step": 196020, "epoch": 2202, "lr": 1.793400487174956e-05}
{"train_loss": 0.08975131064653397, "global_step": 196021, "epoch": 2202, "lr": 1.79335600573123e-05}
{"train_loss": 0.05663555487990379, "global_step": 196022, "epoch": 2202, "lr": 1.7933115247185966e-05}
{"train_loss": 0.0799407958984375, "global_step": 196023, "epoch": 2202, "lr": 1.7932670441370654e-05}
{"train_loss": 0.07767443358898163, "global_step": 196024, "epoch": 2202, "lr": 1.7932225639866384e-05}
{"train_loss": 0.034312255680561066, "global_step": 196025, "epoch": 2202, "lr": 1.793178084267325e-05}
{"train_loss": 0.0466676726937294, "global_step": 196026, "epoch": 2202, "lr": 1.7931336049791288e-05}
{"train_loss": 0.044549621641635895, "global_step": 196027, "epoch": 2202, "lr": 1.7930891261220573e-05}
{"train_loss": 0.08810868114233017, "global_step": 196028, "epoch": 2202, "lr": 1.793044647696115e-05}
{"train_loss": 0.0502304770052433, "global_step": 196029, "epoch": 2202, "lr": 1.7930001697013084e-05}
{"train_loss": 0.033637482672929764, "global_step": 196030, "epoch": 2202, "lr": 1.7929556921376455e-05}
{"train_loss": 0.10584653913974762, "global_step": 196031, "epoch": 2202, "lr": 1.792911215005129e-05}
{"train_loss": 0.07917876541614532, "global_step": 196032, "epoch": 2202, "lr": 1.7928667383037684e-05}
{"train_loss": 0.04821940138936043, "global_step": 196033, "epoch": 2202, "lr": 1.792822262033566e-05}
{"train_loss": 0.03160111978650093, "global_step": 196034, "epoch": 2202, "lr": 1.792777786194531e-05}
{"train_loss": 0.06264273822307587, "global_step": 196035, "epoch": 2202, "lr": 1.7927333107866668e-05}
{"train_loss": 0.036675550043582916, "global_step": 196036, "epoch": 2202, "lr": 1.792688835809982e-05}
{"train_loss": 0.0730222761631012, "global_step": 196037, "epoch": 2202, "lr": 1.7926443612644788e-05}
{"train_loss": 0.1315855085849762, "global_step": 196038, "epoch": 2202, "lr": 1.7925998871501675e-05}
{"train_loss": 0.05343367159366608, "global_step": 196039, "epoch": 2202, "lr": 1.7925554134670508e-05}
{"train_loss": 0.03409280627965927, "global_step": 196040, "epoch": 2202, "lr": 1.7925109402151376e-05}
{"train_loss": 0.0974719449877739, "global_step": 196041, "epoch": 2202, "lr": 1.79246646739443e-05}
{"train_loss": 0.06525790691375732, "global_step": 196042, "epoch": 2202, "lr": 1.792421995004938e-05}
{"train_loss": 0.05357516184449196, "global_step": 196043, "epoch": 2202, "lr": 1.792377523046666e-05}
{"train_loss": 0.05453617498278618, "global_step": 196044, "epoch": 2202, "lr": 1.792333051519618e-05}
{"train_loss": 0.11115724593400955, "global_step": 196045, "epoch": 2202, "lr": 1.7922885804238034e-05}
{"train_loss": 0.07940821349620819, "global_step": 196046, "epoch": 2202, "lr": 1.7922441097592248e-05}
{"train_loss": 0.03703039139509201, "global_step": 196047, "epoch": 2202, "lr": 1.792199639525891e-05}
{"train_loss": 0.04315999150276184, "global_step": 196048, "epoch": 2202, "lr": 1.7921551697238058e-05}
{"train_loss": 0.08140642195940018, "global_step": 196049, "epoch": 2202, "lr": 1.792110700352978e-05}
{"train_loss": 0.06271076947450638, "global_step": 196050, "epoch": 2202, "lr": 1.79206623141341e-05}
{"train_loss": 0.05358486622571945, "global_step": 196051, "epoch": 2202, "lr": 1.792021762905111e-05}
{"train_loss": 0.03896396979689598, "global_step": 196052, "epoch": 2202, "lr": 1.7919772948280843e-05}
{"train_loss": 0.060960203409194946, "global_step": 196053, "epoch": 2202, "lr": 1.791932827182339e-05}
{"train_loss": 0.022130023688077927, "global_step": 196054, "epoch": 2202, "lr": 1.7918883599678766e-05}
{"train_loss": 0.07612427324056625, "global_step": 196055, "epoch": 2202, "lr": 1.791843893184708e-05}
{"train_loss": 0.04828111082315445, "global_step": 196056, "epoch": 2202, "lr": 1.7917994268328353e-05}
{"train_loss": 0.059367403388023376, "global_step": 196057, "epoch": 2202, "lr": 1.791754960912268e-05}
{"train_loss": 0.05462663993239403, "global_step": 196058, "epoch": 2202, "lr": 1.791710495423008e-05}
{"train_loss": 0.07142653316259384, "global_step": 196059, "epoch": 2202, "lr": 1.791666030365065e-05}
{"train_loss": 0.07029057294130325, "global_step": 196060, "epoch": 2202, "lr": 1.7916215657384423e-05}
{"train_loss": 0.08707643300294876, "global_step": 196061, "epoch": 2202, "lr": 1.7915771015431483e-05}
{"train_loss": 0.04453349858522415, "global_step": 196062, "epoch": 2202, "lr": 1.791532637779186e-05}
{"train_loss": 0.06511639803647995, "global_step": 196063, "epoch": 2202, "lr": 1.7914881744465645e-05}
{"train_loss": 0.06302239745855331, "global_step": 196064, "epoch": 2202, "lr": 1.7914437115452864e-05}
{"train_loss": 0.029826601967215538, "global_step": 196065, "epoch": 2202, "lr": 1.79139924907536e-05}
{"train_loss": 0.06529851046422225, "global_step": 196066, "epoch": 2202, "lr": 1.7913547870367925e-05, "val_loss": 8.0318603515625}
{"train_loss": 0.09612838178873062, "global_step": 196067, "epoch": 2203, "lr": 1.7913103254295867e-05}
{"train_loss": 0.036793678998947144, "global_step": 196068, "epoch": 2203, "lr": 1.7912658642537517e-05}
{"train_loss": 0.117581307888031, "global_step": 196069, "epoch": 2203, "lr": 1.79122140350929e-05}
{"train_loss": 0.06207204982638359, "global_step": 196070, "epoch": 2203, "lr": 1.7911769431962116e-05}
{"train_loss": 0.0197746641933918, "global_step": 196071, "epoch": 2203, "lr": 1.7911324833145183e-05}
{"train_loss": 0.09580080211162567, "global_step": 196072, "epoch": 2203, "lr": 1.7910880238642202e-05}
{"train_loss": 0.11625558882951736, "global_step": 196073, "epoch": 2203, "lr": 1.7910435648453194e-05}
{"train_loss": 0.08970165997743607, "global_step": 196074, "epoch": 2203, "lr": 1.790999106257825e-05}
{"train_loss": 0.09554050117731094, "global_step": 196075, "epoch": 2203, "lr": 1.7909546481017405e-05}
{"train_loss": 0.11164310574531555, "global_step": 196076, "epoch": 2203, "lr": 1.790910190377074e-05}
{"train_loss": 0.057885389775037766, "global_step": 196077, "epoch": 2203, "lr": 1.7908657330838292e-05}
{"train_loss": 0.07571941614151001, "global_step": 196078, "epoch": 2203, "lr": 1.790821276222015e-05}
{"train_loss": 0.07523631304502487, "global_step": 196079, "epoch": 2203, "lr": 1.7907768197916347e-05}
{"train_loss": 0.09837989509105682, "global_step": 196080, "epoch": 2203, "lr": 1.7907323637926964e-05}
{"train_loss": 0.01959841698408127, "global_step": 196081, "epoch": 2203, "lr": 1.7906879082252048e-05}
{"train_loss": 0.045349206775426865, "global_step": 196082, "epoch": 2203, "lr": 1.7906434530891647e-05}
{"train_loss": 0.08310733735561371, "global_step": 196083, "epoch": 2203, "lr": 1.7905989983845856e-05}
{"train_loss": 0.028351260349154472, "global_step": 196084, "epoch": 2203, "lr": 1.790554544111469e-05}
{"train_loss": 0.07869354635477066, "global_step": 196085, "epoch": 2203, "lr": 1.790510090269825e-05}
{"train_loss": 0.03245961666107178, "global_step": 196086, "epoch": 2203, "lr": 1.7904656368596566e-05}
{"train_loss": 0.02681371569633484, "global_step": 196087, "epoch": 2203, "lr": 1.7904211838809725e-05}
{"train_loss": 0.054848749190568924, "global_step": 196088, "epoch": 2203, "lr": 1.790376731333775e-05}
{"train_loss": 0.07354475557804108, "global_step": 196089, "epoch": 2203, "lr": 1.7903322792180742e-05}
{"train_loss": 0.04414282739162445, "global_step": 196090, "epoch": 2203, "lr": 1.7902878275338726e-05}
{"train_loss": 0.08548605442047119, "global_step": 196091, "epoch": 2203, "lr": 1.7902433762811794e-05}
{"train_loss": 0.07351233065128326, "global_step": 196092, "epoch": 2203, "lr": 1.790198925459997e-05}
{"train_loss": 0.07144569605588913, "global_step": 196093, "epoch": 2203, "lr": 1.7901544750703348e-05}
{"train_loss": 0.05423682928085327, "global_step": 196094, "epoch": 2203, "lr": 1.7901100251121956e-05}
{"train_loss": 0.07611122727394104, "global_step": 196095, "epoch": 2203, "lr": 1.7900655755855888e-05}
{"train_loss": 0.07591118663549423, "global_step": 196096, "epoch": 2203, "lr": 1.7900211264905165e-05}
{"train_loss": 0.04802779480814934, "global_step": 196097, "epoch": 2203, "lr": 1.789976677826989e-05}
{"train_loss": 0.08553513884544373, "global_step": 196098, "epoch": 2203, "lr": 1.7899322295950076e-05}
{"train_loss": 0.09865562617778778, "global_step": 196099, "epoch": 2203, "lr": 1.7898877817945832e-05}
{"train_loss": 0.1109299585223198, "global_step": 196100, "epoch": 2203, "lr": 1.789843334425717e-05}
{"train_loss": 0.10132209956645966, "global_step": 196101, "epoch": 2203, "lr": 1.7897988874884174e-05}
{"train_loss": 0.09402459859848022, "global_step": 196102, "epoch": 2203, "lr": 1.7897544409826923e-05}
{"train_loss": 0.042912308126688004, "global_step": 196103, "epoch": 2203, "lr": 1.7897099949085434e-05}
{"train_loss": 0.0664600133895874, "global_step": 196104, "epoch": 2203, "lr": 1.78966554926598e-05}
{"train_loss": 0.05324994772672653, "global_step": 196105, "epoch": 2203, "lr": 1.7896211040550066e-05}
{"train_loss": 0.08414743840694427, "global_step": 196106, "epoch": 2203, "lr": 1.7895766592756302e-05}
{"train_loss": 0.09553791582584381, "global_step": 196107, "epoch": 2203, "lr": 1.7895322149278547e-05}
{"train_loss": 0.05662126839160919, "global_step": 196108, "epoch": 2203, "lr": 1.7894877710116892e-05}
{"train_loss": 0.05288698896765709, "global_step": 196109, "epoch": 2203, "lr": 1.7894433275271362e-05}
{"train_loss": 0.06566479057073593, "global_step": 196110, "epoch": 2203, "lr": 1.7893988844742047e-05}
{"train_loss": 0.06989619135856628, "global_step": 196111, "epoch": 2203, "lr": 1.7893544418528985e-05}
{"train_loss": 0.09310884773731232, "global_step": 196112, "epoch": 2203, "lr": 1.7893099996632258e-05}
{"train_loss": 0.1068221777677536, "global_step": 196113, "epoch": 2203, "lr": 1.7892655579051898e-05}
{"train_loss": 0.0871797651052475, "global_step": 196114, "epoch": 2203, "lr": 1.789221116578799e-05}
{"train_loss": 0.06141320616006851, "global_step": 196115, "epoch": 2203, "lr": 1.789176675684057e-05}
{"train_loss": 0.06521189212799072, "global_step": 196116, "epoch": 2203, "lr": 1.7891322352209726e-05}
{"train_loss": 0.07969675213098526, "global_step": 196117, "epoch": 2203, "lr": 1.7890877951895486e-05}
{"train_loss": 0.026154402643442154, "global_step": 196118, "epoch": 2203, "lr": 1.7890433555897946e-05}
{"train_loss": 0.07974991202354431, "global_step": 196119, "epoch": 2203, "lr": 1.7889989164217123e-05}
{"train_loss": 0.14378340542316437, "global_step": 196120, "epoch": 2203, "lr": 1.788954477685312e-05}
{"train_loss": 0.03219540789723396, "global_step": 196121, "epoch": 2203, "lr": 1.7889100393805974e-05}
{"train_loss": 0.12091827392578125, "global_step": 196122, "epoch": 2203, "lr": 1.7888656015075733e-05}
{"train_loss": 0.06244094297289848, "global_step": 196123, "epoch": 2203, "lr": 1.788821164066249e-05}
{"train_loss": 0.1387515664100647, "global_step": 196124, "epoch": 2203, "lr": 1.7887767270566262e-05}
{"train_loss": 0.08553896099328995, "global_step": 196125, "epoch": 2203, "lr": 1.788732290478715e-05}
{"train_loss": 0.028638949617743492, "global_step": 196126, "epoch": 2203, "lr": 1.7886878543325187e-05}
{"train_loss": 0.06097858399152756, "global_step": 196127, "epoch": 2203, "lr": 1.7886434186180452e-05}
{"train_loss": 0.062405940145254135, "global_step": 196128, "epoch": 2203, "lr": 1.788598983335298e-05}
{"train_loss": 0.06800013780593872, "global_step": 196129, "epoch": 2203, "lr": 1.7885545484842858e-05}
{"train_loss": 0.07166192680597305, "global_step": 196130, "epoch": 2203, "lr": 1.7885101140650124e-05}
{"train_loss": 0.03768087550997734, "global_step": 196131, "epoch": 2203, "lr": 1.7884656800774856e-05}
{"train_loss": 0.055045001208782196, "global_step": 196132, "epoch": 2203, "lr": 1.788421246521709e-05}
{"train_loss": 0.03370237722992897, "global_step": 196133, "epoch": 2203, "lr": 1.7883768133976918e-05}
{"train_loss": 0.04555121064186096, "global_step": 196134, "epoch": 2203, "lr": 1.7883323807054363e-05}
{"train_loss": 0.07698741555213928, "global_step": 196135, "epoch": 2203, "lr": 1.788287948444952e-05}
{"train_loss": 0.05235074833035469, "global_step": 196136, "epoch": 2203, "lr": 1.788243516616242e-05}
{"train_loss": 0.02369091659784317, "global_step": 196137, "epoch": 2203, "lr": 1.7881990852193138e-05}
{"train_loss": 0.039575230330228806, "global_step": 196138, "epoch": 2203, "lr": 1.7881546542541737e-05}
{"train_loss": 0.05554421991109848, "global_step": 196139, "epoch": 2203, "lr": 1.7881102237208263e-05}
{"train_loss": 0.11784690618515015, "global_step": 196140, "epoch": 2203, "lr": 1.7880657936192796e-05}
{"train_loss": 0.04093228280544281, "global_step": 196141, "epoch": 2203, "lr": 1.7880213639495365e-05}
{"train_loss": 0.07335400581359863, "global_step": 196142, "epoch": 2203, "lr": 1.7879769347116067e-05}
{"train_loss": 0.06692392379045486, "global_step": 196143, "epoch": 2203, "lr": 1.7879325059054925e-05}
{"train_loss": 0.05637979879975319, "global_step": 196144, "epoch": 2203, "lr": 1.7878880775312036e-05}
{"train_loss": 0.0698414221405983, "global_step": 196145, "epoch": 2203, "lr": 1.787843649588742e-05}
{"train_loss": 0.08444850146770477, "global_step": 196146, "epoch": 2203, "lr": 1.787799222078117e-05}
{"train_loss": 0.06655684858560562, "global_step": 196147, "epoch": 2203, "lr": 1.7877547949993324e-05}
{"train_loss": 0.05097954720258713, "global_step": 196148, "epoch": 2203, "lr": 1.787710368352396e-05}
{"train_loss": 0.08495208621025085, "global_step": 196149, "epoch": 2203, "lr": 1.787665942137311e-05}
{"train_loss": 0.033001698553562164, "global_step": 196150, "epoch": 2203, "lr": 1.7876215163540877e-05}
{"train_loss": 0.04739496484398842, "global_step": 196151, "epoch": 2203, "lr": 1.7875770910027273e-05}
{"train_loss": 0.06511731445789337, "global_step": 196152, "epoch": 2203, "lr": 1.787532666083239e-05}
{"train_loss": 0.05451028048992157, "global_step": 196153, "epoch": 2203, "lr": 1.7874882415956272e-05}
{"train_loss": 0.11259154230356216, "global_step": 196154, "epoch": 2203, "lr": 1.7874438175399e-05}
{"train_loss": 0.0692027813932869, "global_step": 196155, "epoch": 2203, "lr": 1.78739939391606e-05, "val_loss": 8.13588809967041}
{"train_loss": 0.06402208656072617, "global_step": 196156, "epoch": 2204, "lr": 1.7873549707241162e-05}
{"train_loss": 0.12853841483592987, "global_step": 196157, "epoch": 2204, "lr": 1.787310547964072e-05}
{"train_loss": 0.05736764520406723, "global_step": 196158, "epoch": 2204, "lr": 1.787266125635937e-05}
{"train_loss": 0.09423227608203888, "global_step": 196159, "epoch": 2204, "lr": 1.7872217037397143e-05}
{"train_loss": 0.08205053955316544, "global_step": 196160, "epoch": 2204, "lr": 1.7871772822754083e-05}
{"train_loss": 0.0520024411380291, "global_step": 196161, "epoch": 2204, "lr": 1.78713286124303e-05}
{"train_loss": 0.055452946573495865, "global_step": 196162, "epoch": 2204, "lr": 1.7870884406425803e-05}
{"train_loss": 0.11543556302785873, "global_step": 196163, "epoch": 2204, "lr": 1.787044020474069e-05}
{"train_loss": 0.10842731595039368, "global_step": 196164, "epoch": 2204, "lr": 1.786999600737499e-05}
{"train_loss": 0.055010225623846054, "global_step": 196165, "epoch": 2204, "lr": 1.786955181432879e-05}
{"train_loss": 0.039785996079444885, "global_step": 196166, "epoch": 2204, "lr": 1.7869107625602123e-05}
{"train_loss": 0.07182226330041885, "global_step": 196167, "epoch": 2204, "lr": 1.7868663441195077e-05}
{"train_loss": 0.04593876749277115, "global_step": 196168, "epoch": 2204, "lr": 1.786821926110769e-05}
{"train_loss": 0.09079715609550476, "global_step": 196169, "epoch": 2204, "lr": 1.7867775085340033e-05}
{"train_loss": 0.048430271446704865, "global_step": 196170, "epoch": 2204, "lr": 1.7867330913892154e-05}
{"train_loss": 0.012633828446269035, "global_step": 196171, "epoch": 2204, "lr": 1.7866886746764133e-05}
{"train_loss": 0.05348074808716774, "global_step": 196172, "epoch": 2204, "lr": 1.7866442583956007e-05}
{"train_loss": 0.02642422914505005, "global_step": 196173, "epoch": 2204, "lr": 1.7865998425467857e-05}
{"train_loss": 0.07319078594446182, "global_step": 196174, "epoch": 2204, "lr": 1.7865554271299718e-05}
{"train_loss": 0.08659347891807556, "global_step": 196175, "epoch": 2204, "lr": 1.786511012145166e-05}
{"train_loss": 0.03582362085580826, "global_step": 196176, "epoch": 2204, "lr": 1.7864665975923765e-05}
{"train_loss": 0.07093197107315063, "global_step": 196177, "epoch": 2204, "lr": 1.786422183471606e-05}
{"train_loss": 0.09189091622829437, "global_step": 196178, "epoch": 2204, "lr": 1.7863777697828633e-05}
{"train_loss": 0.06807199120521545, "global_step": 196179, "epoch": 2204, "lr": 1.7863333565261514e-05}
{"train_loss": 0.029321320354938507, "global_step": 196180, "epoch": 2204, "lr": 1.7862889437014792e-05}
{"train_loss": 0.04514222964644432, "global_step": 196181, "epoch": 2204, "lr": 1.78624453130885e-05}
{"train_loss": 0.04781137406826019, "global_step": 196182, "epoch": 2204, "lr": 1.7862001193482724e-05}
{"train_loss": 0.04968402534723282, "global_step": 196183, "epoch": 2204, "lr": 1.7861557078197493e-05}
{"train_loss": 0.10250823944807053, "global_step": 196184, "epoch": 2204, "lr": 1.7861112967232903e-05}
{"train_loss": 0.10389969497919083, "global_step": 196185, "epoch": 2204, "lr": 1.7860668860588973e-05}
{"train_loss": 0.08585096150636673, "global_step": 196186, "epoch": 2204, "lr": 1.786022475826581e-05}
{"train_loss": 0.06070129573345184, "global_step": 196187, "epoch": 2204, "lr": 1.7859780660263426e-05}
{"train_loss": 0.07787810266017914, "global_step": 196188, "epoch": 2204, "lr": 1.785933656658192e-05}
{"train_loss": 0.04081390053033829, "global_step": 196189, "epoch": 2204, "lr": 1.7858892477221318e-05}
{"train_loss": 0.050107263028621674, "global_step": 196190, "epoch": 2204, "lr": 1.7858448392181716e-05}
{"train_loss": 0.04370932653546333, "global_step": 196191, "epoch": 2204, "lr": 1.7858004311463127e-05}
{"train_loss": 0.0656185895204544, "global_step": 196192, "epoch": 2204, "lr": 1.7857560235065663e-05}
{"train_loss": 0.06602416187524796, "global_step": 196193, "epoch": 2204, "lr": 1.7857116162989342e-05}
{"train_loss": 0.11763919144868851, "global_step": 196194, "epoch": 2204, "lr": 1.7856672095234257e-05}
{"train_loss": 0.06156046688556671, "global_step": 196195, "epoch": 2204, "lr": 1.7856228031800426e-05}
{"train_loss": 0.08181943744421005, "global_step": 196196, "epoch": 2204, "lr": 1.7855783972687957e-05}
{"train_loss": 0.12081848829984665, "global_step": 196197, "epoch": 2204, "lr": 1.7855339917896867e-05}
{"train_loss": 0.07467474788427353, "global_step": 196198, "epoch": 2204, "lr": 1.785489586742725e-05}
{"train_loss": 0.12141484767198563, "global_step": 196199, "epoch": 2204, "lr": 1.7854451821279157e-05}
{"train_loss": 0.04911939054727554, "global_step": 196200, "epoch": 2204, "lr": 1.7854007779452618e-05}
{"train_loss": 0.026844846084713936, "global_step": 196201, "epoch": 2204, "lr": 1.7853563741947736e-05}
{"train_loss": 0.04331083968281746, "global_step": 196202, "epoch": 2204, "lr": 1.785311970876453e-05}
{"train_loss": 0.06476512551307678, "global_step": 196203, "epoch": 2204, "lr": 1.7852675679903102e-05}
{"train_loss": 0.12623772025108337, "global_step": 196204, "epoch": 2204, "lr": 1.7852231655363472e-05}
{"train_loss": 0.11917173862457275, "global_step": 196205, "epoch": 2204, "lr": 1.7851787635145733e-05}
{"train_loss": 0.02471785433590412, "global_step": 196206, "epoch": 2204, "lr": 1.785134361924991e-05}
{"train_loss": 0.031146937981247902, "global_step": 196207, "epoch": 2204, "lr": 1.7850899607676103e-05}
{"train_loss": 0.04351531341671944, "global_step": 196208, "epoch": 2204, "lr": 1.785045560042433e-05}
{"train_loss": 0.05679688602685928, "global_step": 196209, "epoch": 2204, "lr": 1.7850011597494693e-05}
{"train_loss": 0.058388035744428635, "global_step": 196210, "epoch": 2204, "lr": 1.7849567598887207e-05}
{"train_loss": 0.034193817526102066, "global_step": 196211, "epoch": 2204, "lr": 1.7849123604601958e-05}
{"train_loss": 0.0731651782989502, "global_step": 196212, "epoch": 2204, "lr": 1.7848679614639026e-05}
{"train_loss": 0.04180367663502693, "global_step": 196213, "epoch": 2204, "lr": 1.7848235628998422e-05}
{"train_loss": 0.030294468626379967, "global_step": 196214, "epoch": 2204, "lr": 1.7847791647680244e-05}
{"train_loss": 0.02906513772904873, "global_step": 196215, "epoch": 2204, "lr": 1.7847347670684533e-05}
{"train_loss": 0.03973742574453354, "global_step": 196216, "epoch": 2204, "lr": 1.7846903698011363e-05}
{"train_loss": 0.04525979980826378, "global_step": 196217, "epoch": 2204, "lr": 1.7846459729660768e-05}
{"train_loss": 0.10524879395961761, "global_step": 196218, "epoch": 2204, "lr": 1.7846015765632846e-05}
{"train_loss": 0.04803398996591568, "global_step": 196219, "epoch": 2204, "lr": 1.784557180592762e-05}
{"train_loss": 0.05504852533340454, "global_step": 196220, "epoch": 2204, "lr": 1.784512785054518e-05}
{"train_loss": 0.09011930227279663, "global_step": 196221, "epoch": 2204, "lr": 1.784468389948555e-05}
{"train_loss": 0.05666796863079071, "global_step": 196222, "epoch": 2204, "lr": 1.7844239952748827e-05}
{"train_loss": 0.06173461675643921, "global_step": 196223, "epoch": 2204, "lr": 1.784379601033504e-05}
{"train_loss": 0.05883710831403732, "global_step": 196224, "epoch": 2204, "lr": 1.7843352072244278e-05}
{"train_loss": 0.039583440870046616, "global_step": 196225, "epoch": 2204, "lr": 1.784290813847657e-05}
{"train_loss": 0.025357738137245178, "global_step": 196226, "epoch": 2204, "lr": 1.7842464209032006e-05}
{"train_loss": 0.03454452008008957, "global_step": 196227, "epoch": 2204, "lr": 1.7842020283910615e-05}
{"train_loss": 0.06056255102157593, "global_step": 196228, "epoch": 2204, "lr": 1.7841576363112477e-05}
{"train_loss": 0.046898480504751205, "global_step": 196229, "epoch": 2204, "lr": 1.784113244663766e-05}
{"train_loss": 0.07248228788375854, "global_step": 196230, "epoch": 2204, "lr": 1.7840688534486193e-05}
{"train_loss": 0.07439004629850388, "global_step": 196231, "epoch": 2204, "lr": 1.784024462665817e-05}
{"train_loss": 0.047759901732206345, "global_step": 196232, "epoch": 2204, "lr": 1.783980072315363e-05}
{"train_loss": 0.03409188613295555, "global_step": 196233, "epoch": 2204, "lr": 1.783935682397263e-05}
{"train_loss": 0.061937980353832245, "global_step": 196234, "epoch": 2204, "lr": 1.7838912929115247e-05}
{"train_loss": 0.10242404043674469, "global_step": 196235, "epoch": 2204, "lr": 1.7838469038581518e-05}
{"train_loss": 0.04750531539320946, "global_step": 196236, "epoch": 2204, "lr": 1.7838025152371523e-05}
{"train_loss": 0.06787705421447754, "global_step": 196237, "epoch": 2204, "lr": 1.783758127048532e-05}
{"train_loss": 0.043304331600666046, "global_step": 196238, "epoch": 2204, "lr": 1.7837137392922948e-05}
{"train_loss": 0.08198176324367523, "global_step": 196239, "epoch": 2204, "lr": 1.7836693519684494e-05}
{"train_loss": 0.0802382081747055, "global_step": 196240, "epoch": 2204, "lr": 1.7836249650769987e-05}
{"train_loss": 0.074034184217453, "global_step": 196241, "epoch": 2204, "lr": 1.7835805786179522e-05}
{"train_loss": 0.05622250214219093, "global_step": 196242, "epoch": 2204, "lr": 1.7835361925913123e-05}
{"train_loss": 0.02796224132180214, "global_step": 196243, "epoch": 2204, "lr": 1.7834918069970885e-05}
{"train_loss": 0.06293881797556126, "global_step": 196244, "epoch": 2204, "lr": 1.783447421835283e-05, "val_loss": 8.286877632141113}
{"train_loss": 0.07510018348693848, "global_step": 196245, "epoch": 2205, "lr": 1.7834030371059064e-05}
{"train_loss": 0.045032650232315063, "global_step": 196246, "epoch": 2205, "lr": 1.7833586528089596e-05}
{"train_loss": 0.08219319581985474, "global_step": 196247, "epoch": 2205, "lr": 1.783314268944451e-05}
{"train_loss": 0.06441500037908554, "global_step": 196248, "epoch": 2205, "lr": 1.7832698855123885e-05}
{"train_loss": 0.07537642866373062, "global_step": 196249, "epoch": 2205, "lr": 1.7832255025127746e-05}
{"train_loss": 0.06381231546401978, "global_step": 196250, "epoch": 2205, "lr": 1.7831811199456182e-05}
{"train_loss": 0.07071807980537415, "global_step": 196251, "epoch": 2205, "lr": 1.7831367378109226e-05}
{"train_loss": 0.05132706090807915, "global_step": 196252, "epoch": 2205, "lr": 1.7830923561086966e-05}
{"train_loss": 0.07088737189769745, "global_step": 196253, "epoch": 2205, "lr": 1.783047974838943e-05}
{"train_loss": 0.0706205815076828, "global_step": 196254, "epoch": 2205, "lr": 1.783003594001671e-05}
{"train_loss": 0.0898212268948555, "global_step": 196255, "epoch": 2205, "lr": 1.782959213596883e-05}
{"train_loss": 0.1303291618824005, "global_step": 196256, "epoch": 2205, "lr": 1.7829148336245886e-05}
{"train_loss": 0.07530941069126129, "global_step": 196257, "epoch": 2205, "lr": 1.7828704540847906e-05}
{"train_loss": 0.07042546570301056, "global_step": 196258, "epoch": 2205, "lr": 1.782826074977498e-05}
{"train_loss": 0.08049049228429794, "global_step": 196259, "epoch": 2205, "lr": 1.782781696302714e-05}
{"train_loss": 0.05620119348168373, "global_step": 196260, "epoch": 2205, "lr": 1.7827373180604474e-05}
{"train_loss": 0.07100197672843933, "global_step": 196261, "epoch": 2205, "lr": 1.7826929402507004e-05}
{"train_loss": 0.10011114925146103, "global_step": 196262, "epoch": 2205, "lr": 1.782648562873483e-05}
{"train_loss": 0.08241182565689087, "global_step": 196263, "epoch": 2205, "lr": 1.782604185928798e-05}
{"train_loss": 0.05118561163544655, "global_step": 196264, "epoch": 2205, "lr": 1.7825598094166525e-05}
{"train_loss": 0.06777716428041458, "global_step": 196265, "epoch": 2205, "lr": 1.7825154333370536e-05}
{"train_loss": 0.047447096556425095, "global_step": 196266, "epoch": 2205, "lr": 1.782471057690005e-05}
{"train_loss": 0.040714554488658905, "global_step": 196267, "epoch": 2205, "lr": 1.782426682475516e-05}
{"train_loss": 0.04313036426901817, "global_step": 196268, "epoch": 2205, "lr": 1.7823823076935887e-05}
{"train_loss": 0.05258863419294357, "global_step": 196269, "epoch": 2205, "lr": 1.7823379333442325e-05}
{"train_loss": 0.040057893842458725, "global_step": 196270, "epoch": 2205, "lr": 1.7822935594274515e-05}
{"train_loss": 0.04765947535634041, "global_step": 196271, "epoch": 2205, "lr": 1.7822491859432506e-05}
{"train_loss": 0.07333247363567352, "global_step": 196272, "epoch": 2205, "lr": 1.7822048128916386e-05}
{"train_loss": 0.06195171922445297, "global_step": 196273, "epoch": 2205, "lr": 1.7821604402726182e-05}
{"train_loss": 0.07103855907917023, "global_step": 196274, "epoch": 2205, "lr": 1.782116068086199e-05}
{"train_loss": 0.06850679218769073, "global_step": 196275, "epoch": 2205, "lr": 1.7820716963323846e-05}
{"train_loss": 0.0591110922396183, "global_step": 196276, "epoch": 2205, "lr": 1.7820273250111803e-05}
{"train_loss": 0.10951045900583267, "global_step": 196277, "epoch": 2205, "lr": 1.7819829541225946e-05}
{"train_loss": 0.06008627638220787, "global_step": 196278, "epoch": 2205, "lr": 1.7819385836666304e-05}
{"train_loss": 0.06309592723846436, "global_step": 196279, "epoch": 2205, "lr": 1.781894213643297e-05}
{"train_loss": 0.13164260983467102, "global_step": 196280, "epoch": 2205, "lr": 1.7818498440525973e-05}
{"train_loss": 0.11176259815692902, "global_step": 196281, "epoch": 2205, "lr": 1.78180547489454e-05}
{"train_loss": 0.044213905930519104, "global_step": 196282, "epoch": 2205, "lr": 1.781761106169128e-05}
{"train_loss": 0.03331930935382843, "global_step": 196283, "epoch": 2205, "lr": 1.7817167378763694e-05}
{"train_loss": 0.07350865751504898, "global_step": 196284, "epoch": 2205, "lr": 1.7816723700162707e-05}
{"train_loss": 0.0185908954590559, "global_step": 196285, "epoch": 2205, "lr": 1.781628002588836e-05}
{"train_loss": 0.04751596599817276, "global_step": 196286, "epoch": 2205, "lr": 1.7815836355940736e-05}
{"train_loss": 0.08663574606180191, "global_step": 196287, "epoch": 2205, "lr": 1.7815392690319865e-05}
{"train_loss": 0.03428829833865166, "global_step": 196288, "epoch": 2205, "lr": 1.7814949029025836e-05}
{"train_loss": 0.09552936255931854, "global_step": 196289, "epoch": 2205, "lr": 1.781450537205868e-05}
{"train_loss": 0.08527549356222153, "global_step": 196290, "epoch": 2205, "lr": 1.781406171941849e-05}
{"train_loss": 0.06964538246393204, "global_step": 196291, "epoch": 2205, "lr": 1.7813618071105286e-05}
{"train_loss": 0.016641709953546524, "global_step": 196292, "epoch": 2205, "lr": 1.7813174427119165e-05}
{"train_loss": 0.07096690684556961, "global_step": 196293, "epoch": 2205, "lr": 1.7812730787460163e-05}
{"train_loss": 0.08049378544092178, "global_step": 196294, "epoch": 2205, "lr": 1.7812287152128354e-05}
{"train_loss": 0.05729895457625389, "global_step": 196295, "epoch": 2205, "lr": 1.781184352112378e-05}
{"train_loss": 0.04461640119552612, "global_step": 196296, "epoch": 2205, "lr": 1.7811399894446522e-05}
{"train_loss": 0.02427150309085846, "global_step": 196297, "epoch": 2205, "lr": 1.7810956272096618e-05}
{"train_loss": 0.05532211810350418, "global_step": 196298, "epoch": 2205, "lr": 1.781051265407415e-05}
{"train_loss": 0.04977281764149666, "global_step": 196299, "epoch": 2205, "lr": 1.7810069040379158e-05}
{"train_loss": 0.06930727511644363, "global_step": 196300, "epoch": 2205, "lr": 1.7809625431011717e-05}
{"train_loss": 0.09484976530075073, "global_step": 196301, "epoch": 2205, "lr": 1.7809181825971866e-05}
{"train_loss": 0.06826546788215637, "global_step": 196302, "epoch": 2205, "lr": 1.780873822525968e-05}
{"train_loss": 0.050844062119722366, "global_step": 196303, "epoch": 2205, "lr": 1.780829462887523e-05}
{"train_loss": 0.08487541228532791, "global_step": 196304, "epoch": 2205, "lr": 1.7807851036818552e-05}
{"train_loss": 0.0386340469121933, "global_step": 196305, "epoch": 2205, "lr": 1.780740744908973e-05}
{"train_loss": 0.045314814895391464, "global_step": 196306, "epoch": 2205, "lr": 1.7806963865688793e-05}
{"train_loss": 0.08540011197328568, "global_step": 196307, "epoch": 2205, "lr": 1.780652028661583e-05}
{"train_loss": 0.0841129943728447, "global_step": 196308, "epoch": 2205, "lr": 1.7806076711870893e-05}
{"train_loss": 0.06829052418470383, "global_step": 196309, "epoch": 2205, "lr": 1.7805633141454016e-05}
{"train_loss": 0.04310274496674538, "global_step": 196310, "epoch": 2205, "lr": 1.78051895753653e-05}
{"train_loss": 0.06552747637033463, "global_step": 196311, "epoch": 2205, "lr": 1.780474601360476e-05}
{"train_loss": 0.11155195534229279, "global_step": 196312, "epoch": 2205, "lr": 1.7804302456172505e-05}
{"train_loss": 0.052865613251924515, "global_step": 196313, "epoch": 2205, "lr": 1.780385890306855e-05}
{"train_loss": 0.03833061829209328, "global_step": 196314, "epoch": 2205, "lr": 1.780341535429299e-05}
{"train_loss": 0.036389827728271484, "global_step": 196315, "epoch": 2205, "lr": 1.7802971809845863e-05}
{"train_loss": 0.049016013741493225, "global_step": 196316, "epoch": 2205, "lr": 1.7802528269727225e-05}
{"train_loss": 0.1277996003627777, "global_step": 196317, "epoch": 2205, "lr": 1.780208473393716e-05}
{"train_loss": 0.06368500739336014, "global_step": 196318, "epoch": 2205, "lr": 1.780164120247569e-05}
{"train_loss": 0.039916250854730606, "global_step": 196319, "epoch": 2205, "lr": 1.7801197675342903e-05}
{"train_loss": 0.06937551498413086, "global_step": 196320, "epoch": 2205, "lr": 1.780075415253887e-05}
{"train_loss": 0.0673263892531395, "global_step": 196321, "epoch": 2205, "lr": 1.780031063406361e-05}
{"train_loss": 0.0531708225607872, "global_step": 196322, "epoch": 2205, "lr": 1.7799867119917224e-05}
{"train_loss": 0.059451423585414886, "global_step": 196323, "epoch": 2205, "lr": 1.779942361009974e-05}
{"train_loss": 0.037912528961896896, "global_step": 196324, "epoch": 2205, "lr": 1.7798980104611244e-05}
{"train_loss": 0.055304285138845444, "global_step": 196325, "epoch": 2205, "lr": 1.779853660345177e-05}
{"train_loss": 0.053715575486421585, "global_step": 196326, "epoch": 2205, "lr": 1.7798093106621406e-05}
{"train_loss": 0.044926632195711136, "global_step": 196327, "epoch": 2205, "lr": 1.7797649614120176e-05}
{"train_loss": 0.07509297132492065, "global_step": 196328, "epoch": 2205, "lr": 1.7797206125948178e-05}
{"train_loss": 0.019436324015259743, "global_step": 196329, "epoch": 2205, "lr": 1.7796762642105436e-05}
{"train_loss": 0.027093224227428436, "global_step": 196330, "epoch": 2205, "lr": 1.7796319162592046e-05}
{"train_loss": 0.0563192293047905, "global_step": 196331, "epoch": 2205, "lr": 1.779587568740803e-05}
{"train_loss": 0.07594022899866104, "global_step": 196332, "epoch": 2205, "lr": 1.779543221655348e-05}
{"train_loss": 0.06393997526068367, "global_step": 196333, "epoch": 2205, "lr": 1.779498875002843e-05, "val_loss": 8.228917121887207, "train_action_mse_error": 26.135160446166992}
{"train_loss": 0.028832072392106056, "global_step": 196334, "epoch": 2206, "lr": 1.7794545287832963e-05}
{"train_loss": 0.047163110226392746, "global_step": 196335, "epoch": 2206, "lr": 1.7794101829967113e-05}
{"train_loss": 0.1007177084684372, "global_step": 196336, "epoch": 2206, "lr": 1.7793658376430973e-05}
{"train_loss": 0.025123804807662964, "global_step": 196337, "epoch": 2206, "lr": 1.7793214927224568e-05}
{"train_loss": 0.05276499688625336, "global_step": 196338, "epoch": 2206, "lr": 1.7792771482347964e-05}
{"train_loss": 0.06973956525325775, "global_step": 196339, "epoch": 2206, "lr": 1.7792328041801248e-05}
{"train_loss": 0.027221962809562683, "global_step": 196340, "epoch": 2206, "lr": 1.779188460558445e-05}
{"train_loss": 0.07634537667036057, "global_step": 196341, "epoch": 2206, "lr": 1.7791441173697654e-05}
{"train_loss": 0.052962370216846466, "global_step": 196342, "epoch": 2206, "lr": 1.7790997746140886e-05}
{"train_loss": 0.06436452269554138, "global_step": 196343, "epoch": 2206, "lr": 1.7790554322914245e-05}
{"train_loss": 0.08534054458141327, "global_step": 196344, "epoch": 2206, "lr": 1.7790110904017755e-05}
{"train_loss": 0.06332485377788544, "global_step": 196345, "epoch": 2206, "lr": 1.778966748945151e-05}
{"train_loss": 0.03781753405928612, "global_step": 196346, "epoch": 2206, "lr": 1.7789224079215532e-05}
{"train_loss": 0.040354013442993164, "global_step": 196347, "epoch": 2206, "lr": 1.778878067330992e-05}
{"train_loss": 0.05486351251602173, "global_step": 196348, "epoch": 2206, "lr": 1.778833727173471e-05}
{"train_loss": 0.04126007854938507, "global_step": 196349, "epoch": 2206, "lr": 1.778789387448995e-05}
{"train_loss": 0.06374154984951019, "global_step": 196350, "epoch": 2206, "lr": 1.778745048157574e-05}
{"train_loss": 0.07009075582027435, "global_step": 196351, "epoch": 2206, "lr": 1.778700709299209e-05}
{"train_loss": 0.05098741129040718, "global_step": 196352, "epoch": 2206, "lr": 1.77865637087391e-05}
{"train_loss": 0.09448190033435822, "global_step": 196353, "epoch": 2206, "lr": 1.778612032881682e-05}
{"train_loss": 0.06633102893829346, "global_step": 196354, "epoch": 2206, "lr": 1.7785676953225277e-05}
{"train_loss": 0.0414554625749588, "global_step": 196355, "epoch": 2206, "lr": 1.7785233581964568e-05}
{"train_loss": 0.06509045511484146, "global_step": 196356, "epoch": 2206, "lr": 1.7784790215034752e-05}
{"train_loss": 0.08042895793914795, "global_step": 196357, "epoch": 2206, "lr": 1.7784346852435863e-05}
{"train_loss": 0.02959679253399372, "global_step": 196358, "epoch": 2206, "lr": 1.7783903494167992e-05}
{"train_loss": 0.06868738681077957, "global_step": 196359, "epoch": 2206, "lr": 1.7783460140231167e-05}
{"train_loss": 0.08213723450899124, "global_step": 196360, "epoch": 2206, "lr": 1.778301679062548e-05}
{"train_loss": 0.0763658806681633, "global_step": 196361, "epoch": 2206, "lr": 1.7782573445350954e-05}
{"train_loss": 0.0729822963476181, "global_step": 196362, "epoch": 2206, "lr": 1.7782130104407686e-05}
{"train_loss": 0.03822477161884308, "global_step": 196363, "epoch": 2206, "lr": 1.7781686767795703e-05}
{"train_loss": 0.036514826118946075, "global_step": 196364, "epoch": 2206, "lr": 1.7781243435515095e-05}
{"train_loss": 0.06651678681373596, "global_step": 196365, "epoch": 2206, "lr": 1.7780800107565894e-05}
{"train_loss": 0.048849355429410934, "global_step": 196366, "epoch": 2206, "lr": 1.7780356783948183e-05}
{"train_loss": 0.08986026793718338, "global_step": 196367, "epoch": 2206, "lr": 1.7779913464661995e-05}
{"train_loss": 0.053576044738292694, "global_step": 196368, "epoch": 2206, "lr": 1.777947014970742e-05}
{"train_loss": 0.038617074489593506, "global_step": 196369, "epoch": 2206, "lr": 1.777902683908449e-05}
{"train_loss": 0.09261338412761688, "global_step": 196370, "epoch": 2206, "lr": 1.777858353279329e-05}
{"train_loss": 0.0772763341665268, "global_step": 196371, "epoch": 2206, "lr": 1.777814023083385e-05}
{"train_loss": 0.07920524477958679, "global_step": 196372, "epoch": 2206, "lr": 1.7777696933206268e-05}
{"train_loss": 0.05322318524122238, "global_step": 196373, "epoch": 2206, "lr": 1.7777253639910557e-05}
{"train_loss": 0.07841696590185165, "global_step": 196374, "epoch": 2206, "lr": 1.777681035094681e-05}
{"train_loss": 0.08187989145517349, "global_step": 196375, "epoch": 2206, "lr": 1.7776367066315085e-05}
{"train_loss": 0.04538973048329353, "global_step": 196376, "epoch": 2206, "lr": 1.7775923786015424e-05}
{"train_loss": 0.06892123073339462, "global_step": 196377, "epoch": 2206, "lr": 1.7775480510047914e-05}
{"train_loss": 0.04829227551817894, "global_step": 196378, "epoch": 2206, "lr": 1.7775037238412585e-05}
{"train_loss": 0.0554419569671154, "global_step": 196379, "epoch": 2206, "lr": 1.777459397110952e-05}
{"train_loss": 0.07914292067289352, "global_step": 196380, "epoch": 2206, "lr": 1.7774150708138755e-05}
{"train_loss": 0.08763410896062851, "global_step": 196381, "epoch": 2206, "lr": 1.7773707449500376e-05}
{"train_loss": 0.07338155806064606, "global_step": 196382, "epoch": 2206, "lr": 1.7773264195194416e-05}
{"train_loss": 0.09359954297542572, "global_step": 196383, "epoch": 2206, "lr": 1.7772820945220964e-05}
{"train_loss": 0.06777556240558624, "global_step": 196384, "epoch": 2206, "lr": 1.7772377699580046e-05}
{"train_loss": 0.03745780140161514, "global_step": 196385, "epoch": 2206, "lr": 1.7771934458271755e-05}
{"train_loss": 0.10357388108968735, "global_step": 196386, "epoch": 2206, "lr": 1.7771491221296132e-05}
{"train_loss": 0.059327542781829834, "global_step": 196387, "epoch": 2206, "lr": 1.777104798865323e-05}
{"train_loss": 0.04252650588750839, "global_step": 196388, "epoch": 2206, "lr": 1.777060476034313e-05}
{"train_loss": 0.10565938800573349, "global_step": 196389, "epoch": 2206, "lr": 1.7770161536365858e-05}
{"train_loss": 0.030864374712109566, "global_step": 196390, "epoch": 2206, "lr": 1.776971831672152e-05}
{"train_loss": 0.043990980833768845, "global_step": 196391, "epoch": 2206, "lr": 1.7769275101410133e-05}
{"train_loss": 0.10016678273677826, "global_step": 196392, "epoch": 2206, "lr": 1.7768831890431785e-05}
{"train_loss": 0.04657399281859398, "global_step": 196393, "epoch": 2206, "lr": 1.7768388683786515e-05}
{"train_loss": 0.03414694964885712, "global_step": 196394, "epoch": 2206, "lr": 1.77679454814744e-05}
{"train_loss": 0.1197781041264534, "global_step": 196395, "epoch": 2206, "lr": 1.776750228349548e-05}
{"train_loss": 0.10036741942167282, "global_step": 196396, "epoch": 2206, "lr": 1.7767059089849846e-05}
{"train_loss": 0.050870463252067566, "global_step": 196397, "epoch": 2206, "lr": 1.776661590053752e-05}
{"train_loss": 0.03576165810227394, "global_step": 196398, "epoch": 2206, "lr": 1.7766172715558593e-05}
{"train_loss": 0.07488125562667847, "global_step": 196399, "epoch": 2206, "lr": 1.7765729534913095e-05}
{"train_loss": 0.04129805415868759, "global_step": 196400, "epoch": 2206, "lr": 1.7765286358601125e-05}
{"train_loss": 0.05125169828534126, "global_step": 196401, "epoch": 2206, "lr": 1.7764843186622694e-05}
{"train_loss": 0.07975129038095474, "global_step": 196402, "epoch": 2206, "lr": 1.7764400018977904e-05}
{"train_loss": 0.07635263353586197, "global_step": 196403, "epoch": 2206, "lr": 1.7763956855666786e-05}
{"train_loss": 0.0492730587720871, "global_step": 196404, "epoch": 2206, "lr": 1.7763513696689426e-05}
{"train_loss": 0.05799569562077522, "global_step": 196405, "epoch": 2206, "lr": 1.776307054204585e-05}
{"train_loss": 0.07948756217956543, "global_step": 196406, "epoch": 2206, "lr": 1.776262739173616e-05}
{"train_loss": 0.03709641844034195, "global_step": 196407, "epoch": 2206, "lr": 1.7762184245760366e-05}
{"train_loss": 0.0380076989531517, "global_step": 196408, "epoch": 2206, "lr": 1.776174110411858e-05}
{"train_loss": 0.040931541472673416, "global_step": 196409, "epoch": 2206, "lr": 1.776129796681081e-05}
{"train_loss": 0.07016732543706894, "global_step": 196410, "epoch": 2206, "lr": 1.7760854833837142e-05}
{"train_loss": 0.07917814701795578, "global_step": 196411, "epoch": 2206, "lr": 1.7760411705197653e-05}
{"train_loss": 0.04994671791791916, "global_step": 196412, "epoch": 2206, "lr": 1.7759968580892368e-05}
{"train_loss": 0.028887979686260223, "global_step": 196413, "epoch": 2206, "lr": 1.7759525460921374e-05}
{"train_loss": 0.05572963505983353, "global_step": 196414, "epoch": 2206, "lr": 1.7759082345284705e-05}
{"train_loss": 0.07237471640110016, "global_step": 196415, "epoch": 2206, "lr": 1.775863923398245e-05}
{"train_loss": 0.032399434596300125, "global_step": 196416, "epoch": 2206, "lr": 1.7758196127014638e-05}
{"train_loss": 0.035096313804388046, "global_step": 196417, "epoch": 2206, "lr": 1.7757753024381358e-05}
{"train_loss": 0.06811125576496124, "global_step": 196418, "epoch": 2206, "lr": 1.7757309926082644e-05}
{"train_loss": 0.027005299925804138, "global_step": 196419, "epoch": 2206, "lr": 1.775686683211858e-05}
{"train_loss": 0.06598886847496033, "global_step": 196420, "epoch": 2206, "lr": 1.775642374248919e-05}
{"train_loss": 0.05055792257189751, "global_step": 196421, "epoch": 2206, "lr": 1.775598065719458e-05}
{"train_loss": 0.060912247516968276, "global_step": 196422, "epoch": 2206, "lr": 1.7755537576234767e-05, "val_loss": 8.268021583557129}
{"train_loss": 0.018888283520936966, "global_step": 196423, "epoch": 2207, "lr": 1.7755094499609844e-05}
{"train_loss": 0.05467929318547249, "global_step": 196424, "epoch": 2207, "lr": 1.775465142731986e-05}
{"train_loss": 0.06713078916072845, "global_step": 196425, "epoch": 2207, "lr": 1.7754208359364843e-05}
{"train_loss": 0.10659065842628479, "global_step": 196426, "epoch": 2207, "lr": 1.7753765295744905e-05}
{"train_loss": 0.06480620056390762, "global_step": 196427, "epoch": 2207, "lr": 1.7753322236460064e-05}
{"train_loss": 0.09085380285978317, "global_step": 196428, "epoch": 2207, "lr": 1.7752879181510406e-05}
{"train_loss": 0.09718144685029984, "global_step": 196429, "epoch": 2207, "lr": 1.775243613089597e-05}
{"train_loss": 0.07330165803432465, "global_step": 196430, "epoch": 2207, "lr": 1.7751993084616843e-05}
{"train_loss": 0.08680582791566849, "global_step": 196431, "epoch": 2207, "lr": 1.775155004267305e-05}
{"train_loss": 0.04922999069094658, "global_step": 196432, "epoch": 2207, "lr": 1.7751107005064684e-05}
{"train_loss": 0.07062256336212158, "global_step": 196433, "epoch": 2207, "lr": 1.7750663971791764e-05}
{"train_loss": 0.06696667522192001, "global_step": 196434, "epoch": 2207, "lr": 1.77502209428544e-05}
{"train_loss": 0.06799381971359253, "global_step": 196435, "epoch": 2207, "lr": 1.7749777918252607e-05}
{"train_loss": 0.10644406825304031, "global_step": 196436, "epoch": 2207, "lr": 1.774933489798648e-05}
{"train_loss": 0.06613244861364365, "global_step": 196437, "epoch": 2207, "lr": 1.7748891882056046e-05}
{"train_loss": 0.07564094662666321, "global_step": 196438, "epoch": 2207, "lr": 1.7748448870461393e-05}
{"train_loss": 0.03258461132645607, "global_step": 196439, "epoch": 2207, "lr": 1.7748005863202556e-05}
{"train_loss": 0.08447737991809845, "global_step": 196440, "epoch": 2207, "lr": 1.774756286027962e-05}
{"train_loss": 0.06133822724223137, "global_step": 196441, "epoch": 2207, "lr": 1.7747119861692618e-05}
{"train_loss": 0.08828332275152206, "global_step": 196442, "epoch": 2207, "lr": 1.7746676867441635e-05}
{"train_loss": 0.10538605600595474, "global_step": 196443, "epoch": 2207, "lr": 1.774623387752671e-05}
{"train_loss": 0.12228234112262726, "global_step": 196444, "epoch": 2207, "lr": 1.7745790891947918e-05}
{"train_loss": 0.07563448697328568, "global_step": 196445, "epoch": 2207, "lr": 1.7745347910705294e-05}
{"train_loss": 0.07404038310050964, "global_step": 196446, "epoch": 2207, "lr": 1.7744904933798924e-05}
{"train_loss": 0.08598633855581284, "global_step": 196447, "epoch": 2207, "lr": 1.7744461961228865e-05}
{"train_loss": 0.05826963111758232, "global_step": 196448, "epoch": 2207, "lr": 1.774401899299516e-05}
{"train_loss": 0.04489675536751747, "global_step": 196449, "epoch": 2207, "lr": 1.774357602909789e-05}
{"train_loss": 0.048517581075429916, "global_step": 196450, "epoch": 2207, "lr": 1.7743133069537092e-05}
{"train_loss": 0.059870753437280655, "global_step": 196451, "epoch": 2207, "lr": 1.7742690114312847e-05}
{"train_loss": 0.06765936315059662, "global_step": 196452, "epoch": 2207, "lr": 1.7742247163425195e-05}
{"train_loss": 0.05731569230556488, "global_step": 196453, "epoch": 2207, "lr": 1.774180421687422e-05}
{"train_loss": 0.08600902557373047, "global_step": 196454, "epoch": 2207, "lr": 1.7741361274659946e-05}
{"train_loss": 0.037577368319034576, "global_step": 196455, "epoch": 2207, "lr": 1.774091833678247e-05}
{"train_loss": 0.07774597406387329, "global_step": 196456, "epoch": 2207, "lr": 1.7740475403241817e-05}
{"train_loss": 0.03441975638270378, "global_step": 196457, "epoch": 2207, "lr": 1.7740032474038083e-05}
{"train_loss": 0.014670588076114655, "global_step": 196458, "epoch": 2207, "lr": 1.7739589549171293e-05}
{"train_loss": 0.07934563606977463, "global_step": 196459, "epoch": 2207, "lr": 1.7739146628641535e-05}
{"train_loss": 0.054160188883543015, "global_step": 196460, "epoch": 2207, "lr": 1.773870371244884e-05}
{"train_loss": 0.03391017019748688, "global_step": 196461, "epoch": 2207, "lr": 1.77382608005933e-05}
{"train_loss": 0.039165906608104706, "global_step": 196462, "epoch": 2207, "lr": 1.7737817893074936e-05}
{"train_loss": 0.09539055824279785, "global_step": 196463, "epoch": 2207, "lr": 1.7737374989893856e-05}
{"train_loss": 0.06613688170909882, "global_step": 196464, "epoch": 2207, "lr": 1.7736932091050086e-05}
{"train_loss": 0.08440575748682022, "global_step": 196465, "epoch": 2207, "lr": 1.7736489196543677e-05}
{"train_loss": 0.06453642249107361, "global_step": 196466, "epoch": 2207, "lr": 1.773604630637472e-05}
{"train_loss": 0.07998565584421158, "global_step": 196467, "epoch": 2207, "lr": 1.7735603420543244e-05}
{"train_loss": 0.03478419408202171, "global_step": 196468, "epoch": 2207, "lr": 1.7735160539049335e-05}
{"train_loss": 0.05726705491542816, "global_step": 196469, "epoch": 2207, "lr": 1.773471766189303e-05}
{"train_loss": 0.040238600224256516, "global_step": 196470, "epoch": 2207, "lr": 1.7734274789074412e-05}
{"train_loss": 0.05367313697934151, "global_step": 196471, "epoch": 2207, "lr": 1.7733831920593508e-05}
{"train_loss": 0.029860621318221092, "global_step": 196472, "epoch": 2207, "lr": 1.7733389056450418e-05}
{"train_loss": 0.07328318804502487, "global_step": 196473, "epoch": 2207, "lr": 1.773294619664516e-05}
{"train_loss": 0.05110533535480499, "global_step": 196474, "epoch": 2207, "lr": 1.7732503341177836e-05}
{"train_loss": 0.036541134119033813, "global_step": 196475, "epoch": 2207, "lr": 1.7732060490048462e-05}
{"train_loss": 0.04374397173523903, "global_step": 196476, "epoch": 2207, "lr": 1.7731617643257137e-05}
{"train_loss": 0.07356788963079453, "global_step": 196477, "epoch": 2207, "lr": 1.773117480080389e-05}
{"train_loss": 0.024579353630542755, "global_step": 196478, "epoch": 2207, "lr": 1.7730731962688806e-05}
{"train_loss": 0.027535518631339073, "global_step": 196479, "epoch": 2207, "lr": 1.7730289128911915e-05}
{"train_loss": 0.0505245216190815, "global_step": 196480, "epoch": 2207, "lr": 1.7729846299473313e-05}
{"train_loss": 0.0659690871834755, "global_step": 196481, "epoch": 2207, "lr": 1.7729403474373018e-05}
{"train_loss": 0.05638512223958969, "global_step": 196482, "epoch": 2207, "lr": 1.7728960653611117e-05}
{"train_loss": 0.09332247823476791, "global_step": 196483, "epoch": 2207, "lr": 1.772851783718768e-05}
{"train_loss": 0.0462070032954216, "global_step": 196484, "epoch": 2207, "lr": 1.772807502510273e-05}
{"train_loss": 0.07292632013559341, "global_step": 196485, "epoch": 2207, "lr": 1.7727632217356366e-05}
{"train_loss": 0.038320984691381454, "global_step": 196486, "epoch": 2207, "lr": 1.7727189413948615e-05}
{"train_loss": 0.08891348540782928, "global_step": 196487, "epoch": 2207, "lr": 1.7726746614879556e-05}
{"train_loss": 0.039177656173706055, "global_step": 196488, "epoch": 2207, "lr": 1.7726303820149238e-05}
{"train_loss": 0.059801485389471054, "global_step": 196489, "epoch": 2207, "lr": 1.7725861029757734e-05}
{"train_loss": 0.07711459696292877, "global_step": 196490, "epoch": 2207, "lr": 1.7725418243705085e-05}
{"train_loss": 0.05695527046918869, "global_step": 196491, "epoch": 2207, "lr": 1.7724975461991372e-05}
{"train_loss": 0.03391614928841591, "global_step": 196492, "epoch": 2207, "lr": 1.7724532684616623e-05}
{"train_loss": 0.1069708988070488, "global_step": 196493, "epoch": 2207, "lr": 1.772408991158094e-05}
{"train_loss": 0.049630776047706604, "global_step": 196494, "epoch": 2207, "lr": 1.7723647142884337e-05}
{"train_loss": 0.058114249259233475, "global_step": 196495, "epoch": 2207, "lr": 1.772320437852692e-05}
{"train_loss": 0.04756620153784752, "global_step": 196496, "epoch": 2207, "lr": 1.7722761618508705e-05}
{"train_loss": 0.08709238469600677, "global_step": 196497, "epoch": 2207, "lr": 1.7722318862829784e-05}
{"train_loss": 0.10647809505462646, "global_step": 196498, "epoch": 2207, "lr": 1.7721876111490195e-05}
{"train_loss": 0.0710054412484169, "global_step": 196499, "epoch": 2207, "lr": 1.7721433364490013e-05}
{"train_loss": 0.08466079831123352, "global_step": 196500, "epoch": 2207, "lr": 1.772099062182928e-05}
{"train_loss": 0.03073298931121826, "global_step": 196501, "epoch": 2207, "lr": 1.7720547883508086e-05}
{"train_loss": 0.1037193313241005, "global_step": 196502, "epoch": 2207, "lr": 1.7720105149526466e-05}
{"train_loss": 0.08806878328323364, "global_step": 196503, "epoch": 2207, "lr": 1.771966241988447e-05}
{"train_loss": 0.09648109972476959, "global_step": 196504, "epoch": 2207, "lr": 1.771921969458219e-05}
{"train_loss": 0.09661584347486496, "global_step": 196505, "epoch": 2207, "lr": 1.771877697361965e-05}
{"train_loss": 0.08042845875024796, "global_step": 196506, "epoch": 2207, "lr": 1.7718334256996944e-05}
{"train_loss": 0.06611407548189163, "global_step": 196507, "epoch": 2207, "lr": 1.7717891544714092e-05}
{"train_loss": 0.05745670944452286, "global_step": 196508, "epoch": 2207, "lr": 1.77174488367712e-05}
{"train_loss": 0.0753488689661026, "global_step": 196509, "epoch": 2207, "lr": 1.7717006133168283e-05}
{"train_loss": 0.038495007902383804, "global_step": 196510, "epoch": 2207, "lr": 1.7716563433905436e-05}
{"train_loss": 0.06600985906264756, "global_step": 196511, "epoch": 2207, "lr": 1.7716120738982693e-05, "val_loss": 8.061975479125977}
{"train_loss": 0.044974662363529205, "global_step": 196512, "epoch": 2208, "lr": 1.7715678048400137e-05}
{"train_loss": 0.07528173178434372, "global_step": 196513, "epoch": 2208, "lr": 1.7715235362157795e-05}
{"train_loss": 0.060276951640844345, "global_step": 196514, "epoch": 2208, "lr": 1.771479268025577e-05}
{"train_loss": 0.09540462493896484, "global_step": 196515, "epoch": 2208, "lr": 1.7714350002694074e-05}
{"train_loss": 0.07734963297843933, "global_step": 196516, "epoch": 2208, "lr": 1.7713907329472814e-05}
{"train_loss": 0.058803193271160126, "global_step": 196517, "epoch": 2208, "lr": 1.7713464660592e-05}
{"train_loss": 0.028784101828932762, "global_step": 196518, "epoch": 2208, "lr": 1.7713021996051727e-05}
{"train_loss": 0.0835907906293869, "global_step": 196519, "epoch": 2208, "lr": 1.7712579335852053e-05}
{"train_loss": 0.05712419003248215, "global_step": 196520, "epoch": 2208, "lr": 1.771213667999302e-05}
{"train_loss": 0.06875232607126236, "global_step": 196521, "epoch": 2208, "lr": 1.7711694028474705e-05}
{"train_loss": 0.10449905693531036, "global_step": 196522, "epoch": 2208, "lr": 1.7711251381297143e-05}
{"train_loss": 0.03778911381959915, "global_step": 196523, "epoch": 2208, "lr": 1.7710808738460432e-05}
{"train_loss": 0.04955988749861717, "global_step": 196524, "epoch": 2208, "lr": 1.771036609996459e-05}
{"train_loss": 0.053077444434165955, "global_step": 196525, "epoch": 2208, "lr": 1.7709923465809714e-05}
{"train_loss": 0.09332377463579178, "global_step": 196526, "epoch": 2208, "lr": 1.770948083599583e-05}
{"train_loss": 0.04858949035406113, "global_step": 196527, "epoch": 2208, "lr": 1.7709038210523026e-05}
{"train_loss": 0.05609160289168358, "global_step": 196528, "epoch": 2208, "lr": 1.7708595589391337e-05}
{"train_loss": 0.04546193778514862, "global_step": 196529, "epoch": 2208, "lr": 1.7708152972600845e-05}
{"train_loss": 0.14586524665355682, "global_step": 196530, "epoch": 2208, "lr": 1.770771036015158e-05}
{"train_loss": 0.042465388774871826, "global_step": 196531, "epoch": 2208, "lr": 1.770726775204364e-05}
{"train_loss": 0.05961868539452553, "global_step": 196532, "epoch": 2208, "lr": 1.7706825148277044e-05}
{"train_loss": 0.060139499604701996, "global_step": 196533, "epoch": 2208, "lr": 1.770638254885189e-05}
{"train_loss": 0.07738573849201202, "global_step": 196534, "epoch": 2208, "lr": 1.770593995376821e-05}
{"train_loss": 0.12073744833469391, "global_step": 196535, "epoch": 2208, "lr": 1.770549736302608e-05}
{"train_loss": 0.0665641501545906, "global_step": 196536, "epoch": 2208, "lr": 1.7705054776625534e-05}
{"train_loss": 0.05921465903520584, "global_step": 196537, "epoch": 2208, "lr": 1.7704612194566673e-05}
{"train_loss": 0.07739738374948502, "global_step": 196538, "epoch": 2208, "lr": 1.7704169616849515e-05}
{"train_loss": 0.02675667218863964, "global_step": 196539, "epoch": 2208, "lr": 1.7703727043474154e-05}
{"train_loss": 0.08885570615530014, "global_step": 196540, "epoch": 2208, "lr": 1.7703284474440633e-05}
{"train_loss": 0.10741657763719559, "global_step": 196541, "epoch": 2208, "lr": 1.7702841909748996e-05}
{"train_loss": 0.07920482754707336, "global_step": 196542, "epoch": 2208, "lr": 1.7702399349399335e-05}
{"train_loss": 0.06930699199438095, "global_step": 196543, "epoch": 2208, "lr": 1.7701956793391676e-05}
{"train_loss": 0.06373836100101471, "global_step": 196544, "epoch": 2208, "lr": 1.7701514241726115e-05}
{"train_loss": 0.052045874297618866, "global_step": 196545, "epoch": 2208, "lr": 1.7701071694402665e-05}
{"train_loss": 0.08341199904680252, "global_step": 196546, "epoch": 2208, "lr": 1.7700629151421438e-05}
{"train_loss": 0.030489711090922356, "global_step": 196547, "epoch": 2208, "lr": 1.770018661278245e-05}
{"train_loss": 0.08529822528362274, "global_step": 196548, "epoch": 2208, "lr": 1.769974407848579e-05}
{"train_loss": 0.09877965599298477, "global_step": 196549, "epoch": 2208, "lr": 1.7699301548531494e-05}
{"train_loss": 0.07765423506498337, "global_step": 196550, "epoch": 2208, "lr": 1.769885902291965e-05}
{"train_loss": 0.05932383984327316, "global_step": 196551, "epoch": 2208, "lr": 1.7698416501650283e-05}
{"train_loss": 0.09727238863706589, "global_step": 196552, "epoch": 2208, "lr": 1.769797398472348e-05}
{"train_loss": 0.07251309603452682, "global_step": 196553, "epoch": 2208, "lr": 1.7697531472139285e-05}
{"train_loss": 0.09447882324457169, "global_step": 196554, "epoch": 2208, "lr": 1.7697088963897772e-05}
{"train_loss": 0.06202871352434158, "global_step": 196555, "epoch": 2208, "lr": 1.7696646459998977e-05}
{"train_loss": 0.06047374755144119, "global_step": 196556, "epoch": 2208, "lr": 1.769620396044298e-05}
{"train_loss": 0.11110983788967133, "global_step": 196557, "epoch": 2208, "lr": 1.7695761465229842e-05}
{"train_loss": 0.05571621283888817, "global_step": 196558, "epoch": 2208, "lr": 1.7695318974359603e-05}
{"train_loss": 0.047714054584503174, "global_step": 196559, "epoch": 2208, "lr": 1.7694876487832356e-05}
{"train_loss": 0.09481074661016464, "global_step": 196560, "epoch": 2208, "lr": 1.7694434005648115e-05}
{"train_loss": 0.05787914991378784, "global_step": 196561, "epoch": 2208, "lr": 1.7693991527806984e-05}
{"train_loss": 0.04024820774793625, "global_step": 196562, "epoch": 2208, "lr": 1.7693549054308984e-05}
{"train_loss": 0.0707547664642334, "global_step": 196563, "epoch": 2208, "lr": 1.769310658515421e-05}
{"train_loss": 0.058202020823955536, "global_step": 196564, "epoch": 2208, "lr": 1.769266412034269e-05}
{"train_loss": 0.05795175954699516, "global_step": 196565, "epoch": 2208, "lr": 1.769222165987451e-05}
{"train_loss": 0.08589771389961243, "global_step": 196566, "epoch": 2208, "lr": 1.7691779203749698e-05}
{"train_loss": 0.046949613839387894, "global_step": 196567, "epoch": 2208, "lr": 1.7691336751968356e-05}
{"train_loss": 0.03773735463619232, "global_step": 196568, "epoch": 2208, "lr": 1.7690894304530497e-05}
{"train_loss": 0.027117984369397163, "global_step": 196569, "epoch": 2208, "lr": 1.7690451861436226e-05}
{"train_loss": 0.1253291368484497, "global_step": 196570, "epoch": 2208, "lr": 1.769000942268556e-05}
{"train_loss": 0.05450965464115143, "global_step": 196571, "epoch": 2208, "lr": 1.7689566988278593e-05}
{"train_loss": 0.09958872199058533, "global_step": 196572, "epoch": 2208, "lr": 1.7689124558215358e-05}
{"train_loss": 0.05661710351705551, "global_step": 196573, "epoch": 2208, "lr": 1.7688682132495926e-05}
{"train_loss": 0.03475324809551239, "global_step": 196574, "epoch": 2208, "lr": 1.768823971112037e-05}
{"train_loss": 0.04891166463494301, "global_step": 196575, "epoch": 2208, "lr": 1.7687797294088737e-05}
{"train_loss": 0.09567613154649734, "global_step": 196576, "epoch": 2208, "lr": 1.768735488140107e-05}
{"train_loss": 0.0831519365310669, "global_step": 196577, "epoch": 2208, "lr": 1.7686912473057467e-05}
{"train_loss": 0.07177254557609558, "global_step": 196578, "epoch": 2208, "lr": 1.7686470069057938e-05}
{"train_loss": 0.0740802213549614, "global_step": 196579, "epoch": 2208, "lr": 1.768602766940259e-05}
{"train_loss": 0.099610835313797, "global_step": 196580, "epoch": 2208, "lr": 1.7685585274091464e-05}
{"train_loss": 0.04251693934202194, "global_step": 196581, "epoch": 2208, "lr": 1.7685142883124594e-05}
{"train_loss": 0.050071071833372116, "global_step": 196582, "epoch": 2208, "lr": 1.7684700496502088e-05}
{"train_loss": 0.06573952734470367, "global_step": 196583, "epoch": 2208, "lr": 1.7684258114223956e-05}
{"train_loss": 0.051455359905958176, "global_step": 196584, "epoch": 2208, "lr": 1.7683815736290304e-05}
{"train_loss": 0.08244713395833969, "global_step": 196585, "epoch": 2208, "lr": 1.768337336270115e-05}
{"train_loss": 0.06870397180318832, "global_step": 196586, "epoch": 2208, "lr": 1.7682930993456587e-05}
{"train_loss": 0.051741987466812134, "global_step": 196587, "epoch": 2208, "lr": 1.7682488628556648e-05}
{"train_loss": 0.08940371125936508, "global_step": 196588, "epoch": 2208, "lr": 1.7682046268001418e-05}
{"train_loss": 0.10599876195192337, "global_step": 196589, "epoch": 2208, "lr": 1.768160391179093e-05}
{"train_loss": 0.0693163126707077, "global_step": 196590, "epoch": 2208, "lr": 1.7681161559925268e-05}
{"train_loss": 0.07534532994031906, "global_step": 196591, "epoch": 2208, "lr": 1.768071921240447e-05}
{"train_loss": 0.05002962425351143, "global_step": 196592, "epoch": 2208, "lr": 1.7680276869228597e-05}
{"train_loss": 0.07357403635978699, "global_step": 196593, "epoch": 2208, "lr": 1.7679834530397736e-05}
{"train_loss": 0.09451223164796829, "global_step": 196594, "epoch": 2208, "lr": 1.7679392195911914e-05}
{"train_loss": 0.04667498543858528, "global_step": 196595, "epoch": 2208, "lr": 1.7678949865771215e-05}
{"train_loss": 0.038027532398700714, "global_step": 196596, "epoch": 2208, "lr": 1.7678507539975677e-05}
{"train_loss": 0.05348406732082367, "global_step": 196597, "epoch": 2208, "lr": 1.7678065218525387e-05}
{"train_loss": 0.17145881056785583, "global_step": 196598, "epoch": 2208, "lr": 1.7677622901420365e-05}
{"train_loss": 0.11639811843633652, "global_step": 196599, "epoch": 2208, "lr": 1.7677180588660708e-05}
{"train_loss": 0.0708423557827312, "global_step": 196600, "epoch": 2208, "lr": 1.7676738280246448e-05, "val_loss": 8.167121887207031}
{"train_loss": 0.041775140911340714, "global_step": 196601, "epoch": 2209, "lr": 1.767629597617767e-05}
{"train_loss": 0.13217954337596893, "global_step": 196602, "epoch": 2209, "lr": 1.7675853676454406e-05}
{"train_loss": 0.04223771020770073, "global_step": 196603, "epoch": 2209, "lr": 1.7675411381076745e-05}
{"train_loss": 0.09497474879026413, "global_step": 196604, "epoch": 2209, "lr": 1.767496909004472e-05}
{"train_loss": 0.03733288496732712, "global_step": 196605, "epoch": 2209, "lr": 1.767452680335841e-05}
{"train_loss": 0.07177922129631042, "global_step": 196606, "epoch": 2209, "lr": 1.7674084521017854e-05}
{"train_loss": 0.06334375590085983, "global_step": 196607, "epoch": 2209, "lr": 1.7673642243023136e-05}
{"train_loss": 0.07634717971086502, "global_step": 196608, "epoch": 2209, "lr": 1.767319996937429e-05}
{"train_loss": 0.06289985775947571, "global_step": 196609, "epoch": 2209, "lr": 1.7672757700071386e-05}
{"train_loss": 0.05512459948658943, "global_step": 196610, "epoch": 2209, "lr": 1.76723154351145e-05}
{"train_loss": 0.10595747828483582, "global_step": 196611, "epoch": 2209, "lr": 1.7671873174503667e-05}
{"train_loss": 0.1033734455704689, "global_step": 196612, "epoch": 2209, "lr": 1.767143091823897e-05}
{"train_loss": 0.07112416625022888, "global_step": 196613, "epoch": 2209, "lr": 1.7670988666320455e-05}
{"train_loss": 0.06266079843044281, "global_step": 196614, "epoch": 2209, "lr": 1.767054641874816e-05}
{"train_loss": 0.05718213692307472, "global_step": 196615, "epoch": 2209, "lr": 1.767010417552219e-05}
{"train_loss": 0.06471489369869232, "global_step": 196616, "epoch": 2209, "lr": 1.7669661936642563e-05}
{"train_loss": 0.09836039692163467, "global_step": 196617, "epoch": 2209, "lr": 1.766921970210937e-05}
{"train_loss": 0.07169081270694733, "global_step": 196618, "epoch": 2209, "lr": 1.7668777471922655e-05}
{"train_loss": 0.10038953274488449, "global_step": 196619, "epoch": 2209, "lr": 1.7668335246082462e-05}
{"train_loss": 0.07638709247112274, "global_step": 196620, "epoch": 2209, "lr": 1.7667893024588887e-05}
{"train_loss": 0.10086371004581451, "global_step": 196621, "epoch": 2209, "lr": 1.7667450807441953e-05}
{"train_loss": 0.028167104348540306, "global_step": 196622, "epoch": 2209, "lr": 1.7667008594641755e-05}
{"train_loss": 0.05988384038209915, "global_step": 196623, "epoch": 2209, "lr": 1.766656638618831e-05}
{"train_loss": 0.02378271520137787, "global_step": 196624, "epoch": 2209, "lr": 1.7666124182081722e-05}
{"train_loss": 0.06408900767564774, "global_step": 196625, "epoch": 2209, "lr": 1.7665681982322012e-05}
{"train_loss": 0.03715136647224426, "global_step": 196626, "epoch": 2209, "lr": 1.7665239786909276e-05}
{"train_loss": 0.08328340202569962, "global_step": 196627, "epoch": 2209, "lr": 1.7664797595843534e-05}
{"train_loss": 0.06682366132736206, "global_step": 196628, "epoch": 2209, "lr": 1.7664355409124872e-05}
{"train_loss": 0.07587476074695587, "global_step": 196629, "epoch": 2209, "lr": 1.7663913226753354e-05}
{"train_loss": 0.039047133177518845, "global_step": 196630, "epoch": 2209, "lr": 1.7663471048729015e-05}
{"train_loss": 0.0980830192565918, "global_step": 196631, "epoch": 2209, "lr": 1.766302887505194e-05}
{"train_loss": 0.06522353738546371, "global_step": 196632, "epoch": 2209, "lr": 1.7662586705722167e-05}
{"train_loss": 0.06765840947628021, "global_step": 196633, "epoch": 2209, "lr": 1.7662144540739777e-05}
{"train_loss": 0.07360726594924927, "global_step": 196634, "epoch": 2209, "lr": 1.76617023801048e-05}
{"train_loss": 0.046584878116846085, "global_step": 196635, "epoch": 2209, "lr": 1.7661260223817333e-05}
{"train_loss": 0.05801849067211151, "global_step": 196636, "epoch": 2209, "lr": 1.76608180718774e-05}
{"train_loss": 0.042127929627895355, "global_step": 196637, "epoch": 2209, "lr": 1.766037592428509e-05}
{"train_loss": 0.06133674457669258, "global_step": 196638, "epoch": 2209, "lr": 1.7659933781040428e-05}
{"train_loss": 0.11886116117238998, "global_step": 196639, "epoch": 2209, "lr": 1.7659491642143515e-05}
{"train_loss": 0.05840012803673744, "global_step": 196640, "epoch": 2209, "lr": 1.7659049507594372e-05}
{"train_loss": 0.10881742835044861, "global_step": 196641, "epoch": 2209, "lr": 1.7658607377393093e-05}
{"train_loss": 0.08486955612897873, "global_step": 196642, "epoch": 2209, "lr": 1.7658165251539706e-05}
{"train_loss": 0.10076234489679337, "global_step": 196643, "epoch": 2209, "lr": 1.7657723130034294e-05}
{"train_loss": 0.07967662811279297, "global_step": 196644, "epoch": 2209, "lr": 1.7657281012876892e-05}
{"train_loss": 0.08007217198610306, "global_step": 196645, "epoch": 2209, "lr": 1.7656838900067578e-05}
{"train_loss": 0.08010388910770416, "global_step": 196646, "epoch": 2209, "lr": 1.7656396791606423e-05}
{"train_loss": 0.09623939543962479, "global_step": 196647, "epoch": 2209, "lr": 1.7655954687493454e-05}
{"train_loss": 0.04439055919647217, "global_step": 196648, "epoch": 2209, "lr": 1.7655512587728762e-05}
{"train_loss": 0.10727906972169876, "global_step": 196649, "epoch": 2209, "lr": 1.7655070492312382e-05}
{"train_loss": 0.09972849488258362, "global_step": 196650, "epoch": 2209, "lr": 1.7654628401244395e-05}
{"train_loss": 0.037623245269060135, "global_step": 196651, "epoch": 2209, "lr": 1.7654186314524844e-05}
{"train_loss": 0.05079498142004013, "global_step": 196652, "epoch": 2209, "lr": 1.7653744232153784e-05}
{"train_loss": 0.06857745349407196, "global_step": 196653, "epoch": 2209, "lr": 1.76533021541313e-05}
{"train_loss": 0.08229823410511017, "global_step": 196654, "epoch": 2209, "lr": 1.765286008045742e-05}
{"train_loss": 0.10796106606721878, "global_step": 196655, "epoch": 2209, "lr": 1.7652418011132232e-05}
{"train_loss": 0.06824541091918945, "global_step": 196656, "epoch": 2209, "lr": 1.765197594615577e-05}
{"train_loss": 0.07087542861700058, "global_step": 196657, "epoch": 2209, "lr": 1.7651533885528117e-05}
{"train_loss": 0.06670653820037842, "global_step": 196658, "epoch": 2209, "lr": 1.765109182924932e-05}
{"train_loss": 0.030058221891522408, "global_step": 196659, "epoch": 2209, "lr": 1.765064977731943e-05}
{"train_loss": 0.05265628173947334, "global_step": 196660, "epoch": 2209, "lr": 1.7650207729738527e-05}
{"train_loss": 0.08223246783018112, "global_step": 196661, "epoch": 2209, "lr": 1.764976568650664e-05}
{"train_loss": 0.09312088042497635, "global_step": 196662, "epoch": 2209, "lr": 1.764932364762387e-05}
{"train_loss": 0.07579488307237625, "global_step": 196663, "epoch": 2209, "lr": 1.764888161309024e-05}
{"train_loss": 0.08314167708158493, "global_step": 196664, "epoch": 2209, "lr": 1.764843958290582e-05}
{"train_loss": 0.052932076156139374, "global_step": 196665, "epoch": 2209, "lr": 1.7647997557070682e-05}
{"train_loss": 0.06619931757450104, "global_step": 196666, "epoch": 2209, "lr": 1.764755553558487e-05}
{"train_loss": 0.11219177395105362, "global_step": 196667, "epoch": 2209, "lr": 1.7647113518448465e-05}
{"train_loss": 0.07773245126008987, "global_step": 196668, "epoch": 2209, "lr": 1.7646671505661494e-05}
{"train_loss": 0.09106235951185226, "global_step": 196669, "epoch": 2209, "lr": 1.7646229497224047e-05}
{"train_loss": 0.05621270462870598, "global_step": 196670, "epoch": 2209, "lr": 1.7645787493136155e-05}
{"train_loss": 0.09887522459030151, "global_step": 196671, "epoch": 2209, "lr": 1.7645345493397906e-05}
{"train_loss": 0.07649751007556915, "global_step": 196672, "epoch": 2209, "lr": 1.7644903498009334e-05}
{"train_loss": 0.05301980674266815, "global_step": 196673, "epoch": 2209, "lr": 1.7644461506970527e-05}
{"train_loss": 0.08777458965778351, "global_step": 196674, "epoch": 2209, "lr": 1.7644019520281512e-05}
{"train_loss": 0.07697491347789764, "global_step": 196675, "epoch": 2209, "lr": 1.764357753794238e-05}
{"train_loss": 0.056777097284793854, "global_step": 196676, "epoch": 2209, "lr": 1.7643135559953156e-05}
{"train_loss": 0.0682636946439743, "global_step": 196677, "epoch": 2209, "lr": 1.7642693586313937e-05}
{"train_loss": 0.04273664951324463, "global_step": 196678, "epoch": 2209, "lr": 1.7642251617024747e-05}
{"train_loss": 0.07966459542512894, "global_step": 196679, "epoch": 2209, "lr": 1.7641809652085677e-05}
{"train_loss": 0.10649403929710388, "global_step": 196680, "epoch": 2209, "lr": 1.7641367691496753e-05}
{"train_loss": 0.07150329649448395, "global_step": 196681, "epoch": 2209, "lr": 1.7640925735258074e-05}
{"train_loss": 0.0751781091094017, "global_step": 196682, "epoch": 2209, "lr": 1.7640483783369655e-05}
{"train_loss": 0.0348731204867363, "global_step": 196683, "epoch": 2209, "lr": 1.7640041835831583e-05}
{"train_loss": 0.04692112281918526, "global_step": 196684, "epoch": 2209, "lr": 1.7639599892643927e-05}
{"train_loss": 0.06596460193395615, "global_step": 196685, "epoch": 2209, "lr": 1.7639157953806717e-05}
{"train_loss": 0.0840395838022232, "global_step": 196686, "epoch": 2209, "lr": 1.7638716019320044e-05}
{"train_loss": 0.09301455318927765, "global_step": 196687, "epoch": 2209, "lr": 1.7638274089183935e-05}
{"train_loss": 0.06854914128780365, "global_step": 196688, "epoch": 2209, "lr": 1.7637832163398486e-05}
{"train_loss": 0.07254194711031539, "global_step": 196689, "epoch": 2209, "lr": 1.7637390241963714e-05, "val_loss": 7.987027168273926}
{"train_loss": 0.024611471220850945, "global_step": 196690, "epoch": 2210, "lr": 1.763694832487972e-05}
{"train_loss": 0.08338894695043564, "global_step": 196691, "epoch": 2210, "lr": 1.7636506412146546e-05}
{"train_loss": 0.11370239406824112, "global_step": 196692, "epoch": 2210, "lr": 1.7636064503764226e-05}
{"train_loss": 0.08285261690616608, "global_step": 196693, "epoch": 2210, "lr": 1.7635622599732864e-05}
{"train_loss": 0.04755337908864021, "global_step": 196694, "epoch": 2210, "lr": 1.7635180700052482e-05}
{"train_loss": 0.061632465571165085, "global_step": 196695, "epoch": 2210, "lr": 1.7634738804723173e-05}
{"train_loss": 0.04914076253771782, "global_step": 196696, "epoch": 2210, "lr": 1.7634296913744975e-05}
{"train_loss": 0.043476201593875885, "global_step": 196697, "epoch": 2210, "lr": 1.7633855027117934e-05}
{"train_loss": 0.0571511909365654, "global_step": 196698, "epoch": 2210, "lr": 1.7633413144842148e-05}
{"train_loss": 0.05169588699936867, "global_step": 196699, "epoch": 2210, "lr": 1.7632971266917635e-05}
{"train_loss": 0.05474662408232689, "global_step": 196700, "epoch": 2210, "lr": 1.7632529393344476e-05}
{"train_loss": 0.06854233890771866, "global_step": 196701, "epoch": 2210, "lr": 1.7632087524122742e-05}
{"train_loss": 0.08556535094976425, "global_step": 196702, "epoch": 2210, "lr": 1.7631645659252467e-05}
{"train_loss": 0.04586944356560707, "global_step": 196703, "epoch": 2210, "lr": 1.7631203798733736e-05}
{"train_loss": 0.06416872888803482, "global_step": 196704, "epoch": 2210, "lr": 1.7630761942566582e-05}
{"train_loss": 0.07727106660604477, "global_step": 196705, "epoch": 2210, "lr": 1.7630320090751097e-05}
{"train_loss": 0.04516589269042015, "global_step": 196706, "epoch": 2210, "lr": 1.7629878243287295e-05}
{"train_loss": 0.09317716211080551, "global_step": 196707, "epoch": 2210, "lr": 1.7629436400175288e-05}
{"train_loss": 0.039032503962516785, "global_step": 196708, "epoch": 2210, "lr": 1.7628994561415084e-05}
{"train_loss": 0.05700835585594177, "global_step": 196709, "epoch": 2210, "lr": 1.7628552727006788e-05}
{"train_loss": 0.09775510430335999, "global_step": 196710, "epoch": 2210, "lr": 1.7628110896950417e-05}
{"train_loss": 0.09348131716251373, "global_step": 196711, "epoch": 2210, "lr": 1.762766907124607e-05}
{"train_loss": 0.06397319585084915, "global_step": 196712, "epoch": 2210, "lr": 1.762722724989378e-05}
{"train_loss": 0.0675877183675766, "global_step": 196713, "epoch": 2210, "lr": 1.762678543289362e-05}
{"train_loss": 0.03838152065873146, "global_step": 196714, "epoch": 2210, "lr": 1.7626343620245637e-05}
{"train_loss": 0.08198211342096329, "global_step": 196715, "epoch": 2210, "lr": 1.7625901811949906e-05}
{"train_loss": 0.08836415410041809, "global_step": 196716, "epoch": 2210, "lr": 1.7625460008006467e-05}
{"train_loss": 0.08851855248212814, "global_step": 196717, "epoch": 2210, "lr": 1.7625018208415405e-05}
{"train_loss": 0.062264759093523026, "global_step": 196718, "epoch": 2210, "lr": 1.762457641317675e-05}
{"train_loss": 0.07843980938196182, "global_step": 196719, "epoch": 2210, "lr": 1.7624134622290572e-05}
{"train_loss": 0.04090447723865509, "global_step": 196720, "epoch": 2210, "lr": 1.7623692835756957e-05}
{"train_loss": 0.06584547460079193, "global_step": 196721, "epoch": 2210, "lr": 1.7623251053575922e-05}
{"train_loss": 0.04836749657988548, "global_step": 196722, "epoch": 2210, "lr": 1.762280927574756e-05}
{"train_loss": 0.034764450043439865, "global_step": 196723, "epoch": 2210, "lr": 1.7622367502271902e-05}
{"train_loss": 0.024708790704607964, "global_step": 196724, "epoch": 2210, "lr": 1.7621925733149042e-05}
{"train_loss": 0.05874969810247421, "global_step": 196725, "epoch": 2210, "lr": 1.7621483968379e-05}
{"train_loss": 0.0697881430387497, "global_step": 196726, "epoch": 2210, "lr": 1.7621042207961874e-05}
{"train_loss": 0.05344224348664284, "global_step": 196727, "epoch": 2210, "lr": 1.7620600451897683e-05}
{"train_loss": 0.11317069828510284, "global_step": 196728, "epoch": 2210, "lr": 1.7620158700186528e-05}
{"train_loss": 0.034504447132349014, "global_step": 196729, "epoch": 2210, "lr": 1.7619716952828447e-05}
{"train_loss": 0.057859428226947784, "global_step": 196730, "epoch": 2210, "lr": 1.7619275209823482e-05}
{"train_loss": 0.06819546222686768, "global_step": 196731, "epoch": 2210, "lr": 1.761883347117173e-05}
{"train_loss": 0.05277504026889801, "global_step": 196732, "epoch": 2210, "lr": 1.761839173687321e-05}
{"train_loss": 0.07565995305776596, "global_step": 196733, "epoch": 2210, "lr": 1.7617950006928024e-05}
{"train_loss": 0.05375029519200325, "global_step": 196734, "epoch": 2210, "lr": 1.761750828133621e-05}
{"train_loss": 0.035563964396715164, "global_step": 196735, "epoch": 2210, "lr": 1.7617066560097807e-05}
{"train_loss": 0.08150608092546463, "global_step": 196736, "epoch": 2210, "lr": 1.76166248432129e-05}
{"train_loss": 0.052631329745054245, "global_step": 196737, "epoch": 2210, "lr": 1.761618313068155e-05}
{"train_loss": 0.029364507645368576, "global_step": 196738, "epoch": 2210, "lr": 1.7615741422503802e-05}
{"train_loss": 0.08128441870212555, "global_step": 196739, "epoch": 2210, "lr": 1.761529971867974e-05}
{"train_loss": 0.06546751409769058, "global_step": 196740, "epoch": 2210, "lr": 1.7614858019209384e-05}
{"train_loss": 0.056725915521383286, "global_step": 196741, "epoch": 2210, "lr": 1.7614416324092836e-05}
{"train_loss": 0.08996118605136871, "global_step": 196742, "epoch": 2210, "lr": 1.7613974633330116e-05}
{"train_loss": 0.05981412157416344, "global_step": 196743, "epoch": 2210, "lr": 1.7613532946921318e-05}
{"train_loss": 0.0662432312965393, "global_step": 196744, "epoch": 2210, "lr": 1.7613091264866472e-05}
{"train_loss": 0.08769417554140091, "global_step": 196745, "epoch": 2210, "lr": 1.761264958716567e-05}
{"train_loss": 0.05467773228883743, "global_step": 196746, "epoch": 2210, "lr": 1.761220791381893e-05}
{"train_loss": 0.029507562518119812, "global_step": 196747, "epoch": 2210, "lr": 1.7611766244826356e-05}
{"train_loss": 0.060579366981983185, "global_step": 196748, "epoch": 2210, "lr": 1.7611324580187964e-05}
{"train_loss": 0.06788235902786255, "global_step": 196749, "epoch": 2210, "lr": 1.7610882919903854e-05}
{"train_loss": 0.08139347285032272, "global_step": 196750, "epoch": 2210, "lr": 1.761044126397405e-05}
{"train_loss": 0.05430026352405548, "global_step": 196751, "epoch": 2210, "lr": 1.760999961239864e-05}
{"train_loss": 0.07191409170627594, "global_step": 196752, "epoch": 2210, "lr": 1.760955796517766e-05}
{"train_loss": 0.03820143640041351, "global_step": 196753, "epoch": 2210, "lr": 1.7609116322311193e-05}
{"train_loss": 0.06918807327747345, "global_step": 196754, "epoch": 2210, "lr": 1.760867468379927e-05}
{"train_loss": 0.09179068356752396, "global_step": 196755, "epoch": 2210, "lr": 1.7608233049641964e-05}
{"train_loss": 0.020952142775058746, "global_step": 196756, "epoch": 2210, "lr": 1.7607791419839354e-05}
{"train_loss": 0.060885656625032425, "global_step": 196757, "epoch": 2210, "lr": 1.7607349794391463e-05}
{"train_loss": 0.045935582369565964, "global_step": 196758, "epoch": 2210, "lr": 1.7606908173298392e-05}
{"train_loss": 0.07171417772769928, "global_step": 196759, "epoch": 2210, "lr": 1.7606466556560154e-05}
{"train_loss": 0.03954720497131348, "global_step": 196760, "epoch": 2210, "lr": 1.760602494417685e-05}
{"train_loss": 0.09454315155744553, "global_step": 196761, "epoch": 2210, "lr": 1.7605583336148508e-05}
{"train_loss": 0.06330616772174835, "global_step": 196762, "epoch": 2210, "lr": 1.7605141732475218e-05}
{"train_loss": 0.07101374119520187, "global_step": 196763, "epoch": 2210, "lr": 1.7604700133157e-05}
{"train_loss": 0.03421034663915634, "global_step": 196764, "epoch": 2210, "lr": 1.7604258538193952e-05}
{"train_loss": 0.1164344847202301, "global_step": 196765, "epoch": 2210, "lr": 1.7603816947586105e-05}
{"train_loss": 0.05143672972917557, "global_step": 196766, "epoch": 2210, "lr": 1.7603375361333542e-05}
{"train_loss": 0.08640069514513016, "global_step": 196767, "epoch": 2210, "lr": 1.760293377943631e-05}
{"train_loss": 0.09919803589582443, "global_step": 196768, "epoch": 2210, "lr": 1.7602492201894455e-05}
{"train_loss": 0.07541899383068085, "global_step": 196769, "epoch": 2210, "lr": 1.7602050628708066e-05}
{"train_loss": 0.03739400580525398, "global_step": 196770, "epoch": 2210, "lr": 1.7601609059877168e-05}
{"train_loss": 0.0867699459195137, "global_step": 196771, "epoch": 2210, "lr": 1.7601167495401856e-05}
{"train_loss": 0.03420281410217285, "global_step": 196772, "epoch": 2210, "lr": 1.760072593528216e-05}
{"train_loss": 0.042215194553136826, "global_step": 196773, "epoch": 2210, "lr": 1.760028437951816e-05}
{"train_loss": 0.06527244299650192, "global_step": 196774, "epoch": 2210, "lr": 1.7599842828109898e-05}
{"train_loss": 0.07722261548042297, "global_step": 196775, "epoch": 2210, "lr": 1.7599401281057454e-05}
{"train_loss": 0.06820625811815262, "global_step": 196776, "epoch": 2210, "lr": 1.759895973836086e-05}
{"train_loss": 0.09030943363904953, "global_step": 196777, "epoch": 2210, "lr": 1.759851820002021e-05}
{"train_loss": 0.06397543320160233, "global_step": 196778, "epoch": 2210, "lr": 1.7598076666035523e-05, "val_loss": 8.109126091003418, "train_action_mse_error": 16.858238220214844}
{"train_loss": 0.08303844928741455, "global_step": 196779, "epoch": 2211, "lr": 1.75976351364069e-05}
{"train_loss": 0.07313317805528641, "global_step": 196780, "epoch": 2211, "lr": 1.7597193611134362e-05}
{"train_loss": 0.06559321284294128, "global_step": 196781, "epoch": 2211, "lr": 1.7596752090218004e-05}
{"train_loss": 0.09415098279714584, "global_step": 196782, "epoch": 2211, "lr": 1.7596310573657848e-05}
{"train_loss": 0.02979472279548645, "global_step": 196783, "epoch": 2211, "lr": 1.7595869061453996e-05}
{"train_loss": 0.08461612462997437, "global_step": 196784, "epoch": 2211, "lr": 1.7595427553606458e-05}
{"train_loss": 0.07930024713277817, "global_step": 196785, "epoch": 2211, "lr": 1.7594986050115347e-05}
{"train_loss": 0.0730462297797203, "global_step": 196786, "epoch": 2211, "lr": 1.7594544550980674e-05}
{"train_loss": 0.08522240072488785, "global_step": 196787, "epoch": 2211, "lr": 1.7594103056202533e-05}
{"train_loss": 0.05144472047686577, "global_step": 196788, "epoch": 2211, "lr": 1.7593661565780957e-05}
{"train_loss": 0.036658208817243576, "global_step": 196789, "epoch": 2211, "lr": 1.7593220079716033e-05}
{"train_loss": 0.09998910129070282, "global_step": 196790, "epoch": 2211, "lr": 1.7592778598007794e-05}
{"train_loss": 0.06433715671300888, "global_step": 196791, "epoch": 2211, "lr": 1.7592337120656305e-05}
{"train_loss": 0.03946681320667267, "global_step": 196792, "epoch": 2211, "lr": 1.759189564766165e-05}
{"train_loss": 0.053032662719488144, "global_step": 196793, "epoch": 2211, "lr": 1.7591454179023852e-05}
{"train_loss": 0.07074140757322311, "global_step": 196794, "epoch": 2211, "lr": 1.7591012714743004e-05}
{"train_loss": 0.056771766394376755, "global_step": 196795, "epoch": 2211, "lr": 1.7590571254819137e-05}
{"train_loss": 0.0896955132484436, "global_step": 196796, "epoch": 2211, "lr": 1.7590129799252336e-05}
{"train_loss": 0.12145178020000458, "global_step": 196797, "epoch": 2211, "lr": 1.7589688348042632e-05}
{"train_loss": 0.07724910974502563, "global_step": 196798, "epoch": 2211, "lr": 1.7589246901190116e-05}
{"train_loss": 0.0856444239616394, "global_step": 196799, "epoch": 2211, "lr": 1.7588805458694813e-05}
{"train_loss": 0.04124434292316437, "global_step": 196800, "epoch": 2211, "lr": 1.7588364020556813e-05}
{"train_loss": 0.05843740329146385, "global_step": 196801, "epoch": 2211, "lr": 1.758792258677615e-05}
{"train_loss": 0.1443202942609787, "global_step": 196802, "epoch": 2211, "lr": 1.7587481157352913e-05}
{"train_loss": 0.08011934906244278, "global_step": 196803, "epoch": 2211, "lr": 1.7587039732287125e-05}
{"train_loss": 0.05840779468417168, "global_step": 196804, "epoch": 2211, "lr": 1.7586598311578884e-05}
{"train_loss": 0.02201184630393982, "global_step": 196805, "epoch": 2211, "lr": 1.758615689522821e-05}
{"train_loss": 0.06874938309192657, "global_step": 196806, "epoch": 2211, "lr": 1.7585715483235205e-05}
{"train_loss": 0.07588060200214386, "global_step": 196807, "epoch": 2211, "lr": 1.75852740755999e-05}
{"train_loss": 0.035998567938804626, "global_step": 196808, "epoch": 2211, "lr": 1.7584832672322345e-05}
{"train_loss": 0.053974397480487823, "global_step": 196809, "epoch": 2211, "lr": 1.7584391273402628e-05}
{"train_loss": 0.07236289232969284, "global_step": 196810, "epoch": 2211, "lr": 1.7583949878840776e-05}
{"train_loss": 0.07040750980377197, "global_step": 196811, "epoch": 2211, "lr": 1.7583508488636892e-05}
{"train_loss": 0.11772021651268005, "global_step": 196812, "epoch": 2211, "lr": 1.758306710279099e-05}
{"train_loss": 0.06380025297403336, "global_step": 196813, "epoch": 2211, "lr": 1.7582625721303163e-05}
{"train_loss": 0.12041392177343369, "global_step": 196814, "epoch": 2211, "lr": 1.7582184344173443e-05}
{"train_loss": 0.03350042551755905, "global_step": 196815, "epoch": 2211, "lr": 1.7581742971401917e-05}
{"train_loss": 0.056148093193769455, "global_step": 196816, "epoch": 2211, "lr": 1.7581301602988613e-05}
{"train_loss": 0.06812892854213715, "global_step": 196817, "epoch": 2211, "lr": 1.7580860238933632e-05}
{"train_loss": 0.15820591151714325, "global_step": 196818, "epoch": 2211, "lr": 1.758041887923699e-05}
{"train_loss": 0.05027519911527634, "global_step": 196819, "epoch": 2211, "lr": 1.7579977523898777e-05}
{"train_loss": 0.06716464459896088, "global_step": 196820, "epoch": 2211, "lr": 1.7579536172919027e-05}
{"train_loss": 0.06980384886264801, "global_step": 196821, "epoch": 2211, "lr": 1.757909482629783e-05}
{"train_loss": 0.05732313543558121, "global_step": 196822, "epoch": 2211, "lr": 1.7578653484035212e-05}
{"train_loss": 0.05956569314002991, "global_step": 196823, "epoch": 2211, "lr": 1.757821214613127e-05}
{"train_loss": 0.05638047680258751, "global_step": 196824, "epoch": 2211, "lr": 1.757777081258602e-05}
{"train_loss": 0.04132812097668648, "global_step": 196825, "epoch": 2211, "lr": 1.7577329483399556e-05}
{"train_loss": 0.06978274881839752, "global_step": 196826, "epoch": 2211, "lr": 1.7576888158571918e-05}
{"train_loss": 0.11242149025201797, "global_step": 196827, "epoch": 2211, "lr": 1.7576446838103166e-05}
{"train_loss": 0.14039760828018188, "global_step": 196828, "epoch": 2211, "lr": 1.7576005521993384e-05}
{"train_loss": 0.03372825309634209, "global_step": 196829, "epoch": 2211, "lr": 1.7575564210242596e-05}
{"train_loss": 0.08307396620512009, "global_step": 196830, "epoch": 2211, "lr": 1.7575122902850895e-05}
{"train_loss": 0.03892768919467926, "global_step": 196831, "epoch": 2211, "lr": 1.757468159981831e-05}
{"train_loss": 0.08499547839164734, "global_step": 196832, "epoch": 2211, "lr": 1.7574240301144924e-05}
{"train_loss": 0.11621119827032089, "global_step": 196833, "epoch": 2211, "lr": 1.7573799006830777e-05}
{"train_loss": 0.09997101128101349, "global_step": 196834, "epoch": 2211, "lr": 1.757335771687595e-05}
{"train_loss": 0.05511830374598503, "global_step": 196835, "epoch": 2211, "lr": 1.7572916431280472e-05}
{"train_loss": 0.07562743872404099, "global_step": 196836, "epoch": 2211, "lr": 1.7572475150044436e-05}
{"train_loss": 0.056185804307460785, "global_step": 196837, "epoch": 2211, "lr": 1.7572033873167872e-05}
{"train_loss": 0.09909708797931671, "global_step": 196838, "epoch": 2211, "lr": 1.7571592600650864e-05}
{"train_loss": 0.07251647859811783, "global_step": 196839, "epoch": 2211, "lr": 1.757115133249345e-05}
{"train_loss": 0.03890504688024521, "global_step": 196840, "epoch": 2211, "lr": 1.7570710068695715e-05}
{"train_loss": 0.051121786236763, "global_step": 196841, "epoch": 2211, "lr": 1.7570268809257683e-05}
{"train_loss": 0.08512657135725021, "global_step": 196842, "epoch": 2211, "lr": 1.7569827554179457e-05}
{"train_loss": 0.05420219153165817, "global_step": 196843, "epoch": 2211, "lr": 1.756938630346105e-05}
{"train_loss": 0.05838814005255699, "global_step": 196844, "epoch": 2211, "lr": 1.756894505710256e-05}
{"train_loss": 0.08386310189962387, "global_step": 196845, "epoch": 2211, "lr": 1.756850381510403e-05}
{"train_loss": 0.07631305605173111, "global_step": 196846, "epoch": 2211, "lr": 1.7568062577465506e-05}
{"train_loss": 0.1062542200088501, "global_step": 196847, "epoch": 2211, "lr": 1.7567621344187073e-05}
{"train_loss": 0.041152603924274445, "global_step": 196848, "epoch": 2211, "lr": 1.7567180115268765e-05}
{"train_loss": 0.06710131466388702, "global_step": 196849, "epoch": 2211, "lr": 1.7566738890710672e-05}
{"train_loss": 0.0982353687286377, "global_step": 196850, "epoch": 2211, "lr": 1.7566297670512816e-05}
{"train_loss": 0.08578996360301971, "global_step": 196851, "epoch": 2211, "lr": 1.756585645467529e-05}
{"train_loss": 0.05827798321843147, "global_step": 196852, "epoch": 2211, "lr": 1.7565415243198125e-05}
{"train_loss": 0.13735873997211456, "global_step": 196853, "epoch": 2211, "lr": 1.7564974036081412e-05}
{"train_loss": 0.1132931262254715, "global_step": 196854, "epoch": 2211, "lr": 1.7564532833325174e-05}
{"train_loss": 0.03598141297698021, "global_step": 196855, "epoch": 2211, "lr": 1.7564091634929513e-05}
{"train_loss": 0.08176127076148987, "global_step": 196856, "epoch": 2211, "lr": 1.756365044089444e-05}
{"train_loss": 0.07641427963972092, "global_step": 196857, "epoch": 2211, "lr": 1.7563209251220058e-05}
{"train_loss": 0.08409709483385086, "global_step": 196858, "epoch": 2211, "lr": 1.7562768065906393e-05}
{"train_loss": 0.10498257726430893, "global_step": 196859, "epoch": 2211, "lr": 1.7562326884953532e-05}
{"train_loss": 0.1280730962753296, "global_step": 196860, "epoch": 2211, "lr": 1.756188570836151e-05}
{"train_loss": 0.039737775921821594, "global_step": 196861, "epoch": 2211, "lr": 1.7561444536130407e-05}
{"train_loss": 0.04257936030626297, "global_step": 196862, "epoch": 2211, "lr": 1.756100336826026e-05}
{"train_loss": 0.055428534746170044, "global_step": 196863, "epoch": 2211, "lr": 1.7560562204751135e-05}
{"train_loss": 0.1055235043168068, "global_step": 196864, "epoch": 2211, "lr": 1.7560121045603118e-05}
{"train_loss": 0.04336996003985405, "global_step": 196865, "epoch": 2211, "lr": 1.7559679890816226e-05}
{"train_loss": 0.10251370817422867, "global_step": 196866, "epoch": 2211, "lr": 1.7559238740390564e-05}
{"train_loss": 0.0736262717106369, "global_step": 196867, "epoch": 2211, "lr": 1.755879759432615e-05, "val_loss": 8.202827453613281}
{"train_loss": 0.04022960364818573, "global_step": 196868, "epoch": 2212, "lr": 1.755835645262307e-05}
{"train_loss": 0.0841895118355751, "global_step": 196869, "epoch": 2212, "lr": 1.7557915315281358e-05}
{"train_loss": 0.07856321334838867, "global_step": 196870, "epoch": 2212, "lr": 1.755747418230111e-05}
{"train_loss": 0.046014223247766495, "global_step": 196871, "epoch": 2212, "lr": 1.755703305368234e-05}
{"train_loss": 0.0929950550198555, "global_step": 196872, "epoch": 2212, "lr": 1.755659192942516e-05}
{"train_loss": 0.03108249418437481, "global_step": 196873, "epoch": 2212, "lr": 1.7556150809529577e-05}
{"train_loss": 0.034058380872011185, "global_step": 196874, "epoch": 2212, "lr": 1.7555709693995693e-05}
{"train_loss": 0.1010848730802536, "global_step": 196875, "epoch": 2212, "lr": 1.7555268582823524e-05}
{"train_loss": 0.08318854868412018, "global_step": 196876, "epoch": 2212, "lr": 1.755482747601318e-05}
{"train_loss": 0.04057077318429947, "global_step": 196877, "epoch": 2212, "lr": 1.7554386373564675e-05}
{"train_loss": 0.048348672688007355, "global_step": 196878, "epoch": 2212, "lr": 1.75539452754781e-05}
{"train_loss": 0.06849738210439682, "global_step": 196879, "epoch": 2212, "lr": 1.755350418175349e-05}
{"train_loss": 0.05185728147625923, "global_step": 196880, "epoch": 2212, "lr": 1.755306309239093e-05}
{"train_loss": 0.06960289180278778, "global_step": 196881, "epoch": 2212, "lr": 1.7552622007390453e-05}
{"train_loss": 0.06657499074935913, "global_step": 196882, "epoch": 2212, "lr": 1.7552180926752137e-05}
{"train_loss": 0.0524488240480423, "global_step": 196883, "epoch": 2212, "lr": 1.755173985047604e-05}
{"train_loss": 0.024917948991060257, "global_step": 196884, "epoch": 2212, "lr": 1.7551298778562202e-05}
{"train_loss": 0.06104106456041336, "global_step": 196885, "epoch": 2212, "lr": 1.755085771101071e-05}
{"train_loss": 0.06436818838119507, "global_step": 196886, "epoch": 2212, "lr": 1.7550416647821594e-05}
{"train_loss": 0.06282119452953339, "global_step": 196887, "epoch": 2212, "lr": 1.754997558899495e-05}
{"train_loss": 0.03450934588909149, "global_step": 196888, "epoch": 2212, "lr": 1.7549534534530792e-05}
{"train_loss": 0.040052127093076706, "global_step": 196889, "epoch": 2212, "lr": 1.7549093484429224e-05}
{"train_loss": 0.07926426082849503, "global_step": 196890, "epoch": 2212, "lr": 1.754865243869027e-05}
{"train_loss": 0.06349148601293564, "global_step": 196891, "epoch": 2212, "lr": 1.7548211397314017e-05}
{"train_loss": 0.0962194874882698, "global_step": 196892, "epoch": 2212, "lr": 1.7547770360300497e-05}
{"train_loss": 0.06318236142396927, "global_step": 196893, "epoch": 2212, "lr": 1.7547329327649803e-05}
{"train_loss": 0.08132889121770859, "global_step": 196894, "epoch": 2212, "lr": 1.7546888299361952e-05}
{"train_loss": 0.05600937455892563, "global_step": 196895, "epoch": 2212, "lr": 1.7546447275437046e-05}
{"train_loss": 0.05228830873966217, "global_step": 196896, "epoch": 2212, "lr": 1.754600625587511e-05}
{"train_loss": 0.11232766509056091, "global_step": 196897, "epoch": 2212, "lr": 1.754556524067623e-05}
{"train_loss": 0.06773873418569565, "global_step": 196898, "epoch": 2212, "lr": 1.7545124229840444e-05}
{"train_loss": 0.07763444632291794, "global_step": 196899, "epoch": 2212, "lr": 1.7544683223367814e-05}
{"train_loss": 0.08941401541233063, "global_step": 196900, "epoch": 2212, "lr": 1.754424222125842e-05}
{"train_loss": 0.05877966433763504, "global_step": 196901, "epoch": 2212, "lr": 1.7543801223512295e-05}
{"train_loss": 0.03486297279596329, "global_step": 196902, "epoch": 2212, "lr": 1.754336023012953e-05}
{"train_loss": 0.0652557834982872, "global_step": 196903, "epoch": 2212, "lr": 1.7542919241110144e-05}
{"train_loss": 0.07618970423936844, "global_step": 196904, "epoch": 2212, "lr": 1.7542478256454238e-05}
{"train_loss": 0.09103547036647797, "global_step": 196905, "epoch": 2212, "lr": 1.7542037276161828e-05}
{"train_loss": 0.055479492992162704, "global_step": 196906, "epoch": 2212, "lr": 1.754159630023301e-05}
{"train_loss": 0.12047652900218964, "global_step": 196907, "epoch": 2212, "lr": 1.754115532866782e-05}
{"train_loss": 0.13913923501968384, "global_step": 196908, "epoch": 2212, "lr": 1.7540714361466338e-05}
{"train_loss": 0.10349967330694199, "global_step": 196909, "epoch": 2212, "lr": 1.7540273398628594e-05}
{"train_loss": 0.08565308898687363, "global_step": 196910, "epoch": 2212, "lr": 1.753983244015468e-05}
{"train_loss": 0.025127144530415535, "global_step": 196911, "epoch": 2212, "lr": 1.7539391486044628e-05}
{"train_loss": 0.07291658967733383, "global_step": 196912, "epoch": 2212, "lr": 1.7538950536298527e-05}
{"train_loss": 0.04222291707992554, "global_step": 196913, "epoch": 2212, "lr": 1.7538509590916397e-05}
{"train_loss": 0.045964211225509644, "global_step": 196914, "epoch": 2212, "lr": 1.7538068649898338e-05}
{"train_loss": 0.07521093636751175, "global_step": 196915, "epoch": 2212, "lr": 1.753762771324437e-05}
{"train_loss": 0.049946803599596024, "global_step": 196916, "epoch": 2212, "lr": 1.753718678095459e-05}
{"train_loss": 0.04538640379905701, "global_step": 196917, "epoch": 2212, "lr": 1.7536745853029023e-05}
{"train_loss": 0.037983108311891556, "global_step": 196918, "epoch": 2212, "lr": 1.753630492946776e-05}
{"train_loss": 0.11256330460309982, "global_step": 196919, "epoch": 2212, "lr": 1.7535864010270838e-05}
{"train_loss": 0.05292750522494316, "global_step": 196920, "epoch": 2212, "lr": 1.753542309543833e-05}
{"train_loss": 0.06788754463195801, "global_step": 196921, "epoch": 2212, "lr": 1.753498218497028e-05}
{"train_loss": 0.03347298130393028, "global_step": 196922, "epoch": 2212, "lr": 1.753454127886676e-05}
{"train_loss": 0.09039826691150665, "global_step": 196923, "epoch": 2212, "lr": 1.753410037712784e-05}
{"train_loss": 0.09255548566579819, "global_step": 196924, "epoch": 2212, "lr": 1.7533659479753538e-05}
{"train_loss": 0.06302196532487869, "global_step": 196925, "epoch": 2212, "lr": 1.753321858674396e-05}
{"train_loss": 0.13017244637012482, "global_step": 196926, "epoch": 2212, "lr": 1.7532777698099123e-05}
{"train_loss": 0.08983099460601807, "global_step": 196927, "epoch": 2212, "lr": 1.7532336813819132e-05}
{"train_loss": 0.08460590243339539, "global_step": 196928, "epoch": 2212, "lr": 1.7531895933904003e-05}
{"train_loss": 0.06958696246147156, "global_step": 196929, "epoch": 2212, "lr": 1.7531455058353834e-05}
{"train_loss": 0.07807915657758713, "global_step": 196930, "epoch": 2212, "lr": 1.7531014187168644e-05}
{"train_loss": 0.042414065450429916, "global_step": 196931, "epoch": 2212, "lr": 1.7530573320348532e-05}
{"train_loss": 0.08080268651247025, "global_step": 196932, "epoch": 2212, "lr": 1.7530132457893524e-05}
{"train_loss": 0.0413995087146759, "global_step": 196933, "epoch": 2212, "lr": 1.752969159980371e-05}
{"train_loss": 0.09521476924419403, "global_step": 196934, "epoch": 2212, "lr": 1.7529250746079117e-05}
{"train_loss": 0.04224696755409241, "global_step": 196935, "epoch": 2212, "lr": 1.7528809896719834e-05}
{"train_loss": 0.07292943447828293, "global_step": 196936, "epoch": 2212, "lr": 1.752836905172589e-05}
{"train_loss": 0.09820328652858734, "global_step": 196937, "epoch": 2212, "lr": 1.7527928211097367e-05}
{"train_loss": 0.07819956541061401, "global_step": 196938, "epoch": 2212, "lr": 1.7527487374834328e-05}
{"train_loss": 0.021055208519101143, "global_step": 196939, "epoch": 2212, "lr": 1.752704654293681e-05}
{"train_loss": 0.03900400176644325, "global_step": 196940, "epoch": 2212, "lr": 1.7526605715404898e-05}
{"train_loss": 0.05627115070819855, "global_step": 196941, "epoch": 2212, "lr": 1.7526164892238628e-05}
{"train_loss": 0.05225236341357231, "global_step": 196942, "epoch": 2212, "lr": 1.7525724073438078e-05}
{"train_loss": 0.0783303901553154, "global_step": 196943, "epoch": 2212, "lr": 1.7525283259003288e-05}
{"train_loss": 0.02808348275721073, "global_step": 196944, "epoch": 2212, "lr": 1.7524842448934344e-05}
{"train_loss": 0.04618896543979645, "global_step": 196945, "epoch": 2212, "lr": 1.752440164323127e-05}
{"train_loss": 0.07527747005224228, "global_step": 196946, "epoch": 2212, "lr": 1.7523960841894162e-05}
{"train_loss": 0.05610770359635353, "global_step": 196947, "epoch": 2212, "lr": 1.7523520044923043e-05}
{"train_loss": 0.08329880982637405, "global_step": 196948, "epoch": 2212, "lr": 1.7523079252318007e-05}
{"train_loss": 0.09583567082881927, "global_step": 196949, "epoch": 2212, "lr": 1.7522638464079084e-05}
{"train_loss": 0.03294084593653679, "global_step": 196950, "epoch": 2212, "lr": 1.752219768020636e-05}
{"train_loss": 0.0610220767557621, "global_step": 196951, "epoch": 2212, "lr": 1.7521756900699866e-05}
{"train_loss": 0.05397752299904823, "global_step": 196952, "epoch": 2212, "lr": 1.7521316125559693e-05}
{"train_loss": 0.07999716699123383, "global_step": 196953, "epoch": 2212, "lr": 1.7520875354785864e-05}
{"train_loss": 0.05287899076938629, "global_step": 196954, "epoch": 2212, "lr": 1.752043458837846e-05}
{"train_loss": 0.05627266317605972, "global_step": 196955, "epoch": 2212, "lr": 1.7519993826337556e-05}
{"train_loss": 0.06595005028033524, "global_step": 196956, "epoch": 2212, "lr": 1.751955306866319e-05, "val_loss": 8.159449577331543}
{"train_loss": 0.0412876270711422, "global_step": 196957, "epoch": 2213, "lr": 1.7519112315355407e-05}
{"train_loss": 0.10918296128511429, "global_step": 196958, "epoch": 2213, "lr": 1.7518671566414303e-05}
{"train_loss": 0.052574485540390015, "global_step": 196959, "epoch": 2213, "lr": 1.7518230821839897e-05}
{"train_loss": 0.08111122250556946, "global_step": 196960, "epoch": 2213, "lr": 1.751779008163229e-05}
{"train_loss": 0.06266642361879349, "global_step": 196961, "epoch": 2213, "lr": 1.7517349345791517e-05}
{"train_loss": 0.042976222932338715, "global_step": 196962, "epoch": 2213, "lr": 1.7516908614317625e-05}
{"train_loss": 0.12384676933288574, "global_step": 196963, "epoch": 2213, "lr": 1.7516467887210702e-05}
{"train_loss": 0.08245359361171722, "global_step": 196964, "epoch": 2213, "lr": 1.7516027164470783e-05}
{"train_loss": 0.14332501590251923, "global_step": 196965, "epoch": 2213, "lr": 1.7515586446097952e-05}
{"train_loss": 0.03639235720038414, "global_step": 196966, "epoch": 2213, "lr": 1.751514573209224e-05}
{"train_loss": 0.06384715437889099, "global_step": 196967, "epoch": 2213, "lr": 1.7514705022453736e-05}
{"train_loss": 0.021929549053311348, "global_step": 196968, "epoch": 2213, "lr": 1.7514264317182466e-05}
{"train_loss": 0.0294325090944767, "global_step": 196969, "epoch": 2213, "lr": 1.7513823616278524e-05}
{"train_loss": 0.038408003747463226, "global_step": 196970, "epoch": 2213, "lr": 1.751338291974194e-05}
{"train_loss": 0.0656038150191307, "global_step": 196971, "epoch": 2213, "lr": 1.7512942227572797e-05}
{"train_loss": 0.04485657438635826, "global_step": 196972, "epoch": 2213, "lr": 1.7512501539771127e-05}
{"train_loss": 0.0724606066942215, "global_step": 196973, "epoch": 2213, "lr": 1.751206085633701e-05}
{"train_loss": 0.08777835220098495, "global_step": 196974, "epoch": 2213, "lr": 1.751162017727051e-05}
{"train_loss": 0.052387215197086334, "global_step": 196975, "epoch": 2213, "lr": 1.7511179502571666e-05}
{"train_loss": 0.0738234743475914, "global_step": 196976, "epoch": 2213, "lr": 1.751073883224056e-05}
{"train_loss": 0.03330790624022484, "global_step": 196977, "epoch": 2213, "lr": 1.7510298166277228e-05}
{"train_loss": 0.05392848700284958, "global_step": 196978, "epoch": 2213, "lr": 1.7509857504681755e-05}
{"train_loss": 0.08122708648443222, "global_step": 196979, "epoch": 2213, "lr": 1.7509416847454162e-05}
{"train_loss": 0.06382671743631363, "global_step": 196980, "epoch": 2213, "lr": 1.7508976194594556e-05}
{"train_loss": 0.10710867494344711, "global_step": 196981, "epoch": 2213, "lr": 1.7508535546102957e-05}
{"train_loss": 0.05804320424795151, "global_step": 196982, "epoch": 2213, "lr": 1.750809490197945e-05}
{"train_loss": 0.06648116558790207, "global_step": 196983, "epoch": 2213, "lr": 1.7507654262224073e-05}
{"train_loss": 0.07648338377475739, "global_step": 196984, "epoch": 2213, "lr": 1.750721362683691e-05}
{"train_loss": 0.07702386379241943, "global_step": 196985, "epoch": 2213, "lr": 1.7506772995817988e-05}
{"train_loss": 0.05916742607951164, "global_step": 196986, "epoch": 2213, "lr": 1.75063323691674e-05}
{"train_loss": 0.08341427147388458, "global_step": 196987, "epoch": 2213, "lr": 1.7505891746885172e-05}
{"train_loss": 0.05082174390554428, "global_step": 196988, "epoch": 2213, "lr": 1.75054511289714e-05}
{"train_loss": 0.03360170125961304, "global_step": 196989, "epoch": 2213, "lr": 1.750501051542611e-05}
{"train_loss": 0.03369782492518425, "global_step": 196990, "epoch": 2213, "lr": 1.7504569906249373e-05}
{"train_loss": 0.08759929239749908, "global_step": 196991, "epoch": 2213, "lr": 1.750412930144127e-05}
{"train_loss": 0.059937167912721634, "global_step": 196992, "epoch": 2213, "lr": 1.7503688701001818e-05}
{"train_loss": 0.09033088386058807, "global_step": 196993, "epoch": 2213, "lr": 1.750324810493112e-05}
{"train_loss": 0.06485597789287567, "global_step": 196994, "epoch": 2213, "lr": 1.7502807513229213e-05}
{"train_loss": 0.0199603158980608, "global_step": 196995, "epoch": 2213, "lr": 1.750236692589614e-05}
{"train_loss": 0.054468631744384766, "global_step": 196996, "epoch": 2213, "lr": 1.7501926342931996e-05}
{"train_loss": 0.0725095197558403, "global_step": 196997, "epoch": 2213, "lr": 1.7501485764336806e-05}
{"train_loss": 0.03941454365849495, "global_step": 196998, "epoch": 2213, "lr": 1.750104519011066e-05}
{"train_loss": 0.059821076691150665, "global_step": 196999, "epoch": 2213, "lr": 1.7500604620253585e-05}
{"train_loss": 0.038649335503578186, "global_step": 197000, "epoch": 2213, "lr": 1.750016405476567e-05}
{"train_loss": 0.06781328469514847, "global_step": 197001, "epoch": 2213, "lr": 1.7499723493646974e-05}
{"train_loss": 0.04745011404156685, "global_step": 197002, "epoch": 2213, "lr": 1.749928293689752e-05}
{"train_loss": 0.07915797829627991, "global_step": 197003, "epoch": 2213, "lr": 1.7498842384517404e-05}
{"train_loss": 0.08237693458795547, "global_step": 197004, "epoch": 2213, "lr": 1.749840183650666e-05}
{"train_loss": 0.07075794786214828, "global_step": 197005, "epoch": 2213, "lr": 1.749796129286538e-05}
{"train_loss": 0.07242943346500397, "global_step": 197006, "epoch": 2213, "lr": 1.7497520753593583e-05}
{"train_loss": 0.07409506291151047, "global_step": 197007, "epoch": 2213, "lr": 1.7497080218691365e-05}
{"train_loss": 0.04639729857444763, "global_step": 197008, "epoch": 2213, "lr": 1.749663968815875e-05}
{"train_loss": 0.04311686009168625, "global_step": 197009, "epoch": 2213, "lr": 1.7496199161995818e-05}
{"train_loss": 0.09333977848291397, "global_step": 197010, "epoch": 2213, "lr": 1.749575864020264e-05}
{"train_loss": 0.08192015439271927, "global_step": 197011, "epoch": 2213, "lr": 1.7495318122779246e-05}
{"train_loss": 0.035616952925920486, "global_step": 197012, "epoch": 2213, "lr": 1.7494877609725728e-05}
{"train_loss": 0.06696759164333344, "global_step": 197013, "epoch": 2213, "lr": 1.749443710104211e-05}
{"train_loss": 0.03321298956871033, "global_step": 197014, "epoch": 2213, "lr": 1.7493996596728485e-05}
{"train_loss": 0.08764304220676422, "global_step": 197015, "epoch": 2213, "lr": 1.749355609678488e-05}
{"train_loss": 0.06496013700962067, "global_step": 197016, "epoch": 2213, "lr": 1.7493115601211383e-05}
{"train_loss": 0.069581039249897, "global_step": 197017, "epoch": 2213, "lr": 1.749267511000803e-05}
{"train_loss": 0.04097801074385643, "global_step": 197018, "epoch": 2213, "lr": 1.7492234623174904e-05}
{"train_loss": 0.04417489841580391, "global_step": 197019, "epoch": 2213, "lr": 1.7491794140712038e-05}
{"train_loss": 0.07439128309488297, "global_step": 197020, "epoch": 2213, "lr": 1.7491353662619514e-05}
{"train_loss": 0.09551072120666504, "global_step": 197021, "epoch": 2213, "lr": 1.7490913188897368e-05}
{"train_loss": 0.062063418328762054, "global_step": 197022, "epoch": 2213, "lr": 1.749047271954569e-05}
{"train_loss": 0.050789572298526764, "global_step": 197023, "epoch": 2213, "lr": 1.7490032254564505e-05}
{"train_loss": 0.10170634835958481, "global_step": 197024, "epoch": 2213, "lr": 1.7489591793953904e-05}
{"train_loss": 0.08261673897504807, "global_step": 197025, "epoch": 2213, "lr": 1.7489151337713918e-05}
{"train_loss": 0.08319943398237228, "global_step": 197026, "epoch": 2213, "lr": 1.748871088584461e-05}
{"train_loss": 0.05685991793870926, "global_step": 197027, "epoch": 2213, "lr": 1.7488270438346077e-05}
{"train_loss": 0.06611354649066925, "global_step": 197028, "epoch": 2213, "lr": 1.7487829995218324e-05}
{"train_loss": 0.0881873369216919, "global_step": 197029, "epoch": 2213, "lr": 1.7487389556461457e-05}
{"train_loss": 0.08889146149158478, "global_step": 197030, "epoch": 2213, "lr": 1.7486949122075497e-05}
{"train_loss": 0.04695398733019829, "global_step": 197031, "epoch": 2213, "lr": 1.748650869206054e-05}
{"train_loss": 0.07116327434778214, "global_step": 197032, "epoch": 2213, "lr": 1.7486068266416606e-05}
{"train_loss": 0.06903117895126343, "global_step": 197033, "epoch": 2213, "lr": 1.7485627845143787e-05}
{"train_loss": 0.05582764372229576, "global_step": 197034, "epoch": 2213, "lr": 1.7485187428242132e-05}
{"train_loss": 0.06964942812919617, "global_step": 197035, "epoch": 2213, "lr": 1.7484747015711683e-05}
{"train_loss": 0.12405377626419067, "global_step": 197036, "epoch": 2213, "lr": 1.748430660755253e-05}
{"train_loss": 0.04311439394950867, "global_step": 197037, "epoch": 2213, "lr": 1.74838662037647e-05}
{"train_loss": 0.08366649597883224, "global_step": 197038, "epoch": 2213, "lr": 1.748342580434828e-05}
{"train_loss": 0.05256786569952965, "global_step": 197039, "epoch": 2213, "lr": 1.7482985409303316e-05}
{"train_loss": 0.11091484129428864, "global_step": 197040, "epoch": 2213, "lr": 1.748254501862986e-05}
{"train_loss": 0.027236944064497948, "global_step": 197041, "epoch": 2213, "lr": 1.748210463232799e-05}
{"train_loss": 0.1037275567650795, "global_step": 197042, "epoch": 2213, "lr": 1.7481664250397732e-05}
{"train_loss": 0.10994598269462585, "global_step": 197043, "epoch": 2213, "lr": 1.7481223872839196e-05}
{"train_loss": 0.06277310103178024, "global_step": 197044, "epoch": 2213, "lr": 1.7480783499652393e-05}
{"train_loss": 0.06628483148773065, "global_step": 197045, "epoch": 2213, "lr": 1.7480343130837402e-05, "val_loss": 7.992330074310303}
{"train_loss": 0.05060778558254242, "global_step": 197046, "epoch": 2214, "lr": 1.7479902766394297e-05}
{"train_loss": 0.05904127284884453, "global_step": 197047, "epoch": 2214, "lr": 1.7479462406323104e-05}
{"train_loss": 0.044767919927835464, "global_step": 197048, "epoch": 2214, "lr": 1.7479022050623924e-05}
{"train_loss": 0.12710213661193848, "global_step": 197049, "epoch": 2214, "lr": 1.747858169929677e-05}
{"train_loss": 0.10783209651708603, "global_step": 197050, "epoch": 2214, "lr": 1.747814135234175e-05}
{"train_loss": 0.04621238633990288, "global_step": 197051, "epoch": 2214, "lr": 1.7477701009758873e-05}
{"train_loss": 0.038732461631298065, "global_step": 197052, "epoch": 2214, "lr": 1.747726067154824e-05}
{"train_loss": 0.10444831848144531, "global_step": 197053, "epoch": 2214, "lr": 1.7476820337709878e-05}
{"train_loss": 0.11788686364889145, "global_step": 197054, "epoch": 2214, "lr": 1.7476380008243876e-05}
{"train_loss": 0.07591928541660309, "global_step": 197055, "epoch": 2214, "lr": 1.7475939683150266e-05}
{"train_loss": 0.06775867938995361, "global_step": 197056, "epoch": 2214, "lr": 1.747549936242913e-05}
{"train_loss": 0.06421064585447311, "global_step": 197057, "epoch": 2214, "lr": 1.7475059046080506e-05}
{"train_loss": 0.071748748421669, "global_step": 197058, "epoch": 2214, "lr": 1.7474618734104475e-05}
{"train_loss": 0.04492276534438133, "global_step": 197059, "epoch": 2214, "lr": 1.7474178426501075e-05}
{"train_loss": 0.0926351472735405, "global_step": 197060, "epoch": 2214, "lr": 1.7473738123270388e-05}
{"train_loss": 0.040428224951028824, "global_step": 197061, "epoch": 2214, "lr": 1.7473297824412444e-05}
{"train_loss": 0.046001385897397995, "global_step": 197062, "epoch": 2214, "lr": 1.7472857529927335e-05}
{"train_loss": 0.0640917420387268, "global_step": 197063, "epoch": 2214, "lr": 1.7472417239815087e-05}
{"train_loss": 0.07063037157058716, "global_step": 197064, "epoch": 2214, "lr": 1.7471976954075776e-05}
{"train_loss": 0.10424305498600006, "global_step": 197065, "epoch": 2214, "lr": 1.7471536672709476e-05}
{"train_loss": 0.07627072930335999, "global_step": 197066, "epoch": 2214, "lr": 1.7471096395716217e-05}
{"train_loss": 0.048373207449913025, "global_step": 197067, "epoch": 2214, "lr": 1.747065612309609e-05}
{"train_loss": 0.08590801805257797, "global_step": 197068, "epoch": 2214, "lr": 1.747021585484912e-05}
{"train_loss": 0.04395123943686485, "global_step": 197069, "epoch": 2214, "lr": 1.74697755909754e-05}
{"train_loss": 0.11090104281902313, "global_step": 197070, "epoch": 2214, "lr": 1.7469335331474958e-05}
{"train_loss": 0.1263676881790161, "global_step": 197071, "epoch": 2214, "lr": 1.7468895076347878e-05}
{"train_loss": 0.06896759569644928, "global_step": 197072, "epoch": 2214, "lr": 1.746845482559421e-05}
{"train_loss": 0.06472264975309372, "global_step": 197073, "epoch": 2214, "lr": 1.7468014579214e-05}
{"train_loss": 0.027753999456763268, "global_step": 197074, "epoch": 2214, "lr": 1.746757433720733e-05}
{"train_loss": 0.0414213091135025, "global_step": 197075, "epoch": 2214, "lr": 1.7467134099574233e-05}
{"train_loss": 0.07266789674758911, "global_step": 197076, "epoch": 2214, "lr": 1.7466693866314803e-05}
{"train_loss": 0.03074122965335846, "global_step": 197077, "epoch": 2214, "lr": 1.7466253637429075e-05}
{"train_loss": 0.06476335972547531, "global_step": 197078, "epoch": 2214, "lr": 1.7465813412917097e-05}
{"train_loss": 0.054105885326862335, "global_step": 197079, "epoch": 2214, "lr": 1.7465373192778966e-05}
{"train_loss": 0.04295138642191887, "global_step": 197080, "epoch": 2214, "lr": 1.7464932977014697e-05}
{"train_loss": 0.05656401440501213, "global_step": 197081, "epoch": 2214, "lr": 1.7464492765624374e-05}
{"train_loss": 0.022586317732930183, "global_step": 197082, "epoch": 2214, "lr": 1.7464052558608067e-05}
{"train_loss": 0.03983902186155319, "global_step": 197083, "epoch": 2214, "lr": 1.7463612355965802e-05}
{"train_loss": 0.06666138023138046, "global_step": 197084, "epoch": 2214, "lr": 1.7463172157697683e-05}
{"train_loss": 0.056084319949150085, "global_step": 197085, "epoch": 2214, "lr": 1.746273196380372e-05}
{"train_loss": 0.10562257468700409, "global_step": 197086, "epoch": 2214, "lr": 1.7462291774284017e-05}
{"train_loss": 0.09862282872200012, "global_step": 197087, "epoch": 2214, "lr": 1.7461851589138593e-05}
{"train_loss": 0.09503346681594849, "global_step": 197088, "epoch": 2214, "lr": 1.7461411408367545e-05}
{"train_loss": 0.07565734535455704, "global_step": 197089, "epoch": 2214, "lr": 1.7460971231970898e-05}
{"train_loss": 0.08361680060625076, "global_step": 197090, "epoch": 2214, "lr": 1.746053105994874e-05}
{"train_loss": 0.04922517389059067, "global_step": 197091, "epoch": 2214, "lr": 1.74600908923011e-05}
{"train_loss": 0.038377780467271805, "global_step": 197092, "epoch": 2214, "lr": 1.7459650729028076e-05}
{"train_loss": 0.04577591270208359, "global_step": 197093, "epoch": 2214, "lr": 1.7459210570129685e-05}
{"train_loss": 0.05081099271774292, "global_step": 197094, "epoch": 2214, "lr": 1.7458770415606024e-05}
{"train_loss": 0.07817091047763824, "global_step": 197095, "epoch": 2214, "lr": 1.7458330265457118e-05}
{"train_loss": 0.025723645463585854, "global_step": 197096, "epoch": 2214, "lr": 1.7457890119683062e-05}
{"train_loss": 0.058042362332344055, "global_step": 197097, "epoch": 2214, "lr": 1.7457449978283874e-05}
{"train_loss": 0.0784400999546051, "global_step": 197098, "epoch": 2214, "lr": 1.745700984125966e-05}
{"train_loss": 0.03388997167348862, "global_step": 197099, "epoch": 2214, "lr": 1.7456569708610427e-05}
{"train_loss": 0.06458988040685654, "global_step": 197100, "epoch": 2214, "lr": 1.745612958033627e-05}
{"train_loss": 0.0972089096903801, "global_step": 197101, "epoch": 2214, "lr": 1.7455689456437252e-05}
{"train_loss": 0.033110830932855606, "global_step": 197102, "epoch": 2214, "lr": 1.7455249336913405e-05}
{"train_loss": 0.055753711611032486, "global_step": 197103, "epoch": 2214, "lr": 1.745480922176482e-05}
{"train_loss": 0.10762879252433777, "global_step": 197104, "epoch": 2214, "lr": 1.745436911099152e-05}
{"train_loss": 0.04955863952636719, "global_step": 197105, "epoch": 2214, "lr": 1.74539290045936e-05}
{"train_loss": 0.10565125197172165, "global_step": 197106, "epoch": 2214, "lr": 1.745348890257109e-05}
{"train_loss": 0.08636455237865448, "global_step": 197107, "epoch": 2214, "lr": 1.7453048804924078e-05}
{"train_loss": 0.06814788281917572, "global_step": 197108, "epoch": 2214, "lr": 1.7452608711652584e-05}
{"train_loss": 0.04247028008103371, "global_step": 197109, "epoch": 2214, "lr": 1.7452168622756713e-05}
{"train_loss": 0.11332403123378754, "global_step": 197110, "epoch": 2214, "lr": 1.74517285382365e-05}
{"train_loss": 0.10771124064922333, "global_step": 197111, "epoch": 2214, "lr": 1.745128845809199e-05}
{"train_loss": 0.08423082530498505, "global_step": 197112, "epoch": 2214, "lr": 1.745084838232327e-05}
{"train_loss": 0.06432205438613892, "global_step": 197113, "epoch": 2214, "lr": 1.7450408310930372e-05}
{"train_loss": 0.07579772919416428, "global_step": 197114, "epoch": 2214, "lr": 1.744996824391339e-05}
{"train_loss": 0.06287418305873871, "global_step": 197115, "epoch": 2214, "lr": 1.744952818127234e-05}
{"train_loss": 0.14499010145664215, "global_step": 197116, "epoch": 2214, "lr": 1.7449088123007323e-05}
{"train_loss": 0.0559736043214798, "global_step": 197117, "epoch": 2214, "lr": 1.744864806911837e-05}
{"train_loss": 0.10232634842395782, "global_step": 197118, "epoch": 2214, "lr": 1.7448208019605555e-05}
{"train_loss": 0.05124324560165405, "global_step": 197119, "epoch": 2214, "lr": 1.744776797446892e-05}
{"train_loss": 0.0671989843249321, "global_step": 197120, "epoch": 2214, "lr": 1.7447327933708557e-05}
{"train_loss": 0.055149320513010025, "global_step": 197121, "epoch": 2214, "lr": 1.7446887897324483e-05}
{"train_loss": 0.05670492723584175, "global_step": 197122, "epoch": 2214, "lr": 1.7446447865316795e-05}
{"train_loss": 0.1098078042268753, "global_step": 197123, "epoch": 2214, "lr": 1.7446007837685513e-05}
{"train_loss": 0.0363362692296505, "global_step": 197124, "epoch": 2214, "lr": 1.744556781443074e-05}
{"train_loss": 0.0756465494632721, "global_step": 197125, "epoch": 2214, "lr": 1.74451277955525e-05}
{"train_loss": 0.08656711876392365, "global_step": 197126, "epoch": 2214, "lr": 1.7444687781050883e-05}
{"train_loss": 0.09819580614566803, "global_step": 197127, "epoch": 2214, "lr": 1.744424777092591e-05}
{"train_loss": 0.16595257818698883, "global_step": 197128, "epoch": 2214, "lr": 1.744380776517768e-05}
{"train_loss": 0.04949740692973137, "global_step": 197129, "epoch": 2214, "lr": 1.7443367763806213e-05}
{"train_loss": 0.052809689193964005, "global_step": 197130, "epoch": 2214, "lr": 1.7442927766811608e-05}
{"train_loss": 0.0628783330321312, "global_step": 197131, "epoch": 2214, "lr": 1.744248777419389e-05}
{"train_loss": 0.03571021184325218, "global_step": 197132, "epoch": 2214, "lr": 1.7442047785953143e-05}
{"train_loss": 0.033820606768131256, "global_step": 197133, "epoch": 2214, "lr": 1.7441607802089404e-05}
{"train_loss": 0.06957066023533934, "global_step": 197134, "epoch": 2214, "lr": 1.744116782260276e-05, "val_loss": 8.117204666137695}
{"train_loss": 0.04228901490569115, "global_step": 197135, "epoch": 2215, "lr": 1.7440727847493236e-05}
{"train_loss": 0.06693098694086075, "global_step": 197136, "epoch": 2215, "lr": 1.7440287876760914e-05}
{"train_loss": 0.13726742565631866, "global_step": 197137, "epoch": 2215, "lr": 1.743984791040586e-05}
{"train_loss": 0.06022823601961136, "global_step": 197138, "epoch": 2215, "lr": 1.7439407948428104e-05}
{"train_loss": 0.03788698837161064, "global_step": 197139, "epoch": 2215, "lr": 1.7438967990827744e-05}
{"train_loss": 0.08731204271316528, "global_step": 197140, "epoch": 2215, "lr": 1.7438528037604803e-05}
{"train_loss": 0.07413596659898758, "global_step": 197141, "epoch": 2215, "lr": 1.7438088088759365e-05}
{"train_loss": 0.04577815160155296, "global_step": 197142, "epoch": 2215, "lr": 1.743764814429146e-05}
{"train_loss": 0.06993772089481354, "global_step": 197143, "epoch": 2215, "lr": 1.7437208204201193e-05}
{"train_loss": 0.04629846662282944, "global_step": 197144, "epoch": 2215, "lr": 1.7436768268488575e-05}
{"train_loss": 0.05045955628156662, "global_step": 197145, "epoch": 2215, "lr": 1.74363283371537e-05}
{"train_loss": 0.10813461989164352, "global_step": 197146, "epoch": 2215, "lr": 1.7435888410196603e-05}
{"train_loss": 0.074777752161026, "global_step": 197147, "epoch": 2215, "lr": 1.7435448487617366e-05}
{"train_loss": 0.023634832352399826, "global_step": 197148, "epoch": 2215, "lr": 1.7435008569416022e-05}
{"train_loss": 0.07443525642156601, "global_step": 197149, "epoch": 2215, "lr": 1.7434568655592666e-05}
{"train_loss": 0.061392124742269516, "global_step": 197150, "epoch": 2215, "lr": 1.7434128746147328e-05}
{"train_loss": 0.0868782252073288, "global_step": 197151, "epoch": 2215, "lr": 1.743368884108006e-05}
{"train_loss": 0.03986550495028496, "global_step": 197152, "epoch": 2215, "lr": 1.743324894039095e-05}
{"train_loss": 0.02509317174553871, "global_step": 197153, "epoch": 2215, "lr": 1.743280904408003e-05}
{"train_loss": 0.05785127356648445, "global_step": 197154, "epoch": 2215, "lr": 1.7432369152147383e-05}
{"train_loss": 0.1713201403617859, "global_step": 197155, "epoch": 2215, "lr": 1.7431929264593043e-05}
{"train_loss": 0.09891786426305771, "global_step": 197156, "epoch": 2215, "lr": 1.7431489381417104e-05}
{"train_loss": 0.0919337049126625, "global_step": 197157, "epoch": 2215, "lr": 1.7431049502619583e-05}
{"train_loss": 0.1040545105934143, "global_step": 197158, "epoch": 2215, "lr": 1.743060962820058e-05}
{"train_loss": 0.04941035807132721, "global_step": 197159, "epoch": 2215, "lr": 1.7430169758160116e-05}
{"train_loss": 0.07366057485342026, "global_step": 197160, "epoch": 2215, "lr": 1.742972989249828e-05}
{"train_loss": 0.12611980736255646, "global_step": 197161, "epoch": 2215, "lr": 1.742929003121511e-05}
{"train_loss": 0.05566480755805969, "global_step": 197162, "epoch": 2215, "lr": 1.742885017431069e-05}
{"train_loss": 0.08716440200805664, "global_step": 197163, "epoch": 2215, "lr": 1.742841032178505e-05}
{"train_loss": 0.0662514939904213, "global_step": 197164, "epoch": 2215, "lr": 1.7427970473638277e-05}
{"train_loss": 0.07136854529380798, "global_step": 197165, "epoch": 2215, "lr": 1.7427530629870404e-05}
{"train_loss": 0.054188210517168045, "global_step": 197166, "epoch": 2215, "lr": 1.7427090790481515e-05}
{"train_loss": 0.05198872089385986, "global_step": 197167, "epoch": 2215, "lr": 1.742665095547164e-05}
{"train_loss": 0.05718335136771202, "global_step": 197168, "epoch": 2215, "lr": 1.742621112484087e-05}
{"train_loss": 0.09496819227933884, "global_step": 197169, "epoch": 2215, "lr": 1.7425771298589233e-05}
{"train_loss": 0.04249466955661774, "global_step": 197170, "epoch": 2215, "lr": 1.7425331476716823e-05}
{"train_loss": 0.0651417002081871, "global_step": 197171, "epoch": 2215, "lr": 1.7424891659223664e-05}
{"train_loss": 0.03561292588710785, "global_step": 197172, "epoch": 2215, "lr": 1.742445184610983e-05}
{"train_loss": 0.07300876826047897, "global_step": 197173, "epoch": 2215, "lr": 1.7424012037375397e-05}
{"train_loss": 0.06780430674552917, "global_step": 197174, "epoch": 2215, "lr": 1.742357223302039e-05}
{"train_loss": 0.09283996373414993, "global_step": 197175, "epoch": 2215, "lr": 1.742313243304491e-05}
{"train_loss": 0.05338457226753235, "global_step": 197176, "epoch": 2215, "lr": 1.7422692637448973e-05}
{"train_loss": 0.032647959887981415, "global_step": 197177, "epoch": 2215, "lr": 1.7422252846232673e-05}
{"train_loss": 0.06632418930530548, "global_step": 197178, "epoch": 2215, "lr": 1.7421813059396036e-05}
{"train_loss": 0.09652340412139893, "global_step": 197179, "epoch": 2215, "lr": 1.7421373276939155e-05}
{"train_loss": 0.14588011801242828, "global_step": 197180, "epoch": 2215, "lr": 1.742093349886206e-05}
{"train_loss": 0.0273349080234766, "global_step": 197181, "epoch": 2215, "lr": 1.7420493725164843e-05}
{"train_loss": 0.07182133942842484, "global_step": 197182, "epoch": 2215, "lr": 1.742005395584752e-05}
{"train_loss": 0.048392221331596375, "global_step": 197183, "epoch": 2215, "lr": 1.7419614190910195e-05}
{"train_loss": 0.06519590318202972, "global_step": 197184, "epoch": 2215, "lr": 1.741917443035289e-05}
{"train_loss": 0.06029738858342171, "global_step": 197185, "epoch": 2215, "lr": 1.7418734674175697e-05}
{"train_loss": 0.038064826279878616, "global_step": 197186, "epoch": 2215, "lr": 1.7418294922378638e-05}
{"train_loss": 0.07056792080402374, "global_step": 197187, "epoch": 2215, "lr": 1.7417855174961806e-05}
{"train_loss": 0.06883346289396286, "global_step": 197188, "epoch": 2215, "lr": 1.741741543192525e-05}
{"train_loss": 0.05261824652552605, "global_step": 197189, "epoch": 2215, "lr": 1.741697569326901e-05}
{"train_loss": 0.06848388910293579, "global_step": 197190, "epoch": 2215, "lr": 1.7416535958993173e-05}
{"train_loss": 0.04412390664219856, "global_step": 197191, "epoch": 2215, "lr": 1.7416096229097774e-05}
{"train_loss": 0.06251851469278336, "global_step": 197192, "epoch": 2215, "lr": 1.74156565035829e-05}
{"train_loss": 0.037176281213760376, "global_step": 197193, "epoch": 2215, "lr": 1.7415216782448575e-05}
{"train_loss": 0.04765310883522034, "global_step": 197194, "epoch": 2215, "lr": 1.741477706569489e-05}
{"train_loss": 0.03366706520318985, "global_step": 197195, "epoch": 2215, "lr": 1.741433735332188e-05}
{"train_loss": 0.058263253420591354, "global_step": 197196, "epoch": 2215, "lr": 1.741389764532963e-05}
{"train_loss": 0.08236352354288101, "global_step": 197197, "epoch": 2215, "lr": 1.741345794171817e-05}
{"train_loss": 0.07421712577342987, "global_step": 197198, "epoch": 2215, "lr": 1.7413018242487583e-05}
{"train_loss": 0.02021186240017414, "global_step": 197199, "epoch": 2215, "lr": 1.741257854763791e-05}
{"train_loss": 0.05133318528532982, "global_step": 197200, "epoch": 2215, "lr": 1.741213885716923e-05}
{"train_loss": 0.0730062797665596, "global_step": 197201, "epoch": 2215, "lr": 1.7411699171081575e-05}
{"train_loss": 0.03912757337093353, "global_step": 197202, "epoch": 2215, "lr": 1.7411259489375035e-05}
{"train_loss": 0.048459481447935104, "global_step": 197203, "epoch": 2215, "lr": 1.7410819812049644e-05}
{"train_loss": 0.05259158834815025, "global_step": 197204, "epoch": 2215, "lr": 1.741038013910548e-05}
{"train_loss": 0.0913112461566925, "global_step": 197205, "epoch": 2215, "lr": 1.7409940470542584e-05}
{"train_loss": 0.07960070669651031, "global_step": 197206, "epoch": 2215, "lr": 1.7409500806361034e-05}
{"train_loss": 0.11165159195661545, "global_step": 197207, "epoch": 2215, "lr": 1.7409061146560866e-05}
{"train_loss": 0.04395340010523796, "global_step": 197208, "epoch": 2215, "lr": 1.7408621491142153e-05}
{"train_loss": 0.04657692462205887, "global_step": 197209, "epoch": 2215, "lr": 1.7408181840104972e-05}
{"train_loss": 0.07376261055469513, "global_step": 197210, "epoch": 2215, "lr": 1.7407742193449346e-05}
{"train_loss": 0.08085402101278305, "global_step": 197211, "epoch": 2215, "lr": 1.7407302551175364e-05}
{"train_loss": 0.05455240607261658, "global_step": 197212, "epoch": 2215, "lr": 1.740686291328306e-05}
{"train_loss": 0.03291533514857292, "global_step": 197213, "epoch": 2215, "lr": 1.7406423279772522e-05}
{"train_loss": 0.09210244566202164, "global_step": 197214, "epoch": 2215, "lr": 1.7405983650643777e-05}
{"train_loss": 0.04026990383863449, "global_step": 197215, "epoch": 2215, "lr": 1.7405544025896913e-05}
{"train_loss": 0.03566846251487732, "global_step": 197216, "epoch": 2215, "lr": 1.7405104405531963e-05}
{"train_loss": 0.06201512739062309, "global_step": 197217, "epoch": 2215, "lr": 1.740466478954902e-05}
{"train_loss": 0.1266019344329834, "global_step": 197218, "epoch": 2215, "lr": 1.7404225177948104e-05}
{"train_loss": 0.08205587416887283, "global_step": 197219, "epoch": 2215, "lr": 1.7403785570729302e-05}
{"train_loss": 0.09360969811677933, "global_step": 197220, "epoch": 2215, "lr": 1.7403345967892653e-05}
{"train_loss": 0.0809064656496048, "global_step": 197221, "epoch": 2215, "lr": 1.7402906369438243e-05}
{"train_loss": 0.10315734893083572, "global_step": 197222, "epoch": 2215, "lr": 1.74024667753661e-05}
{"train_loss": 0.06694288024406754, "global_step": 197223, "epoch": 2215, "lr": 1.740202718567631e-05, "val_loss": 7.897563457489014, "train_action_mse_error": 16.11252212524414}
{"train_loss": 0.09574222564697266, "global_step": 197224, "epoch": 2216, "lr": 1.7401587600368907e-05}
{"train_loss": 0.15670396387577057, "global_step": 197225, "epoch": 2216, "lr": 1.740114801944398e-05}
{"train_loss": 0.03137645870447159, "global_step": 197226, "epoch": 2216, "lr": 1.740070844290157e-05}
{"train_loss": 0.05336076766252518, "global_step": 197227, "epoch": 2216, "lr": 1.740026887074172e-05}
{"train_loss": 0.12488183379173279, "global_step": 197228, "epoch": 2216, "lr": 1.7399829302964527e-05}
{"train_loss": 0.044035326689481735, "global_step": 197229, "epoch": 2216, "lr": 1.739938973957001e-05}
{"train_loss": 0.06275948137044907, "global_step": 197230, "epoch": 2216, "lr": 1.7398950180558267e-05}
{"train_loss": 0.08734924346208572, "global_step": 197231, "epoch": 2216, "lr": 1.7398510625929315e-05}
{"train_loss": 0.04216624051332474, "global_step": 197232, "epoch": 2216, "lr": 1.7398071075683257e-05}
{"train_loss": 0.061304543167352676, "global_step": 197233, "epoch": 2216, "lr": 1.7397631529820114e-05}
{"train_loss": 0.0385613888502121, "global_step": 197234, "epoch": 2216, "lr": 1.739719198833998e-05}
{"train_loss": 0.09931638091802597, "global_step": 197235, "epoch": 2216, "lr": 1.7396752451242875e-05}
{"train_loss": 0.07193950563669205, "global_step": 197236, "epoch": 2216, "lr": 1.7396312918528902e-05}
{"train_loss": 0.05582872033119202, "global_step": 197237, "epoch": 2216, "lr": 1.7395873390198077e-05}
{"train_loss": 0.0866035670042038, "global_step": 197238, "epoch": 2216, "lr": 1.7395433866250495e-05}
{"train_loss": 0.09735224395990372, "global_step": 197239, "epoch": 2216, "lr": 1.7394994346686183e-05}
{"train_loss": 0.02654292620718479, "global_step": 197240, "epoch": 2216, "lr": 1.739455483150524e-05}
{"train_loss": 0.04848899319767952, "global_step": 197241, "epoch": 2216, "lr": 1.7394115320707678e-05}
{"train_loss": 0.07657583057880402, "global_step": 197242, "epoch": 2216, "lr": 1.7393675814293597e-05}
{"train_loss": 0.08730128407478333, "global_step": 197243, "epoch": 2216, "lr": 1.7393236312263023e-05}
{"train_loss": 0.1063462346792221, "global_step": 197244, "epoch": 2216, "lr": 1.739279681461603e-05}
{"train_loss": 0.03019736148416996, "global_step": 197245, "epoch": 2216, "lr": 1.7392357321352698e-05}
{"train_loss": 0.08540160208940506, "global_step": 197246, "epoch": 2216, "lr": 1.739191783247305e-05}
{"train_loss": 0.032874517142772675, "global_step": 197247, "epoch": 2216, "lr": 1.7391478347977175e-05}
{"train_loss": 0.04795392230153084, "global_step": 197248, "epoch": 2216, "lr": 1.73910388678651e-05}
{"train_loss": 0.08068446815013885, "global_step": 197249, "epoch": 2216, "lr": 1.7390599392136918e-05}
{"train_loss": 0.07815859466791153, "global_step": 197250, "epoch": 2216, "lr": 1.7390159920792664e-05}
{"train_loss": 0.03897334635257721, "global_step": 197251, "epoch": 2216, "lr": 1.7389720453832415e-05}
{"train_loss": 0.03973441198468208, "global_step": 197252, "epoch": 2216, "lr": 1.738928099125621e-05}
{"train_loss": 0.11384248733520508, "global_step": 197253, "epoch": 2216, "lr": 1.7388841533064126e-05}
{"train_loss": 0.06888072192668915, "global_step": 197254, "epoch": 2216, "lr": 1.7388402079256205e-05}
{"train_loss": 0.0759992003440857, "global_step": 197255, "epoch": 2216, "lr": 1.7387962629832534e-05}
{"train_loss": 0.06050588935613632, "global_step": 197256, "epoch": 2216, "lr": 1.738752318479313e-05}
{"train_loss": 0.025434279814362526, "global_step": 197257, "epoch": 2216, "lr": 1.73870837441381e-05}
{"train_loss": 0.05490386486053467, "global_step": 197258, "epoch": 2216, "lr": 1.738664430786746e-05}
{"train_loss": 0.059164565056562424, "global_step": 197259, "epoch": 2216, "lr": 1.7386204875981305e-05}
{"train_loss": 0.06952322274446487, "global_step": 197260, "epoch": 2216, "lr": 1.7385765448479656e-05}
{"train_loss": 0.07153970748186111, "global_step": 197261, "epoch": 2216, "lr": 1.7385326025362615e-05}
{"train_loss": 0.056439172476530075, "global_step": 197262, "epoch": 2216, "lr": 1.7384886606630203e-05}
{"train_loss": 0.06743080168962479, "global_step": 197263, "epoch": 2216, "lr": 1.738444719228251e-05}
{"train_loss": 0.07545802742242813, "global_step": 197264, "epoch": 2216, "lr": 1.7384007782319566e-05}
{"train_loss": 0.09098983556032181, "global_step": 197265, "epoch": 2216, "lr": 1.738356837674146e-05}
{"train_loss": 0.05831016227602959, "global_step": 197266, "epoch": 2216, "lr": 1.7383128975548235e-05}
{"train_loss": 0.03872663155198097, "global_step": 197267, "epoch": 2216, "lr": 1.7382689578739936e-05}
{"train_loss": 0.06717479974031448, "global_step": 197268, "epoch": 2216, "lr": 1.738225018631665e-05}
{"train_loss": 0.05284876376390457, "global_step": 197269, "epoch": 2216, "lr": 1.738181079827841e-05}
{"train_loss": 0.061912648379802704, "global_step": 197270, "epoch": 2216, "lr": 1.7381371414625304e-05}
{"train_loss": 0.08250303566455841, "global_step": 197271, "epoch": 2216, "lr": 1.7380932035357357e-05}
{"train_loss": 0.09039181470870972, "global_step": 197272, "epoch": 2216, "lr": 1.738049266047466e-05}
{"train_loss": 0.05351562052965164, "global_step": 197273, "epoch": 2216, "lr": 1.7380053289977248e-05}
{"train_loss": 0.0707755759358406, "global_step": 197274, "epoch": 2216, "lr": 1.73796139238652e-05}
{"train_loss": 0.04492131620645523, "global_step": 197275, "epoch": 2216, "lr": 1.7379174562138555e-05}
{"train_loss": 0.05053142085671425, "global_step": 197276, "epoch": 2216, "lr": 1.7378735204797397e-05}
{"train_loss": 0.042793288826942444, "global_step": 197277, "epoch": 2216, "lr": 1.7378295851841752e-05}
{"train_loss": 0.01279633678495884, "global_step": 197278, "epoch": 2216, "lr": 1.7377856503271715e-05}
{"train_loss": 0.0994599461555481, "global_step": 197279, "epoch": 2216, "lr": 1.7377417159087313e-05}
{"train_loss": 0.10252470523118973, "global_step": 197280, "epoch": 2216, "lr": 1.737697781928862e-05}
{"train_loss": 0.0683593899011612, "global_step": 197281, "epoch": 2216, "lr": 1.737653848387571e-05}
{"train_loss": 0.10796569287776947, "global_step": 197282, "epoch": 2216, "lr": 1.7376099152848608e-05}
{"train_loss": 0.07159458845853806, "global_step": 197283, "epoch": 2216, "lr": 1.7375659826207407e-05}
{"train_loss": 0.043248500674963, "global_step": 197284, "epoch": 2216, "lr": 1.737522050395214e-05}
{"train_loss": 0.06228751316666603, "global_step": 197285, "epoch": 2216, "lr": 1.737478118608289e-05}
{"train_loss": 0.04521055519580841, "global_step": 197286, "epoch": 2216, "lr": 1.7374341872599688e-05}
{"train_loss": 0.09359639137983322, "global_step": 197287, "epoch": 2216, "lr": 1.7373902563502626e-05}
{"train_loss": 0.10356064885854721, "global_step": 197288, "epoch": 2216, "lr": 1.737346325879172e-05}
{"train_loss": 0.04244327172636986, "global_step": 197289, "epoch": 2216, "lr": 1.737302395846708e-05}
{"train_loss": 0.04712540656328201, "global_step": 197290, "epoch": 2216, "lr": 1.7372584662528724e-05}
{"train_loss": 0.0950387641787529, "global_step": 197291, "epoch": 2216, "lr": 1.7372145370976733e-05}
{"train_loss": 0.035985615104436874, "global_step": 197292, "epoch": 2216, "lr": 1.737170608381115e-05}
{"train_loss": 0.06124356389045715, "global_step": 197293, "epoch": 2216, "lr": 1.737126680103206e-05}
{"train_loss": 0.034078679978847504, "global_step": 197294, "epoch": 2216, "lr": 1.7370827522639488e-05}
{"train_loss": 0.08048457652330399, "global_step": 197295, "epoch": 2216, "lr": 1.7370388248633528e-05}
{"train_loss": 0.040998704731464386, "global_step": 197296, "epoch": 2216, "lr": 1.73699489790142e-05}
{"train_loss": 0.04696735739707947, "global_step": 197297, "epoch": 2216, "lr": 1.7369509713781605e-05}
{"train_loss": 0.0962994173169136, "global_step": 197298, "epoch": 2216, "lr": 1.7369070452935765e-05}
{"train_loss": 0.06088600680232048, "global_step": 197299, "epoch": 2216, "lr": 1.7368631196476776e-05}
{"train_loss": 0.05170277878642082, "global_step": 197300, "epoch": 2216, "lr": 1.7368191944404656e-05}
{"train_loss": 0.03875173255801201, "global_step": 197301, "epoch": 2216, "lr": 1.7367752696719502e-05}
{"train_loss": 0.0930456593632698, "global_step": 197302, "epoch": 2216, "lr": 1.7367313453421342e-05}
{"train_loss": 0.09660565853118896, "global_step": 197303, "epoch": 2216, "lr": 1.736687421451026e-05}
{"train_loss": 0.036592140793800354, "global_step": 197304, "epoch": 2216, "lr": 1.7366434979986308e-05}
{"train_loss": 0.07559528946876526, "global_step": 197305, "epoch": 2216, "lr": 1.7365995749849522e-05}
{"train_loss": 0.08077575266361237, "global_step": 197306, "epoch": 2216, "lr": 1.7365556524099998e-05}
{"train_loss": 0.08416861295700073, "global_step": 197307, "epoch": 2216, "lr": 1.7365117302737762e-05}
{"train_loss": 0.06765218079090118, "global_step": 197308, "epoch": 2216, "lr": 1.7364678085762902e-05}
{"train_loss": 0.05719141662120819, "global_step": 197309, "epoch": 2216, "lr": 1.7364238873175446e-05}
{"train_loss": 0.03753805160522461, "global_step": 197310, "epoch": 2216, "lr": 1.7363799664975496e-05}
{"train_loss": 0.060826510190963745, "global_step": 197311, "epoch": 2216, "lr": 1.7363360461163057e-05}
{"train_loss": 0.06584935392556565, "global_step": 197312, "epoch": 2216, "lr": 1.736292126173824e-05, "val_loss": 8.169818878173828}
{"train_loss": 0.036003123968839645, "global_step": 197313, "epoch": 2217, "lr": 1.7362482066701062e-05}
{"train_loss": 0.03898850455880165, "global_step": 197314, "epoch": 2217, "lr": 1.7362042876051614e-05}
{"train_loss": 0.03250749409198761, "global_step": 197315, "epoch": 2217, "lr": 1.736160368978993e-05}
{"train_loss": 0.04155964031815529, "global_step": 197316, "epoch": 2217, "lr": 1.7361164507916094e-05}
{"train_loss": 0.03013390675187111, "global_step": 197317, "epoch": 2217, "lr": 1.7360725330430138e-05}
{"train_loss": 0.037227801978588104, "global_step": 197318, "epoch": 2217, "lr": 1.7360286157332127e-05}
{"train_loss": 0.05036788061261177, "global_step": 197319, "epoch": 2217, "lr": 1.7359846988622153e-05}
{"train_loss": 0.06838540732860565, "global_step": 197320, "epoch": 2217, "lr": 1.7359407824300234e-05}
{"train_loss": 0.04727525636553764, "global_step": 197321, "epoch": 2217, "lr": 1.7358968664366454e-05}
{"train_loss": 0.0741553008556366, "global_step": 197322, "epoch": 2217, "lr": 1.735852950882085e-05}
{"train_loss": 0.030396562069654465, "global_step": 197323, "epoch": 2217, "lr": 1.735809035766351e-05}
{"train_loss": 0.055288780480623245, "global_step": 197324, "epoch": 2217, "lr": 1.735765121089446e-05}
{"train_loss": 0.05915180593729019, "global_step": 197325, "epoch": 2217, "lr": 1.7357212068513795e-05}
{"train_loss": 0.0445932112634182, "global_step": 197326, "epoch": 2217, "lr": 1.7356772930521535e-05}
{"train_loss": 0.055924348533153534, "global_step": 197327, "epoch": 2217, "lr": 1.7356333796917778e-05}
{"train_loss": 0.10158412158489227, "global_step": 197328, "epoch": 2217, "lr": 1.7355894667702545e-05}
{"train_loss": 0.08656005561351776, "global_step": 197329, "epoch": 2217, "lr": 1.7355455542875932e-05}
{"train_loss": 0.06947005540132523, "global_step": 197330, "epoch": 2217, "lr": 1.7355016422437963e-05}
{"train_loss": 0.056144166737794876, "global_step": 197331, "epoch": 2217, "lr": 1.735457730638873e-05}
{"train_loss": 0.06713321059942245, "global_step": 197332, "epoch": 2217, "lr": 1.7354138194728264e-05}
{"train_loss": 0.06387259066104889, "global_step": 197333, "epoch": 2217, "lr": 1.735369908745665e-05}
{"train_loss": 0.03780899941921234, "global_step": 197334, "epoch": 2217, "lr": 1.7353259984573917e-05}
{"train_loss": 0.11645817756652832, "global_step": 197335, "epoch": 2217, "lr": 1.735282088608014e-05}
{"train_loss": 0.02980293333530426, "global_step": 197336, "epoch": 2217, "lr": 1.735238179197539e-05}
{"train_loss": 0.12887369096279144, "global_step": 197337, "epoch": 2217, "lr": 1.735194270225972e-05}
{"train_loss": 0.08329454064369202, "global_step": 197338, "epoch": 2217, "lr": 1.7351503616933167e-05}
{"train_loss": 0.027531912550330162, "global_step": 197339, "epoch": 2217, "lr": 1.7351064535995827e-05}
{"train_loss": 0.06438294798135757, "global_step": 197340, "epoch": 2217, "lr": 1.735062545944771e-05}
{"train_loss": 0.04061216115951538, "global_step": 197341, "epoch": 2217, "lr": 1.735018638728893e-05}
{"train_loss": 0.03173178434371948, "global_step": 197342, "epoch": 2217, "lr": 1.73497473195195e-05}
{"train_loss": 0.05020955204963684, "global_step": 197343, "epoch": 2217, "lr": 1.7349308256139513e-05}
{"train_loss": 0.059349846094846725, "global_step": 197344, "epoch": 2217, "lr": 1.7348869197149013e-05}
{"train_loss": 0.06773728877305984, "global_step": 197345, "epoch": 2217, "lr": 1.734843014254805e-05}
{"train_loss": 0.02633853629231453, "global_step": 197346, "epoch": 2217, "lr": 1.73479910923367e-05}
{"train_loss": 0.06911398470401764, "global_step": 197347, "epoch": 2217, "lr": 1.734755204651501e-05}
{"train_loss": 0.0531756617128849, "global_step": 197348, "epoch": 2217, "lr": 1.734711300508305e-05}
{"train_loss": 0.03351006284356117, "global_step": 197349, "epoch": 2217, "lr": 1.7346673968040854e-05}
{"train_loss": 0.08473744243383408, "global_step": 197350, "epoch": 2217, "lr": 1.734623493538853e-05}
{"train_loss": 0.06646610051393509, "global_step": 197351, "epoch": 2217, "lr": 1.7345795907126078e-05}
{"train_loss": 0.13524308800697327, "global_step": 197352, "epoch": 2217, "lr": 1.7345356883253605e-05}
{"train_loss": 0.044617075473070145, "global_step": 197353, "epoch": 2217, "lr": 1.7344917863771132e-05}
{"train_loss": 0.06077209487557411, "global_step": 197354, "epoch": 2217, "lr": 1.7344478848678742e-05}
{"train_loss": 0.028903279453516006, "global_step": 197355, "epoch": 2217, "lr": 1.73440398379765e-05}
{"train_loss": 0.028358621522784233, "global_step": 197356, "epoch": 2217, "lr": 1.734360083166444e-05}
{"train_loss": 0.03516101464629173, "global_step": 197357, "epoch": 2217, "lr": 1.7343161829742654e-05}
{"train_loss": 0.07154566794633865, "global_step": 197358, "epoch": 2217, "lr": 1.734272283221116e-05}
{"train_loss": 0.04639852046966553, "global_step": 197359, "epoch": 2217, "lr": 1.734228383907006e-05}
{"train_loss": 0.1092727854847908, "global_step": 197360, "epoch": 2217, "lr": 1.734184485031937e-05}
{"train_loss": 0.058205075562000275, "global_step": 197361, "epoch": 2217, "lr": 1.7341405865959197e-05}
{"train_loss": 0.050637587904930115, "global_step": 197362, "epoch": 2217, "lr": 1.734096688598955e-05}
{"train_loss": 0.05345672741532326, "global_step": 197363, "epoch": 2217, "lr": 1.7340527910410532e-05}
{"train_loss": 0.1360635906457901, "global_step": 197364, "epoch": 2217, "lr": 1.734008893922216e-05}
{"train_loss": 0.06592003256082535, "global_step": 197365, "epoch": 2217, "lr": 1.7339649972424537e-05}
{"train_loss": 0.08114098757505417, "global_step": 197366, "epoch": 2217, "lr": 1.733921101001768e-05}
{"train_loss": 0.020654942840337753, "global_step": 197367, "epoch": 2217, "lr": 1.7338772052001688e-05}
{"train_loss": 0.02684670314192772, "global_step": 197368, "epoch": 2217, "lr": 1.733833309837658e-05}
{"train_loss": 0.08012419939041138, "global_step": 197369, "epoch": 2217, "lr": 1.7337894149142453e-05}
{"train_loss": 0.06354672461748123, "global_step": 197370, "epoch": 2217, "lr": 1.733745520429933e-05}
{"train_loss": 0.04972809553146362, "global_step": 197371, "epoch": 2217, "lr": 1.7337016263847287e-05}
{"train_loss": 0.048110850155353546, "global_step": 197372, "epoch": 2217, "lr": 1.733657732778641e-05}
{"train_loss": 0.10332602262496948, "global_step": 197373, "epoch": 2217, "lr": 1.7336138396116708e-05}
{"train_loss": 0.05873802676796913, "global_step": 197374, "epoch": 2217, "lr": 1.733569946883828e-05}
{"train_loss": 0.052497901022434235, "global_step": 197375, "epoch": 2217, "lr": 1.7335260545951153e-05}
{"train_loss": 0.06766747683286667, "global_step": 197376, "epoch": 2217, "lr": 1.7334821627455417e-05}
{"train_loss": 0.05374708026647568, "global_step": 197377, "epoch": 2217, "lr": 1.7334382713351123e-05}
{"train_loss": 0.12207640707492828, "global_step": 197378, "epoch": 2217, "lr": 1.73339438036383e-05}
{"train_loss": 0.03588993102312088, "global_step": 197379, "epoch": 2217, "lr": 1.733350489831705e-05}
{"train_loss": 0.11595219373703003, "global_step": 197380, "epoch": 2217, "lr": 1.7333065997387393e-05}
{"train_loss": 0.04707551747560501, "global_step": 197381, "epoch": 2217, "lr": 1.7332627100849424e-05}
{"train_loss": 0.08660834282636642, "global_step": 197382, "epoch": 2217, "lr": 1.7332188208703186e-05}
{"train_loss": 0.06127128750085831, "global_step": 197383, "epoch": 2217, "lr": 1.733174932094872e-05}
{"train_loss": 0.0542227178812027, "global_step": 197384, "epoch": 2217, "lr": 1.7331310437586123e-05}
{"train_loss": 0.04855911433696747, "global_step": 197385, "epoch": 2217, "lr": 1.7330871558615414e-05}
{"train_loss": 0.06822499632835388, "global_step": 197386, "epoch": 2217, "lr": 1.7330432684036686e-05}
{"train_loss": 0.04118316248059273, "global_step": 197387, "epoch": 2217, "lr": 1.7329993813849967e-05}
{"train_loss": 0.08343365788459778, "global_step": 197388, "epoch": 2217, "lr": 1.732955494805535e-05}
{"train_loss": 0.05566861107945442, "global_step": 197389, "epoch": 2217, "lr": 1.7329116086652857e-05}
{"train_loss": 0.08233506232500076, "global_step": 197390, "epoch": 2217, "lr": 1.7328677229642564e-05}
{"train_loss": 0.05158011615276337, "global_step": 197391, "epoch": 2217, "lr": 1.732823837702455e-05}
{"train_loss": 0.03931054845452309, "global_step": 197392, "epoch": 2217, "lr": 1.7327799528798845e-05}
{"train_loss": 0.061312802135944366, "global_step": 197393, "epoch": 2217, "lr": 1.7327360684965532e-05}
{"train_loss": 0.08867660164833069, "global_step": 197394, "epoch": 2217, "lr": 1.732692184552464e-05}
{"train_loss": 0.04836759716272354, "global_step": 197395, "epoch": 2217, "lr": 1.7326483010476264e-05}
{"train_loss": 0.0518798753619194, "global_step": 197396, "epoch": 2217, "lr": 1.7326044179820426e-05}
{"train_loss": 0.05622521787881851, "global_step": 197397, "epoch": 2217, "lr": 1.7325605353557223e-05}
{"train_loss": 0.038674212992191315, "global_step": 197398, "epoch": 2217, "lr": 1.7325166531686676e-05}
{"train_loss": 0.10372089594602585, "global_step": 197399, "epoch": 2217, "lr": 1.7324727714208878e-05}
{"train_loss": 0.0721217691898346, "global_step": 197400, "epoch": 2217, "lr": 1.7324288901123855e-05}
{"train_loss": 0.06063898146319925, "global_step": 197401, "epoch": 2217, "lr": 1.73238500924317e-05, "val_loss": 8.269314765930176}
{"train_loss": 0.04107547178864479, "global_step": 197402, "epoch": 2218, "lr": 1.7323411288132434e-05}
{"train_loss": 0.03744151443243027, "global_step": 197403, "epoch": 2218, "lr": 1.732297248822616e-05}
{"train_loss": 0.05954407900571823, "global_step": 197404, "epoch": 2218, "lr": 1.7322533692712895e-05}
{"train_loss": 0.05183516815304756, "global_step": 197405, "epoch": 2218, "lr": 1.7322094901592733e-05}
{"train_loss": 0.04784715920686722, "global_step": 197406, "epoch": 2218, "lr": 1.7321656114865704e-05}
{"train_loss": 0.06289996951818466, "global_step": 197407, "epoch": 2218, "lr": 1.7321217332531877e-05}
{"train_loss": 0.0785043016076088, "global_step": 197408, "epoch": 2218, "lr": 1.7320778554591327e-05}
{"train_loss": 0.0893859714269638, "global_step": 197409, "epoch": 2218, "lr": 1.7320339781044087e-05}
{"train_loss": 0.0778006985783577, "global_step": 197410, "epoch": 2218, "lr": 1.7319901011890248e-05}
{"train_loss": 0.03587384894490242, "global_step": 197411, "epoch": 2218, "lr": 1.731946224712983e-05}
{"train_loss": 0.053926799446344376, "global_step": 197412, "epoch": 2218, "lr": 1.731902348676293e-05}
{"train_loss": 0.05500102415680885, "global_step": 197413, "epoch": 2218, "lr": 1.731858473078957e-05}
{"train_loss": 0.08071091771125793, "global_step": 197414, "epoch": 2218, "lr": 1.731814597920985e-05}
{"train_loss": 0.08419708907604218, "global_step": 197415, "epoch": 2218, "lr": 1.7317707232023806e-05}
{"train_loss": 0.0406927689909935, "global_step": 197416, "epoch": 2218, "lr": 1.7317268489231475e-05}
{"train_loss": 0.04937879368662834, "global_step": 197417, "epoch": 2218, "lr": 1.7316829750832965e-05}
{"train_loss": 0.06297940760850906, "global_step": 197418, "epoch": 2218, "lr": 1.7316391016828283e-05}
{"train_loss": 0.08206214755773544, "global_step": 197419, "epoch": 2218, "lr": 1.7315952287217536e-05}
{"train_loss": 0.05285535007715225, "global_step": 197420, "epoch": 2218, "lr": 1.7315513562000756e-05}
{"train_loss": 0.05185916647315025, "global_step": 197421, "epoch": 2218, "lr": 1.7315074841177996e-05}
{"train_loss": 0.06275177001953125, "global_step": 197422, "epoch": 2218, "lr": 1.731463612474934e-05}
{"train_loss": 0.0343506783246994, "global_step": 197423, "epoch": 2218, "lr": 1.7314197412714818e-05}
{"train_loss": 0.06724762171506882, "global_step": 197424, "epoch": 2218, "lr": 1.7313758705074517e-05}
{"train_loss": 0.022967344149947166, "global_step": 197425, "epoch": 2218, "lr": 1.731332000182847e-05}
{"train_loss": 0.055242475122213364, "global_step": 197426, "epoch": 2218, "lr": 1.7312881302976746e-05}
{"train_loss": 0.06785434484481812, "global_step": 197427, "epoch": 2218, "lr": 1.7312442608519425e-05}
{"train_loss": 0.052881963551044464, "global_step": 197428, "epoch": 2218, "lr": 1.7312003918456525e-05}
{"train_loss": 0.03504037857055664, "global_step": 197429, "epoch": 2218, "lr": 1.731156523278815e-05}
{"train_loss": 0.07727042585611343, "global_step": 197430, "epoch": 2218, "lr": 1.7311126551514318e-05}
{"train_loss": 0.08546961098909378, "global_step": 197431, "epoch": 2218, "lr": 1.731068787463513e-05}
{"train_loss": 0.054773855954408646, "global_step": 197432, "epoch": 2218, "lr": 1.7310249202150596e-05}
{"train_loss": 0.10877808928489685, "global_step": 197433, "epoch": 2218, "lr": 1.730981053406082e-05}
{"train_loss": 0.05943690240383148, "global_step": 197434, "epoch": 2218, "lr": 1.7309371870365826e-05}
{"train_loss": 0.06572459638118744, "global_step": 197435, "epoch": 2218, "lr": 1.7308933211065703e-05}
{"train_loss": 0.0339416041970253, "global_step": 197436, "epoch": 2218, "lr": 1.730849455616048e-05}
{"train_loss": 0.06608691811561584, "global_step": 197437, "epoch": 2218, "lr": 1.730805590565025e-05}
{"train_loss": 0.09519712626934052, "global_step": 197438, "epoch": 2218, "lr": 1.7307617259535035e-05}
{"train_loss": 0.06683706492185593, "global_step": 197439, "epoch": 2218, "lr": 1.7307178617814928e-05}
{"train_loss": 0.03551860153675079, "global_step": 197440, "epoch": 2218, "lr": 1.7306739980489956e-05}
{"train_loss": 0.10827566683292389, "global_step": 197441, "epoch": 2218, "lr": 1.7306301347560217e-05}
{"train_loss": 0.019793670624494553, "global_step": 197442, "epoch": 2218, "lr": 1.7305862719025722e-05}
{"train_loss": 0.07684096693992615, "global_step": 197443, "epoch": 2218, "lr": 1.7305424094886573e-05}
{"train_loss": 0.07528439164161682, "global_step": 197444, "epoch": 2218, "lr": 1.73049854751428e-05}
{"train_loss": 0.06419260054826736, "global_step": 197445, "epoch": 2218, "lr": 1.7304546859794467e-05}
{"train_loss": 0.0482008121907711, "global_step": 197446, "epoch": 2218, "lr": 1.7304108248841662e-05}
{"train_loss": 0.06945346295833588, "global_step": 197447, "epoch": 2218, "lr": 1.7303669642284403e-05}
{"train_loss": 0.07367566227912903, "global_step": 197448, "epoch": 2218, "lr": 1.730323104012278e-05}
{"train_loss": 0.0652427151799202, "global_step": 197449, "epoch": 2218, "lr": 1.7302792442356825e-05}
{"train_loss": 0.04087791219353676, "global_step": 197450, "epoch": 2218, "lr": 1.730235384898663e-05}
{"train_loss": 0.09468764066696167, "global_step": 197451, "epoch": 2218, "lr": 1.7301915260012215e-05}
{"train_loss": 0.07940996438264847, "global_step": 197452, "epoch": 2218, "lr": 1.7301476675433675e-05}
{"train_loss": 0.06767740100622177, "global_step": 197453, "epoch": 2218, "lr": 1.7301038095251054e-05}
{"train_loss": 0.06233055517077446, "global_step": 197454, "epoch": 2218, "lr": 1.7300599519464394e-05}
{"train_loss": 0.0974351316690445, "global_step": 197455, "epoch": 2218, "lr": 1.730016094807379e-05}
{"train_loss": 0.046914178878068924, "global_step": 197456, "epoch": 2218, "lr": 1.729972238107926e-05}
{"train_loss": 0.06137916073203087, "global_step": 197457, "epoch": 2218, "lr": 1.72992838184809e-05}
{"train_loss": 0.09644938260316849, "global_step": 197458, "epoch": 2218, "lr": 1.7298845260278735e-05}
{"train_loss": 0.07479678094387054, "global_step": 197459, "epoch": 2218, "lr": 1.7298406706472858e-05}
{"train_loss": 0.04456451162695885, "global_step": 197460, "epoch": 2218, "lr": 1.7297968157063316e-05}
{"train_loss": 0.06318586319684982, "global_step": 197461, "epoch": 2218, "lr": 1.7297529612050145e-05}
{"train_loss": 0.07217445969581604, "global_step": 197462, "epoch": 2218, "lr": 1.7297091071433425e-05}
{"train_loss": 0.07320582121610641, "global_step": 197463, "epoch": 2218, "lr": 1.7296652535213226e-05}
{"train_loss": 0.04915191978216171, "global_step": 197464, "epoch": 2218, "lr": 1.7296214003389576e-05}
{"train_loss": 0.05331229791045189, "global_step": 197465, "epoch": 2218, "lr": 1.7295775475962568e-05}
{"train_loss": 0.061123255640268326, "global_step": 197466, "epoch": 2218, "lr": 1.7295336952932227e-05}
{"train_loss": 0.1374935656785965, "global_step": 197467, "epoch": 2218, "lr": 1.729489843429865e-05}
{"train_loss": 0.07580678910017014, "global_step": 197468, "epoch": 2218, "lr": 1.7294459920061858e-05}
{"train_loss": 0.0878671333193779, "global_step": 197469, "epoch": 2218, "lr": 1.7294021410221937e-05}
{"train_loss": 0.07478897273540497, "global_step": 197470, "epoch": 2218, "lr": 1.7293582904778928e-05}
{"train_loss": 0.09453873336315155, "global_step": 197471, "epoch": 2218, "lr": 1.7293144403732914e-05}
{"train_loss": 0.036931801587343216, "global_step": 197472, "epoch": 2218, "lr": 1.7292705907083916e-05}
{"train_loss": 0.056450843811035156, "global_step": 197473, "epoch": 2218, "lr": 1.7292267414832035e-05}
{"train_loss": 0.04471983015537262, "global_step": 197474, "epoch": 2218, "lr": 1.7291828926977295e-05}
{"train_loss": 0.022792702540755272, "global_step": 197475, "epoch": 2218, "lr": 1.7291390443519785e-05}
{"train_loss": 0.08132234960794449, "global_step": 197476, "epoch": 2218, "lr": 1.7290951964459534e-05}
{"train_loss": 0.08637870848178864, "global_step": 197477, "epoch": 2218, "lr": 1.7290513489796628e-05}
{"train_loss": 0.11931120604276657, "global_step": 197478, "epoch": 2218, "lr": 1.72900750195311e-05}
{"train_loss": 0.039736125618219376, "global_step": 197479, "epoch": 2218, "lr": 1.728963655366304e-05}
{"train_loss": 0.04329770430922508, "global_step": 197480, "epoch": 2218, "lr": 1.728919809219247e-05}
{"train_loss": 0.04008525237441063, "global_step": 197481, "epoch": 2218, "lr": 1.7288759635119474e-05}
{"train_loss": 0.059029027819633484, "global_step": 197482, "epoch": 2218, "lr": 1.7288321182444116e-05}
{"train_loss": 0.10938242077827454, "global_step": 197483, "epoch": 2218, "lr": 1.7287882734166432e-05}
{"train_loss": 0.08103732019662857, "global_step": 197484, "epoch": 2218, "lr": 1.728744429028651e-05}
{"train_loss": 0.033515915274620056, "global_step": 197485, "epoch": 2218, "lr": 1.7287005850804378e-05}
{"train_loss": 0.06346549838781357, "global_step": 197486, "epoch": 2218, "lr": 1.7286567415720123e-05}
{"train_loss": 0.061666134744882584, "global_step": 197487, "epoch": 2218, "lr": 1.728612898503377e-05}
{"train_loss": 0.051722634583711624, "global_step": 197488, "epoch": 2218, "lr": 1.728569055874542e-05}
{"train_loss": 0.04633650556206703, "global_step": 197489, "epoch": 2218, "lr": 1.728525213685509e-05}
{"train_loss": 0.063826771139094, "global_step": 197490, "epoch": 2218, "lr": 1.728481371936288e-05, "val_loss": 8.193065643310547}
{"train_loss": 0.031674280762672424, "global_step": 197491, "epoch": 2219, "lr": 1.7284375306268806e-05}
{"train_loss": 0.06000090017914772, "global_step": 197492, "epoch": 2219, "lr": 1.728393689757297e-05}
{"train_loss": 0.03057112917304039, "global_step": 197493, "epoch": 2219, "lr": 1.7283498493275406e-05}
{"train_loss": 0.055057186633348465, "global_step": 197494, "epoch": 2219, "lr": 1.7283060093376165e-05}
{"train_loss": 0.10679136961698532, "global_step": 197495, "epoch": 2219, "lr": 1.7282621697875335e-05}
{"train_loss": 0.053254060447216034, "global_step": 197496, "epoch": 2219, "lr": 1.7282183306772937e-05}
{"train_loss": 0.08886072039604187, "global_step": 197497, "epoch": 2219, "lr": 1.728174492006907e-05}
{"train_loss": 0.05628621578216553, "global_step": 197498, "epoch": 2219, "lr": 1.7281306537763752e-05}
{"train_loss": 0.03539387136697769, "global_step": 197499, "epoch": 2219, "lr": 1.7280868159857087e-05}
{"train_loss": 0.07040450721979141, "global_step": 197500, "epoch": 2219, "lr": 1.728042978634909e-05}
{"train_loss": 0.05079608038067818, "global_step": 197501, "epoch": 2219, "lr": 1.7279991417239854e-05}
{"train_loss": 0.05496928095817566, "global_step": 197502, "epoch": 2219, "lr": 1.7279553052529413e-05}
{"train_loss": 0.041826240718364716, "global_step": 197503, "epoch": 2219, "lr": 1.7279114692217856e-05}
{"train_loss": 0.08156316727399826, "global_step": 197504, "epoch": 2219, "lr": 1.72786763363052e-05}
{"train_loss": 0.06415900588035583, "global_step": 197505, "epoch": 2219, "lr": 1.7278237984791544e-05}
{"train_loss": 0.07967919111251831, "global_step": 197506, "epoch": 2219, "lr": 1.7277799637676923e-05}
{"train_loss": 0.10007458925247192, "global_step": 197507, "epoch": 2219, "lr": 1.7277361294961407e-05}
{"train_loss": 0.03349902108311653, "global_step": 197508, "epoch": 2219, "lr": 1.7276922956645042e-05}
{"train_loss": 0.09678123146295547, "global_step": 197509, "epoch": 2219, "lr": 1.7276484622727912e-05}
{"train_loss": 0.04241640120744705, "global_step": 197510, "epoch": 2219, "lr": 1.727604629321004e-05}
{"train_loss": 0.051832206547260284, "global_step": 197511, "epoch": 2219, "lr": 1.7275607968091523e-05}
{"train_loss": 0.16513743996620178, "global_step": 197512, "epoch": 2219, "lr": 1.7275169647372386e-05}
{"train_loss": 0.051999885588884354, "global_step": 197513, "epoch": 2219, "lr": 1.7274731331052718e-05}
{"train_loss": 0.05146588385105133, "global_step": 197514, "epoch": 2219, "lr": 1.7274293019132552e-05}
{"train_loss": 0.060553889721632004, "global_step": 197515, "epoch": 2219, "lr": 1.727385471161197e-05}
{"train_loss": 0.03711143881082535, "global_step": 197516, "epoch": 2219, "lr": 1.7273416408491007e-05}
{"train_loss": 0.03843279555439949, "global_step": 197517, "epoch": 2219, "lr": 1.727297810976973e-05}
{"train_loss": 0.0698072537779808, "global_step": 197518, "epoch": 2219, "lr": 1.7272539815448218e-05}
{"train_loss": 0.05151478201150894, "global_step": 197519, "epoch": 2219, "lr": 1.72721015255265e-05}
{"train_loss": 0.035288747400045395, "global_step": 197520, "epoch": 2219, "lr": 1.7271663240004666e-05}
{"train_loss": 0.08393155038356781, "global_step": 197521, "epoch": 2219, "lr": 1.7271224958882744e-05}
{"train_loss": 0.04574437439441681, "global_step": 197522, "epoch": 2219, "lr": 1.7270786682160816e-05}
{"train_loss": 0.10558491200208664, "global_step": 197523, "epoch": 2219, "lr": 1.7270348409838922e-05}
{"train_loss": 0.07288328558206558, "global_step": 197524, "epoch": 2219, "lr": 1.7269910141917144e-05}
{"train_loss": 0.08279891312122345, "global_step": 197525, "epoch": 2219, "lr": 1.726947187839551e-05}
{"train_loss": 0.057467084378004074, "global_step": 197526, "epoch": 2219, "lr": 1.7269033619274116e-05}
{"train_loss": 0.055385276675224304, "global_step": 197527, "epoch": 2219, "lr": 1.7268595364552988e-05}
{"train_loss": 0.026219571009278297, "global_step": 197528, "epoch": 2219, "lr": 1.726815711423221e-05}
{"train_loss": 0.042776528745889664, "global_step": 197529, "epoch": 2219, "lr": 1.726771886831181e-05}
{"train_loss": 0.08138618618249893, "global_step": 197530, "epoch": 2219, "lr": 1.7267280626791887e-05}
{"train_loss": 0.1242486909031868, "global_step": 197531, "epoch": 2219, "lr": 1.7266842389672476e-05}
{"train_loss": 0.03967129811644554, "global_step": 197532, "epoch": 2219, "lr": 1.7266404156953632e-05}
{"train_loss": 0.02614750526845455, "global_step": 197533, "epoch": 2219, "lr": 1.7265965928635425e-05}
{"train_loss": 0.0501943938434124, "global_step": 197534, "epoch": 2219, "lr": 1.72655277047179e-05}
{"train_loss": 0.06870096921920776, "global_step": 197535, "epoch": 2219, "lr": 1.726508948520114e-05}
{"train_loss": 0.09097305685281754, "global_step": 197536, "epoch": 2219, "lr": 1.7264651270085176e-05}
{"train_loss": 0.049301065504550934, "global_step": 197537, "epoch": 2219, "lr": 1.72642130593701e-05}
{"train_loss": 0.09767557680606842, "global_step": 197538, "epoch": 2219, "lr": 1.7263774853055932e-05}
{"train_loss": 0.052224356681108475, "global_step": 197539, "epoch": 2219, "lr": 1.7263336651142764e-05}
{"train_loss": 0.07772819697856903, "global_step": 197540, "epoch": 2219, "lr": 1.7262898453630628e-05}
{"train_loss": 0.10553821921348572, "global_step": 197541, "epoch": 2219, "lr": 1.726246026051961e-05}
{"train_loss": 0.03407353162765503, "global_step": 197542, "epoch": 2219, "lr": 1.726202207180974e-05}
{"train_loss": 0.09114563465118408, "global_step": 197543, "epoch": 2219, "lr": 1.726158388750111e-05}
{"train_loss": 0.07350780069828033, "global_step": 197544, "epoch": 2219, "lr": 1.726114570759374e-05}
{"train_loss": 0.02956550009548664, "global_step": 197545, "epoch": 2219, "lr": 1.7260707532087732e-05}
{"train_loss": 0.09141489118337631, "global_step": 197546, "epoch": 2219, "lr": 1.7260269360983104e-05}
{"train_loss": 0.07825314253568649, "global_step": 197547, "epoch": 2219, "lr": 1.725983119427995e-05}
{"train_loss": 0.046072401106357574, "global_step": 197548, "epoch": 2219, "lr": 1.7259393031978293e-05}
{"train_loss": 0.10839243233203888, "global_step": 197549, "epoch": 2219, "lr": 1.7258954874078238e-05}
{"train_loss": 0.056441739201545715, "global_step": 197550, "epoch": 2219, "lr": 1.7258516720579793e-05}
{"train_loss": 0.07125870883464813, "global_step": 197551, "epoch": 2219, "lr": 1.7258078571483056e-05}
{"train_loss": 0.1091899499297142, "global_step": 197552, "epoch": 2219, "lr": 1.725764042678806e-05}
{"train_loss": 0.07530225813388824, "global_step": 197553, "epoch": 2219, "lr": 1.725720228649487e-05}
{"train_loss": 0.05412241816520691, "global_step": 197554, "epoch": 2219, "lr": 1.725676415060357e-05}
{"train_loss": 0.10402484238147736, "global_step": 197555, "epoch": 2219, "lr": 1.7256326019114183e-05}
{"train_loss": 0.05314188450574875, "global_step": 197556, "epoch": 2219, "lr": 1.7255887892026802e-05}
{"train_loss": 0.07216314226388931, "global_step": 197557, "epoch": 2219, "lr": 1.7255449769341447e-05}
{"train_loss": 0.12503425776958466, "global_step": 197558, "epoch": 2219, "lr": 1.7255011651058217e-05}
{"train_loss": 0.049057237803936005, "global_step": 197559, "epoch": 2219, "lr": 1.7254573537177137e-05}
{"train_loss": 0.09468822181224823, "global_step": 197560, "epoch": 2219, "lr": 1.7254135427698293e-05}
{"train_loss": 0.08058148622512817, "global_step": 197561, "epoch": 2219, "lr": 1.7253697322621716e-05}
{"train_loss": 0.03595583140850067, "global_step": 197562, "epoch": 2219, "lr": 1.72532592219475e-05}
{"train_loss": 0.06572262197732925, "global_step": 197563, "epoch": 2219, "lr": 1.725282112567566e-05}
{"train_loss": 0.07580319046974182, "global_step": 197564, "epoch": 2219, "lr": 1.7252383033806302e-05}
{"train_loss": 0.05099347233772278, "global_step": 197565, "epoch": 2219, "lr": 1.7251944946339443e-05}
{"train_loss": 0.11799229681491852, "global_step": 197566, "epoch": 2219, "lr": 1.725150686327518e-05}
{"train_loss": 0.08303126692771912, "global_step": 197567, "epoch": 2219, "lr": 1.7251068784613535e-05}
{"train_loss": 0.09123729914426804, "global_step": 197568, "epoch": 2219, "lr": 1.72506307103546e-05}
{"train_loss": 0.05519162863492966, "global_step": 197569, "epoch": 2219, "lr": 1.7250192640498418e-05}
{"train_loss": 0.03578726574778557, "global_step": 197570, "epoch": 2219, "lr": 1.724975457504503e-05}
{"train_loss": 0.07322873920202255, "global_step": 197571, "epoch": 2219, "lr": 1.7249316513994536e-05}
{"train_loss": 0.09074041992425919, "global_step": 197572, "epoch": 2219, "lr": 1.724887845734695e-05}
{"train_loss": 0.09923011809587479, "global_step": 197573, "epoch": 2219, "lr": 1.7248440405102372e-05}
{"train_loss": 0.07914222031831741, "global_step": 197574, "epoch": 2219, "lr": 1.724800235726083e-05}
{"train_loss": 0.05982581526041031, "global_step": 197575, "epoch": 2219, "lr": 1.7247564313822406e-05}
{"train_loss": 0.061465609818696976, "global_step": 197576, "epoch": 2219, "lr": 1.7247126274787128e-05}
{"train_loss": 0.03596295788884163, "global_step": 197577, "epoch": 2219, "lr": 1.7246688240155097e-05}
{"train_loss": 0.04633639007806778, "global_step": 197578, "epoch": 2219, "lr": 1.7246250209926336e-05}
{"train_loss": 0.06799085235244103, "global_step": 197579, "epoch": 2219, "lr": 1.7245812184100923e-05, "val_loss": 8.514610290527344}
{"train_loss": 0.04694085568189621, "global_step": 197580, "epoch": 2220, "lr": 1.72453741626789e-05}
{"train_loss": 0.03500376641750336, "global_step": 197581, "epoch": 2220, "lr": 1.7244936145660356e-05}
{"train_loss": 0.05467851459980011, "global_step": 197582, "epoch": 2220, "lr": 1.7244498133045317e-05}
{"train_loss": 0.05282900482416153, "global_step": 197583, "epoch": 2220, "lr": 1.7244060124833866e-05}
{"train_loss": 0.08656664192676544, "global_step": 197584, "epoch": 2220, "lr": 1.724362212102604e-05}
{"train_loss": 0.17628167569637299, "global_step": 197585, "epoch": 2220, "lr": 1.7243184121621924e-05}
{"train_loss": 0.05288667604327202, "global_step": 197586, "epoch": 2220, "lr": 1.7242746126621552e-05}
{"train_loss": 0.045246291905641556, "global_step": 197587, "epoch": 2220, "lr": 1.7242308136025005e-05}
{"train_loss": 0.05827352777123451, "global_step": 197588, "epoch": 2220, "lr": 1.724187014983231e-05}
{"train_loss": 0.05195825546979904, "global_step": 197589, "epoch": 2220, "lr": 1.724143216804355e-05}
{"train_loss": 0.06002524867653847, "global_step": 197590, "epoch": 2220, "lr": 1.72409941906588e-05}
{"train_loss": 0.038717929273843765, "global_step": 197591, "epoch": 2220, "lr": 1.7240556217678077e-05}
{"train_loss": 0.049262724816799164, "global_step": 197592, "epoch": 2220, "lr": 1.7240118249101477e-05}
{"train_loss": 0.0957847312092781, "global_step": 197593, "epoch": 2220, "lr": 1.723968028492903e-05}
{"train_loss": 0.08264444768428802, "global_step": 197594, "epoch": 2220, "lr": 1.723924232516083e-05}
{"train_loss": 0.06708096712827682, "global_step": 197595, "epoch": 2220, "lr": 1.7238804369796895e-05}
{"train_loss": 0.05830301716923714, "global_step": 197596, "epoch": 2220, "lr": 1.723836641883732e-05}
{"train_loss": 0.07327854633331299, "global_step": 197597, "epoch": 2220, "lr": 1.723792847228213e-05}
{"train_loss": 0.06589347869157791, "global_step": 197598, "epoch": 2220, "lr": 1.7237490530131418e-05}
{"train_loss": 0.07533364742994308, "global_step": 197599, "epoch": 2220, "lr": 1.7237052592385204e-05}
{"train_loss": 0.04862077161669731, "global_step": 197600, "epoch": 2220, "lr": 1.7236614659043592e-05}
{"train_loss": 0.0762309730052948, "global_step": 197601, "epoch": 2220, "lr": 1.7236176730106602e-05}
{"train_loss": 0.09312403202056885, "global_step": 197602, "epoch": 2220, "lr": 1.7235738805574326e-05}
{"train_loss": 0.04731559753417969, "global_step": 197603, "epoch": 2220, "lr": 1.7235300885446787e-05}
{"train_loss": 0.050733234733343124, "global_step": 197604, "epoch": 2220, "lr": 1.7234862969724077e-05}
{"train_loss": 0.05575697496533394, "global_step": 197605, "epoch": 2220, "lr": 1.7234425058406222e-05}
{"train_loss": 0.08473484963178635, "global_step": 197606, "epoch": 2220, "lr": 1.723398715149332e-05}
{"train_loss": 0.05905427783727646, "global_step": 197607, "epoch": 2220, "lr": 1.7233549248985393e-05}
{"train_loss": 0.06181538850069046, "global_step": 197608, "epoch": 2220, "lr": 1.723311135088253e-05}
{"train_loss": 0.023408975452184677, "global_step": 197609, "epoch": 2220, "lr": 1.7232673457184778e-05}
{"train_loss": 0.028856687247753143, "global_step": 197610, "epoch": 2220, "lr": 1.7232235567892175e-05}
{"train_loss": 0.06908424943685532, "global_step": 197611, "epoch": 2220, "lr": 1.723179768300482e-05}
{"train_loss": 0.09510212391614914, "global_step": 197612, "epoch": 2220, "lr": 1.7231359802522728e-05}
{"train_loss": 0.06475292146205902, "global_step": 197613, "epoch": 2220, "lr": 1.7230921926445998e-05}
{"train_loss": 0.054338451474905014, "global_step": 197614, "epoch": 2220, "lr": 1.7230484054774653e-05}
{"train_loss": 0.03639103099703789, "global_step": 197615, "epoch": 2220, "lr": 1.723004618750879e-05}
{"train_loss": 0.10352331399917603, "global_step": 197616, "epoch": 2220, "lr": 1.7229608324648427e-05}
{"train_loss": 0.0696944072842598, "global_step": 197617, "epoch": 2220, "lr": 1.722917046619367e-05}
{"train_loss": 0.04808668792247772, "global_step": 197618, "epoch": 2220, "lr": 1.722873261214453e-05}
{"train_loss": 0.0967211127281189, "global_step": 197619, "epoch": 2220, "lr": 1.7228294762501095e-05}
{"train_loss": 0.039253585040569305, "global_step": 197620, "epoch": 2220, "lr": 1.7227856917263408e-05}
{"train_loss": 0.05372384935617447, "global_step": 197621, "epoch": 2220, "lr": 1.7227419076431556e-05}
{"train_loss": 0.11569610983133316, "global_step": 197622, "epoch": 2220, "lr": 1.7226981240005557e-05}
{"train_loss": 0.07585816085338593, "global_step": 197623, "epoch": 2220, "lr": 1.7226543407985507e-05}
{"train_loss": 0.09489332139492035, "global_step": 197624, "epoch": 2220, "lr": 1.7226105580371432e-05}
{"train_loss": 0.06145212799310684, "global_step": 197625, "epoch": 2220, "lr": 1.7225667757163412e-05}
{"train_loss": 0.05435850843787193, "global_step": 197626, "epoch": 2220, "lr": 1.7225229938361515e-05}
{"train_loss": 0.06006351485848427, "global_step": 197627, "epoch": 2220, "lr": 1.7224792123965765e-05}
{"train_loss": 0.06377431005239487, "global_step": 197628, "epoch": 2220, "lr": 1.7224354313976267e-05}
{"train_loss": 0.07892260700464249, "global_step": 197629, "epoch": 2220, "lr": 1.7223916508393034e-05}
{"train_loss": 0.05033988505601883, "global_step": 197630, "epoch": 2220, "lr": 1.722347870721616e-05}
{"train_loss": 0.020573627203702927, "global_step": 197631, "epoch": 2220, "lr": 1.7223040910445683e-05}
{"train_loss": 0.04375100135803223, "global_step": 197632, "epoch": 2220, "lr": 1.7222603118081677e-05}
{"train_loss": 0.038307417184114456, "global_step": 197633, "epoch": 2220, "lr": 1.722216533012418e-05}
{"train_loss": 0.11450594663619995, "global_step": 197634, "epoch": 2220, "lr": 1.7221727546573273e-05}
{"train_loss": 0.025218814611434937, "global_step": 197635, "epoch": 2220, "lr": 1.7221289767428995e-05}
{"train_loss": 0.05532969534397125, "global_step": 197636, "epoch": 2220, "lr": 1.7220851992691432e-05}
{"train_loss": 0.07443099468946457, "global_step": 197637, "epoch": 2220, "lr": 1.722041422236061e-05}
{"train_loss": 0.06616850942373276, "global_step": 197638, "epoch": 2220, "lr": 1.7219976456436615e-05}
{"train_loss": 0.028124496340751648, "global_step": 197639, "epoch": 2220, "lr": 1.7219538694919478e-05}
{"train_loss": 0.08890112489461899, "global_step": 197640, "epoch": 2220, "lr": 1.7219100937809295e-05}
{"train_loss": 0.054761115461587906, "global_step": 197641, "epoch": 2220, "lr": 1.7218663185106088e-05}
{"train_loss": 0.0777730643749237, "global_step": 197642, "epoch": 2220, "lr": 1.721822543680995e-05}
{"train_loss": 0.059393253177404404, "global_step": 197643, "epoch": 2220, "lr": 1.72177876929209e-05}
{"train_loss": 0.06729987263679504, "global_step": 197644, "epoch": 2220, "lr": 1.7217349953439038e-05}
{"train_loss": 0.06414955109357834, "global_step": 197645, "epoch": 2220, "lr": 1.721691221836439e-05}
{"train_loss": 0.05596121400594711, "global_step": 197646, "epoch": 2220, "lr": 1.7216474487697043e-05}
{"train_loss": 0.03927123546600342, "global_step": 197647, "epoch": 2220, "lr": 1.7216036761437043e-05}
{"train_loss": 0.07951854169368744, "global_step": 197648, "epoch": 2220, "lr": 1.721559903958443e-05}
{"train_loss": 0.04871874302625656, "global_step": 197649, "epoch": 2220, "lr": 1.7215161322139295e-05}
{"train_loss": 0.06116792932152748, "global_step": 197650, "epoch": 2220, "lr": 1.7214723609101663e-05}
{"train_loss": 0.045386940240859985, "global_step": 197651, "epoch": 2220, "lr": 1.7214285900471634e-05}
{"train_loss": 0.04083865135908127, "global_step": 197652, "epoch": 2220, "lr": 1.7213848196249222e-05}
{"train_loss": 0.06063881516456604, "global_step": 197653, "epoch": 2220, "lr": 1.721341049643453e-05}
{"train_loss": 0.044713668525218964, "global_step": 197654, "epoch": 2220, "lr": 1.721297280102758e-05}
{"train_loss": 0.06635940074920654, "global_step": 197655, "epoch": 2220, "lr": 1.7212535110028456e-05}
{"train_loss": 0.06011933088302612, "global_step": 197656, "epoch": 2220, "lr": 1.7212097423437197e-05}
{"train_loss": 0.0531315952539444, "global_step": 197657, "epoch": 2220, "lr": 1.721165974125388e-05}
{"train_loss": 0.05947386100888252, "global_step": 197658, "epoch": 2220, "lr": 1.721122206347855e-05}
{"train_loss": 0.0571889653801918, "global_step": 197659, "epoch": 2220, "lr": 1.7210784390111284e-05}
{"train_loss": 0.07879593968391418, "global_step": 197660, "epoch": 2220, "lr": 1.721034672115211e-05}
{"train_loss": 0.04658691585063934, "global_step": 197661, "epoch": 2220, "lr": 1.720990905660111e-05}
{"train_loss": 0.1236940249800682, "global_step": 197662, "epoch": 2220, "lr": 1.7209471396458347e-05}
{"train_loss": 0.059642497450113297, "global_step": 197663, "epoch": 2220, "lr": 1.7209033740723856e-05}
{"train_loss": 0.05078698322176933, "global_step": 197664, "epoch": 2220, "lr": 1.7208596089397733e-05}
{"train_loss": 0.032670337706804276, "global_step": 197665, "epoch": 2220, "lr": 1.7208158442479994e-05}
{"train_loss": 0.08470871299505234, "global_step": 197666, "epoch": 2220, "lr": 1.720772079997074e-05}
{"train_loss": 0.10801635682582855, "global_step": 197667, "epoch": 2220, "lr": 1.7207283161869986e-05}
{"train_loss": 0.06345538866151584, "global_step": 197668, "epoch": 2220, "lr": 1.720684552817783e-05, "val_loss": 8.364096641540527, "train_action_mse_error": 18.916095733642578}
{"train_loss": 0.0709141418337822, "global_step": 197669, "epoch": 2221, "lr": 1.7206407898894295e-05}
{"train_loss": 0.03141362592577934, "global_step": 197670, "epoch": 2221, "lr": 1.7205970274019477e-05}
{"train_loss": 0.040668413043022156, "global_step": 197671, "epoch": 2221, "lr": 1.7205532653553408e-05}
{"train_loss": 0.048819649964571, "global_step": 197672, "epoch": 2221, "lr": 1.7205095037496165e-05}
{"train_loss": 0.04958663135766983, "global_step": 197673, "epoch": 2221, "lr": 1.720465742584778e-05}
{"train_loss": 0.09103357046842575, "global_step": 197674, "epoch": 2221, "lr": 1.7204219818608347e-05}
{"train_loss": 0.058484338223934174, "global_step": 197675, "epoch": 2221, "lr": 1.7203782215777893e-05}
{"train_loss": 0.09862081706523895, "global_step": 197676, "epoch": 2221, "lr": 1.7203344617356503e-05}
{"train_loss": 0.06830111891031265, "global_step": 197677, "epoch": 2221, "lr": 1.7202907023344204e-05}
{"train_loss": 0.07529766112565994, "global_step": 197678, "epoch": 2221, "lr": 1.72024694337411e-05}
{"train_loss": 0.07404117286205292, "global_step": 197679, "epoch": 2221, "lr": 1.7202031848547202e-05}
{"train_loss": 0.0661226361989975, "global_step": 197680, "epoch": 2221, "lr": 1.720159426776261e-05}
{"train_loss": 0.05545828863978386, "global_step": 197681, "epoch": 2221, "lr": 1.720115669138735e-05}
{"train_loss": 0.07247845828533173, "global_step": 197682, "epoch": 2221, "lr": 1.7200719119421503e-05}
{"train_loss": 0.10234221816062927, "global_step": 197683, "epoch": 2221, "lr": 1.720028155186511e-05}
{"train_loss": 0.05067414045333862, "global_step": 197684, "epoch": 2221, "lr": 1.719984398871825e-05}
{"train_loss": 0.06186281517148018, "global_step": 197685, "epoch": 2221, "lr": 1.719940642998098e-05}
{"train_loss": 0.0324375219643116, "global_step": 197686, "epoch": 2221, "lr": 1.7198968875653325e-05}
{"train_loss": 0.06900513917207718, "global_step": 197687, "epoch": 2221, "lr": 1.719853132573539e-05}
{"train_loss": 0.10500206798315048, "global_step": 197688, "epoch": 2221, "lr": 1.7198093780227192e-05}
{"train_loss": 0.05710799992084503, "global_step": 197689, "epoch": 2221, "lr": 1.7197656239128835e-05}
{"train_loss": 0.0324784591794014, "global_step": 197690, "epoch": 2221, "lr": 1.7197218702440325e-05}
{"train_loss": 0.0735112726688385, "global_step": 197691, "epoch": 2221, "lr": 1.7196781170161773e-05}
{"train_loss": 0.04147975519299507, "global_step": 197692, "epoch": 2221, "lr": 1.71963436422932e-05}
{"train_loss": 0.09146815538406372, "global_step": 197693, "epoch": 2221, "lr": 1.7195906118834693e-05}
{"train_loss": 0.10831629484891891, "global_step": 197694, "epoch": 2221, "lr": 1.719546859978628e-05}
{"train_loss": 0.05345318466424942, "global_step": 197695, "epoch": 2221, "lr": 1.719503108514805e-05}
{"train_loss": 0.04885882884263992, "global_step": 197696, "epoch": 2221, "lr": 1.7194593574920032e-05}
{"train_loss": 0.02977295219898224, "global_step": 197697, "epoch": 2221, "lr": 1.719415606910232e-05}
{"train_loss": 0.08229678869247437, "global_step": 197698, "epoch": 2221, "lr": 1.719371856769494e-05}
{"train_loss": 0.06593240797519684, "global_step": 197699, "epoch": 2221, "lr": 1.719328107069796e-05}
{"train_loss": 0.05437203124165535, "global_step": 197700, "epoch": 2221, "lr": 1.7192843578111462e-05}
{"train_loss": 0.06474481523036957, "global_step": 197701, "epoch": 2221, "lr": 1.7192406089935466e-05}
{"train_loss": 0.04628273472189903, "global_step": 197702, "epoch": 2221, "lr": 1.7191968606170072e-05}
{"train_loss": 0.05553853139281273, "global_step": 197703, "epoch": 2221, "lr": 1.7191531126815297e-05}
{"train_loss": 0.11169231683015823, "global_step": 197704, "epoch": 2221, "lr": 1.7191093651871243e-05}
{"train_loss": 0.08346983790397644, "global_step": 197705, "epoch": 2221, "lr": 1.7190656181337927e-05}
{"train_loss": 0.06284263730049133, "global_step": 197706, "epoch": 2221, "lr": 1.7190218715215443e-05}
{"train_loss": 0.08710508793592453, "global_step": 197707, "epoch": 2221, "lr": 1.7189781253503823e-05}
{"train_loss": 0.0631934106349945, "global_step": 197708, "epoch": 2221, "lr": 1.718934379620315e-05}
{"train_loss": 0.04378179460763931, "global_step": 197709, "epoch": 2221, "lr": 1.718890634331345e-05}
{"train_loss": 0.041127193719148636, "global_step": 197710, "epoch": 2221, "lr": 1.718846889483482e-05}
{"train_loss": 0.07815394550561905, "global_step": 197711, "epoch": 2221, "lr": 1.718803145076729e-05}
{"train_loss": 0.06623377650976181, "global_step": 197712, "epoch": 2221, "lr": 1.7187594011110936e-05}
{"train_loss": 0.01991199515759945, "global_step": 197713, "epoch": 2221, "lr": 1.7187156575865798e-05}
{"train_loss": 0.08698159456253052, "global_step": 197714, "epoch": 2221, "lr": 1.7186719145031964e-05}
{"train_loss": 0.1122007966041565, "global_step": 197715, "epoch": 2221, "lr": 1.718628171860946e-05}
{"train_loss": 0.046028200536966324, "global_step": 197716, "epoch": 2221, "lr": 1.718584429659836e-05}
{"train_loss": 0.027789391577243805, "global_step": 197717, "epoch": 2221, "lr": 1.718540687899874e-05}
{"train_loss": 0.06861919909715652, "global_step": 197718, "epoch": 2221, "lr": 1.7184969465810625e-05}
{"train_loss": 0.039600200951099396, "global_step": 197719, "epoch": 2221, "lr": 1.71845320570341e-05}
{"train_loss": 0.08890866488218307, "global_step": 197720, "epoch": 2221, "lr": 1.7184094652669224e-05}
{"train_loss": 0.10925634950399399, "global_step": 197721, "epoch": 2221, "lr": 1.718365725271603e-05}
{"train_loss": 0.03594669699668884, "global_step": 197722, "epoch": 2221, "lr": 1.718321985717461e-05}
{"train_loss": 0.060999516397714615, "global_step": 197723, "epoch": 2221, "lr": 1.718278246604499e-05}
{"train_loss": 0.0393889844417572, "global_step": 197724, "epoch": 2221, "lr": 1.7182345079327254e-05}
{"train_loss": 0.06946903467178345, "global_step": 197725, "epoch": 2221, "lr": 1.718190769702146e-05}
{"train_loss": 0.07088784873485565, "global_step": 197726, "epoch": 2221, "lr": 1.7181470319127634e-05}
{"train_loss": 0.051289018243551254, "global_step": 197727, "epoch": 2221, "lr": 1.718103294564588e-05}
{"train_loss": 0.0901440903544426, "global_step": 197728, "epoch": 2221, "lr": 1.718059557657622e-05}
{"train_loss": 0.12720385193824768, "global_step": 197729, "epoch": 2221, "lr": 1.7180158211918747e-05}
{"train_loss": 0.05388389900326729, "global_step": 197730, "epoch": 2221, "lr": 1.717972085167348e-05}
{"train_loss": 0.10520105063915253, "global_step": 197731, "epoch": 2221, "lr": 1.717928349584052e-05}
{"train_loss": 0.03842921927571297, "global_step": 197732, "epoch": 2221, "lr": 1.717884614441989e-05}
{"train_loss": 0.042814914137125015, "global_step": 197733, "epoch": 2221, "lr": 1.7178408797411684e-05}
{"train_loss": 0.11549795418977737, "global_step": 197734, "epoch": 2221, "lr": 1.7177971454815918e-05}
{"train_loss": 0.10888762027025223, "global_step": 197735, "epoch": 2221, "lr": 1.7177534116632672e-05}
{"train_loss": 0.05656379461288452, "global_step": 197736, "epoch": 2221, "lr": 1.7177096782862024e-05}
{"train_loss": 0.09197887033224106, "global_step": 197737, "epoch": 2221, "lr": 1.7176659453504e-05}
{"train_loss": 0.05693364515900612, "global_step": 197738, "epoch": 2221, "lr": 1.7176222128558693e-05}
{"train_loss": 0.05192749574780464, "global_step": 197739, "epoch": 2221, "lr": 1.7175784808026123e-05}
{"train_loss": 0.08225160837173462, "global_step": 197740, "epoch": 2221, "lr": 1.717534749190639e-05}
{"train_loss": 0.07006038725376129, "global_step": 197741, "epoch": 2221, "lr": 1.717491018019951e-05}
{"train_loss": 0.08536908030509949, "global_step": 197742, "epoch": 2221, "lr": 1.7174472872905585e-05}
{"train_loss": 0.03884711489081383, "global_step": 197743, "epoch": 2221, "lr": 1.7174035570024626e-05}
{"train_loss": 0.09225371479988098, "global_step": 197744, "epoch": 2221, "lr": 1.7173598271556746e-05}
{"train_loss": 0.042522720992565155, "global_step": 197745, "epoch": 2221, "lr": 1.717316097750195e-05}
{"train_loss": 0.08153971284627914, "global_step": 197746, "epoch": 2221, "lr": 1.7172723687860344e-05}
{"train_loss": 0.05636182054877281, "global_step": 197747, "epoch": 2221, "lr": 1.7172286402631948e-05}
{"train_loss": 0.05275000259280205, "global_step": 197748, "epoch": 2221, "lr": 1.717184912181685e-05}
{"train_loss": 0.0659070536494255, "global_step": 197749, "epoch": 2221, "lr": 1.7171411845415087e-05}
{"train_loss": 0.06548603624105453, "global_step": 197750, "epoch": 2221, "lr": 1.717097457342674e-05}
{"train_loss": 0.07451171427965164, "global_step": 197751, "epoch": 2221, "lr": 1.7170537305851843e-05}
{"train_loss": 0.06666214764118195, "global_step": 197752, "epoch": 2221, "lr": 1.717010004269046e-05}
{"train_loss": 0.05510479211807251, "global_step": 197753, "epoch": 2221, "lr": 1.716966278394268e-05}
{"train_loss": 0.06568475067615509, "global_step": 197754, "epoch": 2221, "lr": 1.716922552960852e-05}
{"train_loss": 0.08224483579397202, "global_step": 197755, "epoch": 2221, "lr": 1.7168788279688068e-05}
{"train_loss": 0.05055732652544975, "global_step": 197756, "epoch": 2221, "lr": 1.7168351034181363e-05}
{"train_loss": 0.06617657208208287, "global_step": 197757, "epoch": 2221, "lr": 1.7167913793088493e-05, "val_loss": 8.42567253112793}
{"train_loss": 0.028957609087228775, "global_step": 197758, "epoch": 2222, "lr": 1.716747655640949e-05}
{"train_loss": 0.037194203585386276, "global_step": 197759, "epoch": 2222, "lr": 1.71670393241444e-05}
{"train_loss": 0.058901842683553696, "global_step": 197760, "epoch": 2222, "lr": 1.7166602096293326e-05}
{"train_loss": 0.09651299566030502, "global_step": 197761, "epoch": 2222, "lr": 1.716616487285628e-05}
{"train_loss": 0.040553875267505646, "global_step": 197762, "epoch": 2222, "lr": 1.7165727653833363e-05}
{"train_loss": 0.08624870330095291, "global_step": 197763, "epoch": 2222, "lr": 1.7165290439224614e-05}
{"train_loss": 0.057903092354536057, "global_step": 197764, "epoch": 2222, "lr": 1.716485322903007e-05}
{"train_loss": 0.0851447656750679, "global_step": 197765, "epoch": 2222, "lr": 1.7164416023249836e-05}
{"train_loss": 0.041236720979213715, "global_step": 197766, "epoch": 2222, "lr": 1.716397882188392e-05}
{"train_loss": 0.0634322240948677, "global_step": 197767, "epoch": 2222, "lr": 1.7163541624932426e-05}
{"train_loss": 0.05690643936395645, "global_step": 197768, "epoch": 2222, "lr": 1.7163104432395384e-05}
{"train_loss": 0.0635291263461113, "global_step": 197769, "epoch": 2222, "lr": 1.7162667244272874e-05}
{"train_loss": 0.036378879100084305, "global_step": 197770, "epoch": 2222, "lr": 1.7162230060564928e-05}
{"train_loss": 0.04376651719212532, "global_step": 197771, "epoch": 2222, "lr": 1.7161792881271615e-05}
{"train_loss": 0.0849471166729927, "global_step": 197772, "epoch": 2222, "lr": 1.7161355706393018e-05}
{"train_loss": 0.03722655028104782, "global_step": 197773, "epoch": 2222, "lr": 1.716091853592916e-05}
{"train_loss": 0.04633237048983574, "global_step": 197774, "epoch": 2222, "lr": 1.716048136988014e-05}
{"train_loss": 0.06814874708652496, "global_step": 197775, "epoch": 2222, "lr": 1.716004420824597e-05}
{"train_loss": 0.07823391258716583, "global_step": 197776, "epoch": 2222, "lr": 1.7159607051026745e-05}
{"train_loss": 0.052653830498456955, "global_step": 197777, "epoch": 2222, "lr": 1.7159169898222498e-05}
{"train_loss": 0.0750517025589943, "global_step": 197778, "epoch": 2222, "lr": 1.7158732749833318e-05}
{"train_loss": 0.0468783937394619, "global_step": 197779, "epoch": 2222, "lr": 1.7158295605859226e-05}
{"train_loss": 0.05520729720592499, "global_step": 197780, "epoch": 2222, "lr": 1.715785846630032e-05}
{"train_loss": 0.09351321309804916, "global_step": 197781, "epoch": 2222, "lr": 1.715742133115662e-05}
{"train_loss": 0.048078279942274094, "global_step": 197782, "epoch": 2222, "lr": 1.715698420042822e-05}
{"train_loss": 0.034756407141685486, "global_step": 197783, "epoch": 2222, "lr": 1.7156547074115152e-05}
{"train_loss": 0.08304011076688766, "global_step": 197784, "epoch": 2222, "lr": 1.7156109952217503e-05}
{"train_loss": 0.05287322774529457, "global_step": 197785, "epoch": 2222, "lr": 1.7155672834735292e-05}
{"train_loss": 0.04003283753991127, "global_step": 197786, "epoch": 2222, "lr": 1.7155235721668622e-05}
{"train_loss": 0.08627892285585403, "global_step": 197787, "epoch": 2222, "lr": 1.7154798613017515e-05}
{"train_loss": 0.05526227504014969, "global_step": 197788, "epoch": 2222, "lr": 1.7154361508782042e-05}
{"train_loss": 0.0792296975851059, "global_step": 197789, "epoch": 2222, "lr": 1.715392440896228e-05}
{"train_loss": 0.05042283982038498, "global_step": 197790, "epoch": 2222, "lr": 1.715348731355826e-05}
{"train_loss": 0.07920096814632416, "global_step": 197791, "epoch": 2222, "lr": 1.7153050222570067e-05}
{"train_loss": 0.055430181324481964, "global_step": 197792, "epoch": 2222, "lr": 1.715261313599773e-05}
{"train_loss": 0.034828897565603256, "global_step": 197793, "epoch": 2222, "lr": 1.715217605384134e-05}
{"train_loss": 0.08484350144863129, "global_step": 197794, "epoch": 2222, "lr": 1.7151738976100922e-05}
{"train_loss": 0.06990742683410645, "global_step": 197795, "epoch": 2222, "lr": 1.7151301902776574e-05}
{"train_loss": 0.0858164131641388, "global_step": 197796, "epoch": 2222, "lr": 1.7150864833868324e-05}
{"train_loss": 0.11089392751455307, "global_step": 197797, "epoch": 2222, "lr": 1.7150427769376233e-05}
{"train_loss": 0.10497283190488815, "global_step": 197798, "epoch": 2222, "lr": 1.714999070930038e-05}
{"train_loss": 0.07076428085565567, "global_step": 197799, "epoch": 2222, "lr": 1.7149553653640788e-05}
{"train_loss": 0.09567245095968246, "global_step": 197800, "epoch": 2222, "lr": 1.714911660239756e-05}
{"train_loss": 0.10182185471057892, "global_step": 197801, "epoch": 2222, "lr": 1.714867955557072e-05}
{"train_loss": 0.036034051328897476, "global_step": 197802, "epoch": 2222, "lr": 1.7148242513160347e-05}
{"train_loss": 0.035531990230083466, "global_step": 197803, "epoch": 2222, "lr": 1.7147805475166495e-05}
{"train_loss": 0.08464621007442474, "global_step": 197804, "epoch": 2222, "lr": 1.7147368441589206e-05}
{"train_loss": 0.019036713987588882, "global_step": 197805, "epoch": 2222, "lr": 1.714693141242857e-05}
{"train_loss": 0.08344930410385132, "global_step": 197806, "epoch": 2222, "lr": 1.7146494387684615e-05}
{"train_loss": 0.10078155249357224, "global_step": 197807, "epoch": 2222, "lr": 1.7146057367357404e-05}
{"train_loss": 0.054265253245830536, "global_step": 197808, "epoch": 2222, "lr": 1.714562035144703e-05}
{"train_loss": 0.09483516961336136, "global_step": 197809, "epoch": 2222, "lr": 1.714518333995351e-05}
{"train_loss": 0.06373812258243561, "global_step": 197810, "epoch": 2222, "lr": 1.7144746332876933e-05}
{"train_loss": 0.06802552938461304, "global_step": 197811, "epoch": 2222, "lr": 1.714430933021733e-05}
{"train_loss": 0.0370747372508049, "global_step": 197812, "epoch": 2222, "lr": 1.7143872331974788e-05}
{"train_loss": 0.057899389415979385, "global_step": 197813, "epoch": 2222, "lr": 1.7143435338149337e-05}
{"train_loss": 0.09696164727210999, "global_step": 197814, "epoch": 2222, "lr": 1.714299834874107e-05}
{"train_loss": 0.045112043619155884, "global_step": 197815, "epoch": 2222, "lr": 1.7142561363750002e-05}
{"train_loss": 0.07043994963169098, "global_step": 197816, "epoch": 2222, "lr": 1.714212438317624e-05}
{"train_loss": 0.0534018836915493, "global_step": 197817, "epoch": 2222, "lr": 1.7141687407019795e-05}
{"train_loss": 0.08802724629640579, "global_step": 197818, "epoch": 2222, "lr": 1.7141250435280777e-05}
{"train_loss": 0.030860884115099907, "global_step": 197819, "epoch": 2222, "lr": 1.7140813467959194e-05}
{"train_loss": 0.027201035991311073, "global_step": 197820, "epoch": 2222, "lr": 1.7140376505055145e-05}
{"train_loss": 0.02612314745783806, "global_step": 197821, "epoch": 2222, "lr": 1.7139939546568655e-05}
{"train_loss": 0.051662445068359375, "global_step": 197822, "epoch": 2222, "lr": 1.7139502592499817e-05}
{"train_loss": 0.058322176337242126, "global_step": 197823, "epoch": 2222, "lr": 1.7139065642848656e-05}
{"train_loss": 0.055966053158044815, "global_step": 197824, "epoch": 2222, "lr": 1.713862869761526e-05}
{"train_loss": 0.089450903236866, "global_step": 197825, "epoch": 2222, "lr": 1.7138191756799666e-05}
{"train_loss": 0.12129884213209152, "global_step": 197826, "epoch": 2222, "lr": 1.7137754820401935e-05}
{"train_loss": 0.07212767004966736, "global_step": 197827, "epoch": 2222, "lr": 1.7137317888422154e-05}
{"train_loss": 0.10628737509250641, "global_step": 197828, "epoch": 2222, "lr": 1.7136880960860342e-05}
{"train_loss": 0.04028814285993576, "global_step": 197829, "epoch": 2222, "lr": 1.7136444037716586e-05}
{"train_loss": 0.050055962055921555, "global_step": 197830, "epoch": 2222, "lr": 1.7136007118990927e-05}
{"train_loss": 0.08387110382318497, "global_step": 197831, "epoch": 2222, "lr": 1.7135570204683442e-05}
{"train_loss": 0.0843152403831482, "global_step": 197832, "epoch": 2222, "lr": 1.7135133294794166e-05}
{"train_loss": 0.0806998535990715, "global_step": 197833, "epoch": 2222, "lr": 1.7134696389323186e-05}
{"train_loss": 0.06048678979277611, "global_step": 197834, "epoch": 2222, "lr": 1.7134259488270527e-05}
{"train_loss": 0.07581036537885666, "global_step": 197835, "epoch": 2222, "lr": 1.713382259163629e-05}
{"train_loss": 0.0680176243185997, "global_step": 197836, "epoch": 2222, "lr": 1.7133385699420502e-05}
{"train_loss": 0.02325626090168953, "global_step": 197837, "epoch": 2222, "lr": 1.7132948811623217e-05}
{"train_loss": 0.035403888672590256, "global_step": 197838, "epoch": 2222, "lr": 1.713251192824452e-05}
{"train_loss": 0.11936096847057343, "global_step": 197839, "epoch": 2222, "lr": 1.713207504928444e-05}
{"train_loss": 0.06915880739688873, "global_step": 197840, "epoch": 2222, "lr": 1.713163817474307e-05}
{"train_loss": 0.12351197749376297, "global_step": 197841, "epoch": 2222, "lr": 1.7131201304620453e-05}
{"train_loss": 0.06358443200588226, "global_step": 197842, "epoch": 2222, "lr": 1.7130764438916624e-05}
{"train_loss": 0.10526274144649506, "global_step": 197843, "epoch": 2222, "lr": 1.7130327577631667e-05}
{"train_loss": 0.11153754591941833, "global_step": 197844, "epoch": 2222, "lr": 1.712989072076565e-05}
{"train_loss": 0.026005232706665993, "global_step": 197845, "epoch": 2222, "lr": 1.7129453868318608e-05}
{"train_loss": 0.06579952672375051, "global_step": 197846, "epoch": 2222, "lr": 1.7129017020290618e-05, "val_loss": 8.167169570922852}
{"train_loss": 0.04479122534394264, "global_step": 197847, "epoch": 2223, "lr": 1.7128580176681717e-05}
{"train_loss": 0.0830322653055191, "global_step": 197848, "epoch": 2223, "lr": 1.7128143337491998e-05}
{"train_loss": 0.05035687983036041, "global_step": 197849, "epoch": 2223, "lr": 1.712770650272148e-05}
{"train_loss": 0.0938870906829834, "global_step": 197850, "epoch": 2223, "lr": 1.712726967237026e-05}
{"train_loss": 0.05397966876626015, "global_step": 197851, "epoch": 2223, "lr": 1.712683284643836e-05}
{"train_loss": 0.018238402903079987, "global_step": 197852, "epoch": 2223, "lr": 1.7126396024925877e-05}
{"train_loss": 0.0788913294672966, "global_step": 197853, "epoch": 2223, "lr": 1.7125959207832825e-05}
{"train_loss": 0.03941462188959122, "global_step": 197854, "epoch": 2223, "lr": 1.712552239515931e-05}
{"train_loss": 0.08059189468622208, "global_step": 197855, "epoch": 2223, "lr": 1.7125085586905347e-05}
{"train_loss": 0.054338328540325165, "global_step": 197856, "epoch": 2223, "lr": 1.712464878307104e-05}
{"train_loss": 0.07081800699234009, "global_step": 197857, "epoch": 2223, "lr": 1.71242119836564e-05}
{"train_loss": 0.0797823891043663, "global_step": 197858, "epoch": 2223, "lr": 1.7123775188661527e-05}
{"train_loss": 0.04358268529176712, "global_step": 197859, "epoch": 2223, "lr": 1.712333839808644e-05}
{"train_loss": 0.04188962280750275, "global_step": 197860, "epoch": 2223, "lr": 1.7122901611931246e-05}
{"train_loss": 0.07080690562725067, "global_step": 197861, "epoch": 2223, "lr": 1.7122464830195954e-05}
{"train_loss": 0.03935566172003746, "global_step": 197862, "epoch": 2223, "lr": 1.7122028052880646e-05}
{"train_loss": 0.10090067982673645, "global_step": 197863, "epoch": 2223, "lr": 1.71215912799854e-05}
{"train_loss": 0.04154155030846596, "global_step": 197864, "epoch": 2223, "lr": 1.712115451151024e-05}
{"train_loss": 0.09573127329349518, "global_step": 197865, "epoch": 2223, "lr": 1.712071774745525e-05}
{"train_loss": 0.023782702162861824, "global_step": 197866, "epoch": 2223, "lr": 1.712028098782047e-05}
{"train_loss": 0.0262803602963686, "global_step": 197867, "epoch": 2223, "lr": 1.711984423260598e-05}
{"train_loss": 0.08078055828809738, "global_step": 197868, "epoch": 2223, "lr": 1.7119407481811805e-05}
{"train_loss": 0.07696665823459625, "global_step": 197869, "epoch": 2223, "lr": 1.7118970735438052e-05}
{"train_loss": 0.045564696192741394, "global_step": 197870, "epoch": 2223, "lr": 1.7118533993484726e-05}
{"train_loss": 0.03506492078304291, "global_step": 197871, "epoch": 2223, "lr": 1.7118097255951937e-05}
{"train_loss": 0.09030606597661972, "global_step": 197872, "epoch": 2223, "lr": 1.7117660522839697e-05}
{"train_loss": 0.06926050037145615, "global_step": 197873, "epoch": 2223, "lr": 1.711722379414811e-05}
{"train_loss": 0.0619085356593132, "global_step": 197874, "epoch": 2223, "lr": 1.71167870698772e-05}
{"train_loss": 0.042494889348745346, "global_step": 197875, "epoch": 2223, "lr": 1.711635035002703e-05}
{"train_loss": 0.071773461997509, "global_step": 197876, "epoch": 2223, "lr": 1.711591363459768e-05}
{"train_loss": 0.04998979717493057, "global_step": 197877, "epoch": 2223, "lr": 1.7115476923589174e-05}
{"train_loss": 0.045927781611680984, "global_step": 197878, "epoch": 2223, "lr": 1.7115040217001615e-05}
{"train_loss": 0.04787984490394592, "global_step": 197879, "epoch": 2223, "lr": 1.7114603514835016e-05}
{"train_loss": 0.043954163789749146, "global_step": 197880, "epoch": 2223, "lr": 1.711416681708948e-05}
{"train_loss": 0.0868387296795845, "global_step": 197881, "epoch": 2223, "lr": 1.711373012376502e-05}
{"train_loss": 0.05563313513994217, "global_step": 197882, "epoch": 2223, "lr": 1.7113293434861736e-05}
{"train_loss": 0.0550810880959034, "global_step": 197883, "epoch": 2223, "lr": 1.7112856750379657e-05}
{"train_loss": 0.042206380516290665, "global_step": 197884, "epoch": 2223, "lr": 1.711242007031887e-05}
{"train_loss": 0.06479031592607498, "global_step": 197885, "epoch": 2223, "lr": 1.7111983394679394e-05}
{"train_loss": 0.05092008039355278, "global_step": 197886, "epoch": 2223, "lr": 1.711154672346133e-05}
{"train_loss": 0.06528762727975845, "global_step": 197887, "epoch": 2223, "lr": 1.7111110056664702e-05}
{"train_loss": 0.0829731747508049, "global_step": 197888, "epoch": 2223, "lr": 1.7110673394289607e-05}
{"train_loss": 0.049141258001327515, "global_step": 197889, "epoch": 2223, "lr": 1.7110236736336054e-05}
{"train_loss": 0.06799335777759552, "global_step": 197890, "epoch": 2223, "lr": 1.7109800082804152e-05}
{"train_loss": 0.04024394229054451, "global_step": 197891, "epoch": 2223, "lr": 1.710936343369392e-05}
{"train_loss": 0.07120752334594727, "global_step": 197892, "epoch": 2223, "lr": 1.7108926789005447e-05}
{"train_loss": 0.06762664020061493, "global_step": 197893, "epoch": 2223, "lr": 1.7108490148738765e-05}
{"train_loss": 0.13066500425338745, "global_step": 197894, "epoch": 2223, "lr": 1.7108053512893958e-05}
{"train_loss": 0.04215158522129059, "global_step": 197895, "epoch": 2223, "lr": 1.7107616881471056e-05}
{"train_loss": 0.03329402208328247, "global_step": 197896, "epoch": 2223, "lr": 1.7107180254470158e-05}
{"train_loss": 0.030130954459309578, "global_step": 197897, "epoch": 2223, "lr": 1.710674363189127e-05}
{"train_loss": 0.06192493438720703, "global_step": 197898, "epoch": 2223, "lr": 1.7106307013734492e-05}
{"train_loss": 0.13134369254112244, "global_step": 197899, "epoch": 2223, "lr": 1.7105870399999874e-05}
{"train_loss": 0.0660601407289505, "global_step": 197900, "epoch": 2223, "lr": 1.7105433790687463e-05}
{"train_loss": 0.06687861680984497, "global_step": 197901, "epoch": 2223, "lr": 1.710499718579734e-05}
{"train_loss": 0.06633946299552917, "global_step": 197902, "epoch": 2223, "lr": 1.710456058532953e-05}
{"train_loss": 0.04540444537997246, "global_step": 197903, "epoch": 2223, "lr": 1.7104123989284132e-05}
{"train_loss": 0.06133156269788742, "global_step": 197904, "epoch": 2223, "lr": 1.7103687397661167e-05}
{"train_loss": 0.05258411914110184, "global_step": 197905, "epoch": 2223, "lr": 1.710325081046072e-05}
{"train_loss": 0.03985133767127991, "global_step": 197906, "epoch": 2223, "lr": 1.710281422768283e-05}
{"train_loss": 0.069043830037117, "global_step": 197907, "epoch": 2223, "lr": 1.7102377649327578e-05}
{"train_loss": 0.027929121628403664, "global_step": 197908, "epoch": 2223, "lr": 1.7101941075395e-05}
{"train_loss": 0.11847411096096039, "global_step": 197909, "epoch": 2223, "lr": 1.7101504505885173e-05}
{"train_loss": 0.06941358745098114, "global_step": 197910, "epoch": 2223, "lr": 1.7101067940798137e-05}
{"train_loss": 0.04334498196840286, "global_step": 197911, "epoch": 2223, "lr": 1.710063138013398e-05}
{"train_loss": 0.060345835983753204, "global_step": 197912, "epoch": 2223, "lr": 1.710019482389274e-05}
{"train_loss": 0.05495818704366684, "global_step": 197913, "epoch": 2223, "lr": 1.7099758272074457e-05}
{"train_loss": 0.1075412780046463, "global_step": 197914, "epoch": 2223, "lr": 1.709932172467923e-05}
{"train_loss": 0.08604296296834946, "global_step": 197915, "epoch": 2223, "lr": 1.7098885181707082e-05}
{"train_loss": 0.0575689971446991, "global_step": 197916, "epoch": 2223, "lr": 1.7098448643158104e-05}
{"train_loss": 0.09747661650180817, "global_step": 197917, "epoch": 2223, "lr": 1.7098012109032325e-05}
{"train_loss": 0.09598732739686966, "global_step": 197918, "epoch": 2223, "lr": 1.7097575579329826e-05}
{"train_loss": 0.09493262320756912, "global_step": 197919, "epoch": 2223, "lr": 1.709713905405065e-05}
{"train_loss": 0.09606198221445084, "global_step": 197920, "epoch": 2223, "lr": 1.7096702533194874e-05}
{"train_loss": 0.054953597486019135, "global_step": 197921, "epoch": 2223, "lr": 1.709626601676253e-05}
{"train_loss": 0.04062722250819206, "global_step": 197922, "epoch": 2223, "lr": 1.709582950475371e-05}
{"train_loss": 0.07112672179937363, "global_step": 197923, "epoch": 2223, "lr": 1.7095392997168436e-05}
{"train_loss": 0.08900865912437439, "global_step": 197924, "epoch": 2223, "lr": 1.7094956494006797e-05}
{"train_loss": 0.11010057479143143, "global_step": 197925, "epoch": 2223, "lr": 1.709451999526883e-05}
{"train_loss": 0.07761852443218231, "global_step": 197926, "epoch": 2223, "lr": 1.709408350095462e-05}
{"train_loss": 0.054161880165338516, "global_step": 197927, "epoch": 2223, "lr": 1.709364701106419e-05}
{"train_loss": 0.0967312902212143, "global_step": 197928, "epoch": 2223, "lr": 1.7093210525597637e-05}
{"train_loss": 0.07801403105258942, "global_step": 197929, "epoch": 2223, "lr": 1.7092774044554982e-05}
{"train_loss": 0.035504311323165894, "global_step": 197930, "epoch": 2223, "lr": 1.7092337567936323e-05}
{"train_loss": 0.038843754678964615, "global_step": 197931, "epoch": 2223, "lr": 1.709190109574168e-05}
{"train_loss": 0.04786055162549019, "global_step": 197932, "epoch": 2223, "lr": 1.7091464627971144e-05}
{"train_loss": 0.06289685517549515, "global_step": 197933, "epoch": 2223, "lr": 1.7091028164624746e-05}
{"train_loss": 0.06560970097780228, "global_step": 197934, "epoch": 2223, "lr": 1.7090591705702556e-05}
{"train_loss": 0.06383275537845794, "global_step": 197935, "epoch": 2223, "lr": 1.7090155251204654e-05, "val_loss": 8.2507963180542}
{"train_loss": 0.06664992868900299, "global_step": 197936, "epoch": 2224, "lr": 1.708971880113106e-05}
{"train_loss": 0.0505882166326046, "global_step": 197937, "epoch": 2224, "lr": 1.7089282355481866e-05}
{"train_loss": 0.06693574786186218, "global_step": 197938, "epoch": 2224, "lr": 1.7088845914257105e-05}
{"train_loss": 0.0729365348815918, "global_step": 197939, "epoch": 2224, "lr": 1.7088409477456862e-05}
{"train_loss": 0.06074640899896622, "global_step": 197940, "epoch": 2224, "lr": 1.708797304508117e-05}
{"train_loss": 0.1188381239771843, "global_step": 197941, "epoch": 2224, "lr": 1.7087536617130106e-05}
{"train_loss": 0.08854813128709793, "global_step": 197942, "epoch": 2224, "lr": 1.708710019360371e-05}
{"train_loss": 0.06270428746938705, "global_step": 197943, "epoch": 2224, "lr": 1.708666377450207e-05}
{"train_loss": 0.049618784338235855, "global_step": 197944, "epoch": 2224, "lr": 1.7086227359825213e-05}
{"train_loss": 0.06576748937368393, "global_step": 197945, "epoch": 2224, "lr": 1.7085790949573222e-05}
{"train_loss": 0.10999781638383865, "global_step": 197946, "epoch": 2224, "lr": 1.708535454374613e-05}
{"train_loss": 0.04488988220691681, "global_step": 197947, "epoch": 2224, "lr": 1.708491814234403e-05}
{"train_loss": 0.04229363054037094, "global_step": 197948, "epoch": 2224, "lr": 1.7084481745366944e-05}
{"train_loss": 0.05512582138180733, "global_step": 197949, "epoch": 2224, "lr": 1.7084045352814963e-05}
{"train_loss": 0.058668240904808044, "global_step": 197950, "epoch": 2224, "lr": 1.708360896468812e-05}
{"train_loss": 0.05112164840102196, "global_step": 197951, "epoch": 2224, "lr": 1.7083172580986495e-05}
{"train_loss": 0.044517695903778076, "global_step": 197952, "epoch": 2224, "lr": 1.7082736201710136e-05}
{"train_loss": 0.07062721252441406, "global_step": 197953, "epoch": 2224, "lr": 1.7082299826859087e-05}
{"train_loss": 0.06120551750063896, "global_step": 197954, "epoch": 2224, "lr": 1.708186345643344e-05}
{"train_loss": 0.08157858997583389, "global_step": 197955, "epoch": 2224, "lr": 1.7081427090433218e-05}
{"train_loss": 0.10472662001848221, "global_step": 197956, "epoch": 2224, "lr": 1.708099072885851e-05}
{"train_loss": 0.11721698194742203, "global_step": 197957, "epoch": 2224, "lr": 1.708055437170935e-05}
{"train_loss": 0.0827714279294014, "global_step": 197958, "epoch": 2224, "lr": 1.7080118018985825e-05}
{"train_loss": 0.05276575684547424, "global_step": 197959, "epoch": 2224, "lr": 1.707968167068795e-05}
{"train_loss": 0.050217218697071075, "global_step": 197960, "epoch": 2224, "lr": 1.7079245326815842e-05}
{"train_loss": 0.061551932245492935, "global_step": 197961, "epoch": 2224, "lr": 1.7078808987369503e-05}
{"train_loss": 0.06986567378044128, "global_step": 197962, "epoch": 2224, "lr": 1.7078372652349034e-05}
{"train_loss": 0.14574670791625977, "global_step": 197963, "epoch": 2224, "lr": 1.707793632175446e-05}
{"train_loss": 0.08826040476560593, "global_step": 197964, "epoch": 2224, "lr": 1.7077499995585878e-05}
{"train_loss": 0.06615187972784042, "global_step": 197965, "epoch": 2224, "lr": 1.7077063673843304e-05}
{"train_loss": 0.04715525358915329, "global_step": 197966, "epoch": 2224, "lr": 1.7076627356526832e-05}
{"train_loss": 0.08170589804649353, "global_step": 197967, "epoch": 2224, "lr": 1.707619104363649e-05}
{"train_loss": 0.04948551207780838, "global_step": 197968, "epoch": 2224, "lr": 1.707575473517237e-05}
{"train_loss": 0.08845911175012589, "global_step": 197969, "epoch": 2224, "lr": 1.7075318431134503e-05}
{"train_loss": 0.05910283327102661, "global_step": 197970, "epoch": 2224, "lr": 1.707488213152295e-05}
{"train_loss": 0.06829271465539932, "global_step": 197971, "epoch": 2224, "lr": 1.7074445836337798e-05}
{"train_loss": 0.05201878771185875, "global_step": 197972, "epoch": 2224, "lr": 1.707400954557907e-05}
{"train_loss": 0.04094407334923744, "global_step": 197973, "epoch": 2224, "lr": 1.7073573259246856e-05}
{"train_loss": 0.04942702874541283, "global_step": 197974, "epoch": 2224, "lr": 1.707313697734118e-05}
{"train_loss": 0.11777752637863159, "global_step": 197975, "epoch": 2224, "lr": 1.7072700699862138e-05}
{"train_loss": 0.06844450533390045, "global_step": 197976, "epoch": 2224, "lr": 1.7072264426809752e-05}
{"train_loss": 0.0996256172657013, "global_step": 197977, "epoch": 2224, "lr": 1.7071828158184116e-05}
{"train_loss": 0.04934178665280342, "global_step": 197978, "epoch": 2224, "lr": 1.7071391893985257e-05}
{"train_loss": 0.04222114756703377, "global_step": 197979, "epoch": 2224, "lr": 1.707095563421326e-05}
{"train_loss": 0.07323409616947174, "global_step": 197980, "epoch": 2224, "lr": 1.707051937886816e-05}
{"train_loss": 0.029420597478747368, "global_step": 197981, "epoch": 2224, "lr": 1.7070083127950037e-05}
{"train_loss": 0.04616202786564827, "global_step": 197982, "epoch": 2224, "lr": 1.7069646881458927e-05}
{"train_loss": 0.08276808261871338, "global_step": 197983, "epoch": 2224, "lr": 1.7069210639394922e-05}
{"train_loss": 0.09194721281528473, "global_step": 197984, "epoch": 2224, "lr": 1.706877440175804e-05}
{"train_loss": 0.07191646099090576, "global_step": 197985, "epoch": 2224, "lr": 1.7068338168548377e-05}
{"train_loss": 0.06415042281150818, "global_step": 197986, "epoch": 2224, "lr": 1.706790193976596e-05}
{"train_loss": 0.06407994776964188, "global_step": 197987, "epoch": 2224, "lr": 1.7067465715410874e-05}
{"train_loss": 0.05365198105573654, "global_step": 197988, "epoch": 2224, "lr": 1.7067029495483155e-05}
{"train_loss": 0.06285832077264786, "global_step": 197989, "epoch": 2224, "lr": 1.7066593279982883e-05}
{"train_loss": 0.06869731098413467, "global_step": 197990, "epoch": 2224, "lr": 1.706615706891011e-05}
{"train_loss": 0.08928076922893524, "global_step": 197991, "epoch": 2224, "lr": 1.706572086226488e-05}
{"train_loss": 0.08798205107450485, "global_step": 197992, "epoch": 2224, "lr": 1.706528466004727e-05}
{"train_loss": 0.06144409254193306, "global_step": 197993, "epoch": 2224, "lr": 1.7064848462257315e-05}
{"train_loss": 0.07710161060094833, "global_step": 197994, "epoch": 2224, "lr": 1.7064412268895107e-05}
{"train_loss": 0.04305187985301018, "global_step": 197995, "epoch": 2224, "lr": 1.7063976079960675e-05}
{"train_loss": 0.0411515086889267, "global_step": 197996, "epoch": 2224, "lr": 1.7063539895454104e-05}
{"train_loss": 0.03984685242176056, "global_step": 197997, "epoch": 2224, "lr": 1.7063103715375422e-05}
{"train_loss": 0.047475188970565796, "global_step": 197998, "epoch": 2224, "lr": 1.706266753972472e-05}
{"train_loss": 0.034143611788749695, "global_step": 197999, "epoch": 2224, "lr": 1.7062231368502025e-05}
{"train_loss": 0.09853333234786987, "global_step": 198000, "epoch": 2224, "lr": 1.7061795201707425e-05}
{"train_loss": 0.02708045020699501, "global_step": 198001, "epoch": 2224, "lr": 1.706135903934095e-05}
{"train_loss": 0.06586676836013794, "global_step": 198002, "epoch": 2224, "lr": 1.7060922881402696e-05}
{"train_loss": 0.038071416318416595, "global_step": 198003, "epoch": 2224, "lr": 1.7060486727892676e-05}
{"train_loss": 0.06811277568340302, "global_step": 198004, "epoch": 2224, "lr": 1.7060050578810992e-05}
{"train_loss": 0.0381542444229126, "global_step": 198005, "epoch": 2224, "lr": 1.7059614434157667e-05}
{"train_loss": 0.05219279229640961, "global_step": 198006, "epoch": 2224, "lr": 1.7059178293932776e-05}
{"train_loss": 0.12244639545679092, "global_step": 198007, "epoch": 2224, "lr": 1.7058742158136394e-05}
{"train_loss": 0.0431784987449646, "global_step": 198008, "epoch": 2224, "lr": 1.705830602676854e-05}
{"train_loss": 0.06961765140295029, "global_step": 198009, "epoch": 2224, "lr": 1.7057869899829314e-05}
{"train_loss": 0.08589979261159897, "global_step": 198010, "epoch": 2224, "lr": 1.7057433777318745e-05}
{"train_loss": 0.08744167536497116, "global_step": 198011, "epoch": 2224, "lr": 1.705699765923691e-05}
{"train_loss": 0.0675644800066948, "global_step": 198012, "epoch": 2224, "lr": 1.705656154558385e-05}
{"train_loss": 0.058353323489427567, "global_step": 198013, "epoch": 2224, "lr": 1.705612543635965e-05}
{"train_loss": 0.042371973395347595, "global_step": 198014, "epoch": 2224, "lr": 1.7055689331564335e-05}
{"train_loss": 0.04354940727353096, "global_step": 198015, "epoch": 2224, "lr": 1.7055253231198e-05}
{"train_loss": 0.0474301353096962, "global_step": 198016, "epoch": 2224, "lr": 1.7054817135260663e-05}
{"train_loss": 0.054033063352108, "global_step": 198017, "epoch": 2224, "lr": 1.705438104375242e-05}
{"train_loss": 0.07159464061260223, "global_step": 198018, "epoch": 2224, "lr": 1.7053944956673302e-05}
{"train_loss": 0.062170546501874924, "global_step": 198019, "epoch": 2224, "lr": 1.7053508874023395e-05}
{"train_loss": 0.056914810091257095, "global_step": 198020, "epoch": 2224, "lr": 1.7053072795802725e-05}
{"train_loss": 0.0573565736413002, "global_step": 198021, "epoch": 2224, "lr": 1.7052636722011384e-05}
{"train_loss": 0.08239886909723282, "global_step": 198022, "epoch": 2224, "lr": 1.7052200652649402e-05}
{"train_loss": 0.08138375729322433, "global_step": 198023, "epoch": 2224, "lr": 1.7051764587716863e-05}
{"train_loss": 0.06631116311620461, "global_step": 198024, "epoch": 2224, "lr": 1.7051328527213795e-05, "val_loss": 8.234292984008789}
{"train_loss": 0.032893456518650055, "global_step": 198025, "epoch": 2225, "lr": 1.705089247114029e-05}
{"train_loss": 0.12327690422534943, "global_step": 198026, "epoch": 2225, "lr": 1.705045641949638e-05}
{"train_loss": 0.10063178092241287, "global_step": 198027, "epoch": 2225, "lr": 1.7050020372282143e-05}
{"train_loss": 0.07143690437078476, "global_step": 198028, "epoch": 2225, "lr": 1.7049584329497636e-05}
{"train_loss": 0.06589720398187637, "global_step": 198029, "epoch": 2225, "lr": 1.704914829114289e-05}
{"train_loss": 0.09565229713916779, "global_step": 198030, "epoch": 2225, "lr": 1.7048712257218e-05}
{"train_loss": 0.05540924891829491, "global_step": 198031, "epoch": 2225, "lr": 1.7048276227722997e-05}
{"train_loss": 0.05722493305802345, "global_step": 198032, "epoch": 2225, "lr": 1.704784020265797e-05}
{"train_loss": 0.07677224278450012, "global_step": 198033, "epoch": 2225, "lr": 1.7047404182022935e-05}
{"train_loss": 0.09144557267427444, "global_step": 198034, "epoch": 2225, "lr": 1.7046968165818e-05}
{"train_loss": 0.03561465069651604, "global_step": 198035, "epoch": 2225, "lr": 1.7046532154043178e-05}
{"train_loss": 0.020551491528749466, "global_step": 198036, "epoch": 2225, "lr": 1.7046096146698566e-05}
{"train_loss": 0.057140473276376724, "global_step": 198037, "epoch": 2225, "lr": 1.704566014378418e-05}
{"train_loss": 0.08726255595684052, "global_step": 198038, "epoch": 2225, "lr": 1.7045224145300133e-05}
{"train_loss": 0.04763689637184143, "global_step": 198039, "epoch": 2225, "lr": 1.7044788151246427e-05}
{"train_loss": 0.07938967645168304, "global_step": 198040, "epoch": 2225, "lr": 1.7044352161623167e-05}
{"train_loss": 0.054938480257987976, "global_step": 198041, "epoch": 2225, "lr": 1.7043916176430375e-05}
{"train_loss": 0.0255889929831028, "global_step": 198042, "epoch": 2225, "lr": 1.7043480195668133e-05}
{"train_loss": 0.06880377978086472, "global_step": 198043, "epoch": 2225, "lr": 1.70430442193365e-05}
{"train_loss": 0.0998658835887909, "global_step": 198044, "epoch": 2225, "lr": 1.7042608247435522e-05}
{"train_loss": 0.0357072576880455, "global_step": 198045, "epoch": 2225, "lr": 1.7042172279965275e-05}
{"train_loss": 0.0701514482498169, "global_step": 198046, "epoch": 2225, "lr": 1.7041736316925788e-05}
{"train_loss": 0.059092193841934204, "global_step": 198047, "epoch": 2225, "lr": 1.704130035831715e-05}
{"train_loss": 0.036422740668058395, "global_step": 198048, "epoch": 2225, "lr": 1.7040864404139402e-05}
{"train_loss": 0.04240550100803375, "global_step": 198049, "epoch": 2225, "lr": 1.704042845439262e-05}
{"train_loss": 0.06266100704669952, "global_step": 198050, "epoch": 2225, "lr": 1.703999250907683e-05}
{"train_loss": 0.046761613339185715, "global_step": 198051, "epoch": 2225, "lr": 1.7039556568192132e-05}
{"train_loss": 0.04806946963071823, "global_step": 198052, "epoch": 2225, "lr": 1.7039120631738547e-05}
{"train_loss": 0.04644758626818657, "global_step": 198053, "epoch": 2225, "lr": 1.7038684699716167e-05}
{"train_loss": 0.07220368832349777, "global_step": 198054, "epoch": 2225, "lr": 1.703824877212502e-05}
{"train_loss": 0.06619390100240707, "global_step": 198055, "epoch": 2225, "lr": 1.703781284896519e-05}
{"train_loss": 0.08068385720252991, "global_step": 198056, "epoch": 2225, "lr": 1.703737693023671e-05}
{"train_loss": 0.08364784717559814, "global_step": 198057, "epoch": 2225, "lr": 1.703694101593967e-05}
{"train_loss": 0.0812644436955452, "global_step": 198058, "epoch": 2225, "lr": 1.7036505106074098e-05}
{"train_loss": 0.057200491428375244, "global_step": 198059, "epoch": 2225, "lr": 1.7036069200640075e-05}
{"train_loss": 0.07934439182281494, "global_step": 198060, "epoch": 2225, "lr": 1.7035633299637638e-05}
{"train_loss": 0.054887786507606506, "global_step": 198061, "epoch": 2225, "lr": 1.703519740306686e-05}
{"train_loss": 0.06755657494068146, "global_step": 198062, "epoch": 2225, "lr": 1.7034761510927816e-05}
{"train_loss": 0.057041607797145844, "global_step": 198063, "epoch": 2225, "lr": 1.7034325623220543e-05}
{"train_loss": 0.08342364430427551, "global_step": 198064, "epoch": 2225, "lr": 1.7033889739945085e-05}
{"train_loss": 0.06332378089427948, "global_step": 198065, "epoch": 2225, "lr": 1.7033453861101538e-05}
{"train_loss": 0.08066297322511673, "global_step": 198066, "epoch": 2225, "lr": 1.7033017986689924e-05}
{"train_loss": 0.1398889273405075, "global_step": 198067, "epoch": 2225, "lr": 1.703258211671034e-05}
{"train_loss": 0.047253094613552094, "global_step": 198068, "epoch": 2225, "lr": 1.7032146251162818e-05}
{"train_loss": 0.03745489567518234, "global_step": 198069, "epoch": 2225, "lr": 1.7031710390047406e-05}
{"train_loss": 0.055511556565761566, "global_step": 198070, "epoch": 2225, "lr": 1.7031274533364195e-05}
{"train_loss": 0.06995901465415955, "global_step": 198071, "epoch": 2225, "lr": 1.7030838681113213e-05}
{"train_loss": 0.015010911040008068, "global_step": 198072, "epoch": 2225, "lr": 1.703040283329455e-05}
{"train_loss": 0.050379592925310135, "global_step": 198073, "epoch": 2225, "lr": 1.702996698990823e-05}
{"train_loss": 0.07798635214567184, "global_step": 198074, "epoch": 2225, "lr": 1.702953115095434e-05}
{"train_loss": 0.025032108649611473, "global_step": 198075, "epoch": 2225, "lr": 1.702909531643292e-05}
{"train_loss": 0.07748649269342422, "global_step": 198076, "epoch": 2225, "lr": 1.702865948634405e-05}
{"train_loss": 0.046912577003240585, "global_step": 198077, "epoch": 2225, "lr": 1.702822366068776e-05}
{"train_loss": 0.04632194712758064, "global_step": 198078, "epoch": 2225, "lr": 1.7027787839464133e-05}
{"train_loss": 0.11345888674259186, "global_step": 198079, "epoch": 2225, "lr": 1.702735202267321e-05}
{"train_loss": 0.0442793108522892, "global_step": 198080, "epoch": 2225, "lr": 1.7026916210315054e-05}
{"train_loss": 0.05704200640320778, "global_step": 198081, "epoch": 2225, "lr": 1.702648040238974e-05}
{"train_loss": 0.05865902453660965, "global_step": 198082, "epoch": 2225, "lr": 1.70260445988973e-05}
{"train_loss": 0.08030341565608978, "global_step": 198083, "epoch": 2225, "lr": 1.7025608799837824e-05}
{"train_loss": 0.0709121823310852, "global_step": 198084, "epoch": 2225, "lr": 1.7025173005211337e-05}
{"train_loss": 0.047798458486795425, "global_step": 198085, "epoch": 2225, "lr": 1.702473721501793e-05}
{"train_loss": 0.09661493450403214, "global_step": 198086, "epoch": 2225, "lr": 1.7024301429257627e-05}
{"train_loss": 0.04502628371119499, "global_step": 198087, "epoch": 2225, "lr": 1.702386564793052e-05}
{"train_loss": 0.062073152512311935, "global_step": 198088, "epoch": 2225, "lr": 1.7023429871036638e-05}
{"train_loss": 0.028816459700465202, "global_step": 198089, "epoch": 2225, "lr": 1.702299409857607e-05}
{"train_loss": 0.04769685119390488, "global_step": 198090, "epoch": 2225, "lr": 1.7022558330548845e-05}
{"train_loss": 0.04341195523738861, "global_step": 198091, "epoch": 2225, "lr": 1.7022122566955047e-05}
{"train_loss": 0.05073828622698784, "global_step": 198092, "epoch": 2225, "lr": 1.7021686807794707e-05}
{"train_loss": 0.11116942018270493, "global_step": 198093, "epoch": 2225, "lr": 1.7021251053067917e-05}
{"train_loss": 0.06605266779661179, "global_step": 198094, "epoch": 2225, "lr": 1.70208153027747e-05}
{"train_loss": 0.05753874406218529, "global_step": 198095, "epoch": 2225, "lr": 1.702037955691515e-05}
{"train_loss": 0.06591634452342987, "global_step": 198096, "epoch": 2225, "lr": 1.701994381548929e-05}
{"train_loss": 0.08283144980669022, "global_step": 198097, "epoch": 2225, "lr": 1.7019508078497203e-05}
{"train_loss": 0.07069382816553116, "global_step": 198098, "epoch": 2225, "lr": 1.7019072345938953e-05}
{"train_loss": 0.14029647409915924, "global_step": 198099, "epoch": 2225, "lr": 1.701863661781457e-05}
{"train_loss": 0.028217650949954987, "global_step": 198100, "epoch": 2225, "lr": 1.7018200894124146e-05}
{"train_loss": 0.025897569954395294, "global_step": 198101, "epoch": 2225, "lr": 1.7017765174867722e-05}
{"train_loss": 0.07355824112892151, "global_step": 198102, "epoch": 2225, "lr": 1.701732946004534e-05}
{"train_loss": 0.06808110326528549, "global_step": 198103, "epoch": 2225, "lr": 1.7016893749657098e-05}
{"train_loss": 0.0982418805360794, "global_step": 198104, "epoch": 2225, "lr": 1.701645804370301e-05}
{"train_loss": 0.06475254148244858, "global_step": 198105, "epoch": 2225, "lr": 1.7016022342183175e-05}
{"train_loss": 0.07043842226266861, "global_step": 198106, "epoch": 2225, "lr": 1.7015586645097635e-05}
{"train_loss": 0.062333010137081146, "global_step": 198107, "epoch": 2225, "lr": 1.7015150952446434e-05}
{"train_loss": 0.05331587418913841, "global_step": 198108, "epoch": 2225, "lr": 1.7014715264229657e-05}
{"train_loss": 0.0647040456533432, "global_step": 198109, "epoch": 2225, "lr": 1.7014279580447335e-05}
{"train_loss": 0.07072978466749191, "global_step": 198110, "epoch": 2225, "lr": 1.7013843901099552e-05}
{"train_loss": 0.059774480760097504, "global_step": 198111, "epoch": 2225, "lr": 1.7013408226186346e-05}
{"train_loss": 0.046154044568538666, "global_step": 198112, "epoch": 2225, "lr": 1.7012972555707794e-05}
{"train_loss": 0.06355798808502013, "global_step": 198113, "epoch": 2225, "lr": 1.7012536889663938e-05, "val_loss": 8.084756851196289, "train_action_mse_error": 36.299686431884766}
{"train_loss": 0.04291949421167374, "global_step": 198114, "epoch": 2226, "lr": 1.7012101228054855e-05}
{"train_loss": 0.04813055321574211, "global_step": 198115, "epoch": 2226, "lr": 1.7011665570880582e-05}
{"train_loss": 0.08173806965351105, "global_step": 198116, "epoch": 2226, "lr": 1.701122991814118e-05}
{"train_loss": 0.046625856310129166, "global_step": 198117, "epoch": 2226, "lr": 1.7010794269836738e-05}
{"train_loss": 0.07193084806203842, "global_step": 198118, "epoch": 2226, "lr": 1.7010358625967276e-05}
{"train_loss": 0.04809834808111191, "global_step": 198119, "epoch": 2226, "lr": 1.7009922986532885e-05}
{"train_loss": 0.09202025830745697, "global_step": 198120, "epoch": 2226, "lr": 1.7009487351533593e-05}
{"train_loss": 0.07096344977617264, "global_step": 198121, "epoch": 2226, "lr": 1.7009051720969486e-05}
{"train_loss": 0.02872586064040661, "global_step": 198122, "epoch": 2226, "lr": 1.7008616094840597e-05}
{"train_loss": 0.06497976928949356, "global_step": 198123, "epoch": 2226, "lr": 1.7008180473147013e-05}
{"train_loss": 0.08123879879713058, "global_step": 198124, "epoch": 2226, "lr": 1.7007744855888758e-05}
{"train_loss": 0.05233335494995117, "global_step": 198125, "epoch": 2226, "lr": 1.7007309243065927e-05}
{"train_loss": 0.06875362247228622, "global_step": 198126, "epoch": 2226, "lr": 1.700687363467855e-05}
{"train_loss": 0.040442466735839844, "global_step": 198127, "epoch": 2226, "lr": 1.7006438030726707e-05}
{"train_loss": 0.09686082601547241, "global_step": 198128, "epoch": 2226, "lr": 1.700600243121043e-05}
{"train_loss": 0.08252368867397308, "global_step": 198129, "epoch": 2226, "lr": 1.700556683612981e-05}
{"train_loss": 0.04364067316055298, "global_step": 198130, "epoch": 2226, "lr": 1.700513124548488e-05}
{"train_loss": 0.045296378433704376, "global_step": 198131, "epoch": 2226, "lr": 1.7004695659275717e-05}
{"train_loss": 0.045000314712524414, "global_step": 198132, "epoch": 2226, "lr": 1.7004260077502356e-05}
{"train_loss": 0.06546637415885925, "global_step": 198133, "epoch": 2226, "lr": 1.700382450016487e-05}
{"train_loss": 0.03546660393476486, "global_step": 198134, "epoch": 2226, "lr": 1.7003388927263336e-05}
{"train_loss": 0.05390644818544388, "global_step": 198135, "epoch": 2226, "lr": 1.7002953358797774e-05}
{"train_loss": 0.08569198846817017, "global_step": 198136, "epoch": 2226, "lr": 1.7002517794768285e-05}
{"train_loss": 0.056552089750766754, "global_step": 198137, "epoch": 2226, "lr": 1.700208223517488e-05}
{"train_loss": 0.0703921914100647, "global_step": 198138, "epoch": 2226, "lr": 1.700164668001767e-05}
{"train_loss": 0.04498732462525368, "global_step": 198139, "epoch": 2226, "lr": 1.700121112929668e-05}
{"train_loss": 0.10932287573814392, "global_step": 198140, "epoch": 2226, "lr": 1.7000775583011958e-05}
{"train_loss": 0.094791479408741, "global_step": 198141, "epoch": 2226, "lr": 1.70003400411636e-05}
{"train_loss": 0.12471509724855423, "global_step": 198142, "epoch": 2226, "lr": 1.6999904503751623e-05}
{"train_loss": 0.06589754670858383, "global_step": 198143, "epoch": 2226, "lr": 1.6999468970776127e-05}
{"train_loss": 0.08501159399747849, "global_step": 198144, "epoch": 2226, "lr": 1.6999033442237133e-05}
{"train_loss": 0.09287363290786743, "global_step": 198145, "epoch": 2226, "lr": 1.6998597918134728e-05}
{"train_loss": 0.0869225412607193, "global_step": 198146, "epoch": 2226, "lr": 1.6998162398468963e-05}
{"train_loss": 0.10383366793394089, "global_step": 198147, "epoch": 2226, "lr": 1.6997726883239878e-05}
{"train_loss": 0.06701581925153732, "global_step": 198148, "epoch": 2226, "lr": 1.6997291372447566e-05}
{"train_loss": 0.1338232010602951, "global_step": 198149, "epoch": 2226, "lr": 1.6996855866092042e-05}
{"train_loss": 0.03954630345106125, "global_step": 198150, "epoch": 2226, "lr": 1.699642036417341e-05}
{"train_loss": 0.04662569984793663, "global_step": 198151, "epoch": 2226, "lr": 1.699598486669169e-05}
{"train_loss": 0.06889711320400238, "global_step": 198152, "epoch": 2226, "lr": 1.6995549373646952e-05}
{"train_loss": 0.1202039048075676, "global_step": 198153, "epoch": 2226, "lr": 1.6995113885039283e-05}
{"train_loss": 0.06929722428321838, "global_step": 198154, "epoch": 2226, "lr": 1.69946784008687e-05}
{"train_loss": 0.03128126636147499, "global_step": 198155, "epoch": 2226, "lr": 1.6994242921135295e-05}
{"train_loss": 0.07902976870536804, "global_step": 198156, "epoch": 2226, "lr": 1.69938074458391e-05}
{"train_loss": 0.057271115481853485, "global_step": 198157, "epoch": 2226, "lr": 1.6993371974980198e-05}
{"train_loss": 0.05834351107478142, "global_step": 198158, "epoch": 2226, "lr": 1.6992936508558622e-05}
{"train_loss": 0.06437865644693375, "global_step": 198159, "epoch": 2226, "lr": 1.6992501046574455e-05}
{"train_loss": 0.09015921503305435, "global_step": 198160, "epoch": 2226, "lr": 1.699206558902773e-05}
{"train_loss": 0.06503260135650635, "global_step": 198161, "epoch": 2226, "lr": 1.6991630135918536e-05}
{"train_loss": 0.06936103105545044, "global_step": 198162, "epoch": 2226, "lr": 1.6991194687246898e-05}
{"train_loss": 0.07443997263908386, "global_step": 198163, "epoch": 2226, "lr": 1.6990759243012903e-05}
{"train_loss": 0.06428660452365875, "global_step": 198164, "epoch": 2226, "lr": 1.699032380321659e-05}
{"train_loss": 0.0968288779258728, "global_step": 198165, "epoch": 2226, "lr": 1.6989888367858035e-05}
{"train_loss": 0.05545893311500549, "global_step": 198166, "epoch": 2226, "lr": 1.6989452936937278e-05}
{"train_loss": 0.06593073904514313, "global_step": 198167, "epoch": 2226, "lr": 1.6989017510454396e-05}
{"train_loss": 0.10628322511911392, "global_step": 198168, "epoch": 2226, "lr": 1.6988582088409426e-05}
{"train_loss": 0.055013228207826614, "global_step": 198169, "epoch": 2226, "lr": 1.698814667080244e-05}
{"train_loss": 0.04421226680278778, "global_step": 198170, "epoch": 2226, "lr": 1.698771125763351e-05}
{"train_loss": 0.043469980359077454, "global_step": 198171, "epoch": 2226, "lr": 1.698727584890266e-05}
{"train_loss": 0.0488266795873642, "global_step": 198172, "epoch": 2226, "lr": 1.698684044460999e-05}
{"train_loss": 0.08503987640142441, "global_step": 198173, "epoch": 2226, "lr": 1.698640504475552e-05}
{"train_loss": 0.10218197852373123, "global_step": 198174, "epoch": 2226, "lr": 1.698596964933934e-05}
{"train_loss": 0.1045074388384819, "global_step": 198175, "epoch": 2226, "lr": 1.698553425836148e-05}
{"train_loss": 0.030849464237689972, "global_step": 198176, "epoch": 2226, "lr": 1.6985098871822026e-05}
{"train_loss": 0.054885707795619965, "global_step": 198177, "epoch": 2226, "lr": 1.6984663489721008e-05}
{"train_loss": 0.07671000808477402, "global_step": 198178, "epoch": 2226, "lr": 1.6984228112058514e-05}
{"train_loss": 0.04097292199730873, "global_step": 198179, "epoch": 2226, "lr": 1.698379273883459e-05}
{"train_loss": 0.112757109105587, "global_step": 198180, "epoch": 2226, "lr": 1.6983357370049274e-05}
{"train_loss": 0.07520754635334015, "global_step": 198181, "epoch": 2226, "lr": 1.6982922005702666e-05}
{"train_loss": 0.06354440003633499, "global_step": 198182, "epoch": 2226, "lr": 1.6982486645794782e-05}
{"train_loss": 0.07070562243461609, "global_step": 198183, "epoch": 2226, "lr": 1.6982051290325714e-05}
{"train_loss": 0.07588258385658264, "global_step": 198184, "epoch": 2226, "lr": 1.6981615939295507e-05}
{"train_loss": 0.06228012591600418, "global_step": 198185, "epoch": 2226, "lr": 1.6981180592704205e-05}
{"train_loss": 0.041252974420785904, "global_step": 198186, "epoch": 2226, "lr": 1.69807452505519e-05}
{"train_loss": 0.0554782897233963, "global_step": 198187, "epoch": 2226, "lr": 1.698030991283861e-05}
{"train_loss": 0.08617396652698517, "global_step": 198188, "epoch": 2226, "lr": 1.6979874579564414e-05}
{"train_loss": 0.10100529342889786, "global_step": 198189, "epoch": 2226, "lr": 1.697943925072939e-05}
{"train_loss": 0.08026556670665741, "global_step": 198190, "epoch": 2226, "lr": 1.6979003926333564e-05}
{"train_loss": 0.044946737587451935, "global_step": 198191, "epoch": 2226, "lr": 1.6978568606377022e-05}
{"train_loss": 0.09487535059452057, "global_step": 198192, "epoch": 2226, "lr": 1.6978133290859794e-05}
{"train_loss": 0.047078877687454224, "global_step": 198193, "epoch": 2226, "lr": 1.6977697979781964e-05}
{"train_loss": 0.07239063084125519, "global_step": 198194, "epoch": 2226, "lr": 1.6977262673143575e-05}
{"train_loss": 0.022308120504021645, "global_step": 198195, "epoch": 2226, "lr": 1.6976827370944696e-05}
{"train_loss": 0.1261845976114273, "global_step": 198196, "epoch": 2226, "lr": 1.697639207318537e-05}
{"train_loss": 0.04240242391824722, "global_step": 198197, "epoch": 2226, "lr": 1.6975956779865682e-05}
{"train_loss": 0.07970473170280457, "global_step": 198198, "epoch": 2226, "lr": 1.697552149098565e-05}
{"train_loss": 0.031238125637173653, "global_step": 198199, "epoch": 2226, "lr": 1.697508620654538e-05}
{"train_loss": 0.023809513077139854, "global_step": 198200, "epoch": 2226, "lr": 1.6974650926544894e-05}
{"train_loss": 0.05277133360505104, "global_step": 198201, "epoch": 2226, "lr": 1.6974215650984277e-05}
{"train_loss": 0.06833747832962636, "global_step": 198202, "epoch": 2226, "lr": 1.6973780379863552e-05, "val_loss": 8.010455131530762}
{"train_loss": 0.01359778642654419, "global_step": 198203, "epoch": 2227, "lr": 1.697334511318282e-05}
{"train_loss": 0.0479581356048584, "global_step": 198204, "epoch": 2227, "lr": 1.6972909850942105e-05}
{"train_loss": 0.09516473114490509, "global_step": 198205, "epoch": 2227, "lr": 1.6972474593141495e-05}
{"train_loss": 0.052146848291158676, "global_step": 198206, "epoch": 2227, "lr": 1.6972039339781015e-05}
{"train_loss": 0.09296534955501556, "global_step": 198207, "epoch": 2227, "lr": 1.6971604090860743e-05}
{"train_loss": 0.07660599797964096, "global_step": 198208, "epoch": 2227, "lr": 1.6971168846380752e-05}
{"train_loss": 0.07211316376924515, "global_step": 198209, "epoch": 2227, "lr": 1.697073360634107e-05}
{"train_loss": 0.04015342891216278, "global_step": 198210, "epoch": 2227, "lr": 1.6970298370741788e-05}
{"train_loss": 0.024653680622577667, "global_step": 198211, "epoch": 2227, "lr": 1.6969863139582926e-05}
{"train_loss": 0.062244489789009094, "global_step": 198212, "epoch": 2227, "lr": 1.6969427912864583e-05}
{"train_loss": 0.05008924379944801, "global_step": 198213, "epoch": 2227, "lr": 1.6968992690586783e-05}
{"train_loss": 0.060301654040813446, "global_step": 198214, "epoch": 2227, "lr": 1.6968557472749614e-05}
{"train_loss": 0.07840347290039062, "global_step": 198215, "epoch": 2227, "lr": 1.6968122259353097e-05}
{"train_loss": 0.047296810895204544, "global_step": 198216, "epoch": 2227, "lr": 1.696768705039734e-05}
{"train_loss": 0.1013021171092987, "global_step": 198217, "epoch": 2227, "lr": 1.696725184588237e-05}
{"train_loss": 0.03548314794898033, "global_step": 198218, "epoch": 2227, "lr": 1.696681664580823e-05}
{"train_loss": 0.05737787485122681, "global_step": 198219, "epoch": 2227, "lr": 1.6966381450175022e-05}
{"train_loss": 0.07623131573200226, "global_step": 198220, "epoch": 2227, "lr": 1.6965946258982762e-05}
{"train_loss": 0.037029508501291275, "global_step": 198221, "epoch": 2227, "lr": 1.6965511072231543e-05}
{"train_loss": 0.0427316352725029, "global_step": 198222, "epoch": 2227, "lr": 1.696507588992141e-05}
{"train_loss": 0.05416586250066757, "global_step": 198223, "epoch": 2227, "lr": 1.6964640712052406e-05}
{"train_loss": 0.09303218126296997, "global_step": 198224, "epoch": 2227, "lr": 1.6964205538624602e-05}
{"train_loss": 0.05064094811677933, "global_step": 198225, "epoch": 2227, "lr": 1.696377036963807e-05}
{"train_loss": 0.10661715269088745, "global_step": 198226, "epoch": 2227, "lr": 1.6963335205092846e-05}
{"train_loss": 0.08277583122253418, "global_step": 198227, "epoch": 2227, "lr": 1.6962900044989016e-05}
{"train_loss": 0.06449560821056366, "global_step": 198228, "epoch": 2227, "lr": 1.69624648893266e-05}
{"train_loss": 0.04852152615785599, "global_step": 198229, "epoch": 2227, "lr": 1.6962029738105693e-05}
{"train_loss": 0.035312898457050323, "global_step": 198230, "epoch": 2227, "lr": 1.696159459132633e-05}
{"train_loss": 0.03551984950900078, "global_step": 198231, "epoch": 2227, "lr": 1.6961159448988588e-05}
{"train_loss": 0.06860329955816269, "global_step": 198232, "epoch": 2227, "lr": 1.6960724311092502e-05}
{"train_loss": 0.1084296777844429, "global_step": 198233, "epoch": 2227, "lr": 1.6960289177638155e-05}
{"train_loss": 0.06272228062152863, "global_step": 198234, "epoch": 2227, "lr": 1.6959854048625584e-05}
{"train_loss": 0.0625954195857048, "global_step": 198235, "epoch": 2227, "lr": 1.695941892405487e-05}
{"train_loss": 0.04579818993806839, "global_step": 198236, "epoch": 2227, "lr": 1.695898380392605e-05}
{"train_loss": 0.05687246471643448, "global_step": 198237, "epoch": 2227, "lr": 1.69585486882392e-05}
{"train_loss": 0.024554740637540817, "global_step": 198238, "epoch": 2227, "lr": 1.6958113576994356e-05}
{"train_loss": 0.05711543187499046, "global_step": 198239, "epoch": 2227, "lr": 1.6957678470191613e-05}
{"train_loss": 0.1355394572019577, "global_step": 198240, "epoch": 2227, "lr": 1.695724336783098e-05}
{"train_loss": 0.09244745224714279, "global_step": 198241, "epoch": 2227, "lr": 1.6956808269912573e-05}
{"train_loss": 0.07198119163513184, "global_step": 198242, "epoch": 2227, "lr": 1.6956373176436396e-05}
{"train_loss": 0.025604665279388428, "global_step": 198243, "epoch": 2227, "lr": 1.6955938087402533e-05}
{"train_loss": 0.0407307930290699, "global_step": 198244, "epoch": 2227, "lr": 1.6955503002811057e-05}
{"train_loss": 0.06410084664821625, "global_step": 198245, "epoch": 2227, "lr": 1.6955067922661994e-05}
{"train_loss": 0.10028170049190521, "global_step": 198246, "epoch": 2227, "lr": 1.6954632846955436e-05}
{"train_loss": 0.09364401549100876, "global_step": 198247, "epoch": 2227, "lr": 1.695419777569141e-05}
{"train_loss": 0.07497355341911316, "global_step": 198248, "epoch": 2227, "lr": 1.6953762708870008e-05}
{"train_loss": 0.06988774240016937, "global_step": 198249, "epoch": 2227, "lr": 1.6953327646491253e-05}
{"train_loss": 0.08353277295827866, "global_step": 198250, "epoch": 2227, "lr": 1.695289258855523e-05}
{"train_loss": 0.07356858253479004, "global_step": 198251, "epoch": 2227, "lr": 1.6952457535061978e-05}
{"train_loss": 0.06141112372279167, "global_step": 198252, "epoch": 2227, "lr": 1.6952022486011577e-05}
{"train_loss": 0.06936284154653549, "global_step": 198253, "epoch": 2227, "lr": 1.695158744140406e-05}
{"train_loss": 0.0442439429461956, "global_step": 198254, "epoch": 2227, "lr": 1.6951152401239517e-05}
{"train_loss": 0.08099386096000671, "global_step": 198255, "epoch": 2227, "lr": 1.6950717365517986e-05}
{"train_loss": 0.04883769527077675, "global_step": 198256, "epoch": 2227, "lr": 1.6950282334239516e-05}
{"train_loss": 0.04975650832056999, "global_step": 198257, "epoch": 2227, "lr": 1.6949847307404188e-05}
{"train_loss": 0.11425597220659256, "global_step": 198258, "epoch": 2227, "lr": 1.694941228501204e-05}
{"train_loss": 0.08750515431165695, "global_step": 198259, "epoch": 2227, "lr": 1.6948977267063153e-05}
{"train_loss": 0.10447467863559723, "global_step": 198260, "epoch": 2227, "lr": 1.6948542253557558e-05}
{"train_loss": 0.05659930035471916, "global_step": 198261, "epoch": 2227, "lr": 1.6948107244495342e-05}
{"train_loss": 0.06720541417598724, "global_step": 198262, "epoch": 2227, "lr": 1.694767223987654e-05}
{"train_loss": 0.07139480859041214, "global_step": 198263, "epoch": 2227, "lr": 1.694723723970123e-05}
{"train_loss": 0.026690354570746422, "global_step": 198264, "epoch": 2227, "lr": 1.694680224396945e-05}
{"train_loss": 0.05022169649600983, "global_step": 198265, "epoch": 2227, "lr": 1.6946367252681282e-05}
{"train_loss": 0.08341348171234131, "global_step": 198266, "epoch": 2227, "lr": 1.694593226583676e-05}
{"train_loss": 0.10355144739151001, "global_step": 198267, "epoch": 2227, "lr": 1.6945497283435968e-05}
{"train_loss": 0.037447307258844376, "global_step": 198268, "epoch": 2227, "lr": 1.6945062305478937e-05}
{"train_loss": 0.03712117299437523, "global_step": 198269, "epoch": 2227, "lr": 1.6944627331965755e-05}
{"train_loss": 0.04249945655465126, "global_step": 198270, "epoch": 2227, "lr": 1.694419236289645e-05}
{"train_loss": 0.06444626301527023, "global_step": 198271, "epoch": 2227, "lr": 1.6943757398271108e-05}
{"train_loss": 0.0589771531522274, "global_step": 198272, "epoch": 2227, "lr": 1.6943322438089758e-05}
{"train_loss": 0.06548251956701279, "global_step": 198273, "epoch": 2227, "lr": 1.6942887482352495e-05}
{"train_loss": 0.034148454666137695, "global_step": 198274, "epoch": 2227, "lr": 1.694245253105934e-05}
{"train_loss": 0.09077935665845871, "global_step": 198275, "epoch": 2227, "lr": 1.694201758421039e-05}
{"train_loss": 0.12990160286426544, "global_step": 198276, "epoch": 2227, "lr": 1.6941582641805658e-05}
{"train_loss": 0.06034204363822937, "global_step": 198277, "epoch": 2227, "lr": 1.694114770384525e-05}
{"train_loss": 0.045834776014089584, "global_step": 198278, "epoch": 2227, "lr": 1.6940712770329186e-05}
{"train_loss": 0.04407218471169472, "global_step": 198279, "epoch": 2227, "lr": 1.6940277841257536e-05}
{"train_loss": 0.08729960024356842, "global_step": 198280, "epoch": 2227, "lr": 1.6939842916630383e-05}
{"train_loss": 0.041825179010629654, "global_step": 198281, "epoch": 2227, "lr": 1.6939407996447748e-05}
{"train_loss": 0.08214101940393448, "global_step": 198282, "epoch": 2227, "lr": 1.693897308070972e-05}
{"train_loss": 0.06886360049247742, "global_step": 198283, "epoch": 2227, "lr": 1.6938538169416328e-05}
{"train_loss": 0.03945326432585716, "global_step": 198284, "epoch": 2227, "lr": 1.693810326256767e-05}
{"train_loss": 0.07928831875324249, "global_step": 198285, "epoch": 2227, "lr": 1.6937668360163757e-05}
{"train_loss": 0.03260316327214241, "global_step": 198286, "epoch": 2227, "lr": 1.6937233462204688e-05}
{"train_loss": 0.06415430456399918, "global_step": 198287, "epoch": 2227, "lr": 1.6936798568690492e-05}
{"train_loss": 0.05736996978521347, "global_step": 198288, "epoch": 2227, "lr": 1.6936363679621254e-05}
{"train_loss": 0.04781894013285637, "global_step": 198289, "epoch": 2227, "lr": 1.6935928794997e-05}
{"train_loss": 0.07822968065738678, "global_step": 198290, "epoch": 2227, "lr": 1.6935493914817834e-05}
{"train_loss": 0.0638890971921468, "global_step": 198291, "epoch": 2227, "lr": 1.693505903908376e-05, "val_loss": 8.200639724731445}
{"train_loss": 0.06553111225366592, "global_step": 198292, "epoch": 2228, "lr": 1.6934624167794887e-05}
{"train_loss": 0.08278782665729523, "global_step": 198293, "epoch": 2228, "lr": 1.6934189300951238e-05}
{"train_loss": 0.049401577562093735, "global_step": 198294, "epoch": 2228, "lr": 1.6933754438552895e-05}
{"train_loss": 0.047996215522289276, "global_step": 198295, "epoch": 2228, "lr": 1.6933319580599904e-05}
{"train_loss": 0.07582923024892807, "global_step": 198296, "epoch": 2228, "lr": 1.6932884727092314e-05}
{"train_loss": 0.06453466415405273, "global_step": 198297, "epoch": 2228, "lr": 1.693244987803021e-05}
{"train_loss": 0.11792934685945511, "global_step": 198298, "epoch": 2228, "lr": 1.6932015033413618e-05}
{"train_loss": 0.07090337574481964, "global_step": 198299, "epoch": 2228, "lr": 1.693158019324263e-05}
{"train_loss": 0.0490167960524559, "global_step": 198300, "epoch": 2228, "lr": 1.693114535751727e-05}
{"train_loss": 0.04544796794652939, "global_step": 198301, "epoch": 2228, "lr": 1.693071052623763e-05}
{"train_loss": 0.08464986830949783, "global_step": 198302, "epoch": 2228, "lr": 1.6930275699403735e-05}
{"train_loss": 0.05214804410934448, "global_step": 198303, "epoch": 2228, "lr": 1.6929840877015685e-05}
{"train_loss": 0.05773283168673515, "global_step": 198304, "epoch": 2228, "lr": 1.6929406059073495e-05}
{"train_loss": 0.07562941312789917, "global_step": 198305, "epoch": 2228, "lr": 1.6928971245577258e-05}
{"train_loss": 0.1024172380566597, "global_step": 198306, "epoch": 2228, "lr": 1.6928536436527006e-05}
{"train_loss": 0.10324191302061081, "global_step": 198307, "epoch": 2228, "lr": 1.6928101631922816e-05}
{"train_loss": 0.047478217631578445, "global_step": 198308, "epoch": 2228, "lr": 1.6927666831764732e-05}
{"train_loss": 0.07158198207616806, "global_step": 198309, "epoch": 2228, "lr": 1.6927232036052833e-05}
{"train_loss": 0.02131558768451214, "global_step": 198310, "epoch": 2228, "lr": 1.6926797244787145e-05}
{"train_loss": 0.041078321635723114, "global_step": 198311, "epoch": 2228, "lr": 1.692636245796777e-05}
{"train_loss": 0.07953567802906036, "global_step": 198312, "epoch": 2228, "lr": 1.6925927675594723e-05}
{"train_loss": 0.10223756730556488, "global_step": 198313, "epoch": 2228, "lr": 1.6925492897668093e-05}
{"train_loss": 0.07688926160335541, "global_step": 198314, "epoch": 2228, "lr": 1.692505812418792e-05}
{"train_loss": 0.05515853688120842, "global_step": 198315, "epoch": 2228, "lr": 1.692462335515426e-05}
{"train_loss": 0.03380998596549034, "global_step": 198316, "epoch": 2228, "lr": 1.6924188590567207e-05}
{"train_loss": 0.048352405428886414, "global_step": 198317, "epoch": 2228, "lr": 1.6923753830426775e-05}
{"train_loss": 0.06772472709417343, "global_step": 198318, "epoch": 2228, "lr": 1.6923319074733053e-05}
{"train_loss": 0.03813543915748596, "global_step": 198319, "epoch": 2228, "lr": 1.6922884323486077e-05}
{"train_loss": 0.07533037662506104, "global_step": 198320, "epoch": 2228, "lr": 1.692244957668593e-05}
{"train_loss": 0.07867513597011566, "global_step": 198321, "epoch": 2228, "lr": 1.692201483433264e-05}
{"train_loss": 0.04178766533732414, "global_step": 198322, "epoch": 2228, "lr": 1.6921580096426298e-05}
{"train_loss": 0.0529191792011261, "global_step": 198323, "epoch": 2228, "lr": 1.692114536296693e-05}
{"train_loss": 0.09163497388362885, "global_step": 198324, "epoch": 2228, "lr": 1.6920710633954623e-05}
{"train_loss": 0.04240957275032997, "global_step": 198325, "epoch": 2228, "lr": 1.6920275909389412e-05}
{"train_loss": 0.07985994964838028, "global_step": 198326, "epoch": 2228, "lr": 1.6919841189271385e-05}
{"train_loss": 0.05041419342160225, "global_step": 198327, "epoch": 2228, "lr": 1.6919406473600562e-05}
{"train_loss": 0.07521478086709976, "global_step": 198328, "epoch": 2228, "lr": 1.6918971762377038e-05}
{"train_loss": 0.07056065648794174, "global_step": 198329, "epoch": 2228, "lr": 1.691853705560084e-05}
{"train_loss": 0.06428208947181702, "global_step": 198330, "epoch": 2228, "lr": 1.6918102353272057e-05}
{"train_loss": 0.03455212339758873, "global_step": 198331, "epoch": 2228, "lr": 1.691766765539072e-05}
{"train_loss": 0.050925642251968384, "global_step": 198332, "epoch": 2228, "lr": 1.691723296195691e-05}
{"train_loss": 0.053883083164691925, "global_step": 198333, "epoch": 2228, "lr": 1.6916798272970674e-05}
{"train_loss": 0.04389844089746475, "global_step": 198334, "epoch": 2228, "lr": 1.691636358843206e-05}
{"train_loss": 0.07752519845962524, "global_step": 198335, "epoch": 2228, "lr": 1.691592890834115e-05}
{"train_loss": 0.029105374589562416, "global_step": 198336, "epoch": 2228, "lr": 1.6915494232697977e-05}
{"train_loss": 0.04544621706008911, "global_step": 198337, "epoch": 2228, "lr": 1.691505956150262e-05}
{"train_loss": 0.035325393080711365, "global_step": 198338, "epoch": 2228, "lr": 1.6914624894755127e-05}
{"train_loss": 0.08894836157560349, "global_step": 198339, "epoch": 2228, "lr": 1.691419023245557e-05}
{"train_loss": 0.06511557102203369, "global_step": 198340, "epoch": 2228, "lr": 1.6913755574603977e-05}
{"train_loss": 0.0899646133184433, "global_step": 198341, "epoch": 2228, "lr": 1.6913320921200447e-05}
{"train_loss": 0.05105164274573326, "global_step": 198342, "epoch": 2228, "lr": 1.6912886272245e-05}
{"train_loss": 0.05234285071492195, "global_step": 198343, "epoch": 2228, "lr": 1.691245162773773e-05}
{"train_loss": 0.10664001107215881, "global_step": 198344, "epoch": 2228, "lr": 1.691201698767866e-05}
{"train_loss": 0.03409406542778015, "global_step": 198345, "epoch": 2228, "lr": 1.691158235206788e-05}
{"train_loss": 0.04900968819856644, "global_step": 198346, "epoch": 2228, "lr": 1.6911147720905425e-05}
{"train_loss": 0.05805336683988571, "global_step": 198347, "epoch": 2228, "lr": 1.691071309419137e-05}
{"train_loss": 0.055058445781469345, "global_step": 198348, "epoch": 2228, "lr": 1.6910278471925755e-05}
{"train_loss": 0.0763690173625946, "global_step": 198349, "epoch": 2228, "lr": 1.6909843854108664e-05}
{"train_loss": 0.13094227015972137, "global_step": 198350, "epoch": 2228, "lr": 1.6909409240740125e-05}
{"train_loss": 0.031244343146681786, "global_step": 198351, "epoch": 2228, "lr": 1.6908974631820218e-05}
{"train_loss": 0.14393867552280426, "global_step": 198352, "epoch": 2228, "lr": 1.690854002734901e-05}
{"train_loss": 0.07518517971038818, "global_step": 198353, "epoch": 2228, "lr": 1.6908105427326525e-05}
{"train_loss": 0.05701909959316254, "global_step": 198354, "epoch": 2228, "lr": 1.6907670831752865e-05}
{"train_loss": 0.09988653659820557, "global_step": 198355, "epoch": 2228, "lr": 1.6907236240628045e-05}
{"train_loss": 0.06172291561961174, "global_step": 198356, "epoch": 2228, "lr": 1.6906801653952158e-05}
{"train_loss": 0.06068601459264755, "global_step": 198357, "epoch": 2228, "lr": 1.6906367071725233e-05}
{"train_loss": 0.04278034716844559, "global_step": 198358, "epoch": 2228, "lr": 1.6905932493947356e-05}
{"train_loss": 0.029472503811120987, "global_step": 198359, "epoch": 2228, "lr": 1.6905497920618563e-05}
{"train_loss": 0.04355458542704582, "global_step": 198360, "epoch": 2228, "lr": 1.6905063351738936e-05}
{"train_loss": 0.09190300107002258, "global_step": 198361, "epoch": 2228, "lr": 1.6904628787308506e-05}
{"train_loss": 0.05757825821638107, "global_step": 198362, "epoch": 2228, "lr": 1.6904194227327363e-05}
{"train_loss": 0.05919332429766655, "global_step": 198363, "epoch": 2228, "lr": 1.6903759671795525e-05}
{"train_loss": 0.1036851555109024, "global_step": 198364, "epoch": 2228, "lr": 1.6903325120713098e-05}
{"train_loss": 0.04817986488342285, "global_step": 198365, "epoch": 2228, "lr": 1.6902890574080093e-05}
{"train_loss": 0.07489734888076782, "global_step": 198366, "epoch": 2228, "lr": 1.6902456031896607e-05}
{"train_loss": 0.0696697011590004, "global_step": 198367, "epoch": 2228, "lr": 1.6902021494162666e-05}
{"train_loss": 0.052010487765073776, "global_step": 198368, "epoch": 2228, "lr": 1.690158696087837e-05}
{"train_loss": 0.10632819682359695, "global_step": 198369, "epoch": 2228, "lr": 1.6901152432043726e-05}
{"train_loss": 0.0663466528058052, "global_step": 198370, "epoch": 2228, "lr": 1.6900717907658835e-05}
{"train_loss": 0.08710362762212753, "global_step": 198371, "epoch": 2228, "lr": 1.6900283387723743e-05}
{"train_loss": 0.10585711151361465, "global_step": 198372, "epoch": 2228, "lr": 1.6899848872238488e-05}
{"train_loss": 0.022859767079353333, "global_step": 198373, "epoch": 2228, "lr": 1.6899414361203158e-05}
{"train_loss": 0.07779555767774582, "global_step": 198374, "epoch": 2228, "lr": 1.6898979854617786e-05}
{"train_loss": 0.10913759469985962, "global_step": 198375, "epoch": 2228, "lr": 1.689854535248246e-05}
{"train_loss": 0.07443137466907501, "global_step": 198376, "epoch": 2228, "lr": 1.68981108547972e-05}
{"train_loss": 0.08137120306491852, "global_step": 198377, "epoch": 2228, "lr": 1.6897676361562102e-05}
{"train_loss": 0.07989883422851562, "global_step": 198378, "epoch": 2228, "lr": 1.6897241872777193e-05}
{"train_loss": 0.062041860073804855, "global_step": 198379, "epoch": 2228, "lr": 1.6896807388442564e-05}
{"train_loss": 0.06647144168029341, "global_step": 198380, "epoch": 2228, "lr": 1.689637290855824e-05, "val_loss": 8.15101146697998}
{"train_loss": 0.03214870020747185, "global_step": 198381, "epoch": 2229, "lr": 1.689593843312431e-05}
{"train_loss": 0.0866413488984108, "global_step": 198382, "epoch": 2229, "lr": 1.6895503962140797e-05}
{"train_loss": 0.06198202446103096, "global_step": 198383, "epoch": 2229, "lr": 1.6895069495607802e-05}
{"train_loss": 0.03805668652057648, "global_step": 198384, "epoch": 2229, "lr": 1.6894635033525345e-05}
{"train_loss": 0.0435192845761776, "global_step": 198385, "epoch": 2229, "lr": 1.6894200575893515e-05}
{"train_loss": 0.0605890192091465, "global_step": 198386, "epoch": 2229, "lr": 1.689376612271234e-05}
{"train_loss": 0.07172606140375137, "global_step": 198387, "epoch": 2229, "lr": 1.6893331673981895e-05}
{"train_loss": 0.07794485241174698, "global_step": 198388, "epoch": 2229, "lr": 1.6892897229702254e-05}
{"train_loss": 0.02284734696149826, "global_step": 198389, "epoch": 2229, "lr": 1.6892462789873447e-05}
{"train_loss": 0.07811740785837173, "global_step": 198390, "epoch": 2229, "lr": 1.689202835449556e-05}
{"train_loss": 0.038685142993927, "global_step": 198391, "epoch": 2229, "lr": 1.6891593923568615e-05}
{"train_loss": 0.06139242276549339, "global_step": 198392, "epoch": 2229, "lr": 1.6891159497092713e-05}
{"train_loss": 0.044538114219903946, "global_step": 198393, "epoch": 2229, "lr": 1.689072507506787e-05}
{"train_loss": 0.06384049355983734, "global_step": 198394, "epoch": 2229, "lr": 1.6890290657494183e-05}
{"train_loss": 0.026269346475601196, "global_step": 198395, "epoch": 2229, "lr": 1.6889856244371684e-05}
{"train_loss": 0.07663130760192871, "global_step": 198396, "epoch": 2229, "lr": 1.6889421835700448e-05}
{"train_loss": 0.06743475794792175, "global_step": 198397, "epoch": 2229, "lr": 1.6888987431480513e-05}
{"train_loss": 0.024345839396119118, "global_step": 198398, "epoch": 2229, "lr": 1.6888553031711962e-05}
{"train_loss": 0.09185429662466049, "global_step": 198399, "epoch": 2229, "lr": 1.6888118636394828e-05}
{"train_loss": 0.0809529721736908, "global_step": 198400, "epoch": 2229, "lr": 1.68876842455292e-05}
{"train_loss": 0.06492253392934799, "global_step": 198401, "epoch": 2229, "lr": 1.6887249859115107e-05}
{"train_loss": 0.08463207632303238, "global_step": 198402, "epoch": 2229, "lr": 1.688681547715263e-05}
{"train_loss": 0.10014990717172623, "global_step": 198403, "epoch": 2229, "lr": 1.68863810996418e-05}
{"train_loss": 0.07815273106098175, "global_step": 198404, "epoch": 2229, "lr": 1.6885946726582714e-05}
{"train_loss": 0.06304479390382767, "global_step": 198405, "epoch": 2229, "lr": 1.688551235797539e-05}
{"train_loss": 0.08806303888559341, "global_step": 198406, "epoch": 2229, "lr": 1.688507799381992e-05}
{"train_loss": 0.08623173832893372, "global_step": 198407, "epoch": 2229, "lr": 1.688464363411633e-05}
{"train_loss": 0.03938832879066467, "global_step": 198408, "epoch": 2229, "lr": 1.6884209278864715e-05}
{"train_loss": 0.0946301743388176, "global_step": 198409, "epoch": 2229, "lr": 1.6883774928065098e-05}
{"train_loss": 0.07003014534711838, "global_step": 198410, "epoch": 2229, "lr": 1.688334058171757e-05}
{"train_loss": 0.080243781208992, "global_step": 198411, "epoch": 2229, "lr": 1.6882906239822177e-05}
{"train_loss": 0.0801663026213646, "global_step": 198412, "epoch": 2229, "lr": 1.6882471902378955e-05}
{"train_loss": 0.014914829283952713, "global_step": 198413, "epoch": 2229, "lr": 1.688203756938799e-05}
{"train_loss": 0.04194460064172745, "global_step": 198414, "epoch": 2229, "lr": 1.6881603240849325e-05}
{"train_loss": 0.0392879992723465, "global_step": 198415, "epoch": 2229, "lr": 1.688116891676304e-05}
{"train_loss": 0.032319240272045135, "global_step": 198416, "epoch": 2229, "lr": 1.6880734597129156e-05}
{"train_loss": 0.04670418053865433, "global_step": 198417, "epoch": 2229, "lr": 1.6880300281947776e-05}
{"train_loss": 0.04573320224881172, "global_step": 198418, "epoch": 2229, "lr": 1.6879865971218916e-05}
{"train_loss": 0.02526387758553028, "global_step": 198419, "epoch": 2229, "lr": 1.6879431664942667e-05}
{"train_loss": 0.06428595632314682, "global_step": 198420, "epoch": 2229, "lr": 1.6878997363119064e-05}
{"train_loss": 0.08371014893054962, "global_step": 198421, "epoch": 2229, "lr": 1.687856306574819e-05}
{"train_loss": 0.0662115216255188, "global_step": 198422, "epoch": 2229, "lr": 1.6878128772830076e-05}
{"train_loss": 0.04670125991106033, "global_step": 198423, "epoch": 2229, "lr": 1.6877694484364793e-05}
{"train_loss": 0.10498573631048203, "global_step": 198424, "epoch": 2229, "lr": 1.687726020035242e-05}
{"train_loss": 0.08894622325897217, "global_step": 198425, "epoch": 2229, "lr": 1.6876825920792972e-05}
{"train_loss": 0.05454663187265396, "global_step": 198426, "epoch": 2229, "lr": 1.687639164568655e-05}
{"train_loss": 0.01728103496134281, "global_step": 198427, "epoch": 2229, "lr": 1.687595737503318e-05}
{"train_loss": 0.05255800485610962, "global_step": 198428, "epoch": 2229, "lr": 1.6875523108832947e-05}
{"train_loss": 0.08866408467292786, "global_step": 198429, "epoch": 2229, "lr": 1.6875088847085878e-05}
{"train_loss": 0.0436941422522068, "global_step": 198430, "epoch": 2229, "lr": 1.687465458979207e-05}
{"train_loss": 0.07010812312364578, "global_step": 198431, "epoch": 2229, "lr": 1.6874220336951546e-05}
{"train_loss": 0.07105700671672821, "global_step": 198432, "epoch": 2229, "lr": 1.6873786088564392e-05}
{"train_loss": 0.022677671164274216, "global_step": 198433, "epoch": 2229, "lr": 1.6873351844630636e-05}
{"train_loss": 0.10810627788305283, "global_step": 198434, "epoch": 2229, "lr": 1.6872917605150372e-05}
{"train_loss": 0.09263545274734497, "global_step": 198435, "epoch": 2229, "lr": 1.6872483370123625e-05}
{"train_loss": 0.04597223177552223, "global_step": 198436, "epoch": 2229, "lr": 1.687204913955049e-05}
{"train_loss": 0.033203039318323135, "global_step": 198437, "epoch": 2229, "lr": 1.6871614913430977e-05}
{"train_loss": 0.069643534719944, "global_step": 198438, "epoch": 2229, "lr": 1.68711806917652e-05}
{"train_loss": 0.029455574229359627, "global_step": 198439, "epoch": 2229, "lr": 1.6870746474553163e-05}
{"train_loss": 0.05135239288210869, "global_step": 198440, "epoch": 2229, "lr": 1.687031226179497e-05}
{"train_loss": 0.08532662689685822, "global_step": 198441, "epoch": 2229, "lr": 1.6869878053490645e-05}
{"train_loss": 0.11388339102268219, "global_step": 198442, "epoch": 2229, "lr": 1.686944384964026e-05}
{"train_loss": 0.056601863354444504, "global_step": 198443, "epoch": 2229, "lr": 1.686900965024389e-05}
{"train_loss": 0.07450110465288162, "global_step": 198444, "epoch": 2229, "lr": 1.6868575455301576e-05}
{"train_loss": 0.0646941214799881, "global_step": 198445, "epoch": 2229, "lr": 1.6868141264813363e-05}
{"train_loss": 0.03418455272912979, "global_step": 198446, "epoch": 2229, "lr": 1.6867707078779343e-05}
{"train_loss": 0.07134836912155151, "global_step": 198447, "epoch": 2229, "lr": 1.6867272897199538e-05}
{"train_loss": 0.05428308993577957, "global_step": 198448, "epoch": 2229, "lr": 1.686683872007404e-05}
{"train_loss": 0.041303861886262894, "global_step": 198449, "epoch": 2229, "lr": 1.68664045474029e-05}
{"train_loss": 0.042922988533973694, "global_step": 198450, "epoch": 2229, "lr": 1.6865970379186142e-05}
{"train_loss": 0.0786970779299736, "global_step": 198451, "epoch": 2229, "lr": 1.6865536215423865e-05}
{"train_loss": 0.05537611246109009, "global_step": 198452, "epoch": 2229, "lr": 1.6865102056116107e-05}
{"train_loss": 0.051103390753269196, "global_step": 198453, "epoch": 2229, "lr": 1.686466790126294e-05}
{"train_loss": 0.06507423520088196, "global_step": 198454, "epoch": 2229, "lr": 1.68642337508644e-05}
{"train_loss": 0.04452402889728546, "global_step": 198455, "epoch": 2229, "lr": 1.6863799604920577e-05}
{"train_loss": 0.013828722760081291, "global_step": 198456, "epoch": 2229, "lr": 1.6863365463431498e-05}
{"train_loss": 0.0780031606554985, "global_step": 198457, "epoch": 2229, "lr": 1.6862931326397247e-05}
{"train_loss": 0.03359444439411163, "global_step": 198458, "epoch": 2229, "lr": 1.6862497193817857e-05}
{"train_loss": 0.04896905645728111, "global_step": 198459, "epoch": 2229, "lr": 1.6862063065693417e-05}
{"train_loss": 0.045011430978775024, "global_step": 198460, "epoch": 2229, "lr": 1.6861628942023956e-05}
{"train_loss": 0.030072202906012535, "global_step": 198461, "epoch": 2229, "lr": 1.686119482280954e-05}
{"train_loss": 0.07668597251176834, "global_step": 198462, "epoch": 2229, "lr": 1.686076070805025e-05}
{"train_loss": 0.046072620898485184, "global_step": 198463, "epoch": 2229, "lr": 1.6860326597746112e-05}
{"train_loss": 0.0654076486825943, "global_step": 198464, "epoch": 2229, "lr": 1.6859892491897212e-05}
{"train_loss": 0.044070545583963394, "global_step": 198465, "epoch": 2229, "lr": 1.6859458390503587e-05}
{"train_loss": 0.08644883334636688, "global_step": 198466, "epoch": 2229, "lr": 1.685902429356531e-05}
{"train_loss": 0.07743944227695465, "global_step": 198467, "epoch": 2229, "lr": 1.6858590201082424e-05}
{"train_loss": 0.05977478623390198, "global_step": 198468, "epoch": 2229, "lr": 1.6858156113055006e-05}
{"train_loss": 0.05980735770269726, "global_step": 198469, "epoch": 2229, "lr": 1.6857722029483096e-05, "val_loss": 8.393621444702148}
{"train_loss": 0.12098436057567596, "global_step": 198470, "epoch": 2230, "lr": 1.6857287950366772e-05}
{"train_loss": 0.05273835361003876, "global_step": 198471, "epoch": 2230, "lr": 1.6856853875706076e-05}
{"train_loss": 0.04262154921889305, "global_step": 198472, "epoch": 2230, "lr": 1.6856419805501077e-05}
{"train_loss": 0.04498325660824776, "global_step": 198473, "epoch": 2230, "lr": 1.6855985739751818e-05}
{"train_loss": 0.027772827073931694, "global_step": 198474, "epoch": 2230, "lr": 1.6855551678458382e-05}
{"train_loss": 0.04192999377846718, "global_step": 198475, "epoch": 2230, "lr": 1.685511762162079e-05}
{"train_loss": 0.06794261932373047, "global_step": 198476, "epoch": 2230, "lr": 1.6854683569239148e-05}
{"train_loss": 0.03588541969656944, "global_step": 198477, "epoch": 2230, "lr": 1.6854249521313474e-05}
{"train_loss": 0.0669471025466919, "global_step": 198478, "epoch": 2230, "lr": 1.6853815477843844e-05}
{"train_loss": 0.07603656500577927, "global_step": 198479, "epoch": 2230, "lr": 1.6853381438830323e-05}
{"train_loss": 0.04701818525791168, "global_step": 198480, "epoch": 2230, "lr": 1.685294740427295e-05}
{"train_loss": 0.07496815174818039, "global_step": 198481, "epoch": 2230, "lr": 1.6852513374171808e-05}
{"train_loss": 0.053790830075740814, "global_step": 198482, "epoch": 2230, "lr": 1.6852079348526946e-05}
{"train_loss": 0.03749052807688713, "global_step": 198483, "epoch": 2230, "lr": 1.6851645327338396e-05}
{"train_loss": 0.04159910976886749, "global_step": 198484, "epoch": 2230, "lr": 1.685121131060626e-05}
{"train_loss": 0.041730090975761414, "global_step": 198485, "epoch": 2230, "lr": 1.685077729833055e-05}
{"train_loss": 0.07601453363895416, "global_step": 198486, "epoch": 2230, "lr": 1.6850343290511374e-05}
{"train_loss": 0.050660230219364166, "global_step": 198487, "epoch": 2230, "lr": 1.6849909287148757e-05}
{"train_loss": 0.0502852201461792, "global_step": 198488, "epoch": 2230, "lr": 1.6849475288242754e-05}
{"train_loss": 0.08564554899930954, "global_step": 198489, "epoch": 2230, "lr": 1.6849041293793455e-05}
{"train_loss": 0.06980926543474197, "global_step": 198490, "epoch": 2230, "lr": 1.6848607303800878e-05}
{"train_loss": 0.10399138182401657, "global_step": 198491, "epoch": 2230, "lr": 1.6848173318265114e-05}
{"train_loss": 0.06861300766468048, "global_step": 198492, "epoch": 2230, "lr": 1.6847739337186196e-05}
{"train_loss": 0.037439633160829544, "global_step": 198493, "epoch": 2230, "lr": 1.6847305360564213e-05}
{"train_loss": 0.04269156605005264, "global_step": 198494, "epoch": 2230, "lr": 1.6846871388399193e-05}
{"train_loss": 0.09705325216054916, "global_step": 198495, "epoch": 2230, "lr": 1.684643742069122e-05}
{"train_loss": 0.05141136795282364, "global_step": 198496, "epoch": 2230, "lr": 1.6846003457440322e-05}
{"train_loss": 0.03959796205163002, "global_step": 198497, "epoch": 2230, "lr": 1.6845569498646574e-05}
{"train_loss": 0.07338736951351166, "global_step": 198498, "epoch": 2230, "lr": 1.6845135544310055e-05}
{"train_loss": 0.07150383293628693, "global_step": 198499, "epoch": 2230, "lr": 1.6844701594430783e-05}
{"train_loss": 0.08892802894115448, "global_step": 198500, "epoch": 2230, "lr": 1.6844267649008855e-05}
{"train_loss": 0.07798638194799423, "global_step": 198501, "epoch": 2230, "lr": 1.6843833708044293e-05}
{"train_loss": 0.0913194939494133, "global_step": 198502, "epoch": 2230, "lr": 1.6843399771537195e-05}
{"train_loss": 0.029815033078193665, "global_step": 198503, "epoch": 2230, "lr": 1.684296583948758e-05}
{"train_loss": 0.07758999615907669, "global_step": 198504, "epoch": 2230, "lr": 1.6842531911895537e-05}
{"train_loss": 0.047670044004917145, "global_step": 198505, "epoch": 2230, "lr": 1.6842097988761096e-05}
{"train_loss": 0.046108558773994446, "global_step": 198506, "epoch": 2230, "lr": 1.6841664070084346e-05}
{"train_loss": 0.19382844865322113, "global_step": 198507, "epoch": 2230, "lr": 1.6841230155865317e-05}
{"train_loss": 0.09029695391654968, "global_step": 198508, "epoch": 2230, "lr": 1.6840796246104097e-05}
{"train_loss": 0.06262931227684021, "global_step": 198509, "epoch": 2230, "lr": 1.684036234080071e-05}
{"train_loss": 0.04385245591402054, "global_step": 198510, "epoch": 2230, "lr": 1.6839928439955243e-05}
{"train_loss": 0.08187294006347656, "global_step": 198511, "epoch": 2230, "lr": 1.6839494543567735e-05}
{"train_loss": 0.07139647006988525, "global_step": 198512, "epoch": 2230, "lr": 1.6839060651638266e-05}
{"train_loss": 0.030607057735323906, "global_step": 198513, "epoch": 2230, "lr": 1.6838626764166866e-05}
{"train_loss": 0.07850539684295654, "global_step": 198514, "epoch": 2230, "lr": 1.6838192881153602e-05}
{"train_loss": 0.05215616524219513, "global_step": 198515, "epoch": 2230, "lr": 1.6837759002598563e-05}
{"train_loss": 0.13872261345386505, "global_step": 198516, "epoch": 2230, "lr": 1.6837325128501763e-05}
{"train_loss": 0.05088847875595093, "global_step": 198517, "epoch": 2230, "lr": 1.68368912588633e-05}
{"train_loss": 0.05805268883705139, "global_step": 198518, "epoch": 2230, "lr": 1.6836457393683197e-05}
{"train_loss": 0.05248731002211571, "global_step": 198519, "epoch": 2230, "lr": 1.6836023532961542e-05}
{"train_loss": 0.05139743909239769, "global_step": 198520, "epoch": 2230, "lr": 1.683558967669836e-05}
{"train_loss": 0.03297065943479538, "global_step": 198521, "epoch": 2230, "lr": 1.683515582489375e-05}
{"train_loss": 0.13558082282543182, "global_step": 198522, "epoch": 2230, "lr": 1.6834721977547745e-05}
{"train_loss": 0.10021467506885529, "global_step": 198523, "epoch": 2230, "lr": 1.6834288134660396e-05}
{"train_loss": 0.029370460659265518, "global_step": 198524, "epoch": 2230, "lr": 1.6833854296231788e-05}
{"train_loss": 0.08366575092077255, "global_step": 198525, "epoch": 2230, "lr": 1.6833420462261944e-05}
{"train_loss": 0.04805630072951317, "global_step": 198526, "epoch": 2230, "lr": 1.6832986632750964e-05}
{"train_loss": 0.10351302474737167, "global_step": 198527, "epoch": 2230, "lr": 1.6832552807698883e-05}
{"train_loss": 0.11903224140405655, "global_step": 198528, "epoch": 2230, "lr": 1.683211898710574e-05}
{"train_loss": 0.05448620393872261, "global_step": 198529, "epoch": 2230, "lr": 1.6831685170971634e-05}
{"train_loss": 0.10057050734758377, "global_step": 198530, "epoch": 2230, "lr": 1.6831251359296586e-05}
{"train_loss": 0.09593859314918518, "global_step": 198531, "epoch": 2230, "lr": 1.683081755208069e-05}
{"train_loss": 0.10145365446805954, "global_step": 198532, "epoch": 2230, "lr": 1.683038374932397e-05}
{"train_loss": 0.0763632133603096, "global_step": 198533, "epoch": 2230, "lr": 1.6829949951026503e-05}
{"train_loss": 0.05655095353722572, "global_step": 198534, "epoch": 2230, "lr": 1.682951615718836e-05}
{"train_loss": 0.03375667706131935, "global_step": 198535, "epoch": 2230, "lr": 1.682908236780956e-05}
{"train_loss": 0.04408949241042137, "global_step": 198536, "epoch": 2230, "lr": 1.682864858289021e-05}
{"train_loss": 0.05161586403846741, "global_step": 198537, "epoch": 2230, "lr": 1.6828214802430324e-05}
{"train_loss": 0.04322792589664459, "global_step": 198538, "epoch": 2230, "lr": 1.682778102643e-05}
{"train_loss": 0.04706816375255585, "global_step": 198539, "epoch": 2230, "lr": 1.682734725488926e-05}
{"train_loss": 0.07157730311155319, "global_step": 198540, "epoch": 2230, "lr": 1.6826913487808194e-05}
{"train_loss": 0.09457147121429443, "global_step": 198541, "epoch": 2230, "lr": 1.682647972518683e-05}
{"train_loss": 0.06956697255373001, "global_step": 198542, "epoch": 2230, "lr": 1.6826045967025255e-05}
{"train_loss": 0.024620171636343002, "global_step": 198543, "epoch": 2230, "lr": 1.68256122133235e-05}
{"train_loss": 0.07641664892435074, "global_step": 198544, "epoch": 2230, "lr": 1.6825178464081653e-05}
{"train_loss": 0.06506047397851944, "global_step": 198545, "epoch": 2230, "lr": 1.6824744719299734e-05}
{"train_loss": 0.08979269117116928, "global_step": 198546, "epoch": 2230, "lr": 1.6824310978977846e-05}
{"train_loss": 0.08179720491170883, "global_step": 198547, "epoch": 2230, "lr": 1.682387724311601e-05}
{"train_loss": 0.057613346725702286, "global_step": 198548, "epoch": 2230, "lr": 1.6823443511714314e-05}
{"train_loss": 0.07302762567996979, "global_step": 198549, "epoch": 2230, "lr": 1.6823009784772786e-05}
{"train_loss": 0.09290661662817001, "global_step": 198550, "epoch": 2230, "lr": 1.68225760622915e-05}
{"train_loss": 0.0953609049320221, "global_step": 198551, "epoch": 2230, "lr": 1.6822142344270532e-05}
{"train_loss": 0.08650554716587067, "global_step": 198552, "epoch": 2230, "lr": 1.6821708630709898e-05}
{"train_loss": 0.041242580860853195, "global_step": 198553, "epoch": 2230, "lr": 1.6821274921609702e-05}
{"train_loss": 0.061518020927906036, "global_step": 198554, "epoch": 2230, "lr": 1.682084121696997e-05}
{"train_loss": 0.09211303293704987, "global_step": 198555, "epoch": 2230, "lr": 1.682040751679078e-05}
{"train_loss": 0.07543457299470901, "global_step": 198556, "epoch": 2230, "lr": 1.681997382107217e-05}
{"train_loss": 0.11230378597974777, "global_step": 198557, "epoch": 2230, "lr": 1.6819540129814226e-05}
{"train_loss": 0.06835574422324642, "global_step": 198558, "epoch": 2230, "lr": 1.6819106443016974e-05, "val_loss": 8.272802352905273, "train_action_mse_error": 7.834836483001709}
{"train_loss": 0.11405335366725922, "global_step": 198559, "epoch": 2231, "lr": 1.681867276068051e-05}
{"train_loss": 0.08182042837142944, "global_step": 198560, "epoch": 2231, "lr": 1.6818239082804865e-05}
{"train_loss": 0.03955336660146713, "global_step": 198561, "epoch": 2231, "lr": 1.6817805409390086e-05}
{"train_loss": 0.0657375380396843, "global_step": 198562, "epoch": 2231, "lr": 1.6817371740436267e-05}
{"train_loss": 0.0767628401517868, "global_step": 198563, "epoch": 2231, "lr": 1.6816938075943435e-05}
{"train_loss": 0.049976639449596405, "global_step": 198564, "epoch": 2231, "lr": 1.681650441591167e-05}
{"train_loss": 0.06468271464109421, "global_step": 198565, "epoch": 2231, "lr": 1.681607076034103e-05}
{"train_loss": 0.09278029203414917, "global_step": 198566, "epoch": 2231, "lr": 1.6815637109231543e-05}
{"train_loss": 0.04600560665130615, "global_step": 198567, "epoch": 2231, "lr": 1.6815203462583306e-05}
{"train_loss": 0.1161474883556366, "global_step": 198568, "epoch": 2231, "lr": 1.681476982039634e-05}
{"train_loss": 0.1113174557685852, "global_step": 198569, "epoch": 2231, "lr": 1.681433618267073e-05}
{"train_loss": 0.14092275500297546, "global_step": 198570, "epoch": 2231, "lr": 1.6813902549406546e-05}
{"train_loss": 0.06100783497095108, "global_step": 198571, "epoch": 2231, "lr": 1.681346892060381e-05}
{"train_loss": 0.057997990399599075, "global_step": 198572, "epoch": 2231, "lr": 1.6813035296262613e-05}
{"train_loss": 0.03335367888212204, "global_step": 198573, "epoch": 2231, "lr": 1.6812601676382983e-05}
{"train_loss": 0.04598613083362579, "global_step": 198574, "epoch": 2231, "lr": 1.681216806096501e-05}
{"train_loss": 0.06981498003005981, "global_step": 198575, "epoch": 2231, "lr": 1.681173445000872e-05}
{"train_loss": 0.056738581508398056, "global_step": 198576, "epoch": 2231, "lr": 1.6811300843514206e-05}
{"train_loss": 0.0696660727262497, "global_step": 198577, "epoch": 2231, "lr": 1.681086724148149e-05}
{"train_loss": 0.06186467781662941, "global_step": 198578, "epoch": 2231, "lr": 1.6810433643910666e-05}
{"train_loss": 0.06360197067260742, "global_step": 198579, "epoch": 2231, "lr": 1.6810000050801756e-05}
{"train_loss": 0.04167185351252556, "global_step": 198580, "epoch": 2231, "lr": 1.6809566462154857e-05}
{"train_loss": 0.1031903624534607, "global_step": 198581, "epoch": 2231, "lr": 1.680913287796999e-05}
{"train_loss": 0.07509636133909225, "global_step": 198582, "epoch": 2231, "lr": 1.6808699298247244e-05}
{"train_loss": 0.08116093277931213, "global_step": 198583, "epoch": 2231, "lr": 1.680826572298665e-05}
{"train_loss": 0.056653838604688644, "global_step": 198584, "epoch": 2231, "lr": 1.6807832152188296e-05}
{"train_loss": 0.06654704362154007, "global_step": 198585, "epoch": 2231, "lr": 1.6807398585852207e-05}
{"train_loss": 0.05107841640710831, "global_step": 198586, "epoch": 2231, "lr": 1.6806965023978477e-05}
{"train_loss": 0.057040322571992874, "global_step": 198587, "epoch": 2231, "lr": 1.6806531466567127e-05}
{"train_loss": 0.04320614039897919, "global_step": 198588, "epoch": 2231, "lr": 1.680609791361824e-05}
{"train_loss": 0.08186022937297821, "global_step": 198589, "epoch": 2231, "lr": 1.680566436513188e-05}
{"train_loss": 0.06321211904287338, "global_step": 198590, "epoch": 2231, "lr": 1.6805230821108076e-05}
{"train_loss": 0.06296399235725403, "global_step": 198591, "epoch": 2231, "lr": 1.6804797281546925e-05}
{"train_loss": 0.07134733349084854, "global_step": 198592, "epoch": 2231, "lr": 1.6804363746448442e-05}
{"train_loss": 0.053334880620241165, "global_step": 198593, "epoch": 2231, "lr": 1.6803930215812728e-05}
{"train_loss": 0.026631541550159454, "global_step": 198594, "epoch": 2231, "lr": 1.6803496689639807e-05}
{"train_loss": 0.06078791245818138, "global_step": 198595, "epoch": 2231, "lr": 1.6803063167929768e-05}
{"train_loss": 0.05910323187708855, "global_step": 198596, "epoch": 2231, "lr": 1.6802629650682633e-05}
{"train_loss": 0.06592881679534912, "global_step": 198597, "epoch": 2231, "lr": 1.6802196137898498e-05}
{"train_loss": 0.08771783113479614, "global_step": 198598, "epoch": 2231, "lr": 1.68017626295774e-05}
{"train_loss": 0.04811033979058266, "global_step": 198599, "epoch": 2231, "lr": 1.6801329125719385e-05}
{"train_loss": 0.08546271175146103, "global_step": 198600, "epoch": 2231, "lr": 1.6800895626324544e-05}
{"train_loss": 0.061261679977178574, "global_step": 198601, "epoch": 2231, "lr": 1.68004621313929e-05}
{"train_loss": 0.04738686978816986, "global_step": 198602, "epoch": 2231, "lr": 1.6800028640924547e-05}
{"train_loss": 0.09268149733543396, "global_step": 198603, "epoch": 2231, "lr": 1.679959515491951e-05}
{"train_loss": 0.08317718654870987, "global_step": 198604, "epoch": 2231, "lr": 1.6799161673377878e-05}
{"train_loss": 0.05014954134821892, "global_step": 198605, "epoch": 2231, "lr": 1.6798728196299678e-05}
{"train_loss": 0.09543386101722717, "global_step": 198606, "epoch": 2231, "lr": 1.6798294723685e-05}
{"train_loss": 0.06686801463365555, "global_step": 198607, "epoch": 2231, "lr": 1.679786125553387e-05}
{"train_loss": 0.10619798302650452, "global_step": 198608, "epoch": 2231, "lr": 1.6797427791846382e-05}
{"train_loss": 0.06722217798233032, "global_step": 198609, "epoch": 2231, "lr": 1.6796994332622556e-05}
{"train_loss": 0.09622327983379364, "global_step": 198610, "epoch": 2231, "lr": 1.679656087786249e-05}
{"train_loss": 0.09031809121370316, "global_step": 198611, "epoch": 2231, "lr": 1.67961274275662e-05}
{"train_loss": 0.07831668108701706, "global_step": 198612, "epoch": 2231, "lr": 1.6795693981733785e-05}
{"train_loss": 0.12005846202373505, "global_step": 198613, "epoch": 2231, "lr": 1.6795260540365264e-05}
{"train_loss": 0.08068505674600601, "global_step": 198614, "epoch": 2231, "lr": 1.6794827103460735e-05}
{"train_loss": 0.045857254415750504, "global_step": 198615, "epoch": 2231, "lr": 1.679439367102022e-05}
{"train_loss": 0.07599909603595734, "global_step": 198616, "epoch": 2231, "lr": 1.679396024304381e-05}
{"train_loss": 0.04745271056890488, "global_step": 198617, "epoch": 2231, "lr": 1.6793526819531535e-05}
{"train_loss": 0.03455477952957153, "global_step": 198618, "epoch": 2231, "lr": 1.6793093400483477e-05}
{"train_loss": 0.04651195555925369, "global_step": 198619, "epoch": 2231, "lr": 1.6792659985899666e-05}
{"train_loss": 0.04761212691664696, "global_step": 198620, "epoch": 2231, "lr": 1.67922265757802e-05}
{"train_loss": 0.06807217001914978, "global_step": 198621, "epoch": 2231, "lr": 1.6791793170125092e-05}
{"train_loss": 0.0516180545091629, "global_step": 198622, "epoch": 2231, "lr": 1.679135976893444e-05}
{"train_loss": 0.12916401028633118, "global_step": 198623, "epoch": 2231, "lr": 1.6790926372208273e-05}
{"train_loss": 0.13676714897155762, "global_step": 198624, "epoch": 2231, "lr": 1.6790492979946653e-05}
{"train_loss": 0.06481871753931046, "global_step": 198625, "epoch": 2231, "lr": 1.679005959214967e-05}
{"train_loss": 0.023994198068976402, "global_step": 198626, "epoch": 2231, "lr": 1.678962620881734e-05}
{"train_loss": 0.06691808998584747, "global_step": 198627, "epoch": 2231, "lr": 1.6789192829949756e-05}
{"train_loss": 0.08397196233272552, "global_step": 198628, "epoch": 2231, "lr": 1.6788759455546942e-05}
{"train_loss": 0.06879090517759323, "global_step": 198629, "epoch": 2231, "lr": 1.6788326085608998e-05}
{"train_loss": 0.0323210284113884, "global_step": 198630, "epoch": 2231, "lr": 1.6787892720135933e-05}
{"train_loss": 0.03600846603512764, "global_step": 198631, "epoch": 2231, "lr": 1.6787459359127854e-05}
{"train_loss": 0.07132892310619354, "global_step": 198632, "epoch": 2231, "lr": 1.6787026002584777e-05}
{"train_loss": 0.04929393529891968, "global_step": 198633, "epoch": 2231, "lr": 1.6786592650506793e-05}
{"train_loss": 0.06815044581890106, "global_step": 198634, "epoch": 2231, "lr": 1.678615930289394e-05}
{"train_loss": 0.07491045445203781, "global_step": 198635, "epoch": 2231, "lr": 1.678572595974629e-05}
{"train_loss": 0.08290455490350723, "global_step": 198636, "epoch": 2231, "lr": 1.678529262106388e-05}
{"train_loss": 0.06169616058468819, "global_step": 198637, "epoch": 2231, "lr": 1.67848592868468e-05}
{"train_loss": 0.09463503956794739, "global_step": 198638, "epoch": 2231, "lr": 1.678442595709509e-05}
{"train_loss": 0.1205824464559555, "global_step": 198639, "epoch": 2231, "lr": 1.67839926318088e-05}
{"train_loss": 0.06727290898561478, "global_step": 198640, "epoch": 2231, "lr": 1.6783559310988005e-05}
{"train_loss": 0.07013154029846191, "global_step": 198641, "epoch": 2231, "lr": 1.678312599463274e-05}
{"train_loss": 0.030363161116838455, "global_step": 198642, "epoch": 2231, "lr": 1.6782692682743097e-05}
{"train_loss": 0.07111755758523941, "global_step": 198643, "epoch": 2231, "lr": 1.6782259375319098e-05}
{"train_loss": 0.08167487382888794, "global_step": 198644, "epoch": 2231, "lr": 1.678182607236084e-05}
{"train_loss": 0.031438861042261124, "global_step": 198645, "epoch": 2231, "lr": 1.678139277386834e-05}
{"train_loss": 0.0671733096241951, "global_step": 198646, "epoch": 2231, "lr": 1.6780959479841697e-05}
{"train_loss": 0.06914434567344992, "global_step": 198647, "epoch": 2231, "lr": 1.678052619028093e-05, "val_loss": 8.19123363494873}
{"train_loss": 0.10326781868934631, "global_step": 198648, "epoch": 2232, "lr": 1.6780092905186133e-05}
{"train_loss": 0.05114803463220596, "global_step": 198649, "epoch": 2232, "lr": 1.677965962455733e-05}
{"train_loss": 0.06521524488925934, "global_step": 198650, "epoch": 2232, "lr": 1.6779226348394617e-05}
{"train_loss": 0.07525087147951126, "global_step": 198651, "epoch": 2232, "lr": 1.6778793076698014e-05}
{"train_loss": 0.07354560494422913, "global_step": 198652, "epoch": 2232, "lr": 1.6778359809467615e-05}
{"train_loss": 0.0484970323741436, "global_step": 198653, "epoch": 2232, "lr": 1.6777926546703438e-05}
{"train_loss": 0.06055433303117752, "global_step": 198654, "epoch": 2232, "lr": 1.6777493288405588e-05}
{"train_loss": 0.07757698744535446, "global_step": 198655, "epoch": 2232, "lr": 1.677706003457408e-05}
{"train_loss": 0.0481305830180645, "global_step": 198656, "epoch": 2232, "lr": 1.6776626785209004e-05}
{"train_loss": 0.08260703086853027, "global_step": 198657, "epoch": 2232, "lr": 1.6776193540310392e-05}
{"train_loss": 0.08578190952539444, "global_step": 198658, "epoch": 2232, "lr": 1.6775760299878334e-05}
{"train_loss": 0.08122431486845016, "global_step": 198659, "epoch": 2232, "lr": 1.6775327063912853e-05}
{"train_loss": 0.14061644673347473, "global_step": 198660, "epoch": 2232, "lr": 1.6774893832414024e-05}
{"train_loss": 0.06043628603219986, "global_step": 198661, "epoch": 2232, "lr": 1.677446060538192e-05}
{"train_loss": 0.11745844036340714, "global_step": 198662, "epoch": 2232, "lr": 1.677402738281657e-05}
{"train_loss": 0.06390400975942612, "global_step": 198663, "epoch": 2232, "lr": 1.6773594164718064e-05}
{"train_loss": 0.07548549026250839, "global_step": 198664, "epoch": 2232, "lr": 1.6773160951086426e-05}
{"train_loss": 0.047507092356681824, "global_step": 198665, "epoch": 2232, "lr": 1.6772727741921752e-05}
{"train_loss": 0.06540397554636002, "global_step": 198666, "epoch": 2232, "lr": 1.6772294537224052e-05}
{"train_loss": 0.08674655109643936, "global_step": 198667, "epoch": 2232, "lr": 1.6771861336993438e-05}
{"train_loss": 0.07878710329532623, "global_step": 198668, "epoch": 2232, "lr": 1.6771428141229927e-05}
{"train_loss": 0.09010591357946396, "global_step": 198669, "epoch": 2232, "lr": 1.6770994949933606e-05}
{"train_loss": 0.10444693267345428, "global_step": 198670, "epoch": 2232, "lr": 1.6770561763104503e-05}
{"train_loss": 0.06794926524162292, "global_step": 198671, "epoch": 2232, "lr": 1.677012858074271e-05}
{"train_loss": 0.061265572905540466, "global_step": 198672, "epoch": 2232, "lr": 1.6769695402848252e-05}
{"train_loss": 0.034768566489219666, "global_step": 198673, "epoch": 2232, "lr": 1.6769262229421217e-05}
{"train_loss": 0.039561059325933456, "global_step": 198674, "epoch": 2232, "lr": 1.676882906046164e-05}
{"train_loss": 0.03537176549434662, "global_step": 198675, "epoch": 2232, "lr": 1.6768395895969598e-05}
{"train_loss": 0.043650709092617035, "global_step": 198676, "epoch": 2232, "lr": 1.676796273594514e-05}
{"train_loss": 0.05744381621479988, "global_step": 198677, "epoch": 2232, "lr": 1.6767529580388315e-05}
{"train_loss": 0.016908124089241028, "global_step": 198678, "epoch": 2232, "lr": 1.67670964292992e-05}
{"train_loss": 0.08434659242630005, "global_step": 198679, "epoch": 2232, "lr": 1.676666328267783e-05}
{"train_loss": 0.0613270066678524, "global_step": 198680, "epoch": 2232, "lr": 1.676623014052429e-05}
{"train_loss": 0.08883640170097351, "global_step": 198681, "epoch": 2232, "lr": 1.6765797002838617e-05}
{"train_loss": 0.06444322317838669, "global_step": 198682, "epoch": 2232, "lr": 1.6765363869620886e-05}
{"train_loss": 0.029608795419335365, "global_step": 198683, "epoch": 2232, "lr": 1.6764930740871138e-05}
{"train_loss": 0.0231668371707201, "global_step": 198684, "epoch": 2232, "lr": 1.6764497616589452e-05}
{"train_loss": 0.055099006742239, "global_step": 198685, "epoch": 2232, "lr": 1.6764064496775854e-05}
{"train_loss": 0.03670714050531387, "global_step": 198686, "epoch": 2232, "lr": 1.6763631381430444e-05}
{"train_loss": 0.04158603027462959, "global_step": 198687, "epoch": 2232, "lr": 1.6763198270553244e-05}
{"train_loss": 0.07035233825445175, "global_step": 198688, "epoch": 2232, "lr": 1.676276516414434e-05}
{"train_loss": 0.062232572585344315, "global_step": 198689, "epoch": 2232, "lr": 1.6762332062203763e-05}
{"train_loss": 0.06743733584880829, "global_step": 198690, "epoch": 2232, "lr": 1.6761898964731598e-05}
{"train_loss": 0.07790394127368927, "global_step": 198691, "epoch": 2232, "lr": 1.6761465871727877e-05}
{"train_loss": 0.05983273684978485, "global_step": 198692, "epoch": 2232, "lr": 1.6761032783192686e-05}
{"train_loss": 0.03604470565915108, "global_step": 198693, "epoch": 2232, "lr": 1.6760599699126057e-05}
{"train_loss": 0.05252150818705559, "global_step": 198694, "epoch": 2232, "lr": 1.6760166619528073e-05}
{"train_loss": 0.02273738570511341, "global_step": 198695, "epoch": 2232, "lr": 1.675973354439876e-05}
{"train_loss": 0.12168719619512558, "global_step": 198696, "epoch": 2232, "lr": 1.67593004737382e-05}
{"train_loss": 0.036661338061094284, "global_step": 198697, "epoch": 2232, "lr": 1.6758867407546463e-05}
{"train_loss": 0.07474689185619354, "global_step": 198698, "epoch": 2232, "lr": 1.6758434345823578e-05}
{"train_loss": 0.14235888421535492, "global_step": 198699, "epoch": 2232, "lr": 1.6758001288569624e-05}
{"train_loss": 0.037458907812833786, "global_step": 198700, "epoch": 2232, "lr": 1.6757568235784644e-05}
{"train_loss": 0.10208532214164734, "global_step": 198701, "epoch": 2232, "lr": 1.6757135187468715e-05}
{"train_loss": 0.06435022503137589, "global_step": 198702, "epoch": 2232, "lr": 1.6756702143621867e-05}
{"train_loss": 0.039943039417266846, "global_step": 198703, "epoch": 2232, "lr": 1.675626910424419e-05}
{"train_loss": 0.057871948927640915, "global_step": 198704, "epoch": 2232, "lr": 1.6755836069335716e-05}
{"train_loss": 0.0955808162689209, "global_step": 198705, "epoch": 2232, "lr": 1.6755403038896533e-05}
{"train_loss": 0.09784824401140213, "global_step": 198706, "epoch": 2232, "lr": 1.6754970012926664e-05}
{"train_loss": 0.08089322596788406, "global_step": 198707, "epoch": 2232, "lr": 1.6754536991426195e-05}
{"train_loss": 0.04891781508922577, "global_step": 198708, "epoch": 2232, "lr": 1.6754103974395157e-05}
{"train_loss": 0.08406450599431992, "global_step": 198709, "epoch": 2232, "lr": 1.6753670961833647e-05}
{"train_loss": 0.04787806048989296, "global_step": 198710, "epoch": 2232, "lr": 1.675323795374168e-05}
{"train_loss": 0.14259934425354004, "global_step": 198711, "epoch": 2232, "lr": 1.6752804950119348e-05}
{"train_loss": 0.06302924454212189, "global_step": 198712, "epoch": 2232, "lr": 1.6752371950966688e-05}
{"train_loss": 0.0490218885242939, "global_step": 198713, "epoch": 2232, "lr": 1.675193895628378e-05}
{"train_loss": 0.0367080420255661, "global_step": 198714, "epoch": 2232, "lr": 1.675150596607067e-05}
{"train_loss": 0.052256159484386444, "global_step": 198715, "epoch": 2232, "lr": 1.675107298032739e-05}
{"train_loss": 0.06106063351035118, "global_step": 198716, "epoch": 2232, "lr": 1.6750639999054045e-05}
{"train_loss": 0.06384099274873734, "global_step": 198717, "epoch": 2232, "lr": 1.6750207022250658e-05}
{"train_loss": 0.04586648568511009, "global_step": 198718, "epoch": 2232, "lr": 1.674977404991731e-05}
{"train_loss": 0.08328563719987869, "global_step": 198719, "epoch": 2232, "lr": 1.6749341082054038e-05}
{"train_loss": 0.03382585942745209, "global_step": 198720, "epoch": 2232, "lr": 1.6748908118660927e-05}
{"train_loss": 0.050114259123802185, "global_step": 198721, "epoch": 2232, "lr": 1.6748475159738002e-05}
{"train_loss": 0.06666281819343567, "global_step": 198722, "epoch": 2232, "lr": 1.6748042205285358e-05}
{"train_loss": 0.07996803522109985, "global_step": 198723, "epoch": 2232, "lr": 1.674760925530302e-05}
{"train_loss": 0.040160730481147766, "global_step": 198724, "epoch": 2232, "lr": 1.6747176309791074e-05}
{"train_loss": 0.028999757021665573, "global_step": 198725, "epoch": 2232, "lr": 1.6746743368749547e-05}
{"train_loss": 0.07026100158691406, "global_step": 198726, "epoch": 2232, "lr": 1.6746310432178537e-05}
{"train_loss": 0.11349712312221527, "global_step": 198727, "epoch": 2232, "lr": 1.6745877500078056e-05}
{"train_loss": 0.05854581668972969, "global_step": 198728, "epoch": 2232, "lr": 1.674544457244821e-05}
{"train_loss": 0.09000381827354431, "global_step": 198729, "epoch": 2232, "lr": 1.674501164928901e-05}
{"train_loss": 0.04373609274625778, "global_step": 198730, "epoch": 2232, "lr": 1.6744578730600563e-05}
{"train_loss": 0.08280714601278305, "global_step": 198731, "epoch": 2232, "lr": 1.6744145816382878e-05}
{"train_loss": 0.07087116688489914, "global_step": 198732, "epoch": 2232, "lr": 1.6743712906636043e-05}
{"train_loss": 0.09110726416110992, "global_step": 198733, "epoch": 2232, "lr": 1.6743280001360123e-05}
{"train_loss": 0.05341228097677231, "global_step": 198734, "epoch": 2232, "lr": 1.6742847100555147e-05}
{"train_loss": 0.06457079946994781, "global_step": 198735, "epoch": 2232, "lr": 1.6742414204221208e-05}
{"train_loss": 0.06649344994194722, "global_step": 198736, "epoch": 2232, "lr": 1.674198131235833e-05, "val_loss": 8.232812881469727}
{"train_loss": 0.024019213393330574, "global_step": 198737, "epoch": 2233, "lr": 1.67415484249666e-05}
{"train_loss": 0.05497835576534271, "global_step": 198738, "epoch": 2233, "lr": 1.6741115542046055e-05}
{"train_loss": 0.07131994515657425, "global_step": 198739, "epoch": 2233, "lr": 1.6740682663596773e-05}
{"train_loss": 0.08102530241012573, "global_step": 198740, "epoch": 2233, "lr": 1.6740249789618784e-05}
{"train_loss": 0.05113128945231438, "global_step": 198741, "epoch": 2233, "lr": 1.6739816920112177e-05}
{"train_loss": 0.09528589993715286, "global_step": 198742, "epoch": 2233, "lr": 1.6739384055076983e-05}
{"train_loss": 0.09646683186292648, "global_step": 198743, "epoch": 2233, "lr": 1.6738951194513293e-05}
{"train_loss": 0.054468512535095215, "global_step": 198744, "epoch": 2233, "lr": 1.6738518338421126e-05}
{"train_loss": 0.07628588378429413, "global_step": 198745, "epoch": 2233, "lr": 1.6738085486800576e-05}
{"train_loss": 0.08035165816545486, "global_step": 198746, "epoch": 2233, "lr": 1.6737652639651668e-05}
{"train_loss": 0.05767804756760597, "global_step": 198747, "epoch": 2233, "lr": 1.6737219796974494e-05}
{"train_loss": 0.0644233450293541, "global_step": 198748, "epoch": 2233, "lr": 1.673678695876908e-05}
{"train_loss": 0.04499402269721031, "global_step": 198749, "epoch": 2233, "lr": 1.6736354125035513e-05}
{"train_loss": 0.07265940308570862, "global_step": 198750, "epoch": 2233, "lr": 1.6735921295773826e-05}
{"train_loss": 0.06458273530006409, "global_step": 198751, "epoch": 2233, "lr": 1.6735488470984105e-05}
{"train_loss": 0.049687422811985016, "global_step": 198752, "epoch": 2233, "lr": 1.6735055650666372e-05}
{"train_loss": 0.09802082180976868, "global_step": 198753, "epoch": 2233, "lr": 1.6734622834820722e-05}
{"train_loss": 0.09059449285268784, "global_step": 198754, "epoch": 2233, "lr": 1.67341900234472e-05}
{"train_loss": 0.061202358454465866, "global_step": 198755, "epoch": 2233, "lr": 1.673375721654584e-05}
{"train_loss": 0.03822494298219681, "global_step": 198756, "epoch": 2233, "lr": 1.6733324414116743e-05}
{"train_loss": 0.02513190731406212, "global_step": 198757, "epoch": 2233, "lr": 1.6732891616159924e-05}
{"train_loss": 0.0706230029463768, "global_step": 198758, "epoch": 2233, "lr": 1.673245882267548e-05}
{"train_loss": 0.058995798230171204, "global_step": 198759, "epoch": 2233, "lr": 1.6732026033663433e-05}
{"train_loss": 0.08759365975856781, "global_step": 198760, "epoch": 2233, "lr": 1.6731593249123877e-05}
{"train_loss": 0.05906795337796211, "global_step": 198761, "epoch": 2233, "lr": 1.6731160469056838e-05}
{"train_loss": 0.07191719114780426, "global_step": 198762, "epoch": 2233, "lr": 1.6730727693462407e-05}
{"train_loss": 0.05575341358780861, "global_step": 198763, "epoch": 2233, "lr": 1.6730294922340605e-05}
{"train_loss": 0.042042121291160583, "global_step": 198764, "epoch": 2233, "lr": 1.6729862155691527e-05}
{"train_loss": 0.03862255439162254, "global_step": 198765, "epoch": 2233, "lr": 1.6729429393515196e-05}
{"train_loss": 0.04228643700480461, "global_step": 198766, "epoch": 2233, "lr": 1.6728996635811707e-05}
{"train_loss": 0.04061738774180412, "global_step": 198767, "epoch": 2233, "lr": 1.6728563882581077e-05}
{"train_loss": 0.06485313922166824, "global_step": 198768, "epoch": 2233, "lr": 1.6728131133823387e-05}
{"train_loss": 0.07815313339233398, "global_step": 198769, "epoch": 2233, "lr": 1.6727698389538716e-05}
{"train_loss": 0.08907192945480347, "global_step": 198770, "epoch": 2233, "lr": 1.672726564972708e-05}
{"train_loss": 0.04476790875196457, "global_step": 198771, "epoch": 2233, "lr": 1.6726832914388575e-05}
{"train_loss": 0.05725125968456268, "global_step": 198772, "epoch": 2233, "lr": 1.6726400183523227e-05}
{"train_loss": 0.06422434747219086, "global_step": 198773, "epoch": 2233, "lr": 1.672596745713112e-05}
{"train_loss": 0.04604043439030647, "global_step": 198774, "epoch": 2233, "lr": 1.672553473521229e-05}
{"train_loss": 0.06038959324359894, "global_step": 198775, "epoch": 2233, "lr": 1.672510201776682e-05}
{"train_loss": 0.041946202516555786, "global_step": 198776, "epoch": 2233, "lr": 1.6724669304794738e-05}
{"train_loss": 0.04719126224517822, "global_step": 198777, "epoch": 2233, "lr": 1.6724236596296138e-05}
{"train_loss": 0.05140058323740959, "global_step": 198778, "epoch": 2233, "lr": 1.6723803892271038e-05}
{"train_loss": 0.054839812219142914, "global_step": 198779, "epoch": 2233, "lr": 1.672337119271954e-05}
{"train_loss": 0.06118157133460045, "global_step": 198780, "epoch": 2233, "lr": 1.6722938497641654e-05}
{"train_loss": 0.030891159549355507, "global_step": 198781, "epoch": 2233, "lr": 1.6722505807037486e-05}
{"train_loss": 0.04195022955536842, "global_step": 198782, "epoch": 2233, "lr": 1.6722073120907055e-05}
{"train_loss": 0.03142344951629639, "global_step": 198783, "epoch": 2233, "lr": 1.672164043925045e-05}
{"train_loss": 0.057541199028491974, "global_step": 198784, "epoch": 2233, "lr": 1.6721207762067697e-05}
{"train_loss": 0.03049895353615284, "global_step": 198785, "epoch": 2233, "lr": 1.6720775089358892e-05}
{"train_loss": 0.04348715767264366, "global_step": 198786, "epoch": 2233, "lr": 1.672034242112405e-05}
{"train_loss": 0.1121290922164917, "global_step": 198787, "epoch": 2233, "lr": 1.671990975736328e-05}
{"train_loss": 0.07168903201818466, "global_step": 198788, "epoch": 2233, "lr": 1.6719477098076585e-05}
{"train_loss": 0.04235931113362312, "global_step": 198789, "epoch": 2233, "lr": 1.6719044443264075e-05}
{"train_loss": 0.05197451263666153, "global_step": 198790, "epoch": 2233, "lr": 1.6718611792925766e-05}
{"train_loss": 0.06293301284313202, "global_step": 198791, "epoch": 2233, "lr": 1.6718179147061746e-05}
{"train_loss": 0.03054346889257431, "global_step": 198792, "epoch": 2233, "lr": 1.6717746505672066e-05}
{"train_loss": 0.024168072268366814, "global_step": 198793, "epoch": 2233, "lr": 1.6717313868756763e-05}
{"train_loss": 0.07617621123790741, "global_step": 198794, "epoch": 2233, "lr": 1.6716881236315925e-05}
{"train_loss": 0.06986276060342789, "global_step": 198795, "epoch": 2233, "lr": 1.671644860834958e-05}
{"train_loss": 0.15780006349086761, "global_step": 198796, "epoch": 2233, "lr": 1.671601598485782e-05}
{"train_loss": 0.046304672956466675, "global_step": 198797, "epoch": 2233, "lr": 1.671558336584067e-05}
{"train_loss": 0.04697950929403305, "global_step": 198798, "epoch": 2233, "lr": 1.671515075129822e-05}
{"train_loss": 0.05265869200229645, "global_step": 198799, "epoch": 2233, "lr": 1.6714718141230496e-05}
{"train_loss": 0.08730906248092651, "global_step": 198800, "epoch": 2233, "lr": 1.671428553563759e-05}
{"train_loss": 0.04235332831740379, "global_step": 198801, "epoch": 2233, "lr": 1.6713852934519526e-05}
{"train_loss": 0.02836042456328869, "global_step": 198802, "epoch": 2233, "lr": 1.6713420337876394e-05}
{"train_loss": 0.08876407891511917, "global_step": 198803, "epoch": 2233, "lr": 1.671298774570822e-05}
{"train_loss": 0.06446949392557144, "global_step": 198804, "epoch": 2233, "lr": 1.671255515801508e-05}
{"train_loss": 0.11011162400245667, "global_step": 198805, "epoch": 2233, "lr": 1.6712122574797045e-05}
{"train_loss": 0.11465010046958923, "global_step": 198806, "epoch": 2233, "lr": 1.671168999605415e-05}
{"train_loss": 0.08217909187078476, "global_step": 198807, "epoch": 2233, "lr": 1.6711257421786476e-05}
{"train_loss": 0.01856510154902935, "global_step": 198808, "epoch": 2233, "lr": 1.6710824851994045e-05}
{"train_loss": 0.0346909761428833, "global_step": 198809, "epoch": 2233, "lr": 1.6710392286676963e-05}
{"train_loss": 0.03303658962249756, "global_step": 198810, "epoch": 2233, "lr": 1.670995972583524e-05}
{"train_loss": 0.06407774239778519, "global_step": 198811, "epoch": 2233, "lr": 1.670952716946898e-05}
{"train_loss": 0.04817899689078331, "global_step": 198812, "epoch": 2233, "lr": 1.6709094617578196e-05}
{"train_loss": 0.028030145913362503, "global_step": 198813, "epoch": 2233, "lr": 1.670866207016299e-05}
{"train_loss": 0.024257123470306396, "global_step": 198814, "epoch": 2233, "lr": 1.670822952722338e-05}
{"train_loss": 0.03534385561943054, "global_step": 198815, "epoch": 2233, "lr": 1.6707796988759454e-05}
{"train_loss": 0.1011861190199852, "global_step": 198816, "epoch": 2233, "lr": 1.6707364454771248e-05}
{"train_loss": 0.0986960306763649, "global_step": 198817, "epoch": 2233, "lr": 1.6706931925258845e-05}
{"train_loss": 0.026603246107697487, "global_step": 198818, "epoch": 2233, "lr": 1.6706499400222275e-05}
{"train_loss": 0.0743088498711586, "global_step": 198819, "epoch": 2233, "lr": 1.670606687966163e-05}
{"train_loss": 0.07933082431554794, "global_step": 198820, "epoch": 2233, "lr": 1.6705634363576927e-05}
{"train_loss": 0.05919954180717468, "global_step": 198821, "epoch": 2233, "lr": 1.670520185196826e-05}
{"train_loss": 0.08833824098110199, "global_step": 198822, "epoch": 2233, "lr": 1.6704769344835663e-05}
{"train_loss": 0.061175715178251266, "global_step": 198823, "epoch": 2233, "lr": 1.67043368421792e-05}
{"train_loss": 0.08752993494272232, "global_step": 198824, "epoch": 2233, "lr": 1.670390434399895e-05}
{"train_loss": 0.060663843485579065, "global_step": 198825, "epoch": 2233, "lr": 1.6703471850294953e-05, "val_loss": 8.254706382751465}
{"train_loss": 0.0750187411904335, "global_step": 198826, "epoch": 2234, "lr": 1.6703039361067258e-05}
{"train_loss": 0.10038988292217255, "global_step": 198827, "epoch": 2234, "lr": 1.670260687631594e-05}
{"train_loss": 0.1268080174922943, "global_step": 198828, "epoch": 2234, "lr": 1.6702174396041042e-05}
{"train_loss": 0.07107435166835785, "global_step": 198829, "epoch": 2234, "lr": 1.6701741920242647e-05}
{"train_loss": 0.033357035368680954, "global_step": 198830, "epoch": 2234, "lr": 1.6701309448920794e-05}
{"train_loss": 0.053612783551216125, "global_step": 198831, "epoch": 2234, "lr": 1.6700876982075524e-05}
{"train_loss": 0.03487454727292061, "global_step": 198832, "epoch": 2234, "lr": 1.670044451970694e-05}
{"train_loss": 0.04023129492998123, "global_step": 198833, "epoch": 2234, "lr": 1.6700012061815057e-05}
{"train_loss": 0.033870305866003036, "global_step": 198834, "epoch": 2234, "lr": 1.6699579608399962e-05}
{"train_loss": 0.05667565390467644, "global_step": 198835, "epoch": 2234, "lr": 1.6699147159461694e-05}
{"train_loss": 0.051262397319078445, "global_step": 198836, "epoch": 2234, "lr": 1.669871471500033e-05}
{"train_loss": 0.06208641827106476, "global_step": 198837, "epoch": 2234, "lr": 1.6698282275015904e-05}
{"train_loss": 0.05800531059503555, "global_step": 198838, "epoch": 2234, "lr": 1.6697849839508505e-05}
{"train_loss": 0.06882213056087494, "global_step": 198839, "epoch": 2234, "lr": 1.6697417408478156e-05}
{"train_loss": 0.07341690361499786, "global_step": 198840, "epoch": 2234, "lr": 1.6696984981924952e-05}
{"train_loss": 0.04368980973958969, "global_step": 198841, "epoch": 2234, "lr": 1.6696552559848917e-05}
{"train_loss": 0.035502444952726364, "global_step": 198842, "epoch": 2234, "lr": 1.6696120142250123e-05}
{"train_loss": 0.06059786304831505, "global_step": 198843, "epoch": 2234, "lr": 1.6695687729128646e-05}
{"train_loss": 0.0659337192773819, "global_step": 198844, "epoch": 2234, "lr": 1.669525532048451e-05}
{"train_loss": 0.06628595292568207, "global_step": 198845, "epoch": 2234, "lr": 1.6694822916317805e-05}
{"train_loss": 0.07040324062108994, "global_step": 198846, "epoch": 2234, "lr": 1.669439051662856e-05}
{"train_loss": 0.05273144319653511, "global_step": 198847, "epoch": 2234, "lr": 1.669395812141687e-05}
{"train_loss": 0.04254515469074249, "global_step": 198848, "epoch": 2234, "lr": 1.669352573068275e-05}
{"train_loss": 0.0958264023065567, "global_step": 198849, "epoch": 2234, "lr": 1.6693093344426302e-05}
{"train_loss": 0.09872638434171677, "global_step": 198850, "epoch": 2234, "lr": 1.6692660962647538e-05}
{"train_loss": 0.059239551424980164, "global_step": 198851, "epoch": 2234, "lr": 1.669222858534656e-05}
{"train_loss": 0.06656002998352051, "global_step": 198852, "epoch": 2234, "lr": 1.6691796212523393e-05}
{"train_loss": 0.06389284133911133, "global_step": 198853, "epoch": 2234, "lr": 1.6691363844178114e-05}
{"train_loss": 0.08300945162773132, "global_step": 198854, "epoch": 2234, "lr": 1.6690931480310766e-05}
{"train_loss": 0.0950174331665039, "global_step": 198855, "epoch": 2234, "lr": 1.6690499120921434e-05}
{"train_loss": 0.08174275606870651, "global_step": 198856, "epoch": 2234, "lr": 1.6690066766010138e-05}
{"train_loss": 0.05724991485476494, "global_step": 198857, "epoch": 2234, "lr": 1.6689634415576976e-05}
{"train_loss": 0.056596118956804276, "global_step": 198858, "epoch": 2234, "lr": 1.6689202069621966e-05}
{"train_loss": 0.030352987349033356, "global_step": 198859, "epoch": 2234, "lr": 1.6688769728145192e-05}
{"train_loss": 0.04922904446721077, "global_step": 198860, "epoch": 2234, "lr": 1.668833739114672e-05}
{"train_loss": 0.09380386769771576, "global_step": 198861, "epoch": 2234, "lr": 1.668790505862658e-05}
{"train_loss": 0.06512933224439621, "global_step": 198862, "epoch": 2234, "lr": 1.6687472730584863e-05}
{"train_loss": 0.06880002468824387, "global_step": 198863, "epoch": 2234, "lr": 1.668704040702159e-05}
{"train_loss": 0.05139989033341408, "global_step": 198864, "epoch": 2234, "lr": 1.6686608087936856e-05}
{"train_loss": 0.043330203741788864, "global_step": 198865, "epoch": 2234, "lr": 1.6686175773330697e-05}
{"train_loss": 0.07605599611997604, "global_step": 198866, "epoch": 2234, "lr": 1.668574346320317e-05}
{"train_loss": 0.036119453608989716, "global_step": 198867, "epoch": 2234, "lr": 1.6685311157554347e-05}
{"train_loss": 0.0420348159968853, "global_step": 198868, "epoch": 2234, "lr": 1.6684878856384262e-05}
{"train_loss": 0.14860539138317108, "global_step": 198869, "epoch": 2234, "lr": 1.6684446559693007e-05}
{"train_loss": 0.06790585070848465, "global_step": 198870, "epoch": 2234, "lr": 1.6684014267480618e-05}
{"train_loss": 0.028657391667366028, "global_step": 198871, "epoch": 2234, "lr": 1.668358197974715e-05}
{"train_loss": 0.03796159476041794, "global_step": 198872, "epoch": 2234, "lr": 1.668314969649268e-05}
{"train_loss": 0.07312777638435364, "global_step": 198873, "epoch": 2234, "lr": 1.668271741771723e-05}
{"train_loss": 0.06077907606959343, "global_step": 198874, "epoch": 2234, "lr": 1.6682285143420913e-05}
{"train_loss": 0.06630492955446243, "global_step": 198875, "epoch": 2234, "lr": 1.6681852873603733e-05}
{"train_loss": 0.07186190783977509, "global_step": 198876, "epoch": 2234, "lr": 1.6681420608265785e-05}
{"train_loss": 0.05108043923974037, "global_step": 198877, "epoch": 2234, "lr": 1.66809883474071e-05}
{"train_loss": 0.041743867099285126, "global_step": 198878, "epoch": 2234, "lr": 1.6680556091027755e-05}
{"train_loss": 0.06433112174272537, "global_step": 198879, "epoch": 2234, "lr": 1.6680123839127815e-05}
{"train_loss": 0.09888318181037903, "global_step": 198880, "epoch": 2234, "lr": 1.6679691591707307e-05}
{"train_loss": 0.047002192586660385, "global_step": 198881, "epoch": 2234, "lr": 1.667925934876633e-05}
{"train_loss": 0.0539168044924736, "global_step": 198882, "epoch": 2234, "lr": 1.66788271103049e-05}
{"train_loss": 0.05553123727440834, "global_step": 198883, "epoch": 2234, "lr": 1.6678394876323116e-05}
{"train_loss": 0.10685930401086807, "global_step": 198884, "epoch": 2234, "lr": 1.6677962646821e-05}
{"train_loss": 0.0902949646115303, "global_step": 198885, "epoch": 2234, "lr": 1.6677530421798636e-05}
{"train_loss": 0.07483844459056854, "global_step": 198886, "epoch": 2234, "lr": 1.667709820125606e-05}
{"train_loss": 0.07797626405954361, "global_step": 198887, "epoch": 2234, "lr": 1.667666598519336e-05}
{"train_loss": 0.02476581186056137, "global_step": 198888, "epoch": 2234, "lr": 1.6676233773610554e-05}
{"train_loss": 0.05917363241314888, "global_step": 198889, "epoch": 2234, "lr": 1.6675801566507744e-05}
{"train_loss": 0.10256446897983551, "global_step": 198890, "epoch": 2234, "lr": 1.6675369363884946e-05}
{"train_loss": 0.07194271683692932, "global_step": 198891, "epoch": 2234, "lr": 1.667493716574226e-05}
{"train_loss": 0.10732639580965042, "global_step": 198892, "epoch": 2234, "lr": 1.66745049720797e-05}
{"train_loss": 0.04959046095609665, "global_step": 198893, "epoch": 2234, "lr": 1.6674072782897364e-05}
{"train_loss": 0.06791064888238907, "global_step": 198894, "epoch": 2234, "lr": 1.667364059819528e-05}
{"train_loss": 0.06511203944683075, "global_step": 198895, "epoch": 2234, "lr": 1.6673208417973517e-05}
{"train_loss": 0.03188835456967354, "global_step": 198896, "epoch": 2234, "lr": 1.6672776242232153e-05}
{"train_loss": 0.05423992872238159, "global_step": 198897, "epoch": 2234, "lr": 1.6672344070971207e-05}
{"train_loss": 0.07164747267961502, "global_step": 198898, "epoch": 2234, "lr": 1.667191190419078e-05}
{"train_loss": 0.07436022162437439, "global_step": 198899, "epoch": 2234, "lr": 1.667147974189089e-05}
{"train_loss": 0.08585435897111893, "global_step": 198900, "epoch": 2234, "lr": 1.6671047584071625e-05}
{"train_loss": 0.06407768279314041, "global_step": 198901, "epoch": 2234, "lr": 1.6670615430733022e-05}
{"train_loss": 0.10031774640083313, "global_step": 198902, "epoch": 2234, "lr": 1.6670183281875162e-05}
{"train_loss": 0.03284904360771179, "global_step": 198903, "epoch": 2234, "lr": 1.666975113749809e-05}
{"train_loss": 0.06525416672229767, "global_step": 198904, "epoch": 2234, "lr": 1.6669318997601845e-05}
{"train_loss": 0.06270501762628555, "global_step": 198905, "epoch": 2234, "lr": 1.6668886862186523e-05}
{"train_loss": 0.07837188243865967, "global_step": 198906, "epoch": 2234, "lr": 1.6668454731252148e-05}
{"train_loss": 0.12632186710834503, "global_step": 198907, "epoch": 2234, "lr": 1.6668022604798804e-05}
{"train_loss": 0.04982980340719223, "global_step": 198908, "epoch": 2234, "lr": 1.666759048282654e-05}
{"train_loss": 0.0964570865035057, "global_step": 198909, "epoch": 2234, "lr": 1.66671583653354e-05}
{"train_loss": 0.07220770418643951, "global_step": 198910, "epoch": 2234, "lr": 1.6666726252325466e-05}
{"train_loss": 0.11958252638578415, "global_step": 198911, "epoch": 2234, "lr": 1.666629414379677e-05}
{"train_loss": 0.09116048365831375, "global_step": 198912, "epoch": 2234, "lr": 1.66658620397494e-05}
{"train_loss": 0.052195433527231216, "global_step": 198913, "epoch": 2234, "lr": 1.6665429940183387e-05}
{"train_loss": 0.06649960456102082, "global_step": 198914, "epoch": 2234, "lr": 1.666499784509879e-05, "val_loss": 8.197128295898438}
{"train_loss": 0.15446975827217102, "global_step": 198915, "epoch": 2235, "lr": 1.6664565754495703e-05}
{"train_loss": 0.0735241174697876, "global_step": 198916, "epoch": 2235, "lr": 1.6664133668374138e-05}
{"train_loss": 0.07653272151947021, "global_step": 198917, "epoch": 2235, "lr": 1.6663701586734194e-05}
{"train_loss": 0.0751219093799591, "global_step": 198918, "epoch": 2235, "lr": 1.666326950957589e-05}
{"train_loss": 0.06530638039112091, "global_step": 198919, "epoch": 2235, "lr": 1.666283743689932e-05}
{"train_loss": 0.07414234429597855, "global_step": 198920, "epoch": 2235, "lr": 1.666240536870451e-05}
{"train_loss": 0.09647722542285919, "global_step": 198921, "epoch": 2235, "lr": 1.6661973304991553e-05}
{"train_loss": 0.06118664890527725, "global_step": 198922, "epoch": 2235, "lr": 1.6661541245760465e-05}
{"train_loss": 0.08183182030916214, "global_step": 198923, "epoch": 2235, "lr": 1.6661109191011344e-05}
{"train_loss": 0.04023795574903488, "global_step": 198924, "epoch": 2235, "lr": 1.6660677140744213e-05}
{"train_loss": 0.06298813968896866, "global_step": 198925, "epoch": 2235, "lr": 1.6660245094959166e-05}
{"train_loss": 0.08469876646995544, "global_step": 198926, "epoch": 2235, "lr": 1.6659813053656225e-05}
{"train_loss": 0.12643863260746002, "global_step": 198927, "epoch": 2235, "lr": 1.6659381016835484e-05}
{"train_loss": 0.07310985773801804, "global_step": 198928, "epoch": 2235, "lr": 1.6658948984496964e-05}
{"train_loss": 0.10741335153579712, "global_step": 198929, "epoch": 2235, "lr": 1.6658516956640763e-05}
{"train_loss": 0.09461822360754013, "global_step": 198930, "epoch": 2235, "lr": 1.66580849332669e-05}
{"train_loss": 0.04537961259484291, "global_step": 198931, "epoch": 2235, "lr": 1.6657652914375448e-05}
{"train_loss": 0.05048160254955292, "global_step": 198932, "epoch": 2235, "lr": 1.6657220899966488e-05}
{"train_loss": 0.02592284418642521, "global_step": 198933, "epoch": 2235, "lr": 1.6656788890040035e-05}
{"train_loss": 0.0828273743391037, "global_step": 198934, "epoch": 2235, "lr": 1.665635688459619e-05}
{"train_loss": 0.08333912491798401, "global_step": 198935, "epoch": 2235, "lr": 1.6655924883634982e-05}
{"train_loss": 0.10044769942760468, "global_step": 198936, "epoch": 2235, "lr": 1.6655492887156486e-05}
{"train_loss": 0.09424077719449997, "global_step": 198937, "epoch": 2235, "lr": 1.665506089516074e-05}
{"train_loss": 0.033238865435123444, "global_step": 198938, "epoch": 2235, "lr": 1.6654628907647836e-05}
{"train_loss": 0.11604728549718857, "global_step": 198939, "epoch": 2235, "lr": 1.6654196924617787e-05}
{"train_loss": 0.05053115636110306, "global_step": 198940, "epoch": 2235, "lr": 1.6653764946070694e-05}
{"train_loss": 0.09548080712556839, "global_step": 198941, "epoch": 2235, "lr": 1.6653332972006597e-05}
{"train_loss": 0.07363738864660263, "global_step": 198942, "epoch": 2235, "lr": 1.665290100242553e-05}
{"train_loss": 0.060040876269340515, "global_step": 198943, "epoch": 2235, "lr": 1.6652469037327596e-05}
{"train_loss": 0.06068599969148636, "global_step": 198944, "epoch": 2235, "lr": 1.6652037076712817e-05}
{"train_loss": 0.044416073709726334, "global_step": 198945, "epoch": 2235, "lr": 1.6651605120581275e-05}
{"train_loss": 0.05798416957259178, "global_step": 198946, "epoch": 2235, "lr": 1.665117316893301e-05}
{"train_loss": 0.11067092418670654, "global_step": 198947, "epoch": 2235, "lr": 1.6650741221768095e-05}
{"train_loss": 0.07447700947523117, "global_step": 198948, "epoch": 2235, "lr": 1.665030927908659e-05}
{"train_loss": 0.07820301502943039, "global_step": 198949, "epoch": 2235, "lr": 1.6649877340888527e-05}
{"train_loss": 0.11662815511226654, "global_step": 198950, "epoch": 2235, "lr": 1.6649445407173976e-05}
{"train_loss": 0.041492559015750885, "global_step": 198951, "epoch": 2235, "lr": 1.664901347794302e-05}
{"train_loss": 0.05623254179954529, "global_step": 198952, "epoch": 2235, "lr": 1.664858155319568e-05}
{"train_loss": 0.0896320641040802, "global_step": 198953, "epoch": 2235, "lr": 1.6648149632932046e-05}
{"train_loss": 0.0730019137263298, "global_step": 198954, "epoch": 2235, "lr": 1.664771771715215e-05}
{"train_loss": 0.06701583415269852, "global_step": 198955, "epoch": 2235, "lr": 1.664728580585608e-05}
{"train_loss": 0.03780674561858177, "global_step": 198956, "epoch": 2235, "lr": 1.6646853899043856e-05}
{"train_loss": 0.053229156881570816, "global_step": 198957, "epoch": 2235, "lr": 1.6646421996715573e-05}
{"train_loss": 0.1276419460773468, "global_step": 198958, "epoch": 2235, "lr": 1.6645990098871255e-05}
{"train_loss": 0.07971464842557907, "global_step": 198959, "epoch": 2235, "lr": 1.6645558205510993e-05}
{"train_loss": 0.03223785385489464, "global_step": 198960, "epoch": 2235, "lr": 1.6645126316634812e-05}
{"train_loss": 0.05969848483800888, "global_step": 198961, "epoch": 2235, "lr": 1.6644694432242807e-05}
{"train_loss": 0.11228121817111969, "global_step": 198962, "epoch": 2235, "lr": 1.6644262552334998e-05}
{"train_loss": 0.0852968841791153, "global_step": 198963, "epoch": 2235, "lr": 1.6643830676911483e-05}
{"train_loss": 0.09460566937923431, "global_step": 198964, "epoch": 2235, "lr": 1.664339880597227e-05}
{"train_loss": 0.1206643283367157, "global_step": 198965, "epoch": 2235, "lr": 1.6642966939517473e-05}
{"train_loss": 0.035611264407634735, "global_step": 198966, "epoch": 2235, "lr": 1.66425350775471e-05}
{"train_loss": 0.10180284082889557, "global_step": 198967, "epoch": 2235, "lr": 1.664210322006125e-05}
{"train_loss": 0.04315419867634773, "global_step": 198968, "epoch": 2235, "lr": 1.6641671367059946e-05}
{"train_loss": 0.04747069999575615, "global_step": 198969, "epoch": 2235, "lr": 1.6641239518543266e-05}
{"train_loss": 0.050167493522167206, "global_step": 198970, "epoch": 2235, "lr": 1.6640807674511276e-05}
{"train_loss": 0.0572589635848999, "global_step": 198971, "epoch": 2235, "lr": 1.664037583496401e-05}
{"train_loss": 0.06501935422420502, "global_step": 198972, "epoch": 2235, "lr": 1.663994399990155e-05}
{"train_loss": 0.08290419727563858, "global_step": 198973, "epoch": 2235, "lr": 1.6639512169323934e-05}
{"train_loss": 0.037984006106853485, "global_step": 198974, "epoch": 2235, "lr": 1.663908034323124e-05}
{"train_loss": 0.034813348203897476, "global_step": 198975, "epoch": 2235, "lr": 1.66386485216235e-05}
{"train_loss": 0.041514765471220016, "global_step": 198976, "epoch": 2235, "lr": 1.663821670450081e-05}
{"train_loss": 0.07968989759683609, "global_step": 198977, "epoch": 2235, "lr": 1.6637784891863177e-05}
{"train_loss": 0.031135186553001404, "global_step": 198978, "epoch": 2235, "lr": 1.663735308371071e-05}
{"train_loss": 0.053130362182855606, "global_step": 198979, "epoch": 2235, "lr": 1.6636921280043428e-05}
{"train_loss": 0.07725517451763153, "global_step": 198980, "epoch": 2235, "lr": 1.6636489480861423e-05}
{"train_loss": 0.047773756086826324, "global_step": 198981, "epoch": 2235, "lr": 1.6636057686164728e-05}
{"train_loss": 0.021373791620135307, "global_step": 198982, "epoch": 2235, "lr": 1.6635625895953398e-05}
{"train_loss": 0.025193311274051666, "global_step": 198983, "epoch": 2235, "lr": 1.663519411022752e-05}
{"train_loss": 0.04416113346815109, "global_step": 198984, "epoch": 2235, "lr": 1.6634762328987114e-05}
{"train_loss": 0.10304510593414307, "global_step": 198985, "epoch": 2235, "lr": 1.6634330552232274e-05}
{"train_loss": 0.04301087185740471, "global_step": 198986, "epoch": 2235, "lr": 1.6633898779963026e-05}
{"train_loss": 0.06213745102286339, "global_step": 198987, "epoch": 2235, "lr": 1.6633467012179455e-05}
{"train_loss": 0.08757727593183517, "global_step": 198988, "epoch": 2235, "lr": 1.6633035248881594e-05}
{"train_loss": 0.08979605883359909, "global_step": 198989, "epoch": 2235, "lr": 1.663260349006954e-05}
{"train_loss": 0.0933515802025795, "global_step": 198990, "epoch": 2235, "lr": 1.6632171735743295e-05}
{"train_loss": 0.05670445039868355, "global_step": 198991, "epoch": 2235, "lr": 1.663173998590297e-05}
{"train_loss": 0.0908479169011116, "global_step": 198992, "epoch": 2235, "lr": 1.6631308240548588e-05}
{"train_loss": 0.07841658592224121, "global_step": 198993, "epoch": 2235, "lr": 1.6630876499680238e-05}
{"train_loss": 0.04983321204781532, "global_step": 198994, "epoch": 2235, "lr": 1.6630444763297936e-05}
{"train_loss": 0.03921979293227196, "global_step": 198995, "epoch": 2235, "lr": 1.6630013031401785e-05}
{"train_loss": 0.015507147647440434, "global_step": 198996, "epoch": 2235, "lr": 1.6629581303991802e-05}
{"train_loss": 0.05713948234915733, "global_step": 198997, "epoch": 2235, "lr": 1.662914958106809e-05}
{"train_loss": 0.02859446592628956, "global_step": 198998, "epoch": 2235, "lr": 1.6628717862630654e-05}
{"train_loss": 0.08137188851833344, "global_step": 198999, "epoch": 2235, "lr": 1.6628286148679605e-05}
{"train_loss": 0.075999416410923, "global_step": 199000, "epoch": 2235, "lr": 1.6627854439214957e-05}
{"train_loss": 0.07284494489431381, "global_step": 199001, "epoch": 2235, "lr": 1.66274227342368e-05}
{"train_loss": 0.0799049511551857, "global_step": 199002, "epoch": 2235, "lr": 1.662699103374517e-05}
{"train_loss": 0.06947676401083054, "global_step": 199003, "epoch": 2235, "lr": 1.6626559337740143e-05, "val_loss": 8.02747631072998, "train_action_mse_error": 19.933197021484375}
{"train_loss": 0.14143361151218414, "global_step": 199004, "epoch": 2236, "lr": 1.6626127646221758e-05}
{"train_loss": 0.06560664623975754, "global_step": 199005, "epoch": 2236, "lr": 1.6625695959190086e-05}
{"train_loss": 0.08486539125442505, "global_step": 199006, "epoch": 2236, "lr": 1.6625264276645192e-05}
{"train_loss": 0.057284094393253326, "global_step": 199007, "epoch": 2236, "lr": 1.662483259858711e-05}
{"train_loss": 0.10232537984848022, "global_step": 199008, "epoch": 2236, "lr": 1.6624400925015926e-05}
{"train_loss": 0.036176569759845734, "global_step": 199009, "epoch": 2236, "lr": 1.662396925593167e-05}
{"train_loss": 0.023085834458470345, "global_step": 199010, "epoch": 2236, "lr": 1.6623537591334438e-05}
{"train_loss": 0.03123951330780983, "global_step": 199011, "epoch": 2236, "lr": 1.662310593122424e-05}
{"train_loss": 0.041529856622219086, "global_step": 199012, "epoch": 2236, "lr": 1.662267427560118e-05}
{"train_loss": 0.0729278028011322, "global_step": 199013, "epoch": 2236, "lr": 1.6622242624465272e-05}
{"train_loss": 0.07590359449386597, "global_step": 199014, "epoch": 2236, "lr": 1.6621810977816614e-05}
{"train_loss": 0.04941084608435631, "global_step": 199015, "epoch": 2236, "lr": 1.6621379335655236e-05}
{"train_loss": 0.10704133659601212, "global_step": 199016, "epoch": 2236, "lr": 1.662094769798122e-05}
{"train_loss": 0.03880583122372627, "global_step": 199017, "epoch": 2236, "lr": 1.6620516064794593e-05}
{"train_loss": 0.0788392424583435, "global_step": 199018, "epoch": 2236, "lr": 1.662008443609545e-05}
{"train_loss": 0.09070736169815063, "global_step": 199019, "epoch": 2236, "lr": 1.661965281188383e-05}
{"train_loss": 0.16522744297981262, "global_step": 199020, "epoch": 2236, "lr": 1.6619221192159774e-05}
{"train_loss": 0.06806804239749908, "global_step": 199021, "epoch": 2236, "lr": 1.661878957692337e-05}
{"train_loss": 0.04242657870054245, "global_step": 199022, "epoch": 2236, "lr": 1.6618357966174646e-05}
{"train_loss": 0.07905025780200958, "global_step": 199023, "epoch": 2236, "lr": 1.6617926359913698e-05}
{"train_loss": 0.09356753528118134, "global_step": 199024, "epoch": 2236, "lr": 1.6617494758140544e-05}
{"train_loss": 0.02740894816815853, "global_step": 199025, "epoch": 2236, "lr": 1.661706316085528e-05}
{"train_loss": 0.0536552295088768, "global_step": 199026, "epoch": 2236, "lr": 1.661663156805792e-05}
{"train_loss": 0.05145355314016342, "global_step": 199027, "epoch": 2236, "lr": 1.6616199979748576e-05}
{"train_loss": 0.06084361672401428, "global_step": 199028, "epoch": 2236, "lr": 1.661576839592725e-05}
{"train_loss": 0.03851243108510971, "global_step": 199029, "epoch": 2236, "lr": 1.661533681659405e-05}
{"train_loss": 0.05962451174855232, "global_step": 199030, "epoch": 2236, "lr": 1.661490524174899e-05}
{"train_loss": 0.07850854843854904, "global_step": 199031, "epoch": 2236, "lr": 1.6614473671392173e-05}
{"train_loss": 0.05903436616063118, "global_step": 199032, "epoch": 2236, "lr": 1.6614042105523607e-05}
{"train_loss": 0.050522059202194214, "global_step": 199033, "epoch": 2236, "lr": 1.6613610544143398e-05}
{"train_loss": 0.0334753543138504, "global_step": 199034, "epoch": 2236, "lr": 1.6613178987251566e-05}
{"train_loss": 0.0896448940038681, "global_step": 199035, "epoch": 2236, "lr": 1.6612747434848202e-05}
{"train_loss": 0.058594346046447754, "global_step": 199036, "epoch": 2236, "lr": 1.6612315886933327e-05}
{"train_loss": 0.102337546646595, "global_step": 199037, "epoch": 2236, "lr": 1.6611884343507038e-05}
{"train_loss": 0.07232341915369034, "global_step": 199038, "epoch": 2236, "lr": 1.661145280456936e-05}
{"train_loss": 0.03879758343100548, "global_step": 199039, "epoch": 2236, "lr": 1.661102127012038e-05}
{"train_loss": 0.0852842852473259, "global_step": 199040, "epoch": 2236, "lr": 1.661058974016012e-05}
{"train_loss": 0.05304994434118271, "global_step": 199041, "epoch": 2236, "lr": 1.6610158214688663e-05}
{"train_loss": 0.1255132257938385, "global_step": 199042, "epoch": 2236, "lr": 1.6609726693706085e-05}
{"train_loss": 0.0916779562830925, "global_step": 199043, "epoch": 2236, "lr": 1.6609295177212398e-05}
{"train_loss": 0.034364935010671616, "global_step": 199044, "epoch": 2236, "lr": 1.66088636652077e-05}
{"train_loss": 0.04183521121740341, "global_step": 199045, "epoch": 2236, "lr": 1.660843215769202e-05}
{"train_loss": 0.09963193535804749, "global_step": 199046, "epoch": 2236, "lr": 1.660800065466545e-05}
{"train_loss": 0.03157941251993179, "global_step": 199047, "epoch": 2236, "lr": 1.6607569156128005e-05}
{"train_loss": 0.06544694304466248, "global_step": 199048, "epoch": 2236, "lr": 1.6607137662079787e-05}
{"train_loss": 0.04255876690149307, "global_step": 199049, "epoch": 2236, "lr": 1.6606706172520812e-05}
{"train_loss": 0.10594093799591064, "global_step": 199050, "epoch": 2236, "lr": 1.6606274687451174e-05}
{"train_loss": 0.09562503546476364, "global_step": 199051, "epoch": 2236, "lr": 1.66058432068709e-05}
{"train_loss": 0.0630473792552948, "global_step": 199052, "epoch": 2236, "lr": 1.660541173078008e-05}
{"train_loss": 0.049828700721263885, "global_step": 199053, "epoch": 2236, "lr": 1.6604980259178742e-05}
{"train_loss": 0.05412773787975311, "global_step": 199054, "epoch": 2236, "lr": 1.660454879206697e-05}
{"train_loss": 0.06598518043756485, "global_step": 199055, "epoch": 2236, "lr": 1.6604117329444797e-05}
{"train_loss": 0.0637025386095047, "global_step": 199056, "epoch": 2236, "lr": 1.6603685871312304e-05}
{"train_loss": 0.07147171348333359, "global_step": 199057, "epoch": 2236, "lr": 1.660325441766954e-05}
{"train_loss": 0.04005642980337143, "global_step": 199058, "epoch": 2236, "lr": 1.6602822968516547e-05}
{"train_loss": 0.03280831500887871, "global_step": 199059, "epoch": 2236, "lr": 1.6602391523853412e-05}
{"train_loss": 0.059691090136766434, "global_step": 199060, "epoch": 2236, "lr": 1.6601960083680162e-05}
{"train_loss": 0.04401005059480667, "global_step": 199061, "epoch": 2236, "lr": 1.6601528647996888e-05}
{"train_loss": 0.06228528171777725, "global_step": 199062, "epoch": 2236, "lr": 1.6601097216803617e-05}
{"train_loss": 0.06914560496807098, "global_step": 199063, "epoch": 2236, "lr": 1.660066579010044e-05}
{"train_loss": 0.06148035079240799, "global_step": 199064, "epoch": 2236, "lr": 1.6600234367887375e-05}
{"train_loss": 0.08933722972869873, "global_step": 199065, "epoch": 2236, "lr": 1.6599802950164516e-05}
{"train_loss": 0.027596119791269302, "global_step": 199066, "epoch": 2236, "lr": 1.6599371536931895e-05}
{"train_loss": 0.038483746349811554, "global_step": 199067, "epoch": 2236, "lr": 1.65989401281896e-05}
{"train_loss": 0.06848293542861938, "global_step": 199068, "epoch": 2236, "lr": 1.6598508723937646e-05}
{"train_loss": 0.03952661529183388, "global_step": 199069, "epoch": 2236, "lr": 1.6598077324176138e-05}
{"train_loss": 0.04194309189915657, "global_step": 199070, "epoch": 2236, "lr": 1.6597645928905094e-05}
{"train_loss": 0.10513865947723389, "global_step": 199071, "epoch": 2236, "lr": 1.65972145381246e-05}
{"train_loss": 0.07430470734834671, "global_step": 199072, "epoch": 2236, "lr": 1.6596783151834695e-05}
{"train_loss": 0.1130075678229332, "global_step": 199073, "epoch": 2236, "lr": 1.6596351770035455e-05}
{"train_loss": 0.04579533264040947, "global_step": 199074, "epoch": 2236, "lr": 1.6595920392726917e-05}
{"train_loss": 0.07506135106086731, "global_step": 199075, "epoch": 2236, "lr": 1.6595489019909168e-05}
{"train_loss": 0.06720363348722458, "global_step": 199076, "epoch": 2236, "lr": 1.659505765158223e-05}
{"train_loss": 0.07791193574666977, "global_step": 199077, "epoch": 2236, "lr": 1.659462628774618e-05}
{"train_loss": 0.05777589604258537, "global_step": 199078, "epoch": 2236, "lr": 1.6594194928401085e-05}
{"train_loss": 0.07715372741222382, "global_step": 199079, "epoch": 2236, "lr": 1.6593763573546984e-05}
{"train_loss": 0.059727855026721954, "global_step": 199080, "epoch": 2236, "lr": 1.659333222318396e-05}
{"train_loss": 0.061748769134283066, "global_step": 199081, "epoch": 2236, "lr": 1.6592900877312036e-05}
{"train_loss": 0.08679893612861633, "global_step": 199082, "epoch": 2236, "lr": 1.659246953593131e-05}
{"train_loss": 0.05978968366980553, "global_step": 199083, "epoch": 2236, "lr": 1.65920381990418e-05}
{"train_loss": 0.08322716504335403, "global_step": 199084, "epoch": 2236, "lr": 1.65916068666436e-05}
{"train_loss": 0.0741845965385437, "global_step": 199085, "epoch": 2236, "lr": 1.6591175538736732e-05}
{"train_loss": 0.061922065913677216, "global_step": 199086, "epoch": 2236, "lr": 1.6590744215321295e-05}
{"train_loss": 0.08521003276109695, "global_step": 199087, "epoch": 2236, "lr": 1.6590312896397305e-05}
{"train_loss": 0.04923415929079056, "global_step": 199088, "epoch": 2236, "lr": 1.6589881581964862e-05}
{"train_loss": 0.04590243473649025, "global_step": 199089, "epoch": 2236, "lr": 1.6589450272023978e-05}
{"train_loss": 0.05415230989456177, "global_step": 199090, "epoch": 2236, "lr": 1.6589018966574754e-05}
{"train_loss": 0.0383383184671402, "global_step": 199091, "epoch": 2236, "lr": 1.6588587665617216e-05}
{"train_loss": 0.06532105050060187, "global_step": 199092, "epoch": 2236, "lr": 1.6588156369151447e-05, "val_loss": 8.002809524536133}
{"train_loss": 0.05293857678771019, "global_step": 199093, "epoch": 2237, "lr": 1.6587725077177484e-05}
{"train_loss": 0.10615450888872147, "global_step": 199094, "epoch": 2237, "lr": 1.65872937896954e-05}
{"train_loss": 0.11106117069721222, "global_step": 199095, "epoch": 2237, "lr": 1.658686250670524e-05}
{"train_loss": 0.07660818099975586, "global_step": 199096, "epoch": 2237, "lr": 1.6586431228207084e-05}
{"train_loss": 0.0867762342095375, "global_step": 199097, "epoch": 2237, "lr": 1.658599995420097e-05}
{"train_loss": 0.07521764189004898, "global_step": 199098, "epoch": 2237, "lr": 1.6585568684686952e-05}
{"train_loss": 0.0570431612432003, "global_step": 199099, "epoch": 2237, "lr": 1.658513741966511e-05}
{"train_loss": 0.06060338765382767, "global_step": 199100, "epoch": 2237, "lr": 1.6584706159135472e-05}
{"train_loss": 0.06419268250465393, "global_step": 199101, "epoch": 2237, "lr": 1.6584274903098128e-05}
{"train_loss": 0.03754783421754837, "global_step": 199102, "epoch": 2237, "lr": 1.6583843651553104e-05}
{"train_loss": 0.03806116804480553, "global_step": 199103, "epoch": 2237, "lr": 1.6583412404500493e-05}
{"train_loss": 0.01922101527452469, "global_step": 199104, "epoch": 2237, "lr": 1.658298116194032e-05}
{"train_loss": 0.05587229132652283, "global_step": 199105, "epoch": 2237, "lr": 1.6582549923872676e-05}
{"train_loss": 0.05098904296755791, "global_step": 199106, "epoch": 2237, "lr": 1.6582118690297575e-05}
{"train_loss": 0.05488758534193039, "global_step": 199107, "epoch": 2237, "lr": 1.6581687461215127e-05}
{"train_loss": 0.07062657177448273, "global_step": 199108, "epoch": 2237, "lr": 1.658125623662535e-05}
{"train_loss": 0.06202153488993645, "global_step": 199109, "epoch": 2237, "lr": 1.6580825016528324e-05}
{"train_loss": 0.12382517009973526, "global_step": 199110, "epoch": 2237, "lr": 1.658039380092408e-05}
{"train_loss": 0.07459214329719543, "global_step": 199111, "epoch": 2237, "lr": 1.657996258981272e-05}
{"train_loss": 0.06072581186890602, "global_step": 199112, "epoch": 2237, "lr": 1.6579531383194254e-05}
{"train_loss": 0.08057309687137604, "global_step": 199113, "epoch": 2237, "lr": 1.6579100181068765e-05}
{"train_loss": 0.05552070960402489, "global_step": 199114, "epoch": 2237, "lr": 1.6578668983436324e-05}
{"train_loss": 0.04958729073405266, "global_step": 199115, "epoch": 2237, "lr": 1.6578237790296957e-05}
{"train_loss": 0.03580160811543465, "global_step": 199116, "epoch": 2237, "lr": 1.6577806601650757e-05}
{"train_loss": 0.09849020838737488, "global_step": 199117, "epoch": 2237, "lr": 1.657737541749775e-05}
{"train_loss": 0.09557099640369415, "global_step": 199118, "epoch": 2237, "lr": 1.6576944237838016e-05}
{"train_loss": 0.09482534229755402, "global_step": 199119, "epoch": 2237, "lr": 1.6576513062671594e-05}
{"train_loss": 0.04536525160074234, "global_step": 199120, "epoch": 2237, "lr": 1.657608189199857e-05}
{"train_loss": 0.07440289855003357, "global_step": 199121, "epoch": 2237, "lr": 1.657565072581896e-05}
{"train_loss": 0.07741520553827286, "global_step": 199122, "epoch": 2237, "lr": 1.6575219564132873e-05}
{"train_loss": 0.03275023028254509, "global_step": 199123, "epoch": 2237, "lr": 1.6574788406940316e-05}
{"train_loss": 0.07406655699014664, "global_step": 199124, "epoch": 2237, "lr": 1.6574357254241392e-05}
{"train_loss": 0.045257940888404846, "global_step": 199125, "epoch": 2237, "lr": 1.657392610603612e-05}
{"train_loss": 0.04939820617437363, "global_step": 199126, "epoch": 2237, "lr": 1.6573494962324594e-05}
{"train_loss": 0.08446117490530014, "global_step": 199127, "epoch": 2237, "lr": 1.6573063823106832e-05}
{"train_loss": 0.05700967088341713, "global_step": 199128, "epoch": 2237, "lr": 1.657263268838294e-05}
{"train_loss": 0.06756974756717682, "global_step": 199129, "epoch": 2237, "lr": 1.6572201558152926e-05}
{"train_loss": 0.04799278825521469, "global_step": 199130, "epoch": 2237, "lr": 1.657177043241689e-05}
{"train_loss": 0.064018614590168, "global_step": 199131, "epoch": 2237, "lr": 1.657133931117486e-05}
{"train_loss": 0.026996787637472153, "global_step": 199132, "epoch": 2237, "lr": 1.6570908194426916e-05}
{"train_loss": 0.02580706775188446, "global_step": 199133, "epoch": 2237, "lr": 1.6570477082173096e-05}
{"train_loss": 0.0755871906876564, "global_step": 199134, "epoch": 2237, "lr": 1.6570045974413477e-05}
{"train_loss": 0.08394846320152283, "global_step": 199135, "epoch": 2237, "lr": 1.6569614871148114e-05}
{"train_loss": 0.04141176491975784, "global_step": 199136, "epoch": 2237, "lr": 1.656918377237704e-05}
{"train_loss": 0.014492827467620373, "global_step": 199137, "epoch": 2237, "lr": 1.6568752678100352e-05}
{"train_loss": 0.08663936704397202, "global_step": 199138, "epoch": 2237, "lr": 1.6568321588318063e-05}
{"train_loss": 0.05675788223743439, "global_step": 199139, "epoch": 2237, "lr": 1.6567890503030282e-05}
{"train_loss": 0.033232275396585464, "global_step": 199140, "epoch": 2237, "lr": 1.6567459422237018e-05}
{"train_loss": 0.04278040677309036, "global_step": 199141, "epoch": 2237, "lr": 1.6567028345938368e-05}
{"train_loss": 0.10574151575565338, "global_step": 199142, "epoch": 2237, "lr": 1.656659727413436e-05}
{"train_loss": 0.02857108786702156, "global_step": 199143, "epoch": 2237, "lr": 1.6566166206825077e-05}
{"train_loss": 0.065457783639431, "global_step": 199144, "epoch": 2237, "lr": 1.656573514401056e-05}
{"train_loss": 0.022554518654942513, "global_step": 199145, "epoch": 2237, "lr": 1.6565304085690874e-05}
{"train_loss": 0.04470231011509895, "global_step": 199146, "epoch": 2237, "lr": 1.6564873031866068e-05}
{"train_loss": 0.052045851945877075, "global_step": 199147, "epoch": 2237, "lr": 1.6564441982536222e-05}
{"train_loss": 0.07995057106018066, "global_step": 199148, "epoch": 2237, "lr": 1.6564010937701363e-05}
{"train_loss": 0.05074674263596535, "global_step": 199149, "epoch": 2237, "lr": 1.6563579897361566e-05}
{"train_loss": 0.06513052433729172, "global_step": 199150, "epoch": 2237, "lr": 1.6563148861516904e-05}
{"train_loss": 0.06248416379094124, "global_step": 199151, "epoch": 2237, "lr": 1.6562717830167403e-05}
{"train_loss": 0.020951474085450172, "global_step": 199152, "epoch": 2237, "lr": 1.6562286803313153e-05}
{"train_loss": 0.04065794497728348, "global_step": 199153, "epoch": 2237, "lr": 1.6561855780954177e-05}
{"train_loss": 0.05414436757564545, "global_step": 199154, "epoch": 2237, "lr": 1.6561424763090572e-05}
{"train_loss": 0.03381543979048729, "global_step": 199155, "epoch": 2237, "lr": 1.656099374972236e-05}
{"train_loss": 0.023438259959220886, "global_step": 199156, "epoch": 2237, "lr": 1.6560562740849627e-05}
{"train_loss": 0.04228057339787483, "global_step": 199157, "epoch": 2237, "lr": 1.6560131736472406e-05}
{"train_loss": 0.02451189234852791, "global_step": 199158, "epoch": 2237, "lr": 1.6559700736590784e-05}
{"train_loss": 0.07660281658172607, "global_step": 199159, "epoch": 2237, "lr": 1.6559269741204787e-05}
{"train_loss": 0.09133249521255493, "global_step": 199160, "epoch": 2237, "lr": 1.65588387503145e-05}
{"train_loss": 0.05049527436494827, "global_step": 199161, "epoch": 2237, "lr": 1.655840776391996e-05}
{"train_loss": 0.09060560166835785, "global_step": 199162, "epoch": 2237, "lr": 1.6557976782021245e-05}
{"train_loss": 0.057821858674287796, "global_step": 199163, "epoch": 2237, "lr": 1.6557545804618384e-05}
{"train_loss": 0.049058202654123306, "global_step": 199164, "epoch": 2237, "lr": 1.655711483171148e-05}
{"train_loss": 0.04274645820260048, "global_step": 199165, "epoch": 2237, "lr": 1.6556683863300537e-05}
{"train_loss": 0.04104582592844963, "global_step": 199166, "epoch": 2237, "lr": 1.655625289938566e-05}
{"train_loss": 0.036706920713186264, "global_step": 199167, "epoch": 2237, "lr": 1.6555821939966877e-05}
{"train_loss": 0.04964936524629593, "global_step": 199168, "epoch": 2237, "lr": 1.6555390985044267e-05}
{"train_loss": 0.049420349299907684, "global_step": 199169, "epoch": 2237, "lr": 1.655496003461786e-05}
{"train_loss": 0.050508707761764526, "global_step": 199170, "epoch": 2237, "lr": 1.655452908868775e-05}
{"train_loss": 0.06110812723636627, "global_step": 199171, "epoch": 2237, "lr": 1.6554098147253956e-05}
{"train_loss": 0.041585229337215424, "global_step": 199172, "epoch": 2237, "lr": 1.6553667210316577e-05}
{"train_loss": 0.05048718675971031, "global_step": 199173, "epoch": 2237, "lr": 1.6553236277875644e-05}
{"train_loss": 0.058960672467947006, "global_step": 199174, "epoch": 2237, "lr": 1.655280534993121e-05}
{"train_loss": 0.08029494434595108, "global_step": 199175, "epoch": 2237, "lr": 1.655237442648336e-05}
{"train_loss": 0.08160955458879471, "global_step": 199176, "epoch": 2237, "lr": 1.6551943507532114e-05}
{"train_loss": 0.07838320732116699, "global_step": 199177, "epoch": 2237, "lr": 1.6551512593077574e-05}
{"train_loss": 0.07129190117120743, "global_step": 199178, "epoch": 2237, "lr": 1.6551081683119752e-05}
{"train_loss": 0.027890820056200027, "global_step": 199179, "epoch": 2237, "lr": 1.655065077765875e-05}
{"train_loss": 0.06204730272293091, "global_step": 199180, "epoch": 2237, "lr": 1.655021987669459e-05}
{"train_loss": 0.059239658568933436, "global_step": 199181, "epoch": 2237, "lr": 1.654978898022736e-05, "val_loss": 8.323073387145996}
{"train_loss": 0.06634905934333801, "global_step": 199182, "epoch": 2238, "lr": 1.654935808825709e-05}
{"train_loss": 0.04975549504160881, "global_step": 199183, "epoch": 2238, "lr": 1.654892720078386e-05}
{"train_loss": 0.05887990444898605, "global_step": 199184, "epoch": 2238, "lr": 1.654849631780771e-05}
{"train_loss": 0.09037992358207703, "global_step": 199185, "epoch": 2238, "lr": 1.6548065439328704e-05}
{"train_loss": 0.06881427764892578, "global_step": 199186, "epoch": 2238, "lr": 1.654763456534692e-05}
{"train_loss": 0.11376497894525528, "global_step": 199187, "epoch": 2238, "lr": 1.654720369586238e-05}
{"train_loss": 0.0526743046939373, "global_step": 199188, "epoch": 2238, "lr": 1.654677283087518e-05}
{"train_loss": 0.06374833732843399, "global_step": 199189, "epoch": 2238, "lr": 1.654634197038534e-05}
{"train_loss": 0.060810841619968414, "global_step": 199190, "epoch": 2238, "lr": 1.6545911114392952e-05}
{"train_loss": 0.051500074565410614, "global_step": 199191, "epoch": 2238, "lr": 1.6545480262898046e-05}
{"train_loss": 0.028150228783488274, "global_step": 199192, "epoch": 2238, "lr": 1.6545049415900705e-05}
{"train_loss": 0.060243330895900726, "global_step": 199193, "epoch": 2238, "lr": 1.654461857340096e-05}
{"train_loss": 0.07436543703079224, "global_step": 199194, "epoch": 2238, "lr": 1.6544187735398902e-05}
{"train_loss": 0.05281573534011841, "global_step": 199195, "epoch": 2238, "lr": 1.654375690189455e-05}
{"train_loss": 0.07648080587387085, "global_step": 199196, "epoch": 2238, "lr": 1.6543326072888e-05}
{"train_loss": 0.08563961088657379, "global_step": 199197, "epoch": 2238, "lr": 1.6542895248379276e-05}
{"train_loss": 0.05613509193062782, "global_step": 199198, "epoch": 2238, "lr": 1.654246442836847e-05}
{"train_loss": 0.09405988454818726, "global_step": 199199, "epoch": 2238, "lr": 1.65420336128556e-05}
{"train_loss": 0.04140860214829445, "global_step": 199200, "epoch": 2238, "lr": 1.6541602801840768e-05}
{"train_loss": 0.10197529196739197, "global_step": 199201, "epoch": 2238, "lr": 1.654117199532399e-05}
{"train_loss": 0.029745575040578842, "global_step": 199202, "epoch": 2238, "lr": 1.6540741193305358e-05}
{"train_loss": 0.06964197009801865, "global_step": 199203, "epoch": 2238, "lr": 1.6540310395784902e-05}
{"train_loss": 0.0519365519285202, "global_step": 199204, "epoch": 2238, "lr": 1.65398796027627e-05}
{"train_loss": 0.047723282128572464, "global_step": 199205, "epoch": 2238, "lr": 1.653944881423881e-05}
{"train_loss": 0.055047690868377686, "global_step": 199206, "epoch": 2238, "lr": 1.653901803021327e-05}
{"train_loss": 0.0692773386836052, "global_step": 199207, "epoch": 2238, "lr": 1.6538587250686166e-05}
{"train_loss": 0.06329816579818726, "global_step": 199208, "epoch": 2238, "lr": 1.6538156475657545e-05}
{"train_loss": 0.07263723760843277, "global_step": 199209, "epoch": 2238, "lr": 1.6537725705127445e-05}
{"train_loss": 0.05060090124607086, "global_step": 199210, "epoch": 2238, "lr": 1.6537294939095947e-05}
{"train_loss": 0.05877240374684334, "global_step": 199211, "epoch": 2238, "lr": 1.6536864177563092e-05}
{"train_loss": 0.05585672706365585, "global_step": 199212, "epoch": 2238, "lr": 1.6536433420528968e-05}
{"train_loss": 0.040148623287677765, "global_step": 199213, "epoch": 2238, "lr": 1.6536002667993612e-05}
{"train_loss": 0.05846729129552841, "global_step": 199214, "epoch": 2238, "lr": 1.6535571919957062e-05}
{"train_loss": 0.06650488078594208, "global_step": 199215, "epoch": 2238, "lr": 1.653514117641941e-05}
{"train_loss": 0.06969055533409119, "global_step": 199216, "epoch": 2238, "lr": 1.6534710437380692e-05}
{"train_loss": 0.03740351274609566, "global_step": 199217, "epoch": 2238, "lr": 1.6534279702840988e-05}
{"train_loss": 0.06714484840631485, "global_step": 199218, "epoch": 2238, "lr": 1.6533848972800327e-05}
{"train_loss": 0.04403725266456604, "global_step": 199219, "epoch": 2238, "lr": 1.6533418247258793e-05}
{"train_loss": 0.09808821231126785, "global_step": 199220, "epoch": 2238, "lr": 1.653298752621642e-05}
{"train_loss": 0.05819009616971016, "global_step": 199221, "epoch": 2238, "lr": 1.6532556809673293e-05}
{"train_loss": 0.061026040464639664, "global_step": 199222, "epoch": 2238, "lr": 1.6532126097629443e-05}
{"train_loss": 0.03225744143128395, "global_step": 199223, "epoch": 2238, "lr": 1.653169539008494e-05}
{"train_loss": 0.08077012002468109, "global_step": 199224, "epoch": 2238, "lr": 1.6531264687039855e-05}
{"train_loss": 0.040140386670827866, "global_step": 199225, "epoch": 2238, "lr": 1.653083398849422e-05}
{"train_loss": 0.06653684377670288, "global_step": 199226, "epoch": 2238, "lr": 1.6530403294448126e-05}
{"train_loss": 0.05404209718108177, "global_step": 199227, "epoch": 2238, "lr": 1.652997260490159e-05}
{"train_loss": 0.0363486111164093, "global_step": 199228, "epoch": 2238, "lr": 1.6529541919854715e-05}
{"train_loss": 0.052910834550857544, "global_step": 199229, "epoch": 2238, "lr": 1.652911123930751e-05}
{"train_loss": 0.07506603747606277, "global_step": 199230, "epoch": 2238, "lr": 1.652868056326008e-05}
{"train_loss": 0.04251230135560036, "global_step": 199231, "epoch": 2238, "lr": 1.6528249891712437e-05}
{"train_loss": 0.04639726132154465, "global_step": 199232, "epoch": 2238, "lr": 1.6527819224664688e-05}
{"train_loss": 0.07953444123268127, "global_step": 199233, "epoch": 2238, "lr": 1.652738856211685e-05}
{"train_loss": 0.13714277744293213, "global_step": 199234, "epoch": 2238, "lr": 1.6526957904069007e-05}
{"train_loss": 0.02858106978237629, "global_step": 199235, "epoch": 2238, "lr": 1.652652725052119e-05}
{"train_loss": 0.06472554057836533, "global_step": 199236, "epoch": 2238, "lr": 1.652609660147349e-05}
{"train_loss": 0.04241210222244263, "global_step": 199237, "epoch": 2238, "lr": 1.6525665956925936e-05}
{"train_loss": 0.06627810001373291, "global_step": 199238, "epoch": 2238, "lr": 1.652523531687861e-05}
{"train_loss": 0.06694980710744858, "global_step": 199239, "epoch": 2238, "lr": 1.6524804681331547e-05}
{"train_loss": 0.060164760798215866, "global_step": 199240, "epoch": 2238, "lr": 1.652437405028481e-05}
{"train_loss": 0.02057168446481228, "global_step": 199241, "epoch": 2238, "lr": 1.6523943423738485e-05}
{"train_loss": 0.04162529110908508, "global_step": 199242, "epoch": 2238, "lr": 1.6523512801692593e-05}
{"train_loss": 0.050937239080667496, "global_step": 199243, "epoch": 2238, "lr": 1.6523082184147214e-05}
{"train_loss": 0.08555883169174194, "global_step": 199244, "epoch": 2238, "lr": 1.652265157110239e-05}
{"train_loss": 0.03143885359168053, "global_step": 199245, "epoch": 2238, "lr": 1.65222209625582e-05}
{"train_loss": 0.06379097700119019, "global_step": 199246, "epoch": 2238, "lr": 1.6521790358514695e-05}
{"train_loss": 0.057097189128398895, "global_step": 199247, "epoch": 2238, "lr": 1.6521359758971904e-05}
{"train_loss": 0.07655445486307144, "global_step": 199248, "epoch": 2238, "lr": 1.6520929163929928e-05}
{"train_loss": 0.036217499524354935, "global_step": 199249, "epoch": 2238, "lr": 1.6520498573388786e-05}
{"train_loss": 0.10415873676538467, "global_step": 199250, "epoch": 2238, "lr": 1.652006798734857e-05}
{"train_loss": 0.07228299975395203, "global_step": 199251, "epoch": 2238, "lr": 1.651963740580933e-05}
{"train_loss": 0.04019041731953621, "global_step": 199252, "epoch": 2238, "lr": 1.6519206828771096e-05}
{"train_loss": 0.0976000726222992, "global_step": 199253, "epoch": 2238, "lr": 1.6518776256233958e-05}
{"train_loss": 0.09638059884309769, "global_step": 199254, "epoch": 2238, "lr": 1.6518345688197956e-05}
{"train_loss": 0.08474192023277283, "global_step": 199255, "epoch": 2238, "lr": 1.6517915124663163e-05}
{"train_loss": 0.09268287569284439, "global_step": 199256, "epoch": 2238, "lr": 1.6517484565629614e-05}
{"train_loss": 0.04431537911295891, "global_step": 199257, "epoch": 2238, "lr": 1.6517054011097398e-05}
{"train_loss": 0.02491779252886772, "global_step": 199258, "epoch": 2238, "lr": 1.651662346106654e-05}
{"train_loss": 0.05369199812412262, "global_step": 199259, "epoch": 2238, "lr": 1.6516192915537114e-05}
{"train_loss": 0.06163349375128746, "global_step": 199260, "epoch": 2238, "lr": 1.6515762374509192e-05}
{"train_loss": 0.0477273054420948, "global_step": 199261, "epoch": 2238, "lr": 1.6515331837982806e-05}
{"train_loss": 0.05254022032022476, "global_step": 199262, "epoch": 2238, "lr": 1.6514901305958037e-05}
{"train_loss": 0.10658859461545944, "global_step": 199263, "epoch": 2238, "lr": 1.6514470778434914e-05}
{"train_loss": 0.06501978635787964, "global_step": 199264, "epoch": 2238, "lr": 1.6514040255413533e-05}
{"train_loss": 0.10788476467132568, "global_step": 199265, "epoch": 2238, "lr": 1.6513609736893916e-05}
{"train_loss": 0.04063604399561882, "global_step": 199266, "epoch": 2238, "lr": 1.6513179222876146e-05}
{"train_loss": 0.022725876420736313, "global_step": 199267, "epoch": 2238, "lr": 1.651274871336026e-05}
{"train_loss": 0.05396481603384018, "global_step": 199268, "epoch": 2238, "lr": 1.6512318208346338e-05}
{"train_loss": 0.06619440764188766, "global_step": 199269, "epoch": 2238, "lr": 1.6511887707834417e-05}
{"train_loss": 0.061870719044563476, "global_step": 199270, "epoch": 2238, "lr": 1.6511457211824578e-05, "val_loss": 8.350029945373535}
{"train_loss": 0.052398040890693665, "global_step": 199271, "epoch": 2239, "lr": 1.6511026720316848e-05}
{"train_loss": 0.053176362067461014, "global_step": 199272, "epoch": 2239, "lr": 1.6510596233311322e-05}
{"train_loss": 0.055845167487859726, "global_step": 199273, "epoch": 2239, "lr": 1.6510165750808016e-05}
{"train_loss": 0.0588468573987484, "global_step": 199274, "epoch": 2239, "lr": 1.650973527280703e-05}
{"train_loss": 0.04196696728467941, "global_step": 199275, "epoch": 2239, "lr": 1.6509304799308382e-05}
{"train_loss": 0.06447067111730576, "global_step": 199276, "epoch": 2239, "lr": 1.6508874330312152e-05}
{"train_loss": 0.09969853609800339, "global_step": 199277, "epoch": 2239, "lr": 1.6508443865818414e-05}
{"train_loss": 0.12267367541790009, "global_step": 199278, "epoch": 2239, "lr": 1.6508013405827187e-05}
{"train_loss": 0.06843305379152298, "global_step": 199279, "epoch": 2239, "lr": 1.6507582950338575e-05}
{"train_loss": 0.0743856206536293, "global_step": 199280, "epoch": 2239, "lr": 1.6507152499352586e-05}
{"train_loss": 0.06248784437775612, "global_step": 199281, "epoch": 2239, "lr": 1.6506722052869316e-05}
{"train_loss": 0.06076023727655411, "global_step": 199282, "epoch": 2239, "lr": 1.6506291610888798e-05}
{"train_loss": 0.08209466189146042, "global_step": 199283, "epoch": 2239, "lr": 1.6505861173411118e-05}
{"train_loss": 0.032711539417505264, "global_step": 199284, "epoch": 2239, "lr": 1.6505430740436312e-05}
{"train_loss": 0.06549054384231567, "global_step": 199285, "epoch": 2239, "lr": 1.650500031196443e-05}
{"train_loss": 0.084046371281147, "global_step": 199286, "epoch": 2239, "lr": 1.650456988799556e-05}
{"train_loss": 0.030303964391350746, "global_step": 199287, "epoch": 2239, "lr": 1.6504139468529722e-05}
{"train_loss": 0.04205808788537979, "global_step": 199288, "epoch": 2239, "lr": 1.650370905356701e-05}
{"train_loss": 0.02783624269068241, "global_step": 199289, "epoch": 2239, "lr": 1.6503278643107456e-05}
{"train_loss": 0.04440855607390404, "global_step": 199290, "epoch": 2239, "lr": 1.6502848237151135e-05}
{"train_loss": 0.06187174841761589, "global_step": 199291, "epoch": 2239, "lr": 1.6502417835698104e-05}
{"train_loss": 0.06619925051927567, "global_step": 199292, "epoch": 2239, "lr": 1.65019874387484e-05}
{"train_loss": 0.08678575605154037, "global_step": 199293, "epoch": 2239, "lr": 1.6501557046302108e-05}
{"train_loss": 0.08191724866628647, "global_step": 199294, "epoch": 2239, "lr": 1.6501126658359257e-05}
{"train_loss": 0.04012037068605423, "global_step": 199295, "epoch": 2239, "lr": 1.6500696274919924e-05}
{"train_loss": 0.08347057551145554, "global_step": 199296, "epoch": 2239, "lr": 1.650026589598418e-05}
{"train_loss": 0.05448067933320999, "global_step": 199297, "epoch": 2239, "lr": 1.649983552155205e-05}
{"train_loss": 0.042117100208997726, "global_step": 199298, "epoch": 2239, "lr": 1.649940515162362e-05}
{"train_loss": 0.07567325234413147, "global_step": 199299, "epoch": 2239, "lr": 1.6498974786198924e-05}
{"train_loss": 0.09574731439352036, "global_step": 199300, "epoch": 2239, "lr": 1.649854442527805e-05}
{"train_loss": 0.035169340670108795, "global_step": 199301, "epoch": 2239, "lr": 1.649811406886102e-05}
{"train_loss": 0.08732260018587112, "global_step": 199302, "epoch": 2239, "lr": 1.649768371694793e-05}
{"train_loss": 0.06488127261400223, "global_step": 199303, "epoch": 2239, "lr": 1.64972533695388e-05}
{"train_loss": 0.05298596993088722, "global_step": 199304, "epoch": 2239, "lr": 1.6496823026633722e-05}
{"train_loss": 0.05263673514127731, "global_step": 199305, "epoch": 2239, "lr": 1.6496392688232726e-05}
{"train_loss": 0.07045246660709381, "global_step": 199306, "epoch": 2239, "lr": 1.6495962354335893e-05}
{"train_loss": 0.05700770020484924, "global_step": 199307, "epoch": 2239, "lr": 1.649553202494325e-05}
{"train_loss": 0.07340092211961746, "global_step": 199308, "epoch": 2239, "lr": 1.6495101700054892e-05}
{"train_loss": 0.0669880136847496, "global_step": 199309, "epoch": 2239, "lr": 1.649467137967085e-05}
{"train_loss": 0.07013751566410065, "global_step": 199310, "epoch": 2239, "lr": 1.6494241063791204e-05}
{"train_loss": 0.066146619617939, "global_step": 199311, "epoch": 2239, "lr": 1.6493810752415977e-05}
{"train_loss": 0.08934357017278671, "global_step": 199312, "epoch": 2239, "lr": 1.6493380445545255e-05}
{"train_loss": 0.07502378523349762, "global_step": 199313, "epoch": 2239, "lr": 1.6492950143179105e-05}
{"train_loss": 0.0309619028121233, "global_step": 199314, "epoch": 2239, "lr": 1.649251984531755e-05}
{"train_loss": 0.055219706147909164, "global_step": 199315, "epoch": 2239, "lr": 1.6492089551960686e-05}
{"train_loss": 0.030384305864572525, "global_step": 199316, "epoch": 2239, "lr": 1.649165926310854e-05}
{"train_loss": 0.14675737917423248, "global_step": 199317, "epoch": 2239, "lr": 1.6491228978761193e-05}
{"train_loss": 0.1202302798628807, "global_step": 199318, "epoch": 2239, "lr": 1.6490798698918675e-05}
{"train_loss": 0.0803460031747818, "global_step": 199319, "epoch": 2239, "lr": 1.6490368423581075e-05}
{"train_loss": 0.028042003512382507, "global_step": 199320, "epoch": 2239, "lr": 1.648993815274843e-05}
{"train_loss": 0.053550835698843, "global_step": 199321, "epoch": 2239, "lr": 1.6489507886420814e-05}
{"train_loss": 0.04774164780974388, "global_step": 199322, "epoch": 2239, "lr": 1.6489077624598258e-05}
{"train_loss": 0.035692278295755386, "global_step": 199323, "epoch": 2239, "lr": 1.648864736728085e-05}
{"train_loss": 0.07699589431285858, "global_step": 199324, "epoch": 2239, "lr": 1.6488217114468645e-05}
{"train_loss": 0.04253165423870087, "global_step": 199325, "epoch": 2239, "lr": 1.648778686616167e-05}
{"train_loss": 0.05839686095714569, "global_step": 199326, "epoch": 2239, "lr": 1.6487356622360016e-05}
{"train_loss": 0.058071065694093704, "global_step": 199327, "epoch": 2239, "lr": 1.6486926383063716e-05}
{"train_loss": 0.07615749537944794, "global_step": 199328, "epoch": 2239, "lr": 1.6486496148272858e-05}
{"train_loss": 0.067942313849926, "global_step": 199329, "epoch": 2239, "lr": 1.648606591798748e-05}
{"train_loss": 0.0852866992354393, "global_step": 199330, "epoch": 2239, "lr": 1.6485635692207623e-05}
{"train_loss": 0.08315443247556686, "global_step": 199331, "epoch": 2239, "lr": 1.6485205470933368e-05}
{"train_loss": 0.025968968868255615, "global_step": 199332, "epoch": 2239, "lr": 1.648477525416478e-05}
{"train_loss": 0.09203588217496872, "global_step": 199333, "epoch": 2239, "lr": 1.6484345041901894e-05}
{"train_loss": 0.06833111494779587, "global_step": 199334, "epoch": 2239, "lr": 1.6483914834144794e-05}
{"train_loss": 0.06751663982868195, "global_step": 199335, "epoch": 2239, "lr": 1.648348463089351e-05}
{"train_loss": 0.031323038041591644, "global_step": 199336, "epoch": 2239, "lr": 1.6483054432148125e-05}
{"train_loss": 0.038837727159261703, "global_step": 199337, "epoch": 2239, "lr": 1.6482624237908667e-05}
{"train_loss": 0.07382193207740784, "global_step": 199338, "epoch": 2239, "lr": 1.6482194048175232e-05}
{"train_loss": 0.03614645451307297, "global_step": 199339, "epoch": 2239, "lr": 1.648176386294784e-05}
{"train_loss": 0.09605632722377777, "global_step": 199340, "epoch": 2239, "lr": 1.6481333682226585e-05}
{"train_loss": 0.04785183444619179, "global_step": 199341, "epoch": 2239, "lr": 1.6480903506011487e-05}
{"train_loss": 0.044537778943777084, "global_step": 199342, "epoch": 2239, "lr": 1.648047333430264e-05}
{"train_loss": 0.02565564401447773, "global_step": 199343, "epoch": 2239, "lr": 1.6480043167100067e-05}
{"train_loss": 0.09714928269386292, "global_step": 199344, "epoch": 2239, "lr": 1.647961300440386e-05}
{"train_loss": 0.06144874170422554, "global_step": 199345, "epoch": 2239, "lr": 1.6479182846214046e-05}
{"train_loss": 0.07340896874666214, "global_step": 199346, "epoch": 2239, "lr": 1.6478752692530713e-05}
{"train_loss": 0.021819276735186577, "global_step": 199347, "epoch": 2239, "lr": 1.6478322543353887e-05}
{"train_loss": 0.033324725925922394, "global_step": 199348, "epoch": 2239, "lr": 1.6477892398683658e-05}
{"train_loss": 0.11924988031387329, "global_step": 199349, "epoch": 2239, "lr": 1.6477462258520048e-05}
{"train_loss": 0.07769905030727386, "global_step": 199350, "epoch": 2239, "lr": 1.647703212286314e-05}
{"train_loss": 0.0836591050028801, "global_step": 199351, "epoch": 2239, "lr": 1.6476601991713002e-05}
{"train_loss": 0.06530840694904327, "global_step": 199352, "epoch": 2239, "lr": 1.6476171865069657e-05}
{"train_loss": 0.07442740350961685, "global_step": 199353, "epoch": 2239, "lr": 1.64757417429332e-05}
{"train_loss": 0.026668120175600052, "global_step": 199354, "epoch": 2239, "lr": 1.6475311625303657e-05}
{"train_loss": 0.08743877708911896, "global_step": 199355, "epoch": 2239, "lr": 1.6474881512181113e-05}
{"train_loss": 0.08439674228429794, "global_step": 199356, "epoch": 2239, "lr": 1.6474451403565594e-05}
{"train_loss": 0.07780037075281143, "global_step": 199357, "epoch": 2239, "lr": 1.64740212994572e-05}
{"train_loss": 0.05516945570707321, "global_step": 199358, "epoch": 2239, "lr": 1.6473591199855943e-05}
{"train_loss": 0.06391750901853771, "global_step": 199359, "epoch": 2239, "lr": 1.647316110476192e-05, "val_loss": 8.341336250305176}
{"train_loss": 0.041598036885261536, "global_step": 199360, "epoch": 2240, "lr": 1.6472731014175158e-05}
{"train_loss": 0.035857219249010086, "global_step": 199361, "epoch": 2240, "lr": 1.647230092809574e-05}
{"train_loss": 0.07719013094902039, "global_step": 199362, "epoch": 2240, "lr": 1.647187084652372e-05}
{"train_loss": 0.056550804525613785, "global_step": 199363, "epoch": 2240, "lr": 1.6471440769459128e-05}
{"train_loss": 0.08119150996208191, "global_step": 199364, "epoch": 2240, "lr": 1.6471010696902056e-05}
{"train_loss": 0.05283283069729805, "global_step": 199365, "epoch": 2240, "lr": 1.6470580628852534e-05}
{"train_loss": 0.07856333255767822, "global_step": 199366, "epoch": 2240, "lr": 1.6470150565310653e-05}
{"train_loss": 0.03501863032579422, "global_step": 199367, "epoch": 2240, "lr": 1.646972050627643e-05}
{"train_loss": 0.03711618110537529, "global_step": 199368, "epoch": 2240, "lr": 1.6469290451749963e-05}
{"train_loss": 0.040787264704704285, "global_step": 199369, "epoch": 2240, "lr": 1.646886040173128e-05}
{"train_loss": 0.05545647069811821, "global_step": 199370, "epoch": 2240, "lr": 1.646843035622046e-05}
{"train_loss": 0.05753398314118385, "global_step": 199371, "epoch": 2240, "lr": 1.6468000315217537e-05}
{"train_loss": 0.048844121396541595, "global_step": 199372, "epoch": 2240, "lr": 1.6467570278722594e-05}
{"train_loss": 0.08380526304244995, "global_step": 199373, "epoch": 2240, "lr": 1.646714024673567e-05}
{"train_loss": 0.0495147667825222, "global_step": 199374, "epoch": 2240, "lr": 1.6466710219256837e-05}
{"train_loss": 0.028454508632421494, "global_step": 199375, "epoch": 2240, "lr": 1.6466280196286137e-05}
{"train_loss": 0.09354287385940552, "global_step": 199376, "epoch": 2240, "lr": 1.6465850177823648e-05}
{"train_loss": 0.03570991009473801, "global_step": 199377, "epoch": 2240, "lr": 1.6465420163869404e-05}
{"train_loss": 0.04704007878899574, "global_step": 199378, "epoch": 2240, "lr": 1.6464990154423493e-05}
{"train_loss": 0.025497078895568848, "global_step": 199379, "epoch": 2240, "lr": 1.6464560149485935e-05}
{"train_loss": 0.10324746370315552, "global_step": 199380, "epoch": 2240, "lr": 1.646413014905683e-05}
{"train_loss": 0.05603008344769478, "global_step": 199381, "epoch": 2240, "lr": 1.646370015313619e-05}
{"train_loss": 0.06287350505590439, "global_step": 199382, "epoch": 2240, "lr": 1.646327016172412e-05}
{"train_loss": 0.04389379173517227, "global_step": 199383, "epoch": 2240, "lr": 1.646284017482063e-05}
{"train_loss": 0.04331665113568306, "global_step": 199384, "epoch": 2240, "lr": 1.6462410192425825e-05}
{"train_loss": 0.056977152824401855, "global_step": 199385, "epoch": 2240, "lr": 1.6461980214539725e-05}
{"train_loss": 0.024403177201747894, "global_step": 199386, "epoch": 2240, "lr": 1.6461550241162404e-05}
{"train_loss": 0.03552703186869621, "global_step": 199387, "epoch": 2240, "lr": 1.646112027229393e-05}
{"train_loss": 0.027410807088017464, "global_step": 199388, "epoch": 2240, "lr": 1.6460690307934334e-05}
{"train_loss": 0.044497374445199966, "global_step": 199389, "epoch": 2240, "lr": 1.6460260348083707e-05}
{"train_loss": 0.09516897052526474, "global_step": 199390, "epoch": 2240, "lr": 1.6459830392742076e-05}
{"train_loss": 0.06542469561100006, "global_step": 199391, "epoch": 2240, "lr": 1.6459400441909523e-05}
{"train_loss": 0.12269780784845352, "global_step": 199392, "epoch": 2240, "lr": 1.645897049558608e-05}
{"train_loss": 0.05937998369336128, "global_step": 199393, "epoch": 2240, "lr": 1.645854055377184e-05}
{"train_loss": 0.09406612813472748, "global_step": 199394, "epoch": 2240, "lr": 1.645811061646682e-05}
{"train_loss": 0.04514278098940849, "global_step": 199395, "epoch": 2240, "lr": 1.6457680683671112e-05}
{"train_loss": 0.09989290684461594, "global_step": 199396, "epoch": 2240, "lr": 1.645725075538475e-05}
{"train_loss": 0.03403962776064873, "global_step": 199397, "epoch": 2240, "lr": 1.6456820831607816e-05}
{"train_loss": 0.06735597550868988, "global_step": 199398, "epoch": 2240, "lr": 1.6456390912340336e-05}
{"train_loss": 0.02821454033255577, "global_step": 199399, "epoch": 2240, "lr": 1.6455960997582405e-05}
{"train_loss": 0.10443904250860214, "global_step": 199400, "epoch": 2240, "lr": 1.6455531087334054e-05}
{"train_loss": 0.04803195595741272, "global_step": 199401, "epoch": 2240, "lr": 1.645510118159534e-05}
{"train_loss": 0.04298984259366989, "global_step": 199402, "epoch": 2240, "lr": 1.645467128036634e-05}
{"train_loss": 0.038097746670246124, "global_step": 199403, "epoch": 2240, "lr": 1.6454241383647085e-05}
{"train_loss": 0.1214422881603241, "global_step": 199404, "epoch": 2240, "lr": 1.6453811491437666e-05}
{"train_loss": 0.02952061966061592, "global_step": 199405, "epoch": 2240, "lr": 1.6453381603738104e-05}
{"train_loss": 0.07736875861883163, "global_step": 199406, "epoch": 2240, "lr": 1.6452951720548495e-05}
{"train_loss": 0.05559206008911133, "global_step": 199407, "epoch": 2240, "lr": 1.645252184186886e-05}
{"train_loss": 0.05533456429839134, "global_step": 199408, "epoch": 2240, "lr": 1.645209196769929e-05}
{"train_loss": 0.07970985025167465, "global_step": 199409, "epoch": 2240, "lr": 1.6451662098039815e-05}
{"train_loss": 0.09844858944416046, "global_step": 199410, "epoch": 2240, "lr": 1.6451232232890516e-05}
{"train_loss": 0.06301706284284592, "global_step": 199411, "epoch": 2240, "lr": 1.6450802372251428e-05}
{"train_loss": 0.04957167059183121, "global_step": 199412, "epoch": 2240, "lr": 1.6450372516122638e-05}
{"train_loss": 0.02419738471508026, "global_step": 199413, "epoch": 2240, "lr": 1.6449942664504165e-05}
{"train_loss": 0.05671016871929169, "global_step": 199414, "epoch": 2240, "lr": 1.6449512817396107e-05}
{"train_loss": 0.0533185750246048, "global_step": 199415, "epoch": 2240, "lr": 1.644908297479849e-05}
{"train_loss": 0.09282862395048141, "global_step": 199416, "epoch": 2240, "lr": 1.6448653136711394e-05}
{"train_loss": 0.03329108655452728, "global_step": 199417, "epoch": 2240, "lr": 1.6448223303134856e-05}
{"train_loss": 0.045125603675842285, "global_step": 199418, "epoch": 2240, "lr": 1.644779347406896e-05}
{"train_loss": 0.10853655636310577, "global_step": 199419, "epoch": 2240, "lr": 1.6447363649513737e-05}
{"train_loss": 0.06918986886739731, "global_step": 199420, "epoch": 2240, "lr": 1.6446933829469276e-05}
{"train_loss": 0.03425140306353569, "global_step": 199421, "epoch": 2240, "lr": 1.6446504013935593e-05}
{"train_loss": 0.06328253448009491, "global_step": 199422, "epoch": 2240, "lr": 1.6446074202912764e-05}
{"train_loss": 0.08490091562271118, "global_step": 199423, "epoch": 2240, "lr": 1.6445644396400876e-05}
{"train_loss": 0.08851078897714615, "global_step": 199424, "epoch": 2240, "lr": 1.6445214594399944e-05}
{"train_loss": 0.0684177428483963, "global_step": 199425, "epoch": 2240, "lr": 1.6444784796910062e-05}
{"train_loss": 0.08404962718486786, "global_step": 199426, "epoch": 2240, "lr": 1.644435500393125e-05}
{"train_loss": 0.044081997126340866, "global_step": 199427, "epoch": 2240, "lr": 1.64439252154636e-05}
{"train_loss": 0.055798452347517014, "global_step": 199428, "epoch": 2240, "lr": 1.6443495431507144e-05}
{"train_loss": 0.06988637149333954, "global_step": 199429, "epoch": 2240, "lr": 1.644306565206196e-05}
{"train_loss": 0.0615246519446373, "global_step": 199430, "epoch": 2240, "lr": 1.644263587712809e-05}
{"train_loss": 0.053370602428913116, "global_step": 199431, "epoch": 2240, "lr": 1.644220610670561e-05}
{"train_loss": 0.04966723173856735, "global_step": 199432, "epoch": 2240, "lr": 1.644177634079455e-05}
{"train_loss": 0.10092397034168243, "global_step": 199433, "epoch": 2240, "lr": 1.6441346579395002e-05}
{"train_loss": 0.06388361006975174, "global_step": 199434, "epoch": 2240, "lr": 1.6440916822506984e-05}
{"train_loss": 0.07438141107559204, "global_step": 199435, "epoch": 2240, "lr": 1.6440487070130602e-05}
{"train_loss": 0.059758998453617096, "global_step": 199436, "epoch": 2240, "lr": 1.6440057322265866e-05}
{"train_loss": 0.09622995555400848, "global_step": 199437, "epoch": 2240, "lr": 1.643962757891287e-05}
{"train_loss": 0.09051719307899475, "global_step": 199438, "epoch": 2240, "lr": 1.643919784007164e-05}
{"train_loss": 0.09033309668302536, "global_step": 199439, "epoch": 2240, "lr": 1.643876810574227e-05}
{"train_loss": 0.06898988038301468, "global_step": 199440, "epoch": 2240, "lr": 1.64383383759248e-05}
{"train_loss": 0.0744333416223526, "global_step": 199441, "epoch": 2240, "lr": 1.6437908650619267e-05}
{"train_loss": 0.063447967171669, "global_step": 199442, "epoch": 2240, "lr": 1.6437478929825766e-05}
{"train_loss": 0.11745447665452957, "global_step": 199443, "epoch": 2240, "lr": 1.643704921354432e-05}
{"train_loss": 0.040706098079681396, "global_step": 199444, "epoch": 2240, "lr": 1.643661950177502e-05}
{"train_loss": 0.06038195267319679, "global_step": 199445, "epoch": 2240, "lr": 1.6436189794517887e-05}
{"train_loss": 0.048803236335515976, "global_step": 199446, "epoch": 2240, "lr": 1.643576009177302e-05}
{"train_loss": 0.09771964699029922, "global_step": 199447, "epoch": 2240, "lr": 1.643533039354044e-05}
{"train_loss": 0.06227940410961596, "global_step": 199448, "epoch": 2240, "lr": 1.6434900699820238e-05, "val_loss": 8.420199394226074, "train_action_mse_error": 3.4074478149414062}
{"train_loss": 0.06549983471632004, "global_step": 199449, "epoch": 2241, "lr": 1.6434471010612436e-05}
{"train_loss": 0.03354684263467789, "global_step": 199450, "epoch": 2241, "lr": 1.6434041325917127e-05}
{"train_loss": 0.04662986099720001, "global_step": 199451, "epoch": 2241, "lr": 1.6433611645734336e-05}
{"train_loss": 0.024560384452342987, "global_step": 199452, "epoch": 2241, "lr": 1.643318197006415e-05}
{"train_loss": 0.08804304152727127, "global_step": 199453, "epoch": 2241, "lr": 1.6432752298906602e-05}
{"train_loss": 0.031080355867743492, "global_step": 199454, "epoch": 2241, "lr": 1.6432322632261776e-05}
{"train_loss": 0.09749165922403336, "global_step": 199455, "epoch": 2241, "lr": 1.6431892970129697e-05}
{"train_loss": 0.043063562363386154, "global_step": 199456, "epoch": 2241, "lr": 1.643146331251046e-05}
{"train_loss": 0.10829344391822815, "global_step": 199457, "epoch": 2241, "lr": 1.6431033659404083e-05}
{"train_loss": 0.05904877185821533, "global_step": 199458, "epoch": 2241, "lr": 1.6430604010810647e-05}
{"train_loss": 0.08007175475358963, "global_step": 199459, "epoch": 2241, "lr": 1.643017436673023e-05}
{"train_loss": 0.03211279585957527, "global_step": 199460, "epoch": 2241, "lr": 1.6429744727162837e-05}
{"train_loss": 0.05766230821609497, "global_step": 199461, "epoch": 2241, "lr": 1.642931509210858e-05}
{"train_loss": 0.08205483853816986, "global_step": 199462, "epoch": 2241, "lr": 1.6428885461567477e-05}
{"train_loss": 0.08167492598295212, "global_step": 199463, "epoch": 2241, "lr": 1.6428455835539612e-05}
{"train_loss": 0.049927931278944016, "global_step": 199464, "epoch": 2241, "lr": 1.6428026214025023e-05}
{"train_loss": 0.053135551512241364, "global_step": 199465, "epoch": 2241, "lr": 1.6427596597023788e-05}
{"train_loss": 0.05044781416654587, "global_step": 199466, "epoch": 2241, "lr": 1.642716698453594e-05}
{"train_loss": 0.062021221965551376, "global_step": 199467, "epoch": 2241, "lr": 1.6426737376561562e-05}
{"train_loss": 0.06472941488027573, "global_step": 199468, "epoch": 2241, "lr": 1.642630777310069e-05}
{"train_loss": 0.08463255316019058, "global_step": 199469, "epoch": 2241, "lr": 1.6425878174153404e-05}
{"train_loss": 0.12783531844615936, "global_step": 199470, "epoch": 2241, "lr": 1.6425448579719732e-05}
{"train_loss": 0.100275918841362, "global_step": 199471, "epoch": 2241, "lr": 1.6425018989799766e-05}
{"train_loss": 0.09954293817281723, "global_step": 199472, "epoch": 2241, "lr": 1.6424589404393538e-05}
{"train_loss": 0.08779945969581604, "global_step": 199473, "epoch": 2241, "lr": 1.6424159823501124e-05}
{"train_loss": 0.041594989597797394, "global_step": 199474, "epoch": 2241, "lr": 1.6423730247122556e-05}
{"train_loss": 0.06156456470489502, "global_step": 199475, "epoch": 2241, "lr": 1.6423300675257924e-05}
{"train_loss": 0.04486184939742088, "global_step": 199476, "epoch": 2241, "lr": 1.6422871107907256e-05}
{"train_loss": 0.05984862148761749, "global_step": 199477, "epoch": 2241, "lr": 1.6422441545070644e-05}
{"train_loss": 0.10005344450473785, "global_step": 199478, "epoch": 2241, "lr": 1.642201198674812e-05}
{"train_loss": 0.06264634430408478, "global_step": 199479, "epoch": 2241, "lr": 1.642158243293973e-05}
{"train_loss": 0.030519215390086174, "global_step": 199480, "epoch": 2241, "lr": 1.642115288364557e-05}
{"train_loss": 0.0832313671708107, "global_step": 199481, "epoch": 2241, "lr": 1.6420723338865657e-05}
{"train_loss": 0.07939168810844421, "global_step": 199482, "epoch": 2241, "lr": 1.6420293798600085e-05}
{"train_loss": 0.04849003255367279, "global_step": 199483, "epoch": 2241, "lr": 1.641986426284888e-05}
{"train_loss": 0.039890356361866, "global_step": 199484, "epoch": 2241, "lr": 1.6419434731612125e-05}
{"train_loss": 0.05996876582503319, "global_step": 199485, "epoch": 2241, "lr": 1.641900520488986e-05}
{"train_loss": 0.06883099675178528, "global_step": 199486, "epoch": 2241, "lr": 1.6418575682682163e-05}
{"train_loss": 0.05253589153289795, "global_step": 199487, "epoch": 2241, "lr": 1.6418146164989062e-05}
{"train_loss": 0.10552991926670074, "global_step": 199488, "epoch": 2241, "lr": 1.6417716651810654e-05}
{"train_loss": 0.0428834967315197, "global_step": 199489, "epoch": 2241, "lr": 1.6417287143146953e-05}
{"train_loss": 0.04561702534556389, "global_step": 199490, "epoch": 2241, "lr": 1.6416857638998058e-05}
{"train_loss": 0.062356263399124146, "global_step": 199491, "epoch": 2241, "lr": 1.641642813936399e-05}
{"train_loss": 0.05888693779706955, "global_step": 199492, "epoch": 2241, "lr": 1.6415998644244835e-05}
{"train_loss": 0.09162215888500214, "global_step": 199493, "epoch": 2241, "lr": 1.641556915364063e-05}
{"train_loss": 0.05357358232140541, "global_step": 199494, "epoch": 2241, "lr": 1.6415139667551438e-05}
{"train_loss": 0.060745541006326675, "global_step": 199495, "epoch": 2241, "lr": 1.641471018597734e-05}
{"train_loss": 0.03660595789551735, "global_step": 199496, "epoch": 2241, "lr": 1.6414280708918355e-05}
{"train_loss": 0.06930288672447205, "global_step": 199497, "epoch": 2241, "lr": 1.6413851236374584e-05}
{"train_loss": 0.05498562008142471, "global_step": 199498, "epoch": 2241, "lr": 1.6413421768346038e-05}
{"train_loss": 0.062276989221572876, "global_step": 199499, "epoch": 2241, "lr": 1.641299230483282e-05}
{"train_loss": 0.06090264394879341, "global_step": 199500, "epoch": 2241, "lr": 1.6412562845834943e-05}
{"train_loss": 0.054608047008514404, "global_step": 199501, "epoch": 2241, "lr": 1.6412133391352507e-05}
{"train_loss": 0.06254681944847107, "global_step": 199502, "epoch": 2241, "lr": 1.6411703941385535e-05}
{"train_loss": 0.07939153909683228, "global_step": 199503, "epoch": 2241, "lr": 1.641127449593412e-05}
{"train_loss": 0.10266599804162979, "global_step": 199504, "epoch": 2241, "lr": 1.6410845054998275e-05}
{"train_loss": 0.07802309095859528, "global_step": 199505, "epoch": 2241, "lr": 1.6410415618578108e-05}
{"train_loss": 0.07932911068201065, "global_step": 199506, "epoch": 2241, "lr": 1.6409986186673626e-05}
{"train_loss": 0.06950272619724274, "global_step": 199507, "epoch": 2241, "lr": 1.6409556759284934e-05}
{"train_loss": 0.02705548144876957, "global_step": 199508, "epoch": 2241, "lr": 1.640912733641205e-05}
{"train_loss": 0.034541912376880646, "global_step": 199509, "epoch": 2241, "lr": 1.6408697918055067e-05}
{"train_loss": 0.043392181396484375, "global_step": 199510, "epoch": 2241, "lr": 1.6408268504214003e-05}
{"train_loss": 0.06569436937570572, "global_step": 199511, "epoch": 2241, "lr": 1.640783909488896e-05}
{"train_loss": 0.04029759019613266, "global_step": 199512, "epoch": 2241, "lr": 1.6407409690079954e-05}
{"train_loss": 0.05811016634106636, "global_step": 199513, "epoch": 2241, "lr": 1.6406980289787084e-05}
{"train_loss": 0.12232374399900436, "global_step": 199514, "epoch": 2241, "lr": 1.6406550894010363e-05}
{"train_loss": 0.05665672942996025, "global_step": 199515, "epoch": 2241, "lr": 1.6406121502749893e-05}
{"train_loss": 0.07281278818845749, "global_step": 199516, "epoch": 2241, "lr": 1.640569211600571e-05}
{"train_loss": 0.10248658061027527, "global_step": 199517, "epoch": 2241, "lr": 1.6405262733777855e-05}
{"train_loss": 0.05276453495025635, "global_step": 199518, "epoch": 2241, "lr": 1.640483335606642e-05}
{"train_loss": 0.020055348053574562, "global_step": 199519, "epoch": 2241, "lr": 1.6404403982871424e-05}
{"train_loss": 0.05759452283382416, "global_step": 199520, "epoch": 2241, "lr": 1.6403974614192968e-05}
{"train_loss": 0.08752160519361496, "global_step": 199521, "epoch": 2241, "lr": 1.640354525003107e-05}
{"train_loss": 0.0439259372651577, "global_step": 199522, "epoch": 2241, "lr": 1.640311589038582e-05}
{"train_loss": 0.03872399032115936, "global_step": 199523, "epoch": 2241, "lr": 1.640268653525725e-05}
{"train_loss": 0.05026218295097351, "global_step": 199524, "epoch": 2241, "lr": 1.6402257184645447e-05}
{"train_loss": 0.052993807941675186, "global_step": 199525, "epoch": 2241, "lr": 1.640182783855043e-05}
{"train_loss": 0.07577207684516907, "global_step": 199526, "epoch": 2241, "lr": 1.64013984969723e-05}
{"train_loss": 0.057020384818315506, "global_step": 199527, "epoch": 2241, "lr": 1.640096915991107e-05}
{"train_loss": 0.07150473445653915, "global_step": 199528, "epoch": 2241, "lr": 1.6400539827366834e-05}
{"train_loss": 0.039431795477867126, "global_step": 199529, "epoch": 2241, "lr": 1.6400110499339627e-05}
{"train_loss": 0.07559715956449509, "global_step": 199530, "epoch": 2241, "lr": 1.639968117582951e-05}
{"train_loss": 0.04609975218772888, "global_step": 199531, "epoch": 2241, "lr": 1.6399251856836568e-05}
{"train_loss": 0.12185070663690567, "global_step": 199532, "epoch": 2241, "lr": 1.6398822542360813e-05}
{"train_loss": 0.09949693828821182, "global_step": 199533, "epoch": 2241, "lr": 1.6398393232402354e-05}
{"train_loss": 0.08165863901376724, "global_step": 199534, "epoch": 2241, "lr": 1.6397963926961195e-05}
{"train_loss": 0.06975610554218292, "global_step": 199535, "epoch": 2241, "lr": 1.6397534626037443e-05}
{"train_loss": 0.07899284362792969, "global_step": 199536, "epoch": 2241, "lr": 1.6397105329631114e-05}
{"train_loss": 0.06490203235926253, "global_step": 199537, "epoch": 2241, "lr": 1.6396676037742297e-05, "val_loss": 8.179034233093262}
{"train_loss": 0.038625918328762054, "global_step": 199538, "epoch": 2242, "lr": 1.639624675037103e-05}
{"train_loss": 0.05948399752378464, "global_step": 199539, "epoch": 2242, "lr": 1.639581746751739e-05}
{"train_loss": 0.03455587103962898, "global_step": 199540, "epoch": 2242, "lr": 1.6395388189181403e-05}
{"train_loss": 0.02130228281021118, "global_step": 199541, "epoch": 2242, "lr": 1.6394958915363166e-05}
{"train_loss": 0.044373854994773865, "global_step": 199542, "epoch": 2242, "lr": 1.63945296460627e-05}
{"train_loss": 0.06438177824020386, "global_step": 199543, "epoch": 2242, "lr": 1.6394100381280097e-05}
{"train_loss": 0.08917152881622314, "global_step": 199544, "epoch": 2242, "lr": 1.639367112101538e-05}
{"train_loss": 0.06325585395097733, "global_step": 199545, "epoch": 2242, "lr": 1.6393241865268645e-05}
{"train_loss": 0.04788391664624214, "global_step": 199546, "epoch": 2242, "lr": 1.639281261403991e-05}
{"train_loss": 0.1047569066286087, "global_step": 199547, "epoch": 2242, "lr": 1.6392383367329267e-05}
{"train_loss": 0.08359646797180176, "global_step": 199548, "epoch": 2242, "lr": 1.6391954125136744e-05}
{"train_loss": 0.07158469408750534, "global_step": 199549, "epoch": 2242, "lr": 1.6391524887462412e-05}
{"train_loss": 0.0681002140045166, "global_step": 199550, "epoch": 2242, "lr": 1.6391095654306344e-05}
{"train_loss": 0.05826326087117195, "global_step": 199551, "epoch": 2242, "lr": 1.6390666425668584e-05}
{"train_loss": 0.04396602511405945, "global_step": 199552, "epoch": 2242, "lr": 1.6390237201549175e-05}
{"train_loss": 0.05648726969957352, "global_step": 199553, "epoch": 2242, "lr": 1.638980798194821e-05}
{"train_loss": 0.06407362222671509, "global_step": 199554, "epoch": 2242, "lr": 1.63893787668657e-05}
{"train_loss": 0.10634537786245346, "global_step": 199555, "epoch": 2242, "lr": 1.638894955630175e-05}
{"train_loss": 0.07615165412425995, "global_step": 199556, "epoch": 2242, "lr": 1.6388520350256393e-05}
{"train_loss": 0.03702939301729202, "global_step": 199557, "epoch": 2242, "lr": 1.6388091148729677e-05}
{"train_loss": 0.04315565153956413, "global_step": 199558, "epoch": 2242, "lr": 1.6387661951721682e-05}
{"train_loss": 0.043358348309993744, "global_step": 199559, "epoch": 2242, "lr": 1.638723275923245e-05}
{"train_loss": 0.05336796119809151, "global_step": 199560, "epoch": 2242, "lr": 1.6386803571262054e-05}
{"train_loss": 0.062056705355644226, "global_step": 199561, "epoch": 2242, "lr": 1.6386374387810527e-05}
{"train_loss": 0.029426267370581627, "global_step": 199562, "epoch": 2242, "lr": 1.638594520887796e-05}
{"train_loss": 0.0371747761964798, "global_step": 199563, "epoch": 2242, "lr": 1.6385516034464376e-05}
{"train_loss": 0.05225890502333641, "global_step": 199564, "epoch": 2242, "lr": 1.6385086864569872e-05}
{"train_loss": 0.1114121675491333, "global_step": 199565, "epoch": 2242, "lr": 1.638465769919446e-05}
{"train_loss": 0.08279780298471451, "global_step": 199566, "epoch": 2242, "lr": 1.6384228538338218e-05}
{"train_loss": 0.10859248042106628, "global_step": 199567, "epoch": 2242, "lr": 1.6383799382001232e-05}
{"train_loss": 0.11269926279783249, "global_step": 199568, "epoch": 2242, "lr": 1.6383370230183515e-05}
{"train_loss": 0.0902148112654686, "global_step": 199569, "epoch": 2242, "lr": 1.6382941082885156e-05}
{"train_loss": 0.11334586888551712, "global_step": 199570, "epoch": 2242, "lr": 1.638251194010619e-05}
{"train_loss": 0.0883331298828125, "global_step": 199571, "epoch": 2242, "lr": 1.63820828018467e-05}
{"train_loss": 0.11839766055345535, "global_step": 199572, "epoch": 2242, "lr": 1.638165366810671e-05}
{"train_loss": 0.11569250375032425, "global_step": 199573, "epoch": 2242, "lr": 1.638122453888632e-05}
{"train_loss": 0.04394590109586716, "global_step": 199574, "epoch": 2242, "lr": 1.638079541418554e-05}
{"train_loss": 0.052000198513269424, "global_step": 199575, "epoch": 2242, "lr": 1.6380366294004472e-05}
{"train_loss": 0.04953230544924736, "global_step": 199576, "epoch": 2242, "lr": 1.637993717834314e-05}
{"train_loss": 0.05157623067498207, "global_step": 199577, "epoch": 2242, "lr": 1.6379508067201632e-05}
{"train_loss": 0.07798918336629868, "global_step": 199578, "epoch": 2242, "lr": 1.6379078960579974e-05}
{"train_loss": 0.054005131125450134, "global_step": 199579, "epoch": 2242, "lr": 1.637864985847825e-05}
{"train_loss": 0.08073436468839645, "global_step": 199580, "epoch": 2242, "lr": 1.6378220760896496e-05}
{"train_loss": 0.08260804414749146, "global_step": 199581, "epoch": 2242, "lr": 1.6377791667834792e-05}
{"train_loss": 0.08515258878469467, "global_step": 199582, "epoch": 2242, "lr": 1.637736257929317e-05}
{"train_loss": 0.12751080095767975, "global_step": 199583, "epoch": 2242, "lr": 1.6376933495271717e-05}
{"train_loss": 0.014802654273808002, "global_step": 199584, "epoch": 2242, "lr": 1.637650441577046e-05}
{"train_loss": 0.05751817300915718, "global_step": 199585, "epoch": 2242, "lr": 1.637607534078947e-05}
{"train_loss": 0.06134633719921112, "global_step": 199586, "epoch": 2242, "lr": 1.6375646270328826e-05}
{"train_loss": 0.11645234376192093, "global_step": 199587, "epoch": 2242, "lr": 1.6375217204388554e-05}
{"train_loss": 0.07295218110084534, "global_step": 199588, "epoch": 2242, "lr": 1.6374788142968735e-05}
{"train_loss": 0.08801785111427307, "global_step": 199589, "epoch": 2242, "lr": 1.6374359086069418e-05}
{"train_loss": 0.03532686084508896, "global_step": 199590, "epoch": 2242, "lr": 1.637393003369064e-05}
{"train_loss": 0.06816398352384567, "global_step": 199591, "epoch": 2242, "lr": 1.6373500985832498e-05}
{"train_loss": 0.041712503880262375, "global_step": 199592, "epoch": 2242, "lr": 1.6373071942495005e-05}
{"train_loss": 0.061050381511449814, "global_step": 199593, "epoch": 2242, "lr": 1.6372642903678263e-05}
{"train_loss": 0.06605958938598633, "global_step": 199594, "epoch": 2242, "lr": 1.6372213869382312e-05}
{"train_loss": 0.054306674748659134, "global_step": 199595, "epoch": 2242, "lr": 1.637178483960719e-05}
{"train_loss": 0.04391965642571449, "global_step": 199596, "epoch": 2242, "lr": 1.6371355814352984e-05}
{"train_loss": 0.06393197178840637, "global_step": 199597, "epoch": 2242, "lr": 1.6370926793619728e-05}
{"train_loss": 0.11111080646514893, "global_step": 199598, "epoch": 2242, "lr": 1.63704977774075e-05}
{"train_loss": 0.03675932437181473, "global_step": 199599, "epoch": 2242, "lr": 1.637006876571634e-05}
{"train_loss": 0.05463043227791786, "global_step": 199600, "epoch": 2242, "lr": 1.636963975854633e-05}
{"train_loss": 0.09564482420682907, "global_step": 199601, "epoch": 2242, "lr": 1.636921075589749e-05}
{"train_loss": 0.06675548106431961, "global_step": 199602, "epoch": 2242, "lr": 1.6368781757769918e-05}
{"train_loss": 0.11364061385393143, "global_step": 199603, "epoch": 2242, "lr": 1.636835276416364e-05}
{"train_loss": 0.036166686564683914, "global_step": 199604, "epoch": 2242, "lr": 1.6367923775078724e-05}
{"train_loss": 0.05199817568063736, "global_step": 199605, "epoch": 2242, "lr": 1.6367494790515252e-05}
{"train_loss": 0.03995528444647789, "global_step": 199606, "epoch": 2242, "lr": 1.6367065810473242e-05}
{"train_loss": 0.07776530086994171, "global_step": 199607, "epoch": 2242, "lr": 1.6366636834952786e-05}
{"train_loss": 0.05742703005671501, "global_step": 199608, "epoch": 2242, "lr": 1.636620786395391e-05}
{"train_loss": 0.0853179469704628, "global_step": 199609, "epoch": 2242, "lr": 1.6365778897476702e-05}
{"train_loss": 0.10055353492498398, "global_step": 199610, "epoch": 2242, "lr": 1.6365349935521186e-05}
{"train_loss": 0.10478488355875015, "global_step": 199611, "epoch": 2242, "lr": 1.636492097808746e-05}
{"train_loss": 0.028664480894804, "global_step": 199612, "epoch": 2242, "lr": 1.6364492025175544e-05}
{"train_loss": 0.03780796751379967, "global_step": 199613, "epoch": 2242, "lr": 1.6364063076785526e-05}
{"train_loss": 0.052262645214796066, "global_step": 199614, "epoch": 2242, "lr": 1.6363634132917433e-05}
{"train_loss": 0.04345384240150452, "global_step": 199615, "epoch": 2242, "lr": 1.636320519357136e-05}
{"train_loss": 0.08996019512414932, "global_step": 199616, "epoch": 2242, "lr": 1.636277625874732e-05}
{"train_loss": 0.053501687943935394, "global_step": 199617, "epoch": 2242, "lr": 1.6362347328445414e-05}
{"train_loss": 0.1308886855840683, "global_step": 199618, "epoch": 2242, "lr": 1.6361918402665673e-05}
{"train_loss": 0.09127213060855865, "global_step": 199619, "epoch": 2242, "lr": 1.6361489481408166e-05}
{"train_loss": 0.07754043489694595, "global_step": 199620, "epoch": 2242, "lr": 1.6361060564672938e-05}
{"train_loss": 0.08794090151786804, "global_step": 199621, "epoch": 2242, "lr": 1.6360631652460052e-05}
{"train_loss": 0.02323661558330059, "global_step": 199622, "epoch": 2242, "lr": 1.6360202744769587e-05}
{"train_loss": 0.0548248253762722, "global_step": 199623, "epoch": 2242, "lr": 1.6359773841601567e-05}
{"train_loss": 0.0653967335820198, "global_step": 199624, "epoch": 2242, "lr": 1.635934494295608e-05}
{"train_loss": 0.053815875202417374, "global_step": 199625, "epoch": 2242, "lr": 1.635891604883315e-05}
{"train_loss": 0.06801525843486692, "global_step": 199626, "epoch": 2242, "lr": 1.6358487159232877e-05, "val_loss": 8.230406761169434}
{"train_loss": 0.045837488025426865, "global_step": 199627, "epoch": 2243, "lr": 1.6358058274155292e-05}
{"train_loss": 0.03865022212266922, "global_step": 199628, "epoch": 2243, "lr": 1.635762939360044e-05}
{"train_loss": 0.08625487983226776, "global_step": 199629, "epoch": 2243, "lr": 1.635720051756841e-05}
{"train_loss": 0.08807209134101868, "global_step": 199630, "epoch": 2243, "lr": 1.635677164605923e-05}
{"train_loss": 0.0652991533279419, "global_step": 199631, "epoch": 2243, "lr": 1.6356342779072986e-05}
{"train_loss": 0.09885245561599731, "global_step": 199632, "epoch": 2243, "lr": 1.6355913916609724e-05}
{"train_loss": 0.07438323646783829, "global_step": 199633, "epoch": 2243, "lr": 1.6355485058669478e-05}
{"train_loss": 0.07740555703639984, "global_step": 199634, "epoch": 2243, "lr": 1.6355056205252344e-05}
{"train_loss": 0.09074503183364868, "global_step": 199635, "epoch": 2243, "lr": 1.635462735635835e-05}
{"train_loss": 0.05042727291584015, "global_step": 199636, "epoch": 2243, "lr": 1.6354198511987586e-05}
{"train_loss": 0.0480254590511322, "global_step": 199637, "epoch": 2243, "lr": 1.635376967214007e-05}
{"train_loss": 0.07010519504547119, "global_step": 199638, "epoch": 2243, "lr": 1.6353340836815888e-05}
{"train_loss": 0.06455698609352112, "global_step": 199639, "epoch": 2243, "lr": 1.6352912006015083e-05}
{"train_loss": 0.043049219995737076, "global_step": 199640, "epoch": 2243, "lr": 1.6352483179737715e-05}
{"train_loss": 0.0293720830231905, "global_step": 199641, "epoch": 2243, "lr": 1.6352054357983865e-05}
{"train_loss": 0.09022018313407898, "global_step": 199642, "epoch": 2243, "lr": 1.635162554075355e-05}
{"train_loss": 0.07852766662836075, "global_step": 199643, "epoch": 2243, "lr": 1.635119672804687e-05}
{"train_loss": 0.0639975368976593, "global_step": 199644, "epoch": 2243, "lr": 1.635076791986384e-05}
{"train_loss": 0.05720005929470062, "global_step": 199645, "epoch": 2243, "lr": 1.635033911620456e-05}
{"train_loss": 0.07164382189512253, "global_step": 199646, "epoch": 2243, "lr": 1.6349910317069052e-05}
{"train_loss": 0.059995006769895554, "global_step": 199647, "epoch": 2243, "lr": 1.63494815224574e-05}
{"train_loss": 0.0584946870803833, "global_step": 199648, "epoch": 2243, "lr": 1.6349052732369634e-05}
{"train_loss": 0.04948067665100098, "global_step": 199649, "epoch": 2243, "lr": 1.6348623946805847e-05}
{"train_loss": 0.0663992241024971, "global_step": 199650, "epoch": 2243, "lr": 1.6348195165766055e-05}
{"train_loss": 0.06065019965171814, "global_step": 199651, "epoch": 2243, "lr": 1.6347766389250362e-05}
{"train_loss": 0.06839625537395477, "global_step": 199652, "epoch": 2243, "lr": 1.634733761725878e-05}
{"train_loss": 0.05215102806687355, "global_step": 199653, "epoch": 2243, "lr": 1.6346908849791408e-05}
{"train_loss": 0.03464202582836151, "global_step": 199654, "epoch": 2243, "lr": 1.6346480086848265e-05}
{"train_loss": 0.04475412517786026, "global_step": 199655, "epoch": 2243, "lr": 1.6346051328429445e-05}
{"train_loss": 0.06347323954105377, "global_step": 199656, "epoch": 2243, "lr": 1.6345622574534976e-05}
{"train_loss": 0.05111552029848099, "global_step": 199657, "epoch": 2243, "lr": 1.6345193825164923e-05}
{"train_loss": 0.08561954647302628, "global_step": 199658, "epoch": 2243, "lr": 1.6344765080319365e-05}
{"train_loss": 0.03875616565346718, "global_step": 199659, "epoch": 2243, "lr": 1.6344336339998328e-05}
{"train_loss": 0.055242158472537994, "global_step": 199660, "epoch": 2243, "lr": 1.6343907604201897e-05}
{"train_loss": 0.07467572391033173, "global_step": 199661, "epoch": 2243, "lr": 1.6343478872930106e-05}
{"train_loss": 0.025332313030958176, "global_step": 199662, "epoch": 2243, "lr": 1.6343050146183043e-05}
{"train_loss": 0.044086750596761703, "global_step": 199663, "epoch": 2243, "lr": 1.6342621423960725e-05}
{"train_loss": 0.09673800319433212, "global_step": 199664, "epoch": 2243, "lr": 1.634219270626325e-05}
{"train_loss": 0.047364089637994766, "global_step": 199665, "epoch": 2243, "lr": 1.6341763993090635e-05}
{"train_loss": 0.10215224325656891, "global_step": 199666, "epoch": 2243, "lr": 1.6341335284442982e-05}
{"train_loss": 0.07770780473947525, "global_step": 199667, "epoch": 2243, "lr": 1.6340906580320325e-05}
{"train_loss": 0.04958928003907204, "global_step": 199668, "epoch": 2243, "lr": 1.634047788072271e-05}
{"train_loss": 0.06618476659059525, "global_step": 199669, "epoch": 2243, "lr": 1.634004918565022e-05}
{"train_loss": 0.056077469140291214, "global_step": 199670, "epoch": 2243, "lr": 1.633962049510288e-05}
{"train_loss": 0.04936826974153519, "global_step": 199671, "epoch": 2243, "lr": 1.633919180908079e-05}
{"train_loss": 0.09086302667856216, "global_step": 199672, "epoch": 2243, "lr": 1.6338763127583984e-05}
{"train_loss": 0.061857037246227264, "global_step": 199673, "epoch": 2243, "lr": 1.6338334450612504e-05}
{"train_loss": 0.044925421476364136, "global_step": 199674, "epoch": 2243, "lr": 1.633790577816644e-05}
{"train_loss": 0.051918841898441315, "global_step": 199675, "epoch": 2243, "lr": 1.633747711024582e-05}
{"train_loss": 0.0968611091375351, "global_step": 199676, "epoch": 2243, "lr": 1.6337048446850718e-05}
{"train_loss": 0.11267545819282532, "global_step": 199677, "epoch": 2243, "lr": 1.63366197879812e-05}
{"train_loss": 0.0316954180598259, "global_step": 199678, "epoch": 2243, "lr": 1.6336191133637303e-05}
{"train_loss": 0.06753870844841003, "global_step": 199679, "epoch": 2243, "lr": 1.6335762483819107e-05}
{"train_loss": 0.08221456408500671, "global_step": 199680, "epoch": 2243, "lr": 1.633533383852664e-05}
{"train_loss": 0.10246361792087555, "global_step": 199681, "epoch": 2243, "lr": 1.6334905197759996e-05}
{"train_loss": 0.058690328150987625, "global_step": 199682, "epoch": 2243, "lr": 1.6334476561519197e-05}
{"train_loss": 0.06674543768167496, "global_step": 199683, "epoch": 2243, "lr": 1.633404792980433e-05}
{"train_loss": 0.05590091645717621, "global_step": 199684, "epoch": 2243, "lr": 1.6333619302615427e-05}
{"train_loss": 0.02601420134305954, "global_step": 199685, "epoch": 2243, "lr": 1.6333190679952575e-05}
{"train_loss": 0.06519732624292374, "global_step": 199686, "epoch": 2243, "lr": 1.6332762061815797e-05}
{"train_loss": 0.10781469196081161, "global_step": 199687, "epoch": 2243, "lr": 1.633233344820519e-05}
{"train_loss": 0.07112012058496475, "global_step": 199688, "epoch": 2243, "lr": 1.633190483912077e-05}
{"train_loss": 0.046425510197877884, "global_step": 199689, "epoch": 2243, "lr": 1.6331476234562633e-05}
{"train_loss": 0.07951056957244873, "global_step": 199690, "epoch": 2243, "lr": 1.63310476345308e-05}
{"train_loss": 0.04673471301794052, "global_step": 199691, "epoch": 2243, "lr": 1.633061903902537e-05}
{"train_loss": 0.05488861724734306, "global_step": 199692, "epoch": 2243, "lr": 1.6330190448046356e-05}
{"train_loss": 0.03315543383359909, "global_step": 199693, "epoch": 2243, "lr": 1.6329761861593834e-05}
{"train_loss": 0.03582148253917694, "global_step": 199694, "epoch": 2243, "lr": 1.6329333279667887e-05}
{"train_loss": 0.06554900854825974, "global_step": 199695, "epoch": 2243, "lr": 1.632890470226853e-05}
{"train_loss": 0.10348904132843018, "global_step": 199696, "epoch": 2243, "lr": 1.6328476129395866e-05}
{"train_loss": 0.053488608449697495, "global_step": 199697, "epoch": 2243, "lr": 1.632804756104991e-05}
{"train_loss": 0.07895614206790924, "global_step": 199698, "epoch": 2243, "lr": 1.632761899723075e-05}
{"train_loss": 0.06726940721273422, "global_step": 199699, "epoch": 2243, "lr": 1.632719043793841e-05}
{"train_loss": 0.05651000887155533, "global_step": 199700, "epoch": 2243, "lr": 1.632676188317299e-05}
{"train_loss": 0.08236303180456161, "global_step": 199701, "epoch": 2243, "lr": 1.6326333332934513e-05}
{"train_loss": 0.050627805292606354, "global_step": 199702, "epoch": 2243, "lr": 1.6325904787223063e-05}
{"train_loss": 0.0405697375535965, "global_step": 199703, "epoch": 2243, "lr": 1.632547624603867e-05}
{"train_loss": 0.0610840767621994, "global_step": 199704, "epoch": 2243, "lr": 1.6325047709381425e-05}
{"train_loss": 0.09448655694723129, "global_step": 199705, "epoch": 2243, "lr": 1.632461917725136e-05}
{"train_loss": 0.062094058841466904, "global_step": 199706, "epoch": 2243, "lr": 1.632419064964853e-05}
{"train_loss": 0.10841259360313416, "global_step": 199707, "epoch": 2243, "lr": 1.632376212657301e-05}
{"train_loss": 0.07477333396673203, "global_step": 199708, "epoch": 2243, "lr": 1.6323333608024844e-05}
{"train_loss": 0.04599988833069801, "global_step": 199709, "epoch": 2243, "lr": 1.6322905094004102e-05}
{"train_loss": 0.06961974501609802, "global_step": 199710, "epoch": 2243, "lr": 1.6322476584510843e-05}
{"train_loss": 0.06315533071756363, "global_step": 199711, "epoch": 2243, "lr": 1.6322048079545093e-05}
{"train_loss": 0.08325009047985077, "global_step": 199712, "epoch": 2243, "lr": 1.6321619579106938e-05}
{"train_loss": 0.03872397541999817, "global_step": 199713, "epoch": 2243, "lr": 1.6321191083196448e-05}
{"train_loss": 0.09199414402246475, "global_step": 199714, "epoch": 2243, "lr": 1.6320762591813644e-05}
{"train_loss": 0.06535547881732497, "global_step": 199715, "epoch": 2243, "lr": 1.6320334104958624e-05, "val_loss": 8.037657737731934}
{"train_loss": 0.06298495829105377, "global_step": 199716, "epoch": 2244, "lr": 1.6319905622631405e-05}
{"train_loss": 0.06693292409181595, "global_step": 199717, "epoch": 2244, "lr": 1.631947714483208e-05}
{"train_loss": 0.05755431205034256, "global_step": 199718, "epoch": 2244, "lr": 1.6319048671560672e-05}
{"train_loss": 0.05778965726494789, "global_step": 199719, "epoch": 2244, "lr": 1.6318620202817277e-05}
{"train_loss": 0.05948244407773018, "global_step": 199720, "epoch": 2244, "lr": 1.6318191738601913e-05}
{"train_loss": 0.0987471267580986, "global_step": 199721, "epoch": 2244, "lr": 1.631776327891468e-05}
{"train_loss": 0.06816251575946808, "global_step": 199722, "epoch": 2244, "lr": 1.631733482375559e-05}
{"train_loss": 0.10320840775966644, "global_step": 199723, "epoch": 2244, "lr": 1.6316906373124745e-05}
{"train_loss": 0.05424509197473526, "global_step": 199724, "epoch": 2244, "lr": 1.6316477927022167e-05}
{"train_loss": 0.07265675067901611, "global_step": 199725, "epoch": 2244, "lr": 1.6316049485447936e-05}
{"train_loss": 0.06652559340000153, "global_step": 199726, "epoch": 2244, "lr": 1.6315621048402093e-05}
{"train_loss": 0.03650709241628647, "global_step": 199727, "epoch": 2244, "lr": 1.6315192615884718e-05}
{"train_loss": 0.1339668333530426, "global_step": 199728, "epoch": 2244, "lr": 1.6314764187895838e-05}
{"train_loss": 0.055668074637651443, "global_step": 199729, "epoch": 2244, "lr": 1.6314335764435546e-05}
{"train_loss": 0.06305177509784698, "global_step": 199730, "epoch": 2244, "lr": 1.6313907345503864e-05}
{"train_loss": 0.022775638848543167, "global_step": 199731, "epoch": 2244, "lr": 1.631347893110086e-05}
{"train_loss": 0.08742713183164597, "global_step": 199732, "epoch": 2244, "lr": 1.6313050521226625e-05}
{"train_loss": 0.04669879376888275, "global_step": 199733, "epoch": 2244, "lr": 1.6312622115881165e-05}
{"train_loss": 0.049361079931259155, "global_step": 199734, "epoch": 2244, "lr": 1.6312193715064584e-05}
{"train_loss": 0.07326380908489227, "global_step": 199735, "epoch": 2244, "lr": 1.6311765318776902e-05}
{"train_loss": 0.06909558922052383, "global_step": 199736, "epoch": 2244, "lr": 1.6311336927018207e-05}
{"train_loss": 0.054372284561395645, "global_step": 199737, "epoch": 2244, "lr": 1.6310908539788522e-05}
{"train_loss": 0.09438096731901169, "global_step": 199738, "epoch": 2244, "lr": 1.6310480157087943e-05}
{"train_loss": 0.06195232272148132, "global_step": 199739, "epoch": 2244, "lr": 1.6310051778916497e-05}
{"train_loss": 0.14338766038417816, "global_step": 199740, "epoch": 2244, "lr": 1.6309623405274267e-05}
{"train_loss": 0.07493817806243896, "global_step": 199741, "epoch": 2244, "lr": 1.6309195036161286e-05}
{"train_loss": 0.06240410730242729, "global_step": 199742, "epoch": 2244, "lr": 1.630876667157763e-05}
{"train_loss": 0.03715428337454796, "global_step": 199743, "epoch": 2244, "lr": 1.6308338311523358e-05}
{"train_loss": 0.058771051466464996, "global_step": 199744, "epoch": 2244, "lr": 1.6307909955998503e-05}
{"train_loss": 0.04924441874027252, "global_step": 199745, "epoch": 2244, "lr": 1.630748160500315e-05}
{"train_loss": 0.04570472612977028, "global_step": 199746, "epoch": 2244, "lr": 1.630705325853733e-05}
{"train_loss": 0.02684936299920082, "global_step": 199747, "epoch": 2244, "lr": 1.6306624916601132e-05}
{"train_loss": 0.0529048815369606, "global_step": 199748, "epoch": 2244, "lr": 1.6306196579194582e-05}
{"train_loss": 0.033989232033491135, "global_step": 199749, "epoch": 2244, "lr": 1.6305768246317772e-05}
{"train_loss": 0.04012153297662735, "global_step": 199750, "epoch": 2244, "lr": 1.6305339917970725e-05}
{"train_loss": 0.013743279501795769, "global_step": 199751, "epoch": 2244, "lr": 1.6304911594153528e-05}
{"train_loss": 0.05282561480998993, "global_step": 199752, "epoch": 2244, "lr": 1.6304483274866206e-05}
{"train_loss": 0.06316995620727539, "global_step": 199753, "epoch": 2244, "lr": 1.6304054960108857e-05}
{"train_loss": 0.05540473014116287, "global_step": 199754, "epoch": 2244, "lr": 1.6303626649881498e-05}
{"train_loss": 0.04091711342334747, "global_step": 199755, "epoch": 2244, "lr": 1.6303198344184223e-05}
{"train_loss": 0.09457073360681534, "global_step": 199756, "epoch": 2244, "lr": 1.6302770043017057e-05}
{"train_loss": 0.03195058926939964, "global_step": 199757, "epoch": 2244, "lr": 1.630234174638009e-05}
{"train_loss": 0.07294899225234985, "global_step": 199758, "epoch": 2244, "lr": 1.6301913454273344e-05}
{"train_loss": 0.09552143514156342, "global_step": 199759, "epoch": 2244, "lr": 1.6301485166696913e-05}
{"train_loss": 0.12836627662181854, "global_step": 199760, "epoch": 2244, "lr": 1.6301056883650817e-05}
{"train_loss": 0.08204960078001022, "global_step": 199761, "epoch": 2244, "lr": 1.6300628605135148e-05}
{"train_loss": 0.06492729485034943, "global_step": 199762, "epoch": 2244, "lr": 1.6300200331149933e-05}
{"train_loss": 0.06538324058055878, "global_step": 199763, "epoch": 2244, "lr": 1.6299772061695267e-05}
{"train_loss": 0.046094201505184174, "global_step": 199764, "epoch": 2244, "lr": 1.6299343796771167e-05}
{"train_loss": 0.022016700357198715, "global_step": 199765, "epoch": 2244, "lr": 1.629891553637773e-05}
{"train_loss": 0.07818968594074249, "global_step": 199766, "epoch": 2244, "lr": 1.629848728051497e-05}
{"train_loss": 0.1084250956773758, "global_step": 199767, "epoch": 2244, "lr": 1.6298059029182965e-05}
{"train_loss": 0.03497122973203659, "global_step": 199768, "epoch": 2244, "lr": 1.6297630782381802e-05}
{"train_loss": 0.07942962646484375, "global_step": 199769, "epoch": 2244, "lr": 1.629720254011149e-05}
{"train_loss": 0.028238600119948387, "global_step": 199770, "epoch": 2244, "lr": 1.6296774302372126e-05}
{"train_loss": 0.060606878250837326, "global_step": 199771, "epoch": 2244, "lr": 1.629634606916373e-05}
{"train_loss": 0.07875864952802658, "global_step": 199772, "epoch": 2244, "lr": 1.6295917840486396e-05}
{"train_loss": 0.06874503195285797, "global_step": 199773, "epoch": 2244, "lr": 1.6295489616340153e-05}
{"train_loss": 0.06151623651385307, "global_step": 199774, "epoch": 2244, "lr": 1.629506139672508e-05}
{"train_loss": 0.05583859980106354, "global_step": 199775, "epoch": 2244, "lr": 1.6294633181641216e-05}
{"train_loss": 0.08065056055784225, "global_step": 199776, "epoch": 2244, "lr": 1.6294204971088646e-05}
{"train_loss": 0.028850065544247627, "global_step": 199777, "epoch": 2244, "lr": 1.6293776765067386e-05}
{"train_loss": 0.043099112808704376, "global_step": 199778, "epoch": 2244, "lr": 1.629334856357754e-05}
{"train_loss": 0.0673704743385315, "global_step": 199779, "epoch": 2244, "lr": 1.6292920366619118e-05}
{"train_loss": 0.03934809938073158, "global_step": 199780, "epoch": 2244, "lr": 1.629249217419222e-05}
{"train_loss": 0.10018261522054672, "global_step": 199781, "epoch": 2244, "lr": 1.6292063986296875e-05}
{"train_loss": 0.04461650177836418, "global_step": 199782, "epoch": 2244, "lr": 1.6291635802933164e-05}
{"train_loss": 0.029245318844914436, "global_step": 199783, "epoch": 2244, "lr": 1.629120762410113e-05}
{"train_loss": 0.05260802060365677, "global_step": 199784, "epoch": 2244, "lr": 1.629077944980082e-05}
{"train_loss": 0.03198469430208206, "global_step": 199785, "epoch": 2244, "lr": 1.6290351280032316e-05}
{"train_loss": 0.06527262181043625, "global_step": 199786, "epoch": 2244, "lr": 1.6289923114795645e-05}
{"train_loss": 0.06478601694107056, "global_step": 199787, "epoch": 2244, "lr": 1.6289494954090906e-05}
{"train_loss": 0.07800307124853134, "global_step": 199788, "epoch": 2244, "lr": 1.6289066797918118e-05}
{"train_loss": 0.050988201051950455, "global_step": 199789, "epoch": 2244, "lr": 1.6288638646277366e-05}
{"train_loss": 0.04156416654586792, "global_step": 199790, "epoch": 2244, "lr": 1.6288210499168678e-05}
{"train_loss": 0.06901296228170395, "global_step": 199791, "epoch": 2244, "lr": 1.6287782356592147e-05}
{"train_loss": 0.07214996218681335, "global_step": 199792, "epoch": 2244, "lr": 1.6287354218547802e-05}
{"train_loss": 0.05681084096431732, "global_step": 199793, "epoch": 2244, "lr": 1.628692608503572e-05}
{"train_loss": 0.06160719320178032, "global_step": 199794, "epoch": 2244, "lr": 1.6286497956055936e-05}
{"train_loss": 0.11325446516275406, "global_step": 199795, "epoch": 2244, "lr": 1.6286069831608537e-05}
{"train_loss": 0.04604875668883324, "global_step": 199796, "epoch": 2244, "lr": 1.628564171169355e-05}
{"train_loss": 0.08747243136167526, "global_step": 199797, "epoch": 2244, "lr": 1.6285213596311066e-05}
{"train_loss": 0.07257673889398575, "global_step": 199798, "epoch": 2244, "lr": 1.6284785485461107e-05}
{"train_loss": 0.09053901582956314, "global_step": 199799, "epoch": 2244, "lr": 1.628435737914376e-05}
{"train_loss": 0.06803023815155029, "global_step": 199800, "epoch": 2244, "lr": 1.6283929277359056e-05}
{"train_loss": 0.03296723961830139, "global_step": 199801, "epoch": 2244, "lr": 1.6283501180107086e-05}
{"train_loss": 0.04805516451597214, "global_step": 199802, "epoch": 2244, "lr": 1.6283073087387867e-05}
{"train_loss": 0.1049167737364769, "global_step": 199803, "epoch": 2244, "lr": 1.628264499920148e-05}
{"train_loss": 0.06341118662712279, "global_step": 199804, "epoch": 2244, "lr": 1.6282216915547997e-05, "val_loss": 8.30325698852539}
{"train_loss": 0.09075910598039627, "global_step": 199805, "epoch": 2245, "lr": 1.6281788836427442e-05}
{"train_loss": 0.0933433398604393, "global_step": 199806, "epoch": 2245, "lr": 1.628136076183991e-05}
{"train_loss": 0.0701877623796463, "global_step": 199807, "epoch": 2245, "lr": 1.6280932691785418e-05}
{"train_loss": 0.058325424790382385, "global_step": 199808, "epoch": 2245, "lr": 1.628050462626406e-05}
{"train_loss": 0.05245034024119377, "global_step": 199809, "epoch": 2245, "lr": 1.6280076565275858e-05}
{"train_loss": 0.08076626062393188, "global_step": 199810, "epoch": 2245, "lr": 1.627964850882091e-05}
{"train_loss": 0.03916909545660019, "global_step": 199811, "epoch": 2245, "lr": 1.6279220456899234e-05}
{"train_loss": 0.08006314933300018, "global_step": 199812, "epoch": 2245, "lr": 1.6278792409510925e-05}
{"train_loss": 0.059533875435590744, "global_step": 199813, "epoch": 2245, "lr": 1.6278364366656e-05}
{"train_loss": 0.06344059109687805, "global_step": 199814, "epoch": 2245, "lr": 1.627793632833456e-05}
{"train_loss": 0.0625419020652771, "global_step": 199815, "epoch": 2245, "lr": 1.6277508294546617e-05}
{"train_loss": 0.07066786289215088, "global_step": 199816, "epoch": 2245, "lr": 1.6277080265292273e-05}
{"train_loss": 0.08036141842603683, "global_step": 199817, "epoch": 2245, "lr": 1.6276652240571545e-05}
{"train_loss": 0.08715981990098953, "global_step": 199818, "epoch": 2245, "lr": 1.627622422038453e-05}
{"train_loss": 0.05797310918569565, "global_step": 199819, "epoch": 2245, "lr": 1.6275796204731248e-05}
{"train_loss": 0.0754452720284462, "global_step": 199820, "epoch": 2245, "lr": 1.627536819361179e-05}
{"train_loss": 0.060898471623659134, "global_step": 199821, "epoch": 2245, "lr": 1.62749401870262e-05}
{"train_loss": 0.02624528855085373, "global_step": 199822, "epoch": 2245, "lr": 1.627451218497451e-05}
{"train_loss": 0.05501573532819748, "global_step": 199823, "epoch": 2245, "lr": 1.6274084187456824e-05}
{"train_loss": 0.031869761645793915, "global_step": 199824, "epoch": 2245, "lr": 1.6273656194473157e-05}
{"train_loss": 0.07462142407894135, "global_step": 199825, "epoch": 2245, "lr": 1.6273228206023605e-05}
{"train_loss": 0.050637781620025635, "global_step": 199826, "epoch": 2245, "lr": 1.6272800222108188e-05}
{"train_loss": 0.06507673114538193, "global_step": 199827, "epoch": 2245, "lr": 1.6272372242726995e-05}
{"train_loss": 0.038452599197626114, "global_step": 199828, "epoch": 2245, "lr": 1.627194426788006e-05}
{"train_loss": 0.05144994705915451, "global_step": 199829, "epoch": 2245, "lr": 1.6271516297567467e-05}
{"train_loss": 0.0917653813958168, "global_step": 199830, "epoch": 2245, "lr": 1.627108833178924e-05}
{"train_loss": 0.06817395985126495, "global_step": 199831, "epoch": 2245, "lr": 1.627066037054547e-05}
{"train_loss": 0.09143461287021637, "global_step": 199832, "epoch": 2245, "lr": 1.6270232413836178e-05}
{"train_loss": 0.039303265511989594, "global_step": 199833, "epoch": 2245, "lr": 1.6269804461661463e-05}
{"train_loss": 0.07079032808542252, "global_step": 199834, "epoch": 2245, "lr": 1.6269376514021346e-05}
{"train_loss": 0.04395062103867531, "global_step": 199835, "epoch": 2245, "lr": 1.6268948570915915e-05}
{"train_loss": 0.06167047843337059, "global_step": 199836, "epoch": 2245, "lr": 1.6268520632345196e-05}
{"train_loss": 0.04995614290237427, "global_step": 199837, "epoch": 2245, "lr": 1.6268092698309285e-05}
{"train_loss": 0.07070563733577728, "global_step": 199838, "epoch": 2245, "lr": 1.6267664768808192e-05}
{"train_loss": 0.05194256827235222, "global_step": 199839, "epoch": 2245, "lr": 1.626723684384201e-05}
{"train_loss": 0.031088192015886307, "global_step": 199840, "epoch": 2245, "lr": 1.62668089234108e-05}
{"train_loss": 0.050742942839860916, "global_step": 199841, "epoch": 2245, "lr": 1.6266381007514587e-05}
{"train_loss": 0.055376086384058, "global_step": 199842, "epoch": 2245, "lr": 1.6265953096153464e-05}
{"train_loss": 0.06771081686019897, "global_step": 199843, "epoch": 2245, "lr": 1.6265525189327456e-05}
{"train_loss": 0.08168622851371765, "global_step": 199844, "epoch": 2245, "lr": 1.6265097287036657e-05}
{"train_loss": 0.057535190135240555, "global_step": 199845, "epoch": 2245, "lr": 1.6264669389281085e-05}
{"train_loss": 0.07073430716991425, "global_step": 199846, "epoch": 2245, "lr": 1.6264241496060832e-05}
{"train_loss": 0.04339972883462906, "global_step": 199847, "epoch": 2245, "lr": 1.6263813607375926e-05}
{"train_loss": 0.08139084279537201, "global_step": 199848, "epoch": 2245, "lr": 1.626338572322646e-05}
{"train_loss": 0.12036976963281631, "global_step": 199849, "epoch": 2245, "lr": 1.6262957843612448e-05}
{"train_loss": 0.07468538731336594, "global_step": 199850, "epoch": 2245, "lr": 1.626252996853399e-05}
{"train_loss": 0.041201166808605194, "global_step": 199851, "epoch": 2245, "lr": 1.6262102097991104e-05}
{"train_loss": 0.024590112268924713, "global_step": 199852, "epoch": 2245, "lr": 1.6261674231983885e-05}
{"train_loss": 0.06498152017593384, "global_step": 199853, "epoch": 2245, "lr": 1.6261246370512362e-05}
{"train_loss": 0.04579723998904228, "global_step": 199854, "epoch": 2245, "lr": 1.6260818513576616e-05}
{"train_loss": 0.05681923031806946, "global_step": 199855, "epoch": 2245, "lr": 1.6260390661176673e-05}
{"train_loss": 0.06282810866832733, "global_step": 199856, "epoch": 2245, "lr": 1.6259962813312623e-05}
{"train_loss": 0.05255459249019623, "global_step": 199857, "epoch": 2245, "lr": 1.6259534969984498e-05}
{"train_loss": 0.060793738812208176, "global_step": 199858, "epoch": 2245, "lr": 1.6259107131192387e-05}
{"train_loss": 0.015224460512399673, "global_step": 199859, "epoch": 2245, "lr": 1.625867929693632e-05}
{"train_loss": 0.05514946207404137, "global_step": 199860, "epoch": 2245, "lr": 1.6258251467216353e-05}
{"train_loss": 0.05578313022851944, "global_step": 199861, "epoch": 2245, "lr": 1.625782364203256e-05}
{"train_loss": 0.04947476089000702, "global_step": 199862, "epoch": 2245, "lr": 1.625739582138498e-05}
{"train_loss": 0.06593916565179825, "global_step": 199863, "epoch": 2245, "lr": 1.6256968005273705e-05}
{"train_loss": 0.1485442966222763, "global_step": 199864, "epoch": 2245, "lr": 1.6256540193698744e-05}
{"train_loss": 0.12193269282579422, "global_step": 199865, "epoch": 2245, "lr": 1.62561123866602e-05}
{"train_loss": 0.11388766765594482, "global_step": 199866, "epoch": 2245, "lr": 1.6255684584158092e-05}
{"train_loss": 0.021790191531181335, "global_step": 199867, "epoch": 2245, "lr": 1.6255256786192512e-05}
{"train_loss": 0.08563026040792465, "global_step": 199868, "epoch": 2245, "lr": 1.6254828992763487e-05}
{"train_loss": 0.03349001333117485, "global_step": 199869, "epoch": 2245, "lr": 1.6254401203871107e-05}
{"train_loss": 0.06313689053058624, "global_step": 199870, "epoch": 2245, "lr": 1.6253973419515388e-05}
{"train_loss": 0.05324934795498848, "global_step": 199871, "epoch": 2245, "lr": 1.625354563969644e-05}
{"train_loss": 0.017275655642151833, "global_step": 199872, "epoch": 2245, "lr": 1.6253117864414265e-05}
{"train_loss": 0.05500293895602226, "global_step": 199873, "epoch": 2245, "lr": 1.6252690093668966e-05}
{"train_loss": 0.08019676059484482, "global_step": 199874, "epoch": 2245, "lr": 1.625226232746056e-05}
{"train_loss": 0.05072525888681412, "global_step": 199875, "epoch": 2245, "lr": 1.625183456578913e-05}
{"train_loss": 0.03311936557292938, "global_step": 199876, "epoch": 2245, "lr": 1.625140680865475e-05}
{"train_loss": 0.06436225026845932, "global_step": 199877, "epoch": 2245, "lr": 1.6250979056057437e-05}
{"train_loss": 0.01845797710120678, "global_step": 199878, "epoch": 2245, "lr": 1.6250551307997286e-05}
{"train_loss": 0.0920179933309555, "global_step": 199879, "epoch": 2245, "lr": 1.6250123564474318e-05}
{"train_loss": 0.16653519868850708, "global_step": 199880, "epoch": 2245, "lr": 1.624969582548863e-05}
{"train_loss": 0.09345109015703201, "global_step": 199881, "epoch": 2245, "lr": 1.6249268091040236e-05}
{"train_loss": 0.05546592175960541, "global_step": 199882, "epoch": 2245, "lr": 1.6248840361129243e-05}
{"train_loss": 0.0923086404800415, "global_step": 199883, "epoch": 2245, "lr": 1.6248412635755657e-05}
{"train_loss": 0.05435635522007942, "global_step": 199884, "epoch": 2245, "lr": 1.624798491491958e-05}
{"train_loss": 0.05379251018166542, "global_step": 199885, "epoch": 2245, "lr": 1.6247557198621033e-05}
{"train_loss": 0.061163514852523804, "global_step": 199886, "epoch": 2245, "lr": 1.6247129486860106e-05}
{"train_loss": 0.10096138715744019, "global_step": 199887, "epoch": 2245, "lr": 1.624670177963683e-05}
{"train_loss": 0.051180653274059296, "global_step": 199888, "epoch": 2245, "lr": 1.6246274076951286e-05}
{"train_loss": 0.10204585641622543, "global_step": 199889, "epoch": 2245, "lr": 1.6245846378803504e-05}
{"train_loss": 0.04421057552099228, "global_step": 199890, "epoch": 2245, "lr": 1.624541868519357e-05}
{"train_loss": 0.04014987871050835, "global_step": 199891, "epoch": 2245, "lr": 1.624499099612151e-05}
{"train_loss": 0.024865729734301567, "global_step": 199892, "epoch": 2245, "lr": 1.6244563311587425e-05}
{"train_loss": 0.06340100626680958, "global_step": 199893, "epoch": 2245, "lr": 1.624413563159132e-05, "val_loss": 8.112224578857422, "train_action_mse_error": 14.544583320617676}
{"train_loss": 0.03853941708803177, "global_step": 199894, "epoch": 2246, "lr": 1.6243707956133302e-05}
{"train_loss": 0.027514930814504623, "global_step": 199895, "epoch": 2246, "lr": 1.624328028521339e-05}
{"train_loss": 0.0559106208384037, "global_step": 199896, "epoch": 2246, "lr": 1.6242852618831672e-05}
{"train_loss": 0.06947996467351913, "global_step": 199897, "epoch": 2246, "lr": 1.624242495698818e-05}
{"train_loss": 0.04315934330224991, "global_step": 199898, "epoch": 2246, "lr": 1.6241997299682994e-05}
{"train_loss": 0.04773212969303131, "global_step": 199899, "epoch": 2246, "lr": 1.6241569646916154e-05}
{"train_loss": 0.0926305279135704, "global_step": 199900, "epoch": 2246, "lr": 1.6241141998687713e-05}
{"train_loss": 0.022754129022359848, "global_step": 199901, "epoch": 2246, "lr": 1.624071435499776e-05}
{"train_loss": 0.1052427738904953, "global_step": 199902, "epoch": 2246, "lr": 1.624028671584631e-05}
{"train_loss": 0.058365896344184875, "global_step": 199903, "epoch": 2246, "lr": 1.623985908123346e-05}
{"train_loss": 0.0908961147069931, "global_step": 199904, "epoch": 2246, "lr": 1.6239431451159232e-05}
{"train_loss": 0.0870366245508194, "global_step": 199905, "epoch": 2246, "lr": 1.6239003825623715e-05}
{"train_loss": 0.10748394578695297, "global_step": 199906, "epoch": 2246, "lr": 1.6238576204626936e-05}
{"train_loss": 0.05635681375861168, "global_step": 199907, "epoch": 2246, "lr": 1.623814858816899e-05}
{"train_loss": 0.066994309425354, "global_step": 199908, "epoch": 2246, "lr": 1.6237720976249897e-05}
{"train_loss": 0.05362457036972046, "global_step": 199909, "epoch": 2246, "lr": 1.6237293368869745e-05}
{"train_loss": 0.06263427436351776, "global_step": 199910, "epoch": 2246, "lr": 1.623686576602856e-05}
{"train_loss": 0.041161417961120605, "global_step": 199911, "epoch": 2246, "lr": 1.623643816772642e-05}
{"train_loss": 0.03818153217434883, "global_step": 199912, "epoch": 2246, "lr": 1.6236010573963395e-05}
{"train_loss": 0.0822276771068573, "global_step": 199913, "epoch": 2246, "lr": 1.6235582984739506e-05}
{"train_loss": 0.06989598274230957, "global_step": 199914, "epoch": 2246, "lr": 1.6235155400054854e-05}
{"train_loss": 0.08350792527198792, "global_step": 199915, "epoch": 2246, "lr": 1.623472781990945e-05}
{"train_loss": 0.06737800687551498, "global_step": 199916, "epoch": 2246, "lr": 1.6234300244303396e-05}
{"train_loss": 0.09218218177556992, "global_step": 199917, "epoch": 2246, "lr": 1.6233872673236715e-05}
{"train_loss": 0.06917428970336914, "global_step": 199918, "epoch": 2246, "lr": 1.6233445106709495e-05}
{"train_loss": 0.04239194840192795, "global_step": 199919, "epoch": 2246, "lr": 1.6233017544721757e-05}
{"train_loss": 0.05349401384592056, "global_step": 199920, "epoch": 2246, "lr": 1.6232589987273595e-05}
{"train_loss": 0.09825367480516434, "global_step": 199921, "epoch": 2246, "lr": 1.6232162434365034e-05}
{"train_loss": 0.06542739272117615, "global_step": 199922, "epoch": 2246, "lr": 1.6231734885996157e-05}
{"train_loss": 0.08485438674688339, "global_step": 199923, "epoch": 2246, "lr": 1.6231307342167007e-05}
{"train_loss": 0.02662530168890953, "global_step": 199924, "epoch": 2246, "lr": 1.623087980287766e-05}
{"train_loss": 0.07290226966142654, "global_step": 199925, "epoch": 2246, "lr": 1.6230452268128132e-05}
{"train_loss": 0.11976472288370132, "global_step": 199926, "epoch": 2246, "lr": 1.6230024737918537e-05}
{"train_loss": 0.0710807666182518, "global_step": 199927, "epoch": 2246, "lr": 1.6229597212248882e-05}
{"train_loss": 0.06915024667978287, "global_step": 199928, "epoch": 2246, "lr": 1.6229169691119263e-05}
{"train_loss": 0.046571049839258194, "global_step": 199929, "epoch": 2246, "lr": 1.6228742174529703e-05}
{"train_loss": 0.043661653995513916, "global_step": 199930, "epoch": 2246, "lr": 1.622831466248028e-05}
{"train_loss": 0.04442942515015602, "global_step": 199931, "epoch": 2246, "lr": 1.622788715497106e-05}
{"train_loss": 0.05733605846762657, "global_step": 199932, "epoch": 2246, "lr": 1.6227459652002098e-05}
{"train_loss": 0.06744170933961868, "global_step": 199933, "epoch": 2246, "lr": 1.622703215357342e-05}
{"train_loss": 0.05535271018743515, "global_step": 199934, "epoch": 2246, "lr": 1.6226604659685118e-05}
{"train_loss": 0.11701777577400208, "global_step": 199935, "epoch": 2246, "lr": 1.6226177170337226e-05}
{"train_loss": 0.09985014796257019, "global_step": 199936, "epoch": 2246, "lr": 1.622574968552983e-05}
{"train_loss": 0.027590928599238396, "global_step": 199937, "epoch": 2246, "lr": 1.6225322205262967e-05}
{"train_loss": 0.05346538871526718, "global_step": 199938, "epoch": 2246, "lr": 1.6224894729536682e-05}
{"train_loss": 0.04743429645895958, "global_step": 199939, "epoch": 2246, "lr": 1.6224467258351066e-05}
{"train_loss": 0.07276094704866409, "global_step": 199940, "epoch": 2246, "lr": 1.6224039791706147e-05}
{"train_loss": 0.10803591459989548, "global_step": 199941, "epoch": 2246, "lr": 1.6223612329602005e-05}
{"train_loss": 0.08905056864023209, "global_step": 199942, "epoch": 2246, "lr": 1.622318487203867e-05}
{"train_loss": 0.04688388854265213, "global_step": 199943, "epoch": 2246, "lr": 1.6222757419016234e-05}
{"train_loss": 0.0462690070271492, "global_step": 199944, "epoch": 2246, "lr": 1.6222329970534723e-05}
{"train_loss": 0.05901720002293587, "global_step": 199945, "epoch": 2246, "lr": 1.6221902526594217e-05}
{"train_loss": 0.08901713043451309, "global_step": 199946, "epoch": 2246, "lr": 1.6221475087194753e-05}
{"train_loss": 0.032839514315128326, "global_step": 199947, "epoch": 2246, "lr": 1.62210476523364e-05}
{"train_loss": 0.041332848370075226, "global_step": 199948, "epoch": 2246, "lr": 1.6220620222019234e-05}
{"train_loss": 0.10168087482452393, "global_step": 199949, "epoch": 2246, "lr": 1.6220192796243272e-05}
{"train_loss": 0.0767388716340065, "global_step": 199950, "epoch": 2246, "lr": 1.6219765375008612e-05}
{"train_loss": 0.059289783239364624, "global_step": 199951, "epoch": 2246, "lr": 1.621933795831528e-05}
{"train_loss": 0.043793708086013794, "global_step": 199952, "epoch": 2246, "lr": 1.6218910546163357e-05}
{"train_loss": 0.06663345545530319, "global_step": 199953, "epoch": 2246, "lr": 1.6218483138552876e-05}
{"train_loss": 0.04721224308013916, "global_step": 199954, "epoch": 2246, "lr": 1.6218055735483922e-05}
{"train_loss": 0.10772748291492462, "global_step": 199955, "epoch": 2246, "lr": 1.6217628336956525e-05}
{"train_loss": 0.09520736336708069, "global_step": 199956, "epoch": 2246, "lr": 1.6217200942970773e-05}
{"train_loss": 0.04867277666926384, "global_step": 199957, "epoch": 2246, "lr": 1.6216773553526693e-05}
{"train_loss": 0.07895228266716003, "global_step": 199958, "epoch": 2246, "lr": 1.621634616862437e-05}
{"train_loss": 0.07064703851938248, "global_step": 199959, "epoch": 2246, "lr": 1.6215918788263827e-05}
{"train_loss": 0.09637222439050674, "global_step": 199960, "epoch": 2246, "lr": 1.621549141244516e-05}
{"train_loss": 0.030972154811024666, "global_step": 199961, "epoch": 2246, "lr": 1.621506404116839e-05}
{"train_loss": 0.05180010944604874, "global_step": 199962, "epoch": 2246, "lr": 1.621463667443362e-05}
{"train_loss": 0.04531626030802727, "global_step": 199963, "epoch": 2246, "lr": 1.6214209312240853e-05}
{"train_loss": 0.017902029678225517, "global_step": 199964, "epoch": 2246, "lr": 1.6213781954590196e-05}
{"train_loss": 0.04342939704656601, "global_step": 199965, "epoch": 2246, "lr": 1.621335460148166e-05}
{"train_loss": 0.07527744770050049, "global_step": 199966, "epoch": 2246, "lr": 1.621292725291534e-05}
{"train_loss": 0.10274457931518555, "global_step": 199967, "epoch": 2246, "lr": 1.6212499908891287e-05}
{"train_loss": 0.11801649630069733, "global_step": 199968, "epoch": 2246, "lr": 1.6212072569409537e-05}
{"train_loss": 0.03259729593992233, "global_step": 199969, "epoch": 2246, "lr": 1.6211645234470174e-05}
{"train_loss": 0.08043473958969116, "global_step": 199970, "epoch": 2246, "lr": 1.6211217904073252e-05}
{"train_loss": 0.12833911180496216, "global_step": 199971, "epoch": 2246, "lr": 1.6210790578218798e-05}
{"train_loss": 0.045912448316812515, "global_step": 199972, "epoch": 2246, "lr": 1.6210363256906908e-05}
{"train_loss": 0.0639100894331932, "global_step": 199973, "epoch": 2246, "lr": 1.620993594013761e-05}
{"train_loss": 0.03218451142311096, "global_step": 199974, "epoch": 2246, "lr": 1.6209508627910985e-05}
{"train_loss": 0.03262825310230255, "global_step": 199975, "epoch": 2246, "lr": 1.6209081320227082e-05}
{"train_loss": 0.061895281076431274, "global_step": 199976, "epoch": 2246, "lr": 1.6208654017085946e-05}
{"train_loss": 0.08842788636684418, "global_step": 199977, "epoch": 2246, "lr": 1.620822671848765e-05}
{"train_loss": 0.032515596598386765, "global_step": 199978, "epoch": 2246, "lr": 1.6207799424432236e-05}
{"train_loss": 0.04123029485344887, "global_step": 199979, "epoch": 2246, "lr": 1.620737213491979e-05}
{"train_loss": 0.04508255422115326, "global_step": 199980, "epoch": 2246, "lr": 1.6206944849950328e-05}
{"train_loss": 0.06779170781373978, "global_step": 199981, "epoch": 2246, "lr": 1.6206517569523955e-05}
{"train_loss": 0.06517276835491818, "global_step": 199982, "epoch": 2246, "lr": 1.620609029364068e-05, "val_loss": 8.217486381530762}
{"train_loss": 0.04215918108820915, "global_step": 199983, "epoch": 2247, "lr": 1.6205663022300604e-05}
{"train_loss": 0.05483902245759964, "global_step": 199984, "epoch": 2247, "lr": 1.620523575550375e-05}
{"train_loss": 0.06801068782806396, "global_step": 199985, "epoch": 2247, "lr": 1.620480849325019e-05}
{"train_loss": 0.062499504536390305, "global_step": 199986, "epoch": 2247, "lr": 1.6204381235539996e-05}
{"train_loss": 0.05158620327711105, "global_step": 199987, "epoch": 2247, "lr": 1.6203953982373194e-05}
{"train_loss": 0.03069264441728592, "global_step": 199988, "epoch": 2247, "lr": 1.6203526733749873e-05}
{"train_loss": 0.05293918028473854, "global_step": 199989, "epoch": 2247, "lr": 1.6203099489670063e-05}
{"train_loss": 0.07272747159004211, "global_step": 199990, "epoch": 2247, "lr": 1.620267225013385e-05}
{"train_loss": 0.05426279455423355, "global_step": 199991, "epoch": 2247, "lr": 1.620224501514126e-05}
{"train_loss": 0.11455382406711578, "global_step": 199992, "epoch": 2247, "lr": 1.620181778469238e-05}
{"train_loss": 0.0852508619427681, "global_step": 199993, "epoch": 2247, "lr": 1.620139055878724e-05}
{"train_loss": 0.055329687893390656, "global_step": 199994, "epoch": 2247, "lr": 1.6200963337425927e-05}
{"train_loss": 0.059672947973012924, "global_step": 199995, "epoch": 2247, "lr": 1.6200536120608466e-05}
{"train_loss": 0.035326529294252396, "global_step": 199996, "epoch": 2247, "lr": 1.620010890833495e-05}
{"train_loss": 0.12276013195514679, "global_step": 199997, "epoch": 2247, "lr": 1.6199681700605394e-05}
{"train_loss": 0.07337264716625214, "global_step": 199998, "epoch": 2247, "lr": 1.6199254497419902e-05}
{"train_loss": 0.0422203503549099, "global_step": 199999, "epoch": 2247, "lr": 1.6198827298778484e-05}
{"train_loss": 0.03566874563694, "global_step": 200000, "epoch": 2247, "lr": 1.6198400104681245e-05}
{"train_loss": 0.11115197837352753, "global_step": 200001, "epoch": 2247, "lr": 1.6197972915128203e-05}
{"train_loss": 0.03182046115398407, "global_step": 200002, "epoch": 2247, "lr": 1.6197545730119427e-05}
{"train_loss": 0.0820765346288681, "global_step": 200003, "epoch": 2247, "lr": 1.6197118549655e-05}
{"train_loss": 0.054349131882190704, "global_step": 200004, "epoch": 2247, "lr": 1.619669137373494e-05}
{"train_loss": 0.06986121088266373, "global_step": 200005, "epoch": 2247, "lr": 1.619626420235934e-05}
{"train_loss": 0.044264476746320724, "global_step": 200006, "epoch": 2247, "lr": 1.619583703552822e-05}
{"train_loss": 0.03967029228806496, "global_step": 200007, "epoch": 2247, "lr": 1.619540987324168e-05}
{"train_loss": 0.03154979273676872, "global_step": 200008, "epoch": 2247, "lr": 1.6194982715499736e-05}
{"train_loss": 0.0479687824845314, "global_step": 200009, "epoch": 2247, "lr": 1.6194555562302483e-05}
{"train_loss": 0.05211976543068886, "global_step": 200010, "epoch": 2247, "lr": 1.6194128413649957e-05}
{"train_loss": 0.08302831649780273, "global_step": 200011, "epoch": 2247, "lr": 1.61937012695422e-05}
{"train_loss": 0.014180942438542843, "global_step": 200012, "epoch": 2247, "lr": 1.6193274129979308e-05}
{"train_loss": 0.09637987613677979, "global_step": 200013, "epoch": 2247, "lr": 1.6192846994961303e-05}
{"train_loss": 0.09530709683895111, "global_step": 200014, "epoch": 2247, "lr": 1.6192419864488266e-05}
{"train_loss": 0.032631147652864456, "global_step": 200015, "epoch": 2247, "lr": 1.6191992738560256e-05}
{"train_loss": 0.05097462981939316, "global_step": 200016, "epoch": 2247, "lr": 1.61915656171773e-05}
{"train_loss": 0.12544041872024536, "global_step": 200017, "epoch": 2247, "lr": 1.619113850033949e-05}
{"train_loss": 0.07210765033960342, "global_step": 200018, "epoch": 2247, "lr": 1.6190711388046858e-05}
{"train_loss": 0.04202909395098686, "global_step": 200019, "epoch": 2247, "lr": 1.6190284280299485e-05}
{"train_loss": 0.08932074159383774, "global_step": 200020, "epoch": 2247, "lr": 1.61898571770974e-05}
{"train_loss": 0.0679461732506752, "global_step": 200021, "epoch": 2247, "lr": 1.6189430078440682e-05}
{"train_loss": 0.10053839534521103, "global_step": 200022, "epoch": 2247, "lr": 1.6189002984329394e-05}
{"train_loss": 0.09752513468265533, "global_step": 200023, "epoch": 2247, "lr": 1.618857589476357e-05}
{"train_loss": 0.11045350879430771, "global_step": 200024, "epoch": 2247, "lr": 1.618814880974329e-05}
{"train_loss": 0.07153769582509995, "global_step": 200025, "epoch": 2247, "lr": 1.618772172926859e-05}
{"train_loss": 0.1023377850651741, "global_step": 200026, "epoch": 2247, "lr": 1.6187294653339552e-05}
{"train_loss": 0.036349765956401825, "global_step": 200027, "epoch": 2247, "lr": 1.6186867581956204e-05}
{"train_loss": 0.0909675732254982, "global_step": 200028, "epoch": 2247, "lr": 1.618644051511864e-05}
{"train_loss": 0.04907841607928276, "global_step": 200029, "epoch": 2247, "lr": 1.6186013452826876e-05}
{"train_loss": 0.09717492014169693, "global_step": 200030, "epoch": 2247, "lr": 1.618558639508101e-05}
{"train_loss": 0.08558497577905655, "global_step": 200031, "epoch": 2247, "lr": 1.618515934188106e-05}
{"train_loss": 0.08594562858343124, "global_step": 200032, "epoch": 2247, "lr": 1.618473229322712e-05}
{"train_loss": 0.03344759717583656, "global_step": 200033, "epoch": 2247, "lr": 1.618430524911922e-05}
{"train_loss": 0.04823155328631401, "global_step": 200034, "epoch": 2247, "lr": 1.618387820955744e-05}
{"train_loss": 0.055749278515577316, "global_step": 200035, "epoch": 2247, "lr": 1.618345117454181e-05}
{"train_loss": 0.10018927603960037, "global_step": 200036, "epoch": 2247, "lr": 1.6183024144072416e-05}
{"train_loss": 0.0800924301147461, "global_step": 200037, "epoch": 2247, "lr": 1.618259711814929e-05}
{"train_loss": 0.041816096752882004, "global_step": 200038, "epoch": 2247, "lr": 1.6182170096772505e-05}
{"train_loss": 0.06217779219150543, "global_step": 200039, "epoch": 2247, "lr": 1.6181743079942125e-05}
{"train_loss": 0.08049540966749191, "global_step": 200040, "epoch": 2247, "lr": 1.618131606765818e-05}
{"train_loss": 0.05794975161552429, "global_step": 200041, "epoch": 2247, "lr": 1.6180889059920764e-05}
{"train_loss": 0.04965372756123543, "global_step": 200042, "epoch": 2247, "lr": 1.61804620567299e-05}
{"train_loss": 0.027725404128432274, "global_step": 200043, "epoch": 2247, "lr": 1.6180035058085674e-05}
{"train_loss": 0.06992268562316895, "global_step": 200044, "epoch": 2247, "lr": 1.6179608063988117e-05}
{"train_loss": 0.07241687923669815, "global_step": 200045, "epoch": 2247, "lr": 1.617918107443731e-05}
{"train_loss": 0.028221838176250458, "global_step": 200046, "epoch": 2247, "lr": 1.617875408943329e-05}
{"train_loss": 0.07057540863752365, "global_step": 200047, "epoch": 2247, "lr": 1.6178327108976136e-05}
{"train_loss": 0.059051383286714554, "global_step": 200048, "epoch": 2247, "lr": 1.6177900133065898e-05}
{"train_loss": 0.04229866340756416, "global_step": 200049, "epoch": 2247, "lr": 1.617747316170261e-05}
{"train_loss": 0.10912829637527466, "global_step": 200050, "epoch": 2247, "lr": 1.617704619488637e-05}
{"train_loss": 0.052106067538261414, "global_step": 200051, "epoch": 2247, "lr": 1.617661923261719e-05}
{"train_loss": 0.04410505294799805, "global_step": 200052, "epoch": 2247, "lr": 1.617619227489517e-05}
{"train_loss": 0.07116999477148056, "global_step": 200053, "epoch": 2247, "lr": 1.6175765321720348e-05}
{"train_loss": 0.06642220169305801, "global_step": 200054, "epoch": 2247, "lr": 1.617533837309277e-05}
{"train_loss": 0.04909256473183632, "global_step": 200055, "epoch": 2247, "lr": 1.6174911429012518e-05}
{"train_loss": 0.034049469977617264, "global_step": 200056, "epoch": 2247, "lr": 1.6174484489479624e-05}
{"train_loss": 0.04363695904612541, "global_step": 200057, "epoch": 2247, "lr": 1.6174057554494155e-05}
{"train_loss": 0.05559110641479492, "global_step": 200058, "epoch": 2247, "lr": 1.6173630624056192e-05}
{"train_loss": 0.03590329736471176, "global_step": 200059, "epoch": 2247, "lr": 1.617320369816575e-05}
{"train_loss": 0.05222040042281151, "global_step": 200060, "epoch": 2247, "lr": 1.6172776776822928e-05}
{"train_loss": 0.016158368438482285, "global_step": 200061, "epoch": 2247, "lr": 1.6172349860027746e-05}
{"train_loss": 0.03436945378780365, "global_step": 200062, "epoch": 2247, "lr": 1.61719229477803e-05}
{"train_loss": 0.08737901598215103, "global_step": 200063, "epoch": 2247, "lr": 1.61714960400806e-05}
{"train_loss": 0.05322028696537018, "global_step": 200064, "epoch": 2247, "lr": 1.6171069136928758e-05}
{"train_loss": 0.05710369721055031, "global_step": 200065, "epoch": 2247, "lr": 1.617064223832478e-05}
{"train_loss": 0.023208677768707275, "global_step": 200066, "epoch": 2247, "lr": 1.617021534426877e-05}
{"train_loss": 0.0471918061375618, "global_step": 200067, "epoch": 2247, "lr": 1.6169788454760737e-05}
{"train_loss": 0.029800714924931526, "global_step": 200068, "epoch": 2247, "lr": 1.6169361569800783e-05}
{"train_loss": 0.04187244549393654, "global_step": 200069, "epoch": 2247, "lr": 1.616893468938893e-05}
{"train_loss": 0.06259163469076157, "global_step": 200070, "epoch": 2247, "lr": 1.6168507813525275e-05}
{"train_loss": 0.06128578029279964, "global_step": 200071, "epoch": 2247, "lr": 1.616808094220983e-05, "val_loss": 8.25011920928955}
{"train_loss": 0.11894441395998001, "global_step": 200072, "epoch": 2248, "lr": 1.6167654075442685e-05}
{"train_loss": 0.02954855002462864, "global_step": 200073, "epoch": 2248, "lr": 1.616722721322388e-05}
{"train_loss": 0.02527790702879429, "global_step": 200074, "epoch": 2248, "lr": 1.616680035555347e-05}
{"train_loss": 0.07035493105649948, "global_step": 200075, "epoch": 2248, "lr": 1.6166373502431543e-05}
{"train_loss": 0.025998417288064957, "global_step": 200076, "epoch": 2248, "lr": 1.6165946653858115e-05}
{"train_loss": 0.03689523786306381, "global_step": 200077, "epoch": 2248, "lr": 1.6165519809833285e-05}
{"train_loss": 0.09355787187814713, "global_step": 200078, "epoch": 2248, "lr": 1.6165092970357065e-05}
{"train_loss": 0.1389564424753189, "global_step": 200079, "epoch": 2248, "lr": 1.616466613542956e-05}
{"train_loss": 0.05101446062326431, "global_step": 200080, "epoch": 2248, "lr": 1.616423930505078e-05}
{"train_loss": 0.08597097545862198, "global_step": 200081, "epoch": 2248, "lr": 1.6163812479220825e-05}
{"train_loss": 0.0793333426117897, "global_step": 200082, "epoch": 2248, "lr": 1.6163385657939717e-05}
{"train_loss": 0.022260606288909912, "global_step": 200083, "epoch": 2248, "lr": 1.6162958841207542e-05}
{"train_loss": 0.07040350884199142, "global_step": 200084, "epoch": 2248, "lr": 1.6162532029024337e-05}
{"train_loss": 0.07349453866481781, "global_step": 200085, "epoch": 2248, "lr": 1.6162105221390173e-05}
{"train_loss": 0.09286189824342728, "global_step": 200086, "epoch": 2248, "lr": 1.6161678418305108e-05}
{"train_loss": 0.03661803528666496, "global_step": 200087, "epoch": 2248, "lr": 1.6161251619769176e-05}
{"train_loss": 0.03981875628232956, "global_step": 200088, "epoch": 2248, "lr": 1.6160824825782468e-05}
{"train_loss": 0.057067323476076126, "global_step": 200089, "epoch": 2248, "lr": 1.6160398036345005e-05}
{"train_loss": 0.0972871258854866, "global_step": 200090, "epoch": 2248, "lr": 1.615997125145688e-05}
{"train_loss": 0.03910724073648453, "global_step": 200091, "epoch": 2248, "lr": 1.615954447111812e-05}
{"train_loss": 0.030040603131055832, "global_step": 200092, "epoch": 2248, "lr": 1.6159117695328813e-05}
{"train_loss": 0.12255179136991501, "global_step": 200093, "epoch": 2248, "lr": 1.6158690924088986e-05}
{"train_loss": 0.09672196954488754, "global_step": 200094, "epoch": 2248, "lr": 1.6158264157398727e-05}
{"train_loss": 0.064348965883255, "global_step": 200095, "epoch": 2248, "lr": 1.6157837395258056e-05}
{"train_loss": 0.03453421965241432, "global_step": 200096, "epoch": 2248, "lr": 1.615741063766707e-05}
{"train_loss": 0.09349089860916138, "global_step": 200097, "epoch": 2248, "lr": 1.6156983884625792e-05}
{"train_loss": 0.03115518018603325, "global_step": 200098, "epoch": 2248, "lr": 1.6156557136134315e-05}
{"train_loss": 0.07246071100234985, "global_step": 200099, "epoch": 2248, "lr": 1.6156130392192655e-05}
{"train_loss": 0.10228835791349411, "global_step": 200100, "epoch": 2248, "lr": 1.6155703652800906e-05}
{"train_loss": 0.11417209357023239, "global_step": 200101, "epoch": 2248, "lr": 1.6155276917959094e-05}
{"train_loss": 0.040319278836250305, "global_step": 200102, "epoch": 2248, "lr": 1.6154850187667315e-05}
{"train_loss": 0.048605918884277344, "global_step": 200103, "epoch": 2248, "lr": 1.6154423461925584e-05}
{"train_loss": 0.04816220700740814, "global_step": 200104, "epoch": 2248, "lr": 1.615399674073399e-05}
{"train_loss": 0.047693055123090744, "global_step": 200105, "epoch": 2248, "lr": 1.6153570024092567e-05}
{"train_loss": 0.059159476310014725, "global_step": 200106, "epoch": 2248, "lr": 1.6153143312001402e-05}
{"train_loss": 0.06849119067192078, "global_step": 200107, "epoch": 2248, "lr": 1.6152716604460515e-05}
{"train_loss": 0.05402860417962074, "global_step": 200108, "epoch": 2248, "lr": 1.615228990147e-05}
{"train_loss": 0.06182890012860298, "global_step": 200109, "epoch": 2248, "lr": 1.6151863203029883e-05}
{"train_loss": 0.048996537923812866, "global_step": 200110, "epoch": 2248, "lr": 1.6151436509140244e-05}
{"train_loss": 0.061262793838977814, "global_step": 200111, "epoch": 2248, "lr": 1.6151009819801122e-05}
{"train_loss": 0.058518774807453156, "global_step": 200112, "epoch": 2248, "lr": 1.615058313501258e-05}
{"train_loss": 0.0844404399394989, "global_step": 200113, "epoch": 2248, "lr": 1.61501564547747e-05}
{"train_loss": 0.03841644898056984, "global_step": 200114, "epoch": 2248, "lr": 1.6149729779087503e-05}
{"train_loss": 0.05079992860555649, "global_step": 200115, "epoch": 2248, "lr": 1.614930310795108e-05}
{"train_loss": 0.06106732413172722, "global_step": 200116, "epoch": 2248, "lr": 1.614887644136545e-05}
{"train_loss": 0.0359157994389534, "global_step": 200117, "epoch": 2248, "lr": 1.614844977933071e-05}
{"train_loss": 0.061819035559892654, "global_step": 200118, "epoch": 2248, "lr": 1.6148023121846877e-05}
{"train_loss": 0.07602211833000183, "global_step": 200119, "epoch": 2248, "lr": 1.614759646891405e-05}
{"train_loss": 0.06769467890262604, "global_step": 200120, "epoch": 2248, "lr": 1.614716982053225e-05}
{"train_loss": 0.06442899256944656, "global_step": 200121, "epoch": 2248, "lr": 1.6146743176701566e-05}
{"train_loss": 0.10092591494321823, "global_step": 200122, "epoch": 2248, "lr": 1.6146316537422024e-05}
{"train_loss": 0.06670878082513809, "global_step": 200123, "epoch": 2248, "lr": 1.6145889902693712e-05}
{"train_loss": 0.05623411387205124, "global_step": 200124, "epoch": 2248, "lr": 1.6145463272516657e-05}
{"train_loss": 0.047252751886844635, "global_step": 200125, "epoch": 2248, "lr": 1.6145036646890948e-05}
{"train_loss": 0.10780882835388184, "global_step": 200126, "epoch": 2248, "lr": 1.6144610025816624e-05}
{"train_loss": 0.09626385569572449, "global_step": 200127, "epoch": 2248, "lr": 1.6144183409293734e-05}
{"train_loss": 0.07835184782743454, "global_step": 200128, "epoch": 2248, "lr": 1.6143756797322358e-05}
{"train_loss": 0.060561757534742355, "global_step": 200129, "epoch": 2248, "lr": 1.6143330189902524e-05}
{"train_loss": 0.05079660937190056, "global_step": 200130, "epoch": 2248, "lr": 1.6142903587034326e-05}
{"train_loss": 0.06258237361907959, "global_step": 200131, "epoch": 2248, "lr": 1.6142476988717786e-05}
{"train_loss": 0.06017085909843445, "global_step": 200132, "epoch": 2248, "lr": 1.614205039495299e-05}
{"train_loss": 0.0794837549328804, "global_step": 200133, "epoch": 2248, "lr": 1.614162380573997e-05}
{"train_loss": 0.07181902229785919, "global_step": 200134, "epoch": 2248, "lr": 1.6141197221078807e-05}
{"train_loss": 0.05818293243646622, "global_step": 200135, "epoch": 2248, "lr": 1.6140770640969538e-05}
{"train_loss": 0.026872830465435982, "global_step": 200136, "epoch": 2248, "lr": 1.6140344065412237e-05}
{"train_loss": 0.07602637261152267, "global_step": 200137, "epoch": 2248, "lr": 1.613991749440695e-05}
{"train_loss": 0.05130380764603615, "global_step": 200138, "epoch": 2248, "lr": 1.6139490927953745e-05}
{"train_loss": 0.04152954742312431, "global_step": 200139, "epoch": 2248, "lr": 1.6139064366052658e-05}
{"train_loss": 0.054501235485076904, "global_step": 200140, "epoch": 2248, "lr": 1.6138637808703784e-05}
{"train_loss": 0.04467006400227547, "global_step": 200141, "epoch": 2248, "lr": 1.6138211255907133e-05}
{"train_loss": 0.032368991523981094, "global_step": 200142, "epoch": 2248, "lr": 1.6137784707662806e-05}
{"train_loss": 0.04284187778830528, "global_step": 200143, "epoch": 2248, "lr": 1.6137358163970824e-05}
{"train_loss": 0.07385018467903137, "global_step": 200144, "epoch": 2248, "lr": 1.613693162483128e-05}
{"train_loss": 0.038662705570459366, "global_step": 200145, "epoch": 2248, "lr": 1.6136505090244192e-05}
{"train_loss": 0.1012686938047409, "global_step": 200146, "epoch": 2248, "lr": 1.613607856020966e-05}
{"train_loss": 0.05251186341047287, "global_step": 200147, "epoch": 2248, "lr": 1.61356520347277e-05}
{"train_loss": 0.03682286664843559, "global_step": 200148, "epoch": 2248, "lr": 1.613522551379839e-05}
{"train_loss": 0.061432067304849625, "global_step": 200149, "epoch": 2248, "lr": 1.6134798997421802e-05}
{"train_loss": 0.06505037844181061, "global_step": 200150, "epoch": 2248, "lr": 1.613437248559796e-05}
{"train_loss": 0.0645245760679245, "global_step": 200151, "epoch": 2248, "lr": 1.613394597832696e-05}
{"train_loss": 0.07100818306207657, "global_step": 200152, "epoch": 2248, "lr": 1.6133519475608816e-05}
{"train_loss": 0.0199857447296381, "global_step": 200153, "epoch": 2248, "lr": 1.6133092977443625e-05}
{"train_loss": 0.04992692917585373, "global_step": 200154, "epoch": 2248, "lr": 1.6132666483831416e-05}
{"train_loss": 0.04604005441069603, "global_step": 200155, "epoch": 2248, "lr": 1.613223999477227e-05}
{"train_loss": 0.06906629353761673, "global_step": 200156, "epoch": 2248, "lr": 1.6131813510266213e-05}
{"train_loss": 0.08865813165903091, "global_step": 200157, "epoch": 2248, "lr": 1.6131387030313338e-05}
{"train_loss": 0.06075206398963928, "global_step": 200158, "epoch": 2248, "lr": 1.613096055491367e-05}
{"train_loss": 0.05892620608210564, "global_step": 200159, "epoch": 2248, "lr": 1.6130534084067295e-05}
{"train_loss": 0.06248047830683462, "global_step": 200160, "epoch": 2248, "lr": 1.6130107617774243e-05, "val_loss": 8.12010383605957}
{"train_loss": 0.10110697150230408, "global_step": 200161, "epoch": 2249, "lr": 1.6129681156034603e-05}
{"train_loss": 0.07921711355447769, "global_step": 200162, "epoch": 2249, "lr": 1.6129254698848395e-05}
{"train_loss": 0.061994265764951706, "global_step": 200163, "epoch": 2249, "lr": 1.6128828246215717e-05}
{"train_loss": 0.026857221499085426, "global_step": 200164, "epoch": 2249, "lr": 1.61284017981366e-05}
{"train_loss": 0.10013595968484879, "global_step": 200165, "epoch": 2249, "lr": 1.6127975354611093e-05}
{"train_loss": 0.08440801501274109, "global_step": 200166, "epoch": 2249, "lr": 1.6127548915639284e-05}
{"train_loss": 0.04120691120624542, "global_step": 200167, "epoch": 2249, "lr": 1.6127122481221197e-05}
{"train_loss": 0.03234576806426048, "global_step": 200168, "epoch": 2249, "lr": 1.6126696051356917e-05}
{"train_loss": 0.08127301186323166, "global_step": 200169, "epoch": 2249, "lr": 1.6126269626046477e-05}
{"train_loss": 0.06543121486902237, "global_step": 200170, "epoch": 2249, "lr": 1.6125843205289963e-05}
{"train_loss": 0.05560200288891792, "global_step": 200171, "epoch": 2249, "lr": 1.61254167890874e-05}
{"train_loss": 0.08445514738559723, "global_step": 200172, "epoch": 2249, "lr": 1.6124990377438882e-05}
{"train_loss": 0.06675724685192108, "global_step": 200173, "epoch": 2249, "lr": 1.6124563970344424e-05}
{"train_loss": 0.09446366876363754, "global_step": 200174, "epoch": 2249, "lr": 1.6124137567804126e-05}
{"train_loss": 0.046218402683734894, "global_step": 200175, "epoch": 2249, "lr": 1.6123711169818006e-05}
{"train_loss": 0.12514737248420715, "global_step": 200176, "epoch": 2249, "lr": 1.6123284776386156e-05}
{"train_loss": 0.06719793379306793, "global_step": 200177, "epoch": 2249, "lr": 1.6122858387508604e-05}
{"train_loss": 0.1086091473698616, "global_step": 200178, "epoch": 2249, "lr": 1.6122432003185434e-05}
{"train_loss": 0.014087304472923279, "global_step": 200179, "epoch": 2249, "lr": 1.6122005623416676e-05}
{"train_loss": 0.05680035427212715, "global_step": 200180, "epoch": 2249, "lr": 1.6121579248202416e-05}
{"train_loss": 0.0402945913374424, "global_step": 200181, "epoch": 2249, "lr": 1.6121152877542684e-05}
{"train_loss": 0.051720619201660156, "global_step": 200182, "epoch": 2249, "lr": 1.6120726511437563e-05}
{"train_loss": 0.026034193113446236, "global_step": 200183, "epoch": 2249, "lr": 1.6120300149887087e-05}
{"train_loss": 0.06821010261774063, "global_step": 200184, "epoch": 2249, "lr": 1.6119873792891317e-05}
{"train_loss": 0.05431068688631058, "global_step": 200185, "epoch": 2249, "lr": 1.611944744045033e-05}
{"train_loss": 0.07944799959659576, "global_step": 200186, "epoch": 2249, "lr": 1.6119021092564164e-05}
{"train_loss": 0.052217282354831696, "global_step": 200187, "epoch": 2249, "lr": 1.611859474923289e-05}
{"train_loss": 0.09137053042650223, "global_step": 200188, "epoch": 2249, "lr": 1.6118168410456547e-05}
{"train_loss": 0.03851228952407837, "global_step": 200189, "epoch": 2249, "lr": 1.611774207623522e-05}
{"train_loss": 0.11047790199518204, "global_step": 200190, "epoch": 2249, "lr": 1.6117315746568927e-05}
{"train_loss": 0.09602344036102295, "global_step": 200191, "epoch": 2249, "lr": 1.611688942145777e-05}
{"train_loss": 0.029142038896679878, "global_step": 200192, "epoch": 2249, "lr": 1.611646310090177e-05}
{"train_loss": 0.061426080763339996, "global_step": 200193, "epoch": 2249, "lr": 1.6116036784901012e-05}
{"train_loss": 0.055151015520095825, "global_step": 200194, "epoch": 2249, "lr": 1.6115610473455523e-05}
{"train_loss": 0.12660889327526093, "global_step": 200195, "epoch": 2249, "lr": 1.6115184166565394e-05}
{"train_loss": 0.08545389771461487, "global_step": 200196, "epoch": 2249, "lr": 1.611475786423065e-05}
{"train_loss": 0.055700648576021194, "global_step": 200197, "epoch": 2249, "lr": 1.611433156645138e-05}
{"train_loss": 0.08770390599966049, "global_step": 200198, "epoch": 2249, "lr": 1.6113905273227603e-05}
{"train_loss": 0.05983436852693558, "global_step": 200199, "epoch": 2249, "lr": 1.6113478984559423e-05}
{"train_loss": 0.04351668432354927, "global_step": 200200, "epoch": 2249, "lr": 1.611305270044685e-05}
{"train_loss": 0.06836799532175064, "global_step": 200201, "epoch": 2249, "lr": 1.6112626420889987e-05}
{"train_loss": 0.06961128115653992, "global_step": 200202, "epoch": 2249, "lr": 1.611220014588886e-05}
{"train_loss": 0.04399048164486885, "global_step": 200203, "epoch": 2249, "lr": 1.6111773875443527e-05}
{"train_loss": 0.0653131976723671, "global_step": 200204, "epoch": 2249, "lr": 1.611134760955406e-05}
{"train_loss": 0.044587016105651855, "global_step": 200205, "epoch": 2249, "lr": 1.6110921348220502e-05}
{"train_loss": 0.06914600729942322, "global_step": 200206, "epoch": 2249, "lr": 1.611049509144293e-05}
{"train_loss": 0.08335836976766586, "global_step": 200207, "epoch": 2249, "lr": 1.611006883922137e-05}
{"train_loss": 0.0711747482419014, "global_step": 200208, "epoch": 2249, "lr": 1.610964259155592e-05}
{"train_loss": 0.07027889788150787, "global_step": 200209, "epoch": 2249, "lr": 1.61092163484466e-05}
{"train_loss": 0.08134745806455612, "global_step": 200210, "epoch": 2249, "lr": 1.610879010989349e-05}
{"train_loss": 0.05535674840211868, "global_step": 200211, "epoch": 2249, "lr": 1.610836387589663e-05}
{"train_loss": 0.08361005038022995, "global_step": 200212, "epoch": 2249, "lr": 1.6107937646456108e-05}
{"train_loss": 0.08352378010749817, "global_step": 200213, "epoch": 2249, "lr": 1.6107511421571935e-05}
{"train_loss": 0.055477894842624664, "global_step": 200214, "epoch": 2249, "lr": 1.6107085201244216e-05}
{"train_loss": 0.11842025071382523, "global_step": 200215, "epoch": 2249, "lr": 1.6106658985472965e-05}
{"train_loss": 0.056828759610652924, "global_step": 200216, "epoch": 2249, "lr": 1.6106232774258285e-05}
{"train_loss": 0.09549367427825928, "global_step": 200217, "epoch": 2249, "lr": 1.6105806567600184e-05}
{"train_loss": 0.08076881617307663, "global_step": 200218, "epoch": 2249, "lr": 1.6105380365498768e-05}
{"train_loss": 0.08640879392623901, "global_step": 200219, "epoch": 2249, "lr": 1.610495416795405e-05}
{"train_loss": 0.06979304552078247, "global_step": 200220, "epoch": 2249, "lr": 1.6104527974966112e-05}
{"train_loss": 0.1052190437912941, "global_step": 200221, "epoch": 2249, "lr": 1.6104101786535018e-05}
{"train_loss": 0.09929025173187256, "global_step": 200222, "epoch": 2249, "lr": 1.61036756026608e-05}
{"train_loss": 0.10816410928964615, "global_step": 200223, "epoch": 2249, "lr": 1.6103249423343546e-05}
{"train_loss": 0.07887114584445953, "global_step": 200224, "epoch": 2249, "lr": 1.6102823248583283e-05}
{"train_loss": 0.07590217143297195, "global_step": 200225, "epoch": 2249, "lr": 1.6102397078380094e-05}
{"train_loss": 0.08386427909135818, "global_step": 200226, "epoch": 2249, "lr": 1.610197091273401e-05}
{"train_loss": 0.059233758598566055, "global_step": 200227, "epoch": 2249, "lr": 1.610154475164512e-05}
{"train_loss": 0.02769281528890133, "global_step": 200228, "epoch": 2249, "lr": 1.6101118595113446e-05}
{"train_loss": 0.08913864195346832, "global_step": 200229, "epoch": 2249, "lr": 1.610069244313908e-05}
{"train_loss": 0.09374081343412399, "global_step": 200230, "epoch": 2249, "lr": 1.610026629572205e-05}
{"train_loss": 0.0740891844034195, "global_step": 200231, "epoch": 2249, "lr": 1.6099840152862437e-05}
{"train_loss": 0.041698746383190155, "global_step": 200232, "epoch": 2249, "lr": 1.6099414014560278e-05}
{"train_loss": 0.08249776810407639, "global_step": 200233, "epoch": 2249, "lr": 1.609898788081565e-05}
{"train_loss": 0.06676650047302246, "global_step": 200234, "epoch": 2249, "lr": 1.609856175162859e-05}
{"train_loss": 0.019924936816096306, "global_step": 200235, "epoch": 2249, "lr": 1.609813562699917e-05}
{"train_loss": 0.05604005977511406, "global_step": 200236, "epoch": 2249, "lr": 1.6097709506927438e-05}
{"train_loss": 0.06590536236763, "global_step": 200237, "epoch": 2249, "lr": 1.6097283391413463e-05}
{"train_loss": 0.042768605053424835, "global_step": 200238, "epoch": 2249, "lr": 1.609685728045729e-05}
{"train_loss": 0.061152245849370956, "global_step": 200239, "epoch": 2249, "lr": 1.6096431174058984e-05}
{"train_loss": 0.06246616691350937, "global_step": 200240, "epoch": 2249, "lr": 1.6096005072218595e-05}
{"train_loss": 0.05236704647541046, "global_step": 200241, "epoch": 2249, "lr": 1.609557897493619e-05}
{"train_loss": 0.030454423278570175, "global_step": 200242, "epoch": 2249, "lr": 1.6095152882211835e-05}
{"train_loss": 0.07096390426158905, "global_step": 200243, "epoch": 2249, "lr": 1.6094726794045546e-05}
{"train_loss": 0.05678717419505119, "global_step": 200244, "epoch": 2249, "lr": 1.6094300710437425e-05}
{"train_loss": 0.01745016872882843, "global_step": 200245, "epoch": 2249, "lr": 1.60938746313875e-05}
{"train_loss": 0.035575173795223236, "global_step": 200246, "epoch": 2249, "lr": 1.6093448556895857e-05}
{"train_loss": 0.04421224817633629, "global_step": 200247, "epoch": 2249, "lr": 1.6093022486962518e-05}
{"train_loss": 0.04607062414288521, "global_step": 200248, "epoch": 2249, "lr": 1.6092596421587574e-05}
{"train_loss": 0.0670644019552496, "global_step": 200249, "epoch": 2249, "lr": 1.6092170360771054e-05, "val_loss": 8.16845989227295}
{"train_loss": 0.047015558928251266, "global_step": 200250, "epoch": 2250, "lr": 1.609174430451304e-05}
{"train_loss": 0.05294511094689369, "global_step": 200251, "epoch": 2250, "lr": 1.6091318252813563e-05}
{"train_loss": 0.052345652133226395, "global_step": 200252, "epoch": 2250, "lr": 1.6090892205672715e-05}
{"train_loss": 0.059192925691604614, "global_step": 200253, "epoch": 2250, "lr": 1.609046616309051e-05}
{"train_loss": 0.11028515547513962, "global_step": 200254, "epoch": 2250, "lr": 1.609004012506705e-05}
{"train_loss": 0.07780376821756363, "global_step": 200255, "epoch": 2250, "lr": 1.6089614091602355e-05}
{"train_loss": 0.03660855442285538, "global_step": 200256, "epoch": 2250, "lr": 1.6089188062696497e-05}
{"train_loss": 0.04716267064213753, "global_step": 200257, "epoch": 2250, "lr": 1.6088762038349552e-05}
{"train_loss": 0.07725873589515686, "global_step": 200258, "epoch": 2250, "lr": 1.608833601856154e-05}
{"train_loss": 0.08998530358076096, "global_step": 200259, "epoch": 2250, "lr": 1.6087910003332556e-05}
{"train_loss": 0.03380747139453888, "global_step": 200260, "epoch": 2250, "lr": 1.6087483992662622e-05}
{"train_loss": 0.07348786294460297, "global_step": 200261, "epoch": 2250, "lr": 1.608705798655183e-05}
{"train_loss": 0.04890334606170654, "global_step": 200262, "epoch": 2250, "lr": 1.6086631985000206e-05}
{"train_loss": 0.030121982097625732, "global_step": 200263, "epoch": 2250, "lr": 1.608620598800783e-05}
{"train_loss": 0.07038860768079758, "global_step": 200264, "epoch": 2250, "lr": 1.608577999557474e-05}
{"train_loss": 0.04430106282234192, "global_step": 200265, "epoch": 2250, "lr": 1.608535400770102e-05}
{"train_loss": 0.04673079773783684, "global_step": 200266, "epoch": 2250, "lr": 1.608492802438669e-05}
{"train_loss": 0.06723814457654953, "global_step": 200267, "epoch": 2250, "lr": 1.608450204563185e-05}
{"train_loss": 0.12720297276973724, "global_step": 200268, "epoch": 2250, "lr": 1.6084076071436514e-05}
{"train_loss": 0.0967167466878891, "global_step": 200269, "epoch": 2250, "lr": 1.608365010180078e-05}
{"train_loss": 0.03978434577584267, "global_step": 200270, "epoch": 2250, "lr": 1.6083224136724668e-05}
{"train_loss": 0.05332987383008003, "global_step": 200271, "epoch": 2250, "lr": 1.608279817620827e-05}
{"train_loss": 0.09432799369096756, "global_step": 200272, "epoch": 2250, "lr": 1.6082372220251606e-05}
{"train_loss": 0.06955049932003021, "global_step": 200273, "epoch": 2250, "lr": 1.6081946268854775e-05}
{"train_loss": 0.07108699530363083, "global_step": 200274, "epoch": 2250, "lr": 1.6081520322017796e-05}
{"train_loss": 0.06136723980307579, "global_step": 200275, "epoch": 2250, "lr": 1.608109437974076e-05}
{"train_loss": 0.0897182747721672, "global_step": 200276, "epoch": 2250, "lr": 1.608066844202369e-05}
{"train_loss": 0.054834023118019104, "global_step": 200277, "epoch": 2250, "lr": 1.608024250886668e-05}
{"train_loss": 0.09414752572774887, "global_step": 200278, "epoch": 2250, "lr": 1.6079816580269748e-05}
{"train_loss": 0.06309720128774643, "global_step": 200279, "epoch": 2250, "lr": 1.6079390656232985e-05}
{"train_loss": 0.06521055847406387, "global_step": 200280, "epoch": 2250, "lr": 1.6078964736756435e-05}
{"train_loss": 0.051642827689647675, "global_step": 200281, "epoch": 2250, "lr": 1.6078538821840144e-05}
{"train_loss": 0.02968369983136654, "global_step": 200282, "epoch": 2250, "lr": 1.6078112911484193e-05}
{"train_loss": 0.07798619568347931, "global_step": 200283, "epoch": 2250, "lr": 1.6077687005688614e-05}
{"train_loss": 0.06985332071781158, "global_step": 200284, "epoch": 2250, "lr": 1.607726110445349e-05}
{"train_loss": 0.09088060259819031, "global_step": 200285, "epoch": 2250, "lr": 1.6076835207778846e-05}
{"train_loss": 0.07798665016889572, "global_step": 200286, "epoch": 2250, "lr": 1.6076409315664777e-05}
{"train_loss": 0.08423521369695663, "global_step": 200287, "epoch": 2250, "lr": 1.6075983428111302e-05}
{"train_loss": 0.07499208301305771, "global_step": 200288, "epoch": 2250, "lr": 1.6075557545118514e-05}
{"train_loss": 0.03270784765481949, "global_step": 200289, "epoch": 2250, "lr": 1.6075131666686442e-05}
{"train_loss": 0.07958688586950302, "global_step": 200290, "epoch": 2250, "lr": 1.607470579281517e-05}
{"train_loss": 0.08113870769739151, "global_step": 200291, "epoch": 2250, "lr": 1.607427992350472e-05}
{"train_loss": 0.08033425360918045, "global_step": 200292, "epoch": 2250, "lr": 1.6073854058755172e-05}
{"train_loss": 0.05395463854074478, "global_step": 200293, "epoch": 2250, "lr": 1.6073428198566597e-05}
{"train_loss": 0.043931715190410614, "global_step": 200294, "epoch": 2250, "lr": 1.607300234293902e-05}
{"train_loss": 0.09152382612228394, "global_step": 200295, "epoch": 2250, "lr": 1.607257649187253e-05}
{"train_loss": 0.07291081547737122, "global_step": 200296, "epoch": 2250, "lr": 1.6072150645367157e-05}
{"train_loss": 0.05312124639749527, "global_step": 200297, "epoch": 2250, "lr": 1.607172480342298e-05}
{"train_loss": 0.11017472296953201, "global_step": 200298, "epoch": 2250, "lr": 1.6071298966040034e-05}
{"train_loss": 0.13005395233631134, "global_step": 200299, "epoch": 2250, "lr": 1.60708731332184e-05}
{"train_loss": 0.1327560544013977, "global_step": 200300, "epoch": 2250, "lr": 1.6070447304958115e-05}
{"train_loss": 0.03337980434298515, "global_step": 200301, "epoch": 2250, "lr": 1.6070021481259256e-05}
{"train_loss": 0.04866768419742584, "global_step": 200302, "epoch": 2250, "lr": 1.606959566212185e-05}
{"train_loss": 0.08152402937412262, "global_step": 200303, "epoch": 2250, "lr": 1.6069169847545996e-05}
{"train_loss": 0.07387992739677429, "global_step": 200304, "epoch": 2250, "lr": 1.606874403753171e-05}
{"train_loss": 0.08360334485769272, "global_step": 200305, "epoch": 2250, "lr": 1.6068318232079078e-05}
{"train_loss": 0.032329872250556946, "global_step": 200306, "epoch": 2250, "lr": 1.6067892431188136e-05}
{"train_loss": 0.07369886338710785, "global_step": 200307, "epoch": 2250, "lr": 1.6067466634858968e-05}
{"train_loss": 0.08775223046541214, "global_step": 200308, "epoch": 2250, "lr": 1.60670408430916e-05}
{"train_loss": 0.07920272648334503, "global_step": 200309, "epoch": 2250, "lr": 1.606661505588612e-05}
{"train_loss": 0.048462796956300735, "global_step": 200310, "epoch": 2250, "lr": 1.6066189273242555e-05}
{"train_loss": 0.07683650404214859, "global_step": 200311, "epoch": 2250, "lr": 1.606576349516098e-05}
{"train_loss": 0.08333441615104675, "global_step": 200312, "epoch": 2250, "lr": 1.606533772164146e-05}
{"train_loss": 0.08396531641483307, "global_step": 200313, "epoch": 2250, "lr": 1.606491195268404e-05}
{"train_loss": 0.0668533518910408, "global_step": 200314, "epoch": 2250, "lr": 1.606448618828877e-05}
{"train_loss": 0.0244146715849638, "global_step": 200315, "epoch": 2250, "lr": 1.606406042845573e-05}
{"train_loss": 0.055512867867946625, "global_step": 200316, "epoch": 2250, "lr": 1.6063634673184947e-05}
{"train_loss": 0.09661179780960083, "global_step": 200317, "epoch": 2250, "lr": 1.606320892247651e-05}
{"train_loss": 0.0764002650976181, "global_step": 200318, "epoch": 2250, "lr": 1.606278317633046e-05}
{"train_loss": 0.094328872859478, "global_step": 200319, "epoch": 2250, "lr": 1.606235743474684e-05}
{"train_loss": 0.03750363737344742, "global_step": 200320, "epoch": 2250, "lr": 1.606193169772574e-05}
{"train_loss": 0.06000944599509239, "global_step": 200321, "epoch": 2250, "lr": 1.6061505965267182e-05}
{"train_loss": 0.04997658729553223, "global_step": 200322, "epoch": 2250, "lr": 1.6061080237371256e-05}
{"train_loss": 0.09239153563976288, "global_step": 200323, "epoch": 2250, "lr": 1.6060654514037988e-05}
{"train_loss": 0.06782418489456177, "global_step": 200324, "epoch": 2250, "lr": 1.6060228795267466e-05}
{"train_loss": 0.07774461060762405, "global_step": 200325, "epoch": 2250, "lr": 1.605980308105972e-05}
{"train_loss": 0.050234097987413406, "global_step": 200326, "epoch": 2250, "lr": 1.605937737141484e-05}
{"train_loss": 0.05401746928691864, "global_step": 200327, "epoch": 2250, "lr": 1.605895166633284e-05}
{"train_loss": 0.0448145717382431, "global_step": 200328, "epoch": 2250, "lr": 1.6058525965813804e-05}
{"train_loss": 0.05252522975206375, "global_step": 200329, "epoch": 2250, "lr": 1.60581002698578e-05}
{"train_loss": 0.051417797803878784, "global_step": 200330, "epoch": 2250, "lr": 1.6057674578464854e-05}
{"train_loss": 0.06433593481779099, "global_step": 200331, "epoch": 2250, "lr": 1.6057248891635057e-05}
{"train_loss": 0.11535660922527313, "global_step": 200332, "epoch": 2250, "lr": 1.6056823209368433e-05}
{"train_loss": 0.07871273905038834, "global_step": 200333, "epoch": 2250, "lr": 1.6056397531665067e-05}
{"train_loss": 0.06348911672830582, "global_step": 200334, "epoch": 2250, "lr": 1.6055971858524993e-05}
{"train_loss": 0.05514935404062271, "global_step": 200335, "epoch": 2250, "lr": 1.6055546189948295e-05}
{"train_loss": 0.05797922983765602, "global_step": 200336, "epoch": 2250, "lr": 1.6055120525935e-05}
{"train_loss": 0.053056810051202774, "global_step": 200337, "epoch": 2250, "lr": 1.60546948664852e-05}
{"train_loss": 0.06804987742157464, "global_step": 200338, "epoch": 2250, "lr": 1.605426921159891e-05, "train/sim_max_reward_0": 0.6602939811269873, "train/sim_max_reward_1": 0.6961650720463237, "train/sim_max_reward_2": 0.4295389133802492, "train/sim_max_reward_3": 0.7130626263741814, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9988883947564551, "test/sim_max_reward_4500000": 0.991380991369435, "test/sim_max_reward_4500001": 0.9720469595771786, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.4534642662124108, "test/sim_max_reward_4500004": 0.9687055806057453, "test/sim_max_reward_4500005": 0.9900873408763008, "test/sim_max_reward_4500006": 0.4139282367476546, "test/sim_max_reward_4500007": 0.9568134085911901, "test/sim_max_reward_4500008": 0.5683538790088679, "test/sim_max_reward_4500009": 0.05061701189674781, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.42953450688910455, "test/sim_max_reward_4500012": 0.950507844749075, "test/sim_max_reward_4500013": 0.9705186276167245, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9993085080120223, "test/sim_max_reward_4500016": 0.9934915809125038, "test/sim_max_reward_4500017": 0.5172092340042515, "test/sim_max_reward_4500018": 0.9672059017125105, "test/sim_max_reward_4500019": 0.9852564802357945, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.5320112005505634, "test/sim_max_reward_4500022": 0.9906440790350078, "test/sim_max_reward_4500023": 0.6661302312295069, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.985510844007151, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6743512376930534, "test/sim_max_reward_4500029": 0.8144807527773453, "test/sim_max_reward_4500030": 0.9953951217014269, "test/sim_max_reward_4500031": 0.3980520788490153, "test/sim_max_reward_4500032": 0.1786604968942586, "test/sim_max_reward_4500033": 0.94262902993891, "test/sim_max_reward_4500034": 0.9988467900167585, "test/sim_max_reward_4500035": 0.4413882273583823, "test/sim_max_reward_4500036": 0.9667659128216912, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.16263220512360543, "test/sim_max_reward_4500039": 0.8991469053015679, "test/sim_max_reward_4500040": 0.9952808629460064, "test/sim_max_reward_4500041": 0.617627426590909, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.6754632195639108, "test/sim_max_reward_4500044": 0.6854018156834862, "test/sim_max_reward_4500045": 0.5568727425822271, "test/sim_max_reward_4500046": 0.4211328514316566, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7496581646140328, "test/mean_score": 0.6755370878222792, "val_loss": 7.97605037689209, "train_action_mse_error": 31.7097110748291}
{"train_loss": 0.09303896129131317, "global_step": 200339, "epoch": 2251, "lr": 1.6053843561276228e-05}
{"train_loss": 0.07376761734485626, "global_step": 200340, "epoch": 2251, "lr": 1.6053417915517178e-05}
{"train_loss": 0.04106399416923523, "global_step": 200341, "epoch": 2251, "lr": 1.6052992274321847e-05}
{"train_loss": 0.04684145003557205, "global_step": 200342, "epoch": 2251, "lr": 1.6052566637690265e-05}
{"train_loss": 0.05438055098056793, "global_step": 200343, "epoch": 2251, "lr": 1.605214100562251e-05}
{"train_loss": 0.0808594673871994, "global_step": 200344, "epoch": 2251, "lr": 1.605171537811862e-05}
{"train_loss": 0.04792822524905205, "global_step": 200345, "epoch": 2251, "lr": 1.605128975517868e-05}
{"train_loss": 0.10364536941051483, "global_step": 200346, "epoch": 2251, "lr": 1.6050864136802713e-05}
{"train_loss": 0.08135591447353363, "global_step": 200347, "epoch": 2251, "lr": 1.6050438522990795e-05}
{"train_loss": 0.0756947472691536, "global_step": 200348, "epoch": 2251, "lr": 1.6050012913742995e-05}
{"train_loss": 0.0774066373705864, "global_step": 200349, "epoch": 2251, "lr": 1.6049587309059345e-05}
{"train_loss": 0.04786830022931099, "global_step": 200350, "epoch": 2251, "lr": 1.6049161708939925e-05}
{"train_loss": 0.04534861817955971, "global_step": 200351, "epoch": 2251, "lr": 1.6048736113384765e-05}
{"train_loss": 0.07739194482564926, "global_step": 200352, "epoch": 2251, "lr": 1.604831052239396e-05}
{"train_loss": 0.06594548374414444, "global_step": 200353, "epoch": 2251, "lr": 1.6047884935967543e-05}
{"train_loss": 0.05708230286836624, "global_step": 200354, "epoch": 2251, "lr": 1.604745935410556e-05}
{"train_loss": 0.07531332224607468, "global_step": 200355, "epoch": 2251, "lr": 1.6047033776808096e-05}
{"train_loss": 0.060804177075624466, "global_step": 200356, "epoch": 2251, "lr": 1.6046608204075175e-05}
{"train_loss": 0.06937845796346664, "global_step": 200357, "epoch": 2251, "lr": 1.6046182635906898e-05}
{"train_loss": 0.06630362570285797, "global_step": 200358, "epoch": 2251, "lr": 1.6045757072303292e-05}
{"train_loss": 0.12697933614253998, "global_step": 200359, "epoch": 2251, "lr": 1.6045331513264407e-05}
{"train_loss": 0.03791430592536926, "global_step": 200360, "epoch": 2251, "lr": 1.6044905958790325e-05}
{"train_loss": 0.09560802578926086, "global_step": 200361, "epoch": 2251, "lr": 1.604448040888108e-05}
{"train_loss": 0.07289465516805649, "global_step": 200362, "epoch": 2251, "lr": 1.6044054863536756e-05}
{"train_loss": 0.046712201088666916, "global_step": 200363, "epoch": 2251, "lr": 1.6043629322757374e-05}
{"train_loss": 0.07670485228300095, "global_step": 200364, "epoch": 2251, "lr": 1.604320378654304e-05}
{"train_loss": 0.03796519711613655, "global_step": 200365, "epoch": 2251, "lr": 1.6042778254893754e-05}
{"train_loss": 0.060579195618629456, "global_step": 200366, "epoch": 2251, "lr": 1.6042352727809614e-05}
{"train_loss": 0.0655248612165451, "global_step": 200367, "epoch": 2251, "lr": 1.6041927205290673e-05}
{"train_loss": 0.058859217911958694, "global_step": 200368, "epoch": 2251, "lr": 1.604150168733697e-05}
{"train_loss": 0.055952947586774826, "global_step": 200369, "epoch": 2251, "lr": 1.6041076173948582e-05}
{"train_loss": 0.09151608496904373, "global_step": 200370, "epoch": 2251, "lr": 1.604065066512555e-05}
{"train_loss": 0.08624501526355743, "global_step": 200371, "epoch": 2251, "lr": 1.604022516086795e-05}
{"train_loss": 0.04586785286664963, "global_step": 200372, "epoch": 2251, "lr": 1.6039799661175815e-05}
{"train_loss": 0.10614646226167679, "global_step": 200373, "epoch": 2251, "lr": 1.6039374166049232e-05}
{"train_loss": 0.06768685579299927, "global_step": 200374, "epoch": 2251, "lr": 1.6038948675488217e-05}
{"train_loss": 0.038826052099466324, "global_step": 200375, "epoch": 2251, "lr": 1.6038523189492877e-05}
{"train_loss": 0.030057616531848907, "global_step": 200376, "epoch": 2251, "lr": 1.603809770806322e-05}
{"train_loss": 0.038264453411102295, "global_step": 200377, "epoch": 2251, "lr": 1.6037672231199348e-05}
{"train_loss": 0.04923465475440025, "global_step": 200378, "epoch": 2251, "lr": 1.6037246758901282e-05}
{"train_loss": 0.06327655911445618, "global_step": 200379, "epoch": 2251, "lr": 1.6036821291169106e-05}
{"train_loss": 0.038068074733018875, "global_step": 200380, "epoch": 2251, "lr": 1.603639582800285e-05}
{"train_loss": 0.09585418552160263, "global_step": 200381, "epoch": 2251, "lr": 1.6035970369402603e-05}
{"train_loss": 0.07673180103302002, "global_step": 200382, "epoch": 2251, "lr": 1.603554491536839e-05}
{"train_loss": 0.05714324116706848, "global_step": 200383, "epoch": 2251, "lr": 1.6035119465900285e-05}
{"train_loss": 0.05956126004457474, "global_step": 200384, "epoch": 2251, "lr": 1.6034694020998358e-05}
{"train_loss": 0.05090438202023506, "global_step": 200385, "epoch": 2251, "lr": 1.6034268580662643e-05}
{"train_loss": 0.036189571022987366, "global_step": 200386, "epoch": 2251, "lr": 1.6033843144893212e-05}
{"train_loss": 0.01756615936756134, "global_step": 200387, "epoch": 2251, "lr": 1.6033417713690108e-05}
{"train_loss": 0.0648411214351654, "global_step": 200388, "epoch": 2251, "lr": 1.603299228705341e-05}
{"train_loss": 0.06849616020917892, "global_step": 200389, "epoch": 2251, "lr": 1.603256686498315e-05}
{"train_loss": 0.032264504581689835, "global_step": 200390, "epoch": 2251, "lr": 1.603214144747941e-05}
{"train_loss": 0.046520963311195374, "global_step": 200391, "epoch": 2251, "lr": 1.603171603454224e-05}
{"train_loss": 0.06831071525812149, "global_step": 200392, "epoch": 2251, "lr": 1.603129062617167e-05}
{"train_loss": 0.07809477299451828, "global_step": 200393, "epoch": 2251, "lr": 1.6030865222367796e-05}
{"train_loss": 0.05483979731798172, "global_step": 200394, "epoch": 2251, "lr": 1.6030439823130645e-05}
{"train_loss": 0.013862325809895992, "global_step": 200395, "epoch": 2251, "lr": 1.60300144284603e-05}
{"train_loss": 0.06958707422018051, "global_step": 200396, "epoch": 2251, "lr": 1.6029589038356803e-05}
{"train_loss": 0.032937370240688324, "global_step": 200397, "epoch": 2251, "lr": 1.60291636528202e-05}
{"train_loss": 0.07801171392202377, "global_step": 200398, "epoch": 2251, "lr": 1.602873827185058e-05}
{"train_loss": 0.10076480358839035, "global_step": 200399, "epoch": 2251, "lr": 1.6028312895447968e-05}
{"train_loss": 0.07256285846233368, "global_step": 200400, "epoch": 2251, "lr": 1.6027887523612445e-05}
{"train_loss": 0.0860118493437767, "global_step": 200401, "epoch": 2251, "lr": 1.602746215634405e-05}
{"train_loss": 0.094924196600914, "global_step": 200402, "epoch": 2251, "lr": 1.602703679364284e-05}
{"train_loss": 0.03728719800710678, "global_step": 200403, "epoch": 2251, "lr": 1.6026611435508905e-05}
{"train_loss": 0.050513625144958496, "global_step": 200404, "epoch": 2251, "lr": 1.6026186081942256e-05}
{"train_loss": 0.05643915385007858, "global_step": 200405, "epoch": 2251, "lr": 1.6025760732942992e-05}
{"train_loss": 0.07010570913553238, "global_step": 200406, "epoch": 2251, "lr": 1.6025335388511132e-05}
{"train_loss": 0.05066933110356331, "global_step": 200407, "epoch": 2251, "lr": 1.602491004864677e-05}
{"train_loss": 0.03629675507545471, "global_step": 200408, "epoch": 2251, "lr": 1.6024484713349923e-05}
{"train_loss": 0.06324976682662964, "global_step": 200409, "epoch": 2251, "lr": 1.6024059382620692e-05}
{"train_loss": 0.016403354704380035, "global_step": 200410, "epoch": 2251, "lr": 1.6023634056459093e-05}
{"train_loss": 0.0811825767159462, "global_step": 200411, "epoch": 2251, "lr": 1.602320873486522e-05}
{"train_loss": 0.08377208560705185, "global_step": 200412, "epoch": 2251, "lr": 1.60227834178391e-05}
{"train_loss": 0.06419342756271362, "global_step": 200413, "epoch": 2251, "lr": 1.6022358105380815e-05}
{"train_loss": 0.05366472527384758, "global_step": 200414, "epoch": 2251, "lr": 1.602193279749039e-05}
{"train_loss": 0.04083095118403435, "global_step": 200415, "epoch": 2251, "lr": 1.602150749416792e-05}
{"train_loss": 0.11586865782737732, "global_step": 200416, "epoch": 2251, "lr": 1.6021082195413427e-05}
{"train_loss": 0.058421432971954346, "global_step": 200417, "epoch": 2251, "lr": 1.602065690122701e-05}
{"train_loss": 0.06301411986351013, "global_step": 200418, "epoch": 2251, "lr": 1.602023161160868e-05}
{"train_loss": 0.05304345488548279, "global_step": 200419, "epoch": 2251, "lr": 1.601980632655851e-05}
{"train_loss": 0.055924493819475174, "global_step": 200420, "epoch": 2251, "lr": 1.6019381046076586e-05}
{"train_loss": 0.050138916820287704, "global_step": 200421, "epoch": 2251, "lr": 1.6018955770162926e-05}
{"train_loss": 0.06338375806808472, "global_step": 200422, "epoch": 2251, "lr": 1.6018530498817612e-05}
{"train_loss": 0.03798293322324753, "global_step": 200423, "epoch": 2251, "lr": 1.6018105232040687e-05}
{"train_loss": 0.05755668878555298, "global_step": 200424, "epoch": 2251, "lr": 1.6017679969832223e-05}
{"train_loss": 0.10579391568899155, "global_step": 200425, "epoch": 2251, "lr": 1.6017254712192252e-05}
{"train_loss": 0.049809396266937256, "global_step": 200426, "epoch": 2251, "lr": 1.6016829459120865e-05}
{"train_loss": 0.0635045153099332, "global_step": 200427, "epoch": 2251, "lr": 1.6016404210618085e-05, "val_loss": 8.074766159057617}
{"train_loss": 0.057859767228364944, "global_step": 200428, "epoch": 2252, "lr": 1.6015978966684e-05}
{"train_loss": 0.059832196682691574, "global_step": 200429, "epoch": 2252, "lr": 1.601555372731865e-05}
{"train_loss": 0.04487600922584534, "global_step": 200430, "epoch": 2252, "lr": 1.6015128492522085e-05}
{"train_loss": 0.09561394155025482, "global_step": 200431, "epoch": 2252, "lr": 1.6014703262294383e-05}
{"train_loss": 0.06320526450872421, "global_step": 200432, "epoch": 2252, "lr": 1.601427803663557e-05}
{"train_loss": 0.07867937535047531, "global_step": 200433, "epoch": 2252, "lr": 1.601385281554575e-05}
{"train_loss": 0.027915233746170998, "global_step": 200434, "epoch": 2252, "lr": 1.6013427599024945e-05}
{"train_loss": 0.05898354575037956, "global_step": 200435, "epoch": 2252, "lr": 1.6013002387073206e-05}
{"train_loss": 0.07085934281349182, "global_step": 200436, "epoch": 2252, "lr": 1.6012577179690623e-05}
{"train_loss": 0.05055546015501022, "global_step": 200437, "epoch": 2252, "lr": 1.6012151976877214e-05}
{"train_loss": 0.05520843341946602, "global_step": 200438, "epoch": 2252, "lr": 1.6011726778633064e-05}
{"train_loss": 0.05199512094259262, "global_step": 200439, "epoch": 2252, "lr": 1.6011301584958232e-05}
{"train_loss": 0.054335128515958786, "global_step": 200440, "epoch": 2252, "lr": 1.6010876395852754e-05}
{"train_loss": 0.07873464375734329, "global_step": 200441, "epoch": 2252, "lr": 1.6010451211316717e-05}
{"train_loss": 0.03926605358719826, "global_step": 200442, "epoch": 2252, "lr": 1.6010026031350138e-05}
{"train_loss": 0.07136645913124084, "global_step": 200443, "epoch": 2252, "lr": 1.6009600855953122e-05}
{"train_loss": 0.10674623399972916, "global_step": 200444, "epoch": 2252, "lr": 1.6009175685125677e-05}
{"train_loss": 0.07483740150928497, "global_step": 200445, "epoch": 2252, "lr": 1.6008750518867906e-05}
{"train_loss": 0.04595159366726875, "global_step": 200446, "epoch": 2252, "lr": 1.600832535717982e-05}
{"train_loss": 0.06417305022478104, "global_step": 200447, "epoch": 2252, "lr": 1.6007900200061528e-05}
{"train_loss": 0.08591170608997345, "global_step": 200448, "epoch": 2252, "lr": 1.6007475047513033e-05}
{"train_loss": 0.08991655707359314, "global_step": 200449, "epoch": 2252, "lr": 1.6007049899534436e-05}
{"train_loss": 0.10950443893671036, "global_step": 200450, "epoch": 2252, "lr": 1.6006624756125766e-05}
{"train_loss": 0.06014515832066536, "global_step": 200451, "epoch": 2252, "lr": 1.6006199617287105e-05}
{"train_loss": 0.05212031304836273, "global_step": 200452, "epoch": 2252, "lr": 1.6005774483018482e-05}
{"train_loss": 0.08933723717927933, "global_step": 200453, "epoch": 2252, "lr": 1.600534935331998e-05}
{"train_loss": 0.10571976006031036, "global_step": 200454, "epoch": 2252, "lr": 1.600492422819163e-05}
{"train_loss": 0.021796880289912224, "global_step": 200455, "epoch": 2252, "lr": 1.6004499107633503e-05}
{"train_loss": 0.05853576213121414, "global_step": 200456, "epoch": 2252, "lr": 1.6004073991645674e-05}
{"train_loss": 0.06518086791038513, "global_step": 200457, "epoch": 2252, "lr": 1.6003648880228167e-05}
{"train_loss": 0.0272822342813015, "global_step": 200458, "epoch": 2252, "lr": 1.600322377338107e-05}
{"train_loss": 0.11773397773504257, "global_step": 200459, "epoch": 2252, "lr": 1.6002798671104414e-05}
{"train_loss": 0.04579140245914459, "global_step": 200460, "epoch": 2252, "lr": 1.6002373573398277e-05}
{"train_loss": 0.05894523859024048, "global_step": 200461, "epoch": 2252, "lr": 1.6001948480262692e-05}
{"train_loss": 0.05466390401124954, "global_step": 200462, "epoch": 2252, "lr": 1.6001523391697747e-05}
{"train_loss": 0.03993174806237221, "global_step": 200463, "epoch": 2252, "lr": 1.6001098307703468e-05}
{"train_loss": 0.03846539556980133, "global_step": 200464, "epoch": 2252, "lr": 1.600067322827994e-05}
{"train_loss": 0.06863376498222351, "global_step": 200465, "epoch": 2252, "lr": 1.6000248153427195e-05}
{"train_loss": 0.08038803189992905, "global_step": 200466, "epoch": 2252, "lr": 1.5999823083145315e-05}
{"train_loss": 0.051617443561553955, "global_step": 200467, "epoch": 2252, "lr": 1.599939801743433e-05}
{"train_loss": 0.029508540406823158, "global_step": 200468, "epoch": 2252, "lr": 1.5998972956294326e-05}
{"train_loss": 0.04164443910121918, "global_step": 200469, "epoch": 2252, "lr": 1.5998547899725347e-05}
{"train_loss": 0.045214347541332245, "global_step": 200470, "epoch": 2252, "lr": 1.599812284772744e-05}
{"train_loss": 0.07787450402975082, "global_step": 200471, "epoch": 2252, "lr": 1.5997697800300675e-05}
{"train_loss": 0.020144982263445854, "global_step": 200472, "epoch": 2252, "lr": 1.599727275744509e-05}
{"train_loss": 0.08112659305334091, "global_step": 200473, "epoch": 2252, "lr": 1.599684771916078e-05}
{"train_loss": 0.04477623105049133, "global_step": 200474, "epoch": 2252, "lr": 1.599642268544776e-05}
{"train_loss": 0.1394200325012207, "global_step": 200475, "epoch": 2252, "lr": 1.599599765630612e-05}
{"train_loss": 0.05234227702021599, "global_step": 200476, "epoch": 2252, "lr": 1.5995572631735895e-05}
{"train_loss": 0.020821336656808853, "global_step": 200477, "epoch": 2252, "lr": 1.599514761173716e-05}
{"train_loss": 0.07965616881847382, "global_step": 200478, "epoch": 2252, "lr": 1.5994722596309953e-05}
{"train_loss": 0.024054627865552902, "global_step": 200479, "epoch": 2252, "lr": 1.5994297585454356e-05}
{"train_loss": 0.04115470126271248, "global_step": 200480, "epoch": 2252, "lr": 1.599387257917039e-05}
{"train_loss": 0.0929749384522438, "global_step": 200481, "epoch": 2252, "lr": 1.599344757745815e-05}
{"train_loss": 0.03559951111674309, "global_step": 200482, "epoch": 2252, "lr": 1.5993022580317662e-05}
{"train_loss": 0.07709427177906036, "global_step": 200483, "epoch": 2252, "lr": 1.5992597587749015e-05}
{"train_loss": 0.016819853335618973, "global_step": 200484, "epoch": 2252, "lr": 1.599217259975223e-05}
{"train_loss": 0.03783427178859711, "global_step": 200485, "epoch": 2252, "lr": 1.59917476163274e-05}
{"train_loss": 0.06652375310659409, "global_step": 200486, "epoch": 2252, "lr": 1.5991322637474554e-05}
{"train_loss": 0.040985170751810074, "global_step": 200487, "epoch": 2252, "lr": 1.599089766319377e-05}
{"train_loss": 0.024481581524014473, "global_step": 200488, "epoch": 2252, "lr": 1.599047269348508e-05}
{"train_loss": 0.058597687631845474, "global_step": 200489, "epoch": 2252, "lr": 1.5990047728348577e-05}
{"train_loss": 0.07173556089401245, "global_step": 200490, "epoch": 2252, "lr": 1.598962276778428e-05}
{"train_loss": 0.09125453978776932, "global_step": 200491, "epoch": 2252, "lr": 1.5989197811792278e-05}
{"train_loss": 0.06613753736019135, "global_step": 200492, "epoch": 2252, "lr": 1.5988772860372597e-05}
{"train_loss": 0.038208670914173126, "global_step": 200493, "epoch": 2252, "lr": 1.5988347913525307e-05}
{"train_loss": 0.05684295669198036, "global_step": 200494, "epoch": 2252, "lr": 1.598792297125049e-05}
{"train_loss": 0.06902309507131577, "global_step": 200495, "epoch": 2252, "lr": 1.598749803354817e-05}
{"train_loss": 0.037174291908741, "global_step": 200496, "epoch": 2252, "lr": 1.5987073100418427e-05}
{"train_loss": 0.09317157417535782, "global_step": 200497, "epoch": 2252, "lr": 1.598664817186129e-05}
{"train_loss": 0.14995966851711273, "global_step": 200498, "epoch": 2252, "lr": 1.598622324787686e-05}
{"train_loss": 0.08357042819261551, "global_step": 200499, "epoch": 2252, "lr": 1.598579832846514e-05}
{"train_loss": 0.09408248960971832, "global_step": 200500, "epoch": 2252, "lr": 1.5985373413626236e-05}
{"train_loss": 0.03617807850241661, "global_step": 200501, "epoch": 2252, "lr": 1.5984948503360166e-05}
{"train_loss": 0.07358837127685547, "global_step": 200502, "epoch": 2252, "lr": 1.5984523597667022e-05}
{"train_loss": 0.049798253923654556, "global_step": 200503, "epoch": 2252, "lr": 1.5984098696546827e-05}
{"train_loss": 0.07563991099596024, "global_step": 200504, "epoch": 2252, "lr": 1.598367379999967e-05}
{"train_loss": 0.07056434452533722, "global_step": 200505, "epoch": 2252, "lr": 1.5983248908025582e-05}
{"train_loss": 0.10317520052194595, "global_step": 200506, "epoch": 2252, "lr": 1.5982824020624647e-05}
{"train_loss": 0.09687577188014984, "global_step": 200507, "epoch": 2252, "lr": 1.5982399137796904e-05}
{"train_loss": 0.09281891584396362, "global_step": 200508, "epoch": 2252, "lr": 1.5981974259542397e-05}
{"train_loss": 0.04440675675868988, "global_step": 200509, "epoch": 2252, "lr": 1.5981549385861217e-05}
{"train_loss": 0.03965603932738304, "global_step": 200510, "epoch": 2252, "lr": 1.598112451675339e-05}
{"train_loss": 0.12696048617362976, "global_step": 200511, "epoch": 2252, "lr": 1.5980699652218993e-05}
{"train_loss": 0.039998702704906464, "global_step": 200512, "epoch": 2252, "lr": 1.598027479225807e-05}
{"train_loss": 0.05553547665476799, "global_step": 200513, "epoch": 2252, "lr": 1.5979849936870695e-05}
{"train_loss": 0.06433700770139694, "global_step": 200514, "epoch": 2252, "lr": 1.5979425086056897e-05}
{"train_loss": 0.08022739738225937, "global_step": 200515, "epoch": 2252, "lr": 1.5979000239816773e-05}
{"train_loss": 0.06338576355090972, "global_step": 200516, "epoch": 2252, "lr": 1.5978575398150335e-05, "val_loss": 8.040677070617676}
{"train_loss": 0.14406222105026245, "global_step": 200517, "epoch": 2253, "lr": 1.5978150561057686e-05}
{"train_loss": 0.08134244382381439, "global_step": 200518, "epoch": 2253, "lr": 1.5977725728538838e-05}
{"train_loss": 0.04231180623173714, "global_step": 200519, "epoch": 2253, "lr": 1.597730090059389e-05}
{"train_loss": 0.09815993160009384, "global_step": 200520, "epoch": 2253, "lr": 1.597687607722286e-05}
{"train_loss": 0.09292589873075485, "global_step": 200521, "epoch": 2253, "lr": 1.597645125842584e-05}
{"train_loss": 0.0652315616607666, "global_step": 200522, "epoch": 2253, "lr": 1.5976026444202857e-05}
{"train_loss": 0.036010585725307465, "global_step": 200523, "epoch": 2253, "lr": 1.5975601634554e-05}
{"train_loss": 0.018799690529704094, "global_step": 200524, "epoch": 2253, "lr": 1.5975176829479287e-05}
{"train_loss": 0.047533903270959854, "global_step": 200525, "epoch": 2253, "lr": 1.5974752028978817e-05}
{"train_loss": 0.049168653786182404, "global_step": 200526, "epoch": 2253, "lr": 1.597432723305261e-05}
{"train_loss": 0.061502765864133835, "global_step": 200527, "epoch": 2253, "lr": 1.5973902441700762e-05}
{"train_loss": 0.06636488437652588, "global_step": 200528, "epoch": 2253, "lr": 1.5973477654923285e-05}
{"train_loss": 0.029363617300987244, "global_step": 200529, "epoch": 2253, "lr": 1.597305287272026e-05}
{"train_loss": 0.052034035325050354, "global_step": 200530, "epoch": 2253, "lr": 1.5972628095091764e-05}
{"train_loss": 0.04856540635228157, "global_step": 200531, "epoch": 2253, "lr": 1.597220332203781e-05}
{"train_loss": 0.11858447641134262, "global_step": 200532, "epoch": 2253, "lr": 1.5971778553558496e-05}
{"train_loss": 0.13370434939861298, "global_step": 200533, "epoch": 2253, "lr": 1.5971353789653854e-05}
{"train_loss": 0.04306276515126228, "global_step": 200534, "epoch": 2253, "lr": 1.5970929030323955e-05}
{"train_loss": 0.042649850249290466, "global_step": 200535, "epoch": 2253, "lr": 1.5970504275568837e-05}
{"train_loss": 0.04325512796640396, "global_step": 200536, "epoch": 2253, "lr": 1.5970079525388588e-05}
{"train_loss": 0.09484630823135376, "global_step": 200537, "epoch": 2253, "lr": 1.596965477978323e-05}
{"train_loss": 0.09144667536020279, "global_step": 200538, "epoch": 2253, "lr": 1.5969230038752853e-05}
{"train_loss": 0.0657678171992302, "global_step": 200539, "epoch": 2253, "lr": 1.596880530229748e-05}
{"train_loss": 0.10558640956878662, "global_step": 200540, "epoch": 2253, "lr": 1.5968380570417208e-05}
{"train_loss": 0.07127651572227478, "global_step": 200541, "epoch": 2253, "lr": 1.596795584311205e-05}
{"train_loss": 0.07809586822986603, "global_step": 200542, "epoch": 2253, "lr": 1.5967531120382105e-05}
{"train_loss": 0.044946104288101196, "global_step": 200543, "epoch": 2253, "lr": 1.5967106402227393e-05}
{"train_loss": 0.13003034889698029, "global_step": 200544, "epoch": 2253, "lr": 1.5966681688648007e-05}
{"train_loss": 0.03321094438433647, "global_step": 200545, "epoch": 2253, "lr": 1.5966256979643985e-05}
{"train_loss": 0.04154990240931511, "global_step": 200546, "epoch": 2253, "lr": 1.5965832275215363e-05}
{"train_loss": 0.10653337091207504, "global_step": 200547, "epoch": 2253, "lr": 1.5965407575362244e-05}
{"train_loss": 0.07427613437175751, "global_step": 200548, "epoch": 2253, "lr": 1.5964982880084645e-05}
{"train_loss": 0.06072483956813812, "global_step": 200549, "epoch": 2253, "lr": 1.596455818938265e-05}
{"train_loss": 0.07328605651855469, "global_step": 200550, "epoch": 2253, "lr": 1.5964133503256296e-05}
{"train_loss": 0.07362502068281174, "global_step": 200551, "epoch": 2253, "lr": 1.596370882170566e-05}
{"train_loss": 0.061310235410928726, "global_step": 200552, "epoch": 2253, "lr": 1.596328414473077e-05}
{"train_loss": 0.053493596613407135, "global_step": 200553, "epoch": 2253, "lr": 1.5962859472331725e-05}
{"train_loss": 0.08396318554878235, "global_step": 200554, "epoch": 2253, "lr": 1.5962434804508537e-05}
{"train_loss": 0.09376044571399689, "global_step": 200555, "epoch": 2253, "lr": 1.5962010141261302e-05}
{"train_loss": 0.05161253735423088, "global_step": 200556, "epoch": 2253, "lr": 1.5961585482590048e-05}
{"train_loss": 0.06535925716161728, "global_step": 200557, "epoch": 2253, "lr": 1.5961160828494858e-05}
{"train_loss": 0.03727196902036667, "global_step": 200558, "epoch": 2253, "lr": 1.5960736178975754e-05}
{"train_loss": 0.04316534847021103, "global_step": 200559, "epoch": 2253, "lr": 1.5960311534032834e-05}
{"train_loss": 0.013332542032003403, "global_step": 200560, "epoch": 2253, "lr": 1.595988689366612e-05}
{"train_loss": 0.06978678703308105, "global_step": 200561, "epoch": 2253, "lr": 1.5959462257875703e-05}
{"train_loss": 0.11672355234622955, "global_step": 200562, "epoch": 2253, "lr": 1.59590376266616e-05}
{"train_loss": 0.04351316764950752, "global_step": 200563, "epoch": 2253, "lr": 1.595861300002391e-05}
{"train_loss": 0.08236125856637955, "global_step": 200564, "epoch": 2253, "lr": 1.5958188377962652e-05}
{"train_loss": 0.15090543031692505, "global_step": 200565, "epoch": 2253, "lr": 1.5957763760477897e-05}
{"train_loss": 0.070968396961689, "global_step": 200566, "epoch": 2253, "lr": 1.595733914756972e-05}
{"train_loss": 0.08212139457464218, "global_step": 200567, "epoch": 2253, "lr": 1.5956914539238153e-05}
{"train_loss": 0.05133695900440216, "global_step": 200568, "epoch": 2253, "lr": 1.595648993548328e-05}
{"train_loss": 0.07731278985738754, "global_step": 200569, "epoch": 2253, "lr": 1.5956065336305127e-05}
{"train_loss": 0.05700899288058281, "global_step": 200570, "epoch": 2253, "lr": 1.595564074170378e-05}
{"train_loss": 0.037156932055950165, "global_step": 200571, "epoch": 2253, "lr": 1.595521615167927e-05}
{"train_loss": 0.0679248720407486, "global_step": 200572, "epoch": 2253, "lr": 1.5954791566231676e-05}
{"train_loss": 0.05459502711892128, "global_step": 200573, "epoch": 2253, "lr": 1.595436698536103e-05}
{"train_loss": 0.03473545238375664, "global_step": 200574, "epoch": 2253, "lr": 1.5953942409067425e-05}
{"train_loss": 0.048262059688568115, "global_step": 200575, "epoch": 2253, "lr": 1.595351783735088e-05}
{"train_loss": 0.054596349596977234, "global_step": 200576, "epoch": 2253, "lr": 1.5953093270211485e-05}
{"train_loss": 0.0655713826417923, "global_step": 200577, "epoch": 2253, "lr": 1.5952668707649264e-05}
{"train_loss": 0.08940708637237549, "global_step": 200578, "epoch": 2253, "lr": 1.5952244149664308e-05}
{"train_loss": 0.04557682201266289, "global_step": 200579, "epoch": 2253, "lr": 1.5951819596256647e-05}
{"train_loss": 0.07901902496814728, "global_step": 200580, "epoch": 2253, "lr": 1.5951395047426355e-05}
{"train_loss": 0.041077762842178345, "global_step": 200581, "epoch": 2253, "lr": 1.5950970503173473e-05}
{"train_loss": 0.034323591738939285, "global_step": 200582, "epoch": 2253, "lr": 1.5950545963498086e-05}
{"train_loss": 0.0902973860502243, "global_step": 200583, "epoch": 2253, "lr": 1.595012142840021e-05}
{"train_loss": 0.06511718779802322, "global_step": 200584, "epoch": 2253, "lr": 1.5949696897879952e-05}
{"train_loss": 0.0483233705163002, "global_step": 200585, "epoch": 2253, "lr": 1.5949272371937334e-05}
{"train_loss": 0.05302358791232109, "global_step": 200586, "epoch": 2253, "lr": 1.5948847850572407e-05}
{"train_loss": 0.08443666994571686, "global_step": 200587, "epoch": 2253, "lr": 1.5948423333785257e-05}
{"train_loss": 0.07252267003059387, "global_step": 200588, "epoch": 2253, "lr": 1.594799882157591e-05}
{"train_loss": 0.06463680416345596, "global_step": 200589, "epoch": 2253, "lr": 1.594757431394446e-05}
{"train_loss": 0.10427188128232956, "global_step": 200590, "epoch": 2253, "lr": 1.5947149810890928e-05}
{"train_loss": 0.0819820985198021, "global_step": 200591, "epoch": 2253, "lr": 1.5946725312415396e-05}
{"train_loss": 0.07832007110118866, "global_step": 200592, "epoch": 2253, "lr": 1.59463008185179e-05}
{"train_loss": 0.07065461575984955, "global_step": 200593, "epoch": 2253, "lr": 1.594587632919853e-05}
{"train_loss": 0.07101371884346008, "global_step": 200594, "epoch": 2253, "lr": 1.5945451844457294e-05}
{"train_loss": 0.042036641389131546, "global_step": 200595, "epoch": 2253, "lr": 1.5945027364294303e-05}
{"train_loss": 0.07703648507595062, "global_step": 200596, "epoch": 2253, "lr": 1.5944602888709566e-05}
{"train_loss": 0.08959630131721497, "global_step": 200597, "epoch": 2253, "lr": 1.594417841770318e-05}
{"train_loss": 0.03571216017007828, "global_step": 200598, "epoch": 2253, "lr": 1.594375395127517e-05}
{"train_loss": 0.05181436613202095, "global_step": 200599, "epoch": 2253, "lr": 1.5943329489425618e-05}
{"train_loss": 0.05828246846795082, "global_step": 200600, "epoch": 2253, "lr": 1.594290503215456e-05}
{"train_loss": 0.05611776188015938, "global_step": 200601, "epoch": 2253, "lr": 1.594248057946206e-05}
{"train_loss": 0.08186816424131393, "global_step": 200602, "epoch": 2253, "lr": 1.5942056131348197e-05}
{"train_loss": 0.06708617508411407, "global_step": 200603, "epoch": 2253, "lr": 1.594163168781299e-05}
{"train_loss": 0.10169245302677155, "global_step": 200604, "epoch": 2253, "lr": 1.5941207248856537e-05}
{"train_loss": 0.0672705883599734, "global_step": 200605, "epoch": 2253, "lr": 1.594078281447885e-05, "val_loss": 8.149455070495605}
{"train_loss": 0.06238291412591934, "global_step": 200606, "epoch": 2254, "lr": 1.5940358384680032e-05}
{"train_loss": 0.11103630065917969, "global_step": 200607, "epoch": 2254, "lr": 1.59399339594601e-05}
{"train_loss": 0.08491208404302597, "global_step": 200608, "epoch": 2254, "lr": 1.5939509538819142e-05}
{"train_loss": 0.04899788275361061, "global_step": 200609, "epoch": 2254, "lr": 1.5939085122757192e-05}
{"train_loss": 0.07676488161087036, "global_step": 200610, "epoch": 2254, "lr": 1.593866071127433e-05}
{"train_loss": 0.052251994609832764, "global_step": 200611, "epoch": 2254, "lr": 1.593823630437058e-05}
{"train_loss": 0.05839008092880249, "global_step": 200612, "epoch": 2254, "lr": 1.5937811902046035e-05}
{"train_loss": 0.0333121158182621, "global_step": 200613, "epoch": 2254, "lr": 1.5937387504300722e-05}
{"train_loss": 0.05828635394573212, "global_step": 200614, "epoch": 2254, "lr": 1.593696311113473e-05}
{"train_loss": 0.08867737650871277, "global_step": 200615, "epoch": 2254, "lr": 1.5936538722548082e-05}
{"train_loss": 0.07936413586139679, "global_step": 200616, "epoch": 2254, "lr": 1.5936114338540864e-05}
{"train_loss": 0.03793599084019661, "global_step": 200617, "epoch": 2254, "lr": 1.593568995911311e-05}
{"train_loss": 0.05964736267924309, "global_step": 200618, "epoch": 2254, "lr": 1.5935265584264896e-05}
{"train_loss": 0.07829101383686066, "global_step": 200619, "epoch": 2254, "lr": 1.593484121399626e-05}
{"train_loss": 0.02347290888428688, "global_step": 200620, "epoch": 2254, "lr": 1.5934416848307283e-05}
{"train_loss": 0.05904185026884079, "global_step": 200621, "epoch": 2254, "lr": 1.593399248719799e-05}
{"train_loss": 0.06240593641996384, "global_step": 200622, "epoch": 2254, "lr": 1.593356813066848e-05}
{"train_loss": 0.059720270335674286, "global_step": 200623, "epoch": 2254, "lr": 1.5933143778718777e-05}
{"train_loss": 0.082523874938488, "global_step": 200624, "epoch": 2254, "lr": 1.5932719431348943e-05}
{"train_loss": 0.04091648757457733, "global_step": 200625, "epoch": 2254, "lr": 1.5932295088559046e-05}
{"train_loss": 0.07577493786811829, "global_step": 200626, "epoch": 2254, "lr": 1.5931870750349127e-05}
{"train_loss": 0.044873978942632675, "global_step": 200627, "epoch": 2254, "lr": 1.5931446416719265e-05}
{"train_loss": 0.06284491717815399, "global_step": 200628, "epoch": 2254, "lr": 1.593102208766949e-05}
{"train_loss": 0.03661798685789108, "global_step": 200629, "epoch": 2254, "lr": 1.5930597763199885e-05}
{"train_loss": 0.030364317819476128, "global_step": 200630, "epoch": 2254, "lr": 1.5930173443310487e-05}
{"train_loss": 0.036754075437784195, "global_step": 200631, "epoch": 2254, "lr": 1.5929749128001374e-05}
{"train_loss": 0.0694449171423912, "global_step": 200632, "epoch": 2254, "lr": 1.5929324817272572e-05}
{"train_loss": 0.09701364487409592, "global_step": 200633, "epoch": 2254, "lr": 1.592890051112418e-05}
{"train_loss": 0.05803016200661659, "global_step": 200634, "epoch": 2254, "lr": 1.5928476209556215e-05}
{"train_loss": 0.04230089858174324, "global_step": 200635, "epoch": 2254, "lr": 1.5928051912568764e-05}
{"train_loss": 0.12676508724689484, "global_step": 200636, "epoch": 2254, "lr": 1.5927627620161857e-05}
{"train_loss": 0.0413946732878685, "global_step": 200637, "epoch": 2254, "lr": 1.5927203332335566e-05}
{"train_loss": 0.12298841774463654, "global_step": 200638, "epoch": 2254, "lr": 1.5926779049089958e-05}
{"train_loss": 0.10222045332193375, "global_step": 200639, "epoch": 2254, "lr": 1.5926354770425066e-05}
{"train_loss": 0.03343659266829491, "global_step": 200640, "epoch": 2254, "lr": 1.592593049634098e-05}
{"train_loss": 0.07961641252040863, "global_step": 200641, "epoch": 2254, "lr": 1.5925506226837715e-05}
{"train_loss": 0.07726061344146729, "global_step": 200642, "epoch": 2254, "lr": 1.592508196191537e-05}
{"train_loss": 0.05265873670578003, "global_step": 200643, "epoch": 2254, "lr": 1.592465770157397e-05}
{"train_loss": 0.04516386240720749, "global_step": 200644, "epoch": 2254, "lr": 1.5924233445813595e-05}
{"train_loss": 0.049512557685375214, "global_step": 200645, "epoch": 2254, "lr": 1.5923809194634283e-05}
{"train_loss": 0.04510931298136711, "global_step": 200646, "epoch": 2254, "lr": 1.5923384948036107e-05}
{"train_loss": 0.07921247184276581, "global_step": 200647, "epoch": 2254, "lr": 1.5922960706019103e-05}
{"train_loss": 0.07200034707784653, "global_step": 200648, "epoch": 2254, "lr": 1.592253646858336e-05}
{"train_loss": 0.07242366671562195, "global_step": 200649, "epoch": 2254, "lr": 1.5922112235728897e-05}
{"train_loss": 0.045269906520843506, "global_step": 200650, "epoch": 2254, "lr": 1.592168800745581e-05}
{"train_loss": 0.04755176603794098, "global_step": 200651, "epoch": 2254, "lr": 1.5921263783764118e-05}
{"train_loss": 0.03539225459098816, "global_step": 200652, "epoch": 2254, "lr": 1.592083956465391e-05}
{"train_loss": 0.07522394508123398, "global_step": 200653, "epoch": 2254, "lr": 1.5920415350125218e-05}
{"train_loss": 0.059454843401908875, "global_step": 200654, "epoch": 2254, "lr": 1.591999114017813e-05}
{"train_loss": 0.0878598690032959, "global_step": 200655, "epoch": 2254, "lr": 1.5919566934812662e-05}
{"train_loss": 0.07384464144706726, "global_step": 200656, "epoch": 2254, "lr": 1.5919142734028913e-05}
{"train_loss": 0.09058362990617752, "global_step": 200657, "epoch": 2254, "lr": 1.59187185378269e-05}
{"train_loss": 0.05121433734893799, "global_step": 200658, "epoch": 2254, "lr": 1.591829434620672e-05}
{"train_loss": 0.06075994670391083, "global_step": 200659, "epoch": 2254, "lr": 1.591787015916839e-05}
{"train_loss": 0.0730263963341713, "global_step": 200660, "epoch": 2254, "lr": 1.5917445976712e-05}
{"train_loss": 0.06575697660446167, "global_step": 200661, "epoch": 2254, "lr": 1.59170217988376e-05}
{"train_loss": 0.0665673092007637, "global_step": 200662, "epoch": 2254, "lr": 1.5916597625545227e-05}
{"train_loss": 0.06897083669900894, "global_step": 200663, "epoch": 2254, "lr": 1.5916173456834967e-05}
{"train_loss": 0.04897291585803032, "global_step": 200664, "epoch": 2254, "lr": 1.591574929270684e-05}
{"train_loss": 0.04304218664765358, "global_step": 200665, "epoch": 2254, "lr": 1.5915325133160947e-05}
{"train_loss": 0.07337519526481628, "global_step": 200666, "epoch": 2254, "lr": 1.5914900978197306e-05}
{"train_loss": 0.10667160153388977, "global_step": 200667, "epoch": 2254, "lr": 1.5914476827816005e-05}
{"train_loss": 0.04140094295144081, "global_step": 200668, "epoch": 2254, "lr": 1.591405268201707e-05}
{"train_loss": 0.06629833579063416, "global_step": 200669, "epoch": 2254, "lr": 1.5913628540800595e-05}
{"train_loss": 0.08139742165803909, "global_step": 200670, "epoch": 2254, "lr": 1.59132044041666e-05}
{"train_loss": 0.03539181873202324, "global_step": 200671, "epoch": 2254, "lr": 1.5912780272115175e-05}
{"train_loss": 0.0636691004037857, "global_step": 200672, "epoch": 2254, "lr": 1.5912356144646346e-05}
{"train_loss": 0.05949989706277847, "global_step": 200673, "epoch": 2254, "lr": 1.5911932021760178e-05}
{"train_loss": 0.07347031682729721, "global_step": 200674, "epoch": 2254, "lr": 1.5911507903456764e-05}
{"train_loss": 0.011367267929017544, "global_step": 200675, "epoch": 2254, "lr": 1.5911083789736103e-05}
{"train_loss": 0.02300490438938141, "global_step": 200676, "epoch": 2254, "lr": 1.5910659680598306e-05}
{"train_loss": 0.10564994066953659, "global_step": 200677, "epoch": 2254, "lr": 1.5910235576043385e-05}
{"train_loss": 0.07765688002109528, "global_step": 200678, "epoch": 2254, "lr": 1.590981147607144e-05}
{"train_loss": 0.057985227555036545, "global_step": 200679, "epoch": 2254, "lr": 1.5909387380682483e-05}
{"train_loss": 0.10795444995164871, "global_step": 200680, "epoch": 2254, "lr": 1.590896328987661e-05}
{"train_loss": 0.08105708658695221, "global_step": 200681, "epoch": 2254, "lr": 1.5908539203653848e-05}
{"train_loss": 0.07417381554841995, "global_step": 200682, "epoch": 2254, "lr": 1.590811512201428e-05}
{"train_loss": 0.10738764703273773, "global_step": 200683, "epoch": 2254, "lr": 1.590769104495794e-05}
{"train_loss": 0.04304356127977371, "global_step": 200684, "epoch": 2254, "lr": 1.5907266972484906e-05}
{"train_loss": 0.060331765562295914, "global_step": 200685, "epoch": 2254, "lr": 1.590684290459521e-05}
{"train_loss": 0.11944020539522171, "global_step": 200686, "epoch": 2254, "lr": 1.590641884128894e-05}
{"train_loss": 0.04529456049203873, "global_step": 200687, "epoch": 2254, "lr": 1.590599478256612e-05}
{"train_loss": 0.07753079384565353, "global_step": 200688, "epoch": 2254, "lr": 1.5905570728426844e-05}
{"train_loss": 0.05409242957830429, "global_step": 200689, "epoch": 2254, "lr": 1.5905146678871124e-05}
{"train_loss": 0.05771045759320259, "global_step": 200690, "epoch": 2254, "lr": 1.5904722633899065e-05}
{"train_loss": 0.06548105925321579, "global_step": 200691, "epoch": 2254, "lr": 1.590429859351068e-05}
{"train_loss": 0.05788280442357063, "global_step": 200692, "epoch": 2254, "lr": 1.5903874557706043e-05}
{"train_loss": 0.05594975873827934, "global_step": 200693, "epoch": 2254, "lr": 1.5903450526485242e-05}
{"train_loss": 0.06446748755328106, "global_step": 200694, "epoch": 2254, "lr": 1.590302649984828e-05, "val_loss": 8.171486854553223}
{"train_loss": 0.02659205161035061, "global_step": 200695, "epoch": 2255, "lr": 1.5902602477795258e-05}
{"train_loss": 0.0465947762131691, "global_step": 200696, "epoch": 2255, "lr": 1.5902178460326223e-05}
{"train_loss": 0.08338968455791473, "global_step": 200697, "epoch": 2255, "lr": 1.5901754447441202e-05}
{"train_loss": 0.03736436367034912, "global_step": 200698, "epoch": 2255, "lr": 1.590133043914029e-05}
{"train_loss": 0.027593933045864105, "global_step": 200699, "epoch": 2255, "lr": 1.5900906435423512e-05}
{"train_loss": 0.03422164171934128, "global_step": 200700, "epoch": 2255, "lr": 1.5900482436290966e-05}
{"train_loss": 0.06279754638671875, "global_step": 200701, "epoch": 2255, "lr": 1.5900058441742677e-05}
{"train_loss": 0.022412925958633423, "global_step": 200702, "epoch": 2255, "lr": 1.5899634451778696e-05}
{"train_loss": 0.046551525592803955, "global_step": 200703, "epoch": 2255, "lr": 1.5899210466399105e-05}
{"train_loss": 0.0527057945728302, "global_step": 200704, "epoch": 2255, "lr": 1.5898786485603934e-05}
{"train_loss": 0.08613782376050949, "global_step": 200705, "epoch": 2255, "lr": 1.5898362509393278e-05}
{"train_loss": 0.03907764330506325, "global_step": 200706, "epoch": 2255, "lr": 1.5897938537767147e-05}
{"train_loss": 0.10848386585712433, "global_step": 200707, "epoch": 2255, "lr": 1.5897514570725646e-05}
{"train_loss": 0.07279568165540695, "global_step": 200708, "epoch": 2255, "lr": 1.589709060826879e-05}
{"train_loss": 0.04007476195693016, "global_step": 200709, "epoch": 2255, "lr": 1.589666665039665e-05}
{"train_loss": 0.0472785048186779, "global_step": 200710, "epoch": 2255, "lr": 1.5896242697109315e-05}
{"train_loss": 0.07542882114648819, "global_step": 200711, "epoch": 2255, "lr": 1.589581874840679e-05}
{"train_loss": 0.11711230129003525, "global_step": 200712, "epoch": 2255, "lr": 1.589539480428917e-05}
{"train_loss": 0.07143808156251907, "global_step": 200713, "epoch": 2255, "lr": 1.5894970864756486e-05}
{"train_loss": 0.08456848561763763, "global_step": 200714, "epoch": 2255, "lr": 1.589454692980883e-05}
{"train_loss": 0.1679607480764389, "global_step": 200715, "epoch": 2255, "lr": 1.589412299944621e-05}
{"train_loss": 0.0819622278213501, "global_step": 200716, "epoch": 2255, "lr": 1.5893699073668733e-05}
{"train_loss": 0.07195116579532623, "global_step": 200717, "epoch": 2255, "lr": 1.589327515247641e-05}
{"train_loss": 0.043938253074884415, "global_step": 200718, "epoch": 2255, "lr": 1.589285123586934e-05}
{"train_loss": 0.043548863381147385, "global_step": 200719, "epoch": 2255, "lr": 1.589242732384754e-05}
{"train_loss": 0.03172469884157181, "global_step": 200720, "epoch": 2255, "lr": 1.5892003416411112e-05}
{"train_loss": 0.09077095985412598, "global_step": 200721, "epoch": 2255, "lr": 1.589157951356007e-05}
{"train_loss": 0.0759660080075264, "global_step": 200722, "epoch": 2255, "lr": 1.5891155615294502e-05}
{"train_loss": 0.11455302685499191, "global_step": 200723, "epoch": 2255, "lr": 1.5890731721614437e-05}
{"train_loss": 0.027230489999055862, "global_step": 200724, "epoch": 2255, "lr": 1.5890307832519964e-05}
{"train_loss": 0.024585241451859474, "global_step": 200725, "epoch": 2255, "lr": 1.5889883948011108e-05}
{"train_loss": 0.0798797532916069, "global_step": 200726, "epoch": 2255, "lr": 1.588946006808796e-05}
{"train_loss": 0.09122729301452637, "global_step": 200727, "epoch": 2255, "lr": 1.5889036192750535e-05}
{"train_loss": 0.056179244071245193, "global_step": 200728, "epoch": 2255, "lr": 1.588861232199892e-05}
{"train_loss": 0.10388253629207611, "global_step": 200729, "epoch": 2255, "lr": 1.5888188455833175e-05}
{"train_loss": 0.14202766120433807, "global_step": 200730, "epoch": 2255, "lr": 1.588776459425334e-05}
{"train_loss": 0.05318266153335571, "global_step": 200731, "epoch": 2255, "lr": 1.588734073725949e-05}
{"train_loss": 0.14788831770420074, "global_step": 200732, "epoch": 2255, "lr": 1.5886916884851656e-05}
{"train_loss": 0.04266274720430374, "global_step": 200733, "epoch": 2255, "lr": 1.5886493037029924e-05}
{"train_loss": 0.039732854813337326, "global_step": 200734, "epoch": 2255, "lr": 1.588606919379434e-05}
{"train_loss": 0.024646015837788582, "global_step": 200735, "epoch": 2255, "lr": 1.588564535514494e-05}
{"train_loss": 0.06572097539901733, "global_step": 200736, "epoch": 2255, "lr": 1.5885221521081817e-05}
{"train_loss": 0.030580777674913406, "global_step": 200737, "epoch": 2255, "lr": 1.5884797691604993e-05}
{"train_loss": 0.05350276082754135, "global_step": 200738, "epoch": 2255, "lr": 1.5884373866714557e-05}
{"train_loss": 0.0728524774312973, "global_step": 200739, "epoch": 2255, "lr": 1.5883950046410557e-05}
{"train_loss": 0.045950260013341904, "global_step": 200740, "epoch": 2255, "lr": 1.5883526230693024e-05}
{"train_loss": 0.052500225603580475, "global_step": 200741, "epoch": 2255, "lr": 1.5883102419562052e-05}
{"train_loss": 0.03625325858592987, "global_step": 200742, "epoch": 2255, "lr": 1.5882678613017666e-05}
{"train_loss": 0.06564328819513321, "global_step": 200743, "epoch": 2255, "lr": 1.588225481105995e-05}
{"train_loss": 0.06071984022855759, "global_step": 200744, "epoch": 2255, "lr": 1.5881831013688936e-05}
{"train_loss": 0.10079261660575867, "global_step": 200745, "epoch": 2255, "lr": 1.5881407220904708e-05}
{"train_loss": 0.05505219101905823, "global_step": 200746, "epoch": 2255, "lr": 1.5880983432707292e-05}
{"train_loss": 0.08345671743154526, "global_step": 200747, "epoch": 2255, "lr": 1.5880559649096766e-05}
{"train_loss": 0.05225855112075806, "global_step": 200748, "epoch": 2255, "lr": 1.58801358700732e-05}
{"train_loss": 0.0269022136926651, "global_step": 200749, "epoch": 2255, "lr": 1.5879712095636612e-05}
{"train_loss": 0.09105474501848221, "global_step": 200750, "epoch": 2255, "lr": 1.58792883257871e-05}
{"train_loss": 0.06682294607162476, "global_step": 200751, "epoch": 2255, "lr": 1.5878864560524686e-05}
{"train_loss": 0.02627371810376644, "global_step": 200752, "epoch": 2255, "lr": 1.5878440799849458e-05}
{"train_loss": 0.020545827224850655, "global_step": 200753, "epoch": 2255, "lr": 1.5878017043761438e-05}
{"train_loss": 0.08281830698251724, "global_step": 200754, "epoch": 2255, "lr": 1.587759329226073e-05}
{"train_loss": 0.05327007547020912, "global_step": 200755, "epoch": 2255, "lr": 1.5877169545347336e-05}
{"train_loss": 0.0812905952334404, "global_step": 200756, "epoch": 2255, "lr": 1.5876745803021366e-05}
{"train_loss": 0.10264096409082413, "global_step": 200757, "epoch": 2255, "lr": 1.5876322065282828e-05}
{"train_loss": 0.07690676301717758, "global_step": 200758, "epoch": 2255, "lr": 1.5875898332131823e-05}
{"train_loss": 0.06177747994661331, "global_step": 200759, "epoch": 2255, "lr": 1.587547460356837e-05}
{"train_loss": 0.05781767889857292, "global_step": 200760, "epoch": 2255, "lr": 1.5875050879592558e-05}
{"train_loss": 0.04133056476712227, "global_step": 200761, "epoch": 2255, "lr": 1.5874627160204413e-05}
{"train_loss": 0.01910860650241375, "global_step": 200762, "epoch": 2255, "lr": 1.5874203445404028e-05}
{"train_loss": 0.053501397371292114, "global_step": 200763, "epoch": 2255, "lr": 1.5873779735191425e-05}
{"train_loss": 0.1113206297159195, "global_step": 200764, "epoch": 2255, "lr": 1.587335602956667e-05}
{"train_loss": 0.04778917878866196, "global_step": 200765, "epoch": 2255, "lr": 1.587293232852985e-05}
{"train_loss": 0.06446556001901627, "global_step": 200766, "epoch": 2255, "lr": 1.5872508632080972e-05}
{"train_loss": 0.10493960976600647, "global_step": 200767, "epoch": 2255, "lr": 1.5872084940220144e-05}
{"train_loss": 0.058026522397994995, "global_step": 200768, "epoch": 2255, "lr": 1.5871661252947385e-05}
{"train_loss": 0.15868154168128967, "global_step": 200769, "epoch": 2255, "lr": 1.5871237570262776e-05}
{"train_loss": 0.09740705043077469, "global_step": 200770, "epoch": 2255, "lr": 1.587081389216635e-05}
{"train_loss": 0.06796465069055557, "global_step": 200771, "epoch": 2255, "lr": 1.5870390218658187e-05}
{"train_loss": 0.03866685926914215, "global_step": 200772, "epoch": 2255, "lr": 1.5869966549738334e-05}
{"train_loss": 0.04701509699225426, "global_step": 200773, "epoch": 2255, "lr": 1.586954288540684e-05}
{"train_loss": 0.049716174602508545, "global_step": 200774, "epoch": 2255, "lr": 1.5869119225663782e-05}
{"train_loss": 0.05785146728157997, "global_step": 200775, "epoch": 2255, "lr": 1.5868695570509185e-05}
{"train_loss": 0.049732375890016556, "global_step": 200776, "epoch": 2255, "lr": 1.5868271919943144e-05}
{"train_loss": 0.05891452729701996, "global_step": 200777, "epoch": 2255, "lr": 1.58678482739657e-05}
{"train_loss": 0.035359613597393036, "global_step": 200778, "epoch": 2255, "lr": 1.5867424632576892e-05}
{"train_loss": 0.06751668453216553, "global_step": 200779, "epoch": 2255, "lr": 1.586700099577681e-05}
{"train_loss": 0.029127536341547966, "global_step": 200780, "epoch": 2255, "lr": 1.5866577363565476e-05}
{"train_loss": 0.0863611027598381, "global_step": 200781, "epoch": 2255, "lr": 1.586615373594298e-05}
{"train_loss": 0.03907286748290062, "global_step": 200782, "epoch": 2255, "lr": 1.5865730112909348e-05}
{"train_loss": 0.064241477717342, "global_step": 200783, "epoch": 2255, "lr": 1.586530649446465e-05, "val_loss": 8.179905891418457, "train_action_mse_error": 5.880232334136963}
{"train_loss": 0.06955595314502716, "global_step": 200784, "epoch": 2256, "lr": 1.5864882880608966e-05}
{"train_loss": 0.05659845471382141, "global_step": 200785, "epoch": 2256, "lr": 1.5864459271342314e-05}
{"train_loss": 0.025516582652926445, "global_step": 200786, "epoch": 2256, "lr": 1.5864035666664784e-05}
{"train_loss": 0.0617610365152359, "global_step": 200787, "epoch": 2256, "lr": 1.5863612066576405e-05}
{"train_loss": 0.07590925693511963, "global_step": 200788, "epoch": 2256, "lr": 1.5863188471077263e-05}
{"train_loss": 0.05459880083799362, "global_step": 200789, "epoch": 2256, "lr": 1.586276488016738e-05}
{"train_loss": 0.03787335753440857, "global_step": 200790, "epoch": 2256, "lr": 1.5862341293846854e-05}
{"train_loss": 0.0839405432343483, "global_step": 200791, "epoch": 2256, "lr": 1.5861917712115698e-05}
{"train_loss": 0.07597839832305908, "global_step": 200792, "epoch": 2256, "lr": 1.5861494134974013e-05}
{"train_loss": 0.05332440882921219, "global_step": 200793, "epoch": 2256, "lr": 1.5861070562421814e-05}
{"train_loss": 0.07269018888473511, "global_step": 200794, "epoch": 2256, "lr": 1.5860646994459195e-05}
{"train_loss": 0.048948075622320175, "global_step": 200795, "epoch": 2256, "lr": 1.586022343108618e-05}
{"train_loss": 0.044838424772024155, "global_step": 200796, "epoch": 2256, "lr": 1.585979987230286e-05}
{"train_loss": 0.04125947505235672, "global_step": 200797, "epoch": 2256, "lr": 1.5859376318109255e-05}
{"train_loss": 0.03876528888940811, "global_step": 200798, "epoch": 2256, "lr": 1.585895276850546e-05}
{"train_loss": 0.03821362182497978, "global_step": 200799, "epoch": 2256, "lr": 1.5858529223491493e-05}
{"train_loss": 0.047861285507678986, "global_step": 200800, "epoch": 2256, "lr": 1.5858105683067427e-05}
{"train_loss": 0.10562267899513245, "global_step": 200801, "epoch": 2256, "lr": 1.5857682147233344e-05}
{"train_loss": 0.041783735156059265, "global_step": 200802, "epoch": 2256, "lr": 1.585725861598926e-05}
{"train_loss": 0.09167176485061646, "global_step": 200803, "epoch": 2256, "lr": 1.5856835089335277e-05}
{"train_loss": 0.03675767779350281, "global_step": 200804, "epoch": 2256, "lr": 1.5856411567271402e-05}
{"train_loss": 0.039607785642147064, "global_step": 200805, "epoch": 2256, "lr": 1.5855988049797734e-05}
{"train_loss": 0.035574331879615784, "global_step": 200806, "epoch": 2256, "lr": 1.5855564536914293e-05}
{"train_loss": 0.07967652380466461, "global_step": 200807, "epoch": 2256, "lr": 1.5855141028621174e-05}
{"train_loss": 0.1266600340604782, "global_step": 200808, "epoch": 2256, "lr": 1.58547175249184e-05}
{"train_loss": 0.0569583959877491, "global_step": 200809, "epoch": 2256, "lr": 1.585429402580606e-05}
{"train_loss": 0.08478712290525436, "global_step": 200810, "epoch": 2256, "lr": 1.5853870531284176e-05}
{"train_loss": 0.03575831279158592, "global_step": 200811, "epoch": 2256, "lr": 1.5853447041352838e-05}
{"train_loss": 0.06898807734251022, "global_step": 200812, "epoch": 2256, "lr": 1.5853023556012086e-05}
{"train_loss": 0.0732465609908104, "global_step": 200813, "epoch": 2256, "lr": 1.585260007526197e-05}
{"train_loss": 0.1100030243396759, "global_step": 200814, "epoch": 2256, "lr": 1.5852176599102563e-05}
{"train_loss": 0.07841482013463974, "global_step": 200815, "epoch": 2256, "lr": 1.585175312753391e-05}
{"train_loss": 0.11463287472724915, "global_step": 200816, "epoch": 2256, "lr": 1.585132966055608e-05}
{"train_loss": 0.069963738322258, "global_step": 200817, "epoch": 2256, "lr": 1.5850906198169125e-05}
{"train_loss": 0.07967834174633026, "global_step": 200818, "epoch": 2256, "lr": 1.585048274037308e-05}
{"train_loss": 0.03476773947477341, "global_step": 200819, "epoch": 2256, "lr": 1.5850059287168022e-05}
{"train_loss": 0.03174435347318649, "global_step": 200820, "epoch": 2256, "lr": 1.584963583855403e-05}
{"train_loss": 0.033775538206100464, "global_step": 200821, "epoch": 2256, "lr": 1.5849212394531115e-05}
{"train_loss": 0.07593876868486404, "global_step": 200822, "epoch": 2256, "lr": 1.5848788955099376e-05}
{"train_loss": 0.015296394936740398, "global_step": 200823, "epoch": 2256, "lr": 1.5848365520258834e-05}
{"train_loss": 0.037059102207422256, "global_step": 200824, "epoch": 2256, "lr": 1.5847942090009583e-05}
{"train_loss": 0.04717407748103142, "global_step": 200825, "epoch": 2256, "lr": 1.584751866435164e-05}
{"train_loss": 0.10852932929992676, "global_step": 200826, "epoch": 2256, "lr": 1.5847095243285097e-05}
{"train_loss": 0.09375244379043579, "global_step": 200827, "epoch": 2256, "lr": 1.5846671826809984e-05}
{"train_loss": 0.0742521807551384, "global_step": 200828, "epoch": 2256, "lr": 1.5846248414926386e-05}
{"train_loss": 0.04347550496459007, "global_step": 200829, "epoch": 2256, "lr": 1.5845825007634323e-05}
{"train_loss": 0.041473388671875, "global_step": 200830, "epoch": 2256, "lr": 1.584540160493389e-05}
{"train_loss": 0.03019578382372856, "global_step": 200831, "epoch": 2256, "lr": 1.584497820682511e-05}
{"train_loss": 0.039386648684740067, "global_step": 200832, "epoch": 2256, "lr": 1.584455481330807e-05}
{"train_loss": 0.05958158150315285, "global_step": 200833, "epoch": 2256, "lr": 1.5844131424382803e-05}
{"train_loss": 0.11176997423171997, "global_step": 200834, "epoch": 2256, "lr": 1.5843708040049392e-05}
{"train_loss": 0.028945738449692726, "global_step": 200835, "epoch": 2256, "lr": 1.5843284660307862e-05}
{"train_loss": 0.0442027673125267, "global_step": 200836, "epoch": 2256, "lr": 1.5842861285158277e-05}
{"train_loss": 0.08503195643424988, "global_step": 200837, "epoch": 2256, "lr": 1.584243791460073e-05}
{"train_loss": 0.08184503763914108, "global_step": 200838, "epoch": 2256, "lr": 1.584201454863523e-05}
{"train_loss": 0.049268435686826706, "global_step": 200839, "epoch": 2256, "lr": 1.5841591187261874e-05}
{"train_loss": 0.060729339718818665, "global_step": 200840, "epoch": 2256, "lr": 1.584116783048068e-05}
{"train_loss": 0.041170384734869, "global_step": 200841, "epoch": 2256, "lr": 1.5840744478291742e-05}
{"train_loss": 0.07042938470840454, "global_step": 200842, "epoch": 2256, "lr": 1.5840321130695085e-05}
{"train_loss": 0.058464519679546356, "global_step": 200843, "epoch": 2256, "lr": 1.5839897787690794e-05}
{"train_loss": 0.06627687811851501, "global_step": 200844, "epoch": 2256, "lr": 1.5839474449278895e-05}
{"train_loss": 0.05867457017302513, "global_step": 200845, "epoch": 2256, "lr": 1.583905111545948e-05}
{"train_loss": 0.05211319774389267, "global_step": 200846, "epoch": 2256, "lr": 1.5838627786232575e-05}
{"train_loss": 0.0567987859249115, "global_step": 200847, "epoch": 2256, "lr": 1.5838204461598266e-05}
{"train_loss": 0.05720016360282898, "global_step": 200848, "epoch": 2256, "lr": 1.5837781141556574e-05}
{"train_loss": 0.05910244211554527, "global_step": 200849, "epoch": 2256, "lr": 1.5837357826107596e-05}
{"train_loss": 0.0408015139400959, "global_step": 200850, "epoch": 2256, "lr": 1.5836934515251365e-05}
{"train_loss": 0.01233717892318964, "global_step": 200851, "epoch": 2256, "lr": 1.583651120898793e-05}
{"train_loss": 0.0464155413210392, "global_step": 200852, "epoch": 2256, "lr": 1.583608790731737e-05}
{"train_loss": 0.052016306668519974, "global_step": 200853, "epoch": 2256, "lr": 1.583566461023972e-05}
{"train_loss": 0.07242831587791443, "global_step": 200854, "epoch": 2256, "lr": 1.583524131775507e-05}
{"train_loss": 0.06101726368069649, "global_step": 200855, "epoch": 2256, "lr": 1.583481802986343e-05}
{"train_loss": 0.040553249418735504, "global_step": 200856, "epoch": 2256, "lr": 1.58343947465649e-05}
{"train_loss": 0.06366634368896484, "global_step": 200857, "epoch": 2256, "lr": 1.5833971467859504e-05}
{"train_loss": 0.06433825939893723, "global_step": 200858, "epoch": 2256, "lr": 1.5833548193747333e-05}
{"train_loss": 0.06465478986501694, "global_step": 200859, "epoch": 2256, "lr": 1.5833124924228404e-05}
{"train_loss": 0.08357253670692444, "global_step": 200860, "epoch": 2256, "lr": 1.5832701659302814e-05}
{"train_loss": 0.052171122282743454, "global_step": 200861, "epoch": 2256, "lr": 1.5832278398970584e-05}
{"train_loss": 0.07347320765256882, "global_step": 200862, "epoch": 2256, "lr": 1.5831855143231805e-05}
{"train_loss": 0.04546045884490013, "global_step": 200863, "epoch": 2256, "lr": 1.5831431892086495e-05}
{"train_loss": 0.09879797697067261, "global_step": 200864, "epoch": 2256, "lr": 1.5831008645534756e-05}
{"train_loss": 0.06024746969342232, "global_step": 200865, "epoch": 2256, "lr": 1.5830585403576603e-05}
{"train_loss": 0.07302099466323853, "global_step": 200866, "epoch": 2256, "lr": 1.5830162166212126e-05}
{"train_loss": 0.05656810104846954, "global_step": 200867, "epoch": 2256, "lr": 1.5829738933441352e-05}
{"train_loss": 0.06238945573568344, "global_step": 200868, "epoch": 2256, "lr": 1.582931570526437e-05}
{"train_loss": 0.03928584232926369, "global_step": 200869, "epoch": 2256, "lr": 1.58288924816812e-05}
{"train_loss": 0.04867883771657944, "global_step": 200870, "epoch": 2256, "lr": 1.5828469262691942e-05}
{"train_loss": 0.05474607273936272, "global_step": 200871, "epoch": 2256, "lr": 1.5828046048296606e-05}
{"train_loss": 0.060126929428805126, "global_step": 200872, "epoch": 2256, "lr": 1.58276228384953e-05, "val_loss": 8.064746856689453}
{"train_loss": 0.047609396278858185, "global_step": 200873, "epoch": 2257, "lr": 1.5827199633288026e-05}
{"train_loss": 0.03629086911678314, "global_step": 200874, "epoch": 2257, "lr": 1.582677643267487e-05}
{"train_loss": 0.03918467462062836, "global_step": 200875, "epoch": 2257, "lr": 1.582635323665591e-05}
{"train_loss": 0.05594978854060173, "global_step": 200876, "epoch": 2257, "lr": 1.582593004523116e-05}
{"train_loss": 0.05118293687701225, "global_step": 200877, "epoch": 2257, "lr": 1.582550685840071e-05}
{"train_loss": 0.06323657184839249, "global_step": 200878, "epoch": 2257, "lr": 1.5825083676164598e-05}
{"train_loss": 0.03377583250403404, "global_step": 200879, "epoch": 2257, "lr": 1.5824660498522893e-05}
{"train_loss": 0.08708488941192627, "global_step": 200880, "epoch": 2257, "lr": 1.5824237325475633e-05}
{"train_loss": 0.08103278279304504, "global_step": 200881, "epoch": 2257, "lr": 1.5823814157022904e-05}
{"train_loss": 0.05257813632488251, "global_step": 200882, "epoch": 2257, "lr": 1.5823390993164732e-05}
{"train_loss": 0.04896615818142891, "global_step": 200883, "epoch": 2257, "lr": 1.5822967833901203e-05}
{"train_loss": 0.044724926352500916, "global_step": 200884, "epoch": 2257, "lr": 1.582254467923235e-05}
{"train_loss": 0.049143459647893906, "global_step": 200885, "epoch": 2257, "lr": 1.5822121529158246e-05}
{"train_loss": 0.045683737844228745, "global_step": 200886, "epoch": 2257, "lr": 1.5821698383678928e-05}
{"train_loss": 0.07670454680919647, "global_step": 200887, "epoch": 2257, "lr": 1.582127524279448e-05}
{"train_loss": 0.03224994242191315, "global_step": 200888, "epoch": 2257, "lr": 1.5820852106504952e-05}
{"train_loss": 0.0800102949142456, "global_step": 200889, "epoch": 2257, "lr": 1.582042897481037e-05}
{"train_loss": 0.08134863525629044, "global_step": 200890, "epoch": 2257, "lr": 1.5820005847710832e-05}
{"train_loss": 0.08629500865936279, "global_step": 200891, "epoch": 2257, "lr": 1.5819582725206368e-05}
{"train_loss": 0.05407334119081497, "global_step": 200892, "epoch": 2257, "lr": 1.5819159607297057e-05}
{"train_loss": 0.02992943860590458, "global_step": 200893, "epoch": 2257, "lr": 1.5818736493982923e-05}
{"train_loss": 0.05684685334563255, "global_step": 200894, "epoch": 2257, "lr": 1.5818313385264065e-05}
{"train_loss": 0.08184462785720825, "global_step": 200895, "epoch": 2257, "lr": 1.5817890281140497e-05}
{"train_loss": 0.07225566357374191, "global_step": 200896, "epoch": 2257, "lr": 1.5817467181612316e-05}
{"train_loss": 0.0695757120847702, "global_step": 200897, "epoch": 2257, "lr": 1.5817044086679543e-05}
{"train_loss": 0.05375232174992561, "global_step": 200898, "epoch": 2257, "lr": 1.5816620996342276e-05}
{"train_loss": 0.07797999680042267, "global_step": 200899, "epoch": 2257, "lr": 1.581619791060052e-05}
{"train_loss": 0.05393761396408081, "global_step": 200900, "epoch": 2257, "lr": 1.581577482945438e-05}
{"train_loss": 0.07034020870923996, "global_step": 200901, "epoch": 2257, "lr": 1.5815351752903877e-05}
{"train_loss": 0.07252661883831024, "global_step": 200902, "epoch": 2257, "lr": 1.58149286809491e-05}
{"train_loss": 0.08046712726354599, "global_step": 200903, "epoch": 2257, "lr": 1.581450561359007e-05}
{"train_loss": 0.051923226565122604, "global_step": 200904, "epoch": 2257, "lr": 1.581408255082688e-05}
{"train_loss": 0.06063438206911087, "global_step": 200905, "epoch": 2257, "lr": 1.5813659492659555e-05}
{"train_loss": 0.038395971059799194, "global_step": 200906, "epoch": 2257, "lr": 1.581323643908818e-05}
{"train_loss": 0.041586700826883316, "global_step": 200907, "epoch": 2257, "lr": 1.5812813390112792e-05}
{"train_loss": 0.08935314416885376, "global_step": 200908, "epoch": 2257, "lr": 1.581239034573346e-05}
{"train_loss": 0.06130446866154671, "global_step": 200909, "epoch": 2257, "lr": 1.5811967305950222e-05}
{"train_loss": 0.055165715515613556, "global_step": 200910, "epoch": 2257, "lr": 1.581154427076315e-05}
{"train_loss": 0.12174492329359055, "global_step": 200911, "epoch": 2257, "lr": 1.5811121240172315e-05}
{"train_loss": 0.07714906334877014, "global_step": 200912, "epoch": 2257, "lr": 1.5810698214177745e-05}
{"train_loss": 0.054049696773290634, "global_step": 200913, "epoch": 2257, "lr": 1.5810275192779523e-05}
{"train_loss": 0.06924137473106384, "global_step": 200914, "epoch": 2257, "lr": 1.5809852175977676e-05}
{"train_loss": 0.08446208387613297, "global_step": 200915, "epoch": 2257, "lr": 1.5809429163772293e-05}
{"train_loss": 0.07705317437648773, "global_step": 200916, "epoch": 2257, "lr": 1.5809006156163396e-05}
{"train_loss": 0.04418577998876572, "global_step": 200917, "epoch": 2257, "lr": 1.5808583153151085e-05}
{"train_loss": 0.06713147461414337, "global_step": 200918, "epoch": 2257, "lr": 1.5808160154735375e-05}
{"train_loss": 0.03873902186751366, "global_step": 200919, "epoch": 2257, "lr": 1.5807737160916357e-05}
{"train_loss": 0.06539571285247803, "global_step": 200920, "epoch": 2257, "lr": 1.5807314171694055e-05}
{"train_loss": 0.029110293835401535, "global_step": 200921, "epoch": 2257, "lr": 1.5806891187068556e-05}
{"train_loss": 0.065427765250206, "global_step": 200922, "epoch": 2257, "lr": 1.580646820703989e-05}
{"train_loss": 0.04994586855173111, "global_step": 200923, "epoch": 2257, "lr": 1.5806045231608146e-05}
{"train_loss": 0.08135305345058441, "global_step": 200924, "epoch": 2257, "lr": 1.5805622260773343e-05}
{"train_loss": 0.06881055235862732, "global_step": 200925, "epoch": 2257, "lr": 1.5805199294535578e-05}
{"train_loss": 0.10120394825935364, "global_step": 200926, "epoch": 2257, "lr": 1.580477633289487e-05}
{"train_loss": 0.12845782935619354, "global_step": 200927, "epoch": 2257, "lr": 1.580435337585131e-05}
{"train_loss": 0.060226764529943466, "global_step": 200928, "epoch": 2257, "lr": 1.5803930423404935e-05}
{"train_loss": 0.030294811353087425, "global_step": 200929, "epoch": 2257, "lr": 1.580350747555579e-05}
{"train_loss": 0.08998937904834747, "global_step": 200930, "epoch": 2257, "lr": 1.580308453230396e-05}
{"train_loss": 0.056791361421346664, "global_step": 200931, "epoch": 2257, "lr": 1.5802661593649475e-05}
{"train_loss": 0.03003837913274765, "global_step": 200932, "epoch": 2257, "lr": 1.5802238659592423e-05}
{"train_loss": 0.06667307764291763, "global_step": 200933, "epoch": 2257, "lr": 1.5801815730132825e-05}
{"train_loss": 0.029175065457820892, "global_step": 200934, "epoch": 2257, "lr": 1.580139280527077e-05}
{"train_loss": 0.03948912024497986, "global_step": 200935, "epoch": 2257, "lr": 1.5800969885006283e-05}
{"train_loss": 0.05515126511454582, "global_step": 200936, "epoch": 2257, "lr": 1.5800546969339465e-05}
{"train_loss": 0.06680413335561752, "global_step": 200937, "epoch": 2257, "lr": 1.580012405827032e-05}
{"train_loss": 0.04520019516348839, "global_step": 200938, "epoch": 2257, "lr": 1.579970115179895e-05}
{"train_loss": 0.04807508364319801, "global_step": 200939, "epoch": 2257, "lr": 1.579927824992538e-05}
{"train_loss": 0.10416006296873093, "global_step": 200940, "epoch": 2257, "lr": 1.5798855352649698e-05}
{"train_loss": 0.11648285388946533, "global_step": 200941, "epoch": 2257, "lr": 1.5798432459971923e-05}
{"train_loss": 0.046727363020181656, "global_step": 200942, "epoch": 2257, "lr": 1.5798009571892148e-05}
{"train_loss": 0.08487648516893387, "global_step": 200943, "epoch": 2257, "lr": 1.57975866884104e-05}
{"train_loss": 0.08226210623979568, "global_step": 200944, "epoch": 2257, "lr": 1.5797163809526766e-05}
{"train_loss": 0.048281677067279816, "global_step": 200945, "epoch": 2257, "lr": 1.5796740935241268e-05}
{"train_loss": 0.04762255400419235, "global_step": 200946, "epoch": 2257, "lr": 1.5796318065553983e-05}
{"train_loss": 0.07500443607568741, "global_step": 200947, "epoch": 2257, "lr": 1.5795895200464982e-05}
{"train_loss": 0.05694018676877022, "global_step": 200948, "epoch": 2257, "lr": 1.5795472339974286e-05}
{"train_loss": 0.09211073815822601, "global_step": 200949, "epoch": 2257, "lr": 1.579504948408199e-05}
{"train_loss": 0.07628713548183441, "global_step": 200950, "epoch": 2257, "lr": 1.5794626632788124e-05}
{"train_loss": 0.056288812309503555, "global_step": 200951, "epoch": 2257, "lr": 1.579420378609276e-05}
{"train_loss": 0.05869784951210022, "global_step": 200952, "epoch": 2257, "lr": 1.579378094399594e-05}
{"train_loss": 0.06510347127914429, "global_step": 200953, "epoch": 2257, "lr": 1.5793358106497743e-05}
{"train_loss": 0.0386798195540905, "global_step": 200954, "epoch": 2257, "lr": 1.5792935273598193e-05}
{"train_loss": 0.04239097237586975, "global_step": 200955, "epoch": 2257, "lr": 1.579251244529738e-05}
{"train_loss": 0.03519448637962341, "global_step": 200956, "epoch": 2257, "lr": 1.5792089621595336e-05}
{"train_loss": 0.05728141590952873, "global_step": 200957, "epoch": 2257, "lr": 1.579166680249214e-05}
{"train_loss": 0.06571044772863388, "global_step": 200958, "epoch": 2257, "lr": 1.5791243987987825e-05}
{"train_loss": 0.05283017456531525, "global_step": 200959, "epoch": 2257, "lr": 1.5790821178082476e-05}
{"train_loss": 0.12775972485542297, "global_step": 200960, "epoch": 2257, "lr": 1.5790398372776115e-05}
{"train_loss": 0.06266645383968782, "global_step": 200961, "epoch": 2257, "lr": 1.578997557206884e-05, "val_loss": 8.189048767089844}
{"train_loss": 0.04783658683300018, "global_step": 200962, "epoch": 2258, "lr": 1.5789552775960663e-05}
{"train_loss": 0.0363251268863678, "global_step": 200963, "epoch": 2258, "lr": 1.5789129984451683e-05}
{"train_loss": 0.10538089275360107, "global_step": 200964, "epoch": 2258, "lr": 1.5788707197541923e-05}
{"train_loss": 0.048617225140333176, "global_step": 200965, "epoch": 2258, "lr": 1.578828441523147e-05}
{"train_loss": 0.06710966676473618, "global_step": 200966, "epoch": 2258, "lr": 1.5787861637520363e-05}
{"train_loss": 0.07869338989257812, "global_step": 200967, "epoch": 2258, "lr": 1.578743886440865e-05}
{"train_loss": 0.06957433372735977, "global_step": 200968, "epoch": 2258, "lr": 1.5787016095896405e-05}
{"train_loss": 0.05042439326643944, "global_step": 200969, "epoch": 2258, "lr": 1.5786593331983674e-05}
{"train_loss": 0.018772903829813004, "global_step": 200970, "epoch": 2258, "lr": 1.578617057267053e-05}
{"train_loss": 0.06411895155906677, "global_step": 200971, "epoch": 2258, "lr": 1.5785747817957e-05}
{"train_loss": 0.08936192840337753, "global_step": 200972, "epoch": 2258, "lr": 1.5785325067843176e-05}
{"train_loss": 0.07744243741035461, "global_step": 200973, "epoch": 2258, "lr": 1.578490232232908e-05}
{"train_loss": 0.07257462292909622, "global_step": 200974, "epoch": 2258, "lr": 1.5784479581414806e-05}
{"train_loss": 0.0820101946592331, "global_step": 200975, "epoch": 2258, "lr": 1.5784056845100377e-05}
{"train_loss": 0.036099862307310104, "global_step": 200976, "epoch": 2258, "lr": 1.5783634113385876e-05}
{"train_loss": 0.05613373592495918, "global_step": 200977, "epoch": 2258, "lr": 1.5783211386271336e-05}
{"train_loss": 0.07285530120134354, "global_step": 200978, "epoch": 2258, "lr": 1.578278866375684e-05}
{"train_loss": 0.031278930604457855, "global_step": 200979, "epoch": 2258, "lr": 1.578236594584242e-05}
{"train_loss": 0.09548448771238327, "global_step": 200980, "epoch": 2258, "lr": 1.5781943232528152e-05}
{"train_loss": 0.05958830937743187, "global_step": 200981, "epoch": 2258, "lr": 1.5781520523814074e-05}
{"train_loss": 0.05358805134892464, "global_step": 200982, "epoch": 2258, "lr": 1.5781097819700253e-05}
{"train_loss": 0.04593363776803017, "global_step": 200983, "epoch": 2258, "lr": 1.578067512018676e-05}
{"train_loss": 0.03158906474709511, "global_step": 200984, "epoch": 2258, "lr": 1.5780252425273624e-05}
{"train_loss": 0.09952975809574127, "global_step": 200985, "epoch": 2258, "lr": 1.5779829734960934e-05}
{"train_loss": 0.05644134804606438, "global_step": 200986, "epoch": 2258, "lr": 1.577940704924871e-05}
{"train_loss": 0.07816826552152634, "global_step": 200987, "epoch": 2258, "lr": 1.5778984368137045e-05}
{"train_loss": 0.029348844662308693, "global_step": 200988, "epoch": 2258, "lr": 1.577856169162596e-05}
{"train_loss": 0.111969955265522, "global_step": 200989, "epoch": 2258, "lr": 1.577813901971555e-05}
{"train_loss": 0.04562488570809364, "global_step": 200990, "epoch": 2258, "lr": 1.5777716352405832e-05}
{"train_loss": 0.05644896999001503, "global_step": 200991, "epoch": 2258, "lr": 1.5777293689696902e-05}
{"train_loss": 0.029995650053024292, "global_step": 200992, "epoch": 2258, "lr": 1.577687103158878e-05}
{"train_loss": 0.07038184255361557, "global_step": 200993, "epoch": 2258, "lr": 1.577644837808156e-05}
{"train_loss": 0.07158808410167694, "global_step": 200994, "epoch": 2258, "lr": 1.577602572917526e-05}
{"train_loss": 0.07571841031312943, "global_step": 200995, "epoch": 2258, "lr": 1.577560308486997e-05}
{"train_loss": 0.046695683151483536, "global_step": 200996, "epoch": 2258, "lr": 1.577518044516572e-05}
{"train_loss": 0.038483329117298126, "global_step": 200997, "epoch": 2258, "lr": 1.5774757810062594e-05}
{"train_loss": 0.05839793011546135, "global_step": 200998, "epoch": 2258, "lr": 1.5774335179560622e-05}
{"train_loss": 0.06279249489307404, "global_step": 200999, "epoch": 2258, "lr": 1.5773912553659892e-05}
{"train_loss": 0.04592736065387726, "global_step": 201000, "epoch": 2258, "lr": 1.5773489932360418e-05}
{"train_loss": 0.07169846445322037, "global_step": 201001, "epoch": 2258, "lr": 1.5773067315662304e-05}
{"train_loss": 0.02835497073829174, "global_step": 201002, "epoch": 2258, "lr": 1.5772644703565565e-05}
{"train_loss": 0.0734870433807373, "global_step": 201003, "epoch": 2258, "lr": 1.5772222096070295e-05}
{"train_loss": 0.05874095484614372, "global_step": 201004, "epoch": 2258, "lr": 1.5771799493176525e-05}
{"train_loss": 0.09339091926813126, "global_step": 201005, "epoch": 2258, "lr": 1.577137689488431e-05}
{"train_loss": 0.033324360847473145, "global_step": 201006, "epoch": 2258, "lr": 1.577095430119373e-05}
{"train_loss": 0.04725097119808197, "global_step": 201007, "epoch": 2258, "lr": 1.5770531712104813e-05}
{"train_loss": 0.19965344667434692, "global_step": 201008, "epoch": 2258, "lr": 1.5770109127617648e-05}
{"train_loss": 0.04503704980015755, "global_step": 201009, "epoch": 2258, "lr": 1.576968654773226e-05}
{"train_loss": 0.16296301782131195, "global_step": 201010, "epoch": 2258, "lr": 1.5769263972448732e-05}
{"train_loss": 0.1107042133808136, "global_step": 201011, "epoch": 2258, "lr": 1.57688414017671e-05}
{"train_loss": 0.03410235419869423, "global_step": 201012, "epoch": 2258, "lr": 1.5768418835687438e-05}
{"train_loss": 0.07995783537626266, "global_step": 201013, "epoch": 2258, "lr": 1.5767996274209784e-05}
{"train_loss": 0.08303441852331161, "global_step": 201014, "epoch": 2258, "lr": 1.5767573717334216e-05}
{"train_loss": 0.07400497049093246, "global_step": 201015, "epoch": 2258, "lr": 1.5767151165060768e-05}
{"train_loss": 0.0337255522608757, "global_step": 201016, "epoch": 2258, "lr": 1.5766728617389524e-05}
{"train_loss": 0.05700471252202988, "global_step": 201017, "epoch": 2258, "lr": 1.5766306074320512e-05}
{"train_loss": 0.09463369101285934, "global_step": 201018, "epoch": 2258, "lr": 1.5765883535853803e-05}
{"train_loss": 0.09572329372167587, "global_step": 201019, "epoch": 2258, "lr": 1.5765461001989474e-05}
{"train_loss": 0.057650405913591385, "global_step": 201020, "epoch": 2258, "lr": 1.576503847272754e-05}
{"train_loss": 0.11961902678012848, "global_step": 201021, "epoch": 2258, "lr": 1.5764615948068097e-05}
{"train_loss": 0.0698118656873703, "global_step": 201022, "epoch": 2258, "lr": 1.5764193428011164e-05}
{"train_loss": 0.09106151759624481, "global_step": 201023, "epoch": 2258, "lr": 1.5763770912556842e-05}
{"train_loss": 0.021018970757722855, "global_step": 201024, "epoch": 2258, "lr": 1.576334840170514e-05}
{"train_loss": 0.03156774491071701, "global_step": 201025, "epoch": 2258, "lr": 1.5762925895456153e-05}
{"train_loss": 0.07335598021745682, "global_step": 201026, "epoch": 2258, "lr": 1.5762503393809912e-05}
{"train_loss": 0.0570049062371254, "global_step": 201027, "epoch": 2258, "lr": 1.576208089676651e-05}
{"train_loss": 0.039678577333688736, "global_step": 201028, "epoch": 2258, "lr": 1.5761658404325946e-05}
{"train_loss": 0.04013243690133095, "global_step": 201029, "epoch": 2258, "lr": 1.5761235916488338e-05}
{"train_loss": 0.11528391391038895, "global_step": 201030, "epoch": 2258, "lr": 1.5760813433253692e-05}
{"train_loss": 0.12968334555625916, "global_step": 201031, "epoch": 2258, "lr": 1.5760390954622105e-05}
{"train_loss": 0.027682192623615265, "global_step": 201032, "epoch": 2258, "lr": 1.57599684805936e-05}
{"train_loss": 0.05417932942509651, "global_step": 201033, "epoch": 2258, "lr": 1.5759546011168268e-05}
{"train_loss": 0.10711994767189026, "global_step": 201034, "epoch": 2258, "lr": 1.5759123546346127e-05}
{"train_loss": 0.11978159099817276, "global_step": 201035, "epoch": 2258, "lr": 1.5758701086127274e-05}
{"train_loss": 0.09308458864688873, "global_step": 201036, "epoch": 2258, "lr": 1.575827863051173e-05}
{"train_loss": 0.07090239971876144, "global_step": 201037, "epoch": 2258, "lr": 1.575785617949957e-05}
{"train_loss": 0.0430181510746479, "global_step": 201038, "epoch": 2258, "lr": 1.575743373309086e-05}
{"train_loss": 0.0972391739487648, "global_step": 201039, "epoch": 2258, "lr": 1.575701129128565e-05}
{"train_loss": 0.05349574238061905, "global_step": 201040, "epoch": 2258, "lr": 1.5756588854083976e-05}
{"train_loss": 0.0600961372256279, "global_step": 201041, "epoch": 2258, "lr": 1.5756166421485923e-05}
{"train_loss": 0.10507867485284805, "global_step": 201042, "epoch": 2258, "lr": 1.5755743993491523e-05}
{"train_loss": 0.05459102988243103, "global_step": 201043, "epoch": 2258, "lr": 1.5755321570100867e-05}
{"train_loss": 0.09210842847824097, "global_step": 201044, "epoch": 2258, "lr": 1.5754899151313984e-05}
{"train_loss": 0.04328722506761551, "global_step": 201045, "epoch": 2258, "lr": 1.5754476737130925e-05}
{"train_loss": 0.08557569980621338, "global_step": 201046, "epoch": 2258, "lr": 1.575405432755177e-05}
{"train_loss": 0.06384655088186264, "global_step": 201047, "epoch": 2258, "lr": 1.575363192257655e-05}
{"train_loss": 0.06656245142221451, "global_step": 201048, "epoch": 2258, "lr": 1.5753209522205358e-05}
{"train_loss": 0.06206595152616501, "global_step": 201049, "epoch": 2258, "lr": 1.575278712643821e-05}
{"train_loss": 0.06818132307589723, "global_step": 201050, "epoch": 2258, "lr": 1.57523647352752e-05, "val_loss": 8.206023216247559}
{"train_loss": 0.09276804327964783, "global_step": 201051, "epoch": 2259, "lr": 1.5751942348716348e-05}
{"train_loss": 0.04256831482052803, "global_step": 201052, "epoch": 2259, "lr": 1.575151996676175e-05}
{"train_loss": 0.09206700325012207, "global_step": 201053, "epoch": 2259, "lr": 1.5751097589411425e-05}
{"train_loss": 0.05173357203602791, "global_step": 201054, "epoch": 2259, "lr": 1.5750675216665444e-05}
{"train_loss": 0.04251023381948471, "global_step": 201055, "epoch": 2259, "lr": 1.5750252848523887e-05}
{"train_loss": 0.07596243917942047, "global_step": 201056, "epoch": 2259, "lr": 1.5749830484986778e-05}
{"train_loss": 0.07985153794288635, "global_step": 201057, "epoch": 2259, "lr": 1.57494081260542e-05}
{"train_loss": 0.10460095852613449, "global_step": 201058, "epoch": 2259, "lr": 1.5748985771726175e-05}
{"train_loss": 0.06768931448459625, "global_step": 201059, "epoch": 2259, "lr": 1.5748563422002804e-05}
{"train_loss": 0.08088599145412445, "global_step": 201060, "epoch": 2259, "lr": 1.5748141076884105e-05}
{"train_loss": 0.05005158856511116, "global_step": 201061, "epoch": 2259, "lr": 1.5747718736370164e-05}
{"train_loss": 0.045416489243507385, "global_step": 201062, "epoch": 2259, "lr": 1.5747296400461016e-05}
{"train_loss": 0.036994531750679016, "global_step": 201063, "epoch": 2259, "lr": 1.5746874069156736e-05}
{"train_loss": 0.04661218449473381, "global_step": 201064, "epoch": 2259, "lr": 1.5746451742457357e-05}
{"train_loss": 0.08612348139286041, "global_step": 201065, "epoch": 2259, "lr": 1.574602942036296e-05}
{"train_loss": 0.09119798988103867, "global_step": 201066, "epoch": 2259, "lr": 1.5745607102873584e-05}
{"train_loss": 0.10655529052019119, "global_step": 201067, "epoch": 2259, "lr": 1.574518478998931e-05}
{"train_loss": 0.11037667095661163, "global_step": 201068, "epoch": 2259, "lr": 1.5744762481710156e-05}
{"train_loss": 0.04872054234147072, "global_step": 201069, "epoch": 2259, "lr": 1.5744340178036223e-05}
{"train_loss": 0.06235668063163757, "global_step": 201070, "epoch": 2259, "lr": 1.5743917878967523e-05}
{"train_loss": 0.06020433455705643, "global_step": 201071, "epoch": 2259, "lr": 1.5743495584504158e-05}
{"train_loss": 0.09805796295404434, "global_step": 201072, "epoch": 2259, "lr": 1.5743073294646145e-05}
{"train_loss": 0.09924668818712234, "global_step": 201073, "epoch": 2259, "lr": 1.5742651009393555e-05}
{"train_loss": 0.05426109582185745, "global_step": 201074, "epoch": 2259, "lr": 1.5742228728746466e-05}
{"train_loss": 0.047528225928545, "global_step": 201075, "epoch": 2259, "lr": 1.5741806452704906e-05}
{"train_loss": 0.08419176936149597, "global_step": 201076, "epoch": 2259, "lr": 1.5741384181268955e-05}
{"train_loss": 0.07645853608846664, "global_step": 201077, "epoch": 2259, "lr": 1.5740961914438657e-05}
{"train_loss": 0.034625694155693054, "global_step": 201078, "epoch": 2259, "lr": 1.5740539652214048e-05}
{"train_loss": 0.10873900353908539, "global_step": 201079, "epoch": 2259, "lr": 1.5740117394595227e-05}
{"train_loss": 0.04412120208144188, "global_step": 201080, "epoch": 2259, "lr": 1.573969514158222e-05}
{"train_loss": 0.030577221885323524, "global_step": 201081, "epoch": 2259, "lr": 1.5739272893175094e-05}
{"train_loss": 0.07592695951461792, "global_step": 201082, "epoch": 2259, "lr": 1.5738850649373915e-05}
{"train_loss": 0.03516913205385208, "global_step": 201083, "epoch": 2259, "lr": 1.573842841017871e-05}
{"train_loss": 0.03913984075188637, "global_step": 201084, "epoch": 2259, "lr": 1.573800617558958e-05}
{"train_loss": 0.06922118365764618, "global_step": 201085, "epoch": 2259, "lr": 1.5737583945606532e-05}
{"train_loss": 0.04173751175403595, "global_step": 201086, "epoch": 2259, "lr": 1.573716172022967e-05}
{"train_loss": 0.05442414805293083, "global_step": 201087, "epoch": 2259, "lr": 1.5736739499459013e-05}
{"train_loss": 0.12576255202293396, "global_step": 201088, "epoch": 2259, "lr": 1.573631728329465e-05}
{"train_loss": 0.0882938802242279, "global_step": 201089, "epoch": 2259, "lr": 1.5735895071736602e-05}
{"train_loss": 0.052728597074747086, "global_step": 201090, "epoch": 2259, "lr": 1.5735472864784944e-05}
{"train_loss": 0.07212439179420471, "global_step": 201091, "epoch": 2259, "lr": 1.5735050662439753e-05}
{"train_loss": 0.06817852705717087, "global_step": 201092, "epoch": 2259, "lr": 1.5734628464701052e-05}
{"train_loss": 0.10570642352104187, "global_step": 201093, "epoch": 2259, "lr": 1.5734206271568923e-05}
{"train_loss": 0.07359329611063004, "global_step": 201094, "epoch": 2259, "lr": 1.5733784083043403e-05}
{"train_loss": 0.0798688679933548, "global_step": 201095, "epoch": 2259, "lr": 1.573336189912457e-05}
{"train_loss": 0.03441412001848221, "global_step": 201096, "epoch": 2259, "lr": 1.5732939719812457e-05}
{"train_loss": 0.07822435349225998, "global_step": 201097, "epoch": 2259, "lr": 1.5732517545107144e-05}
{"train_loss": 0.03946871683001518, "global_step": 201098, "epoch": 2259, "lr": 1.5732095375008666e-05}
{"train_loss": 0.05965752899646759, "global_step": 201099, "epoch": 2259, "lr": 1.57316732095171e-05}
{"train_loss": 0.06841350346803665, "global_step": 201100, "epoch": 2259, "lr": 1.5731251048632477e-05}
{"train_loss": 0.0838930755853653, "global_step": 201101, "epoch": 2259, "lr": 1.5730828892354894e-05}
{"train_loss": 0.07316040247678757, "global_step": 201102, "epoch": 2259, "lr": 1.5730406740684357e-05}
{"train_loss": 0.07215844094753265, "global_step": 201103, "epoch": 2259, "lr": 1.572998459362097e-05}
{"train_loss": 0.05324944481253624, "global_step": 201104, "epoch": 2259, "lr": 1.5729562451164753e-05}
{"train_loss": 0.052380893379449844, "global_step": 201105, "epoch": 2259, "lr": 1.5729140313315794e-05}
{"train_loss": 0.061240971088409424, "global_step": 201106, "epoch": 2259, "lr": 1.5728718180074116e-05}
{"train_loss": 0.050052907317876816, "global_step": 201107, "epoch": 2259, "lr": 1.5728296051439812e-05}
{"train_loss": 0.04725579917430878, "global_step": 201108, "epoch": 2259, "lr": 1.5727873927412906e-05}
{"train_loss": 0.07830489426851273, "global_step": 201109, "epoch": 2259, "lr": 1.572745180799347e-05}
{"train_loss": 0.04879912734031677, "global_step": 201110, "epoch": 2259, "lr": 1.572702969318158e-05}
{"train_loss": 0.08777919411659241, "global_step": 201111, "epoch": 2259, "lr": 1.5726607582977248e-05}
{"train_loss": 0.1055263876914978, "global_step": 201112, "epoch": 2259, "lr": 1.5726185477380577e-05}
{"train_loss": 0.022826476022601128, "global_step": 201113, "epoch": 2259, "lr": 1.5725763376391582e-05}
{"train_loss": 0.058016180992126465, "global_step": 201114, "epoch": 2259, "lr": 1.5725341280010365e-05}
{"train_loss": 0.045171890407800674, "global_step": 201115, "epoch": 2259, "lr": 1.5724919188236953e-05}
{"train_loss": 0.10550566017627716, "global_step": 201116, "epoch": 2259, "lr": 1.572449710107139e-05}
{"train_loss": 0.07318782806396484, "global_step": 201117, "epoch": 2259, "lr": 1.5724075018513768e-05}
{"train_loss": 0.09325087070465088, "global_step": 201118, "epoch": 2259, "lr": 1.5723652940564116e-05}
{"train_loss": 0.04984530434012413, "global_step": 201119, "epoch": 2259, "lr": 1.572323086722251e-05}
{"train_loss": 0.03632306307554245, "global_step": 201120, "epoch": 2259, "lr": 1.5722808798488998e-05}
{"train_loss": 0.07067368924617767, "global_step": 201121, "epoch": 2259, "lr": 1.5722386734363625e-05}
{"train_loss": 0.052421584725379944, "global_step": 201122, "epoch": 2259, "lr": 1.5721964674846475e-05}
{"train_loss": 0.07273314148187637, "global_step": 201123, "epoch": 2259, "lr": 1.572154261993757e-05}
{"train_loss": 0.0888083204627037, "global_step": 201124, "epoch": 2259, "lr": 1.5721120569636998e-05}
{"train_loss": 0.08587843179702759, "global_step": 201125, "epoch": 2259, "lr": 1.5720698523944794e-05}
{"train_loss": 0.0662921816110611, "global_step": 201126, "epoch": 2259, "lr": 1.5720276482861042e-05}
{"train_loss": 0.2125139832496643, "global_step": 201127, "epoch": 2259, "lr": 1.5719854446385763e-05}
{"train_loss": 0.08465924859046936, "global_step": 201128, "epoch": 2259, "lr": 1.5719432414519027e-05}
{"train_loss": 0.04264718294143677, "global_step": 201129, "epoch": 2259, "lr": 1.571901038726092e-05}
{"train_loss": 0.05295196548104286, "global_step": 201130, "epoch": 2259, "lr": 1.5718588364611447e-05}
{"train_loss": 0.06223466992378235, "global_step": 201131, "epoch": 2259, "lr": 1.571816634657072e-05}
{"train_loss": 0.06952020525932312, "global_step": 201132, "epoch": 2259, "lr": 1.571774433313874e-05}
{"train_loss": 0.021659307181835175, "global_step": 201133, "epoch": 2259, "lr": 1.5717322324315613e-05}
{"train_loss": 0.07068666070699692, "global_step": 201134, "epoch": 2259, "lr": 1.5716900320101362e-05}
{"train_loss": 0.03907448425889015, "global_step": 201135, "epoch": 2259, "lr": 1.571647832049607e-05}
{"train_loss": 0.046662673354148865, "global_step": 201136, "epoch": 2259, "lr": 1.5716056325499762e-05}
{"train_loss": 0.0333629734814167, "global_step": 201137, "epoch": 2259, "lr": 1.5715634335112528e-05}
{"train_loss": 0.06214958429336548, "global_step": 201138, "epoch": 2259, "lr": 1.571521234933439e-05}
{"train_loss": 0.06781496429878674, "global_step": 201139, "epoch": 2259, "lr": 1.5714790368165446e-05, "val_loss": 8.130989074707031}
{"train_loss": 0.0748799592256546, "global_step": 201140, "epoch": 2260, "lr": 1.571436839160571e-05}
{"train_loss": 0.05734088271856308, "global_step": 201141, "epoch": 2260, "lr": 1.571394641965528e-05}
{"train_loss": 0.12618152797222137, "global_step": 201142, "epoch": 2260, "lr": 1.571352445231417e-05}
{"train_loss": 0.09550727903842926, "global_step": 201143, "epoch": 2260, "lr": 1.5713102489582478e-05}
{"train_loss": 0.038827575743198395, "global_step": 201144, "epoch": 2260, "lr": 1.5712680531460226e-05}
{"train_loss": 0.09584660828113556, "global_step": 201145, "epoch": 2260, "lr": 1.5712258577947487e-05}
{"train_loss": 0.0655321329832077, "global_step": 201146, "epoch": 2260, "lr": 1.571183662904433e-05}
{"train_loss": 0.09013330191373825, "global_step": 201147, "epoch": 2260, "lr": 1.5711414684750788e-05}
{"train_loss": 0.07698012888431549, "global_step": 201148, "epoch": 2260, "lr": 1.571099274506694e-05}
{"train_loss": 0.061665866523981094, "global_step": 201149, "epoch": 2260, "lr": 1.5710570809992813e-05}
{"train_loss": 0.04972871020436287, "global_step": 201150, "epoch": 2260, "lr": 1.5710148879528497e-05}
{"train_loss": 0.06384213268756866, "global_step": 201151, "epoch": 2260, "lr": 1.5709726953674027e-05}
{"train_loss": 0.04181941598653793, "global_step": 201152, "epoch": 2260, "lr": 1.5709305032429472e-05}
{"train_loss": 0.06785707920789719, "global_step": 201153, "epoch": 2260, "lr": 1.570888311579487e-05}
{"train_loss": 0.03384587541222572, "global_step": 201154, "epoch": 2260, "lr": 1.5708461203770307e-05}
{"train_loss": 0.11430437117815018, "global_step": 201155, "epoch": 2260, "lr": 1.5708039296355825e-05}
{"train_loss": 0.04681779816746712, "global_step": 201156, "epoch": 2260, "lr": 1.5707617393551465e-05}
{"train_loss": 0.08123720437288284, "global_step": 201157, "epoch": 2260, "lr": 1.5707195495357313e-05}
{"train_loss": 0.052940525114536285, "global_step": 201158, "epoch": 2260, "lr": 1.5706773601773388e-05}
{"train_loss": 0.051118992269039154, "global_step": 201159, "epoch": 2260, "lr": 1.5706351712799792e-05}
{"train_loss": 0.07743269205093384, "global_step": 201160, "epoch": 2260, "lr": 1.570592982843656e-05}
{"train_loss": 0.05143361538648605, "global_step": 201161, "epoch": 2260, "lr": 1.5705507948683728e-05}
{"train_loss": 0.04577740654349327, "global_step": 201162, "epoch": 2260, "lr": 1.570508607354139e-05}
{"train_loss": 0.030135096982121468, "global_step": 201163, "epoch": 2260, "lr": 1.5704664203009566e-05}
{"train_loss": 0.05527881532907486, "global_step": 201164, "epoch": 2260, "lr": 1.5704242337088338e-05}
{"train_loss": 0.0734107494354248, "global_step": 201165, "epoch": 2260, "lr": 1.570382047577777e-05}
{"train_loss": 0.0537676140666008, "global_step": 201166, "epoch": 2260, "lr": 1.5703398619077886e-05}
{"train_loss": 0.05824528634548187, "global_step": 201167, "epoch": 2260, "lr": 1.570297676698878e-05}
{"train_loss": 0.06880373507738113, "global_step": 201168, "epoch": 2260, "lr": 1.5702554919510475e-05}
{"train_loss": 0.07322896271944046, "global_step": 201169, "epoch": 2260, "lr": 1.570213307664306e-05}
{"train_loss": 0.10125694423913956, "global_step": 201170, "epoch": 2260, "lr": 1.5701711238386563e-05}
{"train_loss": 0.03735615685582161, "global_step": 201171, "epoch": 2260, "lr": 1.5701289404741065e-05}
{"train_loss": 0.06353962421417236, "global_step": 201172, "epoch": 2260, "lr": 1.570086757570659e-05}
{"train_loss": 0.05949457734823227, "global_step": 201173, "epoch": 2260, "lr": 1.5700445751283238e-05}
{"train_loss": 0.054414790123701096, "global_step": 201174, "epoch": 2260, "lr": 1.5700023931471024e-05}
{"train_loss": 0.042816027998924255, "global_step": 201175, "epoch": 2260, "lr": 1.569960211627004e-05}
{"train_loss": 0.044320475310087204, "global_step": 201176, "epoch": 2260, "lr": 1.5699180305680312e-05}
{"train_loss": 0.05924174189567566, "global_step": 201177, "epoch": 2260, "lr": 1.5698758499701925e-05}
{"train_loss": 0.0823858231306076, "global_step": 201178, "epoch": 2260, "lr": 1.5698336698334908e-05}
{"train_loss": 0.05830322578549385, "global_step": 201179, "epoch": 2260, "lr": 1.5697914901579347e-05}
{"train_loss": 0.05773685500025749, "global_step": 201180, "epoch": 2260, "lr": 1.569749310943527e-05}
{"train_loss": 0.06486205756664276, "global_step": 201181, "epoch": 2260, "lr": 1.569707132190274e-05}
{"train_loss": 0.04309411346912384, "global_step": 201182, "epoch": 2260, "lr": 1.569664953898184e-05}
{"train_loss": 0.07599378377199173, "global_step": 201183, "epoch": 2260, "lr": 1.5696227760672593e-05}
{"train_loss": 0.0325879231095314, "global_step": 201184, "epoch": 2260, "lr": 1.5695805986975086e-05}
{"train_loss": 0.05147312954068184, "global_step": 201185, "epoch": 2260, "lr": 1.569538421788934e-05}
{"train_loss": 0.03400178626179695, "global_step": 201186, "epoch": 2260, "lr": 1.569496245341545e-05}
{"train_loss": 0.0433105006814003, "global_step": 201187, "epoch": 2260, "lr": 1.569454069355344e-05}
{"train_loss": 0.04837154969573021, "global_step": 201188, "epoch": 2260, "lr": 1.5694118938303392e-05}
{"train_loss": 0.07961660623550415, "global_step": 201189, "epoch": 2260, "lr": 1.569369718766534e-05}
{"train_loss": 0.020378490909934044, "global_step": 201190, "epoch": 2260, "lr": 1.569327544163937e-05}
{"train_loss": 0.09210412204265594, "global_step": 201191, "epoch": 2260, "lr": 1.56928537002255e-05}
{"train_loss": 0.09548147022724152, "global_step": 201192, "epoch": 2260, "lr": 1.5692431963423826e-05}
{"train_loss": 0.06007998436689377, "global_step": 201193, "epoch": 2260, "lr": 1.569201023123439e-05}
{"train_loss": 0.04987018555402756, "global_step": 201194, "epoch": 2260, "lr": 1.5691588503657223e-05}
{"train_loss": 0.05898605287075043, "global_step": 201195, "epoch": 2260, "lr": 1.5691166780692422e-05}
{"train_loss": 0.11155620217323303, "global_step": 201196, "epoch": 2260, "lr": 1.5690745062340007e-05}
{"train_loss": 0.09990276396274567, "global_step": 201197, "epoch": 2260, "lr": 1.5690323348600073e-05}
{"train_loss": 0.039072807878255844, "global_step": 201198, "epoch": 2260, "lr": 1.5689901639472653e-05}
{"train_loss": 0.09641756117343903, "global_step": 201199, "epoch": 2260, "lr": 1.5689479934957795e-05}
{"train_loss": 0.050822775810956955, "global_step": 201200, "epoch": 2260, "lr": 1.568905823505557e-05}
{"train_loss": 0.045816570520401, "global_step": 201201, "epoch": 2260, "lr": 1.5688636539766043e-05}
{"train_loss": 0.05249963328242302, "global_step": 201202, "epoch": 2260, "lr": 1.568821484908925e-05}
{"train_loss": 0.09104081988334656, "global_step": 201203, "epoch": 2260, "lr": 1.568779316302527e-05}
{"train_loss": 0.03633565455675125, "global_step": 201204, "epoch": 2260, "lr": 1.5687371481574137e-05}
{"train_loss": 0.13668404519557953, "global_step": 201205, "epoch": 2260, "lr": 1.5686949804735928e-05}
{"train_loss": 0.06286343932151794, "global_step": 201206, "epoch": 2260, "lr": 1.568652813251068e-05}
{"train_loss": 0.03335390239953995, "global_step": 201207, "epoch": 2260, "lr": 1.5686106464898477e-05}
{"train_loss": 0.04175131022930145, "global_step": 201208, "epoch": 2260, "lr": 1.5685684801899336e-05}
{"train_loss": 0.075296550989151, "global_step": 201209, "epoch": 2260, "lr": 1.5685263143513357e-05}
{"train_loss": 0.06283721327781677, "global_step": 201210, "epoch": 2260, "lr": 1.568484148974056e-05}
{"train_loss": 0.09281059354543686, "global_step": 201211, "epoch": 2260, "lr": 1.5684419840581033e-05}
{"train_loss": 0.09227557480335236, "global_step": 201212, "epoch": 2260, "lr": 1.56839981960348e-05}
{"train_loss": 0.054276153445243835, "global_step": 201213, "epoch": 2260, "lr": 1.5683576556101952e-05}
{"train_loss": 0.0920514240860939, "global_step": 201214, "epoch": 2260, "lr": 1.5683154920782517e-05}
{"train_loss": 0.055153947323560715, "global_step": 201215, "epoch": 2260, "lr": 1.5682733290076572e-05}
{"train_loss": 0.07053142786026001, "global_step": 201216, "epoch": 2260, "lr": 1.5682311663984158e-05}
{"train_loss": 0.07593100517988205, "global_step": 201217, "epoch": 2260, "lr": 1.5681890042505332e-05}
{"train_loss": 0.0630689263343811, "global_step": 201218, "epoch": 2260, "lr": 1.5681468425640176e-05}
{"train_loss": 0.026838980615139008, "global_step": 201219, "epoch": 2260, "lr": 1.5681046813388718e-05}
{"train_loss": 0.0567147471010685, "global_step": 201220, "epoch": 2260, "lr": 1.568062520575103e-05}
{"train_loss": 0.07152383029460907, "global_step": 201221, "epoch": 2260, "lr": 1.5680203602727155e-05}
{"train_loss": 0.07599227875471115, "global_step": 201222, "epoch": 2260, "lr": 1.5679782004317174e-05}
{"train_loss": 0.05575045198202133, "global_step": 201223, "epoch": 2260, "lr": 1.567936041052111e-05}
{"train_loss": 0.06030237674713135, "global_step": 201224, "epoch": 2260, "lr": 1.567893882133905e-05}
{"train_loss": 0.04560841992497444, "global_step": 201225, "epoch": 2260, "lr": 1.5678517236771027e-05}
{"train_loss": 0.06823389232158661, "global_step": 201226, "epoch": 2260, "lr": 1.5678095656817128e-05}
{"train_loss": 0.09326633810997009, "global_step": 201227, "epoch": 2260, "lr": 1.5677674081477373e-05}
{"train_loss": 0.0649499250680543, "global_step": 201228, "epoch": 2260, "lr": 1.567725251075185e-05, "val_loss": 8.291049003601074, "train_action_mse_error": 10.495603561401367}
{"train_loss": 0.05627724155783653, "global_step": 201229, "epoch": 2261, "lr": 1.5676830944640586e-05}
{"train_loss": 0.09353769570589066, "global_step": 201230, "epoch": 2261, "lr": 1.567640938314367e-05}
{"train_loss": 0.04567067697644234, "global_step": 201231, "epoch": 2261, "lr": 1.567598782626114e-05}
{"train_loss": 0.09456390887498856, "global_step": 201232, "epoch": 2261, "lr": 1.567556627399304e-05}
{"train_loss": 0.08272887021303177, "global_step": 201233, "epoch": 2261, "lr": 1.5675144726339462e-05}
{"train_loss": 0.053098972886800766, "global_step": 201234, "epoch": 2261, "lr": 1.5674723183300427e-05}
{"train_loss": 0.0537882000207901, "global_step": 201235, "epoch": 2261, "lr": 1.5674301644876015e-05}
{"train_loss": 0.02732452191412449, "global_step": 201236, "epoch": 2261, "lr": 1.5673880111066268e-05}
{"train_loss": 0.07193614542484283, "global_step": 201237, "epoch": 2261, "lr": 1.567345858187126e-05}
{"train_loss": 0.06154109910130501, "global_step": 201238, "epoch": 2261, "lr": 1.5673037057291023e-05}
{"train_loss": 0.06036732345819473, "global_step": 201239, "epoch": 2261, "lr": 1.567261553732564e-05}
{"train_loss": 0.0519317165017128, "global_step": 201240, "epoch": 2261, "lr": 1.5672194021975144e-05}
{"train_loss": 0.056119099259376526, "global_step": 201241, "epoch": 2261, "lr": 1.567177251123962e-05}
{"train_loss": 0.06854856014251709, "global_step": 201242, "epoch": 2261, "lr": 1.567135100511909e-05}
{"train_loss": 0.06425321102142334, "global_step": 201243, "epoch": 2261, "lr": 1.5670929503613646e-05}
{"train_loss": 0.07305502891540527, "global_step": 201244, "epoch": 2261, "lr": 1.5670508006723304e-05}
{"train_loss": 0.07074085623025894, "global_step": 201245, "epoch": 2261, "lr": 1.5670086514448167e-05}
{"train_loss": 0.12081719189882278, "global_step": 201246, "epoch": 2261, "lr": 1.5669665026788248e-05}
{"train_loss": 0.04125877097249031, "global_step": 201247, "epoch": 2261, "lr": 1.5669243543743646e-05}
{"train_loss": 0.06711498647928238, "global_step": 201248, "epoch": 2261, "lr": 1.5668822065314377e-05}
{"train_loss": 0.04101383686065674, "global_step": 201249, "epoch": 2261, "lr": 1.5668400591500533e-05}
{"train_loss": 0.03848004341125488, "global_step": 201250, "epoch": 2261, "lr": 1.5667979122302134e-05}
{"train_loss": 0.0625535249710083, "global_step": 201251, "epoch": 2261, "lr": 1.566755765771928e-05}
{"train_loss": 0.0925450548529625, "global_step": 201252, "epoch": 2261, "lr": 1.566713619775198e-05}
{"train_loss": 0.05298800393939018, "global_step": 201253, "epoch": 2261, "lr": 1.5666714742400335e-05}
{"train_loss": 0.05581918731331825, "global_step": 201254, "epoch": 2261, "lr": 1.5666293291664367e-05}
{"train_loss": 0.07089453935623169, "global_step": 201255, "epoch": 2261, "lr": 1.5665871845544144e-05}
{"train_loss": 0.0722663626074791, "global_step": 201256, "epoch": 2261, "lr": 1.5665450404039745e-05}
{"train_loss": 0.10387419164180756, "global_step": 201257, "epoch": 2261, "lr": 1.566502896715119e-05}
{"train_loss": 0.033571507781744, "global_step": 201258, "epoch": 2261, "lr": 1.5664607534878573e-05}
{"train_loss": 0.1013101264834404, "global_step": 201259, "epoch": 2261, "lr": 1.5664186107221913e-05}
{"train_loss": 0.12851527333259583, "global_step": 201260, "epoch": 2261, "lr": 1.5663764684181297e-05}
{"train_loss": 0.03960398584604263, "global_step": 201261, "epoch": 2261, "lr": 1.5663343265756757e-05}
{"train_loss": 0.04052840918302536, "global_step": 201262, "epoch": 2261, "lr": 1.566292185194838e-05}
{"train_loss": 0.08860933780670166, "global_step": 201263, "epoch": 2261, "lr": 1.5662500442756185e-05}
{"train_loss": 0.03314243629574776, "global_step": 201264, "epoch": 2261, "lr": 1.5662079038180267e-05}
{"train_loss": 0.06955312937498093, "global_step": 201265, "epoch": 2261, "lr": 1.5661657638220645e-05}
{"train_loss": 0.04259781539440155, "global_step": 201266, "epoch": 2261, "lr": 1.5661236242877414e-05}
{"train_loss": 0.07288049161434174, "global_step": 201267, "epoch": 2261, "lr": 1.566081485215059e-05}
{"train_loss": 0.0705677792429924, "global_step": 201268, "epoch": 2261, "lr": 1.566039346604027e-05}
{"train_loss": 0.029677562415599823, "global_step": 201269, "epoch": 2261, "lr": 1.5659972084546474e-05}
{"train_loss": 0.048764437437057495, "global_step": 201270, "epoch": 2261, "lr": 1.5659550707669297e-05}
{"train_loss": 0.11084040254354477, "global_step": 201271, "epoch": 2261, "lr": 1.5659129335408774e-05}
{"train_loss": 0.029905853793025017, "global_step": 201272, "epoch": 2261, "lr": 1.565870796776494e-05}
{"train_loss": 0.0621495246887207, "global_step": 201273, "epoch": 2261, "lr": 1.5658286604737893e-05}
{"train_loss": 0.09756343066692352, "global_step": 201274, "epoch": 2261, "lr": 1.5657865246327656e-05}
{"train_loss": 0.060268744826316833, "global_step": 201275, "epoch": 2261, "lr": 1.565744389253432e-05}
{"train_loss": 0.11366567015647888, "global_step": 201276, "epoch": 2261, "lr": 1.5657022543357903e-05}
{"train_loss": 0.04946611076593399, "global_step": 201277, "epoch": 2261, "lr": 1.565660119879849e-05}
{"train_loss": 0.1142255887389183, "global_step": 201278, "epoch": 2261, "lr": 1.5656179858856125e-05}
{"train_loss": 0.05116330832242966, "global_step": 201279, "epoch": 2261, "lr": 1.5655758523530877e-05}
{"train_loss": 0.04772268980741501, "global_step": 201280, "epoch": 2261, "lr": 1.5655337192822778e-05}
{"train_loss": 0.08019275218248367, "global_step": 201281, "epoch": 2261, "lr": 1.5654915866731917e-05}
{"train_loss": 0.07275991141796112, "global_step": 201282, "epoch": 2261, "lr": 1.5654494545258318e-05}
{"train_loss": 0.03657156974077225, "global_step": 201283, "epoch": 2261, "lr": 1.565407322840207e-05}
{"train_loss": 0.07920824736356735, "global_step": 201284, "epoch": 2261, "lr": 1.5653651916163197e-05}
{"train_loss": 0.06300567835569382, "global_step": 201285, "epoch": 2261, "lr": 1.5653230608541785e-05}
{"train_loss": 0.04039739444851875, "global_step": 201286, "epoch": 2261, "lr": 1.5652809305537864e-05}
{"train_loss": 0.03692210465669632, "global_step": 201287, "epoch": 2261, "lr": 1.565238800715152e-05}
{"train_loss": 0.054849911481142044, "global_step": 201288, "epoch": 2261, "lr": 1.5651966713382782e-05}
{"train_loss": 0.07484593242406845, "global_step": 201289, "epoch": 2261, "lr": 1.565154542423173e-05}
{"train_loss": 0.04849115386605263, "global_step": 201290, "epoch": 2261, "lr": 1.5651124139698392e-05}
{"train_loss": 0.08496539294719696, "global_step": 201291, "epoch": 2261, "lr": 1.5650702859782844e-05}
{"train_loss": 0.037265680730342865, "global_step": 201292, "epoch": 2261, "lr": 1.5650281584485154e-05}
{"train_loss": 0.0927833542227745, "global_step": 201293, "epoch": 2261, "lr": 1.5649860313805352e-05}
{"train_loss": 0.05117006599903107, "global_step": 201294, "epoch": 2261, "lr": 1.564943904774352e-05}
{"train_loss": 0.04641912132501602, "global_step": 201295, "epoch": 2261, "lr": 1.564901778629969e-05}
{"train_loss": 0.061013732105493546, "global_step": 201296, "epoch": 2261, "lr": 1.5648596529473946e-05}
{"train_loss": 0.07855411618947983, "global_step": 201297, "epoch": 2261, "lr": 1.5648175277266313e-05}
{"train_loss": 0.09735511988401413, "global_step": 201298, "epoch": 2261, "lr": 1.5647754029676877e-05}
{"train_loss": 0.12117920070886612, "global_step": 201299, "epoch": 2261, "lr": 1.564733278670567e-05}
{"train_loss": 0.08361164480447769, "global_step": 201300, "epoch": 2261, "lr": 1.5646911548352777e-05}
{"train_loss": 0.08793441206216812, "global_step": 201301, "epoch": 2261, "lr": 1.5646490314618217e-05}
{"train_loss": 0.09315735846757889, "global_step": 201302, "epoch": 2261, "lr": 1.5646069085502085e-05}
{"train_loss": 0.11357760429382324, "global_step": 201303, "epoch": 2261, "lr": 1.564564786100441e-05}
{"train_loss": 0.04359794408082962, "global_step": 201304, "epoch": 2261, "lr": 1.564522664112527e-05}
{"train_loss": 0.0888611301779747, "global_step": 201305, "epoch": 2261, "lr": 1.5644805425864696e-05}
{"train_loss": 0.03909124433994293, "global_step": 201306, "epoch": 2261, "lr": 1.5644384215222773e-05}
{"train_loss": 0.030890321359038353, "global_step": 201307, "epoch": 2261, "lr": 1.564396300919953e-05}
{"train_loss": 0.04857480153441429, "global_step": 201308, "epoch": 2261, "lr": 1.5643541807795053e-05}
{"train_loss": 0.04845356568694115, "global_step": 201309, "epoch": 2261, "lr": 1.5643120611009382e-05}
{"train_loss": 0.07550253719091415, "global_step": 201310, "epoch": 2261, "lr": 1.5642699418842562e-05}
{"train_loss": 0.11578343063592911, "global_step": 201311, "epoch": 2261, "lr": 1.5642278231294676e-05}
{"train_loss": 0.05479756370186806, "global_step": 201312, "epoch": 2261, "lr": 1.564185704836575e-05}
{"train_loss": 0.07335727661848068, "global_step": 201313, "epoch": 2261, "lr": 1.5641435870055877e-05}
{"train_loss": 0.04682463780045509, "global_step": 201314, "epoch": 2261, "lr": 1.5641014696365074e-05}
{"train_loss": 0.026704788208007812, "global_step": 201315, "epoch": 2261, "lr": 1.5640593527293434e-05}
{"train_loss": 0.05041833594441414, "global_step": 201316, "epoch": 2261, "lr": 1.5640172362840983e-05}
{"train_loss": 0.06607603924244307, "global_step": 201317, "epoch": 2261, "lr": 1.563975120300781e-05, "val_loss": 8.293081283569336}
{"train_loss": 0.04537559300661087, "global_step": 201318, "epoch": 2262, "lr": 1.5639330047793932e-05}
{"train_loss": 0.04273385554552078, "global_step": 201319, "epoch": 2262, "lr": 1.563890889719945e-05}
{"train_loss": 0.06764154136180878, "global_step": 201320, "epoch": 2262, "lr": 1.563848775122438e-05}
{"train_loss": 0.08353961259126663, "global_step": 201321, "epoch": 2262, "lr": 1.5638066609868806e-05}
{"train_loss": 0.04271874576807022, "global_step": 201322, "epoch": 2262, "lr": 1.5637645473132767e-05}
{"train_loss": 0.03537093102931976, "global_step": 201323, "epoch": 2262, "lr": 1.5637224341016344e-05}
{"train_loss": 0.0973387211561203, "global_step": 201324, "epoch": 2262, "lr": 1.5636803213519558e-05}
{"train_loss": 0.06967716664075851, "global_step": 201325, "epoch": 2262, "lr": 1.56363820906425e-05}
{"train_loss": 0.04335146024823189, "global_step": 201326, "epoch": 2262, "lr": 1.5635960972385195e-05}
{"train_loss": 0.0766744539141655, "global_step": 201327, "epoch": 2262, "lr": 1.5635539858747717e-05}
{"train_loss": 0.06482335180044174, "global_step": 201328, "epoch": 2262, "lr": 1.5635118749730144e-05}
{"train_loss": 0.0716792494058609, "global_step": 201329, "epoch": 2262, "lr": 1.5634697645332484e-05}
{"train_loss": 0.04954350367188454, "global_step": 201330, "epoch": 2262, "lr": 1.5634276545554844e-05}
{"train_loss": 0.055170971900224686, "global_step": 201331, "epoch": 2262, "lr": 1.563385545039724e-05}
{"train_loss": 0.09526640176773071, "global_step": 201332, "epoch": 2262, "lr": 1.5633434359859757e-05}
{"train_loss": 0.03211217373609543, "global_step": 201333, "epoch": 2262, "lr": 1.5633013273942425e-05}
{"train_loss": 0.08719910681247711, "global_step": 201334, "epoch": 2262, "lr": 1.5632592192645333e-05}
{"train_loss": 0.0991361141204834, "global_step": 201335, "epoch": 2262, "lr": 1.5632171115968507e-05}
{"train_loss": 0.042469434440135956, "global_step": 201336, "epoch": 2262, "lr": 1.563175004391203e-05}
{"train_loss": 0.06166394427418709, "global_step": 201337, "epoch": 2262, "lr": 1.563132897647593e-05}
{"train_loss": 0.019508637487888336, "global_step": 201338, "epoch": 2262, "lr": 1.5630907913660293e-05}
{"train_loss": 0.07478892803192139, "global_step": 201339, "epoch": 2262, "lr": 1.5630486855465144e-05}
{"train_loss": 0.04433850198984146, "global_step": 201340, "epoch": 2262, "lr": 1.5630065801890576e-05}
{"train_loss": 0.044553790241479874, "global_step": 201341, "epoch": 2262, "lr": 1.5629644752936613e-05}
{"train_loss": 0.06426354497671127, "global_step": 201342, "epoch": 2262, "lr": 1.5629223708603334e-05}
{"train_loss": 0.07747156918048859, "global_step": 201343, "epoch": 2262, "lr": 1.562880266889078e-05}
{"train_loss": 0.09670510143041611, "global_step": 201344, "epoch": 2262, "lr": 1.5628381633799022e-05}
{"train_loss": 0.05473974719643593, "global_step": 201345, "epoch": 2262, "lr": 1.56279606033281e-05}
{"train_loss": 0.04102853685617447, "global_step": 201346, "epoch": 2262, "lr": 1.5627539577478094e-05}
{"train_loss": 0.07865014672279358, "global_step": 201347, "epoch": 2262, "lr": 1.5627118556249044e-05}
{"train_loss": 0.08086738735437393, "global_step": 201348, "epoch": 2262, "lr": 1.5626697539640993e-05}
{"train_loss": 0.08226653933525085, "global_step": 201349, "epoch": 2262, "lr": 1.5626276527654032e-05}
{"train_loss": 0.041634391993284225, "global_step": 201350, "epoch": 2262, "lr": 1.5625855520288185e-05}
{"train_loss": 0.10230247676372528, "global_step": 201351, "epoch": 2262, "lr": 1.5625434517543535e-05}
{"train_loss": 0.07990992069244385, "global_step": 201352, "epoch": 2262, "lr": 1.562501351942011e-05}
{"train_loss": 0.031914301216602325, "global_step": 201353, "epoch": 2262, "lr": 1.5624592525918e-05}
{"train_loss": 0.10252206772565842, "global_step": 201354, "epoch": 2262, "lr": 1.562417153703723e-05}
{"train_loss": 0.08512228727340698, "global_step": 201355, "epoch": 2262, "lr": 1.5623750552777883e-05}
{"train_loss": 0.07734957337379456, "global_step": 201356, "epoch": 2262, "lr": 1.5623329573139995e-05}
{"train_loss": 0.06511878967285156, "global_step": 201357, "epoch": 2262, "lr": 1.5622908598123643e-05}
{"train_loss": 0.057110875844955444, "global_step": 201358, "epoch": 2262, "lr": 1.5622487627728856e-05}
{"train_loss": 0.08241936564445496, "global_step": 201359, "epoch": 2262, "lr": 1.562206666195572e-05}
{"train_loss": 0.03354164585471153, "global_step": 201360, "epoch": 2262, "lr": 1.5621645700804265e-05}
{"train_loss": 0.08420838415622711, "global_step": 201361, "epoch": 2262, "lr": 1.5621224744274577e-05}
{"train_loss": 0.0343673937022686, "global_step": 201362, "epoch": 2262, "lr": 1.5620803792366677e-05}
{"train_loss": 0.043010059744119644, "global_step": 201363, "epoch": 2262, "lr": 1.5620382845080646e-05}
{"train_loss": 0.05226772278547287, "global_step": 201364, "epoch": 2262, "lr": 1.5619961902416545e-05}
{"train_loss": 0.08924189954996109, "global_step": 201365, "epoch": 2262, "lr": 1.561954096437441e-05}
{"train_loss": 0.08994058519601822, "global_step": 201366, "epoch": 2262, "lr": 1.5619120030954327e-05}
{"train_loss": 0.04516319930553436, "global_step": 201367, "epoch": 2262, "lr": 1.5618699102156314e-05}
{"train_loss": 0.0490570142865181, "global_step": 201368, "epoch": 2262, "lr": 1.5618278177980466e-05}
{"train_loss": 0.07014814764261246, "global_step": 201369, "epoch": 2262, "lr": 1.5617857258426805e-05}
{"train_loss": 0.09928803890943527, "global_step": 201370, "epoch": 2262, "lr": 1.5617436343495417e-05}
{"train_loss": 0.04356140270829201, "global_step": 201371, "epoch": 2262, "lr": 1.561701543318633e-05}
{"train_loss": 0.029595116153359413, "global_step": 201372, "epoch": 2262, "lr": 1.561659452749964e-05}
{"train_loss": 0.057314183562994, "global_step": 201373, "epoch": 2262, "lr": 1.5616173626435353e-05}
{"train_loss": 0.12115379422903061, "global_step": 201374, "epoch": 2262, "lr": 1.5615752729993576e-05}
{"train_loss": 0.05802140757441521, "global_step": 201375, "epoch": 2262, "lr": 1.5615331838174325e-05}
{"train_loss": 0.0862153172492981, "global_step": 201376, "epoch": 2262, "lr": 1.5614910950977686e-05}
{"train_loss": 0.07403479516506195, "global_step": 201377, "epoch": 2262, "lr": 1.561449006840369e-05}
{"train_loss": 0.045477572828531265, "global_step": 201378, "epoch": 2262, "lr": 1.5614069190452423e-05}
{"train_loss": 0.0704328790307045, "global_step": 201379, "epoch": 2262, "lr": 1.5613648317123908e-05}
{"train_loss": 0.06808990240097046, "global_step": 201380, "epoch": 2262, "lr": 1.561322744841823e-05}
{"train_loss": 0.11164334416389465, "global_step": 201381, "epoch": 2262, "lr": 1.5612806584335427e-05}
{"train_loss": 0.03907649219036102, "global_step": 201382, "epoch": 2262, "lr": 1.561238572487558e-05}
{"train_loss": 0.038190897554159164, "global_step": 201383, "epoch": 2262, "lr": 1.5611964870038705e-05}
{"train_loss": 0.08433865755796432, "global_step": 201384, "epoch": 2262, "lr": 1.5611544019824896e-05}
{"train_loss": 0.09386046230792999, "global_step": 201385, "epoch": 2262, "lr": 1.561112317423419e-05}
{"train_loss": 0.06634502857923508, "global_step": 201386, "epoch": 2262, "lr": 1.5610702333266656e-05}
{"train_loss": 0.08296490460634232, "global_step": 201387, "epoch": 2262, "lr": 1.561028149692235e-05}
{"train_loss": 0.07667956501245499, "global_step": 201388, "epoch": 2262, "lr": 1.5609860665201308e-05}
{"train_loss": 0.05646958574652672, "global_step": 201389, "epoch": 2262, "lr": 1.560943983810361e-05}
{"train_loss": 0.05563781410455704, "global_step": 201390, "epoch": 2262, "lr": 1.5609019015629294e-05}
{"train_loss": 0.02535349689424038, "global_step": 201391, "epoch": 2262, "lr": 1.560859819777844e-05}
{"train_loss": 0.031154101714491844, "global_step": 201392, "epoch": 2262, "lr": 1.5608177384551075e-05}
{"train_loss": 0.09228121489286423, "global_step": 201393, "epoch": 2262, "lr": 1.5607756575947286e-05}
{"train_loss": 0.056432079523801804, "global_step": 201394, "epoch": 2262, "lr": 1.5607335771967103e-05}
{"train_loss": 0.07375762611627579, "global_step": 201395, "epoch": 2262, "lr": 1.5606914972610602e-05}
{"train_loss": 0.08316970616579056, "global_step": 201396, "epoch": 2262, "lr": 1.5606494177877824e-05}
{"train_loss": 0.07432756572961807, "global_step": 201397, "epoch": 2262, "lr": 1.560607338776885e-05}
{"train_loss": 0.06013674661517143, "global_step": 201398, "epoch": 2262, "lr": 1.56056526022837e-05}
{"train_loss": 0.052807699888944626, "global_step": 201399, "epoch": 2262, "lr": 1.560523182142245e-05}
{"train_loss": 0.059404902160167694, "global_step": 201400, "epoch": 2262, "lr": 1.5604811045185176e-05}
{"train_loss": 0.04211759939789772, "global_step": 201401, "epoch": 2262, "lr": 1.56043902735719e-05}
{"train_loss": 0.04580308869481087, "global_step": 201402, "epoch": 2262, "lr": 1.560396950658271e-05}
{"train_loss": 0.07765712589025497, "global_step": 201403, "epoch": 2262, "lr": 1.5603548744217627e-05}
{"train_loss": 0.09942959994077682, "global_step": 201404, "epoch": 2262, "lr": 1.560312798647675e-05}
{"train_loss": 0.08588279038667679, "global_step": 201405, "epoch": 2262, "lr": 1.56027072333601e-05}
{"train_loss": 0.06566499996135074, "global_step": 201406, "epoch": 2262, "lr": 1.560228648486775e-05, "val_loss": 8.193678855895996}
{"train_loss": 0.07120296359062195, "global_step": 201407, "epoch": 2263, "lr": 1.5601865740999748e-05}
{"train_loss": 0.06254745274782181, "global_step": 201408, "epoch": 2263, "lr": 1.560144500175617e-05}
{"train_loss": 0.061889298260211945, "global_step": 201409, "epoch": 2263, "lr": 1.5601024267137042e-05}
{"train_loss": 0.026245063170790672, "global_step": 201410, "epoch": 2263, "lr": 1.560060353714245e-05}
{"train_loss": 0.04329736530780792, "global_step": 201411, "epoch": 2263, "lr": 1.5600182811772423e-05}
{"train_loss": 0.07073289901018143, "global_step": 201412, "epoch": 2263, "lr": 1.5599762091027055e-05}
{"train_loss": 0.07034918665885925, "global_step": 201413, "epoch": 2263, "lr": 1.559934137490635e-05}
{"train_loss": 0.04737980663776398, "global_step": 201414, "epoch": 2263, "lr": 1.559892066341042e-05}
{"train_loss": 0.03142599016427994, "global_step": 201415, "epoch": 2263, "lr": 1.559849995653928e-05}
{"train_loss": 0.032449495047330856, "global_step": 201416, "epoch": 2263, "lr": 1.5598079254293018e-05}
{"train_loss": 0.09468335658311844, "global_step": 201417, "epoch": 2263, "lr": 1.559765855667166e-05}
{"train_loss": 0.06295089423656464, "global_step": 201418, "epoch": 2263, "lr": 1.559723786367527e-05}
{"train_loss": 0.06370875984430313, "global_step": 201419, "epoch": 2263, "lr": 1.5596817175303934e-05}
{"train_loss": 0.07372775673866272, "global_step": 201420, "epoch": 2263, "lr": 1.5596396491557686e-05}
{"train_loss": 0.06500691920518875, "global_step": 201421, "epoch": 2263, "lr": 1.5595975812436564e-05}
{"train_loss": 0.06736080348491669, "global_step": 201422, "epoch": 2263, "lr": 1.5595555137940664e-05}
{"train_loss": 0.05872524902224541, "global_step": 201423, "epoch": 2263, "lr": 1.5595134468070006e-05}
{"train_loss": 0.055083077400922775, "global_step": 201424, "epoch": 2263, "lr": 1.5594713802824672e-05}
{"train_loss": 0.043299198150634766, "global_step": 201425, "epoch": 2263, "lr": 1.5594293142204718e-05}
{"train_loss": 0.10258379578590393, "global_step": 201426, "epoch": 2263, "lr": 1.559387248621017e-05}
{"train_loss": 0.08230097591876984, "global_step": 201427, "epoch": 2263, "lr": 1.5593451834841126e-05}
{"train_loss": 0.10624470561742783, "global_step": 201428, "epoch": 2263, "lr": 1.5593031188097608e-05}
{"train_loss": 0.0253225676715374, "global_step": 201429, "epoch": 2263, "lr": 1.5592610545979702e-05}
{"train_loss": 0.03226648271083832, "global_step": 201430, "epoch": 2263, "lr": 1.559218990848743e-05}
{"train_loss": 0.05471393093466759, "global_step": 201431, "epoch": 2263, "lr": 1.5591769275620893e-05}
{"train_loss": 0.08955258876085281, "global_step": 201432, "epoch": 2263, "lr": 1.55913486473801e-05}
{"train_loss": 0.07854624837636948, "global_step": 201433, "epoch": 2263, "lr": 1.559092802376515e-05}
{"train_loss": 0.10979820042848587, "global_step": 201434, "epoch": 2263, "lr": 1.5590507404776068e-05}
{"train_loss": 0.06611177325248718, "global_step": 201435, "epoch": 2263, "lr": 1.5590086790412918e-05}
{"train_loss": 0.04923104867339134, "global_step": 201436, "epoch": 2263, "lr": 1.5589666180675778e-05}
{"train_loss": 0.10654673725366592, "global_step": 201437, "epoch": 2263, "lr": 1.558924557556467e-05}
{"train_loss": 0.052608516067266464, "global_step": 201438, "epoch": 2263, "lr": 1.558882497507969e-05}
{"train_loss": 0.08181308954954147, "global_step": 201439, "epoch": 2263, "lr": 1.558840437922085e-05}
{"train_loss": 0.05186723917722702, "global_step": 201440, "epoch": 2263, "lr": 1.5587983787988253e-05}
{"train_loss": 0.06933218985795975, "global_step": 201441, "epoch": 2263, "lr": 1.5587563201381912e-05}
{"train_loss": 0.08118031173944473, "global_step": 201442, "epoch": 2263, "lr": 1.5587142619401916e-05}
{"train_loss": 0.08015986531972885, "global_step": 201443, "epoch": 2263, "lr": 1.5586722042048302e-05}
{"train_loss": 0.07229497283697128, "global_step": 201444, "epoch": 2263, "lr": 1.5586301469321145e-05}
{"train_loss": 0.056797176599502563, "global_step": 201445, "epoch": 2263, "lr": 1.5585880901220472e-05}
{"train_loss": 0.03866522014141083, "global_step": 201446, "epoch": 2263, "lr": 1.5585460337746384e-05}
{"train_loss": 0.04217708855867386, "global_step": 201447, "epoch": 2263, "lr": 1.5585039778898885e-05}
{"train_loss": 0.058146730065345764, "global_step": 201448, "epoch": 2263, "lr": 1.5584619224678076e-05}
{"train_loss": 0.03607963025569916, "global_step": 201449, "epoch": 2263, "lr": 1.5584198675083982e-05}
{"train_loss": 0.058489225804805756, "global_step": 201450, "epoch": 2263, "lr": 1.558377813011669e-05}
{"train_loss": 0.03475340083241463, "global_step": 201451, "epoch": 2263, "lr": 1.5583357589776217e-05}
{"train_loss": 0.08563143759965897, "global_step": 201452, "epoch": 2263, "lr": 1.558293705406266e-05}
{"train_loss": 0.06625596433877945, "global_step": 201453, "epoch": 2263, "lr": 1.5582516522976047e-05}
{"train_loss": 0.0648060142993927, "global_step": 201454, "epoch": 2263, "lr": 1.5582095996516438e-05}
{"train_loss": 0.04776247963309288, "global_step": 201455, "epoch": 2263, "lr": 1.558167547468392e-05}
{"train_loss": 0.06298934668302536, "global_step": 201456, "epoch": 2263, "lr": 1.558125495747851e-05}
{"train_loss": 0.09345006942749023, "global_step": 201457, "epoch": 2263, "lr": 1.558083444490029e-05}
{"train_loss": 0.06028260663151741, "global_step": 201458, "epoch": 2263, "lr": 1.5580413936949313e-05}
{"train_loss": 0.06620632112026215, "global_step": 201459, "epoch": 2263, "lr": 1.5579993433625613e-05}
{"train_loss": 0.07007217407226562, "global_step": 201460, "epoch": 2263, "lr": 1.5579572934929277e-05}
{"train_loss": 0.08250614255666733, "global_step": 201461, "epoch": 2263, "lr": 1.5579152440860327e-05}
{"train_loss": 0.02463398315012455, "global_step": 201462, "epoch": 2263, "lr": 1.5578731951418867e-05}
{"train_loss": 0.0988367423415184, "global_step": 201463, "epoch": 2263, "lr": 1.557831146660492e-05}
{"train_loss": 0.08703496307134628, "global_step": 201464, "epoch": 2263, "lr": 1.5577890986418536e-05}
{"train_loss": 0.05024738609790802, "global_step": 201465, "epoch": 2263, "lr": 1.5577470510859803e-05}
{"train_loss": 0.08432070910930634, "global_step": 201466, "epoch": 2263, "lr": 1.5577050039928744e-05}
{"train_loss": 0.05399782955646515, "global_step": 201467, "epoch": 2263, "lr": 1.557662957362544e-05}
{"train_loss": 0.09168751537799835, "global_step": 201468, "epoch": 2263, "lr": 1.557620911194993e-05}
{"train_loss": 0.06207551062107086, "global_step": 201469, "epoch": 2263, "lr": 1.557578865490229e-05}
{"train_loss": 0.039862073957920074, "global_step": 201470, "epoch": 2263, "lr": 1.5575368202482554e-05}
{"train_loss": 0.07068391144275665, "global_step": 201471, "epoch": 2263, "lr": 1.557494775469079e-05}
{"train_loss": 0.10885000228881836, "global_step": 201472, "epoch": 2263, "lr": 1.5574527311527065e-05}
{"train_loss": 0.06386309862136841, "global_step": 201473, "epoch": 2263, "lr": 1.557410687299141e-05}
{"train_loss": 0.043069250881671906, "global_step": 201474, "epoch": 2263, "lr": 1.5573686439083917e-05}
{"train_loss": 0.026762358844280243, "global_step": 201475, "epoch": 2263, "lr": 1.5573266009804605e-05}
{"train_loss": 0.03324638307094574, "global_step": 201476, "epoch": 2263, "lr": 1.5572845585153566e-05}
{"train_loss": 0.06989887356758118, "global_step": 201477, "epoch": 2263, "lr": 1.5572425165130817e-05}
{"train_loss": 0.0722823515534401, "global_step": 201478, "epoch": 2263, "lr": 1.5572004749736458e-05}
{"train_loss": 0.07181242108345032, "global_step": 201479, "epoch": 2263, "lr": 1.55715843389705e-05}
{"train_loss": 0.09993094950914383, "global_step": 201480, "epoch": 2263, "lr": 1.5571163932833044e-05}
{"train_loss": 0.04185252636671066, "global_step": 201481, "epoch": 2263, "lr": 1.557074353132411e-05}
{"train_loss": 0.03482356294989586, "global_step": 201482, "epoch": 2263, "lr": 1.5570323134443782e-05}
{"train_loss": 0.09600823372602463, "global_step": 201483, "epoch": 2263, "lr": 1.5569902742192095e-05}
{"train_loss": 0.06573909521102905, "global_step": 201484, "epoch": 2263, "lr": 1.5569482354569125e-05}
{"train_loss": 0.047597337514162064, "global_step": 201485, "epoch": 2263, "lr": 1.5569061971574905e-05}
{"train_loss": 0.07686452567577362, "global_step": 201486, "epoch": 2263, "lr": 1.556864159320952e-05}
{"train_loss": 0.06074785813689232, "global_step": 201487, "epoch": 2263, "lr": 1.5568221219472994e-05}
{"train_loss": 0.07014846056699753, "global_step": 201488, "epoch": 2263, "lr": 1.5567800850365416e-05}
{"train_loss": 0.09423169493675232, "global_step": 201489, "epoch": 2263, "lr": 1.5567380485886818e-05}
{"train_loss": 0.05860075354576111, "global_step": 201490, "epoch": 2263, "lr": 1.556696012603726e-05}
{"train_loss": 0.05629044026136398, "global_step": 201491, "epoch": 2263, "lr": 1.5566539770816825e-05}
{"train_loss": 0.0702444463968277, "global_step": 201492, "epoch": 2263, "lr": 1.556611942022553e-05}
{"train_loss": 0.03400615602731705, "global_step": 201493, "epoch": 2263, "lr": 1.5565699074263468e-05}
{"train_loss": 0.05205661430954933, "global_step": 201494, "epoch": 2263, "lr": 1.556527873293066e-05}
{"train_loss": 0.06397756239336529, "global_step": 201495, "epoch": 2263, "lr": 1.55648583962272e-05, "val_loss": 8.333248138427734}
{"train_loss": 0.10243947803974152, "global_step": 201496, "epoch": 2264, "lr": 1.5564438064153103e-05}
{"train_loss": 0.042316731065511703, "global_step": 201497, "epoch": 2264, "lr": 1.556401773670847e-05}
{"train_loss": 0.025776585564017296, "global_step": 201498, "epoch": 2264, "lr": 1.5563597413893332e-05}
{"train_loss": 0.09357836842536926, "global_step": 201499, "epoch": 2264, "lr": 1.556317709570773e-05}
{"train_loss": 0.10371458530426025, "global_step": 201500, "epoch": 2264, "lr": 1.556275678215176e-05}
{"train_loss": 0.08097229897975922, "global_step": 201501, "epoch": 2264, "lr": 1.5562336473225443e-05}
{"train_loss": 0.015659986063838005, "global_step": 201502, "epoch": 2264, "lr": 1.5561916168928857e-05}
{"train_loss": 0.043500859290361404, "global_step": 201503, "epoch": 2264, "lr": 1.556149586926206e-05}
{"train_loss": 0.028448306024074554, "global_step": 201504, "epoch": 2264, "lr": 1.5561075574225083e-05}
{"train_loss": 0.05080033466219902, "global_step": 201505, "epoch": 2264, "lr": 1.5560655283818015e-05}
{"train_loss": 0.06191505864262581, "global_step": 201506, "epoch": 2264, "lr": 1.556023499804088e-05}
{"train_loss": 0.061544016003608704, "global_step": 201507, "epoch": 2264, "lr": 1.5559814716893766e-05}
{"train_loss": 0.07996673882007599, "global_step": 201508, "epoch": 2264, "lr": 1.55593944403767e-05}
{"train_loss": 0.05459825322031975, "global_step": 201509, "epoch": 2264, "lr": 1.5558974168489758e-05}
{"train_loss": 0.07078959047794342, "global_step": 201510, "epoch": 2264, "lr": 1.555855390123301e-05}
{"train_loss": 0.06648288667201996, "global_step": 201511, "epoch": 2264, "lr": 1.555813363860647e-05}
{"train_loss": 0.08928292244672775, "global_step": 201512, "epoch": 2264, "lr": 1.555771338061024e-05}
{"train_loss": 0.054817307740449905, "global_step": 201513, "epoch": 2264, "lr": 1.5557293127244337e-05}
{"train_loss": 0.06249673664569855, "global_step": 201514, "epoch": 2264, "lr": 1.5556872878508854e-05}
{"train_loss": 0.09591900557279587, "global_step": 201515, "epoch": 2264, "lr": 1.5556452634403813e-05}
{"train_loss": 0.10133659839630127, "global_step": 201516, "epoch": 2264, "lr": 1.5556032394929304e-05}
{"train_loss": 0.056675054132938385, "global_step": 201517, "epoch": 2264, "lr": 1.5555612160085348e-05}
{"train_loss": 0.05559365823864937, "global_step": 201518, "epoch": 2264, "lr": 1.5555191929872043e-05}
{"train_loss": 0.08229623734951019, "global_step": 201519, "epoch": 2264, "lr": 1.55547717042894e-05}
{"train_loss": 0.05761627107858658, "global_step": 201520, "epoch": 2264, "lr": 1.5554351483337515e-05}
{"train_loss": 0.0718429833650589, "global_step": 201521, "epoch": 2264, "lr": 1.5553931267016415e-05}
{"train_loss": 0.07285264134407043, "global_step": 201522, "epoch": 2264, "lr": 1.5553511055326182e-05}
{"train_loss": 0.09005914628505707, "global_step": 201523, "epoch": 2264, "lr": 1.5553090848266845e-05}
{"train_loss": 0.09981058537960052, "global_step": 201524, "epoch": 2264, "lr": 1.555267064583849e-05}
{"train_loss": 0.08633125573396683, "global_step": 201525, "epoch": 2264, "lr": 1.555225044804114e-05}
{"train_loss": 0.07415411621332169, "global_step": 201526, "epoch": 2264, "lr": 1.555183025487487e-05}
{"train_loss": 0.06448134034872055, "global_step": 201527, "epoch": 2264, "lr": 1.555141006633976e-05}
{"train_loss": 0.04527992755174637, "global_step": 201528, "epoch": 2264, "lr": 1.5550989882435825e-05}
{"train_loss": 0.0692998543381691, "global_step": 201529, "epoch": 2264, "lr": 1.555056970316315e-05}
{"train_loss": 0.04772833362221718, "global_step": 201530, "epoch": 2264, "lr": 1.555014952852177e-05}
{"train_loss": 0.0965486541390419, "global_step": 201531, "epoch": 2264, "lr": 1.5549729358511768e-05}
{"train_loss": 0.03553274646401405, "global_step": 201532, "epoch": 2264, "lr": 1.5549309193133166e-05}
{"train_loss": 0.04061935469508171, "global_step": 201533, "epoch": 2264, "lr": 1.5548889032386053e-05}
{"train_loss": 0.10377272218465805, "global_step": 201534, "epoch": 2264, "lr": 1.5548468876270462e-05}
{"train_loss": 0.06988000124692917, "global_step": 201535, "epoch": 2264, "lr": 1.5548048724786475e-05}
{"train_loss": 0.09915302693843842, "global_step": 201536, "epoch": 2264, "lr": 1.554762857793413e-05}
{"train_loss": 0.08861980587244034, "global_step": 201537, "epoch": 2264, "lr": 1.554720843571347e-05}
{"train_loss": 0.0749586820602417, "global_step": 201538, "epoch": 2264, "lr": 1.5546788298124588e-05}
{"train_loss": 0.055463891476392746, "global_step": 201539, "epoch": 2264, "lr": 1.55463681651675e-05}
{"train_loss": 0.06379759311676025, "global_step": 201540, "epoch": 2264, "lr": 1.5545948036842306e-05}
{"train_loss": 0.023721592500805855, "global_step": 201541, "epoch": 2264, "lr": 1.554552791314903e-05}
{"train_loss": 0.02708120457828045, "global_step": 201542, "epoch": 2264, "lr": 1.5545107794087734e-05}
{"train_loss": 0.047018807381391525, "global_step": 201543, "epoch": 2264, "lr": 1.5544687679658483e-05}
{"train_loss": 0.10098040848970413, "global_step": 201544, "epoch": 2264, "lr": 1.5544267569861316e-05}
{"train_loss": 0.02718932181596756, "global_step": 201545, "epoch": 2264, "lr": 1.5543847464696306e-05}
{"train_loss": 0.05249428376555443, "global_step": 201546, "epoch": 2264, "lr": 1.5543427364163514e-05}
{"train_loss": 0.05881430208683014, "global_step": 201547, "epoch": 2264, "lr": 1.5543007268262977e-05}
{"train_loss": 0.019520875066518784, "global_step": 201548, "epoch": 2264, "lr": 1.554258717699478e-05}
{"train_loss": 0.08241740614175797, "global_step": 201549, "epoch": 2264, "lr": 1.5542167090358945e-05}
{"train_loss": 0.13222305476665497, "global_step": 201550, "epoch": 2264, "lr": 1.5541747008355562e-05}
{"train_loss": 0.05694058910012245, "global_step": 201551, "epoch": 2264, "lr": 1.554132693098465e-05}
{"train_loss": 0.04620033875107765, "global_step": 201552, "epoch": 2264, "lr": 1.5540906858246307e-05}
{"train_loss": 0.05158524960279465, "global_step": 201553, "epoch": 2264, "lr": 1.5540486790140556e-05}
{"train_loss": 0.05939212441444397, "global_step": 201554, "epoch": 2264, "lr": 1.554006672666748e-05}
{"train_loss": 0.06986103951931, "global_step": 201555, "epoch": 2264, "lr": 1.5539646667827107e-05}
{"train_loss": 0.059230223298072815, "global_step": 201556, "epoch": 2264, "lr": 1.5539226613619523e-05}
{"train_loss": 0.07318500429391861, "global_step": 201557, "epoch": 2264, "lr": 1.553880656404475e-05}
{"train_loss": 0.0481884628534317, "global_step": 201558, "epoch": 2264, "lr": 1.5538386519102887e-05}
{"train_loss": 0.11418499797582626, "global_step": 201559, "epoch": 2264, "lr": 1.553796647879395e-05}
{"train_loss": 0.04662024974822998, "global_step": 201560, "epoch": 2264, "lr": 1.553754644311803e-05}
{"train_loss": 0.023634672164916992, "global_step": 201561, "epoch": 2264, "lr": 1.5537126412075147e-05}
{"train_loss": 0.056381773203611374, "global_step": 201562, "epoch": 2264, "lr": 1.553670638566538e-05}
{"train_loss": 0.04220116138458252, "global_step": 201563, "epoch": 2264, "lr": 1.5536286363888797e-05}
{"train_loss": 0.05272391811013222, "global_step": 201564, "epoch": 2264, "lr": 1.5535866346745427e-05}
{"train_loss": 0.06651367247104645, "global_step": 201565, "epoch": 2264, "lr": 1.5535446334235355e-05}
{"train_loss": 0.058878201991319656, "global_step": 201566, "epoch": 2264, "lr": 1.5535026326358604e-05}
{"train_loss": 0.08512595295906067, "global_step": 201567, "epoch": 2264, "lr": 1.5534606323115268e-05}
{"train_loss": 0.03398493677377701, "global_step": 201568, "epoch": 2264, "lr": 1.553418632450537e-05}
{"train_loss": 0.0462040975689888, "global_step": 201569, "epoch": 2264, "lr": 1.5533766330528987e-05}
{"train_loss": 0.09880564361810684, "global_step": 201570, "epoch": 2264, "lr": 1.553334634118616e-05}
{"train_loss": 0.08329478651285172, "global_step": 201571, "epoch": 2264, "lr": 1.5532926356476973e-05}
{"train_loss": 0.06948684900999069, "global_step": 201572, "epoch": 2264, "lr": 1.5532506376401446e-05}
{"train_loss": 0.06857684999704361, "global_step": 201573, "epoch": 2264, "lr": 1.5532086400959666e-05}
{"train_loss": 0.04426141455769539, "global_step": 201574, "epoch": 2264, "lr": 1.5531666430151675e-05}
{"train_loss": 0.10572004318237305, "global_step": 201575, "epoch": 2264, "lr": 1.5531246463977524e-05}
{"train_loss": 0.04239695891737938, "global_step": 201576, "epoch": 2264, "lr": 1.5530826502437286e-05}
{"train_loss": 0.05052303150296211, "global_step": 201577, "epoch": 2264, "lr": 1.5530406545530996e-05}
{"train_loss": 0.04518994688987732, "global_step": 201578, "epoch": 2264, "lr": 1.5529986593258732e-05}
{"train_loss": 0.0716271623969078, "global_step": 201579, "epoch": 2264, "lr": 1.5529566645620552e-05}
{"train_loss": 0.061343707144260406, "global_step": 201580, "epoch": 2264, "lr": 1.5529146702616475e-05}
{"train_loss": 0.1034855768084526, "global_step": 201581, "epoch": 2264, "lr": 1.5528726764246593e-05}
{"train_loss": 0.06687309592962265, "global_step": 201582, "epoch": 2264, "lr": 1.5528306830510965e-05}
{"train_loss": 0.07988132536411285, "global_step": 201583, "epoch": 2264, "lr": 1.5527886901409622e-05}
{"train_loss": 0.06504132602824254, "global_step": 201584, "epoch": 2264, "lr": 1.5527466976942645e-05, "val_loss": 8.217530250549316}
{"train_loss": 0.041263189166784286, "global_step": 201585, "epoch": 2265, "lr": 1.5527047057110066e-05}
{"train_loss": 0.05810464173555374, "global_step": 201586, "epoch": 2265, "lr": 1.5526627141911976e-05}
{"train_loss": 0.07512938976287842, "global_step": 201587, "epoch": 2265, "lr": 1.5526207231348394e-05}
{"train_loss": 0.07047213613986969, "global_step": 201588, "epoch": 2265, "lr": 1.5525787325419407e-05}
{"train_loss": 0.10107219964265823, "global_step": 201589, "epoch": 2265, "lr": 1.552536742412504e-05}
{"train_loss": 0.05860840901732445, "global_step": 201590, "epoch": 2265, "lr": 1.5524947527465383e-05}
{"train_loss": 0.05671054869890213, "global_step": 201591, "epoch": 2265, "lr": 1.5524527635440465e-05}
{"train_loss": 0.07585233449935913, "global_step": 201592, "epoch": 2265, "lr": 1.5524107748050366e-05}
{"train_loss": 0.07671435922384262, "global_step": 201593, "epoch": 2265, "lr": 1.5523687865295116e-05}
{"train_loss": 0.04259997978806496, "global_step": 201594, "epoch": 2265, "lr": 1.5523267987174807e-05}
{"train_loss": 0.06104324012994766, "global_step": 201595, "epoch": 2265, "lr": 1.5522848113689455e-05}
{"train_loss": 0.06641235202550888, "global_step": 201596, "epoch": 2265, "lr": 1.5522428244839148e-05}
{"train_loss": 0.06504078954458237, "global_step": 201597, "epoch": 2265, "lr": 1.5522008380623914e-05}
{"train_loss": 0.04394666105508804, "global_step": 201598, "epoch": 2265, "lr": 1.5521588521043828e-05}
{"train_loss": 0.08863162249326706, "global_step": 201599, "epoch": 2265, "lr": 1.5521168666098963e-05}
{"train_loss": 0.051348764449357986, "global_step": 201600, "epoch": 2265, "lr": 1.5520748815789337e-05}
{"train_loss": 0.10118023306131363, "global_step": 201601, "epoch": 2265, "lr": 1.5520328970115045e-05}
{"train_loss": 0.039717964828014374, "global_step": 201602, "epoch": 2265, "lr": 1.5519909129076103e-05}
{"train_loss": 0.03621461242437363, "global_step": 201603, "epoch": 2265, "lr": 1.5519489292672614e-05}
{"train_loss": 0.09275483340024948, "global_step": 201604, "epoch": 2265, "lr": 1.5519069460904583e-05}
{"train_loss": 0.03513212502002716, "global_step": 201605, "epoch": 2265, "lr": 1.551864963377212e-05}
{"train_loss": 0.01842588558793068, "global_step": 201606, "epoch": 2265, "lr": 1.551822981127523e-05}
{"train_loss": 0.036440517753362656, "global_step": 201607, "epoch": 2265, "lr": 1.5517809993414014e-05}
{"train_loss": 0.038862694054841995, "global_step": 201608, "epoch": 2265, "lr": 1.551739018018849e-05}
{"train_loss": 0.03324548900127411, "global_step": 201609, "epoch": 2265, "lr": 1.551697037159875e-05}
{"train_loss": 0.03691495209932327, "global_step": 201610, "epoch": 2265, "lr": 1.551655056764481e-05}
{"train_loss": 0.10489016026258469, "global_step": 201611, "epoch": 2265, "lr": 1.5516130768326775e-05}
{"train_loss": 0.04599319025874138, "global_step": 201612, "epoch": 2265, "lr": 1.5515710973644652e-05}
{"train_loss": 0.06450431793928146, "global_step": 201613, "epoch": 2265, "lr": 1.5515291183598548e-05}
{"train_loss": 0.04705966264009476, "global_step": 201614, "epoch": 2265, "lr": 1.5514871398188484e-05}
{"train_loss": 0.0748153105378151, "global_step": 201615, "epoch": 2265, "lr": 1.551445161741451e-05}
{"train_loss": 0.053843848407268524, "global_step": 201616, "epoch": 2265, "lr": 1.5514031841276717e-05}
{"train_loss": 0.039769578725099564, "global_step": 201617, "epoch": 2265, "lr": 1.5513612069775125e-05}
{"train_loss": 0.06699144095182419, "global_step": 201618, "epoch": 2265, "lr": 1.5513192302909825e-05}
{"train_loss": 0.03669823706150055, "global_step": 201619, "epoch": 2265, "lr": 1.5512772540680837e-05}
{"train_loss": 0.08850088715553284, "global_step": 201620, "epoch": 2265, "lr": 1.551235278308825e-05}
{"train_loss": 0.05337835103273392, "global_step": 201621, "epoch": 2265, "lr": 1.5511933030132096e-05}
{"train_loss": 0.03474130854010582, "global_step": 201622, "epoch": 2265, "lr": 1.5511513281812457e-05}
{"train_loss": 0.05150236934423447, "global_step": 201623, "epoch": 2265, "lr": 1.5511093538129358e-05}
{"train_loss": 0.0528719387948513, "global_step": 201624, "epoch": 2265, "lr": 1.5510673799082886e-05}
{"train_loss": 0.04653047025203705, "global_step": 201625, "epoch": 2265, "lr": 1.5510254064673073e-05}
{"train_loss": 0.12517520785331726, "global_step": 201626, "epoch": 2265, "lr": 1.5509834334899997e-05}
{"train_loss": 0.07624942064285278, "global_step": 201627, "epoch": 2265, "lr": 1.550941460976369e-05}
{"train_loss": 0.08216556906700134, "global_step": 201628, "epoch": 2265, "lr": 1.550899488926424e-05}
{"train_loss": 0.04863325506448746, "global_step": 201629, "epoch": 2265, "lr": 1.5508575173401662e-05}
{"train_loss": 0.08412599563598633, "global_step": 201630, "epoch": 2265, "lr": 1.5508155462176062e-05}
{"train_loss": 0.1008777767419815, "global_step": 201631, "epoch": 2265, "lr": 1.5507735755587443e-05}
{"train_loss": 0.05571487173438072, "global_step": 201632, "epoch": 2265, "lr": 1.5507316053635912e-05}
{"train_loss": 0.07606042176485062, "global_step": 201633, "epoch": 2265, "lr": 1.5506896356321483e-05}
{"train_loss": 0.053647011518478394, "global_step": 201634, "epoch": 2265, "lr": 1.550647666364425e-05}
{"train_loss": 0.08182336390018463, "global_step": 201635, "epoch": 2265, "lr": 1.5506056975604228e-05}
{"train_loss": 0.10773572325706482, "global_step": 201636, "epoch": 2265, "lr": 1.5505637292201502e-05}
{"train_loss": 0.07073062658309937, "global_step": 201637, "epoch": 2265, "lr": 1.550521761343614e-05}
{"train_loss": 0.06545312702655792, "global_step": 201638, "epoch": 2265, "lr": 1.5504797939308164e-05}
{"train_loss": 0.03919697925448418, "global_step": 201639, "epoch": 2265, "lr": 1.550437826981766e-05}
{"train_loss": 0.04726685211062431, "global_step": 201640, "epoch": 2265, "lr": 1.5503958604964657e-05}
{"train_loss": 0.043969009071588516, "global_step": 201641, "epoch": 2265, "lr": 1.5503538944749242e-05}
{"train_loss": 0.06678929924964905, "global_step": 201642, "epoch": 2265, "lr": 1.550311928917144e-05}
{"train_loss": 0.07417413592338562, "global_step": 201643, "epoch": 2265, "lr": 1.5502699638231344e-05}
{"train_loss": 0.09333128482103348, "global_step": 201644, "epoch": 2265, "lr": 1.550227999192897e-05}
{"train_loss": 0.07563798129558563, "global_step": 201645, "epoch": 2265, "lr": 1.550186035026441e-05}
{"train_loss": 0.06530478596687317, "global_step": 201646, "epoch": 2265, "lr": 1.5501440713237686e-05}
{"train_loss": 0.09112419933080673, "global_step": 201647, "epoch": 2265, "lr": 1.5501021080848892e-05}
{"train_loss": 0.09438513219356537, "global_step": 201648, "epoch": 2265, "lr": 1.5500601453098047e-05}
{"train_loss": 0.09641112387180328, "global_step": 201649, "epoch": 2265, "lr": 1.5500181829985243e-05}
{"train_loss": 0.05529329553246498, "global_step": 201650, "epoch": 2265, "lr": 1.5499762211510503e-05}
{"train_loss": 0.08614140748977661, "global_step": 201651, "epoch": 2265, "lr": 1.5499342597673916e-05}
{"train_loss": 0.08502545952796936, "global_step": 201652, "epoch": 2265, "lr": 1.549892298847552e-05}
{"train_loss": 0.05806838348507881, "global_step": 201653, "epoch": 2265, "lr": 1.5498503383915358e-05}
{"train_loss": 0.07220148295164108, "global_step": 201654, "epoch": 2265, "lr": 1.5498083783993523e-05}
{"train_loss": 0.03741404786705971, "global_step": 201655, "epoch": 2265, "lr": 1.5497664188710027e-05}
{"train_loss": 0.051097970455884933, "global_step": 201656, "epoch": 2265, "lr": 1.5497244598064968e-05}
{"train_loss": 0.0845702588558197, "global_step": 201657, "epoch": 2265, "lr": 1.549682501205837e-05}
{"train_loss": 0.045821141451597214, "global_step": 201658, "epoch": 2265, "lr": 1.5496405430690314e-05}
{"train_loss": 0.09017787873744965, "global_step": 201659, "epoch": 2265, "lr": 1.549598585396083e-05}
{"train_loss": 0.04743991792201996, "global_step": 201660, "epoch": 2265, "lr": 1.549556628187001e-05}
{"train_loss": 0.06926032900810242, "global_step": 201661, "epoch": 2265, "lr": 1.5495146714417873e-05}
{"train_loss": 0.03898519650101662, "global_step": 201662, "epoch": 2265, "lr": 1.549472715160451e-05}
{"train_loss": 0.0799410343170166, "global_step": 201663, "epoch": 2265, "lr": 1.5494307593429945e-05}
{"train_loss": 0.03509782999753952, "global_step": 201664, "epoch": 2265, "lr": 1.5493888039894267e-05}
{"train_loss": 0.0763823464512825, "global_step": 201665, "epoch": 2265, "lr": 1.5493468490997497e-05}
{"train_loss": 0.032167382538318634, "global_step": 201666, "epoch": 2265, "lr": 1.5493048946739723e-05}
{"train_loss": 0.06970090419054031, "global_step": 201667, "epoch": 2265, "lr": 1.5492629407120973e-05}
{"train_loss": 0.05463288724422455, "global_step": 201668, "epoch": 2265, "lr": 1.549220987214134e-05}
{"train_loss": 0.02862713672220707, "global_step": 201669, "epoch": 2265, "lr": 1.549179034180084e-05}
{"train_loss": 0.08440352231264114, "global_step": 201670, "epoch": 2265, "lr": 1.5491370816099558e-05}
{"train_loss": 0.04460056126117706, "global_step": 201671, "epoch": 2265, "lr": 1.549095129503753e-05}
{"train_loss": 0.09992088377475739, "global_step": 201672, "epoch": 2265, "lr": 1.5490531778614825e-05}
{"train_loss": 0.06374666318715957, "global_step": 201673, "epoch": 2265, "lr": 1.5490112266831508e-05, "val_loss": 8.188508033752441, "train_action_mse_error": 21.958337783813477}
{"train_loss": 0.05574795976281166, "global_step": 201674, "epoch": 2266, "lr": 1.5489692759687614e-05}
{"train_loss": 0.09205681085586548, "global_step": 201675, "epoch": 2266, "lr": 1.548927325718322e-05}
{"train_loss": 0.03733712062239647, "global_step": 201676, "epoch": 2266, "lr": 1.5488853759318367e-05}
{"train_loss": 0.09061011672019958, "global_step": 201677, "epoch": 2266, "lr": 1.5488434266093122e-05}
{"train_loss": 0.08136893808841705, "global_step": 201678, "epoch": 2266, "lr": 1.5488014777507527e-05}
{"train_loss": 0.06900499761104584, "global_step": 201679, "epoch": 2266, "lr": 1.5487595293561664e-05}
{"train_loss": 0.05411459878087044, "global_step": 201680, "epoch": 2266, "lr": 1.548717581425555e-05}
{"train_loss": 0.07050031423568726, "global_step": 201681, "epoch": 2266, "lr": 1.5486756339589287e-05}
{"train_loss": 0.054785098880529404, "global_step": 201682, "epoch": 2266, "lr": 1.5486336869562888e-05}
{"train_loss": 0.04628633335232735, "global_step": 201683, "epoch": 2266, "lr": 1.548591740417645e-05}
{"train_loss": 0.05458137392997742, "global_step": 201684, "epoch": 2266, "lr": 1.5485497943429994e-05}
{"train_loss": 0.08246495574712753, "global_step": 201685, "epoch": 2266, "lr": 1.5485078487323613e-05}
{"train_loss": 0.07151994109153748, "global_step": 201686, "epoch": 2266, "lr": 1.548465903585732e-05}
{"train_loss": 0.05609540268778801, "global_step": 201687, "epoch": 2266, "lr": 1.5484239589031208e-05}
{"train_loss": 0.053369686007499695, "global_step": 201688, "epoch": 2266, "lr": 1.5483820146845308e-05}
{"train_loss": 0.06397654861211777, "global_step": 201689, "epoch": 2266, "lr": 1.54834007092997e-05}
{"train_loss": 0.053411874920129776, "global_step": 201690, "epoch": 2266, "lr": 1.5482981276394432e-05}
{"train_loss": 0.07650880515575409, "global_step": 201691, "epoch": 2266, "lr": 1.5482561848129533e-05}
{"train_loss": 0.044053997844457626, "global_step": 201692, "epoch": 2266, "lr": 1.5482142424505107e-05}
{"train_loss": 0.06063801050186157, "global_step": 201693, "epoch": 2266, "lr": 1.548172300552117e-05}
{"train_loss": 0.025464162230491638, "global_step": 201694, "epoch": 2266, "lr": 1.5481303591177808e-05}
{"train_loss": 0.04714084789156914, "global_step": 201695, "epoch": 2266, "lr": 1.5480884181475046e-05}
{"train_loss": 0.11056018620729446, "global_step": 201696, "epoch": 2266, "lr": 1.5480464776412978e-05}
{"train_loss": 0.12161502242088318, "global_step": 201697, "epoch": 2266, "lr": 1.5480045375991624e-05}
{"train_loss": 0.0884750708937645, "global_step": 201698, "epoch": 2266, "lr": 1.5479625980211072e-05}
{"train_loss": 0.08921653032302856, "global_step": 201699, "epoch": 2266, "lr": 1.547920658907135e-05}
{"train_loss": 0.047776080667972565, "global_step": 201700, "epoch": 2266, "lr": 1.5478787202572537e-05}
{"train_loss": 0.06637608259916306, "global_step": 201701, "epoch": 2266, "lr": 1.5478367820714672e-05}
{"train_loss": 0.09739790111780167, "global_step": 201702, "epoch": 2266, "lr": 1.547794844349783e-05}
{"train_loss": 0.09502425044775009, "global_step": 201703, "epoch": 2266, "lr": 1.5477529070922046e-05}
{"train_loss": 0.022149840369820595, "global_step": 201704, "epoch": 2266, "lr": 1.54771097029874e-05}
{"train_loss": 0.10323745012283325, "global_step": 201705, "epoch": 2266, "lr": 1.5476690339693923e-05}
{"train_loss": 0.06587180495262146, "global_step": 201706, "epoch": 2266, "lr": 1.5476270981041695e-05}
{"train_loss": 0.011270062066614628, "global_step": 201707, "epoch": 2266, "lr": 1.5475851627030754e-05}
{"train_loss": 0.10560456663370132, "global_step": 201708, "epoch": 2266, "lr": 1.5475432277661156e-05}
{"train_loss": 0.057902514934539795, "global_step": 201709, "epoch": 2266, "lr": 1.5475012932932987e-05}
{"train_loss": 0.12388011813163757, "global_step": 201710, "epoch": 2266, "lr": 1.547459359284626e-05}
{"train_loss": 0.07525146752595901, "global_step": 201711, "epoch": 2266, "lr": 1.5474174257401075e-05}
{"train_loss": 0.04452970251441002, "global_step": 201712, "epoch": 2266, "lr": 1.5473754926597444e-05}
{"train_loss": 0.0937257781624794, "global_step": 201713, "epoch": 2266, "lr": 1.5473335600435468e-05}
{"train_loss": 0.06188944727182388, "global_step": 201714, "epoch": 2266, "lr": 1.5472916278915162e-05}
{"train_loss": 0.026695072650909424, "global_step": 201715, "epoch": 2266, "lr": 1.547249696203662e-05}
{"train_loss": 0.07130017131567001, "global_step": 201716, "epoch": 2266, "lr": 1.547207764979986e-05}
{"train_loss": 0.09969282895326614, "global_step": 201717, "epoch": 2266, "lr": 1.5471658342204975e-05}
{"train_loss": 0.05501711741089821, "global_step": 201718, "epoch": 2266, "lr": 1.5471239039251993e-05}
{"train_loss": 0.12781085073947906, "global_step": 201719, "epoch": 2266, "lr": 1.5470819740940996e-05}
{"train_loss": 0.06180708855390549, "global_step": 201720, "epoch": 2266, "lr": 1.547040044727201e-05}
{"train_loss": 0.05053604394197464, "global_step": 201721, "epoch": 2266, "lr": 1.5469981158245122e-05}
{"train_loss": 0.06604636460542679, "global_step": 201722, "epoch": 2266, "lr": 1.546956187386036e-05}
{"train_loss": 0.05438913032412529, "global_step": 201723, "epoch": 2266, "lr": 1.546914259411781e-05}
{"train_loss": 0.08128869533538818, "global_step": 201724, "epoch": 2266, "lr": 1.5468723319017497e-05}
{"train_loss": 0.05761658027768135, "global_step": 201725, "epoch": 2266, "lr": 1.5468304048559508e-05}
{"train_loss": 0.058004386723041534, "global_step": 201726, "epoch": 2266, "lr": 1.546788478274387e-05}
{"train_loss": 0.05967045947909355, "global_step": 201727, "epoch": 2266, "lr": 1.5467465521570673e-05}
{"train_loss": 0.04537776857614517, "global_step": 201728, "epoch": 2266, "lr": 1.5467046265039935e-05}
{"train_loss": 0.026510372757911682, "global_step": 201729, "epoch": 2266, "lr": 1.5466627013151748e-05}
{"train_loss": 0.05900450050830841, "global_step": 201730, "epoch": 2266, "lr": 1.5466207765906148e-05}
{"train_loss": 0.07152877002954483, "global_step": 201731, "epoch": 2266, "lr": 1.5465788523303186e-05}
{"train_loss": 0.042357366532087326, "global_step": 201732, "epoch": 2266, "lr": 1.546536928534294e-05}
{"train_loss": 0.05156330391764641, "global_step": 201733, "epoch": 2266, "lr": 1.546495005202544e-05}
{"train_loss": 0.03793898969888687, "global_step": 201734, "epoch": 2266, "lr": 1.5464530823350776e-05}
{"train_loss": 0.05036338046193123, "global_step": 201735, "epoch": 2266, "lr": 1.5464111599318965e-05}
{"train_loss": 0.11105619370937347, "global_step": 201736, "epoch": 2266, "lr": 1.5463692379930093e-05}
{"train_loss": 0.06109430268406868, "global_step": 201737, "epoch": 2266, "lr": 1.54632731651842e-05}
{"train_loss": 0.025824151933193207, "global_step": 201738, "epoch": 2266, "lr": 1.5462853955081363e-05}
{"train_loss": 0.059998415410518646, "global_step": 201739, "epoch": 2266, "lr": 1.5462434749621606e-05}
{"train_loss": 0.06458266079425812, "global_step": 201740, "epoch": 2266, "lr": 1.5462015548805026e-05}
{"train_loss": 0.06384698301553726, "global_step": 201741, "epoch": 2266, "lr": 1.5461596352631634e-05}
{"train_loss": 0.07034312188625336, "global_step": 201742, "epoch": 2266, "lr": 1.5461177161101527e-05}
{"train_loss": 0.08787631243467331, "global_step": 201743, "epoch": 2266, "lr": 1.546075797421473e-05}
{"train_loss": 0.06335621327161789, "global_step": 201744, "epoch": 2266, "lr": 1.546033879197131e-05}
{"train_loss": 0.05087558925151825, "global_step": 201745, "epoch": 2266, "lr": 1.5459919614371343e-05}
{"train_loss": 0.06131545826792717, "global_step": 201746, "epoch": 2266, "lr": 1.5459500441414852e-05}
{"train_loss": 0.06531505286693573, "global_step": 201747, "epoch": 2266, "lr": 1.545908127310193e-05}
{"train_loss": 0.04511447995901108, "global_step": 201748, "epoch": 2266, "lr": 1.5458662109432592e-05}
{"train_loss": 0.06046314164996147, "global_step": 201749, "epoch": 2266, "lr": 1.5458242950406937e-05}
{"train_loss": 0.06649735569953918, "global_step": 201750, "epoch": 2266, "lr": 1.5457823796024985e-05}
{"train_loss": 0.04102208465337753, "global_step": 201751, "epoch": 2266, "lr": 1.5457404646286823e-05}
{"train_loss": 0.099918894469738, "global_step": 201752, "epoch": 2266, "lr": 1.545698550119247e-05}
{"train_loss": 0.03582512587308884, "global_step": 201753, "epoch": 2266, "lr": 1.5456566360742026e-05}
{"train_loss": 0.04740376025438309, "global_step": 201754, "epoch": 2266, "lr": 1.545614722493551e-05}
{"train_loss": 0.07046200335025787, "global_step": 201755, "epoch": 2266, "lr": 1.5455728093773003e-05}
{"train_loss": 0.06510935723781586, "global_step": 201756, "epoch": 2266, "lr": 1.5455308967254545e-05}
{"train_loss": 0.06633161753416061, "global_step": 201757, "epoch": 2266, "lr": 1.545488984538021e-05}
{"train_loss": 0.11450424045324326, "global_step": 201758, "epoch": 2266, "lr": 1.5454470728150027e-05}
{"train_loss": 0.055347081273794174, "global_step": 201759, "epoch": 2266, "lr": 1.5454051615564092e-05}
{"train_loss": 0.047672662883996964, "global_step": 201760, "epoch": 2266, "lr": 1.545363250762242e-05}
{"train_loss": 0.08173353970050812, "global_step": 201761, "epoch": 2266, "lr": 1.5453213404325094e-05}
{"train_loss": 0.06516133875617486, "global_step": 201762, "epoch": 2266, "lr": 1.5452794305672157e-05, "val_loss": 8.237887382507324}
{"train_loss": 0.048666615039110184, "global_step": 201763, "epoch": 2267, "lr": 1.5452375211663684e-05}
{"train_loss": 0.047705698758363724, "global_step": 201764, "epoch": 2267, "lr": 1.54519561222997e-05}
{"train_loss": 0.0772964134812355, "global_step": 201765, "epoch": 2267, "lr": 1.5451537037580298e-05}
{"train_loss": 0.034304939210414886, "global_step": 201766, "epoch": 2267, "lr": 1.5451117957505497e-05}
{"train_loss": 0.03320133313536644, "global_step": 201767, "epoch": 2267, "lr": 1.5450698882075387e-05}
{"train_loss": 0.06565524637699127, "global_step": 201768, "epoch": 2267, "lr": 1.5450279811290013e-05}
{"train_loss": 0.04216502234339714, "global_step": 201769, "epoch": 2267, "lr": 1.5449860745149415e-05}
{"train_loss": 0.04926936328411102, "global_step": 201770, "epoch": 2267, "lr": 1.5449441683653672e-05}
{"train_loss": 0.11836601793766022, "global_step": 201771, "epoch": 2267, "lr": 1.5449022626802816e-05}
{"train_loss": 0.06569886207580566, "global_step": 201772, "epoch": 2267, "lr": 1.544860357459693e-05}
{"train_loss": 0.0674666315317154, "global_step": 201773, "epoch": 2267, "lr": 1.5448184527036048e-05}
{"train_loss": 0.09992610663175583, "global_step": 201774, "epoch": 2267, "lr": 1.5447765484120254e-05}
{"train_loss": 0.07200281322002411, "global_step": 201775, "epoch": 2267, "lr": 1.5447346445849564e-05}
{"train_loss": 0.04566321149468422, "global_step": 201776, "epoch": 2267, "lr": 1.5446927412224078e-05}
{"train_loss": 0.02462681010365486, "global_step": 201777, "epoch": 2267, "lr": 1.544650838324381e-05}
{"train_loss": 0.0479983389377594, "global_step": 201778, "epoch": 2267, "lr": 1.544608935890886e-05}
{"train_loss": 0.037333205342292786, "global_step": 201779, "epoch": 2267, "lr": 1.5445670339219248e-05}
{"train_loss": 0.07530750334262848, "global_step": 201780, "epoch": 2267, "lr": 1.5445251324175035e-05}
{"train_loss": 0.041816212236881256, "global_step": 201781, "epoch": 2267, "lr": 1.5444832313776314e-05}
{"train_loss": 0.041748613119125366, "global_step": 201782, "epoch": 2267, "lr": 1.5444413308023087e-05}
{"train_loss": 0.07778050005435944, "global_step": 201783, "epoch": 2267, "lr": 1.544399430691546e-05}
{"train_loss": 0.05179116502404213, "global_step": 201784, "epoch": 2267, "lr": 1.544357531045345e-05}
{"train_loss": 0.060557134449481964, "global_step": 201785, "epoch": 2267, "lr": 1.5443156318637147e-05}
{"train_loss": 0.05212802812457085, "global_step": 201786, "epoch": 2267, "lr": 1.5442737331466572e-05}
{"train_loss": 0.04823121055960655, "global_step": 201787, "epoch": 2267, "lr": 1.544231834894182e-05}
{"train_loss": 0.045042749494314194, "global_step": 201788, "epoch": 2267, "lr": 1.5441899371062907e-05}
{"train_loss": 0.0838906317949295, "global_step": 201789, "epoch": 2267, "lr": 1.5441480397829928e-05}
{"train_loss": 0.01946425251662731, "global_step": 201790, "epoch": 2267, "lr": 1.5441061429242898e-05}
{"train_loss": 0.026392528787255287, "global_step": 201791, "epoch": 2267, "lr": 1.5440642465301917e-05}
{"train_loss": 0.10799728333950043, "global_step": 201792, "epoch": 2267, "lr": 1.5440223506007008e-05}
{"train_loss": 0.046069059520959854, "global_step": 201793, "epoch": 2267, "lr": 1.543980455135825e-05}
{"train_loss": 0.06155163794755936, "global_step": 201794, "epoch": 2267, "lr": 1.5439385601355678e-05}
{"train_loss": 0.08531131595373154, "global_step": 201795, "epoch": 2267, "lr": 1.543896665599937e-05}
{"train_loss": 0.10609854757785797, "global_step": 201796, "epoch": 2267, "lr": 1.543854771528936e-05}
{"train_loss": 0.04786824807524681, "global_step": 201797, "epoch": 2267, "lr": 1.543812877922573e-05}
{"train_loss": 0.05664392188191414, "global_step": 201798, "epoch": 2267, "lr": 1.543770984780851e-05}
{"train_loss": 0.07264642417430878, "global_step": 201799, "epoch": 2267, "lr": 1.543729092103776e-05}
{"train_loss": 0.11591950803995132, "global_step": 201800, "epoch": 2267, "lr": 1.543687199891357e-05}
{"train_loss": 0.08985289186239243, "global_step": 201801, "epoch": 2267, "lr": 1.543645308143597e-05}
{"train_loss": 0.09129396080970764, "global_step": 201802, "epoch": 2267, "lr": 1.5436034168604997e-05}
{"train_loss": 0.04697723686695099, "global_step": 201803, "epoch": 2267, "lr": 1.5435615260420743e-05}
{"train_loss": 0.08590304106473923, "global_step": 201804, "epoch": 2267, "lr": 1.543519635688324e-05}
{"train_loss": 0.03861003369092941, "global_step": 201805, "epoch": 2267, "lr": 1.5434777457992565e-05}
{"train_loss": 0.06799209862947464, "global_step": 201806, "epoch": 2267, "lr": 1.5434358563748762e-05}
{"train_loss": 0.04089786484837532, "global_step": 201807, "epoch": 2267, "lr": 1.5433939674151878e-05}
{"train_loss": 0.05449417605996132, "global_step": 201808, "epoch": 2267, "lr": 1.5433520789201988e-05}
{"train_loss": 0.060599129647016525, "global_step": 201809, "epoch": 2267, "lr": 1.5433101908899128e-05}
{"train_loss": 0.09611134976148605, "global_step": 201810, "epoch": 2267, "lr": 1.5432683033243382e-05}
{"train_loss": 0.03833095729351044, "global_step": 201811, "epoch": 2267, "lr": 1.5432264162234778e-05}
{"train_loss": 0.11796973645687103, "global_step": 201812, "epoch": 2267, "lr": 1.5431845295873392e-05}
{"train_loss": 0.07803928852081299, "global_step": 201813, "epoch": 2267, "lr": 1.5431426434159264e-05}
{"train_loss": 0.028868570923805237, "global_step": 201814, "epoch": 2267, "lr": 1.5431007577092472e-05}
{"train_loss": 0.07714491337537766, "global_step": 201815, "epoch": 2267, "lr": 1.5430588724673046e-05}
{"train_loss": 0.09372000396251678, "global_step": 201816, "epoch": 2267, "lr": 1.5430169876901056e-05}
{"train_loss": 0.06182863935828209, "global_step": 201817, "epoch": 2267, "lr": 1.5429751033776572e-05}
{"train_loss": 0.08773908019065857, "global_step": 201818, "epoch": 2267, "lr": 1.5429332195299623e-05}
{"train_loss": 0.08275353908538818, "global_step": 201819, "epoch": 2267, "lr": 1.5428913361470293e-05}
{"train_loss": 0.05844653397798538, "global_step": 201820, "epoch": 2267, "lr": 1.5428494532288613e-05}
{"train_loss": 0.08540358394384384, "global_step": 201821, "epoch": 2267, "lr": 1.5428075707754656e-05}
{"train_loss": 0.06969060003757477, "global_step": 201822, "epoch": 2267, "lr": 1.5427656887868468e-05}
{"train_loss": 0.0504511259496212, "global_step": 201823, "epoch": 2267, "lr": 1.5427238072630118e-05}
{"train_loss": 0.056403230875730515, "global_step": 201824, "epoch": 2267, "lr": 1.5426819262039642e-05}
{"train_loss": 0.02857937105000019, "global_step": 201825, "epoch": 2267, "lr": 1.542640045609713e-05}
{"train_loss": 0.06108640879392624, "global_step": 201826, "epoch": 2267, "lr": 1.5425981654802597e-05}
{"train_loss": 0.04880668967962265, "global_step": 201827, "epoch": 2267, "lr": 1.542556285815613e-05}
{"train_loss": 0.060849543660879135, "global_step": 201828, "epoch": 2267, "lr": 1.5425144066157765e-05}
{"train_loss": 0.06409820914268494, "global_step": 201829, "epoch": 2267, "lr": 1.5424725278807583e-05}
{"train_loss": 0.07136021554470062, "global_step": 201830, "epoch": 2267, "lr": 1.5424306496105605e-05}
{"train_loss": 0.03029724955558777, "global_step": 201831, "epoch": 2267, "lr": 1.5423887718051928e-05}
{"train_loss": 0.05633144825696945, "global_step": 201832, "epoch": 2267, "lr": 1.5423468944646567e-05}
{"train_loss": 0.047849539667367935, "global_step": 201833, "epoch": 2267, "lr": 1.5423050175889625e-05}
{"train_loss": 0.029496613889932632, "global_step": 201834, "epoch": 2267, "lr": 1.542263141178111e-05}
{"train_loss": 0.09484343975782394, "global_step": 201835, "epoch": 2267, "lr": 1.54222126523211e-05}
{"train_loss": 0.07152693718671799, "global_step": 201836, "epoch": 2267, "lr": 1.5421793897509666e-05}
{"train_loss": 0.09761399030685425, "global_step": 201837, "epoch": 2267, "lr": 1.542137514734684e-05}
{"train_loss": 0.0813508927822113, "global_step": 201838, "epoch": 2267, "lr": 1.5420956401832704e-05}
{"train_loss": 0.07984399795532227, "global_step": 201839, "epoch": 2267, "lr": 1.5420537660967277e-05}
{"train_loss": 0.052591048181056976, "global_step": 201840, "epoch": 2267, "lr": 1.542011892475066e-05}
{"train_loss": 0.037929292768239975, "global_step": 201841, "epoch": 2267, "lr": 1.5419700193182883e-05}
{"train_loss": 0.08685768395662308, "global_step": 201842, "epoch": 2267, "lr": 1.541928146626399e-05}
{"train_loss": 0.126109316945076, "global_step": 201843, "epoch": 2267, "lr": 1.541886274399407e-05}
{"train_loss": 0.048986583948135376, "global_step": 201844, "epoch": 2267, "lr": 1.5418444026373145e-05}
{"train_loss": 0.08900202065706253, "global_step": 201845, "epoch": 2267, "lr": 1.541802531340131e-05}
{"train_loss": 0.08131523430347443, "global_step": 201846, "epoch": 2267, "lr": 1.5417606605078595e-05}
{"train_loss": 0.06388475745916367, "global_step": 201847, "epoch": 2267, "lr": 1.5417187901405046e-05}
{"train_loss": 0.0382760614156723, "global_step": 201848, "epoch": 2267, "lr": 1.541676920238075e-05}
{"train_loss": 0.052207741886377335, "global_step": 201849, "epoch": 2267, "lr": 1.5416350508005734e-05}
{"train_loss": 0.06321900337934494, "global_step": 201850, "epoch": 2267, "lr": 1.5415931818280083e-05}
{"train_loss": 0.06436246233793456, "global_step": 201851, "epoch": 2267, "lr": 1.5415513133203823e-05, "val_loss": 8.274985313415527}
{"train_loss": 0.03556341305375099, "global_step": 201852, "epoch": 2268, "lr": 1.5415094452777025e-05}
{"train_loss": 0.12252935767173767, "global_step": 201853, "epoch": 2268, "lr": 1.541467577699976e-05}
{"train_loss": 0.09557993710041046, "global_step": 201854, "epoch": 2268, "lr": 1.5414257105872056e-05}
{"train_loss": 0.10309849679470062, "global_step": 201855, "epoch": 2268, "lr": 1.5413838439394003e-05}
{"train_loss": 0.04560823366045952, "global_step": 201856, "epoch": 2268, "lr": 1.541341977756562e-05}
{"train_loss": 0.027068618685007095, "global_step": 201857, "epoch": 2268, "lr": 1.5413001120386993e-05}
{"train_loss": 0.08512897044420242, "global_step": 201858, "epoch": 2268, "lr": 1.5412582467858157e-05}
{"train_loss": 0.05677161365747452, "global_step": 201859, "epoch": 2268, "lr": 1.5412163819979186e-05}
{"train_loss": 0.023837590590119362, "global_step": 201860, "epoch": 2268, "lr": 1.5411745176750114e-05}
{"train_loss": 0.07661435008049011, "global_step": 201861, "epoch": 2268, "lr": 1.5411326538171033e-05}
{"train_loss": 0.09054325520992279, "global_step": 201862, "epoch": 2268, "lr": 1.5410907904241957e-05}
{"train_loss": 0.06379213184118271, "global_step": 201863, "epoch": 2268, "lr": 1.541048927496298e-05}
{"train_loss": 0.10475222021341324, "global_step": 201864, "epoch": 2268, "lr": 1.5410070650334124e-05}
{"train_loss": 0.028286611661314964, "global_step": 201865, "epoch": 2268, "lr": 1.540965203035548e-05}
{"train_loss": 0.0551561638712883, "global_step": 201866, "epoch": 2268, "lr": 1.5409233415027067e-05}
{"train_loss": 0.09448958188295364, "global_step": 201867, "epoch": 2268, "lr": 1.540881480434898e-05}
{"train_loss": 0.041012492030858994, "global_step": 201868, "epoch": 2268, "lr": 1.5408396198321247e-05}
{"train_loss": 0.0709628015756607, "global_step": 201869, "epoch": 2268, "lr": 1.540797759694394e-05}
{"train_loss": 0.03816390782594681, "global_step": 201870, "epoch": 2268, "lr": 1.5407559000217097e-05}
{"train_loss": 0.05276534706354141, "global_step": 201871, "epoch": 2268, "lr": 1.5407140408140786e-05}
{"train_loss": 0.11271369457244873, "global_step": 201872, "epoch": 2268, "lr": 1.5406721820715075e-05}
{"train_loss": 0.07098223268985748, "global_step": 201873, "epoch": 2268, "lr": 1.5406303237939995e-05}
{"train_loss": 0.026667483150959015, "global_step": 201874, "epoch": 2268, "lr": 1.5405884659815634e-05}
{"train_loss": 0.06341709196567535, "global_step": 201875, "epoch": 2268, "lr": 1.5405466086342013e-05}
{"train_loss": 0.04590853676199913, "global_step": 201876, "epoch": 2268, "lr": 1.5405047517519222e-05}
{"train_loss": 0.12611734867095947, "global_step": 201877, "epoch": 2268, "lr": 1.5404628953347283e-05}
{"train_loss": 0.0707099661231041, "global_step": 201878, "epoch": 2268, "lr": 1.5404210393826286e-05}
{"train_loss": 0.07254014909267426, "global_step": 201879, "epoch": 2268, "lr": 1.5403791838956273e-05}
{"train_loss": 0.08084045350551605, "global_step": 201880, "epoch": 2268, "lr": 1.5403373288737284e-05}
{"train_loss": 0.06570306420326233, "global_step": 201881, "epoch": 2268, "lr": 1.54029547431694e-05}
{"train_loss": 0.1338128000497818, "global_step": 201882, "epoch": 2268, "lr": 1.5402536202252656e-05}
{"train_loss": 0.03934641182422638, "global_step": 201883, "epoch": 2268, "lr": 1.540211766598713e-05}
{"train_loss": 0.06551493704319, "global_step": 201884, "epoch": 2268, "lr": 1.5401699134372876e-05}
{"train_loss": 0.1065034344792366, "global_step": 201885, "epoch": 2268, "lr": 1.5401280607409918e-05}
{"train_loss": 0.04532717540860176, "global_step": 201886, "epoch": 2268, "lr": 1.5400862085098354e-05}
{"train_loss": 0.07276205718517303, "global_step": 201887, "epoch": 2268, "lr": 1.540044356743821e-05}
{"train_loss": 0.05269193649291992, "global_step": 201888, "epoch": 2268, "lr": 1.5400025054429568e-05}
{"train_loss": 0.07244371622800827, "global_step": 201889, "epoch": 2268, "lr": 1.539960654607246e-05}
{"train_loss": 0.06388265639543533, "global_step": 201890, "epoch": 2268, "lr": 1.5399188042366945e-05}
{"train_loss": 0.06675689667463303, "global_step": 201891, "epoch": 2268, "lr": 1.5398769543313107e-05}
{"train_loss": 0.0460030697286129, "global_step": 201892, "epoch": 2268, "lr": 1.5398351048910963e-05}
{"train_loss": 0.09398388862609863, "global_step": 201893, "epoch": 2268, "lr": 1.539793255916061e-05}
{"train_loss": 0.08089569211006165, "global_step": 201894, "epoch": 2268, "lr": 1.539751407406207e-05}
{"train_loss": 0.06009988859295845, "global_step": 201895, "epoch": 2268, "lr": 1.5397095593615423e-05}
{"train_loss": 0.04311503469944, "global_step": 201896, "epoch": 2268, "lr": 1.5396677117820702e-05}
{"train_loss": 0.039445001631975174, "global_step": 201897, "epoch": 2268, "lr": 1.5396258646677985e-05}
{"train_loss": 0.058341849595308304, "global_step": 201898, "epoch": 2268, "lr": 1.539584018018731e-05}
{"train_loss": 0.11382788419723511, "global_step": 201899, "epoch": 2268, "lr": 1.5395421718348757e-05}
{"train_loss": 0.061808060854673386, "global_step": 201900, "epoch": 2268, "lr": 1.5395003261162354e-05}
{"train_loss": 0.08069773763418198, "global_step": 201901, "epoch": 2268, "lr": 1.539458480862818e-05}
{"train_loss": 0.12076928466558456, "global_step": 201902, "epoch": 2268, "lr": 1.5394166360746277e-05}
{"train_loss": 0.05037490651011467, "global_step": 201903, "epoch": 2268, "lr": 1.5393747917516717e-05}
{"train_loss": 0.09315038472414017, "global_step": 201904, "epoch": 2268, "lr": 1.5393329478939526e-05}
{"train_loss": 0.0939633771777153, "global_step": 201905, "epoch": 2268, "lr": 1.5392911045014803e-05}
{"train_loss": 0.05024068430066109, "global_step": 201906, "epoch": 2268, "lr": 1.539249261574256e-05}
{"train_loss": 0.05808698758482933, "global_step": 201907, "epoch": 2268, "lr": 1.5392074191122874e-05}
{"train_loss": 0.0602763406932354, "global_step": 201908, "epoch": 2268, "lr": 1.539165577115582e-05}
{"train_loss": 0.033112533390522, "global_step": 201909, "epoch": 2268, "lr": 1.539123735584142e-05}
{"train_loss": 0.052462171763181686, "global_step": 201910, "epoch": 2268, "lr": 1.5390818945179764e-05}
{"train_loss": 0.03555457293987274, "global_step": 201911, "epoch": 2268, "lr": 1.5390400539170875e-05}
{"train_loss": 0.03957861289381981, "global_step": 201912, "epoch": 2268, "lr": 1.5389982137814833e-05}
{"train_loss": 0.04571551829576492, "global_step": 201913, "epoch": 2268, "lr": 1.538956374111168e-05}
{"train_loss": 0.10173103958368301, "global_step": 201914, "epoch": 2268, "lr": 1.538914534906149e-05}
{"train_loss": 0.03925291448831558, "global_step": 201915, "epoch": 2268, "lr": 1.538872696166429e-05}
{"train_loss": 0.09242881834506989, "global_step": 201916, "epoch": 2268, "lr": 1.5388308578920174e-05}
{"train_loss": 0.08071791380643845, "global_step": 201917, "epoch": 2268, "lr": 1.5387890200829175e-05}
{"train_loss": 0.06276463717222214, "global_step": 201918, "epoch": 2268, "lr": 1.5387471827391336e-05}
{"train_loss": 0.0756407082080841, "global_step": 201919, "epoch": 2268, "lr": 1.5387053458606747e-05}
{"train_loss": 0.04075796157121658, "global_step": 201920, "epoch": 2268, "lr": 1.538663509447543e-05}
{"train_loss": 0.06201031804084778, "global_step": 201921, "epoch": 2268, "lr": 1.538621673499747e-05}
{"train_loss": 0.04035584256052971, "global_step": 201922, "epoch": 2268, "lr": 1.5385798380172922e-05}
{"train_loss": 0.05327402055263519, "global_step": 201923, "epoch": 2268, "lr": 1.5385380030001807e-05}
{"train_loss": 0.04609595239162445, "global_step": 201924, "epoch": 2268, "lr": 1.5384961684484224e-05}
{"train_loss": 0.05090503394603729, "global_step": 201925, "epoch": 2268, "lr": 1.5384543343620195e-05}
{"train_loss": 0.047033149749040604, "global_step": 201926, "epoch": 2268, "lr": 1.5384125007409796e-05}
{"train_loss": 0.0991780236363411, "global_step": 201927, "epoch": 2268, "lr": 1.5383706675853087e-05}
{"train_loss": 0.06370692700147629, "global_step": 201928, "epoch": 2268, "lr": 1.5383288348950103e-05}
{"train_loss": 0.025472491979599, "global_step": 201929, "epoch": 2268, "lr": 1.5382870026700934e-05}
{"train_loss": 0.04569094628095627, "global_step": 201930, "epoch": 2268, "lr": 1.53824517091056e-05}
{"train_loss": 0.03649280592799187, "global_step": 201931, "epoch": 2268, "lr": 1.5382033396164187e-05}
{"train_loss": 0.07914099097251892, "global_step": 201932, "epoch": 2268, "lr": 1.538161508787672e-05}
{"train_loss": 0.11160952597856522, "global_step": 201933, "epoch": 2268, "lr": 1.5381196784243296e-05}
{"train_loss": 0.07944848388433456, "global_step": 201934, "epoch": 2268, "lr": 1.5380778485263925e-05}
{"train_loss": 0.06621591746807098, "global_step": 201935, "epoch": 2268, "lr": 1.538036019093871e-05}
{"train_loss": 0.05264819413423538, "global_step": 201936, "epoch": 2268, "lr": 1.5379941901267658e-05}
{"train_loss": 0.05570929870009422, "global_step": 201937, "epoch": 2268, "lr": 1.5379523616250873e-05}
{"train_loss": 0.07662184536457062, "global_step": 201938, "epoch": 2268, "lr": 1.537910533588837e-05}
{"train_loss": 0.02964147925376892, "global_step": 201939, "epoch": 2268, "lr": 1.537868706018024e-05}
{"train_loss": 0.06562773824742671, "global_step": 201940, "epoch": 2268, "lr": 1.537826878912651e-05, "val_loss": 8.16960620880127}
{"train_loss": 0.09014466404914856, "global_step": 201941, "epoch": 2269, "lr": 1.5377850522727265e-05}
{"train_loss": 0.04559675231575966, "global_step": 201942, "epoch": 2269, "lr": 1.5377432260982534e-05}
{"train_loss": 0.07395793497562408, "global_step": 201943, "epoch": 2269, "lr": 1.537701400389238e-05}
{"train_loss": 0.049226678907871246, "global_step": 201944, "epoch": 2269, "lr": 1.5376595751456884e-05}
{"train_loss": 0.07642208784818649, "global_step": 201945, "epoch": 2269, "lr": 1.5376177503676064e-05}
{"train_loss": 0.04534953087568283, "global_step": 201946, "epoch": 2269, "lr": 1.537575926055001e-05}
{"train_loss": 0.07569471001625061, "global_step": 201947, "epoch": 2269, "lr": 1.5375341022078748e-05}
{"train_loss": 0.028859395533800125, "global_step": 201948, "epoch": 2269, "lr": 1.537492278826237e-05}
{"train_loss": 0.046037815511226654, "global_step": 201949, "epoch": 2269, "lr": 1.537450455910089e-05}
{"train_loss": 0.08534029871225357, "global_step": 201950, "epoch": 2269, "lr": 1.53740863345944e-05}
{"train_loss": 0.05736033245921135, "global_step": 201951, "epoch": 2269, "lr": 1.5373668114742933e-05}
{"train_loss": 0.05490934103727341, "global_step": 201952, "epoch": 2269, "lr": 1.5373249899546565e-05}
{"train_loss": 0.10585156828165054, "global_step": 201953, "epoch": 2269, "lr": 1.537283168900533e-05}
{"train_loss": 0.1113249883055687, "global_step": 201954, "epoch": 2269, "lr": 1.5372413483119307e-05}
{"train_loss": 0.06712846457958221, "global_step": 201955, "epoch": 2269, "lr": 1.537199528188853e-05}
{"train_loss": 0.052377477288246155, "global_step": 201956, "epoch": 2269, "lr": 1.5371577085313078e-05}
{"train_loss": 0.049159761518239975, "global_step": 201957, "epoch": 2269, "lr": 1.5371158893392996e-05}
{"train_loss": 0.054038193076848984, "global_step": 201958, "epoch": 2269, "lr": 1.5370740706128323e-05}
{"train_loss": 0.04085857421159744, "global_step": 201959, "epoch": 2269, "lr": 1.537032252351915e-05}
{"train_loss": 0.07032126933336258, "global_step": 201960, "epoch": 2269, "lr": 1.53699043455655e-05}
{"train_loss": 0.05720601975917816, "global_step": 201961, "epoch": 2269, "lr": 1.5369486172267454e-05}
{"train_loss": 0.08312511444091797, "global_step": 201962, "epoch": 2269, "lr": 1.5369068003625054e-05}
{"train_loss": 0.07440165430307388, "global_step": 201963, "epoch": 2269, "lr": 1.5368649839638364e-05}
{"train_loss": 0.07157070934772491, "global_step": 201964, "epoch": 2269, "lr": 1.5368231680307428e-05}
{"train_loss": 0.04207964614033699, "global_step": 201965, "epoch": 2269, "lr": 1.536781352563233e-05}
{"train_loss": 0.0646059438586235, "global_step": 201966, "epoch": 2269, "lr": 1.5367395375613085e-05}
{"train_loss": 0.058460500091314316, "global_step": 201967, "epoch": 2269, "lr": 1.536697723024979e-05}
{"train_loss": 0.03209972009062767, "global_step": 201968, "epoch": 2269, "lr": 1.5366559089542466e-05}
{"train_loss": 0.038438573479652405, "global_step": 201969, "epoch": 2269, "lr": 1.5366140953491204e-05}
{"train_loss": 0.04282292351126671, "global_step": 201970, "epoch": 2269, "lr": 1.5365722822096024e-05}
{"train_loss": 0.042952172458171844, "global_step": 201971, "epoch": 2269, "lr": 1.5365304695357015e-05}
{"train_loss": 0.0701972171664238, "global_step": 201972, "epoch": 2269, "lr": 1.5364886573274207e-05}
{"train_loss": 0.0789022147655487, "global_step": 201973, "epoch": 2269, "lr": 1.536446845584768e-05}
{"train_loss": 0.05311471223831177, "global_step": 201974, "epoch": 2269, "lr": 1.5364050343077462e-05}
{"train_loss": 0.056911490857601166, "global_step": 201975, "epoch": 2269, "lr": 1.5363632234963644e-05}
{"train_loss": 0.07397893071174622, "global_step": 201976, "epoch": 2269, "lr": 1.536321413150625e-05}
{"train_loss": 0.04849778115749359, "global_step": 201977, "epoch": 2269, "lr": 1.5362796032705362e-05}
{"train_loss": 0.06829214096069336, "global_step": 201978, "epoch": 2269, "lr": 1.5362377938561008e-05}
{"train_loss": 0.054365918040275574, "global_step": 201979, "epoch": 2269, "lr": 1.536195984907326e-05}
{"train_loss": 0.03945908322930336, "global_step": 201980, "epoch": 2269, "lr": 1.536154176424219e-05}
{"train_loss": 0.035125866532325745, "global_step": 201981, "epoch": 2269, "lr": 1.5361123684067824e-05}
{"train_loss": 0.04993794113397598, "global_step": 201982, "epoch": 2269, "lr": 1.5360705608550245e-05}
{"train_loss": 0.07462626695632935, "global_step": 201983, "epoch": 2269, "lr": 1.536028753768948e-05}
{"train_loss": 0.07481119781732559, "global_step": 201984, "epoch": 2269, "lr": 1.5359869471485627e-05}
{"train_loss": 0.07124083489179611, "global_step": 201985, "epoch": 2269, "lr": 1.5359451409938695e-05}
{"train_loss": 0.02979431487619877, "global_step": 201986, "epoch": 2269, "lr": 1.535903335304878e-05}
{"train_loss": 0.051601018756628036, "global_step": 201987, "epoch": 2269, "lr": 1.5358615300815904e-05}
{"train_loss": 0.05646127462387085, "global_step": 201988, "epoch": 2269, "lr": 1.5358197253240157e-05}
{"train_loss": 0.05005739629268646, "global_step": 201989, "epoch": 2269, "lr": 1.5357779210321567e-05}
{"train_loss": 0.08966189622879028, "global_step": 201990, "epoch": 2269, "lr": 1.5357361172060208e-05}
{"train_loss": 0.0916166678071022, "global_step": 201991, "epoch": 2269, "lr": 1.535694313845612e-05}
{"train_loss": 0.06586118787527084, "global_step": 201992, "epoch": 2269, "lr": 1.535652510950938e-05}
{"train_loss": 0.06675013154745102, "global_step": 201993, "epoch": 2269, "lr": 1.5356107085220018e-05}
{"train_loss": 0.049263712018728256, "global_step": 201994, "epoch": 2269, "lr": 1.5355689065588125e-05}
{"train_loss": 0.030469773337244987, "global_step": 201995, "epoch": 2269, "lr": 1.5355271050613735e-05}
{"train_loss": 0.05883632227778435, "global_step": 201996, "epoch": 2269, "lr": 1.535485304029689e-05}
{"train_loss": 0.07166269421577454, "global_step": 201997, "epoch": 2269, "lr": 1.5354435034637677e-05}
{"train_loss": 0.03878609836101532, "global_step": 201998, "epoch": 2269, "lr": 1.5354017033636124e-05}
{"train_loss": 0.0712386965751648, "global_step": 201999, "epoch": 2269, "lr": 1.5353599037292315e-05}
{"train_loss": 0.06791560351848602, "global_step": 202000, "epoch": 2269, "lr": 1.535318104560628e-05}
{"train_loss": 0.07219564914703369, "global_step": 202001, "epoch": 2269, "lr": 1.53527630585781e-05}
{"train_loss": 0.06457951664924622, "global_step": 202002, "epoch": 2269, "lr": 1.5352345076207807e-05}
{"train_loss": 0.03264600783586502, "global_step": 202003, "epoch": 2269, "lr": 1.5351927098495483e-05}
{"train_loss": 0.07893849164247513, "global_step": 202004, "epoch": 2269, "lr": 1.5351509125441156e-05}
{"train_loss": 0.029913349077105522, "global_step": 202005, "epoch": 2269, "lr": 1.535109115704491e-05}
{"train_loss": 0.05576840043067932, "global_step": 202006, "epoch": 2269, "lr": 1.535067319330677e-05}
{"train_loss": 0.07599881291389465, "global_step": 202007, "epoch": 2269, "lr": 1.535025523422683e-05}
{"train_loss": 0.0998610407114029, "global_step": 202008, "epoch": 2269, "lr": 1.5349837279805108e-05}
{"train_loss": 0.05204625055193901, "global_step": 202009, "epoch": 2269, "lr": 1.5349419330041693e-05}
{"train_loss": 0.03754130378365517, "global_step": 202010, "epoch": 2269, "lr": 1.5349001384936607e-05}
{"train_loss": 0.06760574877262115, "global_step": 202011, "epoch": 2269, "lr": 1.5348583444489945e-05}
{"train_loss": 0.055206432938575745, "global_step": 202012, "epoch": 2269, "lr": 1.5348165508701728e-05}
{"train_loss": 0.08129268884658813, "global_step": 202013, "epoch": 2269, "lr": 1.5347747577572042e-05}
{"train_loss": 0.07141625136137009, "global_step": 202014, "epoch": 2269, "lr": 1.5347329651100915e-05}
{"train_loss": 0.10837841778993607, "global_step": 202015, "epoch": 2269, "lr": 1.5346911729288433e-05}
{"train_loss": 0.09101229906082153, "global_step": 202016, "epoch": 2269, "lr": 1.5346493812134614e-05}
{"train_loss": 0.10385196655988693, "global_step": 202017, "epoch": 2269, "lr": 1.5346075899639545e-05}
{"train_loss": 0.05256617069244385, "global_step": 202018, "epoch": 2269, "lr": 1.5345657991803285e-05}
{"train_loss": 0.08169698715209961, "global_step": 202019, "epoch": 2269, "lr": 1.5345240088625867e-05}
{"train_loss": 0.06040070578455925, "global_step": 202020, "epoch": 2269, "lr": 1.5344822190107373e-05}
{"train_loss": 0.03957938775420189, "global_step": 202021, "epoch": 2269, "lr": 1.5344404296247835e-05}
{"train_loss": 0.07963308691978455, "global_step": 202022, "epoch": 2269, "lr": 1.5343986407047323e-05}
{"train_loss": 0.03196478262543678, "global_step": 202023, "epoch": 2269, "lr": 1.534356852250588e-05}
{"train_loss": 0.03327275067567825, "global_step": 202024, "epoch": 2269, "lr": 1.5343150642623588e-05}
{"train_loss": 0.019317757338285446, "global_step": 202025, "epoch": 2269, "lr": 1.5342732767400475e-05}
{"train_loss": 0.08820688724517822, "global_step": 202026, "epoch": 2269, "lr": 1.534231489683662e-05}
{"train_loss": 0.03398789092898369, "global_step": 202027, "epoch": 2269, "lr": 1.5341897030932056e-05}
{"train_loss": 0.03966452181339264, "global_step": 202028, "epoch": 2269, "lr": 1.534147916968686e-05}
{"train_loss": 0.06123049655573422, "global_step": 202029, "epoch": 2269, "lr": 1.5341061313101075e-05, "val_loss": 8.279644012451172}
{"train_loss": 0.0489463284611702, "global_step": 202030, "epoch": 2270, "lr": 1.534064346117477e-05}
{"train_loss": 0.038007598370313644, "global_step": 202031, "epoch": 2270, "lr": 1.5340225613907977e-05}
{"train_loss": 0.062484629452228546, "global_step": 202032, "epoch": 2270, "lr": 1.5339807771300786e-05}
{"train_loss": 0.024163678288459778, "global_step": 202033, "epoch": 2270, "lr": 1.533938993335324e-05}
{"train_loss": 0.07502681761980057, "global_step": 202034, "epoch": 2270, "lr": 1.5338972100065362e-05}
{"train_loss": 0.040341250598430634, "global_step": 202035, "epoch": 2270, "lr": 1.5338554271437268e-05}
{"train_loss": 0.08954343944787979, "global_step": 202036, "epoch": 2270, "lr": 1.5338136447468953e-05}
{"train_loss": 0.03263479843735695, "global_step": 202037, "epoch": 2270, "lr": 1.533771862816053e-05}
{"train_loss": 0.09216634929180145, "global_step": 202038, "epoch": 2270, "lr": 1.5337300813512008e-05}
{"train_loss": 0.037215184420347214, "global_step": 202039, "epoch": 2270, "lr": 1.5336883003523476e-05}
{"train_loss": 0.04786348715424538, "global_step": 202040, "epoch": 2270, "lr": 1.5336465198194965e-05}
{"train_loss": 0.05968187749385834, "global_step": 202041, "epoch": 2270, "lr": 1.533604739752656e-05}
{"train_loss": 0.03974858298897743, "global_step": 202042, "epoch": 2270, "lr": 1.5335629601518288e-05}
{"train_loss": 0.06814340502023697, "global_step": 202043, "epoch": 2270, "lr": 1.533521181017023e-05}
{"train_loss": 0.08314366638660431, "global_step": 202044, "epoch": 2270, "lr": 1.5334794023482414e-05}
{"train_loss": 0.07428271323442459, "global_step": 202045, "epoch": 2270, "lr": 1.533437624145493e-05}
{"train_loss": 0.1412924826145172, "global_step": 202046, "epoch": 2270, "lr": 1.53339584640878e-05}
{"train_loss": 0.05124657601118088, "global_step": 202047, "epoch": 2270, "lr": 1.533354069138111e-05}
{"train_loss": 0.0803183987736702, "global_step": 202048, "epoch": 2270, "lr": 1.5333122923334892e-05}
{"train_loss": 0.04480305314064026, "global_step": 202049, "epoch": 2270, "lr": 1.533270515994923e-05}
{"train_loss": 0.06368458271026611, "global_step": 202050, "epoch": 2270, "lr": 1.5332287401224142e-05}
{"train_loss": 0.057293783873319626, "global_step": 202051, "epoch": 2270, "lr": 1.5331869647159724e-05}
{"train_loss": 0.08219214528799057, "global_step": 202052, "epoch": 2270, "lr": 1.5331451897755994e-05}
{"train_loss": 0.10951391607522964, "global_step": 202053, "epoch": 2270, "lr": 1.5331034153013035e-05}
{"train_loss": 0.05734650045633316, "global_step": 202054, "epoch": 2270, "lr": 1.5330616412930905e-05}
{"train_loss": 0.08983855694532394, "global_step": 202055, "epoch": 2270, "lr": 1.533019867750964e-05}
{"train_loss": 0.041498832404613495, "global_step": 202056, "epoch": 2270, "lr": 1.5329780946749317e-05}
{"train_loss": 0.06867071986198425, "global_step": 202057, "epoch": 2270, "lr": 1.5329363220649973e-05}
{"train_loss": 0.08693967759609222, "global_step": 202058, "epoch": 2270, "lr": 1.532894549921169e-05}
{"train_loss": 0.078219473361969, "global_step": 202059, "epoch": 2270, "lr": 1.532852778243449e-05}
{"train_loss": 0.08982130885124207, "global_step": 202060, "epoch": 2270, "lr": 1.5328110070318457e-05}
{"train_loss": 0.05354876443743706, "global_step": 202061, "epoch": 2270, "lr": 1.5327692362863626e-05}
{"train_loss": 0.04988660290837288, "global_step": 202062, "epoch": 2270, "lr": 1.532727466007008e-05}
{"train_loss": 0.0809633880853653, "global_step": 202063, "epoch": 2270, "lr": 1.5326856961937842e-05}
{"train_loss": 0.1982959806919098, "global_step": 202064, "epoch": 2270, "lr": 1.5326439268467e-05}
{"train_loss": 0.055710893124341965, "global_step": 202065, "epoch": 2270, "lr": 1.5326021579657585e-05}
{"train_loss": 0.07585699111223221, "global_step": 202066, "epoch": 2270, "lr": 1.5325603895509676e-05}
{"train_loss": 0.04209250211715698, "global_step": 202067, "epoch": 2270, "lr": 1.53251862160233e-05}
{"train_loss": 0.047828253358602524, "global_step": 202068, "epoch": 2270, "lr": 1.5324768541198547e-05}
{"train_loss": 0.11201436817646027, "global_step": 202069, "epoch": 2270, "lr": 1.5324350871035444e-05}
{"train_loss": 0.0390593521296978, "global_step": 202070, "epoch": 2270, "lr": 1.5323933205534074e-05}
{"train_loss": 0.053575508296489716, "global_step": 202071, "epoch": 2270, "lr": 1.532351554469446e-05}
{"train_loss": 0.06061186268925667, "global_step": 202072, "epoch": 2270, "lr": 1.532309788851669e-05}
{"train_loss": 0.08389036357402802, "global_step": 202073, "epoch": 2270, "lr": 1.5322680237000815e-05}
{"train_loss": 0.034913115203380585, "global_step": 202074, "epoch": 2270, "lr": 1.5322262590146862e-05}
{"train_loss": 0.020889444276690483, "global_step": 202075, "epoch": 2270, "lr": 1.532184494795492e-05}
{"train_loss": 0.06477776914834976, "global_step": 202076, "epoch": 2270, "lr": 1.5321427310425025e-05}
{"train_loss": 0.03670486807823181, "global_step": 202077, "epoch": 2270, "lr": 1.5321009677557256e-05}
{"train_loss": 0.09736593812704086, "global_step": 202078, "epoch": 2270, "lr": 1.532059204935164e-05}
{"train_loss": 0.06413770467042923, "global_step": 202079, "epoch": 2270, "lr": 1.532017442580826e-05}
{"train_loss": 0.07938393205404282, "global_step": 202080, "epoch": 2270, "lr": 1.5319756806927145e-05}
{"train_loss": 0.0956580638885498, "global_step": 202081, "epoch": 2270, "lr": 1.5319339192708375e-05}
{"train_loss": 0.06769538670778275, "global_step": 202082, "epoch": 2270, "lr": 1.531892158315199e-05}
{"train_loss": 0.05126980319619179, "global_step": 202083, "epoch": 2270, "lr": 1.531850397825807e-05}
{"train_loss": 0.020777184516191483, "global_step": 202084, "epoch": 2270, "lr": 1.5318086378026637e-05}
{"train_loss": 0.0518060065805912, "global_step": 202085, "epoch": 2270, "lr": 1.5317668782457778e-05}
{"train_loss": 0.11094843596220016, "global_step": 202086, "epoch": 2270, "lr": 1.531725119155152e-05}
{"train_loss": 0.1006135493516922, "global_step": 202087, "epoch": 2270, "lr": 1.531683360530795e-05}
{"train_loss": 0.0538545586168766, "global_step": 202088, "epoch": 2270, "lr": 1.5316416023727092e-05}
{"train_loss": 0.030214298516511917, "global_step": 202089, "epoch": 2270, "lr": 1.5315998446809027e-05}
{"train_loss": 0.0753943994641304, "global_step": 202090, "epoch": 2270, "lr": 1.5315580874553815e-05}
{"train_loss": 0.05321620777249336, "global_step": 202091, "epoch": 2270, "lr": 1.5315163306961478e-05}
{"train_loss": 0.07861878722906113, "global_step": 202092, "epoch": 2270, "lr": 1.5314745744032115e-05}
{"train_loss": 0.0568401925265789, "global_step": 202093, "epoch": 2270, "lr": 1.5314328185765747e-05}
{"train_loss": 0.042753975838422775, "global_step": 202094, "epoch": 2270, "lr": 1.5313910632162464e-05}
{"train_loss": 0.03192063793540001, "global_step": 202095, "epoch": 2270, "lr": 1.531349308322228e-05}
{"train_loss": 0.03860672563314438, "global_step": 202096, "epoch": 2270, "lr": 1.5313075538945298e-05}
{"train_loss": 0.059911467134952545, "global_step": 202097, "epoch": 2270, "lr": 1.5312657999331527e-05}
{"train_loss": 0.053763799369335175, "global_step": 202098, "epoch": 2270, "lr": 1.5312240464381068e-05}
{"train_loss": 0.14266437292099, "global_step": 202099, "epoch": 2270, "lr": 1.5311822934093933e-05}
{"train_loss": 0.06866993755102158, "global_step": 202100, "epoch": 2270, "lr": 1.5311405408470218e-05}
{"train_loss": 0.05483036860823631, "global_step": 202101, "epoch": 2270, "lr": 1.5310987887509952e-05}
{"train_loss": 0.06650768220424652, "global_step": 202102, "epoch": 2270, "lr": 1.5310570371213212e-05}
{"train_loss": 0.044026363641023636, "global_step": 202103, "epoch": 2270, "lr": 1.531015285958003e-05}
{"train_loss": 0.1298219859600067, "global_step": 202104, "epoch": 2270, "lr": 1.5309735352610488e-05}
{"train_loss": 0.08420904725790024, "global_step": 202105, "epoch": 2270, "lr": 1.530931785030461e-05}
{"train_loss": 0.06990376114845276, "global_step": 202106, "epoch": 2270, "lr": 1.530890035266249e-05}
{"train_loss": 0.040212176740169525, "global_step": 202107, "epoch": 2270, "lr": 1.5308482859684155e-05}
{"train_loss": 0.14627498388290405, "global_step": 202108, "epoch": 2270, "lr": 1.5308065371369684e-05}
{"train_loss": 0.07318098098039627, "global_step": 202109, "epoch": 2270, "lr": 1.5307647887719105e-05}
{"train_loss": 0.12014046311378479, "global_step": 202110, "epoch": 2270, "lr": 1.5307230408732503e-05}
{"train_loss": 0.03772827982902527, "global_step": 202111, "epoch": 2270, "lr": 1.5306812934409925e-05}
{"train_loss": 0.08822696655988693, "global_step": 202112, "epoch": 2270, "lr": 1.5306395464751406e-05}
{"train_loss": 0.07568565011024475, "global_step": 202113, "epoch": 2270, "lr": 1.530597799975704e-05}
{"train_loss": 0.09769780188798904, "global_step": 202114, "epoch": 2270, "lr": 1.530556053942684e-05}
{"train_loss": 0.02090073749423027, "global_step": 202115, "epoch": 2270, "lr": 1.5305143083760897e-05}
{"train_loss": 0.0612340047955513, "global_step": 202116, "epoch": 2270, "lr": 1.5304725632759244e-05}
{"train_loss": 0.04436425492167473, "global_step": 202117, "epoch": 2270, "lr": 1.530430818642196e-05}
{"train_loss": 0.06766978017148677, "global_step": 202118, "epoch": 2270, "lr": 1.5303890744749077e-05, "val_loss": 8.296748161315918, "train_action_mse_error": 12.970574378967285}
{"train_loss": 0.11158087104558945, "global_step": 202119, "epoch": 2271, "lr": 1.530347330774068e-05}
{"train_loss": 0.04731924459338188, "global_step": 202120, "epoch": 2271, "lr": 1.5303055875396794e-05}
{"train_loss": 0.10565902292728424, "global_step": 202121, "epoch": 2271, "lr": 1.5302638447717498e-05}
{"train_loss": 0.04593070223927498, "global_step": 202122, "epoch": 2271, "lr": 1.530222102470283e-05}
{"train_loss": 0.07630674540996552, "global_step": 202123, "epoch": 2271, "lr": 1.5301803606352867e-05}
{"train_loss": 0.06116306409239769, "global_step": 202124, "epoch": 2271, "lr": 1.530138619266764e-05}
{"train_loss": 0.07063398510217667, "global_step": 202125, "epoch": 2271, "lr": 1.5300968783647212e-05}
{"train_loss": 0.10826608538627625, "global_step": 202126, "epoch": 2271, "lr": 1.5300551379291674e-05}
{"train_loss": 0.03630828857421875, "global_step": 202127, "epoch": 2271, "lr": 1.5300133979601027e-05}
{"train_loss": 0.04152119532227516, "global_step": 202128, "epoch": 2271, "lr": 1.5299716584575373e-05}
{"train_loss": 0.056105706840753555, "global_step": 202129, "epoch": 2271, "lr": 1.529929919421474e-05}
{"train_loss": 0.114161916077137, "global_step": 202130, "epoch": 2271, "lr": 1.52988818085192e-05}
{"train_loss": 0.11109747737646103, "global_step": 202131, "epoch": 2271, "lr": 1.529846442748879e-05}
{"train_loss": 0.09859773516654968, "global_step": 202132, "epoch": 2271, "lr": 1.5298047051123592e-05}
{"train_loss": 0.05660320445895195, "global_step": 202133, "epoch": 2271, "lr": 1.5297629679423636e-05}
{"train_loss": 0.09436054527759552, "global_step": 202134, "epoch": 2271, "lr": 1.529721231238901e-05}
{"train_loss": 0.07226380705833435, "global_step": 202135, "epoch": 2271, "lr": 1.529679495001973e-05}
{"train_loss": 0.05644753947854042, "global_step": 202136, "epoch": 2271, "lr": 1.529637759231589e-05}
{"train_loss": 0.03832545131444931, "global_step": 202137, "epoch": 2271, "lr": 1.5295960239277512e-05}
{"train_loss": 0.06795503944158554, "global_step": 202138, "epoch": 2271, "lr": 1.5295542890904684e-05}
{"train_loss": 0.05426385998725891, "global_step": 202139, "epoch": 2271, "lr": 1.5295125547197437e-05}
{"train_loss": 0.0725400373339653, "global_step": 202140, "epoch": 2271, "lr": 1.529470820815585e-05}
{"train_loss": 0.07430475205183029, "global_step": 202141, "epoch": 2271, "lr": 1.5294290873779947e-05}
{"train_loss": 0.0653141513466835, "global_step": 202142, "epoch": 2271, "lr": 1.5293873544069827e-05}
{"train_loss": 0.05158752202987671, "global_step": 202143, "epoch": 2271, "lr": 1.52934562190255e-05}
{"train_loss": 0.04794862121343613, "global_step": 202144, "epoch": 2271, "lr": 1.5293038898647067e-05}
{"train_loss": 0.10948436707258224, "global_step": 202145, "epoch": 2271, "lr": 1.529262158293454e-05}
{"train_loss": 0.057949721813201904, "global_step": 202146, "epoch": 2271, "lr": 1.5292204271888018e-05}
{"train_loss": 0.06040957570075989, "global_step": 202147, "epoch": 2271, "lr": 1.5291786965507517e-05}
{"train_loss": 0.035815950483083725, "global_step": 202148, "epoch": 2271, "lr": 1.5291369663793126e-05}
{"train_loss": 0.06297861784696579, "global_step": 202149, "epoch": 2271, "lr": 1.529095236674489e-05}
{"train_loss": 0.060478512197732925, "global_step": 202150, "epoch": 2271, "lr": 1.5290535074362843e-05}
{"train_loss": 0.03884795680642128, "global_step": 202151, "epoch": 2271, "lr": 1.529011778664708e-05}
{"train_loss": 0.06799793988466263, "global_step": 202152, "epoch": 2271, "lr": 1.5289700503597622e-05}
{"train_loss": 0.06982304155826569, "global_step": 202153, "epoch": 2271, "lr": 1.5289283225214558e-05}
{"train_loss": 0.04292503371834755, "global_step": 202154, "epoch": 2271, "lr": 1.528886595149791e-05}
{"train_loss": 0.12474770098924637, "global_step": 202155, "epoch": 2271, "lr": 1.5288448682447763e-05}
{"train_loss": 0.04417824372649193, "global_step": 202156, "epoch": 2271, "lr": 1.5288031418064148e-05}
{"train_loss": 0.06303144246339798, "global_step": 202157, "epoch": 2271, "lr": 1.5287614158347152e-05}
{"train_loss": 0.06368043273687363, "global_step": 202158, "epoch": 2271, "lr": 1.528719690329679e-05}
{"train_loss": 0.05140203982591629, "global_step": 202159, "epoch": 2271, "lr": 1.5286779652913162e-05}
{"train_loss": 0.08362782001495361, "global_step": 202160, "epoch": 2271, "lr": 1.5286362407196293e-05}
{"train_loss": 0.131610706448555, "global_step": 202161, "epoch": 2271, "lr": 1.5285945166146243e-05}
{"train_loss": 0.0812619999051094, "global_step": 202162, "epoch": 2271, "lr": 1.5285527929763094e-05}
{"train_loss": 0.06773806363344193, "global_step": 202163, "epoch": 2271, "lr": 1.5285110698046863e-05}
{"train_loss": 0.06816942244768143, "global_step": 202164, "epoch": 2271, "lr": 1.528469347099764e-05}
{"train_loss": 0.06267403811216354, "global_step": 202165, "epoch": 2271, "lr": 1.528427624861546e-05}
{"train_loss": 0.06755658239126205, "global_step": 202166, "epoch": 2271, "lr": 1.5283859030900394e-05}
{"train_loss": 0.12695829570293427, "global_step": 202167, "epoch": 2271, "lr": 1.528344181785247e-05}
{"train_loss": 0.06711618602275848, "global_step": 202168, "epoch": 2271, "lr": 1.528302460947179e-05}
{"train_loss": 0.04196872189640999, "global_step": 202169, "epoch": 2271, "lr": 1.5282607405758364e-05}
{"train_loss": 0.12913388013839722, "global_step": 202170, "epoch": 2271, "lr": 1.528219020671229e-05}
{"train_loss": 0.06376871466636658, "global_step": 202171, "epoch": 2271, "lr": 1.528177301233358e-05}
{"train_loss": 0.05869697406888008, "global_step": 202172, "epoch": 2271, "lr": 1.5281355822622324e-05}
{"train_loss": 0.011570374481379986, "global_step": 202173, "epoch": 2271, "lr": 1.5280938637578557e-05}
{"train_loss": 0.08652376383543015, "global_step": 202174, "epoch": 2271, "lr": 1.5280521457202362e-05}
{"train_loss": 0.08192145824432373, "global_step": 202175, "epoch": 2271, "lr": 1.5280104281493755e-05}
{"train_loss": 0.0567665658891201, "global_step": 202176, "epoch": 2271, "lr": 1.527968711045284e-05}
{"train_loss": 0.041675105690956116, "global_step": 202177, "epoch": 2271, "lr": 1.5279269944079626e-05}
{"train_loss": 0.06158362329006195, "global_step": 202178, "epoch": 2271, "lr": 1.527885278237421e-05}
{"train_loss": 0.046448804438114166, "global_step": 202179, "epoch": 2271, "lr": 1.5278435625336607e-05}
{"train_loss": 0.04042607173323631, "global_step": 202180, "epoch": 2271, "lr": 1.52780184729669e-05}
{"train_loss": 0.0688345804810524, "global_step": 202181, "epoch": 2271, "lr": 1.5277601325265155e-05}
{"train_loss": 0.023990754038095474, "global_step": 202182, "epoch": 2271, "lr": 1.5277184182231396e-05}
{"train_loss": 0.05365762487053871, "global_step": 202183, "epoch": 2271, "lr": 1.5276767043865714e-05}
{"train_loss": 0.11111431568861008, "global_step": 202184, "epoch": 2271, "lr": 1.5276349910168146e-05}
{"train_loss": 0.11066024005413055, "global_step": 202185, "epoch": 2271, "lr": 1.5275932781138742e-05}
{"train_loss": 0.043750565499067307, "global_step": 202186, "epoch": 2271, "lr": 1.527551565677757e-05}
{"train_loss": 0.051203835755586624, "global_step": 202187, "epoch": 2271, "lr": 1.527509853708467e-05}
{"train_loss": 0.0825575739145279, "global_step": 202188, "epoch": 2271, "lr": 1.527468142206013e-05}
{"train_loss": 0.05998046696186066, "global_step": 202189, "epoch": 2271, "lr": 1.527426431170398e-05}
{"train_loss": 0.0484585165977478, "global_step": 202190, "epoch": 2271, "lr": 1.5273847206016266e-05}
{"train_loss": 0.07333523035049438, "global_step": 202191, "epoch": 2271, "lr": 1.5273430104997076e-05}
{"train_loss": 0.04880569130182266, "global_step": 202192, "epoch": 2271, "lr": 1.527301300864644e-05}
{"train_loss": 0.11868429183959961, "global_step": 202193, "epoch": 2271, "lr": 1.5272595916964434e-05}
{"train_loss": 0.11489081382751465, "global_step": 202194, "epoch": 2271, "lr": 1.527217882995109e-05}
{"train_loss": 0.09870562702417374, "global_step": 202195, "epoch": 2271, "lr": 1.5271761747606494e-05}
{"train_loss": 0.05557020381093025, "global_step": 202196, "epoch": 2271, "lr": 1.527134466993067e-05}
{"train_loss": 0.06079917028546333, "global_step": 202197, "epoch": 2271, "lr": 1.5270927596923695e-05}
{"train_loss": 0.09149790555238724, "global_step": 202198, "epoch": 2271, "lr": 1.5270510528585628e-05}
{"train_loss": 0.05226863920688629, "global_step": 202199, "epoch": 2271, "lr": 1.5270093464916512e-05}
{"train_loss": 0.04423690587282181, "global_step": 202200, "epoch": 2271, "lr": 1.5269676405916418e-05}
{"train_loss": 0.08125003427267075, "global_step": 202201, "epoch": 2271, "lr": 1.526925935158538e-05}
{"train_loss": 0.0955401286482811, "global_step": 202202, "epoch": 2271, "lr": 1.526884230192348e-05}
{"train_loss": 0.03272794559597969, "global_step": 202203, "epoch": 2271, "lr": 1.526842525693074e-05}
{"train_loss": 0.06985713541507721, "global_step": 202204, "epoch": 2271, "lr": 1.526800821660726e-05}
{"train_loss": 0.07254017889499664, "global_step": 202205, "epoch": 2271, "lr": 1.5267591180953056e-05}
{"train_loss": 0.08188341557979584, "global_step": 202206, "epoch": 2271, "lr": 1.5267174149968217e-05}
{"train_loss": 0.06999918373741126, "global_step": 202207, "epoch": 2271, "lr": 1.5266757123652763e-05, "val_loss": 8.215676307678223}
{"train_loss": 0.047495726495981216, "global_step": 202208, "epoch": 2272, "lr": 1.5266340102006795e-05}
{"train_loss": 0.12466757744550705, "global_step": 202209, "epoch": 2272, "lr": 1.526592308503032e-05}
{"train_loss": 0.06736131012439728, "global_step": 202210, "epoch": 2272, "lr": 1.526550607272344e-05}
{"train_loss": 0.0330825001001358, "global_step": 202211, "epoch": 2272, "lr": 1.5265089065086168e-05}
{"train_loss": 0.09932415932416916, "global_step": 202212, "epoch": 2272, "lr": 1.52646720621186e-05}
{"train_loss": 0.025294849649071693, "global_step": 202213, "epoch": 2272, "lr": 1.526425506382076e-05}
{"train_loss": 0.028073687106370926, "global_step": 202214, "epoch": 2272, "lr": 1.5263838070192728e-05}
{"train_loss": 0.030698636546730995, "global_step": 202215, "epoch": 2272, "lr": 1.5263421081234537e-05}
{"train_loss": 0.029045701026916504, "global_step": 202216, "epoch": 2272, "lr": 1.5263004096946255e-05}
{"train_loss": 0.07448258996009827, "global_step": 202217, "epoch": 2272, "lr": 1.5262587117327954e-05}
{"train_loss": 0.046405162662267685, "global_step": 202218, "epoch": 2272, "lr": 1.526217014237966e-05}
{"train_loss": 0.07125032693147659, "global_step": 202219, "epoch": 2272, "lr": 1.526175317210146e-05}
{"train_loss": 0.10976850986480713, "global_step": 202220, "epoch": 2272, "lr": 1.5261336206493376e-05}
{"train_loss": 0.03697985038161278, "global_step": 202221, "epoch": 2272, "lr": 1.52609192455555e-05}
{"train_loss": 0.03796517103910446, "global_step": 202222, "epoch": 2272, "lr": 1.526050228928787e-05}
{"train_loss": 0.07207685708999634, "global_step": 202223, "epoch": 2272, "lr": 1.526008533769052e-05}
{"train_loss": 0.10917624831199646, "global_step": 202224, "epoch": 2272, "lr": 1.5259668390763554e-05}
{"train_loss": 0.04897189885377884, "global_step": 202225, "epoch": 2272, "lr": 1.5259251448506982e-05}
{"train_loss": 0.04099005460739136, "global_step": 202226, "epoch": 2272, "lr": 1.5258834510920895e-05}
{"train_loss": 0.09591402113437653, "global_step": 202227, "epoch": 2272, "lr": 1.5258417578005335e-05}
{"train_loss": 0.1028352752327919, "global_step": 202228, "epoch": 2272, "lr": 1.525800064976034e-05}
{"train_loss": 0.052813198417425156, "global_step": 202229, "epoch": 2272, "lr": 1.5257583726186004e-05}
{"train_loss": 0.11181004345417023, "global_step": 202230, "epoch": 2272, "lr": 1.5257166807282342e-05}
{"train_loss": 0.15013152360916138, "global_step": 202231, "epoch": 2272, "lr": 1.5256749893049449e-05}
{"train_loss": 0.05533847585320473, "global_step": 202232, "epoch": 2272, "lr": 1.5256332983487342e-05}
{"train_loss": 0.07331030815839767, "global_step": 202233, "epoch": 2272, "lr": 1.52559160785961e-05}
{"train_loss": 0.10738641768693924, "global_step": 202234, "epoch": 2272, "lr": 1.5255499178375793e-05}
{"train_loss": 0.08582227677106857, "global_step": 202235, "epoch": 2272, "lr": 1.5255082282826444e-05}
{"train_loss": 0.07146012783050537, "global_step": 202236, "epoch": 2272, "lr": 1.5254665391948148e-05}
{"train_loss": 0.05841333791613579, "global_step": 202237, "epoch": 2272, "lr": 1.5254248505740915e-05}
{"train_loss": 0.07381341606378555, "global_step": 202238, "epoch": 2272, "lr": 1.5253831624204845e-05}
{"train_loss": 0.13596685230731964, "global_step": 202239, "epoch": 2272, "lr": 1.5253414747339956e-05}
{"train_loss": 0.04950341582298279, "global_step": 202240, "epoch": 2272, "lr": 1.5252997875146335e-05}
{"train_loss": 0.08431006222963333, "global_step": 202241, "epoch": 2272, "lr": 1.5252581007624011e-05}
{"train_loss": 0.08398322761058807, "global_step": 202242, "epoch": 2272, "lr": 1.5252164144773073e-05}
{"train_loss": 0.04251088574528694, "global_step": 202243, "epoch": 2272, "lr": 1.5251747286593537e-05}
{"train_loss": 0.07334244251251221, "global_step": 202244, "epoch": 2272, "lr": 1.5251330433085503e-05}
{"train_loss": 0.028527624905109406, "global_step": 202245, "epoch": 2272, "lr": 1.5250913584248982e-05}
{"train_loss": 0.09473995864391327, "global_step": 202246, "epoch": 2272, "lr": 1.525049674008407e-05}
{"train_loss": 0.08985290676355362, "global_step": 202247, "epoch": 2272, "lr": 1.525007990059079e-05}
{"train_loss": 0.07312048971652985, "global_step": 202248, "epoch": 2272, "lr": 1.5249663065769226e-05}
{"train_loss": 0.056569039821624756, "global_step": 202249, "epoch": 2272, "lr": 1.5249246235619408e-05}
{"train_loss": 0.06311056017875671, "global_step": 202250, "epoch": 2272, "lr": 1.5248829410141423e-05}
{"train_loss": 0.05440620705485344, "global_step": 202251, "epoch": 2272, "lr": 1.5248412589335293e-05}
{"train_loss": 0.04076920449733734, "global_step": 202252, "epoch": 2272, "lr": 1.5247995773201085e-05}
{"train_loss": 0.11262603849172592, "global_step": 202253, "epoch": 2272, "lr": 1.5247578961738884e-05}
{"train_loss": 0.05093222111463547, "global_step": 202254, "epoch": 2272, "lr": 1.5247162154948702e-05}
{"train_loss": 0.04690560698509216, "global_step": 202255, "epoch": 2272, "lr": 1.5246745352830633e-05}
{"train_loss": 0.06601005792617798, "global_step": 202256, "epoch": 2272, "lr": 1.5246328555384697e-05}
{"train_loss": 0.029865093529224396, "global_step": 202257, "epoch": 2272, "lr": 1.5245911762610987e-05}
{"train_loss": 0.03908088803291321, "global_step": 202258, "epoch": 2272, "lr": 1.5245494974509522e-05}
{"train_loss": 0.051704902201890945, "global_step": 202259, "epoch": 2272, "lr": 1.5245078191080391e-05}
{"train_loss": 0.04346175864338875, "global_step": 202260, "epoch": 2272, "lr": 1.5244661412323641e-05}
{"train_loss": 0.04125262051820755, "global_step": 202261, "epoch": 2272, "lr": 1.5244244638239308e-05}
{"train_loss": 0.044988881796598434, "global_step": 202262, "epoch": 2272, "lr": 1.5243827868827476e-05}
{"train_loss": 0.07307419925928116, "global_step": 202263, "epoch": 2272, "lr": 1.5243411104088168e-05}
{"train_loss": 0.051374055445194244, "global_step": 202264, "epoch": 2272, "lr": 1.5242994344021478e-05}
{"train_loss": 0.0686502680182457, "global_step": 202265, "epoch": 2272, "lr": 1.5242577588627443e-05}
{"train_loss": 0.061520688235759735, "global_step": 202266, "epoch": 2272, "lr": 1.5242160837906105e-05}
{"train_loss": 0.09707418829202652, "global_step": 202267, "epoch": 2272, "lr": 1.5241744091857552e-05}
{"train_loss": 0.05614122748374939, "global_step": 202268, "epoch": 2272, "lr": 1.524132735048181e-05}
{"train_loss": 0.08987016975879669, "global_step": 202269, "epoch": 2272, "lr": 1.5240910613778958e-05}
{"train_loss": 0.08659825474023819, "global_step": 202270, "epoch": 2272, "lr": 1.524049388174903e-05}
{"train_loss": 0.0665399432182312, "global_step": 202271, "epoch": 2272, "lr": 1.524007715439209e-05}
{"train_loss": 0.0239021647721529, "global_step": 202272, "epoch": 2272, "lr": 1.5239660431708214e-05}
{"train_loss": 0.04058003053069115, "global_step": 202273, "epoch": 2272, "lr": 1.5239243713697426e-05}
{"train_loss": 0.06206221505999565, "global_step": 202274, "epoch": 2272, "lr": 1.5238827000359817e-05}
{"train_loss": 0.06671198457479477, "global_step": 202275, "epoch": 2272, "lr": 1.523841029169541e-05}
{"train_loss": 0.09221158921718597, "global_step": 202276, "epoch": 2272, "lr": 1.5237993587704286e-05}
{"train_loss": 0.04643720015883446, "global_step": 202277, "epoch": 2272, "lr": 1.5237576888386479e-05}
{"train_loss": 0.08916259557008743, "global_step": 202278, "epoch": 2272, "lr": 1.5237160193742067e-05}
{"train_loss": 0.07777529209852219, "global_step": 202279, "epoch": 2272, "lr": 1.5236743503771079e-05}
{"train_loss": 0.06369739025831223, "global_step": 202280, "epoch": 2272, "lr": 1.5236326818473606e-05}
{"train_loss": 0.048428114503622055, "global_step": 202281, "epoch": 2272, "lr": 1.5235910137849669e-05}
{"train_loss": 0.05919172614812851, "global_step": 202282, "epoch": 2272, "lr": 1.5235493461899358e-05}
{"train_loss": 0.06721645593643188, "global_step": 202283, "epoch": 2272, "lr": 1.5235076790622688e-05}
{"train_loss": 0.0358063243329525, "global_step": 202284, "epoch": 2272, "lr": 1.5234660124019761e-05}
{"train_loss": 0.09182797372341156, "global_step": 202285, "epoch": 2272, "lr": 1.5234243462090591e-05}
{"train_loss": 0.09653910994529724, "global_step": 202286, "epoch": 2272, "lr": 1.5233826804835271e-05}
{"train_loss": 0.05510738492012024, "global_step": 202287, "epoch": 2272, "lr": 1.5233410152253825e-05}
{"train_loss": 0.056187499314546585, "global_step": 202288, "epoch": 2272, "lr": 1.5232993504346321e-05}
{"train_loss": 0.05414671078324318, "global_step": 202289, "epoch": 2272, "lr": 1.523257686111283e-05}
{"train_loss": 0.06744140386581421, "global_step": 202290, "epoch": 2272, "lr": 1.5232160222553382e-05}
{"train_loss": 0.032182466238737106, "global_step": 202291, "epoch": 2272, "lr": 1.5231743588668062e-05}
{"train_loss": 0.037032727152109146, "global_step": 202292, "epoch": 2272, "lr": 1.5231326959456898e-05}
{"train_loss": 0.035916656255722046, "global_step": 202293, "epoch": 2272, "lr": 1.5230910334919974e-05}
{"train_loss": 0.03807561844587326, "global_step": 202294, "epoch": 2272, "lr": 1.523049371505731e-05}
{"train_loss": 0.04005599766969681, "global_step": 202295, "epoch": 2272, "lr": 1.5230077099869e-05}
{"train_loss": 0.06480411258055253, "global_step": 202296, "epoch": 2272, "lr": 1.5229660489355064e-05, "val_loss": 8.204010009765625}
{"train_loss": 0.08064810186624527, "global_step": 202297, "epoch": 2273, "lr": 1.5229243883515599e-05}
{"train_loss": 0.0988365039229393, "global_step": 202298, "epoch": 2273, "lr": 1.5228827282350616e-05}
{"train_loss": 0.05729968100786209, "global_step": 202299, "epoch": 2273, "lr": 1.5228410685860211e-05}
{"train_loss": 0.14420273900032043, "global_step": 202300, "epoch": 2273, "lr": 1.5227994094044424e-05}
{"train_loss": 0.03668907284736633, "global_step": 202301, "epoch": 2273, "lr": 1.522757750690329e-05}
{"train_loss": 0.08981011807918549, "global_step": 202302, "epoch": 2273, "lr": 1.5227160924436901e-05}
{"train_loss": 0.024157391861081123, "global_step": 202303, "epoch": 2273, "lr": 1.5226744346645282e-05}
{"train_loss": 0.08364623785018921, "global_step": 202304, "epoch": 2273, "lr": 1.522632777352852e-05}
{"train_loss": 0.053921595215797424, "global_step": 202305, "epoch": 2273, "lr": 1.5225911205086657e-05}
{"train_loss": 0.0482589527964592, "global_step": 202306, "epoch": 2273, "lr": 1.5225494641319726e-05}
{"train_loss": 0.10071481764316559, "global_step": 202307, "epoch": 2273, "lr": 1.5225078082227806e-05}
{"train_loss": 0.04055418819189072, "global_step": 202308, "epoch": 2273, "lr": 1.5224661527810963e-05}
{"train_loss": 0.023658648133277893, "global_step": 202309, "epoch": 2273, "lr": 1.5224244978069225e-05}
{"train_loss": 0.037172116339206696, "global_step": 202310, "epoch": 2273, "lr": 1.5223828433002685e-05}
{"train_loss": 0.05872786045074463, "global_step": 202311, "epoch": 2273, "lr": 1.5223411892611356e-05}
{"train_loss": 0.04493015259504318, "global_step": 202312, "epoch": 2273, "lr": 1.5222995356895331e-05}
{"train_loss": 0.09261687844991684, "global_step": 202313, "epoch": 2273, "lr": 1.5222578825854639e-05}
{"train_loss": 0.0862966924905777, "global_step": 202314, "epoch": 2273, "lr": 1.5222162299489357e-05}
{"train_loss": 0.03575742989778519, "global_step": 202315, "epoch": 2273, "lr": 1.5221745777799518e-05}
{"train_loss": 0.04434789717197418, "global_step": 202316, "epoch": 2273, "lr": 1.522132926078521e-05}
{"train_loss": 0.10152341425418854, "global_step": 202317, "epoch": 2273, "lr": 1.5220912748446454e-05}
{"train_loss": 0.07401279360055923, "global_step": 202318, "epoch": 2273, "lr": 1.5220496240783334e-05}
{"train_loss": 0.0495651550590992, "global_step": 202319, "epoch": 2273, "lr": 1.5220079737795883e-05}
{"train_loss": 0.060966506600379944, "global_step": 202320, "epoch": 2273, "lr": 1.5219663239484183e-05}
{"train_loss": 0.02643187902867794, "global_step": 202321, "epoch": 2273, "lr": 1.5219246745848259e-05}
{"train_loss": 0.03602081909775734, "global_step": 202322, "epoch": 2273, "lr": 1.52188302568882e-05}
{"train_loss": 0.08542462438344955, "global_step": 202323, "epoch": 2273, "lr": 1.5218413772604029e-05}
{"train_loss": 0.053787607699632645, "global_step": 202324, "epoch": 2273, "lr": 1.5217997292995817e-05}
{"train_loss": 0.06855480372905731, "global_step": 202325, "epoch": 2273, "lr": 1.5217580818063636e-05}
{"train_loss": 0.036458611488342285, "global_step": 202326, "epoch": 2273, "lr": 1.5217164347807517e-05}
{"train_loss": 0.07355980575084686, "global_step": 202327, "epoch": 2273, "lr": 1.5216747882227539e-05}
{"train_loss": 0.07392475008964539, "global_step": 202328, "epoch": 2273, "lr": 1.5216331421323732e-05}
{"train_loss": 0.06001456081867218, "global_step": 202329, "epoch": 2273, "lr": 1.5215914965096178e-05}
{"train_loss": 0.061746079474687576, "global_step": 202330, "epoch": 2273, "lr": 1.5215498513544902e-05}
{"train_loss": 0.035540301352739334, "global_step": 202331, "epoch": 2273, "lr": 1.521508206667e-05}
{"train_loss": 0.0693400502204895, "global_step": 202332, "epoch": 2273, "lr": 1.5214665624471486e-05}
{"train_loss": 0.08592411875724792, "global_step": 202333, "epoch": 2273, "lr": 1.5214249186949453e-05}
{"train_loss": 0.046953313052654266, "global_step": 202334, "epoch": 2273, "lr": 1.5213832754103923e-05}
{"train_loss": 0.07578194886445999, "global_step": 202335, "epoch": 2273, "lr": 1.5213416325934988e-05}
{"train_loss": 0.08648218214511871, "global_step": 202336, "epoch": 2273, "lr": 1.521299990244267e-05}
{"train_loss": 0.08606772124767303, "global_step": 202337, "epoch": 2273, "lr": 1.5212583483627052e-05}
{"train_loss": 0.022287487983703613, "global_step": 202338, "epoch": 2273, "lr": 1.5212167069488182e-05}
{"train_loss": 0.0838099867105484, "global_step": 202339, "epoch": 2273, "lr": 1.5211750660026098e-05}
{"train_loss": 0.07246746122837067, "global_step": 202340, "epoch": 2273, "lr": 1.5211334255240877e-05}
{"train_loss": 0.027586283162236214, "global_step": 202341, "epoch": 2273, "lr": 1.5210917855132561e-05}
{"train_loss": 0.05139924958348274, "global_step": 202342, "epoch": 2273, "lr": 1.5210501459701227e-05}
{"train_loss": 0.05454382672905922, "global_step": 202343, "epoch": 2273, "lr": 1.52100850689469e-05}
{"train_loss": 0.08679432421922684, "global_step": 202344, "epoch": 2273, "lr": 1.5209668682869672e-05}
{"train_loss": 0.06420828402042389, "global_step": 202345, "epoch": 2273, "lr": 1.5209252301469557e-05}
{"train_loss": 0.049750350415706635, "global_step": 202346, "epoch": 2273, "lr": 1.5208835924746656e-05}
{"train_loss": 0.08031420409679413, "global_step": 202347, "epoch": 2273, "lr": 1.5208419552700987e-05}
{"train_loss": 0.05493322014808655, "global_step": 202348, "epoch": 2273, "lr": 1.520800318533263e-05}
{"train_loss": 0.08337413519620895, "global_step": 202349, "epoch": 2273, "lr": 1.5207586822641629e-05}
{"train_loss": 0.05376579985022545, "global_step": 202350, "epoch": 2273, "lr": 1.5207170464628051e-05}
{"train_loss": 0.01974499598145485, "global_step": 202351, "epoch": 2273, "lr": 1.5206754111291932e-05}
{"train_loss": 0.05417824536561966, "global_step": 202352, "epoch": 2273, "lr": 1.5206337762633354e-05}
{"train_loss": 0.06547155976295471, "global_step": 202353, "epoch": 2273, "lr": 1.5205921418652347e-05}
{"train_loss": 0.06948073208332062, "global_step": 202354, "epoch": 2273, "lr": 1.5205505079348997e-05}
{"train_loss": 0.051331985741853714, "global_step": 202355, "epoch": 2273, "lr": 1.5205088744723322e-05}
{"train_loss": 0.06129984185099602, "global_step": 202356, "epoch": 2273, "lr": 1.5204672414775423e-05}
{"train_loss": 0.17146235704421997, "global_step": 202357, "epoch": 2273, "lr": 1.5204256089505308e-05}
{"train_loss": 0.045492079108953476, "global_step": 202358, "epoch": 2273, "lr": 1.5203839768913076e-05}
{"train_loss": 0.09974822402000427, "global_step": 202359, "epoch": 2273, "lr": 1.5203423452998743e-05}
{"train_loss": 0.025817859917879105, "global_step": 202360, "epoch": 2273, "lr": 1.5203007141762393e-05}
{"train_loss": 0.07592952251434326, "global_step": 202361, "epoch": 2273, "lr": 1.5202590835204083e-05}
{"train_loss": 0.061865080147981644, "global_step": 202362, "epoch": 2273, "lr": 1.5202174533323848e-05}
{"train_loss": 0.07727094739675522, "global_step": 202363, "epoch": 2273, "lr": 1.5201758236121777e-05}
{"train_loss": 0.10115041583776474, "global_step": 202364, "epoch": 2273, "lr": 1.5201341943597885e-05}
{"train_loss": 0.06086049601435661, "global_step": 202365, "epoch": 2273, "lr": 1.5200925655752263e-05}
{"train_loss": 0.021780364215373993, "global_step": 202366, "epoch": 2273, "lr": 1.5200509372584937e-05}
{"train_loss": 0.05411892756819725, "global_step": 202367, "epoch": 2273, "lr": 1.5200093094095997e-05}
{"train_loss": 0.05620915815234184, "global_step": 202368, "epoch": 2273, "lr": 1.5199676820285463e-05}
{"train_loss": 0.061257392168045044, "global_step": 202369, "epoch": 2273, "lr": 1.5199260551153427e-05}
{"train_loss": 0.06907614320516586, "global_step": 202370, "epoch": 2273, "lr": 1.5198844286699903e-05}
{"train_loss": 0.05397501960396767, "global_step": 202371, "epoch": 2273, "lr": 1.5198428026924994e-05}
{"train_loss": 0.09144660830497742, "global_step": 202372, "epoch": 2273, "lr": 1.5198011771828713e-05}
{"train_loss": 0.10307332873344421, "global_step": 202373, "epoch": 2273, "lr": 1.519759552141115e-05}
{"train_loss": 0.055764228105545044, "global_step": 202374, "epoch": 2273, "lr": 1.519717927567233e-05}
{"train_loss": 0.08079703897237778, "global_step": 202375, "epoch": 2273, "lr": 1.5196763034612343e-05}
{"train_loss": 0.030759071931242943, "global_step": 202376, "epoch": 2273, "lr": 1.5196346798231225e-05}
{"train_loss": 0.0752217099070549, "global_step": 202377, "epoch": 2273, "lr": 1.519593056652902e-05}
{"train_loss": 0.06917805224657059, "global_step": 202378, "epoch": 2273, "lr": 1.5195514339505812e-05}
{"train_loss": 0.06113826856017113, "global_step": 202379, "epoch": 2273, "lr": 1.5195098117161627e-05}
{"train_loss": 0.07687828689813614, "global_step": 202380, "epoch": 2273, "lr": 1.5194681899496555e-05}
{"train_loss": 0.05806152895092964, "global_step": 202381, "epoch": 2273, "lr": 1.5194265686510617e-05}
{"train_loss": 0.06983454525470734, "global_step": 202382, "epoch": 2273, "lr": 1.5193849478203903e-05}
{"train_loss": 0.050520021468400955, "global_step": 202383, "epoch": 2273, "lr": 1.5193433274576436e-05}
{"train_loss": 0.08503303676843643, "global_step": 202384, "epoch": 2273, "lr": 1.5193017075628302e-05}
{"train_loss": 0.06485119164827165, "global_step": 202385, "epoch": 2273, "lr": 1.5192600881359526e-05, "val_loss": 8.101285934448242}
{"train_loss": 0.052883826196193695, "global_step": 202386, "epoch": 2274, "lr": 1.5192184691770195e-05}
{"train_loss": 0.0575326606631279, "global_step": 202387, "epoch": 2274, "lr": 1.5191768506860338e-05}
{"train_loss": 0.04569273814558983, "global_step": 202388, "epoch": 2274, "lr": 1.5191352326630037e-05}
{"train_loss": 0.0593746043741703, "global_step": 202389, "epoch": 2274, "lr": 1.5190936151079316e-05}
{"train_loss": 0.05196461081504822, "global_step": 202390, "epoch": 2274, "lr": 1.5190519980208273e-05}
{"train_loss": 0.04880727455019951, "global_step": 202391, "epoch": 2274, "lr": 1.5190103814016921e-05}
{"train_loss": 0.06952910125255585, "global_step": 202392, "epoch": 2274, "lr": 1.5189687652505347e-05}
{"train_loss": 0.07788810133934021, "global_step": 202393, "epoch": 2274, "lr": 1.5189271495673585e-05}
{"train_loss": 0.02108210325241089, "global_step": 202394, "epoch": 2274, "lr": 1.5188855343521713e-05}
{"train_loss": 0.054034579545259476, "global_step": 202395, "epoch": 2274, "lr": 1.5188439196049759e-05}
{"train_loss": 0.06380501389503479, "global_step": 202396, "epoch": 2274, "lr": 1.5188023053257815e-05}
{"train_loss": 0.14149600267410278, "global_step": 202397, "epoch": 2274, "lr": 1.5187606915145897e-05}
{"train_loss": 0.02291693352162838, "global_step": 202398, "epoch": 2274, "lr": 1.5187190781714083e-05}
{"train_loss": 0.04414485767483711, "global_step": 202399, "epoch": 2274, "lr": 1.5186774652962443e-05}
{"train_loss": 0.04281610995531082, "global_step": 202400, "epoch": 2274, "lr": 1.5186358528891004e-05}
{"train_loss": 0.0625523179769516, "global_step": 202401, "epoch": 2274, "lr": 1.5185942409499843e-05}
{"train_loss": 0.03300076723098755, "global_step": 202402, "epoch": 2274, "lr": 1.5185526294788999e-05}
{"train_loss": 0.044289182871580124, "global_step": 202403, "epoch": 2274, "lr": 1.518511018475855e-05}
{"train_loss": 0.08800792694091797, "global_step": 202404, "epoch": 2274, "lr": 1.5184694079408518e-05}
{"train_loss": 0.08357276022434235, "global_step": 202405, "epoch": 2274, "lr": 1.5184277978739003e-05}
{"train_loss": 0.052640099078416824, "global_step": 202406, "epoch": 2274, "lr": 1.5183861882750016e-05}
{"train_loss": 0.1108589917421341, "global_step": 202407, "epoch": 2274, "lr": 1.5183445791441652e-05}
{"train_loss": 0.041333701461553574, "global_step": 202408, "epoch": 2274, "lr": 1.518302970481394e-05}
{"train_loss": 0.04426342993974686, "global_step": 202409, "epoch": 2274, "lr": 1.5182613622866953e-05}
{"train_loss": 0.11370236426591873, "global_step": 202410, "epoch": 2274, "lr": 1.5182197545600723e-05}
{"train_loss": 0.06947366893291473, "global_step": 202411, "epoch": 2274, "lr": 1.5181781473015343e-05}
{"train_loss": 0.057908494025468826, "global_step": 202412, "epoch": 2274, "lr": 1.5181365405110832e-05}
{"train_loss": 0.054429806768894196, "global_step": 202413, "epoch": 2274, "lr": 1.5180949341887275e-05}
{"train_loss": 0.11253044009208679, "global_step": 202414, "epoch": 2274, "lr": 1.51805332833447e-05}
{"train_loss": 0.049609363079071045, "global_step": 202415, "epoch": 2274, "lr": 1.5180117229483193e-05}
{"train_loss": 0.04756064713001251, "global_step": 202416, "epoch": 2274, "lr": 1.5179701180302797e-05}
{"train_loss": 0.042114291340112686, "global_step": 202417, "epoch": 2274, "lr": 1.5179285135803545e-05}
{"train_loss": 0.06792467832565308, "global_step": 202418, "epoch": 2274, "lr": 1.5178869095985531e-05}
{"train_loss": 0.016280923038721085, "global_step": 202419, "epoch": 2274, "lr": 1.5178453060848785e-05}
{"train_loss": 0.08309460431337357, "global_step": 202420, "epoch": 2274, "lr": 1.5178037030393382e-05}
{"train_loss": 0.047147639095783234, "global_step": 202421, "epoch": 2274, "lr": 1.5177621004619353e-05}
{"train_loss": 0.12918151915073395, "global_step": 202422, "epoch": 2274, "lr": 1.5177204983526783e-05}
{"train_loss": 0.06537231057882309, "global_step": 202423, "epoch": 2274, "lr": 1.5176788967115696e-05}
{"train_loss": 0.050048667937517166, "global_step": 202424, "epoch": 2274, "lr": 1.517637295538618e-05}
{"train_loss": 0.05050517991185188, "global_step": 202425, "epoch": 2274, "lr": 1.5175956948338265e-05}
{"train_loss": 0.08098429441452026, "global_step": 202426, "epoch": 2274, "lr": 1.5175540945972033e-05}
{"train_loss": 0.06250424683094025, "global_step": 202427, "epoch": 2274, "lr": 1.5175124948287505e-05}
{"train_loss": 0.06953504681587219, "global_step": 202428, "epoch": 2274, "lr": 1.5174708955284777e-05}
{"train_loss": 0.06890775263309479, "global_step": 202429, "epoch": 2274, "lr": 1.517429296696387e-05}
{"train_loss": 0.06891856342554092, "global_step": 202430, "epoch": 2274, "lr": 1.517387698332487e-05}
{"train_loss": 0.05955789238214493, "global_step": 202431, "epoch": 2274, "lr": 1.5173461004367801e-05}
{"train_loss": 0.06487743556499481, "global_step": 202432, "epoch": 2274, "lr": 1.5173045030092753e-05}
{"train_loss": 0.0296500101685524, "global_step": 202433, "epoch": 2274, "lr": 1.5172629060499744e-05}
{"train_loss": 0.03759770840406418, "global_step": 202434, "epoch": 2274, "lr": 1.5172213095588855e-05}
{"train_loss": 0.05344701558351517, "global_step": 202435, "epoch": 2274, "lr": 1.517179713536015e-05}
{"train_loss": 0.08188547194004059, "global_step": 202436, "epoch": 2274, "lr": 1.517138117981366e-05}
{"train_loss": 0.037251994013786316, "global_step": 202437, "epoch": 2274, "lr": 1.5170965228949469e-05}
{"train_loss": 0.09603667259216309, "global_step": 202438, "epoch": 2274, "lr": 1.51705492827676e-05}
{"train_loss": 0.049407295882701874, "global_step": 202439, "epoch": 2274, "lr": 1.5170133341268144e-05}
{"train_loss": 0.03648100793361664, "global_step": 202440, "epoch": 2274, "lr": 1.5169717404451117e-05}
{"train_loss": 0.06376330554485321, "global_step": 202441, "epoch": 2274, "lr": 1.5169301472316622e-05}
{"train_loss": 0.04437356814742088, "global_step": 202442, "epoch": 2274, "lr": 1.5168885544864669e-05}
{"train_loss": 0.03442094847559929, "global_step": 202443, "epoch": 2274, "lr": 1.516846962209535e-05}
{"train_loss": 0.0692673847079277, "global_step": 202444, "epoch": 2274, "lr": 1.516805370400869e-05}
{"train_loss": 0.08333317935466766, "global_step": 202445, "epoch": 2274, "lr": 1.5167637790604778e-05}
{"train_loss": 0.06250393390655518, "global_step": 202446, "epoch": 2274, "lr": 1.5167221881883637e-05}
{"train_loss": 0.08031885325908661, "global_step": 202447, "epoch": 2274, "lr": 1.516680597784535e-05}
{"train_loss": 0.057071503251791, "global_step": 202448, "epoch": 2274, "lr": 1.5166390078489951e-05}
{"train_loss": 0.09840937703847885, "global_step": 202449, "epoch": 2274, "lr": 1.5165974183817516e-05}
{"train_loss": 0.025404203683137894, "global_step": 202450, "epoch": 2274, "lr": 1.516555829382808e-05}
{"train_loss": 0.08240816742181778, "global_step": 202451, "epoch": 2274, "lr": 1.5165142408521727e-05}
{"train_loss": 0.04832899942994118, "global_step": 202452, "epoch": 2274, "lr": 1.5164726527898477e-05}
{"train_loss": 0.08437161892652512, "global_step": 202453, "epoch": 2274, "lr": 1.5164310651958424e-05}
{"train_loss": 0.1529403179883957, "global_step": 202454, "epoch": 2274, "lr": 1.5163894780701605e-05}
{"train_loss": 0.061028964817523956, "global_step": 202455, "epoch": 2274, "lr": 1.5163478914128055e-05}
{"train_loss": 0.10021049529314041, "global_step": 202456, "epoch": 2274, "lr": 1.516306305223787e-05}
{"train_loss": 0.03832031786441803, "global_step": 202457, "epoch": 2274, "lr": 1.516264719503107e-05}
{"train_loss": 0.09912669658660889, "global_step": 202458, "epoch": 2274, "lr": 1.5162231342507744e-05}
{"train_loss": 0.0889834612607956, "global_step": 202459, "epoch": 2274, "lr": 1.5161815494667914e-05}
{"train_loss": 0.05922478809952736, "global_step": 202460, "epoch": 2274, "lr": 1.516139965151167e-05}
{"train_loss": 0.03469671308994293, "global_step": 202461, "epoch": 2274, "lr": 1.5160983813039036e-05}
{"train_loss": 0.08087745308876038, "global_step": 202462, "epoch": 2274, "lr": 1.5160567979250101e-05}
{"train_loss": 0.08081647008657455, "global_step": 202463, "epoch": 2274, "lr": 1.5160152150144885e-05}
{"train_loss": 0.04147323593497276, "global_step": 202464, "epoch": 2274, "lr": 1.5159736325723477e-05}
{"train_loss": 0.13498811423778534, "global_step": 202465, "epoch": 2274, "lr": 1.5159320505985902e-05}
{"train_loss": 0.024261105805635452, "global_step": 202466, "epoch": 2274, "lr": 1.5158904690932247e-05}
{"train_loss": 0.04033103212714195, "global_step": 202467, "epoch": 2274, "lr": 1.5158488880562538e-05}
{"train_loss": 0.044583454728126526, "global_step": 202468, "epoch": 2274, "lr": 1.5158073074876861e-05}
{"train_loss": 0.04806292802095413, "global_step": 202469, "epoch": 2274, "lr": 1.5157657273875242e-05}
{"train_loss": 0.015194297768175602, "global_step": 202470, "epoch": 2274, "lr": 1.5157241477557748e-05}
{"train_loss": 0.06615757197141647, "global_step": 202471, "epoch": 2274, "lr": 1.5156825685924459e-05}
{"train_loss": 0.06441012769937515, "global_step": 202472, "epoch": 2274, "lr": 1.5156409898975388e-05}
{"train_loss": 0.03697728365659714, "global_step": 202473, "epoch": 2274, "lr": 1.515599411671063e-05}
{"train_loss": 0.06295028110191728, "global_step": 202474, "epoch": 2274, "lr": 1.515557833913021e-05, "val_loss": 8.358654022216797}
{"train_loss": 0.07302530854940414, "global_step": 202475, "epoch": 2275, "lr": 1.5155162566234215e-05}
{"train_loss": 0.04567977041006088, "global_step": 202476, "epoch": 2275, "lr": 1.5154746798022667e-05}
{"train_loss": 0.0681733563542366, "global_step": 202477, "epoch": 2275, "lr": 1.5154331034495655e-05}
{"train_loss": 0.04452264681458473, "global_step": 202478, "epoch": 2275, "lr": 1.5153915275653202e-05}
{"train_loss": 0.03215278685092926, "global_step": 202479, "epoch": 2275, "lr": 1.5153499521495396e-05}
{"train_loss": 0.08028800040483475, "global_step": 202480, "epoch": 2275, "lr": 1.5153083772022258e-05}
{"train_loss": 0.10612880438566208, "global_step": 202481, "epoch": 2275, "lr": 1.5152668027233885e-05}
{"train_loss": 0.06452152878046036, "global_step": 202482, "epoch": 2275, "lr": 1.5152252287130292e-05}
{"train_loss": 0.060009460896253586, "global_step": 202483, "epoch": 2275, "lr": 1.5151836551711574e-05}
{"train_loss": 0.11275482922792435, "global_step": 202484, "epoch": 2275, "lr": 1.5151420820977746e-05}
{"train_loss": 0.061089590191841125, "global_step": 202485, "epoch": 2275, "lr": 1.51510050949289e-05}
{"train_loss": 0.09105757623910904, "global_step": 202486, "epoch": 2275, "lr": 1.515058937356506e-05}
{"train_loss": 0.0444827638566494, "global_step": 202487, "epoch": 2275, "lr": 1.5150173656886318e-05}
{"train_loss": 0.09096022695302963, "global_step": 202488, "epoch": 2275, "lr": 1.5149757944892696e-05}
{"train_loss": 0.07409133017063141, "global_step": 202489, "epoch": 2275, "lr": 1.514934223758428e-05}
{"train_loss": 0.04487220197916031, "global_step": 202490, "epoch": 2275, "lr": 1.5148926534961089e-05}
{"train_loss": 0.04723668098449707, "global_step": 202491, "epoch": 2275, "lr": 1.5148510837023216e-05}
{"train_loss": 0.07506847381591797, "global_step": 202492, "epoch": 2275, "lr": 1.5148095143770708e-05}
{"train_loss": 0.07348547875881195, "global_step": 202493, "epoch": 2275, "lr": 1.5147679455203595e-05}
{"train_loss": 0.05048808827996254, "global_step": 202494, "epoch": 2275, "lr": 1.5147263771321963e-05}
{"train_loss": 0.046894509345293045, "global_step": 202495, "epoch": 2275, "lr": 1.5146848092125843e-05}
{"train_loss": 0.051786236464977264, "global_step": 202496, "epoch": 2275, "lr": 1.5146432417615324e-05}
{"train_loss": 0.05470352992415428, "global_step": 202497, "epoch": 2275, "lr": 1.5146016747790426e-05}
{"train_loss": 0.04688107967376709, "global_step": 202498, "epoch": 2275, "lr": 1.5145601082651234e-05}
{"train_loss": 0.011493583209812641, "global_step": 202499, "epoch": 2275, "lr": 1.5145185422197778e-05}
{"train_loss": 0.06105726957321167, "global_step": 202500, "epoch": 2275, "lr": 1.5144769766430139e-05}
{"train_loss": 0.051892463117837906, "global_step": 202501, "epoch": 2275, "lr": 1.5144354115348346e-05}
{"train_loss": 0.02656993828713894, "global_step": 202502, "epoch": 2275, "lr": 1.514393846895249e-05}
{"train_loss": 0.09641478210687637, "global_step": 202503, "epoch": 2275, "lr": 1.5143522827242585e-05}
{"train_loss": 0.1277260035276413, "global_step": 202504, "epoch": 2275, "lr": 1.5143107190218724e-05}
{"train_loss": 0.0755305290222168, "global_step": 202505, "epoch": 2275, "lr": 1.514269155788094e-05}
{"train_loss": 0.07905282080173492, "global_step": 202506, "epoch": 2275, "lr": 1.5142275930229288e-05}
{"train_loss": 0.05988309532403946, "global_step": 202507, "epoch": 2275, "lr": 1.514186030726385e-05}
{"train_loss": 0.1153840720653534, "global_step": 202508, "epoch": 2275, "lr": 1.5141444688984652e-05}
{"train_loss": 0.047698576003313065, "global_step": 202509, "epoch": 2275, "lr": 1.5141029075391772e-05}
{"train_loss": 0.06399159133434296, "global_step": 202510, "epoch": 2275, "lr": 1.5140613466485243e-05}
{"train_loss": 0.07788330316543579, "global_step": 202511, "epoch": 2275, "lr": 1.5140197862265149e-05}
{"train_loss": 0.10182096809148788, "global_step": 202512, "epoch": 2275, "lr": 1.5139782262731517e-05}
{"train_loss": 0.07458439469337463, "global_step": 202513, "epoch": 2275, "lr": 1.513936666788443e-05}
{"train_loss": 0.04888198524713516, "global_step": 202514, "epoch": 2275, "lr": 1.5138951077723912e-05}
{"train_loss": 0.10205534845590591, "global_step": 202515, "epoch": 2275, "lr": 1.5138535492250055e-05}
{"train_loss": 0.0520305298268795, "global_step": 202516, "epoch": 2275, "lr": 1.5138119911462883e-05}
{"train_loss": 0.06460493803024292, "global_step": 202517, "epoch": 2275, "lr": 1.5137704335362479e-05}
{"train_loss": 0.08670215308666229, "global_step": 202518, "epoch": 2275, "lr": 1.513728876394887e-05}
{"train_loss": 0.03988979011774063, "global_step": 202519, "epoch": 2275, "lr": 1.5136873197222145e-05}
{"train_loss": 0.05882007256150246, "global_step": 202520, "epoch": 2275, "lr": 1.5136457635182327e-05}
{"train_loss": 0.04901982098817825, "global_step": 202521, "epoch": 2275, "lr": 1.5136042077829504e-05}
{"train_loss": 0.04013291373848915, "global_step": 202522, "epoch": 2275, "lr": 1.51356265251637e-05}
{"train_loss": 0.08528749644756317, "global_step": 202523, "epoch": 2275, "lr": 1.5135210977184994e-05}
{"train_loss": 0.040491003543138504, "global_step": 202524, "epoch": 2275, "lr": 1.5134795433893428e-05}
{"train_loss": 0.07490956783294678, "global_step": 202525, "epoch": 2275, "lr": 1.5134379895289058e-05}
{"train_loss": 0.06753350794315338, "global_step": 202526, "epoch": 2275, "lr": 1.5133964361371961e-05}
{"train_loss": 0.028241369873285294, "global_step": 202527, "epoch": 2275, "lr": 1.513354883214218e-05}
{"train_loss": 0.05211244896054268, "global_step": 202528, "epoch": 2275, "lr": 1.5133133307599757e-05}
{"train_loss": 0.07994815707206726, "global_step": 202529, "epoch": 2275, "lr": 1.5132717787744766e-05}
{"train_loss": 0.051652729511260986, "global_step": 202530, "epoch": 2275, "lr": 1.5132302272577248e-05}
{"train_loss": 0.017186634242534637, "global_step": 202531, "epoch": 2275, "lr": 1.5131886762097274e-05}
{"train_loss": 0.0643341988325119, "global_step": 202532, "epoch": 2275, "lr": 1.5131471256304902e-05}
{"train_loss": 0.07066114246845245, "global_step": 202533, "epoch": 2275, "lr": 1.5131055755200158e-05}
{"train_loss": 0.024049686267971992, "global_step": 202534, "epoch": 2275, "lr": 1.5130640258783136e-05}
{"train_loss": 0.09626280516386032, "global_step": 202535, "epoch": 2275, "lr": 1.5130224767053857e-05}
{"train_loss": 0.05975915119051933, "global_step": 202536, "epoch": 2275, "lr": 1.5129809280012414e-05}
{"train_loss": 0.07002688944339752, "global_step": 202537, "epoch": 2275, "lr": 1.5129393797658825e-05}
{"train_loss": 0.03344247490167618, "global_step": 202538, "epoch": 2275, "lr": 1.5128978319993176e-05}
{"train_loss": 0.08048321306705475, "global_step": 202539, "epoch": 2275, "lr": 1.5128562847015499e-05}
{"train_loss": 0.07862064242362976, "global_step": 202540, "epoch": 2275, "lr": 1.5128147378725876e-05}
{"train_loss": 0.05427227541804314, "global_step": 202541, "epoch": 2275, "lr": 1.5127731915124338e-05}
{"train_loss": 0.06393682211637497, "global_step": 202542, "epoch": 2275, "lr": 1.5127316456210943e-05}
{"train_loss": 0.06143632158637047, "global_step": 202543, "epoch": 2275, "lr": 1.5126901001985777e-05}
{"train_loss": 0.05510527268052101, "global_step": 202544, "epoch": 2275, "lr": 1.5126485552448855e-05}
{"train_loss": 0.05302492156624794, "global_step": 202545, "epoch": 2275, "lr": 1.5126070107600271e-05}
{"train_loss": 0.07124263793230057, "global_step": 202546, "epoch": 2275, "lr": 1.5125654667440042e-05}
{"train_loss": 0.04231444373726845, "global_step": 202547, "epoch": 2275, "lr": 1.5125239231968264e-05}
{"train_loss": 0.07987730205059052, "global_step": 202548, "epoch": 2275, "lr": 1.512482380118495e-05}
{"train_loss": 0.06796211004257202, "global_step": 202549, "epoch": 2275, "lr": 1.5124408375090205e-05}
{"train_loss": 0.05990699678659439, "global_step": 202550, "epoch": 2275, "lr": 1.5123992953684036e-05}
{"train_loss": 0.07152505964040756, "global_step": 202551, "epoch": 2275, "lr": 1.5123577536966533e-05}
{"train_loss": 0.06317141652107239, "global_step": 202552, "epoch": 2275, "lr": 1.5123162124937728e-05}
{"train_loss": 0.05226050317287445, "global_step": 202553, "epoch": 2275, "lr": 1.5122746717597708e-05}
{"train_loss": 0.07868009060621262, "global_step": 202554, "epoch": 2275, "lr": 1.5122331314946486e-05}
{"train_loss": 0.04854978621006012, "global_step": 202555, "epoch": 2275, "lr": 1.5121915916984165e-05}
{"train_loss": 0.05056125298142433, "global_step": 202556, "epoch": 2275, "lr": 1.5121500523710758e-05}
{"train_loss": 0.05984240397810936, "global_step": 202557, "epoch": 2275, "lr": 1.5121085135126355e-05}
{"train_loss": 0.08070825040340424, "global_step": 202558, "epoch": 2275, "lr": 1.512066975123098e-05}
{"train_loss": 0.05627657473087311, "global_step": 202559, "epoch": 2275, "lr": 1.5120254372024728e-05}
{"train_loss": 0.05875903740525246, "global_step": 202560, "epoch": 2275, "lr": 1.5119838997507613e-05}
{"train_loss": 0.07820006459951401, "global_step": 202561, "epoch": 2275, "lr": 1.5119423627679707e-05}
{"train_loss": 0.06580325961112976, "global_step": 202562, "epoch": 2275, "lr": 1.511900826254109e-05}
{"train_loss": 0.06344488306033812, "global_step": 202563, "epoch": 2275, "lr": 1.5118592902091777e-05, "val_loss": 8.352550506591797, "train_action_mse_error": 7.774118423461914}
{"train_loss": 0.016072029247879982, "global_step": 202564, "epoch": 2276, "lr": 1.5118177546331862e-05}
{"train_loss": 0.09495754539966583, "global_step": 202565, "epoch": 2276, "lr": 1.5117762195261386e-05}
{"train_loss": 0.09997834265232086, "global_step": 202566, "epoch": 2276, "lr": 1.5117346848880381e-05}
{"train_loss": 0.05648980289697647, "global_step": 202567, "epoch": 2276, "lr": 1.5116931507188942e-05}
{"train_loss": 0.05702821537852287, "global_step": 202568, "epoch": 2276, "lr": 1.5116516170187095e-05}
{"train_loss": 0.07448339462280273, "global_step": 202569, "epoch": 2276, "lr": 1.5116100837874914e-05}
{"train_loss": 0.11988364905118942, "global_step": 202570, "epoch": 2276, "lr": 1.5115685510252453e-05}
{"train_loss": 0.05181030184030533, "global_step": 202571, "epoch": 2276, "lr": 1.5115270187319746e-05}
{"train_loss": 0.04668427258729935, "global_step": 202572, "epoch": 2276, "lr": 1.5114854869076884e-05}
{"train_loss": 0.04798601195216179, "global_step": 202573, "epoch": 2276, "lr": 1.5114439555523886e-05}
{"train_loss": 0.07888425141572952, "global_step": 202574, "epoch": 2276, "lr": 1.5114024246660846e-05}
{"train_loss": 0.08702073246240616, "global_step": 202575, "epoch": 2276, "lr": 1.5113608942487778e-05}
{"train_loss": 0.037785060703754425, "global_step": 202576, "epoch": 2276, "lr": 1.5113193643004781e-05}
{"train_loss": 0.041468095034360886, "global_step": 202577, "epoch": 2276, "lr": 1.5112778348211875e-05}
{"train_loss": 0.05712614208459854, "global_step": 202578, "epoch": 2276, "lr": 1.5112363058109125e-05}
{"train_loss": 0.06649661064147949, "global_step": 202579, "epoch": 2276, "lr": 1.5111947772696611e-05}
{"train_loss": 0.06902449578046799, "global_step": 202580, "epoch": 2276, "lr": 1.5111532491974356e-05}
{"train_loss": 0.0459405891597271, "global_step": 202581, "epoch": 2276, "lr": 1.5111117215942444e-05}
{"train_loss": 0.06339496374130249, "global_step": 202582, "epoch": 2276, "lr": 1.5110701944600902e-05}
{"train_loss": 0.04445108771324158, "global_step": 202583, "epoch": 2276, "lr": 1.511028667794982e-05}
{"train_loss": 0.03376702964305878, "global_step": 202584, "epoch": 2276, "lr": 1.5109871415989219e-05}
{"train_loss": 0.08635153621435165, "global_step": 202585, "epoch": 2276, "lr": 1.5109456158719181e-05}
{"train_loss": 0.07910161465406418, "global_step": 202586, "epoch": 2276, "lr": 1.5109040906139738e-05}
{"train_loss": 0.1039675697684288, "global_step": 202587, "epoch": 2276, "lr": 1.5108625658250975e-05}
{"train_loss": 0.04785003140568733, "global_step": 202588, "epoch": 2276, "lr": 1.5108210415052915e-05}
{"train_loss": 0.08119509369134903, "global_step": 202589, "epoch": 2276, "lr": 1.5107795176545647e-05}
{"train_loss": 0.060583580285310745, "global_step": 202590, "epoch": 2276, "lr": 1.51073799427292e-05}
{"train_loss": 0.062135614454746246, "global_step": 202591, "epoch": 2276, "lr": 1.510696471360365e-05}
{"train_loss": 0.07291208207607269, "global_step": 202592, "epoch": 2276, "lr": 1.5106549489169031e-05}
{"train_loss": 0.0898614227771759, "global_step": 202593, "epoch": 2276, "lr": 1.5106134269425426e-05}
{"train_loss": 0.036625273525714874, "global_step": 202594, "epoch": 2276, "lr": 1.5105719054372858e-05}
{"train_loss": 0.033370550721883774, "global_step": 202595, "epoch": 2276, "lr": 1.5105303844011415e-05}
{"train_loss": 0.05700283870100975, "global_step": 202596, "epoch": 2276, "lr": 1.5104888638341131e-05}
{"train_loss": 0.07689835876226425, "global_step": 202597, "epoch": 2276, "lr": 1.510447343736206e-05}
{"train_loss": 0.06979957222938538, "global_step": 202598, "epoch": 2276, "lr": 1.5104058241074288e-05}
{"train_loss": 0.061347175389528275, "global_step": 202599, "epoch": 2276, "lr": 1.5103643049477834e-05}
{"train_loss": 0.029311127960681915, "global_step": 202600, "epoch": 2276, "lr": 1.5103227862572788e-05}
{"train_loss": 0.046887125819921494, "global_step": 202601, "epoch": 2276, "lr": 1.5102812680359169e-05}
{"train_loss": 0.04101717844605446, "global_step": 202602, "epoch": 2276, "lr": 1.5102397502837069e-05}
{"train_loss": 0.0736609399318695, "global_step": 202603, "epoch": 2276, "lr": 1.510198233000652e-05}
{"train_loss": 0.06118462607264519, "global_step": 202604, "epoch": 2276, "lr": 1.5101567161867575e-05}
{"train_loss": 0.06339362263679504, "global_step": 202605, "epoch": 2276, "lr": 1.5101151998420315e-05}
{"train_loss": 0.09807602316141129, "global_step": 202606, "epoch": 2276, "lr": 1.5100736839664758e-05}
{"train_loss": 0.04680638387799263, "global_step": 202607, "epoch": 2276, "lr": 1.5100321685601004e-05}
{"train_loss": 0.08224833011627197, "global_step": 202608, "epoch": 2276, "lr": 1.509990653622908e-05}
{"train_loss": 0.05180235207080841, "global_step": 202609, "epoch": 2276, "lr": 1.5099491391549037e-05}
{"train_loss": 0.07431939989328384, "global_step": 202610, "epoch": 2276, "lr": 1.5099076251560956e-05}
{"train_loss": 0.07913824915885925, "global_step": 202611, "epoch": 2276, "lr": 1.5098661116264867e-05}
{"train_loss": 0.11334337294101715, "global_step": 202612, "epoch": 2276, "lr": 1.5098245985660847e-05}
{"train_loss": 0.07102189213037491, "global_step": 202613, "epoch": 2276, "lr": 1.509783085974893e-05}
{"train_loss": 0.05935373902320862, "global_step": 202614, "epoch": 2276, "lr": 1.5097415738529185e-05}
{"train_loss": 0.09673523157835007, "global_step": 202615, "epoch": 2276, "lr": 1.5097000622001683e-05}
{"train_loss": 0.048365261405706406, "global_step": 202616, "epoch": 2276, "lr": 1.5096585510166444e-05}
{"train_loss": 0.061444319784641266, "global_step": 202617, "epoch": 2276, "lr": 1.5096170403023563e-05}
{"train_loss": 0.09173091500997543, "global_step": 202618, "epoch": 2276, "lr": 1.509575530057306e-05}
{"train_loss": 0.06988390535116196, "global_step": 202619, "epoch": 2276, "lr": 1.5095340202815028e-05}
{"train_loss": 0.03762426972389221, "global_step": 202620, "epoch": 2276, "lr": 1.509492510974948e-05}
{"train_loss": 0.09464059770107269, "global_step": 202621, "epoch": 2276, "lr": 1.5094510021376513e-05}
{"train_loss": 0.11345823109149933, "global_step": 202622, "epoch": 2276, "lr": 1.5094094937696146e-05}
{"train_loss": 0.06425868719816208, "global_step": 202623, "epoch": 2276, "lr": 1.5093679858708465e-05}
{"train_loss": 0.05092623084783554, "global_step": 202624, "epoch": 2276, "lr": 1.5093264784413502e-05}
{"train_loss": 0.10613270103931427, "global_step": 202625, "epoch": 2276, "lr": 1.5092849714811341e-05}
{"train_loss": 0.06724797934293747, "global_step": 202626, "epoch": 2276, "lr": 1.5092434649901998e-05}
{"train_loss": 0.07392115145921707, "global_step": 202627, "epoch": 2276, "lr": 1.5092019589685574e-05}
{"train_loss": 0.08593524247407913, "global_step": 202628, "epoch": 2276, "lr": 1.5091604534162084e-05}
{"train_loss": 0.061159439384937286, "global_step": 202629, "epoch": 2276, "lr": 1.509118948333162e-05}
{"train_loss": 0.037418171763420105, "global_step": 202630, "epoch": 2276, "lr": 1.5090774437194204e-05}
{"train_loss": 0.04813682287931442, "global_step": 202631, "epoch": 2276, "lr": 1.5090359395749925e-05}
{"train_loss": 0.055360425263643265, "global_step": 202632, "epoch": 2276, "lr": 1.5089944358998804e-05}
{"train_loss": 0.07230459153652191, "global_step": 202633, "epoch": 2276, "lr": 1.5089529326940916e-05}
{"train_loss": 0.10841192305088043, "global_step": 202634, "epoch": 2276, "lr": 1.5089114299576328e-05}
{"train_loss": 0.07037083059549332, "global_step": 202635, "epoch": 2276, "lr": 1.5088699276905066e-05}
{"train_loss": 0.06948980689048767, "global_step": 202636, "epoch": 2276, "lr": 1.5088284258927222e-05}
{"train_loss": 0.06136447563767433, "global_step": 202637, "epoch": 2276, "lr": 1.5087869245642816e-05}
{"train_loss": 0.04721532016992569, "global_step": 202638, "epoch": 2276, "lr": 1.508745423705194e-05}
{"train_loss": 0.07652932405471802, "global_step": 202639, "epoch": 2276, "lr": 1.5087039233154609e-05}
{"train_loss": 0.04881026968359947, "global_step": 202640, "epoch": 2276, "lr": 1.508662423395092e-05}
{"train_loss": 0.08923456072807312, "global_step": 202641, "epoch": 2276, "lr": 1.508620923944089e-05}
{"train_loss": 0.04222157225012779, "global_step": 202642, "epoch": 2276, "lr": 1.5085794249624607e-05}
{"train_loss": 0.09280339628458023, "global_step": 202643, "epoch": 2276, "lr": 1.508537926450212e-05}
{"train_loss": 0.07165844738483429, "global_step": 202644, "epoch": 2276, "lr": 1.508496428407346e-05}
{"train_loss": 0.07779179513454437, "global_step": 202645, "epoch": 2276, "lr": 1.5084549308338714e-05}
{"train_loss": 0.18889889121055603, "global_step": 202646, "epoch": 2276, "lr": 1.5084134337297911e-05}
{"train_loss": 0.1099860891699791, "global_step": 202647, "epoch": 2276, "lr": 1.5083719370951138e-05}
{"train_loss": 0.06728015094995499, "global_step": 202648, "epoch": 2276, "lr": 1.5083304409298432e-05}
{"train_loss": 0.09650889039039612, "global_step": 202649, "epoch": 2276, "lr": 1.5082889452339832e-05}
{"train_loss": 0.06548993289470673, "global_step": 202650, "epoch": 2276, "lr": 1.5082474500075434e-05}
{"train_loss": 0.03551214933395386, "global_step": 202651, "epoch": 2276, "lr": 1.5082059552505252e-05}
{"train_loss": 0.06880126267755299, "global_step": 202652, "epoch": 2276, "lr": 1.5081644609629358e-05, "val_loss": 8.275812149047852}
{"train_loss": 0.05349328741431236, "global_step": 202653, "epoch": 2277, "lr": 1.5081229671447833e-05}
{"train_loss": 0.0554225780069828, "global_step": 202654, "epoch": 2277, "lr": 1.508081473796069e-05}
{"train_loss": 0.054995059967041016, "global_step": 202655, "epoch": 2277, "lr": 1.5080399809168023e-05}
{"train_loss": 0.05089598894119263, "global_step": 202656, "epoch": 2277, "lr": 1.5079984885069859e-05}
{"train_loss": 0.054326266050338745, "global_step": 202657, "epoch": 2277, "lr": 1.5079569965666278e-05}
{"train_loss": 0.07738792896270752, "global_step": 202658, "epoch": 2277, "lr": 1.5079155050957305e-05}
{"train_loss": 0.020918820053339005, "global_step": 202659, "epoch": 2277, "lr": 1.5078740140943031e-05}
{"train_loss": 0.04707365483045578, "global_step": 202660, "epoch": 2277, "lr": 1.5078325235623486e-05}
{"train_loss": 0.07077182084321976, "global_step": 202661, "epoch": 2277, "lr": 1.5077910334998741e-05}
{"train_loss": 0.07807145267724991, "global_step": 202662, "epoch": 2277, "lr": 1.5077495439068834e-05}
{"train_loss": 0.08219014108181, "global_step": 202663, "epoch": 2277, "lr": 1.507708054783385e-05}
{"train_loss": 0.0561562217772007, "global_step": 202664, "epoch": 2277, "lr": 1.5076665661293804e-05}
{"train_loss": 0.0676526352763176, "global_step": 202665, "epoch": 2277, "lr": 1.5076250779448796e-05}
{"train_loss": 0.10399320721626282, "global_step": 202666, "epoch": 2277, "lr": 1.5075835902298847e-05}
{"train_loss": 0.06266511231660843, "global_step": 202667, "epoch": 2277, "lr": 1.5075421029844034e-05}
{"train_loss": 0.02723751775920391, "global_step": 202668, "epoch": 2277, "lr": 1.5075006162084393e-05}
{"train_loss": 0.08439766615629196, "global_step": 202669, "epoch": 2277, "lr": 1.5074591299019996e-05}
{"train_loss": 0.041024766862392426, "global_step": 202670, "epoch": 2277, "lr": 1.5074176440650906e-05}
{"train_loss": 0.10294271260499954, "global_step": 202671, "epoch": 2277, "lr": 1.5073761586977154e-05}
{"train_loss": 0.09393834322690964, "global_step": 202672, "epoch": 2277, "lr": 1.5073346737998822e-05}
{"train_loss": 0.023623749613761902, "global_step": 202673, "epoch": 2277, "lr": 1.5072931893715936e-05}
{"train_loss": 0.03647331893444061, "global_step": 202674, "epoch": 2277, "lr": 1.5072517054128588e-05}
{"train_loss": 0.05338539555668831, "global_step": 202675, "epoch": 2277, "lr": 1.5072102219236794e-05}
{"train_loss": 0.079646036028862, "global_step": 202676, "epoch": 2277, "lr": 1.507168738904065e-05}
{"train_loss": 0.103085957467556, "global_step": 202677, "epoch": 2277, "lr": 1.5071272563540174e-05}
{"train_loss": 0.059768714010715485, "global_step": 202678, "epoch": 2277, "lr": 1.5070857742735455e-05}
{"train_loss": 0.03123495914041996, "global_step": 202679, "epoch": 2277, "lr": 1.5070442926626521e-05}
{"train_loss": 0.05669866129755974, "global_step": 202680, "epoch": 2277, "lr": 1.5070028115213458e-05}
{"train_loss": 0.04787002503871918, "global_step": 202681, "epoch": 2277, "lr": 1.50696133084963e-05}
{"train_loss": 0.0613982193171978, "global_step": 202682, "epoch": 2277, "lr": 1.5069198506475091e-05}
{"train_loss": 0.09915181994438171, "global_step": 202683, "epoch": 2277, "lr": 1.5068783709149914e-05}
{"train_loss": 0.05418001487851143, "global_step": 202684, "epoch": 2277, "lr": 1.5068368916520803e-05}
{"train_loss": 0.05851534381508827, "global_step": 202685, "epoch": 2277, "lr": 1.5067954128587841e-05}
{"train_loss": 0.11166047304868698, "global_step": 202686, "epoch": 2277, "lr": 1.5067539345351061e-05}
{"train_loss": 0.03838130831718445, "global_step": 202687, "epoch": 2277, "lr": 1.5067124566810514e-05}
{"train_loss": 0.03870207071304321, "global_step": 202688, "epoch": 2277, "lr": 1.5066709792966266e-05}
{"train_loss": 0.0552787110209465, "global_step": 202689, "epoch": 2277, "lr": 1.5066295023818388e-05}
{"train_loss": 0.05622097849845886, "global_step": 202690, "epoch": 2277, "lr": 1.5065880259366904e-05}
{"train_loss": 0.04925941303372383, "global_step": 202691, "epoch": 2277, "lr": 1.5065465499611902e-05}
{"train_loss": 0.07756093144416809, "global_step": 202692, "epoch": 2277, "lr": 1.506505074455341e-05}
{"train_loss": 0.0757194459438324, "global_step": 202693, "epoch": 2277, "lr": 1.5064635994191505e-05}
{"train_loss": 0.04952622577548027, "global_step": 202694, "epoch": 2277, "lr": 1.5064221248526227e-05}
{"train_loss": 0.0323978066444397, "global_step": 202695, "epoch": 2277, "lr": 1.5063806507557649e-05}
{"train_loss": 0.05206882208585739, "global_step": 202696, "epoch": 2277, "lr": 1.50633917712858e-05}
{"train_loss": 0.03132965788245201, "global_step": 202697, "epoch": 2277, "lr": 1.5062977039710768e-05}
{"train_loss": 0.0444614514708519, "global_step": 202698, "epoch": 2277, "lr": 1.506256231283258e-05}
{"train_loss": 0.054811060428619385, "global_step": 202699, "epoch": 2277, "lr": 1.506214759065132e-05}
{"train_loss": 0.041266027837991714, "global_step": 202700, "epoch": 2277, "lr": 1.5061732873167011e-05}
{"train_loss": 0.08683506399393082, "global_step": 202701, "epoch": 2277, "lr": 1.5061318160379744e-05}
{"train_loss": 0.06616532057523727, "global_step": 202702, "epoch": 2277, "lr": 1.5060903452289537e-05}
{"train_loss": 0.11116310209035873, "global_step": 202703, "epoch": 2277, "lr": 1.5060488748896489e-05}
{"train_loss": 0.08743806928396225, "global_step": 202704, "epoch": 2277, "lr": 1.5060074050200613e-05}
{"train_loss": 0.026853788644075394, "global_step": 202705, "epoch": 2277, "lr": 1.5059659356201983e-05}
{"train_loss": 0.08916353434324265, "global_step": 202706, "epoch": 2277, "lr": 1.5059244666900668e-05}
{"train_loss": 0.046138472855091095, "global_step": 202707, "epoch": 2277, "lr": 1.5058829982296702e-05}
{"train_loss": 0.04273967072367668, "global_step": 202708, "epoch": 2277, "lr": 1.5058415302390162e-05}
{"train_loss": 0.0623771958053112, "global_step": 202709, "epoch": 2277, "lr": 1.5058000627181079e-05}
{"train_loss": 0.05264074727892876, "global_step": 202710, "epoch": 2277, "lr": 1.505758595666954e-05}
{"train_loss": 0.0563352145254612, "global_step": 202711, "epoch": 2277, "lr": 1.505717129085556e-05}
{"train_loss": 0.09319695830345154, "global_step": 202712, "epoch": 2277, "lr": 1.5056756629739238e-05}
{"train_loss": 0.05485822632908821, "global_step": 202713, "epoch": 2277, "lr": 1.5056341973320592e-05}
{"train_loss": 0.048016924411058426, "global_step": 202714, "epoch": 2277, "lr": 1.505592732159971e-05}
{"train_loss": 0.036879029124975204, "global_step": 202715, "epoch": 2277, "lr": 1.5055512674576622e-05}
{"train_loss": 0.11855069547891617, "global_step": 202716, "epoch": 2277, "lr": 1.5055098032251408e-05}
{"train_loss": 0.01965007744729519, "global_step": 202717, "epoch": 2277, "lr": 1.5054683394624092e-05}
{"train_loss": 0.07839851081371307, "global_step": 202718, "epoch": 2277, "lr": 1.5054268761694762e-05}
{"train_loss": 0.05211982503533363, "global_step": 202719, "epoch": 2277, "lr": 1.5053854133463464e-05}
{"train_loss": 0.0478789359331131, "global_step": 202720, "epoch": 2277, "lr": 1.5053439509930233e-05}
{"train_loss": 0.055783964693546295, "global_step": 202721, "epoch": 2277, "lr": 1.5053024891095153e-05}
{"train_loss": 0.03515450656414032, "global_step": 202722, "epoch": 2277, "lr": 1.5052610276958257e-05}
{"train_loss": 0.07629799097776413, "global_step": 202723, "epoch": 2277, "lr": 1.5052195667519624e-05}
{"train_loss": 0.0464053675532341, "global_step": 202724, "epoch": 2277, "lr": 1.505178106277928e-05}
{"train_loss": 0.0417385958135128, "global_step": 202725, "epoch": 2277, "lr": 1.5051366462737321e-05}
{"train_loss": 0.049696389585733414, "global_step": 202726, "epoch": 2277, "lr": 1.5050951867393759e-05}
{"train_loss": 0.04778652265667915, "global_step": 202727, "epoch": 2277, "lr": 1.5050537276748683e-05}
{"train_loss": 0.05928555876016617, "global_step": 202728, "epoch": 2277, "lr": 1.5050122690802127e-05}
{"train_loss": 0.03956533968448639, "global_step": 202729, "epoch": 2277, "lr": 1.5049708109554167e-05}
{"train_loss": 0.02878989651799202, "global_step": 202730, "epoch": 2277, "lr": 1.5049293533004838e-05}
{"train_loss": 0.04208000749349594, "global_step": 202731, "epoch": 2277, "lr": 1.5048878961154217e-05}
{"train_loss": 0.061897825449705124, "global_step": 202732, "epoch": 2277, "lr": 1.5048464394002332e-05}
{"train_loss": 0.05625423789024353, "global_step": 202733, "epoch": 2277, "lr": 1.5048049831549271e-05}
{"train_loss": 0.1003308966755867, "global_step": 202734, "epoch": 2277, "lr": 1.5047635273795057e-05}
{"train_loss": 0.07056209444999695, "global_step": 202735, "epoch": 2277, "lr": 1.504722072073978e-05}
{"train_loss": 0.050685156136751175, "global_step": 202736, "epoch": 2277, "lr": 1.5046806172383466e-05}
{"train_loss": 0.04861903935670853, "global_step": 202737, "epoch": 2277, "lr": 1.5046391628726191e-05}
{"train_loss": 0.06617538630962372, "global_step": 202738, "epoch": 2277, "lr": 1.5045977089767993e-05}
{"train_loss": 0.07375764846801758, "global_step": 202739, "epoch": 2277, "lr": 1.5045562555508952e-05}
{"train_loss": 0.08145001530647278, "global_step": 202740, "epoch": 2277, "lr": 1.504514802594909e-05}
{"train_loss": 0.05973875503777788, "global_step": 202741, "epoch": 2277, "lr": 1.5044733501088481e-05, "val_loss": 8.251267433166504}
{"train_loss": 0.0720270425081253, "global_step": 202742, "epoch": 2278, "lr": 1.5044318980927203e-05}
{"train_loss": 0.051981423050165176, "global_step": 202743, "epoch": 2278, "lr": 1.504390446546527e-05}
{"train_loss": 0.05157153308391571, "global_step": 202744, "epoch": 2278, "lr": 1.5043489954702777e-05}
{"train_loss": 0.08081258088350296, "global_step": 202745, "epoch": 2278, "lr": 1.5043075448639744e-05}
{"train_loss": 0.08690435439348221, "global_step": 202746, "epoch": 2278, "lr": 1.504266094727626e-05}
{"train_loss": 0.04975207895040512, "global_step": 202747, "epoch": 2278, "lr": 1.5042246450612347e-05}
{"train_loss": 0.05334482342004776, "global_step": 202748, "epoch": 2278, "lr": 1.504183195864809e-05}
{"train_loss": 0.01202346570789814, "global_step": 202749, "epoch": 2278, "lr": 1.5041417471383523e-05}
{"train_loss": 0.046607885509729385, "global_step": 202750, "epoch": 2278, "lr": 1.5041002988818725e-05}
{"train_loss": 0.08853872865438461, "global_step": 202751, "epoch": 2278, "lr": 1.5040588510953718e-05}
{"train_loss": 0.060255300253629684, "global_step": 202752, "epoch": 2278, "lr": 1.50401740377886e-05}
{"train_loss": 0.02336660400032997, "global_step": 202753, "epoch": 2278, "lr": 1.5039759569323386e-05}
{"train_loss": 0.06028158590197563, "global_step": 202754, "epoch": 2278, "lr": 1.5039345105558167e-05}
{"train_loss": 0.129841148853302, "global_step": 202755, "epoch": 2278, "lr": 1.5038930646492965e-05}
{"train_loss": 0.08255331963300705, "global_step": 202756, "epoch": 2278, "lr": 1.5038516192127866e-05}
{"train_loss": 0.04771735146641731, "global_step": 202757, "epoch": 2278, "lr": 1.5038101742462902e-05}
{"train_loss": 0.058080658316612244, "global_step": 202758, "epoch": 2278, "lr": 1.5037687297498154e-05}
{"train_loss": 0.06313569098711014, "global_step": 202759, "epoch": 2278, "lr": 1.503727285723366e-05}
{"train_loss": 0.10127473622560501, "global_step": 202760, "epoch": 2278, "lr": 1.5036858421669464e-05}
{"train_loss": 0.06519287824630737, "global_step": 202761, "epoch": 2278, "lr": 1.5036443990805648e-05}
{"train_loss": 0.049889303743839264, "global_step": 202762, "epoch": 2278, "lr": 1.5036029564642245e-05}
{"train_loss": 0.09626539051532745, "global_step": 202763, "epoch": 2278, "lr": 1.5035615143179337e-05}
{"train_loss": 0.017039654776453972, "global_step": 202764, "epoch": 2278, "lr": 1.5035200726416948e-05}
{"train_loss": 0.0537535697221756, "global_step": 202765, "epoch": 2278, "lr": 1.5034786314355164e-05}
{"train_loss": 0.04258156940340996, "global_step": 202766, "epoch": 2278, "lr": 1.5034371906994015e-05}
{"train_loss": 0.038341931998729706, "global_step": 202767, "epoch": 2278, "lr": 1.5033957504333584e-05}
{"train_loss": 0.07785370200872421, "global_step": 202768, "epoch": 2278, "lr": 1.5033543106373893e-05}
{"train_loss": 0.06729648262262344, "global_step": 202769, "epoch": 2278, "lr": 1.5033128713115036e-05}
{"train_loss": 0.032975856214761734, "global_step": 202770, "epoch": 2278, "lr": 1.5032714324557024e-05}
{"train_loss": 0.046064458787441254, "global_step": 202771, "epoch": 2278, "lr": 1.5032299940699962e-05}
{"train_loss": 0.03731725737452507, "global_step": 202772, "epoch": 2278, "lr": 1.5031885561543862e-05}
{"train_loss": 0.07267531752586365, "global_step": 202773, "epoch": 2278, "lr": 1.5031471187088813e-05}
{"train_loss": 0.05886755511164665, "global_step": 202774, "epoch": 2278, "lr": 1.5031056817334843e-05}
{"train_loss": 0.07169582694768906, "global_step": 202775, "epoch": 2278, "lr": 1.5030642452282034e-05}
{"train_loss": 0.04082166776061058, "global_step": 202776, "epoch": 2278, "lr": 1.5030228091930416e-05}
{"train_loss": 0.050350554287433624, "global_step": 202777, "epoch": 2278, "lr": 1.5029813736280074e-05}
{"train_loss": 0.0885680764913559, "global_step": 202778, "epoch": 2278, "lr": 1.5029399385331028e-05}
{"train_loss": 0.02875410206615925, "global_step": 202779, "epoch": 2278, "lr": 1.5028985039083354e-05}
{"train_loss": 0.024561887606978416, "global_step": 202780, "epoch": 2278, "lr": 1.5028570697537126e-05}
{"train_loss": 0.02818748727440834, "global_step": 202781, "epoch": 2278, "lr": 1.5028156360692364e-05}
{"train_loss": 0.06966041028499603, "global_step": 202782, "epoch": 2278, "lr": 1.5027742028549152e-05}
{"train_loss": 0.040015604346990585, "global_step": 202783, "epoch": 2278, "lr": 1.5027327701107519e-05}
{"train_loss": 0.05484490096569061, "global_step": 202784, "epoch": 2278, "lr": 1.5026913378367551e-05}
{"train_loss": 0.0357745885848999, "global_step": 202785, "epoch": 2278, "lr": 1.5026499060329274e-05}
{"train_loss": 0.07956008613109589, "global_step": 202786, "epoch": 2278, "lr": 1.5026084746992774e-05}
{"train_loss": 0.0699276477098465, "global_step": 202787, "epoch": 2278, "lr": 1.5025670438358075e-05}
{"train_loss": 0.08151497691869736, "global_step": 202788, "epoch": 2278, "lr": 1.5025256134425263e-05}
{"train_loss": 0.06313642114400864, "global_step": 202789, "epoch": 2278, "lr": 1.5024841835194364e-05}
{"train_loss": 0.07796984910964966, "global_step": 202790, "epoch": 2278, "lr": 1.5024427540665464e-05}
{"train_loss": 0.06023910269141197, "global_step": 202791, "epoch": 2278, "lr": 1.5024013250838591e-05}
{"train_loss": 0.07543381303548813, "global_step": 202792, "epoch": 2278, "lr": 1.5023598965713826e-05}
{"train_loss": 0.05990230664610863, "global_step": 202793, "epoch": 2278, "lr": 1.5023184685291198e-05}
{"train_loss": 0.06536344438791275, "global_step": 202794, "epoch": 2278, "lr": 1.5022770409570792e-05}
{"train_loss": 0.040141403675079346, "global_step": 202795, "epoch": 2278, "lr": 1.5022356138552628e-05}
{"train_loss": 0.053940724581480026, "global_step": 202796, "epoch": 2278, "lr": 1.50219418722368e-05}
{"train_loss": 0.08345253765583038, "global_step": 202797, "epoch": 2278, "lr": 1.5021527610623349e-05}
{"train_loss": 0.031939391046762466, "global_step": 202798, "epoch": 2278, "lr": 1.5021113353712307e-05}
{"train_loss": 0.05177481845021248, "global_step": 202799, "epoch": 2278, "lr": 1.5020699101503765e-05}
{"train_loss": 0.08027169108390808, "global_step": 202800, "epoch": 2278, "lr": 1.502028485399775e-05}
{"train_loss": 0.06759153306484222, "global_step": 202801, "epoch": 2278, "lr": 1.5019870611194348e-05}
{"train_loss": 0.02293485403060913, "global_step": 202802, "epoch": 2278, "lr": 1.5019456373093582e-05}
{"train_loss": 0.05262259393930435, "global_step": 202803, "epoch": 2278, "lr": 1.5019042139695538e-05}
{"train_loss": 0.04386496543884277, "global_step": 202804, "epoch": 2278, "lr": 1.5018627911000243e-05}
{"train_loss": 0.04028256610035896, "global_step": 202805, "epoch": 2278, "lr": 1.5018213687007782e-05}
{"train_loss": 0.0564182847738266, "global_step": 202806, "epoch": 2278, "lr": 1.501779946771818e-05}
{"train_loss": 0.06587077677249908, "global_step": 202807, "epoch": 2278, "lr": 1.5017385253131527e-05}
{"train_loss": 0.05512060597538948, "global_step": 202808, "epoch": 2278, "lr": 1.501697104324784e-05}
{"train_loss": 0.038416143506765366, "global_step": 202809, "epoch": 2278, "lr": 1.5016556838067214e-05}
{"train_loss": 0.07667166739702225, "global_step": 202810, "epoch": 2278, "lr": 1.501614263758967e-05}
{"train_loss": 0.13667599856853485, "global_step": 202811, "epoch": 2278, "lr": 1.5015728441815291e-05}
{"train_loss": 0.04029476270079613, "global_step": 202812, "epoch": 2278, "lr": 1.501531425074411e-05}
{"train_loss": 0.07951424270868301, "global_step": 202813, "epoch": 2278, "lr": 1.5014900064376209e-05}
{"train_loss": 0.09811429679393768, "global_step": 202814, "epoch": 2278, "lr": 1.501448588271161e-05}
{"train_loss": 0.04836566001176834, "global_step": 202815, "epoch": 2278, "lr": 1.5014071705750387e-05}
{"train_loss": 0.07080729305744171, "global_step": 202816, "epoch": 2278, "lr": 1.5013657533492615e-05}
{"train_loss": 0.0803096815943718, "global_step": 202817, "epoch": 2278, "lr": 1.5013243365938306e-05}
{"train_loss": 0.1267457753419876, "global_step": 202818, "epoch": 2278, "lr": 1.5012829203087564e-05}
{"train_loss": 0.061677947640419006, "global_step": 202819, "epoch": 2278, "lr": 1.5012415044940403e-05}
{"train_loss": 0.0364786833524704, "global_step": 202820, "epoch": 2278, "lr": 1.5012000891496908e-05}
{"train_loss": 0.06595723330974579, "global_step": 202821, "epoch": 2278, "lr": 1.5011586742757112e-05}
{"train_loss": 0.04619504511356354, "global_step": 202822, "epoch": 2278, "lr": 1.5011172598721096e-05}
{"train_loss": 0.06807518750429153, "global_step": 202823, "epoch": 2278, "lr": 1.5010758459388885e-05}
{"train_loss": 0.05335862934589386, "global_step": 202824, "epoch": 2278, "lr": 1.5010344324760572e-05}
{"train_loss": 0.09590870141983032, "global_step": 202825, "epoch": 2278, "lr": 1.5009930194836169e-05}
{"train_loss": 0.08395615220069885, "global_step": 202826, "epoch": 2278, "lr": 1.5009516069615776e-05}
{"train_loss": 0.05844730883836746, "global_step": 202827, "epoch": 2278, "lr": 1.5009101949099407e-05}
{"train_loss": 0.06107279658317566, "global_step": 202828, "epoch": 2278, "lr": 1.5008687833287155e-05}
{"train_loss": 0.031204717233777046, "global_step": 202829, "epoch": 2278, "lr": 1.500827372217904e-05}
{"train_loss": 0.060431474108207095, "global_step": 202830, "epoch": 2278, "lr": 1.5007859615775155e-05, "val_loss": 8.17455768585205}
{"train_loss": 0.033729519695043564, "global_step": 202831, "epoch": 2279, "lr": 1.5007445514075518e-05}
{"train_loss": 0.07281389832496643, "global_step": 202832, "epoch": 2279, "lr": 1.5007031417080225e-05}
{"train_loss": 0.051327094435691833, "global_step": 202833, "epoch": 2279, "lr": 1.5006617324789291e-05}
{"train_loss": 0.04781051725149155, "global_step": 202834, "epoch": 2279, "lr": 1.5006203237202804e-05}
{"train_loss": 0.06088081747293472, "global_step": 202835, "epoch": 2279, "lr": 1.500578915432081e-05}
{"train_loss": 0.09321994334459305, "global_step": 202836, "epoch": 2279, "lr": 1.5005375076143346e-05}
{"train_loss": 0.0717083215713501, "global_step": 202837, "epoch": 2279, "lr": 1.5004961002670493e-05}
{"train_loss": 0.07955370843410492, "global_step": 202838, "epoch": 2279, "lr": 1.5004546933902286e-05}
{"train_loss": 0.058270715177059174, "global_step": 202839, "epoch": 2279, "lr": 1.5004132869838805e-05}
{"train_loss": 0.008269743993878365, "global_step": 202840, "epoch": 2279, "lr": 1.5003718810480078e-05}
{"train_loss": 0.04887060075998306, "global_step": 202841, "epoch": 2279, "lr": 1.5003304755826186e-05}
{"train_loss": 0.09829426556825638, "global_step": 202842, "epoch": 2279, "lr": 1.5002890705877159e-05}
{"train_loss": 0.1131548285484314, "global_step": 202843, "epoch": 2279, "lr": 1.5002476660633086e-05}
{"train_loss": 0.025810619816184044, "global_step": 202844, "epoch": 2279, "lr": 1.5002062620093987e-05}
{"train_loss": 0.020596541464328766, "global_step": 202845, "epoch": 2279, "lr": 1.5001648584259947e-05}
{"train_loss": 0.03249160200357437, "global_step": 202846, "epoch": 2279, "lr": 1.5001234553130994e-05}
{"train_loss": 0.03171947970986366, "global_step": 202847, "epoch": 2279, "lr": 1.5000820526707216e-05}
{"train_loss": 0.014232417568564415, "global_step": 202848, "epoch": 2279, "lr": 1.5000406504988628e-05}
{"train_loss": 0.06332854181528091, "global_step": 202849, "epoch": 2279, "lr": 1.4999992487975334e-05}
{"train_loss": 0.03703050687909126, "global_step": 202850, "epoch": 2279, "lr": 1.4999578475667341e-05}
{"train_loss": 0.08670379966497421, "global_step": 202851, "epoch": 2279, "lr": 1.4999164468064735e-05}
{"train_loss": 0.08369319885969162, "global_step": 202852, "epoch": 2279, "lr": 1.4998750465167577e-05}
{"train_loss": 0.09150896966457367, "global_step": 202853, "epoch": 2279, "lr": 1.4998336466975893e-05}
{"train_loss": 0.055382244288921356, "global_step": 202854, "epoch": 2279, "lr": 1.4997922473489779e-05}
{"train_loss": 0.06005258485674858, "global_step": 202855, "epoch": 2279, "lr": 1.4997508484709244e-05}
{"train_loss": 0.0728839784860611, "global_step": 202856, "epoch": 2279, "lr": 1.4997094500634384e-05}
{"train_loss": 0.05046498775482178, "global_step": 202857, "epoch": 2279, "lr": 1.4996680521265228e-05}
{"train_loss": 0.06930461525917053, "global_step": 202858, "epoch": 2279, "lr": 1.4996266546601851e-05}
{"train_loss": 0.07006983458995819, "global_step": 202859, "epoch": 2279, "lr": 1.4995852576644288e-05}
{"train_loss": 0.1285359412431717, "global_step": 202860, "epoch": 2279, "lr": 1.4995438611392621e-05}
{"train_loss": 0.08966914564371109, "global_step": 202861, "epoch": 2279, "lr": 1.4995024650846873e-05}
{"train_loss": 0.07914402335882187, "global_step": 202862, "epoch": 2279, "lr": 1.4994610695007133e-05}
{"train_loss": 0.04275615140795708, "global_step": 202863, "epoch": 2279, "lr": 1.4994196743873423e-05}
{"train_loss": 0.09213867783546448, "global_step": 202864, "epoch": 2279, "lr": 1.4993782797445837e-05}
{"train_loss": 0.07697922736406326, "global_step": 202865, "epoch": 2279, "lr": 1.4993368855724393e-05}
{"train_loss": 0.08094902336597443, "global_step": 202866, "epoch": 2279, "lr": 1.499295491870918e-05}
{"train_loss": 0.06823862344026566, "global_step": 202867, "epoch": 2279, "lr": 1.4992540986400222e-05}
{"train_loss": 0.09015047550201416, "global_step": 202868, "epoch": 2279, "lr": 1.4992127058797589e-05}
{"train_loss": 0.0467243567109108, "global_step": 202869, "epoch": 2279, "lr": 1.4991713135901352e-05}
{"train_loss": 0.04487168416380882, "global_step": 202870, "epoch": 2279, "lr": 1.4991299217711558e-05}
{"train_loss": 0.06825143843889236, "global_step": 202871, "epoch": 2279, "lr": 1.4990885304228236e-05}
{"train_loss": 0.029968002811074257, "global_step": 202872, "epoch": 2279, "lr": 1.4990471395451483e-05}
{"train_loss": 0.14136843383312225, "global_step": 202873, "epoch": 2279, "lr": 1.4990057491381316e-05}
{"train_loss": 0.02313619665801525, "global_step": 202874, "epoch": 2279, "lr": 1.4989643592017821e-05}
{"train_loss": 0.10007797926664352, "global_step": 202875, "epoch": 2279, "lr": 1.4989229697361046e-05}
{"train_loss": 0.06893189996480942, "global_step": 202876, "epoch": 2279, "lr": 1.4988815807411027e-05}
{"train_loss": 0.08679370582103729, "global_step": 202877, "epoch": 2279, "lr": 1.498840192216785e-05}
{"train_loss": 0.09708058089017868, "global_step": 202878, "epoch": 2279, "lr": 1.498798804163154e-05}
{"train_loss": 0.10450567305088043, "global_step": 202879, "epoch": 2279, "lr": 1.4987574165802182e-05}
{"train_loss": 0.13411863148212433, "global_step": 202880, "epoch": 2279, "lr": 1.4987160294679803e-05}
{"train_loss": 0.08912386745214462, "global_step": 202881, "epoch": 2279, "lr": 1.4986746428264492e-05}
{"train_loss": 0.06026984378695488, "global_step": 202882, "epoch": 2279, "lr": 1.498633256655627e-05}
{"train_loss": 0.07020790129899979, "global_step": 202883, "epoch": 2279, "lr": 1.4985918709555225e-05}
{"train_loss": 0.049163106828927994, "global_step": 202884, "epoch": 2279, "lr": 1.4985504857261378e-05}
{"train_loss": 0.07248248904943466, "global_step": 202885, "epoch": 2279, "lr": 1.4985091009674817e-05}
{"train_loss": 0.053172994405031204, "global_step": 202886, "epoch": 2279, "lr": 1.4984677166795574e-05}
{"train_loss": 0.08483876287937164, "global_step": 202887, "epoch": 2279, "lr": 1.498426332862371e-05}
{"train_loss": 0.03685326501727104, "global_step": 202888, "epoch": 2279, "lr": 1.4983849495159302e-05}
{"train_loss": 0.025568241253495216, "global_step": 202889, "epoch": 2279, "lr": 1.4983435666402374e-05}
{"train_loss": 0.10451419651508331, "global_step": 202890, "epoch": 2279, "lr": 1.4983021842353012e-05}
{"train_loss": 0.05934314429759979, "global_step": 202891, "epoch": 2279, "lr": 1.4982608023011235e-05}
{"train_loss": 0.09448054432868958, "global_step": 202892, "epoch": 2279, "lr": 1.4982194208377143e-05}
{"train_loss": 0.01944795809686184, "global_step": 202893, "epoch": 2279, "lr": 1.4981780398450746e-05}
{"train_loss": 0.05842660367488861, "global_step": 202894, "epoch": 2279, "lr": 1.4981366593232143e-05}
{"train_loss": 0.0674736499786377, "global_step": 202895, "epoch": 2279, "lr": 1.4980952792721349e-05}
{"train_loss": 0.06448930501937866, "global_step": 202896, "epoch": 2279, "lr": 1.4980538996918453e-05}
{"train_loss": 0.053872935473918915, "global_step": 202897, "epoch": 2279, "lr": 1.4980125205823486e-05}
{"train_loss": 0.0401751734316349, "global_step": 202898, "epoch": 2279, "lr": 1.4979711419436526e-05}
{"train_loss": 0.10020899027585983, "global_step": 202899, "epoch": 2279, "lr": 1.4979297637757605e-05}
{"train_loss": 0.07026803493499756, "global_step": 202900, "epoch": 2279, "lr": 1.4978883860786802e-05}
{"train_loss": 0.0510372593998909, "global_step": 202901, "epoch": 2279, "lr": 1.4978470088524149e-05}
{"train_loss": 0.0795733630657196, "global_step": 202902, "epoch": 2279, "lr": 1.497805632096973e-05}
{"train_loss": 0.07553215324878693, "global_step": 202903, "epoch": 2279, "lr": 1.4977642558123567e-05}
{"train_loss": 0.08904746919870377, "global_step": 202904, "epoch": 2279, "lr": 1.4977228799985748e-05}
{"train_loss": 0.06941933929920197, "global_step": 202905, "epoch": 2279, "lr": 1.4976815046556303e-05}
{"train_loss": 0.08126605302095413, "global_step": 202906, "epoch": 2279, "lr": 1.497640129783529e-05}
{"train_loss": 0.09477050602436066, "global_step": 202907, "epoch": 2279, "lr": 1.4975987553822795e-05}
{"train_loss": 0.06541317701339722, "global_step": 202908, "epoch": 2279, "lr": 1.4975573814518851e-05}
{"train_loss": 0.0629035234451294, "global_step": 202909, "epoch": 2279, "lr": 1.4975160079923495e-05}
{"train_loss": 0.04015020653605461, "global_step": 202910, "epoch": 2279, "lr": 1.4974746350036823e-05}
{"train_loss": 0.06708120554685593, "global_step": 202911, "epoch": 2279, "lr": 1.497433262485885e-05}
{"train_loss": 0.06770806014537811, "global_step": 202912, "epoch": 2279, "lr": 1.4973918904389666e-05}
{"train_loss": 0.04879765212535858, "global_step": 202913, "epoch": 2279, "lr": 1.4973505188629311e-05}
{"train_loss": 0.05238255485892296, "global_step": 202914, "epoch": 2279, "lr": 1.4973091477577827e-05}
{"train_loss": 0.05259789898991585, "global_step": 202915, "epoch": 2279, "lr": 1.4972677771235305e-05}
{"train_loss": 0.039781708270311356, "global_step": 202916, "epoch": 2279, "lr": 1.4972264069601755e-05}
{"train_loss": 0.03428633511066437, "global_step": 202917, "epoch": 2279, "lr": 1.4971850372677276e-05}
{"train_loss": 0.05033301189541817, "global_step": 202918, "epoch": 2279, "lr": 1.4971436680461893e-05}
{"train_loss": 0.06571666731007314, "global_step": 202919, "epoch": 2279, "lr": 1.4971022992955685e-05, "val_loss": 8.251047134399414}
{"train_loss": 0.1535578966140747, "global_step": 202920, "epoch": 2280, "lr": 1.4970609310158684e-05}
{"train_loss": 0.060103047639131546, "global_step": 202921, "epoch": 2280, "lr": 1.4970195632070972e-05}
{"train_loss": 0.08240555971860886, "global_step": 202922, "epoch": 2280, "lr": 1.4969781958692569e-05}
{"train_loss": 0.05183129385113716, "global_step": 202923, "epoch": 2280, "lr": 1.4969368290023556e-05}
{"train_loss": 0.03319384530186653, "global_step": 202924, "epoch": 2280, "lr": 1.4968954626064002e-05}
{"train_loss": 0.04112526774406433, "global_step": 202925, "epoch": 2280, "lr": 1.4968540966813926e-05}
{"train_loss": 0.0713488832116127, "global_step": 202926, "epoch": 2280, "lr": 1.4968127312273423e-05}
{"train_loss": 0.0183675829321146, "global_step": 202927, "epoch": 2280, "lr": 1.4967713662442506e-05}
{"train_loss": 0.07622196525335312, "global_step": 202928, "epoch": 2280, "lr": 1.4967300017321273e-05}
{"train_loss": 0.09750213474035263, "global_step": 202929, "epoch": 2280, "lr": 1.4966886376909745e-05}
{"train_loss": 0.020093746483325958, "global_step": 202930, "epoch": 2280, "lr": 1.4966472741208005e-05}
{"train_loss": 0.0889696329832077, "global_step": 202931, "epoch": 2280, "lr": 1.496605911021608e-05}
{"train_loss": 0.0572577640414238, "global_step": 202932, "epoch": 2280, "lr": 1.4965645483934059e-05}
{"train_loss": 0.062206462025642395, "global_step": 202933, "epoch": 2280, "lr": 1.4965231862361967e-05}
{"train_loss": 0.06499189138412476, "global_step": 202934, "epoch": 2280, "lr": 1.4964818245499884e-05}
{"train_loss": 0.06345388293266296, "global_step": 202935, "epoch": 2280, "lr": 1.4964404633347839e-05}
{"train_loss": 0.04907209798693657, "global_step": 202936, "epoch": 2280, "lr": 1.4963991025905916e-05}
{"train_loss": 0.047880612313747406, "global_step": 202937, "epoch": 2280, "lr": 1.4963577423174146e-05}
{"train_loss": 0.04644852504134178, "global_step": 202938, "epoch": 2280, "lr": 1.4963163825152609e-05}
{"train_loss": 0.10319279879331589, "global_step": 202939, "epoch": 2280, "lr": 1.4962750231841332e-05}
{"train_loss": 0.07773727178573608, "global_step": 202940, "epoch": 2280, "lr": 1.4962336643240404e-05}
{"train_loss": 0.03319490700960159, "global_step": 202941, "epoch": 2280, "lr": 1.4961923059349846e-05}
{"train_loss": 0.07290925085544586, "global_step": 202942, "epoch": 2280, "lr": 1.4961509480169727e-05}
{"train_loss": 0.07402153313159943, "global_step": 202943, "epoch": 2280, "lr": 1.4961095905700129e-05}
{"train_loss": 0.05491289123892784, "global_step": 202944, "epoch": 2280, "lr": 1.4960682335941063e-05}
{"train_loss": 0.10212273895740509, "global_step": 202945, "epoch": 2280, "lr": 1.496026877089262e-05}
{"train_loss": 0.051135070621967316, "global_step": 202946, "epoch": 2280, "lr": 1.4959855210554846e-05}
{"train_loss": 0.037518247961997986, "global_step": 202947, "epoch": 2280, "lr": 1.495944165492778e-05}
{"train_loss": 0.046044182032346725, "global_step": 202948, "epoch": 2280, "lr": 1.4959028104011502e-05}
{"train_loss": 0.0375080332159996, "global_step": 202949, "epoch": 2280, "lr": 1.4958614557806039e-05}
{"train_loss": 0.07412157952785492, "global_step": 202950, "epoch": 2280, "lr": 1.4958201016311479e-05}
{"train_loss": 0.08787651360034943, "global_step": 202951, "epoch": 2280, "lr": 1.4957787479527863e-05}
{"train_loss": 0.02910788543522358, "global_step": 202952, "epoch": 2280, "lr": 1.4957373947455228e-05}
{"train_loss": 0.06783249229192734, "global_step": 202953, "epoch": 2280, "lr": 1.4956960420093662e-05}
{"train_loss": 0.06491327285766602, "global_step": 202954, "epoch": 2280, "lr": 1.495654689744319e-05}
{"train_loss": 0.03518739342689514, "global_step": 202955, "epoch": 2280, "lr": 1.4956133379503906e-05}
{"train_loss": 0.03175409510731697, "global_step": 202956, "epoch": 2280, "lr": 1.495571986627582e-05}
{"train_loss": 0.032533105462789536, "global_step": 202957, "epoch": 2280, "lr": 1.4955306357759024e-05}
{"train_loss": 0.04571027308702469, "global_step": 202958, "epoch": 2280, "lr": 1.495489285395355e-05}
{"train_loss": 0.053087927401065826, "global_step": 202959, "epoch": 2280, "lr": 1.4954479354859462e-05}
{"train_loss": 0.10765944421291351, "global_step": 202960, "epoch": 2280, "lr": 1.495406586047683e-05}
{"train_loss": 0.06934072822332382, "global_step": 202961, "epoch": 2280, "lr": 1.4953652370805682e-05}
{"train_loss": 0.05767631158232689, "global_step": 202962, "epoch": 2280, "lr": 1.4953238885846104e-05}
{"train_loss": 0.08878347277641296, "global_step": 202963, "epoch": 2280, "lr": 1.4952825405598125e-05}
{"train_loss": 0.05539403110742569, "global_step": 202964, "epoch": 2280, "lr": 1.4952411930061822e-05}
{"train_loss": 0.0816667303442955, "global_step": 202965, "epoch": 2280, "lr": 1.4951998459237227e-05}
{"train_loss": 0.08233711868524551, "global_step": 202966, "epoch": 2280, "lr": 1.4951584993124423e-05}
{"train_loss": 0.14553862810134888, "global_step": 202967, "epoch": 2280, "lr": 1.4951171531723434e-05}
{"train_loss": 0.0586157888174057, "global_step": 202968, "epoch": 2280, "lr": 1.4950758075034349e-05}
{"train_loss": 0.11265408992767334, "global_step": 202969, "epoch": 2280, "lr": 1.4950344623057194e-05}
{"train_loss": 0.05854269489645958, "global_step": 202970, "epoch": 2280, "lr": 1.4949931175792053e-05}
{"train_loss": 0.04857591912150383, "global_step": 202971, "epoch": 2280, "lr": 1.494951773323895e-05}
{"train_loss": 0.06127750873565674, "global_step": 202972, "epoch": 2280, "lr": 1.4949104295397975e-05}
{"train_loss": 0.05127246305346489, "global_step": 202973, "epoch": 2280, "lr": 1.4948690862269143e-05}
{"train_loss": 0.05493147671222687, "global_step": 202974, "epoch": 2280, "lr": 1.4948277433852558e-05}
{"train_loss": 0.06725001335144043, "global_step": 202975, "epoch": 2280, "lr": 1.4947864010148227e-05}
{"train_loss": 0.06343860924243927, "global_step": 202976, "epoch": 2280, "lr": 1.4947450591156248e-05}
{"train_loss": 0.07792335748672485, "global_step": 202977, "epoch": 2280, "lr": 1.4947037176876638e-05}
{"train_loss": 0.08124911785125732, "global_step": 202978, "epoch": 2280, "lr": 1.4946623767309476e-05}
{"train_loss": 0.08838286995887756, "global_step": 202979, "epoch": 2280, "lr": 1.4946210362454827e-05}
{"train_loss": 0.06375152617692947, "global_step": 202980, "epoch": 2280, "lr": 1.4945796962312714e-05}
{"train_loss": 0.02553076669573784, "global_step": 202981, "epoch": 2280, "lr": 1.4945383566883232e-05}
{"train_loss": 0.08078354597091675, "global_step": 202982, "epoch": 2280, "lr": 1.4944970176166395e-05}
{"train_loss": 0.09785202145576477, "global_step": 202983, "epoch": 2280, "lr": 1.4944556790162295e-05}
{"train_loss": 0.08183195441961288, "global_step": 202984, "epoch": 2280, "lr": 1.4944143408870958e-05}
{"train_loss": 0.036871667951345444, "global_step": 202985, "epoch": 2280, "lr": 1.4943730032292475e-05}
{"train_loss": 0.13616855442523956, "global_step": 202986, "epoch": 2280, "lr": 1.4943316660426876e-05}
{"train_loss": 0.07147157937288284, "global_step": 202987, "epoch": 2280, "lr": 1.4942903293274201e-05}
{"train_loss": 0.03656116873025894, "global_step": 202988, "epoch": 2280, "lr": 1.4942489930834546e-05}
{"train_loss": 0.09089330583810806, "global_step": 202989, "epoch": 2280, "lr": 1.4942076573107933e-05}
{"train_loss": 0.059806112200021744, "global_step": 202990, "epoch": 2280, "lr": 1.4941663220094437e-05}
{"train_loss": 0.05562837794423103, "global_step": 202991, "epoch": 2280, "lr": 1.4941249871794117e-05}
{"train_loss": 0.10687702894210815, "global_step": 202992, "epoch": 2280, "lr": 1.4940836528206998e-05}
{"train_loss": 0.06284280866384506, "global_step": 202993, "epoch": 2280, "lr": 1.494042318933317e-05}
{"train_loss": 0.08408112078905106, "global_step": 202994, "epoch": 2280, "lr": 1.494000985517266e-05}
{"train_loss": 0.04775765538215637, "global_step": 202995, "epoch": 2280, "lr": 1.4939596525725541e-05}
{"train_loss": 0.062424078583717346, "global_step": 202996, "epoch": 2280, "lr": 1.493918320099188e-05}
{"train_loss": 0.08120227605104446, "global_step": 202997, "epoch": 2280, "lr": 1.4938769880971704e-05}
{"train_loss": 0.05481576919555664, "global_step": 202998, "epoch": 2280, "lr": 1.4938356565665096e-05}
{"train_loss": 0.10844390094280243, "global_step": 202999, "epoch": 2280, "lr": 1.4937943255072084e-05}
{"train_loss": 0.04916255921125412, "global_step": 203000, "epoch": 2280, "lr": 1.4937529949192757e-05}
{"train_loss": 0.04303021356463432, "global_step": 203001, "epoch": 2280, "lr": 1.4937116648027134e-05}
{"train_loss": 0.06517992168664932, "global_step": 203002, "epoch": 2280, "lr": 1.4936703351575304e-05}
{"train_loss": 0.04968245327472687, "global_step": 203003, "epoch": 2280, "lr": 1.4936290059837293e-05}
{"train_loss": 0.07374434918165207, "global_step": 203004, "epoch": 2280, "lr": 1.4935876772813184e-05}
{"train_loss": 0.04386464133858681, "global_step": 203005, "epoch": 2280, "lr": 1.4935463490502999e-05}
{"train_loss": 0.04994738847017288, "global_step": 203006, "epoch": 2280, "lr": 1.4935050212906837e-05}
{"train_loss": 0.10803757607936859, "global_step": 203007, "epoch": 2280, "lr": 1.493463694002471e-05}
{"train_loss": 0.06597245589233516, "global_step": 203008, "epoch": 2280, "lr": 1.4934223671856711e-05, "val_loss": 8.29279613494873, "train_action_mse_error": 22.804943084716797}
{"train_loss": 0.039693355560302734, "global_step": 203009, "epoch": 2281, "lr": 1.493381040840286e-05}
{"train_loss": 0.0704813003540039, "global_step": 203010, "epoch": 2281, "lr": 1.4933397149663248e-05}
{"train_loss": 0.08406762778759003, "global_step": 203011, "epoch": 2281, "lr": 1.4932983895637898e-05}
{"train_loss": 0.0564115084707737, "global_step": 203012, "epoch": 2281, "lr": 1.49325706463269e-05}
{"train_loss": 0.0797426849603653, "global_step": 203013, "epoch": 2281, "lr": 1.4932157401730268e-05}
{"train_loss": 0.029090195894241333, "global_step": 203014, "epoch": 2281, "lr": 1.4931744161848083e-05}
{"train_loss": 0.11148285120725632, "global_step": 203015, "epoch": 2281, "lr": 1.4931330926680415e-05}
{"train_loss": 0.047175630927085876, "global_step": 203016, "epoch": 2281, "lr": 1.4930917696227287e-05}
{"train_loss": 0.10245475172996521, "global_step": 203017, "epoch": 2281, "lr": 1.4930504470488782e-05}
{"train_loss": 0.05448323115706444, "global_step": 203018, "epoch": 2281, "lr": 1.4930091249464928e-05}
{"train_loss": 0.08942113071680069, "global_step": 203019, "epoch": 2281, "lr": 1.4929678033155814e-05}
{"train_loss": 0.030895709991455078, "global_step": 203020, "epoch": 2281, "lr": 1.4929264821561461e-05}
{"train_loss": 0.03667402267456055, "global_step": 203021, "epoch": 2281, "lr": 1.4928851614681955e-05}
{"train_loss": 0.09395453333854675, "global_step": 203022, "epoch": 2281, "lr": 1.492843841251732e-05}
{"train_loss": 0.07693219184875488, "global_step": 203023, "epoch": 2281, "lr": 1.4928025215067649e-05}
{"train_loss": 0.06972504407167435, "global_step": 203024, "epoch": 2281, "lr": 1.4927612022332976e-05}
{"train_loss": 0.03028758615255356, "global_step": 203025, "epoch": 2281, "lr": 1.4927198834313344e-05}
{"train_loss": 0.04600335657596588, "global_step": 203026, "epoch": 2281, "lr": 1.4926785651008835e-05}
{"train_loss": 0.06350640207529068, "global_step": 203027, "epoch": 2281, "lr": 1.492637247241948e-05}
{"train_loss": 0.06295225769281387, "global_step": 203028, "epoch": 2281, "lr": 1.4925959298545362e-05}
{"train_loss": 0.08732245862483978, "global_step": 203029, "epoch": 2281, "lr": 1.492554612938652e-05}
{"train_loss": 0.054864492267370224, "global_step": 203030, "epoch": 2281, "lr": 1.492513296494299e-05}
{"train_loss": 0.06358423084020615, "global_step": 203031, "epoch": 2281, "lr": 1.4924719805214871e-05}
{"train_loss": 0.06410552561283112, "global_step": 203032, "epoch": 2281, "lr": 1.492430665020218e-05}
{"train_loss": 0.023269273340702057, "global_step": 203033, "epoch": 2281, "lr": 1.4923893499904983e-05}
{"train_loss": 0.05349055305123329, "global_step": 203034, "epoch": 2281, "lr": 1.492348035432336e-05}
{"train_loss": 0.04185432568192482, "global_step": 203035, "epoch": 2281, "lr": 1.4923067213457332e-05}
{"train_loss": 0.029852289706468582, "global_step": 203036, "epoch": 2281, "lr": 1.4922654077306986e-05}
{"train_loss": 0.060943979769945145, "global_step": 203037, "epoch": 2281, "lr": 1.4922240945872346e-05}
{"train_loss": 0.03914903104305267, "global_step": 203038, "epoch": 2281, "lr": 1.49218278191535e-05}
{"train_loss": 0.061670489609241486, "global_step": 203039, "epoch": 2281, "lr": 1.4921414697150471e-05}
{"train_loss": 0.08208730816841125, "global_step": 203040, "epoch": 2281, "lr": 1.4921001579863342e-05}
{"train_loss": 0.03597836196422577, "global_step": 203041, "epoch": 2281, "lr": 1.4920588467292146e-05}
{"train_loss": 0.03449263796210289, "global_step": 203042, "epoch": 2281, "lr": 1.492017535943696e-05}
{"train_loss": 0.03168068081140518, "global_step": 203043, "epoch": 2281, "lr": 1.491976225629782e-05}
{"train_loss": 0.03192785009741783, "global_step": 203044, "epoch": 2281, "lr": 1.4919349157874802e-05}
{"train_loss": 0.05076400935649872, "global_step": 203045, "epoch": 2281, "lr": 1.4918936064167932e-05}
{"train_loss": 0.08408497273921967, "global_step": 203046, "epoch": 2281, "lr": 1.4918522975177302e-05}
{"train_loss": 0.06228369101881981, "global_step": 203047, "epoch": 2281, "lr": 1.4918109890902932e-05}
{"train_loss": 0.05519432574510574, "global_step": 203048, "epoch": 2281, "lr": 1.4917696811344912e-05}
{"train_loss": 0.04393918067216873, "global_step": 203049, "epoch": 2281, "lr": 1.4917283736503263e-05}
{"train_loss": 0.04960011690855026, "global_step": 203050, "epoch": 2281, "lr": 1.4916870666378057e-05}
{"train_loss": 0.040395718067884445, "global_step": 203051, "epoch": 2281, "lr": 1.4916457600969369e-05}
{"train_loss": 0.11110824346542358, "global_step": 203052, "epoch": 2281, "lr": 1.4916044540277214e-05}
{"train_loss": 0.06420554220676422, "global_step": 203053, "epoch": 2281, "lr": 1.491563148430169e-05}
{"train_loss": 0.06462367624044418, "global_step": 203054, "epoch": 2281, "lr": 1.4915218433042816e-05}
{"train_loss": 0.0351610966026783, "global_step": 203055, "epoch": 2281, "lr": 1.4914805386500675e-05}
{"train_loss": 0.04670463502407074, "global_step": 203056, "epoch": 2281, "lr": 1.4914392344675298e-05}
{"train_loss": 0.10311269015073776, "global_step": 203057, "epoch": 2281, "lr": 1.4913979307566773e-05}
{"train_loss": 0.07371290028095245, "global_step": 203058, "epoch": 2281, "lr": 1.4913566275175112e-05}
{"train_loss": 0.1012137308716774, "global_step": 203059, "epoch": 2281, "lr": 1.4913153247500416e-05}
{"train_loss": 0.05949705094099045, "global_step": 203060, "epoch": 2281, "lr": 1.4912740224542698e-05}
{"train_loss": 0.06079879403114319, "global_step": 203061, "epoch": 2281, "lr": 1.4912327206302057e-05}
{"train_loss": 0.0816117599606514, "global_step": 203062, "epoch": 2281, "lr": 1.4911914192778519e-05}
{"train_loss": 0.07058659195899963, "global_step": 203063, "epoch": 2281, "lr": 1.4911501183972138e-05}
{"train_loss": 0.07106519490480423, "global_step": 203064, "epoch": 2281, "lr": 1.4911088179882987e-05}
{"train_loss": 0.07699007540941238, "global_step": 203065, "epoch": 2281, "lr": 1.4910675180511103e-05}
{"train_loss": 0.03529682382941246, "global_step": 203066, "epoch": 2281, "lr": 1.491026218585656e-05}
{"train_loss": 0.0848669707775116, "global_step": 203067, "epoch": 2281, "lr": 1.4909849195919407e-05}
{"train_loss": 0.053658127784729004, "global_step": 203068, "epoch": 2281, "lr": 1.4909436210699685e-05}
{"train_loss": 0.0917503833770752, "global_step": 203069, "epoch": 2281, "lr": 1.4909023230197455e-05}
{"train_loss": 0.06549154967069626, "global_step": 203070, "epoch": 2281, "lr": 1.4908610254412802e-05}
{"train_loss": 0.03869881480932236, "global_step": 203071, "epoch": 2281, "lr": 1.4908197283345737e-05}
{"train_loss": 0.06708289682865143, "global_step": 203072, "epoch": 2281, "lr": 1.4907784316996354e-05}
{"train_loss": 0.06815403699874878, "global_step": 203073, "epoch": 2281, "lr": 1.490737135536468e-05}
{"train_loss": 0.04230593889951706, "global_step": 203074, "epoch": 2281, "lr": 1.4906958398450793e-05}
{"train_loss": 0.030957896262407303, "global_step": 203075, "epoch": 2281, "lr": 1.4906545446254728e-05}
{"train_loss": 0.09346871078014374, "global_step": 203076, "epoch": 2281, "lr": 1.490613249877656e-05}
{"train_loss": 0.08128192275762558, "global_step": 203077, "epoch": 2281, "lr": 1.4905719556016324e-05}
{"train_loss": 0.1183861568570137, "global_step": 203078, "epoch": 2281, "lr": 1.4905306617974096e-05}
{"train_loss": 0.03685379400849342, "global_step": 203079, "epoch": 2281, "lr": 1.4904893684649912e-05}
{"train_loss": 0.039581820368766785, "global_step": 203080, "epoch": 2281, "lr": 1.4904480756043848e-05}
{"train_loss": 0.04903902858495712, "global_step": 203081, "epoch": 2281, "lr": 1.4904067832155938e-05}
{"train_loss": 0.02718641795217991, "global_step": 203082, "epoch": 2281, "lr": 1.490365491298626e-05}
{"train_loss": 0.06437089294195175, "global_step": 203083, "epoch": 2281, "lr": 1.4903241998534845e-05}
{"train_loss": 0.04246854409575462, "global_step": 203084, "epoch": 2281, "lr": 1.4902829088801774e-05}
{"train_loss": 0.03766840323805809, "global_step": 203085, "epoch": 2281, "lr": 1.4902416183787076e-05}
{"train_loss": 0.02389156073331833, "global_step": 203086, "epoch": 2281, "lr": 1.4902003283490823e-05}
{"train_loss": 0.09453783929347992, "global_step": 203087, "epoch": 2281, "lr": 1.4901590387913078e-05}
{"train_loss": 0.08743126690387726, "global_step": 203088, "epoch": 2281, "lr": 1.4901177497053876e-05}
{"train_loss": 0.045922666788101196, "global_step": 203089, "epoch": 2281, "lr": 1.4900764610913292e-05}
{"train_loss": 0.0336398109793663, "global_step": 203090, "epoch": 2281, "lr": 1.4900351729491362e-05}
{"train_loss": 0.04538610205054283, "global_step": 203091, "epoch": 2281, "lr": 1.4899938852788165e-05}
{"train_loss": 0.07732010632753372, "global_step": 203092, "epoch": 2281, "lr": 1.4899525980803731e-05}
{"train_loss": 0.06363179534673691, "global_step": 203093, "epoch": 2281, "lr": 1.4899113113538143e-05}
{"train_loss": 0.06276966631412506, "global_step": 203094, "epoch": 2281, "lr": 1.489870025099142e-05}
{"train_loss": 0.03690805658698082, "global_step": 203095, "epoch": 2281, "lr": 1.4898287393163662e-05}
{"train_loss": 0.06104114279150963, "global_step": 203096, "epoch": 2281, "lr": 1.4897874540054884e-05}
{"train_loss": 0.05969757015366903, "global_step": 203097, "epoch": 2281, "lr": 1.4897461691665171e-05, "val_loss": 8.41338062286377}
{"train_loss": 0.05590188130736351, "global_step": 203098, "epoch": 2282, "lr": 1.4897048847994554e-05}
{"train_loss": 0.05054521933197975, "global_step": 203099, "epoch": 2282, "lr": 1.4896636009043113e-05}
{"train_loss": 0.0883663222193718, "global_step": 203100, "epoch": 2282, "lr": 1.4896223174810881e-05}
{"train_loss": 0.07146767526865005, "global_step": 203101, "epoch": 2282, "lr": 1.4895810345297939e-05}
{"train_loss": 0.028424914926290512, "global_step": 203102, "epoch": 2282, "lr": 1.4895397520504323e-05}
{"train_loss": 0.061178408563137054, "global_step": 203103, "epoch": 2282, "lr": 1.489498470043008e-05}
{"train_loss": 0.07875705510377884, "global_step": 203104, "epoch": 2282, "lr": 1.4894571885075293e-05}
{"train_loss": 0.10433436185121536, "global_step": 203105, "epoch": 2282, "lr": 1.489415907443999e-05}
{"train_loss": 0.07765133678913116, "global_step": 203106, "epoch": 2282, "lr": 1.4893746268524256e-05}
{"train_loss": 0.03025742992758751, "global_step": 203107, "epoch": 2282, "lr": 1.4893333467328108e-05}
{"train_loss": 0.0879574567079544, "global_step": 203108, "epoch": 2282, "lr": 1.4892920670851645e-05}
{"train_loss": 0.05824969708919525, "global_step": 203109, "epoch": 2282, "lr": 1.489250787909488e-05}
{"train_loss": 0.04813714325428009, "global_step": 203110, "epoch": 2282, "lr": 1.4892095092057911e-05}
{"train_loss": 0.08341683447360992, "global_step": 203111, "epoch": 2282, "lr": 1.489168230974075e-05}
{"train_loss": 0.04105023294687271, "global_step": 203112, "epoch": 2282, "lr": 1.4891269532143492e-05}
{"train_loss": 0.054834041744470596, "global_step": 203113, "epoch": 2282, "lr": 1.4890856759266164e-05}
{"train_loss": 0.1054459810256958, "global_step": 203114, "epoch": 2282, "lr": 1.4890443991108844e-05}
{"train_loss": 0.06530292332172394, "global_step": 203115, "epoch": 2282, "lr": 1.4890031227671558e-05}
{"train_loss": 0.0746011883020401, "global_step": 203116, "epoch": 2282, "lr": 1.4889618468954402e-05}
{"train_loss": 0.0480184406042099, "global_step": 203117, "epoch": 2282, "lr": 1.4889205714957389e-05}
{"train_loss": 0.039778679609298706, "global_step": 203118, "epoch": 2282, "lr": 1.488879296568061e-05}
{"train_loss": 0.06567125767469406, "global_step": 203119, "epoch": 2282, "lr": 1.4888380221124087e-05}
{"train_loss": 0.044862255454063416, "global_step": 203120, "epoch": 2282, "lr": 1.4887967481287912e-05}
{"train_loss": 0.04150798171758652, "global_step": 203121, "epoch": 2282, "lr": 1.4887554746172105e-05}
{"train_loss": 0.07421477138996124, "global_step": 203122, "epoch": 2282, "lr": 1.488714201577674e-05}
{"train_loss": 0.0755724161863327, "global_step": 203123, "epoch": 2282, "lr": 1.4886729290101886e-05}
{"train_loss": 0.04683838039636612, "global_step": 203124, "epoch": 2282, "lr": 1.4886316569147562e-05}
{"train_loss": 0.07058898359537125, "global_step": 203125, "epoch": 2282, "lr": 1.4885903852913869e-05}
{"train_loss": 0.07907931506633759, "global_step": 203126, "epoch": 2282, "lr": 1.4885491141400821e-05}
{"train_loss": 0.04956640303134918, "global_step": 203127, "epoch": 2282, "lr": 1.4885078434608501e-05}
{"train_loss": 0.09340803325176239, "global_step": 203128, "epoch": 2282, "lr": 1.488466573253694e-05}
{"train_loss": 0.0530107356607914, "global_step": 203129, "epoch": 2282, "lr": 1.4884253035186224e-05}
{"train_loss": 0.09243448823690414, "global_step": 203130, "epoch": 2282, "lr": 1.4883840342556383e-05}
{"train_loss": 0.09250316023826599, "global_step": 203131, "epoch": 2282, "lr": 1.4883427654647492e-05}
{"train_loss": 0.01585923694074154, "global_step": 203132, "epoch": 2282, "lr": 1.488301497145958e-05}
{"train_loss": 0.05412300303578377, "global_step": 203133, "epoch": 2282, "lr": 1.4882602292992731e-05}
{"train_loss": 0.07875305414199829, "global_step": 203134, "epoch": 2282, "lr": 1.4882189619246978e-05}
{"train_loss": 0.0680249035358429, "global_step": 203135, "epoch": 2282, "lr": 1.4881776950222399e-05}
{"train_loss": 0.10934171080589294, "global_step": 203136, "epoch": 2282, "lr": 1.4881364285919025e-05}
{"train_loss": 0.0362822562456131, "global_step": 203137, "epoch": 2282, "lr": 1.4880951626336937e-05}
{"train_loss": 0.07028522342443466, "global_step": 203138, "epoch": 2282, "lr": 1.488053897147616e-05}
{"train_loss": 0.09166745096445084, "global_step": 203139, "epoch": 2282, "lr": 1.4880126321336785e-05}
{"train_loss": 0.05072915554046631, "global_step": 203140, "epoch": 2282, "lr": 1.4879713675918844e-05}
{"train_loss": 0.07573778182268143, "global_step": 203141, "epoch": 2282, "lr": 1.4879301035222387e-05}
{"train_loss": 0.0451088547706604, "global_step": 203142, "epoch": 2282, "lr": 1.4878888399247492e-05}
{"train_loss": 0.08676217496395111, "global_step": 203143, "epoch": 2282, "lr": 1.4878475767994187e-05}
{"train_loss": 0.04170457273721695, "global_step": 203144, "epoch": 2282, "lr": 1.4878063141462562e-05}
{"train_loss": 0.0513104684650898, "global_step": 203145, "epoch": 2282, "lr": 1.4877650519652636e-05}
{"train_loss": 0.03560361638665199, "global_step": 203146, "epoch": 2282, "lr": 1.48772379025645e-05}
{"train_loss": 0.03301636502146721, "global_step": 203147, "epoch": 2282, "lr": 1.4876825290198171e-05}
{"train_loss": 0.06675980240106583, "global_step": 203148, "epoch": 2282, "lr": 1.4876412682553741e-05}
{"train_loss": 0.08654160052537918, "global_step": 203149, "epoch": 2282, "lr": 1.4876000079631236e-05}
{"train_loss": 0.0335441492497921, "global_step": 203150, "epoch": 2282, "lr": 1.4875587481430737e-05}
{"train_loss": 0.04793313145637512, "global_step": 203151, "epoch": 2282, "lr": 1.4875174887952275e-05}
{"train_loss": 0.06612330675125122, "global_step": 203152, "epoch": 2282, "lr": 1.4874762299195933e-05}
{"train_loss": 0.09677112847566605, "global_step": 203153, "epoch": 2282, "lr": 1.4874349715161733e-05}
{"train_loss": 0.07400041818618774, "global_step": 203154, "epoch": 2282, "lr": 1.4873937135849764e-05}
{"train_loss": 0.06713713705539703, "global_step": 203155, "epoch": 2282, "lr": 1.4873524561260055e-05}
{"train_loss": 0.04778621718287468, "global_step": 203156, "epoch": 2282, "lr": 1.4873111991392686e-05}
{"train_loss": 0.06281490623950958, "global_step": 203157, "epoch": 2282, "lr": 1.4872699426247677e-05}
{"train_loss": 0.07890854775905609, "global_step": 203158, "epoch": 2282, "lr": 1.487228686582513e-05}
{"train_loss": 0.04066634923219681, "global_step": 203159, "epoch": 2282, "lr": 1.4871874310125055e-05}
{"train_loss": 0.08069278299808502, "global_step": 203160, "epoch": 2282, "lr": 1.487146175914753e-05}
{"train_loss": 0.04958545044064522, "global_step": 203161, "epoch": 2282, "lr": 1.4871049212892623e-05}
{"train_loss": 0.07524297386407852, "global_step": 203162, "epoch": 2282, "lr": 1.487063667136036e-05}
{"train_loss": 0.05297141149640083, "global_step": 203163, "epoch": 2282, "lr": 1.487022413455083e-05}
{"train_loss": 0.04294775798916817, "global_step": 203164, "epoch": 2282, "lr": 1.4869811602464057e-05}
{"train_loss": 0.03254427760839462, "global_step": 203165, "epoch": 2282, "lr": 1.486939907510012e-05}
{"train_loss": 0.036631908267736435, "global_step": 203166, "epoch": 2282, "lr": 1.4868986552459047e-05}
{"train_loss": 0.039127036929130554, "global_step": 203167, "epoch": 2282, "lr": 1.4868574034540934e-05}
{"train_loss": 0.05021735653281212, "global_step": 203168, "epoch": 2282, "lr": 1.4868161521345787e-05}
{"train_loss": 0.014322501607239246, "global_step": 203169, "epoch": 2282, "lr": 1.486774901287371e-05}
{"train_loss": 0.08309809118509293, "global_step": 203170, "epoch": 2282, "lr": 1.4867336509124719e-05}
{"train_loss": 0.06696747988462448, "global_step": 203171, "epoch": 2282, "lr": 1.4866924010098903e-05}
{"train_loss": 0.09506870061159134, "global_step": 203172, "epoch": 2282, "lr": 1.4866511515796282e-05}
{"train_loss": 0.07776783406734467, "global_step": 203173, "epoch": 2282, "lr": 1.4866099026216951e-05}
{"train_loss": 0.06392610818147659, "global_step": 203174, "epoch": 2282, "lr": 1.4865686541360924e-05}
{"train_loss": 0.04019796475768089, "global_step": 203175, "epoch": 2282, "lr": 1.4865274061228295e-05}
{"train_loss": 0.07081608474254608, "global_step": 203176, "epoch": 2282, "lr": 1.4864861585819085e-05}
{"train_loss": 0.038328249007463455, "global_step": 203177, "epoch": 2282, "lr": 1.4864449115133384e-05}
{"train_loss": 0.10454930365085602, "global_step": 203178, "epoch": 2282, "lr": 1.4864036649171225e-05}
{"train_loss": 0.06797827780246735, "global_step": 203179, "epoch": 2282, "lr": 1.4863624187932657e-05}
{"train_loss": 0.04208570346236229, "global_step": 203180, "epoch": 2282, "lr": 1.4863211731417765e-05}
{"train_loss": 0.06100410223007202, "global_step": 203181, "epoch": 2282, "lr": 1.4862799279626565e-05}
{"train_loss": 0.035322654992341995, "global_step": 203182, "epoch": 2282, "lr": 1.486238683255915e-05}
{"train_loss": 0.043421171605587006, "global_step": 203183, "epoch": 2282, "lr": 1.4861974390215544e-05}
{"train_loss": 0.1099790707230568, "global_step": 203184, "epoch": 2282, "lr": 1.4861561952595831e-05}
{"train_loss": 0.04478796198964119, "global_step": 203185, "epoch": 2282, "lr": 1.4861149519700034e-05}
{"train_loss": 0.0618772597020764, "global_step": 203186, "epoch": 2282, "lr": 1.4860737091528249e-05, "val_loss": 8.34193229675293}
{"train_loss": 0.05275056138634682, "global_step": 203187, "epoch": 2283, "lr": 1.4860324668080494e-05}
{"train_loss": 0.07223434001207352, "global_step": 203188, "epoch": 2283, "lr": 1.4859912249356855e-05}
{"train_loss": 0.04004853591322899, "global_step": 203189, "epoch": 2283, "lr": 1.4859499835357355e-05}
{"train_loss": 0.07344332337379456, "global_step": 203190, "epoch": 2283, "lr": 1.485908742608208e-05}
{"train_loss": 0.07454994320869446, "global_step": 203191, "epoch": 2283, "lr": 1.485867502153106e-05}
{"train_loss": 0.1198691576719284, "global_step": 203192, "epoch": 2283, "lr": 1.4858262621704383e-05}
{"train_loss": 0.05073270946741104, "global_step": 203193, "epoch": 2283, "lr": 1.4857850226602061e-05}
{"train_loss": 0.031569112092256546, "global_step": 203194, "epoch": 2283, "lr": 1.4857437836224193e-05}
{"train_loss": 0.07373533397912979, "global_step": 203195, "epoch": 2283, "lr": 1.4857025450570794e-05}
{"train_loss": 0.06502094119787216, "global_step": 203196, "epoch": 2283, "lr": 1.4856613069641944e-05}
{"train_loss": 0.05279799923300743, "global_step": 203197, "epoch": 2283, "lr": 1.485620069343771e-05}
{"train_loss": 0.09148415923118591, "global_step": 203198, "epoch": 2283, "lr": 1.4855788321958114e-05}
{"train_loss": 0.08744915574789047, "global_step": 203199, "epoch": 2283, "lr": 1.4855375955203244e-05}
{"train_loss": 0.08377958089113235, "global_step": 203200, "epoch": 2283, "lr": 1.4854963593173127e-05}
{"train_loss": 0.03154310584068298, "global_step": 203201, "epoch": 2283, "lr": 1.4854551235867841e-05}
{"train_loss": 0.036689117550849915, "global_step": 203202, "epoch": 2283, "lr": 1.4854138883287422e-05}
{"train_loss": 0.07914343476295471, "global_step": 203203, "epoch": 2283, "lr": 1.485372653543195e-05}
{"train_loss": 0.05758441612124443, "global_step": 203204, "epoch": 2283, "lr": 1.4853314192301453e-05}
{"train_loss": 0.10331352800130844, "global_step": 203205, "epoch": 2283, "lr": 1.4852901853896012e-05}
{"train_loss": 0.1016443744301796, "global_step": 203206, "epoch": 2283, "lr": 1.4852489520215656e-05}
{"train_loss": 0.04528191685676575, "global_step": 203207, "epoch": 2283, "lr": 1.4852077191260472e-05}
{"train_loss": 0.07940758019685745, "global_step": 203208, "epoch": 2283, "lr": 1.4851664867030479e-05}
{"train_loss": 0.07902295887470245, "global_step": 203209, "epoch": 2283, "lr": 1.4851252547525768e-05}
{"train_loss": 0.03669377416372299, "global_step": 203210, "epoch": 2283, "lr": 1.4850840232746361e-05}
{"train_loss": 0.05470152199268341, "global_step": 203211, "epoch": 2283, "lr": 1.4850427922692346e-05}
{"train_loss": 0.05330492928624153, "global_step": 203212, "epoch": 2283, "lr": 1.485001561736375e-05}
{"train_loss": 0.0846511498093605, "global_step": 203213, "epoch": 2283, "lr": 1.4849603316760652e-05}
{"train_loss": 0.03715980798006058, "global_step": 203214, "epoch": 2283, "lr": 1.484919102088308e-05}
{"train_loss": 0.03386364132165909, "global_step": 203215, "epoch": 2283, "lr": 1.4848778729731128e-05}
{"train_loss": 0.08597928285598755, "global_step": 203216, "epoch": 2283, "lr": 1.4848366443304807e-05}
{"train_loss": 0.09446854144334793, "global_step": 203217, "epoch": 2283, "lr": 1.4847954161604217e-05}
{"train_loss": 0.09445788711309433, "global_step": 203218, "epoch": 2283, "lr": 1.4847541884629384e-05}
{"train_loss": 0.08938777446746826, "global_step": 203219, "epoch": 2283, "lr": 1.4847129612380362e-05}
{"train_loss": 0.06793343275785446, "global_step": 203220, "epoch": 2283, "lr": 1.4846717344857224e-05}
{"train_loss": 0.04655013978481293, "global_step": 203221, "epoch": 2283, "lr": 1.4846305082060002e-05}
{"train_loss": 0.02437923476099968, "global_step": 203222, "epoch": 2283, "lr": 1.4845892823988783e-05}
{"train_loss": 0.07796679437160492, "global_step": 203223, "epoch": 2283, "lr": 1.4845480570643593e-05}
{"train_loss": 0.043555695563554764, "global_step": 203224, "epoch": 2283, "lr": 1.4845068322024514e-05}
{"train_loss": 0.10692693293094635, "global_step": 203225, "epoch": 2283, "lr": 1.4844656078131564e-05}
{"train_loss": 0.0960383340716362, "global_step": 203226, "epoch": 2283, "lr": 1.4844243838964843e-05}
{"train_loss": 0.04806194826960564, "global_step": 203227, "epoch": 2283, "lr": 1.4843831604524372e-05}
{"train_loss": 0.07375723123550415, "global_step": 203228, "epoch": 2283, "lr": 1.484341937481023e-05}
{"train_loss": 0.061835262924432755, "global_step": 203229, "epoch": 2283, "lr": 1.4843007149822447e-05}
{"train_loss": 0.0513930469751358, "global_step": 203230, "epoch": 2283, "lr": 1.484259492956111e-05}
{"train_loss": 0.0461265854537487, "global_step": 203231, "epoch": 2283, "lr": 1.4842182714026242e-05}
{"train_loss": 0.10079173743724823, "global_step": 203232, "epoch": 2283, "lr": 1.4841770503217917e-05}
{"train_loss": 0.06571385264396667, "global_step": 203233, "epoch": 2283, "lr": 1.4841358297136198e-05}
{"train_loss": 0.054929692298173904, "global_step": 203234, "epoch": 2283, "lr": 1.4840946095781117e-05}
{"train_loss": 0.08213844150304794, "global_step": 203235, "epoch": 2283, "lr": 1.4840533899152753e-05}
{"train_loss": 0.03811173886060715, "global_step": 203236, "epoch": 2283, "lr": 1.4840121707251142e-05}
{"train_loss": 0.07303958386182785, "global_step": 203237, "epoch": 2283, "lr": 1.4839709520076362e-05}
{"train_loss": 0.060163550078868866, "global_step": 203238, "epoch": 2283, "lr": 1.4839297337628438e-05}
{"train_loss": 0.05653304606676102, "global_step": 203239, "epoch": 2283, "lr": 1.4838885159907456e-05}
{"train_loss": 0.06099032610654831, "global_step": 203240, "epoch": 2283, "lr": 1.4838472986913442e-05}
{"train_loss": 0.05096710845828056, "global_step": 203241, "epoch": 2283, "lr": 1.4838060818646481e-05}
{"train_loss": 0.05846215412020683, "global_step": 203242, "epoch": 2283, "lr": 1.48376486551066e-05}
{"train_loss": 0.11574437469244003, "global_step": 203243, "epoch": 2283, "lr": 1.4837236496293888e-05}
{"train_loss": 0.06871400773525238, "global_step": 203244, "epoch": 2283, "lr": 1.483682434220836e-05}
{"train_loss": 0.07699660956859589, "global_step": 203245, "epoch": 2283, "lr": 1.4836412192850114e-05}
{"train_loss": 0.03179582580924034, "global_step": 203246, "epoch": 2283, "lr": 1.4836000048219161e-05}
{"train_loss": 0.06543342024087906, "global_step": 203247, "epoch": 2283, "lr": 1.4835587908315602e-05}
{"train_loss": 0.027384191751480103, "global_step": 203248, "epoch": 2283, "lr": 1.4835175773139448e-05}
{"train_loss": 0.06513621658086777, "global_step": 203249, "epoch": 2283, "lr": 1.483476364269078e-05}
{"train_loss": 0.06294330209493637, "global_step": 203250, "epoch": 2283, "lr": 1.4834351516969664e-05}
{"train_loss": 0.04076642170548439, "global_step": 203251, "epoch": 2283, "lr": 1.4833939395976138e-05}
{"train_loss": 0.06623909622430801, "global_step": 203252, "epoch": 2283, "lr": 1.4833527279710247e-05}
{"train_loss": 0.06422173231840134, "global_step": 203253, "epoch": 2283, "lr": 1.4833115168172074e-05}
{"train_loss": 0.09219866245985031, "global_step": 203254, "epoch": 2283, "lr": 1.4832703061361646e-05}
{"train_loss": 0.045366913080215454, "global_step": 203255, "epoch": 2283, "lr": 1.483229095927905e-05}
{"train_loss": 0.06366777420043945, "global_step": 203256, "epoch": 2283, "lr": 1.483187886192432e-05}
{"train_loss": 0.08791551738977432, "global_step": 203257, "epoch": 2283, "lr": 1.4831466769297504e-05}
{"train_loss": 0.03558638319373131, "global_step": 203258, "epoch": 2283, "lr": 1.4831054681398682e-05}
{"train_loss": 0.09481833875179291, "global_step": 203259, "epoch": 2283, "lr": 1.4830642598227878e-05}
{"train_loss": 0.056713588535785675, "global_step": 203260, "epoch": 2283, "lr": 1.4830230519785183e-05}
{"train_loss": 0.09280586242675781, "global_step": 203261, "epoch": 2283, "lr": 1.482981844607062e-05}
{"train_loss": 0.08932200819253922, "global_step": 203262, "epoch": 2283, "lr": 1.4829406377084271e-05}
{"train_loss": 0.09755968302488327, "global_step": 203263, "epoch": 2283, "lr": 1.4828994312826166e-05}
{"train_loss": 0.07163366675376892, "global_step": 203264, "epoch": 2283, "lr": 1.4828582253296392e-05}
{"train_loss": 0.06241614744067192, "global_step": 203265, "epoch": 2283, "lr": 1.4828170198494967e-05}
{"train_loss": 0.08507005870342255, "global_step": 203266, "epoch": 2283, "lr": 1.482775814842199e-05}
{"train_loss": 0.08567089587450027, "global_step": 203267, "epoch": 2283, "lr": 1.4827346103077466e-05}
{"train_loss": 0.06989777833223343, "global_step": 203268, "epoch": 2283, "lr": 1.4826934062461483e-05}
{"train_loss": 0.11213026195764542, "global_step": 203269, "epoch": 2283, "lr": 1.4826522026574101e-05}
{"train_loss": 0.04410257562994957, "global_step": 203270, "epoch": 2283, "lr": 1.4826109995415355e-05}
{"train_loss": 0.03664310276508331, "global_step": 203271, "epoch": 2283, "lr": 1.482569796898532e-05}
{"train_loss": 0.0427556112408638, "global_step": 203272, "epoch": 2283, "lr": 1.4825285947284029e-05}
{"train_loss": 0.04129279404878616, "global_step": 203273, "epoch": 2283, "lr": 1.4824873930311562e-05}
{"train_loss": 0.08723889291286469, "global_step": 203274, "epoch": 2283, "lr": 1.482446191806795e-05}
{"train_loss": 0.06642800983920526, "global_step": 203275, "epoch": 2283, "lr": 1.4824049910553272e-05, "val_loss": 8.524861335754395}
{"train_loss": 0.05076977610588074, "global_step": 203276, "epoch": 2284, "lr": 1.4823637907767557e-05}
{"train_loss": 0.06531146168708801, "global_step": 203277, "epoch": 2284, "lr": 1.4823225909710897e-05}
{"train_loss": 0.09032319486141205, "global_step": 203278, "epoch": 2284, "lr": 1.4822813916383304e-05}
{"train_loss": 0.04752819240093231, "global_step": 203279, "epoch": 2284, "lr": 1.4822401927784869e-05}
{"train_loss": 0.05539274960756302, "global_step": 203280, "epoch": 2284, "lr": 1.482198994391562e-05}
{"train_loss": 0.06765340268611908, "global_step": 203281, "epoch": 2284, "lr": 1.482157796477564e-05}
{"train_loss": 0.056312087923288345, "global_step": 203282, "epoch": 2284, "lr": 1.482116599036496e-05}
{"train_loss": 0.08969596028327942, "global_step": 203283, "epoch": 2284, "lr": 1.4820754020683653e-05}
{"train_loss": 0.0480697937309742, "global_step": 203284, "epoch": 2284, "lr": 1.4820342055731757e-05}
{"train_loss": 0.04855218529701233, "global_step": 203285, "epoch": 2284, "lr": 1.4819930095509349e-05}
{"train_loss": 0.05605769529938698, "global_step": 203286, "epoch": 2284, "lr": 1.4819518140016464e-05}
{"train_loss": 0.03298244625329971, "global_step": 203287, "epoch": 2284, "lr": 1.4819106189253157e-05}
{"train_loss": 0.024883875623345375, "global_step": 203288, "epoch": 2284, "lr": 1.4818694243219512e-05}
{"train_loss": 0.11902511864900589, "global_step": 203289, "epoch": 2284, "lr": 1.4818282301915564e-05}
{"train_loss": 0.05656031146645546, "global_step": 203290, "epoch": 2284, "lr": 1.4817870365341357e-05}
{"train_loss": 0.06067601218819618, "global_step": 203291, "epoch": 2284, "lr": 1.4817458433496973e-05}
{"train_loss": 0.08190737664699554, "global_step": 203292, "epoch": 2284, "lr": 1.4817046506382432e-05}
{"train_loss": 0.02555704675614834, "global_step": 203293, "epoch": 2284, "lr": 1.4816634583997829e-05}
{"train_loss": 0.05371071770787239, "global_step": 203294, "epoch": 2284, "lr": 1.4816222666343205e-05}
{"train_loss": 0.11064541339874268, "global_step": 203295, "epoch": 2284, "lr": 1.481581075341859e-05}
{"train_loss": 0.06382589787244797, "global_step": 203296, "epoch": 2284, "lr": 1.4815398845224076e-05}
{"train_loss": 0.07412978261709213, "global_step": 203297, "epoch": 2284, "lr": 1.4814986941759689e-05}
{"train_loss": 0.0635138526558876, "global_step": 203298, "epoch": 2284, "lr": 1.4814575043025513e-05}
{"train_loss": 0.047723349183797836, "global_step": 203299, "epoch": 2284, "lr": 1.4814163149021576e-05}
{"train_loss": 0.04909062758088112, "global_step": 203300, "epoch": 2284, "lr": 1.481375125974796e-05}
{"train_loss": 0.021173356100916862, "global_step": 203301, "epoch": 2284, "lr": 1.4813339375204688e-05}
{"train_loss": 0.10308045893907547, "global_step": 203302, "epoch": 2284, "lr": 1.4812927495391855e-05}
{"train_loss": 0.07879270613193512, "global_step": 203303, "epoch": 2284, "lr": 1.4812515620309475e-05}
{"train_loss": 0.10323110222816467, "global_step": 203304, "epoch": 2284, "lr": 1.4812103749957623e-05}
{"train_loss": 0.07138202339410782, "global_step": 203305, "epoch": 2284, "lr": 1.4811691884336371e-05}
{"train_loss": 0.04209854081273079, "global_step": 203306, "epoch": 2284, "lr": 1.481128002344574e-05}
{"train_loss": 0.033821821212768555, "global_step": 203307, "epoch": 2284, "lr": 1.4810868167285824e-05}
{"train_loss": 0.0639863982796669, "global_step": 203308, "epoch": 2284, "lr": 1.481045631585664e-05}
{"train_loss": 0.03965998813509941, "global_step": 203309, "epoch": 2284, "lr": 1.4810044469158279e-05}
{"train_loss": 0.05899228900671005, "global_step": 203310, "epoch": 2284, "lr": 1.4809632627190762e-05}
{"train_loss": 0.042006976902484894, "global_step": 203311, "epoch": 2284, "lr": 1.480922078995417e-05}
{"train_loss": 0.027179596945643425, "global_step": 203312, "epoch": 2284, "lr": 1.4808808957448545e-05}
{"train_loss": 0.06199564039707184, "global_step": 203313, "epoch": 2284, "lr": 1.4808397129673952e-05}
{"train_loss": 0.016623780131340027, "global_step": 203314, "epoch": 2284, "lr": 1.4807985306630435e-05}
{"train_loss": 0.0808653011918068, "global_step": 203315, "epoch": 2284, "lr": 1.480757348831806e-05}
{"train_loss": 0.09196804463863373, "global_step": 203316, "epoch": 2284, "lr": 1.4807161674736869e-05}
{"train_loss": 0.07119052857160568, "global_step": 203317, "epoch": 2284, "lr": 1.4806749865886943e-05}
{"train_loss": 0.08833230286836624, "global_step": 203318, "epoch": 2284, "lr": 1.4806338061768304e-05}
{"train_loss": 0.08935151249170303, "global_step": 203319, "epoch": 2284, "lr": 1.4805926262381037e-05}
{"train_loss": 0.09416864812374115, "global_step": 203320, "epoch": 2284, "lr": 1.4805514467725174e-05}
{"train_loss": 0.08914021402597427, "global_step": 203321, "epoch": 2284, "lr": 1.4805102677800791e-05}
{"train_loss": 0.018321596086025238, "global_step": 203322, "epoch": 2284, "lr": 1.480469089260792e-05}
{"train_loss": 0.08390960097312927, "global_step": 203323, "epoch": 2284, "lr": 1.4804279112146624e-05}
{"train_loss": 0.08449945598840714, "global_step": 203324, "epoch": 2284, "lr": 1.4803867336416987e-05}
{"train_loss": 0.07133512198925018, "global_step": 203325, "epoch": 2284, "lr": 1.4803455565419022e-05}
{"train_loss": 0.06678763031959534, "global_step": 203326, "epoch": 2284, "lr": 1.4803043799152815e-05}
{"train_loss": 0.04854070395231247, "global_step": 203327, "epoch": 2284, "lr": 1.4802632037618402e-05}
{"train_loss": 0.03478311374783516, "global_step": 203328, "epoch": 2284, "lr": 1.4802220280815859e-05}
{"train_loss": 0.07733291387557983, "global_step": 203329, "epoch": 2284, "lr": 1.4801808528745226e-05}
{"train_loss": 0.09226840734481812, "global_step": 203330, "epoch": 2284, "lr": 1.4801396781406547e-05}
{"train_loss": 0.027592888101935387, "global_step": 203331, "epoch": 2284, "lr": 1.4800985038799903e-05}
{"train_loss": 0.07548859715461731, "global_step": 203332, "epoch": 2284, "lr": 1.480057330092533e-05}
{"train_loss": 0.03339634835720062, "global_step": 203333, "epoch": 2284, "lr": 1.48001615677829e-05}
{"train_loss": 0.071536123752594, "global_step": 203334, "epoch": 2284, "lr": 1.4799749839372661e-05}
{"train_loss": 0.1096344143152237, "global_step": 203335, "epoch": 2284, "lr": 1.479933811569465e-05}
{"train_loss": 0.047152165323495865, "global_step": 203336, "epoch": 2284, "lr": 1.4798926396748958e-05}
{"train_loss": 0.03794845938682556, "global_step": 203337, "epoch": 2284, "lr": 1.4798514682535602e-05}
{"train_loss": 0.07196369022130966, "global_step": 203338, "epoch": 2284, "lr": 1.4798102973054672e-05}
{"train_loss": 0.06513170152902603, "global_step": 203339, "epoch": 2284, "lr": 1.4797691268306196e-05}
{"train_loss": 0.04125319421291351, "global_step": 203340, "epoch": 2284, "lr": 1.479727956829024e-05}
{"train_loss": 0.059347543865442276, "global_step": 203341, "epoch": 2284, "lr": 1.4796867873006876e-05}
{"train_loss": 0.05504275858402252, "global_step": 203342, "epoch": 2284, "lr": 1.4796456182456126e-05}
{"train_loss": 0.07953505963087082, "global_step": 203343, "epoch": 2284, "lr": 1.4796044496638084e-05}
{"train_loss": 0.07776691764593124, "global_step": 203344, "epoch": 2284, "lr": 1.4795632815552763e-05}
{"train_loss": 0.05460913106799126, "global_step": 203345, "epoch": 2284, "lr": 1.479522113920026e-05}
{"train_loss": 0.04698372259736061, "global_step": 203346, "epoch": 2284, "lr": 1.4794809467580595e-05}
{"train_loss": 0.06865767389535904, "global_step": 203347, "epoch": 2284, "lr": 1.4794397800693849e-05}
{"train_loss": 0.07949032634496689, "global_step": 203348, "epoch": 2284, "lr": 1.4793986138540056e-05}
{"train_loss": 0.08166170120239258, "global_step": 203349, "epoch": 2284, "lr": 1.4793574481119299e-05}
{"train_loss": 0.0394061915576458, "global_step": 203350, "epoch": 2284, "lr": 1.4793162828431601e-05}
{"train_loss": 0.05151300132274628, "global_step": 203351, "epoch": 2284, "lr": 1.4792751180477049e-05}
{"train_loss": 0.0668473169207573, "global_step": 203352, "epoch": 2284, "lr": 1.4792339537255662e-05}
{"train_loss": 0.05097884312272072, "global_step": 203353, "epoch": 2284, "lr": 1.4791927898767538e-05}
{"train_loss": 0.0699835866689682, "global_step": 203354, "epoch": 2284, "lr": 1.479151626501269e-05}
{"train_loss": 0.07004883140325546, "global_step": 203355, "epoch": 2284, "lr": 1.479110463599121e-05}
{"train_loss": 0.06630060076713562, "global_step": 203356, "epoch": 2284, "lr": 1.4790693011703122e-05}
{"train_loss": 0.06420785188674927, "global_step": 203357, "epoch": 2284, "lr": 1.4790281392148514e-05}
{"train_loss": 0.042257487773895264, "global_step": 203358, "epoch": 2284, "lr": 1.4789869777327403e-05}
{"train_loss": 0.060165517032146454, "global_step": 203359, "epoch": 2284, "lr": 1.4789458167239867e-05}
{"train_loss": 0.06892334669828415, "global_step": 203360, "epoch": 2284, "lr": 1.4789046561885978e-05}
{"train_loss": 0.08359380811452866, "global_step": 203361, "epoch": 2284, "lr": 1.4788634961265751e-05}
{"train_loss": 0.09563261270523071, "global_step": 203362, "epoch": 2284, "lr": 1.4788223365379283e-05}
{"train_loss": 0.07169579714536667, "global_step": 203363, "epoch": 2284, "lr": 1.4787811774226596e-05}
{"train_loss": 0.0634754134153717, "global_step": 203364, "epoch": 2284, "lr": 1.4787400187807765e-05, "val_loss": 8.228317260742188}
{"train_loss": 0.0796104222536087, "global_step": 203365, "epoch": 2285, "lr": 1.4786988606122831e-05}
{"train_loss": 0.04453146830201149, "global_step": 203366, "epoch": 2285, "lr": 1.4786577029171873e-05}
{"train_loss": 0.11289612948894501, "global_step": 203367, "epoch": 2285, "lr": 1.4786165456954926e-05}
{"train_loss": 0.09537719190120697, "global_step": 203368, "epoch": 2285, "lr": 1.4785753889472037e-05}
{"train_loss": 0.06273844093084335, "global_step": 203369, "epoch": 2285, "lr": 1.478534232672329e-05}
{"train_loss": 0.025073610246181488, "global_step": 203370, "epoch": 2285, "lr": 1.4784930768708705e-05}
{"train_loss": 0.06850318610668182, "global_step": 203371, "epoch": 2285, "lr": 1.4784519215428377e-05}
{"train_loss": 0.05620964244008064, "global_step": 203372, "epoch": 2285, "lr": 1.4784107666882342e-05}
{"train_loss": 0.04797477647662163, "global_step": 203373, "epoch": 2285, "lr": 1.4783696123070634e-05}
{"train_loss": 0.057835653424263, "global_step": 203374, "epoch": 2285, "lr": 1.4783284583993346e-05}
{"train_loss": 0.06686846911907196, "global_step": 203375, "epoch": 2285, "lr": 1.4782873049650503e-05}
{"train_loss": 0.08773083984851837, "global_step": 203376, "epoch": 2285, "lr": 1.4782461520042168e-05}
{"train_loss": 0.09452572464942932, "global_step": 203377, "epoch": 2285, "lr": 1.4782049995168423e-05}
{"train_loss": 0.07356313616037369, "global_step": 203378, "epoch": 2285, "lr": 1.4781638475029285e-05}
{"train_loss": 0.08531472831964493, "global_step": 203379, "epoch": 2285, "lr": 1.4781226959624844e-05}
{"train_loss": 0.03560062497854233, "global_step": 203380, "epoch": 2285, "lr": 1.4780815448955121e-05}
{"train_loss": 0.10166274756193161, "global_step": 203381, "epoch": 2285, "lr": 1.47804039430202e-05}
{"train_loss": 0.04472718760371208, "global_step": 203382, "epoch": 2285, "lr": 1.4779992441820112e-05}
{"train_loss": 0.057667218148708344, "global_step": 203383, "epoch": 2285, "lr": 1.4779580945354943e-05}
{"train_loss": 0.09169790148735046, "global_step": 203384, "epoch": 2285, "lr": 1.4779169453624708e-05}
{"train_loss": 0.07793349772691727, "global_step": 203385, "epoch": 2285, "lr": 1.4778757966629502e-05}
{"train_loss": 0.05542458966374397, "global_step": 203386, "epoch": 2285, "lr": 1.4778346484369348e-05}
{"train_loss": 0.04174356907606125, "global_step": 203387, "epoch": 2285, "lr": 1.4777935006844335e-05}
{"train_loss": 0.12154411524534225, "global_step": 203388, "epoch": 2285, "lr": 1.4777523534054483e-05}
{"train_loss": 0.02481110207736492, "global_step": 203389, "epoch": 2285, "lr": 1.4777112065999876e-05}
{"train_loss": 0.06344591826200485, "global_step": 203390, "epoch": 2285, "lr": 1.4776700602680542e-05}
{"train_loss": 0.08900655806064606, "global_step": 203391, "epoch": 2285, "lr": 1.4776289144096567e-05}
{"train_loss": 0.04069674387574196, "global_step": 203392, "epoch": 2285, "lr": 1.4775877690247974e-05}
{"train_loss": 0.04519437626004219, "global_step": 203393, "epoch": 2285, "lr": 1.4775466241134855e-05}
{"train_loss": 0.016485102474689484, "global_step": 203394, "epoch": 2285, "lr": 1.4775054796757226e-05}
{"train_loss": 0.053141988813877106, "global_step": 203395, "epoch": 2285, "lr": 1.4774643357115159e-05}
{"train_loss": 0.13303141295909882, "global_step": 203396, "epoch": 2285, "lr": 1.477423192220873e-05}
{"train_loss": 0.06845945864915848, "global_step": 203397, "epoch": 2285, "lr": 1.4773820492037965e-05}
{"train_loss": 0.04860442876815796, "global_step": 203398, "epoch": 2285, "lr": 1.4773409066602939e-05}
{"train_loss": 0.030929680913686752, "global_step": 203399, "epoch": 2285, "lr": 1.4772997645903686e-05}
{"train_loss": 0.03884882852435112, "global_step": 203400, "epoch": 2285, "lr": 1.4772586229940294e-05}
{"train_loss": 0.13653574883937836, "global_step": 203401, "epoch": 2285, "lr": 1.4772174818712775e-05}
{"train_loss": 0.030131585896015167, "global_step": 203402, "epoch": 2285, "lr": 1.4771763412221229e-05}
{"train_loss": 0.05838533118367195, "global_step": 203403, "epoch": 2285, "lr": 1.4771352010465672e-05}
{"train_loss": 0.06315978616476059, "global_step": 203404, "epoch": 2285, "lr": 1.4770940613446194e-05}
{"train_loss": 0.041884057223796844, "global_step": 203405, "epoch": 2285, "lr": 1.4770529221162831e-05}
{"train_loss": 0.05793571099638939, "global_step": 203406, "epoch": 2285, "lr": 1.4770117833615631e-05}
{"train_loss": 0.039403125643730164, "global_step": 203407, "epoch": 2285, "lr": 1.4769706450804672e-05}
{"train_loss": 0.07154128700494766, "global_step": 203408, "epoch": 2285, "lr": 1.4769295072729983e-05}
{"train_loss": 0.06406737864017487, "global_step": 203409, "epoch": 2285, "lr": 1.4768883699391644e-05}
{"train_loss": 0.06776570528745651, "global_step": 203410, "epoch": 2285, "lr": 1.4768472330789701e-05}
{"train_loss": 0.08272607624530792, "global_step": 203411, "epoch": 2285, "lr": 1.476806096692419e-05}
{"train_loss": 0.0793188214302063, "global_step": 203412, "epoch": 2285, "lr": 1.4767649607795203e-05}
{"train_loss": 0.07033295929431915, "global_step": 203413, "epoch": 2285, "lr": 1.476723825340276e-05}
{"train_loss": 0.028941571712493896, "global_step": 203414, "epoch": 2285, "lr": 1.4766826903746933e-05}
{"train_loss": 0.04975719377398491, "global_step": 203415, "epoch": 2285, "lr": 1.4766415558827789e-05}
{"train_loss": 0.058704786002635956, "global_step": 203416, "epoch": 2285, "lr": 1.4766004218645357e-05}
{"train_loss": 0.04811513423919678, "global_step": 203417, "epoch": 2285, "lr": 1.4765592883199719e-05}
{"train_loss": 0.09809114784002304, "global_step": 203418, "epoch": 2285, "lr": 1.476518155249091e-05}
{"train_loss": 0.07583748549222946, "global_step": 203419, "epoch": 2285, "lr": 1.4764770226518998e-05}
{"train_loss": 0.06447931379079819, "global_step": 203420, "epoch": 2285, "lr": 1.4764358905284025e-05}
{"train_loss": 0.11550762504339218, "global_step": 203421, "epoch": 2285, "lr": 1.476394758878607e-05}
{"train_loss": 0.05218064785003662, "global_step": 203422, "epoch": 2285, "lr": 1.4763536277025153e-05}
{"train_loss": 0.11572767794132233, "global_step": 203423, "epoch": 2285, "lr": 1.4763124970001369e-05}
{"train_loss": 0.07133093476295471, "global_step": 203424, "epoch": 2285, "lr": 1.4762713667714733e-05}
{"train_loss": 0.055881135165691376, "global_step": 203425, "epoch": 2285, "lr": 1.4762302370165343e-05}
{"train_loss": 0.04256058484315872, "global_step": 203426, "epoch": 2285, "lr": 1.4761891077353207e-05}
{"train_loss": 0.03057808056473732, "global_step": 203427, "epoch": 2285, "lr": 1.4761479789278426e-05}
{"train_loss": 0.04772164672613144, "global_step": 203428, "epoch": 2285, "lr": 1.4761068505941022e-05}
{"train_loss": 0.051130734384059906, "global_step": 203429, "epoch": 2285, "lr": 1.4760657227341073e-05}
{"train_loss": 0.1101466715335846, "global_step": 203430, "epoch": 2285, "lr": 1.4760245953478613e-05}
{"train_loss": 0.04115388169884682, "global_step": 203431, "epoch": 2285, "lr": 1.4759834684353707e-05}
{"train_loss": 0.09523051232099533, "global_step": 203432, "epoch": 2285, "lr": 1.4759423419966429e-05}
{"train_loss": 0.05602390319108963, "global_step": 203433, "epoch": 2285, "lr": 1.47590121603168e-05}
{"train_loss": 0.08271843940019608, "global_step": 203434, "epoch": 2285, "lr": 1.4758600905404906e-05}
{"train_loss": 0.0626615509390831, "global_step": 203435, "epoch": 2285, "lr": 1.475818965523077e-05}
{"train_loss": 0.04195322468876839, "global_step": 203436, "epoch": 2285, "lr": 1.4757778409794487e-05}
{"train_loss": 0.09069665521383286, "global_step": 203437, "epoch": 2285, "lr": 1.4757367169096076e-05}
{"train_loss": 0.10317710787057877, "global_step": 203438, "epoch": 2285, "lr": 1.4756955933135618e-05}
{"train_loss": 0.05501489341259003, "global_step": 203439, "epoch": 2285, "lr": 1.475654470191315e-05}
{"train_loss": 0.06954627484083176, "global_step": 203440, "epoch": 2285, "lr": 1.475613347542874e-05}
{"train_loss": 0.06660129129886627, "global_step": 203441, "epoch": 2285, "lr": 1.4755722253682431e-05}
{"train_loss": 0.0333297960460186, "global_step": 203442, "epoch": 2285, "lr": 1.4755311036674302e-05}
{"train_loss": 0.050667840987443924, "global_step": 203443, "epoch": 2285, "lr": 1.4754899824404373e-05}
{"train_loss": 0.042797766625881195, "global_step": 203444, "epoch": 2285, "lr": 1.475448861687273e-05}
{"train_loss": 0.051220301538705826, "global_step": 203445, "epoch": 2285, "lr": 1.475407741407942e-05}
{"train_loss": 0.09880045801401138, "global_step": 203446, "epoch": 2285, "lr": 1.475366621602448e-05}
{"train_loss": 0.03676377236843109, "global_step": 203447, "epoch": 2285, "lr": 1.4753255022707996e-05}
{"train_loss": 0.04077088460326195, "global_step": 203448, "epoch": 2285, "lr": 1.4752843834129992e-05}
{"train_loss": 0.059053242206573486, "global_step": 203449, "epoch": 2285, "lr": 1.4752432650290548e-05}
{"train_loss": 0.06553293019533157, "global_step": 203450, "epoch": 2285, "lr": 1.4752021471189703e-05}
{"train_loss": 0.030978236347436905, "global_step": 203451, "epoch": 2285, "lr": 1.4751610296827528e-05}
{"train_loss": 0.06478267163038254, "global_step": 203452, "epoch": 2285, "lr": 1.4751199127204058e-05}
{"train_loss": 0.06443843009013138, "global_step": 203453, "epoch": 2285, "lr": 1.4750787962319373e-05, "val_loss": 8.040529251098633, "train_action_mse_error": 14.451838493347168}
{"train_loss": 0.06113214045763016, "global_step": 203454, "epoch": 2286, "lr": 1.47503768021735e-05}
{"train_loss": 0.05633404850959778, "global_step": 203455, "epoch": 2286, "lr": 1.4749965646766522e-05}
{"train_loss": 0.047384876757860184, "global_step": 203456, "epoch": 2286, "lr": 1.4749554496098472e-05}
{"train_loss": 0.07032207399606705, "global_step": 203457, "epoch": 2286, "lr": 1.4749143350169426e-05}
{"train_loss": 0.060067545622587204, "global_step": 203458, "epoch": 2286, "lr": 1.4748732208979416e-05}
{"train_loss": 0.033264148980379105, "global_step": 203459, "epoch": 2286, "lr": 1.474832107252852e-05}
{"train_loss": 0.04654041677713394, "global_step": 203460, "epoch": 2286, "lr": 1.4747909940816768e-05}
{"train_loss": 0.07799500226974487, "global_step": 203461, "epoch": 2286, "lr": 1.4747498813844247e-05}
{"train_loss": 0.07475508004426956, "global_step": 203462, "epoch": 2286, "lr": 1.4747087691610978e-05}
{"train_loss": 0.04646385833621025, "global_step": 203463, "epoch": 2286, "lr": 1.4746676574117053e-05}
{"train_loss": 0.032797351479530334, "global_step": 203464, "epoch": 2286, "lr": 1.474626546136249e-05}
{"train_loss": 0.09118594229221344, "global_step": 203465, "epoch": 2286, "lr": 1.4745854353347376e-05}
{"train_loss": 0.06885432451963425, "global_step": 203466, "epoch": 2286, "lr": 1.4745443250071738e-05}
{"train_loss": 0.06083815544843674, "global_step": 203467, "epoch": 2286, "lr": 1.4745032151535642e-05}
{"train_loss": 0.018588678911328316, "global_step": 203468, "epoch": 2286, "lr": 1.4744621057739167e-05}
{"train_loss": 0.047973327338695526, "global_step": 203469, "epoch": 2286, "lr": 1.4744209968682332e-05}
{"train_loss": 0.04995143786072731, "global_step": 203470, "epoch": 2286, "lr": 1.4743798884365224e-05}
{"train_loss": 0.026363614946603775, "global_step": 203471, "epoch": 2286, "lr": 1.4743387804787861e-05}
{"train_loss": 0.02128317579627037, "global_step": 203472, "epoch": 2286, "lr": 1.4742976729950341e-05}
{"train_loss": 0.0869760513305664, "global_step": 203473, "epoch": 2286, "lr": 1.474256565985268e-05}
{"train_loss": 0.03645964339375496, "global_step": 203474, "epoch": 2286, "lr": 1.4742154594494967e-05}
{"train_loss": 0.06358163058757782, "global_step": 203475, "epoch": 2286, "lr": 1.4741743533877229e-05}
{"train_loss": 0.05289469286799431, "global_step": 203476, "epoch": 2286, "lr": 1.4741332477999547e-05}
{"train_loss": 0.023816172033548355, "global_step": 203477, "epoch": 2286, "lr": 1.4740921426861948e-05}
{"train_loss": 0.0592222586274147, "global_step": 203478, "epoch": 2286, "lr": 1.4740510380464518e-05}
{"train_loss": 0.07098573446273804, "global_step": 203479, "epoch": 2286, "lr": 1.4740099338807284e-05}
{"train_loss": 0.01948312483727932, "global_step": 203480, "epoch": 2286, "lr": 1.4739688301890326e-05}
{"train_loss": 0.06215602532029152, "global_step": 203481, "epoch": 2286, "lr": 1.473927726971367e-05}
{"train_loss": 0.03949218615889549, "global_step": 203482, "epoch": 2286, "lr": 1.4738866242277406e-05}
{"train_loss": 0.03592953830957413, "global_step": 203483, "epoch": 2286, "lr": 1.473845521958157e-05}
{"train_loss": 0.030818071216344833, "global_step": 203484, "epoch": 2286, "lr": 1.4738044201626205e-05}
{"train_loss": 0.1461067497730255, "global_step": 203485, "epoch": 2286, "lr": 1.4737633188411398e-05}
{"train_loss": 0.05385790392756462, "global_step": 203486, "epoch": 2286, "lr": 1.4737222179937171e-05}
{"train_loss": 0.06463958323001862, "global_step": 203487, "epoch": 2286, "lr": 1.4736811176203608e-05}
{"train_loss": 0.1310877799987793, "global_step": 203488, "epoch": 2286, "lr": 1.4736400177210734e-05}
{"train_loss": 0.07517226040363312, "global_step": 203489, "epoch": 2286, "lr": 1.4735989182958638e-05}
{"train_loss": 0.059012189507484436, "global_step": 203490, "epoch": 2286, "lr": 1.4735578193447341e-05}
{"train_loss": 0.06353873014450073, "global_step": 203491, "epoch": 2286, "lr": 1.4735167208676936e-05}
{"train_loss": 0.04875677078962326, "global_step": 203492, "epoch": 2286, "lr": 1.473475622864744e-05}
{"train_loss": 0.05795879662036896, "global_step": 203493, "epoch": 2286, "lr": 1.4734345253358939e-05}
{"train_loss": 0.0791664719581604, "global_step": 203494, "epoch": 2286, "lr": 1.4733934282811463e-05}
{"train_loss": 0.07070903480052948, "global_step": 203495, "epoch": 2286, "lr": 1.4733523317005094e-05}
{"train_loss": 0.038163989782333374, "global_step": 203496, "epoch": 2286, "lr": 1.473311235593986e-05}
{"train_loss": 0.04387251287698746, "global_step": 203497, "epoch": 2286, "lr": 1.473270139961584e-05}
{"train_loss": 0.0515570193529129, "global_step": 203498, "epoch": 2286, "lr": 1.4732290448033065e-05}
{"train_loss": 0.06929688900709152, "global_step": 203499, "epoch": 2286, "lr": 1.473187950119162e-05}
{"train_loss": 0.07643559575080872, "global_step": 203500, "epoch": 2286, "lr": 1.4731468559091527e-05}
{"train_loss": 0.11799157410860062, "global_step": 203501, "epoch": 2286, "lr": 1.473105762173288e-05}
{"train_loss": 0.037419285625219345, "global_step": 203502, "epoch": 2286, "lr": 1.473064668911569e-05}
{"train_loss": 0.09726593643426895, "global_step": 203503, "epoch": 2286, "lr": 1.4730235761240035e-05}
{"train_loss": 0.09726054966449738, "global_step": 203504, "epoch": 2286, "lr": 1.4729824838105987e-05}
{"train_loss": 0.03863009065389633, "global_step": 203505, "epoch": 2286, "lr": 1.4729413919713569e-05}
{"train_loss": 0.031247634440660477, "global_step": 203506, "epoch": 2286, "lr": 1.472900300606287e-05}
{"train_loss": 0.08047298341989517, "global_step": 203507, "epoch": 2286, "lr": 1.4728592097153905e-05}
{"train_loss": 0.11645683646202087, "global_step": 203508, "epoch": 2286, "lr": 1.4728181192986767e-05}
{"train_loss": 0.1042744442820549, "global_step": 203509, "epoch": 2286, "lr": 1.4727770293561483e-05}
{"train_loss": 0.033387232571840286, "global_step": 203510, "epoch": 2286, "lr": 1.4727359398878137e-05}
{"train_loss": 0.06485451012849808, "global_step": 203511, "epoch": 2286, "lr": 1.472694850893675e-05}
{"train_loss": 0.03190276771783829, "global_step": 203512, "epoch": 2286, "lr": 1.4726537623737407e-05}
{"train_loss": 0.08357684314250946, "global_step": 203513, "epoch": 2286, "lr": 1.4726126743280138e-05}
{"train_loss": 0.04472055658698082, "global_step": 203514, "epoch": 2286, "lr": 1.4725715867565027e-05}
{"train_loss": 0.09040011465549469, "global_step": 203515, "epoch": 2286, "lr": 1.4725304996592098e-05}
{"train_loss": 0.06828776746988297, "global_step": 203516, "epoch": 2286, "lr": 1.4724894130361439e-05}
{"train_loss": 0.03835981339216232, "global_step": 203517, "epoch": 2286, "lr": 1.4724483268873073e-05}
{"train_loss": 0.04804716631770134, "global_step": 203518, "epoch": 2286, "lr": 1.4724072412127083e-05}
{"train_loss": 0.07419660687446594, "global_step": 203519, "epoch": 2286, "lr": 1.4723661560123497e-05}
{"train_loss": 0.05610746517777443, "global_step": 203520, "epoch": 2286, "lr": 1.4723250712862402e-05}
{"train_loss": 0.09252358973026276, "global_step": 203521, "epoch": 2286, "lr": 1.4722839870343835e-05}
{"train_loss": 0.06963688880205154, "global_step": 203522, "epoch": 2286, "lr": 1.4722429032567837e-05}
{"train_loss": 0.059774599969387054, "global_step": 203523, "epoch": 2286, "lr": 1.4722018199534492e-05}
{"train_loss": 0.07501041889190674, "global_step": 203524, "epoch": 2286, "lr": 1.4721607371243828e-05}
{"train_loss": 0.06862979382276535, "global_step": 203525, "epoch": 2286, "lr": 1.4721196547695927e-05}
{"train_loss": 0.03371356800198555, "global_step": 203526, "epoch": 2286, "lr": 1.4720785728890818e-05}
{"train_loss": 0.07919543981552124, "global_step": 203527, "epoch": 2286, "lr": 1.4720374914828583e-05}
{"train_loss": 0.09606251120567322, "global_step": 203528, "epoch": 2286, "lr": 1.4719964105509249e-05}
{"train_loss": 0.052178800106048584, "global_step": 203529, "epoch": 2286, "lr": 1.4719553300932904e-05}
{"train_loss": 0.060194872319698334, "global_step": 203530, "epoch": 2286, "lr": 1.4719142501099565e-05}
{"train_loss": 0.05207223445177078, "global_step": 203531, "epoch": 2286, "lr": 1.4718731706009326e-05}
{"train_loss": 0.05927812308073044, "global_step": 203532, "epoch": 2286, "lr": 1.4718320915662204e-05}
{"train_loss": 0.051997266709804535, "global_step": 203533, "epoch": 2286, "lr": 1.4717910130058293e-05}
{"train_loss": 0.03543563559651375, "global_step": 203534, "epoch": 2286, "lr": 1.4717499349197616e-05}
{"train_loss": 0.08145666122436523, "global_step": 203535, "epoch": 2286, "lr": 1.4717088573080251e-05}
{"train_loss": 0.07715187221765518, "global_step": 203536, "epoch": 2286, "lr": 1.471667780170623e-05}
{"train_loss": 0.06846390664577484, "global_step": 203537, "epoch": 2286, "lr": 1.4716267035075637e-05}
{"train_loss": 0.02877362258732319, "global_step": 203538, "epoch": 2286, "lr": 1.47158562731885e-05}
{"train_loss": 0.025649171322584152, "global_step": 203539, "epoch": 2286, "lr": 1.4715445516044895e-05}
{"train_loss": 0.0448351688683033, "global_step": 203540, "epoch": 2286, "lr": 1.4715034763644858e-05}
{"train_loss": 0.08567524701356888, "global_step": 203541, "epoch": 2286, "lr": 1.4714624015988449e-05}
{"train_loss": 0.06067263697054279, "global_step": 203542, "epoch": 2286, "lr": 1.471421327307575e-05, "val_loss": 8.315241813659668}
{"train_loss": 0.07749209553003311, "global_step": 203543, "epoch": 2287, "lr": 1.4713802534906778e-05}
{"train_loss": 0.08798257261514664, "global_step": 203544, "epoch": 2287, "lr": 1.4713391801481619e-05}
{"train_loss": 0.05141349881887436, "global_step": 203545, "epoch": 2287, "lr": 1.4712981072800297e-05}
{"train_loss": 0.04077611491084099, "global_step": 203546, "epoch": 2287, "lr": 1.4712570348862903e-05}
{"train_loss": 0.035057246685028076, "global_step": 203547, "epoch": 2287, "lr": 1.4712159629669459e-05}
{"train_loss": 0.05481700599193573, "global_step": 203548, "epoch": 2287, "lr": 1.4711748915220053e-05}
{"train_loss": 0.10922016948461533, "global_step": 203549, "epoch": 2287, "lr": 1.4711338205514703e-05}
{"train_loss": 0.042738959193229675, "global_step": 203550, "epoch": 2287, "lr": 1.47109275005535e-05}
{"train_loss": 0.08736103028059006, "global_step": 203551, "epoch": 2287, "lr": 1.4710516800336466e-05}
{"train_loss": 0.0618116594851017, "global_step": 203552, "epoch": 2287, "lr": 1.4710106104863685e-05}
{"train_loss": 0.04295992851257324, "global_step": 203553, "epoch": 2287, "lr": 1.4709695414135195e-05}
{"train_loss": 0.07835458964109421, "global_step": 203554, "epoch": 2287, "lr": 1.4709284728151063e-05}
{"train_loss": 0.040277738124132156, "global_step": 203555, "epoch": 2287, "lr": 1.4708874046911325e-05}
{"train_loss": 0.055284373462200165, "global_step": 203556, "epoch": 2287, "lr": 1.4708463370416065e-05}
{"train_loss": 0.049951378256082535, "global_step": 203557, "epoch": 2287, "lr": 1.4708052698665304e-05}
{"train_loss": 0.03208132088184357, "global_step": 203558, "epoch": 2287, "lr": 1.470764203165913e-05}
{"train_loss": 0.0721265971660614, "global_step": 203559, "epoch": 2287, "lr": 1.4707231369397572e-05}
{"train_loss": 0.057087261229753494, "global_step": 203560, "epoch": 2287, "lr": 1.4706820711880708e-05}
{"train_loss": 0.08608308434486389, "global_step": 203561, "epoch": 2287, "lr": 1.4706410059108583e-05}
{"train_loss": 0.05574485659599304, "global_step": 203562, "epoch": 2287, "lr": 1.4705999411081239e-05}
{"train_loss": 0.07511210441589355, "global_step": 203563, "epoch": 2287, "lr": 1.4705588767798755e-05}
{"train_loss": 0.07067976891994476, "global_step": 203564, "epoch": 2287, "lr": 1.470517812926116e-05}
{"train_loss": 0.09243504703044891, "global_step": 203565, "epoch": 2287, "lr": 1.4704767495468536e-05}
{"train_loss": 0.020268328487873077, "global_step": 203566, "epoch": 2287, "lr": 1.4704356866420915e-05}
{"train_loss": 0.045901644974946976, "global_step": 203567, "epoch": 2287, "lr": 1.4703946242118377e-05}
{"train_loss": 0.04019416123628616, "global_step": 203568, "epoch": 2287, "lr": 1.4703535622560949e-05}
{"train_loss": 0.05710740387439728, "global_step": 203569, "epoch": 2287, "lr": 1.4703125007748713e-05}
{"train_loss": 0.05417482927441597, "global_step": 203570, "epoch": 2287, "lr": 1.4702714397681693e-05}
{"train_loss": 0.0524301677942276, "global_step": 203571, "epoch": 2287, "lr": 1.4702303792359984e-05}
{"train_loss": 0.05583812668919563, "global_step": 203572, "epoch": 2287, "lr": 1.4701893191783606e-05}
{"train_loss": 0.07704608142375946, "global_step": 203573, "epoch": 2287, "lr": 1.4701482595952643e-05}
{"train_loss": 0.036222413182258606, "global_step": 203574, "epoch": 2287, "lr": 1.4701072004867116e-05}
{"train_loss": 0.06845678389072418, "global_step": 203575, "epoch": 2287, "lr": 1.4700661418527117e-05}
{"train_loss": 0.052094921469688416, "global_step": 203576, "epoch": 2287, "lr": 1.4700250836932666e-05}
{"train_loss": 0.05481453612446785, "global_step": 203577, "epoch": 2287, "lr": 1.4699840260083842e-05}
{"train_loss": 0.0623319037258625, "global_step": 203578, "epoch": 2287, "lr": 1.4699429687980703e-05}
{"train_loss": 0.0601281113922596, "global_step": 203579, "epoch": 2287, "lr": 1.4699019120623287e-05}
{"train_loss": 0.03991993889212608, "global_step": 203580, "epoch": 2287, "lr": 1.4698608558011667e-05}
{"train_loss": 0.10266999900341034, "global_step": 203581, "epoch": 2287, "lr": 1.4698198000145875e-05}
{"train_loss": 0.05992920696735382, "global_step": 203582, "epoch": 2287, "lr": 1.4697787447025996e-05}
{"train_loss": 0.03204473853111267, "global_step": 203583, "epoch": 2287, "lr": 1.4697376898652054e-05}
{"train_loss": 0.054027900099754333, "global_step": 203584, "epoch": 2287, "lr": 1.4696966355024138e-05}
{"train_loss": 0.09985596686601639, "global_step": 203585, "epoch": 2287, "lr": 1.4696555816142265e-05}
{"train_loss": 0.12309477478265762, "global_step": 203586, "epoch": 2287, "lr": 1.4696145282006524e-05}
{"train_loss": 0.1538279503583908, "global_step": 203587, "epoch": 2287, "lr": 1.4695734752616946e-05}
{"train_loss": 0.10579928755760193, "global_step": 203588, "epoch": 2287, "lr": 1.4695324227973606e-05}
{"train_loss": 0.09915100038051605, "global_step": 203589, "epoch": 2287, "lr": 1.4694913708076536e-05}
{"train_loss": 0.04706725478172302, "global_step": 203590, "epoch": 2287, "lr": 1.4694503192925817e-05}
{"train_loss": 0.05779005214571953, "global_step": 203591, "epoch": 2287, "lr": 1.4694092682521482e-05}
{"train_loss": 0.052150510251522064, "global_step": 203592, "epoch": 2287, "lr": 1.469368217686361e-05}
{"train_loss": 0.03480752557516098, "global_step": 203593, "epoch": 2287, "lr": 1.4693271675952225e-05}
{"train_loss": 0.0371641181409359, "global_step": 203594, "epoch": 2287, "lr": 1.4692861179787415e-05}
{"train_loss": 0.05704612284898758, "global_step": 203595, "epoch": 2287, "lr": 1.4692450688369202e-05}
{"train_loss": 0.04361908510327339, "global_step": 203596, "epoch": 2287, "lr": 1.4692040201697676e-05}
{"train_loss": 0.06312918663024902, "global_step": 203597, "epoch": 2287, "lr": 1.4691629719772865e-05}
{"train_loss": 0.05419932305812836, "global_step": 203598, "epoch": 2287, "lr": 1.4691219242594838e-05}
{"train_loss": 0.06483162939548492, "global_step": 203599, "epoch": 2287, "lr": 1.4690808770163656e-05}
{"train_loss": 0.04231495037674904, "global_step": 203600, "epoch": 2287, "lr": 1.4690398302479347e-05}
{"train_loss": 0.09522488713264465, "global_step": 203601, "epoch": 2287, "lr": 1.4689987839541997e-05}
{"train_loss": 0.06844714283943176, "global_step": 203602, "epoch": 2287, "lr": 1.4689577381351632e-05}
{"train_loss": 0.04852760583162308, "global_step": 203603, "epoch": 2287, "lr": 1.4689166927908338e-05}
{"train_loss": 0.08405066281557083, "global_step": 203604, "epoch": 2287, "lr": 1.468875647921214e-05}
{"train_loss": 0.05315622687339783, "global_step": 203605, "epoch": 2287, "lr": 1.4688346035263129e-05}
{"train_loss": 0.0982639417052269, "global_step": 203606, "epoch": 2287, "lr": 1.4687935596061315e-05}
{"train_loss": 0.058018241077661514, "global_step": 203607, "epoch": 2287, "lr": 1.4687525161606801e-05}
{"train_loss": 0.0441259890794754, "global_step": 203608, "epoch": 2287, "lr": 1.4687114731899598e-05}
{"train_loss": 0.06917813420295715, "global_step": 203609, "epoch": 2287, "lr": 1.4686704306939802e-05}
{"train_loss": 0.04973694309592247, "global_step": 203610, "epoch": 2287, "lr": 1.4686293886727426e-05}
{"train_loss": 0.054749518632888794, "global_step": 203611, "epoch": 2287, "lr": 1.4685883471262569e-05}
{"train_loss": 0.05966120585799217, "global_step": 203612, "epoch": 2287, "lr": 1.4685473060545246e-05}
{"train_loss": 0.03781836852431297, "global_step": 203613, "epoch": 2287, "lr": 1.4685062654575532e-05}
{"train_loss": 0.06334526836872101, "global_step": 203614, "epoch": 2287, "lr": 1.4684652253353498e-05}
{"train_loss": 0.07442894577980042, "global_step": 203615, "epoch": 2287, "lr": 1.4684241856879166e-05}
{"train_loss": 0.04951685294508934, "global_step": 203616, "epoch": 2287, "lr": 1.468383146515262e-05}
{"train_loss": 0.07592332363128662, "global_step": 203617, "epoch": 2287, "lr": 1.4683421078173887e-05}
{"train_loss": 0.10931359231472015, "global_step": 203618, "epoch": 2287, "lr": 1.4683010695943056e-05}
{"train_loss": 0.09572634100914001, "global_step": 203619, "epoch": 2287, "lr": 1.4682600318460148e-05}
{"train_loss": 0.048593319952487946, "global_step": 203620, "epoch": 2287, "lr": 1.468218994572525e-05}
{"train_loss": 0.057621538639068604, "global_step": 203621, "epoch": 2287, "lr": 1.4681779577738381e-05}
{"train_loss": 0.08619490265846252, "global_step": 203622, "epoch": 2287, "lr": 1.4681369214499635e-05}
{"train_loss": 0.06725893914699554, "global_step": 203623, "epoch": 2287, "lr": 1.4680958856009036e-05}
{"train_loss": 0.07646769285202026, "global_step": 203624, "epoch": 2287, "lr": 1.468054850226666e-05}
{"train_loss": 0.07188927382230759, "global_step": 203625, "epoch": 2287, "lr": 1.4680138153272544e-05}
{"train_loss": 0.07561618834733963, "global_step": 203626, "epoch": 2287, "lr": 1.4679727809026767e-05}
{"train_loss": 0.08993043005466461, "global_step": 203627, "epoch": 2287, "lr": 1.4679317469529351e-05}
{"train_loss": 0.039276786148548126, "global_step": 203628, "epoch": 2287, "lr": 1.467890713478039e-05}
{"train_loss": 0.08849208056926727, "global_step": 203629, "epoch": 2287, "lr": 1.46784968047799e-05}
{"train_loss": 0.06723318248987198, "global_step": 203630, "epoch": 2287, "lr": 1.4678086479527963e-05}
{"train_loss": 0.06431882468502173, "global_step": 203631, "epoch": 2287, "lr": 1.4677676159024633e-05, "val_loss": 8.258026123046875}
{"train_loss": 0.08782964199781418, "global_step": 203632, "epoch": 2288, "lr": 1.4677265843269966e-05}
{"train_loss": 0.1036115512251854, "global_step": 203633, "epoch": 2288, "lr": 1.4676855532263995e-05}
{"train_loss": 0.06334928423166275, "global_step": 203634, "epoch": 2288, "lr": 1.4676445226006802e-05}
{"train_loss": 0.0607149600982666, "global_step": 203635, "epoch": 2288, "lr": 1.4676034924498416e-05}
{"train_loss": 0.045497484505176544, "global_step": 203636, "epoch": 2288, "lr": 1.467562462773892e-05}
{"train_loss": 0.05643375217914581, "global_step": 203637, "epoch": 2288, "lr": 1.467521433572836e-05}
{"train_loss": 0.11658034473657608, "global_step": 203638, "epoch": 2288, "lr": 1.4674804048466768e-05}
{"train_loss": 0.0294156763702631, "global_step": 203639, "epoch": 2288, "lr": 1.4674393765954236e-05}
{"train_loss": 0.06466361880302429, "global_step": 203640, "epoch": 2288, "lr": 1.4673983488190784e-05}
{"train_loss": 0.09120708703994751, "global_step": 203641, "epoch": 2288, "lr": 1.46735732151765e-05}
{"train_loss": 0.0789223238825798, "global_step": 203642, "epoch": 2288, "lr": 1.4673162946911406e-05}
{"train_loss": 0.025892753154039383, "global_step": 203643, "epoch": 2288, "lr": 1.467275268339559e-05}
{"train_loss": 0.11634984612464905, "global_step": 203644, "epoch": 2288, "lr": 1.4672342424629082e-05}
{"train_loss": 0.125588521361351, "global_step": 203645, "epoch": 2288, "lr": 1.4671932170611956e-05}
{"train_loss": 0.025287648662924767, "global_step": 203646, "epoch": 2288, "lr": 1.467152192134425e-05}
{"train_loss": 0.07733578234910965, "global_step": 203647, "epoch": 2288, "lr": 1.4671111676826038e-05}
{"train_loss": 0.0584278404712677, "global_step": 203648, "epoch": 2288, "lr": 1.4670701437057348e-05}
{"train_loss": 0.06664570420980453, "global_step": 203649, "epoch": 2288, "lr": 1.467029120203825e-05}
{"train_loss": 0.03199157118797302, "global_step": 203650, "epoch": 2288, "lr": 1.4669880971768823e-05}
{"train_loss": 0.05258103460073471, "global_step": 203651, "epoch": 2288, "lr": 1.4669470746249081e-05}
{"train_loss": 0.0345856249332428, "global_step": 203652, "epoch": 2288, "lr": 1.4669060525479112e-05}
{"train_loss": 0.03715299814939499, "global_step": 203653, "epoch": 2288, "lr": 1.4668650309458943e-05}
{"train_loss": 0.10954754799604416, "global_step": 203654, "epoch": 2288, "lr": 1.4668240098188656e-05}
{"train_loss": 0.051639385521411896, "global_step": 203655, "epoch": 2288, "lr": 1.4667829891668283e-05}
{"train_loss": 0.0652036964893341, "global_step": 203656, "epoch": 2288, "lr": 1.4667419689897905e-05}
{"train_loss": 0.028652340173721313, "global_step": 203657, "epoch": 2288, "lr": 1.4667009492877542e-05}
{"train_loss": 0.0394713319838047, "global_step": 203658, "epoch": 2288, "lr": 1.4666599300607287e-05}
{"train_loss": 0.08551569283008575, "global_step": 203659, "epoch": 2288, "lr": 1.4666189113087159e-05}
{"train_loss": 0.0921742171049118, "global_step": 203660, "epoch": 2288, "lr": 1.4665778930317248e-05}
{"train_loss": 0.07164962589740753, "global_step": 203661, "epoch": 2288, "lr": 1.4665368752297582e-05}
{"train_loss": 0.06395956873893738, "global_step": 203662, "epoch": 2288, "lr": 1.4664958579028238e-05}
{"train_loss": 0.0386023186147213, "global_step": 203663, "epoch": 2288, "lr": 1.4664548410509237e-05}
{"train_loss": 0.05662725493311882, "global_step": 203664, "epoch": 2288, "lr": 1.4664138246740684e-05}
{"train_loss": 0.020919376984238625, "global_step": 203665, "epoch": 2288, "lr": 1.4663728087722584e-05}
{"train_loss": 0.07050041109323502, "global_step": 203666, "epoch": 2288, "lr": 1.4663317933455033e-05}
{"train_loss": 0.04163528233766556, "global_step": 203667, "epoch": 2288, "lr": 1.4662907783938052e-05}
{"train_loss": 0.05251311510801315, "global_step": 203668, "epoch": 2288, "lr": 1.4662497639171713e-05}
{"train_loss": 0.055655840784311295, "global_step": 203669, "epoch": 2288, "lr": 1.4662087499156085e-05}
{"train_loss": 0.06051712483167648, "global_step": 203670, "epoch": 2288, "lr": 1.4661677363891196e-05}
{"train_loss": 0.055198729038238525, "global_step": 203671, "epoch": 2288, "lr": 1.466126723337713e-05}
{"train_loss": 0.05744920298457146, "global_step": 203672, "epoch": 2288, "lr": 1.4660857107613923e-05}
{"train_loss": 0.10141804069280624, "global_step": 203673, "epoch": 2288, "lr": 1.4660446986601617e-05}
{"train_loss": 0.028589121997356415, "global_step": 203674, "epoch": 2288, "lr": 1.46600368703403e-05}
{"train_loss": 0.027700118720531464, "global_step": 203675, "epoch": 2288, "lr": 1.4659626758829992e-05}
{"train_loss": 0.0661599338054657, "global_step": 203676, "epoch": 2288, "lr": 1.4659216652070785e-05}
{"train_loss": 0.03707552328705788, "global_step": 203677, "epoch": 2288, "lr": 1.4658806550062716e-05}
{"train_loss": 0.025922441855072975, "global_step": 203678, "epoch": 2288, "lr": 1.4658396452805822e-05}
{"train_loss": 0.05164118856191635, "global_step": 203679, "epoch": 2288, "lr": 1.4657986360300197e-05}
{"train_loss": 0.06689160317182541, "global_step": 203680, "epoch": 2288, "lr": 1.4657576272545854e-05}
{"train_loss": 0.04634474217891693, "global_step": 203681, "epoch": 2288, "lr": 1.4657166189542881e-05}
{"train_loss": 0.10639004409313202, "global_step": 203682, "epoch": 2288, "lr": 1.4656756111291315e-05}
{"train_loss": 0.06780492514371872, "global_step": 203683, "epoch": 2288, "lr": 1.4656346037791225e-05}
{"train_loss": 0.05475202575325966, "global_step": 203684, "epoch": 2288, "lr": 1.4655935969042645e-05}
{"train_loss": 0.05261523276567459, "global_step": 203685, "epoch": 2288, "lr": 1.4655525905045647e-05}
{"train_loss": 0.08967384696006775, "global_step": 203686, "epoch": 2288, "lr": 1.4655115845800293e-05}
{"train_loss": 0.09187138825654984, "global_step": 203687, "epoch": 2288, "lr": 1.4654705791306617e-05}
{"train_loss": 0.08665941655635834, "global_step": 203688, "epoch": 2288, "lr": 1.46542957415647e-05}
{"train_loss": 0.03921667858958244, "global_step": 203689, "epoch": 2288, "lr": 1.4653885696574566e-05}
{"train_loss": 0.03248712047934532, "global_step": 203690, "epoch": 2288, "lr": 1.4653475656336301e-05}
{"train_loss": 0.06315787136554718, "global_step": 203691, "epoch": 2288, "lr": 1.4653065620849932e-05}
{"train_loss": 0.0797422006726265, "global_step": 203692, "epoch": 2288, "lr": 1.4652655590115538e-05}
{"train_loss": 0.08422301709651947, "global_step": 203693, "epoch": 2288, "lr": 1.4652245564133155e-05}
{"train_loss": 0.038623616099357605, "global_step": 203694, "epoch": 2288, "lr": 1.4651835542902859e-05}
{"train_loss": 0.08189987391233444, "global_step": 203695, "epoch": 2288, "lr": 1.4651425526424673e-05}
{"train_loss": 0.06974447518587112, "global_step": 203696, "epoch": 2288, "lr": 1.465101551469869e-05}
{"train_loss": 0.03594126179814339, "global_step": 203697, "epoch": 2288, "lr": 1.4650605507724935e-05}
{"train_loss": 0.06889080256223679, "global_step": 203698, "epoch": 2288, "lr": 1.4650195505503488e-05}
{"train_loss": 0.05015094205737114, "global_step": 203699, "epoch": 2288, "lr": 1.464978550803438e-05}
{"train_loss": 0.040409088134765625, "global_step": 203700, "epoch": 2288, "lr": 1.4649375515317687e-05}
{"train_loss": 0.03521944209933281, "global_step": 203701, "epoch": 2288, "lr": 1.4648965527353442e-05}
{"train_loss": 0.06865627318620682, "global_step": 203702, "epoch": 2288, "lr": 1.4648555544141728e-05}
{"train_loss": 0.07755528390407562, "global_step": 203703, "epoch": 2288, "lr": 1.464814556568257e-05}
{"train_loss": 0.03409967198967934, "global_step": 203704, "epoch": 2288, "lr": 1.4647735591976036e-05}
{"train_loss": 0.045996591448783875, "global_step": 203705, "epoch": 2288, "lr": 1.46473256230222e-05}
{"train_loss": 0.061661917716264725, "global_step": 203706, "epoch": 2288, "lr": 1.4646915658821086e-05}
{"train_loss": 0.06377459317445755, "global_step": 203707, "epoch": 2288, "lr": 1.4646505699372775e-05}
{"train_loss": 0.0980287566781044, "global_step": 203708, "epoch": 2288, "lr": 1.4646095744677301e-05}
{"train_loss": 0.0374285951256752, "global_step": 203709, "epoch": 2288, "lr": 1.4645685794734738e-05}
{"train_loss": 0.06233133375644684, "global_step": 203710, "epoch": 2288, "lr": 1.4645275849545137e-05}
{"train_loss": 0.10061255842447281, "global_step": 203711, "epoch": 2288, "lr": 1.464486590910853e-05}
{"train_loss": 0.11725613474845886, "global_step": 203712, "epoch": 2288, "lr": 1.4644455973425003e-05}
{"train_loss": 0.10764920711517334, "global_step": 203713, "epoch": 2288, "lr": 1.464404604249458e-05}
{"train_loss": 0.09106766432523727, "global_step": 203714, "epoch": 2288, "lr": 1.4643636116317354e-05}
{"train_loss": 0.054069194942712784, "global_step": 203715, "epoch": 2288, "lr": 1.4643226194893362e-05}
{"train_loss": 0.09692058712244034, "global_step": 203716, "epoch": 2288, "lr": 1.4642816278222637e-05}
{"train_loss": 0.044961702078580856, "global_step": 203717, "epoch": 2288, "lr": 1.4642406366305279e-05}
{"train_loss": 0.08152295649051666, "global_step": 203718, "epoch": 2288, "lr": 1.4641996459141294e-05}
{"train_loss": 0.019224518910050392, "global_step": 203719, "epoch": 2288, "lr": 1.4641586556730779e-05}
{"train_loss": 0.06335501910679125, "global_step": 203720, "epoch": 2288, "lr": 1.4641176659073757e-05, "val_loss": 8.282301902770996}
{"train_loss": 0.09570734947919846, "global_step": 203721, "epoch": 2289, "lr": 1.4640766766170294e-05}
{"train_loss": 0.07891253381967545, "global_step": 203722, "epoch": 2289, "lr": 1.4640356878020473e-05}
{"train_loss": 0.052403297275304794, "global_step": 203723, "epoch": 2289, "lr": 1.4639946994624304e-05}
{"train_loss": 0.06290671974420547, "global_step": 203724, "epoch": 2289, "lr": 1.4639537115981877e-05}
{"train_loss": 0.03239979222416878, "global_step": 203725, "epoch": 2289, "lr": 1.463912724209322e-05}
{"train_loss": 0.058648187667131424, "global_step": 203726, "epoch": 2289, "lr": 1.4638717372958416e-05}
{"train_loss": 0.07467808574438095, "global_step": 203727, "epoch": 2289, "lr": 1.4638307508577492e-05}
{"train_loss": 0.07156409323215485, "global_step": 203728, "epoch": 2289, "lr": 1.4637897648950533e-05}
{"train_loss": 0.08205381780862808, "global_step": 203729, "epoch": 2289, "lr": 1.4637487794077558e-05}
{"train_loss": 0.03302786126732826, "global_step": 203730, "epoch": 2289, "lr": 1.4637077943958661e-05}
{"train_loss": 0.06559145450592041, "global_step": 203731, "epoch": 2289, "lr": 1.4636668098593858e-05}
{"train_loss": 0.04370805248618126, "global_step": 203732, "epoch": 2289, "lr": 1.4636258257983243e-05}
{"train_loss": 0.08049376308917999, "global_step": 203733, "epoch": 2289, "lr": 1.4635848422126836e-05}
{"train_loss": 0.03557245433330536, "global_step": 203734, "epoch": 2289, "lr": 1.4635438591024724e-05}
{"train_loss": 0.03365078940987587, "global_step": 203735, "epoch": 2289, "lr": 1.4635028764676928e-05}
{"train_loss": 0.05078161507844925, "global_step": 203736, "epoch": 2289, "lr": 1.4634618943083539e-05}
{"train_loss": 0.0626363605260849, "global_step": 203737, "epoch": 2289, "lr": 1.463420912624458e-05}
{"train_loss": 0.032759618014097214, "global_step": 203738, "epoch": 2289, "lr": 1.4633799314160135e-05}
{"train_loss": 0.047214578837156296, "global_step": 203739, "epoch": 2289, "lr": 1.463338950683023e-05}
{"train_loss": 0.06418778747320175, "global_step": 203740, "epoch": 2289, "lr": 1.4632979704254934e-05}
{"train_loss": 0.0714218020439148, "global_step": 203741, "epoch": 2289, "lr": 1.4632569906434317e-05}
{"train_loss": 0.060704320669174194, "global_step": 203742, "epoch": 2289, "lr": 1.4632160113368403e-05}
{"train_loss": 0.08935070037841797, "global_step": 203743, "epoch": 2289, "lr": 1.4631750325057287e-05}
{"train_loss": 0.08982498198747635, "global_step": 203744, "epoch": 2289, "lr": 1.4631340541500976e-05}
{"train_loss": 0.03651265799999237, "global_step": 203745, "epoch": 2289, "lr": 1.4630930762699573e-05}
{"train_loss": 0.05540638417005539, "global_step": 203746, "epoch": 2289, "lr": 1.463052098865309e-05}
{"train_loss": 0.032647956162691116, "global_step": 203747, "epoch": 2289, "lr": 1.463011121936162e-05}
{"train_loss": 0.04758269339799881, "global_step": 203748, "epoch": 2289, "lr": 1.4629701454825201e-05}
{"train_loss": 0.058841194957494736, "global_step": 203749, "epoch": 2289, "lr": 1.4629291695043874e-05}
{"train_loss": 0.04617413133382797, "global_step": 203750, "epoch": 2289, "lr": 1.462888194001772e-05}
{"train_loss": 0.05400538817048073, "global_step": 203751, "epoch": 2289, "lr": 1.4628472189746767e-05}
{"train_loss": 0.0582907609641552, "global_step": 203752, "epoch": 2289, "lr": 1.46280624442311e-05}
{"train_loss": 0.05281529948115349, "global_step": 203753, "epoch": 2289, "lr": 1.4627652703470757e-05}
{"train_loss": 0.0688425675034523, "global_step": 203754, "epoch": 2289, "lr": 1.4627242967465782e-05}
{"train_loss": 0.13374337553977966, "global_step": 203755, "epoch": 2289, "lr": 1.4626833236216258e-05}
{"train_loss": 0.05370344966650009, "global_step": 203756, "epoch": 2289, "lr": 1.4626423509722209e-05}
{"train_loss": 0.10205764323472977, "global_step": 203757, "epoch": 2289, "lr": 1.4626013787983706e-05}
{"train_loss": 0.022395770996809006, "global_step": 203758, "epoch": 2289, "lr": 1.4625604071000821e-05}
{"train_loss": 0.0651545375585556, "global_step": 203759, "epoch": 2289, "lr": 1.4625194358773575e-05}
{"train_loss": 0.04263631999492645, "global_step": 203760, "epoch": 2289, "lr": 1.462478465130206e-05}
{"train_loss": 0.11231919378042221, "global_step": 203761, "epoch": 2289, "lr": 1.4624374948586294e-05}
{"train_loss": 0.058634694665670395, "global_step": 203762, "epoch": 2289, "lr": 1.4623965250626365e-05}
{"train_loss": 0.07617983222007751, "global_step": 203763, "epoch": 2289, "lr": 1.4623555557422296e-05}
{"train_loss": 0.06848524510860443, "global_step": 203764, "epoch": 2289, "lr": 1.4623145868974176e-05}
{"train_loss": 0.024026455357670784, "global_step": 203765, "epoch": 2289, "lr": 1.4622736185282026e-05}
{"train_loss": 0.07387331873178482, "global_step": 203766, "epoch": 2289, "lr": 1.462232650634593e-05}
{"train_loss": 0.0690433606505394, "global_step": 203767, "epoch": 2289, "lr": 1.4621916832165922e-05}
{"train_loss": 0.06563255935907364, "global_step": 203768, "epoch": 2289, "lr": 1.4621507162742076e-05}
{"train_loss": 0.040717002004384995, "global_step": 203769, "epoch": 2289, "lr": 1.4621097498074426e-05}
{"train_loss": 0.04200226068496704, "global_step": 203770, "epoch": 2289, "lr": 1.4620687838163055e-05}
{"train_loss": 0.02655370719730854, "global_step": 203771, "epoch": 2289, "lr": 1.462027818300798e-05}
{"train_loss": 0.062045373022556305, "global_step": 203772, "epoch": 2289, "lr": 1.4619868532609299e-05}
{"train_loss": 0.036768317222595215, "global_step": 203773, "epoch": 2289, "lr": 1.4619458886967025e-05}
{"train_loss": 0.023833956569433212, "global_step": 203774, "epoch": 2289, "lr": 1.4619049246081251e-05}
{"train_loss": 0.08443696796894073, "global_step": 203775, "epoch": 2289, "lr": 1.4618639609952e-05}
{"train_loss": 0.04337523505091667, "global_step": 203776, "epoch": 2289, "lr": 1.461822997857934e-05}
{"train_loss": 0.05757822468876839, "global_step": 203777, "epoch": 2289, "lr": 1.4617820351963346e-05}
{"train_loss": 0.01272876188158989, "global_step": 203778, "epoch": 2289, "lr": 1.4617410730104036e-05}
{"train_loss": 0.06254758685827255, "global_step": 203779, "epoch": 2289, "lr": 1.46170011130015e-05}
{"train_loss": 0.04338660463690758, "global_step": 203780, "epoch": 2289, "lr": 1.4616591500655763e-05}
{"train_loss": 0.04372924938797951, "global_step": 203781, "epoch": 2289, "lr": 1.4616181893066905e-05}
{"train_loss": 0.05653093755245209, "global_step": 203782, "epoch": 2289, "lr": 1.4615772290234958e-05}
{"train_loss": 0.07676707208156586, "global_step": 203783, "epoch": 2289, "lr": 1.4615362692160007e-05}
{"train_loss": 0.05886128917336464, "global_step": 203784, "epoch": 2289, "lr": 1.4614953098842072e-05}
{"train_loss": 0.04183829203248024, "global_step": 203785, "epoch": 2289, "lr": 1.4614543510281243e-05}
{"train_loss": 0.035697903484106064, "global_step": 203786, "epoch": 2289, "lr": 1.4614133926477536e-05}
{"train_loss": 0.057263318449258804, "global_step": 203787, "epoch": 2289, "lr": 1.4613724347431046e-05}
{"train_loss": 0.06989002972841263, "global_step": 203788, "epoch": 2289, "lr": 1.4613314773141812e-05}
{"train_loss": 0.06083081290125847, "global_step": 203789, "epoch": 2289, "lr": 1.4612905203609867e-05}
{"train_loss": 0.027872830629348755, "global_step": 203790, "epoch": 2289, "lr": 1.4612495638835299e-05}
{"train_loss": 0.043010056018829346, "global_step": 203791, "epoch": 2289, "lr": 1.4612086078818137e-05}
{"train_loss": 0.049562033265829086, "global_step": 203792, "epoch": 2289, "lr": 1.4611676523558472e-05}
{"train_loss": 0.0344684012234211, "global_step": 203793, "epoch": 2289, "lr": 1.4611266973056326e-05}
{"train_loss": 0.034233495593070984, "global_step": 203794, "epoch": 2289, "lr": 1.4610857427311747e-05}
{"train_loss": 0.070573590695858, "global_step": 203795, "epoch": 2289, "lr": 1.4610447886324814e-05}
{"train_loss": 0.06353696435689926, "global_step": 203796, "epoch": 2289, "lr": 1.461003835009559e-05}
{"train_loss": 0.06954881548881531, "global_step": 203797, "epoch": 2289, "lr": 1.4609628818624094e-05}
{"train_loss": 0.06045924127101898, "global_step": 203798, "epoch": 2289, "lr": 1.460921929191042e-05}
{"train_loss": 0.01259309146553278, "global_step": 203799, "epoch": 2289, "lr": 1.460880976995459e-05}
{"train_loss": 0.06960989534854889, "global_step": 203800, "epoch": 2289, "lr": 1.4608400252756687e-05}
{"train_loss": 0.10221162438392639, "global_step": 203801, "epoch": 2289, "lr": 1.4607990740316746e-05}
{"train_loss": 0.055253636091947556, "global_step": 203802, "epoch": 2289, "lr": 1.4607581232634838e-05}
{"train_loss": 0.05790942907333374, "global_step": 203803, "epoch": 2289, "lr": 1.4607171729710995e-05}
{"train_loss": 0.024270683526992798, "global_step": 203804, "epoch": 2289, "lr": 1.46067622315453e-05}
{"train_loss": 0.10252583771944046, "global_step": 203805, "epoch": 2289, "lr": 1.460635273813778e-05}
{"train_loss": 0.05450489744544029, "global_step": 203806, "epoch": 2289, "lr": 1.4605943249488524e-05}
{"train_loss": 0.0498393215239048, "global_step": 203807, "epoch": 2289, "lr": 1.4605533765597546e-05}
{"train_loss": 0.0892900750041008, "global_step": 203808, "epoch": 2289, "lr": 1.4605124286464944e-05}
{"train_loss": 0.05749506542061487, "global_step": 203809, "epoch": 2289, "lr": 1.460471481209073e-05, "val_loss": 8.412713050842285}
{"train_loss": 0.05791182816028595, "global_step": 203810, "epoch": 2290, "lr": 1.4604305342475e-05}
{"train_loss": 0.05813612788915634, "global_step": 203811, "epoch": 2290, "lr": 1.4603895877617774e-05}
{"train_loss": 0.039290446788072586, "global_step": 203812, "epoch": 2290, "lr": 1.4603486417519125e-05}
{"train_loss": 0.09551753848791122, "global_step": 203813, "epoch": 2290, "lr": 1.4603076962179113e-05}
{"train_loss": 0.08097951114177704, "global_step": 203814, "epoch": 2290, "lr": 1.4602667511597778e-05}
{"train_loss": 0.08652744442224503, "global_step": 203815, "epoch": 2290, "lr": 1.46022580657752e-05}
{"train_loss": 0.09752639383077621, "global_step": 203816, "epoch": 2290, "lr": 1.4601848624711396e-05}
{"train_loss": 0.03549383208155632, "global_step": 203817, "epoch": 2290, "lr": 1.4601439188406457e-05}
{"train_loss": 0.038569118827581406, "global_step": 203818, "epoch": 2290, "lr": 1.4601029756860412e-05}
{"train_loss": 0.0694473534822464, "global_step": 203819, "epoch": 2290, "lr": 1.4600620330073339e-05}
{"train_loss": 0.04481463134288788, "global_step": 203820, "epoch": 2290, "lr": 1.460021090804527e-05}
{"train_loss": 0.03777894750237465, "global_step": 203821, "epoch": 2290, "lr": 1.4599801490776283e-05}
{"train_loss": 0.10428670793771744, "global_step": 203822, "epoch": 2290, "lr": 1.4599392078266404e-05}
{"train_loss": 0.03391431272029877, "global_step": 203823, "epoch": 2290, "lr": 1.4598982670515726e-05}
{"train_loss": 0.10962266474962234, "global_step": 203824, "epoch": 2290, "lr": 1.4598573267524263e-05}
{"train_loss": 0.044152285903692245, "global_step": 203825, "epoch": 2290, "lr": 1.4598163869292108e-05}
{"train_loss": 0.08417822420597076, "global_step": 203826, "epoch": 2290, "lr": 1.4597754475819298e-05}
{"train_loss": 0.03490191325545311, "global_step": 203827, "epoch": 2290, "lr": 1.4597345087105873e-05}
{"train_loss": 0.036933012306690216, "global_step": 203828, "epoch": 2290, "lr": 1.459693570315192e-05}
{"train_loss": 0.026608752086758614, "global_step": 203829, "epoch": 2290, "lr": 1.459652632395746e-05}
{"train_loss": 0.06730056554079056, "global_step": 203830, "epoch": 2290, "lr": 1.4596116949522587e-05}
{"train_loss": 0.08266431093215942, "global_step": 203831, "epoch": 2290, "lr": 1.4595707579847311e-05}
{"train_loss": 0.04369809851050377, "global_step": 203832, "epoch": 2290, "lr": 1.4595298214931729e-05}
{"train_loss": 0.02556684985756874, "global_step": 203833, "epoch": 2290, "lr": 1.4594888854775862e-05}
{"train_loss": 0.07238253951072693, "global_step": 203834, "epoch": 2290, "lr": 1.4594479499379799e-05}
{"train_loss": 0.10606447607278824, "global_step": 203835, "epoch": 2290, "lr": 1.459407014874356e-05}
{"train_loss": 0.07012223452329636, "global_step": 203836, "epoch": 2290, "lr": 1.4593660802867231e-05}
{"train_loss": 0.04930052533745766, "global_step": 203837, "epoch": 2290, "lr": 1.4593251461750833e-05}
{"train_loss": 0.07491268217563629, "global_step": 203838, "epoch": 2290, "lr": 1.4592842125394463e-05}
{"train_loss": 0.04300321638584137, "global_step": 203839, "epoch": 2290, "lr": 1.4592432793798138e-05}
{"train_loss": 0.04915208742022514, "global_step": 203840, "epoch": 2290, "lr": 1.459202346696194e-05}
{"train_loss": 0.05389469489455223, "global_step": 203841, "epoch": 2290, "lr": 1.4591614144885901e-05}
{"train_loss": 0.06154424324631691, "global_step": 203842, "epoch": 2290, "lr": 1.4591204827570105e-05}
{"train_loss": 0.0769655629992485, "global_step": 203843, "epoch": 2290, "lr": 1.4590795515014566e-05}
{"train_loss": 0.04717827960848808, "global_step": 203844, "epoch": 2290, "lr": 1.4590386207219387e-05}
{"train_loss": 0.027347221970558167, "global_step": 203845, "epoch": 2290, "lr": 1.4589976904184583e-05}
{"train_loss": 0.06938527524471283, "global_step": 203846, "epoch": 2290, "lr": 1.4589567605910238e-05}
{"train_loss": 0.07658735662698746, "global_step": 203847, "epoch": 2290, "lr": 1.4589158312396379e-05}
{"train_loss": 0.09233630448579788, "global_step": 203848, "epoch": 2290, "lr": 1.4588749023643072e-05}
{"train_loss": 0.020408036187291145, "global_step": 203849, "epoch": 2290, "lr": 1.4588339739650397e-05}
{"train_loss": 0.06675250828266144, "global_step": 203850, "epoch": 2290, "lr": 1.458793046041837e-05}
{"train_loss": 0.08975154906511307, "global_step": 203851, "epoch": 2290, "lr": 1.458752118594708e-05}
{"train_loss": 0.06758473068475723, "global_step": 203852, "epoch": 2290, "lr": 1.4587111916236552e-05}
{"train_loss": 0.044205788522958755, "global_step": 203853, "epoch": 2290, "lr": 1.4586702651286871e-05}
{"train_loss": 0.04911740496754646, "global_step": 203854, "epoch": 2290, "lr": 1.458629339109806e-05}
{"train_loss": 0.0814499631524086, "global_step": 203855, "epoch": 2290, "lr": 1.4585884135670202e-05}
{"train_loss": 0.03611958399415016, "global_step": 203856, "epoch": 2290, "lr": 1.458547488500333e-05}
{"train_loss": 0.043198492377996445, "global_step": 203857, "epoch": 2290, "lr": 1.4585065639097522e-05}
{"train_loss": 0.07737307250499725, "global_step": 203858, "epoch": 2290, "lr": 1.4584656397952807e-05}
{"train_loss": 0.11714595556259155, "global_step": 203859, "epoch": 2290, "lr": 1.4584247161569265e-05}
{"train_loss": 0.09566724300384521, "global_step": 203860, "epoch": 2290, "lr": 1.4583837929946926e-05}
{"train_loss": 0.01779058761894703, "global_step": 203861, "epoch": 2290, "lr": 1.4583428703085878e-05}
{"train_loss": 0.049389924854040146, "global_step": 203862, "epoch": 2290, "lr": 1.4583019480986132e-05}
{"train_loss": 0.055903516709804535, "global_step": 203863, "epoch": 2290, "lr": 1.458261026364779e-05}
{"train_loss": 0.07333646714687347, "global_step": 203864, "epoch": 2290, "lr": 1.4582201051070881e-05}
{"train_loss": 0.020363081246614456, "global_step": 203865, "epoch": 2290, "lr": 1.4581791843255448e-05}
{"train_loss": 0.07134179770946503, "global_step": 203866, "epoch": 2290, "lr": 1.4581382640201579e-05}
{"train_loss": 0.10439041256904602, "global_step": 203867, "epoch": 2290, "lr": 1.4580973441909295e-05}
{"train_loss": 0.06696367263793945, "global_step": 203868, "epoch": 2290, "lr": 1.458056424837868e-05}
{"train_loss": 0.037735920399427414, "global_step": 203869, "epoch": 2290, "lr": 1.4580155059609767e-05}
{"train_loss": 0.07908264547586441, "global_step": 203870, "epoch": 2290, "lr": 1.4579745875602635e-05}
{"train_loss": 0.06973923742771149, "global_step": 203871, "epoch": 2290, "lr": 1.4579336696357304e-05}
{"train_loss": 0.08187269419431686, "global_step": 203872, "epoch": 2290, "lr": 1.4578927521873869e-05}
{"train_loss": 0.044640008360147476, "global_step": 203873, "epoch": 2290, "lr": 1.4578518352152349e-05}
{"train_loss": 0.056121066212654114, "global_step": 203874, "epoch": 2290, "lr": 1.457810918719283e-05}
{"train_loss": 0.06147143617272377, "global_step": 203875, "epoch": 2290, "lr": 1.4577700026995334e-05}
{"train_loss": 0.054745133966207504, "global_step": 203876, "epoch": 2290, "lr": 1.4577290871559956e-05}
{"train_loss": 0.11605477333068848, "global_step": 203877, "epoch": 2290, "lr": 1.457688172088671e-05}
{"train_loss": 0.07362772524356842, "global_step": 203878, "epoch": 2290, "lr": 1.4576472574975685e-05}
{"train_loss": 0.03900624439120293, "global_step": 203879, "epoch": 2290, "lr": 1.4576063433826914e-05}
{"train_loss": 0.062370821833610535, "global_step": 203880, "epoch": 2290, "lr": 1.4575654297440467e-05}
{"train_loss": 0.07487734407186508, "global_step": 203881, "epoch": 2290, "lr": 1.4575245165816376e-05}
{"train_loss": 0.10941099375486374, "global_step": 203882, "epoch": 2290, "lr": 1.4574836038954732e-05}
{"train_loss": 0.024568861350417137, "global_step": 203883, "epoch": 2290, "lr": 1.4574426916855549e-05}
{"train_loss": 0.0753258690237999, "global_step": 203884, "epoch": 2290, "lr": 1.4574017799518908e-05}
{"train_loss": 0.09141960740089417, "global_step": 203885, "epoch": 2290, "lr": 1.4573608686944873e-05}
{"train_loss": 0.06631068140268326, "global_step": 203886, "epoch": 2290, "lr": 1.4573199579133462e-05}
{"train_loss": 0.07031741738319397, "global_step": 203887, "epoch": 2290, "lr": 1.4572790476084775e-05}
{"train_loss": 0.051459185779094696, "global_step": 203888, "epoch": 2290, "lr": 1.4572381377798827e-05}
{"train_loss": 0.06110158935189247, "global_step": 203889, "epoch": 2290, "lr": 1.457197228427571e-05}
{"train_loss": 0.028329025954008102, "global_step": 203890, "epoch": 2290, "lr": 1.4571563195515436e-05}
{"train_loss": 0.08468105643987656, "global_step": 203891, "epoch": 2290, "lr": 1.4571154111518104e-05}
{"train_loss": 0.068398118019104, "global_step": 203892, "epoch": 2290, "lr": 1.4570745032283733e-05}
{"train_loss": 0.03783395141363144, "global_step": 203893, "epoch": 2290, "lr": 1.457033595781241e-05}
{"train_loss": 0.04421386495232582, "global_step": 203894, "epoch": 2290, "lr": 1.4569926888104152e-05}
{"train_loss": 0.0706624686717987, "global_step": 203895, "epoch": 2290, "lr": 1.4569517823159058e-05}
{"train_loss": 0.058003734797239304, "global_step": 203896, "epoch": 2290, "lr": 1.456910876297714e-05}
{"train_loss": 0.06877688318490982, "global_step": 203897, "epoch": 2290, "lr": 1.4568699707558492e-05}
{"train_loss": 0.06315411321735115, "global_step": 203898, "epoch": 2290, "lr": 1.4568290656903133e-05, "val_loss": 8.435929298400879, "train_action_mse_error": 27.105148315429688}
{"train_loss": 0.07177705317735672, "global_step": 203899, "epoch": 2291, "lr": 1.4567881611011153e-05}
{"train_loss": 0.06956234574317932, "global_step": 203900, "epoch": 2291, "lr": 1.4567472569882573e-05}
{"train_loss": 0.03765040636062622, "global_step": 203901, "epoch": 2291, "lr": 1.4567063533517483e-05}
{"train_loss": 0.025478636845946312, "global_step": 203902, "epoch": 2291, "lr": 1.45666545019159e-05}
{"train_loss": 0.04835577681660652, "global_step": 203903, "epoch": 2291, "lr": 1.4566245475077916e-05}
{"train_loss": 0.0940072238445282, "global_step": 203904, "epoch": 2291, "lr": 1.4565836453003568e-05}
{"train_loss": 0.027430007234215736, "global_step": 203905, "epoch": 2291, "lr": 1.456542743569289e-05}
{"train_loss": 0.0595235601067543, "global_step": 203906, "epoch": 2291, "lr": 1.4565018423145983e-05}
{"train_loss": 0.05944210663437843, "global_step": 203907, "epoch": 2291, "lr": 1.4564609415362857e-05}
{"train_loss": 0.04838521033525467, "global_step": 203908, "epoch": 2291, "lr": 1.4564200412343604e-05}
{"train_loss": 0.05531136319041252, "global_step": 203909, "epoch": 2291, "lr": 1.4563791414088246e-05}
{"train_loss": 0.03708330914378166, "global_step": 203910, "epoch": 2291, "lr": 1.4563382420596872e-05}
{"train_loss": 0.12011358886957169, "global_step": 203911, "epoch": 2291, "lr": 1.4562973431869503e-05}
{"train_loss": 0.05249260365962982, "global_step": 203912, "epoch": 2291, "lr": 1.4562564447906223e-05}
{"train_loss": 0.054901935160160065, "global_step": 203913, "epoch": 2291, "lr": 1.4562155468707062e-05}
{"train_loss": 0.06069919466972351, "global_step": 203914, "epoch": 2291, "lr": 1.45617464942721e-05}
{"train_loss": 0.06328562647104263, "global_step": 203915, "epoch": 2291, "lr": 1.4561337524601365e-05}
{"train_loss": 0.022367946803569794, "global_step": 203916, "epoch": 2291, "lr": 1.4560928559694941e-05}
{"train_loss": 0.05662824958562851, "global_step": 203917, "epoch": 2291, "lr": 1.4560519599552857e-05}
{"train_loss": 0.12792955338954926, "global_step": 203918, "epoch": 2291, "lr": 1.456011064417519e-05}
{"train_loss": 0.07484204322099686, "global_step": 203919, "epoch": 2291, "lr": 1.4559701693561973e-05}
{"train_loss": 0.03426923602819443, "global_step": 203920, "epoch": 2291, "lr": 1.4559292747713282e-05}
{"train_loss": 0.08114874362945557, "global_step": 203921, "epoch": 2291, "lr": 1.4558883806629153e-05}
{"train_loss": 0.05600310117006302, "global_step": 203922, "epoch": 2291, "lr": 1.4558474870309647e-05}
{"train_loss": 0.05604751408100128, "global_step": 203923, "epoch": 2291, "lr": 1.4558065938754833e-05}
{"train_loss": 0.08669842779636383, "global_step": 203924, "epoch": 2291, "lr": 1.4557657011964749e-05}
{"train_loss": 0.030094092711806297, "global_step": 203925, "epoch": 2291, "lr": 1.4557248089939462e-05}
{"train_loss": 0.028207097202539444, "global_step": 203926, "epoch": 2291, "lr": 1.4556839172679009e-05}
{"train_loss": 0.1061982810497284, "global_step": 203927, "epoch": 2291, "lr": 1.4556430260183479e-05}
{"train_loss": 0.0831947848200798, "global_step": 203928, "epoch": 2291, "lr": 1.4556021352452881e-05}
{"train_loss": 0.03177123889327049, "global_step": 203929, "epoch": 2291, "lr": 1.455561244948731e-05}
{"train_loss": 0.0610068179666996, "global_step": 203930, "epoch": 2291, "lr": 1.4555203551286795e-05}
{"train_loss": 0.02621448040008545, "global_step": 203931, "epoch": 2291, "lr": 1.4554794657851412e-05}
{"train_loss": 0.08529806137084961, "global_step": 203932, "epoch": 2291, "lr": 1.455438576918119e-05}
{"train_loss": 0.07602722197771072, "global_step": 203933, "epoch": 2291, "lr": 1.4553976885276216e-05}
{"train_loss": 0.07675397396087646, "global_step": 203934, "epoch": 2291, "lr": 1.4553568006136508e-05}
{"train_loss": 0.08418373763561249, "global_step": 203935, "epoch": 2291, "lr": 1.4553159131762161e-05}
{"train_loss": 0.053823426365852356, "global_step": 203936, "epoch": 2291, "lr": 1.4552750262153193e-05}
{"train_loss": 0.04130275920033455, "global_step": 203937, "epoch": 2291, "lr": 1.4552341397309688e-05}
{"train_loss": 0.08518426865339279, "global_step": 203938, "epoch": 2291, "lr": 1.4551932537231678e-05}
{"train_loss": 0.05516393482685089, "global_step": 203939, "epoch": 2291, "lr": 1.4551523681919244e-05}
{"train_loss": 0.06525786966085434, "global_step": 203940, "epoch": 2291, "lr": 1.455111483137241e-05}
{"train_loss": 0.061074148863554, "global_step": 203941, "epoch": 2291, "lr": 1.4550705985591257e-05}
{"train_loss": 0.07066450268030167, "global_step": 203942, "epoch": 2291, "lr": 1.4550297144575831e-05}
{"train_loss": 0.058616798371076584, "global_step": 203943, "epoch": 2291, "lr": 1.4549888308326171e-05}
{"train_loss": 0.05142219364643097, "global_step": 203944, "epoch": 2291, "lr": 1.4549479476842364e-05}
{"train_loss": 0.06936787068843842, "global_step": 203945, "epoch": 2291, "lr": 1.454907065012443e-05}
{"train_loss": 0.10135958343744278, "global_step": 203946, "epoch": 2291, "lr": 1.4548661828172454e-05}
{"train_loss": 0.07694652676582336, "global_step": 203947, "epoch": 2291, "lr": 1.4548253010986469e-05}
{"train_loss": 0.11839140951633453, "global_step": 203948, "epoch": 2291, "lr": 1.4547844198566551e-05}
{"train_loss": 0.04597528651356697, "global_step": 203949, "epoch": 2291, "lr": 1.4547435390912728e-05}
{"train_loss": 0.058513760566711426, "global_step": 203950, "epoch": 2291, "lr": 1.4547026588025086e-05}
{"train_loss": 0.037034086883068085, "global_step": 203951, "epoch": 2291, "lr": 1.454661778990365e-05}
{"train_loss": 0.035997580736875534, "global_step": 203952, "epoch": 2291, "lr": 1.4546208996548504e-05}
{"train_loss": 0.06480889022350311, "global_step": 203953, "epoch": 2291, "lr": 1.4545800207959676e-05}
{"train_loss": 0.03700655326247215, "global_step": 203954, "epoch": 2291, "lr": 1.4545391424137239e-05}
{"train_loss": 0.07210681587457657, "global_step": 203955, "epoch": 2291, "lr": 1.4544982645081234e-05}
{"train_loss": 0.0973891019821167, "global_step": 203956, "epoch": 2291, "lr": 1.454457387079174e-05}
{"train_loss": 0.052942704409360886, "global_step": 203957, "epoch": 2291, "lr": 1.4544165101268775e-05}
{"train_loss": 0.045971717685461044, "global_step": 203958, "epoch": 2291, "lr": 1.4543756336512421e-05}
{"train_loss": 0.08135196566581726, "global_step": 203959, "epoch": 2291, "lr": 1.4543347576522737e-05}
{"train_loss": 0.010545547120273113, "global_step": 203960, "epoch": 2291, "lr": 1.4542938821299756e-05}
{"train_loss": 0.05654902011156082, "global_step": 203961, "epoch": 2291, "lr": 1.4542530070843558e-05}
{"train_loss": 0.06258143484592438, "global_step": 203962, "epoch": 2291, "lr": 1.4542121325154168e-05}
{"train_loss": 0.07955633848905563, "global_step": 203963, "epoch": 2291, "lr": 1.4541712584231676e-05}
{"train_loss": 0.0462179109454155, "global_step": 203964, "epoch": 2291, "lr": 1.4541303848076105e-05}
{"train_loss": 0.0579683817923069, "global_step": 203965, "epoch": 2291, "lr": 1.4540895116687536e-05}
{"train_loss": 0.06960290670394897, "global_step": 203966, "epoch": 2291, "lr": 1.4540486390065993e-05}
{"train_loss": 0.08598855137825012, "global_step": 203967, "epoch": 2291, "lr": 1.4540077668211567e-05}
{"train_loss": 0.15622898936271667, "global_step": 203968, "epoch": 2291, "lr": 1.453966895112428e-05}
{"train_loss": 0.02635243348777294, "global_step": 203969, "epoch": 2291, "lr": 1.453926023880422e-05}
{"train_loss": 0.047401558607816696, "global_step": 203970, "epoch": 2291, "lr": 1.4538851531251407e-05}
{"train_loss": 0.055783677846193314, "global_step": 203971, "epoch": 2291, "lr": 1.4538442828465932e-05}
{"train_loss": 0.03300206735730171, "global_step": 203972, "epoch": 2291, "lr": 1.4538034130447809e-05}
{"train_loss": 0.04382120817899704, "global_step": 203973, "epoch": 2291, "lr": 1.453762543719714e-05}
{"train_loss": 0.031014014035463333, "global_step": 203974, "epoch": 2291, "lr": 1.4537216748713927e-05}
{"train_loss": 0.017805451527237892, "global_step": 203975, "epoch": 2291, "lr": 1.453680806499828e-05}
{"train_loss": 0.0947517454624176, "global_step": 203976, "epoch": 2291, "lr": 1.4536399386050203e-05}
{"train_loss": 0.0874604806303978, "global_step": 203977, "epoch": 2291, "lr": 1.4535990711869796e-05}
{"train_loss": 0.061947789043188095, "global_step": 203978, "epoch": 2291, "lr": 1.4535582042457075e-05}
{"train_loss": 0.07300173491239548, "global_step": 203979, "epoch": 2291, "lr": 1.4535173377812128e-05}
{"train_loss": 0.024308741092681885, "global_step": 203980, "epoch": 2291, "lr": 1.4534764717934996e-05}
{"train_loss": 0.07963234931230545, "global_step": 203981, "epoch": 2291, "lr": 1.4534356062825715e-05}
{"train_loss": 0.05420071631669998, "global_step": 203982, "epoch": 2291, "lr": 1.4533947412484378e-05}
{"train_loss": 0.011969520710408688, "global_step": 203983, "epoch": 2291, "lr": 1.4533538766910998e-05}
{"train_loss": 0.03505855053663254, "global_step": 203984, "epoch": 2291, "lr": 1.453313012610567e-05}
{"train_loss": 0.038628950715065, "global_step": 203985, "epoch": 2291, "lr": 1.4532721490068413e-05}
{"train_loss": 0.08180921524763107, "global_step": 203986, "epoch": 2291, "lr": 1.4532312858799318e-05}
{"train_loss": 0.06028210329875518, "global_step": 203987, "epoch": 2291, "lr": 1.4531904232298404e-05, "val_loss": 8.272122383117676}
{"train_loss": 0.12080613523721695, "global_step": 203988, "epoch": 2292, "lr": 1.453149561056576e-05}
{"train_loss": 0.035457998514175415, "global_step": 203989, "epoch": 2292, "lr": 1.4531086993601406e-05}
{"train_loss": 0.04379831627011299, "global_step": 203990, "epoch": 2292, "lr": 1.4530678381405432e-05}
{"train_loss": 0.10185903310775757, "global_step": 203991, "epoch": 2292, "lr": 1.4530269773977862e-05}
{"train_loss": 0.050751376897096634, "global_step": 203992, "epoch": 2292, "lr": 1.4529861171318776e-05}
{"train_loss": 0.06636898964643478, "global_step": 203993, "epoch": 2292, "lr": 1.4529452573428204e-05}
{"train_loss": 0.05629567801952362, "global_step": 203994, "epoch": 2292, "lr": 1.4529043980306218e-05}
{"train_loss": 0.11455176770687103, "global_step": 203995, "epoch": 2292, "lr": 1.4528635391952883e-05}
{"train_loss": 0.050718359649181366, "global_step": 203996, "epoch": 2292, "lr": 1.4528226808368223e-05}
{"train_loss": 0.03798957169055939, "global_step": 203997, "epoch": 2292, "lr": 1.4527818229552326e-05}
{"train_loss": 0.054925981909036636, "global_step": 203998, "epoch": 2292, "lr": 1.452740965550522e-05}
{"train_loss": 0.041340336203575134, "global_step": 203999, "epoch": 2292, "lr": 1.4527001086226988e-05}
{"train_loss": 0.06445658206939697, "global_step": 204000, "epoch": 2292, "lr": 1.4526592521717647e-05}
{"train_loss": 0.06224486231803894, "global_step": 204001, "epoch": 2292, "lr": 1.452618396197729e-05}
{"train_loss": 0.07739190757274628, "global_step": 204002, "epoch": 2292, "lr": 1.4525775407005942e-05}
{"train_loss": 0.03766944631934166, "global_step": 204003, "epoch": 2292, "lr": 1.4525366856803684e-05}
{"train_loss": 0.05767526105046272, "global_step": 204004, "epoch": 2292, "lr": 1.4524958311370545e-05}
{"train_loss": 0.04693718999624252, "global_step": 204005, "epoch": 2292, "lr": 1.4524549770706614e-05}
{"train_loss": 0.06659501045942307, "global_step": 204006, "epoch": 2292, "lr": 1.45241412348119e-05}
{"train_loss": 0.11377757042646408, "global_step": 204007, "epoch": 2292, "lr": 1.4523732703686504e-05}
{"train_loss": 0.043121516704559326, "global_step": 204008, "epoch": 2292, "lr": 1.4523324177330438e-05}
{"train_loss": 0.09389525651931763, "global_step": 204009, "epoch": 2292, "lr": 1.4522915655743803e-05}
{"train_loss": 0.14825455844402313, "global_step": 204010, "epoch": 2292, "lr": 1.4522507138926605e-05}
{"train_loss": 0.06372695416212082, "global_step": 204011, "epoch": 2292, "lr": 1.4522098626878927e-05}
{"train_loss": 0.05781395360827446, "global_step": 204012, "epoch": 2292, "lr": 1.452169011960084e-05}
{"train_loss": 0.0410771444439888, "global_step": 204013, "epoch": 2292, "lr": 1.4521281617092358e-05}
{"train_loss": 0.09436231851577759, "global_step": 204014, "epoch": 2292, "lr": 1.452087311935358e-05}
{"train_loss": 0.07084357738494873, "global_step": 204015, "epoch": 2292, "lr": 1.4520464626384528e-05}
{"train_loss": 0.028118891641497612, "global_step": 204016, "epoch": 2292, "lr": 1.4520056138185261e-05}
{"train_loss": 0.040429338812828064, "global_step": 204017, "epoch": 2292, "lr": 1.4519647654755853e-05}
{"train_loss": 0.045450154691934586, "global_step": 204018, "epoch": 2292, "lr": 1.4519239176096327e-05}
{"train_loss": 0.07011791318655014, "global_step": 204019, "epoch": 2292, "lr": 1.451883070220677e-05}
{"train_loss": 0.09409958869218826, "global_step": 204020, "epoch": 2292, "lr": 1.4518422233087231e-05}
{"train_loss": 0.08896545320749283, "global_step": 204021, "epoch": 2292, "lr": 1.4518013768737737e-05}
{"train_loss": 0.046375956386327744, "global_step": 204022, "epoch": 2292, "lr": 1.4517605309158383e-05}
{"train_loss": 0.0728912279009819, "global_step": 204023, "epoch": 2292, "lr": 1.4517196854349185e-05}
{"train_loss": 0.058618564158678055, "global_step": 204024, "epoch": 2292, "lr": 1.4516788404310239e-05}
{"train_loss": 0.06282268464565277, "global_step": 204025, "epoch": 2292, "lr": 1.451637995904156e-05}
{"train_loss": 0.02770231105387211, "global_step": 204026, "epoch": 2292, "lr": 1.4515971518543237e-05}
{"train_loss": 0.06879103928804398, "global_step": 204027, "epoch": 2292, "lr": 1.4515563082815293e-05}
{"train_loss": 0.01806354895234108, "global_step": 204028, "epoch": 2292, "lr": 1.451515465185781e-05}
{"train_loss": 0.0562259703874588, "global_step": 204029, "epoch": 2292, "lr": 1.4514746225670822e-05}
{"train_loss": 0.0747578963637352, "global_step": 204030, "epoch": 2292, "lr": 1.451433780425439e-05}
{"train_loss": 0.08382014185190201, "global_step": 204031, "epoch": 2292, "lr": 1.4513929387608593e-05}
{"train_loss": 0.06826615333557129, "global_step": 204032, "epoch": 2292, "lr": 1.4513520975733446e-05}
{"train_loss": 0.0716630071401596, "global_step": 204033, "epoch": 2292, "lr": 1.451311256862904e-05}
{"train_loss": 0.04983897507190704, "global_step": 204034, "epoch": 2292, "lr": 1.4512704166295399e-05}
{"train_loss": 0.11212125420570374, "global_step": 204035, "epoch": 2292, "lr": 1.451229576873261e-05}
{"train_loss": 0.021261973306536674, "global_step": 204036, "epoch": 2292, "lr": 1.4511887375940691e-05}
{"train_loss": 0.06829345226287842, "global_step": 204037, "epoch": 2292, "lr": 1.4511478987919735e-05}
{"train_loss": 0.046134915202856064, "global_step": 204038, "epoch": 2292, "lr": 1.4511070604669757e-05}
{"train_loss": 0.08462011814117432, "global_step": 204039, "epoch": 2292, "lr": 1.4510662226190858e-05}
{"train_loss": 0.05295049026608467, "global_step": 204040, "epoch": 2292, "lr": 1.451025385248304e-05}
{"train_loss": 0.07614336162805557, "global_step": 204041, "epoch": 2292, "lr": 1.4509845483546414e-05}
{"train_loss": 0.03494991362094879, "global_step": 204042, "epoch": 2292, "lr": 1.4509437119380981e-05}
{"train_loss": 0.05147513374686241, "global_step": 204043, "epoch": 2292, "lr": 1.4509028759986842e-05}
{"train_loss": 0.061252396553754807, "global_step": 204044, "epoch": 2292, "lr": 1.4508620405364015e-05}
{"train_loss": 0.051687948405742645, "global_step": 204045, "epoch": 2292, "lr": 1.4508212055512588e-05}
{"train_loss": 0.07037393748760223, "global_step": 204046, "epoch": 2292, "lr": 1.4507803710432583e-05}
{"train_loss": 0.07444853335618973, "global_step": 204047, "epoch": 2292, "lr": 1.4507395370124088e-05}
{"train_loss": 0.07287727296352386, "global_step": 204048, "epoch": 2292, "lr": 1.4506987034587121e-05}
{"train_loss": 0.03745070844888687, "global_step": 204049, "epoch": 2292, "lr": 1.4506578703821766e-05}
{"train_loss": 0.058288950473070145, "global_step": 204050, "epoch": 2292, "lr": 1.4506170377828077e-05}
{"train_loss": 0.08262018114328384, "global_step": 204051, "epoch": 2292, "lr": 1.4505762056606087e-05}
{"train_loss": 0.07840965688228607, "global_step": 204052, "epoch": 2292, "lr": 1.450535374015588e-05}
{"train_loss": 0.06762412190437317, "global_step": 204053, "epoch": 2292, "lr": 1.4504945428477495e-05}
{"train_loss": 0.08954279869794846, "global_step": 204054, "epoch": 2292, "lr": 1.450453712157097e-05}
{"train_loss": 0.05395538732409477, "global_step": 204055, "epoch": 2292, "lr": 1.4504128819436396e-05}
{"train_loss": 0.039126571267843246, "global_step": 204056, "epoch": 2292, "lr": 1.4503720522073793e-05}
{"train_loss": 0.06888799369335175, "global_step": 204057, "epoch": 2292, "lr": 1.4503312229483252e-05}
{"train_loss": 0.07561895251274109, "global_step": 204058, "epoch": 2292, "lr": 1.4502903941664803e-05}
{"train_loss": 0.07670074701309204, "global_step": 204059, "epoch": 2292, "lr": 1.4502495658618493e-05}
{"train_loss": 0.05839552357792854, "global_step": 204060, "epoch": 2292, "lr": 1.4502087380344403e-05}
{"train_loss": 0.05518566444516182, "global_step": 204061, "epoch": 2292, "lr": 1.4501679106842559e-05}
{"train_loss": 0.042977042496204376, "global_step": 204062, "epoch": 2292, "lr": 1.450127083811305e-05}
{"train_loss": 0.05751141905784607, "global_step": 204063, "epoch": 2292, "lr": 1.4500862574155894e-05}
{"train_loss": 0.03311992809176445, "global_step": 204064, "epoch": 2292, "lr": 1.4500454314971179e-05}
{"train_loss": 0.06679059565067291, "global_step": 204065, "epoch": 2292, "lr": 1.450004606055893e-05}
{"train_loss": 0.05277447775006294, "global_step": 204066, "epoch": 2292, "lr": 1.449963781091922e-05}
{"train_loss": 0.07916701585054398, "global_step": 204067, "epoch": 2292, "lr": 1.449922956605212e-05}
{"train_loss": 0.04545500874519348, "global_step": 204068, "epoch": 2292, "lr": 1.449882132595764e-05}
{"train_loss": 0.07171200215816498, "global_step": 204069, "epoch": 2292, "lr": 1.4498413090635887e-05}
{"train_loss": 0.05535445734858513, "global_step": 204070, "epoch": 2292, "lr": 1.4498004860086867e-05}
{"train_loss": 0.06832613050937653, "global_step": 204071, "epoch": 2292, "lr": 1.4497596634310672e-05}
{"train_loss": 0.08300174027681351, "global_step": 204072, "epoch": 2292, "lr": 1.4497188413307332e-05}
{"train_loss": 0.016618920490145683, "global_step": 204073, "epoch": 2292, "lr": 1.4496780197076925e-05}
{"train_loss": 0.037137579172849655, "global_step": 204074, "epoch": 2292, "lr": 1.4496371985619479e-05}
{"train_loss": 0.053279679268598557, "global_step": 204075, "epoch": 2292, "lr": 1.4495963778935078e-05}
{"train_loss": 0.0622270538816961, "global_step": 204076, "epoch": 2292, "lr": 1.4495555577023746e-05, "val_loss": 8.281455993652344}
{"train_loss": 0.012352711521089077, "global_step": 204077, "epoch": 2293, "lr": 1.449514737988557e-05}
{"train_loss": 0.052640266716480255, "global_step": 204078, "epoch": 2293, "lr": 1.4494739187520572e-05}
{"train_loss": 0.06211536005139351, "global_step": 204079, "epoch": 2293, "lr": 1.449433099992884e-05}
{"train_loss": 0.06230451539158821, "global_step": 204080, "epoch": 2293, "lr": 1.4493922817110401e-05}
{"train_loss": 0.06309378147125244, "global_step": 204081, "epoch": 2293, "lr": 1.4493514639065331e-05}
{"train_loss": 0.035751305520534515, "global_step": 204082, "epoch": 2293, "lr": 1.449310646579366e-05}
{"train_loss": 0.012415292672812939, "global_step": 204083, "epoch": 2293, "lr": 1.4492698297295477e-05}
{"train_loss": 0.07562731206417084, "global_step": 204084, "epoch": 2293, "lr": 1.4492290133570802e-05}
{"train_loss": 0.07249073684215546, "global_step": 204085, "epoch": 2293, "lr": 1.4491881974619703e-05}
{"train_loss": 0.06208950653672218, "global_step": 204086, "epoch": 2293, "lr": 1.4491473820442253e-05}
{"train_loss": 0.08075999468564987, "global_step": 204087, "epoch": 2293, "lr": 1.4491065671038478e-05}
{"train_loss": 0.03720336779952049, "global_step": 204088, "epoch": 2293, "lr": 1.4490657526408463e-05}
{"train_loss": 0.04135829210281372, "global_step": 204089, "epoch": 2293, "lr": 1.449024938655223e-05}
{"train_loss": 0.039193131029605865, "global_step": 204090, "epoch": 2293, "lr": 1.4489841251469866e-05}
{"train_loss": 0.11453825980424881, "global_step": 204091, "epoch": 2293, "lr": 1.4489433121161405e-05}
{"train_loss": 0.11920653283596039, "global_step": 204092, "epoch": 2293, "lr": 1.4489024995626899e-05}
{"train_loss": 0.036337848752737045, "global_step": 204093, "epoch": 2293, "lr": 1.4488616874866424e-05}
{"train_loss": 0.070103220641613, "global_step": 204094, "epoch": 2293, "lr": 1.448820875888e-05}
{"train_loss": 0.043178290128707886, "global_step": 204095, "epoch": 2293, "lr": 1.4487800647667726e-05}
{"train_loss": 0.08483314514160156, "global_step": 204096, "epoch": 2293, "lr": 1.4487392541229633e-05}
{"train_loss": 0.03942251577973366, "global_step": 204097, "epoch": 2293, "lr": 1.4486984439565759e-05}
{"train_loss": 0.06220943108201027, "global_step": 204098, "epoch": 2293, "lr": 1.4486576342676195e-05}
{"train_loss": 0.07878024876117706, "global_step": 204099, "epoch": 2293, "lr": 1.4486168250560961e-05}
{"train_loss": 0.06904589384794235, "global_step": 204100, "epoch": 2293, "lr": 1.4485760163220147e-05}
{"train_loss": 0.042405400425195694, "global_step": 204101, "epoch": 2293, "lr": 1.448535208065377e-05}
{"train_loss": 0.03328379616141319, "global_step": 204102, "epoch": 2293, "lr": 1.448494400286191e-05}
{"train_loss": 0.05762813612818718, "global_step": 204103, "epoch": 2293, "lr": 1.4484535929844628e-05}
{"train_loss": 0.05974609777331352, "global_step": 204104, "epoch": 2293, "lr": 1.4484127861601954e-05}
{"train_loss": 0.05736807361245155, "global_step": 204105, "epoch": 2293, "lr": 1.4483719798133971e-05}
{"train_loss": 0.0669311136007309, "global_step": 204106, "epoch": 2293, "lr": 1.4483311739440702e-05}
{"train_loss": 0.05898350477218628, "global_step": 204107, "epoch": 2293, "lr": 1.4482903685522231e-05}
{"train_loss": 0.07077842205762863, "global_step": 204108, "epoch": 2293, "lr": 1.448249563637859e-05}
{"train_loss": 0.044694654643535614, "global_step": 204109, "epoch": 2293, "lr": 1.4482087592009863e-05}
{"train_loss": 0.07698649913072586, "global_step": 204110, "epoch": 2293, "lr": 1.4481679552416067e-05}
{"train_loss": 0.05516605079174042, "global_step": 204111, "epoch": 2293, "lr": 1.448127151759729e-05}
{"train_loss": 0.06434149295091629, "global_step": 204112, "epoch": 2293, "lr": 1.4480863487553558e-05}
{"train_loss": 0.05306483432650566, "global_step": 204113, "epoch": 2293, "lr": 1.4480455462284958e-05}
{"train_loss": 0.021041864529252052, "global_step": 204114, "epoch": 2293, "lr": 1.4480047441791517e-05}
{"train_loss": 0.027264360338449478, "global_step": 204115, "epoch": 2293, "lr": 1.447963942607331e-05}
{"train_loss": 0.057374004274606705, "global_step": 204116, "epoch": 2293, "lr": 1.4479231415130368e-05}
{"train_loss": 0.06742986291646957, "global_step": 204117, "epoch": 2293, "lr": 1.4478823408962778e-05}
{"train_loss": 0.08213432878255844, "global_step": 204118, "epoch": 2293, "lr": 1.4478415407570561e-05}
{"train_loss": 0.05061400309205055, "global_step": 204119, "epoch": 2293, "lr": 1.4478007410953803e-05}
{"train_loss": 0.024653788655996323, "global_step": 204120, "epoch": 2293, "lr": 1.4477599419112532e-05}
{"train_loss": 0.06402558833360672, "global_step": 204121, "epoch": 2293, "lr": 1.4477191432046805e-05}
{"train_loss": 0.059608642011880875, "global_step": 204122, "epoch": 2293, "lr": 1.447678344975671e-05}
{"train_loss": 0.06332993507385254, "global_step": 204123, "epoch": 2293, "lr": 1.4476375472242265e-05}
{"train_loss": 0.0631377100944519, "global_step": 204124, "epoch": 2293, "lr": 1.4475967499503545e-05}
{"train_loss": 0.03380238264799118, "global_step": 204125, "epoch": 2293, "lr": 1.447555953154059e-05}
{"train_loss": 0.02768794260919094, "global_step": 204126, "epoch": 2293, "lr": 1.4475151568353479e-05}
{"train_loss": 0.040414873510599136, "global_step": 204127, "epoch": 2293, "lr": 1.4474743609942232e-05}
{"train_loss": 0.04078930616378784, "global_step": 204128, "epoch": 2293, "lr": 1.4474335656306937e-05}
{"train_loss": 0.05654308199882507, "global_step": 204129, "epoch": 2293, "lr": 1.447392770744762e-05}
{"train_loss": 0.04212331399321556, "global_step": 204130, "epoch": 2293, "lr": 1.4473519763364368e-05}
{"train_loss": 0.07354888319969177, "global_step": 204131, "epoch": 2293, "lr": 1.4473111824057218e-05}
{"train_loss": 0.07320111244916916, "global_step": 204132, "epoch": 2293, "lr": 1.4472703889526213e-05}
{"train_loss": 0.07586856186389923, "global_step": 204133, "epoch": 2293, "lr": 1.4472295959771432e-05}
{"train_loss": 0.03853698447346687, "global_step": 204134, "epoch": 2293, "lr": 1.4471888034792896e-05}
{"train_loss": 0.04455430805683136, "global_step": 204135, "epoch": 2293, "lr": 1.4471480114590707e-05}
{"train_loss": 0.05828755721449852, "global_step": 204136, "epoch": 2293, "lr": 1.4471072199164888e-05}
{"train_loss": 0.0549822673201561, "global_step": 204137, "epoch": 2293, "lr": 1.447066428851549e-05}
{"train_loss": 0.0403965599834919, "global_step": 204138, "epoch": 2293, "lr": 1.4470256382642572e-05}
{"train_loss": 0.04908633977174759, "global_step": 204139, "epoch": 2293, "lr": 1.4469848481546216e-05}
{"train_loss": 0.08128658682107925, "global_step": 204140, "epoch": 2293, "lr": 1.446944058522644e-05}
{"train_loss": 0.0329333171248436, "global_step": 204141, "epoch": 2293, "lr": 1.4469032693683327e-05}
{"train_loss": 0.0876990407705307, "global_step": 204142, "epoch": 2293, "lr": 1.4468624806916902e-05}
{"train_loss": 0.1153717190027237, "global_step": 204143, "epoch": 2293, "lr": 1.4468216924927258e-05}
{"train_loss": 0.05182774364948273, "global_step": 204144, "epoch": 2293, "lr": 1.4467809047714415e-05}
{"train_loss": 0.08508588373661041, "global_step": 204145, "epoch": 2293, "lr": 1.446740117527845e-05}
{"train_loss": 0.05471215397119522, "global_step": 204146, "epoch": 2293, "lr": 1.4466993307619398e-05}
{"train_loss": 0.04245191812515259, "global_step": 204147, "epoch": 2293, "lr": 1.4466585444737346e-05}
{"train_loss": 0.05976690724492073, "global_step": 204148, "epoch": 2293, "lr": 1.4466177586632306e-05}
{"train_loss": 0.07227492332458496, "global_step": 204149, "epoch": 2293, "lr": 1.4465769733304374e-05}
{"train_loss": 0.06632855534553528, "global_step": 204150, "epoch": 2293, "lr": 1.446536188475357e-05}
{"train_loss": 0.03791501000523567, "global_step": 204151, "epoch": 2293, "lr": 1.4464954040979983e-05}
{"train_loss": 0.07516475021839142, "global_step": 204152, "epoch": 2293, "lr": 1.446454620198363e-05}
{"train_loss": 0.10652618110179901, "global_step": 204153, "epoch": 2293, "lr": 1.44641383677646e-05}
{"train_loss": 0.07968773692846298, "global_step": 204154, "epoch": 2293, "lr": 1.4463730538322923e-05}
{"train_loss": 0.08807897567749023, "global_step": 204155, "epoch": 2293, "lr": 1.4463322713658678e-05}
{"train_loss": 0.031242305412888527, "global_step": 204156, "epoch": 2293, "lr": 1.446291489377189e-05}
{"train_loss": 0.1246412843465805, "global_step": 204157, "epoch": 2293, "lr": 1.4462507078662635e-05}
{"train_loss": 0.051714006811380386, "global_step": 204158, "epoch": 2293, "lr": 1.446209926833097e-05}
{"train_loss": 0.04303290322422981, "global_step": 204159, "epoch": 2293, "lr": 1.4461691462776933e-05}
{"train_loss": 0.04928254336118698, "global_step": 204160, "epoch": 2293, "lr": 1.4461283662000602e-05}
{"train_loss": 0.03374524042010307, "global_step": 204161, "epoch": 2293, "lr": 1.4460875866002e-05}
{"train_loss": 0.04536174237728119, "global_step": 204162, "epoch": 2293, "lr": 1.446046807478122e-05}
{"train_loss": 0.05476263910531998, "global_step": 204163, "epoch": 2293, "lr": 1.4460060288338278e-05}
{"train_loss": 0.08568727970123291, "global_step": 204164, "epoch": 2293, "lr": 1.4459652506673266e-05}
{"train_loss": 0.05870106920934795, "global_step": 204165, "epoch": 2293, "lr": 1.4459244729786204e-05, "val_loss": 8.290593147277832}
{"train_loss": 0.07134751230478287, "global_step": 204166, "epoch": 2294, "lr": 1.445883695767718e-05}
{"train_loss": 0.026548082008957863, "global_step": 204167, "epoch": 2294, "lr": 1.4458429190346213e-05}
{"train_loss": 0.05529087781906128, "global_step": 204168, "epoch": 2294, "lr": 1.4458021427793394e-05}
{"train_loss": 0.08212584257125854, "global_step": 204169, "epoch": 2294, "lr": 1.4457613670018766e-05}
{"train_loss": 0.11518529802560806, "global_step": 204170, "epoch": 2294, "lr": 1.445720591702236e-05}
{"train_loss": 0.03978660702705383, "global_step": 204171, "epoch": 2294, "lr": 1.4456798168804259e-05}
{"train_loss": 0.049720533192157745, "global_step": 204172, "epoch": 2294, "lr": 1.4456390425364502e-05}
{"train_loss": 0.007499458268284798, "global_step": 204173, "epoch": 2294, "lr": 1.4455982686703157e-05}
{"train_loss": 0.080680251121521, "global_step": 204174, "epoch": 2294, "lr": 1.4455574952820278e-05}
{"train_loss": 0.1399170607328415, "global_step": 204175, "epoch": 2294, "lr": 1.4455167223715899e-05}
{"train_loss": 0.09522680938243866, "global_step": 204176, "epoch": 2294, "lr": 1.4454759499390086e-05}
{"train_loss": 0.11786752939224243, "global_step": 204177, "epoch": 2294, "lr": 1.4454351779842912e-05}
{"train_loss": 0.023413119837641716, "global_step": 204178, "epoch": 2294, "lr": 1.4453944065074403e-05}
{"train_loss": 0.05671938508749008, "global_step": 204179, "epoch": 2294, "lr": 1.4453536355084646e-05}
{"train_loss": 0.06109508126974106, "global_step": 204180, "epoch": 2294, "lr": 1.445312864987366e-05}
{"train_loss": 0.08415956050157547, "global_step": 204181, "epoch": 2294, "lr": 1.4452720949441534e-05}
{"train_loss": 0.07910162210464478, "global_step": 204182, "epoch": 2294, "lr": 1.4452313253788285e-05}
{"train_loss": 0.034107498824596405, "global_step": 204183, "epoch": 2294, "lr": 1.4451905562914014e-05}
{"train_loss": 0.02517954260110855, "global_step": 204184, "epoch": 2294, "lr": 1.4451497876818731e-05}
{"train_loss": 0.09873643517494202, "global_step": 204185, "epoch": 2294, "lr": 1.4451090195502527e-05}
{"train_loss": 0.028948364779353142, "global_step": 204186, "epoch": 2294, "lr": 1.4450682518965424e-05}
{"train_loss": 0.04988779127597809, "global_step": 204187, "epoch": 2294, "lr": 1.4450274847207512e-05}
{"train_loss": 0.09311307966709137, "global_step": 204188, "epoch": 2294, "lr": 1.4449867180228804e-05}
{"train_loss": 0.08357679843902588, "global_step": 204189, "epoch": 2294, "lr": 1.4449459518029402e-05}
{"train_loss": 0.09637648612260818, "global_step": 204190, "epoch": 2294, "lr": 1.444905186060932e-05}
{"train_loss": 0.0744030773639679, "global_step": 204191, "epoch": 2294, "lr": 1.4448644207968643e-05}
{"train_loss": 0.1306730955839157, "global_step": 204192, "epoch": 2294, "lr": 1.4448236560107397e-05}
{"train_loss": 0.031200360506772995, "global_step": 204193, "epoch": 2294, "lr": 1.4447828917025652e-05}
{"train_loss": 0.02213549241423607, "global_step": 204194, "epoch": 2294, "lr": 1.4447421278723483e-05}
{"train_loss": 0.08532676100730896, "global_step": 204195, "epoch": 2294, "lr": 1.4447013645200901e-05}
{"train_loss": 0.01568029075860977, "global_step": 204196, "epoch": 2294, "lr": 1.4446606016458009e-05}
{"train_loss": 0.08624086529016495, "global_step": 204197, "epoch": 2294, "lr": 1.4446198392494814e-05}
{"train_loss": 0.04712506756186485, "global_step": 204198, "epoch": 2294, "lr": 1.4445790773311419e-05}
{"train_loss": 0.04701079800724983, "global_step": 204199, "epoch": 2294, "lr": 1.4445383158907832e-05}
{"train_loss": 0.0473288893699646, "global_step": 204200, "epoch": 2294, "lr": 1.4444975549284146e-05}
{"train_loss": 0.09534449130296707, "global_step": 204201, "epoch": 2294, "lr": 1.4444567944440385e-05}
{"train_loss": 0.08243168890476227, "global_step": 204202, "epoch": 2294, "lr": 1.4444160344376634e-05}
{"train_loss": 0.06378660351037979, "global_step": 204203, "epoch": 2294, "lr": 1.4443752749092915e-05}
{"train_loss": 0.11703091859817505, "global_step": 204204, "epoch": 2294, "lr": 1.4443345158589317e-05}
{"train_loss": 0.07752564549446106, "global_step": 204205, "epoch": 2294, "lr": 1.4442937572865866e-05}
{"train_loss": 0.12299364060163498, "global_step": 204206, "epoch": 2294, "lr": 1.4442529991922638e-05}
{"train_loss": 0.09377845376729965, "global_step": 204207, "epoch": 2294, "lr": 1.4442122415759685e-05}
{"train_loss": 0.031054720282554626, "global_step": 204208, "epoch": 2294, "lr": 1.4441714844377036e-05}
{"train_loss": 0.10202687233686447, "global_step": 204209, "epoch": 2294, "lr": 1.444130727777478e-05}
{"train_loss": 0.053267329931259155, "global_step": 204210, "epoch": 2294, "lr": 1.4440899715952938e-05}
{"train_loss": 0.06653595715761185, "global_step": 204211, "epoch": 2294, "lr": 1.4440492158911605e-05}
{"train_loss": 0.020921500399708748, "global_step": 204212, "epoch": 2294, "lr": 1.44400846066508e-05}
{"train_loss": 0.05062223598361015, "global_step": 204213, "epoch": 2294, "lr": 1.4439677059170603e-05}
{"train_loss": 0.061729416251182556, "global_step": 204214, "epoch": 2294, "lr": 1.4439269516471044e-05}
{"train_loss": 0.0446791872382164, "global_step": 204215, "epoch": 2294, "lr": 1.443886197855221e-05}
{"train_loss": 0.05201096087694168, "global_step": 204216, "epoch": 2294, "lr": 1.4438454445414118e-05}
{"train_loss": 0.060623276978731155, "global_step": 204217, "epoch": 2294, "lr": 1.4438046917056858e-05}
{"train_loss": 0.08964061737060547, "global_step": 204218, "epoch": 2294, "lr": 1.4437639393480457e-05}
{"train_loss": 0.06573908776044846, "global_step": 204219, "epoch": 2294, "lr": 1.4437231874684993e-05}
{"train_loss": 0.0709766373038292, "global_step": 204220, "epoch": 2294, "lr": 1.4436824360670498e-05}
{"train_loss": 0.03826360031962395, "global_step": 204221, "epoch": 2294, "lr": 1.4436416851437051e-05}
{"train_loss": 0.022759458050131798, "global_step": 204222, "epoch": 2294, "lr": 1.443600934698468e-05}
{"train_loss": 0.0630217045545578, "global_step": 204223, "epoch": 2294, "lr": 1.4435601847313473e-05}
{"train_loss": 0.0852954313158989, "global_step": 204224, "epoch": 2294, "lr": 1.4435194352423447e-05}
{"train_loss": 0.04016464203596115, "global_step": 204225, "epoch": 2294, "lr": 1.4434786862314692e-05}
{"train_loss": 0.09310010820627213, "global_step": 204226, "epoch": 2294, "lr": 1.4434379376987229e-05}
{"train_loss": 0.030068304389715195, "global_step": 204227, "epoch": 2294, "lr": 1.4433971896441152e-05}
{"train_loss": 0.037338174879550934, "global_step": 204228, "epoch": 2294, "lr": 1.443356442067647e-05}
{"train_loss": 0.02012256160378456, "global_step": 204229, "epoch": 2294, "lr": 1.4433156949693267e-05}
{"train_loss": 0.07164838910102844, "global_step": 204230, "epoch": 2294, "lr": 1.4432749483491603e-05}
{"train_loss": 0.09095999598503113, "global_step": 204231, "epoch": 2294, "lr": 1.4432342022071516e-05}
{"train_loss": 0.08112117648124695, "global_step": 204232, "epoch": 2294, "lr": 1.4431934565433076e-05}
{"train_loss": 0.07044155150651932, "global_step": 204233, "epoch": 2294, "lr": 1.443152711357631e-05}
{"train_loss": 0.10582006722688675, "global_step": 204234, "epoch": 2294, "lr": 1.4431119666501314e-05}
{"train_loss": 0.09111811965703964, "global_step": 204235, "epoch": 2294, "lr": 1.4430712224208103e-05}
{"train_loss": 0.09542166441679001, "global_step": 204236, "epoch": 2294, "lr": 1.4430304786696763e-05}
{"train_loss": 0.051239170134067535, "global_step": 204237, "epoch": 2294, "lr": 1.442989735396732e-05}
{"train_loss": 0.07136581838130951, "global_step": 204238, "epoch": 2294, "lr": 1.4429489926019862e-05}
{"train_loss": 0.07122694700956345, "global_step": 204239, "epoch": 2294, "lr": 1.4429082502854407e-05}
{"train_loss": 0.0835094302892685, "global_step": 204240, "epoch": 2294, "lr": 1.4428675084471045e-05}
{"train_loss": 0.0777985006570816, "global_step": 204241, "epoch": 2294, "lr": 1.44282676708698e-05}
{"train_loss": 0.10789988934993744, "global_step": 204242, "epoch": 2294, "lr": 1.4427860262050752e-05}
{"train_loss": 0.057175278663635254, "global_step": 204243, "epoch": 2294, "lr": 1.4427452858013934e-05}
{"train_loss": 0.06220478564500809, "global_step": 204244, "epoch": 2294, "lr": 1.4427045458759425e-05}
{"train_loss": 0.08221490681171417, "global_step": 204245, "epoch": 2294, "lr": 1.4426638064287252e-05}
{"train_loss": 0.045561231672763824, "global_step": 204246, "epoch": 2294, "lr": 1.4426230674597496e-05}
{"train_loss": 0.05312182009220123, "global_step": 204247, "epoch": 2294, "lr": 1.4425823289690199e-05}
{"train_loss": 0.07569637149572372, "global_step": 204248, "epoch": 2294, "lr": 1.4425415909565404e-05}
{"train_loss": 0.07672841846942902, "global_step": 204249, "epoch": 2294, "lr": 1.4425008534223194e-05}
{"train_loss": 0.05882633849978447, "global_step": 204250, "epoch": 2294, "lr": 1.4424601163663593e-05}
{"train_loss": 0.043468862771987915, "global_step": 204251, "epoch": 2294, "lr": 1.4424193797886687e-05}
{"train_loss": 0.04922041669487953, "global_step": 204252, "epoch": 2294, "lr": 1.4423786436892495e-05}
{"train_loss": 0.05400179326534271, "global_step": 204253, "epoch": 2294, "lr": 1.442337908068111e-05}
{"train_loss": 0.06628957135456331, "global_step": 204254, "epoch": 2294, "lr": 1.442297172925255e-05, "val_loss": 8.325721740722656}
{"train_loss": 0.034267548471689224, "global_step": 204255, "epoch": 2295, "lr": 1.4422564382606906e-05}
{"train_loss": 0.08045829087495804, "global_step": 204256, "epoch": 2295, "lr": 1.44221570407442e-05}
{"train_loss": 0.05793405696749687, "global_step": 204257, "epoch": 2295, "lr": 1.4421749703664512e-05}
{"train_loss": 0.03524192422628403, "global_step": 204258, "epoch": 2295, "lr": 1.4421342371367874e-05}
{"train_loss": 0.0693851038813591, "global_step": 204259, "epoch": 2295, "lr": 1.4420935043854367e-05}
{"train_loss": 0.04915430396795273, "global_step": 204260, "epoch": 2295, "lr": 1.4420527721124016e-05}
{"train_loss": 0.07682301104068756, "global_step": 204261, "epoch": 2295, "lr": 1.4420120403176906e-05}
{"train_loss": 0.038035422563552856, "global_step": 204262, "epoch": 2295, "lr": 1.4419713090013066e-05}
{"train_loss": 0.06924912333488464, "global_step": 204263, "epoch": 2295, "lr": 1.4419305781632574e-05}
{"train_loss": 0.1302419900894165, "global_step": 204264, "epoch": 2295, "lr": 1.441889847803546e-05}
{"train_loss": 0.13175125420093536, "global_step": 204265, "epoch": 2295, "lr": 1.4418491179221788e-05}
{"train_loss": 0.09253070503473282, "global_step": 204266, "epoch": 2295, "lr": 1.4418083885191636e-05}
{"train_loss": 0.060803256928920746, "global_step": 204267, "epoch": 2295, "lr": 1.4417676595945018e-05}
{"train_loss": 0.06301327049732208, "global_step": 204268, "epoch": 2295, "lr": 1.441726931148203e-05}
{"train_loss": 0.0887274220585823, "global_step": 204269, "epoch": 2295, "lr": 1.4416862031802686e-05}
{"train_loss": 0.09490251541137695, "global_step": 204270, "epoch": 2295, "lr": 1.4416454756907077e-05}
{"train_loss": 0.07132072746753693, "global_step": 204271, "epoch": 2295, "lr": 1.4416047486795231e-05}
{"train_loss": 0.03814921900629997, "global_step": 204272, "epoch": 2295, "lr": 1.4415640221467224e-05}
{"train_loss": 0.06515218317508698, "global_step": 204273, "epoch": 2295, "lr": 1.4415232960923091e-05}
{"train_loss": 0.05918782576918602, "global_step": 204274, "epoch": 2295, "lr": 1.441482570516291e-05}
{"train_loss": 0.04358820617198944, "global_step": 204275, "epoch": 2295, "lr": 1.4414418454186706e-05}
{"train_loss": 0.06502071022987366, "global_step": 204276, "epoch": 2295, "lr": 1.4414011207994566e-05}
{"train_loss": 0.05159236118197441, "global_step": 204277, "epoch": 2295, "lr": 1.4413603966586508e-05}
{"train_loss": 0.061439044773578644, "global_step": 204278, "epoch": 2295, "lr": 1.4413196729962625e-05}
{"train_loss": 0.061136554926633835, "global_step": 204279, "epoch": 2295, "lr": 1.4412789498122941e-05}
{"train_loss": 0.058962687849998474, "global_step": 204280, "epoch": 2295, "lr": 1.4412382271067537e-05}
{"train_loss": 0.04397987946867943, "global_step": 204281, "epoch": 2295, "lr": 1.4411975048796438e-05}
{"train_loss": 0.03574708476662636, "global_step": 204282, "epoch": 2295, "lr": 1.4411567831309736e-05}
{"train_loss": 0.06990965455770493, "global_step": 204283, "epoch": 2295, "lr": 1.4411160618607444e-05}
{"train_loss": 0.07888588309288025, "global_step": 204284, "epoch": 2295, "lr": 1.4410753410689653e-05}
{"train_loss": 0.05820626765489578, "global_step": 204285, "epoch": 2295, "lr": 1.4410346207556407e-05}
{"train_loss": 0.11336394399404526, "global_step": 204286, "epoch": 2295, "lr": 1.440993900920774e-05}
{"train_loss": 0.08327897638082504, "global_step": 204287, "epoch": 2295, "lr": 1.4409531815643734e-05}
{"train_loss": 0.03793487325310707, "global_step": 204288, "epoch": 2295, "lr": 1.4409124626864418e-05}
{"train_loss": 0.05126035958528519, "global_step": 204289, "epoch": 2295, "lr": 1.4408717442869878e-05}
{"train_loss": 0.07795582711696625, "global_step": 204290, "epoch": 2295, "lr": 1.4408310263660141e-05}
{"train_loss": 0.03262593597173691, "global_step": 204291, "epoch": 2295, "lr": 1.4407903089235287e-05}
{"train_loss": 0.02487487718462944, "global_step": 204292, "epoch": 2295, "lr": 1.4407495919595338e-05}
{"train_loss": 0.05988297238945961, "global_step": 204293, "epoch": 2295, "lr": 1.4407088754740383e-05}
{"train_loss": 0.08408208191394806, "global_step": 204294, "epoch": 2295, "lr": 1.4406681594670446e-05}
{"train_loss": 0.03697289526462555, "global_step": 204295, "epoch": 2295, "lr": 1.4406274439385609e-05}
{"train_loss": 0.05483941733837128, "global_step": 204296, "epoch": 2295, "lr": 1.4405867288885904e-05}
{"train_loss": 0.06756868958473206, "global_step": 204297, "epoch": 2295, "lr": 1.4405460143171413e-05}
{"train_loss": 0.05990622192621231, "global_step": 204298, "epoch": 2295, "lr": 1.4405053002242152e-05}
{"train_loss": 0.03518814221024513, "global_step": 204299, "epoch": 2295, "lr": 1.4404645866098221e-05}
{"train_loss": 0.06223224103450775, "global_step": 204300, "epoch": 2295, "lr": 1.4404238734739629e-05}
{"train_loss": 0.037996795028448105, "global_step": 204301, "epoch": 2295, "lr": 1.4403831608166474e-05}
{"train_loss": 0.07462409883737564, "global_step": 204302, "epoch": 2295, "lr": 1.4403424486378769e-05}
{"train_loss": 0.0812801867723465, "global_step": 204303, "epoch": 2295, "lr": 1.4403017369376598e-05}
{"train_loss": 0.09890098124742508, "global_step": 204304, "epoch": 2295, "lr": 1.4402610257160015e-05}
{"train_loss": 0.07324358820915222, "global_step": 204305, "epoch": 2295, "lr": 1.4402203149729054e-05}
{"train_loss": 0.04646329954266548, "global_step": 204306, "epoch": 2295, "lr": 1.44017960470838e-05}
{"train_loss": 0.03562154993414879, "global_step": 204307, "epoch": 2295, "lr": 1.4401388949224275e-05}
{"train_loss": 0.11639343202114105, "global_step": 204308, "epoch": 2295, "lr": 1.4400981856150564e-05}
{"train_loss": 0.04845067486166954, "global_step": 204309, "epoch": 2295, "lr": 1.440057476786269e-05}
{"train_loss": 0.08828076720237732, "global_step": 204310, "epoch": 2295, "lr": 1.4400167684360744e-05}
{"train_loss": 0.06591000407934189, "global_step": 204311, "epoch": 2295, "lr": 1.4399760605644747e-05}
{"train_loss": 0.062286049127578735, "global_step": 204312, "epoch": 2295, "lr": 1.4399353531714781e-05}
{"train_loss": 0.031998373568058014, "global_step": 204313, "epoch": 2295, "lr": 1.4398946462570872e-05}
{"train_loss": 0.12375637143850327, "global_step": 204314, "epoch": 2295, "lr": 1.4398539398213107e-05}
{"train_loss": 0.02521931566298008, "global_step": 204315, "epoch": 2295, "lr": 1.4398132338641507e-05}
{"train_loss": 0.09755008667707443, "global_step": 204316, "epoch": 2295, "lr": 1.4397725283856161e-05}
{"train_loss": 0.11190028488636017, "global_step": 204317, "epoch": 2295, "lr": 1.4397318233857094e-05}
{"train_loss": 0.05565384775400162, "global_step": 204318, "epoch": 2295, "lr": 1.4396911188644386e-05}
{"train_loss": 0.060926321893930435, "global_step": 204319, "epoch": 2295, "lr": 1.4396504148218065e-05}
{"train_loss": 0.06659816950559616, "global_step": 204320, "epoch": 2295, "lr": 1.4396097112578217e-05}
{"train_loss": 0.06669756025075912, "global_step": 204321, "epoch": 2295, "lr": 1.439569008172486e-05}
{"train_loss": 0.05077981576323509, "global_step": 204322, "epoch": 2295, "lr": 1.4395283055658087e-05}
{"train_loss": 0.03927040845155716, "global_step": 204323, "epoch": 2295, "lr": 1.4394876034377936e-05}
{"train_loss": 0.06862499564886093, "global_step": 204324, "epoch": 2295, "lr": 1.4394469017884438e-05}
{"train_loss": 0.021647684276103973, "global_step": 204325, "epoch": 2295, "lr": 1.4394062006177694e-05}
{"train_loss": 0.0896630585193634, "global_step": 204326, "epoch": 2295, "lr": 1.439365499925771e-05}
{"train_loss": 0.03926023840904236, "global_step": 204327, "epoch": 2295, "lr": 1.4393247997124587e-05}
{"train_loss": 0.06140945106744766, "global_step": 204328, "epoch": 2295, "lr": 1.439284099977834e-05}
{"train_loss": 0.04244398698210716, "global_step": 204329, "epoch": 2295, "lr": 1.4392434007219058e-05}
{"train_loss": 0.0826350674033165, "global_step": 204330, "epoch": 2295, "lr": 1.4392027019446763e-05}
{"train_loss": 0.103010393679142, "global_step": 204331, "epoch": 2295, "lr": 1.4391620036461539e-05}
{"train_loss": 0.09609133750200272, "global_step": 204332, "epoch": 2295, "lr": 1.439121305826342e-05}
{"train_loss": 0.02335154078900814, "global_step": 204333, "epoch": 2295, "lr": 1.4390806084852477e-05}
{"train_loss": 0.03460715711116791, "global_step": 204334, "epoch": 2295, "lr": 1.4390399116228742e-05}
{"train_loss": 0.10165518522262573, "global_step": 204335, "epoch": 2295, "lr": 1.43899921523923e-05}
{"train_loss": 0.08797766268253326, "global_step": 204336, "epoch": 2295, "lr": 1.4389585193343174e-05}
{"train_loss": 0.050034623593091965, "global_step": 204337, "epoch": 2295, "lr": 1.4389178239081447e-05}
{"train_loss": 0.06847217679023743, "global_step": 204338, "epoch": 2295, "lr": 1.438877128960715e-05}
{"train_loss": 0.029670080170035362, "global_step": 204339, "epoch": 2295, "lr": 1.4388364344920347e-05}
{"train_loss": 0.028234120458364487, "global_step": 204340, "epoch": 2295, "lr": 1.4387957405021107e-05}
{"train_loss": 0.06002258136868477, "global_step": 204341, "epoch": 2295, "lr": 1.4387550469909462e-05}
{"train_loss": 0.08741723001003265, "global_step": 204342, "epoch": 2295, "lr": 1.4387143539585484e-05}
{"train_loss": 0.06459543584019281, "global_step": 204343, "epoch": 2295, "lr": 1.4386736614049212e-05, "val_loss": 8.337514877319336, "train_action_mse_error": 27.86821174621582}
{"train_loss": 0.05825895816087723, "global_step": 204344, "epoch": 2296, "lr": 1.4386329693300727e-05}
{"train_loss": 0.04261918365955353, "global_step": 204345, "epoch": 2296, "lr": 1.4385922777340044e-05}
{"train_loss": 0.024758661165833473, "global_step": 204346, "epoch": 2296, "lr": 1.4385515866167255e-05}
{"train_loss": 0.04682891070842743, "global_step": 204347, "epoch": 2296, "lr": 1.438510895978239e-05}
{"train_loss": 0.09844643622636795, "global_step": 204348, "epoch": 2296, "lr": 1.4384702058185522e-05}
{"train_loss": 0.08351968228816986, "global_step": 204349, "epoch": 2296, "lr": 1.4384295161376682e-05}
{"train_loss": 0.08524004369974136, "global_step": 204350, "epoch": 2296, "lr": 1.438388826935596e-05}
{"train_loss": 0.08163699507713318, "global_step": 204351, "epoch": 2296, "lr": 1.4383481382123371e-05}
{"train_loss": 0.07954694330692291, "global_step": 204352, "epoch": 2296, "lr": 1.4383074499679007e-05}
{"train_loss": 0.08442958444356918, "global_step": 204353, "epoch": 2296, "lr": 1.4382667622022883e-05}
{"train_loss": 0.03993236646056175, "global_step": 204354, "epoch": 2296, "lr": 1.4382260749155097e-05}
{"train_loss": 0.0640859603881836, "global_step": 204355, "epoch": 2296, "lr": 1.4381853881075662e-05}
{"train_loss": 0.08163793385028839, "global_step": 204356, "epoch": 2296, "lr": 1.4381447017784656e-05}
{"train_loss": 0.05517244711518288, "global_step": 204357, "epoch": 2296, "lr": 1.4381040159282145e-05}
{"train_loss": 0.04841002821922302, "global_step": 204358, "epoch": 2296, "lr": 1.438063330556817e-05}
{"train_loss": 0.057530563324689865, "global_step": 204359, "epoch": 2296, "lr": 1.4380226456642764e-05}
{"train_loss": 0.07792574167251587, "global_step": 204360, "epoch": 2296, "lr": 1.4379819612506024e-05}
{"train_loss": 0.07610352337360382, "global_step": 204361, "epoch": 2296, "lr": 1.4379412773157964e-05}
{"train_loss": 0.025018831714987755, "global_step": 204362, "epoch": 2296, "lr": 1.4379005938598672e-05}
{"train_loss": 0.050778746604919434, "global_step": 204363, "epoch": 2296, "lr": 1.4378599108828195e-05}
{"train_loss": 0.07524912804365158, "global_step": 204364, "epoch": 2296, "lr": 1.4378192283846559e-05}
{"train_loss": 0.07185181975364685, "global_step": 204365, "epoch": 2296, "lr": 1.4377785463653859e-05}
{"train_loss": 0.1405368447303772, "global_step": 204366, "epoch": 2296, "lr": 1.4377378648250117e-05}
{"train_loss": 0.1163405030965805, "global_step": 204367, "epoch": 2296, "lr": 1.4376971837635417e-05}
{"train_loss": 0.10485520958900452, "global_step": 204368, "epoch": 2296, "lr": 1.4376565031809785e-05}
{"train_loss": 0.05858849734067917, "global_step": 204369, "epoch": 2296, "lr": 1.4376158230773301e-05}
{"train_loss": 0.1175355389714241, "global_step": 204370, "epoch": 2296, "lr": 1.4375751434525998e-05}
{"train_loss": 0.10551799088716507, "global_step": 204371, "epoch": 2296, "lr": 1.437534464306795e-05}
{"train_loss": 0.09540379047393799, "global_step": 204372, "epoch": 2296, "lr": 1.4374937856399195e-05}
{"train_loss": 0.05851441249251366, "global_step": 204373, "epoch": 2296, "lr": 1.4374531074519803e-05}
{"train_loss": 0.08536404371261597, "global_step": 204374, "epoch": 2296, "lr": 1.4374124297429808e-05}
{"train_loss": 0.0890984907746315, "global_step": 204375, "epoch": 2296, "lr": 1.4373717525129283e-05}
{"train_loss": 0.030362820252776146, "global_step": 204376, "epoch": 2296, "lr": 1.4373310757618292e-05}
{"train_loss": 0.03879111632704735, "global_step": 204377, "epoch": 2296, "lr": 1.4372903994896852e-05}
{"train_loss": 0.0806078091263771, "global_step": 204378, "epoch": 2296, "lr": 1.437249723696506e-05}
{"train_loss": 0.02824597805738449, "global_step": 204379, "epoch": 2296, "lr": 1.4372090483822943e-05}
{"train_loss": 0.09014071524143219, "global_step": 204380, "epoch": 2296, "lr": 1.437168373547057e-05}
{"train_loss": 0.05583634972572327, "global_step": 204381, "epoch": 2296, "lr": 1.4371276991907979e-05}
{"train_loss": 0.02447214163839817, "global_step": 204382, "epoch": 2296, "lr": 1.4370870253135249e-05}
{"train_loss": 0.06680970638990402, "global_step": 204383, "epoch": 2296, "lr": 1.437046351915241e-05}
{"train_loss": 0.041345465928316116, "global_step": 204384, "epoch": 2296, "lr": 1.437005678995954e-05}
{"train_loss": 0.021985750645399094, "global_step": 204385, "epoch": 2296, "lr": 1.4369650065556666e-05}
{"train_loss": 0.039124250411987305, "global_step": 204386, "epoch": 2296, "lr": 1.4369243345943872e-05}
{"train_loss": 0.053931884467601776, "global_step": 204387, "epoch": 2296, "lr": 1.4368836631121185e-05}
{"train_loss": 0.05961687117815018, "global_step": 204388, "epoch": 2296, "lr": 1.4368429921088694e-05}
{"train_loss": 0.04927023500204086, "global_step": 204389, "epoch": 2296, "lr": 1.436802321584641e-05}
{"train_loss": 0.06758323311805725, "global_step": 204390, "epoch": 2296, "lr": 1.4367616515394434e-05}
{"train_loss": 0.0606231763958931, "global_step": 204391, "epoch": 2296, "lr": 1.4367209819732774e-05}
{"train_loss": 0.03752371296286583, "global_step": 204392, "epoch": 2296, "lr": 1.4366803128861517e-05}
{"train_loss": 0.09366047382354736, "global_step": 204393, "epoch": 2296, "lr": 1.4366396442780716e-05}
{"train_loss": 0.04306681826710701, "global_step": 204394, "epoch": 2296, "lr": 1.4365989761490412e-05}
{"train_loss": 0.0437416210770607, "global_step": 204395, "epoch": 2296, "lr": 1.4365583084990675e-05}
{"train_loss": 0.04776475578546524, "global_step": 204396, "epoch": 2296, "lr": 1.4365176413281556e-05}
{"train_loss": 0.07362854480743408, "global_step": 204397, "epoch": 2296, "lr": 1.4364769746363082e-05}
{"train_loss": 0.045377474278211594, "global_step": 204398, "epoch": 2296, "lr": 1.4364363084235354e-05}
{"train_loss": 0.0465204194188118, "global_step": 204399, "epoch": 2296, "lr": 1.4363956426898384e-05}
{"train_loss": 0.03457879275083542, "global_step": 204400, "epoch": 2296, "lr": 1.4363549774352258e-05}
{"train_loss": 0.08713341504335403, "global_step": 204401, "epoch": 2296, "lr": 1.4363143126597023e-05}
{"train_loss": 0.07312198728322983, "global_step": 204402, "epoch": 2296, "lr": 1.436273648363271e-05}
{"train_loss": 0.09251490235328674, "global_step": 204403, "epoch": 2296, "lr": 1.436232984545941e-05}
{"train_loss": 0.03151608631014824, "global_step": 204404, "epoch": 2296, "lr": 1.436192321207715e-05}
{"train_loss": 0.07826051861047745, "global_step": 204405, "epoch": 2296, "lr": 1.4361516583486001e-05}
{"train_loss": 0.08634321391582489, "global_step": 204406, "epoch": 2296, "lr": 1.4361109959686003e-05}
{"train_loss": 0.03813226893544197, "global_step": 204407, "epoch": 2296, "lr": 1.4360703340677234e-05}
{"train_loss": 0.07198340445756912, "global_step": 204408, "epoch": 2296, "lr": 1.4360296726459716e-05}
{"train_loss": 0.05429914593696594, "global_step": 204409, "epoch": 2296, "lr": 1.4359890117033542e-05}
{"train_loss": 0.09530562162399292, "global_step": 204410, "epoch": 2296, "lr": 1.4359483512398724e-05}
{"train_loss": 0.06469137966632843, "global_step": 204411, "epoch": 2296, "lr": 1.4359076912555342e-05}
{"train_loss": 0.05098137632012367, "global_step": 204412, "epoch": 2296, "lr": 1.4358670317503465e-05}
{"train_loss": 0.07023601979017258, "global_step": 204413, "epoch": 2296, "lr": 1.4358263727243115e-05}
{"train_loss": 0.10461319237947464, "global_step": 204414, "epoch": 2296, "lr": 1.4357857141774372e-05}
{"train_loss": 0.13439682126045227, "global_step": 204415, "epoch": 2296, "lr": 1.4357450561097275e-05}
{"train_loss": 0.0921933576464653, "global_step": 204416, "epoch": 2296, "lr": 1.4357043985211893e-05}
{"train_loss": 0.07065248489379883, "global_step": 204417, "epoch": 2296, "lr": 1.4356637414118257e-05}
{"train_loss": 0.07757911086082458, "global_step": 204418, "epoch": 2296, "lr": 1.4356230847816455e-05}
{"train_loss": 0.1138727143406868, "global_step": 204419, "epoch": 2296, "lr": 1.4355824286306502e-05}
{"train_loss": 0.04522738605737686, "global_step": 204420, "epoch": 2296, "lr": 1.4355417729588493e-05}
{"train_loss": 0.0585470050573349, "global_step": 204421, "epoch": 2296, "lr": 1.4355011177662448e-05}
{"train_loss": 0.02668537199497223, "global_step": 204422, "epoch": 2296, "lr": 1.4354604630528457e-05}
{"train_loss": 0.04603783041238785, "global_step": 204423, "epoch": 2296, "lr": 1.4354198088186533e-05}
{"train_loss": 0.06777207553386688, "global_step": 204424, "epoch": 2296, "lr": 1.435379155063677e-05}
{"train_loss": 0.054412707686424255, "global_step": 204425, "epoch": 2296, "lr": 1.435338501787919e-05}
{"train_loss": 0.07362373918294907, "global_step": 204426, "epoch": 2296, "lr": 1.4352978489913881e-05}
{"train_loss": 0.04645319655537605, "global_step": 204427, "epoch": 2296, "lr": 1.4352571966740864e-05}
{"train_loss": 0.0603947639465332, "global_step": 204428, "epoch": 2296, "lr": 1.4352165448360222e-05}
{"train_loss": 0.026891520246863365, "global_step": 204429, "epoch": 2296, "lr": 1.435175893477198e-05}
{"train_loss": 0.06302998214960098, "global_step": 204430, "epoch": 2296, "lr": 1.4351352425976211e-05}
{"train_loss": 0.08263999968767166, "global_step": 204431, "epoch": 2296, "lr": 1.4350945921972986e-05}
{"train_loss": 0.06568145088516594, "global_step": 204432, "epoch": 2296, "lr": 1.4350539422762322e-05, "val_loss": 8.343477249145508}
{"train_loss": 0.09370768815279007, "global_step": 204433, "epoch": 2297, "lr": 1.4350132928344311e-05}
{"train_loss": 0.03630812466144562, "global_step": 204434, "epoch": 2297, "lr": 1.434972643871899e-05}
{"train_loss": 0.06145143508911133, "global_step": 204435, "epoch": 2297, "lr": 1.4349319953886403e-05}
{"train_loss": 0.07587406039237976, "global_step": 204436, "epoch": 2297, "lr": 1.4348913473846626e-05}
{"train_loss": 0.07537481188774109, "global_step": 204437, "epoch": 2297, "lr": 1.4348506998599686e-05}
{"train_loss": 0.07817043364048004, "global_step": 204438, "epoch": 2297, "lr": 1.4348100528145675e-05}
{"train_loss": 0.03447406366467476, "global_step": 204439, "epoch": 2297, "lr": 1.4347694062484623e-05}
{"train_loss": 0.05303821712732315, "global_step": 204440, "epoch": 2297, "lr": 1.4347287601616577e-05}
{"train_loss": 0.14072056114673615, "global_step": 204441, "epoch": 2297, "lr": 1.4346881145541619e-05}
{"train_loss": 0.07676500082015991, "global_step": 204442, "epoch": 2297, "lr": 1.434647469425977e-05}
{"train_loss": 0.04158179089426994, "global_step": 204443, "epoch": 2297, "lr": 1.4346068247771115e-05}
{"train_loss": 0.09067424386739731, "global_step": 204444, "epoch": 2297, "lr": 1.434566180607569e-05}
{"train_loss": 0.08640973269939423, "global_step": 204445, "epoch": 2297, "lr": 1.4345255369173565e-05}
{"train_loss": 0.05922560766339302, "global_step": 204446, "epoch": 2297, "lr": 1.4344848937064775e-05}
{"train_loss": 0.06234915181994438, "global_step": 204447, "epoch": 2297, "lr": 1.4344442509749384e-05}
{"train_loss": 0.047756824642419815, "global_step": 204448, "epoch": 2297, "lr": 1.4344036087227463e-05}
{"train_loss": 0.05115920677781105, "global_step": 204449, "epoch": 2297, "lr": 1.4343629669499036e-05}
{"train_loss": 0.04217476025223732, "global_step": 204450, "epoch": 2297, "lr": 1.4343223256564187e-05}
{"train_loss": 0.054407693445682526, "global_step": 204451, "epoch": 2297, "lr": 1.4342816848422947e-05}
{"train_loss": 0.048043180257081985, "global_step": 204452, "epoch": 2297, "lr": 1.4342410445075387e-05}
{"train_loss": 0.0903673991560936, "global_step": 204453, "epoch": 2297, "lr": 1.4342004046521552e-05}
{"train_loss": 0.06736019998788834, "global_step": 204454, "epoch": 2297, "lr": 1.4341597652761508e-05}
{"train_loss": 0.021970365196466446, "global_step": 204455, "epoch": 2297, "lr": 1.4341191263795284e-05}
{"train_loss": 0.0703732967376709, "global_step": 204456, "epoch": 2297, "lr": 1.434078487962297e-05}
{"train_loss": 0.11716824769973755, "global_step": 204457, "epoch": 2297, "lr": 1.434037850024459e-05}
{"train_loss": 0.03360534831881523, "global_step": 204458, "epoch": 2297, "lr": 1.4339972125660228e-05}
{"train_loss": 0.05310969799757004, "global_step": 204459, "epoch": 2297, "lr": 1.43395657558699e-05}
{"train_loss": 0.027556974440813065, "global_step": 204460, "epoch": 2297, "lr": 1.4339159390873702e-05}
{"train_loss": 0.08578114211559296, "global_step": 204461, "epoch": 2297, "lr": 1.4338753030671653e-05}
{"train_loss": 0.04803670197725296, "global_step": 204462, "epoch": 2297, "lr": 1.4338346675263841e-05}
{"train_loss": 0.057739563286304474, "global_step": 204463, "epoch": 2297, "lr": 1.433794032465029e-05}
{"train_loss": 0.04626912251114845, "global_step": 204464, "epoch": 2297, "lr": 1.433753397883108e-05}
{"train_loss": 0.02381077967584133, "global_step": 204465, "epoch": 2297, "lr": 1.4337127637806242e-05}
{"train_loss": 0.05001990124583244, "global_step": 204466, "epoch": 2297, "lr": 1.4336721301575839e-05}
{"train_loss": 0.07300201058387756, "global_step": 204467, "epoch": 2297, "lr": 1.4336314970139947e-05}
{"train_loss": 0.06161430850625038, "global_step": 204468, "epoch": 2297, "lr": 1.4335908643498585e-05}
{"train_loss": 0.06758082658052444, "global_step": 204469, "epoch": 2297, "lr": 1.433550232165184e-05}
{"train_loss": 0.046603698283433914, "global_step": 204470, "epoch": 2297, "lr": 1.4335096004599741e-05}
{"train_loss": 0.08017482608556747, "global_step": 204471, "epoch": 2297, "lr": 1.4334689692342368e-05}
{"train_loss": 0.09222165495157242, "global_step": 204472, "epoch": 2297, "lr": 1.4334283384879743e-05}
{"train_loss": 0.04600798338651657, "global_step": 204473, "epoch": 2297, "lr": 1.4333877082211955e-05}
{"train_loss": 0.10836965590715408, "global_step": 204474, "epoch": 2297, "lr": 1.4333470784339042e-05}
{"train_loss": 0.08975106477737427, "global_step": 204475, "epoch": 2297, "lr": 1.4333064491261044e-05}
{"train_loss": 0.053655244410037994, "global_step": 204476, "epoch": 2297, "lr": 1.433265820297805e-05}
{"train_loss": 0.06574466824531555, "global_step": 204477, "epoch": 2297, "lr": 1.4332251919490074e-05}
{"train_loss": 0.04253091663122177, "global_step": 204478, "epoch": 2297, "lr": 1.4331845640797215e-05}
{"train_loss": 0.06553443521261215, "global_step": 204479, "epoch": 2297, "lr": 1.4331439366899502e-05}
{"train_loss": 0.06223992630839348, "global_step": 204480, "epoch": 2297, "lr": 1.433103309779697e-05}
{"train_loss": 0.07321435958147049, "global_step": 204481, "epoch": 2297, "lr": 1.4330626833489718e-05}
{"train_loss": 0.04234147071838379, "global_step": 204482, "epoch": 2297, "lr": 1.4330220573977765e-05}
{"train_loss": 0.14747604727745056, "global_step": 204483, "epoch": 2297, "lr": 1.432981431926117e-05}
{"train_loss": 0.038170669227838516, "global_step": 204484, "epoch": 2297, "lr": 1.4329408069340022e-05}
{"train_loss": 0.0748339295387268, "global_step": 204485, "epoch": 2297, "lr": 1.4329001824214327e-05}
{"train_loss": 0.07608731091022491, "global_step": 204486, "epoch": 2297, "lr": 1.4328595583884186e-05}
{"train_loss": 0.08882622420787811, "global_step": 204487, "epoch": 2297, "lr": 1.4328189348349607e-05}
{"train_loss": 0.11184849590063095, "global_step": 204488, "epoch": 2297, "lr": 1.4327783117610688e-05}
{"train_loss": 0.061601702123880386, "global_step": 204489, "epoch": 2297, "lr": 1.4327376891667449e-05}
{"train_loss": 0.027625856921076775, "global_step": 204490, "epoch": 2297, "lr": 1.4326970670519973e-05}
{"train_loss": 0.10650404542684555, "global_step": 204491, "epoch": 2297, "lr": 1.432656445416829e-05}
{"train_loss": 0.11211255937814713, "global_step": 204492, "epoch": 2297, "lr": 1.4326158242612474e-05}
{"train_loss": 0.07170998305082321, "global_step": 204493, "epoch": 2297, "lr": 1.4325752035852558e-05}
{"train_loss": 0.03378397598862648, "global_step": 204494, "epoch": 2297, "lr": 1.4325345833888625e-05}
{"train_loss": 0.03357308357954025, "global_step": 204495, "epoch": 2297, "lr": 1.4324939636720701e-05}
{"train_loss": 0.08438433706760406, "global_step": 204496, "epoch": 2297, "lr": 1.432453344434887e-05}
{"train_loss": 0.10524197667837143, "global_step": 204497, "epoch": 2297, "lr": 1.4324127256773157e-05}
{"train_loss": 0.05830935388803482, "global_step": 204498, "epoch": 2297, "lr": 1.432372107399364e-05}
{"train_loss": 0.06299211829900742, "global_step": 204499, "epoch": 2297, "lr": 1.4323314896010353e-05}
{"train_loss": 0.05036046728491783, "global_step": 204500, "epoch": 2297, "lr": 1.4322908722823375e-05}
{"train_loss": 0.0456916019320488, "global_step": 204501, "epoch": 2297, "lr": 1.4322502554432732e-05}
{"train_loss": 0.0673273503780365, "global_step": 204502, "epoch": 2297, "lr": 1.4322096390838497e-05}
{"train_loss": 0.07642045617103577, "global_step": 204503, "epoch": 2297, "lr": 1.432169023204073e-05}
{"train_loss": 0.04870361462235451, "global_step": 204504, "epoch": 2297, "lr": 1.4321284078039465e-05}
{"train_loss": 0.029315732419490814, "global_step": 204505, "epoch": 2297, "lr": 1.4320877928834786e-05}
{"train_loss": 0.06949802488088608, "global_step": 204506, "epoch": 2297, "lr": 1.432047178442671e-05}
{"train_loss": 0.09145432710647583, "global_step": 204507, "epoch": 2297, "lr": 1.4320065644815328e-05}
{"train_loss": 0.07105528563261032, "global_step": 204508, "epoch": 2297, "lr": 1.4319659510000666e-05}
{"train_loss": 0.06347155570983887, "global_step": 204509, "epoch": 2297, "lr": 1.4319253379982805e-05}
{"train_loss": 0.06537783145904541, "global_step": 204510, "epoch": 2297, "lr": 1.4318847254761769e-05}
{"train_loss": 0.07410812377929688, "global_step": 204511, "epoch": 2297, "lr": 1.4318441134337645e-05}
{"train_loss": 0.0948534831404686, "global_step": 204512, "epoch": 2297, "lr": 1.4318035018710473e-05}
{"train_loss": 0.11566348373889923, "global_step": 204513, "epoch": 2297, "lr": 1.4317628907880287e-05}
{"train_loss": 0.046758994460105896, "global_step": 204514, "epoch": 2297, "lr": 1.431722280184718e-05}
{"train_loss": 0.0679023340344429, "global_step": 204515, "epoch": 2297, "lr": 1.4316816700611175e-05}
{"train_loss": 0.07717366516590118, "global_step": 204516, "epoch": 2297, "lr": 1.4316410604172348e-05}
{"train_loss": 0.07538609951734543, "global_step": 204517, "epoch": 2297, "lr": 1.4316004512530745e-05}
{"train_loss": 0.028638826683163643, "global_step": 204518, "epoch": 2297, "lr": 1.4315598425686411e-05}
{"train_loss": 0.06543996930122375, "global_step": 204519, "epoch": 2297, "lr": 1.4315192343639406e-05}
{"train_loss": 0.07525186240673065, "global_step": 204520, "epoch": 2297, "lr": 1.43147862663898e-05}
{"train_loss": 0.06685961475281903, "global_step": 204521, "epoch": 2297, "lr": 1.4314380193937626e-05, "val_loss": 8.26148509979248}
{"train_loss": 0.04146004095673561, "global_step": 204522, "epoch": 2298, "lr": 1.4313974126282964e-05}
{"train_loss": 0.08536074310541153, "global_step": 204523, "epoch": 2298, "lr": 1.4313568063425836e-05}
{"train_loss": 0.11134045571088791, "global_step": 204524, "epoch": 2298, "lr": 1.4313162005366326e-05}
{"train_loss": 0.04213644564151764, "global_step": 204525, "epoch": 2298, "lr": 1.4312755952104467e-05}
{"train_loss": 0.05162997543811798, "global_step": 204526, "epoch": 2298, "lr": 1.4312349903640338e-05}
{"train_loss": 0.08017174154520035, "global_step": 204527, "epoch": 2298, "lr": 1.431194385997396e-05}
{"train_loss": 0.034848082810640335, "global_step": 204528, "epoch": 2298, "lr": 1.431153782110542e-05}
{"train_loss": 0.0439845472574234, "global_step": 204529, "epoch": 2298, "lr": 1.4311131787034748e-05}
{"train_loss": 0.05986235290765762, "global_step": 204530, "epoch": 2298, "lr": 1.4310725757762022e-05}
{"train_loss": 0.07075954228639603, "global_step": 204531, "epoch": 2298, "lr": 1.4310319733287269e-05}
{"train_loss": 0.05821552872657776, "global_step": 204532, "epoch": 2298, "lr": 1.4309913713610573e-05}
{"train_loss": 0.09556233882904053, "global_step": 204533, "epoch": 2298, "lr": 1.430950769873196e-05}
{"train_loss": 0.056060418486595154, "global_step": 204534, "epoch": 2298, "lr": 1.4309101688651516e-05}
{"train_loss": 0.0486912801861763, "global_step": 204535, "epoch": 2298, "lr": 1.430869568336926e-05}
{"train_loss": 0.04725445434451103, "global_step": 204536, "epoch": 2298, "lr": 1.4308289682885284e-05}
{"train_loss": 0.07029670476913452, "global_step": 204537, "epoch": 2298, "lr": 1.4307883687199603e-05}
{"train_loss": 0.07325290888547897, "global_step": 204538, "epoch": 2298, "lr": 1.4307477696312293e-05}
{"train_loss": 0.06547959893941879, "global_step": 204539, "epoch": 2298, "lr": 1.4307071710223424e-05}
{"train_loss": 0.06263095885515213, "global_step": 204540, "epoch": 2298, "lr": 1.4306665728933021e-05}
{"train_loss": 0.04995978623628616, "global_step": 204541, "epoch": 2298, "lr": 1.4306259752441165e-05}
{"train_loss": 0.09121426939964294, "global_step": 204542, "epoch": 2298, "lr": 1.4305853780747885e-05}
{"train_loss": 0.043411560356616974, "global_step": 204543, "epoch": 2298, "lr": 1.4305447813853256e-05}
{"train_loss": 0.0785600021481514, "global_step": 204544, "epoch": 2298, "lr": 1.4305041851757317e-05}
{"train_loss": 0.08121347427368164, "global_step": 204545, "epoch": 2298, "lr": 1.4304635894460139e-05}
{"train_loss": 0.06517146527767181, "global_step": 204546, "epoch": 2298, "lr": 1.4304229941961755e-05}
{"train_loss": 0.08083304017782211, "global_step": 204547, "epoch": 2298, "lr": 1.430382399426225e-05}
{"train_loss": 0.02030208706855774, "global_step": 204548, "epoch": 2298, "lr": 1.4303418051361644e-05}
{"train_loss": 0.06067672744393349, "global_step": 204549, "epoch": 2298, "lr": 1.4303012113260022e-05}
{"train_loss": 0.06195615604519844, "global_step": 204550, "epoch": 2298, "lr": 1.4302606179957429e-05}
{"train_loss": 0.06326643377542496, "global_step": 204551, "epoch": 2298, "lr": 1.4302200251453896e-05}
{"train_loss": 0.041825953871011734, "global_step": 204552, "epoch": 2298, "lr": 1.4301794327749512e-05}
{"train_loss": 0.10097217559814453, "global_step": 204553, "epoch": 2298, "lr": 1.4301388408844302e-05}
{"train_loss": 0.01352923084050417, "global_step": 204554, "epoch": 2298, "lr": 1.4300982494738351e-05}
{"train_loss": 0.10522521287202835, "global_step": 204555, "epoch": 2298, "lr": 1.4300576585431696e-05}
{"train_loss": 0.07394136488437653, "global_step": 204556, "epoch": 2298, "lr": 1.4300170680924386e-05}
{"train_loss": 0.04500504210591316, "global_step": 204557, "epoch": 2298, "lr": 1.4299764781216474e-05}
{"train_loss": 0.09508343040943146, "global_step": 204558, "epoch": 2298, "lr": 1.429935888630804e-05}
{"train_loss": 0.045967645943164825, "global_step": 204559, "epoch": 2298, "lr": 1.429895299619911e-05}
{"train_loss": 0.038496602326631546, "global_step": 204560, "epoch": 2298, "lr": 1.4298547110889764e-05}
{"train_loss": 0.08397519588470459, "global_step": 204561, "epoch": 2298, "lr": 1.429814123038003e-05}
{"train_loss": 0.11457348614931107, "global_step": 204562, "epoch": 2298, "lr": 1.4297735354669989e-05}
{"train_loss": 0.05304602533578873, "global_step": 204563, "epoch": 2298, "lr": 1.429732948375967e-05}
{"train_loss": 0.05071088299155235, "global_step": 204564, "epoch": 2298, "lr": 1.4296923617649149e-05}
{"train_loss": 0.0699876993894577, "global_step": 204565, "epoch": 2298, "lr": 1.4296517756338462e-05}
{"train_loss": 0.05186860263347626, "global_step": 204566, "epoch": 2298, "lr": 1.4296111899827686e-05}
{"train_loss": 0.08458125591278076, "global_step": 204567, "epoch": 2298, "lr": 1.4295706048116847e-05}
{"train_loss": 0.06736735254526138, "global_step": 204568, "epoch": 2298, "lr": 1.429530020120603e-05}
{"train_loss": 0.04770168289542198, "global_step": 204569, "epoch": 2298, "lr": 1.4294894359095262e-05}
{"train_loss": 0.04740867018699646, "global_step": 204570, "epoch": 2298, "lr": 1.4294488521784622e-05}
{"train_loss": 0.0790279284119606, "global_step": 204571, "epoch": 2298, "lr": 1.4294082689274142e-05}
{"train_loss": 0.058086082339286804, "global_step": 204572, "epoch": 2298, "lr": 1.4293676861563898e-05}
{"train_loss": 0.021359959617257118, "global_step": 204573, "epoch": 2298, "lr": 1.4293271038653921e-05}
{"train_loss": 0.04290709272027016, "global_step": 204574, "epoch": 2298, "lr": 1.4292865220544277e-05}
{"train_loss": 0.05108045041561127, "global_step": 204575, "epoch": 2298, "lr": 1.4292459407235042e-05}
{"train_loss": 0.036315735429525375, "global_step": 204576, "epoch": 2298, "lr": 1.429205359872623e-05}
{"train_loss": 0.10775120556354523, "global_step": 204577, "epoch": 2298, "lr": 1.4291647795017937e-05}
{"train_loss": 0.05898822471499443, "global_step": 204578, "epoch": 2298, "lr": 1.4291241996110177e-05}
{"train_loss": 0.07449369877576828, "global_step": 204579, "epoch": 2298, "lr": 1.4290836202003039e-05}
{"train_loss": 0.042418017983436584, "global_step": 204580, "epoch": 2298, "lr": 1.4290430412696548e-05}
{"train_loss": 0.05719975754618645, "global_step": 204581, "epoch": 2298, "lr": 1.429002462819079e-05}
{"train_loss": 0.04871055483818054, "global_step": 204582, "epoch": 2298, "lr": 1.4289618848485787e-05}
{"train_loss": 0.0850159302353859, "global_step": 204583, "epoch": 2298, "lr": 1.4289213073581626e-05}
{"train_loss": 0.040460314601659775, "global_step": 204584, "epoch": 2298, "lr": 1.4288807303478329e-05}
{"train_loss": 0.0955086275935173, "global_step": 204585, "epoch": 2298, "lr": 1.4288401538175983e-05}
{"train_loss": 0.04889867827296257, "global_step": 204586, "epoch": 2298, "lr": 1.4287995777674607e-05}
{"train_loss": 0.050996750593185425, "global_step": 204587, "epoch": 2298, "lr": 1.4287590021974296e-05}
{"train_loss": 0.05059683322906494, "global_step": 204588, "epoch": 2298, "lr": 1.4287184271075065e-05}
{"train_loss": 0.06517208367586136, "global_step": 204589, "epoch": 2298, "lr": 1.4286778524977e-05}
{"train_loss": 0.0821971446275711, "global_step": 204590, "epoch": 2298, "lr": 1.4286372783680146e-05}
{"train_loss": 0.11111553013324738, "global_step": 204591, "epoch": 2298, "lr": 1.4285967047184534e-05}
{"train_loss": 0.06354072690010071, "global_step": 204592, "epoch": 2298, "lr": 1.428556131549026e-05}
{"train_loss": 0.08540858328342438, "global_step": 204593, "epoch": 2298, "lr": 1.4285155588597337e-05}
{"train_loss": 0.07988572865724564, "global_step": 204594, "epoch": 2298, "lr": 1.4284749866505853e-05}
{"train_loss": 0.0552738793194294, "global_step": 204595, "epoch": 2298, "lr": 1.428434414921584e-05}
{"train_loss": 0.05785214528441429, "global_step": 204596, "epoch": 2298, "lr": 1.4283938436727373e-05}
{"train_loss": 0.047829605638980865, "global_step": 204597, "epoch": 2298, "lr": 1.4283532729040477e-05}
{"train_loss": 0.043433092534542084, "global_step": 204598, "epoch": 2298, "lr": 1.4283127026155247e-05}
{"train_loss": 0.027438770979642868, "global_step": 204599, "epoch": 2298, "lr": 1.4282721328071697e-05}
{"train_loss": 0.027162082493305206, "global_step": 204600, "epoch": 2298, "lr": 1.4282315634789911e-05}
{"train_loss": 0.028078559786081314, "global_step": 204601, "epoch": 2298, "lr": 1.4281909946309924e-05}
{"train_loss": 0.08459918200969696, "global_step": 204602, "epoch": 2298, "lr": 1.4281504262631806e-05}
{"train_loss": 0.09042314440011978, "global_step": 204603, "epoch": 2298, "lr": 1.4281098583755597e-05}
{"train_loss": 0.035851046442985535, "global_step": 204604, "epoch": 2298, "lr": 1.4280692909681371e-05}
{"train_loss": 0.10040291398763657, "global_step": 204605, "epoch": 2298, "lr": 1.4280287240409157e-05}
{"train_loss": 0.027672085911035538, "global_step": 204606, "epoch": 2298, "lr": 1.4279881575939036e-05}
{"train_loss": 0.052165232598781586, "global_step": 204607, "epoch": 2298, "lr": 1.4279475916271034e-05}
{"train_loss": 0.06809390336275101, "global_step": 204608, "epoch": 2298, "lr": 1.4279070261405236e-05}
{"train_loss": 0.06871933490037918, "global_step": 204609, "epoch": 2298, "lr": 1.4278664611341668e-05}
{"train_loss": 0.06252601729140858, "global_step": 204610, "epoch": 2298, "lr": 1.42782589660804e-05, "val_loss": 8.30085277557373}
{"train_loss": 0.04536940157413483, "global_step": 204611, "epoch": 2299, "lr": 1.4277853325621494e-05}
{"train_loss": 0.0786801129579544, "global_step": 204612, "epoch": 2299, "lr": 1.4277447689964985e-05}
{"train_loss": 0.06727742403745651, "global_step": 204613, "epoch": 2299, "lr": 1.427704205911095e-05}
{"train_loss": 0.08078596740961075, "global_step": 204614, "epoch": 2299, "lr": 1.4276636433059421e-05}
{"train_loss": 0.0717242956161499, "global_step": 204615, "epoch": 2299, "lr": 1.427623081181047e-05}
{"train_loss": 0.028248673304915428, "global_step": 204616, "epoch": 2299, "lr": 1.4275825195364134e-05}
{"train_loss": 0.08653150498867035, "global_step": 204617, "epoch": 2299, "lr": 1.4275419583720495e-05}
{"train_loss": 0.020149758085608482, "global_step": 204618, "epoch": 2299, "lr": 1.4275013976879575e-05}
{"train_loss": 0.08234794437885284, "global_step": 204619, "epoch": 2299, "lr": 1.4274608374841458e-05}
{"train_loss": 0.018053948879241943, "global_step": 204620, "epoch": 2299, "lr": 1.4274202777606166e-05}
{"train_loss": 0.05738486349582672, "global_step": 204621, "epoch": 2299, "lr": 1.4273797185173793e-05}
{"train_loss": 0.0804540142416954, "global_step": 204622, "epoch": 2299, "lr": 1.427339159754435e-05}
{"train_loss": 0.07598591595888138, "global_step": 204623, "epoch": 2299, "lr": 1.4272986014717937e-05}
{"train_loss": 0.0661473274230957, "global_step": 204624, "epoch": 2299, "lr": 1.4272580436694566e-05}
{"train_loss": 0.08320242911577225, "global_step": 204625, "epoch": 2299, "lr": 1.4272174863474325e-05}
{"train_loss": 0.07028346508741379, "global_step": 204626, "epoch": 2299, "lr": 1.4271769295057242e-05}
{"train_loss": 0.06128578633069992, "global_step": 204627, "epoch": 2299, "lr": 1.42713637314434e-05}
{"train_loss": 0.06260967254638672, "global_step": 204628, "epoch": 2299, "lr": 1.4270958172632838e-05}
{"train_loss": 0.046601440757513046, "global_step": 204629, "epoch": 2299, "lr": 1.4270552618625594e-05}
{"train_loss": 0.04894502833485603, "global_step": 204630, "epoch": 2299, "lr": 1.4270147069421752e-05}
{"train_loss": 0.0804453194141388, "global_step": 204631, "epoch": 2299, "lr": 1.426974152502134e-05}
{"train_loss": 0.041519634425640106, "global_step": 204632, "epoch": 2299, "lr": 1.4269335985424438e-05}
{"train_loss": 0.0873849019408226, "global_step": 204633, "epoch": 2299, "lr": 1.4268930450631074e-05}
{"train_loss": 0.0857628583908081, "global_step": 204634, "epoch": 2299, "lr": 1.4268524920641335e-05}
{"train_loss": 0.0939655527472496, "global_step": 204635, "epoch": 2299, "lr": 1.4268119395455243e-05}
{"train_loss": 0.026508387178182602, "global_step": 204636, "epoch": 2299, "lr": 1.426771387507288e-05}
{"train_loss": 0.06464587152004242, "global_step": 204637, "epoch": 2299, "lr": 1.426730835949427e-05}
{"train_loss": 0.05111129581928253, "global_step": 204638, "epoch": 2299, "lr": 1.42669028487195e-05}
{"train_loss": 0.05853409692645073, "global_step": 204639, "epoch": 2299, "lr": 1.4266497342748597e-05}
{"train_loss": 0.0723690614104271, "global_step": 204640, "epoch": 2299, "lr": 1.426609184158164e-05}
{"train_loss": 0.051616765558719635, "global_step": 204641, "epoch": 2299, "lr": 1.426568634521866e-05}
{"train_loss": 0.0689270868897438, "global_step": 204642, "epoch": 2299, "lr": 1.426528085365974e-05}
{"train_loss": 0.0809214636683464, "global_step": 204643, "epoch": 2299, "lr": 1.4264875366904896e-05}
{"train_loss": 0.05766146630048752, "global_step": 204644, "epoch": 2299, "lr": 1.4264469884954223e-05}
{"train_loss": 0.048406485468149185, "global_step": 204645, "epoch": 2299, "lr": 1.4264064407807737e-05}
{"train_loss": 0.052964258939027786, "global_step": 204646, "epoch": 2299, "lr": 1.4263658935465518e-05}
{"train_loss": 0.04631185159087181, "global_step": 204647, "epoch": 2299, "lr": 1.4263253467927628e-05}
{"train_loss": 0.047654423862695694, "global_step": 204648, "epoch": 2299, "lr": 1.426284800519409e-05}
{"train_loss": 0.025782158598303795, "global_step": 204649, "epoch": 2299, "lr": 1.4262442547264993e-05}
{"train_loss": 0.031162461265921593, "global_step": 204650, "epoch": 2299, "lr": 1.4262037094140363e-05}
{"train_loss": 0.05589490756392479, "global_step": 204651, "epoch": 2299, "lr": 1.4261631645820278e-05}
{"train_loss": 0.04451795294880867, "global_step": 204652, "epoch": 2299, "lr": 1.4261226202304767e-05}
{"train_loss": 0.09925131499767303, "global_step": 204653, "epoch": 2299, "lr": 1.4260820763593912e-05}
{"train_loss": 0.07531114667654037, "global_step": 204654, "epoch": 2299, "lr": 1.4260415329687738e-05}
{"train_loss": 0.08366887271404266, "global_step": 204655, "epoch": 2299, "lr": 1.4260009900586335e-05}
{"train_loss": 0.06587535887956619, "global_step": 204656, "epoch": 2299, "lr": 1.4259604476289724e-05}
{"train_loss": 0.021634697914123535, "global_step": 204657, "epoch": 2299, "lr": 1.4259199056797984e-05}
{"train_loss": 0.07734175026416779, "global_step": 204658, "epoch": 2299, "lr": 1.4258793642111146e-05}
{"train_loss": 0.056804902851581573, "global_step": 204659, "epoch": 2299, "lr": 1.4258388232229291e-05}
{"train_loss": 0.08087920397520065, "global_step": 204660, "epoch": 2299, "lr": 1.4257982827152444e-05}
{"train_loss": 0.03748776763677597, "global_step": 204661, "epoch": 2299, "lr": 1.4257577426880698e-05}
{"train_loss": 0.032473813742399216, "global_step": 204662, "epoch": 2299, "lr": 1.4257172031414061e-05}
{"train_loss": 0.05941036716103554, "global_step": 204663, "epoch": 2299, "lr": 1.4256766640752634e-05}
{"train_loss": 0.038450099527835846, "global_step": 204664, "epoch": 2299, "lr": 1.4256361254896428e-05}
{"train_loss": 0.03689592331647873, "global_step": 204665, "epoch": 2299, "lr": 1.4255955873845539e-05}
{"train_loss": 0.043993495404720306, "global_step": 204666, "epoch": 2299, "lr": 1.4255550497599996e-05}
{"train_loss": 0.025677958503365517, "global_step": 204667, "epoch": 2299, "lr": 1.4255145126159847e-05}
{"train_loss": 0.06928359717130661, "global_step": 204668, "epoch": 2299, "lr": 1.4254739759525171e-05}
{"train_loss": 0.08137279748916626, "global_step": 204669, "epoch": 2299, "lr": 1.4254334397695995e-05}
{"train_loss": 0.07454866915941238, "global_step": 204670, "epoch": 2299, "lr": 1.4253929040672403e-05}
{"train_loss": 0.0720418244600296, "global_step": 204671, "epoch": 2299, "lr": 1.4253523688454423e-05}
{"train_loss": 0.05971153825521469, "global_step": 204672, "epoch": 2299, "lr": 1.4253118341042133e-05}
{"train_loss": 0.023029664531350136, "global_step": 204673, "epoch": 2299, "lr": 1.4252712998435558e-05}
{"train_loss": 0.05652828514575958, "global_step": 204674, "epoch": 2299, "lr": 1.4252307660634789e-05}
{"train_loss": 0.12338393926620483, "global_step": 204675, "epoch": 2299, "lr": 1.4251902327639844e-05}
{"train_loss": 0.053211700171232224, "global_step": 204676, "epoch": 2299, "lr": 1.4251496999450813e-05}
{"train_loss": 0.04531349241733551, "global_step": 204677, "epoch": 2299, "lr": 1.4251091676067712e-05}
{"train_loss": 0.05325319245457649, "global_step": 204678, "epoch": 2299, "lr": 1.4250686357490639e-05}
{"train_loss": 0.04352831467986107, "global_step": 204679, "epoch": 2299, "lr": 1.4250281043719605e-05}
{"train_loss": 0.0669814720749855, "global_step": 204680, "epoch": 2299, "lr": 1.4249875734754702e-05}
{"train_loss": 0.0547143779695034, "global_step": 204681, "epoch": 2299, "lr": 1.4249470430595952e-05}
{"train_loss": 0.08031704276800156, "global_step": 204682, "epoch": 2299, "lr": 1.4249065131243438e-05}
{"train_loss": 0.06257887929677963, "global_step": 204683, "epoch": 2299, "lr": 1.4248659836697186e-05}
{"train_loss": 0.09907111525535583, "global_step": 204684, "epoch": 2299, "lr": 1.4248254546957273e-05}
{"train_loss": 0.048384834080934525, "global_step": 204685, "epoch": 2299, "lr": 1.4247849262023755e-05}
{"train_loss": 0.045664746314287186, "global_step": 204686, "epoch": 2299, "lr": 1.4247443981896664e-05}
{"train_loss": 0.06465142220258713, "global_step": 204687, "epoch": 2299, "lr": 1.4247038706576082e-05}
{"train_loss": 0.07778435200452805, "global_step": 204688, "epoch": 2299, "lr": 1.424663343606204e-05}
{"train_loss": 0.058652155101299286, "global_step": 204689, "epoch": 2299, "lr": 1.4246228170354613e-05}
{"train_loss": 0.05111999064683914, "global_step": 204690, "epoch": 2299, "lr": 1.4245822909453833e-05}
{"train_loss": 0.021157464012503624, "global_step": 204691, "epoch": 2299, "lr": 1.4245417653359778e-05}
{"train_loss": 0.07109629362821579, "global_step": 204692, "epoch": 2299, "lr": 1.4245012402072477e-05}
{"train_loss": 0.0813135877251625, "global_step": 204693, "epoch": 2299, "lr": 1.4244607155592016e-05}
{"train_loss": 0.03226067125797272, "global_step": 204694, "epoch": 2299, "lr": 1.4244201913918409e-05}
{"train_loss": 0.046147603541612625, "global_step": 204695, "epoch": 2299, "lr": 1.4243796677051758e-05}
{"train_loss": 0.09610260277986526, "global_step": 204696, "epoch": 2299, "lr": 1.4243391444992071e-05}
{"train_loss": 0.04611817002296448, "global_step": 204697, "epoch": 2299, "lr": 1.424298621773944e-05}
{"train_loss": 0.0684361681342125, "global_step": 204698, "epoch": 2299, "lr": 1.4242580995293891e-05}
{"train_loss": 0.059638095726625304, "global_step": 204699, "epoch": 2299, "lr": 1.4242175777655503e-05, "val_loss": 8.3245849609375}
{"train_loss": 0.05147204175591469, "global_step": 204700, "epoch": 2300, "lr": 1.4241770564824308e-05}
{"train_loss": 0.05471952259540558, "global_step": 204701, "epoch": 2300, "lr": 1.4241365356800384e-05}
{"train_loss": 0.02201198786497116, "global_step": 204702, "epoch": 2300, "lr": 1.424096015358376e-05}
{"train_loss": 0.03137999400496483, "global_step": 204703, "epoch": 2300, "lr": 1.4240554955174512e-05}
{"train_loss": 0.04242406412959099, "global_step": 204704, "epoch": 2300, "lr": 1.4240149761572675e-05}
{"train_loss": 0.035238154232501984, "global_step": 204705, "epoch": 2300, "lr": 1.423974457277833e-05}
{"train_loss": 0.07323530316352844, "global_step": 204706, "epoch": 2300, "lr": 1.4239339388791512e-05}
{"train_loss": 0.07455778867006302, "global_step": 204707, "epoch": 2300, "lr": 1.4238934209612265e-05}
{"train_loss": 0.040494486689567566, "global_step": 204708, "epoch": 2300, "lr": 1.4238529035240672e-05}
{"train_loss": 0.08441770076751709, "global_step": 204709, "epoch": 2300, "lr": 1.423812386567676e-05}
{"train_loss": 0.07761882245540619, "global_step": 204710, "epoch": 2300, "lr": 1.4237718700920611e-05}
{"train_loss": 0.046632543206214905, "global_step": 204711, "epoch": 2300, "lr": 1.4237313540972248e-05}
{"train_loss": 0.06959086656570435, "global_step": 204712, "epoch": 2300, "lr": 1.423690838583176e-05}
{"train_loss": 0.07412014901638031, "global_step": 204713, "epoch": 2300, "lr": 1.4236503235499166e-05}
{"train_loss": 0.0631999671459198, "global_step": 204714, "epoch": 2300, "lr": 1.4236098089974554e-05}
{"train_loss": 0.04981027543544769, "global_step": 204715, "epoch": 2300, "lr": 1.423569294925795e-05}
{"train_loss": 0.05506459251046181, "global_step": 204716, "epoch": 2300, "lr": 1.4235287813349435e-05}
{"train_loss": 0.0382196269929409, "global_step": 204717, "epoch": 2300, "lr": 1.4234882682249034e-05}
{"train_loss": 0.030642462894320488, "global_step": 204718, "epoch": 2300, "lr": 1.4234477555956832e-05}
{"train_loss": 0.08979960530996323, "global_step": 204719, "epoch": 2300, "lr": 1.423407243447285e-05}
{"train_loss": 0.043838754296302795, "global_step": 204720, "epoch": 2300, "lr": 1.4233667317797167e-05}
{"train_loss": 0.062772735953331, "global_step": 204721, "epoch": 2300, "lr": 1.4233262205929843e-05}
{"train_loss": 0.03814765065908432, "global_step": 204722, "epoch": 2300, "lr": 1.4232857098870911e-05}
{"train_loss": 0.013642559759318829, "global_step": 204723, "epoch": 2300, "lr": 1.4232451996620444e-05}
{"train_loss": 0.060315970331430435, "global_step": 204724, "epoch": 2300, "lr": 1.4232046899178475e-05}
{"train_loss": 0.07017247378826141, "global_step": 204725, "epoch": 2300, "lr": 1.4231641806545087e-05}
{"train_loss": 0.05620681494474411, "global_step": 204726, "epoch": 2300, "lr": 1.4231236718720304e-05}
{"train_loss": 0.056984562426805496, "global_step": 204727, "epoch": 2300, "lr": 1.4230831635704211e-05}
{"train_loss": 0.07602483779191971, "global_step": 204728, "epoch": 2300, "lr": 1.423042655749683e-05}
{"train_loss": 0.08814658224582672, "global_step": 204729, "epoch": 2300, "lr": 1.4230021484098249e-05}
{"train_loss": 0.03702868893742561, "global_step": 204730, "epoch": 2300, "lr": 1.422961641550849e-05}
{"train_loss": 0.05067732185125351, "global_step": 204731, "epoch": 2300, "lr": 1.4229211351727634e-05}
{"train_loss": 0.054059844464063644, "global_step": 204732, "epoch": 2300, "lr": 1.4228806292755715e-05}
{"train_loss": 0.03790891915559769, "global_step": 204733, "epoch": 2300, "lr": 1.4228401238592808e-05}
{"train_loss": 0.060298994183540344, "global_step": 204734, "epoch": 2300, "lr": 1.4227996189238945e-05}
{"train_loss": 0.04411900043487549, "global_step": 204735, "epoch": 2300, "lr": 1.4227591144694208e-05}
{"train_loss": 0.06733421981334686, "global_step": 204736, "epoch": 2300, "lr": 1.4227186104958617e-05}
{"train_loss": 0.08439061790704727, "global_step": 204737, "epoch": 2300, "lr": 1.422678107003224e-05}
{"train_loss": 0.039076101034879684, "global_step": 204738, "epoch": 2300, "lr": 1.4226376039915163e-05}
{"train_loss": 0.03254883363842964, "global_step": 204739, "epoch": 2300, "lr": 1.4225971014607403e-05}
{"train_loss": 0.05023498460650444, "global_step": 204740, "epoch": 2300, "lr": 1.422556599410902e-05}
{"train_loss": 0.06424595415592194, "global_step": 204741, "epoch": 2300, "lr": 1.4225160978420083e-05}
{"train_loss": 0.053046710789203644, "global_step": 204742, "epoch": 2300, "lr": 1.4224755967540621e-05}
{"train_loss": 0.06880254298448563, "global_step": 204743, "epoch": 2300, "lr": 1.4224350961470722e-05}
{"train_loss": 0.04858267679810524, "global_step": 204744, "epoch": 2300, "lr": 1.4223945960210423e-05}
{"train_loss": 0.04099768400192261, "global_step": 204745, "epoch": 2300, "lr": 1.4223540963759757e-05}
{"train_loss": 0.07159625738859177, "global_step": 204746, "epoch": 2300, "lr": 1.4223135972118822e-05}
{"train_loss": 0.11699077486991882, "global_step": 204747, "epoch": 2300, "lr": 1.4222730985287636e-05}
{"train_loss": 0.05066239461302757, "global_step": 204748, "epoch": 2300, "lr": 1.4222326003266278e-05}
{"train_loss": 0.06209147348999977, "global_step": 204749, "epoch": 2300, "lr": 1.422192102605478e-05}
{"train_loss": 0.08222551643848419, "global_step": 204750, "epoch": 2300, "lr": 1.4221516053653221e-05}
{"train_loss": 0.04843561351299286, "global_step": 204751, "epoch": 2300, "lr": 1.4221111086061634e-05}
{"train_loss": 0.06679397821426392, "global_step": 204752, "epoch": 2300, "lr": 1.4220706123280091e-05}
{"train_loss": 0.03816807642579079, "global_step": 204753, "epoch": 2300, "lr": 1.4220301165308625e-05}
{"train_loss": 0.021580111235380173, "global_step": 204754, "epoch": 2300, "lr": 1.4219896212147321e-05}
{"train_loss": 0.03713659197092056, "global_step": 204755, "epoch": 2300, "lr": 1.4219491263796198e-05}
{"train_loss": 0.03201805055141449, "global_step": 204756, "epoch": 2300, "lr": 1.4219086320255332e-05}
{"train_loss": 0.05403488874435425, "global_step": 204757, "epoch": 2300, "lr": 1.4218681381524784e-05}
{"train_loss": 0.06674958765506744, "global_step": 204758, "epoch": 2300, "lr": 1.4218276447604583e-05}
{"train_loss": 0.0716337189078331, "global_step": 204759, "epoch": 2300, "lr": 1.4217871518494819e-05}
{"train_loss": 0.06555211544036865, "global_step": 204760, "epoch": 2300, "lr": 1.4217466594195506e-05}
{"train_loss": 0.03733430430293083, "global_step": 204761, "epoch": 2300, "lr": 1.4217061674706738e-05}
{"train_loss": 0.1508079171180725, "global_step": 204762, "epoch": 2300, "lr": 1.421665676002853e-05}
{"train_loss": 0.07347694039344788, "global_step": 204763, "epoch": 2300, "lr": 1.4216251850160978e-05}
{"train_loss": 0.04750154912471771, "global_step": 204764, "epoch": 2300, "lr": 1.4215846945104095e-05}
{"train_loss": 0.15912482142448425, "global_step": 204765, "epoch": 2300, "lr": 1.4215442044857968e-05}
{"train_loss": 0.10391777008771896, "global_step": 204766, "epoch": 2300, "lr": 1.4215037149422627e-05}
{"train_loss": 0.04090888798236847, "global_step": 204767, "epoch": 2300, "lr": 1.4214632258798149e-05}
{"train_loss": 0.05136755481362343, "global_step": 204768, "epoch": 2300, "lr": 1.4214227372984567e-05}
{"train_loss": 0.0619688555598259, "global_step": 204769, "epoch": 2300, "lr": 1.421382249198196e-05}
{"train_loss": 0.043724436312913895, "global_step": 204770, "epoch": 2300, "lr": 1.4213417615790354e-05}
{"train_loss": 0.0493459589779377, "global_step": 204771, "epoch": 2300, "lr": 1.421301274440983e-05}
{"train_loss": 0.1005048006772995, "global_step": 204772, "epoch": 2300, "lr": 1.4212607877840422e-05}
{"train_loss": 0.046488985419273376, "global_step": 204773, "epoch": 2300, "lr": 1.4212203016082187e-05}
{"train_loss": 0.041502099484205246, "global_step": 204774, "epoch": 2300, "lr": 1.4211798159135203e-05}
{"train_loss": 0.07370343804359436, "global_step": 204775, "epoch": 2300, "lr": 1.4211393306999487e-05}
{"train_loss": 0.04053505137562752, "global_step": 204776, "epoch": 2300, "lr": 1.4210988459675135e-05}
{"train_loss": 0.07710816711187363, "global_step": 204777, "epoch": 2300, "lr": 1.421058361716217e-05}
{"train_loss": 0.03530418500304222, "global_step": 204778, "epoch": 2300, "lr": 1.4210178779460648e-05}
{"train_loss": 0.05879149213433266, "global_step": 204779, "epoch": 2300, "lr": 1.4209773946570642e-05}
{"train_loss": 0.08425839990377426, "global_step": 204780, "epoch": 2300, "lr": 1.4209369118492183e-05}
{"train_loss": 0.12321125715970993, "global_step": 204781, "epoch": 2300, "lr": 1.4208964295225357e-05}
{"train_loss": 0.03635652735829353, "global_step": 204782, "epoch": 2300, "lr": 1.4208559476770195e-05}
{"train_loss": 0.06814251095056534, "global_step": 204783, "epoch": 2300, "lr": 1.4208154663126737e-05}
{"train_loss": 0.03881000354886055, "global_step": 204784, "epoch": 2300, "lr": 1.4207749854295078e-05}
{"train_loss": 0.08607722818851471, "global_step": 204785, "epoch": 2300, "lr": 1.4207345050275234e-05}
{"train_loss": 0.06407129764556885, "global_step": 204786, "epoch": 2300, "lr": 1.4206940251067292e-05}
{"train_loss": 0.07088865339756012, "global_step": 204787, "epoch": 2300, "lr": 1.4206535456671272e-05}
{"train_loss": 0.06019496814128053, "global_step": 204788, "epoch": 2300, "lr": 1.4206130667087264e-05, "train/sim_max_reward_0": 0.6480532484366673, "train/sim_max_reward_1": 0.9241871654955466, "train/sim_max_reward_2": 0.6184139874553822, "train/sim_max_reward_3": 0.6607927909611392, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.6144586489299244, "test/sim_max_reward_4500000": 1.0, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.931598544118808, "test/sim_max_reward_4500003": 0.46929933920963396, "test/sim_max_reward_4500004": 0.9198154735522523, "test/sim_max_reward_4500005": 0.37513816410621986, "test/sim_max_reward_4500006": 0.9523680736764593, "test/sim_max_reward_4500007": 0.46118884924183584, "test/sim_max_reward_4500008": 0.11604520644090471, "test/sim_max_reward_4500009": 0.05006421541117228, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.21562600374790444, "test/sim_max_reward_4500014": 0.988116401757774, "test/sim_max_reward_4500015": 0.9578759351761038, "test/sim_max_reward_4500016": 1.0, "test/sim_max_reward_4500017": 0.5588740555450665, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.952888924557963, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9597629566082869, "test/sim_max_reward_4500022": 0.9756699845781047, "test/sim_max_reward_4500023": 0.5012767979073568, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9963776183942537, "test/sim_max_reward_4500026": 0.2200071257563209, "test/sim_max_reward_4500027": 0.9585015265415584, "test/sim_max_reward_4500028": 0.6890265686240484, "test/sim_max_reward_4500029": 0.54519619751721, "test/sim_max_reward_4500030": 0.9551376068448201, "test/sim_max_reward_4500031": 0.5106713125108918, "test/sim_max_reward_4500032": 0.5732107469729465, "test/sim_max_reward_4500033": 0.7147402088412739, "test/sim_max_reward_4500034": 0.923304534301528, "test/sim_max_reward_4500035": 0.6938549942351029, "test/sim_max_reward_4500036": 0.9698277492453167, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.162601504236713, "test/sim_max_reward_4500039": 0.9647246690936128, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6096644890741556, "test/sim_max_reward_4500042": 0.015491447644703723, "test/sim_max_reward_4500043": 0.4208898807306349, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.3494152340650646, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9864594875169047, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7443176402131101, "test/mean_score": 0.5928942365556581, "val_loss": 8.37083911895752, "train_action_mse_error": 16.994653701782227}
{"train_loss": 0.05567365884780884, "global_step": 204789, "epoch": 2301, "lr": 1.420572588231529e-05}
{"train_loss": 0.06020308658480644, "global_step": 204790, "epoch": 2301, "lr": 1.420532110235544e-05}
{"train_loss": 0.03999553248286247, "global_step": 204791, "epoch": 2301, "lr": 1.4204916327207729e-05}
{"train_loss": 0.047300785779953, "global_step": 204792, "epoch": 2301, "lr": 1.420451155687223e-05}
{"train_loss": 0.04758049175143242, "global_step": 204793, "epoch": 2301, "lr": 1.4204106791349015e-05}
{"train_loss": 0.07232096046209335, "global_step": 204794, "epoch": 2301, "lr": 1.4203702030638105e-05}
{"train_loss": 0.08553427457809448, "global_step": 204795, "epoch": 2301, "lr": 1.4203297274739585e-05}
{"train_loss": 0.07633505761623383, "global_step": 204796, "epoch": 2301, "lr": 1.4202892523653483e-05}
{"train_loss": 0.04394320398569107, "global_step": 204797, "epoch": 2301, "lr": 1.420248777737988e-05}
{"train_loss": 0.10313240438699722, "global_step": 204798, "epoch": 2301, "lr": 1.42020830359188e-05}
{"train_loss": 0.07218301296234131, "global_step": 204799, "epoch": 2301, "lr": 1.4201678299270326e-05}
{"train_loss": 0.13948169350624084, "global_step": 204800, "epoch": 2301, "lr": 1.4201273567434492e-05}
{"train_loss": 0.025116074830293655, "global_step": 204801, "epoch": 2301, "lr": 1.4200868840411375e-05}
{"train_loss": 0.058387842029333115, "global_step": 204802, "epoch": 2301, "lr": 1.4200464118200995e-05}
{"train_loss": 0.08146000653505325, "global_step": 204803, "epoch": 2301, "lr": 1.4200059400803439e-05}
{"train_loss": 0.13689914345741272, "global_step": 204804, "epoch": 2301, "lr": 1.4199654688218738e-05}
{"train_loss": 0.05960102379322052, "global_step": 204805, "epoch": 2301, "lr": 1.419924998044697e-05}
{"train_loss": 0.06787335872650146, "global_step": 204806, "epoch": 2301, "lr": 1.4198845277488165e-05}
{"train_loss": 0.06083153188228607, "global_step": 204807, "epoch": 2301, "lr": 1.4198440579342403e-05}
{"train_loss": 0.08932730555534363, "global_step": 204808, "epoch": 2301, "lr": 1.4198035886009708e-05}
{"train_loss": 0.07694380730390549, "global_step": 204809, "epoch": 2301, "lr": 1.419763119749017e-05}
{"train_loss": 0.08745065331459045, "global_step": 204810, "epoch": 2301, "lr": 1.41972265137838e-05}
{"train_loss": 0.07136041671037674, "global_step": 204811, "epoch": 2301, "lr": 1.4196821834890683e-05}
{"train_loss": 0.0436333492398262, "global_step": 204812, "epoch": 2301, "lr": 1.4196417160810882e-05}
{"train_loss": 0.057752810418605804, "global_step": 204813, "epoch": 2301, "lr": 1.419601249154442e-05}
{"train_loss": 0.08879673480987549, "global_step": 204814, "epoch": 2301, "lr": 1.4195607827091384e-05}
{"train_loss": 0.06805845350027084, "global_step": 204815, "epoch": 2301, "lr": 1.4195203167451797e-05}
{"train_loss": 0.08654318749904633, "global_step": 204816, "epoch": 2301, "lr": 1.419479851262574e-05}
{"train_loss": 0.07994362711906433, "global_step": 204817, "epoch": 2301, "lr": 1.419439386261326e-05}
{"train_loss": 0.14808008074760437, "global_step": 204818, "epoch": 2301, "lr": 1.4193989217414389e-05}
{"train_loss": 0.07472316920757294, "global_step": 204819, "epoch": 2301, "lr": 1.4193584577029217e-05}
{"train_loss": 0.04674750566482544, "global_step": 204820, "epoch": 2301, "lr": 1.4193179941457769e-05}
{"train_loss": 0.0674893781542778, "global_step": 204821, "epoch": 2301, "lr": 1.4192775310700124e-05}
{"train_loss": 0.035437267273664474, "global_step": 204822, "epoch": 2301, "lr": 1.4192370684756323e-05}
{"train_loss": 0.05790415033698082, "global_step": 204823, "epoch": 2301, "lr": 1.4191966063626405e-05}
{"train_loss": 0.08579325675964355, "global_step": 204824, "epoch": 2301, "lr": 1.4191561447310459e-05}
{"train_loss": 0.11146488040685654, "global_step": 204825, "epoch": 2301, "lr": 1.41911568358085e-05}
{"train_loss": 0.06222355365753174, "global_step": 204826, "epoch": 2301, "lr": 1.4190752229120623e-05}
{"train_loss": 0.0199624914675951, "global_step": 204827, "epoch": 2301, "lr": 1.4190347627246848e-05}
{"train_loss": 0.05961380526423454, "global_step": 204828, "epoch": 2301, "lr": 1.4189943030187247e-05}
{"train_loss": 0.13856464624404907, "global_step": 204829, "epoch": 2301, "lr": 1.418953843794188e-05}
{"train_loss": 0.0561014749109745, "global_step": 204830, "epoch": 2301, "lr": 1.4189133850510783e-05}
{"train_loss": 0.08322501927614212, "global_step": 204831, "epoch": 2301, "lr": 1.4188729267894035e-05}
{"train_loss": 0.05448223650455475, "global_step": 204832, "epoch": 2301, "lr": 1.4188324690091659e-05}
{"train_loss": 0.064192034304142, "global_step": 204833, "epoch": 2301, "lr": 1.4187920117103743e-05}
{"train_loss": 0.08096392452716827, "global_step": 204834, "epoch": 2301, "lr": 1.4187515548930308e-05}
{"train_loss": 0.058809954673051834, "global_step": 204835, "epoch": 2301, "lr": 1.4187110985571444e-05}
{"train_loss": 0.055155545473098755, "global_step": 204836, "epoch": 2301, "lr": 1.4186706427027168e-05}
{"train_loss": 0.15036144852638245, "global_step": 204837, "epoch": 2301, "lr": 1.418630187329757e-05}
{"train_loss": 0.07213998585939407, "global_step": 204838, "epoch": 2301, "lr": 1.418589732438267e-05}
{"train_loss": 0.09211903065443039, "global_step": 204839, "epoch": 2301, "lr": 1.4185492780282556e-05}
{"train_loss": 0.030506912618875504, "global_step": 204840, "epoch": 2301, "lr": 1.4185088240997246e-05}
{"train_loss": 0.11317436397075653, "global_step": 204841, "epoch": 2301, "lr": 1.4184683706526836e-05}
{"train_loss": 0.061035122722387314, "global_step": 204842, "epoch": 2301, "lr": 1.418427917687134e-05}
{"train_loss": 0.07741126418113708, "global_step": 204843, "epoch": 2301, "lr": 1.4183874652030849e-05}
{"train_loss": 0.08456023782491684, "global_step": 204844, "epoch": 2301, "lr": 1.4183470132005382e-05}
{"train_loss": 0.08111664652824402, "global_step": 204845, "epoch": 2301, "lr": 1.4183065616795027e-05}
{"train_loss": 0.04885018244385719, "global_step": 204846, "epoch": 2301, "lr": 1.4182661106399809e-05}
{"train_loss": 0.043784089386463165, "global_step": 204847, "epoch": 2301, "lr": 1.4182256600819793e-05}
{"train_loss": 0.06207653880119324, "global_step": 204848, "epoch": 2301, "lr": 1.4181852100055048e-05}
{"train_loss": 0.06003870815038681, "global_step": 204849, "epoch": 2301, "lr": 1.4181447604105607e-05}
{"train_loss": 0.0968179851770401, "global_step": 204850, "epoch": 2301, "lr": 1.4181043112971548e-05}
{"train_loss": 0.10644443333148956, "global_step": 204851, "epoch": 2301, "lr": 1.4180638626652893e-05}
{"train_loss": 0.04268793761730194, "global_step": 204852, "epoch": 2301, "lr": 1.4180234145149734e-05}
{"train_loss": 0.05196605250239372, "global_step": 204853, "epoch": 2301, "lr": 1.4179829668462085e-05}
{"train_loss": 0.03477371856570244, "global_step": 204854, "epoch": 2301, "lr": 1.417942519659004e-05}
{"train_loss": 0.07536159455776215, "global_step": 204855, "epoch": 2301, "lr": 1.4179020729533637e-05}
{"train_loss": 0.05618898198008537, "global_step": 204856, "epoch": 2301, "lr": 1.417861626729291e-05}
{"train_loss": 0.07978405803442001, "global_step": 204857, "epoch": 2301, "lr": 1.4178211809867948e-05}
{"train_loss": 0.10453791171312332, "global_step": 204858, "epoch": 2301, "lr": 1.417780735725877e-05}
{"train_loss": 0.034230850636959076, "global_step": 204859, "epoch": 2301, "lr": 1.417740290946547e-05}
{"train_loss": 0.016468850895762444, "global_step": 204860, "epoch": 2301, "lr": 1.4176998466488079e-05}
{"train_loss": 0.06454532593488693, "global_step": 204861, "epoch": 2301, "lr": 1.417659402832664e-05}
{"train_loss": 0.06097140163183212, "global_step": 204862, "epoch": 2301, "lr": 1.4176189594981231e-05}
{"train_loss": 0.08513836562633514, "global_step": 204863, "epoch": 2301, "lr": 1.4175785166451887e-05}
{"train_loss": 0.030542103573679924, "global_step": 204864, "epoch": 2301, "lr": 1.417538074273867e-05}
{"train_loss": 0.0501134991645813, "global_step": 204865, "epoch": 2301, "lr": 1.4174976323841655e-05}
{"train_loss": 0.07807454466819763, "global_step": 204866, "epoch": 2301, "lr": 1.4174571909760859e-05}
{"train_loss": 0.05620477721095085, "global_step": 204867, "epoch": 2301, "lr": 1.4174167500496371e-05}
{"train_loss": 0.09822076559066772, "global_step": 204868, "epoch": 2301, "lr": 1.4173763096048215e-05}
{"train_loss": 0.1274552345275879, "global_step": 204869, "epoch": 2301, "lr": 1.4173358696416471e-05}
{"train_loss": 0.04324635863304138, "global_step": 204870, "epoch": 2301, "lr": 1.4172954301601177e-05}
{"train_loss": 0.04422196373343468, "global_step": 204871, "epoch": 2301, "lr": 1.41725499116024e-05}
{"train_loss": 0.07982746511697769, "global_step": 204872, "epoch": 2301, "lr": 1.4172145526420171e-05}
{"train_loss": 0.040792156010866165, "global_step": 204873, "epoch": 2301, "lr": 1.417174114605458e-05}
{"train_loss": 0.053769707679748535, "global_step": 204874, "epoch": 2301, "lr": 1.4171336770505644e-05}
{"train_loss": 0.0927368775010109, "global_step": 204875, "epoch": 2301, "lr": 1.4170932399773452e-05}
{"train_loss": 0.054537784308195114, "global_step": 204876, "epoch": 2301, "lr": 1.4170528033858022e-05}
{"train_loss": 0.06988955705604527, "global_step": 204877, "epoch": 2301, "lr": 1.4170123672759445e-05, "val_loss": 8.421721458435059}
{"train_loss": 0.05973600223660469, "global_step": 204878, "epoch": 2302, "lr": 1.4169719316477748e-05}
{"train_loss": 0.10012955218553543, "global_step": 204879, "epoch": 2302, "lr": 1.4169314965013004e-05}
{"train_loss": 0.07460028678178787, "global_step": 204880, "epoch": 2302, "lr": 1.4168910618365244e-05}
{"train_loss": 0.09868355840444565, "global_step": 204881, "epoch": 2302, "lr": 1.4168506276534554e-05}
{"train_loss": 0.07338035106658936, "global_step": 204882, "epoch": 2302, "lr": 1.4168101939520961e-05}
{"train_loss": 0.03901329264044762, "global_step": 204883, "epoch": 2302, "lr": 1.4167697607324526e-05}
{"train_loss": 0.07224790751934052, "global_step": 204884, "epoch": 2302, "lr": 1.416729327994532e-05}
{"train_loss": 0.10759387910366058, "global_step": 204885, "epoch": 2302, "lr": 1.4166888957383373e-05}
{"train_loss": 0.04374629631638527, "global_step": 204886, "epoch": 2302, "lr": 1.4166484639638767e-05}
{"train_loss": 0.05541161820292473, "global_step": 204887, "epoch": 2302, "lr": 1.4166080326711523e-05}
{"train_loss": 0.033094316720962524, "global_step": 204888, "epoch": 2302, "lr": 1.4165676018601726e-05}
{"train_loss": 0.12142961472272873, "global_step": 204889, "epoch": 2302, "lr": 1.4165271715309409e-05}
{"train_loss": 0.10524161905050278, "global_step": 204890, "epoch": 2302, "lr": 1.4164867416834643e-05}
{"train_loss": 0.08203443884849548, "global_step": 204891, "epoch": 2302, "lr": 1.4164463123177463e-05}
{"train_loss": 0.08321808278560638, "global_step": 204892, "epoch": 2302, "lr": 1.4164058834337945e-05}
{"train_loss": 0.04095035046339035, "global_step": 204893, "epoch": 2302, "lr": 1.4163654550316136e-05}
{"train_loss": 0.051267169415950775, "global_step": 204894, "epoch": 2302, "lr": 1.4163250271112067e-05}
{"train_loss": 0.038764115422964096, "global_step": 204895, "epoch": 2302, "lr": 1.4162845996725832e-05}
{"train_loss": 0.04687220975756645, "global_step": 204896, "epoch": 2302, "lr": 1.416244172715745e-05}
{"train_loss": 0.05795937404036522, "global_step": 204897, "epoch": 2302, "lr": 1.4162037462407003e-05}
{"train_loss": 0.11780830472707748, "global_step": 204898, "epoch": 2302, "lr": 1.4161633202474534e-05}
{"train_loss": 0.030655255541205406, "global_step": 204899, "epoch": 2302, "lr": 1.4161228947360084e-05}
{"train_loss": 0.05360115319490433, "global_step": 204900, "epoch": 2302, "lr": 1.4160824697063718e-05}
{"train_loss": 0.11554567515850067, "global_step": 204901, "epoch": 2302, "lr": 1.416042045158551e-05}
{"train_loss": 0.020699214190244675, "global_step": 204902, "epoch": 2302, "lr": 1.416001621092548e-05}
{"train_loss": 0.08378579467535019, "global_step": 204903, "epoch": 2302, "lr": 1.4159611975083715e-05}
{"train_loss": 0.01378893107175827, "global_step": 204904, "epoch": 2302, "lr": 1.4159207744060232e-05}
{"train_loss": 0.08841466158628464, "global_step": 204905, "epoch": 2302, "lr": 1.415880351785513e-05}
{"train_loss": 0.07919018715620041, "global_step": 204906, "epoch": 2302, "lr": 1.4158399296468422e-05}
{"train_loss": 0.032037124037742615, "global_step": 204907, "epoch": 2302, "lr": 1.4157995079900193e-05}
{"train_loss": 0.05781977251172066, "global_step": 204908, "epoch": 2302, "lr": 1.4157590868150472e-05}
{"train_loss": 0.04438953101634979, "global_step": 204909, "epoch": 2302, "lr": 1.415718666121934e-05}
{"train_loss": 0.0493868887424469, "global_step": 204910, "epoch": 2302, "lr": 1.4156782459106827e-05}
{"train_loss": 0.024140967056155205, "global_step": 204911, "epoch": 2302, "lr": 1.4156378261813009e-05}
{"train_loss": 0.045401424169540405, "global_step": 204912, "epoch": 2302, "lr": 1.4155974069337912e-05}
{"train_loss": 0.06708057969808578, "global_step": 204913, "epoch": 2302, "lr": 1.4155569881681623e-05}
{"train_loss": 0.1259012669324875, "global_step": 204914, "epoch": 2302, "lr": 1.415516569884417e-05}
{"train_loss": 0.032333578914403915, "global_step": 204915, "epoch": 2302, "lr": 1.4154761520825626e-05}
{"train_loss": 0.07788262516260147, "global_step": 204916, "epoch": 2302, "lr": 1.415435734762603e-05}
{"train_loss": 0.03434367477893829, "global_step": 204917, "epoch": 2302, "lr": 1.4153953179245455e-05}
{"train_loss": 0.10831660777330399, "global_step": 204918, "epoch": 2302, "lr": 1.4153549015683931e-05}
{"train_loss": 0.08529070764780045, "global_step": 204919, "epoch": 2302, "lr": 1.4153144856941526e-05}
{"train_loss": 0.05472508445382118, "global_step": 204920, "epoch": 2302, "lr": 1.4152740703018302e-05}
{"train_loss": 0.03200250864028931, "global_step": 204921, "epoch": 2302, "lr": 1.4152336553914297e-05}
{"train_loss": 0.10570796579122543, "global_step": 204922, "epoch": 2302, "lr": 1.4151932409629587e-05}
{"train_loss": 0.05190140753984451, "global_step": 204923, "epoch": 2302, "lr": 1.4151528270164194e-05}
{"train_loss": 0.06442221999168396, "global_step": 204924, "epoch": 2302, "lr": 1.4151124135518212e-05}
{"train_loss": 0.037827931344509125, "global_step": 204925, "epoch": 2302, "lr": 1.4150720005691653e-05}
{"train_loss": 0.03647785633802414, "global_step": 204926, "epoch": 2302, "lr": 1.4150315880684611e-05}
{"train_loss": 0.0635601282119751, "global_step": 204927, "epoch": 2302, "lr": 1.4149911760497109e-05}
{"train_loss": 0.04440109059214592, "global_step": 204928, "epoch": 2302, "lr": 1.4149507645129229e-05}
{"train_loss": 0.03657282143831253, "global_step": 204929, "epoch": 2302, "lr": 1.4149103534580994e-05}
{"train_loss": 0.028640462085604668, "global_step": 204930, "epoch": 2302, "lr": 1.4148699428852491e-05}
{"train_loss": 0.026478974148631096, "global_step": 204931, "epoch": 2302, "lr": 1.4148295327943745e-05}
{"train_loss": 0.05984354019165039, "global_step": 204932, "epoch": 2302, "lr": 1.4147891231854838e-05}
{"train_loss": 0.05628647655248642, "global_step": 204933, "epoch": 2302, "lr": 1.4147487140585808e-05}
{"train_loss": 0.06773771345615387, "global_step": 204934, "epoch": 2302, "lr": 1.4147083054136695e-05}
{"train_loss": 0.04232929274439812, "global_step": 204935, "epoch": 2302, "lr": 1.4146678972507588e-05}
{"train_loss": 0.0277620367705822, "global_step": 204936, "epoch": 2302, "lr": 1.4146274895698508e-05}
{"train_loss": 0.03661326691508293, "global_step": 204937, "epoch": 2302, "lr": 1.4145870823709533e-05}
{"train_loss": 0.05990521237254143, "global_step": 204938, "epoch": 2302, "lr": 1.4145466756540705e-05}
{"train_loss": 0.08037468045949936, "global_step": 204939, "epoch": 2302, "lr": 1.4145062694192085e-05}
{"train_loss": 0.043762434273958206, "global_step": 204940, "epoch": 2302, "lr": 1.4144658636663716e-05}
{"train_loss": 0.09143995493650436, "global_step": 204941, "epoch": 2302, "lr": 1.4144254583955675e-05}
{"train_loss": 0.08581523597240448, "global_step": 204942, "epoch": 2302, "lr": 1.4143850536067987e-05}
{"train_loss": 0.06312442570924759, "global_step": 204943, "epoch": 2302, "lr": 1.414344649300074e-05}
{"train_loss": 0.06783759593963623, "global_step": 204944, "epoch": 2302, "lr": 1.4143042454753946e-05}
{"train_loss": 0.04651877284049988, "global_step": 204945, "epoch": 2302, "lr": 1.4142638421327703e-05}
{"train_loss": 0.04595046862959862, "global_step": 204946, "epoch": 2302, "lr": 1.4142234392722031e-05}
{"train_loss": 0.05406063050031662, "global_step": 204947, "epoch": 2302, "lr": 1.4141830368937014e-05}
{"train_loss": 0.04733733832836151, "global_step": 204948, "epoch": 2302, "lr": 1.4141426349972675e-05}
{"train_loss": 0.052657246589660645, "global_step": 204949, "epoch": 2302, "lr": 1.4141022335829096e-05}
{"train_loss": 0.053342293947935104, "global_step": 204950, "epoch": 2302, "lr": 1.4140618326506306e-05}
{"train_loss": 0.02343793585896492, "global_step": 204951, "epoch": 2302, "lr": 1.4140214322004392e-05}
{"train_loss": 0.05515710264444351, "global_step": 204952, "epoch": 2302, "lr": 1.4139810322323366e-05}
{"train_loss": 0.03312798589468002, "global_step": 204953, "epoch": 2302, "lr": 1.4139406327463329e-05}
{"train_loss": 0.08861783146858215, "global_step": 204954, "epoch": 2302, "lr": 1.4139002337424289e-05}
{"train_loss": 0.03896763548254967, "global_step": 204955, "epoch": 2302, "lr": 1.4138598352206328e-05}
{"train_loss": 0.09812407195568085, "global_step": 204956, "epoch": 2302, "lr": 1.413819437180951e-05}
{"train_loss": 0.11372315883636475, "global_step": 204957, "epoch": 2302, "lr": 1.4137790396233857e-05}
{"train_loss": 0.0958297923207283, "global_step": 204958, "epoch": 2302, "lr": 1.413738642547946e-05}
{"train_loss": 0.05892084166407585, "global_step": 204959, "epoch": 2302, "lr": 1.4136982459546339e-05}
{"train_loss": 0.10445199906826019, "global_step": 204960, "epoch": 2302, "lr": 1.4136578498434577e-05}
{"train_loss": 0.06752189993858337, "global_step": 204961, "epoch": 2302, "lr": 1.4136174542144198e-05}
{"train_loss": 0.06539802998304367, "global_step": 204962, "epoch": 2302, "lr": 1.413577059067529e-05}
{"train_loss": 0.11050651222467422, "global_step": 204963, "epoch": 2302, "lr": 1.4135366644027875e-05}
{"train_loss": 0.06343826651573181, "global_step": 204964, "epoch": 2302, "lr": 1.4134962702202037e-05}
{"train_loss": 0.02263471856713295, "global_step": 204965, "epoch": 2302, "lr": 1.4134558765197803e-05}
{"train_loss": 0.06177465173970448, "global_step": 204966, "epoch": 2302, "lr": 1.4134154833015257e-05, "val_loss": 8.385100364685059}
{"train_loss": 0.1028052344918251, "global_step": 204967, "epoch": 2303, "lr": 1.4133750905654419e-05}
{"train_loss": 0.06410147249698639, "global_step": 204968, "epoch": 2303, "lr": 1.4133346983115375e-05}
{"train_loss": 0.04269684851169586, "global_step": 204969, "epoch": 2303, "lr": 1.4132943065398153e-05}
{"train_loss": 0.052991077303886414, "global_step": 204970, "epoch": 2303, "lr": 1.4132539152502833e-05}
{"train_loss": 0.10811460018157959, "global_step": 204971, "epoch": 2303, "lr": 1.4132135244429457e-05}
{"train_loss": 0.08142761141061783, "global_step": 204972, "epoch": 2303, "lr": 1.4131731341178061e-05}
{"train_loss": 0.12810084223747253, "global_step": 204973, "epoch": 2303, "lr": 1.4131327442748738e-05}
{"train_loss": 0.04877970367670059, "global_step": 204974, "epoch": 2303, "lr": 1.41309235491415e-05}
{"train_loss": 0.03088925965130329, "global_step": 204975, "epoch": 2303, "lr": 1.413051966035644e-05}
{"train_loss": 0.10142093151807785, "global_step": 204976, "epoch": 2303, "lr": 1.4130115776393577e-05}
{"train_loss": 0.05444585531949997, "global_step": 204977, "epoch": 2303, "lr": 1.4129711897252995e-05}
{"train_loss": 0.06665905565023422, "global_step": 204978, "epoch": 2303, "lr": 1.4129308022934729e-05}
{"train_loss": 0.049411989748477936, "global_step": 204979, "epoch": 2303, "lr": 1.4128904153438848e-05}
{"train_loss": 0.05631132796406746, "global_step": 204980, "epoch": 2303, "lr": 1.4128500288765389e-05}
{"train_loss": 0.05672995373606682, "global_step": 204981, "epoch": 2303, "lr": 1.4128096428914428e-05}
{"train_loss": 0.048624180257320404, "global_step": 204982, "epoch": 2303, "lr": 1.4127692573885993e-05}
{"train_loss": 0.0848119705915451, "global_step": 204983, "epoch": 2303, "lr": 1.4127288723680166e-05}
{"train_loss": 0.0810210332274437, "global_step": 204984, "epoch": 2303, "lr": 1.4126884878296975e-05}
{"train_loss": 0.10054588317871094, "global_step": 204985, "epoch": 2303, "lr": 1.41264810377365e-05}
{"train_loss": 0.09842908382415771, "global_step": 204986, "epoch": 2303, "lr": 1.4126077201998766e-05}
{"train_loss": 0.046144433319568634, "global_step": 204987, "epoch": 2303, "lr": 1.4125673371083864e-05}
{"train_loss": 0.025786707177758217, "global_step": 204988, "epoch": 2303, "lr": 1.412526954499181e-05}
{"train_loss": 0.0659593790769577, "global_step": 204989, "epoch": 2303, "lr": 1.412486572372269e-05}
{"train_loss": 0.03950176388025284, "global_step": 204990, "epoch": 2303, "lr": 1.4124461907276531e-05}
{"train_loss": 0.0252962876111269, "global_step": 204991, "epoch": 2303, "lr": 1.4124058095653402e-05}
{"train_loss": 0.039654336869716644, "global_step": 204992, "epoch": 2303, "lr": 1.4123654288853372e-05}
{"train_loss": 0.06545620411634445, "global_step": 204993, "epoch": 2303, "lr": 1.412325048687646e-05}
{"train_loss": 0.05426403507590294, "global_step": 204994, "epoch": 2303, "lr": 1.4122846689722758e-05}
{"train_loss": 0.0664881244301796, "global_step": 204995, "epoch": 2303, "lr": 1.4122442897392285e-05}
{"train_loss": 0.1451510787010193, "global_step": 204996, "epoch": 2303, "lr": 1.4122039109885127e-05}
{"train_loss": 0.049140915274620056, "global_step": 204997, "epoch": 2303, "lr": 1.412163532720131e-05}
{"train_loss": 0.06657936424016953, "global_step": 204998, "epoch": 2303, "lr": 1.4121231549340918e-05}
{"train_loss": 0.07415345311164856, "global_step": 204999, "epoch": 2303, "lr": 1.4120827776303974e-05}
{"train_loss": 0.07310128211975098, "global_step": 205000, "epoch": 2303, "lr": 1.412042400809056e-05}
{"train_loss": 0.06041903421282768, "global_step": 205001, "epoch": 2303, "lr": 1.4120020244700705e-05}
{"train_loss": 0.04529164358973503, "global_step": 205002, "epoch": 2303, "lr": 1.411961648613449e-05}
{"train_loss": 0.03579343482851982, "global_step": 205003, "epoch": 2303, "lr": 1.411921273239194e-05}
{"train_loss": 0.07055383175611496, "global_step": 205004, "epoch": 2303, "lr": 1.4118808983473136e-05}
{"train_loss": 0.05043020471930504, "global_step": 205005, "epoch": 2303, "lr": 1.4118405239378113e-05}
{"train_loss": 0.07924459874629974, "global_step": 205006, "epoch": 2303, "lr": 1.4118001500106941e-05}
{"train_loss": 0.08377739787101746, "global_step": 205007, "epoch": 2303, "lr": 1.4117597765659658e-05}
{"train_loss": 0.06145563721656799, "global_step": 205008, "epoch": 2303, "lr": 1.4117194036036335e-05}
{"train_loss": 0.034053657203912735, "global_step": 205009, "epoch": 2303, "lr": 1.4116790311237026e-05}
{"train_loss": 0.059037864208221436, "global_step": 205010, "epoch": 2303, "lr": 1.4116386591261755e-05}
{"train_loss": 0.03020709753036499, "global_step": 205011, "epoch": 2303, "lr": 1.4115982876110617e-05}
{"train_loss": 0.06769035011529922, "global_step": 205012, "epoch": 2303, "lr": 1.4115579165783633e-05}
{"train_loss": 0.11261157691478729, "global_step": 205013, "epoch": 2303, "lr": 1.4115175460280883e-05}
{"train_loss": 0.060058679431676865, "global_step": 205014, "epoch": 2303, "lr": 1.4114771759602403e-05}
{"train_loss": 0.0607185997068882, "global_step": 205015, "epoch": 2303, "lr": 1.4114368063748263e-05}
{"train_loss": 0.0632224753499031, "global_step": 205016, "epoch": 2303, "lr": 1.411396437271849e-05}
{"train_loss": 0.038203589618206024, "global_step": 205017, "epoch": 2303, "lr": 1.4113560686513182e-05}
{"train_loss": 0.023724079132080078, "global_step": 205018, "epoch": 2303, "lr": 1.4113157005132349e-05}
{"train_loss": 0.08040569722652435, "global_step": 205019, "epoch": 2303, "lr": 1.4112753328576079e-05}
{"train_loss": 0.06455963104963303, "global_step": 205020, "epoch": 2303, "lr": 1.4112349656844399e-05}
{"train_loss": 0.02724120393395424, "global_step": 205021, "epoch": 2303, "lr": 1.4111945989937386e-05}
{"train_loss": 0.042627908289432526, "global_step": 205022, "epoch": 2303, "lr": 1.4111542327855077e-05}
{"train_loss": 0.044898275285959244, "global_step": 205023, "epoch": 2303, "lr": 1.411113867059754e-05}
{"train_loss": 0.035712990909814835, "global_step": 205024, "epoch": 2303, "lr": 1.4110735018164817e-05}
{"train_loss": 0.04763535410165787, "global_step": 205025, "epoch": 2303, "lr": 1.4110331370556978e-05}
{"train_loss": 0.03813149780035019, "global_step": 205026, "epoch": 2303, "lr": 1.4109927727774053e-05}
{"train_loss": 0.07840574532747269, "global_step": 205027, "epoch": 2303, "lr": 1.4109524089816113e-05}
{"train_loss": 0.07188266515731812, "global_step": 205028, "epoch": 2303, "lr": 1.4109120456683223e-05}
{"train_loss": 0.07181955873966217, "global_step": 205029, "epoch": 2303, "lr": 1.4108716828375406e-05}
{"train_loss": 0.0857449322938919, "global_step": 205030, "epoch": 2303, "lr": 1.4108313204892754e-05}
{"train_loss": 0.07747501134872437, "global_step": 205031, "epoch": 2303, "lr": 1.4107909586235284e-05}
{"train_loss": 0.0778970941901207, "global_step": 205032, "epoch": 2303, "lr": 1.410750597240309e-05}
{"train_loss": 0.08561639487743378, "global_step": 205033, "epoch": 2303, "lr": 1.4107102363396185e-05}
{"train_loss": 0.0475027933716774, "global_step": 205034, "epoch": 2303, "lr": 1.4106698759214654e-05}
{"train_loss": 0.06703023612499237, "global_step": 205035, "epoch": 2303, "lr": 1.410629515985853e-05}
{"train_loss": 0.04177215322852135, "global_step": 205036, "epoch": 2303, "lr": 1.4105891565327889e-05}
{"train_loss": 0.08054739981889725, "global_step": 205037, "epoch": 2303, "lr": 1.410548797562276e-05}
{"train_loss": 0.016576012596488, "global_step": 205038, "epoch": 2303, "lr": 1.4105084390743228e-05}
{"train_loss": 0.06157386675477028, "global_step": 205039, "epoch": 2303, "lr": 1.4104680810689307e-05}
{"train_loss": 0.028173867613077164, "global_step": 205040, "epoch": 2303, "lr": 1.41042772354611e-05}
{"train_loss": 0.028211023658514023, "global_step": 205041, "epoch": 2303, "lr": 1.4103873665058608e-05}
{"train_loss": 0.02990780398249626, "global_step": 205042, "epoch": 2303, "lr": 1.4103470099481935e-05}
{"train_loss": 0.08061881363391876, "global_step": 205043, "epoch": 2303, "lr": 1.4103066538731102e-05}
{"train_loss": 0.07826357334852219, "global_step": 205044, "epoch": 2303, "lr": 1.410266298280618e-05}
{"train_loss": 0.07012949138879776, "global_step": 205045, "epoch": 2303, "lr": 1.410225943170721e-05}
{"train_loss": 0.01795867644250393, "global_step": 205046, "epoch": 2303, "lr": 1.4101855885434268e-05}
{"train_loss": 0.027476564049720764, "global_step": 205047, "epoch": 2303, "lr": 1.4101452343987375e-05}
{"train_loss": 0.08382310718297958, "global_step": 205048, "epoch": 2303, "lr": 1.4101048807366619e-05}
{"train_loss": 0.08809719979763031, "global_step": 205049, "epoch": 2303, "lr": 1.4100645275572044e-05}
{"train_loss": 0.043537527322769165, "global_step": 205050, "epoch": 2303, "lr": 1.4100241748603682e-05}
{"train_loss": 0.048468995839357376, "global_step": 205051, "epoch": 2303, "lr": 1.409983822646162e-05}
{"train_loss": 0.04154049977660179, "global_step": 205052, "epoch": 2303, "lr": 1.409943470914588e-05}
{"train_loss": 0.06451799720525742, "global_step": 205053, "epoch": 2303, "lr": 1.4099031196656547e-05}
{"train_loss": 0.16818739473819733, "global_step": 205054, "epoch": 2303, "lr": 1.4098627688993655e-05}
{"train_loss": 0.06219898501222723, "global_step": 205055, "epoch": 2303, "lr": 1.4098224186157271e-05, "val_loss": 8.415571212768555}
{"train_loss": 0.03208088502287865, "global_step": 205056, "epoch": 2304, "lr": 1.4097820688147434e-05}
{"train_loss": 0.010075491853058338, "global_step": 205057, "epoch": 2304, "lr": 1.409741719496422e-05}
{"train_loss": 0.05707168206572533, "global_step": 205058, "epoch": 2304, "lr": 1.4097013706607659e-05}
{"train_loss": 0.04304991662502289, "global_step": 205059, "epoch": 2304, "lr": 1.4096610223077828e-05}
{"train_loss": 0.04000008851289749, "global_step": 205060, "epoch": 2304, "lr": 1.4096206744374757e-05}
{"train_loss": 0.051604315638542175, "global_step": 205061, "epoch": 2304, "lr": 1.4095803270498525e-05}
{"train_loss": 0.08802127093076706, "global_step": 205062, "epoch": 2304, "lr": 1.4095399801449166e-05}
{"train_loss": 0.03961450234055519, "global_step": 205063, "epoch": 2304, "lr": 1.4094996337226751e-05}
{"train_loss": 0.06417939811944962, "global_step": 205064, "epoch": 2304, "lr": 1.4094592877831315e-05}
{"train_loss": 0.05438155680894852, "global_step": 205065, "epoch": 2304, "lr": 1.409418942326292e-05}
{"train_loss": 0.029482241719961166, "global_step": 205066, "epoch": 2304, "lr": 1.4093785973521646e-05}
{"train_loss": 0.09242940694093704, "global_step": 205067, "epoch": 2304, "lr": 1.4093382528607501e-05}
{"train_loss": 0.034681279212236404, "global_step": 205068, "epoch": 2304, "lr": 1.4092979088520586e-05}
{"train_loss": 0.053904611617326736, "global_step": 205069, "epoch": 2304, "lr": 1.4092575653260914e-05}
{"train_loss": 0.04476432874798775, "global_step": 205070, "epoch": 2304, "lr": 1.409217222282857e-05}
{"train_loss": 0.038486652076244354, "global_step": 205071, "epoch": 2304, "lr": 1.4091768797223582e-05}
{"train_loss": 0.0336046926677227, "global_step": 205072, "epoch": 2304, "lr": 1.4091365376446036e-05}
{"train_loss": 0.08342840522527695, "global_step": 205073, "epoch": 2304, "lr": 1.4090961960495953e-05}
{"train_loss": 0.07540708035230637, "global_step": 205074, "epoch": 2304, "lr": 1.4090558549373412e-05}
{"train_loss": 0.0518222339451313, "global_step": 205075, "epoch": 2304, "lr": 1.4090155143078449e-05}
{"train_loss": 0.06542368233203888, "global_step": 205076, "epoch": 2304, "lr": 1.408975174161114e-05}
{"train_loss": 0.0984518975019455, "global_step": 205077, "epoch": 2304, "lr": 1.4089348344971509e-05}
{"train_loss": 0.03120400197803974, "global_step": 205078, "epoch": 2304, "lr": 1.4088944953159643e-05}
{"train_loss": 0.0180943813174963, "global_step": 205079, "epoch": 2304, "lr": 1.4088541566175566e-05}
{"train_loss": 0.08705521374940872, "global_step": 205080, "epoch": 2304, "lr": 1.4088138184019367e-05}
{"train_loss": 0.08111537247896194, "global_step": 205081, "epoch": 2304, "lr": 1.4087734806691055e-05}
{"train_loss": 0.09388688951730728, "global_step": 205082, "epoch": 2304, "lr": 1.4087331434190732e-05}
{"train_loss": 0.04739604890346527, "global_step": 205083, "epoch": 2304, "lr": 1.4086928066518413e-05}
{"train_loss": 0.042583562433719635, "global_step": 205084, "epoch": 2304, "lr": 1.4086524703674187e-05}
{"train_loss": 0.09450457990169525, "global_step": 205085, "epoch": 2304, "lr": 1.408612134565807e-05}
{"train_loss": 0.07285229116678238, "global_step": 205086, "epoch": 2304, "lr": 1.4085717992470149e-05}
{"train_loss": 0.0606817901134491, "global_step": 205087, "epoch": 2304, "lr": 1.408531464411047e-05}
{"train_loss": 0.08600278943777084, "global_step": 205088, "epoch": 2304, "lr": 1.4084911300579069e-05}
{"train_loss": 0.048526663333177567, "global_step": 205089, "epoch": 2304, "lr": 1.4084507961876025e-05}
{"train_loss": 0.05456285551190376, "global_step": 205090, "epoch": 2304, "lr": 1.4084104628001371e-05}
{"train_loss": 0.11594732850790024, "global_step": 205091, "epoch": 2304, "lr": 1.4083701298955183e-05}
{"train_loss": 0.0822540670633316, "global_step": 205092, "epoch": 2304, "lr": 1.4083297974737485e-05}
{"train_loss": 0.07967831194400787, "global_step": 205093, "epoch": 2304, "lr": 1.4082894655348372e-05}
{"train_loss": 0.050221752375364304, "global_step": 205094, "epoch": 2304, "lr": 1.4082491340787856e-05}
{"train_loss": 0.05157416686415672, "global_step": 205095, "epoch": 2304, "lr": 1.4082088031056028e-05}
{"train_loss": 0.05896180123090744, "global_step": 205096, "epoch": 2304, "lr": 1.408168472615291e-05}
{"train_loss": 0.054041799157857895, "global_step": 205097, "epoch": 2304, "lr": 1.408128142607859e-05}
{"train_loss": 0.06182142719626427, "global_step": 205098, "epoch": 2304, "lr": 1.4080878130833086e-05}
{"train_loss": 0.05053218826651573, "global_step": 205099, "epoch": 2304, "lr": 1.4080474840416486e-05}
{"train_loss": 0.05553152784705162, "global_step": 205100, "epoch": 2304, "lr": 1.4080071554828812e-05}
{"train_loss": 0.05078084021806717, "global_step": 205101, "epoch": 2304, "lr": 1.4079668274070135e-05}
{"train_loss": 0.060352396219968796, "global_step": 205102, "epoch": 2304, "lr": 1.4079264998140524e-05}
{"train_loss": 0.027559855952858925, "global_step": 205103, "epoch": 2304, "lr": 1.4078861727040005e-05}
{"train_loss": 0.025954103097319603, "global_step": 205104, "epoch": 2304, "lr": 1.4078458460768657e-05}
{"train_loss": 0.025065777823328972, "global_step": 205105, "epoch": 2304, "lr": 1.4078055199326507e-05}
{"train_loss": 0.07769769430160522, "global_step": 205106, "epoch": 2304, "lr": 1.4077651942713643e-05}
{"train_loss": 0.06188090518116951, "global_step": 205107, "epoch": 2304, "lr": 1.4077248690930084e-05}
{"train_loss": 0.028833163902163506, "global_step": 205108, "epoch": 2304, "lr": 1.4076845443975916e-05}
{"train_loss": 0.058039214462041855, "global_step": 205109, "epoch": 2304, "lr": 1.4076442201851164e-05}
{"train_loss": 0.03943491727113724, "global_step": 205110, "epoch": 2304, "lr": 1.407603896455591e-05}
{"train_loss": 0.04200131446123123, "global_step": 205111, "epoch": 2304, "lr": 1.4075635732090182e-05}
{"train_loss": 0.08714888989925385, "global_step": 205112, "epoch": 2304, "lr": 1.4075232504454061e-05}
{"train_loss": 0.043249476701021194, "global_step": 205113, "epoch": 2304, "lr": 1.4074829281647573e-05}
{"train_loss": 0.05113937705755234, "global_step": 205114, "epoch": 2304, "lr": 1.4074426063670797e-05}
{"train_loss": 0.10156349837779999, "global_step": 205115, "epoch": 2304, "lr": 1.4074022850523766e-05}
{"train_loss": 0.09007250517606735, "global_step": 205116, "epoch": 2304, "lr": 1.407361964220656e-05}
{"train_loss": 0.04079383239150047, "global_step": 205117, "epoch": 2304, "lr": 1.40732164387192e-05}
{"train_loss": 0.08171916753053665, "global_step": 205118, "epoch": 2304, "lr": 1.407281324006176e-05}
{"train_loss": 0.04929535835981369, "global_step": 205119, "epoch": 2304, "lr": 1.4072410046234307e-05}
{"train_loss": 0.04976474493741989, "global_step": 205120, "epoch": 2304, "lr": 1.407200685723688e-05}
{"train_loss": 0.08266941457986832, "global_step": 205121, "epoch": 2304, "lr": 1.4071603673069523e-05}
{"train_loss": 0.09308094531297684, "global_step": 205122, "epoch": 2304, "lr": 1.4071200493732312e-05}
{"train_loss": 0.03544578701257706, "global_step": 205123, "epoch": 2304, "lr": 1.407079731922527e-05}
{"train_loss": 0.10477633029222488, "global_step": 205124, "epoch": 2304, "lr": 1.4070394149548493e-05}
{"train_loss": 0.049382828176021576, "global_step": 205125, "epoch": 2304, "lr": 1.4069990984702014e-05}
{"train_loss": 0.03325040265917778, "global_step": 205126, "epoch": 2304, "lr": 1.406958782468587e-05}
{"train_loss": 0.05493245646357536, "global_step": 205127, "epoch": 2304, "lr": 1.4069184669500146e-05}
{"train_loss": 0.0689498633146286, "global_step": 205128, "epoch": 2304, "lr": 1.4068781519144868e-05}
{"train_loss": 0.045411717146635056, "global_step": 205129, "epoch": 2304, "lr": 1.4068378373620123e-05}
{"train_loss": 0.09787439554929733, "global_step": 205130, "epoch": 2304, "lr": 1.4067975232925929e-05}
{"train_loss": 0.09919285029172897, "global_step": 205131, "epoch": 2304, "lr": 1.406757209706237e-05}
{"train_loss": 0.04833870753645897, "global_step": 205132, "epoch": 2304, "lr": 1.4067168966029476e-05}
{"train_loss": 0.051758673042058945, "global_step": 205133, "epoch": 2304, "lr": 1.4066765839827328e-05}
{"train_loss": 0.05686400458216667, "global_step": 205134, "epoch": 2304, "lr": 1.406636271845595e-05}
{"train_loss": 0.08186200261116028, "global_step": 205135, "epoch": 2304, "lr": 1.4065959601915424e-05}
{"train_loss": 0.04040531441569328, "global_step": 205136, "epoch": 2304, "lr": 1.4065556490205777e-05}
{"train_loss": 0.08218298107385635, "global_step": 205137, "epoch": 2304, "lr": 1.4065153383327085e-05}
{"train_loss": 0.10089574754238129, "global_step": 205138, "epoch": 2304, "lr": 1.4064750281279404e-05}
{"train_loss": 0.0849265307188034, "global_step": 205139, "epoch": 2304, "lr": 1.4064347184062765e-05}
{"train_loss": 0.06592787057161331, "global_step": 205140, "epoch": 2304, "lr": 1.406394409167725e-05}
{"train_loss": 0.03594151511788368, "global_step": 205141, "epoch": 2304, "lr": 1.4063541004122888e-05}
{"train_loss": 0.04052601009607315, "global_step": 205142, "epoch": 2304, "lr": 1.406313792139976e-05}
{"train_loss": 0.06923835724592209, "global_step": 205143, "epoch": 2304, "lr": 1.4062734843507886e-05}
{"train_loss": 0.05995595770145065, "global_step": 205144, "epoch": 2304, "lr": 1.4062331770447356e-05, "val_loss": 8.233586311340332}
{"train_loss": 0.0657256618142128, "global_step": 205145, "epoch": 2305, "lr": 1.4061928702218198e-05}
{"train_loss": 0.06482986360788345, "global_step": 205146, "epoch": 2305, "lr": 1.4061525638820483e-05}
{"train_loss": 0.03847578912973404, "global_step": 205147, "epoch": 2305, "lr": 1.4061122580254249e-05}
{"train_loss": 0.09129640460014343, "global_step": 205148, "epoch": 2305, "lr": 1.4060719526519567e-05}
{"train_loss": 0.07517439126968384, "global_step": 205149, "epoch": 2305, "lr": 1.4060316477616475e-05}
{"train_loss": 0.060096170753240585, "global_step": 205150, "epoch": 2305, "lr": 1.4059913433545047e-05}
{"train_loss": 0.0686197280883789, "global_step": 205151, "epoch": 2305, "lr": 1.4059510394305315e-05}
{"train_loss": 0.09358488023281097, "global_step": 205152, "epoch": 2305, "lr": 1.405910735989735e-05}
{"train_loss": 0.026332912966609, "global_step": 205153, "epoch": 2305, "lr": 1.4058704330321193e-05}
{"train_loss": 0.07116398215293884, "global_step": 205154, "epoch": 2305, "lr": 1.4058301305576899e-05}
{"train_loss": 0.09466711431741714, "global_step": 205155, "epoch": 2305, "lr": 1.4057898285664551e-05}
{"train_loss": 0.041622839868068695, "global_step": 205156, "epoch": 2305, "lr": 1.4057495270584158e-05}
{"train_loss": 0.0965583473443985, "global_step": 205157, "epoch": 2305, "lr": 1.4057092260335813e-05}
{"train_loss": 0.06899349391460419, "global_step": 205158, "epoch": 2305, "lr": 1.4056689254919541e-05}
{"train_loss": 0.0797000601887703, "global_step": 205159, "epoch": 2305, "lr": 1.4056286254335427e-05}
{"train_loss": 0.024614514783024788, "global_step": 205160, "epoch": 2305, "lr": 1.4055883258583503e-05}
{"train_loss": 0.10142546892166138, "global_step": 205161, "epoch": 2305, "lr": 1.4055480267663812e-05}
{"train_loss": 0.04917639121413231, "global_step": 205162, "epoch": 2305, "lr": 1.405507728157644e-05}
{"train_loss": 0.07405155152082443, "global_step": 205163, "epoch": 2305, "lr": 1.4054674300321408e-05}
{"train_loss": 0.07481427490711212, "global_step": 205164, "epoch": 2305, "lr": 1.4054271323898805e-05}
{"train_loss": 0.04957727715373039, "global_step": 205165, "epoch": 2305, "lr": 1.4053868352308664e-05}
{"train_loss": 0.09387025982141495, "global_step": 205166, "epoch": 2305, "lr": 1.4053465385551028e-05}
{"train_loss": 0.061277151107788086, "global_step": 205167, "epoch": 2305, "lr": 1.4053062423625985e-05}
{"train_loss": 0.07197434455156326, "global_step": 205168, "epoch": 2305, "lr": 1.4052659466533546e-05}
{"train_loss": 0.04782624542713165, "global_step": 205169, "epoch": 2305, "lr": 1.405225651427381e-05}
{"train_loss": 0.053910061717033386, "global_step": 205170, "epoch": 2305, "lr": 1.4051853566846796e-05}
{"train_loss": 0.07343697547912598, "global_step": 205171, "epoch": 2305, "lr": 1.4051450624252582e-05}
{"train_loss": 0.0665605440735817, "global_step": 205172, "epoch": 2305, "lr": 1.4051047686491203e-05}
{"train_loss": 0.05012792721390724, "global_step": 205173, "epoch": 2305, "lr": 1.405064475356272e-05}
{"train_loss": 0.043474599719047546, "global_step": 205174, "epoch": 2305, "lr": 1.4050241825467202e-05}
{"train_loss": 0.10166075825691223, "global_step": 205175, "epoch": 2305, "lr": 1.4049838902204677e-05}
{"train_loss": 0.04264167323708534, "global_step": 205176, "epoch": 2305, "lr": 1.4049435983775233e-05}
{"train_loss": 0.04509272426366806, "global_step": 205177, "epoch": 2305, "lr": 1.4049033070178884e-05}
{"train_loss": 0.055904705077409744, "global_step": 205178, "epoch": 2305, "lr": 1.4048630161415721e-05}
{"train_loss": 0.08567260205745697, "global_step": 205179, "epoch": 2305, "lr": 1.4048227257485764e-05}
{"train_loss": 0.049704816192388535, "global_step": 205180, "epoch": 2305, "lr": 1.40478243583891e-05}
{"train_loss": 0.04821639880537987, "global_step": 205181, "epoch": 2305, "lr": 1.4047421464125753e-05}
{"train_loss": 0.043915875256061554, "global_step": 205182, "epoch": 2305, "lr": 1.4047018574695814e-05}
{"train_loss": 0.0672982707619667, "global_step": 205183, "epoch": 2305, "lr": 1.404661569009929e-05}
{"train_loss": 0.04857689514756203, "global_step": 205184, "epoch": 2305, "lr": 1.4046212810336284e-05}
{"train_loss": 0.048212964087724686, "global_step": 205185, "epoch": 2305, "lr": 1.4045809935406807e-05}
{"train_loss": 0.03202343359589577, "global_step": 205186, "epoch": 2305, "lr": 1.4045407065310945e-05}
{"train_loss": 0.030143385753035545, "global_step": 205187, "epoch": 2305, "lr": 1.4045004200048727e-05}
{"train_loss": 0.038272783160209656, "global_step": 205188, "epoch": 2305, "lr": 1.4044601339620239e-05}
{"train_loss": 0.04074733704328537, "global_step": 205189, "epoch": 2305, "lr": 1.4044198484025495e-05}
{"train_loss": 0.07004717737436295, "global_step": 205190, "epoch": 2305, "lr": 1.404379563326459e-05}
{"train_loss": 0.03177294507622719, "global_step": 205191, "epoch": 2305, "lr": 1.404339278733754e-05}
{"train_loss": 0.08487568795681, "global_step": 205192, "epoch": 2305, "lr": 1.4042989946244423e-05}
{"train_loss": 0.0445544458925724, "global_step": 205193, "epoch": 2305, "lr": 1.4042587109985295e-05}
{"train_loss": 0.0819360688328743, "global_step": 205194, "epoch": 2305, "lr": 1.4042184278560194e-05}
{"train_loss": 0.04546119645237923, "global_step": 205195, "epoch": 2305, "lr": 1.4041781451969194e-05}
{"train_loss": 0.029916469007730484, "global_step": 205196, "epoch": 2305, "lr": 1.4041378630212325e-05}
{"train_loss": 0.03664753586053848, "global_step": 205197, "epoch": 2305, "lr": 1.404097581328967e-05}
{"train_loss": 0.07207086682319641, "global_step": 205198, "epoch": 2305, "lr": 1.4040573001201267e-05}
{"train_loss": 0.042032621800899506, "global_step": 205199, "epoch": 2305, "lr": 1.4040170193947156e-05}
{"train_loss": 0.059602897614240646, "global_step": 205200, "epoch": 2305, "lr": 1.403976739152742e-05}
{"train_loss": 0.09675689041614532, "global_step": 205201, "epoch": 2305, "lr": 1.4039364593942084e-05}
{"train_loss": 0.037706490606069565, "global_step": 205202, "epoch": 2305, "lr": 1.4038961801191237e-05}
{"train_loss": 0.06078745797276497, "global_step": 205203, "epoch": 2305, "lr": 1.4038559013274905e-05}
{"train_loss": 0.1353549361228943, "global_step": 205204, "epoch": 2305, "lr": 1.4038156230193138e-05}
{"train_loss": 0.03209736943244934, "global_step": 205205, "epoch": 2305, "lr": 1.4037753451946023e-05}
{"train_loss": 0.0493779256939888, "global_step": 205206, "epoch": 2305, "lr": 1.4037350678533573e-05}
{"train_loss": 0.05369899421930313, "global_step": 205207, "epoch": 2305, "lr": 1.4036947909955877e-05}
{"train_loss": 0.07677178084850311, "global_step": 205208, "epoch": 2305, "lr": 1.4036545146212966e-05}
{"train_loss": 0.12933407723903656, "global_step": 205209, "epoch": 2305, "lr": 1.4036142387304901e-05}
{"train_loss": 0.08602621406316757, "global_step": 205210, "epoch": 2305, "lr": 1.4035739633231748e-05}
{"train_loss": 0.05007877200841904, "global_step": 205211, "epoch": 2305, "lr": 1.4035336883993544e-05}
{"train_loss": 0.05307517945766449, "global_step": 205212, "epoch": 2305, "lr": 1.403493413959036e-05}
{"train_loss": 0.059723593294620514, "global_step": 205213, "epoch": 2305, "lr": 1.4034531400022227e-05}
{"train_loss": 0.04746006056666374, "global_step": 205214, "epoch": 2305, "lr": 1.4034128665289231e-05}
{"train_loss": 0.10028159618377686, "global_step": 205215, "epoch": 2305, "lr": 1.403372593539139e-05}
{"train_loss": 0.05125688016414642, "global_step": 205216, "epoch": 2305, "lr": 1.4033323210328792e-05}
{"train_loss": 0.07211153209209442, "global_step": 205217, "epoch": 2305, "lr": 1.403292049010146e-05}
{"train_loss": 0.06474417448043823, "global_step": 205218, "epoch": 2305, "lr": 1.4032517774709481e-05}
{"train_loss": 0.051911044865846634, "global_step": 205219, "epoch": 2305, "lr": 1.4032115064152873e-05}
{"train_loss": 0.06904454529285431, "global_step": 205220, "epoch": 2305, "lr": 1.4031712358431725e-05}
{"train_loss": 0.06819996237754822, "global_step": 205221, "epoch": 2305, "lr": 1.403130965754606e-05}
{"train_loss": 0.03769323602318764, "global_step": 205222, "epoch": 2305, "lr": 1.403090696149596e-05}
{"train_loss": 0.11497629433870316, "global_step": 205223, "epoch": 2305, "lr": 1.4030504270281452e-05}
{"train_loss": 0.05358053371310234, "global_step": 205224, "epoch": 2305, "lr": 1.4030101583902622e-05}
{"train_loss": 0.07838486135005951, "global_step": 205225, "epoch": 2305, "lr": 1.402969890235949e-05}
{"train_loss": 0.033091623336076736, "global_step": 205226, "epoch": 2305, "lr": 1.4029296225652144e-05}
{"train_loss": 0.10141296684741974, "global_step": 205227, "epoch": 2305, "lr": 1.4028893553780603e-05}
{"train_loss": 0.06096427142620087, "global_step": 205228, "epoch": 2305, "lr": 1.4028490886744939e-05}
{"train_loss": 0.04086672142148018, "global_step": 205229, "epoch": 2305, "lr": 1.4028088224545222e-05}
{"train_loss": 0.043051350861787796, "global_step": 205230, "epoch": 2305, "lr": 1.402768556718147e-05}
{"train_loss": 0.04053715243935585, "global_step": 205231, "epoch": 2305, "lr": 1.4027282914653777e-05}
{"train_loss": 0.049380019307136536, "global_step": 205232, "epoch": 2305, "lr": 1.402688026696216e-05}
{"train_loss": 0.0623297548821468, "global_step": 205233, "epoch": 2305, "lr": 1.402647762410671e-05, "val_loss": 8.5574312210083, "train_action_mse_error": 21.017728805541992}
{"train_loss": 0.12079519778490067, "global_step": 205234, "epoch": 2306, "lr": 1.4026074986087446e-05}
{"train_loss": 0.043796107172966, "global_step": 205235, "epoch": 2306, "lr": 1.4025672352904445e-05}
{"train_loss": 0.08119592815637589, "global_step": 205236, "epoch": 2306, "lr": 1.402526972455776e-05}
{"train_loss": 0.0421789176762104, "global_step": 205237, "epoch": 2306, "lr": 1.4024867101047424e-05}
{"train_loss": 0.028859583660960197, "global_step": 205238, "epoch": 2306, "lr": 1.4024464482373523e-05}
{"train_loss": 0.06575526297092438, "global_step": 205239, "epoch": 2306, "lr": 1.4024061868536076e-05}
{"train_loss": 0.06814621388912201, "global_step": 205240, "epoch": 2306, "lr": 1.4023659259535171e-05}
{"train_loss": 0.06378709524869919, "global_step": 205241, "epoch": 2306, "lr": 1.4023256655370848e-05}
{"train_loss": 0.0694168210029602, "global_step": 205242, "epoch": 2306, "lr": 1.4022854056043144e-05}
{"train_loss": 0.08247942477464676, "global_step": 205243, "epoch": 2306, "lr": 1.4022451461552144e-05}
{"train_loss": 0.05840340256690979, "global_step": 205244, "epoch": 2306, "lr": 1.4022048871897869e-05}
{"train_loss": 0.06884277611970901, "global_step": 205245, "epoch": 2306, "lr": 1.4021646287080391e-05}
{"train_loss": 0.11062236875295639, "global_step": 205246, "epoch": 2306, "lr": 1.4021243707099786e-05}
{"train_loss": 0.0876869335770607, "global_step": 205247, "epoch": 2306, "lr": 1.4020841131956065e-05}
{"train_loss": 0.03609062731266022, "global_step": 205248, "epoch": 2306, "lr": 1.4020438561649319e-05}
{"train_loss": 0.1017698273062706, "global_step": 205249, "epoch": 2306, "lr": 1.402003599617957e-05}
{"train_loss": 0.05462675914168358, "global_step": 205250, "epoch": 2306, "lr": 1.4019633435546909e-05}
{"train_loss": 0.09241530299186707, "global_step": 205251, "epoch": 2306, "lr": 1.401923087975135e-05}
{"train_loss": 0.027495993301272392, "global_step": 205252, "epoch": 2306, "lr": 1.4018828328792983e-05}
{"train_loss": 0.04213326424360275, "global_step": 205253, "epoch": 2306, "lr": 1.4018425782671834e-05}
{"train_loss": 0.034433513879776, "global_step": 205254, "epoch": 2306, "lr": 1.401802324138798e-05}
{"train_loss": 0.04645581170916557, "global_step": 205255, "epoch": 2306, "lr": 1.4017620704941453e-05}
{"train_loss": 0.041301827877759933, "global_step": 205256, "epoch": 2306, "lr": 1.401721817333233e-05}
{"train_loss": 0.044721364974975586, "global_step": 205257, "epoch": 2306, "lr": 1.4016815646560639e-05}
{"train_loss": 0.11516465246677399, "global_step": 205258, "epoch": 2306, "lr": 1.4016413124626465e-05}
{"train_loss": 0.08112072944641113, "global_step": 205259, "epoch": 2306, "lr": 1.4016010607529828e-05}
{"train_loss": 0.024655045941472054, "global_step": 205260, "epoch": 2306, "lr": 1.401560809527082e-05}
{"train_loss": 0.08275213092565536, "global_step": 205261, "epoch": 2306, "lr": 1.4015205587849456e-05}
{"train_loss": 0.040327757596969604, "global_step": 205262, "epoch": 2306, "lr": 1.4014803085265826e-05}
{"train_loss": 0.06850526481866837, "global_step": 205263, "epoch": 2306, "lr": 1.401440058751995e-05}
{"train_loss": 0.044101692736148834, "global_step": 205264, "epoch": 2306, "lr": 1.4013998094611896e-05}
{"train_loss": 0.06017270311713219, "global_step": 205265, "epoch": 2306, "lr": 1.401359560654174e-05}
{"train_loss": 0.06803703308105469, "global_step": 205266, "epoch": 2306, "lr": 1.4013193123309504e-05}
{"train_loss": 0.09034930169582367, "global_step": 205267, "epoch": 2306, "lr": 1.4012790644915268e-05}
{"train_loss": 0.08108875155448914, "global_step": 205268, "epoch": 2306, "lr": 1.4012388171359065e-05}
{"train_loss": 0.05561162158846855, "global_step": 205269, "epoch": 2306, "lr": 1.4011985702640967e-05}
{"train_loss": 0.11504503339529037, "global_step": 205270, "epoch": 2306, "lr": 1.4011583238761005e-05}
{"train_loss": 0.0847783163189888, "global_step": 205271, "epoch": 2306, "lr": 1.4011180779719264e-05}
{"train_loss": 0.13166719675064087, "global_step": 205272, "epoch": 2306, "lr": 1.401077832551576e-05}
{"train_loss": 0.03163132444024086, "global_step": 205273, "epoch": 2306, "lr": 1.4010375876150588e-05}
{"train_loss": 0.05712105333805084, "global_step": 205274, "epoch": 2306, "lr": 1.4009973431623769e-05}
{"train_loss": 0.05081351101398468, "global_step": 205275, "epoch": 2306, "lr": 1.4009570991935378e-05}
{"train_loss": 0.10236912965774536, "global_step": 205276, "epoch": 2306, "lr": 1.4009168557085467e-05}
{"train_loss": 0.06213799864053726, "global_step": 205277, "epoch": 2306, "lr": 1.4008766127074069e-05}
{"train_loss": 0.06722591817378998, "global_step": 205278, "epoch": 2306, "lr": 1.4008363701901272e-05}
{"train_loss": 0.07836464047431946, "global_step": 205279, "epoch": 2306, "lr": 1.400796128156709e-05}
{"train_loss": 0.08051258325576782, "global_step": 205280, "epoch": 2306, "lr": 1.4007558866071618e-05}
{"train_loss": 0.045366231352090836, "global_step": 205281, "epoch": 2306, "lr": 1.4007156455414877e-05}
{"train_loss": 0.01896977610886097, "global_step": 205282, "epoch": 2306, "lr": 1.4006754049596949e-05}
{"train_loss": 0.0486624650657177, "global_step": 205283, "epoch": 2306, "lr": 1.4006351648617854e-05}
{"train_loss": 0.025076834484934807, "global_step": 205284, "epoch": 2306, "lr": 1.400594925247769e-05}
{"train_loss": 0.03765684738755226, "global_step": 205285, "epoch": 2306, "lr": 1.4005546861176465e-05}
{"train_loss": 0.10227204859256744, "global_step": 205286, "epoch": 2306, "lr": 1.4005144474714277e-05}
{"train_loss": 0.016462884843349457, "global_step": 205287, "epoch": 2306, "lr": 1.4004742093091138e-05}
{"train_loss": 0.07817981392145157, "global_step": 205288, "epoch": 2306, "lr": 1.4004339716307141e-05}
{"train_loss": 0.041128408163785934, "global_step": 205289, "epoch": 2306, "lr": 1.4003937344362306e-05}
{"train_loss": 0.03815890848636627, "global_step": 205290, "epoch": 2306, "lr": 1.4003534977256721e-05}
{"train_loss": 0.06272785365581512, "global_step": 205291, "epoch": 2306, "lr": 1.4003132614990405e-05}
{"train_loss": 0.061129067093133926, "global_step": 205292, "epoch": 2306, "lr": 1.400273025756344e-05}
{"train_loss": 0.08809170871973038, "global_step": 205293, "epoch": 2306, "lr": 1.400232790497586e-05}
{"train_loss": 0.07016465812921524, "global_step": 205294, "epoch": 2306, "lr": 1.400192555722774e-05}
{"train_loss": 0.113557830452919, "global_step": 205295, "epoch": 2306, "lr": 1.4001523214319106e-05}
{"train_loss": 0.06148247793316841, "global_step": 205296, "epoch": 2306, "lr": 1.4001120876250046e-05}
{"train_loss": 0.03454829007387161, "global_step": 205297, "epoch": 2306, "lr": 1.400071854302058e-05}
{"train_loss": 0.07758964598178864, "global_step": 205298, "epoch": 2306, "lr": 1.4000316214630792e-05}
{"train_loss": 0.054022859781980515, "global_step": 205299, "epoch": 2306, "lr": 1.3999913891080712e-05}
{"train_loss": 0.06147311255335808, "global_step": 205300, "epoch": 2306, "lr": 1.3999511572370405e-05}
{"train_loss": 0.06726530939340591, "global_step": 205301, "epoch": 2306, "lr": 1.3999109258499936e-05}
{"train_loss": 0.040691569447517395, "global_step": 205302, "epoch": 2306, "lr": 1.3998706949469338e-05}
{"train_loss": 0.04250652715563774, "global_step": 205303, "epoch": 2306, "lr": 1.3998304645278687e-05}
{"train_loss": 0.051673926413059235, "global_step": 205304, "epoch": 2306, "lr": 1.399790234592801e-05}
{"train_loss": 0.05617402866482735, "global_step": 205305, "epoch": 2306, "lr": 1.399750005141739e-05}
{"train_loss": 0.04970788210630417, "global_step": 205306, "epoch": 2306, "lr": 1.3997097761746852e-05}
{"train_loss": 0.04469113051891327, "global_step": 205307, "epoch": 2306, "lr": 1.3996695476916482e-05}
{"train_loss": 0.07237669080495834, "global_step": 205308, "epoch": 2306, "lr": 1.3996293196926303e-05}
{"train_loss": 0.06451424956321716, "global_step": 205309, "epoch": 2306, "lr": 1.39958909217764e-05}
{"train_loss": 0.0587625615298748, "global_step": 205310, "epoch": 2306, "lr": 1.3995488651466793e-05}
{"train_loss": 0.0795140266418457, "global_step": 205311, "epoch": 2306, "lr": 1.3995086385997569e-05}
{"train_loss": 0.049365874379873276, "global_step": 205312, "epoch": 2306, "lr": 1.3994684125368751e-05}
{"train_loss": 0.07182803004980087, "global_step": 205313, "epoch": 2306, "lr": 1.3994281869580428e-05}
{"train_loss": 0.0551178902387619, "global_step": 205314, "epoch": 2306, "lr": 1.3993879618632626e-05}
{"train_loss": 0.09512858837842941, "global_step": 205315, "epoch": 2306, "lr": 1.3993477372525405e-05}
{"train_loss": 0.0688665583729744, "global_step": 205316, "epoch": 2306, "lr": 1.399307513125883e-05}
{"train_loss": 0.07315133512020111, "global_step": 205317, "epoch": 2306, "lr": 1.3992672894832932e-05}
{"train_loss": 0.04546134173870087, "global_step": 205318, "epoch": 2306, "lr": 1.3992270663247797e-05}
{"train_loss": 0.08575984835624695, "global_step": 205319, "epoch": 2306, "lr": 1.3991868436503447e-05}
{"train_loss": 0.04447166994214058, "global_step": 205320, "epoch": 2306, "lr": 1.3991466214599968e-05}
{"train_loss": 0.03962548449635506, "global_step": 205321, "epoch": 2306, "lr": 1.399106399753738e-05}
{"train_loss": 0.06303088392099637, "global_step": 205322, "epoch": 2306, "lr": 1.3990661785315767e-05, "val_loss": 8.3986234664917}
{"train_loss": 0.054786019027233124, "global_step": 205323, "epoch": 2307, "lr": 1.399025957793516e-05}
{"train_loss": 0.0564887672662735, "global_step": 205324, "epoch": 2307, "lr": 1.3989857375395638e-05}
{"train_loss": 0.060023508965969086, "global_step": 205325, "epoch": 2307, "lr": 1.3989455177697225e-05}
{"train_loss": 0.05315318703651428, "global_step": 205326, "epoch": 2307, "lr": 1.3989052984840007e-05}
{"train_loss": 0.062149468809366226, "global_step": 205327, "epoch": 2307, "lr": 1.3988650796824004e-05}
{"train_loss": 0.05639426037669182, "global_step": 205328, "epoch": 2307, "lr": 1.3988248613649308e-05}
{"train_loss": 0.06939581036567688, "global_step": 205329, "epoch": 2307, "lr": 1.3987846435315938e-05}
{"train_loss": 0.048267804086208344, "global_step": 205330, "epoch": 2307, "lr": 1.3987444261823974e-05}
{"train_loss": 0.055293161422014236, "global_step": 205331, "epoch": 2307, "lr": 1.3987042093173441e-05}
{"train_loss": 0.08447308093309402, "global_step": 205332, "epoch": 2307, "lr": 1.3986639929364432e-05}
{"train_loss": 0.0676003098487854, "global_step": 205333, "epoch": 2307, "lr": 1.3986237770396965e-05}
{"train_loss": 0.04214882105588913, "global_step": 205334, "epoch": 2307, "lr": 1.3985835616271126e-05}
{"train_loss": 0.044164977967739105, "global_step": 205335, "epoch": 2307, "lr": 1.3985433466986937e-05}
{"train_loss": 0.06181071698665619, "global_step": 205336, "epoch": 2307, "lr": 1.3985031322544462e-05}
{"train_loss": 0.051693882793188095, "global_step": 205337, "epoch": 2307, "lr": 1.3984629182943781e-05}
{"train_loss": 0.07271081954240799, "global_step": 205338, "epoch": 2307, "lr": 1.3984227048184905e-05}
{"train_loss": 0.07816073298454285, "global_step": 205339, "epoch": 2307, "lr": 1.3983824918267934e-05}
{"train_loss": 0.07951080054044724, "global_step": 205340, "epoch": 2307, "lr": 1.398342279319288e-05}
{"train_loss": 0.047240037471055984, "global_step": 205341, "epoch": 2307, "lr": 1.3983020672959834e-05}
{"train_loss": 0.05943030118942261, "global_step": 205342, "epoch": 2307, "lr": 1.3982618557568816e-05}
{"train_loss": 0.025807354599237442, "global_step": 205343, "epoch": 2307, "lr": 1.3982216447019908e-05}
{"train_loss": 0.10985822975635529, "global_step": 205344, "epoch": 2307, "lr": 1.3981814341313138e-05}
{"train_loss": 0.0657380148768425, "global_step": 205345, "epoch": 2307, "lr": 1.3981412240448594e-05}
{"train_loss": 0.04376284033060074, "global_step": 205346, "epoch": 2307, "lr": 1.3981010144426288e-05}
{"train_loss": 0.051723264157772064, "global_step": 205347, "epoch": 2307, "lr": 1.3980608053246314e-05}
{"train_loss": 0.05462288483977318, "global_step": 205348, "epoch": 2307, "lr": 1.3980205966908693e-05}
{"train_loss": 0.04401706904172897, "global_step": 205349, "epoch": 2307, "lr": 1.3979803885413512e-05}
{"train_loss": 0.06156237795948982, "global_step": 205350, "epoch": 2307, "lr": 1.397940180876079e-05}
{"train_loss": 0.03179354965686798, "global_step": 205351, "epoch": 2307, "lr": 1.397899973695061e-05}
{"train_loss": 0.10927055776119232, "global_step": 205352, "epoch": 2307, "lr": 1.3978597669983018e-05}
{"train_loss": 0.08081194758415222, "global_step": 205353, "epoch": 2307, "lr": 1.3978195607858052e-05}
{"train_loss": 0.08937551081180573, "global_step": 205354, "epoch": 2307, "lr": 1.3977793550575791e-05}
{"train_loss": 0.05203096196055412, "global_step": 205355, "epoch": 2307, "lr": 1.3977391498136261e-05}
{"train_loss": 0.09921727329492569, "global_step": 205356, "epoch": 2307, "lr": 1.3976989450539545e-05}
{"train_loss": 0.02997300960123539, "global_step": 205357, "epoch": 2307, "lr": 1.3976587407785675e-05}
{"train_loss": 0.09869188070297241, "global_step": 205358, "epoch": 2307, "lr": 1.3976185369874723e-05}
{"train_loss": 0.07727059721946716, "global_step": 205359, "epoch": 2307, "lr": 1.397578333680672e-05}
{"train_loss": 0.09520771354436874, "global_step": 205360, "epoch": 2307, "lr": 1.3975381308581748e-05}
{"train_loss": 0.04967857152223587, "global_step": 205361, "epoch": 2307, "lr": 1.3974979285199836e-05}
{"train_loss": 0.049111392349004745, "global_step": 205362, "epoch": 2307, "lr": 1.397457726666106e-05}
{"train_loss": 0.028041889891028404, "global_step": 205363, "epoch": 2307, "lr": 1.3974175252965449e-05}
{"train_loss": 0.0341232605278492, "global_step": 205364, "epoch": 2307, "lr": 1.3973773244113087e-05}
{"train_loss": 0.08697670698165894, "global_step": 205365, "epoch": 2307, "lr": 1.3973371240103994e-05}
{"train_loss": 0.06807731091976166, "global_step": 205366, "epoch": 2307, "lr": 1.3972969240938261e-05}
{"train_loss": 0.11372653394937515, "global_step": 205367, "epoch": 2307, "lr": 1.3972567246615908e-05}
{"train_loss": 0.053377665579319, "global_step": 205368, "epoch": 2307, "lr": 1.3972165257137015e-05}
{"train_loss": 0.0713360607624054, "global_step": 205369, "epoch": 2307, "lr": 1.3971763272501615e-05}
{"train_loss": 0.06385351717472076, "global_step": 205370, "epoch": 2307, "lr": 1.3971361292709783e-05}
{"train_loss": 0.05754623934626579, "global_step": 205371, "epoch": 2307, "lr": 1.397095931776155e-05}
{"train_loss": 0.0754055604338646, "global_step": 205372, "epoch": 2307, "lr": 1.3970557347656983e-05}
{"train_loss": 0.08129271119832993, "global_step": 205373, "epoch": 2307, "lr": 1.397015538239615e-05}
{"train_loss": 0.04637722671031952, "global_step": 205374, "epoch": 2307, "lr": 1.3969753421979076e-05}
{"train_loss": 0.08120675384998322, "global_step": 205375, "epoch": 2307, "lr": 1.3969351466405844e-05}
{"train_loss": 0.06778193265199661, "global_step": 205376, "epoch": 2307, "lr": 1.3968949515676477e-05}
{"train_loss": 0.0731281116604805, "global_step": 205377, "epoch": 2307, "lr": 1.3968547569791063e-05}
{"train_loss": 0.07372181117534637, "global_step": 205378, "epoch": 2307, "lr": 1.3968145628749623e-05}
{"train_loss": 0.0697428360581398, "global_step": 205379, "epoch": 2307, "lr": 1.3967743692552238e-05}
{"train_loss": 0.12325403094291687, "global_step": 205380, "epoch": 2307, "lr": 1.3967341761198943e-05}
{"train_loss": 0.04031297191977501, "global_step": 205381, "epoch": 2307, "lr": 1.3966939834689807e-05}
{"train_loss": 0.10701754689216614, "global_step": 205382, "epoch": 2307, "lr": 1.3966537913024868e-05}
{"train_loss": 0.0927657037973404, "global_step": 205383, "epoch": 2307, "lr": 1.39661359962042e-05}
{"train_loss": 0.12543998658657074, "global_step": 205384, "epoch": 2307, "lr": 1.3965734084227834e-05}
{"train_loss": 0.06419774889945984, "global_step": 205385, "epoch": 2307, "lr": 1.3965332177095847e-05}
{"train_loss": 0.05093924701213837, "global_step": 205386, "epoch": 2307, "lr": 1.3964930274808269e-05}
{"train_loss": 0.03868426755070686, "global_step": 205387, "epoch": 2307, "lr": 1.3964528377365177e-05}
{"train_loss": 0.051750071346759796, "global_step": 205388, "epoch": 2307, "lr": 1.3964126484766604e-05}
{"train_loss": 0.05542297661304474, "global_step": 205389, "epoch": 2307, "lr": 1.3963724597012633e-05}
{"train_loss": 0.029803812503814697, "global_step": 205390, "epoch": 2307, "lr": 1.3963322714103284e-05}
{"train_loss": 0.040191639214754105, "global_step": 205391, "epoch": 2307, "lr": 1.396292083603864e-05}
{"train_loss": 0.026790864765644073, "global_step": 205392, "epoch": 2307, "lr": 1.3962518962818744e-05}
{"train_loss": 0.04591083154082298, "global_step": 205393, "epoch": 2307, "lr": 1.3962117094443628e-05}
{"train_loss": 0.1491757482290268, "global_step": 205394, "epoch": 2307, "lr": 1.3961715230913385e-05}
{"train_loss": 0.045347750186920166, "global_step": 205395, "epoch": 2307, "lr": 1.3961313372228036e-05}
{"train_loss": 0.03182467445731163, "global_step": 205396, "epoch": 2307, "lr": 1.3960911518387664e-05}
{"train_loss": 0.11918941140174866, "global_step": 205397, "epoch": 2307, "lr": 1.396050966939229e-05}
{"train_loss": 0.042805999517440796, "global_step": 205398, "epoch": 2307, "lr": 1.3960107825242003e-05}
{"train_loss": 0.043614450842142105, "global_step": 205399, "epoch": 2307, "lr": 1.3959705985936828e-05}
{"train_loss": 0.10017239302396774, "global_step": 205400, "epoch": 2307, "lr": 1.3959304151476843e-05}
{"train_loss": 0.10693855583667755, "global_step": 205401, "epoch": 2307, "lr": 1.3958902321862078e-05}
{"train_loss": 0.07883398234844208, "global_step": 205402, "epoch": 2307, "lr": 1.3958500497092609e-05}
{"train_loss": 0.09263814240694046, "global_step": 205403, "epoch": 2307, "lr": 1.395809867716847e-05}
{"train_loss": 0.04330693930387497, "global_step": 205404, "epoch": 2307, "lr": 1.3957696862089736e-05}
{"train_loss": 0.07857474684715271, "global_step": 205405, "epoch": 2307, "lr": 1.395729505185644e-05}
{"train_loss": 0.05761272460222244, "global_step": 205406, "epoch": 2307, "lr": 1.395689324646866e-05}
{"train_loss": 0.07797101140022278, "global_step": 205407, "epoch": 2307, "lr": 1.3956491445926422e-05}
{"train_loss": 0.14321209490299225, "global_step": 205408, "epoch": 2307, "lr": 1.3956089650229792e-05}
{"train_loss": 0.08080495893955231, "global_step": 205409, "epoch": 2307, "lr": 1.3955687859378846e-05}
{"train_loss": 0.1004277914762497, "global_step": 205410, "epoch": 2307, "lr": 1.3955286073373597e-05}
{"train_loss": 0.06801203872715489, "global_step": 205411, "epoch": 2307, "lr": 1.3954884292214143e-05, "val_loss": 8.460880279541016}
{"train_loss": 0.0674971342086792, "global_step": 205412, "epoch": 2308, "lr": 1.3954482515900496e-05}
{"train_loss": 0.07529525458812714, "global_step": 205413, "epoch": 2308, "lr": 1.3954080744432745e-05}
{"train_loss": 0.07750100642442703, "global_step": 205414, "epoch": 2308, "lr": 1.3953678977810914e-05}
{"train_loss": 0.040641527622938156, "global_step": 205415, "epoch": 2308, "lr": 1.3953277216035082e-05}
{"train_loss": 0.04706436023116112, "global_step": 205416, "epoch": 2308, "lr": 1.3952875459105286e-05}
{"train_loss": 0.04729315638542175, "global_step": 205417, "epoch": 2308, "lr": 1.3952473707021596e-05}
{"train_loss": 0.08580967783927917, "global_step": 205418, "epoch": 2308, "lr": 1.3952071959784041e-05}
{"train_loss": 0.09182145446538925, "global_step": 205419, "epoch": 2308, "lr": 1.3951670217392709e-05}
{"train_loss": 0.04213714599609375, "global_step": 205420, "epoch": 2308, "lr": 1.3951268479847618e-05}
{"train_loss": 0.06278078258037567, "global_step": 205421, "epoch": 2308, "lr": 1.3950866747148854e-05}
{"train_loss": 0.07866077870130539, "global_step": 205422, "epoch": 2308, "lr": 1.3950465019296443e-05}
{"train_loss": 0.05454274266958237, "global_step": 205423, "epoch": 2308, "lr": 1.3950063296290466e-05}
{"train_loss": 0.06696140766143799, "global_step": 205424, "epoch": 2308, "lr": 1.3949661578130952e-05}
{"train_loss": 0.09390760958194733, "global_step": 205425, "epoch": 2308, "lr": 1.3949259864817981e-05}
{"train_loss": 0.05931558087468147, "global_step": 205426, "epoch": 2308, "lr": 1.3948858156351574e-05}
{"train_loss": 0.0811021476984024, "global_step": 205427, "epoch": 2308, "lr": 1.3948456452731822e-05}
{"train_loss": 0.06288851052522659, "global_step": 205428, "epoch": 2308, "lr": 1.3948054753958745e-05}
{"train_loss": 0.042088381946086884, "global_step": 205429, "epoch": 2308, "lr": 1.3947653060032429e-05}
{"train_loss": 0.04849448800086975, "global_step": 205430, "epoch": 2308, "lr": 1.3947251370952908e-05}
{"train_loss": 0.04305025562644005, "global_step": 205431, "epoch": 2308, "lr": 1.3946849686720226e-05}
{"train_loss": 0.03480983152985573, "global_step": 205432, "epoch": 2308, "lr": 1.3946448007334467e-05}
{"train_loss": 0.08776554465293884, "global_step": 205433, "epoch": 2308, "lr": 1.3946046332795654e-05}
{"train_loss": 0.06478239595890045, "global_step": 205434, "epoch": 2308, "lr": 1.3945644663103869e-05}
{"train_loss": 0.05477483198046684, "global_step": 205435, "epoch": 2308, "lr": 1.394524299825914e-05}
{"train_loss": 0.09363503009080887, "global_step": 205436, "epoch": 2308, "lr": 1.3944841338261543e-05}
{"train_loss": 0.04783454164862633, "global_step": 205437, "epoch": 2308, "lr": 1.394443968311111e-05}
{"train_loss": 0.11325588077306747, "global_step": 205438, "epoch": 2308, "lr": 1.3944038032807927e-05}
{"train_loss": 0.0670594573020935, "global_step": 205439, "epoch": 2308, "lr": 1.394363638735201e-05}
{"train_loss": 0.0329616442322731, "global_step": 205440, "epoch": 2308, "lr": 1.394323474674345e-05}
{"train_loss": 0.03208388760685921, "global_step": 205441, "epoch": 2308, "lr": 1.3942833110982262e-05}
{"train_loss": 0.045942846685647964, "global_step": 205442, "epoch": 2308, "lr": 1.394243148006854e-05}
{"train_loss": 0.07267452031373978, "global_step": 205443, "epoch": 2308, "lr": 1.3942029854002303e-05}
{"train_loss": 0.07523594796657562, "global_step": 205444, "epoch": 2308, "lr": 1.3941628232783633e-05}
{"train_loss": 0.035196997225284576, "global_step": 205445, "epoch": 2308, "lr": 1.3941226616412556e-05}
{"train_loss": 0.09046655148267746, "global_step": 205446, "epoch": 2308, "lr": 1.3940825004889146e-05}
{"train_loss": 0.09713786095380783, "global_step": 205447, "epoch": 2308, "lr": 1.3940423398213464e-05}
{"train_loss": 0.07704665511846542, "global_step": 205448, "epoch": 2308, "lr": 1.3940021796385538e-05}
{"train_loss": 0.06142836809158325, "global_step": 205449, "epoch": 2308, "lr": 1.3939620199405456e-05}
{"train_loss": 0.030652271583676338, "global_step": 205450, "epoch": 2308, "lr": 1.393921860727323e-05}
{"train_loss": 0.0761805921792984, "global_step": 205451, "epoch": 2308, "lr": 1.3938817019988959e-05}
{"train_loss": 0.0833837240934372, "global_step": 205452, "epoch": 2308, "lr": 1.3938415437552655e-05}
{"train_loss": 0.04705093801021576, "global_step": 205453, "epoch": 2308, "lr": 1.3938013859964404e-05}
{"train_loss": 0.06280248612165451, "global_step": 205454, "epoch": 2308, "lr": 1.3937612287224239e-05}
{"train_loss": 0.07251075655221939, "global_step": 205455, "epoch": 2308, "lr": 1.3937210719332234e-05}
{"train_loss": 0.05077345296740532, "global_step": 205456, "epoch": 2308, "lr": 1.3936809156288415e-05}
{"train_loss": 0.031749919056892395, "global_step": 205457, "epoch": 2308, "lr": 1.3936407598092871e-05}
{"train_loss": 0.057769693434238434, "global_step": 205458, "epoch": 2308, "lr": 1.3936006044745625e-05}
{"train_loss": 0.08429333567619324, "global_step": 205459, "epoch": 2308, "lr": 1.3935604496246751e-05}
{"train_loss": 0.046424418687820435, "global_step": 205460, "epoch": 2308, "lr": 1.3935202952596282e-05}
{"train_loss": 0.02891959622502327, "global_step": 205461, "epoch": 2308, "lr": 1.3934801413794308e-05}
{"train_loss": 0.046502720564603806, "global_step": 205462, "epoch": 2308, "lr": 1.3934399879840838e-05}
{"train_loss": 0.010857859626412392, "global_step": 205463, "epoch": 2308, "lr": 1.3933998350735966e-05}
{"train_loss": 0.0673895925283432, "global_step": 205464, "epoch": 2308, "lr": 1.3933596826479712e-05}
{"train_loss": 0.06769376993179321, "global_step": 205465, "epoch": 2308, "lr": 1.3933195307072161e-05}
{"train_loss": 0.06451570987701416, "global_step": 205466, "epoch": 2308, "lr": 1.3932793792513339e-05}
{"train_loss": 0.06852959096431732, "global_step": 205467, "epoch": 2308, "lr": 1.3932392282803326e-05}
{"train_loss": 0.08678842335939407, "global_step": 205468, "epoch": 2308, "lr": 1.3931990777942166e-05}
{"train_loss": 0.07330182939767838, "global_step": 205469, "epoch": 2308, "lr": 1.3931589277929891e-05}
{"train_loss": 0.043363168835639954, "global_step": 205470, "epoch": 2308, "lr": 1.3931187782766592e-05}
{"train_loss": 0.06483768671751022, "global_step": 205471, "epoch": 2308, "lr": 1.3930786292452291e-05}
{"train_loss": 0.06833072751760483, "global_step": 205472, "epoch": 2308, "lr": 1.3930384806987068e-05}
{"train_loss": 0.041075605899095535, "global_step": 205473, "epoch": 2308, "lr": 1.3929983326370955e-05}
{"train_loss": 0.07221349328756332, "global_step": 205474, "epoch": 2308, "lr": 1.3929581850604029e-05}
{"train_loss": 0.04922936484217644, "global_step": 205475, "epoch": 2308, "lr": 1.3929180379686312e-05}
{"train_loss": 0.05330610275268555, "global_step": 205476, "epoch": 2308, "lr": 1.3928778913617897e-05}
{"train_loss": 0.06622602045536041, "global_step": 205477, "epoch": 2308, "lr": 1.39283774523988e-05}
{"train_loss": 0.032838933169841766, "global_step": 205478, "epoch": 2308, "lr": 1.3927975996029113e-05}
{"train_loss": 0.0436856783926487, "global_step": 205479, "epoch": 2308, "lr": 1.3927574544508853e-05}
{"train_loss": 0.0960332378745079, "global_step": 205480, "epoch": 2308, "lr": 1.3927173097838103e-05}
{"train_loss": 0.06557818502187729, "global_step": 205481, "epoch": 2308, "lr": 1.392677165601689e-05}
{"train_loss": 0.05409526824951172, "global_step": 205482, "epoch": 2308, "lr": 1.3926370219045287e-05}
{"train_loss": 0.07001601159572601, "global_step": 205483, "epoch": 2308, "lr": 1.3925968786923355e-05}
{"train_loss": 0.041363973170518875, "global_step": 205484, "epoch": 2308, "lr": 1.392556735965112e-05}
{"train_loss": 0.049167450517416, "global_step": 205485, "epoch": 2308, "lr": 1.3925165937228674e-05}
{"train_loss": 0.07630008459091187, "global_step": 205486, "epoch": 2308, "lr": 1.392476451965603e-05}
{"train_loss": 0.02896224707365036, "global_step": 205487, "epoch": 2308, "lr": 1.392436310693328e-05}
{"train_loss": 0.09533044695854187, "global_step": 205488, "epoch": 2308, "lr": 1.392396169906044e-05}
{"train_loss": 0.08545177429914474, "global_step": 205489, "epoch": 2308, "lr": 1.3923560296037602e-05}
{"train_loss": 0.036937907338142395, "global_step": 205490, "epoch": 2308, "lr": 1.3923158897864786e-05}
{"train_loss": 0.05373823642730713, "global_step": 205491, "epoch": 2308, "lr": 1.3922757504542077e-05}
{"train_loss": 0.10325421392917633, "global_step": 205492, "epoch": 2308, "lr": 1.3922356116069501e-05}
{"train_loss": 0.06814458221197128, "global_step": 205493, "epoch": 2308, "lr": 1.3921954732447134e-05}
{"train_loss": 0.10421149432659149, "global_step": 205494, "epoch": 2308, "lr": 1.392155335367501e-05}
{"train_loss": 0.045874644070863724, "global_step": 205495, "epoch": 2308, "lr": 1.3921151979753206e-05}
{"train_loss": 0.02550668455660343, "global_step": 205496, "epoch": 2308, "lr": 1.3920750610681748e-05}
{"train_loss": 0.04984854906797409, "global_step": 205497, "epoch": 2308, "lr": 1.3920349246460723e-05}
{"train_loss": 0.07079598307609558, "global_step": 205498, "epoch": 2308, "lr": 1.391994788709015e-05}
{"train_loss": 0.022573595866560936, "global_step": 205499, "epoch": 2308, "lr": 1.3919546532570099e-05}
{"train_loss": 0.06112469435575303, "global_step": 205500, "epoch": 2308, "lr": 1.3919145182900644e-05, "val_loss": 8.377345085144043}
{"train_loss": 0.04323487728834152, "global_step": 205501, "epoch": 2309, "lr": 1.3918743838081805e-05}
{"train_loss": 0.05065689980983734, "global_step": 205502, "epoch": 2309, "lr": 1.3918342498113663e-05}
{"train_loss": 0.05700695887207985, "global_step": 205503, "epoch": 2309, "lr": 1.3917941162996263e-05}
{"train_loss": 0.09127258509397507, "global_step": 205504, "epoch": 2309, "lr": 1.3917539832729637e-05}
{"train_loss": 0.04880824685096741, "global_step": 205505, "epoch": 2309, "lr": 1.3917138507313877e-05}
{"train_loss": 0.06051909551024437, "global_step": 205506, "epoch": 2309, "lr": 1.3916737186749e-05}
{"train_loss": 0.07588034123182297, "global_step": 205507, "epoch": 2309, "lr": 1.3916335871035096e-05}
{"train_loss": 0.15011009573936462, "global_step": 205508, "epoch": 2309, "lr": 1.39159345601722e-05}
{"train_loss": 0.08743763715028763, "global_step": 205509, "epoch": 2309, "lr": 1.3915533254160352e-05}
{"train_loss": 0.03355340287089348, "global_step": 205510, "epoch": 2309, "lr": 1.3915131952999633e-05}
{"train_loss": 0.07122331857681274, "global_step": 205511, "epoch": 2309, "lr": 1.3914730656690073e-05}
{"train_loss": 0.09964701533317566, "global_step": 205512, "epoch": 2309, "lr": 1.3914329365231748e-05}
{"train_loss": 0.04500419273972511, "global_step": 205513, "epoch": 2309, "lr": 1.3913928078624694e-05}
{"train_loss": 0.10462790727615356, "global_step": 205514, "epoch": 2309, "lr": 1.391352679686898e-05}
{"train_loss": 0.03216617926955223, "global_step": 205515, "epoch": 2309, "lr": 1.3913125519964643e-05}
{"train_loss": 0.07040440291166306, "global_step": 205516, "epoch": 2309, "lr": 1.3912724247911757e-05}
{"train_loss": 0.04759370908141136, "global_step": 205517, "epoch": 2309, "lr": 1.3912322980710351e-05}
{"train_loss": 0.09302148222923279, "global_step": 205518, "epoch": 2309, "lr": 1.3911921718360498e-05}
{"train_loss": 0.08991187810897827, "global_step": 205519, "epoch": 2309, "lr": 1.391152046086226e-05}
{"train_loss": 0.06568239629268646, "global_step": 205520, "epoch": 2309, "lr": 1.3911119208215662e-05}
{"train_loss": 0.06060688942670822, "global_step": 205521, "epoch": 2309, "lr": 1.3910717960420788e-05}
{"train_loss": 0.054140690714120865, "global_step": 205522, "epoch": 2309, "lr": 1.3910316717477667e-05}
{"train_loss": 0.0538264736533165, "global_step": 205523, "epoch": 2309, "lr": 1.3909915479386377e-05}
{"train_loss": 0.05480235815048218, "global_step": 205524, "epoch": 2309, "lr": 1.3909514246146942e-05}
{"train_loss": 0.10044268518686295, "global_step": 205525, "epoch": 2309, "lr": 1.3909113017759451e-05}
{"train_loss": 0.052274271845817566, "global_step": 205526, "epoch": 2309, "lr": 1.3908711794223922e-05}
{"train_loss": 0.0585012286901474, "global_step": 205527, "epoch": 2309, "lr": 1.390831057554044e-05}
{"train_loss": 0.04632030427455902, "global_step": 205528, "epoch": 2309, "lr": 1.3907909361709038e-05}
{"train_loss": 0.07611629366874695, "global_step": 205529, "epoch": 2309, "lr": 1.3907508152729787e-05}
{"train_loss": 0.07945110648870468, "global_step": 205530, "epoch": 2309, "lr": 1.3907106948602717e-05}
{"train_loss": 0.02436167374253273, "global_step": 205531, "epoch": 2309, "lr": 1.3906705749327914e-05}
{"train_loss": 0.06543773412704468, "global_step": 205532, "epoch": 2309, "lr": 1.3906304554905397e-05}
{"train_loss": 0.11953829973936081, "global_step": 205533, "epoch": 2309, "lr": 1.3905903365335254e-05}
{"train_loss": 0.06880924105644226, "global_step": 205534, "epoch": 2309, "lr": 1.3905502180617507e-05}
{"train_loss": 0.060527049005031586, "global_step": 205535, "epoch": 2309, "lr": 1.3905101000752224e-05}
{"train_loss": 0.042290572077035904, "global_step": 205536, "epoch": 2309, "lr": 1.390469982573948e-05}
{"train_loss": 0.02392527088522911, "global_step": 205537, "epoch": 2309, "lr": 1.3904298655579289e-05}
{"train_loss": 0.08279599249362946, "global_step": 205538, "epoch": 2309, "lr": 1.3903897490271744e-05}
{"train_loss": 0.05505891144275665, "global_step": 205539, "epoch": 2309, "lr": 1.3903496329816861e-05}
{"train_loss": 0.05932249873876572, "global_step": 205540, "epoch": 2309, "lr": 1.390309517421473e-05}
{"train_loss": 0.08880601823329926, "global_step": 205541, "epoch": 2309, "lr": 1.390269402346539e-05}
{"train_loss": 0.11282944679260254, "global_step": 205542, "epoch": 2309, "lr": 1.3902292877568873e-05}
{"train_loss": 0.09204686433076859, "global_step": 205543, "epoch": 2309, "lr": 1.3901891736525275e-05}
{"train_loss": 0.04087112471461296, "global_step": 205544, "epoch": 2309, "lr": 1.3901490600334605e-05}
{"train_loss": 0.03948105126619339, "global_step": 205545, "epoch": 2309, "lr": 1.3901089468996959e-05}
{"train_loss": 0.06232940033078194, "global_step": 205546, "epoch": 2309, "lr": 1.390068834251237e-05}
{"train_loss": 0.04505078122019768, "global_step": 205547, "epoch": 2309, "lr": 1.390028722088088e-05}
{"train_loss": 0.05090256407856941, "global_step": 205548, "epoch": 2309, "lr": 1.3899886104102566e-05}
{"train_loss": 0.03389181196689606, "global_step": 205549, "epoch": 2309, "lr": 1.3899484992177465e-05}
{"train_loss": 0.05510810390114784, "global_step": 205550, "epoch": 2309, "lr": 1.389908388510565e-05}
{"train_loss": 0.08570527285337448, "global_step": 205551, "epoch": 2309, "lr": 1.3898682782887146e-05}
{"train_loss": 0.07433933019638062, "global_step": 205552, "epoch": 2309, "lr": 1.3898281685522042e-05}
{"train_loss": 0.09095185995101929, "global_step": 205553, "epoch": 2309, "lr": 1.389788059301036e-05}
{"train_loss": 0.09645170718431473, "global_step": 205554, "epoch": 2309, "lr": 1.3897479505352162e-05}
{"train_loss": 0.11952177435159683, "global_step": 205555, "epoch": 2309, "lr": 1.3897078422547527e-05}
{"train_loss": 0.05541345849633217, "global_step": 205556, "epoch": 2309, "lr": 1.3896677344596477e-05}
{"train_loss": 0.046872373670339584, "global_step": 205557, "epoch": 2309, "lr": 1.3896276271499092e-05}
{"train_loss": 0.053590066730976105, "global_step": 205558, "epoch": 2309, "lr": 1.3895875203255393e-05}
{"train_loss": 0.084710493683815, "global_step": 205559, "epoch": 2309, "lr": 1.3895474139865472e-05}
{"train_loss": 0.12976151704788208, "global_step": 205560, "epoch": 2309, "lr": 1.3895073081329352e-05}
{"train_loss": 0.09803919494152069, "global_step": 205561, "epoch": 2309, "lr": 1.3894672027647109e-05}
{"train_loss": 0.06334079802036285, "global_step": 205562, "epoch": 2309, "lr": 1.389427097881878e-05}
{"train_loss": 0.04078473895788193, "global_step": 205563, "epoch": 2309, "lr": 1.3893869934844434e-05}
{"train_loss": 0.0679059624671936, "global_step": 205564, "epoch": 2309, "lr": 1.3893468895724105e-05}
{"train_loss": 0.05194558948278427, "global_step": 205565, "epoch": 2309, "lr": 1.3893067861457877e-05}
{"train_loss": 0.04751535505056381, "global_step": 205566, "epoch": 2309, "lr": 1.3892666832045763e-05}
{"train_loss": 0.03348333016037941, "global_step": 205567, "epoch": 2309, "lr": 1.3892265807487864e-05}
{"train_loss": 0.044496502727270126, "global_step": 205568, "epoch": 2309, "lr": 1.3891864787784187e-05}
{"train_loss": 0.057436857372522354, "global_step": 205569, "epoch": 2309, "lr": 1.3891463772934826e-05}
{"train_loss": 0.034207798540592194, "global_step": 205570, "epoch": 2309, "lr": 1.3891062762939805e-05}
{"train_loss": 0.11512473970651627, "global_step": 205571, "epoch": 2309, "lr": 1.3890661757799206e-05}
{"train_loss": 0.051894765347242355, "global_step": 205572, "epoch": 2309, "lr": 1.3890260757513051e-05}
{"train_loss": 0.0732978880405426, "global_step": 205573, "epoch": 2309, "lr": 1.388985976208141e-05}
{"train_loss": 0.06373009085655212, "global_step": 205574, "epoch": 2309, "lr": 1.3889458771504349e-05}
{"train_loss": 0.06942097842693329, "global_step": 205575, "epoch": 2309, "lr": 1.38890577857819e-05}
{"train_loss": 0.0872124433517456, "global_step": 205576, "epoch": 2309, "lr": 1.388865680491414e-05}
{"train_loss": 0.11697643995285034, "global_step": 205577, "epoch": 2309, "lr": 1.3888255828901093e-05}
{"train_loss": 0.04728792607784271, "global_step": 205578, "epoch": 2309, "lr": 1.388785485774285e-05}
{"train_loss": 0.06571490317583084, "global_step": 205579, "epoch": 2309, "lr": 1.388745389143944e-05}
{"train_loss": 0.016415411606431007, "global_step": 205580, "epoch": 2309, "lr": 1.3887052929990907e-05}
{"train_loss": 0.07322226464748383, "global_step": 205581, "epoch": 2309, "lr": 1.388665197339734e-05}
{"train_loss": 0.040254879742860794, "global_step": 205582, "epoch": 2309, "lr": 1.3886251021658753e-05}
{"train_loss": 0.05113212391734123, "global_step": 205583, "epoch": 2309, "lr": 1.3885850074775237e-05}
{"train_loss": 0.08722557127475739, "global_step": 205584, "epoch": 2309, "lr": 1.3885449132746826e-05}
{"train_loss": 0.08128301054239273, "global_step": 205585, "epoch": 2309, "lr": 1.3885048195573558e-05}
{"train_loss": 0.041900258511304855, "global_step": 205586, "epoch": 2309, "lr": 1.3884647263255524e-05}
{"train_loss": 0.07728283852338791, "global_step": 205587, "epoch": 2309, "lr": 1.3884246335792745e-05}
{"train_loss": 0.057126253843307495, "global_step": 205588, "epoch": 2309, "lr": 1.3883845413185303e-05}
{"train_loss": 0.0667389462377583, "global_step": 205589, "epoch": 2309, "lr": 1.3883444495433224e-05, "val_loss": 8.460733413696289}
{"train_loss": 0.01987569034099579, "global_step": 205590, "epoch": 2310, "lr": 1.388304358253657e-05}
{"train_loss": 0.048299696296453476, "global_step": 205591, "epoch": 2310, "lr": 1.3882642674495422e-05}
{"train_loss": 0.05579885467886925, "global_step": 205592, "epoch": 2310, "lr": 1.3882241771309795e-05}
{"train_loss": 0.05599811300635338, "global_step": 205593, "epoch": 2310, "lr": 1.3881840872979773e-05}
{"train_loss": 0.11933576315641403, "global_step": 205594, "epoch": 2310, "lr": 1.3881439979505384e-05}
{"train_loss": 0.029902132228016853, "global_step": 205595, "epoch": 2310, "lr": 1.3881039090886711e-05}
{"train_loss": 0.040443211793899536, "global_step": 205596, "epoch": 2310, "lr": 1.388063820712378e-05}
{"train_loss": 0.0834321677684784, "global_step": 205597, "epoch": 2310, "lr": 1.3880237328216667e-05}
{"train_loss": 0.033850789070129395, "global_step": 205598, "epoch": 2310, "lr": 1.3879836454165401e-05}
{"train_loss": 0.06231246516108513, "global_step": 205599, "epoch": 2310, "lr": 1.3879435584970068e-05}
{"train_loss": 0.07710051536560059, "global_step": 205600, "epoch": 2310, "lr": 1.387903472063069e-05}
{"train_loss": 0.03904134780168533, "global_step": 205601, "epoch": 2310, "lr": 1.387863386114735e-05}
{"train_loss": 0.034459903836250305, "global_step": 205602, "epoch": 2310, "lr": 1.3878233006520069e-05}
{"train_loss": 0.06573964655399323, "global_step": 205603, "epoch": 2310, "lr": 1.3877832156748938e-05}
{"train_loss": 0.10126124322414398, "global_step": 205604, "epoch": 2310, "lr": 1.3877431311833977e-05}
{"train_loss": 0.04679242894053459, "global_step": 205605, "epoch": 2310, "lr": 1.3877030471775271e-05}
{"train_loss": 0.1450420320034027, "global_step": 205606, "epoch": 2310, "lr": 1.387662963657284e-05}
{"train_loss": 0.04061339050531387, "global_step": 205607, "epoch": 2310, "lr": 1.3876228806226777e-05}
{"train_loss": 0.07911726087331772, "global_step": 205608, "epoch": 2310, "lr": 1.3875827980737094e-05}
{"train_loss": 0.029616396874189377, "global_step": 205609, "epoch": 2310, "lr": 1.3875427160103865e-05}
{"train_loss": 0.06764236092567444, "global_step": 205610, "epoch": 2310, "lr": 1.3875026344327163e-05}
{"train_loss": 0.0627983883023262, "global_step": 205611, "epoch": 2310, "lr": 1.3874625533407004e-05}
{"train_loss": 0.08408568054437637, "global_step": 205612, "epoch": 2310, "lr": 1.3874224727343483e-05}
{"train_loss": 0.08914589881896973, "global_step": 205613, "epoch": 2310, "lr": 1.387382392613661e-05}
{"train_loss": 0.08725519478321075, "global_step": 205614, "epoch": 2310, "lr": 1.3873423129786477e-05}
{"train_loss": 0.06736776977777481, "global_step": 205615, "epoch": 2310, "lr": 1.3873022338293112e-05}
{"train_loss": 0.08796095848083496, "global_step": 205616, "epoch": 2310, "lr": 1.387262155165659e-05}
{"train_loss": 0.06479611992835999, "global_step": 205617, "epoch": 2310, "lr": 1.3872220769876942e-05}
{"train_loss": 0.07647901028394699, "global_step": 205618, "epoch": 2310, "lr": 1.3871819992954244e-05}
{"train_loss": 0.03652050718665123, "global_step": 205619, "epoch": 2310, "lr": 1.3871419220888543e-05}
{"train_loss": 0.0579981654882431, "global_step": 205620, "epoch": 2310, "lr": 1.3871018453679873e-05}
{"train_loss": 0.08918896317481995, "global_step": 205621, "epoch": 2310, "lr": 1.3870617691328319e-05}
{"train_loss": 0.10728122293949127, "global_step": 205622, "epoch": 2310, "lr": 1.3870216933833907e-05}
{"train_loss": 0.052184898406267166, "global_step": 205623, "epoch": 2310, "lr": 1.3869816181196715e-05}
{"train_loss": 0.0646691769361496, "global_step": 205624, "epoch": 2310, "lr": 1.3869415433416789e-05}
{"train_loss": 0.060520440340042114, "global_step": 205625, "epoch": 2310, "lr": 1.3869014690494164e-05}
{"train_loss": 0.06430142372846603, "global_step": 205626, "epoch": 2310, "lr": 1.386861395242891e-05}
{"train_loss": 0.05232946574687958, "global_step": 205627, "epoch": 2310, "lr": 1.38682132192211e-05}
{"train_loss": 0.03540400415658951, "global_step": 205628, "epoch": 2310, "lr": 1.3867812490870747e-05}
{"train_loss": 0.052100107073783875, "global_step": 205629, "epoch": 2310, "lr": 1.386741176737794e-05}
{"train_loss": 0.03885447233915329, "global_step": 205630, "epoch": 2310, "lr": 1.3867011048742707e-05}
{"train_loss": 0.048011213541030884, "global_step": 205631, "epoch": 2310, "lr": 1.386661033496513e-05}
{"train_loss": 0.11912377178668976, "global_step": 205632, "epoch": 2310, "lr": 1.3866209626045228e-05}
{"train_loss": 0.050485409796237946, "global_step": 205633, "epoch": 2310, "lr": 1.3865808921983092e-05}
{"train_loss": 0.047999367117881775, "global_step": 205634, "epoch": 2310, "lr": 1.3865408222778742e-05}
{"train_loss": 0.06169375777244568, "global_step": 205635, "epoch": 2310, "lr": 1.386500752843226e-05}
{"train_loss": 0.06097187101840973, "global_step": 205636, "epoch": 2310, "lr": 1.3864606838943678e-05}
{"train_loss": 0.020258242264389992, "global_step": 205637, "epoch": 2310, "lr": 1.386420615431307e-05}
{"train_loss": 0.08562368899583817, "global_step": 205638, "epoch": 2310, "lr": 1.3863805474540464e-05}
{"train_loss": 0.10462290048599243, "global_step": 205639, "epoch": 2310, "lr": 1.3863404799625946e-05}
{"train_loss": 0.02988273836672306, "global_step": 205640, "epoch": 2310, "lr": 1.3863004129569534e-05}
{"train_loss": 0.040507860481739044, "global_step": 205641, "epoch": 2310, "lr": 1.386260346437132e-05}
{"train_loss": 0.05444609001278877, "global_step": 205642, "epoch": 2310, "lr": 1.3862202804031322e-05}
{"train_loss": 0.04830194637179375, "global_step": 205643, "epoch": 2310, "lr": 1.3861802148549623e-05}
{"train_loss": 0.05257854610681534, "global_step": 205644, "epoch": 2310, "lr": 1.3861401497926251e-05}
{"train_loss": 0.08027289807796478, "global_step": 205645, "epoch": 2310, "lr": 1.3861000852161276e-05}
{"train_loss": 0.02817457728087902, "global_step": 205646, "epoch": 2310, "lr": 1.386060021125476e-05}
{"train_loss": 0.09740783274173737, "global_step": 205647, "epoch": 2310, "lr": 1.3860199575206734e-05}
{"train_loss": 0.08372317999601364, "global_step": 205648, "epoch": 2310, "lr": 1.3859798944017276e-05}
{"train_loss": 0.06465510278940201, "global_step": 205649, "epoch": 2310, "lr": 1.3859398317686411e-05}
{"train_loss": 0.050881315022706985, "global_step": 205650, "epoch": 2310, "lr": 1.3858997696214232e-05}
{"train_loss": 0.09053396433591843, "global_step": 205651, "epoch": 2310, "lr": 1.385859707960075e-05}
{"train_loss": 0.07873666286468506, "global_step": 205652, "epoch": 2310, "lr": 1.3858196467846053e-05}
{"train_loss": 0.04895246401429176, "global_step": 205653, "epoch": 2310, "lr": 1.3857795860950167e-05}
{"train_loss": 0.0796494111418724, "global_step": 205654, "epoch": 2310, "lr": 1.3857395258913174e-05}
{"train_loss": 0.04628020152449608, "global_step": 205655, "epoch": 2310, "lr": 1.3856994661735101e-05}
{"train_loss": 0.05190788209438324, "global_step": 205656, "epoch": 2310, "lr": 1.385659406941603e-05}
{"train_loss": 0.05313365161418915, "global_step": 205657, "epoch": 2310, "lr": 1.3856193481955993e-05}
{"train_loss": 0.03706284612417221, "global_step": 205658, "epoch": 2310, "lr": 1.3855792899355042e-05}
{"train_loss": 0.045220643281936646, "global_step": 205659, "epoch": 2310, "lr": 1.3855392321613253e-05}
{"train_loss": 0.03510813042521477, "global_step": 205660, "epoch": 2310, "lr": 1.385499174873065e-05}
{"train_loss": 0.0415598526597023, "global_step": 205661, "epoch": 2310, "lr": 1.3854591180707317e-05}
{"train_loss": 0.12046156823635101, "global_step": 205662, "epoch": 2310, "lr": 1.3854190617543278e-05}
{"train_loss": 0.051707006990909576, "global_step": 205663, "epoch": 2310, "lr": 1.385379005923862e-05}
{"train_loss": 0.079477958381176, "global_step": 205664, "epoch": 2310, "lr": 1.3853389505793363e-05}
{"train_loss": 0.059760384261608124, "global_step": 205665, "epoch": 2310, "lr": 1.3852988957207592e-05}
{"train_loss": 0.03923708572983742, "global_step": 205666, "epoch": 2310, "lr": 1.3852588413481333e-05}
{"train_loss": 0.04145948588848114, "global_step": 205667, "epoch": 2310, "lr": 1.3852187874614665e-05}
{"train_loss": 0.06679505854845047, "global_step": 205668, "epoch": 2310, "lr": 1.385178734060762e-05}
{"train_loss": 0.09156383574008942, "global_step": 205669, "epoch": 2310, "lr": 1.3851386811460272e-05}
{"train_loss": 0.02550777979195118, "global_step": 205670, "epoch": 2310, "lr": 1.3850986287172646e-05}
{"train_loss": 0.04756033048033714, "global_step": 205671, "epoch": 2310, "lr": 1.3850585767744834e-05}
{"train_loss": 0.012897941283881664, "global_step": 205672, "epoch": 2310, "lr": 1.3850185253176856e-05}
{"train_loss": 0.07473695278167725, "global_step": 205673, "epoch": 2310, "lr": 1.3849784743468797e-05}
{"train_loss": 0.06279338896274567, "global_step": 205674, "epoch": 2310, "lr": 1.3849384238620672e-05}
{"train_loss": 0.0687331110239029, "global_step": 205675, "epoch": 2310, "lr": 1.3848983738632575e-05}
{"train_loss": 0.05299517512321472, "global_step": 205676, "epoch": 2310, "lr": 1.3848583243504525e-05}
{"train_loss": 0.09607094526290894, "global_step": 205677, "epoch": 2310, "lr": 1.384818275323661e-05}
{"train_loss": 0.061721483528028045, "global_step": 205678, "epoch": 2310, "lr": 1.3847782267828852e-05, "val_loss": 8.553674697875977, "train_action_mse_error": 27.674152374267578}
{"train_loss": 0.05697431415319443, "global_step": 205679, "epoch": 2311, "lr": 1.3847381787281333e-05}
{"train_loss": 0.04979905113577843, "global_step": 205680, "epoch": 2311, "lr": 1.3846981311594076e-05}
{"train_loss": 0.13924847543239594, "global_step": 205681, "epoch": 2311, "lr": 1.3846580840767154e-05}
{"train_loss": 0.06671535223722458, "global_step": 205682, "epoch": 2311, "lr": 1.3846180374800632e-05}
{"train_loss": 0.09867747873067856, "global_step": 205683, "epoch": 2311, "lr": 1.3845779913694535e-05}
{"train_loss": 0.08034338802099228, "global_step": 205684, "epoch": 2311, "lr": 1.3845379457448948e-05}
{"train_loss": 0.029985161498188972, "global_step": 205685, "epoch": 2311, "lr": 1.3844979006063897e-05}
{"train_loss": 0.09085334837436676, "global_step": 205686, "epoch": 2311, "lr": 1.3844578559539461e-05}
{"train_loss": 0.053535547107458115, "global_step": 205687, "epoch": 2311, "lr": 1.3844178117875661e-05}
{"train_loss": 0.06518373638391495, "global_step": 205688, "epoch": 2311, "lr": 1.384377768107259e-05}
{"train_loss": 0.054512497037649155, "global_step": 205689, "epoch": 2311, "lr": 1.3843377249130268e-05}
{"train_loss": 0.08831798285245895, "global_step": 205690, "epoch": 2311, "lr": 1.3842976822048776e-05}
{"train_loss": 0.12278038263320923, "global_step": 205691, "epoch": 2311, "lr": 1.3842576399828145e-05}
{"train_loss": 0.07427071034908295, "global_step": 205692, "epoch": 2311, "lr": 1.3842175982468448e-05}
{"train_loss": 0.06110375002026558, "global_step": 205693, "epoch": 2311, "lr": 1.3841775569969717e-05}
{"train_loss": 0.08019544929265976, "global_step": 205694, "epoch": 2311, "lr": 1.3841375162332032e-05}
{"train_loss": 0.12408135086297989, "global_step": 205695, "epoch": 2311, "lr": 1.3840974759555436e-05}
{"train_loss": 0.08254911750555038, "global_step": 205696, "epoch": 2311, "lr": 1.3840574361639963e-05}
{"train_loss": 0.06477053463459015, "global_step": 205697, "epoch": 2311, "lr": 1.3840173968585702e-05}
{"train_loss": 0.053073450922966, "global_step": 205698, "epoch": 2311, "lr": 1.3839773580392673e-05}
{"train_loss": 0.06119023263454437, "global_step": 205699, "epoch": 2311, "lr": 1.3839373197060956e-05}
{"train_loss": 0.05544231832027435, "global_step": 205700, "epoch": 2311, "lr": 1.3838972818590585e-05}
{"train_loss": 0.0803980752825737, "global_step": 205701, "epoch": 2311, "lr": 1.3838572444981635e-05}
{"train_loss": 0.09470850974321365, "global_step": 205702, "epoch": 2311, "lr": 1.3838172076234135e-05}
{"train_loss": 0.09036799520254135, "global_step": 205703, "epoch": 2311, "lr": 1.3837771712348168e-05}
{"train_loss": 0.06011121720075607, "global_step": 205704, "epoch": 2311, "lr": 1.3837371353323758e-05}
{"train_loss": 0.02942444011569023, "global_step": 205705, "epoch": 2311, "lr": 1.383697099916098e-05}
{"train_loss": 0.08376961946487427, "global_step": 205706, "epoch": 2311, "lr": 1.3836570649859876e-05}
{"train_loss": 0.05348380282521248, "global_step": 205707, "epoch": 2311, "lr": 1.3836170305420509e-05}
{"train_loss": 0.09878768771886826, "global_step": 205708, "epoch": 2311, "lr": 1.383576996584292e-05}
{"train_loss": 0.046200964599847794, "global_step": 205709, "epoch": 2311, "lr": 1.3835369631127188e-05}
{"train_loss": 0.06089213117957115, "global_step": 205710, "epoch": 2311, "lr": 1.3834969301273327e-05}
{"train_loss": 0.09578774124383926, "global_step": 205711, "epoch": 2311, "lr": 1.383456897628143e-05}
{"train_loss": 0.04531770199537277, "global_step": 205712, "epoch": 2311, "lr": 1.3834168656151524e-05}
{"train_loss": 0.05365382879972458, "global_step": 205713, "epoch": 2311, "lr": 1.383376834088369e-05}
{"train_loss": 0.06268033385276794, "global_step": 205714, "epoch": 2311, "lr": 1.3833368030477944e-05}
{"train_loss": 0.0821654200553894, "global_step": 205715, "epoch": 2311, "lr": 1.3832967724934375e-05}
{"train_loss": 0.05442342534661293, "global_step": 205716, "epoch": 2311, "lr": 1.3832567424253013e-05}
{"train_loss": 0.06431595981121063, "global_step": 205717, "epoch": 2311, "lr": 1.3832167128433921e-05}
{"train_loss": 0.06219731271266937, "global_step": 205718, "epoch": 2311, "lr": 1.3831766837477168e-05}
{"train_loss": 0.0591813288629055, "global_step": 205719, "epoch": 2311, "lr": 1.3831366551382775e-05}
{"train_loss": 0.0460253469645977, "global_step": 205720, "epoch": 2311, "lr": 1.383096627015083e-05}
{"train_loss": 0.0960519015789032, "global_step": 205721, "epoch": 2311, "lr": 1.3830565993781358e-05}
{"train_loss": 0.05622730031609535, "global_step": 205722, "epoch": 2311, "lr": 1.3830165722274435e-05}
{"train_loss": 0.08114133775234222, "global_step": 205723, "epoch": 2311, "lr": 1.3829765455630095e-05}
{"train_loss": 0.0768745020031929, "global_step": 205724, "epoch": 2311, "lr": 1.3829365193848414e-05}
{"train_loss": 0.06889773160219193, "global_step": 205725, "epoch": 2311, "lr": 1.3828964936929422e-05}
{"train_loss": 0.04299040511250496, "global_step": 205726, "epoch": 2311, "lr": 1.3828564684873202e-05}
{"train_loss": 0.14882291853427887, "global_step": 205727, "epoch": 2311, "lr": 1.3828164437679775e-05}
{"train_loss": 0.03711865097284317, "global_step": 205728, "epoch": 2311, "lr": 1.3827764195349218e-05}
{"train_loss": 0.05140085518360138, "global_step": 205729, "epoch": 2311, "lr": 1.3827363957881572e-05}
{"train_loss": 0.03413001820445061, "global_step": 205730, "epoch": 2311, "lr": 1.3826963725276903e-05}
{"train_loss": 0.08168881386518478, "global_step": 205731, "epoch": 2311, "lr": 1.382656349753525e-05}
{"train_loss": 0.035762906074523926, "global_step": 205732, "epoch": 2311, "lr": 1.3826163274656685e-05}
{"train_loss": 0.09165570139884949, "global_step": 205733, "epoch": 2311, "lr": 1.3825763056641238e-05}
{"train_loss": 0.020526401698589325, "global_step": 205734, "epoch": 2311, "lr": 1.3825362843488988e-05}
{"train_loss": 0.08150026202201843, "global_step": 205735, "epoch": 2311, "lr": 1.3824962635199983e-05}
{"train_loss": 0.029825635254383087, "global_step": 205736, "epoch": 2311, "lr": 1.3824562431774251e-05}
{"train_loss": 0.07858560979366302, "global_step": 205737, "epoch": 2311, "lr": 1.3824162233211886e-05}
{"train_loss": 0.05965500324964523, "global_step": 205738, "epoch": 2311, "lr": 1.38237620395129e-05}
{"train_loss": 0.04082728177309036, "global_step": 205739, "epoch": 2311, "lr": 1.3823361850677385e-05}
{"train_loss": 0.0451335571706295, "global_step": 205740, "epoch": 2311, "lr": 1.3822961666705363e-05}
{"train_loss": 0.08122451603412628, "global_step": 205741, "epoch": 2311, "lr": 1.3822561487596919e-05}
{"train_loss": 0.04946131631731987, "global_step": 205742, "epoch": 2311, "lr": 1.3822161313352078e-05}
{"train_loss": 0.05329880490899086, "global_step": 205743, "epoch": 2311, "lr": 1.3821761143970919e-05}
{"train_loss": 0.0754329115152359, "global_step": 205744, "epoch": 2311, "lr": 1.3821360979453469e-05}
{"train_loss": 0.029853586107492447, "global_step": 205745, "epoch": 2311, "lr": 1.382096081979981e-05}
{"train_loss": 0.03884880989789963, "global_step": 205746, "epoch": 2311, "lr": 1.3820560665009963e-05}
{"train_loss": 0.0722028836607933, "global_step": 205747, "epoch": 2311, "lr": 1.3820160515084024e-05}
{"train_loss": 0.06452209502458572, "global_step": 205748, "epoch": 2311, "lr": 1.3819760370022e-05}
{"train_loss": 0.07439723610877991, "global_step": 205749, "epoch": 2311, "lr": 1.3819360229823992e-05}
{"train_loss": 0.025747517123818398, "global_step": 205750, "epoch": 2311, "lr": 1.3818960094490008e-05}
{"train_loss": 0.033291418105363846, "global_step": 205751, "epoch": 2311, "lr": 1.3818559964020145e-05}
{"train_loss": 0.05336582288146019, "global_step": 205752, "epoch": 2311, "lr": 1.3818159838414413e-05}
{"train_loss": 0.07806310802698135, "global_step": 205753, "epoch": 2311, "lr": 1.3817759717672896e-05}
{"train_loss": 0.0217733234167099, "global_step": 205754, "epoch": 2311, "lr": 1.3817359601795649e-05}
{"train_loss": 0.08626096695661545, "global_step": 205755, "epoch": 2311, "lr": 1.3816959490782705e-05}
{"train_loss": 0.07252408564090729, "global_step": 205756, "epoch": 2311, "lr": 1.3816559384634143e-05}
{"train_loss": 0.05532484129071236, "global_step": 205757, "epoch": 2311, "lr": 1.3816159283349995e-05}
{"train_loss": 0.05430570989847183, "global_step": 205758, "epoch": 2311, "lr": 1.3815759186930333e-05}
{"train_loss": 0.06950666755437851, "global_step": 205759, "epoch": 2311, "lr": 1.3815359095375186e-05}
{"train_loss": 0.09637483954429626, "global_step": 205760, "epoch": 2311, "lr": 1.3814959008684636e-05}
{"train_loss": 0.07213771343231201, "global_step": 205761, "epoch": 2311, "lr": 1.3814558926858711e-05}
{"train_loss": 0.03035706840455532, "global_step": 205762, "epoch": 2311, "lr": 1.381415884989749e-05}
{"train_loss": 0.06786643713712692, "global_step": 205763, "epoch": 2311, "lr": 1.3813758777801006e-05}
{"train_loss": 0.030255720019340515, "global_step": 205764, "epoch": 2311, "lr": 1.381335871056933e-05}
{"train_loss": 0.15554651618003845, "global_step": 205765, "epoch": 2311, "lr": 1.3812958648202495e-05}
{"train_loss": 0.04477445036172867, "global_step": 205766, "epoch": 2311, "lr": 1.381255859070058e-05}
{"train_loss": 0.06571485556327225, "global_step": 205767, "epoch": 2311, "lr": 1.3812158538063614e-05, "val_loss": 8.518594741821289}
{"train_loss": 0.06474994868040085, "global_step": 205768, "epoch": 2312, "lr": 1.3811758490291671e-05}
{"train_loss": 0.057156916707754135, "global_step": 205769, "epoch": 2312, "lr": 1.3811358447384786e-05}
{"train_loss": 0.05761216580867767, "global_step": 205770, "epoch": 2312, "lr": 1.3810958409343033e-05}
{"train_loss": 0.1143561452627182, "global_step": 205771, "epoch": 2312, "lr": 1.3810558376166449e-05}
{"train_loss": 0.04293687641620636, "global_step": 205772, "epoch": 2312, "lr": 1.3810158347855102e-05}
{"train_loss": 0.08831492066383362, "global_step": 205773, "epoch": 2312, "lr": 1.380975832440904e-05}
{"train_loss": 0.028163719922304153, "global_step": 205774, "epoch": 2312, "lr": 1.3809358305828302e-05}
{"train_loss": 0.07977444678544998, "global_step": 205775, "epoch": 2312, "lr": 1.3808958292112967e-05}
{"train_loss": 0.03659548982977867, "global_step": 205776, "epoch": 2312, "lr": 1.3808558283263063e-05}
{"train_loss": 0.0358504019677639, "global_step": 205777, "epoch": 2312, "lr": 1.3808158279278671e-05}
{"train_loss": 0.0432063527405262, "global_step": 205778, "epoch": 2312, "lr": 1.3807758280159815e-05}
{"train_loss": 0.026068292558193207, "global_step": 205779, "epoch": 2312, "lr": 1.380735828590658e-05}
{"train_loss": 0.07974708080291748, "global_step": 205780, "epoch": 2312, "lr": 1.3806958296518991e-05}
{"train_loss": 0.05168263986706734, "global_step": 205781, "epoch": 2312, "lr": 1.3806558311997132e-05}
{"train_loss": 0.03797323256731033, "global_step": 205782, "epoch": 2312, "lr": 1.3806158332341024e-05}
{"train_loss": 0.046284351497888565, "global_step": 205783, "epoch": 2312, "lr": 1.3805758357550746e-05}
{"train_loss": 0.043913476169109344, "global_step": 205784, "epoch": 2312, "lr": 1.3805358387626333e-05}
{"train_loss": 0.14310364425182343, "global_step": 205785, "epoch": 2312, "lr": 1.3804958422567865e-05}
{"train_loss": 0.011600260622799397, "global_step": 205786, "epoch": 2312, "lr": 1.380455846237536e-05}
{"train_loss": 0.07746875286102295, "global_step": 205787, "epoch": 2312, "lr": 1.3804158507048909e-05}
{"train_loss": 0.07984725385904312, "global_step": 205788, "epoch": 2312, "lr": 1.3803758556588531e-05}
{"train_loss": 0.07619216293096542, "global_step": 205789, "epoch": 2312, "lr": 1.3803358610994298e-05}
{"train_loss": 0.06154393404722214, "global_step": 205790, "epoch": 2312, "lr": 1.380295867026628e-05}
{"train_loss": 0.08725059777498245, "global_step": 205791, "epoch": 2312, "lr": 1.3802558734404491e-05}
{"train_loss": 0.025991221889853477, "global_step": 205792, "epoch": 2312, "lr": 1.3802158803409026e-05}
{"train_loss": 0.050171297043561935, "global_step": 205793, "epoch": 2312, "lr": 1.3801758877279908e-05}
{"train_loss": 0.018786685541272163, "global_step": 205794, "epoch": 2312, "lr": 1.3801358956017214e-05}
{"train_loss": 0.05544227734208107, "global_step": 205795, "epoch": 2312, "lr": 1.3800959039620969e-05}
{"train_loss": 0.06473352760076523, "global_step": 205796, "epoch": 2312, "lr": 1.3800559128091256e-05}
{"train_loss": 0.027071308344602585, "global_step": 205797, "epoch": 2312, "lr": 1.380015922142811e-05}
{"train_loss": 0.04869784414768219, "global_step": 205798, "epoch": 2312, "lr": 1.37997593196316e-05}
{"train_loss": 0.036396101117134094, "global_step": 205799, "epoch": 2312, "lr": 1.3799359422701758e-05}
{"train_loss": 0.033409006893634796, "global_step": 205800, "epoch": 2312, "lr": 1.3798959530638666e-05}
{"train_loss": 0.06665383279323578, "global_step": 205801, "epoch": 2312, "lr": 1.3798559643442349e-05}
{"train_loss": 0.06869518011808395, "global_step": 205802, "epoch": 2312, "lr": 1.3798159761112889e-05}
{"train_loss": 0.06294785439968109, "global_step": 205803, "epoch": 2312, "lr": 1.3797759883650307e-05}
{"train_loss": 0.030698595568537712, "global_step": 205804, "epoch": 2312, "lr": 1.3797360011054695e-05}
{"train_loss": 0.05299900472164154, "global_step": 205805, "epoch": 2312, "lr": 1.3796960143326071e-05}
{"train_loss": 0.11311101913452148, "global_step": 205806, "epoch": 2312, "lr": 1.3796560280464516e-05}
{"train_loss": 0.04675687849521637, "global_step": 205807, "epoch": 2312, "lr": 1.3796160422470062e-05}
{"train_loss": 0.0691455826163292, "global_step": 205808, "epoch": 2312, "lr": 1.379576056934278e-05}
{"train_loss": 0.009729234501719475, "global_step": 205809, "epoch": 2312, "lr": 1.3795360721082706e-05}
{"train_loss": 0.06434454768896103, "global_step": 205810, "epoch": 2312, "lr": 1.3794960877689922e-05}
{"train_loss": 0.05835074186325073, "global_step": 205811, "epoch": 2312, "lr": 1.379456103916446e-05}
{"train_loss": 0.08579205721616745, "global_step": 205812, "epoch": 2312, "lr": 1.3794161205506367e-05}
{"train_loss": 0.10090141743421555, "global_step": 205813, "epoch": 2312, "lr": 1.3793761376715725e-05}
{"train_loss": 0.06672155112028122, "global_step": 205814, "epoch": 2312, "lr": 1.379336155279255e-05}
{"train_loss": 0.06537764519453049, "global_step": 205815, "epoch": 2312, "lr": 1.379296173373693e-05}
{"train_loss": 0.07222788035869598, "global_step": 205816, "epoch": 2312, "lr": 1.3792561919548897e-05}
{"train_loss": 0.08624693006277084, "global_step": 205817, "epoch": 2312, "lr": 1.3792162110228518e-05}
{"train_loss": 0.055934324860572815, "global_step": 205818, "epoch": 2312, "lr": 1.3791762305775835e-05}
{"train_loss": 0.1012924313545227, "global_step": 205819, "epoch": 2312, "lr": 1.3791362506190919e-05}
{"train_loss": 0.07841426134109497, "global_step": 205820, "epoch": 2312, "lr": 1.3790962711473798e-05}
{"train_loss": 0.07216066122055054, "global_step": 205821, "epoch": 2312, "lr": 1.3790562921624562e-05}
{"train_loss": 0.10305105149745941, "global_step": 205822, "epoch": 2312, "lr": 1.3790163136643219e-05}
{"train_loss": 0.07142957299947739, "global_step": 205823, "epoch": 2312, "lr": 1.3789763356529868e-05}
{"train_loss": 0.026252301409840584, "global_step": 205824, "epoch": 2312, "lr": 1.3789363581284526e-05}
{"train_loss": 0.04653170332312584, "global_step": 205825, "epoch": 2312, "lr": 1.378896381090728e-05}
{"train_loss": 0.06210020184516907, "global_step": 205826, "epoch": 2312, "lr": 1.3788564045398144e-05}
{"train_loss": 0.032150596380233765, "global_step": 205827, "epoch": 2312, "lr": 1.37881642847572e-05}
{"train_loss": 0.11096175760030746, "global_step": 205828, "epoch": 2312, "lr": 1.3787764528984509e-05}
{"train_loss": 0.10535617917776108, "global_step": 205829, "epoch": 2312, "lr": 1.3787364778080097e-05}
{"train_loss": 0.023976387456059456, "global_step": 205830, "epoch": 2312, "lr": 1.3786965032044052e-05}
{"train_loss": 0.07133415341377258, "global_step": 205831, "epoch": 2312, "lr": 1.3786565290876385e-05}
{"train_loss": 0.06825720518827438, "global_step": 205832, "epoch": 2312, "lr": 1.3786165554577197e-05}
{"train_loss": 0.056396372616291046, "global_step": 205833, "epoch": 2312, "lr": 1.3785765823146496e-05}
{"train_loss": 0.05585559457540512, "global_step": 205834, "epoch": 2312, "lr": 1.3785366096584374e-05}
{"train_loss": 0.049057915806770325, "global_step": 205835, "epoch": 2312, "lr": 1.3784966374890856e-05}
{"train_loss": 0.06133190169930458, "global_step": 205836, "epoch": 2312, "lr": 1.3784566658066018e-05}
{"train_loss": 0.04773940518498421, "global_step": 205837, "epoch": 2312, "lr": 1.378416694610989e-05}
{"train_loss": 0.08901051431894302, "global_step": 205838, "epoch": 2312, "lr": 1.3783767239022555e-05}
{"train_loss": 0.055895861238241196, "global_step": 205839, "epoch": 2312, "lr": 1.3783367536804038e-05}
{"train_loss": 0.055077724158763885, "global_step": 205840, "epoch": 2312, "lr": 1.378296783945442e-05}
{"train_loss": 0.05525176227092743, "global_step": 205841, "epoch": 2312, "lr": 1.3782568146973723e-05}
{"train_loss": 0.06948278099298477, "global_step": 205842, "epoch": 2312, "lr": 1.3782168459362038e-05}
{"train_loss": 0.07069294899702072, "global_step": 205843, "epoch": 2312, "lr": 1.3781768776619381e-05}
{"train_loss": 0.07169567793607712, "global_step": 205844, "epoch": 2312, "lr": 1.3781369098745822e-05}
{"train_loss": 0.04058465361595154, "global_step": 205845, "epoch": 2312, "lr": 1.378096942574143e-05}
{"train_loss": 0.055265143513679504, "global_step": 205846, "epoch": 2312, "lr": 1.3780569757606248e-05}
{"train_loss": 0.05301119387149811, "global_step": 205847, "epoch": 2312, "lr": 1.3780170094340317e-05}
{"train_loss": 0.08216216415166855, "global_step": 205848, "epoch": 2312, "lr": 1.3779770435943712e-05}
{"train_loss": 0.07984461635351181, "global_step": 205849, "epoch": 2312, "lr": 1.3779370782416456e-05}
{"train_loss": 0.078009232878685, "global_step": 205850, "epoch": 2312, "lr": 1.3778971133758639e-05}
{"train_loss": 0.031395383179187775, "global_step": 205851, "epoch": 2312, "lr": 1.3778571489970304e-05}
{"train_loss": 0.0954141914844513, "global_step": 205852, "epoch": 2312, "lr": 1.3778171851051475e-05}
{"train_loss": 0.0665167048573494, "global_step": 205853, "epoch": 2312, "lr": 1.3777772217002249e-05}
{"train_loss": 0.03360400348901749, "global_step": 205854, "epoch": 2312, "lr": 1.3777372587822645e-05}
{"train_loss": 0.03771550953388214, "global_step": 205855, "epoch": 2312, "lr": 1.3776972963512746e-05}
{"train_loss": 0.061294989608060776, "global_step": 205856, "epoch": 2312, "lr": 1.377657334407258e-05, "val_loss": 8.4193754196167}
{"train_loss": 0.08044470846652985, "global_step": 205857, "epoch": 2313, "lr": 1.3776173729502223e-05}
{"train_loss": 0.03340670466423035, "global_step": 205858, "epoch": 2313, "lr": 1.3775774119801704e-05}
{"train_loss": 0.0631716251373291, "global_step": 205859, "epoch": 2313, "lr": 1.3775374514971101e-05}
{"train_loss": 0.03519916161894798, "global_step": 205860, "epoch": 2313, "lr": 1.3774974915010446e-05}
{"train_loss": 0.061325784772634506, "global_step": 205861, "epoch": 2313, "lr": 1.3774575319919825e-05}
{"train_loss": 0.02393219619989395, "global_step": 205862, "epoch": 2313, "lr": 1.3774175729699246e-05}
{"train_loss": 0.027830755338072777, "global_step": 205863, "epoch": 2313, "lr": 1.3773776144348793e-05}
{"train_loss": 0.021934153512120247, "global_step": 205864, "epoch": 2313, "lr": 1.3773376563868529e-05}
{"train_loss": 0.07001539319753647, "global_step": 205865, "epoch": 2313, "lr": 1.3772976988258479e-05}
{"train_loss": 0.0834444910287857, "global_step": 205866, "epoch": 2313, "lr": 1.3772577417518723e-05}
{"train_loss": 0.044692810624837875, "global_step": 205867, "epoch": 2313, "lr": 1.3772177851649287e-05}
{"train_loss": 0.02262730523943901, "global_step": 205868, "epoch": 2313, "lr": 1.3771778290650261e-05}
{"train_loss": 0.05753353238105774, "global_step": 205869, "epoch": 2313, "lr": 1.3771378734521656e-05}
{"train_loss": 0.11450778692960739, "global_step": 205870, "epoch": 2313, "lr": 1.3770979183263566e-05}
{"train_loss": 0.056796859949827194, "global_step": 205871, "epoch": 2313, "lr": 1.3770579636876013e-05}
{"train_loss": 0.06641606241464615, "global_step": 205872, "epoch": 2313, "lr": 1.3770180095359075e-05}
{"train_loss": 0.04189887270331383, "global_step": 205873, "epoch": 2313, "lr": 1.376978055871278e-05}
{"train_loss": 0.05255706235766411, "global_step": 205874, "epoch": 2313, "lr": 1.3769381026937212e-05}
{"train_loss": 0.10013887286186218, "global_step": 205875, "epoch": 2313, "lr": 1.3768981500032396e-05}
{"train_loss": 0.022390687838196754, "global_step": 205876, "epoch": 2313, "lr": 1.3768581977998412e-05}
{"train_loss": 0.04871809482574463, "global_step": 205877, "epoch": 2313, "lr": 1.3768182460835289e-05}
{"train_loss": 0.08854614943265915, "global_step": 205878, "epoch": 2313, "lr": 1.3767782948543102e-05}
{"train_loss": 0.0581396222114563, "global_step": 205879, "epoch": 2313, "lr": 1.3767383441121884e-05}
{"train_loss": 0.09419173002243042, "global_step": 205880, "epoch": 2313, "lr": 1.3766983938571699e-05}
{"train_loss": 0.0552932545542717, "global_step": 205881, "epoch": 2313, "lr": 1.376658444089261e-05}
{"train_loss": 0.06851600110530853, "global_step": 205882, "epoch": 2313, "lr": 1.3766184948084654e-05}
{"train_loss": 0.060915619134902954, "global_step": 205883, "epoch": 2313, "lr": 1.3765785460147906e-05}
{"train_loss": 0.025123395025730133, "global_step": 205884, "epoch": 2313, "lr": 1.3765385977082406e-05}
{"train_loss": 0.05920817330479622, "global_step": 205885, "epoch": 2313, "lr": 1.3764986498888194e-05}
{"train_loss": 0.07508774101734161, "global_step": 205886, "epoch": 2313, "lr": 1.3764587025565357e-05}
{"train_loss": 0.05068901926279068, "global_step": 205887, "epoch": 2313, "lr": 1.3764187557113906e-05}
{"train_loss": 0.0714869275689125, "global_step": 205888, "epoch": 2313, "lr": 1.3763788093533941e-05}
{"train_loss": 0.049638569355010986, "global_step": 205889, "epoch": 2313, "lr": 1.3763388634825491e-05}
{"train_loss": 0.08120205253362656, "global_step": 205890, "epoch": 2313, "lr": 1.3762989180988595e-05}
{"train_loss": 0.09284082800149918, "global_step": 205891, "epoch": 2313, "lr": 1.376258973202334e-05}
{"train_loss": 0.04717118665575981, "global_step": 205892, "epoch": 2313, "lr": 1.3762190287929744e-05}
{"train_loss": 0.0640062615275383, "global_step": 205893, "epoch": 2313, "lr": 1.3761790848707901e-05}
{"train_loss": 0.04304938763380051, "global_step": 205894, "epoch": 2313, "lr": 1.3761391414357822e-05}
{"train_loss": 0.05956408381462097, "global_step": 205895, "epoch": 2313, "lr": 1.37609919848796e-05}
{"train_loss": 0.05933813005685806, "global_step": 205896, "epoch": 2313, "lr": 1.3760592560273261e-05}
{"train_loss": 0.0835341364145279, "global_step": 205897, "epoch": 2313, "lr": 1.376019314053888e-05}
{"train_loss": 0.049790360033512115, "global_step": 205898, "epoch": 2313, "lr": 1.3759793725676485e-05}
{"train_loss": 0.06964434683322906, "global_step": 205899, "epoch": 2313, "lr": 1.3759394315686141e-05}
{"train_loss": 0.08138269186019897, "global_step": 205900, "epoch": 2313, "lr": 1.3758994910567918e-05}
{"train_loss": 0.03342173993587494, "global_step": 205901, "epoch": 2313, "lr": 1.3758595510321847e-05}
{"train_loss": 0.056661754846572876, "global_step": 205902, "epoch": 2313, "lr": 1.3758196114948003e-05}
{"train_loss": 0.046506673097610474, "global_step": 205903, "epoch": 2313, "lr": 1.3757796724446415e-05}
{"train_loss": 0.06233564019203186, "global_step": 205904, "epoch": 2313, "lr": 1.3757397338817163e-05}
{"train_loss": 0.07648459821939468, "global_step": 205905, "epoch": 2313, "lr": 1.3756997958060275e-05}
{"train_loss": 0.03712251037359238, "global_step": 205906, "epoch": 2313, "lr": 1.3756598582175829e-05}
{"train_loss": 0.0516875796020031, "global_step": 205907, "epoch": 2313, "lr": 1.3756199211163851e-05}
{"train_loss": 0.06835956126451492, "global_step": 205908, "epoch": 2313, "lr": 1.3755799845024426e-05}
{"train_loss": 0.10866613686084747, "global_step": 205909, "epoch": 2313, "lr": 1.3755400483757574e-05}
{"train_loss": 0.0876035988330841, "global_step": 205910, "epoch": 2313, "lr": 1.3755001127363386e-05}
{"train_loss": 0.039462555199861526, "global_step": 205911, "epoch": 2313, "lr": 1.375460177584188e-05}
{"train_loss": 0.07303228229284286, "global_step": 205912, "epoch": 2313, "lr": 1.3754202429193141e-05}
{"train_loss": 0.055747244507074356, "global_step": 205913, "epoch": 2313, "lr": 1.3753803087417194e-05}
{"train_loss": 0.07773049920797348, "global_step": 205914, "epoch": 2313, "lr": 1.3753403750514122e-05}
{"train_loss": 0.072933629155159, "global_step": 205915, "epoch": 2313, "lr": 1.3753004418483944e-05}
{"train_loss": 0.05447665974497795, "global_step": 205916, "epoch": 2313, "lr": 1.3752605091326736e-05}
{"train_loss": 0.07700386643409729, "global_step": 205917, "epoch": 2313, "lr": 1.3752205769042564e-05}
{"train_loss": 0.03466169536113739, "global_step": 205918, "epoch": 2313, "lr": 1.375180645163145e-05}
{"train_loss": 0.0831964835524559, "global_step": 205919, "epoch": 2313, "lr": 1.375140713909348e-05}
{"train_loss": 0.04925234988331795, "global_step": 205920, "epoch": 2313, "lr": 1.3751007831428674e-05}
{"train_loss": 0.06532890349626541, "global_step": 205921, "epoch": 2313, "lr": 1.3750608528637122e-05}
{"train_loss": 0.05743491277098656, "global_step": 205922, "epoch": 2313, "lr": 1.3750209230718853e-05}
{"train_loss": 0.04598725214600563, "global_step": 205923, "epoch": 2313, "lr": 1.3749809937673914e-05}
{"train_loss": 0.044989146292209625, "global_step": 205924, "epoch": 2313, "lr": 1.3749410649502387e-05}
{"train_loss": 0.05112694948911667, "global_step": 205925, "epoch": 2313, "lr": 1.3749011366204295e-05}
{"train_loss": 0.08584797382354736, "global_step": 205926, "epoch": 2313, "lr": 1.3748612087779722e-05}
{"train_loss": 0.03619211167097092, "global_step": 205927, "epoch": 2313, "lr": 1.3748212814228706e-05}
{"train_loss": 0.08302777260541916, "global_step": 205928, "epoch": 2313, "lr": 1.3747813545551287e-05}
{"train_loss": 0.04427479952573776, "global_step": 205929, "epoch": 2313, "lr": 1.3747414281747544e-05}
{"train_loss": 0.09456346929073334, "global_step": 205930, "epoch": 2313, "lr": 1.3747015022817506e-05}
{"train_loss": 0.08360562473535538, "global_step": 205931, "epoch": 2313, "lr": 1.3746615768761256e-05}
{"train_loss": 0.08375056833028793, "global_step": 205932, "epoch": 2313, "lr": 1.3746216519578813e-05}
{"train_loss": 0.05407026782631874, "global_step": 205933, "epoch": 2313, "lr": 1.3745817275270273e-05}
{"train_loss": 0.059725649654865265, "global_step": 205934, "epoch": 2313, "lr": 1.3745418035835645e-05}
{"train_loss": 0.04363993555307388, "global_step": 205935, "epoch": 2313, "lr": 1.3745018801275e-05}
{"train_loss": 0.04889119789004326, "global_step": 205936, "epoch": 2313, "lr": 1.3744619571588418e-05}
{"train_loss": 0.06275755167007446, "global_step": 205937, "epoch": 2313, "lr": 1.374422034677591e-05}
{"train_loss": 0.048992011696100235, "global_step": 205938, "epoch": 2313, "lr": 1.3743821126837564e-05}
{"train_loss": 0.041077084839344025, "global_step": 205939, "epoch": 2313, "lr": 1.3743421911773407e-05}
{"train_loss": 0.04516103118658066, "global_step": 205940, "epoch": 2313, "lr": 1.3743022701583518e-05}
{"train_loss": 0.09740202128887177, "global_step": 205941, "epoch": 2313, "lr": 1.3742623496267926e-05}
{"train_loss": 0.05338622257113457, "global_step": 205942, "epoch": 2313, "lr": 1.3742224295826712e-05}
{"train_loss": 0.03289460763335228, "global_step": 205943, "epoch": 2313, "lr": 1.3741825100259898e-05}
{"train_loss": 0.05094754695892334, "global_step": 205944, "epoch": 2313, "lr": 1.3741425909567562e-05}
{"train_loss": 0.060013749312316435, "global_step": 205945, "epoch": 2313, "lr": 1.3741026723749745e-05, "val_loss": 8.428889274597168}
{"train_loss": 0.0855945274233818, "global_step": 205946, "epoch": 2314, "lr": 1.3740627542806511e-05}
{"train_loss": 0.05073848366737366, "global_step": 205947, "epoch": 2314, "lr": 1.3740228366737894e-05}
{"train_loss": 0.05286506563425064, "global_step": 205948, "epoch": 2314, "lr": 1.373982919554398e-05}
{"train_loss": 0.05789618939161301, "global_step": 205949, "epoch": 2314, "lr": 1.3739430029224793e-05}
{"train_loss": 0.041063860058784485, "global_step": 205950, "epoch": 2314, "lr": 1.3739030867780407e-05}
{"train_loss": 0.07547089457511902, "global_step": 205951, "epoch": 2314, "lr": 1.3738631711210847e-05}
{"train_loss": 0.05494765564799309, "global_step": 205952, "epoch": 2314, "lr": 1.3738232559516207e-05}
{"train_loss": 0.05417223647236824, "global_step": 205953, "epoch": 2314, "lr": 1.3737833412696505e-05}
{"train_loss": 0.01775895245373249, "global_step": 205954, "epoch": 2314, "lr": 1.373743427075181e-05}
{"train_loss": 0.0861428827047348, "global_step": 205955, "epoch": 2314, "lr": 1.3737035133682185e-05}
{"train_loss": 0.1072314977645874, "global_step": 205956, "epoch": 2314, "lr": 1.3736636001487663e-05}
{"train_loss": 0.07090708613395691, "global_step": 205957, "epoch": 2314, "lr": 1.3736236874168318e-05}
{"train_loss": 0.08955613523721695, "global_step": 205958, "epoch": 2314, "lr": 1.3735837751724184e-05}
{"train_loss": 0.07050053775310516, "global_step": 205959, "epoch": 2314, "lr": 1.3735438634155339e-05}
{"train_loss": 0.09195930510759354, "global_step": 205960, "epoch": 2314, "lr": 1.3735039521461807e-05}
{"train_loss": 0.05487498268485069, "global_step": 205961, "epoch": 2314, "lr": 1.3734640413643668e-05}
{"train_loss": 0.0812826007604599, "global_step": 205962, "epoch": 2314, "lr": 1.3734241310700969e-05}
{"train_loss": 0.045133013278245926, "global_step": 205963, "epoch": 2314, "lr": 1.3733842212633746e-05}
{"train_loss": 0.021004613488912582, "global_step": 205964, "epoch": 2314, "lr": 1.3733443119442075e-05}
{"train_loss": 0.04407527297735214, "global_step": 205965, "epoch": 2314, "lr": 1.3733044031125992e-05}
{"train_loss": 0.05392857640981674, "global_step": 205966, "epoch": 2314, "lr": 1.3732644947685569e-05}
{"train_loss": 0.0383235327899456, "global_step": 205967, "epoch": 2314, "lr": 1.3732245869120853e-05}
{"train_loss": 0.0619671531021595, "global_step": 205968, "epoch": 2314, "lr": 1.3731846795431873e-05}
{"train_loss": 0.06767192482948303, "global_step": 205969, "epoch": 2314, "lr": 1.3731447726618729e-05}
{"train_loss": 0.03441939502954483, "global_step": 205970, "epoch": 2314, "lr": 1.3731048662681428e-05}
{"train_loss": 0.06059650704264641, "global_step": 205971, "epoch": 2314, "lr": 1.373064960362005e-05}
{"train_loss": 0.054307106882333755, "global_step": 205972, "epoch": 2314, "lr": 1.3730250549434653e-05}
{"train_loss": 0.045046452432870865, "global_step": 205973, "epoch": 2314, "lr": 1.3729851500125274e-05}
{"train_loss": 0.03732839971780777, "global_step": 205974, "epoch": 2314, "lr": 1.3729452455691982e-05}
{"train_loss": 0.04099778085947037, "global_step": 205975, "epoch": 2314, "lr": 1.3729053416134812e-05}
{"train_loss": 0.07297071069478989, "global_step": 205976, "epoch": 2314, "lr": 1.3728654381453848e-05}
{"train_loss": 0.03859710320830345, "global_step": 205977, "epoch": 2314, "lr": 1.3728255351649105e-05}
{"train_loss": 0.054763589054346085, "global_step": 205978, "epoch": 2314, "lr": 1.3727856326720668e-05}
{"train_loss": 0.033390045166015625, "global_step": 205979, "epoch": 2314, "lr": 1.3727457306668568e-05}
{"train_loss": 0.12277643382549286, "global_step": 205980, "epoch": 2314, "lr": 1.3727058291492883e-05}
{"train_loss": 0.03580503165721893, "global_step": 205981, "epoch": 2314, "lr": 1.3726659281193637e-05}
{"train_loss": 0.06662850081920624, "global_step": 205982, "epoch": 2314, "lr": 1.3726260275770914e-05}
{"train_loss": 0.02838199958205223, "global_step": 205983, "epoch": 2314, "lr": 1.3725861275224739e-05}
{"train_loss": 0.024247214198112488, "global_step": 205984, "epoch": 2314, "lr": 1.3725462279555196e-05}
{"train_loss": 0.06868387013673782, "global_step": 205985, "epoch": 2314, "lr": 1.3725063288762308e-05}
{"train_loss": 0.0526677630841732, "global_step": 205986, "epoch": 2314, "lr": 1.3724664302846157e-05}
{"train_loss": 0.060598913580179214, "global_step": 205987, "epoch": 2314, "lr": 1.3724265321806767e-05}
{"train_loss": 0.0647202804684639, "global_step": 205988, "epoch": 2314, "lr": 1.3723866345644221e-05}
{"train_loss": 0.029470378533005714, "global_step": 205989, "epoch": 2314, "lr": 1.3723467374358545e-05}
{"train_loss": 0.013764538802206516, "global_step": 205990, "epoch": 2314, "lr": 1.3723068407949812e-05}
{"train_loss": 0.057845376431941986, "global_step": 205991, "epoch": 2314, "lr": 1.372266944641808e-05}
{"train_loss": 0.06210551783442497, "global_step": 205992, "epoch": 2314, "lr": 1.3722270489763378e-05}
{"train_loss": 0.06674327701330185, "global_step": 205993, "epoch": 2314, "lr": 1.3721871537985792e-05}
{"train_loss": 0.081489197909832, "global_step": 205994, "epoch": 2314, "lr": 1.3721472591085338e-05}
{"train_loss": 0.12555468082427979, "global_step": 205995, "epoch": 2314, "lr": 1.372107364906211e-05}
{"train_loss": 0.10252273827791214, "global_step": 205996, "epoch": 2314, "lr": 1.3720674711916126e-05}
{"train_loss": 0.079204261302948, "global_step": 205997, "epoch": 2314, "lr": 1.3720275779647468e-05}
{"train_loss": 0.07350118458271027, "global_step": 205998, "epoch": 2314, "lr": 1.3719876852256164e-05}
{"train_loss": 0.05357077717781067, "global_step": 205999, "epoch": 2314, "lr": 1.3719477929742291e-05}
{"train_loss": 0.04980340600013733, "global_step": 206000, "epoch": 2314, "lr": 1.3719079012105901e-05}
{"train_loss": 0.11348705738782883, "global_step": 206001, "epoch": 2314, "lr": 1.3718680099347014e-05}
{"train_loss": 0.07438510656356812, "global_step": 206002, "epoch": 2314, "lr": 1.3718281191465726e-05}
{"train_loss": 0.0791497454047203, "global_step": 206003, "epoch": 2314, "lr": 1.3717882288462058e-05}
{"train_loss": 0.09711042791604996, "global_step": 206004, "epoch": 2314, "lr": 1.37174833903361e-05}
{"train_loss": 0.11134205013513565, "global_step": 206005, "epoch": 2314, "lr": 1.371708449708788e-05}
{"train_loss": 0.034873925149440765, "global_step": 206006, "epoch": 2314, "lr": 1.3716685608717433e-05}
{"train_loss": 0.14861443638801575, "global_step": 206007, "epoch": 2314, "lr": 1.3716286725224842e-05}
{"train_loss": 0.04740719497203827, "global_step": 206008, "epoch": 2314, "lr": 1.371588784661017e-05}
{"train_loss": 0.06865382194519043, "global_step": 206009, "epoch": 2314, "lr": 1.3715488972873442e-05}
{"train_loss": 0.060216911137104034, "global_step": 206010, "epoch": 2314, "lr": 1.3715090104014732e-05}
{"train_loss": 0.06628412753343582, "global_step": 206011, "epoch": 2314, "lr": 1.3714691240034078e-05}
{"train_loss": 0.05838233605027199, "global_step": 206012, "epoch": 2314, "lr": 1.3714292380931548e-05}
{"train_loss": 0.07764822244644165, "global_step": 206013, "epoch": 2314, "lr": 1.3713893526707184e-05}
{"train_loss": 0.047470785677433014, "global_step": 206014, "epoch": 2314, "lr": 1.3713494677361056e-05}
{"train_loss": 0.047108523547649384, "global_step": 206015, "epoch": 2314, "lr": 1.3713095832893186e-05}
{"train_loss": 0.03997752070426941, "global_step": 206016, "epoch": 2314, "lr": 1.371269699330367e-05}
{"train_loss": 0.10723870247602463, "global_step": 206017, "epoch": 2314, "lr": 1.3712298158592518e-05}
{"train_loss": 0.06755336374044418, "global_step": 206018, "epoch": 2314, "lr": 1.3711899328759825e-05}
{"train_loss": 0.08523785322904587, "global_step": 206019, "epoch": 2314, "lr": 1.3711500503805608e-05}
{"train_loss": 0.05888400599360466, "global_step": 206020, "epoch": 2314, "lr": 1.3711101683729955e-05}
{"train_loss": 0.05568457022309303, "global_step": 206021, "epoch": 2314, "lr": 1.3710702868532888e-05}
{"train_loss": 0.04504956305027008, "global_step": 206022, "epoch": 2314, "lr": 1.3710304058214485e-05}
{"train_loss": 0.057445913553237915, "global_step": 206023, "epoch": 2314, "lr": 1.3709905252774773e-05}
{"train_loss": 0.04783288761973381, "global_step": 206024, "epoch": 2314, "lr": 1.3709506452213844e-05}
{"train_loss": 0.03948361426591873, "global_step": 206025, "epoch": 2314, "lr": 1.3709107656531712e-05}
{"train_loss": 0.07576311379671097, "global_step": 206026, "epoch": 2314, "lr": 1.3708708865728448e-05}
{"train_loss": 0.05898743495345116, "global_step": 206027, "epoch": 2314, "lr": 1.3708310079804115e-05}
{"train_loss": 0.08740969747304916, "global_step": 206028, "epoch": 2314, "lr": 1.3707911298758752e-05}
{"train_loss": 0.06084951013326645, "global_step": 206029, "epoch": 2314, "lr": 1.3707512522592425e-05}
{"train_loss": 0.04272814095020294, "global_step": 206030, "epoch": 2314, "lr": 1.3707113751305167e-05}
{"train_loss": 0.05594376474618912, "global_step": 206031, "epoch": 2314, "lr": 1.3706714984897067e-05}
{"train_loss": 0.043336812406778336, "global_step": 206032, "epoch": 2314, "lr": 1.3706316223368138e-05}
{"train_loss": 0.08526808768510818, "global_step": 206033, "epoch": 2314, "lr": 1.3705917466718466e-05}
{"train_loss": 0.06265198980363902, "global_step": 206034, "epoch": 2314, "lr": 1.3705518714948073e-05, "val_loss": 8.498907089233398}
{"train_loss": 0.05813268572092056, "global_step": 206035, "epoch": 2315, "lr": 1.3705119968057051e-05}
{"train_loss": 0.08332375437021255, "global_step": 206036, "epoch": 2315, "lr": 1.3704721226045414e-05}
{"train_loss": 0.052059076726436615, "global_step": 206037, "epoch": 2315, "lr": 1.3704322488913252e-05}
{"train_loss": 0.09501301497220993, "global_step": 206038, "epoch": 2315, "lr": 1.3703923756660602e-05}
{"train_loss": 0.04457299783825874, "global_step": 206039, "epoch": 2315, "lr": 1.3703525029287501e-05}
{"train_loss": 0.055097006261348724, "global_step": 206040, "epoch": 2315, "lr": 1.3703126306794035e-05}
{"train_loss": 0.02972446009516716, "global_step": 206041, "epoch": 2315, "lr": 1.3702727589180225e-05}
{"train_loss": 0.05064418911933899, "global_step": 206042, "epoch": 2315, "lr": 1.3702328876446157e-05}
{"train_loss": 0.05949598178267479, "global_step": 206043, "epoch": 2315, "lr": 1.3701930168591853e-05}
{"train_loss": 0.01472114771604538, "global_step": 206044, "epoch": 2315, "lr": 1.3701531465617395e-05}
{"train_loss": 0.08467685431241989, "global_step": 206045, "epoch": 2315, "lr": 1.3701132767522812e-05}
{"train_loss": 0.059632472693920135, "global_step": 206046, "epoch": 2315, "lr": 1.3700734074308185e-05}
{"train_loss": 0.08313024789094925, "global_step": 206047, "epoch": 2315, "lr": 1.3700335385973535e-05}
{"train_loss": 0.05240019038319588, "global_step": 206048, "epoch": 2315, "lr": 1.3699936702518944e-05}
{"train_loss": 0.06558059900999069, "global_step": 206049, "epoch": 2315, "lr": 1.3699538023944441e-05}
{"train_loss": 0.06028963625431061, "global_step": 206050, "epoch": 2315, "lr": 1.3699139350250112e-05}
{"train_loss": 0.04124068096280098, "global_step": 206051, "epoch": 2315, "lr": 1.3698740681435972e-05}
{"train_loss": 0.059241801500320435, "global_step": 206052, "epoch": 2315, "lr": 1.3698342017502114e-05}
{"train_loss": 0.022296208888292313, "global_step": 206053, "epoch": 2315, "lr": 1.369794335844855e-05}
{"train_loss": 0.04028654843568802, "global_step": 206054, "epoch": 2315, "lr": 1.3697544704275372e-05}
{"train_loss": 0.039906371384859085, "global_step": 206055, "epoch": 2315, "lr": 1.3697146054982602e-05}
{"train_loss": 0.08477114140987396, "global_step": 206056, "epoch": 2315, "lr": 1.3696747410570326e-05}
{"train_loss": 0.10147470235824585, "global_step": 206057, "epoch": 2315, "lr": 1.3696348771038558e-05}
{"train_loss": 0.05796017497777939, "global_step": 206058, "epoch": 2315, "lr": 1.3695950136387397e-05}
{"train_loss": 0.09630683809518814, "global_step": 206059, "epoch": 2315, "lr": 1.3695551506616854e-05}
{"train_loss": 0.060520101338624954, "global_step": 206060, "epoch": 2315, "lr": 1.3695152881727013e-05}
{"train_loss": 0.07208199799060822, "global_step": 206061, "epoch": 2315, "lr": 1.3694754261717908e-05}
{"train_loss": 0.08136510103940964, "global_step": 206062, "epoch": 2315, "lr": 1.3694355646589596e-05}
{"train_loss": 0.05670229718089104, "global_step": 206063, "epoch": 2315, "lr": 1.3693957036342154e-05}
{"train_loss": 0.008088916540145874, "global_step": 206064, "epoch": 2315, "lr": 1.3693558430975595e-05}
{"train_loss": 0.07607721537351608, "global_step": 206065, "epoch": 2315, "lr": 1.3693159830490015e-05}
{"train_loss": 0.0694098174571991, "global_step": 206066, "epoch": 2315, "lr": 1.3692761234885432e-05}
{"train_loss": 0.05923204496502876, "global_step": 206067, "epoch": 2315, "lr": 1.3692362644161933e-05}
{"train_loss": 0.07840970158576965, "global_step": 206068, "epoch": 2315, "lr": 1.3691964058319535e-05}
{"train_loss": 0.05347517877817154, "global_step": 206069, "epoch": 2315, "lr": 1.369156547735832e-05}
{"train_loss": 0.07293611764907837, "global_step": 206070, "epoch": 2315, "lr": 1.3691166901278324e-05}
{"train_loss": 0.06646472215652466, "global_step": 206071, "epoch": 2315, "lr": 1.3690768330079617e-05}
{"train_loss": 0.08252005279064178, "global_step": 206072, "epoch": 2315, "lr": 1.3690369763762235e-05}
{"train_loss": 0.050255995243787766, "global_step": 206073, "epoch": 2315, "lr": 1.3689971202326252e-05}
{"train_loss": 0.1741519272327423, "global_step": 206074, "epoch": 2315, "lr": 1.3689572645771692e-05}
{"train_loss": 0.06405449658632278, "global_step": 206075, "epoch": 2315, "lr": 1.3689174094098645e-05}
{"train_loss": 0.046876005828380585, "global_step": 206076, "epoch": 2315, "lr": 1.368877554730713e-05}
{"train_loss": 0.0756622925400734, "global_step": 206077, "epoch": 2315, "lr": 1.368837700539723e-05}
{"train_loss": 0.07613285630941391, "global_step": 206078, "epoch": 2315, "lr": 1.3687978468368989e-05}
{"train_loss": 0.030289893969893456, "global_step": 206079, "epoch": 2315, "lr": 1.3687579936222439e-05}
{"train_loss": 0.022349713370203972, "global_step": 206080, "epoch": 2315, "lr": 1.3687181408957666e-05}
{"train_loss": 0.06989402323961258, "global_step": 206081, "epoch": 2315, "lr": 1.3686782886574695e-05}
{"train_loss": 0.025217551738023758, "global_step": 206082, "epoch": 2315, "lr": 1.3686384369073607e-05}
{"train_loss": 0.11007802188396454, "global_step": 206083, "epoch": 2315, "lr": 1.368598585645443e-05}
{"train_loss": 0.060586243867874146, "global_step": 206084, "epoch": 2315, "lr": 1.3685587348717243e-05}
{"train_loss": 0.08613386750221252, "global_step": 206085, "epoch": 2315, "lr": 1.3685188845862068e-05}
{"train_loss": 0.033581409603357315, "global_step": 206086, "epoch": 2315, "lr": 1.3684790347888999e-05}
{"train_loss": 0.052788231521844864, "global_step": 206087, "epoch": 2315, "lr": 1.3684391854798045e-05}
{"train_loss": 0.05786198750138283, "global_step": 206088, "epoch": 2315, "lr": 1.3683993366589299e-05}
{"train_loss": 0.06267097592353821, "global_step": 206089, "epoch": 2315, "lr": 1.3683594883262785e-05}
{"train_loss": 0.04687456786632538, "global_step": 206090, "epoch": 2315, "lr": 1.3683196404818583e-05}
{"train_loss": 0.06210000440478325, "global_step": 206091, "epoch": 2315, "lr": 1.3682797931256714e-05}
{"train_loss": 0.07569090276956558, "global_step": 206092, "epoch": 2315, "lr": 1.3682399462577266e-05}
{"train_loss": 0.08618999272584915, "global_step": 206093, "epoch": 2315, "lr": 1.3682000998780264e-05}
{"train_loss": 0.043448176234960556, "global_step": 206094, "epoch": 2315, "lr": 1.368160253986579e-05}
{"train_loss": 0.04711048677563667, "global_step": 206095, "epoch": 2315, "lr": 1.3681204085833865e-05}
{"train_loss": 0.07617013901472092, "global_step": 206096, "epoch": 2315, "lr": 1.368080563668458e-05}
{"train_loss": 0.1275569349527359, "global_step": 206097, "epoch": 2315, "lr": 1.3680407192417948e-05}
{"train_loss": 0.08234556019306183, "global_step": 206098, "epoch": 2315, "lr": 1.368000875303404e-05}
{"train_loss": 0.05058731511235237, "global_step": 206099, "epoch": 2315, "lr": 1.3679610318532931e-05}
{"train_loss": 0.03693271428346634, "global_step": 206100, "epoch": 2315, "lr": 1.367921188891464e-05}
{"train_loss": 0.03794049099087715, "global_step": 206101, "epoch": 2315, "lr": 1.3678813464179252e-05}
{"train_loss": 0.03691253438591957, "global_step": 206102, "epoch": 2315, "lr": 1.3678415044326786e-05}
{"train_loss": 0.06410392373800278, "global_step": 206103, "epoch": 2315, "lr": 1.3678016629357337e-05}
{"train_loss": 0.021631291136145592, "global_step": 206104, "epoch": 2315, "lr": 1.3677618219270916e-05}
{"train_loss": 0.06418055295944214, "global_step": 206105, "epoch": 2315, "lr": 1.367721981406761e-05}
{"train_loss": 0.057278431951999664, "global_step": 206106, "epoch": 2315, "lr": 1.3676821413747448e-05}
{"train_loss": 0.07105402648448944, "global_step": 206107, "epoch": 2315, "lr": 1.3676423018310507e-05}
{"train_loss": 0.046619314700365067, "global_step": 206108, "epoch": 2315, "lr": 1.3676024627756812e-05}
{"train_loss": 0.058440882712602615, "global_step": 206109, "epoch": 2315, "lr": 1.3675626242086447e-05}
{"train_loss": 0.03547821193933487, "global_step": 206110, "epoch": 2315, "lr": 1.3675227861299445e-05}
{"train_loss": 0.10988536477088928, "global_step": 206111, "epoch": 2315, "lr": 1.3674829485395874e-05}
{"train_loss": 0.02544279582798481, "global_step": 206112, "epoch": 2315, "lr": 1.3674431114375762e-05}
{"train_loss": 0.05371558666229248, "global_step": 206113, "epoch": 2315, "lr": 1.3674032748239201e-05}
{"train_loss": 0.07442441582679749, "global_step": 206114, "epoch": 2315, "lr": 1.3673634386986206e-05}
{"train_loss": 0.05650629848241806, "global_step": 206115, "epoch": 2315, "lr": 1.367323603061686e-05}
{"train_loss": 0.05294041708111763, "global_step": 206116, "epoch": 2315, "lr": 1.3672837679131207e-05}
{"train_loss": 0.056266527622938156, "global_step": 206117, "epoch": 2315, "lr": 1.3672439332529285e-05}
{"train_loss": 0.04676639288663864, "global_step": 206118, "epoch": 2315, "lr": 1.3672040990811174e-05}
{"train_loss": 0.13766300678253174, "global_step": 206119, "epoch": 2315, "lr": 1.3671642653976901e-05}
{"train_loss": 0.05427899584174156, "global_step": 206120, "epoch": 2315, "lr": 1.3671244322026544e-05}
{"train_loss": 0.03392405062913895, "global_step": 206121, "epoch": 2315, "lr": 1.3670845994960136e-05}
{"train_loss": 0.15103596448898315, "global_step": 206122, "epoch": 2315, "lr": 1.3670447672777754e-05}
{"train_loss": 0.062439461395646746, "global_step": 206123, "epoch": 2315, "lr": 1.367004935547942e-05, "val_loss": 8.389241218566895, "train_action_mse_error": 8.840550422668457}
{"train_loss": 0.061782583594322205, "global_step": 206124, "epoch": 2316, "lr": 1.3669651043065218e-05}
{"train_loss": 0.09461679309606552, "global_step": 206125, "epoch": 2316, "lr": 1.3669252735535176e-05}
{"train_loss": 0.0776064395904541, "global_step": 206126, "epoch": 2316, "lr": 1.3668854432889377e-05}
{"train_loss": 0.05778498202562332, "global_step": 206127, "epoch": 2316, "lr": 1.3668456135127838e-05}
{"train_loss": 0.033652711659669876, "global_step": 206128, "epoch": 2316, "lr": 1.3668057842250653e-05}
{"train_loss": 0.07535960525274277, "global_step": 206129, "epoch": 2316, "lr": 1.366765955425784e-05}
{"train_loss": 0.07032030820846558, "global_step": 206130, "epoch": 2316, "lr": 1.3667261271149479e-05}
{"train_loss": 0.08410200476646423, "global_step": 206131, "epoch": 2316, "lr": 1.3666862992925594e-05}
{"train_loss": 0.10573582351207733, "global_step": 206132, "epoch": 2316, "lr": 1.3666464719586275e-05}
{"train_loss": 0.04663442447781563, "global_step": 206133, "epoch": 2316, "lr": 1.3666066451131543e-05}
{"train_loss": 0.05087951570749283, "global_step": 206134, "epoch": 2316, "lr": 1.3665668187561464e-05}
{"train_loss": 0.07851950079202652, "global_step": 206135, "epoch": 2316, "lr": 1.3665269928876106e-05}
{"train_loss": 0.06388985365629196, "global_step": 206136, "epoch": 2316, "lr": 1.3664871675075497e-05}
{"train_loss": 0.04844938591122627, "global_step": 206137, "epoch": 2316, "lr": 1.3664473426159718e-05}
{"train_loss": 0.0397515669465065, "global_step": 206138, "epoch": 2316, "lr": 1.3664075182128794e-05}
{"train_loss": 0.03849845752120018, "global_step": 206139, "epoch": 2316, "lr": 1.3663676942982805e-05}
{"train_loss": 0.08678214997053146, "global_step": 206140, "epoch": 2316, "lr": 1.3663278708721777e-05}
{"train_loss": 0.13535453379154205, "global_step": 206141, "epoch": 2316, "lr": 1.3662880479345797e-05}
{"train_loss": 0.05826723948121071, "global_step": 206142, "epoch": 2316, "lr": 1.3662482254854881e-05}
{"train_loss": 0.05477535352110863, "global_step": 206143, "epoch": 2316, "lr": 1.3662084035249117e-05}
{"train_loss": 0.07935895025730133, "global_step": 206144, "epoch": 2316, "lr": 1.366168582052853e-05}
{"train_loss": 0.03856813162565231, "global_step": 206145, "epoch": 2316, "lr": 1.3661287610693196e-05}
{"train_loss": 0.07501684129238129, "global_step": 206146, "epoch": 2316, "lr": 1.3660889405743149e-05}
{"train_loss": 0.08033747225999832, "global_step": 206147, "epoch": 2316, "lr": 1.3660491205678466e-05}
{"train_loss": 0.03226000443100929, "global_step": 206148, "epoch": 2316, "lr": 1.3660093010499169e-05}
{"train_loss": 0.07676304131746292, "global_step": 206149, "epoch": 2316, "lr": 1.3659694820205348e-05}
{"train_loss": 0.09896917641162872, "global_step": 206150, "epoch": 2316, "lr": 1.3659296634797025e-05}
{"train_loss": 0.055758025497198105, "global_step": 206151, "epoch": 2316, "lr": 1.3658898454274283e-05}
{"train_loss": 0.04259588569402695, "global_step": 206152, "epoch": 2316, "lr": 1.3658500278637137e-05}
{"train_loss": 0.1023099422454834, "global_step": 206153, "epoch": 2316, "lr": 1.3658102107885685e-05}
{"train_loss": 0.06525713205337524, "global_step": 206154, "epoch": 2316, "lr": 1.365770394201995e-05}
{"train_loss": 0.091373972594738, "global_step": 206155, "epoch": 2316, "lr": 1.3657305781039986e-05}
{"train_loss": 0.06567130982875824, "global_step": 206156, "epoch": 2316, "lr": 1.3656907624945864e-05}
{"train_loss": 0.04978158324956894, "global_step": 206157, "epoch": 2316, "lr": 1.3656509473737616e-05}
{"train_loss": 0.08982910960912704, "global_step": 206158, "epoch": 2316, "lr": 1.3656111327415322e-05}
{"train_loss": 0.05977052450180054, "global_step": 206159, "epoch": 2316, "lr": 1.3655713185979009e-05}
{"train_loss": 0.010902293026447296, "global_step": 206160, "epoch": 2316, "lr": 1.3655315049428752e-05}
{"train_loss": 0.03537917882204056, "global_step": 206161, "epoch": 2316, "lr": 1.3654916917764582e-05}
{"train_loss": 0.05682055652141571, "global_step": 206162, "epoch": 2316, "lr": 1.3654518790986582e-05}
{"train_loss": 0.06478733569383621, "global_step": 206163, "epoch": 2316, "lr": 1.3654120669094766e-05}
{"train_loss": 0.07517607510089874, "global_step": 206164, "epoch": 2316, "lr": 1.3653722552089237e-05}
{"train_loss": 0.09766244888305664, "global_step": 206165, "epoch": 2316, "lr": 1.365332443997e-05}
{"train_loss": 0.07926128804683685, "global_step": 206166, "epoch": 2316, "lr": 1.3652926332737148e-05}
{"train_loss": 0.10230839997529984, "global_step": 206167, "epoch": 2316, "lr": 1.3652528230390698e-05}
{"train_loss": 0.040877748280763626, "global_step": 206168, "epoch": 2316, "lr": 1.365213013293074e-05}
{"train_loss": 0.07804691791534424, "global_step": 206169, "epoch": 2316, "lr": 1.36517320403573e-05}
{"train_loss": 0.05834977701306343, "global_step": 206170, "epoch": 2316, "lr": 1.3651333952670437e-05}
{"train_loss": 0.0945718064904213, "global_step": 206171, "epoch": 2316, "lr": 1.3650935869870223e-05}
{"train_loss": 0.05717577785253525, "global_step": 206172, "epoch": 2316, "lr": 1.3650537791956685e-05}
{"train_loss": 0.021454181522130966, "global_step": 206173, "epoch": 2316, "lr": 1.3650139718929905e-05}
{"train_loss": 0.03479892015457153, "global_step": 206174, "epoch": 2316, "lr": 1.3649741650789904e-05}
{"train_loss": 0.07109417021274567, "global_step": 206175, "epoch": 2316, "lr": 1.3649343587536766e-05}
{"train_loss": 0.04558410495519638, "global_step": 206176, "epoch": 2316, "lr": 1.3648945529170516e-05}
{"train_loss": 0.07226858288049698, "global_step": 206177, "epoch": 2316, "lr": 1.3648547475691242e-05}
{"train_loss": 0.03906899318099022, "global_step": 206178, "epoch": 2316, "lr": 1.3648149427098955e-05}
{"train_loss": 0.10748572647571564, "global_step": 206179, "epoch": 2316, "lr": 1.3647751383393754e-05}
{"train_loss": 0.04800349473953247, "global_step": 206180, "epoch": 2316, "lr": 1.3647353344575648e-05}
{"train_loss": 0.11512243747711182, "global_step": 206181, "epoch": 2316, "lr": 1.3646955310644733e-05}
{"train_loss": 0.05675647407770157, "global_step": 206182, "epoch": 2316, "lr": 1.3646557281601024e-05}
{"train_loss": 0.10477268695831299, "global_step": 206183, "epoch": 2316, "lr": 1.3646159257444607e-05}
{"train_loss": 0.05369144678115845, "global_step": 206184, "epoch": 2316, "lr": 1.3645761238175503e-05}
{"train_loss": 0.06664685904979706, "global_step": 206185, "epoch": 2316, "lr": 1.3645363223793806e-05}
{"train_loss": 0.03873778134584427, "global_step": 206186, "epoch": 2316, "lr": 1.3644965214299527e-05}
{"train_loss": 0.07951264083385468, "global_step": 206187, "epoch": 2316, "lr": 1.364456720969275e-05}
{"train_loss": 0.0556427426636219, "global_step": 206188, "epoch": 2316, "lr": 1.3644169209973507e-05}
{"train_loss": 0.03265540674328804, "global_step": 206189, "epoch": 2316, "lr": 1.3643771215141881e-05}
{"train_loss": 0.0807117149233818, "global_step": 206190, "epoch": 2316, "lr": 1.3643373225197887e-05}
{"train_loss": 0.024413766339421272, "global_step": 206191, "epoch": 2316, "lr": 1.3642975240141614e-05}
{"train_loss": 0.09017910808324814, "global_step": 206192, "epoch": 2316, "lr": 1.3642577259973088e-05}
{"train_loss": 0.05838233232498169, "global_step": 206193, "epoch": 2316, "lr": 1.3642179284692385e-05}
{"train_loss": 0.0499456524848938, "global_step": 206194, "epoch": 2316, "lr": 1.364178131429955e-05}
{"train_loss": 0.07447103410959244, "global_step": 206195, "epoch": 2316, "lr": 1.3641383348794618e-05}
{"train_loss": 0.05072760581970215, "global_step": 206196, "epoch": 2316, "lr": 1.3640985388177679e-05}
{"train_loss": 0.05820606276392937, "global_step": 206197, "epoch": 2316, "lr": 1.3640587432448748e-05}
{"train_loss": 0.09130872786045074, "global_step": 206198, "epoch": 2316, "lr": 1.3640189481607906e-05}
{"train_loss": 0.028386516496539116, "global_step": 206199, "epoch": 2316, "lr": 1.3639791535655189e-05}
{"train_loss": 0.05662081390619278, "global_step": 206200, "epoch": 2316, "lr": 1.3639393594590677e-05}
{"train_loss": 0.030977588146924973, "global_step": 206201, "epoch": 2316, "lr": 1.3638995658414382e-05}
{"train_loss": 0.06983255594968796, "global_step": 206202, "epoch": 2316, "lr": 1.3638597727126401e-05}
{"train_loss": 0.08503194898366928, "global_step": 206203, "epoch": 2316, "lr": 1.3638199800726754e-05}
{"train_loss": 0.10501983016729355, "global_step": 206204, "epoch": 2316, "lr": 1.3637801879215516e-05}
{"train_loss": 0.04506832733750343, "global_step": 206205, "epoch": 2316, "lr": 1.3637403962592721e-05}
{"train_loss": 0.07822039723396301, "global_step": 206206, "epoch": 2316, "lr": 1.3637006050858453e-05}
{"train_loss": 0.05306147783994675, "global_step": 206207, "epoch": 2316, "lr": 1.3636608144012724e-05}
{"train_loss": 0.12520833313465118, "global_step": 206208, "epoch": 2316, "lr": 1.3636210242055613e-05}
{"train_loss": 0.0612289197742939, "global_step": 206209, "epoch": 2316, "lr": 1.3635812344987187e-05}
{"train_loss": 0.05156972259283066, "global_step": 206210, "epoch": 2316, "lr": 1.3635414452807466e-05}
{"train_loss": 0.03980676829814911, "global_step": 206211, "epoch": 2316, "lr": 1.3635016565516539e-05}
{"train_loss": 0.06633118933506227, "global_step": 206212, "epoch": 2316, "lr": 1.3634618683114419e-05, "val_loss": 8.527212142944336}
{"train_loss": 0.07671253383159637, "global_step": 206213, "epoch": 2317, "lr": 1.36342208056012e-05}
{"train_loss": 0.04196775332093239, "global_step": 206214, "epoch": 2317, "lr": 1.3633822932976903e-05}
{"train_loss": 0.08976640552282333, "global_step": 206215, "epoch": 2317, "lr": 1.3633425065241606e-05}
{"train_loss": 0.052643127739429474, "global_step": 206216, "epoch": 2317, "lr": 1.3633027202395343e-05}
{"train_loss": 0.03660336881875992, "global_step": 206217, "epoch": 2317, "lr": 1.3632629344438186e-05}
{"train_loss": 0.08839254081249237, "global_step": 206218, "epoch": 2317, "lr": 1.3632231491370168e-05}
{"train_loss": 0.06156797707080841, "global_step": 206219, "epoch": 2317, "lr": 1.3631833643191366e-05}
{"train_loss": 0.02585839293897152, "global_step": 206220, "epoch": 2317, "lr": 1.3631435799901803e-05}
{"train_loss": 0.12834514677524567, "global_step": 206221, "epoch": 2317, "lr": 1.363103796150157e-05}
{"train_loss": 0.05651324987411499, "global_step": 206222, "epoch": 2317, "lr": 1.3630640127990685e-05}
{"train_loss": 0.057129260152578354, "global_step": 206223, "epoch": 2317, "lr": 1.3630242299369228e-05}
{"train_loss": 0.05833260715007782, "global_step": 206224, "epoch": 2317, "lr": 1.3629844475637226e-05}
{"train_loss": 0.05033095180988312, "global_step": 206225, "epoch": 2317, "lr": 1.3629446656794753e-05}
{"train_loss": 0.026316208764910698, "global_step": 206226, "epoch": 2317, "lr": 1.3629048842841873e-05}
{"train_loss": 0.06089043617248535, "global_step": 206227, "epoch": 2317, "lr": 1.3628651033778622e-05}
{"train_loss": 0.055887844413518906, "global_step": 206228, "epoch": 2317, "lr": 1.3628253229605043e-05}
{"train_loss": 0.039224956184625626, "global_step": 206229, "epoch": 2317, "lr": 1.3627855430321212e-05}
{"train_loss": 0.05044621229171753, "global_step": 206230, "epoch": 2317, "lr": 1.362745763592716e-05}
{"train_loss": 0.021534021943807602, "global_step": 206231, "epoch": 2317, "lr": 1.3627059846422968e-05}
{"train_loss": 0.022409122437238693, "global_step": 206232, "epoch": 2317, "lr": 1.3626662061808676e-05}
{"train_loss": 0.06852809339761734, "global_step": 206233, "epoch": 2317, "lr": 1.3626264282084316e-05}
{"train_loss": 0.036722332239151, "global_step": 206234, "epoch": 2317, "lr": 1.3625866507249985e-05}
{"train_loss": 0.025828933343291283, "global_step": 206235, "epoch": 2317, "lr": 1.3625468737305691e-05}
{"train_loss": 0.052390120923519135, "global_step": 206236, "epoch": 2317, "lr": 1.3625070972251524e-05}
{"train_loss": 0.07026426494121552, "global_step": 206237, "epoch": 2317, "lr": 1.3624673212087514e-05}
{"train_loss": 0.05123085528612137, "global_step": 206238, "epoch": 2317, "lr": 1.3624275456813734e-05}
{"train_loss": 0.09467857331037521, "global_step": 206239, "epoch": 2317, "lr": 1.362387770643021e-05}
{"train_loss": 0.0256721880286932, "global_step": 206240, "epoch": 2317, "lr": 1.362347996093703e-05}
{"train_loss": 0.05461600422859192, "global_step": 206241, "epoch": 2317, "lr": 1.3623082220334216e-05}
{"train_loss": 0.05259590968489647, "global_step": 206242, "epoch": 2317, "lr": 1.3622684484621844e-05}
{"train_loss": 0.03265496715903282, "global_step": 206243, "epoch": 2317, "lr": 1.3622286753799951e-05}
{"train_loss": 0.05657729133963585, "global_step": 206244, "epoch": 2317, "lr": 1.3621889027868596e-05}
{"train_loss": 0.06465853750705719, "global_step": 206245, "epoch": 2317, "lr": 1.362149130682785e-05}
{"train_loss": 0.06360868364572525, "global_step": 206246, "epoch": 2317, "lr": 1.3621093590677736e-05}
{"train_loss": 0.05063991621136665, "global_step": 206247, "epoch": 2317, "lr": 1.3620695879418333e-05}
{"train_loss": 0.08905445784330368, "global_step": 206248, "epoch": 2317, "lr": 1.3620298173049673e-05}
{"train_loss": 0.05786169320344925, "global_step": 206249, "epoch": 2317, "lr": 1.3619900471571839e-05}
{"train_loss": 0.08545573800802231, "global_step": 206250, "epoch": 2317, "lr": 1.3619502774984844e-05}
{"train_loss": 0.06221497803926468, "global_step": 206251, "epoch": 2317, "lr": 1.361910508328878e-05}
{"train_loss": 0.02969108521938324, "global_step": 206252, "epoch": 2317, "lr": 1.3618707396483671e-05}
{"train_loss": 0.04930266737937927, "global_step": 206253, "epoch": 2317, "lr": 1.36183097145696e-05}
{"train_loss": 0.07237965613603592, "global_step": 206254, "epoch": 2317, "lr": 1.3617912037546587e-05}
{"train_loss": 0.09644346684217453, "global_step": 206255, "epoch": 2317, "lr": 1.361751436541472e-05}
{"train_loss": 0.04098609462380409, "global_step": 206256, "epoch": 2317, "lr": 1.3617116698174014e-05}
{"train_loss": 0.05112839862704277, "global_step": 206257, "epoch": 2317, "lr": 1.361671903582456e-05}
{"train_loss": 0.046062104403972626, "global_step": 206258, "epoch": 2317, "lr": 1.3616321378366381e-05}
{"train_loss": 0.07403609901666641, "global_step": 206259, "epoch": 2317, "lr": 1.3615923725799562e-05}
{"train_loss": 0.05439639836549759, "global_step": 206260, "epoch": 2317, "lr": 1.3615526078124119e-05}
{"train_loss": 0.0641484260559082, "global_step": 206261, "epoch": 2317, "lr": 1.3615128435340125e-05}
{"train_loss": 0.07386703789234161, "global_step": 206262, "epoch": 2317, "lr": 1.361473079744765e-05}
{"train_loss": 0.04383813962340355, "global_step": 206263, "epoch": 2317, "lr": 1.3614333164446718e-05}
{"train_loss": 0.06469082832336426, "global_step": 206264, "epoch": 2317, "lr": 1.3613935536337407e-05}
{"train_loss": 0.08412671089172363, "global_step": 206265, "epoch": 2317, "lr": 1.3613537913119761e-05}
{"train_loss": 0.07027710229158401, "global_step": 206266, "epoch": 2317, "lr": 1.3613140294793813e-05}
{"train_loss": 0.10463041812181473, "global_step": 206267, "epoch": 2317, "lr": 1.3612742681359653e-05}
{"train_loss": 0.08748070150613785, "global_step": 206268, "epoch": 2317, "lr": 1.3612345072817301e-05}
{"train_loss": 0.021141178905963898, "global_step": 206269, "epoch": 2317, "lr": 1.3611947469166841e-05}
{"train_loss": 0.058654844760894775, "global_step": 206270, "epoch": 2317, "lr": 1.3611549870408313e-05}
{"train_loss": 0.07402414083480835, "global_step": 206271, "epoch": 2317, "lr": 1.3611152276541749e-05}
{"train_loss": 0.10889818519353867, "global_step": 206272, "epoch": 2317, "lr": 1.361075468756724e-05}
{"train_loss": 0.05018620193004608, "global_step": 206273, "epoch": 2317, "lr": 1.3610357103484805e-05}
{"train_loss": 0.06821025162935257, "global_step": 206274, "epoch": 2317, "lr": 1.360995952429453e-05}
{"train_loss": 0.05598360300064087, "global_step": 206275, "epoch": 2317, "lr": 1.3609561949996435e-05}
{"train_loss": 0.04604146257042885, "global_step": 206276, "epoch": 2317, "lr": 1.360916438059061e-05}
{"train_loss": 0.032667022198438644, "global_step": 206277, "epoch": 2317, "lr": 1.3608766816077072e-05}
{"train_loss": 0.040777649730443954, "global_step": 206278, "epoch": 2317, "lr": 1.3608369256455905e-05}
{"train_loss": 0.0515630766749382, "global_step": 206279, "epoch": 2317, "lr": 1.360797170172714e-05}
{"train_loss": 0.03935505822300911, "global_step": 206280, "epoch": 2317, "lr": 1.3607574151890835e-05}
{"train_loss": 0.10375931113958359, "global_step": 206281, "epoch": 2317, "lr": 1.3607176606947064e-05}
{"train_loss": 0.06479579210281372, "global_step": 206282, "epoch": 2317, "lr": 1.3606779066895847e-05}
{"train_loss": 0.06513255834579468, "global_step": 206283, "epoch": 2317, "lr": 1.3606381531737273e-05}
{"train_loss": 0.06618630141019821, "global_step": 206284, "epoch": 2317, "lr": 1.3605984001471362e-05}
{"train_loss": 0.07451584190130234, "global_step": 206285, "epoch": 2317, "lr": 1.3605586476098198e-05}
{"train_loss": 0.06470898538827896, "global_step": 206286, "epoch": 2317, "lr": 1.3605188955617804e-05}
{"train_loss": 0.028291407972574234, "global_step": 206287, "epoch": 2317, "lr": 1.3604791440030263e-05}
{"train_loss": 0.03775118291378021, "global_step": 206288, "epoch": 2317, "lr": 1.3604393929335602e-05}
{"train_loss": 0.05320487171411514, "global_step": 206289, "epoch": 2317, "lr": 1.3603996423533898e-05}
{"train_loss": 0.06318943202495575, "global_step": 206290, "epoch": 2317, "lr": 1.3603598922625182e-05}
{"train_loss": 0.04700037091970444, "global_step": 206291, "epoch": 2317, "lr": 1.360320142660953e-05}
{"train_loss": 0.06438329070806503, "global_step": 206292, "epoch": 2317, "lr": 1.3602803935486968e-05}
{"train_loss": 0.044574856758117676, "global_step": 206293, "epoch": 2317, "lr": 1.3602406449257581e-05}
{"train_loss": 0.0889645367860794, "global_step": 206294, "epoch": 2317, "lr": 1.3602008967921398e-05}
{"train_loss": 0.07605230063199997, "global_step": 206295, "epoch": 2317, "lr": 1.3601611491478488e-05}
{"train_loss": 0.103840172290802, "global_step": 206296, "epoch": 2317, "lr": 1.3601214019928887e-05}
{"train_loss": 0.07621598243713379, "global_step": 206297, "epoch": 2317, "lr": 1.360081655327266e-05}
{"train_loss": 0.07217997312545776, "global_step": 206298, "epoch": 2317, "lr": 1.3600419091509875e-05}
{"train_loss": 0.02800723724067211, "global_step": 206299, "epoch": 2317, "lr": 1.3600021634640553e-05}
{"train_loss": 0.052248258143663406, "global_step": 206300, "epoch": 2317, "lr": 1.3599624182664778e-05}
{"train_loss": 0.05936827198759224, "global_step": 206301, "epoch": 2317, "lr": 1.3599226735582576e-05, "val_loss": 8.319464683532715}
{"train_loss": 0.09403108805418015, "global_step": 206302, "epoch": 2318, "lr": 1.3598829293394027e-05}
{"train_loss": 0.029994072392582893, "global_step": 206303, "epoch": 2318, "lr": 1.3598431856099164e-05}
{"train_loss": 0.05775246024131775, "global_step": 206304, "epoch": 2318, "lr": 1.3598034423698058e-05}
{"train_loss": 0.09000103175640106, "global_step": 206305, "epoch": 2318, "lr": 1.3597636996190755e-05}
{"train_loss": 0.055364955216646194, "global_step": 206306, "epoch": 2318, "lr": 1.3597239573577285e-05}
{"train_loss": 0.07973229885101318, "global_step": 206307, "epoch": 2318, "lr": 1.3596842155857742e-05}
{"train_loss": 0.11297838389873505, "global_step": 206308, "epoch": 2318, "lr": 1.3596444743032144e-05}
{"train_loss": 0.009247583337128162, "global_step": 206309, "epoch": 2318, "lr": 1.3596047335100576e-05}
{"train_loss": 0.05734200403094292, "global_step": 206310, "epoch": 2318, "lr": 1.3595649932063076e-05}
{"train_loss": 0.049110203981399536, "global_step": 206311, "epoch": 2318, "lr": 1.3595252533919684e-05}
{"train_loss": 0.03278709575533867, "global_step": 206312, "epoch": 2318, "lr": 1.3594855140670476e-05}
{"train_loss": 0.04463760554790497, "global_step": 206313, "epoch": 2318, "lr": 1.3594457752315488e-05}
{"train_loss": 0.033483654260635376, "global_step": 206314, "epoch": 2318, "lr": 1.3594060368854788e-05}
{"train_loss": 0.09625107795000076, "global_step": 206315, "epoch": 2318, "lr": 1.3593662990288414e-05}
{"train_loss": 0.08549532294273376, "global_step": 206316, "epoch": 2318, "lr": 1.359326561661643e-05}
{"train_loss": 0.06732509285211563, "global_step": 206317, "epoch": 2318, "lr": 1.3592868247838897e-05}
{"train_loss": 0.0556180402636528, "global_step": 206318, "epoch": 2318, "lr": 1.3592470883955849e-05}
{"train_loss": 0.06321212649345398, "global_step": 206319, "epoch": 2318, "lr": 1.3592073524967364e-05}
{"train_loss": 0.08930403739213943, "global_step": 206320, "epoch": 2318, "lr": 1.3591676170873462e-05}
{"train_loss": 0.06197427213191986, "global_step": 206321, "epoch": 2318, "lr": 1.359127882167423e-05}
{"train_loss": 0.04186360910534859, "global_step": 206322, "epoch": 2318, "lr": 1.3590881477369694e-05}
{"train_loss": 0.06654073297977448, "global_step": 206323, "epoch": 2318, "lr": 1.3590484137959935e-05}
{"train_loss": 0.08273206651210785, "global_step": 206324, "epoch": 2318, "lr": 1.359008680344498e-05}
{"train_loss": 0.06803681701421738, "global_step": 206325, "epoch": 2318, "lr": 1.3589689473824908e-05}
{"train_loss": 0.11126109212636948, "global_step": 206326, "epoch": 2318, "lr": 1.3589292149099742e-05}
{"train_loss": 0.08454267680644989, "global_step": 206327, "epoch": 2318, "lr": 1.3588894829269567e-05}
{"train_loss": 0.08155917376279831, "global_step": 206328, "epoch": 2318, "lr": 1.3588497514334403e-05}
{"train_loss": 0.01883523166179657, "global_step": 206329, "epoch": 2318, "lr": 1.3588100204294341e-05}
{"train_loss": 0.024344226345419884, "global_step": 206330, "epoch": 2318, "lr": 1.3587702899149396e-05}
{"train_loss": 0.07221704721450806, "global_step": 206331, "epoch": 2318, "lr": 1.3587305598899663e-05}
{"train_loss": 0.0374012365937233, "global_step": 206332, "epoch": 2318, "lr": 1.3586908303545148e-05}
{"train_loss": 0.057276416569948196, "global_step": 206333, "epoch": 2318, "lr": 1.358651101308595e-05}
{"train_loss": 0.05648515373468399, "global_step": 206334, "epoch": 2318, "lr": 1.3586113727522088e-05}
{"train_loss": 0.04860847443342209, "global_step": 206335, "epoch": 2318, "lr": 1.358571644685363e-05}
{"train_loss": 0.07674700766801834, "global_step": 206336, "epoch": 2318, "lr": 1.358531917108064e-05}
{"train_loss": 0.06157299503684044, "global_step": 206337, "epoch": 2318, "lr": 1.3584921900203145e-05}
{"train_loss": 0.039997633546590805, "global_step": 206338, "epoch": 2318, "lr": 1.358452463422123e-05}
{"train_loss": 0.1124272421002388, "global_step": 206339, "epoch": 2318, "lr": 1.358412737313492e-05}
{"train_loss": 0.05736887454986572, "global_step": 206340, "epoch": 2318, "lr": 1.3583730116944288e-05}
{"train_loss": 0.034558821469545364, "global_step": 206341, "epoch": 2318, "lr": 1.3583332865649368e-05}
{"train_loss": 0.10036144405603409, "global_step": 206342, "epoch": 2318, "lr": 1.3582935619250242e-05}
{"train_loss": 0.03646161034703255, "global_step": 206343, "epoch": 2318, "lr": 1.3582538377746945e-05}
{"train_loss": 0.04013983905315399, "global_step": 206344, "epoch": 2318, "lr": 1.3582141141139515e-05}
{"train_loss": 0.07825027406215668, "global_step": 206345, "epoch": 2318, "lr": 1.358174390942804e-05}
{"train_loss": 0.08736956864595413, "global_step": 206346, "epoch": 2318, "lr": 1.3581346682612538e-05}
{"train_loss": 0.04781928285956383, "global_step": 206347, "epoch": 2318, "lr": 1.3580949460693098e-05}
{"train_loss": 0.029691096395254135, "global_step": 206348, "epoch": 2318, "lr": 1.3580552243669753e-05}
{"train_loss": 0.05658058449625969, "global_step": 206349, "epoch": 2318, "lr": 1.3580155031542546e-05}
{"train_loss": 0.04073147475719452, "global_step": 206350, "epoch": 2318, "lr": 1.3579757824311557e-05}
{"train_loss": 0.08395262062549591, "global_step": 206351, "epoch": 2318, "lr": 1.3579360621976811e-05}
{"train_loss": 0.028961408883333206, "global_step": 206352, "epoch": 2318, "lr": 1.3578963424538377e-05}
{"train_loss": 0.03057251125574112, "global_step": 206353, "epoch": 2318, "lr": 1.357856623199632e-05}
{"train_loss": 0.0995296761393547, "global_step": 206354, "epoch": 2318, "lr": 1.3578169044350669e-05}
{"train_loss": 0.029964042827486992, "global_step": 206355, "epoch": 2318, "lr": 1.3577771861601502e-05}
{"train_loss": 0.03612211346626282, "global_step": 206356, "epoch": 2318, "lr": 1.3577374683748844e-05}
{"train_loss": 0.07625985890626907, "global_step": 206357, "epoch": 2318, "lr": 1.3576977510792783e-05}
{"train_loss": 0.06746171414852142, "global_step": 206358, "epoch": 2318, "lr": 1.3576580342733336e-05}
{"train_loss": 0.12263328582048416, "global_step": 206359, "epoch": 2318, "lr": 1.3576183179570584e-05}
{"train_loss": 0.047237034887075424, "global_step": 206360, "epoch": 2318, "lr": 1.3575786021304559e-05}
{"train_loss": 0.0579247772693634, "global_step": 206361, "epoch": 2318, "lr": 1.3575388867935346e-05}
{"train_loss": 0.06602340191602707, "global_step": 206362, "epoch": 2318, "lr": 1.357499171946296e-05}
{"train_loss": 0.061478354036808014, "global_step": 206363, "epoch": 2318, "lr": 1.3574594575887483e-05}
{"train_loss": 0.06384525448083878, "global_step": 206364, "epoch": 2318, "lr": 1.3574197437208947e-05}
{"train_loss": 0.044175952672958374, "global_step": 206365, "epoch": 2318, "lr": 1.3573800303427426e-05}
{"train_loss": 0.12219356000423431, "global_step": 206366, "epoch": 2318, "lr": 1.3573403174542954e-05}
{"train_loss": 0.02794024907052517, "global_step": 206367, "epoch": 2318, "lr": 1.3573006050555609e-05}
{"train_loss": 0.06400828063488007, "global_step": 206368, "epoch": 2318, "lr": 1.3572608931465413e-05}
{"train_loss": 0.07951496541500092, "global_step": 206369, "epoch": 2318, "lr": 1.3572211817272452e-05}
{"train_loss": 0.08615484833717346, "global_step": 206370, "epoch": 2318, "lr": 1.3571814707976744e-05}
{"train_loss": 0.03226771950721741, "global_step": 206371, "epoch": 2318, "lr": 1.3571417603578363e-05}
{"train_loss": 0.05557188391685486, "global_step": 206372, "epoch": 2318, "lr": 1.3571020504077375e-05}
{"train_loss": 0.02851090207695961, "global_step": 206373, "epoch": 2318, "lr": 1.3570623409473809e-05}
{"train_loss": 0.05559883266687393, "global_step": 206374, "epoch": 2318, "lr": 1.3570226319767737e-05}
{"train_loss": 0.06947163492441177, "global_step": 206375, "epoch": 2318, "lr": 1.3569829234959197e-05}
{"train_loss": 0.04117963835597038, "global_step": 206376, "epoch": 2318, "lr": 1.3569432155048261e-05}
{"train_loss": 0.06310373544692993, "global_step": 206377, "epoch": 2318, "lr": 1.3569035080034953e-05}
{"train_loss": 0.03207330405712128, "global_step": 206378, "epoch": 2318, "lr": 1.356863800991936e-05}
{"train_loss": 0.05085315555334091, "global_step": 206379, "epoch": 2318, "lr": 1.356824094470151e-05}
{"train_loss": 0.08403851091861725, "global_step": 206380, "epoch": 2318, "lr": 1.3567843884381476e-05}
{"train_loss": 0.049655668437480927, "global_step": 206381, "epoch": 2318, "lr": 1.3567446828959301e-05}
{"train_loss": 0.07335373014211655, "global_step": 206382, "epoch": 2318, "lr": 1.3567049778435027e-05}
{"train_loss": 0.07083351910114288, "global_step": 206383, "epoch": 2318, "lr": 1.3566652732808732e-05}
{"train_loss": 0.08347291499376297, "global_step": 206384, "epoch": 2318, "lr": 1.3566255692080437e-05}
{"train_loss": 0.06415791809558868, "global_step": 206385, "epoch": 2318, "lr": 1.3565858656250235e-05}
{"train_loss": 0.06200327351689339, "global_step": 206386, "epoch": 2318, "lr": 1.3565461625318159e-05}
{"train_loss": 0.07711801677942276, "global_step": 206387, "epoch": 2318, "lr": 1.356506459928425e-05}
{"train_loss": 0.0515853613615036, "global_step": 206388, "epoch": 2318, "lr": 1.3564667578148571e-05}
{"train_loss": 0.098564513027668, "global_step": 206389, "epoch": 2318, "lr": 1.356427056191119e-05}
{"train_loss": 0.061893203463196086, "global_step": 206390, "epoch": 2318, "lr": 1.3563873550572143e-05, "val_loss": 8.504831314086914}
{"train_loss": 0.06804560124874115, "global_step": 206391, "epoch": 2319, "lr": 1.3563476544131498e-05}
{"train_loss": 0.03275926038622856, "global_step": 206392, "epoch": 2319, "lr": 1.356307954258928e-05}
{"train_loss": 0.05897112563252449, "global_step": 206393, "epoch": 2319, "lr": 1.3562682545945581e-05}
{"train_loss": 0.03360268473625183, "global_step": 206394, "epoch": 2319, "lr": 1.3562285554200426e-05}
{"train_loss": 0.07028520107269287, "global_step": 206395, "epoch": 2319, "lr": 1.3561888567353886e-05}
{"train_loss": 0.111347496509552, "global_step": 206396, "epoch": 2319, "lr": 1.3561491585405994e-05}
{"train_loss": 0.05592816695570946, "global_step": 206397, "epoch": 2319, "lr": 1.3561094608356828e-05}
{"train_loss": 0.04173874855041504, "global_step": 206398, "epoch": 2319, "lr": 1.3560697636206415e-05}
{"train_loss": 0.0472484827041626, "global_step": 206399, "epoch": 2319, "lr": 1.3560300668954834e-05}
{"train_loss": 0.054398585110902786, "global_step": 206400, "epoch": 2319, "lr": 1.3559903706602117e-05}
{"train_loss": 0.05882386863231659, "global_step": 206401, "epoch": 2319, "lr": 1.3559506749148337e-05}
{"train_loss": 0.06237703934311867, "global_step": 206402, "epoch": 2319, "lr": 1.3559109796593523e-05}
{"train_loss": 0.08409509807825089, "global_step": 206403, "epoch": 2319, "lr": 1.355871284893776e-05}
{"train_loss": 0.052476461976766586, "global_step": 206404, "epoch": 2319, "lr": 1.3558315906181068e-05}
{"train_loss": 0.1123906597495079, "global_step": 206405, "epoch": 2319, "lr": 1.3557918968323529e-05}
{"train_loss": 0.0736769363284111, "global_step": 206406, "epoch": 2319, "lr": 1.3557522035365173e-05}
{"train_loss": 0.05239582061767578, "global_step": 206407, "epoch": 2319, "lr": 1.3557125107306062e-05}
{"train_loss": 0.08286521583795547, "global_step": 206408, "epoch": 2319, "lr": 1.3556728184146267e-05}
{"train_loss": 0.08847576379776001, "global_step": 206409, "epoch": 2319, "lr": 1.3556331265885808e-05}
{"train_loss": 0.05599014461040497, "global_step": 206410, "epoch": 2319, "lr": 1.3555934352524768e-05}
{"train_loss": 0.11150519549846649, "global_step": 206411, "epoch": 2319, "lr": 1.3555537444063183e-05}
{"train_loss": 0.041977617889642715, "global_step": 206412, "epoch": 2319, "lr": 1.3555140540501122e-05}
{"train_loss": 0.09463536739349365, "global_step": 206413, "epoch": 2319, "lr": 1.3554743641838613e-05}
{"train_loss": 0.08454003185033798, "global_step": 206414, "epoch": 2319, "lr": 1.3554346748075735e-05}
{"train_loss": 0.06361963599920273, "global_step": 206415, "epoch": 2319, "lr": 1.3553949859212522e-05}
{"train_loss": 0.042078178375959396, "global_step": 206416, "epoch": 2319, "lr": 1.355355297524905e-05}
{"train_loss": 0.06344988197088242, "global_step": 206417, "epoch": 2319, "lr": 1.3553156096185344e-05}
{"train_loss": 0.0360255129635334, "global_step": 206418, "epoch": 2319, "lr": 1.3552759222021482e-05}
{"train_loss": 0.02860621176660061, "global_step": 206419, "epoch": 2319, "lr": 1.3552362352757498e-05}
{"train_loss": 0.04985758662223816, "global_step": 206420, "epoch": 2319, "lr": 1.355196548839347e-05}
{"train_loss": 0.038408152759075165, "global_step": 206421, "epoch": 2319, "lr": 1.3551568628929434e-05}
{"train_loss": 0.05745959281921387, "global_step": 206422, "epoch": 2319, "lr": 1.3551171774365429e-05}
{"train_loss": 0.053120654076337814, "global_step": 206423, "epoch": 2319, "lr": 1.3550774924701543e-05}
{"train_loss": 0.1112968847155571, "global_step": 206424, "epoch": 2319, "lr": 1.3550378079937792e-05}
{"train_loss": 0.02863946370780468, "global_step": 206425, "epoch": 2319, "lr": 1.3549981240074266e-05}
{"train_loss": 0.06561209261417389, "global_step": 206426, "epoch": 2319, "lr": 1.3549584405110987e-05}
{"train_loss": 0.048290058970451355, "global_step": 206427, "epoch": 2319, "lr": 1.3549187575048034e-05}
{"train_loss": 0.05058537423610687, "global_step": 206428, "epoch": 2319, "lr": 1.3548790749885437e-05}
{"train_loss": 0.0336301289498806, "global_step": 206429, "epoch": 2319, "lr": 1.354839392962327e-05}
{"train_loss": 0.0708787739276886, "global_step": 206430, "epoch": 2319, "lr": 1.3547997114261568e-05}
{"train_loss": 0.06591062992811203, "global_step": 206431, "epoch": 2319, "lr": 1.3547600303800407e-05}
{"train_loss": 0.04009800776839256, "global_step": 206432, "epoch": 2319, "lr": 1.3547203498239807e-05}
{"train_loss": 0.06767634302377701, "global_step": 206433, "epoch": 2319, "lr": 1.354680669757986e-05}
{"train_loss": 0.12007171660661697, "global_step": 206434, "epoch": 2319, "lr": 1.354640990182059e-05}
{"train_loss": 0.042202576994895935, "global_step": 206435, "epoch": 2319, "lr": 1.354601311096207e-05}
{"train_loss": 0.0540865920484066, "global_step": 206436, "epoch": 2319, "lr": 1.3545616325004334e-05}
{"train_loss": 0.058464787900447845, "global_step": 206437, "epoch": 2319, "lr": 1.3545219543947458e-05}
{"train_loss": 0.0455562025308609, "global_step": 206438, "epoch": 2319, "lr": 1.3544822767791465e-05}
{"train_loss": 0.031417544931173325, "global_step": 206439, "epoch": 2319, "lr": 1.3544425996536441e-05}
{"train_loss": 0.033159371465444565, "global_step": 206440, "epoch": 2319, "lr": 1.3544029230182415e-05}
{"train_loss": 0.09204217046499252, "global_step": 206441, "epoch": 2319, "lr": 1.3543632468729462e-05}
{"train_loss": 0.03714986518025398, "global_step": 206442, "epoch": 2319, "lr": 1.354323571217761e-05}
{"train_loss": 0.049314409494400024, "global_step": 206443, "epoch": 2319, "lr": 1.3542838960526921e-05}
{"train_loss": 0.07655297964811325, "global_step": 206444, "epoch": 2319, "lr": 1.3542442213777473e-05}
{"train_loss": 0.03286121413111687, "global_step": 206445, "epoch": 2319, "lr": 1.3542045471929282e-05}
{"train_loss": 0.051128238439559937, "global_step": 206446, "epoch": 2319, "lr": 1.3541648734982437e-05}
{"train_loss": 0.060508523136377335, "global_step": 206447, "epoch": 2319, "lr": 1.3541252002936955e-05}
{"train_loss": 0.06855899095535278, "global_step": 206448, "epoch": 2319, "lr": 1.354085527579292e-05}
{"train_loss": 0.04497702047228813, "global_step": 206449, "epoch": 2319, "lr": 1.3540458553550361e-05}
{"train_loss": 0.09242191910743713, "global_step": 206450, "epoch": 2319, "lr": 1.3540061836209356e-05}
{"train_loss": 0.07389869540929794, "global_step": 206451, "epoch": 2319, "lr": 1.353966512376993e-05}
{"train_loss": 0.08809234946966171, "global_step": 206452, "epoch": 2319, "lr": 1.353926841623217e-05}
{"train_loss": 0.02958867698907852, "global_step": 206453, "epoch": 2319, "lr": 1.353887171359609e-05}
{"train_loss": 0.05883430689573288, "global_step": 206454, "epoch": 2319, "lr": 1.3538475015861784e-05}
{"train_loss": 0.05451497435569763, "global_step": 206455, "epoch": 2319, "lr": 1.3538078323029269e-05}
{"train_loss": 0.06899368762969971, "global_step": 206456, "epoch": 2319, "lr": 1.3537681635098631e-05}
{"train_loss": 0.06981078535318375, "global_step": 206457, "epoch": 2319, "lr": 1.3537284952069896e-05}
{"train_loss": 0.03880506381392479, "global_step": 206458, "epoch": 2319, "lr": 1.3536888273943138e-05}
{"train_loss": 0.040230415761470795, "global_step": 206459, "epoch": 2319, "lr": 1.3536491600718404e-05}
{"train_loss": 0.0615914911031723, "global_step": 206460, "epoch": 2319, "lr": 1.3536094932395726e-05}
{"train_loss": 0.031559646129608154, "global_step": 206461, "epoch": 2319, "lr": 1.3535698268975195e-05}
{"train_loss": 0.02625158801674843, "global_step": 206462, "epoch": 2319, "lr": 1.3535301610456825e-05}
{"train_loss": 0.0534842424094677, "global_step": 206463, "epoch": 2319, "lr": 1.3534904956840705e-05}
{"train_loss": 0.07199955731630325, "global_step": 206464, "epoch": 2319, "lr": 1.3534508308126858e-05}
{"train_loss": 0.04254201799631119, "global_step": 206465, "epoch": 2319, "lr": 1.3534111664315369e-05}
{"train_loss": 0.09148125350475311, "global_step": 206466, "epoch": 2319, "lr": 1.3533715025406257e-05}
{"train_loss": 0.04761115089058876, "global_step": 206467, "epoch": 2319, "lr": 1.3533318391399612e-05}
{"train_loss": 0.04579214006662369, "global_step": 206468, "epoch": 2319, "lr": 1.3532921762295447e-05}
{"train_loss": 0.07374853640794754, "global_step": 206469, "epoch": 2319, "lr": 1.3532525138093855e-05}
{"train_loss": 0.08639239519834518, "global_step": 206470, "epoch": 2319, "lr": 1.3532128518794856e-05}
{"train_loss": 0.058377377688884735, "global_step": 206471, "epoch": 2319, "lr": 1.353173190439853e-05}
{"train_loss": 0.03711215406656265, "global_step": 206472, "epoch": 2319, "lr": 1.3531335294904906e-05}
{"train_loss": 0.04244331270456314, "global_step": 206473, "epoch": 2319, "lr": 1.353093869031406e-05}
{"train_loss": 0.031104115769267082, "global_step": 206474, "epoch": 2319, "lr": 1.3530542090626025e-05}
{"train_loss": 0.06455126404762268, "global_step": 206475, "epoch": 2319, "lr": 1.3530145495840879e-05}
{"train_loss": 0.11410719156265259, "global_step": 206476, "epoch": 2319, "lr": 1.3529748905958639e-05}
{"train_loss": 0.05347328633069992, "global_step": 206477, "epoch": 2319, "lr": 1.35293523209794e-05}
{"train_loss": 0.1020687073469162, "global_step": 206478, "epoch": 2319, "lr": 1.3528955740903182e-05}
{"train_loss": 0.060206983594244784, "global_step": 206479, "epoch": 2319, "lr": 1.3528559165730049e-05, "val_loss": 8.46157169342041}
{"train_loss": 0.08230696618556976, "global_step": 206480, "epoch": 2320, "lr": 1.352816259546007e-05}
{"train_loss": 0.06053612753748894, "global_step": 206481, "epoch": 2320, "lr": 1.352776603009327e-05}
{"train_loss": 0.05222364887595177, "global_step": 206482, "epoch": 2320, "lr": 1.3527369469629736e-05}
{"train_loss": 0.1180032417178154, "global_step": 206483, "epoch": 2320, "lr": 1.352697291406948e-05}
{"train_loss": 0.08897857367992401, "global_step": 206484, "epoch": 2320, "lr": 1.3526576363412603e-05}
{"train_loss": 0.0368337519466877, "global_step": 206485, "epoch": 2320, "lr": 1.3526179817659112e-05}
{"train_loss": 0.07636190950870514, "global_step": 206486, "epoch": 2320, "lr": 1.3525783276809095e-05}
{"train_loss": 0.03440345823764801, "global_step": 206487, "epoch": 2320, "lr": 1.3525386740862583e-05}
{"train_loss": 0.054971594363451004, "global_step": 206488, "epoch": 2320, "lr": 1.3524990209819649e-05}
{"train_loss": 0.0531855970621109, "global_step": 206489, "epoch": 2320, "lr": 1.352459368368032e-05}
{"train_loss": 0.04174911975860596, "global_step": 206490, "epoch": 2320, "lr": 1.3524197162444685e-05}
{"train_loss": 0.04203059524297714, "global_step": 206491, "epoch": 2320, "lr": 1.3523800646112755e-05}
{"train_loss": 0.0652916356921196, "global_step": 206492, "epoch": 2320, "lr": 1.3523404134684624e-05}
{"train_loss": 0.07754799723625183, "global_step": 206493, "epoch": 2320, "lr": 1.3523007628160311e-05}
{"train_loss": 0.05233781784772873, "global_step": 206494, "epoch": 2320, "lr": 1.3522611126539902e-05}
{"train_loss": 0.0631679818034172, "global_step": 206495, "epoch": 2320, "lr": 1.3522214629823415e-05}
{"train_loss": 0.035740409046411514, "global_step": 206496, "epoch": 2320, "lr": 1.3521818138010938e-05}
{"train_loss": 0.07555415481328964, "global_step": 206497, "epoch": 2320, "lr": 1.3521421651102512e-05}
{"train_loss": 0.01962796039879322, "global_step": 206498, "epoch": 2320, "lr": 1.3521025169098168e-05}
{"train_loss": 0.07305677235126495, "global_step": 206499, "epoch": 2320, "lr": 1.3520628691997988e-05}
{"train_loss": 0.06451024860143661, "global_step": 206500, "epoch": 2320, "lr": 1.352023221980201e-05}
{"train_loss": 0.0654435008764267, "global_step": 206501, "epoch": 2320, "lr": 1.3519835752510296e-05}
{"train_loss": 0.08012036234140396, "global_step": 206502, "epoch": 2320, "lr": 1.3519439290122887e-05}
{"train_loss": 0.11427604407072067, "global_step": 206503, "epoch": 2320, "lr": 1.351904283263986e-05}
{"train_loss": 0.05460464581847191, "global_step": 206504, "epoch": 2320, "lr": 1.3518646380061234e-05}
{"train_loss": 0.03617396578192711, "global_step": 206505, "epoch": 2320, "lr": 1.35182499323871e-05}
{"train_loss": 0.03810655698180199, "global_step": 206506, "epoch": 2320, "lr": 1.3517853489617478e-05}
{"train_loss": 0.0696500763297081, "global_step": 206507, "epoch": 2320, "lr": 1.3517457051752447e-05}
{"train_loss": 0.08904297649860382, "global_step": 206508, "epoch": 2320, "lr": 1.3517060618792037e-05}
{"train_loss": 0.0785137414932251, "global_step": 206509, "epoch": 2320, "lr": 1.3516664190736328e-05}
{"train_loss": 0.06318138539791107, "global_step": 206510, "epoch": 2320, "lr": 1.3516267767585344e-05}
{"train_loss": 0.06770438700914383, "global_step": 206511, "epoch": 2320, "lr": 1.351587134933917e-05}
{"train_loss": 0.10434438288211823, "global_step": 206512, "epoch": 2320, "lr": 1.3515474935997824e-05}
{"train_loss": 0.06796025484800339, "global_step": 206513, "epoch": 2320, "lr": 1.3515078527561398e-05}
{"train_loss": 0.04797015339136124, "global_step": 206514, "epoch": 2320, "lr": 1.3514682124029903e-05}
{"train_loss": 0.04816354438662529, "global_step": 206515, "epoch": 2320, "lr": 1.351428572540342e-05}
{"train_loss": 0.09383832663297653, "global_step": 206516, "epoch": 2320, "lr": 1.3513889331682006e-05}
{"train_loss": 0.047613389790058136, "global_step": 206517, "epoch": 2320, "lr": 1.3513492942865696e-05}
{"train_loss": 0.03935028612613678, "global_step": 206518, "epoch": 2320, "lr": 1.3513096558954564e-05}
{"train_loss": 0.08402755856513977, "global_step": 206519, "epoch": 2320, "lr": 1.3512700179948635e-05}
{"train_loss": 0.100413978099823, "global_step": 206520, "epoch": 2320, "lr": 1.3512303805847997e-05}
{"train_loss": 0.090485118329525, "global_step": 206521, "epoch": 2320, "lr": 1.3511907436652672e-05}
{"train_loss": 0.03503541648387909, "global_step": 206522, "epoch": 2320, "lr": 1.3511511072362737e-05}
{"train_loss": 0.08988142758607864, "global_step": 206523, "epoch": 2320, "lr": 1.3511114712978218e-05}
{"train_loss": 0.04203663021326065, "global_step": 206524, "epoch": 2320, "lr": 1.3510718358499202e-05}
{"train_loss": 0.07085846364498138, "global_step": 206525, "epoch": 2320, "lr": 1.3510322008925708e-05}
{"train_loss": 0.06559980660676956, "global_step": 206526, "epoch": 2320, "lr": 1.3509925664257827e-05}
{"train_loss": 0.0445755198597908, "global_step": 206527, "epoch": 2320, "lr": 1.3509529324495574e-05}
{"train_loss": 0.036694858223199844, "global_step": 206528, "epoch": 2320, "lr": 1.3509132989639035e-05}
{"train_loss": 0.03760990872979164, "global_step": 206529, "epoch": 2320, "lr": 1.3508736659688231e-05}
{"train_loss": 0.06293317675590515, "global_step": 206530, "epoch": 2320, "lr": 1.350834033464325e-05}
{"train_loss": 0.05859465152025223, "global_step": 206531, "epoch": 2320, "lr": 1.3507944014504115e-05}
{"train_loss": 0.05621591582894325, "global_step": 206532, "epoch": 2320, "lr": 1.35075476992709e-05}
{"train_loss": 0.0860808789730072, "global_step": 206533, "epoch": 2320, "lr": 1.3507151388943646e-05}
{"train_loss": 0.032591260969638824, "global_step": 206534, "epoch": 2320, "lr": 1.3506755083522415e-05}
{"train_loss": 0.02681867778301239, "global_step": 206535, "epoch": 2320, "lr": 1.3506358783007245e-05}
{"train_loss": 0.09046153724193573, "global_step": 206536, "epoch": 2320, "lr": 1.3505962487398216e-05}
{"train_loss": 0.06456156075000763, "global_step": 206537, "epoch": 2320, "lr": 1.3505566196695369e-05}
{"train_loss": 0.0697304755449295, "global_step": 206538, "epoch": 2320, "lr": 1.3505169910898734e-05}
{"train_loss": 0.03429475426673889, "global_step": 206539, "epoch": 2320, "lr": 1.35047736300084e-05}
{"train_loss": 0.02511000819504261, "global_step": 206540, "epoch": 2320, "lr": 1.3504377354024389e-05}
{"train_loss": 0.09586718678474426, "global_step": 206541, "epoch": 2320, "lr": 1.3503981082946787e-05}
{"train_loss": 0.08020323514938354, "global_step": 206542, "epoch": 2320, "lr": 1.3503584816775611e-05}
{"train_loss": 0.03972736746072769, "global_step": 206543, "epoch": 2320, "lr": 1.3503188555510955e-05}
{"train_loss": 0.018956486135721207, "global_step": 206544, "epoch": 2320, "lr": 1.3502792299152828e-05}
{"train_loss": 0.047074854373931885, "global_step": 206545, "epoch": 2320, "lr": 1.3502396047701327e-05}
{"train_loss": 0.0573972649872303, "global_step": 206546, "epoch": 2320, "lr": 1.3501999801156461e-05}
{"train_loss": 0.04547808691859245, "global_step": 206547, "epoch": 2320, "lr": 1.3501603559518328e-05}
{"train_loss": 0.041868798434734344, "global_step": 206548, "epoch": 2320, "lr": 1.3501207322786946e-05}
{"train_loss": 0.08986599743366241, "global_step": 206549, "epoch": 2320, "lr": 1.350081109096239e-05}
{"train_loss": 0.03508380800485611, "global_step": 206550, "epoch": 2320, "lr": 1.3500414864044696e-05}
{"train_loss": 0.05480608716607094, "global_step": 206551, "epoch": 2320, "lr": 1.3500018642033924e-05}
{"train_loss": 0.05640614032745361, "global_step": 206552, "epoch": 2320, "lr": 1.3499622424930142e-05}
{"train_loss": 0.08150461316108704, "global_step": 206553, "epoch": 2320, "lr": 1.3499226212733379e-05}
{"train_loss": 0.07820651680231094, "global_step": 206554, "epoch": 2320, "lr": 1.3498830005443714e-05}
{"train_loss": 0.0531092993915081, "global_step": 206555, "epoch": 2320, "lr": 1.3498433803061173e-05}
{"train_loss": 0.07465925067663193, "global_step": 206556, "epoch": 2320, "lr": 1.349803760558584e-05}
{"train_loss": 0.07043575495481491, "global_step": 206557, "epoch": 2320, "lr": 1.349764141301773e-05}
{"train_loss": 0.07048187404870987, "global_step": 206558, "epoch": 2320, "lr": 1.3497245225356936e-05}
{"train_loss": 0.05843915790319443, "global_step": 206559, "epoch": 2320, "lr": 1.3496849042603477e-05}
{"train_loss": 0.10342036187648773, "global_step": 206560, "epoch": 2320, "lr": 1.3496452864757436e-05}
{"train_loss": 0.019541889429092407, "global_step": 206561, "epoch": 2320, "lr": 1.3496056691818838e-05}
{"train_loss": 0.03821958228945732, "global_step": 206562, "epoch": 2320, "lr": 1.3495660523787762e-05}
{"train_loss": 0.048014428466558456, "global_step": 206563, "epoch": 2320, "lr": 1.3495264360664234e-05}
{"train_loss": 0.04536646977066994, "global_step": 206564, "epoch": 2320, "lr": 1.3494868202448341e-05}
{"train_loss": 0.0803411677479744, "global_step": 206565, "epoch": 2320, "lr": 1.3494472049140106e-05}
{"train_loss": 0.08960937708616257, "global_step": 206566, "epoch": 2320, "lr": 1.3494075900739605e-05}
{"train_loss": 0.07037647813558578, "global_step": 206567, "epoch": 2320, "lr": 1.3493679757246864e-05}
{"train_loss": 0.06165902513299096, "global_step": 206568, "epoch": 2320, "lr": 1.3493283618661972e-05, "val_loss": 8.453692436218262, "train_action_mse_error": 34.20783996582031}
{"train_loss": 0.04914724826812744, "global_step": 206569, "epoch": 2321, "lr": 1.3492887484984945e-05}
{"train_loss": 0.07210138440132141, "global_step": 206570, "epoch": 2321, "lr": 1.349249135621587e-05}
{"train_loss": 0.020259134471416473, "global_step": 206571, "epoch": 2321, "lr": 1.3492095232354773e-05}
{"train_loss": 0.0576210655272007, "global_step": 206572, "epoch": 2321, "lr": 1.3491699113401729e-05}
{"train_loss": 0.023098062723875046, "global_step": 206573, "epoch": 2321, "lr": 1.3491302999356769e-05}
{"train_loss": 0.035815849900245667, "global_step": 206574, "epoch": 2321, "lr": 1.3490906890219974e-05}
{"train_loss": 0.10202974826097488, "global_step": 206575, "epoch": 2321, "lr": 1.3490510785991379e-05}
{"train_loss": 0.045905761420726776, "global_step": 206576, "epoch": 2321, "lr": 1.3490114686671025e-05}
{"train_loss": 0.039745792746543884, "global_step": 206577, "epoch": 2321, "lr": 1.3489718592259e-05}
{"train_loss": 0.03442832827568054, "global_step": 206578, "epoch": 2321, "lr": 1.3489322502755313e-05}
{"train_loss": 0.035469066351652145, "global_step": 206579, "epoch": 2321, "lr": 1.3488926418160064e-05}
{"train_loss": 0.02485000155866146, "global_step": 206580, "epoch": 2321, "lr": 1.3488530338473265e-05}
{"train_loss": 0.06705072522163391, "global_step": 206581, "epoch": 2321, "lr": 1.3488134263695007e-05}
{"train_loss": 0.05162816867232323, "global_step": 206582, "epoch": 2321, "lr": 1.3487738193825306e-05}
{"train_loss": 0.08530454337596893, "global_step": 206583, "epoch": 2321, "lr": 1.348734212886425e-05}
{"train_loss": 0.08559466898441315, "global_step": 206584, "epoch": 2321, "lr": 1.3486946068811863e-05}
{"train_loss": 0.062381453812122345, "global_step": 206585, "epoch": 2321, "lr": 1.348655001366822e-05}
{"train_loss": 0.10625379532575607, "global_step": 206586, "epoch": 2321, "lr": 1.3486153963433356e-05}
{"train_loss": 0.12064676731824875, "global_step": 206587, "epoch": 2321, "lr": 1.3485757918107344e-05}
{"train_loss": 0.05000048875808716, "global_step": 206588, "epoch": 2321, "lr": 1.3485361877690217e-05}
{"train_loss": 0.13367076218128204, "global_step": 206589, "epoch": 2321, "lr": 1.3484965842182035e-05}
{"train_loss": 0.05878123641014099, "global_step": 206590, "epoch": 2321, "lr": 1.348456981158287e-05}
{"train_loss": 0.04108831658959389, "global_step": 206591, "epoch": 2321, "lr": 1.3484173785892745e-05}
{"train_loss": 0.04567590355873108, "global_step": 206592, "epoch": 2321, "lr": 1.3483777765111737e-05}
{"train_loss": 0.051438476890325546, "global_step": 206593, "epoch": 2321, "lr": 1.3483381749239882e-05}
{"train_loss": 0.07045096904039383, "global_step": 206594, "epoch": 2321, "lr": 1.3482985738277254e-05}
{"train_loss": 0.06567859649658203, "global_step": 206595, "epoch": 2321, "lr": 1.3482589732223877e-05}
{"train_loss": 0.02661616913974285, "global_step": 206596, "epoch": 2321, "lr": 1.348219373107984e-05}
{"train_loss": 0.05299997329711914, "global_step": 206597, "epoch": 2321, "lr": 1.3481797734845158e-05}
{"train_loss": 0.049461763352155685, "global_step": 206598, "epoch": 2321, "lr": 1.3481401743519923e-05}
{"train_loss": 0.05612270161509514, "global_step": 206599, "epoch": 2321, "lr": 1.3481005757104148e-05}
{"train_loss": 0.030318012461066246, "global_step": 206600, "epoch": 2321, "lr": 1.3480609775597925e-05}
{"train_loss": 0.02355450764298439, "global_step": 206601, "epoch": 2321, "lr": 1.3480213799001274e-05}
{"train_loss": 0.07955481112003326, "global_step": 206602, "epoch": 2321, "lr": 1.3479817827314273e-05}
{"train_loss": 0.05626271292567253, "global_step": 206603, "epoch": 2321, "lr": 1.3479421860536956e-05}
{"train_loss": 0.05004552751779556, "global_step": 206604, "epoch": 2321, "lr": 1.3479025898669395e-05}
{"train_loss": 0.032355550676584244, "global_step": 206605, "epoch": 2321, "lr": 1.3478629941711623e-05}
{"train_loss": 0.05937434732913971, "global_step": 206606, "epoch": 2321, "lr": 1.3478233989663708e-05}
{"train_loss": 0.09380493313074112, "global_step": 206607, "epoch": 2321, "lr": 1.3477838042525708e-05}
{"train_loss": 0.09721719473600388, "global_step": 206608, "epoch": 2321, "lr": 1.3477442100297672e-05}
{"train_loss": 0.06196392700076103, "global_step": 206609, "epoch": 2321, "lr": 1.3477046162979628e-05}
{"train_loss": 0.0587538406252861, "global_step": 206610, "epoch": 2321, "lr": 1.347665023057167e-05}
{"train_loss": 0.09860776364803314, "global_step": 206611, "epoch": 2321, "lr": 1.3476254303073816e-05}
{"train_loss": 0.042484063655138016, "global_step": 206612, "epoch": 2321, "lr": 1.3475858380486145e-05}
{"train_loss": 0.06477631628513336, "global_step": 206613, "epoch": 2321, "lr": 1.3475462462808702e-05}
{"train_loss": 0.08884088695049286, "global_step": 206614, "epoch": 2321, "lr": 1.3475066550041527e-05}
{"train_loss": 0.024094130843877792, "global_step": 206615, "epoch": 2321, "lr": 1.3474670642184695e-05}
{"train_loss": 0.049684830009937286, "global_step": 206616, "epoch": 2321, "lr": 1.347427473923823e-05}
{"train_loss": 0.10743269324302673, "global_step": 206617, "epoch": 2321, "lr": 1.3473878841202226e-05}
{"train_loss": 0.06541778147220612, "global_step": 206618, "epoch": 2321, "lr": 1.3473482948076693e-05}
{"train_loss": 0.0412379615008831, "global_step": 206619, "epoch": 2321, "lr": 1.3473087059861722e-05}
{"train_loss": 0.04551364481449127, "global_step": 206620, "epoch": 2321, "lr": 1.3472691176557334e-05}
{"train_loss": 0.04216312617063522, "global_step": 206621, "epoch": 2321, "lr": 1.3472295298163611e-05}
{"train_loss": 0.06169392541050911, "global_step": 206622, "epoch": 2321, "lr": 1.3471899424680584e-05}
{"train_loss": 0.05770706757903099, "global_step": 206623, "epoch": 2321, "lr": 1.3471503556108322e-05}
{"train_loss": 0.039716750383377075, "global_step": 206624, "epoch": 2321, "lr": 1.347110769244686e-05}
{"train_loss": 0.06375652551651001, "global_step": 206625, "epoch": 2321, "lr": 1.3470711833696264e-05}
{"train_loss": 0.04828814044594765, "global_step": 206626, "epoch": 2321, "lr": 1.3470315979856602e-05}
{"train_loss": 0.0867694541811943, "global_step": 206627, "epoch": 2321, "lr": 1.3469920130927888e-05}
{"train_loss": 0.051923662424087524, "global_step": 206628, "epoch": 2321, "lr": 1.3469524286910218e-05}
{"train_loss": 0.0557863675057888, "global_step": 206629, "epoch": 2321, "lr": 1.3469128447803608e-05}
{"train_loss": 0.05719839408993721, "global_step": 206630, "epoch": 2321, "lr": 1.3468732613608148e-05}
{"train_loss": 0.05843879282474518, "global_step": 206631, "epoch": 2321, "lr": 1.3468336784323854e-05}
{"train_loss": 0.11546095460653305, "global_step": 206632, "epoch": 2321, "lr": 1.346794095995081e-05}
{"train_loss": 0.06082293391227722, "global_step": 206633, "epoch": 2321, "lr": 1.3467545140489041e-05}
{"train_loss": 0.05116858705878258, "global_step": 206634, "epoch": 2321, "lr": 1.3467149325938633e-05}
{"train_loss": 0.033545006066560745, "global_step": 206635, "epoch": 2321, "lr": 1.3466753516299606e-05}
{"train_loss": 0.04735999181866646, "global_step": 206636, "epoch": 2321, "lr": 1.3466357711572041e-05}
{"train_loss": 0.04978252947330475, "global_step": 206637, "epoch": 2321, "lr": 1.346596191175597e-05}
{"train_loss": 0.07700076699256897, "global_step": 206638, "epoch": 2321, "lr": 1.3465566116851464e-05}
{"train_loss": 0.09827546030282974, "global_step": 206639, "epoch": 2321, "lr": 1.3465170326858556e-05}
{"train_loss": 0.038067884743213654, "global_step": 206640, "epoch": 2321, "lr": 1.3464774541777326e-05}
{"train_loss": 0.05209778994321823, "global_step": 206641, "epoch": 2321, "lr": 1.3464378761607793e-05}
{"train_loss": 0.03291253373026848, "global_step": 206642, "epoch": 2321, "lr": 1.3463982986350032e-05}
{"train_loss": 0.07128796726465225, "global_step": 206643, "epoch": 2321, "lr": 1.3463587216004103e-05}
{"train_loss": 0.05422234907746315, "global_step": 206644, "epoch": 2321, "lr": 1.3463191450570039e-05}
{"train_loss": 0.055338941514492035, "global_step": 206645, "epoch": 2321, "lr": 1.3462795690047919e-05}
{"train_loss": 0.11302851140499115, "global_step": 206646, "epoch": 2321, "lr": 1.3462399934437764e-05}
{"train_loss": 0.054481666535139084, "global_step": 206647, "epoch": 2321, "lr": 1.3462004183739663e-05}
{"train_loss": 0.06791514158248901, "global_step": 206648, "epoch": 2321, "lr": 1.3461608437953644e-05}
{"train_loss": 0.07656780630350113, "global_step": 206649, "epoch": 2321, "lr": 1.3461212697079756e-05}
{"train_loss": 0.06337106227874756, "global_step": 206650, "epoch": 2321, "lr": 1.3460816961118077e-05}
{"train_loss": 0.08049016445875168, "global_step": 206651, "epoch": 2321, "lr": 1.3460421230068626e-05}
{"train_loss": 0.09210072457790375, "global_step": 206652, "epoch": 2321, "lr": 1.3460025503931495e-05}
{"train_loss": 0.049331553280353546, "global_step": 206653, "epoch": 2321, "lr": 1.3459629782706722e-05}
{"train_loss": 0.06795895099639893, "global_step": 206654, "epoch": 2321, "lr": 1.3459234066394338e-05}
{"train_loss": 0.02545599825680256, "global_step": 206655, "epoch": 2321, "lr": 1.345883835499443e-05}
{"train_loss": 0.10669920593500137, "global_step": 206656, "epoch": 2321, "lr": 1.345844264850702e-05}
{"train_loss": 0.06040019422685832, "global_step": 206657, "epoch": 2321, "lr": 1.345804694693219e-05, "val_loss": 8.442341804504395}
{"train_loss": 0.06981541961431503, "global_step": 206658, "epoch": 2322, "lr": 1.3457651250269965e-05}
{"train_loss": 0.05231371149420738, "global_step": 206659, "epoch": 2322, "lr": 1.3457255558520431e-05}
{"train_loss": 0.07378415018320084, "global_step": 206660, "epoch": 2322, "lr": 1.3456859871683608e-05}
{"train_loss": 0.12083239108324051, "global_step": 206661, "epoch": 2322, "lr": 1.3456464189759565e-05}
{"train_loss": 0.05756281688809395, "global_step": 206662, "epoch": 2322, "lr": 1.3456068512748371e-05}
{"train_loss": 0.013089695014059544, "global_step": 206663, "epoch": 2322, "lr": 1.3455672840650046e-05}
{"train_loss": 0.07129164040088654, "global_step": 206664, "epoch": 2322, "lr": 1.3455277173464675e-05}
{"train_loss": 0.046689894050359726, "global_step": 206665, "epoch": 2322, "lr": 1.3454881511192285e-05}
{"train_loss": 0.05510937422513962, "global_step": 206666, "epoch": 2322, "lr": 1.3454485853832955e-05}
{"train_loss": 0.03032781556248665, "global_step": 206667, "epoch": 2322, "lr": 1.3454090201386704e-05}
{"train_loss": 0.14200931787490845, "global_step": 206668, "epoch": 2322, "lr": 1.3453694553853624e-05}
{"train_loss": 0.021529532968997955, "global_step": 206669, "epoch": 2322, "lr": 1.3453298911233737e-05}
{"train_loss": 0.060791973024606705, "global_step": 206670, "epoch": 2322, "lr": 1.3452903273527118e-05}
{"train_loss": 0.08041129261255264, "global_step": 206671, "epoch": 2322, "lr": 1.3452507640733797e-05}
{"train_loss": 0.055455513298511505, "global_step": 206672, "epoch": 2322, "lr": 1.3452112012853863e-05}
{"train_loss": 0.06689353287220001, "global_step": 206673, "epoch": 2322, "lr": 1.3451716389887325e-05}
{"train_loss": 0.05574430525302887, "global_step": 206674, "epoch": 2322, "lr": 1.3451320771834274e-05}
{"train_loss": 0.03500036150217056, "global_step": 206675, "epoch": 2322, "lr": 1.3450925158694733e-05}
{"train_loss": 0.12379873543977737, "global_step": 206676, "epoch": 2322, "lr": 1.3450529550468783e-05}
{"train_loss": 0.09698813408613205, "global_step": 206677, "epoch": 2322, "lr": 1.3450133947156452e-05}
{"train_loss": 0.051568616181612015, "global_step": 206678, "epoch": 2322, "lr": 1.3449738348757801e-05}
{"train_loss": 0.08881344646215439, "global_step": 206679, "epoch": 2322, "lr": 1.344934275527291e-05}
{"train_loss": 0.042897433042526245, "global_step": 206680, "epoch": 2322, "lr": 1.3448947166701786e-05}
{"train_loss": 0.07150820642709732, "global_step": 206681, "epoch": 2322, "lr": 1.3448551583044522e-05}
{"train_loss": 0.026148749515414238, "global_step": 206682, "epoch": 2322, "lr": 1.3448156004301144e-05}
{"train_loss": 0.05557931214570999, "global_step": 206683, "epoch": 2322, "lr": 1.3447760430471728e-05}
{"train_loss": 0.04745735973119736, "global_step": 206684, "epoch": 2322, "lr": 1.34473648615563e-05}
{"train_loss": 0.0436692014336586, "global_step": 206685, "epoch": 2322, "lr": 1.3446969297554945e-05}
{"train_loss": 0.023852096870541573, "global_step": 206686, "epoch": 2322, "lr": 1.3446573738467693e-05}
{"train_loss": 0.05134033411741257, "global_step": 206687, "epoch": 2322, "lr": 1.3446178184294594e-05}
{"train_loss": 0.08732956647872925, "global_step": 206688, "epoch": 2322, "lr": 1.344578263503573e-05}
{"train_loss": 0.11238041520118713, "global_step": 206689, "epoch": 2322, "lr": 1.3445387090691114e-05}
{"train_loss": 0.10635059326887131, "global_step": 206690, "epoch": 2322, "lr": 1.3444991551260837e-05}
{"train_loss": 0.056211553514003754, "global_step": 206691, "epoch": 2322, "lr": 1.3444596016744937e-05}
{"train_loss": 0.10321664810180664, "global_step": 206692, "epoch": 2322, "lr": 1.344420048714345e-05}
{"train_loss": 0.04524388536810875, "global_step": 206693, "epoch": 2322, "lr": 1.3443804962456458e-05}
{"train_loss": 0.08770370483398438, "global_step": 206694, "epoch": 2322, "lr": 1.3443409442683986e-05}
{"train_loss": 0.07332388311624527, "global_step": 206695, "epoch": 2322, "lr": 1.3443013927826115e-05}
{"train_loss": 0.08315470069646835, "global_step": 206696, "epoch": 2322, "lr": 1.3442618417882874e-05}
{"train_loss": 0.0787329450249672, "global_step": 206697, "epoch": 2322, "lr": 1.3442222912854324e-05}
{"train_loss": 0.059390317648649216, "global_step": 206698, "epoch": 2322, "lr": 1.3441827412740537e-05}
{"train_loss": 0.05510730668902397, "global_step": 206699, "epoch": 2322, "lr": 1.3441431917541536e-05}
{"train_loss": 0.08758067339658737, "global_step": 206700, "epoch": 2322, "lr": 1.3441036427257409e-05}
{"train_loss": 0.07585703581571579, "global_step": 206701, "epoch": 2322, "lr": 1.3440640941888166e-05}
{"train_loss": 0.053319867700338364, "global_step": 206702, "epoch": 2322, "lr": 1.3440245461433904e-05}
{"train_loss": 0.05724423751235008, "global_step": 206703, "epoch": 2322, "lr": 1.3439849985894636e-05}
{"train_loss": 0.06936697661876678, "global_step": 206704, "epoch": 2322, "lr": 1.3439454515270455e-05}
{"train_loss": 0.06748679280281067, "global_step": 206705, "epoch": 2322, "lr": 1.3439059049561376e-05}
{"train_loss": 0.05075010284781456, "global_step": 206706, "epoch": 2322, "lr": 1.3438663588767487e-05}
{"train_loss": 0.07707683742046356, "global_step": 206707, "epoch": 2322, "lr": 1.3438268132888804e-05}
{"train_loss": 0.08416526764631271, "global_step": 206708, "epoch": 2322, "lr": 1.3437872681925418e-05}
{"train_loss": 0.08941821753978729, "global_step": 206709, "epoch": 2322, "lr": 1.343747723587735e-05}
{"train_loss": 0.11303967982530594, "global_step": 206710, "epoch": 2322, "lr": 1.3437081794744682e-05}
{"train_loss": 0.055106598883867264, "global_step": 206711, "epoch": 2322, "lr": 1.343668635852744e-05}
{"train_loss": 0.09446072578430176, "global_step": 206712, "epoch": 2322, "lr": 1.3436290927225703e-05}
{"train_loss": 0.039116691797971725, "global_step": 206713, "epoch": 2322, "lr": 1.3435895500839497e-05}
{"train_loss": 0.04477534070611, "global_step": 206714, "epoch": 2322, "lr": 1.3435500079368902e-05}
{"train_loss": 0.06925641745328903, "global_step": 206715, "epoch": 2322, "lr": 1.343510466281394e-05}
{"train_loss": 0.0947546660900116, "global_step": 206716, "epoch": 2322, "lr": 1.3434709251174688e-05}
{"train_loss": 0.08207908272743225, "global_step": 206717, "epoch": 2322, "lr": 1.3434313844451208e-05}
{"train_loss": 0.05778408795595169, "global_step": 206718, "epoch": 2322, "lr": 1.3433918442643522e-05}
{"train_loss": 0.0678405836224556, "global_step": 206719, "epoch": 2322, "lr": 1.3433523045751716e-05}
{"train_loss": 0.11669666320085526, "global_step": 206720, "epoch": 2322, "lr": 1.3433127653775811e-05}
{"train_loss": 0.041443776339292526, "global_step": 206721, "epoch": 2322, "lr": 1.3432732266715898e-05}
{"train_loss": 0.11080282181501389, "global_step": 206722, "epoch": 2322, "lr": 1.343233688457199e-05}
{"train_loss": 0.08601006120443344, "global_step": 206723, "epoch": 2322, "lr": 1.343194150734417e-05}
{"train_loss": 0.08893019706010818, "global_step": 206724, "epoch": 2322, "lr": 1.343154613503248e-05}
{"train_loss": 0.0445416197180748, "global_step": 206725, "epoch": 2322, "lr": 1.3431150767636957e-05}
{"train_loss": 0.08310557901859283, "global_step": 206726, "epoch": 2322, "lr": 1.343075540515769e-05}
{"train_loss": 0.037376124411821365, "global_step": 206727, "epoch": 2322, "lr": 1.3430360047594691e-05}
{"train_loss": 0.094149149954319, "global_step": 206728, "epoch": 2322, "lr": 1.3429964694948055e-05}
{"train_loss": 0.13453680276870728, "global_step": 206729, "epoch": 2322, "lr": 1.3429569347217814e-05}
{"train_loss": 0.04057653248310089, "global_step": 206730, "epoch": 2322, "lr": 1.3429174004404004e-05}
{"train_loss": 0.05614421144127846, "global_step": 206731, "epoch": 2322, "lr": 1.3428778666506713e-05}
{"train_loss": 0.03816639631986618, "global_step": 206732, "epoch": 2322, "lr": 1.3428383333525962e-05}
{"train_loss": 0.09309542179107666, "global_step": 206733, "epoch": 2322, "lr": 1.3427988005461817e-05}
{"train_loss": 0.065538689494133, "global_step": 206734, "epoch": 2322, "lr": 1.342759268231435e-05}
{"train_loss": 0.06035633757710457, "global_step": 206735, "epoch": 2322, "lr": 1.3427197364083583e-05}
{"train_loss": 0.042540039867162704, "global_step": 206736, "epoch": 2322, "lr": 1.3426802050769598e-05}
{"train_loss": 0.030643029138445854, "global_step": 206737, "epoch": 2322, "lr": 1.3426406742372417e-05}
{"train_loss": 0.058109745383262634, "global_step": 206738, "epoch": 2322, "lr": 1.3426011438892127e-05}
{"train_loss": 0.04950148984789848, "global_step": 206739, "epoch": 2322, "lr": 1.3425616140328746e-05}
{"train_loss": 0.06700780242681503, "global_step": 206740, "epoch": 2322, "lr": 1.3425220846682363e-05}
{"train_loss": 0.04533760994672775, "global_step": 206741, "epoch": 2322, "lr": 1.3424825557952997e-05}
{"train_loss": 0.032477688044309616, "global_step": 206742, "epoch": 2322, "lr": 1.3424430274140731e-05}
{"train_loss": 0.045209068804979324, "global_step": 206743, "epoch": 2322, "lr": 1.3424034995245594e-05}
{"train_loss": 0.08834777772426605, "global_step": 206744, "epoch": 2322, "lr": 1.3423639721267662e-05}
{"train_loss": 0.036911316215991974, "global_step": 206745, "epoch": 2322, "lr": 1.3423244452206963e-05}
{"train_loss": 0.06701337457163616, "global_step": 206746, "epoch": 2322, "lr": 1.3422849188063575e-05, "val_loss": 8.299344062805176}
{"train_loss": 0.08835813403129578, "global_step": 206747, "epoch": 2323, "lr": 1.3422453928837525e-05}
{"train_loss": 0.03330773115158081, "global_step": 206748, "epoch": 2323, "lr": 1.3422058674528892e-05}
{"train_loss": 0.0320606604218483, "global_step": 206749, "epoch": 2323, "lr": 1.342166342513771e-05}
{"train_loss": 0.06275688111782074, "global_step": 206750, "epoch": 2323, "lr": 1.342126818066405e-05}
{"train_loss": 0.03945312276482582, "global_step": 206751, "epoch": 2323, "lr": 1.3420872941107937e-05}
{"train_loss": 0.042719896882772446, "global_step": 206752, "epoch": 2323, "lr": 1.3420477706469442e-05}
{"train_loss": 0.03304009884595871, "global_step": 206753, "epoch": 2323, "lr": 1.3420082476748635e-05}
{"train_loss": 0.052685219794511795, "global_step": 206754, "epoch": 2323, "lr": 1.341968725194554e-05}
{"train_loss": 0.052611805498600006, "global_step": 206755, "epoch": 2323, "lr": 1.3419292032060238e-05}
{"train_loss": 0.09035767614841461, "global_step": 206756, "epoch": 2323, "lr": 1.3418896817092747e-05}
{"train_loss": 0.06186254695057869, "global_step": 206757, "epoch": 2323, "lr": 1.3418501607043155e-05}
{"train_loss": 0.07268646359443665, "global_step": 206758, "epoch": 2323, "lr": 1.3418106401911485e-05}
{"train_loss": 0.04789319261908531, "global_step": 206759, "epoch": 2323, "lr": 1.3417711201697819e-05}
{"train_loss": 0.07355981320142746, "global_step": 206760, "epoch": 2323, "lr": 1.3417316006402182e-05}
{"train_loss": 0.04542339965701103, "global_step": 206761, "epoch": 2323, "lr": 1.3416920816024658e-05}
{"train_loss": 0.030608924105763435, "global_step": 206762, "epoch": 2323, "lr": 1.3416525630565263e-05}
{"train_loss": 0.10009512305259705, "global_step": 206763, "epoch": 2323, "lr": 1.3416130450024089e-05}
{"train_loss": 0.12255162000656128, "global_step": 206764, "epoch": 2323, "lr": 1.341573527440117e-05}
{"train_loss": 0.04849465936422348, "global_step": 206765, "epoch": 2323, "lr": 1.3415340103696544e-05}
{"train_loss": 0.09699149429798126, "global_step": 206766, "epoch": 2323, "lr": 1.3414944937910295e-05}
{"train_loss": 0.040467504411935806, "global_step": 206767, "epoch": 2323, "lr": 1.3414549777042446e-05}
{"train_loss": 0.025230765342712402, "global_step": 206768, "epoch": 2323, "lr": 1.3414154621093079e-05}
{"train_loss": 0.038376983255147934, "global_step": 206769, "epoch": 2323, "lr": 1.3413759470062215e-05}
{"train_loss": 0.08175180107355118, "global_step": 206770, "epoch": 2323, "lr": 1.3413364323949945e-05}
{"train_loss": 0.04000142961740494, "global_step": 206771, "epoch": 2323, "lr": 1.3412969182756285e-05}
{"train_loss": 0.04975311830639839, "global_step": 206772, "epoch": 2323, "lr": 1.3412574046481319e-05}
{"train_loss": 0.033265773206949234, "global_step": 206773, "epoch": 2323, "lr": 1.3412178915125074e-05}
{"train_loss": 0.06966879218816757, "global_step": 206774, "epoch": 2323, "lr": 1.3411783788687632e-05}
{"train_loss": 0.027561593800783157, "global_step": 206775, "epoch": 2323, "lr": 1.3411388667169012e-05}
{"train_loss": 0.08116895705461502, "global_step": 206776, "epoch": 2323, "lr": 1.3410993550569302e-05}
{"train_loss": 0.06239032745361328, "global_step": 206777, "epoch": 2323, "lr": 1.3410598438888522e-05}
{"train_loss": 0.09615354984998703, "global_step": 206778, "epoch": 2323, "lr": 1.3410203332126752e-05}
{"train_loss": 0.09430159628391266, "global_step": 206779, "epoch": 2323, "lr": 1.3409808230284027e-05}
{"train_loss": 0.0654253214597702, "global_step": 206780, "epoch": 2323, "lr": 1.3409413133360416e-05}
{"train_loss": 0.06350050121545792, "global_step": 206781, "epoch": 2323, "lr": 1.3409018041355947e-05}
{"train_loss": 0.05859086662530899, "global_step": 206782, "epoch": 2323, "lr": 1.340862295427071e-05}
{"train_loss": 0.07782665640115738, "global_step": 206783, "epoch": 2323, "lr": 1.3408227872104723e-05}
{"train_loss": 0.046465933322906494, "global_step": 206784, "epoch": 2323, "lr": 1.3407832794858065e-05}
{"train_loss": 0.026183951646089554, "global_step": 206785, "epoch": 2323, "lr": 1.3407437722530764e-05}
{"train_loss": 0.04512356221675873, "global_step": 206786, "epoch": 2323, "lr": 1.3407042655122904e-05}
{"train_loss": 0.06113136559724808, "global_step": 206787, "epoch": 2323, "lr": 1.3406647592634503e-05}
{"train_loss": 0.052738770842552185, "global_step": 206788, "epoch": 2323, "lr": 1.3406252535065633e-05}
{"train_loss": 0.03858547285199165, "global_step": 206789, "epoch": 2323, "lr": 1.3405857482416362e-05}
{"train_loss": 0.07904959470033646, "global_step": 206790, "epoch": 2323, "lr": 1.340546243468671e-05}
{"train_loss": 0.02158733457326889, "global_step": 206791, "epoch": 2323, "lr": 1.3405067391876769e-05}
{"train_loss": 0.038847666233778, "global_step": 206792, "epoch": 2323, "lr": 1.3404672353986553e-05}
{"train_loss": 0.06933081895112991, "global_step": 206793, "epoch": 2323, "lr": 1.3404277321016146e-05}
{"train_loss": 0.049090005457401276, "global_step": 206794, "epoch": 2323, "lr": 1.3403882292965569e-05}
{"train_loss": 0.023232173174619675, "global_step": 206795, "epoch": 2323, "lr": 1.3403487269834913e-05}
{"train_loss": 0.05181365832686424, "global_step": 206796, "epoch": 2323, "lr": 1.34030922516242e-05}
{"train_loss": 0.03747323155403137, "global_step": 206797, "epoch": 2323, "lr": 1.3402697238333506e-05}
{"train_loss": 0.03349959850311279, "global_step": 206798, "epoch": 2323, "lr": 1.340230222996286e-05}
{"train_loss": 0.06585631519556046, "global_step": 206799, "epoch": 2323, "lr": 1.3401907226512345e-05}
{"train_loss": 0.07633552700281143, "global_step": 206800, "epoch": 2323, "lr": 1.3401512227981982e-05}
{"train_loss": 0.04988257586956024, "global_step": 206801, "epoch": 2323, "lr": 1.340111723437185e-05}
{"train_loss": 0.04727410897612572, "global_step": 206802, "epoch": 2323, "lr": 1.3400722245681997e-05}
{"train_loss": 0.07261864840984344, "global_step": 206803, "epoch": 2323, "lr": 1.3400327261912453e-05}
{"train_loss": 0.04705887660384178, "global_step": 206804, "epoch": 2323, "lr": 1.3399932283063304e-05}
{"train_loss": 0.061103127896785736, "global_step": 206805, "epoch": 2323, "lr": 1.3399537309134575e-05}
{"train_loss": 0.07112104445695877, "global_step": 206806, "epoch": 2323, "lr": 1.3399142340126348e-05}
{"train_loss": 0.04687052220106125, "global_step": 206807, "epoch": 2323, "lr": 1.339874737603864e-05}
{"train_loss": 0.07479662448167801, "global_step": 206808, "epoch": 2323, "lr": 1.3398352416871545e-05}
{"train_loss": 0.053741246461868286, "global_step": 206809, "epoch": 2323, "lr": 1.3397957462625078e-05}
{"train_loss": 0.09139138460159302, "global_step": 206810, "epoch": 2323, "lr": 1.3397562513299323e-05}
{"train_loss": 0.06364738196134567, "global_step": 206811, "epoch": 2323, "lr": 1.3397167568894303e-05}
{"train_loss": 0.06964103132486343, "global_step": 206812, "epoch": 2323, "lr": 1.3396772629410109e-05}
{"train_loss": 0.04781293123960495, "global_step": 206813, "epoch": 2323, "lr": 1.339637769484675e-05}
{"train_loss": 0.08558627218008041, "global_step": 206814, "epoch": 2323, "lr": 1.3395982765204324e-05}
{"train_loss": 0.058760035783052444, "global_step": 206815, "epoch": 2323, "lr": 1.3395587840482844e-05}
{"train_loss": 0.037341102957725525, "global_step": 206816, "epoch": 2323, "lr": 1.3395192920682398e-05}
{"train_loss": 0.03143629431724548, "global_step": 206817, "epoch": 2323, "lr": 1.3394798005803006e-05}
{"train_loss": 0.059182602912187576, "global_step": 206818, "epoch": 2323, "lr": 1.339440309584475e-05}
{"train_loss": 0.0721295177936554, "global_step": 206819, "epoch": 2323, "lr": 1.339400819080766e-05}
{"train_loss": 0.03390858322381973, "global_step": 206820, "epoch": 2323, "lr": 1.339361329069181e-05}
{"train_loss": 0.0797145664691925, "global_step": 206821, "epoch": 2323, "lr": 1.3393218395497232e-05}
{"train_loss": 0.046956684440374374, "global_step": 206822, "epoch": 2323, "lr": 1.3392823505224e-05}
{"train_loss": 0.05602556839585304, "global_step": 206823, "epoch": 2323, "lr": 1.3392428619872144e-05}
{"train_loss": 0.045078862458467484, "global_step": 206824, "epoch": 2323, "lr": 1.3392033739441728e-05}
{"train_loss": 0.05845898762345314, "global_step": 206825, "epoch": 2323, "lr": 1.3391638863932821e-05}
{"train_loss": 0.03674572706222534, "global_step": 206826, "epoch": 2323, "lr": 1.3391243993345448e-05}
{"train_loss": 0.04993138089776039, "global_step": 206827, "epoch": 2323, "lr": 1.3390849127679695e-05}
{"train_loss": 0.01879575103521347, "global_step": 206828, "epoch": 2323, "lr": 1.3390454266935577e-05}
{"train_loss": 0.03185209631919861, "global_step": 206829, "epoch": 2323, "lr": 1.3390059411113182e-05}
{"train_loss": 0.07176652550697327, "global_step": 206830, "epoch": 2323, "lr": 1.3389664560212534e-05}
{"train_loss": 0.08473429083824158, "global_step": 206831, "epoch": 2323, "lr": 1.3389269714233715e-05}
{"train_loss": 0.04629917070269585, "global_step": 206832, "epoch": 2323, "lr": 1.3388874873176743e-05}
{"train_loss": 0.050048913806676865, "global_step": 206833, "epoch": 2323, "lr": 1.3388480037041705e-05}
{"train_loss": 0.06860259920358658, "global_step": 206834, "epoch": 2323, "lr": 1.3388085205828632e-05}
{"train_loss": 0.05607503857672884, "global_step": 206835, "epoch": 2323, "lr": 1.3387690379537593e-05, "val_loss": 8.507147789001465}
{"train_loss": 0.05836961045861244, "global_step": 206836, "epoch": 2324, "lr": 1.3387295558168621e-05}
{"train_loss": 0.04189946502447128, "global_step": 206837, "epoch": 2324, "lr": 1.3386900741721797e-05}
{"train_loss": 0.07701928168535233, "global_step": 206838, "epoch": 2324, "lr": 1.338650593019714e-05}
{"train_loss": 0.021724261343479156, "global_step": 206839, "epoch": 2324, "lr": 1.3386111123594735e-05}
{"train_loss": 0.10170646011829376, "global_step": 206840, "epoch": 2324, "lr": 1.3385716321914626e-05}
{"train_loss": 0.07342106103897095, "global_step": 206841, "epoch": 2324, "lr": 1.338532152515684e-05}
{"train_loss": 0.06199269741773605, "global_step": 206842, "epoch": 2324, "lr": 1.3384926733321473e-05}
{"train_loss": 0.0917167067527771, "global_step": 206843, "epoch": 2324, "lr": 1.3384531946408535e-05}
{"train_loss": 0.07467002421617508, "global_step": 206844, "epoch": 2324, "lr": 1.338413716441812e-05}
{"train_loss": 0.0737585723400116, "global_step": 206845, "epoch": 2324, "lr": 1.3383742387350246e-05}
{"train_loss": 0.05444587767124176, "global_step": 206846, "epoch": 2324, "lr": 1.3383347615204995e-05}
{"train_loss": 0.05694557726383209, "global_step": 206847, "epoch": 2324, "lr": 1.3382952847982389e-05}
{"train_loss": 0.0712873637676239, "global_step": 206848, "epoch": 2324, "lr": 1.3382558085682517e-05}
{"train_loss": 0.05431964248418808, "global_step": 206849, "epoch": 2324, "lr": 1.3382163328305402e-05}
{"train_loss": 0.07516452670097351, "global_step": 206850, "epoch": 2324, "lr": 1.3381768575851116e-05}
{"train_loss": 0.02645479142665863, "global_step": 206851, "epoch": 2324, "lr": 1.3381373828319699e-05}
{"train_loss": 0.06198740005493164, "global_step": 206852, "epoch": 2324, "lr": 1.338097908571122e-05}
{"train_loss": 0.07838450372219086, "global_step": 206853, "epoch": 2324, "lr": 1.3380584348025704e-05}
{"train_loss": 0.040328290313482285, "global_step": 206854, "epoch": 2324, "lr": 1.338018961526324e-05}
{"train_loss": 0.03492957726120949, "global_step": 206855, "epoch": 2324, "lr": 1.3379794887423853e-05}
{"train_loss": 0.028351884335279465, "global_step": 206856, "epoch": 2324, "lr": 1.3379400164507611e-05}
{"train_loss": 0.05298013240098953, "global_step": 206857, "epoch": 2324, "lr": 1.3379005446514554e-05}
{"train_loss": 0.07331393659114838, "global_step": 206858, "epoch": 2324, "lr": 1.3378610733444757e-05}
{"train_loss": 0.0569879375398159, "global_step": 206859, "epoch": 2324, "lr": 1.3378216025298245e-05}
{"train_loss": 0.021319514140486717, "global_step": 206860, "epoch": 2324, "lr": 1.3377821322075084e-05}
{"train_loss": 0.09028071910142899, "global_step": 206861, "epoch": 2324, "lr": 1.3377426623775347e-05}
{"train_loss": 0.036162350326776505, "global_step": 206862, "epoch": 2324, "lr": 1.3377031930399054e-05}
{"train_loss": 0.04000077024102211, "global_step": 206863, "epoch": 2324, "lr": 1.3376637241946282e-05}
{"train_loss": 0.04526733607053757, "global_step": 206864, "epoch": 2324, "lr": 1.3376242558417062e-05}
{"train_loss": 0.04357655718922615, "global_step": 206865, "epoch": 2324, "lr": 1.3375847879811476e-05}
{"train_loss": 0.055117860436439514, "global_step": 206866, "epoch": 2324, "lr": 1.3375453206129546e-05}
{"train_loss": 0.04042783007025719, "global_step": 206867, "epoch": 2324, "lr": 1.3375058537371354e-05}
{"train_loss": 0.051330920308828354, "global_step": 206868, "epoch": 2324, "lr": 1.3374663873536925e-05}
{"train_loss": 0.0891081765294075, "global_step": 206869, "epoch": 2324, "lr": 1.337426921462634e-05}
{"train_loss": 0.029184559360146523, "global_step": 206870, "epoch": 2324, "lr": 1.3373874560639627e-05}
{"train_loss": 0.031880881637334824, "global_step": 206871, "epoch": 2324, "lr": 1.3373479911576859e-05}
{"train_loss": 0.0945708155632019, "global_step": 206872, "epoch": 2324, "lr": 1.3373085267438068e-05}
{"train_loss": 0.04922680929303169, "global_step": 206873, "epoch": 2324, "lr": 1.3372690628223333e-05}
{"train_loss": 0.08231069892644882, "global_step": 206874, "epoch": 2324, "lr": 1.3372295993932682e-05}
{"train_loss": 0.08037085831165314, "global_step": 206875, "epoch": 2324, "lr": 1.3371901364566192e-05}
{"train_loss": 0.06061464920639992, "global_step": 206876, "epoch": 2324, "lr": 1.337150674012389e-05}
{"train_loss": 0.06516747176647186, "global_step": 206877, "epoch": 2324, "lr": 1.3371112120605856e-05}
{"train_loss": 0.07338248193264008, "global_step": 206878, "epoch": 2324, "lr": 1.337071750601211e-05}
{"train_loss": 0.04951338469982147, "global_step": 206879, "epoch": 2324, "lr": 1.3370322896342747e-05}
{"train_loss": 0.06453152745962143, "global_step": 206880, "epoch": 2324, "lr": 1.3369928291597794e-05}
{"train_loss": 0.05609520152211189, "global_step": 206881, "epoch": 2324, "lr": 1.3369533691777292e-05}
{"train_loss": 0.05576662719249725, "global_step": 206882, "epoch": 2324, "lr": 1.3369139096881328e-05}
{"train_loss": 0.0366155281662941, "global_step": 206883, "epoch": 2324, "lr": 1.336874450690992e-05}
{"train_loss": 0.04361552372574806, "global_step": 206884, "epoch": 2324, "lr": 1.3368349921863149e-05}
{"train_loss": 0.0911162868142128, "global_step": 206885, "epoch": 2324, "lr": 1.3367955341741045e-05}
{"train_loss": 0.06239272281527519, "global_step": 206886, "epoch": 2324, "lr": 1.336756076654369e-05}
{"train_loss": 0.057633206248283386, "global_step": 206887, "epoch": 2324, "lr": 1.3367166196271103e-05}
{"train_loss": 0.0345064215362072, "global_step": 206888, "epoch": 2324, "lr": 1.336677163092337e-05}
{"train_loss": 0.029862055554986, "global_step": 206889, "epoch": 2324, "lr": 1.3366377070500514e-05}
{"train_loss": 0.06894891709089279, "global_step": 206890, "epoch": 2324, "lr": 1.3365982515002618e-05}
{"train_loss": 0.05471552535891533, "global_step": 206891, "epoch": 2324, "lr": 1.3365587964429698e-05}
{"train_loss": 0.06259852647781372, "global_step": 206892, "epoch": 2324, "lr": 1.3365193418781847e-05}
{"train_loss": 0.04962710663676262, "global_step": 206893, "epoch": 2324, "lr": 1.3364798878059087e-05}
{"train_loss": 0.07189217209815979, "global_step": 206894, "epoch": 2324, "lr": 1.3364404342261494e-05}
{"train_loss": 0.1014305129647255, "global_step": 206895, "epoch": 2324, "lr": 1.3364009811389095e-05}
{"train_loss": 0.14286385476589203, "global_step": 206896, "epoch": 2324, "lr": 1.3363615285441961e-05}
{"train_loss": 0.05048090219497681, "global_step": 206897, "epoch": 2324, "lr": 1.3363220764420153e-05}
{"train_loss": 0.015002305619418621, "global_step": 206898, "epoch": 2324, "lr": 1.3362826248323707e-05}
{"train_loss": 0.06390684843063354, "global_step": 206899, "epoch": 2324, "lr": 1.3362431737152692e-05}
{"train_loss": 0.04375528171658516, "global_step": 206900, "epoch": 2324, "lr": 1.3362037230907137e-05}
{"train_loss": 0.03776394575834274, "global_step": 206901, "epoch": 2324, "lr": 1.336164272958712e-05}
{"train_loss": 0.016071509569883347, "global_step": 206902, "epoch": 2324, "lr": 1.3361248233192675e-05}
{"train_loss": 0.024691825732588768, "global_step": 206903, "epoch": 2324, "lr": 1.3360853741723878e-05}
{"train_loss": 0.025710148736834526, "global_step": 206904, "epoch": 2324, "lr": 1.3360459255180752e-05}
{"train_loss": 0.08267588913440704, "global_step": 206905, "epoch": 2324, "lr": 1.336006477356338e-05}
{"train_loss": 0.09375597536563873, "global_step": 206906, "epoch": 2324, "lr": 1.3359670296871785e-05}
{"train_loss": 0.03458334133028984, "global_step": 206907, "epoch": 2324, "lr": 1.3359275825106049e-05}
{"train_loss": 0.03599271923303604, "global_step": 206908, "epoch": 2324, "lr": 1.3358881358266202e-05}
{"train_loss": 0.09748265892267227, "global_step": 206909, "epoch": 2324, "lr": 1.3358486896352317e-05}
{"train_loss": 0.0912923663854599, "global_step": 206910, "epoch": 2324, "lr": 1.3358092439364423e-05}
{"train_loss": 0.06107188016176224, "global_step": 206911, "epoch": 2324, "lr": 1.3357697987302604e-05}
{"train_loss": 0.01973431557416916, "global_step": 206912, "epoch": 2324, "lr": 1.3357303540166876e-05}
{"train_loss": 0.08779396861791611, "global_step": 206913, "epoch": 2324, "lr": 1.3356909097957331e-05}
{"train_loss": 0.024635570123791695, "global_step": 206914, "epoch": 2324, "lr": 1.3356514660673991e-05}
{"train_loss": 0.0624123141169548, "global_step": 206915, "epoch": 2324, "lr": 1.3356120228316932e-05}
{"train_loss": 0.050391390919685364, "global_step": 206916, "epoch": 2324, "lr": 1.3355725800886182e-05}
{"train_loss": 0.0662812888622284, "global_step": 206917, "epoch": 2324, "lr": 1.3355331378381819e-05}
{"train_loss": 0.05750579014420509, "global_step": 206918, "epoch": 2324, "lr": 1.3354936960803894e-05}
{"train_loss": 0.04657447338104248, "global_step": 206919, "epoch": 2324, "lr": 1.3354542548152427e-05}
{"train_loss": 0.08621607720851898, "global_step": 206920, "epoch": 2324, "lr": 1.3354148140427513e-05}
{"train_loss": 0.050436101853847504, "global_step": 206921, "epoch": 2324, "lr": 1.3353753737629176e-05}
{"train_loss": 0.07931835204362869, "global_step": 206922, "epoch": 2324, "lr": 1.3353359339757488e-05}
{"train_loss": 0.05705566331744194, "global_step": 206923, "epoch": 2324, "lr": 1.3352964946812485e-05}
{"train_loss": 0.057702755000902696, "global_step": 206924, "epoch": 2324, "lr": 1.3352570558794242e-05, "val_loss": 8.493194580078125}
{"train_loss": 0.03774259611964226, "global_step": 206925, "epoch": 2325, "lr": 1.3352176175702785e-05}
{"train_loss": 0.042139116674661636, "global_step": 206926, "epoch": 2325, "lr": 1.3351781797538193e-05}
{"train_loss": 0.052308108657598495, "global_step": 206927, "epoch": 2325, "lr": 1.33513874243005e-05}
{"train_loss": 0.041262831538915634, "global_step": 206928, "epoch": 2325, "lr": 1.3350993055989775e-05}
{"train_loss": 0.09377073496580124, "global_step": 206929, "epoch": 2325, "lr": 1.3350598692606047e-05}
{"train_loss": 0.06309781968593597, "global_step": 206930, "epoch": 2325, "lr": 1.3350204334149403e-05}
{"train_loss": 0.0740785300731659, "global_step": 206931, "epoch": 2325, "lr": 1.334980998061986e-05}
{"train_loss": 0.026423487812280655, "global_step": 206932, "epoch": 2325, "lr": 1.3349415632017487e-05}
{"train_loss": 0.07914137095212936, "global_step": 206933, "epoch": 2325, "lr": 1.3349021288342356e-05}
{"train_loss": 0.02026107907295227, "global_step": 206934, "epoch": 2325, "lr": 1.3348626949594484e-05}
{"train_loss": 0.048818644136190414, "global_step": 206935, "epoch": 2325, "lr": 1.3348232615773965e-05}
{"train_loss": 0.0715753510594368, "global_step": 206936, "epoch": 2325, "lr": 1.3347838286880805e-05}
{"train_loss": 0.0988113060593605, "global_step": 206937, "epoch": 2325, "lr": 1.33474439629151e-05}
{"train_loss": 0.03711798042058945, "global_step": 206938, "epoch": 2325, "lr": 1.3347049643876869e-05}
{"train_loss": 0.02967747300863266, "global_step": 206939, "epoch": 2325, "lr": 1.33466553297662e-05}
{"train_loss": 0.0698191449046135, "global_step": 206940, "epoch": 2325, "lr": 1.3346261020583107e-05}
{"train_loss": 0.041130997240543365, "global_step": 206941, "epoch": 2325, "lr": 1.3345866716327677e-05}
{"train_loss": 0.08471344411373138, "global_step": 206942, "epoch": 2325, "lr": 1.3345472416999937e-05}
{"train_loss": 0.02849331870675087, "global_step": 206943, "epoch": 2325, "lr": 1.3345078122599964e-05}
{"train_loss": 0.04540598392486572, "global_step": 206944, "epoch": 2325, "lr": 1.3344683833127786e-05}
{"train_loss": 0.028630517423152924, "global_step": 206945, "epoch": 2325, "lr": 1.3344289548583477e-05}
{"train_loss": 0.03916037827730179, "global_step": 206946, "epoch": 2325, "lr": 1.3343895268967072e-05}
{"train_loss": 0.035501670092344284, "global_step": 206947, "epoch": 2325, "lr": 1.3343500994278646e-05}
{"train_loss": 0.06087109446525574, "global_step": 206948, "epoch": 2325, "lr": 1.3343106724518228e-05}
{"train_loss": 0.06917479634284973, "global_step": 206949, "epoch": 2325, "lr": 1.3342712459685897e-05}
{"train_loss": 0.03431469202041626, "global_step": 206950, "epoch": 2325, "lr": 1.3342318199781672e-05}
{"train_loss": 0.0754893571138382, "global_step": 206951, "epoch": 2325, "lr": 1.3341923944805645e-05}
{"train_loss": 0.05634284019470215, "global_step": 206952, "epoch": 2325, "lr": 1.3341529694757831e-05}
{"train_loss": 0.06434284150600433, "global_step": 206953, "epoch": 2325, "lr": 1.334113544963832e-05}
{"train_loss": 0.087580107152462, "global_step": 206954, "epoch": 2325, "lr": 1.3340741209447127e-05}
{"train_loss": 0.06706966459751129, "global_step": 206955, "epoch": 2325, "lr": 1.3340346974184342e-05}
{"train_loss": 0.05015577748417854, "global_step": 206956, "epoch": 2325, "lr": 1.3339952743850004e-05}
{"train_loss": 0.07113075256347656, "global_step": 206957, "epoch": 2325, "lr": 1.3339558518444146e-05}
{"train_loss": 0.05979933589696884, "global_step": 206958, "epoch": 2325, "lr": 1.3339164297966849e-05}
{"train_loss": 0.03035612218081951, "global_step": 206959, "epoch": 2325, "lr": 1.333877008241814e-05}
{"train_loss": 0.05903839319944382, "global_step": 206960, "epoch": 2325, "lr": 1.3338375871798103e-05}
{"train_loss": 0.044799529016017914, "global_step": 206961, "epoch": 2325, "lr": 1.3337981666106764e-05}
{"train_loss": 0.05026670917868614, "global_step": 206962, "epoch": 2325, "lr": 1.3337587465344192e-05}
{"train_loss": 0.04018079489469528, "global_step": 206963, "epoch": 2325, "lr": 1.3337193269510428e-05}
{"train_loss": 0.047331780195236206, "global_step": 206964, "epoch": 2325, "lr": 1.3336799078605545e-05}
{"train_loss": 0.08264246582984924, "global_step": 206965, "epoch": 2325, "lr": 1.3336404892629562e-05}
{"train_loss": 0.06668412685394287, "global_step": 206966, "epoch": 2325, "lr": 1.333601071158257e-05}
{"train_loss": 0.06879410147666931, "global_step": 206967, "epoch": 2325, "lr": 1.333561653546459e-05}
{"train_loss": 0.07493484020233154, "global_step": 206968, "epoch": 2325, "lr": 1.3335222364275707e-05}
{"train_loss": 0.04806901887059212, "global_step": 206969, "epoch": 2325, "lr": 1.3334828198015942e-05}
{"train_loss": 0.040200814604759216, "global_step": 206970, "epoch": 2325, "lr": 1.3334434036685356e-05}
{"train_loss": 0.08847516030073166, "global_step": 206971, "epoch": 2325, "lr": 1.3334039880284033e-05}
{"train_loss": 0.08609670400619507, "global_step": 206972, "epoch": 2325, "lr": 1.3333645728811978e-05}
{"train_loss": 0.0628710389137268, "global_step": 206973, "epoch": 2325, "lr": 1.3333251582269285e-05}
{"train_loss": 0.07772107422351837, "global_step": 206974, "epoch": 2325, "lr": 1.3332857440655977e-05}
{"train_loss": 0.03221501410007477, "global_step": 206975, "epoch": 2325, "lr": 1.333246330397213e-05}
{"train_loss": 0.07100985944271088, "global_step": 206976, "epoch": 2325, "lr": 1.3332069172217776e-05}
{"train_loss": 0.08793844282627106, "global_step": 206977, "epoch": 2325, "lr": 1.3331675045392988e-05}
{"train_loss": 0.02272997796535492, "global_step": 206978, "epoch": 2325, "lr": 1.3331280923497797e-05}
{"train_loss": 0.09737411141395569, "global_step": 206979, "epoch": 2325, "lr": 1.3330886806532283e-05}
{"train_loss": 0.03510388731956482, "global_step": 206980, "epoch": 2325, "lr": 1.3330492694496471e-05}
{"train_loss": 0.02427142858505249, "global_step": 206981, "epoch": 2325, "lr": 1.3330098587390445e-05}
{"train_loss": 0.0641760304570198, "global_step": 206982, "epoch": 2325, "lr": 1.3329704485214218e-05}
{"train_loss": 0.03613560274243355, "global_step": 206983, "epoch": 2325, "lr": 1.3329310387967885e-05}
{"train_loss": 0.0586002953350544, "global_step": 206984, "epoch": 2325, "lr": 1.332891629565146e-05}
{"train_loss": 0.043297816067934036, "global_step": 206985, "epoch": 2325, "lr": 1.3328522208265037e-05}
{"train_loss": 0.047501519322395325, "global_step": 206986, "epoch": 2325, "lr": 1.3328128125808632e-05}
{"train_loss": 0.07756590843200684, "global_step": 206987, "epoch": 2325, "lr": 1.3327734048282309e-05}
{"train_loss": 0.03845541179180145, "global_step": 206988, "epoch": 2325, "lr": 1.3327339975686137e-05}
{"train_loss": 0.05164249613881111, "global_step": 206989, "epoch": 2325, "lr": 1.332694590802015e-05}
{"train_loss": 0.03988104313611984, "global_step": 206990, "epoch": 2325, "lr": 1.3326551845284419e-05}
{"train_loss": 0.025990957394242287, "global_step": 206991, "epoch": 2325, "lr": 1.3326157787478983e-05}
{"train_loss": 0.0608355738222599, "global_step": 206992, "epoch": 2325, "lr": 1.3325763734603885e-05}
{"train_loss": 0.054958269000053406, "global_step": 206993, "epoch": 2325, "lr": 1.3325369686659212e-05}
{"train_loss": 0.028587888926267624, "global_step": 206994, "epoch": 2325, "lr": 1.3324975643644977e-05}
{"train_loss": 0.038042452186346054, "global_step": 206995, "epoch": 2325, "lr": 1.3324581605561265e-05}
{"train_loss": 0.06920217722654343, "global_step": 206996, "epoch": 2325, "lr": 1.332418757240812e-05}
{"train_loss": 0.09112837165594101, "global_step": 206997, "epoch": 2325, "lr": 1.332379354418557e-05}
{"train_loss": 0.05556780472397804, "global_step": 206998, "epoch": 2325, "lr": 1.332339952089371e-05}
{"train_loss": 0.045896194875240326, "global_step": 206999, "epoch": 2325, "lr": 1.3323005502532553e-05}
{"train_loss": 0.06476130336523056, "global_step": 207000, "epoch": 2325, "lr": 1.3322611489102188e-05}
{"train_loss": 0.08532722294330597, "global_step": 207001, "epoch": 2325, "lr": 1.3322217480602633e-05}
{"train_loss": 0.049092210829257965, "global_step": 207002, "epoch": 2325, "lr": 1.3321823477033978e-05}
{"train_loss": 0.03299442306160927, "global_step": 207003, "epoch": 2325, "lr": 1.3321429478396242e-05}
{"train_loss": 0.10430461168289185, "global_step": 207004, "epoch": 2325, "lr": 1.3321035484689504e-05}
{"train_loss": 0.08435613662004471, "global_step": 207005, "epoch": 2325, "lr": 1.3320641495913794e-05}
{"train_loss": 0.0818195641040802, "global_step": 207006, "epoch": 2325, "lr": 1.3320247512069172e-05}
{"train_loss": 0.08109643310308456, "global_step": 207007, "epoch": 2325, "lr": 1.3319853533155713e-05}
{"train_loss": 0.0409262515604496, "global_step": 207008, "epoch": 2325, "lr": 1.3319459559173441e-05}
{"train_loss": 0.033980704843997955, "global_step": 207009, "epoch": 2325, "lr": 1.3319065590122432e-05}
{"train_loss": 0.1072002723813057, "global_step": 207010, "epoch": 2325, "lr": 1.3318671626002716e-05}
{"train_loss": 0.07866235822439194, "global_step": 207011, "epoch": 2325, "lr": 1.3318277666814366e-05}
{"train_loss": 0.04953572526574135, "global_step": 207012, "epoch": 2325, "lr": 1.3317883712557416e-05}
{"train_loss": 0.057701217157117436, "global_step": 207013, "epoch": 2325, "lr": 1.3317489763231943e-05, "val_loss": 8.531047821044922, "train_action_mse_error": 20.073780059814453}
{"train_loss": 0.06315886229276657, "global_step": 207014, "epoch": 2326, "lr": 1.3317095818837976e-05}
{"train_loss": 0.03651891648769379, "global_step": 207015, "epoch": 2326, "lr": 1.3316701879375588e-05}
{"train_loss": 0.04537999629974365, "global_step": 207016, "epoch": 2326, "lr": 1.3316307944844808e-05}
{"train_loss": 0.04773163050413132, "global_step": 207017, "epoch": 2326, "lr": 1.3315914015245723e-05}
{"train_loss": 0.02692892774939537, "global_step": 207018, "epoch": 2326, "lr": 1.3315520090578344e-05}
{"train_loss": 0.049368519335985184, "global_step": 207019, "epoch": 2326, "lr": 1.3315126170842768e-05}
{"train_loss": 0.05381612852215767, "global_step": 207020, "epoch": 2326, "lr": 1.3314732256039002e-05}
{"train_loss": 0.0824730172753334, "global_step": 207021, "epoch": 2326, "lr": 1.3314338346167149e-05}
{"train_loss": 0.011998334899544716, "global_step": 207022, "epoch": 2326, "lr": 1.3313944441227216e-05}
{"train_loss": 0.11710865050554276, "global_step": 207023, "epoch": 2326, "lr": 1.3313550541219278e-05}
{"train_loss": 0.048111043870449066, "global_step": 207024, "epoch": 2326, "lr": 1.3313156646143398e-05}
{"train_loss": 0.02853277139365673, "global_step": 207025, "epoch": 2326, "lr": 1.3312762755999603e-05}
{"train_loss": 0.13817694783210754, "global_step": 207026, "epoch": 2326, "lr": 1.3312368870787972e-05}
{"train_loss": 0.08419366180896759, "global_step": 207027, "epoch": 2326, "lr": 1.3311974990508536e-05}
{"train_loss": 0.030536523088812828, "global_step": 207028, "epoch": 2326, "lr": 1.331158111516137e-05}
{"train_loss": 0.06647266447544098, "global_step": 207029, "epoch": 2326, "lr": 1.3311187244746514e-05}
{"train_loss": 0.08220706135034561, "global_step": 207030, "epoch": 2326, "lr": 1.3310793379264009e-05}
{"train_loss": 0.12341703474521637, "global_step": 207031, "epoch": 2326, "lr": 1.3310399518713933e-05}
{"train_loss": 0.08019965142011642, "global_step": 207032, "epoch": 2326, "lr": 1.3310005663096314e-05}
{"train_loss": 0.07851595431566238, "global_step": 207033, "epoch": 2326, "lr": 1.3309611812411232e-05}
{"train_loss": 0.07371673732995987, "global_step": 207034, "epoch": 2326, "lr": 1.3309217966658727e-05}
{"train_loss": 0.06196998059749603, "global_step": 207035, "epoch": 2326, "lr": 1.330882412583883e-05}
{"train_loss": 0.09296026825904846, "global_step": 207036, "epoch": 2326, "lr": 1.3308430289951634e-05}
{"train_loss": 0.06943678110837936, "global_step": 207037, "epoch": 2326, "lr": 1.3308036458997158e-05}
{"train_loss": 0.04894939810037613, "global_step": 207038, "epoch": 2326, "lr": 1.3307642632975481e-05}
{"train_loss": 0.057737305760383606, "global_step": 207039, "epoch": 2326, "lr": 1.3307248811886636e-05}
{"train_loss": 0.06530400365591049, "global_step": 207040, "epoch": 2326, "lr": 1.3306854995730695e-05}
{"train_loss": 0.018276600167155266, "global_step": 207041, "epoch": 2326, "lr": 1.330646118450768e-05}
{"train_loss": 0.026582062244415283, "global_step": 207042, "epoch": 2326, "lr": 1.3306067378217668e-05}
{"train_loss": 0.05180193856358528, "global_step": 207043, "epoch": 2326, "lr": 1.3305673576860723e-05}
{"train_loss": 0.07540464401245117, "global_step": 207044, "epoch": 2326, "lr": 1.330527978043687e-05}
{"train_loss": 0.024504590779542923, "global_step": 207045, "epoch": 2326, "lr": 1.3304885988946186e-05}
{"train_loss": 0.05717198923230171, "global_step": 207046, "epoch": 2326, "lr": 1.3304492202388701e-05}
{"train_loss": 0.07291677594184875, "global_step": 207047, "epoch": 2326, "lr": 1.3304098420764494e-05}
{"train_loss": 0.030439460650086403, "global_step": 207048, "epoch": 2326, "lr": 1.3303704644073589e-05}
{"train_loss": 0.035138245671987534, "global_step": 207049, "epoch": 2326, "lr": 1.3303310872316067e-05}
{"train_loss": 0.06334031373262405, "global_step": 207050, "epoch": 2326, "lr": 1.3302917105491957e-05}
{"train_loss": 0.06923587620258331, "global_step": 207051, "epoch": 2326, "lr": 1.330252334360133e-05}
{"train_loss": 0.1035526841878891, "global_step": 207052, "epoch": 2326, "lr": 1.3302129586644225e-05}
{"train_loss": 0.04676821827888489, "global_step": 207053, "epoch": 2326, "lr": 1.330173583462071e-05}
{"train_loss": 0.057380445301532745, "global_step": 207054, "epoch": 2326, "lr": 1.3301342087530815e-05}
{"train_loss": 0.018191207200288773, "global_step": 207055, "epoch": 2326, "lr": 1.3300948345374626e-05}
{"train_loss": 0.06919027119874954, "global_step": 207056, "epoch": 2326, "lr": 1.330055460815216e-05}
{"train_loss": 0.07207119464874268, "global_step": 207057, "epoch": 2326, "lr": 1.3300160875863505e-05}
{"train_loss": 0.05677183344960213, "global_step": 207058, "epoch": 2326, "lr": 1.3299767148508675e-05}
{"train_loss": 0.08443363755941391, "global_step": 207059, "epoch": 2326, "lr": 1.3299373426087747e-05}
{"train_loss": 0.06654134392738342, "global_step": 207060, "epoch": 2326, "lr": 1.3298979708600784e-05}
{"train_loss": 0.054751574993133545, "global_step": 207061, "epoch": 2326, "lr": 1.3298585996047814e-05}
{"train_loss": 0.05455905199050903, "global_step": 207062, "epoch": 2326, "lr": 1.3298192288428917e-05}
{"train_loss": 0.10727599263191223, "global_step": 207063, "epoch": 2326, "lr": 1.3297798585744114e-05}
{"train_loss": 0.08217332512140274, "global_step": 207064, "epoch": 2326, "lr": 1.3297404887993492e-05}
{"train_loss": 0.014432162046432495, "global_step": 207065, "epoch": 2326, "lr": 1.3297011195177067e-05}
{"train_loss": 0.02502618357539177, "global_step": 207066, "epoch": 2326, "lr": 1.329661750729493e-05}
{"train_loss": 0.05731457471847534, "global_step": 207067, "epoch": 2326, "lr": 1.3296223824347116e-05}
{"train_loss": 0.030619777739048004, "global_step": 207068, "epoch": 2326, "lr": 1.329583014633366e-05}
{"train_loss": 0.03805461898446083, "global_step": 207069, "epoch": 2326, "lr": 1.3295436473254646e-05}
{"train_loss": 0.07542569935321808, "global_step": 207070, "epoch": 2326, "lr": 1.3295042805110102e-05}
{"train_loss": 0.06706642359495163, "global_step": 207071, "epoch": 2326, "lr": 1.3294649141900106e-05}
{"train_loss": 0.09435661882162094, "global_step": 207072, "epoch": 2326, "lr": 1.3294255483624696e-05}
{"train_loss": 0.05601155757904053, "global_step": 207073, "epoch": 2326, "lr": 1.329386183028391e-05}
{"train_loss": 0.04064080864191055, "global_step": 207074, "epoch": 2326, "lr": 1.3293468181877833e-05}
{"train_loss": 0.060503970831632614, "global_step": 207075, "epoch": 2326, "lr": 1.3293074538406486e-05}
{"train_loss": 0.09883929044008255, "global_step": 207076, "epoch": 2326, "lr": 1.3292680899869959e-05}
{"train_loss": 0.09109524637460709, "global_step": 207077, "epoch": 2326, "lr": 1.329228726626826e-05}
{"train_loss": 0.06321628391742706, "global_step": 207078, "epoch": 2326, "lr": 1.329189363760147e-05}
{"train_loss": 0.048552487045526505, "global_step": 207079, "epoch": 2326, "lr": 1.3291500013869651e-05}
{"train_loss": 0.06618665903806686, "global_step": 207080, "epoch": 2326, "lr": 1.3291106395072828e-05}
{"train_loss": 0.046629399061203, "global_step": 207081, "epoch": 2326, "lr": 1.3290712781211085e-05}
{"train_loss": 0.13571389019489288, "global_step": 207082, "epoch": 2326, "lr": 1.3290319172284444e-05}
{"train_loss": 0.06220407411456108, "global_step": 207083, "epoch": 2326, "lr": 1.3289925568292983e-05}
{"train_loss": 0.09087514132261276, "global_step": 207084, "epoch": 2326, "lr": 1.3289531969236729e-05}
{"train_loss": 0.0170125812292099, "global_step": 207085, "epoch": 2326, "lr": 1.3289138375115768e-05}
{"train_loss": 0.10965794324874878, "global_step": 207086, "epoch": 2326, "lr": 1.3288744785930119e-05}
{"train_loss": 0.05939444527029991, "global_step": 207087, "epoch": 2326, "lr": 1.3288351201679867e-05}
{"train_loss": 0.041449472308158875, "global_step": 207088, "epoch": 2326, "lr": 1.3287957622365027e-05}
{"train_loss": 0.042898066341876984, "global_step": 207089, "epoch": 2326, "lr": 1.3287564047985695e-05}
{"train_loss": 0.07534782588481903, "global_step": 207090, "epoch": 2326, "lr": 1.3287170478541889e-05}
{"train_loss": 0.06894002109766006, "global_step": 207091, "epoch": 2326, "lr": 1.3286776914033688e-05}
{"train_loss": 0.03683158755302429, "global_step": 207092, "epoch": 2326, "lr": 1.3286383354461119e-05}
{"train_loss": 0.04270306974649429, "global_step": 207093, "epoch": 2326, "lr": 1.3285989799824262e-05}
{"train_loss": 0.06335371732711792, "global_step": 207094, "epoch": 2326, "lr": 1.3285596250123138e-05}
{"train_loss": 0.07200875133275986, "global_step": 207095, "epoch": 2326, "lr": 1.3285202705357835e-05}
{"train_loss": 0.021490884944796562, "global_step": 207096, "epoch": 2326, "lr": 1.3284809165528378e-05}
{"train_loss": 0.13374154269695282, "global_step": 207097, "epoch": 2326, "lr": 1.3284415630634827e-05}
{"train_loss": 0.02482130192220211, "global_step": 207098, "epoch": 2326, "lr": 1.328402210067725e-05}
{"train_loss": 0.10896237194538116, "global_step": 207099, "epoch": 2326, "lr": 1.3283628575655682e-05}
{"train_loss": 0.051306042820215225, "global_step": 207100, "epoch": 2326, "lr": 1.3283235055570192e-05}
{"train_loss": 0.04834631830453873, "global_step": 207101, "epoch": 2326, "lr": 1.328284154042081e-05}
{"train_loss": 0.06219656934982605, "global_step": 207102, "epoch": 2326, "lr": 1.3282448030207617e-05, "val_loss": 8.612311363220215}
{"train_loss": 0.04571538791060448, "global_step": 207103, "epoch": 2327, "lr": 1.3282054524930638e-05}
{"train_loss": 0.06234139949083328, "global_step": 207104, "epoch": 2327, "lr": 1.3281661024589953e-05}
{"train_loss": 0.054043710231781006, "global_step": 207105, "epoch": 2327, "lr": 1.3281267529185586e-05}
{"train_loss": 0.06584562361240387, "global_step": 207106, "epoch": 2327, "lr": 1.3280874038717623e-05}
{"train_loss": 0.056898050010204315, "global_step": 207107, "epoch": 2327, "lr": 1.3280480553186098e-05}
{"train_loss": 0.06101250648498535, "global_step": 207108, "epoch": 2327, "lr": 1.3280087072591047e-05}
{"train_loss": 0.04482249170541763, "global_step": 207109, "epoch": 2327, "lr": 1.3279693596932558e-05}
{"train_loss": 0.04526114463806152, "global_step": 207110, "epoch": 2327, "lr": 1.3279300126210654e-05}
{"train_loss": 0.036389365792274475, "global_step": 207111, "epoch": 2327, "lr": 1.3278906660425411e-05}
{"train_loss": 0.04164140298962593, "global_step": 207112, "epoch": 2327, "lr": 1.327851319957687e-05}
{"train_loss": 0.06690127402544022, "global_step": 207113, "epoch": 2327, "lr": 1.3278119743665074e-05}
{"train_loss": 0.0768309012055397, "global_step": 207114, "epoch": 2327, "lr": 1.3277726292690085e-05}
{"train_loss": 0.08735185116529465, "global_step": 207115, "epoch": 2327, "lr": 1.3277332846651974e-05}
{"train_loss": 0.03958350419998169, "global_step": 207116, "epoch": 2327, "lr": 1.3276939405550764e-05}
{"train_loss": 0.07482553273439407, "global_step": 207117, "epoch": 2327, "lr": 1.3276545969386538e-05}
{"train_loss": 0.0903795063495636, "global_step": 207118, "epoch": 2327, "lr": 1.3276152538159314e-05}
{"train_loss": 0.032912805676460266, "global_step": 207119, "epoch": 2327, "lr": 1.3275759111869185e-05}
{"train_loss": 0.09662944078445435, "global_step": 207120, "epoch": 2327, "lr": 1.3275365690516162e-05}
{"train_loss": 0.11808627843856812, "global_step": 207121, "epoch": 2327, "lr": 1.3274972274100334e-05}
{"train_loss": 0.11120342463254929, "global_step": 207122, "epoch": 2327, "lr": 1.3274578862621723e-05}
{"train_loss": 0.03283904492855072, "global_step": 207123, "epoch": 2327, "lr": 1.3274185456080418e-05}
{"train_loss": 0.03611082583665848, "global_step": 207124, "epoch": 2327, "lr": 1.327379205447643e-05}
{"train_loss": 0.12417751550674438, "global_step": 207125, "epoch": 2327, "lr": 1.327339865780985e-05}
{"train_loss": 0.05034605413675308, "global_step": 207126, "epoch": 2327, "lr": 1.3273005266080696e-05}
{"train_loss": 0.04484059289097786, "global_step": 207127, "epoch": 2327, "lr": 1.327261187928906e-05}
{"train_loss": 0.07321557402610779, "global_step": 207128, "epoch": 2327, "lr": 1.3272218497434953e-05}
{"train_loss": 0.03949225693941116, "global_step": 207129, "epoch": 2327, "lr": 1.3271825120518466e-05}
{"train_loss": 0.09870108962059021, "global_step": 207130, "epoch": 2327, "lr": 1.327143174853962e-05}
{"train_loss": 0.04459668695926666, "global_step": 207131, "epoch": 2327, "lr": 1.3271038381498496e-05}
{"train_loss": 0.06523653864860535, "global_step": 207132, "epoch": 2327, "lr": 1.3270645019395118e-05}
{"train_loss": 0.09681106358766556, "global_step": 207133, "epoch": 2327, "lr": 1.3270251662229555e-05}
{"train_loss": 0.042627718299627304, "global_step": 207134, "epoch": 2327, "lr": 1.3269858310001877e-05}
{"train_loss": 0.06875845044851303, "global_step": 207135, "epoch": 2327, "lr": 1.3269464962712097e-05}
{"train_loss": 0.06406545639038086, "global_step": 207136, "epoch": 2327, "lr": 1.3269071620360307e-05}
{"train_loss": 0.04162309691309929, "global_step": 207137, "epoch": 2327, "lr": 1.3268678282946533e-05}
{"train_loss": 0.039764486253261566, "global_step": 207138, "epoch": 2327, "lr": 1.326828495047085e-05}
{"train_loss": 0.05482413247227669, "global_step": 207139, "epoch": 2327, "lr": 1.3267891622933282e-05}
{"train_loss": 0.03695642203092575, "global_step": 207140, "epoch": 2327, "lr": 1.3267498300333908e-05}
{"train_loss": 0.05936911702156067, "global_step": 207141, "epoch": 2327, "lr": 1.3267104982672763e-05}
{"train_loss": 0.055335838347673416, "global_step": 207142, "epoch": 2327, "lr": 1.3266711669949921e-05}
{"train_loss": 0.06439433991909027, "global_step": 207143, "epoch": 2327, "lr": 1.326631836216541e-05}
{"train_loss": 0.09007787704467773, "global_step": 207144, "epoch": 2327, "lr": 1.3265925059319306e-05}
{"train_loss": 0.04500662162899971, "global_step": 207145, "epoch": 2327, "lr": 1.3265531761411654e-05}
{"train_loss": 0.13089041411876678, "global_step": 207146, "epoch": 2327, "lr": 1.3265138468442485e-05}
{"train_loss": 0.06458910554647446, "global_step": 207147, "epoch": 2327, "lr": 1.3264745180411885e-05}
{"train_loss": 0.08609113842248917, "global_step": 207148, "epoch": 2327, "lr": 1.3264351897319883e-05}
{"train_loss": 0.1327984631061554, "global_step": 207149, "epoch": 2327, "lr": 1.326395861916655e-05}
{"train_loss": 0.10124550014734268, "global_step": 207150, "epoch": 2327, "lr": 1.3263565345951917e-05}
{"train_loss": 0.08387815952301025, "global_step": 207151, "epoch": 2327, "lr": 1.326317207767606e-05}
{"train_loss": 0.07184102386236191, "global_step": 207152, "epoch": 2327, "lr": 1.3262778814339016e-05}
{"train_loss": 0.06142277643084526, "global_step": 207153, "epoch": 2327, "lr": 1.3262385555940853e-05}
{"train_loss": 0.048069607466459274, "global_step": 207154, "epoch": 2327, "lr": 1.3261992302481602e-05}
{"train_loss": 0.02407979965209961, "global_step": 207155, "epoch": 2327, "lr": 1.3261599053961343e-05}
{"train_loss": 0.06242407485842705, "global_step": 207156, "epoch": 2327, "lr": 1.3261205810380095e-05}
{"train_loss": 0.05503719672560692, "global_step": 207157, "epoch": 2327, "lr": 1.3260812571737952e-05}
{"train_loss": 0.0797118991613388, "global_step": 207158, "epoch": 2327, "lr": 1.3260419338034925e-05}
{"train_loss": 0.1059534102678299, "global_step": 207159, "epoch": 2327, "lr": 1.3260026109271107e-05}
{"train_loss": 0.03075314499437809, "global_step": 207160, "epoch": 2327, "lr": 1.325963288544651e-05}
{"train_loss": 0.04711099714040756, "global_step": 207161, "epoch": 2327, "lr": 1.3259239666561224e-05}
{"train_loss": 0.06217115744948387, "global_step": 207162, "epoch": 2327, "lr": 1.325884645261527e-05}
{"train_loss": 0.02486741542816162, "global_step": 207163, "epoch": 2327, "lr": 1.3258453243608732e-05}
{"train_loss": 0.09363734722137451, "global_step": 207164, "epoch": 2327, "lr": 1.325806003954163e-05}
{"train_loss": 0.04151010885834694, "global_step": 207165, "epoch": 2327, "lr": 1.3257666840414056e-05}
{"train_loss": 0.11176352202892303, "global_step": 207166, "epoch": 2327, "lr": 1.325727364622602e-05}
{"train_loss": 0.11071263998746872, "global_step": 207167, "epoch": 2327, "lr": 1.3256880456977611e-05}
{"train_loss": 0.033294834196567535, "global_step": 207168, "epoch": 2327, "lr": 1.3256487272668854e-05}
{"train_loss": 0.061753470450639725, "global_step": 207169, "epoch": 2327, "lr": 1.3256094093299814e-05}
{"train_loss": 0.0620199516415596, "global_step": 207170, "epoch": 2327, "lr": 1.3255700918870556e-05}
{"train_loss": 0.06347067654132843, "global_step": 207171, "epoch": 2327, "lr": 1.3255307749381108e-05}
{"train_loss": 0.0717545822262764, "global_step": 207172, "epoch": 2327, "lr": 1.3254914584831551e-05}
{"train_loss": 0.056364286690950394, "global_step": 207173, "epoch": 2327, "lr": 1.325452142522191e-05}
{"train_loss": 0.07162393629550934, "global_step": 207174, "epoch": 2327, "lr": 1.3254128270552263e-05}
{"train_loss": 0.06849046796560287, "global_step": 207175, "epoch": 2327, "lr": 1.3253735120822642e-05}
{"train_loss": 0.12132599204778671, "global_step": 207176, "epoch": 2327, "lr": 1.3253341976033119e-05}
{"train_loss": 0.04228649660944939, "global_step": 207177, "epoch": 2327, "lr": 1.3252948836183715e-05}
{"train_loss": 0.11448553949594498, "global_step": 207178, "epoch": 2327, "lr": 1.3252555701274528e-05}
{"train_loss": 0.05736488103866577, "global_step": 207179, "epoch": 2327, "lr": 1.3252162571305565e-05}
{"train_loss": 0.060796406120061874, "global_step": 207180, "epoch": 2327, "lr": 1.3251769446276924e-05}
{"train_loss": 0.0987541675567627, "global_step": 207181, "epoch": 2327, "lr": 1.3251376326188615e-05}
{"train_loss": 0.0180529672652483, "global_step": 207182, "epoch": 2327, "lr": 1.3250983211040725e-05}
{"train_loss": 0.06198813393712044, "global_step": 207183, "epoch": 2327, "lr": 1.3250590100833293e-05}
{"train_loss": 0.04129108041524887, "global_step": 207184, "epoch": 2327, "lr": 1.3250196995566356e-05}
{"train_loss": 0.04272445663809776, "global_step": 207185, "epoch": 2327, "lr": 1.3249803895239997e-05}
{"train_loss": 0.043975330889225006, "global_step": 207186, "epoch": 2327, "lr": 1.324941079985424e-05}
{"train_loss": 0.0616585910320282, "global_step": 207187, "epoch": 2327, "lr": 1.3249017709409168e-05}
{"train_loss": 0.061677344143390656, "global_step": 207188, "epoch": 2327, "lr": 1.3248624623904799e-05}
{"train_loss": 0.04713869467377663, "global_step": 207189, "epoch": 2327, "lr": 1.3248231543341222e-05}
{"train_loss": 0.07030216604471207, "global_step": 207190, "epoch": 2327, "lr": 1.3247838467718454e-05}
{"train_loss": 0.06604179170694244, "global_step": 207191, "epoch": 2327, "lr": 1.3247445397036585e-05, "val_loss": 8.473398208618164}
{"train_loss": 0.05285324156284332, "global_step": 207192, "epoch": 2328, "lr": 1.3247052331295634e-05}
{"train_loss": 0.03782503679394722, "global_step": 207193, "epoch": 2328, "lr": 1.3246659270495682e-05}
{"train_loss": 0.04244649410247803, "global_step": 207194, "epoch": 2328, "lr": 1.3246266214636754e-05}
{"train_loss": 0.03323948383331299, "global_step": 207195, "epoch": 2328, "lr": 1.324587316371893e-05}
{"train_loss": 0.06872399151325226, "global_step": 207196, "epoch": 2328, "lr": 1.3245480117742243e-05}
{"train_loss": 0.06819470971822739, "global_step": 207197, "epoch": 2328, "lr": 1.3245087076706764e-05}
{"train_loss": 0.051118817180395126, "global_step": 207198, "epoch": 2328, "lr": 1.3244694040612515e-05}
{"train_loss": 0.03701281547546387, "global_step": 207199, "epoch": 2328, "lr": 1.3244301009459592e-05}
{"train_loss": 0.1362868994474411, "global_step": 207200, "epoch": 2328, "lr": 1.3243907983248005e-05}
{"train_loss": 0.048387687653303146, "global_step": 207201, "epoch": 2328, "lr": 1.3243514961977844e-05}
{"train_loss": 0.08967088907957077, "global_step": 207202, "epoch": 2328, "lr": 1.3243121945649129e-05}
{"train_loss": 0.06930182129144669, "global_step": 207203, "epoch": 2328, "lr": 1.324272893426194e-05}
{"train_loss": 0.10748651623725891, "global_step": 207204, "epoch": 2328, "lr": 1.3242335927816307e-05}
{"train_loss": 0.055746376514434814, "global_step": 207205, "epoch": 2328, "lr": 1.324194292631229e-05}
{"train_loss": 0.04058077558875084, "global_step": 207206, "epoch": 2328, "lr": 1.3241549929749964e-05}
{"train_loss": 0.025936881080269814, "global_step": 207207, "epoch": 2328, "lr": 1.3241156938129351e-05}
{"train_loss": 0.12531106173992157, "global_step": 207208, "epoch": 2328, "lr": 1.3240763951450525e-05}
{"train_loss": 0.10009890794754028, "global_step": 207209, "epoch": 2328, "lr": 1.3240370969713517e-05}
{"train_loss": 0.039640430361032486, "global_step": 207210, "epoch": 2328, "lr": 1.3239977992918412e-05}
{"train_loss": 0.04317992180585861, "global_step": 207211, "epoch": 2328, "lr": 1.3239585021065227e-05}
{"train_loss": 0.04760923981666565, "global_step": 207212, "epoch": 2328, "lr": 1.3239192054154049e-05}
{"train_loss": 0.04119778424501419, "global_step": 207213, "epoch": 2328, "lr": 1.3238799092184894e-05}
{"train_loss": 0.04781609773635864, "global_step": 207214, "epoch": 2328, "lr": 1.3238406135157849e-05}
{"train_loss": 0.07131481170654297, "global_step": 207215, "epoch": 2328, "lr": 1.3238013183072939e-05}
{"train_loss": 0.06967537850141525, "global_step": 207216, "epoch": 2328, "lr": 1.3237620235930248e-05}
{"train_loss": 0.0528283566236496, "global_step": 207217, "epoch": 2328, "lr": 1.3237227293729792e-05}
{"train_loss": 0.011772751808166504, "global_step": 207218, "epoch": 2328, "lr": 1.323683435647166e-05}
{"train_loss": 0.056836195290088654, "global_step": 207219, "epoch": 2328, "lr": 1.3236441424155871e-05}
{"train_loss": 0.09322810173034668, "global_step": 207220, "epoch": 2328, "lr": 1.323604849678251e-05}
{"train_loss": 0.058081064373254776, "global_step": 207221, "epoch": 2328, "lr": 1.3235655574351603e-05}
{"train_loss": 0.04980692267417908, "global_step": 207222, "epoch": 2328, "lr": 1.3235262656863223e-05}
{"train_loss": 0.05995761603116989, "global_step": 207223, "epoch": 2328, "lr": 1.3234869744317413e-05}
{"train_loss": 0.08330795913934708, "global_step": 207224, "epoch": 2328, "lr": 1.3234476836714215e-05}
{"train_loss": 0.05529914051294327, "global_step": 207225, "epoch": 2328, "lr": 1.323408393405371e-05}
{"train_loss": 0.06300023198127747, "global_step": 207226, "epoch": 2328, "lr": 1.3233691036335916e-05}
{"train_loss": 0.06462416797876358, "global_step": 207227, "epoch": 2328, "lr": 1.3233298143560924e-05}
{"train_loss": 0.0698522999882698, "global_step": 207228, "epoch": 2328, "lr": 1.3232905255728744e-05}
{"train_loss": 0.04994687810540199, "global_step": 207229, "epoch": 2328, "lr": 1.3232512372839474e-05}
{"train_loss": 0.0996604710817337, "global_step": 207230, "epoch": 2328, "lr": 1.3232119494893125e-05}
{"train_loss": 0.10912522673606873, "global_step": 207231, "epoch": 2328, "lr": 1.323172662188979e-05}
{"train_loss": 0.036915384232997894, "global_step": 207232, "epoch": 2328, "lr": 1.3231333753829484e-05}
{"train_loss": 0.04036874324083328, "global_step": 207233, "epoch": 2328, "lr": 1.323094089071229e-05}
{"train_loss": 0.06658767908811569, "global_step": 207234, "epoch": 2328, "lr": 1.3230548032538231e-05}
{"train_loss": 0.05255606397986412, "global_step": 207235, "epoch": 2328, "lr": 1.3230155179307396e-05}
{"train_loss": 0.03270476311445236, "global_step": 207236, "epoch": 2328, "lr": 1.3229762331019801e-05}
{"train_loss": 0.04704776406288147, "global_step": 207237, "epoch": 2328, "lr": 1.322936948767553e-05}
{"train_loss": 0.06205199658870697, "global_step": 207238, "epoch": 2328, "lr": 1.322897664927461e-05}
{"train_loss": 0.05934998393058777, "global_step": 207239, "epoch": 2328, "lr": 1.322858381581712e-05}
{"train_loss": 0.08377200365066528, "global_step": 207240, "epoch": 2328, "lr": 1.3228190987303085e-05}
{"train_loss": 0.08576890826225281, "global_step": 207241, "epoch": 2328, "lr": 1.3227798163732568e-05}
{"train_loss": 0.08472398668527603, "global_step": 207242, "epoch": 2328, "lr": 1.3227405345105643e-05}
{"train_loss": 0.061983540654182434, "global_step": 207243, "epoch": 2328, "lr": 1.3227012531422328e-05}
{"train_loss": 0.0655900239944458, "global_step": 207244, "epoch": 2328, "lr": 1.322661972268271e-05}
{"train_loss": 0.05757421255111694, "global_step": 207245, "epoch": 2328, "lr": 1.3226226918886808e-05}
{"train_loss": 0.03570666164159775, "global_step": 207246, "epoch": 2328, "lr": 1.3225834120034709e-05}
{"train_loss": 0.04867497459053993, "global_step": 207247, "epoch": 2328, "lr": 1.322544132612643e-05}
{"train_loss": 0.04154089093208313, "global_step": 207248, "epoch": 2328, "lr": 1.322504853716206e-05}
{"train_loss": 0.0593104362487793, "global_step": 207249, "epoch": 2328, "lr": 1.3224655753141624e-05}
{"train_loss": 0.06200713291764259, "global_step": 207250, "epoch": 2328, "lr": 1.3224262974065193e-05}
{"train_loss": 0.06532168388366699, "global_step": 207251, "epoch": 2328, "lr": 1.3223870199932798e-05}
{"train_loss": 0.08279236406087875, "global_step": 207252, "epoch": 2328, "lr": 1.322347743074452e-05}
{"train_loss": 0.07342077791690826, "global_step": 207253, "epoch": 2328, "lr": 1.322308466650038e-05}
{"train_loss": 0.06076822429895401, "global_step": 207254, "epoch": 2328, "lr": 1.3222691907200469e-05}
{"train_loss": 0.04814288765192032, "global_step": 207255, "epoch": 2328, "lr": 1.3222299152844798e-05}
{"train_loss": 0.033785343170166016, "global_step": 207256, "epoch": 2328, "lr": 1.3221906403433459e-05}
{"train_loss": 0.02525581605732441, "global_step": 207257, "epoch": 2328, "lr": 1.3221513658966472e-05}
{"train_loss": 0.039820581674575806, "global_step": 207258, "epoch": 2328, "lr": 1.3221120919443913e-05}
{"train_loss": 0.08212307840585709, "global_step": 207259, "epoch": 2328, "lr": 1.3220728184865817e-05}
{"train_loss": 0.1366097331047058, "global_step": 207260, "epoch": 2328, "lr": 1.3220335455232264e-05}
{"train_loss": 0.07380911707878113, "global_step": 207261, "epoch": 2328, "lr": 1.3219942730543283e-05}
{"train_loss": 0.033498939126729965, "global_step": 207262, "epoch": 2328, "lr": 1.3219550010798914e-05}
{"train_loss": 0.08986254781484604, "global_step": 207263, "epoch": 2328, "lr": 1.3219157295999251e-05}
{"train_loss": 0.04978211224079132, "global_step": 207264, "epoch": 2328, "lr": 1.3218764586144305e-05}
{"train_loss": 0.09778106212615967, "global_step": 207265, "epoch": 2328, "lr": 1.3218371881234165e-05}
{"train_loss": 0.051568858325481415, "global_step": 207266, "epoch": 2328, "lr": 1.3217979181268848e-05}
{"train_loss": 0.05623259022831917, "global_step": 207267, "epoch": 2328, "lr": 1.3217586486248446e-05}
{"train_loss": 0.08651552349328995, "global_step": 207268, "epoch": 2328, "lr": 1.3217193796172971e-05}
{"train_loss": 0.043593630194664, "global_step": 207269, "epoch": 2328, "lr": 1.3216801111042516e-05}
{"train_loss": 0.05679042637348175, "global_step": 207270, "epoch": 2328, "lr": 1.3216408430857091e-05}
{"train_loss": 0.06263972073793411, "global_step": 207271, "epoch": 2328, "lr": 1.3216015755616789e-05}
{"train_loss": 0.10604537278413773, "global_step": 207272, "epoch": 2328, "lr": 1.3215623085321638e-05}
{"train_loss": 0.12603260576725006, "global_step": 207273, "epoch": 2328, "lr": 1.3215230419971702e-05}
{"train_loss": 0.05485471338033676, "global_step": 207274, "epoch": 2328, "lr": 1.3214837759567023e-05}
{"train_loss": 0.06765330582857132, "global_step": 207275, "epoch": 2328, "lr": 1.3214445104107676e-05}
{"train_loss": 0.019227532669901848, "global_step": 207276, "epoch": 2328, "lr": 1.3214052453593678e-05}
{"train_loss": 0.07840389013290405, "global_step": 207277, "epoch": 2328, "lr": 1.3213659808025103e-05}
{"train_loss": 0.07948057353496552, "global_step": 207278, "epoch": 2328, "lr": 1.321326716740202e-05}
{"train_loss": 0.03195585310459137, "global_step": 207279, "epoch": 2328, "lr": 1.321287453172445e-05}
{"train_loss": 0.06285159626787298, "global_step": 207280, "epoch": 2328, "lr": 1.3212481900992479e-05, "val_loss": 8.454073905944824}
{"train_loss": 0.090842105448246, "global_step": 207281, "epoch": 2329, "lr": 1.3212089275206119e-05}
{"train_loss": 0.040161024779081345, "global_step": 207282, "epoch": 2329, "lr": 1.321169665436547e-05}
{"train_loss": 0.0738217830657959, "global_step": 207283, "epoch": 2329, "lr": 1.3211304038470535e-05}
{"train_loss": 0.05184868350625038, "global_step": 207284, "epoch": 2329, "lr": 1.321091142752141e-05}
{"train_loss": 0.03414059430360794, "global_step": 207285, "epoch": 2329, "lr": 1.321051882151812e-05}
{"train_loss": 0.059570200741291046, "global_step": 207286, "epoch": 2329, "lr": 1.3210126220460733e-05}
{"train_loss": 0.03586956113576889, "global_step": 207287, "epoch": 2329, "lr": 1.3209733624349285e-05}
{"train_loss": 0.05040859803557396, "global_step": 207288, "epoch": 2329, "lr": 1.3209341033183859e-05}
{"train_loss": 0.059043318033218384, "global_step": 207289, "epoch": 2329, "lr": 1.320894844696447e-05}
{"train_loss": 0.0871962457895279, "global_step": 207290, "epoch": 2329, "lr": 1.3208555865691207e-05}
{"train_loss": 0.07545916736125946, "global_step": 207291, "epoch": 2329, "lr": 1.3208163289364089e-05}
{"train_loss": 0.05605826899409294, "global_step": 207292, "epoch": 2329, "lr": 1.3207770717983197e-05}
{"train_loss": 0.06356675177812576, "global_step": 207293, "epoch": 2329, "lr": 1.3207378151548561e-05}
{"train_loss": 0.10315535962581635, "global_step": 207294, "epoch": 2329, "lr": 1.3206985590060256e-05}
{"train_loss": 0.0937732607126236, "global_step": 207295, "epoch": 2329, "lr": 1.320659303351831e-05}
{"train_loss": 0.08919312804937363, "global_step": 207296, "epoch": 2329, "lr": 1.3206200481922803e-05}
{"train_loss": 0.04277268424630165, "global_step": 207297, "epoch": 2329, "lr": 1.3205807935273762e-05}
{"train_loss": 0.047529056668281555, "global_step": 207298, "epoch": 2329, "lr": 1.3205415393571264e-05}
{"train_loss": 0.0735231265425682, "global_step": 207299, "epoch": 2329, "lr": 1.3205022856815353e-05}
{"train_loss": 0.0878603532910347, "global_step": 207300, "epoch": 2329, "lr": 1.3204630325006057e-05}
{"train_loss": 0.028727179393172264, "global_step": 207301, "epoch": 2329, "lr": 1.320423779814347e-05}
{"train_loss": 0.04462199658155441, "global_step": 207302, "epoch": 2329, "lr": 1.3203845276227605e-05}
{"train_loss": 0.0466955192387104, "global_step": 207303, "epoch": 2329, "lr": 1.3203452759258555e-05}
{"train_loss": 0.06084263324737549, "global_step": 207304, "epoch": 2329, "lr": 1.3203060247236332e-05}
{"train_loss": 0.045483946800231934, "global_step": 207305, "epoch": 2329, "lr": 1.320266774016103e-05}
{"train_loss": 0.07289502769708633, "global_step": 207306, "epoch": 2329, "lr": 1.3202275238032662e-05}
{"train_loss": 0.05208032205700874, "global_step": 207307, "epoch": 2329, "lr": 1.3201882740851312e-05}
{"train_loss": 0.054367050528526306, "global_step": 207308, "epoch": 2329, "lr": 1.320149024861701e-05}
{"train_loss": 0.1153562068939209, "global_step": 207309, "epoch": 2329, "lr": 1.3201097761329833e-05}
{"train_loss": 0.043660808354616165, "global_step": 207310, "epoch": 2329, "lr": 1.3200705278989806e-05}
{"train_loss": 0.07272662967443466, "global_step": 207311, "epoch": 2329, "lr": 1.320031280159701e-05}
{"train_loss": 0.09754115343093872, "global_step": 207312, "epoch": 2329, "lr": 1.3199920329151466e-05}
{"train_loss": 0.02968444675207138, "global_step": 207313, "epoch": 2329, "lr": 1.319952786165325e-05}
{"train_loss": 0.06018951162695885, "global_step": 207314, "epoch": 2329, "lr": 1.3199135399102419e-05}
{"train_loss": 0.04429282620549202, "global_step": 207315, "epoch": 2329, "lr": 1.3198742941498997e-05}
{"train_loss": 0.09789819270372391, "global_step": 207316, "epoch": 2329, "lr": 1.3198350488843075e-05}
{"train_loss": 0.07019014656543732, "global_step": 207317, "epoch": 2329, "lr": 1.3197958041134673e-05}
{"train_loss": 0.0817403793334961, "global_step": 207318, "epoch": 2329, "lr": 1.3197565598373874e-05}
{"train_loss": 0.0342353992164135, "global_step": 207319, "epoch": 2329, "lr": 1.3197173160560695e-05}
{"train_loss": 0.08370789140462875, "global_step": 207320, "epoch": 2329, "lr": 1.3196780727695218e-05}
{"train_loss": 0.06942028552293777, "global_step": 207321, "epoch": 2329, "lr": 1.319638829977748e-05}
{"train_loss": 0.07714192569255829, "global_step": 207322, "epoch": 2329, "lr": 1.3195995876807543e-05}
{"train_loss": 0.042823322117328644, "global_step": 207323, "epoch": 2329, "lr": 1.319560345878545e-05}
{"train_loss": 0.045451752841472626, "global_step": 207324, "epoch": 2329, "lr": 1.319521104571127e-05}
{"train_loss": 0.04941478744149208, "global_step": 207325, "epoch": 2329, "lr": 1.3194818637585032e-05}
{"train_loss": 0.06360981613397598, "global_step": 207326, "epoch": 2329, "lr": 1.3194426234406815e-05}
{"train_loss": 0.059520378708839417, "global_step": 207327, "epoch": 2329, "lr": 1.3194033836176644e-05}
{"train_loss": 0.09260918945074081, "global_step": 207328, "epoch": 2329, "lr": 1.3193641442894606e-05}
{"train_loss": 0.040512602776288986, "global_step": 207329, "epoch": 2329, "lr": 1.3193249054560714e-05}
{"train_loss": 0.03515179455280304, "global_step": 207330, "epoch": 2329, "lr": 1.3192856671175058e-05}
{"train_loss": 0.043703168630599976, "global_step": 207331, "epoch": 2329, "lr": 1.319246429273766e-05}
{"train_loss": 0.08411168307065964, "global_step": 207332, "epoch": 2329, "lr": 1.31920719192486e-05}
{"train_loss": 0.08105763047933578, "global_step": 207333, "epoch": 2329, "lr": 1.3191679550707904e-05}
{"train_loss": 0.039598025381565094, "global_step": 207334, "epoch": 2329, "lr": 1.3191287187115653e-05}
{"train_loss": 0.04573517665266991, "global_step": 207335, "epoch": 2329, "lr": 1.3190894828471866e-05}
{"train_loss": 0.04371042177081108, "global_step": 207336, "epoch": 2329, "lr": 1.3190502474776633e-05}
{"train_loss": 0.06472305208444595, "global_step": 207337, "epoch": 2329, "lr": 1.3190110126029975e-05}
{"train_loss": 0.028664959594607353, "global_step": 207338, "epoch": 2329, "lr": 1.3189717782231969e-05}
{"train_loss": 0.07370202243328094, "global_step": 207339, "epoch": 2329, "lr": 1.318932544338266e-05}
{"train_loss": 0.07079721242189407, "global_step": 207340, "epoch": 2329, "lr": 1.3188933109482077e-05}
{"train_loss": 0.05023094639182091, "global_step": 207341, "epoch": 2329, "lr": 1.3188540780530312e-05}
{"train_loss": 0.07486114650964737, "global_step": 207342, "epoch": 2329, "lr": 1.3188148456527389e-05}
{"train_loss": 0.07080911844968796, "global_step": 207343, "epoch": 2329, "lr": 1.3187756137473379e-05}
{"train_loss": 0.05041899159550667, "global_step": 207344, "epoch": 2329, "lr": 1.3187363823368314e-05}
{"train_loss": 0.07323388755321503, "global_step": 207345, "epoch": 2329, "lr": 1.318697151421227e-05}
{"train_loss": 0.03194262459874153, "global_step": 207346, "epoch": 2329, "lr": 1.318657921000528e-05}
{"train_loss": 0.03835165500640869, "global_step": 207347, "epoch": 2329, "lr": 1.3186186910747417e-05}
{"train_loss": 0.0611424520611763, "global_step": 207348, "epoch": 2329, "lr": 1.3185794616438712e-05}
{"train_loss": 0.09873268008232117, "global_step": 207349, "epoch": 2329, "lr": 1.3185402327079238e-05}
{"train_loss": 0.08518246561288834, "global_step": 207350, "epoch": 2329, "lr": 1.3185010042669028e-05}
{"train_loss": 0.05140573903918266, "global_step": 207351, "epoch": 2329, "lr": 1.318461776320814e-05}
{"train_loss": 0.027717314660549164, "global_step": 207352, "epoch": 2329, "lr": 1.318422548869665e-05}
{"train_loss": 0.12827445566654205, "global_step": 207353, "epoch": 2329, "lr": 1.3183833219134573e-05}
{"train_loss": 0.09468898177146912, "global_step": 207354, "epoch": 2329, "lr": 1.3183440954521997e-05}
{"train_loss": 0.0867290124297142, "global_step": 207355, "epoch": 2329, "lr": 1.3183048694858945e-05}
{"train_loss": 0.052324458956718445, "global_step": 207356, "epoch": 2329, "lr": 1.3182656440145502e-05}
{"train_loss": 0.06769685447216034, "global_step": 207357, "epoch": 2329, "lr": 1.3182264190381682e-05}
{"train_loss": 0.04105738550424576, "global_step": 207358, "epoch": 2329, "lr": 1.3181871945567576e-05}
{"train_loss": 0.042944446206092834, "global_step": 207359, "epoch": 2329, "lr": 1.3181479705703198e-05}
{"train_loss": 0.0729859322309494, "global_step": 207360, "epoch": 2329, "lr": 1.318108747078864e-05}
{"train_loss": 0.0677139163017273, "global_step": 207361, "epoch": 2329, "lr": 1.318069524082392e-05}
{"train_loss": 0.05499370023608208, "global_step": 207362, "epoch": 2329, "lr": 1.3180303015809126e-05}
{"train_loss": 0.07675628364086151, "global_step": 207363, "epoch": 2329, "lr": 1.3179910795744272e-05}
{"train_loss": 0.08768240362405777, "global_step": 207364, "epoch": 2329, "lr": 1.3179518580629446e-05}
{"train_loss": 0.10192769765853882, "global_step": 207365, "epoch": 2329, "lr": 1.3179126370464672e-05}
{"train_loss": 0.09003143757581711, "global_step": 207366, "epoch": 2329, "lr": 1.3178734165250029e-05}
{"train_loss": 0.05194306746125221, "global_step": 207367, "epoch": 2329, "lr": 1.317834196498554e-05}
{"train_loss": 0.09613737463951111, "global_step": 207368, "epoch": 2329, "lr": 1.3177949769671277e-05}
{"train_loss": 0.06443349464555805, "global_step": 207369, "epoch": 2329, "lr": 1.3177557579307303e-05, "val_loss": 8.535935401916504}
{"train_loss": 0.13811394572257996, "global_step": 207370, "epoch": 2330, "lr": 1.3177165393893642e-05}
{"train_loss": 0.03449490666389465, "global_step": 207371, "epoch": 2330, "lr": 1.3176773213430382e-05}
{"train_loss": 0.07848852872848511, "global_step": 207372, "epoch": 2330, "lr": 1.3176381037917551e-05}
{"train_loss": 0.037374451756477356, "global_step": 207373, "epoch": 2330, "lr": 1.3175988867355193e-05}
{"train_loss": 0.05635644868016243, "global_step": 207374, "epoch": 2330, "lr": 1.3175596701743392e-05}
{"train_loss": 0.05350848287343979, "global_step": 207375, "epoch": 2330, "lr": 1.3175204541082165e-05}
{"train_loss": 0.056383367627859116, "global_step": 207376, "epoch": 2330, "lr": 1.31748123853716e-05}
{"train_loss": 0.044055208563804626, "global_step": 207377, "epoch": 2330, "lr": 1.3174420234611728e-05}
{"train_loss": 0.02995411306619644, "global_step": 207378, "epoch": 2330, "lr": 1.3174028088802598e-05}
{"train_loss": 0.043512944132089615, "global_step": 207379, "epoch": 2330, "lr": 1.317363594794428e-05}
{"train_loss": 0.08067862689495087, "global_step": 207380, "epoch": 2330, "lr": 1.3173243812036806e-05}
{"train_loss": 0.036409031599760056, "global_step": 207381, "epoch": 2330, "lr": 1.3172851681080256e-05}
{"train_loss": 0.053738296031951904, "global_step": 207382, "epoch": 2330, "lr": 1.317245955507465e-05}
{"train_loss": 0.08137381821870804, "global_step": 207383, "epoch": 2330, "lr": 1.3172067434020074e-05}
{"train_loss": 0.0539245642721653, "global_step": 207384, "epoch": 2330, "lr": 1.3171675317916554e-05}
{"train_loss": 0.07647594809532166, "global_step": 207385, "epoch": 2330, "lr": 1.3171283206764162e-05}
{"train_loss": 0.0904134213924408, "global_step": 207386, "epoch": 2330, "lr": 1.3170891100562933e-05}
{"train_loss": 0.1127779483795166, "global_step": 207387, "epoch": 2330, "lr": 1.3170498999312925e-05}
{"train_loss": 0.07632912695407867, "global_step": 207388, "epoch": 2330, "lr": 1.3170106903014206e-05}
{"train_loss": 0.0719640776515007, "global_step": 207389, "epoch": 2330, "lr": 1.3169714811666806e-05}
{"train_loss": 0.04608675464987755, "global_step": 207390, "epoch": 2330, "lr": 1.3169322725270805e-05}
{"train_loss": 0.06781104952096939, "global_step": 207391, "epoch": 2330, "lr": 1.3168930643826222e-05}
{"train_loss": 0.06562449783086777, "global_step": 207392, "epoch": 2330, "lr": 1.3168538567333144e-05}
{"train_loss": 0.058671049773693085, "global_step": 207393, "epoch": 2330, "lr": 1.3168146495791594e-05}
{"train_loss": 0.07566539943218231, "global_step": 207394, "epoch": 2330, "lr": 1.316775442920165e-05}
{"train_loss": 0.04603482037782669, "global_step": 207395, "epoch": 2330, "lr": 1.316736236756334e-05}
{"train_loss": 0.02737141214311123, "global_step": 207396, "epoch": 2330, "lr": 1.3166970310876742e-05}
{"train_loss": 0.06426123529672623, "global_step": 207397, "epoch": 2330, "lr": 1.3166578259141882e-05}
{"train_loss": 0.03859757259488106, "global_step": 207398, "epoch": 2330, "lr": 1.3166186212358839e-05}
{"train_loss": 0.024106500670313835, "global_step": 207399, "epoch": 2330, "lr": 1.316579417052764e-05}
{"train_loss": 0.08107393980026245, "global_step": 207400, "epoch": 2330, "lr": 1.3165402133648368e-05}
{"train_loss": 0.0478263720870018, "global_step": 207401, "epoch": 2330, "lr": 1.3165010101721043e-05}
{"train_loss": 0.0766683965921402, "global_step": 207402, "epoch": 2330, "lr": 1.316461807474575e-05}
{"train_loss": 0.02092241682112217, "global_step": 207403, "epoch": 2330, "lr": 1.3164226052722507e-05}
{"train_loss": 0.07134883105754852, "global_step": 207404, "epoch": 2330, "lr": 1.3163834035651385e-05}
{"train_loss": 0.0741836205124855, "global_step": 207405, "epoch": 2330, "lr": 1.3163442023532452e-05}
{"train_loss": 0.10481525957584381, "global_step": 207406, "epoch": 2330, "lr": 1.316305001636573e-05}
{"train_loss": 0.09820950031280518, "global_step": 207407, "epoch": 2330, "lr": 1.3162658014151302e-05}
{"train_loss": 0.054099954664707184, "global_step": 207408, "epoch": 2330, "lr": 1.3162266016889191e-05}
{"train_loss": 0.1231609359383583, "global_step": 207409, "epoch": 2330, "lr": 1.3161874024579479e-05}
{"train_loss": 0.056391987949609756, "global_step": 207410, "epoch": 2330, "lr": 1.3161482037222206e-05}
{"train_loss": 0.05849718302488327, "global_step": 207411, "epoch": 2330, "lr": 1.3161090054817404e-05}
{"train_loss": 0.06386900693178177, "global_step": 207412, "epoch": 2330, "lr": 1.3160698077365164e-05}
{"train_loss": 0.03952302783727646, "global_step": 207413, "epoch": 2330, "lr": 1.31603061048655e-05}
{"train_loss": 0.055579110980033875, "global_step": 207414, "epoch": 2330, "lr": 1.3159914137318502e-05}
{"train_loss": 0.05521128699183464, "global_step": 207415, "epoch": 2330, "lr": 1.3159522174724203e-05}
{"train_loss": 0.04268096759915352, "global_step": 207416, "epoch": 2330, "lr": 1.3159130217082643e-05}
{"train_loss": 0.050361115485429764, "global_step": 207417, "epoch": 2330, "lr": 1.31587382643939e-05}
{"train_loss": 0.053214915096759796, "global_step": 207418, "epoch": 2330, "lr": 1.3158346316658004e-05}
{"train_loss": 0.08007121831178665, "global_step": 207419, "epoch": 2330, "lr": 1.3157954373875036e-05}
{"train_loss": 0.043118152767419815, "global_step": 207420, "epoch": 2330, "lr": 1.3157562436045013e-05}
{"train_loss": 0.06378338485956192, "global_step": 207421, "epoch": 2330, "lr": 1.3157170503168026e-05}
{"train_loss": 0.05269123241305351, "global_step": 207422, "epoch": 2330, "lr": 1.3156778575244088e-05}
{"train_loss": 0.102726049721241, "global_step": 207423, "epoch": 2330, "lr": 1.3156386652273278e-05}
{"train_loss": 0.10081310570240021, "global_step": 207424, "epoch": 2330, "lr": 1.3155994734255656e-05}
{"train_loss": 0.08717335760593414, "global_step": 207425, "epoch": 2330, "lr": 1.3155602821191243e-05}
{"train_loss": 0.06505218893289566, "global_step": 207426, "epoch": 2330, "lr": 1.315521091308013e-05}
{"train_loss": 0.06949271261692047, "global_step": 207427, "epoch": 2330, "lr": 1.315481900992233e-05}
{"train_loss": 0.09338848292827606, "global_step": 207428, "epoch": 2330, "lr": 1.3154427111717938e-05}
{"train_loss": 0.0324542410671711, "global_step": 207429, "epoch": 2330, "lr": 1.3154035218466958e-05}
{"train_loss": 0.0648433268070221, "global_step": 207430, "epoch": 2330, "lr": 1.315364333016949e-05}
{"train_loss": 0.06930269300937653, "global_step": 207431, "epoch": 2330, "lr": 1.3153251446825554e-05}
{"train_loss": 0.07397282123565674, "global_step": 207432, "epoch": 2330, "lr": 1.3152859568435227e-05}
{"train_loss": 0.040112074464559555, "global_step": 207433, "epoch": 2330, "lr": 1.3152467694998532e-05}
{"train_loss": 0.05503588542342186, "global_step": 207434, "epoch": 2330, "lr": 1.3152075826515553e-05}
{"train_loss": 0.05078525468707085, "global_step": 207435, "epoch": 2330, "lr": 1.3151683962986316e-05}
{"train_loss": 0.049003489315509796, "global_step": 207436, "epoch": 2330, "lr": 1.31512921044109e-05}
{"train_loss": 0.07055312395095825, "global_step": 207437, "epoch": 2330, "lr": 1.3150900250789327e-05}
{"train_loss": 0.04653967171907425, "global_step": 207438, "epoch": 2330, "lr": 1.315050840212168e-05}
{"train_loss": 0.06483034044504166, "global_step": 207439, "epoch": 2330, "lr": 1.3150116558407988e-05}
{"train_loss": 0.06114092469215393, "global_step": 207440, "epoch": 2330, "lr": 1.314972471964831e-05}
{"train_loss": 0.038460712879896164, "global_step": 207441, "epoch": 2330, "lr": 1.3149332885842718e-05}
{"train_loss": 0.040866535156965256, "global_step": 207442, "epoch": 2330, "lr": 1.3148941056991237e-05}
{"train_loss": 0.045917410403490067, "global_step": 207443, "epoch": 2330, "lr": 1.3148549233093943e-05}
{"train_loss": 0.05045822635293007, "global_step": 207444, "epoch": 2330, "lr": 1.3148157414150864e-05}
{"train_loss": 0.05271058902144432, "global_step": 207445, "epoch": 2330, "lr": 1.3147765600162082e-05}
{"train_loss": 0.04327428713440895, "global_step": 207446, "epoch": 2330, "lr": 1.3147373791127616e-05}
{"train_loss": 0.043179482221603394, "global_step": 207447, "epoch": 2330, "lr": 1.3146981987047552e-05}
{"train_loss": 0.045621607452631, "global_step": 207448, "epoch": 2330, "lr": 1.3146590187921915e-05}
{"train_loss": 0.04596428573131561, "global_step": 207449, "epoch": 2330, "lr": 1.3146198393750781e-05}
{"train_loss": 0.07223371416330338, "global_step": 207450, "epoch": 2330, "lr": 1.3145806604534194e-05}
{"train_loss": 0.044850219041109085, "global_step": 207451, "epoch": 2330, "lr": 1.314541482027219e-05}
{"train_loss": 0.041866324841976166, "global_step": 207452, "epoch": 2330, "lr": 1.314502304096485e-05}
{"train_loss": 0.07110840082168579, "global_step": 207453, "epoch": 2330, "lr": 1.3144631266612195e-05}
{"train_loss": 0.06927778571844101, "global_step": 207454, "epoch": 2330, "lr": 1.3144239497214312e-05}
{"train_loss": 0.0668950080871582, "global_step": 207455, "epoch": 2330, "lr": 1.3143847732771236e-05}
{"train_loss": 0.08469832688570023, "global_step": 207456, "epoch": 2330, "lr": 1.3143455973283009e-05}
{"train_loss": 0.0892966240644455, "global_step": 207457, "epoch": 2330, "lr": 1.3143064218749707e-05}
{"train_loss": 0.06133175214354911, "global_step": 207458, "epoch": 2330, "lr": 1.3142672469171352e-05, "val_loss": 8.472001075744629, "train_action_mse_error": 20.752216339111328}
{"train_loss": 0.08650150895118713, "global_step": 207459, "epoch": 2331, "lr": 1.3142280724548022e-05}
{"train_loss": 0.03288491070270538, "global_step": 207460, "epoch": 2331, "lr": 1.3141888984879774e-05}
{"train_loss": 0.03209745138883591, "global_step": 207461, "epoch": 2331, "lr": 1.314149725016664e-05}
{"train_loss": 0.10612750798463821, "global_step": 207462, "epoch": 2331, "lr": 1.3141105520408692e-05}
{"train_loss": 0.07811804860830307, "global_step": 207463, "epoch": 2331, "lr": 1.3140713795605958e-05}
{"train_loss": 0.04317851737141609, "global_step": 207464, "epoch": 2331, "lr": 1.3140322075758527e-05}
{"train_loss": 0.04163865000009537, "global_step": 207465, "epoch": 2331, "lr": 1.313993036086641e-05}
{"train_loss": 0.05746283754706383, "global_step": 207466, "epoch": 2331, "lr": 1.3139538650929694e-05}
{"train_loss": 0.0475781112909317, "global_step": 207467, "epoch": 2331, "lr": 1.3139146945948405e-05}
{"train_loss": 0.029503317549824715, "global_step": 207468, "epoch": 2331, "lr": 1.3138755245922624e-05}
{"train_loss": 0.03589530289173126, "global_step": 207469, "epoch": 2331, "lr": 1.3138363550852373e-05}
{"train_loss": 0.05433112382888794, "global_step": 207470, "epoch": 2331, "lr": 1.3137971860737736e-05}
{"train_loss": 0.04962281510233879, "global_step": 207471, "epoch": 2331, "lr": 1.3137580175578728e-05}
{"train_loss": 0.049225542694330215, "global_step": 207472, "epoch": 2331, "lr": 1.3137188495375446e-05}
{"train_loss": 0.0786479040980339, "global_step": 207473, "epoch": 2331, "lr": 1.31367968201279e-05}
{"train_loss": 0.06153665855526924, "global_step": 207474, "epoch": 2331, "lr": 1.3136405149836178e-05}
{"train_loss": 0.08242247998714447, "global_step": 207475, "epoch": 2331, "lr": 1.3136013484500303e-05}
{"train_loss": 0.09047964215278625, "global_step": 207476, "epoch": 2331, "lr": 1.313562182412036e-05}
{"train_loss": 0.038034114986658096, "global_step": 207477, "epoch": 2331, "lr": 1.313523016869636e-05}
{"train_loss": 0.10157507658004761, "global_step": 207478, "epoch": 2331, "lr": 1.3134838518228387e-05}
{"train_loss": 0.04487273842096329, "global_step": 207479, "epoch": 2331, "lr": 1.3134446872716499e-05}
{"train_loss": 0.05678986757993698, "global_step": 207480, "epoch": 2331, "lr": 1.313405523216072e-05}
{"train_loss": 0.13157500326633453, "global_step": 207481, "epoch": 2331, "lr": 1.313366359656113e-05}
{"train_loss": 0.030450426042079926, "global_step": 207482, "epoch": 2331, "lr": 1.3133271965917754e-05}
{"train_loss": 0.07849591970443726, "global_step": 207483, "epoch": 2331, "lr": 1.313288034023068e-05}
{"train_loss": 0.04343278706073761, "global_step": 207484, "epoch": 2331, "lr": 1.3132488719499925e-05}
{"train_loss": 0.037023648619651794, "global_step": 207485, "epoch": 2331, "lr": 1.3132097103725565e-05}
{"train_loss": 0.048916492611169815, "global_step": 207486, "epoch": 2331, "lr": 1.3131705492907637e-05}
{"train_loss": 0.03414815664291382, "global_step": 207487, "epoch": 2331, "lr": 1.3131313887046221e-05}
{"train_loss": 0.020896226167678833, "global_step": 207488, "epoch": 2331, "lr": 1.3130922286141345e-05}
{"train_loss": 0.024773381650447845, "global_step": 207489, "epoch": 2331, "lr": 1.313053069019305e-05}
{"train_loss": 0.0464644730091095, "global_step": 207490, "epoch": 2331, "lr": 1.3130139099201422e-05}
{"train_loss": 0.09399566799402237, "global_step": 207491, "epoch": 2331, "lr": 1.3129747513166484e-05}
{"train_loss": 0.08932175487279892, "global_step": 207492, "epoch": 2331, "lr": 1.3129355932088317e-05}
{"train_loss": 0.046570636332035065, "global_step": 207493, "epoch": 2331, "lr": 1.312896435596696e-05}
{"train_loss": 0.058686062693595886, "global_step": 207494, "epoch": 2331, "lr": 1.3128572784802445e-05}
{"train_loss": 0.08407286554574966, "global_step": 207495, "epoch": 2331, "lr": 1.3128181218594849e-05}
{"train_loss": 0.1022195965051651, "global_step": 207496, "epoch": 2331, "lr": 1.312778965734423e-05}
{"train_loss": 0.04698428511619568, "global_step": 207497, "epoch": 2331, "lr": 1.312739810105062e-05}
{"train_loss": 0.10331081598997116, "global_step": 207498, "epoch": 2331, "lr": 1.3127006549714094e-05}
{"train_loss": 0.04272119700908661, "global_step": 207499, "epoch": 2331, "lr": 1.312661500333468e-05}
{"train_loss": 0.030512012541294098, "global_step": 207500, "epoch": 2331, "lr": 1.3126223461912456e-05}
{"train_loss": 0.026668651029467583, "global_step": 207501, "epoch": 2331, "lr": 1.3125831925447446e-05}
{"train_loss": 0.06705936044454575, "global_step": 207502, "epoch": 2331, "lr": 1.3125440393939736e-05}
{"train_loss": 0.03454507142305374, "global_step": 207503, "epoch": 2331, "lr": 1.3125048867389345e-05}
{"train_loss": 0.0511801578104496, "global_step": 207504, "epoch": 2331, "lr": 1.3124657345796355e-05}
{"train_loss": 0.04585960507392883, "global_step": 207505, "epoch": 2331, "lr": 1.3124265829160793e-05}
{"train_loss": 0.037280190736055374, "global_step": 207506, "epoch": 2331, "lr": 1.3123874317482737e-05}
{"train_loss": 0.07869908958673477, "global_step": 207507, "epoch": 2331, "lr": 1.3123482810762217e-05}
{"train_loss": 0.047578826546669006, "global_step": 207508, "epoch": 2331, "lr": 1.3123091308999303e-05}
{"train_loss": 0.13561442494392395, "global_step": 207509, "epoch": 2331, "lr": 1.3122699812194028e-05}
{"train_loss": 0.06890004873275757, "global_step": 207510, "epoch": 2331, "lr": 1.3122308320346471e-05}
{"train_loss": 0.08779410272836685, "global_step": 207511, "epoch": 2331, "lr": 1.3121916833456655e-05}
{"train_loss": 0.0871894359588623, "global_step": 207512, "epoch": 2331, "lr": 1.3121525351524666e-05}
{"train_loss": 0.05118877440690994, "global_step": 207513, "epoch": 2331, "lr": 1.3121133874550518e-05}
{"train_loss": 0.03914486616849899, "global_step": 207514, "epoch": 2331, "lr": 1.3120742402534286e-05}
{"train_loss": 0.05775303393602371, "global_step": 207515, "epoch": 2331, "lr": 1.3120350935476039e-05}
{"train_loss": 0.02564973384141922, "global_step": 207516, "epoch": 2331, "lr": 1.3119959473375792e-05}
{"train_loss": 0.032977256923913956, "global_step": 207517, "epoch": 2331, "lr": 1.3119568016233635e-05}
{"train_loss": 0.03932629153132439, "global_step": 207518, "epoch": 2331, "lr": 1.3119176564049584e-05}
{"train_loss": 0.04525773972272873, "global_step": 207519, "epoch": 2331, "lr": 1.3118785116823729e-05}
{"train_loss": 0.08452458679676056, "global_step": 207520, "epoch": 2331, "lr": 1.3118393674556084e-05}
{"train_loss": 0.06821929663419724, "global_step": 207521, "epoch": 2331, "lr": 1.311800223724674e-05}
{"train_loss": 0.052781231701374054, "global_step": 207522, "epoch": 2331, "lr": 1.311761080489572e-05}
{"train_loss": 0.05325733870267868, "global_step": 207523, "epoch": 2331, "lr": 1.3117219377503098e-05}
{"train_loss": 0.02187744714319706, "global_step": 207524, "epoch": 2331, "lr": 1.31168279550689e-05}
{"train_loss": 0.08732228726148605, "global_step": 207525, "epoch": 2331, "lr": 1.3116436537593213e-05}
{"train_loss": 0.05003617703914642, "global_step": 207526, "epoch": 2331, "lr": 1.311604512507607e-05}
{"train_loss": 0.035766251385211945, "global_step": 207527, "epoch": 2331, "lr": 1.3115653717517512e-05}
{"train_loss": 0.058416206389665604, "global_step": 207528, "epoch": 2331, "lr": 1.3115262314917614e-05}
{"train_loss": 0.05911429226398468, "global_step": 207529, "epoch": 2331, "lr": 1.3114870917276412e-05}
{"train_loss": 0.045321274548769, "global_step": 207530, "epoch": 2331, "lr": 1.3114479524593976e-05}
{"train_loss": 0.06639213114976883, "global_step": 207531, "epoch": 2331, "lr": 1.3114088136870339e-05}
{"train_loss": 0.060437317937612534, "global_step": 207532, "epoch": 2331, "lr": 1.311369675410557e-05}
{"train_loss": 0.07033640146255493, "global_step": 207533, "epoch": 2331, "lr": 1.3113305376299707e-05}
{"train_loss": 0.07178708165884018, "global_step": 207534, "epoch": 2331, "lr": 1.3112914003452826e-05}
{"train_loss": 0.06633970886468887, "global_step": 207535, "epoch": 2331, "lr": 1.3112522635564945e-05}
{"train_loss": 0.02537524327635765, "global_step": 207536, "epoch": 2331, "lr": 1.3112131272636152e-05}
{"train_loss": 0.06723611056804657, "global_step": 207537, "epoch": 2331, "lr": 1.3111739914666465e-05}
{"train_loss": 0.03871430084109306, "global_step": 207538, "epoch": 2331, "lr": 1.3111348561655972e-05}
{"train_loss": 0.07385382056236267, "global_step": 207539, "epoch": 2331, "lr": 1.3110957213604697e-05}
{"train_loss": 0.06854111701250076, "global_step": 207540, "epoch": 2331, "lr": 1.311056587051272e-05}
{"train_loss": 0.06466083973646164, "global_step": 207541, "epoch": 2331, "lr": 1.311017453238006e-05}
{"train_loss": 0.06496798992156982, "global_step": 207542, "epoch": 2331, "lr": 1.3109783199206799e-05}
{"train_loss": 0.04668083041906357, "global_step": 207543, "epoch": 2331, "lr": 1.3109391870992965e-05}
{"train_loss": 0.041162796318531036, "global_step": 207544, "epoch": 2331, "lr": 1.3109000547738642e-05}
{"train_loss": 0.0772438794374466, "global_step": 207545, "epoch": 2331, "lr": 1.3108609229443847e-05}
{"train_loss": 0.031281616538763046, "global_step": 207546, "epoch": 2331, "lr": 1.3108217916108668e-05}
{"train_loss": 0.057789260772674274, "global_step": 207547, "epoch": 2331, "lr": 1.310782660773312e-05, "val_loss": 8.47760009765625}
{"train_loss": 0.025742415338754654, "global_step": 207548, "epoch": 2332, "lr": 1.310743530431729e-05}
{"train_loss": 0.04260097071528435, "global_step": 207549, "epoch": 2332, "lr": 1.3107044005861202e-05}
{"train_loss": 0.0963452011346817, "global_step": 207550, "epoch": 2332, "lr": 1.3106652712364925e-05}
{"train_loss": 0.09031213819980621, "global_step": 207551, "epoch": 2332, "lr": 1.3106261423828519e-05}
{"train_loss": 0.07690580934286118, "global_step": 207552, "epoch": 2332, "lr": 1.3105870140252014e-05}
{"train_loss": 0.044213611632585526, "global_step": 207553, "epoch": 2332, "lr": 1.310547886163549e-05}
{"train_loss": 0.09705111384391785, "global_step": 207554, "epoch": 2332, "lr": 1.3105087587978971e-05}
{"train_loss": 0.0592401959002018, "global_step": 207555, "epoch": 2332, "lr": 1.3104696319282533e-05}
{"train_loss": 0.07471385598182678, "global_step": 207556, "epoch": 2332, "lr": 1.3104305055546213e-05}
{"train_loss": 0.07864489406347275, "global_step": 207557, "epoch": 2332, "lr": 1.3103913796770074e-05}
{"train_loss": 0.04855680465698242, "global_step": 207558, "epoch": 2332, "lr": 1.3103522542954154e-05}
{"train_loss": 0.03412517160177231, "global_step": 207559, "epoch": 2332, "lr": 1.3103131294098531e-05}
{"train_loss": 0.024145880714058876, "global_step": 207560, "epoch": 2332, "lr": 1.3102740050203228e-05}
{"train_loss": 0.09299340844154358, "global_step": 207561, "epoch": 2332, "lr": 1.3102348811268327e-05}
{"train_loss": 0.08421630412340164, "global_step": 207562, "epoch": 2332, "lr": 1.310195757729385e-05}
{"train_loss": 0.04011460766196251, "global_step": 207563, "epoch": 2332, "lr": 1.310156634827988e-05}
{"train_loss": 0.047525882720947266, "global_step": 207564, "epoch": 2332, "lr": 1.3101175124226439e-05}
{"train_loss": 0.11579429358243942, "global_step": 207565, "epoch": 2332, "lr": 1.3100783905133612e-05}
{"train_loss": 0.09082488715648651, "global_step": 207566, "epoch": 2332, "lr": 1.3100392691001434e-05}
{"train_loss": 0.06615807116031647, "global_step": 207567, "epoch": 2332, "lr": 1.3100001481829943e-05}
{"train_loss": 0.07199710607528687, "global_step": 207568, "epoch": 2332, "lr": 1.3099610277619222e-05}
{"train_loss": 0.03364413231611252, "global_step": 207569, "epoch": 2332, "lr": 1.3099219078369291e-05}
{"train_loss": 0.07314454019069672, "global_step": 207570, "epoch": 2332, "lr": 1.3098827884080238e-05}
{"train_loss": 0.056607555598020554, "global_step": 207571, "epoch": 2332, "lr": 1.3098436694752087e-05}
{"train_loss": 0.046160224825143814, "global_step": 207572, "epoch": 2332, "lr": 1.309804551038491e-05}
{"train_loss": 0.0632622167468071, "global_step": 207573, "epoch": 2332, "lr": 1.3097654330978737e-05}
{"train_loss": 0.1281823366880417, "global_step": 207574, "epoch": 2332, "lr": 1.3097263156533652e-05}
{"train_loss": 0.022717321291565895, "global_step": 207575, "epoch": 2332, "lr": 1.3096871987049674e-05}
{"train_loss": 0.054415274411439896, "global_step": 207576, "epoch": 2332, "lr": 1.3096480822526886e-05}
{"train_loss": 0.011218832805752754, "global_step": 207577, "epoch": 2332, "lr": 1.309608966296531e-05}
{"train_loss": 0.07788349688053131, "global_step": 207578, "epoch": 2332, "lr": 1.3095698508365035e-05}
{"train_loss": 0.06124669685959816, "global_step": 207579, "epoch": 2332, "lr": 1.3095307358726072e-05}
{"train_loss": 0.05393432825803757, "global_step": 207580, "epoch": 2332, "lr": 1.3094916214048513e-05}
{"train_loss": 0.04104579612612724, "global_step": 207581, "epoch": 2332, "lr": 1.309452507433238e-05}
{"train_loss": 0.04184702783823013, "global_step": 207582, "epoch": 2332, "lr": 1.3094133939577752e-05}
{"train_loss": 0.08482381701469421, "global_step": 207583, "epoch": 2332, "lr": 1.3093742809784654e-05}
{"train_loss": 0.04660765081644058, "global_step": 207584, "epoch": 2332, "lr": 1.3093351684953164e-05}
{"train_loss": 0.0703866183757782, "global_step": 207585, "epoch": 2332, "lr": 1.3092960565083311e-05}
{"train_loss": 0.04443484917283058, "global_step": 207586, "epoch": 2332, "lr": 1.3092569450175163e-05}
{"train_loss": 0.042950090020895004, "global_step": 207587, "epoch": 2332, "lr": 1.3092178340228777e-05}
{"train_loss": 0.08909004926681519, "global_step": 207588, "epoch": 2332, "lr": 1.3091787235244185e-05}
{"train_loss": 0.030980471521615982, "global_step": 207589, "epoch": 2332, "lr": 1.3091396135221468e-05}
{"train_loss": 0.04113641381263733, "global_step": 207590, "epoch": 2332, "lr": 1.3091005040160648e-05}
{"train_loss": 0.06059372425079346, "global_step": 207591, "epoch": 2332, "lr": 1.3090613950061808e-05}
{"train_loss": 0.026815520599484444, "global_step": 207592, "epoch": 2332, "lr": 1.3090222864924967e-05}
{"train_loss": 0.040227584540843964, "global_step": 207593, "epoch": 2332, "lr": 1.3089831784750218e-05}
{"train_loss": 0.03522321954369545, "global_step": 207594, "epoch": 2332, "lr": 1.308944070953757e-05}
{"train_loss": 0.056678734719753265, "global_step": 207595, "epoch": 2332, "lr": 1.3089049639287116e-05}
{"train_loss": 0.061156630516052246, "global_step": 207596, "epoch": 2332, "lr": 1.3088658573998869e-05}
{"train_loss": 0.05953992158174515, "global_step": 207597, "epoch": 2332, "lr": 1.308826751367292e-05}
{"train_loss": 0.03890679031610489, "global_step": 207598, "epoch": 2332, "lr": 1.3087876458309289e-05}
{"train_loss": 0.13466355204582214, "global_step": 207599, "epoch": 2332, "lr": 1.3087485407908057e-05}
{"train_loss": 0.06599219888448715, "global_step": 207600, "epoch": 2332, "lr": 1.308709436246925e-05}
{"train_loss": 0.08753049373626709, "global_step": 207601, "epoch": 2332, "lr": 1.3086703321992944e-05}
{"train_loss": 0.04588162526488304, "global_step": 207602, "epoch": 2332, "lr": 1.3086312286479169e-05}
{"train_loss": 0.10271865129470825, "global_step": 207603, "epoch": 2332, "lr": 1.3085921255928002e-05}
{"train_loss": 0.08617333322763443, "global_step": 207604, "epoch": 2332, "lr": 1.3085530230339482e-05}
{"train_loss": 0.019587276503443718, "global_step": 207605, "epoch": 2332, "lr": 1.3085139209713655e-05}
{"train_loss": 0.04748767614364624, "global_step": 207606, "epoch": 2332, "lr": 1.308474819405059e-05}
{"train_loss": 0.0454990454018116, "global_step": 207607, "epoch": 2332, "lr": 1.3084357183350316e-05}
{"train_loss": 0.0530831553041935, "global_step": 207608, "epoch": 2332, "lr": 1.3083966177612916e-05}
{"train_loss": 0.037348903715610504, "global_step": 207609, "epoch": 2332, "lr": 1.3083575176838409e-05}
{"train_loss": 0.04579480364918709, "global_step": 207610, "epoch": 2332, "lr": 1.3083184181026887e-05}
{"train_loss": 0.0447191447019577, "global_step": 207611, "epoch": 2332, "lr": 1.308279319017836e-05}
{"train_loss": 0.06096359342336655, "global_step": 207612, "epoch": 2332, "lr": 1.3082402204292921e-05}
{"train_loss": 0.09339281916618347, "global_step": 207613, "epoch": 2332, "lr": 1.3082011223370582e-05}
{"train_loss": 0.044188860803842545, "global_step": 207614, "epoch": 2332, "lr": 1.3081620247411436e-05}
{"train_loss": 0.05282213166356087, "global_step": 207615, "epoch": 2332, "lr": 1.3081229276415503e-05}
{"train_loss": 0.04326914995908737, "global_step": 207616, "epoch": 2332, "lr": 1.308083831038286e-05}
{"train_loss": 0.02717715874314308, "global_step": 207617, "epoch": 2332, "lr": 1.3080447349313534e-05}
{"train_loss": 0.017299726605415344, "global_step": 207618, "epoch": 2332, "lr": 1.3080056393207607e-05}
{"train_loss": 0.04271926358342171, "global_step": 207619, "epoch": 2332, "lr": 1.3079665442065103e-05}
{"train_loss": 0.0736641064286232, "global_step": 207620, "epoch": 2332, "lr": 1.3079274495886102e-05}
{"train_loss": 0.04262274131178856, "global_step": 207621, "epoch": 2332, "lr": 1.307888355467063e-05}
{"train_loss": 0.048964761197566986, "global_step": 207622, "epoch": 2332, "lr": 1.3078492618418748e-05}
{"train_loss": 0.07005852460861206, "global_step": 207623, "epoch": 2332, "lr": 1.3078101687130529e-05}
{"train_loss": 0.05334991216659546, "global_step": 207624, "epoch": 2332, "lr": 1.3077710760805995e-05}
{"train_loss": 0.06211641803383827, "global_step": 207625, "epoch": 2332, "lr": 1.3077319839445224e-05}
{"train_loss": 0.04236132651567459, "global_step": 207626, "epoch": 2332, "lr": 1.3076928923048249e-05}
{"train_loss": 0.05070788785815239, "global_step": 207627, "epoch": 2332, "lr": 1.307653801161514e-05}
{"train_loss": 0.032751016318798065, "global_step": 207628, "epoch": 2332, "lr": 1.3076147105145925e-05}
{"train_loss": 0.03240890055894852, "global_step": 207629, "epoch": 2332, "lr": 1.3075756203640694e-05}
{"train_loss": 0.03521846979856491, "global_step": 207630, "epoch": 2332, "lr": 1.3075365307099457e-05}
{"train_loss": 0.05568874254822731, "global_step": 207631, "epoch": 2332, "lr": 1.30749744155223e-05}
{"train_loss": 0.0386834554374218, "global_step": 207632, "epoch": 2332, "lr": 1.3074583528909251e-05}
{"train_loss": 0.09475310891866684, "global_step": 207633, "epoch": 2332, "lr": 1.3074192647260392e-05}
{"train_loss": 0.06660665571689606, "global_step": 207634, "epoch": 2332, "lr": 1.3073801770575738e-05}
{"train_loss": 0.09167364984750748, "global_step": 207635, "epoch": 2332, "lr": 1.3073410898855382e-05}
{"train_loss": 0.05865790621701921, "global_step": 207636, "epoch": 2332, "lr": 1.3073020032099336e-05, "val_loss": 8.493820190429688}
{"train_loss": 0.05860786885023117, "global_step": 207637, "epoch": 2333, "lr": 1.3072629170307687e-05}
{"train_loss": 0.047423575073480606, "global_step": 207638, "epoch": 2333, "lr": 1.3072238313480462e-05}
{"train_loss": 0.051405057311058044, "global_step": 207639, "epoch": 2333, "lr": 1.3071847461617736e-05}
{"train_loss": 0.03003772906959057, "global_step": 207640, "epoch": 2333, "lr": 1.3071456614719535e-05}
{"train_loss": 0.04904358088970184, "global_step": 207641, "epoch": 2333, "lr": 1.3071065772785939e-05}
{"train_loss": 0.023734591901302338, "global_step": 207642, "epoch": 2333, "lr": 1.3070674935816995e-05}
{"train_loss": 0.05015597864985466, "global_step": 207643, "epoch": 2333, "lr": 1.307028410381273e-05}
{"train_loss": 0.013421209529042244, "global_step": 207644, "epoch": 2333, "lr": 1.3069893276773232e-05}
{"train_loss": 0.04656321182847023, "global_step": 207645, "epoch": 2333, "lr": 1.3069502454698517e-05}
{"train_loss": 0.06081932783126831, "global_step": 207646, "epoch": 2333, "lr": 1.3069111637588677e-05}
{"train_loss": 0.08076690882444382, "global_step": 207647, "epoch": 2333, "lr": 1.3068720825443726e-05}
{"train_loss": 0.04796663671731949, "global_step": 207648, "epoch": 2333, "lr": 1.3068330018263752e-05}
{"train_loss": 0.05046086013317108, "global_step": 207649, "epoch": 2333, "lr": 1.306793921604878e-05}
{"train_loss": 0.11788696050643921, "global_step": 207650, "epoch": 2333, "lr": 1.306754841879888e-05}
{"train_loss": 0.0700160339474678, "global_step": 207651, "epoch": 2333, "lr": 1.3067157626514092e-05}
{"train_loss": 0.04968342185020447, "global_step": 207652, "epoch": 2333, "lr": 1.3066766839194483e-05}
{"train_loss": 0.04679059237241745, "global_step": 207653, "epoch": 2333, "lr": 1.3066376056840085e-05}
{"train_loss": 0.07665835320949554, "global_step": 207654, "epoch": 2333, "lr": 1.3065985279450976e-05}
{"train_loss": 0.0297467652708292, "global_step": 207655, "epoch": 2333, "lr": 1.3065594507027184e-05}
{"train_loss": 0.1016073152422905, "global_step": 207656, "epoch": 2333, "lr": 1.3065203739568783e-05}
{"train_loss": 0.05949866771697998, "global_step": 207657, "epoch": 2333, "lr": 1.3064812977075802e-05}
{"train_loss": 0.03909017890691757, "global_step": 207658, "epoch": 2333, "lr": 1.3064422219548306e-05}
{"train_loss": 0.14891476929187775, "global_step": 207659, "epoch": 2333, "lr": 1.3064031466986366e-05}
{"train_loss": 0.035469602793455124, "global_step": 207660, "epoch": 2333, "lr": 1.3063640719389997e-05}
{"train_loss": 0.020124943926930428, "global_step": 207661, "epoch": 2333, "lr": 1.306324997675929e-05}
{"train_loss": 0.09647145867347717, "global_step": 207662, "epoch": 2333, "lr": 1.3062859239094261e-05}
{"train_loss": 0.09942129254341125, "global_step": 207663, "epoch": 2333, "lr": 1.3062468506394998e-05}
{"train_loss": 0.03939703106880188, "global_step": 207664, "epoch": 2333, "lr": 1.3062077778661519e-05}
{"train_loss": 0.03407004475593567, "global_step": 207665, "epoch": 2333, "lr": 1.3061687055893912e-05}
{"train_loss": 0.09818283468484879, "global_step": 207666, "epoch": 2333, "lr": 1.3061296338092193e-05}
{"train_loss": 0.032190170139074326, "global_step": 207667, "epoch": 2333, "lr": 1.3060905625256448e-05}
{"train_loss": 0.04447532445192337, "global_step": 207668, "epoch": 2333, "lr": 1.3060514917386702e-05}
{"train_loss": 0.050274237990379333, "global_step": 207669, "epoch": 2333, "lr": 1.3060124214483028e-05}
{"train_loss": 0.09013708680868149, "global_step": 207670, "epoch": 2333, "lr": 1.3059733516545458e-05}
{"train_loss": 0.04773857071995735, "global_step": 207671, "epoch": 2333, "lr": 1.3059342823574072e-05}
{"train_loss": 0.08373849093914032, "global_step": 207672, "epoch": 2333, "lr": 1.3058952135568892e-05}
{"train_loss": 0.07915724068880081, "global_step": 207673, "epoch": 2333, "lr": 1.3058561452530006e-05}
{"train_loss": 0.05696301534771919, "global_step": 207674, "epoch": 2333, "lr": 1.3058170774457424e-05}
{"train_loss": 0.06493955850601196, "global_step": 207675, "epoch": 2333, "lr": 1.3057780101351235e-05}
{"train_loss": 0.07388686388731003, "global_step": 207676, "epoch": 2333, "lr": 1.3057389433211465e-05}
{"train_loss": 0.048778802156448364, "global_step": 207677, "epoch": 2333, "lr": 1.3056998770038193e-05}
{"train_loss": 0.044358689337968826, "global_step": 207678, "epoch": 2333, "lr": 1.3056608111831447e-05}
{"train_loss": 0.0528680793941021, "global_step": 207679, "epoch": 2333, "lr": 1.3056217458591297e-05}
{"train_loss": 0.07357487082481384, "global_step": 207680, "epoch": 2333, "lr": 1.3055826810317778e-05}
{"train_loss": 0.04929385334253311, "global_step": 207681, "epoch": 2333, "lr": 1.3055436167010966e-05}
{"train_loss": 0.09423964470624924, "global_step": 207682, "epoch": 2333, "lr": 1.30550455286709e-05}
{"train_loss": 0.03077748790383339, "global_step": 207683, "epoch": 2333, "lr": 1.305465489529762e-05}
{"train_loss": 0.059784092009067535, "global_step": 207684, "epoch": 2333, "lr": 1.3054264266891208e-05}
{"train_loss": 0.06617192178964615, "global_step": 207685, "epoch": 2333, "lr": 1.305387364345168e-05}
{"train_loss": 0.06217155233025551, "global_step": 207686, "epoch": 2333, "lr": 1.3053483024979123e-05}
{"train_loss": 0.05719120800495148, "global_step": 207687, "epoch": 2333, "lr": 1.3053092411473566e-05}
{"train_loss": 0.05860899016261101, "global_step": 207688, "epoch": 2333, "lr": 1.305270180293508e-05}
{"train_loss": 0.057278458029031754, "global_step": 207689, "epoch": 2333, "lr": 1.3052311199363697e-05}
{"train_loss": 0.05901245027780533, "global_step": 207690, "epoch": 2333, "lr": 1.3051920600759493e-05}
{"train_loss": 0.06945217400789261, "global_step": 207691, "epoch": 2333, "lr": 1.3051530007122492e-05}
{"train_loss": 0.03636360540986061, "global_step": 207692, "epoch": 2333, "lr": 1.305113941845278e-05}
{"train_loss": 0.05562847480177879, "global_step": 207693, "epoch": 2333, "lr": 1.3050748834750376e-05}
{"train_loss": 0.09540298581123352, "global_step": 207694, "epoch": 2333, "lr": 1.3050358256015349e-05}
{"train_loss": 0.05754876509308815, "global_step": 207695, "epoch": 2333, "lr": 1.304996768224776e-05}
{"train_loss": 0.02409185841679573, "global_step": 207696, "epoch": 2333, "lr": 1.3049577113447647e-05}
{"train_loss": 0.08407262712717056, "global_step": 207697, "epoch": 2333, "lr": 1.3049186549615079e-05}
{"train_loss": 0.03623669221997261, "global_step": 207698, "epoch": 2333, "lr": 1.3048795990750079e-05}
{"train_loss": 0.1274886429309845, "global_step": 207699, "epoch": 2333, "lr": 1.3048405436852734e-05}
{"train_loss": 0.09441030025482178, "global_step": 207700, "epoch": 2333, "lr": 1.304801488792307e-05}
{"train_loss": 0.05155705288052559, "global_step": 207701, "epoch": 2333, "lr": 1.3047624343961161e-05}
{"train_loss": 0.029803793877363205, "global_step": 207702, "epoch": 2333, "lr": 1.3047233804967034e-05}
{"train_loss": 0.0521785169839859, "global_step": 207703, "epoch": 2333, "lr": 1.3046843270940772e-05}
{"train_loss": 0.08694905787706375, "global_step": 207704, "epoch": 2333, "lr": 1.3046452741882397e-05}
{"train_loss": 0.026581885293126106, "global_step": 207705, "epoch": 2333, "lr": 1.3046062217791993e-05}
{"train_loss": 0.12239374965429306, "global_step": 207706, "epoch": 2333, "lr": 1.3045671698669575e-05}
{"train_loss": 0.041751086711883545, "global_step": 207707, "epoch": 2333, "lr": 1.3045281184515235e-05}
{"train_loss": 0.053162284195423126, "global_step": 207708, "epoch": 2333, "lr": 1.3044890675328991e-05}
{"train_loss": 0.04410674422979355, "global_step": 207709, "epoch": 2333, "lr": 1.304450017111093e-05}
{"train_loss": 0.1054287925362587, "global_step": 207710, "epoch": 2333, "lr": 1.3044109671861066e-05}
{"train_loss": 0.06641628593206406, "global_step": 207711, "epoch": 2333, "lr": 1.3043719177579484e-05}
{"train_loss": 0.09637348353862762, "global_step": 207712, "epoch": 2333, "lr": 1.304332868826621e-05}
{"train_loss": 0.10290037840604782, "global_step": 207713, "epoch": 2333, "lr": 1.3042938203921313e-05}
{"train_loss": 0.08174008131027222, "global_step": 207714, "epoch": 2333, "lr": 1.304254772454485e-05}
{"train_loss": 0.052693452686071396, "global_step": 207715, "epoch": 2333, "lr": 1.3042157250136877e-05}
{"train_loss": 0.045415062457323074, "global_step": 207716, "epoch": 2333, "lr": 1.3041766780697413e-05}
{"train_loss": 0.02483014389872551, "global_step": 207717, "epoch": 2333, "lr": 1.304137631622655e-05}
{"train_loss": 0.11711174994707108, "global_step": 207718, "epoch": 2333, "lr": 1.3040985856724309e-05}
{"train_loss": 0.08970771729946136, "global_step": 207719, "epoch": 2333, "lr": 1.3040595402190769e-05}
{"train_loss": 0.031911566853523254, "global_step": 207720, "epoch": 2333, "lr": 1.3040204952625972e-05}
{"train_loss": 0.07541092485189438, "global_step": 207721, "epoch": 2333, "lr": 1.3039814508029957e-05}
{"train_loss": 0.05400233715772629, "global_step": 207722, "epoch": 2333, "lr": 1.30394240684028e-05}
{"train_loss": 0.055342935025691986, "global_step": 207723, "epoch": 2333, "lr": 1.3039033633744529e-05}
{"train_loss": 0.047208644449710846, "global_step": 207724, "epoch": 2333, "lr": 1.3038643204055223e-05}
{"train_loss": 0.06159845318938239, "global_step": 207725, "epoch": 2333, "lr": 1.3038252779334908e-05, "val_loss": 8.555802345275879}
{"train_loss": 0.07195369899272919, "global_step": 207726, "epoch": 2334, "lr": 1.3037862359583664e-05}
{"train_loss": 0.038557007908821106, "global_step": 207727, "epoch": 2334, "lr": 1.303747194480151e-05}
{"train_loss": 0.03917112946510315, "global_step": 207728, "epoch": 2334, "lr": 1.3037081534988538e-05}
{"train_loss": 0.06434234231710434, "global_step": 207729, "epoch": 2334, "lr": 1.3036691130144763e-05}
{"train_loss": 0.05864515155553818, "global_step": 207730, "epoch": 2334, "lr": 1.3036300730270268e-05}
{"train_loss": 0.04882464185357094, "global_step": 207731, "epoch": 2334, "lr": 1.3035910335365076e-05}
{"train_loss": 0.058603253215551376, "global_step": 207732, "epoch": 2334, "lr": 1.3035519945429252e-05}
{"train_loss": 0.06317387521266937, "global_step": 207733, "epoch": 2334, "lr": 1.3035129560462878e-05}
{"train_loss": 0.043913308531045914, "global_step": 207734, "epoch": 2334, "lr": 1.3034739180465954e-05}
{"train_loss": 0.0795486569404602, "global_step": 207735, "epoch": 2334, "lr": 1.3034348805438579e-05}
{"train_loss": 0.07330498844385147, "global_step": 207736, "epoch": 2334, "lr": 1.3033958435380772e-05}
{"train_loss": 0.07056144624948502, "global_step": 207737, "epoch": 2334, "lr": 1.3033568070292612e-05}
{"train_loss": 0.06378757953643799, "global_step": 207738, "epoch": 2334, "lr": 1.3033177710174122e-05}
{"train_loss": 0.0704246386885643, "global_step": 207739, "epoch": 2334, "lr": 1.3032787355025384e-05}
{"train_loss": 0.010228754952549934, "global_step": 207740, "epoch": 2334, "lr": 1.3032397004846425e-05}
{"train_loss": 0.0530092678964138, "global_step": 207741, "epoch": 2334, "lr": 1.3032006659637324e-05}
{"train_loss": 0.06329528242349625, "global_step": 207742, "epoch": 2334, "lr": 1.3031616319398104e-05}
{"train_loss": 0.03765255957841873, "global_step": 207743, "epoch": 2334, "lr": 1.3031225984128848e-05}
{"train_loss": 0.08752810209989548, "global_step": 207744, "epoch": 2334, "lr": 1.3030835653829582e-05}
{"train_loss": 0.08750644326210022, "global_step": 207745, "epoch": 2334, "lr": 1.303044532850038e-05}
{"train_loss": 0.03898013383150101, "global_step": 207746, "epoch": 2334, "lr": 1.3030055008141268e-05}
{"train_loss": 0.08722054213285446, "global_step": 207747, "epoch": 2334, "lr": 1.302966469275233e-05}
{"train_loss": 0.06090667098760605, "global_step": 207748, "epoch": 2334, "lr": 1.3029274382333589e-05}
{"train_loss": 0.06645910441875458, "global_step": 207749, "epoch": 2334, "lr": 1.3028884076885112e-05}
{"train_loss": 0.0739157497882843, "global_step": 207750, "epoch": 2334, "lr": 1.3028493776406964e-05}
{"train_loss": 0.07219251245260239, "global_step": 207751, "epoch": 2334, "lr": 1.3028103480899173e-05}
{"train_loss": 0.06809955835342407, "global_step": 207752, "epoch": 2334, "lr": 1.3027713190361818e-05}
{"train_loss": 0.1345365345478058, "global_step": 207753, "epoch": 2334, "lr": 1.3027322904794931e-05}
{"train_loss": 0.05765055865049362, "global_step": 207754, "epoch": 2334, "lr": 1.3026932624198557e-05}
{"train_loss": 0.015682004392147064, "global_step": 207755, "epoch": 2334, "lr": 1.3026542348572779e-05}
{"train_loss": 0.06999950855970383, "global_step": 207756, "epoch": 2334, "lr": 1.3026152077917613e-05}
{"train_loss": 0.052785150706768036, "global_step": 207757, "epoch": 2334, "lr": 1.302576181223315e-05}
{"train_loss": 0.05938471481204033, "global_step": 207758, "epoch": 2334, "lr": 1.3025371551519416e-05}
{"train_loss": 0.057675495743751526, "global_step": 207759, "epoch": 2334, "lr": 1.3024981295776461e-05}
{"train_loss": 0.04477057233452797, "global_step": 207760, "epoch": 2334, "lr": 1.3024591045004359e-05}
{"train_loss": 0.0348505824804306, "global_step": 207761, "epoch": 2334, "lr": 1.3024200799203134e-05}
{"train_loss": 0.0662059411406517, "global_step": 207762, "epoch": 2334, "lr": 1.3023810558372873e-05}
{"train_loss": 0.04823692888021469, "global_step": 207763, "epoch": 2334, "lr": 1.3023420322513591e-05}
{"train_loss": 0.09465327113866806, "global_step": 207764, "epoch": 2334, "lr": 1.3023030091625376e-05}
{"train_loss": 0.06084461137652397, "global_step": 207765, "epoch": 2334, "lr": 1.3022639865708253e-05}
{"train_loss": 0.0489167682826519, "global_step": 207766, "epoch": 2334, "lr": 1.3022249644762297e-05}
{"train_loss": 0.09109634906053543, "global_step": 207767, "epoch": 2334, "lr": 1.302185942878753e-05}
{"train_loss": 0.08285690099000931, "global_step": 207768, "epoch": 2334, "lr": 1.3021469217784032e-05}
{"train_loss": 0.10164214670658112, "global_step": 207769, "epoch": 2334, "lr": 1.3021079011751853e-05}
{"train_loss": 0.06720093637704849, "global_step": 207770, "epoch": 2334, "lr": 1.3020688810691029e-05}
{"train_loss": 0.13368728756904602, "global_step": 207771, "epoch": 2334, "lr": 1.3020298614601639e-05}
{"train_loss": 0.10933027416467667, "global_step": 207772, "epoch": 2334, "lr": 1.3019908423483701e-05}
{"train_loss": 0.07676281780004501, "global_step": 207773, "epoch": 2334, "lr": 1.3019518237337303e-05}
{"train_loss": 0.07262296974658966, "global_step": 207774, "epoch": 2334, "lr": 1.301912805616246e-05}
{"train_loss": 0.03614698722958565, "global_step": 207775, "epoch": 2334, "lr": 1.3018737879959263e-05}
{"train_loss": 0.059107471257448196, "global_step": 207776, "epoch": 2334, "lr": 1.301834770872773e-05}
{"train_loss": 0.05215296894311905, "global_step": 207777, "epoch": 2334, "lr": 1.3017957542467945e-05}
{"train_loss": 0.049575112760066986, "global_step": 207778, "epoch": 2334, "lr": 1.3017567381179929e-05}
{"train_loss": 0.0739927887916565, "global_step": 207779, "epoch": 2334, "lr": 1.301717722486377e-05}
{"train_loss": 0.056582771241664886, "global_step": 207780, "epoch": 2334, "lr": 1.3016787073519482e-05}
{"train_loss": 0.06670268625020981, "global_step": 207781, "epoch": 2334, "lr": 1.3016396927147152e-05}
{"train_loss": 0.08391059190034866, "global_step": 207782, "epoch": 2334, "lr": 1.3016006785746798e-05}
{"train_loss": 0.05231542885303497, "global_step": 207783, "epoch": 2334, "lr": 1.3015616649318513e-05}
{"train_loss": 0.09858059883117676, "global_step": 207784, "epoch": 2334, "lr": 1.3015226517862305e-05}
{"train_loss": 0.02877933718264103, "global_step": 207785, "epoch": 2334, "lr": 1.3014836391378254e-05}
{"train_loss": 0.06736667454242706, "global_step": 207786, "epoch": 2334, "lr": 1.3014446269866421e-05}
{"train_loss": 0.13093772530555725, "global_step": 207787, "epoch": 2334, "lr": 1.3014056153326832e-05}
{"train_loss": 0.09026714414358139, "global_step": 207788, "epoch": 2334, "lr": 1.3013666041759565e-05}
{"train_loss": 0.042897067964076996, "global_step": 207789, "epoch": 2334, "lr": 1.3013275935164643e-05}
{"train_loss": 0.038765233010053635, "global_step": 207790, "epoch": 2334, "lr": 1.3012885833542155e-05}
{"train_loss": 0.08976548910140991, "global_step": 207791, "epoch": 2334, "lr": 1.3012495736892116e-05}
{"train_loss": 0.08797438442707062, "global_step": 207792, "epoch": 2334, "lr": 1.3012105645214612e-05}
{"train_loss": 0.06863048672676086, "global_step": 207793, "epoch": 2334, "lr": 1.301171555850968e-05}
{"train_loss": 0.04830152913928032, "global_step": 207794, "epoch": 2334, "lr": 1.3011325476777353e-05}
{"train_loss": 0.08541105687618256, "global_step": 207795, "epoch": 2334, "lr": 1.3010935400017722e-05}
{"train_loss": 0.06389010697603226, "global_step": 207796, "epoch": 2334, "lr": 1.30105453282308e-05}
{"train_loss": 0.05939264968037605, "global_step": 207797, "epoch": 2334, "lr": 1.3010155261416678e-05}
{"train_loss": 0.06780940294265747, "global_step": 207798, "epoch": 2334, "lr": 1.300976519957539e-05}
{"train_loss": 0.01608244702219963, "global_step": 207799, "epoch": 2334, "lr": 1.3009375142706975e-05}
{"train_loss": 0.027929456904530525, "global_step": 207800, "epoch": 2334, "lr": 1.300898509081151e-05}
{"train_loss": 0.05860741436481476, "global_step": 207801, "epoch": 2334, "lr": 1.3008595043889021e-05}
{"train_loss": 0.05257149413228035, "global_step": 207802, "epoch": 2334, "lr": 1.3008205001939594e-05}
{"train_loss": 0.04148546978831291, "global_step": 207803, "epoch": 2334, "lr": 1.3007814964963245e-05}
{"train_loss": 0.04269314929842949, "global_step": 207804, "epoch": 2334, "lr": 1.3007424932960044e-05}
{"train_loss": 0.06213727593421936, "global_step": 207805, "epoch": 2334, "lr": 1.300703490593006e-05}
{"train_loss": 0.10565191507339478, "global_step": 207806, "epoch": 2334, "lr": 1.3006644883873314e-05}
{"train_loss": 0.029308060184121132, "global_step": 207807, "epoch": 2334, "lr": 1.3006254866789891e-05}
{"train_loss": 0.06203122437000275, "global_step": 207808, "epoch": 2334, "lr": 1.3005864854679812e-05}
{"train_loss": 0.06170814856886864, "global_step": 207809, "epoch": 2334, "lr": 1.3005474847543152e-05}
{"train_loss": 0.06948500871658325, "global_step": 207810, "epoch": 2334, "lr": 1.3005084845379945e-05}
{"train_loss": 0.05003565922379494, "global_step": 207811, "epoch": 2334, "lr": 1.3004694848190268e-05}
{"train_loss": 0.0624021515250206, "global_step": 207812, "epoch": 2334, "lr": 1.3004304855974142e-05}
{"train_loss": 0.04233081638813019, "global_step": 207813, "epoch": 2334, "lr": 1.3003914868731654e-05}
{"train_loss": 0.063300785200482, "global_step": 207814, "epoch": 2334, "lr": 1.300352488646282e-05, "val_loss": 8.480648040771484}
{"train_loss": 0.05673235282301903, "global_step": 207815, "epoch": 2335, "lr": 1.300313490916773e-05}
{"train_loss": 0.08598896116018295, "global_step": 207816, "epoch": 2335, "lr": 1.30027449368464e-05}
{"train_loss": 0.05320126190781593, "global_step": 207817, "epoch": 2335, "lr": 1.3002354969498914e-05}
{"train_loss": 0.04048960283398628, "global_step": 207818, "epoch": 2335, "lr": 1.30019650071253e-05}
{"train_loss": 0.06175192818045616, "global_step": 207819, "epoch": 2335, "lr": 1.3001575049725634e-05}
{"train_loss": 0.0639808177947998, "global_step": 207820, "epoch": 2335, "lr": 1.300118509729994e-05}
{"train_loss": 0.024979474022984505, "global_step": 207821, "epoch": 2335, "lr": 1.3000795149848283e-05}
{"train_loss": 0.06323462724685669, "global_step": 207822, "epoch": 2335, "lr": 1.3000405207370735e-05}
{"train_loss": 0.08326070755720139, "global_step": 207823, "epoch": 2335, "lr": 1.3000015269867316e-05}
{"train_loss": 0.036675114184617996, "global_step": 207824, "epoch": 2335, "lr": 1.2999625337338112e-05}
{"train_loss": 0.08872183412313461, "global_step": 207825, "epoch": 2335, "lr": 1.2999235409783139e-05}
{"train_loss": 0.056328192353248596, "global_step": 207826, "epoch": 2335, "lr": 1.2998845487202489e-05}
{"train_loss": 0.08357807993888855, "global_step": 207827, "epoch": 2335, "lr": 1.2998455569596168e-05}
{"train_loss": 0.06519287824630737, "global_step": 207828, "epoch": 2335, "lr": 1.2998065656964276e-05}
{"train_loss": 0.031892526894807816, "global_step": 207829, "epoch": 2335, "lr": 1.2997675749306825e-05}
{"train_loss": 0.04893350973725319, "global_step": 207830, "epoch": 2335, "lr": 1.2997285846623902e-05}
{"train_loss": 0.0696970671415329, "global_step": 207831, "epoch": 2335, "lr": 1.2996895948915545e-05}
{"train_loss": 0.03933329880237579, "global_step": 207832, "epoch": 2335, "lr": 1.2996506056181784e-05}
{"train_loss": 0.04777150973677635, "global_step": 207833, "epoch": 2335, "lr": 1.2996116168422711e-05}
{"train_loss": 0.04364722967147827, "global_step": 207834, "epoch": 2335, "lr": 1.2995726285638344e-05}
{"train_loss": 0.08607892692089081, "global_step": 207835, "epoch": 2335, "lr": 1.2995336407828762e-05}
{"train_loss": 0.078206367790699, "global_step": 207836, "epoch": 2335, "lr": 1.2994946534994013e-05}
{"train_loss": 0.10260537266731262, "global_step": 207837, "epoch": 2335, "lr": 1.2994556667134122e-05}
{"train_loss": 0.09220349788665771, "global_step": 207838, "epoch": 2335, "lr": 1.2994166804249175e-05}
{"train_loss": 0.03185124322772026, "global_step": 207839, "epoch": 2335, "lr": 1.2993776946339199e-05}
{"train_loss": 0.05992873013019562, "global_step": 207840, "epoch": 2335, "lr": 1.2993387093404257e-05}
{"train_loss": 0.040729109197854996, "global_step": 207841, "epoch": 2335, "lr": 1.2992997245444421e-05}
{"train_loss": 0.05190848559141159, "global_step": 207842, "epoch": 2335, "lr": 1.2992607402459711e-05}
{"train_loss": 0.0860477164387703, "global_step": 207843, "epoch": 2335, "lr": 1.2992217564450204e-05}
{"train_loss": 0.026140740141272545, "global_step": 207844, "epoch": 2335, "lr": 1.2991827731415929e-05}
{"train_loss": 0.01732700876891613, "global_step": 207845, "epoch": 2335, "lr": 1.2991437903356967e-05}
{"train_loss": 0.10439682006835938, "global_step": 207846, "epoch": 2335, "lr": 1.2991048080273344e-05}
{"train_loss": 0.04920336976647377, "global_step": 207847, "epoch": 2335, "lr": 1.2990658262165135e-05}
{"train_loss": 0.06808394193649292, "global_step": 207848, "epoch": 2335, "lr": 1.2990268449032366e-05}
{"train_loss": 0.04363355413079262, "global_step": 207849, "epoch": 2335, "lr": 1.298987864087512e-05}
{"train_loss": 0.04054943472146988, "global_step": 207850, "epoch": 2335, "lr": 1.298948883769342e-05}
{"train_loss": 0.040028080344200134, "global_step": 207851, "epoch": 2335, "lr": 1.2989099039487346e-05}
{"train_loss": 0.08571231365203857, "global_step": 207852, "epoch": 2335, "lr": 1.298870924625692e-05}
{"train_loss": 0.07884976267814636, "global_step": 207853, "epoch": 2335, "lr": 1.298831945800223e-05}
{"train_loss": 0.04738469421863556, "global_step": 207854, "epoch": 2335, "lr": 1.2987929674723287e-05}
{"train_loss": 0.03189585730433464, "global_step": 207855, "epoch": 2335, "lr": 1.2987539896420186e-05}
{"train_loss": 0.03790448606014252, "global_step": 207856, "epoch": 2335, "lr": 1.2987150123092945e-05}
{"train_loss": 0.02034229226410389, "global_step": 207857, "epoch": 2335, "lr": 1.2986760354741645e-05}
{"train_loss": 0.08116070181131363, "global_step": 207858, "epoch": 2335, "lr": 1.2986370591366304e-05}
{"train_loss": 0.04728521406650543, "global_step": 207859, "epoch": 2335, "lr": 1.2985980832966999e-05}
{"train_loss": 0.03857316076755524, "global_step": 207860, "epoch": 2335, "lr": 1.298559107954379e-05}
{"train_loss": 0.07203204184770584, "global_step": 207861, "epoch": 2335, "lr": 1.2985201331096709e-05}
{"train_loss": 0.030308587476611137, "global_step": 207862, "epoch": 2335, "lr": 1.2984811587625822e-05}
{"train_loss": 0.036879125982522964, "global_step": 207863, "epoch": 2335, "lr": 1.2984421849131168e-05}
{"train_loss": 0.05288234353065491, "global_step": 207864, "epoch": 2335, "lr": 1.298403211561282e-05}
{"train_loss": 0.025353929027915, "global_step": 207865, "epoch": 2335, "lr": 1.29836423870708e-05}
{"train_loss": 0.06472788751125336, "global_step": 207866, "epoch": 2335, "lr": 1.2983252663505197e-05}
{"train_loss": 0.03919937461614609, "global_step": 207867, "epoch": 2335, "lr": 1.298286294491603e-05}
{"train_loss": 0.1194397360086441, "global_step": 207868, "epoch": 2335, "lr": 1.2982473231303377e-05}
{"train_loss": 0.05140256881713867, "global_step": 207869, "epoch": 2335, "lr": 1.2982083522667281e-05}
{"train_loss": 0.05679604038596153, "global_step": 207870, "epoch": 2335, "lr": 1.2981693819007779e-05}
{"train_loss": 0.07976718991994858, "global_step": 207871, "epoch": 2335, "lr": 1.298130412032495e-05}
{"train_loss": 0.05347627401351929, "global_step": 207872, "epoch": 2335, "lr": 1.2980914426618818e-05}
{"train_loss": 0.046577129513025284, "global_step": 207873, "epoch": 2335, "lr": 1.2980524737889472e-05}
{"train_loss": 0.04548307880759239, "global_step": 207874, "epoch": 2335, "lr": 1.2980135054136939e-05}
{"train_loss": 0.06598076969385147, "global_step": 207875, "epoch": 2335, "lr": 1.2979745375361258e-05}
{"train_loss": 0.07299599051475525, "global_step": 207876, "epoch": 2335, "lr": 1.2979355701562501e-05}
{"train_loss": 0.08212875574827194, "global_step": 207877, "epoch": 2335, "lr": 1.2978966032740736e-05}
{"train_loss": 0.008224746212363243, "global_step": 207878, "epoch": 2335, "lr": 1.2978576368895978e-05}
{"train_loss": 0.07109949737787247, "global_step": 207879, "epoch": 2335, "lr": 1.2978186710028318e-05}
{"train_loss": 0.07118362933397293, "global_step": 207880, "epoch": 2335, "lr": 1.2977797056137769e-05}
{"train_loss": 0.024997031316161156, "global_step": 207881, "epoch": 2335, "lr": 1.297740740722443e-05}
{"train_loss": 0.06307440251111984, "global_step": 207882, "epoch": 2335, "lr": 1.2977017763288302e-05}
{"train_loss": 0.09574636071920395, "global_step": 207883, "epoch": 2335, "lr": 1.2976628124329482e-05}
{"train_loss": 0.09003772586584091, "global_step": 207884, "epoch": 2335, "lr": 1.2976238490347987e-05}
{"train_loss": 0.04799214377999306, "global_step": 207885, "epoch": 2335, "lr": 1.2975848861343898e-05}
{"train_loss": 0.07364650070667267, "global_step": 207886, "epoch": 2335, "lr": 1.2975459237317245e-05}
{"train_loss": 0.0771118700504303, "global_step": 207887, "epoch": 2335, "lr": 1.2975069618268104e-05}
{"train_loss": 0.08058229088783264, "global_step": 207888, "epoch": 2335, "lr": 1.29746800041965e-05}
{"train_loss": 0.051012054085731506, "global_step": 207889, "epoch": 2335, "lr": 1.2974290395102506e-05}
{"train_loss": 0.04241275414824486, "global_step": 207890, "epoch": 2335, "lr": 1.2973900790986159e-05}
{"train_loss": 0.06596653163433075, "global_step": 207891, "epoch": 2335, "lr": 1.2973511191847532e-05}
{"train_loss": 0.08501294255256653, "global_step": 207892, "epoch": 2335, "lr": 1.2973121597686655e-05}
{"train_loss": 0.03764001280069351, "global_step": 207893, "epoch": 2335, "lr": 1.2972732008503602e-05}
{"train_loss": 0.03710632771253586, "global_step": 207894, "epoch": 2335, "lr": 1.29723424242984e-05}
{"train_loss": 0.05608329176902771, "global_step": 207895, "epoch": 2335, "lr": 1.2971952845071111e-05}
{"train_loss": 0.08077482879161835, "global_step": 207896, "epoch": 2335, "lr": 1.2971563270821812e-05}
{"train_loss": 0.032581668347120285, "global_step": 207897, "epoch": 2335, "lr": 1.297117370155052e-05}
{"train_loss": 0.07893799245357513, "global_step": 207898, "epoch": 2335, "lr": 1.2970784137257314e-05}
{"train_loss": 0.071705661714077, "global_step": 207899, "epoch": 2335, "lr": 1.2970394577942219e-05}
{"train_loss": 0.044935937970876694, "global_step": 207900, "epoch": 2335, "lr": 1.2970005023605319e-05}
{"train_loss": 0.030057527124881744, "global_step": 207901, "epoch": 2335, "lr": 1.2969615474246639e-05}
{"train_loss": 0.09826924651861191, "global_step": 207902, "epoch": 2335, "lr": 1.2969225929866257e-05}
{"train_loss": 0.05834247146764498, "global_step": 207903, "epoch": 2335, "lr": 1.2968836390464195e-05, "val_loss": 8.618486404418945, "train_action_mse_error": 8.007990837097168}
{"train_loss": 0.04820273816585541, "global_step": 207904, "epoch": 2336, "lr": 1.2968446856040533e-05}
{"train_loss": 0.0435560867190361, "global_step": 207905, "epoch": 2336, "lr": 1.29680573265953e-05}
{"train_loss": 0.06728879362344742, "global_step": 207906, "epoch": 2336, "lr": 1.296766780212858e-05}
{"train_loss": 0.0572538860142231, "global_step": 207907, "epoch": 2336, "lr": 1.2967278282640388e-05}
{"train_loss": 0.07980720698833466, "global_step": 207908, "epoch": 2336, "lr": 1.2966888768130803e-05}
{"train_loss": 0.04062492027878761, "global_step": 207909, "epoch": 2336, "lr": 1.2966499258599873e-05}
{"train_loss": 0.06164711341261864, "global_step": 207910, "epoch": 2336, "lr": 1.2966109754047634e-05}
{"train_loss": 0.0299825482070446, "global_step": 207911, "epoch": 2336, "lr": 1.2965720254474162e-05}
{"train_loss": 0.061291977763175964, "global_step": 207912, "epoch": 2336, "lr": 1.2965330759879479e-05}
{"train_loss": 0.06802399456501007, "global_step": 207913, "epoch": 2336, "lr": 1.2964941270263676e-05}
{"train_loss": 0.04540999233722687, "global_step": 207914, "epoch": 2336, "lr": 1.2964551785626772e-05}
{"train_loss": 0.06113513559103012, "global_step": 207915, "epoch": 2336, "lr": 1.296416230596884e-05}
{"train_loss": 0.040210578590631485, "global_step": 207916, "epoch": 2336, "lr": 1.2963772831289916e-05}
{"train_loss": 0.03445851057767868, "global_step": 207917, "epoch": 2336, "lr": 1.2963383361590077e-05}
{"train_loss": 0.04437590762972832, "global_step": 207918, "epoch": 2336, "lr": 1.296299389686934e-05}
{"train_loss": 0.054992493242025375, "global_step": 207919, "epoch": 2336, "lr": 1.2962604437127796e-05}
{"train_loss": 0.11442957073450089, "global_step": 207920, "epoch": 2336, "lr": 1.2962214982365462e-05}
{"train_loss": 0.06710579246282578, "global_step": 207921, "epoch": 2336, "lr": 1.2961825532582422e-05}
{"train_loss": 0.07629036158323288, "global_step": 207922, "epoch": 2336, "lr": 1.2961436087778695e-05}
{"train_loss": 0.052048806101083755, "global_step": 207923, "epoch": 2336, "lr": 1.2961046647954368e-05}
{"train_loss": 0.06563683599233627, "global_step": 207924, "epoch": 2336, "lr": 1.2960657213109461e-05}
{"train_loss": 0.056635625660419464, "global_step": 207925, "epoch": 2336, "lr": 1.2960267783244061e-05}
{"train_loss": 0.03634269908070564, "global_step": 207926, "epoch": 2336, "lr": 1.2959878358358186e-05}
{"train_loss": 0.07471401244401932, "global_step": 207927, "epoch": 2336, "lr": 1.295948893845192e-05}
{"train_loss": 0.03168568015098572, "global_step": 207928, "epoch": 2336, "lr": 1.2959099523525281e-05}
{"train_loss": 0.060731470584869385, "global_step": 207929, "epoch": 2336, "lr": 1.2958710113578354e-05}
{"train_loss": 0.0582939013838768, "global_step": 207930, "epoch": 2336, "lr": 1.2958320708611166e-05}
{"train_loss": 0.06937630474567413, "global_step": 207931, "epoch": 2336, "lr": 1.2957931308623778e-05}
{"train_loss": 0.013602588325738907, "global_step": 207932, "epoch": 2336, "lr": 1.2957541913616255e-05}
{"train_loss": 0.08904866129159927, "global_step": 207933, "epoch": 2336, "lr": 1.2957152523588628e-05}
{"train_loss": 0.10358217358589172, "global_step": 207934, "epoch": 2336, "lr": 1.2956763138540978e-05}
{"train_loss": 0.047156061977148056, "global_step": 207935, "epoch": 2336, "lr": 1.2956373758473323e-05}
{"train_loss": 0.06195422634482384, "global_step": 207936, "epoch": 2336, "lr": 1.2955984383385744e-05}
{"train_loss": 0.08097393065690994, "global_step": 207937, "epoch": 2336, "lr": 1.2955595013278271e-05}
{"train_loss": 0.042927250266075134, "global_step": 207938, "epoch": 2336, "lr": 1.2955205648150981e-05}
{"train_loss": 0.05905716493725777, "global_step": 207939, "epoch": 2336, "lr": 1.2954816288003896e-05}
{"train_loss": 0.07890977710485458, "global_step": 207940, "epoch": 2336, "lr": 1.2954426932837094e-05}
{"train_loss": 0.03742402046918869, "global_step": 207941, "epoch": 2336, "lr": 1.2954037582650608e-05}
{"train_loss": 0.07048940658569336, "global_step": 207942, "epoch": 2336, "lr": 1.2953648237444516e-05}
{"train_loss": 0.08343136310577393, "global_step": 207943, "epoch": 2336, "lr": 1.295325889721884e-05}
{"train_loss": 0.06884320080280304, "global_step": 207944, "epoch": 2336, "lr": 1.2952869561973663e-05}
{"train_loss": 0.06256357580423355, "global_step": 207945, "epoch": 2336, "lr": 1.2952480231709002e-05}
{"train_loss": 0.06203027814626694, "global_step": 207946, "epoch": 2336, "lr": 1.2952090906424941e-05}
{"train_loss": 0.02704641781747341, "global_step": 207947, "epoch": 2336, "lr": 1.2951701586121529e-05}
{"train_loss": 0.02765795588493347, "global_step": 207948, "epoch": 2336, "lr": 1.2951312270798788e-05}
{"train_loss": 0.04268902540206909, "global_step": 207949, "epoch": 2336, "lr": 1.2950922960456807e-05}
{"train_loss": 0.11536397784948349, "global_step": 207950, "epoch": 2336, "lr": 1.295053365509561e-05}
{"train_loss": 0.11495713144540787, "global_step": 207951, "epoch": 2336, "lr": 1.295014435471527e-05}
{"train_loss": 0.09051260352134705, "global_step": 207952, "epoch": 2336, "lr": 1.2949755059315826e-05}
{"train_loss": 0.031851381063461304, "global_step": 207953, "epoch": 2336, "lr": 1.2949365768897342e-05}
{"train_loss": 0.058421242982149124, "global_step": 207954, "epoch": 2336, "lr": 1.2948976483459857e-05}
{"train_loss": 0.0922112762928009, "global_step": 207955, "epoch": 2336, "lr": 1.2948587203003443e-05}
{"train_loss": 0.045685335993766785, "global_step": 207956, "epoch": 2336, "lr": 1.2948197927528122e-05}
{"train_loss": 0.06571794301271439, "global_step": 207957, "epoch": 2336, "lr": 1.2947808657033982e-05}
{"train_loss": 0.06372806429862976, "global_step": 207958, "epoch": 2336, "lr": 1.2947419391521043e-05}
{"train_loss": 0.05195778235793114, "global_step": 207959, "epoch": 2336, "lr": 1.2947030130989385e-05}
{"train_loss": 0.03323185443878174, "global_step": 207960, "epoch": 2336, "lr": 1.2946640875439038e-05}
{"train_loss": 0.05398472025990486, "global_step": 207961, "epoch": 2336, "lr": 1.294625162487007e-05}
{"train_loss": 0.07253004610538483, "global_step": 207962, "epoch": 2336, "lr": 1.2945862379282514e-05}
{"train_loss": 0.0508197657763958, "global_step": 207963, "epoch": 2336, "lr": 1.2945473138676455e-05}
{"train_loss": 0.043534498661756516, "global_step": 207964, "epoch": 2336, "lr": 1.2945083903051908e-05}
{"train_loss": 0.037121549248695374, "global_step": 207965, "epoch": 2336, "lr": 1.2944694672408952e-05}
{"train_loss": 0.05844702571630478, "global_step": 207966, "epoch": 2336, "lr": 1.2944305446747624e-05}
{"train_loss": 0.059104401618242264, "global_step": 207967, "epoch": 2336, "lr": 1.2943916226067976e-05}
{"train_loss": 0.07398319989442825, "global_step": 207968, "epoch": 2336, "lr": 1.2943527010370088e-05}
{"train_loss": 0.05094116926193237, "global_step": 207969, "epoch": 2336, "lr": 1.294313779965397e-05}
{"train_loss": 0.041552986949682236, "global_step": 207970, "epoch": 2336, "lr": 1.2942748593919713e-05}
{"train_loss": 0.05103794112801552, "global_step": 207971, "epoch": 2336, "lr": 1.2942359393167335e-05}
{"train_loss": 0.022699855268001556, "global_step": 207972, "epoch": 2336, "lr": 1.2941970197396924e-05}
{"train_loss": 0.11560580134391785, "global_step": 207973, "epoch": 2336, "lr": 1.2941581006608495e-05}
{"train_loss": 0.05842195451259613, "global_step": 207974, "epoch": 2336, "lr": 1.2941191820802134e-05}
{"train_loss": 0.04098525643348694, "global_step": 207975, "epoch": 2336, "lr": 1.2940802639977866e-05}
{"train_loss": 0.04212101176381111, "global_step": 207976, "epoch": 2336, "lr": 1.294041346413577e-05}
{"train_loss": 0.04339810460805893, "global_step": 207977, "epoch": 2336, "lr": 1.2940024293275865e-05}
{"train_loss": 0.06458855420351028, "global_step": 207978, "epoch": 2336, "lr": 1.2939635127398242e-05}
{"train_loss": 0.03717294707894325, "global_step": 207979, "epoch": 2336, "lr": 1.2939245966502917e-05}
{"train_loss": 0.04898429661989212, "global_step": 207980, "epoch": 2336, "lr": 1.2938856810589972e-05}
{"train_loss": 0.07457254081964493, "global_step": 207981, "epoch": 2336, "lr": 1.2938467659659432e-05}
{"train_loss": 0.05512245371937752, "global_step": 207982, "epoch": 2336, "lr": 1.2938078513711377e-05}
{"train_loss": 0.05013519898056984, "global_step": 207983, "epoch": 2336, "lr": 1.2937689372745832e-05}
{"train_loss": 0.03660709410905838, "global_step": 207984, "epoch": 2336, "lr": 1.2937300236762872e-05}
{"train_loss": 0.050909481942653656, "global_step": 207985, "epoch": 2336, "lr": 1.293691110576255e-05}
{"train_loss": 0.04073229059576988, "global_step": 207986, "epoch": 2336, "lr": 1.2936521979744887e-05}
{"train_loss": 0.06152743846178055, "global_step": 207987, "epoch": 2336, "lr": 1.2936132858709976e-05}
{"train_loss": 0.08548201620578766, "global_step": 207988, "epoch": 2336, "lr": 1.293574374265783e-05}
{"train_loss": 0.0566355399787426, "global_step": 207989, "epoch": 2336, "lr": 1.2935354631588542e-05}
{"train_loss": 0.03428097069263458, "global_step": 207990, "epoch": 2336, "lr": 1.2934965525502124e-05}
{"train_loss": 0.06281273812055588, "global_step": 207991, "epoch": 2336, "lr": 1.293457642439867e-05}
{"train_loss": 0.057918016181400654, "global_step": 207992, "epoch": 2336, "lr": 1.2934187328278191e-05, "val_loss": 8.597482681274414}
{"train_loss": 0.04061532020568848, "global_step": 207993, "epoch": 2337, "lr": 1.2933798237140771e-05}
{"train_loss": 0.035296350717544556, "global_step": 207994, "epoch": 2337, "lr": 1.2933409150986437e-05}
{"train_loss": 0.06727694720029831, "global_step": 207995, "epoch": 2337, "lr": 1.2933020069815272e-05}
{"train_loss": 0.11967943608760834, "global_step": 207996, "epoch": 2337, "lr": 1.2932630993627292e-05}
{"train_loss": 0.05426507070660591, "global_step": 207997, "epoch": 2337, "lr": 1.2932241922422588e-05}
{"train_loss": 0.04575316980481148, "global_step": 207998, "epoch": 2337, "lr": 1.293185285620117e-05}
{"train_loss": 0.06508681923151016, "global_step": 207999, "epoch": 2337, "lr": 1.2931463794963134e-05}
{"train_loss": 0.05779584124684334, "global_step": 208000, "epoch": 2337, "lr": 1.2931074738708493e-05}
{"train_loss": 0.006614345591515303, "global_step": 208001, "epoch": 2337, "lr": 1.2930685687437332e-05}
{"train_loss": 0.04593803361058235, "global_step": 208002, "epoch": 2337, "lr": 1.2930296641149675e-05}
{"train_loss": 0.07673438638448715, "global_step": 208003, "epoch": 2337, "lr": 1.2929907599845586e-05}
{"train_loss": 0.04174663871526718, "global_step": 208004, "epoch": 2337, "lr": 1.292951856352514e-05}
{"train_loss": 0.0758928582072258, "global_step": 208005, "epoch": 2337, "lr": 1.2929129532188345e-05}
{"train_loss": 0.023176800459623337, "global_step": 208006, "epoch": 2337, "lr": 1.2928740505835296e-05}
{"train_loss": 0.05058197304606438, "global_step": 208007, "epoch": 2337, "lr": 1.292835148446601e-05}
{"train_loss": 0.04410088062286377, "global_step": 208008, "epoch": 2337, "lr": 1.2927962468080573e-05}
{"train_loss": 0.05971707031130791, "global_step": 208009, "epoch": 2337, "lr": 1.2927573456678999e-05}
{"train_loss": 0.09106200188398361, "global_step": 208010, "epoch": 2337, "lr": 1.2927184450261381e-05}
{"train_loss": 0.05619271099567413, "global_step": 208011, "epoch": 2337, "lr": 1.2926795448827733e-05}
{"train_loss": 0.08586092293262482, "global_step": 208012, "epoch": 2337, "lr": 1.2926406452378142e-05}
{"train_loss": 0.03287870064377785, "global_step": 208013, "epoch": 2337, "lr": 1.2926017460912625e-05}
{"train_loss": 0.04982488229870796, "global_step": 208014, "epoch": 2337, "lr": 1.2925628474431272e-05}
{"train_loss": 0.09318453818559647, "global_step": 208015, "epoch": 2337, "lr": 1.2925239492934105e-05}
{"train_loss": 0.05480000749230385, "global_step": 208016, "epoch": 2337, "lr": 1.29248505164212e-05}
{"train_loss": 0.051115889102220535, "global_step": 208017, "epoch": 2337, "lr": 1.292446154489258e-05}
{"train_loss": 0.044037118554115295, "global_step": 208018, "epoch": 2337, "lr": 1.2924072578348329e-05}
{"train_loss": 0.050392452627420425, "global_step": 208019, "epoch": 2337, "lr": 1.2923683616788469e-05}
{"train_loss": 0.10699129849672318, "global_step": 208020, "epoch": 2337, "lr": 1.2923294660213087e-05}
{"train_loss": 0.12658953666687012, "global_step": 208021, "epoch": 2337, "lr": 1.2922905708622197e-05}
{"train_loss": 0.04795411601662636, "global_step": 208022, "epoch": 2337, "lr": 1.2922516762015885e-05}
{"train_loss": 0.019862744957208633, "global_step": 208023, "epoch": 2337, "lr": 1.2922127820394176e-05}
{"train_loss": 0.058012641966342926, "global_step": 208024, "epoch": 2337, "lr": 1.2921738883757151e-05}
{"train_loss": 0.06985238194465637, "global_step": 208025, "epoch": 2337, "lr": 1.2921349952104843e-05}
{"train_loss": 0.06665028631687164, "global_step": 208026, "epoch": 2337, "lr": 1.2920961025437294e-05}
{"train_loss": 0.05806902050971985, "global_step": 208027, "epoch": 2337, "lr": 1.2920572103754585e-05}
{"train_loss": 0.055003080517053604, "global_step": 208028, "epoch": 2337, "lr": 1.2920183187056739e-05}
{"train_loss": 0.08106342703104019, "global_step": 208029, "epoch": 2337, "lr": 1.2919794275343838e-05}
{"train_loss": 0.046154290437698364, "global_step": 208030, "epoch": 2337, "lr": 1.2919405368615905e-05}
{"train_loss": 0.017420925199985504, "global_step": 208031, "epoch": 2337, "lr": 1.2919016466873019e-05}
{"train_loss": 0.0572456419467926, "global_step": 208032, "epoch": 2337, "lr": 1.2918627570115204e-05}
{"train_loss": 0.03003562055528164, "global_step": 208033, "epoch": 2337, "lr": 1.2918238678342543e-05}
{"train_loss": 0.0620277114212513, "global_step": 208034, "epoch": 2337, "lr": 1.2917849791555053e-05}
{"train_loss": 0.0639030709862709, "global_step": 208035, "epoch": 2337, "lr": 1.2917460909752827e-05}
{"train_loss": 0.04968457296490669, "global_step": 208036, "epoch": 2337, "lr": 1.2917072032935879e-05}
{"train_loss": 0.10835375636816025, "global_step": 208037, "epoch": 2337, "lr": 1.2916683161104297e-05}
{"train_loss": 0.041116438806056976, "global_step": 208038, "epoch": 2337, "lr": 1.2916294294258096e-05}
{"train_loss": 0.09610587358474731, "global_step": 208039, "epoch": 2337, "lr": 1.2915905432397351e-05}
{"train_loss": 0.04671168327331543, "global_step": 208040, "epoch": 2337, "lr": 1.2915516575522119e-05}
{"train_loss": 0.07616080343723297, "global_step": 208041, "epoch": 2337, "lr": 1.2915127723632436e-05}
{"train_loss": 0.03729266673326492, "global_step": 208042, "epoch": 2337, "lr": 1.2914738876728371e-05}
{"train_loss": 0.03746059164404869, "global_step": 208043, "epoch": 2337, "lr": 1.2914350034809959e-05}
{"train_loss": 0.07375046610832214, "global_step": 208044, "epoch": 2337, "lr": 1.2913961197877273e-05}
{"train_loss": 0.07948707044124603, "global_step": 208045, "epoch": 2337, "lr": 1.2913572365930337e-05}
{"train_loss": 0.02882613055408001, "global_step": 208046, "epoch": 2337, "lr": 1.2913183538969232e-05}
{"train_loss": 0.06579512357711792, "global_step": 208047, "epoch": 2337, "lr": 1.2912794716993987e-05}
{"train_loss": 0.0714116245508194, "global_step": 208048, "epoch": 2337, "lr": 1.2912405900004676e-05}
{"train_loss": 0.06478045135736465, "global_step": 208049, "epoch": 2337, "lr": 1.2912017088001327e-05}
{"train_loss": 0.09640755504369736, "global_step": 208050, "epoch": 2337, "lr": 1.2911628280984017e-05}
{"train_loss": 0.13726042211055756, "global_step": 208051, "epoch": 2337, "lr": 1.2911239478952775e-05}
{"train_loss": 0.06237545609474182, "global_step": 208052, "epoch": 2337, "lr": 1.2910850681907678e-05}
{"train_loss": 0.03815261647105217, "global_step": 208053, "epoch": 2337, "lr": 1.2910461889848751e-05}
{"train_loss": 0.05107719823718071, "global_step": 208054, "epoch": 2337, "lr": 1.2910073102776077e-05}
{"train_loss": 0.06165790557861328, "global_step": 208055, "epoch": 2337, "lr": 1.2909684320689675e-05}
{"train_loss": 0.0817684531211853, "global_step": 208056, "epoch": 2337, "lr": 1.2909295543589628e-05}
{"train_loss": 0.07454972714185715, "global_step": 208057, "epoch": 2337, "lr": 1.2908906771475959e-05}
{"train_loss": 0.09047164022922516, "global_step": 208058, "epoch": 2337, "lr": 1.2908518004348751e-05}
{"train_loss": 0.04188554733991623, "global_step": 208059, "epoch": 2337, "lr": 1.2908129242208028e-05}
{"train_loss": 0.05639905855059624, "global_step": 208060, "epoch": 2337, "lr": 1.2907740485053866e-05}
{"train_loss": 0.07982437312602997, "global_step": 208061, "epoch": 2337, "lr": 1.2907351732886296e-05}
{"train_loss": 0.11087038367986679, "global_step": 208062, "epoch": 2337, "lr": 1.290696298570539e-05}
{"train_loss": 0.016582997515797615, "global_step": 208063, "epoch": 2337, "lr": 1.2906574243511193e-05}
{"train_loss": 0.04650698974728584, "global_step": 208064, "epoch": 2337, "lr": 1.2906185506303742e-05}
{"train_loss": 0.06347939372062683, "global_step": 208065, "epoch": 2337, "lr": 1.2905796774083117e-05}
{"train_loss": 0.0600413978099823, "global_step": 208066, "epoch": 2337, "lr": 1.2905408046849338e-05}
{"train_loss": 0.053349900990724564, "global_step": 208067, "epoch": 2337, "lr": 1.2905019324602491e-05}
{"train_loss": 0.025861196219921112, "global_step": 208068, "epoch": 2337, "lr": 1.29046306073426e-05}
{"train_loss": 0.06224548816680908, "global_step": 208069, "epoch": 2337, "lr": 1.290424189506974e-05}
{"train_loss": 0.11204034835100174, "global_step": 208070, "epoch": 2337, "lr": 1.2903853187783938e-05}
{"train_loss": 0.024821097031235695, "global_step": 208071, "epoch": 2337, "lr": 1.2903464485485273e-05}
{"train_loss": 0.12148603051900864, "global_step": 208072, "epoch": 2337, "lr": 1.2903075788173775e-05}
{"train_loss": 0.09163719415664673, "global_step": 208073, "epoch": 2337, "lr": 1.2902687095849519e-05}
{"train_loss": 0.07150363177061081, "global_step": 208074, "epoch": 2337, "lr": 1.2902298408512531e-05}
{"train_loss": 0.08893553167581558, "global_step": 208075, "epoch": 2337, "lr": 1.2901909726162876e-05}
{"train_loss": 0.05058346688747406, "global_step": 208076, "epoch": 2337, "lr": 1.2901521048800625e-05}
{"train_loss": 0.025188874453306198, "global_step": 208077, "epoch": 2337, "lr": 1.2901132376425796e-05}
{"train_loss": 0.09382876008749008, "global_step": 208078, "epoch": 2337, "lr": 1.290074370903847e-05}
{"train_loss": 0.04159841313958168, "global_step": 208079, "epoch": 2337, "lr": 1.2900355046638674e-05}
{"train_loss": 0.11462654173374176, "global_step": 208080, "epoch": 2337, "lr": 1.2899966389226487e-05}
{"train_loss": 0.06263143225478825, "global_step": 208081, "epoch": 2337, "lr": 1.2899577736801933e-05, "val_loss": 8.477402687072754}
{"train_loss": 0.08940953761339188, "global_step": 208082, "epoch": 2338, "lr": 1.2899189089365093e-05}
{"train_loss": 0.04972551017999649, "global_step": 208083, "epoch": 2338, "lr": 1.2898800446915987e-05}
{"train_loss": 0.025149323046207428, "global_step": 208084, "epoch": 2338, "lr": 1.2898411809454702e-05}
{"train_loss": 0.04408863186836243, "global_step": 208085, "epoch": 2338, "lr": 1.2898023176981262e-05}
{"train_loss": 0.0428994745016098, "global_step": 208086, "epoch": 2338, "lr": 1.2897634549495741e-05}
{"train_loss": 0.033134885132312775, "global_step": 208087, "epoch": 2338, "lr": 1.2897245926998169e-05}
{"train_loss": 0.10073407739400864, "global_step": 208088, "epoch": 2338, "lr": 1.2896857309488625e-05}
{"train_loss": 0.05355380102992058, "global_step": 208089, "epoch": 2338, "lr": 1.2896468696967129e-05}
{"train_loss": 0.05267202481627464, "global_step": 208090, "epoch": 2338, "lr": 1.2896080089433766e-05}
{"train_loss": 0.07714661955833435, "global_step": 208091, "epoch": 2338, "lr": 1.2895691486888561e-05}
{"train_loss": 0.10137028247117996, "global_step": 208092, "epoch": 2338, "lr": 1.289530288933159e-05}
{"train_loss": 0.0999482050538063, "global_step": 208093, "epoch": 2338, "lr": 1.2894914296762878e-05}
{"train_loss": 0.11219800263643265, "global_step": 208094, "epoch": 2338, "lr": 1.2894525709182492e-05}
{"train_loss": 0.0694466084241867, "global_step": 208095, "epoch": 2338, "lr": 1.2894137126590499e-05}
{"train_loss": 0.07771191745996475, "global_step": 208096, "epoch": 2338, "lr": 1.2893748548986945e-05}
{"train_loss": 0.061261024326086044, "global_step": 208097, "epoch": 2338, "lr": 1.289335997637185e-05}
{"train_loss": 0.04540461301803589, "global_step": 208098, "epoch": 2338, "lr": 1.2892971408745307e-05}
{"train_loss": 0.05736536532640457, "global_step": 208099, "epoch": 2338, "lr": 1.2892582846107343e-05}
{"train_loss": 0.04096328094601631, "global_step": 208100, "epoch": 2338, "lr": 1.2892194288458026e-05}
{"train_loss": 0.10280019789934158, "global_step": 208101, "epoch": 2338, "lr": 1.2891805735797407e-05}
{"train_loss": 0.038380302488803864, "global_step": 208102, "epoch": 2338, "lr": 1.2891417188125516e-05}
{"train_loss": 0.09556588530540466, "global_step": 208103, "epoch": 2338, "lr": 1.2891028645442432e-05}
{"train_loss": 0.03384842351078987, "global_step": 208104, "epoch": 2338, "lr": 1.2890640107748186e-05}
{"train_loss": 0.07327218353748322, "global_step": 208105, "epoch": 2338, "lr": 1.2890251575042855e-05}
{"train_loss": 0.03502167761325836, "global_step": 208106, "epoch": 2338, "lr": 1.2889863047326467e-05}
{"train_loss": 0.037189170718193054, "global_step": 208107, "epoch": 2338, "lr": 1.2889474524599093e-05}
{"train_loss": 0.04928481578826904, "global_step": 208108, "epoch": 2338, "lr": 1.2889086006860767e-05}
{"train_loss": 0.06447836011648178, "global_step": 208109, "epoch": 2338, "lr": 1.2888697494111563e-05}
{"train_loss": 0.05860956758260727, "global_step": 208110, "epoch": 2338, "lr": 1.2888308986351505e-05}
{"train_loss": 0.04504729062318802, "global_step": 208111, "epoch": 2338, "lr": 1.2887920483580678e-05}
{"train_loss": 0.0795280933380127, "global_step": 208112, "epoch": 2338, "lr": 1.2887531985799105e-05}
{"train_loss": 0.02879341132938862, "global_step": 208113, "epoch": 2338, "lr": 1.2887143493006848e-05}
{"train_loss": 0.12787364423274994, "global_step": 208114, "epoch": 2338, "lr": 1.2886755005203983e-05}
{"train_loss": 0.07612729072570801, "global_step": 208115, "epoch": 2338, "lr": 1.2886366522390519e-05}
{"train_loss": 0.05347588285803795, "global_step": 208116, "epoch": 2338, "lr": 1.2885978044566549e-05}
{"train_loss": 0.08007129281759262, "global_step": 208117, "epoch": 2338, "lr": 1.2885589571732093e-05}
{"train_loss": 0.08690926432609558, "global_step": 208118, "epoch": 2338, "lr": 1.2885201103887228e-05}
{"train_loss": 0.034064989537000656, "global_step": 208119, "epoch": 2338, "lr": 1.2884812641031985e-05}
{"train_loss": 0.08797746896743774, "global_step": 208120, "epoch": 2338, "lr": 1.2884424183166439e-05}
{"train_loss": 0.11204074323177338, "global_step": 208121, "epoch": 2338, "lr": 1.2884035730290617e-05}
{"train_loss": 0.06109244376420975, "global_step": 208122, "epoch": 2338, "lr": 1.2883647282404598e-05}
{"train_loss": 0.05152090638875961, "global_step": 208123, "epoch": 2338, "lr": 1.2883258839508406e-05}
{"train_loss": 0.07128199189901352, "global_step": 208124, "epoch": 2338, "lr": 1.2882870401602121e-05}
{"train_loss": 0.0291291493922472, "global_step": 208125, "epoch": 2338, "lr": 1.2882481968685761e-05}
{"train_loss": 0.03419291228055954, "global_step": 208126, "epoch": 2338, "lr": 1.288209354075942e-05}
{"train_loss": 0.04082551971077919, "global_step": 208127, "epoch": 2338, "lr": 1.2881705117823117e-05}
{"train_loss": 0.04104279726743698, "global_step": 208128, "epoch": 2338, "lr": 1.2881316699876928e-05}
{"train_loss": 0.07020207494497299, "global_step": 208129, "epoch": 2338, "lr": 1.288092828692088e-05}
{"train_loss": 0.06249649077653885, "global_step": 208130, "epoch": 2338, "lr": 1.2880539878955034e-05}
{"train_loss": 0.03842499107122421, "global_step": 208131, "epoch": 2338, "lr": 1.2880151475979468e-05}
{"train_loss": 0.071351058781147, "global_step": 208132, "epoch": 2338, "lr": 1.2879763077994194e-05}
{"train_loss": 0.04475900158286095, "global_step": 208133, "epoch": 2338, "lr": 1.2879374684999301e-05}
{"train_loss": 0.0708998441696167, "global_step": 208134, "epoch": 2338, "lr": 1.2878986296994822e-05}
{"train_loss": 0.10144183039665222, "global_step": 208135, "epoch": 2338, "lr": 1.2878597913980794e-05}
{"train_loss": 0.07063804566860199, "global_step": 208136, "epoch": 2338, "lr": 1.2878209535957308e-05}
{"train_loss": 0.06567555665969849, "global_step": 208137, "epoch": 2338, "lr": 1.2877821162924375e-05}
{"train_loss": 0.05050809308886528, "global_step": 208138, "epoch": 2338, "lr": 1.2877432794882078e-05}
{"train_loss": 0.07331439107656479, "global_step": 208139, "epoch": 2338, "lr": 1.2877044431830449e-05}
{"train_loss": 0.07643742114305496, "global_step": 208140, "epoch": 2338, "lr": 1.2876656073769561e-05}
{"train_loss": 0.07742013037204742, "global_step": 208141, "epoch": 2338, "lr": 1.2876267720699452e-05}
{"train_loss": 0.06229792535305023, "global_step": 208142, "epoch": 2338, "lr": 1.2875879372620164e-05}
{"train_loss": 0.04202280566096306, "global_step": 208143, "epoch": 2338, "lr": 1.2875491029531778e-05}
{"train_loss": 0.05143989622592926, "global_step": 208144, "epoch": 2338, "lr": 1.287510269143431e-05}
{"train_loss": 0.07192795723676682, "global_step": 208145, "epoch": 2338, "lr": 1.287471435832785e-05}
{"train_loss": 0.05329427123069763, "global_step": 208146, "epoch": 2338, "lr": 1.287432603021242e-05}
{"train_loss": 0.06089404225349426, "global_step": 208147, "epoch": 2338, "lr": 1.2873937707088097e-05}
{"train_loss": 0.04808121919631958, "global_step": 208148, "epoch": 2338, "lr": 1.2873549388954902e-05}
{"train_loss": 0.06582501530647278, "global_step": 208149, "epoch": 2338, "lr": 1.2873161075812907e-05}
{"train_loss": 0.08429571241140366, "global_step": 208150, "epoch": 2338, "lr": 1.2872772767662177e-05}
{"train_loss": 0.07289017736911774, "global_step": 208151, "epoch": 2338, "lr": 1.287238446450274e-05}
{"train_loss": 0.02787434682250023, "global_step": 208152, "epoch": 2338, "lr": 1.287199616633467e-05}
{"train_loss": 0.08677329868078232, "global_step": 208153, "epoch": 2338, "lr": 1.2871607873157998e-05}
{"train_loss": 0.05327282473444939, "global_step": 208154, "epoch": 2338, "lr": 1.2871219584972793e-05}
{"train_loss": 0.04132834076881409, "global_step": 208155, "epoch": 2338, "lr": 1.2870831301779091e-05}
{"train_loss": 0.08135129511356354, "global_step": 208156, "epoch": 2338, "lr": 1.2870443023576966e-05}
{"train_loss": 0.08797208219766617, "global_step": 208157, "epoch": 2338, "lr": 1.2870054750366439e-05}
{"train_loss": 0.06625297665596008, "global_step": 208158, "epoch": 2338, "lr": 1.2869666482147601e-05}
{"train_loss": 0.038792166858911514, "global_step": 208159, "epoch": 2338, "lr": 1.2869278218920467e-05}
{"train_loss": 0.05577971786260605, "global_step": 208160, "epoch": 2338, "lr": 1.286888996068512e-05}
{"train_loss": 0.05695117264986038, "global_step": 208161, "epoch": 2338, "lr": 1.2868501707441582e-05}
{"train_loss": 0.06464697420597076, "global_step": 208162, "epoch": 2338, "lr": 1.286811345918994e-05}
{"train_loss": 0.02413475513458252, "global_step": 208163, "epoch": 2338, "lr": 1.286772521593021e-05}
{"train_loss": 0.054880835115909576, "global_step": 208164, "epoch": 2338, "lr": 1.2867336977662475e-05}
{"train_loss": 0.041292138397693634, "global_step": 208165, "epoch": 2338, "lr": 1.2866948744386764e-05}
{"train_loss": 0.0436195433139801, "global_step": 208166, "epoch": 2338, "lr": 1.2866560516103138e-05}
{"train_loss": 0.08635394275188446, "global_step": 208167, "epoch": 2338, "lr": 1.2866172292811662e-05}
{"train_loss": 0.06959119439125061, "global_step": 208168, "epoch": 2338, "lr": 1.2865784074512361e-05}
{"train_loss": 0.10957701504230499, "global_step": 208169, "epoch": 2338, "lr": 1.286539586120532e-05}
{"train_loss": 0.06322473701968621, "global_step": 208170, "epoch": 2338, "lr": 1.286500765289056e-05, "val_loss": 8.448151588439941}
{"train_loss": 0.030390240252017975, "global_step": 208171, "epoch": 2339, "lr": 1.286461944956816e-05}
{"train_loss": 0.04352188855409622, "global_step": 208172, "epoch": 2339, "lr": 1.2864231251238145e-05}
{"train_loss": 0.06118083372712135, "global_step": 208173, "epoch": 2339, "lr": 1.2863843057900599e-05}
{"train_loss": 0.08084976673126221, "global_step": 208174, "epoch": 2339, "lr": 1.2863454869555552e-05}
{"train_loss": 0.05219775438308716, "global_step": 208175, "epoch": 2339, "lr": 1.2863066686203052e-05}
{"train_loss": 0.02904639206826687, "global_step": 208176, "epoch": 2339, "lr": 1.2862678507843169e-05}
{"train_loss": 0.08249714970588684, "global_step": 208177, "epoch": 2339, "lr": 1.2862290334475935e-05}
{"train_loss": 0.029184797778725624, "global_step": 208178, "epoch": 2339, "lr": 1.2861902166101425e-05}
{"train_loss": 0.026646429672837257, "global_step": 208179, "epoch": 2339, "lr": 1.2861514002719683e-05}
{"train_loss": 0.0318206250667572, "global_step": 208180, "epoch": 2339, "lr": 1.286112584433074e-05}
{"train_loss": 0.055268578231334686, "global_step": 208181, "epoch": 2339, "lr": 1.2860737690934682e-05}
{"train_loss": 0.09010317176580429, "global_step": 208182, "epoch": 2339, "lr": 1.2860349542531535e-05}
{"train_loss": 0.03440668806433678, "global_step": 208183, "epoch": 2339, "lr": 1.2859961399121368e-05}
{"train_loss": 0.08626087009906769, "global_step": 208184, "epoch": 2339, "lr": 1.2859573260704216e-05}
{"train_loss": 0.06632775813341141, "global_step": 208185, "epoch": 2339, "lr": 1.2859185127280144e-05}
{"train_loss": 0.06810815632343292, "global_step": 208186, "epoch": 2339, "lr": 1.2858796998849216e-05}
{"train_loss": 0.06185804679989815, "global_step": 208187, "epoch": 2339, "lr": 1.2858408875411448e-05}
{"train_loss": 0.08195802569389343, "global_step": 208188, "epoch": 2339, "lr": 1.2858020756966937e-05}
{"train_loss": 0.020124996080994606, "global_step": 208189, "epoch": 2339, "lr": 1.2857632643515693e-05}
{"train_loss": 0.055300138890743256, "global_step": 208190, "epoch": 2339, "lr": 1.2857244535057806e-05}
{"train_loss": 0.06982023268938065, "global_step": 208191, "epoch": 2339, "lr": 1.2856856431593295e-05}
{"train_loss": 0.07117892056703568, "global_step": 208192, "epoch": 2339, "lr": 1.2856468333122234e-05}
{"train_loss": 0.06039898842573166, "global_step": 208193, "epoch": 2339, "lr": 1.2856080239644663e-05}
{"train_loss": 0.02653796412050724, "global_step": 208194, "epoch": 2339, "lr": 1.2855692151160647e-05}
{"train_loss": 0.04513734579086304, "global_step": 208195, "epoch": 2339, "lr": 1.285530406767022e-05}
{"train_loss": 0.045973967760801315, "global_step": 208196, "epoch": 2339, "lr": 1.285491598917346e-05}
{"train_loss": 0.04279034957289696, "global_step": 208197, "epoch": 2339, "lr": 1.285452791567039e-05}
{"train_loss": 0.05315912514925003, "global_step": 208198, "epoch": 2339, "lr": 1.2854139847161084e-05}
{"train_loss": 0.11909986287355423, "global_step": 208199, "epoch": 2339, "lr": 1.2853751783645579e-05}
{"train_loss": 0.04424242302775383, "global_step": 208200, "epoch": 2339, "lr": 1.2853363725123946e-05}
{"train_loss": 0.04460699111223221, "global_step": 208201, "epoch": 2339, "lr": 1.2852975671596207e-05}
{"train_loss": 0.07486709952354431, "global_step": 208202, "epoch": 2339, "lr": 1.2852587623062439e-05}
{"train_loss": 0.05633250251412392, "global_step": 208203, "epoch": 2339, "lr": 1.28521995795227e-05}
{"train_loss": 0.04488338902592659, "global_step": 208204, "epoch": 2339, "lr": 1.2851811540977016e-05}
{"train_loss": 0.041715361177921295, "global_step": 208205, "epoch": 2339, "lr": 1.2851423507425465e-05}
{"train_loss": 0.06377765536308289, "global_step": 208206, "epoch": 2339, "lr": 1.2851035478868073e-05}
{"train_loss": 0.05006512627005577, "global_step": 208207, "epoch": 2339, "lr": 1.2850647455304921e-05}
{"train_loss": 0.04022368788719177, "global_step": 208208, "epoch": 2339, "lr": 1.2850259436736035e-05}
{"train_loss": 0.05032980814576149, "global_step": 208209, "epoch": 2339, "lr": 1.2849871423161492e-05}
{"train_loss": 0.06307883560657501, "global_step": 208210, "epoch": 2339, "lr": 1.2849483414581314e-05}
{"train_loss": 0.06508123874664307, "global_step": 208211, "epoch": 2339, "lr": 1.2849095410995594e-05}
{"train_loss": 0.0558289960026741, "global_step": 208212, "epoch": 2339, "lr": 1.2848707412404349e-05}
{"train_loss": 0.04187720641493797, "global_step": 208213, "epoch": 2339, "lr": 1.284831941880763e-05}
{"train_loss": 0.05482659116387367, "global_step": 208214, "epoch": 2339, "lr": 1.2847931430205523e-05}
{"train_loss": 0.034877970814704895, "global_step": 208215, "epoch": 2339, "lr": 1.2847543446598038e-05}
{"train_loss": 0.08076314628124237, "global_step": 208216, "epoch": 2339, "lr": 1.2847155467985266e-05}
{"train_loss": 0.038994546979665756, "global_step": 208217, "epoch": 2339, "lr": 1.284676749436724e-05}
{"train_loss": 0.031060004606842995, "global_step": 208218, "epoch": 2339, "lr": 1.2846379525743996e-05}
{"train_loss": 0.04050325229763985, "global_step": 208219, "epoch": 2339, "lr": 1.2845991562115622e-05}
{"train_loss": 0.06371179223060608, "global_step": 208220, "epoch": 2339, "lr": 1.2845603603482136e-05}
{"train_loss": 0.1289619654417038, "global_step": 208221, "epoch": 2339, "lr": 1.28452156498436e-05}
{"train_loss": 0.027497392147779465, "global_step": 208222, "epoch": 2339, "lr": 1.2844827701200096e-05}
{"train_loss": 0.0474960021674633, "global_step": 208223, "epoch": 2339, "lr": 1.284443975755163e-05}
{"train_loss": 0.07987794280052185, "global_step": 208224, "epoch": 2339, "lr": 1.2844051818898294e-05}
{"train_loss": 0.08783114701509476, "global_step": 208225, "epoch": 2339, "lr": 1.2843663885240109e-05}
{"train_loss": 0.06024400144815445, "global_step": 208226, "epoch": 2339, "lr": 1.2843275956577155e-05}
{"train_loss": 0.04073233902454376, "global_step": 208227, "epoch": 2339, "lr": 1.284288803290945e-05}
{"train_loss": 0.07478684186935425, "global_step": 208228, "epoch": 2339, "lr": 1.2842500114237088e-05}
{"train_loss": 0.06734809279441833, "global_step": 208229, "epoch": 2339, "lr": 1.284211220056008e-05}
{"train_loss": 0.04753490909934044, "global_step": 208230, "epoch": 2339, "lr": 1.2841724291878509e-05}
{"train_loss": 0.08877750486135483, "global_step": 208231, "epoch": 2339, "lr": 1.2841336388192405e-05}
{"train_loss": 0.03893115371465683, "global_step": 208232, "epoch": 2339, "lr": 1.2840948489501847e-05}
{"train_loss": 0.034299906343221664, "global_step": 208233, "epoch": 2339, "lr": 1.284056059580685e-05}
{"train_loss": 0.08394721150398254, "global_step": 208234, "epoch": 2339, "lr": 1.2840172707107507e-05}
{"train_loss": 0.0769919604063034, "global_step": 208235, "epoch": 2339, "lr": 1.2839784823403828e-05}
{"train_loss": 0.09509950131177902, "global_step": 208236, "epoch": 2339, "lr": 1.2839396944695909e-05}
{"train_loss": 0.08581037819385529, "global_step": 208237, "epoch": 2339, "lr": 1.283900907098376e-05}
{"train_loss": 0.0818379819393158, "global_step": 208238, "epoch": 2339, "lr": 1.283862120226747e-05}
{"train_loss": 0.057772208005189896, "global_step": 208239, "epoch": 2339, "lr": 1.2838233338547062e-05}
{"train_loss": 0.040703557431697845, "global_step": 208240, "epoch": 2339, "lr": 1.2837845479822602e-05}
{"train_loss": 0.11701967567205429, "global_step": 208241, "epoch": 2339, "lr": 1.2837457626094152e-05}
{"train_loss": 0.08622942864894867, "global_step": 208242, "epoch": 2339, "lr": 1.2837069777361738e-05}
{"train_loss": 0.04640841856598854, "global_step": 208243, "epoch": 2339, "lr": 1.2836681933625439e-05}
{"train_loss": 0.06526535749435425, "global_step": 208244, "epoch": 2339, "lr": 1.2836294094885288e-05}
{"train_loss": 0.08497560024261475, "global_step": 208245, "epoch": 2339, "lr": 1.283590626114135e-05}
{"train_loss": 0.03117230348289013, "global_step": 208246, "epoch": 2339, "lr": 1.2835518432393662e-05}
{"train_loss": 0.057154569774866104, "global_step": 208247, "epoch": 2339, "lr": 1.2835130608642304e-05}
{"train_loss": 0.05816403776407242, "global_step": 208248, "epoch": 2339, "lr": 1.283474278988729e-05}
{"train_loss": 0.0930297002196312, "global_step": 208249, "epoch": 2339, "lr": 1.2834354976128705e-05}
{"train_loss": 0.09562239050865173, "global_step": 208250, "epoch": 2339, "lr": 1.283396716736658e-05}
{"train_loss": 0.04643501713871956, "global_step": 208251, "epoch": 2339, "lr": 1.2833579363600984e-05}
{"train_loss": 0.0487358495593071, "global_step": 208252, "epoch": 2339, "lr": 1.2833191564831959e-05}
{"train_loss": 0.11502261459827423, "global_step": 208253, "epoch": 2339, "lr": 1.2832803771059549e-05}
{"train_loss": 0.05377776175737381, "global_step": 208254, "epoch": 2339, "lr": 1.2832415982283825e-05}
{"train_loss": 0.058438435196876526, "global_step": 208255, "epoch": 2339, "lr": 1.2832028198504819e-05}
{"train_loss": 0.04133370891213417, "global_step": 208256, "epoch": 2339, "lr": 1.2831640419722607e-05}
{"train_loss": 0.05464133620262146, "global_step": 208257, "epoch": 2339, "lr": 1.2831252645937214e-05}
{"train_loss": 0.1195637658238411, "global_step": 208258, "epoch": 2339, "lr": 1.2830864877148718e-05}
{"train_loss": 0.06104584467293841, "global_step": 208259, "epoch": 2339, "lr": 1.2830477113357148e-05, "val_loss": 8.464517593383789}
{"train_loss": 0.07213321328163147, "global_step": 208260, "epoch": 2340, "lr": 1.283008935456258e-05}
{"train_loss": 0.05832206457853317, "global_step": 208261, "epoch": 2340, "lr": 1.2829701600765038e-05}
{"train_loss": 0.08947183191776276, "global_step": 208262, "epoch": 2340, "lr": 1.2829313851964603e-05}
{"train_loss": 0.08162452280521393, "global_step": 208263, "epoch": 2340, "lr": 1.2828926108161304e-05}
{"train_loss": 0.05133943632245064, "global_step": 208264, "epoch": 2340, "lr": 1.2828538369355215e-05}
{"train_loss": 0.07221836596727371, "global_step": 208265, "epoch": 2340, "lr": 1.2828150635546359e-05}
{"train_loss": 0.13468098640441895, "global_step": 208266, "epoch": 2340, "lr": 1.2827762906734825e-05}
{"train_loss": 0.06027328968048096, "global_step": 208267, "epoch": 2340, "lr": 1.2827375182920626e-05}
{"train_loss": 0.02897617034614086, "global_step": 208268, "epoch": 2340, "lr": 1.2826987464103852e-05}
{"train_loss": 0.06805527955293655, "global_step": 208269, "epoch": 2340, "lr": 1.2826599750284518e-05}
{"train_loss": 0.08401757478713989, "global_step": 208270, "epoch": 2340, "lr": 1.282621204146271e-05}
{"train_loss": 0.10684332996606827, "global_step": 208271, "epoch": 2340, "lr": 1.2825824337638453e-05}
{"train_loss": 0.03658312186598778, "global_step": 208272, "epoch": 2340, "lr": 1.2825436638811821e-05}
{"train_loss": 0.05235354229807854, "global_step": 208273, "epoch": 2340, "lr": 1.2825048944982843e-05}
{"train_loss": 0.037694841623306274, "global_step": 208274, "epoch": 2340, "lr": 1.2824661256151599e-05}
{"train_loss": 0.07879625260829926, "global_step": 208275, "epoch": 2340, "lr": 1.2824273572318112e-05}
{"train_loss": 0.04867471382021904, "global_step": 208276, "epoch": 2340, "lr": 1.2823885893482452e-05}
{"train_loss": 0.057690560817718506, "global_step": 208277, "epoch": 2340, "lr": 1.2823498219644676e-05}
{"train_loss": 0.034782491624355316, "global_step": 208278, "epoch": 2340, "lr": 1.2823110550804812e-05}
{"train_loss": 0.04977979511022568, "global_step": 208279, "epoch": 2340, "lr": 1.2822722886962946e-05}
{"train_loss": 0.05051971226930618, "global_step": 208280, "epoch": 2340, "lr": 1.2822335228119098e-05}
{"train_loss": 0.07220260053873062, "global_step": 208281, "epoch": 2340, "lr": 1.2821947574273346e-05}
{"train_loss": 0.05668247118592262, "global_step": 208282, "epoch": 2340, "lr": 1.2821559925425718e-05}
{"train_loss": 0.05476947873830795, "global_step": 208283, "epoch": 2340, "lr": 1.2821172281576293e-05}
{"train_loss": 0.08125629276037216, "global_step": 208284, "epoch": 2340, "lr": 1.2820784642725093e-05}
{"train_loss": 0.10938520729541779, "global_step": 208285, "epoch": 2340, "lr": 1.2820397008872198e-05}
{"train_loss": 0.0838010162115097, "global_step": 208286, "epoch": 2340, "lr": 1.2820009380017634e-05}
{"train_loss": 0.03413565456867218, "global_step": 208287, "epoch": 2340, "lr": 1.2819621756161476e-05}
{"train_loss": 0.06051173433661461, "global_step": 208288, "epoch": 2340, "lr": 1.281923413730376e-05}
{"train_loss": 0.08470014482736588, "global_step": 208289, "epoch": 2340, "lr": 1.2818846523444556e-05}
{"train_loss": 0.07003587484359741, "global_step": 208290, "epoch": 2340, "lr": 1.2818458914583902e-05}
{"train_loss": 0.04609784856438637, "global_step": 208291, "epoch": 2340, "lr": 1.2818071310721847e-05}
{"train_loss": 0.02819579280912876, "global_step": 208292, "epoch": 2340, "lr": 1.2817683711858459e-05}
{"train_loss": 0.08232532441616058, "global_step": 208293, "epoch": 2340, "lr": 1.281729611799376e-05}
{"train_loss": 0.07766306400299072, "global_step": 208294, "epoch": 2340, "lr": 1.2816908529127841e-05}
{"train_loss": 0.07552073895931244, "global_step": 208295, "epoch": 2340, "lr": 1.2816520945260724e-05}
{"train_loss": 0.05750715360045433, "global_step": 208296, "epoch": 2340, "lr": 1.2816133366392485e-05}
{"train_loss": 0.05711287632584572, "global_step": 208297, "epoch": 2340, "lr": 1.2815745792523148e-05}
{"train_loss": 0.02123217098414898, "global_step": 208298, "epoch": 2340, "lr": 1.2815358223652796e-05}
{"train_loss": 0.033497367054224014, "global_step": 208299, "epoch": 2340, "lr": 1.281497065978145e-05}
{"train_loss": 0.03183211386203766, "global_step": 208300, "epoch": 2340, "lr": 1.28145831009092e-05}
{"train_loss": 0.036702342331409454, "global_step": 208301, "epoch": 2340, "lr": 1.281419554703605e-05}
{"train_loss": 0.017229609191417694, "global_step": 208302, "epoch": 2340, "lr": 1.2813807998162102e-05}
{"train_loss": 0.056768275797367096, "global_step": 208303, "epoch": 2340, "lr": 1.2813420454287368e-05}
{"train_loss": 0.04518119990825653, "global_step": 208304, "epoch": 2340, "lr": 1.2813032915411927e-05}
{"train_loss": 0.03999217227101326, "global_step": 208305, "epoch": 2340, "lr": 1.281264538153581e-05}
{"train_loss": 0.06494232267141342, "global_step": 208306, "epoch": 2340, "lr": 1.2812257852659087e-05}
{"train_loss": 0.06197024881839752, "global_step": 208307, "epoch": 2340, "lr": 1.2811870328781795e-05}
{"train_loss": 0.05474239960312843, "global_step": 208308, "epoch": 2340, "lr": 1.2811482809904007e-05}
{"train_loss": 0.04997645691037178, "global_step": 208309, "epoch": 2340, "lr": 1.281109529602575e-05}
{"train_loss": 0.04756224527955055, "global_step": 208310, "epoch": 2340, "lr": 1.2810707787147098e-05}
{"train_loss": 0.04399869963526726, "global_step": 208311, "epoch": 2340, "lr": 1.2810320283268078e-05}
{"train_loss": 0.03650981932878494, "global_step": 208312, "epoch": 2340, "lr": 1.2809932784388762e-05}
{"train_loss": 0.04214194789528847, "global_step": 208313, "epoch": 2340, "lr": 1.2809545290509205e-05}
{"train_loss": 0.06326141953468323, "global_step": 208314, "epoch": 2340, "lr": 1.280915780162944e-05}
{"train_loss": 0.047322582453489304, "global_step": 208315, "epoch": 2340, "lr": 1.2808770317749546e-05}
{"train_loss": 0.0577523447573185, "global_step": 208316, "epoch": 2340, "lr": 1.2808382838869549e-05}
{"train_loss": 0.03431028500199318, "global_step": 208317, "epoch": 2340, "lr": 1.2807995364989522e-05}
{"train_loss": 0.05021163448691368, "global_step": 208318, "epoch": 2340, "lr": 1.280760789610949e-05}
{"train_loss": 0.06789598613977432, "global_step": 208319, "epoch": 2340, "lr": 1.2807220432229539e-05}
{"train_loss": 0.06346923857927322, "global_step": 208320, "epoch": 2340, "lr": 1.2806832973349687e-05}
{"train_loss": 0.037765227258205414, "global_step": 208321, "epoch": 2340, "lr": 1.2806445519470017e-05}
{"train_loss": 0.06473320722579956, "global_step": 208322, "epoch": 2340, "lr": 1.280605807059056e-05}
{"train_loss": 0.05621713772416115, "global_step": 208323, "epoch": 2340, "lr": 1.280567062671138e-05}
{"train_loss": 0.04828374460339546, "global_step": 208324, "epoch": 2340, "lr": 1.2805283187832517e-05}
{"train_loss": 0.05168360471725464, "global_step": 208325, "epoch": 2340, "lr": 1.2804895753954043e-05}
{"train_loss": 0.06504794210195541, "global_step": 208326, "epoch": 2340, "lr": 1.2804508325075981e-05}
{"train_loss": 0.07068685442209244, "global_step": 208327, "epoch": 2340, "lr": 1.2804120901198413e-05}
{"train_loss": 0.05115889757871628, "global_step": 208328, "epoch": 2340, "lr": 1.2803733482321385e-05}
{"train_loss": 0.048212241381406784, "global_step": 208329, "epoch": 2340, "lr": 1.2803346068444916e-05}
{"train_loss": 0.03241707384586334, "global_step": 208330, "epoch": 2340, "lr": 1.2802958659569109e-05}
{"train_loss": 0.04066617786884308, "global_step": 208331, "epoch": 2340, "lr": 1.2802571255693969e-05}
{"train_loss": 0.052194952964782715, "global_step": 208332, "epoch": 2340, "lr": 1.2802183856819589e-05}
{"train_loss": 0.047712959349155426, "global_step": 208333, "epoch": 2340, "lr": 1.280179646294598e-05}
{"train_loss": 0.06697995960712433, "global_step": 208334, "epoch": 2340, "lr": 1.280140907407324e-05}
{"train_loss": 0.032983992248773575, "global_step": 208335, "epoch": 2340, "lr": 1.2801021690201376e-05}
{"train_loss": 0.08538215607404709, "global_step": 208336, "epoch": 2340, "lr": 1.2800634311330479e-05}
{"train_loss": 0.04496388137340546, "global_step": 208337, "epoch": 2340, "lr": 1.280024693746057e-05}
{"train_loss": 0.06642934679985046, "global_step": 208338, "epoch": 2340, "lr": 1.2799859568591727e-05}
{"train_loss": 0.07191261649131775, "global_step": 208339, "epoch": 2340, "lr": 1.2799472204723972e-05}
{"train_loss": 0.06379618495702744, "global_step": 208340, "epoch": 2340, "lr": 1.2799084845857395e-05}
{"train_loss": 0.04490203782916069, "global_step": 208341, "epoch": 2340, "lr": 1.2798697491992006e-05}
{"train_loss": 0.06450352817773819, "global_step": 208342, "epoch": 2340, "lr": 1.27983101431279e-05}
{"train_loss": 0.05649478733539581, "global_step": 208343, "epoch": 2340, "lr": 1.2797922799265088e-05}
{"train_loss": 0.05636495351791382, "global_step": 208344, "epoch": 2340, "lr": 1.2797535460403659e-05}
{"train_loss": 0.048189274966716766, "global_step": 208345, "epoch": 2340, "lr": 1.2797148126543635e-05}
{"train_loss": 0.07198301702737808, "global_step": 208346, "epoch": 2340, "lr": 1.2796760797685092e-05}
{"train_loss": 0.06817398965358734, "global_step": 208347, "epoch": 2340, "lr": 1.279637347382806e-05}
{"train_loss": 0.05733039345215546, "global_step": 208348, "epoch": 2340, "lr": 1.27959861549726e-05, "val_loss": 8.537254333496094, "train_action_mse_error": 22.37505340576172}
{"train_loss": 0.12158932536840439, "global_step": 208349, "epoch": 2341, "lr": 1.2795598841118783e-05}
{"train_loss": 0.052285462617874146, "global_step": 208350, "epoch": 2341, "lr": 1.2795211532266626e-05}
{"train_loss": 0.08308154344558716, "global_step": 208351, "epoch": 2341, "lr": 1.279482422841622e-05}
{"train_loss": 0.04197578877210617, "global_step": 208352, "epoch": 2341, "lr": 1.2794436929567578e-05}
{"train_loss": 0.03414949029684067, "global_step": 208353, "epoch": 2341, "lr": 1.2794049635720784e-05}
{"train_loss": 0.040161848068237305, "global_step": 208354, "epoch": 2341, "lr": 1.2793662346875862e-05}
{"train_loss": 0.05663000047206879, "global_step": 208355, "epoch": 2341, "lr": 1.2793275063032895e-05}
{"train_loss": 0.07483227550983429, "global_step": 208356, "epoch": 2341, "lr": 1.2792887784191898e-05}
{"train_loss": 0.0837005004286766, "global_step": 208357, "epoch": 2341, "lr": 1.2792500510352967e-05}
{"train_loss": 0.025401460006833076, "global_step": 208358, "epoch": 2341, "lr": 1.2792113241516113e-05}
{"train_loss": 0.07254460453987122, "global_step": 208359, "epoch": 2341, "lr": 1.2791725977681418e-05}
{"train_loss": 0.09784630686044693, "global_step": 208360, "epoch": 2341, "lr": 1.2791338718848912e-05}
{"train_loss": 0.06366381049156189, "global_step": 208361, "epoch": 2341, "lr": 1.2790951465018668e-05}
{"train_loss": 0.056691303849220276, "global_step": 208362, "epoch": 2341, "lr": 1.2790564216190715e-05}
{"train_loss": 0.05000660568475723, "global_step": 208363, "epoch": 2341, "lr": 1.279017697236513e-05}
{"train_loss": 0.07360794395208359, "global_step": 208364, "epoch": 2341, "lr": 1.2789789733541934e-05}
{"train_loss": 0.07736669480800629, "global_step": 208365, "epoch": 2341, "lr": 1.278940249972122e-05}
{"train_loss": 0.053094007074832916, "global_step": 208366, "epoch": 2341, "lr": 1.2789015270902999e-05}
{"train_loss": 0.04587060585618019, "global_step": 208367, "epoch": 2341, "lr": 1.2788628047087352e-05}
{"train_loss": 0.04392968490719795, "global_step": 208368, "epoch": 2341, "lr": 1.2788240828274322e-05}
{"train_loss": 0.06464413553476334, "global_step": 208369, "epoch": 2341, "lr": 1.2787853614463951e-05}
{"train_loss": 0.06388700753450394, "global_step": 208370, "epoch": 2341, "lr": 1.2787466405656306e-05}
{"train_loss": 0.06977872550487518, "global_step": 208371, "epoch": 2341, "lr": 1.278707920185142e-05}
{"train_loss": 0.044460613280534744, "global_step": 208372, "epoch": 2341, "lr": 1.2786692003049377e-05}
{"train_loss": 0.06199050694704056, "global_step": 208373, "epoch": 2341, "lr": 1.2786304809250193e-05}
{"train_loss": 0.09159727394580841, "global_step": 208374, "epoch": 2341, "lr": 1.2785917620453952e-05}
{"train_loss": 0.07114596664905548, "global_step": 208375, "epoch": 2341, "lr": 1.278553043666067e-05}
{"train_loss": 0.03568234294652939, "global_step": 208376, "epoch": 2341, "lr": 1.2785143257870441e-05}
{"train_loss": 0.08810243010520935, "global_step": 208377, "epoch": 2341, "lr": 1.2784756084083283e-05}
{"train_loss": 0.02758077159523964, "global_step": 208378, "epoch": 2341, "lr": 1.2784368915299266e-05}
{"train_loss": 0.023007670417428017, "global_step": 208379, "epoch": 2341, "lr": 1.2783981751518432e-05}
{"train_loss": 0.09197498857975006, "global_step": 208380, "epoch": 2341, "lr": 1.278359459274085e-05}
{"train_loss": 0.08919268101453781, "global_step": 208381, "epoch": 2341, "lr": 1.2783207438966543e-05}
{"train_loss": 0.05517374351620674, "global_step": 208382, "epoch": 2341, "lr": 1.2782820290195596e-05}
{"train_loss": 0.09445907920598984, "global_step": 208383, "epoch": 2341, "lr": 1.278243314642803e-05}
{"train_loss": 0.06234072148799896, "global_step": 208384, "epoch": 2341, "lr": 1.2782046007663912e-05}
{"train_loss": 0.0381644181907177, "global_step": 208385, "epoch": 2341, "lr": 1.2781658873903313e-05}
{"train_loss": 0.044467486441135406, "global_step": 208386, "epoch": 2341, "lr": 1.2781271745146246e-05}
{"train_loss": 0.046288441866636276, "global_step": 208387, "epoch": 2341, "lr": 1.2780884621392802e-05}
{"train_loss": 0.05001160502433777, "global_step": 208388, "epoch": 2341, "lr": 1.2780497502642996e-05}
{"train_loss": 0.03880215063691139, "global_step": 208389, "epoch": 2341, "lr": 1.2780110388896921e-05}
{"train_loss": 0.0466066412627697, "global_step": 208390, "epoch": 2341, "lr": 1.2779723280154582e-05}
{"train_loss": 0.06911733001470566, "global_step": 208391, "epoch": 2341, "lr": 1.2779336176416078e-05}
{"train_loss": 0.0495358370244503, "global_step": 208392, "epoch": 2341, "lr": 1.2778949077681418e-05}
{"train_loss": 0.04466787353157997, "global_step": 208393, "epoch": 2341, "lr": 1.2778561983950693e-05}
{"train_loss": 0.045124202966690063, "global_step": 208394, "epoch": 2341, "lr": 1.277817489522392e-05}
{"train_loss": 0.04869677126407623, "global_step": 208395, "epoch": 2341, "lr": 1.2777787811501185e-05}
{"train_loss": 0.08863567560911179, "global_step": 208396, "epoch": 2341, "lr": 1.2777400732782508e-05}
{"train_loss": 0.0316765159368515, "global_step": 208397, "epoch": 2341, "lr": 1.2777013659067966e-05}
{"train_loss": 0.043999169021844864, "global_step": 208398, "epoch": 2341, "lr": 1.2776626590357588e-05}
{"train_loss": 0.12460993230342865, "global_step": 208399, "epoch": 2341, "lr": 1.2776239526651457e-05}
{"train_loss": 0.052285172045230865, "global_step": 208400, "epoch": 2341, "lr": 1.2775852467949589e-05}
{"train_loss": 0.013569968752563, "global_step": 208401, "epoch": 2341, "lr": 1.2775465414252069e-05}
{"train_loss": 0.11308753490447998, "global_step": 208402, "epoch": 2341, "lr": 1.2775078365558917e-05}
{"train_loss": 0.03108508139848709, "global_step": 208403, "epoch": 2341, "lr": 1.2774691321870225e-05}
{"train_loss": 0.053083788603544235, "global_step": 208404, "epoch": 2341, "lr": 1.2774304283186001e-05}
{"train_loss": 0.021749738603830338, "global_step": 208405, "epoch": 2341, "lr": 1.277391724950633e-05}
{"train_loss": 0.13445362448692322, "global_step": 208406, "epoch": 2341, "lr": 1.2773530220831258e-05}
{"train_loss": 0.05247298255562782, "global_step": 208407, "epoch": 2341, "lr": 1.2773143197160814e-05}
{"train_loss": 0.06639247387647629, "global_step": 208408, "epoch": 2341, "lr": 1.2772756178495082e-05}
{"train_loss": 0.055165089666843414, "global_step": 208409, "epoch": 2341, "lr": 1.2772369164834085e-05}
{"train_loss": 0.03622965142130852, "global_step": 208410, "epoch": 2341, "lr": 1.2771982156177903e-05}
{"train_loss": 0.04876835271716118, "global_step": 208411, "epoch": 2341, "lr": 1.277159515252656e-05}
{"train_loss": 0.08295087516307831, "global_step": 208412, "epoch": 2341, "lr": 1.2771208153880138e-05}
{"train_loss": 0.09822305291891098, "global_step": 208413, "epoch": 2341, "lr": 1.2770821160238654e-05}
{"train_loss": 0.03095972165465355, "global_step": 208414, "epoch": 2341, "lr": 1.2770434171602197e-05}
{"train_loss": 0.07769261300563812, "global_step": 208415, "epoch": 2341, "lr": 1.277004718797079e-05}
{"train_loss": 0.054605741053819656, "global_step": 208416, "epoch": 2341, "lr": 1.2769660209344508e-05}
{"train_loss": 0.05102073401212692, "global_step": 208417, "epoch": 2341, "lr": 1.2769273235723378e-05}
{"train_loss": 0.07433026283979416, "global_step": 208418, "epoch": 2341, "lr": 1.2768886267107477e-05}
{"train_loss": 0.05572444945573807, "global_step": 208419, "epoch": 2341, "lr": 1.2768499303496833e-05}
{"train_loss": 0.03509664908051491, "global_step": 208420, "epoch": 2341, "lr": 1.2768112344891508e-05}
{"train_loss": 0.08494101464748383, "global_step": 208421, "epoch": 2341, "lr": 1.2767725391291574e-05}
{"train_loss": 0.0965343713760376, "global_step": 208422, "epoch": 2341, "lr": 1.2767338442697051e-05}
{"train_loss": 0.05727793648838997, "global_step": 208423, "epoch": 2341, "lr": 1.276695149910802e-05}
{"train_loss": 0.04830698296427727, "global_step": 208424, "epoch": 2341, "lr": 1.2766564560524502e-05}
{"train_loss": 0.0695127472281456, "global_step": 208425, "epoch": 2341, "lr": 1.2766177626946585e-05}
{"train_loss": 0.023143645375967026, "global_step": 208426, "epoch": 2341, "lr": 1.276579069837428e-05}
{"train_loss": 0.0584835410118103, "global_step": 208427, "epoch": 2341, "lr": 1.2765403774807682e-05}
{"train_loss": 0.09535571187734604, "global_step": 208428, "epoch": 2341, "lr": 1.2765016856246803e-05}
{"train_loss": 0.03231367468833923, "global_step": 208429, "epoch": 2341, "lr": 1.276462994269173e-05}
{"train_loss": 0.028772959485650063, "global_step": 208430, "epoch": 2341, "lr": 1.2764243034142482e-05}
{"train_loss": 0.027895240113139153, "global_step": 208431, "epoch": 2341, "lr": 1.2763856130599144e-05}
{"train_loss": 0.05102924630045891, "global_step": 208432, "epoch": 2341, "lr": 1.2763469232061736e-05}
{"train_loss": 0.050699397921562195, "global_step": 208433, "epoch": 2341, "lr": 1.2763082338530342e-05}
{"train_loss": 0.028036272153258324, "global_step": 208434, "epoch": 2341, "lr": 1.276269545000498e-05}
{"train_loss": 0.017328275367617607, "global_step": 208435, "epoch": 2341, "lr": 1.2762308566485737e-05}
{"train_loss": 0.11785080283880234, "global_step": 208436, "epoch": 2341, "lr": 1.2761921687972633e-05}
{"train_loss": 0.0598857460041227, "global_step": 208437, "epoch": 2341, "lr": 1.2761534814465748e-05, "val_loss": 8.555689811706543}
{"train_loss": 0.07301253080368042, "global_step": 208438, "epoch": 2342, "lr": 1.2761147945965107e-05}
{"train_loss": 0.03360600024461746, "global_step": 208439, "epoch": 2342, "lr": 1.276076108247079e-05}
{"train_loss": 0.04108777269721031, "global_step": 208440, "epoch": 2342, "lr": 1.2760374223982824e-05}
{"train_loss": 0.06590629369020462, "global_step": 208441, "epoch": 2342, "lr": 1.275998737050128e-05}
{"train_loss": 0.03435410186648369, "global_step": 208442, "epoch": 2342, "lr": 1.2759600522026194e-05}
{"train_loss": 0.05201344192028046, "global_step": 208443, "epoch": 2342, "lr": 1.2759213678557635e-05}
{"train_loss": 0.06343430280685425, "global_step": 208444, "epoch": 2342, "lr": 1.2758826840095644e-05}
{"train_loss": 0.06156911700963974, "global_step": 208445, "epoch": 2342, "lr": 1.2758440006640265e-05}
{"train_loss": 0.042291656136512756, "global_step": 208446, "epoch": 2342, "lr": 1.2758053178191575e-05}
{"train_loss": 0.05356568098068237, "global_step": 208447, "epoch": 2342, "lr": 1.2757666354749592e-05}
{"train_loss": 0.03068491443991661, "global_step": 208448, "epoch": 2342, "lr": 1.2757279536314404e-05}
{"train_loss": 0.0901767760515213, "global_step": 208449, "epoch": 2342, "lr": 1.2756892722886033e-05}
{"train_loss": 0.044579826295375824, "global_step": 208450, "epoch": 2342, "lr": 1.2756505914464557e-05}
{"train_loss": 0.04612189158797264, "global_step": 208451, "epoch": 2342, "lr": 1.2756119111050002e-05}
{"train_loss": 0.0948236882686615, "global_step": 208452, "epoch": 2342, "lr": 1.2755732312642444e-05}
{"train_loss": 0.06605252623558044, "global_step": 208453, "epoch": 2342, "lr": 1.2755345519241908e-05}
{"train_loss": 0.0861993059515953, "global_step": 208454, "epoch": 2342, "lr": 1.2754958730848481e-05}
{"train_loss": 0.05017279088497162, "global_step": 208455, "epoch": 2342, "lr": 1.275457194746218e-05}
{"train_loss": 0.07371048629283905, "global_step": 208456, "epoch": 2342, "lr": 1.2754185169083072e-05}
{"train_loss": 0.05110662430524826, "global_step": 208457, "epoch": 2342, "lr": 1.2753798395711226e-05}
{"train_loss": 0.0632731020450592, "global_step": 208458, "epoch": 2342, "lr": 1.2753411627346657e-05}
{"train_loss": 0.0877753272652626, "global_step": 208459, "epoch": 2342, "lr": 1.275302486398946e-05}
{"train_loss": 0.03687646612524986, "global_step": 208460, "epoch": 2342, "lr": 1.2752638105639646e-05}
{"train_loss": 0.0743899792432785, "global_step": 208461, "epoch": 2342, "lr": 1.27522513522973e-05}
{"train_loss": 0.04251471534371376, "global_step": 208462, "epoch": 2342, "lr": 1.2751864603962449e-05}
{"train_loss": 0.0692516565322876, "global_step": 208463, "epoch": 2342, "lr": 1.275147786063517e-05}
{"train_loss": 0.03481642156839371, "global_step": 208464, "epoch": 2342, "lr": 1.2751091122315484e-05}
{"train_loss": 0.03357915207743645, "global_step": 208465, "epoch": 2342, "lr": 1.2750704389003477e-05}
{"train_loss": 0.050888750702142715, "global_step": 208466, "epoch": 2342, "lr": 1.2750317660699162e-05}
{"train_loss": 0.0321367047727108, "global_step": 208467, "epoch": 2342, "lr": 1.2749930937402632e-05}
{"train_loss": 0.03178470954298973, "global_step": 208468, "epoch": 2342, "lr": 1.2749544219113907e-05}
{"train_loss": 0.08128892630338669, "global_step": 208469, "epoch": 2342, "lr": 1.2749157505833066e-05}
{"train_loss": 0.03834174945950508, "global_step": 208470, "epoch": 2342, "lr": 1.2748770797560133e-05}
{"train_loss": 0.056832000613212585, "global_step": 208471, "epoch": 2342, "lr": 1.2748384094295185e-05}
{"train_loss": 0.10557243973016739, "global_step": 208472, "epoch": 2342, "lr": 1.2747997396038248e-05}
{"train_loss": 0.06995291262865067, "global_step": 208473, "epoch": 2342, "lr": 1.2747610702789408e-05}
{"train_loss": 0.022016311064362526, "global_step": 208474, "epoch": 2342, "lr": 1.2747224014548681e-05}
{"train_loss": 0.14405730366706848, "global_step": 208475, "epoch": 2342, "lr": 1.274683733131613e-05}
{"train_loss": 0.024389080703258514, "global_step": 208476, "epoch": 2342, "lr": 1.2746450653091836e-05}
{"train_loss": 0.028293456882238388, "global_step": 208477, "epoch": 2342, "lr": 1.2746063979875822e-05}
{"train_loss": 0.06665372848510742, "global_step": 208478, "epoch": 2342, "lr": 1.274567731166813e-05}
{"train_loss": 0.07003246992826462, "global_step": 208479, "epoch": 2342, "lr": 1.274529064846885e-05}
{"train_loss": 0.07991820573806763, "global_step": 208480, "epoch": 2342, "lr": 1.274490399027799e-05}
{"train_loss": 0.08728983253240585, "global_step": 208481, "epoch": 2342, "lr": 1.2744517337095641e-05}
{"train_loss": 0.04655072093009949, "global_step": 208482, "epoch": 2342, "lr": 1.2744130688921818e-05}
{"train_loss": 0.032257676124572754, "global_step": 208483, "epoch": 2342, "lr": 1.2743744045756611e-05}
{"train_loss": 0.08424687385559082, "global_step": 208484, "epoch": 2342, "lr": 1.274335740760006e-05}
{"train_loss": 0.09099584817886353, "global_step": 208485, "epoch": 2342, "lr": 1.2742970774452184e-05}
{"train_loss": 0.03746574744582176, "global_step": 208486, "epoch": 2342, "lr": 1.2742584146313085e-05}
{"train_loss": 0.026151584461331367, "global_step": 208487, "epoch": 2342, "lr": 1.2742197523182775e-05}
{"train_loss": 0.057846538722515106, "global_step": 208488, "epoch": 2342, "lr": 1.2741810905061336e-05}
{"train_loss": 0.03426887467503548, "global_step": 208489, "epoch": 2342, "lr": 1.274142429194879e-05}
{"train_loss": 0.050587575882673264, "global_step": 208490, "epoch": 2342, "lr": 1.274103768384522e-05}
{"train_loss": 0.0812128335237503, "global_step": 208491, "epoch": 2342, "lr": 1.2740651080750648e-05}
{"train_loss": 0.035611074417829514, "global_step": 208492, "epoch": 2342, "lr": 1.274026448266516e-05}
{"train_loss": 0.11600110679864883, "global_step": 208493, "epoch": 2342, "lr": 1.2739877889588775e-05}
{"train_loss": 0.023766612634062767, "global_step": 208494, "epoch": 2342, "lr": 1.2739491301521555e-05}
{"train_loss": 0.0487888902425766, "global_step": 208495, "epoch": 2342, "lr": 1.2739104718463574e-05}
{"train_loss": 0.047585971653461456, "global_step": 208496, "epoch": 2342, "lr": 1.2738718140414852e-05}
{"train_loss": 0.025145381689071655, "global_step": 208497, "epoch": 2342, "lr": 1.2738331567375466e-05}
{"train_loss": 0.07516314834356308, "global_step": 208498, "epoch": 2342, "lr": 1.2737944999345447e-05}
{"train_loss": 0.028299972414970398, "global_step": 208499, "epoch": 2342, "lr": 1.273755843632487e-05}
{"train_loss": 0.059719283133745193, "global_step": 208500, "epoch": 2342, "lr": 1.2737171878313758e-05}
{"train_loss": 0.059963203966617584, "global_step": 208501, "epoch": 2342, "lr": 1.2736785325312194e-05}
{"train_loss": 0.07769586890935898, "global_step": 208502, "epoch": 2342, "lr": 1.2736398777320203e-05}
{"train_loss": 0.07320339232683182, "global_step": 208503, "epoch": 2342, "lr": 1.2736012234337863e-05}
{"train_loss": 0.0617012120783329, "global_step": 208504, "epoch": 2342, "lr": 1.2735625696365194e-05}
{"train_loss": 0.054334014654159546, "global_step": 208505, "epoch": 2342, "lr": 1.2735239163402284e-05}
{"train_loss": 0.021274138242006302, "global_step": 208506, "epoch": 2342, "lr": 1.2734852635449151e-05}
{"train_loss": 0.04117588698863983, "global_step": 208507, "epoch": 2342, "lr": 1.2734466112505877e-05}
{"train_loss": 0.05080719292163849, "global_step": 208508, "epoch": 2342, "lr": 1.2734079594572485e-05}
{"train_loss": 0.05129743367433548, "global_step": 208509, "epoch": 2342, "lr": 1.2733693081649055e-05}
{"train_loss": 0.08086314797401428, "global_step": 208510, "epoch": 2342, "lr": 1.2733306573735616e-05}
{"train_loss": 0.06959731876850128, "global_step": 208511, "epoch": 2342, "lr": 1.2732920070832227e-05}
{"train_loss": 0.060922037810087204, "global_step": 208512, "epoch": 2342, "lr": 1.273253357293896e-05}
{"train_loss": 0.08825601637363434, "global_step": 208513, "epoch": 2342, "lr": 1.2732147080055828e-05}
{"train_loss": 0.03567599877715111, "global_step": 208514, "epoch": 2342, "lr": 1.2731760592182924e-05}
{"train_loss": 0.06619089096784592, "global_step": 208515, "epoch": 2342, "lr": 1.2731374109320265e-05}
{"train_loss": 0.0711798220872879, "global_step": 208516, "epoch": 2342, "lr": 1.2730987631467933e-05}
{"train_loss": 0.06514936685562134, "global_step": 208517, "epoch": 2342, "lr": 1.2730601158625972e-05}
{"train_loss": 0.07432714849710464, "global_step": 208518, "epoch": 2342, "lr": 1.2730214690794407e-05}
{"train_loss": 0.07999251037836075, "global_step": 208519, "epoch": 2342, "lr": 1.2729828227973324e-05}
{"train_loss": 0.08128485083580017, "global_step": 208520, "epoch": 2342, "lr": 1.2729441770162748e-05}
{"train_loss": 0.06924168765544891, "global_step": 208521, "epoch": 2342, "lr": 1.2729055317362754e-05}
{"train_loss": 0.046517424285411835, "global_step": 208522, "epoch": 2342, "lr": 1.272866886957339e-05}
{"train_loss": 0.020131554454565048, "global_step": 208523, "epoch": 2342, "lr": 1.2728282426794685e-05}
{"train_loss": 0.06087028607726097, "global_step": 208524, "epoch": 2342, "lr": 1.2727895989026728e-05}
{"train_loss": 0.06772956252098083, "global_step": 208525, "epoch": 2342, "lr": 1.272750955626953e-05}
{"train_loss": 0.05776213810601261, "global_step": 208526, "epoch": 2342, "lr": 1.2727123128523182e-05, "val_loss": 8.457571983337402}
{"train_loss": 0.07348834723234177, "global_step": 208527, "epoch": 2343, "lr": 1.2726736705787701e-05}
{"train_loss": 0.07162865251302719, "global_step": 208528, "epoch": 2343, "lr": 1.2726350288063177e-05}
{"train_loss": 0.03991890698671341, "global_step": 208529, "epoch": 2343, "lr": 1.2725963875349616e-05}
{"train_loss": 0.029672574251890182, "global_step": 208530, "epoch": 2343, "lr": 1.2725577467647105e-05}
{"train_loss": 0.04922664165496826, "global_step": 208531, "epoch": 2343, "lr": 1.2725191064955694e-05}
{"train_loss": 0.09722337126731873, "global_step": 208532, "epoch": 2343, "lr": 1.2724804667275413e-05}
{"train_loss": 0.07074900716543198, "global_step": 208533, "epoch": 2343, "lr": 1.2724418274606342e-05}
{"train_loss": 0.05665288865566254, "global_step": 208534, "epoch": 2343, "lr": 1.2724031886948506e-05}
{"train_loss": 0.044331274926662445, "global_step": 208535, "epoch": 2343, "lr": 1.2723645504301985e-05}
{"train_loss": 0.04620377719402313, "global_step": 208536, "epoch": 2343, "lr": 1.2723259126666798e-05}
{"train_loss": 0.025604892522096634, "global_step": 208537, "epoch": 2343, "lr": 1.272287275404303e-05}
{"train_loss": 0.08400977402925491, "global_step": 208538, "epoch": 2343, "lr": 1.2722486386430704e-05}
{"train_loss": 0.06306132674217224, "global_step": 208539, "epoch": 2343, "lr": 1.2722100023829903e-05}
{"train_loss": 0.047059960663318634, "global_step": 208540, "epoch": 2343, "lr": 1.2721713666240642e-05}
{"train_loss": 0.05560217425227165, "global_step": 208541, "epoch": 2343, "lr": 1.2721327313663006e-05}
{"train_loss": 0.07736881822347641, "global_step": 208542, "epoch": 2343, "lr": 1.2720940966097023e-05}
{"train_loss": 0.059014104306697845, "global_step": 208543, "epoch": 2343, "lr": 1.2720554623542763e-05}
{"train_loss": 0.030922843143343925, "global_step": 208544, "epoch": 2343, "lr": 1.2720168286000261e-05}
{"train_loss": 0.07486945390701294, "global_step": 208545, "epoch": 2343, "lr": 1.2719781953469589e-05}
{"train_loss": 0.10476402938365936, "global_step": 208546, "epoch": 2343, "lr": 1.2719395625950775e-05}
{"train_loss": 0.11698482930660248, "global_step": 208547, "epoch": 2343, "lr": 1.2719009303443884e-05}
{"train_loss": 0.033489733934402466, "global_step": 208548, "epoch": 2343, "lr": 1.2718622985948986e-05}
{"train_loss": 0.08615215867757797, "global_step": 208549, "epoch": 2343, "lr": 1.2718236673466094e-05}
{"train_loss": 0.04189906641840935, "global_step": 208550, "epoch": 2343, "lr": 1.2717850365995299e-05}
{"train_loss": 0.03422502428293228, "global_step": 208551, "epoch": 2343, "lr": 1.2717464063536622e-05}
{"train_loss": 0.03759829327464104, "global_step": 208552, "epoch": 2343, "lr": 1.2717077766090136e-05}
{"train_loss": 0.10225147008895874, "global_step": 208553, "epoch": 2343, "lr": 1.2716691473655872e-05}
{"train_loss": 0.039940495043992996, "global_step": 208554, "epoch": 2343, "lr": 1.2716305186233913e-05}
{"train_loss": 0.04066658020019531, "global_step": 208555, "epoch": 2343, "lr": 1.2715918903824286e-05}
{"train_loss": 0.04725080356001854, "global_step": 208556, "epoch": 2343, "lr": 1.2715532626427041e-05}
{"train_loss": 0.056673020124435425, "global_step": 208557, "epoch": 2343, "lr": 1.271514635404225e-05}
{"train_loss": 0.07168306410312653, "global_step": 208558, "epoch": 2343, "lr": 1.271476008666994e-05}
{"train_loss": 0.09638679772615433, "global_step": 208559, "epoch": 2343, "lr": 1.2714373824310188e-05}
{"train_loss": 0.09022723138332367, "global_step": 208560, "epoch": 2343, "lr": 1.2713987566963036e-05}
{"train_loss": 0.04650082439184189, "global_step": 208561, "epoch": 2343, "lr": 1.271360131462852e-05}
{"train_loss": 0.033516716212034225, "global_step": 208562, "epoch": 2343, "lr": 1.2713215067306717e-05}
{"train_loss": 0.08074581623077393, "global_step": 208563, "epoch": 2343, "lr": 1.2712828824997652e-05}
{"train_loss": 0.04485946521162987, "global_step": 208564, "epoch": 2343, "lr": 1.2712442587701412e-05}
{"train_loss": 0.10942105203866959, "global_step": 208565, "epoch": 2343, "lr": 1.2712056355418018e-05}
{"train_loss": 0.058038633316755295, "global_step": 208566, "epoch": 2343, "lr": 1.2711670128147523e-05}
{"train_loss": 0.05172596126794815, "global_step": 208567, "epoch": 2343, "lr": 1.2711283905890014e-05}
{"train_loss": 0.06047436594963074, "global_step": 208568, "epoch": 2343, "lr": 1.2710897688645502e-05}
{"train_loss": 0.053152505308389664, "global_step": 208569, "epoch": 2343, "lr": 1.2710511476414067e-05}
{"train_loss": 0.03315196931362152, "global_step": 208570, "epoch": 2343, "lr": 1.2710125269195733e-05}
{"train_loss": 0.03380622714757919, "global_step": 208571, "epoch": 2343, "lr": 1.2709739066990584e-05}
{"train_loss": 0.02865883708000183, "global_step": 208572, "epoch": 2343, "lr": 1.2709352869798646e-05}
{"train_loss": 0.047405365854501724, "global_step": 208573, "epoch": 2343, "lr": 1.2708966677619988e-05}
{"train_loss": 0.04614327847957611, "global_step": 208574, "epoch": 2343, "lr": 1.2708580490454647e-05}
{"train_loss": 0.06682559847831726, "global_step": 208575, "epoch": 2343, "lr": 1.2708194308302696e-05}
{"train_loss": 0.1228804737329483, "global_step": 208576, "epoch": 2343, "lr": 1.2707808131164155e-05}
{"train_loss": 0.06261560320854187, "global_step": 208577, "epoch": 2343, "lr": 1.2707421959039111e-05}
{"train_loss": 0.040823228657245636, "global_step": 208578, "epoch": 2343, "lr": 1.270703579192759e-05}
{"train_loss": 0.028680752962827682, "global_step": 208579, "epoch": 2343, "lr": 1.2706649629829664e-05}
{"train_loss": 0.0891055092215538, "global_step": 208580, "epoch": 2343, "lr": 1.270626347274536e-05}
{"train_loss": 0.13036957383155823, "global_step": 208581, "epoch": 2343, "lr": 1.2705877320674758e-05}
{"train_loss": 0.04846098646521568, "global_step": 208582, "epoch": 2343, "lr": 1.2705491173617884e-05}
{"train_loss": 0.07496678829193115, "global_step": 208583, "epoch": 2343, "lr": 1.2705105031574798e-05}
{"train_loss": 0.051667798310518265, "global_step": 208584, "epoch": 2343, "lr": 1.2704718894545575e-05}
{"train_loss": 0.03206752613186836, "global_step": 208585, "epoch": 2343, "lr": 1.2704332762530236e-05}
{"train_loss": 0.05221846327185631, "global_step": 208586, "epoch": 2343, "lr": 1.2703946635528852e-05}
{"train_loss": 0.06473085284233093, "global_step": 208587, "epoch": 2343, "lr": 1.2703560513541457e-05}
{"train_loss": 0.07138992846012115, "global_step": 208588, "epoch": 2343, "lr": 1.270317439656813e-05}
{"train_loss": 0.05176877602934837, "global_step": 208589, "epoch": 2343, "lr": 1.2702788284608891e-05}
{"train_loss": 0.043122462928295135, "global_step": 208590, "epoch": 2343, "lr": 1.2702402177663818e-05}
{"train_loss": 0.03364730253815651, "global_step": 208591, "epoch": 2343, "lr": 1.2702016075732941e-05}
{"train_loss": 0.02999035082757473, "global_step": 208592, "epoch": 2343, "lr": 1.2701629978816337e-05}
{"train_loss": 0.06879101693630219, "global_step": 208593, "epoch": 2343, "lr": 1.2701243886914032e-05}
{"train_loss": 0.03562292084097862, "global_step": 208594, "epoch": 2343, "lr": 1.2700857800026106e-05}
{"train_loss": 0.06710159778594971, "global_step": 208595, "epoch": 2343, "lr": 1.270047171815259e-05}
{"train_loss": 0.026742588728666306, "global_step": 208596, "epoch": 2343, "lr": 1.270008564129353e-05}
{"train_loss": 0.05479150265455246, "global_step": 208597, "epoch": 2343, "lr": 1.2699699569449002e-05}
{"train_loss": 0.09658519923686981, "global_step": 208598, "epoch": 2343, "lr": 1.2699313502619032e-05}
{"train_loss": 0.046689894050359726, "global_step": 208599, "epoch": 2343, "lr": 1.2698927440803698e-05}
{"train_loss": 0.10116270184516907, "global_step": 208600, "epoch": 2343, "lr": 1.2698541384003038e-05}
{"train_loss": 0.03230580314993858, "global_step": 208601, "epoch": 2343, "lr": 1.2698155332217088e-05}
{"train_loss": 0.042631786316633224, "global_step": 208602, "epoch": 2343, "lr": 1.2697769285445921e-05}
{"train_loss": 0.08003557473421097, "global_step": 208603, "epoch": 2343, "lr": 1.2697383243689592e-05}
{"train_loss": 0.10829389840364456, "global_step": 208604, "epoch": 2343, "lr": 1.2696997206948136e-05}
{"train_loss": 0.029101846739649773, "global_step": 208605, "epoch": 2343, "lr": 1.2696611175221628e-05}
{"train_loss": 0.049309100955724716, "global_step": 208606, "epoch": 2343, "lr": 1.2696225148510094e-05}
{"train_loss": 0.0796111673116684, "global_step": 208607, "epoch": 2343, "lr": 1.2695839126813609e-05}
{"train_loss": 0.053482428193092346, "global_step": 208608, "epoch": 2343, "lr": 1.26954531101322e-05}
{"train_loss": 0.023975685238838196, "global_step": 208609, "epoch": 2343, "lr": 1.2695067098465947e-05}
{"train_loss": 0.047177501022815704, "global_step": 208610, "epoch": 2343, "lr": 1.269468109181487e-05}
{"train_loss": 0.08098756521940231, "global_step": 208611, "epoch": 2343, "lr": 1.2694295090179059e-05}
{"train_loss": 0.07817104458808899, "global_step": 208612, "epoch": 2343, "lr": 1.2693909093558526e-05}
{"train_loss": 0.07999920845031738, "global_step": 208613, "epoch": 2343, "lr": 1.2693523101953359e-05}
{"train_loss": 0.08007338643074036, "global_step": 208614, "epoch": 2343, "lr": 1.2693137115363578e-05}
{"train_loss": 0.06005062834684099, "global_step": 208615, "epoch": 2343, "lr": 1.2692751133789266e-05, "val_loss": 8.619696617126465}
{"train_loss": 0.07524822652339935, "global_step": 208616, "epoch": 2344, "lr": 1.269236515723044e-05}
{"train_loss": 0.0676819384098053, "global_step": 208617, "epoch": 2344, "lr": 1.2691979185687187e-05}
{"train_loss": 0.05937616899609566, "global_step": 208618, "epoch": 2344, "lr": 1.2691593219159531e-05}
{"train_loss": 0.04694461449980736, "global_step": 208619, "epoch": 2344, "lr": 1.2691207257647547e-05}
{"train_loss": 0.05520633980631828, "global_step": 208620, "epoch": 2344, "lr": 1.2690821301151263e-05}
{"train_loss": 0.08341731876134872, "global_step": 208621, "epoch": 2344, "lr": 1.2690435349670742e-05}
{"train_loss": 0.03431588411331177, "global_step": 208622, "epoch": 2344, "lr": 1.2690049403206056e-05}
{"train_loss": 0.04644119739532471, "global_step": 208623, "epoch": 2344, "lr": 1.2689663461757218e-05}
{"train_loss": 0.09541231393814087, "global_step": 208624, "epoch": 2344, "lr": 1.2689277525324317e-05}
{"train_loss": 0.050142064690589905, "global_step": 208625, "epoch": 2344, "lr": 1.268889159390737e-05}
{"train_loss": 0.04837800934910774, "global_step": 208626, "epoch": 2344, "lr": 1.2688505667506466e-05}
{"train_loss": 0.07017103582620621, "global_step": 208627, "epoch": 2344, "lr": 1.2688119746121625e-05}
{"train_loss": 0.09258528053760529, "global_step": 208628, "epoch": 2344, "lr": 1.2687733829752919e-05}
{"train_loss": 0.0559261254966259, "global_step": 208629, "epoch": 2344, "lr": 1.2687347918400382e-05}
{"train_loss": 0.05391376465559006, "global_step": 208630, "epoch": 2344, "lr": 1.268696201206409e-05}
{"train_loss": 0.1450842171907425, "global_step": 208631, "epoch": 2344, "lr": 1.2686576110744063e-05}
{"train_loss": 0.06213759630918503, "global_step": 208632, "epoch": 2344, "lr": 1.268619021444039e-05}
{"train_loss": 0.04185301065444946, "global_step": 208633, "epoch": 2344, "lr": 1.2685804323153106e-05}
{"train_loss": 0.024293268099427223, "global_step": 208634, "epoch": 2344, "lr": 1.268541843688224e-05}
{"train_loss": 0.02578594908118248, "global_step": 208635, "epoch": 2344, "lr": 1.2685032555627884e-05}
{"train_loss": 0.06499837338924408, "global_step": 208636, "epoch": 2344, "lr": 1.2684646679390055e-05}
{"train_loss": 0.07394598424434662, "global_step": 208637, "epoch": 2344, "lr": 1.2684260808168835e-05}
{"train_loss": 0.08717449754476547, "global_step": 208638, "epoch": 2344, "lr": 1.2683874941964247e-05}
{"train_loss": 0.05330384150147438, "global_step": 208639, "epoch": 2344, "lr": 1.2683489080776373e-05}
{"train_loss": 0.07148361206054688, "global_step": 208640, "epoch": 2344, "lr": 1.268310322460523e-05}
{"train_loss": 0.0445866622030735, "global_step": 208641, "epoch": 2344, "lr": 1.2682717373450909e-05}
{"train_loss": 0.06023018807172775, "global_step": 208642, "epoch": 2344, "lr": 1.2682331527313423e-05}
{"train_loss": 0.0805051401257515, "global_step": 208643, "epoch": 2344, "lr": 1.2681945686192859e-05}
{"train_loss": 0.057193394750356674, "global_step": 208644, "epoch": 2344, "lr": 1.2681559850089237e-05}
{"train_loss": 0.059234827756881714, "global_step": 208645, "epoch": 2344, "lr": 1.2681174019002639e-05}
{"train_loss": 0.040096621960401535, "global_step": 208646, "epoch": 2344, "lr": 1.268078819293309e-05}
{"train_loss": 0.05247009918093681, "global_step": 208647, "epoch": 2344, "lr": 1.2680402371880668e-05}
{"train_loss": 0.05222056433558464, "global_step": 208648, "epoch": 2344, "lr": 1.2680016555845398e-05}
{"train_loss": 0.06659102439880371, "global_step": 208649, "epoch": 2344, "lr": 1.2679630744827353e-05}
{"train_loss": 0.07958345860242844, "global_step": 208650, "epoch": 2344, "lr": 1.2679244938826568e-05}
{"train_loss": 0.034931913018226624, "global_step": 208651, "epoch": 2344, "lr": 1.2678859137843119e-05}
{"train_loss": 0.025421731173992157, "global_step": 208652, "epoch": 2344, "lr": 1.2678473341877023e-05}
{"train_loss": 0.035556942224502563, "global_step": 208653, "epoch": 2344, "lr": 1.267808755092837e-05}
{"train_loss": 0.07293051481246948, "global_step": 208654, "epoch": 2344, "lr": 1.2677701764997174e-05}
{"train_loss": 0.045014768838882446, "global_step": 208655, "epoch": 2344, "lr": 1.2677315984083526e-05}
{"train_loss": 0.07873648405075073, "global_step": 208656, "epoch": 2344, "lr": 1.2676930208187443e-05}
{"train_loss": 0.09802121669054031, "global_step": 208657, "epoch": 2344, "lr": 1.267654443730899e-05}
{"train_loss": 0.025484532117843628, "global_step": 208658, "epoch": 2344, "lr": 1.2676158671448236e-05}
{"train_loss": 0.08468308299779892, "global_step": 208659, "epoch": 2344, "lr": 1.2675772910605199e-05}
{"train_loss": 0.0327138677239418, "global_step": 208660, "epoch": 2344, "lr": 1.2675387154779967e-05}
{"train_loss": 0.051914114505052567, "global_step": 208661, "epoch": 2344, "lr": 1.2675001403972565e-05}
{"train_loss": 0.05375898629426956, "global_step": 208662, "epoch": 2344, "lr": 1.2674615658183059e-05}
{"train_loss": 0.06774980574846268, "global_step": 208663, "epoch": 2344, "lr": 1.267422991741149e-05}
{"train_loss": 0.05402378737926483, "global_step": 208664, "epoch": 2344, "lr": 1.2673844181657924e-05}
{"train_loss": 0.06390031427145004, "global_step": 208665, "epoch": 2344, "lr": 1.2673458450922392e-05}
{"train_loss": 0.030822964385151863, "global_step": 208666, "epoch": 2344, "lr": 1.2673072725204977e-05}
{"train_loss": 0.031822413206100464, "global_step": 208667, "epoch": 2344, "lr": 1.267268700450569e-05}
{"train_loss": 0.04287780448794365, "global_step": 208668, "epoch": 2344, "lr": 1.2672301288824628e-05}
{"train_loss": 0.07381046563386917, "global_step": 208669, "epoch": 2344, "lr": 1.2671915578161802e-05}
{"train_loss": 0.04633783921599388, "global_step": 208670, "epoch": 2344, "lr": 1.2671529872517295e-05}
{"train_loss": 0.0859609842300415, "global_step": 208671, "epoch": 2344, "lr": 1.2671144171891147e-05}
{"train_loss": 0.0904569923877716, "global_step": 208672, "epoch": 2344, "lr": 1.2670758476283395e-05}
{"train_loss": 0.057334087789058685, "global_step": 208673, "epoch": 2344, "lr": 1.2670372785694118e-05}
{"train_loss": 0.038526635617017746, "global_step": 208674, "epoch": 2344, "lr": 1.2669987100123343e-05}
{"train_loss": 0.03356921672821045, "global_step": 208675, "epoch": 2344, "lr": 1.2669601419571142e-05}
{"train_loss": 0.0609322264790535, "global_step": 208676, "epoch": 2344, "lr": 1.2669215744037549e-05}
{"train_loss": 0.08703333139419556, "global_step": 208677, "epoch": 2344, "lr": 1.2668830073522636e-05}
{"train_loss": 0.030781831592321396, "global_step": 208678, "epoch": 2344, "lr": 1.266844440802643e-05}
{"train_loss": 0.08211489021778107, "global_step": 208679, "epoch": 2344, "lr": 1.2668058747549016e-05}
{"train_loss": 0.01962222531437874, "global_step": 208680, "epoch": 2344, "lr": 1.2667673092090405e-05}
{"train_loss": 0.09585828334093094, "global_step": 208681, "epoch": 2344, "lr": 1.2667287441650688e-05}
{"train_loss": 0.06542710214853287, "global_step": 208682, "epoch": 2344, "lr": 1.2666901796229885e-05}
{"train_loss": 0.07166869193315506, "global_step": 208683, "epoch": 2344, "lr": 1.2666516155828074e-05}
{"train_loss": 0.06793477386236191, "global_step": 208684, "epoch": 2344, "lr": 1.2666130520445286e-05}
{"train_loss": 0.03031117096543312, "global_step": 208685, "epoch": 2344, "lr": 1.2665744890081588e-05}
{"train_loss": 0.03113238699734211, "global_step": 208686, "epoch": 2344, "lr": 1.2665359264737014e-05}
{"train_loss": 0.0617772601544857, "global_step": 208687, "epoch": 2344, "lr": 1.2664973644411649e-05}
{"train_loss": 0.04220602288842201, "global_step": 208688, "epoch": 2344, "lr": 1.26645880291055e-05}
{"train_loss": 0.06415125727653503, "global_step": 208689, "epoch": 2344, "lr": 1.266420241881866e-05}
{"train_loss": 0.07001008093357086, "global_step": 208690, "epoch": 2344, "lr": 1.2663816813551149e-05}
{"train_loss": 0.08394373953342438, "global_step": 208691, "epoch": 2344, "lr": 1.2663431213303045e-05}
{"train_loss": 0.08001530170440674, "global_step": 208692, "epoch": 2344, "lr": 1.2663045618074377e-05}
{"train_loss": 0.03593455255031586, "global_step": 208693, "epoch": 2344, "lr": 1.2662660027865203e-05}
{"train_loss": 0.03296119347214699, "global_step": 208694, "epoch": 2344, "lr": 1.2662274442675598e-05}
{"train_loss": 0.08892248570919037, "global_step": 208695, "epoch": 2344, "lr": 1.2661888862505578e-05}
{"train_loss": 0.05032913014292717, "global_step": 208696, "epoch": 2344, "lr": 1.2661503287355225e-05}
{"train_loss": 0.031933050602674484, "global_step": 208697, "epoch": 2344, "lr": 1.2661117717224563e-05}
{"train_loss": 0.09963511675596237, "global_step": 208698, "epoch": 2344, "lr": 1.2660732152113675e-05}
{"train_loss": 0.02233782596886158, "global_step": 208699, "epoch": 2344, "lr": 1.2660346592022581e-05}
{"train_loss": 0.045259568840265274, "global_step": 208700, "epoch": 2344, "lr": 1.2659961036951368e-05}
{"train_loss": 0.08749202638864517, "global_step": 208701, "epoch": 2344, "lr": 1.2659575486900049e-05}
{"train_loss": 0.053512297570705414, "global_step": 208702, "epoch": 2344, "lr": 1.265918994186871e-05}
{"train_loss": 0.03384529799222946, "global_step": 208703, "epoch": 2344, "lr": 1.2658804401857371e-05}
{"train_loss": 0.058689409804143264, "global_step": 208704, "epoch": 2344, "lr": 1.2658418866866117e-05, "val_loss": 8.54753589630127}
{"train_loss": 0.0539601594209671, "global_step": 208705, "epoch": 2345, "lr": 1.265803333689497e-05}
{"train_loss": 0.03413606807589531, "global_step": 208706, "epoch": 2345, "lr": 1.2657647811944012e-05}
{"train_loss": 0.03423610329627991, "global_step": 208707, "epoch": 2345, "lr": 1.2657262292013261e-05}
{"train_loss": 0.07191979885101318, "global_step": 208708, "epoch": 2345, "lr": 1.26568767771028e-05}
{"train_loss": 0.04108284041285515, "global_step": 208709, "epoch": 2345, "lr": 1.265649126721265e-05}
{"train_loss": 0.05872376635670662, "global_step": 208710, "epoch": 2345, "lr": 1.2656105762342901e-05}
{"train_loss": 0.09284559637308121, "global_step": 208711, "epoch": 2345, "lr": 1.2655720262493575e-05}
{"train_loss": 0.11368970572948456, "global_step": 208712, "epoch": 2345, "lr": 1.2655334767664722e-05}
{"train_loss": 0.0968475267291069, "global_step": 208713, "epoch": 2345, "lr": 1.2654949277856421e-05}
{"train_loss": 0.028408514335751534, "global_step": 208714, "epoch": 2345, "lr": 1.2654563793068691e-05}
{"train_loss": 0.0593770295381546, "global_step": 208715, "epoch": 2345, "lr": 1.2654178313301612e-05}
{"train_loss": 0.037811487913131714, "global_step": 208716, "epoch": 2345, "lr": 1.2653792838555207e-05}
{"train_loss": 0.06848512589931488, "global_step": 208717, "epoch": 2345, "lr": 1.2653407368829562e-05}
{"train_loss": 0.10641198605298996, "global_step": 208718, "epoch": 2345, "lr": 1.2653021904124696e-05}
{"train_loss": 0.12034401297569275, "global_step": 208719, "epoch": 2345, "lr": 1.2652636444440691e-05}
{"train_loss": 0.09438876807689667, "global_step": 208720, "epoch": 2345, "lr": 1.265225098977757e-05}
{"train_loss": 0.027502883225679398, "global_step": 208721, "epoch": 2345, "lr": 1.2651865540135415e-05}
{"train_loss": 0.07837525755167007, "global_step": 208722, "epoch": 2345, "lr": 1.2651480095514245e-05}
{"train_loss": 0.051142457872629166, "global_step": 208723, "epoch": 2345, "lr": 1.2651094655914142e-05}
{"train_loss": 0.011661011725664139, "global_step": 208724, "epoch": 2345, "lr": 1.2650709221335127e-05}
{"train_loss": 0.04451100155711174, "global_step": 208725, "epoch": 2345, "lr": 1.2650323791777291e-05}
{"train_loss": 0.052941448986530304, "global_step": 208726, "epoch": 2345, "lr": 1.2649938367240644e-05}
{"train_loss": 0.0327119380235672, "global_step": 208727, "epoch": 2345, "lr": 1.2649552947725268e-05}
{"train_loss": 0.03702646121382713, "global_step": 208728, "epoch": 2345, "lr": 1.2649167533231198e-05}
{"train_loss": 0.06964633613824844, "global_step": 208729, "epoch": 2345, "lr": 1.2648782123758485e-05}
{"train_loss": 0.05566130205988884, "global_step": 208730, "epoch": 2345, "lr": 1.2648396719307209e-05}
{"train_loss": 0.0929449200630188, "global_step": 208731, "epoch": 2345, "lr": 1.2648011319877385e-05}
{"train_loss": 0.05184953287243843, "global_step": 208732, "epoch": 2345, "lr": 1.2647625925469092e-05}
{"train_loss": 0.07622029632329941, "global_step": 208733, "epoch": 2345, "lr": 1.2647240536082356e-05}
{"train_loss": 0.06362564861774445, "global_step": 208734, "epoch": 2345, "lr": 1.2646855151717263e-05}
{"train_loss": 0.04619808495044708, "global_step": 208735, "epoch": 2345, "lr": 1.2646469772373825e-05}
{"train_loss": 0.07610538601875305, "global_step": 208736, "epoch": 2345, "lr": 1.2646084398052128e-05}
{"train_loss": 0.04421508312225342, "global_step": 208737, "epoch": 2345, "lr": 1.26456990287522e-05}
{"train_loss": 0.023045098409056664, "global_step": 208738, "epoch": 2345, "lr": 1.2645313664474113e-05}
{"train_loss": 0.06679847836494446, "global_step": 208739, "epoch": 2345, "lr": 1.2644928305217895e-05}
{"train_loss": 0.04600498825311661, "global_step": 208740, "epoch": 2345, "lr": 1.264454295098363e-05}
{"train_loss": 0.03927404433488846, "global_step": 208741, "epoch": 2345, "lr": 1.2644157601771328e-05}
{"train_loss": 0.07338602840900421, "global_step": 208742, "epoch": 2345, "lr": 1.2643772257581082e-05}
{"train_loss": 0.06955686956644058, "global_step": 208743, "epoch": 2345, "lr": 1.2643386918412913e-05}
{"train_loss": 0.03922116756439209, "global_step": 208744, "epoch": 2345, "lr": 1.2643001584266895e-05}
{"train_loss": 0.03431619703769684, "global_step": 208745, "epoch": 2345, "lr": 1.2642616255143058e-05}
{"train_loss": 0.06628147512674332, "global_step": 208746, "epoch": 2345, "lr": 1.2642230931041483e-05}
{"train_loss": 0.042693089693784714, "global_step": 208747, "epoch": 2345, "lr": 1.2641845611962188e-05}
{"train_loss": 0.03396470099687576, "global_step": 208748, "epoch": 2345, "lr": 1.2641460297905255e-05}
{"train_loss": 0.058845460414886475, "global_step": 208749, "epoch": 2345, "lr": 1.2641074988870727e-05}
{"train_loss": 0.04909162595868111, "global_step": 208750, "epoch": 2345, "lr": 1.264068968485863e-05}
{"train_loss": 0.07169870287179947, "global_step": 208751, "epoch": 2345, "lr": 1.2640304385869057e-05}
{"train_loss": 0.03769850358366966, "global_step": 208752, "epoch": 2345, "lr": 1.2639919091902026e-05}
{"train_loss": 0.0855691060423851, "global_step": 208753, "epoch": 2345, "lr": 1.2639533802957615e-05}
{"train_loss": 0.04044234752655029, "global_step": 208754, "epoch": 2345, "lr": 1.2639148519035842e-05}
{"train_loss": 0.02223903499543667, "global_step": 208755, "epoch": 2345, "lr": 1.2638763240136802e-05}
{"train_loss": 0.03428196534514427, "global_step": 208756, "epoch": 2345, "lr": 1.2638377966260512e-05}
{"train_loss": 0.0640890970826149, "global_step": 208757, "epoch": 2345, "lr": 1.2637992697407047e-05}
{"train_loss": 0.035555120557546616, "global_step": 208758, "epoch": 2345, "lr": 1.2637607433576437e-05}
{"train_loss": 0.10901546478271484, "global_step": 208759, "epoch": 2345, "lr": 1.2637222174768754e-05}
{"train_loss": 0.10152189433574677, "global_step": 208760, "epoch": 2345, "lr": 1.263683692098403e-05}
{"train_loss": 0.03878668323159218, "global_step": 208761, "epoch": 2345, "lr": 1.2636451672222343e-05}
{"train_loss": 0.09689085185527802, "global_step": 208762, "epoch": 2345, "lr": 1.2636066428483717e-05}
{"train_loss": 0.019044451415538788, "global_step": 208763, "epoch": 2345, "lr": 1.2635681189768228e-05}
{"train_loss": 0.02709377184510231, "global_step": 208764, "epoch": 2345, "lr": 1.2635295956075904e-05}
{"train_loss": 0.10703963041305542, "global_step": 208765, "epoch": 2345, "lr": 1.2634910727406807e-05}
{"train_loss": 0.05761629343032837, "global_step": 208766, "epoch": 2345, "lr": 1.2634525503761008e-05}
{"train_loss": 0.07784826308488846, "global_step": 208767, "epoch": 2345, "lr": 1.2634140285138523e-05}
{"train_loss": 0.07461380958557129, "global_step": 208768, "epoch": 2345, "lr": 1.2633755071539438e-05}
{"train_loss": 0.05511883273720741, "global_step": 208769, "epoch": 2345, "lr": 1.2633369862963779e-05}
{"train_loss": 0.0685662031173706, "global_step": 208770, "epoch": 2345, "lr": 1.2632984659411618e-05}
{"train_loss": 0.04264095798134804, "global_step": 208771, "epoch": 2345, "lr": 1.2632599460882982e-05}
{"train_loss": 0.10654361546039581, "global_step": 208772, "epoch": 2345, "lr": 1.2632214267377956e-05}
{"train_loss": 0.05368948355317116, "global_step": 208773, "epoch": 2345, "lr": 1.2631829078896556e-05}
{"train_loss": 0.05835152789950371, "global_step": 208774, "epoch": 2345, "lr": 1.2631443895438866e-05}
{"train_loss": 0.07500166445970535, "global_step": 208775, "epoch": 2345, "lr": 1.2631058717004913e-05}
{"train_loss": 0.08051719516515732, "global_step": 208776, "epoch": 2345, "lr": 1.2630673543594768e-05}
{"train_loss": 0.08416561037302017, "global_step": 208777, "epoch": 2345, "lr": 1.263028837520846e-05}
{"train_loss": 0.07542401552200317, "global_step": 208778, "epoch": 2345, "lr": 1.2629903211846066e-05}
{"train_loss": 0.07627597451210022, "global_step": 208779, "epoch": 2345, "lr": 1.2629518053507616e-05}
{"train_loss": 0.08856305480003357, "global_step": 208780, "epoch": 2345, "lr": 1.262913290019318e-05}
{"train_loss": 0.055151764303445816, "global_step": 208781, "epoch": 2345, "lr": 1.2628747751902792e-05}
{"train_loss": 0.046896226704120636, "global_step": 208782, "epoch": 2345, "lr": 1.2628362608636529e-05}
{"train_loss": 0.06177287921309471, "global_step": 208783, "epoch": 2345, "lr": 1.2627977470394408e-05}
{"train_loss": 0.020447805523872375, "global_step": 208784, "epoch": 2345, "lr": 1.2627592337176519e-05}
{"train_loss": 0.040112972259521484, "global_step": 208785, "epoch": 2345, "lr": 1.2627207208982883e-05}
{"train_loss": 0.02113250084221363, "global_step": 208786, "epoch": 2345, "lr": 1.2626822085813573e-05}
{"train_loss": 0.05256449803709984, "global_step": 208787, "epoch": 2345, "lr": 1.2626436967668632e-05}
{"train_loss": 0.051266975700855255, "global_step": 208788, "epoch": 2345, "lr": 1.2626051854548093e-05}
{"train_loss": 0.0521162748336792, "global_step": 208789, "epoch": 2345, "lr": 1.2625666746452047e-05}
{"train_loss": 0.054185524582862854, "global_step": 208790, "epoch": 2345, "lr": 1.2625281643380505e-05}
{"train_loss": 0.08083871006965637, "global_step": 208791, "epoch": 2345, "lr": 1.2624896545333553e-05}
{"train_loss": 0.046191491186618805, "global_step": 208792, "epoch": 2345, "lr": 1.2624511452311216e-05}
{"train_loss": 0.0587263305702906, "global_step": 208793, "epoch": 2345, "lr": 1.2624126364313571e-05, "val_loss": 8.483556747436523, "train_action_mse_error": 35.66658020019531}
{"train_loss": 0.04523222893476486, "global_step": 208794, "epoch": 2346, "lr": 1.2623741281340645e-05}
{"train_loss": 0.09317462891340256, "global_step": 208795, "epoch": 2346, "lr": 1.262335620339251e-05}
{"train_loss": 0.028910305351018906, "global_step": 208796, "epoch": 2346, "lr": 1.2622971130469201e-05}
{"train_loss": 0.09080302715301514, "global_step": 208797, "epoch": 2346, "lr": 1.262258606257079e-05}
{"train_loss": 0.03423917293548584, "global_step": 208798, "epoch": 2346, "lr": 1.2622200999697303e-05}
{"train_loss": 0.05907939746975899, "global_step": 208799, "epoch": 2346, "lr": 1.2621815941848819e-05}
{"train_loss": 0.06481857597827911, "global_step": 208800, "epoch": 2346, "lr": 1.2621430889025359e-05}
{"train_loss": 0.05099649727344513, "global_step": 208801, "epoch": 2346, "lr": 1.2621045841226997e-05}
{"train_loss": 0.0823572501540184, "global_step": 208802, "epoch": 2346, "lr": 1.2620660798453793e-05}
{"train_loss": 0.0352107435464859, "global_step": 208803, "epoch": 2346, "lr": 1.2620275760705768e-05}
{"train_loss": 0.03428377956151962, "global_step": 208804, "epoch": 2346, "lr": 1.2619890727983009e-05}
{"train_loss": 0.06185978651046753, "global_step": 208805, "epoch": 2346, "lr": 1.2619505700285538e-05}
{"train_loss": 0.030388079583644867, "global_step": 208806, "epoch": 2346, "lr": 1.2619120677613428e-05}
{"train_loss": 0.038449108600616455, "global_step": 208807, "epoch": 2346, "lr": 1.2618735659966708e-05}
{"train_loss": 0.04248608276247978, "global_step": 208808, "epoch": 2346, "lr": 1.2618350647345461e-05}
{"train_loss": 0.0735960304737091, "global_step": 208809, "epoch": 2346, "lr": 1.2617965639749707e-05}
{"train_loss": 0.09303569048643112, "global_step": 208810, "epoch": 2346, "lr": 1.2617580637179521e-05}
{"train_loss": 0.04376835376024246, "global_step": 208811, "epoch": 2346, "lr": 1.2617195639634933e-05}
{"train_loss": 0.09809911996126175, "global_step": 208812, "epoch": 2346, "lr": 1.2616810647116023e-05}
{"train_loss": 0.08456408977508545, "global_step": 208813, "epoch": 2346, "lr": 1.2616425659622817e-05}
{"train_loss": 0.0397782102227211, "global_step": 208814, "epoch": 2346, "lr": 1.2616040677155382e-05}
{"train_loss": 0.0385250486433506, "global_step": 208815, "epoch": 2346, "lr": 1.2615655699713757e-05}
{"train_loss": 0.06130320578813553, "global_step": 208816, "epoch": 2346, "lr": 1.2615270727298017e-05}
{"train_loss": 0.03893360495567322, "global_step": 208817, "epoch": 2346, "lr": 1.261488575990818e-05}
{"train_loss": 0.1040911003947258, "global_step": 208818, "epoch": 2346, "lr": 1.2614500797544332e-05}
{"train_loss": 0.058564379811286926, "global_step": 208819, "epoch": 2346, "lr": 1.2614115840206491e-05}
{"train_loss": 0.0785399004817009, "global_step": 208820, "epoch": 2346, "lr": 1.2613730887894748e-05}
{"train_loss": 0.07476375252008438, "global_step": 208821, "epoch": 2346, "lr": 1.2613345940609112e-05}
{"train_loss": 0.03531738370656967, "global_step": 208822, "epoch": 2346, "lr": 1.261296099834967e-05}
{"train_loss": 0.08367698639631271, "global_step": 208823, "epoch": 2346, "lr": 1.2612576061116449e-05}
{"train_loss": 0.04639320820569992, "global_step": 208824, "epoch": 2346, "lr": 1.2612191128909529e-05}
{"train_loss": 0.05291196331381798, "global_step": 208825, "epoch": 2346, "lr": 1.2611806201728926e-05}
{"train_loss": 0.023037873208522797, "global_step": 208826, "epoch": 2346, "lr": 1.2611421279574725e-05}
{"train_loss": 0.055158209055662155, "global_step": 208827, "epoch": 2346, "lr": 1.261103636244696e-05}
{"train_loss": 0.04085148498415947, "global_step": 208828, "epoch": 2346, "lr": 1.2610651450345673e-05}
{"train_loss": 0.02385988086462021, "global_step": 208829, "epoch": 2346, "lr": 1.2610266543270943e-05}
{"train_loss": 0.05655503645539284, "global_step": 208830, "epoch": 2346, "lr": 1.2609881641222793e-05}
{"train_loss": 0.03066953271627426, "global_step": 208831, "epoch": 2346, "lr": 1.2609496744201298e-05}
{"train_loss": 0.03525266423821449, "global_step": 208832, "epoch": 2346, "lr": 1.260911185220649e-05}
{"train_loss": 0.09092868864536285, "global_step": 208833, "epoch": 2346, "lr": 1.260872696523845e-05}
{"train_loss": 0.0698036476969719, "global_step": 208834, "epoch": 2346, "lr": 1.2608342083297192e-05}
{"train_loss": 0.08650145679712296, "global_step": 208835, "epoch": 2346, "lr": 1.2607957206382797e-05}
{"train_loss": 0.04171912372112274, "global_step": 208836, "epoch": 2346, "lr": 1.2607572334495299e-05}
{"train_loss": 0.06947368383407593, "global_step": 208837, "epoch": 2346, "lr": 1.2607187467634757e-05}
{"train_loss": 0.04623568430542946, "global_step": 208838, "epoch": 2346, "lr": 1.2606802605801232e-05}
{"train_loss": 0.08266745507717133, "global_step": 208839, "epoch": 2346, "lr": 1.2606417748994754e-05}
{"train_loss": 0.0354088619351387, "global_step": 208840, "epoch": 2346, "lr": 1.2606032897215402e-05}
{"train_loss": 0.06406978517770767, "global_step": 208841, "epoch": 2346, "lr": 1.2605648050463204e-05}
{"train_loss": 0.06400318443775177, "global_step": 208842, "epoch": 2346, "lr": 1.2605263208738227e-05}
{"train_loss": 0.039814144372940063, "global_step": 208843, "epoch": 2346, "lr": 1.2604878372040508e-05}
{"train_loss": 0.03906583786010742, "global_step": 208844, "epoch": 2346, "lr": 1.2604493540370122e-05}
{"train_loss": 0.08354266732931137, "global_step": 208845, "epoch": 2346, "lr": 1.2604108713727087e-05}
{"train_loss": 0.08813992887735367, "global_step": 208846, "epoch": 2346, "lr": 1.260372389211149e-05}
{"train_loss": 0.059150006622076035, "global_step": 208847, "epoch": 2346, "lr": 1.2603339075523352e-05}
{"train_loss": 0.04028813913464546, "global_step": 208848, "epoch": 2346, "lr": 1.2602954263962758e-05}
{"train_loss": 0.030026119202375412, "global_step": 208849, "epoch": 2346, "lr": 1.260256945742972e-05}
{"train_loss": 0.08913005888462067, "global_step": 208850, "epoch": 2346, "lr": 1.2602184655924326e-05}
{"train_loss": 0.10536742210388184, "global_step": 208851, "epoch": 2346, "lr": 1.26017998594466e-05}
{"train_loss": 0.06919675320386887, "global_step": 208852, "epoch": 2346, "lr": 1.2601415067996624e-05}
{"train_loss": 0.09785681217908859, "global_step": 208853, "epoch": 2346, "lr": 1.2601030281574411e-05}
{"train_loss": 0.04452339932322502, "global_step": 208854, "epoch": 2346, "lr": 1.2600645500180053e-05}
{"train_loss": 0.04809385538101196, "global_step": 208855, "epoch": 2346, "lr": 1.2600260723813568e-05}
{"train_loss": 0.06019644811749458, "global_step": 208856, "epoch": 2346, "lr": 1.2599875952475021e-05}
{"train_loss": 0.0930560976266861, "global_step": 208857, "epoch": 2346, "lr": 1.2599491186164475e-05}
{"train_loss": 0.06658504158258438, "global_step": 208858, "epoch": 2346, "lr": 1.2599106424881963e-05}
{"train_loss": 0.01999565027654171, "global_step": 208859, "epoch": 2346, "lr": 1.2598721668627555e-05}
{"train_loss": 0.06040308624505997, "global_step": 208860, "epoch": 2346, "lr": 1.2598336917401298e-05}
{"train_loss": 0.044946882873773575, "global_step": 208861, "epoch": 2346, "lr": 1.259795217120322e-05}
{"train_loss": 0.026889942586421967, "global_step": 208862, "epoch": 2346, "lr": 1.2597567430033407e-05}
{"train_loss": 0.042527057230472565, "global_step": 208863, "epoch": 2346, "lr": 1.2597182693891885e-05}
{"train_loss": 0.030905406922101974, "global_step": 208864, "epoch": 2346, "lr": 1.2596797962778722e-05}
{"train_loss": 0.07247280329465866, "global_step": 208865, "epoch": 2346, "lr": 1.2596413236693971e-05}
{"train_loss": 0.05030127614736557, "global_step": 208866, "epoch": 2346, "lr": 1.2596028515637659e-05}
{"train_loss": 0.03708307817578316, "global_step": 208867, "epoch": 2346, "lr": 1.2595643799609869e-05}
{"train_loss": 0.0276216808706522, "global_step": 208868, "epoch": 2346, "lr": 1.2595259088610628e-05}
{"train_loss": 0.08207526803016663, "global_step": 208869, "epoch": 2346, "lr": 1.259487438264001e-05}
{"train_loss": 0.11370225250720978, "global_step": 208870, "epoch": 2346, "lr": 1.2594489681698041e-05}
{"train_loss": 0.12449057400226593, "global_step": 208871, "epoch": 2346, "lr": 1.2594104985784799e-05}
{"train_loss": 0.052014756947755814, "global_step": 208872, "epoch": 2346, "lr": 1.2593720294900313e-05}
{"train_loss": 0.07339964807033539, "global_step": 208873, "epoch": 2346, "lr": 1.2593335609044659e-05}
{"train_loss": 0.04566369205713272, "global_step": 208874, "epoch": 2346, "lr": 1.2592950928217861e-05}
{"train_loss": 0.0794011577963829, "global_step": 208875, "epoch": 2346, "lr": 1.259256625241998e-05}
{"train_loss": 0.07119537144899368, "global_step": 208876, "epoch": 2346, "lr": 1.2592181581651092e-05}
{"train_loss": 0.06071089953184128, "global_step": 208877, "epoch": 2346, "lr": 1.2591796915911213e-05}
{"train_loss": 0.08568211644887924, "global_step": 208878, "epoch": 2346, "lr": 1.259141225520043e-05}
{"train_loss": 0.09041329473257065, "global_step": 208879, "epoch": 2346, "lr": 1.2591027599518757e-05}
{"train_loss": 0.031542032957077026, "global_step": 208880, "epoch": 2346, "lr": 1.2590642948866276e-05}
{"train_loss": 0.08447088301181793, "global_step": 208881, "epoch": 2346, "lr": 1.2590258303243018e-05}
{"train_loss": 0.05983152366086338, "global_step": 208882, "epoch": 2346, "lr": 1.2589873662649054e-05, "val_loss": 8.596892356872559}
{"train_loss": 0.07933895289897919, "global_step": 208883, "epoch": 2347, "lr": 1.258948902708441e-05}
{"train_loss": 0.0574054941534996, "global_step": 208884, "epoch": 2347, "lr": 1.2589104396549173e-05}
{"train_loss": 0.05167178437113762, "global_step": 208885, "epoch": 2347, "lr": 1.258871977104336e-05}
{"train_loss": 0.03797725588083267, "global_step": 208886, "epoch": 2347, "lr": 1.2588335150567049e-05}
{"train_loss": 0.04566221684217453, "global_step": 208887, "epoch": 2347, "lr": 1.2587950535120268e-05}
{"train_loss": 0.05047166347503662, "global_step": 208888, "epoch": 2347, "lr": 1.2587565924703092e-05}
{"train_loss": 0.060985028743743896, "global_step": 208889, "epoch": 2347, "lr": 1.2587181319315555e-05}
{"train_loss": 0.053486939519643784, "global_step": 208890, "epoch": 2347, "lr": 1.2586796718957722e-05}
{"train_loss": 0.05157581344246864, "global_step": 208891, "epoch": 2347, "lr": 1.2586412123629625e-05}
{"train_loss": 0.02127276360988617, "global_step": 208892, "epoch": 2347, "lr": 1.2586027533331336e-05}
{"train_loss": 0.0565694235265255, "global_step": 208893, "epoch": 2347, "lr": 1.2585642948062909e-05}
{"train_loss": 0.09823621809482574, "global_step": 208894, "epoch": 2347, "lr": 1.258525836782437e-05}
{"train_loss": 0.11594774574041367, "global_step": 208895, "epoch": 2347, "lr": 1.2584873792615803e-05}
{"train_loss": 0.08373261988162994, "global_step": 208896, "epoch": 2347, "lr": 1.258448922243723e-05}
{"train_loss": 0.08339148759841919, "global_step": 208897, "epoch": 2347, "lr": 1.2584104657288725e-05}
{"train_loss": 0.029437366873025894, "global_step": 208898, "epoch": 2347, "lr": 1.2583720097170336e-05}
{"train_loss": 0.04293579235672951, "global_step": 208899, "epoch": 2347, "lr": 1.2583335542082097e-05}
{"train_loss": 0.04042235389351845, "global_step": 208900, "epoch": 2347, "lr": 1.258295099202409e-05}
{"train_loss": 0.0764082819223404, "global_step": 208901, "epoch": 2347, "lr": 1.2582566446996324e-05}
{"train_loss": 0.07746583968400955, "global_step": 208902, "epoch": 2347, "lr": 1.2582181906998902e-05}
{"train_loss": 0.026777509599924088, "global_step": 208903, "epoch": 2347, "lr": 1.258179737203184e-05}
{"train_loss": 0.04016182944178581, "global_step": 208904, "epoch": 2347, "lr": 1.2581412842095191e-05}
{"train_loss": 0.04477023333311081, "global_step": 208905, "epoch": 2347, "lr": 1.2581028317189026e-05}
{"train_loss": 0.06862663477659225, "global_step": 208906, "epoch": 2347, "lr": 1.2580643797313374e-05}
{"train_loss": 0.06516362726688385, "global_step": 208907, "epoch": 2347, "lr": 1.258025928246831e-05}
{"train_loss": 0.04646241292357445, "global_step": 208908, "epoch": 2347, "lr": 1.257987477265386e-05}
{"train_loss": 0.07849971204996109, "global_step": 208909, "epoch": 2347, "lr": 1.2579490267870109e-05}
{"train_loss": 0.04585729166865349, "global_step": 208910, "epoch": 2347, "lr": 1.2579105768117066e-05}
{"train_loss": 0.11761172860860825, "global_step": 208911, "epoch": 2347, "lr": 1.2578721273394816e-05}
{"train_loss": 0.06006913632154465, "global_step": 208912, "epoch": 2347, "lr": 1.2578336783703409e-05}
{"train_loss": 0.05323314294219017, "global_step": 208913, "epoch": 2347, "lr": 1.2577952299042877e-05}
{"train_loss": 0.054789166897535324, "global_step": 208914, "epoch": 2347, "lr": 1.2577567819413294e-05}
{"train_loss": 0.02239062637090683, "global_step": 208915, "epoch": 2347, "lr": 1.2577183344814691e-05}
{"train_loss": 0.044280003756284714, "global_step": 208916, "epoch": 2347, "lr": 1.2576798875247143e-05}
{"train_loss": 0.08055586367845535, "global_step": 208917, "epoch": 2347, "lr": 1.2576414410710669e-05}
{"train_loss": 0.09130600839853287, "global_step": 208918, "epoch": 2347, "lr": 1.257602995120536e-05}
{"train_loss": 0.0872400775551796, "global_step": 208919, "epoch": 2347, "lr": 1.257564549673123e-05}
{"train_loss": 0.0926583781838417, "global_step": 208920, "epoch": 2347, "lr": 1.2575261047288362e-05}
{"train_loss": 0.05746112018823624, "global_step": 208921, "epoch": 2347, "lr": 1.2574876602876784e-05}
{"train_loss": 0.029859865084290504, "global_step": 208922, "epoch": 2347, "lr": 1.257449216349657e-05}
{"train_loss": 0.10726414620876312, "global_step": 208923, "epoch": 2347, "lr": 1.2574107729147745e-05}
{"train_loss": 0.03345029801130295, "global_step": 208924, "epoch": 2347, "lr": 1.2573723299830387e-05}
{"train_loss": 0.07405871152877808, "global_step": 208925, "epoch": 2347, "lr": 1.2573338875544521e-05}
{"train_loss": 0.06331869959831238, "global_step": 208926, "epoch": 2347, "lr": 1.257295445629023e-05}
{"train_loss": 0.0812942236661911, "global_step": 208927, "epoch": 2347, "lr": 1.2572570042067539e-05}
{"train_loss": 0.06341522186994553, "global_step": 208928, "epoch": 2347, "lr": 1.2572185632876505e-05}
{"train_loss": 0.06612364947795868, "global_step": 208929, "epoch": 2347, "lr": 1.25718012287172e-05}
{"train_loss": 0.0638231560587883, "global_step": 208930, "epoch": 2347, "lr": 1.2571416829589644e-05}
{"train_loss": 0.06755241006612778, "global_step": 208931, "epoch": 2347, "lr": 1.2571032435493924e-05}
{"train_loss": 0.06565696001052856, "global_step": 208932, "epoch": 2347, "lr": 1.2570648046430051e-05}
{"train_loss": 0.0309278666973114, "global_step": 208933, "epoch": 2347, "lr": 1.257026366239812e-05}
{"train_loss": 0.11692114919424057, "global_step": 208934, "epoch": 2347, "lr": 1.256987928339814e-05}
{"train_loss": 0.06659519672393799, "global_step": 208935, "epoch": 2347, "lr": 1.2569494909430201e-05}
{"train_loss": 0.03728165104985237, "global_step": 208936, "epoch": 2347, "lr": 1.2569110540494323e-05}
{"train_loss": 0.06220921874046326, "global_step": 208937, "epoch": 2347, "lr": 1.2568726176590584e-05}
{"train_loss": 0.05184759944677353, "global_step": 208938, "epoch": 2347, "lr": 1.2568341817719026e-05}
{"train_loss": 0.03174171224236488, "global_step": 208939, "epoch": 2347, "lr": 1.2567957463879682e-05}
{"train_loss": 0.05158734321594238, "global_step": 208940, "epoch": 2347, "lr": 1.2567573115072634e-05}
{"train_loss": 0.07345786690711975, "global_step": 208941, "epoch": 2347, "lr": 1.2567188771297905e-05}
{"train_loss": 0.07429514825344086, "global_step": 208942, "epoch": 2347, "lr": 1.256680443255558e-05}
{"train_loss": 0.017072768881917, "global_step": 208943, "epoch": 2347, "lr": 1.2566420098845687e-05}
{"train_loss": 0.025809742510318756, "global_step": 208944, "epoch": 2347, "lr": 1.2566035770168272e-05}
{"train_loss": 0.08354880660772324, "global_step": 208945, "epoch": 2347, "lr": 1.2565651446523408e-05}
{"train_loss": 0.0652567520737648, "global_step": 208946, "epoch": 2347, "lr": 1.2565267127911124e-05}
{"train_loss": 0.02445339784026146, "global_step": 208947, "epoch": 2347, "lr": 1.2564882814331485e-05}
{"train_loss": 0.09721803665161133, "global_step": 208948, "epoch": 2347, "lr": 1.256449850578455e-05}
{"train_loss": 0.046829212456941605, "global_step": 208949, "epoch": 2347, "lr": 1.2564114202270355e-05}
{"train_loss": 0.1636446714401245, "global_step": 208950, "epoch": 2347, "lr": 1.2563729903788968e-05}
{"train_loss": 0.05506698042154312, "global_step": 208951, "epoch": 2347, "lr": 1.2563345610340415e-05}
{"train_loss": 0.051812414079904556, "global_step": 208952, "epoch": 2347, "lr": 1.2562961321924782e-05}
{"train_loss": 0.04742894321680069, "global_step": 208953, "epoch": 2347, "lr": 1.2562577038542089e-05}
{"train_loss": 0.07559458166360855, "global_step": 208954, "epoch": 2347, "lr": 1.2562192760192415e-05}
{"train_loss": 0.052590493112802505, "global_step": 208955, "epoch": 2347, "lr": 1.256180848687578e-05}
{"train_loss": 0.041281793266534805, "global_step": 208956, "epoch": 2347, "lr": 1.2561424218592272e-05}
{"train_loss": 0.030819926410913467, "global_step": 208957, "epoch": 2347, "lr": 1.2561039955341907e-05}
{"train_loss": 0.08379139751195908, "global_step": 208958, "epoch": 2347, "lr": 1.2560655697124768e-05}
{"train_loss": 0.04663714021444321, "global_step": 208959, "epoch": 2347, "lr": 1.256027144394088e-05}
{"train_loss": 0.04202880337834358, "global_step": 208960, "epoch": 2347, "lr": 1.2559887195790322e-05}
{"train_loss": 0.08763179928064346, "global_step": 208961, "epoch": 2347, "lr": 1.2559502952673114e-05}
{"train_loss": 0.06817197054624557, "global_step": 208962, "epoch": 2347, "lr": 1.2559118714589341e-05}
{"train_loss": 0.06274557113647461, "global_step": 208963, "epoch": 2347, "lr": 1.255873448153902e-05}
{"train_loss": 0.07715285569429398, "global_step": 208964, "epoch": 2347, "lr": 1.2558350253522228e-05}
{"train_loss": 0.05724693834781647, "global_step": 208965, "epoch": 2347, "lr": 1.255796603053902e-05}
{"train_loss": 0.07888979464769363, "global_step": 208966, "epoch": 2347, "lr": 1.255758181258942e-05}
{"train_loss": 0.08137025684118271, "global_step": 208967, "epoch": 2347, "lr": 1.2557197599673515e-05}
{"train_loss": 0.15542617440223694, "global_step": 208968, "epoch": 2347, "lr": 1.2556813391791323e-05}
{"train_loss": 0.062140434980392456, "global_step": 208969, "epoch": 2347, "lr": 1.2556429188942925e-05}
{"train_loss": 0.119561567902565, "global_step": 208970, "epoch": 2347, "lr": 1.255604499112834e-05}
{"train_loss": 0.06426548949453269, "global_step": 208971, "epoch": 2347, "lr": 1.2555660798347663e-05, "val_loss": 8.56678295135498}
{"train_loss": 0.047410864382982254, "global_step": 208972, "epoch": 2348, "lr": 1.2555276610600897e-05}
{"train_loss": 0.08987194299697876, "global_step": 208973, "epoch": 2348, "lr": 1.2554892427888138e-05}
{"train_loss": 0.01759207248687744, "global_step": 208974, "epoch": 2348, "lr": 1.25545082502094e-05}
{"train_loss": 0.03285631537437439, "global_step": 208975, "epoch": 2348, "lr": 1.2554124077564767e-05}
{"train_loss": 0.0680040493607521, "global_step": 208976, "epoch": 2348, "lr": 1.2553739909954276e-05}
{"train_loss": 0.05231226235628128, "global_step": 208977, "epoch": 2348, "lr": 1.2553355747377965e-05}
{"train_loss": 0.0444660410284996, "global_step": 208978, "epoch": 2348, "lr": 1.255297158983591e-05}
{"train_loss": 0.06425456702709198, "global_step": 208979, "epoch": 2348, "lr": 1.2552587437328139e-05}
{"train_loss": 0.07046416401863098, "global_step": 208980, "epoch": 2348, "lr": 1.255220328985473e-05}
{"train_loss": 0.0505012646317482, "global_step": 208981, "epoch": 2348, "lr": 1.2551819147415723e-05}
{"train_loss": 0.03472936898469925, "global_step": 208982, "epoch": 2348, "lr": 1.2551435010011154e-05}
{"train_loss": 0.023809928447008133, "global_step": 208983, "epoch": 2348, "lr": 1.2551050877641085e-05}
{"train_loss": 0.04298106208443642, "global_step": 208984, "epoch": 2348, "lr": 1.2550666750305585e-05}
{"train_loss": 0.020125236362218857, "global_step": 208985, "epoch": 2348, "lr": 1.2550282628004677e-05}
{"train_loss": 0.061651863157749176, "global_step": 208986, "epoch": 2348, "lr": 1.2549898510738439e-05}
{"train_loss": 0.06328250467777252, "global_step": 208987, "epoch": 2348, "lr": 1.2549514398506902e-05}
{"train_loss": 0.13140340149402618, "global_step": 208988, "epoch": 2348, "lr": 1.254913029131014e-05}
{"train_loss": 0.06211642175912857, "global_step": 208989, "epoch": 2348, "lr": 1.2548746189148174e-05}
{"train_loss": 0.08215415477752686, "global_step": 208990, "epoch": 2348, "lr": 1.2548362092021088e-05}
{"train_loss": 0.0456983745098114, "global_step": 208991, "epoch": 2348, "lr": 1.2547977999928906e-05}
{"train_loss": 0.07561367005109787, "global_step": 208992, "epoch": 2348, "lr": 1.2547593912871703e-05}
{"train_loss": 0.06979051232337952, "global_step": 208993, "epoch": 2348, "lr": 1.2547209830849504e-05}
{"train_loss": 0.03658151626586914, "global_step": 208994, "epoch": 2348, "lr": 1.2546825753862396e-05}
{"train_loss": 0.054353103041648865, "global_step": 208995, "epoch": 2348, "lr": 1.2546441681910392e-05}
{"train_loss": 0.062473926693201065, "global_step": 208996, "epoch": 2348, "lr": 1.2546057614993578e-05}
{"train_loss": 0.11335354298353195, "global_step": 208997, "epoch": 2348, "lr": 1.2545673553111974e-05}
{"train_loss": 0.05547823756933212, "global_step": 208998, "epoch": 2348, "lr": 1.2545289496265667e-05}
{"train_loss": 0.04355783015489578, "global_step": 208999, "epoch": 2348, "lr": 1.2544905444454668e-05}
{"train_loss": 0.02751436084508896, "global_step": 209000, "epoch": 2348, "lr": 1.2544521397679071e-05}
{"train_loss": 0.035957496613264084, "global_step": 209001, "epoch": 2348, "lr": 1.254413735593889e-05}
{"train_loss": 0.08648595958948135, "global_step": 209002, "epoch": 2348, "lr": 1.2543753319234192e-05}
{"train_loss": 0.05135403200984001, "global_step": 209003, "epoch": 2348, "lr": 1.2543369287565044e-05}
{"train_loss": 0.06035195291042328, "global_step": 209004, "epoch": 2348, "lr": 1.2542985260931466e-05}
{"train_loss": 0.09643998742103577, "global_step": 209005, "epoch": 2348, "lr": 1.2542601239333545e-05}
{"train_loss": 0.054938867688179016, "global_step": 209006, "epoch": 2348, "lr": 1.2542217222771301e-05}
{"train_loss": 0.09579670429229736, "global_step": 209007, "epoch": 2348, "lr": 1.2541833211244814e-05}
{"train_loss": 0.03473856672644615, "global_step": 209008, "epoch": 2348, "lr": 1.2541449204754108e-05}
{"train_loss": 0.04505511745810509, "global_step": 209009, "epoch": 2348, "lr": 1.2541065203299258e-05}
{"train_loss": 0.06566090136766434, "global_step": 209010, "epoch": 2348, "lr": 1.2540681206880294e-05}
{"train_loss": 0.032451122999191284, "global_step": 209011, "epoch": 2348, "lr": 1.2540297215497294e-05}
{"train_loss": 0.07381223887205124, "global_step": 209012, "epoch": 2348, "lr": 1.2539913229150273e-05}
{"train_loss": 0.059494175016880035, "global_step": 209013, "epoch": 2348, "lr": 1.2539529247839327e-05}
{"train_loss": 0.08258217573165894, "global_step": 209014, "epoch": 2348, "lr": 1.2539145271564478e-05}
{"train_loss": 0.009818296879529953, "global_step": 209015, "epoch": 2348, "lr": 1.2538761300325775e-05}
{"train_loss": 0.07877045124769211, "global_step": 209016, "epoch": 2348, "lr": 1.2538377334123292e-05}
{"train_loss": 0.05142153799533844, "global_step": 209017, "epoch": 2348, "lr": 1.2537993372957052e-05}
{"train_loss": 0.05769360810518265, "global_step": 209018, "epoch": 2348, "lr": 1.2537609416827139e-05}
{"train_loss": 0.10369016230106354, "global_step": 209019, "epoch": 2348, "lr": 1.253722546573357e-05}
{"train_loss": 0.05670800805091858, "global_step": 209020, "epoch": 2348, "lr": 1.2536841519676428e-05}
{"train_loss": 0.06848900765180588, "global_step": 209021, "epoch": 2348, "lr": 1.2536457578655742e-05}
{"train_loss": 0.07075619697570801, "global_step": 209022, "epoch": 2348, "lr": 1.2536073642671586e-05}
{"train_loss": 0.059034544974565506, "global_step": 209023, "epoch": 2348, "lr": 1.253568971172398e-05}
{"train_loss": 0.05342988297343254, "global_step": 209024, "epoch": 2348, "lr": 1.2535305785813011e-05}
{"train_loss": 0.04278912767767906, "global_step": 209025, "epoch": 2348, "lr": 1.2534921864938699e-05}
{"train_loss": 0.06557001173496246, "global_step": 209026, "epoch": 2348, "lr": 1.2534537949101122e-05}
{"train_loss": 0.03107011877000332, "global_step": 209027, "epoch": 2348, "lr": 1.2534154038300311e-05}
{"train_loss": 0.06125606596469879, "global_step": 209028, "epoch": 2348, "lr": 1.2533770132536333e-05}
{"train_loss": 0.05474449694156647, "global_step": 209029, "epoch": 2348, "lr": 1.2533386231809224e-05}
{"train_loss": 0.07206983864307404, "global_step": 209030, "epoch": 2348, "lr": 1.2533002336119059e-05}
{"train_loss": 0.054681144654750824, "global_step": 209031, "epoch": 2348, "lr": 1.253261844546586e-05}
{"train_loss": 0.09202057868242264, "global_step": 209032, "epoch": 2348, "lr": 1.2532234559849704e-05}
{"train_loss": 0.07808688282966614, "global_step": 209033, "epoch": 2348, "lr": 1.2531850679270624e-05}
{"train_loss": 0.062403302639722824, "global_step": 209034, "epoch": 2348, "lr": 1.2531466803728692e-05}
{"train_loss": 0.05917825922369957, "global_step": 209035, "epoch": 2348, "lr": 1.2531082933223932e-05}
{"train_loss": 0.0496593713760376, "global_step": 209036, "epoch": 2348, "lr": 1.253069906775643e-05}
{"train_loss": 0.041662395000457764, "global_step": 209037, "epoch": 2348, "lr": 1.2530315207326198e-05}
{"train_loss": 0.045664072036743164, "global_step": 209038, "epoch": 2348, "lr": 1.2529931351933311e-05}
{"train_loss": 0.033198971301317215, "global_step": 209039, "epoch": 2348, "lr": 1.2529547501577832e-05}
{"train_loss": 0.05389014631509781, "global_step": 209040, "epoch": 2348, "lr": 1.2529163656259785e-05}
{"train_loss": 0.082210473716259, "global_step": 209041, "epoch": 2348, "lr": 1.2528779815979253e-05}
{"train_loss": 0.03197537735104561, "global_step": 209042, "epoch": 2348, "lr": 1.2528395980736252e-05}
{"train_loss": 0.061631083488464355, "global_step": 209043, "epoch": 2348, "lr": 1.2528012150530866e-05}
{"train_loss": 0.046245768666267395, "global_step": 209044, "epoch": 2348, "lr": 1.2527628325363117e-05}
{"train_loss": 0.14513888955116272, "global_step": 209045, "epoch": 2348, "lr": 1.2527244505233087e-05}
{"train_loss": 0.030906688421964645, "global_step": 209046, "epoch": 2348, "lr": 1.2526860690140796e-05}
{"train_loss": 0.0509146973490715, "global_step": 209047, "epoch": 2348, "lr": 1.252647688008633e-05}
{"train_loss": 0.03853990510106087, "global_step": 209048, "epoch": 2348, "lr": 1.2526093075069706e-05}
{"train_loss": 0.07228624075651169, "global_step": 209049, "epoch": 2348, "lr": 1.2525709275091008e-05}
{"train_loss": 0.0307314470410347, "global_step": 209050, "epoch": 2348, "lr": 1.2525325480150252e-05}
{"train_loss": 0.10979925841093063, "global_step": 209051, "epoch": 2348, "lr": 1.2524941690247533e-05}
{"train_loss": 0.052951548248529434, "global_step": 209052, "epoch": 2348, "lr": 1.2524557905382856e-05}
{"train_loss": 0.060654111206531525, "global_step": 209053, "epoch": 2348, "lr": 1.2524174125556315e-05}
{"train_loss": 0.048981960862874985, "global_step": 209054, "epoch": 2348, "lr": 1.2523790350767939e-05}
{"train_loss": 0.047426145523786545, "global_step": 209055, "epoch": 2348, "lr": 1.2523406581017765e-05}
{"train_loss": 0.04493933171033859, "global_step": 209056, "epoch": 2348, "lr": 1.2523022816305886e-05}
{"train_loss": 0.03595738112926483, "global_step": 209057, "epoch": 2348, "lr": 1.2522639056632307e-05}
{"train_loss": 0.06700172275304794, "global_step": 209058, "epoch": 2348, "lr": 1.2522255301997121e-05}
{"train_loss": 0.040353380143642426, "global_step": 209059, "epoch": 2348, "lr": 1.2521871552400343e-05}
{"train_loss": 0.05808355129752936, "global_step": 209060, "epoch": 2348, "lr": 1.2521487807842059e-05, "val_loss": 8.510065078735352}
{"train_loss": 0.03809734806418419, "global_step": 209061, "epoch": 2349, "lr": 1.2521104068322293e-05}
{"train_loss": 0.07898161560297012, "global_step": 209062, "epoch": 2349, "lr": 1.2520720333841117e-05}
{"train_loss": 0.04856511577963829, "global_step": 209063, "epoch": 2349, "lr": 1.2520336604398558e-05}
{"train_loss": 0.04084780067205429, "global_step": 209064, "epoch": 2349, "lr": 1.2519952879994702e-05}
{"train_loss": 0.04826125130057335, "global_step": 209065, "epoch": 2349, "lr": 1.2519569160629568e-05}
{"train_loss": 0.06535772979259491, "global_step": 209066, "epoch": 2349, "lr": 1.2519185446303233e-05}
{"train_loss": 0.07104747742414474, "global_step": 209067, "epoch": 2349, "lr": 1.2518801737015723e-05}
{"train_loss": 0.06046370789408684, "global_step": 209068, "epoch": 2349, "lr": 1.2518418032767115e-05}
{"train_loss": 0.029045237228274345, "global_step": 209069, "epoch": 2349, "lr": 1.2518034333557438e-05}
{"train_loss": 0.077601857483387, "global_step": 209070, "epoch": 2349, "lr": 1.251765063938677e-05}
{"train_loss": 0.05005969479680061, "global_step": 209071, "epoch": 2349, "lr": 1.2517266950255125e-05}
{"train_loss": 0.08377842605113983, "global_step": 209072, "epoch": 2349, "lr": 1.25168832661626e-05}
{"train_loss": 0.039553530514240265, "global_step": 209073, "epoch": 2349, "lr": 1.2516499587109204e-05}
{"train_loss": 0.032347749918699265, "global_step": 209074, "epoch": 2349, "lr": 1.2516115913095005e-05}
{"train_loss": 0.05296103656291962, "global_step": 209075, "epoch": 2349, "lr": 1.2515732244120076e-05}
{"train_loss": 0.05995994061231613, "global_step": 209076, "epoch": 2349, "lr": 1.2515348580184433e-05}
{"train_loss": 0.041040629148483276, "global_step": 209077, "epoch": 2349, "lr": 1.2514964921288163e-05}
{"train_loss": 0.060817353427410126, "global_step": 209078, "epoch": 2349, "lr": 1.251458126743128e-05}
{"train_loss": 0.05670551210641861, "global_step": 209079, "epoch": 2349, "lr": 1.2514197618613877e-05}
{"train_loss": 0.02016080915927887, "global_step": 209080, "epoch": 2349, "lr": 1.2513813974835958e-05}
{"train_loss": 0.03396814689040184, "global_step": 209081, "epoch": 2349, "lr": 1.251343033609762e-05}
{"train_loss": 0.04483571648597717, "global_step": 209082, "epoch": 2349, "lr": 1.2513046702398878e-05}
{"train_loss": 0.0683722272515297, "global_step": 209083, "epoch": 2349, "lr": 1.2512663073739817e-05}
{"train_loss": 0.047633904963731766, "global_step": 209084, "epoch": 2349, "lr": 1.2512279450120457e-05}
{"train_loss": 0.07453981041908264, "global_step": 209085, "epoch": 2349, "lr": 1.2511895831540877e-05}
{"train_loss": 0.010033655911684036, "global_step": 209086, "epoch": 2349, "lr": 1.2511512218001103e-05}
{"train_loss": 0.03778033331036568, "global_step": 209087, "epoch": 2349, "lr": 1.251112860950121e-05}
{"train_loss": 0.051386330276727676, "global_step": 209088, "epoch": 2349, "lr": 1.251074500604123e-05}
{"train_loss": 0.05885209143161774, "global_step": 209089, "epoch": 2349, "lr": 1.2510361407621234e-05}
{"train_loss": 0.09591501206159592, "global_step": 209090, "epoch": 2349, "lr": 1.2509977814241247e-05}
{"train_loss": 0.1268497258424759, "global_step": 209091, "epoch": 2349, "lr": 1.2509594225901356e-05}
{"train_loss": 0.12299930304288864, "global_step": 209092, "epoch": 2349, "lr": 1.2509210642601592e-05}
{"train_loss": 0.05760053172707558, "global_step": 209093, "epoch": 2349, "lr": 1.2508827064341988e-05}
{"train_loss": 0.06348209083080292, "global_step": 209094, "epoch": 2349, "lr": 1.2508443491122635e-05}
{"train_loss": 0.02185683511197567, "global_step": 209095, "epoch": 2349, "lr": 1.250805992294355e-05}
{"train_loss": 0.04964796453714371, "global_step": 209096, "epoch": 2349, "lr": 1.250767635980481e-05}
{"train_loss": 0.05426112562417984, "global_step": 209097, "epoch": 2349, "lr": 1.2507292801706443e-05}
{"train_loss": 0.10691697895526886, "global_step": 209098, "epoch": 2349, "lr": 1.250690924864853e-05}
{"train_loss": 0.08342118561267853, "global_step": 209099, "epoch": 2349, "lr": 1.2506525700631095e-05}
{"train_loss": 0.06059304252266884, "global_step": 209100, "epoch": 2349, "lr": 1.2506142157654211e-05}
{"train_loss": 0.015011342242360115, "global_step": 209101, "epoch": 2349, "lr": 1.25057586197179e-05}
{"train_loss": 0.06646999716758728, "global_step": 209102, "epoch": 2349, "lr": 1.2505375086822252e-05}
{"train_loss": 0.05226478353142738, "global_step": 209103, "epoch": 2349, "lr": 1.2504991558967288e-05}
{"train_loss": 0.09801315516233444, "global_step": 209104, "epoch": 2349, "lr": 1.2504608036153081e-05}
{"train_loss": 0.0519857257604599, "global_step": 209105, "epoch": 2349, "lr": 1.250422451837966e-05}
{"train_loss": 0.09020935744047165, "global_step": 209106, "epoch": 2349, "lr": 1.2503841005647098e-05}
{"train_loss": 0.043826811015605927, "global_step": 209107, "epoch": 2349, "lr": 1.2503457497955429e-05}
{"train_loss": 0.046890221536159515, "global_step": 209108, "epoch": 2349, "lr": 1.2503073995304726e-05}
{"train_loss": 0.038978494703769684, "global_step": 209109, "epoch": 2349, "lr": 1.2502690497695014e-05}
{"train_loss": 0.04918584227561951, "global_step": 209110, "epoch": 2349, "lr": 1.2502307005126357e-05}
{"train_loss": 0.05624113604426384, "global_step": 209111, "epoch": 2349, "lr": 1.2501923517598823e-05}
{"train_loss": 0.03563733398914337, "global_step": 209112, "epoch": 2349, "lr": 1.2501540035112436e-05}
{"train_loss": 0.04506959021091461, "global_step": 209113, "epoch": 2349, "lr": 1.2501156557667271e-05}
{"train_loss": 0.08654375374317169, "global_step": 209114, "epoch": 2349, "lr": 1.2500773085263351e-05}
{"train_loss": 0.0704180896282196, "global_step": 209115, "epoch": 2349, "lr": 1.2500389617900765e-05}
{"train_loss": 0.07509123533964157, "global_step": 209116, "epoch": 2349, "lr": 1.250000615557953e-05}
{"train_loss": 0.03898170590400696, "global_step": 209117, "epoch": 2349, "lr": 1.2499622698299723e-05}
{"train_loss": 0.07835877686738968, "global_step": 209118, "epoch": 2349, "lr": 1.2499239246061373e-05}
{"train_loss": 0.051200199872255325, "global_step": 209119, "epoch": 2349, "lr": 1.2498855798864555e-05}
{"train_loss": 0.06739424169063568, "global_step": 209120, "epoch": 2349, "lr": 1.24984723567093e-05}
{"train_loss": 0.15663591027259827, "global_step": 209121, "epoch": 2349, "lr": 1.249808891959568e-05}
{"train_loss": 0.07746321707963943, "global_step": 209122, "epoch": 2349, "lr": 1.2497705487523714e-05}
{"train_loss": 0.054385095834732056, "global_step": 209123, "epoch": 2349, "lr": 1.2497322060493499e-05}
{"train_loss": 0.03923947736620903, "global_step": 209124, "epoch": 2349, "lr": 1.249693863850504e-05}
{"train_loss": 0.10933798551559448, "global_step": 209125, "epoch": 2349, "lr": 1.2496555221558432e-05}
{"train_loss": 0.027851376682519913, "global_step": 209126, "epoch": 2349, "lr": 1.2496171809653685e-05}
{"train_loss": 0.04059026762843132, "global_step": 209127, "epoch": 2349, "lr": 1.2495788402790887e-05}
{"train_loss": 0.0683940201997757, "global_step": 209128, "epoch": 2349, "lr": 1.2495405000970057e-05}
{"train_loss": 0.05570051819086075, "global_step": 209129, "epoch": 2349, "lr": 1.2495021604191277e-05}
{"train_loss": 0.029573332518339157, "global_step": 209130, "epoch": 2349, "lr": 1.2494638212454584e-05}
{"train_loss": 0.05583331361413002, "global_step": 209131, "epoch": 2349, "lr": 1.2494254825760015e-05}
{"train_loss": 0.039466742426157, "global_step": 209132, "epoch": 2349, "lr": 1.2493871444107658e-05}
{"train_loss": 0.07100388407707214, "global_step": 209133, "epoch": 2349, "lr": 1.2493488067497522e-05}
{"train_loss": 0.06417983770370483, "global_step": 209134, "epoch": 2349, "lr": 1.2493104695929691e-05}
{"train_loss": 0.07317221164703369, "global_step": 209135, "epoch": 2349, "lr": 1.2492721329404195e-05}
{"train_loss": 0.1065450981259346, "global_step": 209136, "epoch": 2349, "lr": 1.2492337967921109e-05}
{"train_loss": 0.10431903600692749, "global_step": 209137, "epoch": 2349, "lr": 1.2491954611480461e-05}
{"train_loss": 0.06512459367513657, "global_step": 209138, "epoch": 2349, "lr": 1.2491571260082318e-05}
{"train_loss": 0.0918525978922844, "global_step": 209139, "epoch": 2349, "lr": 1.249118791372672e-05}
{"train_loss": 0.09981981664896011, "global_step": 209140, "epoch": 2349, "lr": 1.2490804572413734e-05}
{"train_loss": 0.051194824278354645, "global_step": 209141, "epoch": 2349, "lr": 1.2490421236143391e-05}
{"train_loss": 0.055852971971035004, "global_step": 209142, "epoch": 2349, "lr": 1.2490037904915764e-05}
{"train_loss": 0.055281106382608414, "global_step": 209143, "epoch": 2349, "lr": 1.2489654578730881e-05}
{"train_loss": 0.07353928685188293, "global_step": 209144, "epoch": 2349, "lr": 1.2489271257588825e-05}
{"train_loss": 0.07846378535032272, "global_step": 209145, "epoch": 2349, "lr": 1.2488887941489607e-05}
{"train_loss": 0.04753753915429115, "global_step": 209146, "epoch": 2349, "lr": 1.2488504630433306e-05}
{"train_loss": 0.10758182406425476, "global_step": 209147, "epoch": 2349, "lr": 1.2488121324419988e-05}
{"train_loss": 0.03875744342803955, "global_step": 209148, "epoch": 2349, "lr": 1.2487738023449663e-05}
{"train_loss": 0.06120603251155843, "global_step": 209149, "epoch": 2349, "lr": 1.2487354727522422e-05, "val_loss": 8.352021217346191}
{"train_loss": 0.08190633356571198, "global_step": 209150, "epoch": 2350, "lr": 1.2486971436638284e-05}
{"train_loss": 0.14730176329612732, "global_step": 209151, "epoch": 2350, "lr": 1.2486588150797329e-05}
{"train_loss": 0.053894881159067154, "global_step": 209152, "epoch": 2350, "lr": 1.248620486999958e-05}
{"train_loss": 0.06169448792934418, "global_step": 209153, "epoch": 2350, "lr": 1.2485821594245118e-05}
{"train_loss": 0.020338008180260658, "global_step": 209154, "epoch": 2350, "lr": 1.2485438323533966e-05}
{"train_loss": 0.06777261942625046, "global_step": 209155, "epoch": 2350, "lr": 1.2485055057866208e-05}
{"train_loss": 0.054618950933218, "global_step": 209156, "epoch": 2350, "lr": 1.2484671797241853e-05}
{"train_loss": 0.04200753942131996, "global_step": 209157, "epoch": 2350, "lr": 1.2484288541660999e-05}
{"train_loss": 0.037403225898742676, "global_step": 209158, "epoch": 2350, "lr": 1.248390529112366e-05}
{"train_loss": 0.045899342745542526, "global_step": 209159, "epoch": 2350, "lr": 1.2483522045629914e-05}
{"train_loss": 0.03583506494760513, "global_step": 209160, "epoch": 2350, "lr": 1.2483138805179789e-05}
{"train_loss": 0.06951815634965897, "global_step": 209161, "epoch": 2350, "lr": 1.2482755569773363e-05}
{"train_loss": 0.0613427460193634, "global_step": 209162, "epoch": 2350, "lr": 1.2482372339410663e-05}
{"train_loss": 0.03929446265101433, "global_step": 209163, "epoch": 2350, "lr": 1.2481989114091758e-05}
{"train_loss": 0.0451846718788147, "global_step": 209164, "epoch": 2350, "lr": 1.2481605893816684e-05}
{"train_loss": 0.03688592091202736, "global_step": 209165, "epoch": 2350, "lr": 1.2481222678585508e-05}
{"train_loss": 0.042233239859342575, "global_step": 209166, "epoch": 2350, "lr": 1.2480839468398265e-05}
{"train_loss": 0.03903747722506523, "global_step": 209167, "epoch": 2350, "lr": 1.2480456263255031e-05}
{"train_loss": 0.037616752088069916, "global_step": 209168, "epoch": 2350, "lr": 1.2480073063155829e-05}
{"train_loss": 0.06629931181669235, "global_step": 209169, "epoch": 2350, "lr": 1.247968986810073e-05}
{"train_loss": 0.0383286289870739, "global_step": 209170, "epoch": 2350, "lr": 1.2479306678089785e-05}
{"train_loss": 0.04977884516119957, "global_step": 209171, "epoch": 2350, "lr": 1.2478923493123029e-05}
{"train_loss": 0.039904896169900894, "global_step": 209172, "epoch": 2350, "lr": 1.2478540313200533e-05}
{"train_loss": 0.07205384224653244, "global_step": 209173, "epoch": 2350, "lr": 1.2478157138322328e-05}
{"train_loss": 0.059388406574726105, "global_step": 209174, "epoch": 2350, "lr": 1.2477773968488494e-05}
{"train_loss": 0.04793474078178406, "global_step": 209175, "epoch": 2350, "lr": 1.2477390803699046e-05}
{"train_loss": 0.06927972286939621, "global_step": 209176, "epoch": 2350, "lr": 1.2477007643954076e-05}
{"train_loss": 0.03163813054561615, "global_step": 209177, "epoch": 2350, "lr": 1.2476624489253597e-05}
{"train_loss": 0.01408236101269722, "global_step": 209178, "epoch": 2350, "lr": 1.2476241339597694e-05}
{"train_loss": 0.03280842304229736, "global_step": 209179, "epoch": 2350, "lr": 1.2475858194986383e-05}
{"train_loss": 0.019084498286247253, "global_step": 209180, "epoch": 2350, "lr": 1.2475475055419756e-05}
{"train_loss": 0.026661179959774017, "global_step": 209181, "epoch": 2350, "lr": 1.2475091920897831e-05}
{"train_loss": 0.024602193385362625, "global_step": 209182, "epoch": 2350, "lr": 1.247470879142067e-05}
{"train_loss": 0.06285329163074493, "global_step": 209183, "epoch": 2350, "lr": 1.247432566698834e-05}
{"train_loss": 0.0476379469037056, "global_step": 209184, "epoch": 2350, "lr": 1.247394254760087e-05}
{"train_loss": 0.06629307568073273, "global_step": 209185, "epoch": 2350, "lr": 1.247355943325833e-05}
{"train_loss": 0.04414384439587593, "global_step": 209186, "epoch": 2350, "lr": 1.247317632396075e-05}
{"train_loss": 0.06480646133422852, "global_step": 209187, "epoch": 2350, "lr": 1.2472793219708206e-05}
{"train_loss": 0.03406596556305885, "global_step": 209188, "epoch": 2350, "lr": 1.2472410120500727e-05}
{"train_loss": 0.07806983590126038, "global_step": 209189, "epoch": 2350, "lr": 1.247202702633839e-05}
{"train_loss": 0.11343804746866226, "global_step": 209190, "epoch": 2350, "lr": 1.2471643937221216e-05}
{"train_loss": 0.027774177491664886, "global_step": 209191, "epoch": 2350, "lr": 1.2471260853149285e-05}
{"train_loss": 0.05169191211462021, "global_step": 209192, "epoch": 2350, "lr": 1.2470877774122624e-05}
{"train_loss": 0.049835238605737686, "global_step": 209193, "epoch": 2350, "lr": 1.2470494700141306e-05}
{"train_loss": 0.05373986065387726, "global_step": 209194, "epoch": 2350, "lr": 1.247011163120536e-05}
{"train_loss": 0.04563932493329048, "global_step": 209195, "epoch": 2350, "lr": 1.246972856731487e-05}
{"train_loss": 0.06835059821605682, "global_step": 209196, "epoch": 2350, "lr": 1.2469345508469844e-05}
{"train_loss": 0.07139205187559128, "global_step": 209197, "epoch": 2350, "lr": 1.2468962454670379e-05}
{"train_loss": 0.11359795182943344, "global_step": 209198, "epoch": 2350, "lr": 1.2468579405916486e-05}
{"train_loss": 0.10019196569919586, "global_step": 209199, "epoch": 2350, "lr": 1.246819636220825e-05}
{"train_loss": 0.028590353205800056, "global_step": 209200, "epoch": 2350, "lr": 1.2467813323545696e-05}
{"train_loss": 0.04937681183218956, "global_step": 209201, "epoch": 2350, "lr": 1.2467430289928884e-05}
{"train_loss": 0.0488426573574543, "global_step": 209202, "epoch": 2350, "lr": 1.2467047261357884e-05}
{"train_loss": 0.03134765475988388, "global_step": 209203, "epoch": 2350, "lr": 1.2466664237832732e-05}
{"train_loss": 0.032926689833402634, "global_step": 209204, "epoch": 2350, "lr": 1.2466281219353465e-05}
{"train_loss": 0.029317090287804604, "global_step": 209205, "epoch": 2350, "lr": 1.2465898205920162e-05}
{"train_loss": 0.051355306059122086, "global_step": 209206, "epoch": 2350, "lr": 1.2465515197532846e-05}
{"train_loss": 0.059851113706827164, "global_step": 209207, "epoch": 2350, "lr": 1.2465132194191603e-05}
{"train_loss": 0.05550595372915268, "global_step": 209208, "epoch": 2350, "lr": 1.2464749195896464e-05}
{"train_loss": 0.049615275114774704, "global_step": 209209, "epoch": 2350, "lr": 1.2464366202647465e-05}
{"train_loss": 0.06226228550076485, "global_step": 209210, "epoch": 2350, "lr": 1.2463983214444691e-05}
{"train_loss": 0.052806440740823746, "global_step": 209211, "epoch": 2350, "lr": 1.2463600231288164e-05}
{"train_loss": 0.10865338891744614, "global_step": 209212, "epoch": 2350, "lr": 1.2463217253177961e-05}
{"train_loss": 0.06139802187681198, "global_step": 209213, "epoch": 2350, "lr": 1.2462834280114106e-05}
{"train_loss": 0.06275881826877594, "global_step": 209214, "epoch": 2350, "lr": 1.246245131209668e-05}
{"train_loss": 0.0617564395070076, "global_step": 209215, "epoch": 2350, "lr": 1.2462068349125705e-05}
{"train_loss": 0.049636874347925186, "global_step": 209216, "epoch": 2350, "lr": 1.246168539120126e-05}
{"train_loss": 0.056300900876522064, "global_step": 209217, "epoch": 2350, "lr": 1.2461302438323374e-05}
{"train_loss": 0.028641272336244583, "global_step": 209218, "epoch": 2350, "lr": 1.2460919490492107e-05}
{"train_loss": 0.05908143147826195, "global_step": 209219, "epoch": 2350, "lr": 1.2460536547707525e-05}
{"train_loss": 0.06360269337892532, "global_step": 209220, "epoch": 2350, "lr": 1.2460153609969655e-05}
{"train_loss": 0.09635341912508011, "global_step": 209221, "epoch": 2350, "lr": 1.2459770677278571e-05}
{"train_loss": 0.03983423486351967, "global_step": 209222, "epoch": 2350, "lr": 1.24593877496343e-05}
{"train_loss": 0.027629463002085686, "global_step": 209223, "epoch": 2350, "lr": 1.2459004827036918e-05}
{"train_loss": 0.0654732808470726, "global_step": 209224, "epoch": 2350, "lr": 1.2458621909486457e-05}
{"train_loss": 0.051332440227270126, "global_step": 209225, "epoch": 2350, "lr": 1.2458238996982984e-05}
{"train_loss": 0.061126478016376495, "global_step": 209226, "epoch": 2350, "lr": 1.2457856089526537e-05}
{"train_loss": 0.1010756567120552, "global_step": 209227, "epoch": 2350, "lr": 1.245747318711718e-05}
{"train_loss": 0.0677667111158371, "global_step": 209228, "epoch": 2350, "lr": 1.2457090289754947e-05}
{"train_loss": 0.07690498977899551, "global_step": 209229, "epoch": 2350, "lr": 1.245670739743992e-05}
{"train_loss": 0.05229734629392624, "global_step": 209230, "epoch": 2350, "lr": 1.2456324510172112e-05}
{"train_loss": 0.05252249538898468, "global_step": 209231, "epoch": 2350, "lr": 1.2455941627951607e-05}
{"train_loss": 0.07003188133239746, "global_step": 209232, "epoch": 2350, "lr": 1.2455558750778429e-05}
{"train_loss": 0.038085486739873886, "global_step": 209233, "epoch": 2350, "lr": 1.2455175878652659e-05}
{"train_loss": 0.08117196708917618, "global_step": 209234, "epoch": 2350, "lr": 1.2454793011574318e-05}
{"train_loss": 0.05801953002810478, "global_step": 209235, "epoch": 2350, "lr": 1.2454410149543489e-05}
{"train_loss": 0.054522864520549774, "global_step": 209236, "epoch": 2350, "lr": 1.2454027292560188e-05}
{"train_loss": 0.061005521565675735, "global_step": 209237, "epoch": 2350, "lr": 1.245364444062449e-05}
{"train_loss": 0.05539076256283214, "global_step": 209238, "epoch": 2350, "lr": 1.2453261593736453e-05, "train/sim_max_reward_0": 0.6847692665343511, "train/sim_max_reward_1": 0.9761223978823951, "train/sim_max_reward_2": 0.9969920175599563, "train/sim_max_reward_3": 0.6878548831075856, "train/sim_max_reward_4": 0.40781614025392965, "train/sim_max_reward_5": 1.0, "test/sim_max_reward_4500000": 0.4177081881305215, "test/sim_max_reward_4500001": 0.9942299694504437, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.5995543775964409, "test/sim_max_reward_4500004": 0.49476864833509543, "test/sim_max_reward_4500005": 0.3623724089170001, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9958920083889651, "test/sim_max_reward_4500008": 0.4756749236759819, "test/sim_max_reward_4500009": 0.04938363252894031, "test/sim_max_reward_4500010": 0.33825581985383024, "test/sim_max_reward_4500011": 0.9806722166915098, "test/sim_max_reward_4500012": 0.9745452593332402, "test/sim_max_reward_4500013": 0.6165749665575391, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9855048529450169, "test/sim_max_reward_4500016": 0.9718687004114764, "test/sim_max_reward_4500017": 0.0, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.7581367379901668, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 1.0, "test/sim_max_reward_4500023": 1.0, "test/sim_max_reward_4500024": 0.5831340210244297, "test/sim_max_reward_4500025": 0.748363473643715, "test/sim_max_reward_4500026": 0.9675338374249463, "test/sim_max_reward_4500027": 0.8975639736410502, "test/sim_max_reward_4500028": 0.6310927159197539, "test/sim_max_reward_4500029": 0.2599444762283043, "test/sim_max_reward_4500030": 0.8812564866411128, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6703657704092153, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.728507017967603, "test/sim_max_reward_4500036": 0.0803378706207045, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.8228706467880033, "test/sim_max_reward_4500039": 0.9527127691015691, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6240942237357102, "test/sim_max_reward_4500042": 0.4852325376754678, "test/sim_max_reward_4500043": 0.009539120693109769, "test/sim_max_reward_4500044": 0.19923787323852105, "test/sim_max_reward_4500045": 0.9591213580492092, "test/sim_max_reward_4500046": 0.3982817172342309, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.969990160740576, "test/sim_max_reward_4500049": 0.99789144907282, "train/mean_score": 0.7922591175563696, "test/mean_score": 0.6576442842131245, "val_loss": 8.593356132507324, "train_action_mse_error": 10.941225051879883}
{"train_loss": 0.051953140646219254, "global_step": 209239, "epoch": 2351, "lr": 1.2452878751896108e-05}
{"train_loss": 0.06448396295309067, "global_step": 209240, "epoch": 2351, "lr": 1.2452495915103523e-05}
{"train_loss": 0.06235520914196968, "global_step": 209241, "epoch": 2351, "lr": 1.2452113083358746e-05}
{"train_loss": 0.09729281067848206, "global_step": 209242, "epoch": 2351, "lr": 1.2451730256661808e-05}
{"train_loss": 0.034550417214632034, "global_step": 209243, "epoch": 2351, "lr": 1.2451347435012795e-05}
{"train_loss": 0.041803691536188126, "global_step": 209244, "epoch": 2351, "lr": 1.2450964618411726e-05}
{"train_loss": 0.06512216478586197, "global_step": 209245, "epoch": 2351, "lr": 1.245058180685868e-05}
{"train_loss": 0.05643683671951294, "global_step": 209246, "epoch": 2351, "lr": 1.2450199000353696e-05}
{"train_loss": 0.06728854030370712, "global_step": 209247, "epoch": 2351, "lr": 1.2449816198896813e-05}
{"train_loss": 0.11026795208454132, "global_step": 209248, "epoch": 2351, "lr": 1.2449433402488108e-05}
{"train_loss": 0.053505152463912964, "global_step": 209249, "epoch": 2351, "lr": 1.2449050611127604e-05}
{"train_loss": 0.023671874776482582, "global_step": 209250, "epoch": 2351, "lr": 1.244866782481538e-05}
{"train_loss": 0.035366594791412354, "global_step": 209251, "epoch": 2351, "lr": 1.244828504355146e-05}
{"train_loss": 0.06906035542488098, "global_step": 209252, "epoch": 2351, "lr": 1.2447902267335931e-05}
{"train_loss": 0.03067632205784321, "global_step": 209253, "epoch": 2351, "lr": 1.2447519496168803e-05}
{"train_loss": 0.029182469472289085, "global_step": 209254, "epoch": 2351, "lr": 1.2447136730050164e-05}
{"train_loss": 0.06284149736166, "global_step": 209255, "epoch": 2351, "lr": 1.2446753968980036e-05}
{"train_loss": 0.06800578534603119, "global_step": 209256, "epoch": 2351, "lr": 1.2446371212958485e-05}
{"train_loss": 0.049826037138700485, "global_step": 209257, "epoch": 2351, "lr": 1.2445988461985576e-05}
{"train_loss": 0.05737660825252533, "global_step": 209258, "epoch": 2351, "lr": 1.2445605716061332e-05}
{"train_loss": 0.03258384019136429, "global_step": 209259, "epoch": 2351, "lr": 1.2445222975185833e-05}
{"train_loss": 0.0198531374335289, "global_step": 209260, "epoch": 2351, "lr": 1.2444840239359101e-05}
{"train_loss": 0.0579240508377552, "global_step": 209261, "epoch": 2351, "lr": 1.2444457508581214e-05}
{"train_loss": 0.08507853746414185, "global_step": 209262, "epoch": 2351, "lr": 1.24440747828522e-05}
{"train_loss": 0.051617756485939026, "global_step": 209263, "epoch": 2351, "lr": 1.2443692062172141e-05}
{"train_loss": 0.04504542052745819, "global_step": 209264, "epoch": 2351, "lr": 1.244330934654105e-05}
{"train_loss": 0.08603496104478836, "global_step": 209265, "epoch": 2351, "lr": 1.2442926635959012e-05}
{"train_loss": 0.030153734609484673, "global_step": 209266, "epoch": 2351, "lr": 1.2442543930426054e-05}
{"train_loss": 0.029966970905661583, "global_step": 209267, "epoch": 2351, "lr": 1.2442161229942251e-05}
{"train_loss": 0.02777845598757267, "global_step": 209268, "epoch": 2351, "lr": 1.2441778534507625e-05}
{"train_loss": 0.11316494643688202, "global_step": 209269, "epoch": 2351, "lr": 1.2441395844122266e-05}
{"train_loss": 0.08221685886383057, "global_step": 209270, "epoch": 2351, "lr": 1.244101315878618e-05}
{"train_loss": 0.05500080808997154, "global_step": 209271, "epoch": 2351, "lr": 1.2440630478499465e-05}
{"train_loss": 0.06166287511587143, "global_step": 209272, "epoch": 2351, "lr": 1.2440247803262129e-05}
{"train_loss": 0.07290282845497131, "global_step": 209273, "epoch": 2351, "lr": 1.2439865133074246e-05}
{"train_loss": 0.07342192530632019, "global_step": 209274, "epoch": 2351, "lr": 1.2439482467935875e-05}
{"train_loss": 0.06222471594810486, "global_step": 209275, "epoch": 2351, "lr": 1.2439099807847048e-05}
{"train_loss": 0.041176170110702515, "global_step": 209276, "epoch": 2351, "lr": 1.2438717152807844e-05}
{"train_loss": 0.07365680485963821, "global_step": 209277, "epoch": 2351, "lr": 1.2438334502818277e-05}
{"train_loss": 0.05089767649769783, "global_step": 209278, "epoch": 2351, "lr": 1.2437951857878433e-05}
{"train_loss": 0.0549095943570137, "global_step": 209279, "epoch": 2351, "lr": 1.2437569217988353e-05}
{"train_loss": 0.11127831786870956, "global_step": 209280, "epoch": 2351, "lr": 1.2437186583148064e-05}
{"train_loss": 0.051706135272979736, "global_step": 209281, "epoch": 2351, "lr": 1.2436803953357657e-05}
{"train_loss": 0.044613808393478394, "global_step": 209282, "epoch": 2351, "lr": 1.2436421328617143e-05}
{"train_loss": 0.0663713663816452, "global_step": 209283, "epoch": 2351, "lr": 1.2436038708926612e-05}
{"train_loss": 0.040861353278160095, "global_step": 209284, "epoch": 2351, "lr": 1.2435656094286085e-05}
{"train_loss": 0.05015424266457558, "global_step": 209285, "epoch": 2351, "lr": 1.2435273484695636e-05}
{"train_loss": 0.0746426060795784, "global_step": 209286, "epoch": 2351, "lr": 1.2434890880155315e-05}
{"train_loss": 0.0480957068502903, "global_step": 209287, "epoch": 2351, "lr": 1.243450828066514e-05}
{"train_loss": 0.05315563082695007, "global_step": 209288, "epoch": 2351, "lr": 1.2434125686225213e-05}
{"train_loss": 0.04444314166903496, "global_step": 209289, "epoch": 2351, "lr": 1.2433743096835538e-05}
{"train_loss": 0.061843741685152054, "global_step": 209290, "epoch": 2351, "lr": 1.2433360512496206e-05}
{"train_loss": 0.04290531203150749, "global_step": 209291, "epoch": 2351, "lr": 1.2432977933207235e-05}
{"train_loss": 0.02917989157140255, "global_step": 209292, "epoch": 2351, "lr": 1.243259535896869e-05}
{"train_loss": 0.03697764873504639, "global_step": 209293, "epoch": 2351, "lr": 1.2432212789780639e-05}
{"train_loss": 0.05151378735899925, "global_step": 209294, "epoch": 2351, "lr": 1.243183022564311e-05}
{"train_loss": 0.06875760108232498, "global_step": 209295, "epoch": 2351, "lr": 1.2431447666556174e-05}
{"train_loss": 0.05614427849650383, "global_step": 209296, "epoch": 2351, "lr": 1.2431065112519858e-05}
{"train_loss": 0.04430075362324715, "global_step": 209297, "epoch": 2351, "lr": 1.2430682563534245e-05}
{"train_loss": 0.044695895165205, "global_step": 209298, "epoch": 2351, "lr": 1.243030001959935e-05}
{"train_loss": 0.08599371463060379, "global_step": 209299, "epoch": 2351, "lr": 1.242991748071526e-05}
{"train_loss": 0.03434969112277031, "global_step": 209300, "epoch": 2351, "lr": 1.2429534946881994e-05}
{"train_loss": 0.0743471011519432, "global_step": 209301, "epoch": 2351, "lr": 1.2429152418099639e-05}
{"train_loss": 0.051918644458055496, "global_step": 209302, "epoch": 2351, "lr": 1.2428769894368203e-05}
{"train_loss": 0.04871212691068649, "global_step": 209303, "epoch": 2351, "lr": 1.2428387375687783e-05}
{"train_loss": 0.03250093758106232, "global_step": 209304, "epoch": 2351, "lr": 1.242800486205839e-05}
{"train_loss": 0.09688500314950943, "global_step": 209305, "epoch": 2351, "lr": 1.2427622353480106e-05}
{"train_loss": 0.0565512552857399, "global_step": 209306, "epoch": 2351, "lr": 1.242723984995296e-05}
{"train_loss": 0.08252495527267456, "global_step": 209307, "epoch": 2351, "lr": 1.2426857351477028e-05}
{"train_loss": 0.030732031911611557, "global_step": 209308, "epoch": 2351, "lr": 1.2426474858052328e-05}
{"train_loss": 0.042745985090732574, "global_step": 209309, "epoch": 2351, "lr": 1.2426092369678938e-05}
{"train_loss": 0.08438104391098022, "global_step": 209310, "epoch": 2351, "lr": 1.2425709886356913e-05}
{"train_loss": 0.11484314501285553, "global_step": 209311, "epoch": 2351, "lr": 1.2425327408086274e-05}
{"train_loss": 0.029130499809980392, "global_step": 209312, "epoch": 2351, "lr": 1.242494493486711e-05}
{"train_loss": 0.11146113276481628, "global_step": 209313, "epoch": 2351, "lr": 1.242456246669944e-05}
{"train_loss": 0.10350973159074783, "global_step": 209314, "epoch": 2351, "lr": 1.2424180003583346e-05}
{"train_loss": 0.04825223237276077, "global_step": 209315, "epoch": 2351, "lr": 1.2423797545518845e-05}
{"train_loss": 0.1003979966044426, "global_step": 209316, "epoch": 2351, "lr": 1.2423415092506025e-05}
{"train_loss": 0.052822258323431015, "global_step": 209317, "epoch": 2351, "lr": 1.2423032644544902e-05}
{"train_loss": 0.056553371250629425, "global_step": 209318, "epoch": 2351, "lr": 1.242265020163556e-05}
{"train_loss": 0.07565058767795563, "global_step": 209319, "epoch": 2351, "lr": 1.2422267763778034e-05}
{"train_loss": 0.027046440169215202, "global_step": 209320, "epoch": 2351, "lr": 1.242188533097236e-05}
{"train_loss": 0.02977358177304268, "global_step": 209321, "epoch": 2351, "lr": 1.2421502903218623e-05}
{"train_loss": 0.04477287083864212, "global_step": 209322, "epoch": 2351, "lr": 1.2421120480516841e-05}
{"train_loss": 0.04705355688929558, "global_step": 209323, "epoch": 2351, "lr": 1.2420738062867094e-05}
{"train_loss": 0.055155616253614426, "global_step": 209324, "epoch": 2351, "lr": 1.2420355650269422e-05}
{"train_loss": 0.050351083278656006, "global_step": 209325, "epoch": 2351, "lr": 1.241997324272386e-05}
{"train_loss": 0.03365851193666458, "global_step": 209326, "epoch": 2351, "lr": 1.2419590840230488e-05}
{"train_loss": 0.0574456626193577, "global_step": 209327, "epoch": 2351, "lr": 1.2419208442789333e-05, "val_loss": 8.49958324432373}
{"train_loss": 0.032677873969078064, "global_step": 209328, "epoch": 2352, "lr": 1.2418826050400456e-05}
{"train_loss": 0.04108894243836403, "global_step": 209329, "epoch": 2352, "lr": 1.2418443663063922e-05}
{"train_loss": 0.07984785735607147, "global_step": 209330, "epoch": 2352, "lr": 1.2418061280779757e-05}
{"train_loss": 0.055121082812547684, "global_step": 209331, "epoch": 2352, "lr": 1.241767890354804e-05}
{"train_loss": 0.07130096852779388, "global_step": 209332, "epoch": 2352, "lr": 1.2417296531368794e-05}
{"train_loss": 0.10048729181289673, "global_step": 209333, "epoch": 2352, "lr": 1.2416914164242099e-05}
{"train_loss": 0.11300475150346756, "global_step": 209334, "epoch": 2352, "lr": 1.241653180216798e-05}
{"train_loss": 0.07374658435583115, "global_step": 209335, "epoch": 2352, "lr": 1.2416149445146514e-05}
{"train_loss": 0.0535537526011467, "global_step": 209336, "epoch": 2352, "lr": 1.241576709317772e-05}
{"train_loss": 0.09787869453430176, "global_step": 209337, "epoch": 2352, "lr": 1.2415384746261688e-05}
{"train_loss": 0.04089319705963135, "global_step": 209338, "epoch": 2352, "lr": 1.241500240439843e-05}
{"train_loss": 0.047315314412117004, "global_step": 209339, "epoch": 2352, "lr": 1.2414620067588035e-05}
{"train_loss": 0.051104918122291565, "global_step": 209340, "epoch": 2352, "lr": 1.241423773583052e-05}
{"train_loss": 0.024898264557123184, "global_step": 209341, "epoch": 2352, "lr": 1.2413855409125963e-05}
{"train_loss": 0.024267887696623802, "global_step": 209342, "epoch": 2352, "lr": 1.2413473087474397e-05}
{"train_loss": 0.05907672643661499, "global_step": 209343, "epoch": 2352, "lr": 1.241309077087589e-05}
{"train_loss": 0.0528612844645977, "global_step": 209344, "epoch": 2352, "lr": 1.2412708459330474e-05}
{"train_loss": 0.10188409686088562, "global_step": 209345, "epoch": 2352, "lr": 1.241232615283821e-05}
{"train_loss": 0.03515281528234482, "global_step": 209346, "epoch": 2352, "lr": 1.2411943851399166e-05}
{"train_loss": 0.07561888545751572, "global_step": 209347, "epoch": 2352, "lr": 1.2411561555013368e-05}
{"train_loss": 0.053907498717308044, "global_step": 209348, "epoch": 2352, "lr": 1.2411179263680884e-05}
{"train_loss": 0.025020316243171692, "global_step": 209349, "epoch": 2352, "lr": 1.241079697740175e-05}
{"train_loss": 0.07555820047855377, "global_step": 209350, "epoch": 2352, "lr": 1.2410414696176038e-05}
{"train_loss": 0.012327387928962708, "global_step": 209351, "epoch": 2352, "lr": 1.2410032420003776e-05}
{"train_loss": 0.0703568160533905, "global_step": 209352, "epoch": 2352, "lr": 1.2409650148885038e-05}
{"train_loss": 0.13086667656898499, "global_step": 209353, "epoch": 2352, "lr": 1.2409267882819847e-05}
{"train_loss": 0.06495147198438644, "global_step": 209354, "epoch": 2352, "lr": 1.2408885621808292e-05}
{"train_loss": 0.0787380188703537, "global_step": 209355, "epoch": 2352, "lr": 1.2408503365850387e-05}
{"train_loss": 0.04755788668990135, "global_step": 209356, "epoch": 2352, "lr": 1.240812111494622e-05}
{"train_loss": 0.04159333556890488, "global_step": 209357, "epoch": 2352, "lr": 1.2407738869095815e-05}
{"train_loss": 0.06220143660902977, "global_step": 209358, "epoch": 2352, "lr": 1.2407356628299222e-05}
{"train_loss": 0.1387869119644165, "global_step": 209359, "epoch": 2352, "lr": 1.2406974392556514e-05}
{"train_loss": 0.05298894643783569, "global_step": 209360, "epoch": 2352, "lr": 1.2406592161867719e-05}
{"train_loss": 0.048767730593681335, "global_step": 209361, "epoch": 2352, "lr": 1.2406209936232909e-05}
{"train_loss": 0.0945425033569336, "global_step": 209362, "epoch": 2352, "lr": 1.240582771565213e-05}
{"train_loss": 0.08135250955820084, "global_step": 209363, "epoch": 2352, "lr": 1.2405445500125413e-05}
{"train_loss": 0.07328404486179352, "global_step": 209364, "epoch": 2352, "lr": 1.2405063289652825e-05}
{"train_loss": 0.02235816791653633, "global_step": 209365, "epoch": 2352, "lr": 1.2404681084234431e-05}
{"train_loss": 0.03472013771533966, "global_step": 209366, "epoch": 2352, "lr": 1.2404298883870258e-05}
{"train_loss": 0.04488560929894447, "global_step": 209367, "epoch": 2352, "lr": 1.2403916688560385e-05}
{"train_loss": 0.04317085072398186, "global_step": 209368, "epoch": 2352, "lr": 1.2403534498304831e-05}
{"train_loss": 0.02193155325949192, "global_step": 209369, "epoch": 2352, "lr": 1.2403152313103678e-05}
{"train_loss": 0.03923763707280159, "global_step": 209370, "epoch": 2352, "lr": 1.2402770132956943e-05}
{"train_loss": 0.1043902039527893, "global_step": 209371, "epoch": 2352, "lr": 1.240238795786472e-05}
{"train_loss": 0.04870950058102608, "global_step": 209372, "epoch": 2352, "lr": 1.2402005787827014e-05}
{"train_loss": 0.03421658277511597, "global_step": 209373, "epoch": 2352, "lr": 1.2401623622843922e-05}
{"train_loss": 0.07425317913293839, "global_step": 209374, "epoch": 2352, "lr": 1.2401241462915459e-05}
{"train_loss": 0.06729113310575485, "global_step": 209375, "epoch": 2352, "lr": 1.2400859308041701e-05}
{"train_loss": 0.05453915894031525, "global_step": 209376, "epoch": 2352, "lr": 1.2400477158222678e-05}
{"train_loss": 0.04419894888997078, "global_step": 209377, "epoch": 2352, "lr": 1.2400095013458469e-05}
{"train_loss": 0.0388169139623642, "global_step": 209378, "epoch": 2352, "lr": 1.2399712873749092e-05}
{"train_loss": 0.09274467825889587, "global_step": 209379, "epoch": 2352, "lr": 1.2399330739094628e-05}
{"train_loss": 0.09857559949159622, "global_step": 209380, "epoch": 2352, "lr": 1.2398948609495104e-05}
{"train_loss": 0.05685582384467125, "global_step": 209381, "epoch": 2352, "lr": 1.2398566484950596e-05}
{"train_loss": 0.04423343017697334, "global_step": 209382, "epoch": 2352, "lr": 1.2398184365461124e-05}
{"train_loss": 0.07448415458202362, "global_step": 209383, "epoch": 2352, "lr": 1.2397802251026763e-05}
{"train_loss": 0.053282491862773895, "global_step": 209384, "epoch": 2352, "lr": 1.2397420141647576e-05}
{"train_loss": 0.03116190806031227, "global_step": 209385, "epoch": 2352, "lr": 1.2397038037323577e-05}
{"train_loss": 0.054908230900764465, "global_step": 209386, "epoch": 2352, "lr": 1.2396655938054857e-05}
{"train_loss": 0.06531928479671478, "global_step": 209387, "epoch": 2352, "lr": 1.2396273843841434e-05}
{"train_loss": 0.03481669723987579, "global_step": 209388, "epoch": 2352, "lr": 1.239589175468338e-05}
{"train_loss": 0.059143904596567154, "global_step": 209389, "epoch": 2352, "lr": 1.2395509670580734e-05}
{"train_loss": 0.03185277059674263, "global_step": 209390, "epoch": 2352, "lr": 1.2395127591533568e-05}
{"train_loss": 0.03605860099196434, "global_step": 209391, "epoch": 2352, "lr": 1.2394745517541901e-05}
{"train_loss": 0.06535699218511581, "global_step": 209392, "epoch": 2352, "lr": 1.2394363448605816e-05}
{"train_loss": 0.028318103402853012, "global_step": 209393, "epoch": 2352, "lr": 1.239398138472534e-05}
{"train_loss": 0.059830356389284134, "global_step": 209394, "epoch": 2352, "lr": 1.2393599325900546e-05}
{"train_loss": 0.08899841457605362, "global_step": 209395, "epoch": 2352, "lr": 1.2393217272131457e-05}
{"train_loss": 0.04114082455635071, "global_step": 209396, "epoch": 2352, "lr": 1.2392835223418159e-05}
{"train_loss": 0.05063755065202713, "global_step": 209397, "epoch": 2352, "lr": 1.2392453179760688e-05}
{"train_loss": 0.07508960366249084, "global_step": 209398, "epoch": 2352, "lr": 1.2392071141159078e-05}
{"train_loss": 0.016817405819892883, "global_step": 209399, "epoch": 2352, "lr": 1.2391689107613408e-05}
{"train_loss": 0.1065274327993393, "global_step": 209400, "epoch": 2352, "lr": 1.2391307079123704e-05}
{"train_loss": 0.10785973072052002, "global_step": 209401, "epoch": 2352, "lr": 1.2390925055690045e-05}
{"train_loss": 0.047582369297742844, "global_step": 209402, "epoch": 2352, "lr": 1.2390543037312452e-05}
{"train_loss": 0.07751397043466568, "global_step": 209403, "epoch": 2352, "lr": 1.239016102399101e-05}
{"train_loss": 0.032160770148038864, "global_step": 209404, "epoch": 2352, "lr": 1.2389779015725733e-05}
{"train_loss": 0.06335043162107468, "global_step": 209405, "epoch": 2352, "lr": 1.2389397012516713e-05}
{"train_loss": 0.04967507719993591, "global_step": 209406, "epoch": 2352, "lr": 1.2389015014363958e-05}
{"train_loss": 0.0509553924202919, "global_step": 209407, "epoch": 2352, "lr": 1.238863302126756e-05}
{"train_loss": 0.045938339084386826, "global_step": 209408, "epoch": 2352, "lr": 1.2388251033227538e-05}
{"train_loss": 0.08521407097578049, "global_step": 209409, "epoch": 2352, "lr": 1.2387869050243972e-05}
{"train_loss": 0.07008016854524612, "global_step": 209410, "epoch": 2352, "lr": 1.2387487072316879e-05}
{"train_loss": 0.06403730809688568, "global_step": 209411, "epoch": 2352, "lr": 1.238710509944635e-05}
{"train_loss": 0.026677433401346207, "global_step": 209412, "epoch": 2352, "lr": 1.2386723131632399e-05}
{"train_loss": 0.05442480742931366, "global_step": 209413, "epoch": 2352, "lr": 1.238634116887511e-05}
{"train_loss": 0.037177156656980515, "global_step": 209414, "epoch": 2352, "lr": 1.2385959211174503e-05}
{"train_loss": 0.06954579055309296, "global_step": 209415, "epoch": 2352, "lr": 1.2385577258530662e-05}
{"train_loss": 0.05932501130057185, "global_step": 209416, "epoch": 2352, "lr": 1.2385195310943609e-05, "val_loss": 8.545516014099121}
{"train_loss": 0.055776748806238174, "global_step": 209417, "epoch": 2353, "lr": 1.2384813368413417e-05}
{"train_loss": 0.0443306490778923, "global_step": 209418, "epoch": 2353, "lr": 1.2384431430940119e-05}
{"train_loss": 0.061461709439754486, "global_step": 209419, "epoch": 2353, "lr": 1.2384049498523775e-05}
{"train_loss": 0.03595544770359993, "global_step": 209420, "epoch": 2353, "lr": 1.2383667571164448e-05}
{"train_loss": 0.05288968235254288, "global_step": 209421, "epoch": 2353, "lr": 1.238328564886217e-05}
{"train_loss": 0.07537926733493805, "global_step": 209422, "epoch": 2353, "lr": 1.2382903731617006e-05}
{"train_loss": 0.08001535385847092, "global_step": 209423, "epoch": 2353, "lr": 1.2382521819428993e-05}
{"train_loss": 0.10727989673614502, "global_step": 209424, "epoch": 2353, "lr": 1.2382139912298208e-05}
{"train_loss": 0.010120281018316746, "global_step": 209425, "epoch": 2353, "lr": 1.238175801022467e-05}
{"train_loss": 0.036126963794231415, "global_step": 209426, "epoch": 2353, "lr": 1.238137611320846e-05}
{"train_loss": 0.05333101376891136, "global_step": 209427, "epoch": 2353, "lr": 1.2380994221249603e-05}
{"train_loss": 0.06784109026193619, "global_step": 209428, "epoch": 2353, "lr": 1.2380612334348174e-05}
{"train_loss": 0.06233619526028633, "global_step": 209429, "epoch": 2353, "lr": 1.2380230452504198e-05}
{"train_loss": 0.08221735805273056, "global_step": 209430, "epoch": 2353, "lr": 1.2379848575717756e-05}
{"train_loss": 0.06055852025747299, "global_step": 209431, "epoch": 2353, "lr": 1.2379466703988874e-05}
{"train_loss": 0.05645519867539406, "global_step": 209432, "epoch": 2353, "lr": 1.2379084837317628e-05}
{"train_loss": 0.019117502495646477, "global_step": 209433, "epoch": 2353, "lr": 1.2378702975704038e-05}
{"train_loss": 0.09429854154586792, "global_step": 209434, "epoch": 2353, "lr": 1.237832111914819e-05}
{"train_loss": 0.04971372336149216, "global_step": 209435, "epoch": 2353, "lr": 1.2377939267650118e-05}
{"train_loss": 0.03998280689120293, "global_step": 209436, "epoch": 2353, "lr": 1.2377557421209856e-05}
{"train_loss": 0.052303023636341095, "global_step": 209437, "epoch": 2353, "lr": 1.2377175579827494e-05}
{"train_loss": 0.027279436588287354, "global_step": 209438, "epoch": 2353, "lr": 1.237679374350304e-05}
{"train_loss": 0.04439888149499893, "global_step": 209439, "epoch": 2353, "lr": 1.2376411912236585e-05}
{"train_loss": 0.0698738619685173, "global_step": 209440, "epoch": 2353, "lr": 1.2376030086028151e-05}
{"train_loss": 0.031362131237983704, "global_step": 209441, "epoch": 2353, "lr": 1.2375648264877814e-05}
{"train_loss": 0.07502162456512451, "global_step": 209442, "epoch": 2353, "lr": 1.23752664487856e-05}
{"train_loss": 0.03806813433766365, "global_step": 209443, "epoch": 2353, "lr": 1.2374884637751582e-05}
{"train_loss": 0.05443766340613365, "global_step": 209444, "epoch": 2353, "lr": 1.2374502831775792e-05}
{"train_loss": 0.07373115420341492, "global_step": 209445, "epoch": 2353, "lr": 1.2374121030858305e-05}
{"train_loss": 0.0322621613740921, "global_step": 209446, "epoch": 2353, "lr": 1.2373739234999144e-05}
{"train_loss": 0.07189091295003891, "global_step": 209447, "epoch": 2353, "lr": 1.2373357444198385e-05}
{"train_loss": 0.03834085911512375, "global_step": 209448, "epoch": 2353, "lr": 1.2372975658456059e-05}
{"train_loss": 0.04945700243115425, "global_step": 209449, "epoch": 2353, "lr": 1.2372593877772249e-05}
{"train_loss": 0.08200273662805557, "global_step": 209450, "epoch": 2353, "lr": 1.2372212102146964e-05}
{"train_loss": 0.09075203537940979, "global_step": 209451, "epoch": 2353, "lr": 1.2371830331580286e-05}
{"train_loss": 0.0326196625828743, "global_step": 209452, "epoch": 2353, "lr": 1.2371448566072252e-05}
{"train_loss": 0.08531732857227325, "global_step": 209453, "epoch": 2353, "lr": 1.2371066805622927e-05}
{"train_loss": 0.1138697937130928, "global_step": 209454, "epoch": 2353, "lr": 1.237068505023234e-05}
{"train_loss": 0.05252944678068161, "global_step": 209455, "epoch": 2353, "lr": 1.2370303299900554e-05}
{"train_loss": 0.0456557534635067, "global_step": 209456, "epoch": 2353, "lr": 1.2369921554627639e-05}
{"train_loss": 0.07087065279483795, "global_step": 209457, "epoch": 2353, "lr": 1.2369539814413616e-05}
{"train_loss": 0.04849379137158394, "global_step": 209458, "epoch": 2353, "lr": 1.2369158079258564e-05}
{"train_loss": 0.11480630189180374, "global_step": 209459, "epoch": 2353, "lr": 1.2368776349162504e-05}
{"train_loss": 0.02405332773923874, "global_step": 209460, "epoch": 2353, "lr": 1.2368394624125518e-05}
{"train_loss": 0.041352491825819016, "global_step": 209461, "epoch": 2353, "lr": 1.2368012904147624e-05}
{"train_loss": 0.07018724083900452, "global_step": 209462, "epoch": 2353, "lr": 1.2367631189228912e-05}
{"train_loss": 0.06930080056190491, "global_step": 209463, "epoch": 2353, "lr": 1.2367249479369397e-05}
{"train_loss": 0.08911820501089096, "global_step": 209464, "epoch": 2353, "lr": 1.2366867774569163e-05}
{"train_loss": 0.0892823114991188, "global_step": 209465, "epoch": 2353, "lr": 1.2366486074828232e-05}
{"train_loss": 0.03344236686825752, "global_step": 209466, "epoch": 2353, "lr": 1.2366104380146675e-05}
{"train_loss": 0.056366268545389175, "global_step": 209467, "epoch": 2353, "lr": 1.2365722690524528e-05}
{"train_loss": 0.04017135128378868, "global_step": 209468, "epoch": 2353, "lr": 1.2365341005961867e-05}
{"train_loss": 0.09138068556785583, "global_step": 209469, "epoch": 2353, "lr": 1.2364959326458708e-05}
{"train_loss": 0.04288741201162338, "global_step": 209470, "epoch": 2353, "lr": 1.2364577652015142e-05}
{"train_loss": 0.09009582549333572, "global_step": 209471, "epoch": 2353, "lr": 1.2364195982631178e-05}
{"train_loss": 0.09343676269054413, "global_step": 209472, "epoch": 2353, "lr": 1.236381431830691e-05}
{"train_loss": 0.04247526824474335, "global_step": 209473, "epoch": 2353, "lr": 1.2363432659042362e-05}
{"train_loss": 0.06943830102682114, "global_step": 209474, "epoch": 2353, "lr": 1.2363051004837584e-05}
{"train_loss": 0.08498868346214294, "global_step": 209475, "epoch": 2353, "lr": 1.2362669355692646e-05}
{"train_loss": 0.06190349534153938, "global_step": 209476, "epoch": 2353, "lr": 1.2362287711607573e-05}
{"train_loss": 0.008831489831209183, "global_step": 209477, "epoch": 2353, "lr": 1.2361906072582447e-05}
{"train_loss": 0.05181538313627243, "global_step": 209478, "epoch": 2353, "lr": 1.236152443861729e-05}
{"train_loss": 0.05394808202981949, "global_step": 209479, "epoch": 2353, "lr": 1.2361142809712183e-05}
{"train_loss": 0.027024034410715103, "global_step": 209480, "epoch": 2353, "lr": 1.2360761185867142e-05}
{"train_loss": 0.07848013937473297, "global_step": 209481, "epoch": 2353, "lr": 1.2360379567082258e-05}
{"train_loss": 0.06789293885231018, "global_step": 209482, "epoch": 2353, "lr": 1.235999795335755e-05}
{"train_loss": 0.08006452769041061, "global_step": 209483, "epoch": 2353, "lr": 1.2359616344693087e-05}
{"train_loss": 0.026101339608430862, "global_step": 209484, "epoch": 2353, "lr": 1.235923474108891e-05}
{"train_loss": 0.04288167878985405, "global_step": 209485, "epoch": 2353, "lr": 1.235885314254508e-05}
{"train_loss": 0.038859277963638306, "global_step": 209486, "epoch": 2353, "lr": 1.2358471549061635e-05}
{"train_loss": 0.0715114027261734, "global_step": 209487, "epoch": 2353, "lr": 1.2358089960638647e-05}
{"train_loss": 0.08226335793733597, "global_step": 209488, "epoch": 2353, "lr": 1.2357708377276144e-05}
{"train_loss": 0.048668865114450455, "global_step": 209489, "epoch": 2353, "lr": 1.23573267989742e-05}
{"train_loss": 0.05082640051841736, "global_step": 209490, "epoch": 2353, "lr": 1.235694522573284e-05}
{"train_loss": 0.041040707379579544, "global_step": 209491, "epoch": 2353, "lr": 1.2356563657552128e-05}
{"train_loss": 0.08228478580713272, "global_step": 209492, "epoch": 2353, "lr": 1.2356182094432133e-05}
{"train_loss": 0.09334965795278549, "global_step": 209493, "epoch": 2353, "lr": 1.235580053637288e-05}
{"train_loss": 0.048310596495866776, "global_step": 209494, "epoch": 2353, "lr": 1.235541898337444e-05}
{"train_loss": 0.035272639244794846, "global_step": 209495, "epoch": 2353, "lr": 1.2355037435436845e-05}
{"train_loss": 0.07277219742536545, "global_step": 209496, "epoch": 2353, "lr": 1.235465589256017e-05}
{"train_loss": 0.10064440965652466, "global_step": 209497, "epoch": 2353, "lr": 1.2354274354744432e-05}
{"train_loss": 0.10782912373542786, "global_step": 209498, "epoch": 2353, "lr": 1.2353892821989727e-05}
{"train_loss": 0.052920859307050705, "global_step": 209499, "epoch": 2353, "lr": 1.2353511294296061e-05}
{"train_loss": 0.051426537334918976, "global_step": 209500, "epoch": 2353, "lr": 1.2353129771663523e-05}
{"train_loss": 0.028609367087483406, "global_step": 209501, "epoch": 2353, "lr": 1.2352748254092133e-05}
{"train_loss": 0.03716234862804413, "global_step": 209502, "epoch": 2353, "lr": 1.2352366741581973e-05}
{"train_loss": 0.02020839974284172, "global_step": 209503, "epoch": 2353, "lr": 1.2351985234133063e-05}
{"train_loss": 0.04763779416680336, "global_step": 209504, "epoch": 2353, "lr": 1.2351603731745487e-05}
{"train_loss": 0.058799569076450356, "global_step": 209505, "epoch": 2353, "lr": 1.2351222234419263e-05, "val_loss": 8.588297843933105}
{"train_loss": 0.03838123753666878, "global_step": 209506, "epoch": 2354, "lr": 1.2350840742154468e-05}
{"train_loss": 0.09849947690963745, "global_step": 209507, "epoch": 2354, "lr": 1.2350459254951136e-05}
{"train_loss": 0.049602679908275604, "global_step": 209508, "epoch": 2354, "lr": 1.2350077772809338e-05}
{"train_loss": 0.07386508584022522, "global_step": 209509, "epoch": 2354, "lr": 1.2349696295729096e-05}
{"train_loss": 0.07794513553380966, "global_step": 209510, "epoch": 2354, "lr": 1.2349314823710495e-05}
{"train_loss": 0.05287563428282738, "global_step": 209511, "epoch": 2354, "lr": 1.2348933356753555e-05}
{"train_loss": 0.044993069022893906, "global_step": 209512, "epoch": 2354, "lr": 1.2348551894858357e-05}
{"train_loss": 0.058045368641614914, "global_step": 209513, "epoch": 2354, "lr": 1.2348170438024936e-05}
{"train_loss": 0.02363753877580166, "global_step": 209514, "epoch": 2354, "lr": 1.2347788986253328e-05}
{"train_loss": 0.10440172255039215, "global_step": 209515, "epoch": 2354, "lr": 1.234740753954362e-05}
{"train_loss": 0.08863963186740875, "global_step": 209516, "epoch": 2354, "lr": 1.2347026097895824e-05}
{"train_loss": 0.04637463018298149, "global_step": 209517, "epoch": 2354, "lr": 1.2346644661310037e-05}
{"train_loss": 0.03719286993145943, "global_step": 209518, "epoch": 2354, "lr": 1.2346263229786259e-05}
{"train_loss": 0.07642035186290741, "global_step": 209519, "epoch": 2354, "lr": 1.2345881803324588e-05}
{"train_loss": 0.09998949617147446, "global_step": 209520, "epoch": 2354, "lr": 1.2345500381925036e-05}
{"train_loss": 0.06829561293125153, "global_step": 209521, "epoch": 2354, "lr": 1.2345118965587687e-05}
{"train_loss": 0.03708291053771973, "global_step": 209522, "epoch": 2354, "lr": 1.2344737554312564e-05}
{"train_loss": 0.042494434863328934, "global_step": 209523, "epoch": 2354, "lr": 1.234435614809975e-05}
{"train_loss": 0.05178697407245636, "global_step": 209524, "epoch": 2354, "lr": 1.234397474694926e-05}
{"train_loss": 0.06638830155134201, "global_step": 209525, "epoch": 2354, "lr": 1.2343593350861176e-05}
{"train_loss": 0.08631841093301773, "global_step": 209526, "epoch": 2354, "lr": 1.2343211959835527e-05}
{"train_loss": 0.07822203636169434, "global_step": 209527, "epoch": 2354, "lr": 1.234283057387237e-05}
{"train_loss": 0.026484141126275063, "global_step": 209528, "epoch": 2354, "lr": 1.2342449192971778e-05}
{"train_loss": 0.061224792152643204, "global_step": 209529, "epoch": 2354, "lr": 1.2342067817133773e-05}
{"train_loss": 0.02054603397846222, "global_step": 209530, "epoch": 2354, "lr": 1.234168644635843e-05}
{"train_loss": 0.10368435829877853, "global_step": 209531, "epoch": 2354, "lr": 1.2341305080645772e-05}
{"train_loss": 0.0557909831404686, "global_step": 209532, "epoch": 2354, "lr": 1.2340923719995885e-05}
{"train_loss": 0.08521237969398499, "global_step": 209533, "epoch": 2354, "lr": 1.234054236440878e-05}
{"train_loss": 0.03795212507247925, "global_step": 209534, "epoch": 2354, "lr": 1.234016101388455e-05}
{"train_loss": 0.05598379671573639, "global_step": 209535, "epoch": 2354, "lr": 1.2339779668423212e-05}
{"train_loss": 0.04818182811141014, "global_step": 209536, "epoch": 2354, "lr": 1.2339398328024842e-05}
{"train_loss": 0.06773965060710907, "global_step": 209537, "epoch": 2354, "lr": 1.2339016992689474e-05}
{"train_loss": 0.0893704816699028, "global_step": 209538, "epoch": 2354, "lr": 1.2338635662417175e-05}
{"train_loss": 0.09003859758377075, "global_step": 209539, "epoch": 2354, "lr": 1.2338254337207971e-05}
{"train_loss": 0.05855291336774826, "global_step": 209540, "epoch": 2354, "lr": 1.2337873017061946e-05}
{"train_loss": 0.062103912234306335, "global_step": 209541, "epoch": 2354, "lr": 1.2337491701979126e-05}
{"train_loss": 0.0634894147515297, "global_step": 209542, "epoch": 2354, "lr": 1.2337110391959578e-05}
{"train_loss": 0.08325786143541336, "global_step": 209543, "epoch": 2354, "lr": 1.2336729087003334e-05}
{"train_loss": 0.04190540686249733, "global_step": 209544, "epoch": 2354, "lr": 1.2336347787110475e-05}
{"train_loss": 0.06600430607795715, "global_step": 209545, "epoch": 2354, "lr": 1.2335966492281021e-05}
{"train_loss": 0.03938506916165352, "global_step": 209546, "epoch": 2354, "lr": 1.2335585202515049e-05}
{"train_loss": 0.1029093936085701, "global_step": 209547, "epoch": 2354, "lr": 1.2335203917812582e-05}
{"train_loss": 0.026267172768712044, "global_step": 209548, "epoch": 2354, "lr": 1.2334822638173704e-05}
{"train_loss": 0.12704595923423767, "global_step": 209549, "epoch": 2354, "lr": 1.2334441363598437e-05}
{"train_loss": 0.04654944688081741, "global_step": 209550, "epoch": 2354, "lr": 1.2334060094086863e-05}
{"train_loss": 0.04968293756246567, "global_step": 209551, "epoch": 2354, "lr": 1.2333678829639012e-05}
{"train_loss": 0.07617705315351486, "global_step": 209552, "epoch": 2354, "lr": 1.2333297570254921e-05}
{"train_loss": 0.04330438748002052, "global_step": 209553, "epoch": 2354, "lr": 1.2332916315934679e-05}
{"train_loss": 0.02632972225546837, "global_step": 209554, "epoch": 2354, "lr": 1.2332535066678296e-05}
{"train_loss": 0.05532525107264519, "global_step": 209555, "epoch": 2354, "lr": 1.2332153822485865e-05}
{"train_loss": 0.0351017564535141, "global_step": 209556, "epoch": 2354, "lr": 1.2331772583357404e-05}
{"train_loss": 0.07169640064239502, "global_step": 209557, "epoch": 2354, "lr": 1.2331391349292986e-05}
{"train_loss": 0.06951192766427994, "global_step": 209558, "epoch": 2354, "lr": 1.2331010120292641e-05}
{"train_loss": 0.015988469123840332, "global_step": 209559, "epoch": 2354, "lr": 1.233062889635645e-05}
{"train_loss": 0.03778734803199768, "global_step": 209560, "epoch": 2354, "lr": 1.2330247677484425e-05}
{"train_loss": 0.04018067941069603, "global_step": 209561, "epoch": 2354, "lr": 1.232986646367666e-05}
{"train_loss": 0.04367030784487724, "global_step": 209562, "epoch": 2354, "lr": 1.2329485254933171e-05}
{"train_loss": 0.061479926109313965, "global_step": 209563, "epoch": 2354, "lr": 1.232910405125402e-05}
{"train_loss": 0.06959368288516998, "global_step": 209564, "epoch": 2354, "lr": 1.232872285263928e-05}
{"train_loss": 0.08016317337751389, "global_step": 209565, "epoch": 2354, "lr": 1.2328341659088966e-05}
{"train_loss": 0.10890262573957443, "global_step": 209566, "epoch": 2354, "lr": 1.2327960470603162e-05}
{"train_loss": 0.04356222599744797, "global_step": 209567, "epoch": 2354, "lr": 1.2327579287181889e-05}
{"train_loss": 0.047808531671762466, "global_step": 209568, "epoch": 2354, "lr": 1.2327198108825234e-05}
{"train_loss": 0.05225349962711334, "global_step": 209569, "epoch": 2354, "lr": 1.2326816935533209e-05}
{"train_loss": 0.09370852261781693, "global_step": 209570, "epoch": 2354, "lr": 1.23264357673059e-05}
{"train_loss": 0.08366110920906067, "global_step": 209571, "epoch": 2354, "lr": 1.2326054604143327e-05}
{"train_loss": 0.09131838381290436, "global_step": 209572, "epoch": 2354, "lr": 1.2325673446045572e-05}
{"train_loss": 0.095257468521595, "global_step": 209573, "epoch": 2354, "lr": 1.2325292293012657e-05}
{"train_loss": 0.052196111530065536, "global_step": 209574, "epoch": 2354, "lr": 1.232491114504466e-05}
{"train_loss": 0.07310508191585541, "global_step": 209575, "epoch": 2354, "lr": 1.232453000214161e-05}
{"train_loss": 0.04011917859315872, "global_step": 209576, "epoch": 2354, "lr": 1.2324148864303575e-05}
{"train_loss": 0.09719906747341156, "global_step": 209577, "epoch": 2354, "lr": 1.232376773153059e-05}
{"train_loss": 0.04797399789094925, "global_step": 209578, "epoch": 2354, "lr": 1.2323386603822728e-05}
{"train_loss": 0.039093296974897385, "global_step": 209579, "epoch": 2354, "lr": 1.2323005481180011e-05}
{"train_loss": 0.11798836290836334, "global_step": 209580, "epoch": 2354, "lr": 1.2322624363602525e-05}
{"train_loss": 0.05890825018286705, "global_step": 209581, "epoch": 2354, "lr": 1.2322243251090288e-05}
{"train_loss": 0.09930074959993362, "global_step": 209582, "epoch": 2354, "lr": 1.2321862143643365e-05}
{"train_loss": 0.0788731575012207, "global_step": 209583, "epoch": 2354, "lr": 1.2321481041261818e-05}
{"train_loss": 0.01588316820561886, "global_step": 209584, "epoch": 2354, "lr": 1.2321099943945697e-05}
{"train_loss": 0.0475914441049099, "global_step": 209585, "epoch": 2354, "lr": 1.2320718851695023e-05}
{"train_loss": 0.02988998405635357, "global_step": 209586, "epoch": 2354, "lr": 1.2320337764509887e-05}
{"train_loss": 0.07371017336845398, "global_step": 209587, "epoch": 2354, "lr": 1.231995668239031e-05}
{"train_loss": 0.03425304964184761, "global_step": 209588, "epoch": 2354, "lr": 1.231957560533637e-05}
{"train_loss": 0.09386350214481354, "global_step": 209589, "epoch": 2354, "lr": 1.23191945333481e-05}
{"train_loss": 0.05913865566253662, "global_step": 209590, "epoch": 2354, "lr": 1.2318813466425543e-05}
{"train_loss": 0.07934019714593887, "global_step": 209591, "epoch": 2354, "lr": 1.2318432404568775e-05}
{"train_loss": 0.04632717743515968, "global_step": 209592, "epoch": 2354, "lr": 1.231805134777782e-05}
{"train_loss": 0.06917218863964081, "global_step": 209593, "epoch": 2354, "lr": 1.2317670296052764e-05}
{"train_loss": 0.062353626835379705, "global_step": 209594, "epoch": 2354, "lr": 1.2317289249393615e-05, "val_loss": 8.432557106018066}
{"train_loss": 0.1104610487818718, "global_step": 209595, "epoch": 2355, "lr": 1.2316908207800464e-05}
{"train_loss": 0.0505756214261055, "global_step": 209596, "epoch": 2355, "lr": 1.2316527171273335e-05}
{"train_loss": 0.12596163153648376, "global_step": 209597, "epoch": 2355, "lr": 1.2316146139812302e-05}
{"train_loss": 0.06670229136943817, "global_step": 209598, "epoch": 2355, "lr": 1.2315765113417387e-05}
{"train_loss": 0.05657360702753067, "global_step": 209599, "epoch": 2355, "lr": 1.2315384092088662e-05}
{"train_loss": 0.056722912937402725, "global_step": 209600, "epoch": 2355, "lr": 1.2315003075826188e-05}
{"train_loss": 0.06033426895737648, "global_step": 209601, "epoch": 2355, "lr": 1.2314622064629982e-05}
{"train_loss": 0.09685519337654114, "global_step": 209602, "epoch": 2355, "lr": 1.2314241058500136e-05}
{"train_loss": 0.08587329834699631, "global_step": 209603, "epoch": 2355, "lr": 1.2313860057436665e-05}
{"train_loss": 0.03247186541557312, "global_step": 209604, "epoch": 2355, "lr": 1.2313479061439653e-05}
{"train_loss": 0.0422620065510273, "global_step": 209605, "epoch": 2355, "lr": 1.2313098070509117e-05}
{"train_loss": 0.09995028376579285, "global_step": 209606, "epoch": 2355, "lr": 1.2312717084645142e-05}
{"train_loss": 0.06920682638883591, "global_step": 209607, "epoch": 2355, "lr": 1.231233610384775e-05}
{"train_loss": 0.11101971566677094, "global_step": 209608, "epoch": 2355, "lr": 1.2311955128117015e-05}
{"train_loss": 0.053596630692481995, "global_step": 209609, "epoch": 2355, "lr": 1.2311574157452965e-05}
{"train_loss": 0.0719466283917427, "global_step": 209610, "epoch": 2355, "lr": 1.231119319185568e-05}
{"train_loss": 0.04075304791331291, "global_step": 209611, "epoch": 2355, "lr": 1.2310812231325181e-05}
{"train_loss": 0.05728213116526604, "global_step": 209612, "epoch": 2355, "lr": 1.2310431275861551e-05}
{"train_loss": 0.05132101848721504, "global_step": 209613, "epoch": 2355, "lr": 1.2310050325464806e-05}
{"train_loss": 0.07788845896720886, "global_step": 209614, "epoch": 2355, "lr": 1.2309669380135031e-05}
{"train_loss": 0.04741469770669937, "global_step": 209615, "epoch": 2355, "lr": 1.2309288439872246e-05}
{"train_loss": 0.06569449603557587, "global_step": 209616, "epoch": 2355, "lr": 1.230890750467653e-05}
{"train_loss": 0.05307180434465408, "global_step": 209617, "epoch": 2355, "lr": 1.2308526574547913e-05}
{"train_loss": 0.05318174138665199, "global_step": 209618, "epoch": 2355, "lr": 1.2308145649486453e-05}
{"train_loss": 0.030368899926543236, "global_step": 209619, "epoch": 2355, "lr": 1.2307764729492222e-05}
{"train_loss": 0.05640774965286255, "global_step": 209620, "epoch": 2355, "lr": 1.2307383814565233e-05}
{"train_loss": 0.06266666203737259, "global_step": 209621, "epoch": 2355, "lr": 1.2307002904705572e-05}
{"train_loss": 0.06576872617006302, "global_step": 209622, "epoch": 2355, "lr": 1.2306621999913276e-05}
{"train_loss": 0.12111935764551163, "global_step": 209623, "epoch": 2355, "lr": 1.2306241100188376e-05}
{"train_loss": 0.048783618956804276, "global_step": 209624, "epoch": 2355, "lr": 1.2305860205530961e-05}
{"train_loss": 0.04437991604208946, "global_step": 209625, "epoch": 2355, "lr": 1.2305479315941053e-05}
{"train_loss": 0.07136179506778717, "global_step": 209626, "epoch": 2355, "lr": 1.2305098431418727e-05}
{"train_loss": 0.088260218501091, "global_step": 209627, "epoch": 2355, "lr": 1.2304717551964001e-05}
{"train_loss": 0.04771324247121811, "global_step": 209628, "epoch": 2355, "lr": 1.2304336677576966e-05}
{"train_loss": 0.05623640865087509, "global_step": 209629, "epoch": 2355, "lr": 1.230395580825765e-05}
{"train_loss": 0.026917748153209686, "global_step": 209630, "epoch": 2355, "lr": 1.2303574944006096e-05}
{"train_loss": 0.05784536525607109, "global_step": 209631, "epoch": 2355, "lr": 1.2303194084822379e-05}
{"train_loss": 0.069040447473526, "global_step": 209632, "epoch": 2355, "lr": 1.2302813230706522e-05}
{"train_loss": 0.020965905860066414, "global_step": 209633, "epoch": 2355, "lr": 1.2302432381658613e-05}
{"train_loss": 0.07169149070978165, "global_step": 209634, "epoch": 2355, "lr": 1.2302051537678666e-05}
{"train_loss": 0.029642248526215553, "global_step": 209635, "epoch": 2355, "lr": 1.230167069876676e-05}
{"train_loss": 0.04642476513981819, "global_step": 209636, "epoch": 2355, "lr": 1.2301289864922922e-05}
{"train_loss": 0.06924557685852051, "global_step": 209637, "epoch": 2355, "lr": 1.2300909036147218e-05}
{"train_loss": 0.06458499282598495, "global_step": 209638, "epoch": 2355, "lr": 1.230052821243971e-05}
{"train_loss": 0.0992560014128685, "global_step": 209639, "epoch": 2355, "lr": 1.2300147393800421e-05}
{"train_loss": 0.06621433794498444, "global_step": 209640, "epoch": 2355, "lr": 1.2299766580229433e-05}
{"train_loss": 0.06440972536802292, "global_step": 209641, "epoch": 2355, "lr": 1.2299385771726768e-05}
{"train_loss": 0.0713944360613823, "global_step": 209642, "epoch": 2355, "lr": 1.2299004968292505e-05}
{"train_loss": 0.01972644403576851, "global_step": 209643, "epoch": 2355, "lr": 1.2298624169926664e-05}
{"train_loss": 0.0782010480761528, "global_step": 209644, "epoch": 2355, "lr": 1.2298243376629332e-05}
{"train_loss": 0.041022252291440964, "global_step": 209645, "epoch": 2355, "lr": 1.2297862588400528e-05}
{"train_loss": 0.07735870033502579, "global_step": 209646, "epoch": 2355, "lr": 1.229748180524033e-05}
{"train_loss": 0.05120972916483879, "global_step": 209647, "epoch": 2355, "lr": 1.229710102714876e-05}
{"train_loss": 0.07203389704227448, "global_step": 209648, "epoch": 2355, "lr": 1.2296720254125898e-05}
{"train_loss": 0.05564916878938675, "global_step": 209649, "epoch": 2355, "lr": 1.2296339486171771e-05}
{"train_loss": 0.0983610674738884, "global_step": 209650, "epoch": 2355, "lr": 1.2295958723286456e-05}
{"train_loss": 0.05541791021823883, "global_step": 209651, "epoch": 2355, "lr": 1.2295577965469978e-05}
{"train_loss": 0.06509780883789062, "global_step": 209652, "epoch": 2355, "lr": 1.229519721272241e-05}
{"train_loss": 0.020230470225214958, "global_step": 209653, "epoch": 2355, "lr": 1.2294816465043778e-05}
{"train_loss": 0.03673073276877403, "global_step": 209654, "epoch": 2355, "lr": 1.229443572243415e-05}
{"train_loss": 0.03515256196260452, "global_step": 209655, "epoch": 2355, "lr": 1.2294054984893594e-05}
{"train_loss": 0.07723352313041687, "global_step": 209656, "epoch": 2355, "lr": 1.2293674252422121e-05}
{"train_loss": 0.04955420643091202, "global_step": 209657, "epoch": 2355, "lr": 1.2293293525019827e-05}
{"train_loss": 0.05013225972652435, "global_step": 209658, "epoch": 2355, "lr": 1.2292912802686719e-05}
{"train_loss": 0.0459771603345871, "global_step": 209659, "epoch": 2355, "lr": 1.2292532085422887e-05}
{"train_loss": 0.07167229056358337, "global_step": 209660, "epoch": 2355, "lr": 1.229215137322835e-05}
{"train_loss": 0.06318369507789612, "global_step": 209661, "epoch": 2355, "lr": 1.2291770666103185e-05}
{"train_loss": 0.045757848769426346, "global_step": 209662, "epoch": 2355, "lr": 1.2291389964047434e-05}
{"train_loss": 0.06542588025331497, "global_step": 209663, "epoch": 2355, "lr": 1.2291009267061132e-05}
{"train_loss": 0.07041963189840317, "global_step": 209664, "epoch": 2355, "lr": 1.2290628575144359e-05}
{"train_loss": 0.04878653958439827, "global_step": 209665, "epoch": 2355, "lr": 1.2290247888297135e-05}
{"train_loss": 0.07812851667404175, "global_step": 209666, "epoch": 2355, "lr": 1.228986720651954e-05}
{"train_loss": 0.05240826681256294, "global_step": 209667, "epoch": 2355, "lr": 1.2289486529811617e-05}
{"train_loss": 0.07246677577495575, "global_step": 209668, "epoch": 2355, "lr": 1.2289105858173395e-05}
{"train_loss": 0.0753830298781395, "global_step": 209669, "epoch": 2355, "lr": 1.2288725191604961e-05}
{"train_loss": 0.060888372361660004, "global_step": 209670, "epoch": 2355, "lr": 1.2288344530106333e-05}
{"train_loss": 0.07681697607040405, "global_step": 209671, "epoch": 2355, "lr": 1.2287963873677589e-05}
{"train_loss": 0.020215021446347237, "global_step": 209672, "epoch": 2355, "lr": 1.2287583222318755e-05}
{"train_loss": 0.06912428140640259, "global_step": 209673, "epoch": 2355, "lr": 1.22872025760299e-05}
{"train_loss": 0.07129786163568497, "global_step": 209674, "epoch": 2355, "lr": 1.2286821934811083e-05}
{"train_loss": 0.10214779525995255, "global_step": 209675, "epoch": 2355, "lr": 1.2286441298662327e-05}
{"train_loss": 0.04834550991654396, "global_step": 209676, "epoch": 2355, "lr": 1.2286060667583715e-05}
{"train_loss": 0.05267568677663803, "global_step": 209677, "epoch": 2355, "lr": 1.228568004157527e-05}
{"train_loss": 0.03280360624194145, "global_step": 209678, "epoch": 2355, "lr": 1.2285299420637064e-05}
{"train_loss": 0.038816217333078384, "global_step": 209679, "epoch": 2355, "lr": 1.228491880476913e-05}
{"train_loss": 0.07420287281274796, "global_step": 209680, "epoch": 2355, "lr": 1.2284538193971546e-05}
{"train_loss": 0.028976138681173325, "global_step": 209681, "epoch": 2355, "lr": 1.2284157588244327e-05}
{"train_loss": 0.05303690582513809, "global_step": 209682, "epoch": 2355, "lr": 1.2283776987587558e-05}
{"train_loss": 0.06205309209612648, "global_step": 209683, "epoch": 2355, "lr": 1.2283396392001262e-05, "val_loss": 8.617470741271973, "train_action_mse_error": 29.840412139892578}
{"train_loss": 0.08361320197582245, "global_step": 209684, "epoch": 2356, "lr": 1.228301580148552e-05}
{"train_loss": 0.05566032975912094, "global_step": 209685, "epoch": 2356, "lr": 1.228263521604035e-05}
{"train_loss": 0.06483950465917587, "global_step": 209686, "epoch": 2356, "lr": 1.2282254635665835e-05}
{"train_loss": 0.05842890217900276, "global_step": 209687, "epoch": 2356, "lr": 1.2281874060361998e-05}
{"train_loss": 0.059342339634895325, "global_step": 209688, "epoch": 2356, "lr": 1.2281493490128915e-05}
{"train_loss": 0.07830443233251572, "global_step": 209689, "epoch": 2356, "lr": 1.2281112924966615e-05}
{"train_loss": 0.06985892355442047, "global_step": 209690, "epoch": 2356, "lr": 1.2280732364875158e-05}
{"train_loss": 0.12553659081459045, "global_step": 209691, "epoch": 2356, "lr": 1.2280351809854612e-05}
{"train_loss": 0.07719023525714874, "global_step": 209692, "epoch": 2356, "lr": 1.2279971259904994e-05}
{"train_loss": 0.051624178886413574, "global_step": 209693, "epoch": 2356, "lr": 1.2279590715026396e-05}
{"train_loss": 0.0323808528482914, "global_step": 209694, "epoch": 2356, "lr": 1.2279210175218824e-05}
{"train_loss": 0.04378531500697136, "global_step": 209695, "epoch": 2356, "lr": 1.2278829640482375e-05}
{"train_loss": 0.0846211165189743, "global_step": 209696, "epoch": 2356, "lr": 1.2278449110817058e-05}
{"train_loss": 0.06329093873500824, "global_step": 209697, "epoch": 2356, "lr": 1.2278068586222957e-05}
{"train_loss": 0.04898178577423096, "global_step": 209698, "epoch": 2356, "lr": 1.2277688066700099e-05}
{"train_loss": 0.08848883211612701, "global_step": 209699, "epoch": 2356, "lr": 1.2277307552248562e-05}
{"train_loss": 0.07688075304031372, "global_step": 209700, "epoch": 2356, "lr": 1.227692704286838e-05}
{"train_loss": 0.03217560425400734, "global_step": 209701, "epoch": 2356, "lr": 1.2276546538559592e-05}
{"train_loss": 0.033199869096279144, "global_step": 209702, "epoch": 2356, "lr": 1.2276166039322273e-05}
{"train_loss": 0.08383473753929138, "global_step": 209703, "epoch": 2356, "lr": 1.2275785545156449e-05}
{"train_loss": 0.11653120815753937, "global_step": 209704, "epoch": 2356, "lr": 1.2275405056062205e-05}
{"train_loss": 0.07695389539003372, "global_step": 209705, "epoch": 2356, "lr": 1.2275024572039567e-05}
{"train_loss": 0.0516485869884491, "global_step": 209706, "epoch": 2356, "lr": 1.2274644093088583e-05}
{"train_loss": 0.06664231419563293, "global_step": 209707, "epoch": 2356, "lr": 1.2274263619209325e-05}
{"train_loss": 0.04849470779299736, "global_step": 209708, "epoch": 2356, "lr": 1.2273883150401821e-05}
{"train_loss": 0.07746977359056473, "global_step": 209709, "epoch": 2356, "lr": 1.2273502686666133e-05}
{"train_loss": 0.10627244412899017, "global_step": 209710, "epoch": 2356, "lr": 1.2273122228002321e-05}
{"train_loss": 0.052485715597867966, "global_step": 209711, "epoch": 2356, "lr": 1.2272741774410418e-05}
{"train_loss": 0.09287581592798233, "global_step": 209712, "epoch": 2356, "lr": 1.2272361325890502e-05}
{"train_loss": 0.09245816618204117, "global_step": 209713, "epoch": 2356, "lr": 1.2271980882442591e-05}
{"train_loss": 0.06571529060602188, "global_step": 209714, "epoch": 2356, "lr": 1.2271600444066766e-05}
{"train_loss": 0.06556802988052368, "global_step": 209715, "epoch": 2356, "lr": 1.2271220010763051e-05}
{"train_loss": 0.04149804264307022, "global_step": 209716, "epoch": 2356, "lr": 1.2270839582531523e-05}
{"train_loss": 0.07579158246517181, "global_step": 209717, "epoch": 2356, "lr": 1.2270459159372205e-05}
{"train_loss": 0.10167603939771652, "global_step": 209718, "epoch": 2356, "lr": 1.2270078741285184e-05}
{"train_loss": 0.029923317953944206, "global_step": 209719, "epoch": 2356, "lr": 1.2269698328270473e-05}
{"train_loss": 0.04300528019666672, "global_step": 209720, "epoch": 2356, "lr": 1.2269317920328154e-05}
{"train_loss": 0.05043742060661316, "global_step": 209721, "epoch": 2356, "lr": 1.2268937517458257e-05}
{"train_loss": 0.06795341521501541, "global_step": 209722, "epoch": 2356, "lr": 1.2268557119660851e-05}
{"train_loss": 0.05497545376420021, "global_step": 209723, "epoch": 2356, "lr": 1.2268176726935966e-05}
{"train_loss": 0.06580527126789093, "global_step": 209724, "epoch": 2356, "lr": 1.2267796339283672e-05}
{"train_loss": 0.04189348220825195, "global_step": 209725, "epoch": 2356, "lr": 1.2267415956704003e-05}
{"train_loss": 0.061889275908470154, "global_step": 209726, "epoch": 2356, "lr": 1.2267035579197023e-05}
{"train_loss": 0.042340319603681564, "global_step": 209727, "epoch": 2356, "lr": 1.2266655206762795e-05}
{"train_loss": 0.06417930126190186, "global_step": 209728, "epoch": 2356, "lr": 1.2266274839401337e-05}
{"train_loss": 0.04893937334418297, "global_step": 209729, "epoch": 2356, "lr": 1.2265894477112733e-05}
{"train_loss": 0.0499233677983284, "global_step": 209730, "epoch": 2356, "lr": 1.2265514119897009e-05}
{"train_loss": 0.07117652893066406, "global_step": 209731, "epoch": 2356, "lr": 1.2265133767754234e-05}
{"train_loss": 0.0775555893778801, "global_step": 209732, "epoch": 2356, "lr": 1.2264753420684444e-05}
{"train_loss": 0.040639232844114304, "global_step": 209733, "epoch": 2356, "lr": 1.2264373078687708e-05}
{"train_loss": 0.04958978295326233, "global_step": 209734, "epoch": 2356, "lr": 1.2263992741764058e-05}
{"train_loss": 0.06100144237279892, "global_step": 209735, "epoch": 2356, "lr": 1.2263612409913567e-05}
{"train_loss": 0.045762572437524796, "global_step": 209736, "epoch": 2356, "lr": 1.2263232083136256e-05}
{"train_loss": 0.0497545562684536, "global_step": 209737, "epoch": 2356, "lr": 1.226285176143221e-05}
{"train_loss": 0.06504283845424652, "global_step": 209738, "epoch": 2356, "lr": 1.2262471444801448e-05}
{"train_loss": 0.07503626495599747, "global_step": 209739, "epoch": 2356, "lr": 1.2262091133244052e-05}
{"train_loss": 0.04969333857297897, "global_step": 209740, "epoch": 2356, "lr": 1.226171082676006e-05}
{"train_loss": 0.08643116801977158, "global_step": 209741, "epoch": 2356, "lr": 1.2261330525349501e-05}
{"train_loss": 0.048707906156778336, "global_step": 209742, "epoch": 2356, "lr": 1.2260950229012464e-05}
{"train_loss": 0.09583922475576401, "global_step": 209743, "epoch": 2356, "lr": 1.2260569937748967e-05}
{"train_loss": 0.05654362589120865, "global_step": 209744, "epoch": 2356, "lr": 1.226018965155909e-05}
{"train_loss": 0.08414523303508759, "global_step": 209745, "epoch": 2356, "lr": 1.225980937044286e-05}
{"train_loss": 0.03550991415977478, "global_step": 209746, "epoch": 2356, "lr": 1.2259429094400349e-05}
{"train_loss": 0.04355616122484207, "global_step": 209747, "epoch": 2356, "lr": 1.2259048823431584e-05}
{"train_loss": 0.012497053481638432, "global_step": 209748, "epoch": 2356, "lr": 1.2258668557536645e-05}
{"train_loss": 0.04724786803126335, "global_step": 209749, "epoch": 2356, "lr": 1.2258288296715558e-05}
{"train_loss": 0.03506296128034592, "global_step": 209750, "epoch": 2356, "lr": 1.225790804096839e-05}
{"train_loss": 0.05201474204659462, "global_step": 209751, "epoch": 2356, "lr": 1.2257527790295182e-05}
{"train_loss": 0.060348134487867355, "global_step": 209752, "epoch": 2356, "lr": 1.2257147544695996e-05}
{"train_loss": 0.04581828415393829, "global_step": 209753, "epoch": 2356, "lr": 1.2256767304170864e-05}
{"train_loss": 0.030702803283929825, "global_step": 209754, "epoch": 2356, "lr": 1.2256387068719866e-05}
{"train_loss": 0.10947690904140472, "global_step": 209755, "epoch": 2356, "lr": 1.225600683834302e-05}
{"train_loss": 0.09027278423309326, "global_step": 209756, "epoch": 2356, "lr": 1.2255626613040405e-05}
{"train_loss": 0.12009704113006592, "global_step": 209757, "epoch": 2356, "lr": 1.2255246392812053e-05}
{"train_loss": 0.038051050156354904, "global_step": 209758, "epoch": 2356, "lr": 1.2254866177658036e-05}
{"train_loss": 0.05828419327735901, "global_step": 209759, "epoch": 2356, "lr": 1.2254485967578377e-05}
{"train_loss": 0.05460192635655403, "global_step": 209760, "epoch": 2356, "lr": 1.2254105762573153e-05}
{"train_loss": 0.010954124853014946, "global_step": 209761, "epoch": 2356, "lr": 1.2253725562642393e-05}
{"train_loss": 0.039756812155246735, "global_step": 209762, "epoch": 2356, "lr": 1.2253345367786173e-05}
{"train_loss": 0.03704609349370003, "global_step": 209763, "epoch": 2356, "lr": 1.2252965178004521e-05}
{"train_loss": 0.059831660240888596, "global_step": 209764, "epoch": 2356, "lr": 1.2252584993297495e-05}
{"train_loss": 0.058568358421325684, "global_step": 209765, "epoch": 2356, "lr": 1.2252204813665164e-05}
{"train_loss": 0.08396393060684204, "global_step": 209766, "epoch": 2356, "lr": 1.2251824639107546e-05}
{"train_loss": 0.051437702029943466, "global_step": 209767, "epoch": 2356, "lr": 1.2251444469624724e-05}
{"train_loss": 0.08607861399650574, "global_step": 209768, "epoch": 2356, "lr": 1.2251064305216725e-05}
{"train_loss": 0.044024351984262466, "global_step": 209769, "epoch": 2356, "lr": 1.2250684145883623e-05}
{"train_loss": 0.03785143420100212, "global_step": 209770, "epoch": 2356, "lr": 1.2250303991625444e-05}
{"train_loss": 0.07386834174394608, "global_step": 209771, "epoch": 2356, "lr": 1.2249923842442263e-05}
{"train_loss": 0.062494695741222815, "global_step": 209772, "epoch": 2356, "lr": 1.2249543698334104e-05, "val_loss": 8.580883026123047}
{"train_loss": 0.06289077550172806, "global_step": 209773, "epoch": 2357, "lr": 1.2249163559301052e-05}
{"train_loss": 0.08541985601186752, "global_step": 209774, "epoch": 2357, "lr": 1.2248783425343124e-05}
{"train_loss": 0.08144886046648026, "global_step": 209775, "epoch": 2357, "lr": 1.2248403296460403e-05}
{"train_loss": 0.03684273362159729, "global_step": 209776, "epoch": 2357, "lr": 1.2248023172652906e-05}
{"train_loss": 0.0813901498913765, "global_step": 209777, "epoch": 2357, "lr": 1.224764305392072e-05}
{"train_loss": 0.06525683403015137, "global_step": 209778, "epoch": 2357, "lr": 1.2247262940263876e-05}
{"train_loss": 0.031064976006746292, "global_step": 209779, "epoch": 2357, "lr": 1.2246882831682415e-05}
{"train_loss": 0.06564515084028244, "global_step": 209780, "epoch": 2357, "lr": 1.2246502728176413e-05}
{"train_loss": 0.01776289753615856, "global_step": 209781, "epoch": 2357, "lr": 1.2246122629745892e-05}
{"train_loss": 0.035423967987298965, "global_step": 209782, "epoch": 2357, "lr": 1.2245742536390936e-05}
{"train_loss": 0.05564800277352333, "global_step": 209783, "epoch": 2357, "lr": 1.2245362448111568e-05}
{"train_loss": 0.026053374633193016, "global_step": 209784, "epoch": 2357, "lr": 1.2244982364907865e-05}
{"train_loss": 0.0523109957575798, "global_step": 209785, "epoch": 2357, "lr": 1.2244602286779843e-05}
{"train_loss": 0.029507271945476532, "global_step": 209786, "epoch": 2357, "lr": 1.2244222213727596e-05}
{"train_loss": 0.03174206614494324, "global_step": 209787, "epoch": 2357, "lr": 1.2243842145751139e-05}
{"train_loss": 0.09644491225481033, "global_step": 209788, "epoch": 2357, "lr": 1.2243462082850549e-05}
{"train_loss": 0.039372414350509644, "global_step": 209789, "epoch": 2357, "lr": 1.2243082025025848e-05}
{"train_loss": 0.043450307101011276, "global_step": 209790, "epoch": 2357, "lr": 1.2242701972277121e-05}
{"train_loss": 0.10219048708677292, "global_step": 209791, "epoch": 2357, "lr": 1.224232192460439e-05}
{"train_loss": 0.0826135128736496, "global_step": 209792, "epoch": 2357, "lr": 1.224194188200773e-05}
{"train_loss": 0.06006106734275818, "global_step": 209793, "epoch": 2357, "lr": 1.2241561844487171e-05}
{"train_loss": 0.051680147647857666, "global_step": 209794, "epoch": 2357, "lr": 1.2241181812042784e-05}
{"train_loss": 0.028499405831098557, "global_step": 209795, "epoch": 2357, "lr": 1.2240801784674594e-05}
{"train_loss": 0.06478939950466156, "global_step": 209796, "epoch": 2357, "lr": 1.2240421762382687e-05}
{"train_loss": 0.10400297492742538, "global_step": 209797, "epoch": 2357, "lr": 1.224004174516708e-05}
{"train_loss": 0.03579559177160263, "global_step": 209798, "epoch": 2357, "lr": 1.2239661733027851e-05}
{"train_loss": 0.03572652116417885, "global_step": 209799, "epoch": 2357, "lr": 1.2239281725965023e-05}
{"train_loss": 0.03708238527178764, "global_step": 209800, "epoch": 2357, "lr": 1.2238901723978669e-05}
{"train_loss": 0.06196656823158264, "global_step": 209801, "epoch": 2357, "lr": 1.2238521727068847e-05}
{"train_loss": 0.08725525438785553, "global_step": 209802, "epoch": 2357, "lr": 1.223814173523558e-05}
{"train_loss": 0.043420761823654175, "global_step": 209803, "epoch": 2357, "lr": 1.223776174847895e-05}
{"train_loss": 0.06321196258068085, "global_step": 209804, "epoch": 2357, "lr": 1.2237381766798978e-05}
{"train_loss": 0.01451824139803648, "global_step": 209805, "epoch": 2357, "lr": 1.223700179019574e-05}
{"train_loss": 0.02589287981390953, "global_step": 209806, "epoch": 2357, "lr": 1.223662181866927e-05}
{"train_loss": 0.021675726398825645, "global_step": 209807, "epoch": 2357, "lr": 1.2236241852219632e-05}
{"train_loss": 0.034517258405685425, "global_step": 209808, "epoch": 2357, "lr": 1.223586189084686e-05}
{"train_loss": 0.03284011036157608, "global_step": 209809, "epoch": 2357, "lr": 1.2235481934551035e-05}
{"train_loss": 0.12159883230924606, "global_step": 209810, "epoch": 2357, "lr": 1.223510198333217e-05}
{"train_loss": 0.06317393481731415, "global_step": 209811, "epoch": 2357, "lr": 1.2234722037190356e-05}
{"train_loss": 0.08072055131196976, "global_step": 209812, "epoch": 2357, "lr": 1.22343420961256e-05}
{"train_loss": 0.06455563008785248, "global_step": 209813, "epoch": 2357, "lr": 1.2233962160137997e-05}
{"train_loss": 0.053344979882240295, "global_step": 209814, "epoch": 2357, "lr": 1.2233582229227558e-05}
{"train_loss": 0.082069531083107, "global_step": 209815, "epoch": 2357, "lr": 1.223320230339437e-05}
{"train_loss": 0.09962999820709229, "global_step": 209816, "epoch": 2357, "lr": 1.223282238263847e-05}
{"train_loss": 0.018297599628567696, "global_step": 209817, "epoch": 2357, "lr": 1.2232442466959892e-05}
{"train_loss": 0.04378264397382736, "global_step": 209818, "epoch": 2357, "lr": 1.2232062556358715e-05}
{"train_loss": 0.02448396384716034, "global_step": 209819, "epoch": 2357, "lr": 1.2231682650834964e-05}
{"train_loss": 0.0892362892627716, "global_step": 209820, "epoch": 2357, "lr": 1.2231302750388712e-05}
{"train_loss": 0.08813665807247162, "global_step": 209821, "epoch": 2357, "lr": 1.2230922855019994e-05}
{"train_loss": 0.01648910716176033, "global_step": 209822, "epoch": 2357, "lr": 1.2230542964728876e-05}
{"train_loss": 0.10146857053041458, "global_step": 209823, "epoch": 2357, "lr": 1.2230163079515388e-05}
{"train_loss": 0.01867298223078251, "global_step": 209824, "epoch": 2357, "lr": 1.2229783199379614e-05}
{"train_loss": 0.04119105637073517, "global_step": 209825, "epoch": 2357, "lr": 1.2229403324321565e-05}
{"train_loss": 0.061020057648420334, "global_step": 209826, "epoch": 2357, "lr": 1.2229023454341331e-05}
{"train_loss": 0.048731740564107895, "global_step": 209827, "epoch": 2357, "lr": 1.2228643589438926e-05}
{"train_loss": 0.07294368743896484, "global_step": 209828, "epoch": 2357, "lr": 1.2228263729614436e-05}
{"train_loss": 0.0541219599545002, "global_step": 209829, "epoch": 2357, "lr": 1.2227883874867878e-05}
{"train_loss": 0.07438720762729645, "global_step": 209830, "epoch": 2357, "lr": 1.2227504025199344e-05}
{"train_loss": 0.016929563134908676, "global_step": 209831, "epoch": 2357, "lr": 1.222712418060884e-05}
{"train_loss": 0.03977226838469505, "global_step": 209832, "epoch": 2357, "lr": 1.2226744341096452e-05}
{"train_loss": 0.06155886873602867, "global_step": 209833, "epoch": 2357, "lr": 1.2226364506662207e-05}
{"train_loss": 0.024348517879843712, "global_step": 209834, "epoch": 2357, "lr": 1.2225984677306178e-05}
{"train_loss": 0.03862001746892929, "global_step": 209835, "epoch": 2357, "lr": 1.2225604853028393e-05}
{"train_loss": 0.035612624138593674, "global_step": 209836, "epoch": 2357, "lr": 1.2225225033828918e-05}
{"train_loss": 0.05523868650197983, "global_step": 209837, "epoch": 2357, "lr": 1.2224845219707809e-05}
{"train_loss": 0.08363227546215057, "global_step": 209838, "epoch": 2357, "lr": 1.2224465410665097e-05}
{"train_loss": 0.10074803978204727, "global_step": 209839, "epoch": 2357, "lr": 1.2224085606700863e-05}
{"train_loss": 0.06887643039226532, "global_step": 209840, "epoch": 2357, "lr": 1.2223705807815122e-05}
{"train_loss": 0.062022678554058075, "global_step": 209841, "epoch": 2357, "lr": 1.2223326014007963e-05}
{"train_loss": 0.03592325747013092, "global_step": 209842, "epoch": 2357, "lr": 1.2222946225279397e-05}
{"train_loss": 0.06356455385684967, "global_step": 209843, "epoch": 2357, "lr": 1.2222566441629513e-05}
{"train_loss": 0.06930910795927048, "global_step": 209844, "epoch": 2357, "lr": 1.2222186663058328e-05}
{"train_loss": 0.025087758898735046, "global_step": 209845, "epoch": 2357, "lr": 1.2221806889565923e-05}
{"train_loss": 0.04352913051843643, "global_step": 209846, "epoch": 2357, "lr": 1.2221427121152324e-05}
{"train_loss": 0.045042891055345535, "global_step": 209847, "epoch": 2357, "lr": 1.222104735781761e-05}
{"train_loss": 0.08053795248270035, "global_step": 209848, "epoch": 2357, "lr": 1.2220667599561797e-05}
{"train_loss": 0.04633866623044014, "global_step": 209849, "epoch": 2357, "lr": 1.222028784638497e-05}
{"train_loss": 0.05720629170536995, "global_step": 209850, "epoch": 2357, "lr": 1.2219908098287154e-05}
{"train_loss": 0.06644835323095322, "global_step": 209851, "epoch": 2357, "lr": 1.2219528355268422e-05}
{"train_loss": 0.05019133910536766, "global_step": 209852, "epoch": 2357, "lr": 1.2219148617328802e-05}
{"train_loss": 0.049512073397636414, "global_step": 209853, "epoch": 2357, "lr": 1.2218768884468368e-05}
{"train_loss": 0.03092150203883648, "global_step": 209854, "epoch": 2357, "lr": 1.2218389156687143e-05}
{"train_loss": 0.0296842772513628, "global_step": 209855, "epoch": 2357, "lr": 1.2218009433985217e-05}
{"train_loss": 0.033826109021902084, "global_step": 209856, "epoch": 2357, "lr": 1.2217629716362617e-05}
{"train_loss": 0.08108901977539062, "global_step": 209857, "epoch": 2357, "lr": 1.221725000381938e-05}
{"train_loss": 0.030654657632112503, "global_step": 209858, "epoch": 2357, "lr": 1.2216870296355588e-05}
{"train_loss": 0.06072165071964264, "global_step": 209859, "epoch": 2357, "lr": 1.2216490593971263e-05}
{"train_loss": 0.05733827129006386, "global_step": 209860, "epoch": 2357, "lr": 1.2216110896666483e-05}
{"train_loss": 0.054431249282919286, "global_step": 209861, "epoch": 2357, "lr": 1.2215731204441272e-05, "val_loss": 8.542319297790527}
{"train_loss": 0.0737832710146904, "global_step": 209862, "epoch": 2358, "lr": 1.2215351517295714e-05}
{"train_loss": 0.03609941899776459, "global_step": 209863, "epoch": 2358, "lr": 1.2214971835229828e-05}
{"train_loss": 0.019516371190547943, "global_step": 209864, "epoch": 2358, "lr": 1.2214592158243688e-05}
{"train_loss": 0.028041476383805275, "global_step": 209865, "epoch": 2358, "lr": 1.2214212486337323e-05}
{"train_loss": 0.07507099956274033, "global_step": 209866, "epoch": 2358, "lr": 1.2213832819510806e-05}
{"train_loss": 0.09277866780757904, "global_step": 209867, "epoch": 2358, "lr": 1.2213453157764171e-05}
{"train_loss": 0.06358732283115387, "global_step": 209868, "epoch": 2358, "lr": 1.2213073501097488e-05}
{"train_loss": 0.046098873019218445, "global_step": 209869, "epoch": 2358, "lr": 1.2212693849510786e-05}
{"train_loss": 0.03997389227151871, "global_step": 209870, "epoch": 2358, "lr": 1.2212314203004138e-05}
{"train_loss": 0.07789814472198486, "global_step": 209871, "epoch": 2358, "lr": 1.221193456157757e-05}
{"train_loss": 0.048839740455150604, "global_step": 209872, "epoch": 2358, "lr": 1.2211554925231145e-05}
{"train_loss": 0.0826268270611763, "global_step": 209873, "epoch": 2358, "lr": 1.2211175293964933e-05}
{"train_loss": 0.09302636235952377, "global_step": 209874, "epoch": 2358, "lr": 1.221079566777895e-05}
{"train_loss": 0.05656926706433296, "global_step": 209875, "epoch": 2358, "lr": 1.2210416046673284e-05}
{"train_loss": 0.04357420280575752, "global_step": 209876, "epoch": 2358, "lr": 1.2210036430647947e-05}
{"train_loss": 0.04390430077910423, "global_step": 209877, "epoch": 2358, "lr": 1.220965681970303e-05}
{"train_loss": 0.09523213654756546, "global_step": 209878, "epoch": 2358, "lr": 1.2209277213838549e-05}
{"train_loss": 0.03609920293092728, "global_step": 209879, "epoch": 2358, "lr": 1.2208897613054583e-05}
{"train_loss": 0.057134777307510376, "global_step": 209880, "epoch": 2358, "lr": 1.2208518017351156e-05}
{"train_loss": 0.05698148533701897, "global_step": 209881, "epoch": 2358, "lr": 1.2208138426728344e-05}
{"train_loss": 0.06781657040119171, "global_step": 209882, "epoch": 2358, "lr": 1.2207758841186179e-05}
{"train_loss": 0.09343726933002472, "global_step": 209883, "epoch": 2358, "lr": 1.2207379260724727e-05}
{"train_loss": 0.05056124925613403, "global_step": 209884, "epoch": 2358, "lr": 1.2206999685344023e-05}
{"train_loss": 0.03720252960920334, "global_step": 209885, "epoch": 2358, "lr": 1.2206620115044143e-05}
{"train_loss": 0.05096134543418884, "global_step": 209886, "epoch": 2358, "lr": 1.2206240549825105e-05}
{"train_loss": 0.04731680080294609, "global_step": 209887, "epoch": 2358, "lr": 1.220586098968699e-05}
{"train_loss": 0.052763570100069046, "global_step": 209888, "epoch": 2358, "lr": 1.2205481434629829e-05}
{"train_loss": 0.08427462726831436, "global_step": 209889, "epoch": 2358, "lr": 1.2205101884653686e-05}
{"train_loss": 0.08037838339805603, "global_step": 209890, "epoch": 2358, "lr": 1.22047223397586e-05}
{"train_loss": 0.09260941296815872, "global_step": 209891, "epoch": 2358, "lr": 1.220434279994464e-05}
{"train_loss": 0.0798395425081253, "global_step": 209892, "epoch": 2358, "lr": 1.2203963265211826e-05}
{"train_loss": 0.05578700453042984, "global_step": 209893, "epoch": 2358, "lr": 1.2203583735560248e-05}
{"train_loss": 0.06313513219356537, "global_step": 209894, "epoch": 2358, "lr": 1.2203204210989939e-05}
{"train_loss": 0.07640710473060608, "global_step": 209895, "epoch": 2358, "lr": 1.220282469150093e-05}
{"train_loss": 0.035188496112823486, "global_step": 209896, "epoch": 2358, "lr": 1.2202445177093308e-05}
{"train_loss": 0.06619378179311752, "global_step": 209897, "epoch": 2358, "lr": 1.2202065667767088e-05}
{"train_loss": 0.04191511124372482, "global_step": 209898, "epoch": 2358, "lr": 1.2201686163522357e-05}
{"train_loss": 0.09848640114068985, "global_step": 209899, "epoch": 2358, "lr": 1.2201306664359136e-05}
{"train_loss": 0.1067446619272232, "global_step": 209900, "epoch": 2358, "lr": 1.2200927170277498e-05}
{"train_loss": 0.04575256630778313, "global_step": 209901, "epoch": 2358, "lr": 1.2200547681277474e-05}
{"train_loss": 0.0432814285159111, "global_step": 209902, "epoch": 2358, "lr": 1.220016819735914e-05}
{"train_loss": 0.0501566044986248, "global_step": 209903, "epoch": 2358, "lr": 1.2199788718522515e-05}
{"train_loss": 0.05588405206799507, "global_step": 209904, "epoch": 2358, "lr": 1.2199409244767685e-05}
{"train_loss": 0.053259726613759995, "global_step": 209905, "epoch": 2358, "lr": 1.2199029776094672e-05}
{"train_loss": 0.10161549597978592, "global_step": 209906, "epoch": 2358, "lr": 1.2198650312503545e-05}
{"train_loss": 0.062302373349666595, "global_step": 209907, "epoch": 2358, "lr": 1.2198270853994343e-05}
{"train_loss": 0.061723995953798294, "global_step": 209908, "epoch": 2358, "lr": 1.2197891400567118e-05}
{"train_loss": 0.06664618849754333, "global_step": 209909, "epoch": 2358, "lr": 1.219751195222194e-05}
{"train_loss": 0.08169408142566681, "global_step": 209910, "epoch": 2358, "lr": 1.2197132508958836e-05}
{"train_loss": 0.047802455723285675, "global_step": 209911, "epoch": 2358, "lr": 1.2196753070777878e-05}
{"train_loss": 0.08279705047607422, "global_step": 209912, "epoch": 2358, "lr": 1.2196373637679086e-05}
{"train_loss": 0.03317483887076378, "global_step": 209913, "epoch": 2358, "lr": 1.2195994209662554e-05}
{"train_loss": 0.03239668160676956, "global_step": 209914, "epoch": 2358, "lr": 1.2195614786728293e-05}
{"train_loss": 0.1331438273191452, "global_step": 209915, "epoch": 2358, "lr": 1.2195235368876384e-05}
{"train_loss": 0.0929143950343132, "global_step": 209916, "epoch": 2358, "lr": 1.2194855956106848e-05}
{"train_loss": 0.047816403210163116, "global_step": 209917, "epoch": 2358, "lr": 1.219447654841977e-05}
{"train_loss": 0.10080510377883911, "global_step": 209918, "epoch": 2358, "lr": 1.219409714581517e-05}
{"train_loss": 0.05789956450462341, "global_step": 209919, "epoch": 2358, "lr": 1.2193717748293127e-05}
{"train_loss": 0.08829952031373978, "global_step": 209920, "epoch": 2358, "lr": 1.2193338355853662e-05}
{"train_loss": 0.03491911664605141, "global_step": 209921, "epoch": 2358, "lr": 1.2192958968496854e-05}
{"train_loss": 0.0444202721118927, "global_step": 209922, "epoch": 2358, "lr": 1.219257958622273e-05}
{"train_loss": 0.040018413215875626, "global_step": 209923, "epoch": 2358, "lr": 1.2192200209031368e-05}
{"train_loss": 0.056534551084041595, "global_step": 209924, "epoch": 2358, "lr": 1.2191820836922791e-05}
{"train_loss": 0.033635854721069336, "global_step": 209925, "epoch": 2358, "lr": 1.2191441469897074e-05}
{"train_loss": 0.05866459012031555, "global_step": 209926, "epoch": 2358, "lr": 1.2191062107954243e-05}
{"train_loss": 0.0883176177740097, "global_step": 209927, "epoch": 2358, "lr": 1.219068275109438e-05}
{"train_loss": 0.043700456619262695, "global_step": 209928, "epoch": 2358, "lr": 1.2190303399317504e-05}
{"train_loss": 0.059758394956588745, "global_step": 209929, "epoch": 2358, "lr": 1.2189924052623691e-05}
{"train_loss": 0.02953966148197651, "global_step": 209930, "epoch": 2358, "lr": 1.2189544711012974e-05}
{"train_loss": 0.032580066472291946, "global_step": 209931, "epoch": 2358, "lr": 1.2189165374485424e-05}
{"train_loss": 0.06387918442487717, "global_step": 209932, "epoch": 2358, "lr": 1.2188786043041079e-05}
{"train_loss": 0.062156323343515396, "global_step": 209933, "epoch": 2358, "lr": 1.2188406716679974e-05}
{"train_loss": 0.028170451521873474, "global_step": 209934, "epoch": 2358, "lr": 1.2188027395402191e-05}
{"train_loss": 0.0404958613216877, "global_step": 209935, "epoch": 2358, "lr": 1.2187648079207758e-05}
{"train_loss": 0.07218685746192932, "global_step": 209936, "epoch": 2358, "lr": 1.2187268768096744e-05}
{"train_loss": 0.061999108642339706, "global_step": 209937, "epoch": 2358, "lr": 1.2186889462069178e-05}
{"train_loss": 0.029476072639226913, "global_step": 209938, "epoch": 2358, "lr": 1.2186510161125142e-05}
{"train_loss": 0.02732834778726101, "global_step": 209939, "epoch": 2358, "lr": 1.218613086526465e-05}
{"train_loss": 0.055261094123125076, "global_step": 209940, "epoch": 2358, "lr": 1.2185751574487786e-05}
{"train_loss": 0.016006488353013992, "global_step": 209941, "epoch": 2358, "lr": 1.2185372288794572e-05}
{"train_loss": 0.05955236032605171, "global_step": 209942, "epoch": 2358, "lr": 1.2184993008185091e-05}
{"train_loss": 0.05722640082240105, "global_step": 209943, "epoch": 2358, "lr": 1.2184613732659361e-05}
{"train_loss": 0.03985106572508812, "global_step": 209944, "epoch": 2358, "lr": 1.2184234462217447e-05}
{"train_loss": 0.08316916972398758, "global_step": 209945, "epoch": 2358, "lr": 1.2183855196859423e-05}
{"train_loss": 0.07428762316703796, "global_step": 209946, "epoch": 2358, "lr": 1.2183475936585298e-05}
{"train_loss": 0.04820011556148529, "global_step": 209947, "epoch": 2358, "lr": 1.2183096681395162e-05}
{"train_loss": 0.025676211342215538, "global_step": 209948, "epoch": 2358, "lr": 1.2182717431289032e-05}
{"train_loss": 0.04791800305247307, "global_step": 209949, "epoch": 2358, "lr": 1.2182338186266984e-05}
{"train_loss": 0.058913393427481815, "global_step": 209950, "epoch": 2358, "lr": 1.2181958946329052e-05, "val_loss": 8.499143600463867}
{"train_loss": 0.04633758217096329, "global_step": 209951, "epoch": 2359, "lr": 1.2181579711475305e-05}
{"train_loss": 0.052651286125183105, "global_step": 209952, "epoch": 2359, "lr": 1.2181200481705773e-05}
{"train_loss": 0.06877656280994415, "global_step": 209953, "epoch": 2359, "lr": 1.2180821257020531e-05}
{"train_loss": 0.030832257121801376, "global_step": 209954, "epoch": 2359, "lr": 1.2180442037419597e-05}
{"train_loss": 0.036367397755384445, "global_step": 209955, "epoch": 2359, "lr": 1.218006282290306e-05}
{"train_loss": 0.040889106690883636, "global_step": 209956, "epoch": 2359, "lr": 1.2179683613470939e-05}
{"train_loss": 0.06970297545194626, "global_step": 209957, "epoch": 2359, "lr": 1.2179304409123315e-05}
{"train_loss": 0.05464217811822891, "global_step": 209958, "epoch": 2359, "lr": 1.2178925209860203e-05}
{"train_loss": 0.04018028452992439, "global_step": 209959, "epoch": 2359, "lr": 1.2178546015681691e-05}
{"train_loss": 0.052817393094301224, "global_step": 209960, "epoch": 2359, "lr": 1.2178166826587795e-05}
{"train_loss": 0.05810709670186043, "global_step": 209961, "epoch": 2359, "lr": 1.2177787642578603e-05}
{"train_loss": 0.047167371958494186, "global_step": 209962, "epoch": 2359, "lr": 1.2177408463654128e-05}
{"train_loss": 0.06846856325864792, "global_step": 209963, "epoch": 2359, "lr": 1.217702928981444e-05}
{"train_loss": 0.08170340955257416, "global_step": 209964, "epoch": 2359, "lr": 1.2176650121059607e-05}
{"train_loss": 0.09097815304994583, "global_step": 209965, "epoch": 2359, "lr": 1.2176270957389657e-05}
{"train_loss": 0.043609242886304855, "global_step": 209966, "epoch": 2359, "lr": 1.217589179880464e-05}
{"train_loss": 0.05857923626899719, "global_step": 209967, "epoch": 2359, "lr": 1.217551264530462e-05}
{"train_loss": 0.05836661532521248, "global_step": 209968, "epoch": 2359, "lr": 1.2175133496889629e-05}
{"train_loss": 0.07585491240024567, "global_step": 209969, "epoch": 2359, "lr": 1.2174754353559748e-05}
{"train_loss": 0.06295312196016312, "global_step": 209970, "epoch": 2359, "lr": 1.2174375215314992e-05}
{"train_loss": 0.10581149160861969, "global_step": 209971, "epoch": 2359, "lr": 1.2173996082155448e-05}
{"train_loss": 0.13192327320575714, "global_step": 209972, "epoch": 2359, "lr": 1.2173616954081146e-05}
{"train_loss": 0.09410975873470306, "global_step": 209973, "epoch": 2359, "lr": 1.2173237831092126e-05}
{"train_loss": 0.0817224457859993, "global_step": 209974, "epoch": 2359, "lr": 1.2172858713188467e-05}
{"train_loss": 0.05255267396569252, "global_step": 209975, "epoch": 2359, "lr": 1.2172479600370195e-05}
{"train_loss": 0.03758135437965393, "global_step": 209976, "epoch": 2359, "lr": 1.2172100492637384e-05}
{"train_loss": 0.09814365953207016, "global_step": 209977, "epoch": 2359, "lr": 1.2171721389990059e-05}
{"train_loss": 0.07276501506567001, "global_step": 209978, "epoch": 2359, "lr": 1.2171342292428301e-05}
{"train_loss": 0.04163867607712746, "global_step": 209979, "epoch": 2359, "lr": 1.2170963199952129e-05}
{"train_loss": 0.027871545404195786, "global_step": 209980, "epoch": 2359, "lr": 1.2170584112561607e-05}
{"train_loss": 0.04787096753716469, "global_step": 209981, "epoch": 2359, "lr": 1.217020503025681e-05}
{"train_loss": 0.05013464391231537, "global_step": 209982, "epoch": 2359, "lr": 1.216982595303775e-05}
{"train_loss": 0.03578256070613861, "global_step": 209983, "epoch": 2359, "lr": 1.2169446880904506e-05}
{"train_loss": 0.03976687416434288, "global_step": 209984, "epoch": 2359, "lr": 1.216906781385711e-05}
{"train_loss": 0.07172050327062607, "global_step": 209985, "epoch": 2359, "lr": 1.2168688751895634e-05}
{"train_loss": 0.0574059784412384, "global_step": 209986, "epoch": 2359, "lr": 1.2168309695020103e-05}
{"train_loss": 0.0659872367978096, "global_step": 209987, "epoch": 2359, "lr": 1.2167930643230596e-05}
{"train_loss": 0.061246540397405624, "global_step": 209988, "epoch": 2359, "lr": 1.2167551596527132e-05}
{"train_loss": 0.0620950311422348, "global_step": 209989, "epoch": 2359, "lr": 1.2167172554909801e-05}
{"train_loss": 0.06101253628730774, "global_step": 209990, "epoch": 2359, "lr": 1.2166793518378612e-05}
{"train_loss": 0.052724651992321014, "global_step": 209991, "epoch": 2359, "lr": 1.2166414486933653e-05}
{"train_loss": 0.09535489231348038, "global_step": 209992, "epoch": 2359, "lr": 1.2166035460574943e-05}
{"train_loss": 0.05039434880018234, "global_step": 209993, "epoch": 2359, "lr": 1.2165656439302565e-05}
{"train_loss": 0.04431314021348953, "global_step": 209994, "epoch": 2359, "lr": 1.216527742311654e-05}
{"train_loss": 0.10225829482078552, "global_step": 209995, "epoch": 2359, "lr": 1.2164898412016945e-05}
{"train_loss": 0.025057697668671608, "global_step": 209996, "epoch": 2359, "lr": 1.2164519406003805e-05}
{"train_loss": 0.08430788666009903, "global_step": 209997, "epoch": 2359, "lr": 1.2164140405077195e-05}
{"train_loss": 0.0398431159555912, "global_step": 209998, "epoch": 2359, "lr": 1.216376140923714e-05}
{"train_loss": 0.06994762271642685, "global_step": 209999, "epoch": 2359, "lr": 1.2163382418483709e-05}
{"train_loss": 0.06964823603630066, "global_step": 210000, "epoch": 2359, "lr": 1.2163003432816967e-05}
{"train_loss": 0.03163266181945801, "global_step": 210001, "epoch": 2359, "lr": 1.2162624452236932e-05}
{"train_loss": 0.056421224027872086, "global_step": 210002, "epoch": 2359, "lr": 1.2162245476743683e-05}
{"train_loss": 0.032703906297683716, "global_step": 210003, "epoch": 2359, "lr": 1.2161866506337244e-05}
{"train_loss": 0.04571959376335144, "global_step": 210004, "epoch": 2359, "lr": 1.2161487541017697e-05}
{"train_loss": 0.0791759341955185, "global_step": 210005, "epoch": 2359, "lr": 1.216110858078508e-05}
{"train_loss": 0.09060943126678467, "global_step": 210006, "epoch": 2359, "lr": 1.2160729625639422e-05}
{"train_loss": 0.05577947199344635, "global_step": 210007, "epoch": 2359, "lr": 1.2160350675580812e-05}
{"train_loss": 0.039377644658088684, "global_step": 210008, "epoch": 2359, "lr": 1.2159971730609259e-05}
{"train_loss": 0.06706558167934418, "global_step": 210009, "epoch": 2359, "lr": 1.2159592790724856e-05}
{"train_loss": 0.10693816840648651, "global_step": 210010, "epoch": 2359, "lr": 1.2159213855927636e-05}
{"train_loss": 0.03255566209554672, "global_step": 210011, "epoch": 2359, "lr": 1.2158834926217633e-05}
{"train_loss": 0.06343825161457062, "global_step": 210012, "epoch": 2359, "lr": 1.2158456001594926e-05}
{"train_loss": 0.06517594307661057, "global_step": 210013, "epoch": 2359, "lr": 1.2158077082059537e-05}
{"train_loss": 0.05114100128412247, "global_step": 210014, "epoch": 2359, "lr": 1.2157698167611554e-05}
{"train_loss": 0.03968913480639458, "global_step": 210015, "epoch": 2359, "lr": 1.2157319258250988e-05}
{"train_loss": 0.08036963641643524, "global_step": 210016, "epoch": 2359, "lr": 1.2156940353977925e-05}
{"train_loss": 0.04940363019704819, "global_step": 210017, "epoch": 2359, "lr": 1.2156561454792387e-05}
{"train_loss": 0.11267200112342834, "global_step": 210018, "epoch": 2359, "lr": 1.215618256069444e-05}
{"train_loss": 0.06947061419487, "global_step": 210019, "epoch": 2359, "lr": 1.215580367168414e-05}
{"train_loss": 0.0674743726849556, "global_step": 210020, "epoch": 2359, "lr": 1.2155424787761522e-05}
{"train_loss": 0.059016890823841095, "global_step": 210021, "epoch": 2359, "lr": 1.215504590892666e-05}
{"train_loss": 0.036978770047426224, "global_step": 210022, "epoch": 2359, "lr": 1.215466703517958e-05}
{"train_loss": 0.052737146615982056, "global_step": 210023, "epoch": 2359, "lr": 1.215428816652035e-05}
{"train_loss": 0.03535761684179306, "global_step": 210024, "epoch": 2359, "lr": 1.2153909302949006e-05}
{"train_loss": 0.04142230749130249, "global_step": 210025, "epoch": 2359, "lr": 1.2153530444465616e-05}
{"train_loss": 0.026871932670474052, "global_step": 210026, "epoch": 2359, "lr": 1.2153151591070216e-05}
{"train_loss": 0.07442111521959305, "global_step": 210027, "epoch": 2359, "lr": 1.2152772742762875e-05}
{"train_loss": 0.05005325376987457, "global_step": 210028, "epoch": 2359, "lr": 1.2152393899543613e-05}
{"train_loss": 0.04116823524236679, "global_step": 210029, "epoch": 2359, "lr": 1.2152015061412519e-05}
{"train_loss": 0.06860454380512238, "global_step": 210030, "epoch": 2359, "lr": 1.2151636228369613e-05}
{"train_loss": 0.07341504842042923, "global_step": 210031, "epoch": 2359, "lr": 1.2151257400414972e-05}
{"train_loss": 0.0441962406039238, "global_step": 210032, "epoch": 2359, "lr": 1.2150878577548613e-05}
{"train_loss": 0.09835175424814224, "global_step": 210033, "epoch": 2359, "lr": 1.2150499759770628e-05}
{"train_loss": 0.015822743996977806, "global_step": 210034, "epoch": 2359, "lr": 1.2150120947081029e-05}
{"train_loss": 0.050646282732486725, "global_step": 210035, "epoch": 2359, "lr": 1.2149742139479885e-05}
{"train_loss": 0.0904829278588295, "global_step": 210036, "epoch": 2359, "lr": 1.2149363336967262e-05}
{"train_loss": 0.11602097749710083, "global_step": 210037, "epoch": 2359, "lr": 1.2148984539543185e-05}
{"train_loss": 0.028245992958545685, "global_step": 210038, "epoch": 2359, "lr": 1.2148605747207725e-05}
{"train_loss": 0.05991288718212856, "global_step": 210039, "epoch": 2359, "lr": 1.2148226959960912e-05, "val_loss": 8.396774291992188}
{"train_loss": 0.05519500747323036, "global_step": 210040, "epoch": 2360, "lr": 1.2147848177802823e-05}
{"train_loss": 0.08154332637786865, "global_step": 210041, "epoch": 2360, "lr": 1.214746940073348e-05}
{"train_loss": 0.09142860770225525, "global_step": 210042, "epoch": 2360, "lr": 1.214709062875296e-05}
{"train_loss": 0.062307246029376984, "global_step": 210043, "epoch": 2360, "lr": 1.2146711861861304e-05}
{"train_loss": 0.04910712689161301, "global_step": 210044, "epoch": 2360, "lr": 1.214633310005855e-05}
{"train_loss": 0.07323774695396423, "global_step": 210045, "epoch": 2360, "lr": 1.2145954343344767e-05}
{"train_loss": 0.04933200776576996, "global_step": 210046, "epoch": 2360, "lr": 1.2145575591719988e-05}
{"train_loss": 0.053926028311252594, "global_step": 210047, "epoch": 2360, "lr": 1.2145196845184292e-05}
{"train_loss": 0.036253899335861206, "global_step": 210048, "epoch": 2360, "lr": 1.214481810373771e-05}
{"train_loss": 0.05705953761935234, "global_step": 210049, "epoch": 2360, "lr": 1.2144439367380283e-05}
{"train_loss": 0.07447151094675064, "global_step": 210050, "epoch": 2360, "lr": 1.2144060636112092e-05}
{"train_loss": 0.06371729075908661, "global_step": 210051, "epoch": 2360, "lr": 1.2143681909933151e-05}
{"train_loss": 0.0822795033454895, "global_step": 210052, "epoch": 2360, "lr": 1.2143303188843547e-05}
{"train_loss": 0.025519294664263725, "global_step": 210053, "epoch": 2360, "lr": 1.2142924472843303e-05}
{"train_loss": 0.07655755430459976, "global_step": 210054, "epoch": 2360, "lr": 1.2142545761932478e-05}
{"train_loss": 0.07956419885158539, "global_step": 210055, "epoch": 2360, "lr": 1.2142167056111142e-05}
{"train_loss": 0.054962754249572754, "global_step": 210056, "epoch": 2360, "lr": 1.2141788355379314e-05}
{"train_loss": 0.06754493713378906, "global_step": 210057, "epoch": 2360, "lr": 1.2141409659737074e-05}
{"train_loss": 0.06296539306640625, "global_step": 210058, "epoch": 2360, "lr": 1.2141030969184453e-05}
{"train_loss": 0.046845365315675735, "global_step": 210059, "epoch": 2360, "lr": 1.2140652283721515e-05}
{"train_loss": 0.05824227258563042, "global_step": 210060, "epoch": 2360, "lr": 1.2140273603348296e-05}
{"train_loss": 0.09525059163570404, "global_step": 210061, "epoch": 2360, "lr": 1.2139894928064865e-05}
{"train_loss": 0.05413211137056351, "global_step": 210062, "epoch": 2360, "lr": 1.2139516257871253e-05}
{"train_loss": 0.03634008392691612, "global_step": 210063, "epoch": 2360, "lr": 1.2139137592767536e-05}
{"train_loss": 0.05448111891746521, "global_step": 210064, "epoch": 2360, "lr": 1.2138758932753735e-05}
{"train_loss": 0.04906725883483887, "global_step": 210065, "epoch": 2360, "lr": 1.2138380277829936e-05}
{"train_loss": 0.05273739993572235, "global_step": 210066, "epoch": 2360, "lr": 1.2138001627996149e-05}
{"train_loss": 0.07868514209985733, "global_step": 210067, "epoch": 2360, "lr": 1.2137622983252461e-05}
{"train_loss": 0.10619011521339417, "global_step": 210068, "epoch": 2360, "lr": 1.2137244343598896e-05}
{"train_loss": 0.04103904962539673, "global_step": 210069, "epoch": 2360, "lr": 1.2136865709035534e-05}
{"train_loss": 0.05007229372859001, "global_step": 210070, "epoch": 2360, "lr": 1.2136487079562392e-05}
{"train_loss": 0.043642617762088776, "global_step": 210071, "epoch": 2360, "lr": 1.2136108455179535e-05}
{"train_loss": 0.06321427971124649, "global_step": 210072, "epoch": 2360, "lr": 1.2135729835887033e-05}
{"train_loss": 0.04658370465040207, "global_step": 210073, "epoch": 2360, "lr": 1.2135351221684909e-05}
{"train_loss": 0.05596079304814339, "global_step": 210074, "epoch": 2360, "lr": 1.2134972612573236e-05}
{"train_loss": 0.048910416662693024, "global_step": 210075, "epoch": 2360, "lr": 1.2134594008552042e-05}
{"train_loss": 0.03191505745053291, "global_step": 210076, "epoch": 2360, "lr": 1.2134215409621402e-05}
{"train_loss": 0.03465826436877251, "global_step": 210077, "epoch": 2360, "lr": 1.2133836815781347e-05}
{"train_loss": 0.09011916816234589, "global_step": 210078, "epoch": 2360, "lr": 1.2133458227031947e-05}
{"train_loss": 0.0742441788315773, "global_step": 210079, "epoch": 2360, "lr": 1.2133079643373225e-05}
{"train_loss": 0.03039640374481678, "global_step": 210080, "epoch": 2360, "lr": 1.2132701064805269e-05}
{"train_loss": 0.06726202368736267, "global_step": 210081, "epoch": 2360, "lr": 1.2132322491328103e-05}
{"train_loss": 0.05661090835928917, "global_step": 210082, "epoch": 2360, "lr": 1.2131943922941774e-05}
{"train_loss": 0.05574243515729904, "global_step": 210083, "epoch": 2360, "lr": 1.2131565359646358e-05}
{"train_loss": 0.03950851783156395, "global_step": 210084, "epoch": 2360, "lr": 1.213118680144188e-05}
{"train_loss": 0.05997411906719208, "global_step": 210085, "epoch": 2360, "lr": 1.2130808248328413e-05}
{"train_loss": 0.07951464504003525, "global_step": 210086, "epoch": 2360, "lr": 1.2130429700305984e-05}
{"train_loss": 0.039500895887613297, "global_step": 210087, "epoch": 2360, "lr": 1.2130051157374666e-05}
{"train_loss": 0.039729904383420944, "global_step": 210088, "epoch": 2360, "lr": 1.212967261953451e-05}
{"train_loss": 0.06056984141469002, "global_step": 210089, "epoch": 2360, "lr": 1.2129294086785537e-05}
{"train_loss": 0.05547897517681122, "global_step": 210090, "epoch": 2360, "lr": 1.2128915559127823e-05}
{"train_loss": 0.03651493042707443, "global_step": 210091, "epoch": 2360, "lr": 1.2128537036561427e-05}
{"train_loss": 0.03334367647767067, "global_step": 210092, "epoch": 2360, "lr": 1.2128158519086375e-05}
{"train_loss": 0.10335210710763931, "global_step": 210093, "epoch": 2360, "lr": 1.2127780006702744e-05}
{"train_loss": 0.0738634318113327, "global_step": 210094, "epoch": 2360, "lr": 1.2127401499410557e-05}
{"train_loss": 0.029659923166036606, "global_step": 210095, "epoch": 2360, "lr": 1.2127022997209891e-05}
{"train_loss": 0.10484666377305984, "global_step": 210096, "epoch": 2360, "lr": 1.2126644500100775e-05}
{"train_loss": 0.05298079550266266, "global_step": 210097, "epoch": 2360, "lr": 1.2126266008083281e-05}
{"train_loss": 0.057981979101896286, "global_step": 210098, "epoch": 2360, "lr": 1.2125887521157436e-05}
{"train_loss": 0.05775878578424454, "global_step": 210099, "epoch": 2360, "lr": 1.2125509039323319e-05}
{"train_loss": 0.015998942777514458, "global_step": 210100, "epoch": 2360, "lr": 1.212513056258095e-05}
{"train_loss": 0.048548102378845215, "global_step": 210101, "epoch": 2360, "lr": 1.2124752090930413e-05}
{"train_loss": 0.06817612797021866, "global_step": 210102, "epoch": 2360, "lr": 1.212437362437172e-05}
{"train_loss": 0.0577348992228508, "global_step": 210103, "epoch": 2360, "lr": 1.2123995162904967e-05}
{"train_loss": 0.03750147297978401, "global_step": 210104, "epoch": 2360, "lr": 1.212361670653016e-05}
{"train_loss": 0.02293424867093563, "global_step": 210105, "epoch": 2360, "lr": 1.2123238255247388e-05}
{"train_loss": 0.05623861029744148, "global_step": 210106, "epoch": 2360, "lr": 1.2122859809056669e-05}
{"train_loss": 0.06554772704839706, "global_step": 210107, "epoch": 2360, "lr": 1.2122481367958072e-05}
{"train_loss": 0.09167788177728653, "global_step": 210108, "epoch": 2360, "lr": 1.2122102931951657e-05}
{"train_loss": 0.04941285029053688, "global_step": 210109, "epoch": 2360, "lr": 1.2121724501037451e-05}
{"train_loss": 0.05436556786298752, "global_step": 210110, "epoch": 2360, "lr": 1.2121346075215534e-05}
{"train_loss": 0.05915414169430733, "global_step": 210111, "epoch": 2360, "lr": 1.2120967654485927e-05}
{"train_loss": 0.10392508655786514, "global_step": 210112, "epoch": 2360, "lr": 1.21205892388487e-05}
{"train_loss": 0.057900428771972656, "global_step": 210113, "epoch": 2360, "lr": 1.212021082830389e-05}
{"train_loss": 0.07455383241176605, "global_step": 210114, "epoch": 2360, "lr": 1.2119832422851574e-05}
{"train_loss": 0.062340445816516876, "global_step": 210115, "epoch": 2360, "lr": 1.2119454022491766e-05}
{"train_loss": 0.07348783314228058, "global_step": 210116, "epoch": 2360, "lr": 1.2119075627224547e-05}
{"train_loss": 0.07421635091304779, "global_step": 210117, "epoch": 2360, "lr": 1.2118697237049948e-05}
{"train_loss": 0.07974927127361298, "global_step": 210118, "epoch": 2360, "lr": 1.211831885196804e-05}
{"train_loss": 0.04877246916294098, "global_step": 210119, "epoch": 2360, "lr": 1.211794047197885e-05}
{"train_loss": 0.07017534971237183, "global_step": 210120, "epoch": 2360, "lr": 1.2117562097082453e-05}
{"train_loss": 0.11821122467517853, "global_step": 210121, "epoch": 2360, "lr": 1.2117183727278885e-05}
{"train_loss": 0.05177842080593109, "global_step": 210122, "epoch": 2360, "lr": 1.211680536256819e-05}
{"train_loss": 0.09135637432336807, "global_step": 210123, "epoch": 2360, "lr": 1.2116427002950443e-05}
{"train_loss": 0.04061760753393173, "global_step": 210124, "epoch": 2360, "lr": 1.2116048648425665e-05}
{"train_loss": 0.10398276150226593, "global_step": 210125, "epoch": 2360, "lr": 1.2115670298993943e-05}
{"train_loss": 0.05234995484352112, "global_step": 210126, "epoch": 2360, "lr": 1.2115291954655284e-05}
{"train_loss": 0.014655202627182007, "global_step": 210127, "epoch": 2360, "lr": 1.2114913615409784e-05}
{"train_loss": 0.06082884691069635, "global_step": 210128, "epoch": 2360, "lr": 1.2114535281257449e-05, "val_loss": 8.604632377624512, "train_action_mse_error": 16.8026065826416}
{"train_loss": 0.03805595263838768, "global_step": 210129, "epoch": 2361, "lr": 1.2114156952198374e-05}
{"train_loss": 0.0231581162661314, "global_step": 210130, "epoch": 2361, "lr": 1.2113778628232576e-05}
{"train_loss": 0.07899127155542374, "global_step": 210131, "epoch": 2361, "lr": 1.2113400309360129e-05}
{"train_loss": 0.06381135433912277, "global_step": 210132, "epoch": 2361, "lr": 1.2113021995581059e-05}
{"train_loss": 0.08275416493415833, "global_step": 210133, "epoch": 2361, "lr": 1.2112643686895447e-05}
{"train_loss": 0.043128807097673416, "global_step": 210134, "epoch": 2361, "lr": 1.2112265383303312e-05}
{"train_loss": 0.02547428570687771, "global_step": 210135, "epoch": 2361, "lr": 1.211188708480474e-05}
{"train_loss": 0.08719026297330856, "global_step": 210136, "epoch": 2361, "lr": 1.2111508791399745e-05}
{"train_loss": 0.03912901505827904, "global_step": 210137, "epoch": 2361, "lr": 1.2111130503088414e-05}
{"train_loss": 0.08156375586986542, "global_step": 210138, "epoch": 2361, "lr": 1.2110752219870757e-05}
{"train_loss": 0.039845727384090424, "global_step": 210139, "epoch": 2361, "lr": 1.2110373941746872e-05}
{"train_loss": 0.054675422608852386, "global_step": 210140, "epoch": 2361, "lr": 1.2109995668716768e-05}
{"train_loss": 0.05744965746998787, "global_step": 210141, "epoch": 2361, "lr": 1.2109617400780527e-05}
{"train_loss": 0.048259854316711426, "global_step": 210142, "epoch": 2361, "lr": 1.2109239137938166e-05}
{"train_loss": 0.12810048460960388, "global_step": 210143, "epoch": 2361, "lr": 1.2108860880189777e-05}
{"train_loss": 0.05132792890071869, "global_step": 210144, "epoch": 2361, "lr": 1.2108482627535372e-05}
{"train_loss": 0.04888016730546951, "global_step": 210145, "epoch": 2361, "lr": 1.2108104379975022e-05}
{"train_loss": 0.037676066160202026, "global_step": 210146, "epoch": 2361, "lr": 1.2107726137508791e-05}
{"train_loss": 0.04697949439287186, "global_step": 210147, "epoch": 2361, "lr": 1.2107347900136696e-05}
{"train_loss": 0.029081124812364578, "global_step": 210148, "epoch": 2361, "lr": 1.2106969667858826e-05}
{"train_loss": 0.07639805227518082, "global_step": 210149, "epoch": 2361, "lr": 1.2106591440675192e-05}
{"train_loss": 0.033951032906770706, "global_step": 210150, "epoch": 2361, "lr": 1.2106213218585883e-05}
{"train_loss": 0.11176182329654694, "global_step": 210151, "epoch": 2361, "lr": 1.2105835001590915e-05}
{"train_loss": 0.04197703301906586, "global_step": 210152, "epoch": 2361, "lr": 1.2105456789690373e-05}
{"train_loss": 0.05431226268410683, "global_step": 210153, "epoch": 2361, "lr": 1.2105078582884271e-05}
{"train_loss": 0.06291568279266357, "global_step": 210154, "epoch": 2361, "lr": 1.2104700381172701e-05}
{"train_loss": 0.041885677725076675, "global_step": 210155, "epoch": 2361, "lr": 1.2104322184555673e-05}
{"train_loss": 0.010621430352330208, "global_step": 210156, "epoch": 2361, "lr": 1.2103943993033272e-05}
{"train_loss": 0.07275107502937317, "global_step": 210157, "epoch": 2361, "lr": 1.2103565806605521e-05}
{"train_loss": 0.053686365485191345, "global_step": 210158, "epoch": 2361, "lr": 1.2103187625272494e-05}
{"train_loss": 0.042282044887542725, "global_step": 210159, "epoch": 2361, "lr": 1.2102809449034236e-05}
{"train_loss": 0.09236189723014832, "global_step": 210160, "epoch": 2361, "lr": 1.210243127789078e-05}
{"train_loss": 0.04455633461475372, "global_step": 210161, "epoch": 2361, "lr": 1.2102053111842198e-05}
{"train_loss": 0.05592246353626251, "global_step": 210162, "epoch": 2361, "lr": 1.2101674950888525e-05}
{"train_loss": 0.06715826690196991, "global_step": 210163, "epoch": 2361, "lr": 1.2101296795029832e-05}
{"train_loss": 0.06375785171985626, "global_step": 210164, "epoch": 2361, "lr": 1.210091864426614e-05}
{"train_loss": 0.061221614480018616, "global_step": 210165, "epoch": 2361, "lr": 1.2100540498597535e-05}
{"train_loss": 0.08713093400001526, "global_step": 210166, "epoch": 2361, "lr": 1.2100162358024036e-05}
{"train_loss": 0.06123529002070427, "global_step": 210167, "epoch": 2361, "lr": 1.2099784222545723e-05}
{"train_loss": 0.023660864681005478, "global_step": 210168, "epoch": 2361, "lr": 1.2099406092162619e-05}
{"train_loss": 0.07018473744392395, "global_step": 210169, "epoch": 2361, "lr": 1.2099027966874798e-05}
{"train_loss": 0.04109346494078636, "global_step": 210170, "epoch": 2361, "lr": 1.2098649846682291e-05}
{"train_loss": 0.04943918436765671, "global_step": 210171, "epoch": 2361, "lr": 1.209827173158517e-05}
{"train_loss": 0.04548345506191254, "global_step": 210172, "epoch": 2361, "lr": 1.2097893621583462e-05}
{"train_loss": 0.0653042271733284, "global_step": 210173, "epoch": 2361, "lr": 1.2097515516677243e-05}
{"train_loss": 0.07110420614480972, "global_step": 210174, "epoch": 2361, "lr": 1.209713741686654e-05}
{"train_loss": 0.03983103856444359, "global_step": 210175, "epoch": 2361, "lr": 1.2096759322151424e-05}
{"train_loss": 0.024943655356764793, "global_step": 210176, "epoch": 2361, "lr": 1.2096381232531928e-05}
{"train_loss": 0.05793963745236397, "global_step": 210177, "epoch": 2361, "lr": 1.2096003148008122e-05}
{"train_loss": 0.07553916424512863, "global_step": 210178, "epoch": 2361, "lr": 1.209562506858003e-05}
{"train_loss": 0.10912802815437317, "global_step": 210179, "epoch": 2361, "lr": 1.209524699424774e-05}
{"train_loss": 0.11765486001968384, "global_step": 210180, "epoch": 2361, "lr": 1.2094868925011265e-05}
{"train_loss": 0.047388963401317596, "global_step": 210181, "epoch": 2361, "lr": 1.2094490860870672e-05}
{"train_loss": 0.0401618629693985, "global_step": 210182, "epoch": 2361, "lr": 1.2094112801826024e-05}
{"train_loss": 0.06676285713911057, "global_step": 210183, "epoch": 2361, "lr": 1.209373474787735e-05}
{"train_loss": 0.07584945857524872, "global_step": 210184, "epoch": 2361, "lr": 1.2093356699024727e-05}
{"train_loss": 0.11229139566421509, "global_step": 210185, "epoch": 2361, "lr": 1.2092978655268172e-05}
{"train_loss": 0.024125386029481888, "global_step": 210186, "epoch": 2361, "lr": 1.209260061660777e-05}
{"train_loss": 0.06111661344766617, "global_step": 210187, "epoch": 2361, "lr": 1.2092222583043545e-05}
{"train_loss": 0.02904510498046875, "global_step": 210188, "epoch": 2361, "lr": 1.2091844554575566e-05}
{"train_loss": 0.037026721984148026, "global_step": 210189, "epoch": 2361, "lr": 1.2091466531203865e-05}
{"train_loss": 0.051803845912218094, "global_step": 210190, "epoch": 2361, "lr": 1.2091088512928522e-05}
{"train_loss": 0.04179499298334122, "global_step": 210191, "epoch": 2361, "lr": 1.2090710499749552e-05}
{"train_loss": 0.10416224598884583, "global_step": 210192, "epoch": 2361, "lr": 1.209033249166704e-05}
{"train_loss": 0.05053817108273506, "global_step": 210193, "epoch": 2361, "lr": 1.2089954488681004e-05}
{"train_loss": 0.05745553970336914, "global_step": 210194, "epoch": 2361, "lr": 1.2089576490791531e-05}
{"train_loss": 0.030259503051638603, "global_step": 210195, "epoch": 2361, "lr": 1.2089198497998633e-05}
{"train_loss": 0.03359927609562874, "global_step": 210196, "epoch": 2361, "lr": 1.2088820510302395e-05}
{"train_loss": 0.06932487338781357, "global_step": 210197, "epoch": 2361, "lr": 1.2088442527702842e-05}
{"train_loss": 0.05839530751109123, "global_step": 210198, "epoch": 2361, "lr": 1.2088064550200046e-05}
{"train_loss": 0.15140074491500854, "global_step": 210199, "epoch": 2361, "lr": 1.208768657779405e-05}
{"train_loss": 0.05621764808893204, "global_step": 210200, "epoch": 2361, "lr": 1.2087308610484887e-05}
{"train_loss": 0.09069669246673584, "global_step": 210201, "epoch": 2361, "lr": 1.2086930648272643e-05}
{"train_loss": 0.05356407165527344, "global_step": 210202, "epoch": 2361, "lr": 1.2086552691157327e-05}
{"train_loss": 0.032445404678583145, "global_step": 210203, "epoch": 2361, "lr": 1.2086174739139034e-05}
{"train_loss": 0.054924171417951584, "global_step": 210204, "epoch": 2361, "lr": 1.2085796792217769e-05}
{"train_loss": 0.04915865138173103, "global_step": 210205, "epoch": 2361, "lr": 1.2085418850393625e-05}
{"train_loss": 0.034469012171030045, "global_step": 210206, "epoch": 2361, "lr": 1.2085040913666623e-05}
{"train_loss": 0.07435163110494614, "global_step": 210207, "epoch": 2361, "lr": 1.2084662982036837e-05}
{"train_loss": 0.06609917432069778, "global_step": 210208, "epoch": 2361, "lr": 1.2084285055504291e-05}
{"train_loss": 0.03310711681842804, "global_step": 210209, "epoch": 2361, "lr": 1.2083907134069067e-05}
{"train_loss": 0.04415685683488846, "global_step": 210210, "epoch": 2361, "lr": 1.2083529217731183e-05}
{"train_loss": 0.05928774178028107, "global_step": 210211, "epoch": 2361, "lr": 1.2083151306490726e-05}
{"train_loss": 0.024109724909067154, "global_step": 210212, "epoch": 2361, "lr": 1.2082773400347709e-05}
{"train_loss": 0.01784617267549038, "global_step": 210213, "epoch": 2361, "lr": 1.2082395499302213e-05}
{"train_loss": 0.03282764181494713, "global_step": 210214, "epoch": 2361, "lr": 1.2082017603354262e-05}
{"train_loss": 0.04648834466934204, "global_step": 210215, "epoch": 2361, "lr": 1.2081639712503939e-05}
{"train_loss": 0.07091614603996277, "global_step": 210216, "epoch": 2361, "lr": 1.208126182675126e-05}
{"train_loss": 0.05690870145231151, "global_step": 210217, "epoch": 2361, "lr": 1.20808839460963e-05, "val_loss": 8.69342041015625}
{"train_loss": 0.04473720118403435, "global_step": 210218, "epoch": 2362, "lr": 1.208050607053911e-05}
{"train_loss": 0.04375101625919342, "global_step": 210219, "epoch": 2362, "lr": 1.208012820007972e-05}
{"train_loss": 0.06785869598388672, "global_step": 210220, "epoch": 2362, "lr": 1.2079750334718216e-05}
{"train_loss": 0.07418345659971237, "global_step": 210221, "epoch": 2362, "lr": 1.2079372474454604e-05}
{"train_loss": 0.06475245952606201, "global_step": 210222, "epoch": 2362, "lr": 1.207899461928898e-05}
{"train_loss": 0.0525948591530323, "global_step": 210223, "epoch": 2362, "lr": 1.2078616769221352e-05}
{"train_loss": 0.1140744760632515, "global_step": 210224, "epoch": 2362, "lr": 1.2078238924251811e-05}
{"train_loss": 0.07375260442495346, "global_step": 210225, "epoch": 2362, "lr": 1.207786108438037e-05}
{"train_loss": 0.04333973675966263, "global_step": 210226, "epoch": 2362, "lr": 1.2077483249607119e-05}
{"train_loss": 0.046140141785144806, "global_step": 210227, "epoch": 2362, "lr": 1.2077105419932073e-05}
{"train_loss": 0.057171352207660675, "global_step": 210228, "epoch": 2362, "lr": 1.2076727595355308e-05}
{"train_loss": 0.07869117707014084, "global_step": 210229, "epoch": 2362, "lr": 1.2076349775876849e-05}
{"train_loss": 0.030187826603651047, "global_step": 210230, "epoch": 2362, "lr": 1.207597196149678e-05}
{"train_loss": 0.013971894979476929, "global_step": 210231, "epoch": 2362, "lr": 1.207559415221512e-05}
{"train_loss": 0.04984400048851967, "global_step": 210232, "epoch": 2362, "lr": 1.2075216348031948e-05}
{"train_loss": 0.09092046320438385, "global_step": 210233, "epoch": 2362, "lr": 1.2074838548947287e-05}
{"train_loss": 0.08987286686897278, "global_step": 210234, "epoch": 2362, "lr": 1.2074460754961215e-05}
{"train_loss": 0.0486474484205246, "global_step": 210235, "epoch": 2362, "lr": 1.2074082966073757e-05}
{"train_loss": 0.06350068747997284, "global_step": 210236, "epoch": 2362, "lr": 1.2073705182284988e-05}
{"train_loss": 0.024310747161507607, "global_step": 210237, "epoch": 2362, "lr": 1.2073327403594952e-05}
{"train_loss": 0.08674795925617218, "global_step": 210238, "epoch": 2362, "lr": 1.2072949630003677e-05}
{"train_loss": 0.05182211101055145, "global_step": 210239, "epoch": 2362, "lr": 1.2072571861511245e-05}
{"train_loss": 0.045408789068460464, "global_step": 210240, "epoch": 2362, "lr": 1.2072194098117684e-05}
{"train_loss": 0.02669946290552616, "global_step": 210241, "epoch": 2362, "lr": 1.2071816339823061e-05}
{"train_loss": 0.06425975263118744, "global_step": 210242, "epoch": 2362, "lr": 1.207143858662741e-05}
{"train_loss": 0.02481275051832199, "global_step": 210243, "epoch": 2362, "lr": 1.2071060838530806e-05}
{"train_loss": 0.13088877499103546, "global_step": 210244, "epoch": 2362, "lr": 1.2070683095533275e-05}
{"train_loss": 0.0698549747467041, "global_step": 210245, "epoch": 2362, "lr": 1.2070305357634887e-05}
{"train_loss": 0.06353013217449188, "global_step": 210246, "epoch": 2362, "lr": 1.206992762483567e-05}
{"train_loss": 0.07328041642904282, "global_step": 210247, "epoch": 2362, "lr": 1.2069549897135708e-05}
{"train_loss": 0.049412742257118225, "global_step": 210248, "epoch": 2362, "lr": 1.2069172174535014e-05}
{"train_loss": 0.049065761268138885, "global_step": 210249, "epoch": 2362, "lr": 1.2068794457033671e-05}
{"train_loss": 0.05454133078455925, "global_step": 210250, "epoch": 2362, "lr": 1.2068416744631705e-05}
{"train_loss": 0.051248449832201004, "global_step": 210251, "epoch": 2362, "lr": 1.2068039037329193e-05}
{"train_loss": 0.034024927765131, "global_step": 210252, "epoch": 2362, "lr": 1.2067661335126151e-05}
{"train_loss": 0.0505484938621521, "global_step": 210253, "epoch": 2362, "lr": 1.2067283638022659e-05}
{"train_loss": 0.048219919204711914, "global_step": 210254, "epoch": 2362, "lr": 1.2066905946018764e-05}
{"train_loss": 0.09516584873199463, "global_step": 210255, "epoch": 2362, "lr": 1.20665282591145e-05}
{"train_loss": 0.05486416444182396, "global_step": 210256, "epoch": 2362, "lr": 1.2066150577309943e-05}
{"train_loss": 0.11444002389907837, "global_step": 210257, "epoch": 2362, "lr": 1.2065772900605115e-05}
{"train_loss": 0.06500961631536484, "global_step": 210258, "epoch": 2362, "lr": 1.2065395229000098e-05}
{"train_loss": 0.05908069759607315, "global_step": 210259, "epoch": 2362, "lr": 1.2065017562494913e-05}
{"train_loss": 0.080291248857975, "global_step": 210260, "epoch": 2362, "lr": 1.206463990108963e-05}
{"train_loss": 0.06512170284986496, "global_step": 210261, "epoch": 2362, "lr": 1.2064262244784285e-05}
{"train_loss": 0.08673785626888275, "global_step": 210262, "epoch": 2362, "lr": 1.2063884593578956e-05}
{"train_loss": 0.03873695433139801, "global_step": 210263, "epoch": 2362, "lr": 1.2063506947473657e-05}
{"train_loss": 0.04133472964167595, "global_step": 210264, "epoch": 2362, "lr": 1.2063129306468468e-05}
{"train_loss": 0.04033837467432022, "global_step": 210265, "epoch": 2362, "lr": 1.2062751670563421e-05}
{"train_loss": 0.08079931885004044, "global_step": 210266, "epoch": 2362, "lr": 1.2062374039758584e-05}
{"train_loss": 0.05890268832445145, "global_step": 210267, "epoch": 2362, "lr": 1.2061996414053989e-05}
{"train_loss": 0.08458387851715088, "global_step": 210268, "epoch": 2362, "lr": 1.2061618793449708e-05}
{"train_loss": 0.06919629126787186, "global_step": 210269, "epoch": 2362, "lr": 1.2061241177945765e-05}
{"train_loss": 0.07394155859947205, "global_step": 210270, "epoch": 2362, "lr": 1.2060863567542241e-05}
{"train_loss": 0.057383984327316284, "global_step": 210271, "epoch": 2362, "lr": 1.2060485962239153e-05}
{"train_loss": 0.057674720883369446, "global_step": 210272, "epoch": 2362, "lr": 1.2060108362036588e-05}
{"train_loss": 0.04943528398871422, "global_step": 210273, "epoch": 2362, "lr": 1.2059730766934569e-05}
{"train_loss": 0.06151275336742401, "global_step": 210274, "epoch": 2362, "lr": 1.2059353176933163e-05}
{"train_loss": 0.06279463320970535, "global_step": 210275, "epoch": 2362, "lr": 1.2058975592032417e-05}
{"train_loss": 0.04270021244883537, "global_step": 210276, "epoch": 2362, "lr": 1.205859801223237e-05}
{"train_loss": 0.06444519758224487, "global_step": 210277, "epoch": 2362, "lr": 1.205822043753309e-05}
{"train_loss": 0.04538824409246445, "global_step": 210278, "epoch": 2362, "lr": 1.205784286793461e-05}
{"train_loss": 0.05047282949090004, "global_step": 210279, "epoch": 2362, "lr": 1.2057465303437004e-05}
{"train_loss": 0.030207674950361252, "global_step": 210280, "epoch": 2362, "lr": 1.2057087744040291e-05}
{"train_loss": 0.06177244335412979, "global_step": 210281, "epoch": 2362, "lr": 1.2056710189744559e-05}
{"train_loss": 0.04165961220860481, "global_step": 210282, "epoch": 2362, "lr": 1.2056332640549823e-05}
{"train_loss": 0.04591449350118637, "global_step": 210283, "epoch": 2362, "lr": 1.2055955096456168e-05}
{"train_loss": 0.07671412825584412, "global_step": 210284, "epoch": 2362, "lr": 1.205557755746361e-05}
{"train_loss": 0.026146842166781425, "global_step": 210285, "epoch": 2362, "lr": 1.2055200023572233e-05}
{"train_loss": 0.02562956139445305, "global_step": 210286, "epoch": 2362, "lr": 1.2054822494782059e-05}
{"train_loss": 0.05479845404624939, "global_step": 210287, "epoch": 2362, "lr": 1.2054444971093165e-05}
{"train_loss": 0.05754883959889412, "global_step": 210288, "epoch": 2362, "lr": 1.2054067452505568e-05}
{"train_loss": 0.0672236755490303, "global_step": 210289, "epoch": 2362, "lr": 1.2053689939019347e-05}
{"train_loss": 0.07160244137048721, "global_step": 210290, "epoch": 2362, "lr": 1.2053312430634555e-05}
{"train_loss": 0.04624159261584282, "global_step": 210291, "epoch": 2362, "lr": 1.2052934927351223e-05}
{"train_loss": 0.09439801424741745, "global_step": 210292, "epoch": 2362, "lr": 1.205255742916942e-05}
{"train_loss": 0.11196812242269516, "global_step": 210293, "epoch": 2362, "lr": 1.2052179936089175e-05}
{"train_loss": 0.0354941189289093, "global_step": 210294, "epoch": 2362, "lr": 1.2051802448110566e-05}
{"train_loss": 0.05484601482748985, "global_step": 210295, "epoch": 2362, "lr": 1.205142496523362e-05}
{"train_loss": 0.039445847272872925, "global_step": 210296, "epoch": 2362, "lr": 1.2051047487458405e-05}
{"train_loss": 0.061037205159664154, "global_step": 210297, "epoch": 2362, "lr": 1.2050670014784954e-05}
{"train_loss": 0.0770426020026207, "global_step": 210298, "epoch": 2362, "lr": 1.2050292547213344e-05}
{"train_loss": 0.05054914951324463, "global_step": 210299, "epoch": 2362, "lr": 1.204991508474359e-05}
{"train_loss": 0.08649399876594543, "global_step": 210300, "epoch": 2362, "lr": 1.2049537627375784e-05}
{"train_loss": 0.06597551703453064, "global_step": 210301, "epoch": 2362, "lr": 1.2049160175109935e-05}
{"train_loss": 0.025912443175911903, "global_step": 210302, "epoch": 2362, "lr": 1.2048782727946128e-05}
{"train_loss": 0.042051270604133606, "global_step": 210303, "epoch": 2362, "lr": 1.204840528588439e-05}
{"train_loss": 0.0765698179602623, "global_step": 210304, "epoch": 2362, "lr": 1.2048027848924793e-05}
{"train_loss": 0.025538360700011253, "global_step": 210305, "epoch": 2362, "lr": 1.2047650417067358e-05}
{"train_loss": 0.05904141853280952, "global_step": 210306, "epoch": 2362, "lr": 1.204727299031217e-05, "val_loss": 8.689847946166992}
{"train_loss": 0.038339484483003616, "global_step": 210307, "epoch": 2363, "lr": 1.204689556865925e-05}
{"train_loss": 0.08935368061065674, "global_step": 210308, "epoch": 2363, "lr": 1.2046518152108677e-05}
{"train_loss": 0.04014737531542778, "global_step": 210309, "epoch": 2363, "lr": 1.2046140740660477e-05}
{"train_loss": 0.08994422107934952, "global_step": 210310, "epoch": 2363, "lr": 1.2045763334314719e-05}
{"train_loss": 0.06534646451473236, "global_step": 210311, "epoch": 2363, "lr": 1.2045385933071434e-05}
{"train_loss": 0.06497427821159363, "global_step": 210312, "epoch": 2363, "lr": 1.2045008536930696e-05}
{"train_loss": 0.09448353946208954, "global_step": 210313, "epoch": 2363, "lr": 1.2044631145892533e-05}
{"train_loss": 0.047708746045827866, "global_step": 210314, "epoch": 2363, "lr": 1.2044253759957019e-05}
{"train_loss": 0.04821779206395149, "global_step": 210315, "epoch": 2363, "lr": 1.2043876379124191e-05}
{"train_loss": 0.082102470099926, "global_step": 210316, "epoch": 2363, "lr": 1.2043499003394093e-05}
{"train_loss": 0.06391868740320206, "global_step": 210317, "epoch": 2363, "lr": 1.204312163276679e-05}
{"train_loss": 0.07136251777410507, "global_step": 210318, "epoch": 2363, "lr": 1.2042744267242318e-05}
{"train_loss": 0.021756790578365326, "global_step": 210319, "epoch": 2363, "lr": 1.2042366906820751e-05}
{"train_loss": 0.05195210501551628, "global_step": 210320, "epoch": 2363, "lr": 1.204198955150211e-05}
{"train_loss": 0.04651792719960213, "global_step": 210321, "epoch": 2363, "lr": 1.2041612201286473e-05}
{"train_loss": 0.041931092739105225, "global_step": 210322, "epoch": 2363, "lr": 1.2041234856173866e-05}
{"train_loss": 0.04033976420760155, "global_step": 210323, "epoch": 2363, "lr": 1.2040857516164367e-05}
{"train_loss": 0.06245410442352295, "global_step": 210324, "epoch": 2363, "lr": 1.2040480181257996e-05}
{"train_loss": 0.052747052162885666, "global_step": 210325, "epoch": 2363, "lr": 1.204010285145482e-05}
{"train_loss": 0.04753296077251434, "global_step": 210326, "epoch": 2363, "lr": 1.2039725526754902e-05}
{"train_loss": 0.05188293755054474, "global_step": 210327, "epoch": 2363, "lr": 1.2039348207158268e-05}
{"train_loss": 0.04735751822590828, "global_step": 210328, "epoch": 2363, "lr": 1.2038970892664996e-05}
{"train_loss": 0.05514451116323471, "global_step": 210329, "epoch": 2363, "lr": 1.203859358327511e-05}
{"train_loss": 0.054367344826459885, "global_step": 210330, "epoch": 2363, "lr": 1.2038216278988685e-05}
{"train_loss": 0.04578687995672226, "global_step": 210331, "epoch": 2363, "lr": 1.2037838979805744e-05}
{"train_loss": 0.029791658744215965, "global_step": 210332, "epoch": 2363, "lr": 1.2037461685726364e-05}
{"train_loss": 0.03596492111682892, "global_step": 210333, "epoch": 2363, "lr": 1.203708439675057e-05}
{"train_loss": 0.06917102634906769, "global_step": 210334, "epoch": 2363, "lr": 1.2036707112878448e-05}
{"train_loss": 0.07526785880327225, "global_step": 210335, "epoch": 2363, "lr": 1.203632983411001e-05}
{"train_loss": 0.04332534968852997, "global_step": 210336, "epoch": 2363, "lr": 1.203595256044534e-05}
{"train_loss": 0.09337730705738068, "global_step": 210337, "epoch": 2363, "lr": 1.2035575291884454e-05}
{"train_loss": 0.04753201827406883, "global_step": 210338, "epoch": 2363, "lr": 1.2035198028427446e-05}
{"train_loss": 0.0395924374461174, "global_step": 210339, "epoch": 2363, "lr": 1.2034820770074324e-05}
{"train_loss": 0.043847646564245224, "global_step": 210340, "epoch": 2363, "lr": 1.2034443516825166e-05}
{"train_loss": 0.03143463283777237, "global_step": 210341, "epoch": 2363, "lr": 1.203406626868001e-05}
{"train_loss": 0.08768345415592194, "global_step": 210342, "epoch": 2363, "lr": 1.2033689025638923e-05}
{"train_loss": 0.041492752730846405, "global_step": 210343, "epoch": 2363, "lr": 1.2033311787701924e-05}
{"train_loss": 0.05986710637807846, "global_step": 210344, "epoch": 2363, "lr": 1.2032934554869086e-05}
{"train_loss": 0.04913710802793503, "global_step": 210345, "epoch": 2363, "lr": 1.2032557327140475e-05}
{"train_loss": 0.06677263230085373, "global_step": 210346, "epoch": 2363, "lr": 1.2032180104516105e-05}
{"train_loss": 0.08869738131761551, "global_step": 210347, "epoch": 2363, "lr": 1.2031802886996063e-05}
{"train_loss": 0.052194129675626755, "global_step": 210348, "epoch": 2363, "lr": 1.2031425674580381e-05}
{"train_loss": 0.08880758285522461, "global_step": 210349, "epoch": 2363, "lr": 1.2031048467269097e-05}
{"train_loss": 0.03870757296681404, "global_step": 210350, "epoch": 2363, "lr": 1.2030671265062288e-05}
{"train_loss": 0.07576516270637512, "global_step": 210351, "epoch": 2363, "lr": 1.203029406795998e-05}
{"train_loss": 0.05510365962982178, "global_step": 210352, "epoch": 2363, "lr": 1.202991687596225e-05}
{"train_loss": 0.05751977115869522, "global_step": 210353, "epoch": 2363, "lr": 1.2029539689069137e-05}
{"train_loss": 0.08386380970478058, "global_step": 210354, "epoch": 2363, "lr": 1.2029162507280672e-05}
{"train_loss": 0.04138569161295891, "global_step": 210355, "epoch": 2363, "lr": 1.2028785330596942e-05}
{"train_loss": 0.04547136276960373, "global_step": 210356, "epoch": 2363, "lr": 1.2028408159017963e-05}
{"train_loss": 0.018185589462518692, "global_step": 210357, "epoch": 2363, "lr": 1.2028030992543815e-05}
{"train_loss": 0.04846000671386719, "global_step": 210358, "epoch": 2363, "lr": 1.2027653831174518e-05}
{"train_loss": 0.08188003301620483, "global_step": 210359, "epoch": 2363, "lr": 1.2027276674910159e-05}
{"train_loss": 0.033227451145648956, "global_step": 210360, "epoch": 2363, "lr": 1.2026899523750757e-05}
{"train_loss": 0.09464949369430542, "global_step": 210361, "epoch": 2363, "lr": 1.202652237769637e-05}
{"train_loss": 0.048140283674001694, "global_step": 210362, "epoch": 2363, "lr": 1.2026145236747072e-05}
{"train_loss": 0.06207072734832764, "global_step": 210363, "epoch": 2363, "lr": 1.202576810090288e-05}
{"train_loss": 0.05057023465633392, "global_step": 210364, "epoch": 2363, "lr": 1.2025390970163875e-05}
{"train_loss": 0.04993031173944473, "global_step": 210365, "epoch": 2363, "lr": 1.2025013844530082e-05}
{"train_loss": 0.06495830416679382, "global_step": 210366, "epoch": 2363, "lr": 1.2024636724001576e-05}
{"train_loss": 0.07182827591896057, "global_step": 210367, "epoch": 2363, "lr": 1.202425960857838e-05}
{"train_loss": 0.07323602586984634, "global_step": 210368, "epoch": 2363, "lr": 1.202388249826057e-05}
{"train_loss": 0.08530933409929276, "global_step": 210369, "epoch": 2363, "lr": 1.2023505393048178e-05}
{"train_loss": 0.03987419977784157, "global_step": 210370, "epoch": 2363, "lr": 1.2023128292941271e-05}
{"train_loss": 0.06073763221502304, "global_step": 210371, "epoch": 2363, "lr": 1.2022751197939886e-05}
{"train_loss": 0.1036066859960556, "global_step": 210372, "epoch": 2363, "lr": 1.2022374108044088e-05}
{"train_loss": 0.062176257371902466, "global_step": 210373, "epoch": 2363, "lr": 1.20219970232539e-05}
{"train_loss": 0.0626518502831459, "global_step": 210374, "epoch": 2363, "lr": 1.2021619943569411e-05}
{"train_loss": 0.048734258860349655, "global_step": 210375, "epoch": 2363, "lr": 1.202124286899064e-05}
{"train_loss": 0.046775512397289276, "global_step": 210376, "epoch": 2363, "lr": 1.2020865799517661e-05}
{"train_loss": 0.08830537647008896, "global_step": 210377, "epoch": 2363, "lr": 1.2020488735150498e-05}
{"train_loss": 0.0642869770526886, "global_step": 210378, "epoch": 2363, "lr": 1.2020111675889234e-05}
{"train_loss": 0.048569206148386, "global_step": 210379, "epoch": 2363, "lr": 1.2019734621733892e-05}
{"train_loss": 0.12481352686882019, "global_step": 210380, "epoch": 2363, "lr": 1.2019357572684532e-05}
{"train_loss": 0.07267559319734573, "global_step": 210381, "epoch": 2363, "lr": 1.201898052874122e-05}
{"train_loss": 0.1287725269794464, "global_step": 210382, "epoch": 2363, "lr": 1.2018603489903974e-05}
{"train_loss": 0.07024399191141129, "global_step": 210383, "epoch": 2363, "lr": 1.2018226456172883e-05}
{"train_loss": 0.07112360745668411, "global_step": 210384, "epoch": 2363, "lr": 1.2017849427547962e-05}
{"train_loss": 0.12643659114837646, "global_step": 210385, "epoch": 2363, "lr": 1.2017472404029295e-05}
{"train_loss": 0.11242746561765671, "global_step": 210386, "epoch": 2363, "lr": 1.2017095385616916e-05}
{"train_loss": 0.048007428646087646, "global_step": 210387, "epoch": 2363, "lr": 1.201671837231086e-05}
{"train_loss": 0.04569756239652634, "global_step": 210388, "epoch": 2363, "lr": 1.2016341364111206e-05}
{"train_loss": 0.045410607010126114, "global_step": 210389, "epoch": 2363, "lr": 1.2015964361017985e-05}
{"train_loss": 0.033080898225307465, "global_step": 210390, "epoch": 2363, "lr": 1.201558736303126e-05}
{"train_loss": 0.07875732332468033, "global_step": 210391, "epoch": 2363, "lr": 1.2015210370151081e-05}
{"train_loss": 0.08080235123634338, "global_step": 210392, "epoch": 2363, "lr": 1.2014833382377477e-05}
{"train_loss": 0.09150301665067673, "global_step": 210393, "epoch": 2363, "lr": 1.2014456399710533e-05}
{"train_loss": 0.04615398123860359, "global_step": 210394, "epoch": 2363, "lr": 1.2014079422150265e-05}
{"train_loss": 0.06205989493664061, "global_step": 210395, "epoch": 2363, "lr": 1.2013702449696756e-05, "val_loss": 8.570648193359375}
{"train_loss": 0.04514850303530693, "global_step": 210396, "epoch": 2364, "lr": 1.201332548235003e-05}
{"train_loss": 0.0595172680914402, "global_step": 210397, "epoch": 2364, "lr": 1.2012948520110162e-05}
{"train_loss": 0.04471004009246826, "global_step": 210398, "epoch": 2364, "lr": 1.2012571562977177e-05}
{"train_loss": 0.08479321748018265, "global_step": 210399, "epoch": 2364, "lr": 1.2012194610951133e-05}
{"train_loss": 0.04511428251862526, "global_step": 210400, "epoch": 2364, "lr": 1.2011817664032105e-05}
{"train_loss": 0.09067361056804657, "global_step": 210401, "epoch": 2364, "lr": 1.2011440722220113e-05}
{"train_loss": 0.05800572782754898, "global_step": 210402, "epoch": 2364, "lr": 1.2011063785515231e-05}
{"train_loss": 0.04157537594437599, "global_step": 210403, "epoch": 2364, "lr": 1.2010686853917485e-05}
{"train_loss": 0.09820031374692917, "global_step": 210404, "epoch": 2364, "lr": 1.2010309927426956e-05}
{"train_loss": 0.11396472901105881, "global_step": 210405, "epoch": 2364, "lr": 1.2009933006043662e-05}
{"train_loss": 0.08233831822872162, "global_step": 210406, "epoch": 2364, "lr": 1.2009556089767682e-05}
{"train_loss": 0.03932873159646988, "global_step": 210407, "epoch": 2364, "lr": 1.200917917859904e-05}
{"train_loss": 0.04622127488255501, "global_step": 210408, "epoch": 2364, "lr": 1.200880227253781e-05}
{"train_loss": 0.053449176251888275, "global_step": 210409, "epoch": 2364, "lr": 1.2008425371584032e-05}
{"train_loss": 0.07982264459133148, "global_step": 210410, "epoch": 2364, "lr": 1.200804847573776e-05}
{"train_loss": 0.09469226747751236, "global_step": 210411, "epoch": 2364, "lr": 1.2007671584999037e-05}
{"train_loss": 0.09059108048677444, "global_step": 210412, "epoch": 2364, "lr": 1.2007294699367933e-05}
{"train_loss": 0.03874935209751129, "global_step": 210413, "epoch": 2364, "lr": 1.2006917818844465e-05}
{"train_loss": 0.02585817500948906, "global_step": 210414, "epoch": 2364, "lr": 1.2006540943428724e-05}
{"train_loss": 0.057678356766700745, "global_step": 210415, "epoch": 2364, "lr": 1.2006164073120724e-05}
{"train_loss": 0.04607747122645378, "global_step": 210416, "epoch": 2364, "lr": 1.2005787207920533e-05}
{"train_loss": 0.031680867075920105, "global_step": 210417, "epoch": 2364, "lr": 1.200541034782821e-05}
{"train_loss": 0.03647659346461296, "global_step": 210418, "epoch": 2364, "lr": 1.2005033492843787e-05}
{"train_loss": 0.05157479643821716, "global_step": 210419, "epoch": 2364, "lr": 1.2004656642967338e-05}
{"train_loss": 0.0643516257405281, "global_step": 210420, "epoch": 2364, "lr": 1.2004279798198886e-05}
{"train_loss": 0.03301620855927467, "global_step": 210421, "epoch": 2364, "lr": 1.200390295853851e-05}
{"train_loss": 0.05647987499833107, "global_step": 210422, "epoch": 2364, "lr": 1.200352612398623e-05}
{"train_loss": 0.02454381063580513, "global_step": 210423, "epoch": 2364, "lr": 1.2003149294542127e-05}
{"train_loss": 0.09369207918643951, "global_step": 210424, "epoch": 2364, "lr": 1.2002772470206242e-05}
{"train_loss": 0.06591901928186417, "global_step": 210425, "epoch": 2364, "lr": 1.2002395650978603e-05}
{"train_loss": 0.03811997175216675, "global_step": 210426, "epoch": 2364, "lr": 1.2002018836859291e-05}
{"train_loss": 0.05330409109592438, "global_step": 210427, "epoch": 2364, "lr": 1.2001642027848331e-05}
{"train_loss": 0.06674664467573166, "global_step": 210428, "epoch": 2364, "lr": 1.2001265223945807e-05}
{"train_loss": 0.05000418424606323, "global_step": 210429, "epoch": 2364, "lr": 1.200088842515173e-05}
{"train_loss": 0.06433659791946411, "global_step": 210430, "epoch": 2364, "lr": 1.2000511631466188e-05}
{"train_loss": 0.04059824347496033, "global_step": 210431, "epoch": 2364, "lr": 1.200013484288921e-05}
{"train_loss": 0.07547002285718918, "global_step": 210432, "epoch": 2364, "lr": 1.199975805942084e-05}
{"train_loss": 0.05848376452922821, "global_step": 210433, "epoch": 2364, "lr": 1.199938128106115e-05}
{"train_loss": 0.03599535673856735, "global_step": 210434, "epoch": 2364, "lr": 1.199900450781017e-05}
{"train_loss": 0.07778337597846985, "global_step": 210435, "epoch": 2364, "lr": 1.1998627739667956e-05}
{"train_loss": 0.12882454693317413, "global_step": 210436, "epoch": 2364, "lr": 1.1998250976634584e-05}
{"train_loss": 0.05820050090551376, "global_step": 210437, "epoch": 2364, "lr": 1.1997874218710065e-05}
{"train_loss": 0.06256614625453949, "global_step": 210438, "epoch": 2364, "lr": 1.1997497465894486e-05}
{"train_loss": 0.047608256340026855, "global_step": 210439, "epoch": 2364, "lr": 1.1997120718187859e-05}
{"train_loss": 0.015229438431560993, "global_step": 210440, "epoch": 2364, "lr": 1.199674397559028e-05}
{"train_loss": 0.032023195177316666, "global_step": 210441, "epoch": 2364, "lr": 1.1996367238101752e-05}
{"train_loss": 0.05097842961549759, "global_step": 210442, "epoch": 2364, "lr": 1.199599050572237e-05}
{"train_loss": 0.06347756832838058, "global_step": 210443, "epoch": 2364, "lr": 1.1995613778452147e-05}
{"train_loss": 0.0510646253824234, "global_step": 210444, "epoch": 2364, "lr": 1.1995237056291165e-05}
{"train_loss": 0.014567029662430286, "global_step": 210445, "epoch": 2364, "lr": 1.1994860339239445e-05}
{"train_loss": 0.040611110627651215, "global_step": 210446, "epoch": 2364, "lr": 1.199448362729707e-05}
{"train_loss": 0.012008978985249996, "global_step": 210447, "epoch": 2364, "lr": 1.199410692046406e-05}
{"train_loss": 0.014534817077219486, "global_step": 210448, "epoch": 2364, "lr": 1.1993730218740489e-05}
{"train_loss": 0.03604607284069061, "global_step": 210449, "epoch": 2364, "lr": 1.1993353522126383e-05}
{"train_loss": 0.02607228048145771, "global_step": 210450, "epoch": 2364, "lr": 1.1992976830621827e-05}
{"train_loss": 0.04459504038095474, "global_step": 210451, "epoch": 2364, "lr": 1.1992600144226834e-05}
{"train_loss": 0.09313389658927917, "global_step": 210452, "epoch": 2364, "lr": 1.1992223462941471e-05}
{"train_loss": 0.06681458652019501, "global_step": 210453, "epoch": 2364, "lr": 1.199184678676581e-05}
{"train_loss": 0.024587372317910194, "global_step": 210454, "epoch": 2364, "lr": 1.1991470115699866e-05}
{"train_loss": 0.07168130576610565, "global_step": 210455, "epoch": 2364, "lr": 1.1991093449743717e-05}
{"train_loss": 0.05409101024270058, "global_step": 210456, "epoch": 2364, "lr": 1.1990716788897389e-05}
{"train_loss": 0.046303119510412216, "global_step": 210457, "epoch": 2364, "lr": 1.1990340133160959e-05}
{"train_loss": 0.06276004016399384, "global_step": 210458, "epoch": 2364, "lr": 1.1989963482534456e-05}
{"train_loss": 0.054701924324035645, "global_step": 210459, "epoch": 2364, "lr": 1.1989586837017946e-05}
{"train_loss": 0.053888432681560516, "global_step": 210460, "epoch": 2364, "lr": 1.1989210196611461e-05}
{"train_loss": 0.049569256603717804, "global_step": 210461, "epoch": 2364, "lr": 1.1988833561315076e-05}
{"train_loss": 0.06262143701314926, "global_step": 210462, "epoch": 2364, "lr": 1.1988456931128821e-05}
{"train_loss": 0.048833318054676056, "global_step": 210463, "epoch": 2364, "lr": 1.1988080306052762e-05}
{"train_loss": 0.013233209028840065, "global_step": 210464, "epoch": 2364, "lr": 1.1987703686086948e-05}
{"train_loss": 0.06202474609017372, "global_step": 210465, "epoch": 2364, "lr": 1.1987327071231409e-05}
{"train_loss": 0.07837396115064621, "global_step": 210466, "epoch": 2364, "lr": 1.1986950461486218e-05}
{"train_loss": 0.04306390509009361, "global_step": 210467, "epoch": 2364, "lr": 1.1986573856851414e-05}
{"train_loss": 0.05168424919247627, "global_step": 210468, "epoch": 2364, "lr": 1.1986197257327058e-05}
{"train_loss": 0.05766493082046509, "global_step": 210469, "epoch": 2364, "lr": 1.19858206629132e-05}
{"train_loss": 0.04021814465522766, "global_step": 210470, "epoch": 2364, "lr": 1.1985444073609864e-05}
{"train_loss": 0.03736685961484909, "global_step": 210471, "epoch": 2364, "lr": 1.1985067489417129e-05}
{"train_loss": 0.043947961181402206, "global_step": 210472, "epoch": 2364, "lr": 1.198469091033505e-05}
{"train_loss": 0.06651842594146729, "global_step": 210473, "epoch": 2364, "lr": 1.198431433636365e-05}
{"train_loss": 0.06651776283979416, "global_step": 210474, "epoch": 2364, "lr": 1.1983937767503011e-05}
{"train_loss": 0.08923059701919556, "global_step": 210475, "epoch": 2364, "lr": 1.1983561203753152e-05}
{"train_loss": 0.035385895520448685, "global_step": 210476, "epoch": 2364, "lr": 1.1983184645114159e-05}
{"train_loss": 0.06893587112426758, "global_step": 210477, "epoch": 2364, "lr": 1.1982808091586046e-05}
{"train_loss": 0.11412056535482407, "global_step": 210478, "epoch": 2364, "lr": 1.1982431543168892e-05}
{"train_loss": 0.050920434296131134, "global_step": 210479, "epoch": 2364, "lr": 1.198205499986273e-05}
{"train_loss": 0.06654946506023407, "global_step": 210480, "epoch": 2364, "lr": 1.1981678461667628e-05}
{"train_loss": 0.06220962852239609, "global_step": 210481, "epoch": 2364, "lr": 1.198130192858361e-05}
{"train_loss": 0.06321869790554047, "global_step": 210482, "epoch": 2364, "lr": 1.198092540061076e-05}
{"train_loss": 0.12493164092302322, "global_step": 210483, "epoch": 2364, "lr": 1.1980548877749092e-05}
{"train_loss": 0.056268350629324324, "global_step": 210484, "epoch": 2364, "lr": 1.1980172359998692e-05, "val_loss": 8.63845443725586}
{"train_loss": 0.04908197373151779, "global_step": 210485, "epoch": 2365, "lr": 1.1979795847359582e-05}
{"train_loss": 0.0771404281258583, "global_step": 210486, "epoch": 2365, "lr": 1.1979419339831844e-05}
{"train_loss": 0.023733071982860565, "global_step": 210487, "epoch": 2365, "lr": 1.1979042837415488e-05}
{"train_loss": 0.042509883642196655, "global_step": 210488, "epoch": 2365, "lr": 1.1978666340110589e-05}
{"train_loss": 0.06647608429193497, "global_step": 210489, "epoch": 2365, "lr": 1.1978289847917207e-05}
{"train_loss": 0.08457326889038086, "global_step": 210490, "epoch": 2365, "lr": 1.197791336083537e-05}
{"train_loss": 0.06266328692436218, "global_step": 210491, "epoch": 2365, "lr": 1.1977536878865147e-05}
{"train_loss": 0.07446568459272385, "global_step": 210492, "epoch": 2365, "lr": 1.197716040200657e-05}
{"train_loss": 0.07373036444187164, "global_step": 210493, "epoch": 2365, "lr": 1.1976783930259717e-05}
{"train_loss": 0.0916556715965271, "global_step": 210494, "epoch": 2365, "lr": 1.1976407463624605e-05}
{"train_loss": 0.0756002813577652, "global_step": 210495, "epoch": 2365, "lr": 1.1976031002101318e-05}
{"train_loss": 0.08671557903289795, "global_step": 210496, "epoch": 2365, "lr": 1.1975654545689874e-05}
{"train_loss": 0.03653557598590851, "global_step": 210497, "epoch": 2365, "lr": 1.1975278094390357e-05}
{"train_loss": 0.07196082174777985, "global_step": 210498, "epoch": 2365, "lr": 1.1974901648202785e-05}
{"train_loss": 0.028579071164131165, "global_step": 210499, "epoch": 2365, "lr": 1.1974525207127235e-05}
{"train_loss": 0.07330364733934402, "global_step": 210500, "epoch": 2365, "lr": 1.1974148771163734e-05}
{"train_loss": 0.08022390305995941, "global_step": 210501, "epoch": 2365, "lr": 1.1973772340312362e-05}
{"train_loss": 0.11880416423082352, "global_step": 210502, "epoch": 2365, "lr": 1.197339591457315e-05}
{"train_loss": 0.035366252064704895, "global_step": 210503, "epoch": 2365, "lr": 1.1973019493946136e-05}
{"train_loss": 0.05752578005194664, "global_step": 210504, "epoch": 2365, "lr": 1.1972643078431406e-05}
{"train_loss": 0.1224452406167984, "global_step": 210505, "epoch": 2365, "lr": 1.1972266668028969e-05}
{"train_loss": 0.08445005863904953, "global_step": 210506, "epoch": 2365, "lr": 1.1971890262738916e-05}
{"train_loss": 0.1477876752614975, "global_step": 210507, "epoch": 2365, "lr": 1.1971513862561263e-05}
{"train_loss": 0.07344911992549896, "global_step": 210508, "epoch": 2365, "lr": 1.197113746749609e-05}
{"train_loss": 0.02953704260289669, "global_step": 210509, "epoch": 2365, "lr": 1.1970761077543424e-05}
{"train_loss": 0.08066366612911224, "global_step": 210510, "epoch": 2365, "lr": 1.197038469270334e-05}
{"train_loss": 0.031477946788072586, "global_step": 210511, "epoch": 2365, "lr": 1.1970008312975856e-05}
{"train_loss": 0.061450205743312836, "global_step": 210512, "epoch": 2365, "lr": 1.1969631938361054e-05}
{"train_loss": 0.053322963416576385, "global_step": 210513, "epoch": 2365, "lr": 1.1969255568858961e-05}
{"train_loss": 0.08789006620645523, "global_step": 210514, "epoch": 2365, "lr": 1.1968879204469651e-05}
{"train_loss": 0.06478337943553925, "global_step": 210515, "epoch": 2365, "lr": 1.1968502845193147e-05}
{"train_loss": 0.07183724641799927, "global_step": 210516, "epoch": 2365, "lr": 1.196812649102953e-05}
{"train_loss": 0.025288870558142662, "global_step": 210517, "epoch": 2365, "lr": 1.1967750141978818e-05}
{"train_loss": 0.06281504034996033, "global_step": 210518, "epoch": 2365, "lr": 1.1967373798041093e-05}
{"train_loss": 0.11508035659790039, "global_step": 210519, "epoch": 2365, "lr": 1.1966997459216377e-05}
{"train_loss": 0.06315697729587555, "global_step": 210520, "epoch": 2365, "lr": 1.196662112550475e-05}
{"train_loss": 0.04859035834670067, "global_step": 210521, "epoch": 2365, "lr": 1.1966244796906235e-05}
{"train_loss": 0.08785401284694672, "global_step": 210522, "epoch": 2365, "lr": 1.196586847342091e-05}
{"train_loss": 0.023003486916422844, "global_step": 210523, "epoch": 2365, "lr": 1.196549215504879e-05}
{"train_loss": 0.07620184123516083, "global_step": 210524, "epoch": 2365, "lr": 1.1965115841789965e-05}
{"train_loss": 0.05517926439642906, "global_step": 210525, "epoch": 2365, "lr": 1.1964739533644447e-05}
{"train_loss": 0.06016286462545395, "global_step": 210526, "epoch": 2365, "lr": 1.1964363230612313e-05}
{"train_loss": 0.08596765995025635, "global_step": 210527, "epoch": 2365, "lr": 1.1963986932693616e-05}
{"train_loss": 0.06556109338998795, "global_step": 210528, "epoch": 2365, "lr": 1.1963610639888385e-05}
{"train_loss": 0.037013888359069824, "global_step": 210529, "epoch": 2365, "lr": 1.19632343521967e-05}
{"train_loss": 0.1266942322254181, "global_step": 210530, "epoch": 2365, "lr": 1.1962858069618577e-05}
{"train_loss": 0.032619841396808624, "global_step": 210531, "epoch": 2365, "lr": 1.1962481792154095e-05}
{"train_loss": 0.0801813080906868, "global_step": 210532, "epoch": 2365, "lr": 1.1962105519803285e-05}
{"train_loss": 0.04234940931200981, "global_step": 210533, "epoch": 2365, "lr": 1.1961729252566218e-05}
{"train_loss": 0.08652058243751526, "global_step": 210534, "epoch": 2365, "lr": 1.196135299044292e-05}
{"train_loss": 0.09662416577339172, "global_step": 210535, "epoch": 2365, "lr": 1.1960976733433471e-05}
{"train_loss": 0.09411650896072388, "global_step": 210536, "epoch": 2365, "lr": 1.1960600481537887e-05}
{"train_loss": 0.0542854443192482, "global_step": 210537, "epoch": 2365, "lr": 1.196022423475625e-05}
{"train_loss": 0.07612114399671555, "global_step": 210538, "epoch": 2365, "lr": 1.1959847993088585e-05}
{"train_loss": 0.09197372198104858, "global_step": 210539, "epoch": 2365, "lr": 1.1959471756534968e-05}
{"train_loss": 0.06159968674182892, "global_step": 210540, "epoch": 2365, "lr": 1.1959095525095426e-05}
{"train_loss": 0.038922734558582306, "global_step": 210541, "epoch": 2365, "lr": 1.1958719298770033e-05}
{"train_loss": 0.03170369938015938, "global_step": 210542, "epoch": 2365, "lr": 1.195834307755882e-05}
{"train_loss": 0.055314917117357254, "global_step": 210543, "epoch": 2365, "lr": 1.1957966861461838e-05}
{"train_loss": 0.04799845814704895, "global_step": 210544, "epoch": 2365, "lr": 1.195759065047915e-05}
{"train_loss": 0.014659272506833076, "global_step": 210545, "epoch": 2365, "lr": 1.1957214444610792e-05}
{"train_loss": 0.033991970121860504, "global_step": 210546, "epoch": 2365, "lr": 1.1956838243856838e-05}
{"train_loss": 0.04146447777748108, "global_step": 210547, "epoch": 2365, "lr": 1.1956462048217304e-05}
{"train_loss": 0.0565677173435688, "global_step": 210548, "epoch": 2365, "lr": 1.1956085857692279e-05}
{"train_loss": 0.04443105310201645, "global_step": 210549, "epoch": 2365, "lr": 1.1955709672281779e-05}
{"train_loss": 0.04836316406726837, "global_step": 210550, "epoch": 2365, "lr": 1.1955333491985881e-05}
{"train_loss": 0.06735589355230331, "global_step": 210551, "epoch": 2365, "lr": 1.1954957316804616e-05}
{"train_loss": 0.0704193264245987, "global_step": 210552, "epoch": 2365, "lr": 1.1954581146738054e-05}
{"train_loss": 0.028597427532076836, "global_step": 210553, "epoch": 2365, "lr": 1.1954204981786222e-05}
{"train_loss": 0.05145883932709694, "global_step": 210554, "epoch": 2365, "lr": 1.1953828821949197e-05}
{"train_loss": 0.051940325647592545, "global_step": 210555, "epoch": 2365, "lr": 1.1953452667227e-05}
{"train_loss": 0.031278517097234726, "global_step": 210556, "epoch": 2365, "lr": 1.1953076517619715e-05}
{"train_loss": 0.026667773723602295, "global_step": 210557, "epoch": 2365, "lr": 1.1952700373127356e-05}
{"train_loss": 0.02992384508252144, "global_step": 210558, "epoch": 2365, "lr": 1.1952324233750012e-05}
{"train_loss": 0.047996457666158676, "global_step": 210559, "epoch": 2365, "lr": 1.1951948099487697e-05}
{"train_loss": 0.05771617218852043, "global_step": 210560, "epoch": 2365, "lr": 1.1951571970340497e-05}
{"train_loss": 0.043639566749334335, "global_step": 210561, "epoch": 2365, "lr": 1.1951195846308427e-05}
{"train_loss": 0.08404523134231567, "global_step": 210562, "epoch": 2365, "lr": 1.1950819727391554e-05}
{"train_loss": 0.029631609097123146, "global_step": 210563, "epoch": 2365, "lr": 1.1950443613589946e-05}
{"train_loss": 0.06981876492500305, "global_step": 210564, "epoch": 2365, "lr": 1.1950067504903623e-05}
{"train_loss": 0.044170744717121124, "global_step": 210565, "epoch": 2365, "lr": 1.1949691401332663e-05}
{"train_loss": 0.07338124513626099, "global_step": 210566, "epoch": 2365, "lr": 1.194931530287709e-05}
{"train_loss": 0.10817320644855499, "global_step": 210567, "epoch": 2365, "lr": 1.194893920953698e-05}
{"train_loss": 0.07083743065595627, "global_step": 210568, "epoch": 2365, "lr": 1.1948563121312362e-05}
{"train_loss": 0.04922755807638168, "global_step": 210569, "epoch": 2365, "lr": 1.1948187038203307e-05}
{"train_loss": 0.07296190410852432, "global_step": 210570, "epoch": 2365, "lr": 1.1947810960209837e-05}
{"train_loss": 0.032636649906635284, "global_step": 210571, "epoch": 2365, "lr": 1.1947434887332037e-05}
{"train_loss": 0.043445657938718796, "global_step": 210572, "epoch": 2365, "lr": 1.1947058819569928e-05}
{"train_loss": 0.06188130977281024, "global_step": 210573, "epoch": 2365, "lr": 1.1946682756923583e-05, "val_loss": 8.656915664672852, "train_action_mse_error": 16.035354614257812}
{"train_loss": 0.05790659785270691, "global_step": 210574, "epoch": 2366, "lr": 1.1946306699393034e-05}
{"train_loss": 0.05455004423856735, "global_step": 210575, "epoch": 2366, "lr": 1.1945930646978348e-05}
{"train_loss": 0.02220245636999607, "global_step": 210576, "epoch": 2366, "lr": 1.1945554599679559e-05}
{"train_loss": 0.03867201879620552, "global_step": 210577, "epoch": 2366, "lr": 1.1945178557496739e-05}
{"train_loss": 0.05340230464935303, "global_step": 210578, "epoch": 2366, "lr": 1.194480252042991e-05}
{"train_loss": 0.04333752021193504, "global_step": 210579, "epoch": 2366, "lr": 1.1944426488479155e-05}
{"train_loss": 0.05914418771862984, "global_step": 210580, "epoch": 2366, "lr": 1.1944050461644508e-05}
{"train_loss": 0.05932636186480522, "global_step": 210581, "epoch": 2366, "lr": 1.1943674439926007e-05}
{"train_loss": 0.060002315789461136, "global_step": 210582, "epoch": 2366, "lr": 1.1943298423323724e-05}
{"train_loss": 0.03812526538968086, "global_step": 210583, "epoch": 2366, "lr": 1.1942922411837687e-05}
{"train_loss": 0.038052886724472046, "global_step": 210584, "epoch": 2366, "lr": 1.194254640546798e-05}
{"train_loss": 0.018891844898462296, "global_step": 210585, "epoch": 2366, "lr": 1.1942170404214614e-05}
{"train_loss": 0.019723741337656975, "global_step": 210586, "epoch": 2366, "lr": 1.1941794408077679e-05}
{"train_loss": 0.06917227059602737, "global_step": 210587, "epoch": 2366, "lr": 1.1941418417057188e-05}
{"train_loss": 0.050611820071935654, "global_step": 210588, "epoch": 2366, "lr": 1.194104243115322e-05}
{"train_loss": 0.03485001251101494, "global_step": 210589, "epoch": 2366, "lr": 1.1940666450365806e-05}
{"train_loss": 0.023674894124269485, "global_step": 210590, "epoch": 2366, "lr": 1.194029047469502e-05}
{"train_loss": 0.08441327512264252, "global_step": 210591, "epoch": 2366, "lr": 1.1939914504140882e-05}
{"train_loss": 0.04131411015987396, "global_step": 210592, "epoch": 2366, "lr": 1.1939538538703477e-05}
{"train_loss": 0.09628742933273315, "global_step": 210593, "epoch": 2366, "lr": 1.1939162578382818e-05}
{"train_loss": 0.04734138026833534, "global_step": 210594, "epoch": 2366, "lr": 1.1938786623178988e-05}
{"train_loss": 0.027755098417401314, "global_step": 210595, "epoch": 2366, "lr": 1.1938410673092016e-05}
{"train_loss": 0.04176913574337959, "global_step": 210596, "epoch": 2366, "lr": 1.1938034728121966e-05}
{"train_loss": 0.0653219148516655, "global_step": 210597, "epoch": 2366, "lr": 1.1937658788268879e-05}
{"train_loss": 0.04554345831274986, "global_step": 210598, "epoch": 2366, "lr": 1.1937282853532804e-05}
{"train_loss": 0.04621802270412445, "global_step": 210599, "epoch": 2366, "lr": 1.193690692391381e-05}
{"train_loss": 0.0546875074505806, "global_step": 210600, "epoch": 2366, "lr": 1.1936530999411927e-05}
{"train_loss": 0.03631427511572838, "global_step": 210601, "epoch": 2366, "lr": 1.1936155080027222e-05}
{"train_loss": 0.06189554184675217, "global_step": 210602, "epoch": 2366, "lr": 1.1935779165759726e-05}
{"train_loss": 0.02824411913752556, "global_step": 210603, "epoch": 2366, "lr": 1.1935403256609511e-05}
{"train_loss": 0.0705045834183693, "global_step": 210604, "epoch": 2366, "lr": 1.1935027352576611e-05}
{"train_loss": 0.050866659730672836, "global_step": 210605, "epoch": 2366, "lr": 1.1934651453661089e-05}
{"train_loss": 0.04690079763531685, "global_step": 210606, "epoch": 2366, "lr": 1.1934275559862978e-05}
{"train_loss": 0.05052339658141136, "global_step": 210607, "epoch": 2366, "lr": 1.1933899671182353e-05}
{"train_loss": 0.05175047367811203, "global_step": 210608, "epoch": 2366, "lr": 1.1933523787619239e-05}
{"train_loss": 0.033519428223371506, "global_step": 210609, "epoch": 2366, "lr": 1.1933147909173708e-05}
{"train_loss": 0.06975506246089935, "global_step": 210610, "epoch": 2366, "lr": 1.1932772035845791e-05}
{"train_loss": 0.04021378606557846, "global_step": 210611, "epoch": 2366, "lr": 1.1932396167635562e-05}
{"train_loss": 0.06661565601825714, "global_step": 210612, "epoch": 2366, "lr": 1.1932020304543046e-05}
{"train_loss": 0.04373147711157799, "global_step": 210613, "epoch": 2366, "lr": 1.193164444656832e-05}
{"train_loss": 0.04676837846636772, "global_step": 210614, "epoch": 2366, "lr": 1.1931268593711403e-05}
{"train_loss": 0.07362636178731918, "global_step": 210615, "epoch": 2366, "lr": 1.1930892745972377e-05}
{"train_loss": 0.05842970311641693, "global_step": 210616, "epoch": 2366, "lr": 1.193051690335127e-05}
{"train_loss": 0.07003246992826462, "global_step": 210617, "epoch": 2366, "lr": 1.193014106584815e-05}
{"train_loss": 0.035038262605667114, "global_step": 210618, "epoch": 2366, "lr": 1.1929765233463058e-05}
{"train_loss": 0.0477290078997612, "global_step": 210619, "epoch": 2366, "lr": 1.1929389406196033e-05}
{"train_loss": 0.04404519498348236, "global_step": 210620, "epoch": 2366, "lr": 1.1929013584047155e-05}
{"train_loss": 0.05915876477956772, "global_step": 210621, "epoch": 2366, "lr": 1.1928637767016437e-05}
{"train_loss": 0.03499449044466019, "global_step": 210622, "epoch": 2366, "lr": 1.1928261955103969e-05}
{"train_loss": 0.028141433373093605, "global_step": 210623, "epoch": 2366, "lr": 1.1927886148309763e-05}
{"train_loss": 0.07767123728990555, "global_step": 210624, "epoch": 2366, "lr": 1.1927510346633908e-05}
{"train_loss": 0.08005527406930923, "global_step": 210625, "epoch": 2366, "lr": 1.192713455007642e-05}
{"train_loss": 0.041796863079071045, "global_step": 210626, "epoch": 2366, "lr": 1.1926758758637374e-05}
{"train_loss": 0.040065325796604156, "global_step": 210627, "epoch": 2366, "lr": 1.1926382972316802e-05}
{"train_loss": 0.09383319318294525, "global_step": 210628, "epoch": 2366, "lr": 1.1926007191114775e-05}
{"train_loss": 0.05499037727713585, "global_step": 210629, "epoch": 2366, "lr": 1.1925631415031319e-05}
{"train_loss": 0.04751138389110565, "global_step": 210630, "epoch": 2366, "lr": 1.1925255644066514e-05}
{"train_loss": 0.053695306181907654, "global_step": 210631, "epoch": 2366, "lr": 1.192487987822038e-05}
{"train_loss": 0.03277498111128807, "global_step": 210632, "epoch": 2366, "lr": 1.192450411749299e-05}
{"train_loss": 0.05523229390382767, "global_step": 210633, "epoch": 2366, "lr": 1.192412836188438e-05}
{"train_loss": 0.04572887718677521, "global_step": 210634, "epoch": 2366, "lr": 1.1923752611394602e-05}
{"train_loss": 0.032083529978990555, "global_step": 210635, "epoch": 2366, "lr": 1.1923376866023729e-05}
{"train_loss": 0.0424095056951046, "global_step": 210636, "epoch": 2366, "lr": 1.1923001125771777e-05}
{"train_loss": 0.11113234609365463, "global_step": 210637, "epoch": 2366, "lr": 1.1922625390638825e-05}
{"train_loss": 0.06175726652145386, "global_step": 210638, "epoch": 2366, "lr": 1.1922249660624902e-05}
{"train_loss": 0.06205349788069725, "global_step": 210639, "epoch": 2366, "lr": 1.1921873935730082e-05}
{"train_loss": 0.07819276303052902, "global_step": 210640, "epoch": 2366, "lr": 1.1921498215954386e-05}
{"train_loss": 0.027274660766124725, "global_step": 210641, "epoch": 2366, "lr": 1.1921122501297898e-05}
{"train_loss": 0.08417408913373947, "global_step": 210642, "epoch": 2366, "lr": 1.1920746791760634e-05}
{"train_loss": 0.15042586624622345, "global_step": 210643, "epoch": 2366, "lr": 1.1920371087342674e-05}
{"train_loss": 0.06512797623872757, "global_step": 210644, "epoch": 2366, "lr": 1.1919995388044047e-05}
{"train_loss": 0.025391381233930588, "global_step": 210645, "epoch": 2366, "lr": 1.1919619693864825e-05}
{"train_loss": 0.02903515286743641, "global_step": 210646, "epoch": 2366, "lr": 1.191924400480503e-05}
{"train_loss": 0.04670211300253868, "global_step": 210647, "epoch": 2366, "lr": 1.1918868320864745e-05}
{"train_loss": 0.03828120604157448, "global_step": 210648, "epoch": 2366, "lr": 1.1918492642043989e-05}
{"train_loss": 0.03262858837842941, "global_step": 210649, "epoch": 2366, "lr": 1.191811696834284e-05}
{"train_loss": 0.06407645344734192, "global_step": 210650, "epoch": 2366, "lr": 1.1917741299761321e-05}
{"train_loss": 0.05238691344857216, "global_step": 210651, "epoch": 2366, "lr": 1.1917365636299515e-05}
{"train_loss": 0.058049947023391724, "global_step": 210652, "epoch": 2366, "lr": 1.1916989977957439e-05}
{"train_loss": 0.06042730063199997, "global_step": 210653, "epoch": 2366, "lr": 1.191661432473517e-05}
{"train_loss": 0.05214795470237732, "global_step": 210654, "epoch": 2366, "lr": 1.1916238676632741e-05}
{"train_loss": 0.07318561524152756, "global_step": 210655, "epoch": 2366, "lr": 1.1915863033650221e-05}
{"train_loss": 0.08584687113761902, "global_step": 210656, "epoch": 2366, "lr": 1.1915487395787633e-05}
{"train_loss": 0.06607335060834885, "global_step": 210657, "epoch": 2366, "lr": 1.1915111763045062e-05}
{"train_loss": 0.10814826190471649, "global_step": 210658, "epoch": 2366, "lr": 1.1914736135422533e-05}
{"train_loss": 0.08980630338191986, "global_step": 210659, "epoch": 2366, "lr": 1.1914360512920098e-05}
{"train_loss": 0.0923057496547699, "global_step": 210660, "epoch": 2366, "lr": 1.1913984895537822e-05}
{"train_loss": 0.04596508666872978, "global_step": 210661, "epoch": 2366, "lr": 1.1913609283275734e-05}
{"train_loss": 0.05431244712783379, "global_step": 210662, "epoch": 2366, "lr": 1.1913233676133906e-05, "val_loss": 8.631231307983398}
{"train_loss": 0.02847442775964737, "global_step": 210663, "epoch": 2367, "lr": 1.191285807411237e-05}
{"train_loss": 0.0377788282930851, "global_step": 210664, "epoch": 2367, "lr": 1.19124824772112e-05}
{"train_loss": 0.0771588459610939, "global_step": 210665, "epoch": 2367, "lr": 1.1912106885430418e-05}
{"train_loss": 0.07372170686721802, "global_step": 210666, "epoch": 2367, "lr": 1.1911731298770108e-05}
{"train_loss": 0.047810185700654984, "global_step": 210667, "epoch": 2367, "lr": 1.1911355717230277e-05}
{"train_loss": 0.06013638153672218, "global_step": 210668, "epoch": 2367, "lr": 1.1910980140811023e-05}
{"train_loss": 0.06139369681477547, "global_step": 210669, "epoch": 2367, "lr": 1.1910604569512357e-05}
{"train_loss": 0.0607050396502018, "global_step": 210670, "epoch": 2367, "lr": 1.1910229003334344e-05}
{"train_loss": 0.06227494031190872, "global_step": 210671, "epoch": 2367, "lr": 1.1909853442277053e-05}
{"train_loss": 0.11928229033946991, "global_step": 210672, "epoch": 2367, "lr": 1.1909477886340503e-05}
{"train_loss": 0.065085768699646, "global_step": 210673, "epoch": 2367, "lr": 1.190910233552477e-05}
{"train_loss": 0.07179563492536545, "global_step": 210674, "epoch": 2367, "lr": 1.1908726789829883e-05}
{"train_loss": 0.09072189033031464, "global_step": 210675, "epoch": 2367, "lr": 1.1908351249255916e-05}
{"train_loss": 0.10663575679063797, "global_step": 210676, "epoch": 2367, "lr": 1.19079757138029e-05}
{"train_loss": 0.05239991843700409, "global_step": 210677, "epoch": 2367, "lr": 1.1907600183470897e-05}
{"train_loss": 0.03925611078739166, "global_step": 210678, "epoch": 2367, "lr": 1.1907224658259941e-05}
{"train_loss": 0.09291873872280121, "global_step": 210679, "epoch": 2367, "lr": 1.1906849138170112e-05}
{"train_loss": 0.05905858427286148, "global_step": 210680, "epoch": 2367, "lr": 1.1906473623201424e-05}
{"train_loss": 0.06665587425231934, "global_step": 210681, "epoch": 2367, "lr": 1.1906098113353965e-05}
{"train_loss": 0.11278167366981506, "global_step": 210682, "epoch": 2367, "lr": 1.190572260862775e-05}
{"train_loss": 0.06571544706821442, "global_step": 210683, "epoch": 2367, "lr": 1.1905347109022864e-05}
{"train_loss": 0.0346364825963974, "global_step": 210684, "epoch": 2367, "lr": 1.1904971614539323e-05}
{"train_loss": 0.04007047042250633, "global_step": 210685, "epoch": 2367, "lr": 1.190459612517721e-05}
{"train_loss": 0.06990276277065277, "global_step": 210686, "epoch": 2367, "lr": 1.1904220640936541e-05}
{"train_loss": 0.07667456567287445, "global_step": 210687, "epoch": 2367, "lr": 1.19038451618174e-05}
{"train_loss": 0.07470845431089401, "global_step": 210688, "epoch": 2367, "lr": 1.1903469687819812e-05}
{"train_loss": 0.021532300859689713, "global_step": 210689, "epoch": 2367, "lr": 1.1903094218943839e-05}
{"train_loss": 0.0966646745800972, "global_step": 210690, "epoch": 2367, "lr": 1.190271875518954e-05}
{"train_loss": 0.04891545698046684, "global_step": 210691, "epoch": 2367, "lr": 1.1902343296556962e-05}
{"train_loss": 0.03935559466481209, "global_step": 210692, "epoch": 2367, "lr": 1.190196784304613e-05}
{"train_loss": 0.042657624930143356, "global_step": 210693, "epoch": 2367, "lr": 1.1901592394657129e-05}
{"train_loss": 0.02837466262280941, "global_step": 210694, "epoch": 2367, "lr": 1.1901216951389982e-05}
{"train_loss": 0.048537224531173706, "global_step": 210695, "epoch": 2367, "lr": 1.1900841513244771e-05}
{"train_loss": 0.033228542655706406, "global_step": 210696, "epoch": 2367, "lr": 1.1900466080221522e-05}
{"train_loss": 0.06505583971738815, "global_step": 210697, "epoch": 2367, "lr": 1.1900090652320279e-05}
{"train_loss": 0.06048431620001793, "global_step": 210698, "epoch": 2367, "lr": 1.1899715229541114e-05}
{"train_loss": 0.04006826505064964, "global_step": 210699, "epoch": 2367, "lr": 1.1899339811884058e-05}
{"train_loss": 0.10115306824445724, "global_step": 210700, "epoch": 2367, "lr": 1.1898964399349183e-05}
{"train_loss": 0.08844194561243057, "global_step": 210701, "epoch": 2367, "lr": 1.1898588991936515e-05}
{"train_loss": 0.03809629753232002, "global_step": 210702, "epoch": 2367, "lr": 1.189821358964613e-05}
{"train_loss": 0.02726193517446518, "global_step": 210703, "epoch": 2367, "lr": 1.1897838192478056e-05}
{"train_loss": 0.07474032789468765, "global_step": 210704, "epoch": 2367, "lr": 1.1897462800432363e-05}
{"train_loss": 0.07363983243703842, "global_step": 210705, "epoch": 2367, "lr": 1.1897087413509079e-05}
{"train_loss": 0.03658835217356682, "global_step": 210706, "epoch": 2367, "lr": 1.1896712031708269e-05}
{"train_loss": 0.034006331115961075, "global_step": 210707, "epoch": 2367, "lr": 1.1896336655029994e-05}
{"train_loss": 0.03821472451090813, "global_step": 210708, "epoch": 2367, "lr": 1.1895961283474278e-05}
{"train_loss": 0.10779757797718048, "global_step": 210709, "epoch": 2367, "lr": 1.1895585917041197e-05}
{"train_loss": 0.06349899619817734, "global_step": 210710, "epoch": 2367, "lr": 1.189521055573078e-05}
{"train_loss": 0.10052546113729477, "global_step": 210711, "epoch": 2367, "lr": 1.1894835199543103e-05}
{"train_loss": 0.021246101707220078, "global_step": 210712, "epoch": 2367, "lr": 1.1894459848478179e-05}
{"train_loss": 0.05140344798564911, "global_step": 210713, "epoch": 2367, "lr": 1.18940845025361e-05}
{"train_loss": 0.03747900202870369, "global_step": 210714, "epoch": 2367, "lr": 1.1893709161716888e-05}
{"train_loss": 0.015782158821821213, "global_step": 210715, "epoch": 2367, "lr": 1.1893333826020608e-05}
{"train_loss": 0.04668516293168068, "global_step": 210716, "epoch": 2367, "lr": 1.1892958495447292e-05}
{"train_loss": 0.04901910200715065, "global_step": 210717, "epoch": 2367, "lr": 1.1892583169997018e-05}
{"train_loss": 0.10208971053361893, "global_step": 210718, "epoch": 2367, "lr": 1.1892207849669812e-05}
{"train_loss": 0.02920554019510746, "global_step": 210719, "epoch": 2367, "lr": 1.1891832534465746e-05}
{"train_loss": 0.04027509316802025, "global_step": 210720, "epoch": 2367, "lr": 1.1891457224384844e-05}
{"train_loss": 0.028569016605615616, "global_step": 210721, "epoch": 2367, "lr": 1.1891081919427183e-05}
{"train_loss": 0.035487014800310135, "global_step": 210722, "epoch": 2367, "lr": 1.1890706619592789e-05}
{"train_loss": 0.01494717225432396, "global_step": 210723, "epoch": 2367, "lr": 1.1890331324881737e-05}
{"train_loss": 0.059808481484651566, "global_step": 210724, "epoch": 2367, "lr": 1.1889956035294053e-05}
{"train_loss": 0.0937100276350975, "global_step": 210725, "epoch": 2367, "lr": 1.1889580750829804e-05}
{"train_loss": 0.061624474823474884, "global_step": 210726, "epoch": 2367, "lr": 1.1889205471489045e-05}
{"train_loss": 0.10340601205825806, "global_step": 210727, "epoch": 2367, "lr": 1.1888830197271806e-05}
{"train_loss": 0.05087388679385185, "global_step": 210728, "epoch": 2367, "lr": 1.188845492817816e-05}
{"train_loss": 0.03657088428735733, "global_step": 210729, "epoch": 2367, "lr": 1.1888079664208151e-05}
{"train_loss": 0.07684096693992615, "global_step": 210730, "epoch": 2367, "lr": 1.188770440536181e-05}
{"train_loss": 0.07767851650714874, "global_step": 210731, "epoch": 2367, "lr": 1.1887329151639221e-05}
{"train_loss": 0.015433390624821186, "global_step": 210732, "epoch": 2367, "lr": 1.1886953903040393e-05}
{"train_loss": 0.034818824380636215, "global_step": 210733, "epoch": 2367, "lr": 1.1886578659565422e-05}
{"train_loss": 0.06884263455867767, "global_step": 210734, "epoch": 2367, "lr": 1.1886203421214331e-05}
{"train_loss": 0.04164603352546692, "global_step": 210735, "epoch": 2367, "lr": 1.1885828187987164e-05}
{"train_loss": 0.05244898423552513, "global_step": 210736, "epoch": 2367, "lr": 1.188545295988399e-05}
{"train_loss": 0.04358234629034996, "global_step": 210737, "epoch": 2367, "lr": 1.1885077736904849e-05}
{"train_loss": 0.04145457595586777, "global_step": 210738, "epoch": 2367, "lr": 1.18847025190498e-05}
{"train_loss": 0.07521052658557892, "global_step": 210739, "epoch": 2367, "lr": 1.1884327306318882e-05}
{"train_loss": 0.04563684016466141, "global_step": 210740, "epoch": 2367, "lr": 1.188395209871216e-05}
{"train_loss": 0.030169231817126274, "global_step": 210741, "epoch": 2367, "lr": 1.1883576896229665e-05}
{"train_loss": 0.045916371047496796, "global_step": 210742, "epoch": 2367, "lr": 1.1883201698871455e-05}
{"train_loss": 0.06049268692731857, "global_step": 210743, "epoch": 2367, "lr": 1.1882826506637601e-05}
{"train_loss": 0.04253048077225685, "global_step": 210744, "epoch": 2367, "lr": 1.188245131952812e-05}
{"train_loss": 0.05330401286482811, "global_step": 210745, "epoch": 2367, "lr": 1.188207613754309e-05}
{"train_loss": 0.05050197243690491, "global_step": 210746, "epoch": 2367, "lr": 1.1881700960682541e-05}
{"train_loss": 0.05620725452899933, "global_step": 210747, "epoch": 2367, "lr": 1.1881325788946546e-05}
{"train_loss": 0.028956864029169083, "global_step": 210748, "epoch": 2367, "lr": 1.188095062233513e-05}
{"train_loss": 0.04099632799625397, "global_step": 210749, "epoch": 2367, "lr": 1.188057546084837e-05}
{"train_loss": 0.015982145443558693, "global_step": 210750, "epoch": 2367, "lr": 1.1880200304486283e-05}
{"train_loss": 0.05586126246855835, "global_step": 210751, "epoch": 2367, "lr": 1.1879825153248958e-05, "val_loss": 8.606359481811523}
{"train_loss": 0.03932713717222214, "global_step": 210752, "epoch": 2368, "lr": 1.1879450007136405e-05}
{"train_loss": 0.060117222368717194, "global_step": 210753, "epoch": 2368, "lr": 1.1879074866148716e-05}
{"train_loss": 0.0186154805123806, "global_step": 210754, "epoch": 2368, "lr": 1.1878699730285902e-05}
{"train_loss": 0.036347903311252594, "global_step": 210755, "epoch": 2368, "lr": 1.1878324599548052e-05}
{"train_loss": 0.046878378838300705, "global_step": 210756, "epoch": 2368, "lr": 1.1877949473935174e-05}
{"train_loss": 0.03288830816745758, "global_step": 210757, "epoch": 2368, "lr": 1.1877574353447362e-05}
{"train_loss": 0.07019894570112228, "global_step": 210758, "epoch": 2368, "lr": 1.187719923808463e-05}
{"train_loss": 0.0816858559846878, "global_step": 210759, "epoch": 2368, "lr": 1.1876824127847059e-05}
{"train_loss": 0.03338630869984627, "global_step": 210760, "epoch": 2368, "lr": 1.1876449022734665e-05}
{"train_loss": 0.07991962134838104, "global_step": 210761, "epoch": 2368, "lr": 1.187607392274752e-05}
{"train_loss": 0.0538431778550148, "global_step": 210762, "epoch": 2368, "lr": 1.187569882788569e-05}
{"train_loss": 0.06971267610788345, "global_step": 210763, "epoch": 2368, "lr": 1.187532373814919e-05}
{"train_loss": 0.07043638080358505, "global_step": 210764, "epoch": 2368, "lr": 1.1874948653538105e-05}
{"train_loss": 0.06740930676460266, "global_step": 210765, "epoch": 2368, "lr": 1.1874573574052456e-05}
{"train_loss": 0.06828298419713974, "global_step": 210766, "epoch": 2368, "lr": 1.1874198499692313e-05}
{"train_loss": 0.0541485995054245, "global_step": 210767, "epoch": 2368, "lr": 1.1873823430457725e-05}
{"train_loss": 0.08401918411254883, "global_step": 210768, "epoch": 2368, "lr": 1.1873448366348721e-05}
{"train_loss": 0.05103648453950882, "global_step": 210769, "epoch": 2368, "lr": 1.187307330736538e-05}
{"train_loss": 0.06610485166311264, "global_step": 210770, "epoch": 2368, "lr": 1.1872698253507725e-05}
{"train_loss": 0.04461490362882614, "global_step": 210771, "epoch": 2368, "lr": 1.1872323204775837e-05}
{"train_loss": 0.038780514150857925, "global_step": 210772, "epoch": 2368, "lr": 1.1871948161169738e-05}
{"train_loss": 0.04726194962859154, "global_step": 210773, "epoch": 2368, "lr": 1.18715731226895e-05}
{"train_loss": 0.0353233702480793, "global_step": 210774, "epoch": 2368, "lr": 1.1871198089335168e-05}
{"train_loss": 0.053420089185237885, "global_step": 210775, "epoch": 2368, "lr": 1.1870823061106772e-05}
{"train_loss": 0.03180242329835892, "global_step": 210776, "epoch": 2368, "lr": 1.18704480380044e-05}
{"train_loss": 0.05619499459862709, "global_step": 210777, "epoch": 2368, "lr": 1.1870073020028061e-05}
{"train_loss": 0.0682545080780983, "global_step": 210778, "epoch": 2368, "lr": 1.1869698007177842e-05}
{"train_loss": 0.040889233350753784, "global_step": 210779, "epoch": 2368, "lr": 1.1869322999453763e-05}
{"train_loss": 0.07370607554912567, "global_step": 210780, "epoch": 2368, "lr": 1.1868947996855889e-05}
{"train_loss": 0.06122437119483948, "global_step": 210781, "epoch": 2368, "lr": 1.1868572999384281e-05}
{"train_loss": 0.10433178395032883, "global_step": 210782, "epoch": 2368, "lr": 1.1868198007038966e-05}
{"train_loss": 0.043930087238550186, "global_step": 210783, "epoch": 2368, "lr": 1.1867823019820023e-05}
{"train_loss": 0.06721128523349762, "global_step": 210784, "epoch": 2368, "lr": 1.1867448037727475e-05}
{"train_loss": 0.0593000166118145, "global_step": 210785, "epoch": 2368, "lr": 1.1867073060761392e-05}
{"train_loss": 0.05059279501438141, "global_step": 210786, "epoch": 2368, "lr": 1.1866698088921807e-05}
{"train_loss": 0.055433645844459534, "global_step": 210787, "epoch": 2368, "lr": 1.1866323122208789e-05}
{"train_loss": 0.050521112978458405, "global_step": 210788, "epoch": 2368, "lr": 1.186594816062237e-05}
{"train_loss": 0.04200536757707596, "global_step": 210789, "epoch": 2368, "lr": 1.1865573204162623e-05}
{"train_loss": 0.07637452334165573, "global_step": 210790, "epoch": 2368, "lr": 1.1865198252829567e-05}
{"train_loss": 0.052048467099666595, "global_step": 210791, "epoch": 2368, "lr": 1.1864823306623286e-05}
{"train_loss": 0.037481263279914856, "global_step": 210792, "epoch": 2368, "lr": 1.1864448365543801e-05}
{"train_loss": 0.08595834672451019, "global_step": 210793, "epoch": 2368, "lr": 1.1864073429591189e-05}
{"train_loss": 0.04317367076873779, "global_step": 210794, "epoch": 2368, "lr": 1.1863698498765474e-05}
{"train_loss": 0.07434304058551788, "global_step": 210795, "epoch": 2368, "lr": 1.1863323573066737e-05}
{"train_loss": 0.0959230437874794, "global_step": 210796, "epoch": 2368, "lr": 1.1862948652494993e-05}
{"train_loss": 0.0851578563451767, "global_step": 210797, "epoch": 2368, "lr": 1.1862573737050315e-05}
{"train_loss": 0.030095728114247322, "global_step": 210798, "epoch": 2368, "lr": 1.1862198826732757e-05}
{"train_loss": 0.06894121319055557, "global_step": 210799, "epoch": 2368, "lr": 1.1861823921542353e-05}
{"train_loss": 0.034043215215206146, "global_step": 210800, "epoch": 2368, "lr": 1.186144902147917e-05}
{"train_loss": 0.0449567511677742, "global_step": 210801, "epoch": 2368, "lr": 1.1861074126543242e-05}
{"train_loss": 0.033435821533203125, "global_step": 210802, "epoch": 2368, "lr": 1.186069923673464e-05}
{"train_loss": 0.04907418042421341, "global_step": 210803, "epoch": 2368, "lr": 1.1860324352053393e-05}
{"train_loss": 0.03366146981716156, "global_step": 210804, "epoch": 2368, "lr": 1.1859949472499564e-05}
{"train_loss": 0.05372925475239754, "global_step": 210805, "epoch": 2368, "lr": 1.1859574598073198e-05}
{"train_loss": 0.03124314174056053, "global_step": 210806, "epoch": 2368, "lr": 1.1859199728774351e-05}
{"train_loss": 0.024518344551324844, "global_step": 210807, "epoch": 2368, "lr": 1.1858824864603075e-05}
{"train_loss": 0.038202740252017975, "global_step": 210808, "epoch": 2368, "lr": 1.1858450005559402e-05}
{"train_loss": 0.05395622178912163, "global_step": 210809, "epoch": 2368, "lr": 1.1858075151643405e-05}
{"train_loss": 0.046401247382164, "global_step": 210810, "epoch": 2368, "lr": 1.1857700302855112e-05}
{"train_loss": 0.0812886655330658, "global_step": 210811, "epoch": 2368, "lr": 1.1857325459194602e-05}
{"train_loss": 0.05440706014633179, "global_step": 210812, "epoch": 2368, "lr": 1.1856950620661912e-05}
{"train_loss": 0.052169330418109894, "global_step": 210813, "epoch": 2368, "lr": 1.1856575787257074e-05}
{"train_loss": 0.05716836079955101, "global_step": 210814, "epoch": 2368, "lr": 1.1856200958980167e-05}
{"train_loss": 0.036423880606889725, "global_step": 210815, "epoch": 2368, "lr": 1.1855826135831217e-05}
{"train_loss": 0.028581641614437103, "global_step": 210816, "epoch": 2368, "lr": 1.1855451317810285e-05}
{"train_loss": 0.03342900052666664, "global_step": 210817, "epoch": 2368, "lr": 1.1855076504917435e-05}
{"train_loss": 0.04942991957068443, "global_step": 210818, "epoch": 2368, "lr": 1.1854701697152693e-05}
{"train_loss": 0.013339039869606495, "global_step": 210819, "epoch": 2368, "lr": 1.1854326894516138e-05}
{"train_loss": 0.0836872011423111, "global_step": 210820, "epoch": 2368, "lr": 1.1853952097007787e-05}
{"train_loss": 0.040726833045482635, "global_step": 210821, "epoch": 2368, "lr": 1.1853577304627717e-05}
{"train_loss": 0.07625868171453476, "global_step": 210822, "epoch": 2368, "lr": 1.1853202517375961e-05}
{"train_loss": 0.0357118621468544, "global_step": 210823, "epoch": 2368, "lr": 1.1852827735252586e-05}
{"train_loss": 0.0530528798699379, "global_step": 210824, "epoch": 2368, "lr": 1.1852452958257625e-05}
{"train_loss": 0.07522343844175339, "global_step": 210825, "epoch": 2368, "lr": 1.1852078186391146e-05}
{"train_loss": 0.05924012139439583, "global_step": 210826, "epoch": 2368, "lr": 1.1851703419653182e-05}
{"train_loss": 0.055532678961753845, "global_step": 210827, "epoch": 2368, "lr": 1.1851328658043797e-05}
{"train_loss": 0.04363628476858139, "global_step": 210828, "epoch": 2368, "lr": 1.1850953901563028e-05}
{"train_loss": 0.06595239788293839, "global_step": 210829, "epoch": 2368, "lr": 1.185057915021095e-05}
{"train_loss": 0.030303554609417915, "global_step": 210830, "epoch": 2368, "lr": 1.1850204403987575e-05}
{"train_loss": 0.05031050741672516, "global_step": 210831, "epoch": 2368, "lr": 1.1849829662892991e-05}
{"train_loss": 0.05235026404261589, "global_step": 210832, "epoch": 2368, "lr": 1.1849454926927222e-05}
{"train_loss": 0.1457318216562271, "global_step": 210833, "epoch": 2368, "lr": 1.1849080196090328e-05}
{"train_loss": 0.06342010200023651, "global_step": 210834, "epoch": 2368, "lr": 1.184870547038237e-05}
{"train_loss": 0.014423551969230175, "global_step": 210835, "epoch": 2368, "lr": 1.1848330749803377e-05}
{"train_loss": 0.07135464251041412, "global_step": 210836, "epoch": 2368, "lr": 1.1847956034353424e-05}
{"train_loss": 0.04533668980002403, "global_step": 210837, "epoch": 2368, "lr": 1.1847581324032537e-05}
{"train_loss": 0.08597251027822495, "global_step": 210838, "epoch": 2368, "lr": 1.1847206618840794e-05}
{"train_loss": 0.056419737637043, "global_step": 210839, "epoch": 2368, "lr": 1.1846831918778211e-05}
{"train_loss": 0.05432338183850385, "global_step": 210840, "epoch": 2368, "lr": 1.184645722384487e-05, "val_loss": 8.617769241333008}
{"train_loss": 0.057323165237903595, "global_step": 210841, "epoch": 2369, "lr": 1.1846082534040793e-05}
{"train_loss": 0.06957115232944489, "global_step": 210842, "epoch": 2369, "lr": 1.184570784936606e-05}
{"train_loss": 0.05423545837402344, "global_step": 210843, "epoch": 2369, "lr": 1.1845333169820694e-05}
{"train_loss": 0.046912044286727905, "global_step": 210844, "epoch": 2369, "lr": 1.1844958495404772e-05}
{"train_loss": 0.01999560557305813, "global_step": 210845, "epoch": 2369, "lr": 1.1844583826118327e-05}
{"train_loss": 0.022451108321547508, "global_step": 210846, "epoch": 2369, "lr": 1.1844209161961406e-05}
{"train_loss": 0.05390504375100136, "global_step": 210847, "epoch": 2369, "lr": 1.1843834502934076e-05}
{"train_loss": 0.027169400826096535, "global_step": 210848, "epoch": 2369, "lr": 1.1843459849036359e-05}
{"train_loss": 0.0654997006058693, "global_step": 210849, "epoch": 2369, "lr": 1.1843085200268344e-05}
{"train_loss": 0.03137318417429924, "global_step": 210850, "epoch": 2369, "lr": 1.1842710556630054e-05}
{"train_loss": 0.012261324562132359, "global_step": 210851, "epoch": 2369, "lr": 1.1842335918121538e-05}
{"train_loss": 0.027510305866599083, "global_step": 210852, "epoch": 2369, "lr": 1.1841961284742852e-05}
{"train_loss": 0.047347333282232285, "global_step": 210853, "epoch": 2369, "lr": 1.1841586656494069e-05}
{"train_loss": 0.056919049471616745, "global_step": 210854, "epoch": 2369, "lr": 1.18412120333752e-05}
{"train_loss": 0.11177786439657211, "global_step": 210855, "epoch": 2369, "lr": 1.184083741538633e-05}
{"train_loss": 0.043169230222702026, "global_step": 210856, "epoch": 2369, "lr": 1.184046280252748e-05}
{"train_loss": 0.04927825182676315, "global_step": 210857, "epoch": 2369, "lr": 1.1840088194798732e-05}
{"train_loss": 0.09096986055374146, "global_step": 210858, "epoch": 2369, "lr": 1.1839713592200102e-05}
{"train_loss": 0.07616124302148819, "global_step": 210859, "epoch": 2369, "lr": 1.1839338994731674e-05}
{"train_loss": 0.04332941025495529, "global_step": 210860, "epoch": 2369, "lr": 1.1838964402393465e-05}
{"train_loss": 0.05984468013048172, "global_step": 210861, "epoch": 2369, "lr": 1.1838589815185557e-05}
{"train_loss": 0.1533307135105133, "global_step": 210862, "epoch": 2369, "lr": 1.1838215233107969e-05}
{"train_loss": 0.07062803208827972, "global_step": 210863, "epoch": 2369, "lr": 1.1837840656160788e-05}
{"train_loss": 0.04732075333595276, "global_step": 210864, "epoch": 2369, "lr": 1.1837466084344024e-05}
{"train_loss": 0.08777663856744766, "global_step": 210865, "epoch": 2369, "lr": 1.1837091517657766e-05}
{"train_loss": 0.08001695573329926, "global_step": 210866, "epoch": 2369, "lr": 1.1836716956102029e-05}
{"train_loss": 0.04243433102965355, "global_step": 210867, "epoch": 2369, "lr": 1.1836342399676893e-05}
{"train_loss": 0.037732917815446854, "global_step": 210868, "epoch": 2369, "lr": 1.1835967848382384e-05}
{"train_loss": 0.03855067491531372, "global_step": 210869, "epoch": 2369, "lr": 1.1835593302218567e-05}
{"train_loss": 0.038674190640449524, "global_step": 210870, "epoch": 2369, "lr": 1.1835218761185496e-05}
{"train_loss": 0.06690052151679993, "global_step": 210871, "epoch": 2369, "lr": 1.1834844225283204e-05}
{"train_loss": 0.05969598889350891, "global_step": 210872, "epoch": 2369, "lr": 1.1834469694511768e-05}
{"train_loss": 0.05399012938141823, "global_step": 210873, "epoch": 2369, "lr": 1.1834095168871206e-05}
{"train_loss": 0.0721190869808197, "global_step": 210874, "epoch": 2369, "lr": 1.1833720648361601e-05}
{"train_loss": 0.12363824248313904, "global_step": 210875, "epoch": 2369, "lr": 1.1833346132982975e-05}
{"train_loss": 0.13001810014247894, "global_step": 210876, "epoch": 2369, "lr": 1.18329716227354e-05}
{"train_loss": 0.0674089789390564, "global_step": 210877, "epoch": 2369, "lr": 1.1832597117618905e-05}
{"train_loss": 0.04998782277107239, "global_step": 210878, "epoch": 2369, "lr": 1.1832222617633566e-05}
{"train_loss": 0.06007326766848564, "global_step": 210879, "epoch": 2369, "lr": 1.1831848122779405e-05}
{"train_loss": 0.06357330083847046, "global_step": 210880, "epoch": 2369, "lr": 1.18314736330565e-05}
{"train_loss": 0.1266918033361435, "global_step": 210881, "epoch": 2369, "lr": 1.1831099148464875e-05}
{"train_loss": 0.08072012662887573, "global_step": 210882, "epoch": 2369, "lr": 1.1830724669004607e-05}
{"train_loss": 0.08639127761125565, "global_step": 210883, "epoch": 2369, "lr": 1.1830350194675716e-05}
{"train_loss": 0.07856888324022293, "global_step": 210884, "epoch": 2369, "lr": 1.1829975725478287e-05}
{"train_loss": 0.05519035458564758, "global_step": 210885, "epoch": 2369, "lr": 1.1829601261412354e-05}
{"train_loss": 0.059752222150564194, "global_step": 210886, "epoch": 2369, "lr": 1.182922680247795e-05}
{"train_loss": 0.03339942917227745, "global_step": 210887, "epoch": 2369, "lr": 1.1828852348675151e-05}
{"train_loss": 0.03572464361786842, "global_step": 210888, "epoch": 2369, "lr": 1.1828477900003986e-05}
{"train_loss": 0.0827414020895958, "global_step": 210889, "epoch": 2369, "lr": 1.1828103456464535e-05}
{"train_loss": 0.0812152847647667, "global_step": 210890, "epoch": 2369, "lr": 1.1827729018056816e-05}
{"train_loss": 0.053809478878974915, "global_step": 210891, "epoch": 2369, "lr": 1.1827354584780903e-05}
{"train_loss": 0.09367148578166962, "global_step": 210892, "epoch": 2369, "lr": 1.1826980156636825e-05}
{"train_loss": 0.09407836198806763, "global_step": 210893, "epoch": 2369, "lr": 1.1826605733624658e-05}
{"train_loss": 0.04868980869650841, "global_step": 210894, "epoch": 2369, "lr": 1.1826231315744423e-05}
{"train_loss": 0.07813486456871033, "global_step": 210895, "epoch": 2369, "lr": 1.1825856902996202e-05}
{"train_loss": 0.09805430471897125, "global_step": 210896, "epoch": 2369, "lr": 1.1825482495380019e-05}
{"train_loss": 0.094412662088871, "global_step": 210897, "epoch": 2369, "lr": 1.1825108092895943e-05}
{"train_loss": 0.059704408049583435, "global_step": 210898, "epoch": 2369, "lr": 1.1824733695544004e-05}
{"train_loss": 0.08982309699058533, "global_step": 210899, "epoch": 2369, "lr": 1.1824359303324278e-05}
{"train_loss": 0.07380834966897964, "global_step": 210900, "epoch": 2369, "lr": 1.1823984916236786e-05}
{"train_loss": 0.04640020430088043, "global_step": 210901, "epoch": 2369, "lr": 1.1823610534281609e-05}
{"train_loss": 0.11787031590938568, "global_step": 210902, "epoch": 2369, "lr": 1.1823236157458767e-05}
{"train_loss": 0.058403380215168, "global_step": 210903, "epoch": 2369, "lr": 1.1822861785768347e-05}
{"train_loss": 0.08925024420022964, "global_step": 210904, "epoch": 2369, "lr": 1.1822487419210355e-05}
{"train_loss": 0.09009970724582672, "global_step": 210905, "epoch": 2369, "lr": 1.1822113057784885e-05}
{"train_loss": 0.042917314916849136, "global_step": 210906, "epoch": 2369, "lr": 1.1821738701491946e-05}
{"train_loss": 0.05061552673578262, "global_step": 210907, "epoch": 2369, "lr": 1.1821364350331609e-05}
{"train_loss": 0.04626667872071266, "global_step": 210908, "epoch": 2369, "lr": 1.1820990004303944e-05}
{"train_loss": 0.07231555134057999, "global_step": 210909, "epoch": 2369, "lr": 1.1820615663408963e-05}
{"train_loss": 0.06366454064846039, "global_step": 210910, "epoch": 2369, "lr": 1.1820241327646753e-05}
{"train_loss": 0.09759894758462906, "global_step": 210911, "epoch": 2369, "lr": 1.1819866997017326e-05}
{"train_loss": 0.025880249217152596, "global_step": 210912, "epoch": 2369, "lr": 1.181949267152077e-05}
{"train_loss": 0.04498881474137306, "global_step": 210913, "epoch": 2369, "lr": 1.1819118351157105e-05}
{"train_loss": 0.05059705302119255, "global_step": 210914, "epoch": 2369, "lr": 1.181874403592641e-05}
{"train_loss": 0.05920124053955078, "global_step": 210915, "epoch": 2369, "lr": 1.1818369725828704e-05}
{"train_loss": 0.035750314593315125, "global_step": 210916, "epoch": 2369, "lr": 1.1817995420864063e-05}
{"train_loss": 0.08429483324289322, "global_step": 210917, "epoch": 2369, "lr": 1.1817621121032518e-05}
{"train_loss": 0.09842006862163544, "global_step": 210918, "epoch": 2369, "lr": 1.1817246826334139e-05}
{"train_loss": 0.03128047659993172, "global_step": 210919, "epoch": 2369, "lr": 1.1816872536768953e-05}
{"train_loss": 0.03548470139503479, "global_step": 210920, "epoch": 2369, "lr": 1.1816498252337038e-05}
{"train_loss": 0.03058578260242939, "global_step": 210921, "epoch": 2369, "lr": 1.1816123973038412e-05}
{"train_loss": 0.06307727098464966, "global_step": 210922, "epoch": 2369, "lr": 1.1815749698873163e-05}
{"train_loss": 0.10230764746665955, "global_step": 210923, "epoch": 2369, "lr": 1.1815375429841313e-05}
{"train_loss": 0.056042544543743134, "global_step": 210924, "epoch": 2369, "lr": 1.1815001165942913e-05}
{"train_loss": 0.059540797024965286, "global_step": 210925, "epoch": 2369, "lr": 1.1814626907178034e-05}
{"train_loss": 0.06072603911161423, "global_step": 210926, "epoch": 2369, "lr": 1.1814252653546699e-05}
{"train_loss": 0.10166486352682114, "global_step": 210927, "epoch": 2369, "lr": 1.1813878405048984e-05}
{"train_loss": 0.06314461678266525, "global_step": 210928, "epoch": 2369, "lr": 1.1813504161684912e-05}
{"train_loss": 0.06447823451350579, "global_step": 210929, "epoch": 2369, "lr": 1.1813129923454563e-05, "val_loss": 8.597267150878906}
{"train_loss": 0.051141392439603806, "global_step": 210930, "epoch": 2370, "lr": 1.1812755690357963e-05}
{"train_loss": 0.0655178353190422, "global_step": 210931, "epoch": 2370, "lr": 1.1812381462395183e-05}
{"train_loss": 0.08917933702468872, "global_step": 210932, "epoch": 2370, "lr": 1.1812007239566253e-05}
{"train_loss": 0.028705846518278122, "global_step": 210933, "epoch": 2370, "lr": 1.1811633021871244e-05}
{"train_loss": 0.030739177018404007, "global_step": 210934, "epoch": 2370, "lr": 1.1811258809310182e-05}
{"train_loss": 0.04262692853808403, "global_step": 210935, "epoch": 2370, "lr": 1.1810884601883149e-05}
{"train_loss": 0.07414425909519196, "global_step": 210936, "epoch": 2370, "lr": 1.1810510399590158e-05}
{"train_loss": 0.08198706060647964, "global_step": 210937, "epoch": 2370, "lr": 1.1810136202431294e-05}
{"train_loss": 0.06270679831504822, "global_step": 210938, "epoch": 2370, "lr": 1.1809762010406578e-05}
{"train_loss": 0.027864452451467514, "global_step": 210939, "epoch": 2370, "lr": 1.1809387823516087e-05}
{"train_loss": 0.07322068512439728, "global_step": 210940, "epoch": 2370, "lr": 1.1809013641759848e-05}
{"train_loss": 0.0625821053981781, "global_step": 210941, "epoch": 2370, "lr": 1.1808639465137927e-05}
{"train_loss": 0.08040109276771545, "global_step": 210942, "epoch": 2370, "lr": 1.180826529365036e-05}
{"train_loss": 0.07587619125843048, "global_step": 210943, "epoch": 2370, "lr": 1.180789112729721e-05}
{"train_loss": 0.08878489583730698, "global_step": 210944, "epoch": 2370, "lr": 1.1807516966078536e-05}
{"train_loss": 0.0267450213432312, "global_step": 210945, "epoch": 2370, "lr": 1.1807142809994359e-05}
{"train_loss": 0.06446996331214905, "global_step": 210946, "epoch": 2370, "lr": 1.180676865904476e-05}
{"train_loss": 0.05367144197225571, "global_step": 210947, "epoch": 2370, "lr": 1.1806394513229763e-05}
{"train_loss": 0.057769790291786194, "global_step": 210948, "epoch": 2370, "lr": 1.1806020372549447e-05}
{"train_loss": 0.08926807343959808, "global_step": 210949, "epoch": 2370, "lr": 1.1805646237003825e-05}
{"train_loss": 0.05546743422746658, "global_step": 210950, "epoch": 2370, "lr": 1.1805272106592991e-05}
{"train_loss": 0.06490852683782578, "global_step": 210951, "epoch": 2370, "lr": 1.1804897981316954e-05}
{"train_loss": 0.07508283853530884, "global_step": 210952, "epoch": 2370, "lr": 1.1804523861175799e-05}
{"train_loss": 0.09831298887729645, "global_step": 210953, "epoch": 2370, "lr": 1.1804149746169541e-05}
{"train_loss": 0.03287459537386894, "global_step": 210954, "epoch": 2370, "lr": 1.180377563629827e-05}
{"train_loss": 0.06775996088981628, "global_step": 210955, "epoch": 2370, "lr": 1.1803401531561998e-05}
{"train_loss": 0.06578255444765091, "global_step": 210956, "epoch": 2370, "lr": 1.1803027431960805e-05}
{"train_loss": 0.049400147050619125, "global_step": 210957, "epoch": 2370, "lr": 1.1802653337494724e-05}
{"train_loss": 0.04198994114995003, "global_step": 210958, "epoch": 2370, "lr": 1.1802279248163816e-05}
{"train_loss": 0.048591237515211105, "global_step": 210959, "epoch": 2370, "lr": 1.1801905163968113e-05}
{"train_loss": 0.0636894628405571, "global_step": 210960, "epoch": 2370, "lr": 1.1801531084907696e-05}
{"train_loss": 0.13540485501289368, "global_step": 210961, "epoch": 2370, "lr": 1.18011570109826e-05}
{"train_loss": 0.0412047803401947, "global_step": 210962, "epoch": 2370, "lr": 1.180078294219285e-05}
{"train_loss": 0.05064774677157402, "global_step": 210963, "epoch": 2370, "lr": 1.1800408878538544e-05}
{"train_loss": 0.057825952768325806, "global_step": 210964, "epoch": 2370, "lr": 1.1800034820019684e-05}
{"train_loss": 0.04939696937799454, "global_step": 210965, "epoch": 2370, "lr": 1.1799660766636362e-05}
{"train_loss": 0.07696563005447388, "global_step": 210966, "epoch": 2370, "lr": 1.1799286718388598e-05}
{"train_loss": 0.07231387495994568, "global_step": 210967, "epoch": 2370, "lr": 1.1798912675276464e-05}
{"train_loss": 0.053829874843358994, "global_step": 210968, "epoch": 2370, "lr": 1.1798538637299988e-05}
{"train_loss": 0.06609748303890228, "global_step": 210969, "epoch": 2370, "lr": 1.1798164604459243e-05}
{"train_loss": 0.06528738886117935, "global_step": 210970, "epoch": 2370, "lr": 1.1797790576754265e-05}
{"train_loss": 0.047484252601861954, "global_step": 210971, "epoch": 2370, "lr": 1.1797416554185114e-05}
{"train_loss": 0.03827473148703575, "global_step": 210972, "epoch": 2370, "lr": 1.1797042536751823e-05}
{"train_loss": 0.0441448912024498, "global_step": 210973, "epoch": 2370, "lr": 1.1796668524454469e-05}
{"train_loss": 0.08317451179027557, "global_step": 210974, "epoch": 2370, "lr": 1.1796294517293071e-05}
{"train_loss": 0.053615156561136246, "global_step": 210975, "epoch": 2370, "lr": 1.179592051526771e-05}
{"train_loss": 0.06521633267402649, "global_step": 210976, "epoch": 2370, "lr": 1.1795546518378409e-05}
{"train_loss": 0.05175260081887245, "global_step": 210977, "epoch": 2370, "lr": 1.1795172526625247e-05}
{"train_loss": 0.0370892770588398, "global_step": 210978, "epoch": 2370, "lr": 1.1794798540008245e-05}
{"train_loss": 0.05551053211092949, "global_step": 210979, "epoch": 2370, "lr": 1.1794424558527461e-05}
{"train_loss": 0.03770244121551514, "global_step": 210980, "epoch": 2370, "lr": 1.179405058218297e-05}
{"train_loss": 0.07877495884895325, "global_step": 210981, "epoch": 2370, "lr": 1.1793676610974785e-05}
{"train_loss": 0.05407930910587311, "global_step": 210982, "epoch": 2370, "lr": 1.1793302644902993e-05}
{"train_loss": 0.05450083687901497, "global_step": 210983, "epoch": 2370, "lr": 1.1792928683967607e-05}
{"train_loss": 0.1071566492319107, "global_step": 210984, "epoch": 2370, "lr": 1.1792554728168714e-05}
{"train_loss": 0.07843794673681259, "global_step": 210985, "epoch": 2370, "lr": 1.1792180777506334e-05}
{"train_loss": 0.07132933288812637, "global_step": 210986, "epoch": 2370, "lr": 1.1791806831980539e-05}
{"train_loss": 0.055828310549259186, "global_step": 210987, "epoch": 2370, "lr": 1.1791432891591359e-05}
{"train_loss": 0.034396931529045105, "global_step": 210988, "epoch": 2370, "lr": 1.1791058956338868e-05}
{"train_loss": 0.06318694353103638, "global_step": 210989, "epoch": 2370, "lr": 1.1790685026223086e-05}
{"train_loss": 0.020241176709532738, "global_step": 210990, "epoch": 2370, "lr": 1.17903111012441e-05}
{"train_loss": 0.04953896626830101, "global_step": 210991, "epoch": 2370, "lr": 1.1789937181401923e-05}
{"train_loss": 0.038130275905132294, "global_step": 210992, "epoch": 2370, "lr": 1.178956326669664e-05}
{"train_loss": 0.059214502573013306, "global_step": 210993, "epoch": 2370, "lr": 1.1789189357128271e-05}
{"train_loss": 0.033892352133989334, "global_step": 210994, "epoch": 2370, "lr": 1.1788815452696895e-05}
{"train_loss": 0.11016958951950073, "global_step": 210995, "epoch": 2370, "lr": 1.178844155340253e-05}
{"train_loss": 0.03889377787709236, "global_step": 210996, "epoch": 2370, "lr": 1.1788067659245256e-05}
{"train_loss": 0.03791537880897522, "global_step": 210997, "epoch": 2370, "lr": 1.1787693770225094e-05}
{"train_loss": 0.04701361805200577, "global_step": 210998, "epoch": 2370, "lr": 1.178731988634213e-05}
{"train_loss": 0.015406893566250801, "global_step": 210999, "epoch": 2370, "lr": 1.1786946007596383e-05}
{"train_loss": 0.04810408502817154, "global_step": 211000, "epoch": 2370, "lr": 1.1786572133987927e-05}
{"train_loss": 0.05051451548933983, "global_step": 211001, "epoch": 2370, "lr": 1.1786198265516801e-05}
{"train_loss": 0.007941946387290955, "global_step": 211002, "epoch": 2370, "lr": 1.1785824402183038e-05}
{"train_loss": 0.05842147767543793, "global_step": 211003, "epoch": 2370, "lr": 1.178545054398672e-05}
{"train_loss": 0.04022783786058426, "global_step": 211004, "epoch": 2370, "lr": 1.1785076690927871e-05}
{"train_loss": 0.03951512277126312, "global_step": 211005, "epoch": 2370, "lr": 1.1784702843006567e-05}
{"train_loss": 0.08497074246406555, "global_step": 211006, "epoch": 2370, "lr": 1.1784329000222827e-05}
{"train_loss": 0.04495953768491745, "global_step": 211007, "epoch": 2370, "lr": 1.1783955162576732e-05}
{"train_loss": 0.03742407634854317, "global_step": 211008, "epoch": 2370, "lr": 1.1783581330068306e-05}
{"train_loss": 0.04764512553811073, "global_step": 211009, "epoch": 2370, "lr": 1.1783207502697624e-05}
{"train_loss": 0.07293929904699326, "global_step": 211010, "epoch": 2370, "lr": 1.1782833680464711e-05}
{"train_loss": 0.05833570286631584, "global_step": 211011, "epoch": 2370, "lr": 1.1782459863369649e-05}
{"train_loss": 0.05157152935862541, "global_step": 211012, "epoch": 2370, "lr": 1.178208605141245e-05}
{"train_loss": 0.03605519235134125, "global_step": 211013, "epoch": 2370, "lr": 1.1781712244593197e-05}
{"train_loss": 0.07670148462057114, "global_step": 211014, "epoch": 2370, "lr": 1.1781338442911916e-05}
{"train_loss": 0.0963122546672821, "global_step": 211015, "epoch": 2370, "lr": 1.1780964646368665e-05}
{"train_loss": 0.03361950442194939, "global_step": 211016, "epoch": 2370, "lr": 1.1780590854963513e-05}
{"train_loss": 0.0753588080406189, "global_step": 211017, "epoch": 2370, "lr": 1.178021706869648e-05}
{"train_loss": 0.05751996604579218, "global_step": 211018, "epoch": 2370, "lr": 1.177984328756765e-05, "val_loss": 8.590022087097168, "train_action_mse_error": 24.340747833251953}
{"train_loss": 0.08688516914844513, "global_step": 211019, "epoch": 2371, "lr": 1.1779469511577034e-05}
{"train_loss": 0.02905985899269581, "global_step": 211020, "epoch": 2371, "lr": 1.1779095740724722e-05}
{"train_loss": 0.015642616897821426, "global_step": 211021, "epoch": 2371, "lr": 1.1778721975010725e-05}
{"train_loss": 0.04470048099756241, "global_step": 211022, "epoch": 2371, "lr": 1.177834821443513e-05}
{"train_loss": 0.07734457403421402, "global_step": 211023, "epoch": 2371, "lr": 1.1777974458997959e-05}
{"train_loss": 0.04294745624065399, "global_step": 211024, "epoch": 2371, "lr": 1.1777600708699283e-05}
{"train_loss": 0.08505343645811081, "global_step": 211025, "epoch": 2371, "lr": 1.1777226963539128e-05}
{"train_loss": 0.04362772777676582, "global_step": 211026, "epoch": 2371, "lr": 1.1776853223517575e-05}
{"train_loss": 0.06244527921080589, "global_step": 211027, "epoch": 2371, "lr": 1.1776479488634644e-05}
{"train_loss": 0.04648105055093765, "global_step": 211028, "epoch": 2371, "lr": 1.1776105758890415e-05}
{"train_loss": 0.05372002348303795, "global_step": 211029, "epoch": 2371, "lr": 1.177573203428491e-05}
{"train_loss": 0.08589700609445572, "global_step": 211030, "epoch": 2371, "lr": 1.1775358314818202e-05}
{"train_loss": 0.04323585703969002, "global_step": 211031, "epoch": 2371, "lr": 1.177498460049032e-05}
{"train_loss": 0.07340148091316223, "global_step": 211032, "epoch": 2371, "lr": 1.1774610891301335e-05}
{"train_loss": 0.05891622230410576, "global_step": 211033, "epoch": 2371, "lr": 1.1774237187251274e-05}
{"train_loss": 0.03340385854244232, "global_step": 211034, "epoch": 2371, "lr": 1.177386348834022e-05}
{"train_loss": 0.03926598280668259, "global_step": 211035, "epoch": 2371, "lr": 1.1773489794568187e-05}
{"train_loss": 0.09725232422351837, "global_step": 211036, "epoch": 2371, "lr": 1.1773116105935256e-05}
{"train_loss": 0.06560305505990982, "global_step": 211037, "epoch": 2371, "lr": 1.1772742422441451e-05}
{"train_loss": 0.040096063166856766, "global_step": 211038, "epoch": 2371, "lr": 1.1772368744086848e-05}
{"train_loss": 0.09594841301441193, "global_step": 211039, "epoch": 2371, "lr": 1.1771995070871489e-05}
{"train_loss": 0.07263250648975372, "global_step": 211040, "epoch": 2371, "lr": 1.1771621402795397e-05}
{"train_loss": 0.07944756001234055, "global_step": 211041, "epoch": 2371, "lr": 1.1771247739858665e-05}
{"train_loss": 0.04451984167098999, "global_step": 211042, "epoch": 2371, "lr": 1.1770874082061306e-05}
{"train_loss": 0.07125439494848251, "global_step": 211043, "epoch": 2371, "lr": 1.1770500429403402e-05}
{"train_loss": 0.050256725400686264, "global_step": 211044, "epoch": 2371, "lr": 1.1770126781884972e-05}
{"train_loss": 0.052879564464092255, "global_step": 211045, "epoch": 2371, "lr": 1.1769753139506101e-05}
{"train_loss": 0.0379309356212616, "global_step": 211046, "epoch": 2371, "lr": 1.1769379502266803e-05}
{"train_loss": 0.07924868166446686, "global_step": 211047, "epoch": 2371, "lr": 1.1769005870167161e-05}
{"train_loss": 0.06456498801708221, "global_step": 211048, "epoch": 2371, "lr": 1.1768632243207195e-05}
{"train_loss": 0.06337916105985641, "global_step": 211049, "epoch": 2371, "lr": 1.1768258621386986e-05}
{"train_loss": 0.0433952733874321, "global_step": 211050, "epoch": 2371, "lr": 1.1767885004706552e-05}
{"train_loss": 0.03485485538840294, "global_step": 211051, "epoch": 2371, "lr": 1.1767511393165964e-05}
{"train_loss": 0.0467425100505352, "global_step": 211052, "epoch": 2371, "lr": 1.1767137786765282e-05}
{"train_loss": 0.05458100140094757, "global_step": 211053, "epoch": 2371, "lr": 1.1766764185504526e-05}
{"train_loss": 0.03725913166999817, "global_step": 211054, "epoch": 2371, "lr": 1.1766390589383774e-05}
{"train_loss": 0.021300507709383965, "global_step": 211055, "epoch": 2371, "lr": 1.1766016998403052e-05}
{"train_loss": 0.053298454731702805, "global_step": 211056, "epoch": 2371, "lr": 1.1765643412562437e-05}
{"train_loss": 0.03500836342573166, "global_step": 211057, "epoch": 2371, "lr": 1.1765269831861958e-05}
{"train_loss": 0.06049920246005058, "global_step": 211058, "epoch": 2371, "lr": 1.1764896256301677e-05}
{"train_loss": 0.05869784578680992, "global_step": 211059, "epoch": 2371, "lr": 1.1764522685881635e-05}
{"train_loss": 0.04196331650018692, "global_step": 211060, "epoch": 2371, "lr": 1.1764149120601891e-05}
{"train_loss": 0.053081244230270386, "global_step": 211061, "epoch": 2371, "lr": 1.1763775560462486e-05}
{"train_loss": 0.049577366560697556, "global_step": 211062, "epoch": 2371, "lr": 1.1763402005463487e-05}
{"train_loss": 0.049225568771362305, "global_step": 211063, "epoch": 2371, "lr": 1.1763028455604913e-05}
{"train_loss": 0.09453663229942322, "global_step": 211064, "epoch": 2371, "lr": 1.176265491088685e-05}
{"train_loss": 0.041839636862277985, "global_step": 211065, "epoch": 2371, "lr": 1.1762281371309326e-05}
{"train_loss": 0.08454529196023941, "global_step": 211066, "epoch": 2371, "lr": 1.1761907836872398e-05}
{"train_loss": 0.1112496629357338, "global_step": 211067, "epoch": 2371, "lr": 1.176153430757611e-05}
{"train_loss": 0.05099308490753174, "global_step": 211068, "epoch": 2371, "lr": 1.1761160783420533e-05}
{"train_loss": 0.03590529039502144, "global_step": 211069, "epoch": 2371, "lr": 1.1760787264405682e-05}
{"train_loss": 0.038103096187114716, "global_step": 211070, "epoch": 2371, "lr": 1.176041375053163e-05}
{"train_loss": 0.07274596393108368, "global_step": 211071, "epoch": 2371, "lr": 1.1760040241798442e-05}
{"train_loss": 0.0639517679810524, "global_step": 211072, "epoch": 2371, "lr": 1.1759666738206143e-05}
{"train_loss": 0.05905446410179138, "global_step": 211073, "epoch": 2371, "lr": 1.1759293239754782e-05}
{"train_loss": 0.033060915768146515, "global_step": 211074, "epoch": 2371, "lr": 1.175891974644443e-05}
{"train_loss": 0.09381908923387527, "global_step": 211075, "epoch": 2371, "lr": 1.1758546258275111e-05}
{"train_loss": 0.07224734127521515, "global_step": 211076, "epoch": 2371, "lr": 1.1758172775246905e-05}
{"train_loss": 0.054219815880060196, "global_step": 211077, "epoch": 2371, "lr": 1.1757799297359846e-05}
{"train_loss": 0.08090778440237045, "global_step": 211078, "epoch": 2371, "lr": 1.1757425824613971e-05}
{"train_loss": 0.07227031141519547, "global_step": 211079, "epoch": 2371, "lr": 1.1757052357009358e-05}
{"train_loss": 0.07685358077287674, "global_step": 211080, "epoch": 2371, "lr": 1.1756678894546026e-05}
{"train_loss": 0.10839932411909103, "global_step": 211081, "epoch": 2371, "lr": 1.1756305437224059e-05}
{"train_loss": 0.060940179973840714, "global_step": 211082, "epoch": 2371, "lr": 1.1755931985043478e-05}
{"train_loss": 0.07058979570865631, "global_step": 211083, "epoch": 2371, "lr": 1.1755558538004357e-05}
{"train_loss": 0.0394548624753952, "global_step": 211084, "epoch": 2371, "lr": 1.1755185096106724e-05}
{"train_loss": 0.05347313731908798, "global_step": 211085, "epoch": 2371, "lr": 1.1754811659350651e-05}
{"train_loss": 0.07164033502340317, "global_step": 211086, "epoch": 2371, "lr": 1.1754438227736164e-05}
{"train_loss": 0.017543334513902664, "global_step": 211087, "epoch": 2371, "lr": 1.1754064801263332e-05}
{"train_loss": 0.02394131012260914, "global_step": 211088, "epoch": 2371, "lr": 1.1753691379932208e-05}
{"train_loss": 0.03486672788858414, "global_step": 211089, "epoch": 2371, "lr": 1.1753317963742821e-05}
{"train_loss": 0.05154058337211609, "global_step": 211090, "epoch": 2371, "lr": 1.175294455269525e-05}
{"train_loss": 0.03263610228896141, "global_step": 211091, "epoch": 2371, "lr": 1.175257114678951e-05}
{"train_loss": 0.07356210798025131, "global_step": 211092, "epoch": 2371, "lr": 1.1752197746025689e-05}
{"train_loss": 0.09901046007871628, "global_step": 211093, "epoch": 2371, "lr": 1.1751824350403805e-05}
{"train_loss": 0.07532782852649689, "global_step": 211094, "epoch": 2371, "lr": 1.1751450959923937e-05}
{"train_loss": 0.05922778323292732, "global_step": 211095, "epoch": 2371, "lr": 1.1751077574586106e-05}
{"train_loss": 0.05680574104189873, "global_step": 211096, "epoch": 2371, "lr": 1.1750704194390389e-05}
{"train_loss": 0.09372343122959137, "global_step": 211097, "epoch": 2371, "lr": 1.1750330819336808e-05}
{"train_loss": 0.052030298858881, "global_step": 211098, "epoch": 2371, "lr": 1.1749957449425448e-05}
{"train_loss": 0.08408722281455994, "global_step": 211099, "epoch": 2371, "lr": 1.1749584084656323e-05}
{"train_loss": 0.06700626015663147, "global_step": 211100, "epoch": 2371, "lr": 1.1749210725029513e-05}
{"train_loss": 0.03149687498807907, "global_step": 211101, "epoch": 2371, "lr": 1.174883737054504e-05}
{"train_loss": 0.05628076195716858, "global_step": 211102, "epoch": 2371, "lr": 1.1748464021202987e-05}
{"train_loss": 0.055620066821575165, "global_step": 211103, "epoch": 2371, "lr": 1.174809067700337e-05}
{"train_loss": 0.04118639975786209, "global_step": 211104, "epoch": 2371, "lr": 1.1747717337946274e-05}
{"train_loss": 0.06933486461639404, "global_step": 211105, "epoch": 2371, "lr": 1.1747344004031713e-05}
{"train_loss": 0.04672500118613243, "global_step": 211106, "epoch": 2371, "lr": 1.174697067525976e-05}
{"train_loss": 0.0581294960143526, "global_step": 211107, "epoch": 2371, "lr": 1.1746597351630473e-05, "val_loss": 8.780056953430176}
{"train_loss": 0.052897170186042786, "global_step": 211108, "epoch": 2372, "lr": 1.174622403314387e-05}
{"train_loss": 0.07170510292053223, "global_step": 211109, "epoch": 2372, "lr": 1.1745850719800044e-05}
{"train_loss": 0.05586697906255722, "global_step": 211110, "epoch": 2372, "lr": 1.1745477411599015e-05}
{"train_loss": 0.035217154771089554, "global_step": 211111, "epoch": 2372, "lr": 1.1745104108540828e-05}
{"train_loss": 0.07116346061229706, "global_step": 211112, "epoch": 2372, "lr": 1.1744730810625565e-05}
{"train_loss": 0.047171369194984436, "global_step": 211113, "epoch": 2372, "lr": 1.1744357517853238e-05}
{"train_loss": 0.05315128341317177, "global_step": 211114, "epoch": 2372, "lr": 1.1743984230223925e-05}
{"train_loss": 0.03436780720949173, "global_step": 211115, "epoch": 2372, "lr": 1.174361094773766e-05}
{"train_loss": 0.07685616612434387, "global_step": 211116, "epoch": 2372, "lr": 1.1743237670394513e-05}
{"train_loss": 0.0421534925699234, "global_step": 211117, "epoch": 2372, "lr": 1.1742864398194519e-05}
{"train_loss": 0.04002316668629646, "global_step": 211118, "epoch": 2372, "lr": 1.1742491131137722e-05}
{"train_loss": 0.055612824857234955, "global_step": 211119, "epoch": 2372, "lr": 1.174211786922419e-05}
{"train_loss": 0.03527364507317543, "global_step": 211120, "epoch": 2372, "lr": 1.1741744612453947e-05}
{"train_loss": 0.08586239814758301, "global_step": 211121, "epoch": 2372, "lr": 1.174137136082708e-05}
{"train_loss": 0.053147993981838226, "global_step": 211122, "epoch": 2372, "lr": 1.1740998114343604e-05}
{"train_loss": 0.0956653580069542, "global_step": 211123, "epoch": 2372, "lr": 1.1740624873003587e-05}
{"train_loss": 0.08819498121738434, "global_step": 211124, "epoch": 2372, "lr": 1.1740251636807087e-05}
{"train_loss": 0.0742955282330513, "global_step": 211125, "epoch": 2372, "lr": 1.1739878405754129e-05}
{"train_loss": 0.03395550325512886, "global_step": 211126, "epoch": 2372, "lr": 1.1739505179844795e-05}
{"train_loss": 0.029956242069602013, "global_step": 211127, "epoch": 2372, "lr": 1.1739131959079103e-05}
{"train_loss": 0.03954463452100754, "global_step": 211128, "epoch": 2372, "lr": 1.1738758743457129e-05}
{"train_loss": 0.06749038398265839, "global_step": 211129, "epoch": 2372, "lr": 1.1738385532978902e-05}
{"train_loss": 0.04706259071826935, "global_step": 211130, "epoch": 2372, "lr": 1.1738012327644499e-05}
{"train_loss": 0.09265831857919693, "global_step": 211131, "epoch": 2372, "lr": 1.1737639127453936e-05}
{"train_loss": 0.04111324995756149, "global_step": 211132, "epoch": 2372, "lr": 1.1737265932407293e-05}
{"train_loss": 0.051862932741642, "global_step": 211133, "epoch": 2372, "lr": 1.1736892742504602e-05}
{"train_loss": 0.04656615108251572, "global_step": 211134, "epoch": 2372, "lr": 1.173651955774593e-05}
{"train_loss": 0.023336632177233696, "global_step": 211135, "epoch": 2372, "lr": 1.1736146378131297e-05}
{"train_loss": 0.0358746275305748, "global_step": 211136, "epoch": 2372, "lr": 1.1735773203660793e-05}
{"train_loss": 0.042100418359041214, "global_step": 211137, "epoch": 2372, "lr": 1.1735400034334431e-05}
{"train_loss": 0.07791604846715927, "global_step": 211138, "epoch": 2372, "lr": 1.1735026870152293e-05}
{"train_loss": 0.11620993912220001, "global_step": 211139, "epoch": 2372, "lr": 1.1734653711114401e-05}
{"train_loss": 0.07473886758089066, "global_step": 211140, "epoch": 2372, "lr": 1.1734280557220833e-05}
{"train_loss": 0.08271557837724686, "global_step": 211141, "epoch": 2372, "lr": 1.1733907408471612e-05}
{"train_loss": 0.10833992809057236, "global_step": 211142, "epoch": 2372, "lr": 1.1733534264866796e-05}
{"train_loss": 0.04698697850108147, "global_step": 211143, "epoch": 2372, "lr": 1.1733161126406455e-05}
{"train_loss": 0.08685524016618729, "global_step": 211144, "epoch": 2372, "lr": 1.1732787993090611e-05}
{"train_loss": 0.07965560257434845, "global_step": 211145, "epoch": 2372, "lr": 1.1732414864919344e-05}
{"train_loss": 0.020302996039390564, "global_step": 211146, "epoch": 2372, "lr": 1.1732041741892668e-05}
{"train_loss": 0.08218000829219818, "global_step": 211147, "epoch": 2372, "lr": 1.1731668624010672e-05}
{"train_loss": 0.03678162023425102, "global_step": 211148, "epoch": 2372, "lr": 1.1731295511273372e-05}
{"train_loss": 0.08873818069696426, "global_step": 211149, "epoch": 2372, "lr": 1.1730922403680844e-05}
{"train_loss": 0.06019863113760948, "global_step": 211150, "epoch": 2372, "lr": 1.1730549301233129e-05}
{"train_loss": 0.037437811493873596, "global_step": 211151, "epoch": 2372, "lr": 1.1730176203930259e-05}
{"train_loss": 0.061083387583494186, "global_step": 211152, "epoch": 2372, "lr": 1.1729803111772313e-05}
{"train_loss": 0.03313290700316429, "global_step": 211153, "epoch": 2372, "lr": 1.1729430024759319e-05}
{"train_loss": 0.05679737403988838, "global_step": 211154, "epoch": 2372, "lr": 1.1729056942891348e-05}
{"train_loss": 0.020280202850699425, "global_step": 211155, "epoch": 2372, "lr": 1.1728683866168438e-05}
{"train_loss": 0.10352211445569992, "global_step": 211156, "epoch": 2372, "lr": 1.172831079459063e-05}
{"train_loss": 0.07836096733808517, "global_step": 211157, "epoch": 2372, "lr": 1.1727937728157995e-05}
{"train_loss": 0.08407274633646011, "global_step": 211158, "epoch": 2372, "lr": 1.1727564666870562e-05}
{"train_loss": 0.08815288543701172, "global_step": 211159, "epoch": 2372, "lr": 1.17271916107284e-05}
{"train_loss": 0.05930868163704872, "global_step": 211160, "epoch": 2372, "lr": 1.172681855973154e-05}
{"train_loss": 0.0493323877453804, "global_step": 211161, "epoch": 2372, "lr": 1.172644551388004e-05}
{"train_loss": 0.06795237958431244, "global_step": 211162, "epoch": 2372, "lr": 1.1726072473173971e-05}
{"train_loss": 0.04383060708642006, "global_step": 211163, "epoch": 2372, "lr": 1.1725699437613352e-05}
{"train_loss": 0.05988481640815735, "global_step": 211164, "epoch": 2372, "lr": 1.1725326407198256e-05}
{"train_loss": 0.030975013971328735, "global_step": 211165, "epoch": 2372, "lr": 1.1724953381928711e-05}
{"train_loss": 0.08241242915391922, "global_step": 211166, "epoch": 2372, "lr": 1.1724580361804794e-05}
{"train_loss": 0.07796638458967209, "global_step": 211167, "epoch": 2372, "lr": 1.1724207346826522e-05}
{"train_loss": 0.08566795289516449, "global_step": 211168, "epoch": 2372, "lr": 1.1723834336993983e-05}
{"train_loss": 0.04781078174710274, "global_step": 211169, "epoch": 2372, "lr": 1.1723461332307189e-05}
{"train_loss": 0.05907401815056801, "global_step": 211170, "epoch": 2372, "lr": 1.172308833276623e-05}
{"train_loss": 0.07298187166452408, "global_step": 211171, "epoch": 2372, "lr": 1.1722715338371116e-05}
{"train_loss": 0.06670477241277695, "global_step": 211172, "epoch": 2372, "lr": 1.1722342349121934e-05}
{"train_loss": 0.05239760875701904, "global_step": 211173, "epoch": 2372, "lr": 1.1721969365018704e-05}
{"train_loss": 0.042154859751462936, "global_step": 211174, "epoch": 2372, "lr": 1.1721596386061495e-05}
{"train_loss": 0.06791020929813385, "global_step": 211175, "epoch": 2372, "lr": 1.1721223412250343e-05}
{"train_loss": 0.031001603230834007, "global_step": 211176, "epoch": 2372, "lr": 1.1720850443585318e-05}
{"train_loss": 0.01739232800900936, "global_step": 211177, "epoch": 2372, "lr": 1.1720477480066444e-05}
{"train_loss": 0.0494573675096035, "global_step": 211178, "epoch": 2372, "lr": 1.1720104521693786e-05}
{"train_loss": 0.035159289836883545, "global_step": 211179, "epoch": 2372, "lr": 1.1719731568467406e-05}
{"train_loss": 0.06273618340492249, "global_step": 211180, "epoch": 2372, "lr": 1.1719358620387327e-05}
{"train_loss": 0.07025480270385742, "global_step": 211181, "epoch": 2372, "lr": 1.171898567745363e-05}
{"train_loss": 0.05211132392287254, "global_step": 211182, "epoch": 2372, "lr": 1.1718612739666335e-05}
{"train_loss": 0.10588773339986801, "global_step": 211183, "epoch": 2372, "lr": 1.1718239807025517e-05}
{"train_loss": 0.04811498895287514, "global_step": 211184, "epoch": 2372, "lr": 1.1717866879531203e-05}
{"train_loss": 0.06620355695486069, "global_step": 211185, "epoch": 2372, "lr": 1.1717493957183473e-05}
{"train_loss": 0.03264939412474632, "global_step": 211186, "epoch": 2372, "lr": 1.1717121039982338e-05}
{"train_loss": 0.02956480346620083, "global_step": 211187, "epoch": 2372, "lr": 1.171674812792789e-05}
{"train_loss": 0.05183503031730652, "global_step": 211188, "epoch": 2372, "lr": 1.1716375221020154e-05}
{"train_loss": 0.07847569137811661, "global_step": 211189, "epoch": 2372, "lr": 1.1716002319259172e-05}
{"train_loss": 0.055590130388736725, "global_step": 211190, "epoch": 2372, "lr": 1.1715629422645018e-05}
{"train_loss": 0.040115710347890854, "global_step": 211191, "epoch": 2372, "lr": 1.1715256531177721e-05}
{"train_loss": 0.04565935581922531, "global_step": 211192, "epoch": 2372, "lr": 1.1714883644857356e-05}
{"train_loss": 0.06407105177640915, "global_step": 211193, "epoch": 2372, "lr": 1.1714510763683955e-05}
{"train_loss": 0.0429605133831501, "global_step": 211194, "epoch": 2372, "lr": 1.1714137887657562e-05}
{"train_loss": 0.07284953445196152, "global_step": 211195, "epoch": 2372, "lr": 1.1713765016778244e-05}
{"train_loss": 0.05845532700251997, "global_step": 211196, "epoch": 2372, "lr": 1.1713392151046037e-05, "val_loss": 8.658297538757324}
{"train_loss": 0.04875195771455765, "global_step": 211197, "epoch": 2373, "lr": 1.171301929046099e-05}
{"train_loss": 0.04612315446138382, "global_step": 211198, "epoch": 2373, "lr": 1.1712646435023184e-05}
{"train_loss": 0.08343490958213806, "global_step": 211199, "epoch": 2373, "lr": 1.1712273584732625e-05}
{"train_loss": 0.047177985310554504, "global_step": 211200, "epoch": 2373, "lr": 1.1711900739589398e-05}
{"train_loss": 0.039033811539411545, "global_step": 211201, "epoch": 2373, "lr": 1.1711527899593527e-05}
{"train_loss": 0.05869914963841438, "global_step": 211202, "epoch": 2373, "lr": 1.1711155064745089e-05}
{"train_loss": 0.07817411422729492, "global_step": 211203, "epoch": 2373, "lr": 1.1710782235044105e-05}
{"train_loss": 0.03232042118906975, "global_step": 211204, "epoch": 2373, "lr": 1.1710409410490659e-05}
{"train_loss": 0.0734376311302185, "global_step": 211205, "epoch": 2373, "lr": 1.1710036591084761e-05}
{"train_loss": 0.07761198282241821, "global_step": 211206, "epoch": 2373, "lr": 1.1709663776826496e-05}
{"train_loss": 0.05160877853631973, "global_step": 211207, "epoch": 2373, "lr": 1.1709290967715885e-05}
{"train_loss": 0.10540620237588882, "global_step": 211208, "epoch": 2373, "lr": 1.1708918163753014e-05}
{"train_loss": 0.03195580840110779, "global_step": 211209, "epoch": 2373, "lr": 1.170854536493789e-05}
{"train_loss": 0.03430665656924248, "global_step": 211210, "epoch": 2373, "lr": 1.1708172571270603e-05}
{"train_loss": 0.07710415124893188, "global_step": 211211, "epoch": 2373, "lr": 1.1707799782751172e-05}
{"train_loss": 0.03584709390997887, "global_step": 211212, "epoch": 2373, "lr": 1.1707426999379678e-05}
{"train_loss": 0.07916014641523361, "global_step": 211213, "epoch": 2373, "lr": 1.1707054221156133e-05}
{"train_loss": 0.04024672135710716, "global_step": 211214, "epoch": 2373, "lr": 1.1706681448080614e-05}
{"train_loss": 0.05599058419466019, "global_step": 211215, "epoch": 2373, "lr": 1.1706308680153173e-05}
{"train_loss": 0.04691964387893677, "global_step": 211216, "epoch": 2373, "lr": 1.170593591737384e-05}
{"train_loss": 0.07965993136167526, "global_step": 211217, "epoch": 2373, "lr": 1.1705563159742695e-05}
{"train_loss": 0.03374544531106949, "global_step": 211218, "epoch": 2373, "lr": 1.1705190407259753e-05}
{"train_loss": 0.02387405000627041, "global_step": 211219, "epoch": 2373, "lr": 1.1704817659925094e-05}
{"train_loss": 0.032169509679079056, "global_step": 211220, "epoch": 2373, "lr": 1.1704444917738744e-05}
{"train_loss": 0.04732801020145416, "global_step": 211221, "epoch": 2373, "lr": 1.1704072180700776e-05}
{"train_loss": 0.07325159013271332, "global_step": 211222, "epoch": 2373, "lr": 1.1703699448811217e-05}
{"train_loss": 0.09621801227331161, "global_step": 211223, "epoch": 2373, "lr": 1.1703326722070146e-05}
{"train_loss": 0.05695278197526932, "global_step": 211224, "epoch": 2373, "lr": 1.1702954000477578e-05}
{"train_loss": 0.027785135433077812, "global_step": 211225, "epoch": 2373, "lr": 1.1702581284033593e-05}
{"train_loss": 0.03343519940972328, "global_step": 211226, "epoch": 2373, "lr": 1.1702208572738232e-05}
{"train_loss": 0.04661178216338158, "global_step": 211227, "epoch": 2373, "lr": 1.1701835866591527e-05}
{"train_loss": 0.06981775909662247, "global_step": 211228, "epoch": 2373, "lr": 1.1701463165593562e-05}
{"train_loss": 0.04811704158782959, "global_step": 211229, "epoch": 2373, "lr": 1.1701090469744347e-05}
{"train_loss": 0.044322095811367035, "global_step": 211230, "epoch": 2373, "lr": 1.1700717779043974e-05}
{"train_loss": 0.059782058000564575, "global_step": 211231, "epoch": 2373, "lr": 1.1700345093492454e-05}
{"train_loss": 0.01573716662824154, "global_step": 211232, "epoch": 2373, "lr": 1.169997241308987e-05}
{"train_loss": 0.09800171852111816, "global_step": 211233, "epoch": 2373, "lr": 1.1699599737836248e-05}
{"train_loss": 0.07726332545280457, "global_step": 211234, "epoch": 2373, "lr": 1.169922706773166e-05}
{"train_loss": 0.04723102226853371, "global_step": 211235, "epoch": 2373, "lr": 1.1698854402776133e-05}
{"train_loss": 0.04159426689147949, "global_step": 211236, "epoch": 2373, "lr": 1.1698481742969735e-05}
{"train_loss": 0.05759043991565704, "global_step": 211237, "epoch": 2373, "lr": 1.1698109088312504e-05}
{"train_loss": 0.0961652547121048, "global_step": 211238, "epoch": 2373, "lr": 1.1697736438804502e-05}
{"train_loss": 0.05358950048685074, "global_step": 211239, "epoch": 2373, "lr": 1.1697363794445758e-05}
{"train_loss": 0.04226788133382797, "global_step": 211240, "epoch": 2373, "lr": 1.1696991155236359e-05}
{"train_loss": 0.06461355835199356, "global_step": 211241, "epoch": 2373, "lr": 1.1696618521176311e-05}
{"train_loss": 0.08836303651332855, "global_step": 211242, "epoch": 2373, "lr": 1.16962458922657e-05}
{"train_loss": 0.04612228274345398, "global_step": 211243, "epoch": 2373, "lr": 1.1695873268504554e-05}
{"train_loss": 0.032092899084091187, "global_step": 211244, "epoch": 2373, "lr": 1.1695500649892942e-05}
{"train_loss": 0.061837464570999146, "global_step": 211245, "epoch": 2373, "lr": 1.1695128036430886e-05}
{"train_loss": 0.06591702997684479, "global_step": 211246, "epoch": 2373, "lr": 1.169475542811847e-05}
{"train_loss": 0.03969030827283859, "global_step": 211247, "epoch": 2373, "lr": 1.1694382824955713e-05}
{"train_loss": 0.013861979357898235, "global_step": 211248, "epoch": 2373, "lr": 1.1694010226942692e-05}
{"train_loss": 0.04489830136299133, "global_step": 211249, "epoch": 2373, "lr": 1.1693637634079429e-05}
{"train_loss": 0.04374488443136215, "global_step": 211250, "epoch": 2373, "lr": 1.1693265046365997e-05}
{"train_loss": 0.03585473820567131, "global_step": 211251, "epoch": 2373, "lr": 1.1692892463802446e-05}
{"train_loss": 0.07454831153154373, "global_step": 211252, "epoch": 2373, "lr": 1.1692519886388809e-05}
{"train_loss": 0.015506439842283726, "global_step": 211253, "epoch": 2373, "lr": 1.1692147314125156e-05}
{"train_loss": 0.03292522579431534, "global_step": 211254, "epoch": 2373, "lr": 1.1691774747011513e-05}
{"train_loss": 0.0712786540389061, "global_step": 211255, "epoch": 2373, "lr": 1.1691402185047956e-05}
{"train_loss": 0.0446917787194252, "global_step": 211256, "epoch": 2373, "lr": 1.1691029628234513e-05}
{"train_loss": 0.03821692615747452, "global_step": 211257, "epoch": 2373, "lr": 1.169065707657126e-05}
{"train_loss": 0.05389299988746643, "global_step": 211258, "epoch": 2373, "lr": 1.1690284530058216e-05}
{"train_loss": 0.08212502300739288, "global_step": 211259, "epoch": 2373, "lr": 1.1689911988695457e-05}
{"train_loss": 0.03401857614517212, "global_step": 211260, "epoch": 2373, "lr": 1.1689539452483012e-05}
{"train_loss": 0.08769537508487701, "global_step": 211261, "epoch": 2373, "lr": 1.1689166921420951e-05}
{"train_loss": 0.06071611866354942, "global_step": 211262, "epoch": 2373, "lr": 1.16887943955093e-05}
{"train_loss": 0.010251631028950214, "global_step": 211263, "epoch": 2373, "lr": 1.1688421874748145e-05}
{"train_loss": 0.03574591130018234, "global_step": 211264, "epoch": 2373, "lr": 1.1688049359137494e-05}
{"train_loss": 0.040577422827482224, "global_step": 211265, "epoch": 2373, "lr": 1.1687676848677437e-05}
{"train_loss": 0.06146254390478134, "global_step": 211266, "epoch": 2373, "lr": 1.1687304343368e-05}
{"train_loss": 0.07160867005586624, "global_step": 211267, "epoch": 2373, "lr": 1.1686931843209225e-05}
{"train_loss": 0.05156441032886505, "global_step": 211268, "epoch": 2373, "lr": 1.1686559348201193e-05}
{"train_loss": 0.06761746108531952, "global_step": 211269, "epoch": 2373, "lr": 1.1686186858343917e-05}
{"train_loss": 0.050075169652700424, "global_step": 211270, "epoch": 2373, "lr": 1.1685814373637482e-05}
{"train_loss": 0.03631148114800453, "global_step": 211271, "epoch": 2373, "lr": 1.1685441894081911e-05}
{"train_loss": 0.12785880267620087, "global_step": 211272, "epoch": 2373, "lr": 1.1685069419677281e-05}
{"train_loss": 0.05083158239722252, "global_step": 211273, "epoch": 2373, "lr": 1.1684696950423607e-05}
{"train_loss": 0.04716423898935318, "global_step": 211274, "epoch": 2373, "lr": 1.1684324486320975e-05}
{"train_loss": 0.09204263985157013, "global_step": 211275, "epoch": 2373, "lr": 1.1683952027369399e-05}
{"train_loss": 0.046762630343437195, "global_step": 211276, "epoch": 2373, "lr": 1.168357957356897e-05}
{"train_loss": 0.048306480050086975, "global_step": 211277, "epoch": 2373, "lr": 1.1683207124919698e-05}
{"train_loss": 0.0486372672021389, "global_step": 211278, "epoch": 2373, "lr": 1.1682834681421668e-05}
{"train_loss": 0.0611368864774704, "global_step": 211279, "epoch": 2373, "lr": 1.1682462243074898e-05}
{"train_loss": 0.03354480117559433, "global_step": 211280, "epoch": 2373, "lr": 1.1682089809879471e-05}
{"train_loss": 0.039856597781181335, "global_step": 211281, "epoch": 2373, "lr": 1.1681717381835405e-05}
{"train_loss": 0.057500507682561874, "global_step": 211282, "epoch": 2373, "lr": 1.168134495894278e-05}
{"train_loss": 0.04662727937102318, "global_step": 211283, "epoch": 2373, "lr": 1.1680972541201617e-05}
{"train_loss": 0.03489382937550545, "global_step": 211284, "epoch": 2373, "lr": 1.168060012861199e-05}
{"train_loss": 0.054320637293662245, "global_step": 211285, "epoch": 2373, "lr": 1.168022772117393e-05, "val_loss": 8.644850730895996}
{"train_loss": 0.04655861854553223, "global_step": 211286, "epoch": 2374, "lr": 1.1679855318887506e-05}
{"train_loss": 0.06549876183271408, "global_step": 211287, "epoch": 2374, "lr": 1.1679482921752749e-05}
{"train_loss": 0.0350116603076458, "global_step": 211288, "epoch": 2374, "lr": 1.1679110529769715e-05}
{"train_loss": 0.06939207017421722, "global_step": 211289, "epoch": 2374, "lr": 1.1678738142938473e-05}
{"train_loss": 0.020255329087376595, "global_step": 211290, "epoch": 2374, "lr": 1.1678365761259048e-05}
{"train_loss": 0.03819936141371727, "global_step": 211291, "epoch": 2374, "lr": 1.1677993384731506e-05}
{"train_loss": 0.04887431114912033, "global_step": 211292, "epoch": 2374, "lr": 1.1677621013355882e-05}
{"train_loss": 0.056314948946237564, "global_step": 211293, "epoch": 2374, "lr": 1.1677248647132244e-05}
{"train_loss": 0.10102369636297226, "global_step": 211294, "epoch": 2374, "lr": 1.1676876286060624e-05}
{"train_loss": 0.05002212151885033, "global_step": 211295, "epoch": 2374, "lr": 1.1676503930141092e-05}
{"train_loss": 0.0500614307820797, "global_step": 211296, "epoch": 2374, "lr": 1.1676131579373673e-05}
{"train_loss": 0.04110042750835419, "global_step": 211297, "epoch": 2374, "lr": 1.1675759233758443e-05}
{"train_loss": 0.03903444483876228, "global_step": 211298, "epoch": 2374, "lr": 1.167538689329543e-05}
{"train_loss": 0.040480729192495346, "global_step": 211299, "epoch": 2374, "lr": 1.167501455798471e-05}
{"train_loss": 0.04157340154051781, "global_step": 211300, "epoch": 2374, "lr": 1.16746422278263e-05}
{"train_loss": 0.07060014456510544, "global_step": 211301, "epoch": 2374, "lr": 1.1674269902820278e-05}
{"train_loss": 0.012764481827616692, "global_step": 211302, "epoch": 2374, "lr": 1.167389758296667e-05}
{"train_loss": 0.0739947035908699, "global_step": 211303, "epoch": 2374, "lr": 1.1673525268265556e-05}
{"train_loss": 0.07145500928163528, "global_step": 211304, "epoch": 2374, "lr": 1.1673152958716971e-05}
{"train_loss": 0.056439582258462906, "global_step": 211305, "epoch": 2374, "lr": 1.1672780654320947e-05}
{"train_loss": 0.041575539857149124, "global_step": 211306, "epoch": 2374, "lr": 1.1672408355077564e-05}
{"train_loss": 0.06860756129026413, "global_step": 211307, "epoch": 2374, "lr": 1.1672036060986846e-05}
{"train_loss": 0.05840447545051575, "global_step": 211308, "epoch": 2374, "lr": 1.1671663772048869e-05}
{"train_loss": 0.04791847988963127, "global_step": 211309, "epoch": 2374, "lr": 1.1671291488263658e-05}
{"train_loss": 0.026685092598199844, "global_step": 211310, "epoch": 2374, "lr": 1.1670919209631287e-05}
{"train_loss": 0.040179189294576645, "global_step": 211311, "epoch": 2374, "lr": 1.1670546936151778e-05}
{"train_loss": 0.052650731056928635, "global_step": 211312, "epoch": 2374, "lr": 1.1670174667825212e-05}
{"train_loss": 0.016505707055330276, "global_step": 211313, "epoch": 2374, "lr": 1.1669802404651608e-05}
{"train_loss": 0.02790699154138565, "global_step": 211314, "epoch": 2374, "lr": 1.166943014663105e-05}
{"train_loss": 0.03477847948670387, "global_step": 211315, "epoch": 2374, "lr": 1.1669057893763552e-05}
{"train_loss": 0.055674925446510315, "global_step": 211316, "epoch": 2374, "lr": 1.16686856460492e-05}
{"train_loss": 0.040938470512628555, "global_step": 211317, "epoch": 2374, "lr": 1.1668313403488007e-05}
{"train_loss": 0.04399995505809784, "global_step": 211318, "epoch": 2374, "lr": 1.166794116608006e-05}
{"train_loss": 0.05280395597219467, "global_step": 211319, "epoch": 2374, "lr": 1.1667568933825373e-05}
{"train_loss": 0.09549787640571594, "global_step": 211320, "epoch": 2374, "lr": 1.1667196706724031e-05}
{"train_loss": 0.06039445102214813, "global_step": 211321, "epoch": 2374, "lr": 1.1666824484776057e-05}
{"train_loss": 0.0694100633263588, "global_step": 211322, "epoch": 2374, "lr": 1.166645226798152e-05}
{"train_loss": 0.10519235581159592, "global_step": 211323, "epoch": 2374, "lr": 1.1666080056340444e-05}
{"train_loss": 0.06911300867795944, "global_step": 211324, "epoch": 2374, "lr": 1.1665707849852903e-05}
{"train_loss": 0.041997164487838745, "global_step": 211325, "epoch": 2374, "lr": 1.1665335648518954e-05}
{"train_loss": 0.08023828268051147, "global_step": 211326, "epoch": 2374, "lr": 1.1664963452338618e-05}
{"train_loss": 0.07165997475385666, "global_step": 211327, "epoch": 2374, "lr": 1.1664591261311974e-05}
{"train_loss": 0.0392482690513134, "global_step": 211328, "epoch": 2374, "lr": 1.1664219075439042e-05}
{"train_loss": 0.04829514026641846, "global_step": 211329, "epoch": 2374, "lr": 1.166384689471991e-05}
{"train_loss": 0.012386688962578773, "global_step": 211330, "epoch": 2374, "lr": 1.1663474719154582e-05}
{"train_loss": 0.05668218433856964, "global_step": 211331, "epoch": 2374, "lr": 1.1663102548743155e-05}
{"train_loss": 0.05646519735455513, "global_step": 211332, "epoch": 2374, "lr": 1.1662730383485637e-05}
{"train_loss": 0.06033218279480934, "global_step": 211333, "epoch": 2374, "lr": 1.166235822338212e-05}
{"train_loss": 0.03762488812208176, "global_step": 211334, "epoch": 2374, "lr": 1.1661986068432606e-05}
{"train_loss": 0.03381940349936485, "global_step": 211335, "epoch": 2374, "lr": 1.1661613918637193e-05}
{"train_loss": 0.014943825080990791, "global_step": 211336, "epoch": 2374, "lr": 1.1661241773995897e-05}
{"train_loss": 0.044769011437892914, "global_step": 211337, "epoch": 2374, "lr": 1.1660869634508787e-05}
{"train_loss": 0.05675916001200676, "global_step": 211338, "epoch": 2374, "lr": 1.1660497500175893e-05}
{"train_loss": 0.06747855991125107, "global_step": 211339, "epoch": 2374, "lr": 1.1660125370997294e-05}
{"train_loss": 0.07575133442878723, "global_step": 211340, "epoch": 2374, "lr": 1.165975324697301e-05}
{"train_loss": 0.09479052573442459, "global_step": 211341, "epoch": 2374, "lr": 1.1659381128103113e-05}
{"train_loss": 0.04013587161898613, "global_step": 211342, "epoch": 2374, "lr": 1.165900901438764e-05}
{"train_loss": 0.044914014637470245, "global_step": 211343, "epoch": 2374, "lr": 1.1658636905826653e-05}
{"train_loss": 0.05713341012597084, "global_step": 211344, "epoch": 2374, "lr": 1.16582648024202e-05}
{"train_loss": 0.04970142990350723, "global_step": 211345, "epoch": 2374, "lr": 1.1657892704168305e-05}
{"train_loss": 0.04284433647990227, "global_step": 211346, "epoch": 2374, "lr": 1.165752061107106e-05}
{"train_loss": 0.04118217155337334, "global_step": 211347, "epoch": 2374, "lr": 1.1657148523128475e-05}
{"train_loss": 0.028828497976064682, "global_step": 211348, "epoch": 2374, "lr": 1.165677644034064e-05}
{"train_loss": 0.08288148045539856, "global_step": 211349, "epoch": 2374, "lr": 1.1656404362707563e-05}
{"train_loss": 0.0793110802769661, "global_step": 211350, "epoch": 2374, "lr": 1.165603229022933e-05}
{"train_loss": 0.0803389772772789, "global_step": 211351, "epoch": 2374, "lr": 1.1655660222905963e-05}
{"train_loss": 0.0623663067817688, "global_step": 211352, "epoch": 2374, "lr": 1.1655288160737543e-05}
{"train_loss": 0.06090022251009941, "global_step": 211353, "epoch": 2374, "lr": 1.1654916103724083e-05}
{"train_loss": 0.06626366078853607, "global_step": 211354, "epoch": 2374, "lr": 1.1654544051865667e-05}
{"train_loss": 0.03874679282307625, "global_step": 211355, "epoch": 2374, "lr": 1.1654172005162318e-05}
{"train_loss": 0.11413414031267166, "global_step": 211356, "epoch": 2374, "lr": 1.1653799963614114e-05}
{"train_loss": 0.06086024269461632, "global_step": 211357, "epoch": 2374, "lr": 1.1653427927221072e-05}
{"train_loss": 0.0608484260737896, "global_step": 211358, "epoch": 2374, "lr": 1.1653055895983273e-05}
{"train_loss": 0.027973640710115433, "global_step": 211359, "epoch": 2374, "lr": 1.1652683869900737e-05}
{"train_loss": 0.04380464553833008, "global_step": 211360, "epoch": 2374, "lr": 1.1652311848973536e-05}
{"train_loss": 0.07254569232463837, "global_step": 211361, "epoch": 2374, "lr": 1.1651939833201725e-05}
{"train_loss": 0.04499732330441475, "global_step": 211362, "epoch": 2374, "lr": 1.165156782258533e-05}
{"train_loss": 0.07056832313537598, "global_step": 211363, "epoch": 2374, "lr": 1.1651195817124433e-05}
{"train_loss": 0.08190639317035675, "global_step": 211364, "epoch": 2374, "lr": 1.1650823816819045e-05}
{"train_loss": 0.0567016564309597, "global_step": 211365, "epoch": 2374, "lr": 1.1650451821669251e-05}
{"train_loss": 0.0960899069905281, "global_step": 211366, "epoch": 2374, "lr": 1.1650079831675076e-05}
{"train_loss": 0.03944507986307144, "global_step": 211367, "epoch": 2374, "lr": 1.1649707846836589e-05}
{"train_loss": 0.07238519191741943, "global_step": 211368, "epoch": 2374, "lr": 1.1649335867153821e-05}
{"train_loss": 0.09584499895572662, "global_step": 211369, "epoch": 2374, "lr": 1.1648963892626846e-05}
{"train_loss": 0.08440469950437546, "global_step": 211370, "epoch": 2374, "lr": 1.1648591923255686e-05}
{"train_loss": 0.07296595722436905, "global_step": 211371, "epoch": 2374, "lr": 1.1648219959040419e-05}
{"train_loss": 0.041437797248363495, "global_step": 211372, "epoch": 2374, "lr": 1.1647847999981066e-05}
{"train_loss": 0.05361001193523407, "global_step": 211373, "epoch": 2374, "lr": 1.164747604607771e-05}
{"train_loss": 0.0552543725483538, "global_step": 211374, "epoch": 2374, "lr": 1.1647104097330364e-05, "val_loss": 8.697515487670898}
{"train_loss": 0.03950296342372894, "global_step": 211375, "epoch": 2375, "lr": 1.1646732153739115e-05}
{"train_loss": 0.05386364832520485, "global_step": 211376, "epoch": 2375, "lr": 1.164636021530398e-05}
{"train_loss": 0.0398108996450901, "global_step": 211377, "epoch": 2375, "lr": 1.1645988282025045e-05}
{"train_loss": 0.0413220189511776, "global_step": 211378, "epoch": 2375, "lr": 1.1645616353902317e-05}
{"train_loss": 0.05261308699846268, "global_step": 211379, "epoch": 2375, "lr": 1.1645244430935886e-05}
{"train_loss": 0.028175625950098038, "global_step": 211380, "epoch": 2375, "lr": 1.164487251312577e-05}
{"train_loss": 0.061356931924819946, "global_step": 211381, "epoch": 2375, "lr": 1.1644500600472048e-05}
{"train_loss": 0.06550688296556473, "global_step": 211382, "epoch": 2375, "lr": 1.1644128692974753e-05}
{"train_loss": 0.019604535773396492, "global_step": 211383, "epoch": 2375, "lr": 1.1643756790633926e-05}
{"train_loss": 0.034634578973054886, "global_step": 211384, "epoch": 2375, "lr": 1.1643384893449638e-05}
{"train_loss": 0.05679609626531601, "global_step": 211385, "epoch": 2375, "lr": 1.164301300142192e-05}
{"train_loss": 0.05873474106192589, "global_step": 211386, "epoch": 2375, "lr": 1.1642641114550845e-05}
{"train_loss": 0.011628640815615654, "global_step": 211387, "epoch": 2375, "lr": 1.1642269232836434e-05}
{"train_loss": 0.0308286901563406, "global_step": 211388, "epoch": 2375, "lr": 1.1641897356278769e-05}
{"train_loss": 0.08254469931125641, "global_step": 211389, "epoch": 2375, "lr": 1.1641525484877863e-05}
{"train_loss": 0.03879115730524063, "global_step": 211390, "epoch": 2375, "lr": 1.164115361863381e-05}
{"train_loss": 0.04845995455980301, "global_step": 211391, "epoch": 2375, "lr": 1.1640781757546615e-05}
{"train_loss": 0.03234391286969185, "global_step": 211392, "epoch": 2375, "lr": 1.1640409901616366e-05}
{"train_loss": 0.04434794560074806, "global_step": 211393, "epoch": 2375, "lr": 1.1640038050843082e-05}
{"train_loss": 0.042834777384996414, "global_step": 211394, "epoch": 2375, "lr": 1.163966620522684e-05}
{"train_loss": 0.04273973032832146, "global_step": 211395, "epoch": 2375, "lr": 1.1639294364767667e-05}
{"train_loss": 0.0553896389901638, "global_step": 211396, "epoch": 2375, "lr": 1.1638922529465618e-05}
{"train_loss": 0.0317629799246788, "global_step": 211397, "epoch": 2375, "lr": 1.1638550699320766e-05}
{"train_loss": 0.02934667281806469, "global_step": 211398, "epoch": 2375, "lr": 1.1638178874333127e-05}
{"train_loss": 0.02407686784863472, "global_step": 211399, "epoch": 2375, "lr": 1.163780705450278e-05}
{"train_loss": 0.07889319211244583, "global_step": 211400, "epoch": 2375, "lr": 1.1637435239829753e-05}
{"train_loss": 0.04584628343582153, "global_step": 211401, "epoch": 2375, "lr": 1.1637063430314116e-05}
{"train_loss": 0.035749033093452454, "global_step": 211402, "epoch": 2375, "lr": 1.1636691625955898e-05}
{"train_loss": 0.031053004786372185, "global_step": 211403, "epoch": 2375, "lr": 1.1636319826755171e-05}
{"train_loss": 0.04168974235653877, "global_step": 211404, "epoch": 2375, "lr": 1.1635948032711958e-05}
{"train_loss": 0.04366634786128998, "global_step": 211405, "epoch": 2375, "lr": 1.163557624382634e-05}
{"train_loss": 0.08480043709278107, "global_step": 211406, "epoch": 2375, "lr": 1.1635204460098337e-05}
{"train_loss": 0.05615433678030968, "global_step": 211407, "epoch": 2375, "lr": 1.1634832681528024e-05}
{"train_loss": 0.0688377171754837, "global_step": 211408, "epoch": 2375, "lr": 1.163446090811543e-05}
{"train_loss": 0.06462465971708298, "global_step": 211409, "epoch": 2375, "lr": 1.1634089139860626e-05}
{"train_loss": 0.07901416718959808, "global_step": 211410, "epoch": 2375, "lr": 1.1633717376763641e-05}
{"train_loss": 0.023159697651863098, "global_step": 211411, "epoch": 2375, "lr": 1.1633345618824549e-05}
{"train_loss": 0.026256410405039787, "global_step": 211412, "epoch": 2375, "lr": 1.1632973866043368e-05}
{"train_loss": 0.022244790568947792, "global_step": 211413, "epoch": 2375, "lr": 1.1632602118420183e-05}
{"train_loss": 0.03953422233462334, "global_step": 211414, "epoch": 2375, "lr": 1.1632230375955012e-05}
{"train_loss": 0.03935837000608444, "global_step": 211415, "epoch": 2375, "lr": 1.163185863864793e-05}
{"train_loss": 0.059880491346120834, "global_step": 211416, "epoch": 2375, "lr": 1.1631486906498967e-05}
{"train_loss": 0.040668144822120667, "global_step": 211417, "epoch": 2375, "lr": 1.1631115179508202e-05}
{"train_loss": 0.04580414295196533, "global_step": 211418, "epoch": 2375, "lr": 1.1630743457675642e-05}
{"train_loss": 0.07962951809167862, "global_step": 211419, "epoch": 2375, "lr": 1.1630371741001384e-05}
{"train_loss": 0.011593041010200977, "global_step": 211420, "epoch": 2375, "lr": 1.163000002948545e-05}
{"train_loss": 0.09306497871875763, "global_step": 211421, "epoch": 2375, "lr": 1.1629628323127878e-05}
{"train_loss": 0.06940855085849762, "global_step": 211422, "epoch": 2375, "lr": 1.1629256621928753e-05}
{"train_loss": 0.060133934020996094, "global_step": 211423, "epoch": 2375, "lr": 1.1628884925888089e-05}
{"train_loss": 0.0905829519033432, "global_step": 211424, "epoch": 2375, "lr": 1.1628513235005972e-05}
{"train_loss": 0.03642448037862778, "global_step": 211425, "epoch": 2375, "lr": 1.1628141549282418e-05}
{"train_loss": 0.035270292311906815, "global_step": 211426, "epoch": 2375, "lr": 1.1627769868717508e-05}
{"train_loss": 0.07263627648353577, "global_step": 211427, "epoch": 2375, "lr": 1.1627398193311261e-05}
{"train_loss": 0.09036745876073837, "global_step": 211428, "epoch": 2375, "lr": 1.1627026523063762e-05}
{"train_loss": 0.06855897605419159, "global_step": 211429, "epoch": 2375, "lr": 1.1626654857975023e-05}
{"train_loss": 0.10083175450563431, "global_step": 211430, "epoch": 2375, "lr": 1.1626283198045123e-05}
{"train_loss": 0.028158968314528465, "global_step": 211431, "epoch": 2375, "lr": 1.1625911543274093e-05}
{"train_loss": 0.07624578475952148, "global_step": 211432, "epoch": 2375, "lr": 1.1625539893661985e-05}
{"train_loss": 0.01857682317495346, "global_step": 211433, "epoch": 2375, "lr": 1.1625168249208878e-05}
{"train_loss": 0.09607318788766861, "global_step": 211434, "epoch": 2375, "lr": 1.162479660991478e-05}
{"train_loss": 0.03206887096166611, "global_step": 211435, "epoch": 2375, "lr": 1.1624424975779774e-05}
{"train_loss": 0.04805913195014, "global_step": 211436, "epoch": 2375, "lr": 1.1624053346803887e-05}
{"train_loss": 0.05553552508354187, "global_step": 211437, "epoch": 2375, "lr": 1.162368172298719e-05}
{"train_loss": 0.014850586652755737, "global_step": 211438, "epoch": 2375, "lr": 1.1623310104329704e-05}
{"train_loss": 0.02845514938235283, "global_step": 211439, "epoch": 2375, "lr": 1.1622938490831514e-05}
{"train_loss": 0.05648662522435188, "global_step": 211440, "epoch": 2375, "lr": 1.1622566882492636e-05}
{"train_loss": 0.1175672635436058, "global_step": 211441, "epoch": 2375, "lr": 1.1622195279313153e-05}
{"train_loss": 0.046453043818473816, "global_step": 211442, "epoch": 2375, "lr": 1.1621823681293081e-05}
{"train_loss": 0.03493568301200867, "global_step": 211443, "epoch": 2375, "lr": 1.1621452088432499e-05}
{"train_loss": 0.026684623211622238, "global_step": 211444, "epoch": 2375, "lr": 1.1621080500731435e-05}
{"train_loss": 0.04473361000418663, "global_step": 211445, "epoch": 2375, "lr": 1.1620708918189965e-05}
{"train_loss": 0.05237611010670662, "global_step": 211446, "epoch": 2375, "lr": 1.1620337340808107e-05}
{"train_loss": 0.06447397172451019, "global_step": 211447, "epoch": 2375, "lr": 1.161996576858594e-05}
{"train_loss": 0.07650532573461533, "global_step": 211448, "epoch": 2375, "lr": 1.1619594201523482e-05}
{"train_loss": 0.06544557213783264, "global_step": 211449, "epoch": 2375, "lr": 1.1619222639620825e-05}
{"train_loss": 0.04137784615159035, "global_step": 211450, "epoch": 2375, "lr": 1.1618851082877973e-05}
{"train_loss": 0.03928336128592491, "global_step": 211451, "epoch": 2375, "lr": 1.1618479531295007e-05}
{"train_loss": 0.02131045237183571, "global_step": 211452, "epoch": 2375, "lr": 1.1618107984871978e-05}
{"train_loss": 0.05317053198814392, "global_step": 211453, "epoch": 2375, "lr": 1.1617736443608923e-05}
{"train_loss": 0.0701960101723671, "global_step": 211454, "epoch": 2375, "lr": 1.161736490750589e-05}
{"train_loss": 0.034468263387680054, "global_step": 211455, "epoch": 2375, "lr": 1.161699337656294e-05}
{"train_loss": 0.07544973492622375, "global_step": 211456, "epoch": 2375, "lr": 1.1616621850780106e-05}
{"train_loss": 0.0405537448823452, "global_step": 211457, "epoch": 2375, "lr": 1.1616250330157469e-05}
{"train_loss": 0.020534371957182884, "global_step": 211458, "epoch": 2375, "lr": 1.1615878814695042e-05}
{"train_loss": 0.04092054441571236, "global_step": 211459, "epoch": 2375, "lr": 1.1615507304392902e-05}
{"train_loss": 0.014307946898043156, "global_step": 211460, "epoch": 2375, "lr": 1.1615135799251093e-05}
{"train_loss": 0.09460998326539993, "global_step": 211461, "epoch": 2375, "lr": 1.1614764299269648e-05}
{"train_loss": 0.04185473918914795, "global_step": 211462, "epoch": 2375, "lr": 1.1614392804448644e-05}
{"train_loss": 0.04981117687198553, "global_step": 211463, "epoch": 2375, "lr": 1.1614021314788099e-05, "val_loss": 8.734375, "train_action_mse_error": 15.76382064819336}
{"train_loss": 0.052973754703998566, "global_step": 211464, "epoch": 2376, "lr": 1.1613649830288098e-05}
{"train_loss": 0.061495378613471985, "global_step": 211465, "epoch": 2376, "lr": 1.1613278350948658e-05}
{"train_loss": 0.05681534856557846, "global_step": 211466, "epoch": 2376, "lr": 1.1612906876769863e-05}
{"train_loss": 0.051527444273233414, "global_step": 211467, "epoch": 2376, "lr": 1.1612535407751729e-05}
{"train_loss": 0.07860118895769119, "global_step": 211468, "epoch": 2376, "lr": 1.1612163943894322e-05}
{"train_loss": 0.09125502407550812, "global_step": 211469, "epoch": 2376, "lr": 1.1611792485197704e-05}
{"train_loss": 0.061834629625082016, "global_step": 211470, "epoch": 2376, "lr": 1.1611421031661901e-05}
{"train_loss": 0.05721179023385048, "global_step": 211471, "epoch": 2376, "lr": 1.1611049583286988e-05}
{"train_loss": 0.10527238249778748, "global_step": 211472, "epoch": 2376, "lr": 1.1610678140072983e-05}
{"train_loss": 0.05240636691451073, "global_step": 211473, "epoch": 2376, "lr": 1.1610306702019974e-05}
{"train_loss": 0.023354938253760338, "global_step": 211474, "epoch": 2376, "lr": 1.1609935269127974e-05}
{"train_loss": 0.054049715399742126, "global_step": 211475, "epoch": 2376, "lr": 1.1609563841397065e-05}
{"train_loss": 0.025057177990674973, "global_step": 211476, "epoch": 2376, "lr": 1.1609192418827269e-05}
{"train_loss": 0.03971880301833153, "global_step": 211477, "epoch": 2376, "lr": 1.1608821001418662e-05}
{"train_loss": 0.04638274386525154, "global_step": 211478, "epoch": 2376, "lr": 1.1608449589171272e-05}
{"train_loss": 0.03329227492213249, "global_step": 211479, "epoch": 2376, "lr": 1.1608078182085163e-05}
{"train_loss": 0.06137721985578537, "global_step": 211480, "epoch": 2376, "lr": 1.160770678016037e-05}
{"train_loss": 0.04064914956688881, "global_step": 211481, "epoch": 2376, "lr": 1.1607335383396972e-05}
{"train_loss": 0.039363108575344086, "global_step": 211482, "epoch": 2376, "lr": 1.1606963991794984e-05}
{"train_loss": 0.05482713133096695, "global_step": 211483, "epoch": 2376, "lr": 1.1606592605354483e-05}
{"train_loss": 0.07994310557842255, "global_step": 211484, "epoch": 2376, "lr": 1.1606221224075492e-05}
{"train_loss": 0.10284583270549774, "global_step": 211485, "epoch": 2376, "lr": 1.1605849847958094e-05}
{"train_loss": 0.06066305562853813, "global_step": 211486, "epoch": 2376, "lr": 1.1605478477002307e-05}
{"train_loss": 0.04497945308685303, "global_step": 211487, "epoch": 2376, "lr": 1.1605107111208192e-05}
{"train_loss": 0.06830962002277374, "global_step": 211488, "epoch": 2376, "lr": 1.1604735750575818e-05}
{"train_loss": 0.059759367257356644, "global_step": 211489, "epoch": 2376, "lr": 1.160436439510521e-05}
{"train_loss": 0.08567085862159729, "global_step": 211490, "epoch": 2376, "lr": 1.1603993044796435e-05}
{"train_loss": 0.060728758573532104, "global_step": 211491, "epoch": 2376, "lr": 1.1603621699649525e-05}
{"train_loss": 0.07775086909532547, "global_step": 211492, "epoch": 2376, "lr": 1.1603250359664553e-05}
{"train_loss": 0.046564359217882156, "global_step": 211493, "epoch": 2376, "lr": 1.1602879024841557e-05}
{"train_loss": 0.06961087137460709, "global_step": 211494, "epoch": 2376, "lr": 1.160250769518057e-05}
{"train_loss": 0.05784818157553673, "global_step": 211495, "epoch": 2376, "lr": 1.160213637068167e-05}
{"train_loss": 0.055478714406490326, "global_step": 211496, "epoch": 2376, "lr": 1.1601765051344886e-05}
{"train_loss": 0.055270902812480927, "global_step": 211497, "epoch": 2376, "lr": 1.1601393737170285e-05}
{"train_loss": 0.048657577484846115, "global_step": 211498, "epoch": 2376, "lr": 1.1601022428157916e-05}
{"train_loss": 0.03891468048095703, "global_step": 211499, "epoch": 2376, "lr": 1.1600651124307799e-05}
{"train_loss": 0.07263527810573578, "global_step": 211500, "epoch": 2376, "lr": 1.1600279825620024e-05}
{"train_loss": 0.04304885491728783, "global_step": 211501, "epoch": 2376, "lr": 1.159990853209461e-05}
{"train_loss": 0.06810234487056732, "global_step": 211502, "epoch": 2376, "lr": 1.159953724373164e-05}
{"train_loss": 0.03413984552025795, "global_step": 211503, "epoch": 2376, "lr": 1.1599165960531122e-05}
{"train_loss": 0.06255602836608887, "global_step": 211504, "epoch": 2376, "lr": 1.159879468249313e-05}
{"train_loss": 0.08184175193309784, "global_step": 211505, "epoch": 2376, "lr": 1.1598423409617725e-05}
{"train_loss": 0.040740467607975006, "global_step": 211506, "epoch": 2376, "lr": 1.1598052141904931e-05}
{"train_loss": 0.07423409074544907, "global_step": 211507, "epoch": 2376, "lr": 1.1597680879354828e-05}
{"train_loss": 0.04035654664039612, "global_step": 211508, "epoch": 2376, "lr": 1.1597309621967428e-05}
{"train_loss": 0.03565141186118126, "global_step": 211509, "epoch": 2376, "lr": 1.1596938369742822e-05}
{"train_loss": 0.04829220101237297, "global_step": 211510, "epoch": 2376, "lr": 1.1596567122681023e-05}
{"train_loss": 0.04692558944225311, "global_step": 211511, "epoch": 2376, "lr": 1.1596195880782113e-05}
{"train_loss": 0.058411192148923874, "global_step": 211512, "epoch": 2376, "lr": 1.159582464404611e-05}
{"train_loss": 0.05183195695281029, "global_step": 211513, "epoch": 2376, "lr": 1.1595453412473094e-05}
{"train_loss": 0.03292515501379967, "global_step": 211514, "epoch": 2376, "lr": 1.1595082186063088e-05}
{"train_loss": 0.04647618904709816, "global_step": 211515, "epoch": 2376, "lr": 1.1594710964816175e-05}
{"train_loss": 0.05314696207642555, "global_step": 211516, "epoch": 2376, "lr": 1.1594339748732363e-05}
{"train_loss": 0.06066671013832092, "global_step": 211517, "epoch": 2376, "lr": 1.1593968537811739e-05}
{"train_loss": 0.04152364283800125, "global_step": 211518, "epoch": 2376, "lr": 1.1593597332054329e-05}
{"train_loss": 0.09607553482055664, "global_step": 211519, "epoch": 2376, "lr": 1.15932261314602e-05}
{"train_loss": 0.02683866396546364, "global_step": 211520, "epoch": 2376, "lr": 1.1592854936029384e-05}
{"train_loss": 0.03517748787999153, "global_step": 211521, "epoch": 2376, "lr": 1.159248374576195e-05}
{"train_loss": 0.06251592934131622, "global_step": 211522, "epoch": 2376, "lr": 1.1592112560657925e-05}
{"train_loss": 0.10014422982931137, "global_step": 211523, "epoch": 2376, "lr": 1.1591741380717374e-05}
{"train_loss": 0.05978171154856682, "global_step": 211524, "epoch": 2376, "lr": 1.159137020594036e-05}
{"train_loss": 0.013842177577316761, "global_step": 211525, "epoch": 2376, "lr": 1.1590999036326906e-05}
{"train_loss": 0.023157380521297455, "global_step": 211526, "epoch": 2376, "lr": 1.1590627871877085e-05}
{"train_loss": 0.03392554074525833, "global_step": 211527, "epoch": 2376, "lr": 1.1590256712590925e-05}
{"train_loss": 0.039210524410009384, "global_step": 211528, "epoch": 2376, "lr": 1.1589885558468493e-05}
{"train_loss": 0.052765052765607834, "global_step": 211529, "epoch": 2376, "lr": 1.1589514409509828e-05}
{"train_loss": 0.05503159388899803, "global_step": 211530, "epoch": 2376, "lr": 1.1589143265714991e-05}
{"train_loss": 0.08118893206119537, "global_step": 211531, "epoch": 2376, "lr": 1.158877212708403e-05}
{"train_loss": 0.08687940984964371, "global_step": 211532, "epoch": 2376, "lr": 1.1588400993616977e-05}
{"train_loss": 0.05210826173424721, "global_step": 211533, "epoch": 2376, "lr": 1.158802986531391e-05}
{"train_loss": 0.06932417303323746, "global_step": 211534, "epoch": 2376, "lr": 1.1587658742174845e-05}
{"train_loss": 0.07587041705846786, "global_step": 211535, "epoch": 2376, "lr": 1.1587287624199871e-05}
{"train_loss": 0.04492955282330513, "global_step": 211536, "epoch": 2376, "lr": 1.1586916511389013e-05}
{"train_loss": 0.06067328527569771, "global_step": 211537, "epoch": 2376, "lr": 1.1586545403742316e-05}
{"train_loss": 0.07295666635036469, "global_step": 211538, "epoch": 2376, "lr": 1.158617430125985e-05}
{"train_loss": 0.041617970913648605, "global_step": 211539, "epoch": 2376, "lr": 1.1585803203941642e-05}
{"train_loss": 0.03736003488302231, "global_step": 211540, "epoch": 2376, "lr": 1.158543211178777e-05}
{"train_loss": 0.08446203172206879, "global_step": 211541, "epoch": 2376, "lr": 1.1585061024798255e-05}
{"train_loss": 0.025300413370132446, "global_step": 211542, "epoch": 2376, "lr": 1.158468994297316e-05}
{"train_loss": 0.014630184508860111, "global_step": 211543, "epoch": 2376, "lr": 1.1584318866312548e-05}
{"train_loss": 0.06512514501810074, "global_step": 211544, "epoch": 2376, "lr": 1.158394779481644e-05}
{"train_loss": 0.07912462949752808, "global_step": 211545, "epoch": 2376, "lr": 1.1583576728484918e-05}
{"train_loss": 0.08272319287061691, "global_step": 211546, "epoch": 2376, "lr": 1.1583205667318004e-05}
{"train_loss": 0.09728296101093292, "global_step": 211547, "epoch": 2376, "lr": 1.1582834611315774e-05}
{"train_loss": 0.060333672910928726, "global_step": 211548, "epoch": 2376, "lr": 1.1582463560478245e-05}
{"train_loss": 0.03158950433135033, "global_step": 211549, "epoch": 2376, "lr": 1.1582092514805504e-05}
{"train_loss": 0.04989905282855034, "global_step": 211550, "epoch": 2376, "lr": 1.1581721474297563e-05}
{"train_loss": 0.020341521129012108, "global_step": 211551, "epoch": 2376, "lr": 1.1581350438954514e-05}
{"train_loss": 0.05636633717025934, "global_step": 211552, "epoch": 2376, "lr": 1.158097940877636e-05, "val_loss": 8.690256118774414}
{"train_loss": 0.04580315202474594, "global_step": 211553, "epoch": 2377, "lr": 1.1580608383763197e-05}
{"train_loss": 0.0617431104183197, "global_step": 211554, "epoch": 2377, "lr": 1.1580237363915037e-05}
{"train_loss": 0.031656570732593536, "global_step": 211555, "epoch": 2377, "lr": 1.1579866349231955e-05}
{"train_loss": 0.03550141677260399, "global_step": 211556, "epoch": 2377, "lr": 1.1579495339713981e-05}
{"train_loss": 0.06500867009162903, "global_step": 211557, "epoch": 2377, "lr": 1.157912433536119e-05}
{"train_loss": 0.036973290145397186, "global_step": 211558, "epoch": 2377, "lr": 1.1578753336173604e-05}
{"train_loss": 0.11356198042631149, "global_step": 211559, "epoch": 2377, "lr": 1.157838234215129e-05}
{"train_loss": 0.057579826563596725, "global_step": 211560, "epoch": 2377, "lr": 1.15780113532943e-05}
{"train_loss": 0.08369193226099014, "global_step": 211561, "epoch": 2377, "lr": 1.1577640369602671e-05}
{"train_loss": 0.038935378193855286, "global_step": 211562, "epoch": 2377, "lr": 1.1577269391076467e-05}
{"train_loss": 0.03850407153367996, "global_step": 211563, "epoch": 2377, "lr": 1.1576898417715726e-05}
{"train_loss": 0.04765378683805466, "global_step": 211564, "epoch": 2377, "lr": 1.1576527449520508e-05}
{"train_loss": 0.03422269970178604, "global_step": 211565, "epoch": 2377, "lr": 1.1576156486490846e-05}
{"train_loss": 0.12416257709264755, "global_step": 211566, "epoch": 2377, "lr": 1.157578552862682e-05}
{"train_loss": 0.09360311180353165, "global_step": 211567, "epoch": 2377, "lr": 1.1575414575928446e-05}
{"train_loss": 0.04462926834821701, "global_step": 211568, "epoch": 2377, "lr": 1.1575043628395798e-05}
{"train_loss": 0.027269016951322556, "global_step": 211569, "epoch": 2377, "lr": 1.157467268602892e-05}
{"train_loss": 0.07770581543445587, "global_step": 211570, "epoch": 2377, "lr": 1.1574301748827842e-05}
{"train_loss": 0.09379059821367264, "global_step": 211571, "epoch": 2377, "lr": 1.157393081679265e-05}
{"train_loss": 0.11399008333683014, "global_step": 211572, "epoch": 2377, "lr": 1.1573559889923358e-05}
{"train_loss": 0.09610029309988022, "global_step": 211573, "epoch": 2377, "lr": 1.157318896822005e-05}
{"train_loss": 0.08800923824310303, "global_step": 211574, "epoch": 2377, "lr": 1.1572818051682738e-05}
{"train_loss": 0.021411797031760216, "global_step": 211575, "epoch": 2377, "lr": 1.157244714031151e-05}
{"train_loss": 0.025736913084983826, "global_step": 211576, "epoch": 2377, "lr": 1.15720762341064e-05}
{"train_loss": 0.09358783811330795, "global_step": 211577, "epoch": 2377, "lr": 1.157170533306744e-05}
{"train_loss": 0.04404204711318016, "global_step": 211578, "epoch": 2377, "lr": 1.1571334437194692e-05}
{"train_loss": 0.12353628873825073, "global_step": 211579, "epoch": 2377, "lr": 1.1570963546488228e-05}
{"train_loss": 0.033453088253736496, "global_step": 211580, "epoch": 2377, "lr": 1.1570592660948066e-05}
{"train_loss": 0.07255799323320389, "global_step": 211581, "epoch": 2377, "lr": 1.157022178057428e-05}
{"train_loss": 0.05156590789556503, "global_step": 211582, "epoch": 2377, "lr": 1.1569850905366897e-05}
{"train_loss": 0.04746144264936447, "global_step": 211583, "epoch": 2377, "lr": 1.1569480035325997e-05}
{"train_loss": 0.025107864290475845, "global_step": 211584, "epoch": 2377, "lr": 1.1569109170451591e-05}
{"train_loss": 0.07747343182563782, "global_step": 211585, "epoch": 2377, "lr": 1.1568738310743771e-05}
{"train_loss": 0.015354681760072708, "global_step": 211586, "epoch": 2377, "lr": 1.156836745620255e-05}
{"train_loss": 0.08688496798276901, "global_step": 211587, "epoch": 2377, "lr": 1.1567996606828002e-05}
{"train_loss": 0.04121806472539902, "global_step": 211588, "epoch": 2377, "lr": 1.156762576262016e-05}
{"train_loss": 0.07999813556671143, "global_step": 211589, "epoch": 2377, "lr": 1.1567254923579096e-05}
{"train_loss": 0.04895550757646561, "global_step": 211590, "epoch": 2377, "lr": 1.1566884089704827e-05}
{"train_loss": 0.05077340453863144, "global_step": 211591, "epoch": 2377, "lr": 1.1566513260997442e-05}
{"train_loss": 0.03779836744070053, "global_step": 211592, "epoch": 2377, "lr": 1.1566142437456956e-05}
{"train_loss": 0.1019468605518341, "global_step": 211593, "epoch": 2377, "lr": 1.1565771619083443e-05}
{"train_loss": 0.06951043754816055, "global_step": 211594, "epoch": 2377, "lr": 1.1565400805876935e-05}
{"train_loss": 0.0368955172598362, "global_step": 211595, "epoch": 2377, "lr": 1.156502999783749e-05}
{"train_loss": 0.029423534870147705, "global_step": 211596, "epoch": 2377, "lr": 1.1564659194965166e-05}
{"train_loss": 0.049691375344991684, "global_step": 211597, "epoch": 2377, "lr": 1.1564288397259998e-05}
{"train_loss": 0.04344065114855766, "global_step": 211598, "epoch": 2377, "lr": 1.1563917604722052e-05}
{"train_loss": 0.11850039660930634, "global_step": 211599, "epoch": 2377, "lr": 1.1563546817351362e-05}
{"train_loss": 0.05844448506832123, "global_step": 211600, "epoch": 2377, "lr": 1.1563176035147994e-05}
{"train_loss": 0.06063824146986008, "global_step": 211601, "epoch": 2377, "lr": 1.1562805258111975e-05}
{"train_loss": 0.04970606416463852, "global_step": 211602, "epoch": 2377, "lr": 1.156243448624338e-05}
{"train_loss": 0.0479767806828022, "global_step": 211603, "epoch": 2377, "lr": 1.1562063719542232e-05}
{"train_loss": 0.04173673316836357, "global_step": 211604, "epoch": 2377, "lr": 1.1561692958008618e-05}
{"train_loss": 0.081510528922081, "global_step": 211605, "epoch": 2377, "lr": 1.1561322201642543e-05}
{"train_loss": 0.13763196766376495, "global_step": 211606, "epoch": 2377, "lr": 1.1560951450444103e-05}
{"train_loss": 0.12331259995698929, "global_step": 211607, "epoch": 2377, "lr": 1.1560580704413305e-05}
{"train_loss": 0.09159926325082779, "global_step": 211608, "epoch": 2377, "lr": 1.1560209963550228e-05}
{"train_loss": 0.05647493526339531, "global_step": 211609, "epoch": 2377, "lr": 1.1559839227854918e-05}
{"train_loss": 0.039500899612903595, "global_step": 211610, "epoch": 2377, "lr": 1.1559468497327403e-05}
{"train_loss": 0.06944113969802856, "global_step": 211611, "epoch": 2377, "lr": 1.1559097771967764e-05}
{"train_loss": 0.08120277523994446, "global_step": 211612, "epoch": 2377, "lr": 1.1558727051776024e-05}
{"train_loss": 0.04591226950287819, "global_step": 211613, "epoch": 2377, "lr": 1.1558356336752257e-05}
{"train_loss": 0.03364820405840874, "global_step": 211614, "epoch": 2377, "lr": 1.1557985626896483e-05}
{"train_loss": 0.1140923872590065, "global_step": 211615, "epoch": 2377, "lr": 1.1557614922208788e-05}
{"train_loss": 0.057864367961883545, "global_step": 211616, "epoch": 2377, "lr": 1.1557244222689189e-05}
{"train_loss": 0.06664888560771942, "global_step": 211617, "epoch": 2377, "lr": 1.1556873528337759e-05}
{"train_loss": 0.0623759850859642, "global_step": 211618, "epoch": 2377, "lr": 1.1556502839154532e-05}
{"train_loss": 0.030227908864617348, "global_step": 211619, "epoch": 2377, "lr": 1.155613215513957e-05}
{"train_loss": 0.06548570096492767, "global_step": 211620, "epoch": 2377, "lr": 1.1555761476292908e-05}
{"train_loss": 0.055252961814403534, "global_step": 211621, "epoch": 2377, "lr": 1.1555390802614625e-05}
{"train_loss": 0.06408004462718964, "global_step": 211622, "epoch": 2377, "lr": 1.1555020134104728e-05}
{"train_loss": 0.058637119829654694, "global_step": 211623, "epoch": 2377, "lr": 1.1554649470763308e-05}
{"train_loss": 0.03619895875453949, "global_step": 211624, "epoch": 2377, "lr": 1.155427881259038e-05}
{"train_loss": 0.036927420645952225, "global_step": 211625, "epoch": 2377, "lr": 1.155390815958603e-05}
{"train_loss": 0.06435134261846542, "global_step": 211626, "epoch": 2377, "lr": 1.1553537511750273e-05}
{"train_loss": 0.03881698474287987, "global_step": 211627, "epoch": 2377, "lr": 1.1553166869083187e-05}
{"train_loss": 0.068557970225811, "global_step": 211628, "epoch": 2377, "lr": 1.1552796231584794e-05}
{"train_loss": 0.0438864603638649, "global_step": 211629, "epoch": 2377, "lr": 1.1552425599255173e-05}
{"train_loss": 0.038754940032958984, "global_step": 211630, "epoch": 2377, "lr": 1.1552054972094345e-05}
{"train_loss": 0.05272899940609932, "global_step": 211631, "epoch": 2377, "lr": 1.1551684350102376e-05}
{"train_loss": 0.06917429715394974, "global_step": 211632, "epoch": 2377, "lr": 1.1551313733279329e-05}
{"train_loss": 0.04287539795041084, "global_step": 211633, "epoch": 2377, "lr": 1.1550943121625225e-05}
{"train_loss": 0.034340932965278625, "global_step": 211634, "epoch": 2377, "lr": 1.155057251514014e-05}
{"train_loss": 0.04549698904156685, "global_step": 211635, "epoch": 2377, "lr": 1.1550201913824098e-05}
{"train_loss": 0.08347507566213608, "global_step": 211636, "epoch": 2377, "lr": 1.1549831317677185e-05}
{"train_loss": 0.06421355158090591, "global_step": 211637, "epoch": 2377, "lr": 1.1549460726699407e-05}
{"train_loss": 0.037700068205595016, "global_step": 211638, "epoch": 2377, "lr": 1.154909014089085e-05}
{"train_loss": 0.03461039438843727, "global_step": 211639, "epoch": 2377, "lr": 1.1548719560251536e-05}
{"train_loss": 0.038100335747003555, "global_step": 211640, "epoch": 2377, "lr": 1.1548348984781537e-05}
{"train_loss": 0.06123165597908952, "global_step": 211641, "epoch": 2377, "lr": 1.1547978414480887e-05, "val_loss": 8.758393287658691}
{"train_loss": 0.016935035586357117, "global_step": 211642, "epoch": 2378, "lr": 1.1547607849349657e-05}
{"train_loss": 0.05767830088734627, "global_step": 211643, "epoch": 2378, "lr": 1.1547237289387863e-05}
{"train_loss": 0.08225282281637192, "global_step": 211644, "epoch": 2378, "lr": 1.154686673459559e-05}
{"train_loss": 0.056445568799972534, "global_step": 211645, "epoch": 2378, "lr": 1.154649618497286e-05}
{"train_loss": 0.09204293042421341, "global_step": 211646, "epoch": 2378, "lr": 1.1546125640519745e-05}
{"train_loss": 0.03191423416137695, "global_step": 211647, "epoch": 2378, "lr": 1.154575510123629e-05}
{"train_loss": 0.06147315353155136, "global_step": 211648, "epoch": 2378, "lr": 1.1545384567122525e-05}
{"train_loss": 0.03868826851248741, "global_step": 211649, "epoch": 2378, "lr": 1.1545014038178525e-05}
{"train_loss": 0.08563646674156189, "global_step": 211650, "epoch": 2378, "lr": 1.1544643514404312e-05}
{"train_loss": 0.06769654154777527, "global_step": 211651, "epoch": 2378, "lr": 1.1544272995799977e-05}
{"train_loss": 0.067423515021801, "global_step": 211652, "epoch": 2378, "lr": 1.1543902482365526e-05}
{"train_loss": 0.07636339217424393, "global_step": 211653, "epoch": 2378, "lr": 1.154353197410104e-05}
{"train_loss": 0.09682833403348923, "global_step": 211654, "epoch": 2378, "lr": 1.1543161471006548e-05}
{"train_loss": 0.04407630115747452, "global_step": 211655, "epoch": 2378, "lr": 1.1542790973082119e-05}
{"train_loss": 0.06543532758951187, "global_step": 211656, "epoch": 2378, "lr": 1.1542420480327781e-05}
{"train_loss": 0.05576753616333008, "global_step": 211657, "epoch": 2378, "lr": 1.1542049992743613e-05}
{"train_loss": 0.03026781976222992, "global_step": 211658, "epoch": 2378, "lr": 1.1541679510329628e-05}
{"train_loss": 0.044001899659633636, "global_step": 211659, "epoch": 2378, "lr": 1.1541309033085912e-05}
{"train_loss": 0.08525033295154572, "global_step": 211660, "epoch": 2378, "lr": 1.1540938561012488e-05}
{"train_loss": 0.07290631532669067, "global_step": 211661, "epoch": 2378, "lr": 1.1540568094109428e-05}
{"train_loss": 0.05101726949214935, "global_step": 211662, "epoch": 2378, "lr": 1.1540197632376753e-05}
{"train_loss": 0.09680274873971939, "global_step": 211663, "epoch": 2378, "lr": 1.1539827175814544e-05}
{"train_loss": 0.03854895010590553, "global_step": 211664, "epoch": 2378, "lr": 1.1539456724422826e-05}
{"train_loss": 0.11481350660324097, "global_step": 211665, "epoch": 2378, "lr": 1.1539086278201672e-05}
{"train_loss": 0.08732903748750687, "global_step": 211666, "epoch": 2378, "lr": 1.1538715837151105e-05}
{"train_loss": 0.08585356175899506, "global_step": 211667, "epoch": 2378, "lr": 1.15383454012712e-05}
{"train_loss": 0.06407543271780014, "global_step": 211668, "epoch": 2378, "lr": 1.1537974970561982e-05}
{"train_loss": 0.03631264716386795, "global_step": 211669, "epoch": 2378, "lr": 1.1537604545023516e-05}
{"train_loss": 0.06441082060337067, "global_step": 211670, "epoch": 2378, "lr": 1.1537234124655865e-05}
{"train_loss": 0.04046962410211563, "global_step": 211671, "epoch": 2378, "lr": 1.153686370945905e-05}
{"train_loss": 0.10368137806653976, "global_step": 211672, "epoch": 2378, "lr": 1.1536493299433148e-05}
{"train_loss": 0.03773662820458412, "global_step": 211673, "epoch": 2378, "lr": 1.1536122894578184e-05}
{"train_loss": 0.04211462289094925, "global_step": 211674, "epoch": 2378, "lr": 1.153575249489423e-05}
{"train_loss": 0.053366631269454956, "global_step": 211675, "epoch": 2378, "lr": 1.153538210038131e-05}
{"train_loss": 0.02731078304350376, "global_step": 211676, "epoch": 2378, "lr": 1.1535011711039506e-05}
{"train_loss": 0.025776918977499008, "global_step": 211677, "epoch": 2378, "lr": 1.1534641326868839e-05}
{"train_loss": 0.055724963545799255, "global_step": 211678, "epoch": 2378, "lr": 1.153427094786938e-05}
{"train_loss": 0.057162582874298096, "global_step": 211679, "epoch": 2378, "lr": 1.1533900574041157e-05}
{"train_loss": 0.06877486407756805, "global_step": 211680, "epoch": 2378, "lr": 1.1533530205384247e-05}
{"train_loss": 0.04670815169811249, "global_step": 211681, "epoch": 2378, "lr": 1.1533159841898672e-05}
{"train_loss": 0.09312178194522858, "global_step": 211682, "epoch": 2378, "lr": 1.153278948358451e-05}
{"train_loss": 0.06615234911441803, "global_step": 211683, "epoch": 2378, "lr": 1.153241913044178e-05}
{"train_loss": 0.06837144494056702, "global_step": 211684, "epoch": 2378, "lr": 1.1532048782470562e-05}
{"train_loss": 0.10547464340925217, "global_step": 211685, "epoch": 2378, "lr": 1.1531678439670872e-05}
{"train_loss": 0.027743011713027954, "global_step": 211686, "epoch": 2378, "lr": 1.1531308102042799e-05}
{"train_loss": 0.09927670657634735, "global_step": 211687, "epoch": 2378, "lr": 1.153093776958637e-05}
{"train_loss": 0.03472752496600151, "global_step": 211688, "epoch": 2378, "lr": 1.1530567442301626e-05}
{"train_loss": 0.041911348700523376, "global_step": 211689, "epoch": 2378, "lr": 1.1530197120188641e-05}
{"train_loss": 0.08261227607727051, "global_step": 211690, "epoch": 2378, "lr": 1.1529826803247434e-05}
{"train_loss": 0.05467751622200012, "global_step": 211691, "epoch": 2378, "lr": 1.1529456491478091e-05}
{"train_loss": 0.0171924140304327, "global_step": 211692, "epoch": 2378, "lr": 1.1529086184880632e-05}
{"train_loss": 0.057944536209106445, "global_step": 211693, "epoch": 2378, "lr": 1.1528715883455126e-05}
{"train_loss": 0.12069162726402283, "global_step": 211694, "epoch": 2378, "lr": 1.1528345587201605e-05}
{"train_loss": 0.04421565681695938, "global_step": 211695, "epoch": 2378, "lr": 1.1527975296120147e-05}
{"train_loss": 0.04147488996386528, "global_step": 211696, "epoch": 2378, "lr": 1.1527605010210762e-05}
{"train_loss": 0.08628999441862106, "global_step": 211697, "epoch": 2378, "lr": 1.1527234729473541e-05}
{"train_loss": 0.0588306300342083, "global_step": 211698, "epoch": 2378, "lr": 1.1526864453908498e-05}
{"train_loss": 0.08366720378398895, "global_step": 211699, "epoch": 2378, "lr": 1.1526494183515718e-05}
{"train_loss": 0.11402832716703415, "global_step": 211700, "epoch": 2378, "lr": 1.1526123918295212e-05}
{"train_loss": 0.05573883280158043, "global_step": 211701, "epoch": 2378, "lr": 1.1525753658247063e-05}
{"train_loss": 0.06748928129673004, "global_step": 211702, "epoch": 2378, "lr": 1.1525383403371298e-05}
{"train_loss": 0.05553114041686058, "global_step": 211703, "epoch": 2378, "lr": 1.152501315366799e-05}
{"train_loss": 0.03204290568828583, "global_step": 211704, "epoch": 2378, "lr": 1.1524642909137157e-05}
{"train_loss": 0.03864465653896332, "global_step": 211705, "epoch": 2378, "lr": 1.1524272669778868e-05}
{"train_loss": 0.04808706417679787, "global_step": 211706, "epoch": 2378, "lr": 1.1523902435593187e-05}
{"train_loss": 0.071364626288414, "global_step": 211707, "epoch": 2378, "lr": 1.1523532206580135e-05}
{"train_loss": 0.0708836019039154, "global_step": 211708, "epoch": 2378, "lr": 1.1523161982739789e-05}
{"train_loss": 0.04127245023846626, "global_step": 211709, "epoch": 2378, "lr": 1.1522791764072172e-05}
{"train_loss": 0.036007799208164215, "global_step": 211710, "epoch": 2378, "lr": 1.1522421550577356e-05}
{"train_loss": 0.05011148750782013, "global_step": 211711, "epoch": 2378, "lr": 1.1522051342255374e-05}
{"train_loss": 0.09894955158233643, "global_step": 211712, "epoch": 2378, "lr": 1.1521681139106295e-05}
{"train_loss": 0.06866741925477982, "global_step": 211713, "epoch": 2378, "lr": 1.1521310941130137e-05}
{"train_loss": 0.08624224364757538, "global_step": 211714, "epoch": 2378, "lr": 1.1520940748326991e-05}
{"train_loss": 0.02824726514518261, "global_step": 211715, "epoch": 2378, "lr": 1.1520570560696869e-05}
{"train_loss": 0.0463864803314209, "global_step": 211716, "epoch": 2378, "lr": 1.1520200378239849e-05}
{"train_loss": 0.08294227719306946, "global_step": 211717, "epoch": 2378, "lr": 1.1519830200955956e-05}
{"train_loss": 0.03844978287816048, "global_step": 211718, "epoch": 2378, "lr": 1.151946002884527e-05}
{"train_loss": 0.12027174234390259, "global_step": 211719, "epoch": 2378, "lr": 1.1519089861907806e-05}
{"train_loss": 0.0658152624964714, "global_step": 211720, "epoch": 2378, "lr": 1.1518719700143649e-05}
{"train_loss": 0.05546994134783745, "global_step": 211721, "epoch": 2378, "lr": 1.1518349543552815e-05}
{"train_loss": 0.038594700396060944, "global_step": 211722, "epoch": 2378, "lr": 1.151797939213538e-05}
{"train_loss": 0.06437195092439651, "global_step": 211723, "epoch": 2378, "lr": 1.1517609245891375e-05}
{"train_loss": 0.07719246298074722, "global_step": 211724, "epoch": 2378, "lr": 1.151723910482087e-05}
{"train_loss": 0.07778290659189224, "global_step": 211725, "epoch": 2378, "lr": 1.1516868968923904e-05}
{"train_loss": 0.06299304217100143, "global_step": 211726, "epoch": 2378, "lr": 1.1516498838200513e-05}
{"train_loss": 0.04286816343665123, "global_step": 211727, "epoch": 2378, "lr": 1.1516128712650775e-05}
{"train_loss": 0.04332764819264412, "global_step": 211728, "epoch": 2378, "lr": 1.1515758592274706e-05}
{"train_loss": 0.08138036727905273, "global_step": 211729, "epoch": 2378, "lr": 1.1515388477072392e-05}
{"train_loss": 0.062153898870174806, "global_step": 211730, "epoch": 2378, "lr": 1.151501836704385e-05, "val_loss": 8.762970924377441}
{"train_loss": 0.025549758225679398, "global_step": 211731, "epoch": 2379, "lr": 1.151464826218916e-05}
{"train_loss": 0.06091225519776344, "global_step": 211732, "epoch": 2379, "lr": 1.1514278162508335e-05}
{"train_loss": 0.04584885388612747, "global_step": 211733, "epoch": 2379, "lr": 1.1513908068001467e-05}
{"train_loss": 0.06123783439397812, "global_step": 211734, "epoch": 2379, "lr": 1.1513537978668571e-05}
{"train_loss": 0.07338447868824005, "global_step": 211735, "epoch": 2379, "lr": 1.1513167894509725e-05}
{"train_loss": 0.04305027052760124, "global_step": 211736, "epoch": 2379, "lr": 1.1512797815524946e-05}
{"train_loss": 0.041170667856931686, "global_step": 211737, "epoch": 2379, "lr": 1.1512427741714321e-05}
{"train_loss": 0.09137749671936035, "global_step": 211738, "epoch": 2379, "lr": 1.1512057673077864e-05}
{"train_loss": 0.03952272981405258, "global_step": 211739, "epoch": 2379, "lr": 1.1511687609615656e-05}
{"train_loss": 0.10202634334564209, "global_step": 211740, "epoch": 2379, "lr": 1.1511317551327722e-05}
{"train_loss": 0.09319997578859329, "global_step": 211741, "epoch": 2379, "lr": 1.151094749821412e-05}
{"train_loss": 0.06481420993804932, "global_step": 211742, "epoch": 2379, "lr": 1.1510577450274912e-05}
{"train_loss": 0.020821239799261093, "global_step": 211743, "epoch": 2379, "lr": 1.1510207407510132e-05}
{"train_loss": 0.0344979427754879, "global_step": 211744, "epoch": 2379, "lr": 1.150983736991984e-05}
{"train_loss": 0.06128571927547455, "global_step": 211745, "epoch": 2379, "lr": 1.1509467337504076e-05}
{"train_loss": 0.024962494149804115, "global_step": 211746, "epoch": 2379, "lr": 1.1509097310262907e-05}
{"train_loss": 0.05560275539755821, "global_step": 211747, "epoch": 2379, "lr": 1.1508727288196353e-05}
{"train_loss": 0.05043758824467659, "global_step": 211748, "epoch": 2379, "lr": 1.1508357271304498e-05}
{"train_loss": 0.07616548985242844, "global_step": 211749, "epoch": 2379, "lr": 1.150798725958736e-05}
{"train_loss": 0.028344128280878067, "global_step": 211750, "epoch": 2379, "lr": 1.1507617253045016e-05}
{"train_loss": 0.025463733822107315, "global_step": 211751, "epoch": 2379, "lr": 1.1507247251677494e-05}
{"train_loss": 0.030741671100258827, "global_step": 211752, "epoch": 2379, "lr": 1.1506877255484866e-05}
{"train_loss": 0.028556518256664276, "global_step": 211753, "epoch": 2379, "lr": 1.1506507264467154e-05}
{"train_loss": 0.03551387041807175, "global_step": 211754, "epoch": 2379, "lr": 1.1506137278624435e-05}
{"train_loss": 0.04595297574996948, "global_step": 211755, "epoch": 2379, "lr": 1.1505767297956738e-05}
{"train_loss": 0.04655871540307999, "global_step": 211756, "epoch": 2379, "lr": 1.1505397322464134e-05}
{"train_loss": 0.09506171196699142, "global_step": 211757, "epoch": 2379, "lr": 1.1505027352146641e-05}
{"train_loss": 0.05001598224043846, "global_step": 211758, "epoch": 2379, "lr": 1.1504657387004347e-05}
{"train_loss": 0.09143145382404327, "global_step": 211759, "epoch": 2379, "lr": 1.150428742703727e-05}
{"train_loss": 0.03881547227501869, "global_step": 211760, "epoch": 2379, "lr": 1.1503917472245484e-05}
{"train_loss": 0.06503375619649887, "global_step": 211761, "epoch": 2379, "lr": 1.150354752262901e-05}
{"train_loss": 0.06641101092100143, "global_step": 211762, "epoch": 2379, "lr": 1.1503177578187935e-05}
{"train_loss": 0.05434570088982582, "global_step": 211763, "epoch": 2379, "lr": 1.1502807638922285e-05}
{"train_loss": 0.09007471799850464, "global_step": 211764, "epoch": 2379, "lr": 1.1502437704832097e-05}
{"train_loss": 0.06367177516222, "global_step": 211765, "epoch": 2379, "lr": 1.1502067775917453e-05}
{"train_loss": 0.051830220967531204, "global_step": 211766, "epoch": 2379, "lr": 1.1501697852178373e-05}
{"train_loss": 0.07618463784456253, "global_step": 211767, "epoch": 2379, "lr": 1.1501327933614936e-05}
{"train_loss": 0.07645400613546371, "global_step": 211768, "epoch": 2379, "lr": 1.150095802022716e-05}
{"train_loss": 0.03937298804521561, "global_step": 211769, "epoch": 2379, "lr": 1.1500588112015125e-05}
{"train_loss": 0.11792127788066864, "global_step": 211770, "epoch": 2379, "lr": 1.1500218208978857e-05}
{"train_loss": 0.040284641087055206, "global_step": 211771, "epoch": 2379, "lr": 1.1499848311118428e-05}
{"train_loss": 0.06519751995801926, "global_step": 211772, "epoch": 2379, "lr": 1.1499478418433856e-05}
{"train_loss": 0.049876801669597626, "global_step": 211773, "epoch": 2379, "lr": 1.1499108530925228e-05}
{"train_loss": 0.04614633321762085, "global_step": 211774, "epoch": 2379, "lr": 1.1498738648592562e-05}
{"train_loss": 0.03014783002436161, "global_step": 211775, "epoch": 2379, "lr": 1.1498368771435935e-05}
{"train_loss": 0.05021641403436661, "global_step": 211776, "epoch": 2379, "lr": 1.1497998899455364e-05}
{"train_loss": 0.038266345858573914, "global_step": 211777, "epoch": 2379, "lr": 1.1497629032650925e-05}
{"train_loss": 0.07601585984230042, "global_step": 211778, "epoch": 2379, "lr": 1.1497259171022673e-05}
{"train_loss": 0.04914543405175209, "global_step": 211779, "epoch": 2379, "lr": 1.149688931457063e-05}
{"train_loss": 0.05001474916934967, "global_step": 211780, "epoch": 2379, "lr": 1.1496519463294875e-05}
{"train_loss": 0.06158705800771713, "global_step": 211781, "epoch": 2379, "lr": 1.1496149617195428e-05}
{"train_loss": 0.05553784593939781, "global_step": 211782, "epoch": 2379, "lr": 1.1495779776272369e-05}
{"train_loss": 0.10767820477485657, "global_step": 211783, "epoch": 2379, "lr": 1.149540994052572e-05}
{"train_loss": 0.04682448133826256, "global_step": 211784, "epoch": 2379, "lr": 1.1495040109955558e-05}
{"train_loss": 0.038858648389577866, "global_step": 211785, "epoch": 2379, "lr": 1.1494670284561904e-05}
{"train_loss": 0.0687589943408966, "global_step": 211786, "epoch": 2379, "lr": 1.1494300464344837e-05}
{"train_loss": 0.060121770948171616, "global_step": 211787, "epoch": 2379, "lr": 1.1493930649304374e-05}
{"train_loss": 0.06397518515586853, "global_step": 211788, "epoch": 2379, "lr": 1.1493560839440603e-05}
{"train_loss": 0.06248330697417259, "global_step": 211789, "epoch": 2379, "lr": 1.1493191034753537e-05}
{"train_loss": 0.08960561454296112, "global_step": 211790, "epoch": 2379, "lr": 1.1492821235243256e-05}
{"train_loss": 0.03145349398255348, "global_step": 211791, "epoch": 2379, "lr": 1.149245144090978e-05}
{"train_loss": 0.07007380574941635, "global_step": 211792, "epoch": 2379, "lr": 1.1492081651753189e-05}
{"train_loss": 0.060344334691762924, "global_step": 211793, "epoch": 2379, "lr": 1.1491711867773508e-05}
{"train_loss": 0.07020585983991623, "global_step": 211794, "epoch": 2379, "lr": 1.1491342088970807e-05}
{"train_loss": 0.08772887289524078, "global_step": 211795, "epoch": 2379, "lr": 1.149097231534511e-05}
{"train_loss": 0.05636582896113396, "global_step": 211796, "epoch": 2379, "lr": 1.14906025468965e-05}
{"train_loss": 0.020299090072512627, "global_step": 211797, "epoch": 2379, "lr": 1.1490232783624994e-05}
{"train_loss": 0.053090114146471024, "global_step": 211798, "epoch": 2379, "lr": 1.1489863025530673e-05}
{"train_loss": 0.052090123295784, "global_step": 211799, "epoch": 2379, "lr": 1.1489493272613556e-05}
{"train_loss": 0.04305939003825188, "global_step": 211800, "epoch": 2379, "lr": 1.1489123524873718e-05}
{"train_loss": 0.056877583265304565, "global_step": 211801, "epoch": 2379, "lr": 1.1488753782311191e-05}
{"train_loss": 0.022430097684264183, "global_step": 211802, "epoch": 2379, "lr": 1.1488384044926038e-05}
{"train_loss": 0.07528570294380188, "global_step": 211803, "epoch": 2379, "lr": 1.1488014312718304e-05}
{"train_loss": 0.05169534683227539, "global_step": 211804, "epoch": 2379, "lr": 1.1487644585688023e-05}
{"train_loss": 0.045805662870407104, "global_step": 211805, "epoch": 2379, "lr": 1.1487274863835279e-05}
{"train_loss": 0.07559118419885635, "global_step": 211806, "epoch": 2379, "lr": 1.1486905147160083e-05}
{"train_loss": 0.10961228609085083, "global_step": 211807, "epoch": 2379, "lr": 1.1486535435662516e-05}
{"train_loss": 0.0422590896487236, "global_step": 211808, "epoch": 2379, "lr": 1.1486165729342601e-05}
{"train_loss": 0.04904027655720711, "global_step": 211809, "epoch": 2379, "lr": 1.1485796028200414e-05}
{"train_loss": 0.045205097645521164, "global_step": 211810, "epoch": 2379, "lr": 1.1485426332235983e-05}
{"train_loss": 0.05866455286741257, "global_step": 211811, "epoch": 2379, "lr": 1.1485056641449376e-05}
{"train_loss": 0.06311827152967453, "global_step": 211812, "epoch": 2379, "lr": 1.1484686955840623e-05}
{"train_loss": 0.030762260779738426, "global_step": 211813, "epoch": 2379, "lr": 1.1484317275409784e-05}
{"train_loss": 0.027503397315740585, "global_step": 211814, "epoch": 2379, "lr": 1.1483947600156925e-05}
{"train_loss": 0.05264296755194664, "global_step": 211815, "epoch": 2379, "lr": 1.1483577930082063e-05}
{"train_loss": 0.0716172531247139, "global_step": 211816, "epoch": 2379, "lr": 1.1483208265185274e-05}
{"train_loss": 0.04682265594601631, "global_step": 211817, "epoch": 2379, "lr": 1.148283860546659e-05}
{"train_loss": 0.05023610219359398, "global_step": 211818, "epoch": 2379, "lr": 1.1482468950926085e-05}
{"train_loss": 0.0562967299326752, "global_step": 211819, "epoch": 2379, "lr": 1.1482099301563775e-05, "val_loss": 8.811409950256348}
{"train_loss": 0.06506457179784775, "global_step": 211820, "epoch": 2380, "lr": 1.1481729657379741e-05}
{"train_loss": 0.044686488807201385, "global_step": 211821, "epoch": 2380, "lr": 1.1481360018374005e-05}
{"train_loss": 0.08025937527418137, "global_step": 211822, "epoch": 2380, "lr": 1.1480990384546648e-05}
{"train_loss": 0.1124752089381218, "global_step": 211823, "epoch": 2380, "lr": 1.1480620755897687e-05}
{"train_loss": 0.019163623452186584, "global_step": 211824, "epoch": 2380, "lr": 1.14802511324272e-05}
{"train_loss": 0.05450362712144852, "global_step": 211825, "epoch": 2380, "lr": 1.147988151413521e-05}
{"train_loss": 0.05883083492517471, "global_step": 211826, "epoch": 2380, "lr": 1.1479511901021795e-05}
{"train_loss": 0.04470545053482056, "global_step": 211827, "epoch": 2380, "lr": 1.1479142293086976e-05}
{"train_loss": 0.05570126324892044, "global_step": 211828, "epoch": 2380, "lr": 1.1478772690330825e-05}
{"train_loss": 0.059428367763757706, "global_step": 211829, "epoch": 2380, "lr": 1.1478403092753376e-05}
{"train_loss": 0.10216475278139114, "global_step": 211830, "epoch": 2380, "lr": 1.1478033500354701e-05}
{"train_loss": 0.07130587846040726, "global_step": 211831, "epoch": 2380, "lr": 1.1477663913134818e-05}
{"train_loss": 0.059701524674892426, "global_step": 211832, "epoch": 2380, "lr": 1.1477294331093791e-05}
{"train_loss": 0.02823460102081299, "global_step": 211833, "epoch": 2380, "lr": 1.1476924754231693e-05}
{"train_loss": 0.032772354781627655, "global_step": 211834, "epoch": 2380, "lr": 1.1476555182548537e-05}
{"train_loss": 0.04690713435411453, "global_step": 211835, "epoch": 2380, "lr": 1.1476185616044399e-05}
{"train_loss": 0.047911062836647034, "global_step": 211836, "epoch": 2380, "lr": 1.1475816054719318e-05}
{"train_loss": 0.07380133122205734, "global_step": 211837, "epoch": 2380, "lr": 1.1475446498573333e-05}
{"train_loss": 0.07199158519506454, "global_step": 211838, "epoch": 2380, "lr": 1.1475076947606523e-05}
{"train_loss": 0.03937980532646179, "global_step": 211839, "epoch": 2380, "lr": 1.1474707401818896e-05}
{"train_loss": 0.03841057047247887, "global_step": 211840, "epoch": 2380, "lr": 1.1474337861210543e-05}
{"train_loss": 0.055876702070236206, "global_step": 211841, "epoch": 2380, "lr": 1.1473968325781497e-05}
{"train_loss": 0.03745081275701523, "global_step": 211842, "epoch": 2380, "lr": 1.1473598795531793e-05}
{"train_loss": 0.04825318604707718, "global_step": 211843, "epoch": 2380, "lr": 1.1473229270461504e-05}
{"train_loss": 0.11526205390691757, "global_step": 211844, "epoch": 2380, "lr": 1.1472859750570658e-05}
{"train_loss": 0.0449456088244915, "global_step": 211845, "epoch": 2380, "lr": 1.1472490235859335e-05}
{"train_loss": 0.035261206328868866, "global_step": 211846, "epoch": 2380, "lr": 1.1472120726327545e-05}
{"train_loss": 0.06696111708879471, "global_step": 211847, "epoch": 2380, "lr": 1.147175122197538e-05}
{"train_loss": 0.03624367341399193, "global_step": 211848, "epoch": 2380, "lr": 1.1471381722802848e-05}
{"train_loss": 0.043751854449510574, "global_step": 211849, "epoch": 2380, "lr": 1.1471012228810018e-05}
{"train_loss": 0.07315054535865784, "global_step": 211850, "epoch": 2380, "lr": 1.1470642739996961e-05}
{"train_loss": 0.08365906029939651, "global_step": 211851, "epoch": 2380, "lr": 1.147027325636369e-05}
{"train_loss": 0.11865819245576859, "global_step": 211852, "epoch": 2380, "lr": 1.1469903777910285e-05}
{"train_loss": 0.05677832290530205, "global_step": 211853, "epoch": 2380, "lr": 1.1469534304636764e-05}
{"train_loss": 0.027498474344611168, "global_step": 211854, "epoch": 2380, "lr": 1.1469164836543217e-05}
{"train_loss": 0.043758977204561234, "global_step": 211855, "epoch": 2380, "lr": 1.1468795373629649e-05}
{"train_loss": 0.08469444513320923, "global_step": 211856, "epoch": 2380, "lr": 1.1468425915896154e-05}
{"train_loss": 0.030259080231189728, "global_step": 211857, "epoch": 2380, "lr": 1.1468056463342735e-05}
{"train_loss": 0.045008495450019836, "global_step": 211858, "epoch": 2380, "lr": 1.1467687015969486e-05}
{"train_loss": 0.08073191344738007, "global_step": 211859, "epoch": 2380, "lr": 1.1467317573776426e-05}
{"train_loss": 0.08585938811302185, "global_step": 211860, "epoch": 2380, "lr": 1.1466948136763622e-05}
{"train_loss": 0.06580645591020584, "global_step": 211861, "epoch": 2380, "lr": 1.1466578704931108e-05}
{"train_loss": 0.042663514614105225, "global_step": 211862, "epoch": 2380, "lr": 1.1466209278278955e-05}
{"train_loss": 0.04410345107316971, "global_step": 211863, "epoch": 2380, "lr": 1.1465839856807187e-05}
{"train_loss": 0.11780146509408951, "global_step": 211864, "epoch": 2380, "lr": 1.1465470440515885e-05}
{"train_loss": 0.06184384971857071, "global_step": 211865, "epoch": 2380, "lr": 1.1465101029405061e-05}
{"train_loss": 0.05105919763445854, "global_step": 211866, "epoch": 2380, "lr": 1.14647316234748e-05}
{"train_loss": 0.06532811373472214, "global_step": 211867, "epoch": 2380, "lr": 1.1464362222725127e-05}
{"train_loss": 0.07925455272197723, "global_step": 211868, "epoch": 2380, "lr": 1.1463992827156095e-05}
{"train_loss": 0.012302999384701252, "global_step": 211869, "epoch": 2380, "lr": 1.1463623436767778e-05}
{"train_loss": 0.05104423314332962, "global_step": 211870, "epoch": 2380, "lr": 1.146325405156019e-05}
{"train_loss": 0.07049960643053055, "global_step": 211871, "epoch": 2380, "lr": 1.1462884671533414e-05}
{"train_loss": 0.06432228535413742, "global_step": 211872, "epoch": 2380, "lr": 1.1462515296687464e-05}
{"train_loss": 0.030850548297166824, "global_step": 211873, "epoch": 2380, "lr": 1.1462145927022427e-05}
{"train_loss": 0.041917864233255386, "global_step": 211874, "epoch": 2380, "lr": 1.1461776562538335e-05}
{"train_loss": 0.0870780497789383, "global_step": 211875, "epoch": 2380, "lr": 1.1461407203235224e-05}
{"train_loss": 0.03129546716809273, "global_step": 211876, "epoch": 2380, "lr": 1.146103784911317e-05}
{"train_loss": 0.06910621374845505, "global_step": 211877, "epoch": 2380, "lr": 1.1460668500172194e-05}
{"train_loss": 0.013112244196236134, "global_step": 211878, "epoch": 2380, "lr": 1.1460299156412379e-05}
{"train_loss": 0.04295056685805321, "global_step": 211879, "epoch": 2380, "lr": 1.145992981783376e-05}
{"train_loss": 0.09365750104188919, "global_step": 211880, "epoch": 2380, "lr": 1.145956048443636e-05}
{"train_loss": 0.10811205208301544, "global_step": 211881, "epoch": 2380, "lr": 1.1459191156220273e-05}
{"train_loss": 0.08665606379508972, "global_step": 211882, "epoch": 2380, "lr": 1.1458821833185512e-05}
{"train_loss": 0.0794818103313446, "global_step": 211883, "epoch": 2380, "lr": 1.1458452515332157e-05}
{"train_loss": 0.027624430134892464, "global_step": 211884, "epoch": 2380, "lr": 1.145808320266023e-05}
{"train_loss": 0.03580009564757347, "global_step": 211885, "epoch": 2380, "lr": 1.1457713895169792e-05}
{"train_loss": 0.04298974201083183, "global_step": 211886, "epoch": 2380, "lr": 1.1457344592860913e-05}
{"train_loss": 0.023323366418480873, "global_step": 211887, "epoch": 2380, "lr": 1.1456975295733607e-05}
{"train_loss": 0.03965390473604202, "global_step": 211888, "epoch": 2380, "lr": 1.1456606003787956e-05}
{"train_loss": 0.03596208617091179, "global_step": 211889, "epoch": 2380, "lr": 1.1456236717023977e-05}
{"train_loss": 0.06380783021450043, "global_step": 211890, "epoch": 2380, "lr": 1.1455867435441754e-05}
{"train_loss": 0.04165095090866089, "global_step": 211891, "epoch": 2380, "lr": 1.1455498159041305e-05}
{"train_loss": 0.03323354944586754, "global_step": 211892, "epoch": 2380, "lr": 1.1455128887822708e-05}
{"train_loss": 0.07692346721887589, "global_step": 211893, "epoch": 2380, "lr": 1.1454759621785987e-05}
{"train_loss": 0.05426718294620514, "global_step": 211894, "epoch": 2380, "lr": 1.1454390360931216e-05}
{"train_loss": 0.040921520441770554, "global_step": 211895, "epoch": 2380, "lr": 1.1454021105258423e-05}
{"train_loss": 0.04978236183524132, "global_step": 211896, "epoch": 2380, "lr": 1.1453651854767677e-05}
{"train_loss": 0.05970215052366257, "global_step": 211897, "epoch": 2380, "lr": 1.1453282609459005e-05}
{"train_loss": 0.018408963456749916, "global_step": 211898, "epoch": 2380, "lr": 1.145291336933248e-05}
{"train_loss": 0.046837713569402695, "global_step": 211899, "epoch": 2380, "lr": 1.1452544134388133e-05}
{"train_loss": 0.03317505493760109, "global_step": 211900, "epoch": 2380, "lr": 1.1452174904626028e-05}
{"train_loss": 0.06280071288347244, "global_step": 211901, "epoch": 2380, "lr": 1.1451805680046202e-05}
{"train_loss": 0.029104216024279594, "global_step": 211902, "epoch": 2380, "lr": 1.1451436460648718e-05}
{"train_loss": 0.03375361114740372, "global_step": 211903, "epoch": 2380, "lr": 1.1451067246433612e-05}
{"train_loss": 0.06924287229776382, "global_step": 211904, "epoch": 2380, "lr": 1.1450698037400936e-05}
{"train_loss": 0.043130725622177124, "global_step": 211905, "epoch": 2380, "lr": 1.1450328833550755e-05}
{"train_loss": 0.031025145202875137, "global_step": 211906, "epoch": 2380, "lr": 1.1449959634883095e-05}
{"train_loss": 0.021861759945750237, "global_step": 211907, "epoch": 2380, "lr": 1.1449590441398034e-05}
{"train_loss": 0.05541825777945224, "global_step": 211908, "epoch": 2380, "lr": 1.1449221253095594e-05, "val_loss": 8.83968448638916, "train_action_mse_error": 9.825708389282227}
{"train_loss": 0.058613553643226624, "global_step": 211909, "epoch": 2381, "lr": 1.1448852069975847e-05}
{"train_loss": 0.08501880615949631, "global_step": 211910, "epoch": 2381, "lr": 1.144848289203882e-05}
{"train_loss": 0.025125760585069656, "global_step": 211911, "epoch": 2381, "lr": 1.1448113719284587e-05}
{"train_loss": 0.032460130751132965, "global_step": 211912, "epoch": 2381, "lr": 1.144774455171318e-05}
{"train_loss": 0.038214121013879776, "global_step": 211913, "epoch": 2381, "lr": 1.144737538932465e-05}
{"train_loss": 0.05511997640132904, "global_step": 211914, "epoch": 2381, "lr": 1.1447006232119056e-05}
{"train_loss": 0.041508834809064865, "global_step": 211915, "epoch": 2381, "lr": 1.1446637080096434e-05}
{"train_loss": 0.02032921090722084, "global_step": 211916, "epoch": 2381, "lr": 1.144626793325685e-05}
{"train_loss": 0.08534053713083267, "global_step": 211917, "epoch": 2381, "lr": 1.1445898791600334e-05}
{"train_loss": 0.08576300740242004, "global_step": 211918, "epoch": 2381, "lr": 1.1445529655126962e-05}
{"train_loss": 0.09522396326065063, "global_step": 211919, "epoch": 2381, "lr": 1.1445160523836773e-05}
{"train_loss": 0.05564971640706062, "global_step": 211920, "epoch": 2381, "lr": 1.1444791397729792e-05}
{"train_loss": 0.03612762689590454, "global_step": 211921, "epoch": 2381, "lr": 1.1444422276806105e-05}
{"train_loss": 0.07068796455860138, "global_step": 211922, "epoch": 2381, "lr": 1.1444053161065732e-05}
{"train_loss": 0.08018268644809723, "global_step": 211923, "epoch": 2381, "lr": 1.1443684050508735e-05}
{"train_loss": 0.02561519853770733, "global_step": 211924, "epoch": 2381, "lr": 1.1443314945135174e-05}
{"train_loss": 0.06082911789417267, "global_step": 211925, "epoch": 2381, "lr": 1.144294584494508e-05}
{"train_loss": 0.08764538168907166, "global_step": 211926, "epoch": 2381, "lr": 1.1442576749938527e-05}
{"train_loss": 0.06211518496274948, "global_step": 211927, "epoch": 2381, "lr": 1.1442207660115538e-05}
{"train_loss": 0.049154140055179596, "global_step": 211928, "epoch": 2381, "lr": 1.1441838575476182e-05}
{"train_loss": 0.040719423443078995, "global_step": 211929, "epoch": 2381, "lr": 1.144146949602049e-05}
{"train_loss": 0.031199980527162552, "global_step": 211930, "epoch": 2381, "lr": 1.1441100421748536e-05}
{"train_loss": 0.06101939082145691, "global_step": 211931, "epoch": 2381, "lr": 1.1440731352660345e-05}
{"train_loss": 0.07394873350858688, "global_step": 211932, "epoch": 2381, "lr": 1.144036228875599e-05}
{"train_loss": 0.04616018012166023, "global_step": 211933, "epoch": 2381, "lr": 1.143999323003549e-05}
{"train_loss": 0.015560610219836235, "global_step": 211934, "epoch": 2381, "lr": 1.143962417649893e-05}
{"train_loss": 0.053387872874736786, "global_step": 211935, "epoch": 2381, "lr": 1.1439255128146331e-05}
{"train_loss": 0.04785336181521416, "global_step": 211936, "epoch": 2381, "lr": 1.1438886084977767e-05}
{"train_loss": 0.04154045879840851, "global_step": 211937, "epoch": 2381, "lr": 1.1438517046993258e-05}
{"train_loss": 0.04815375432372093, "global_step": 211938, "epoch": 2381, "lr": 1.1438148014192884e-05}
{"train_loss": 0.06586413830518723, "global_step": 211939, "epoch": 2381, "lr": 1.1437778986576674e-05}
{"train_loss": 0.05340390279889107, "global_step": 211940, "epoch": 2381, "lr": 1.143740996414468e-05}
{"train_loss": 0.0409478135406971, "global_step": 211941, "epoch": 2381, "lr": 1.1437040946896965e-05}
{"train_loss": 0.036673255264759064, "global_step": 211942, "epoch": 2381, "lr": 1.1436671934833564e-05}
{"train_loss": 0.05510426685214043, "global_step": 211943, "epoch": 2381, "lr": 1.1436302927954545e-05}
{"train_loss": 0.08303865790367126, "global_step": 211944, "epoch": 2381, "lr": 1.1435933926259929e-05}
{"train_loss": 0.03653945401310921, "global_step": 211945, "epoch": 2381, "lr": 1.1435564929749793e-05}
{"train_loss": 0.04143071919679642, "global_step": 211946, "epoch": 2381, "lr": 1.1435195938424165e-05}
{"train_loss": 0.04197615385055542, "global_step": 211947, "epoch": 2381, "lr": 1.1434826952283117e-05}
{"train_loss": 0.027899691835045815, "global_step": 211948, "epoch": 2381, "lr": 1.1434457971326673e-05}
{"train_loss": 0.09449964016675949, "global_step": 211949, "epoch": 2381, "lr": 1.1434088995554915e-05}
{"train_loss": 0.025421520695090294, "global_step": 211950, "epoch": 2381, "lr": 1.1433720024967853e-05}
{"train_loss": 0.03579334169626236, "global_step": 211951, "epoch": 2381, "lr": 1.1433351059565572e-05}
{"train_loss": 0.05565958842635155, "global_step": 211952, "epoch": 2381, "lr": 1.1432982099348111e-05}
{"train_loss": 0.04061095044016838, "global_step": 211953, "epoch": 2381, "lr": 1.1432613144315497e-05}
{"train_loss": 0.052602436393499374, "global_step": 211954, "epoch": 2381, "lr": 1.1432244194467817e-05}
{"train_loss": 0.05695085972547531, "global_step": 211955, "epoch": 2381, "lr": 1.1431875249805085e-05}
{"train_loss": 0.1005425900220871, "global_step": 211956, "epoch": 2381, "lr": 1.1431506310327378e-05}
{"train_loss": 0.0523458868265152, "global_step": 211957, "epoch": 2381, "lr": 1.143113737603474e-05}
{"train_loss": 0.061574652791023254, "global_step": 211958, "epoch": 2381, "lr": 1.1430768446927199e-05}
{"train_loss": 0.029458515346050262, "global_step": 211959, "epoch": 2381, "lr": 1.143039952300482e-05}
{"train_loss": 0.051836952567100525, "global_step": 211960, "epoch": 2381, "lr": 1.1430030604267671e-05}
{"train_loss": 0.07842352986335754, "global_step": 211961, "epoch": 2381, "lr": 1.142966169071577e-05}
{"train_loss": 0.04414411261677742, "global_step": 211962, "epoch": 2381, "lr": 1.1429292782349194e-05}
{"train_loss": 0.028069116175174713, "global_step": 211963, "epoch": 2381, "lr": 1.1428923879167964e-05}
{"train_loss": 0.02343175932765007, "global_step": 211964, "epoch": 2381, "lr": 1.1428554981172156e-05}
{"train_loss": 0.07555356621742249, "global_step": 211965, "epoch": 2381, "lr": 1.1428186088361803e-05}
{"train_loss": 0.03169532120227814, "global_step": 211966, "epoch": 2381, "lr": 1.1427817200736968e-05}
{"train_loss": 0.07024949789047241, "global_step": 211967, "epoch": 2381, "lr": 1.1427448318297678e-05}
{"train_loss": 0.03645310550928116, "global_step": 211968, "epoch": 2381, "lr": 1.142707944104402e-05}
{"train_loss": 0.07612021267414093, "global_step": 211969, "epoch": 2381, "lr": 1.1426710568975996e-05}
{"train_loss": 0.07138822972774506, "global_step": 211970, "epoch": 2381, "lr": 1.14263417020937e-05}
{"train_loss": 0.044198594987392426, "global_step": 211971, "epoch": 2381, "lr": 1.1425972840397148e-05}
{"train_loss": 0.06039628013968468, "global_step": 211972, "epoch": 2381, "lr": 1.1425603983886418e-05}
{"train_loss": 0.04439849033951759, "global_step": 211973, "epoch": 2381, "lr": 1.1425235132561535e-05}
{"train_loss": 0.09789909422397614, "global_step": 211974, "epoch": 2381, "lr": 1.142486628642257e-05}
{"train_loss": 0.10671523213386536, "global_step": 211975, "epoch": 2381, "lr": 1.1424497445469546e-05}
{"train_loss": 0.018360568210482597, "global_step": 211976, "epoch": 2381, "lr": 1.1424128609702528e-05}
{"train_loss": 0.028965268284082413, "global_step": 211977, "epoch": 2381, "lr": 1.142375977912158e-05}
{"train_loss": 0.06550151854753494, "global_step": 211978, "epoch": 2381, "lr": 1.1423390953726726e-05}
{"train_loss": 0.024490775540471077, "global_step": 211979, "epoch": 2381, "lr": 1.1423022133518035e-05}
{"train_loss": 0.03161536529660225, "global_step": 211980, "epoch": 2381, "lr": 1.1422653318495542e-05}
{"train_loss": 0.0749041810631752, "global_step": 211981, "epoch": 2381, "lr": 1.142228450865931e-05}
{"train_loss": 0.039965223520994186, "global_step": 211982, "epoch": 2381, "lr": 1.1421915704009373e-05}
{"train_loss": 0.07067380845546722, "global_step": 211983, "epoch": 2381, "lr": 1.14215469045458e-05}
{"train_loss": 0.03788482025265694, "global_step": 211984, "epoch": 2381, "lr": 1.1421178110268616e-05}
{"train_loss": 0.03987889736890793, "global_step": 211985, "epoch": 2381, "lr": 1.14208093211779e-05}
{"train_loss": 0.04190107807517052, "global_step": 211986, "epoch": 2381, "lr": 1.1420440537273674e-05}
{"train_loss": 0.022416502237319946, "global_step": 211987, "epoch": 2381, "lr": 1.142007175855601e-05}
{"train_loss": 0.14650671184062958, "global_step": 211988, "epoch": 2381, "lr": 1.1419702985024932e-05}
{"train_loss": 0.10173261910676956, "global_step": 211989, "epoch": 2381, "lr": 1.1419334216680522e-05}
{"train_loss": 0.08162488043308258, "global_step": 211990, "epoch": 2381, "lr": 1.1418965453522812e-05}
{"train_loss": 0.09615933150053024, "global_step": 211991, "epoch": 2381, "lr": 1.1418596695551837e-05}
{"train_loss": 0.04822666198015213, "global_step": 211992, "epoch": 2381, "lr": 1.1418227942767673e-05}
{"train_loss": 0.04598149284720421, "global_step": 211993, "epoch": 2381, "lr": 1.141785919517035e-05}
{"train_loss": 0.09586922824382782, "global_step": 211994, "epoch": 2381, "lr": 1.1417490452759932e-05}
{"train_loss": 0.055020250380039215, "global_step": 211995, "epoch": 2381, "lr": 1.1417121715536454e-05}
{"train_loss": 0.08344537764787674, "global_step": 211996, "epoch": 2381, "lr": 1.141675298349999e-05}
{"train_loss": 0.055318327097410566, "global_step": 211997, "epoch": 2381, "lr": 1.1416384256650554e-05, "val_loss": 8.872732162475586}
{"train_loss": 0.06402274966239929, "global_step": 211998, "epoch": 2382, "lr": 1.1416015534988229e-05}
{"train_loss": 0.06944551318883896, "global_step": 211999, "epoch": 2382, "lr": 1.1415646818513037e-05}
{"train_loss": 0.0870872512459755, "global_step": 212000, "epoch": 2382, "lr": 1.1415278107225058e-05}
{"train_loss": 0.0349908284842968, "global_step": 212001, "epoch": 2382, "lr": 1.1414909401124307e-05}
{"train_loss": 0.050249431282281876, "global_step": 212002, "epoch": 2382, "lr": 1.1414540700210863e-05}
{"train_loss": 0.041005391627550125, "global_step": 212003, "epoch": 2382, "lr": 1.1414172004484751e-05}
{"train_loss": 0.043790534138679504, "global_step": 212004, "epoch": 2382, "lr": 1.1413803313946054e-05}
{"train_loss": 0.07931821048259735, "global_step": 212005, "epoch": 2382, "lr": 1.1413434628594783e-05}
{"train_loss": 0.06913863122463226, "global_step": 212006, "epoch": 2382, "lr": 1.1413065948431018e-05}
{"train_loss": 0.029185185208916664, "global_step": 212007, "epoch": 2382, "lr": 1.141269727345478e-05}
{"train_loss": 0.03814587742090225, "global_step": 212008, "epoch": 2382, "lr": 1.141232860366615e-05}
{"train_loss": 0.04979543387889862, "global_step": 212009, "epoch": 2382, "lr": 1.1411959939065154e-05}
{"train_loss": 0.046848997473716736, "global_step": 212010, "epoch": 2382, "lr": 1.1411591279651856e-05}
{"train_loss": 0.058631978929042816, "global_step": 212011, "epoch": 2382, "lr": 1.1411222625426287e-05}
{"train_loss": 0.08256031572818756, "global_step": 212012, "epoch": 2382, "lr": 1.1410853976388508e-05}
{"train_loss": 0.07138596475124359, "global_step": 212013, "epoch": 2382, "lr": 1.141048533253859e-05}
{"train_loss": 0.07173997908830643, "global_step": 212014, "epoch": 2382, "lr": 1.1410116693876537e-05}
{"train_loss": 0.0632888451218605, "global_step": 212015, "epoch": 2382, "lr": 1.1409748060402448e-05}
{"train_loss": 0.051463790237903595, "global_step": 212016, "epoch": 2382, "lr": 1.140937943211633e-05}
{"train_loss": 0.04276922345161438, "global_step": 212017, "epoch": 2382, "lr": 1.140901080901826e-05}
{"train_loss": 0.043626196682453156, "global_step": 212018, "epoch": 2382, "lr": 1.1408642191108271e-05}
{"train_loss": 0.054913077503442764, "global_step": 212019, "epoch": 2382, "lr": 1.1408273578386431e-05}
{"train_loss": 0.03545958548784256, "global_step": 212020, "epoch": 2382, "lr": 1.1407904970852761e-05}
{"train_loss": 0.07291105389595032, "global_step": 212021, "epoch": 2382, "lr": 1.140753636850735e-05}
{"train_loss": 0.05195087939500809, "global_step": 212022, "epoch": 2382, "lr": 1.1407167771350202e-05}
{"train_loss": 0.036246296018362045, "global_step": 212023, "epoch": 2382, "lr": 1.1406799179381412e-05}
{"train_loss": 0.03443945199251175, "global_step": 212024, "epoch": 2382, "lr": 1.1406430592600986e-05}
{"train_loss": 0.027881884947419167, "global_step": 212025, "epoch": 2382, "lr": 1.1406062011009016e-05}
{"train_loss": 0.04017994925379753, "global_step": 212026, "epoch": 2382, "lr": 1.1405693434605513e-05}
{"train_loss": 0.07020585983991623, "global_step": 212027, "epoch": 2382, "lr": 1.140532486339056e-05}
{"train_loss": 0.06180576980113983, "global_step": 212028, "epoch": 2382, "lr": 1.1404956297364189e-05}
{"train_loss": 0.02137676812708378, "global_step": 212029, "epoch": 2382, "lr": 1.1404587736526434e-05}
{"train_loss": 0.04370657727122307, "global_step": 212030, "epoch": 2382, "lr": 1.140421918087738e-05}
{"train_loss": 0.05085176229476929, "global_step": 212031, "epoch": 2382, "lr": 1.1403850630417045e-05}
{"train_loss": 0.051742810755968094, "global_step": 212032, "epoch": 2382, "lr": 1.1403482085145501e-05}
{"train_loss": 0.07095237821340561, "global_step": 212033, "epoch": 2382, "lr": 1.1403113545062783e-05}
{"train_loss": 0.05439697578549385, "global_step": 212034, "epoch": 2382, "lr": 1.1402745010168952e-05}
{"train_loss": 0.05357060208916664, "global_step": 212035, "epoch": 2382, "lr": 1.1402376480464039e-05}
{"train_loss": 0.05604167282581329, "global_step": 212036, "epoch": 2382, "lr": 1.1402007955948124e-05}
{"train_loss": 0.0431952141225338, "global_step": 212037, "epoch": 2382, "lr": 1.140163943662122e-05}
{"train_loss": 0.062515489757061, "global_step": 212038, "epoch": 2382, "lr": 1.140127092248341e-05}
{"train_loss": 0.05461231619119644, "global_step": 212039, "epoch": 2382, "lr": 1.1400902413534714e-05}
{"train_loss": 0.053304024040699005, "global_step": 212040, "epoch": 2382, "lr": 1.1400533909775213e-05}
{"train_loss": 0.13600178062915802, "global_step": 212041, "epoch": 2382, "lr": 1.1400165411204927e-05}
{"train_loss": 0.029370475560426712, "global_step": 212042, "epoch": 2382, "lr": 1.1399796917823929e-05}
{"train_loss": 0.06475221365690231, "global_step": 212043, "epoch": 2382, "lr": 1.1399428429632243e-05}
{"train_loss": 0.07905799895524979, "global_step": 212044, "epoch": 2382, "lr": 1.1399059946629948e-05}
{"train_loss": 0.02307945489883423, "global_step": 212045, "epoch": 2382, "lr": 1.1398691468817068e-05}
{"train_loss": 0.027571508660912514, "global_step": 212046, "epoch": 2382, "lr": 1.1398322996193673e-05}
{"train_loss": 0.09507034718990326, "global_step": 212047, "epoch": 2382, "lr": 1.1397954528759791e-05}
{"train_loss": 0.03763880208134651, "global_step": 212048, "epoch": 2382, "lr": 1.1397586066515497e-05}
{"train_loss": 0.06970570236444473, "global_step": 212049, "epoch": 2382, "lr": 1.139721760946082e-05}
{"train_loss": 0.04191388934850693, "global_step": 212050, "epoch": 2382, "lr": 1.1396849157595807e-05}
{"train_loss": 0.070539690554142, "global_step": 212051, "epoch": 2382, "lr": 1.1396480710920537e-05}
{"train_loss": 0.07010531425476074, "global_step": 212052, "epoch": 2382, "lr": 1.1396112269435021e-05}
{"train_loss": 0.07077454030513763, "global_step": 212053, "epoch": 2382, "lr": 1.1395743833139345e-05}
{"train_loss": 0.10171565413475037, "global_step": 212054, "epoch": 2382, "lr": 1.1395375402033525e-05}
{"train_loss": 0.053008921444416046, "global_step": 212055, "epoch": 2382, "lr": 1.139500697611764e-05}
{"train_loss": 0.048976294696331024, "global_step": 212056, "epoch": 2382, "lr": 1.1394638555391712e-05}
{"train_loss": 0.016013875603675842, "global_step": 212057, "epoch": 2382, "lr": 1.1394270139855817e-05}
{"train_loss": 0.10457154363393784, "global_step": 212058, "epoch": 2382, "lr": 1.1393901729509981e-05}
{"train_loss": 0.0461418479681015, "global_step": 212059, "epoch": 2382, "lr": 1.1393533324354278e-05}
{"train_loss": 0.0426051989197731, "global_step": 212060, "epoch": 2382, "lr": 1.1393164924388727e-05}
{"train_loss": 0.09055028855800629, "global_step": 212061, "epoch": 2382, "lr": 1.139279652961341e-05}
{"train_loss": 0.04619751498103142, "global_step": 212062, "epoch": 2382, "lr": 1.1392428140028349e-05}
{"train_loss": 0.05326807126402855, "global_step": 212063, "epoch": 2382, "lr": 1.1392059755633617e-05}
{"train_loss": 0.09034301340579987, "global_step": 212064, "epoch": 2382, "lr": 1.1391691376429236e-05}
{"train_loss": 0.05421651899814606, "global_step": 212065, "epoch": 2382, "lr": 1.1391323002415289e-05}
{"train_loss": 0.05603686720132828, "global_step": 212066, "epoch": 2382, "lr": 1.1390954633591794e-05}
{"train_loss": 0.054461151361465454, "global_step": 212067, "epoch": 2382, "lr": 1.1390586269958825e-05}
{"train_loss": 0.027790464460849762, "global_step": 212068, "epoch": 2382, "lr": 1.1390217911516427e-05}
{"train_loss": 0.04314449429512024, "global_step": 212069, "epoch": 2382, "lr": 1.1389849558264626e-05}
{"train_loss": 0.08978915959596634, "global_step": 212070, "epoch": 2382, "lr": 1.13894812102035e-05}
{"train_loss": 0.08328688144683838, "global_step": 212071, "epoch": 2382, "lr": 1.138911286733308e-05}
{"train_loss": 0.046174462884664536, "global_step": 212072, "epoch": 2382, "lr": 1.1388744529653434e-05}
{"train_loss": 0.03985171765089035, "global_step": 212073, "epoch": 2382, "lr": 1.1388376197164584e-05}
{"train_loss": 0.0693715289235115, "global_step": 212074, "epoch": 2382, "lr": 1.1388007869866619e-05}
{"train_loss": 0.036189910024404526, "global_step": 212075, "epoch": 2382, "lr": 1.1387639547759543e-05}
{"train_loss": 0.03027571551501751, "global_step": 212076, "epoch": 2382, "lr": 1.1387271230843443e-05}
{"train_loss": 0.0588068887591362, "global_step": 212077, "epoch": 2382, "lr": 1.1386902919118348e-05}
{"train_loss": 0.0731261745095253, "global_step": 212078, "epoch": 2382, "lr": 1.138653461258432e-05}
{"train_loss": 0.03198167309165001, "global_step": 212079, "epoch": 2382, "lr": 1.1386166311241386e-05}
{"train_loss": 0.056473687291145325, "global_step": 212080, "epoch": 2382, "lr": 1.1385798015089628e-05}
{"train_loss": 0.04823646694421768, "global_step": 212081, "epoch": 2382, "lr": 1.1385429724129065e-05}
{"train_loss": 0.08646275103092194, "global_step": 212082, "epoch": 2382, "lr": 1.138506143835978e-05}
{"train_loss": 0.08983653783798218, "global_step": 212083, "epoch": 2382, "lr": 1.1384693157781778e-05}
{"train_loss": 0.09013428539037704, "global_step": 212084, "epoch": 2382, "lr": 1.1384324882395153e-05}
{"train_loss": 0.05225754156708717, "global_step": 212085, "epoch": 2382, "lr": 1.138395661219992e-05}
{"train_loss": 0.057202666097979867, "global_step": 212086, "epoch": 2382, "lr": 1.1383588347196145e-05, "val_loss": 8.857392311096191}
{"train_loss": 0.03998178616166115, "global_step": 212087, "epoch": 2383, "lr": 1.1383220087383889e-05}
{"train_loss": 0.0433642640709877, "global_step": 212088, "epoch": 2383, "lr": 1.1382851832763175e-05}
{"train_loss": 0.02635246142745018, "global_step": 212089, "epoch": 2383, "lr": 1.1382483583334081e-05}
{"train_loss": 0.05157756432890892, "global_step": 212090, "epoch": 2383, "lr": 1.1382115339096622e-05}
{"train_loss": 0.020357297733426094, "global_step": 212091, "epoch": 2383, "lr": 1.1381747100050888e-05}
{"train_loss": 0.03620991110801697, "global_step": 212092, "epoch": 2383, "lr": 1.138137886619689e-05}
{"train_loss": 0.048830997198820114, "global_step": 212093, "epoch": 2383, "lr": 1.1381010637534705e-05}
{"train_loss": 0.05267231911420822, "global_step": 212094, "epoch": 2383, "lr": 1.1380642414064362e-05}
{"train_loss": 0.1101892963051796, "global_step": 212095, "epoch": 2383, "lr": 1.1380274195785934e-05}
{"train_loss": 0.10326308012008667, "global_step": 212096, "epoch": 2383, "lr": 1.1379905982699446e-05}
{"train_loss": 0.03189438581466675, "global_step": 212097, "epoch": 2383, "lr": 1.1379537774804972e-05}
{"train_loss": 0.06046760827302933, "global_step": 212098, "epoch": 2383, "lr": 1.1379169572102532e-05}
{"train_loss": 0.053906746208667755, "global_step": 212099, "epoch": 2383, "lr": 1.1378801374592208e-05}
{"train_loss": 0.03630717098712921, "global_step": 212100, "epoch": 2383, "lr": 1.1378433182274022e-05}
{"train_loss": 0.06471353024244308, "global_step": 212101, "epoch": 2383, "lr": 1.1378064995148048e-05}
{"train_loss": 0.05871118605136871, "global_step": 212102, "epoch": 2383, "lr": 1.1377696813214305e-05}
{"train_loss": 0.07946757972240448, "global_step": 212103, "epoch": 2383, "lr": 1.1377328636472873e-05}
{"train_loss": 0.028358764946460724, "global_step": 212104, "epoch": 2383, "lr": 1.1376960464923781e-05}
{"train_loss": 0.039257459342479706, "global_step": 212105, "epoch": 2383, "lr": 1.137659229856709e-05}
{"train_loss": 0.09183470904827118, "global_step": 212106, "epoch": 2383, "lr": 1.1376224137402852e-05}
{"train_loss": 0.10540422797203064, "global_step": 212107, "epoch": 2383, "lr": 1.13758559814311e-05}
{"train_loss": 0.04494194686412811, "global_step": 212108, "epoch": 2383, "lr": 1.1375487830651898e-05}
{"train_loss": 0.08908442407846451, "global_step": 212109, "epoch": 2383, "lr": 1.1375119685065283e-05}
{"train_loss": 0.05832483246922493, "global_step": 212110, "epoch": 2383, "lr": 1.1374751544671326e-05}
{"train_loss": 0.06666844338178635, "global_step": 212111, "epoch": 2383, "lr": 1.1374383409470046e-05}
{"train_loss": 0.059765592217445374, "global_step": 212112, "epoch": 2383, "lr": 1.1374015279461525e-05}
{"train_loss": 0.07740627229213715, "global_step": 212113, "epoch": 2383, "lr": 1.1373647154645784e-05}
{"train_loss": 0.05660165473818779, "global_step": 212114, "epoch": 2383, "lr": 1.1373279035022899e-05}
{"train_loss": 0.045922525227069855, "global_step": 212115, "epoch": 2383, "lr": 1.1372910920592888e-05}
{"train_loss": 0.038160912692546844, "global_step": 212116, "epoch": 2383, "lr": 1.1372542811355835e-05}
{"train_loss": 0.036272499710321426, "global_step": 212117, "epoch": 2383, "lr": 1.137217470731176e-05}
{"train_loss": 0.0747172012925148, "global_step": 212118, "epoch": 2383, "lr": 1.1371806608460739e-05}
{"train_loss": 0.040816210210323334, "global_step": 212119, "epoch": 2383, "lr": 1.1371438514802796e-05}
{"train_loss": 0.07202503085136414, "global_step": 212120, "epoch": 2383, "lr": 1.1371070426338005e-05}
{"train_loss": 0.05012111738324165, "global_step": 212121, "epoch": 2383, "lr": 1.1370702343066385e-05}
{"train_loss": 0.02150776982307434, "global_step": 212122, "epoch": 2383, "lr": 1.1370334264988003e-05}
{"train_loss": 0.04566846787929535, "global_step": 212123, "epoch": 2383, "lr": 1.1369966192102927e-05}
{"train_loss": 0.012572266161441803, "global_step": 212124, "epoch": 2383, "lr": 1.1369598124411174e-05}
{"train_loss": 0.08541442453861237, "global_step": 212125, "epoch": 2383, "lr": 1.1369230061912816e-05}
{"train_loss": 0.06664659082889557, "global_step": 212126, "epoch": 2383, "lr": 1.1368862004607888e-05}
{"train_loss": 0.06123040243983269, "global_step": 212127, "epoch": 2383, "lr": 1.1368493952496456e-05}
{"train_loss": 0.07103389501571655, "global_step": 212128, "epoch": 2383, "lr": 1.1368125905578547e-05}
{"train_loss": 0.0512482114136219, "global_step": 212129, "epoch": 2383, "lr": 1.1367757863854234e-05}
{"train_loss": 0.06999298930168152, "global_step": 212130, "epoch": 2383, "lr": 1.1367389827323543e-05}
{"train_loss": 0.0961180105805397, "global_step": 212131, "epoch": 2383, "lr": 1.136702179598655e-05}
{"train_loss": 0.04821157082915306, "global_step": 212132, "epoch": 2383, "lr": 1.1366653769843272e-05}
{"train_loss": 0.09310708194971085, "global_step": 212133, "epoch": 2383, "lr": 1.136628574889379e-05}
{"train_loss": 0.07485762983560562, "global_step": 212134, "epoch": 2383, "lr": 1.1365917733138131e-05}
{"train_loss": 0.07984918355941772, "global_step": 212135, "epoch": 2383, "lr": 1.1365549722576369e-05}
{"train_loss": 0.06238658353686333, "global_step": 212136, "epoch": 2383, "lr": 1.1365181717208518e-05}
{"train_loss": 0.058866068720817566, "global_step": 212137, "epoch": 2383, "lr": 1.1364813717034661e-05}
{"train_loss": 0.07324551790952682, "global_step": 212138, "epoch": 2383, "lr": 1.1364445722054822e-05}
{"train_loss": 0.059648238122463226, "global_step": 212139, "epoch": 2383, "lr": 1.1364077732269079e-05}
{"train_loss": 0.020459692925214767, "global_step": 212140, "epoch": 2383, "lr": 1.1363709747677449e-05}
{"train_loss": 0.06624747812747955, "global_step": 212141, "epoch": 2383, "lr": 1.1363341768280012e-05}
{"train_loss": 0.048992954194545746, "global_step": 212142, "epoch": 2383, "lr": 1.1362973794076787e-05}
{"train_loss": 0.02766772173345089, "global_step": 212143, "epoch": 2383, "lr": 1.1362605825067857e-05}
{"train_loss": 0.0445588193833828, "global_step": 212144, "epoch": 2383, "lr": 1.1362237861253239e-05}
{"train_loss": 0.04821833595633507, "global_step": 212145, "epoch": 2383, "lr": 1.1361869902633004e-05}
{"train_loss": 0.05984232947230339, "global_step": 212146, "epoch": 2383, "lr": 1.1361501949207203e-05}
{"train_loss": 0.05831669643521309, "global_step": 212147, "epoch": 2383, "lr": 1.136113400097586e-05}
{"train_loss": 0.06786900758743286, "global_step": 212148, "epoch": 2383, "lr": 1.1360766057939054e-05}
{"train_loss": 0.04878080636262894, "global_step": 212149, "epoch": 2383, "lr": 1.136039812009681e-05}
{"train_loss": 0.034335844218730927, "global_step": 212150, "epoch": 2383, "lr": 1.1360030187449206e-05}
{"train_loss": 0.07163777202367783, "global_step": 212151, "epoch": 2383, "lr": 1.1359662259996257e-05}
{"train_loss": 0.04685945808887482, "global_step": 212152, "epoch": 2383, "lr": 1.1359294337738047e-05}
{"train_loss": 0.06697189062833786, "global_step": 212153, "epoch": 2383, "lr": 1.1358926420674593e-05}
{"train_loss": 0.12485391646623611, "global_step": 212154, "epoch": 2383, "lr": 1.1358558508805973e-05}
{"train_loss": 0.07715052366256714, "global_step": 212155, "epoch": 2383, "lr": 1.1358190602132212e-05}
{"train_loss": 0.10456427931785583, "global_step": 212156, "epoch": 2383, "lr": 1.1357822700653387e-05}
{"train_loss": 0.0781545341014862, "global_step": 212157, "epoch": 2383, "lr": 1.1357454804369516e-05}
{"train_loss": 0.07373562455177307, "global_step": 212158, "epoch": 2383, "lr": 1.1357086913280663e-05}
{"train_loss": 0.0668306052684784, "global_step": 212159, "epoch": 2383, "lr": 1.1356719027386892e-05}
{"train_loss": 0.031032292172312737, "global_step": 212160, "epoch": 2383, "lr": 1.1356351146688227e-05}
{"train_loss": 0.0600430853664875, "global_step": 212161, "epoch": 2383, "lr": 1.1355983271184745e-05}
{"train_loss": 0.051562558859586716, "global_step": 212162, "epoch": 2383, "lr": 1.1355615400876463e-05}
{"train_loss": 0.06044155731797218, "global_step": 212163, "epoch": 2383, "lr": 1.1355247535763463e-05}
{"train_loss": 0.08199388533830643, "global_step": 212164, "epoch": 2383, "lr": 1.1354879675845764e-05}
{"train_loss": 0.11340904235839844, "global_step": 212165, "epoch": 2383, "lr": 1.1354511821123447e-05}
{"train_loss": 0.056772276759147644, "global_step": 212166, "epoch": 2383, "lr": 1.1354143971596531e-05}
{"train_loss": 0.03803427144885063, "global_step": 212167, "epoch": 2383, "lr": 1.1353776127265092e-05}
{"train_loss": 0.03419013321399689, "global_step": 212168, "epoch": 2383, "lr": 1.1353408288129152e-05}
{"train_loss": 0.044328704476356506, "global_step": 212169, "epoch": 2383, "lr": 1.1353040454188791e-05}
{"train_loss": 0.06437547504901886, "global_step": 212170, "epoch": 2383, "lr": 1.1352672625444028e-05}
{"train_loss": 0.0819198414683342, "global_step": 212171, "epoch": 2383, "lr": 1.1352304801894937e-05}
{"train_loss": 0.07242578268051147, "global_step": 212172, "epoch": 2383, "lr": 1.135193698354155e-05}
{"train_loss": 0.09441215544939041, "global_step": 212173, "epoch": 2383, "lr": 1.135156917038393e-05}
{"train_loss": 0.043946824967861176, "global_step": 212174, "epoch": 2383, "lr": 1.1351201362422115e-05}
{"train_loss": 0.06044174012926857, "global_step": 212175, "epoch": 2383, "lr": 1.1350833559656171e-05, "val_loss": 8.693859100341797}
{"train_loss": 0.03938993066549301, "global_step": 212176, "epoch": 2384, "lr": 1.135046576208612e-05}
{"train_loss": 0.07631529122591019, "global_step": 212177, "epoch": 2384, "lr": 1.135009796971203e-05}
{"train_loss": 0.04101015254855156, "global_step": 212178, "epoch": 2384, "lr": 1.1349730182533963e-05}
{"train_loss": 0.044207483530044556, "global_step": 212179, "epoch": 2384, "lr": 1.134936240055195e-05}
{"train_loss": 0.03870460018515587, "global_step": 212180, "epoch": 2384, "lr": 1.1348994623766034e-05}
{"train_loss": 0.07990323007106781, "global_step": 212181, "epoch": 2384, "lr": 1.1348626852176287e-05}
{"train_loss": 0.048847585916519165, "global_step": 212182, "epoch": 2384, "lr": 1.134825908578273e-05}
{"train_loss": 0.058807745575904846, "global_step": 212183, "epoch": 2384, "lr": 1.1347891324585447e-05}
{"train_loss": 0.10783287137746811, "global_step": 212184, "epoch": 2384, "lr": 1.1347523568584466e-05}
{"train_loss": 0.06823727488517761, "global_step": 212185, "epoch": 2384, "lr": 1.1347155817779825e-05}
{"train_loss": 0.0687926784157753, "global_step": 212186, "epoch": 2384, "lr": 1.1346788072171605e-05}
{"train_loss": 0.03082461841404438, "global_step": 212187, "epoch": 2384, "lr": 1.1346420331759822e-05}
{"train_loss": 0.015436341986060143, "global_step": 212188, "epoch": 2384, "lr": 1.1346052596544554e-05}
{"train_loss": 0.08653336018323898, "global_step": 212189, "epoch": 2384, "lr": 1.1345684866525825e-05}
{"train_loss": 0.08175325393676758, "global_step": 212190, "epoch": 2384, "lr": 1.1345317141703716e-05}
{"train_loss": 0.12192733585834503, "global_step": 212191, "epoch": 2384, "lr": 1.1344949422078238e-05}
{"train_loss": 0.0798235535621643, "global_step": 212192, "epoch": 2384, "lr": 1.1344581707649476e-05}
{"train_loss": 0.15316429734230042, "global_step": 212193, "epoch": 2384, "lr": 1.1344213998417452e-05}
{"train_loss": 0.06790956854820251, "global_step": 212194, "epoch": 2384, "lr": 1.134384629438222e-05}
{"train_loss": 0.04836725443601608, "global_step": 212195, "epoch": 2384, "lr": 1.134347859554386e-05}
{"train_loss": 0.0909658744931221, "global_step": 212196, "epoch": 2384, "lr": 1.1343110901902382e-05}
{"train_loss": 0.033053867518901825, "global_step": 212197, "epoch": 2384, "lr": 1.1342743213457863e-05}
{"train_loss": 0.05238533020019531, "global_step": 212198, "epoch": 2384, "lr": 1.1342375530210331e-05}
{"train_loss": 0.07831161469221115, "global_step": 212199, "epoch": 2384, "lr": 1.1342007852159853e-05}
{"train_loss": 0.06829246878623962, "global_step": 212200, "epoch": 2384, "lr": 1.1341640179306463e-05}
{"train_loss": 0.05850793421268463, "global_step": 212201, "epoch": 2384, "lr": 1.134127251165023e-05}
{"train_loss": 0.0579565092921257, "global_step": 212202, "epoch": 2384, "lr": 1.1340904849191175e-05}
{"train_loss": 0.05647186189889908, "global_step": 212203, "epoch": 2384, "lr": 1.1340537191929379e-05}
{"train_loss": 0.025959651917219162, "global_step": 212204, "epoch": 2384, "lr": 1.1340169539864864e-05}
{"train_loss": 0.04860897362232208, "global_step": 212205, "epoch": 2384, "lr": 1.1339801892997709e-05}
{"train_loss": 0.07026955485343933, "global_step": 212206, "epoch": 2384, "lr": 1.133943425132793e-05}
{"train_loss": 0.06497856229543686, "global_step": 212207, "epoch": 2384, "lr": 1.1339066614855604e-05}
{"train_loss": 0.0897686779499054, "global_step": 212208, "epoch": 2384, "lr": 1.1338698983580759e-05}
{"train_loss": 0.04104167968034744, "global_step": 212209, "epoch": 2384, "lr": 1.1338331357503468e-05}
{"train_loss": 0.06235151365399361, "global_step": 212210, "epoch": 2384, "lr": 1.1337963736623752e-05}
{"train_loss": 0.11401714384555817, "global_step": 212211, "epoch": 2384, "lr": 1.133759612094169e-05}
{"train_loss": 0.09352696686983109, "global_step": 212212, "epoch": 2384, "lr": 1.1337228510457305e-05}
{"train_loss": 0.06358499079942703, "global_step": 212213, "epoch": 2384, "lr": 1.1336860905170655e-05}
{"train_loss": 0.0555044487118721, "global_step": 212214, "epoch": 2384, "lr": 1.1336493305081808e-05}
{"train_loss": 0.042374204844236374, "global_step": 212215, "epoch": 2384, "lr": 1.1336125710190787e-05}
{"train_loss": 0.030877141281962395, "global_step": 212216, "epoch": 2384, "lr": 1.1335758120497665e-05}
{"train_loss": 0.06972505897283554, "global_step": 212217, "epoch": 2384, "lr": 1.1335390536002477e-05}
{"train_loss": 0.1027505025267601, "global_step": 212218, "epoch": 2384, "lr": 1.133502295670526e-05}
{"train_loss": 0.08087754994630814, "global_step": 212219, "epoch": 2384, "lr": 1.1334655382606097e-05}
{"train_loss": 0.066238634288311, "global_step": 212220, "epoch": 2384, "lr": 1.1334287813705003e-05}
{"train_loss": 0.07648777961730957, "global_step": 212221, "epoch": 2384, "lr": 1.133392025000205e-05}
{"train_loss": 0.06137707456946373, "global_step": 212222, "epoch": 2384, "lr": 1.1333552691497284e-05}
{"train_loss": 0.07150423526763916, "global_step": 212223, "epoch": 2384, "lr": 1.1333185138190738e-05}
{"train_loss": 0.058877211064100266, "global_step": 212224, "epoch": 2384, "lr": 1.133281759008249e-05}
{"train_loss": 0.05703715607523918, "global_step": 212225, "epoch": 2384, "lr": 1.1332450047172554e-05}
{"train_loss": 0.07876881957054138, "global_step": 212226, "epoch": 2384, "lr": 1.1332082509461017e-05}
{"train_loss": 0.06578100472688675, "global_step": 212227, "epoch": 2384, "lr": 1.1331714976947893e-05}
{"train_loss": 0.07133223861455917, "global_step": 212228, "epoch": 2384, "lr": 1.1331347449633267e-05}
{"train_loss": 0.015684667974710464, "global_step": 212229, "epoch": 2384, "lr": 1.1330979927517155e-05}
{"train_loss": 0.02942838706076145, "global_step": 212230, "epoch": 2384, "lr": 1.1330612410599617e-05}
{"train_loss": 0.04503074660897255, "global_step": 212231, "epoch": 2384, "lr": 1.1330244898880726e-05}
{"train_loss": 0.030192241072654724, "global_step": 212232, "epoch": 2384, "lr": 1.1329877392360495e-05}
{"train_loss": 0.0951157808303833, "global_step": 212233, "epoch": 2384, "lr": 1.1329509891039008e-05}
{"train_loss": 0.05123661458492279, "global_step": 212234, "epoch": 2384, "lr": 1.132914239491628e-05}
{"train_loss": 0.023879654705524445, "global_step": 212235, "epoch": 2384, "lr": 1.1328774903992389e-05}
{"train_loss": 0.0661264955997467, "global_step": 212236, "epoch": 2384, "lr": 1.1328407418267367e-05}
{"train_loss": 0.03937782719731331, "global_step": 212237, "epoch": 2384, "lr": 1.1328039937741274e-05}
{"train_loss": 0.0328802652657032, "global_step": 212238, "epoch": 2384, "lr": 1.1327672462414146e-05}
{"train_loss": 0.06018298864364624, "global_step": 212239, "epoch": 2384, "lr": 1.1327304992286059e-05}
{"train_loss": 0.04234199598431587, "global_step": 212240, "epoch": 2384, "lr": 1.1326937527357024e-05}
{"train_loss": 0.045048557221889496, "global_step": 212241, "epoch": 2384, "lr": 1.1326570067627123e-05}
{"train_loss": 0.06865532696247101, "global_step": 212242, "epoch": 2384, "lr": 1.1326202613096387e-05}
{"train_loss": 0.041776712983846664, "global_step": 212243, "epoch": 2384, "lr": 1.132583516376488e-05}
{"train_loss": 0.040522608906030655, "global_step": 212244, "epoch": 2384, "lr": 1.1325467719632633e-05}
{"train_loss": 0.010706765577197075, "global_step": 212245, "epoch": 2384, "lr": 1.1325100280699713e-05}
{"train_loss": 0.08268315345048904, "global_step": 212246, "epoch": 2384, "lr": 1.1324732846966151e-05}
{"train_loss": 0.05858384072780609, "global_step": 212247, "epoch": 2384, "lr": 1.1324365418432025e-05}
{"train_loss": 0.03715699166059494, "global_step": 212248, "epoch": 2384, "lr": 1.1323997995097351e-05}
{"train_loss": 0.05481373146176338, "global_step": 212249, "epoch": 2384, "lr": 1.132363057696219e-05}
{"train_loss": 0.13839665055274963, "global_step": 212250, "epoch": 2384, "lr": 1.1323263164026615e-05}
{"train_loss": 0.036851610988378525, "global_step": 212251, "epoch": 2384, "lr": 1.132289575629064e-05}
{"train_loss": 0.018333258107304573, "global_step": 212252, "epoch": 2384, "lr": 1.132252835375434e-05}
{"train_loss": 0.0552515871822834, "global_step": 212253, "epoch": 2384, "lr": 1.1322160956417749e-05}
{"train_loss": 0.07892323285341263, "global_step": 212254, "epoch": 2384, "lr": 1.132179356428093e-05}
{"train_loss": 0.04405174404382706, "global_step": 212255, "epoch": 2384, "lr": 1.1321426177343924e-05}
{"train_loss": 0.0861392691731453, "global_step": 212256, "epoch": 2384, "lr": 1.132105879560677e-05}
{"train_loss": 0.1461549550294876, "global_step": 212257, "epoch": 2384, "lr": 1.132069141906954e-05}
{"train_loss": 0.01616167277097702, "global_step": 212258, "epoch": 2384, "lr": 1.1320324047732261e-05}
{"train_loss": 0.15160231292247772, "global_step": 212259, "epoch": 2384, "lr": 1.1319956681595002e-05}
{"train_loss": 0.05792033672332764, "global_step": 212260, "epoch": 2384, "lr": 1.1319589320657792e-05}
{"train_loss": 0.05820605903863907, "global_step": 212261, "epoch": 2384, "lr": 1.1319221964920707e-05}
{"train_loss": 0.03587052971124649, "global_step": 212262, "epoch": 2384, "lr": 1.1318854614383778e-05}
{"train_loss": 0.05794597044587135, "global_step": 212263, "epoch": 2384, "lr": 1.1318487269047045e-05}
{"train_loss": 0.06277524391084574, "global_step": 212264, "epoch": 2384, "lr": 1.1318119928910587e-05, "val_loss": 8.761412620544434}
{"train_loss": 0.04550290107727051, "global_step": 212265, "epoch": 2385, "lr": 1.1317752593974418e-05}
{"train_loss": 0.02639176696538925, "global_step": 212266, "epoch": 2385, "lr": 1.1317385264238623e-05}
{"train_loss": 0.10682955384254456, "global_step": 212267, "epoch": 2385, "lr": 1.1317017939703217e-05}
{"train_loss": 0.06489275395870209, "global_step": 212268, "epoch": 2385, "lr": 1.1316650620368268e-05}
{"train_loss": 0.03016204945743084, "global_step": 212269, "epoch": 2385, "lr": 1.1316283306233838e-05}
{"train_loss": 0.049833305180072784, "global_step": 212270, "epoch": 2385, "lr": 1.1315915997299947e-05}
{"train_loss": 0.06077076494693756, "global_step": 212271, "epoch": 2385, "lr": 1.1315548693566674e-05}
{"train_loss": 0.019516633823513985, "global_step": 212272, "epoch": 2385, "lr": 1.1315181395034041e-05}
{"train_loss": 0.04149181395769119, "global_step": 212273, "epoch": 2385, "lr": 1.1314814101702127e-05}
{"train_loss": 0.06530685722827911, "global_step": 212274, "epoch": 2385, "lr": 1.1314446813570945e-05}
{"train_loss": 0.05012291669845581, "global_step": 212275, "epoch": 2385, "lr": 1.1314079530640581e-05}
{"train_loss": 0.06698282808065414, "global_step": 212276, "epoch": 2385, "lr": 1.1313712252911052e-05}
{"train_loss": 0.038273848593235016, "global_step": 212277, "epoch": 2385, "lr": 1.131334498038244e-05}
{"train_loss": 0.07058878242969513, "global_step": 212278, "epoch": 2385, "lr": 1.1312977713054756e-05}
{"train_loss": 0.07782583683729172, "global_step": 212279, "epoch": 2385, "lr": 1.1312610450928091e-05}
{"train_loss": 0.04908443242311478, "global_step": 212280, "epoch": 2385, "lr": 1.1312243194002459e-05}
{"train_loss": 0.07387189567089081, "global_step": 212281, "epoch": 2385, "lr": 1.1311875942277939e-05}
{"train_loss": 0.0331314317882061, "global_step": 212282, "epoch": 2385, "lr": 1.1311508695754552e-05}
{"train_loss": 0.06775709241628647, "global_step": 212283, "epoch": 2385, "lr": 1.1311141454432373e-05}
{"train_loss": 0.05530908331274986, "global_step": 212284, "epoch": 2385, "lr": 1.1310774218311427e-05}
{"train_loss": 0.04707004502415657, "global_step": 212285, "epoch": 2385, "lr": 1.1310406987391776e-05}
{"train_loss": 0.03163129836320877, "global_step": 212286, "epoch": 2385, "lr": 1.131003976167348e-05}
{"train_loss": 0.040330808609724045, "global_step": 212287, "epoch": 2385, "lr": 1.1309672541156568e-05}
{"train_loss": 0.061394672840833664, "global_step": 212288, "epoch": 2385, "lr": 1.1309305325841112e-05}
{"train_loss": 0.04814939945936203, "global_step": 212289, "epoch": 2385, "lr": 1.130893811572713e-05}
{"train_loss": 0.06933292746543884, "global_step": 212290, "epoch": 2385, "lr": 1.130857091081471e-05}
{"train_loss": 0.07846735417842865, "global_step": 212291, "epoch": 2385, "lr": 1.1308203711103866e-05}
{"train_loss": 0.05347288027405739, "global_step": 212292, "epoch": 2385, "lr": 1.1307836516594672e-05}
{"train_loss": 0.0423872210085392, "global_step": 212293, "epoch": 2385, "lr": 1.1307469327287156e-05}
{"train_loss": 0.03319916874170303, "global_step": 212294, "epoch": 2385, "lr": 1.1307102143181397e-05}
{"train_loss": 0.07378707081079483, "global_step": 212295, "epoch": 2385, "lr": 1.1306734964277421e-05}
{"train_loss": 0.02002653293311596, "global_step": 212296, "epoch": 2385, "lr": 1.1306367790575273e-05}
{"train_loss": 0.03539721295237541, "global_step": 212297, "epoch": 2385, "lr": 1.130600062207502e-05}
{"train_loss": 0.050310805439949036, "global_step": 212298, "epoch": 2385, "lr": 1.1305633458776693e-05}
{"train_loss": 0.07333652675151825, "global_step": 212299, "epoch": 2385, "lr": 1.1305266300680367e-05}
{"train_loss": 0.020642155781388283, "global_step": 212300, "epoch": 2385, "lr": 1.130489914778608e-05}
{"train_loss": 0.043770574033260345, "global_step": 212301, "epoch": 2385, "lr": 1.1304532000093859e-05}
{"train_loss": 0.06551439315080643, "global_step": 212302, "epoch": 2385, "lr": 1.1304164857603789e-05}
{"train_loss": 0.05084646865725517, "global_step": 212303, "epoch": 2385, "lr": 1.1303797720315883e-05}
{"train_loss": 0.029095493257045746, "global_step": 212304, "epoch": 2385, "lr": 1.1303430588230212e-05}
{"train_loss": 0.05881604924798012, "global_step": 212305, "epoch": 2385, "lr": 1.1303063461346841e-05}
{"train_loss": 0.07200313359498978, "global_step": 212306, "epoch": 2385, "lr": 1.130269633966578e-05}
{"train_loss": 0.05080171674489975, "global_step": 212307, "epoch": 2385, "lr": 1.130232922318712e-05}
{"train_loss": 0.10475306957960129, "global_step": 212308, "epoch": 2385, "lr": 1.1301962111910875e-05}
{"train_loss": 0.0334639810025692, "global_step": 212309, "epoch": 2385, "lr": 1.1301595005837123e-05}
{"train_loss": 0.05513545870780945, "global_step": 212310, "epoch": 2385, "lr": 1.130122790496589e-05}
{"train_loss": 0.08554375171661377, "global_step": 212311, "epoch": 2385, "lr": 1.1300860809297243e-05}
{"train_loss": 0.028464974835515022, "global_step": 212312, "epoch": 2385, "lr": 1.1300493718831213e-05}
{"train_loss": 0.03633370250463486, "global_step": 212313, "epoch": 2385, "lr": 1.130012663356787e-05}
{"train_loss": 0.0686117485165596, "global_step": 212314, "epoch": 2385, "lr": 1.1299759553507245e-05}
{"train_loss": 0.03996794670820236, "global_step": 212315, "epoch": 2385, "lr": 1.1299392478649406e-05}
{"train_loss": 0.05525883287191391, "global_step": 212316, "epoch": 2385, "lr": 1.129902540899438e-05}
{"train_loss": 0.054857395589351654, "global_step": 212317, "epoch": 2385, "lr": 1.129865834454224e-05}
{"train_loss": 0.020460069179534912, "global_step": 212318, "epoch": 2385, "lr": 1.1298291285293011e-05}
{"train_loss": 0.06058238074183464, "global_step": 212319, "epoch": 2385, "lr": 1.1297924231246765e-05}
{"train_loss": 0.04970044270157814, "global_step": 212320, "epoch": 2385, "lr": 1.1297557182403535e-05}
{"train_loss": 0.039753202348947525, "global_step": 212321, "epoch": 2385, "lr": 1.1297190138763369e-05}
{"train_loss": 0.08436048030853271, "global_step": 212322, "epoch": 2385, "lr": 1.1296823100326343e-05}
{"train_loss": 0.05711182951927185, "global_step": 212323, "epoch": 2385, "lr": 1.129645606709247e-05}
{"train_loss": 0.07617034763097763, "global_step": 212324, "epoch": 2385, "lr": 1.1296089039061836e-05}
{"train_loss": 0.06831427663564682, "global_step": 212325, "epoch": 2385, "lr": 1.1295722016234457e-05}
{"train_loss": 0.029866397380828857, "global_step": 212326, "epoch": 2385, "lr": 1.1295354998610402e-05}
{"train_loss": 0.0332181453704834, "global_step": 212327, "epoch": 2385, "lr": 1.1294987986189709e-05}
{"train_loss": 0.030132165178656578, "global_step": 212328, "epoch": 2385, "lr": 1.1294620978972442e-05}
{"train_loss": 0.04740653559565544, "global_step": 212329, "epoch": 2385, "lr": 1.1294253976958629e-05}
{"train_loss": 0.030275877565145493, "global_step": 212330, "epoch": 2385, "lr": 1.129388698014835e-05}
{"train_loss": 0.0691642016172409, "global_step": 212331, "epoch": 2385, "lr": 1.1293519988541617e-05}
{"train_loss": 0.08139882236719131, "global_step": 212332, "epoch": 2385, "lr": 1.1293153002138518e-05}
{"train_loss": 0.07752921432256699, "global_step": 212333, "epoch": 2385, "lr": 1.1292786020939078e-05}
{"train_loss": 0.0540277324616909, "global_step": 212334, "epoch": 2385, "lr": 1.1292419044943337e-05}
{"train_loss": 0.052894242107868195, "global_step": 212335, "epoch": 2385, "lr": 1.1292052074151372e-05}
{"train_loss": 0.05512411147356033, "global_step": 212336, "epoch": 2385, "lr": 1.129168510856321e-05}
{"train_loss": 0.08987358957529068, "global_step": 212337, "epoch": 2385, "lr": 1.129131814817892e-05}
{"train_loss": 0.1096864640712738, "global_step": 212338, "epoch": 2385, "lr": 1.1290951192998539e-05}
{"train_loss": 0.05219045281410217, "global_step": 212339, "epoch": 2385, "lr": 1.1290584243022107e-05}
{"train_loss": 0.03423512727022171, "global_step": 212340, "epoch": 2385, "lr": 1.1290217298249678e-05}
{"train_loss": 0.07613607496023178, "global_step": 212341, "epoch": 2385, "lr": 1.1289850358681326e-05}
{"train_loss": 0.03439132496714592, "global_step": 212342, "epoch": 2385, "lr": 1.1289483424317065e-05}
{"train_loss": 0.045245714485645294, "global_step": 212343, "epoch": 2385, "lr": 1.1289116495156981e-05}
{"train_loss": 0.11078402400016785, "global_step": 212344, "epoch": 2385, "lr": 1.1288749571201084e-05}
{"train_loss": 0.09046852588653564, "global_step": 212345, "epoch": 2385, "lr": 1.1288382652449458e-05}
{"train_loss": 0.09105933457612991, "global_step": 212346, "epoch": 2385, "lr": 1.1288015738902124e-05}
{"train_loss": 0.05924268811941147, "global_step": 212347, "epoch": 2385, "lr": 1.128764883055916e-05}
{"train_loss": 0.0755000188946724, "global_step": 212348, "epoch": 2385, "lr": 1.1287281927420585e-05}
{"train_loss": 0.13046713173389435, "global_step": 212349, "epoch": 2385, "lr": 1.1286915029486479e-05}
{"train_loss": 0.09816751629114151, "global_step": 212350, "epoch": 2385, "lr": 1.1286548136756858e-05}
{"train_loss": 0.042732492089271545, "global_step": 212351, "epoch": 2385, "lr": 1.1286181249231803e-05}
{"train_loss": 0.10733425617218018, "global_step": 212352, "epoch": 2385, "lr": 1.1285814366911335e-05}
{"train_loss": 0.05774260563461968, "global_step": 212353, "epoch": 2385, "lr": 1.128544748979553e-05, "val_loss": 8.76423168182373, "train_action_mse_error": 22.306949615478516}
{"train_loss": 0.0979730635881424, "global_step": 212354, "epoch": 2386, "lr": 1.1285080617884414e-05}
{"train_loss": 0.02963441237807274, "global_step": 212355, "epoch": 2386, "lr": 1.1284713751178062e-05}
{"train_loss": 0.05871288478374481, "global_step": 212356, "epoch": 2386, "lr": 1.128434688967649e-05}
{"train_loss": 0.07061977684497833, "global_step": 212357, "epoch": 2386, "lr": 1.1283980033379765e-05}
{"train_loss": 0.0459887869656086, "global_step": 212358, "epoch": 2386, "lr": 1.1283613182287955e-05}
{"train_loss": 0.07366514205932617, "global_step": 212359, "epoch": 2386, "lr": 1.1283246336401076e-05}
{"train_loss": 0.060560427606105804, "global_step": 212360, "epoch": 2386, "lr": 1.1282879495719206e-05}
{"train_loss": 0.0878499373793602, "global_step": 212361, "epoch": 2386, "lr": 1.1282512660242367e-05}
{"train_loss": 0.06471763551235199, "global_step": 212362, "epoch": 2386, "lr": 1.1282145829970642e-05}
{"train_loss": 0.050468847155570984, "global_step": 212363, "epoch": 2386, "lr": 1.1281779004904041e-05}
{"train_loss": 0.03482164442539215, "global_step": 212364, "epoch": 2386, "lr": 1.1281412185042644e-05}
{"train_loss": 0.05294651910662651, "global_step": 212365, "epoch": 2386, "lr": 1.1281045370386478e-05}
{"train_loss": 0.06928472965955734, "global_step": 212366, "epoch": 2386, "lr": 1.128067856093562e-05}
{"train_loss": 0.03277885168790817, "global_step": 212367, "epoch": 2386, "lr": 1.1280311756690088e-05}
{"train_loss": 0.022366715595126152, "global_step": 212368, "epoch": 2386, "lr": 1.1279944957649957e-05}
{"train_loss": 0.028375914320349693, "global_step": 212369, "epoch": 2386, "lr": 1.1279578163815258e-05}
{"train_loss": 0.05990367382764816, "global_step": 212370, "epoch": 2386, "lr": 1.1279211375186055e-05}
{"train_loss": 0.09155412018299103, "global_step": 212371, "epoch": 2386, "lr": 1.1278844591762394e-05}
{"train_loss": 0.03753121942281723, "global_step": 212372, "epoch": 2386, "lr": 1.1278477813544303e-05}
{"train_loss": 0.05964517965912819, "global_step": 212373, "epoch": 2386, "lr": 1.1278111040531863e-05}
{"train_loss": 0.1094776839017868, "global_step": 212374, "epoch": 2386, "lr": 1.12777442727251e-05}
{"train_loss": 0.07989063113927841, "global_step": 212375, "epoch": 2386, "lr": 1.1277377510124082e-05}
{"train_loss": 0.048998042941093445, "global_step": 212376, "epoch": 2386, "lr": 1.1277010752728834e-05}
{"train_loss": 0.05042941868305206, "global_step": 212377, "epoch": 2386, "lr": 1.1276644000539438e-05}
{"train_loss": 0.035153213888406754, "global_step": 212378, "epoch": 2386, "lr": 1.1276277253555906e-05}
{"train_loss": 0.08166231960058212, "global_step": 212379, "epoch": 2386, "lr": 1.1275910511778325e-05}
{"train_loss": 0.039662666618824005, "global_step": 212380, "epoch": 2386, "lr": 1.127554377520671e-05}
{"train_loss": 0.06417497992515564, "global_step": 212381, "epoch": 2386, "lr": 1.127517704384114e-05}
{"train_loss": 0.06122858077287674, "global_step": 212382, "epoch": 2386, "lr": 1.127481031768164e-05}
{"train_loss": 0.065879687666893, "global_step": 212383, "epoch": 2386, "lr": 1.1274443596728278e-05}
{"train_loss": 0.05327422916889191, "global_step": 212384, "epoch": 2386, "lr": 1.1274076880981083e-05}
{"train_loss": 0.056809645146131516, "global_step": 212385, "epoch": 2386, "lr": 1.1273710170440133e-05}
{"train_loss": 0.041653938591480255, "global_step": 212386, "epoch": 2386, "lr": 1.1273343465105441e-05}
{"train_loss": 0.03127123415470123, "global_step": 212387, "epoch": 2386, "lr": 1.1272976764977094e-05}
{"train_loss": 0.08659727871417999, "global_step": 212388, "epoch": 2386, "lr": 1.1272610070055112e-05}
{"train_loss": 0.08659259974956512, "global_step": 212389, "epoch": 2386, "lr": 1.1272243380339565e-05}
{"train_loss": 0.05058487877249718, "global_step": 212390, "epoch": 2386, "lr": 1.127187669583048e-05}
{"train_loss": 0.05754195898771286, "global_step": 212391, "epoch": 2386, "lr": 1.1271510016527931e-05}
{"train_loss": 0.08407788723707199, "global_step": 212392, "epoch": 2386, "lr": 1.1271143342431945e-05}
{"train_loss": 0.04399913176894188, "global_step": 212393, "epoch": 2386, "lr": 1.1270776673542593e-05}
{"train_loss": 0.033048391342163086, "global_step": 212394, "epoch": 2386, "lr": 1.1270410009859906e-05}
{"train_loss": 0.0727837085723877, "global_step": 212395, "epoch": 2386, "lr": 1.127004335138393e-05}
{"train_loss": 0.06256124377250671, "global_step": 212396, "epoch": 2386, "lr": 1.1269676698114746e-05}
{"train_loss": 0.04210754483938217, "global_step": 212397, "epoch": 2386, "lr": 1.1269310050052367e-05}
{"train_loss": 0.06230190768837929, "global_step": 212398, "epoch": 2386, "lr": 1.1268943407196868e-05}
{"train_loss": 0.05635443329811096, "global_step": 212399, "epoch": 2386, "lr": 1.1268576769548272e-05}
{"train_loss": 0.051133349537849426, "global_step": 212400, "epoch": 2386, "lr": 1.1268210137106666e-05}
{"train_loss": 0.04349445551633835, "global_step": 212401, "epoch": 2386, "lr": 1.1267843509872055e-05}
{"train_loss": 0.07326415181159973, "global_step": 212402, "epoch": 2386, "lr": 1.126747688784453e-05}
{"train_loss": 0.06004675477743149, "global_step": 212403, "epoch": 2386, "lr": 1.12671102710241e-05}
{"train_loss": 0.060071032494306564, "global_step": 212404, "epoch": 2386, "lr": 1.1266743659410856e-05}
{"train_loss": 0.030754685401916504, "global_step": 212405, "epoch": 2386, "lr": 1.1266377053004806e-05}
{"train_loss": 0.04707660898566246, "global_step": 212406, "epoch": 2386, "lr": 1.1266010451806042e-05}
{"train_loss": 0.0894479751586914, "global_step": 212407, "epoch": 2386, "lr": 1.1265643855814567e-05}
{"train_loss": 0.034741804003715515, "global_step": 212408, "epoch": 2386, "lr": 1.1265277265030472e-05}
{"train_loss": 0.02267884649336338, "global_step": 212409, "epoch": 2386, "lr": 1.1264910679453777e-05}
{"train_loss": 0.09998737275600433, "global_step": 212410, "epoch": 2386, "lr": 1.1264544099084556e-05}
{"train_loss": 0.09342657774686813, "global_step": 212411, "epoch": 2386, "lr": 1.1264177523922843e-05}
{"train_loss": 0.042743660509586334, "global_step": 212412, "epoch": 2386, "lr": 1.1263810953968673e-05}
{"train_loss": 0.07798145711421967, "global_step": 212413, "epoch": 2386, "lr": 1.1263444389222132e-05}
{"train_loss": 0.029964542016386986, "global_step": 212414, "epoch": 2386, "lr": 1.1263077829683227e-05}
{"train_loss": 0.06616820394992828, "global_step": 212415, "epoch": 2386, "lr": 1.126271127535205e-05}
{"train_loss": 0.07439015060663223, "global_step": 212416, "epoch": 2386, "lr": 1.1262344726228613e-05}
{"train_loss": 0.01568184606730938, "global_step": 212417, "epoch": 2386, "lr": 1.1261978182312993e-05}
{"train_loss": 0.09837798029184341, "global_step": 212418, "epoch": 2386, "lr": 1.126161164360522e-05}
{"train_loss": 0.031615179032087326, "global_step": 212419, "epoch": 2386, "lr": 1.1261245110105357e-05}
{"train_loss": 0.06542760878801346, "global_step": 212420, "epoch": 2386, "lr": 1.1260878581813439e-05}
{"train_loss": 0.06329892575740814, "global_step": 212421, "epoch": 2386, "lr": 1.1260512058729534e-05}
{"train_loss": 0.06265249103307724, "global_step": 212422, "epoch": 2386, "lr": 1.1260145540853673e-05}
{"train_loss": 0.08084814250469208, "global_step": 212423, "epoch": 2386, "lr": 1.1259779028185919e-05}
{"train_loss": 0.05045752599835396, "global_step": 212424, "epoch": 2386, "lr": 1.1259412520726309e-05}
{"train_loss": 0.04619879648089409, "global_step": 212425, "epoch": 2386, "lr": 1.1259046018474906e-05}
{"train_loss": 0.040230944752693176, "global_step": 212426, "epoch": 2386, "lr": 1.1258679521431741e-05}
{"train_loss": 0.05140829458832741, "global_step": 212427, "epoch": 2386, "lr": 1.125831302959689e-05}
{"train_loss": 0.0703405812382698, "global_step": 212428, "epoch": 2386, "lr": 1.1257946542970376e-05}
{"train_loss": 0.04929600656032562, "global_step": 212429, "epoch": 2386, "lr": 1.1257580061552264e-05}
{"train_loss": 0.05372229591012001, "global_step": 212430, "epoch": 2386, "lr": 1.1257213585342591e-05}
{"train_loss": 0.07077236473560333, "global_step": 212431, "epoch": 2386, "lr": 1.1256847114341412e-05}
{"train_loss": 0.06914404034614563, "global_step": 212432, "epoch": 2386, "lr": 1.1256480648548789e-05}
{"train_loss": 0.062247321009635925, "global_step": 212433, "epoch": 2386, "lr": 1.125611418796475e-05}
{"train_loss": 0.056889381259679794, "global_step": 212434, "epoch": 2386, "lr": 1.1255747732589367e-05}
{"train_loss": 0.07193802297115326, "global_step": 212435, "epoch": 2386, "lr": 1.125538128242266e-05}
{"train_loss": 0.06729435920715332, "global_step": 212436, "epoch": 2386, "lr": 1.1255014837464712e-05}
{"train_loss": 0.07211171835660934, "global_step": 212437, "epoch": 2386, "lr": 1.1254648397715539e-05}
{"train_loss": 0.06694744527339935, "global_step": 212438, "epoch": 2386, "lr": 1.1254281963175218e-05}
{"train_loss": 0.02707325480878353, "global_step": 212439, "epoch": 2386, "lr": 1.1253915533843779e-05}
{"train_loss": 0.09067018330097198, "global_step": 212440, "epoch": 2386, "lr": 1.125354910972129e-05}
{"train_loss": 0.03195545822381973, "global_step": 212441, "epoch": 2386, "lr": 1.1253182690807773e-05}
{"train_loss": 0.058247871194662675, "global_step": 212442, "epoch": 2386, "lr": 1.1252816277103306e-05, "val_loss": 8.822556495666504}
{"train_loss": 0.05805206671357155, "global_step": 212443, "epoch": 2387, "lr": 1.1252449868607916e-05}
{"train_loss": 0.02434314228594303, "global_step": 212444, "epoch": 2387, "lr": 1.1252083465321672e-05}
{"train_loss": 0.04650231450796127, "global_step": 212445, "epoch": 2387, "lr": 1.12517170672446e-05}
{"train_loss": 0.07876387238502502, "global_step": 212446, "epoch": 2387, "lr": 1.1251350674376775e-05}
{"train_loss": 0.04435324668884277, "global_step": 212447, "epoch": 2387, "lr": 1.1250984286718224e-05}
{"train_loss": 0.05214798450469971, "global_step": 212448, "epoch": 2387, "lr": 1.1250617904269017e-05}
{"train_loss": 0.034063614904880524, "global_step": 212449, "epoch": 2387, "lr": 1.1250251527029193e-05}
{"train_loss": 0.0654822438955307, "global_step": 212450, "epoch": 2387, "lr": 1.124988515499879e-05}
{"train_loss": 0.059143468737602234, "global_step": 212451, "epoch": 2387, "lr": 1.1249518788177876e-05}
{"train_loss": 0.05889945477247238, "global_step": 212452, "epoch": 2387, "lr": 1.1249152426566478e-05}
{"train_loss": 0.06054965779185295, "global_step": 212453, "epoch": 2387, "lr": 1.124878607016468e-05}
{"train_loss": 0.03228997439146042, "global_step": 212454, "epoch": 2387, "lr": 1.1248419718972492e-05}
{"train_loss": 0.047380998730659485, "global_step": 212455, "epoch": 2387, "lr": 1.1248053372989992e-05}
{"train_loss": 0.05129052326083183, "global_step": 212456, "epoch": 2387, "lr": 1.1247687032217208e-05}
{"train_loss": 0.04194292053580284, "global_step": 212457, "epoch": 2387, "lr": 1.1247320696654212e-05}
{"train_loss": 0.037516288459300995, "global_step": 212458, "epoch": 2387, "lr": 1.1246954366301033e-05}
{"train_loss": 0.0402626171708107, "global_step": 212459, "epoch": 2387, "lr": 1.1246588041157741e-05}
{"train_loss": 0.06701745837926865, "global_step": 212460, "epoch": 2387, "lr": 1.1246221721224353e-05}
{"train_loss": 0.07136785984039307, "global_step": 212461, "epoch": 2387, "lr": 1.124585540650096e-05}
{"train_loss": 0.07834377139806747, "global_step": 212462, "epoch": 2387, "lr": 1.1245489096987572e-05}
{"train_loss": 0.11923448741436005, "global_step": 212463, "epoch": 2387, "lr": 1.1245122792684264e-05}
{"train_loss": 0.046000074595212936, "global_step": 212464, "epoch": 2387, "lr": 1.1244756493591068e-05}
{"train_loss": 0.05266905948519707, "global_step": 212465, "epoch": 2387, "lr": 1.1244390199708061e-05}
{"train_loss": 0.05707775801420212, "global_step": 212466, "epoch": 2387, "lr": 1.124402391103525e-05}
{"train_loss": 0.038595691323280334, "global_step": 212467, "epoch": 2387, "lr": 1.1243657627572717e-05}
{"train_loss": 0.08536360412836075, "global_step": 212468, "epoch": 2387, "lr": 1.1243291349320512e-05}
{"train_loss": 0.048001714050769806, "global_step": 212469, "epoch": 2387, "lr": 1.1242925076278655e-05}
{"train_loss": 0.041698943823575974, "global_step": 212470, "epoch": 2387, "lr": 1.1242558808447234e-05}
{"train_loss": 0.04323675110936165, "global_step": 212471, "epoch": 2387, "lr": 1.1242192545826263e-05}
{"train_loss": 0.044266872107982635, "global_step": 212472, "epoch": 2387, "lr": 1.1241826288415819e-05}
{"train_loss": 0.07482738047838211, "global_step": 212473, "epoch": 2387, "lr": 1.1241460036215928e-05}
{"train_loss": 0.08960361033678055, "global_step": 212474, "epoch": 2387, "lr": 1.1241093789226664e-05}
{"train_loss": 0.08848684281110764, "global_step": 212475, "epoch": 2387, "lr": 1.1240727547448049e-05}
{"train_loss": 0.06663908064365387, "global_step": 212476, "epoch": 2387, "lr": 1.124036131088016e-05}
{"train_loss": 0.053060200065374374, "global_step": 212477, "epoch": 2387, "lr": 1.1239995079523013e-05}
{"train_loss": 0.047538161277770996, "global_step": 212478, "epoch": 2387, "lr": 1.1239628853376694e-05}
{"train_loss": 0.0567689910531044, "global_step": 212479, "epoch": 2387, "lr": 1.1239262632441221e-05}
{"train_loss": 0.0672997459769249, "global_step": 212480, "epoch": 2387, "lr": 1.1238896416716672e-05}
{"train_loss": 0.06314309686422348, "global_step": 212481, "epoch": 2387, "lr": 1.1238530206203062e-05}
{"train_loss": 0.017365235835313797, "global_step": 212482, "epoch": 2387, "lr": 1.123816400090048e-05}
{"train_loss": 0.07025345414876938, "global_step": 212483, "epoch": 2387, "lr": 1.1237797800808936e-05}
{"train_loss": 0.09174478054046631, "global_step": 212484, "epoch": 2387, "lr": 1.1237431605928512e-05}
{"train_loss": 0.04840191453695297, "global_step": 212485, "epoch": 2387, "lr": 1.123706541625923e-05}
{"train_loss": 0.09266425669193268, "global_step": 212486, "epoch": 2387, "lr": 1.1236699231801162e-05}
{"train_loss": 0.04886455461382866, "global_step": 212487, "epoch": 2387, "lr": 1.1236333052554337e-05}
{"train_loss": 0.0610528327524662, "global_step": 212488, "epoch": 2387, "lr": 1.1235966878518828e-05}
{"train_loss": 0.05505929887294769, "global_step": 212489, "epoch": 2387, "lr": 1.123560070969467e-05}
{"train_loss": 0.05463835969567299, "global_step": 212490, "epoch": 2387, "lr": 1.1235234546081902e-05}
{"train_loss": 0.10213710367679596, "global_step": 212491, "epoch": 2387, "lr": 1.1234868387680597e-05}
{"train_loss": 0.04789189621806145, "global_step": 212492, "epoch": 2387, "lr": 1.123450223449078e-05}
{"train_loss": 0.0550064779818058, "global_step": 212493, "epoch": 2387, "lr": 1.123413608651252e-05}
{"train_loss": 0.08142687380313873, "global_step": 212494, "epoch": 2387, "lr": 1.1233769943745847e-05}
{"train_loss": 0.0552857406437397, "global_step": 212495, "epoch": 2387, "lr": 1.1233403806190834e-05}
{"train_loss": 0.027776027098298073, "global_step": 212496, "epoch": 2387, "lr": 1.1233037673847502e-05}
{"train_loss": 0.06340428441762924, "global_step": 212497, "epoch": 2387, "lr": 1.1232671546715934e-05}
{"train_loss": 0.03574074059724808, "global_step": 212498, "epoch": 2387, "lr": 1.1232305424796142e-05}
{"train_loss": 0.05928260460495949, "global_step": 212499, "epoch": 2387, "lr": 1.1231939308088207e-05}
{"train_loss": 0.05274137109518051, "global_step": 212500, "epoch": 2387, "lr": 1.1231573196592155e-05}
{"train_loss": 0.08734162151813507, "global_step": 212501, "epoch": 2387, "lr": 1.123120709030806e-05}
{"train_loss": 0.017441364005208015, "global_step": 212502, "epoch": 2387, "lr": 1.123084098923594e-05}
{"train_loss": 0.07410572469234467, "global_step": 212503, "epoch": 2387, "lr": 1.1230474893375864e-05}
{"train_loss": 0.05974184721708298, "global_step": 212504, "epoch": 2387, "lr": 1.123010880272789e-05}
{"train_loss": 0.09153979271650314, "global_step": 212505, "epoch": 2387, "lr": 1.1229742717292042e-05}
{"train_loss": 0.045036010444164276, "global_step": 212506, "epoch": 2387, "lr": 1.1229376637068395e-05}
{"train_loss": 0.059026703238487244, "global_step": 212507, "epoch": 2387, "lr": 1.1229010562056974e-05}
{"train_loss": 0.07404161989688873, "global_step": 212508, "epoch": 2387, "lr": 1.1228644492257856e-05}
{"train_loss": 0.03712213411927223, "global_step": 212509, "epoch": 2387, "lr": 1.1228278427671057e-05}
{"train_loss": 0.03268129751086235, "global_step": 212510, "epoch": 2387, "lr": 1.1227912368296655e-05}
{"train_loss": 0.05158388614654541, "global_step": 212511, "epoch": 2387, "lr": 1.1227546314134679e-05}
{"train_loss": 0.05933019146323204, "global_step": 212512, "epoch": 2387, "lr": 1.12271802651852e-05}
{"train_loss": 0.06008010730147362, "global_step": 212513, "epoch": 2387, "lr": 1.1226814221448233e-05}
{"train_loss": 0.06897471845149994, "global_step": 212514, "epoch": 2387, "lr": 1.1226448182923871e-05}
{"train_loss": 0.04068060219287872, "global_step": 212515, "epoch": 2387, "lr": 1.1226082149612122e-05}
{"train_loss": 0.0900692567229271, "global_step": 212516, "epoch": 2387, "lr": 1.1225716121513068e-05}
{"train_loss": 0.04049071669578552, "global_step": 212517, "epoch": 2387, "lr": 1.122535009862673e-05}
{"train_loss": 0.05314153805375099, "global_step": 212518, "epoch": 2387, "lr": 1.1224984080953188e-05}
{"train_loss": 0.03756817430257797, "global_step": 212519, "epoch": 2387, "lr": 1.122461806849246e-05}
{"train_loss": 0.05238610878586769, "global_step": 212520, "epoch": 2387, "lr": 1.1224252061244616e-05}
{"train_loss": 0.042328204959630966, "global_step": 212521, "epoch": 2387, "lr": 1.1223886059209693e-05}
{"train_loss": 0.10676842927932739, "global_step": 212522, "epoch": 2387, "lr": 1.1223520062387754e-05}
{"train_loss": 0.06915082037448883, "global_step": 212523, "epoch": 2387, "lr": 1.1223154070778829e-05}
{"train_loss": 0.06447872519493103, "global_step": 212524, "epoch": 2387, "lr": 1.1222788084382995e-05}
{"train_loss": 0.06003952398896217, "global_step": 212525, "epoch": 2387, "lr": 1.122242210320027e-05}
{"train_loss": 0.047540340572595596, "global_step": 212526, "epoch": 2387, "lr": 1.1222056127230729e-05}
{"train_loss": 0.07056216895580292, "global_step": 212527, "epoch": 2387, "lr": 1.1221690156474408e-05}
{"train_loss": 0.04140295460820198, "global_step": 212528, "epoch": 2387, "lr": 1.1221324190931348e-05}
{"train_loss": 0.11211949586868286, "global_step": 212529, "epoch": 2387, "lr": 1.122095823060162e-05}
{"train_loss": 0.034080397337675095, "global_step": 212530, "epoch": 2387, "lr": 1.1220592275485247e-05}
{"train_loss": 0.058637013240309246, "global_step": 212531, "epoch": 2387, "lr": 1.1220226325582312e-05, "val_loss": 8.708638191223145}
{"train_loss": 0.05851706862449646, "global_step": 212532, "epoch": 2388, "lr": 1.1219860380892827e-05}
{"train_loss": 0.053102847188711166, "global_step": 212533, "epoch": 2388, "lr": 1.1219494441416872e-05}
{"train_loss": 0.031372394412755966, "global_step": 212534, "epoch": 2388, "lr": 1.121912850715447e-05}
{"train_loss": 0.05313149094581604, "global_step": 212535, "epoch": 2388, "lr": 1.1218762578105697e-05}
{"train_loss": 0.05810987204313278, "global_step": 212536, "epoch": 2388, "lr": 1.121839665427058e-05}
{"train_loss": 0.13546298444271088, "global_step": 212537, "epoch": 2388, "lr": 1.1218030735649182e-05}
{"train_loss": 0.07545589655637741, "global_step": 212538, "epoch": 2388, "lr": 1.1217664822241542e-05}
{"train_loss": 0.08739431947469711, "global_step": 212539, "epoch": 2388, "lr": 1.1217298914047709e-05}
{"train_loss": 0.08160391449928284, "global_step": 212540, "epoch": 2388, "lr": 1.1216933011067749e-05}
{"train_loss": 0.06413422524929047, "global_step": 212541, "epoch": 2388, "lr": 1.121656711330169e-05}
{"train_loss": 0.030034521594643593, "global_step": 212542, "epoch": 2388, "lr": 1.1216201220749606e-05}
{"train_loss": 0.04414569213986397, "global_step": 212543, "epoch": 2388, "lr": 1.1215835333411512e-05}
{"train_loss": 0.07853361964225769, "global_step": 212544, "epoch": 2388, "lr": 1.1215469451287497e-05}
{"train_loss": 0.02770382910966873, "global_step": 212545, "epoch": 2388, "lr": 1.1215103574377572e-05}
{"train_loss": 0.06752064824104309, "global_step": 212546, "epoch": 2388, "lr": 1.1214737702681815e-05}
{"train_loss": 0.09532642364501953, "global_step": 212547, "epoch": 2388, "lr": 1.121437183620025e-05}
{"train_loss": 0.05803906172513962, "global_step": 212548, "epoch": 2388, "lr": 1.1214005974932956e-05}
{"train_loss": 0.030974458903074265, "global_step": 212549, "epoch": 2388, "lr": 1.1213640118879953e-05}
{"train_loss": 0.03574264794588089, "global_step": 212550, "epoch": 2388, "lr": 1.1213274268041318e-05}
{"train_loss": 0.04971238225698471, "global_step": 212551, "epoch": 2388, "lr": 1.1212908422417068e-05}
{"train_loss": 0.06243418529629707, "global_step": 212552, "epoch": 2388, "lr": 1.1212542582007285e-05}
{"train_loss": 0.03010365180671215, "global_step": 212553, "epoch": 2388, "lr": 1.1212176746811992e-05}
{"train_loss": 0.08667605370283127, "global_step": 212554, "epoch": 2388, "lr": 1.1211810916831261e-05}
{"train_loss": 0.03958740830421448, "global_step": 212555, "epoch": 2388, "lr": 1.1211445092065114e-05}
{"train_loss": 0.06157556548714638, "global_step": 212556, "epoch": 2388, "lr": 1.1211079272513631e-05}
{"train_loss": 0.08325903862714767, "global_step": 212557, "epoch": 2388, "lr": 1.121071345817683e-05}
{"train_loss": 0.0552574023604393, "global_step": 212558, "epoch": 2388, "lr": 1.1210347649054775e-05}
{"train_loss": 0.034361470490694046, "global_step": 212559, "epoch": 2388, "lr": 1.1209981845147532e-05}
{"train_loss": 0.07115574926137924, "global_step": 212560, "epoch": 2388, "lr": 1.1209616046455135e-05}
{"train_loss": 0.07110133767127991, "global_step": 212561, "epoch": 2388, "lr": 1.120925025297762e-05}
{"train_loss": 0.04661867395043373, "global_step": 212562, "epoch": 2388, "lr": 1.1208884464715058e-05}
{"train_loss": 0.03654943406581879, "global_step": 212563, "epoch": 2388, "lr": 1.120851868166748e-05}
{"train_loss": 0.088554248213768, "global_step": 212564, "epoch": 2388, "lr": 1.120815290383495e-05}
{"train_loss": 0.059915587306022644, "global_step": 212565, "epoch": 2388, "lr": 1.1207787131217517e-05}
{"train_loss": 0.04081711173057556, "global_step": 212566, "epoch": 2388, "lr": 1.1207421363815208e-05}
{"train_loss": 0.04820748046040535, "global_step": 212567, "epoch": 2388, "lr": 1.1207055601628102e-05}
{"train_loss": 0.0531277060508728, "global_step": 212568, "epoch": 2388, "lr": 1.120668984465622e-05}
{"train_loss": 0.019052566960453987, "global_step": 212569, "epoch": 2388, "lr": 1.1206324092899645e-05}
{"train_loss": 0.03740527853369713, "global_step": 212570, "epoch": 2388, "lr": 1.1205958346358387e-05}
{"train_loss": 0.03153664618730545, "global_step": 212571, "epoch": 2388, "lr": 1.1205592605032534e-05}
{"train_loss": 0.054175566881895065, "global_step": 212572, "epoch": 2388, "lr": 1.1205226868922098e-05}
{"train_loss": 0.055100660771131516, "global_step": 212573, "epoch": 2388, "lr": 1.1204861138027161e-05}
{"train_loss": 0.0745985135436058, "global_step": 212574, "epoch": 2388, "lr": 1.1204495412347743e-05}
{"train_loss": 0.03342047333717346, "global_step": 212575, "epoch": 2388, "lr": 1.1204129691883914e-05}
{"train_loss": 0.05392993614077568, "global_step": 212576, "epoch": 2388, "lr": 1.1203763976635722e-05}
{"train_loss": 0.059900231659412384, "global_step": 212577, "epoch": 2388, "lr": 1.1203398266603205e-05}
{"train_loss": 0.031266409903764725, "global_step": 212578, "epoch": 2388, "lr": 1.1203032561786425e-05}
{"train_loss": 0.056761015206575394, "global_step": 212579, "epoch": 2388, "lr": 1.1202666862185419e-05}
{"train_loss": 0.05614578723907471, "global_step": 212580, "epoch": 2388, "lr": 1.1202301167800245e-05}
{"train_loss": 0.05297726392745972, "global_step": 212581, "epoch": 2388, "lr": 1.1201935478630943e-05}
{"train_loss": 0.09683454781770706, "global_step": 212582, "epoch": 2388, "lr": 1.120156979467758e-05}
{"train_loss": 0.06618857383728027, "global_step": 212583, "epoch": 2388, "lr": 1.1201204115940179e-05}
{"train_loss": 0.04625902324914932, "global_step": 212584, "epoch": 2388, "lr": 1.1200838442418815e-05}
{"train_loss": 0.01424821000546217, "global_step": 212585, "epoch": 2388, "lr": 1.1200472774113513e-05}
{"train_loss": 0.04932951554656029, "global_step": 212586, "epoch": 2388, "lr": 1.1200107111024344e-05}
{"train_loss": 0.07758395373821259, "global_step": 212587, "epoch": 2388, "lr": 1.1199741453151342e-05}
{"train_loss": 0.053543876856565475, "global_step": 212588, "epoch": 2388, "lr": 1.1199375800494572e-05}
{"train_loss": 0.054571010172367096, "global_step": 212589, "epoch": 2388, "lr": 1.1199010153054057e-05}
{"train_loss": 0.04870585724711418, "global_step": 212590, "epoch": 2388, "lr": 1.1198644510829875e-05}
{"train_loss": 0.09778027981519699, "global_step": 212591, "epoch": 2388, "lr": 1.1198278873822054e-05}
{"train_loss": 0.09975562244653702, "global_step": 212592, "epoch": 2388, "lr": 1.1197913242030661e-05}
{"train_loss": 0.0382598377764225, "global_step": 212593, "epoch": 2388, "lr": 1.1197547615455723e-05}
{"train_loss": 0.07944079488515854, "global_step": 212594, "epoch": 2388, "lr": 1.1197181994097306e-05}
{"train_loss": 0.11352230608463287, "global_step": 212595, "epoch": 2388, "lr": 1.119681637795546e-05}
{"train_loss": 0.043145377188920975, "global_step": 212596, "epoch": 2388, "lr": 1.1196450767030225e-05}
{"train_loss": 0.09973004460334778, "global_step": 212597, "epoch": 2388, "lr": 1.1196085161321662e-05}
{"train_loss": 0.06406890600919724, "global_step": 212598, "epoch": 2388, "lr": 1.1195719560829815e-05}
{"train_loss": 0.02667025476694107, "global_step": 212599, "epoch": 2388, "lr": 1.1195353965554711e-05}
{"train_loss": 0.036088138818740845, "global_step": 212600, "epoch": 2388, "lr": 1.119498837549644e-05}
{"train_loss": 0.05027017369866371, "global_step": 212601, "epoch": 2388, "lr": 1.1194622790655013e-05}
{"train_loss": 0.05326421931385994, "global_step": 212602, "epoch": 2388, "lr": 1.1194257211030506e-05}
{"train_loss": 0.05049557238817215, "global_step": 212603, "epoch": 2388, "lr": 1.119389163662295e-05}
{"train_loss": 0.06965550035238266, "global_step": 212604, "epoch": 2388, "lr": 1.1193526067432413e-05}
{"train_loss": 0.01096538919955492, "global_step": 212605, "epoch": 2388, "lr": 1.1193160503458939e-05}
{"train_loss": 0.06625117361545563, "global_step": 212606, "epoch": 2388, "lr": 1.119279494470255e-05}
{"train_loss": 0.05153454467654228, "global_step": 212607, "epoch": 2388, "lr": 1.1192429391163334e-05}
{"train_loss": 0.036553334444761276, "global_step": 212608, "epoch": 2388, "lr": 1.1192063842841311e-05}
{"train_loss": 0.0564228892326355, "global_step": 212609, "epoch": 2388, "lr": 1.1191698299736553e-05}
{"train_loss": 0.038043852895498276, "global_step": 212610, "epoch": 2388, "lr": 1.1191332761849088e-05}
{"train_loss": 0.06698353588581085, "global_step": 212611, "epoch": 2388, "lr": 1.1190967229178972e-05}
{"train_loss": 0.09029192477464676, "global_step": 212612, "epoch": 2388, "lr": 1.1190601701726272e-05}
{"train_loss": 0.05595974996685982, "global_step": 212613, "epoch": 2388, "lr": 1.119023617949101e-05}
{"train_loss": 0.045045290142297745, "global_step": 212614, "epoch": 2388, "lr": 1.1189870662473256e-05}
{"train_loss": 0.12887141108512878, "global_step": 212615, "epoch": 2388, "lr": 1.118950515067304e-05}
{"train_loss": 0.05797477439045906, "global_step": 212616, "epoch": 2388, "lr": 1.118913964409044e-05}
{"train_loss": 0.07771417498588562, "global_step": 212617, "epoch": 2388, "lr": 1.1188774142725466e-05}
{"train_loss": 0.028202224522829056, "global_step": 212618, "epoch": 2388, "lr": 1.1188408646578207e-05}
{"train_loss": 0.07923078536987305, "global_step": 212619, "epoch": 2388, "lr": 1.118804315564868e-05}
{"train_loss": 0.05829383095914728, "global_step": 212620, "epoch": 2388, "lr": 1.1187677669936964e-05, "val_loss": 8.799421310424805}
{"train_loss": 0.04376111924648285, "global_step": 212621, "epoch": 2389, "lr": 1.1187312189443072e-05}
{"train_loss": 0.06557545065879822, "global_step": 212622, "epoch": 2389, "lr": 1.118694671416709e-05}
{"train_loss": 0.07788033038377762, "global_step": 212623, "epoch": 2389, "lr": 1.1186581244109034e-05}
{"train_loss": 0.04663563519716263, "global_step": 212624, "epoch": 2389, "lr": 1.1186215779268982e-05}
{"train_loss": 0.0738852471113205, "global_step": 212625, "epoch": 2389, "lr": 1.1185850319646957e-05}
{"train_loss": 0.0661480650305748, "global_step": 212626, "epoch": 2389, "lr": 1.118548486524304e-05}
{"train_loss": 0.06057746335864067, "global_step": 212627, "epoch": 2389, "lr": 1.1185119416057243e-05}
{"train_loss": 0.017985187470912933, "global_step": 212628, "epoch": 2389, "lr": 1.118475397208965e-05}
{"train_loss": 0.09173145145177841, "global_step": 212629, "epoch": 2389, "lr": 1.1184388533340278e-05}
{"train_loss": 0.07411989569664001, "global_step": 212630, "epoch": 2389, "lr": 1.1184023099809193e-05}
{"train_loss": 0.13653665781021118, "global_step": 212631, "epoch": 2389, "lr": 1.1183657671496457e-05}
{"train_loss": 0.05272180214524269, "global_step": 212632, "epoch": 2389, "lr": 1.1183292248402088e-05}
{"train_loss": 0.05895928665995598, "global_step": 212633, "epoch": 2389, "lr": 1.118292683052617e-05}
{"train_loss": 0.0553344190120697, "global_step": 212634, "epoch": 2389, "lr": 1.1182561417868714e-05}
{"train_loss": 0.04000204801559448, "global_step": 212635, "epoch": 2389, "lr": 1.1182196010429813e-05}
{"train_loss": 0.06716000288724899, "global_step": 212636, "epoch": 2389, "lr": 1.118183060820947e-05}
{"train_loss": 0.04608828201889992, "global_step": 212637, "epoch": 2389, "lr": 1.1181465211207776e-05}
{"train_loss": 0.07084468752145767, "global_step": 212638, "epoch": 2389, "lr": 1.1181099819424761e-05}
{"train_loss": 0.04635856673121452, "global_step": 212639, "epoch": 2389, "lr": 1.1180734432860457e-05}
{"train_loss": 0.026435352861881256, "global_step": 212640, "epoch": 2389, "lr": 1.1180369051514944e-05}
{"train_loss": 0.028487127274274826, "global_step": 212641, "epoch": 2389, "lr": 1.1180003675388246e-05}
{"train_loss": 0.03744938224554062, "global_step": 212642, "epoch": 2389, "lr": 1.117963830448044e-05}
{"train_loss": 0.053421687334775925, "global_step": 212643, "epoch": 2389, "lr": 1.1179272938791552e-05}
{"train_loss": 0.09406784921884537, "global_step": 212644, "epoch": 2389, "lr": 1.117890757832163e-05}
{"train_loss": 0.034477461129426956, "global_step": 212645, "epoch": 2389, "lr": 1.1178542223070738e-05}
{"train_loss": 0.061107419431209564, "global_step": 212646, "epoch": 2389, "lr": 1.1178176873038909e-05}
{"train_loss": 0.06702615320682526, "global_step": 212647, "epoch": 2389, "lr": 1.1177811528226212e-05}
{"train_loss": 0.08856672793626785, "global_step": 212648, "epoch": 2389, "lr": 1.1177446188632674e-05}
{"train_loss": 0.024762164801359177, "global_step": 212649, "epoch": 2389, "lr": 1.1177080854258349e-05}
{"train_loss": 0.05852032080292702, "global_step": 212650, "epoch": 2389, "lr": 1.1176715525103309e-05}
{"train_loss": 0.05138244107365608, "global_step": 212651, "epoch": 2389, "lr": 1.1176350201167579e-05}
{"train_loss": 0.06407294422388077, "global_step": 212652, "epoch": 2389, "lr": 1.1175984882451224e-05}
{"train_loss": 0.017845721915364265, "global_step": 212653, "epoch": 2389, "lr": 1.117561956895427e-05}
{"train_loss": 0.028269143775105476, "global_step": 212654, "epoch": 2389, "lr": 1.1175254260676792e-05}
{"train_loss": 0.04545995965600014, "global_step": 212655, "epoch": 2389, "lr": 1.1174888957618817e-05}
{"train_loss": 0.03360294550657272, "global_step": 212656, "epoch": 2389, "lr": 1.1174523659780423e-05}
{"train_loss": 0.06091901287436485, "global_step": 212657, "epoch": 2389, "lr": 1.1174158367161624e-05}
{"train_loss": 0.046717505902051926, "global_step": 212658, "epoch": 2389, "lr": 1.1173793079762501e-05}
{"train_loss": 0.10944327712059021, "global_step": 212659, "epoch": 2389, "lr": 1.117342779758307e-05}
{"train_loss": 0.051438651978969574, "global_step": 212660, "epoch": 2389, "lr": 1.1173062520623418e-05}
{"train_loss": 0.03527899459004402, "global_step": 212661, "epoch": 2389, "lr": 1.1172697248883556e-05}
{"train_loss": 0.06880675256252289, "global_step": 212662, "epoch": 2389, "lr": 1.117233198236357e-05}
{"train_loss": 0.057983580976724625, "global_step": 212663, "epoch": 2389, "lr": 1.117196672106347e-05}
{"train_loss": 0.09429492801427841, "global_step": 212664, "epoch": 2389, "lr": 1.1171601464983345e-05}
{"train_loss": 0.046287454664707184, "global_step": 212665, "epoch": 2389, "lr": 1.117123621412321e-05}
{"train_loss": 0.05571664869785309, "global_step": 212666, "epoch": 2389, "lr": 1.1170870968483133e-05}
{"train_loss": 0.04566699638962746, "global_step": 212667, "epoch": 2389, "lr": 1.1170505728063168e-05}
{"train_loss": 0.03507787361741066, "global_step": 212668, "epoch": 2389, "lr": 1.117014049286334e-05}
{"train_loss": 0.06122167035937309, "global_step": 212669, "epoch": 2389, "lr": 1.116977526288373e-05}
{"train_loss": 0.07714217156171799, "global_step": 212670, "epoch": 2389, "lr": 1.1169410038124356e-05}
{"train_loss": 0.03210071474313736, "global_step": 212671, "epoch": 2389, "lr": 1.11690448185853e-05}
{"train_loss": 0.04003572463989258, "global_step": 212672, "epoch": 2389, "lr": 1.1168679604266574e-05}
{"train_loss": 0.07459277659654617, "global_step": 212673, "epoch": 2389, "lr": 1.116831439516826e-05}
{"train_loss": 0.07074147462844849, "global_step": 212674, "epoch": 2389, "lr": 1.1167949191290378e-05}
{"train_loss": 0.039391931146383286, "global_step": 212675, "epoch": 2389, "lr": 1.1167583992633008e-05}
{"train_loss": 0.05857783928513527, "global_step": 212676, "epoch": 2389, "lr": 1.1167218799196184e-05}
{"train_loss": 0.10880381613969803, "global_step": 212677, "epoch": 2389, "lr": 1.1166853610979938e-05}
{"train_loss": 0.05103975161910057, "global_step": 212678, "epoch": 2389, "lr": 1.1166488427984345e-05}
{"train_loss": 0.06392493098974228, "global_step": 212679, "epoch": 2389, "lr": 1.1166123250209437e-05}
{"train_loss": 0.0488295815885067, "global_step": 212680, "epoch": 2389, "lr": 1.116575807765528e-05}
{"train_loss": 0.07682184129953384, "global_step": 212681, "epoch": 2389, "lr": 1.116539291032192e-05}
{"train_loss": 0.03854519873857498, "global_step": 212682, "epoch": 2389, "lr": 1.1165027748209377e-05}
{"train_loss": 0.03561893478035927, "global_step": 212683, "epoch": 2389, "lr": 1.1164662591317742e-05}
{"train_loss": 0.10525727272033691, "global_step": 212684, "epoch": 2389, "lr": 1.1164297439647031e-05}
{"train_loss": 0.053779326379299164, "global_step": 212685, "epoch": 2389, "lr": 1.1163932293197305e-05}
{"train_loss": 0.0603424608707428, "global_step": 212686, "epoch": 2389, "lr": 1.116356715196863e-05}
{"train_loss": 0.08104559034109116, "global_step": 212687, "epoch": 2389, "lr": 1.1163202015961027e-05}
{"train_loss": 0.1325794756412506, "global_step": 212688, "epoch": 2389, "lr": 1.1162836885174571e-05}
{"train_loss": 0.042356185615062714, "global_step": 212689, "epoch": 2389, "lr": 1.1162471759609284e-05}
{"train_loss": 0.06960313022136688, "global_step": 212690, "epoch": 2389, "lr": 1.1162106639265241e-05}
{"train_loss": 0.03821767121553421, "global_step": 212691, "epoch": 2389, "lr": 1.1161741524142466e-05}
{"train_loss": 0.025715995579957962, "global_step": 212692, "epoch": 2389, "lr": 1.1161376414241037e-05}
{"train_loss": 0.0823151022195816, "global_step": 212693, "epoch": 2389, "lr": 1.1161011309560976e-05}
{"train_loss": 0.0499374121427536, "global_step": 212694, "epoch": 2389, "lr": 1.1160646210102354e-05}
{"train_loss": 0.11038146167993546, "global_step": 212695, "epoch": 2389, "lr": 1.1160281115865195e-05}
{"train_loss": 0.09695564210414886, "global_step": 212696, "epoch": 2389, "lr": 1.1159916026849582e-05}
{"train_loss": 0.07659230381250381, "global_step": 212697, "epoch": 2389, "lr": 1.115955094305553e-05}
{"train_loss": 0.04636813700199127, "global_step": 212698, "epoch": 2389, "lr": 1.1159185864483112e-05}
{"train_loss": 0.06138335540890694, "global_step": 212699, "epoch": 2389, "lr": 1.1158820791132358e-05}
{"train_loss": 0.02621602453291416, "global_step": 212700, "epoch": 2389, "lr": 1.1158455723003341e-05}
{"train_loss": 0.06616036593914032, "global_step": 212701, "epoch": 2389, "lr": 1.1158090660096076e-05}
{"train_loss": 0.06804574280977249, "global_step": 212702, "epoch": 2389, "lr": 1.115772560241064e-05}
{"train_loss": 0.05387714505195618, "global_step": 212703, "epoch": 2389, "lr": 1.1157360549947088e-05}
{"train_loss": 0.05358818173408508, "global_step": 212704, "epoch": 2389, "lr": 1.1156995502705436e-05}
{"train_loss": 0.03774970769882202, "global_step": 212705, "epoch": 2389, "lr": 1.1156630460685768e-05}
{"train_loss": 0.0597037710249424, "global_step": 212706, "epoch": 2389, "lr": 1.1156265423888107e-05}
{"train_loss": 0.10715362429618835, "global_step": 212707, "epoch": 2389, "lr": 1.115590039231253e-05}
{"train_loss": 0.06986816227436066, "global_step": 212708, "epoch": 2389, "lr": 1.1155535365959046e-05}
{"train_loss": 0.05984130842883265, "global_step": 212709, "epoch": 2389, "lr": 1.1155170344827747e-05, "val_loss": 8.757317543029785}
{"train_loss": 0.09993620216846466, "global_step": 212710, "epoch": 2390, "lr": 1.1154805328918644e-05}
{"train_loss": 0.08763713389635086, "global_step": 212711, "epoch": 2390, "lr": 1.1154440318231819e-05}
{"train_loss": 0.07172682136297226, "global_step": 212712, "epoch": 2390, "lr": 1.1154075312767292e-05}
{"train_loss": 0.05908939987421036, "global_step": 212713, "epoch": 2390, "lr": 1.1153710312525139e-05}
{"train_loss": 0.016149034723639488, "global_step": 212714, "epoch": 2390, "lr": 1.1153345317505398e-05}
{"train_loss": 0.07026764005422592, "global_step": 212715, "epoch": 2390, "lr": 1.1152980327708106e-05}
{"train_loss": 0.06399337202310562, "global_step": 212716, "epoch": 2390, "lr": 1.1152615343133333e-05}
{"train_loss": 0.06598657369613647, "global_step": 212717, "epoch": 2390, "lr": 1.1152250363781103e-05}
{"train_loss": 0.07566994428634644, "global_step": 212718, "epoch": 2390, "lr": 1.115188538965149e-05}
{"train_loss": 0.040198810398578644, "global_step": 212719, "epoch": 2390, "lr": 1.115152042074452e-05}
{"train_loss": 0.03411073610186577, "global_step": 212720, "epoch": 2390, "lr": 1.1151155457060275e-05}
{"train_loss": 0.05236165225505829, "global_step": 212721, "epoch": 2390, "lr": 1.1150790498598763e-05}
{"train_loss": 0.08856614679098129, "global_step": 212722, "epoch": 2390, "lr": 1.1150425545360072e-05}
{"train_loss": 0.06352483481168747, "global_step": 212723, "epoch": 2390, "lr": 1.1150060597344213e-05}
{"train_loss": 0.04389170929789543, "global_step": 212724, "epoch": 2390, "lr": 1.1149695654551273e-05}
{"train_loss": 0.04455036669969559, "global_step": 212725, "epoch": 2390, "lr": 1.1149330716981271e-05}
{"train_loss": 0.07912906259298325, "global_step": 212726, "epoch": 2390, "lr": 1.1148965784634275e-05}
{"train_loss": 0.06635330617427826, "global_step": 212727, "epoch": 2390, "lr": 1.1148600857510317e-05}
{"train_loss": 0.04039634019136429, "global_step": 212728, "epoch": 2390, "lr": 1.1148235935609475e-05}
{"train_loss": 0.07932639122009277, "global_step": 212729, "epoch": 2390, "lr": 1.1147871018931755e-05}
{"train_loss": 0.08600947260856628, "global_step": 212730, "epoch": 2390, "lr": 1.1147506107477252e-05}
{"train_loss": 0.04380732774734497, "global_step": 212731, "epoch": 2390, "lr": 1.1147141201245976e-05}
{"train_loss": 0.12395444512367249, "global_step": 212732, "epoch": 2390, "lr": 1.114677630023801e-05}
{"train_loss": 0.07969881594181061, "global_step": 212733, "epoch": 2390, "lr": 1.1146411404453373e-05}
{"train_loss": 0.02017497643828392, "global_step": 212734, "epoch": 2390, "lr": 1.1146046513892133e-05}
{"train_loss": 0.08284619450569153, "global_step": 212735, "epoch": 2390, "lr": 1.1145681628554328e-05}
{"train_loss": 0.06303519010543823, "global_step": 212736, "epoch": 2390, "lr": 1.1145316748440027e-05}
{"train_loss": 0.07733042538166046, "global_step": 212737, "epoch": 2390, "lr": 1.1144951873549243e-05}
{"train_loss": 0.024519342929124832, "global_step": 212738, "epoch": 2390, "lr": 1.1144587003882051e-05}
{"train_loss": 0.050355151295661926, "global_step": 212739, "epoch": 2390, "lr": 1.1144222139438509e-05}
{"train_loss": 0.05006173998117447, "global_step": 212740, "epoch": 2390, "lr": 1.1143857280218644e-05}
{"train_loss": 0.07233382016420364, "global_step": 212741, "epoch": 2390, "lr": 1.1143492426222524e-05}
{"train_loss": 0.04391256719827652, "global_step": 212742, "epoch": 2390, "lr": 1.1143127577450169e-05}
{"train_loss": 0.05822738632559776, "global_step": 212743, "epoch": 2390, "lr": 1.1142762733901669e-05}
{"train_loss": 0.03778041526675224, "global_step": 212744, "epoch": 2390, "lr": 1.1142397895577034e-05}
{"train_loss": 0.06991036236286163, "global_step": 212745, "epoch": 2390, "lr": 1.1142033062476342e-05}
{"train_loss": 0.072611965239048, "global_step": 212746, "epoch": 2390, "lr": 1.1141668234599616e-05}
{"train_loss": 0.05250716581940651, "global_step": 212747, "epoch": 2390, "lr": 1.1141303411946935e-05}
{"train_loss": 0.13779863715171814, "global_step": 212748, "epoch": 2390, "lr": 1.1140938594518318e-05}
{"train_loss": 0.058524154126644135, "global_step": 212749, "epoch": 2390, "lr": 1.114057378231384e-05}
{"train_loss": 0.0690324455499649, "global_step": 212750, "epoch": 2390, "lr": 1.1140208975333527e-05}
{"train_loss": 0.050727374851703644, "global_step": 212751, "epoch": 2390, "lr": 1.1139844173577453e-05}
{"train_loss": 0.04088246449828148, "global_step": 212752, "epoch": 2390, "lr": 1.1139479377045635e-05}
{"train_loss": 0.06365616619586945, "global_step": 212753, "epoch": 2390, "lr": 1.1139114585738158e-05}
{"train_loss": 0.0420086532831192, "global_step": 212754, "epoch": 2390, "lr": 1.1138749799655057e-05}
{"train_loss": 0.07312054187059402, "global_step": 212755, "epoch": 2390, "lr": 1.1138385018796355e-05}
{"train_loss": 0.05960127338767052, "global_step": 212756, "epoch": 2390, "lr": 1.1138020243162145e-05}
{"train_loss": 0.09810647368431091, "global_step": 212757, "epoch": 2390, "lr": 1.113765547275244e-05}
{"train_loss": 0.029721690341830254, "global_step": 212758, "epoch": 2390, "lr": 1.113729070756731e-05}
{"train_loss": 0.030752331018447876, "global_step": 212759, "epoch": 2390, "lr": 1.1136925947606792e-05}
{"train_loss": 0.04312676191329956, "global_step": 212760, "epoch": 2390, "lr": 1.1136561192870948e-05}
{"train_loss": 0.0703980103135109, "global_step": 212761, "epoch": 2390, "lr": 1.1136196443359809e-05}
{"train_loss": 0.09329739958047867, "global_step": 212762, "epoch": 2390, "lr": 1.1135831699073451e-05}
{"train_loss": 0.0770254135131836, "global_step": 212763, "epoch": 2390, "lr": 1.1135466960011891e-05}
{"train_loss": 0.06936068087816238, "global_step": 212764, "epoch": 2390, "lr": 1.1135102226175208e-05}
{"train_loss": 0.05783407762646675, "global_step": 212765, "epoch": 2390, "lr": 1.113473749756342e-05}
{"train_loss": 0.036970168352127075, "global_step": 212766, "epoch": 2390, "lr": 1.113437277417661e-05}
{"train_loss": 0.04997057095170021, "global_step": 212767, "epoch": 2390, "lr": 1.1134008056014794e-05}
{"train_loss": 0.049229491502046585, "global_step": 212768, "epoch": 2390, "lr": 1.1133643343078055e-05}
{"train_loss": 0.054195091128349304, "global_step": 212769, "epoch": 2390, "lr": 1.1133278635366407e-05}
{"train_loss": 0.05481290817260742, "global_step": 212770, "epoch": 2390, "lr": 1.113291393287993e-05}
{"train_loss": 0.0464121513068676, "global_step": 212771, "epoch": 2390, "lr": 1.113254923561864e-05}
{"train_loss": 0.036310743540525436, "global_step": 212772, "epoch": 2390, "lr": 1.1132184543582624e-05}
{"train_loss": 0.06712844967842102, "global_step": 212773, "epoch": 2390, "lr": 1.1131819856771896e-05}
{"train_loss": 0.03402542695403099, "global_step": 212774, "epoch": 2390, "lr": 1.113145517518654e-05}
{"train_loss": 0.09743081778287888, "global_step": 212775, "epoch": 2390, "lr": 1.1131090498826568e-05}
{"train_loss": 0.10598991066217422, "global_step": 212776, "epoch": 2390, "lr": 1.1130725827692046e-05}
{"train_loss": 0.06146139279007912, "global_step": 212777, "epoch": 2390, "lr": 1.1130361161783043e-05}
{"train_loss": 0.02658483013510704, "global_step": 212778, "epoch": 2390, "lr": 1.1129996501099571e-05}
{"train_loss": 0.04997869208455086, "global_step": 212779, "epoch": 2390, "lr": 1.1129631845641713e-05}
{"train_loss": 0.08043146878480911, "global_step": 212780, "epoch": 2390, "lr": 1.1129267195409488e-05}
{"train_loss": 0.039169009774923325, "global_step": 212781, "epoch": 2390, "lr": 1.1128902550402975e-05}
{"train_loss": 0.06315705925226212, "global_step": 212782, "epoch": 2390, "lr": 1.1128537910622189e-05}
{"train_loss": 0.05478272959589958, "global_step": 212783, "epoch": 2390, "lr": 1.1128173276067217e-05}
{"train_loss": 0.07091190665960312, "global_step": 212784, "epoch": 2390, "lr": 1.1127808646738069e-05}
{"train_loss": 0.07954992353916168, "global_step": 212785, "epoch": 2390, "lr": 1.1127444022634831e-05}
{"train_loss": 0.06658992916345596, "global_step": 212786, "epoch": 2390, "lr": 1.1127079403757517e-05}
{"train_loss": 0.05942274630069733, "global_step": 212787, "epoch": 2390, "lr": 1.1126714790106207e-05}
{"train_loss": 0.05557584762573242, "global_step": 212788, "epoch": 2390, "lr": 1.1126350181680928e-05}
{"train_loss": 0.06519010663032532, "global_step": 212789, "epoch": 2390, "lr": 1.1125985578481746e-05}
{"train_loss": 0.09536072611808777, "global_step": 212790, "epoch": 2390, "lr": 1.1125620980508689e-05}
{"train_loss": 0.0527055524289608, "global_step": 212791, "epoch": 2390, "lr": 1.1125256387761834e-05}
{"train_loss": 0.040107183158397675, "global_step": 212792, "epoch": 2390, "lr": 1.112489180024121e-05}
{"train_loss": 0.04872267693281174, "global_step": 212793, "epoch": 2390, "lr": 1.1124527217946862e-05}
{"train_loss": 0.02953207679092884, "global_step": 212794, "epoch": 2390, "lr": 1.112416264087886e-05}
{"train_loss": 0.0683048889040947, "global_step": 212795, "epoch": 2390, "lr": 1.1123798069037222e-05}
{"train_loss": 0.05665440484881401, "global_step": 212796, "epoch": 2390, "lr": 1.1123433502422037e-05}
{"train_loss": 0.03921952843666077, "global_step": 212797, "epoch": 2390, "lr": 1.1123068941033316e-05}
{"train_loss": 0.06070368069360096, "global_step": 212798, "epoch": 2390, "lr": 1.1122704384871136e-05, "val_loss": 8.749163627624512, "train_action_mse_error": 21.46822738647461}
{"train_loss": 0.03744428604841232, "global_step": 212799, "epoch": 2391, "lr": 1.112233983393552e-05}
{"train_loss": 0.09692329168319702, "global_step": 212800, "epoch": 2391, "lr": 1.112197528822655e-05}
{"train_loss": 0.04340306669473648, "global_step": 212801, "epoch": 2391, "lr": 1.1121610747744238e-05}
{"train_loss": 0.11044520884752274, "global_step": 212802, "epoch": 2391, "lr": 1.1121246212488668e-05}
{"train_loss": 0.03346799314022064, "global_step": 212803, "epoch": 2391, "lr": 1.1120881682459855e-05}
{"train_loss": 0.07299711555242538, "global_step": 212804, "epoch": 2391, "lr": 1.1120517157657885e-05}
{"train_loss": 0.023114658892154694, "global_step": 212805, "epoch": 2391, "lr": 1.112015263808277e-05}
{"train_loss": 0.0291578471660614, "global_step": 212806, "epoch": 2391, "lr": 1.1119788123734592e-05}
{"train_loss": 0.08613529801368713, "global_step": 212807, "epoch": 2391, "lr": 1.1119423614613373e-05}
{"train_loss": 0.06091487407684326, "global_step": 212808, "epoch": 2391, "lr": 1.1119059110719183e-05}
{"train_loss": 0.033505551517009735, "global_step": 212809, "epoch": 2391, "lr": 1.111869461205205e-05}
{"train_loss": 0.060881730169057846, "global_step": 212810, "epoch": 2391, "lr": 1.1118330118612047e-05}
{"train_loss": 0.052059732377529144, "global_step": 212811, "epoch": 2391, "lr": 1.1117965630399203e-05}
{"train_loss": 0.04830082878470421, "global_step": 212812, "epoch": 2391, "lr": 1.1117601147413565e-05}
{"train_loss": 0.08939937502145767, "global_step": 212813, "epoch": 2391, "lr": 1.1117236669655213e-05}
{"train_loss": 0.09157254546880722, "global_step": 212814, "epoch": 2391, "lr": 1.1116872197124156e-05}
{"train_loss": 0.07834097743034363, "global_step": 212815, "epoch": 2391, "lr": 1.111650772982048e-05}
{"train_loss": 0.06783750653266907, "global_step": 212816, "epoch": 2391, "lr": 1.1116143267744195e-05}
{"train_loss": 0.020480645820498466, "global_step": 212817, "epoch": 2391, "lr": 1.1115778810895389e-05}
{"train_loss": 0.03811512887477875, "global_step": 212818, "epoch": 2391, "lr": 1.1115414359274078e-05}
{"train_loss": 0.05847010016441345, "global_step": 212819, "epoch": 2391, "lr": 1.1115049912880343e-05}
{"train_loss": 0.023142913356423378, "global_step": 212820, "epoch": 2391, "lr": 1.1114685471714198e-05}
{"train_loss": 0.01953730173408985, "global_step": 212821, "epoch": 2391, "lr": 1.1114321035775721e-05}
{"train_loss": 0.025521092116832733, "global_step": 212822, "epoch": 2391, "lr": 1.111395660506494e-05}
{"train_loss": 0.04261935129761696, "global_step": 212823, "epoch": 2391, "lr": 1.1113592179581928e-05}
{"train_loss": 0.040093421936035156, "global_step": 212824, "epoch": 2391, "lr": 1.11132277593267e-05}
{"train_loss": 0.032091107219457626, "global_step": 212825, "epoch": 2391, "lr": 1.1112863344299346e-05}
{"train_loss": 0.0588068962097168, "global_step": 212826, "epoch": 2391, "lr": 1.1112498934499877e-05}
{"train_loss": 0.08881508558988571, "global_step": 212827, "epoch": 2391, "lr": 1.111213452992837e-05}
{"train_loss": 0.022342704236507416, "global_step": 212828, "epoch": 2391, "lr": 1.1111770130584847e-05}
{"train_loss": 0.049183204770088196, "global_step": 212829, "epoch": 2391, "lr": 1.1111405736469393e-05}
{"train_loss": 0.07591712474822998, "global_step": 212830, "epoch": 2391, "lr": 1.111104134758203e-05}
{"train_loss": 0.06971338391304016, "global_step": 212831, "epoch": 2391, "lr": 1.1110676963922806e-05}
{"train_loss": 0.10811756551265717, "global_step": 212832, "epoch": 2391, "lr": 1.1110312585491783e-05}
{"train_loss": 0.06778600811958313, "global_step": 212833, "epoch": 2391, "lr": 1.1109948212288996e-05}
{"train_loss": 0.09934516996145248, "global_step": 212834, "epoch": 2391, "lr": 1.110958384431452e-05}
{"train_loss": 0.060038063675165176, "global_step": 212835, "epoch": 2391, "lr": 1.1109219481568367e-05}
{"train_loss": 0.0676339715719223, "global_step": 212836, "epoch": 2391, "lr": 1.1108855124050621e-05}
{"train_loss": 0.08351507782936096, "global_step": 212837, "epoch": 2391, "lr": 1.1108490771761298e-05}
{"train_loss": 0.05819307640194893, "global_step": 212838, "epoch": 2391, "lr": 1.1108126424700482e-05}
{"train_loss": 0.050472769886255264, "global_step": 212839, "epoch": 2391, "lr": 1.110776208286819e-05}
{"train_loss": 0.020016228780150414, "global_step": 212840, "epoch": 2391, "lr": 1.1107397746264504e-05}
{"train_loss": 0.04865250363945961, "global_step": 212841, "epoch": 2391, "lr": 1.1107033414889434e-05}
{"train_loss": 0.02749921940267086, "global_step": 212842, "epoch": 2391, "lr": 1.1106669088743066e-05}
{"train_loss": 0.055459752678871155, "global_step": 212843, "epoch": 2391, "lr": 1.1106304767825416e-05}
{"train_loss": 0.07893268018960953, "global_step": 212844, "epoch": 2391, "lr": 1.1105940452136566e-05}
{"train_loss": 0.06537600606679916, "global_step": 212845, "epoch": 2391, "lr": 1.1105576141676532e-05}
{"train_loss": 0.0549989752471447, "global_step": 212846, "epoch": 2391, "lr": 1.110521183644539e-05}
{"train_loss": 0.07681453227996826, "global_step": 212847, "epoch": 2391, "lr": 1.110484753644317e-05}
{"train_loss": 0.050495121628046036, "global_step": 212848, "epoch": 2391, "lr": 1.1104483241669927e-05}
{"train_loss": 0.0849907398223877, "global_step": 212849, "epoch": 2391, "lr": 1.1104118952125725e-05}
{"train_loss": 0.032863326370716095, "global_step": 212850, "epoch": 2391, "lr": 1.110375466781059e-05}
{"train_loss": 0.044185567647218704, "global_step": 212851, "epoch": 2391, "lr": 1.1103390388724594e-05}
{"train_loss": 0.052120011299848557, "global_step": 212852, "epoch": 2391, "lr": 1.1103026114867754e-05}
{"train_loss": 0.044111452996730804, "global_step": 212853, "epoch": 2391, "lr": 1.110266184624016e-05}
{"train_loss": 0.060790080577135086, "global_step": 212854, "epoch": 2391, "lr": 1.110229758284182e-05}
{"train_loss": 0.017375530675053596, "global_step": 212855, "epoch": 2391, "lr": 1.1101933324672814e-05}
{"train_loss": 0.06263595074415207, "global_step": 212856, "epoch": 2391, "lr": 1.110156907173317e-05}
{"train_loss": 0.038867056369781494, "global_step": 212857, "epoch": 2391, "lr": 1.1101204824022959e-05}
{"train_loss": 0.07338591665029526, "global_step": 212858, "epoch": 2391, "lr": 1.11008405815422e-05}
{"train_loss": 0.04177254065871239, "global_step": 212859, "epoch": 2391, "lr": 1.1100476344290977e-05}
{"train_loss": 0.03563380986452103, "global_step": 212860, "epoch": 2391, "lr": 1.1100112112269301e-05}
{"train_loss": 0.05833886191248894, "global_step": 212861, "epoch": 2391, "lr": 1.1099747885477262e-05}
{"train_loss": 0.0465974323451519, "global_step": 212862, "epoch": 2391, "lr": 1.1099383663914869e-05}
{"train_loss": 0.04136163368821144, "global_step": 212863, "epoch": 2391, "lr": 1.1099019447582205e-05}
{"train_loss": 0.0580606646835804, "global_step": 212864, "epoch": 2391, "lr": 1.1098655236479288e-05}
{"train_loss": 0.03652683272957802, "global_step": 212865, "epoch": 2391, "lr": 1.1098291030606201e-05}
{"train_loss": 0.10934454947710037, "global_step": 212866, "epoch": 2391, "lr": 1.1097926829962957e-05}
{"train_loss": 0.06888864934444427, "global_step": 212867, "epoch": 2391, "lr": 1.1097562634549635e-05}
{"train_loss": 0.081944040954113, "global_step": 212868, "epoch": 2391, "lr": 1.1097198444366263e-05}
{"train_loss": 0.03207799792289734, "global_step": 212869, "epoch": 2391, "lr": 1.1096834259412908e-05}
{"train_loss": 0.014652141369879246, "global_step": 212870, "epoch": 2391, "lr": 1.1096470079689608e-05}
{"train_loss": 0.07742435485124588, "global_step": 212871, "epoch": 2391, "lr": 1.1096105905196402e-05}
{"train_loss": 0.034274131059646606, "global_step": 212872, "epoch": 2391, "lr": 1.1095741735933368e-05}
{"train_loss": 0.04859279468655586, "global_step": 212873, "epoch": 2391, "lr": 1.1095377571900518e-05}
{"train_loss": 0.04355509206652641, "global_step": 212874, "epoch": 2391, "lr": 1.1095013413097938e-05}
{"train_loss": 0.05764453858137131, "global_step": 212875, "epoch": 2391, "lr": 1.1094649259525641e-05}
{"train_loss": 0.058977894484996796, "global_step": 212876, "epoch": 2391, "lr": 1.109428511118371e-05}
{"train_loss": 0.0822405070066452, "global_step": 212877, "epoch": 2391, "lr": 1.1093920968072169e-05}
{"train_loss": 0.05901654064655304, "global_step": 212878, "epoch": 2391, "lr": 1.1093556830191088e-05}
{"train_loss": 0.052776094526052475, "global_step": 212879, "epoch": 2391, "lr": 1.1093192697540482e-05}
{"train_loss": 0.07386627048254013, "global_step": 212880, "epoch": 2391, "lr": 1.1092828570120445e-05}
{"train_loss": 0.0887213870882988, "global_step": 212881, "epoch": 2391, "lr": 1.109246444793099e-05}
{"train_loss": 0.08872091770172119, "global_step": 212882, "epoch": 2391, "lr": 1.1092100330972188e-05}
{"train_loss": 0.06645999103784561, "global_step": 212883, "epoch": 2391, "lr": 1.1091736219244064e-05}
{"train_loss": 0.017070062458515167, "global_step": 212884, "epoch": 2391, "lr": 1.1091372112746684e-05}
{"train_loss": 0.06533791124820709, "global_step": 212885, "epoch": 2391, "lr": 1.1091008011480108e-05}
{"train_loss": 0.05786183848977089, "global_step": 212886, "epoch": 2391, "lr": 1.109064391544436e-05}
{"train_loss": 0.05668441236563278, "global_step": 212887, "epoch": 2391, "lr": 1.109027982463951e-05, "val_loss": 8.82643985748291}
{"train_loss": 0.06856773048639297, "global_step": 212888, "epoch": 2392, "lr": 1.1089915739065587e-05}
{"train_loss": 0.04498035088181496, "global_step": 212889, "epoch": 2392, "lr": 1.1089551658722668e-05}
{"train_loss": 0.0314740426838398, "global_step": 212890, "epoch": 2392, "lr": 1.1089187583610766e-05}
{"train_loss": 0.06238958239555359, "global_step": 212891, "epoch": 2392, "lr": 1.1088823513729962e-05}
{"train_loss": 0.1018085703253746, "global_step": 212892, "epoch": 2392, "lr": 1.1088459449080279e-05}
{"train_loss": 0.06908301264047623, "global_step": 212893, "epoch": 2392, "lr": 1.108809538966179e-05}
{"train_loss": 0.025096731260418892, "global_step": 212894, "epoch": 2392, "lr": 1.1087731335474521e-05}
{"train_loss": 0.0774066150188446, "global_step": 212895, "epoch": 2392, "lr": 1.1087367286518547e-05}
{"train_loss": 0.03372710198163986, "global_step": 212896, "epoch": 2392, "lr": 1.1087003242793887e-05}
{"train_loss": 0.04499439522624016, "global_step": 212897, "epoch": 2392, "lr": 1.108663920430062e-05}
{"train_loss": 0.02413114719092846, "global_step": 212898, "epoch": 2392, "lr": 1.1086275171038762e-05}
{"train_loss": 0.05261444300413132, "global_step": 212899, "epoch": 2392, "lr": 1.1085911143008398e-05}
{"train_loss": 0.06251655519008636, "global_step": 212900, "epoch": 2392, "lr": 1.1085547120209545e-05}
{"train_loss": 0.044865213334560394, "global_step": 212901, "epoch": 2392, "lr": 1.1085183102642277e-05}
{"train_loss": 0.04032767191529274, "global_step": 212902, "epoch": 2392, "lr": 1.1084819090306619e-05}
{"train_loss": 0.07817550748586655, "global_step": 212903, "epoch": 2392, "lr": 1.108445508320265e-05}
{"train_loss": 0.021579895168542862, "global_step": 212904, "epoch": 2392, "lr": 1.1084091081330384e-05}
{"train_loss": 0.08634408563375473, "global_step": 212905, "epoch": 2392, "lr": 1.1083727084689904e-05}
{"train_loss": 0.09724686294794083, "global_step": 212906, "epoch": 2392, "lr": 1.108336309328123e-05}
{"train_loss": 0.05722326412796974, "global_step": 212907, "epoch": 2392, "lr": 1.108299910710443e-05}
{"train_loss": 0.029056083410978317, "global_step": 212908, "epoch": 2392, "lr": 1.1082635126159552e-05}
{"train_loss": 0.01717674545943737, "global_step": 212909, "epoch": 2392, "lr": 1.1082271150446627e-05}
{"train_loss": 0.05099824070930481, "global_step": 212910, "epoch": 2392, "lr": 1.1081907179965728e-05}
{"train_loss": 0.07615786045789719, "global_step": 212911, "epoch": 2392, "lr": 1.1081543214716877e-05}
{"train_loss": 0.057932525873184204, "global_step": 212912, "epoch": 2392, "lr": 1.1081179254700152e-05}
{"train_loss": 0.0834907740354538, "global_step": 212913, "epoch": 2392, "lr": 1.1080815299915576e-05}
{"train_loss": 0.037905286997556686, "global_step": 212914, "epoch": 2392, "lr": 1.1080451350363225e-05}
{"train_loss": 0.0382239855825901, "global_step": 212915, "epoch": 2392, "lr": 1.1080087406043116e-05}
{"train_loss": 0.05795637518167496, "global_step": 212916, "epoch": 2392, "lr": 1.1079723466955333e-05}
{"train_loss": 0.047405797988176346, "global_step": 212917, "epoch": 2392, "lr": 1.1079359533099893e-05}
{"train_loss": 0.018001999706029892, "global_step": 212918, "epoch": 2392, "lr": 1.1078995604476867e-05}
{"train_loss": 0.08160914480686188, "global_step": 212919, "epoch": 2392, "lr": 1.1078631681086281e-05}
{"train_loss": 0.0873090922832489, "global_step": 212920, "epoch": 2392, "lr": 1.1078267762928207e-05}
{"train_loss": 0.039568640291690826, "global_step": 212921, "epoch": 2392, "lr": 1.1077903850002691e-05}
{"train_loss": 0.059119753539562225, "global_step": 212922, "epoch": 2392, "lr": 1.1077539942309766e-05}
{"train_loss": 0.054883502423763275, "global_step": 212923, "epoch": 2392, "lr": 1.1077176039849508e-05}
{"train_loss": 0.04372546449303627, "global_step": 212924, "epoch": 2392, "lr": 1.1076812142621934e-05}
{"train_loss": 0.03347949683666229, "global_step": 212925, "epoch": 2392, "lr": 1.107644825062712e-05}
{"train_loss": 0.07816173881292343, "global_step": 212926, "epoch": 2392, "lr": 1.1076084363865092e-05}
{"train_loss": 0.030001377686858177, "global_step": 212927, "epoch": 2392, "lr": 1.1075720482335921e-05}
{"train_loss": 0.028164055198431015, "global_step": 212928, "epoch": 2392, "lr": 1.1075356606039633e-05}
{"train_loss": 0.06013339385390282, "global_step": 212929, "epoch": 2392, "lr": 1.1074992734976297e-05}
{"train_loss": 0.11839456856250763, "global_step": 212930, "epoch": 2392, "lr": 1.1074628869145947e-05}
{"train_loss": 0.06870803982019424, "global_step": 212931, "epoch": 2392, "lr": 1.107426500854865e-05}
{"train_loss": 0.032342176884412766, "global_step": 212932, "epoch": 2392, "lr": 1.107390115318443e-05}
{"train_loss": 0.04356741905212402, "global_step": 212933, "epoch": 2392, "lr": 1.1073537303053367e-05}
{"train_loss": 0.08877286314964294, "global_step": 212934, "epoch": 2392, "lr": 1.1073173458155473e-05}
{"train_loss": 0.07015436887741089, "global_step": 212935, "epoch": 2392, "lr": 1.1072809618490832e-05}
{"train_loss": 0.06381300836801529, "global_step": 212936, "epoch": 2392, "lr": 1.1072445784059465e-05}
{"train_loss": 0.05152835696935654, "global_step": 212937, "epoch": 2392, "lr": 1.1072081954861446e-05}
{"train_loss": 0.07361486554145813, "global_step": 212938, "epoch": 2392, "lr": 1.1071718130896796e-05}
{"train_loss": 0.07527749240398407, "global_step": 212939, "epoch": 2392, "lr": 1.1071354312165578e-05}
{"train_loss": 0.027287280187010765, "global_step": 212940, "epoch": 2392, "lr": 1.107099049866786e-05}
{"train_loss": 0.0411243662238121, "global_step": 212941, "epoch": 2392, "lr": 1.1070626690403668e-05}
{"train_loss": 0.08997203409671783, "global_step": 212942, "epoch": 2392, "lr": 1.1070262887373045e-05}
{"train_loss": 0.014615096151828766, "global_step": 212943, "epoch": 2392, "lr": 1.106989908957607e-05}
{"train_loss": 0.03432938829064369, "global_step": 212944, "epoch": 2392, "lr": 1.106953529701275e-05}
{"train_loss": 0.03759235143661499, "global_step": 212945, "epoch": 2392, "lr": 1.1069171509683173e-05}
{"train_loss": 0.07749149203300476, "global_step": 212946, "epoch": 2392, "lr": 1.106880772758736e-05}
{"train_loss": 0.03001217357814312, "global_step": 212947, "epoch": 2392, "lr": 1.1068443950725382e-05}
{"train_loss": 0.06139059737324715, "global_step": 212948, "epoch": 2392, "lr": 1.1068080179097279e-05}
{"train_loss": 0.06189871206879616, "global_step": 212949, "epoch": 2392, "lr": 1.1067716412703083e-05}
{"train_loss": 0.07332557439804077, "global_step": 212950, "epoch": 2392, "lr": 1.1067352651542873e-05}
{"train_loss": 0.04402429237961769, "global_step": 212951, "epoch": 2392, "lr": 1.1066988895616665e-05}
{"train_loss": 0.016799909994006157, "global_step": 212952, "epoch": 2392, "lr": 1.1066625144924542e-05}
{"train_loss": 0.07105589658021927, "global_step": 212953, "epoch": 2392, "lr": 1.1066261399466526e-05}
{"train_loss": 0.07780960947275162, "global_step": 212954, "epoch": 2392, "lr": 1.106589765924269e-05}
{"train_loss": 0.053626466542482376, "global_step": 212955, "epoch": 2392, "lr": 1.1065533924253053e-05}
{"train_loss": 0.09101627767086029, "global_step": 212956, "epoch": 2392, "lr": 1.1065170194497682e-05}
{"train_loss": 0.08340702205896378, "global_step": 212957, "epoch": 2392, "lr": 1.1064806469976635e-05}
{"train_loss": 0.08441723138093948, "global_step": 212958, "epoch": 2392, "lr": 1.1064442750689941e-05}
{"train_loss": 0.0528888925909996, "global_step": 212959, "epoch": 2392, "lr": 1.106407903663767e-05}
{"train_loss": 0.06261590868234634, "global_step": 212960, "epoch": 2392, "lr": 1.1063715327819846e-05}
{"train_loss": 0.03376227244734764, "global_step": 212961, "epoch": 2392, "lr": 1.1063351624236545e-05}
{"train_loss": 0.07999000698328018, "global_step": 212962, "epoch": 2392, "lr": 1.1062987925887786e-05}
{"train_loss": 0.037352073937654495, "global_step": 212963, "epoch": 2392, "lr": 1.106262423277365e-05}
{"train_loss": 0.06271939724683762, "global_step": 212964, "epoch": 2392, "lr": 1.1062260544894148e-05}
{"train_loss": 0.04343384504318237, "global_step": 212965, "epoch": 2392, "lr": 1.1061896862249371e-05}
{"train_loss": 0.06871134042739868, "global_step": 212966, "epoch": 2392, "lr": 1.1061533184839334e-05}
{"train_loss": 0.020793750882148743, "global_step": 212967, "epoch": 2392, "lr": 1.1061169512664115e-05}
{"train_loss": 0.06616926193237305, "global_step": 212968, "epoch": 2392, "lr": 1.1060805845723727e-05}
{"train_loss": 0.04965437203645706, "global_step": 212969, "epoch": 2392, "lr": 1.106044218401826e-05}
{"train_loss": 0.02889176644384861, "global_step": 212970, "epoch": 2392, "lr": 1.1060078527547724e-05}
{"train_loss": 0.05045011639595032, "global_step": 212971, "epoch": 2392, "lr": 1.10597148763122e-05}
{"train_loss": 0.03430674597620964, "global_step": 212972, "epoch": 2392, "lr": 1.105935123031171e-05}
{"train_loss": 0.052681077271699905, "global_step": 212973, "epoch": 2392, "lr": 1.1058987589546327e-05}
{"train_loss": 0.06659922748804092, "global_step": 212974, "epoch": 2392, "lr": 1.1058623954016072e-05}
{"train_loss": 0.027813738211989403, "global_step": 212975, "epoch": 2392, "lr": 1.105826032372102e-05}
{"train_loss": 0.055094026259324524, "global_step": 212976, "epoch": 2392, "lr": 1.1057896698661219e-05, "val_loss": 8.832311630249023}
{"train_loss": 0.04077005013823509, "global_step": 212977, "epoch": 2393, "lr": 1.1057533078836691e-05}
{"train_loss": 0.026378151029348373, "global_step": 212978, "epoch": 2393, "lr": 1.1057169464247525e-05}
{"train_loss": 0.08215658366680145, "global_step": 212979, "epoch": 2393, "lr": 1.1056805854893726e-05}
{"train_loss": 0.09062570333480835, "global_step": 212980, "epoch": 2393, "lr": 1.1056442250775384e-05}
{"train_loss": 0.03598439320921898, "global_step": 212981, "epoch": 2393, "lr": 1.1056078651892527e-05}
{"train_loss": 0.03873604163527489, "global_step": 212982, "epoch": 2393, "lr": 1.1055715058245192e-05}
{"train_loss": 0.030346740037202835, "global_step": 212983, "epoch": 2393, "lr": 1.1055351469833453e-05}
{"train_loss": 0.04934466630220413, "global_step": 212984, "epoch": 2393, "lr": 1.1054987886657336e-05}
{"train_loss": 0.08028886467218399, "global_step": 212985, "epoch": 2393, "lr": 1.1054624308716916e-05}
{"train_loss": 0.046769723296165466, "global_step": 212986, "epoch": 2393, "lr": 1.1054260736012223e-05}
{"train_loss": 0.08572670817375183, "global_step": 212987, "epoch": 2393, "lr": 1.1053897168543303e-05}
{"train_loss": 0.05830182880163193, "global_step": 212988, "epoch": 2393, "lr": 1.1053533606310218e-05}
{"train_loss": 0.04891545698046684, "global_step": 212989, "epoch": 2393, "lr": 1.1053170049313e-05}
{"train_loss": 0.08527097851037979, "global_step": 212990, "epoch": 2393, "lr": 1.1052806497551727e-05}
{"train_loss": 0.028655461966991425, "global_step": 212991, "epoch": 2393, "lr": 1.1052442951026409e-05}
{"train_loss": 0.023607071489095688, "global_step": 212992, "epoch": 2393, "lr": 1.1052079409737121e-05}
{"train_loss": 0.07282419502735138, "global_step": 212993, "epoch": 2393, "lr": 1.1051715873683916e-05}
{"train_loss": 0.030201541259884834, "global_step": 212994, "epoch": 2393, "lr": 1.1051352342866817e-05}
{"train_loss": 0.04885803163051605, "global_step": 212995, "epoch": 2393, "lr": 1.1050988817285906e-05}
{"train_loss": 0.09996812045574188, "global_step": 212996, "epoch": 2393, "lr": 1.1050625296941204e-05}
{"train_loss": 0.04145921766757965, "global_step": 212997, "epoch": 2393, "lr": 1.1050261781832783e-05}
{"train_loss": 0.06499087065458298, "global_step": 212998, "epoch": 2393, "lr": 1.1049898271960663e-05}
{"train_loss": 0.026261840015649796, "global_step": 212999, "epoch": 2393, "lr": 1.104953476732492e-05}
{"train_loss": 0.045291490852832794, "global_step": 213000, "epoch": 2393, "lr": 1.1049171267925584e-05}
{"train_loss": 0.06522238254547119, "global_step": 213001, "epoch": 2393, "lr": 1.1048807773762726e-05}
{"train_loss": 0.05998140573501587, "global_step": 213002, "epoch": 2393, "lr": 1.1048444284836368e-05}
{"train_loss": 0.06160736456513405, "global_step": 213003, "epoch": 2393, "lr": 1.1048080801146583e-05}
{"train_loss": 0.040136802941560745, "global_step": 213004, "epoch": 2393, "lr": 1.1047717322693396e-05}
{"train_loss": 0.03420166298747063, "global_step": 213005, "epoch": 2393, "lr": 1.1047353849476883e-05}
{"train_loss": 0.1142011508345604, "global_step": 213006, "epoch": 2393, "lr": 1.1046990381497064e-05}
{"train_loss": 0.06552894413471222, "global_step": 213007, "epoch": 2393, "lr": 1.1046626918754017e-05}
{"train_loss": 0.04458688572049141, "global_step": 213008, "epoch": 2393, "lr": 1.1046263461247758e-05}
{"train_loss": 0.049915608018636703, "global_step": 213009, "epoch": 2393, "lr": 1.1045900008978372e-05}
{"train_loss": 0.03363966569304466, "global_step": 213010, "epoch": 2393, "lr": 1.104553656194588e-05}
{"train_loss": 0.08628761023283005, "global_step": 213011, "epoch": 2393, "lr": 1.1045173120150331e-05}
{"train_loss": 0.06193152815103531, "global_step": 213012, "epoch": 2393, "lr": 1.1044809683591805e-05}
{"train_loss": 0.04259248077869415, "global_step": 213013, "epoch": 2393, "lr": 1.1044446252270313e-05}
{"train_loss": 0.039457980543375015, "global_step": 213014, "epoch": 2393, "lr": 1.1044082826185936e-05}
{"train_loss": 0.0914948359131813, "global_step": 213015, "epoch": 2393, "lr": 1.1043719405338694e-05}
{"train_loss": 0.06574825942516327, "global_step": 213016, "epoch": 2393, "lr": 1.104335598972866e-05}
{"train_loss": 0.03496209904551506, "global_step": 213017, "epoch": 2393, "lr": 1.1042992579355859e-05}
{"train_loss": 0.10443057119846344, "global_step": 213018, "epoch": 2393, "lr": 1.104262917422037e-05}
{"train_loss": 0.03841181844472885, "global_step": 213019, "epoch": 2393, "lr": 1.104226577432222e-05}
{"train_loss": 0.06606722623109818, "global_step": 213020, "epoch": 2393, "lr": 1.104190237966145e-05}
{"train_loss": 0.08294732868671417, "global_step": 213021, "epoch": 2393, "lr": 1.1041538990238138e-05}
{"train_loss": 0.04694013297557831, "global_step": 213022, "epoch": 2393, "lr": 1.1041175606052301e-05}
{"train_loss": 0.03770153224468231, "global_step": 213023, "epoch": 2393, "lr": 1.1040812227104013e-05}
{"train_loss": 0.05657172575592995, "global_step": 213024, "epoch": 2393, "lr": 1.1040448853393314e-05}
{"train_loss": 0.046790704131126404, "global_step": 213025, "epoch": 2393, "lr": 1.1040085484920244e-05}
{"train_loss": 0.06510467827320099, "global_step": 213026, "epoch": 2393, "lr": 1.103972212168487e-05}
{"train_loss": 0.06722152233123779, "global_step": 213027, "epoch": 2393, "lr": 1.1039358763687212e-05}
{"train_loss": 0.07239767909049988, "global_step": 213028, "epoch": 2393, "lr": 1.1038995410927355e-05}
{"train_loss": 0.021558938547968864, "global_step": 213029, "epoch": 2393, "lr": 1.103863206340532e-05}
{"train_loss": 0.0504780150949955, "global_step": 213030, "epoch": 2393, "lr": 1.1038268721121164e-05}
{"train_loss": 0.03837822005152702, "global_step": 213031, "epoch": 2393, "lr": 1.1037905384074949e-05}
{"train_loss": 0.05822380632162094, "global_step": 213032, "epoch": 2393, "lr": 1.1037542052266702e-05}
{"train_loss": 0.03415980190038681, "global_step": 213033, "epoch": 2393, "lr": 1.1037178725696495e-05}
{"train_loss": 0.08133934438228607, "global_step": 213034, "epoch": 2393, "lr": 1.1036815404364347e-05}
{"train_loss": 0.05605066567659378, "global_step": 213035, "epoch": 2393, "lr": 1.1036452088270344e-05}
{"train_loss": 0.058013688772916794, "global_step": 213036, "epoch": 2393, "lr": 1.10360887774145e-05}
{"train_loss": 0.03455265983939171, "global_step": 213037, "epoch": 2393, "lr": 1.1035725471796888e-05}
{"train_loss": 0.04110786318778992, "global_step": 213038, "epoch": 2393, "lr": 1.1035362171417541e-05}
{"train_loss": 0.040704403072595596, "global_step": 213039, "epoch": 2393, "lr": 1.1034998876276526e-05}
{"train_loss": 0.06908980756998062, "global_step": 213040, "epoch": 2393, "lr": 1.1034635586373865e-05}
{"train_loss": 0.07814320176839828, "global_step": 213041, "epoch": 2393, "lr": 1.1034272301709636e-05}
{"train_loss": 0.07613449543714523, "global_step": 213042, "epoch": 2393, "lr": 1.1033909022283862e-05}
{"train_loss": 0.08919329196214676, "global_step": 213043, "epoch": 2393, "lr": 1.1033545748096612e-05}
{"train_loss": 0.06063109636306763, "global_step": 213044, "epoch": 2393, "lr": 1.103318247914792e-05}
{"train_loss": 0.04281478375196457, "global_step": 213045, "epoch": 2393, "lr": 1.1032819215437855e-05}
{"train_loss": 0.03676620125770569, "global_step": 213046, "epoch": 2393, "lr": 1.1032455956966436e-05}
{"train_loss": 0.05438968166708946, "global_step": 213047, "epoch": 2393, "lr": 1.1032092703733731e-05}
{"train_loss": 0.08400560170412064, "global_step": 213048, "epoch": 2393, "lr": 1.1031729455739798e-05}
{"train_loss": 0.06466919183731079, "global_step": 213049, "epoch": 2393, "lr": 1.1031366212984662e-05}
{"train_loss": 0.049533188343048096, "global_step": 213050, "epoch": 2393, "lr": 1.1031002975468396e-05}
{"train_loss": 0.06255649030208588, "global_step": 213051, "epoch": 2393, "lr": 1.1030639743191023e-05}
{"train_loss": 0.0299846064299345, "global_step": 213052, "epoch": 2393, "lr": 1.1030276516152621e-05}
{"train_loss": 0.03616062551736832, "global_step": 213053, "epoch": 2393, "lr": 1.1029913294353205e-05}
{"train_loss": 0.021666845306754112, "global_step": 213054, "epoch": 2393, "lr": 1.1029550077792861e-05}
{"train_loss": 0.04330870136618614, "global_step": 213055, "epoch": 2393, "lr": 1.1029186866471602e-05}
{"train_loss": 0.09833786636590958, "global_step": 213056, "epoch": 2393, "lr": 1.1028823660389508e-05}
{"train_loss": 0.0704408586025238, "global_step": 213057, "epoch": 2393, "lr": 1.102846045954662e-05}
{"train_loss": 0.07243701815605164, "global_step": 213058, "epoch": 2393, "lr": 1.102809726394296e-05}
{"train_loss": 0.08854718506336212, "global_step": 213059, "epoch": 2393, "lr": 1.102773407357861e-05}
{"train_loss": 0.04024910926818848, "global_step": 213060, "epoch": 2393, "lr": 1.1027370888453597e-05}
{"train_loss": 0.05509980767965317, "global_step": 213061, "epoch": 2393, "lr": 1.1027007708567987e-05}
{"train_loss": 0.027781780809164047, "global_step": 213062, "epoch": 2393, "lr": 1.1026644533921815e-05}
{"train_loss": 0.08281832933425903, "global_step": 213063, "epoch": 2393, "lr": 1.1026281364515145e-05}
{"train_loss": 0.062360506504774094, "global_step": 213064, "epoch": 2393, "lr": 1.1025918200348012e-05}
{"train_loss": 0.05629773994677522, "global_step": 213065, "epoch": 2393, "lr": 1.102555504142046e-05, "val_loss": 8.884384155273438}
{"train_loss": 0.01830296777188778, "global_step": 213066, "epoch": 2394, "lr": 1.102519188773255e-05}
{"train_loss": 0.06242727115750313, "global_step": 213067, "epoch": 2394, "lr": 1.1024828739284337e-05}
{"train_loss": 0.057162877172231674, "global_step": 213068, "epoch": 2394, "lr": 1.102446559607585e-05}
{"train_loss": 0.06057410687208176, "global_step": 213069, "epoch": 2394, "lr": 1.102410245810716e-05}
{"train_loss": 0.06470373272895813, "global_step": 213070, "epoch": 2394, "lr": 1.102373932537829e-05}
{"train_loss": 0.08009012788534164, "global_step": 213071, "epoch": 2394, "lr": 1.1023376197889324e-05}
{"train_loss": 0.022714490070939064, "global_step": 213072, "epoch": 2394, "lr": 1.102301307564027e-05}
{"train_loss": 0.08987784385681152, "global_step": 213073, "epoch": 2394, "lr": 1.102264995863121e-05}
{"train_loss": 0.13348528742790222, "global_step": 213074, "epoch": 2394, "lr": 1.1022286846862168e-05}
{"train_loss": 0.08491892367601395, "global_step": 213075, "epoch": 2394, "lr": 1.1021923740333218e-05}
{"train_loss": 0.06859086453914642, "global_step": 213076, "epoch": 2394, "lr": 1.1021560639044382e-05}
{"train_loss": 0.07159588485956192, "global_step": 213077, "epoch": 2394, "lr": 1.1021197542995732e-05}
{"train_loss": 0.05461261793971062, "global_step": 213078, "epoch": 2394, "lr": 1.1020834452187295e-05}
{"train_loss": 0.05392852798104286, "global_step": 213079, "epoch": 2394, "lr": 1.1020471366619151e-05}
{"train_loss": 0.08295353502035141, "global_step": 213080, "epoch": 2394, "lr": 1.1020108286291309e-05}
{"train_loss": 0.08521848171949387, "global_step": 213081, "epoch": 2394, "lr": 1.1019745211203852e-05}
{"train_loss": 0.060054536908864975, "global_step": 213082, "epoch": 2394, "lr": 1.1019382141356804e-05}
{"train_loss": 0.08376564085483551, "global_step": 213083, "epoch": 2394, "lr": 1.1019019076750226e-05}
{"train_loss": 0.021816615015268326, "global_step": 213084, "epoch": 2394, "lr": 1.1018656017384177e-05}
{"train_loss": 0.0491509810090065, "global_step": 213085, "epoch": 2394, "lr": 1.1018292963258686e-05}
{"train_loss": 0.06801919639110565, "global_step": 213086, "epoch": 2394, "lr": 1.1017929914373821e-05}
{"train_loss": 0.0354011096060276, "global_step": 213087, "epoch": 2394, "lr": 1.1017566870729602e-05}
{"train_loss": 0.03013620898127556, "global_step": 213088, "epoch": 2394, "lr": 1.1017203832326118e-05}
{"train_loss": 0.04007844254374504, "global_step": 213089, "epoch": 2394, "lr": 1.1016840799163375e-05}
{"train_loss": 0.06831110268831253, "global_step": 213090, "epoch": 2394, "lr": 1.1016477771241456e-05}
{"train_loss": 0.07486977428197861, "global_step": 213091, "epoch": 2394, "lr": 1.1016114748560386e-05}
{"train_loss": 0.09818938374519348, "global_step": 213092, "epoch": 2394, "lr": 1.1015751731120239e-05}
{"train_loss": 0.07773171365261078, "global_step": 213093, "epoch": 2394, "lr": 1.1015388718921032e-05}
{"train_loss": 0.053114935755729675, "global_step": 213094, "epoch": 2394, "lr": 1.1015025711962851e-05}
{"train_loss": 0.06191904842853546, "global_step": 213095, "epoch": 2394, "lr": 1.1014662710245704e-05}
{"train_loss": 0.06159405782818794, "global_step": 213096, "epoch": 2394, "lr": 1.1014299713769677e-05}
{"train_loss": 0.05478396639227867, "global_step": 213097, "epoch": 2394, "lr": 1.1013936722534801e-05}
{"train_loss": 0.04273223131895065, "global_step": 213098, "epoch": 2394, "lr": 1.1013573736541111e-05}
{"train_loss": 0.03538117557764053, "global_step": 213099, "epoch": 2394, "lr": 1.1013210755788688e-05}
{"train_loss": 0.056691259145736694, "global_step": 213100, "epoch": 2394, "lr": 1.1012847780277551e-05}
{"train_loss": 0.04420071095228195, "global_step": 213101, "epoch": 2394, "lr": 1.1012484810007773e-05}
{"train_loss": 0.04470377787947655, "global_step": 213102, "epoch": 2394, "lr": 1.1012121844979373e-05}
{"train_loss": 0.047937776893377304, "global_step": 213103, "epoch": 2394, "lr": 1.1011758885192436e-05}
{"train_loss": 0.05596978962421417, "global_step": 213104, "epoch": 2394, "lr": 1.1011395930646984e-05}
{"train_loss": 0.06016898155212402, "global_step": 213105, "epoch": 2394, "lr": 1.1011032981343078e-05}
{"train_loss": 0.05415726453065872, "global_step": 213106, "epoch": 2394, "lr": 1.1010670037280757e-05}
{"train_loss": 0.05443842336535454, "global_step": 213107, "epoch": 2394, "lr": 1.1010307098460088e-05}
{"train_loss": 0.035938333719968796, "global_step": 213108, "epoch": 2394, "lr": 1.1009944164881092e-05}
{"train_loss": 0.07065030932426453, "global_step": 213109, "epoch": 2394, "lr": 1.1009581236543854e-05}
{"train_loss": 0.09005557745695114, "global_step": 213110, "epoch": 2394, "lr": 1.1009218313448383e-05}
{"train_loss": 0.03642355278134346, "global_step": 213111, "epoch": 2394, "lr": 1.1008855395594763e-05}
{"train_loss": 0.04309990629553795, "global_step": 213112, "epoch": 2394, "lr": 1.1008492482983012e-05}
{"train_loss": 0.015078967437148094, "global_step": 213113, "epoch": 2394, "lr": 1.1008129575613208e-05}
{"train_loss": 0.072813980281353, "global_step": 213114, "epoch": 2394, "lr": 1.1007766673485375e-05}
{"train_loss": 0.04096752032637596, "global_step": 213115, "epoch": 2394, "lr": 1.1007403776599583e-05}
{"train_loss": 0.048591405153274536, "global_step": 213116, "epoch": 2394, "lr": 1.100704088495586e-05}
{"train_loss": 0.041608333587646484, "global_step": 213117, "epoch": 2394, "lr": 1.1006677998554276e-05}
{"train_loss": 0.03426550701260567, "global_step": 213118, "epoch": 2394, "lr": 1.1006315117394856e-05}
{"train_loss": 0.027755131945014, "global_step": 213119, "epoch": 2394, "lr": 1.1005952241477658e-05}
{"train_loss": 0.04438086971640587, "global_step": 213120, "epoch": 2394, "lr": 1.1005589370802755e-05}
{"train_loss": 0.08285076916217804, "global_step": 213121, "epoch": 2394, "lr": 1.1005226505370153e-05}
{"train_loss": 0.05792776495218277, "global_step": 213122, "epoch": 2394, "lr": 1.1004863645179942e-05}
{"train_loss": 0.06038990616798401, "global_step": 213123, "epoch": 2394, "lr": 1.1004500790232142e-05}
{"train_loss": 0.05641421675682068, "global_step": 213124, "epoch": 2394, "lr": 1.100413794052682e-05}
{"train_loss": 0.061580218374729156, "global_step": 213125, "epoch": 2394, "lr": 1.1003775096064e-05}
{"train_loss": 0.05068468302488327, "global_step": 213126, "epoch": 2394, "lr": 1.1003412256843764e-05}
{"train_loss": 0.055326323956251144, "global_step": 213127, "epoch": 2394, "lr": 1.1003049422866136e-05}
{"train_loss": 0.06235329806804657, "global_step": 213128, "epoch": 2394, "lr": 1.1002686594131178e-05}
{"train_loss": 0.033842407166957855, "global_step": 213129, "epoch": 2394, "lr": 1.1002323770638923e-05}
{"train_loss": 0.021799227222800255, "global_step": 213130, "epoch": 2394, "lr": 1.1001960952389446e-05}
{"train_loss": 0.023067209869623184, "global_step": 213131, "epoch": 2394, "lr": 1.1001598139382762e-05}
{"train_loss": 0.07016930729150772, "global_step": 213132, "epoch": 2394, "lr": 1.1001235331618953e-05}
{"train_loss": 0.05296602100133896, "global_step": 213133, "epoch": 2394, "lr": 1.1000872529098038e-05}
{"train_loss": 0.05357086658477783, "global_step": 213134, "epoch": 2394, "lr": 1.1000509731820097e-05}
{"train_loss": 0.05023794621229172, "global_step": 213135, "epoch": 2394, "lr": 1.1000146939785161e-05}
{"train_loss": 0.07192172110080719, "global_step": 213136, "epoch": 2394, "lr": 1.099978415299327e-05}
{"train_loss": 0.08187046647071838, "global_step": 213137, "epoch": 2394, "lr": 1.099942137144449e-05}
{"train_loss": 0.041194211691617966, "global_step": 213138, "epoch": 2394, "lr": 1.099905859513885e-05}
{"train_loss": 0.05408598110079765, "global_step": 213139, "epoch": 2394, "lr": 1.0998695824076426e-05}
{"train_loss": 0.039793916046619415, "global_step": 213140, "epoch": 2394, "lr": 1.0998333058257237e-05}
{"train_loss": 0.04536521062254906, "global_step": 213141, "epoch": 2394, "lr": 1.0997970297681364e-05}
{"train_loss": 0.04416206106543541, "global_step": 213142, "epoch": 2394, "lr": 1.0997607542348825e-05}
{"train_loss": 0.06469666957855225, "global_step": 213143, "epoch": 2394, "lr": 1.0997244792259693e-05}
{"train_loss": 0.07455594837665558, "global_step": 213144, "epoch": 2394, "lr": 1.0996882047413993e-05}
{"train_loss": 0.08331973850727081, "global_step": 213145, "epoch": 2394, "lr": 1.0996519307811798e-05}
{"train_loss": 0.057236943393945694, "global_step": 213146, "epoch": 2394, "lr": 1.0996156573453136e-05}
{"train_loss": 0.048759575933218, "global_step": 213147, "epoch": 2394, "lr": 1.0995793844338076e-05}
{"train_loss": 0.060710977762937546, "global_step": 213148, "epoch": 2394, "lr": 1.0995431120466643e-05}
{"train_loss": 0.017559785395860672, "global_step": 213149, "epoch": 2394, "lr": 1.0995068401838915e-05}
{"train_loss": 0.03371476009488106, "global_step": 213150, "epoch": 2394, "lr": 1.0994705688454909e-05}
{"train_loss": 0.03126443922519684, "global_step": 213151, "epoch": 2394, "lr": 1.0994342980314704e-05}
{"train_loss": 0.09382552653551102, "global_step": 213152, "epoch": 2394, "lr": 1.099398027741832e-05}
{"train_loss": 0.026201467961072922, "global_step": 213153, "epoch": 2394, "lr": 1.0993617579765835e-05}
{"train_loss": 0.056030162788006696, "global_step": 213154, "epoch": 2394, "lr": 1.0993254887357268e-05, "val_loss": 8.816628456115723}
{"train_loss": 0.045802533626556396, "global_step": 213155, "epoch": 2395, "lr": 1.0992892200192696e-05}
{"train_loss": 0.06331654638051987, "global_step": 213156, "epoch": 2395, "lr": 1.0992529518272138e-05}
{"train_loss": 0.03616400435566902, "global_step": 213157, "epoch": 2395, "lr": 1.0992166841595663e-05}
{"train_loss": 0.09115003049373627, "global_step": 213158, "epoch": 2395, "lr": 1.0991804170163328e-05}
{"train_loss": 0.04528174549341202, "global_step": 213159, "epoch": 2395, "lr": 1.0991441503975152e-05}
{"train_loss": 0.04338213801383972, "global_step": 213160, "epoch": 2395, "lr": 1.0991078843031222e-05}
{"train_loss": 0.03634489327669144, "global_step": 213161, "epoch": 2395, "lr": 1.0990716187331545e-05}
{"train_loss": 0.026391664519906044, "global_step": 213162, "epoch": 2395, "lr": 1.0990353536876208e-05}
{"train_loss": 0.0685180202126503, "global_step": 213163, "epoch": 2395, "lr": 1.098999089166523e-05}
{"train_loss": 0.033854614943265915, "global_step": 213164, "epoch": 2395, "lr": 1.0989628251698685e-05}
{"train_loss": 0.080592580139637, "global_step": 213165, "epoch": 2395, "lr": 1.0989265616976596e-05}
{"train_loss": 0.06849143654108047, "global_step": 213166, "epoch": 2395, "lr": 1.098890298749904e-05}
{"train_loss": 0.0353848934173584, "global_step": 213167, "epoch": 2395, "lr": 1.0988540363266032e-05}
{"train_loss": 0.051460977643728256, "global_step": 213168, "epoch": 2395, "lr": 1.0988177744277656e-05}
{"train_loss": 0.09881655871868134, "global_step": 213169, "epoch": 2395, "lr": 1.098781513053393e-05}
{"train_loss": 0.07797524333000183, "global_step": 213170, "epoch": 2395, "lr": 1.098745252203493e-05}
{"train_loss": 0.06015457585453987, "global_step": 213171, "epoch": 2395, "lr": 1.098708991878068e-05}
{"train_loss": 0.052106283605098724, "global_step": 213172, "epoch": 2395, "lr": 1.098672732077125e-05}
{"train_loss": 0.0978902131319046, "global_step": 213173, "epoch": 2395, "lr": 1.0986364728006682e-05}
{"train_loss": 0.03445643186569214, "global_step": 213174, "epoch": 2395, "lr": 1.0986002140487006e-05}
{"train_loss": 0.07485134154558182, "global_step": 213175, "epoch": 2395, "lr": 1.0985639558212302e-05}
{"train_loss": 0.10762886703014374, "global_step": 213176, "epoch": 2395, "lr": 1.098527698118259e-05}
{"train_loss": 0.06713379174470901, "global_step": 213177, "epoch": 2395, "lr": 1.0984914409397944e-05}
{"train_loss": 0.07102304697036743, "global_step": 213178, "epoch": 2395, "lr": 1.0984551842858387e-05}
{"train_loss": 0.08338510245084763, "global_step": 213179, "epoch": 2395, "lr": 1.0984189281563995e-05}
{"train_loss": 0.11434338241815567, "global_step": 213180, "epoch": 2395, "lr": 1.098382672551479e-05}
{"train_loss": 0.0552181638777256, "global_step": 213181, "epoch": 2395, "lr": 1.0983464174710851e-05}
{"train_loss": 0.06458670645952225, "global_step": 213182, "epoch": 2395, "lr": 1.0983101629152188e-05}
{"train_loss": 0.05524042993783951, "global_step": 213183, "epoch": 2395, "lr": 1.098273908883889e-05}
{"train_loss": 0.033169519156217575, "global_step": 213184, "epoch": 2395, "lr": 1.0982376553770974e-05}
{"train_loss": 0.053868427872657776, "global_step": 213185, "epoch": 2395, "lr": 1.0982014023948517e-05}
{"train_loss": 0.05021819844841957, "global_step": 213186, "epoch": 2395, "lr": 1.0981651499371536e-05}
{"train_loss": 0.04923433065414429, "global_step": 213187, "epoch": 2395, "lr": 1.098128898004011e-05}
{"train_loss": 0.05534787103533745, "global_step": 213188, "epoch": 2395, "lr": 1.098092646595426e-05}
{"train_loss": 0.0861581563949585, "global_step": 213189, "epoch": 2395, "lr": 1.0980563957114066e-05}
{"train_loss": 0.052658990025520325, "global_step": 213190, "epoch": 2395, "lr": 1.0980201453519546e-05}
{"train_loss": 0.02781766653060913, "global_step": 213191, "epoch": 2395, "lr": 1.0979838955170774e-05}
{"train_loss": 0.039200715720653534, "global_step": 213192, "epoch": 2395, "lr": 1.0979476462067772e-05}
{"train_loss": 0.05809573084115982, "global_step": 213193, "epoch": 2395, "lr": 1.097911397421061e-05}
{"train_loss": 0.13551630079746246, "global_step": 213194, "epoch": 2395, "lr": 1.0978751491599337e-05}
{"train_loss": 0.05756109952926636, "global_step": 213195, "epoch": 2395, "lr": 1.097838901423398e-05}
{"train_loss": 0.08932331204414368, "global_step": 213196, "epoch": 2395, "lr": 1.0978026542114622e-05}
{"train_loss": 0.05613604933023453, "global_step": 213197, "epoch": 2395, "lr": 1.097766407524128e-05}
{"train_loss": 0.0575980506837368, "global_step": 213198, "epoch": 2395, "lr": 1.0977301613614027e-05}
{"train_loss": 0.08502782881259918, "global_step": 213199, "epoch": 2395, "lr": 1.0976939157232884e-05}
{"train_loss": 0.09814745932817459, "global_step": 213200, "epoch": 2395, "lr": 1.0976576706097936e-05}
{"train_loss": 0.02785574272274971, "global_step": 213201, "epoch": 2395, "lr": 1.0976214260209189e-05}
{"train_loss": 0.08711767941713333, "global_step": 213202, "epoch": 2395, "lr": 1.0975851819566735e-05}
{"train_loss": 0.06851846724748611, "global_step": 213203, "epoch": 2395, "lr": 1.0975489384170584e-05}
{"train_loss": 0.09218885749578476, "global_step": 213204, "epoch": 2395, "lr": 1.097512695402082e-05}
{"train_loss": 0.0703718289732933, "global_step": 213205, "epoch": 2395, "lr": 1.0974764529117459e-05}
{"train_loss": 0.043571583926677704, "global_step": 213206, "epoch": 2395, "lr": 1.097440210946058e-05}
{"train_loss": 0.03633536398410797, "global_step": 213207, "epoch": 2395, "lr": 1.0974039695050204e-05}
{"train_loss": 0.04680637642741203, "global_step": 213208, "epoch": 2395, "lr": 1.0973677285886408e-05}
{"train_loss": 0.04357745498418808, "global_step": 213209, "epoch": 2395, "lr": 1.0973314881969204e-05}
{"train_loss": 0.02826291136443615, "global_step": 213210, "epoch": 2395, "lr": 1.0972952483298682e-05}
{"train_loss": 0.03718983381986618, "global_step": 213211, "epoch": 2395, "lr": 1.0972590089874856e-05}
{"train_loss": 0.062493376433849335, "global_step": 213212, "epoch": 2395, "lr": 1.0972227701697806e-05}
{"train_loss": 0.08776672184467316, "global_step": 213213, "epoch": 2395, "lr": 1.097186531876756e-05}
{"train_loss": 0.042492691427469254, "global_step": 213214, "epoch": 2395, "lr": 1.097150294108416e-05}
{"train_loss": 0.029368076473474503, "global_step": 213215, "epoch": 2395, "lr": 1.0971140568647675e-05}
{"train_loss": 0.08012591302394867, "global_step": 213216, "epoch": 2395, "lr": 1.0970778201458136e-05}
{"train_loss": 0.0409323014318943, "global_step": 213217, "epoch": 2395, "lr": 1.0970415839515613e-05}
{"train_loss": 0.05338221788406372, "global_step": 213218, "epoch": 2395, "lr": 1.0970053482820125e-05}
{"train_loss": 0.030816903337836266, "global_step": 213219, "epoch": 2395, "lr": 1.0969691131371757e-05}
{"train_loss": 0.04267625883221626, "global_step": 213220, "epoch": 2395, "lr": 1.0969328785170518e-05}
{"train_loss": 0.06035659834742546, "global_step": 213221, "epoch": 2395, "lr": 1.0968966444216493e-05}
{"train_loss": 0.04407283663749695, "global_step": 213222, "epoch": 2395, "lr": 1.0968604108509706e-05}
{"train_loss": 0.04464954882860184, "global_step": 213223, "epoch": 2395, "lr": 1.096824177805022e-05}
{"train_loss": 0.04769156128168106, "global_step": 213224, "epoch": 2395, "lr": 1.096787945283807e-05}
{"train_loss": 0.07342508435249329, "global_step": 213225, "epoch": 2395, "lr": 1.0967517132873328e-05}
{"train_loss": 0.041029371321201324, "global_step": 213226, "epoch": 2395, "lr": 1.0967154818156012e-05}
{"train_loss": 0.06413616240024567, "global_step": 213227, "epoch": 2395, "lr": 1.0966792508686197e-05}
{"train_loss": 0.07224655151367188, "global_step": 213228, "epoch": 2395, "lr": 1.0966430204463907e-05}
{"train_loss": 0.10954843461513519, "global_step": 213229, "epoch": 2395, "lr": 1.0966067905489208e-05}
{"train_loss": 0.10249356925487518, "global_step": 213230, "epoch": 2395, "lr": 1.0965705611762156e-05}
{"train_loss": 0.03214188292622566, "global_step": 213231, "epoch": 2395, "lr": 1.0965343323282778e-05}
{"train_loss": 0.06903812289237976, "global_step": 213232, "epoch": 2395, "lr": 1.0964981040051147e-05}
{"train_loss": 0.08589780330657959, "global_step": 213233, "epoch": 2395, "lr": 1.0964618762067285e-05}
{"train_loss": 0.052987661212682724, "global_step": 213234, "epoch": 2395, "lr": 1.096425648933127e-05}
{"train_loss": 0.04086494818329811, "global_step": 213235, "epoch": 2395, "lr": 1.0963894221843118e-05}
{"train_loss": 0.02493429370224476, "global_step": 213236, "epoch": 2395, "lr": 1.0963531959602907e-05}
{"train_loss": 0.048239532858133316, "global_step": 213237, "epoch": 2395, "lr": 1.0963169702610666e-05}
{"train_loss": 0.05380459874868393, "global_step": 213238, "epoch": 2395, "lr": 1.0962807450866458e-05}
{"train_loss": 0.02419048361480236, "global_step": 213239, "epoch": 2395, "lr": 1.0962445204370314e-05}
{"train_loss": 0.026949908584356308, "global_step": 213240, "epoch": 2395, "lr": 1.0962082963122311e-05}
{"train_loss": 0.08926661312580109, "global_step": 213241, "epoch": 2395, "lr": 1.096172072712246e-05}
{"train_loss": 0.04502223804593086, "global_step": 213242, "epoch": 2395, "lr": 1.0961358496370845e-05}
{"train_loss": 0.05927918694411101, "global_step": 213243, "epoch": 2395, "lr": 1.0960996270867485e-05, "val_loss": 8.877765655517578, "train_action_mse_error": 7.718162536621094}
{"train_loss": 0.07578118145465851, "global_step": 213244, "epoch": 2396, "lr": 1.0960634050612461e-05}
{"train_loss": 0.019670972600579262, "global_step": 213245, "epoch": 2396, "lr": 1.096027183560579e-05}
{"train_loss": 0.07148542255163193, "global_step": 213246, "epoch": 2396, "lr": 1.0959909625847542e-05}
{"train_loss": 0.054635606706142426, "global_step": 213247, "epoch": 2396, "lr": 1.0959547421337752e-05}
{"train_loss": 0.03470081835985184, "global_step": 213248, "epoch": 2396, "lr": 1.0959185222076485e-05}
{"train_loss": 0.038347721099853516, "global_step": 213249, "epoch": 2396, "lr": 1.0958823028063764e-05}
{"train_loss": 0.07514630258083344, "global_step": 213250, "epoch": 2396, "lr": 1.0958460839299673e-05}
{"train_loss": 0.0829656794667244, "global_step": 213251, "epoch": 2396, "lr": 1.0958098655784238e-05}
{"train_loss": 0.045833516865968704, "global_step": 213252, "epoch": 2396, "lr": 1.0957736477517494e-05}
{"train_loss": 0.0571165457367897, "global_step": 213253, "epoch": 2396, "lr": 1.0957374304499524e-05}
{"train_loss": 0.04410095512866974, "global_step": 213254, "epoch": 2396, "lr": 1.0957012136730343e-05}
{"train_loss": 0.11924383789300919, "global_step": 213255, "epoch": 2396, "lr": 1.0956649974210032e-05}
{"train_loss": 0.029974035918712616, "global_step": 213256, "epoch": 2396, "lr": 1.095628781693861e-05}
{"train_loss": 0.07056385278701782, "global_step": 213257, "epoch": 2396, "lr": 1.095592566491615e-05}
{"train_loss": 0.05001198127865791, "global_step": 213258, "epoch": 2396, "lr": 1.095556351814268e-05}
{"train_loss": 0.03878534212708473, "global_step": 213259, "epoch": 2396, "lr": 1.0955201376618268e-05}
{"train_loss": 0.0357368066906929, "global_step": 213260, "epoch": 2396, "lr": 1.0954839240342946e-05}
{"train_loss": 0.05032753571867943, "global_step": 213261, "epoch": 2396, "lr": 1.095447710931678e-05}
{"train_loss": 0.06883350759744644, "global_step": 213262, "epoch": 2396, "lr": 1.0954114983539793e-05}
{"train_loss": 0.10000724345445633, "global_step": 213263, "epoch": 2396, "lr": 1.0953752863012062e-05}
{"train_loss": 0.02650984190404415, "global_step": 213264, "epoch": 2396, "lr": 1.0953390747733611e-05}
{"train_loss": 0.08413859456777573, "global_step": 213265, "epoch": 2396, "lr": 1.0953028637704504e-05}
{"train_loss": 0.04834529012441635, "global_step": 213266, "epoch": 2396, "lr": 1.0952666532924799e-05}
{"train_loss": 0.05944574251770973, "global_step": 213267, "epoch": 2396, "lr": 1.0952304433394517e-05}
{"train_loss": 0.08362625539302826, "global_step": 213268, "epoch": 2396, "lr": 1.0951942339113736e-05}
{"train_loss": 0.08122406154870987, "global_step": 213269, "epoch": 2396, "lr": 1.0951580250082472e-05}
{"train_loss": 0.06352786719799042, "global_step": 213270, "epoch": 2396, "lr": 1.0951218166300815e-05}
{"train_loss": 0.044888611882925034, "global_step": 213271, "epoch": 2396, "lr": 1.0950856087768768e-05}
{"train_loss": 0.058552999049425125, "global_step": 213272, "epoch": 2396, "lr": 1.0950494014486423e-05}
{"train_loss": 0.1635880470275879, "global_step": 213273, "epoch": 2396, "lr": 1.095013194645379e-05}
{"train_loss": 0.031831249594688416, "global_step": 213274, "epoch": 2396, "lr": 1.0949769883670952e-05}
{"train_loss": 0.05383807048201561, "global_step": 213275, "epoch": 2396, "lr": 1.0949407826137925e-05}
{"train_loss": 0.04798276722431183, "global_step": 213276, "epoch": 2396, "lr": 1.0949045773854789e-05}
{"train_loss": 0.035645242780447006, "global_step": 213277, "epoch": 2396, "lr": 1.0948683726821568e-05}
{"train_loss": 0.049931347370147705, "global_step": 213278, "epoch": 2396, "lr": 1.0948321685038332e-05}
{"train_loss": 0.08845842629671097, "global_step": 213279, "epoch": 2396, "lr": 1.0947959648505101e-05}
{"train_loss": 0.020636465400457382, "global_step": 213280, "epoch": 2396, "lr": 1.0947597617221955e-05}
{"train_loss": 0.03799791634082794, "global_step": 213281, "epoch": 2396, "lr": 1.094723559118892e-05}
{"train_loss": 0.09297061711549759, "global_step": 213282, "epoch": 2396, "lr": 1.0946873570406064e-05}
{"train_loss": 0.06368684023618698, "global_step": 213283, "epoch": 2396, "lr": 1.0946511554873412e-05}
{"train_loss": 0.02976061962544918, "global_step": 213284, "epoch": 2396, "lr": 1.0946149544591039e-05}
{"train_loss": 0.034246087074279785, "global_step": 213285, "epoch": 2396, "lr": 1.094578753955896e-05}
{"train_loss": 0.05435681715607643, "global_step": 213286, "epoch": 2396, "lr": 1.0945425539777265e-05}
{"train_loss": 0.07041557878255844, "global_step": 213287, "epoch": 2396, "lr": 1.0945063545245965e-05}
{"train_loss": 0.02750374563038349, "global_step": 213288, "epoch": 2396, "lr": 1.0944701555965142e-05}
{"train_loss": 0.07077709585428238, "global_step": 213289, "epoch": 2396, "lr": 1.0944339571934814e-05}
{"train_loss": 0.07675588130950928, "global_step": 213290, "epoch": 2396, "lr": 1.0943977593155053e-05}
{"train_loss": 0.05403652414679527, "global_step": 213291, "epoch": 2396, "lr": 1.0943615619625903e-05}
{"train_loss": 0.07059844583272934, "global_step": 213292, "epoch": 2396, "lr": 1.094325365134739e-05}
{"train_loss": 0.06031282618641853, "global_step": 213293, "epoch": 2396, "lr": 1.0942891688319595e-05}
{"train_loss": 0.030765347182750702, "global_step": 213294, "epoch": 2396, "lr": 1.0942529730542539e-05}
{"train_loss": 0.05974849686026573, "global_step": 213295, "epoch": 2396, "lr": 1.09421677780163e-05}
{"train_loss": 0.0425969623029232, "global_step": 213296, "epoch": 2396, "lr": 1.0941805830740891e-05}
{"train_loss": 0.02940054051578045, "global_step": 213297, "epoch": 2396, "lr": 1.0941443888716402e-05}
{"train_loss": 0.037317391484975815, "global_step": 213298, "epoch": 2396, "lr": 1.0941081951942839e-05}
{"train_loss": 0.17139485478401184, "global_step": 213299, "epoch": 2396, "lr": 1.0940720020420292e-05}
{"train_loss": 0.03850090876221657, "global_step": 213300, "epoch": 2396, "lr": 1.0940358094148767e-05}
{"train_loss": 0.0504608228802681, "global_step": 213301, "epoch": 2396, "lr": 1.0939996173128341e-05}
{"train_loss": 0.055612653493881226, "global_step": 213302, "epoch": 2396, "lr": 1.0939634257359066e-05}
{"train_loss": 0.04464846849441528, "global_step": 213303, "epoch": 2396, "lr": 1.0939272346840973e-05}
{"train_loss": 0.04184873402118683, "global_step": 213304, "epoch": 2396, "lr": 1.0938910441574129e-05}
{"train_loss": 0.034108489751815796, "global_step": 213305, "epoch": 2396, "lr": 1.0938548541558558e-05}
{"train_loss": 0.12306397408246994, "global_step": 213306, "epoch": 2396, "lr": 1.0938186646794335e-05}
{"train_loss": 0.053534019738435745, "global_step": 213307, "epoch": 2396, "lr": 1.093782475728149e-05}
{"train_loss": 0.07031558454036713, "global_step": 213308, "epoch": 2396, "lr": 1.0937462873020088e-05}
{"train_loss": 0.1018037348985672, "global_step": 213309, "epoch": 2396, "lr": 1.0937100994010152e-05}
{"train_loss": 0.037485599517822266, "global_step": 213310, "epoch": 2396, "lr": 1.093673912025176e-05}
{"train_loss": 0.04374634101986885, "global_step": 213311, "epoch": 2396, "lr": 1.0936377251744939e-05}
{"train_loss": 0.0883232057094574, "global_step": 213312, "epoch": 2396, "lr": 1.0936015388489756e-05}
{"train_loss": 0.062284503132104874, "global_step": 213313, "epoch": 2396, "lr": 1.0935653530486234e-05}
{"train_loss": 0.05085555464029312, "global_step": 213314, "epoch": 2396, "lr": 1.0935291677734455e-05}
{"train_loss": 0.05940249562263489, "global_step": 213315, "epoch": 2396, "lr": 1.0934929830234436e-05}
{"train_loss": 0.06901644915342331, "global_step": 213316, "epoch": 2396, "lr": 1.093456798798625e-05}
{"train_loss": 0.05855974555015564, "global_step": 213317, "epoch": 2396, "lr": 1.0934206150989922e-05}
{"train_loss": 0.10299104452133179, "global_step": 213318, "epoch": 2396, "lr": 1.093384431924553e-05}
{"train_loss": 0.03208192437887192, "global_step": 213319, "epoch": 2396, "lr": 1.0933482492753094e-05}
{"train_loss": 0.0531986765563488, "global_step": 213320, "epoch": 2396, "lr": 1.093312067151267e-05}
{"train_loss": 0.028943171724677086, "global_step": 213321, "epoch": 2396, "lr": 1.0932758855524327e-05}
{"train_loss": 0.051834866404533386, "global_step": 213322, "epoch": 2396, "lr": 1.093239704478809e-05}
{"train_loss": 0.06953490525484085, "global_step": 213323, "epoch": 2396, "lr": 1.0932035239304023e-05}
{"train_loss": 0.05402017757296562, "global_step": 213324, "epoch": 2396, "lr": 1.0931673439072171e-05}
{"train_loss": 0.04400273412466049, "global_step": 213325, "epoch": 2396, "lr": 1.0931311644092567e-05}
{"train_loss": 0.050890401005744934, "global_step": 213326, "epoch": 2396, "lr": 1.093094985436528e-05}
{"train_loss": 0.053622934967279434, "global_step": 213327, "epoch": 2396, "lr": 1.093058806989034e-05}
{"train_loss": 0.05187766253948212, "global_step": 213328, "epoch": 2396, "lr": 1.0930226290667821e-05}
{"train_loss": 0.03695749118924141, "global_step": 213329, "epoch": 2396, "lr": 1.0929864516697758e-05}
{"train_loss": 0.0750821977853775, "global_step": 213330, "epoch": 2396, "lr": 1.0929502747980186e-05}
{"train_loss": 0.0619286485016346, "global_step": 213331, "epoch": 2396, "lr": 1.0929140984515174e-05}
{"train_loss": 0.058465805951129185, "global_step": 213332, "epoch": 2396, "lr": 1.0928779226302755e-05, "val_loss": 8.773675918579102}
{"train_loss": 0.04219413921236992, "global_step": 213333, "epoch": 2397, "lr": 1.0928417473342994e-05}
{"train_loss": 0.05328286439180374, "global_step": 213334, "epoch": 2397, "lr": 1.0928055725635923e-05}
{"train_loss": 0.03879360854625702, "global_step": 213335, "epoch": 2397, "lr": 1.0927693983181608e-05}
{"train_loss": 0.06442741304636002, "global_step": 213336, "epoch": 2397, "lr": 1.0927332245980076e-05}
{"train_loss": 0.0016820705495774746, "global_step": 213337, "epoch": 2397, "lr": 1.0926970514031387e-05}
{"train_loss": 0.044484782963991165, "global_step": 213338, "epoch": 2397, "lr": 1.0926608787335602e-05}
{"train_loss": 0.042721036821603775, "global_step": 213339, "epoch": 2397, "lr": 1.0926247065892747e-05}
{"train_loss": 0.09403455257415771, "global_step": 213340, "epoch": 2397, "lr": 1.0925885349702898e-05}
{"train_loss": 0.07607212662696838, "global_step": 213341, "epoch": 2397, "lr": 1.0925523638766072e-05}
{"train_loss": 0.053957682102918625, "global_step": 213342, "epoch": 2397, "lr": 1.0925161933082345e-05}
{"train_loss": 0.08687909692525864, "global_step": 213343, "epoch": 2397, "lr": 1.0924800232651744e-05}
{"train_loss": 0.07376338541507721, "global_step": 213344, "epoch": 2397, "lr": 1.0924438537474335e-05}
{"train_loss": 0.026485508307814598, "global_step": 213345, "epoch": 2397, "lr": 1.0924076847550153e-05}
{"train_loss": 0.07229963690042496, "global_step": 213346, "epoch": 2397, "lr": 1.0923715162879262e-05}
{"train_loss": 0.07181164622306824, "global_step": 213347, "epoch": 2397, "lr": 1.0923353483461684e-05}
{"train_loss": 0.03401268273591995, "global_step": 213348, "epoch": 2397, "lr": 1.0922991809297506e-05}
{"train_loss": 0.09602462500333786, "global_step": 213349, "epoch": 2397, "lr": 1.0922630140386736e-05}
{"train_loss": 0.04626763239502907, "global_step": 213350, "epoch": 2397, "lr": 1.0922268476729458e-05}
{"train_loss": 0.024129195138812065, "global_step": 213351, "epoch": 2397, "lr": 1.0921906818325689e-05}
{"train_loss": 0.0591406486928463, "global_step": 213352, "epoch": 2397, "lr": 1.0921545165175513e-05}
{"train_loss": 0.027554428204894066, "global_step": 213353, "epoch": 2397, "lr": 1.0921183517278938e-05}
{"train_loss": 0.046864286065101624, "global_step": 213354, "epoch": 2397, "lr": 1.092082187463605e-05}
{"train_loss": 0.0572480782866478, "global_step": 213355, "epoch": 2397, "lr": 1.0920460237246872e-05}
{"train_loss": 0.04564570635557175, "global_step": 213356, "epoch": 2397, "lr": 1.0920098605111456e-05}
{"train_loss": 0.0453927107155323, "global_step": 213357, "epoch": 2397, "lr": 1.0919736978229872e-05}
{"train_loss": 0.05177445709705353, "global_step": 213358, "epoch": 2397, "lr": 1.091937535660214e-05}
{"train_loss": 0.030627569183707237, "global_step": 213359, "epoch": 2397, "lr": 1.0919013740228334e-05}
{"train_loss": 0.1059819683432579, "global_step": 213360, "epoch": 2397, "lr": 1.0918652129108475e-05}
{"train_loss": 0.11224213242530823, "global_step": 213361, "epoch": 2397, "lr": 1.091829052324264e-05}
{"train_loss": 0.04496646299958229, "global_step": 213362, "epoch": 2397, "lr": 1.091792892263087e-05}
{"train_loss": 0.03868824988603592, "global_step": 213363, "epoch": 2397, "lr": 1.0917567327273192e-05}
{"train_loss": 0.09241000562906265, "global_step": 213364, "epoch": 2397, "lr": 1.0917205737169684e-05}
{"train_loss": 0.03810868412256241, "global_step": 213365, "epoch": 2397, "lr": 1.0916844152320365e-05}
{"train_loss": 0.04437939077615738, "global_step": 213366, "epoch": 2397, "lr": 1.0916482572725317e-05}
{"train_loss": 0.10238602012395859, "global_step": 213367, "epoch": 2397, "lr": 1.0916120998384572e-05}
{"train_loss": 0.044513922184705734, "global_step": 213368, "epoch": 2397, "lr": 1.0915759429298167e-05}
{"train_loss": 0.03743796795606613, "global_step": 213369, "epoch": 2397, "lr": 1.091539786546617e-05}
{"train_loss": 0.013898889534175396, "global_step": 213370, "epoch": 2397, "lr": 1.091503630688861e-05}
{"train_loss": 0.09843102842569351, "global_step": 213371, "epoch": 2397, "lr": 1.0914674753565556e-05}
{"train_loss": 0.06529521942138672, "global_step": 213372, "epoch": 2397, "lr": 1.0914313205497039e-05}
{"train_loss": 0.10722578316926956, "global_step": 213373, "epoch": 2397, "lr": 1.091395166268312e-05}
{"train_loss": 0.025030899792909622, "global_step": 213374, "epoch": 2397, "lr": 1.0913590125123852e-05}
{"train_loss": 0.038027774542570114, "global_step": 213375, "epoch": 2397, "lr": 1.0913228592819258e-05}
{"train_loss": 0.0390777550637722, "global_step": 213376, "epoch": 2397, "lr": 1.0912867065769421e-05}
{"train_loss": 0.0171731598675251, "global_step": 213377, "epoch": 2397, "lr": 1.0912505543974367e-05}
{"train_loss": 0.026737172156572342, "global_step": 213378, "epoch": 2397, "lr": 1.091214402743415e-05}
{"train_loss": 0.10340868681669235, "global_step": 213379, "epoch": 2397, "lr": 1.0911782516148817e-05}
{"train_loss": 0.049464162439107895, "global_step": 213380, "epoch": 2397, "lr": 1.0911421010118428e-05}
{"train_loss": 0.07076583802700043, "global_step": 213381, "epoch": 2397, "lr": 1.0911059509343008e-05}
{"train_loss": 0.0671161562204361, "global_step": 213382, "epoch": 2397, "lr": 1.0910698013822629e-05}
{"train_loss": 0.02352861315011978, "global_step": 213383, "epoch": 2397, "lr": 1.0910336523557324e-05}
{"train_loss": 0.07406877726316452, "global_step": 213384, "epoch": 2397, "lr": 1.0909975038547154e-05}
{"train_loss": 0.016851091757416725, "global_step": 213385, "epoch": 2397, "lr": 1.0909613558792154e-05}
{"train_loss": 0.02824592962861061, "global_step": 213386, "epoch": 2397, "lr": 1.0909252084292393e-05}
{"train_loss": 0.013359741307795048, "global_step": 213387, "epoch": 2397, "lr": 1.090889061504789e-05}
{"train_loss": 0.04340730234980583, "global_step": 213388, "epoch": 2397, "lr": 1.0908529151058722e-05}
{"train_loss": 0.04375442862510681, "global_step": 213389, "epoch": 2397, "lr": 1.0908167692324917e-05}
{"train_loss": 0.026593264192342758, "global_step": 213390, "epoch": 2397, "lr": 1.0907806238846546e-05}
{"train_loss": 0.02628590539097786, "global_step": 213391, "epoch": 2397, "lr": 1.0907444790623628e-05}
{"train_loss": 0.06032203882932663, "global_step": 213392, "epoch": 2397, "lr": 1.0907083347656228e-05}
{"train_loss": 0.04611596092581749, "global_step": 213393, "epoch": 2397, "lr": 1.0906721909944406e-05}
{"train_loss": 0.04606529697775841, "global_step": 213394, "epoch": 2397, "lr": 1.0906360477488187e-05}
{"train_loss": 0.054747454822063446, "global_step": 213395, "epoch": 2397, "lr": 1.0905999050287646e-05}
{"train_loss": 0.04386403039097786, "global_step": 213396, "epoch": 2397, "lr": 1.0905637628342802e-05}
{"train_loss": 0.025520505383610725, "global_step": 213397, "epoch": 2397, "lr": 1.0905276211653736e-05}
{"train_loss": 0.07787246257066727, "global_step": 213398, "epoch": 2397, "lr": 1.0904914800220461e-05}
{"train_loss": 0.06870083510875702, "global_step": 213399, "epoch": 2397, "lr": 1.090455339404306e-05}
{"train_loss": 0.08872295171022415, "global_step": 213400, "epoch": 2397, "lr": 1.090419199312156e-05}
{"train_loss": 0.05170104280114174, "global_step": 213401, "epoch": 2397, "lr": 1.0903830597456005e-05}
{"train_loss": 0.0506211519241333, "global_step": 213402, "epoch": 2397, "lr": 1.0903469207046469e-05}
{"train_loss": 0.05482247471809387, "global_step": 213403, "epoch": 2397, "lr": 1.0903107821892972e-05}
{"train_loss": 0.051163870841264725, "global_step": 213404, "epoch": 2397, "lr": 1.0902746441995582e-05}
{"train_loss": 0.0315980389714241, "global_step": 213405, "epoch": 2397, "lr": 1.0902385067354337e-05}
{"train_loss": 0.06403049826622009, "global_step": 213406, "epoch": 2397, "lr": 1.0902023697969294e-05}
{"train_loss": 0.039572231471538544, "global_step": 213407, "epoch": 2397, "lr": 1.0901662333840506e-05}
{"train_loss": 0.06619825214147568, "global_step": 213408, "epoch": 2397, "lr": 1.0901300974967992e-05}
{"train_loss": 0.058371227234601974, "global_step": 213409, "epoch": 2397, "lr": 1.0900939621351841e-05}
{"train_loss": 0.023733876645565033, "global_step": 213410, "epoch": 2397, "lr": 1.0900578272992068e-05}
{"train_loss": 0.04929901286959648, "global_step": 213411, "epoch": 2397, "lr": 1.0900216929888729e-05}
{"train_loss": 0.05179901421070099, "global_step": 213412, "epoch": 2397, "lr": 1.0899855592041902e-05}
{"train_loss": 0.06391846388578415, "global_step": 213413, "epoch": 2397, "lr": 1.0899494259451592e-05}
{"train_loss": 0.06558579206466675, "global_step": 213414, "epoch": 2397, "lr": 1.0899132932117884e-05}
{"train_loss": 0.047375570982694626, "global_step": 213415, "epoch": 2397, "lr": 1.0898771610040798e-05}
{"train_loss": 0.05412356182932854, "global_step": 213416, "epoch": 2397, "lr": 1.0898410293220412e-05}
{"train_loss": 0.01954728551208973, "global_step": 213417, "epoch": 2397, "lr": 1.0898048981656745e-05}
{"train_loss": 0.0662202388048172, "global_step": 213418, "epoch": 2397, "lr": 1.0897687675349865e-05}
{"train_loss": 0.018909044563770294, "global_step": 213419, "epoch": 2397, "lr": 1.0897326374299805e-05}
{"train_loss": 0.06754326820373535, "global_step": 213420, "epoch": 2397, "lr": 1.0896965078506637e-05}
{"train_loss": 0.05265850816026665, "global_step": 213421, "epoch": 2397, "lr": 1.0896603787970383e-05, "val_loss": 8.784008979797363}
{"train_loss": 0.06545303016901016, "global_step": 213422, "epoch": 2398, "lr": 1.0896242502691117e-05}
{"train_loss": 0.10242493450641632, "global_step": 213423, "epoch": 2398, "lr": 1.0895881222668858e-05}
{"train_loss": 0.05479785054922104, "global_step": 213424, "epoch": 2398, "lr": 1.0895519947903688e-05}
{"train_loss": 0.02493276819586754, "global_step": 213425, "epoch": 2398, "lr": 1.0895158678395623e-05}
{"train_loss": 0.07025706022977829, "global_step": 213426, "epoch": 2398, "lr": 1.0894797414144741e-05}
{"train_loss": 0.05187593400478363, "global_step": 213427, "epoch": 2398, "lr": 1.0894436155151066e-05}
{"train_loss": 0.048861585557460785, "global_step": 213428, "epoch": 2398, "lr": 1.0894074901414653e-05}
{"train_loss": 0.11026567965745926, "global_step": 213429, "epoch": 2398, "lr": 1.0893713652935572e-05}
{"train_loss": 0.04911806434392929, "global_step": 213430, "epoch": 2398, "lr": 1.0893352409713841e-05}
{"train_loss": 0.05582317337393761, "global_step": 213431, "epoch": 2398, "lr": 1.0892991171749534e-05}
{"train_loss": 0.03607177734375, "global_step": 213432, "epoch": 2398, "lr": 1.0892629939042676e-05}
{"train_loss": 0.04251990467309952, "global_step": 213433, "epoch": 2398, "lr": 1.089226871159334e-05}
{"train_loss": 0.03839670866727829, "global_step": 213434, "epoch": 2398, "lr": 1.0891907489401548e-05}
{"train_loss": 0.0628054216504097, "global_step": 213435, "epoch": 2398, "lr": 1.0891546272467373e-05}
{"train_loss": 0.10081006586551666, "global_step": 213436, "epoch": 2398, "lr": 1.0891185060790842e-05}
{"train_loss": 0.08407159149646759, "global_step": 213437, "epoch": 2398, "lr": 1.0890823854372034e-05}
{"train_loss": 0.0705782100558281, "global_step": 213438, "epoch": 2398, "lr": 1.089046265321096e-05}
{"train_loss": 0.02797204628586769, "global_step": 213439, "epoch": 2398, "lr": 1.0890101457307695e-05}
{"train_loss": 0.06389285624027252, "global_step": 213440, "epoch": 2398, "lr": 1.0889740266662285e-05}
{"train_loss": 0.060557276010513306, "global_step": 213441, "epoch": 2398, "lr": 1.0889379081274758e-05}
{"train_loss": 0.04354438558220863, "global_step": 213442, "epoch": 2398, "lr": 1.0889017901145188e-05}
{"train_loss": 0.018036818131804466, "global_step": 213443, "epoch": 2398, "lr": 1.0888656726273599e-05}
{"train_loss": 0.05825001001358032, "global_step": 213444, "epoch": 2398, "lr": 1.0888295556660072e-05}
{"train_loss": 0.05288248136639595, "global_step": 213445, "epoch": 2398, "lr": 1.0887934392304633e-05}
{"train_loss": 0.0651354268193245, "global_step": 213446, "epoch": 2398, "lr": 1.0887573233207327e-05}
{"train_loss": 0.030085325241088867, "global_step": 213447, "epoch": 2398, "lr": 1.0887212079368203e-05}
{"train_loss": 0.049777474254369736, "global_step": 213448, "epoch": 2398, "lr": 1.0886850930787335e-05}
{"train_loss": 0.061793163418769836, "global_step": 213449, "epoch": 2398, "lr": 1.0886489787464738e-05}
{"train_loss": 0.04963432624936104, "global_step": 213450, "epoch": 2398, "lr": 1.0886128649400485e-05}
{"train_loss": 0.059035949409008026, "global_step": 213451, "epoch": 2398, "lr": 1.0885767516594608e-05}
{"train_loss": 0.0749538466334343, "global_step": 213452, "epoch": 2398, "lr": 1.088540638904717e-05}
{"train_loss": 0.08539328724145889, "global_step": 213453, "epoch": 2398, "lr": 1.0885045266758203e-05}
{"train_loss": 0.07790789008140564, "global_step": 213454, "epoch": 2398, "lr": 1.088468414972778e-05}
{"train_loss": 0.010960469022393227, "global_step": 213455, "epoch": 2398, "lr": 1.0884323037955919e-05}
{"train_loss": 0.05177876353263855, "global_step": 213456, "epoch": 2398, "lr": 1.0883961931442699e-05}
{"train_loss": 0.08161173015832901, "global_step": 213457, "epoch": 2398, "lr": 1.0883600830188134e-05}
{"train_loss": 0.03624413534998894, "global_step": 213458, "epoch": 2398, "lr": 1.0883239734192314e-05}
{"train_loss": 0.029566403478384018, "global_step": 213459, "epoch": 2398, "lr": 1.0882878643455246e-05}
{"train_loss": 0.10301565378904343, "global_step": 213460, "epoch": 2398, "lr": 1.0882517557977012e-05}
{"train_loss": 0.054920878261327744, "global_step": 213461, "epoch": 2398, "lr": 1.0882156477757638e-05}
{"train_loss": 0.037997372448444366, "global_step": 213462, "epoch": 2398, "lr": 1.088179540279719e-05}
{"train_loss": 0.031035035848617554, "global_step": 213463, "epoch": 2398, "lr": 1.0881434333095697e-05}
{"train_loss": 0.037064481526613235, "global_step": 213464, "epoch": 2398, "lr": 1.0881073268653219e-05}
{"train_loss": 0.040020521730184555, "global_step": 213465, "epoch": 2398, "lr": 1.0880712209469811e-05}
{"train_loss": 0.08959127217531204, "global_step": 213466, "epoch": 2398, "lr": 1.088035115554551e-05}
{"train_loss": 0.0534999780356884, "global_step": 213467, "epoch": 2398, "lr": 1.0879990106880378e-05}
{"train_loss": 0.05105637013912201, "global_step": 213468, "epoch": 2398, "lr": 1.087962906347444e-05}
{"train_loss": 0.061752013862133026, "global_step": 213469, "epoch": 2398, "lr": 1.0879268025327776e-05}
{"train_loss": 0.04285343736410141, "global_step": 213470, "epoch": 2398, "lr": 1.0878906992440401e-05}
{"train_loss": 0.09974753856658936, "global_step": 213471, "epoch": 2398, "lr": 1.0878545964812398e-05}
{"train_loss": 0.041080161929130554, "global_step": 213472, "epoch": 2398, "lr": 1.0878184942443775e-05}
{"train_loss": 0.043013352900743484, "global_step": 213473, "epoch": 2398, "lr": 1.0877823925334623e-05}
{"train_loss": 0.05515465512871742, "global_step": 213474, "epoch": 2398, "lr": 1.0877462913484959e-05}
{"train_loss": 0.035692859441041946, "global_step": 213475, "epoch": 2398, "lr": 1.0877101906894855e-05}
{"train_loss": 0.07077082246541977, "global_step": 213476, "epoch": 2398, "lr": 1.0876740905564337e-05}
{"train_loss": 0.06188777834177017, "global_step": 213477, "epoch": 2398, "lr": 1.0876379909493473e-05}
{"train_loss": 0.04751851037144661, "global_step": 213478, "epoch": 2398, "lr": 1.0876018918682302e-05}
{"train_loss": 0.13504579663276672, "global_step": 213479, "epoch": 2398, "lr": 1.0875657933130862e-05}
{"train_loss": 0.01872285269200802, "global_step": 213480, "epoch": 2398, "lr": 1.087529695283923e-05}
{"train_loss": 0.06209097430109978, "global_step": 213481, "epoch": 2398, "lr": 1.0874935977807422e-05}
{"train_loss": 0.04964018240571022, "global_step": 213482, "epoch": 2398, "lr": 1.0874575008035514e-05}
{"train_loss": 0.07388176023960114, "global_step": 213483, "epoch": 2398, "lr": 1.087421404352353e-05}
{"train_loss": 0.03670912981033325, "global_step": 213484, "epoch": 2398, "lr": 1.087385308427154e-05}
{"train_loss": 0.07296670228242874, "global_step": 213485, "epoch": 2398, "lr": 1.0873492130279578e-05}
{"train_loss": 0.05267558991909027, "global_step": 213486, "epoch": 2398, "lr": 1.0873131181547708e-05}
{"train_loss": 0.034123487770557404, "global_step": 213487, "epoch": 2398, "lr": 1.0872770238075952e-05}
{"train_loss": 0.0650387555360794, "global_step": 213488, "epoch": 2398, "lr": 1.0872409299864394e-05}
{"train_loss": 0.06095163896679878, "global_step": 213489, "epoch": 2398, "lr": 1.0872048366913046e-05}
{"train_loss": 0.045512761920690536, "global_step": 213490, "epoch": 2398, "lr": 1.087168743922199e-05}
{"train_loss": 0.11959724128246307, "global_step": 213491, "epoch": 2398, "lr": 1.0871326516791242e-05}
{"train_loss": 0.06619889289140701, "global_step": 213492, "epoch": 2398, "lr": 1.0870965599620885e-05}
{"train_loss": 0.03386400267481804, "global_step": 213493, "epoch": 2398, "lr": 1.0870604687710934e-05}
{"train_loss": 0.043721459805965424, "global_step": 213494, "epoch": 2398, "lr": 1.0870243781061468e-05}
{"train_loss": 0.03630281239748001, "global_step": 213495, "epoch": 2398, "lr": 1.0869882879672504e-05}
{"train_loss": 0.027338597923517227, "global_step": 213496, "epoch": 2398, "lr": 1.0869521983544129e-05}
{"train_loss": 0.014528798870742321, "global_step": 213497, "epoch": 2398, "lr": 1.086916109267635e-05}
{"train_loss": 0.07181114703416824, "global_step": 213498, "epoch": 2398, "lr": 1.0868800207069247e-05}
{"train_loss": 0.06553474813699722, "global_step": 213499, "epoch": 2398, "lr": 1.0868439326722845e-05}
{"train_loss": 0.03419262543320656, "global_step": 213500, "epoch": 2398, "lr": 1.0868078451637209e-05}
{"train_loss": 0.0779263973236084, "global_step": 213501, "epoch": 2398, "lr": 1.0867717581812392e-05}
{"train_loss": 0.0390346460044384, "global_step": 213502, "epoch": 2398, "lr": 1.0867356717248418e-05}
{"train_loss": 0.027407683432102203, "global_step": 213503, "epoch": 2398, "lr": 1.0866995857945367e-05}
{"train_loss": 0.08780130743980408, "global_step": 213504, "epoch": 2398, "lr": 1.0866635003903258e-05}
{"train_loss": 0.026721904054284096, "global_step": 213505, "epoch": 2398, "lr": 1.0866274155122168e-05}
{"train_loss": 0.042194511741399765, "global_step": 213506, "epoch": 2398, "lr": 1.0865913311602116e-05}
{"train_loss": 0.05901104956865311, "global_step": 213507, "epoch": 2398, "lr": 1.0865552473343177e-05}
{"train_loss": 0.026243263855576515, "global_step": 213508, "epoch": 2398, "lr": 1.0865191640345374e-05}
{"train_loss": 0.05724416673183441, "global_step": 213509, "epoch": 2398, "lr": 1.0864830812608785e-05}
{"train_loss": 0.05538670401601644, "global_step": 213510, "epoch": 2398, "lr": 1.0864469990133425e-05, "val_loss": 8.991483688354492}
{"train_loss": 0.05210382491350174, "global_step": 213511, "epoch": 2399, "lr": 1.0864109172919379e-05}
{"train_loss": 0.0855591669678688, "global_step": 213512, "epoch": 2399, "lr": 1.0863748360966659e-05}
{"train_loss": 0.11539079993963242, "global_step": 213513, "epoch": 2399, "lr": 1.0863387554275345e-05}
{"train_loss": 0.041766922920942307, "global_step": 213514, "epoch": 2399, "lr": 1.0863026752845457e-05}
{"train_loss": 0.031034046784043312, "global_step": 213515, "epoch": 2399, "lr": 1.0862665956677076e-05}
{"train_loss": 0.04349137097597122, "global_step": 213516, "epoch": 2399, "lr": 1.0862305165770231e-05}
{"train_loss": 0.11248461902141571, "global_step": 213517, "epoch": 2399, "lr": 1.0861944380124955e-05}
{"train_loss": 0.08466488867998123, "global_step": 213518, "epoch": 2399, "lr": 1.0861583599741332e-05}
{"train_loss": 0.04723694548010826, "global_step": 213519, "epoch": 2399, "lr": 1.0861222824619372e-05}
{"train_loss": 0.07489623129367828, "global_step": 213520, "epoch": 2399, "lr": 1.0860862054759163e-05}
{"train_loss": 0.02947959117591381, "global_step": 213521, "epoch": 2399, "lr": 1.0860501290160718e-05}
{"train_loss": 0.05415960028767586, "global_step": 213522, "epoch": 2399, "lr": 1.0860140530824114e-05}
{"train_loss": 0.0951259657740593, "global_step": 213523, "epoch": 2399, "lr": 1.0859779776749374e-05}
{"train_loss": 0.02549188956618309, "global_step": 213524, "epoch": 2399, "lr": 1.0859419027936573e-05}
{"train_loss": 0.03436128795146942, "global_step": 213525, "epoch": 2399, "lr": 1.0859058284385738e-05}
{"train_loss": 0.04927332326769829, "global_step": 213526, "epoch": 2399, "lr": 1.0858697546096935e-05}
{"train_loss": 0.08673445880413055, "global_step": 213527, "epoch": 2399, "lr": 1.0858336813070186e-05}
{"train_loss": 0.03747153282165527, "global_step": 213528, "epoch": 2399, "lr": 1.0857976085305577e-05}
{"train_loss": 0.06692097336053848, "global_step": 213529, "epoch": 2399, "lr": 1.0857615362803119e-05}
{"train_loss": 0.026775849983096123, "global_step": 213530, "epoch": 2399, "lr": 1.0857254645562887e-05}
{"train_loss": 0.06024007871747017, "global_step": 213531, "epoch": 2399, "lr": 1.085689393358491e-05}
{"train_loss": 0.08763178437948227, "global_step": 213532, "epoch": 2399, "lr": 1.0856533226869264e-05}
{"train_loss": 0.049227312207221985, "global_step": 213533, "epoch": 2399, "lr": 1.0856172525415965e-05}
{"train_loss": 0.11002085357904434, "global_step": 213534, "epoch": 2399, "lr": 1.0855811829225083e-05}
{"train_loss": 0.09373186528682709, "global_step": 213535, "epoch": 2399, "lr": 1.0855451138296652e-05}
{"train_loss": 0.08498021960258484, "global_step": 213536, "epoch": 2399, "lr": 1.0855090452630746e-05}
{"train_loss": 0.036883413791656494, "global_step": 213537, "epoch": 2399, "lr": 1.0854729772227374e-05}
{"train_loss": 0.025691675022244453, "global_step": 213538, "epoch": 2399, "lr": 1.0854369097086609e-05}
{"train_loss": 0.05553017556667328, "global_step": 213539, "epoch": 2399, "lr": 1.0854008427208512e-05}
{"train_loss": 0.05502961575984955, "global_step": 213540, "epoch": 2399, "lr": 1.0853647762593106e-05}
{"train_loss": 0.04144666716456413, "global_step": 213541, "epoch": 2399, "lr": 1.0853287103240456e-05}
{"train_loss": 0.07440883666276932, "global_step": 213542, "epoch": 2399, "lr": 1.0852926449150596e-05}
{"train_loss": 0.08258421719074249, "global_step": 213543, "epoch": 2399, "lr": 1.08525658003236e-05}
{"train_loss": 0.045329831540584564, "global_step": 213544, "epoch": 2399, "lr": 1.0852205156759476e-05}
{"train_loss": 0.04301195964217186, "global_step": 213545, "epoch": 2399, "lr": 1.0851844518458315e-05}
{"train_loss": 0.06283017992973328, "global_step": 213546, "epoch": 2399, "lr": 1.0851483885420133e-05}
{"train_loss": 0.039189085364341736, "global_step": 213547, "epoch": 2399, "lr": 1.0851123257645006e-05}
{"train_loss": 0.07144535332918167, "global_step": 213548, "epoch": 2399, "lr": 1.0850762635132955e-05}
{"train_loss": 0.0486098937690258, "global_step": 213549, "epoch": 2399, "lr": 1.0850402017884055e-05}
{"train_loss": 0.06108470261096954, "global_step": 213550, "epoch": 2399, "lr": 1.0850041405898325e-05}
{"train_loss": 0.03391076624393463, "global_step": 213551, "epoch": 2399, "lr": 1.0849680799175849e-05}
{"train_loss": 0.033197544515132904, "global_step": 213552, "epoch": 2399, "lr": 1.0849320197716634e-05}
{"train_loss": 0.06294465065002441, "global_step": 213553, "epoch": 2399, "lr": 1.084895960152077e-05}
{"train_loss": 0.08858741819858551, "global_step": 213554, "epoch": 2399, "lr": 1.0848599010588273e-05}
{"train_loss": 0.045405540615320206, "global_step": 213555, "epoch": 2399, "lr": 1.0848238424919216e-05}
{"train_loss": 0.05307505279779434, "global_step": 213556, "epoch": 2399, "lr": 1.0847877844513638e-05}
{"train_loss": 0.07093829661607742, "global_step": 213557, "epoch": 2399, "lr": 1.084751726937157e-05}
{"train_loss": 0.0705903097987175, "global_step": 213558, "epoch": 2399, "lr": 1.0847156699493094e-05}
{"train_loss": 0.07763273268938065, "global_step": 213559, "epoch": 2399, "lr": 1.0846796134878223e-05}
{"train_loss": 0.06847350299358368, "global_step": 213560, "epoch": 2399, "lr": 1.0846435575527037e-05}
{"train_loss": 0.04688585177063942, "global_step": 213561, "epoch": 2399, "lr": 1.0846075021439555e-05}
{"train_loss": 0.0671057403087616, "global_step": 213562, "epoch": 2399, "lr": 1.0845714472615858e-05}
{"train_loss": 0.04692574962973595, "global_step": 213563, "epoch": 2399, "lr": 1.0845353929055963e-05}
{"train_loss": 0.09211757034063339, "global_step": 213564, "epoch": 2399, "lr": 1.084499339075995e-05}
{"train_loss": 0.04337187856435776, "global_step": 213565, "epoch": 2399, "lr": 1.0844632857727833e-05}
{"train_loss": 0.07976140826940536, "global_step": 213566, "epoch": 2399, "lr": 1.0844272329959688e-05}
{"train_loss": 0.04450671002268791, "global_step": 213567, "epoch": 2399, "lr": 1.0843911807455543e-05}
{"train_loss": 0.0864042118191719, "global_step": 213568, "epoch": 2399, "lr": 1.084355129021547e-05}
{"train_loss": 0.07586158812046051, "global_step": 213569, "epoch": 2399, "lr": 1.0843190778239492e-05}
{"train_loss": 0.04428613930940628, "global_step": 213570, "epoch": 2399, "lr": 1.0842830271527688e-05}
{"train_loss": 0.06399285048246384, "global_step": 213571, "epoch": 2399, "lr": 1.0842469770080065e-05}
{"train_loss": 0.029916103929281235, "global_step": 213572, "epoch": 2399, "lr": 1.0842109273896717e-05}
{"train_loss": 0.06748379021883011, "global_step": 213573, "epoch": 2399, "lr": 1.084174878297765e-05}
{"train_loss": 0.034589484333992004, "global_step": 213574, "epoch": 2399, "lr": 1.0841388297322935e-05}
{"train_loss": 0.060088373720645905, "global_step": 213575, "epoch": 2399, "lr": 1.0841027816932631e-05}
{"train_loss": 0.041639093309640884, "global_step": 213576, "epoch": 2399, "lr": 1.084066734180676e-05}
{"train_loss": 0.02984207682311535, "global_step": 213577, "epoch": 2399, "lr": 1.0840306871945399e-05}
{"train_loss": 0.1057015210390091, "global_step": 213578, "epoch": 2399, "lr": 1.0839946407348566e-05}
{"train_loss": 0.06077883765101433, "global_step": 213579, "epoch": 2399, "lr": 1.083958594801634e-05}
{"train_loss": 0.04581410065293312, "global_step": 213580, "epoch": 2399, "lr": 1.083922549394874e-05}
{"train_loss": 0.038300350308418274, "global_step": 213581, "epoch": 2399, "lr": 1.0838865045145841e-05}
{"train_loss": 0.04839978367090225, "global_step": 213582, "epoch": 2399, "lr": 1.0838504601607669e-05}
{"train_loss": 0.050207193940877914, "global_step": 213583, "epoch": 2399, "lr": 1.0838144163334296e-05}
{"train_loss": 0.07491449266672134, "global_step": 213584, "epoch": 2399, "lr": 1.083778373032574e-05}
{"train_loss": 0.04422834515571594, "global_step": 213585, "epoch": 2399, "lr": 1.0837423302582083e-05}
{"train_loss": 0.042201876640319824, "global_step": 213586, "epoch": 2399, "lr": 1.0837062880103344e-05}
{"train_loss": 0.057194046676158905, "global_step": 213587, "epoch": 2399, "lr": 1.0836702462889597e-05}
{"train_loss": 0.0439872182905674, "global_step": 213588, "epoch": 2399, "lr": 1.0836342050940868e-05}
{"train_loss": 0.0871325358748436, "global_step": 213589, "epoch": 2399, "lr": 1.0835981644257221e-05}
{"train_loss": 0.06829706579446793, "global_step": 213590, "epoch": 2399, "lr": 1.0835621242838689e-05}
{"train_loss": 0.06124202534556389, "global_step": 213591, "epoch": 2399, "lr": 1.0835260846685347e-05}
{"train_loss": 0.03662104159593582, "global_step": 213592, "epoch": 2399, "lr": 1.0834900455797215e-05}
{"train_loss": 0.038159504532814026, "global_step": 213593, "epoch": 2399, "lr": 1.083454007017436e-05}
{"train_loss": 0.05061771348118782, "global_step": 213594, "epoch": 2399, "lr": 1.0834179689816825e-05}
{"train_loss": 0.113092340528965, "global_step": 213595, "epoch": 2399, "lr": 1.0833819314724648e-05}
{"train_loss": 0.09268064796924591, "global_step": 213596, "epoch": 2399, "lr": 1.0833458944897895e-05}
{"train_loss": 0.06815352290868759, "global_step": 213597, "epoch": 2399, "lr": 1.0833098580336598e-05}
{"train_loss": 0.022349543869495392, "global_step": 213598, "epoch": 2399, "lr": 1.0832738221040823e-05}
{"train_loss": 0.05913564883005083, "global_step": 213599, "epoch": 2399, "lr": 1.0832377867010596e-05, "val_loss": 8.946951866149902}
{"train_loss": 0.05006943270564079, "global_step": 213600, "epoch": 2400, "lr": 1.0832017518245996e-05}
{"train_loss": 0.036477651447057724, "global_step": 213601, "epoch": 2400, "lr": 1.0831657174747035e-05}
{"train_loss": 0.09512584656476974, "global_step": 213602, "epoch": 2400, "lr": 1.0831296836513799e-05}
{"train_loss": 0.057484738528728485, "global_step": 213603, "epoch": 2400, "lr": 1.0830936503546296e-05}
{"train_loss": 0.13033078610897064, "global_step": 213604, "epoch": 2400, "lr": 1.0830576175844614e-05}
{"train_loss": 0.04919407144188881, "global_step": 213605, "epoch": 2400, "lr": 1.0830215853408775e-05}
{"train_loss": 0.07695033401250839, "global_step": 213606, "epoch": 2400, "lr": 1.0829855536238842e-05}
{"train_loss": 0.026102911680936813, "global_step": 213607, "epoch": 2400, "lr": 1.0829495224334851e-05}
{"train_loss": 0.04426182061433792, "global_step": 213608, "epoch": 2400, "lr": 1.0829134917696865e-05}
{"train_loss": 0.071988545358181, "global_step": 213609, "epoch": 2400, "lr": 1.082877461632491e-05}
{"train_loss": 0.03617328405380249, "global_step": 213610, "epoch": 2400, "lr": 1.0828414320219055e-05}
{"train_loss": 0.07804698497056961, "global_step": 213611, "epoch": 2400, "lr": 1.0828054029379353e-05}
{"train_loss": 0.0481354184448719, "global_step": 213612, "epoch": 2400, "lr": 1.0827693743805829e-05}
{"train_loss": 0.08344421535730362, "global_step": 213613, "epoch": 2400, "lr": 1.082733346349855e-05}
{"train_loss": 0.025508839637041092, "global_step": 213614, "epoch": 2400, "lr": 1.0826973188457556e-05}
{"train_loss": 0.06041836738586426, "global_step": 213615, "epoch": 2400, "lr": 1.0826612918682905e-05}
{"train_loss": 0.05372278019785881, "global_step": 213616, "epoch": 2400, "lr": 1.082625265417463e-05}
{"train_loss": 0.0663788840174675, "global_step": 213617, "epoch": 2400, "lr": 1.0825892394932797e-05}
{"train_loss": 0.12177524715662003, "global_step": 213618, "epoch": 2400, "lr": 1.0825532140957429e-05}
{"train_loss": 0.04953780025243759, "global_step": 213619, "epoch": 2400, "lr": 1.082517189224861e-05}
{"train_loss": 0.09970506280660629, "global_step": 213620, "epoch": 2400, "lr": 1.0824811648806355e-05}
{"train_loss": 0.09472664445638657, "global_step": 213621, "epoch": 2400, "lr": 1.0824451410630736e-05}
{"train_loss": 0.08542608469724655, "global_step": 213622, "epoch": 2400, "lr": 1.0824091177721784e-05}
{"train_loss": 0.0734780803322792, "global_step": 213623, "epoch": 2400, "lr": 1.0823730950079564e-05}
{"train_loss": 0.031681228429079056, "global_step": 213624, "epoch": 2400, "lr": 1.0823370727704108e-05}
{"train_loss": 0.046762045472860336, "global_step": 213625, "epoch": 2400, "lr": 1.0823010510595477e-05}
{"train_loss": 0.033830318599939346, "global_step": 213626, "epoch": 2400, "lr": 1.0822650298753712e-05}
{"train_loss": 0.054274529218673706, "global_step": 213627, "epoch": 2400, "lr": 1.0822290092178871e-05}
{"train_loss": 0.05579652264714241, "global_step": 213628, "epoch": 2400, "lr": 1.082192989087098e-05}
{"train_loss": 0.055373530834913254, "global_step": 213629, "epoch": 2400, "lr": 1.0821569694830126e-05}
{"train_loss": 0.04090941697359085, "global_step": 213630, "epoch": 2400, "lr": 1.0821209504056312e-05}
{"train_loss": 0.048013146966695786, "global_step": 213631, "epoch": 2400, "lr": 1.0820849318549625e-05}
{"train_loss": 0.04101961851119995, "global_step": 213632, "epoch": 2400, "lr": 1.0820489138310081e-05}
{"train_loss": 0.06337244808673859, "global_step": 213633, "epoch": 2400, "lr": 1.082012896333776e-05}
{"train_loss": 0.05406879261136055, "global_step": 213634, "epoch": 2400, "lr": 1.08197687936327e-05}
{"train_loss": 0.07414457947015762, "global_step": 213635, "epoch": 2400, "lr": 1.0819408629194927e-05}
{"train_loss": 0.05568708851933479, "global_step": 213636, "epoch": 2400, "lr": 1.0819048470024522e-05}
{"train_loss": 0.08803635090589523, "global_step": 213637, "epoch": 2400, "lr": 1.08186883161215e-05}
{"train_loss": 0.10456553101539612, "global_step": 213638, "epoch": 2400, "lr": 1.0818328167485947e-05}
{"train_loss": 0.04755221679806709, "global_step": 213639, "epoch": 2400, "lr": 1.081796802411788e-05}
{"train_loss": 0.027150364592671394, "global_step": 213640, "epoch": 2400, "lr": 1.0817607886017366e-05}
{"train_loss": 0.10458844900131226, "global_step": 213641, "epoch": 2400, "lr": 1.081724775318444e-05}
{"train_loss": 0.034492701292037964, "global_step": 213642, "epoch": 2400, "lr": 1.0816887625619166e-05}
{"train_loss": 0.03994131460785866, "global_step": 213643, "epoch": 2400, "lr": 1.0816527503321571e-05}
{"train_loss": 0.0562257282435894, "global_step": 213644, "epoch": 2400, "lr": 1.0816167386291737e-05}
{"train_loss": 0.017127886414527893, "global_step": 213645, "epoch": 2400, "lr": 1.0815807274529671e-05}
{"train_loss": 0.04433528706431389, "global_step": 213646, "epoch": 2400, "lr": 1.0815447168035441e-05}
{"train_loss": 0.024885380640625954, "global_step": 213647, "epoch": 2400, "lr": 1.0815087066809116e-05}
{"train_loss": 0.0354607068002224, "global_step": 213648, "epoch": 2400, "lr": 1.0814726970850708e-05}
{"train_loss": 0.06275705993175507, "global_step": 213649, "epoch": 2400, "lr": 1.0814366880160299e-05}
{"train_loss": 0.08574549108743668, "global_step": 213650, "epoch": 2400, "lr": 1.0814006794737902e-05}
{"train_loss": 0.08448299020528793, "global_step": 213651, "epoch": 2400, "lr": 1.0813646714583603e-05}
{"train_loss": 0.05011216923594475, "global_step": 213652, "epoch": 2400, "lr": 1.0813286639697412e-05}
{"train_loss": 0.06531737744808197, "global_step": 213653, "epoch": 2400, "lr": 1.0812926570079412e-05}
{"train_loss": 0.041154645383358, "global_step": 213654, "epoch": 2400, "lr": 1.0812566505729626e-05}
{"train_loss": 0.08081096410751343, "global_step": 213655, "epoch": 2400, "lr": 1.0812206446648126e-05}
{"train_loss": 0.06120123714208603, "global_step": 213656, "epoch": 2400, "lr": 1.0811846392834934e-05}
{"train_loss": 0.06034224480390549, "global_step": 213657, "epoch": 2400, "lr": 1.0811486344290123e-05}
{"train_loss": 0.061134226620197296, "global_step": 213658, "epoch": 2400, "lr": 1.0811126301013719e-05}
{"train_loss": 0.09005018323659897, "global_step": 213659, "epoch": 2400, "lr": 1.0810766263005789e-05}
{"train_loss": 0.03243830054998398, "global_step": 213660, "epoch": 2400, "lr": 1.0810406230266368e-05}
{"train_loss": 0.06626825034618378, "global_step": 213661, "epoch": 2400, "lr": 1.0810046202795521e-05}
{"train_loss": 0.04409446194767952, "global_step": 213662, "epoch": 2400, "lr": 1.080968618059327e-05}
{"train_loss": 0.02471357211470604, "global_step": 213663, "epoch": 2400, "lr": 1.0809326163659694e-05}
{"train_loss": 0.04341810569167137, "global_step": 213664, "epoch": 2400, "lr": 1.0808966151994816e-05}
{"train_loss": 0.09045836329460144, "global_step": 213665, "epoch": 2400, "lr": 1.0808606145598687e-05}
{"train_loss": 0.06032867357134819, "global_step": 213666, "epoch": 2400, "lr": 1.0808246144471385e-05}
{"train_loss": 0.05140195041894913, "global_step": 213667, "epoch": 2400, "lr": 1.0807886148612934e-05}
{"train_loss": 0.053378503769636154, "global_step": 213668, "epoch": 2400, "lr": 1.0807526158023368e-05}
{"train_loss": 0.059152308851480484, "global_step": 213669, "epoch": 2400, "lr": 1.080716617270277e-05}
{"train_loss": 0.060564156621694565, "global_step": 213670, "epoch": 2400, "lr": 1.0806806192651153e-05}
{"train_loss": 0.09209919720888138, "global_step": 213671, "epoch": 2400, "lr": 1.0806446217868598e-05}
{"train_loss": 0.06770998984575272, "global_step": 213672, "epoch": 2400, "lr": 1.080608624835514e-05}
{"train_loss": 0.046966709196567535, "global_step": 213673, "epoch": 2400, "lr": 1.0805726284110817e-05}
{"train_loss": 0.027985522523522377, "global_step": 213674, "epoch": 2400, "lr": 1.0805366325135696e-05}
{"train_loss": 0.05926915630698204, "global_step": 213675, "epoch": 2400, "lr": 1.0805006371429799e-05}
{"train_loss": 0.05699826776981354, "global_step": 213676, "epoch": 2400, "lr": 1.080464642299321e-05}
{"train_loss": 0.12190835922956467, "global_step": 213677, "epoch": 2400, "lr": 1.0804286479825942e-05}
{"train_loss": 0.046647269278764725, "global_step": 213678, "epoch": 2400, "lr": 1.0803926541928072e-05}
{"train_loss": 0.022896673530340195, "global_step": 213679, "epoch": 2400, "lr": 1.0803566609299625e-05}
{"train_loss": 0.04231544956564903, "global_step": 213680, "epoch": 2400, "lr": 1.0803206681940676e-05}
{"train_loss": 0.021756548434495926, "global_step": 213681, "epoch": 2400, "lr": 1.0802846759851243e-05}
{"train_loss": 0.12182270735502243, "global_step": 213682, "epoch": 2400, "lr": 1.080248684303139e-05}
{"train_loss": 0.05072014778852463, "global_step": 213683, "epoch": 2400, "lr": 1.0802126931481177e-05}
{"train_loss": 0.05650429427623749, "global_step": 213684, "epoch": 2400, "lr": 1.0801767025200626e-05}
{"train_loss": 0.04484007880091667, "global_step": 213685, "epoch": 2400, "lr": 1.0801407124189816e-05}
{"train_loss": 0.09638616442680359, "global_step": 213686, "epoch": 2400, "lr": 1.0801047228448763e-05}
{"train_loss": 0.049648646265268326, "global_step": 213687, "epoch": 2400, "lr": 1.0800687337977545e-05}
{"train_loss": 0.059725791822825926, "global_step": 213688, "epoch": 2400, "lr": 1.0800327452776188e-05, "train/sim_max_reward_0": 0.5509626556492229, "train/sim_max_reward_1": 0.9769938244168578, "train/sim_max_reward_2": 0.45539568089114696, "train/sim_max_reward_3": 0.5216726332079465, "train/sim_max_reward_4": 0.9070504616497246, "train/sim_max_reward_5": 0.5872720270543714, "test/sim_max_reward_4500000": 0.43198065355262333, "test/sim_max_reward_4500001": 0.39461645520635164, "test/sim_max_reward_4500002": 0.9799072709728632, "test/sim_max_reward_4500003": 0.5543250909534634, "test/sim_max_reward_4500004": 0.6059415368263945, "test/sim_max_reward_4500005": 0.39537741211832933, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.990919403954641, "test/sim_max_reward_4500008": 0.5120938986284502, "test/sim_max_reward_4500009": 0.05066133513303226, "test/sim_max_reward_4500010": 0.8841285558211197, "test/sim_max_reward_4500011": 0.9862493349843253, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.451228917842622, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9400864847538699, "test/sim_max_reward_4500016": 0.18299521092958235, "test/sim_max_reward_4500017": 0.32551532749676665, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9720177842735451, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.5493815211742239, "test/sim_max_reward_4500023": 0.7176007707214179, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.4138181470823304, "test/sim_max_reward_4500026": 0.49473794694330087, "test/sim_max_reward_4500027": 0.98125034236307, "test/sim_max_reward_4500028": 0.6578718711042801, "test/sim_max_reward_4500029": 0.539817728729648, "test/sim_max_reward_4500030": 0.5531214566792999, "test/sim_max_reward_4500031": 0.8242395454334244, "test/sim_max_reward_4500032": 0.49720756377676734, "test/sim_max_reward_4500033": 0.9933219172229969, "test/sim_max_reward_4500034": 0.39405200394497897, "test/sim_max_reward_4500035": 0.3014570116062118, "test/sim_max_reward_4500036": 0.06946556327721526, "test/sim_max_reward_4500037": 0.9835600369105381, "test/sim_max_reward_4500038": 0.8588907660470967, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9785682292833316, "test/sim_max_reward_4500041": 0.7121040394803405, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.8702657456903905, "test/sim_max_reward_4500044": 1.0, "test/sim_max_reward_4500045": 0.4051213704767077, "test/sim_max_reward_4500046": 0.6485470733128609, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.6665578804782117, "test/mean_score": 0.6220489064941682, "val_loss": 8.831318855285645, "train_action_mse_error": 15.452778816223145}
{"train_loss": 0.03683874383568764, "global_step": 213689, "epoch": 2401, "lr": 1.0799967572844755e-05}
{"train_loss": 0.046850934624671936, "global_step": 213690, "epoch": 2401, "lr": 1.0799607698183278e-05}
{"train_loss": 0.036128465086221695, "global_step": 213691, "epoch": 2401, "lr": 1.079924782879183e-05}
{"train_loss": 0.032689064741134644, "global_step": 213692, "epoch": 2401, "lr": 1.0798887964670435e-05}
{"train_loss": 0.06006157398223877, "global_step": 213693, "epoch": 2401, "lr": 1.0798528105819162e-05}
{"train_loss": 0.05951635539531708, "global_step": 213694, "epoch": 2401, "lr": 1.079816825223804e-05}
{"train_loss": 0.0758308619260788, "global_step": 213695, "epoch": 2401, "lr": 1.0797808403927135e-05}
{"train_loss": 0.053623396903276443, "global_step": 213696, "epoch": 2401, "lr": 1.0797448560886475e-05}
{"train_loss": 0.06970831006765366, "global_step": 213697, "epoch": 2401, "lr": 1.0797088723116133e-05}
{"train_loss": 0.09338889271020889, "global_step": 213698, "epoch": 2401, "lr": 1.079672889061613e-05}
{"train_loss": 0.04842010512948036, "global_step": 213699, "epoch": 2401, "lr": 1.079636906338654e-05}
{"train_loss": 0.03578934073448181, "global_step": 213700, "epoch": 2401, "lr": 1.0796009241427396e-05}
{"train_loss": 0.07636315375566483, "global_step": 213701, "epoch": 2401, "lr": 1.0795649424738747e-05}
{"train_loss": 0.03233415633440018, "global_step": 213702, "epoch": 2401, "lr": 1.0795289613320653e-05}
{"train_loss": 0.018773086369037628, "global_step": 213703, "epoch": 2401, "lr": 1.0794929807173149e-05}
{"train_loss": 0.06877825409173965, "global_step": 213704, "epoch": 2401, "lr": 1.0794570006296295e-05}
{"train_loss": 0.03496026247739792, "global_step": 213705, "epoch": 2401, "lr": 1.0794210210690137e-05}
{"train_loss": 0.04168928414583206, "global_step": 213706, "epoch": 2401, "lr": 1.0793850420354706e-05}
{"train_loss": 0.06907354295253754, "global_step": 213707, "epoch": 2401, "lr": 1.0793490635290076e-05}
{"train_loss": 0.047615259885787964, "global_step": 213708, "epoch": 2401, "lr": 1.0793130855496269e-05}
{"train_loss": 0.03418315201997757, "global_step": 213709, "epoch": 2401, "lr": 1.0792771080973368e-05}
{"train_loss": 0.08468330651521683, "global_step": 213710, "epoch": 2401, "lr": 1.0792411311721396e-05}
{"train_loss": 0.07667825371026993, "global_step": 213711, "epoch": 2401, "lr": 1.079205154774039e-05}
{"train_loss": 0.0907161757349968, "global_step": 213712, "epoch": 2401, "lr": 1.0791691789030434e-05}
{"train_loss": 0.0424153171479702, "global_step": 213713, "epoch": 2401, "lr": 1.0791332035591539e-05}
{"train_loss": 0.059774257242679596, "global_step": 213714, "epoch": 2401, "lr": 1.0790972287423789e-05}
{"train_loss": 0.0680866613984108, "global_step": 213715, "epoch": 2401, "lr": 1.0790612544527201e-05}
{"train_loss": 0.05496455729007721, "global_step": 213716, "epoch": 2401, "lr": 1.0790252806901845e-05}
{"train_loss": 0.04628269374370575, "global_step": 213717, "epoch": 2401, "lr": 1.0789893074547752e-05}
{"train_loss": 0.06924234330654144, "global_step": 213718, "epoch": 2401, "lr": 1.0789533347464981e-05}
{"train_loss": 0.05141950771212578, "global_step": 213719, "epoch": 2401, "lr": 1.0789173625653593e-05}
{"train_loss": 0.036771975457668304, "global_step": 213720, "epoch": 2401, "lr": 1.078881390911361e-05}
{"train_loss": 0.01768983155488968, "global_step": 213721, "epoch": 2401, "lr": 1.0788454197845105e-05}
{"train_loss": 0.07080358266830444, "global_step": 213722, "epoch": 2401, "lr": 1.07880944918481e-05}
{"train_loss": 0.05251521244645119, "global_step": 213723, "epoch": 2401, "lr": 1.0787734791122672e-05}
{"train_loss": 0.0886353850364685, "global_step": 213724, "epoch": 2401, "lr": 1.0787375095668839e-05}
{"train_loss": 0.04321005567908287, "global_step": 213725, "epoch": 2401, "lr": 1.0787015405486684e-05}
{"train_loss": 0.055308207869529724, "global_step": 213726, "epoch": 2401, "lr": 1.0786655720576222e-05}
{"train_loss": 0.05093744769692421, "global_step": 213727, "epoch": 2401, "lr": 1.0786296040937527e-05}
{"train_loss": 0.034435659646987915, "global_step": 213728, "epoch": 2401, "lr": 1.0785936366570626e-05}
{"train_loss": 0.030792254954576492, "global_step": 213729, "epoch": 2401, "lr": 1.078557669747559e-05}
{"train_loss": 0.061723414808511734, "global_step": 213730, "epoch": 2401, "lr": 1.0785217033652446e-05}
{"train_loss": 0.02344776876270771, "global_step": 213731, "epoch": 2401, "lr": 1.078485737510126e-05}
{"train_loss": 0.036653947085142136, "global_step": 213732, "epoch": 2401, "lr": 1.0784497721822057e-05}
{"train_loss": 0.04190153628587723, "global_step": 213733, "epoch": 2401, "lr": 1.0784138073814914e-05}
{"train_loss": 0.06364384293556213, "global_step": 213734, "epoch": 2401, "lr": 1.0783778431079856e-05}
{"train_loss": 0.06179272383451462, "global_step": 213735, "epoch": 2401, "lr": 1.0783418793616956e-05}
{"train_loss": 0.067777119576931, "global_step": 213736, "epoch": 2401, "lr": 1.078305916142623e-05}
{"train_loss": 0.0761176124215126, "global_step": 213737, "epoch": 2401, "lr": 1.0782699534507746e-05}
{"train_loss": 0.06055227667093277, "global_step": 213738, "epoch": 2401, "lr": 1.0782339912861561e-05}
{"train_loss": 0.04908561334013939, "global_step": 213739, "epoch": 2401, "lr": 1.0781980296487698e-05}
{"train_loss": 0.07014641165733337, "global_step": 213740, "epoch": 2401, "lr": 1.0781620685386235e-05}
{"train_loss": 0.022962357848882675, "global_step": 213741, "epoch": 2401, "lr": 1.0781261079557193e-05}
{"train_loss": 0.07933363318443298, "global_step": 213742, "epoch": 2401, "lr": 1.0780901479000643e-05}
{"train_loss": 0.05636678636074066, "global_step": 213743, "epoch": 2401, "lr": 1.0780541883716628e-05}
{"train_loss": 0.03010384365916252, "global_step": 213744, "epoch": 2401, "lr": 1.078018229370517e-05}
{"train_loss": 0.032111141830682755, "global_step": 213745, "epoch": 2401, "lr": 1.0779822708966358e-05}
{"train_loss": 0.02980049140751362, "global_step": 213746, "epoch": 2401, "lr": 1.0779463129500205e-05}
{"train_loss": 0.05902762711048126, "global_step": 213747, "epoch": 2401, "lr": 1.077910355530679e-05}
{"train_loss": 0.08757492899894714, "global_step": 213748, "epoch": 2401, "lr": 1.077874398638613e-05}
{"train_loss": 0.0808696523308754, "global_step": 213749, "epoch": 2401, "lr": 1.0778384422738307e-05}
{"train_loss": 0.07232420146465302, "global_step": 213750, "epoch": 2401, "lr": 1.0778024864363351e-05}
{"train_loss": 0.08521371334791183, "global_step": 213751, "epoch": 2401, "lr": 1.0777665311261297e-05}
{"train_loss": 0.04385115206241608, "global_step": 213752, "epoch": 2401, "lr": 1.0777305763432221e-05}
{"train_loss": 0.0918964296579361, "global_step": 213753, "epoch": 2401, "lr": 1.077694622087615e-05}
{"train_loss": 0.03643302991986275, "global_step": 213754, "epoch": 2401, "lr": 1.0776586683593132e-05}
{"train_loss": 0.05673367902636528, "global_step": 213755, "epoch": 2401, "lr": 1.0776227151583245e-05}
{"train_loss": 0.03542890027165413, "global_step": 213756, "epoch": 2401, "lr": 1.0775867624846498e-05}
{"train_loss": 0.04647308960556984, "global_step": 213757, "epoch": 2401, "lr": 1.0775508103382969e-05}
{"train_loss": 0.023393284529447556, "global_step": 213758, "epoch": 2401, "lr": 1.0775148587192684e-05}
{"train_loss": 0.07805303484201431, "global_step": 213759, "epoch": 2401, "lr": 1.077478907627572e-05}
{"train_loss": 0.06497856974601746, "global_step": 213760, "epoch": 2401, "lr": 1.0774429570632089e-05}
{"train_loss": 0.05408572405576706, "global_step": 213761, "epoch": 2401, "lr": 1.0774070070261871e-05}
{"train_loss": 0.14112627506256104, "global_step": 213762, "epoch": 2401, "lr": 1.077371057516509e-05}
{"train_loss": 0.08231490850448608, "global_step": 213763, "epoch": 2401, "lr": 1.077335108534182e-05}
{"train_loss": 0.06143563985824585, "global_step": 213764, "epoch": 2401, "lr": 1.0772991600792081e-05}
{"train_loss": 0.07823433727025986, "global_step": 213765, "epoch": 2401, "lr": 1.0772632121515946e-05}
{"train_loss": 0.0872332751750946, "global_step": 213766, "epoch": 2401, "lr": 1.0772272647513442e-05}
{"train_loss": 0.027034690603613853, "global_step": 213767, "epoch": 2401, "lr": 1.0771913178784637e-05}
{"train_loss": 0.06126710772514343, "global_step": 213768, "epoch": 2401, "lr": 1.0771553715329563e-05}
{"train_loss": 0.03549756109714508, "global_step": 213769, "epoch": 2401, "lr": 1.0771194257148281e-05}
{"train_loss": 0.04921332746744156, "global_step": 213770, "epoch": 2401, "lr": 1.0770834804240826e-05}
{"train_loss": 0.039236392825841904, "global_step": 213771, "epoch": 2401, "lr": 1.0770475356607269e-05}
{"train_loss": 0.06056658551096916, "global_step": 213772, "epoch": 2401, "lr": 1.0770115914247625e-05}
{"train_loss": 0.03861416503787041, "global_step": 213773, "epoch": 2401, "lr": 1.0769756477161963e-05}
{"train_loss": 0.02637200430035591, "global_step": 213774, "epoch": 2401, "lr": 1.0769397045350343e-05}
{"train_loss": 0.09158802777528763, "global_step": 213775, "epoch": 2401, "lr": 1.0769037618812784e-05}
{"train_loss": 0.04210234060883522, "global_step": 213776, "epoch": 2401, "lr": 1.0768678197549365e-05}
{"train_loss": 0.056093279669961235, "global_step": 213777, "epoch": 2401, "lr": 1.0768318781560105e-05, "val_loss": 8.719595909118652}
{"train_loss": 0.04763947054743767, "global_step": 213778, "epoch": 2402, "lr": 1.0767959370845082e-05}
{"train_loss": 0.07691174745559692, "global_step": 213779, "epoch": 2402, "lr": 1.0767599965404313e-05}
{"train_loss": 0.04525895044207573, "global_step": 213780, "epoch": 2402, "lr": 1.0767240565237874e-05}
{"train_loss": 0.05577918514609337, "global_step": 213781, "epoch": 2402, "lr": 1.0766881170345788e-05}
{"train_loss": 0.0704612284898758, "global_step": 213782, "epoch": 2402, "lr": 1.0766521780728133e-05}
{"train_loss": 0.0864952877163887, "global_step": 213783, "epoch": 2402, "lr": 1.0766162396384944e-05}
{"train_loss": 0.08646424859762192, "global_step": 213784, "epoch": 2402, "lr": 1.0765803017316245e-05}
{"train_loss": 0.05587725341320038, "global_step": 213785, "epoch": 2402, "lr": 1.0765443643522122e-05}
{"train_loss": 0.08784298598766327, "global_step": 213786, "epoch": 2402, "lr": 1.0765084275002596e-05}
{"train_loss": 0.05183925852179527, "global_step": 213787, "epoch": 2402, "lr": 1.076472491175774e-05}
{"train_loss": 0.10691840201616287, "global_step": 213788, "epoch": 2402, "lr": 1.0764365553787587e-05}
{"train_loss": 0.03813258931040764, "global_step": 213789, "epoch": 2402, "lr": 1.076400620109217e-05}
{"train_loss": 0.044265083968639374, "global_step": 213790, "epoch": 2402, "lr": 1.0763646853671572e-05}
{"train_loss": 0.03488527238368988, "global_step": 213791, "epoch": 2402, "lr": 1.0763287511525804e-05}
{"train_loss": 0.07364398241043091, "global_step": 213792, "epoch": 2402, "lr": 1.0762928174654941e-05}
{"train_loss": 0.0748319998383522, "global_step": 213793, "epoch": 2402, "lr": 1.0762568843059035e-05}
{"train_loss": 0.07268151640892029, "global_step": 213794, "epoch": 2402, "lr": 1.0762209516738109e-05}
{"train_loss": 0.044058360159397125, "global_step": 213795, "epoch": 2402, "lr": 1.0761850195692235e-05}
{"train_loss": 0.05671701580286026, "global_step": 213796, "epoch": 2402, "lr": 1.0761490879921449e-05}
{"train_loss": 0.051201216876506805, "global_step": 213797, "epoch": 2402, "lr": 1.076113156942581e-05}
{"train_loss": 0.045657817274332047, "global_step": 213798, "epoch": 2402, "lr": 1.0760772264205343e-05}
{"train_loss": 0.01055586151778698, "global_step": 213799, "epoch": 2402, "lr": 1.0760412964260124e-05}
{"train_loss": 0.0886150449514389, "global_step": 213800, "epoch": 2402, "lr": 1.076005366959018e-05}
{"train_loss": 0.08131495118141174, "global_step": 213801, "epoch": 2402, "lr": 1.0759694380195583e-05}
{"train_loss": 0.028717560693621635, "global_step": 213802, "epoch": 2402, "lr": 1.0759335096076351e-05}
{"train_loss": 0.07483257353305817, "global_step": 213803, "epoch": 2402, "lr": 1.0758975817232564e-05}
{"train_loss": 0.02455080859363079, "global_step": 213804, "epoch": 2402, "lr": 1.0758616543664236e-05}
{"train_loss": 0.02393193542957306, "global_step": 213805, "epoch": 2402, "lr": 1.0758257275371452e-05}
{"train_loss": 0.11811856925487518, "global_step": 213806, "epoch": 2402, "lr": 1.0757898012354228e-05}
{"train_loss": 0.08412414789199829, "global_step": 213807, "epoch": 2402, "lr": 1.0757538754612639e-05}
{"train_loss": 0.03542829304933548, "global_step": 213808, "epoch": 2402, "lr": 1.0757179502146703e-05}
{"train_loss": 0.05074974149465561, "global_step": 213809, "epoch": 2402, "lr": 1.075682025495649e-05}
{"train_loss": 0.04604802653193474, "global_step": 213810, "epoch": 2402, "lr": 1.0756461013042057e-05}
{"train_loss": 0.02059127390384674, "global_step": 213811, "epoch": 2402, "lr": 1.0756101776403426e-05}
{"train_loss": 0.03969699516892433, "global_step": 213812, "epoch": 2402, "lr": 1.0755742545040676e-05}
{"train_loss": 0.04120172932744026, "global_step": 213813, "epoch": 2402, "lr": 1.0755383318953821e-05}
{"train_loss": 0.06533171236515045, "global_step": 213814, "epoch": 2402, "lr": 1.0755024098142941e-05}
{"train_loss": 0.1225048303604126, "global_step": 213815, "epoch": 2402, "lr": 1.075466488260805e-05}
{"train_loss": 0.06087624654173851, "global_step": 213816, "epoch": 2402, "lr": 1.0754305672349236e-05}
{"train_loss": 0.03563912957906723, "global_step": 213817, "epoch": 2402, "lr": 1.0753946467366515e-05}
{"train_loss": 0.0414588525891304, "global_step": 213818, "epoch": 2402, "lr": 1.075358726765996e-05}
{"train_loss": 0.09454761445522308, "global_step": 213819, "epoch": 2402, "lr": 1.0753228073229587e-05}
{"train_loss": 0.047898899763822556, "global_step": 213820, "epoch": 2402, "lr": 1.0752868884075484e-05}
{"train_loss": 0.043983373790979385, "global_step": 213821, "epoch": 2402, "lr": 1.0752509700197677e-05}
{"train_loss": 0.06981099396944046, "global_step": 213822, "epoch": 2402, "lr": 1.0752150521596204e-05}
{"train_loss": 0.04542886093258858, "global_step": 213823, "epoch": 2402, "lr": 1.075179134827114e-05}
{"train_loss": 0.04428037628531456, "global_step": 213824, "epoch": 2402, "lr": 1.075143218022251e-05}
{"train_loss": 0.09866020828485489, "global_step": 213825, "epoch": 2402, "lr": 1.075107301745038e-05}
{"train_loss": 0.06533072143793106, "global_step": 213826, "epoch": 2402, "lr": 1.0750713859954786e-05}
{"train_loss": 0.034747008234262466, "global_step": 213827, "epoch": 2402, "lr": 1.075035470773577e-05}
{"train_loss": 0.11066730320453644, "global_step": 213828, "epoch": 2402, "lr": 1.0749995560793396e-05}
{"train_loss": 0.06628953665494919, "global_step": 213829, "epoch": 2402, "lr": 1.0749636419127713e-05}
{"train_loss": 0.05761969834566116, "global_step": 213830, "epoch": 2402, "lr": 1.0749277282738752e-05}
{"train_loss": 0.045421600341796875, "global_step": 213831, "epoch": 2402, "lr": 1.0748918151626586e-05}
{"train_loss": 0.05580669641494751, "global_step": 213832, "epoch": 2402, "lr": 1.0748559025791239e-05}
{"train_loss": 0.06147261708974838, "global_step": 213833, "epoch": 2402, "lr": 1.0748199905232781e-05}
{"train_loss": 0.03950436785817146, "global_step": 213834, "epoch": 2402, "lr": 1.074784078995123e-05}
{"train_loss": 0.03618929162621498, "global_step": 213835, "epoch": 2402, "lr": 1.0747481679946676e-05}
{"train_loss": 0.02081279642879963, "global_step": 213836, "epoch": 2402, "lr": 1.0747122575219127e-05}
{"train_loss": 0.09326677024364471, "global_step": 213837, "epoch": 2402, "lr": 1.0746763475768661e-05}
{"train_loss": 0.09515827149152756, "global_step": 213838, "epoch": 2402, "lr": 1.0746404381595304e-05}
{"train_loss": 0.10878065228462219, "global_step": 213839, "epoch": 2402, "lr": 1.0746045292699125e-05}
{"train_loss": 0.03137241303920746, "global_step": 213840, "epoch": 2402, "lr": 1.0745686209080147e-05}
{"train_loss": 0.017946777865290642, "global_step": 213841, "epoch": 2402, "lr": 1.0745327130738454e-05}
{"train_loss": 0.05392339825630188, "global_step": 213842, "epoch": 2402, "lr": 1.0744968057674049e-05}
{"train_loss": 0.06475936621427536, "global_step": 213843, "epoch": 2402, "lr": 1.0744608989887028e-05}
{"train_loss": 0.04121309146285057, "global_step": 213844, "epoch": 2402, "lr": 1.0744249927377397e-05}
{"train_loss": 0.05577971413731575, "global_step": 213845, "epoch": 2402, "lr": 1.0743890870145224e-05}
{"train_loss": 0.026796387508511543, "global_step": 213846, "epoch": 2402, "lr": 1.0743531818190572e-05}
{"train_loss": 0.07847357541322708, "global_step": 213847, "epoch": 2402, "lr": 1.0743172771513454e-05}
{"train_loss": 0.0723392590880394, "global_step": 213848, "epoch": 2402, "lr": 1.074281373011396e-05}
{"train_loss": 0.06028395891189575, "global_step": 213849, "epoch": 2402, "lr": 1.0742454693992098e-05}
{"train_loss": 0.04529760032892227, "global_step": 213850, "epoch": 2402, "lr": 1.074209566314795e-05}
{"train_loss": 0.06473980844020844, "global_step": 213851, "epoch": 2402, "lr": 1.074173663758153e-05}
{"train_loss": 0.0714091956615448, "global_step": 213852, "epoch": 2402, "lr": 1.0741377617292925e-05}
{"train_loss": 0.02396119013428688, "global_step": 213853, "epoch": 2402, "lr": 1.0741018602282144e-05}
{"train_loss": 0.05231456086039543, "global_step": 213854, "epoch": 2402, "lr": 1.0740659592549274e-05}
{"train_loss": 0.07703833281993866, "global_step": 213855, "epoch": 2402, "lr": 1.0740300588094326e-05}
{"train_loss": 0.059377532452344894, "global_step": 213856, "epoch": 2402, "lr": 1.0739941588917379e-05}
{"train_loss": 0.08295457065105438, "global_step": 213857, "epoch": 2402, "lr": 1.073958259501846e-05}
{"train_loss": 0.059097353368997574, "global_step": 213858, "epoch": 2402, "lr": 1.0739223606397636e-05}
{"train_loss": 0.04731529578566551, "global_step": 213859, "epoch": 2402, "lr": 1.0738864623054945e-05}
{"train_loss": 0.08007985353469849, "global_step": 213860, "epoch": 2402, "lr": 1.0738505644990422e-05}
{"train_loss": 0.06980898976325989, "global_step": 213861, "epoch": 2402, "lr": 1.0738146672204141e-05}
{"train_loss": 0.06850795447826385, "global_step": 213862, "epoch": 2402, "lr": 1.0737787704696123e-05}
{"train_loss": 0.01639558933675289, "global_step": 213863, "epoch": 2402, "lr": 1.0737428742466449e-05}
{"train_loss": 0.05133088305592537, "global_step": 213864, "epoch": 2402, "lr": 1.073706978551513e-05}
{"train_loss": 0.020151546224951744, "global_step": 213865, "epoch": 2402, "lr": 1.0736710833842251e-05}
{"train_loss": 0.05800314494565632, "global_step": 213866, "epoch": 2402, "lr": 1.0736351887447827e-05, "val_loss": 8.865200996398926}
{"train_loss": 0.026738470420241356, "global_step": 213867, "epoch": 2403, "lr": 1.0735992946331936e-05}
{"train_loss": 0.11512184143066406, "global_step": 213868, "epoch": 2403, "lr": 1.07356340104946e-05}
{"train_loss": 0.03206070140004158, "global_step": 213869, "epoch": 2403, "lr": 1.0735275079935892e-05}
{"train_loss": 0.05000605434179306, "global_step": 213870, "epoch": 2403, "lr": 1.0734916154655832e-05}
{"train_loss": 0.07418396323919296, "global_step": 213871, "epoch": 2403, "lr": 1.0734557234654503e-05}
{"train_loss": 0.04592379182577133, "global_step": 213872, "epoch": 2403, "lr": 1.0734198319931916e-05}
{"train_loss": 0.11795251816511154, "global_step": 213873, "epoch": 2403, "lr": 1.073383941048815e-05}
{"train_loss": 0.08466649800539017, "global_step": 213874, "epoch": 2403, "lr": 1.073348050632323e-05}
{"train_loss": 0.07030370831489563, "global_step": 213875, "epoch": 2403, "lr": 1.0733121607437225e-05}
{"train_loss": 0.10828182101249695, "global_step": 213876, "epoch": 2403, "lr": 1.0732762713830163e-05}
{"train_loss": 0.026463527232408524, "global_step": 213877, "epoch": 2403, "lr": 1.073240382550212e-05}
{"train_loss": 0.0655190721154213, "global_step": 213878, "epoch": 2403, "lr": 1.0732044942453106e-05}
{"train_loss": 0.055746614933013916, "global_step": 213879, "epoch": 2403, "lr": 1.0731686064683204e-05}
{"train_loss": 0.03636659309267998, "global_step": 213880, "epoch": 2403, "lr": 1.0731327192192437e-05}
{"train_loss": 0.03780636191368103, "global_step": 213881, "epoch": 2403, "lr": 1.0730968324980867e-05}
{"train_loss": 0.06290824711322784, "global_step": 213882, "epoch": 2403, "lr": 1.0730609463048546e-05}
{"train_loss": 0.07064713537693024, "global_step": 213883, "epoch": 2403, "lr": 1.073025060639551e-05}
{"train_loss": 0.07740908861160278, "global_step": 213884, "epoch": 2403, "lr": 1.072989175502182e-05}
{"train_loss": 0.03450099006295204, "global_step": 213885, "epoch": 2403, "lr": 1.0729532908927509e-05}
{"train_loss": 0.03960281237959862, "global_step": 213886, "epoch": 2403, "lr": 1.0729174068112646e-05}
{"train_loss": 0.049812331795692444, "global_step": 213887, "epoch": 2403, "lr": 1.0728815232577255e-05}
{"train_loss": 0.0421326719224453, "global_step": 213888, "epoch": 2403, "lr": 1.0728456402321407e-05}
{"train_loss": 0.09479176998138428, "global_step": 213889, "epoch": 2403, "lr": 1.0728097577345126e-05}
{"train_loss": 0.018224777653813362, "global_step": 213890, "epoch": 2403, "lr": 1.0727738757648487e-05}
{"train_loss": 0.07346706092357635, "global_step": 213891, "epoch": 2403, "lr": 1.0727379943231513e-05}
{"train_loss": 0.053871773183345795, "global_step": 213892, "epoch": 2403, "lr": 1.0727021134094272e-05}
{"train_loss": 0.05233238637447357, "global_step": 213893, "epoch": 2403, "lr": 1.0726662330236797e-05}
{"train_loss": 0.08504214882850647, "global_step": 213894, "epoch": 2403, "lr": 1.0726303531659155e-05}
{"train_loss": 0.04801982641220093, "global_step": 213895, "epoch": 2403, "lr": 1.0725944738361372e-05}
{"train_loss": 0.0575401596724987, "global_step": 213896, "epoch": 2403, "lr": 1.072558595034352e-05}
{"train_loss": 0.0657779648900032, "global_step": 213897, "epoch": 2403, "lr": 1.0725227167605617e-05}
{"train_loss": 0.055869001895189285, "global_step": 213898, "epoch": 2403, "lr": 1.0724868390147747e-05}
{"train_loss": 0.022229067981243134, "global_step": 213899, "epoch": 2403, "lr": 1.0724509617969936e-05}
{"train_loss": 0.050907477736473083, "global_step": 213900, "epoch": 2403, "lr": 1.0724150851072223e-05}
{"train_loss": 0.05289524793624878, "global_step": 213901, "epoch": 2403, "lr": 1.0723792089454682e-05}
{"train_loss": 0.05251891165971756, "global_step": 213902, "epoch": 2403, "lr": 1.0723433333117333e-05}
{"train_loss": 0.059072282165288925, "global_step": 213903, "epoch": 2403, "lr": 1.0723074582060262e-05}
{"train_loss": 0.10948387533426285, "global_step": 213904, "epoch": 2403, "lr": 1.072271583628347e-05}
{"train_loss": 0.03406287729740143, "global_step": 213905, "epoch": 2403, "lr": 1.072235709578705e-05}
{"train_loss": 0.045870330184698105, "global_step": 213906, "epoch": 2403, "lr": 1.0721998360571017e-05}
{"train_loss": 0.04530995339155197, "global_step": 213907, "epoch": 2403, "lr": 1.0721639630635444e-05}
{"train_loss": 0.04522007331252098, "global_step": 213908, "epoch": 2403, "lr": 1.0721280905980357e-05}
{"train_loss": 0.04224291443824768, "global_step": 213909, "epoch": 2403, "lr": 1.0720922186605825e-05}
{"train_loss": 0.023923106491565704, "global_step": 213910, "epoch": 2403, "lr": 1.0720563472511874e-05}
{"train_loss": 0.03114870749413967, "global_step": 213911, "epoch": 2403, "lr": 1.0720204763698582e-05}
{"train_loss": 0.05394180491566658, "global_step": 213912, "epoch": 2403, "lr": 1.0719846060165961e-05}
{"train_loss": 0.03640105947852135, "global_step": 213913, "epoch": 2403, "lr": 1.0719487361914094e-05}
{"train_loss": 0.08464832603931427, "global_step": 213914, "epoch": 2403, "lr": 1.0719128668943002e-05}
{"train_loss": 0.049685802310705185, "global_step": 213915, "epoch": 2403, "lr": 1.0718769981252751e-05}
{"train_loss": 0.057110220193862915, "global_step": 213916, "epoch": 2403, "lr": 1.0718411298843372e-05}
{"train_loss": 0.032108329236507416, "global_step": 213917, "epoch": 2403, "lr": 1.0718052621714941e-05}
{"train_loss": 0.061746787279844284, "global_step": 213918, "epoch": 2403, "lr": 1.0717693949867475e-05}
{"train_loss": 0.035233091562986374, "global_step": 213919, "epoch": 2403, "lr": 1.0717335283301034e-05}
{"train_loss": 0.057217150926589966, "global_step": 213920, "epoch": 2403, "lr": 1.0716976622015679e-05}
{"train_loss": 0.05050604045391083, "global_step": 213921, "epoch": 2403, "lr": 1.0716617966011439e-05}
{"train_loss": 0.03434846177697182, "global_step": 213922, "epoch": 2403, "lr": 1.0716259315288384e-05}
{"train_loss": 0.03994552046060562, "global_step": 213923, "epoch": 2403, "lr": 1.0715900669846535e-05}
{"train_loss": 0.1317550241947174, "global_step": 213924, "epoch": 2403, "lr": 1.0715542029685965e-05}
{"train_loss": 0.0838407576084137, "global_step": 213925, "epoch": 2403, "lr": 1.0715183394806699e-05}
{"train_loss": 0.031160993501544, "global_step": 213926, "epoch": 2403, "lr": 1.0714824765208814e-05}
{"train_loss": 0.05222642794251442, "global_step": 213927, "epoch": 2403, "lr": 1.0714466140892326e-05}
{"train_loss": 0.08599647879600525, "global_step": 213928, "epoch": 2403, "lr": 1.0714107521857314e-05}
{"train_loss": 0.06692952662706375, "global_step": 213929, "epoch": 2403, "lr": 1.07137489081038e-05}
{"train_loss": 0.04093644767999649, "global_step": 213930, "epoch": 2403, "lr": 1.0713390299631853e-05}
{"train_loss": 0.12444816529750824, "global_step": 213931, "epoch": 2403, "lr": 1.07130316964415e-05}
{"train_loss": 0.06067511811852455, "global_step": 213932, "epoch": 2403, "lr": 1.0712673098532816e-05}
{"train_loss": 0.05504612997174263, "global_step": 213933, "epoch": 2403, "lr": 1.0712314505905819e-05}
{"train_loss": 0.06322772800922394, "global_step": 213934, "epoch": 2403, "lr": 1.0711955918560584e-05}
{"train_loss": 0.06376592814922333, "global_step": 213935, "epoch": 2403, "lr": 1.0711597336497143e-05}
{"train_loss": 0.0208877120167017, "global_step": 213936, "epoch": 2403, "lr": 1.0711238759715552e-05}
{"train_loss": 0.07152789831161499, "global_step": 213937, "epoch": 2403, "lr": 1.071088018821586e-05}
{"train_loss": 0.04806268587708473, "global_step": 213938, "epoch": 2403, "lr": 1.0710521621998098e-05}
{"train_loss": 0.06957278400659561, "global_step": 213939, "epoch": 2403, "lr": 1.071016306106234e-05}
{"train_loss": 0.04090023413300514, "global_step": 213940, "epoch": 2403, "lr": 1.0709804505408611e-05}
{"train_loss": 0.07523905485868454, "global_step": 213941, "epoch": 2403, "lr": 1.0709445955036984e-05}
{"train_loss": 0.042757630348205566, "global_step": 213942, "epoch": 2403, "lr": 1.0709087409947477e-05}
{"train_loss": 0.07209798693656921, "global_step": 213943, "epoch": 2403, "lr": 1.0708728870140167e-05}
{"train_loss": 0.0827108770608902, "global_step": 213944, "epoch": 2403, "lr": 1.0708370335615076e-05}
{"train_loss": 0.03429531306028366, "global_step": 213945, "epoch": 2403, "lr": 1.070801180637228e-05}
{"train_loss": 0.06732255965471268, "global_step": 213946, "epoch": 2403, "lr": 1.07076532824118e-05}
{"train_loss": 0.0376112274825573, "global_step": 213947, "epoch": 2403, "lr": 1.0707294763733711e-05}
{"train_loss": 0.056511614471673965, "global_step": 213948, "epoch": 2403, "lr": 1.0706936250338029e-05}
{"train_loss": 0.02167978323996067, "global_step": 213949, "epoch": 2403, "lr": 1.0706577742224838e-05}
{"train_loss": 0.10744263976812363, "global_step": 213950, "epoch": 2403, "lr": 1.0706219239394155e-05}
{"train_loss": 0.04956076294183731, "global_step": 213951, "epoch": 2403, "lr": 1.0705860741846057e-05}
{"train_loss": 0.041450757533311844, "global_step": 213952, "epoch": 2403, "lr": 1.0705502249580562e-05}
{"train_loss": 0.06295975297689438, "global_step": 213953, "epoch": 2403, "lr": 1.0705143762597746e-05}
{"train_loss": 0.05357075855135918, "global_step": 213954, "epoch": 2403, "lr": 1.0704785280897633e-05}
{"train_loss": 0.057354632304625563, "global_step": 213955, "epoch": 2403, "lr": 1.0704426804480284e-05, "val_loss": 8.843094825744629}
{"train_loss": 0.022482579573988914, "global_step": 213956, "epoch": 2404, "lr": 1.0704068333345756e-05}
{"train_loss": 0.06665272265672684, "global_step": 213957, "epoch": 2404, "lr": 1.0703709867494072e-05}
{"train_loss": 0.04176773130893707, "global_step": 213958, "epoch": 2404, "lr": 1.0703351406925317e-05}
{"train_loss": 0.06845498085021973, "global_step": 213959, "epoch": 2404, "lr": 1.0702992951639496e-05}
{"train_loss": 0.04336559772491455, "global_step": 213960, "epoch": 2404, "lr": 1.0702634501636694e-05}
{"train_loss": 0.03111342340707779, "global_step": 213961, "epoch": 2404, "lr": 1.0702276056916933e-05}
{"train_loss": 0.10055024176836014, "global_step": 213962, "epoch": 2404, "lr": 1.0701917617480278e-05}
{"train_loss": 0.04709790647029877, "global_step": 213963, "epoch": 2404, "lr": 1.0701559183326765e-05}
{"train_loss": 0.04885879158973694, "global_step": 213964, "epoch": 2404, "lr": 1.0701200754456458e-05}
{"train_loss": 0.044080980122089386, "global_step": 213965, "epoch": 2404, "lr": 1.0700842330869388e-05}
{"train_loss": 0.05189891532063484, "global_step": 213966, "epoch": 2404, "lr": 1.0700483912565617e-05}
{"train_loss": 0.059303998947143555, "global_step": 213967, "epoch": 2404, "lr": 1.0700125499545176e-05}
{"train_loss": 0.09106596559286118, "global_step": 213968, "epoch": 2404, "lr": 1.0699767091808138e-05}
{"train_loss": 0.046066269278526306, "global_step": 213969, "epoch": 2404, "lr": 1.0699408689354523e-05}
{"train_loss": 0.031456757336854935, "global_step": 213970, "epoch": 2404, "lr": 1.0699050292184409e-05}
{"train_loss": 0.0553240105509758, "global_step": 213971, "epoch": 2404, "lr": 1.0698691900297813e-05}
{"train_loss": 0.03321879729628563, "global_step": 213972, "epoch": 2404, "lr": 1.0698333513694814e-05}
{"train_loss": 0.03963784873485565, "global_step": 213973, "epoch": 2404, "lr": 1.0697975132375432e-05}
{"train_loss": 0.08845911175012589, "global_step": 213974, "epoch": 2404, "lr": 1.069761675633974e-05}
{"train_loss": 0.08351106941699982, "global_step": 213975, "epoch": 2404, "lr": 1.0697258385587777e-05}
{"train_loss": 0.10434356331825256, "global_step": 213976, "epoch": 2404, "lr": 1.0696900020119571e-05}
{"train_loss": 0.09114430099725723, "global_step": 213977, "epoch": 2404, "lr": 1.0696541659935206e-05}
{"train_loss": 0.10809168219566345, "global_step": 213978, "epoch": 2404, "lr": 1.0696183305034695e-05}
{"train_loss": 0.07661072909832001, "global_step": 213979, "epoch": 2404, "lr": 1.0695824955418122e-05}
{"train_loss": 0.06254646182060242, "global_step": 213980, "epoch": 2404, "lr": 1.0695466611085497e-05}
{"train_loss": 0.06434950232505798, "global_step": 213981, "epoch": 2404, "lr": 1.0695108272036902e-05}
{"train_loss": 0.0318295992910862, "global_step": 213982, "epoch": 2404, "lr": 1.0694749938272358e-05}
{"train_loss": 0.04530133306980133, "global_step": 213983, "epoch": 2404, "lr": 1.0694391609791937e-05}
{"train_loss": 0.0878717377781868, "global_step": 213984, "epoch": 2404, "lr": 1.0694033286595661e-05}
{"train_loss": 0.07052277773618698, "global_step": 213985, "epoch": 2404, "lr": 1.0693674968683615e-05}
{"train_loss": 0.0541263073682785, "global_step": 213986, "epoch": 2404, "lr": 1.0693316656055802e-05}
{"train_loss": 0.046366963535547256, "global_step": 213987, "epoch": 2404, "lr": 1.0692958348712312e-05}
{"train_loss": 0.05490822345018387, "global_step": 213988, "epoch": 2404, "lr": 1.0692600046653162e-05}
{"train_loss": 0.05300559476017952, "global_step": 213989, "epoch": 2404, "lr": 1.0692241749878429e-05}
{"train_loss": 0.06721769273281097, "global_step": 213990, "epoch": 2404, "lr": 1.0691883458388124e-05}
{"train_loss": 0.10438309609889984, "global_step": 213991, "epoch": 2404, "lr": 1.069152517218232e-05}
{"train_loss": 0.06092388182878494, "global_step": 213992, "epoch": 2404, "lr": 1.0691166891261073e-05}
{"train_loss": 0.06487271934747696, "global_step": 213993, "epoch": 2404, "lr": 1.069080861562441e-05}
{"train_loss": 0.07994270324707031, "global_step": 213994, "epoch": 2404, "lr": 1.0690450345272402e-05}
{"train_loss": 0.022709732875227928, "global_step": 213995, "epoch": 2404, "lr": 1.0690092080205066e-05}
{"train_loss": 0.024019712582230568, "global_step": 213996, "epoch": 2404, "lr": 1.0689733820422481e-05}
{"train_loss": 0.05966973304748535, "global_step": 213997, "epoch": 2404, "lr": 1.0689375565924675e-05}
{"train_loss": 0.028050199151039124, "global_step": 213998, "epoch": 2404, "lr": 1.0689017316711713e-05}
{"train_loss": 0.03218318521976471, "global_step": 213999, "epoch": 2404, "lr": 1.068865907278362e-05}
{"train_loss": 0.06510741263628006, "global_step": 214000, "epoch": 2404, "lr": 1.068830083414047e-05}
{"train_loss": 0.04463939741253853, "global_step": 214001, "epoch": 2404, "lr": 1.0687942600782285e-05}
{"train_loss": 0.026854215189814568, "global_step": 214002, "epoch": 2404, "lr": 1.0687584372709147e-05}
{"train_loss": 0.06991662085056305, "global_step": 214003, "epoch": 2404, "lr": 1.0687226149921065e-05}
{"train_loss": 0.05590758100152016, "global_step": 214004, "epoch": 2404, "lr": 1.0686867932418121e-05}
{"train_loss": 0.05432644858956337, "global_step": 214005, "epoch": 2404, "lr": 1.0686509720200332e-05}
{"train_loss": 0.043070126324892044, "global_step": 214006, "epoch": 2404, "lr": 1.0686151513267779e-05}
{"train_loss": 0.03744065761566162, "global_step": 214007, "epoch": 2404, "lr": 1.0685793311620478e-05}
{"train_loss": 0.07586804777383804, "global_step": 214008, "epoch": 2404, "lr": 1.0685435115258496e-05}
{"train_loss": 0.03407050669193268, "global_step": 214009, "epoch": 2404, "lr": 1.0685076924181892e-05}
{"train_loss": 0.059900764375925064, "global_step": 214010, "epoch": 2404, "lr": 1.0684718738390698e-05}
{"train_loss": 0.020597312599420547, "global_step": 214011, "epoch": 2404, "lr": 1.0684360557884953e-05}
{"train_loss": 0.05444314330816269, "global_step": 214012, "epoch": 2404, "lr": 1.0684002382664727e-05}
{"train_loss": 0.05367559194564819, "global_step": 214013, "epoch": 2404, "lr": 1.0683644212730044e-05}
{"train_loss": 0.056727707386016846, "global_step": 214014, "epoch": 2404, "lr": 1.0683286048080981e-05}
{"train_loss": 0.034446898847818375, "global_step": 214015, "epoch": 2404, "lr": 1.0682927888717576e-05}
{"train_loss": 0.04795420169830322, "global_step": 214016, "epoch": 2404, "lr": 1.0682569734639853e-05}
{"train_loss": 0.026287442073225975, "global_step": 214017, "epoch": 2404, "lr": 1.068221158584789e-05}
{"train_loss": 0.0336451381444931, "global_step": 214018, "epoch": 2404, "lr": 1.0681853442341716e-05}
{"train_loss": 0.051947616040706635, "global_step": 214019, "epoch": 2404, "lr": 1.0681495304121397e-05}
{"train_loss": 0.03170070797204971, "global_step": 214020, "epoch": 2404, "lr": 1.0681137171186961e-05}
{"train_loss": 0.05132424458861351, "global_step": 214021, "epoch": 2404, "lr": 1.068077904353848e-05}
{"train_loss": 0.09033630043268204, "global_step": 214022, "epoch": 2404, "lr": 1.0680420921175977e-05}
{"train_loss": 0.04205843061208725, "global_step": 214023, "epoch": 2404, "lr": 1.0680062804099523e-05}
{"train_loss": 0.06439989805221558, "global_step": 214024, "epoch": 2404, "lr": 1.0679704692309145e-05}
{"train_loss": 0.08131859451532364, "global_step": 214025, "epoch": 2404, "lr": 1.0679346585804912e-05}
{"train_loss": 0.059815846383571625, "global_step": 214026, "epoch": 2404, "lr": 1.0678988484586849e-05}
{"train_loss": 0.030140897259116173, "global_step": 214027, "epoch": 2404, "lr": 1.0678630388655015e-05}
{"train_loss": 0.04911106824874878, "global_step": 214028, "epoch": 2404, "lr": 1.0678272298009478e-05}
{"train_loss": 0.05104055255651474, "global_step": 214029, "epoch": 2404, "lr": 1.0677914212650252e-05}
{"train_loss": 0.042357996106147766, "global_step": 214030, "epoch": 2404, "lr": 1.0677556132577416e-05}
{"train_loss": 0.018467461690306664, "global_step": 214031, "epoch": 2404, "lr": 1.0677198057790987e-05}
{"train_loss": 0.058161135762929916, "global_step": 214032, "epoch": 2404, "lr": 1.0676839988291044e-05}
{"train_loss": 0.021824447438120842, "global_step": 214033, "epoch": 2404, "lr": 1.0676481924077609e-05}
{"train_loss": 0.09080272167921066, "global_step": 214034, "epoch": 2404, "lr": 1.067612386515075e-05}
{"train_loss": 0.07091190665960312, "global_step": 214035, "epoch": 2404, "lr": 1.0675765811510496e-05}
{"train_loss": 0.059896115213632584, "global_step": 214036, "epoch": 2404, "lr": 1.067540776315692e-05}
{"train_loss": 0.06816216558218002, "global_step": 214037, "epoch": 2404, "lr": 1.0675049720090047e-05}
{"train_loss": 0.060109738260507584, "global_step": 214038, "epoch": 2404, "lr": 1.067469168230994e-05}
{"train_loss": 0.08128169924020767, "global_step": 214039, "epoch": 2404, "lr": 1.0674333649816632e-05}
{"train_loss": 0.06082235276699066, "global_step": 214040, "epoch": 2404, "lr": 1.0673975622610194e-05}
{"train_loss": 0.04186633974313736, "global_step": 214041, "epoch": 2404, "lr": 1.067361760069065e-05}
{"train_loss": 0.0691438838839531, "global_step": 214042, "epoch": 2404, "lr": 1.0673259584058065e-05}
{"train_loss": 0.056000664830207825, "global_step": 214043, "epoch": 2404, "lr": 1.0672901572712474e-05}
{"train_loss": 0.0552085353208057, "global_step": 214044, "epoch": 2404, "lr": 1.0672543566653941e-05, "val_loss": 8.897992134094238}
{"train_loss": 0.07009314745664597, "global_step": 214045, "epoch": 2405, "lr": 1.0672185565882492e-05}
{"train_loss": 0.0435284823179245, "global_step": 214046, "epoch": 2405, "lr": 1.0671827570398191e-05}
{"train_loss": 0.031784530729055405, "global_step": 214047, "epoch": 2405, "lr": 1.0671469580201099e-05}
{"train_loss": 0.06957483291625977, "global_step": 214048, "epoch": 2405, "lr": 1.0671111595291244e-05}
{"train_loss": 0.09537987411022186, "global_step": 214049, "epoch": 2405, "lr": 1.0670753615668667e-05}
{"train_loss": 0.03605359420180321, "global_step": 214050, "epoch": 2405, "lr": 1.0670395641333442e-05}
{"train_loss": 0.06652545183897018, "global_step": 214051, "epoch": 2405, "lr": 1.0670037672285587e-05}
{"train_loss": 0.09619301557540894, "global_step": 214052, "epoch": 2405, "lr": 1.0669679708525183e-05}
{"train_loss": 0.05678600072860718, "global_step": 214053, "epoch": 2405, "lr": 1.0669321750052258e-05}
{"train_loss": 0.043693460524082184, "global_step": 214054, "epoch": 2405, "lr": 1.066896379686685e-05}
{"train_loss": 0.06538672745227814, "global_step": 214055, "epoch": 2405, "lr": 1.0668605848969038e-05}
{"train_loss": 0.06915251165628433, "global_step": 214056, "epoch": 2405, "lr": 1.0668247906358836e-05}
{"train_loss": 0.06157131865620613, "global_step": 214057, "epoch": 2405, "lr": 1.066788996903632e-05}
{"train_loss": 0.04508441314101219, "global_step": 214058, "epoch": 2405, "lr": 1.0667532037001514e-05}
{"train_loss": 0.11140529811382294, "global_step": 214059, "epoch": 2405, "lr": 1.0667174110254497e-05}
{"train_loss": 0.035445135086774826, "global_step": 214060, "epoch": 2405, "lr": 1.0666816188795282e-05}
{"train_loss": 0.02941022999584675, "global_step": 214061, "epoch": 2405, "lr": 1.0666458272623948e-05}
{"train_loss": 0.051511555910110474, "global_step": 214062, "epoch": 2405, "lr": 1.0666100361740517e-05}
{"train_loss": 0.05228199437260628, "global_step": 214063, "epoch": 2405, "lr": 1.0665742456145045e-05}
{"train_loss": 0.05878850072622299, "global_step": 214064, "epoch": 2405, "lr": 1.0665384555837605e-05}
{"train_loss": 0.07928062230348587, "global_step": 214065, "epoch": 2405, "lr": 1.0665026660818205e-05}
{"train_loss": 0.07664402574300766, "global_step": 214066, "epoch": 2405, "lr": 1.0664668771086927e-05}
{"train_loss": 0.05825043097138405, "global_step": 214067, "epoch": 2405, "lr": 1.0664310886643797e-05}
{"train_loss": 0.031337931752204895, "global_step": 214068, "epoch": 2405, "lr": 1.066395300748888e-05}
{"train_loss": 0.043533869087696075, "global_step": 214069, "epoch": 2405, "lr": 1.06635951336222e-05}
{"train_loss": 0.09320138394832611, "global_step": 214070, "epoch": 2405, "lr": 1.0663237265043835e-05}
{"train_loss": 0.08635468780994415, "global_step": 214071, "epoch": 2405, "lr": 1.0662879401753805e-05}
{"train_loss": 0.02901633270084858, "global_step": 214072, "epoch": 2405, "lr": 1.0662521543752191e-05}
{"train_loss": 0.052294597029685974, "global_step": 214073, "epoch": 2405, "lr": 1.0662163691039e-05}
{"train_loss": 0.0718211829662323, "global_step": 214074, "epoch": 2405, "lr": 1.066180584361432e-05}
{"train_loss": 0.0765807256102562, "global_step": 214075, "epoch": 2405, "lr": 1.0661448001478163e-05}
{"train_loss": 0.07460901886224747, "global_step": 214076, "epoch": 2405, "lr": 1.066109016463061e-05}
{"train_loss": 0.05153903365135193, "global_step": 214077, "epoch": 2405, "lr": 1.0660732333071682e-05}
{"train_loss": 0.06265682727098465, "global_step": 214078, "epoch": 2405, "lr": 1.0660374506801452e-05}
{"train_loss": 0.035051487386226654, "global_step": 214079, "epoch": 2405, "lr": 1.066001668581994e-05}
{"train_loss": 0.08276166021823883, "global_step": 214080, "epoch": 2405, "lr": 1.0659658870127227e-05}
{"train_loss": 0.04341772571206093, "global_step": 214081, "epoch": 2405, "lr": 1.065930105972332e-05}
{"train_loss": 0.08479934185743332, "global_step": 214082, "epoch": 2405, "lr": 1.06589432546083e-05}
{"train_loss": 0.05386665463447571, "global_step": 214083, "epoch": 2405, "lr": 1.0658585454782215e-05}
{"train_loss": 0.0536758191883564, "global_step": 214084, "epoch": 2405, "lr": 1.065822766024509e-05}
{"train_loss": 0.06355500966310501, "global_step": 214085, "epoch": 2405, "lr": 1.0657869870997005e-05}
{"train_loss": 0.0746229812502861, "global_step": 214086, "epoch": 2405, "lr": 1.0657512087037986e-05}
{"train_loss": 0.06942330300807953, "global_step": 214087, "epoch": 2405, "lr": 1.0657154308368073e-05}
{"train_loss": 0.043390944600105286, "global_step": 214088, "epoch": 2405, "lr": 1.0656796534987334e-05}
{"train_loss": 0.08478132635354996, "global_step": 214089, "epoch": 2405, "lr": 1.0656438766895805e-05}
{"train_loss": 0.07444418221712112, "global_step": 214090, "epoch": 2405, "lr": 1.0656081004093543e-05}
{"train_loss": 0.06879397481679916, "global_step": 214091, "epoch": 2405, "lr": 1.0655723246580584e-05}
{"train_loss": 0.007318525109440088, "global_step": 214092, "epoch": 2405, "lr": 1.0655365494356994e-05}
{"train_loss": 0.056969426572322845, "global_step": 214093, "epoch": 2405, "lr": 1.0655007747422812e-05}
{"train_loss": 0.09794794023036957, "global_step": 214094, "epoch": 2405, "lr": 1.0654650005778072e-05}
{"train_loss": 0.04755016788840294, "global_step": 214095, "epoch": 2405, "lr": 1.0654292269422844e-05}
{"train_loss": 0.06859929114580154, "global_step": 214096, "epoch": 2405, "lr": 1.0653934538357158e-05}
{"train_loss": 0.04893181473016739, "global_step": 214097, "epoch": 2405, "lr": 1.0653576812581085e-05}
{"train_loss": 0.07720475643873215, "global_step": 214098, "epoch": 2405, "lr": 1.065321909209464e-05}
{"train_loss": 0.056956518441438675, "global_step": 214099, "epoch": 2405, "lr": 1.0652861376897894e-05}
{"train_loss": 0.055240098387002945, "global_step": 214100, "epoch": 2405, "lr": 1.0652503666990904e-05}
{"train_loss": 0.07185155153274536, "global_step": 214101, "epoch": 2405, "lr": 1.0652145962373694e-05}
{"train_loss": 0.07333488762378693, "global_step": 214102, "epoch": 2405, "lr": 1.065178826304633e-05}
{"train_loss": 0.05867992714047432, "global_step": 214103, "epoch": 2405, "lr": 1.0651430569008846e-05}
{"train_loss": 0.07118108868598938, "global_step": 214104, "epoch": 2405, "lr": 1.065107288026131e-05}
{"train_loss": 0.04976382851600647, "global_step": 214105, "epoch": 2405, "lr": 1.0650715196803745e-05}
{"train_loss": 0.07560468465089798, "global_step": 214106, "epoch": 2405, "lr": 1.0650357518636223e-05}
{"train_loss": 0.02526988834142685, "global_step": 214107, "epoch": 2405, "lr": 1.0649999845758768e-05}
{"train_loss": 0.03303917124867439, "global_step": 214108, "epoch": 2405, "lr": 1.064964217817146e-05}
{"train_loss": 0.08285773545503616, "global_step": 214109, "epoch": 2405, "lr": 1.0649284515874308e-05}
{"train_loss": 0.045070018619298935, "global_step": 214110, "epoch": 2405, "lr": 1.0648926858867392e-05}
{"train_loss": 0.11767356097698212, "global_step": 214111, "epoch": 2405, "lr": 1.0648569207150739e-05}
{"train_loss": 0.05286554619669914, "global_step": 214112, "epoch": 2405, "lr": 1.064821156072442e-05}
{"train_loss": 0.11614904552698135, "global_step": 214113, "epoch": 2405, "lr": 1.0647853919588451e-05}
{"train_loss": 0.07860799878835678, "global_step": 214114, "epoch": 2405, "lr": 1.0647496283742913e-05}
{"train_loss": 0.0992879569530487, "global_step": 214115, "epoch": 2405, "lr": 1.0647138653187833e-05}
{"train_loss": 0.03714089095592499, "global_step": 214116, "epoch": 2405, "lr": 1.0646781027923275e-05}
{"train_loss": 0.031591594219207764, "global_step": 214117, "epoch": 2405, "lr": 1.0646423407949263e-05}
{"train_loss": 0.05531095713376999, "global_step": 214118, "epoch": 2405, "lr": 1.0646065793265858e-05}
{"train_loss": 0.03708953782916069, "global_step": 214119, "epoch": 2405, "lr": 1.0645708183873127e-05}
{"train_loss": 0.016278719529509544, "global_step": 214120, "epoch": 2405, "lr": 1.0645350579771086e-05}
{"train_loss": 0.09374630451202393, "global_step": 214121, "epoch": 2405, "lr": 1.0644992980959812e-05}
{"train_loss": 0.0442565456032753, "global_step": 214122, "epoch": 2405, "lr": 1.0644635387439322e-05}
{"train_loss": 0.058256275951862335, "global_step": 214123, "epoch": 2405, "lr": 1.0644277799209695e-05}
{"train_loss": 0.04098520055413246, "global_step": 214124, "epoch": 2405, "lr": 1.0643920216270959e-05}
{"train_loss": 0.07666909694671631, "global_step": 214125, "epoch": 2405, "lr": 1.0643562638623173e-05}
{"train_loss": 0.06215506047010422, "global_step": 214126, "epoch": 2405, "lr": 1.0643205066266381e-05}
{"train_loss": 0.05245188996195793, "global_step": 214127, "epoch": 2405, "lr": 1.0642847499200625e-05}
{"train_loss": 0.028207950294017792, "global_step": 214128, "epoch": 2405, "lr": 1.0642489937425964e-05}
{"train_loss": 0.09804797172546387, "global_step": 214129, "epoch": 2405, "lr": 1.0642132380942433e-05}
{"train_loss": 0.025609558448195457, "global_step": 214130, "epoch": 2405, "lr": 1.0641774829750095e-05}
{"train_loss": 0.053838104009628296, "global_step": 214131, "epoch": 2405, "lr": 1.0641417283848992e-05}
{"train_loss": 0.03086048550903797, "global_step": 214132, "epoch": 2405, "lr": 1.0641059743239163e-05}
{"train_loss": 0.06059456026976865, "global_step": 214133, "epoch": 2405, "lr": 1.0640702207920671e-05, "val_loss": 8.792997360229492, "train_action_mse_error": 7.2241034507751465}
{"train_loss": 0.034750260412693024, "global_step": 214134, "epoch": 2406, "lr": 1.0640344677893543e-05}
{"train_loss": 0.049492768943309784, "global_step": 214135, "epoch": 2406, "lr": 1.0639987153157849e-05}
{"train_loss": 0.06449344754219055, "global_step": 214136, "epoch": 2406, "lr": 1.0639629633713632e-05}
{"train_loss": 0.05849885568022728, "global_step": 214137, "epoch": 2406, "lr": 1.0639272119560933e-05}
{"train_loss": 0.03180476278066635, "global_step": 214138, "epoch": 2406, "lr": 1.0638914610699812e-05}
{"train_loss": 0.10080402344465256, "global_step": 214139, "epoch": 2406, "lr": 1.0638557107130299e-05}
{"train_loss": 0.11839497089385986, "global_step": 214140, "epoch": 2406, "lr": 1.0638199608852461e-05}
{"train_loss": 0.03696947172284126, "global_step": 214141, "epoch": 2406, "lr": 1.0637842115866326e-05}
{"train_loss": 0.06949861347675323, "global_step": 214142, "epoch": 2406, "lr": 1.0637484628171967e-05}
{"train_loss": 0.04920657351613045, "global_step": 214143, "epoch": 2406, "lr": 1.0637127145769404e-05}
{"train_loss": 0.07993597537279129, "global_step": 214144, "epoch": 2406, "lr": 1.0636769668658713e-05}
{"train_loss": 0.11928246170282364, "global_step": 214145, "epoch": 2406, "lr": 1.0636412196839918e-05}
{"train_loss": 0.0656294897198677, "global_step": 214146, "epoch": 2406, "lr": 1.0636054730313088e-05}
{"train_loss": 0.03017924167215824, "global_step": 214147, "epoch": 2406, "lr": 1.0635697269078248e-05}
{"train_loss": 0.04413828253746033, "global_step": 214148, "epoch": 2406, "lr": 1.0635339813135476e-05}
{"train_loss": 0.07032833248376846, "global_step": 214149, "epoch": 2406, "lr": 1.0634982362484785e-05}
{"train_loss": 0.06743980944156647, "global_step": 214150, "epoch": 2406, "lr": 1.0634624917126257e-05}
{"train_loss": 0.01952480338513851, "global_step": 214151, "epoch": 2406, "lr": 1.0634267477059905e-05}
{"train_loss": 0.052568498998880386, "global_step": 214152, "epoch": 2406, "lr": 1.0633910042285817e-05}
{"train_loss": 0.03611353412270546, "global_step": 214153, "epoch": 2406, "lr": 1.0633552612803998e-05}
{"train_loss": 0.05144425481557846, "global_step": 214154, "epoch": 2406, "lr": 1.0633195188614525e-05}
{"train_loss": 0.05237486958503723, "global_step": 214155, "epoch": 2406, "lr": 1.0632837769717452e-05}
{"train_loss": 0.04468562453985214, "global_step": 214156, "epoch": 2406, "lr": 1.0632480356112796e-05}
{"train_loss": 0.07517959922552109, "global_step": 214157, "epoch": 2406, "lr": 1.0632122947800643e-05}
{"train_loss": 0.059453874826431274, "global_step": 214158, "epoch": 2406, "lr": 1.0631765544781003e-05}
{"train_loss": 0.0531274788081646, "global_step": 214159, "epoch": 2406, "lr": 1.0631408147053956e-05}
{"train_loss": 0.03642890974879265, "global_step": 214160, "epoch": 2406, "lr": 1.0631050754619525e-05}
{"train_loss": 0.07095130532979965, "global_step": 214161, "epoch": 2406, "lr": 1.0630693367477779e-05}
{"train_loss": 0.05644688755273819, "global_step": 214162, "epoch": 2406, "lr": 1.0630335985628748e-05}
{"train_loss": 0.06330884248018265, "global_step": 214163, "epoch": 2406, "lr": 1.06299786090725e-05}
{"train_loss": 0.07689421623945236, "global_step": 214164, "epoch": 2406, "lr": 1.0629621237809074e-05}
{"train_loss": 0.06033770740032196, "global_step": 214165, "epoch": 2406, "lr": 1.0629263871838496e-05}
{"train_loss": 0.08457823842763901, "global_step": 214166, "epoch": 2406, "lr": 1.0628906511160852e-05}
{"train_loss": 0.07365284115076065, "global_step": 214167, "epoch": 2406, "lr": 1.0628549155776158e-05}
{"train_loss": 0.0449521504342556, "global_step": 214168, "epoch": 2406, "lr": 1.0628191805684485e-05}
{"train_loss": 0.02693452686071396, "global_step": 214169, "epoch": 2406, "lr": 1.0627834460885876e-05}
{"train_loss": 0.07623164355754852, "global_step": 214170, "epoch": 2406, "lr": 1.0627477121380358e-05}
{"train_loss": 0.043304771184921265, "global_step": 214171, "epoch": 2406, "lr": 1.062711978716801e-05}
{"train_loss": 0.07537233829498291, "global_step": 214172, "epoch": 2406, "lr": 1.0626762458248857e-05}
{"train_loss": 0.07106268405914307, "global_step": 214173, "epoch": 2406, "lr": 1.0626405134622952e-05}
{"train_loss": 0.037683211266994476, "global_step": 214174, "epoch": 2406, "lr": 1.0626047816290357e-05}
{"train_loss": 0.060084931552410126, "global_step": 214175, "epoch": 2406, "lr": 1.0625690503251102e-05}
{"train_loss": 0.02305489592254162, "global_step": 214176, "epoch": 2406, "lr": 1.0625333195505255e-05}
{"train_loss": 0.03570351004600525, "global_step": 214177, "epoch": 2406, "lr": 1.0624975893052841e-05}
{"train_loss": 0.058929964900016785, "global_step": 214178, "epoch": 2406, "lr": 1.0624618595893925e-05}
{"train_loss": 0.12058575451374054, "global_step": 214179, "epoch": 2406, "lr": 1.0624261304028544e-05}
{"train_loss": 0.07191389799118042, "global_step": 214180, "epoch": 2406, "lr": 1.0623904017456759e-05}
{"train_loss": 0.043797872960567474, "global_step": 214181, "epoch": 2406, "lr": 1.0623546736178603e-05}
{"train_loss": 0.03833235055208206, "global_step": 214182, "epoch": 2406, "lr": 1.0623189460194139e-05}
{"train_loss": 0.07529852539300919, "global_step": 214183, "epoch": 2406, "lr": 1.0622832189503395e-05}
{"train_loss": 0.09791845828294754, "global_step": 214184, "epoch": 2406, "lr": 1.0622474924106446e-05}
{"train_loss": 0.05199354887008667, "global_step": 214185, "epoch": 2406, "lr": 1.0622117664003306e-05}
{"train_loss": 0.024772342294454575, "global_step": 214186, "epoch": 2406, "lr": 1.0621760409194064e-05}
{"train_loss": 0.06587877869606018, "global_step": 214187, "epoch": 2406, "lr": 1.0621403159678727e-05}
{"train_loss": 0.052496954798698425, "global_step": 214188, "epoch": 2406, "lr": 1.0621045915457379e-05}
{"train_loss": 0.04484059661626816, "global_step": 214189, "epoch": 2406, "lr": 1.0620688676530038e-05}
{"train_loss": 0.031581077724695206, "global_step": 214190, "epoch": 2406, "lr": 1.0620331442896764e-05}
{"train_loss": 0.04746472090482712, "global_step": 214191, "epoch": 2406, "lr": 1.061997421455762e-05}
{"train_loss": 0.07798445969820023, "global_step": 214192, "epoch": 2406, "lr": 1.0619616991512632e-05}
{"train_loss": 0.0523604117333889, "global_step": 214193, "epoch": 2406, "lr": 1.0619259773761864e-05}
{"train_loss": 0.05378144979476929, "global_step": 214194, "epoch": 2406, "lr": 1.0618902561305344e-05}
{"train_loss": 0.10321106016635895, "global_step": 214195, "epoch": 2406, "lr": 1.0618545354143145e-05}
{"train_loss": 0.05350523442029953, "global_step": 214196, "epoch": 2406, "lr": 1.0618188152275294e-05}
{"train_loss": 0.03503014147281647, "global_step": 214197, "epoch": 2406, "lr": 1.0617830955701857e-05}
{"train_loss": 0.07420533150434494, "global_step": 214198, "epoch": 2406, "lr": 1.0617473764422858e-05}
{"train_loss": 0.09050780534744263, "global_step": 214199, "epoch": 2406, "lr": 1.061711657843838e-05}
{"train_loss": 0.031843580305576324, "global_step": 214200, "epoch": 2406, "lr": 1.0616759397748433e-05}
{"train_loss": 0.07002591341733932, "global_step": 214201, "epoch": 2406, "lr": 1.0616402222353095e-05}
{"train_loss": 0.08567684143781662, "global_step": 214202, "epoch": 2406, "lr": 1.06160450522524e-05}
{"train_loss": 0.06576487421989441, "global_step": 214203, "epoch": 2406, "lr": 1.0615687887446385e-05}
{"train_loss": 0.028727592900395393, "global_step": 214204, "epoch": 2406, "lr": 1.061533072793513e-05}
{"train_loss": 0.07228836417198181, "global_step": 214205, "epoch": 2406, "lr": 1.0614973573718645e-05}
{"train_loss": 0.05455672740936279, "global_step": 214206, "epoch": 2406, "lr": 1.0614616424797008e-05}
{"train_loss": 0.0464189387857914, "global_step": 214207, "epoch": 2406, "lr": 1.0614259281170246e-05}
{"train_loss": 0.07351826876401901, "global_step": 214208, "epoch": 2406, "lr": 1.0613902142838433e-05}
{"train_loss": 0.03547033295035362, "global_step": 214209, "epoch": 2406, "lr": 1.0613545009801585e-05}
{"train_loss": 0.07382001727819443, "global_step": 214210, "epoch": 2406, "lr": 1.0613187882059777e-05}
{"train_loss": 0.09623963385820389, "global_step": 214211, "epoch": 2406, "lr": 1.0612830759613035e-05}
{"train_loss": 0.03531598672270775, "global_step": 214212, "epoch": 2406, "lr": 1.061247364246143e-05}
{"train_loss": 0.04588540643453598, "global_step": 214213, "epoch": 2406, "lr": 1.0612116530604982e-05}
{"train_loss": 0.048151999711990356, "global_step": 214214, "epoch": 2406, "lr": 1.061175942404377e-05}
{"train_loss": 0.07285573333501816, "global_step": 214215, "epoch": 2406, "lr": 1.0611402322777814e-05}
{"train_loss": 0.06940745562314987, "global_step": 214216, "epoch": 2406, "lr": 1.0611045226807192e-05}
{"train_loss": 0.022799557074904442, "global_step": 214217, "epoch": 2406, "lr": 1.0610688136131919e-05}
{"train_loss": 0.10726390033960342, "global_step": 214218, "epoch": 2406, "lr": 1.061033105075207e-05}
{"train_loss": 0.042723704129457474, "global_step": 214219, "epoch": 2406, "lr": 1.060997397066767e-05}
{"train_loss": 0.06903564184904099, "global_step": 214220, "epoch": 2406, "lr": 1.0609616895878794e-05}
{"train_loss": 0.057107072323560715, "global_step": 214221, "epoch": 2406, "lr": 1.0609259826385458e-05}
{"train_loss": 0.059398869261815306, "global_step": 214222, "epoch": 2406, "lr": 1.0608902762187739e-05, "val_loss": 8.874432563781738}
{"train_loss": 0.03422390669584274, "global_step": 214223, "epoch": 2407, "lr": 1.0608545703285666e-05}
{"train_loss": 0.07607172429561615, "global_step": 214224, "epoch": 2407, "lr": 1.0608188649679302e-05}
{"train_loss": 0.05636117234826088, "global_step": 214225, "epoch": 2407, "lr": 1.0607831601368678e-05}
{"train_loss": 0.052802469581365585, "global_step": 214226, "epoch": 2407, "lr": 1.0607474558353846e-05}
{"train_loss": 0.04193021357059479, "global_step": 214227, "epoch": 2407, "lr": 1.0607117520634875e-05}
{"train_loss": 0.0661311224102974, "global_step": 214228, "epoch": 2407, "lr": 1.0606760488211782e-05}
{"train_loss": 0.03003958985209465, "global_step": 214229, "epoch": 2407, "lr": 1.0606403461084647e-05}
{"train_loss": 0.07910068333148956, "global_step": 214230, "epoch": 2407, "lr": 1.060604643925348e-05}
{"train_loss": 0.09559818357229233, "global_step": 214231, "epoch": 2407, "lr": 1.0605689422718373e-05}
{"train_loss": 0.043647486716508865, "global_step": 214232, "epoch": 2407, "lr": 1.060533241147933e-05}
{"train_loss": 0.03505496308207512, "global_step": 214233, "epoch": 2407, "lr": 1.0604975405536438e-05}
{"train_loss": 0.0764554962515831, "global_step": 214234, "epoch": 2407, "lr": 1.060461840488971e-05}
{"train_loss": 0.07529213279485703, "global_step": 214235, "epoch": 2407, "lr": 1.0604261409539228e-05}
{"train_loss": 0.026309669017791748, "global_step": 214236, "epoch": 2407, "lr": 1.0603904419485006e-05}
{"train_loss": 0.0926014631986618, "global_step": 214237, "epoch": 2407, "lr": 1.0603547434727124e-05}
{"train_loss": 0.05555909126996994, "global_step": 214238, "epoch": 2407, "lr": 1.06031904552656e-05}
{"train_loss": 0.11335192620754242, "global_step": 214239, "epoch": 2407, "lr": 1.060283348110051e-05}
{"train_loss": 0.04870595782995224, "global_step": 214240, "epoch": 2407, "lr": 1.0602476512231879e-05}
{"train_loss": 0.07267290353775024, "global_step": 214241, "epoch": 2407, "lr": 1.0602119548659778e-05}
{"train_loss": 0.058815084397792816, "global_step": 214242, "epoch": 2407, "lr": 1.060176259038424e-05}
{"train_loss": 0.039620909839868546, "global_step": 214243, "epoch": 2407, "lr": 1.0601405637405298e-05}
{"train_loss": 0.037025392055511475, "global_step": 214244, "epoch": 2407, "lr": 1.0601048689723037e-05}
{"train_loss": 0.04393259435892105, "global_step": 214245, "epoch": 2407, "lr": 1.0600691747337466e-05}
{"train_loss": 0.04558824375271797, "global_step": 214246, "epoch": 2407, "lr": 1.0600334810248668e-05}
{"train_loss": 0.054875485599040985, "global_step": 214247, "epoch": 2407, "lr": 1.0599977878456657e-05}
{"train_loss": 0.1234956756234169, "global_step": 214248, "epoch": 2407, "lr": 1.059962095196152e-05}
{"train_loss": 0.09407903999090195, "global_step": 214249, "epoch": 2407, "lr": 1.0599264030763261e-05}
{"train_loss": 0.03617578744888306, "global_step": 214250, "epoch": 2407, "lr": 1.0598907114861973e-05}
{"train_loss": 0.03947500139474869, "global_step": 214251, "epoch": 2407, "lr": 1.0598550204257662e-05}
{"train_loss": 0.05220606178045273, "global_step": 214252, "epoch": 2407, "lr": 1.0598193298950415e-05}
{"train_loss": 0.06647361814975739, "global_step": 214253, "epoch": 2407, "lr": 1.0597836398940242e-05}
{"train_loss": 0.030619360506534576, "global_step": 214254, "epoch": 2407, "lr": 1.0597479504227225e-05}
{"train_loss": 0.053908392786979675, "global_step": 214255, "epoch": 2407, "lr": 1.0597122614811383e-05}
{"train_loss": 0.06580869108438492, "global_step": 214256, "epoch": 2407, "lr": 1.0596765730692793e-05}
{"train_loss": 0.07852774113416672, "global_step": 214257, "epoch": 2407, "lr": 1.059640885187147e-05}
{"train_loss": 0.04023100808262825, "global_step": 214258, "epoch": 2407, "lr": 1.0596051978347494e-05}
{"train_loss": 0.07173974066972733, "global_step": 214259, "epoch": 2407, "lr": 1.059569511012089e-05}
{"train_loss": 0.10699442028999329, "global_step": 214260, "epoch": 2407, "lr": 1.0595338247191721e-05}
{"train_loss": 0.06951189786195755, "global_step": 214261, "epoch": 2407, "lr": 1.0594981389560021e-05}
{"train_loss": 0.06403147429227829, "global_step": 214262, "epoch": 2407, "lr": 1.059462453722584e-05}
{"train_loss": 0.06213171035051346, "global_step": 214263, "epoch": 2407, "lr": 1.0594267690189246e-05}
{"train_loss": 0.027044547721743584, "global_step": 214264, "epoch": 2407, "lr": 1.0593910848450261e-05}
{"train_loss": 0.06301332265138626, "global_step": 214265, "epoch": 2407, "lr": 1.0593554012008961e-05}
{"train_loss": 0.04075797274708748, "global_step": 214266, "epoch": 2407, "lr": 1.059319718086536e-05}
{"train_loss": 0.0944063737988472, "global_step": 214267, "epoch": 2407, "lr": 1.0592840355019534e-05}
{"train_loss": 0.11387240141630173, "global_step": 214268, "epoch": 2407, "lr": 1.0592483534471514e-05}
{"train_loss": 0.05558457225561142, "global_step": 214269, "epoch": 2407, "lr": 1.0592126719221362e-05}
{"train_loss": 0.0648110955953598, "global_step": 214270, "epoch": 2407, "lr": 1.0591769909269106e-05}
{"train_loss": 0.044700004160404205, "global_step": 214271, "epoch": 2407, "lr": 1.0591413104614822e-05}
{"train_loss": 0.06644406169652939, "global_step": 214272, "epoch": 2407, "lr": 1.0591056305258528e-05}
{"train_loss": 0.03053896501660347, "global_step": 214273, "epoch": 2407, "lr": 1.0590699511200297e-05}
{"train_loss": 0.06558206677436829, "global_step": 214274, "epoch": 2407, "lr": 1.0590342722440155e-05}
{"train_loss": 0.044466689229011536, "global_step": 214275, "epoch": 2407, "lr": 1.0589985938978175e-05}
{"train_loss": 0.052420154213905334, "global_step": 214276, "epoch": 2407, "lr": 1.0589629160814374e-05}
{"train_loss": 0.03950951248407364, "global_step": 214277, "epoch": 2407, "lr": 1.0589272387948834e-05}
{"train_loss": 0.04025488346815109, "global_step": 214278, "epoch": 2407, "lr": 1.0588915620381573e-05}
{"train_loss": 0.038562458008527756, "global_step": 214279, "epoch": 2407, "lr": 1.0588558858112662e-05}
{"train_loss": 0.03220473974943161, "global_step": 214280, "epoch": 2407, "lr": 1.0588202101142141e-05}
{"train_loss": 0.08840501308441162, "global_step": 214281, "epoch": 2407, "lr": 1.0587845349470043e-05}
{"train_loss": 0.07657516002655029, "global_step": 214282, "epoch": 2407, "lr": 1.058748860309644e-05}
{"train_loss": 0.03587306663393974, "global_step": 214283, "epoch": 2407, "lr": 1.058713186202136e-05}
{"train_loss": 0.051816269755363464, "global_step": 214284, "epoch": 2407, "lr": 1.0586775126244869e-05}
{"train_loss": 0.02361316978931427, "global_step": 214285, "epoch": 2407, "lr": 1.058641839576699e-05}
{"train_loss": 0.0901593267917633, "global_step": 214286, "epoch": 2407, "lr": 1.0586061670587804e-05}
{"train_loss": 0.03261498361825943, "global_step": 214287, "epoch": 2407, "lr": 1.0585704950707326e-05}
{"train_loss": 0.038203101605176926, "global_step": 214288, "epoch": 2407, "lr": 1.0585348236125636e-05}
{"train_loss": 0.02274726703763008, "global_step": 214289, "epoch": 2407, "lr": 1.0584991526842753e-05}
{"train_loss": 0.05441056564450264, "global_step": 214290, "epoch": 2407, "lr": 1.0584634822858747e-05}
{"train_loss": 0.09020725637674332, "global_step": 214291, "epoch": 2407, "lr": 1.0584278124173647e-05}
{"train_loss": 0.04728282243013382, "global_step": 214292, "epoch": 2407, "lr": 1.058392143078752e-05}
{"train_loss": 0.036502864211797714, "global_step": 214293, "epoch": 2407, "lr": 1.0583564742700392e-05}
{"train_loss": 0.04953017458319664, "global_step": 214294, "epoch": 2407, "lr": 1.0583208059912337e-05}
{"train_loss": 0.06595846265554428, "global_step": 214295, "epoch": 2407, "lr": 1.0582851382423376e-05}
{"train_loss": 0.05950341746211052, "global_step": 214296, "epoch": 2407, "lr": 1.0582494710233582e-05}
{"train_loss": 0.08065468817949295, "global_step": 214297, "epoch": 2407, "lr": 1.0582138043342981e-05}
{"train_loss": 0.05354855954647064, "global_step": 214298, "epoch": 2407, "lr": 1.0581781381751643e-05}
{"train_loss": 0.08938058465719223, "global_step": 214299, "epoch": 2407, "lr": 1.0581424725459588e-05}
{"train_loss": 0.06111353635787964, "global_step": 214300, "epoch": 2407, "lr": 1.0581068074466881e-05}
{"train_loss": 0.0356990285217762, "global_step": 214301, "epoch": 2407, "lr": 1.0580711428773587e-05}
{"train_loss": 0.03246039152145386, "global_step": 214302, "epoch": 2407, "lr": 1.0580354788379721e-05}
{"train_loss": 0.04853164777159691, "global_step": 214303, "epoch": 2407, "lr": 1.0579998153285353e-05}
{"train_loss": 0.10602615773677826, "global_step": 214304, "epoch": 2407, "lr": 1.0579641523490519e-05}
{"train_loss": 0.03348700329661369, "global_step": 214305, "epoch": 2407, "lr": 1.0579284898995278e-05}
{"train_loss": 0.032337818294763565, "global_step": 214306, "epoch": 2407, "lr": 1.057892827979966e-05}
{"train_loss": 0.0516887903213501, "global_step": 214307, "epoch": 2407, "lr": 1.0578571665903742e-05}
{"train_loss": 0.057650987058877945, "global_step": 214308, "epoch": 2407, "lr": 1.057821505730754e-05}
{"train_loss": 0.08039706200361252, "global_step": 214309, "epoch": 2407, "lr": 1.0577858454011124e-05}
{"train_loss": 0.035835474729537964, "global_step": 214310, "epoch": 2407, "lr": 1.0577501856014527e-05}
{"train_loss": 0.05793678971907396, "global_step": 214311, "epoch": 2407, "lr": 1.0577145263317817e-05, "val_loss": 8.902045249938965}
{"train_loss": 0.012619990855455399, "global_step": 214312, "epoch": 2408, "lr": 1.0576788675921012e-05}
{"train_loss": 0.05454159155488014, "global_step": 214313, "epoch": 2408, "lr": 1.0576432093824196e-05}
{"train_loss": 0.05049663037061691, "global_step": 214314, "epoch": 2408, "lr": 1.0576075517027383e-05}
{"train_loss": 0.07954476773738861, "global_step": 214315, "epoch": 2408, "lr": 1.057571894553065e-05}
{"train_loss": 0.07400628179311752, "global_step": 214316, "epoch": 2408, "lr": 1.0575362379334013e-05}
{"train_loss": 0.07748888432979584, "global_step": 214317, "epoch": 2408, "lr": 1.0575005818437555e-05}
{"train_loss": 0.04576685652136803, "global_step": 214318, "epoch": 2408, "lr": 1.0574649262841308e-05}
{"train_loss": 0.07691401243209839, "global_step": 214319, "epoch": 2408, "lr": 1.057429271254531e-05}
{"train_loss": 0.0830037072300911, "global_step": 214320, "epoch": 2408, "lr": 1.057393616754962e-05}
{"train_loss": 0.03350706025958061, "global_step": 214321, "epoch": 2408, "lr": 1.0573579627854279e-05}
{"train_loss": 0.031155897304415703, "global_step": 214322, "epoch": 2408, "lr": 1.057322309345935e-05}
{"train_loss": 0.06605236232280731, "global_step": 214323, "epoch": 2408, "lr": 1.0572866564364858e-05}
{"train_loss": 0.06264164298772812, "global_step": 214324, "epoch": 2408, "lr": 1.0572510040570877e-05}
{"train_loss": 0.03345499187707901, "global_step": 214325, "epoch": 2408, "lr": 1.0572153522077427e-05}
{"train_loss": 0.04593740403652191, "global_step": 214326, "epoch": 2408, "lr": 1.0571797008884582e-05}
{"train_loss": 0.06268679350614548, "global_step": 214327, "epoch": 2408, "lr": 1.057144050099237e-05}
{"train_loss": 0.026503756642341614, "global_step": 214328, "epoch": 2408, "lr": 1.0571083998400855e-05}
{"train_loss": 0.030950086191296577, "global_step": 214329, "epoch": 2408, "lr": 1.0570727501110068e-05}
{"train_loss": 0.057342275977134705, "global_step": 214330, "epoch": 2408, "lr": 1.0570371009120079e-05}
{"train_loss": 0.012944789603352547, "global_step": 214331, "epoch": 2408, "lr": 1.0570014522430905e-05}
{"train_loss": 0.06879220902919769, "global_step": 214332, "epoch": 2408, "lr": 1.0569658041042629e-05}
{"train_loss": 0.05862119048833847, "global_step": 214333, "epoch": 2408, "lr": 1.056930156495527e-05}
{"train_loss": 0.10289876908063889, "global_step": 214334, "epoch": 2408, "lr": 1.0568945094168903e-05}
{"train_loss": 0.05392773449420929, "global_step": 214335, "epoch": 2408, "lr": 1.056858862868354e-05}
{"train_loss": 0.07608912885189056, "global_step": 214336, "epoch": 2408, "lr": 1.0568232168499254e-05}
{"train_loss": 0.0803179144859314, "global_step": 214337, "epoch": 2408, "lr": 1.0567875713616104e-05}
{"train_loss": 0.08896823972463608, "global_step": 214338, "epoch": 2408, "lr": 1.0567519264034104e-05}
{"train_loss": 0.05756566300988197, "global_step": 214339, "epoch": 2408, "lr": 1.056716281975334e-05}
{"train_loss": 0.06312938779592514, "global_step": 214340, "epoch": 2408, "lr": 1.0566806380773819e-05}
{"train_loss": 0.03699250519275665, "global_step": 214341, "epoch": 2408, "lr": 1.056644994709563e-05}
{"train_loss": 0.055019453167915344, "global_step": 214342, "epoch": 2408, "lr": 1.0566093518718789e-05}
{"train_loss": 0.062092125415802, "global_step": 214343, "epoch": 2408, "lr": 1.0565737095643363e-05}
{"train_loss": 0.07197976857423782, "global_step": 214344, "epoch": 2408, "lr": 1.0565380677869386e-05}
{"train_loss": 0.07938376814126968, "global_step": 214345, "epoch": 2408, "lr": 1.0565024265396922e-05}
{"train_loss": 0.017374275252223015, "global_step": 214346, "epoch": 2408, "lr": 1.0564667858225997e-05}
{"train_loss": 0.05289757251739502, "global_step": 214347, "epoch": 2408, "lr": 1.056431145635669e-05}
{"train_loss": 0.07453444600105286, "global_step": 214348, "epoch": 2408, "lr": 1.0563955059789011e-05}
{"train_loss": 0.019006861373782158, "global_step": 214349, "epoch": 2408, "lr": 1.0563598668523045e-05}
{"train_loss": 0.07440072298049927, "global_step": 214350, "epoch": 2408, "lr": 1.0563242282558804e-05}
{"train_loss": 0.09243190288543701, "global_step": 214351, "epoch": 2408, "lr": 1.0562885901896375e-05}
{"train_loss": 0.029578382149338722, "global_step": 214352, "epoch": 2408, "lr": 1.0562529526535763e-05}
{"train_loss": 0.059102028608322144, "global_step": 214353, "epoch": 2408, "lr": 1.0562173156477062e-05}
{"train_loss": 0.06779024004936218, "global_step": 214354, "epoch": 2408, "lr": 1.0561816791720274e-05}
{"train_loss": 0.05162513628602028, "global_step": 214355, "epoch": 2408, "lr": 1.0561460432265485e-05}
{"train_loss": 0.047339923679828644, "global_step": 214356, "epoch": 2408, "lr": 1.0561104078112717e-05}
{"train_loss": 0.044918205589056015, "global_step": 214357, "epoch": 2408, "lr": 1.0560747729262037e-05}
{"train_loss": 0.07135208696126938, "global_step": 214358, "epoch": 2408, "lr": 1.0560391385713486e-05}
{"train_loss": 0.09345169365406036, "global_step": 214359, "epoch": 2408, "lr": 1.05600350474671e-05}
{"train_loss": 0.09765388071537018, "global_step": 214360, "epoch": 2408, "lr": 1.0559678714522942e-05}
{"train_loss": 0.038896072655916214, "global_step": 214361, "epoch": 2408, "lr": 1.055932238688105e-05}
{"train_loss": 0.049592893570661545, "global_step": 214362, "epoch": 2408, "lr": 1.0558966064541482e-05}
{"train_loss": 0.09953949600458145, "global_step": 214363, "epoch": 2408, "lr": 1.0558609747504267e-05}
{"train_loss": 0.05877039209008217, "global_step": 214364, "epoch": 2408, "lr": 1.0558253435769483e-05}
{"train_loss": 0.07352027297019958, "global_step": 214365, "epoch": 2408, "lr": 1.0557897129337146e-05}
{"train_loss": 0.038312237709760666, "global_step": 214366, "epoch": 2408, "lr": 1.0557540828207335e-05}
{"train_loss": 0.03130375221371651, "global_step": 214367, "epoch": 2408, "lr": 1.0557184532380066e-05}
{"train_loss": 0.07627720385789871, "global_step": 214368, "epoch": 2408, "lr": 1.0556828241855422e-05}
{"train_loss": 0.018558183684945107, "global_step": 214369, "epoch": 2408, "lr": 1.0556471956633407e-05}
{"train_loss": 0.07292643934488297, "global_step": 214370, "epoch": 2408, "lr": 1.0556115676714118e-05}
{"train_loss": 0.09259966015815735, "global_step": 214371, "epoch": 2408, "lr": 1.055575940209756e-05}
{"train_loss": 0.07843226939439774, "global_step": 214372, "epoch": 2408, "lr": 1.0555403132783797e-05}
{"train_loss": 0.034985072910785675, "global_step": 214373, "epoch": 2408, "lr": 1.0555046868772894e-05}
{"train_loss": 0.05140687897801399, "global_step": 214374, "epoch": 2408, "lr": 1.0554690610064877e-05}
{"train_loss": 0.02613467536866665, "global_step": 214375, "epoch": 2408, "lr": 1.0554334356659807e-05}
{"train_loss": 0.019474292173981667, "global_step": 214376, "epoch": 2408, "lr": 1.0553978108557717e-05}
{"train_loss": 0.055730316787958145, "global_step": 214377, "epoch": 2408, "lr": 1.0553621865758673e-05}
{"train_loss": 0.0846625566482544, "global_step": 214378, "epoch": 2408, "lr": 1.0553265628262704e-05}
{"train_loss": 0.028958013281226158, "global_step": 214379, "epoch": 2408, "lr": 1.0552909396069882e-05}
{"train_loss": 0.0394950695335865, "global_step": 214380, "epoch": 2408, "lr": 1.0552553169180224e-05}
{"train_loss": 0.045853376388549805, "global_step": 214381, "epoch": 2408, "lr": 1.0552196947593806e-05}
{"train_loss": 0.07294369488954544, "global_step": 214382, "epoch": 2408, "lr": 1.0551840731310658e-05}
{"train_loss": 0.04666190966963768, "global_step": 214383, "epoch": 2408, "lr": 1.055148452033084e-05}
{"train_loss": 0.07592745125293732, "global_step": 214384, "epoch": 2408, "lr": 1.0551128314654384e-05}
{"train_loss": 0.054058197885751724, "global_step": 214385, "epoch": 2408, "lr": 1.0550772114281366e-05}
{"train_loss": 0.06241977587342262, "global_step": 214386, "epoch": 2408, "lr": 1.0550415919211799e-05}
{"train_loss": 0.05030805990099907, "global_step": 214387, "epoch": 2408, "lr": 1.0550059729445755e-05}
{"train_loss": 0.036959972232580185, "global_step": 214388, "epoch": 2408, "lr": 1.054970354498327e-05}
{"train_loss": 0.09632015228271484, "global_step": 214389, "epoch": 2408, "lr": 1.0549347365824391e-05}
{"train_loss": 0.0727546289563179, "global_step": 214390, "epoch": 2408, "lr": 1.0548991191969188e-05}
{"train_loss": 0.041981518268585205, "global_step": 214391, "epoch": 2408, "lr": 1.0548635023417691e-05}
{"train_loss": 0.0447801873087883, "global_step": 214392, "epoch": 2408, "lr": 1.0548278860169935e-05}
{"train_loss": 0.06337310373783112, "global_step": 214393, "epoch": 2408, "lr": 1.0547922702225998e-05}
{"train_loss": 0.05256051942706108, "global_step": 214394, "epoch": 2408, "lr": 1.0547566549585897e-05}
{"train_loss": 0.06466581672430038, "global_step": 214395, "epoch": 2408, "lr": 1.0547210402249708e-05}
{"train_loss": 0.054975930601358414, "global_step": 214396, "epoch": 2408, "lr": 1.0546854260217464e-05}
{"train_loss": 0.056355398148298264, "global_step": 214397, "epoch": 2408, "lr": 1.0546498123489202e-05}
{"train_loss": 0.06891689449548721, "global_step": 214398, "epoch": 2408, "lr": 1.0546141992064996e-05}
{"train_loss": 0.03991574048995972, "global_step": 214399, "epoch": 2408, "lr": 1.0545785865944868e-05}
{"train_loss": 0.05714284863029973, "global_step": 214400, "epoch": 2408, "lr": 1.0545429745128894e-05, "val_loss": 8.998056411743164}
{"train_loss": 0.060681454837322235, "global_step": 214401, "epoch": 2409, "lr": 1.0545073629617086e-05}
{"train_loss": 0.05945279821753502, "global_step": 214402, "epoch": 2409, "lr": 1.054471751940953e-05}
{"train_loss": 0.08188483864068985, "global_step": 214403, "epoch": 2409, "lr": 1.0544361414506244e-05}
{"train_loss": 0.06460653990507126, "global_step": 214404, "epoch": 2409, "lr": 1.0544005314907297e-05}
{"train_loss": 0.06655284762382507, "global_step": 214405, "epoch": 2409, "lr": 1.0543649220612717e-05}
{"train_loss": 0.02183249592781067, "global_step": 214406, "epoch": 2409, "lr": 1.0543293131622578e-05}
{"train_loss": 0.033810071647167206, "global_step": 214407, "epoch": 2409, "lr": 1.0542937047936891e-05}
{"train_loss": 0.04074915871024132, "global_step": 214408, "epoch": 2409, "lr": 1.0542580969555733e-05}
{"train_loss": 0.061823777854442596, "global_step": 214409, "epoch": 2409, "lr": 1.0542224896479152e-05}
{"train_loss": 0.05142860859632492, "global_step": 214410, "epoch": 2409, "lr": 1.054186882870718e-05}
{"train_loss": 0.022074513137340546, "global_step": 214411, "epoch": 2409, "lr": 1.054151276623988e-05}
{"train_loss": 0.05313245207071304, "global_step": 214412, "epoch": 2409, "lr": 1.0541156709077282e-05}
{"train_loss": 0.05572718381881714, "global_step": 214413, "epoch": 2409, "lr": 1.0540800657219463e-05}
{"train_loss": 0.05334543064236641, "global_step": 214414, "epoch": 2409, "lr": 1.0540444610666433e-05}
{"train_loss": 0.11146100610494614, "global_step": 214415, "epoch": 2409, "lr": 1.0540088569418271e-05}
{"train_loss": 0.043894365429878235, "global_step": 214416, "epoch": 2409, "lr": 1.0539732533475005e-05}
{"train_loss": 0.041816242039203644, "global_step": 214417, "epoch": 2409, "lr": 1.0539376502836702e-05}
{"train_loss": 0.05253211036324501, "global_step": 214418, "epoch": 2409, "lr": 1.0539020477503386e-05}
{"train_loss": 0.04151346907019615, "global_step": 214419, "epoch": 2409, "lr": 1.0538664457475134e-05}
{"train_loss": 0.11736781150102615, "global_step": 214420, "epoch": 2409, "lr": 1.0538308442751959e-05}
{"train_loss": 0.046990811824798584, "global_step": 214421, "epoch": 2409, "lr": 1.0537952433333941e-05}
{"train_loss": 0.052499011158943176, "global_step": 214422, "epoch": 2409, "lr": 1.0537596429221103e-05}
{"train_loss": 0.05890672653913498, "global_step": 214423, "epoch": 2409, "lr": 1.053724043041352e-05}
{"train_loss": 0.061695195734500885, "global_step": 214424, "epoch": 2409, "lr": 1.0536884436911209e-05}
{"train_loss": 0.05536074563860893, "global_step": 214425, "epoch": 2409, "lr": 1.0536528448714245e-05}
{"train_loss": 0.05999571457505226, "global_step": 214426, "epoch": 2409, "lr": 1.0536172465822653e-05}
{"train_loss": 0.07560461759567261, "global_step": 214427, "epoch": 2409, "lr": 1.0535816488236488e-05}
{"train_loss": 0.03576577082276344, "global_step": 214428, "epoch": 2409, "lr": 1.0535460515955813e-05}
{"train_loss": 0.07719989866018295, "global_step": 214429, "epoch": 2409, "lr": 1.0535104548980667e-05}
{"train_loss": 0.04322446882724762, "global_step": 214430, "epoch": 2409, "lr": 1.0534748587311082e-05}
{"train_loss": 0.06329602748155594, "global_step": 214431, "epoch": 2409, "lr": 1.0534392630947138e-05}
{"train_loss": 0.04720986634492874, "global_step": 214432, "epoch": 2409, "lr": 1.0534036679888842e-05}
{"train_loss": 0.04312746971845627, "global_step": 214433, "epoch": 2409, "lr": 1.053368073413628e-05}
{"train_loss": 0.047774575650691986, "global_step": 214434, "epoch": 2409, "lr": 1.0533324793689476e-05}
{"train_loss": 0.106405109167099, "global_step": 214435, "epoch": 2409, "lr": 1.0532968858548492e-05}
{"train_loss": 0.03477548062801361, "global_step": 214436, "epoch": 2409, "lr": 1.053261292871337e-05}
{"train_loss": 0.03957084193825722, "global_step": 214437, "epoch": 2409, "lr": 1.0532257004184143e-05}
{"train_loss": 0.087476447224617, "global_step": 214438, "epoch": 2409, "lr": 1.0531901084960889e-05}
{"train_loss": 0.04879915714263916, "global_step": 214439, "epoch": 2409, "lr": 1.0531545171043627e-05}
{"train_loss": 0.08186385780572891, "global_step": 214440, "epoch": 2409, "lr": 1.0531189262432428e-05}
{"train_loss": 0.06572644412517548, "global_step": 214441, "epoch": 2409, "lr": 1.0530833359127317e-05}
{"train_loss": 0.05792580172419548, "global_step": 214442, "epoch": 2409, "lr": 1.053047746112837e-05}
{"train_loss": 0.09041524678468704, "global_step": 214443, "epoch": 2409, "lr": 1.0530121568435603e-05}
{"train_loss": 0.05804940685629845, "global_step": 214444, "epoch": 2409, "lr": 1.052976568104908e-05}
{"train_loss": 0.06809317320585251, "global_step": 214445, "epoch": 2409, "lr": 1.0529409798968865e-05}
{"train_loss": 0.04711867496371269, "global_step": 214446, "epoch": 2409, "lr": 1.0529053922194975e-05}
{"train_loss": 0.09528139233589172, "global_step": 214447, "epoch": 2409, "lr": 1.0528698050727486e-05}
{"train_loss": 0.07702254503965378, "global_step": 214448, "epoch": 2409, "lr": 1.052834218456642e-05}
{"train_loss": 0.06793545186519623, "global_step": 214449, "epoch": 2409, "lr": 1.0527986323711852e-05}
{"train_loss": 0.06097973883152008, "global_step": 214450, "epoch": 2409, "lr": 1.0527630468163796e-05}
{"train_loss": 0.09211748093366623, "global_step": 214451, "epoch": 2409, "lr": 1.052727461792234e-05}
{"train_loss": 0.049921371042728424, "global_step": 214452, "epoch": 2409, "lr": 1.0526918772987494e-05}
{"train_loss": 0.05875829607248306, "global_step": 214453, "epoch": 2409, "lr": 1.0526562933359335e-05}
{"train_loss": 0.05338677763938904, "global_step": 214454, "epoch": 2409, "lr": 1.0526207099037882e-05}
{"train_loss": 0.073380246758461, "global_step": 214455, "epoch": 2409, "lr": 1.0525851270023219e-05}
{"train_loss": 0.04585592821240425, "global_step": 214456, "epoch": 2409, "lr": 1.0525495446315359e-05}
{"train_loss": 0.03908893093466759, "global_step": 214457, "epoch": 2409, "lr": 1.0525139627914377e-05}
{"train_loss": 0.07681203633546829, "global_step": 214458, "epoch": 2409, "lr": 1.0524783814820294e-05}
{"train_loss": 0.05576065927743912, "global_step": 214459, "epoch": 2409, "lr": 1.0524428007033188e-05}
{"train_loss": 0.07477715611457825, "global_step": 214460, "epoch": 2409, "lr": 1.0524072204553076e-05}
{"train_loss": 0.061319127678871155, "global_step": 214461, "epoch": 2409, "lr": 1.0523716407380041e-05}
{"train_loss": 0.015432257205247879, "global_step": 214462, "epoch": 2409, "lr": 1.0523360615514095e-05}
{"train_loss": 0.021613622084259987, "global_step": 214463, "epoch": 2409, "lr": 1.0523004828955296e-05}
{"train_loss": 0.07701491564512253, "global_step": 214464, "epoch": 2409, "lr": 1.052264904770372e-05}
{"train_loss": 0.06064788997173309, "global_step": 214465, "epoch": 2409, "lr": 1.052229327175937e-05}
{"train_loss": 0.0770513266324997, "global_step": 214466, "epoch": 2409, "lr": 1.0521937501122337e-05}
{"train_loss": 0.05020132660865784, "global_step": 214467, "epoch": 2409, "lr": 1.0521581735792634e-05}
{"train_loss": 0.0455310195684433, "global_step": 214468, "epoch": 2409, "lr": 1.0521225975770338e-05}
{"train_loss": 0.10012741386890411, "global_step": 214469, "epoch": 2409, "lr": 1.052087022105548e-05}
{"train_loss": 0.02614522911608219, "global_step": 214470, "epoch": 2409, "lr": 1.0520514471648091e-05}
{"train_loss": 0.0770610123872757, "global_step": 214471, "epoch": 2409, "lr": 1.0520158727548257e-05}
{"train_loss": 0.057016078382730484, "global_step": 214472, "epoch": 2409, "lr": 1.0519802988755994e-05}
{"train_loss": 0.06321413069963455, "global_step": 214473, "epoch": 2409, "lr": 1.0519447255271375e-05}
{"train_loss": 0.05453936755657196, "global_step": 214474, "epoch": 2409, "lr": 1.0519091527094432e-05}
{"train_loss": 0.05242731049656868, "global_step": 214475, "epoch": 2409, "lr": 1.0518735804225205e-05}
{"train_loss": 0.08122360706329346, "global_step": 214476, "epoch": 2409, "lr": 1.0518380086663765e-05}
{"train_loss": 0.06812364608049393, "global_step": 214477, "epoch": 2409, "lr": 1.0518024374410136e-05}
{"train_loss": 0.06591812521219254, "global_step": 214478, "epoch": 2409, "lr": 1.0517668667464387e-05}
{"train_loss": 0.030584953725337982, "global_step": 214479, "epoch": 2409, "lr": 1.0517312965826548e-05}
{"train_loss": 0.0948944166302681, "global_step": 214480, "epoch": 2409, "lr": 1.0516957269496675e-05}
{"train_loss": 0.06943035125732422, "global_step": 214481, "epoch": 2409, "lr": 1.0516601578474822e-05}
{"train_loss": 0.04045894369482994, "global_step": 214482, "epoch": 2409, "lr": 1.0516245892761023e-05}
{"train_loss": 0.03745012730360031, "global_step": 214483, "epoch": 2409, "lr": 1.0515890212355346e-05}
{"train_loss": 0.13244950771331787, "global_step": 214484, "epoch": 2409, "lr": 1.0515534537257815e-05}
{"train_loss": 0.052938636392354965, "global_step": 214485, "epoch": 2409, "lr": 1.05151788674685e-05}
{"train_loss": 0.07579904794692993, "global_step": 214486, "epoch": 2409, "lr": 1.0514823202987423e-05}
{"train_loss": 0.05524963513016701, "global_step": 214487, "epoch": 2409, "lr": 1.0514467543814665e-05}
{"train_loss": 0.06404431164264679, "global_step": 214488, "epoch": 2409, "lr": 1.051411188995024e-05}
{"train_loss": 0.060423343327273143, "global_step": 214489, "epoch": 2409, "lr": 1.0513756241394224e-05, "val_loss": 8.884016036987305}
{"train_loss": 0.07096646726131439, "global_step": 214490, "epoch": 2410, "lr": 1.0513400598146639e-05}
{"train_loss": 0.06072816252708435, "global_step": 214491, "epoch": 2410, "lr": 1.0513044960207558e-05}
{"train_loss": 0.07426238805055618, "global_step": 214492, "epoch": 2410, "lr": 1.0512689327577002e-05}
{"train_loss": 0.04874136298894882, "global_step": 214493, "epoch": 2410, "lr": 1.0512333700255051e-05}
{"train_loss": 0.08471059054136276, "global_step": 214494, "epoch": 2410, "lr": 1.051197807824172e-05}
{"train_loss": 0.083197221159935, "global_step": 214495, "epoch": 2410, "lr": 1.0511622461537085e-05}
{"train_loss": 0.046885956078767776, "global_step": 214496, "epoch": 2410, "lr": 1.0511266850141171e-05}
{"train_loss": 0.07374631613492966, "global_step": 214497, "epoch": 2410, "lr": 1.0510911244054051e-05}
{"train_loss": 0.07388999313116074, "global_step": 214498, "epoch": 2410, "lr": 1.0510555643275737e-05}
{"train_loss": 0.04088958352804184, "global_step": 214499, "epoch": 2410, "lr": 1.05102000478063e-05}
{"train_loss": 0.0700480192899704, "global_step": 214500, "epoch": 2410, "lr": 1.05098444576458e-05}
{"train_loss": 0.03820914402604103, "global_step": 214501, "epoch": 2410, "lr": 1.050948887279426e-05}
{"train_loss": 0.09627415239810944, "global_step": 214502, "epoch": 2410, "lr": 1.0509133293251744e-05}
{"train_loss": 0.06490971148014069, "global_step": 214503, "epoch": 2410, "lr": 1.0508777719018287e-05}
{"train_loss": 0.013827953487634659, "global_step": 214504, "epoch": 2410, "lr": 1.0508422150093955e-05}
{"train_loss": 0.0896555557847023, "global_step": 214505, "epoch": 2410, "lr": 1.0508066586478771e-05}
{"train_loss": 0.054017093032598495, "global_step": 214506, "epoch": 2410, "lr": 1.0507711028172807e-05}
{"train_loss": 0.06686906516551971, "global_step": 214507, "epoch": 2410, "lr": 1.05073554751761e-05}
{"train_loss": 0.014347144402563572, "global_step": 214508, "epoch": 2410, "lr": 1.0506999927488687e-05}
{"train_loss": 0.037712570279836655, "global_step": 214509, "epoch": 2410, "lr": 1.0506644385110637e-05}
{"train_loss": 0.035797588527202606, "global_step": 214510, "epoch": 2410, "lr": 1.0506288848041979e-05}
{"train_loss": 0.04365266487002373, "global_step": 214511, "epoch": 2410, "lr": 1.050593331628278e-05}
{"train_loss": 0.04978154972195625, "global_step": 214512, "epoch": 2410, "lr": 1.0505577789833072e-05}
{"train_loss": 0.12172691524028778, "global_step": 214513, "epoch": 2410, "lr": 1.0505222268692899e-05}
{"train_loss": 0.03795779123902321, "global_step": 214514, "epoch": 2410, "lr": 1.050486675286233e-05}
{"train_loss": 0.06402353942394257, "global_step": 214515, "epoch": 2410, "lr": 1.0504511242341385e-05}
{"train_loss": 0.08978253602981567, "global_step": 214516, "epoch": 2410, "lr": 1.0504155737130128e-05}
{"train_loss": 0.047298822551965714, "global_step": 214517, "epoch": 2410, "lr": 1.0503800237228622e-05}
{"train_loss": 0.06208597123622894, "global_step": 214518, "epoch": 2410, "lr": 1.0503444742636886e-05}
{"train_loss": 0.06277085840702057, "global_step": 214519, "epoch": 2410, "lr": 1.0503089253354993e-05}
{"train_loss": 0.09192150086164474, "global_step": 214520, "epoch": 2410, "lr": 1.050273376938296e-05}
{"train_loss": 0.08288391679525375, "global_step": 214521, "epoch": 2410, "lr": 1.0502378290720872e-05}
{"train_loss": 0.04751184210181236, "global_step": 214522, "epoch": 2410, "lr": 1.050202281736874e-05}
{"train_loss": 0.05553898215293884, "global_step": 214523, "epoch": 2410, "lr": 1.0501667349326644e-05}
{"train_loss": 0.05817243456840515, "global_step": 214524, "epoch": 2410, "lr": 1.0501311886594606e-05}
{"train_loss": 0.07917685061693192, "global_step": 214525, "epoch": 2410, "lr": 1.0500956429172698e-05}
{"train_loss": 0.053809478878974915, "global_step": 214526, "epoch": 2410, "lr": 1.0500600977060942e-05}
{"train_loss": 0.058357711881399155, "global_step": 214527, "epoch": 2410, "lr": 1.0500245530259411e-05}
{"train_loss": 0.14759708940982819, "global_step": 214528, "epoch": 2410, "lr": 1.0499890088768128e-05}
{"train_loss": 0.06070321798324585, "global_step": 214529, "epoch": 2410, "lr": 1.0499534652587167e-05}
{"train_loss": 0.09826916456222534, "global_step": 214530, "epoch": 2410, "lr": 1.049917922171655e-05}
{"train_loss": 0.0411854051053524, "global_step": 214531, "epoch": 2410, "lr": 1.0498823796156348e-05}
{"train_loss": 0.07959417998790741, "global_step": 214532, "epoch": 2410, "lr": 1.0498468375906583e-05}
{"train_loss": 0.034045301377773285, "global_step": 214533, "epoch": 2410, "lr": 1.0498112960967332e-05}
{"train_loss": 0.05913228541612625, "global_step": 214534, "epoch": 2410, "lr": 1.0497757551338616e-05}
{"train_loss": 0.03416202962398529, "global_step": 214535, "epoch": 2410, "lr": 1.0497402147020496e-05}
{"train_loss": 0.043772947043180466, "global_step": 214536, "epoch": 2410, "lr": 1.049704674801303e-05}
{"train_loss": 0.04266904294490814, "global_step": 214537, "epoch": 2410, "lr": 1.0496691354316241e-05}
{"train_loss": 0.035322677344083786, "global_step": 214538, "epoch": 2410, "lr": 1.0496335965930204e-05}
{"train_loss": 0.03810989484190941, "global_step": 214539, "epoch": 2410, "lr": 1.0495980582854943e-05}
{"train_loss": 0.03213251382112503, "global_step": 214540, "epoch": 2410, "lr": 1.0495625205090526e-05}
{"train_loss": 0.06043139100074768, "global_step": 214541, "epoch": 2410, "lr": 1.0495269832636978e-05}
{"train_loss": 0.023961544036865234, "global_step": 214542, "epoch": 2410, "lr": 1.0494914465494371e-05}
{"train_loss": 0.04505224525928497, "global_step": 214543, "epoch": 2410, "lr": 1.0494559103662732e-05}
{"train_loss": 0.07889832556247711, "global_step": 214544, "epoch": 2410, "lr": 1.0494203747142128e-05}
{"train_loss": 0.07052744925022125, "global_step": 214545, "epoch": 2410, "lr": 1.0493848395932599e-05}
{"train_loss": 0.13393135368824005, "global_step": 214546, "epoch": 2410, "lr": 1.0493493050034175e-05}
{"train_loss": 0.06980367749929428, "global_step": 214547, "epoch": 2410, "lr": 1.0493137709446932e-05}
{"train_loss": 0.034182894974946976, "global_step": 214548, "epoch": 2410, "lr": 1.0492782374170895e-05}
{"train_loss": 0.03713928908109665, "global_step": 214549, "epoch": 2410, "lr": 1.0492427044206133e-05}
{"train_loss": 0.06874796003103256, "global_step": 214550, "epoch": 2410, "lr": 1.049207171955267e-05}
{"train_loss": 0.08542121946811676, "global_step": 214551, "epoch": 2410, "lr": 1.0491716400210583e-05}
{"train_loss": 0.06605948507785797, "global_step": 214552, "epoch": 2410, "lr": 1.0491361086179901e-05}
{"train_loss": 0.04200580716133118, "global_step": 214553, "epoch": 2410, "lr": 1.0491005777460662e-05}
{"train_loss": 0.02558964304625988, "global_step": 214554, "epoch": 2410, "lr": 1.0490650474052922e-05}
{"train_loss": 0.05200837180018425, "global_step": 214555, "epoch": 2410, "lr": 1.0490295175956744e-05}
{"train_loss": 0.05488264560699463, "global_step": 214556, "epoch": 2410, "lr": 1.0489939883172156e-05}
{"train_loss": 0.03057974949479103, "global_step": 214557, "epoch": 2410, "lr": 1.0489584595699225e-05}
{"train_loss": 0.026754435151815414, "global_step": 214558, "epoch": 2410, "lr": 1.0489229313537974e-05}
{"train_loss": 0.058949537575244904, "global_step": 214559, "epoch": 2410, "lr": 1.048887403668848e-05}
{"train_loss": 0.035493139177560806, "global_step": 214560, "epoch": 2410, "lr": 1.0488518765150762e-05}
{"train_loss": 0.031093118712306023, "global_step": 214561, "epoch": 2410, "lr": 1.048816349892489e-05}
{"train_loss": 0.02690410241484642, "global_step": 214562, "epoch": 2410, "lr": 1.0487808238010893e-05}
{"train_loss": 0.10325843095779419, "global_step": 214563, "epoch": 2410, "lr": 1.0487452982408846e-05}
{"train_loss": 0.06157834455370903, "global_step": 214564, "epoch": 2410, "lr": 1.048709773211876e-05}
{"train_loss": 0.05591481178998947, "global_step": 214565, "epoch": 2410, "lr": 1.0486742487140715e-05}
{"train_loss": 0.06210979446768761, "global_step": 214566, "epoch": 2410, "lr": 1.0486387247474738e-05}
{"train_loss": 0.07788893580436707, "global_step": 214567, "epoch": 2410, "lr": 1.0486032013120895e-05}
{"train_loss": 0.0763552188873291, "global_step": 214568, "epoch": 2410, "lr": 1.0485676784079207e-05}
{"train_loss": 0.04459667205810547, "global_step": 214569, "epoch": 2410, "lr": 1.0485321560349754e-05}
{"train_loss": 0.04012491554021835, "global_step": 214570, "epoch": 2410, "lr": 1.0484966341932556e-05}
{"train_loss": 0.015697738155722618, "global_step": 214571, "epoch": 2410, "lr": 1.048461112882767e-05}
{"train_loss": 0.05984927713871002, "global_step": 214572, "epoch": 2410, "lr": 1.0484255921035164e-05}
{"train_loss": 0.08639076352119446, "global_step": 214573, "epoch": 2410, "lr": 1.048390071855505e-05}
{"train_loss": 0.05798909440636635, "global_step": 214574, "epoch": 2410, "lr": 1.048354552138741e-05}
{"train_loss": 0.026506250724196434, "global_step": 214575, "epoch": 2410, "lr": 1.048319032953226e-05}
{"train_loss": 0.030834663659334183, "global_step": 214576, "epoch": 2410, "lr": 1.0482835142989683e-05}
{"train_loss": 0.06732521206140518, "global_step": 214577, "epoch": 2410, "lr": 1.048247996175969e-05}
{"train_loss": 0.05840874661095022, "global_step": 214578, "epoch": 2410, "lr": 1.0482124785842357e-05, "val_loss": 8.821316719055176, "train_action_mse_error": 10.575584411621094}
{"train_loss": 0.01297798939049244, "global_step": 214579, "epoch": 2411, "lr": 1.0481769615237708e-05}
{"train_loss": 0.046451982110738754, "global_step": 214580, "epoch": 2411, "lr": 1.0481414449945815e-05}
{"train_loss": 0.041856586933135986, "global_step": 214581, "epoch": 2411, "lr": 1.0481059289966705e-05}
{"train_loss": 0.08322553336620331, "global_step": 214582, "epoch": 2411, "lr": 1.0480704135300446e-05}
{"train_loss": 0.051583752036094666, "global_step": 214583, "epoch": 2411, "lr": 1.0480348985947064e-05}
{"train_loss": 0.022401751950383186, "global_step": 214584, "epoch": 2411, "lr": 1.0479993841906626e-05}
{"train_loss": 0.07325855642557144, "global_step": 214585, "epoch": 2411, "lr": 1.0479638703179173e-05}
{"train_loss": 0.05375709384679794, "global_step": 214586, "epoch": 2411, "lr": 1.0479283569764742e-05}
{"train_loss": 0.03672323375940323, "global_step": 214587, "epoch": 2411, "lr": 1.04789284416634e-05}
{"train_loss": 0.02508571185171604, "global_step": 214588, "epoch": 2411, "lr": 1.0478573318875168e-05}
{"train_loss": 0.05906311422586441, "global_step": 214589, "epoch": 2411, "lr": 1.0478218201400125e-05}
{"train_loss": 0.049684010446071625, "global_step": 214590, "epoch": 2411, "lr": 1.0477863089238293e-05}
{"train_loss": 0.044705912470817566, "global_step": 214591, "epoch": 2411, "lr": 1.0477507982389744e-05}
{"train_loss": 0.03640105202794075, "global_step": 214592, "epoch": 2411, "lr": 1.0477152880854496e-05}
{"train_loss": 0.027063384652137756, "global_step": 214593, "epoch": 2411, "lr": 1.0476797784632624e-05}
{"train_loss": 0.02263098582625389, "global_step": 214594, "epoch": 2411, "lr": 1.0476442693724158e-05}
{"train_loss": 0.06418407708406448, "global_step": 214595, "epoch": 2411, "lr": 1.0476087608129165e-05}
{"train_loss": 0.02767143025994301, "global_step": 214596, "epoch": 2411, "lr": 1.0475732527847664e-05}
{"train_loss": 0.07397451996803284, "global_step": 214597, "epoch": 2411, "lr": 1.0475377452879731e-05}
{"train_loss": 0.07351695746183395, "global_step": 214598, "epoch": 2411, "lr": 1.0475022383225392e-05}
{"train_loss": 0.023265590891242027, "global_step": 214599, "epoch": 2411, "lr": 1.0474667318884718e-05}
{"train_loss": 0.0961979478597641, "global_step": 214600, "epoch": 2411, "lr": 1.0474312259857726e-05}
{"train_loss": 0.02805332839488983, "global_step": 214601, "epoch": 2411, "lr": 1.0473957206144496e-05}
{"train_loss": 0.018032461404800415, "global_step": 214602, "epoch": 2411, "lr": 1.047360215774505e-05}
{"train_loss": 0.05502444878220558, "global_step": 214603, "epoch": 2411, "lr": 1.0473247114659462e-05}
{"train_loss": 0.058501213788986206, "global_step": 214604, "epoch": 2411, "lr": 1.0472892076887746e-05}
{"train_loss": 0.06128546968102455, "global_step": 214605, "epoch": 2411, "lr": 1.0472537044429981e-05}
{"train_loss": 0.0725504457950592, "global_step": 214606, "epoch": 2411, "lr": 1.0472182017286191e-05}
{"train_loss": 0.03954099863767624, "global_step": 214607, "epoch": 2411, "lr": 1.047182699545643e-05}
{"train_loss": 0.01965481787919998, "global_step": 214608, "epoch": 2411, "lr": 1.0471471978940767e-05}
{"train_loss": 0.044961411505937576, "global_step": 214609, "epoch": 2411, "lr": 1.0471116967739219e-05}
{"train_loss": 0.04387481510639191, "global_step": 214610, "epoch": 2411, "lr": 1.0470761961851861e-05}
{"train_loss": 0.03915146365761757, "global_step": 214611, "epoch": 2411, "lr": 1.0470406961278712e-05}
{"train_loss": 0.04629987105727196, "global_step": 214612, "epoch": 2411, "lr": 1.047005196601985e-05}
{"train_loss": 0.034947898238897324, "global_step": 214613, "epoch": 2411, "lr": 1.0469696976075294e-05}
{"train_loss": 0.040675658732652664, "global_step": 214614, "epoch": 2411, "lr": 1.046934199144512e-05}
{"train_loss": 0.05624406412243843, "global_step": 214615, "epoch": 2411, "lr": 1.0468987012129344e-05}
{"train_loss": 0.08789331465959549, "global_step": 214616, "epoch": 2411, "lr": 1.046863203812805e-05}
{"train_loss": 0.06570860743522644, "global_step": 214617, "epoch": 2411, "lr": 1.0468277069441252e-05}
{"train_loss": 0.08439537137746811, "global_step": 214618, "epoch": 2411, "lr": 1.0467922106069022e-05}
{"train_loss": 0.027090853080153465, "global_step": 214619, "epoch": 2411, "lr": 1.0467567148011387e-05}
{"train_loss": 0.06272827833890915, "global_step": 214620, "epoch": 2411, "lr": 1.0467212195268416e-05}
{"train_loss": 0.07075593620538712, "global_step": 214621, "epoch": 2411, "lr": 1.046685724784014e-05}
{"train_loss": 0.07731397449970245, "global_step": 214622, "epoch": 2411, "lr": 1.0466502305726616e-05}
{"train_loss": 0.020748496055603027, "global_step": 214623, "epoch": 2411, "lr": 1.0466147368927897e-05}
{"train_loss": 0.05008798837661743, "global_step": 214624, "epoch": 2411, "lr": 1.0465792437444012e-05}
{"train_loss": 0.028369871899485588, "global_step": 214625, "epoch": 2411, "lr": 1.0465437511275027e-05}
{"train_loss": 0.08261144161224365, "global_step": 214626, "epoch": 2411, "lr": 1.0465082590420967e-05}
{"train_loss": 0.08276841044425964, "global_step": 214627, "epoch": 2411, "lr": 1.0464727674881914e-05}
{"train_loss": 0.06006719917058945, "global_step": 214628, "epoch": 2411, "lr": 1.0464372764657876e-05}
{"train_loss": 0.025482308119535446, "global_step": 214629, "epoch": 2411, "lr": 1.0464017859748943e-05}
{"train_loss": 0.029646843671798706, "global_step": 214630, "epoch": 2411, "lr": 1.0463662960155125e-05}
{"train_loss": 0.08738129585981369, "global_step": 214631, "epoch": 2411, "lr": 1.0463308065876492e-05}
{"train_loss": 0.06148151680827141, "global_step": 214632, "epoch": 2411, "lr": 1.0462953176913081e-05}
{"train_loss": 0.02797076478600502, "global_step": 214633, "epoch": 2411, "lr": 1.0462598293264953e-05}
{"train_loss": 0.04168860241770744, "global_step": 214634, "epoch": 2411, "lr": 1.0462243414932132e-05}
{"train_loss": 0.1020650789141655, "global_step": 214635, "epoch": 2411, "lr": 1.04618885419147e-05}
{"train_loss": 0.06116819009184837, "global_step": 214636, "epoch": 2411, "lr": 1.0461533674212664e-05}
{"train_loss": 0.06130724027752876, "global_step": 214637, "epoch": 2411, "lr": 1.0461178811826111e-05}
{"train_loss": 0.04362662509083748, "global_step": 214638, "epoch": 2411, "lr": 1.0460823954755056e-05}
{"train_loss": 0.06248319521546364, "global_step": 214639, "epoch": 2411, "lr": 1.0460469102999577e-05}
{"train_loss": 0.10182501375675201, "global_step": 214640, "epoch": 2411, "lr": 1.0460114256559688e-05}
{"train_loss": 0.05627688392996788, "global_step": 214641, "epoch": 2411, "lr": 1.0459759415435472e-05}
{"train_loss": 0.04275808110833168, "global_step": 214642, "epoch": 2411, "lr": 1.0459404579626947e-05}
{"train_loss": 0.05117719620466232, "global_step": 214643, "epoch": 2411, "lr": 1.045904974913417e-05}
{"train_loss": 0.08067461103200912, "global_step": 214644, "epoch": 2411, "lr": 1.0458694923957207e-05}
{"train_loss": 0.024801230058073997, "global_step": 214645, "epoch": 2411, "lr": 1.0458340104096076e-05}
{"train_loss": 0.04401933029294014, "global_step": 214646, "epoch": 2411, "lr": 1.0457985289550853e-05}
{"train_loss": 0.06956769526004791, "global_step": 214647, "epoch": 2411, "lr": 1.0457630480321556e-05}
{"train_loss": 0.023517269641160965, "global_step": 214648, "epoch": 2411, "lr": 1.0457275676408262e-05}
{"train_loss": 0.018479667603969574, "global_step": 214649, "epoch": 2411, "lr": 1.0456920877811e-05}
{"train_loss": 0.03477027639746666, "global_step": 214650, "epoch": 2411, "lr": 1.0456566084529829e-05}
{"train_loss": 0.08689603954553604, "global_step": 214651, "epoch": 2411, "lr": 1.0456211296564778e-05}
{"train_loss": 0.08766978234052658, "global_step": 214652, "epoch": 2411, "lr": 1.0455856513915923e-05}
{"train_loss": 0.027326876297593117, "global_step": 214653, "epoch": 2411, "lr": 1.045550173658328e-05}
{"train_loss": 0.05319257825613022, "global_step": 214654, "epoch": 2411, "lr": 1.045514696456693e-05}
{"train_loss": 0.03427093103528023, "global_step": 214655, "epoch": 2411, "lr": 1.0454792197866886e-05}
{"train_loss": 0.041987720876932144, "global_step": 214656, "epoch": 2411, "lr": 1.0454437436483227e-05}
{"train_loss": 0.02876707725226879, "global_step": 214657, "epoch": 2411, "lr": 1.0454082680415977e-05}
{"train_loss": 0.0718337893486023, "global_step": 214658, "epoch": 2411, "lr": 1.0453727929665202e-05}
{"train_loss": 0.046163786202669144, "global_step": 214659, "epoch": 2411, "lr": 1.0453373184230935e-05}
{"train_loss": 0.06261182576417923, "global_step": 214660, "epoch": 2411, "lr": 1.0453018444113239e-05}
{"train_loss": 0.027822643518447876, "global_step": 214661, "epoch": 2411, "lr": 1.0452663709312156e-05}
{"train_loss": 0.07420294731855392, "global_step": 214662, "epoch": 2411, "lr": 1.0452308979827708e-05}
{"train_loss": 0.0544683076441288, "global_step": 214663, "epoch": 2411, "lr": 1.0451954255659985e-05}
{"train_loss": 0.05051034688949585, "global_step": 214664, "epoch": 2411, "lr": 1.0451599536809003e-05}
{"train_loss": 0.04039034992456436, "global_step": 214665, "epoch": 2411, "lr": 1.0451244823274831e-05}
{"train_loss": 0.023966431617736816, "global_step": 214666, "epoch": 2411, "lr": 1.0450890115057494e-05}
{"train_loss": 0.05098894399622183, "global_step": 214667, "epoch": 2411, "lr": 1.0450535412157069e-05, "val_loss": 8.941311836242676}
{"train_loss": 0.0793478861451149, "global_step": 214668, "epoch": 2412, "lr": 1.0450180714573571e-05}
{"train_loss": 0.09201838821172714, "global_step": 214669, "epoch": 2412, "lr": 1.0449826022307079e-05}
{"train_loss": 0.049361780285835266, "global_step": 214670, "epoch": 2412, "lr": 1.0449471335357614e-05}
{"train_loss": 0.05199034884572029, "global_step": 214671, "epoch": 2412, "lr": 1.044911665372525e-05}
{"train_loss": 0.06179044395685196, "global_step": 214672, "epoch": 2412, "lr": 1.0448761977409999e-05}
{"train_loss": 0.035344623029232025, "global_step": 214673, "epoch": 2412, "lr": 1.044840730641195e-05}
{"train_loss": 0.061163630336523056, "global_step": 214674, "epoch": 2412, "lr": 1.0448052640731115e-05}
{"train_loss": 0.07377973943948746, "global_step": 214675, "epoch": 2412, "lr": 1.0447697980367571e-05}
{"train_loss": 0.10673680901527405, "global_step": 214676, "epoch": 2412, "lr": 1.044734332532134e-05}
{"train_loss": 0.03754403069615364, "global_step": 214677, "epoch": 2412, "lr": 1.0446988675592495e-05}
{"train_loss": 0.06357807666063309, "global_step": 214678, "epoch": 2412, "lr": 1.0446634031181057e-05}
{"train_loss": 0.04034191370010376, "global_step": 214679, "epoch": 2412, "lr": 1.0446279392087099e-05}
{"train_loss": 0.06354766339063644, "global_step": 214680, "epoch": 2412, "lr": 1.0445924758310644e-05}
{"train_loss": 0.06822829693555832, "global_step": 214681, "epoch": 2412, "lr": 1.044557012985175e-05}
{"train_loss": 0.07095858454704285, "global_step": 214682, "epoch": 2412, "lr": 1.0445215506710482e-05}
{"train_loss": 0.09943734854459763, "global_step": 214683, "epoch": 2412, "lr": 1.0444860888886859e-05}
{"train_loss": 0.06341098994016647, "global_step": 214684, "epoch": 2412, "lr": 1.0444506276380955e-05}
{"train_loss": 0.07495015859603882, "global_step": 214685, "epoch": 2412, "lr": 1.0444151669192791e-05}
{"train_loss": 0.016906678676605225, "global_step": 214686, "epoch": 2412, "lr": 1.0443797067322448e-05}
{"train_loss": 0.05415263772010803, "global_step": 214687, "epoch": 2412, "lr": 1.0443442470769932e-05}
{"train_loss": 0.07577630877494812, "global_step": 214688, "epoch": 2412, "lr": 1.0443087879535336e-05}
{"train_loss": 0.06268011033535004, "global_step": 214689, "epoch": 2412, "lr": 1.0442733293618661e-05}
{"train_loss": 0.03822708502411842, "global_step": 214690, "epoch": 2412, "lr": 1.044237871302e-05}
{"train_loss": 0.017707133665680885, "global_step": 214691, "epoch": 2412, "lr": 1.0442024137739365e-05}
{"train_loss": 0.06873441487550735, "global_step": 214692, "epoch": 2412, "lr": 1.0441669567776829e-05}
{"train_loss": 0.053757596760988235, "global_step": 214693, "epoch": 2412, "lr": 1.0441315003132413e-05}
{"train_loss": 0.03076155297458172, "global_step": 214694, "epoch": 2412, "lr": 1.0440960443806197e-05}
{"train_loss": 0.036607205867767334, "global_step": 214695, "epoch": 2412, "lr": 1.0440605889798194e-05}
{"train_loss": 0.11533564329147339, "global_step": 214696, "epoch": 2412, "lr": 1.0440251341108487e-05}
{"train_loss": 0.039388734847307205, "global_step": 214697, "epoch": 2412, "lr": 1.0439896797737092e-05}
{"train_loss": 0.043794937431812286, "global_step": 214698, "epoch": 2412, "lr": 1.0439542259684081e-05}
{"train_loss": 0.03931855037808418, "global_step": 214699, "epoch": 2412, "lr": 1.0439187726949479e-05}
{"train_loss": 0.07069354504346848, "global_step": 214700, "epoch": 2412, "lr": 1.0438833199533365e-05}
{"train_loss": 0.03459790721535683, "global_step": 214701, "epoch": 2412, "lr": 1.0438478677435758e-05}
{"train_loss": 0.055681150406599045, "global_step": 214702, "epoch": 2412, "lr": 1.0438124160656708e-05}
{"train_loss": 0.04958935081958771, "global_step": 214703, "epoch": 2412, "lr": 1.0437769649196284e-05}
{"train_loss": 0.07036945223808289, "global_step": 214704, "epoch": 2412, "lr": 1.0437415143054507e-05}
{"train_loss": 0.07926714420318604, "global_step": 214705, "epoch": 2412, "lr": 1.043706064223145e-05}
{"train_loss": 0.04899575933814049, "global_step": 214706, "epoch": 2412, "lr": 1.0436706146727132e-05}
{"train_loss": 0.09939359873533249, "global_step": 214707, "epoch": 2412, "lr": 1.0436351656541632e-05}
{"train_loss": 0.04845963045954704, "global_step": 214708, "epoch": 2412, "lr": 1.0435997171674966e-05}
{"train_loss": 0.05440797656774521, "global_step": 214709, "epoch": 2412, "lr": 1.0435642692127212e-05}
{"train_loss": 0.03713058680295944, "global_step": 214710, "epoch": 2412, "lr": 1.0435288217898393e-05}
{"train_loss": 0.07553036510944366, "global_step": 214711, "epoch": 2412, "lr": 1.043493374898858e-05}
{"train_loss": 0.04575146734714508, "global_step": 214712, "epoch": 2412, "lr": 1.0434579285397795e-05}
{"train_loss": 0.05388671159744263, "global_step": 214713, "epoch": 2412, "lr": 1.0434224827126111e-05}
{"train_loss": 0.03181689977645874, "global_step": 214714, "epoch": 2412, "lr": 1.0433870374173548e-05}
{"train_loss": 0.061651330441236496, "global_step": 214715, "epoch": 2412, "lr": 1.0433515926540188e-05}
{"train_loss": 0.0394253171980381, "global_step": 214716, "epoch": 2412, "lr": 1.043316148422604e-05}
{"train_loss": 0.019358044490218163, "global_step": 214717, "epoch": 2412, "lr": 1.0432807047231174e-05}
{"train_loss": 0.05116116628050804, "global_step": 214718, "epoch": 2412, "lr": 1.0432452615555649e-05}
{"train_loss": 0.059087883681058884, "global_step": 214719, "epoch": 2412, "lr": 1.043209818919948e-05}
{"train_loss": 0.04574337974190712, "global_step": 214720, "epoch": 2412, "lr": 1.043174376816275e-05}
{"train_loss": 0.045262765139341354, "global_step": 214721, "epoch": 2412, "lr": 1.0431389352445481e-05}
{"train_loss": 0.026614800095558167, "global_step": 214722, "epoch": 2412, "lr": 1.0431034942047735e-05}
{"train_loss": 0.05709176883101463, "global_step": 214723, "epoch": 2412, "lr": 1.0430680536969544e-05}
{"train_loss": 0.046511463820934296, "global_step": 214724, "epoch": 2412, "lr": 1.0430326137210983e-05}
{"train_loss": 0.0437871553003788, "global_step": 214725, "epoch": 2412, "lr": 1.0429971742772066e-05}
{"train_loss": 0.06507223099470139, "global_step": 214726, "epoch": 2412, "lr": 1.0429617353652865e-05}
{"train_loss": 0.04307681694626808, "global_step": 214727, "epoch": 2412, "lr": 1.0429262969853415e-05}
{"train_loss": 0.03170812129974365, "global_step": 214728, "epoch": 2412, "lr": 1.0428908591373777e-05}
{"train_loss": 0.057057008147239685, "global_step": 214729, "epoch": 2412, "lr": 1.0428554218213976e-05}
{"train_loss": 0.0940786600112915, "global_step": 214730, "epoch": 2412, "lr": 1.0428199850374087e-05}
{"train_loss": 0.08804850280284882, "global_step": 214731, "epoch": 2412, "lr": 1.0427845487854132e-05}
{"train_loss": 0.02668762020766735, "global_step": 214732, "epoch": 2412, "lr": 1.0427491130654182e-05}
{"train_loss": 0.048626337200403214, "global_step": 214733, "epoch": 2412, "lr": 1.0427136778774266e-05}
{"train_loss": 0.06290923058986664, "global_step": 214734, "epoch": 2412, "lr": 1.0426782432214449e-05}
{"train_loss": 0.057866405695676804, "global_step": 214735, "epoch": 2412, "lr": 1.0426428090974754e-05}
{"train_loss": 0.022963382303714752, "global_step": 214736, "epoch": 2412, "lr": 1.042607375505526e-05}
{"train_loss": 0.0353006012737751, "global_step": 214737, "epoch": 2412, "lr": 1.0425719424455982e-05}
{"train_loss": 0.060405850410461426, "global_step": 214738, "epoch": 2412, "lr": 1.0425365099176998e-05}
{"train_loss": 0.044201239943504333, "global_step": 214739, "epoch": 2412, "lr": 1.0425010779218341e-05}
{"train_loss": 0.1141180619597435, "global_step": 214740, "epoch": 2412, "lr": 1.0424656464580046e-05}
{"train_loss": 0.023591140285134315, "global_step": 214741, "epoch": 2412, "lr": 1.0424302155262183e-05}
{"train_loss": 0.033207640051841736, "global_step": 214742, "epoch": 2412, "lr": 1.0423947851264782e-05}
{"train_loss": 0.05129372328519821, "global_step": 214743, "epoch": 2412, "lr": 1.0423593552587912e-05}
{"train_loss": 0.10068131238222122, "global_step": 214744, "epoch": 2412, "lr": 1.0423239259231593e-05}
{"train_loss": 0.0517466701567173, "global_step": 214745, "epoch": 2412, "lr": 1.0422884971195901e-05}
{"train_loss": 0.10557428747415543, "global_step": 214746, "epoch": 2412, "lr": 1.0422530688480852e-05}
{"train_loss": 0.025883490219712257, "global_step": 214747, "epoch": 2412, "lr": 1.0422176411086532e-05}
{"train_loss": 0.03616999834775925, "global_step": 214748, "epoch": 2412, "lr": 1.042182213901295e-05}
{"train_loss": 0.08677360415458679, "global_step": 214749, "epoch": 2412, "lr": 1.0421467872260182e-05}
{"train_loss": 0.0189838707447052, "global_step": 214750, "epoch": 2412, "lr": 1.0421113610828259e-05}
{"train_loss": 0.04150259494781494, "global_step": 214751, "epoch": 2412, "lr": 1.0420759354717247e-05}
{"train_loss": 0.05276628956198692, "global_step": 214752, "epoch": 2412, "lr": 1.0420405103927161e-05}
{"train_loss": 0.053342267870903015, "global_step": 214753, "epoch": 2412, "lr": 1.0420050858458075e-05}
{"train_loss": 0.07436251640319824, "global_step": 214754, "epoch": 2412, "lr": 1.0419696618310043e-05}
{"train_loss": 0.0679268091917038, "global_step": 214755, "epoch": 2412, "lr": 1.0419342383483088e-05}
{"train_loss": 0.05556717699163415, "global_step": 214756, "epoch": 2412, "lr": 1.0418988153977283e-05, "val_loss": 8.882479667663574}
{"train_loss": 0.035075683146715164, "global_step": 214757, "epoch": 2413, "lr": 1.0418633929792648e-05}
{"train_loss": 0.03319321200251579, "global_step": 214758, "epoch": 2413, "lr": 1.0418279710929263e-05}
{"train_loss": 0.050487954169511795, "global_step": 214759, "epoch": 2413, "lr": 1.0417925497387143e-05}
{"train_loss": 0.0429142564535141, "global_step": 214760, "epoch": 2413, "lr": 1.0417571289166361e-05}
{"train_loss": 0.08865109086036682, "global_step": 214761, "epoch": 2413, "lr": 1.0417217086266939e-05}
{"train_loss": 0.01726798340678215, "global_step": 214762, "epoch": 2413, "lr": 1.0416862888688955e-05}
{"train_loss": 0.07268873602151871, "global_step": 214763, "epoch": 2413, "lr": 1.0416508696432432e-05}
{"train_loss": 0.04074840620160103, "global_step": 214764, "epoch": 2413, "lr": 1.0416154509497439e-05}
{"train_loss": 0.0650298148393631, "global_step": 214765, "epoch": 2413, "lr": 1.0415800327883995e-05}
{"train_loss": 0.05936356261372566, "global_step": 214766, "epoch": 2413, "lr": 1.0415446151592178e-05}
{"train_loss": 0.030287083238363266, "global_step": 214767, "epoch": 2413, "lr": 1.041509198062201e-05}
{"train_loss": 0.03359859436750412, "global_step": 214768, "epoch": 2413, "lr": 1.0414737814973563e-05}
{"train_loss": 0.045925404876470566, "global_step": 214769, "epoch": 2413, "lr": 1.0414383654646859e-05}
{"train_loss": 0.07770520448684692, "global_step": 214770, "epoch": 2413, "lr": 1.0414029499641959e-05}
{"train_loss": 0.05388599634170532, "global_step": 214771, "epoch": 2413, "lr": 1.0413675349958923e-05}
{"train_loss": 0.04169357195496559, "global_step": 214772, "epoch": 2413, "lr": 1.0413321205597781e-05}
{"train_loss": 0.05774940922856331, "global_step": 214773, "epoch": 2413, "lr": 1.0412967066558582e-05}
{"train_loss": 0.056448791176080704, "global_step": 214774, "epoch": 2413, "lr": 1.0412612932841381e-05}
{"train_loss": 0.025008196011185646, "global_step": 214775, "epoch": 2413, "lr": 1.0412258804446211e-05}
{"train_loss": 0.07824978977441788, "global_step": 214776, "epoch": 2413, "lr": 1.0411904681373147e-05}
{"train_loss": 0.013629038818180561, "global_step": 214777, "epoch": 2413, "lr": 1.0411550563622218e-05}
{"train_loss": 0.03509459272027016, "global_step": 214778, "epoch": 2413, "lr": 1.0411196451193462e-05}
{"train_loss": 0.039411868900060654, "global_step": 214779, "epoch": 2413, "lr": 1.0410842344086952e-05}
{"train_loss": 0.06864506751298904, "global_step": 214780, "epoch": 2413, "lr": 1.0410488242302708e-05}
{"train_loss": 0.049799688160419464, "global_step": 214781, "epoch": 2413, "lr": 1.04101341458408e-05}
{"train_loss": 0.06491854041814804, "global_step": 214782, "epoch": 2413, "lr": 1.0409780054701257e-05}
{"train_loss": 0.025221912190318108, "global_step": 214783, "epoch": 2413, "lr": 1.0409425968884152e-05}
{"train_loss": 0.09932810813188553, "global_step": 214784, "epoch": 2413, "lr": 1.04090718883895e-05}
{"train_loss": 0.08883639425039291, "global_step": 214785, "epoch": 2413, "lr": 1.0408717813217378e-05}
{"train_loss": 0.06487452238798141, "global_step": 214786, "epoch": 2413, "lr": 1.0408363743367811e-05}
{"train_loss": 0.08977384865283966, "global_step": 214787, "epoch": 2413, "lr": 1.0408009678840868e-05}
{"train_loss": 0.0345650389790535, "global_step": 214788, "epoch": 2413, "lr": 1.0407655619636575e-05}
{"train_loss": 0.059994909912347794, "global_step": 214789, "epoch": 2413, "lr": 1.040730156575499e-05}
{"train_loss": 0.02816324681043625, "global_step": 214790, "epoch": 2413, "lr": 1.040694751719617e-05}
{"train_loss": 0.023623500019311905, "global_step": 214791, "epoch": 2413, "lr": 1.0406593473960147e-05}
{"train_loss": 0.03516962006688118, "global_step": 214792, "epoch": 2413, "lr": 1.0406239436046983e-05}
{"train_loss": 0.05939554423093796, "global_step": 214793, "epoch": 2413, "lr": 1.0405885403456706e-05}
{"train_loss": 0.06070508062839508, "global_step": 214794, "epoch": 2413, "lr": 1.0405531376189387e-05}
{"train_loss": 0.06281805038452148, "global_step": 214795, "epoch": 2413, "lr": 1.0405177354245049e-05}
{"train_loss": 0.039630260318517685, "global_step": 214796, "epoch": 2413, "lr": 1.0404823337623765e-05}
{"train_loss": 0.028890177607536316, "global_step": 214797, "epoch": 2413, "lr": 1.0404469326325555e-05}
{"train_loss": 0.048867300152778625, "global_step": 214798, "epoch": 2413, "lr": 1.0404115320350499e-05}
{"train_loss": 0.03454028442502022, "global_step": 214799, "epoch": 2413, "lr": 1.0403761319698614e-05}
{"train_loss": 0.08394812047481537, "global_step": 214800, "epoch": 2413, "lr": 1.0403407324369973e-05}
{"train_loss": 0.061218418180942535, "global_step": 214801, "epoch": 2413, "lr": 1.0403053334364593e-05}
{"train_loss": 0.03684556111693382, "global_step": 214802, "epoch": 2413, "lr": 1.040269934968256e-05}
{"train_loss": 0.03927106037735939, "global_step": 214803, "epoch": 2413, "lr": 1.0402345370323885e-05}
{"train_loss": 0.04954933747649193, "global_step": 214804, "epoch": 2413, "lr": 1.0401991396288645e-05}
{"train_loss": 0.04317310452461243, "global_step": 214805, "epoch": 2413, "lr": 1.040163742757686e-05}
{"train_loss": 0.056475285440683365, "global_step": 214806, "epoch": 2413, "lr": 1.040128346418861e-05}
{"train_loss": 0.05880476161837578, "global_step": 214807, "epoch": 2413, "lr": 1.0400929506123907e-05}
{"train_loss": 0.05228574201464653, "global_step": 214808, "epoch": 2413, "lr": 1.0400575553382818e-05}
{"train_loss": 0.061653271317481995, "global_step": 214809, "epoch": 2413, "lr": 1.0400221605965405e-05}
{"train_loss": 0.07046160846948624, "global_step": 214810, "epoch": 2413, "lr": 1.0399867663871681e-05}
{"train_loss": 0.054168738424777985, "global_step": 214811, "epoch": 2413, "lr": 1.0399513727101733e-05}
{"train_loss": 0.04373760521411896, "global_step": 214812, "epoch": 2413, "lr": 1.0399159795655583e-05}
{"train_loss": 0.03132516145706177, "global_step": 214813, "epoch": 2413, "lr": 1.0398805869533268e-05}
{"train_loss": 0.03364627808332443, "global_step": 214814, "epoch": 2413, "lr": 1.0398451948734867e-05}
{"train_loss": 0.03462481126189232, "global_step": 214815, "epoch": 2413, "lr": 1.0398098033260394e-05}
{"train_loss": 0.07661198824644089, "global_step": 214816, "epoch": 2413, "lr": 1.0397744123109938e-05}
{"train_loss": 0.11658123135566711, "global_step": 214817, "epoch": 2413, "lr": 1.039739021828351e-05}
{"train_loss": 0.06100364774465561, "global_step": 214818, "epoch": 2413, "lr": 1.0397036318781162e-05}
{"train_loss": 0.07979678362607956, "global_step": 214819, "epoch": 2413, "lr": 1.0396682424602965e-05}
{"train_loss": 0.0240138228982687, "global_step": 214820, "epoch": 2413, "lr": 1.039632853574894e-05}
{"train_loss": 0.04545857757329941, "global_step": 214821, "epoch": 2413, "lr": 1.0395974652219153e-05}
{"train_loss": 0.07862725108861923, "global_step": 214822, "epoch": 2413, "lr": 1.039562077401363e-05}
{"train_loss": 0.03080129809677601, "global_step": 214823, "epoch": 2413, "lr": 1.0395266901132456e-05}
{"train_loss": 0.09154795110225677, "global_step": 214824, "epoch": 2413, "lr": 1.0394913033575632e-05}
{"train_loss": 0.036694418638944626, "global_step": 214825, "epoch": 2413, "lr": 1.0394559171343237e-05}
{"train_loss": 0.028284471482038498, "global_step": 214826, "epoch": 2413, "lr": 1.0394205314435318e-05}
{"train_loss": 0.1307687610387802, "global_step": 214827, "epoch": 2413, "lr": 1.0393851462851906e-05}
{"train_loss": 0.05803457275032997, "global_step": 214828, "epoch": 2413, "lr": 1.0393497616593073e-05}
{"train_loss": 0.0478234700858593, "global_step": 214829, "epoch": 2413, "lr": 1.039314377565883e-05}
{"train_loss": 0.08447938412427902, "global_step": 214830, "epoch": 2413, "lr": 1.0392789940049269e-05}
{"train_loss": 0.03942999988794327, "global_step": 214831, "epoch": 2413, "lr": 1.0392436109764397e-05}
{"train_loss": 0.07843592762947083, "global_step": 214832, "epoch": 2413, "lr": 1.0392082284804294e-05}
{"train_loss": 0.06107393279671669, "global_step": 214833, "epoch": 2413, "lr": 1.0391728465168976e-05}
{"train_loss": 0.08848772943019867, "global_step": 214834, "epoch": 2413, "lr": 1.0391374650858522e-05}
{"train_loss": 0.051427796483039856, "global_step": 214835, "epoch": 2413, "lr": 1.039102084187295e-05}
{"train_loss": 0.017292946577072144, "global_step": 214836, "epoch": 2413, "lr": 1.0390667038212343e-05}
{"train_loss": 0.045796122401952744, "global_step": 214837, "epoch": 2413, "lr": 1.0390313239876708e-05}
{"train_loss": 0.023771503940224648, "global_step": 214838, "epoch": 2413, "lr": 1.0389959446866126e-05}
{"train_loss": 0.03162216395139694, "global_step": 214839, "epoch": 2413, "lr": 1.0389605659180623e-05}
{"train_loss": 0.09293447434902191, "global_step": 214840, "epoch": 2413, "lr": 1.0389251876820267e-05}
{"train_loss": 0.06085362285375595, "global_step": 214841, "epoch": 2413, "lr": 1.0388898099785071e-05}
{"train_loss": 0.05928892642259598, "global_step": 214842, "epoch": 2413, "lr": 1.0388544328075128e-05}
{"train_loss": 0.08158786594867706, "global_step": 214843, "epoch": 2413, "lr": 1.0388190561690447e-05}
{"train_loss": 0.01855548657476902, "global_step": 214844, "epoch": 2413, "lr": 1.038783680063109e-05}
{"train_loss": 0.053385611926990276, "global_step": 214845, "epoch": 2413, "lr": 1.0387483044897117e-05, "val_loss": 8.94122314453125}
{"train_loss": 0.06113606318831444, "global_step": 214846, "epoch": 2414, "lr": 1.0387129294488551e-05}
{"train_loss": 0.07972165197134018, "global_step": 214847, "epoch": 2414, "lr": 1.038677554940547e-05}
{"train_loss": 0.07765580713748932, "global_step": 214848, "epoch": 2414, "lr": 1.038642180964789e-05}
{"train_loss": 0.07494637370109558, "global_step": 214849, "epoch": 2414, "lr": 1.0386068075215881e-05}
{"train_loss": 0.05189107358455658, "global_step": 214850, "epoch": 2414, "lr": 1.0385714346109488e-05}
{"train_loss": 0.06706796586513519, "global_step": 214851, "epoch": 2414, "lr": 1.0385360622328738e-05}
{"train_loss": 0.014066449366509914, "global_step": 214852, "epoch": 2414, "lr": 1.0385006903873706e-05}
{"train_loss": 0.11994972825050354, "global_step": 214853, "epoch": 2414, "lr": 1.0384653190744414e-05}
{"train_loss": 0.058897629380226135, "global_step": 214854, "epoch": 2414, "lr": 1.0384299482940934e-05}
{"train_loss": 0.04085524007678032, "global_step": 214855, "epoch": 2414, "lr": 1.0383945780463305e-05}
{"train_loss": 0.03610530123114586, "global_step": 214856, "epoch": 2414, "lr": 1.0383592083311555e-05}
{"train_loss": 0.07560060918331146, "global_step": 214857, "epoch": 2414, "lr": 1.0383238391485767e-05}
{"train_loss": 0.06663258373737335, "global_step": 214858, "epoch": 2414, "lr": 1.0382884704985952e-05}
{"train_loss": 0.05484681576490402, "global_step": 214859, "epoch": 2414, "lr": 1.0382531023812192e-05}
{"train_loss": 0.03721252456307411, "global_step": 214860, "epoch": 2414, "lr": 1.03821773479645e-05}
{"train_loss": 0.021991364657878876, "global_step": 214861, "epoch": 2414, "lr": 1.0381823677442942e-05}
{"train_loss": 0.057972121983766556, "global_step": 214862, "epoch": 2414, "lr": 1.038147001224758e-05}
{"train_loss": 0.15416020154953003, "global_step": 214863, "epoch": 2414, "lr": 1.0381116352378434e-05}
{"train_loss": 0.030035261064767838, "global_step": 214864, "epoch": 2414, "lr": 1.0380762697835578e-05}
{"train_loss": 0.039726778864860535, "global_step": 214865, "epoch": 2414, "lr": 1.0380409048619028e-05}
{"train_loss": 0.05529015138745308, "global_step": 214866, "epoch": 2414, "lr": 1.0380055404728867e-05}
{"train_loss": 0.05210498347878456, "global_step": 214867, "epoch": 2414, "lr": 1.0379701766165107e-05}
{"train_loss": 0.05198053643107414, "global_step": 214868, "epoch": 2414, "lr": 1.037934813292783e-05}
{"train_loss": 0.04732483625411987, "global_step": 214869, "epoch": 2414, "lr": 1.0378994505017054e-05}
{"train_loss": 0.061576247215270996, "global_step": 214870, "epoch": 2414, "lr": 1.037864088243285e-05}
{"train_loss": 0.06699245423078537, "global_step": 214871, "epoch": 2414, "lr": 1.0378287265175241e-05}
{"train_loss": 0.08394075930118561, "global_step": 214872, "epoch": 2414, "lr": 1.0377933653244304e-05}
{"train_loss": 0.050806786864995956, "global_step": 214873, "epoch": 2414, "lr": 1.0377580046640056e-05}
{"train_loss": 0.031527601182460785, "global_step": 214874, "epoch": 2414, "lr": 1.0377226445362575e-05}
{"train_loss": 0.06233401596546173, "global_step": 214875, "epoch": 2414, "lr": 1.0376872849411879e-05}
{"train_loss": 0.08174613118171692, "global_step": 214876, "epoch": 2414, "lr": 1.0376519258788037e-05}
{"train_loss": 0.04278354346752167, "global_step": 214877, "epoch": 2414, "lr": 1.0376165673491084e-05}
{"train_loss": 0.10008972883224487, "global_step": 214878, "epoch": 2414, "lr": 1.0375812093521082e-05}
{"train_loss": 0.018326133489608765, "global_step": 214879, "epoch": 2414, "lr": 1.0375458518878056e-05}
{"train_loss": 0.08844250440597534, "global_step": 214880, "epoch": 2414, "lr": 1.0375104949562065e-05}
{"train_loss": 0.06920690089464188, "global_step": 214881, "epoch": 2414, "lr": 1.0374751385573172e-05}
{"train_loss": 0.06569907069206238, "global_step": 214882, "epoch": 2414, "lr": 1.0374397826911402e-05}
{"train_loss": 0.09854059666395187, "global_step": 214883, "epoch": 2414, "lr": 1.0374044273576822e-05}
{"train_loss": 0.08327636867761612, "global_step": 214884, "epoch": 2414, "lr": 1.0373690725569452e-05}
{"train_loss": 0.030181491747498512, "global_step": 214885, "epoch": 2414, "lr": 1.0373337182889376e-05}
{"train_loss": 0.07580714672803879, "global_step": 214886, "epoch": 2414, "lr": 1.0372983645536605e-05}
{"train_loss": 0.04903469979763031, "global_step": 214887, "epoch": 2414, "lr": 1.0372630113511212e-05}
{"train_loss": 0.12469757348299026, "global_step": 214888, "epoch": 2414, "lr": 1.037227658681324e-05}
{"train_loss": 0.05026141181588173, "global_step": 214889, "epoch": 2414, "lr": 1.0371923065442724e-05}
{"train_loss": 0.0316300205886364, "global_step": 214890, "epoch": 2414, "lr": 1.037156954939973e-05}
{"train_loss": 0.052652936428785324, "global_step": 214891, "epoch": 2414, "lr": 1.0371216038684278e-05}
{"train_loss": 0.043147262185811996, "global_step": 214892, "epoch": 2414, "lr": 1.037086253329645e-05}
{"train_loss": 0.06293787807226181, "global_step": 214893, "epoch": 2414, "lr": 1.0370509033236265e-05}
{"train_loss": 0.05868234857916832, "global_step": 214894, "epoch": 2414, "lr": 1.037015553850379e-05}
{"train_loss": 0.05051671713590622, "global_step": 214895, "epoch": 2414, "lr": 1.0369802049099065e-05}
{"train_loss": 0.06323011219501495, "global_step": 214896, "epoch": 2414, "lr": 1.0369448565022127e-05}
{"train_loss": 0.03476584702730179, "global_step": 214897, "epoch": 2414, "lr": 1.0369095086273035e-05}
{"train_loss": 0.07264616340398788, "global_step": 214898, "epoch": 2414, "lr": 1.0368741612851846e-05}
{"train_loss": 0.07901807874441147, "global_step": 214899, "epoch": 2414, "lr": 1.0368388144758584e-05}
{"train_loss": 0.044885262846946716, "global_step": 214900, "epoch": 2414, "lr": 1.0368034681993322e-05}
{"train_loss": 0.08048053830862045, "global_step": 214901, "epoch": 2414, "lr": 1.036768122455608e-05}
{"train_loss": 0.07115202397108078, "global_step": 214902, "epoch": 2414, "lr": 1.0367327772446933e-05}
{"train_loss": 0.029909435659646988, "global_step": 214903, "epoch": 2414, "lr": 1.0366974325665908e-05}
{"train_loss": 0.07018324732780457, "global_step": 214904, "epoch": 2414, "lr": 1.0366620884213069e-05}
{"train_loss": 0.03841252624988556, "global_step": 214905, "epoch": 2414, "lr": 1.0366267448088441e-05}
{"train_loss": 0.07745067775249481, "global_step": 214906, "epoch": 2414, "lr": 1.03659140172921e-05}
{"train_loss": 0.013808663003146648, "global_step": 214907, "epoch": 2414, "lr": 1.0365560591824063e-05}
{"train_loss": 0.05498584732413292, "global_step": 214908, "epoch": 2414, "lr": 1.036520717168441e-05}
{"train_loss": 0.06627580523490906, "global_step": 214909, "epoch": 2414, "lr": 1.036485375687316e-05}
{"train_loss": 0.0377851203083992, "global_step": 214910, "epoch": 2414, "lr": 1.0364500347390382e-05}
{"train_loss": 0.05993323028087616, "global_step": 214911, "epoch": 2414, "lr": 1.0364146943236102e-05}
{"train_loss": 0.06952445209026337, "global_step": 214912, "epoch": 2414, "lr": 1.0363793544410394e-05}
{"train_loss": 0.02695385366678238, "global_step": 214913, "epoch": 2414, "lr": 1.0363440150913278e-05}
{"train_loss": 0.04501783475279808, "global_step": 214914, "epoch": 2414, "lr": 1.0363086762744823e-05}
{"train_loss": 0.06838572770357132, "global_step": 214915, "epoch": 2414, "lr": 1.036273337990506e-05}
{"train_loss": 0.07955252379179001, "global_step": 214916, "epoch": 2414, "lr": 1.036238000239404e-05}
{"train_loss": 0.04563881829380989, "global_step": 214917, "epoch": 2414, "lr": 1.036202663021183e-05}
{"train_loss": 0.026300648227334023, "global_step": 214918, "epoch": 2414, "lr": 1.0361673263358446e-05}
{"train_loss": 0.03836942836642265, "global_step": 214919, "epoch": 2414, "lr": 1.0361319901833972e-05}
{"train_loss": 0.03479604050517082, "global_step": 214920, "epoch": 2414, "lr": 1.0360966545638418e-05}
{"train_loss": 0.06516788899898529, "global_step": 214921, "epoch": 2414, "lr": 1.0360613194771863e-05}
{"train_loss": 0.009094194509088993, "global_step": 214922, "epoch": 2414, "lr": 1.0360259849234322e-05}
{"train_loss": 0.050510603934526443, "global_step": 214923, "epoch": 2414, "lr": 1.035990650902588e-05}
{"train_loss": 0.03940773755311966, "global_step": 214924, "epoch": 2414, "lr": 1.0359553174146552e-05}
{"train_loss": 0.07943601906299591, "global_step": 214925, "epoch": 2414, "lr": 1.0359199844596413e-05}
{"train_loss": 0.08345836400985718, "global_step": 214926, "epoch": 2414, "lr": 1.0358846520375481e-05}
{"train_loss": 0.061971504241228104, "global_step": 214927, "epoch": 2414, "lr": 1.0358493201483832e-05}
{"train_loss": 0.041779760271310806, "global_step": 214928, "epoch": 2414, "lr": 1.0358139887921498e-05}
{"train_loss": 0.10695324093103409, "global_step": 214929, "epoch": 2414, "lr": 1.0357786579688522e-05}
{"train_loss": 0.07647668570280075, "global_step": 214930, "epoch": 2414, "lr": 1.0357433276784967e-05}
{"train_loss": 0.05551975220441818, "global_step": 214931, "epoch": 2414, "lr": 1.035707997921086e-05}
{"train_loss": 0.04555564001202583, "global_step": 214932, "epoch": 2414, "lr": 1.0356726686966273e-05}
{"train_loss": 0.05120180547237396, "global_step": 214933, "epoch": 2414, "lr": 1.035637340005124e-05}
{"train_loss": 0.05826872893807928, "global_step": 214934, "epoch": 2414, "lr": 1.0356020118465798e-05, "val_loss": 8.916194915771484}
{"train_loss": 0.051346421241760254, "global_step": 214935, "epoch": 2415, "lr": 1.0355666842210004e-05}
{"train_loss": 0.04484659805893898, "global_step": 214936, "epoch": 2415, "lr": 1.0355313571283925e-05}
{"train_loss": 0.051437102258205414, "global_step": 214937, "epoch": 2415, "lr": 1.0354960305687572e-05}
{"train_loss": 0.018101943656802177, "global_step": 214938, "epoch": 2415, "lr": 1.0354607045421034e-05}
{"train_loss": 0.03362482041120529, "global_step": 214939, "epoch": 2415, "lr": 1.0354253790484314e-05}
{"train_loss": 0.0817948654294014, "global_step": 214940, "epoch": 2415, "lr": 1.0353900540877492e-05}
{"train_loss": 0.03217504173517227, "global_step": 214941, "epoch": 2415, "lr": 1.0353547296600603e-05}
{"train_loss": 0.03432672843337059, "global_step": 214942, "epoch": 2415, "lr": 1.0353194057653699e-05}
{"train_loss": 0.04550224542617798, "global_step": 214943, "epoch": 2415, "lr": 1.0352840824036814e-05}
{"train_loss": 0.05383794754743576, "global_step": 214944, "epoch": 2415, "lr": 1.0352487595750021e-05}
{"train_loss": 0.07018199563026428, "global_step": 214945, "epoch": 2415, "lr": 1.0352134372793342e-05}
{"train_loss": 0.06105351448059082, "global_step": 214946, "epoch": 2415, "lr": 1.0351781155166845e-05}
{"train_loss": 0.0511779859662056, "global_step": 214947, "epoch": 2415, "lr": 1.0351427942870556e-05}
{"train_loss": 0.06780777871608734, "global_step": 214948, "epoch": 2415, "lr": 1.0351074735904548e-05}
{"train_loss": 0.0711682066321373, "global_step": 214949, "epoch": 2415, "lr": 1.0350721534268842e-05}
{"train_loss": 0.025842513889074326, "global_step": 214950, "epoch": 2415, "lr": 1.0350368337963506e-05}
{"train_loss": 0.03517689183354378, "global_step": 214951, "epoch": 2415, "lr": 1.0350015146988574e-05}
{"train_loss": 0.028312334790825844, "global_step": 214952, "epoch": 2415, "lr": 1.0349661961344092e-05}
{"train_loss": 0.023882267996668816, "global_step": 214953, "epoch": 2415, "lr": 1.0349308781030137e-05}
{"train_loss": 0.07099933177232742, "global_step": 214954, "epoch": 2415, "lr": 1.0348955606046713e-05}
{"train_loss": 0.055055730044841766, "global_step": 214955, "epoch": 2415, "lr": 1.0348602436393905e-05}
{"train_loss": 0.052276238799095154, "global_step": 214956, "epoch": 2415, "lr": 1.034824927207173e-05}
{"train_loss": 0.08181833475828171, "global_step": 214957, "epoch": 2415, "lr": 1.0347896113080263e-05}
{"train_loss": 0.07061034440994263, "global_step": 214958, "epoch": 2415, "lr": 1.0347542959419525e-05}
{"train_loss": 0.08369079232215881, "global_step": 214959, "epoch": 2415, "lr": 1.0347189811089592e-05}
{"train_loss": 0.03391953557729721, "global_step": 214960, "epoch": 2415, "lr": 1.0346836668090482e-05}
{"train_loss": 0.04579942300915718, "global_step": 214961, "epoch": 2415, "lr": 1.0346483530422269e-05}
{"train_loss": 0.0795813649892807, "global_step": 214962, "epoch": 2415, "lr": 1.034613039808497e-05}
{"train_loss": 0.0767090767621994, "global_step": 214963, "epoch": 2415, "lr": 1.0345777271078672e-05}
{"train_loss": 0.0748334527015686, "global_step": 214964, "epoch": 2415, "lr": 1.0345424149403383e-05}
{"train_loss": 0.11481600254774094, "global_step": 214965, "epoch": 2415, "lr": 1.0345071033059179e-05}
{"train_loss": 0.044853027909994125, "global_step": 214966, "epoch": 2415, "lr": 1.03447179220461e-05}
{"train_loss": 0.05122361332178116, "global_step": 214967, "epoch": 2415, "lr": 1.0344364816364177e-05}
{"train_loss": 0.047419823706150055, "global_step": 214968, "epoch": 2415, "lr": 1.0344011716013486e-05}
{"train_loss": 0.10357147455215454, "global_step": 214969, "epoch": 2415, "lr": 1.0343658620994046e-05}
{"train_loss": 0.050500303506851196, "global_step": 214970, "epoch": 2415, "lr": 1.0343305531305924e-05}
{"train_loss": 0.044535622000694275, "global_step": 214971, "epoch": 2415, "lr": 1.0342952446949155e-05}
{"train_loss": 0.0358455665409565, "global_step": 214972, "epoch": 2415, "lr": 1.0342599367923806e-05}
{"train_loss": 0.04207799583673477, "global_step": 214973, "epoch": 2415, "lr": 1.0342246294229896e-05}
{"train_loss": 0.05129125714302063, "global_step": 214974, "epoch": 2415, "lr": 1.03418932258675e-05}
{"train_loss": 0.0616544634103775, "global_step": 214975, "epoch": 2415, "lr": 1.0341540162836643e-05}
{"train_loss": 0.039623159915208817, "global_step": 214976, "epoch": 2415, "lr": 1.034118710513739e-05}
{"train_loss": 0.025661420077085495, "global_step": 214977, "epoch": 2415, "lr": 1.034083405276977e-05}
{"train_loss": 0.04350215196609497, "global_step": 214978, "epoch": 2415, "lr": 1.0340481005733859e-05}
{"train_loss": 0.07828544080257416, "global_step": 214979, "epoch": 2415, "lr": 1.0340127964029673e-05}
{"train_loss": 0.06066048517823219, "global_step": 214980, "epoch": 2415, "lr": 1.0339774927657281e-05}
{"train_loss": 0.05503350496292114, "global_step": 214981, "epoch": 2415, "lr": 1.0339421896616713e-05}
{"train_loss": 0.052629463374614716, "global_step": 214982, "epoch": 2415, "lr": 1.0339068870908041e-05}
{"train_loss": 0.05485039949417114, "global_step": 214983, "epoch": 2415, "lr": 1.0338715850531284e-05}
{"train_loss": 0.04676223546266556, "global_step": 214984, "epoch": 2415, "lr": 1.0338362835486515e-05}
{"train_loss": 0.0356675423681736, "global_step": 214985, "epoch": 2415, "lr": 1.0338009825773758e-05}
{"train_loss": 0.07703433930873871, "global_step": 214986, "epoch": 2415, "lr": 1.0337656821393083e-05}
{"train_loss": 0.09846862405538559, "global_step": 214987, "epoch": 2415, "lr": 1.0337303822344519e-05}
{"train_loss": 0.027465706691145897, "global_step": 214988, "epoch": 2415, "lr": 1.0336950828628117e-05}
{"train_loss": 0.08155663311481476, "global_step": 214989, "epoch": 2415, "lr": 1.033659784024394e-05}
{"train_loss": 0.0703512504696846, "global_step": 214990, "epoch": 2415, "lr": 1.0336244857192013e-05}
{"train_loss": 0.055304672569036484, "global_step": 214991, "epoch": 2415, "lr": 1.033589187947241e-05}
{"train_loss": 0.09723272174596786, "global_step": 214992, "epoch": 2415, "lr": 1.033553890708514e-05}
{"train_loss": 0.03644471988081932, "global_step": 214993, "epoch": 2415, "lr": 1.0335185940030295e-05}
{"train_loss": 0.057156603783369064, "global_step": 214994, "epoch": 2415, "lr": 1.033483297830789e-05}
{"train_loss": 0.05902744457125664, "global_step": 214995, "epoch": 2415, "lr": 1.0334480021917992e-05}
{"train_loss": 0.0693775936961174, "global_step": 214996, "epoch": 2415, "lr": 1.0334127070860627e-05}
{"train_loss": 0.03669555485248566, "global_step": 214997, "epoch": 2415, "lr": 1.0333774125135865e-05}
{"train_loss": 0.07506129890680313, "global_step": 214998, "epoch": 2415, "lr": 1.0333421184743736e-05}
{"train_loss": 0.037119060754776, "global_step": 214999, "epoch": 2415, "lr": 1.0333068249684308e-05}
{"train_loss": 0.07393423467874527, "global_step": 215000, "epoch": 2415, "lr": 1.0332715319957598e-05}
{"train_loss": 0.03146664798259735, "global_step": 215001, "epoch": 2415, "lr": 1.0332362395563689e-05}
{"train_loss": 0.04593246802687645, "global_step": 215002, "epoch": 2415, "lr": 1.033200947650259e-05}
{"train_loss": 0.033433400094509125, "global_step": 215003, "epoch": 2415, "lr": 1.0331656562774388e-05}
{"train_loss": 0.04813072085380554, "global_step": 215004, "epoch": 2415, "lr": 1.033130365437911e-05}
{"train_loss": 0.03151897341012955, "global_step": 215005, "epoch": 2415, "lr": 1.0330950751316792e-05}
{"train_loss": 0.07347527891397476, "global_step": 215006, "epoch": 2415, "lr": 1.033059785358751e-05}
{"train_loss": 0.06834296882152557, "global_step": 215007, "epoch": 2415, "lr": 1.0330244961191276e-05}
{"train_loss": 0.05354318767786026, "global_step": 215008, "epoch": 2415, "lr": 1.0329892074128179e-05}
{"train_loss": 0.05048263818025589, "global_step": 215009, "epoch": 2415, "lr": 1.0329539192398225e-05}
{"train_loss": 0.058847296983003616, "global_step": 215010, "epoch": 2415, "lr": 1.0329186316001493e-05}
{"train_loss": 0.10298413038253784, "global_step": 215011, "epoch": 2415, "lr": 1.0328833444938007e-05}
{"train_loss": 0.0719555914402008, "global_step": 215012, "epoch": 2415, "lr": 1.032848057920784e-05}
{"train_loss": 0.05182665213942528, "global_step": 215013, "epoch": 2415, "lr": 1.032812771881101e-05}
{"train_loss": 0.06537369638681412, "global_step": 215014, "epoch": 2415, "lr": 1.0327774863747598e-05}
{"train_loss": 0.04783095419406891, "global_step": 215015, "epoch": 2415, "lr": 1.0327422014017613e-05}
{"train_loss": 0.041668374091386795, "global_step": 215016, "epoch": 2415, "lr": 1.032706916962114e-05}
{"train_loss": 0.0383889339864254, "global_step": 215017, "epoch": 2415, "lr": 1.0326716330558195e-05}
{"train_loss": 0.07588502764701843, "global_step": 215018, "epoch": 2415, "lr": 1.0326363496828851e-05}
{"train_loss": 0.0071633122861385345, "global_step": 215019, "epoch": 2415, "lr": 1.0326010668433133e-05}
{"train_loss": 0.038426853716373444, "global_step": 215020, "epoch": 2415, "lr": 1.0325657845371112e-05}
{"train_loss": 0.04135453701019287, "global_step": 215021, "epoch": 2415, "lr": 1.0325305027642806e-05}
{"train_loss": 0.08179763704538345, "global_step": 215022, "epoch": 2415, "lr": 1.0324952215248296e-05}
{"train_loss": 0.055123959364516015, "global_step": 215023, "epoch": 2415, "lr": 1.0324599408187601e-05, "val_loss": 8.837520599365234, "train_action_mse_error": 14.983878135681152}
{"train_loss": 0.06516715884208679, "global_step": 215024, "epoch": 2416, "lr": 1.0324246606460775e-05}
{"train_loss": 0.05847250670194626, "global_step": 215025, "epoch": 2416, "lr": 1.0323893810067886e-05}
{"train_loss": 0.10018717497587204, "global_step": 215026, "epoch": 2416, "lr": 1.0323541019008953e-05}
{"train_loss": 0.06692009419202805, "global_step": 215027, "epoch": 2416, "lr": 1.0323188233284053e-05}
{"train_loss": 0.074726901948452, "global_step": 215028, "epoch": 2416, "lr": 1.0322835452893199e-05}
{"train_loss": 0.050090037286281586, "global_step": 215029, "epoch": 2416, "lr": 1.032248267783647e-05}
{"train_loss": 0.06626075506210327, "global_step": 215030, "epoch": 2416, "lr": 1.0322129908113887e-05}
{"train_loss": 0.07840525358915329, "global_step": 215031, "epoch": 2416, "lr": 1.0321777143725525e-05}
{"train_loss": 0.009103837423026562, "global_step": 215032, "epoch": 2416, "lr": 1.0321424384671402e-05}
{"train_loss": 0.0406154990196228, "global_step": 215033, "epoch": 2416, "lr": 1.0321071630951596e-05}
{"train_loss": 0.07660852372646332, "global_step": 215034, "epoch": 2416, "lr": 1.0320718882566117e-05}
{"train_loss": 0.03876161202788353, "global_step": 215035, "epoch": 2416, "lr": 1.0320366139515053e-05}
{"train_loss": 0.07021085172891617, "global_step": 215036, "epoch": 2416, "lr": 1.0320013401798417e-05}
{"train_loss": 0.09299282729625702, "global_step": 215037, "epoch": 2416, "lr": 1.0319660669416287e-05}
{"train_loss": 0.10161793977022171, "global_step": 215038, "epoch": 2416, "lr": 1.0319307942368678e-05}
{"train_loss": 0.05523844435811043, "global_step": 215039, "epoch": 2416, "lr": 1.0318955220655673e-05}
{"train_loss": 0.039318718016147614, "global_step": 215040, "epoch": 2416, "lr": 1.0318602504277281e-05}
{"train_loss": 0.047719310969114304, "global_step": 215041, "epoch": 2416, "lr": 1.0318249793233587e-05}
{"train_loss": 0.010817828588187695, "global_step": 215042, "epoch": 2416, "lr": 1.0317897087524609e-05}
{"train_loss": 0.051220789551734924, "global_step": 215043, "epoch": 2416, "lr": 1.0317544387150413e-05}
{"train_loss": 0.07457724213600159, "global_step": 215044, "epoch": 2416, "lr": 1.0317191692111045e-05}
{"train_loss": 0.03527175262570381, "global_step": 215045, "epoch": 2416, "lr": 1.0316839002406532e-05}
{"train_loss": 0.03257904201745987, "global_step": 215046, "epoch": 2416, "lr": 1.031648631803695e-05}
{"train_loss": 0.11872658878564835, "global_step": 215047, "epoch": 2416, "lr": 1.0316133639002317e-05}
{"train_loss": 0.038089316338300705, "global_step": 215048, "epoch": 2416, "lr": 1.0315780965302713e-05}
{"train_loss": 0.015609745867550373, "global_step": 215049, "epoch": 2416, "lr": 1.0315428296938151e-05}
{"train_loss": 0.04275742918252945, "global_step": 215050, "epoch": 2416, "lr": 1.0315075633908716e-05}
{"train_loss": 0.037955477833747864, "global_step": 215051, "epoch": 2416, "lr": 1.0314722976214415e-05}
{"train_loss": 0.05703720450401306, "global_step": 215052, "epoch": 2416, "lr": 1.0314370323855333e-05}
{"train_loss": 0.024930419400334358, "global_step": 215053, "epoch": 2416, "lr": 1.0314017676831483e-05}
{"train_loss": 0.04136186093091965, "global_step": 215054, "epoch": 2416, "lr": 1.031366503514295e-05}
{"train_loss": 0.07114807516336441, "global_step": 215055, "epoch": 2416, "lr": 1.0313312398789743e-05}
{"train_loss": 0.0880347192287445, "global_step": 215056, "epoch": 2416, "lr": 1.031295976777194e-05}
{"train_loss": 0.046969927847385406, "global_step": 215057, "epoch": 2416, "lr": 1.0312607142089564e-05}
{"train_loss": 0.023510389029979706, "global_step": 215058, "epoch": 2416, "lr": 1.0312254521742687e-05}
{"train_loss": 0.11394192278385162, "global_step": 215059, "epoch": 2416, "lr": 1.0311901906731331e-05}
{"train_loss": 0.08543436229228973, "global_step": 215060, "epoch": 2416, "lr": 1.0311549297055573e-05}
{"train_loss": 0.031005408614873886, "global_step": 215061, "epoch": 2416, "lr": 1.0311196692715424e-05}
{"train_loss": 0.07743509113788605, "global_step": 215062, "epoch": 2416, "lr": 1.0310844093710953e-05}
{"train_loss": 0.04554950073361397, "global_step": 215063, "epoch": 2416, "lr": 1.0310491500042219e-05}
{"train_loss": 0.06226334720849991, "global_step": 215064, "epoch": 2416, "lr": 1.0310138911709244e-05}
{"train_loss": 0.0412902757525444, "global_step": 215065, "epoch": 2416, "lr": 1.0309786328712095e-05}
{"train_loss": 0.04550223797559738, "global_step": 215066, "epoch": 2416, "lr": 1.0309433751050801e-05}
{"train_loss": 0.05195214971899986, "global_step": 215067, "epoch": 2416, "lr": 1.0309081178725433e-05}
{"train_loss": 0.038877878338098526, "global_step": 215068, "epoch": 2416, "lr": 1.0308728611736012e-05}
{"train_loss": 0.03266831859946251, "global_step": 215069, "epoch": 2416, "lr": 1.0308376050082612e-05}
{"train_loss": 0.0697941780090332, "global_step": 215070, "epoch": 2416, "lr": 1.0308023493765256e-05}
{"train_loss": 0.052008263766765594, "global_step": 215071, "epoch": 2416, "lr": 1.0307670942784014e-05}
{"train_loss": 0.07974373549222946, "global_step": 215072, "epoch": 2416, "lr": 1.030731839713891e-05}
{"train_loss": 0.01717926561832428, "global_step": 215073, "epoch": 2416, "lr": 1.0306965856830014e-05}
{"train_loss": 0.07621563971042633, "global_step": 215074, "epoch": 2416, "lr": 1.0306613321857356e-05}
{"train_loss": 0.05043255537748337, "global_step": 215075, "epoch": 2416, "lr": 1.0306260792220995e-05}
{"train_loss": 0.033702436834573746, "global_step": 215076, "epoch": 2416, "lr": 1.0305908267920966e-05}
{"train_loss": 0.10850638896226883, "global_step": 215077, "epoch": 2416, "lr": 1.0305555748957335e-05}
{"train_loss": 0.034029826521873474, "global_step": 215078, "epoch": 2416, "lr": 1.0305203235330124e-05}
{"train_loss": 0.0335506908595562, "global_step": 215079, "epoch": 2416, "lr": 1.0304850727039406e-05}
{"train_loss": 0.03370191901922226, "global_step": 215080, "epoch": 2416, "lr": 1.0304498224085208e-05}
{"train_loss": 0.04753836616873741, "global_step": 215081, "epoch": 2416, "lr": 1.0304145726467595e-05}
{"train_loss": 0.04607028886675835, "global_step": 215082, "epoch": 2416, "lr": 1.0303793234186609e-05}
{"train_loss": 0.06656844168901443, "global_step": 215083, "epoch": 2416, "lr": 1.0303440747242283e-05}
{"train_loss": 0.09532703459262848, "global_step": 215084, "epoch": 2416, "lr": 1.0303088265634686e-05}
{"train_loss": 0.04158906638622284, "global_step": 215085, "epoch": 2416, "lr": 1.0302735789363837e-05}
{"train_loss": 0.04587244614958763, "global_step": 215086, "epoch": 2416, "lr": 1.030238331842982e-05}
{"train_loss": 0.033417221158742905, "global_step": 215087, "epoch": 2416, "lr": 1.030203085283265e-05}
{"train_loss": 0.04888468608260155, "global_step": 215088, "epoch": 2416, "lr": 1.03016783925724e-05}
{"train_loss": 0.047617167234420776, "global_step": 215089, "epoch": 2416, "lr": 1.0301325937649099e-05}
{"train_loss": 0.04756985604763031, "global_step": 215090, "epoch": 2416, "lr": 1.0300973488062805e-05}
{"train_loss": 0.06386759877204895, "global_step": 215091, "epoch": 2416, "lr": 1.0300621043813552e-05}
{"train_loss": 0.04617279767990112, "global_step": 215092, "epoch": 2416, "lr": 1.0300268604901408e-05}
{"train_loss": 0.03053351305425167, "global_step": 215093, "epoch": 2416, "lr": 1.0299916171326401e-05}
{"train_loss": 0.0773424282670021, "global_step": 215094, "epoch": 2416, "lr": 1.0299563743088592e-05}
{"train_loss": 0.04908899590373039, "global_step": 215095, "epoch": 2416, "lr": 1.0299211320188012e-05}
{"train_loss": 0.05459362640976906, "global_step": 215096, "epoch": 2416, "lr": 1.0298858902624737e-05}
{"train_loss": 0.047452643513679504, "global_step": 215097, "epoch": 2416, "lr": 1.0298506490398779e-05}
{"train_loss": 0.05862821638584137, "global_step": 215098, "epoch": 2416, "lr": 1.0298154083510203e-05}
{"train_loss": 0.0434376522898674, "global_step": 215099, "epoch": 2416, "lr": 1.0297801681959075e-05}
{"train_loss": 0.02925685979425907, "global_step": 215100, "epoch": 2416, "lr": 1.0297449285745402e-05}
{"train_loss": 0.05362210050225258, "global_step": 215101, "epoch": 2416, "lr": 1.0297096894869274e-05}
{"train_loss": 0.0379827506840229, "global_step": 215102, "epoch": 2416, "lr": 1.0296744509330697e-05}
{"train_loss": 0.020521625876426697, "global_step": 215103, "epoch": 2416, "lr": 1.0296392129129756e-05}
{"train_loss": 0.046928685158491135, "global_step": 215104, "epoch": 2416, "lr": 1.0296039754266467e-05}
{"train_loss": 0.048946622759103775, "global_step": 215105, "epoch": 2416, "lr": 1.0295687384740904e-05}
{"train_loss": 0.05411093682050705, "global_step": 215106, "epoch": 2416, "lr": 1.0295335020553094e-05}
{"train_loss": 0.02533242106437683, "global_step": 215107, "epoch": 2416, "lr": 1.02949826617031e-05}
{"train_loss": 0.05486094206571579, "global_step": 215108, "epoch": 2416, "lr": 1.029463030819095e-05}
{"train_loss": 0.08230265229940414, "global_step": 215109, "epoch": 2416, "lr": 1.0294277960016713e-05}
{"train_loss": 0.08610648661851883, "global_step": 215110, "epoch": 2416, "lr": 1.0293925617180416e-05}
{"train_loss": 0.01484007015824318, "global_step": 215111, "epoch": 2416, "lr": 1.0293573279682134e-05}
{"train_loss": 0.053335574455559254, "global_step": 215112, "epoch": 2416, "lr": 1.0293220947521876e-05, "val_loss": 8.860459327697754}
{"train_loss": 0.029234353452920914, "global_step": 215113, "epoch": 2417, "lr": 1.029286862069973e-05}
{"train_loss": 0.047603607177734375, "global_step": 215114, "epoch": 2417, "lr": 1.0292516299215704e-05}
{"train_loss": 0.017599327489733696, "global_step": 215115, "epoch": 2417, "lr": 1.0292163983069886e-05}
{"train_loss": 0.04662537947297096, "global_step": 215116, "epoch": 2417, "lr": 1.0291811672262287e-05}
{"train_loss": 0.029634106904268265, "global_step": 215117, "epoch": 2417, "lr": 1.0291459366792982e-05}
{"train_loss": 0.0364459790289402, "global_step": 215118, "epoch": 2417, "lr": 1.0291107066661992e-05}
{"train_loss": 0.05049322172999382, "global_step": 215119, "epoch": 2417, "lr": 1.0290754771869392e-05}
{"train_loss": 0.047676797956228256, "global_step": 215120, "epoch": 2417, "lr": 1.0290402482415217e-05}
{"train_loss": 0.07221002131700516, "global_step": 215121, "epoch": 2417, "lr": 1.02900501982995e-05}
{"train_loss": 0.10092316567897797, "global_step": 215122, "epoch": 2417, "lr": 1.0289697919522317e-05}
{"train_loss": 0.0587085522711277, "global_step": 215123, "epoch": 2417, "lr": 1.0289345646083687e-05}
{"train_loss": 0.06561519205570221, "global_step": 215124, "epoch": 2417, "lr": 1.028899337798368e-05}
{"train_loss": 0.040215883404016495, "global_step": 215125, "epoch": 2417, "lr": 1.0288641115222324e-05}
{"train_loss": 0.07519278675317764, "global_step": 215126, "epoch": 2417, "lr": 1.0288288857799688e-05}
{"train_loss": 0.023219456896185875, "global_step": 215127, "epoch": 2417, "lr": 1.0287936605715792e-05}
{"train_loss": 0.036963559687137604, "global_step": 215128, "epoch": 2417, "lr": 1.0287584358970714e-05}
{"train_loss": 0.03607868403196335, "global_step": 215129, "epoch": 2417, "lr": 1.0287232117564476e-05}
{"train_loss": 0.07170359045267105, "global_step": 215130, "epoch": 2417, "lr": 1.0286879881497146e-05}
{"train_loss": 0.10260578989982605, "global_step": 215131, "epoch": 2417, "lr": 1.0286527650768747e-05}
{"train_loss": 0.05963076278567314, "global_step": 215132, "epoch": 2417, "lr": 1.0286175425379358e-05}
{"train_loss": 0.04080614820122719, "global_step": 215133, "epoch": 2417, "lr": 1.0285823205328988e-05}
{"train_loss": 0.10312096029520035, "global_step": 215134, "epoch": 2417, "lr": 1.0285470990617713e-05}
{"train_loss": 0.042705051600933075, "global_step": 215135, "epoch": 2417, "lr": 1.0285118781245579e-05}
{"train_loss": 0.09114840626716614, "global_step": 215136, "epoch": 2417, "lr": 1.0284766577212618e-05}
{"train_loss": 0.07526560127735138, "global_step": 215137, "epoch": 2417, "lr": 1.0284414378518898e-05}
{"train_loss": 0.08400566875934601, "global_step": 215138, "epoch": 2417, "lr": 1.028406218516444e-05}
{"train_loss": 0.07205317169427872, "global_step": 215139, "epoch": 2417, "lr": 1.0283709997149321e-05}
{"train_loss": 0.07136854529380798, "global_step": 215140, "epoch": 2417, "lr": 1.0283357814473554e-05}
{"train_loss": 0.04023455083370209, "global_step": 215141, "epoch": 2417, "lr": 1.0283005637137227e-05}
{"train_loss": 0.05399244278669357, "global_step": 215142, "epoch": 2417, "lr": 1.0282653465140346e-05}
{"train_loss": 0.026306793093681335, "global_step": 215143, "epoch": 2417, "lr": 1.0282301298482993e-05}
{"train_loss": 0.019465995952486992, "global_step": 215144, "epoch": 2417, "lr": 1.0281949137165191e-05}
{"train_loss": 0.07447326928377151, "global_step": 215145, "epoch": 2417, "lr": 1.0281596981187009e-05}
{"train_loss": 0.05509093403816223, "global_step": 215146, "epoch": 2417, "lr": 1.0281244830548464e-05}
{"train_loss": 0.03390055149793625, "global_step": 215147, "epoch": 2417, "lr": 1.0280892685249637e-05}
{"train_loss": 0.05692123994231224, "global_step": 215148, "epoch": 2417, "lr": 1.0280540545290552e-05}
{"train_loss": 0.03927740082144737, "global_step": 215149, "epoch": 2417, "lr": 1.028018841067127e-05}
{"train_loss": 0.0642041489481926, "global_step": 215150, "epoch": 2417, "lr": 1.0279836281391824e-05}
{"train_loss": 0.021803775802254677, "global_step": 215151, "epoch": 2417, "lr": 1.0279484157452268e-05}
{"train_loss": 0.06049163639545441, "global_step": 215152, "epoch": 2417, "lr": 1.0279132038852668e-05}
{"train_loss": 0.09063522517681122, "global_step": 215153, "epoch": 2417, "lr": 1.0278779925593035e-05}
{"train_loss": 0.06539083272218704, "global_step": 215154, "epoch": 2417, "lr": 1.0278427817673452e-05}
{"train_loss": 0.03426995500922203, "global_step": 215155, "epoch": 2417, "lr": 1.0278075715093955e-05}
{"train_loss": 0.058995168656110764, "global_step": 215156, "epoch": 2417, "lr": 1.0277723617854568e-05}
{"train_loss": 0.04504966363310814, "global_step": 215157, "epoch": 2417, "lr": 1.0277371525955371e-05}
{"train_loss": 0.02185959555208683, "global_step": 215158, "epoch": 2417, "lr": 1.0277019439396385e-05}
{"train_loss": 0.08311545848846436, "global_step": 215159, "epoch": 2417, "lr": 1.0276667358177683e-05}
{"train_loss": 0.051686253398656845, "global_step": 215160, "epoch": 2417, "lr": 1.0276315282299298e-05}
{"train_loss": 0.033957645297050476, "global_step": 215161, "epoch": 2417, "lr": 1.0275963211761264e-05}
{"train_loss": 0.05038461461663246, "global_step": 215162, "epoch": 2417, "lr": 1.0275611146563658e-05}
{"train_loss": 0.05533765256404877, "global_step": 215163, "epoch": 2417, "lr": 1.0275259086706496e-05}
{"train_loss": 0.08497448265552521, "global_step": 215164, "epoch": 2417, "lr": 1.0274907032189857e-05}
{"train_loss": 0.03801541030406952, "global_step": 215165, "epoch": 2417, "lr": 1.0274554983013757e-05}
{"train_loss": 0.03673188015818596, "global_step": 215166, "epoch": 2417, "lr": 1.0274202939178273e-05}
{"train_loss": 0.038438666611909866, "global_step": 215167, "epoch": 2417, "lr": 1.0273850900683424e-05}
{"train_loss": 0.07732225954532623, "global_step": 215168, "epoch": 2417, "lr": 1.0273498867529285e-05}
{"train_loss": 0.0704708993434906, "global_step": 215169, "epoch": 2417, "lr": 1.0273146839715881e-05}
{"train_loss": 0.04073016718029976, "global_step": 215170, "epoch": 2417, "lr": 1.0272794817243264e-05}
{"train_loss": 0.03321722894906998, "global_step": 215171, "epoch": 2417, "lr": 1.0272442800111498e-05}
{"train_loss": 0.03127836063504219, "global_step": 215172, "epoch": 2417, "lr": 1.027209078832061e-05}
{"train_loss": 0.0665900856256485, "global_step": 215173, "epoch": 2417, "lr": 1.0271738781870666e-05}
{"train_loss": 0.04688598960638046, "global_step": 215174, "epoch": 2417, "lr": 1.0271386780761682e-05}
{"train_loss": 0.07634749263525009, "global_step": 215175, "epoch": 2417, "lr": 1.0271034784993749e-05}
{"train_loss": 0.036628223955631256, "global_step": 215176, "epoch": 2417, "lr": 1.0270682794566872e-05}
{"train_loss": 0.04004966840147972, "global_step": 215177, "epoch": 2417, "lr": 1.0270330809481132e-05}
{"train_loss": 0.04405638203024864, "global_step": 215178, "epoch": 2417, "lr": 1.0269978829736555e-05}
{"train_loss": 0.03186995908617973, "global_step": 215179, "epoch": 2417, "lr": 1.0269626855333198e-05}
{"train_loss": 0.10928093641996384, "global_step": 215180, "epoch": 2417, "lr": 1.0269274886271097e-05}
{"train_loss": 0.04667028412222862, "global_step": 215181, "epoch": 2417, "lr": 1.0268922922550323e-05}
{"train_loss": 0.04510199651122093, "global_step": 215182, "epoch": 2417, "lr": 1.0268570964170894e-05}
{"train_loss": 0.05113685876131058, "global_step": 215183, "epoch": 2417, "lr": 1.0268219011132879e-05}
{"train_loss": 0.04937266185879707, "global_step": 215184, "epoch": 2417, "lr": 1.026786706343631e-05}
{"train_loss": 0.08595900237560272, "global_step": 215185, "epoch": 2417, "lr": 1.0267515121081256e-05}
{"train_loss": 0.04545117914676666, "global_step": 215186, "epoch": 2417, "lr": 1.0267163184067735e-05}
{"train_loss": 0.04357955604791641, "global_step": 215187, "epoch": 2417, "lr": 1.026681125239582e-05}
{"train_loss": 0.12266207486391068, "global_step": 215188, "epoch": 2417, "lr": 1.0266459326065537e-05}
{"train_loss": 0.09147815406322479, "global_step": 215189, "epoch": 2417, "lr": 1.0266107405076947e-05}
{"train_loss": 0.04969931021332741, "global_step": 215190, "epoch": 2417, "lr": 1.0265755489430107e-05}
{"train_loss": 0.10545435547828674, "global_step": 215191, "epoch": 2417, "lr": 1.0265403579125039e-05}
{"train_loss": 0.08404005318880081, "global_step": 215192, "epoch": 2417, "lr": 1.0265051674161814e-05}
{"train_loss": 0.03414708003401756, "global_step": 215193, "epoch": 2417, "lr": 1.0264699774540465e-05}
{"train_loss": 0.05073624476790428, "global_step": 215194, "epoch": 2417, "lr": 1.0264347880261038e-05}
{"train_loss": 0.09249851107597351, "global_step": 215195, "epoch": 2417, "lr": 1.0263995991323594e-05}
{"train_loss": 0.03235911577939987, "global_step": 215196, "epoch": 2417, "lr": 1.0263644107728154e-05}
{"train_loss": 0.0835287943482399, "global_step": 215197, "epoch": 2417, "lr": 1.0263292229474802e-05}
{"train_loss": 0.061428725719451904, "global_step": 215198, "epoch": 2417, "lr": 1.0262940356563566e-05}
{"train_loss": 0.0826842337846756, "global_step": 215199, "epoch": 2417, "lr": 1.026258848899448e-05}
{"train_loss": 0.08925478160381317, "global_step": 215200, "epoch": 2417, "lr": 1.0262236626767613e-05}
{"train_loss": 0.05681832948846094, "global_step": 215201, "epoch": 2417, "lr": 1.0261884769882996e-05, "val_loss": 8.878107070922852}
{"train_loss": 0.0537094920873642, "global_step": 215202, "epoch": 2418, "lr": 1.0261532918340695e-05}
{"train_loss": 0.09260676801204681, "global_step": 215203, "epoch": 2418, "lr": 1.0261181072140735e-05}
{"train_loss": 0.040608473122119904, "global_step": 215204, "epoch": 2418, "lr": 1.0260829231283186e-05}
{"train_loss": 0.06112160161137581, "global_step": 215205, "epoch": 2418, "lr": 1.026047739576807e-05}
{"train_loss": 0.060605693608522415, "global_step": 215206, "epoch": 2418, "lr": 1.0260125565595453e-05}
{"train_loss": 0.07621146738529205, "global_step": 215207, "epoch": 2418, "lr": 1.0259773740765389e-05}
{"train_loss": 0.08558111637830734, "global_step": 215208, "epoch": 2418, "lr": 1.02594219212779e-05}
{"train_loss": 0.044360220432281494, "global_step": 215209, "epoch": 2418, "lr": 1.0259070107133062e-05}
{"train_loss": 0.025289444252848625, "global_step": 215210, "epoch": 2418, "lr": 1.0258718298330889e-05}
{"train_loss": 0.03854738920927048, "global_step": 215211, "epoch": 2418, "lr": 1.0258366494871463e-05}
{"train_loss": 0.06613793969154358, "global_step": 215212, "epoch": 2418, "lr": 1.0258014696754809e-05}
{"train_loss": 0.027219120413064957, "global_step": 215213, "epoch": 2418, "lr": 1.0257662903980986e-05}
{"train_loss": 0.05067228153347969, "global_step": 215214, "epoch": 2418, "lr": 1.0257311116550022e-05}
{"train_loss": 0.0902751088142395, "global_step": 215215, "epoch": 2418, "lr": 1.0256959334461996e-05}
{"train_loss": 0.044166337698698044, "global_step": 215216, "epoch": 2418, "lr": 1.0256607557716925e-05}
{"train_loss": 0.055164825171232224, "global_step": 215217, "epoch": 2418, "lr": 1.0256255786314878e-05}
{"train_loss": 0.08684328943490982, "global_step": 215218, "epoch": 2418, "lr": 1.025590402025588e-05}
{"train_loss": 0.051975347101688385, "global_step": 215219, "epoch": 2418, "lr": 1.0255552259540008e-05}
{"train_loss": 0.056550949811935425, "global_step": 215220, "epoch": 2418, "lr": 1.0255200504167273e-05}
{"train_loss": 0.054354868829250336, "global_step": 215221, "epoch": 2418, "lr": 1.0254848754137758e-05}
{"train_loss": 0.055399950593709946, "global_step": 215222, "epoch": 2418, "lr": 1.0254497009451487e-05}
{"train_loss": 0.07807572185993195, "global_step": 215223, "epoch": 2418, "lr": 1.0254145270108518e-05}
{"train_loss": 0.04398059472441673, "global_step": 215224, "epoch": 2418, "lr": 1.0253793536108886e-05}
{"train_loss": 0.09574443846940994, "global_step": 215225, "epoch": 2418, "lr": 1.0253441807452646e-05}
{"train_loss": 0.04414737597107887, "global_step": 215226, "epoch": 2418, "lr": 1.0253090084139866e-05}
{"train_loss": 0.05643489211797714, "global_step": 215227, "epoch": 2418, "lr": 1.025273836617055e-05}
{"train_loss": 0.09746462851762772, "global_step": 215228, "epoch": 2418, "lr": 1.0252386653544788e-05}
{"train_loss": 0.036817971616983414, "global_step": 215229, "epoch": 2418, "lr": 1.0252034946262595e-05}
{"train_loss": 0.037747420370578766, "global_step": 215230, "epoch": 2418, "lr": 1.0251683244324045e-05}
{"train_loss": 0.07745779305696487, "global_step": 215231, "epoch": 2418, "lr": 1.025133154772917e-05}
{"train_loss": 0.041430652141571045, "global_step": 215232, "epoch": 2418, "lr": 1.0250979856478005e-05}
{"train_loss": 0.038000017404556274, "global_step": 215233, "epoch": 2418, "lr": 1.0250628170570626e-05}
{"train_loss": 0.03787335380911827, "global_step": 215234, "epoch": 2418, "lr": 1.0250276490007054e-05}
{"train_loss": 0.06257888674736023, "global_step": 215235, "epoch": 2418, "lr": 1.0249924814787359e-05}
{"train_loss": 0.055888671427965164, "global_step": 215236, "epoch": 2418, "lr": 1.0249573144911563e-05}
{"train_loss": 0.08099915832281113, "global_step": 215237, "epoch": 2418, "lr": 1.024922148037974e-05}
{"train_loss": 0.03441436588764191, "global_step": 215238, "epoch": 2418, "lr": 1.0248869821191925e-05}
{"train_loss": 0.0943790003657341, "global_step": 215239, "epoch": 2418, "lr": 1.0248518167348153e-05}
{"train_loss": 0.09261664748191833, "global_step": 215240, "epoch": 2418, "lr": 1.02481665188485e-05}
{"train_loss": 0.045748744159936905, "global_step": 215241, "epoch": 2418, "lr": 1.024781487569298e-05}
{"train_loss": 0.034701231867074966, "global_step": 215242, "epoch": 2418, "lr": 1.0247463237881655e-05}
{"train_loss": 0.06193864345550537, "global_step": 215243, "epoch": 2418, "lr": 1.0247111605414589e-05}
{"train_loss": 0.1374841183423996, "global_step": 215244, "epoch": 2418, "lr": 1.0246759978291803e-05}
{"train_loss": 0.058873359113931656, "global_step": 215245, "epoch": 2418, "lr": 1.0246408356513365e-05}
{"train_loss": 0.05286877974867821, "global_step": 215246, "epoch": 2418, "lr": 1.0246056740079301e-05}
{"train_loss": 0.03801403194665909, "global_step": 215247, "epoch": 2418, "lr": 1.0245705128989685e-05}
{"train_loss": 0.09574980288743973, "global_step": 215248, "epoch": 2418, "lr": 1.0245353523244534e-05}
{"train_loss": 0.10742419958114624, "global_step": 215249, "epoch": 2418, "lr": 1.0245001922843928e-05}
{"train_loss": 0.09189576655626297, "global_step": 215250, "epoch": 2418, "lr": 1.024465032778788e-05}
{"train_loss": 0.0815666913986206, "global_step": 215251, "epoch": 2418, "lr": 1.0244298738076464e-05}
{"train_loss": 0.060125865042209625, "global_step": 215252, "epoch": 2418, "lr": 1.0243947153709709e-05}
{"train_loss": 0.07450681179761887, "global_step": 215253, "epoch": 2418, "lr": 1.0243595574687686e-05}
{"train_loss": 0.049429893493652344, "global_step": 215254, "epoch": 2418, "lr": 1.024324400101041e-05}
{"train_loss": 0.0801682248711586, "global_step": 215255, "epoch": 2418, "lr": 1.0242892432677958e-05}
{"train_loss": 0.09494779258966446, "global_step": 215256, "epoch": 2418, "lr": 1.0242540869690354e-05}
{"train_loss": 0.11121083050966263, "global_step": 215257, "epoch": 2418, "lr": 1.0242189312047668e-05}
{"train_loss": 0.03294662758708, "global_step": 215258, "epoch": 2418, "lr": 1.0241837759749923e-05}
{"train_loss": 0.06627339124679565, "global_step": 215259, "epoch": 2418, "lr": 1.024148621279719e-05}
{"train_loss": 0.025659669190645218, "global_step": 215260, "epoch": 2418, "lr": 1.0241134671189489e-05}
{"train_loss": 0.05462002381682396, "global_step": 215261, "epoch": 2418, "lr": 1.0240783134926884e-05}
{"train_loss": 0.09833382070064545, "global_step": 215262, "epoch": 2418, "lr": 1.0240431604009438e-05}
{"train_loss": 0.06704128533601761, "global_step": 215263, "epoch": 2418, "lr": 1.0240080078437164e-05}
{"train_loss": 0.03207411989569664, "global_step": 215264, "epoch": 2418, "lr": 1.0239728558210143e-05}
{"train_loss": 0.05215480178594589, "global_step": 215265, "epoch": 2418, "lr": 1.0239377043328391e-05}
{"train_loss": 0.05641160160303116, "global_step": 215266, "epoch": 2418, "lr": 1.0239025533791984e-05}
{"train_loss": 0.039383769035339355, "global_step": 215267, "epoch": 2418, "lr": 1.0238674029600942e-05}
{"train_loss": 0.05158579349517822, "global_step": 215268, "epoch": 2418, "lr": 1.023832253075534e-05}
{"train_loss": 0.03538532182574272, "global_step": 215269, "epoch": 2418, "lr": 1.0237971037255196e-05}
{"train_loss": 0.035667963325977325, "global_step": 215270, "epoch": 2418, "lr": 1.0237619549100586e-05}
{"train_loss": 0.04543719068169594, "global_step": 215271, "epoch": 2418, "lr": 1.0237268066291545e-05}
{"train_loss": 0.03224740922451019, "global_step": 215272, "epoch": 2418, "lr": 1.0236916588828105e-05}
{"train_loss": 0.018253251910209656, "global_step": 215273, "epoch": 2418, "lr": 1.023656511671034e-05}
{"train_loss": 0.06205236166715622, "global_step": 215274, "epoch": 2418, "lr": 1.0236213649938265e-05}
{"train_loss": 0.08255259692668915, "global_step": 215275, "epoch": 2418, "lr": 1.0235862188511969e-05}
{"train_loss": 0.06561566889286041, "global_step": 215276, "epoch": 2418, "lr": 1.023551073243147e-05}
{"train_loss": 0.06630399823188782, "global_step": 215277, "epoch": 2418, "lr": 1.023515928169681e-05}
{"train_loss": 0.05153395235538483, "global_step": 215278, "epoch": 2418, "lr": 1.0234807836308053e-05}
{"train_loss": 0.07166419178247452, "global_step": 215279, "epoch": 2418, "lr": 1.0234456396265246e-05}
{"train_loss": 0.08450993150472641, "global_step": 215280, "epoch": 2418, "lr": 1.0234104961568424e-05}
{"train_loss": 0.050262413918972015, "global_step": 215281, "epoch": 2418, "lr": 1.0233753532217655e-05}
{"train_loss": 0.05288821458816528, "global_step": 215282, "epoch": 2418, "lr": 1.0233402108212958e-05}
{"train_loss": 0.07858793437480927, "global_step": 215283, "epoch": 2418, "lr": 1.0233050689554413e-05}
{"train_loss": 0.03189224377274513, "global_step": 215284, "epoch": 2418, "lr": 1.023269927624203e-05}
{"train_loss": 0.0735970214009285, "global_step": 215285, "epoch": 2418, "lr": 1.0232347868275894e-05}
{"train_loss": 0.05147698149085045, "global_step": 215286, "epoch": 2418, "lr": 1.023199646565602e-05}
{"train_loss": 0.04452977702021599, "global_step": 215287, "epoch": 2418, "lr": 1.0231645068382485e-05}
{"train_loss": 0.029544668272137642, "global_step": 215288, "epoch": 2418, "lr": 1.02312936764553e-05}
{"train_loss": 0.09328018873929977, "global_step": 215289, "epoch": 2418, "lr": 1.0230942289874552e-05}
{"train_loss": 0.060827839533599576, "global_step": 215290, "epoch": 2418, "lr": 1.0230590908640253e-05, "val_loss": 8.867704391479492}
{"train_loss": 0.11233194172382355, "global_step": 215291, "epoch": 2419, "lr": 1.0230239532752484e-05}
{"train_loss": 0.05449801683425903, "global_step": 215292, "epoch": 2419, "lr": 1.0229888162211254e-05}
{"train_loss": 0.09469660371541977, "global_step": 215293, "epoch": 2419, "lr": 1.0229536797016647e-05}
{"train_loss": 0.037942372262477875, "global_step": 215294, "epoch": 2419, "lr": 1.0229185437168687e-05}
{"train_loss": 0.07918785512447357, "global_step": 215295, "epoch": 2419, "lr": 1.0228834082667431e-05}
{"train_loss": 0.04792803153395653, "global_step": 215296, "epoch": 2419, "lr": 1.0228482733512917e-05}
{"train_loss": 0.0566849522292614, "global_step": 215297, "epoch": 2419, "lr": 1.0228131389705198e-05}
{"train_loss": 0.07720233500003815, "global_step": 215298, "epoch": 2419, "lr": 1.0227780051244334e-05}
{"train_loss": 0.03522585704922676, "global_step": 215299, "epoch": 2419, "lr": 1.022742871813035e-05}
{"train_loss": 0.03544103354215622, "global_step": 215300, "epoch": 2419, "lr": 1.0227077390363316e-05}
{"train_loss": 0.039124391973018646, "global_step": 215301, "epoch": 2419, "lr": 1.0226726067943248e-05}
{"train_loss": 0.018568648025393486, "global_step": 215302, "epoch": 2419, "lr": 1.0226374750870232e-05}
{"train_loss": 0.02212490700185299, "global_step": 215303, "epoch": 2419, "lr": 1.0226023439144279e-05}
{"train_loss": 0.03349192813038826, "global_step": 215304, "epoch": 2419, "lr": 1.0225672132765463e-05}
{"train_loss": 0.029347678646445274, "global_step": 215305, "epoch": 2419, "lr": 1.022532083173381e-05}
{"train_loss": 0.042945992201566696, "global_step": 215306, "epoch": 2419, "lr": 1.0224969536049395e-05}
{"train_loss": 0.06798011809587479, "global_step": 215307, "epoch": 2419, "lr": 1.0224618245712231e-05}
{"train_loss": 0.06249978393316269, "global_step": 215308, "epoch": 2419, "lr": 1.0224266960722396e-05}
{"train_loss": 0.06965844333171844, "global_step": 215309, "epoch": 2419, "lr": 1.0223915681079921e-05}
{"train_loss": 0.0653422400355339, "global_step": 215310, "epoch": 2419, "lr": 1.0223564406784842e-05}
{"train_loss": 0.04645036906003952, "global_step": 215311, "epoch": 2419, "lr": 1.0223213137837234e-05}
{"train_loss": 0.05002610757946968, "global_step": 215312, "epoch": 2419, "lr": 1.0222861874237121e-05}
{"train_loss": 0.050337277352809906, "global_step": 215313, "epoch": 2419, "lr": 1.022251061598457e-05}
{"train_loss": 0.05906970053911209, "global_step": 215314, "epoch": 2419, "lr": 1.022215936307962e-05}
{"train_loss": 0.031077798455953598, "global_step": 215315, "epoch": 2419, "lr": 1.0221808115522297e-05}
{"train_loss": 0.07280221581459045, "global_step": 215316, "epoch": 2419, "lr": 1.0221456873312673e-05}
{"train_loss": 0.034410662949085236, "global_step": 215317, "epoch": 2419, "lr": 1.02211056364508e-05}
{"train_loss": 0.031069587916135788, "global_step": 215318, "epoch": 2419, "lr": 1.02207544049367e-05}
{"train_loss": 0.05443442612886429, "global_step": 215319, "epoch": 2419, "lr": 1.0220403178770448e-05}
{"train_loss": 0.07495176792144775, "global_step": 215320, "epoch": 2419, "lr": 1.0220051957952064e-05}
{"train_loss": 0.04235895723104477, "global_step": 215321, "epoch": 2419, "lr": 1.0219700742481626e-05}
{"train_loss": 0.011869113892316818, "global_step": 215322, "epoch": 2419, "lr": 1.0219349532359152e-05}
{"train_loss": 0.11659795045852661, "global_step": 215323, "epoch": 2419, "lr": 1.0218998327584711e-05}
{"train_loss": 0.12344352900981903, "global_step": 215324, "epoch": 2419, "lr": 1.0218647128158325e-05}
{"train_loss": 0.05219366028904915, "global_step": 215325, "epoch": 2419, "lr": 1.021829593408008e-05}
{"train_loss": 0.07447753846645355, "global_step": 215326, "epoch": 2419, "lr": 1.0217944745349977e-05}
{"train_loss": 0.054257120937108994, "global_step": 215327, "epoch": 2419, "lr": 1.0217593561968103e-05}
{"train_loss": 0.020868537947535515, "global_step": 215328, "epoch": 2419, "lr": 1.0217242383934478e-05}
{"train_loss": 0.03488299623131752, "global_step": 215329, "epoch": 2419, "lr": 1.0216891211249174e-05}
{"train_loss": 0.06750574707984924, "global_step": 215330, "epoch": 2419, "lr": 1.0216540043912205e-05}
{"train_loss": 0.04544451832771301, "global_step": 215331, "epoch": 2419, "lr": 1.0216188881923655e-05}
{"train_loss": 0.021985139697790146, "global_step": 215332, "epoch": 2419, "lr": 1.021583772528354e-05}
{"train_loss": 0.08314430713653564, "global_step": 215333, "epoch": 2419, "lr": 1.0215486573991916e-05}
{"train_loss": 0.09044679254293442, "global_step": 215334, "epoch": 2419, "lr": 1.0215135428048856e-05}
{"train_loss": 0.05009131133556366, "global_step": 215335, "epoch": 2419, "lr": 1.0214784287454365e-05}
{"train_loss": 0.05659347400069237, "global_step": 215336, "epoch": 2419, "lr": 1.021443315220853e-05}
{"train_loss": 0.03383710980415344, "global_step": 215337, "epoch": 2419, "lr": 1.0214082022311366e-05}
{"train_loss": 0.0425042100250721, "global_step": 215338, "epoch": 2419, "lr": 1.0213730897762947e-05}
{"train_loss": 0.06450746953487396, "global_step": 215339, "epoch": 2419, "lr": 1.0213379778563292e-05}
{"train_loss": 0.042947277426719666, "global_step": 215340, "epoch": 2419, "lr": 1.0213028664712477e-05}
{"train_loss": 0.06713283807039261, "global_step": 215341, "epoch": 2419, "lr": 1.021267755621052e-05}
{"train_loss": 0.06225845590233803, "global_step": 215342, "epoch": 2419, "lr": 1.0212326453057496e-05}
{"train_loss": 0.04381566867232323, "global_step": 215343, "epoch": 2419, "lr": 1.0211975355253434e-05}
{"train_loss": 0.05765202268958092, "global_step": 215344, "epoch": 2419, "lr": 1.0211624262798392e-05}
{"train_loss": 0.03724363073706627, "global_step": 215345, "epoch": 2419, "lr": 1.0211273175692403e-05}
{"train_loss": 0.028048817068338394, "global_step": 215346, "epoch": 2419, "lr": 1.0210922093935538e-05}
{"train_loss": 0.07673952728509903, "global_step": 215347, "epoch": 2419, "lr": 1.0210571017527826e-05}
{"train_loss": 0.02659754827618599, "global_step": 215348, "epoch": 2419, "lr": 1.0210219946469307e-05}
{"train_loss": 0.04419197142124176, "global_step": 215349, "epoch": 2419, "lr": 1.0209868880760049e-05}
{"train_loss": 0.013914297334849834, "global_step": 215350, "epoch": 2419, "lr": 1.0209517820400077e-05}
{"train_loss": 0.050957582890987396, "global_step": 215351, "epoch": 2419, "lr": 1.0209166765389467e-05}
{"train_loss": 0.03380254656076431, "global_step": 215352, "epoch": 2419, "lr": 1.0208815715728231e-05}
{"train_loss": 0.1008601039648056, "global_step": 215353, "epoch": 2419, "lr": 1.0208464671416451e-05}
{"train_loss": 0.08282537758350372, "global_step": 215354, "epoch": 2419, "lr": 1.0208113632454147e-05}
{"train_loss": 0.06998942047357559, "global_step": 215355, "epoch": 2419, "lr": 1.0207762598841386e-05}
{"train_loss": 0.08258921653032303, "global_step": 215356, "epoch": 2419, "lr": 1.0207411570578197e-05}
{"train_loss": 0.053027089685201645, "global_step": 215357, "epoch": 2419, "lr": 1.020706054766465e-05}
{"train_loss": 0.09521893411874771, "global_step": 215358, "epoch": 2419, "lr": 1.020670953010076e-05}
{"train_loss": 0.037901319563388824, "global_step": 215359, "epoch": 2419, "lr": 1.0206358517886611e-05}
{"train_loss": 0.08177662640810013, "global_step": 215360, "epoch": 2419, "lr": 1.020600751102222e-05}
{"train_loss": 0.06349147856235504, "global_step": 215361, "epoch": 2419, "lr": 1.0205656509507656e-05}
{"train_loss": 0.03867518529295921, "global_step": 215362, "epoch": 2419, "lr": 1.0205305513342945e-05}
{"train_loss": 0.05277078226208687, "global_step": 215363, "epoch": 2419, "lr": 1.0204954522528154e-05}
{"train_loss": 0.08081207424402237, "global_step": 215364, "epoch": 2419, "lr": 1.0204603537063312e-05}
{"train_loss": 0.07817995548248291, "global_step": 215365, "epoch": 2419, "lr": 1.020425255694849e-05}
{"train_loss": 0.08270999044179916, "global_step": 215366, "epoch": 2419, "lr": 1.0203901582183712e-05}
{"train_loss": 0.05270811542868614, "global_step": 215367, "epoch": 2419, "lr": 1.0203550612769042e-05}
{"train_loss": 0.04504650458693504, "global_step": 215368, "epoch": 2419, "lr": 1.0203199648704509e-05}
{"train_loss": 0.03596290200948715, "global_step": 215369, "epoch": 2419, "lr": 1.0202848689990169e-05}
{"train_loss": 0.08159878104925156, "global_step": 215370, "epoch": 2419, "lr": 1.0202497736626081e-05}
{"train_loss": 0.031173136085271835, "global_step": 215371, "epoch": 2419, "lr": 1.0202146788612276e-05}
{"train_loss": 0.08194634318351746, "global_step": 215372, "epoch": 2419, "lr": 1.0201795845948814e-05}
{"train_loss": 0.03755225986242294, "global_step": 215373, "epoch": 2419, "lr": 1.0201444908635726e-05}
{"train_loss": 0.052490636706352234, "global_step": 215374, "epoch": 2419, "lr": 1.0201093976673082e-05}
{"train_loss": 0.03960871696472168, "global_step": 215375, "epoch": 2419, "lr": 1.0200743050060902e-05}
{"train_loss": 0.06316869705915451, "global_step": 215376, "epoch": 2419, "lr": 1.0200392128799257e-05}
{"train_loss": 0.0662367045879364, "global_step": 215377, "epoch": 2419, "lr": 1.0200041212888178e-05}
{"train_loss": 0.06037085875868797, "global_step": 215378, "epoch": 2419, "lr": 1.0199690302327725e-05}
{"train_loss": 0.05537326231054711, "global_step": 215379, "epoch": 2419, "lr": 1.019933939711793e-05, "val_loss": 8.932095527648926}
{"train_loss": 0.08467727899551392, "global_step": 215380, "epoch": 2420, "lr": 1.019898849725886e-05}
{"train_loss": 0.15357190370559692, "global_step": 215381, "epoch": 2420, "lr": 1.0198637602750538e-05}
{"train_loss": 0.08045738935470581, "global_step": 215382, "epoch": 2420, "lr": 1.0198286713593036e-05}
{"train_loss": 0.04469548538327217, "global_step": 215383, "epoch": 2420, "lr": 1.0197935829786376e-05}
{"train_loss": 0.04704854637384415, "global_step": 215384, "epoch": 2420, "lr": 1.019758495133063e-05}
{"train_loss": 0.07881399244070053, "global_step": 215385, "epoch": 2420, "lr": 1.0197234078225826e-05}
{"train_loss": 0.06400511413812637, "global_step": 215386, "epoch": 2420, "lr": 1.019688321047203e-05}
{"train_loss": 0.06163142994046211, "global_step": 215387, "epoch": 2420, "lr": 1.0196532348069276e-05}
{"train_loss": 0.06579168885946274, "global_step": 215388, "epoch": 2420, "lr": 1.0196181491017599e-05}
{"train_loss": 0.03552259877324104, "global_step": 215389, "epoch": 2420, "lr": 1.0195830639317077e-05}
{"train_loss": 0.042985618114471436, "global_step": 215390, "epoch": 2420, "lr": 1.0195479792967722e-05}
{"train_loss": 0.045253023505210876, "global_step": 215391, "epoch": 2420, "lr": 1.0195128951969618e-05}
{"train_loss": 0.054824598133563995, "global_step": 215392, "epoch": 2420, "lr": 1.0194778116322778e-05}
{"train_loss": 0.02099839225411415, "global_step": 215393, "epoch": 2420, "lr": 1.019442728602728e-05}
{"train_loss": 0.08226040750741959, "global_step": 215394, "epoch": 2420, "lr": 1.0194076461083141e-05}
{"train_loss": 0.04060041904449463, "global_step": 215395, "epoch": 2420, "lr": 1.0193725641490437e-05}
{"train_loss": 0.013918992131948471, "global_step": 215396, "epoch": 2420, "lr": 1.0193374827249185e-05}
{"train_loss": 0.04551035538315773, "global_step": 215397, "epoch": 2420, "lr": 1.0193024018359466e-05}
{"train_loss": 0.039754558354616165, "global_step": 215398, "epoch": 2420, "lr": 1.0192673214821297e-05}
{"train_loss": 0.03910142183303833, "global_step": 215399, "epoch": 2420, "lr": 1.0192322416634747e-05}
{"train_loss": 0.06607397645711899, "global_step": 215400, "epoch": 2420, "lr": 1.0191971623799846e-05}
{"train_loss": 0.031228970736265182, "global_step": 215401, "epoch": 2420, "lr": 1.0191620836316661e-05}
{"train_loss": 0.07912489026784897, "global_step": 215402, "epoch": 2420, "lr": 1.0191270054185214e-05}
{"train_loss": 0.06030812859535217, "global_step": 215403, "epoch": 2420, "lr": 1.0190919277405575e-05}
{"train_loss": 0.04129909723997116, "global_step": 215404, "epoch": 2420, "lr": 1.0190568505977772e-05}
{"train_loss": 0.06883799284696579, "global_step": 215405, "epoch": 2420, "lr": 1.0190217739901864e-05}
{"train_loss": 0.04109600931406021, "global_step": 215406, "epoch": 2420, "lr": 1.0189866979177903e-05}
{"train_loss": 0.10188791900873184, "global_step": 215407, "epoch": 2420, "lr": 1.0189516223805924e-05}
{"train_loss": 0.042594458907842636, "global_step": 215408, "epoch": 2420, "lr": 1.0189165473785983e-05}
{"train_loss": 0.057152796536684036, "global_step": 215409, "epoch": 2420, "lr": 1.0188814729118117e-05}
{"train_loss": 0.04842448979616165, "global_step": 215410, "epoch": 2420, "lr": 1.0188463989802393e-05}
{"train_loss": 0.10594138503074646, "global_step": 215411, "epoch": 2420, "lr": 1.0188113255838826e-05}
{"train_loss": 0.06480876356363297, "global_step": 215412, "epoch": 2420, "lr": 1.0187762527227501e-05}
{"train_loss": 0.06741846352815628, "global_step": 215413, "epoch": 2420, "lr": 1.0187411803968434e-05}
{"train_loss": 0.07652762532234192, "global_step": 215414, "epoch": 2420, "lr": 1.018706108606169e-05}
{"train_loss": 0.03468696027994156, "global_step": 215415, "epoch": 2420, "lr": 1.0186710373507302e-05}
{"train_loss": 0.019218629226088524, "global_step": 215416, "epoch": 2420, "lr": 1.018635966630534e-05}
{"train_loss": 0.07571426033973694, "global_step": 215417, "epoch": 2420, "lr": 1.0186008964455818e-05}
{"train_loss": 0.06351537257432938, "global_step": 215418, "epoch": 2420, "lr": 1.018565826795882e-05}
{"train_loss": 0.049690332263708115, "global_step": 215419, "epoch": 2420, "lr": 1.0185307576814363e-05}
{"train_loss": 0.032869450747966766, "global_step": 215420, "epoch": 2420, "lr": 1.0184956891022512e-05}
{"train_loss": 0.047772228717803955, "global_step": 215421, "epoch": 2420, "lr": 1.0184606210583303e-05}
{"train_loss": 0.07209900766611099, "global_step": 215422, "epoch": 2420, "lr": 1.0184255535496794e-05}
{"train_loss": 0.036730434745550156, "global_step": 215423, "epoch": 2420, "lr": 1.0183904865763023e-05}
{"train_loss": 0.05962304398417473, "global_step": 215424, "epoch": 2420, "lr": 1.0183554201382045e-05}
{"train_loss": 0.02707444317638874, "global_step": 215425, "epoch": 2420, "lr": 1.0183203542353908e-05}
{"train_loss": 0.0794002041220665, "global_step": 215426, "epoch": 2420, "lr": 1.018285288867864e-05}
{"train_loss": 0.0791141539812088, "global_step": 215427, "epoch": 2420, "lr": 1.0182502240356312e-05}
{"train_loss": 0.06579340249300003, "global_step": 215428, "epoch": 2420, "lr": 1.0182151597386952e-05}
{"train_loss": 0.09625931084156036, "global_step": 215429, "epoch": 2420, "lr": 1.0181800959770627e-05}
{"train_loss": 0.05546991154551506, "global_step": 215430, "epoch": 2420, "lr": 1.0181450327507363e-05}
{"train_loss": 0.06981901079416275, "global_step": 215431, "epoch": 2420, "lr": 1.018109970059723e-05}
{"train_loss": 0.046805888414382935, "global_step": 215432, "epoch": 2420, "lr": 1.0180749079040247e-05}
{"train_loss": 0.06731847673654556, "global_step": 215433, "epoch": 2420, "lr": 1.0180398462836493e-05}
{"train_loss": 0.05297211557626724, "global_step": 215434, "epoch": 2420, "lr": 1.0180047851985985e-05}
{"train_loss": 0.038764551281929016, "global_step": 215435, "epoch": 2420, "lr": 1.01796972464888e-05}
{"train_loss": 0.04643350839614868, "global_step": 215436, "epoch": 2420, "lr": 1.017934664634495e-05}
{"train_loss": 0.06524693965911865, "global_step": 215437, "epoch": 2420, "lr": 1.0178996051554517e-05}
{"train_loss": 0.06855452060699463, "global_step": 215438, "epoch": 2420, "lr": 1.0178645462117526e-05}
{"train_loss": 0.05160906910896301, "global_step": 215439, "epoch": 2420, "lr": 1.0178294878034039e-05}
{"train_loss": 0.08320735394954681, "global_step": 215440, "epoch": 2420, "lr": 1.0177944299304083e-05}
{"train_loss": 0.054478906095027924, "global_step": 215441, "epoch": 2420, "lr": 1.0177593725927725e-05}
{"train_loss": 0.06589531153440475, "global_step": 215442, "epoch": 2420, "lr": 1.0177243157904992e-05}
{"train_loss": 0.030680833384394646, "global_step": 215443, "epoch": 2420, "lr": 1.0176892595235948e-05}
{"train_loss": 0.04226032271981239, "global_step": 215444, "epoch": 2420, "lr": 1.0176542037920645e-05}
{"train_loss": 0.08262911438941956, "global_step": 215445, "epoch": 2420, "lr": 1.0176191485959114e-05}
{"train_loss": 0.014443447813391685, "global_step": 215446, "epoch": 2420, "lr": 1.0175840939351417e-05}
{"train_loss": 0.048153992742300034, "global_step": 215447, "epoch": 2420, "lr": 1.0175490398097576e-05}
{"train_loss": 0.07362574338912964, "global_step": 215448, "epoch": 2420, "lr": 1.017513986219767e-05}
{"train_loss": 0.0642593652009964, "global_step": 215449, "epoch": 2420, "lr": 1.0174789331651718e-05}
{"train_loss": 0.06355834752321243, "global_step": 215450, "epoch": 2420, "lr": 1.0174438806459796e-05}
{"train_loss": 0.06666333228349686, "global_step": 215451, "epoch": 2420, "lr": 1.0174088286621924e-05}
{"train_loss": 0.034154798835515976, "global_step": 215452, "epoch": 2420, "lr": 1.017373777213817e-05}
{"train_loss": 0.04035123810172081, "global_step": 215453, "epoch": 2420, "lr": 1.0173387263008561e-05}
{"train_loss": 0.03861299902200699, "global_step": 215454, "epoch": 2420, "lr": 1.0173036759233162e-05}
{"train_loss": 0.05579805001616478, "global_step": 215455, "epoch": 2420, "lr": 1.0172686260812004e-05}
{"train_loss": 0.03741211071610451, "global_step": 215456, "epoch": 2420, "lr": 1.0172335767745156e-05}
{"train_loss": 0.04654563218355179, "global_step": 215457, "epoch": 2420, "lr": 1.0171985280032637e-05}
{"train_loss": 0.045408256351947784, "global_step": 215458, "epoch": 2420, "lr": 1.0171634797674524e-05}
{"train_loss": 0.06153818964958191, "global_step": 215459, "epoch": 2420, "lr": 1.0171284320670838e-05}
{"train_loss": 0.09578856825828552, "global_step": 215460, "epoch": 2420, "lr": 1.0170933849021647e-05}
{"train_loss": 0.02142329514026642, "global_step": 215461, "epoch": 2420, "lr": 1.0170583382726973e-05}
{"train_loss": 0.0727429986000061, "global_step": 215462, "epoch": 2420, "lr": 1.0170232921786899e-05}
{"train_loss": 0.04590795561671257, "global_step": 215463, "epoch": 2420, "lr": 1.0169882466201447e-05}
{"train_loss": 0.053835008293390274, "global_step": 215464, "epoch": 2420, "lr": 1.0169532015970656e-05}
{"train_loss": 0.05600416660308838, "global_step": 215465, "epoch": 2420, "lr": 1.01691815710946e-05}
{"train_loss": 0.03590617701411247, "global_step": 215466, "epoch": 2420, "lr": 1.0168831131573297e-05}
{"train_loss": 0.07925277203321457, "global_step": 215467, "epoch": 2420, "lr": 1.0168480697406823e-05}
{"train_loss": 0.05761896216132668, "global_step": 215468, "epoch": 2420, "lr": 1.01681302685952e-05, "val_loss": 8.820808410644531, "train_action_mse_error": 15.313497543334961}
{"train_loss": 0.09630923718214035, "global_step": 215469, "epoch": 2421, "lr": 1.0167779845138497e-05}
{"train_loss": 0.022504407912492752, "global_step": 215470, "epoch": 2421, "lr": 1.016742942703674e-05}
{"train_loss": 0.058053068816661835, "global_step": 215471, "epoch": 2421, "lr": 1.0167079014289999e-05}
{"train_loss": 0.02768518216907978, "global_step": 215472, "epoch": 2421, "lr": 1.0166728606898296e-05}
{"train_loss": 0.07091320306062698, "global_step": 215473, "epoch": 2421, "lr": 1.0166378204861704e-05}
{"train_loss": 0.06761705130338669, "global_step": 215474, "epoch": 2421, "lr": 1.0166027808180245e-05}
{"train_loss": 0.07276391237974167, "global_step": 215475, "epoch": 2421, "lr": 1.0165677416853986e-05}
{"train_loss": 0.01717296987771988, "global_step": 215476, "epoch": 2421, "lr": 1.016532703088296e-05}
{"train_loss": 0.061479996889829636, "global_step": 215477, "epoch": 2421, "lr": 1.0164976650267227e-05}
{"train_loss": 0.03816831111907959, "global_step": 215478, "epoch": 2421, "lr": 1.0164626275006816e-05}
{"train_loss": 0.11203090846538544, "global_step": 215479, "epoch": 2421, "lr": 1.016427590510179e-05}
{"train_loss": 0.04225962981581688, "global_step": 215480, "epoch": 2421, "lr": 1.01639255405522e-05}
{"train_loss": 0.03327028080821037, "global_step": 215481, "epoch": 2421, "lr": 1.016357518135807e-05}
{"train_loss": 0.08444985747337341, "global_step": 215482, "epoch": 2421, "lr": 1.016322482751948e-05}
{"train_loss": 0.04118015617132187, "global_step": 215483, "epoch": 2421, "lr": 1.0162874479036444e-05}
{"train_loss": 0.041992444545030594, "global_step": 215484, "epoch": 2421, "lr": 1.0162524135909035e-05}
{"train_loss": 0.035292770713567734, "global_step": 215485, "epoch": 2421, "lr": 1.0162173798137281e-05}
{"train_loss": 0.07792840152978897, "global_step": 215486, "epoch": 2421, "lr": 1.0161823465721243e-05}
{"train_loss": 0.027759820222854614, "global_step": 215487, "epoch": 2421, "lr": 1.0161473138660954e-05}
{"train_loss": 0.030367078259587288, "global_step": 215488, "epoch": 2421, "lr": 1.0161122816956481e-05}
{"train_loss": 0.04313152655959129, "global_step": 215489, "epoch": 2421, "lr": 1.0160772500607845e-05}
{"train_loss": 0.06071218103170395, "global_step": 215490, "epoch": 2421, "lr": 1.0160422189615126e-05}
{"train_loss": 0.03205243498086929, "global_step": 215491, "epoch": 2421, "lr": 1.0160071883978334e-05}
{"train_loss": 0.1082356795668602, "global_step": 215492, "epoch": 2421, "lr": 1.015972158369755e-05}
{"train_loss": 0.03690720349550247, "global_step": 215493, "epoch": 2421, "lr": 1.0159371288772795e-05}
{"train_loss": 0.028790870681405067, "global_step": 215494, "epoch": 2421, "lr": 1.015902099920414e-05}
{"train_loss": 0.03847845643758774, "global_step": 215495, "epoch": 2421, "lr": 1.0158670714991603e-05}
{"train_loss": 0.06934770941734314, "global_step": 215496, "epoch": 2421, "lr": 1.0158320436135244e-05}
{"train_loss": 0.08924617618322372, "global_step": 215497, "epoch": 2421, "lr": 1.0157970162635133e-05}
{"train_loss": 0.06475332379341125, "global_step": 215498, "epoch": 2421, "lr": 1.0157619894491293e-05}
{"train_loss": 0.08703812956809998, "global_step": 215499, "epoch": 2421, "lr": 1.0157269631703764e-05}
{"train_loss": 0.0699675902724266, "global_step": 215500, "epoch": 2421, "lr": 1.0156919374272617e-05}
{"train_loss": 0.051283471286296844, "global_step": 215501, "epoch": 2421, "lr": 1.0156569122197873e-05}
{"train_loss": 0.05996031314134598, "global_step": 215502, "epoch": 2421, "lr": 1.0156218875479606e-05}
{"train_loss": 0.06546405702829361, "global_step": 215503, "epoch": 2421, "lr": 1.015586863411785e-05}
{"train_loss": 0.035731248557567596, "global_step": 215504, "epoch": 2421, "lr": 1.015551839811264e-05}
{"train_loss": 0.023382531479001045, "global_step": 215505, "epoch": 2421, "lr": 1.0155168167464046e-05}
{"train_loss": 0.03219878301024437, "global_step": 215506, "epoch": 2421, "lr": 1.015481794217209e-05}
{"train_loss": 0.0510534904897213, "global_step": 215507, "epoch": 2421, "lr": 1.015446772223685e-05}
{"train_loss": 0.04731694236397743, "global_step": 215508, "epoch": 2421, "lr": 1.015411750765834e-05}
{"train_loss": 0.03801623359322548, "global_step": 215509, "epoch": 2421, "lr": 1.015376729843664e-05}
{"train_loss": 0.07490614056587219, "global_step": 215510, "epoch": 2421, "lr": 1.0153417094571765e-05}
{"train_loss": 0.03594100475311279, "global_step": 215511, "epoch": 2421, "lr": 1.015306689606379e-05}
{"train_loss": 0.024822071194648743, "global_step": 215512, "epoch": 2421, "lr": 1.0152716702912735e-05}
{"train_loss": 0.07631615549325943, "global_step": 215513, "epoch": 2421, "lr": 1.0152366515118677e-05}
{"train_loss": 0.035459570586681366, "global_step": 215514, "epoch": 2421, "lr": 1.0152016332681635e-05}
{"train_loss": 0.17012259364128113, "global_step": 215515, "epoch": 2421, "lr": 1.0151666155601663e-05}
{"train_loss": 0.0669603943824768, "global_step": 215516, "epoch": 2421, "lr": 1.0151315983878835e-05}
{"train_loss": 0.10670880228281021, "global_step": 215517, "epoch": 2421, "lr": 1.0150965817513159e-05}
{"train_loss": 0.032909929752349854, "global_step": 215518, "epoch": 2421, "lr": 1.0150615656504714e-05}
{"train_loss": 0.07820736616849899, "global_step": 215519, "epoch": 2421, "lr": 1.0150265500853524e-05}
{"train_loss": 0.03776552528142929, "global_step": 215520, "epoch": 2421, "lr": 1.0149915350559652e-05}
{"train_loss": 0.034483376890420914, "global_step": 215521, "epoch": 2421, "lr": 1.014956520562313e-05}
{"train_loss": 0.021017463877797127, "global_step": 215522, "epoch": 2421, "lr": 1.0149215066044022e-05}
{"train_loss": 0.058840833604335785, "global_step": 215523, "epoch": 2421, "lr": 1.0148864931822355e-05}
{"train_loss": 0.12434963136911392, "global_step": 215524, "epoch": 2421, "lr": 1.0148514802958197e-05}
{"train_loss": 0.046117160469293594, "global_step": 215525, "epoch": 2421, "lr": 1.0148164679451577e-05}
{"train_loss": 0.08299248665571213, "global_step": 215526, "epoch": 2421, "lr": 1.0147814561302566e-05}
{"train_loss": 0.05735879763960838, "global_step": 215527, "epoch": 2421, "lr": 1.014746444851118e-05}
{"train_loss": 0.07629809528589249, "global_step": 215528, "epoch": 2421, "lr": 1.0147114341077491e-05}
{"train_loss": 0.09041000157594681, "global_step": 215529, "epoch": 2421, "lr": 1.0146764239001528e-05}
{"train_loss": 0.057683300226926804, "global_step": 215530, "epoch": 2421, "lr": 1.0146414142283356e-05}
{"train_loss": 0.0847582072019577, "global_step": 215531, "epoch": 2421, "lr": 1.0146064050923005e-05}
{"train_loss": 0.03752823919057846, "global_step": 215532, "epoch": 2421, "lr": 1.0145713964920522e-05}
{"train_loss": 0.07514400035142899, "global_step": 215533, "epoch": 2421, "lr": 1.0145363884275983e-05}
{"train_loss": 0.0910075381398201, "global_step": 215534, "epoch": 2421, "lr": 1.0145013808989395e-05}
{"train_loss": 0.06039274111390114, "global_step": 215535, "epoch": 2421, "lr": 1.0144663739060844e-05}
{"train_loss": 0.0433693453669548, "global_step": 215536, "epoch": 2421, "lr": 1.0144313674490351e-05}
{"train_loss": 0.053117502480745316, "global_step": 215537, "epoch": 2421, "lr": 1.0143963615277962e-05}
{"train_loss": 0.06003299355506897, "global_step": 215538, "epoch": 2421, "lr": 1.0143613561423737e-05}
{"train_loss": 0.04854532703757286, "global_step": 215539, "epoch": 2421, "lr": 1.0143263512927708e-05}
{"train_loss": 0.03472523391246796, "global_step": 215540, "epoch": 2421, "lr": 1.014291346978995e-05}
{"train_loss": 0.0630037859082222, "global_step": 215541, "epoch": 2421, "lr": 1.0142563432010482e-05}
{"train_loss": 0.04432598501443863, "global_step": 215542, "epoch": 2421, "lr": 1.0142213399589356e-05}
{"train_loss": 0.01937863416969776, "global_step": 215543, "epoch": 2421, "lr": 1.0141863372526627e-05}
{"train_loss": 0.05545860528945923, "global_step": 215544, "epoch": 2421, "lr": 1.0141513350822334e-05}
{"train_loss": 0.08273304998874664, "global_step": 215545, "epoch": 2421, "lr": 1.0141163334476539e-05}
{"train_loss": 0.030460044741630554, "global_step": 215546, "epoch": 2421, "lr": 1.0140813323489267e-05}
{"train_loss": 0.05691462382674217, "global_step": 215547, "epoch": 2421, "lr": 1.0140463317860588e-05}
{"train_loss": 0.0731363520026207, "global_step": 215548, "epoch": 2421, "lr": 1.0140113317590526e-05}
{"train_loss": 0.08245573192834854, "global_step": 215549, "epoch": 2421, "lr": 1.0139763322679152e-05}
{"train_loss": 0.02661861851811409, "global_step": 215550, "epoch": 2421, "lr": 1.0139413333126485e-05}
{"train_loss": 0.03710979223251343, "global_step": 215551, "epoch": 2421, "lr": 1.0139063348932593e-05}
{"train_loss": 0.03361279144883156, "global_step": 215552, "epoch": 2421, "lr": 1.013871337009753e-05}
{"train_loss": 0.03430934622883797, "global_step": 215553, "epoch": 2421, "lr": 1.013836339662132e-05}
{"train_loss": 0.028624413534998894, "global_step": 215554, "epoch": 2421, "lr": 1.0138013428504028e-05}
{"train_loss": 0.04830988869071007, "global_step": 215555, "epoch": 2421, "lr": 1.0137663465745689e-05}
{"train_loss": 0.037363581359386444, "global_step": 215556, "epoch": 2421, "lr": 1.0137313508346363e-05}
{"train_loss": 0.05604787453423055, "global_step": 215557, "epoch": 2421, "lr": 1.0136963556306078e-05, "val_loss": 8.732821464538574}
{"train_loss": 0.05310796946287155, "global_step": 215558, "epoch": 2422, "lr": 1.0136613609624907e-05}
{"train_loss": 0.02599511854350567, "global_step": 215559, "epoch": 2422, "lr": 1.0136263668302865e-05}
{"train_loss": 0.03947373107075691, "global_step": 215560, "epoch": 2422, "lr": 1.013591373234003e-05}
{"train_loss": 0.060907769948244095, "global_step": 215561, "epoch": 2422, "lr": 1.0135563801736426e-05}
{"train_loss": 0.03050347790122032, "global_step": 215562, "epoch": 2422, "lr": 1.0135213876492123e-05}
{"train_loss": 0.08712690323591232, "global_step": 215563, "epoch": 2422, "lr": 1.0134863956607137e-05}
{"train_loss": 0.06509218364953995, "global_step": 215564, "epoch": 2422, "lr": 1.013451404208155e-05}
{"train_loss": 0.0333966463804245, "global_step": 215565, "epoch": 2422, "lr": 1.0134164132915375e-05}
{"train_loss": 0.036029376089572906, "global_step": 215566, "epoch": 2422, "lr": 1.013381422910869e-05}
{"train_loss": 0.037904348224401474, "global_step": 215567, "epoch": 2422, "lr": 1.0133464330661519e-05}
{"train_loss": 0.020678414031863213, "global_step": 215568, "epoch": 2422, "lr": 1.0133114437573927e-05}
{"train_loss": 0.04536331072449684, "global_step": 215569, "epoch": 2422, "lr": 1.0132764549845936e-05}
{"train_loss": 0.05711205676198006, "global_step": 215570, "epoch": 2422, "lr": 1.0132414667477614e-05}
{"train_loss": 0.046728577464818954, "global_step": 215571, "epoch": 2422, "lr": 1.0132064790469014e-05}
{"train_loss": 0.04397435113787651, "global_step": 215572, "epoch": 2422, "lr": 1.0131714918820157e-05}
{"train_loss": 0.10764248669147491, "global_step": 215573, "epoch": 2422, "lr": 1.0131365052531122e-05}
{"train_loss": 0.08808163553476334, "global_step": 215574, "epoch": 2422, "lr": 1.0131015191601939e-05}
{"train_loss": 0.058047227561473846, "global_step": 215575, "epoch": 2422, "lr": 1.0130665336032641e-05}
{"train_loss": 0.07198335230350494, "global_step": 215576, "epoch": 2422, "lr": 1.01303154858233e-05}
{"train_loss": 0.06421651691198349, "global_step": 215577, "epoch": 2422, "lr": 1.0129965640973937e-05}
{"train_loss": 0.026767484843730927, "global_step": 215578, "epoch": 2422, "lr": 1.0129615801484633e-05}
{"train_loss": 0.049890950322151184, "global_step": 215579, "epoch": 2422, "lr": 1.01292659673554e-05}
{"train_loss": 0.03625248372554779, "global_step": 215580, "epoch": 2422, "lr": 1.0128916138586314e-05}
{"train_loss": 0.07300567626953125, "global_step": 215581, "epoch": 2422, "lr": 1.0128566315177413e-05}
{"train_loss": 0.05347739905118942, "global_step": 215582, "epoch": 2422, "lr": 1.0128216497128723e-05}
{"train_loss": 0.08014088869094849, "global_step": 215583, "epoch": 2422, "lr": 1.0127866684440323e-05}
{"train_loss": 0.047419410198926926, "global_step": 215584, "epoch": 2422, "lr": 1.012751687711223e-05}
{"train_loss": 0.03612293675541878, "global_step": 215585, "epoch": 2422, "lr": 1.0127167075144523e-05}
{"train_loss": 0.04071418568491936, "global_step": 215586, "epoch": 2422, "lr": 1.0126817278537221e-05}
{"train_loss": 0.05141488090157509, "global_step": 215587, "epoch": 2422, "lr": 1.0126467487290375e-05}
{"train_loss": 0.05693348869681358, "global_step": 215588, "epoch": 2422, "lr": 1.012611770140406e-05}
{"train_loss": 0.06696299463510513, "global_step": 215589, "epoch": 2422, "lr": 1.0125767920878288e-05}
{"train_loss": 0.07443901151418686, "global_step": 215590, "epoch": 2422, "lr": 1.0125418145713129e-05}
{"train_loss": 0.08010154217481613, "global_step": 215591, "epoch": 2422, "lr": 1.0125068375908614e-05}
{"train_loss": 0.06728707253932953, "global_step": 215592, "epoch": 2422, "lr": 1.0124718611464806e-05}
{"train_loss": 0.06349273771047592, "global_step": 215593, "epoch": 2422, "lr": 1.0124368852381733e-05}
{"train_loss": 0.08033496886491776, "global_step": 215594, "epoch": 2422, "lr": 1.0124019098659466e-05}
{"train_loss": 0.05188918486237526, "global_step": 215595, "epoch": 2422, "lr": 1.012366935029802e-05}
{"train_loss": 0.03500908985733986, "global_step": 215596, "epoch": 2422, "lr": 1.0123319607297482e-05}
{"train_loss": 0.09362328052520752, "global_step": 215597, "epoch": 2422, "lr": 1.0122969869657861e-05}
{"train_loss": 0.09442935883998871, "global_step": 215598, "epoch": 2422, "lr": 1.0122620137379235e-05}
{"train_loss": 0.023838231340050697, "global_step": 215599, "epoch": 2422, "lr": 1.012227041046162e-05}
{"train_loss": 0.07786177098751068, "global_step": 215600, "epoch": 2422, "lr": 1.0121920688905096e-05}
{"train_loss": 0.029080362990498543, "global_step": 215601, "epoch": 2422, "lr": 1.0121570972709681e-05}
{"train_loss": 0.04891102761030197, "global_step": 215602, "epoch": 2422, "lr": 1.0121221261875446e-05}
{"train_loss": 0.04771018773317337, "global_step": 215603, "epoch": 2422, "lr": 1.0120871556402411e-05}
{"train_loss": 0.03507538512349129, "global_step": 215604, "epoch": 2422, "lr": 1.012052185629066e-05}
{"train_loss": 0.039224475622177124, "global_step": 215605, "epoch": 2422, "lr": 1.0120172161540198e-05}
{"train_loss": 0.024163048714399338, "global_step": 215606, "epoch": 2422, "lr": 1.0119822472151097e-05}
{"train_loss": 0.0868188738822937, "global_step": 215607, "epoch": 2422, "lr": 1.0119472788123407e-05}
{"train_loss": 0.03064872696995735, "global_step": 215608, "epoch": 2422, "lr": 1.0119123109457162e-05}
{"train_loss": 0.03808728978037834, "global_step": 215609, "epoch": 2422, "lr": 1.0118773436152423e-05}
{"train_loss": 0.04847371205687523, "global_step": 215610, "epoch": 2422, "lr": 1.0118423768209217e-05}
{"train_loss": 0.07203736901283264, "global_step": 215611, "epoch": 2422, "lr": 1.0118074105627617e-05}
{"train_loss": 0.05084769055247307, "global_step": 215612, "epoch": 2422, "lr": 1.0117724448407645e-05}
{"train_loss": 0.061076413840055466, "global_step": 215613, "epoch": 2422, "lr": 1.0117374796549366e-05}
{"train_loss": 0.025020191445946693, "global_step": 215614, "epoch": 2422, "lr": 1.0117025150052828e-05}
{"train_loss": 0.03769763186573982, "global_step": 215615, "epoch": 2422, "lr": 1.011667550891805e-05}
{"train_loss": 0.04270806536078453, "global_step": 215616, "epoch": 2422, "lr": 1.0116325873145116e-05}
{"train_loss": 0.07714899629354477, "global_step": 215617, "epoch": 2422, "lr": 1.0115976242734037e-05}
{"train_loss": 0.033957719802856445, "global_step": 215618, "epoch": 2422, "lr": 1.0115626617684899e-05}
{"train_loss": 0.026239914819598198, "global_step": 215619, "epoch": 2422, "lr": 1.0115276997997724e-05}
{"train_loss": 0.065274178981781, "global_step": 215620, "epoch": 2422, "lr": 1.0114927383672551e-05}
{"train_loss": 0.09997154772281647, "global_step": 215621, "epoch": 2422, "lr": 1.0114577774709454e-05}
{"train_loss": 0.03400837630033493, "global_step": 215622, "epoch": 2422, "lr": 1.0114228171108459e-05}
{"train_loss": 0.07772661745548248, "global_step": 215623, "epoch": 2422, "lr": 1.0113878572869612e-05}
{"train_loss": 0.03618386387825012, "global_step": 215624, "epoch": 2422, "lr": 1.0113528979992982e-05}
{"train_loss": 0.013328013010323048, "global_step": 215625, "epoch": 2422, "lr": 1.0113179392478589e-05}
{"train_loss": 0.0587962381541729, "global_step": 215626, "epoch": 2422, "lr": 1.0112829810326507e-05}
{"train_loss": 0.03980274125933647, "global_step": 215627, "epoch": 2422, "lr": 1.0112480233536759e-05}
{"train_loss": 0.07853188365697861, "global_step": 215628, "epoch": 2422, "lr": 1.0112130662109408e-05}
{"train_loss": 0.0351603738963604, "global_step": 215629, "epoch": 2422, "lr": 1.0111781096044487e-05}
{"train_loss": 0.02225082740187645, "global_step": 215630, "epoch": 2422, "lr": 1.0111431535342064e-05}
{"train_loss": 0.06343182176351547, "global_step": 215631, "epoch": 2422, "lr": 1.0111081980002156e-05}
{"train_loss": 0.05388113856315613, "global_step": 215632, "epoch": 2422, "lr": 1.0110732430024845e-05}
{"train_loss": 0.0938740074634552, "global_step": 215633, "epoch": 2422, "lr": 1.0110382885410141e-05}
{"train_loss": 0.033000100404024124, "global_step": 215634, "epoch": 2422, "lr": 1.0110033346158127e-05}
{"train_loss": 0.10044826567173004, "global_step": 215635, "epoch": 2422, "lr": 1.0109683812268817e-05}
{"train_loss": 0.037060320377349854, "global_step": 215636, "epoch": 2422, "lr": 1.010933428374229e-05}
{"train_loss": 0.07836820185184479, "global_step": 215637, "epoch": 2422, "lr": 1.010898476057856e-05}
{"train_loss": 0.07412590086460114, "global_step": 215638, "epoch": 2422, "lr": 1.0108635242777704e-05}
{"train_loss": 0.07065287232398987, "global_step": 215639, "epoch": 2422, "lr": 1.0108285730339746e-05}
{"train_loss": 0.0672530084848404, "global_step": 215640, "epoch": 2422, "lr": 1.0107936223264753e-05}
{"train_loss": 0.05266944319009781, "global_step": 215641, "epoch": 2422, "lr": 1.0107586721552753e-05}
{"train_loss": 0.0744340792298317, "global_step": 215642, "epoch": 2422, "lr": 1.0107237225203797e-05}
{"train_loss": 0.059675607830286026, "global_step": 215643, "epoch": 2422, "lr": 1.0106887734217951e-05}
{"train_loss": 0.012298382818698883, "global_step": 215644, "epoch": 2422, "lr": 1.0106538248595238e-05}
{"train_loss": 0.022322915494441986, "global_step": 215645, "epoch": 2422, "lr": 1.0106188768335723e-05}
{"train_loss": 0.053894865180083205, "global_step": 215646, "epoch": 2422, "lr": 1.0105839293439434e-05, "val_loss": 8.623357772827148}
{"train_loss": 0.09598288685083389, "global_step": 215647, "epoch": 2423, "lr": 1.0105489823906444e-05}
{"train_loss": 0.0366269052028656, "global_step": 215648, "epoch": 2423, "lr": 1.0105140359736765e-05}
{"train_loss": 0.030843663960695267, "global_step": 215649, "epoch": 2423, "lr": 1.0104790900930483e-05}
{"train_loss": 0.06927241384983063, "global_step": 215650, "epoch": 2423, "lr": 1.0104441447487611e-05}
{"train_loss": 0.08771459758281708, "global_step": 215651, "epoch": 2423, "lr": 1.0104091999408222e-05}
{"train_loss": 0.058640044182538986, "global_step": 215652, "epoch": 2423, "lr": 1.0103742556692353e-05}
{"train_loss": 0.04794999212026596, "global_step": 215653, "epoch": 2423, "lr": 1.0103393119340037e-05}
{"train_loss": 0.07338457554578781, "global_step": 215654, "epoch": 2423, "lr": 1.0103043687351344e-05}
{"train_loss": 0.023525331169366837, "global_step": 215655, "epoch": 2423, "lr": 1.0102694260726298e-05}
{"train_loss": 0.0493004247546196, "global_step": 215656, "epoch": 2423, "lr": 1.0102344839464972e-05}
{"train_loss": 0.08527172356843948, "global_step": 215657, "epoch": 2423, "lr": 1.0101995423567401e-05}
{"train_loss": 0.04116223007440567, "global_step": 215658, "epoch": 2423, "lr": 1.0101646013033617e-05}
{"train_loss": 0.06001591682434082, "global_step": 215659, "epoch": 2423, "lr": 1.010129660786368e-05}
{"train_loss": 0.04963443800806999, "global_step": 215660, "epoch": 2423, "lr": 1.0100947208057649e-05}
{"train_loss": 0.05997442826628685, "global_step": 215661, "epoch": 2423, "lr": 1.0100597813615547e-05}
{"train_loss": 0.04312792420387268, "global_step": 215662, "epoch": 2423, "lr": 1.0100248424537451e-05}
{"train_loss": 0.06433223932981491, "global_step": 215663, "epoch": 2423, "lr": 1.0099899040823368e-05}
{"train_loss": 0.05464216694235802, "global_step": 215664, "epoch": 2423, "lr": 1.009954966247339e-05}
{"train_loss": 0.08720454573631287, "global_step": 215665, "epoch": 2423, "lr": 1.0099200289487521e-05}
{"train_loss": 0.0699787363409996, "global_step": 215666, "epoch": 2423, "lr": 1.0098850921865844e-05}
{"train_loss": 0.06838638335466385, "global_step": 215667, "epoch": 2423, "lr": 1.0098501559608376e-05}
{"train_loss": 0.055569279938936234, "global_step": 215668, "epoch": 2423, "lr": 1.0098152202715194e-05}
{"train_loss": 0.039214055985212326, "global_step": 215669, "epoch": 2423, "lr": 1.0097802851186311e-05}
{"train_loss": 0.020096968859434128, "global_step": 215670, "epoch": 2423, "lr": 1.0097453505021808e-05}
{"train_loss": 0.0644872710108757, "global_step": 215671, "epoch": 2423, "lr": 1.0097104164221704e-05}
{"train_loss": 0.03843393549323082, "global_step": 215672, "epoch": 2423, "lr": 1.009675482878607e-05}
{"train_loss": 0.07643050700426102, "global_step": 215673, "epoch": 2423, "lr": 1.0096405498714928e-05}
{"train_loss": 0.08421730250120163, "global_step": 215674, "epoch": 2423, "lr": 1.0096056174008356e-05}
{"train_loss": 0.06355904042720795, "global_step": 215675, "epoch": 2423, "lr": 1.0095706854666364e-05}
{"train_loss": 0.07669380307197571, "global_step": 215676, "epoch": 2423, "lr": 1.0095357540689032e-05}
{"train_loss": 0.027340639382600784, "global_step": 215677, "epoch": 2423, "lr": 1.0095008232076381e-05}
{"train_loss": 0.0490712933242321, "global_step": 215678, "epoch": 2423, "lr": 1.0094658928828466e-05}
{"train_loss": 0.053888339549303055, "global_step": 215679, "epoch": 2423, "lr": 1.0094309630945353e-05}
{"train_loss": 0.05795671418309212, "global_step": 215680, "epoch": 2423, "lr": 1.0093960338427061e-05}
{"train_loss": 0.050638530403375626, "global_step": 215681, "epoch": 2423, "lr": 1.009361105127366e-05}
{"train_loss": 0.047826845198869705, "global_step": 215682, "epoch": 2423, "lr": 1.0093261769485179e-05}
{"train_loss": 0.04740314185619354, "global_step": 215683, "epoch": 2423, "lr": 1.0092912493061684e-05}
{"train_loss": 0.0629367083311081, "global_step": 215684, "epoch": 2423, "lr": 1.0092563222003193e-05}
{"train_loss": 0.035679690539836884, "global_step": 215685, "epoch": 2423, "lr": 1.0092213956309788e-05}
{"train_loss": 0.0329262912273407, "global_step": 215686, "epoch": 2423, "lr": 1.0091864695981484e-05}
{"train_loss": 0.050092894583940506, "global_step": 215687, "epoch": 2423, "lr": 1.0091515441018357e-05}
{"train_loss": 0.08021572232246399, "global_step": 215688, "epoch": 2423, "lr": 1.0091166191420426e-05}
{"train_loss": 0.07061433792114258, "global_step": 215689, "epoch": 2423, "lr": 1.0090816947187758e-05}
{"train_loss": 0.11981291323900223, "global_step": 215690, "epoch": 2423, "lr": 1.00904677083204e-05}
{"train_loss": 0.059999044984579086, "global_step": 215691, "epoch": 2423, "lr": 1.0090118474818377e-05}
{"train_loss": 0.10536996275186539, "global_step": 215692, "epoch": 2423, "lr": 1.008976924668177e-05}
{"train_loss": 0.12376869469881058, "global_step": 215693, "epoch": 2423, "lr": 1.0089420023910585e-05}
{"train_loss": 0.06279221177101135, "global_step": 215694, "epoch": 2423, "lr": 1.0089070806504909e-05}
{"train_loss": 0.034386441111564636, "global_step": 215695, "epoch": 2423, "lr": 1.0088721594464756e-05}
{"train_loss": 0.07760293781757355, "global_step": 215696, "epoch": 2423, "lr": 1.0088372387790201e-05}
{"train_loss": 0.055196363478899, "global_step": 215697, "epoch": 2423, "lr": 1.0088023186481266e-05}
{"train_loss": 0.032983534038066864, "global_step": 215698, "epoch": 2423, "lr": 1.008767399053802e-05}
{"train_loss": 0.04220595583319664, "global_step": 215699, "epoch": 2423, "lr": 1.008732479996049e-05}
{"train_loss": 0.055793359875679016, "global_step": 215700, "epoch": 2423, "lr": 1.0086975614748744e-05}
{"train_loss": 0.040682628750801086, "global_step": 215701, "epoch": 2423, "lr": 1.0086626434902808e-05}
{"train_loss": 0.07185067236423492, "global_step": 215702, "epoch": 2423, "lr": 1.008627726042275e-05}
{"train_loss": 0.04644160717725754, "global_step": 215703, "epoch": 2423, "lr": 1.0085928091308589e-05}
{"train_loss": 0.03172295168042183, "global_step": 215704, "epoch": 2423, "lr": 1.0085578927560407e-05}
{"train_loss": 0.05082259327173233, "global_step": 215705, "epoch": 2423, "lr": 1.0085229769178218e-05}
{"train_loss": 0.06932875514030457, "global_step": 215706, "epoch": 2423, "lr": 1.008488061616209e-05}
{"train_loss": 0.051727425307035446, "global_step": 215707, "epoch": 2423, "lr": 1.008453146851206e-05}
{"train_loss": 0.091131791472435, "global_step": 215708, "epoch": 2423, "lr": 1.0084182326228187e-05}
{"train_loss": 0.06447212398052216, "global_step": 215709, "epoch": 2423, "lr": 1.0083833189310493e-05}
{"train_loss": 0.05848365277051926, "global_step": 215710, "epoch": 2423, "lr": 1.0083484057759057e-05}
{"train_loss": 0.03564038127660751, "global_step": 215711, "epoch": 2423, "lr": 1.0083134931573896e-05}
{"train_loss": 0.04831193760037422, "global_step": 215712, "epoch": 2423, "lr": 1.0082785810755085e-05}
{"train_loss": 0.08028910309076309, "global_step": 215713, "epoch": 2423, "lr": 1.0082436695302643e-05}
{"train_loss": 0.046256206929683685, "global_step": 215714, "epoch": 2423, "lr": 1.0082087585216633e-05}
{"train_loss": 0.05609358102083206, "global_step": 215715, "epoch": 2423, "lr": 1.0081738480497111e-05}
{"train_loss": 0.047435611486434937, "global_step": 215716, "epoch": 2423, "lr": 1.0081389381144102e-05}
{"train_loss": 0.03726547583937645, "global_step": 215717, "epoch": 2423, "lr": 1.0081040287157672e-05}
{"train_loss": 0.06280749291181564, "global_step": 215718, "epoch": 2423, "lr": 1.008069119853785e-05}
{"train_loss": 0.07565829902887344, "global_step": 215719, "epoch": 2423, "lr": 1.0080342115284708e-05}
{"train_loss": 0.05683254450559616, "global_step": 215720, "epoch": 2423, "lr": 1.007999303739826e-05}
{"train_loss": 0.045743733644485474, "global_step": 215721, "epoch": 2423, "lr": 1.0079643964878583e-05}
{"train_loss": 0.10349877178668976, "global_step": 215722, "epoch": 2423, "lr": 1.00792948977257e-05}
{"train_loss": 0.05452323332428932, "global_step": 215723, "epoch": 2423, "lr": 1.0078945835939684e-05}
{"train_loss": 0.06702571362257004, "global_step": 215724, "epoch": 2423, "lr": 1.007859677952055e-05}
{"train_loss": 0.03985092043876648, "global_step": 215725, "epoch": 2423, "lr": 1.0078247728468382e-05}
{"train_loss": 0.06297194212675095, "global_step": 215726, "epoch": 2423, "lr": 1.0077898682783188e-05}
{"train_loss": 0.06613412499427795, "global_step": 215727, "epoch": 2423, "lr": 1.0077549642465045e-05}
{"train_loss": 0.06271543353796005, "global_step": 215728, "epoch": 2423, "lr": 1.007720060751398e-05}
{"train_loss": 0.06167619675397873, "global_step": 215729, "epoch": 2423, "lr": 1.0076851577930063e-05}
{"train_loss": 0.027589866891503334, "global_step": 215730, "epoch": 2423, "lr": 1.0076502553713323e-05}
{"train_loss": 0.07960186898708344, "global_step": 215731, "epoch": 2423, "lr": 1.0076153534863803e-05}
{"train_loss": 0.08381600677967072, "global_step": 215732, "epoch": 2423, "lr": 1.0075804521381566e-05}
{"train_loss": 0.0306243933737278, "global_step": 215733, "epoch": 2423, "lr": 1.007545551326664e-05}
{"train_loss": 0.04634741321206093, "global_step": 215734, "epoch": 2423, "lr": 1.0075106510519094e-05}
{"train_loss": 0.05864340537803227, "global_step": 215735, "epoch": 2423, "lr": 1.0074757513138949e-05, "val_loss": 8.813021659851074}
{"train_loss": 0.07986844331026077, "global_step": 215736, "epoch": 2424, "lr": 1.0074408521126284e-05}
{"train_loss": 0.03935125097632408, "global_step": 215737, "epoch": 2424, "lr": 1.0074059534481107e-05}
{"train_loss": 0.06802289932966232, "global_step": 215738, "epoch": 2424, "lr": 1.0073710553203503e-05}
{"train_loss": 0.04666217416524887, "global_step": 215739, "epoch": 2424, "lr": 1.0073361577293488e-05}
{"train_loss": 0.03696342185139656, "global_step": 215740, "epoch": 2424, "lr": 1.0073012606751141e-05}
{"train_loss": 0.04126579314470291, "global_step": 215741, "epoch": 2424, "lr": 1.0072663641576474e-05}
{"train_loss": 0.06951891630887985, "global_step": 215742, "epoch": 2424, "lr": 1.0072314681769562e-05}
{"train_loss": 0.03497384861111641, "global_step": 215743, "epoch": 2424, "lr": 1.007196572733043e-05}
{"train_loss": 0.03414258733391762, "global_step": 215744, "epoch": 2424, "lr": 1.0071616778259151e-05}
{"train_loss": 0.06776128709316254, "global_step": 215745, "epoch": 2424, "lr": 1.0071267834555741e-05}
{"train_loss": 0.02802908606827259, "global_step": 215746, "epoch": 2424, "lr": 1.0070918896220272e-05}
{"train_loss": 0.03596720099449158, "global_step": 215747, "epoch": 2424, "lr": 1.0070569963252774e-05}
{"train_loss": 0.06586314737796783, "global_step": 215748, "epoch": 2424, "lr": 1.007022103565331e-05}
{"train_loss": 0.08169529587030411, "global_step": 215749, "epoch": 2424, "lr": 1.0069872113421908e-05}
{"train_loss": 0.06356817483901978, "global_step": 215750, "epoch": 2424, "lr": 1.0069523196558622e-05}
{"train_loss": 0.03211818262934685, "global_step": 215751, "epoch": 2424, "lr": 1.0069174285063521e-05}
{"train_loss": 0.028300374746322632, "global_step": 215752, "epoch": 2424, "lr": 1.0068825378936614e-05}
{"train_loss": 0.03935084119439125, "global_step": 215753, "epoch": 2424, "lr": 1.006847647817798e-05}
{"train_loss": 0.04329814016819, "global_step": 215754, "epoch": 2424, "lr": 1.006812758278764e-05}
{"train_loss": 0.07097682356834412, "global_step": 215755, "epoch": 2424, "lr": 1.0067778692765668e-05}
{"train_loss": 0.04194340854883194, "global_step": 215756, "epoch": 2424, "lr": 1.0067429808112084e-05}
{"train_loss": 0.05693675950169563, "global_step": 215757, "epoch": 2424, "lr": 1.006708092882696e-05}
{"train_loss": 0.06470323354005814, "global_step": 215758, "epoch": 2424, "lr": 1.0066732054910316e-05}
{"train_loss": 0.04869898036122322, "global_step": 215759, "epoch": 2424, "lr": 1.0066383186362227e-05}
{"train_loss": 0.10083804279565811, "global_step": 215760, "epoch": 2424, "lr": 1.006603432318271e-05}
{"train_loss": 0.03866084665060043, "global_step": 215761, "epoch": 2424, "lr": 1.0065685465371844e-05}
{"train_loss": 0.06539831310510635, "global_step": 215762, "epoch": 2424, "lr": 1.006533661292965e-05}
{"train_loss": 0.048849280923604965, "global_step": 215763, "epoch": 2424, "lr": 1.0064987765856193e-05}
{"train_loss": 0.10397923737764359, "global_step": 215764, "epoch": 2424, "lr": 1.0064638924151498e-05}
{"train_loss": 0.05564136430621147, "global_step": 215765, "epoch": 2424, "lr": 1.0064290087815642e-05}
{"train_loss": 0.04291179031133652, "global_step": 215766, "epoch": 2424, "lr": 1.006394125684864e-05}
{"train_loss": 0.023411480709910393, "global_step": 215767, "epoch": 2424, "lr": 1.0063592431250568e-05}
{"train_loss": 0.04404044523835182, "global_step": 215768, "epoch": 2424, "lr": 1.0063243611021462e-05}
{"train_loss": 0.041065022349357605, "global_step": 215769, "epoch": 2424, "lr": 1.0062894796161348e-05}
{"train_loss": 0.0725979283452034, "global_step": 215770, "epoch": 2424, "lr": 1.0062545986670308e-05}
{"train_loss": 0.06048067659139633, "global_step": 215771, "epoch": 2424, "lr": 1.0062197182548355e-05}
{"train_loss": 0.057802919298410416, "global_step": 215772, "epoch": 2424, "lr": 1.0061848383795569e-05}
{"train_loss": 0.12560570240020752, "global_step": 215773, "epoch": 2424, "lr": 1.0061499590411965e-05}
{"train_loss": 0.07200935482978821, "global_step": 215774, "epoch": 2424, "lr": 1.0061150802397624e-05}
{"train_loss": 0.05839958414435387, "global_step": 215775, "epoch": 2424, "lr": 1.0060802019752558e-05}
{"train_loss": 0.05989154055714607, "global_step": 215776, "epoch": 2424, "lr": 1.0060453242476842e-05}
{"train_loss": 0.04709063842892647, "global_step": 215777, "epoch": 2424, "lr": 1.0060104470570502e-05}
{"train_loss": 0.05172313004732132, "global_step": 215778, "epoch": 2424, "lr": 1.0059755704033608e-05}
{"train_loss": 0.01693510077893734, "global_step": 215779, "epoch": 2424, "lr": 1.0059406942866172e-05}
{"train_loss": 0.06745696812868118, "global_step": 215780, "epoch": 2424, "lr": 1.0059058187068282e-05}
{"train_loss": 0.0639626681804657, "global_step": 215781, "epoch": 2424, "lr": 1.0058709436639952e-05}
{"train_loss": 0.03377077355980873, "global_step": 215782, "epoch": 2424, "lr": 1.0058360691581248e-05}
{"train_loss": 0.062477704137563705, "global_step": 215783, "epoch": 2424, "lr": 1.0058011951892205e-05}
{"train_loss": 0.03971821814775467, "global_step": 215784, "epoch": 2424, "lr": 1.0057663217572883e-05}
{"train_loss": 0.08350690454244614, "global_step": 215785, "epoch": 2424, "lr": 1.0057314488623315e-05}
{"train_loss": 0.042183373123407364, "global_step": 215786, "epoch": 2424, "lr": 1.0056965765043547e-05}
{"train_loss": 0.037509508430957794, "global_step": 215787, "epoch": 2424, "lr": 1.0056617046833649e-05}
{"train_loss": 0.0727686882019043, "global_step": 215788, "epoch": 2424, "lr": 1.005626833399364e-05}
{"train_loss": 0.04982541874051094, "global_step": 215789, "epoch": 2424, "lr": 1.0055919626523591e-05}
{"train_loss": 0.05709541589021683, "global_step": 215790, "epoch": 2424, "lr": 1.0055570924423524e-05}
{"train_loss": 0.030254865065217018, "global_step": 215791, "epoch": 2424, "lr": 1.0055222227693512e-05}
{"train_loss": 0.09870412200689316, "global_step": 215792, "epoch": 2424, "lr": 1.0054873536333575e-05}
{"train_loss": 0.04123896732926369, "global_step": 215793, "epoch": 2424, "lr": 1.0054524850343788e-05}
{"train_loss": 0.04813216254115105, "global_step": 215794, "epoch": 2424, "lr": 1.0054176169724166e-05}
{"train_loss": 0.15002204477787018, "global_step": 215795, "epoch": 2424, "lr": 1.0053827494474788e-05}
{"train_loss": 0.04574403539299965, "global_step": 215796, "epoch": 2424, "lr": 1.0053478824595674e-05}
{"train_loss": 0.07011517882347107, "global_step": 215797, "epoch": 2424, "lr": 1.0053130160086898e-05}
{"train_loss": 0.09103929996490479, "global_step": 215798, "epoch": 2424, "lr": 1.0052781500948478e-05}
{"train_loss": 0.04976964369416237, "global_step": 215799, "epoch": 2424, "lr": 1.0052432847180481e-05}
{"train_loss": 0.09341564774513245, "global_step": 215800, "epoch": 2424, "lr": 1.005208419878294e-05}
{"train_loss": 0.04885876178741455, "global_step": 215801, "epoch": 2424, "lr": 1.0051735555755926e-05}
{"train_loss": 0.059042349457740784, "global_step": 215802, "epoch": 2424, "lr": 1.0051386918099454e-05}
{"train_loss": 0.07726309448480606, "global_step": 215803, "epoch": 2424, "lr": 1.0051038285813597e-05}
{"train_loss": 0.0752076655626297, "global_step": 215804, "epoch": 2424, "lr": 1.0050689658898376e-05}
{"train_loss": 0.0548148974776268, "global_step": 215805, "epoch": 2424, "lr": 1.0050341037353866e-05}
{"train_loss": 0.029342129826545715, "global_step": 215806, "epoch": 2424, "lr": 1.0049992421180104e-05}
{"train_loss": 0.07288558036088943, "global_step": 215807, "epoch": 2424, "lr": 1.004964381037712e-05}
{"train_loss": 0.030123943462967873, "global_step": 215808, "epoch": 2424, "lr": 1.0049295204944987e-05}
{"train_loss": 0.0613093301653862, "global_step": 215809, "epoch": 2424, "lr": 1.004894660488373e-05}
{"train_loss": 0.044466715306043625, "global_step": 215810, "epoch": 2424, "lr": 1.0048598010193412e-05}
{"train_loss": 0.06878655403852463, "global_step": 215811, "epoch": 2424, "lr": 1.0048249420874062e-05}
{"train_loss": 0.050080448389053345, "global_step": 215812, "epoch": 2424, "lr": 1.004790083692575e-05}
{"train_loss": 0.06187933683395386, "global_step": 215813, "epoch": 2424, "lr": 1.0047552258348502e-05}
{"train_loss": 0.07866475731134415, "global_step": 215814, "epoch": 2424, "lr": 1.0047203685142382e-05}
{"train_loss": 0.07141800224781036, "global_step": 215815, "epoch": 2424, "lr": 1.004685511730742e-05}
{"train_loss": 0.05562597140669823, "global_step": 215816, "epoch": 2424, "lr": 1.0046506554843681e-05}
{"train_loss": 0.031778961420059204, "global_step": 215817, "epoch": 2424, "lr": 1.0046157997751187e-05}
{"train_loss": 0.057033978402614594, "global_step": 215818, "epoch": 2424, "lr": 1.0045809446030019e-05}
{"train_loss": 0.07996022701263428, "global_step": 215819, "epoch": 2424, "lr": 1.0045460899680186e-05}
{"train_loss": 0.09341692924499512, "global_step": 215820, "epoch": 2424, "lr": 1.0045112358701774e-05}
{"train_loss": 0.0725012868642807, "global_step": 215821, "epoch": 2424, "lr": 1.0044763823094795e-05}
{"train_loss": 0.12349872291088104, "global_step": 215822, "epoch": 2424, "lr": 1.0044415292859317e-05}
{"train_loss": 0.07540957629680634, "global_step": 215823, "epoch": 2424, "lr": 1.0044066767995374e-05}
{"train_loss": 0.05923178061591775, "global_step": 215824, "epoch": 2424, "lr": 1.0043718248503015e-05, "val_loss": 8.714173316955566}
{"train_loss": 0.02974768914282322, "global_step": 215825, "epoch": 2425, "lr": 1.0043369734382308e-05}
{"train_loss": 0.08466960489749908, "global_step": 215826, "epoch": 2425, "lr": 1.0043021225633264e-05}
{"train_loss": 0.07804212719202042, "global_step": 215827, "epoch": 2425, "lr": 1.004267272225597e-05}
{"train_loss": 0.06488733738660812, "global_step": 215828, "epoch": 2425, "lr": 1.0042324224250432e-05}
{"train_loss": 0.012568868696689606, "global_step": 215829, "epoch": 2425, "lr": 1.0041975731616732e-05}
{"train_loss": 0.026929538697004318, "global_step": 215830, "epoch": 2425, "lr": 1.0041627244354885e-05}
{"train_loss": 0.07830578088760376, "global_step": 215831, "epoch": 2425, "lr": 1.0041278762464973e-05}
{"train_loss": 0.10322052985429764, "global_step": 215832, "epoch": 2425, "lr": 1.004093028594701e-05}
{"train_loss": 0.07080092281103134, "global_step": 215833, "epoch": 2425, "lr": 1.0040581814801065e-05}
{"train_loss": 0.05469411984086037, "global_step": 215834, "epoch": 2425, "lr": 1.0040233349027168e-05}
{"train_loss": 0.05127149447798729, "global_step": 215835, "epoch": 2425, "lr": 1.0039884888625384e-05}
{"train_loss": 0.05420256406068802, "global_step": 215836, "epoch": 2425, "lr": 1.0039536433595742e-05}
{"train_loss": 0.028630008921027184, "global_step": 215837, "epoch": 2425, "lr": 1.0039187983938308e-05}
{"train_loss": 0.0470031276345253, "global_step": 215838, "epoch": 2425, "lr": 1.0038839539653105e-05}
{"train_loss": 0.03894776105880737, "global_step": 215839, "epoch": 2425, "lr": 1.0038491100740204e-05}
{"train_loss": 0.10122859477996826, "global_step": 215840, "epoch": 2425, "lr": 1.0038142667199634e-05}
{"train_loss": 0.05384492129087448, "global_step": 215841, "epoch": 2425, "lr": 1.0037794239031457e-05}
{"train_loss": 0.04189632833003998, "global_step": 215842, "epoch": 2425, "lr": 1.0037445816235696e-05}
{"train_loss": 0.07427478581666946, "global_step": 215843, "epoch": 2425, "lr": 1.0037097398812433e-05}
{"train_loss": 0.05181727558374405, "global_step": 215844, "epoch": 2425, "lr": 1.0036748986761679e-05}
{"train_loss": 0.02141338773071766, "global_step": 215845, "epoch": 2425, "lr": 1.003640058008351e-05}
{"train_loss": 0.06459084898233414, "global_step": 215846, "epoch": 2425, "lr": 1.0036052178777961e-05}
{"train_loss": 0.04382708668708801, "global_step": 215847, "epoch": 2425, "lr": 1.0035703782845063e-05}
{"train_loss": 0.023937981575727463, "global_step": 215848, "epoch": 2425, "lr": 1.0035355392284889e-05}
{"train_loss": 0.08030475676059723, "global_step": 215849, "epoch": 2425, "lr": 1.0035007007097464e-05}
{"train_loss": 0.040655024349689484, "global_step": 215850, "epoch": 2425, "lr": 1.0034658627282855e-05}
{"train_loss": 0.12119746208190918, "global_step": 215851, "epoch": 2425, "lr": 1.0034310252841089e-05}
{"train_loss": 0.05540098994970322, "global_step": 215852, "epoch": 2425, "lr": 1.0033961883772236e-05}
{"train_loss": 0.07911205291748047, "global_step": 215853, "epoch": 2425, "lr": 1.0033613520076312e-05}
{"train_loss": 0.02848157286643982, "global_step": 215854, "epoch": 2425, "lr": 1.0033265161753403e-05}
{"train_loss": 0.06254032999277115, "global_step": 215855, "epoch": 2425, "lr": 1.0032916808803517e-05}
{"train_loss": 0.05220439285039902, "global_step": 215856, "epoch": 2425, "lr": 1.0032568461226733e-05}
{"train_loss": 0.08575168997049332, "global_step": 215857, "epoch": 2425, "lr": 1.0032220119023067e-05}
{"train_loss": 0.0449443943798542, "global_step": 215858, "epoch": 2425, "lr": 1.0031871782192592e-05}
{"train_loss": 0.06168943643569946, "global_step": 215859, "epoch": 2425, "lr": 1.0031523450735337e-05}
{"train_loss": 0.05665654316544533, "global_step": 215860, "epoch": 2425, "lr": 1.0031175124651354e-05}
{"train_loss": 0.09895071387290955, "global_step": 215861, "epoch": 2425, "lr": 1.0030826803940707e-05}
{"train_loss": 0.061663366854190826, "global_step": 215862, "epoch": 2425, "lr": 1.0030478488603417e-05}
{"train_loss": 0.06146695837378502, "global_step": 215863, "epoch": 2425, "lr": 1.0030130178639551e-05}
{"train_loss": 0.07790544629096985, "global_step": 215864, "epoch": 2425, "lr": 1.0029781874049137e-05}
{"train_loss": 0.09398933500051498, "global_step": 215865, "epoch": 2425, "lr": 1.0029433574832241e-05}
{"train_loss": 0.0663389191031456, "global_step": 215866, "epoch": 2425, "lr": 1.0029085280988892e-05}
{"train_loss": 0.07466885447502136, "global_step": 215867, "epoch": 2425, "lr": 1.0028736992519155e-05}
{"train_loss": 0.020693818107247353, "global_step": 215868, "epoch": 2425, "lr": 1.0028388709423054e-05}
{"train_loss": 0.03775191679596901, "global_step": 215869, "epoch": 2425, "lr": 1.0028040431700664e-05}
{"train_loss": 0.05045223981142044, "global_step": 215870, "epoch": 2425, "lr": 1.0027692159352004e-05}
{"train_loss": 0.047436248511075974, "global_step": 215871, "epoch": 2425, "lr": 1.0027343892377144e-05}
{"train_loss": 0.038177940994501114, "global_step": 215872, "epoch": 2425, "lr": 1.0026995630776109e-05}
{"train_loss": 0.0484648123383522, "global_step": 215873, "epoch": 2425, "lr": 1.002664737454897e-05}
{"train_loss": 0.046034812927246094, "global_step": 215874, "epoch": 2425, "lr": 1.0026299123695748e-05}
{"train_loss": 0.07862065732479095, "global_step": 215875, "epoch": 2425, "lr": 1.0025950878216516e-05}
{"train_loss": 0.022808346897363663, "global_step": 215876, "epoch": 2425, "lr": 1.0025602638111298e-05}
{"train_loss": 0.03583874925971031, "global_step": 215877, "epoch": 2425, "lr": 1.0025254403380147e-05}
{"train_loss": 0.034971386194229126, "global_step": 215878, "epoch": 2425, "lr": 1.0024906174023125e-05}
{"train_loss": 0.04213881865143776, "global_step": 215879, "epoch": 2425, "lr": 1.002455795004027e-05}
{"train_loss": 0.06309696286916733, "global_step": 215880, "epoch": 2425, "lr": 1.0024209731431611e-05}
{"train_loss": 0.03120127134025097, "global_step": 215881, "epoch": 2425, "lr": 1.0023861518197224e-05}
{"train_loss": 0.0360243134200573, "global_step": 215882, "epoch": 2425, "lr": 1.002351331033713e-05}
{"train_loss": 0.06118940934538841, "global_step": 215883, "epoch": 2425, "lr": 1.0023165107851396e-05}
{"train_loss": 0.040114875882864, "global_step": 215884, "epoch": 2425, "lr": 1.0022816910740063e-05}
{"train_loss": 0.033949002623558044, "global_step": 215885, "epoch": 2425, "lr": 1.0022468719003158e-05}
{"train_loss": 0.07906675338745117, "global_step": 215886, "epoch": 2425, "lr": 1.0022120532640767e-05}
{"train_loss": 0.06427036225795746, "global_step": 215887, "epoch": 2425, "lr": 1.0021772351652892e-05}
{"train_loss": 0.0338108167052269, "global_step": 215888, "epoch": 2425, "lr": 1.0021424176039622e-05}
{"train_loss": 0.070476233959198, "global_step": 215889, "epoch": 2425, "lr": 1.0021076005800967e-05}
{"train_loss": 0.05823395028710365, "global_step": 215890, "epoch": 2425, "lr": 1.0020727840937005e-05}
{"train_loss": 0.1110236644744873, "global_step": 215891, "epoch": 2425, "lr": 1.0020379681447756e-05}
{"train_loss": 0.08942372351884842, "global_step": 215892, "epoch": 2425, "lr": 1.0020031527333296e-05}
{"train_loss": 0.022523000836372375, "global_step": 215893, "epoch": 2425, "lr": 1.0019683378593641e-05}
{"train_loss": 0.05416034534573555, "global_step": 215894, "epoch": 2425, "lr": 1.001933523522886e-05}
{"train_loss": 0.06432338058948517, "global_step": 215895, "epoch": 2425, "lr": 1.0018987097238985e-05}
{"train_loss": 0.10730300843715668, "global_step": 215896, "epoch": 2425, "lr": 1.0018638964624066e-05}
{"train_loss": 0.07002811133861542, "global_step": 215897, "epoch": 2425, "lr": 1.001829083738417e-05}
{"train_loss": 0.06270947307348251, "global_step": 215898, "epoch": 2425, "lr": 1.0017942715519313e-05}
{"train_loss": 0.06510058045387268, "global_step": 215899, "epoch": 2425, "lr": 1.0017594599029573e-05}
{"train_loss": 0.06094396486878395, "global_step": 215900, "epoch": 2425, "lr": 1.0017246487914961e-05}
{"train_loss": 0.0571993887424469, "global_step": 215901, "epoch": 2425, "lr": 1.001689838217556e-05}
{"train_loss": 0.043821997940540314, "global_step": 215902, "epoch": 2425, "lr": 1.0016550281811382e-05}
{"train_loss": 0.06225891038775444, "global_step": 215903, "epoch": 2425, "lr": 1.001620218682251e-05}
{"train_loss": 0.08281897008419037, "global_step": 215904, "epoch": 2425, "lr": 1.001585409720896e-05}
{"train_loss": 0.05220579728484154, "global_step": 215905, "epoch": 2425, "lr": 1.0015506012970799e-05}
{"train_loss": 0.046254437416791916, "global_step": 215906, "epoch": 2425, "lr": 1.0015157934108055e-05}
{"train_loss": 0.0595097579061985, "global_step": 215907, "epoch": 2425, "lr": 1.00148098606208e-05}
{"train_loss": 0.07469728589057922, "global_step": 215908, "epoch": 2425, "lr": 1.001446179250905e-05}
{"train_loss": 0.0645972341299057, "global_step": 215909, "epoch": 2425, "lr": 1.0014113729772884e-05}
{"train_loss": 0.030913449823856354, "global_step": 215910, "epoch": 2425, "lr": 1.0013765672412318e-05}
{"train_loss": 0.06470026820898056, "global_step": 215911, "epoch": 2425, "lr": 1.0013417620427429e-05}
{"train_loss": 0.015990493819117546, "global_step": 215912, "epoch": 2425, "lr": 1.0013069573818234e-05}
{"train_loss": 0.05750334945120168, "global_step": 215913, "epoch": 2425, "lr": 1.0012721532584795e-05, "val_loss": 8.999377250671387, "train_action_mse_error": 20.999780654907227}
{"train_loss": 0.03745915740728378, "global_step": 215914, "epoch": 2426, "lr": 1.0012373496727173e-05}
{"train_loss": 0.0364808551967144, "global_step": 215915, "epoch": 2426, "lr": 1.0012025466245384e-05}
{"train_loss": 0.031810782849788666, "global_step": 215916, "epoch": 2426, "lr": 1.0011677441139505e-05}
{"train_loss": 0.05311070755124092, "global_step": 215917, "epoch": 2426, "lr": 1.001132942140957e-05}
{"train_loss": 0.05185731500387192, "global_step": 215918, "epoch": 2426, "lr": 1.0010981407055609e-05}
{"train_loss": 0.07836993038654327, "global_step": 215919, "epoch": 2426, "lr": 1.0010633398077701e-05}
{"train_loss": 0.08164032548666, "global_step": 215920, "epoch": 2426, "lr": 1.001028539447586e-05}
{"train_loss": 0.0624275766313076, "global_step": 215921, "epoch": 2426, "lr": 1.0009937396250163e-05}
{"train_loss": 0.07186818867921829, "global_step": 215922, "epoch": 2426, "lr": 1.0009589403400637e-05}
{"train_loss": 0.05872555449604988, "global_step": 215923, "epoch": 2426, "lr": 1.0009241415927329e-05}
{"train_loss": 0.04885140061378479, "global_step": 215924, "epoch": 2426, "lr": 1.0008893433830302e-05}
{"train_loss": 0.03745407238602638, "global_step": 215925, "epoch": 2426, "lr": 1.0008545457109576e-05}
{"train_loss": 0.005302656441926956, "global_step": 215926, "epoch": 2426, "lr": 1.000819748576523e-05}
{"train_loss": 0.025768907740712166, "global_step": 215927, "epoch": 2426, "lr": 1.000784951979728e-05}
{"train_loss": 0.033696580678224564, "global_step": 215928, "epoch": 2426, "lr": 1.0007501559205802e-05}
{"train_loss": 0.042470380663871765, "global_step": 215929, "epoch": 2426, "lr": 1.000715360399081e-05}
{"train_loss": 0.0977901890873909, "global_step": 215930, "epoch": 2426, "lr": 1.0006805654152385e-05}
{"train_loss": 0.10061554610729218, "global_step": 215931, "epoch": 2426, "lr": 1.0006457709690542e-05}
{"train_loss": 0.05248432978987694, "global_step": 215932, "epoch": 2426, "lr": 1.000610977060535e-05}
{"train_loss": 0.03986752778291702, "global_step": 215933, "epoch": 2426, "lr": 1.0005761836896855e-05}
{"train_loss": 0.08605445176362991, "global_step": 215934, "epoch": 2426, "lr": 1.0005413908565086e-05}
{"train_loss": 0.06513495743274689, "global_step": 215935, "epoch": 2426, "lr": 1.0005065985610119e-05}
{"train_loss": 0.07209265232086182, "global_step": 215936, "epoch": 2426, "lr": 1.0004718068031965e-05}
{"train_loss": 0.04401301220059395, "global_step": 215937, "epoch": 2426, "lr": 1.0004370155830706e-05}
{"train_loss": 0.0709604024887085, "global_step": 215938, "epoch": 2426, "lr": 1.0004022249006356e-05}
{"train_loss": 0.035253871232271194, "global_step": 215939, "epoch": 2426, "lr": 1.0003674347558995e-05}
{"train_loss": 0.03885331377387047, "global_step": 215940, "epoch": 2426, "lr": 1.0003326451488636e-05}
{"train_loss": 0.03315430507063866, "global_step": 215941, "epoch": 2426, "lr": 1.0002978560795356e-05}
{"train_loss": 0.11168263852596283, "global_step": 215942, "epoch": 2426, "lr": 1.0002630675479173e-05}
{"train_loss": 0.06295903772115707, "global_step": 215943, "epoch": 2426, "lr": 1.0002282795540163e-05}
{"train_loss": 0.058403097093105316, "global_step": 215944, "epoch": 2426, "lr": 1.000193492097835e-05}
{"train_loss": 0.03567307069897652, "global_step": 215945, "epoch": 2426, "lr": 1.0001587051793798e-05}
{"train_loss": 0.09419755637645721, "global_step": 215946, "epoch": 2426, "lr": 1.0001239187986533e-05}
{"train_loss": 0.04143724963068962, "global_step": 215947, "epoch": 2426, "lr": 1.0000891329556628e-05}
{"train_loss": 0.04490441828966141, "global_step": 215948, "epoch": 2426, "lr": 1.0000543476504098e-05}
{"train_loss": 0.08498355746269226, "global_step": 215949, "epoch": 2426, "lr": 1.000019562882903e-05}
{"train_loss": 0.06254886835813522, "global_step": 215950, "epoch": 2426, "lr": 9.999847786531424e-06}
{"train_loss": 0.06663485616445541, "global_step": 215951, "epoch": 2426, "lr": 9.999499949611356e-06}
{"train_loss": 0.046241696923971176, "global_step": 215952, "epoch": 2426, "lr": 9.999152118068882e-06}
{"train_loss": 0.04158302769064903, "global_step": 215953, "epoch": 2426, "lr": 9.998804291904017e-06}
{"train_loss": 0.09613700211048126, "global_step": 215954, "epoch": 2426, "lr": 9.998456471116846e-06}
{"train_loss": 0.06643982976675034, "global_step": 215955, "epoch": 2426, "lr": 9.998108655707378e-06}
{"train_loss": 0.07052946835756302, "global_step": 215956, "epoch": 2426, "lr": 9.997760845675685e-06}
{"train_loss": 0.06018245220184326, "global_step": 215957, "epoch": 2426, "lr": 9.997413041021813e-06}
{"train_loss": 0.08114678412675858, "global_step": 215958, "epoch": 2426, "lr": 9.997065241745784e-06}
{"train_loss": 0.02494506537914276, "global_step": 215959, "epoch": 2426, "lr": 9.996717447847675e-06}
{"train_loss": 0.09945673495531082, "global_step": 215960, "epoch": 2426, "lr": 9.996369659327509e-06}
{"train_loss": 0.04905780032277107, "global_step": 215961, "epoch": 2426, "lr": 9.996021876185358e-06}
{"train_loss": 0.05245174840092659, "global_step": 215962, "epoch": 2426, "lr": 9.995674098421248e-06}
{"train_loss": 0.06299217790365219, "global_step": 215963, "epoch": 2426, "lr": 9.995326326035221e-06}
{"train_loss": 0.04182143881917, "global_step": 215964, "epoch": 2426, "lr": 9.994978559027352e-06}
{"train_loss": 0.025485048070549965, "global_step": 215965, "epoch": 2426, "lr": 9.994630797397653e-06}
{"train_loss": 0.034570369869470596, "global_step": 215966, "epoch": 2426, "lr": 9.994283041146201e-06}
{"train_loss": 0.07165423035621643, "global_step": 215967, "epoch": 2426, "lr": 9.99393529027302e-06}
{"train_loss": 0.06183755397796631, "global_step": 215968, "epoch": 2426, "lr": 9.993587544778166e-06}
{"train_loss": 0.049792949110269547, "global_step": 215969, "epoch": 2426, "lr": 9.993239804661702e-06}
{"train_loss": 0.04105767607688904, "global_step": 215970, "epoch": 2426, "lr": 9.992892069923642e-06}
{"train_loss": 0.07759415358304977, "global_step": 215971, "epoch": 2426, "lr": 9.992544340564064e-06}
{"train_loss": 0.06768396496772766, "global_step": 215972, "epoch": 2426, "lr": 9.992196616582988e-06}
{"train_loss": 0.05134029686450958, "global_step": 215973, "epoch": 2426, "lr": 9.991848897980488e-06}
{"train_loss": 0.05176784098148346, "global_step": 215974, "epoch": 2426, "lr": 9.991501184756585e-06}
{"train_loss": 0.06087241694331169, "global_step": 215975, "epoch": 2426, "lr": 9.991153476911347e-06}
{"train_loss": 0.04028443619608879, "global_step": 215976, "epoch": 2426, "lr": 9.9908057744448e-06}
{"train_loss": 0.04759879410266876, "global_step": 215977, "epoch": 2426, "lr": 9.990458077357012e-06}
{"train_loss": 0.05748876556754112, "global_step": 215978, "epoch": 2426, "lr": 9.990110385648006e-06}
{"train_loss": 0.019793309271335602, "global_step": 215979, "epoch": 2426, "lr": 9.989762699317857e-06}
{"train_loss": 0.06428775191307068, "global_step": 215980, "epoch": 2426, "lr": 9.989415018366582e-06}
{"train_loss": 0.029363295063376427, "global_step": 215981, "epoch": 2426, "lr": 9.989067342794256e-06}
{"train_loss": 0.024271298199892044, "global_step": 215982, "epoch": 2426, "lr": 9.9887196726009e-06}
{"train_loss": 0.05233903229236603, "global_step": 215983, "epoch": 2426, "lr": 9.988372007786583e-06}
{"train_loss": 0.07630924135446548, "global_step": 215984, "epoch": 2426, "lr": 9.988024348351333e-06}
{"train_loss": 0.0431029237806797, "global_step": 215985, "epoch": 2426, "lr": 9.987676694295218e-06}
{"train_loss": 0.056634627282619476, "global_step": 215986, "epoch": 2426, "lr": 9.987329045618255e-06}
{"train_loss": 0.0511820986866951, "global_step": 215987, "epoch": 2426, "lr": 9.986981402320512e-06}
{"train_loss": 0.054359033703804016, "global_step": 215988, "epoch": 2426, "lr": 9.986633764402043e-06}
{"train_loss": 0.02537863329052925, "global_step": 215989, "epoch": 2426, "lr": 9.98628613186287e-06}
{"train_loss": 0.0741332545876503, "global_step": 215990, "epoch": 2426, "lr": 9.985938504703068e-06}
{"train_loss": 0.07250002771615982, "global_step": 215991, "epoch": 2426, "lr": 9.985590882922657e-06}
{"train_loss": 0.04095673933625221, "global_step": 215992, "epoch": 2426, "lr": 9.985243266521705e-06}
{"train_loss": 0.03895445913076401, "global_step": 215993, "epoch": 2426, "lr": 9.984895655500237e-06}
{"train_loss": 0.04190962761640549, "global_step": 215994, "epoch": 2426, "lr": 9.984548049858328e-06}
{"train_loss": 0.03388301655650139, "global_step": 215995, "epoch": 2426, "lr": 9.984200449596003e-06}
{"train_loss": 0.03614483028650284, "global_step": 215996, "epoch": 2426, "lr": 9.983852854713305e-06}
{"train_loss": 0.10225050151348114, "global_step": 215997, "epoch": 2426, "lr": 9.983505265210307e-06}
{"train_loss": 0.01835177093744278, "global_step": 215998, "epoch": 2426, "lr": 9.983157681087018e-06}
{"train_loss": 0.07804889976978302, "global_step": 215999, "epoch": 2426, "lr": 9.982810102343526e-06}
{"train_loss": 0.035393793135881424, "global_step": 216000, "epoch": 2426, "lr": 9.982462528979852e-06}
{"train_loss": 0.03788895159959793, "global_step": 216001, "epoch": 2426, "lr": 9.982114960996042e-06}
{"train_loss": 0.05497180521906762, "global_step": 216002, "epoch": 2426, "lr": 9.981767398392156e-06, "val_loss": 8.935633659362793}
{"train_loss": 0.05348099023103714, "global_step": 216003, "epoch": 2427, "lr": 9.981419841168221e-06}
{"train_loss": 0.0624869167804718, "global_step": 216004, "epoch": 2427, "lr": 9.9810722893243e-06}
{"train_loss": 0.05487428978085518, "global_step": 216005, "epoch": 2427, "lr": 9.980724742860448e-06}
{"train_loss": 0.05143878981471062, "global_step": 216006, "epoch": 2427, "lr": 9.980377201776681e-06}
{"train_loss": 0.09260567277669907, "global_step": 216007, "epoch": 2427, "lr": 9.980029666073081e-06}
{"train_loss": 0.04279714822769165, "global_step": 216008, "epoch": 2427, "lr": 9.979682135749669e-06}
{"train_loss": 0.07209739089012146, "global_step": 216009, "epoch": 2427, "lr": 9.979334610806513e-06}
{"train_loss": 0.05864543467760086, "global_step": 216010, "epoch": 2427, "lr": 9.97898709124363e-06}
{"train_loss": 0.054373741149902344, "global_step": 216011, "epoch": 2427, "lr": 9.9786395770611e-06}
{"train_loss": 0.03342204540967941, "global_step": 216012, "epoch": 2427, "lr": 9.978292068258943e-06}
{"train_loss": 0.06896138191223145, "global_step": 216013, "epoch": 2427, "lr": 9.977944564837228e-06}
{"train_loss": 0.0234241746366024, "global_step": 216014, "epoch": 2427, "lr": 9.977597066795979e-06}
{"train_loss": 0.04056679084897041, "global_step": 216015, "epoch": 2427, "lr": 9.977249574135267e-06}
{"train_loss": 0.0745539590716362, "global_step": 216016, "epoch": 2427, "lr": 9.97690208685511e-06}
{"train_loss": 0.08173862844705582, "global_step": 216017, "epoch": 2427, "lr": 9.976554604955584e-06}
{"train_loss": 0.08493649214506149, "global_step": 216018, "epoch": 2427, "lr": 9.97620712843671e-06}
{"train_loss": 0.050197675824165344, "global_step": 216019, "epoch": 2427, "lr": 9.975859657298565e-06}
{"train_loss": 0.06324096024036407, "global_step": 216020, "epoch": 2427, "lr": 9.975512191541159e-06}
{"train_loss": 0.04476587101817131, "global_step": 216021, "epoch": 2427, "lr": 9.975164731164572e-06}
{"train_loss": 0.026655377820134163, "global_step": 216022, "epoch": 2427, "lr": 9.974817276168824e-06}
{"train_loss": 0.08240266144275665, "global_step": 216023, "epoch": 2427, "lr": 9.974469826553974e-06}
{"train_loss": 0.026873286813497543, "global_step": 216024, "epoch": 2427, "lr": 9.974122382320083e-06}
{"train_loss": 0.03672952950000763, "global_step": 216025, "epoch": 2427, "lr": 9.973774943467169e-06}
{"train_loss": 0.055497996509075165, "global_step": 216026, "epoch": 2427, "lr": 9.973427509995304e-06}
{"train_loss": 0.05205155536532402, "global_step": 216027, "epoch": 2427, "lr": 9.973080081904513e-06}
{"train_loss": 0.06815177947282791, "global_step": 216028, "epoch": 2427, "lr": 9.972732659194866e-06}
{"train_loss": 0.05207717418670654, "global_step": 216029, "epoch": 2427, "lr": 9.972385241866384e-06}
{"train_loss": 0.022264599800109863, "global_step": 216030, "epoch": 2427, "lr": 9.97203782991914e-06}
{"train_loss": 0.05342501774430275, "global_step": 216031, "epoch": 2427, "lr": 9.971690423353158e-06}
{"train_loss": 0.03992309048771858, "global_step": 216032, "epoch": 2427, "lr": 9.971343022168505e-06}
{"train_loss": 0.07242733985185623, "global_step": 216033, "epoch": 2427, "lr": 9.970995626365214e-06}
{"train_loss": 0.0770145133137703, "global_step": 216034, "epoch": 2427, "lr": 9.970648235943325e-06}
{"train_loss": 0.03130241110920906, "global_step": 216035, "epoch": 2427, "lr": 9.970300850902909e-06}
{"train_loss": 0.034262798726558685, "global_step": 216036, "epoch": 2427, "lr": 9.96995347124398e-06}
{"train_loss": 0.05162210017442703, "global_step": 216037, "epoch": 2427, "lr": 9.969606096966622e-06}
{"train_loss": 0.03356431424617767, "global_step": 216038, "epoch": 2427, "lr": 9.969258728070846e-06}
{"train_loss": 0.09703703969717026, "global_step": 216039, "epoch": 2427, "lr": 9.968911364556726e-06}
{"train_loss": 0.09954556077718735, "global_step": 216040, "epoch": 2427, "lr": 9.968564006424292e-06}
{"train_loss": 0.03202444314956665, "global_step": 216041, "epoch": 2427, "lr": 9.9682166536736e-06}
{"train_loss": 0.0536850206553936, "global_step": 216042, "epoch": 2427, "lr": 9.96786930630469e-06}
{"train_loss": 0.05011634901165962, "global_step": 216043, "epoch": 2427, "lr": 9.967521964317617e-06}
{"train_loss": 0.06933135539293289, "global_step": 216044, "epoch": 2427, "lr": 9.967174627712416e-06}
{"train_loss": 0.029731381684541702, "global_step": 216045, "epoch": 2427, "lr": 9.966827296489151e-06}
{"train_loss": 0.03392519801855087, "global_step": 216046, "epoch": 2427, "lr": 9.96647997064784e-06}
{"train_loss": 0.11831067502498627, "global_step": 216047, "epoch": 2427, "lr": 9.96613265018857e-06}
{"train_loss": 0.06328660249710083, "global_step": 216048, "epoch": 2427, "lr": 9.96578533511135e-06}
{"train_loss": 0.10654591023921967, "global_step": 216049, "epoch": 2427, "lr": 9.965438025416252e-06}
{"train_loss": 0.0587555393576622, "global_step": 216050, "epoch": 2427, "lr": 9.965090721103298e-06}
{"train_loss": 0.10485441237688065, "global_step": 216051, "epoch": 2427, "lr": 9.964743422172568e-06}
{"train_loss": 0.04175802320241928, "global_step": 216052, "epoch": 2427, "lr": 9.964396128624075e-06}
{"train_loss": 0.08248455077409744, "global_step": 216053, "epoch": 2427, "lr": 9.964048840457896e-06}
{"train_loss": 0.08573298156261444, "global_step": 216054, "epoch": 2427, "lr": 9.963701557674048e-06}
{"train_loss": 0.06719435751438141, "global_step": 216055, "epoch": 2427, "lr": 9.963354280272602e-06}
{"train_loss": 0.04583611711859703, "global_step": 216056, "epoch": 2427, "lr": 9.963007008253589e-06}
{"train_loss": 0.05182958394289017, "global_step": 216057, "epoch": 2427, "lr": 9.96265974161707e-06}
{"train_loss": 0.025246739387512207, "global_step": 216058, "epoch": 2427, "lr": 9.962312480363073e-06}
{"train_loss": 0.04929810389876366, "global_step": 216059, "epoch": 2427, "lr": 9.961965224491649e-06}
{"train_loss": 0.05057023838162422, "global_step": 216060, "epoch": 2427, "lr": 9.961617974002869e-06}
{"train_loss": 0.07034744322299957, "global_step": 216061, "epoch": 2427, "lr": 9.961270728896749e-06}
{"train_loss": 0.05601058900356293, "global_step": 216062, "epoch": 2427, "lr": 9.960923489173363e-06}
{"train_loss": 0.0430382676422596, "global_step": 216063, "epoch": 2427, "lr": 9.960576254832727e-06}
{"train_loss": 0.049256790429353714, "global_step": 216064, "epoch": 2427, "lr": 9.960229025874913e-06}
{"train_loss": 0.048584144562482834, "global_step": 216065, "epoch": 2427, "lr": 9.959881802299947e-06}
{"train_loss": 0.03737235814332962, "global_step": 216066, "epoch": 2427, "lr": 9.959534584107905e-06}
{"train_loss": 0.026835419237613678, "global_step": 216067, "epoch": 2427, "lr": 9.959187371298795e-06}
{"train_loss": 0.061510369181632996, "global_step": 216068, "epoch": 2427, "lr": 9.958840163872701e-06}
{"train_loss": 0.04677395895123482, "global_step": 216069, "epoch": 2427, "lr": 9.958492961829641e-06}
{"train_loss": 0.039648376405239105, "global_step": 216070, "epoch": 2427, "lr": 9.958145765169685e-06}
{"train_loss": 0.052780598402023315, "global_step": 216071, "epoch": 2427, "lr": 9.957798573892856e-06}
{"train_loss": 0.029429538175463676, "global_step": 216072, "epoch": 2427, "lr": 9.957451387999229e-06}
{"train_loss": 0.016607817262411118, "global_step": 216073, "epoch": 2427, "lr": 9.957104207488832e-06}
{"train_loss": 0.05040031671524048, "global_step": 216074, "epoch": 2427, "lr": 9.956757032361697e-06}
{"train_loss": 0.0672505795955658, "global_step": 216075, "epoch": 2427, "lr": 9.956409862617904e-06}
{"train_loss": 0.06756812334060669, "global_step": 216076, "epoch": 2427, "lr": 9.956062698257474e-06}
{"train_loss": 0.05410824716091156, "global_step": 216077, "epoch": 2427, "lr": 9.955715539280475e-06}
{"train_loss": 0.03639715909957886, "global_step": 216078, "epoch": 2427, "lr": 9.955368385686926e-06}
{"train_loss": 0.08071175217628479, "global_step": 216079, "epoch": 2427, "lr": 9.955021237476903e-06}
{"train_loss": 0.0335737019777298, "global_step": 216080, "epoch": 2427, "lr": 9.95467409465043e-06}
{"train_loss": 0.03586190938949585, "global_step": 216081, "epoch": 2427, "lr": 9.954326957207571e-06}
{"train_loss": 0.025439102202653885, "global_step": 216082, "epoch": 2427, "lr": 9.953979825148358e-06}
{"train_loss": 0.07604142278432846, "global_step": 216083, "epoch": 2427, "lr": 9.953632698472848e-06}
{"train_loss": 0.04350372031331062, "global_step": 216084, "epoch": 2427, "lr": 9.953285577181077e-06}
{"train_loss": 0.0476900115609169, "global_step": 216085, "epoch": 2427, "lr": 9.952938461273114e-06}
{"train_loss": 0.044636115431785583, "global_step": 216086, "epoch": 2427, "lr": 9.952591350748975e-06}
{"train_loss": 0.08431892842054367, "global_step": 216087, "epoch": 2427, "lr": 9.952244245608733e-06}
{"train_loss": 0.1152176484465599, "global_step": 216088, "epoch": 2427, "lr": 9.951897145852413e-06}
{"train_loss": 0.09207810461521149, "global_step": 216089, "epoch": 2427, "lr": 9.951550051480085e-06}
{"train_loss": 0.03528397157788277, "global_step": 216090, "epoch": 2427, "lr": 9.951202962491769e-06}
{"train_loss": 0.05570305272769392, "global_step": 216091, "epoch": 2427, "lr": 9.950855878887539e-06, "val_loss": 8.88073444366455}
{"train_loss": 0.03317198157310486, "global_step": 216092, "epoch": 2428, "lr": 9.950508800667413e-06}
{"train_loss": 0.053142886608839035, "global_step": 216093, "epoch": 2428, "lr": 9.95016172783147e-06}
{"train_loss": 0.06495212763547897, "global_step": 216094, "epoch": 2428, "lr": 9.949814660379724e-06}
{"train_loss": 0.03958394005894661, "global_step": 216095, "epoch": 2428, "lr": 9.94946759831224e-06}
{"train_loss": 0.06651870161294937, "global_step": 216096, "epoch": 2428, "lr": 9.94912054162907e-06}
{"train_loss": 0.040098465979099274, "global_step": 216097, "epoch": 2428, "lr": 9.948773490330243e-06}
{"train_loss": 0.07193906605243683, "global_step": 216098, "epoch": 2428, "lr": 9.948426444415831e-06}
{"train_loss": 0.05674164369702339, "global_step": 216099, "epoch": 2428, "lr": 9.948079403885846e-06}
{"train_loss": 0.026034170761704445, "global_step": 216100, "epoch": 2428, "lr": 9.94773236874037e-06}
{"train_loss": 0.08591865748167038, "global_step": 216101, "epoch": 2428, "lr": 9.94738533897942e-06}
{"train_loss": 0.035666029900312424, "global_step": 216102, "epoch": 2428, "lr": 9.947038314603068e-06}
{"train_loss": 0.02791343629360199, "global_step": 216103, "epoch": 2428, "lr": 9.946691295611337e-06}
{"train_loss": 0.03701991215348244, "global_step": 216104, "epoch": 2428, "lr": 9.946344282004294e-06}
{"train_loss": 0.056754469871520996, "global_step": 216105, "epoch": 2428, "lr": 9.945997273781971e-06}
{"train_loss": 0.040062353014945984, "global_step": 216106, "epoch": 2428, "lr": 9.945650270944429e-06}
{"train_loss": 0.06678294390439987, "global_step": 216107, "epoch": 2428, "lr": 9.945303273491691e-06}
{"train_loss": 0.06536547839641571, "global_step": 216108, "epoch": 2428, "lr": 9.944956281423834e-06}
{"train_loss": 0.08255971968173981, "global_step": 216109, "epoch": 2428, "lr": 9.944609294740882e-06}
{"train_loss": 0.06577182561159134, "global_step": 216110, "epoch": 2428, "lr": 9.944262313442892e-06}
{"train_loss": 0.06907274574041367, "global_step": 216111, "epoch": 2428, "lr": 9.943915337529913e-06}
{"train_loss": 0.030704151839017868, "global_step": 216112, "epoch": 2428, "lr": 9.943568367001976e-06}
{"train_loss": 0.07321476936340332, "global_step": 216113, "epoch": 2428, "lr": 9.943221401859148e-06}
{"train_loss": 0.0426233634352684, "global_step": 216114, "epoch": 2428, "lr": 9.942874442101452e-06}
{"train_loss": 0.03977485001087189, "global_step": 216115, "epoch": 2428, "lr": 9.94252748772896e-06}
{"train_loss": 0.04668840393424034, "global_step": 216116, "epoch": 2428, "lr": 9.942180538741697e-06}
{"train_loss": 0.060956791043281555, "global_step": 216117, "epoch": 2428, "lr": 9.941833595139732e-06}
{"train_loss": 0.09917058795690536, "global_step": 216118, "epoch": 2428, "lr": 9.941486656923083e-06}
{"train_loss": 0.01703011617064476, "global_step": 216119, "epoch": 2428, "lr": 9.94113972409183e-06}
{"train_loss": 0.06562788039445877, "global_step": 216120, "epoch": 2428, "lr": 9.940792796645987e-06}
{"train_loss": 0.04467583820223808, "global_step": 216121, "epoch": 2428, "lr": 9.94044587458563e-06}
{"train_loss": 0.1063060313463211, "global_step": 216122, "epoch": 2428, "lr": 9.940098957910781e-06}
{"train_loss": 0.07907980680465698, "global_step": 216123, "epoch": 2428, "lr": 9.93975204662151e-06}
{"train_loss": 0.042037233710289, "global_step": 216124, "epoch": 2428, "lr": 9.939405140717833e-06}
{"train_loss": 0.06841152906417847, "global_step": 216125, "epoch": 2428, "lr": 9.939058240199834e-06}
{"train_loss": 0.06122933328151703, "global_step": 216126, "epoch": 2428, "lr": 9.938711345067525e-06}
{"train_loss": 0.059065427631139755, "global_step": 216127, "epoch": 2428, "lr": 9.938364455320981e-06}
{"train_loss": 0.11425786465406418, "global_step": 216128, "epoch": 2428, "lr": 9.938017570960217e-06}
{"train_loss": 0.03180750459432602, "global_step": 216129, "epoch": 2428, "lr": 9.937670691985317e-06}
{"train_loss": 0.06695699691772461, "global_step": 216130, "epoch": 2428, "lr": 9.9373238183963e-06}
{"train_loss": 0.0625448077917099, "global_step": 216131, "epoch": 2428, "lr": 9.936976950193216e-06}
{"train_loss": 0.051824744790792465, "global_step": 216132, "epoch": 2428, "lr": 9.93663008737613e-06}
{"train_loss": 0.05520797148346901, "global_step": 216133, "epoch": 2428, "lr": 9.936283229945064e-06}
{"train_loss": 0.03909194841980934, "global_step": 216134, "epoch": 2428, "lr": 9.93593637790009e-06}
{"train_loss": 0.04377613589167595, "global_step": 216135, "epoch": 2428, "lr": 9.935589531241224e-06}
{"train_loss": 0.025942973792552948, "global_step": 216136, "epoch": 2428, "lr": 9.935242689968549e-06}
{"train_loss": 0.09605063498020172, "global_step": 216137, "epoch": 2428, "lr": 9.934895854082077e-06}
{"train_loss": 0.05014267936348915, "global_step": 216138, "epoch": 2428, "lr": 9.93454902358188e-06}
{"train_loss": 0.051860056817531586, "global_step": 216139, "epoch": 2428, "lr": 9.934202198467984e-06}
{"train_loss": 0.014777175150811672, "global_step": 216140, "epoch": 2428, "lr": 9.933855378740459e-06}
{"train_loss": 0.0877847895026207, "global_step": 216141, "epoch": 2428, "lr": 9.933508564399324e-06}
{"train_loss": 0.053743332624435425, "global_step": 216142, "epoch": 2428, "lr": 9.933161755444654e-06}
{"train_loss": 0.06532670557498932, "global_step": 216143, "epoch": 2428, "lr": 9.932814951876474e-06}
{"train_loss": 0.04219644144177437, "global_step": 216144, "epoch": 2428, "lr": 9.932468153694846e-06}
{"train_loss": 0.022816741839051247, "global_step": 216145, "epoch": 2428, "lr": 9.932121360899799e-06}
{"train_loss": 0.04846677556633949, "global_step": 216146, "epoch": 2428, "lr": 9.931774573491403e-06}
{"train_loss": 0.03263051435351372, "global_step": 216147, "epoch": 2428, "lr": 9.931427791469678e-06}
{"train_loss": 0.04713946953415871, "global_step": 216148, "epoch": 2428, "lr": 9.931081014834693e-06}
{"train_loss": 0.048425815999507904, "global_step": 216149, "epoch": 2428, "lr": 9.930734243586492e-06}
{"train_loss": 0.045285001397132874, "global_step": 216150, "epoch": 2428, "lr": 9.9303874777251e-06}
{"train_loss": 0.024290921166539192, "global_step": 216151, "epoch": 2428, "lr": 9.930040717250589e-06}
{"train_loss": 0.02159283682703972, "global_step": 216152, "epoch": 2428, "lr": 9.929693962162989e-06}
{"train_loss": 0.06185120716691017, "global_step": 216153, "epoch": 2428, "lr": 9.929347212462359e-06}
{"train_loss": 0.05456848815083504, "global_step": 216154, "epoch": 2428, "lr": 9.929000468148731e-06}
{"train_loss": 0.08904016762971878, "global_step": 216155, "epoch": 2428, "lr": 9.928653729222176e-06}
{"train_loss": 0.11140527576208115, "global_step": 216156, "epoch": 2428, "lr": 9.928306995682707e-06}
{"train_loss": 0.08383731544017792, "global_step": 216157, "epoch": 2428, "lr": 9.927960267530406e-06}
{"train_loss": 0.036866702139377594, "global_step": 216158, "epoch": 2428, "lr": 9.927613544765285e-06}
{"train_loss": 0.05504555255174637, "global_step": 216159, "epoch": 2428, "lr": 9.927266827387427e-06}
{"train_loss": 0.055480048060417175, "global_step": 216160, "epoch": 2428, "lr": 9.926920115396843e-06}
{"train_loss": 0.04892841726541519, "global_step": 216161, "epoch": 2428, "lr": 9.92657340879361e-06}
{"train_loss": 0.0547705739736557, "global_step": 216162, "epoch": 2428, "lr": 9.926226707577745e-06}
{"train_loss": 0.013960957527160645, "global_step": 216163, "epoch": 2428, "lr": 9.925880011749322e-06}
{"train_loss": 0.06039128080010414, "global_step": 216164, "epoch": 2428, "lr": 9.925533321308368e-06}
{"train_loss": 0.052079472690820694, "global_step": 216165, "epoch": 2428, "lr": 9.925186636254952e-06}
{"train_loss": 0.017927926033735275, "global_step": 216166, "epoch": 2428, "lr": 9.924839956589088e-06}
{"train_loss": 0.03314764052629471, "global_step": 216167, "epoch": 2428, "lr": 9.924493282310842e-06}
{"train_loss": 0.05797518789768219, "global_step": 216168, "epoch": 2428, "lr": 9.92414661342027e-06}
{"train_loss": 0.035562239587306976, "global_step": 216169, "epoch": 2428, "lr": 9.923799949917401e-06}
{"train_loss": 0.06484821438789368, "global_step": 216170, "epoch": 2428, "lr": 9.923453291802299e-06}
{"train_loss": 0.01874651201069355, "global_step": 216171, "epoch": 2428, "lr": 9.923106639074986e-06}
{"train_loss": 0.07400042563676834, "global_step": 216172, "epoch": 2428, "lr": 9.922759991735542e-06}
{"train_loss": 0.044212065637111664, "global_step": 216173, "epoch": 2428, "lr": 9.92241334978397e-06}
{"train_loss": 0.04709077998995781, "global_step": 216174, "epoch": 2428, "lr": 9.922066713220363e-06}
{"train_loss": 0.04094192013144493, "global_step": 216175, "epoch": 2428, "lr": 9.921720082044733e-06}
{"train_loss": 0.10365325212478638, "global_step": 216176, "epoch": 2428, "lr": 9.921373456257149e-06}
{"train_loss": 0.029133273288607597, "global_step": 216177, "epoch": 2428, "lr": 9.921026835857633e-06}
{"train_loss": 0.02664833515882492, "global_step": 216178, "epoch": 2428, "lr": 9.920680220846262e-06}
{"train_loss": 0.03915015235543251, "global_step": 216179, "epoch": 2428, "lr": 9.920333611223053e-06}
{"train_loss": 0.05345666346799457, "global_step": 216180, "epoch": 2428, "lr": 9.91998700698808e-06, "val_loss": 9.002823829650879}
{"train_loss": 0.034141622483730316, "global_step": 216181, "epoch": 2429, "lr": 9.919640408141367e-06}
{"train_loss": 0.0583227276802063, "global_step": 216182, "epoch": 2429, "lr": 9.919293814682978e-06}
{"train_loss": 0.08181153237819672, "global_step": 216183, "epoch": 2429, "lr": 9.918947226612941e-06}
{"train_loss": 0.08911547809839249, "global_step": 216184, "epoch": 2429, "lr": 9.918600643931325e-06}
{"train_loss": 0.016512243077158928, "global_step": 216185, "epoch": 2429, "lr": 9.918254066638149e-06}
{"train_loss": 0.03978633135557175, "global_step": 216186, "epoch": 2429, "lr": 9.917907494733492e-06}
{"train_loss": 0.03608192503452301, "global_step": 216187, "epoch": 2429, "lr": 9.917560928217368e-06}
{"train_loss": 0.03579090163111687, "global_step": 216188, "epoch": 2429, "lr": 9.917214367089855e-06}
{"train_loss": 0.02678004652261734, "global_step": 216189, "epoch": 2429, "lr": 9.916867811350983e-06}
{"train_loss": 0.046242810785770416, "global_step": 216190, "epoch": 2429, "lr": 9.916521261000788e-06}
{"train_loss": 0.03921671211719513, "global_step": 216191, "epoch": 2429, "lr": 9.91617471603934e-06}
{"train_loss": 0.06503446400165558, "global_step": 216192, "epoch": 2429, "lr": 9.915828176466657e-06}
{"train_loss": 0.10310595482587814, "global_step": 216193, "epoch": 2429, "lr": 9.91548164228282e-06}
{"train_loss": 0.06429104506969452, "global_step": 216194, "epoch": 2429, "lr": 9.915135113487844e-06}
{"train_loss": 0.01891779713332653, "global_step": 216195, "epoch": 2429, "lr": 9.914788590081802e-06}
{"train_loss": 0.05338849127292633, "global_step": 216196, "epoch": 2429, "lr": 9.914442072064717e-06}
{"train_loss": 0.024218305945396423, "global_step": 216197, "epoch": 2429, "lr": 9.91409555943666e-06}
{"train_loss": 0.07668139785528183, "global_step": 216198, "epoch": 2429, "lr": 9.913749052197647e-06}
{"train_loss": 0.030590271577239037, "global_step": 216199, "epoch": 2429, "lr": 9.913402550347756e-06}
{"train_loss": 0.06649397313594818, "global_step": 216200, "epoch": 2429, "lr": 9.91305605388701e-06}
{"train_loss": 0.041341084986925125, "global_step": 216201, "epoch": 2429, "lr": 9.912709562815476e-06}
{"train_loss": 0.042899392545223236, "global_step": 216202, "epoch": 2429, "lr": 9.912363077133174e-06}
{"train_loss": 0.05124203488230705, "global_step": 216203, "epoch": 2429, "lr": 9.912016596840185e-06}
{"train_loss": 0.07381194084882736, "global_step": 216204, "epoch": 2429, "lr": 9.911670121936517e-06}
{"train_loss": 0.047188837081193924, "global_step": 216205, "epoch": 2429, "lr": 9.91132365242224e-06}
{"train_loss": 0.0777694433927536, "global_step": 216206, "epoch": 2429, "lr": 9.91097718829741e-06}
{"train_loss": 0.06005116552114487, "global_step": 216207, "epoch": 2429, "lr": 9.910630729562047e-06}
{"train_loss": 0.05879302695393562, "global_step": 216208, "epoch": 2429, "lr": 9.910284276216225e-06}
{"train_loss": 0.06425222009420395, "global_step": 216209, "epoch": 2429, "lr": 9.909937828259964e-06}
{"train_loss": 0.01675744168460369, "global_step": 216210, "epoch": 2429, "lr": 9.909591385693335e-06}
{"train_loss": 0.04726520553231239, "global_step": 216211, "epoch": 2429, "lr": 9.909244948516355e-06}
{"train_loss": 0.05564567446708679, "global_step": 216212, "epoch": 2429, "lr": 9.90889851672911e-06}
{"train_loss": 0.05631828308105469, "global_step": 216213, "epoch": 2429, "lr": 9.908552090331607e-06}
{"train_loss": 0.017730506137013435, "global_step": 216214, "epoch": 2429, "lr": 9.908205669323922e-06}
{"train_loss": 0.07050618529319763, "global_step": 216215, "epoch": 2429, "lr": 9.907859253706082e-06}
{"train_loss": 0.02962314337491989, "global_step": 216216, "epoch": 2429, "lr": 9.90751284347815e-06}
{"train_loss": 0.06652096658945084, "global_step": 216217, "epoch": 2429, "lr": 9.907166438640153e-06}
{"train_loss": 0.029365049675107002, "global_step": 216218, "epoch": 2429, "lr": 9.906820039192166e-06}
{"train_loss": 0.04280529543757439, "global_step": 216219, "epoch": 2429, "lr": 9.906473645134201e-06}
{"train_loss": 0.06842508912086487, "global_step": 216220, "epoch": 2429, "lr": 9.906127256466335e-06}
{"train_loss": 0.05294012278318405, "global_step": 216221, "epoch": 2429, "lr": 9.90578087318859e-06}
{"train_loss": 0.06513417512178421, "global_step": 216222, "epoch": 2429, "lr": 9.905434495301035e-06}
{"train_loss": 0.03951910138130188, "global_step": 216223, "epoch": 2429, "lr": 9.905088122803696e-06}
{"train_loss": 0.04756490886211395, "global_step": 216224, "epoch": 2429, "lr": 9.90474175569664e-06}
{"train_loss": 0.10178111493587494, "global_step": 216225, "epoch": 2429, "lr": 9.904395393979887e-06}
{"train_loss": 0.08915978670120239, "global_step": 216226, "epoch": 2429, "lr": 9.904049037653518e-06}
{"train_loss": 0.01478084921836853, "global_step": 216227, "epoch": 2429, "lr": 9.90370268671756e-06}
{"train_loss": 0.07186259329319, "global_step": 216228, "epoch": 2429, "lr": 9.903356341172042e-06}
{"train_loss": 0.0674799382686615, "global_step": 216229, "epoch": 2429, "lr": 9.903010001017048e-06}
{"train_loss": 0.0479133240878582, "global_step": 216230, "epoch": 2429, "lr": 9.902663666252587e-06}
{"train_loss": 0.13750027120113373, "global_step": 216231, "epoch": 2429, "lr": 9.902317336878742e-06}
{"train_loss": 0.051581647247076035, "global_step": 216232, "epoch": 2429, "lr": 9.901971012895528e-06}
{"train_loss": 0.05720246210694313, "global_step": 216233, "epoch": 2429, "lr": 9.90162469430302e-06}
{"train_loss": 0.1171988695859909, "global_step": 216234, "epoch": 2429, "lr": 9.90127838110123e-06}
{"train_loss": 0.039310384541749954, "global_step": 216235, "epoch": 2429, "lr": 9.900932073290242e-06}
{"train_loss": 0.05893754959106445, "global_step": 216236, "epoch": 2429, "lr": 9.900585770870074e-06}
{"train_loss": 0.042523905634880066, "global_step": 216237, "epoch": 2429, "lr": 9.900239473840794e-06}
{"train_loss": 0.10475742071866989, "global_step": 216238, "epoch": 2429, "lr": 9.899893182202424e-06}
{"train_loss": 0.027672918513417244, "global_step": 216239, "epoch": 2429, "lr": 9.899546895955043e-06}
{"train_loss": 0.011336023919284344, "global_step": 216240, "epoch": 2429, "lr": 9.899200615098658e-06}
{"train_loss": 0.10538414120674133, "global_step": 216241, "epoch": 2429, "lr": 9.89885433963334e-06}
{"train_loss": 0.03327807039022446, "global_step": 216242, "epoch": 2429, "lr": 9.898508069559147e-06}
{"train_loss": 0.06498216837644577, "global_step": 216243, "epoch": 2429, "lr": 9.898161804876093e-06}
{"train_loss": 0.04442107304930687, "global_step": 216244, "epoch": 2429, "lr": 9.89781554558426e-06}
{"train_loss": 0.03364177793264389, "global_step": 216245, "epoch": 2429, "lr": 9.897469291683664e-06}
{"train_loss": 0.044703416526317596, "global_step": 216246, "epoch": 2429, "lr": 9.897123043174377e-06}
{"train_loss": 0.039924558252096176, "global_step": 216247, "epoch": 2429, "lr": 9.896776800056418e-06}
{"train_loss": 0.04991267994046211, "global_step": 216248, "epoch": 2429, "lr": 9.896430562329867e-06}
{"train_loss": 0.061868298798799515, "global_step": 216249, "epoch": 2429, "lr": 9.896084329994732e-06}
{"train_loss": 0.024667847901582718, "global_step": 216250, "epoch": 2429, "lr": 9.8957381030511e-06}
{"train_loss": 0.05347028374671936, "global_step": 216251, "epoch": 2429, "lr": 9.895391881498978e-06}
{"train_loss": 0.07932321727275848, "global_step": 216252, "epoch": 2429, "lr": 9.895045665338448e-06}
{"train_loss": 0.013984712772071362, "global_step": 216253, "epoch": 2429, "lr": 9.894699454569522e-06}
{"train_loss": 0.05378475785255432, "global_step": 216254, "epoch": 2429, "lr": 9.894353249192285e-06}
{"train_loss": 0.056204654276371, "global_step": 216255, "epoch": 2429, "lr": 9.894007049206749e-06}
{"train_loss": 0.02968813292682171, "global_step": 216256, "epoch": 2429, "lr": 9.893660854612985e-06}
{"train_loss": 0.029514899477362633, "global_step": 216257, "epoch": 2429, "lr": 9.893314665411018e-06}
{"train_loss": 0.07919102907180786, "global_step": 216258, "epoch": 2429, "lr": 9.892968481600907e-06}
{"train_loss": 0.04565242677927017, "global_step": 216259, "epoch": 2429, "lr": 9.892622303182708e-06}
{"train_loss": 0.02404729835689068, "global_step": 216260, "epoch": 2429, "lr": 9.892276130156458e-06}
{"train_loss": 0.08576074242591858, "global_step": 216261, "epoch": 2429, "lr": 9.89192996252219e-06}
{"train_loss": 0.06957297027111053, "global_step": 216262, "epoch": 2429, "lr": 9.891583800279975e-06}
{"train_loss": 0.03758646547794342, "global_step": 216263, "epoch": 2429, "lr": 9.891237643429834e-06}
{"train_loss": 0.057079385966062546, "global_step": 216264, "epoch": 2429, "lr": 9.890891491971843e-06}
{"train_loss": 0.06464167684316635, "global_step": 216265, "epoch": 2429, "lr": 9.890545345906032e-06}
{"train_loss": 0.06285549700260162, "global_step": 216266, "epoch": 2429, "lr": 9.890199205232437e-06}
{"train_loss": 0.024749692529439926, "global_step": 216267, "epoch": 2429, "lr": 9.889853069951126e-06}
{"train_loss": 0.100067138671875, "global_step": 216268, "epoch": 2429, "lr": 9.889506940062126e-06}
{"train_loss": 0.05395885952486751, "global_step": 216269, "epoch": 2429, "lr": 9.889160815565501e-06, "val_loss": 8.805206298828125}
{"train_loss": 0.041333332657814026, "global_step": 216270, "epoch": 2430, "lr": 9.88881469646128e-06}
{"train_loss": 0.07013609260320663, "global_step": 216271, "epoch": 2430, "lr": 9.888468582749533e-06}
{"train_loss": 0.02627212181687355, "global_step": 216272, "epoch": 2430, "lr": 9.888122474430272e-06}
{"train_loss": 0.05680389329791069, "global_step": 216273, "epoch": 2430, "lr": 9.887776371503587e-06}
{"train_loss": 0.05893813818693161, "global_step": 216274, "epoch": 2430, "lr": 9.887430273969483e-06}
{"train_loss": 0.03352492302656174, "global_step": 216275, "epoch": 2430, "lr": 9.887084181828038e-06}
{"train_loss": 0.03980008885264397, "global_step": 216276, "epoch": 2430, "lr": 9.886738095079272e-06}
{"train_loss": 0.05903053283691406, "global_step": 216277, "epoch": 2430, "lr": 9.88639201372325e-06}
{"train_loss": 0.06836120784282684, "global_step": 216278, "epoch": 2430, "lr": 9.886045937760025e-06}
{"train_loss": 0.018009137362241745, "global_step": 216279, "epoch": 2430, "lr": 9.885699867189612e-06}
{"train_loss": 0.043815284967422485, "global_step": 216280, "epoch": 2430, "lr": 9.8853538020121e-06}
{"train_loss": 0.0591520220041275, "global_step": 216281, "epoch": 2430, "lr": 9.885007742227498e-06}
{"train_loss": 0.07339905947446823, "global_step": 216282, "epoch": 2430, "lr": 9.884661687835878e-06}
{"train_loss": 0.10244226455688477, "global_step": 216283, "epoch": 2430, "lr": 9.884315638837266e-06}
{"train_loss": 0.08570980280637741, "global_step": 216284, "epoch": 2430, "lr": 9.883969595231728e-06}
{"train_loss": 0.04371120408177376, "global_step": 216285, "epoch": 2430, "lr": 9.883623557019289e-06}
{"train_loss": 0.02806241065263748, "global_step": 216286, "epoch": 2430, "lr": 9.883277524200024e-06}
{"train_loss": 0.06478717923164368, "global_step": 216287, "epoch": 2430, "lr": 9.88293149677395e-06}
{"train_loss": 0.01516819093376398, "global_step": 216288, "epoch": 2430, "lr": 9.88258547474114e-06}
{"train_loss": 0.01768563501536846, "global_step": 216289, "epoch": 2430, "lr": 9.882239458101612e-06}
{"train_loss": 0.036123160272836685, "global_step": 216290, "epoch": 2430, "lr": 9.881893446855445e-06}
{"train_loss": 0.030756672844290733, "global_step": 216291, "epoch": 2430, "lr": 9.881547441002653e-06}
{"train_loss": 0.06346835941076279, "global_step": 216292, "epoch": 2430, "lr": 9.881201440543309e-06}
{"train_loss": 0.059591736644506454, "global_step": 216293, "epoch": 2430, "lr": 9.880855445477438e-06}
{"train_loss": 0.039247073233127594, "global_step": 216294, "epoch": 2430, "lr": 9.880509455805098e-06}
{"train_loss": 0.0587305948138237, "global_step": 216295, "epoch": 2430, "lr": 9.880163471526343e-06}
{"train_loss": 0.04130661115050316, "global_step": 216296, "epoch": 2430, "lr": 9.879817492641202e-06}
{"train_loss": 0.05111983045935631, "global_step": 216297, "epoch": 2430, "lr": 9.87947151914974e-06}
{"train_loss": 0.11683566868305206, "global_step": 216298, "epoch": 2430, "lr": 9.879125551051988e-06}
{"train_loss": 0.05257077515125275, "global_step": 216299, "epoch": 2430, "lr": 9.878779588348003e-06}
{"train_loss": 0.059360530227422714, "global_step": 216300, "epoch": 2430, "lr": 9.878433631037832e-06}
{"train_loss": 0.03950870782136917, "global_step": 216301, "epoch": 2430, "lr": 9.8780876791215e-06}
{"train_loss": 0.06482941657304764, "global_step": 216302, "epoch": 2430, "lr": 9.877741732599088e-06}
{"train_loss": 0.03363910689949989, "global_step": 216303, "epoch": 2430, "lr": 9.877395791470613e-06}
{"train_loss": 0.022533148527145386, "global_step": 216304, "epoch": 2430, "lr": 9.877049855736143e-06}
{"train_loss": 0.043973006308078766, "global_step": 216305, "epoch": 2430, "lr": 9.876703925395714e-06}
{"train_loss": 0.05916093662381172, "global_step": 216306, "epoch": 2430, "lr": 9.876358000449359e-06}
{"train_loss": 0.08635903149843216, "global_step": 216307, "epoch": 2430, "lr": 9.876012080897152e-06}
{"train_loss": 0.04344097152352333, "global_step": 216308, "epoch": 2430, "lr": 9.875666166739111e-06}
{"train_loss": 0.03488263487815857, "global_step": 216309, "epoch": 2430, "lr": 9.875320257975317e-06}
{"train_loss": 0.03533181920647621, "global_step": 216310, "epoch": 2430, "lr": 9.874974354605781e-06}
{"train_loss": 0.055821191519498825, "global_step": 216311, "epoch": 2430, "lr": 9.874628456630575e-06}
{"train_loss": 0.026118144392967224, "global_step": 216312, "epoch": 2430, "lr": 9.874282564049731e-06}
{"train_loss": 0.032152172178030014, "global_step": 216313, "epoch": 2430, "lr": 9.873936676863294e-06}
{"train_loss": 0.05763518437743187, "global_step": 216314, "epoch": 2430, "lr": 9.87359079507133e-06}
{"train_loss": 0.010245764628052711, "global_step": 216315, "epoch": 2430, "lr": 9.873244918673863e-06}
{"train_loss": 0.026745913550257683, "global_step": 216316, "epoch": 2430, "lr": 9.872899047670964e-06}
{"train_loss": 0.04906785488128662, "global_step": 216317, "epoch": 2430, "lr": 9.87255318206265e-06}
{"train_loss": 0.0438104011118412, "global_step": 216318, "epoch": 2430, "lr": 9.872207321848992e-06}
{"train_loss": 0.045333653688430786, "global_step": 216319, "epoch": 2430, "lr": 9.871861467030014e-06}
{"train_loss": 0.054798685014247894, "global_step": 216320, "epoch": 2430, "lr": 9.871515617605793e-06}
{"train_loss": 0.11087609827518463, "global_step": 216321, "epoch": 2430, "lr": 9.871169773576339e-06}
{"train_loss": 0.08933039009571075, "global_step": 216322, "epoch": 2430, "lr": 9.870823934941731e-06}
{"train_loss": 0.015573467127978802, "global_step": 216323, "epoch": 2430, "lr": 9.87047810170199e-06}
{"train_loss": 0.06340006738901138, "global_step": 216324, "epoch": 2430, "lr": 9.870132273857185e-06}
{"train_loss": 0.052718643099069595, "global_step": 216325, "epoch": 2430, "lr": 9.869786451407342e-06}
{"train_loss": 0.05241304263472557, "global_step": 216326, "epoch": 2430, "lr": 9.869440634352529e-06}
{"train_loss": 0.07130510360002518, "global_step": 216327, "epoch": 2430, "lr": 9.86909482269277e-06}
{"train_loss": 0.0509457066655159, "global_step": 216328, "epoch": 2430, "lr": 9.868749016428131e-06}
{"train_loss": 0.0551350936293602, "global_step": 216329, "epoch": 2430, "lr": 9.868403215558637e-06}
{"train_loss": 0.06481727957725525, "global_step": 216330, "epoch": 2430, "lr": 9.868057420084364e-06}
{"train_loss": 0.030777376145124435, "global_step": 216331, "epoch": 2430, "lr": 9.867711630005322e-06}
{"train_loss": 0.0623343326151371, "global_step": 216332, "epoch": 2430, "lr": 9.867365845321586e-06}
{"train_loss": 0.07083014398813248, "global_step": 216333, "epoch": 2430, "lr": 9.8670200660332e-06}
{"train_loss": 0.06746145337820053, "global_step": 216334, "epoch": 2430, "lr": 9.866674292140188e-06}
{"train_loss": 0.03699808195233345, "global_step": 216335, "epoch": 2430, "lr": 9.866328523642631e-06}
{"train_loss": 0.018211986869573593, "global_step": 216336, "epoch": 2430, "lr": 9.865982760540544e-06}
{"train_loss": 0.04604584723711014, "global_step": 216337, "epoch": 2430, "lr": 9.865637002834e-06}
{"train_loss": 0.06057796999812126, "global_step": 216338, "epoch": 2430, "lr": 9.865291250523029e-06}
{"train_loss": 0.07798732817173004, "global_step": 216339, "epoch": 2430, "lr": 9.86494550360767e-06}
{"train_loss": 0.06168190762400627, "global_step": 216340, "epoch": 2430, "lr": 9.86459976208799e-06}
{"train_loss": 0.024865172803401947, "global_step": 216341, "epoch": 2430, "lr": 9.864254025964015e-06}
{"train_loss": 0.06549142301082611, "global_step": 216342, "epoch": 2430, "lr": 9.863908295235813e-06}
{"train_loss": 0.059900104999542236, "global_step": 216343, "epoch": 2430, "lr": 9.863562569903417e-06}
{"train_loss": 0.051759395748376846, "global_step": 216344, "epoch": 2430, "lr": 9.863216849966867e-06}
{"train_loss": 0.029964959248900414, "global_step": 216345, "epoch": 2430, "lr": 9.862871135426232e-06}
{"train_loss": 0.05304018780589104, "global_step": 216346, "epoch": 2430, "lr": 9.862525426281527e-06}
{"train_loss": 0.04067147523164749, "global_step": 216347, "epoch": 2430, "lr": 9.862179722532832e-06}
{"train_loss": 0.049941349774599075, "global_step": 216348, "epoch": 2430, "lr": 9.861834024180167e-06}
{"train_loss": 0.07018709182739258, "global_step": 216349, "epoch": 2430, "lr": 9.861488331223583e-06}
{"train_loss": 0.07352783530950546, "global_step": 216350, "epoch": 2430, "lr": 9.861142643663152e-06}
{"train_loss": 0.03853773698210716, "global_step": 216351, "epoch": 2430, "lr": 9.860796961498885e-06}
{"train_loss": 0.08237684518098831, "global_step": 216352, "epoch": 2430, "lr": 9.86045128473086e-06}
{"train_loss": 0.055316533893346786, "global_step": 216353, "epoch": 2430, "lr": 9.860105613359094e-06}
{"train_loss": 0.045524630695581436, "global_step": 216354, "epoch": 2430, "lr": 9.859759947383657e-06}
{"train_loss": 0.045262813568115234, "global_step": 216355, "epoch": 2430, "lr": 9.859414286804581e-06}
{"train_loss": 0.04472043365240097, "global_step": 216356, "epoch": 2430, "lr": 9.85906863162192e-06}
{"train_loss": 0.056239914149045944, "global_step": 216357, "epoch": 2430, "lr": 9.858722981835716e-06}
{"train_loss": 0.051576219603754164, "global_step": 216358, "epoch": 2430, "lr": 9.858377337446023e-06, "val_loss": 8.933089256286621, "train_action_mse_error": 20.985445022583008}
{"train_loss": 0.07642930001020432, "global_step": 216359, "epoch": 2431, "lr": 9.85803169845287e-06}
{"train_loss": 0.07168348878622055, "global_step": 216360, "epoch": 2431, "lr": 9.85768606485633e-06}
{"train_loss": 0.08371253311634064, "global_step": 216361, "epoch": 2431, "lr": 9.85734043665642e-06}
{"train_loss": 0.07450143992900848, "global_step": 216362, "epoch": 2431, "lr": 9.856994813853216e-06}
{"train_loss": 0.06314125657081604, "global_step": 216363, "epoch": 2431, "lr": 9.856649196446738e-06}
{"train_loss": 0.09711313247680664, "global_step": 216364, "epoch": 2431, "lr": 9.856303584437054e-06}
{"train_loss": 0.0627136081457138, "global_step": 216365, "epoch": 2431, "lr": 9.855957977824186e-06}
{"train_loss": 0.018217917531728745, "global_step": 216366, "epoch": 2431, "lr": 9.855612376608215e-06}
{"train_loss": 0.052318695932626724, "global_step": 216367, "epoch": 2431, "lr": 9.855266780789152e-06}
{"train_loss": 0.06275942176580429, "global_step": 216368, "epoch": 2431, "lr": 9.854921190367055e-06}
{"train_loss": 0.02530217170715332, "global_step": 216369, "epoch": 2431, "lr": 9.854575605341993e-06}
{"train_loss": 0.06487035006284714, "global_step": 216370, "epoch": 2431, "lr": 9.854230025713979e-06}
{"train_loss": 0.06471502035856247, "global_step": 216371, "epoch": 2431, "lr": 9.853884451483086e-06}
{"train_loss": 0.04009527340531349, "global_step": 216372, "epoch": 2431, "lr": 9.853538882649338e-06}
{"train_loss": 0.08970458805561066, "global_step": 216373, "epoch": 2431, "lr": 9.853193319212806e-06}
{"train_loss": 0.07068802416324615, "global_step": 216374, "epoch": 2431, "lr": 9.852847761173505e-06}
{"train_loss": 0.052271462976932526, "global_step": 216375, "epoch": 2431, "lr": 9.852502208531517e-06}
{"train_loss": 0.040358420461416245, "global_step": 216376, "epoch": 2431, "lr": 9.85215666128687e-06}
{"train_loss": 0.041775938123464584, "global_step": 216377, "epoch": 2431, "lr": 9.851811119439597e-06}
{"train_loss": 0.06366613507270813, "global_step": 216378, "epoch": 2431, "lr": 9.851465582989772e-06}
{"train_loss": 0.07180774956941605, "global_step": 216379, "epoch": 2431, "lr": 9.851120051937412e-06}
{"train_loss": 0.1192937046289444, "global_step": 216380, "epoch": 2431, "lr": 9.850774526282596e-06}
{"train_loss": 0.05485859513282776, "global_step": 216381, "epoch": 2431, "lr": 9.85042900602534e-06}
{"train_loss": 0.04341316595673561, "global_step": 216382, "epoch": 2431, "lr": 9.850083491165723e-06}
{"train_loss": 0.021718887612223625, "global_step": 216383, "epoch": 2431, "lr": 9.849737981703765e-06}
{"train_loss": 0.0622580461204052, "global_step": 216384, "epoch": 2431, "lr": 9.849392477639513e-06}
{"train_loss": 0.12027188390493393, "global_step": 216385, "epoch": 2431, "lr": 9.84904697897302e-06}
{"train_loss": 0.045001450926065445, "global_step": 216386, "epoch": 2431, "lr": 9.848701485704342e-06}
{"train_loss": 0.10403086990118027, "global_step": 216387, "epoch": 2431, "lr": 9.848355997833508e-06}
{"train_loss": 0.05117162689566612, "global_step": 216388, "epoch": 2431, "lr": 9.848010515360583e-06}
{"train_loss": 0.0432555191218853, "global_step": 216389, "epoch": 2431, "lr": 9.847665038285591e-06}
{"train_loss": 0.042794421315193176, "global_step": 216390, "epoch": 2431, "lr": 9.847319566608604e-06}
{"train_loss": 0.04538765922188759, "global_step": 216391, "epoch": 2431, "lr": 9.846974100329648e-06}
{"train_loss": 0.05852629244327545, "global_step": 216392, "epoch": 2431, "lr": 9.846628639448785e-06}
{"train_loss": 0.05513733625411987, "global_step": 216393, "epoch": 2431, "lr": 9.846283183966038e-06}
{"train_loss": 0.03854190185666084, "global_step": 216394, "epoch": 2431, "lr": 9.845937733881488e-06}
{"train_loss": 0.07839169353246689, "global_step": 216395, "epoch": 2431, "lr": 9.845592289195143e-06}
{"train_loss": 0.06046643480658531, "global_step": 216396, "epoch": 2431, "lr": 9.845246849907085e-06}
{"train_loss": 0.053953055292367935, "global_step": 216397, "epoch": 2431, "lr": 9.844901416017332e-06}
{"train_loss": 0.016499962657690048, "global_step": 216398, "epoch": 2431, "lr": 9.844555987525954e-06}
{"train_loss": 0.07686500996351242, "global_step": 216399, "epoch": 2431, "lr": 9.844210564432976e-06}
{"train_loss": 0.06309235095977783, "global_step": 216400, "epoch": 2431, "lr": 9.843865146738463e-06}
{"train_loss": 0.08775050193071365, "global_step": 216401, "epoch": 2431, "lr": 9.843519734442441e-06}
{"train_loss": 0.03641144931316376, "global_step": 216402, "epoch": 2431, "lr": 9.843174327544986e-06}
{"train_loss": 0.04237119480967522, "global_step": 216403, "epoch": 2431, "lr": 9.842828926046105e-06}
{"train_loss": 0.01727648451924324, "global_step": 216404, "epoch": 2431, "lr": 9.842483529945873e-06}
{"train_loss": 0.05027817189693451, "global_step": 216405, "epoch": 2431, "lr": 9.842138139244338e-06}
{"train_loss": 0.06749165803194046, "global_step": 216406, "epoch": 2431, "lr": 9.84179275394153e-06}
{"train_loss": 0.04586086794734001, "global_step": 216407, "epoch": 2431, "lr": 9.841447374037516e-06}
{"train_loss": 0.04806794971227646, "global_step": 216408, "epoch": 2431, "lr": 9.841101999532314e-06}
{"train_loss": 0.04894724488258362, "global_step": 216409, "epoch": 2431, "lr": 9.840756630426001e-06}
{"train_loss": 0.09792857617139816, "global_step": 216410, "epoch": 2431, "lr": 9.840411266718596e-06}
{"train_loss": 0.026868097484111786, "global_step": 216411, "epoch": 2431, "lr": 9.840065908410168e-06}
{"train_loss": 0.09603036195039749, "global_step": 216412, "epoch": 2431, "lr": 9.839720555500747e-06}
{"train_loss": 0.02172040566802025, "global_step": 216413, "epoch": 2431, "lr": 9.839375207990393e-06}
{"train_loss": 0.0357871912419796, "global_step": 216414, "epoch": 2431, "lr": 9.839029865879135e-06}
{"train_loss": 0.056726712733507156, "global_step": 216415, "epoch": 2431, "lr": 9.838684529167041e-06}
{"train_loss": 0.03157522529363632, "global_step": 216416, "epoch": 2431, "lr": 9.838339197854152e-06}
{"train_loss": 0.07564812153577805, "global_step": 216417, "epoch": 2431, "lr": 9.837993871940494e-06}
{"train_loss": 0.02989649400115013, "global_step": 216418, "epoch": 2431, "lr": 9.837648551426138e-06}
{"train_loss": 0.04621962085366249, "global_step": 216419, "epoch": 2431, "lr": 9.837303236311108e-06}
{"train_loss": 0.0763883963227272, "global_step": 216420, "epoch": 2431, "lr": 9.836957926595475e-06}
{"train_loss": 0.043417979031801224, "global_step": 216421, "epoch": 2431, "lr": 9.836612622279263e-06}
{"train_loss": 0.07611808180809021, "global_step": 216422, "epoch": 2431, "lr": 9.836267323362541e-06}
{"train_loss": 0.03822259232401848, "global_step": 216423, "epoch": 2431, "lr": 9.83592202984533e-06}
{"train_loss": 0.02977747656404972, "global_step": 216424, "epoch": 2431, "lr": 9.835576741727703e-06}
{"train_loss": 0.06188758835196495, "global_step": 216425, "epoch": 2431, "lr": 9.83523145900968e-06}
{"train_loss": 0.07296054065227509, "global_step": 216426, "epoch": 2431, "lr": 9.834886181691338e-06}
{"train_loss": 0.09896095097064972, "global_step": 216427, "epoch": 2431, "lr": 9.834540909772688e-06}
{"train_loss": 0.01930714026093483, "global_step": 216428, "epoch": 2431, "lr": 9.834195643253808e-06}
{"train_loss": 0.030514776706695557, "global_step": 216429, "epoch": 2431, "lr": 9.833850382134719e-06}
{"train_loss": 0.04375670477747917, "global_step": 216430, "epoch": 2431, "lr": 9.833505126415494e-06}
{"train_loss": 0.042064301669597626, "global_step": 216431, "epoch": 2431, "lr": 9.833159876096149e-06}
{"train_loss": 0.06588833034038544, "global_step": 216432, "epoch": 2431, "lr": 9.832814631176757e-06}
{"train_loss": 0.07971017062664032, "global_step": 216433, "epoch": 2431, "lr": 9.832469391657345e-06}
{"train_loss": 0.11031060665845871, "global_step": 216434, "epoch": 2431, "lr": 9.83212415753798e-06}
{"train_loss": 0.10100124031305313, "global_step": 216435, "epoch": 2431, "lr": 9.83177892881868e-06}
{"train_loss": 0.03879985958337784, "global_step": 216436, "epoch": 2431, "lr": 9.831433705499521e-06}
{"train_loss": 0.026978865265846252, "global_step": 216437, "epoch": 2431, "lr": 9.831088487580525e-06}
{"train_loss": 0.07886108756065369, "global_step": 216438, "epoch": 2431, "lr": 9.830743275061766e-06}
{"train_loss": 0.02260989509522915, "global_step": 216439, "epoch": 2431, "lr": 9.830398067943258e-06}
{"train_loss": 0.029842102900147438, "global_step": 216440, "epoch": 2431, "lr": 9.830052866225065e-06}
{"train_loss": 0.06594303995370865, "global_step": 216441, "epoch": 2431, "lr": 9.829707669907239e-06}
{"train_loss": 0.026046177372336388, "global_step": 216442, "epoch": 2431, "lr": 9.829362478989812e-06}
{"train_loss": 0.06102616339921951, "global_step": 216443, "epoch": 2431, "lr": 9.829017293472847e-06}
{"train_loss": 0.07004896551370621, "global_step": 216444, "epoch": 2431, "lr": 9.82867211335637e-06}
{"train_loss": 0.05862521007657051, "global_step": 216445, "epoch": 2431, "lr": 9.828326938640453e-06}
{"train_loss": 0.0322042815387249, "global_step": 216446, "epoch": 2431, "lr": 9.827981769325112e-06}
{"train_loss": 0.05697643537116184, "global_step": 216447, "epoch": 2431, "lr": 9.827636605410423e-06, "val_loss": 8.904874801635742}
{"train_loss": 0.0469338521361351, "global_step": 216448, "epoch": 2432, "lr": 9.827291446896403e-06}
{"train_loss": 0.03950153663754463, "global_step": 216449, "epoch": 2432, "lr": 9.826946293783135e-06}
{"train_loss": 0.03296440467238426, "global_step": 216450, "epoch": 2432, "lr": 9.826601146070625e-06}
{"train_loss": 0.03333228826522827, "global_step": 216451, "epoch": 2432, "lr": 9.826256003758948e-06}
{"train_loss": 0.05335274338722229, "global_step": 216452, "epoch": 2432, "lr": 9.825910866848137e-06}
{"train_loss": 0.053681302815675735, "global_step": 216453, "epoch": 2432, "lr": 9.825565735338249e-06}
{"train_loss": 0.0397682823240757, "global_step": 216454, "epoch": 2432, "lr": 9.82522060922933e-06}
{"train_loss": 0.08695465326309204, "global_step": 216455, "epoch": 2432, "lr": 9.824875488521402e-06}
{"train_loss": 0.05458280071616173, "global_step": 216456, "epoch": 2432, "lr": 9.824530373214547e-06}
{"train_loss": 0.06852488964796066, "global_step": 216457, "epoch": 2432, "lr": 9.824185263308778e-06}
{"train_loss": 0.027446629479527473, "global_step": 216458, "epoch": 2432, "lr": 9.823840158804177e-06}
{"train_loss": 0.04853987321257591, "global_step": 216459, "epoch": 2432, "lr": 9.823495059700749e-06}
{"train_loss": 0.1051710695028305, "global_step": 216460, "epoch": 2432, "lr": 9.823149965998579e-06}
{"train_loss": 0.051861200481653214, "global_step": 216461, "epoch": 2432, "lr": 9.822804877697684e-06}
{"train_loss": 0.0764898732304573, "global_step": 216462, "epoch": 2432, "lr": 9.822459794798139e-06}
{"train_loss": 0.03158644959330559, "global_step": 216463, "epoch": 2432, "lr": 9.822114717299957e-06}
{"train_loss": 0.07007483392953873, "global_step": 216464, "epoch": 2432, "lr": 9.821769645203216e-06}
{"train_loss": 0.04732438549399376, "global_step": 216465, "epoch": 2432, "lr": 9.821424578507937e-06}
{"train_loss": 0.050532907247543335, "global_step": 216466, "epoch": 2432, "lr": 9.821079517214193e-06}
{"train_loss": 0.03932223096489906, "global_step": 216467, "epoch": 2432, "lr": 9.820734461321996e-06}
{"train_loss": 0.03422055393457413, "global_step": 216468, "epoch": 2432, "lr": 9.820389410831432e-06}
{"train_loss": 0.05457872897386551, "global_step": 216469, "epoch": 2432, "lr": 9.820044365742509e-06}
{"train_loss": 0.024367745965719223, "global_step": 216470, "epoch": 2432, "lr": 9.819699326055303e-06}
{"train_loss": 0.05483505129814148, "global_step": 216471, "epoch": 2432, "lr": 9.819354291769838e-06}
{"train_loss": 0.04530656337738037, "global_step": 216472, "epoch": 2432, "lr": 9.819009262886186e-06}
{"train_loss": 0.03432239592075348, "global_step": 216473, "epoch": 2432, "lr": 9.818664239404363e-06}
{"train_loss": 0.042163729667663574, "global_step": 216474, "epoch": 2432, "lr": 9.81831922132444e-06}
{"train_loss": 0.042661357671022415, "global_step": 216475, "epoch": 2432, "lr": 9.817974208646447e-06}
{"train_loss": 0.09867151081562042, "global_step": 216476, "epoch": 2432, "lr": 9.817629201370438e-06}
{"train_loss": 0.07043728232383728, "global_step": 216477, "epoch": 2432, "lr": 9.817284199496469e-06}
{"train_loss": 0.0738554522395134, "global_step": 216478, "epoch": 2432, "lr": 9.81693920302456e-06}
{"train_loss": 0.07031510770320892, "global_step": 216479, "epoch": 2432, "lr": 9.816594211954795e-06}
{"train_loss": 0.06855440139770508, "global_step": 216480, "epoch": 2432, "lr": 9.816249226287183e-06}
{"train_loss": 0.04698135331273079, "global_step": 216481, "epoch": 2432, "lr": 9.815904246021795e-06}
{"train_loss": 0.06941477954387665, "global_step": 216482, "epoch": 2432, "lr": 9.815559271158664e-06}
{"train_loss": 0.02671206369996071, "global_step": 216483, "epoch": 2432, "lr": 9.81521430169785e-06}
{"train_loss": 0.06048063933849335, "global_step": 216484, "epoch": 2432, "lr": 9.814869337639376e-06}
{"train_loss": 0.07984360307455063, "global_step": 216485, "epoch": 2432, "lr": 9.81452437898332e-06}
{"train_loss": 0.04245459660887718, "global_step": 216486, "epoch": 2432, "lr": 9.814179425729692e-06}
{"train_loss": 0.08188177645206451, "global_step": 216487, "epoch": 2432, "lr": 9.813834477878574e-06}
{"train_loss": 0.04126398265361786, "global_step": 216488, "epoch": 2432, "lr": 9.813489535429987e-06}
{"train_loss": 0.07090393453836441, "global_step": 216489, "epoch": 2432, "lr": 9.813144598383994e-06}
{"train_loss": 0.10790824145078659, "global_step": 216490, "epoch": 2432, "lr": 9.812799666740623e-06}
{"train_loss": 0.05760035663843155, "global_step": 216491, "epoch": 2432, "lr": 9.812454740499949e-06}
{"train_loss": 0.0665140226483345, "global_step": 216492, "epoch": 2432, "lr": 9.812109819661996e-06}
{"train_loss": 0.06985831260681152, "global_step": 216493, "epoch": 2432, "lr": 9.8117649042268e-06}
{"train_loss": 0.05306466296315193, "global_step": 216494, "epoch": 2432, "lr": 9.811419994194438e-06}
{"train_loss": 0.02254301682114601, "global_step": 216495, "epoch": 2432, "lr": 9.81107508956493e-06}
{"train_loss": 0.05207028239965439, "global_step": 216496, "epoch": 2432, "lr": 9.81073019033834e-06}
{"train_loss": 0.06397690623998642, "global_step": 216497, "epoch": 2432, "lr": 9.8103852965147e-06}
{"train_loss": 0.0337022989988327, "global_step": 216498, "epoch": 2432, "lr": 9.810040408094074e-06}
{"train_loss": 0.09212593734264374, "global_step": 216499, "epoch": 2432, "lr": 9.809695525076485e-06}
{"train_loss": 0.027987167239189148, "global_step": 216500, "epoch": 2432, "lr": 9.809350647462006e-06}
{"train_loss": 0.0782083049416542, "global_step": 216501, "epoch": 2432, "lr": 9.809005775250663e-06}
{"train_loss": 0.021835021674633026, "global_step": 216502, "epoch": 2432, "lr": 9.808660908442513e-06}
{"train_loss": 0.037596628069877625, "global_step": 216503, "epoch": 2432, "lr": 9.80831604703759e-06}
{"train_loss": 0.035204436630010605, "global_step": 216504, "epoch": 2432, "lr": 9.807971191035964e-06}
{"train_loss": 0.026453683152794838, "global_step": 216505, "epoch": 2432, "lr": 9.807626340437648e-06}
{"train_loss": 0.0646762028336525, "global_step": 216506, "epoch": 2432, "lr": 9.807281495242725e-06}
{"train_loss": 0.058345552533864975, "global_step": 216507, "epoch": 2432, "lr": 9.806936655451209e-06}
{"train_loss": 0.05034567043185234, "global_step": 216508, "epoch": 2432, "lr": 9.806591821063172e-06}
{"train_loss": 0.04130081832408905, "global_step": 216509, "epoch": 2432, "lr": 9.806246992078639e-06}
{"train_loss": 0.06378009915351868, "global_step": 216510, "epoch": 2432, "lr": 9.805902168497671e-06}
{"train_loss": 0.056787796318531036, "global_step": 216511, "epoch": 2432, "lr": 9.805557350320305e-06}
{"train_loss": 0.08539814502000809, "global_step": 216512, "epoch": 2432, "lr": 9.805212537546587e-06}
{"train_loss": 0.03498075529932976, "global_step": 216513, "epoch": 2432, "lr": 9.804867730176586e-06}
{"train_loss": 0.03935809060931206, "global_step": 216514, "epoch": 2432, "lr": 9.804522928210314e-06}
{"train_loss": 0.07517921179533005, "global_step": 216515, "epoch": 2432, "lr": 9.804178131647846e-06}
{"train_loss": 0.09896527230739594, "global_step": 216516, "epoch": 2432, "lr": 9.803833340489204e-06}
{"train_loss": 0.07555904984474182, "global_step": 216517, "epoch": 2432, "lr": 9.803488554734463e-06}
{"train_loss": 0.07052063941955566, "global_step": 216518, "epoch": 2432, "lr": 9.803143774383638e-06}
{"train_loss": 0.04003176838159561, "global_step": 216519, "epoch": 2432, "lr": 9.802798999436802e-06}
{"train_loss": 0.04244910180568695, "global_step": 216520, "epoch": 2432, "lr": 9.802454229893982e-06}
{"train_loss": 0.05383862927556038, "global_step": 216521, "epoch": 2432, "lr": 9.802109465755239e-06}
{"train_loss": 0.04734930768609047, "global_step": 216522, "epoch": 2432, "lr": 9.8017647070206e-06}
{"train_loss": 0.048555370420217514, "global_step": 216523, "epoch": 2432, "lr": 9.80141995369014e-06}
{"train_loss": 0.0375639833509922, "global_step": 216524, "epoch": 2432, "lr": 9.801075205763876e-06}
{"train_loss": 0.06681422144174576, "global_step": 216525, "epoch": 2432, "lr": 9.800730463241875e-06}
{"train_loss": 0.06088883802294731, "global_step": 216526, "epoch": 2432, "lr": 9.800385726124167e-06}
{"train_loss": 0.031534019857645035, "global_step": 216527, "epoch": 2432, "lr": 9.800040994410819e-06}
{"train_loss": 0.04051133617758751, "global_step": 216528, "epoch": 2432, "lr": 9.799696268101855e-06}
{"train_loss": 0.04952773451805115, "global_step": 216529, "epoch": 2432, "lr": 9.799351547197344e-06}
{"train_loss": 0.06508440524339676, "global_step": 216530, "epoch": 2432, "lr": 9.799006831697304e-06}
{"train_loss": 0.03765907511115074, "global_step": 216531, "epoch": 2432, "lr": 9.798662121601814e-06}
{"train_loss": 0.05603025108575821, "global_step": 216532, "epoch": 2432, "lr": 9.798317416910906e-06}
{"train_loss": 0.04788382723927498, "global_step": 216533, "epoch": 2432, "lr": 9.797972717624609e-06}
{"train_loss": 0.021450569853186607, "global_step": 216534, "epoch": 2432, "lr": 9.797628023742993e-06}
{"train_loss": 0.022525396198034286, "global_step": 216535, "epoch": 2432, "lr": 9.797283335266088e-06}
{"train_loss": 0.053622302567858375, "global_step": 216536, "epoch": 2432, "lr": 9.79693865219396e-06, "val_loss": 8.85579776763916}
{"train_loss": 0.019708145409822464, "global_step": 216537, "epoch": 2433, "lr": 9.79659397452663e-06}
{"train_loss": 0.07215463370084763, "global_step": 216538, "epoch": 2433, "lr": 9.796249302264172e-06}
{"train_loss": 0.041299618780612946, "global_step": 216539, "epoch": 2433, "lr": 9.795904635406605e-06}
{"train_loss": 0.054383534938097, "global_step": 216540, "epoch": 2433, "lr": 9.795559973954005e-06}
{"train_loss": 0.04996665194630623, "global_step": 216541, "epoch": 2433, "lr": 9.795215317906381e-06}
{"train_loss": 0.05486099049448967, "global_step": 216542, "epoch": 2433, "lr": 9.794870667263817e-06}
{"train_loss": 0.06495564430952072, "global_step": 216543, "epoch": 2433, "lr": 9.794526022026335e-06}
{"train_loss": 0.021695969626307487, "global_step": 216544, "epoch": 2433, "lr": 9.794181382193995e-06}
{"train_loss": 0.044371847063302994, "global_step": 216545, "epoch": 2433, "lr": 9.793836747766826e-06}
{"train_loss": 0.048476144671440125, "global_step": 216546, "epoch": 2433, "lr": 9.793492118744902e-06}
{"train_loss": 0.06543410569429398, "global_step": 216547, "epoch": 2433, "lr": 9.793147495128235e-06}
{"train_loss": 0.041607461869716644, "global_step": 216548, "epoch": 2433, "lr": 9.79280287691689e-06}
{"train_loss": 0.03051375038921833, "global_step": 216549, "epoch": 2433, "lr": 9.792458264110926e-06}
{"train_loss": 0.07318270206451416, "global_step": 216550, "epoch": 2433, "lr": 9.792113656710366e-06}
{"train_loss": 0.07691769301891327, "global_step": 216551, "epoch": 2433, "lr": 9.791769054715278e-06}
{"train_loss": 0.03320791944861412, "global_step": 216552, "epoch": 2433, "lr": 9.79142445812568e-06}
{"train_loss": 0.04584386944770813, "global_step": 216553, "epoch": 2433, "lr": 9.79107986694165e-06}
{"train_loss": 0.07850969582796097, "global_step": 216554, "epoch": 2433, "lr": 9.79073528116321e-06}
{"train_loss": 0.03797562047839165, "global_step": 216555, "epoch": 2433, "lr": 9.790390700790425e-06}
{"train_loss": 0.07649575173854828, "global_step": 216556, "epoch": 2433, "lr": 9.790046125823316e-06}
{"train_loss": 0.05338417738676071, "global_step": 216557, "epoch": 2433, "lr": 9.789701556261959e-06}
{"train_loss": 0.033005811274051666, "global_step": 216558, "epoch": 2433, "lr": 9.78935699210638e-06}
{"train_loss": 0.015318742021918297, "global_step": 216559, "epoch": 2433, "lr": 9.789012433356642e-06}
{"train_loss": 0.1587364226579666, "global_step": 216560, "epoch": 2433, "lr": 9.788667880012764e-06}
{"train_loss": 0.06949938088655472, "global_step": 216561, "epoch": 2433, "lr": 9.788323332074827e-06}
{"train_loss": 0.07208012789487839, "global_step": 216562, "epoch": 2433, "lr": 9.787978789542846e-06}
{"train_loss": 0.03486078605055809, "global_step": 216563, "epoch": 2433, "lr": 9.787634252416895e-06}
{"train_loss": 0.08706464618444443, "global_step": 216564, "epoch": 2433, "lr": 9.78728972069699e-06}
{"train_loss": 0.06139166280627251, "global_step": 216565, "epoch": 2433, "lr": 9.786945194383207e-06}
{"train_loss": 0.06165141239762306, "global_step": 216566, "epoch": 2433, "lr": 9.786600673475565e-06}
{"train_loss": 0.05407675355672836, "global_step": 216567, "epoch": 2433, "lr": 9.786256157974145e-06}
{"train_loss": 0.049338433891534805, "global_step": 216568, "epoch": 2433, "lr": 9.785911647878954e-06}
{"train_loss": 0.021002011373639107, "global_step": 216569, "epoch": 2433, "lr": 9.785567143190072e-06}
{"train_loss": 0.03818771243095398, "global_step": 216570, "epoch": 2433, "lr": 9.785222643907527e-06}
{"train_loss": 0.03687097877264023, "global_step": 216571, "epoch": 2433, "lr": 9.784878150031356e-06}
{"train_loss": 0.050164416432380676, "global_step": 216572, "epoch": 2433, "lr": 9.784533661561634e-06}
{"train_loss": 0.0628967136144638, "global_step": 216573, "epoch": 2433, "lr": 9.784189178498377e-06}
{"train_loss": 0.05262859910726547, "global_step": 216574, "epoch": 2433, "lr": 9.78384470084166e-06}
{"train_loss": 0.08722417801618576, "global_step": 216575, "epoch": 2433, "lr": 9.783500228591503e-06}
{"train_loss": 0.031072089448571205, "global_step": 216576, "epoch": 2433, "lr": 9.783155761747975e-06}
{"train_loss": 0.11496413499116898, "global_step": 216577, "epoch": 2433, "lr": 9.782811300311095e-06}
{"train_loss": 0.0518748015165329, "global_step": 216578, "epoch": 2433, "lr": 9.782466844280941e-06}
{"train_loss": 0.03470675274729729, "global_step": 216579, "epoch": 2433, "lr": 9.782122393657533e-06}
{"train_loss": 0.05413380637764931, "global_step": 216580, "epoch": 2433, "lr": 9.78177794844094e-06}
{"train_loss": 0.04696047306060791, "global_step": 216581, "epoch": 2433, "lr": 9.781433508631189e-06}
{"train_loss": 0.048111625015735626, "global_step": 216582, "epoch": 2433, "lr": 9.781089074228338e-06}
{"train_loss": 0.03178514167666435, "global_step": 216583, "epoch": 2433, "lr": 9.780744645232426e-06}
{"train_loss": 0.041101448237895966, "global_step": 216584, "epoch": 2433, "lr": 9.78040022164351e-06}
{"train_loss": 0.02499435655772686, "global_step": 216585, "epoch": 2433, "lr": 9.780055803461613e-06}
{"train_loss": 0.1073726937174797, "global_step": 216586, "epoch": 2433, "lr": 9.779711390686802e-06}
{"train_loss": 0.046941258013248444, "global_step": 216587, "epoch": 2433, "lr": 9.779366983319132e-06}
{"train_loss": 0.058095090091228485, "global_step": 216588, "epoch": 2433, "lr": 9.779022581358621e-06}
{"train_loss": 0.02120043896138668, "global_step": 216589, "epoch": 2433, "lr": 9.778678184805345e-06}
{"train_loss": 0.030336350202560425, "global_step": 216590, "epoch": 2433, "lr": 9.778333793659322e-06}
{"train_loss": 0.06361858546733856, "global_step": 216591, "epoch": 2433, "lr": 9.777989407920624e-06}
{"train_loss": 0.08070184290409088, "global_step": 216592, "epoch": 2433, "lr": 9.777645027589272e-06}
{"train_loss": 0.040305525064468384, "global_step": 216593, "epoch": 2433, "lr": 9.777300652665339e-06}
{"train_loss": 0.04425032436847687, "global_step": 216594, "epoch": 2433, "lr": 9.776956283148841e-06}
{"train_loss": 0.07076852023601532, "global_step": 216595, "epoch": 2433, "lr": 9.776611919039857e-06}
{"train_loss": 0.07117018103599548, "global_step": 216596, "epoch": 2433, "lr": 9.776267560338403e-06}
{"train_loss": 0.0206720232963562, "global_step": 216597, "epoch": 2433, "lr": 9.775923207044558e-06}
{"train_loss": 0.04113392531871796, "global_step": 216598, "epoch": 2433, "lr": 9.77557885915833e-06}
{"train_loss": 0.025613704696297646, "global_step": 216599, "epoch": 2433, "lr": 9.775234516679804e-06}
{"train_loss": 0.06120859831571579, "global_step": 216600, "epoch": 2433, "lr": 9.774890179608986e-06}
{"train_loss": 0.06704115867614746, "global_step": 216601, "epoch": 2433, "lr": 9.774545847945965e-06}
{"train_loss": 0.02972766011953354, "global_step": 216602, "epoch": 2433, "lr": 9.774201521690751e-06}
{"train_loss": 0.04999633878469467, "global_step": 216603, "epoch": 2433, "lr": 9.773857200843416e-06}
{"train_loss": 0.06391827017068863, "global_step": 216604, "epoch": 2433, "lr": 9.773512885403985e-06}
{"train_loss": 0.04693203419446945, "global_step": 216605, "epoch": 2433, "lr": 9.773168575372526e-06}
{"train_loss": 0.15875403583049774, "global_step": 216606, "epoch": 2433, "lr": 9.772824270749065e-06}
{"train_loss": 0.04736972972750664, "global_step": 216607, "epoch": 2433, "lr": 9.772479971533666e-06}
{"train_loss": 0.023353679105639458, "global_step": 216608, "epoch": 2433, "lr": 9.772135677726369e-06}
{"train_loss": 0.0346437431871891, "global_step": 216609, "epoch": 2433, "lr": 9.771791389327206e-06}
{"train_loss": 0.02845294401049614, "global_step": 216610, "epoch": 2433, "lr": 9.771447106336246e-06}
{"train_loss": 0.06468593329191208, "global_step": 216611, "epoch": 2433, "lr": 9.77110282875351e-06}
{"train_loss": 0.07124742120504379, "global_step": 216612, "epoch": 2433, "lr": 9.770758556579075e-06}
{"train_loss": 0.053127191960811615, "global_step": 216613, "epoch": 2433, "lr": 9.770414289812957e-06}
{"train_loss": 0.0941992849111557, "global_step": 216614, "epoch": 2433, "lr": 9.770070028455231e-06}
{"train_loss": 0.046839069575071335, "global_step": 216615, "epoch": 2433, "lr": 9.769725772505917e-06}
{"train_loss": 0.05258466303348541, "global_step": 216616, "epoch": 2433, "lr": 9.769381521965082e-06}
{"train_loss": 0.06257440149784088, "global_step": 216617, "epoch": 2433, "lr": 9.769037276832755e-06}
{"train_loss": 0.07449270039796829, "global_step": 216618, "epoch": 2433, "lr": 9.768693037108994e-06}
{"train_loss": 0.06227077916264534, "global_step": 216619, "epoch": 2433, "lr": 9.768348802793837e-06}
{"train_loss": 0.040209054946899414, "global_step": 216620, "epoch": 2433, "lr": 9.768004573887346e-06}
{"train_loss": 0.06683249026536942, "global_step": 216621, "epoch": 2433, "lr": 9.767660350389546e-06}
{"train_loss": 0.03429455682635307, "global_step": 216622, "epoch": 2433, "lr": 9.767316132300491e-06}
{"train_loss": 0.05373260751366615, "global_step": 216623, "epoch": 2433, "lr": 9.766971919620244e-06}
{"train_loss": 0.06085658445954323, "global_step": 216624, "epoch": 2433, "lr": 9.766627712348825e-06}
{"train_loss": 0.05401182465590118, "global_step": 216625, "epoch": 2433, "lr": 9.766283510486307e-06, "val_loss": 9.086296081542969}
{"train_loss": 0.07060140371322632, "global_step": 216626, "epoch": 2434, "lr": 9.765939314032702e-06}
{"train_loss": 0.02840043045580387, "global_step": 216627, "epoch": 2434, "lr": 9.765595122988098e-06}
{"train_loss": 0.07844705134630203, "global_step": 216628, "epoch": 2434, "lr": 9.765250937352498e-06}
{"train_loss": 0.02524624951183796, "global_step": 216629, "epoch": 2434, "lr": 9.764906757125985e-06}
{"train_loss": 0.05182768404483795, "global_step": 216630, "epoch": 2434, "lr": 9.764562582308579e-06}
{"train_loss": 0.0810183435678482, "global_step": 216631, "epoch": 2434, "lr": 9.76421841290035e-06}
{"train_loss": 0.05547849088907242, "global_step": 216632, "epoch": 2434, "lr": 9.763874248901316e-06}
{"train_loss": 0.03386463224887848, "global_step": 216633, "epoch": 2434, "lr": 9.763530090311557e-06}
{"train_loss": 0.06910243630409241, "global_step": 216634, "epoch": 2434, "lr": 9.763185937131087e-06}
{"train_loss": 0.10724738985300064, "global_step": 216635, "epoch": 2434, "lr": 9.762841789359973e-06}
{"train_loss": 0.055868517607450485, "global_step": 216636, "epoch": 2434, "lr": 9.762497646998248e-06}
{"train_loss": 0.08337056636810303, "global_step": 216637, "epoch": 2434, "lr": 9.762153510045974e-06}
{"train_loss": 0.0579109862446785, "global_step": 216638, "epoch": 2434, "lr": 9.761809378503178e-06}
{"train_loss": 0.07963329553604126, "global_step": 216639, "epoch": 2434, "lr": 9.761465252369911e-06}
{"train_loss": 0.05665275454521179, "global_step": 216640, "epoch": 2434, "lr": 9.761121131646245e-06}
{"train_loss": 0.06327680498361588, "global_step": 216641, "epoch": 2434, "lr": 9.760777016332185e-06}
{"train_loss": 0.0930781215429306, "global_step": 216642, "epoch": 2434, "lr": 9.760432906427814e-06}
{"train_loss": 0.07096360623836517, "global_step": 216643, "epoch": 2434, "lr": 9.760088801933165e-06}
{"train_loss": 0.06233879178762436, "global_step": 216644, "epoch": 2434, "lr": 9.759744702848268e-06}
{"train_loss": 0.0919291228055954, "global_step": 216645, "epoch": 2434, "lr": 9.759400609173192e-06}
{"train_loss": 0.07513651996850967, "global_step": 216646, "epoch": 2434, "lr": 9.759056520907967e-06}
{"train_loss": 0.04154420271515846, "global_step": 216647, "epoch": 2434, "lr": 9.758712438052653e-06}
{"train_loss": 0.052772633731365204, "global_step": 216648, "epoch": 2434, "lr": 9.758368360607295e-06}
{"train_loss": 0.07410275936126709, "global_step": 216649, "epoch": 2434, "lr": 9.75802428857192e-06}
{"train_loss": 0.025509852916002274, "global_step": 216650, "epoch": 2434, "lr": 9.757680221946596e-06}
{"train_loss": 0.049112819135189056, "global_step": 216651, "epoch": 2434, "lr": 9.757336160731357e-06}
{"train_loss": 0.11437845975160599, "global_step": 216652, "epoch": 2434, "lr": 9.75699210492626e-06}
{"train_loss": 0.02703762799501419, "global_step": 216653, "epoch": 2434, "lr": 9.756648054531331e-06}
{"train_loss": 0.04384172707796097, "global_step": 216654, "epoch": 2434, "lr": 9.756304009546647e-06}
{"train_loss": 0.01789155974984169, "global_step": 216655, "epoch": 2434, "lr": 9.755959969972223e-06}
{"train_loss": 0.07386172562837601, "global_step": 216656, "epoch": 2434, "lr": 9.755615935808132e-06}
{"train_loss": 0.08284195512533188, "global_step": 216657, "epoch": 2434, "lr": 9.755271907054397e-06}
{"train_loss": 0.06402578949928284, "global_step": 216658, "epoch": 2434, "lr": 9.754927883711074e-06}
{"train_loss": 0.08165288716554642, "global_step": 216659, "epoch": 2434, "lr": 9.754583865778227e-06}
{"train_loss": 0.08177995681762695, "global_step": 216660, "epoch": 2434, "lr": 9.75423985325587e-06}
{"train_loss": 0.029274392873048782, "global_step": 216661, "epoch": 2434, "lr": 9.753895846144074e-06}
{"train_loss": 0.05210381746292114, "global_step": 216662, "epoch": 2434, "lr": 9.753551844442865e-06}
{"train_loss": 0.03766767680644989, "global_step": 216663, "epoch": 2434, "lr": 9.753207848152317e-06}
{"train_loss": 0.08461503684520721, "global_step": 216664, "epoch": 2434, "lr": 9.752863857272442e-06}
{"train_loss": 0.1049969345331192, "global_step": 216665, "epoch": 2434, "lr": 9.752519871803323e-06}
{"train_loss": 0.04374345391988754, "global_step": 216666, "epoch": 2434, "lr": 9.752175891744968e-06}
{"train_loss": 0.045457445085048676, "global_step": 216667, "epoch": 2434, "lr": 9.75183191709746e-06}
{"train_loss": 0.06853841245174408, "global_step": 216668, "epoch": 2434, "lr": 9.751487947860811e-06}
{"train_loss": 0.01902052015066147, "global_step": 216669, "epoch": 2434, "lr": 9.751143984035099e-06}
{"train_loss": 0.026158040389418602, "global_step": 216670, "epoch": 2434, "lr": 9.750800025620344e-06}
{"train_loss": 0.05603493005037308, "global_step": 216671, "epoch": 2434, "lr": 9.750456072616615e-06}
{"train_loss": 0.04477258026599884, "global_step": 216672, "epoch": 2434, "lr": 9.75011212502393e-06}
{"train_loss": 0.0369715616106987, "global_step": 216673, "epoch": 2434, "lr": 9.749768182842367e-06}
{"train_loss": 0.06046618893742561, "global_step": 216674, "epoch": 2434, "lr": 9.749424246071947e-06}
{"train_loss": 0.05392872169613838, "global_step": 216675, "epoch": 2434, "lr": 9.749080314712721e-06}
{"train_loss": 0.03282981738448143, "global_step": 216676, "epoch": 2434, "lr": 9.748736388764757e-06}
{"train_loss": 0.0223566722124815, "global_step": 216677, "epoch": 2434, "lr": 9.748392468228073e-06}
{"train_loss": 0.0890866369009018, "global_step": 216678, "epoch": 2434, "lr": 9.748048553102735e-06}
{"train_loss": 0.11208934336900711, "global_step": 216679, "epoch": 2434, "lr": 9.747704643388772e-06}
{"train_loss": 0.10616131871938705, "global_step": 216680, "epoch": 2434, "lr": 9.747360739086253e-06}
{"train_loss": 0.04072826728224754, "global_step": 216681, "epoch": 2434, "lr": 9.747016840195212e-06}
{"train_loss": 0.04253578558564186, "global_step": 216682, "epoch": 2434, "lr": 9.746672946715679e-06}
{"train_loss": 0.03823341056704521, "global_step": 216683, "epoch": 2434, "lr": 9.746329058647724e-06}
{"train_loss": 0.03785755857825279, "global_step": 216684, "epoch": 2434, "lr": 9.74598517599138e-06}
{"train_loss": 0.0621425099670887, "global_step": 216685, "epoch": 2434, "lr": 9.745641298746705e-06}
{"train_loss": 0.028797194361686707, "global_step": 216686, "epoch": 2434, "lr": 9.745297426913735e-06}
{"train_loss": 0.013065053150057793, "global_step": 216687, "epoch": 2434, "lr": 9.744953560492515e-06}
{"train_loss": 0.07113344967365265, "global_step": 216688, "epoch": 2434, "lr": 9.744609699483099e-06}
{"train_loss": 0.04112182557582855, "global_step": 216689, "epoch": 2434, "lr": 9.744265843885525e-06}
{"train_loss": 0.04215060919523239, "global_step": 216690, "epoch": 2434, "lr": 9.74392199369985e-06}
{"train_loss": 0.06841344386339188, "global_step": 216691, "epoch": 2434, "lr": 9.743578148926103e-06}
{"train_loss": 0.0694398283958435, "global_step": 216692, "epoch": 2434, "lr": 9.743234309564359e-06}
{"train_loss": 0.046534907072782516, "global_step": 216693, "epoch": 2434, "lr": 9.742890475614625e-06}
{"train_loss": 0.061970606446266174, "global_step": 216694, "epoch": 2434, "lr": 9.742546647076977e-06}
{"train_loss": 0.04758121818304062, "global_step": 216695, "epoch": 2434, "lr": 9.742202823951462e-06}
{"train_loss": 0.06923231482505798, "global_step": 216696, "epoch": 2434, "lr": 9.741859006238101e-06}
{"train_loss": 0.05117545649409294, "global_step": 216697, "epoch": 2434, "lr": 9.741515193936973e-06}
{"train_loss": 0.053182896226644516, "global_step": 216698, "epoch": 2434, "lr": 9.741171387048098e-06}
{"train_loss": 0.1282491385936737, "global_step": 216699, "epoch": 2434, "lr": 9.740827585571538e-06}
{"train_loss": 0.046275146305561066, "global_step": 216700, "epoch": 2434, "lr": 9.740483789507326e-06}
{"train_loss": 0.04939938336610794, "global_step": 216701, "epoch": 2434, "lr": 9.740139998855524e-06}
{"train_loss": 0.07998443394899368, "global_step": 216702, "epoch": 2434, "lr": 9.73979621361616e-06}
{"train_loss": 0.03353963419795036, "global_step": 216703, "epoch": 2434, "lr": 9.739452433789299e-06}
{"train_loss": 0.07670392096042633, "global_step": 216704, "epoch": 2434, "lr": 9.739108659374968e-06}
{"train_loss": 0.060605503618717194, "global_step": 216705, "epoch": 2434, "lr": 9.738764890373232e-06}
{"train_loss": 0.022256355732679367, "global_step": 216706, "epoch": 2434, "lr": 9.738421126784119e-06}
{"train_loss": 0.04433545470237732, "global_step": 216707, "epoch": 2434, "lr": 9.738077368607696e-06}
{"train_loss": 0.11116448044776917, "global_step": 216708, "epoch": 2434, "lr": 9.737733615843987e-06}
{"train_loss": 0.07694930583238602, "global_step": 216709, "epoch": 2434, "lr": 9.737389868493058e-06}
{"train_loss": 0.07434751093387604, "global_step": 216710, "epoch": 2434, "lr": 9.737046126554938e-06}
{"train_loss": 0.04079177603125572, "global_step": 216711, "epoch": 2434, "lr": 9.73670239002969e-06}
{"train_loss": 0.0162766445428133, "global_step": 216712, "epoch": 2434, "lr": 9.736358658917344e-06}
{"train_loss": 0.033644840121269226, "global_step": 216713, "epoch": 2434, "lr": 9.736014933217946e-06}
{"train_loss": 0.05790896247025956, "global_step": 216714, "epoch": 2434, "lr": 9.735671212931569e-06, "val_loss": 8.865569114685059}
{"train_loss": 0.052758727222681046, "global_step": 216715, "epoch": 2435, "lr": 9.735327498058228e-06}
{"train_loss": 0.07705073058605194, "global_step": 216716, "epoch": 2435, "lr": 9.73498378859799e-06}
{"train_loss": 0.07145324349403381, "global_step": 216717, "epoch": 2435, "lr": 9.734640084550884e-06}
{"train_loss": 0.06348872184753418, "global_step": 216718, "epoch": 2435, "lr": 9.734296385916974e-06}
{"train_loss": 0.09708309918642044, "global_step": 216719, "epoch": 2435, "lr": 9.733952692696296e-06}
{"train_loss": 0.046644844114780426, "global_step": 216720, "epoch": 2435, "lr": 9.733609004888889e-06}
{"train_loss": 0.046034038066864014, "global_step": 216721, "epoch": 2435, "lr": 9.733265322494816e-06}
{"train_loss": 0.07338578999042511, "global_step": 216722, "epoch": 2435, "lr": 9.73292164551411e-06}
{"train_loss": 0.030743326991796494, "global_step": 216723, "epoch": 2435, "lr": 9.732577973946827e-06}
{"train_loss": 0.07876858115196228, "global_step": 216724, "epoch": 2435, "lr": 9.732234307793009e-06}
{"train_loss": 0.04327383264899254, "global_step": 216725, "epoch": 2435, "lr": 9.731890647052688e-06}
{"train_loss": 0.03056424856185913, "global_step": 216726, "epoch": 2435, "lr": 9.731546991725943e-06}
{"train_loss": 0.054682765156030655, "global_step": 216727, "epoch": 2435, "lr": 9.731203341812778e-06}
{"train_loss": 0.12245143204927444, "global_step": 216728, "epoch": 2435, "lr": 9.730859697313282e-06}
{"train_loss": 0.03567349165678024, "global_step": 216729, "epoch": 2435, "lr": 9.730516058227467e-06}
{"train_loss": 0.052132971584796906, "global_step": 216730, "epoch": 2435, "lr": 9.730172424555389e-06}
{"train_loss": 0.07122161239385605, "global_step": 216731, "epoch": 2435, "lr": 9.729828796297118e-06}
{"train_loss": 0.06514829397201538, "global_step": 216732, "epoch": 2435, "lr": 9.729485173452662e-06}
{"train_loss": 0.07251439243555069, "global_step": 216733, "epoch": 2435, "lr": 9.729141556022103e-06}
{"train_loss": 0.038787152618169785, "global_step": 216734, "epoch": 2435, "lr": 9.728797944005452e-06}
{"train_loss": 0.035973675549030304, "global_step": 216735, "epoch": 2435, "lr": 9.728454337402792e-06}
{"train_loss": 0.042387623339891434, "global_step": 216736, "epoch": 2435, "lr": 9.728110736214136e-06}
{"train_loss": 0.04193764552474022, "global_step": 216737, "epoch": 2435, "lr": 9.727767140439553e-06}
{"train_loss": 0.051421329379081726, "global_step": 216738, "epoch": 2435, "lr": 9.727423550079073e-06}
{"train_loss": 0.0625842809677124, "global_step": 216739, "epoch": 2435, "lr": 9.727079965132762e-06}
{"train_loss": 0.03326968103647232, "global_step": 216740, "epoch": 2435, "lr": 9.726736385600643e-06}
{"train_loss": 0.11944299191236496, "global_step": 216741, "epoch": 2435, "lr": 9.726392811482782e-06}
{"train_loss": 0.09516734629869461, "global_step": 216742, "epoch": 2435, "lr": 9.726049242779206e-06}
{"train_loss": 0.0690956637263298, "global_step": 216743, "epoch": 2435, "lr": 9.725705679489982e-06}
{"train_loss": 0.05997886136174202, "global_step": 216744, "epoch": 2435, "lr": 9.725362121615138e-06}
{"train_loss": 0.10149198025465012, "global_step": 216745, "epoch": 2435, "lr": 9.725018569154743e-06}
{"train_loss": 0.04788028076291084, "global_step": 216746, "epoch": 2435, "lr": 9.724675022108809e-06}
{"train_loss": 0.022168850526213646, "global_step": 216747, "epoch": 2435, "lr": 9.724331480477416e-06}
{"train_loss": 0.06497126817703247, "global_step": 216748, "epoch": 2435, "lr": 9.723987944260581e-06}
{"train_loss": 0.05434055253863335, "global_step": 216749, "epoch": 2435, "lr": 9.723644413458372e-06}
{"train_loss": 0.04788753017783165, "global_step": 216750, "epoch": 2435, "lr": 9.723300888070835e-06}
{"train_loss": 0.048339080065488815, "global_step": 216751, "epoch": 2435, "lr": 9.722957368097995e-06}
{"train_loss": 0.035887300968170166, "global_step": 216752, "epoch": 2435, "lr": 9.72261385353993e-06}
{"train_loss": 0.07889313995838165, "global_step": 216753, "epoch": 2435, "lr": 9.722270344396655e-06}
{"train_loss": 0.02119891718029976, "global_step": 216754, "epoch": 2435, "lr": 9.721926840668244e-06}
{"train_loss": 0.04663339629769325, "global_step": 216755, "epoch": 2435, "lr": 9.721583342354712e-06}
{"train_loss": 0.04078977555036545, "global_step": 216756, "epoch": 2435, "lr": 9.721239849456137e-06}
{"train_loss": 0.07972074300050735, "global_step": 216757, "epoch": 2435, "lr": 9.720896361972536e-06}
{"train_loss": 0.058203667402267456, "global_step": 216758, "epoch": 2435, "lr": 9.720552879903982e-06}
{"train_loss": 0.05797129124403, "global_step": 216759, "epoch": 2435, "lr": 9.720209403250513e-06}
{"train_loss": 0.04884462058544159, "global_step": 216760, "epoch": 2435, "lr": 9.719865932012157e-06}
{"train_loss": 0.053078263998031616, "global_step": 216761, "epoch": 2435, "lr": 9.719522466188985e-06}
{"train_loss": 0.05013073980808258, "global_step": 216762, "epoch": 2435, "lr": 9.719179005781021e-06}
{"train_loss": 0.03704618662595749, "global_step": 216763, "epoch": 2435, "lr": 9.71883555078833e-06}
{"train_loss": 0.035309117287397385, "global_step": 216764, "epoch": 2435, "lr": 9.718492101210957e-06}
{"train_loss": 0.07123434543609619, "global_step": 216765, "epoch": 2435, "lr": 9.718148657048926e-06}
{"train_loss": 0.0682702288031578, "global_step": 216766, "epoch": 2435, "lr": 9.7178052183023e-06}
{"train_loss": 0.047830600291490555, "global_step": 216767, "epoch": 2435, "lr": 9.717461784971138e-06}
{"train_loss": 0.056940510869026184, "global_step": 216768, "epoch": 2435, "lr": 9.71711835705546e-06}
{"train_loss": 0.08183382451534271, "global_step": 216769, "epoch": 2435, "lr": 9.716774934555334e-06}
{"train_loss": 0.08311839401721954, "global_step": 216770, "epoch": 2435, "lr": 9.716431517470787e-06}
{"train_loss": 0.04750755801796913, "global_step": 216771, "epoch": 2435, "lr": 9.716088105801885e-06}
{"train_loss": 0.06697100400924683, "global_step": 216772, "epoch": 2435, "lr": 9.71574469954865e-06}
{"train_loss": 0.03981131687760353, "global_step": 216773, "epoch": 2435, "lr": 9.715401298711157e-06}
{"train_loss": 0.04261375591158867, "global_step": 216774, "epoch": 2435, "lr": 9.715057903289427e-06}
{"train_loss": 0.09298557043075562, "global_step": 216775, "epoch": 2435, "lr": 9.714714513283523e-06}
{"train_loss": 0.04915214702486992, "global_step": 216776, "epoch": 2435, "lr": 9.714371128693479e-06}
{"train_loss": 0.04190737009048462, "global_step": 216777, "epoch": 2435, "lr": 9.714027749519356e-06}
{"train_loss": 0.0718773603439331, "global_step": 216778, "epoch": 2435, "lr": 9.713684375761179e-06}
{"train_loss": 0.07949715852737427, "global_step": 216779, "epoch": 2435, "lr": 9.713341007419014e-06}
{"train_loss": 0.030935194343328476, "global_step": 216780, "epoch": 2435, "lr": 9.712997644492894e-06}
{"train_loss": 0.09731744974851608, "global_step": 216781, "epoch": 2435, "lr": 9.71265428698288e-06}
{"train_loss": 0.06124074012041092, "global_step": 216782, "epoch": 2435, "lr": 9.712310934888996e-06}
{"train_loss": 0.05505693703889847, "global_step": 216783, "epoch": 2435, "lr": 9.711967588211313e-06}
{"train_loss": 0.07478628307580948, "global_step": 216784, "epoch": 2435, "lr": 9.711624246949851e-06}
{"train_loss": 0.1028333231806755, "global_step": 216785, "epoch": 2435, "lr": 9.711280911104669e-06}
{"train_loss": 0.055017128586769104, "global_step": 216786, "epoch": 2435, "lr": 9.710937580675832e-06}
{"train_loss": 0.08108815550804138, "global_step": 216787, "epoch": 2435, "lr": 9.710594255663353e-06}
{"train_loss": 0.04014037549495697, "global_step": 216788, "epoch": 2435, "lr": 9.710250936067305e-06}
{"train_loss": 0.03775480017066002, "global_step": 216789, "epoch": 2435, "lr": 9.70990762188771e-06}
{"train_loss": 0.09035340696573257, "global_step": 216790, "epoch": 2435, "lr": 9.709564313124642e-06}
{"train_loss": 0.03683781251311302, "global_step": 216791, "epoch": 2435, "lr": 9.709221009778119e-06}
{"train_loss": 0.04740283638238907, "global_step": 216792, "epoch": 2435, "lr": 9.708877711848213e-06}
{"train_loss": 0.07063139230012894, "global_step": 216793, "epoch": 2435, "lr": 9.708534419334941e-06}
{"train_loss": 0.0669044777750969, "global_step": 216794, "epoch": 2435, "lr": 9.70819113223838e-06}
{"train_loss": 0.10589129477739334, "global_step": 216795, "epoch": 2435, "lr": 9.707847850558555e-06}
{"train_loss": 0.0385853536427021, "global_step": 216796, "epoch": 2435, "lr": 9.707504574295524e-06}
{"train_loss": 0.054843682795763016, "global_step": 216797, "epoch": 2435, "lr": 9.70716130344933e-06}
{"train_loss": 0.05124968662858009, "global_step": 216798, "epoch": 2435, "lr": 9.706818038020005e-06}
{"train_loss": 0.05112985894083977, "global_step": 216799, "epoch": 2435, "lr": 9.706474778007619e-06}
{"train_loss": 0.11211985349655151, "global_step": 216800, "epoch": 2435, "lr": 9.706131523412194e-06}
{"train_loss": 0.049857452511787415, "global_step": 216801, "epoch": 2435, "lr": 9.705788274233801e-06}
{"train_loss": 0.09440304338932037, "global_step": 216802, "epoch": 2435, "lr": 9.70544503047246e-06}
{"train_loss": 0.06042815211197634, "global_step": 216803, "epoch": 2435, "lr": 9.705101792128247e-06, "val_loss": 8.879590034484863, "train_action_mse_error": 17.55429458618164}
{"train_loss": 0.0531216599047184, "global_step": 216804, "epoch": 2436, "lr": 9.704758559201176e-06}
{"train_loss": 0.04190206900238991, "global_step": 216805, "epoch": 2436, "lr": 9.704415331691325e-06}
{"train_loss": 0.04192858189344406, "global_step": 216806, "epoch": 2436, "lr": 9.704072109598705e-06}
{"train_loss": 0.07087919116020203, "global_step": 216807, "epoch": 2436, "lr": 9.703728892923403e-06}
{"train_loss": 0.057320594787597656, "global_step": 216808, "epoch": 2436, "lr": 9.703385681665422e-06}
{"train_loss": 0.05470249056816101, "global_step": 216809, "epoch": 2436, "lr": 9.703042475824848e-06}
{"train_loss": 0.07380610704421997, "global_step": 216810, "epoch": 2436, "lr": 9.702699275401694e-06}
{"train_loss": 0.06224110350012779, "global_step": 216811, "epoch": 2436, "lr": 9.702356080396036e-06}
{"train_loss": 0.04936728999018669, "global_step": 216812, "epoch": 2436, "lr": 9.702012890807889e-06}
{"train_loss": 0.0931771770119667, "global_step": 216813, "epoch": 2436, "lr": 9.701669706637329e-06}
{"train_loss": 0.04949934780597687, "global_step": 216814, "epoch": 2436, "lr": 9.701326527884374e-06}
{"train_loss": 0.05168715491890907, "global_step": 216815, "epoch": 2436, "lr": 9.700983354549104e-06}
{"train_loss": 0.03912058100104332, "global_step": 216816, "epoch": 2436, "lr": 9.700640186631527e-06}
{"train_loss": 0.1063970997929573, "global_step": 216817, "epoch": 2436, "lr": 9.70029702413172e-06}
{"train_loss": 0.029263081029057503, "global_step": 216818, "epoch": 2436, "lr": 9.699953867049705e-06}
{"train_loss": 0.04003801569342613, "global_step": 216819, "epoch": 2436, "lr": 9.699610715385554e-06}
{"train_loss": 0.12529899179935455, "global_step": 216820, "epoch": 2436, "lr": 9.699267569139286e-06}
{"train_loss": 0.050124529749155045, "global_step": 216821, "epoch": 2436, "lr": 9.698924428310963e-06}
{"train_loss": 0.07639545202255249, "global_step": 216822, "epoch": 2436, "lr": 9.698581292900638e-06}
{"train_loss": 0.06647492945194244, "global_step": 216823, "epoch": 2436, "lr": 9.698238162908336e-06}
{"train_loss": 0.04389258846640587, "global_step": 216824, "epoch": 2436, "lr": 9.697895038334126e-06}
{"train_loss": 0.0821552500128746, "global_step": 216825, "epoch": 2436, "lr": 9.697551919178028e-06}
{"train_loss": 0.03415083512663841, "global_step": 216826, "epoch": 2436, "lr": 9.697208805440117e-06}
{"train_loss": 0.035926204174757004, "global_step": 216827, "epoch": 2436, "lr": 9.696865697120416e-06}
{"train_loss": 0.08043640851974487, "global_step": 216828, "epoch": 2436, "lr": 9.696522594218993e-06}
{"train_loss": 0.06345093250274658, "global_step": 216829, "epoch": 2436, "lr": 9.696179496735863e-06}
{"train_loss": 0.05133425444364548, "global_step": 216830, "epoch": 2436, "lr": 9.695836404671105e-06}
{"train_loss": 0.03801342099905014, "global_step": 216831, "epoch": 2436, "lr": 9.69549331802474e-06}
{"train_loss": 0.07773160934448242, "global_step": 216832, "epoch": 2436, "lr": 9.695150236796835e-06}
{"train_loss": 0.06288772076368332, "global_step": 216833, "epoch": 2436, "lr": 9.694807160987417e-06}
{"train_loss": 0.05944085493683815, "global_step": 216834, "epoch": 2436, "lr": 9.69446409059655e-06}
{"train_loss": 0.06741436570882797, "global_step": 216835, "epoch": 2436, "lr": 9.694121025624275e-06}
{"train_loss": 0.051685694605112076, "global_step": 216836, "epoch": 2436, "lr": 9.693777966070616e-06}
{"train_loss": 0.06846485286951065, "global_step": 216837, "epoch": 2436, "lr": 9.693434911935657e-06}
{"train_loss": 0.06135103106498718, "global_step": 216838, "epoch": 2436, "lr": 9.693091863219405e-06}
{"train_loss": 0.07765466719865799, "global_step": 216839, "epoch": 2436, "lr": 9.692748819921937e-06}
{"train_loss": 0.019671030342578888, "global_step": 216840, "epoch": 2436, "lr": 9.69240578204328e-06}
{"train_loss": 0.022952022030949593, "global_step": 216841, "epoch": 2436, "lr": 9.692062749583503e-06}
{"train_loss": 0.02629389427602291, "global_step": 216842, "epoch": 2436, "lr": 9.691719722542619e-06}
{"train_loss": 0.0476686991751194, "global_step": 216843, "epoch": 2436, "lr": 9.691376700920707e-06}
{"train_loss": 0.04640376940369606, "global_step": 216844, "epoch": 2436, "lr": 9.691033684717781e-06}
{"train_loss": 0.0857100784778595, "global_step": 216845, "epoch": 2436, "lr": 9.690690673933922e-06}
{"train_loss": 0.06437689065933228, "global_step": 216846, "epoch": 2436, "lr": 9.690347668569144e-06}
{"train_loss": 0.08047954738140106, "global_step": 216847, "epoch": 2436, "lr": 9.69000466862352e-06}
{"train_loss": 0.05487849563360214, "global_step": 216848, "epoch": 2436, "lr": 9.689661674097073e-06}
{"train_loss": 0.06142660975456238, "global_step": 216849, "epoch": 2436, "lr": 9.689318684989867e-06}
{"train_loss": 0.025503575801849365, "global_step": 216850, "epoch": 2436, "lr": 9.688975701301933e-06}
{"train_loss": 0.026665426790714264, "global_step": 216851, "epoch": 2436, "lr": 9.688632723033336e-06}
{"train_loss": 0.06421325355768204, "global_step": 216852, "epoch": 2436, "lr": 9.688289750184098e-06}
{"train_loss": 0.0887276753783226, "global_step": 216853, "epoch": 2436, "lr": 9.687946782754292e-06}
{"train_loss": 0.07377099245786667, "global_step": 216854, "epoch": 2436, "lr": 9.687603820743935e-06}
{"train_loss": 0.09718339890241623, "global_step": 216855, "epoch": 2436, "lr": 9.687260864153103e-06}
{"train_loss": 0.0922580361366272, "global_step": 216856, "epoch": 2436, "lr": 9.686917912981813e-06}
{"train_loss": 0.04437483847141266, "global_step": 216857, "epoch": 2436, "lr": 9.686574967230128e-06}
{"train_loss": 0.0476621650159359, "global_step": 216858, "epoch": 2436, "lr": 9.686232026898102e-06}
{"train_loss": 0.11297806352376938, "global_step": 216859, "epoch": 2436, "lr": 9.685889091985762e-06}
{"train_loss": 0.09415391087532043, "global_step": 216860, "epoch": 2436, "lr": 9.685546162493169e-06}
{"train_loss": 0.04284488037228584, "global_step": 216861, "epoch": 2436, "lr": 9.685203238420354e-06}
{"train_loss": 0.04363236203789711, "global_step": 216862, "epoch": 2436, "lr": 9.684860319767386e-06}
{"train_loss": 0.03400273248553276, "global_step": 216863, "epoch": 2436, "lr": 9.684517406534283e-06}
{"train_loss": 0.07438473403453827, "global_step": 216864, "epoch": 2436, "lr": 9.68417449872112e-06}
{"train_loss": 0.0539417564868927, "global_step": 216865, "epoch": 2436, "lr": 9.683831596327914e-06}
{"train_loss": 0.05825048312544823, "global_step": 216866, "epoch": 2436, "lr": 9.683488699354737e-06}
{"train_loss": 0.08545132726430893, "global_step": 216867, "epoch": 2436, "lr": 9.683145807801614e-06}
{"train_loss": 0.01660841330885887, "global_step": 216868, "epoch": 2436, "lr": 9.68280292166861e-06}
{"train_loss": 0.060403309762477875, "global_step": 216869, "epoch": 2436, "lr": 9.682460040955749e-06}
{"train_loss": 0.0506058931350708, "global_step": 216870, "epoch": 2436, "lr": 9.682117165663107e-06}
{"train_loss": 0.04497689753770828, "global_step": 216871, "epoch": 2436, "lr": 9.681774295790697e-06}
{"train_loss": 0.02227947488427162, "global_step": 216872, "epoch": 2436, "lr": 9.681431431338594e-06}
{"train_loss": 0.09418435394763947, "global_step": 216873, "epoch": 2436, "lr": 9.681088572306818e-06}
{"train_loss": 0.05097230151295662, "global_step": 216874, "epoch": 2436, "lr": 9.680745718695438e-06}
{"train_loss": 0.06869478523731232, "global_step": 216875, "epoch": 2436, "lr": 9.680402870504496e-06}
{"train_loss": 0.03638199344277382, "global_step": 216876, "epoch": 2436, "lr": 9.680060027734017e-06}
{"train_loss": 0.11439680308103561, "global_step": 216877, "epoch": 2436, "lr": 9.679717190384075e-06}
{"train_loss": 0.021369583904743195, "global_step": 216878, "epoch": 2436, "lr": 9.67937435845469e-06}
{"train_loss": 0.05844942107796669, "global_step": 216879, "epoch": 2436, "lr": 9.679031531945937e-06}
{"train_loss": 0.034257207065820694, "global_step": 216880, "epoch": 2436, "lr": 9.67868871085783e-06}
{"train_loss": 0.059723109006881714, "global_step": 216881, "epoch": 2436, "lr": 9.67834589519045e-06}
{"train_loss": 0.07785488665103912, "global_step": 216882, "epoch": 2436, "lr": 9.678003084943809e-06}
{"train_loss": 0.04759861156344414, "global_step": 216883, "epoch": 2436, "lr": 9.677660280117985e-06}
{"train_loss": 0.04953716695308685, "global_step": 216884, "epoch": 2436, "lr": 9.677317480712988e-06}
{"train_loss": 0.023323990404605865, "global_step": 216885, "epoch": 2436, "lr": 9.676974686728907e-06}
{"train_loss": 0.050809960812330246, "global_step": 216886, "epoch": 2436, "lr": 9.676631898165744e-06}
{"train_loss": 0.05062570050358772, "global_step": 216887, "epoch": 2436, "lr": 9.676289115023585e-06}
{"train_loss": 0.04590502381324768, "global_step": 216888, "epoch": 2436, "lr": 9.675946337302439e-06}
{"train_loss": 0.042916178703308105, "global_step": 216889, "epoch": 2436, "lr": 9.67560356500239e-06}
{"train_loss": 0.07563917338848114, "global_step": 216890, "epoch": 2436, "lr": 9.675260798123448e-06}
{"train_loss": 0.08886808902025223, "global_step": 216891, "epoch": 2436, "lr": 9.674918036665687e-06}
{"train_loss": 0.05837049601034502, "global_step": 216892, "epoch": 2436, "lr": 9.674575280629128e-06, "val_loss": 8.858633041381836}
{"train_loss": 0.04651773348450661, "global_step": 216893, "epoch": 2437, "lr": 9.674232530013832e-06}
{"train_loss": 0.1003384068608284, "global_step": 216894, "epoch": 2437, "lr": 9.673889784819862e-06}
{"train_loss": 0.019673891365528107, "global_step": 216895, "epoch": 2437, "lr": 9.673547045047226e-06}
{"train_loss": 0.061561278998851776, "global_step": 216896, "epoch": 2437, "lr": 9.67320431069601e-06}
{"train_loss": 0.0524459071457386, "global_step": 216897, "epoch": 2437, "lr": 9.672861581766223e-06}
{"train_loss": 0.02316232956945896, "global_step": 216898, "epoch": 2437, "lr": 9.67251885825794e-06}
{"train_loss": 0.0505535751581192, "global_step": 216899, "epoch": 2437, "lr": 9.672176140171186e-06}
{"train_loss": 0.0830870270729065, "global_step": 216900, "epoch": 2437, "lr": 9.671833427506027e-06}
{"train_loss": 0.06193090230226517, "global_step": 216901, "epoch": 2437, "lr": 9.671490720262488e-06}
{"train_loss": 0.053593333810567856, "global_step": 216902, "epoch": 2437, "lr": 9.671148018440635e-06}
{"train_loss": 0.061121709644794464, "global_step": 216903, "epoch": 2437, "lr": 9.670805322040494e-06}
{"train_loss": 0.06083047017455101, "global_step": 216904, "epoch": 2437, "lr": 9.670462631062132e-06}
{"train_loss": 0.051857590675354004, "global_step": 216905, "epoch": 2437, "lr": 9.670119945505573e-06}
{"train_loss": 0.11789533495903015, "global_step": 216906, "epoch": 2437, "lr": 9.669777265370888e-06}
{"train_loss": 0.054568350315093994, "global_step": 216907, "epoch": 2437, "lr": 9.6694345906581e-06}
{"train_loss": 0.06745753437280655, "global_step": 216908, "epoch": 2437, "lr": 9.669091921367274e-06}
{"train_loss": 0.0617852583527565, "global_step": 216909, "epoch": 2437, "lr": 9.668749257498433e-06}
{"train_loss": 0.06536229699850082, "global_step": 216910, "epoch": 2437, "lr": 9.668406599051655e-06}
{"train_loss": 0.030190467834472656, "global_step": 216911, "epoch": 2437, "lr": 9.66806394602695e-06}
{"train_loss": 0.051537808030843735, "global_step": 216912, "epoch": 2437, "lr": 9.667721298424398e-06}
{"train_loss": 0.03729917109012604, "global_step": 216913, "epoch": 2437, "lr": 9.66737865624403e-06}
{"train_loss": 0.06299512088298798, "global_step": 216914, "epoch": 2437, "lr": 9.667036019485875e-06}
{"train_loss": 0.03650345653295517, "global_step": 216915, "epoch": 2437, "lr": 9.666693388150012e-06}
{"train_loss": 0.07702438533306122, "global_step": 216916, "epoch": 2437, "lr": 9.66635076223646e-06}
{"train_loss": 0.02746862918138504, "global_step": 216917, "epoch": 2437, "lr": 9.66600814174528e-06}
{"train_loss": 0.02745913900434971, "global_step": 216918, "epoch": 2437, "lr": 9.66566552667651e-06}
{"train_loss": 0.055381644517183304, "global_step": 216919, "epoch": 2437, "lr": 9.665322917030206e-06}
{"train_loss": 0.09923121333122253, "global_step": 216920, "epoch": 2437, "lr": 9.6649803128064e-06}
{"train_loss": 0.0724288821220398, "global_step": 216921, "epoch": 2437, "lr": 9.664637714005154e-06}
{"train_loss": 0.05467327684164047, "global_step": 216922, "epoch": 2437, "lr": 9.664295120626494e-06}
{"train_loss": 0.06501978635787964, "global_step": 216923, "epoch": 2437, "lr": 9.663952532670489e-06}
{"train_loss": 0.038640346378088, "global_step": 216924, "epoch": 2437, "lr": 9.663609950137165e-06}
{"train_loss": 0.043083805590867996, "global_step": 216925, "epoch": 2437, "lr": 9.663267373026585e-06}
{"train_loss": 0.03771203011274338, "global_step": 216926, "epoch": 2437, "lr": 9.66292480133878e-06}
{"train_loss": 0.03996836766600609, "global_step": 216927, "epoch": 2437, "lr": 9.662582235073813e-06}
{"train_loss": 0.07130260765552521, "global_step": 216928, "epoch": 2437, "lr": 9.66223967423171e-06}
{"train_loss": 0.02861589752137661, "global_step": 216929, "epoch": 2437, "lr": 9.661897118812524e-06}
{"train_loss": 0.09929452836513519, "global_step": 216930, "epoch": 2437, "lr": 9.661554568816317e-06}
{"train_loss": 0.05322430655360222, "global_step": 216931, "epoch": 2437, "lr": 9.66121202424311e-06}
{"train_loss": 0.043448422104120255, "global_step": 216932, "epoch": 2437, "lr": 9.660869485092977e-06}
{"train_loss": 0.053601015359163284, "global_step": 216933, "epoch": 2437, "lr": 9.660526951365934e-06}
{"train_loss": 0.05156026780605316, "global_step": 216934, "epoch": 2437, "lr": 9.660184423062051e-06}
{"train_loss": 0.03173208609223366, "global_step": 216935, "epoch": 2437, "lr": 9.659841900181355e-06}
{"train_loss": 0.08076196163892746, "global_step": 216936, "epoch": 2437, "lr": 9.659499382723913e-06}
{"train_loss": 0.06551233679056168, "global_step": 216937, "epoch": 2437, "lr": 9.659156870689751e-06}
{"train_loss": 0.0578828863799572, "global_step": 216938, "epoch": 2437, "lr": 9.658814364078933e-06}
{"train_loss": 0.028979385271668434, "global_step": 216939, "epoch": 2437, "lr": 9.658471862891482e-06}
{"train_loss": 0.04305865243077278, "global_step": 216940, "epoch": 2437, "lr": 9.658129367127472e-06}
{"train_loss": 0.05392739176750183, "global_step": 216941, "epoch": 2437, "lr": 9.657786876786917e-06}
{"train_loss": 0.06522920727729797, "global_step": 216942, "epoch": 2437, "lr": 9.657444391869902e-06}
{"train_loss": 0.04021511971950531, "global_step": 216943, "epoch": 2437, "lr": 9.657101912376432e-06}
{"train_loss": 0.05452200770378113, "global_step": 216944, "epoch": 2437, "lr": 9.65675943830659e-06}
{"train_loss": 0.07241903990507126, "global_step": 216945, "epoch": 2437, "lr": 9.656416969660392e-06}
{"train_loss": 0.04128342494368553, "global_step": 216946, "epoch": 2437, "lr": 9.656074506437907e-06}
{"train_loss": 0.029652822762727737, "global_step": 216947, "epoch": 2437, "lr": 9.65573204863916e-06}
{"train_loss": 0.029485413804650307, "global_step": 216948, "epoch": 2437, "lr": 9.655389596264224e-06}
{"train_loss": 0.06133421137928963, "global_step": 216949, "epoch": 2437, "lr": 9.655047149313113e-06}
{"train_loss": 0.03807304799556732, "global_step": 216950, "epoch": 2437, "lr": 9.654704707785905e-06}
{"train_loss": 0.029057620093226433, "global_step": 216951, "epoch": 2437, "lr": 9.654362271682626e-06}
{"train_loss": 0.04877782613039017, "global_step": 216952, "epoch": 2437, "lr": 9.654019841003314e-06}
{"train_loss": 0.03548462688922882, "global_step": 216953, "epoch": 2437, "lr": 9.653677415748042e-06}
{"train_loss": 0.05082046985626221, "global_step": 216954, "epoch": 2437, "lr": 9.653334995916829e-06}
{"train_loss": 0.051903631538152695, "global_step": 216955, "epoch": 2437, "lr": 9.652992581509745e-06}
{"train_loss": 0.12953916192054749, "global_step": 216956, "epoch": 2437, "lr": 9.652650172526811e-06}
{"train_loss": 0.04145987704396248, "global_step": 216957, "epoch": 2437, "lr": 9.6523077689681e-06}
{"train_loss": 0.05774080380797386, "global_step": 216958, "epoch": 2437, "lr": 9.651965370833632e-06}
{"train_loss": 0.06010695546865463, "global_step": 216959, "epoch": 2437, "lr": 9.651622978123476e-06}
{"train_loss": 0.04622140899300575, "global_step": 216960, "epoch": 2437, "lr": 9.651280590837658e-06}
{"train_loss": 0.019802648574113846, "global_step": 216961, "epoch": 2437, "lr": 9.650938208976245e-06}
{"train_loss": 0.04333047568798065, "global_step": 216962, "epoch": 2437, "lr": 9.650595832539255e-06}
{"train_loss": 0.06550060957670212, "global_step": 216963, "epoch": 2437, "lr": 9.65025346152677e-06}
{"train_loss": 0.03640090301632881, "global_step": 216964, "epoch": 2437, "lr": 9.649911095938801e-06}
{"train_loss": 0.08153728395700455, "global_step": 216965, "epoch": 2437, "lr": 9.649568735775421e-06}
{"train_loss": 0.07101554423570633, "global_step": 216966, "epoch": 2437, "lr": 9.649226381036652e-06}
{"train_loss": 0.05641768500208855, "global_step": 216967, "epoch": 2437, "lr": 9.648884031722554e-06}
{"train_loss": 0.04844513535499573, "global_step": 216968, "epoch": 2437, "lr": 9.648541687833184e-06}
{"train_loss": 0.0845983624458313, "global_step": 216969, "epoch": 2437, "lr": 9.648199349368564e-06}
{"train_loss": 0.02794543094933033, "global_step": 216970, "epoch": 2437, "lr": 9.647857016328764e-06}
{"train_loss": 0.050250664353370667, "global_step": 216971, "epoch": 2437, "lr": 9.647514688713804e-06}
{"train_loss": 0.07614043354988098, "global_step": 216972, "epoch": 2437, "lr": 9.64717236652376e-06}
{"train_loss": 0.05782756209373474, "global_step": 216973, "epoch": 2437, "lr": 9.646830049758643e-06}
{"train_loss": 0.11011771857738495, "global_step": 216974, "epoch": 2437, "lr": 9.646487738418535e-06}
{"train_loss": 0.03560780733823776, "global_step": 216975, "epoch": 2437, "lr": 9.646145432503451e-06}
{"train_loss": 0.05059477314352989, "global_step": 216976, "epoch": 2437, "lr": 9.645803132013464e-06}
{"train_loss": 0.06083820387721062, "global_step": 216977, "epoch": 2437, "lr": 9.645460836948595e-06}
{"train_loss": 0.04697515815496445, "global_step": 216978, "epoch": 2437, "lr": 9.645118547308912e-06}
{"train_loss": 0.02887945994734764, "global_step": 216979, "epoch": 2437, "lr": 9.644776263094441e-06}
{"train_loss": 0.07665546238422394, "global_step": 216980, "epoch": 2437, "lr": 9.64443398430525e-06}
{"train_loss": 0.055007563276069885, "global_step": 216981, "epoch": 2437, "lr": 9.644091710941361e-06, "val_loss": 8.775015830993652}
{"train_loss": 0.030072597786784172, "global_step": 216982, "epoch": 2438, "lr": 9.643749443002842e-06}
{"train_loss": 0.026195108890533447, "global_step": 216983, "epoch": 2438, "lr": 9.643407180489716e-06}
{"train_loss": 0.14558061957359314, "global_step": 216984, "epoch": 2438, "lr": 9.643064923402045e-06}
{"train_loss": 0.05310855805873871, "global_step": 216985, "epoch": 2438, "lr": 9.642722671739884e-06}
{"train_loss": 0.0417422316968441, "global_step": 216986, "epoch": 2438, "lr": 9.642380425503267e-06}
{"train_loss": 0.05612530559301376, "global_step": 216987, "epoch": 2438, "lr": 9.642038184692226e-06}
{"train_loss": 0.0637970045208931, "global_step": 216988, "epoch": 2438, "lr": 9.641695949306838e-06}
{"train_loss": 0.032624147832393646, "global_step": 216989, "epoch": 2438, "lr": 9.641353719347118e-06}
{"train_loss": 0.0751919373869896, "global_step": 216990, "epoch": 2438, "lr": 9.641011494813134e-06}
{"train_loss": 0.019922545179724693, "global_step": 216991, "epoch": 2438, "lr": 9.640669275704927e-06}
{"train_loss": 0.04295223951339722, "global_step": 216992, "epoch": 2438, "lr": 9.640327062022531e-06}
{"train_loss": 0.05404140427708626, "global_step": 216993, "epoch": 2438, "lr": 9.63998485376601e-06}
{"train_loss": 0.02725929208099842, "global_step": 216994, "epoch": 2438, "lr": 9.639642650935387e-06}
{"train_loss": 0.050686657428741455, "global_step": 216995, "epoch": 2438, "lr": 9.639300453530736e-06}
{"train_loss": 0.05557382106781006, "global_step": 216996, "epoch": 2438, "lr": 9.638958261552079e-06}
{"train_loss": 0.055955372750759125, "global_step": 216997, "epoch": 2438, "lr": 9.638616074999479e-06}
{"train_loss": 0.0546867810189724, "global_step": 216998, "epoch": 2438, "lr": 9.638273893872968e-06}
{"train_loss": 0.024329153820872307, "global_step": 216999, "epoch": 2438, "lr": 9.63793171817261e-06}
{"train_loss": 0.04228942468762398, "global_step": 217000, "epoch": 2438, "lr": 9.63758954789843e-06}
{"train_loss": 0.04946211725473404, "global_step": 217001, "epoch": 2438, "lr": 9.637247383050491e-06}
{"train_loss": 0.019805915653705597, "global_step": 217002, "epoch": 2438, "lr": 9.636905223628822e-06}
{"train_loss": 0.03575143963098526, "global_step": 217003, "epoch": 2438, "lr": 9.636563069633481e-06}
{"train_loss": 0.054469034075737, "global_step": 217004, "epoch": 2438, "lr": 9.636220921064525e-06}
{"train_loss": 0.031089844182133675, "global_step": 217005, "epoch": 2438, "lr": 9.635878777921976e-06}
{"train_loss": 0.06545502692461014, "global_step": 217006, "epoch": 2438, "lr": 9.6355366402059e-06}
{"train_loss": 0.039492666721343994, "global_step": 217007, "epoch": 2438, "lr": 9.635194507916323e-06}
{"train_loss": 0.038549862802028656, "global_step": 217008, "epoch": 2438, "lr": 9.634852381053311e-06}
{"train_loss": 0.09435822069644928, "global_step": 217009, "epoch": 2438, "lr": 9.63451025961689e-06}
{"train_loss": 0.03989630565047264, "global_step": 217010, "epoch": 2438, "lr": 9.634168143607136e-06}
{"train_loss": 0.03629400581121445, "global_step": 217011, "epoch": 2438, "lr": 9.633826033024056e-06}
{"train_loss": 0.06212354451417923, "global_step": 217012, "epoch": 2438, "lr": 9.633483927867731e-06}
{"train_loss": 0.10453657060861588, "global_step": 217013, "epoch": 2438, "lr": 9.633141828138181e-06}
{"train_loss": 0.04177550971508026, "global_step": 217014, "epoch": 2438, "lr": 9.632799733835479e-06}
{"train_loss": 0.045340705662965775, "global_step": 217015, "epoch": 2438, "lr": 9.632457644959636e-06}
{"train_loss": 0.037794098258018494, "global_step": 217016, "epoch": 2438, "lr": 9.632115561510734e-06}
{"train_loss": 0.0348934531211853, "global_step": 217017, "epoch": 2438, "lr": 9.631773483488787e-06}
{"train_loss": 0.08570481836795807, "global_step": 217018, "epoch": 2438, "lr": 9.63143141089387e-06}
{"train_loss": 0.03154934570193291, "global_step": 217019, "epoch": 2438, "lr": 9.631089343726002e-06}
{"train_loss": 0.08501832187175751, "global_step": 217020, "epoch": 2438, "lr": 9.630747281985236e-06}
{"train_loss": 0.05321884527802467, "global_step": 217021, "epoch": 2438, "lr": 9.630405225671646e-06}
{"train_loss": 0.062372662127017975, "global_step": 217022, "epoch": 2438, "lr": 9.630063174785237e-06}
{"train_loss": 0.046038269996643066, "global_step": 217023, "epoch": 2438, "lr": 9.629721129326091e-06}
{"train_loss": 0.05900375917553902, "global_step": 217024, "epoch": 2438, "lr": 9.629379089294232e-06}
{"train_loss": 0.044022150337696075, "global_step": 217025, "epoch": 2438, "lr": 9.629037054689705e-06}
{"train_loss": 0.0647311732172966, "global_step": 217026, "epoch": 2438, "lr": 9.62869502551257e-06}
{"train_loss": 0.03945130482316017, "global_step": 217027, "epoch": 2438, "lr": 9.628353001762852e-06}
{"train_loss": 0.12278398871421814, "global_step": 217028, "epoch": 2438, "lr": 9.628010983440628e-06}
{"train_loss": 0.0680709257721901, "global_step": 217029, "epoch": 2438, "lr": 9.627668970545916e-06}
{"train_loss": 0.02762601338326931, "global_step": 217030, "epoch": 2438, "lr": 9.627326963078765e-06}
{"train_loss": 0.051905252039432526, "global_step": 217031, "epoch": 2438, "lr": 9.626984961039242e-06}
{"train_loss": 0.03851659595966339, "global_step": 217032, "epoch": 2438, "lr": 9.626642964427362e-06}
{"train_loss": 0.06053581088781357, "global_step": 217033, "epoch": 2438, "lr": 9.626300973243207e-06}
{"train_loss": 0.09404032677412033, "global_step": 217034, "epoch": 2438, "lr": 9.625958987486782e-06}
{"train_loss": 0.06637093424797058, "global_step": 217035, "epoch": 2438, "lr": 9.625617007158177e-06}
{"train_loss": 0.07622458040714264, "global_step": 217036, "epoch": 2438, "lr": 9.625275032257397e-06}
{"train_loss": 0.038652654737234116, "global_step": 217037, "epoch": 2438, "lr": 9.624933062784518e-06}
{"train_loss": 0.034123510122299194, "global_step": 217038, "epoch": 2438, "lr": 9.624591098739565e-06}
{"train_loss": 0.07812961935997009, "global_step": 217039, "epoch": 2438, "lr": 9.624249140122593e-06}
{"train_loss": 0.043524593114852905, "global_step": 217040, "epoch": 2438, "lr": 9.623907186933661e-06}
{"train_loss": 0.08643374592065811, "global_step": 217041, "epoch": 2438, "lr": 9.623565239172793e-06}
{"train_loss": 0.06078142672777176, "global_step": 217042, "epoch": 2438, "lr": 9.623223296840056e-06}
{"train_loss": 0.024655353277921677, "global_step": 217043, "epoch": 2438, "lr": 9.62288135993547e-06}
{"train_loss": 0.05850900709629059, "global_step": 217044, "epoch": 2438, "lr": 9.622539428459109e-06}
{"train_loss": 0.03118618205189705, "global_step": 217045, "epoch": 2438, "lr": 9.62219750241099e-06}
{"train_loss": 0.015751605853438377, "global_step": 217046, "epoch": 2438, "lr": 9.621855581791189e-06}
{"train_loss": 0.061393868178129196, "global_step": 217047, "epoch": 2438, "lr": 9.621513666599723e-06}
{"train_loss": 0.045811351388692856, "global_step": 217048, "epoch": 2438, "lr": 9.621171756836672e-06}
{"train_loss": 0.032695528119802475, "global_step": 217049, "epoch": 2438, "lr": 9.620829852502045e-06}
{"train_loss": 0.10556570440530777, "global_step": 217050, "epoch": 2438, "lr": 9.620487953595914e-06}
{"train_loss": 0.0944250077009201, "global_step": 217051, "epoch": 2438, "lr": 9.620146060118306e-06}
{"train_loss": 0.04767559468746185, "global_step": 217052, "epoch": 2438, "lr": 9.619804172069292e-06}
{"train_loss": 0.057484034448862076, "global_step": 217053, "epoch": 2438, "lr": 9.619462289448894e-06}
{"train_loss": 0.05429280921816826, "global_step": 217054, "epoch": 2438, "lr": 9.619120412257172e-06}
{"train_loss": 0.10013771802186966, "global_step": 217055, "epoch": 2438, "lr": 9.618778540494161e-06}
{"train_loss": 0.023591551929712296, "global_step": 217056, "epoch": 2438, "lr": 9.618436674159909e-06}
{"train_loss": 0.04921222850680351, "global_step": 217057, "epoch": 2438, "lr": 9.61809481325448e-06}
{"train_loss": 0.07925798743963242, "global_step": 217058, "epoch": 2438, "lr": 9.617752957777892e-06}
{"train_loss": 0.07301939278841019, "global_step": 217059, "epoch": 2438, "lr": 9.617411107730218e-06}
{"train_loss": 0.02802838385105133, "global_step": 217060, "epoch": 2438, "lr": 9.617069263111478e-06}
{"train_loss": 0.036065056920051575, "global_step": 217061, "epoch": 2438, "lr": 9.616727423921744e-06}
{"train_loss": 0.04952773079276085, "global_step": 217062, "epoch": 2438, "lr": 9.616385590161048e-06}
{"train_loss": 0.05902034044265747, "global_step": 217063, "epoch": 2438, "lr": 9.616043761829424e-06}
{"train_loss": 0.0355934202671051, "global_step": 217064, "epoch": 2438, "lr": 9.615701938926946e-06}
{"train_loss": 0.037997715175151825, "global_step": 217065, "epoch": 2438, "lr": 9.615360121453632e-06}
{"train_loss": 0.022709503769874573, "global_step": 217066, "epoch": 2438, "lr": 9.615018309409551e-06}
{"train_loss": 0.06830751895904541, "global_step": 217067, "epoch": 2438, "lr": 9.61467650279474e-06}
{"train_loss": 0.04994718357920647, "global_step": 217068, "epoch": 2438, "lr": 9.614334701609228e-06}
{"train_loss": 0.041870374232530594, "global_step": 217069, "epoch": 2438, "lr": 9.613992905853092e-06}
{"train_loss": 0.05284370787525445, "global_step": 217070, "epoch": 2438, "lr": 9.613651115526351e-06, "val_loss": 8.839493751525879}
{"train_loss": 0.0661262571811676, "global_step": 217071, "epoch": 2439, "lr": 9.613309330629072e-06}
{"train_loss": 0.08036623895168304, "global_step": 217072, "epoch": 2439, "lr": 9.612967551161278e-06}
{"train_loss": 0.08138925582170486, "global_step": 217073, "epoch": 2439, "lr": 9.61262577712304e-06}
{"train_loss": 0.045755527913570404, "global_step": 217074, "epoch": 2439, "lr": 9.612284008514383e-06}
{"train_loss": 0.04927879944443703, "global_step": 217075, "epoch": 2439, "lr": 9.611942245335365e-06}
{"train_loss": 0.011185084469616413, "global_step": 217076, "epoch": 2439, "lr": 9.611600487586037e-06}
{"train_loss": 0.07315707206726074, "global_step": 217077, "epoch": 2439, "lr": 9.611258735266425e-06}
{"train_loss": 0.05162933096289635, "global_step": 217078, "epoch": 2439, "lr": 9.6109169883766e-06}
{"train_loss": 0.04441641643643379, "global_step": 217079, "epoch": 2439, "lr": 9.610575246916581e-06}
{"train_loss": 0.023897036910057068, "global_step": 217080, "epoch": 2439, "lr": 9.610233510886447e-06}
{"train_loss": 0.017695853486657143, "global_step": 217081, "epoch": 2439, "lr": 9.609891780286202e-06}
{"train_loss": 0.044106610119342804, "global_step": 217082, "epoch": 2439, "lr": 9.609550055115934e-06}
{"train_loss": 0.01389388833194971, "global_step": 217083, "epoch": 2439, "lr": 9.609208335375658e-06}
{"train_loss": 0.041243404150009155, "global_step": 217084, "epoch": 2439, "lr": 9.608866621065437e-06}
{"train_loss": 0.05029592290520668, "global_step": 217085, "epoch": 2439, "lr": 9.608524912185307e-06}
{"train_loss": 0.028244175016880035, "global_step": 217086, "epoch": 2439, "lr": 9.608183208735328e-06}
{"train_loss": 0.08110504597425461, "global_step": 217087, "epoch": 2439, "lr": 9.60784151071552e-06}
{"train_loss": 0.05565357953310013, "global_step": 217088, "epoch": 2439, "lr": 9.607499818125964e-06}
{"train_loss": 0.07003986090421677, "global_step": 217089, "epoch": 2439, "lr": 9.607158130966676e-06}
{"train_loss": 0.044100966304540634, "global_step": 217090, "epoch": 2439, "lr": 9.606816449237721e-06}
{"train_loss": 0.03968854248523712, "global_step": 217091, "epoch": 2439, "lr": 9.606474772939122e-06}
{"train_loss": 0.059557076543569565, "global_step": 217092, "epoch": 2439, "lr": 9.606133102070958e-06}
{"train_loss": 0.03672624006867409, "global_step": 217093, "epoch": 2439, "lr": 9.605791436633244e-06}
{"train_loss": 0.06062588468194008, "global_step": 217094, "epoch": 2439, "lr": 9.605449776626041e-06}
{"train_loss": 0.05080527439713478, "global_step": 217095, "epoch": 2439, "lr": 9.6051081220494e-06}
{"train_loss": 0.04851097613573074, "global_step": 217096, "epoch": 2439, "lr": 9.604766472903348e-06}
{"train_loss": 0.0612761452794075, "global_step": 217097, "epoch": 2439, "lr": 9.604424829187958e-06}
{"train_loss": 0.0669996589422226, "global_step": 217098, "epoch": 2439, "lr": 9.60408319090325e-06}
{"train_loss": 0.05221715196967125, "global_step": 217099, "epoch": 2439, "lr": 9.60374155804929e-06}
{"train_loss": 0.04135465249419212, "global_step": 217100, "epoch": 2439, "lr": 9.6033999306261e-06}
{"train_loss": 0.07858475297689438, "global_step": 217101, "epoch": 2439, "lr": 9.60305830863376e-06}
{"train_loss": 0.07276744395494461, "global_step": 217102, "epoch": 2439, "lr": 9.602716692072289e-06}
{"train_loss": 0.08155566453933716, "global_step": 217103, "epoch": 2439, "lr": 9.602375080941733e-06}
{"train_loss": 0.050841618329286575, "global_step": 217104, "epoch": 2439, "lr": 9.602033475242156e-06}
{"train_loss": 0.04140680655837059, "global_step": 217105, "epoch": 2439, "lr": 9.601691874973578e-06}
{"train_loss": 0.07003498077392578, "global_step": 217106, "epoch": 2439, "lr": 9.601350280136079e-06}
{"train_loss": 0.057889655232429504, "global_step": 217107, "epoch": 2439, "lr": 9.601008690729685e-06}
{"train_loss": 0.055211275815963745, "global_step": 217108, "epoch": 2439, "lr": 9.600667106754424e-06}
{"train_loss": 0.03642457723617554, "global_step": 217109, "epoch": 2439, "lr": 9.60032552821038e-06}
{"train_loss": 0.06953901797533035, "global_step": 217110, "epoch": 2439, "lr": 9.599983955097563e-06}
{"train_loss": 0.03029879927635193, "global_step": 217111, "epoch": 2439, "lr": 9.599642387416036e-06}
{"train_loss": 0.024828894063830376, "global_step": 217112, "epoch": 2439, "lr": 9.599300825165858e-06}
{"train_loss": 0.016372790560126305, "global_step": 217113, "epoch": 2439, "lr": 9.598959268347051e-06}
{"train_loss": 0.05684478208422661, "global_step": 217114, "epoch": 2439, "lr": 9.598617716959684e-06}
{"train_loss": 0.05689712241292, "global_step": 217115, "epoch": 2439, "lr": 9.598276171003772e-06}
{"train_loss": 0.04185612127184868, "global_step": 217116, "epoch": 2439, "lr": 9.597934630479399e-06}
{"train_loss": 0.03961542248725891, "global_step": 217117, "epoch": 2439, "lr": 9.597593095386576e-06}
{"train_loss": 0.11166571080684662, "global_step": 217118, "epoch": 2439, "lr": 9.597251565725379e-06}
{"train_loss": 0.07494820654392242, "global_step": 217119, "epoch": 2439, "lr": 9.59691004149582e-06}
{"train_loss": 0.086497463285923, "global_step": 217120, "epoch": 2439, "lr": 9.596568522697985e-06}
{"train_loss": 0.05258265137672424, "global_step": 217121, "epoch": 2439, "lr": 9.596227009331882e-06}
{"train_loss": 0.07846245914697647, "global_step": 217122, "epoch": 2439, "lr": 9.595885501397583e-06}
{"train_loss": 0.015439013950526714, "global_step": 217123, "epoch": 2439, "lr": 9.595543998895119e-06}
{"train_loss": 0.09819153696298599, "global_step": 217124, "epoch": 2439, "lr": 9.595202501824552e-06}
{"train_loss": 0.04670456424355507, "global_step": 217125, "epoch": 2439, "lr": 9.594861010185902e-06}
{"train_loss": 0.04836881533265114, "global_step": 217126, "epoch": 2439, "lr": 9.594519523979245e-06}
{"train_loss": 0.05491500347852707, "global_step": 217127, "epoch": 2439, "lr": 9.5941780432046e-06}
{"train_loss": 0.07182886451482773, "global_step": 217128, "epoch": 2439, "lr": 9.593836567862036e-06}
{"train_loss": 0.08989100903272629, "global_step": 217129, "epoch": 2439, "lr": 9.593495097951576e-06}
{"train_loss": 0.026533376425504684, "global_step": 217130, "epoch": 2439, "lr": 9.593153633473284e-06}
{"train_loss": 0.04746519401669502, "global_step": 217131, "epoch": 2439, "lr": 9.592812174427206e-06}
{"train_loss": 0.024917563423514366, "global_step": 217132, "epoch": 2439, "lr": 9.592470720813374e-06}
{"train_loss": 0.031533364206552505, "global_step": 217133, "epoch": 2439, "lr": 9.59212927263185e-06}
{"train_loss": 0.08329544961452484, "global_step": 217134, "epoch": 2439, "lr": 9.59178782988266e-06}
{"train_loss": 0.06730574369430542, "global_step": 217135, "epoch": 2439, "lr": 9.591446392565878e-06}
{"train_loss": 0.08404109627008438, "global_step": 217136, "epoch": 2439, "lr": 9.591104960681512e-06}
{"train_loss": 0.07260493189096451, "global_step": 217137, "epoch": 2439, "lr": 9.590763534229652e-06}
{"train_loss": 0.027019077911973, "global_step": 217138, "epoch": 2439, "lr": 9.590422113210306e-06}
{"train_loss": 0.07320386916399002, "global_step": 217139, "epoch": 2439, "lr": 9.590080697623544e-06}
{"train_loss": 0.07692107558250427, "global_step": 217140, "epoch": 2439, "lr": 9.589739287469407e-06}
{"train_loss": 0.04922711104154587, "global_step": 217141, "epoch": 2439, "lr": 9.589397882747924e-06}
{"train_loss": 0.10780242085456848, "global_step": 217142, "epoch": 2439, "lr": 9.589056483459163e-06}
{"train_loss": 0.07186195999383926, "global_step": 217143, "epoch": 2439, "lr": 9.588715089603151e-06}
{"train_loss": 0.05429675430059433, "global_step": 217144, "epoch": 2439, "lr": 9.58837370117996e-06}
{"train_loss": 0.06953810155391693, "global_step": 217145, "epoch": 2439, "lr": 9.588032318189615e-06}
{"train_loss": 0.10054133832454681, "global_step": 217146, "epoch": 2439, "lr": 9.587690940632155e-06}
{"train_loss": 0.027654079720377922, "global_step": 217147, "epoch": 2439, "lr": 9.587349568507636e-06}
{"train_loss": 0.10297468304634094, "global_step": 217148, "epoch": 2439, "lr": 9.587008201816117e-06}
{"train_loss": 0.08066996186971664, "global_step": 217149, "epoch": 2439, "lr": 9.586666840557624e-06}
{"train_loss": 0.03481553867459297, "global_step": 217150, "epoch": 2439, "lr": 9.586325484732222e-06}
{"train_loss": 0.022646114230155945, "global_step": 217151, "epoch": 2439, "lr": 9.585984134339931e-06}
{"train_loss": 0.0949658676981926, "global_step": 217152, "epoch": 2439, "lr": 9.585642789380827e-06}
{"train_loss": 0.045694056898355484, "global_step": 217153, "epoch": 2439, "lr": 9.585301449854928e-06}
{"train_loss": 0.06002728268504143, "global_step": 217154, "epoch": 2439, "lr": 9.584960115762303e-06}
{"train_loss": 0.05512112379074097, "global_step": 217155, "epoch": 2439, "lr": 9.584618787102978e-06}
{"train_loss": 0.054646700620651245, "global_step": 217156, "epoch": 2439, "lr": 9.584277463877017e-06}
{"train_loss": 0.037022463977336884, "global_step": 217157, "epoch": 2439, "lr": 9.583936146084449e-06}
{"train_loss": 0.059999287128448486, "global_step": 217158, "epoch": 2439, "lr": 9.583594833725334e-06}
{"train_loss": 0.056336783594713453, "global_step": 217159, "epoch": 2439, "lr": 9.583253526799707e-06, "val_loss": 8.789473533630371}
{"train_loss": 0.03955437242984772, "global_step": 217160, "epoch": 2440, "lr": 9.582912225307627e-06}
{"train_loss": 0.04706153646111488, "global_step": 217161, "epoch": 2440, "lr": 9.582570929249125e-06}
{"train_loss": 0.06128987669944763, "global_step": 217162, "epoch": 2440, "lr": 9.582229638624262e-06}
{"train_loss": 0.08046887814998627, "global_step": 217163, "epoch": 2440, "lr": 9.58188835343306e-06}
{"train_loss": 0.046286892145872116, "global_step": 217164, "epoch": 2440, "lr": 9.581547073675602e-06}
{"train_loss": 0.023249788209795952, "global_step": 217165, "epoch": 2440, "lr": 9.581205799351888e-06}
{"train_loss": 0.01524126622825861, "global_step": 217166, "epoch": 2440, "lr": 9.580864530461996e-06}
{"train_loss": 0.06856951117515564, "global_step": 217167, "epoch": 2440, "lr": 9.580523267005975e-06}
{"train_loss": 0.03748142346739769, "global_step": 217168, "epoch": 2440, "lr": 9.580182008983851e-06}
{"train_loss": 0.07713568210601807, "global_step": 217169, "epoch": 2440, "lr": 9.579840756395686e-06}
{"train_loss": 0.03377996012568474, "global_step": 217170, "epoch": 2440, "lr": 9.579499509241508e-06}
{"train_loss": 0.04248543828725815, "global_step": 217171, "epoch": 2440, "lr": 9.579158267521387e-06}
{"train_loss": 0.030523626133799553, "global_step": 217172, "epoch": 2440, "lr": 9.578817031235343e-06}
{"train_loss": 0.09055023640394211, "global_step": 217173, "epoch": 2440, "lr": 9.578475800383446e-06}
{"train_loss": 0.05746128037571907, "global_step": 217174, "epoch": 2440, "lr": 9.57813457496572e-06}
{"train_loss": 0.056842874735593796, "global_step": 217175, "epoch": 2440, "lr": 9.577793354982229e-06}
{"train_loss": 0.04563473165035248, "global_step": 217176, "epoch": 2440, "lr": 9.577452140433002e-06}
{"train_loss": 0.06365972012281418, "global_step": 217177, "epoch": 2440, "lr": 9.577110931318107e-06}
{"train_loss": 0.07914026826620102, "global_step": 217178, "epoch": 2440, "lr": 9.576769727637575e-06}
{"train_loss": 0.023683104664087296, "global_step": 217179, "epoch": 2440, "lr": 9.57642852939144e-06}
{"train_loss": 0.03952682018280029, "global_step": 217180, "epoch": 2440, "lr": 9.576087336579776e-06}
{"train_loss": 0.05380503460764885, "global_step": 217181, "epoch": 2440, "lr": 9.575746149202602e-06}
{"train_loss": 0.04235207661986351, "global_step": 217182, "epoch": 2440, "lr": 9.575404967259988e-06}
{"train_loss": 0.07709105312824249, "global_step": 217183, "epoch": 2440, "lr": 9.575063790751953e-06}
{"train_loss": 0.03798248991370201, "global_step": 217184, "epoch": 2440, "lr": 9.574722619678578e-06}
{"train_loss": 0.05294978246092796, "global_step": 217185, "epoch": 2440, "lr": 9.574381454039866e-06}
{"train_loss": 0.07403657585382462, "global_step": 217186, "epoch": 2440, "lr": 9.574040293835907e-06}
{"train_loss": 0.03458919748663902, "global_step": 217187, "epoch": 2440, "lr": 9.573699139066711e-06}
{"train_loss": 0.06913308799266815, "global_step": 217188, "epoch": 2440, "lr": 9.573357989732345e-06}
{"train_loss": 0.07461874186992645, "global_step": 217189, "epoch": 2440, "lr": 9.573016845832844e-06}
{"train_loss": 0.08404303342103958, "global_step": 217190, "epoch": 2440, "lr": 9.572675707368267e-06}
{"train_loss": 0.10081033408641815, "global_step": 217191, "epoch": 2440, "lr": 9.572334574338638e-06}
{"train_loss": 0.02329922653734684, "global_step": 217192, "epoch": 2440, "lr": 9.571993446744026e-06}
{"train_loss": 0.05042693763971329, "global_step": 217193, "epoch": 2440, "lr": 9.571652324584457e-06}
{"train_loss": 0.031926948577165604, "global_step": 217194, "epoch": 2440, "lr": 9.571311207860001e-06}
{"train_loss": 0.06392331421375275, "global_step": 217195, "epoch": 2440, "lr": 9.570970096570675e-06}
{"train_loss": 0.038543589413166046, "global_step": 217196, "epoch": 2440, "lr": 9.570628990716552e-06}
{"train_loss": 0.0638430267572403, "global_step": 217197, "epoch": 2440, "lr": 9.570287890297652e-06}
{"train_loss": 0.07394278794527054, "global_step": 217198, "epoch": 2440, "lr": 9.56994679531405e-06}
{"train_loss": 0.04899857938289642, "global_step": 217199, "epoch": 2440, "lr": 9.56960570576576e-06}
{"train_loss": 0.09055563062429428, "global_step": 217200, "epoch": 2440, "lr": 9.569264621652857e-06}
{"train_loss": 0.03371480107307434, "global_step": 217201, "epoch": 2440, "lr": 9.56892354297536e-06}
{"train_loss": 0.07228315621614456, "global_step": 217202, "epoch": 2440, "lr": 9.568582469733328e-06}
{"train_loss": 0.10361815989017487, "global_step": 217203, "epoch": 2440, "lr": 9.568241401926825e-06}
{"train_loss": 0.029944581910967827, "global_step": 217204, "epoch": 2440, "lr": 9.567900339555858e-06}
{"train_loss": 0.06551367044448853, "global_step": 217205, "epoch": 2440, "lr": 9.567559282620514e-06}
{"train_loss": 0.08759713172912598, "global_step": 217206, "epoch": 2440, "lr": 9.567218231120807e-06}
{"train_loss": 0.08976183086633682, "global_step": 217207, "epoch": 2440, "lr": 9.566877185056799e-06}
{"train_loss": 0.09570332616567612, "global_step": 217208, "epoch": 2440, "lr": 9.566536144428522e-06}
{"train_loss": 0.09262717515230179, "global_step": 217209, "epoch": 2440, "lr": 9.566195109236049e-06}
{"train_loss": 0.03439922630786896, "global_step": 217210, "epoch": 2440, "lr": 9.565854079479392e-06}
{"train_loss": 0.07628334313631058, "global_step": 217211, "epoch": 2440, "lr": 9.565513055158625e-06}
{"train_loss": 0.06732789427042007, "global_step": 217212, "epoch": 2440, "lr": 9.565172036273772e-06}
{"train_loss": 0.05920185521245003, "global_step": 217213, "epoch": 2440, "lr": 9.564831022824899e-06}
{"train_loss": 0.03262762352824211, "global_step": 217214, "epoch": 2440, "lr": 9.564490014812034e-06}
{"train_loss": 0.07016404718160629, "global_step": 217215, "epoch": 2440, "lr": 9.564149012235235e-06}
{"train_loss": 0.09115526080131531, "global_step": 217216, "epoch": 2440, "lr": 9.563808015094539e-06}
{"train_loss": 0.05650021508336067, "global_step": 217217, "epoch": 2440, "lr": 9.563467023389999e-06}
{"train_loss": 0.062040168792009354, "global_step": 217218, "epoch": 2440, "lr": 9.563126037121667e-06}
{"train_loss": 0.07705989480018616, "global_step": 217219, "epoch": 2440, "lr": 9.562785056289564e-06}
{"train_loss": 0.09214729070663452, "global_step": 217220, "epoch": 2440, "lr": 9.562444080893763e-06}
{"train_loss": 0.03756090998649597, "global_step": 217221, "epoch": 2440, "lr": 9.56210311093429e-06}
{"train_loss": 0.11390810459852219, "global_step": 217222, "epoch": 2440, "lr": 9.561762146411207e-06}
{"train_loss": 0.025278355926275253, "global_step": 217223, "epoch": 2440, "lr": 9.561421187324544e-06}
{"train_loss": 0.08685243874788284, "global_step": 217224, "epoch": 2440, "lr": 9.561080233674363e-06}
{"train_loss": 0.057641394436359406, "global_step": 217225, "epoch": 2440, "lr": 9.560739285460691e-06}
{"train_loss": 0.09543820470571518, "global_step": 217226, "epoch": 2440, "lr": 9.560398342683596e-06}
{"train_loss": 0.07004030048847198, "global_step": 217227, "epoch": 2440, "lr": 9.560057405343103e-06}
{"train_loss": 0.04977819696068764, "global_step": 217228, "epoch": 2440, "lr": 9.559716473439279e-06}
{"train_loss": 0.05209947004914284, "global_step": 217229, "epoch": 2440, "lr": 9.559375546972144e-06}
{"train_loss": 0.08486251533031464, "global_step": 217230, "epoch": 2440, "lr": 9.559034625941777e-06}
{"train_loss": 0.021989921107888222, "global_step": 217231, "epoch": 2440, "lr": 9.558693710348182e-06}
{"train_loss": 0.021700970828533173, "global_step": 217232, "epoch": 2440, "lr": 9.55835280019145e-06}
{"train_loss": 0.05338781699538231, "global_step": 217233, "epoch": 2440, "lr": 9.558011895471586e-06}
{"train_loss": 0.05696815252304077, "global_step": 217234, "epoch": 2440, "lr": 9.557670996188668e-06}
{"train_loss": 0.10865885019302368, "global_step": 217235, "epoch": 2440, "lr": 9.557330102342715e-06}
{"train_loss": 0.04290325939655304, "global_step": 217236, "epoch": 2440, "lr": 9.556989213933804e-06}
{"train_loss": 0.045581307262182236, "global_step": 217237, "epoch": 2440, "lr": 9.556648330961943e-06}
{"train_loss": 0.06065314635634422, "global_step": 217238, "epoch": 2440, "lr": 9.556307453427199e-06}
{"train_loss": 0.04897715523838997, "global_step": 217239, "epoch": 2440, "lr": 9.555966581329633e-06}
{"train_loss": 0.01999245025217533, "global_step": 217240, "epoch": 2440, "lr": 9.555625714669259e-06}
{"train_loss": 0.028054865077137947, "global_step": 217241, "epoch": 2440, "lr": 9.55528485344615e-06}
{"train_loss": 0.09279494732618332, "global_step": 217242, "epoch": 2440, "lr": 9.554943997660332e-06}
{"train_loss": 0.06648470461368561, "global_step": 217243, "epoch": 2440, "lr": 9.554603147311869e-06}
{"train_loss": 0.048097409307956696, "global_step": 217244, "epoch": 2440, "lr": 9.55426230240078e-06}
{"train_loss": 0.046875324100255966, "global_step": 217245, "epoch": 2440, "lr": 9.553921462927146e-06}
{"train_loss": 0.05446116253733635, "global_step": 217246, "epoch": 2440, "lr": 9.553580628890979e-06}
{"train_loss": 0.024226056411862373, "global_step": 217247, "epoch": 2440, "lr": 9.553239800292352e-06}
{"train_loss": 0.058137560692312344, "global_step": 217248, "epoch": 2440, "lr": 9.552898977131286e-06, "val_loss": 8.63136100769043, "train_action_mse_error": 9.867977142333984}
{"train_loss": 0.04564886540174484, "global_step": 217249, "epoch": 2441, "lr": 9.552558159407854e-06}
{"train_loss": 0.06992442160844803, "global_step": 217250, "epoch": 2441, "lr": 9.552217347122072e-06}
{"train_loss": 0.027300070971250534, "global_step": 217251, "epoch": 2441, "lr": 9.551876540274018e-06}
{"train_loss": 0.04983213171362877, "global_step": 217252, "epoch": 2441, "lr": 9.55153573886371e-06}
{"train_loss": 0.07661452144384384, "global_step": 217253, "epoch": 2441, "lr": 9.551194942891211e-06}
{"train_loss": 0.06072733551263809, "global_step": 217254, "epoch": 2441, "lr": 9.550854152356553e-06}
{"train_loss": 0.040707673877477646, "global_step": 217255, "epoch": 2441, "lr": 9.5505133672598e-06}
{"train_loss": 0.031030569225549698, "global_step": 217256, "epoch": 2441, "lr": 9.550172587600991e-06}
{"train_loss": 0.02683955430984497, "global_step": 217257, "epoch": 2441, "lr": 9.54983181338015e-06}
{"train_loss": 0.10021194070577621, "global_step": 217258, "epoch": 2441, "lr": 9.549491044597358e-06}
{"train_loss": 0.03876363858580589, "global_step": 217259, "epoch": 2441, "lr": 9.549150281252633e-06}
{"train_loss": 0.027010569348931313, "global_step": 217260, "epoch": 2441, "lr": 9.548809523346042e-06}
{"train_loss": 0.05979663133621216, "global_step": 217261, "epoch": 2441, "lr": 9.548468770877605e-06}
{"train_loss": 0.052708499133586884, "global_step": 217262, "epoch": 2441, "lr": 9.548128023847402e-06}
{"train_loss": 0.07632603496313095, "global_step": 217263, "epoch": 2441, "lr": 9.547787282255444e-06}
{"train_loss": 0.03195269778370857, "global_step": 217264, "epoch": 2441, "lr": 9.547446546101801e-06}
{"train_loss": 0.06708909571170807, "global_step": 217265, "epoch": 2441, "lr": 9.547105815386503e-06}
{"train_loss": 0.038592591881752014, "global_step": 217266, "epoch": 2441, "lr": 9.546765090109615e-06}
{"train_loss": 0.03304203227162361, "global_step": 217267, "epoch": 2441, "lr": 9.546424370271162e-06}
{"train_loss": 0.042844973504543304, "global_step": 217268, "epoch": 2441, "lr": 9.546083655871213e-06}
{"train_loss": 0.0482928492128849, "global_step": 217269, "epoch": 2441, "lr": 9.54574294690978e-06}
{"train_loss": 0.030973445624113083, "global_step": 217270, "epoch": 2441, "lr": 9.545402243386947e-06}
{"train_loss": 0.06870675086975098, "global_step": 217271, "epoch": 2441, "lr": 9.545061545302725e-06}
{"train_loss": 0.0508815161883831, "global_step": 217272, "epoch": 2441, "lr": 9.54472085265719e-06}
{"train_loss": 0.0726907029747963, "global_step": 217273, "epoch": 2441, "lr": 9.544380165450368e-06}
{"train_loss": 0.10176191478967667, "global_step": 217274, "epoch": 2441, "lr": 9.544039483682305e-06}
{"train_loss": 0.05638987943530083, "global_step": 217275, "epoch": 2441, "lr": 9.543698807353063e-06}
{"train_loss": 0.0590665228664875, "global_step": 217276, "epoch": 2441, "lr": 9.543358136462672e-06}
{"train_loss": 0.056041110306978226, "global_step": 217277, "epoch": 2441, "lr": 9.54301747101119e-06}
{"train_loss": 0.05867264047265053, "global_step": 217278, "epoch": 2441, "lr": 9.542676810998646e-06}
{"train_loss": 0.060284655541181564, "global_step": 217279, "epoch": 2441, "lr": 9.542336156425108e-06}
{"train_loss": 0.04706910252571106, "global_step": 217280, "epoch": 2441, "lr": 9.541995507290596e-06}
{"train_loss": 0.0531654991209507, "global_step": 217281, "epoch": 2441, "lr": 9.541654863595184e-06}
{"train_loss": 0.059358254075050354, "global_step": 217282, "epoch": 2441, "lr": 9.541314225338893e-06}
{"train_loss": 0.05635165423154831, "global_step": 217283, "epoch": 2441, "lr": 9.54097359252179e-06}
{"train_loss": 0.0742521584033966, "global_step": 217284, "epoch": 2441, "lr": 9.540632965143898e-06}
{"train_loss": 0.04556197673082352, "global_step": 217285, "epoch": 2441, "lr": 9.540292343205288e-06}
{"train_loss": 0.039599012583494186, "global_step": 217286, "epoch": 2441, "lr": 9.539951726705975e-06}
{"train_loss": 0.034217506647109985, "global_step": 217287, "epoch": 2441, "lr": 9.539611115646036e-06}
{"train_loss": 0.03718600794672966, "global_step": 217288, "epoch": 2441, "lr": 9.539270510025493e-06}
{"train_loss": 0.04515257105231285, "global_step": 217289, "epoch": 2441, "lr": 9.538929909844413e-06}
{"train_loss": 0.04069914668798447, "global_step": 217290, "epoch": 2441, "lr": 9.53858931510282e-06}
{"train_loss": 0.042505454272031784, "global_step": 217291, "epoch": 2441, "lr": 9.538248725800785e-06}
{"train_loss": 0.06243576481938362, "global_step": 217292, "epoch": 2441, "lr": 9.537908141938323e-06}
{"train_loss": 0.03648686408996582, "global_step": 217293, "epoch": 2441, "lr": 9.537567563515509e-06}
{"train_loss": 0.05720677226781845, "global_step": 217294, "epoch": 2441, "lr": 9.537226990532377e-06}
{"train_loss": 0.017271358519792557, "global_step": 217295, "epoch": 2441, "lr": 9.53688642298896e-06}
{"train_loss": 0.06472989916801453, "global_step": 217296, "epoch": 2441, "lr": 9.536545860885326e-06}
{"train_loss": 0.026874450966715813, "global_step": 217297, "epoch": 2441, "lr": 9.536205304221496e-06}
{"train_loss": 0.03716122359037399, "global_step": 217298, "epoch": 2441, "lr": 9.535864752997548e-06}
{"train_loss": 0.0349414199590683, "global_step": 217299, "epoch": 2441, "lr": 9.535524207213493e-06}
{"train_loss": 0.041388027369976044, "global_step": 217300, "epoch": 2441, "lr": 9.535183666869407e-06}
{"train_loss": 0.07372363656759262, "global_step": 217301, "epoch": 2441, "lr": 9.53484313196531e-06}
{"train_loss": 0.059560686349868774, "global_step": 217302, "epoch": 2441, "lr": 9.534502602501271e-06}
{"train_loss": 0.046648528426885605, "global_step": 217303, "epoch": 2441, "lr": 9.534162078477315e-06}
{"train_loss": 0.08651115000247955, "global_step": 217304, "epoch": 2441, "lr": 9.533821559893514e-06}
{"train_loss": 0.04495785012841225, "global_step": 217305, "epoch": 2441, "lr": 9.533481046749875e-06}
{"train_loss": 0.08687130361795425, "global_step": 217306, "epoch": 2441, "lr": 9.533140539046487e-06}
{"train_loss": 0.04242851585149765, "global_step": 217307, "epoch": 2441, "lr": 9.532800036783362e-06}
{"train_loss": 0.16021931171417236, "global_step": 217308, "epoch": 2441, "lr": 9.53245953996057e-06}
{"train_loss": 0.03384162113070488, "global_step": 217309, "epoch": 2441, "lr": 9.53211904857813e-06}
{"train_loss": 0.09063473343849182, "global_step": 217310, "epoch": 2441, "lr": 9.531778562636106e-06}
{"train_loss": 0.04320451617240906, "global_step": 217311, "epoch": 2441, "lr": 9.531438082134553e-06}
{"train_loss": 0.036987677216529846, "global_step": 217312, "epoch": 2441, "lr": 9.531097607073492e-06}
{"train_loss": 0.03921668231487274, "global_step": 217313, "epoch": 2441, "lr": 9.530757137452994e-06}
{"train_loss": 0.05761623755097389, "global_step": 217314, "epoch": 2441, "lr": 9.53041667327308e-06}
{"train_loss": 0.04553726688027382, "global_step": 217315, "epoch": 2441, "lr": 9.530076214533823e-06}
{"train_loss": 0.04543378949165344, "global_step": 217316, "epoch": 2441, "lr": 9.529735761235237e-06}
{"train_loss": 0.078825443983078, "global_step": 217317, "epoch": 2441, "lr": 9.529395313377403e-06}
{"train_loss": 0.06982897967100143, "global_step": 217318, "epoch": 2441, "lr": 9.52905487096033e-06}
{"train_loss": 0.09242944419384003, "global_step": 217319, "epoch": 2441, "lr": 9.528714433984099e-06}
{"train_loss": 0.06762101501226425, "global_step": 217320, "epoch": 2441, "lr": 9.528374002448725e-06}
{"train_loss": 0.04507088661193848, "global_step": 217321, "epoch": 2441, "lr": 9.528033576354283e-06}
{"train_loss": 0.029485125094652176, "global_step": 217322, "epoch": 2441, "lr": 9.52769315570079e-06}
{"train_loss": 0.1122039407491684, "global_step": 217323, "epoch": 2441, "lr": 9.527352740488315e-06}
{"train_loss": 0.040468811988830566, "global_step": 217324, "epoch": 2441, "lr": 9.527012330716884e-06}
{"train_loss": 0.0656522810459137, "global_step": 217325, "epoch": 2441, "lr": 9.526671926386565e-06}
{"train_loss": 0.0396517813205719, "global_step": 217326, "epoch": 2441, "lr": 9.526331527497378e-06}
{"train_loss": 0.021040165796875954, "global_step": 217327, "epoch": 2441, "lr": 9.525991134049395e-06}
{"train_loss": 0.064071886241436, "global_step": 217328, "epoch": 2441, "lr": 9.525650746042636e-06}
{"train_loss": 0.05761025846004486, "global_step": 217329, "epoch": 2441, "lr": 9.525310363477175e-06}
{"train_loss": 0.023753970861434937, "global_step": 217330, "epoch": 2441, "lr": 9.524969986353033e-06}
{"train_loss": 0.06660553812980652, "global_step": 217331, "epoch": 2441, "lr": 9.524629614670271e-06}
{"train_loss": 0.06002892926335335, "global_step": 217332, "epoch": 2441, "lr": 9.524289248428918e-06}
{"train_loss": 0.09975416213274002, "global_step": 217333, "epoch": 2441, "lr": 9.523948887629048e-06}
{"train_loss": 0.034080177545547485, "global_step": 217334, "epoch": 2441, "lr": 9.52360853227069e-06}
{"train_loss": 0.05413336306810379, "global_step": 217335, "epoch": 2441, "lr": 9.523268182353873e-06}
{"train_loss": 0.03714346885681152, "global_step": 217336, "epoch": 2441, "lr": 9.52292783787867e-06}
{"train_loss": 0.054237326690822506, "global_step": 217337, "epoch": 2441, "lr": 9.52258749884511e-06, "val_loss": 8.81495189666748}
{"train_loss": 0.020763061940670013, "global_step": 217338, "epoch": 2442, "lr": 9.52224716525325e-06}
{"train_loss": 0.045511163771152496, "global_step": 217339, "epoch": 2442, "lr": 9.521906837103123e-06}
{"train_loss": 0.054243408143520355, "global_step": 217340, "epoch": 2442, "lr": 9.521566514394792e-06}
{"train_loss": 0.09417892247438431, "global_step": 217341, "epoch": 2442, "lr": 9.521226197128286e-06}
{"train_loss": 0.071306973695755, "global_step": 217342, "epoch": 2442, "lr": 9.520885885303666e-06}
{"train_loss": 0.02431541495025158, "global_step": 217343, "epoch": 2442, "lr": 9.520545578920952e-06}
{"train_loss": 0.029576389119029045, "global_step": 217344, "epoch": 2442, "lr": 9.520205277980226e-06}
{"train_loss": 0.09691648185253143, "global_step": 217345, "epoch": 2442, "lr": 9.519864982481503e-06}
{"train_loss": 0.05797145143151283, "global_step": 217346, "epoch": 2442, "lr": 9.519524692424853e-06}
{"train_loss": 0.06881581246852875, "global_step": 217347, "epoch": 2442, "lr": 9.519184407810294e-06}
{"train_loss": 0.02243215963244438, "global_step": 217348, "epoch": 2442, "lr": 9.518844128637889e-06}
{"train_loss": 0.027148770168423653, "global_step": 217349, "epoch": 2442, "lr": 9.518503854907696e-06}
{"train_loss": 0.09545031189918518, "global_step": 217350, "epoch": 2442, "lr": 9.518163586619732e-06}
{"train_loss": 0.0579153411090374, "global_step": 217351, "epoch": 2442, "lr": 9.517823323774072e-06}
{"train_loss": 0.06963961571455002, "global_step": 217352, "epoch": 2442, "lr": 9.517483066370735e-06}
{"train_loss": 0.040499623864889145, "global_step": 217353, "epoch": 2442, "lr": 9.51714281440979e-06}
{"train_loss": 0.04953162744641304, "global_step": 217354, "epoch": 2442, "lr": 9.516802567891258e-06}
{"train_loss": 0.041169967502355576, "global_step": 217355, "epoch": 2442, "lr": 9.516462326815212e-06}
{"train_loss": 0.057892125099897385, "global_step": 217356, "epoch": 2442, "lr": 9.516122091181672e-06}
{"train_loss": 0.06025533378124237, "global_step": 217357, "epoch": 2442, "lr": 9.515781860990708e-06}
{"train_loss": 0.0704929456114769, "global_step": 217358, "epoch": 2442, "lr": 9.515441636242345e-06}
{"train_loss": 0.025778790935873985, "global_step": 217359, "epoch": 2442, "lr": 9.515101416936645e-06}
{"train_loss": 0.03395563364028931, "global_step": 217360, "epoch": 2442, "lr": 9.514761203073636e-06}
{"train_loss": 0.07738813012838364, "global_step": 217361, "epoch": 2442, "lr": 9.514420994653383e-06}
{"train_loss": 0.09087986499071121, "global_step": 217362, "epoch": 2442, "lr": 9.514080791675912e-06}
{"train_loss": 0.04207102209329605, "global_step": 217363, "epoch": 2442, "lr": 9.513740594141296e-06}
{"train_loss": 0.05670570582151413, "global_step": 217364, "epoch": 2442, "lr": 9.51340040204955e-06}
{"train_loss": 0.0362161323428154, "global_step": 217365, "epoch": 2442, "lr": 9.513060215400727e-06}
{"train_loss": 0.021076757460832596, "global_step": 217366, "epoch": 2442, "lr": 9.512720034194899e-06}
{"train_loss": 0.03322937712073326, "global_step": 217367, "epoch": 2442, "lr": 9.512379858432091e-06}
{"train_loss": 0.07093183696269989, "global_step": 217368, "epoch": 2442, "lr": 9.512039688112339e-06}
{"train_loss": 0.08048228174448013, "global_step": 217369, "epoch": 2442, "lr": 9.511699523235707e-06}
{"train_loss": 0.026098761707544327, "global_step": 217370, "epoch": 2442, "lr": 9.511359363802224e-06}
{"train_loss": 0.0974690169095993, "global_step": 217371, "epoch": 2442, "lr": 9.511019209811961e-06}
{"train_loss": 0.03850111365318298, "global_step": 217372, "epoch": 2442, "lr": 9.510679061264944e-06}
{"train_loss": 0.01902605965733528, "global_step": 217373, "epoch": 2442, "lr": 9.510338918161211e-06}
{"train_loss": 0.05704597011208534, "global_step": 217374, "epoch": 2442, "lr": 9.509998780500833e-06}
{"train_loss": 0.07627802342176437, "global_step": 217375, "epoch": 2442, "lr": 9.509658648283832e-06}
{"train_loss": 0.06606239080429077, "global_step": 217376, "epoch": 2442, "lr": 9.509318521510269e-06}
{"train_loss": 0.040665313601493835, "global_step": 217377, "epoch": 2442, "lr": 9.50897840018018e-06}
{"train_loss": 0.03495582193136215, "global_step": 217378, "epoch": 2442, "lr": 9.508638284293625e-06}
{"train_loss": 0.01046079583466053, "global_step": 217379, "epoch": 2442, "lr": 9.508298173850622e-06}
{"train_loss": 0.044535618275403976, "global_step": 217380, "epoch": 2442, "lr": 9.507958068851252e-06}
{"train_loss": 0.029678674414753914, "global_step": 217381, "epoch": 2442, "lr": 9.507617969295535e-06}
{"train_loss": 0.08079790323972702, "global_step": 217382, "epoch": 2442, "lr": 9.507277875183535e-06}
{"train_loss": 0.027845963835716248, "global_step": 217383, "epoch": 2442, "lr": 9.506937786515274e-06}
{"train_loss": 0.05585351958870888, "global_step": 217384, "epoch": 2442, "lr": 9.506597703290809e-06}
{"train_loss": 0.0712355226278305, "global_step": 217385, "epoch": 2442, "lr": 9.506257625510206e-06}
{"train_loss": 0.07608772814273834, "global_step": 217386, "epoch": 2442, "lr": 9.505917553173477e-06}
{"train_loss": 0.021581027656793594, "global_step": 217387, "epoch": 2442, "lr": 9.505577486280698e-06}
{"train_loss": 0.06919354200363159, "global_step": 217388, "epoch": 2442, "lr": 9.505237424831887e-06}
{"train_loss": 0.040567152202129364, "global_step": 217389, "epoch": 2442, "lr": 9.50489736882712e-06}
{"train_loss": 0.06633695960044861, "global_step": 217390, "epoch": 2442, "lr": 9.504557318266405e-06}
{"train_loss": 0.02856423892080784, "global_step": 217391, "epoch": 2442, "lr": 9.504217273149829e-06}
{"train_loss": 0.022852914407849312, "global_step": 217392, "epoch": 2442, "lr": 9.503877233477404e-06}
{"train_loss": 0.11427134275436401, "global_step": 217393, "epoch": 2442, "lr": 9.503537199249196e-06}
{"train_loss": 0.027037741616368294, "global_step": 217394, "epoch": 2442, "lr": 9.50319717046524e-06}
{"train_loss": 0.07078918814659119, "global_step": 217395, "epoch": 2442, "lr": 9.502857147125594e-06}
{"train_loss": 0.03867243602871895, "global_step": 217396, "epoch": 2442, "lr": 9.502517129230276e-06}
{"train_loss": 0.07391109317541122, "global_step": 217397, "epoch": 2442, "lr": 9.502177116779371e-06}
{"train_loss": 0.034211963415145874, "global_step": 217398, "epoch": 2442, "lr": 9.501837109772894e-06}
{"train_loss": 0.05102793872356415, "global_step": 217399, "epoch": 2442, "lr": 9.501497108210911e-06}
{"train_loss": 0.03052995353937149, "global_step": 217400, "epoch": 2442, "lr": 9.501157112093446e-06}
{"train_loss": 0.036406468600034714, "global_step": 217401, "epoch": 2442, "lr": 9.500817121420552e-06}
{"train_loss": 0.028783291578292847, "global_step": 217402, "epoch": 2442, "lr": 9.500477136192298e-06}
{"train_loss": 0.012339928187429905, "global_step": 217403, "epoch": 2442, "lr": 9.500137156408695e-06}
{"train_loss": 0.037200987339019775, "global_step": 217404, "epoch": 2442, "lr": 9.49979718206982e-06}
{"train_loss": 0.07428525388240814, "global_step": 217405, "epoch": 2442, "lr": 9.499457213175706e-06}
{"train_loss": 0.06362795829772949, "global_step": 217406, "epoch": 2442, "lr": 9.49911724972638e-06}
{"train_loss": 0.04389144107699394, "global_step": 217407, "epoch": 2442, "lr": 9.498777291721916e-06}
{"train_loss": 0.05804944038391113, "global_step": 217408, "epoch": 2442, "lr": 9.498437339162336e-06}
{"train_loss": 0.08264344930648804, "global_step": 217409, "epoch": 2442, "lr": 9.49809739204771e-06}
{"train_loss": 0.042892973870038986, "global_step": 217410, "epoch": 2442, "lr": 9.49775745037807e-06}
{"train_loss": 0.06441020965576172, "global_step": 217411, "epoch": 2442, "lr": 9.497417514153456e-06}
{"train_loss": 0.09088713675737381, "global_step": 217412, "epoch": 2442, "lr": 9.497077583373925e-06}
{"train_loss": 0.0677136480808258, "global_step": 217413, "epoch": 2442, "lr": 9.496737658039512e-06}
{"train_loss": 0.09186552464962006, "global_step": 217414, "epoch": 2442, "lr": 9.496397738150282e-06}
{"train_loss": 0.04384433850646019, "global_step": 217415, "epoch": 2442, "lr": 9.49605782370625e-06}
{"train_loss": 0.037972040474414825, "global_step": 217416, "epoch": 2442, "lr": 9.495717914707491e-06}
{"train_loss": 0.032833606004714966, "global_step": 217417, "epoch": 2442, "lr": 9.495378011154032e-06}
{"train_loss": 0.0606757253408432, "global_step": 217418, "epoch": 2442, "lr": 9.495038113045934e-06}
{"train_loss": 0.047753848135471344, "global_step": 217419, "epoch": 2442, "lr": 9.494698220383225e-06}
{"train_loss": 0.043026141822338104, "global_step": 217420, "epoch": 2442, "lr": 9.49435833316596e-06}
{"train_loss": 0.08642999082803726, "global_step": 217421, "epoch": 2442, "lr": 9.494018451394198e-06}
{"train_loss": 0.04752276465296745, "global_step": 217422, "epoch": 2442, "lr": 9.493678575067954e-06}
{"train_loss": 0.026674164459109306, "global_step": 217423, "epoch": 2442, "lr": 9.49333870418731e-06}
{"train_loss": 0.052850671112537384, "global_step": 217424, "epoch": 2442, "lr": 9.492998838752277e-06}
{"train_loss": 0.05246494337916374, "global_step": 217425, "epoch": 2442, "lr": 9.49265897876293e-06}
{"train_loss": 0.05182358332689893, "global_step": 217426, "epoch": 2442, "lr": 9.49231912421929e-06, "val_loss": 8.735888481140137}
{"train_loss": 0.05358610674738884, "global_step": 217427, "epoch": 2443, "lr": 9.491979275121422e-06}
{"train_loss": 0.04545614868402481, "global_step": 217428, "epoch": 2443, "lr": 9.491639431469357e-06}
{"train_loss": 0.04895433783531189, "global_step": 217429, "epoch": 2443, "lr": 9.491299593263153e-06}
{"train_loss": 0.05766342952847481, "global_step": 217430, "epoch": 2443, "lr": 9.490959760502844e-06}
{"train_loss": 0.051709387451410294, "global_step": 217431, "epoch": 2443, "lr": 9.490619933188494e-06}
{"train_loss": 0.05784081295132637, "global_step": 217432, "epoch": 2443, "lr": 9.49028011132012e-06}
{"train_loss": 0.08228645473718643, "global_step": 217433, "epoch": 2443, "lr": 9.489940294897798e-06}
{"train_loss": 0.05388698726892471, "global_step": 217434, "epoch": 2443, "lr": 9.489600483921545e-06}
{"train_loss": 0.04996490105986595, "global_step": 217435, "epoch": 2443, "lr": 9.489260678391437e-06}
{"train_loss": 0.10079837590456009, "global_step": 217436, "epoch": 2443, "lr": 9.488920878307494e-06}
{"train_loss": 0.058639079332351685, "global_step": 217437, "epoch": 2443, "lr": 9.488581083669768e-06}
{"train_loss": 0.08191514015197754, "global_step": 217438, "epoch": 2443, "lr": 9.488241294478323e-06}
{"train_loss": 0.05202990770339966, "global_step": 217439, "epoch": 2443, "lr": 9.487901510733176e-06}
{"train_loss": 0.045493233948946, "global_step": 217440, "epoch": 2443, "lr": 9.487561732434398e-06}
{"train_loss": 0.06877272576093674, "global_step": 217441, "epoch": 2443, "lr": 9.487221959582015e-06}
{"train_loss": 0.03250299021601677, "global_step": 217442, "epoch": 2443, "lr": 9.486882192176088e-06}
{"train_loss": 0.09548511356115341, "global_step": 217443, "epoch": 2443, "lr": 9.48654243021665e-06}
{"train_loss": 0.06934850662946701, "global_step": 217444, "epoch": 2443, "lr": 9.486202673703765e-06}
{"train_loss": 0.06343290209770203, "global_step": 217445, "epoch": 2443, "lr": 9.485862922637461e-06}
{"train_loss": 0.08354262262582779, "global_step": 217446, "epoch": 2443, "lr": 9.48552317701778e-06}
{"train_loss": 0.03041021339595318, "global_step": 217447, "epoch": 2443, "lr": 9.485183436844785e-06}
{"train_loss": 0.0329921692609787, "global_step": 217448, "epoch": 2443, "lr": 9.484843702118507e-06}
{"train_loss": 0.03445940092206001, "global_step": 217449, "epoch": 2443, "lr": 9.484503972839004e-06}
{"train_loss": 0.04728212580084801, "global_step": 217450, "epoch": 2443, "lr": 9.484164249006322e-06}
{"train_loss": 0.08825206011533737, "global_step": 217451, "epoch": 2443, "lr": 9.483824530620483e-06}
{"train_loss": 0.0791676938533783, "global_step": 217452, "epoch": 2443, "lr": 9.483484817681564e-06}
{"train_loss": 0.060323309153318405, "global_step": 217453, "epoch": 2443, "lr": 9.483145110189585e-06}
{"train_loss": 0.02962278015911579, "global_step": 217454, "epoch": 2443, "lr": 9.482805408144613e-06}
{"train_loss": 0.04117131233215332, "global_step": 217455, "epoch": 2443, "lr": 9.482465711546673e-06}
{"train_loss": 0.052175138145685196, "global_step": 217456, "epoch": 2443, "lr": 9.482126020395826e-06}
{"train_loss": 0.035452377051115036, "global_step": 217457, "epoch": 2443, "lr": 9.481786334692123e-06}
{"train_loss": 0.03939172998070717, "global_step": 217458, "epoch": 2443, "lr": 9.481446654435583e-06}
{"train_loss": 0.050614822655916214, "global_step": 217459, "epoch": 2443, "lr": 9.481106979626287e-06}
{"train_loss": 0.04159871116280556, "global_step": 217460, "epoch": 2443, "lr": 9.480767310264249e-06}
{"train_loss": 0.03380066528916359, "global_step": 217461, "epoch": 2443, "lr": 9.480427646349538e-06}
{"train_loss": 0.024789683520793915, "global_step": 217462, "epoch": 2443, "lr": 9.480087987882175e-06}
{"train_loss": 0.07029296457767487, "global_step": 217463, "epoch": 2443, "lr": 9.479748334862237e-06}
{"train_loss": 0.028734367340803146, "global_step": 217464, "epoch": 2443, "lr": 9.479408687289743e-06}
{"train_loss": 0.06308192014694214, "global_step": 217465, "epoch": 2443, "lr": 9.479069045164756e-06}
{"train_loss": 0.09395266324281693, "global_step": 217466, "epoch": 2443, "lr": 9.478729408487303e-06}
{"train_loss": 0.1086231991648674, "global_step": 217467, "epoch": 2443, "lr": 9.478389777257452e-06}
{"train_loss": 0.034749358892440796, "global_step": 217468, "epoch": 2443, "lr": 9.478050151475227e-06}
{"train_loss": 0.01580452360212803, "global_step": 217469, "epoch": 2443, "lr": 9.477710531140698e-06}
{"train_loss": 0.11020978540182114, "global_step": 217470, "epoch": 2443, "lr": 9.47737091625388e-06}
{"train_loss": 0.04044218361377716, "global_step": 217471, "epoch": 2443, "lr": 9.477031306814855e-06}
{"train_loss": 0.0746937170624733, "global_step": 217472, "epoch": 2443, "lr": 9.476691702823632e-06}
{"train_loss": 0.048481062054634094, "global_step": 217473, "epoch": 2443, "lr": 9.476352104280284e-06}
{"train_loss": 0.02615603804588318, "global_step": 217474, "epoch": 2443, "lr": 9.47601251118484e-06}
{"train_loss": 0.033867545425891876, "global_step": 217475, "epoch": 2443, "lr": 9.475672923537349e-06}
{"train_loss": 0.09128978848457336, "global_step": 217476, "epoch": 2443, "lr": 9.475333341337872e-06}
{"train_loss": 0.05292735993862152, "global_step": 217477, "epoch": 2443, "lr": 9.474993764586432e-06}
{"train_loss": 0.06934673339128494, "global_step": 217478, "epoch": 2443, "lr": 9.474654193283095e-06}
{"train_loss": 0.06158796697854996, "global_step": 217479, "epoch": 2443, "lr": 9.474314627427889e-06}
{"train_loss": 0.02382219210267067, "global_step": 217480, "epoch": 2443, "lr": 9.473975067020873e-06}
{"train_loss": 0.03541753813624382, "global_step": 217481, "epoch": 2443, "lr": 9.473635512062079e-06}
{"train_loss": 0.0626259595155716, "global_step": 217482, "epoch": 2443, "lr": 9.473295962551576e-06}
{"train_loss": 0.05590293928980827, "global_step": 217483, "epoch": 2443, "lr": 9.472956418489388e-06}
{"train_loss": 0.031619634479284286, "global_step": 217484, "epoch": 2443, "lr": 9.472616879875557e-06}
{"train_loss": 0.06335606426000595, "global_step": 217485, "epoch": 2443, "lr": 9.472277346710151e-06}
{"train_loss": 0.06558053195476532, "global_step": 217486, "epoch": 2443, "lr": 9.471937818993193e-06}
{"train_loss": 0.07858055084943771, "global_step": 217487, "epoch": 2443, "lr": 9.47159829672475e-06}
{"train_loss": 0.030112259089946747, "global_step": 217488, "epoch": 2443, "lr": 9.471258779904856e-06}
{"train_loss": 0.06148418411612511, "global_step": 217489, "epoch": 2443, "lr": 9.470919268533545e-06}
{"train_loss": 0.07576307654380798, "global_step": 217490, "epoch": 2443, "lr": 9.470579762610893e-06}
{"train_loss": 0.04247215390205383, "global_step": 217491, "epoch": 2443, "lr": 9.47024026213691e-06}
{"train_loss": 0.02664250321686268, "global_step": 217492, "epoch": 2443, "lr": 9.469900767111656e-06}
{"train_loss": 0.04803228750824928, "global_step": 217493, "epoch": 2443, "lr": 9.469561277535193e-06}
{"train_loss": 0.04922006279230118, "global_step": 217494, "epoch": 2443, "lr": 9.469221793407545e-06}
{"train_loss": 0.045996006578207016, "global_step": 217495, "epoch": 2443, "lr": 9.468882314728777e-06}
{"train_loss": 0.07383686304092407, "global_step": 217496, "epoch": 2443, "lr": 9.468542841498913e-06}
{"train_loss": 0.02298511005938053, "global_step": 217497, "epoch": 2443, "lr": 9.468203373718021e-06}
{"train_loss": 0.033006682991981506, "global_step": 217498, "epoch": 2443, "lr": 9.467863911386121e-06}
{"train_loss": 0.05275779590010643, "global_step": 217499, "epoch": 2443, "lr": 9.467524454503284e-06}
{"train_loss": 0.04401952400803566, "global_step": 217500, "epoch": 2443, "lr": 9.467185003069534e-06}
{"train_loss": 0.06987076252698898, "global_step": 217501, "epoch": 2443, "lr": 9.466845557084942e-06}
{"train_loss": 0.043898291885852814, "global_step": 217502, "epoch": 2443, "lr": 9.466506116549522e-06}
{"train_loss": 0.07710281759500504, "global_step": 217503, "epoch": 2443, "lr": 9.46616668146335e-06}
{"train_loss": 0.05244113877415657, "global_step": 217504, "epoch": 2443, "lr": 9.465827251826448e-06}
{"train_loss": 0.05684470757842064, "global_step": 217505, "epoch": 2443, "lr": 9.46548782763888e-06}
{"train_loss": 0.09621348977088928, "global_step": 217506, "epoch": 2443, "lr": 9.465148408900671e-06}
{"train_loss": 0.06388983130455017, "global_step": 217507, "epoch": 2443, "lr": 9.464808995611896e-06}
{"train_loss": 0.07276798039674759, "global_step": 217508, "epoch": 2443, "lr": 9.464469587772568e-06}
{"train_loss": 0.089572474360466, "global_step": 217509, "epoch": 2443, "lr": 9.464130185382758e-06}
{"train_loss": 0.01033824309706688, "global_step": 217510, "epoch": 2443, "lr": 9.463790788442489e-06}
{"train_loss": 0.1134895607829094, "global_step": 217511, "epoch": 2443, "lr": 9.463451396951822e-06}
{"train_loss": 0.0502912700176239, "global_step": 217512, "epoch": 2443, "lr": 9.463112010910813e-06}
{"train_loss": 0.05530072748661041, "global_step": 217513, "epoch": 2443, "lr": 9.462772630319478e-06}
{"train_loss": 0.02280338667333126, "global_step": 217514, "epoch": 2443, "lr": 9.462433255177894e-06}
{"train_loss": 0.055474138829145536, "global_step": 217515, "epoch": 2443, "lr": 9.46209388548608e-06, "val_loss": 8.82341480255127}
{"train_loss": 0.04780665040016174, "global_step": 217516, "epoch": 2444, "lr": 9.461754521244103e-06}
{"train_loss": 0.060580190271139145, "global_step": 217517, "epoch": 2444, "lr": 9.461415162451991e-06}
{"train_loss": 0.04014671966433525, "global_step": 217518, "epoch": 2444, "lr": 9.46107580910981e-06}
{"train_loss": 0.011646823957562447, "global_step": 217519, "epoch": 2444, "lr": 9.460736461217579e-06}
{"train_loss": 0.05825623497366905, "global_step": 217520, "epoch": 2444, "lr": 9.460397118775371e-06}
{"train_loss": 0.047150444239377975, "global_step": 217521, "epoch": 2444, "lr": 9.460057781783216e-06}
{"train_loss": 0.06577417999505997, "global_step": 217522, "epoch": 2444, "lr": 9.45971845024115e-06}
{"train_loss": 0.06243794038891792, "global_step": 217523, "epoch": 2444, "lr": 9.459379124149249e-06}
{"train_loss": 0.07308340072631836, "global_step": 217524, "epoch": 2444, "lr": 9.45903980350752e-06}
{"train_loss": 0.04089759290218353, "global_step": 217525, "epoch": 2444, "lr": 9.458700488316047e-06}
{"train_loss": 0.040511488914489746, "global_step": 217526, "epoch": 2444, "lr": 9.458361178574855e-06}
{"train_loss": 0.021461348980665207, "global_step": 217527, "epoch": 2444, "lr": 9.458021874283978e-06}
{"train_loss": 0.08710475265979767, "global_step": 217528, "epoch": 2444, "lr": 9.457682575443478e-06}
{"train_loss": 0.07840652763843536, "global_step": 217529, "epoch": 2444, "lr": 9.45734328205341e-06}
{"train_loss": 0.078050896525383, "global_step": 217530, "epoch": 2444, "lr": 9.457003994113794e-06}
{"train_loss": 0.12402762472629547, "global_step": 217531, "epoch": 2444, "lr": 9.456664711624707e-06}
{"train_loss": 0.05524193495512009, "global_step": 217532, "epoch": 2444, "lr": 9.45632543458616e-06}
{"train_loss": 0.04895281791687012, "global_step": 217533, "epoch": 2444, "lr": 9.455986162998232e-06}
{"train_loss": 0.04101558402180672, "global_step": 217534, "epoch": 2444, "lr": 9.455646896860942e-06}
{"train_loss": 0.0574987530708313, "global_step": 217535, "epoch": 2444, "lr": 9.455307636174354e-06}
{"train_loss": 0.022800810635089874, "global_step": 217536, "epoch": 2444, "lr": 9.454968380938495e-06}
{"train_loss": 0.07235663384199142, "global_step": 217537, "epoch": 2444, "lr": 9.454629131153432e-06}
{"train_loss": 0.08771362900733948, "global_step": 217538, "epoch": 2444, "lr": 9.45428988681919e-06}
{"train_loss": 0.08854592591524124, "global_step": 217539, "epoch": 2444, "lr": 9.453950647935838e-06}
{"train_loss": 0.06784562766551971, "global_step": 217540, "epoch": 2444, "lr": 9.453611414503393e-06}
{"train_loss": 0.04421072080731392, "global_step": 217541, "epoch": 2444, "lr": 9.453272186521927e-06}
{"train_loss": 0.019312631338834763, "global_step": 217542, "epoch": 2444, "lr": 9.452932963991461e-06}
{"train_loss": 0.018932240083813667, "global_step": 217543, "epoch": 2444, "lr": 9.45259374691207e-06}
{"train_loss": 0.047458112239837646, "global_step": 217544, "epoch": 2444, "lr": 9.452254535283772e-06}
{"train_loss": 0.07464105635881424, "global_step": 217545, "epoch": 2444, "lr": 9.451915329106636e-06}
{"train_loss": 0.06575866043567657, "global_step": 217546, "epoch": 2444, "lr": 9.45157612838068e-06}
{"train_loss": 0.07552427053451538, "global_step": 217547, "epoch": 2444, "lr": 9.451236933105967e-06}
{"train_loss": 0.07606086134910583, "global_step": 217548, "epoch": 2444, "lr": 9.450897743282554e-06}
{"train_loss": 0.07418005168437958, "global_step": 217549, "epoch": 2444, "lr": 9.450558558910461e-06}
{"train_loss": 0.04374154657125473, "global_step": 217550, "epoch": 2444, "lr": 9.450219379989761e-06}
{"train_loss": 0.0526108480989933, "global_step": 217551, "epoch": 2444, "lr": 9.449880206520467e-06}
{"train_loss": 0.10102437436580658, "global_step": 217552, "epoch": 2444, "lr": 9.449541038502657e-06}
{"train_loss": 0.0805005356669426, "global_step": 217553, "epoch": 2444, "lr": 9.449201875936353e-06}
{"train_loss": 0.05621091648936272, "global_step": 217554, "epoch": 2444, "lr": 9.448862718821622e-06}
{"train_loss": 0.08134492486715317, "global_step": 217555, "epoch": 2444, "lr": 9.44852356715848e-06}
{"train_loss": 0.0840812548995018, "global_step": 217556, "epoch": 2444, "lr": 9.448184420947004e-06}
{"train_loss": 0.04336164891719818, "global_step": 217557, "epoch": 2444, "lr": 9.447845280187212e-06}
{"train_loss": 0.07687467336654663, "global_step": 217558, "epoch": 2444, "lr": 9.447506144879175e-06}
{"train_loss": 0.0624382421374321, "global_step": 217559, "epoch": 2444, "lr": 9.447167015022917e-06}
{"train_loss": 0.05342007800936699, "global_step": 217560, "epoch": 2444, "lr": 9.446827890618504e-06}
{"train_loss": 0.10121692717075348, "global_step": 217561, "epoch": 2444, "lr": 9.446488771665974e-06}
{"train_loss": 0.05442395433783531, "global_step": 217562, "epoch": 2444, "lr": 9.446149658165354e-06}
{"train_loss": 0.0652899518609047, "global_step": 217563, "epoch": 2444, "lr": 9.445810550116712e-06}
{"train_loss": 0.04871216043829918, "global_step": 217564, "epoch": 2444, "lr": 9.445471447520082e-06}
{"train_loss": 0.06579455733299255, "global_step": 217565, "epoch": 2444, "lr": 9.445132350375524e-06}
{"train_loss": 0.09328024834394455, "global_step": 217566, "epoch": 2444, "lr": 9.44479325868306e-06}
{"train_loss": 0.1036372110247612, "global_step": 217567, "epoch": 2444, "lr": 9.444454172442762e-06}
{"train_loss": 0.13135629892349243, "global_step": 217568, "epoch": 2444, "lr": 9.444115091654655e-06}
{"train_loss": 0.057796280831098557, "global_step": 217569, "epoch": 2444, "lr": 9.4437760163188e-06}
{"train_loss": 0.12246407568454742, "global_step": 217570, "epoch": 2444, "lr": 9.443436946435218e-06}
{"train_loss": 0.052085984498262405, "global_step": 217571, "epoch": 2444, "lr": 9.443097882003993e-06}
{"train_loss": 0.06994780898094177, "global_step": 217572, "epoch": 2444, "lr": 9.442758823025132e-06}
{"train_loss": 0.029907792806625366, "global_step": 217573, "epoch": 2444, "lr": 9.442419769498712e-06}
{"train_loss": 0.04600639268755913, "global_step": 217574, "epoch": 2444, "lr": 9.44208072142475e-06}
{"train_loss": 0.06351879984140396, "global_step": 217575, "epoch": 2444, "lr": 9.441741678803324e-06}
{"train_loss": 0.08510927855968475, "global_step": 217576, "epoch": 2444, "lr": 9.441402641634445e-06}
{"train_loss": 0.06434343755245209, "global_step": 217577, "epoch": 2444, "lr": 9.441063609918183e-06}
{"train_loss": 0.030690032988786697, "global_step": 217578, "epoch": 2444, "lr": 9.44072458365457e-06}
{"train_loss": 0.02321021631360054, "global_step": 217579, "epoch": 2444, "lr": 9.440385562843667e-06}
{"train_loss": 0.0760665088891983, "global_step": 217580, "epoch": 2444, "lr": 9.440046547485499e-06}
{"train_loss": 0.05308153107762337, "global_step": 217581, "epoch": 2444, "lr": 9.439707537580129e-06}
{"train_loss": 0.09501379728317261, "global_step": 217582, "epoch": 2444, "lr": 9.439368533127591e-06}
{"train_loss": 0.058503657579422, "global_step": 217583, "epoch": 2444, "lr": 9.43902953412793e-06}
{"train_loss": 0.03454747423529625, "global_step": 217584, "epoch": 2444, "lr": 9.438690540581214e-06}
{"train_loss": 0.03901506960391998, "global_step": 217585, "epoch": 2444, "lr": 9.438351552487457e-06}
{"train_loss": 0.07902145385742188, "global_step": 217586, "epoch": 2444, "lr": 9.438012569846733e-06}
{"train_loss": 0.0482061468064785, "global_step": 217587, "epoch": 2444, "lr": 9.437673592659057e-06}
{"train_loss": 0.08834274858236313, "global_step": 217588, "epoch": 2444, "lr": 9.43733462092451e-06}
{"train_loss": 0.0391630120575428, "global_step": 217589, "epoch": 2444, "lr": 9.436995654643099e-06}
{"train_loss": 0.0533950999379158, "global_step": 217590, "epoch": 2444, "lr": 9.436656693814911e-06}
{"train_loss": 0.09821035712957382, "global_step": 217591, "epoch": 2444, "lr": 9.436317738439948e-06}
{"train_loss": 0.03957720473408699, "global_step": 217592, "epoch": 2444, "lr": 9.435978788518296e-06}
{"train_loss": 0.059885021299123764, "global_step": 217593, "epoch": 2444, "lr": 9.43563984404997e-06}
{"train_loss": 0.07336060702800751, "global_step": 217594, "epoch": 2444, "lr": 9.435300905035039e-06}
{"train_loss": 0.04857799783349037, "global_step": 217595, "epoch": 2444, "lr": 9.434961971473528e-06}
{"train_loss": 0.07944846153259277, "global_step": 217596, "epoch": 2444, "lr": 9.434623043365497e-06}
{"train_loss": 0.08099967241287231, "global_step": 217597, "epoch": 2444, "lr": 9.434284120710984e-06}
{"train_loss": 0.04256197065114975, "global_step": 217598, "epoch": 2444, "lr": 9.433945203510041e-06}
{"train_loss": 0.040569156408309937, "global_step": 217599, "epoch": 2444, "lr": 9.433606291762715e-06}
{"train_loss": 0.038698356598615646, "global_step": 217600, "epoch": 2444, "lr": 9.43326738546903e-06}
{"train_loss": 0.02958417870104313, "global_step": 217601, "epoch": 2444, "lr": 9.432928484629061e-06}
{"train_loss": 0.0633101686835289, "global_step": 217602, "epoch": 2444, "lr": 9.43258958924283e-06}
{"train_loss": 0.03803412988781929, "global_step": 217603, "epoch": 2444, "lr": 9.432250699310407e-06}
{"train_loss": 0.06147879694871018, "global_step": 217604, "epoch": 2444, "lr": 9.431911814831806e-06, "val_loss": 8.83962631225586}
{"train_loss": 0.05372875556349754, "global_step": 217605, "epoch": 2445, "lr": 9.431572935807104e-06}
{"train_loss": 0.04868894815444946, "global_step": 217606, "epoch": 2445, "lr": 9.431234062236322e-06}
{"train_loss": 0.030813368037343025, "global_step": 217607, "epoch": 2445, "lr": 9.430895194119527e-06}
{"train_loss": 0.05716431513428688, "global_step": 217608, "epoch": 2445, "lr": 9.43055633145674e-06}
{"train_loss": 0.04332432150840759, "global_step": 217609, "epoch": 2445, "lr": 9.430217474248038e-06}
{"train_loss": 0.04955762252211571, "global_step": 217610, "epoch": 2445, "lr": 9.429878622493426e-06}
{"train_loss": 0.04460931196808815, "global_step": 217611, "epoch": 2445, "lr": 9.429539776192997e-06}
{"train_loss": 0.020222144201397896, "global_step": 217612, "epoch": 2445, "lr": 9.42920093534675e-06}
{"train_loss": 0.0465189628303051, "global_step": 217613, "epoch": 2445, "lr": 9.428862099954772e-06}
{"train_loss": 0.0555475652217865, "global_step": 217614, "epoch": 2445, "lr": 9.42852327001707e-06}
{"train_loss": 0.04279680550098419, "global_step": 217615, "epoch": 2445, "lr": 9.428184445533728e-06}
{"train_loss": 0.042388491332530975, "global_step": 217616, "epoch": 2445, "lr": 9.427845626504756e-06}
{"train_loss": 0.04013162851333618, "global_step": 217617, "epoch": 2445, "lr": 9.427506812930226e-06}
{"train_loss": 0.05134989321231842, "global_step": 217618, "epoch": 2445, "lr": 9.42716800481016e-06}
{"train_loss": 0.022375844419002533, "global_step": 217619, "epoch": 2445, "lr": 9.426829202144622e-06}
{"train_loss": 0.07669824361801147, "global_step": 217620, "epoch": 2445, "lr": 9.426490404933663e-06}
{"train_loss": 0.028829745948314667, "global_step": 217621, "epoch": 2445, "lr": 9.426151613177304e-06}
{"train_loss": 0.04576719179749489, "global_step": 217622, "epoch": 2445, "lr": 9.425812826875613e-06}
{"train_loss": 0.04935018718242645, "global_step": 217623, "epoch": 2445, "lr": 9.425474046028621e-06}
{"train_loss": 0.04889802262187004, "global_step": 217624, "epoch": 2445, "lr": 9.425135270636393e-06}
{"train_loss": 0.12199732661247253, "global_step": 217625, "epoch": 2445, "lr": 9.42479650069894e-06}
{"train_loss": 0.04964643344283104, "global_step": 217626, "epoch": 2445, "lr": 9.424457736216347e-06}
{"train_loss": 0.0774187371134758, "global_step": 217627, "epoch": 2445, "lr": 9.424118977188629e-06}
{"train_loss": 0.03366556763648987, "global_step": 217628, "epoch": 2445, "lr": 9.423780223615858e-06}
{"train_loss": 0.035484567284584045, "global_step": 217629, "epoch": 2445, "lr": 9.423441475498046e-06}
{"train_loss": 0.053276434540748596, "global_step": 217630, "epoch": 2445, "lr": 9.423102732835276e-06}
{"train_loss": 0.07334114611148834, "global_step": 217631, "epoch": 2445, "lr": 9.422763995627559e-06}
{"train_loss": 0.034619979560375214, "global_step": 217632, "epoch": 2445, "lr": 9.422425263874973e-06}
{"train_loss": 0.06444259732961655, "global_step": 217633, "epoch": 2445, "lr": 9.422086537577535e-06}
{"train_loss": 0.06354213505983353, "global_step": 217634, "epoch": 2445, "lr": 9.42174781673531e-06}
{"train_loss": 0.0769062489271164, "global_step": 217635, "epoch": 2445, "lr": 9.421409101348328e-06}
{"train_loss": 0.0872824490070343, "global_step": 217636, "epoch": 2445, "lr": 9.421070391416654e-06}
{"train_loss": 0.03399508446455002, "global_step": 217637, "epoch": 2445, "lr": 9.420731686940326e-06}
{"train_loss": 0.055769212543964386, "global_step": 217638, "epoch": 2445, "lr": 9.420392987919368e-06}
{"train_loss": 0.0548931248486042, "global_step": 217639, "epoch": 2445, "lr": 9.420054294353858e-06}
{"train_loss": 0.08651581406593323, "global_step": 217640, "epoch": 2445, "lr": 9.419715606243817e-06}
{"train_loss": 0.03233269229531288, "global_step": 217641, "epoch": 2445, "lr": 9.419376923589312e-06}
{"train_loss": 0.04778432846069336, "global_step": 217642, "epoch": 2445, "lr": 9.419038246390366e-06}
{"train_loss": 0.1246456429362297, "global_step": 217643, "epoch": 2445, "lr": 9.418699574647044e-06}
{"train_loss": 0.10149238258600235, "global_step": 217644, "epoch": 2445, "lr": 9.418360908359374e-06}
{"train_loss": 0.042337000370025635, "global_step": 217645, "epoch": 2445, "lr": 9.418022247527426e-06}
{"train_loss": 0.06833020597696304, "global_step": 217646, "epoch": 2445, "lr": 9.417683592151211e-06}
{"train_loss": 0.03275753930211067, "global_step": 217647, "epoch": 2445, "lr": 9.41734494223081e-06}
{"train_loss": 0.06129264831542969, "global_step": 217648, "epoch": 2445, "lr": 9.41700629776624e-06}
{"train_loss": 0.04625754803419113, "global_step": 217649, "epoch": 2445, "lr": 9.416667658757571e-06}
{"train_loss": 0.05116013437509537, "global_step": 217650, "epoch": 2445, "lr": 9.416329025204823e-06}
{"train_loss": 0.08063197880983353, "global_step": 217651, "epoch": 2445, "lr": 9.415990397108065e-06}
{"train_loss": 0.05921023711562157, "global_step": 217652, "epoch": 2445, "lr": 9.415651774467326e-06}
{"train_loss": 0.03917066752910614, "global_step": 217653, "epoch": 2445, "lr": 9.415313157282668e-06}
{"train_loss": 0.06830134242773056, "global_step": 217654, "epoch": 2445, "lr": 9.414974545554112e-06}
{"train_loss": 0.04482152685523033, "global_step": 217655, "epoch": 2445, "lr": 9.414635939281719e-06}
{"train_loss": 0.03440384194254875, "global_step": 217656, "epoch": 2445, "lr": 9.414297338465549e-06}
{"train_loss": 0.014010842889547348, "global_step": 217657, "epoch": 2445, "lr": 9.413958743105616e-06}
{"train_loss": 0.07587862014770508, "global_step": 217658, "epoch": 2445, "lr": 9.413620153201996e-06}
{"train_loss": 0.035936139523983, "global_step": 217659, "epoch": 2445, "lr": 9.413281568754706e-06}
{"train_loss": 0.04121352732181549, "global_step": 217660, "epoch": 2445, "lr": 9.412942989763817e-06}
{"train_loss": 0.06180265173316002, "global_step": 217661, "epoch": 2445, "lr": 9.412604416229354e-06}
{"train_loss": 0.03890320286154747, "global_step": 217662, "epoch": 2445, "lr": 9.41226584815138e-06}
{"train_loss": 0.06520123779773712, "global_step": 217663, "epoch": 2445, "lr": 9.411927285529925e-06}
{"train_loss": 0.037017446011304855, "global_step": 217664, "epoch": 2445, "lr": 9.411588728365051e-06}
{"train_loss": 0.06187742203474045, "global_step": 217665, "epoch": 2445, "lr": 9.411250176656782e-06}
{"train_loss": 0.07592692971229553, "global_step": 217666, "epoch": 2445, "lr": 9.41091163040519e-06}
{"train_loss": 0.03827628493309021, "global_step": 217667, "epoch": 2445, "lr": 9.410573089610291e-06}
{"train_loss": 0.025440510362386703, "global_step": 217668, "epoch": 2445, "lr": 9.410234554272157e-06}
{"train_loss": 0.031120387837290764, "global_step": 217669, "epoch": 2445, "lr": 9.409896024390814e-06}
{"train_loss": 0.06188853830099106, "global_step": 217670, "epoch": 2445, "lr": 9.409557499966326e-06}
{"train_loss": 0.06883690506219864, "global_step": 217671, "epoch": 2445, "lr": 9.409218980998718e-06}
{"train_loss": 0.048925310373306274, "global_step": 217672, "epoch": 2445, "lr": 9.408880467488057e-06}
{"train_loss": 0.0361458882689476, "global_step": 217673, "epoch": 2445, "lr": 9.408541959434364e-06}
{"train_loss": 0.06800858676433563, "global_step": 217674, "epoch": 2445, "lr": 9.40820345683771e-06}
{"train_loss": 0.0248645581305027, "global_step": 217675, "epoch": 2445, "lr": 9.407864959698114e-06}
{"train_loss": 0.07927366346120834, "global_step": 217676, "epoch": 2445, "lr": 9.407526468015648e-06}
{"train_loss": 0.053070761263370514, "global_step": 217677, "epoch": 2445, "lr": 9.40718798179035e-06}
{"train_loss": 0.07489905506372452, "global_step": 217678, "epoch": 2445, "lr": 9.406849501022247e-06}
{"train_loss": 0.050027597695589066, "global_step": 217679, "epoch": 2445, "lr": 9.406511025711411e-06}
{"train_loss": 0.06986186653375626, "global_step": 217680, "epoch": 2445, "lr": 9.406172555857862e-06}
{"train_loss": 0.028241202235221863, "global_step": 217681, "epoch": 2445, "lr": 9.405834091461668e-06}
{"train_loss": 0.03056006133556366, "global_step": 217682, "epoch": 2445, "lr": 9.405495632522854e-06}
{"train_loss": 0.09097743779420853, "global_step": 217683, "epoch": 2445, "lr": 9.405157179041491e-06}
{"train_loss": 0.06565191596746445, "global_step": 217684, "epoch": 2445, "lr": 9.404818731017595e-06}
{"train_loss": 0.025510001927614212, "global_step": 217685, "epoch": 2445, "lr": 9.404480288451239e-06}
{"train_loss": 0.033718109130859375, "global_step": 217686, "epoch": 2445, "lr": 9.404141851342447e-06}
{"train_loss": 0.05820612236857414, "global_step": 217687, "epoch": 2445, "lr": 9.40380341969128e-06}
{"train_loss": 0.048320312052965164, "global_step": 217688, "epoch": 2445, "lr": 9.403464993497768e-06}
{"train_loss": 0.04671349376440048, "global_step": 217689, "epoch": 2445, "lr": 9.403126572761977e-06}
{"train_loss": 0.03355233743786812, "global_step": 217690, "epoch": 2445, "lr": 9.402788157483928e-06}
{"train_loss": 0.05723416432738304, "global_step": 217691, "epoch": 2445, "lr": 9.402449747663678e-06}
{"train_loss": 0.05895177274942398, "global_step": 217692, "epoch": 2445, "lr": 9.402111343301289e-06}
{"train_loss": 0.05288016220575638, "global_step": 217693, "epoch": 2445, "lr": 9.401772944396774e-06, "val_loss": 8.952062606811523, "train_action_mse_error": 10.439729690551758}
{"train_loss": 0.0988994762301445, "global_step": 217694, "epoch": 2446, "lr": 9.401434550950216e-06}
{"train_loss": 0.06784871220588684, "global_step": 217695, "epoch": 2446, "lr": 9.40109616296162e-06}
{"train_loss": 0.05648685619235039, "global_step": 217696, "epoch": 2446, "lr": 9.400757780431069e-06}
{"train_loss": 0.037738025188446045, "global_step": 217697, "epoch": 2446, "lr": 9.40041940335858e-06}
{"train_loss": 0.039188455790281296, "global_step": 217698, "epoch": 2446, "lr": 9.400081031744218e-06}
{"train_loss": 0.04352499917149544, "global_step": 217699, "epoch": 2446, "lr": 9.399742665588013e-06}
{"train_loss": 0.05265972763299942, "global_step": 217700, "epoch": 2446, "lr": 9.399404304890025e-06}
{"train_loss": 0.0325477309525013, "global_step": 217701, "epoch": 2446, "lr": 9.399065949650281e-06}
{"train_loss": 0.07496324181556702, "global_step": 217702, "epoch": 2446, "lr": 9.398727599868851e-06}
{"train_loss": 0.10054680705070496, "global_step": 217703, "epoch": 2446, "lr": 9.398389255545754e-06}
{"train_loss": 0.05305520072579384, "global_step": 217704, "epoch": 2446, "lr": 9.398050916681062e-06}
{"train_loss": 0.03179397061467171, "global_step": 217705, "epoch": 2446, "lr": 9.397712583274798e-06}
{"train_loss": 0.019750092178583145, "global_step": 217706, "epoch": 2446, "lr": 9.397374255327025e-06}
{"train_loss": 0.03930789977312088, "global_step": 217707, "epoch": 2446, "lr": 9.397035932837767e-06}
{"train_loss": 0.029098862782120705, "global_step": 217708, "epoch": 2446, "lr": 9.3966976158071e-06}
{"train_loss": 0.06299935281276703, "global_step": 217709, "epoch": 2446, "lr": 9.39635930423503e-06}
{"train_loss": 0.05553125590085983, "global_step": 217710, "epoch": 2446, "lr": 9.396020998121645e-06}
{"train_loss": 0.04696058854460716, "global_step": 217711, "epoch": 2446, "lr": 9.395682697466957e-06}
{"train_loss": 0.03151050582528114, "global_step": 217712, "epoch": 2446, "lr": 9.395344402271034e-06}
{"train_loss": 0.07746855914592743, "global_step": 217713, "epoch": 2446, "lr": 9.395006112533904e-06}
{"train_loss": 0.04418107494711876, "global_step": 217714, "epoch": 2446, "lr": 9.394667828255633e-06}
{"train_loss": 0.08264203369617462, "global_step": 217715, "epoch": 2446, "lr": 9.394329549436248e-06}
{"train_loss": 0.07023430615663528, "global_step": 217716, "epoch": 2446, "lr": 9.393991276075792e-06}
{"train_loss": 0.030207857489585876, "global_step": 217717, "epoch": 2446, "lr": 9.39365300817433e-06}
{"train_loss": 0.03543431684374809, "global_step": 217718, "epoch": 2446, "lr": 9.393314745731885e-06}
{"train_loss": 0.09509018063545227, "global_step": 217719, "epoch": 2446, "lr": 9.392976488748528e-06}
{"train_loss": 0.030496416613459587, "global_step": 217720, "epoch": 2446, "lr": 9.392638237224277e-06}
{"train_loss": 0.056655190885066986, "global_step": 217721, "epoch": 2446, "lr": 9.392299991159198e-06}
{"train_loss": 0.018496468663215637, "global_step": 217722, "epoch": 2446, "lr": 9.391961750553324e-06}
{"train_loss": 0.05222637578845024, "global_step": 217723, "epoch": 2446, "lr": 9.391623515406717e-06}
{"train_loss": 0.040111687034368515, "global_step": 217724, "epoch": 2446, "lr": 9.391285285719397e-06}
{"train_loss": 0.05070999637246132, "global_step": 217725, "epoch": 2446, "lr": 9.390947061491435e-06}
{"train_loss": 0.04359273612499237, "global_step": 217726, "epoch": 2446, "lr": 9.390608842722854e-06}
{"train_loss": 0.07928712666034698, "global_step": 217727, "epoch": 2446, "lr": 9.390270629413722e-06}
{"train_loss": 0.05381506681442261, "global_step": 217728, "epoch": 2446, "lr": 9.389932421564063e-06}
{"train_loss": 0.04432625323534012, "global_step": 217729, "epoch": 2446, "lr": 9.38959421917393e-06}
{"train_loss": 0.0913027971982956, "global_step": 217730, "epoch": 2446, "lr": 9.389256022243386e-06}
{"train_loss": 0.04300375282764435, "global_step": 217731, "epoch": 2446, "lr": 9.388917830772453e-06}
{"train_loss": 0.03302610293030739, "global_step": 217732, "epoch": 2446, "lr": 9.388579644761193e-06}
{"train_loss": 0.019247574731707573, "global_step": 217733, "epoch": 2446, "lr": 9.38824146420963e-06}
{"train_loss": 0.05754329636693001, "global_step": 217734, "epoch": 2446, "lr": 9.387903289117832e-06}
{"train_loss": 0.07119181752204895, "global_step": 217735, "epoch": 2446, "lr": 9.387565119485831e-06}
{"train_loss": 0.04222160577774048, "global_step": 217736, "epoch": 2446, "lr": 9.387226955313682e-06}
{"train_loss": 0.06261519342660904, "global_step": 217737, "epoch": 2446, "lr": 9.386888796601417e-06}
{"train_loss": 0.06723155081272125, "global_step": 217738, "epoch": 2446, "lr": 9.386550643349107e-06}
{"train_loss": 0.04267093166708946, "global_step": 217739, "epoch": 2446, "lr": 9.386212495556763e-06}
{"train_loss": 0.044137630611658096, "global_step": 217740, "epoch": 2446, "lr": 9.385874353224455e-06}
{"train_loss": 0.022986099123954773, "global_step": 217741, "epoch": 2446, "lr": 9.385536216352214e-06}
{"train_loss": 0.07813408970832825, "global_step": 217742, "epoch": 2446, "lr": 9.385198084940106e-06}
{"train_loss": 0.056531280279159546, "global_step": 217743, "epoch": 2446, "lr": 9.384859958988152e-06}
{"train_loss": 0.04898795485496521, "global_step": 217744, "epoch": 2446, "lr": 9.384521838496412e-06}
{"train_loss": 0.028504105284810066, "global_step": 217745, "epoch": 2446, "lr": 9.384183723464924e-06}
{"train_loss": 0.06602177768945694, "global_step": 217746, "epoch": 2446, "lr": 9.383845613893738e-06}
{"train_loss": 0.048306141048669815, "global_step": 217747, "epoch": 2446, "lr": 9.383507509782907e-06}
{"train_loss": 0.021814128383994102, "global_step": 217748, "epoch": 2446, "lr": 9.38316941113247e-06}
{"train_loss": 0.021417267620563507, "global_step": 217749, "epoch": 2446, "lr": 9.382831317942459e-06}
{"train_loss": 0.04934541881084442, "global_step": 217750, "epoch": 2446, "lr": 9.382493230212941e-06}
{"train_loss": 0.05930459126830101, "global_step": 217751, "epoch": 2446, "lr": 9.382155147943943e-06}
{"train_loss": 0.04709934815764427, "global_step": 217752, "epoch": 2446, "lr": 9.38181707113553e-06}
{"train_loss": 0.02070843242108822, "global_step": 217753, "epoch": 2446, "lr": 9.381478999787734e-06}
{"train_loss": 0.08195406943559647, "global_step": 217754, "epoch": 2446, "lr": 9.381140933900596e-06}
{"train_loss": 0.027824658900499344, "global_step": 217755, "epoch": 2446, "lr": 9.380802873474176e-06}
{"train_loss": 0.08491148799657822, "global_step": 217756, "epoch": 2446, "lr": 9.380464818508506e-06}
{"train_loss": 0.029914408922195435, "global_step": 217757, "epoch": 2446, "lr": 9.380126769003645e-06}
{"train_loss": 0.03817512467503548, "global_step": 217758, "epoch": 2446, "lr": 9.379788724959614e-06}
{"train_loss": 0.05822377651929855, "global_step": 217759, "epoch": 2446, "lr": 9.379450686376495e-06}
{"train_loss": 0.04701598733663559, "global_step": 217760, "epoch": 2446, "lr": 9.379112653254302e-06}
{"train_loss": 0.05531533434987068, "global_step": 217761, "epoch": 2446, "lr": 9.378774625593101e-06}
{"train_loss": 0.06916546076536179, "global_step": 217762, "epoch": 2446, "lr": 9.378436603392915e-06}
{"train_loss": 0.11010091751813889, "global_step": 217763, "epoch": 2446, "lr": 9.378098586653817e-06}
{"train_loss": 0.03327953815460205, "global_step": 217764, "epoch": 2446, "lr": 9.377760575375826e-06}
{"train_loss": 0.026579897850751877, "global_step": 217765, "epoch": 2446, "lr": 9.377422569559002e-06}
{"train_loss": 0.04577535018324852, "global_step": 217766, "epoch": 2446, "lr": 9.377084569203398e-06}
{"train_loss": 0.04157742112874985, "global_step": 217767, "epoch": 2446, "lr": 9.376746574309042e-06}
{"train_loss": 0.04755738377571106, "global_step": 217768, "epoch": 2446, "lr": 9.376408584875995e-06}
{"train_loss": 0.0679488256573677, "global_step": 217769, "epoch": 2446, "lr": 9.376070600904285e-06}
{"train_loss": 0.03532770648598671, "global_step": 217770, "epoch": 2446, "lr": 9.375732622393979e-06}
{"train_loss": 0.024940073490142822, "global_step": 217771, "epoch": 2446, "lr": 9.375394649345099e-06}
{"train_loss": 0.04266104847192764, "global_step": 217772, "epoch": 2446, "lr": 9.375056681757711e-06}
{"train_loss": 0.05210774764418602, "global_step": 217773, "epoch": 2446, "lr": 9.374718719631843e-06}
{"train_loss": 0.05667027458548546, "global_step": 217774, "epoch": 2446, "lr": 9.374380762967555e-06}
{"train_loss": 0.04779535159468651, "global_step": 217775, "epoch": 2446, "lr": 9.374042811764883e-06}
{"train_loss": 0.09667657315731049, "global_step": 217776, "epoch": 2446, "lr": 9.37370486602388e-06}
{"train_loss": 0.08764634281396866, "global_step": 217777, "epoch": 2446, "lr": 9.373366925744581e-06}
{"train_loss": 0.05324976146221161, "global_step": 217778, "epoch": 2446, "lr": 9.373028990927046e-06}
{"train_loss": 0.03027639165520668, "global_step": 217779, "epoch": 2446, "lr": 9.372691061571299e-06}
{"train_loss": 0.06600897014141083, "global_step": 217780, "epoch": 2446, "lr": 9.372353137677415e-06}
{"train_loss": 0.06820283830165863, "global_step": 217781, "epoch": 2446, "lr": 9.372015219245412e-06}
{"train_loss": 0.052347640140672745, "global_step": 217782, "epoch": 2446, "lr": 9.371677306275339e-06, "val_loss": 8.861324310302734}
{"train_loss": 0.011685164645314217, "global_step": 217783, "epoch": 2447, "lr": 9.371339398767265e-06}
{"train_loss": 0.05629797652363777, "global_step": 217784, "epoch": 2447, "lr": 9.37100149672121e-06}
{"train_loss": 0.042533475905656815, "global_step": 217785, "epoch": 2447, "lr": 9.370663600137236e-06}
{"train_loss": 0.060737382620573044, "global_step": 217786, "epoch": 2447, "lr": 9.37032570901537e-06}
{"train_loss": 0.028499694541096687, "global_step": 217787, "epoch": 2447, "lr": 9.36998782335568e-06}
{"train_loss": 0.06510491669178009, "global_step": 217788, "epoch": 2447, "lr": 9.369649943158203e-06}
{"train_loss": 0.03031264990568161, "global_step": 217789, "epoch": 2447, "lr": 9.369312068422969e-06}
{"train_loss": 0.040496401488780975, "global_step": 217790, "epoch": 2447, "lr": 9.36897419915005e-06}
{"train_loss": 0.058765411376953125, "global_step": 217791, "epoch": 2447, "lr": 9.368636335339458e-06}
{"train_loss": 0.04378972202539444, "global_step": 217792, "epoch": 2447, "lr": 9.368298476991272e-06}
{"train_loss": 0.06787708401679993, "global_step": 217793, "epoch": 2447, "lr": 9.367960624105527e-06}
{"train_loss": 0.04524251073598862, "global_step": 217794, "epoch": 2447, "lr": 9.36762277668225e-06}
{"train_loss": 0.06901282072067261, "global_step": 217795, "epoch": 2447, "lr": 9.36728493472151e-06}
{"train_loss": 0.0458914078772068, "global_step": 217796, "epoch": 2447, "lr": 9.366947098223333e-06}
{"train_loss": 0.07867974042892456, "global_step": 217797, "epoch": 2447, "lr": 9.366609267187792e-06}
{"train_loss": 0.06506264209747314, "global_step": 217798, "epoch": 2447, "lr": 9.366271441614899e-06}
{"train_loss": 0.09312279522418976, "global_step": 217799, "epoch": 2447, "lr": 9.365933621504724e-06}
{"train_loss": 0.06373903155326843, "global_step": 217800, "epoch": 2447, "lr": 9.365595806857297e-06}
{"train_loss": 0.04525606334209442, "global_step": 217801, "epoch": 2447, "lr": 9.365257997672667e-06}
{"train_loss": 0.033572468906641006, "global_step": 217802, "epoch": 2447, "lr": 9.3649201939509e-06}
{"train_loss": 0.041569456458091736, "global_step": 217803, "epoch": 2447, "lr": 9.364582395692006e-06}
{"train_loss": 0.061114352196455, "global_step": 217804, "epoch": 2447, "lr": 9.364244602896066e-06}
{"train_loss": 0.07129020988941193, "global_step": 217805, "epoch": 2447, "lr": 9.36390681556309e-06}
{"train_loss": 0.04768747091293335, "global_step": 217806, "epoch": 2447, "lr": 9.36356903369316e-06}
{"train_loss": 0.06037634611129761, "global_step": 217807, "epoch": 2447, "lr": 9.363231257286287e-06}
{"train_loss": 0.057304538786411285, "global_step": 217808, "epoch": 2447, "lr": 9.362893486342544e-06}
{"train_loss": 0.04011594504117966, "global_step": 217809, "epoch": 2447, "lr": 9.362555720861955e-06}
{"train_loss": 0.04584047198295593, "global_step": 217810, "epoch": 2447, "lr": 9.362217960844583e-06}
{"train_loss": 0.03953079506754875, "global_step": 217811, "epoch": 2447, "lr": 9.361880206290457e-06}
{"train_loss": 0.059971705079078674, "global_step": 217812, "epoch": 2447, "lr": 9.36154245719964e-06}
{"train_loss": 0.0429416187107563, "global_step": 217813, "epoch": 2447, "lr": 9.361204713572158e-06}
{"train_loss": 0.04264030605554581, "global_step": 217814, "epoch": 2447, "lr": 9.360866975408084e-06}
{"train_loss": 0.054775066673755646, "global_step": 217815, "epoch": 2447, "lr": 9.360529242707427e-06}
{"train_loss": 0.046531107276678085, "global_step": 217816, "epoch": 2447, "lr": 9.360191515470268e-06}
{"train_loss": 0.025100458413362503, "global_step": 217817, "epoch": 2447, "lr": 9.359853793696616e-06}
{"train_loss": 0.028755666688084602, "global_step": 217818, "epoch": 2447, "lr": 9.359516077386542e-06}
{"train_loss": 0.08170638233423233, "global_step": 217819, "epoch": 2447, "lr": 9.3591783665401e-06}
{"train_loss": 0.08183975517749786, "global_step": 217820, "epoch": 2447, "lr": 9.358840661157309e-06}
{"train_loss": 0.018360871821641922, "global_step": 217821, "epoch": 2447, "lr": 9.358502961238236e-06}
{"train_loss": 0.05182120203971863, "global_step": 217822, "epoch": 2447, "lr": 9.358165266782904e-06}
{"train_loss": 0.08811827003955841, "global_step": 217823, "epoch": 2447, "lr": 9.357827577791385e-06}
{"train_loss": 0.012906950898468494, "global_step": 217824, "epoch": 2447, "lr": 9.357489894263699e-06}
{"train_loss": 0.04457215964794159, "global_step": 217825, "epoch": 2447, "lr": 9.357152216199911e-06}
{"train_loss": 0.06388961523771286, "global_step": 217826, "epoch": 2447, "lr": 9.356814543600062e-06}
{"train_loss": 0.06335649639368057, "global_step": 217827, "epoch": 2447, "lr": 9.356476876464181e-06}
{"train_loss": 0.06474240869283676, "global_step": 217828, "epoch": 2447, "lr": 9.356139214792343e-06}
{"train_loss": 0.03212852030992508, "global_step": 217829, "epoch": 2447, "lr": 9.35580155858456e-06}
{"train_loss": 0.12765932083129883, "global_step": 217830, "epoch": 2447, "lr": 9.355463907840906e-06}
{"train_loss": 0.08111737668514252, "global_step": 217831, "epoch": 2447, "lr": 9.355126262561414e-06}
{"train_loss": 0.04824769124388695, "global_step": 217832, "epoch": 2447, "lr": 9.354788622746125e-06}
{"train_loss": 0.03487560153007507, "global_step": 217833, "epoch": 2447, "lr": 9.354450988395092e-06}
{"train_loss": 0.03183930739760399, "global_step": 217834, "epoch": 2447, "lr": 9.354113359508348e-06}
{"train_loss": 0.09358516335487366, "global_step": 217835, "epoch": 2447, "lr": 9.353775736085963e-06}
{"train_loss": 0.06805530935525894, "global_step": 217836, "epoch": 2447, "lr": 9.353438118127955e-06}
{"train_loss": 0.08141843974590302, "global_step": 217837, "epoch": 2447, "lr": 9.353100505634377e-06}
{"train_loss": 0.1240951344370842, "global_step": 217838, "epoch": 2447, "lr": 9.3527628986053e-06}
{"train_loss": 0.028980469331145287, "global_step": 217839, "epoch": 2447, "lr": 9.352425297040728e-06}
{"train_loss": 0.04789280518889427, "global_step": 217840, "epoch": 2447, "lr": 9.352087700940742e-06}
{"train_loss": 0.06475047022104263, "global_step": 217841, "epoch": 2447, "lr": 9.351750110305363e-06}
{"train_loss": 0.046947211027145386, "global_step": 217842, "epoch": 2447, "lr": 9.351412525134657e-06}
{"train_loss": 0.024783404543995857, "global_step": 217843, "epoch": 2447, "lr": 9.35107494542864e-06}
{"train_loss": 0.02989215962588787, "global_step": 217844, "epoch": 2447, "lr": 9.350737371187396e-06}
{"train_loss": 0.07303228974342346, "global_step": 217845, "epoch": 2447, "lr": 9.350399802410931e-06}
{"train_loss": 0.029722122475504875, "global_step": 217846, "epoch": 2447, "lr": 9.350062239099327e-06}
{"train_loss": 0.07345904409885406, "global_step": 217847, "epoch": 2447, "lr": 9.349724681252598e-06}
{"train_loss": 0.06933143734931946, "global_step": 217848, "epoch": 2447, "lr": 9.349387128870813e-06}
{"train_loss": 0.0584648959338665, "global_step": 217849, "epoch": 2447, "lr": 9.349049581953995e-06}
{"train_loss": 0.049599871039390564, "global_step": 217850, "epoch": 2447, "lr": 9.348712040502217e-06}
{"train_loss": 0.05546574667096138, "global_step": 217851, "epoch": 2447, "lr": 9.348374504515494e-06}
{"train_loss": 0.05859573557972908, "global_step": 217852, "epoch": 2447, "lr": 9.3480369739939e-06}
{"train_loss": 0.05714786425232887, "global_step": 217853, "epoch": 2447, "lr": 9.347699448937458e-06}
{"train_loss": 0.025906480848789215, "global_step": 217854, "epoch": 2447, "lr": 9.347361929346231e-06}
{"train_loss": 0.0590895377099514, "global_step": 217855, "epoch": 2447, "lr": 9.347024415220246e-06}
{"train_loss": 0.057805780321359634, "global_step": 217856, "epoch": 2447, "lr": 9.346686906559555e-06}
{"train_loss": 0.07613387703895569, "global_step": 217857, "epoch": 2447, "lr": 9.346349403364218e-06}
{"train_loss": 0.043719787150621414, "global_step": 217858, "epoch": 2447, "lr": 9.346011905634261e-06}
{"train_loss": 0.022026926279067993, "global_step": 217859, "epoch": 2447, "lr": 9.345674413369742e-06}
{"train_loss": 0.05300004035234451, "global_step": 217860, "epoch": 2447, "lr": 9.345336926570697e-06}
{"train_loss": 0.07083321362733841, "global_step": 217861, "epoch": 2447, "lr": 9.344999445237185e-06}
{"train_loss": 0.05155056715011597, "global_step": 217862, "epoch": 2447, "lr": 9.344661969369229e-06}
{"train_loss": 0.029999706894159317, "global_step": 217863, "epoch": 2447, "lr": 9.344324498966906e-06}
{"train_loss": 0.05183273181319237, "global_step": 217864, "epoch": 2447, "lr": 9.343987034030238e-06}
{"train_loss": 0.03968548774719238, "global_step": 217865, "epoch": 2447, "lr": 9.34364957455926e-06}
{"train_loss": 0.03970344737172127, "global_step": 217866, "epoch": 2447, "lr": 9.34331212055405e-06}
{"train_loss": 0.06819170713424683, "global_step": 217867, "epoch": 2447, "lr": 9.342974672014621e-06}
{"train_loss": 0.043013304471969604, "global_step": 217868, "epoch": 2447, "lr": 9.342637228941048e-06}
{"train_loss": 0.07165171951055527, "global_step": 217869, "epoch": 2447, "lr": 9.342299791333364e-06}
{"train_loss": 0.06062709912657738, "global_step": 217870, "epoch": 2447, "lr": 9.341962359191598e-06}
{"train_loss": 0.0535723694568772, "global_step": 217871, "epoch": 2447, "lr": 9.341624932515818e-06, "val_loss": 8.829678535461426}
{"train_loss": 0.0459149070084095, "global_step": 217872, "epoch": 2448, "lr": 9.341287511306052e-06}
{"train_loss": 0.027739742770791054, "global_step": 217873, "epoch": 2448, "lr": 9.340950095562356e-06}
{"train_loss": 0.021076200529932976, "global_step": 217874, "epoch": 2448, "lr": 9.340612685284784e-06}
{"train_loss": 0.040894970297813416, "global_step": 217875, "epoch": 2448, "lr": 9.340275280473359e-06}
{"train_loss": 0.01982247084379196, "global_step": 217876, "epoch": 2448, "lr": 9.339937881128147e-06}
{"train_loss": 0.056698355823755264, "global_step": 217877, "epoch": 2448, "lr": 9.339600487249179e-06}
{"train_loss": 0.0396641306579113, "global_step": 217878, "epoch": 2448, "lr": 9.339263098836515e-06}
{"train_loss": 0.0352713018655777, "global_step": 217879, "epoch": 2448, "lr": 9.338925715890178e-06}
{"train_loss": 0.028928715735673904, "global_step": 217880, "epoch": 2448, "lr": 9.338588338410243e-06}
{"train_loss": 0.04834940284490585, "global_step": 217881, "epoch": 2448, "lr": 9.338250966396722e-06}
{"train_loss": 0.019892176613211632, "global_step": 217882, "epoch": 2448, "lr": 9.337913599849695e-06}
{"train_loss": 0.0658683180809021, "global_step": 217883, "epoch": 2448, "lr": 9.337576238769175e-06}
{"train_loss": 0.03527448698878288, "global_step": 217884, "epoch": 2448, "lr": 9.337238883155236e-06}
{"train_loss": 0.0764373242855072, "global_step": 217885, "epoch": 2448, "lr": 9.33690153300789e-06}
{"train_loss": 0.08564406633377075, "global_step": 217886, "epoch": 2448, "lr": 9.336564188327223e-06}
{"train_loss": 0.0685514435172081, "global_step": 217887, "epoch": 2448, "lr": 9.336226849113245e-06}
{"train_loss": 0.07289881259202957, "global_step": 217888, "epoch": 2448, "lr": 9.335889515366026e-06}
{"train_loss": 0.07344085723161697, "global_step": 217889, "epoch": 2448, "lr": 9.335552187085595e-06}
{"train_loss": 0.08184841275215149, "global_step": 217890, "epoch": 2448, "lr": 9.335214864272007e-06}
{"train_loss": 0.02718513458967209, "global_step": 217891, "epoch": 2448, "lr": 9.334877546925296e-06}
{"train_loss": 0.11037008464336395, "global_step": 217892, "epoch": 2448, "lr": 9.334540235045513e-06}
{"train_loss": 0.05396793410181999, "global_step": 217893, "epoch": 2448, "lr": 9.334202928632718e-06}
{"train_loss": 0.07325252890586853, "global_step": 217894, "epoch": 2448, "lr": 9.333865627686933e-06}
{"train_loss": 0.05809417739510536, "global_step": 217895, "epoch": 2448, "lr": 9.333528332208224e-06}
{"train_loss": 0.05436643585562706, "global_step": 217896, "epoch": 2448, "lr": 9.333191042196621e-06}
{"train_loss": 0.04647529870271683, "global_step": 217897, "epoch": 2448, "lr": 9.332853757652182e-06}
{"train_loss": 0.05728389695286751, "global_step": 217898, "epoch": 2448, "lr": 9.332516478574933e-06}
{"train_loss": 0.05382788926362991, "global_step": 217899, "epoch": 2448, "lr": 9.332179204964942e-06}
{"train_loss": 0.043259892612695694, "global_step": 217900, "epoch": 2448, "lr": 9.331841936822232e-06}
{"train_loss": 0.05932672321796417, "global_step": 217901, "epoch": 2448, "lr": 9.331504674146873e-06}
{"train_loss": 0.04635594040155411, "global_step": 217902, "epoch": 2448, "lr": 9.331167416938884e-06}
{"train_loss": 0.03521645441651344, "global_step": 217903, "epoch": 2448, "lr": 9.33083016519834e-06}
{"train_loss": 0.061021607369184494, "global_step": 217904, "epoch": 2448, "lr": 9.330492918925271e-06}
{"train_loss": 0.1215561032295227, "global_step": 217905, "epoch": 2448, "lr": 9.330155678119707e-06}
{"train_loss": 0.08022787421941757, "global_step": 217906, "epoch": 2448, "lr": 9.329818442781719e-06}
{"train_loss": 0.09290818125009537, "global_step": 217907, "epoch": 2448, "lr": 9.329481212911329e-06}
{"train_loss": 0.048083458095788956, "global_step": 217908, "epoch": 2448, "lr": 9.32914398850861e-06}
{"train_loss": 0.05567984655499458, "global_step": 217909, "epoch": 2448, "lr": 9.328806769573572e-06}
{"train_loss": 0.04778096824884415, "global_step": 217910, "epoch": 2448, "lr": 9.3284695561063e-06}
{"train_loss": 0.0609191469848156, "global_step": 217911, "epoch": 2448, "lr": 9.328132348106805e-06}
{"train_loss": 0.016926661133766174, "global_step": 217912, "epoch": 2448, "lr": 9.327795145575159e-06}
{"train_loss": 0.08396591991186142, "global_step": 217913, "epoch": 2448, "lr": 9.327457948511381e-06}
{"train_loss": 0.04914528876543045, "global_step": 217914, "epoch": 2448, "lr": 9.327120756915547e-06}
{"train_loss": 0.06732223182916641, "global_step": 217915, "epoch": 2448, "lr": 9.326783570787673e-06}
{"train_loss": 0.0533665306866169, "global_step": 217916, "epoch": 2448, "lr": 9.32644639012783e-06}
{"train_loss": 0.04461030289530754, "global_step": 217917, "epoch": 2448, "lr": 9.326109214936035e-06}
{"train_loss": 0.038823358714580536, "global_step": 217918, "epoch": 2448, "lr": 9.32577204521236e-06}
{"train_loss": 0.02375592477619648, "global_step": 217919, "epoch": 2448, "lr": 9.325434880956829e-06}
{"train_loss": 0.038514189422130585, "global_step": 217920, "epoch": 2448, "lr": 9.325097722169513e-06}
{"train_loss": 0.060314588248729706, "global_step": 217921, "epoch": 2448, "lr": 9.324760568850422e-06}
{"train_loss": 0.02379697374999523, "global_step": 217922, "epoch": 2448, "lr": 9.32442342099964e-06}
{"train_loss": 0.01135703269392252, "global_step": 217923, "epoch": 2448, "lr": 9.324086278617178e-06}
{"train_loss": 0.04939612001180649, "global_step": 217924, "epoch": 2448, "lr": 9.323749141703114e-06}
{"train_loss": 0.026966245844960213, "global_step": 217925, "epoch": 2448, "lr": 9.32341201025746e-06}
{"train_loss": 0.06522273272275925, "global_step": 217926, "epoch": 2448, "lr": 9.323074884280285e-06}
{"train_loss": 0.06983605772256851, "global_step": 217927, "epoch": 2448, "lr": 9.322737763771622e-06}
{"train_loss": 0.07788141816854477, "global_step": 217928, "epoch": 2448, "lr": 9.322400648731522e-06}
{"train_loss": 0.01481266226619482, "global_step": 217929, "epoch": 2448, "lr": 9.322063539160036e-06}
{"train_loss": 0.02944459393620491, "global_step": 217930, "epoch": 2448, "lr": 9.321726435057193e-06}
{"train_loss": 0.04414346441626549, "global_step": 217931, "epoch": 2448, "lr": 9.321389336423059e-06}
{"train_loss": 0.05929824337363243, "global_step": 217932, "epoch": 2448, "lr": 9.32105224325766e-06}
{"train_loss": 0.06424111872911453, "global_step": 217933, "epoch": 2448, "lr": 9.320715155561055e-06}
{"train_loss": 0.04229716956615448, "global_step": 217934, "epoch": 2448, "lr": 9.32037807333328e-06}
{"train_loss": 0.044228482991456985, "global_step": 217935, "epoch": 2448, "lr": 9.320040996574391e-06}
{"train_loss": 0.029834147542715073, "global_step": 217936, "epoch": 2448, "lr": 9.319703925284417e-06}
{"train_loss": 0.0589563325047493, "global_step": 217937, "epoch": 2448, "lr": 9.319366859463425e-06}
{"train_loss": 0.08533386141061783, "global_step": 217938, "epoch": 2448, "lr": 9.319029799111434e-06}
{"train_loss": 0.0918697640299797, "global_step": 217939, "epoch": 2448, "lr": 9.318692744228519e-06}
{"train_loss": 0.040403056889772415, "global_step": 217940, "epoch": 2448, "lr": 9.318355694814695e-06}
{"train_loss": 0.08140172064304352, "global_step": 217941, "epoch": 2448, "lr": 9.318018650870036e-06}
{"train_loss": 0.09503796696662903, "global_step": 217942, "epoch": 2448, "lr": 9.317681612394574e-06}
{"train_loss": 0.09736528992652893, "global_step": 217943, "epoch": 2448, "lr": 9.317344579388337e-06}
{"train_loss": 0.0390758290886879, "global_step": 217944, "epoch": 2448, "lr": 9.317007551851403e-06}
{"train_loss": 0.07104368507862091, "global_step": 217945, "epoch": 2448, "lr": 9.316670529783789e-06}
{"train_loss": 0.04421485960483551, "global_step": 217946, "epoch": 2448, "lr": 9.316333513185566e-06}
{"train_loss": 0.03383105620741844, "global_step": 217947, "epoch": 2448, "lr": 9.315996502056756e-06}
{"train_loss": 0.06570786237716675, "global_step": 217948, "epoch": 2448, "lr": 9.315659496397422e-06}
{"train_loss": 0.04790356010198593, "global_step": 217949, "epoch": 2448, "lr": 9.31532249620759e-06}
{"train_loss": 0.10807851701974869, "global_step": 217950, "epoch": 2448, "lr": 9.314985501487334e-06}
{"train_loss": 0.12302131950855255, "global_step": 217951, "epoch": 2448, "lr": 9.314648512236663e-06}
{"train_loss": 0.05696965008974075, "global_step": 217952, "epoch": 2448, "lr": 9.31431152845566e-06}
{"train_loss": 0.06359875202178955, "global_step": 217953, "epoch": 2448, "lr": 9.313974550144338e-06}
{"train_loss": 0.0428333654999733, "global_step": 217954, "epoch": 2448, "lr": 9.313637577302764e-06}
{"train_loss": 0.05745099112391472, "global_step": 217955, "epoch": 2448, "lr": 9.313300609930969e-06}
{"train_loss": 0.03129876032471657, "global_step": 217956, "epoch": 2448, "lr": 9.312963648029017e-06}
{"train_loss": 0.07927480340003967, "global_step": 217957, "epoch": 2448, "lr": 9.312626691596927e-06}
{"train_loss": 0.09465813636779785, "global_step": 217958, "epoch": 2448, "lr": 9.312289740634766e-06}
{"train_loss": 0.038505177944898605, "global_step": 217959, "epoch": 2448, "lr": 9.311952795142564e-06}
{"train_loss": 0.055395349023047456, "global_step": 217960, "epoch": 2448, "lr": 9.311615855120387e-06, "val_loss": 8.78558349609375}
{"train_loss": 0.04503251984715462, "global_step": 217961, "epoch": 2449, "lr": 9.311278920568256e-06}
{"train_loss": 0.08285816013813019, "global_step": 217962, "epoch": 2449, "lr": 9.310941991486239e-06}
{"train_loss": 0.033006731420755386, "global_step": 217963, "epoch": 2449, "lr": 9.310605067874357e-06}
{"train_loss": 0.06864984333515167, "global_step": 217964, "epoch": 2449, "lr": 9.310268149732666e-06}
{"train_loss": 0.09310358762741089, "global_step": 217965, "epoch": 2449, "lr": 9.30993123706123e-06}
{"train_loss": 0.06385806947946548, "global_step": 217966, "epoch": 2449, "lr": 9.309594329860061e-06}
{"train_loss": 0.05677440017461777, "global_step": 217967, "epoch": 2449, "lr": 9.309257428129237e-06}
{"train_loss": 0.08096910268068314, "global_step": 217968, "epoch": 2449, "lr": 9.308920531868775e-06}
{"train_loss": 0.1339665949344635, "global_step": 217969, "epoch": 2449, "lr": 9.308583641078744e-06}
{"train_loss": 0.06588771194219589, "global_step": 217970, "epoch": 2449, "lr": 9.308246755759165e-06}
{"train_loss": 0.07485198974609375, "global_step": 217971, "epoch": 2449, "lr": 9.307909875910109e-06}
{"train_loss": 0.1218995749950409, "global_step": 217972, "epoch": 2449, "lr": 9.307573001531595e-06}
{"train_loss": 0.02694842219352722, "global_step": 217973, "epoch": 2449, "lr": 9.307236132623697e-06}
{"train_loss": 0.03479460999369621, "global_step": 217974, "epoch": 2449, "lr": 9.306899269186432e-06}
{"train_loss": 0.03952264413237572, "global_step": 217975, "epoch": 2449, "lr": 9.306562411219872e-06}
{"train_loss": 0.0513533391058445, "global_step": 217976, "epoch": 2449, "lr": 9.30622555872403e-06}
{"train_loss": 0.05824418365955353, "global_step": 217977, "epoch": 2449, "lr": 9.30588871169899e-06}
{"train_loss": 0.05840863287448883, "global_step": 217978, "epoch": 2449, "lr": 9.305551870144757e-06}
{"train_loss": 0.058179937303066254, "global_step": 217979, "epoch": 2449, "lr": 9.305215034061415e-06}
{"train_loss": 0.0860644057393074, "global_step": 217980, "epoch": 2449, "lr": 9.30487820344899e-06}
{"train_loss": 0.043640706688165665, "global_step": 217981, "epoch": 2449, "lr": 9.304541378307519e-06}
{"train_loss": 0.042199473828077316, "global_step": 217982, "epoch": 2449, "lr": 9.304204558637064e-06}
{"train_loss": 0.040153175592422485, "global_step": 217983, "epoch": 2449, "lr": 9.303867744437649e-06}
{"train_loss": 0.042743369936943054, "global_step": 217984, "epoch": 2449, "lr": 9.303530935709349e-06}
{"train_loss": 0.06827402859926224, "global_step": 217985, "epoch": 2449, "lr": 9.303194132452181e-06}
{"train_loss": 0.035389117896556854, "global_step": 217986, "epoch": 2449, "lr": 9.30285733466621e-06}
{"train_loss": 0.05263343080878258, "global_step": 217987, "epoch": 2449, "lr": 9.302520542351467e-06}
{"train_loss": 0.041427116841077805, "global_step": 217988, "epoch": 2449, "lr": 9.302183755508016e-06}
{"train_loss": 0.024084249511361122, "global_step": 217989, "epoch": 2449, "lr": 9.301846974135874e-06}
{"train_loss": 0.06908248364925385, "global_step": 217990, "epoch": 2449, "lr": 9.30151019823512e-06}
{"train_loss": 0.05982893332839012, "global_step": 217991, "epoch": 2449, "lr": 9.301173427805765e-06}
{"train_loss": 0.045381102710962296, "global_step": 217992, "epoch": 2449, "lr": 9.300836662847884e-06}
{"train_loss": 0.02191215753555298, "global_step": 217993, "epoch": 2449, "lr": 9.300499903361498e-06}
{"train_loss": 0.03915465250611305, "global_step": 217994, "epoch": 2449, "lr": 9.300163149346674e-06}
{"train_loss": 0.03378874063491821, "global_step": 217995, "epoch": 2449, "lr": 9.29982640080344e-06}
{"train_loss": 0.075165756046772, "global_step": 217996, "epoch": 2449, "lr": 9.299489657731857e-06}
{"train_loss": 0.04986925795674324, "global_step": 217997, "epoch": 2449, "lr": 9.299152920131948e-06}
{"train_loss": 0.09474771469831467, "global_step": 217998, "epoch": 2449, "lr": 9.29881618800379e-06}
{"train_loss": 0.025598224252462387, "global_step": 217999, "epoch": 2449, "lr": 9.298479461347393e-06}
{"train_loss": 0.027830902487039566, "global_step": 218000, "epoch": 2449, "lr": 9.298142740162819e-06}
{"train_loss": 0.07404235005378723, "global_step": 218001, "epoch": 2449, "lr": 9.297806024450129e-06}
{"train_loss": 0.04202435538172722, "global_step": 218002, "epoch": 2449, "lr": 9.297469314209339e-06}
{"train_loss": 0.03842711076140404, "global_step": 218003, "epoch": 2449, "lr": 9.297132609440518e-06}
{"train_loss": 0.016383096575737, "global_step": 218004, "epoch": 2449, "lr": 9.29679591014369e-06}
{"train_loss": 0.03455808386206627, "global_step": 218005, "epoch": 2449, "lr": 9.296459216318926e-06}
{"train_loss": 0.049913763999938965, "global_step": 218006, "epoch": 2449, "lr": 9.296122527966245e-06}
{"train_loss": 0.032840926200151443, "global_step": 218007, "epoch": 2449, "lr": 9.295785845085719e-06}
{"train_loss": 0.0556410513818264, "global_step": 218008, "epoch": 2449, "lr": 9.295449167677361e-06}
{"train_loss": 0.07434920221567154, "global_step": 218009, "epoch": 2449, "lr": 9.295112495741248e-06}
{"train_loss": 0.09570638090372086, "global_step": 218010, "epoch": 2449, "lr": 9.294775829277396e-06}
{"train_loss": 0.0749797448515892, "global_step": 218011, "epoch": 2449, "lr": 9.294439168285879e-06}
{"train_loss": 0.04918970167636871, "global_step": 218012, "epoch": 2449, "lr": 9.294102512766718e-06}
{"train_loss": 0.040190231055021286, "global_step": 218013, "epoch": 2449, "lr": 9.293765862719984e-06}
{"train_loss": 0.0652228444814682, "global_step": 218014, "epoch": 2449, "lr": 9.293429218145688e-06}
{"train_loss": 0.05923861637711525, "global_step": 218015, "epoch": 2449, "lr": 9.29309257904391e-06}
{"train_loss": 0.10062204301357269, "global_step": 218016, "epoch": 2449, "lr": 9.292755945414666e-06}
{"train_loss": 0.06736829876899719, "global_step": 218017, "epoch": 2449, "lr": 9.292419317258028e-06}
{"train_loss": 0.015350054018199444, "global_step": 218018, "epoch": 2449, "lr": 9.292082694574018e-06}
{"train_loss": 0.04085917770862579, "global_step": 218019, "epoch": 2449, "lr": 9.291746077362701e-06}
{"train_loss": 0.06181963533163071, "global_step": 218020, "epoch": 2449, "lr": 9.291409465624113e-06}
{"train_loss": 0.054800212383270264, "global_step": 218021, "epoch": 2449, "lr": 9.291072859358285e-06}
{"train_loss": 0.056074243038892746, "global_step": 218022, "epoch": 2449, "lr": 9.290736258565292e-06}
{"train_loss": 0.09293647110462189, "global_step": 218023, "epoch": 2449, "lr": 9.290399663245147e-06}
{"train_loss": 0.05107101798057556, "global_step": 218024, "epoch": 2449, "lr": 9.290063073397925e-06}
{"train_loss": 0.06078072637319565, "global_step": 218025, "epoch": 2449, "lr": 9.289726489023648e-06}
{"train_loss": 0.0435117669403553, "global_step": 218026, "epoch": 2449, "lr": 9.289389910122381e-06}
{"train_loss": 0.0745488777756691, "global_step": 218027, "epoch": 2449, "lr": 9.289053336694148e-06}
{"train_loss": 0.06527016311883926, "global_step": 218028, "epoch": 2449, "lr": 9.28871676873902e-06}
{"train_loss": 0.05909062176942825, "global_step": 218029, "epoch": 2449, "lr": 9.288380206257013e-06}
{"train_loss": 0.09542311728000641, "global_step": 218030, "epoch": 2449, "lr": 9.288043649248195e-06}
{"train_loss": 0.0588933527469635, "global_step": 218031, "epoch": 2449, "lr": 9.2877070977126e-06}
{"train_loss": 0.0617353655397892, "global_step": 218032, "epoch": 2449, "lr": 9.287370551650281e-06}
{"train_loss": 0.020593101158738136, "global_step": 218033, "epoch": 2449, "lr": 9.287034011061274e-06}
{"train_loss": 0.038044173270463943, "global_step": 218034, "epoch": 2449, "lr": 9.286697475945638e-06}
{"train_loss": 0.0818634033203125, "global_step": 218035, "epoch": 2449, "lr": 9.286360946303396e-06}
{"train_loss": 0.11035246402025223, "global_step": 218036, "epoch": 2449, "lr": 9.286024422134603e-06}
{"train_loss": 0.025987209752202034, "global_step": 218037, "epoch": 2449, "lr": 9.285687903439327e-06}
{"train_loss": 0.054703570902347565, "global_step": 218038, "epoch": 2449, "lr": 9.285351390217577e-06}
{"train_loss": 0.03514684736728668, "global_step": 218039, "epoch": 2449, "lr": 9.285014882469428e-06}
{"train_loss": 0.054046712815761566, "global_step": 218040, "epoch": 2449, "lr": 9.2846783801949e-06}
{"train_loss": 0.04819129407405853, "global_step": 218041, "epoch": 2449, "lr": 9.284341883394065e-06}
{"train_loss": 0.052346449345350266, "global_step": 218042, "epoch": 2449, "lr": 9.284005392066941e-06}
{"train_loss": 0.027620485052466393, "global_step": 218043, "epoch": 2449, "lr": 9.283668906213599e-06}
{"train_loss": 0.07620993256568909, "global_step": 218044, "epoch": 2449, "lr": 9.283332425834063e-06}
{"train_loss": 0.0750378891825676, "global_step": 218045, "epoch": 2449, "lr": 9.282995950928391e-06}
{"train_loss": 0.03141145408153534, "global_step": 218046, "epoch": 2449, "lr": 9.282659481496619e-06}
{"train_loss": 0.0566658079624176, "global_step": 218047, "epoch": 2449, "lr": 9.282323017538802e-06}
{"train_loss": 0.04869476705789566, "global_step": 218048, "epoch": 2449, "lr": 9.281986559054973e-06}
{"train_loss": 0.056113250501286446, "global_step": 218049, "epoch": 2449, "lr": 9.281650106045197e-06, "val_loss": 8.905315399169922}
{"train_loss": 0.09528202563524246, "global_step": 218050, "epoch": 2450, "lr": 9.281313658509494e-06}
{"train_loss": 0.029889678582549095, "global_step": 218051, "epoch": 2450, "lr": 9.280977216447933e-06}
{"train_loss": 0.03127634525299072, "global_step": 218052, "epoch": 2450, "lr": 9.28064077986054e-06}
{"train_loss": 0.0605895109474659, "global_step": 218053, "epoch": 2450, "lr": 9.280304348747377e-06}
{"train_loss": 0.04614030197262764, "global_step": 218054, "epoch": 2450, "lr": 9.279967923108468e-06}
{"train_loss": 0.04387689009308815, "global_step": 218055, "epoch": 2450, "lr": 9.279631502943881e-06}
{"train_loss": 0.06705113500356674, "global_step": 218056, "epoch": 2450, "lr": 9.279295088253643e-06}
{"train_loss": 0.03943512216210365, "global_step": 218057, "epoch": 2450, "lr": 9.278958679037818e-06}
{"train_loss": 0.01911887153983116, "global_step": 218058, "epoch": 2450, "lr": 9.278622275296445e-06}
{"train_loss": 0.07521543651819229, "global_step": 218059, "epoch": 2450, "lr": 9.278285877029546e-06}
{"train_loss": 0.07634023576974869, "global_step": 218060, "epoch": 2450, "lr": 9.277949484237203e-06}
{"train_loss": 0.02333931252360344, "global_step": 218061, "epoch": 2450, "lr": 9.27761309691943e-06}
{"train_loss": 0.07426615059375763, "global_step": 218062, "epoch": 2450, "lr": 9.2772767150763e-06}
{"train_loss": 0.017648078501224518, "global_step": 218063, "epoch": 2450, "lr": 9.276940338707824e-06}
{"train_loss": 0.022807830944657326, "global_step": 218064, "epoch": 2450, "lr": 9.276603967814084e-06}
{"train_loss": 0.04725048691034317, "global_step": 218065, "epoch": 2450, "lr": 9.276267602395095e-06}
{"train_loss": 0.041478224098682404, "global_step": 218066, "epoch": 2450, "lr": 9.27593124245093e-06}
{"train_loss": 0.036944590508937836, "global_step": 218067, "epoch": 2450, "lr": 9.275594887981603e-06}
{"train_loss": 0.02416335605084896, "global_step": 218068, "epoch": 2450, "lr": 9.27525853898719e-06}
{"train_loss": 0.08222519606351852, "global_step": 218069, "epoch": 2450, "lr": 9.274922195467716e-06}
{"train_loss": 0.12426324933767319, "global_step": 218070, "epoch": 2450, "lr": 9.27458585742324e-06}
{"train_loss": 0.065603107213974, "global_step": 218071, "epoch": 2450, "lr": 9.274249524853784e-06}
{"train_loss": 0.029819471761584282, "global_step": 218072, "epoch": 2450, "lr": 9.27391319775941e-06}
{"train_loss": 0.10122215002775192, "global_step": 218073, "epoch": 2450, "lr": 9.273576876140177e-06}
{"train_loss": 0.045975856482982635, "global_step": 218074, "epoch": 2450, "lr": 9.273240559996105e-06}
{"train_loss": 0.030014805495738983, "global_step": 218075, "epoch": 2450, "lr": 9.272904249327258e-06}
{"train_loss": 0.06531067937612534, "global_step": 218076, "epoch": 2450, "lr": 9.272567944133658e-06}
{"train_loss": 0.03889879956841469, "global_step": 218077, "epoch": 2450, "lr": 9.27223164441538e-06}
{"train_loss": 0.04624307528138161, "global_step": 218078, "epoch": 2450, "lr": 9.271895350172444e-06}
{"train_loss": 0.07757914066314697, "global_step": 218079, "epoch": 2450, "lr": 9.271559061404916e-06}
{"train_loss": 0.07620859146118164, "global_step": 218080, "epoch": 2450, "lr": 9.271222778112814e-06}
{"train_loss": 0.034170858561992645, "global_step": 218081, "epoch": 2450, "lr": 9.270886500296217e-06}
{"train_loss": 0.036501213908195496, "global_step": 218082, "epoch": 2450, "lr": 9.270550227955139e-06}
{"train_loss": 0.05398331210017204, "global_step": 218083, "epoch": 2450, "lr": 9.270213961089652e-06}
{"train_loss": 0.02494823932647705, "global_step": 218084, "epoch": 2450, "lr": 9.269877699699775e-06}
{"train_loss": 0.04886801540851593, "global_step": 218085, "epoch": 2450, "lr": 9.26954144378558e-06}
{"train_loss": 0.03854142129421234, "global_step": 218086, "epoch": 2450, "lr": 9.269205193347086e-06}
{"train_loss": 0.0936904177069664, "global_step": 218087, "epoch": 2450, "lr": 9.268868948384362e-06}
{"train_loss": 0.07505787909030914, "global_step": 218088, "epoch": 2450, "lr": 9.268532708897432e-06}
{"train_loss": 0.022343479096889496, "global_step": 218089, "epoch": 2450, "lr": 9.26819647488636e-06}
{"train_loss": 0.05970784276723862, "global_step": 218090, "epoch": 2450, "lr": 9.267860246351173e-06}
{"train_loss": 0.06139111518859863, "global_step": 218091, "epoch": 2450, "lr": 9.267524023291935e-06}
{"train_loss": 0.05699196457862854, "global_step": 218092, "epoch": 2450, "lr": 9.26718780570867e-06}
{"train_loss": 0.016846075654029846, "global_step": 218093, "epoch": 2450, "lr": 9.266851593601455e-06}
{"train_loss": 0.04126672446727753, "global_step": 218094, "epoch": 2450, "lr": 9.266515386970298e-06}
{"train_loss": 0.06386669725179672, "global_step": 218095, "epoch": 2450, "lr": 9.266179185815271e-06}
{"train_loss": 0.08561854064464569, "global_step": 218096, "epoch": 2450, "lr": 9.265842990136414e-06}
{"train_loss": 0.0161396823823452, "global_step": 218097, "epoch": 2450, "lr": 9.265506799933754e-06}
{"train_loss": 0.04376772418618202, "global_step": 218098, "epoch": 2450, "lr": 9.265170615207363e-06}
{"train_loss": 0.058813292533159256, "global_step": 218099, "epoch": 2450, "lr": 9.26483443595726e-06}
{"train_loss": 0.09360068291425705, "global_step": 218100, "epoch": 2450, "lr": 9.264498262183513e-06}
{"train_loss": 0.04136025905609131, "global_step": 218101, "epoch": 2450, "lr": 9.264162093886148e-06}
{"train_loss": 0.10387339442968369, "global_step": 218102, "epoch": 2450, "lr": 9.263825931065234e-06}
{"train_loss": 0.014410704374313354, "global_step": 218103, "epoch": 2450, "lr": 9.26348977372079e-06}
{"train_loss": 0.022795524448156357, "global_step": 218104, "epoch": 2450, "lr": 9.263153621852883e-06}
{"train_loss": 0.0911969393491745, "global_step": 218105, "epoch": 2450, "lr": 9.262817475461539e-06}
{"train_loss": 0.08039697259664536, "global_step": 218106, "epoch": 2450, "lr": 9.262481334546824e-06}
{"train_loss": 0.021283982321619987, "global_step": 218107, "epoch": 2450, "lr": 9.262145199108757e-06}
{"train_loss": 0.07335252314805984, "global_step": 218108, "epoch": 2450, "lr": 9.261809069147409e-06}
{"train_loss": 0.03988925740122795, "global_step": 218109, "epoch": 2450, "lr": 9.261472944662808e-06}
{"train_loss": 0.039833880960941315, "global_step": 218110, "epoch": 2450, "lr": 9.261136825655003e-06}
{"train_loss": 0.0821654424071312, "global_step": 218111, "epoch": 2450, "lr": 9.260800712124051e-06}
{"train_loss": 0.08488177508115768, "global_step": 218112, "epoch": 2450, "lr": 9.26046460406998e-06}
{"train_loss": 0.07351209968328476, "global_step": 218113, "epoch": 2450, "lr": 9.260128501492854e-06}
{"train_loss": 0.09618271142244339, "global_step": 218114, "epoch": 2450, "lr": 9.259792404392693e-06}
{"train_loss": 0.05844494327902794, "global_step": 218115, "epoch": 2450, "lr": 9.259456312769572e-06}
{"train_loss": 0.04272855818271637, "global_step": 218116, "epoch": 2450, "lr": 9.259120226623503e-06}
{"train_loss": 0.01761776953935623, "global_step": 218117, "epoch": 2450, "lr": 9.258784145954563e-06}
{"train_loss": 0.02920408733189106, "global_step": 218118, "epoch": 2450, "lr": 9.258448070762771e-06}
{"train_loss": 0.01934473030269146, "global_step": 218119, "epoch": 2450, "lr": 9.258112001048196e-06}
{"train_loss": 0.027656257152557373, "global_step": 218120, "epoch": 2450, "lr": 9.257775936810863e-06}
{"train_loss": 0.04270510748028755, "global_step": 218121, "epoch": 2450, "lr": 9.257439878050834e-06}
{"train_loss": 0.060721974819898605, "global_step": 218122, "epoch": 2450, "lr": 9.257103824768137e-06}
{"train_loss": 0.06924301385879517, "global_step": 218123, "epoch": 2450, "lr": 9.256767776962832e-06}
{"train_loss": 0.03114907443523407, "global_step": 218124, "epoch": 2450, "lr": 9.256431734634951e-06}
{"train_loss": 0.09044855087995529, "global_step": 218125, "epoch": 2450, "lr": 9.256095697784555e-06}
{"train_loss": 0.06139708310365677, "global_step": 218126, "epoch": 2450, "lr": 9.255759666411667e-06}
{"train_loss": 0.048095326870679855, "global_step": 218127, "epoch": 2450, "lr": 9.255423640516353e-06}
{"train_loss": 0.056149110198020935, "global_step": 218128, "epoch": 2450, "lr": 9.255087620098657e-06}
{"train_loss": 0.024448690935969353, "global_step": 218129, "epoch": 2450, "lr": 9.254751605158607e-06}
{"train_loss": 0.03239181637763977, "global_step": 218130, "epoch": 2450, "lr": 9.25441559569627e-06}
{"train_loss": 0.078346848487854, "global_step": 218131, "epoch": 2450, "lr": 9.254079591711678e-06}
{"train_loss": 0.023882322013378143, "global_step": 218132, "epoch": 2450, "lr": 9.253743593204873e-06}
{"train_loss": 0.025020137429237366, "global_step": 218133, "epoch": 2450, "lr": 9.253407600175912e-06}
{"train_loss": 0.036931443959474564, "global_step": 218134, "epoch": 2450, "lr": 9.253071612624824e-06}
{"train_loss": 0.04819830134510994, "global_step": 218135, "epoch": 2450, "lr": 9.252735630551673e-06}
{"train_loss": 0.1138644814491272, "global_step": 218136, "epoch": 2450, "lr": 9.2523996539565e-06}
{"train_loss": 0.05258253961801529, "global_step": 218137, "epoch": 2450, "lr": 9.252063682839328e-06}
{"train_loss": 0.05288219903961996, "global_step": 218138, "epoch": 2450, "lr": 9.251727717200237e-06, "train/sim_max_reward_0": 0.5479566092312659, "train/sim_max_reward_1": 1.0, "train/sim_max_reward_2": 1.0, "train/sim_max_reward_3": 0.667589630900224, "train/sim_max_reward_4": 0.9754046026196944, "train/sim_max_reward_5": 0.9577600044112377, "test/sim_max_reward_4500000": 0.40782174138481136, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.0010873035596013336, "test/sim_max_reward_4500003": 0.419767835558792, "test/sim_max_reward_4500004": 0.654665077766219, "test/sim_max_reward_4500005": 0.9766009494802624, "test/sim_max_reward_4500006": 0.8376011161285227, "test/sim_max_reward_4500007": 0.9606865090993764, "test/sim_max_reward_4500008": 0.3200312890959637, "test/sim_max_reward_4500009": 0.04996892848898143, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.8650315625084912, "test/sim_max_reward_4500013": 0.4841606731871254, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9436878296189171, "test/sim_max_reward_4500016": 0.2031230781924523, "test/sim_max_reward_4500017": 1.0, "test/sim_max_reward_4500018": 0.408326335360322, "test/sim_max_reward_4500019": 0.39498001003589106, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9613195072207562, "test/sim_max_reward_4500022": 0.926637359141918, "test/sim_max_reward_4500023": 0.9497164981164373, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.41874087317238257, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 0.9618404155951813, "test/sim_max_reward_4500028": 0.9788914682673887, "test/sim_max_reward_4500029": 0.423690569545515, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.40680567365544956, "test/sim_max_reward_4500033": 0.932152185380134, "test/sim_max_reward_4500034": 0.939847427483378, "test/sim_max_reward_4500035": 0.733017724626116, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.9175104483067344, "test/sim_max_reward_4500038": 0.1621177317927209, "test/sim_max_reward_4500039": 0.995237100546938, "test/sim_max_reward_4500040": 0.8052544798645193, "test/sim_max_reward_4500041": 0.5715987800711693, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.24679314898924434, "test/sim_max_reward_4500044": 0.6914342214626226, "test/sim_max_reward_4500045": 0.9835214020687548, "test/sim_max_reward_4500046": 0.4930632209164596, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9992812398292754, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.8581184745270704, "test/mean_score": 0.6085202343103765, "val_loss": 8.79011344909668, "train_action_mse_error": 15.23810863494873}
{"train_loss": 0.03286126255989075, "global_step": 218139, "epoch": 2451, "lr": 9.25139175703924e-06}
{"train_loss": 0.027070825919508934, "global_step": 218140, "epoch": 2451, "lr": 9.251055802356407e-06}
{"train_loss": 0.06359591335058212, "global_step": 218141, "epoch": 2451, "lr": 9.250719853151762e-06}
{"train_loss": 0.08661625534296036, "global_step": 218142, "epoch": 2451, "lr": 9.250383909425376e-06}
{"train_loss": 0.1217857301235199, "global_step": 218143, "epoch": 2451, "lr": 9.250047971177262e-06}
{"train_loss": 0.10273614525794983, "global_step": 218144, "epoch": 2451, "lr": 9.2497120384075e-06}
{"train_loss": 0.03780120238661766, "global_step": 218145, "epoch": 2451, "lr": 9.249376111116099e-06}
{"train_loss": 0.05686556175351143, "global_step": 218146, "epoch": 2451, "lr": 9.249040189303126e-06}
{"train_loss": 0.04832462966442108, "global_step": 218147, "epoch": 2451, "lr": 9.248704272968633e-06}
{"train_loss": 0.03331318125128746, "global_step": 218148, "epoch": 2451, "lr": 9.248368362112647e-06}
{"train_loss": 0.022619184106588364, "global_step": 218149, "epoch": 2451, "lr": 9.248032456735228e-06}
{"train_loss": 0.04463435336947441, "global_step": 218150, "epoch": 2451, "lr": 9.247696556836405e-06}
{"train_loss": 0.04459303990006447, "global_step": 218151, "epoch": 2451, "lr": 9.24736066241625e-06}
{"train_loss": 0.050161756575107574, "global_step": 218152, "epoch": 2451, "lr": 9.247024773474772e-06}
{"train_loss": 0.07078517973423004, "global_step": 218153, "epoch": 2451, "lr": 9.246688890012046e-06}
{"train_loss": 0.08868799358606339, "global_step": 218154, "epoch": 2451, "lr": 9.246353012028096e-06}
{"train_loss": 0.015473449602723122, "global_step": 218155, "epoch": 2451, "lr": 9.246017139522995e-06}
{"train_loss": 0.06131080165505409, "global_step": 218156, "epoch": 2451, "lr": 9.245681272496748e-06}
{"train_loss": 0.09017135947942734, "global_step": 218157, "epoch": 2451, "lr": 9.245345410949441e-06}
{"train_loss": 0.06909779459238052, "global_step": 218158, "epoch": 2451, "lr": 9.245009554881084e-06}
{"train_loss": 0.035881832242012024, "global_step": 218159, "epoch": 2451, "lr": 9.244673704291757e-06}
{"train_loss": 0.037408433854579926, "global_step": 218160, "epoch": 2451, "lr": 9.244337859181468e-06}
{"train_loss": 0.026185132563114166, "global_step": 218161, "epoch": 2451, "lr": 9.244002019550297e-06}
{"train_loss": 0.08500979095697403, "global_step": 218162, "epoch": 2451, "lr": 9.243666185398264e-06}
{"train_loss": 0.054208915680646896, "global_step": 218163, "epoch": 2451, "lr": 9.243330356725416e-06}
{"train_loss": 0.060686081647872925, "global_step": 218164, "epoch": 2451, "lr": 9.242994533531823e-06}
{"train_loss": 0.09797000139951706, "global_step": 218165, "epoch": 2451, "lr": 9.242658715817498e-06}
{"train_loss": 0.03182782977819443, "global_step": 218166, "epoch": 2451, "lr": 9.242322903582518e-06}
{"train_loss": 0.12166967988014221, "global_step": 218167, "epoch": 2451, "lr": 9.241987096826893e-06}
{"train_loss": 0.047800179570913315, "global_step": 218168, "epoch": 2451, "lr": 9.241651295550696e-06}
{"train_loss": 0.05636502057313919, "global_step": 218169, "epoch": 2451, "lr": 9.241315499753972e-06}
{"train_loss": 0.08311128616333008, "global_step": 218170, "epoch": 2451, "lr": 9.240979709436737e-06}
{"train_loss": 0.04325331002473831, "global_step": 218171, "epoch": 2451, "lr": 9.240643924599068e-06}
{"train_loss": 0.04313125088810921, "global_step": 218172, "epoch": 2451, "lr": 9.240308145240984e-06}
{"train_loss": 0.04883041977882385, "global_step": 218173, "epoch": 2451, "lr": 9.23997237136256e-06}
{"train_loss": 0.05448535457253456, "global_step": 218174, "epoch": 2451, "lr": 9.239636602963824e-06}
{"train_loss": 0.11160492897033691, "global_step": 218175, "epoch": 2451, "lr": 9.239300840044807e-06}
{"train_loss": 0.015331166796386242, "global_step": 218176, "epoch": 2451, "lr": 9.238965082605589e-06}
{"train_loss": 0.07694512605667114, "global_step": 218177, "epoch": 2451, "lr": 9.238629330646176e-06}
{"train_loss": 0.027700142934918404, "global_step": 218178, "epoch": 2451, "lr": 9.238293584166647e-06}
{"train_loss": 0.0267178937792778, "global_step": 218179, "epoch": 2451, "lr": 9.237957843167017e-06}
{"train_loss": 0.05888812243938446, "global_step": 218180, "epoch": 2451, "lr": 9.237622107647365e-06}
{"train_loss": 0.05552850663661957, "global_step": 218181, "epoch": 2451, "lr": 9.237286377607702e-06}
{"train_loss": 0.03790571913123131, "global_step": 218182, "epoch": 2451, "lr": 9.236950653048087e-06}
{"train_loss": 0.11649040132761002, "global_step": 218183, "epoch": 2451, "lr": 9.23661493396858e-06}
{"train_loss": 0.07238531857728958, "global_step": 218184, "epoch": 2451, "lr": 9.236279220369203e-06}
{"train_loss": 0.0702173188328743, "global_step": 218185, "epoch": 2451, "lr": 9.235943512250022e-06}
{"train_loss": 0.024560490623116493, "global_step": 218186, "epoch": 2451, "lr": 9.235607809611063e-06}
{"train_loss": 0.031780317425727844, "global_step": 218187, "epoch": 2451, "lr": 9.235272112452392e-06}
{"train_loss": 0.03613489866256714, "global_step": 218188, "epoch": 2451, "lr": 9.234936420774027e-06}
{"train_loss": 0.0515933558344841, "global_step": 218189, "epoch": 2451, "lr": 9.234600734576038e-06}
{"train_loss": 0.050133612006902695, "global_step": 218190, "epoch": 2451, "lr": 9.234265053858454e-06}
{"train_loss": 0.0707075297832489, "global_step": 218191, "epoch": 2451, "lr": 9.233929378621331e-06}
{"train_loss": 0.04902658239006996, "global_step": 218192, "epoch": 2451, "lr": 9.233593708864703e-06}
{"train_loss": 0.06613563746213913, "global_step": 218193, "epoch": 2451, "lr": 9.233258044588633e-06}
{"train_loss": 0.08239902555942535, "global_step": 218194, "epoch": 2451, "lr": 9.232922385793136e-06}
{"train_loss": 0.05454300343990326, "global_step": 218195, "epoch": 2451, "lr": 9.232586732478292e-06}
{"train_loss": 0.03128271549940109, "global_step": 218196, "epoch": 2451, "lr": 9.232251084644122e-06}
{"train_loss": 0.06981794536113739, "global_step": 218197, "epoch": 2451, "lr": 9.23191544229069e-06}
{"train_loss": 0.06844877451658249, "global_step": 218198, "epoch": 2451, "lr": 9.231579805418011e-06}
{"train_loss": 0.13599923253059387, "global_step": 218199, "epoch": 2451, "lr": 9.231244174026154e-06}
{"train_loss": 0.032895058393478394, "global_step": 218200, "epoch": 2451, "lr": 9.230908548115175e-06}
{"train_loss": 0.06807012110948563, "global_step": 218201, "epoch": 2451, "lr": 9.230572927685083e-06}
{"train_loss": 0.06936433166265488, "global_step": 218202, "epoch": 2451, "lr": 9.230237312735961e-06}
{"train_loss": 0.051663171499967575, "global_step": 218203, "epoch": 2451, "lr": 9.229901703267824e-06}
{"train_loss": 0.06416840106248856, "global_step": 218204, "epoch": 2451, "lr": 9.22956609928074e-06}
{"train_loss": 0.04205729812383652, "global_step": 218205, "epoch": 2451, "lr": 9.229230500774738e-06}
{"train_loss": 0.03733568638563156, "global_step": 218206, "epoch": 2451, "lr": 9.228894907749874e-06}
{"train_loss": 0.04344620183110237, "global_step": 218207, "epoch": 2451, "lr": 9.228559320206192e-06}
{"train_loss": 0.07269549369812012, "global_step": 218208, "epoch": 2451, "lr": 9.228223738143722e-06}
{"train_loss": 0.03913608938455582, "global_step": 218209, "epoch": 2451, "lr": 9.227888161562526e-06}
{"train_loss": 0.06719130277633667, "global_step": 218210, "epoch": 2451, "lr": 9.227552590462636e-06}
{"train_loss": 0.04353408142924309, "global_step": 218211, "epoch": 2451, "lr": 9.22721702484412e-06}
{"train_loss": 0.04921247437596321, "global_step": 218212, "epoch": 2451, "lr": 9.226881464707004e-06}
{"train_loss": 0.038483742624521255, "global_step": 218213, "epoch": 2451, "lr": 9.226545910051321e-06}
{"train_loss": 0.07669533044099808, "global_step": 218214, "epoch": 2451, "lr": 9.226210360877153e-06}
{"train_loss": 0.04581158980727196, "global_step": 218215, "epoch": 2451, "lr": 9.225874817184504e-06}
{"train_loss": 0.025029338896274567, "global_step": 218216, "epoch": 2451, "lr": 9.225539278973456e-06}
{"train_loss": 0.032157860696315765, "global_step": 218217, "epoch": 2451, "lr": 9.225203746244022e-06}
{"train_loss": 0.052854716777801514, "global_step": 218218, "epoch": 2451, "lr": 9.224868218996263e-06}
{"train_loss": 0.05930464714765549, "global_step": 218219, "epoch": 2451, "lr": 9.224532697230238e-06}
{"train_loss": 0.06877364963293076, "global_step": 218220, "epoch": 2451, "lr": 9.224197180945959e-06}
{"train_loss": 0.027200443670153618, "global_step": 218221, "epoch": 2451, "lr": 9.223861670143508e-06}
{"train_loss": 0.07013723254203796, "global_step": 218222, "epoch": 2451, "lr": 9.2235261648229e-06}
{"train_loss": 0.0566193051636219, "global_step": 218223, "epoch": 2451, "lr": 9.223190664984204e-06}
{"train_loss": 0.05878951773047447, "global_step": 218224, "epoch": 2451, "lr": 9.222855170627436e-06}
{"train_loss": 0.04245469719171524, "global_step": 218225, "epoch": 2451, "lr": 9.222519681752673e-06}
{"train_loss": 0.044441673904657364, "global_step": 218226, "epoch": 2451, "lr": 9.222184198359935e-06}
{"train_loss": 0.05648448633218414, "global_step": 218227, "epoch": 2451, "lr": 9.22184872044929e-06, "val_loss": 8.797581672668457}
{"train_loss": 0.04179142788052559, "global_step": 218228, "epoch": 2452, "lr": 9.221513248020753e-06}
{"train_loss": 0.0215594545006752, "global_step": 218229, "epoch": 2452, "lr": 9.221177781074403e-06}
{"train_loss": 0.04333268851041794, "global_step": 218230, "epoch": 2452, "lr": 9.220842319610257e-06}
{"train_loss": 0.03493184968829155, "global_step": 218231, "epoch": 2452, "lr": 9.220506863628382e-06}
{"train_loss": 0.09240657836198807, "global_step": 218232, "epoch": 2452, "lr": 9.220171413128803e-06}
{"train_loss": 0.038018181920051575, "global_step": 218233, "epoch": 2452, "lr": 9.219835968111584e-06}
{"train_loss": 0.04739386588335037, "global_step": 218234, "epoch": 2452, "lr": 9.219500528576747e-06}
{"train_loss": 0.04473040625452995, "global_step": 218235, "epoch": 2452, "lr": 9.219165094524368e-06}
{"train_loss": 0.03437843173742294, "global_step": 218236, "epoch": 2452, "lr": 9.218829665954459e-06}
{"train_loss": 0.015231043100357056, "global_step": 218237, "epoch": 2452, "lr": 9.218494242867087e-06}
{"train_loss": 0.06712452322244644, "global_step": 218238, "epoch": 2452, "lr": 9.218158825262296e-06}
{"train_loss": 0.030595878139138222, "global_step": 218239, "epoch": 2452, "lr": 9.21782341314012e-06}
{"train_loss": 0.04586898535490036, "global_step": 218240, "epoch": 2452, "lr": 9.217488006500624e-06}
{"train_loss": 0.053262174129486084, "global_step": 218241, "epoch": 2452, "lr": 9.217152605343821e-06}
{"train_loss": 0.09615906327962875, "global_step": 218242, "epoch": 2452, "lr": 9.216817209669792e-06}
{"train_loss": 0.044193387031555176, "global_step": 218243, "epoch": 2452, "lr": 9.21648181947855e-06}
{"train_loss": 0.0688183531165123, "global_step": 218244, "epoch": 2452, "lr": 9.216146434770167e-06}
{"train_loss": 0.0401165634393692, "global_step": 218245, "epoch": 2452, "lr": 9.215811055544666e-06}
{"train_loss": 0.06399040669202805, "global_step": 218246, "epoch": 2452, "lr": 9.215475681802117e-06}
{"train_loss": 0.08378176391124725, "global_step": 218247, "epoch": 2452, "lr": 9.215140313542548e-06}
{"train_loss": 0.02217850089073181, "global_step": 218248, "epoch": 2452, "lr": 9.214804950765992e-06}
{"train_loss": 0.07524475455284119, "global_step": 218249, "epoch": 2452, "lr": 9.21446959347252e-06}
{"train_loss": 0.04806733876466751, "global_step": 218250, "epoch": 2452, "lr": 9.214134241662154e-06}
{"train_loss": 0.06983568519353867, "global_step": 218251, "epoch": 2452, "lr": 9.213798895334963e-06}
{"train_loss": 0.050376053899526596, "global_step": 218252, "epoch": 2452, "lr": 9.21346355449098e-06}
{"train_loss": 0.015643570572137833, "global_step": 218253, "epoch": 2452, "lr": 9.21312821913024e-06}
{"train_loss": 0.052414607256650925, "global_step": 218254, "epoch": 2452, "lr": 9.212792889252792e-06}
{"train_loss": 0.08025479316711426, "global_step": 218255, "epoch": 2452, "lr": 9.212457564858706e-06}
{"train_loss": 0.05903613939881325, "global_step": 218256, "epoch": 2452, "lr": 9.212122245947991e-06}
{"train_loss": 0.0317130945622921, "global_step": 218257, "epoch": 2452, "lr": 9.211786932520722e-06}
{"train_loss": 0.047263551503419876, "global_step": 218258, "epoch": 2452, "lr": 9.211451624576917e-06}
{"train_loss": 0.044633444398641586, "global_step": 218259, "epoch": 2452, "lr": 9.211116322116653e-06}
{"train_loss": 0.022237688302993774, "global_step": 218260, "epoch": 2452, "lr": 9.210781025139943e-06}
{"train_loss": 0.07202053815126419, "global_step": 218261, "epoch": 2452, "lr": 9.210445733646856e-06}
{"train_loss": 0.04423528164625168, "global_step": 218262, "epoch": 2452, "lr": 9.210110447637415e-06}
{"train_loss": 0.06497158855199814, "global_step": 218263, "epoch": 2452, "lr": 9.209775167111689e-06}
{"train_loss": 0.05017739161849022, "global_step": 218264, "epoch": 2452, "lr": 9.209439892069705e-06}
{"train_loss": 0.07026400417089462, "global_step": 218265, "epoch": 2452, "lr": 9.209104622511522e-06}
{"train_loss": 0.06333059072494507, "global_step": 218266, "epoch": 2452, "lr": 9.208769358437163e-06}
{"train_loss": 0.04953029006719589, "global_step": 218267, "epoch": 2452, "lr": 9.208434099846708e-06}
{"train_loss": 0.05175582319498062, "global_step": 218268, "epoch": 2452, "lr": 9.208098846740166e-06}
{"train_loss": 0.06691353768110275, "global_step": 218269, "epoch": 2452, "lr": 9.207763599117609e-06}
{"train_loss": 0.05285903066396713, "global_step": 218270, "epoch": 2452, "lr": 9.207428356979058e-06}
{"train_loss": 0.05484982952475548, "global_step": 218271, "epoch": 2452, "lr": 9.207093120324584e-06}
{"train_loss": 0.041272636502981186, "global_step": 218272, "epoch": 2452, "lr": 9.20675788915421e-06}
{"train_loss": 0.024613026529550552, "global_step": 218273, "epoch": 2452, "lr": 9.206422663467984e-06}
{"train_loss": 0.07534560561180115, "global_step": 218274, "epoch": 2452, "lr": 9.206087443265976e-06}
{"train_loss": 0.08755034953355789, "global_step": 218275, "epoch": 2452, "lr": 9.205752228548198e-06}
{"train_loss": 0.06609389185905457, "global_step": 218276, "epoch": 2452, "lr": 9.205417019314721e-06}
{"train_loss": 0.07886230945587158, "global_step": 218277, "epoch": 2452, "lr": 9.205081815565564e-06}
{"train_loss": 0.0753873959183693, "global_step": 218278, "epoch": 2452, "lr": 9.204746617300803e-06}
{"train_loss": 0.06171320751309395, "global_step": 218279, "epoch": 2452, "lr": 9.204411424520455e-06}
{"train_loss": 0.04539817199110985, "global_step": 218280, "epoch": 2452, "lr": 9.204076237224585e-06}
{"train_loss": 0.08542915433645248, "global_step": 218281, "epoch": 2452, "lr": 9.203741055413217e-06}
{"train_loss": 0.07256270945072174, "global_step": 218282, "epoch": 2452, "lr": 9.203405879086425e-06}
{"train_loss": 0.05125465616583824, "global_step": 218283, "epoch": 2452, "lr": 9.203070708244222e-06}
{"train_loss": 0.039527326822280884, "global_step": 218284, "epoch": 2452, "lr": 9.202735542886682e-06}
{"train_loss": 0.042097751051187515, "global_step": 218285, "epoch": 2452, "lr": 9.202400383013837e-06}
{"train_loss": 0.02161242999136448, "global_step": 218286, "epoch": 2452, "lr": 9.202065228625723e-06}
{"train_loss": 0.029529310762882233, "global_step": 218287, "epoch": 2452, "lr": 9.201730079722403e-06}
{"train_loss": 0.037394050508737564, "global_step": 218288, "epoch": 2452, "lr": 9.201394936303898e-06}
{"train_loss": 0.058510515838861465, "global_step": 218289, "epoch": 2452, "lr": 9.201059798370288e-06}
{"train_loss": 0.041283704340457916, "global_step": 218290, "epoch": 2452, "lr": 9.200724665921578e-06}
{"train_loss": 0.1188848465681076, "global_step": 218291, "epoch": 2452, "lr": 9.20038953895785e-06}
{"train_loss": 0.06147165596485138, "global_step": 218292, "epoch": 2452, "lr": 9.20005441747912e-06}
{"train_loss": 0.037468552589416504, "global_step": 218293, "epoch": 2452, "lr": 9.19971930148546e-06}
{"train_loss": 0.07779524475336075, "global_step": 218294, "epoch": 2452, "lr": 9.199384190976885e-06}
{"train_loss": 0.09070304036140442, "global_step": 218295, "epoch": 2452, "lr": 9.199049085953465e-06}
{"train_loss": 0.062166355550289154, "global_step": 218296, "epoch": 2452, "lr": 9.19871398641523e-06}
{"train_loss": 0.07463505119085312, "global_step": 218297, "epoch": 2452, "lr": 9.198378892362235e-06}
{"train_loss": 0.03361906111240387, "global_step": 218298, "epoch": 2452, "lr": 9.198043803794516e-06}
{"train_loss": 0.040306128561496735, "global_step": 218299, "epoch": 2452, "lr": 9.197708720712129e-06}
{"train_loss": 0.04057026654481888, "global_step": 218300, "epoch": 2452, "lr": 9.197373643115104e-06}
{"train_loss": 0.046212729066610336, "global_step": 218301, "epoch": 2452, "lr": 9.197038571003508e-06}
{"train_loss": 0.060889724642038345, "global_step": 218302, "epoch": 2452, "lr": 9.196703504377357e-06}
{"train_loss": 0.08290789276361465, "global_step": 218303, "epoch": 2452, "lr": 9.196368443236725e-06}
{"train_loss": 0.06164347752928734, "global_step": 218304, "epoch": 2452, "lr": 9.196033387581631e-06}
{"train_loss": 0.027462273836135864, "global_step": 218305, "epoch": 2452, "lr": 9.195698337412151e-06}
{"train_loss": 0.07611382752656937, "global_step": 218306, "epoch": 2452, "lr": 9.195363292728294e-06}
{"train_loss": 0.044089484959840775, "global_step": 218307, "epoch": 2452, "lr": 9.195028253530135e-06}
{"train_loss": 0.09771338105201721, "global_step": 218308, "epoch": 2452, "lr": 9.194693219817702e-06}
{"train_loss": 0.04203896224498749, "global_step": 218309, "epoch": 2452, "lr": 9.194358191591034e-06}
{"train_loss": 0.05145428702235222, "global_step": 218310, "epoch": 2452, "lr": 9.194023168850207e-06}
{"train_loss": 0.030610430985689163, "global_step": 218311, "epoch": 2452, "lr": 9.193688151595237e-06}
{"train_loss": 0.08406370133161545, "global_step": 218312, "epoch": 2452, "lr": 9.193353139826184e-06}
{"train_loss": 0.05564161762595177, "global_step": 218313, "epoch": 2452, "lr": 9.193018133543075e-06}
{"train_loss": 0.03869419917464256, "global_step": 218314, "epoch": 2452, "lr": 9.192683132745983e-06}
{"train_loss": 0.033395297825336456, "global_step": 218315, "epoch": 2452, "lr": 9.192348137434926e-06}
{"train_loss": 0.055203402519560935, "global_step": 218316, "epoch": 2452, "lr": 9.192013147609969e-06, "val_loss": 8.830221176147461}
{"train_loss": 0.04401657357811928, "global_step": 218317, "epoch": 2453, "lr": 9.191678163271134e-06}
{"train_loss": 0.09443651139736176, "global_step": 218318, "epoch": 2453, "lr": 9.191343184418499e-06}
{"train_loss": 0.03099966049194336, "global_step": 218319, "epoch": 2453, "lr": 9.191008211052076e-06}
{"train_loss": 0.019613191485404968, "global_step": 218320, "epoch": 2453, "lr": 9.190673243171937e-06}
{"train_loss": 0.04599408432841301, "global_step": 218321, "epoch": 2453, "lr": 9.1903382807781e-06}
{"train_loss": 0.0833146795630455, "global_step": 218322, "epoch": 2453, "lr": 9.190003323870638e-06}
{"train_loss": 0.03257396072149277, "global_step": 218323, "epoch": 2453, "lr": 9.189668372449584e-06}
{"train_loss": 0.03508453071117401, "global_step": 218324, "epoch": 2453, "lr": 9.189333426514967e-06}
{"train_loss": 0.050041306763887405, "global_step": 218325, "epoch": 2453, "lr": 9.188998486066862e-06}
{"train_loss": 0.029351772740483284, "global_step": 218326, "epoch": 2453, "lr": 9.188663551105286e-06}
{"train_loss": 0.03587259352207184, "global_step": 218327, "epoch": 2453, "lr": 9.18832862163031e-06}
{"train_loss": 0.024149566888809204, "global_step": 218328, "epoch": 2453, "lr": 9.18799369764195e-06}
{"train_loss": 0.05187901854515076, "global_step": 218329, "epoch": 2453, "lr": 9.187658779140279e-06}
{"train_loss": 0.0454864576458931, "global_step": 218330, "epoch": 2453, "lr": 9.18732386612532e-06}
{"train_loss": 0.04965127259492874, "global_step": 218331, "epoch": 2453, "lr": 9.186988958597143e-06}
{"train_loss": 0.058603428304195404, "global_step": 218332, "epoch": 2453, "lr": 9.18665405655576e-06}
{"train_loss": 0.007414640858769417, "global_step": 218333, "epoch": 2453, "lr": 9.18631916000125e-06}
{"train_loss": 0.09258189052343369, "global_step": 218334, "epoch": 2453, "lr": 9.185984268933622e-06}
{"train_loss": 0.03599642962217331, "global_step": 218335, "epoch": 2453, "lr": 9.185649383352963e-06}
{"train_loss": 0.05228576436638832, "global_step": 218336, "epoch": 2453, "lr": 9.185314503259279e-06}
{"train_loss": 0.07143440842628479, "global_step": 218337, "epoch": 2453, "lr": 9.184979628652646e-06}
{"train_loss": 0.07575817406177521, "global_step": 218338, "epoch": 2453, "lr": 9.184644759533079e-06}
{"train_loss": 0.020615678280591965, "global_step": 218339, "epoch": 2453, "lr": 9.184309895900656e-06}
{"train_loss": 0.03575802221894264, "global_step": 218340, "epoch": 2453, "lr": 9.183975037755393e-06}
{"train_loss": 0.027446554973721504, "global_step": 218341, "epoch": 2453, "lr": 9.183640185097359e-06}
{"train_loss": 0.016983024775981903, "global_step": 218342, "epoch": 2453, "lr": 9.183305337926569e-06}
{"train_loss": 0.06791188567876816, "global_step": 218343, "epoch": 2453, "lr": 9.182970496243103e-06}
{"train_loss": 0.038273729383945465, "global_step": 218344, "epoch": 2453, "lr": 9.182635660046978e-06}
{"train_loss": 0.07946238666772842, "global_step": 218345, "epoch": 2453, "lr": 9.182300829338248e-06}
{"train_loss": 0.04138675704598427, "global_step": 218346, "epoch": 2453, "lr": 9.181966004116971e-06}
{"train_loss": 0.07846280932426453, "global_step": 218347, "epoch": 2453, "lr": 9.181631184383172e-06}
{"train_loss": 0.05568024516105652, "global_step": 218348, "epoch": 2453, "lr": 9.181296370136915e-06}
{"train_loss": 0.02989637292921543, "global_step": 218349, "epoch": 2453, "lr": 9.180961561378227e-06}
{"train_loss": 0.06259354948997498, "global_step": 218350, "epoch": 2453, "lr": 9.180626758107169e-06}
{"train_loss": 0.058994557708501816, "global_step": 218351, "epoch": 2453, "lr": 9.180291960323766e-06}
{"train_loss": 0.07036685943603516, "global_step": 218352, "epoch": 2453, "lr": 9.179957168028092e-06}
{"train_loss": 0.05871430039405823, "global_step": 218353, "epoch": 2453, "lr": 9.179622381220159e-06}
{"train_loss": 0.08687145262956619, "global_step": 218354, "epoch": 2453, "lr": 9.179287599900044e-06}
{"train_loss": 0.05260119587182999, "global_step": 218355, "epoch": 2453, "lr": 9.17895282406776e-06}
{"train_loss": 0.043504245579242706, "global_step": 218356, "epoch": 2453, "lr": 9.178618053723381e-06}
{"train_loss": 0.0898333415389061, "global_step": 218357, "epoch": 2453, "lr": 9.178283288866929e-06}
{"train_loss": 0.09085364639759064, "global_step": 218358, "epoch": 2453, "lr": 9.17794852949847e-06}
{"train_loss": 0.037052467465400696, "global_step": 218359, "epoch": 2453, "lr": 9.177613775618026e-06}
{"train_loss": 0.04468570649623871, "global_step": 218360, "epoch": 2453, "lr": 9.177279027225672e-06}
{"train_loss": 0.01437669899314642, "global_step": 218361, "epoch": 2453, "lr": 9.176944284321415e-06}
{"train_loss": 0.0870809406042099, "global_step": 218362, "epoch": 2453, "lr": 9.176609546905334e-06}
{"train_loss": 0.10062707960605621, "global_step": 218363, "epoch": 2453, "lr": 9.176274814977464e-06}
{"train_loss": 0.044623978435993195, "global_step": 218364, "epoch": 2453, "lr": 9.175940088537832e-06}
{"train_loss": 0.04387316852807999, "global_step": 218365, "epoch": 2453, "lr": 9.17560536758651e-06}
{"train_loss": 0.048848580569028854, "global_step": 218366, "epoch": 2453, "lr": 9.175270652123518e-06}
{"train_loss": 0.058405451476573944, "global_step": 218367, "epoch": 2453, "lr": 9.174935942148926e-06}
{"train_loss": 0.04936196655035019, "global_step": 218368, "epoch": 2453, "lr": 9.17460123766275e-06}
{"train_loss": 0.05328736826777458, "global_step": 218369, "epoch": 2453, "lr": 9.174266538665066e-06}
{"train_loss": 0.025004610419273376, "global_step": 218370, "epoch": 2453, "lr": 9.173931845155892e-06}
{"train_loss": 0.021822845563292503, "global_step": 218371, "epoch": 2453, "lr": 9.1735971571353e-06}
{"train_loss": 0.048701539635658264, "global_step": 218372, "epoch": 2453, "lr": 9.173262474603306e-06}
{"train_loss": 0.05475868657231331, "global_step": 218373, "epoch": 2453, "lr": 9.172927797559983e-06}
{"train_loss": 0.06280559301376343, "global_step": 218374, "epoch": 2453, "lr": 9.172593126005346e-06}
{"train_loss": 0.06581643968820572, "global_step": 218375, "epoch": 2453, "lr": 9.172258459939475e-06}
{"train_loss": 0.0921555683016777, "global_step": 218376, "epoch": 2453, "lr": 9.171923799362381e-06}
{"train_loss": 0.03651689738035202, "global_step": 218377, "epoch": 2453, "lr": 9.171589144274135e-06}
{"train_loss": 0.07838612794876099, "global_step": 218378, "epoch": 2453, "lr": 9.171254494674758e-06}
{"train_loss": 0.057052288204431534, "global_step": 218379, "epoch": 2453, "lr": 9.170919850564325e-06}
{"train_loss": 0.06747902184724808, "global_step": 218380, "epoch": 2453, "lr": 9.170585211942845e-06}
{"train_loss": 0.06224306300282478, "global_step": 218381, "epoch": 2453, "lr": 9.170250578810385e-06}
{"train_loss": 0.09445887804031372, "global_step": 218382, "epoch": 2453, "lr": 9.169915951167001e-06}
{"train_loss": 0.061812423169612885, "global_step": 218383, "epoch": 2453, "lr": 9.16958132901271e-06}
{"train_loss": 0.05836586281657219, "global_step": 218384, "epoch": 2453, "lr": 9.16924671234759e-06}
{"train_loss": 0.07759039849042892, "global_step": 218385, "epoch": 2453, "lr": 9.16891210117165e-06}
{"train_loss": 0.03283510357141495, "global_step": 218386, "epoch": 2453, "lr": 9.168577495484964e-06}
{"train_loss": 0.05296788364648819, "global_step": 218387, "epoch": 2453, "lr": 9.168242895287548e-06}
{"train_loss": 0.03812701627612114, "global_step": 218388, "epoch": 2453, "lr": 9.167908300579487e-06}
{"train_loss": 0.05004023388028145, "global_step": 218389, "epoch": 2453, "lr": 9.167573711360783e-06}
{"train_loss": 0.02945972979068756, "global_step": 218390, "epoch": 2453, "lr": 9.167239127631517e-06}
{"train_loss": 0.0629463717341423, "global_step": 218391, "epoch": 2453, "lr": 9.166904549391703e-06}
{"train_loss": 0.05746949464082718, "global_step": 218392, "epoch": 2453, "lr": 9.166569976641414e-06}
{"train_loss": 0.028482714667916298, "global_step": 218393, "epoch": 2453, "lr": 9.166235409380669e-06}
{"train_loss": 0.06349418312311172, "global_step": 218394, "epoch": 2453, "lr": 9.165900847609538e-06}
{"train_loss": 0.08770526200532913, "global_step": 218395, "epoch": 2453, "lr": 9.165566291328042e-06}
{"train_loss": 0.06375430524349213, "global_step": 218396, "epoch": 2453, "lr": 9.165231740536257e-06}
{"train_loss": 0.03764987736940384, "global_step": 218397, "epoch": 2453, "lr": 9.16489719523419e-06}
{"train_loss": 0.041198741644620895, "global_step": 218398, "epoch": 2453, "lr": 9.164562655421916e-06}
{"train_loss": 0.04134312644600868, "global_step": 218399, "epoch": 2453, "lr": 9.164228121099461e-06}
{"train_loss": 0.024695612490177155, "global_step": 218400, "epoch": 2453, "lr": 9.163893592266887e-06}
{"train_loss": 0.048169974237680435, "global_step": 218401, "epoch": 2453, "lr": 9.163559068924227e-06}
{"train_loss": 0.05017993226647377, "global_step": 218402, "epoch": 2453, "lr": 9.16322455107152e-06}
{"train_loss": 0.035776279866695404, "global_step": 218403, "epoch": 2453, "lr": 9.162890038708832e-06}
{"train_loss": 0.050201766192913055, "global_step": 218404, "epoch": 2453, "lr": 9.162555531836187e-06}
{"train_loss": 0.052236431466645736, "global_step": 218405, "epoch": 2453, "lr": 9.162221030453644e-06, "val_loss": 8.864937782287598}
{"train_loss": 0.043307431042194366, "global_step": 218406, "epoch": 2454, "lr": 9.161886534561237e-06}
{"train_loss": 0.05673685669898987, "global_step": 218407, "epoch": 2454, "lr": 9.161552044159022e-06}
{"train_loss": 0.03892786055803299, "global_step": 218408, "epoch": 2454, "lr": 9.161217559247032e-06}
{"train_loss": 0.02546841651201248, "global_step": 218409, "epoch": 2454, "lr": 9.160883079825328e-06}
{"train_loss": 0.06644151359796524, "global_step": 218410, "epoch": 2454, "lr": 9.160548605893932e-06}
{"train_loss": 0.071311354637146, "global_step": 218411, "epoch": 2454, "lr": 9.160214137452916e-06}
{"train_loss": 0.12085957080125809, "global_step": 218412, "epoch": 2454, "lr": 9.159879674502303e-06}
{"train_loss": 0.06342664361000061, "global_step": 218413, "epoch": 2454, "lr": 9.159545217042154e-06}
{"train_loss": 0.04805907979607582, "global_step": 218414, "epoch": 2454, "lr": 9.159210765072495e-06}
{"train_loss": 0.03594566509127617, "global_step": 218415, "epoch": 2454, "lr": 9.158876318593395e-06}
{"train_loss": 0.07239937037229538, "global_step": 218416, "epoch": 2454, "lr": 9.158541877604876e-06}
{"train_loss": 0.042447686195373535, "global_step": 218417, "epoch": 2454, "lr": 9.158207442106992e-06}
{"train_loss": 0.021220508962869644, "global_step": 218418, "epoch": 2454, "lr": 9.1578730120998e-06}
{"train_loss": 0.06923258304595947, "global_step": 218419, "epoch": 2454, "lr": 9.15753858758332e-06}
{"train_loss": 0.04951877519488335, "global_step": 218420, "epoch": 2454, "lr": 9.157204168557626e-06}
{"train_loss": 0.03561298921704292, "global_step": 218421, "epoch": 2454, "lr": 9.15686975502274e-06}
{"train_loss": 0.06597186625003815, "global_step": 218422, "epoch": 2454, "lr": 9.156535346978722e-06}
{"train_loss": 0.09685993194580078, "global_step": 218423, "epoch": 2454, "lr": 9.156200944425597e-06}
{"train_loss": 0.0397624745965004, "global_step": 218424, "epoch": 2454, "lr": 9.15586654736344e-06}
{"train_loss": 0.06512022763490677, "global_step": 218425, "epoch": 2454, "lr": 9.155532155792263e-06}
{"train_loss": 0.027384458109736443, "global_step": 218426, "epoch": 2454, "lr": 9.155197769712138e-06}
{"train_loss": 0.0289468914270401, "global_step": 218427, "epoch": 2454, "lr": 9.154863389123092e-06}
{"train_loss": 0.04306250438094139, "global_step": 218428, "epoch": 2454, "lr": 9.154529014025181e-06}
{"train_loss": 0.06613573431968689, "global_step": 218429, "epoch": 2454, "lr": 9.15419464441844e-06}
{"train_loss": 0.04081886634230614, "global_step": 218430, "epoch": 2454, "lr": 9.153860280302934e-06}
{"train_loss": 0.04841587692499161, "global_step": 218431, "epoch": 2454, "lr": 9.153525921678674e-06}
{"train_loss": 0.04672960191965103, "global_step": 218432, "epoch": 2454, "lr": 9.153191568545744e-06}
{"train_loss": 0.019910303875803947, "global_step": 218433, "epoch": 2454, "lr": 9.152857220904149e-06}
{"train_loss": 0.0778948962688446, "global_step": 218434, "epoch": 2454, "lr": 9.152522878753971e-06}
{"train_loss": 0.03426329046487808, "global_step": 218435, "epoch": 2454, "lr": 9.15218854209523e-06}
{"train_loss": 0.10975898802280426, "global_step": 218436, "epoch": 2454, "lr": 9.15185421092799e-06}
{"train_loss": 0.08463040739297867, "global_step": 218437, "epoch": 2454, "lr": 9.151519885252269e-06}
{"train_loss": 0.046101558953523636, "global_step": 218438, "epoch": 2454, "lr": 9.151185565068143e-06}
{"train_loss": 0.05913036689162254, "global_step": 218439, "epoch": 2454, "lr": 9.150851250375641e-06}
{"train_loss": 0.03459089249372482, "global_step": 218440, "epoch": 2454, "lr": 9.150516941174797e-06}
{"train_loss": 0.03891564905643463, "global_step": 218441, "epoch": 2454, "lr": 9.150182637465681e-06}
{"train_loss": 0.07397940754890442, "global_step": 218442, "epoch": 2454, "lr": 9.149848339248318e-06}
{"train_loss": 0.09508183598518372, "global_step": 218443, "epoch": 2454, "lr": 9.149514046522767e-06}
{"train_loss": 0.05333636701107025, "global_step": 218444, "epoch": 2454, "lr": 9.149179759289051e-06}
{"train_loss": 0.042035844177007675, "global_step": 218445, "epoch": 2454, "lr": 9.148845477547246e-06}
{"train_loss": 0.048007264733314514, "global_step": 218446, "epoch": 2454, "lr": 9.148511201297365e-06}
{"train_loss": 0.05584467202425003, "global_step": 218447, "epoch": 2454, "lr": 9.148176930539487e-06}
{"train_loss": 0.026452580466866493, "global_step": 218448, "epoch": 2454, "lr": 9.147842665273626e-06}
{"train_loss": 0.046755362302064896, "global_step": 218449, "epoch": 2454, "lr": 9.147508405499844e-06}
{"train_loss": 0.030306344851851463, "global_step": 218450, "epoch": 2454, "lr": 9.147174151218175e-06}
{"train_loss": 0.08108137547969818, "global_step": 218451, "epoch": 2454, "lr": 9.146839902428684e-06}
{"train_loss": 0.09483242779970169, "global_step": 218452, "epoch": 2454, "lr": 9.146505659131388e-06}
{"train_loss": 0.03202616050839424, "global_step": 218453, "epoch": 2454, "lr": 9.14617142132635e-06}
{"train_loss": 0.08043599128723145, "global_step": 218454, "epoch": 2454, "lr": 9.145837189013618e-06}
{"train_loss": 0.041999321430921555, "global_step": 218455, "epoch": 2454, "lr": 9.14550296219322e-06}
{"train_loss": 0.018182171508669853, "global_step": 218456, "epoch": 2454, "lr": 9.145168740865224e-06}
{"train_loss": 0.031151482835412025, "global_step": 218457, "epoch": 2454, "lr": 9.14483452502965e-06}
{"train_loss": 0.013388775289058685, "global_step": 218458, "epoch": 2454, "lr": 9.144500314686565e-06}
{"train_loss": 0.06799635291099548, "global_step": 218459, "epoch": 2454, "lr": 9.144166109835994e-06}
{"train_loss": 0.07016397267580032, "global_step": 218460, "epoch": 2454, "lr": 9.143831910478007e-06}
{"train_loss": 0.05502134561538696, "global_step": 218461, "epoch": 2454, "lr": 9.143497716612614e-06}
{"train_loss": 0.03432415425777435, "global_step": 218462, "epoch": 2454, "lr": 9.143163528239902e-06}
{"train_loss": 0.05140732228755951, "global_step": 218463, "epoch": 2454, "lr": 9.14282934535987e-06}
{"train_loss": 0.029026949778199196, "global_step": 218464, "epoch": 2454, "lr": 9.142495167972609e-06}
{"train_loss": 0.06001613661646843, "global_step": 218465, "epoch": 2454, "lr": 9.142160996078126e-06}
{"train_loss": 0.06134551391005516, "global_step": 218466, "epoch": 2454, "lr": 9.141826829676492e-06}
{"train_loss": 0.049822427332401276, "global_step": 218467, "epoch": 2454, "lr": 9.141492668767727e-06}
{"train_loss": 0.059687741100788116, "global_step": 218468, "epoch": 2454, "lr": 9.141158513351905e-06}
{"train_loss": 0.060794949531555176, "global_step": 218469, "epoch": 2454, "lr": 9.140824363429045e-06}
{"train_loss": 0.04506639763712883, "global_step": 218470, "epoch": 2454, "lr": 9.140490218999214e-06}
{"train_loss": 0.05965622141957283, "global_step": 218471, "epoch": 2454, "lr": 9.140156080062434e-06}
{"train_loss": 0.031232355162501335, "global_step": 218472, "epoch": 2454, "lr": 9.139821946618766e-06}
{"train_loss": 0.06140449643135071, "global_step": 218473, "epoch": 2454, "lr": 9.139487818668257e-06}
{"train_loss": 0.07355374842882156, "global_step": 218474, "epoch": 2454, "lr": 9.13915369621095e-06}
{"train_loss": 0.058684684336185455, "global_step": 218475, "epoch": 2454, "lr": 9.13881957924687e-06}
{"train_loss": 0.0775127187371254, "global_step": 218476, "epoch": 2454, "lr": 9.138485467776098e-06}
{"train_loss": 0.040858909487724304, "global_step": 218477, "epoch": 2454, "lr": 9.138151361798635e-06}
{"train_loss": 0.12321698665618896, "global_step": 218478, "epoch": 2454, "lr": 9.13781726131457e-06}
{"train_loss": 0.06400728225708008, "global_step": 218479, "epoch": 2454, "lr": 9.137483166323924e-06}
{"train_loss": 0.08731896430253983, "global_step": 218480, "epoch": 2454, "lr": 9.137149076826734e-06}
{"train_loss": 0.05469666048884392, "global_step": 218481, "epoch": 2454, "lr": 9.136814992823068e-06}
{"train_loss": 0.03128194436430931, "global_step": 218482, "epoch": 2454, "lr": 9.13648091431295e-06}
{"train_loss": 0.0679226815700531, "global_step": 218483, "epoch": 2454, "lr": 9.136146841296438e-06}
{"train_loss": 0.08125784248113632, "global_step": 218484, "epoch": 2454, "lr": 9.135812773773572e-06}
{"train_loss": 0.037387486547231674, "global_step": 218485, "epoch": 2454, "lr": 9.1354787117444e-06}
{"train_loss": 0.039945583790540695, "global_step": 218486, "epoch": 2454, "lr": 9.13514465520896e-06}
{"train_loss": 0.08916806429624557, "global_step": 218487, "epoch": 2454, "lr": 9.13481060416731e-06}
{"train_loss": 0.045940715819597244, "global_step": 218488, "epoch": 2454, "lr": 9.134476558619475e-06}
{"train_loss": 0.08475259691476822, "global_step": 218489, "epoch": 2454, "lr": 9.134142518565525e-06}
{"train_loss": 0.06608670949935913, "global_step": 218490, "epoch": 2454, "lr": 9.133808484005479e-06}
{"train_loss": 0.044208038598299026, "global_step": 218491, "epoch": 2454, "lr": 9.133474454939394e-06}
{"train_loss": 0.06929050385951996, "global_step": 218492, "epoch": 2454, "lr": 9.133140431367326e-06}
{"train_loss": 0.06450030952692032, "global_step": 218493, "epoch": 2454, "lr": 9.132806413289296e-06}
{"train_loss": 0.05520192792199636, "global_step": 218494, "epoch": 2454, "lr": 9.132472400705378e-06, "val_loss": 8.75658130645752}
{"train_loss": 0.03284299001097679, "global_step": 218495, "epoch": 2455, "lr": 9.132138393615586e-06}
{"train_loss": 0.07789834588766098, "global_step": 218496, "epoch": 2455, "lr": 9.131804392019994e-06}
{"train_loss": 0.03719059005379677, "global_step": 218497, "epoch": 2455, "lr": 9.131470395918622e-06}
{"train_loss": 0.08749780058860779, "global_step": 218498, "epoch": 2455, "lr": 9.131136405311535e-06}
{"train_loss": 0.050429992377758026, "global_step": 218499, "epoch": 2455, "lr": 9.130802420198759e-06}
{"train_loss": 0.05230855569243431, "global_step": 218500, "epoch": 2455, "lr": 9.130468440580358e-06}
{"train_loss": 0.030364342033863068, "global_step": 218501, "epoch": 2455, "lr": 9.130134466456352e-06}
{"train_loss": 0.025604404509067535, "global_step": 218502, "epoch": 2455, "lr": 9.129800497826818e-06}
{"train_loss": 0.05082877725362778, "global_step": 218503, "epoch": 2455, "lr": 9.129466534691772e-06}
{"train_loss": 0.04000547155737877, "global_step": 218504, "epoch": 2455, "lr": 9.129132577051286e-06}
{"train_loss": 0.05980125442147255, "global_step": 218505, "epoch": 2455, "lr": 9.128798624905371e-06}
{"train_loss": 0.031049152836203575, "global_step": 218506, "epoch": 2455, "lr": 9.128464678254111e-06}
{"train_loss": 0.04923844337463379, "global_step": 218507, "epoch": 2455, "lr": 9.128130737097512e-06}
{"train_loss": 0.028873905539512634, "global_step": 218508, "epoch": 2455, "lr": 9.127796801435639e-06}
{"train_loss": 0.029142912477254868, "global_step": 218509, "epoch": 2455, "lr": 9.12746287126855e-06}
{"train_loss": 0.0327795073390007, "global_step": 218510, "epoch": 2455, "lr": 9.127128946596264e-06}
{"train_loss": 0.05031991004943848, "global_step": 218511, "epoch": 2455, "lr": 9.126795027418845e-06}
{"train_loss": 0.046032387763261795, "global_step": 218512, "epoch": 2455, "lr": 9.126461113736335e-06}
{"train_loss": 0.020999576896429062, "global_step": 218513, "epoch": 2455, "lr": 9.126127205548756e-06}
{"train_loss": 0.07349284738302231, "global_step": 218514, "epoch": 2455, "lr": 9.125793302856189e-06}
{"train_loss": 0.04259142279624939, "global_step": 218515, "epoch": 2455, "lr": 9.125459405658648e-06}
{"train_loss": 0.0736229345202446, "global_step": 218516, "epoch": 2455, "lr": 9.125125513956201e-06}
{"train_loss": 0.03968745097517967, "global_step": 218517, "epoch": 2455, "lr": 9.12479162774888e-06}
{"train_loss": 0.05386606603860855, "global_step": 218518, "epoch": 2455, "lr": 9.124457747036725e-06}
{"train_loss": 0.059083353728055954, "global_step": 218519, "epoch": 2455, "lr": 9.1241238718198e-06}
{"train_loss": 0.07263288646936417, "global_step": 218520, "epoch": 2455, "lr": 9.123790002098126e-06}
{"train_loss": 0.10960415005683899, "global_step": 218521, "epoch": 2455, "lr": 9.123456137871773e-06}
{"train_loss": 0.04962317645549774, "global_step": 218522, "epoch": 2455, "lr": 9.123122279140756e-06}
{"train_loss": 0.08721061795949936, "global_step": 218523, "epoch": 2455, "lr": 9.122788425905155e-06}
{"train_loss": 0.014447267167270184, "global_step": 218524, "epoch": 2455, "lr": 9.12245457816498e-06}
{"train_loss": 0.049914080649614334, "global_step": 218525, "epoch": 2455, "lr": 9.122120735920309e-06}
{"train_loss": 0.025161312893033028, "global_step": 218526, "epoch": 2455, "lr": 9.121786899171159e-06}
{"train_loss": 0.028723765164613724, "global_step": 218527, "epoch": 2455, "lr": 9.121453067917584e-06}
{"train_loss": 0.08100444078445435, "global_step": 218528, "epoch": 2455, "lr": 9.121119242159642e-06}
{"train_loss": 0.031248511746525764, "global_step": 218529, "epoch": 2455, "lr": 9.12078542189736e-06}
{"train_loss": 0.053298696875572205, "global_step": 218530, "epoch": 2455, "lr": 9.120451607130798e-06}
{"train_loss": 0.06491611897945404, "global_step": 218531, "epoch": 2455, "lr": 9.120117797859983e-06}
{"train_loss": 0.13824617862701416, "global_step": 218532, "epoch": 2455, "lr": 9.119783994084979e-06}
{"train_loss": 0.11259583383798599, "global_step": 218533, "epoch": 2455, "lr": 9.119450195805812e-06}
{"train_loss": 0.030523631721735, "global_step": 218534, "epoch": 2455, "lr": 9.119116403022548e-06}
{"train_loss": 0.06417956948280334, "global_step": 218535, "epoch": 2455, "lr": 9.118782615735211e-06}
{"train_loss": 0.04588548466563225, "global_step": 218536, "epoch": 2455, "lr": 9.118448833943866e-06}
{"train_loss": 0.03245322033762932, "global_step": 218537, "epoch": 2455, "lr": 9.118115057648535e-06}
{"train_loss": 0.036471132189035416, "global_step": 218538, "epoch": 2455, "lr": 9.117781286849285e-06}
{"train_loss": 0.027493257075548172, "global_step": 218539, "epoch": 2455, "lr": 9.117447521546141e-06}
{"train_loss": 0.04672408103942871, "global_step": 218540, "epoch": 2455, "lr": 9.11711376173917e-06}
{"train_loss": 0.05393263325095177, "global_step": 218541, "epoch": 2455, "lr": 9.116780007428393e-06}
{"train_loss": 0.07665553689002991, "global_step": 218542, "epoch": 2455, "lr": 9.11644625861388e-06}
{"train_loss": 0.07775600254535675, "global_step": 218543, "epoch": 2455, "lr": 9.11611251529565e-06}
{"train_loss": 0.0653378814458847, "global_step": 218544, "epoch": 2455, "lr": 9.11577877747376e-06}
{"train_loss": 0.09917140752077103, "global_step": 218545, "epoch": 2455, "lr": 9.115445045148263e-06}
{"train_loss": 0.015758832916617393, "global_step": 218546, "epoch": 2455, "lr": 9.115111318319192e-06}
{"train_loss": 0.10576090961694717, "global_step": 218547, "epoch": 2455, "lr": 9.114777596986602e-06}
{"train_loss": 0.01700749062001705, "global_step": 218548, "epoch": 2455, "lr": 9.11444388115052e-06}
{"train_loss": 0.09025873988866806, "global_step": 218549, "epoch": 2455, "lr": 9.11411017081102e-06}
{"train_loss": 0.05363737419247627, "global_step": 218550, "epoch": 2455, "lr": 9.113776465968122e-06}
{"train_loss": 0.06443259865045547, "global_step": 218551, "epoch": 2455, "lr": 9.113442766621871e-06}
{"train_loss": 0.12267586588859558, "global_step": 218552, "epoch": 2455, "lr": 9.11310907277233e-06}
{"train_loss": 0.03447923809289932, "global_step": 218553, "epoch": 2455, "lr": 9.11277538441952e-06}
{"train_loss": 0.02553695999085903, "global_step": 218554, "epoch": 2455, "lr": 9.112441701563512e-06}
{"train_loss": 0.0241080392152071, "global_step": 218555, "epoch": 2455, "lr": 9.112108024204341e-06}
{"train_loss": 0.034632835537195206, "global_step": 218556, "epoch": 2455, "lr": 9.111774352342033e-06}
{"train_loss": 0.02497696317732334, "global_step": 218557, "epoch": 2455, "lr": 9.111440685976663e-06}
{"train_loss": 0.05243802070617676, "global_step": 218558, "epoch": 2455, "lr": 9.111107025108251e-06}
{"train_loss": 0.06535819917917252, "global_step": 218559, "epoch": 2455, "lr": 9.110773369736858e-06}
{"train_loss": 0.050696298480033875, "global_step": 218560, "epoch": 2455, "lr": 9.110439719862517e-06}
{"train_loss": 0.024635417386889458, "global_step": 218561, "epoch": 2455, "lr": 9.110106075485287e-06}
{"train_loss": 0.022832175716757774, "global_step": 218562, "epoch": 2455, "lr": 9.109772436605197e-06}
{"train_loss": 0.058723293244838715, "global_step": 218563, "epoch": 2455, "lr": 9.109438803222293e-06}
{"train_loss": 0.06904718279838562, "global_step": 218564, "epoch": 2455, "lr": 9.109105175336646e-06}
{"train_loss": 0.03442530333995819, "global_step": 218565, "epoch": 2455, "lr": 9.108771552948264e-06}
{"train_loss": 0.045891862362623215, "global_step": 218566, "epoch": 2455, "lr": 9.108437936057223e-06}
{"train_loss": 0.0483534149825573, "global_step": 218567, "epoch": 2455, "lr": 9.10810432466354e-06}
{"train_loss": 0.04189728945493698, "global_step": 218568, "epoch": 2455, "lr": 9.107770718767294e-06}
{"train_loss": 0.04667428508400917, "global_step": 218569, "epoch": 2455, "lr": 9.107437118368489e-06}
{"train_loss": 0.06911761313676834, "global_step": 218570, "epoch": 2455, "lr": 9.107103523467208e-06}
{"train_loss": 0.06811623275279999, "global_step": 218571, "epoch": 2455, "lr": 9.106769934063464e-06}
{"train_loss": 0.08111932873725891, "global_step": 218572, "epoch": 2455, "lr": 9.106436350157327e-06}
{"train_loss": 0.076134592294693, "global_step": 218573, "epoch": 2455, "lr": 9.10610277174882e-06}
{"train_loss": 0.042958635836839676, "global_step": 218574, "epoch": 2455, "lr": 9.105769198838015e-06}
{"train_loss": 0.11640606820583344, "global_step": 218575, "epoch": 2455, "lr": 9.105435631424924e-06}
{"train_loss": 0.10114119946956635, "global_step": 218576, "epoch": 2455, "lr": 9.105102069509625e-06}
{"train_loss": 0.03357456624507904, "global_step": 218577, "epoch": 2455, "lr": 9.104768513092133e-06}
{"train_loss": 0.11063651740550995, "global_step": 218578, "epoch": 2455, "lr": 9.104434962172514e-06}
{"train_loss": 0.06448054313659668, "global_step": 218579, "epoch": 2455, "lr": 9.104101416750799e-06}
{"train_loss": 0.04966137558221817, "global_step": 218580, "epoch": 2455, "lr": 9.103767876827035e-06}
{"train_loss": 0.096488818526268, "global_step": 218581, "epoch": 2455, "lr": 9.10343434240129e-06}
{"train_loss": 0.05115141347050667, "global_step": 218582, "epoch": 2455, "lr": 9.103100813473569e-06}
{"train_loss": 0.0557680811753936, "global_step": 218583, "epoch": 2455, "lr": 9.102767290043956e-06, "val_loss": 8.887689590454102, "train_action_mse_error": 9.75082015991211}
{"train_loss": 0.04349163547158241, "global_step": 218584, "epoch": 2456, "lr": 9.102433772112463e-06}
{"train_loss": 0.05989599600434303, "global_step": 218585, "epoch": 2456, "lr": 9.102100259679164e-06}
{"train_loss": 0.03953426331281662, "global_step": 218586, "epoch": 2456, "lr": 9.101766752744074e-06}
{"train_loss": 0.03471734747290611, "global_step": 218587, "epoch": 2456, "lr": 9.101433251307261e-06}
{"train_loss": 0.09116369485855103, "global_step": 218588, "epoch": 2456, "lr": 9.101099755368758e-06}
{"train_loss": 0.03656582161784172, "global_step": 218589, "epoch": 2456, "lr": 9.100766264928622e-06}
{"train_loss": 0.031725168228149414, "global_step": 218590, "epoch": 2456, "lr": 9.100432779986889e-06}
{"train_loss": 0.027598127722740173, "global_step": 218591, "epoch": 2456, "lr": 9.10009930054359e-06}
{"train_loss": 0.023792196065187454, "global_step": 218592, "epoch": 2456, "lr": 9.099765826598799e-06}
{"train_loss": 0.042593538761138916, "global_step": 218593, "epoch": 2456, "lr": 9.099432358152537e-06}
{"train_loss": 0.06454121321439743, "global_step": 218594, "epoch": 2456, "lr": 9.099098895204867e-06}
{"train_loss": 0.03966335579752922, "global_step": 218595, "epoch": 2456, "lr": 9.098765437755823e-06}
{"train_loss": 0.044118769466876984, "global_step": 218596, "epoch": 2456, "lr": 9.098431985805439e-06}
{"train_loss": 0.057792503386735916, "global_step": 218597, "epoch": 2456, "lr": 9.098098539353789e-06}
{"train_loss": 0.06376069784164429, "global_step": 218598, "epoch": 2456, "lr": 9.097765098400885e-06}
{"train_loss": 0.05357516556978226, "global_step": 218599, "epoch": 2456, "lr": 9.097431662946787e-06}
{"train_loss": 0.04408211261034012, "global_step": 218600, "epoch": 2456, "lr": 9.097098232991558e-06}
{"train_loss": 0.0717218667268753, "global_step": 218601, "epoch": 2456, "lr": 9.096764808535207e-06}
{"train_loss": 0.07397910207509995, "global_step": 218602, "epoch": 2456, "lr": 9.096431389577815e-06}
{"train_loss": 0.029397308826446533, "global_step": 218603, "epoch": 2456, "lr": 9.096097976119394e-06}
{"train_loss": 0.013680771924555302, "global_step": 218604, "epoch": 2456, "lr": 9.09576456816002e-06}
{"train_loss": 0.033629994839429855, "global_step": 218605, "epoch": 2456, "lr": 9.095431165699708e-06}
{"train_loss": 0.05108502134680748, "global_step": 218606, "epoch": 2456, "lr": 9.095097768738525e-06}
{"train_loss": 0.07823708653450012, "global_step": 218607, "epoch": 2456, "lr": 9.094764377276499e-06}
{"train_loss": 0.026443341746926308, "global_step": 218608, "epoch": 2456, "lr": 9.094430991313696e-06}
{"train_loss": 0.03237081319093704, "global_step": 218609, "epoch": 2456, "lr": 9.094097610850133e-06}
{"train_loss": 0.048478953540325165, "global_step": 218610, "epoch": 2456, "lr": 9.093764235885888e-06}
{"train_loss": 0.10423065721988678, "global_step": 218611, "epoch": 2456, "lr": 9.093430866420971e-06}
{"train_loss": 0.04573407024145126, "global_step": 218612, "epoch": 2456, "lr": 9.093097502455456e-06}
{"train_loss": 0.07511996477842331, "global_step": 218613, "epoch": 2456, "lr": 9.092764143989364e-06}
{"train_loss": 0.018238577991724014, "global_step": 218614, "epoch": 2456, "lr": 9.09243079102276e-06}
{"train_loss": 0.10214685648679733, "global_step": 218615, "epoch": 2456, "lr": 9.092097443555675e-06}
{"train_loss": 0.10186369717121124, "global_step": 218616, "epoch": 2456, "lr": 9.091764101588168e-06}
{"train_loss": 0.057817492634058, "global_step": 218617, "epoch": 2456, "lr": 9.091430765120263e-06}
{"train_loss": 0.034021925181150436, "global_step": 218618, "epoch": 2456, "lr": 9.091097434152013e-06}
{"train_loss": 0.03959633782505989, "global_step": 218619, "epoch": 2456, "lr": 9.090764108683481e-06}
{"train_loss": 0.027900220826268196, "global_step": 218620, "epoch": 2456, "lr": 9.090430788714682e-06}
{"train_loss": 0.06854091584682465, "global_step": 218621, "epoch": 2456, "lr": 9.090097474245695e-06}
{"train_loss": 0.05148965120315552, "global_step": 218622, "epoch": 2456, "lr": 9.089764165276527e-06}
{"train_loss": 0.0335012786090374, "global_step": 218623, "epoch": 2456, "lr": 9.089430861807257e-06}
{"train_loss": 0.11354739218950272, "global_step": 218624, "epoch": 2456, "lr": 9.0890975638379e-06}
{"train_loss": 0.04625789076089859, "global_step": 218625, "epoch": 2456, "lr": 9.088764271368533e-06}
{"train_loss": 0.06952378898859024, "global_step": 218626, "epoch": 2456, "lr": 9.088430984399165e-06}
{"train_loss": 0.08981306105852127, "global_step": 218627, "epoch": 2456, "lr": 9.08809770292987e-06}
{"train_loss": 0.03138567507266998, "global_step": 218628, "epoch": 2456, "lr": 9.087764426960687e-06}
{"train_loss": 0.04943070188164711, "global_step": 218629, "epoch": 2456, "lr": 9.087431156491638e-06}
{"train_loss": 0.060796841979026794, "global_step": 218630, "epoch": 2456, "lr": 9.0870978915228e-06}
{"train_loss": 0.08280841261148453, "global_step": 218631, "epoch": 2456, "lr": 9.086764632054195e-06}
{"train_loss": 0.026639610528945923, "global_step": 218632, "epoch": 2456, "lr": 9.08643137808588e-06}
{"train_loss": 0.03848971053957939, "global_step": 218633, "epoch": 2456, "lr": 9.086098129617903e-06}
{"train_loss": 0.05331459641456604, "global_step": 218634, "epoch": 2456, "lr": 9.085764886650283e-06}
{"train_loss": 0.04372333735227585, "global_step": 218635, "epoch": 2456, "lr": 9.085431649183086e-06}
{"train_loss": 0.10225758701562881, "global_step": 218636, "epoch": 2456, "lr": 9.08509841721637e-06}
{"train_loss": 0.10672815144062042, "global_step": 218637, "epoch": 2456, "lr": 9.084765190750143e-06}
{"train_loss": 0.09858245402574539, "global_step": 218638, "epoch": 2456, "lr": 9.08443196978449e-06}
{"train_loss": 0.1112242341041565, "global_step": 218639, "epoch": 2456, "lr": 9.08409875431942e-06}
{"train_loss": 0.058703262358903885, "global_step": 218640, "epoch": 2456, "lr": 9.08376554435501e-06}
{"train_loss": 0.0587662011384964, "global_step": 218641, "epoch": 2456, "lr": 9.083432339891273e-06}
{"train_loss": 0.06342439353466034, "global_step": 218642, "epoch": 2456, "lr": 9.083099140928286e-06}
{"train_loss": 0.07010313123464584, "global_step": 218643, "epoch": 2456, "lr": 9.082765947466065e-06}
{"train_loss": 0.06803690642118454, "global_step": 218644, "epoch": 2456, "lr": 9.082432759504678e-06}
{"train_loss": 0.09381690621376038, "global_step": 218645, "epoch": 2456, "lr": 9.082099577044146e-06}
{"train_loss": 0.07218010723590851, "global_step": 218646, "epoch": 2456, "lr": 9.08176640008454e-06}
{"train_loss": 0.04697481915354729, "global_step": 218647, "epoch": 2456, "lr": 9.08143322862588e-06}
{"train_loss": 0.03525897115468979, "global_step": 218648, "epoch": 2456, "lr": 9.081100062668236e-06}
{"train_loss": 0.06421112269163132, "global_step": 218649, "epoch": 2456, "lr": 9.08076690221163e-06}
{"train_loss": 0.06399008631706238, "global_step": 218650, "epoch": 2456, "lr": 9.080433747256122e-06}
{"train_loss": 0.05510687083005905, "global_step": 218651, "epoch": 2456, "lr": 9.080100597801739e-06}
{"train_loss": 0.06585821509361267, "global_step": 218652, "epoch": 2456, "lr": 9.079767453848553e-06}
{"train_loss": 0.05221666023135185, "global_step": 218653, "epoch": 2456, "lr": 9.079434315396578e-06}
{"train_loss": 0.05340403690934181, "global_step": 218654, "epoch": 2456, "lr": 9.079101182445876e-06}
{"train_loss": 0.020069371908903122, "global_step": 218655, "epoch": 2456, "lr": 9.078768054996505e-06}
{"train_loss": 0.09051328897476196, "global_step": 218656, "epoch": 2456, "lr": 9.07843493304848e-06}
{"train_loss": 0.0401461124420166, "global_step": 218657, "epoch": 2456, "lr": 9.078101816601876e-06}
{"train_loss": 0.07170307636260986, "global_step": 218658, "epoch": 2456, "lr": 9.077768705656708e-06}
{"train_loss": 0.03840821981430054, "global_step": 218659, "epoch": 2456, "lr": 9.077435600213047e-06}
{"train_loss": 0.09145691990852356, "global_step": 218660, "epoch": 2456, "lr": 9.077102500270912e-06}
{"train_loss": 0.05339474603533745, "global_step": 218661, "epoch": 2456, "lr": 9.076769405830376e-06}
{"train_loss": 0.05234217271208763, "global_step": 218662, "epoch": 2456, "lr": 9.076436316891456e-06}
{"train_loss": 0.03620095178484917, "global_step": 218663, "epoch": 2456, "lr": 9.076103233454224e-06}
{"train_loss": 0.02794189378619194, "global_step": 218664, "epoch": 2456, "lr": 9.075770155518698e-06}
{"train_loss": 0.04889129474759102, "global_step": 218665, "epoch": 2456, "lr": 9.075437083084953e-06}
{"train_loss": 0.04107348248362541, "global_step": 218666, "epoch": 2456, "lr": 9.075104016153013e-06}
{"train_loss": 0.02714717946946621, "global_step": 218667, "epoch": 2456, "lr": 9.074770954722912e-06}
{"train_loss": 0.05738436058163643, "global_step": 218668, "epoch": 2456, "lr": 9.074437898794724e-06}
{"train_loss": 0.06096900254487991, "global_step": 218669, "epoch": 2456, "lr": 9.074104848368465e-06}
{"train_loss": 0.07523791491985321, "global_step": 218670, "epoch": 2456, "lr": 9.073771803444208e-06}
{"train_loss": 0.033855561167001724, "global_step": 218671, "epoch": 2456, "lr": 9.07343876402197e-06}
{"train_loss": 0.056049021808535195, "global_step": 218672, "epoch": 2456, "lr": 9.073105730101823e-06, "val_loss": 8.851486206054688}
{"train_loss": 0.02587510645389557, "global_step": 218673, "epoch": 2457, "lr": 9.072772701683785e-06}
{"train_loss": 0.04609169811010361, "global_step": 218674, "epoch": 2457, "lr": 9.072439678767924e-06}
{"train_loss": 0.03240339830517769, "global_step": 218675, "epoch": 2457, "lr": 9.072106661354262e-06}
{"train_loss": 0.04922085255384445, "global_step": 218676, "epoch": 2457, "lr": 9.071773649442873e-06}
{"train_loss": 0.04088447615504265, "global_step": 218677, "epoch": 2457, "lr": 9.071440643033773e-06}
{"train_loss": 0.08936262875795364, "global_step": 218678, "epoch": 2457, "lr": 9.07110764212703e-06}
{"train_loss": 0.06513228267431259, "global_step": 218679, "epoch": 2457, "lr": 9.070774646722662e-06}
{"train_loss": 0.03145977109670639, "global_step": 218680, "epoch": 2457, "lr": 9.070441656820744e-06}
{"train_loss": 0.07645756751298904, "global_step": 218681, "epoch": 2457, "lr": 9.070108672421296e-06}
{"train_loss": 0.07564841210842133, "global_step": 218682, "epoch": 2457, "lr": 9.069775693524384e-06}
{"train_loss": 0.05664621293544769, "global_step": 218683, "epoch": 2457, "lr": 9.069442720130029e-06}
{"train_loss": 0.07338766008615494, "global_step": 218684, "epoch": 2457, "lr": 9.069109752238302e-06}
{"train_loss": 0.0699930191040039, "global_step": 218685, "epoch": 2457, "lr": 9.06877678984922e-06}
{"train_loss": 0.0432438962161541, "global_step": 218686, "epoch": 2457, "lr": 9.068443832962853e-06}
{"train_loss": 0.03049168735742569, "global_step": 218687, "epoch": 2457, "lr": 9.068110881579228e-06}
{"train_loss": 0.13773755729198456, "global_step": 218688, "epoch": 2457, "lr": 9.067777935698401e-06}
{"train_loss": 0.06525728106498718, "global_step": 218689, "epoch": 2457, "lr": 9.067444995320406e-06}
{"train_loss": 0.04391370713710785, "global_step": 218690, "epoch": 2457, "lr": 9.067112060445294e-06}
{"train_loss": 0.06893733888864517, "global_step": 218691, "epoch": 2457, "lr": 9.066779131073123e-06}
{"train_loss": 0.06983484327793121, "global_step": 218692, "epoch": 2457, "lr": 9.066446207203911e-06}
{"train_loss": 0.015123679302632809, "global_step": 218693, "epoch": 2457, "lr": 9.06611328883773e-06}
{"train_loss": 0.0920916274189949, "global_step": 218694, "epoch": 2457, "lr": 9.065780375974592e-06}
{"train_loss": 0.019670818001031876, "global_step": 218695, "epoch": 2457, "lr": 9.06544746861458e-06}
{"train_loss": 0.0665043517947197, "global_step": 218696, "epoch": 2457, "lr": 9.065114566757704e-06}
{"train_loss": 0.057673849165439606, "global_step": 218697, "epoch": 2457, "lr": 9.064781670404043e-06}
{"train_loss": 0.0586303249001503, "global_step": 218698, "epoch": 2457, "lr": 9.064448779553602e-06}
{"train_loss": 0.06108780950307846, "global_step": 218699, "epoch": 2457, "lr": 9.064115894206465e-06}
{"train_loss": 0.05165667459368706, "global_step": 218700, "epoch": 2457, "lr": 9.063783014362648e-06}
{"train_loss": 0.019313087686896324, "global_step": 218701, "epoch": 2457, "lr": 9.063450140022212e-06}
{"train_loss": 0.07066983729600906, "global_step": 218702, "epoch": 2457, "lr": 9.06311727118519e-06}
{"train_loss": 0.06213865429162979, "global_step": 218703, "epoch": 2457, "lr": 9.062784407851643e-06}
{"train_loss": 0.11131177097558975, "global_step": 218704, "epoch": 2457, "lr": 9.062451550021589e-06}
{"train_loss": 0.03406013920903206, "global_step": 218705, "epoch": 2457, "lr": 9.06211869769511e-06}
{"train_loss": 0.055558450520038605, "global_step": 218706, "epoch": 2457, "lr": 9.061785850872223e-06}
{"train_loss": 0.0790194571018219, "global_step": 218707, "epoch": 2457, "lr": 9.061453009552972e-06}
{"train_loss": 0.026670964434742928, "global_step": 218708, "epoch": 2457, "lr": 9.06112017373742e-06}
{"train_loss": 0.04784837365150452, "global_step": 218709, "epoch": 2457, "lr": 9.060787343425593e-06}
{"train_loss": 0.03133363649249077, "global_step": 218710, "epoch": 2457, "lr": 9.06045451861755e-06}
{"train_loss": 0.060164690017700195, "global_step": 218711, "epoch": 2457, "lr": 9.060121699313322e-06}
{"train_loss": 0.07243230193853378, "global_step": 218712, "epoch": 2457, "lr": 9.059788885512971e-06}
{"train_loss": 0.07646279782056808, "global_step": 218713, "epoch": 2457, "lr": 9.059456077216521e-06}
{"train_loss": 0.04265321046113968, "global_step": 218714, "epoch": 2457, "lr": 9.059123274424042e-06}
{"train_loss": 0.03486209735274315, "global_step": 218715, "epoch": 2457, "lr": 9.058790477135553e-06}
{"train_loss": 0.039262596517801285, "global_step": 218716, "epoch": 2457, "lr": 9.058457685351124e-06}
{"train_loss": 0.07314524799585342, "global_step": 218717, "epoch": 2457, "lr": 9.058124899070768e-06}
{"train_loss": 0.08735723793506622, "global_step": 218718, "epoch": 2457, "lr": 9.057792118294566e-06}
{"train_loss": 0.10025504976511002, "global_step": 218719, "epoch": 2457, "lr": 9.057459343022528e-06}
{"train_loss": 0.04432518407702446, "global_step": 218720, "epoch": 2457, "lr": 9.057126573254732e-06}
{"train_loss": 0.03296967223286629, "global_step": 218721, "epoch": 2457, "lr": 9.056793808991194e-06}
{"train_loss": 0.04225398227572441, "global_step": 218722, "epoch": 2457, "lr": 9.056461050231985e-06}
{"train_loss": 0.084757499396801, "global_step": 218723, "epoch": 2457, "lr": 9.056128296977118e-06}
{"train_loss": 0.056306514889001846, "global_step": 218724, "epoch": 2457, "lr": 9.05579554922667e-06}
{"train_loss": 0.06723298132419586, "global_step": 218725, "epoch": 2457, "lr": 9.055462806980658e-06}
{"train_loss": 0.014780621975660324, "global_step": 218726, "epoch": 2457, "lr": 9.055130070239143e-06}
{"train_loss": 0.024617372080683708, "global_step": 218727, "epoch": 2457, "lr": 9.054797339002174e-06}
{"train_loss": 0.027650944888591766, "global_step": 218728, "epoch": 2457, "lr": 9.054464613269781e-06}
{"train_loss": 0.05292242020368576, "global_step": 218729, "epoch": 2457, "lr": 9.054131893042029e-06}
{"train_loss": 0.026569455862045288, "global_step": 218730, "epoch": 2457, "lr": 9.053799178318933e-06}
{"train_loss": 0.07045061886310577, "global_step": 218731, "epoch": 2457, "lr": 9.053466469100569e-06}
{"train_loss": 0.04352898523211479, "global_step": 218732, "epoch": 2457, "lr": 9.053133765386957e-06}
{"train_loss": 0.031658533960580826, "global_step": 218733, "epoch": 2457, "lr": 9.052801067178163e-06}
{"train_loss": 0.059640005230903625, "global_step": 218734, "epoch": 2457, "lr": 9.05246837447421e-06}
{"train_loss": 0.113246388733387, "global_step": 218735, "epoch": 2457, "lr": 9.052135687275164e-06}
{"train_loss": 0.08073033392429352, "global_step": 218736, "epoch": 2457, "lr": 9.051803005581044e-06}
{"train_loss": 0.04409148916602135, "global_step": 218737, "epoch": 2457, "lr": 9.051470329391925e-06}
{"train_loss": 0.03535112738609314, "global_step": 218738, "epoch": 2457, "lr": 9.051137658707826e-06}
{"train_loss": 0.05523759499192238, "global_step": 218739, "epoch": 2457, "lr": 9.050804993528816e-06}
{"train_loss": 0.03917372599244118, "global_step": 218740, "epoch": 2457, "lr": 9.05047233385491e-06}
{"train_loss": 0.10475079715251923, "global_step": 218741, "epoch": 2457, "lr": 9.050139679686181e-06}
{"train_loss": 0.05867401137948036, "global_step": 218742, "epoch": 2457, "lr": 9.04980703102265e-06}
{"train_loss": 0.060354121029376984, "global_step": 218743, "epoch": 2457, "lr": 9.049474387864387e-06}
{"train_loss": 0.04789944738149643, "global_step": 218744, "epoch": 2457, "lr": 9.049141750211427e-06}
{"train_loss": 0.03665357455611229, "global_step": 218745, "epoch": 2457, "lr": 9.048809118063794e-06}
{"train_loss": 0.035777900367975235, "global_step": 218746, "epoch": 2457, "lr": 9.04847649142156e-06}
{"train_loss": 0.04348078370094299, "global_step": 218747, "epoch": 2457, "lr": 9.048143870284747e-06}
{"train_loss": 0.03962056711316109, "global_step": 218748, "epoch": 2457, "lr": 9.047811254653427e-06}
{"train_loss": 0.04978007450699806, "global_step": 218749, "epoch": 2457, "lr": 9.047478644527618e-06}
{"train_loss": 0.06807258725166321, "global_step": 218750, "epoch": 2457, "lr": 9.047146039907389e-06}
{"train_loss": 0.06399678438901901, "global_step": 218751, "epoch": 2457, "lr": 9.046813440792756e-06}
{"train_loss": 0.06340552866458893, "global_step": 218752, "epoch": 2457, "lr": 9.046480847183791e-06}
{"train_loss": 0.04768790677189827, "global_step": 218753, "epoch": 2457, "lr": 9.046148259080522e-06}
{"train_loss": 0.04774995893239975, "global_step": 218754, "epoch": 2457, "lr": 9.045815676483006e-06}
{"train_loss": 0.09224911779165268, "global_step": 218755, "epoch": 2457, "lr": 9.045483099391271e-06}
{"train_loss": 0.04865477234125137, "global_step": 218756, "epoch": 2457, "lr": 9.045150527805385e-06}
{"train_loss": 0.06496957689523697, "global_step": 218757, "epoch": 2457, "lr": 9.044817961725366e-06}
{"train_loss": 0.0454082190990448, "global_step": 218758, "epoch": 2457, "lr": 9.044485401151282e-06}
{"train_loss": 0.06312655657529831, "global_step": 218759, "epoch": 2457, "lr": 9.044152846083153e-06}
{"train_loss": 0.048537954688072205, "global_step": 218760, "epoch": 2457, "lr": 9.04382029652106e-06}
{"train_loss": 0.056226600599841456, "global_step": 218761, "epoch": 2457, "lr": 9.043487752465007e-06, "val_loss": 8.70278549194336}
{"train_loss": 0.07059624791145325, "global_step": 218762, "epoch": 2458, "lr": 9.043155213915061e-06}
{"train_loss": 0.0749385878443718, "global_step": 218763, "epoch": 2458, "lr": 9.042822680871271e-06}
{"train_loss": 0.1084914356470108, "global_step": 218764, "epoch": 2458, "lr": 9.042490153333666e-06}
{"train_loss": 0.06165396049618721, "global_step": 218765, "epoch": 2458, "lr": 9.042157631302311e-06}
{"train_loss": 0.07570254057645798, "global_step": 218766, "epoch": 2458, "lr": 9.041825114777225e-06}
{"train_loss": 0.08096218854188919, "global_step": 218767, "epoch": 2458, "lr": 9.041492603758484e-06}
{"train_loss": 0.1383720338344574, "global_step": 218768, "epoch": 2458, "lr": 9.0411600982461e-06}
{"train_loss": 0.123695507645607, "global_step": 218769, "epoch": 2458, "lr": 9.040827598240142e-06}
{"train_loss": 0.05682070925831795, "global_step": 218770, "epoch": 2458, "lr": 9.040495103740632e-06}
{"train_loss": 0.05718530714511871, "global_step": 218771, "epoch": 2458, "lr": 9.040162614747644e-06}
{"train_loss": 0.06137879937887192, "global_step": 218772, "epoch": 2458, "lr": 9.039830131261196e-06}
{"train_loss": 0.039140161126852036, "global_step": 218773, "epoch": 2458, "lr": 9.039497653281354e-06}
{"train_loss": 0.056722383946180344, "global_step": 218774, "epoch": 2458, "lr": 9.03916518080814e-06}
{"train_loss": 0.06620769947767258, "global_step": 218775, "epoch": 2458, "lr": 9.038832713841622e-06}
{"train_loss": 0.07822037488222122, "global_step": 218776, "epoch": 2458, "lr": 9.03850025238182e-06}
{"train_loss": 0.09117301553487778, "global_step": 218777, "epoch": 2458, "lr": 9.03816779642881e-06}
{"train_loss": 0.05311652645468712, "global_step": 218778, "epoch": 2458, "lr": 9.037835345982604e-06}
{"train_loss": 0.0463734045624733, "global_step": 218779, "epoch": 2458, "lr": 9.037502901043276e-06}
{"train_loss": 0.01876857504248619, "global_step": 218780, "epoch": 2458, "lr": 9.037170461610839e-06}
{"train_loss": 0.025983041152358055, "global_step": 218781, "epoch": 2458, "lr": 9.036838027685374e-06}
{"train_loss": 0.06078340858221054, "global_step": 218782, "epoch": 2458, "lr": 9.036505599266904e-06}
{"train_loss": 0.053845953196287155, "global_step": 218783, "epoch": 2458, "lr": 9.036173176355461e-06}
{"train_loss": 0.037877313792705536, "global_step": 218784, "epoch": 2458, "lr": 9.03584075895112e-06}
{"train_loss": 0.044639743864536285, "global_step": 218785, "epoch": 2458, "lr": 9.0355083470539e-06}
{"train_loss": 0.0751926600933075, "global_step": 218786, "epoch": 2458, "lr": 9.03517594066387e-06}
{"train_loss": 0.09620904177427292, "global_step": 218787, "epoch": 2458, "lr": 9.034843539781046e-06}
{"train_loss": 0.09930500388145447, "global_step": 218788, "epoch": 2458, "lr": 9.0345111444055e-06}
{"train_loss": 0.034576673060655594, "global_step": 218789, "epoch": 2458, "lr": 9.034178754537253e-06}
{"train_loss": 0.039536215364933014, "global_step": 218790, "epoch": 2458, "lr": 9.033846370176374e-06}
{"train_loss": 0.09866750240325928, "global_step": 218791, "epoch": 2458, "lr": 9.033513991322883e-06}
{"train_loss": 0.031049566343426704, "global_step": 218792, "epoch": 2458, "lr": 9.033181617976854e-06}
{"train_loss": 0.0391545295715332, "global_step": 218793, "epoch": 2458, "lr": 9.032849250138297e-06}
{"train_loss": 0.03896055370569229, "global_step": 218794, "epoch": 2458, "lr": 9.032516887807285e-06}
{"train_loss": 0.030295483767986298, "global_step": 218795, "epoch": 2458, "lr": 9.032184530983839e-06}
{"train_loss": 0.022219320759177208, "global_step": 218796, "epoch": 2458, "lr": 9.031852179668032e-06}
{"train_loss": 0.04290752485394478, "global_step": 218797, "epoch": 2458, "lr": 9.031519833859881e-06}
{"train_loss": 0.04027852043509483, "global_step": 218798, "epoch": 2458, "lr": 9.03118749355944e-06}
{"train_loss": 0.05566529929637909, "global_step": 218799, "epoch": 2458, "lr": 9.030855158766772e-06}
{"train_loss": 0.07587581127882004, "global_step": 218800, "epoch": 2458, "lr": 9.03052282948189e-06}
{"train_loss": 0.0843655914068222, "global_step": 218801, "epoch": 2458, "lr": 9.030190505704871e-06}
{"train_loss": 0.040553171187639236, "global_step": 218802, "epoch": 2458, "lr": 9.02985818743573e-06}
{"train_loss": 0.036042992025613785, "global_step": 218803, "epoch": 2458, "lr": 9.029525874674538e-06}
{"train_loss": 0.03193514421582222, "global_step": 218804, "epoch": 2458, "lr": 9.029193567421312e-06}
{"train_loss": 0.04453254118561745, "global_step": 218805, "epoch": 2458, "lr": 9.028861265676125e-06}
{"train_loss": 0.08420472592115402, "global_step": 218806, "epoch": 2458, "lr": 9.028528969438998e-06}
{"train_loss": 0.04621542990207672, "global_step": 218807, "epoch": 2458, "lr": 9.02819667871e-06}
{"train_loss": 0.042143188416957855, "global_step": 218808, "epoch": 2458, "lr": 9.027864393489144e-06}
{"train_loss": 0.09879665821790695, "global_step": 218809, "epoch": 2458, "lr": 9.027532113776505e-06}
{"train_loss": 0.06042398884892464, "global_step": 218810, "epoch": 2458, "lr": 9.027199839572104e-06}
{"train_loss": 0.09488356858491898, "global_step": 218811, "epoch": 2458, "lr": 9.026867570876013e-06}
{"train_loss": 0.05447806045413017, "global_step": 218812, "epoch": 2458, "lr": 9.026535307688238e-06}
{"train_loss": 0.13966526091098785, "global_step": 218813, "epoch": 2458, "lr": 9.02620305000887e-06}
{"train_loss": 0.07130080461502075, "global_step": 218814, "epoch": 2458, "lr": 9.02587079783791e-06}
{"train_loss": 0.06473122537136078, "global_step": 218815, "epoch": 2458, "lr": 9.025538551175432e-06}
{"train_loss": 0.0539378896355629, "global_step": 218816, "epoch": 2458, "lr": 9.025206310021466e-06}
{"train_loss": 0.07043816149234772, "global_step": 218817, "epoch": 2458, "lr": 9.024874074376072e-06}
{"train_loss": 0.03671877086162567, "global_step": 218818, "epoch": 2458, "lr": 9.02454184423927e-06}
{"train_loss": 0.024595338851213455, "global_step": 218819, "epoch": 2458, "lr": 9.024209619611129e-06}
{"train_loss": 0.020591378211975098, "global_step": 218820, "epoch": 2458, "lr": 9.023877400491676e-06}
{"train_loss": 0.07179640233516693, "global_step": 218821, "epoch": 2458, "lr": 9.023545186880978e-06}
{"train_loss": 0.08617573231458664, "global_step": 218822, "epoch": 2458, "lr": 9.023212978779056e-06}
{"train_loss": 0.06930780410766602, "global_step": 218823, "epoch": 2458, "lr": 9.022880776185954e-06}
{"train_loss": 0.052370335906744, "global_step": 218824, "epoch": 2458, "lr": 9.022548579101742e-06}
{"train_loss": 0.09595311433076859, "global_step": 218825, "epoch": 2458, "lr": 9.022216387526439e-06}
{"train_loss": 0.04799816384911537, "global_step": 218826, "epoch": 2458, "lr": 9.021884201460107e-06}
{"train_loss": 0.042014118283987045, "global_step": 218827, "epoch": 2458, "lr": 9.021552020902768e-06}
{"train_loss": 0.045572683215141296, "global_step": 218828, "epoch": 2458, "lr": 9.0212198458545e-06}
{"train_loss": 0.06491014361381531, "global_step": 218829, "epoch": 2458, "lr": 9.020887676315315e-06}
{"train_loss": 0.06414450705051422, "global_step": 218830, "epoch": 2458, "lr": 9.020555512285284e-06}
{"train_loss": 0.057234376668930054, "global_step": 218831, "epoch": 2458, "lr": 9.020223353764429e-06}
{"train_loss": 0.023351524025201797, "global_step": 218832, "epoch": 2458, "lr": 9.019891200752818e-06}
{"train_loss": 0.056102558970451355, "global_step": 218833, "epoch": 2458, "lr": 9.01955905325047e-06}
{"train_loss": 0.07416853308677673, "global_step": 218834, "epoch": 2458, "lr": 9.019226911257445e-06}
{"train_loss": 0.04820607602596283, "global_step": 218835, "epoch": 2458, "lr": 9.018894774773794e-06}
{"train_loss": 0.03453686460852623, "global_step": 218836, "epoch": 2458, "lr": 9.018562643799545e-06}
{"train_loss": 0.04482681304216385, "global_step": 218837, "epoch": 2458, "lr": 9.018230518334758e-06}
{"train_loss": 0.05520636588335037, "global_step": 218838, "epoch": 2458, "lr": 9.01789839837946e-06}
{"train_loss": 0.03833847492933273, "global_step": 218839, "epoch": 2458, "lr": 9.017566283933721e-06}
{"train_loss": 0.08598202466964722, "global_step": 218840, "epoch": 2458, "lr": 9.017234174997553e-06}
{"train_loss": 0.04561120271682739, "global_step": 218841, "epoch": 2458, "lr": 9.016902071571032e-06}
{"train_loss": 0.10513558983802795, "global_step": 218842, "epoch": 2458, "lr": 9.016569973654183e-06}
{"train_loss": 0.06235503777861595, "global_step": 218843, "epoch": 2458, "lr": 9.016237881247064e-06}
{"train_loss": 0.04955852031707764, "global_step": 218844, "epoch": 2458, "lr": 9.0159057943497e-06}
{"train_loss": 0.03812630474567413, "global_step": 218845, "epoch": 2458, "lr": 9.015573712962162e-06}
{"train_loss": 0.06987849622964859, "global_step": 218846, "epoch": 2458, "lr": 9.015241637084464e-06}
{"train_loss": 0.05991306155920029, "global_step": 218847, "epoch": 2458, "lr": 9.014909566716684e-06}
{"train_loss": 0.06682382524013519, "global_step": 218848, "epoch": 2458, "lr": 9.014577501858834e-06}
{"train_loss": 0.03156578168272972, "global_step": 218849, "epoch": 2458, "lr": 9.014245442510988e-06}
{"train_loss": 0.059966179950351126, "global_step": 218850, "epoch": 2458, "lr": 9.013913388673168e-06, "val_loss": 8.843339920043945}
{"train_loss": 0.048405442386865616, "global_step": 218851, "epoch": 2459, "lr": 9.013581340345439e-06}
{"train_loss": 0.038222793489694595, "global_step": 218852, "epoch": 2459, "lr": 9.01324929752782e-06}
{"train_loss": 0.06310416013002396, "global_step": 218853, "epoch": 2459, "lr": 9.012917260220371e-06}
{"train_loss": 0.030578389763832092, "global_step": 218854, "epoch": 2459, "lr": 9.012585228423149e-06}
{"train_loss": 0.026051770895719528, "global_step": 218855, "epoch": 2459, "lr": 9.012253202136185e-06}
{"train_loss": 0.04419389367103577, "global_step": 218856, "epoch": 2459, "lr": 9.011921181359512e-06}
{"train_loss": 0.059695784002542496, "global_step": 218857, "epoch": 2459, "lr": 9.011589166093198e-06}
{"train_loss": 0.04141053557395935, "global_step": 218858, "epoch": 2459, "lr": 9.011257156337261e-06}
{"train_loss": 0.04082654044032097, "global_step": 218859, "epoch": 2459, "lr": 9.010925152091776e-06}
{"train_loss": 0.047459542751312256, "global_step": 218860, "epoch": 2459, "lr": 9.010593153356773e-06}
{"train_loss": 0.03824569284915924, "global_step": 218861, "epoch": 2459, "lr": 9.010261160132282e-06}
{"train_loss": 0.03597567602992058, "global_step": 218862, "epoch": 2459, "lr": 9.00992917241838e-06}
{"train_loss": 0.04525812715291977, "global_step": 218863, "epoch": 2459, "lr": 9.009597190215075e-06}
{"train_loss": 0.05620918422937393, "global_step": 218864, "epoch": 2459, "lr": 9.009265213522445e-06}
{"train_loss": 0.058726076036691666, "global_step": 218865, "epoch": 2459, "lr": 9.008933242340507e-06}
{"train_loss": 0.0842408686876297, "global_step": 218866, "epoch": 2459, "lr": 9.008601276669331e-06}
{"train_loss": 0.042213357985019684, "global_step": 218867, "epoch": 2459, "lr": 9.008269316508933e-06}
{"train_loss": 0.036223817616701126, "global_step": 218868, "epoch": 2459, "lr": 9.007937361859392e-06}
{"train_loss": 0.0886513888835907, "global_step": 218869, "epoch": 2459, "lr": 9.007605412720715e-06}
{"train_loss": 0.02879103273153305, "global_step": 218870, "epoch": 2459, "lr": 9.007273469092986e-06}
{"train_loss": 0.07519508898258209, "global_step": 218871, "epoch": 2459, "lr": 9.00694153097621e-06}
{"train_loss": 0.017251286655664444, "global_step": 218872, "epoch": 2459, "lr": 9.006609598370453e-06}
{"train_loss": 0.08469296991825104, "global_step": 218873, "epoch": 2459, "lr": 9.00627767127577e-06}
{"train_loss": 0.02469755709171295, "global_step": 218874, "epoch": 2459, "lr": 9.005945749692185e-06}
{"train_loss": 0.041334472596645355, "global_step": 218875, "epoch": 2459, "lr": 9.005613833619758e-06}
{"train_loss": 0.02318560890853405, "global_step": 218876, "epoch": 2459, "lr": 9.005281923058518e-06}
{"train_loss": 0.05990142375230789, "global_step": 218877, "epoch": 2459, "lr": 9.00495001800853e-06}
{"train_loss": 0.06409533321857452, "global_step": 218878, "epoch": 2459, "lr": 9.004618118469815e-06}
{"train_loss": 0.021232884377241135, "global_step": 218879, "epoch": 2459, "lr": 9.004286224442438e-06}
{"train_loss": 0.030642352998256683, "global_step": 218880, "epoch": 2459, "lr": 9.003954335926423e-06}
{"train_loss": 0.05103529617190361, "global_step": 218881, "epoch": 2459, "lr": 9.003622452921846e-06}
{"train_loss": 0.03496497496962547, "global_step": 218882, "epoch": 2459, "lr": 9.003290575428714e-06}
{"train_loss": 0.05501510202884674, "global_step": 218883, "epoch": 2459, "lr": 9.0029587034471e-06}
{"train_loss": 0.03467017784714699, "global_step": 218884, "epoch": 2459, "lr": 9.00262683697703e-06}
{"train_loss": 0.02700391225516796, "global_step": 218885, "epoch": 2459, "lr": 9.002294976018571e-06}
{"train_loss": 0.09833677858114243, "global_step": 218886, "epoch": 2459, "lr": 9.001963120571738e-06}
{"train_loss": 0.04762774333357811, "global_step": 218887, "epoch": 2459, "lr": 9.001631270636607e-06}
{"train_loss": 0.039879970252513885, "global_step": 218888, "epoch": 2459, "lr": 9.00129942621319e-06}
{"train_loss": 0.03615724295377731, "global_step": 218889, "epoch": 2459, "lr": 9.000967587301551e-06}
{"train_loss": 0.038849737495183945, "global_step": 218890, "epoch": 2459, "lr": 9.000635753901743e-06}
{"train_loss": 0.073372483253479, "global_step": 218891, "epoch": 2459, "lr": 9.000303926013791e-06}
{"train_loss": 0.09320113062858582, "global_step": 218892, "epoch": 2459, "lr": 8.999972103637761e-06}
{"train_loss": 0.06350508332252502, "global_step": 218893, "epoch": 2459, "lr": 8.99964028677368e-06}
{"train_loss": 0.0721617192029953, "global_step": 218894, "epoch": 2459, "lr": 8.999308475421591e-06}
{"train_loss": 0.07632069289684296, "global_step": 218895, "epoch": 2459, "lr": 8.998976669581555e-06}
{"train_loss": 0.04075988754630089, "global_step": 218896, "epoch": 2459, "lr": 8.998644869253592e-06}
{"train_loss": 0.031175700947642326, "global_step": 218897, "epoch": 2459, "lr": 8.998313074437776e-06}
{"train_loss": 0.047786012291908264, "global_step": 218898, "epoch": 2459, "lr": 8.997981285134139e-06}
{"train_loss": 0.10057695209980011, "global_step": 218899, "epoch": 2459, "lr": 8.997649501342709e-06}
{"train_loss": 0.04087219759821892, "global_step": 218900, "epoch": 2459, "lr": 8.997317723063559e-06}
{"train_loss": 0.08674183487892151, "global_step": 218901, "epoch": 2459, "lr": 8.996985950296705e-06}
{"train_loss": 0.06454862654209137, "global_step": 218902, "epoch": 2459, "lr": 8.99665418304222e-06}
{"train_loss": 0.02108658291399479, "global_step": 218903, "epoch": 2459, "lr": 8.996322421300125e-06}
{"train_loss": 0.05017087236046791, "global_step": 218904, "epoch": 2459, "lr": 8.995990665070486e-06}
{"train_loss": 0.06667587161064148, "global_step": 218905, "epoch": 2459, "lr": 8.995658914353323e-06}
{"train_loss": 0.06429129838943481, "global_step": 218906, "epoch": 2459, "lr": 8.995327169148704e-06}
{"train_loss": 0.03720099478960037, "global_step": 218907, "epoch": 2459, "lr": 8.994995429456654e-06}
{"train_loss": 0.034944579005241394, "global_step": 218908, "epoch": 2459, "lr": 8.994663695277228e-06}
{"train_loss": 0.030668223276734352, "global_step": 218909, "epoch": 2459, "lr": 8.99433196661048e-06}
{"train_loss": 0.022834639996290207, "global_step": 218910, "epoch": 2459, "lr": 8.994000243456435e-06}
{"train_loss": 0.08560102432966232, "global_step": 218911, "epoch": 2459, "lr": 8.993668525815152e-06}
{"train_loss": 0.06629177182912827, "global_step": 218912, "epoch": 2459, "lr": 8.993336813686665e-06}
{"train_loss": 0.08202891796827316, "global_step": 218913, "epoch": 2459, "lr": 8.993005107071035e-06}
{"train_loss": 0.04673971235752106, "global_step": 218914, "epoch": 2459, "lr": 8.992673405968283e-06}
{"train_loss": 0.029581528156995773, "global_step": 218915, "epoch": 2459, "lr": 8.992341710378477e-06}
{"train_loss": 0.04213954508304596, "global_step": 218916, "epoch": 2459, "lr": 8.992010020301639e-06}
{"train_loss": 0.043322645127773285, "global_step": 218917, "epoch": 2459, "lr": 8.991678335737842e-06}
{"train_loss": 0.04068262502551079, "global_step": 218918, "epoch": 2459, "lr": 8.991346656687094e-06}
{"train_loss": 0.06777841597795486, "global_step": 218919, "epoch": 2459, "lr": 8.991014983149482e-06}
{"train_loss": 0.053516607731580734, "global_step": 218920, "epoch": 2459, "lr": 8.990683315125008e-06}
{"train_loss": 0.04627301171422005, "global_step": 218921, "epoch": 2459, "lr": 8.990351652613748e-06}
{"train_loss": 0.04014093801379204, "global_step": 218922, "epoch": 2459, "lr": 8.990019995615729e-06}
{"train_loss": 0.07778697460889816, "global_step": 218923, "epoch": 2459, "lr": 8.98968834413101e-06}
{"train_loss": 0.04624778777360916, "global_step": 218924, "epoch": 2459, "lr": 8.989356698159618e-06}
{"train_loss": 0.02879651077091694, "global_step": 218925, "epoch": 2459, "lr": 8.989025057701605e-06}
{"train_loss": 0.08876235783100128, "global_step": 218926, "epoch": 2459, "lr": 8.988693422757032e-06}
{"train_loss": 0.04599430412054062, "global_step": 218927, "epoch": 2459, "lr": 8.988361793325917e-06}
{"train_loss": 0.06278975307941437, "global_step": 218928, "epoch": 2459, "lr": 8.988030169408329e-06}
{"train_loss": 0.06723985821008682, "global_step": 218929, "epoch": 2459, "lr": 8.987698551004287e-06}
{"train_loss": 0.04527634009718895, "global_step": 218930, "epoch": 2459, "lr": 8.987366938113861e-06}
{"train_loss": 0.07557430863380432, "global_step": 218931, "epoch": 2459, "lr": 8.98703533073707e-06}
{"train_loss": 0.06306225061416626, "global_step": 218932, "epoch": 2459, "lr": 8.986703728873985e-06}
{"train_loss": 0.037522368133068085, "global_step": 218933, "epoch": 2459, "lr": 8.986372132524645e-06}
{"train_loss": 0.05951239541172981, "global_step": 218934, "epoch": 2459, "lr": 8.986040541689067e-06}
{"train_loss": 0.04637801647186279, "global_step": 218935, "epoch": 2459, "lr": 8.985708956367328e-06}
{"train_loss": 0.07327438145875931, "global_step": 218936, "epoch": 2459, "lr": 8.985377376559456e-06}
{"train_loss": 0.0659589171409607, "global_step": 218937, "epoch": 2459, "lr": 8.985045802265507e-06}
{"train_loss": 0.053183604031801224, "global_step": 218938, "epoch": 2459, "lr": 8.98471423348552e-06}
{"train_loss": 0.05185348105229688, "global_step": 218939, "epoch": 2459, "lr": 8.984382670219521e-06, "val_loss": 8.829462051391602}
{"train_loss": 0.06036084517836571, "global_step": 218940, "epoch": 2460, "lr": 8.98405111246759e-06}
{"train_loss": 0.05241279676556587, "global_step": 218941, "epoch": 2460, "lr": 8.983719560229736e-06}
{"train_loss": 0.08851831406354904, "global_step": 218942, "epoch": 2460, "lr": 8.983388013506039e-06}
{"train_loss": 0.05256836116313934, "global_step": 218943, "epoch": 2460, "lr": 8.983056472296513e-06}
{"train_loss": 0.05865049362182617, "global_step": 218944, "epoch": 2460, "lr": 8.982724936601212e-06}
{"train_loss": 0.037883441895246506, "global_step": 218945, "epoch": 2460, "lr": 8.982393406420198e-06}
{"train_loss": 0.049835093319416046, "global_step": 218946, "epoch": 2460, "lr": 8.982061881753483e-06}
{"train_loss": 0.04807337373495102, "global_step": 218947, "epoch": 2460, "lr": 8.981730362601148e-06}
{"train_loss": 0.09344279766082764, "global_step": 218948, "epoch": 2460, "lr": 8.981398848963208e-06}
{"train_loss": 0.056513335555791855, "global_step": 218949, "epoch": 2460, "lr": 8.981067340839726e-06}
{"train_loss": 0.03253931552171707, "global_step": 218950, "epoch": 2460, "lr": 8.980735838230736e-06}
{"train_loss": 0.048691608011722565, "global_step": 218951, "epoch": 2460, "lr": 8.98040434113629e-06}
{"train_loss": 0.025017376989126205, "global_step": 218952, "epoch": 2460, "lr": 8.98007284955642e-06}
{"train_loss": 0.07179082185029984, "global_step": 218953, "epoch": 2460, "lr": 8.97974136349119e-06}
{"train_loss": 0.039054784923791885, "global_step": 218954, "epoch": 2460, "lr": 8.979409882940626e-06}
{"train_loss": 0.0468132384121418, "global_step": 218955, "epoch": 2460, "lr": 8.979078407904789e-06}
{"train_loss": 0.09362784028053284, "global_step": 218956, "epoch": 2460, "lr": 8.978746938383702e-06}
{"train_loss": 0.05185385048389435, "global_step": 218957, "epoch": 2460, "lr": 8.978415474377439e-06}
{"train_loss": 0.051583532243967056, "global_step": 218958, "epoch": 2460, "lr": 8.978084015886007e-06}
{"train_loss": 0.06143539026379585, "global_step": 218959, "epoch": 2460, "lr": 8.977752562909492e-06}
{"train_loss": 0.03612547740340233, "global_step": 218960, "epoch": 2460, "lr": 8.977421115447905e-06}
{"train_loss": 0.031079420819878578, "global_step": 218961, "epoch": 2460, "lr": 8.9770896735013e-06}
{"train_loss": 0.054933831095695496, "global_step": 218962, "epoch": 2460, "lr": 8.976758237069738e-06}
{"train_loss": 0.02632707916200161, "global_step": 218963, "epoch": 2460, "lr": 8.976426806153244e-06}
{"train_loss": 0.0632704347372055, "global_step": 218964, "epoch": 2460, "lr": 8.976095380751876e-06}
{"train_loss": 0.03576147183775902, "global_step": 218965, "epoch": 2460, "lr": 8.975763960865663e-06}
{"train_loss": 0.11159680038690567, "global_step": 218966, "epoch": 2460, "lr": 8.975432546494666e-06}
{"train_loss": 0.010431844741106033, "global_step": 218967, "epoch": 2460, "lr": 8.975101137638914e-06}
{"train_loss": 0.03719194233417511, "global_step": 218968, "epoch": 2460, "lr": 8.974769734298472e-06}
{"train_loss": 0.03943730518221855, "global_step": 218969, "epoch": 2460, "lr": 8.974438336473356e-06}
{"train_loss": 0.056773025542497635, "global_step": 218970, "epoch": 2460, "lr": 8.974106944163646e-06}
{"train_loss": 0.050292324274778366, "global_step": 218971, "epoch": 2460, "lr": 8.973775557369362e-06}
{"train_loss": 0.03164227306842804, "global_step": 218972, "epoch": 2460, "lr": 8.97344417609054e-06}
{"train_loss": 0.04629606008529663, "global_step": 218973, "epoch": 2460, "lr": 8.973112800327254e-06}
{"train_loss": 0.057668544352054596, "global_step": 218974, "epoch": 2460, "lr": 8.972781430079518e-06}
{"train_loss": 0.08911822736263275, "global_step": 218975, "epoch": 2460, "lr": 8.97245006534741e-06}
{"train_loss": 0.04839329048991203, "global_step": 218976, "epoch": 2460, "lr": 8.97211870613095e-06}
{"train_loss": 0.10029207915067673, "global_step": 218977, "epoch": 2460, "lr": 8.971787352430178e-06}
{"train_loss": 0.055437877774238586, "global_step": 218978, "epoch": 2460, "lr": 8.97145600424516e-06}
{"train_loss": 0.07064002007246017, "global_step": 218979, "epoch": 2460, "lr": 8.971124661575919e-06}
{"train_loss": 0.07685844600200653, "global_step": 218980, "epoch": 2460, "lr": 8.97079332442251e-06}
{"train_loss": 0.05289726331830025, "global_step": 218981, "epoch": 2460, "lr": 8.970461992784995e-06}
{"train_loss": 0.03626840561628342, "global_step": 218982, "epoch": 2460, "lr": 8.970130666663384e-06}
{"train_loss": 0.06839066743850708, "global_step": 218983, "epoch": 2460, "lr": 8.96979934605775e-06}
{"train_loss": 0.10818266868591309, "global_step": 218984, "epoch": 2460, "lr": 8.969468030968114e-06}
{"train_loss": 0.014615417458117008, "global_step": 218985, "epoch": 2460, "lr": 8.96913672139455e-06}
{"train_loss": 0.08728925883769989, "global_step": 218986, "epoch": 2460, "lr": 8.968805417337067e-06}
{"train_loss": 0.04839051142334938, "global_step": 218987, "epoch": 2460, "lr": 8.968474118795745e-06}
{"train_loss": 0.019457800313830376, "global_step": 218988, "epoch": 2460, "lr": 8.968142825770599e-06}
{"train_loss": 0.0833362564444542, "global_step": 218989, "epoch": 2460, "lr": 8.967811538261695e-06}
{"train_loss": 0.10188556462526321, "global_step": 218990, "epoch": 2460, "lr": 8.967480256269056e-06}
{"train_loss": 0.04994797706604004, "global_step": 218991, "epoch": 2460, "lr": 8.967148979792755e-06}
{"train_loss": 0.050399478524923325, "global_step": 218992, "epoch": 2460, "lr": 8.966817708832808e-06}
{"train_loss": 0.07128491252660751, "global_step": 218993, "epoch": 2460, "lr": 8.966486443389282e-06}
{"train_loss": 0.09866395592689514, "global_step": 218994, "epoch": 2460, "lr": 8.9661551834622e-06}
{"train_loss": 0.10951922833919525, "global_step": 218995, "epoch": 2460, "lr": 8.965823929051631e-06}
{"train_loss": 0.0676206573843956, "global_step": 218996, "epoch": 2460, "lr": 8.965492680157594e-06}
{"train_loss": 0.045266129076480865, "global_step": 218997, "epoch": 2460, "lr": 8.965161436780161e-06}
{"train_loss": 0.06518392264842987, "global_step": 218998, "epoch": 2460, "lr": 8.96483019891935e-06}
{"train_loss": 0.07275316119194031, "global_step": 218999, "epoch": 2460, "lr": 8.964498966575213e-06}
{"train_loss": 0.07090417295694351, "global_step": 219000, "epoch": 2460, "lr": 8.96416773974782e-06}
{"train_loss": 0.03482166677713394, "global_step": 219001, "epoch": 2460, "lr": 8.963836518437173e-06}
{"train_loss": 0.09688456356525421, "global_step": 219002, "epoch": 2460, "lr": 8.963505302643354e-06}
{"train_loss": 0.1000908613204956, "global_step": 219003, "epoch": 2460, "lr": 8.963174092366378e-06}
{"train_loss": 0.055490974336862564, "global_step": 219004, "epoch": 2460, "lr": 8.962842887606315e-06}
{"train_loss": 0.04478742182254791, "global_step": 219005, "epoch": 2460, "lr": 8.96251168836319e-06}
{"train_loss": 0.05908520147204399, "global_step": 219006, "epoch": 2460, "lr": 8.962180494637062e-06}
{"train_loss": 0.06194000691175461, "global_step": 219007, "epoch": 2460, "lr": 8.96184930642796e-06}
{"train_loss": 0.02923429198563099, "global_step": 219008, "epoch": 2460, "lr": 8.961518123735947e-06}
{"train_loss": 0.06421365588903427, "global_step": 219009, "epoch": 2460, "lr": 8.961186946561057e-06}
{"train_loss": 0.03135194256901741, "global_step": 219010, "epoch": 2460, "lr": 8.960855774903326e-06}
{"train_loss": 0.05220147222280502, "global_step": 219011, "epoch": 2460, "lr": 8.960524608762822e-06}
{"train_loss": 0.041543249040842056, "global_step": 219012, "epoch": 2460, "lr": 8.960193448139554e-06}
{"train_loss": 0.018791941925883293, "global_step": 219013, "epoch": 2460, "lr": 8.959862293033606e-06}
{"train_loss": 0.04450450837612152, "global_step": 219014, "epoch": 2460, "lr": 8.959531143445005e-06}
{"train_loss": 0.08083915710449219, "global_step": 219015, "epoch": 2460, "lr": 8.959199999373779e-06}
{"train_loss": 0.024596622213721275, "global_step": 219016, "epoch": 2460, "lr": 8.958868860819991e-06}
{"train_loss": 0.09014252573251724, "global_step": 219017, "epoch": 2460, "lr": 8.9585377277837e-06}
{"train_loss": 0.06679242849349976, "global_step": 219018, "epoch": 2460, "lr": 8.958206600264912e-06}
{"train_loss": 0.08956655859947205, "global_step": 219019, "epoch": 2460, "lr": 8.95787547826371e-06}
{"train_loss": 0.05685945227742195, "global_step": 219020, "epoch": 2460, "lr": 8.957544361780107e-06}
{"train_loss": 0.03864440694451332, "global_step": 219021, "epoch": 2460, "lr": 8.957213250814177e-06}
{"train_loss": 0.03019678220152855, "global_step": 219022, "epoch": 2460, "lr": 8.956882145365936e-06}
{"train_loss": 0.06892840564250946, "global_step": 219023, "epoch": 2460, "lr": 8.956551045435451e-06}
{"train_loss": 0.04570811614394188, "global_step": 219024, "epoch": 2460, "lr": 8.956219951022748e-06}
{"train_loss": 0.026037897914648056, "global_step": 219025, "epoch": 2460, "lr": 8.955888862127898e-06}
{"train_loss": 0.009944235906004906, "global_step": 219026, "epoch": 2460, "lr": 8.955557778750911e-06}
{"train_loss": 0.03269009664654732, "global_step": 219027, "epoch": 2460, "lr": 8.95522670089186e-06}
{"train_loss": 0.05650184846619207, "global_step": 219028, "epoch": 2460, "lr": 8.954895628550769e-06, "val_loss": 8.7501802444458, "train_action_mse_error": 26.389530181884766}
{"train_loss": 0.05996780842542648, "global_step": 219029, "epoch": 2461, "lr": 8.954564561727702e-06}
{"train_loss": 0.06012105569243431, "global_step": 219030, "epoch": 2461, "lr": 8.954233500422681e-06}
{"train_loss": 0.048869289457798004, "global_step": 219031, "epoch": 2461, "lr": 8.953902444635781e-06}
{"train_loss": 0.054907359182834625, "global_step": 219032, "epoch": 2461, "lr": 8.953571394367011e-06}
{"train_loss": 0.05813731253147125, "global_step": 219033, "epoch": 2461, "lr": 8.953240349616442e-06}
{"train_loss": 0.06042002514004707, "global_step": 219034, "epoch": 2461, "lr": 8.952909310384105e-06}
{"train_loss": 0.08664570748806, "global_step": 219035, "epoch": 2461, "lr": 8.952578276670042e-06}
{"train_loss": 0.04427086189389229, "global_step": 219036, "epoch": 2461, "lr": 8.95224724847432e-06}
{"train_loss": 0.08007153868675232, "global_step": 219037, "epoch": 2461, "lr": 8.951916225796958e-06}
{"train_loss": 0.10450275242328644, "global_step": 219038, "epoch": 2461, "lr": 8.951585208638025e-06}
{"train_loss": 0.03177523985505104, "global_step": 219039, "epoch": 2461, "lr": 8.951254196997533e-06}
{"train_loss": 0.09293840080499649, "global_step": 219040, "epoch": 2461, "lr": 8.95092319087556e-06}
{"train_loss": 0.0579611212015152, "global_step": 219041, "epoch": 2461, "lr": 8.950592190272123e-06}
{"train_loss": 0.026102038100361824, "global_step": 219042, "epoch": 2461, "lr": 8.950261195187287e-06}
{"train_loss": 0.10189189016819, "global_step": 219043, "epoch": 2461, "lr": 8.949930205621076e-06}
{"train_loss": 0.046368036419153214, "global_step": 219044, "epoch": 2461, "lr": 8.949599221573562e-06}
{"train_loss": 0.04226653650403023, "global_step": 219045, "epoch": 2461, "lr": 8.949268243044762e-06}
{"train_loss": 0.07773522287607193, "global_step": 219046, "epoch": 2461, "lr": 8.948937270034747e-06}
{"train_loss": 0.11021754145622253, "global_step": 219047, "epoch": 2461, "lr": 8.948606302543527e-06}
{"train_loss": 0.05508396774530411, "global_step": 219048, "epoch": 2461, "lr": 8.948275340571183e-06}
{"train_loss": 0.08599797636270523, "global_step": 219049, "epoch": 2461, "lr": 8.947944384117746e-06}
{"train_loss": 0.056986428797245026, "global_step": 219050, "epoch": 2461, "lr": 8.947613433183238e-06}
{"train_loss": 0.04079336300492287, "global_step": 219051, "epoch": 2461, "lr": 8.947282487767743e-06}
{"train_loss": 0.09120682626962662, "global_step": 219052, "epoch": 2461, "lr": 8.946951547871268e-06}
{"train_loss": 0.09479434043169022, "global_step": 219053, "epoch": 2461, "lr": 8.94662061349389e-06}
{"train_loss": 0.07051496207714081, "global_step": 219054, "epoch": 2461, "lr": 8.946289684635628e-06}
{"train_loss": 0.05307428166270256, "global_step": 219055, "epoch": 2461, "lr": 8.945958761296547e-06}
{"train_loss": 0.03435873985290527, "global_step": 219056, "epoch": 2461, "lr": 8.945627843476668e-06}
{"train_loss": 0.08299034833908081, "global_step": 219057, "epoch": 2461, "lr": 8.945296931176061e-06}
{"train_loss": 0.025531897321343422, "global_step": 219058, "epoch": 2461, "lr": 8.944966024394746e-06}
{"train_loss": 0.019339099526405334, "global_step": 219059, "epoch": 2461, "lr": 8.944635123132795e-06}
{"train_loss": 0.04387544468045235, "global_step": 219060, "epoch": 2461, "lr": 8.94430422739022e-06}
{"train_loss": 0.033002860844135284, "global_step": 219061, "epoch": 2461, "lr": 8.9439733371671e-06}
{"train_loss": 0.017743943259119987, "global_step": 219062, "epoch": 2461, "lr": 8.943642452463452e-06}
{"train_loss": 0.06262185424566269, "global_step": 219063, "epoch": 2461, "lr": 8.943311573279339e-06}
{"train_loss": 0.032405465841293335, "global_step": 219064, "epoch": 2461, "lr": 8.942980699614783e-06}
{"train_loss": 0.01589561253786087, "global_step": 219065, "epoch": 2461, "lr": 8.942649831469862e-06}
{"train_loss": 0.05578986927866936, "global_step": 219066, "epoch": 2461, "lr": 8.942318968844583e-06}
{"train_loss": 0.04392249137163162, "global_step": 219067, "epoch": 2461, "lr": 8.941988111739025e-06}
{"train_loss": 0.0736742913722992, "global_step": 219068, "epoch": 2461, "lr": 8.941657260153202e-06}
{"train_loss": 0.040739040821790695, "global_step": 219069, "epoch": 2461, "lr": 8.941326414087187e-06}
{"train_loss": 0.030716078355908394, "global_step": 219070, "epoch": 2461, "lr": 8.940995573540995e-06}
{"train_loss": 0.034641850739717484, "global_step": 219071, "epoch": 2461, "lr": 8.940664738514686e-06}
{"train_loss": 0.051943209022283554, "global_step": 219072, "epoch": 2461, "lr": 8.940333909008319e-06}
{"train_loss": 0.03421718254685402, "global_step": 219073, "epoch": 2461, "lr": 8.940003085021909e-06}
{"train_loss": 0.08747461438179016, "global_step": 219074, "epoch": 2461, "lr": 8.939672266555526e-06}
{"train_loss": 0.07409439980983734, "global_step": 219075, "epoch": 2461, "lr": 8.939341453609196e-06}
{"train_loss": 0.03192171826958656, "global_step": 219076, "epoch": 2461, "lr": 8.939010646182983e-06}
{"train_loss": 0.07713967561721802, "global_step": 219077, "epoch": 2461, "lr": 8.938679844276904e-06}
{"train_loss": 0.0854199156165123, "global_step": 219078, "epoch": 2461, "lr": 8.938349047891032e-06}
{"train_loss": 0.044248294085264206, "global_step": 219079, "epoch": 2461, "lr": 8.938018257025388e-06}
{"train_loss": 0.03603113070130348, "global_step": 219080, "epoch": 2461, "lr": 8.937687471680034e-06}
{"train_loss": 0.08370888233184814, "global_step": 219081, "epoch": 2461, "lr": 8.937356691855003e-06}
{"train_loss": 0.022158073261380196, "global_step": 219082, "epoch": 2461, "lr": 8.937025917550352e-06}
{"train_loss": 0.02733713388442993, "global_step": 219083, "epoch": 2461, "lr": 8.936695148766106e-06}
{"train_loss": 0.09138847887516022, "global_step": 219084, "epoch": 2461, "lr": 8.936364385502332e-06}
{"train_loss": 0.026497570797801018, "global_step": 219085, "epoch": 2461, "lr": 8.936033627759049e-06}
{"train_loss": 0.10665197670459747, "global_step": 219086, "epoch": 2461, "lr": 8.935702875536333e-06}
{"train_loss": 0.042667221277952194, "global_step": 219087, "epoch": 2461, "lr": 8.935372128834207e-06}
{"train_loss": 0.037547096610069275, "global_step": 219088, "epoch": 2461, "lr": 8.93504138765271e-06}
{"train_loss": 0.09452158957719803, "global_step": 219089, "epoch": 2461, "lr": 8.934710651991907e-06}
{"train_loss": 0.039680659770965576, "global_step": 219090, "epoch": 2461, "lr": 8.934379921851821e-06}
{"train_loss": 0.01787479780614376, "global_step": 219091, "epoch": 2461, "lr": 8.93404919723252e-06}
{"train_loss": 0.019121374934911728, "global_step": 219092, "epoch": 2461, "lr": 8.933718478134024e-06}
{"train_loss": 0.03203433379530907, "global_step": 219093, "epoch": 2461, "lr": 8.933387764556395e-06}
{"train_loss": 0.04839245229959488, "global_step": 219094, "epoch": 2461, "lr": 8.933057056499666e-06}
{"train_loss": 0.033019401133060455, "global_step": 219095, "epoch": 2461, "lr": 8.932726353963894e-06}
{"train_loss": 0.06626918166875839, "global_step": 219096, "epoch": 2461, "lr": 8.932395656949105e-06}
{"train_loss": 0.04095982387661934, "global_step": 219097, "epoch": 2461, "lr": 8.932064965455372e-06}
{"train_loss": 0.06430107355117798, "global_step": 219098, "epoch": 2461, "lr": 8.931734279482706e-06}
{"train_loss": 0.06259047985076904, "global_step": 219099, "epoch": 2461, "lr": 8.931403599031179e-06}
{"train_loss": 0.07109114527702332, "global_step": 219100, "epoch": 2461, "lr": 8.931072924100814e-06}
{"train_loss": 0.041373249143362045, "global_step": 219101, "epoch": 2461, "lr": 8.930742254691681e-06}
{"train_loss": 0.08649439364671707, "global_step": 219102, "epoch": 2461, "lr": 8.930411590803794e-06}
{"train_loss": 0.049856942147016525, "global_step": 219103, "epoch": 2461, "lr": 8.930080932437224e-06}
{"train_loss": 0.05404341220855713, "global_step": 219104, "epoch": 2461, "lr": 8.929750279591992e-06}
{"train_loss": 0.05252544954419136, "global_step": 219105, "epoch": 2461, "lr": 8.929419632268165e-06}
{"train_loss": 0.023633798584342003, "global_step": 219106, "epoch": 2461, "lr": 8.929088990465767e-06}
{"train_loss": 0.03685637190937996, "global_step": 219107, "epoch": 2461, "lr": 8.928758354184852e-06}
{"train_loss": 0.03962308168411255, "global_step": 219108, "epoch": 2461, "lr": 8.928427723425481e-06}
{"train_loss": 0.03371892869472504, "global_step": 219109, "epoch": 2461, "lr": 8.928097098187665e-06}
{"train_loss": 0.04924115538597107, "global_step": 219110, "epoch": 2461, "lr": 8.927766478471478e-06}
{"train_loss": 0.08527681231498718, "global_step": 219111, "epoch": 2461, "lr": 8.92743586427694e-06}
{"train_loss": 0.09176840633153915, "global_step": 219112, "epoch": 2461, "lr": 8.927105255604123e-06}
{"train_loss": 0.005292570684105158, "global_step": 219113, "epoch": 2461, "lr": 8.926774652453041e-06}
{"train_loss": 0.08980028331279755, "global_step": 219114, "epoch": 2461, "lr": 8.926444054823769e-06}
{"train_loss": 0.07498092204332352, "global_step": 219115, "epoch": 2461, "lr": 8.926113462716318e-06}
{"train_loss": 0.06468802690505981, "global_step": 219116, "epoch": 2461, "lr": 8.925782876130768e-06}
{"train_loss": 0.05515227494552062, "global_step": 219117, "epoch": 2461, "lr": 8.925452295067132e-06, "val_loss": 8.845270156860352}
{"train_loss": 0.041596669703722, "global_step": 219118, "epoch": 2462, "lr": 8.92512171952548e-06}
{"train_loss": 0.0508037805557251, "global_step": 219119, "epoch": 2462, "lr": 8.924791149505835e-06}
{"train_loss": 0.04495710879564285, "global_step": 219120, "epoch": 2462, "lr": 8.92446058500826e-06}
{"train_loss": 0.052913159132003784, "global_step": 219121, "epoch": 2462, "lr": 8.924130026032779e-06}
{"train_loss": 0.04675624519586563, "global_step": 219122, "epoch": 2462, "lr": 8.923799472579464e-06}
{"train_loss": 0.039729584008455276, "global_step": 219123, "epoch": 2462, "lr": 8.923468924648325e-06}
{"train_loss": 0.06570747494697571, "global_step": 219124, "epoch": 2462, "lr": 8.923138382239443e-06}
{"train_loss": 0.039266761392354965, "global_step": 219125, "epoch": 2462, "lr": 8.922807845352843e-06}
{"train_loss": 0.0348542295396328, "global_step": 219126, "epoch": 2462, "lr": 8.922477313988553e-06}
{"train_loss": 0.060544393956661224, "global_step": 219127, "epoch": 2462, "lr": 8.92214678814665e-06}
{"train_loss": 0.07681941241025925, "global_step": 219128, "epoch": 2462, "lr": 8.92181626782716e-06}
{"train_loss": 0.056973036378622055, "global_step": 219129, "epoch": 2462, "lr": 8.921485753030134e-06}
{"train_loss": 0.022192008793354034, "global_step": 219130, "epoch": 2462, "lr": 8.921155243755603e-06}
{"train_loss": 0.07702451944351196, "global_step": 219131, "epoch": 2462, "lr": 8.920824740003636e-06}
{"train_loss": 0.09262873977422714, "global_step": 219132, "epoch": 2462, "lr": 8.920494241774253e-06}
{"train_loss": 0.07281313836574554, "global_step": 219133, "epoch": 2462, "lr": 8.920163749067517e-06}
{"train_loss": 0.0382622629404068, "global_step": 219134, "epoch": 2462, "lr": 8.919833261883454e-06}
{"train_loss": 0.05424301326274872, "global_step": 219135, "epoch": 2462, "lr": 8.919502780222128e-06}
{"train_loss": 0.07811430096626282, "global_step": 219136, "epoch": 2462, "lr": 8.919172304083567e-06}
{"train_loss": 0.041347213089466095, "global_step": 219137, "epoch": 2462, "lr": 8.918841833467834e-06}
{"train_loss": 0.04717731475830078, "global_step": 219138, "epoch": 2462, "lr": 8.918511368374948e-06}
{"train_loss": 0.04413410276174545, "global_step": 219139, "epoch": 2462, "lr": 8.918180908804979e-06}
{"train_loss": 0.0639302209019661, "global_step": 219140, "epoch": 2462, "lr": 8.917850454757947e-06}
{"train_loss": 0.03879310190677643, "global_step": 219141, "epoch": 2462, "lr": 8.917520006233926e-06}
{"train_loss": 0.027588550001382828, "global_step": 219142, "epoch": 2462, "lr": 8.917189563232925e-06}
{"train_loss": 0.051512591540813446, "global_step": 219143, "epoch": 2462, "lr": 8.916859125755011e-06}
{"train_loss": 0.06996693462133408, "global_step": 219144, "epoch": 2462, "lr": 8.916528693800235e-06}
{"train_loss": 0.05413658171892166, "global_step": 219145, "epoch": 2462, "lr": 8.91619826736862e-06}
{"train_loss": 0.02587078884243965, "global_step": 219146, "epoch": 2462, "lr": 8.915867846460235e-06}
{"train_loss": 0.03259720280766487, "global_step": 219147, "epoch": 2462, "lr": 8.9155374310751e-06}
{"train_loss": 0.06906643509864807, "global_step": 219148, "epoch": 2462, "lr": 8.91520702121328e-06}
{"train_loss": 0.027806423604488373, "global_step": 219149, "epoch": 2462, "lr": 8.914876616874796e-06}
{"train_loss": 0.03217381611466408, "global_step": 219150, "epoch": 2462, "lr": 8.914546218059716e-06}
{"train_loss": 0.023637639358639717, "global_step": 219151, "epoch": 2462, "lr": 8.914215824768069e-06}
{"train_loss": 0.06223779544234276, "global_step": 219152, "epoch": 2462, "lr": 8.913885436999914e-06}
{"train_loss": 0.037140004336833954, "global_step": 219153, "epoch": 2462, "lr": 8.913555054755273e-06}
{"train_loss": 0.07080164551734924, "global_step": 219154, "epoch": 2462, "lr": 8.913224678034215e-06}
{"train_loss": 0.06861509382724762, "global_step": 219155, "epoch": 2462, "lr": 8.912894306836766e-06}
{"train_loss": 0.049885593354701996, "global_step": 219156, "epoch": 2462, "lr": 8.912563941162987e-06}
{"train_loss": 0.038313232362270355, "global_step": 219157, "epoch": 2462, "lr": 8.912233581012902e-06}
{"train_loss": 0.057094424962997437, "global_step": 219158, "epoch": 2462, "lr": 8.911903226386575e-06}
{"train_loss": 0.09714058041572571, "global_step": 219159, "epoch": 2462, "lr": 8.91157287728403e-06}
{"train_loss": 0.04892212525010109, "global_step": 219160, "epoch": 2462, "lr": 8.911242533705338e-06}
{"train_loss": 0.045356884598731995, "global_step": 219161, "epoch": 2462, "lr": 8.910912195650517e-06}
{"train_loss": 0.05337192118167877, "global_step": 219162, "epoch": 2462, "lr": 8.910581863119633e-06}
{"train_loss": 0.056981898844242096, "global_step": 219163, "epoch": 2462, "lr": 8.910251536112707e-06}
{"train_loss": 0.03522169589996338, "global_step": 219164, "epoch": 2462, "lr": 8.909921214629812e-06}
{"train_loss": 0.06054500490427017, "global_step": 219165, "epoch": 2462, "lr": 8.909590898670977e-06}
{"train_loss": 0.08080198615789413, "global_step": 219166, "epoch": 2462, "lr": 8.909260588236235e-06}
{"train_loss": 0.021927284076809883, "global_step": 219167, "epoch": 2462, "lr": 8.908930283325656e-06}
{"train_loss": 0.07331356406211853, "global_step": 219168, "epoch": 2462, "lr": 8.908599983939254e-06}
{"train_loss": 0.028698384761810303, "global_step": 219169, "epoch": 2462, "lr": 8.908269690077103e-06}
{"train_loss": 0.05507355555891991, "global_step": 219170, "epoch": 2462, "lr": 8.907939401739224e-06}
{"train_loss": 0.037317968904972076, "global_step": 219171, "epoch": 2462, "lr": 8.907609118925681e-06}
{"train_loss": 0.028854714706540108, "global_step": 219172, "epoch": 2462, "lr": 8.907278841636502e-06}
{"train_loss": 0.05192571505904198, "global_step": 219173, "epoch": 2462, "lr": 8.906948569871743e-06}
{"train_loss": 0.023361090570688248, "global_step": 219174, "epoch": 2462, "lr": 8.906618303631436e-06}
{"train_loss": 0.0368955135345459, "global_step": 219175, "epoch": 2462, "lr": 8.906288042915645e-06}
{"train_loss": 0.0757739320397377, "global_step": 219176, "epoch": 2462, "lr": 8.905957787724395e-06}
{"train_loss": 0.11057190597057343, "global_step": 219177, "epoch": 2462, "lr": 8.905627538057743e-06}
{"train_loss": 0.03417688235640526, "global_step": 219178, "epoch": 2462, "lr": 8.905297293915721e-06}
{"train_loss": 0.08274605125188828, "global_step": 219179, "epoch": 2462, "lr": 8.90496705529838e-06}
{"train_loss": 0.009960346855223179, "global_step": 219180, "epoch": 2462, "lr": 8.904636822205776e-06}
{"train_loss": 0.04209328442811966, "global_step": 219181, "epoch": 2462, "lr": 8.904306594637929e-06}
{"train_loss": 0.049496129155159, "global_step": 219182, "epoch": 2462, "lr": 8.903976372594914e-06}
{"train_loss": 0.05372818931937218, "global_step": 219183, "epoch": 2462, "lr": 8.903646156076745e-06}
{"train_loss": 0.04514453560113907, "global_step": 219184, "epoch": 2462, "lr": 8.90331594508349e-06}
{"train_loss": 0.060785744339227676, "global_step": 219185, "epoch": 2462, "lr": 8.902985739615171e-06}
{"train_loss": 0.03516203537583351, "global_step": 219186, "epoch": 2462, "lr": 8.90265553967186e-06}
{"train_loss": 0.06485212594270706, "global_step": 219187, "epoch": 2462, "lr": 8.90232534525357e-06}
{"train_loss": 0.041631728410720825, "global_step": 219188, "epoch": 2462, "lr": 8.901995156360376e-06}
{"train_loss": 0.050314102321863174, "global_step": 219189, "epoch": 2462, "lr": 8.901664972992296e-06}
{"train_loss": 0.09175144135951996, "global_step": 219190, "epoch": 2462, "lr": 8.901334795149396e-06}
{"train_loss": 0.06304112821817398, "global_step": 219191, "epoch": 2462, "lr": 8.901004622831704e-06}
{"train_loss": 0.04662345349788666, "global_step": 219192, "epoch": 2462, "lr": 8.900674456039276e-06}
{"train_loss": 0.024876095354557037, "global_step": 219193, "epoch": 2462, "lr": 8.900344294772145e-06}
{"train_loss": 0.029311874881386757, "global_step": 219194, "epoch": 2462, "lr": 8.900014139030371e-06}
{"train_loss": 0.08275413513183594, "global_step": 219195, "epoch": 2462, "lr": 8.899683988813978e-06}
{"train_loss": 0.06232558190822601, "global_step": 219196, "epoch": 2462, "lr": 8.899353844123031e-06}
{"train_loss": 0.026003671810030937, "global_step": 219197, "epoch": 2462, "lr": 8.899023704957555e-06}
{"train_loss": 0.05459189414978027, "global_step": 219198, "epoch": 2462, "lr": 8.898693571317618e-06}
{"train_loss": 0.05100657418370247, "global_step": 219199, "epoch": 2462, "lr": 8.898363443203239e-06}
{"train_loss": 0.0292963907122612, "global_step": 219200, "epoch": 2462, "lr": 8.898033320614485e-06}
{"train_loss": 0.015208651311695576, "global_step": 219201, "epoch": 2462, "lr": 8.897703203551371e-06}
{"train_loss": 0.09492862969636917, "global_step": 219202, "epoch": 2462, "lr": 8.897373092013984e-06}
{"train_loss": 0.019008096307516098, "global_step": 219203, "epoch": 2462, "lr": 8.897042986002336e-06}
{"train_loss": 0.027616193518042564, "global_step": 219204, "epoch": 2462, "lr": 8.896712885516466e-06}
{"train_loss": 0.034319907426834106, "global_step": 219205, "epoch": 2462, "lr": 8.89638279055645e-06}
{"train_loss": 0.05030917361629813, "global_step": 219206, "epoch": 2462, "lr": 8.896052701122298e-06, "val_loss": 9.032869338989258}
{"train_loss": 0.09040310978889465, "global_step": 219207, "epoch": 2463, "lr": 8.895722617214086e-06}
{"train_loss": 0.04046707972884178, "global_step": 219208, "epoch": 2463, "lr": 8.895392538831826e-06}
{"train_loss": 0.052855320274829865, "global_step": 219209, "epoch": 2463, "lr": 8.895062465975596e-06}
{"train_loss": 0.041496265679597855, "global_step": 219210, "epoch": 2463, "lr": 8.894732398645412e-06}
{"train_loss": 0.07549653202295303, "global_step": 219211, "epoch": 2463, "lr": 8.89440233684134e-06}
{"train_loss": 0.06635013222694397, "global_step": 219212, "epoch": 2463, "lr": 8.894072280563404e-06}
{"train_loss": 0.04716794192790985, "global_step": 219213, "epoch": 2463, "lr": 8.89374222981167e-06}
{"train_loss": 0.04247787967324257, "global_step": 219214, "epoch": 2463, "lr": 8.893412184586158e-06}
{"train_loss": 0.03893042728304863, "global_step": 219215, "epoch": 2463, "lr": 8.893082144886927e-06}
{"train_loss": 0.06680340319871902, "global_step": 219216, "epoch": 2463, "lr": 8.892752110714036e-06}
{"train_loss": 0.03535458445549011, "global_step": 219217, "epoch": 2463, "lr": 8.892422082067498e-06}
{"train_loss": 0.06484795361757278, "global_step": 219218, "epoch": 2463, "lr": 8.892092058947382e-06}
{"train_loss": 0.045835476368665695, "global_step": 219219, "epoch": 2463, "lr": 8.891762041353718e-06}
{"train_loss": 0.07479368895292282, "global_step": 219220, "epoch": 2463, "lr": 8.891432029286562e-06}
{"train_loss": 0.04521380737423897, "global_step": 219221, "epoch": 2463, "lr": 8.891102022745939e-06}
{"train_loss": 0.02252211980521679, "global_step": 219222, "epoch": 2463, "lr": 8.890772021731924e-06}
{"train_loss": 0.07703729718923569, "global_step": 219223, "epoch": 2463, "lr": 8.890442026244527e-06}
{"train_loss": 0.037696756422519684, "global_step": 219224, "epoch": 2463, "lr": 8.890112036283821e-06}
{"train_loss": 0.022151969373226166, "global_step": 219225, "epoch": 2463, "lr": 8.889782051849833e-06}
{"train_loss": 0.05251021683216095, "global_step": 219226, "epoch": 2463, "lr": 8.889452072942617e-06}
{"train_loss": 0.047623924911022186, "global_step": 219227, "epoch": 2463, "lr": 8.889122099562202e-06}
{"train_loss": 0.05046441778540611, "global_step": 219228, "epoch": 2463, "lr": 8.888792131708656e-06}
{"train_loss": 0.07634863257408142, "global_step": 219229, "epoch": 2463, "lr": 8.888462169382e-06}
{"train_loss": 0.05417575314640999, "global_step": 219230, "epoch": 2463, "lr": 8.888132212582306e-06}
{"train_loss": 0.06613993644714355, "global_step": 219231, "epoch": 2463, "lr": 8.88780226130958e-06}
{"train_loss": 0.0772424042224884, "global_step": 219232, "epoch": 2463, "lr": 8.887472315563904e-06}
{"train_loss": 0.06660576164722443, "global_step": 219233, "epoch": 2463, "lr": 8.887142375345298e-06}
{"train_loss": 0.038204386830329895, "global_step": 219234, "epoch": 2463, "lr": 8.886812440653814e-06}
{"train_loss": 0.045727554708719254, "global_step": 219235, "epoch": 2463, "lr": 8.886482511489503e-06}
{"train_loss": 0.0654546469449997, "global_step": 219236, "epoch": 2463, "lr": 8.886152587852409e-06}
{"train_loss": 0.08418945968151093, "global_step": 219237, "epoch": 2463, "lr": 8.885822669742555e-06}
{"train_loss": 0.10902341455221176, "global_step": 219238, "epoch": 2463, "lr": 8.885492757160014e-06}
{"train_loss": 0.08282168954610825, "global_step": 219239, "epoch": 2463, "lr": 8.885162850104806e-06}
{"train_loss": 0.03191712498664856, "global_step": 219240, "epoch": 2463, "lr": 8.884832948577e-06}
{"train_loss": 0.046516645699739456, "global_step": 219241, "epoch": 2463, "lr": 8.884503052576625e-06}
{"train_loss": 0.043924733996391296, "global_step": 219242, "epoch": 2463, "lr": 8.884173162103715e-06}
{"train_loss": 0.019794277846813202, "global_step": 219243, "epoch": 2463, "lr": 8.883843277158344e-06}
{"train_loss": 0.026046015322208405, "global_step": 219244, "epoch": 2463, "lr": 8.883513397740522e-06}
{"train_loss": 0.039336927235126495, "global_step": 219245, "epoch": 2463, "lr": 8.883183523850325e-06}
{"train_loss": 0.038988422602415085, "global_step": 219246, "epoch": 2463, "lr": 8.882853655487767e-06}
{"train_loss": 0.08410847187042236, "global_step": 219247, "epoch": 2463, "lr": 8.882523792652925e-06}
{"train_loss": 0.04243944212794304, "global_step": 219248, "epoch": 2463, "lr": 8.882193935345812e-06}
{"train_loss": 0.05178987234830856, "global_step": 219249, "epoch": 2463, "lr": 8.881864083566499e-06}
{"train_loss": 0.04004206880927086, "global_step": 219250, "epoch": 2463, "lr": 8.881534237315004e-06}
{"train_loss": 0.06903531402349472, "global_step": 219251, "epoch": 2463, "lr": 8.881204396591397e-06}
{"train_loss": 0.07468721270561218, "global_step": 219252, "epoch": 2463, "lr": 8.880874561395703e-06}
{"train_loss": 0.03958163037896156, "global_step": 219253, "epoch": 2463, "lr": 8.880544731727975e-06}
{"train_loss": 0.06369998306035995, "global_step": 219254, "epoch": 2463, "lr": 8.880214907588263e-06}
{"train_loss": 0.03229203447699547, "global_step": 219255, "epoch": 2463, "lr": 8.879885088976596e-06}
{"train_loss": 0.05260669067502022, "global_step": 219256, "epoch": 2463, "lr": 8.879555275893043e-06}
{"train_loss": 0.07882711291313171, "global_step": 219257, "epoch": 2463, "lr": 8.879225468337615e-06}
{"train_loss": 0.06660465896129608, "global_step": 219258, "epoch": 2463, "lr": 8.878895666310388e-06}
{"train_loss": 0.06623416393995285, "global_step": 219259, "epoch": 2463, "lr": 8.878565869811384e-06}
{"train_loss": 0.07423053681850433, "global_step": 219260, "epoch": 2463, "lr": 8.878236078840662e-06}
{"train_loss": 0.02734876424074173, "global_step": 219261, "epoch": 2463, "lr": 8.877906293398247e-06}
{"train_loss": 0.0800805613398552, "global_step": 219262, "epoch": 2463, "lr": 8.877576513484214e-06}
{"train_loss": 0.040363989770412445, "global_step": 219263, "epoch": 2463, "lr": 8.877246739098577e-06}
{"train_loss": 0.06492608040571213, "global_step": 219264, "epoch": 2463, "lr": 8.8769169702414e-06}
{"train_loss": 0.0954991951584816, "global_step": 219265, "epoch": 2463, "lr": 8.876587206912712e-06}
{"train_loss": 0.017360324040055275, "global_step": 219266, "epoch": 2463, "lr": 8.87625744911258e-06}
{"train_loss": 0.052140869200229645, "global_step": 219267, "epoch": 2463, "lr": 8.875927696841014e-06}
{"train_loss": 0.07918966561555862, "global_step": 219268, "epoch": 2463, "lr": 8.875597950098097e-06}
{"train_loss": 0.04327407106757164, "global_step": 219269, "epoch": 2463, "lr": 8.875268208883841e-06}
{"train_loss": 0.02709040977060795, "global_step": 219270, "epoch": 2463, "lr": 8.874938473198303e-06}
{"train_loss": 0.03912663832306862, "global_step": 219271, "epoch": 2463, "lr": 8.87460874304154e-06}
{"train_loss": 0.0644022673368454, "global_step": 219272, "epoch": 2463, "lr": 8.874279018413573e-06}
{"train_loss": 0.06093250960111618, "global_step": 219273, "epoch": 2463, "lr": 8.873949299314471e-06}
{"train_loss": 0.09417202323675156, "global_step": 219274, "epoch": 2463, "lr": 8.873619585744253e-06}
{"train_loss": 0.06659842282533646, "global_step": 219275, "epoch": 2463, "lr": 8.873289877702989e-06}
{"train_loss": 0.06952083855867386, "global_step": 219276, "epoch": 2463, "lr": 8.872960175190708e-06}
{"train_loss": 0.04132732376456261, "global_step": 219277, "epoch": 2463, "lr": 8.872630478207444e-06}
{"train_loss": 0.054458338767290115, "global_step": 219278, "epoch": 2463, "lr": 8.872300786753268e-06}
{"train_loss": 0.0451473668217659, "global_step": 219279, "epoch": 2463, "lr": 8.871971100828191e-06}
{"train_loss": 0.03584260120987892, "global_step": 219280, "epoch": 2463, "lr": 8.871641420432297e-06}
{"train_loss": 0.03367387503385544, "global_step": 219281, "epoch": 2463, "lr": 8.871311745565608e-06}
{"train_loss": 0.04665594920516014, "global_step": 219282, "epoch": 2463, "lr": 8.870982076228152e-06}
{"train_loss": 0.025326542556285858, "global_step": 219283, "epoch": 2463, "lr": 8.870652412420005e-06}
{"train_loss": 0.05402672290802002, "global_step": 219284, "epoch": 2463, "lr": 8.87032275414119e-06}
{"train_loss": 0.04284781962633133, "global_step": 219285, "epoch": 2463, "lr": 8.869993101391766e-06}
{"train_loss": 0.09866964817047119, "global_step": 219286, "epoch": 2463, "lr": 8.869663454171761e-06}
{"train_loss": 0.041546858847141266, "global_step": 219287, "epoch": 2463, "lr": 8.86933381248124e-06}
{"train_loss": 0.042282477021217346, "global_step": 219288, "epoch": 2463, "lr": 8.869004176320222e-06}
{"train_loss": 0.07650936394929886, "global_step": 219289, "epoch": 2463, "lr": 8.868674545688765e-06}
{"train_loss": 0.045992400497198105, "global_step": 219290, "epoch": 2463, "lr": 8.86834492058693e-06}
{"train_loss": 0.08918058127164841, "global_step": 219291, "epoch": 2463, "lr": 8.868015301014726e-06}
{"train_loss": 0.04319556802511215, "global_step": 219292, "epoch": 2463, "lr": 8.867685686972232e-06}
{"train_loss": 0.05148479714989662, "global_step": 219293, "epoch": 2463, "lr": 8.867356078459466e-06}
{"train_loss": 0.037097200751304626, "global_step": 219294, "epoch": 2463, "lr": 8.867026475476491e-06}
{"train_loss": 0.054361560622627814, "global_step": 219295, "epoch": 2463, "lr": 8.866696878023333e-06, "val_loss": 8.938393592834473}
{"train_loss": 0.05022833123803139, "global_step": 219296, "epoch": 2464, "lr": 8.866367286100059e-06}
{"train_loss": 0.13022735714912415, "global_step": 219297, "epoch": 2464, "lr": 8.866037699706686e-06}
{"train_loss": 0.06474090367555618, "global_step": 219298, "epoch": 2464, "lr": 8.865708118843285e-06}
{"train_loss": 0.0399661585688591, "global_step": 219299, "epoch": 2464, "lr": 8.865378543509878e-06}
{"train_loss": 0.04779934883117676, "global_step": 219300, "epoch": 2464, "lr": 8.86504897370653e-06}
{"train_loss": 0.08078583329916, "global_step": 219301, "epoch": 2464, "lr": 8.864719409433264e-06}
{"train_loss": 0.04977252706885338, "global_step": 219302, "epoch": 2464, "lr": 8.864389850690147e-06}
{"train_loss": 0.0927966833114624, "global_step": 219303, "epoch": 2464, "lr": 8.8640602974772e-06}
{"train_loss": 0.03815265744924545, "global_step": 219304, "epoch": 2464, "lr": 8.863730749794491e-06}
{"train_loss": 0.04387067258358002, "global_step": 219305, "epoch": 2464, "lr": 8.863401207642036e-06}
{"train_loss": 0.07491065561771393, "global_step": 219306, "epoch": 2464, "lr": 8.863071671019901e-06}
{"train_loss": 0.04422678053379059, "global_step": 219307, "epoch": 2464, "lr": 8.862742139928132e-06}
{"train_loss": 0.053076863288879395, "global_step": 219308, "epoch": 2464, "lr": 8.862412614366755e-06}
{"train_loss": 0.04093511775135994, "global_step": 219309, "epoch": 2464, "lr": 8.862083094335838e-06}
{"train_loss": 0.07615973055362701, "global_step": 219310, "epoch": 2464, "lr": 8.861753579835403e-06}
{"train_loss": 0.05411257594823837, "global_step": 219311, "epoch": 2464, "lr": 8.861424070865509e-06}
{"train_loss": 0.10452329367399216, "global_step": 219312, "epoch": 2464, "lr": 8.861094567426186e-06}
{"train_loss": 0.05149996280670166, "global_step": 219313, "epoch": 2464, "lr": 8.860765069517508e-06}
{"train_loss": 0.057591430842876434, "global_step": 219314, "epoch": 2464, "lr": 8.860435577139487e-06}
{"train_loss": 0.07515173405408859, "global_step": 219315, "epoch": 2464, "lr": 8.860106090292175e-06}
{"train_loss": 0.0721234530210495, "global_step": 219316, "epoch": 2464, "lr": 8.859776608975629e-06}
{"train_loss": 0.05853595957159996, "global_step": 219317, "epoch": 2464, "lr": 8.859447133189868e-06}
{"train_loss": 0.028442315757274628, "global_step": 219318, "epoch": 2464, "lr": 8.859117662934974e-06}
{"train_loss": 0.08800698816776276, "global_step": 219319, "epoch": 2464, "lr": 8.858788198210966e-06}
{"train_loss": 0.03576728329062462, "global_step": 219320, "epoch": 2464, "lr": 8.858458739017878e-06}
{"train_loss": 0.06857211142778397, "global_step": 219321, "epoch": 2464, "lr": 8.858129285355781e-06}
{"train_loss": 0.06924030184745789, "global_step": 219322, "epoch": 2464, "lr": 8.857799837224694e-06}
{"train_loss": 0.030707689002156258, "global_step": 219323, "epoch": 2464, "lr": 8.857470394624696e-06}
{"train_loss": 0.04552360996603966, "global_step": 219324, "epoch": 2464, "lr": 8.857140957555788e-06}
{"train_loss": 0.044414717704057693, "global_step": 219325, "epoch": 2464, "lr": 8.856811526018038e-06}
{"train_loss": 0.07493744045495987, "global_step": 219326, "epoch": 2464, "lr": 8.856482100011504e-06}
{"train_loss": 0.0577513612806797, "global_step": 219327, "epoch": 2464, "lr": 8.856152679536196e-06}
{"train_loss": 0.06453436613082886, "global_step": 219328, "epoch": 2464, "lr": 8.855823264592195e-06}
{"train_loss": 0.04594247415661812, "global_step": 219329, "epoch": 2464, "lr": 8.855493855179508e-06}
{"train_loss": 0.08564218878746033, "global_step": 219330, "epoch": 2464, "lr": 8.85516445129822e-06}
{"train_loss": 0.04652358219027519, "global_step": 219331, "epoch": 2464, "lr": 8.854835052948336e-06}
{"train_loss": 0.09367581456899643, "global_step": 219332, "epoch": 2464, "lr": 8.85450566012993e-06}
{"train_loss": 0.08217638731002808, "global_step": 219333, "epoch": 2464, "lr": 8.854176272843024e-06}
{"train_loss": 0.05754256993532181, "global_step": 219334, "epoch": 2464, "lr": 8.85384689108768e-06}
{"train_loss": 0.07214441895484924, "global_step": 219335, "epoch": 2464, "lr": 8.853517514863924e-06}
{"train_loss": 0.05428309366106987, "global_step": 219336, "epoch": 2464, "lr": 8.853188144171825e-06}
{"train_loss": 0.060913894325494766, "global_step": 219337, "epoch": 2464, "lr": 8.852858779011402e-06}
{"train_loss": 0.03841002285480499, "global_step": 219338, "epoch": 2464, "lr": 8.85252941938272e-06}
{"train_loss": 0.09502101689577103, "global_step": 219339, "epoch": 2464, "lr": 8.85220006528581e-06}
{"train_loss": 0.05678949132561684, "global_step": 219340, "epoch": 2464, "lr": 8.851870716720723e-06}
{"train_loss": 0.062016211450099945, "global_step": 219341, "epoch": 2464, "lr": 8.851541373687494e-06}
{"train_loss": 0.05732090026140213, "global_step": 219342, "epoch": 2464, "lr": 8.851212036186168e-06}
{"train_loss": 0.025749515742063522, "global_step": 219343, "epoch": 2464, "lr": 8.850882704216812e-06}
{"train_loss": 0.09760657697916031, "global_step": 219344, "epoch": 2464, "lr": 8.850553377779437e-06}
{"train_loss": 0.05015268176794052, "global_step": 219345, "epoch": 2464, "lr": 8.850224056874118e-06}
{"train_loss": 0.08229104429483414, "global_step": 219346, "epoch": 2464, "lr": 8.849894741500875e-06}
{"train_loss": 0.029432469978928566, "global_step": 219347, "epoch": 2464, "lr": 8.849565431659774e-06}
{"train_loss": 0.11333596706390381, "global_step": 219348, "epoch": 2464, "lr": 8.84923612735083e-06}
{"train_loss": 0.07270317524671555, "global_step": 219349, "epoch": 2464, "lr": 8.848906828574121e-06}
{"train_loss": 0.025044936686754227, "global_step": 219350, "epoch": 2464, "lr": 8.84857753532966e-06}
{"train_loss": 0.04125913232564926, "global_step": 219351, "epoch": 2464, "lr": 8.848248247617525e-06}
{"train_loss": 0.058513082563877106, "global_step": 219352, "epoch": 2464, "lr": 8.847918965437734e-06}
{"train_loss": 0.03917402774095535, "global_step": 219353, "epoch": 2464, "lr": 8.847589688790331e-06}
{"train_loss": 0.033385682851076126, "global_step": 219354, "epoch": 2464, "lr": 8.84726041767538e-06}
{"train_loss": 0.06106581538915634, "global_step": 219355, "epoch": 2464, "lr": 8.846931152092897e-06}
{"train_loss": 0.09043791890144348, "global_step": 219356, "epoch": 2464, "lr": 8.846601892042955e-06}
{"train_loss": 0.04511244595050812, "global_step": 219357, "epoch": 2464, "lr": 8.846272637525587e-06}
{"train_loss": 0.0651284009218216, "global_step": 219358, "epoch": 2464, "lr": 8.845943388540829e-06}
{"train_loss": 0.05778186023235321, "global_step": 219359, "epoch": 2464, "lr": 8.845614145088738e-06}
{"train_loss": 0.06315336376428604, "global_step": 219360, "epoch": 2464, "lr": 8.845284907169343e-06}
{"train_loss": 0.03370443359017372, "global_step": 219361, "epoch": 2464, "lr": 8.844955674782695e-06}
{"train_loss": 0.029676305130124092, "global_step": 219362, "epoch": 2464, "lr": 8.84462644792886e-06}
{"train_loss": 0.03239058330655098, "global_step": 219363, "epoch": 2464, "lr": 8.844297226607845e-06}
{"train_loss": 0.041344061493873596, "global_step": 219364, "epoch": 2464, "lr": 8.843968010819726e-06}
{"train_loss": 0.038823287934064865, "global_step": 219365, "epoch": 2464, "lr": 8.843638800564525e-06}
{"train_loss": 0.09225714206695557, "global_step": 219366, "epoch": 2464, "lr": 8.843309595842303e-06}
{"train_loss": 0.033459845930337906, "global_step": 219367, "epoch": 2464, "lr": 8.842980396653088e-06}
{"train_loss": 0.10518340766429901, "global_step": 219368, "epoch": 2464, "lr": 8.842651202996943e-06}
{"train_loss": 0.07051616162061691, "global_step": 219369, "epoch": 2464, "lr": 8.842322014873889e-06}
{"train_loss": 0.03419382870197296, "global_step": 219370, "epoch": 2464, "lr": 8.841992832283995e-06}
{"train_loss": 0.057519249618053436, "global_step": 219371, "epoch": 2464, "lr": 8.841663655227284e-06}
{"train_loss": 0.0953376442193985, "global_step": 219372, "epoch": 2464, "lr": 8.841334483703817e-06}
{"train_loss": 0.03177601471543312, "global_step": 219373, "epoch": 2464, "lr": 8.841005317713625e-06}
{"train_loss": 0.027948491275310516, "global_step": 219374, "epoch": 2464, "lr": 8.840676157256767e-06}
{"train_loss": 0.048182278871536255, "global_step": 219375, "epoch": 2464, "lr": 8.840347002333265e-06}
{"train_loss": 0.027141286060214043, "global_step": 219376, "epoch": 2464, "lr": 8.840017852943189e-06}
{"train_loss": 0.05312043055891991, "global_step": 219377, "epoch": 2464, "lr": 8.839688709086562e-06}
{"train_loss": 0.058947253972291946, "global_step": 219378, "epoch": 2464, "lr": 8.839359570763444e-06}
{"train_loss": 0.066816546022892, "global_step": 219379, "epoch": 2464, "lr": 8.839030437973866e-06}
{"train_loss": 0.06899740546941757, "global_step": 219380, "epoch": 2464, "lr": 8.838701310717874e-06}
{"train_loss": 0.06876696646213531, "global_step": 219381, "epoch": 2464, "lr": 8.83837218899553e-06}
{"train_loss": 0.03290407732129097, "global_step": 219382, "epoch": 2464, "lr": 8.838043072806856e-06}
{"train_loss": 0.04514306038618088, "global_step": 219383, "epoch": 2464, "lr": 8.837713962151917e-06}
{"train_loss": 0.05833711609076918, "global_step": 219384, "epoch": 2464, "lr": 8.837384857030734e-06, "val_loss": 8.90861701965332}
{"train_loss": 0.06210440397262573, "global_step": 219385, "epoch": 2465, "lr": 8.837055757443374e-06}
{"train_loss": 0.08649536222219467, "global_step": 219386, "epoch": 2465, "lr": 8.836726663389856e-06}
{"train_loss": 0.08659054338932037, "global_step": 219387, "epoch": 2465, "lr": 8.836397574870254e-06}
{"train_loss": 0.021999265998601913, "global_step": 219388, "epoch": 2465, "lr": 8.836068491884581e-06}
{"train_loss": 0.017714299261569977, "global_step": 219389, "epoch": 2465, "lr": 8.835739414432908e-06}
{"train_loss": 0.08990633487701416, "global_step": 219390, "epoch": 2465, "lr": 8.835410342515261e-06}
{"train_loss": 0.08741015940904617, "global_step": 219391, "epoch": 2465, "lr": 8.835081276131702e-06}
{"train_loss": 0.020432323217391968, "global_step": 219392, "epoch": 2465, "lr": 8.83475221528226e-06}
{"train_loss": 0.036038462072610855, "global_step": 219393, "epoch": 2465, "lr": 8.834423159966976e-06}
{"train_loss": 0.023482853546738625, "global_step": 219394, "epoch": 2465, "lr": 8.834094110185914e-06}
{"train_loss": 0.043064918369054794, "global_step": 219395, "epoch": 2465, "lr": 8.833765065939092e-06}
{"train_loss": 0.0609857402741909, "global_step": 219396, "epoch": 2465, "lr": 8.833436027226583e-06}
{"train_loss": 0.066870778799057, "global_step": 219397, "epoch": 2465, "lr": 8.833106994048407e-06}
{"train_loss": 0.07218477129936218, "global_step": 219398, "epoch": 2465, "lr": 8.832777966404626e-06}
{"train_loss": 0.04474569484591484, "global_step": 219399, "epoch": 2465, "lr": 8.832448944295262e-06}
{"train_loss": 0.06439196318387985, "global_step": 219400, "epoch": 2465, "lr": 8.832119927720394e-06}
{"train_loss": 0.04917686805129051, "global_step": 219401, "epoch": 2465, "lr": 8.831790916680026e-06}
{"train_loss": 0.05419223755598068, "global_step": 219402, "epoch": 2465, "lr": 8.831461911174238e-06}
{"train_loss": 0.03342655673623085, "global_step": 219403, "epoch": 2465, "lr": 8.831132911203044e-06}
{"train_loss": 0.04949464648962021, "global_step": 219404, "epoch": 2465, "lr": 8.830803916766517e-06}
{"train_loss": 0.04943617060780525, "global_step": 219405, "epoch": 2465, "lr": 8.830474927864668e-06}
{"train_loss": 0.042119938880205154, "global_step": 219406, "epoch": 2465, "lr": 8.830145944497581e-06}
{"train_loss": 0.016055084764957428, "global_step": 219407, "epoch": 2465, "lr": 8.82981696666526e-06}
{"train_loss": 0.061082370579242706, "global_step": 219408, "epoch": 2465, "lr": 8.829487994367786e-06}
{"train_loss": 0.038794465363025665, "global_step": 219409, "epoch": 2465, "lr": 8.829159027605166e-06}
{"train_loss": 0.05421429127454758, "global_step": 219410, "epoch": 2465, "lr": 8.82883006637748e-06}
{"train_loss": 0.11451291292905807, "global_step": 219411, "epoch": 2465, "lr": 8.828501110684744e-06}
{"train_loss": 0.0500817745923996, "global_step": 219412, "epoch": 2465, "lr": 8.828172160527026e-06}
{"train_loss": 0.02463238686323166, "global_step": 219413, "epoch": 2465, "lr": 8.827843215904346e-06}
{"train_loss": 0.03910689055919647, "global_step": 219414, "epoch": 2465, "lr": 8.827514276816779e-06}
{"train_loss": 0.04652708023786545, "global_step": 219415, "epoch": 2465, "lr": 8.827185343264332e-06}
{"train_loss": 0.028370292857289314, "global_step": 219416, "epoch": 2465, "lr": 8.82685641524707e-06}
{"train_loss": 0.04633047431707382, "global_step": 219417, "epoch": 2465, "lr": 8.826527492765047e-06}
{"train_loss": 0.05744602531194687, "global_step": 219418, "epoch": 2465, "lr": 8.826198575818285e-06}
{"train_loss": 0.09632853418588638, "global_step": 219419, "epoch": 2465, "lr": 8.825869664406843e-06}
{"train_loss": 0.12557077407836914, "global_step": 219420, "epoch": 2465, "lr": 8.825540758530759e-06}
{"train_loss": 0.0974179208278656, "global_step": 219421, "epoch": 2465, "lr": 8.825211858190085e-06}
{"train_loss": 0.06636463850736618, "global_step": 219422, "epoch": 2465, "lr": 8.82488296338485e-06}
{"train_loss": 0.06891775131225586, "global_step": 219423, "epoch": 2465, "lr": 8.824554074115121e-06}
{"train_loss": 0.06126121059060097, "global_step": 219424, "epoch": 2465, "lr": 8.824225190380914e-06}
{"train_loss": 0.03488590940833092, "global_step": 219425, "epoch": 2465, "lr": 8.8238963121823e-06}
{"train_loss": 0.06326068192720413, "global_step": 219426, "epoch": 2465, "lr": 8.823567439519304e-06}
{"train_loss": 0.07583596557378769, "global_step": 219427, "epoch": 2465, "lr": 8.823238572391983e-06}
{"train_loss": 0.03678947687149048, "global_step": 219428, "epoch": 2465, "lr": 8.82290971080037e-06}
{"train_loss": 0.10129789263010025, "global_step": 219429, "epoch": 2465, "lr": 8.822580854744523e-06}
{"train_loss": 0.07034271210432053, "global_step": 219430, "epoch": 2465, "lr": 8.82225200422448e-06}
{"train_loss": 0.04661386460065842, "global_step": 219431, "epoch": 2465, "lr": 8.82192315924027e-06}
{"train_loss": 0.06219712644815445, "global_step": 219432, "epoch": 2465, "lr": 8.821594319791959e-06}
{"train_loss": 0.07063373178243637, "global_step": 219433, "epoch": 2465, "lr": 8.821265485879577e-06}
{"train_loss": 0.06764861196279526, "global_step": 219434, "epoch": 2465, "lr": 8.820936657503182e-06}
{"train_loss": 0.09568386524915695, "global_step": 219435, "epoch": 2465, "lr": 8.820607834662797e-06}
{"train_loss": 0.059688400477170944, "global_step": 219436, "epoch": 2465, "lr": 8.820279017358495e-06}
{"train_loss": 0.06032756716012955, "global_step": 219437, "epoch": 2465, "lr": 8.819950205590294e-06}
{"train_loss": 0.08871139585971832, "global_step": 219438, "epoch": 2465, "lr": 8.819621399358257e-06}
{"train_loss": 0.0578702837228775, "global_step": 219439, "epoch": 2465, "lr": 8.81929259866241e-06}
{"train_loss": 0.0250434298068285, "global_step": 219440, "epoch": 2465, "lr": 8.818963803502816e-06}
{"train_loss": 0.04368656128644943, "global_step": 219441, "epoch": 2465, "lr": 8.8186350138795e-06}
{"train_loss": 0.07466927915811539, "global_step": 219442, "epoch": 2465, "lr": 8.818306229792534e-06}
{"train_loss": 0.059545304626226425, "global_step": 219443, "epoch": 2465, "lr": 8.817977451241927e-06}
{"train_loss": 0.06075965613126755, "global_step": 219444, "epoch": 2465, "lr": 8.817648678227753e-06}
{"train_loss": 0.030260683968663216, "global_step": 219445, "epoch": 2465, "lr": 8.817319910750033e-06}
{"train_loss": 0.09353344142436981, "global_step": 219446, "epoch": 2465, "lr": 8.816991148808834e-06}
{"train_loss": 0.036233268678188324, "global_step": 219447, "epoch": 2465, "lr": 8.816662392404178e-06}
{"train_loss": 0.03754833713173866, "global_step": 219448, "epoch": 2465, "lr": 8.816333641536135e-06}
{"train_loss": 0.07399808615446091, "global_step": 219449, "epoch": 2465, "lr": 8.816004896204715e-06}
{"train_loss": 0.039274513721466064, "global_step": 219450, "epoch": 2465, "lr": 8.815676156409996e-06}
{"train_loss": 0.041547659784555435, "global_step": 219451, "epoch": 2465, "lr": 8.815347422151999e-06}
{"train_loss": 0.05540988966822624, "global_step": 219452, "epoch": 2465, "lr": 8.815018693430772e-06}
{"train_loss": 0.038617439568042755, "global_step": 219453, "epoch": 2465, "lr": 8.814689970246375e-06}
{"train_loss": 0.017516586929559708, "global_step": 219454, "epoch": 2465, "lr": 8.81436125259883e-06}
{"train_loss": 0.06394654512405396, "global_step": 219455, "epoch": 2465, "lr": 8.814032540488203e-06}
{"train_loss": 0.03692694008350372, "global_step": 219456, "epoch": 2465, "lr": 8.813703833914521e-06}
{"train_loss": 0.030560903251171112, "global_step": 219457, "epoch": 2465, "lr": 8.813375132877844e-06}
{"train_loss": 0.03968390077352524, "global_step": 219458, "epoch": 2465, "lr": 8.813046437378193e-06}
{"train_loss": 0.030088135972619057, "global_step": 219459, "epoch": 2465, "lr": 8.812717747415638e-06}
{"train_loss": 0.04680943116545677, "global_step": 219460, "epoch": 2465, "lr": 8.8123890629902e-06}
{"train_loss": 0.06804969906806946, "global_step": 219461, "epoch": 2465, "lr": 8.812060384101944e-06}
{"train_loss": 0.05409765988588333, "global_step": 219462, "epoch": 2465, "lr": 8.811731710750898e-06}
{"train_loss": 0.053736377507448196, "global_step": 219463, "epoch": 2465, "lr": 8.811403042937116e-06}
{"train_loss": 0.04466943070292473, "global_step": 219464, "epoch": 2465, "lr": 8.811074380660633e-06}
{"train_loss": 0.032127365469932556, "global_step": 219465, "epoch": 2465, "lr": 8.81074572392151e-06}
{"train_loss": 0.06704015284776688, "global_step": 219466, "epoch": 2465, "lr": 8.810417072719768e-06}
{"train_loss": 0.014658062718808651, "global_step": 219467, "epoch": 2465, "lr": 8.81008842705548e-06}
{"train_loss": 0.027492353692650795, "global_step": 219468, "epoch": 2465, "lr": 8.809759786928668e-06}
{"train_loss": 0.03323101997375488, "global_step": 219469, "epoch": 2465, "lr": 8.80943115233937e-06}
{"train_loss": 0.03350386768579483, "global_step": 219470, "epoch": 2465, "lr": 8.809102523287655e-06}
{"train_loss": 0.018411727622151375, "global_step": 219471, "epoch": 2465, "lr": 8.808773899773542e-06}
{"train_loss": 0.06795267760753632, "global_step": 219472, "epoch": 2465, "lr": 8.8084452817971e-06}
{"train_loss": 0.05364424527068151, "global_step": 219473, "epoch": 2465, "lr": 8.808116669358346e-06, "val_loss": 8.927749633789062, "train_action_mse_error": 23.488792419433594}
{"train_loss": 0.05431350693106651, "global_step": 219474, "epoch": 2466, "lr": 8.807788062457351e-06}
{"train_loss": 0.07418564707040787, "global_step": 219475, "epoch": 2466, "lr": 8.807459461094136e-06}
{"train_loss": 0.02747529372572899, "global_step": 219476, "epoch": 2466, "lr": 8.80713086526877e-06}
{"train_loss": 0.02306561917066574, "global_step": 219477, "epoch": 2466, "lr": 8.806802274981268e-06}
{"train_loss": 0.04191123694181442, "global_step": 219478, "epoch": 2466, "lr": 8.806473690231703e-06}
{"train_loss": 0.05668701231479645, "global_step": 219479, "epoch": 2466, "lr": 8.806145111020092e-06}
{"train_loss": 0.05183211341500282, "global_step": 219480, "epoch": 2466, "lr": 8.805816537346512e-06}
{"train_loss": 0.04987870901823044, "global_step": 219481, "epoch": 2466, "lr": 8.805487969210968e-06}
{"train_loss": 0.0468468964099884, "global_step": 219482, "epoch": 2466, "lr": 8.80515940661354e-06}
{"train_loss": 0.03617691993713379, "global_step": 219483, "epoch": 2466, "lr": 8.80483084955424e-06}
{"train_loss": 0.044130366295576096, "global_step": 219484, "epoch": 2466, "lr": 8.80450229803314e-06}
{"train_loss": 0.07540824264287949, "global_step": 219485, "epoch": 2466, "lr": 8.804173752050265e-06}
{"train_loss": 0.0480157807469368, "global_step": 219486, "epoch": 2466, "lr": 8.803845211605676e-06}
{"train_loss": 0.036571647971868515, "global_step": 219487, "epoch": 2466, "lr": 8.803516676699397e-06}
{"train_loss": 0.12893731892108917, "global_step": 219488, "epoch": 2466, "lr": 8.803188147331482e-06}
{"train_loss": 0.01969912275671959, "global_step": 219489, "epoch": 2466, "lr": 8.80285962350199e-06}
{"train_loss": 0.0420965850353241, "global_step": 219490, "epoch": 2466, "lr": 8.802531105210937e-06}
{"train_loss": 0.039710696786642075, "global_step": 219491, "epoch": 2466, "lr": 8.802202592458397e-06}
{"train_loss": 0.05596516653895378, "global_step": 219492, "epoch": 2466, "lr": 8.801874085244383e-06}
{"train_loss": 0.05808170512318611, "global_step": 219493, "epoch": 2466, "lr": 8.80154558356897e-06}
{"train_loss": 0.08207550644874573, "global_step": 219494, "epoch": 2466, "lr": 8.801217087432174e-06}
{"train_loss": 0.05539071187376976, "global_step": 219495, "epoch": 2466, "lr": 8.800888596834061e-06}
{"train_loss": 0.06497044116258621, "global_step": 219496, "epoch": 2466, "lr": 8.800560111774658e-06}
{"train_loss": 0.0849260687828064, "global_step": 219497, "epoch": 2466, "lr": 8.80023163225403e-06}
{"train_loss": 0.06257851421833038, "global_step": 219498, "epoch": 2466, "lr": 8.799903158272194e-06}
{"train_loss": 0.048755351454019547, "global_step": 219499, "epoch": 2466, "lr": 8.799574689829226e-06}
{"train_loss": 0.05092809349298477, "global_step": 219500, "epoch": 2466, "lr": 8.799246226925134e-06}
{"train_loss": 0.08387767523527145, "global_step": 219501, "epoch": 2466, "lr": 8.79891776956e-06}
{"train_loss": 0.048796407878398895, "global_step": 219502, "epoch": 2466, "lr": 8.79858931773383e-06}
{"train_loss": 0.05116351693868637, "global_step": 219503, "epoch": 2466, "lr": 8.798260871446706e-06}
{"train_loss": 0.12249954789876938, "global_step": 219504, "epoch": 2466, "lr": 8.79793243069864e-06}
{"train_loss": 0.0407976433634758, "global_step": 219505, "epoch": 2466, "lr": 8.7976039954897e-06}
{"train_loss": 0.0934167429804802, "global_step": 219506, "epoch": 2466, "lr": 8.797275565819907e-06}
{"train_loss": 0.022574717178940773, "global_step": 219507, "epoch": 2466, "lr": 8.796947141689333e-06}
{"train_loss": 0.03642917424440384, "global_step": 219508, "epoch": 2466, "lr": 8.79661872309801e-06}
{"train_loss": 0.04056679084897041, "global_step": 219509, "epoch": 2466, "lr": 8.796290310045963e-06}
{"train_loss": 0.05432239919900894, "global_step": 219510, "epoch": 2466, "lr": 8.795961902533262e-06}
{"train_loss": 0.12982694804668427, "global_step": 219511, "epoch": 2466, "lr": 8.795633500559935e-06}
{"train_loss": 0.04549342393875122, "global_step": 219512, "epoch": 2466, "lr": 8.795305104126039e-06}
{"train_loss": 0.04554329067468643, "global_step": 219513, "epoch": 2466, "lr": 8.794976713231607e-06}
{"train_loss": 0.06374123692512512, "global_step": 219514, "epoch": 2466, "lr": 8.794648327876698e-06}
{"train_loss": 0.06547779589891434, "global_step": 219515, "epoch": 2466, "lr": 8.79431994806133e-06}
{"train_loss": 0.044691961258649826, "global_step": 219516, "epoch": 2466, "lr": 8.793991573785581e-06}
{"train_loss": 0.060584429651498795, "global_step": 219517, "epoch": 2466, "lr": 8.793663205049462e-06}
{"train_loss": 0.06139931455254555, "global_step": 219518, "epoch": 2466, "lr": 8.793334841853047e-06}
{"train_loss": 0.05553535744547844, "global_step": 219519, "epoch": 2466, "lr": 8.793006484196354e-06}
{"train_loss": 0.05499206483364105, "global_step": 219520, "epoch": 2466, "lr": 8.792678132079451e-06}
{"train_loss": 0.043223731219768524, "global_step": 219521, "epoch": 2466, "lr": 8.792349785502358e-06}
{"train_loss": 0.05721471086144447, "global_step": 219522, "epoch": 2466, "lr": 8.792021444465137e-06}
{"train_loss": 0.04677413031458855, "global_step": 219523, "epoch": 2466, "lr": 8.791693108967824e-06}
{"train_loss": 0.05293595790863037, "global_step": 219524, "epoch": 2466, "lr": 8.791364779010459e-06}
{"train_loss": 0.06379491090774536, "global_step": 219525, "epoch": 2466, "lr": 8.79103645459311e-06}
{"train_loss": 0.06232455372810364, "global_step": 219526, "epoch": 2466, "lr": 8.790708135715787e-06}
{"train_loss": 0.060255154967308044, "global_step": 219527, "epoch": 2466, "lr": 8.790379822378563e-06}
{"train_loss": 0.04724542051553726, "global_step": 219528, "epoch": 2466, "lr": 8.790051514581465e-06}
{"train_loss": 0.0705895945429802, "global_step": 219529, "epoch": 2466, "lr": 8.789723212324547e-06}
{"train_loss": 0.05327724665403366, "global_step": 219530, "epoch": 2466, "lr": 8.78939491560784e-06}
{"train_loss": 0.06268427520990372, "global_step": 219531, "epoch": 2466, "lr": 8.789066624431409e-06}
{"train_loss": 0.045637551695108414, "global_step": 219532, "epoch": 2466, "lr": 8.788738338795271e-06}
{"train_loss": 0.039992015808820724, "global_step": 219533, "epoch": 2466, "lr": 8.788410058699498e-06}
{"train_loss": 0.05763629078865051, "global_step": 219534, "epoch": 2466, "lr": 8.788081784144108e-06}
{"train_loss": 0.07171056419610977, "global_step": 219535, "epoch": 2466, "lr": 8.78775351512917e-06}
{"train_loss": 0.09832040965557098, "global_step": 219536, "epoch": 2466, "lr": 8.787425251654707e-06}
{"train_loss": 0.06443385034799576, "global_step": 219537, "epoch": 2466, "lr": 8.787096993720783e-06}
{"train_loss": 0.043830756098032, "global_step": 219538, "epoch": 2466, "lr": 8.786768741327423e-06}
{"train_loss": 0.0517054907977581, "global_step": 219539, "epoch": 2466, "lr": 8.786440494474684e-06}
{"train_loss": 0.06291928887367249, "global_step": 219540, "epoch": 2466, "lr": 8.786112253162598e-06}
{"train_loss": 0.05562492460012436, "global_step": 219541, "epoch": 2466, "lr": 8.785784017391229e-06}
{"train_loss": 0.04304828494787216, "global_step": 219542, "epoch": 2466, "lr": 8.785455787160596e-06}
{"train_loss": 0.04348093643784523, "global_step": 219543, "epoch": 2466, "lr": 8.785127562470769e-06}
{"train_loss": 0.07982779294252396, "global_step": 219544, "epoch": 2466, "lr": 8.784799343321771e-06}
{"train_loss": 0.10784775018692017, "global_step": 219545, "epoch": 2466, "lr": 8.784471129713661e-06}
{"train_loss": 0.03877422586083412, "global_step": 219546, "epoch": 2466, "lr": 8.784142921646476e-06}
{"train_loss": 0.054136596620082855, "global_step": 219547, "epoch": 2466, "lr": 8.783814719120248e-06}
{"train_loss": 0.056731343269348145, "global_step": 219548, "epoch": 2466, "lr": 8.783486522135054e-06}
{"train_loss": 0.03241259977221489, "global_step": 219549, "epoch": 2466, "lr": 8.7831583306909e-06}
{"train_loss": 0.09042593091726303, "global_step": 219550, "epoch": 2466, "lr": 8.782830144787857e-06}
{"train_loss": 0.05890665948390961, "global_step": 219551, "epoch": 2466, "lr": 8.782501964425954e-06}
{"train_loss": 0.034533414989709854, "global_step": 219552, "epoch": 2466, "lr": 8.782173789605247e-06}
{"train_loss": 0.07031583786010742, "global_step": 219553, "epoch": 2466, "lr": 8.781845620325767e-06}
{"train_loss": 0.06135506182909012, "global_step": 219554, "epoch": 2466, "lr": 8.78151745658758e-06}
{"train_loss": 0.07039965689182281, "global_step": 219555, "epoch": 2466, "lr": 8.781189298390696e-06}
{"train_loss": 0.06129813566803932, "global_step": 219556, "epoch": 2466, "lr": 8.780861145735198e-06}
{"train_loss": 0.09534215927124023, "global_step": 219557, "epoch": 2466, "lr": 8.780532998621093e-06}
{"train_loss": 0.036142729222774506, "global_step": 219558, "epoch": 2466, "lr": 8.780204857048457e-06}
{"train_loss": 0.049940865486860275, "global_step": 219559, "epoch": 2466, "lr": 8.779876721017311e-06}
{"train_loss": 0.03506475314497948, "global_step": 219560, "epoch": 2466, "lr": 8.779548590527704e-06}
{"train_loss": 0.07231815159320831, "global_step": 219561, "epoch": 2466, "lr": 8.779220465579696e-06}
{"train_loss": 0.057305745801396586, "global_step": 219562, "epoch": 2466, "lr": 8.778892346173312e-06, "val_loss": 8.754487991333008}
{"train_loss": 0.04734381288290024, "global_step": 219563, "epoch": 2467, "lr": 8.778564232308612e-06}
{"train_loss": 0.09271819144487381, "global_step": 219564, "epoch": 2467, "lr": 8.778236123985617e-06}
{"train_loss": 0.05734170600771904, "global_step": 219565, "epoch": 2467, "lr": 8.777908021204406e-06}
{"train_loss": 0.07113813608884811, "global_step": 219566, "epoch": 2467, "lr": 8.777579923964985e-06}
{"train_loss": 0.06448664516210556, "global_step": 219567, "epoch": 2467, "lr": 8.77725183226743e-06}
{"train_loss": 0.03201838210225105, "global_step": 219568, "epoch": 2467, "lr": 8.776923746111753e-06}
{"train_loss": 0.026989838108420372, "global_step": 219569, "epoch": 2467, "lr": 8.776595665498038e-06}
{"train_loss": 0.0768405869603157, "global_step": 219570, "epoch": 2467, "lr": 8.77626759042629e-06}
{"train_loss": 0.048897020518779755, "global_step": 219571, "epoch": 2467, "lr": 8.77593952089658e-06}
{"train_loss": 0.08356045931577682, "global_step": 219572, "epoch": 2467, "lr": 8.775611456908933e-06}
{"train_loss": 0.06011494621634483, "global_step": 219573, "epoch": 2467, "lr": 8.775283398463418e-06}
{"train_loss": 0.043382879346609116, "global_step": 219574, "epoch": 2467, "lr": 8.774955345560049e-06}
{"train_loss": 0.036530643701553345, "global_step": 219575, "epoch": 2467, "lr": 8.774627298198896e-06}
{"train_loss": 0.05747852101922035, "global_step": 219576, "epoch": 2467, "lr": 8.77429925637998e-06}
{"train_loss": 0.06655070185661316, "global_step": 219577, "epoch": 2467, "lr": 8.773971220103366e-06}
{"train_loss": 0.07817868143320084, "global_step": 219578, "epoch": 2467, "lr": 8.773643189369085e-06}
{"train_loss": 0.06229551509022713, "global_step": 219579, "epoch": 2467, "lr": 8.773315164177193e-06}
{"train_loss": 0.025736238807439804, "global_step": 219580, "epoch": 2467, "lr": 8.77298714452771e-06}
{"train_loss": 0.020936820656061172, "global_step": 219581, "epoch": 2467, "lr": 8.772659130420713e-06}
{"train_loss": 0.03606406971812248, "global_step": 219582, "epoch": 2467, "lr": 8.77233112185622e-06}
{"train_loss": 0.09153753519058228, "global_step": 219583, "epoch": 2467, "lr": 8.772003118834292e-06}
{"train_loss": 0.04630591720342636, "global_step": 219584, "epoch": 2467, "lr": 8.771675121354967e-06}
{"train_loss": 0.041463688015937805, "global_step": 219585, "epoch": 2467, "lr": 8.771347129418273e-06}
{"train_loss": 0.04814990237355232, "global_step": 219586, "epoch": 2467, "lr": 8.771019143024284e-06}
{"train_loss": 0.05191062018275261, "global_step": 219587, "epoch": 2467, "lr": 8.770691162173017e-06}
{"train_loss": 0.02710958756506443, "global_step": 219588, "epoch": 2467, "lr": 8.770363186864539e-06}
{"train_loss": 0.030105609446763992, "global_step": 219589, "epoch": 2467, "lr": 8.770035217098876e-06}
{"train_loss": 0.020216844975948334, "global_step": 219590, "epoch": 2467, "lr": 8.769707252876087e-06}
{"train_loss": 0.020262496545910835, "global_step": 219591, "epoch": 2467, "lr": 8.769379294196195e-06}
{"train_loss": 0.08746892213821411, "global_step": 219592, "epoch": 2467, "lr": 8.76905134105927e-06}
{"train_loss": 0.04740024358034134, "global_step": 219593, "epoch": 2467, "lr": 8.768723393465333e-06}
{"train_loss": 0.07565135508775711, "global_step": 219594, "epoch": 2467, "lr": 8.768395451414451e-06}
{"train_loss": 0.03815164789557457, "global_step": 219595, "epoch": 2467, "lr": 8.76806751490664e-06}
{"train_loss": 0.10488937050104141, "global_step": 219596, "epoch": 2467, "lr": 8.767739583941963e-06}
{"train_loss": 0.04414179548621178, "global_step": 219597, "epoch": 2467, "lr": 8.767411658520474e-06}
{"train_loss": 0.0969497337937355, "global_step": 219598, "epoch": 2467, "lr": 8.76708373864219e-06}
{"train_loss": 0.03726695105433464, "global_step": 219599, "epoch": 2467, "lr": 8.766755824307183e-06}
{"train_loss": 0.05041665956377983, "global_step": 219600, "epoch": 2467, "lr": 8.766427915515468e-06}
{"train_loss": 0.05800318345427513, "global_step": 219601, "epoch": 2467, "lr": 8.766100012267115e-06}
{"train_loss": 0.03547823801636696, "global_step": 219602, "epoch": 2467, "lr": 8.765772114562148e-06}
{"train_loss": 0.05806506797671318, "global_step": 219603, "epoch": 2467, "lr": 8.765444222400631e-06}
{"train_loss": 0.04231061041355133, "global_step": 219604, "epoch": 2467, "lr": 8.765116335782592e-06}
{"train_loss": 0.043946389108896255, "global_step": 219605, "epoch": 2467, "lr": 8.764788454708084e-06}
{"train_loss": 0.06850789487361908, "global_step": 219606, "epoch": 2467, "lr": 8.764460579177142e-06}
{"train_loss": 0.04269896075129509, "global_step": 219607, "epoch": 2467, "lr": 8.764132709189821e-06}
{"train_loss": 0.05331677198410034, "global_step": 219608, "epoch": 2467, "lr": 8.763804844746155e-06}
{"train_loss": 0.04187241569161415, "global_step": 219609, "epoch": 2467, "lr": 8.763476985846197e-06}
{"train_loss": 0.0687694102525711, "global_step": 219610, "epoch": 2467, "lr": 8.763149132489978e-06}
{"train_loss": 0.04033692926168442, "global_step": 219611, "epoch": 2467, "lr": 8.762821284677564e-06}
{"train_loss": 0.05337440222501755, "global_step": 219612, "epoch": 2467, "lr": 8.762493442408975e-06}
{"train_loss": 0.06985756754875183, "global_step": 219613, "epoch": 2467, "lr": 8.76216560568428e-06}
{"train_loss": 0.060484401881694794, "global_step": 219614, "epoch": 2467, "lr": 8.761837774503496e-06}
{"train_loss": 0.046245645731687546, "global_step": 219615, "epoch": 2467, "lr": 8.761509948866676e-06}
{"train_loss": 0.03375448286533356, "global_step": 219616, "epoch": 2467, "lr": 8.761182128773881e-06}
{"train_loss": 0.04994826018810272, "global_step": 219617, "epoch": 2467, "lr": 8.760854314225136e-06}
{"train_loss": 0.08575321733951569, "global_step": 219618, "epoch": 2467, "lr": 8.760526505220496e-06}
{"train_loss": 0.057350996881723404, "global_step": 219619, "epoch": 2467, "lr": 8.760198701760008e-06}
{"train_loss": 0.09191841632127762, "global_step": 219620, "epoch": 2467, "lr": 8.759870903843693e-06}
{"train_loss": 0.05897046998143196, "global_step": 219621, "epoch": 2467, "lr": 8.759543111471624e-06}
{"train_loss": 0.0554286353290081, "global_step": 219622, "epoch": 2467, "lr": 8.759215324643815e-06}
{"train_loss": 0.0720263421535492, "global_step": 219623, "epoch": 2467, "lr": 8.758887543360344e-06}
{"train_loss": 0.05091087520122528, "global_step": 219624, "epoch": 2467, "lr": 8.758559767621239e-06}
{"train_loss": 0.04072089120745659, "global_step": 219625, "epoch": 2467, "lr": 8.758231997426525e-06}
{"train_loss": 0.0134441452100873, "global_step": 219626, "epoch": 2467, "lr": 8.757904232776281e-06}
{"train_loss": 0.07251659780740738, "global_step": 219627, "epoch": 2467, "lr": 8.757576473670521e-06}
{"train_loss": 0.030402295291423798, "global_step": 219628, "epoch": 2467, "lr": 8.757248720109312e-06}
{"train_loss": 0.05342420935630798, "global_step": 219629, "epoch": 2467, "lr": 8.75692097209268e-06}
{"train_loss": 0.06430576741695404, "global_step": 219630, "epoch": 2467, "lr": 8.756593229620685e-06}
{"train_loss": 0.07800737768411636, "global_step": 219631, "epoch": 2467, "lr": 8.756265492693355e-06}
{"train_loss": 0.05045289918780327, "global_step": 219632, "epoch": 2467, "lr": 8.755937761310756e-06}
{"train_loss": 0.050677813589572906, "global_step": 219633, "epoch": 2467, "lr": 8.7556100354729e-06}
{"train_loss": 0.05439039319753647, "global_step": 219634, "epoch": 2467, "lr": 8.755282315179853e-06}
{"train_loss": 0.04044928774237633, "global_step": 219635, "epoch": 2467, "lr": 8.754954600431669e-06}
{"train_loss": 0.009416282176971436, "global_step": 219636, "epoch": 2467, "lr": 8.754626891228362e-06}
{"train_loss": 0.05204702168703079, "global_step": 219637, "epoch": 2467, "lr": 8.754299187570009e-06}
{"train_loss": 0.07397767156362534, "global_step": 219638, "epoch": 2467, "lr": 8.753971489456625e-06}
{"train_loss": 0.01586742140352726, "global_step": 219639, "epoch": 2467, "lr": 8.753643796888278e-06}
{"train_loss": 0.05696038156747818, "global_step": 219640, "epoch": 2467, "lr": 8.75331610986499e-06}
{"train_loss": 0.06374213844537735, "global_step": 219641, "epoch": 2467, "lr": 8.752988428386826e-06}
{"train_loss": 0.03720448538661003, "global_step": 219642, "epoch": 2467, "lr": 8.752660752453811e-06}
{"train_loss": 0.06124937906861305, "global_step": 219643, "epoch": 2467, "lr": 8.752333082066011e-06}
{"train_loss": 0.041349802166223526, "global_step": 219644, "epoch": 2467, "lr": 8.752005417223441e-06}
{"train_loss": 0.053429294377565384, "global_step": 219645, "epoch": 2467, "lr": 8.751677757926174e-06}
{"train_loss": 0.04880748316645622, "global_step": 219646, "epoch": 2467, "lr": 8.751350104174233e-06}
{"train_loss": 0.02964578941464424, "global_step": 219647, "epoch": 2467, "lr": 8.751022455967678e-06}
{"train_loss": 0.07809621095657349, "global_step": 219648, "epoch": 2467, "lr": 8.75069481330653e-06}
{"train_loss": 0.04552793502807617, "global_step": 219649, "epoch": 2467, "lr": 8.750367176190866e-06}
{"train_loss": 0.05286078527569771, "global_step": 219650, "epoch": 2467, "lr": 8.750039544620704e-06}
{"train_loss": 0.05284328475134092, "global_step": 219651, "epoch": 2467, "lr": 8.749711918596088e-06, "val_loss": 8.954366683959961}
{"train_loss": 0.11044055223464966, "global_step": 219652, "epoch": 2468, "lr": 8.749384298117085e-06}
{"train_loss": 0.02705577202141285, "global_step": 219653, "epoch": 2468, "lr": 8.749056683183716e-06}
{"train_loss": 0.051143188029527664, "global_step": 219654, "epoch": 2468, "lr": 8.748729073796042e-06}
{"train_loss": 0.06322496384382248, "global_step": 219655, "epoch": 2468, "lr": 8.748401469954088e-06}
{"train_loss": 0.04768235608935356, "global_step": 219656, "epoch": 2468, "lr": 8.74807387165792e-06}
{"train_loss": 0.07124143838882446, "global_step": 219657, "epoch": 2468, "lr": 8.74774627890757e-06}
{"train_loss": 0.04691188782453537, "global_step": 219658, "epoch": 2468, "lr": 8.747418691703069e-06}
{"train_loss": 0.06573064625263214, "global_step": 219659, "epoch": 2468, "lr": 8.747091110044491e-06}
{"train_loss": 0.05486425757408142, "global_step": 219660, "epoch": 2468, "lr": 8.746763533931851e-06}
{"train_loss": 0.03275666385889053, "global_step": 219661, "epoch": 2468, "lr": 8.746435963365219e-06}
{"train_loss": 0.05644197016954422, "global_step": 219662, "epoch": 2468, "lr": 8.74610839834462e-06}
{"train_loss": 0.05745798349380493, "global_step": 219663, "epoch": 2468, "lr": 8.7457808388701e-06}
{"train_loss": 0.048388686031103134, "global_step": 219664, "epoch": 2468, "lr": 8.745453284941713e-06}
{"train_loss": 0.037947386503219604, "global_step": 219665, "epoch": 2468, "lr": 8.745125736559484e-06}
{"train_loss": 0.02051917091012001, "global_step": 219666, "epoch": 2468, "lr": 8.744798193723485e-06}
{"train_loss": 0.06762835383415222, "global_step": 219667, "epoch": 2468, "lr": 8.744470656433734e-06}
{"train_loss": 0.06106288731098175, "global_step": 219668, "epoch": 2468, "lr": 8.744143124690296e-06}
{"train_loss": 0.023103192448616028, "global_step": 219669, "epoch": 2468, "lr": 8.743815598493194e-06}
{"train_loss": 0.06739351153373718, "global_step": 219670, "epoch": 2468, "lr": 8.743488077842483e-06}
{"train_loss": 0.09789615124464035, "global_step": 219671, "epoch": 2468, "lr": 8.743160562738217e-06}
{"train_loss": 0.06477787345647812, "global_step": 219672, "epoch": 2468, "lr": 8.742833053180422e-06}
{"train_loss": 0.10680966079235077, "global_step": 219673, "epoch": 2468, "lr": 8.742505549169161e-06}
{"train_loss": 0.07370078563690186, "global_step": 219674, "epoch": 2468, "lr": 8.742178050704453e-06}
{"train_loss": 0.07174044847488403, "global_step": 219675, "epoch": 2468, "lr": 8.741850557786375e-06}
{"train_loss": 0.03147754818201065, "global_step": 219676, "epoch": 2468, "lr": 8.74152307041493e-06}
{"train_loss": 0.032895319163799286, "global_step": 219677, "epoch": 2468, "lr": 8.741195588590207e-06}
{"train_loss": 0.02514861710369587, "global_step": 219678, "epoch": 2468, "lr": 8.740868112312205e-06}
{"train_loss": 0.059422291815280914, "global_step": 219679, "epoch": 2468, "lr": 8.740540641581013e-06}
{"train_loss": 0.02932065725326538, "global_step": 219680, "epoch": 2468, "lr": 8.740213176396633e-06}
{"train_loss": 0.03635864704847336, "global_step": 219681, "epoch": 2468, "lr": 8.739885716759144e-06}
{"train_loss": 0.07542616128921509, "global_step": 219682, "epoch": 2468, "lr": 8.739558262668563e-06}
{"train_loss": 0.07264169305562973, "global_step": 219683, "epoch": 2468, "lr": 8.739230814124955e-06}
{"train_loss": 0.013594675809144974, "global_step": 219684, "epoch": 2468, "lr": 8.738903371128343e-06}
{"train_loss": 0.0684879943728447, "global_step": 219685, "epoch": 2468, "lr": 8.738575933678795e-06}
{"train_loss": 0.06739144027233124, "global_step": 219686, "epoch": 2468, "lr": 8.738248501776324e-06}
{"train_loss": 0.03516209125518799, "global_step": 219687, "epoch": 2468, "lr": 8.737921075421002e-06}
{"train_loss": 0.04436497762799263, "global_step": 219688, "epoch": 2468, "lr": 8.737593654612874e-06}
{"train_loss": 0.060589373111724854, "global_step": 219689, "epoch": 2468, "lr": 8.73726623935196e-06}
{"train_loss": 0.05792777240276337, "global_step": 219690, "epoch": 2468, "lr": 8.736938829638331e-06}
{"train_loss": 0.01856636069715023, "global_step": 219691, "epoch": 2468, "lr": 8.736611425472002e-06}
{"train_loss": 0.06847617030143738, "global_step": 219692, "epoch": 2468, "lr": 8.736284026853048e-06}
{"train_loss": 0.06478744000196457, "global_step": 219693, "epoch": 2468, "lr": 8.735956633781483e-06}
{"train_loss": 0.03673344478011131, "global_step": 219694, "epoch": 2468, "lr": 8.735629246257376e-06}
{"train_loss": 0.07471943646669388, "global_step": 219695, "epoch": 2468, "lr": 8.735301864280764e-06}
{"train_loss": 0.025857727974653244, "global_step": 219696, "epoch": 2468, "lr": 8.734974487851678e-06}
{"train_loss": 0.03523540496826172, "global_step": 219697, "epoch": 2468, "lr": 8.734647116970184e-06}
{"train_loss": 0.039272259920835495, "global_step": 219698, "epoch": 2468, "lr": 8.7343197516363e-06}
{"train_loss": 0.07370155304670334, "global_step": 219699, "epoch": 2468, "lr": 8.733992391850094e-06}
{"train_loss": 0.1375581920146942, "global_step": 219700, "epoch": 2468, "lr": 8.733665037611598e-06}
{"train_loss": 0.04789353534579277, "global_step": 219701, "epoch": 2468, "lr": 8.73333768892085e-06}
{"train_loss": 0.04862269386649132, "global_step": 219702, "epoch": 2468, "lr": 8.733010345777914e-06}
{"train_loss": 0.07127682119607925, "global_step": 219703, "epoch": 2468, "lr": 8.732683008182806e-06}
{"train_loss": 0.03388677164912224, "global_step": 219704, "epoch": 2468, "lr": 8.7323556761356e-06}
{"train_loss": 0.04846487566828728, "global_step": 219705, "epoch": 2468, "lr": 8.732028349636312e-06}
{"train_loss": 0.07547187060117722, "global_step": 219706, "epoch": 2468, "lr": 8.731701028685002e-06}
{"train_loss": 0.027232065796852112, "global_step": 219707, "epoch": 2468, "lr": 8.731373713281727e-06}
{"train_loss": 0.02832881733775139, "global_step": 219708, "epoch": 2468, "lr": 8.731046403426496e-06}
{"train_loss": 0.05085710063576698, "global_step": 219709, "epoch": 2468, "lr": 8.730719099119388e-06}
{"train_loss": 0.07172810286283493, "global_step": 219710, "epoch": 2468, "lr": 8.730391800360421e-06}
{"train_loss": 0.03289259597659111, "global_step": 219711, "epoch": 2468, "lr": 8.730064507149665e-06}
{"train_loss": 0.04490162432193756, "global_step": 219712, "epoch": 2468, "lr": 8.729737219487133e-06}
{"train_loss": 0.06293375045061111, "global_step": 219713, "epoch": 2468, "lr": 8.729409937372895e-06}
{"train_loss": 0.01535588689148426, "global_step": 219714, "epoch": 2468, "lr": 8.729082660806975e-06}
{"train_loss": 0.038390375673770905, "global_step": 219715, "epoch": 2468, "lr": 8.72875538978944e-06}
{"train_loss": 0.05310540273785591, "global_step": 219716, "epoch": 2468, "lr": 8.728428124320309e-06}
{"train_loss": 0.03427645191550255, "global_step": 219717, "epoch": 2468, "lr": 8.728100864399647e-06}
{"train_loss": 0.09303073585033417, "global_step": 219718, "epoch": 2468, "lr": 8.727773610027478e-06}
{"train_loss": 0.049036018550395966, "global_step": 219719, "epoch": 2468, "lr": 8.727446361203868e-06}
{"train_loss": 0.05769777670502663, "global_step": 219720, "epoch": 2468, "lr": 8.72711911792884e-06}
{"train_loss": 0.0717441737651825, "global_step": 219721, "epoch": 2468, "lr": 8.726791880202455e-06}
{"train_loss": 0.0374467596411705, "global_step": 219722, "epoch": 2468, "lr": 8.726464648024745e-06}
{"train_loss": 0.046840280294418335, "global_step": 219723, "epoch": 2468, "lr": 8.726137421395753e-06}
{"train_loss": 0.02861960604786873, "global_step": 219724, "epoch": 2468, "lr": 8.725810200315543e-06}
{"train_loss": 0.026548946276307106, "global_step": 219725, "epoch": 2468, "lr": 8.725482984784134e-06}
{"train_loss": 0.11078537255525589, "global_step": 219726, "epoch": 2468, "lr": 8.72515577480159e-06}
{"train_loss": 0.018230341374874115, "global_step": 219727, "epoch": 2468, "lr": 8.724828570367938e-06}
{"train_loss": 0.03661170229315758, "global_step": 219728, "epoch": 2468, "lr": 8.724501371483235e-06}
{"train_loss": 0.05090620368719101, "global_step": 219729, "epoch": 2468, "lr": 8.724174178147515e-06}
{"train_loss": 0.06417769193649292, "global_step": 219730, "epoch": 2468, "lr": 8.723846990360834e-06}
{"train_loss": 0.07169479876756668, "global_step": 219731, "epoch": 2468, "lr": 8.723519808123216e-06}
{"train_loss": 0.06996139883995056, "global_step": 219732, "epoch": 2468, "lr": 8.723192631434735e-06}
{"train_loss": 0.09238509833812714, "global_step": 219733, "epoch": 2468, "lr": 8.722865460295398e-06}
{"train_loss": 0.04086419939994812, "global_step": 219734, "epoch": 2468, "lr": 8.722538294705286e-06}
{"train_loss": 0.04508987069129944, "global_step": 219735, "epoch": 2468, "lr": 8.722211134664426e-06}
{"train_loss": 0.06209680810570717, "global_step": 219736, "epoch": 2468, "lr": 8.721883980172846e-06}
{"train_loss": 0.04992786794900894, "global_step": 219737, "epoch": 2468, "lr": 8.72155683123062e-06}
{"train_loss": 0.11848883330821991, "global_step": 219738, "epoch": 2468, "lr": 8.721229687837767e-06}
{"train_loss": 0.02437315694987774, "global_step": 219739, "epoch": 2468, "lr": 8.720902549994353e-06}
{"train_loss": 0.053949106986937896, "global_step": 219740, "epoch": 2468, "lr": 8.72057541770041e-06, "val_loss": 8.934879302978516}
{"train_loss": 0.05558890476822853, "global_step": 219741, "epoch": 2469, "lr": 8.720248290955968e-06}
{"train_loss": 0.041948333382606506, "global_step": 219742, "epoch": 2469, "lr": 8.719921169761086e-06}
{"train_loss": 0.15631961822509766, "global_step": 219743, "epoch": 2469, "lr": 8.719594054115821e-06}
{"train_loss": 0.0768127292394638, "global_step": 219744, "epoch": 2469, "lr": 8.719266944020193e-06}
{"train_loss": 0.039937954396009445, "global_step": 219745, "epoch": 2469, "lr": 8.718939839474266e-06}
{"train_loss": 0.043120089918375015, "global_step": 219746, "epoch": 2469, "lr": 8.718612740478065e-06}
{"train_loss": 0.05862784758210182, "global_step": 219747, "epoch": 2469, "lr": 8.718285647031655e-06}
{"train_loss": 0.11531238257884979, "global_step": 219748, "epoch": 2469, "lr": 8.717958559135053e-06}
{"train_loss": 0.051087744534015656, "global_step": 219749, "epoch": 2469, "lr": 8.71763147678833e-06}
{"train_loss": 0.0568571612238884, "global_step": 219750, "epoch": 2469, "lr": 8.71730439999151e-06}
{"train_loss": 0.016886387020349503, "global_step": 219751, "epoch": 2469, "lr": 8.716977328744658e-06}
{"train_loss": 0.04715878143906593, "global_step": 219752, "epoch": 2469, "lr": 8.716650263047788e-06}
{"train_loss": 0.03108537197113037, "global_step": 219753, "epoch": 2469, "lr": 8.716323202900977e-06}
{"train_loss": 0.09030146151781082, "global_step": 219754, "epoch": 2469, "lr": 8.715996148304239e-06}
{"train_loss": 0.05579136312007904, "global_step": 219755, "epoch": 2469, "lr": 8.715669099257645e-06}
{"train_loss": 0.03340485692024231, "global_step": 219756, "epoch": 2469, "lr": 8.715342055761217e-06}
{"train_loss": 0.050926778465509415, "global_step": 219757, "epoch": 2469, "lr": 8.715015017815015e-06}
{"train_loss": 0.06391731649637222, "global_step": 219758, "epoch": 2469, "lr": 8.714687985419068e-06}
{"train_loss": 0.07160905748605728, "global_step": 219759, "epoch": 2469, "lr": 8.714360958573437e-06}
{"train_loss": 0.06917324662208557, "global_step": 219760, "epoch": 2469, "lr": 8.714033937278148e-06}
{"train_loss": 0.06185237318277359, "global_step": 219761, "epoch": 2469, "lr": 8.71370692153325e-06}
{"train_loss": 0.11309579759836197, "global_step": 219762, "epoch": 2469, "lr": 8.713379911338809e-06}
{"train_loss": 0.07934478670358658, "global_step": 219763, "epoch": 2469, "lr": 8.713052906694834e-06}
{"train_loss": 0.03217677026987076, "global_step": 219764, "epoch": 2469, "lr": 8.712725907601399e-06}
{"train_loss": 0.03158773109316826, "global_step": 219765, "epoch": 2469, "lr": 8.712398914058523e-06}
{"train_loss": 0.03547647222876549, "global_step": 219766, "epoch": 2469, "lr": 8.712071926066273e-06}
{"train_loss": 0.11553570628166199, "global_step": 219767, "epoch": 2469, "lr": 8.711744943624672e-06}
{"train_loss": 0.005428472999483347, "global_step": 219768, "epoch": 2469, "lr": 8.71141796673378e-06}
{"train_loss": 0.08145184814929962, "global_step": 219769, "epoch": 2469, "lr": 8.711090995393629e-06}
{"train_loss": 0.07920976728200912, "global_step": 219770, "epoch": 2469, "lr": 8.710764029604279e-06}
{"train_loss": 0.04343632236123085, "global_step": 219771, "epoch": 2469, "lr": 8.71043706936575e-06}
{"train_loss": 0.059444211423397064, "global_step": 219772, "epoch": 2469, "lr": 8.71011011467811e-06}
{"train_loss": 0.02513456717133522, "global_step": 219773, "epoch": 2469, "lr": 8.709783165541396e-06}
{"train_loss": 0.07029003649950027, "global_step": 219774, "epoch": 2469, "lr": 8.709456221955636e-06}
{"train_loss": 0.03169574588537216, "global_step": 219775, "epoch": 2469, "lr": 8.709129283920897e-06}
{"train_loss": 0.039004649966955185, "global_step": 219776, "epoch": 2469, "lr": 8.708802351437201e-06}
{"train_loss": 0.05439523980021477, "global_step": 219777, "epoch": 2469, "lr": 8.708475424504613e-06}
{"train_loss": 0.05023913457989693, "global_step": 219778, "epoch": 2469, "lr": 8.70814850312316e-06}
{"train_loss": 0.04613383486866951, "global_step": 219779, "epoch": 2469, "lr": 8.707821587292902e-06}
{"train_loss": 0.08836974203586578, "global_step": 219780, "epoch": 2469, "lr": 8.707494677013861e-06}
{"train_loss": 0.09647310525178909, "global_step": 219781, "epoch": 2469, "lr": 8.707167772286106e-06}
{"train_loss": 0.06482210010290146, "global_step": 219782, "epoch": 2469, "lr": 8.706840873109662e-06}
{"train_loss": 0.03683082014322281, "global_step": 219783, "epoch": 2469, "lr": 8.706513979484588e-06}
{"train_loss": 0.09312280267477036, "global_step": 219784, "epoch": 2469, "lr": 8.706187091410905e-06}
{"train_loss": 0.04035688191652298, "global_step": 219785, "epoch": 2469, "lr": 8.705860208888689e-06}
{"train_loss": 0.0848795548081398, "global_step": 219786, "epoch": 2469, "lr": 8.705533331917953e-06}
{"train_loss": 0.04319837689399719, "global_step": 219787, "epoch": 2469, "lr": 8.705206460498766e-06}
{"train_loss": 0.035896413028240204, "global_step": 219788, "epoch": 2469, "lr": 8.704879594631149e-06}
{"train_loss": 0.06383170187473297, "global_step": 219789, "epoch": 2469, "lr": 8.70455273431517e-06}
{"train_loss": 0.05423382297158241, "global_step": 219790, "epoch": 2469, "lr": 8.704225879550848e-06}
{"train_loss": 0.037307050079107285, "global_step": 219791, "epoch": 2469, "lr": 8.703899030338252e-06}
{"train_loss": 0.047883812338113785, "global_step": 219792, "epoch": 2469, "lr": 8.703572186677406e-06}
{"train_loss": 0.039322033524513245, "global_step": 219793, "epoch": 2469, "lr": 8.703245348568367e-06}
{"train_loss": 0.037504132837057114, "global_step": 219794, "epoch": 2469, "lr": 8.70291851601116e-06}
{"train_loss": 0.03723343461751938, "global_step": 219795, "epoch": 2469, "lr": 8.702591689005862e-06}
{"train_loss": 0.09251277893781662, "global_step": 219796, "epoch": 2469, "lr": 8.70226486755248e-06}
{"train_loss": 0.051900628954172134, "global_step": 219797, "epoch": 2469, "lr": 8.701938051651076e-06}
{"train_loss": 0.01577414572238922, "global_step": 219798, "epoch": 2469, "lr": 8.701611241301704e-06}
{"train_loss": 0.046708498150110245, "global_step": 219799, "epoch": 2469, "lr": 8.701284436504386e-06}
{"train_loss": 0.022232763469219208, "global_step": 219800, "epoch": 2469, "lr": 8.700957637259188e-06}
{"train_loss": 0.07866330444812775, "global_step": 219801, "epoch": 2469, "lr": 8.700630843566133e-06}
{"train_loss": 0.03926052898168564, "global_step": 219802, "epoch": 2469, "lr": 8.700304055425285e-06}
{"train_loss": 0.03805943578481674, "global_step": 219803, "epoch": 2469, "lr": 8.699977272836663e-06}
{"train_loss": 0.018997712060809135, "global_step": 219804, "epoch": 2469, "lr": 8.699650495800343e-06}
{"train_loss": 0.025333616882562637, "global_step": 219805, "epoch": 2469, "lr": 8.699323724316338e-06}
{"train_loss": 0.09198816865682602, "global_step": 219806, "epoch": 2469, "lr": 8.698996958384715e-06}
{"train_loss": 0.08327464014291763, "global_step": 219807, "epoch": 2469, "lr": 8.698670198005498e-06}
{"train_loss": 0.14534100890159607, "global_step": 219808, "epoch": 2469, "lr": 8.698343443178758e-06}
{"train_loss": 0.05051754415035248, "global_step": 219809, "epoch": 2469, "lr": 8.698016693904504e-06}
{"train_loss": 0.08668926358222961, "global_step": 219810, "epoch": 2469, "lr": 8.697689950182813e-06}
{"train_loss": 0.04789428040385246, "global_step": 219811, "epoch": 2469, "lr": 8.697363212013715e-06}
{"train_loss": 0.06340216845273972, "global_step": 219812, "epoch": 2469, "lr": 8.697036479397241e-06}
{"train_loss": 0.012904015369713306, "global_step": 219813, "epoch": 2469, "lr": 8.696709752333455e-06}
{"train_loss": 0.05823841318488121, "global_step": 219814, "epoch": 2469, "lr": 8.696383030822386e-06}
{"train_loss": 0.06827140599489212, "global_step": 219815, "epoch": 2469, "lr": 8.696056314864092e-06}
{"train_loss": 0.07343003153800964, "global_step": 219816, "epoch": 2469, "lr": 8.695729604458603e-06}
{"train_loss": 0.03624311089515686, "global_step": 219817, "epoch": 2469, "lr": 8.695402899605975e-06}
{"train_loss": 0.045680928975343704, "global_step": 219818, "epoch": 2469, "lr": 8.69507620030624e-06}
{"train_loss": 0.04192382097244263, "global_step": 219819, "epoch": 2469, "lr": 8.694749506559462e-06}
{"train_loss": 0.0369170680642128, "global_step": 219820, "epoch": 2469, "lr": 8.694422818365655e-06}
{"train_loss": 0.05367400869727135, "global_step": 219821, "epoch": 2469, "lr": 8.694096135724895e-06}
{"train_loss": 0.0826820358633995, "global_step": 219822, "epoch": 2469, "lr": 8.693769458637197e-06}
{"train_loss": 0.055354923009872437, "global_step": 219823, "epoch": 2469, "lr": 8.693442787102629e-06}
{"train_loss": 0.09592701494693756, "global_step": 219824, "epoch": 2469, "lr": 8.693116121121214e-06}
{"train_loss": 0.08769971877336502, "global_step": 219825, "epoch": 2469, "lr": 8.692789460693018e-06}
{"train_loss": 0.07143528014421463, "global_step": 219826, "epoch": 2469, "lr": 8.692462805818058e-06}
{"train_loss": 0.037320204079151154, "global_step": 219827, "epoch": 2469, "lr": 8.692136156496406e-06}
{"train_loss": 0.03960971534252167, "global_step": 219828, "epoch": 2469, "lr": 8.691809512728083e-06}
{"train_loss": 0.058064163039867465, "global_step": 219829, "epoch": 2469, "lr": 8.691482874513156e-06, "val_loss": 8.94017505645752}
{"train_loss": 0.03093426115810871, "global_step": 219830, "epoch": 2470, "lr": 8.691156241851639e-06}
{"train_loss": 0.04779301956295967, "global_step": 219831, "epoch": 2470, "lr": 8.690829614743607e-06}
{"train_loss": 0.04433528706431389, "global_step": 219832, "epoch": 2470, "lr": 8.690502993189077e-06}
{"train_loss": 0.08173083513975143, "global_step": 219833, "epoch": 2470, "lr": 8.690176377188108e-06}
{"train_loss": 0.05553489923477173, "global_step": 219834, "epoch": 2470, "lr": 8.689849766740748e-06}
{"train_loss": 0.015062461607158184, "global_step": 219835, "epoch": 2470, "lr": 8.689523161847024e-06}
{"train_loss": 0.06666023284196854, "global_step": 219836, "epoch": 2470, "lr": 8.689196562507007e-06}
{"train_loss": 0.08136419206857681, "global_step": 219837, "epoch": 2470, "lr": 8.688869968720708e-06}
{"train_loss": 0.030379150062799454, "global_step": 219838, "epoch": 2470, "lr": 8.6885433804882e-06}
{"train_loss": 0.032960716634988785, "global_step": 219839, "epoch": 2470, "lr": 8.688216797809501e-06}
{"train_loss": 0.05895271524786949, "global_step": 219840, "epoch": 2470, "lr": 8.68789022068468e-06}
{"train_loss": 0.03877948224544525, "global_step": 219841, "epoch": 2470, "lr": 8.687563649113755e-06}
{"train_loss": 0.03354068845510483, "global_step": 219842, "epoch": 2470, "lr": 8.687237083096793e-06}
{"train_loss": 0.01921660266816616, "global_step": 219843, "epoch": 2470, "lr": 8.686910522633823e-06}
{"train_loss": 0.025813044980168343, "global_step": 219844, "epoch": 2470, "lr": 8.686583967724904e-06}
{"train_loss": 0.08519779145717621, "global_step": 219845, "epoch": 2470, "lr": 8.686257418370058e-06}
{"train_loss": 0.0733659490942955, "global_step": 219846, "epoch": 2470, "lr": 8.685930874569353e-06}
{"train_loss": 0.035184621810913086, "global_step": 219847, "epoch": 2470, "lr": 8.685604336322812e-06}
{"train_loss": 0.042623914778232574, "global_step": 219848, "epoch": 2470, "lr": 8.685277803630493e-06}
{"train_loss": 0.04160083085298538, "global_step": 219849, "epoch": 2470, "lr": 8.684951276492432e-06}
{"train_loss": 0.08594348281621933, "global_step": 219850, "epoch": 2470, "lr": 8.684624754908682e-06}
{"train_loss": 0.021225789561867714, "global_step": 219851, "epoch": 2470, "lr": 8.68429823887928e-06}
{"train_loss": 0.06075320392847061, "global_step": 219852, "epoch": 2470, "lr": 8.683971728404261e-06}
{"train_loss": 0.06712611019611359, "global_step": 219853, "epoch": 2470, "lr": 8.68364522348369e-06}
{"train_loss": 0.030118810012936592, "global_step": 219854, "epoch": 2470, "lr": 8.683318724117584e-06}
{"train_loss": 0.036276254802942276, "global_step": 219855, "epoch": 2470, "lr": 8.68299223030602e-06}
{"train_loss": 0.050767287611961365, "global_step": 219856, "epoch": 2470, "lr": 8.682665742049007e-06}
{"train_loss": 0.06958245486021042, "global_step": 219857, "epoch": 2470, "lr": 8.682339259346616e-06}
{"train_loss": 0.09064622968435287, "global_step": 219858, "epoch": 2470, "lr": 8.682012782198872e-06}
{"train_loss": 0.05396454408764839, "global_step": 219859, "epoch": 2470, "lr": 8.681686310605847e-06}
{"train_loss": 0.051874905824661255, "global_step": 219860, "epoch": 2470, "lr": 8.681359844567543e-06}
{"train_loss": 0.04091918095946312, "global_step": 219861, "epoch": 2470, "lr": 8.681033384084041e-06}
{"train_loss": 0.08446735143661499, "global_step": 219862, "epoch": 2470, "lr": 8.680706929155363e-06}
{"train_loss": 0.1521412432193756, "global_step": 219863, "epoch": 2470, "lr": 8.680380479781568e-06}
{"train_loss": 0.07184799015522003, "global_step": 219864, "epoch": 2470, "lr": 8.680054035962682e-06}
{"train_loss": 0.041021060198545456, "global_step": 219865, "epoch": 2470, "lr": 8.679727597698772e-06}
{"train_loss": 0.04005199298262596, "global_step": 219866, "epoch": 2470, "lr": 8.67940116498986e-06}
{"train_loss": 0.02656787633895874, "global_step": 219867, "epoch": 2470, "lr": 8.679074737836001e-06}
{"train_loss": 0.02866305038332939, "global_step": 219868, "epoch": 2470, "lr": 8.678748316237234e-06}
{"train_loss": 0.05860253795981407, "global_step": 219869, "epoch": 2470, "lr": 8.678421900193601e-06}
{"train_loss": 0.066444031894207, "global_step": 219870, "epoch": 2470, "lr": 8.678095489705162e-06}
{"train_loss": 0.049703653901815414, "global_step": 219871, "epoch": 2470, "lr": 8.677769084771942e-06}
{"train_loss": 0.053264789283275604, "global_step": 219872, "epoch": 2470, "lr": 8.677442685394e-06}
{"train_loss": 0.05309407040476799, "global_step": 219873, "epoch": 2470, "lr": 8.677116291571364e-06}
{"train_loss": 0.09362192451953888, "global_step": 219874, "epoch": 2470, "lr": 8.676789903304095e-06}
{"train_loss": 0.035599641501903534, "global_step": 219875, "epoch": 2470, "lr": 8.676463520592215e-06}
{"train_loss": 0.04172012582421303, "global_step": 219876, "epoch": 2470, "lr": 8.676137143435798e-06}
{"train_loss": 0.06383143365383148, "global_step": 219877, "epoch": 2470, "lr": 8.675810771834853e-06}
{"train_loss": 0.11731214821338654, "global_step": 219878, "epoch": 2470, "lr": 8.675484405789458e-06}
{"train_loss": 0.046368278563022614, "global_step": 219879, "epoch": 2470, "lr": 8.675158045299625e-06}
{"train_loss": 0.05429928004741669, "global_step": 219880, "epoch": 2470, "lr": 8.674831690365425e-06}
{"train_loss": 0.052283428609371185, "global_step": 219881, "epoch": 2470, "lr": 8.674505340986882e-06}
{"train_loss": 0.06375053524971008, "global_step": 219882, "epoch": 2470, "lr": 8.674178997164056e-06}
{"train_loss": 0.05618954077363014, "global_step": 219883, "epoch": 2470, "lr": 8.673852658896974e-06}
{"train_loss": 0.022875849157571793, "global_step": 219884, "epoch": 2470, "lr": 8.673526326185698e-06}
{"train_loss": 0.06490407139062881, "global_step": 219885, "epoch": 2470, "lr": 8.673199999030251e-06}
{"train_loss": 0.03251189738512039, "global_step": 219886, "epoch": 2470, "lr": 8.672873677430704e-06}
{"train_loss": 0.03997866064310074, "global_step": 219887, "epoch": 2470, "lr": 8.672547361387073e-06}
{"train_loss": 0.07255863398313522, "global_step": 219888, "epoch": 2470, "lr": 8.672221050899427e-06}
{"train_loss": 0.07374195754528046, "global_step": 219889, "epoch": 2470, "lr": 8.671894745967791e-06}
{"train_loss": 0.0693945661187172, "global_step": 219890, "epoch": 2470, "lr": 8.67156844659221e-06}
{"train_loss": 0.06587237864732742, "global_step": 219891, "epoch": 2470, "lr": 8.67124215277274e-06}
{"train_loss": 0.06356503069400787, "global_step": 219892, "epoch": 2470, "lr": 8.670915864509405e-06}
{"train_loss": 0.05301149562001228, "global_step": 219893, "epoch": 2470, "lr": 8.670589581802279e-06}
{"train_loss": 0.03477022424340248, "global_step": 219894, "epoch": 2470, "lr": 8.670263304651372e-06}
{"train_loss": 0.042848654091358185, "global_step": 219895, "epoch": 2470, "lr": 8.669937033056758e-06}
{"train_loss": 0.033646926283836365, "global_step": 219896, "epoch": 2470, "lr": 8.669610767018454e-06}
{"train_loss": 0.03772096708416939, "global_step": 219897, "epoch": 2470, "lr": 8.669284506536529e-06}
{"train_loss": 0.05558934062719345, "global_step": 219898, "epoch": 2470, "lr": 8.66895825161101e-06}
{"train_loss": 0.05729285627603531, "global_step": 219899, "epoch": 2470, "lr": 8.66863200224195e-06}
{"train_loss": 0.054083939641714096, "global_step": 219900, "epoch": 2470, "lr": 8.66830575842938e-06}
{"train_loss": 0.046551816165447235, "global_step": 219901, "epoch": 2470, "lr": 8.667979520173358e-06}
{"train_loss": 0.10584324598312378, "global_step": 219902, "epoch": 2470, "lr": 8.66765328747392e-06}
{"train_loss": 0.04517378658056259, "global_step": 219903, "epoch": 2470, "lr": 8.66732706033112e-06}
{"train_loss": 0.03089282102882862, "global_step": 219904, "epoch": 2470, "lr": 8.667000838744982e-06}
{"train_loss": 0.05542916804552078, "global_step": 219905, "epoch": 2470, "lr": 8.66667462271556e-06}
{"train_loss": 0.06221865490078926, "global_step": 219906, "epoch": 2470, "lr": 8.666348412242914e-06}
{"train_loss": 0.04089009389281273, "global_step": 219907, "epoch": 2470, "lr": 8.666022207327062e-06}
{"train_loss": 0.06139856576919556, "global_step": 219908, "epoch": 2470, "lr": 8.665696007968071e-06}
{"train_loss": 0.026658212766051292, "global_step": 219909, "epoch": 2470, "lr": 8.665369814165958e-06}
{"train_loss": 0.049561355262994766, "global_step": 219910, "epoch": 2470, "lr": 8.665043625920799e-06}
{"train_loss": 0.03849659860134125, "global_step": 219911, "epoch": 2470, "lr": 8.664717443232606e-06}
{"train_loss": 0.029731150716543198, "global_step": 219912, "epoch": 2470, "lr": 8.664391266101452e-06}
{"train_loss": 0.07872987538576126, "global_step": 219913, "epoch": 2470, "lr": 8.664065094527357e-06}
{"train_loss": 0.058036528527736664, "global_step": 219914, "epoch": 2470, "lr": 8.663738928510383e-06}
{"train_loss": 0.05604847893118858, "global_step": 219915, "epoch": 2470, "lr": 8.663412768050556e-06}
{"train_loss": 0.10244126617908478, "global_step": 219916, "epoch": 2470, "lr": 8.663086613147937e-06}
{"train_loss": 0.05109599605202675, "global_step": 219917, "epoch": 2470, "lr": 8.662760463802555e-06}
{"train_loss": 0.053863761350093936, "global_step": 219918, "epoch": 2470, "lr": 8.662434320014467e-06, "val_loss": 8.999448776245117, "train_action_mse_error": 18.2919864654541}
{"train_loss": 0.044565264135599136, "global_step": 219919, "epoch": 2471, "lr": 8.662108181783707e-06}
{"train_loss": 0.054920803755521774, "global_step": 219920, "epoch": 2471, "lr": 8.661782049110328e-06}
{"train_loss": 0.060350820422172546, "global_step": 219921, "epoch": 2471, "lr": 8.66145592199436e-06}
{"train_loss": 0.03267582878470421, "global_step": 219922, "epoch": 2471, "lr": 8.66112980043587e-06}
{"train_loss": 0.0730244517326355, "global_step": 219923, "epoch": 2471, "lr": 8.660803684434871e-06}
{"train_loss": 0.08758341521024704, "global_step": 219924, "epoch": 2471, "lr": 8.660477573991433e-06}
{"train_loss": 0.07831741869449615, "global_step": 219925, "epoch": 2471, "lr": 8.660151469105582e-06}
{"train_loss": 0.03126990422606468, "global_step": 219926, "epoch": 2471, "lr": 8.65982536977738e-06}
{"train_loss": 0.03829280287027359, "global_step": 219927, "epoch": 2471, "lr": 8.659499276006861e-06}
{"train_loss": 0.0694938525557518, "global_step": 219928, "epoch": 2471, "lr": 8.659173187794056e-06}
{"train_loss": 0.03612906485795975, "global_step": 219929, "epoch": 2471, "lr": 8.658847105139035e-06}
{"train_loss": 0.05410323292016983, "global_step": 219930, "epoch": 2471, "lr": 8.658521028041816e-06}
{"train_loss": 0.08049701899290085, "global_step": 219931, "epoch": 2471, "lr": 8.658194956502464e-06}
{"train_loss": 0.037433020770549774, "global_step": 219932, "epoch": 2471, "lr": 8.657868890521004e-06}
{"train_loss": 0.06471209973096848, "global_step": 219933, "epoch": 2471, "lr": 8.657542830097504e-06}
{"train_loss": 0.05613621696829796, "global_step": 219934, "epoch": 2471, "lr": 8.657216775231975e-06}
{"train_loss": 0.03044787049293518, "global_step": 219935, "epoch": 2471, "lr": 8.6568907259245e-06}
{"train_loss": 0.038026392459869385, "global_step": 219936, "epoch": 2471, "lr": 8.656564682175083e-06}
{"train_loss": 0.0867542251944542, "global_step": 219937, "epoch": 2471, "lr": 8.656238643983799e-06}
{"train_loss": 0.05730403587222099, "global_step": 219938, "epoch": 2471, "lr": 8.655912611350669e-06}
{"train_loss": 0.03858386352658272, "global_step": 219939, "epoch": 2471, "lr": 8.655586584275765e-06}
{"train_loss": 0.016612417995929718, "global_step": 219940, "epoch": 2471, "lr": 8.655260562759094e-06}
{"train_loss": 0.07545340806245804, "global_step": 219941, "epoch": 2471, "lr": 8.654934546800725e-06}
{"train_loss": 0.04724162444472313, "global_step": 219942, "epoch": 2471, "lr": 8.654608536400704e-06}
{"train_loss": 0.05976761505007744, "global_step": 219943, "epoch": 2471, "lr": 8.654282531559055e-06}
{"train_loss": 0.033467214554548264, "global_step": 219944, "epoch": 2471, "lr": 8.653956532275848e-06}
{"train_loss": 0.03686130791902542, "global_step": 219945, "epoch": 2471, "lr": 8.6536305385511e-06}
{"train_loss": 0.03877638280391693, "global_step": 219946, "epoch": 2471, "lr": 8.653304550384884e-06}
{"train_loss": 0.021291999146342278, "global_step": 219947, "epoch": 2471, "lr": 8.65297856777721e-06}
{"train_loss": 0.07051582634449005, "global_step": 219948, "epoch": 2471, "lr": 8.65265259072815e-06}
{"train_loss": 0.05185478553175926, "global_step": 219949, "epoch": 2471, "lr": 8.652326619237727e-06}
{"train_loss": 0.07620008289813995, "global_step": 219950, "epoch": 2471, "lr": 8.652000653306009e-06}
{"train_loss": 0.06522984057664871, "global_step": 219951, "epoch": 2471, "lr": 8.65167469293301e-06}
{"train_loss": 0.05222299322485924, "global_step": 219952, "epoch": 2471, "lr": 8.651348738118802e-06}
{"train_loss": 0.10724427551031113, "global_step": 219953, "epoch": 2471, "lr": 8.651022788863406e-06}
{"train_loss": 0.03575598821043968, "global_step": 219954, "epoch": 2471, "lr": 8.650696845166884e-06}
{"train_loss": 0.0337996631860733, "global_step": 219955, "epoch": 2471, "lr": 8.650370907029259e-06}
{"train_loss": 0.02599717117846012, "global_step": 219956, "epoch": 2471, "lr": 8.650044974450605e-06}
{"train_loss": 0.10192200541496277, "global_step": 219957, "epoch": 2471, "lr": 8.649719047430938e-06}
{"train_loss": 0.0640292763710022, "global_step": 219958, "epoch": 2471, "lr": 8.649393125970318e-06}
{"train_loss": 0.04312127083539963, "global_step": 219959, "epoch": 2471, "lr": 8.649067210068773e-06}
{"train_loss": 0.031887881457805634, "global_step": 219960, "epoch": 2471, "lr": 8.64874129972636e-06}
{"train_loss": 0.06969084590673447, "global_step": 219961, "epoch": 2471, "lr": 8.648415394943126e-06}
{"train_loss": 0.09930772334337234, "global_step": 219962, "epoch": 2471, "lr": 8.648089495719113e-06}
{"train_loss": 0.03777754306793213, "global_step": 219963, "epoch": 2471, "lr": 8.647763602054349e-06}
{"train_loss": 0.031040215864777565, "global_step": 219964, "epoch": 2471, "lr": 8.647437713948897e-06}
{"train_loss": 0.06791219115257263, "global_step": 219965, "epoch": 2471, "lr": 8.647111831402788e-06}
{"train_loss": 0.045377716422080994, "global_step": 219966, "epoch": 2471, "lr": 8.646785954416075e-06}
{"train_loss": 0.01656223274767399, "global_step": 219967, "epoch": 2471, "lr": 8.6464600829888e-06}
{"train_loss": 0.06152702495455742, "global_step": 219968, "epoch": 2471, "lr": 8.646134217120994e-06}
{"train_loss": 0.09958495944738388, "global_step": 219969, "epoch": 2471, "lr": 8.645808356812724e-06}
{"train_loss": 0.02464430220425129, "global_step": 219970, "epoch": 2471, "lr": 8.645482502064007e-06}
{"train_loss": 0.0326697938144207, "global_step": 219971, "epoch": 2471, "lr": 8.645156652874914e-06}
{"train_loss": 0.05719263106584549, "global_step": 219972, "epoch": 2471, "lr": 8.644830809245463e-06}
{"train_loss": 0.042425282299518585, "global_step": 219973, "epoch": 2471, "lr": 8.644504971175722e-06}
{"train_loss": 0.03639572486281395, "global_step": 219974, "epoch": 2471, "lr": 8.644179138665714e-06}
{"train_loss": 0.056955475360155106, "global_step": 219975, "epoch": 2471, "lr": 8.643853311715499e-06}
{"train_loss": 0.056460171937942505, "global_step": 219976, "epoch": 2471, "lr": 8.643527490325105e-06}
{"train_loss": 0.11622001975774765, "global_step": 219977, "epoch": 2471, "lr": 8.643201674494584e-06}
{"train_loss": 0.043657489120960236, "global_step": 219978, "epoch": 2471, "lr": 8.642875864223992e-06}
{"train_loss": 0.047218408435583115, "global_step": 219979, "epoch": 2471, "lr": 8.642550059513354e-06}
{"train_loss": 0.04411081224679947, "global_step": 219980, "epoch": 2471, "lr": 8.64222426036273e-06}
{"train_loss": 0.08786814659833908, "global_step": 219981, "epoch": 2471, "lr": 8.641898466772141e-06}
{"train_loss": 0.021142680197954178, "global_step": 219982, "epoch": 2471, "lr": 8.641572678741655e-06}
{"train_loss": 0.05646037682890892, "global_step": 219983, "epoch": 2471, "lr": 8.641246896271298e-06}
{"train_loss": 0.030285494402050972, "global_step": 219984, "epoch": 2471, "lr": 8.640921119361133e-06}
{"train_loss": 0.03853553533554077, "global_step": 219985, "epoch": 2471, "lr": 8.640595348011177e-06}
{"train_loss": 0.04293372109532356, "global_step": 219986, "epoch": 2471, "lr": 8.640269582221506e-06}
{"train_loss": 0.05311901494860649, "global_step": 219987, "epoch": 2471, "lr": 8.639943821992125e-06}
{"train_loss": 0.06414683163166046, "global_step": 219988, "epoch": 2471, "lr": 8.63961806732312e-06}
{"train_loss": 0.030363695695996284, "global_step": 219989, "epoch": 2471, "lr": 8.639292318214498e-06}
{"train_loss": 0.0126302819699049, "global_step": 219990, "epoch": 2471, "lr": 8.638966574666335e-06}
{"train_loss": 0.03456537425518036, "global_step": 219991, "epoch": 2471, "lr": 8.638640836678647e-06}
{"train_loss": 0.0269489623606205, "global_step": 219992, "epoch": 2471, "lr": 8.638315104251498e-06}
{"train_loss": 0.05874750390648842, "global_step": 219993, "epoch": 2471, "lr": 8.637989377384909e-06}
{"train_loss": 0.062149982899427414, "global_step": 219994, "epoch": 2471, "lr": 8.637663656078953e-06}
{"train_loss": 0.05498211085796356, "global_step": 219995, "epoch": 2471, "lr": 8.63733794033365e-06}
{"train_loss": 0.042421985417604446, "global_step": 219996, "epoch": 2471, "lr": 8.63701223014905e-06}
{"train_loss": 0.029588988050818443, "global_step": 219997, "epoch": 2471, "lr": 8.636686525525212e-06}
{"train_loss": 0.04507897421717644, "global_step": 219998, "epoch": 2471, "lr": 8.636360826462154e-06}
{"train_loss": 0.026592936366796494, "global_step": 219999, "epoch": 2471, "lr": 8.636035132959947e-06}
{"train_loss": 0.036651190370321274, "global_step": 220000, "epoch": 2471, "lr": 8.635709445018624e-06}
{"train_loss": 0.07025622576475143, "global_step": 220001, "epoch": 2471, "lr": 8.63538376263821e-06}
{"train_loss": 0.041585564613342285, "global_step": 220002, "epoch": 2471, "lr": 8.635058085818776e-06}
{"train_loss": 0.07498256862163544, "global_step": 220003, "epoch": 2471, "lr": 8.634732414560343e-06}
{"train_loss": 0.015292029827833176, "global_step": 220004, "epoch": 2471, "lr": 8.634406748862977e-06}
{"train_loss": 0.04160040244460106, "global_step": 220005, "epoch": 2471, "lr": 8.634081088726714e-06}
{"train_loss": 0.05892390012741089, "global_step": 220006, "epoch": 2471, "lr": 8.633755434151581e-06}
{"train_loss": 0.05099403085919579, "global_step": 220007, "epoch": 2471, "lr": 8.633429785137643e-06, "val_loss": 8.990646362304688}
{"train_loss": 0.05592397227883339, "global_step": 220008, "epoch": 2472, "lr": 8.633104141684929e-06}
{"train_loss": 0.09544035792350769, "global_step": 220009, "epoch": 2472, "lr": 8.632778503793504e-06}
{"train_loss": 0.0722912922501564, "global_step": 220010, "epoch": 2472, "lr": 8.632452871463381e-06}
{"train_loss": 0.060667265206575394, "global_step": 220011, "epoch": 2472, "lr": 8.632127244694632e-06}
{"train_loss": 0.048729799687862396, "global_step": 220012, "epoch": 2472, "lr": 8.631801623487278e-06}
{"train_loss": 0.034827716648578644, "global_step": 220013, "epoch": 2472, "lr": 8.631476007841394e-06}
{"train_loss": 0.05249646306037903, "global_step": 220014, "epoch": 2472, "lr": 8.631150397756981e-06}
{"train_loss": 0.06876876205205917, "global_step": 220015, "epoch": 2472, "lr": 8.63082479323411e-06}
{"train_loss": 0.027585344389081, "global_step": 220016, "epoch": 2472, "lr": 8.630499194272834e-06}
{"train_loss": 0.022920291870832443, "global_step": 220017, "epoch": 2472, "lr": 8.63017360087317e-06}
{"train_loss": 0.04688366875052452, "global_step": 220018, "epoch": 2472, "lr": 8.629848013035186e-06}
{"train_loss": 0.073415108025074, "global_step": 220019, "epoch": 2472, "lr": 8.629522430758901e-06}
{"train_loss": 0.07766258716583252, "global_step": 220020, "epoch": 2472, "lr": 8.62919685404439e-06}
{"train_loss": 0.03702617064118385, "global_step": 220021, "epoch": 2472, "lr": 8.628871282891666e-06}
{"train_loss": 0.06780127435922623, "global_step": 220022, "epoch": 2472, "lr": 8.628545717300795e-06}
{"train_loss": 0.02734501101076603, "global_step": 220023, "epoch": 2472, "lr": 8.628220157271798e-06}
{"train_loss": 0.03727341070771217, "global_step": 220024, "epoch": 2472, "lr": 8.627894602804749e-06}
{"train_loss": 0.03665357083082199, "global_step": 220025, "epoch": 2472, "lr": 8.62756905389966e-06}
{"train_loss": 0.020100319758057594, "global_step": 220026, "epoch": 2472, "lr": 8.627243510556604e-06}
{"train_loss": 0.1151818037033081, "global_step": 220027, "epoch": 2472, "lr": 8.6269179727756e-06}
{"train_loss": 0.0924665629863739, "global_step": 220028, "epoch": 2472, "lr": 8.626592440556713e-06}
{"train_loss": 0.06177142634987831, "global_step": 220029, "epoch": 2472, "lr": 8.62626691389996e-06}
{"train_loss": 0.04475383833050728, "global_step": 220030, "epoch": 2472, "lr": 8.62594139280542e-06}
{"train_loss": 0.03717435523867607, "global_step": 220031, "epoch": 2472, "lr": 8.625615877273103e-06}
{"train_loss": 0.017695486545562744, "global_step": 220032, "epoch": 2472, "lr": 8.625290367303069e-06}
{"train_loss": 0.08774745464324951, "global_step": 220033, "epoch": 2472, "lr": 8.624964862895373e-06}
{"train_loss": 0.07267812639474869, "global_step": 220034, "epoch": 2472, "lr": 8.62463936405003e-06}
{"train_loss": 0.051645658910274506, "global_step": 220035, "epoch": 2472, "lr": 8.624313870767114e-06}
{"train_loss": 0.05535933002829552, "global_step": 220036, "epoch": 2472, "lr": 8.623988383046643e-06}
{"train_loss": 0.045200444757938385, "global_step": 220037, "epoch": 2472, "lr": 8.62366290088869e-06}
{"train_loss": 0.068975068628788, "global_step": 220038, "epoch": 2472, "lr": 8.623337424293276e-06}
{"train_loss": 0.07799733430147171, "global_step": 220039, "epoch": 2472, "lr": 8.623011953260435e-06}
{"train_loss": 0.04650377109646797, "global_step": 220040, "epoch": 2472, "lr": 8.622686487790243e-06}
{"train_loss": 0.05687524378299713, "global_step": 220041, "epoch": 2472, "lr": 8.622361027882714e-06}
{"train_loss": 0.10862979292869568, "global_step": 220042, "epoch": 2472, "lr": 8.622035573537912e-06}
{"train_loss": 0.04062611609697342, "global_step": 220043, "epoch": 2472, "lr": 8.621710124755878e-06}
{"train_loss": 0.06207123398780823, "global_step": 220044, "epoch": 2472, "lr": 8.621384681536638e-06}
{"train_loss": 0.07807108014822006, "global_step": 220045, "epoch": 2472, "lr": 8.621059243880253e-06}
{"train_loss": 0.031063079833984375, "global_step": 220046, "epoch": 2472, "lr": 8.620733811786758e-06}
{"train_loss": 0.05038430914282799, "global_step": 220047, "epoch": 2472, "lr": 8.620408385256212e-06}
{"train_loss": 0.05799895152449608, "global_step": 220048, "epoch": 2472, "lr": 8.620082964288634e-06}
{"train_loss": 0.06518959999084473, "global_step": 220049, "epoch": 2472, "lr": 8.619757548884095e-06}
{"train_loss": 0.031532496213912964, "global_step": 220050, "epoch": 2472, "lr": 8.61943213904261e-06}
{"train_loss": 0.07471468299627304, "global_step": 220051, "epoch": 2472, "lr": 8.619106734764243e-06}
{"train_loss": 0.06274040043354034, "global_step": 220052, "epoch": 2472, "lr": 8.618781336049042e-06}
{"train_loss": 0.04880869388580322, "global_step": 220053, "epoch": 2472, "lr": 8.61845594289703e-06}
{"train_loss": 0.02694038115441799, "global_step": 220054, "epoch": 2472, "lr": 8.618130555308273e-06}
{"train_loss": 0.08392158895730972, "global_step": 220055, "epoch": 2472, "lr": 8.617805173282795e-06}
{"train_loss": 0.032511286437511444, "global_step": 220056, "epoch": 2472, "lr": 8.617479796820654e-06}
{"train_loss": 0.036196861416101456, "global_step": 220057, "epoch": 2472, "lr": 8.617154425921887e-06}
{"train_loss": 0.10343455523252487, "global_step": 220058, "epoch": 2472, "lr": 8.616829060586545e-06}
{"train_loss": 0.03265100717544556, "global_step": 220059, "epoch": 2472, "lr": 8.616503700814655e-06}
{"train_loss": 0.05806949734687805, "global_step": 220060, "epoch": 2472, "lr": 8.616178346606279e-06}
{"train_loss": 0.0772884413599968, "global_step": 220061, "epoch": 2472, "lr": 8.615852997961444e-06}
{"train_loss": 0.04506438970565796, "global_step": 220062, "epoch": 2472, "lr": 8.615527654880218e-06}
{"train_loss": 0.05455368384718895, "global_step": 220063, "epoch": 2472, "lr": 8.615202317362619e-06}
{"train_loss": 0.03325232118368149, "global_step": 220064, "epoch": 2472, "lr": 8.61487698540871e-06}
{"train_loss": 0.05754788592457771, "global_step": 220065, "epoch": 2472, "lr": 8.614551659018516e-06}
{"train_loss": 0.06596756726503372, "global_step": 220066, "epoch": 2472, "lr": 8.6142263381921e-06}
{"train_loss": 0.04006282985210419, "global_step": 220067, "epoch": 2472, "lr": 8.613901022929489e-06}
{"train_loss": 0.0746232196688652, "global_step": 220068, "epoch": 2472, "lr": 8.613575713230732e-06}
{"train_loss": 0.06957890838384628, "global_step": 220069, "epoch": 2472, "lr": 8.613250409095891e-06}
{"train_loss": 0.08736348152160645, "global_step": 220070, "epoch": 2472, "lr": 8.612925110524977e-06}
{"train_loss": 0.08753284811973572, "global_step": 220071, "epoch": 2472, "lr": 8.612599817518064e-06}
{"train_loss": 0.05005417391657829, "global_step": 220072, "epoch": 2472, "lr": 8.612274530075171e-06}
{"train_loss": 0.042355384677648544, "global_step": 220073, "epoch": 2472, "lr": 8.611949248196366e-06}
{"train_loss": 0.052013035863637924, "global_step": 220074, "epoch": 2472, "lr": 8.611623971881672e-06}
{"train_loss": 0.05910446494817734, "global_step": 220075, "epoch": 2472, "lr": 8.611298701131149e-06}
{"train_loss": 0.08325673639774323, "global_step": 220076, "epoch": 2472, "lr": 8.61097343594482e-06}
{"train_loss": 0.06487978994846344, "global_step": 220077, "epoch": 2472, "lr": 8.61064817632275e-06}
{"train_loss": 0.07292230427265167, "global_step": 220078, "epoch": 2472, "lr": 8.61032292226498e-06}
{"train_loss": 0.017975738272070885, "global_step": 220079, "epoch": 2472, "lr": 8.60999767377153e-06}
{"train_loss": 0.045652419328689575, "global_step": 220080, "epoch": 2472, "lr": 8.609672430842475e-06}
{"train_loss": 0.06475531309843063, "global_step": 220081, "epoch": 2472, "lr": 8.609347193477835e-06}
{"train_loss": 0.06461215019226074, "global_step": 220082, "epoch": 2472, "lr": 8.609021961677677e-06}
{"train_loss": 0.06173194572329521, "global_step": 220083, "epoch": 2472, "lr": 8.608696735442028e-06}
{"train_loss": 0.10314562171697617, "global_step": 220084, "epoch": 2472, "lr": 8.608371514770924e-06}
{"train_loss": 0.06672859191894531, "global_step": 220085, "epoch": 2472, "lr": 8.608046299664435e-06}
{"train_loss": 0.05311329662799835, "global_step": 220086, "epoch": 2472, "lr": 8.607721090122572e-06}
{"train_loss": 0.05320444703102112, "global_step": 220087, "epoch": 2472, "lr": 8.607395886145403e-06}
{"train_loss": 0.030818728730082512, "global_step": 220088, "epoch": 2472, "lr": 8.607070687732971e-06}
{"train_loss": 0.05929454043507576, "global_step": 220089, "epoch": 2472, "lr": 8.606745494885305e-06}
{"train_loss": 0.05309867858886719, "global_step": 220090, "epoch": 2472, "lr": 8.60642030760247e-06}
{"train_loss": 0.07777228951454163, "global_step": 220091, "epoch": 2472, "lr": 8.606095125884483e-06}
{"train_loss": 0.05286455899477005, "global_step": 220092, "epoch": 2472, "lr": 8.605769949731418e-06}
{"train_loss": 0.08930446207523346, "global_step": 220093, "epoch": 2472, "lr": 8.605444779143285e-06}
{"train_loss": 0.015316144563257694, "global_step": 220094, "epoch": 2472, "lr": 8.605119614120161e-06}
{"train_loss": 0.013586167246103287, "global_step": 220095, "epoch": 2472, "lr": 8.604794454662063e-06}
{"train_loss": 0.0569145389508163, "global_step": 220096, "epoch": 2472, "lr": 8.604469300769057e-06, "val_loss": 8.942183494567871}
{"train_loss": 0.05004597455263138, "global_step": 220097, "epoch": 2473, "lr": 8.604144152441163e-06}
{"train_loss": 0.0184654388576746, "global_step": 220098, "epoch": 2473, "lr": 8.603819009678448e-06}
{"train_loss": 0.07991740107536316, "global_step": 220099, "epoch": 2473, "lr": 8.603493872480933e-06}
{"train_loss": 0.06887791305780411, "global_step": 220100, "epoch": 2473, "lr": 8.603168740848689e-06}
{"train_loss": 0.037606872618198395, "global_step": 220101, "epoch": 2473, "lr": 8.60284361478173e-06}
{"train_loss": 0.0751940906047821, "global_step": 220102, "epoch": 2473, "lr": 8.602518494280125e-06}
{"train_loss": 0.0665956437587738, "global_step": 220103, "epoch": 2473, "lr": 8.602193379343898e-06}
{"train_loss": 0.07781665772199631, "global_step": 220104, "epoch": 2473, "lr": 8.6018682699731e-06}
{"train_loss": 0.035843200981616974, "global_step": 220105, "epoch": 2473, "lr": 8.601543166167786e-06}
{"train_loss": 0.05436643585562706, "global_step": 220106, "epoch": 2473, "lr": 8.601218067927979e-06}
{"train_loss": 0.040075864642858505, "global_step": 220107, "epoch": 2473, "lr": 8.60089297525375e-06}
{"train_loss": 0.027975112199783325, "global_step": 220108, "epoch": 2473, "lr": 8.600567888145111e-06}
{"train_loss": 0.04317548871040344, "global_step": 220109, "epoch": 2473, "lr": 8.600242806602133e-06}
{"train_loss": 0.07159676402807236, "global_step": 220110, "epoch": 2473, "lr": 8.599917730624835e-06}
{"train_loss": 0.05141706392168999, "global_step": 220111, "epoch": 2473, "lr": 8.599592660213284e-06}
{"train_loss": 0.06989467144012451, "global_step": 220112, "epoch": 2473, "lr": 8.599267595367506e-06}
{"train_loss": 0.06148891523480415, "global_step": 220113, "epoch": 2473, "lr": 8.598942536087562e-06}
{"train_loss": 0.06116628274321556, "global_step": 220114, "epoch": 2473, "lr": 8.598617482373472e-06}
{"train_loss": 0.050179801881313324, "global_step": 220115, "epoch": 2473, "lr": 8.598292434225308e-06}
{"train_loss": 0.04239987954497337, "global_step": 220116, "epoch": 2473, "lr": 8.597967391643098e-06}
{"train_loss": 0.06917158514261246, "global_step": 220117, "epoch": 2473, "lr": 8.597642354626873e-06}
{"train_loss": 0.0884617269039154, "global_step": 220118, "epoch": 2473, "lr": 8.597317323176701e-06}
{"train_loss": 0.04478618875145912, "global_step": 220119, "epoch": 2473, "lr": 8.596992297292605e-06}
{"train_loss": 0.05300499498844147, "global_step": 220120, "epoch": 2473, "lr": 8.59666727697465e-06}
{"train_loss": 0.11478158831596375, "global_step": 220121, "epoch": 2473, "lr": 8.596342262222873e-06}
{"train_loss": 0.05283308029174805, "global_step": 220122, "epoch": 2473, "lr": 8.596017253037297e-06}
{"train_loss": 0.03310830518603325, "global_step": 220123, "epoch": 2473, "lr": 8.59569224941798e-06}
{"train_loss": 0.028839845210313797, "global_step": 220124, "epoch": 2473, "lr": 8.595367251364983e-06}
{"train_loss": 0.03049784153699875, "global_step": 220125, "epoch": 2473, "lr": 8.595042258878322e-06}
{"train_loss": 0.04586421325802803, "global_step": 220126, "epoch": 2473, "lr": 8.594717271958063e-06}
{"train_loss": 0.051042746752500534, "global_step": 220127, "epoch": 2473, "lr": 8.594392290604231e-06}
{"train_loss": 0.05060246214270592, "global_step": 220128, "epoch": 2473, "lr": 8.594067314816889e-06}
{"train_loss": 0.02654554881155491, "global_step": 220129, "epoch": 2473, "lr": 8.593742344596056e-06}
{"train_loss": 0.04723693057894707, "global_step": 220130, "epoch": 2473, "lr": 8.593417379941803e-06}
{"train_loss": 0.030467813834547997, "global_step": 220131, "epoch": 2473, "lr": 8.593092420854148e-06}
{"train_loss": 0.03203628584742546, "global_step": 220132, "epoch": 2473, "lr": 8.592767467333157e-06}
{"train_loss": 0.10045348852872849, "global_step": 220133, "epoch": 2473, "lr": 8.592442519378857e-06}
{"train_loss": 0.04683315381407738, "global_step": 220134, "epoch": 2473, "lr": 8.592117576991304e-06}
{"train_loss": 0.03984133526682854, "global_step": 220135, "epoch": 2473, "lr": 8.591792640170526e-06}
{"train_loss": 0.036631956696510315, "global_step": 220136, "epoch": 2473, "lr": 8.591467708916596e-06}
{"train_loss": 0.026249846443533897, "global_step": 220137, "epoch": 2473, "lr": 8.591142783229516e-06}
{"train_loss": 0.08372015506029129, "global_step": 220138, "epoch": 2473, "lr": 8.590817863109369e-06}
{"train_loss": 0.07905901223421097, "global_step": 220139, "epoch": 2473, "lr": 8.590492948556173e-06}
{"train_loss": 0.03933192789554596, "global_step": 220140, "epoch": 2473, "lr": 8.590168039569985e-06}
{"train_loss": 0.030296968296170235, "global_step": 220141, "epoch": 2473, "lr": 8.589843136150837e-06}
{"train_loss": 0.0683259442448616, "global_step": 220142, "epoch": 2473, "lr": 8.589518238298783e-06}
{"train_loss": 0.0677499920129776, "global_step": 220143, "epoch": 2473, "lr": 8.589193346013869e-06}
{"train_loss": 0.026359891518950462, "global_step": 220144, "epoch": 2473, "lr": 8.588868459296124e-06}
{"train_loss": 0.09240642935037613, "global_step": 220145, "epoch": 2473, "lr": 8.588543578145613e-06}
{"train_loss": 0.061812032014131546, "global_step": 220146, "epoch": 2473, "lr": 8.588218702562357e-06}
{"train_loss": 0.0379524901509285, "global_step": 220147, "epoch": 2473, "lr": 8.587893832546418e-06}
{"train_loss": 0.05737408995628357, "global_step": 220148, "epoch": 2473, "lr": 8.587568968097826e-06}
{"train_loss": 0.08274146169424057, "global_step": 220149, "epoch": 2473, "lr": 8.587244109216635e-06}
{"train_loss": 0.04758583754301071, "global_step": 220150, "epoch": 2473, "lr": 8.586919255902876e-06}
{"train_loss": 0.045353665947914124, "global_step": 220151, "epoch": 2473, "lr": 8.586594408156617e-06}
{"train_loss": 0.0322144515812397, "global_step": 220152, "epoch": 2473, "lr": 8.586269565977867e-06}
{"train_loss": 0.0328647755086422, "global_step": 220153, "epoch": 2473, "lr": 8.585944729366707e-06}
{"train_loss": 0.04086097702383995, "global_step": 220154, "epoch": 2473, "lr": 8.585619898323155e-06}
{"train_loss": 0.08840632438659668, "global_step": 220155, "epoch": 2473, "lr": 8.585295072847255e-06}
{"train_loss": 0.02955409698188305, "global_step": 220156, "epoch": 2473, "lr": 8.584970252939067e-06}
{"train_loss": 0.07919491082429886, "global_step": 220157, "epoch": 2473, "lr": 8.584645438598616e-06}
{"train_loss": 0.0641857460141182, "global_step": 220158, "epoch": 2473, "lr": 8.584320629825959e-06}
{"train_loss": 0.08306149393320084, "global_step": 220159, "epoch": 2473, "lr": 8.583995826621128e-06}
{"train_loss": 0.044436872005462646, "global_step": 220160, "epoch": 2473, "lr": 8.583671028984186e-06}
{"train_loss": 0.03763910010457039, "global_step": 220161, "epoch": 2473, "lr": 8.583346236915158e-06}
{"train_loss": 0.028254453092813492, "global_step": 220162, "epoch": 2473, "lr": 8.583021450414098e-06}
{"train_loss": 0.044296570122241974, "global_step": 220163, "epoch": 2473, "lr": 8.582696669481039e-06}
{"train_loss": 0.05139552056789398, "global_step": 220164, "epoch": 2473, "lr": 8.582371894116043e-06}
{"train_loss": 0.045110344886779785, "global_step": 220165, "epoch": 2473, "lr": 8.582047124319126e-06}
{"train_loss": 0.03553280979394913, "global_step": 220166, "epoch": 2473, "lr": 8.58172236009036e-06}
{"train_loss": 0.06669791042804718, "global_step": 220167, "epoch": 2473, "lr": 8.58139760142977e-06}
{"train_loss": 0.031666506081819534, "global_step": 220168, "epoch": 2473, "lr": 8.581072848337419e-06}
{"train_loss": 0.02631259709596634, "global_step": 220169, "epoch": 2473, "lr": 8.580748100813318e-06}
{"train_loss": 0.08882161229848862, "global_step": 220170, "epoch": 2473, "lr": 8.580423358857547e-06}
{"train_loss": 0.04984724521636963, "global_step": 220171, "epoch": 2473, "lr": 8.580098622470122e-06}
{"train_loss": 0.04765810817480087, "global_step": 220172, "epoch": 2473, "lr": 8.57977389165111e-06}
{"train_loss": 0.07801560312509537, "global_step": 220173, "epoch": 2473, "lr": 8.579449166400532e-06}
{"train_loss": 0.034303247928619385, "global_step": 220174, "epoch": 2473, "lr": 8.57912444671845e-06}
{"train_loss": 0.0976550504565239, "global_step": 220175, "epoch": 2473, "lr": 8.578799732604892e-06}
{"train_loss": 0.05213351920247078, "global_step": 220176, "epoch": 2473, "lr": 8.578475024059918e-06}
{"train_loss": 0.05382209271192551, "global_step": 220177, "epoch": 2473, "lr": 8.578150321083556e-06}
{"train_loss": 0.06220097467303276, "global_step": 220178, "epoch": 2473, "lr": 8.57782562367585e-06}
{"train_loss": 0.08759912103414536, "global_step": 220179, "epoch": 2473, "lr": 8.577500931836868e-06}
{"train_loss": 0.029010454192757607, "global_step": 220180, "epoch": 2473, "lr": 8.577176245566621e-06}
{"train_loss": 0.0600724034011364, "global_step": 220181, "epoch": 2473, "lr": 8.57685156486518e-06}
{"train_loss": 0.08849819749593735, "global_step": 220182, "epoch": 2473, "lr": 8.576526889732567e-06}
{"train_loss": 0.03352304920554161, "global_step": 220183, "epoch": 2473, "lr": 8.576202220168843e-06}
{"train_loss": 0.028001142665743828, "global_step": 220184, "epoch": 2473, "lr": 8.57587755617404e-06}
{"train_loss": 0.05353955872189463, "global_step": 220185, "epoch": 2473, "lr": 8.575552897748206e-06, "val_loss": 8.949858665466309}
{"train_loss": 0.02678832970559597, "global_step": 220186, "epoch": 2474, "lr": 8.575228244891381e-06}
{"train_loss": 0.11355523765087128, "global_step": 220187, "epoch": 2474, "lr": 8.574903597603618e-06}
{"train_loss": 0.05347391963005066, "global_step": 220188, "epoch": 2474, "lr": 8.574578955884944e-06}
{"train_loss": 0.02063179761171341, "global_step": 220189, "epoch": 2474, "lr": 8.574254319735425e-06}
{"train_loss": 0.07641411572694778, "global_step": 220190, "epoch": 2474, "lr": 8.573929689155086e-06}
{"train_loss": 0.05381966754794121, "global_step": 220191, "epoch": 2474, "lr": 8.57360506414398e-06}
{"train_loss": 0.04279880225658417, "global_step": 220192, "epoch": 2474, "lr": 8.573280444702142e-06}
{"train_loss": 0.036343399435281754, "global_step": 220193, "epoch": 2474, "lr": 8.572955830829631e-06}
{"train_loss": 0.030262725427746773, "global_step": 220194, "epoch": 2474, "lr": 8.572631222526483e-06}
{"train_loss": 0.06520511209964752, "global_step": 220195, "epoch": 2474, "lr": 8.572306619792724e-06}
{"train_loss": 0.029850749298930168, "global_step": 220196, "epoch": 2474, "lr": 8.571982022628434e-06}
{"train_loss": 0.03977476432919502, "global_step": 220197, "epoch": 2474, "lr": 8.571657431033614e-06}
{"train_loss": 0.09307871758937836, "global_step": 220198, "epoch": 2474, "lr": 8.571332845008345e-06}
{"train_loss": 0.05813965946435928, "global_step": 220199, "epoch": 2474, "lr": 8.571008264552648e-06}
{"train_loss": 0.036328624933958054, "global_step": 220200, "epoch": 2474, "lr": 8.570683689666581e-06}
{"train_loss": 0.05629308521747589, "global_step": 220201, "epoch": 2474, "lr": 8.57035912035017e-06}
{"train_loss": 0.05662728473544121, "global_step": 220202, "epoch": 2474, "lr": 8.570034556603478e-06}
{"train_loss": 0.03960777074098587, "global_step": 220203, "epoch": 2474, "lr": 8.569709998426534e-06}
{"train_loss": 0.07550852000713348, "global_step": 220204, "epoch": 2474, "lr": 8.569385445819395e-06}
{"train_loss": 0.0671846866607666, "global_step": 220205, "epoch": 2474, "lr": 8.56906089878209e-06}
{"train_loss": 0.04194073751568794, "global_step": 220206, "epoch": 2474, "lr": 8.568736357314677e-06}
{"train_loss": 0.062649667263031, "global_step": 220207, "epoch": 2474, "lr": 8.568411821417183e-06}
{"train_loss": 0.04114799201488495, "global_step": 220208, "epoch": 2474, "lr": 8.568087291089672e-06}
{"train_loss": 0.025309789925813675, "global_step": 220209, "epoch": 2474, "lr": 8.567762766332165e-06}
{"train_loss": 0.08925670385360718, "global_step": 220210, "epoch": 2474, "lr": 8.567438247144726e-06}
{"train_loss": 0.0622326023876667, "global_step": 220211, "epoch": 2474, "lr": 8.567113733527382e-06}
{"train_loss": 0.03444657474756241, "global_step": 220212, "epoch": 2474, "lr": 8.566789225480198e-06}
{"train_loss": 0.03383143991231918, "global_step": 220213, "epoch": 2474, "lr": 8.566464723003187e-06}
{"train_loss": 0.02435821108520031, "global_step": 220214, "epoch": 2474, "lr": 8.566140226096414e-06}
{"train_loss": 0.04807182028889656, "global_step": 220215, "epoch": 2474, "lr": 8.56581573475993e-06}
{"train_loss": 0.07582143694162369, "global_step": 220216, "epoch": 2474, "lr": 8.565491248993751e-06}
{"train_loss": 0.04334726557135582, "global_step": 220217, "epoch": 2474, "lr": 8.565166768797954e-06}
{"train_loss": 0.03398546203970909, "global_step": 220218, "epoch": 2474, "lr": 8.564842294172548e-06}
{"train_loss": 0.04561866074800491, "global_step": 220219, "epoch": 2474, "lr": 8.564517825117614e-06}
{"train_loss": 0.0329643189907074, "global_step": 220220, "epoch": 2474, "lr": 8.564193361633155e-06}
{"train_loss": 0.04485725238919258, "global_step": 220221, "epoch": 2474, "lr": 8.563868903719252e-06}
{"train_loss": 0.05879530310630798, "global_step": 220222, "epoch": 2474, "lr": 8.563544451375922e-06}
{"train_loss": 0.09475012123584747, "global_step": 220223, "epoch": 2474, "lr": 8.563220004603224e-06}
{"train_loss": 0.024764686822891235, "global_step": 220224, "epoch": 2474, "lr": 8.562895563401185e-06}
{"train_loss": 0.03707347810268402, "global_step": 220225, "epoch": 2474, "lr": 8.562571127769876e-06}
{"train_loss": 0.04192192479968071, "global_step": 220226, "epoch": 2474, "lr": 8.562246697709309e-06}
{"train_loss": 0.02792065404355526, "global_step": 220227, "epoch": 2474, "lr": 8.561922273219558e-06}
{"train_loss": 0.061635762453079224, "global_step": 220228, "epoch": 2474, "lr": 8.561597854300635e-06}
{"train_loss": 0.04947919398546219, "global_step": 220229, "epoch": 2474, "lr": 8.561273440952616e-06}
{"train_loss": 0.023431044071912766, "global_step": 220230, "epoch": 2474, "lr": 8.560949033175514e-06}
{"train_loss": 0.03721092641353607, "global_step": 220231, "epoch": 2474, "lr": 8.5606246309694e-06}
{"train_loss": 0.052212897688150406, "global_step": 220232, "epoch": 2474, "lr": 8.560300234334306e-06}
{"train_loss": 0.038115493953228, "global_step": 220233, "epoch": 2474, "lr": 8.559975843270262e-06}
{"train_loss": 0.06240672618150711, "global_step": 220234, "epoch": 2474, "lr": 8.559651457777335e-06}
{"train_loss": 0.03917301073670387, "global_step": 220235, "epoch": 2474, "lr": 8.559327077855545e-06}
{"train_loss": 0.04481302946805954, "global_step": 220236, "epoch": 2474, "lr": 8.559002703504965e-06}
{"train_loss": 0.06686649471521378, "global_step": 220237, "epoch": 2474, "lr": 8.558678334725606e-06}
{"train_loss": 0.0536818690598011, "global_step": 220238, "epoch": 2474, "lr": 8.558353971517541e-06}
{"train_loss": 0.0956074669957161, "global_step": 220239, "epoch": 2474, "lr": 8.558029613880785e-06}
{"train_loss": 0.02668784186244011, "global_step": 220240, "epoch": 2474, "lr": 8.557705261815414e-06}
{"train_loss": 0.03671209514141083, "global_step": 220241, "epoch": 2474, "lr": 8.557380915321439e-06}
{"train_loss": 0.01413300447165966, "global_step": 220242, "epoch": 2474, "lr": 8.557056574398936e-06}
{"train_loss": 0.061822712421417236, "global_step": 220243, "epoch": 2474, "lr": 8.556732239047915e-06}
{"train_loss": 0.04998987913131714, "global_step": 220244, "epoch": 2474, "lr": 8.55640790926845e-06}
{"train_loss": 0.09646875411272049, "global_step": 220245, "epoch": 2474, "lr": 8.556083585060554e-06}
{"train_loss": 0.04405655711889267, "global_step": 220246, "epoch": 2474, "lr": 8.555759266424306e-06}
{"train_loss": 0.06360359489917755, "global_step": 220247, "epoch": 2474, "lr": 8.555434953359715e-06}
{"train_loss": 0.061868924647569656, "global_step": 220248, "epoch": 2474, "lr": 8.555110645866854e-06}
{"train_loss": 0.04807784780859947, "global_step": 220249, "epoch": 2474, "lr": 8.554786343945743e-06}
{"train_loss": 0.07042306661605835, "global_step": 220250, "epoch": 2474, "lr": 8.554462047596435e-06}
{"train_loss": 0.038636766374111176, "global_step": 220251, "epoch": 2474, "lr": 8.554137756818981e-06}
{"train_loss": 0.05949468910694122, "global_step": 220252, "epoch": 2474, "lr": 8.553813471613415e-06}
{"train_loss": 0.03891075402498245, "global_step": 220253, "epoch": 2474, "lr": 8.55348919197979e-06}
{"train_loss": 0.06965798884630203, "global_step": 220254, "epoch": 2474, "lr": 8.553164917918132e-06}
{"train_loss": 0.03379806503653526, "global_step": 220255, "epoch": 2474, "lr": 8.552840649428506e-06}
{"train_loss": 0.05918790027499199, "global_step": 220256, "epoch": 2474, "lr": 8.55251638651094e-06}
{"train_loss": 0.038093049079179764, "global_step": 220257, "epoch": 2474, "lr": 8.552192129165488e-06}
{"train_loss": 0.028264980763196945, "global_step": 220258, "epoch": 2474, "lr": 8.551867877392178e-06}
{"train_loss": 0.05814804136753082, "global_step": 220259, "epoch": 2474, "lr": 8.551543631191079e-06}
{"train_loss": 0.05186023935675621, "global_step": 220260, "epoch": 2474, "lr": 8.551219390562204e-06}
{"train_loss": 0.05249837413430214, "global_step": 220261, "epoch": 2474, "lr": 8.550895155505628e-06}
{"train_loss": 0.11121588200330734, "global_step": 220262, "epoch": 2474, "lr": 8.550570926021367e-06}
{"train_loss": 0.07340439409017563, "global_step": 220263, "epoch": 2474, "lr": 8.550246702109482e-06}
{"train_loss": 0.04960690438747406, "global_step": 220264, "epoch": 2474, "lr": 8.549922483770006e-06}
{"train_loss": 0.04945258051156998, "global_step": 220265, "epoch": 2474, "lr": 8.549598271003e-06}
{"train_loss": 0.056239932775497437, "global_step": 220266, "epoch": 2474, "lr": 8.549274063808483e-06}
{"train_loss": 0.037831321358680725, "global_step": 220267, "epoch": 2474, "lr": 8.548949862186517e-06}
{"train_loss": 0.06084061041474342, "global_step": 220268, "epoch": 2474, "lr": 8.548625666137133e-06}
{"train_loss": 0.07860907167196274, "global_step": 220269, "epoch": 2474, "lr": 8.548301475660391e-06}
{"train_loss": 0.04079829528927803, "global_step": 220270, "epoch": 2474, "lr": 8.547977290756326e-06}
{"train_loss": 0.03162079676985741, "global_step": 220271, "epoch": 2474, "lr": 8.54765311142497e-06}
{"train_loss": 0.07396168261766434, "global_step": 220272, "epoch": 2474, "lr": 8.547328937666383e-06}
{"train_loss": 0.06947294622659683, "global_step": 220273, "epoch": 2474, "lr": 8.547004769480593e-06}
{"train_loss": 0.05145719722750482, "global_step": 220274, "epoch": 2474, "lr": 8.54668060686767e-06, "val_loss": 8.90253734588623}
{"train_loss": 0.03170732408761978, "global_step": 220275, "epoch": 2475, "lr": 8.546356449827619e-06}
{"train_loss": 0.058337632566690445, "global_step": 220276, "epoch": 2475, "lr": 8.546032298360524e-06}
{"train_loss": 0.019900916144251823, "global_step": 220277, "epoch": 2475, "lr": 8.545708152466397e-06}
{"train_loss": 0.029757339507341385, "global_step": 220278, "epoch": 2475, "lr": 8.545384012145303e-06}
{"train_loss": 0.05193087086081505, "global_step": 220279, "epoch": 2475, "lr": 8.545059877397266e-06}
{"train_loss": 0.09828352183103561, "global_step": 220280, "epoch": 2475, "lr": 8.544735748222355e-06}
{"train_loss": 0.042203448712825775, "global_step": 220281, "epoch": 2475, "lr": 8.544411624620585e-06}
{"train_loss": 0.046955566853284836, "global_step": 220282, "epoch": 2475, "lr": 8.544087506592025e-06}
{"train_loss": 0.09480353444814682, "global_step": 220283, "epoch": 2475, "lr": 8.5437633941367e-06}
{"train_loss": 0.09332679212093353, "global_step": 220284, "epoch": 2475, "lr": 8.543439287254668e-06}
{"train_loss": 0.020190484821796417, "global_step": 220285, "epoch": 2475, "lr": 8.543115185945955e-06}
{"train_loss": 0.07477319240570068, "global_step": 220286, "epoch": 2475, "lr": 8.54279109021061e-06}
{"train_loss": 0.038796767592430115, "global_step": 220287, "epoch": 2475, "lr": 8.542467000048699e-06}
{"train_loss": 0.037471797317266464, "global_step": 220288, "epoch": 2475, "lr": 8.542142915460234e-06}
{"train_loss": 0.014034867286682129, "global_step": 220289, "epoch": 2475, "lr": 8.541818836445287e-06}
{"train_loss": 0.0809435099363327, "global_step": 220290, "epoch": 2475, "lr": 8.541494763003872e-06}
{"train_loss": 0.0658322423696518, "global_step": 220291, "epoch": 2475, "lr": 8.541170695136064e-06}
{"train_loss": 0.05359981581568718, "global_step": 220292, "epoch": 2475, "lr": 8.540846632841876e-06}
{"train_loss": 0.03743309527635574, "global_step": 220293, "epoch": 2475, "lr": 8.540522576121374e-06}
{"train_loss": 0.06028256192803383, "global_step": 220294, "epoch": 2475, "lr": 8.540198524974586e-06}
{"train_loss": 0.019235022366046906, "global_step": 220295, "epoch": 2475, "lr": 8.539874479401571e-06}
{"train_loss": 0.06339465081691742, "global_step": 220296, "epoch": 2475, "lr": 8.539550439402355e-06}
{"train_loss": 0.05624614283442497, "global_step": 220297, "epoch": 2475, "lr": 8.539226404977008e-06}
{"train_loss": 0.02603977918624878, "global_step": 220298, "epoch": 2475, "lr": 8.538902376125535e-06}
{"train_loss": 0.06396698951721191, "global_step": 220299, "epoch": 2475, "lr": 8.53857835284802e-06}
{"train_loss": 0.0617942251265049, "global_step": 220300, "epoch": 2475, "lr": 8.538254335144474e-06}
{"train_loss": 0.04001957178115845, "global_step": 220301, "epoch": 2475, "lr": 8.53793032301497e-06}
{"train_loss": 0.033088862895965576, "global_step": 220302, "epoch": 2475, "lr": 8.537606316459518e-06}
{"train_loss": 0.010023939423263073, "global_step": 220303, "epoch": 2475, "lr": 8.537282315478195e-06}
{"train_loss": 0.055748920887708664, "global_step": 220304, "epoch": 2475, "lr": 8.536958320071015e-06}
{"train_loss": 0.07318147271871567, "global_step": 220305, "epoch": 2475, "lr": 8.536634330238046e-06}
{"train_loss": 0.058537259697914124, "global_step": 220306, "epoch": 2475, "lr": 8.536310345979315e-06}
{"train_loss": 0.02608594484627247, "global_step": 220307, "epoch": 2475, "lr": 8.53598636729488e-06}
{"train_loss": 0.10333733260631561, "global_step": 220308, "epoch": 2475, "lr": 8.535662394184762e-06}
{"train_loss": 0.06873352080583572, "global_step": 220309, "epoch": 2475, "lr": 8.535338426649037e-06}
{"train_loss": 0.10310538858175278, "global_step": 220310, "epoch": 2475, "lr": 8.535014464687723e-06}
{"train_loss": 0.05423871800303459, "global_step": 220311, "epoch": 2475, "lr": 8.534690508300864e-06}
{"train_loss": 0.05466350540518761, "global_step": 220312, "epoch": 2475, "lr": 8.534366557488522e-06}
{"train_loss": 0.07873085141181946, "global_step": 220313, "epoch": 2475, "lr": 8.534042612250715e-06}
{"train_loss": 0.03844860941171646, "global_step": 220314, "epoch": 2475, "lr": 8.533718672587515e-06}
{"train_loss": 0.0284697487950325, "global_step": 220315, "epoch": 2475, "lr": 8.533394738498935e-06}
{"train_loss": 0.014204753562808037, "global_step": 220316, "epoch": 2475, "lr": 8.53307080998505e-06}
{"train_loss": 0.06060964614152908, "global_step": 220317, "epoch": 2475, "lr": 8.532746887045879e-06}
{"train_loss": 0.07978767156600952, "global_step": 220318, "epoch": 2475, "lr": 8.53242296968148e-06}
{"train_loss": 0.03967251256108284, "global_step": 220319, "epoch": 2475, "lr": 8.532099057891884e-06}
{"train_loss": 0.028727436438202858, "global_step": 220320, "epoch": 2475, "lr": 8.531775151677152e-06}
{"train_loss": 0.03214690834283829, "global_step": 220321, "epoch": 2475, "lr": 8.531451251037303e-06}
{"train_loss": 0.04726950079202652, "global_step": 220322, "epoch": 2475, "lr": 8.531127355972402e-06}
{"train_loss": 0.10027322173118591, "global_step": 220323, "epoch": 2475, "lr": 8.53080346648249e-06}
{"train_loss": 0.048029638826847076, "global_step": 220324, "epoch": 2475, "lr": 8.530479582567602e-06}
{"train_loss": 0.040230054408311844, "global_step": 220325, "epoch": 2475, "lr": 8.530155704227789e-06}
{"train_loss": 0.06850600242614746, "global_step": 220326, "epoch": 2475, "lr": 8.529831831463086e-06}
{"train_loss": 0.048055484890937805, "global_step": 220327, "epoch": 2475, "lr": 8.529507964273548e-06}
{"train_loss": 0.07181784510612488, "global_step": 220328, "epoch": 2475, "lr": 8.529184102659205e-06}
{"train_loss": 0.03626806288957596, "global_step": 220329, "epoch": 2475, "lr": 8.528860246620119e-06}
{"train_loss": 0.03901737183332443, "global_step": 220330, "epoch": 2475, "lr": 8.528536396156306e-06}
{"train_loss": 0.0388869047164917, "global_step": 220331, "epoch": 2475, "lr": 8.528212551267845e-06}
{"train_loss": 0.07531971484422684, "global_step": 220332, "epoch": 2475, "lr": 8.52788871195474e-06}
{"train_loss": 0.03095223195850849, "global_step": 220333, "epoch": 2475, "lr": 8.527564878217071e-06}
{"train_loss": 0.05317981168627739, "global_step": 220334, "epoch": 2475, "lr": 8.527241050054857e-06}
{"train_loss": 0.05638541653752327, "global_step": 220335, "epoch": 2475, "lr": 8.526917227468156e-06}
{"train_loss": 0.045725785195827484, "global_step": 220336, "epoch": 2475, "lr": 8.526593410456996e-06}
{"train_loss": 0.07668878138065338, "global_step": 220337, "epoch": 2475, "lr": 8.526269599021441e-06}
{"train_loss": 0.0334189310669899, "global_step": 220338, "epoch": 2475, "lr": 8.525945793161517e-06}
{"train_loss": 0.05364111810922623, "global_step": 220339, "epoch": 2475, "lr": 8.52562199287728e-06}
{"train_loss": 0.038355883210897446, "global_step": 220340, "epoch": 2475, "lr": 8.525298198168758e-06}
{"train_loss": 0.051003940403461456, "global_step": 220341, "epoch": 2475, "lr": 8.524974409036001e-06}
{"train_loss": 0.06429556012153625, "global_step": 220342, "epoch": 2475, "lr": 8.524650625479069e-06}
{"train_loss": 0.037940457463264465, "global_step": 220343, "epoch": 2475, "lr": 8.524326847497994e-06}
{"train_loss": 0.05733416602015495, "global_step": 220344, "epoch": 2475, "lr": 8.524003075092806e-06}
{"train_loss": 0.07047884166240692, "global_step": 220345, "epoch": 2475, "lr": 8.52367930826357e-06}
{"train_loss": 0.04625299945473671, "global_step": 220346, "epoch": 2475, "lr": 8.523355547010309e-06}
{"train_loss": 0.08318216353654861, "global_step": 220347, "epoch": 2475, "lr": 8.523031791333092e-06}
{"train_loss": 0.09524840861558914, "global_step": 220348, "epoch": 2475, "lr": 8.522708041231942e-06}
{"train_loss": 0.06447882205247879, "global_step": 220349, "epoch": 2475, "lr": 8.522384296706903e-06}
{"train_loss": 0.017379142343997955, "global_step": 220350, "epoch": 2475, "lr": 8.522060557758033e-06}
{"train_loss": 0.03710557892918587, "global_step": 220351, "epoch": 2475, "lr": 8.521736824385352e-06}
{"train_loss": 0.03979838266968727, "global_step": 220352, "epoch": 2475, "lr": 8.521413096588932e-06}
{"train_loss": 0.012819347903132439, "global_step": 220353, "epoch": 2475, "lr": 8.52108937436879e-06}
{"train_loss": 0.0863301008939743, "global_step": 220354, "epoch": 2475, "lr": 8.520765657724994e-06}
{"train_loss": 0.05423165485262871, "global_step": 220355, "epoch": 2475, "lr": 8.520441946657565e-06}
{"train_loss": 0.06276662647724152, "global_step": 220356, "epoch": 2475, "lr": 8.52011824116657e-06}
{"train_loss": 0.016811978071928024, "global_step": 220357, "epoch": 2475, "lr": 8.519794541252023e-06}
{"train_loss": 0.039438921958208084, "global_step": 220358, "epoch": 2475, "lr": 8.519470846913985e-06}
{"train_loss": 0.05491074547171593, "global_step": 220359, "epoch": 2475, "lr": 8.519147158152513e-06}
{"train_loss": 0.06684163957834244, "global_step": 220360, "epoch": 2475, "lr": 8.518823474967625e-06}
{"train_loss": 0.10353393852710724, "global_step": 220361, "epoch": 2475, "lr": 8.518499797359387e-06}
{"train_loss": 0.09907539933919907, "global_step": 220362, "epoch": 2475, "lr": 8.518176125327815e-06}
{"train_loss": 0.05303481668024586, "global_step": 220363, "epoch": 2475, "lr": 8.517852458872983e-06, "val_loss": 8.92237663269043, "train_action_mse_error": 20.40974998474121}
{"train_loss": 0.04821893572807312, "global_step": 220364, "epoch": 2476, "lr": 8.517528797994906e-06}
{"train_loss": 0.04127303883433342, "global_step": 220365, "epoch": 2476, "lr": 8.517205142693658e-06}
{"train_loss": 0.05354902148246765, "global_step": 220366, "epoch": 2476, "lr": 8.516881492969253e-06}
{"train_loss": 0.054019905626773834, "global_step": 220367, "epoch": 2476, "lr": 8.51655784882176e-06}
{"train_loss": 0.04787634685635567, "global_step": 220368, "epoch": 2476, "lr": 8.516234210251195e-06}
{"train_loss": 0.06428684294223785, "global_step": 220369, "epoch": 2476, "lr": 8.515910577257629e-06}
{"train_loss": 0.037916991859674454, "global_step": 220370, "epoch": 2476, "lr": 8.515586949841087e-06}
{"train_loss": 0.10529707372188568, "global_step": 220371, "epoch": 2476, "lr": 8.515263328001627e-06}
{"train_loss": 0.04233214259147644, "global_step": 220372, "epoch": 2476, "lr": 8.514939711739272e-06}
{"train_loss": 0.017841022461652756, "global_step": 220373, "epoch": 2476, "lr": 8.51461610105409e-06}
{"train_loss": 0.025966424494981766, "global_step": 220374, "epoch": 2476, "lr": 8.514292495946097e-06}
{"train_loss": 0.049756281077861786, "global_step": 220375, "epoch": 2476, "lr": 8.513968896415369e-06}
{"train_loss": 0.044452447444200516, "global_step": 220376, "epoch": 2476, "lr": 8.513645302461914e-06}
{"train_loss": 0.041449133306741714, "global_step": 220377, "epoch": 2476, "lr": 8.513321714085798e-06}
{"train_loss": 0.05608369782567024, "global_step": 220378, "epoch": 2476, "lr": 8.512998131287075e-06}
{"train_loss": 0.053412482142448425, "global_step": 220379, "epoch": 2476, "lr": 8.512674554065758e-06}
{"train_loss": 0.06937072426080704, "global_step": 220380, "epoch": 2476, "lr": 8.512350982421918e-06}
{"train_loss": 0.07607974112033844, "global_step": 220381, "epoch": 2476, "lr": 8.512027416355589e-06}
{"train_loss": 0.08483138680458069, "global_step": 220382, "epoch": 2476, "lr": 8.511703855866798e-06}
{"train_loss": 0.06791581958532333, "global_step": 220383, "epoch": 2476, "lr": 8.511380300955619e-06}
{"train_loss": 0.04714961349964142, "global_step": 220384, "epoch": 2476, "lr": 8.51105675162206e-06}
{"train_loss": 0.06646852195262909, "global_step": 220385, "epoch": 2476, "lr": 8.5107332078662e-06}
{"train_loss": 0.05621185898780823, "global_step": 220386, "epoch": 2476, "lr": 8.510409669688069e-06}
{"train_loss": 0.034689679741859436, "global_step": 220387, "epoch": 2476, "lr": 8.510086137087692e-06}
{"train_loss": 0.02953307330608368, "global_step": 220388, "epoch": 2476, "lr": 8.509762610065142e-06}
{"train_loss": 0.05036741867661476, "global_step": 220389, "epoch": 2476, "lr": 8.509439088620436e-06}
{"train_loss": 0.04754497855901718, "global_step": 220390, "epoch": 2476, "lr": 8.50911557275364e-06}
{"train_loss": 0.06012223660945892, "global_step": 220391, "epoch": 2476, "lr": 8.508792062464777e-06}
{"train_loss": 0.07120799273252487, "global_step": 220392, "epoch": 2476, "lr": 8.508468557753912e-06}
{"train_loss": 0.07094816863536835, "global_step": 220393, "epoch": 2476, "lr": 8.50814505862107e-06}
{"train_loss": 0.0757398009300232, "global_step": 220394, "epoch": 2476, "lr": 8.50782156506631e-06}
{"train_loss": 0.024065768346190453, "global_step": 220395, "epoch": 2476, "lr": 8.50749807708966e-06}
{"train_loss": 0.05574783310294151, "global_step": 220396, "epoch": 2476, "lr": 8.507174594691164e-06}
{"train_loss": 0.09524565190076828, "global_step": 220397, "epoch": 2476, "lr": 8.50685111787089e-06}
{"train_loss": 0.054460808634757996, "global_step": 220398, "epoch": 2476, "lr": 8.506527646628853e-06}
{"train_loss": 0.039734598249197006, "global_step": 220399, "epoch": 2476, "lr": 8.506204180965116e-06}
{"train_loss": 0.016814490780234337, "global_step": 220400, "epoch": 2476, "lr": 8.505880720879705e-06}
{"train_loss": 0.0511382594704628, "global_step": 220401, "epoch": 2476, "lr": 8.505557266372683e-06}
{"train_loss": 0.07704618573188782, "global_step": 220402, "epoch": 2476, "lr": 8.505233817444069e-06}
{"train_loss": 0.06250202655792236, "global_step": 220403, "epoch": 2476, "lr": 8.504910374093939e-06}
{"train_loss": 0.08996336907148361, "global_step": 220404, "epoch": 2476, "lr": 8.5045869363223e-06}
{"train_loss": 0.049548011273145676, "global_step": 220405, "epoch": 2476, "lr": 8.504263504129229e-06}
{"train_loss": 0.05345628410577774, "global_step": 220406, "epoch": 2476, "lr": 8.503940077514739e-06}
{"train_loss": 0.08511190116405487, "global_step": 220407, "epoch": 2476, "lr": 8.503616656478903e-06}
{"train_loss": 0.06889713555574417, "global_step": 220408, "epoch": 2476, "lr": 8.503293241021737e-06}
{"train_loss": 0.07024344056844711, "global_step": 220409, "epoch": 2476, "lr": 8.502969831143309e-06}
{"train_loss": 0.0296329315751791, "global_step": 220410, "epoch": 2476, "lr": 8.50264642684364e-06}
{"train_loss": 0.04256041720509529, "global_step": 220411, "epoch": 2476, "lr": 8.5023230281228e-06}
{"train_loss": 0.07936768233776093, "global_step": 220412, "epoch": 2476, "lr": 8.501999634980795e-06}
{"train_loss": 0.08866360038518906, "global_step": 220413, "epoch": 2476, "lr": 8.5016762474177e-06}
{"train_loss": 0.07108248025178909, "global_step": 220414, "epoch": 2476, "lr": 8.501352865433554e-06}
{"train_loss": 0.037119437009096146, "global_step": 220415, "epoch": 2476, "lr": 8.501029489028388e-06}
{"train_loss": 0.03353865072131157, "global_step": 220416, "epoch": 2476, "lr": 8.500706118202261e-06}
{"train_loss": 0.07438947260379791, "global_step": 220417, "epoch": 2476, "lr": 8.5003827529552e-06}
{"train_loss": 0.05115598067641258, "global_step": 220418, "epoch": 2476, "lr": 8.500059393287268e-06}
{"train_loss": 0.07369482517242432, "global_step": 220419, "epoch": 2476, "lr": 8.499736039198485e-06}
{"train_loss": 0.03565454110503197, "global_step": 220420, "epoch": 2476, "lr": 8.499412690688919e-06}
{"train_loss": 0.05243203416466713, "global_step": 220421, "epoch": 2476, "lr": 8.499089347758599e-06}
{"train_loss": 0.04749031364917755, "global_step": 220422, "epoch": 2476, "lr": 8.498766010407561e-06}
{"train_loss": 0.04755401983857155, "global_step": 220423, "epoch": 2476, "lr": 8.498442678635865e-06}
{"train_loss": 0.0747418999671936, "global_step": 220424, "epoch": 2476, "lr": 8.49811935244354e-06}
{"train_loss": 0.053955093026161194, "global_step": 220425, "epoch": 2476, "lr": 8.497796031830652e-06}
{"train_loss": 0.03433483839035034, "global_step": 220426, "epoch": 2476, "lr": 8.497472716797222e-06}
{"train_loss": 0.04660208895802498, "global_step": 220427, "epoch": 2476, "lr": 8.497149407343297e-06}
{"train_loss": 0.03989816829562187, "global_step": 220428, "epoch": 2476, "lr": 8.49682610346893e-06}
{"train_loss": 0.04509100317955017, "global_step": 220429, "epoch": 2476, "lr": 8.49650280517415e-06}
{"train_loss": 0.07311436533927917, "global_step": 220430, "epoch": 2476, "lr": 8.496179512459018e-06}
{"train_loss": 0.07976435124874115, "global_step": 220431, "epoch": 2476, "lr": 8.495856225323562e-06}
{"train_loss": 0.07394973933696747, "global_step": 220432, "epoch": 2476, "lr": 8.495532943767831e-06}
{"train_loss": 0.04555274918675423, "global_step": 220433, "epoch": 2476, "lr": 8.495209667791876e-06}
{"train_loss": 0.037151042371988297, "global_step": 220434, "epoch": 2476, "lr": 8.49488639739573e-06}
{"train_loss": 0.060113582760095596, "global_step": 220435, "epoch": 2476, "lr": 8.494563132579447e-06}
{"train_loss": 0.019250506535172462, "global_step": 220436, "epoch": 2476, "lr": 8.494239873343057e-06}
{"train_loss": 0.08720852434635162, "global_step": 220437, "epoch": 2476, "lr": 8.493916619686615e-06}
{"train_loss": 0.04872792586684227, "global_step": 220438, "epoch": 2476, "lr": 8.493593371610153e-06}
{"train_loss": 0.05134304612874985, "global_step": 220439, "epoch": 2476, "lr": 8.493270129113734e-06}
{"train_loss": 0.025174042209982872, "global_step": 220440, "epoch": 2476, "lr": 8.492946892197378e-06}
{"train_loss": 0.02932140976190567, "global_step": 220441, "epoch": 2476, "lr": 8.492623660861148e-06}
{"train_loss": 0.03966649994254112, "global_step": 220442, "epoch": 2476, "lr": 8.492300435105072e-06}
{"train_loss": 0.02996310032904148, "global_step": 220443, "epoch": 2476, "lr": 8.49197721492921e-06}
{"train_loss": 0.04300633817911148, "global_step": 220444, "epoch": 2476, "lr": 8.491654000333576e-06}
{"train_loss": 0.07500416040420532, "global_step": 220445, "epoch": 2476, "lr": 8.491330791318252e-06}
{"train_loss": 0.05035421997308731, "global_step": 220446, "epoch": 2476, "lr": 8.491007587883254e-06}
{"train_loss": 0.02995350770652294, "global_step": 220447, "epoch": 2476, "lr": 8.490684390028641e-06}
{"train_loss": 0.03746449947357178, "global_step": 220448, "epoch": 2476, "lr": 8.490361197754437e-06}
{"train_loss": 0.07819852977991104, "global_step": 220449, "epoch": 2476, "lr": 8.490038011060703e-06}
{"train_loss": 0.06029800325632095, "global_step": 220450, "epoch": 2476, "lr": 8.489714829947487e-06}
{"train_loss": 0.04383121803402901, "global_step": 220451, "epoch": 2476, "lr": 8.489391654414808e-06}
{"train_loss": 0.05410557443255119, "global_step": 220452, "epoch": 2476, "lr": 8.489068484462743e-06, "val_loss": 9.007098197937012}
{"train_loss": 0.05772482231259346, "global_step": 220453, "epoch": 2477, "lr": 8.488745320091296e-06}
{"train_loss": 0.08032342046499252, "global_step": 220454, "epoch": 2477, "lr": 8.488422161300552e-06}
{"train_loss": 0.014342413283884525, "global_step": 220455, "epoch": 2477, "lr": 8.488099008090516e-06}
{"train_loss": 0.07962923496961594, "global_step": 220456, "epoch": 2477, "lr": 8.48777586046126e-06}
{"train_loss": 0.046799175441265106, "global_step": 220457, "epoch": 2477, "lr": 8.487452718412809e-06}
{"train_loss": 0.033900320529937744, "global_step": 220458, "epoch": 2477, "lr": 8.487129581945224e-06}
{"train_loss": 0.05039827525615692, "global_step": 220459, "epoch": 2477, "lr": 8.486806451058537e-06}
{"train_loss": 0.06984876841306686, "global_step": 220460, "epoch": 2477, "lr": 8.486483325752786e-06}
{"train_loss": 0.03881177678704262, "global_step": 220461, "epoch": 2477, "lr": 8.486160206028026e-06}
{"train_loss": 0.08135800808668137, "global_step": 220462, "epoch": 2477, "lr": 8.485837091884291e-06}
{"train_loss": 0.03915762901306152, "global_step": 220463, "epoch": 2477, "lr": 8.485513983321635e-06}
{"train_loss": 0.05039970204234123, "global_step": 220464, "epoch": 2477, "lr": 8.4851908803401e-06}
{"train_loss": 0.05961927771568298, "global_step": 220465, "epoch": 2477, "lr": 8.48486778293971e-06}
{"train_loss": 0.05048529803752899, "global_step": 220466, "epoch": 2477, "lr": 8.484544691120538e-06}
{"train_loss": 0.14325091242790222, "global_step": 220467, "epoch": 2477, "lr": 8.484221604882602e-06}
{"train_loss": 0.06321626901626587, "global_step": 220468, "epoch": 2477, "lr": 8.483898524225953e-06}
{"train_loss": 0.05675078183412552, "global_step": 220469, "epoch": 2477, "lr": 8.48357544915065e-06}
{"train_loss": 0.021993577480316162, "global_step": 220470, "epoch": 2477, "lr": 8.483252379656714e-06}
{"train_loss": 0.05839064344763756, "global_step": 220471, "epoch": 2477, "lr": 8.482929315744209e-06}
{"train_loss": 0.054116807878017426, "global_step": 220472, "epoch": 2477, "lr": 8.48260625741316e-06}
{"train_loss": 0.06056377664208412, "global_step": 220473, "epoch": 2477, "lr": 8.482283204663626e-06}
{"train_loss": 0.07791455090045929, "global_step": 220474, "epoch": 2477, "lr": 8.481960157495632e-06}
{"train_loss": 0.028160791844129562, "global_step": 220475, "epoch": 2477, "lr": 8.481637115909247e-06}
{"train_loss": 0.036302562803030014, "global_step": 220476, "epoch": 2477, "lr": 8.48131407990449e-06}
{"train_loss": 0.03467718884348869, "global_step": 220477, "epoch": 2477, "lr": 8.480991049481418e-06}
{"train_loss": 0.03881130367517471, "global_step": 220478, "epoch": 2477, "lr": 8.480668024640066e-06}
{"train_loss": 0.09067327529191971, "global_step": 220479, "epoch": 2477, "lr": 8.480345005380491e-06}
{"train_loss": 0.06761885434389114, "global_step": 220480, "epoch": 2477, "lr": 8.480021991702713e-06}
{"train_loss": 0.029786957427859306, "global_step": 220481, "epoch": 2477, "lr": 8.479698983606809e-06}
{"train_loss": 0.04215296357870102, "global_step": 220482, "epoch": 2477, "lr": 8.479375981092785e-06}
{"train_loss": 0.049071792513132095, "global_step": 220483, "epoch": 2477, "lr": 8.479052984160717e-06}
{"train_loss": 0.05201157554984093, "global_step": 220484, "epoch": 2477, "lr": 8.478729992810624e-06}
{"train_loss": 0.03799479454755783, "global_step": 220485, "epoch": 2477, "lr": 8.478407007042555e-06}
{"train_loss": 0.09701722115278244, "global_step": 220486, "epoch": 2477, "lr": 8.478084026856576e-06}
{"train_loss": 0.062282536178827286, "global_step": 220487, "epoch": 2477, "lr": 8.477761052252697e-06}
{"train_loss": 0.06082125008106232, "global_step": 220488, "epoch": 2477, "lr": 8.477438083230987e-06}
{"train_loss": 0.025646429508924484, "global_step": 220489, "epoch": 2477, "lr": 8.477115119791474e-06}
{"train_loss": 0.06339571624994278, "global_step": 220490, "epoch": 2477, "lr": 8.476792161934217e-06}
{"train_loss": 0.06126008182764053, "global_step": 220491, "epoch": 2477, "lr": 8.476469209659233e-06}
{"train_loss": 0.0571560300886631, "global_step": 220492, "epoch": 2477, "lr": 8.476146262966594e-06}
{"train_loss": 0.07145733386278152, "global_step": 220493, "epoch": 2477, "lr": 8.475823321856319e-06}
{"train_loss": 0.0759793147444725, "global_step": 220494, "epoch": 2477, "lr": 8.475500386328478e-06}
{"train_loss": 0.026607327163219452, "global_step": 220495, "epoch": 2477, "lr": 8.475177456383087e-06}
{"train_loss": 0.027384314686059952, "global_step": 220496, "epoch": 2477, "lr": 8.474854532020215e-06}
{"train_loss": 0.0755893662571907, "global_step": 220497, "epoch": 2477, "lr": 8.474531613239895e-06}
{"train_loss": 0.07916993647813797, "global_step": 220498, "epoch": 2477, "lr": 8.474208700042153e-06}
{"train_loss": 0.06687590479850769, "global_step": 220499, "epoch": 2477, "lr": 8.473885792427055e-06}
{"train_loss": 0.030172143131494522, "global_step": 220500, "epoch": 2477, "lr": 8.473562890394631e-06}
{"train_loss": 0.05532268434762955, "global_step": 220501, "epoch": 2477, "lr": 8.473239993944942e-06}
{"train_loss": 0.054500456899404526, "global_step": 220502, "epoch": 2477, "lr": 8.472917103078016e-06}
{"train_loss": 0.031159119680523872, "global_step": 220503, "epoch": 2477, "lr": 8.472594217793895e-06}
{"train_loss": 0.06423673778772354, "global_step": 220504, "epoch": 2477, "lr": 8.472271338092619e-06}
{"train_loss": 0.09292460232973099, "global_step": 220505, "epoch": 2477, "lr": 8.471948463974256e-06}
{"train_loss": 0.052067339420318604, "global_step": 220506, "epoch": 2477, "lr": 8.471625595438826e-06}
{"train_loss": 0.05046798288822174, "global_step": 220507, "epoch": 2477, "lr": 8.471302732486386e-06}
{"train_loss": 0.07920683920383453, "global_step": 220508, "epoch": 2477, "lr": 8.470979875116963e-06}
{"train_loss": 0.08590478450059891, "global_step": 220509, "epoch": 2477, "lr": 8.470657023330619e-06}
{"train_loss": 0.04016688093543053, "global_step": 220510, "epoch": 2477, "lr": 8.470334177127382e-06}
{"train_loss": 0.0567442961037159, "global_step": 220511, "epoch": 2477, "lr": 8.470011336507312e-06}
{"train_loss": 0.100867360830307, "global_step": 220512, "epoch": 2477, "lr": 8.46968850147043e-06}
{"train_loss": 0.05944883078336716, "global_step": 220513, "epoch": 2477, "lr": 8.469365672016804e-06}
{"train_loss": 0.10997795313596725, "global_step": 220514, "epoch": 2477, "lr": 8.469042848146458e-06}
{"train_loss": 0.06227261945605278, "global_step": 220515, "epoch": 2477, "lr": 8.46872002985945e-06}
{"train_loss": 0.03721494600176811, "global_step": 220516, "epoch": 2477, "lr": 8.46839721715581e-06}
{"train_loss": 0.036252211779356, "global_step": 220517, "epoch": 2477, "lr": 8.468074410035592e-06}
{"train_loss": 0.059487227350473404, "global_step": 220518, "epoch": 2477, "lr": 8.46775160849883e-06}
{"train_loss": 0.04128606617450714, "global_step": 220519, "epoch": 2477, "lr": 8.46742881254558e-06}
{"train_loss": 0.05142366886138916, "global_step": 220520, "epoch": 2477, "lr": 8.467106022175868e-06}
{"train_loss": 0.04294693097472191, "global_step": 220521, "epoch": 2477, "lr": 8.466783237389764e-06}
{"train_loss": 0.041240524500608444, "global_step": 220522, "epoch": 2477, "lr": 8.466460458187275e-06}
{"train_loss": 0.03787265717983246, "global_step": 220523, "epoch": 2477, "lr": 8.46613768456847e-06}
{"train_loss": 0.04437408596277237, "global_step": 220524, "epoch": 2477, "lr": 8.465814916533393e-06}
{"train_loss": 0.024457480758428574, "global_step": 220525, "epoch": 2477, "lr": 8.465492154082071e-06}
{"train_loss": 0.052542515099048615, "global_step": 220526, "epoch": 2477, "lr": 8.465169397214573e-06}
{"train_loss": 0.027076948434114456, "global_step": 220527, "epoch": 2477, "lr": 8.464846645930913e-06}
{"train_loss": 0.12174807488918304, "global_step": 220528, "epoch": 2477, "lr": 8.464523900231158e-06}
{"train_loss": 0.03629108518362045, "global_step": 220529, "epoch": 2477, "lr": 8.464201160115331e-06}
{"train_loss": 0.07132599502801895, "global_step": 220530, "epoch": 2477, "lr": 8.463878425583493e-06}
{"train_loss": 0.07952883839607239, "global_step": 220531, "epoch": 2477, "lr": 8.463555696635678e-06}
{"train_loss": 0.04839903488755226, "global_step": 220532, "epoch": 2477, "lr": 8.463232973271934e-06}
{"train_loss": 0.05457659810781479, "global_step": 220533, "epoch": 2477, "lr": 8.462910255492295e-06}
{"train_loss": 0.022653212770819664, "global_step": 220534, "epoch": 2477, "lr": 8.462587543296824e-06}
{"train_loss": 0.04869277775287628, "global_step": 220535, "epoch": 2477, "lr": 8.46226483668554e-06}
{"train_loss": 0.05140530318021774, "global_step": 220536, "epoch": 2477, "lr": 8.461942135658507e-06}
{"train_loss": 0.02017834223806858, "global_step": 220537, "epoch": 2477, "lr": 8.461619440215763e-06}
{"train_loss": 0.0672554299235344, "global_step": 220538, "epoch": 2477, "lr": 8.461296750357334e-06}
{"train_loss": 0.0719219297170639, "global_step": 220539, "epoch": 2477, "lr": 8.460974066083294e-06}
{"train_loss": 0.03636446222662926, "global_step": 220540, "epoch": 2477, "lr": 8.460651387393654e-06}
{"train_loss": 0.05557796473123049, "global_step": 220541, "epoch": 2477, "lr": 8.460328714288484e-06, "val_loss": 8.972883224487305}
{"train_loss": 0.030940746888518333, "global_step": 220542, "epoch": 2478, "lr": 8.460006046767804e-06}
{"train_loss": 0.06004481390118599, "global_step": 220543, "epoch": 2478, "lr": 8.459683384831684e-06}
{"train_loss": 0.03353402763605118, "global_step": 220544, "epoch": 2478, "lr": 8.459360728480137e-06}
{"train_loss": 0.059232134371995926, "global_step": 220545, "epoch": 2478, "lr": 8.459038077713239e-06}
{"train_loss": 0.047907184809446335, "global_step": 220546, "epoch": 2478, "lr": 8.458715432531005e-06}
{"train_loss": 0.048263415694236755, "global_step": 220547, "epoch": 2478, "lr": 8.4583927929335e-06}
{"train_loss": 0.05748118460178375, "global_step": 220548, "epoch": 2478, "lr": 8.45807015892075e-06}
{"train_loss": 0.06241981312632561, "global_step": 220549, "epoch": 2478, "lr": 8.457747530492815e-06}
{"train_loss": 0.0772496685385704, "global_step": 220550, "epoch": 2478, "lr": 8.457424907649719e-06}
{"train_loss": 0.06719143688678741, "global_step": 220551, "epoch": 2478, "lr": 8.457102290391527e-06}
{"train_loss": 0.02862136997282505, "global_step": 220552, "epoch": 2478, "lr": 8.456779678718257e-06}
{"train_loss": 0.030590713024139404, "global_step": 220553, "epoch": 2478, "lr": 8.45645707262998e-06}
{"train_loss": 0.051355890929698944, "global_step": 220554, "epoch": 2478, "lr": 8.456134472126715e-06}
{"train_loss": 0.03816555067896843, "global_step": 220555, "epoch": 2478, "lr": 8.455811877208531e-06}
{"train_loss": 0.05718276649713516, "global_step": 220556, "epoch": 2478, "lr": 8.45548928787544e-06}
{"train_loss": 0.06242770701646805, "global_step": 220557, "epoch": 2478, "lr": 8.455166704127515e-06}
{"train_loss": 0.042397648096084595, "global_step": 220558, "epoch": 2478, "lr": 8.454844125964778e-06}
{"train_loss": 0.027626246213912964, "global_step": 220559, "epoch": 2478, "lr": 8.45452155338728e-06}
{"train_loss": 0.05037521570920944, "global_step": 220560, "epoch": 2478, "lr": 8.454198986395074e-06}
{"train_loss": 0.08586355298757553, "global_step": 220561, "epoch": 2478, "lr": 8.453876424988183e-06}
{"train_loss": 0.06500889360904694, "global_step": 220562, "epoch": 2478, "lr": 8.453553869166675e-06}
{"train_loss": 0.0424920991063118, "global_step": 220563, "epoch": 2478, "lr": 8.453231318930565e-06}
{"train_loss": 0.10096724331378937, "global_step": 220564, "epoch": 2478, "lr": 8.452908774279932e-06}
{"train_loss": 0.0832512229681015, "global_step": 220565, "epoch": 2478, "lr": 8.452586235214782e-06}
{"train_loss": 0.06760521233081818, "global_step": 220566, "epoch": 2478, "lr": 8.452263701735186e-06}
{"train_loss": 0.033482398837804794, "global_step": 220567, "epoch": 2478, "lr": 8.451941173841166e-06}
{"train_loss": 0.057296186685562134, "global_step": 220568, "epoch": 2478, "lr": 8.45161865153279e-06}
{"train_loss": 0.08008505403995514, "global_step": 220569, "epoch": 2478, "lr": 8.451296134810071e-06}
{"train_loss": 0.021238695830106735, "global_step": 220570, "epoch": 2478, "lr": 8.450973623673087e-06}
{"train_loss": 0.04570341855287552, "global_step": 220571, "epoch": 2478, "lr": 8.450651118121844e-06}
{"train_loss": 0.026001280173659325, "global_step": 220572, "epoch": 2478, "lr": 8.450328618156423e-06}
{"train_loss": 0.03045455366373062, "global_step": 220573, "epoch": 2478, "lr": 8.450006123776833e-06}
{"train_loss": 0.06967011094093323, "global_step": 220574, "epoch": 2478, "lr": 8.449683634983146e-06}
{"train_loss": 0.0646035224199295, "global_step": 220575, "epoch": 2478, "lr": 8.449361151775393e-06}
{"train_loss": 0.07825852185487747, "global_step": 220576, "epoch": 2478, "lr": 8.449038674153603e-06}
{"train_loss": 0.039775386452674866, "global_step": 220577, "epoch": 2478, "lr": 8.448716202117851e-06}
{"train_loss": 0.06994961202144623, "global_step": 220578, "epoch": 2478, "lr": 8.448393735668148e-06}
{"train_loss": 0.10205035656690598, "global_step": 220579, "epoch": 2478, "lr": 8.44807127480456e-06}
{"train_loss": 0.05108955129981041, "global_step": 220580, "epoch": 2478, "lr": 8.447748819527113e-06}
{"train_loss": 0.06851056963205338, "global_step": 220581, "epoch": 2478, "lr": 8.447426369835875e-06}
{"train_loss": 0.04938687011599541, "global_step": 220582, "epoch": 2478, "lr": 8.447103925730859e-06}
{"train_loss": 0.07346105575561523, "global_step": 220583, "epoch": 2478, "lr": 8.446781487212135e-06}
{"train_loss": 0.04393098130822182, "global_step": 220584, "epoch": 2478, "lr": 8.44645905427972e-06}
{"train_loss": 0.09953872859477997, "global_step": 220585, "epoch": 2478, "lr": 8.446136626933693e-06}
{"train_loss": 0.07507789134979248, "global_step": 220586, "epoch": 2478, "lr": 8.445814205174058e-06}
{"train_loss": 0.05491168051958084, "global_step": 220587, "epoch": 2478, "lr": 8.445491789000886e-06}
{"train_loss": 0.07962915301322937, "global_step": 220588, "epoch": 2478, "lr": 8.445169378414202e-06}
{"train_loss": 0.08040178567171097, "global_step": 220589, "epoch": 2478, "lr": 8.444846973414073e-06}
{"train_loss": 0.04174744337797165, "global_step": 220590, "epoch": 2478, "lr": 8.444524574000517e-06}
{"train_loss": 0.06828270852565765, "global_step": 220591, "epoch": 2478, "lr": 8.444202180173595e-06}
{"train_loss": 0.07647746801376343, "global_step": 220592, "epoch": 2478, "lr": 8.443879791933335e-06}
{"train_loss": 0.03557848930358887, "global_step": 220593, "epoch": 2478, "lr": 8.4435574092798e-06}
{"train_loss": 0.09066883474588394, "global_step": 220594, "epoch": 2478, "lr": 8.44323503221301e-06}
{"train_loss": 0.03312354162335396, "global_step": 220595, "epoch": 2478, "lr": 8.44291266073302e-06}
{"train_loss": 0.05411319062113762, "global_step": 220596, "epoch": 2478, "lr": 8.442590294839886e-06}
{"train_loss": 0.10820107907056808, "global_step": 220597, "epoch": 2478, "lr": 8.442267934533626e-06}
{"train_loss": 0.09040344506502151, "global_step": 220598, "epoch": 2478, "lr": 8.441945579814314e-06}
{"train_loss": 0.05334748700261116, "global_step": 220599, "epoch": 2478, "lr": 8.441623230681961e-06}
{"train_loss": 0.09115244448184967, "global_step": 220600, "epoch": 2478, "lr": 8.441300887136634e-06}
{"train_loss": 0.04450593888759613, "global_step": 220601, "epoch": 2478, "lr": 8.440978549178357e-06}
{"train_loss": 0.09580642729997635, "global_step": 220602, "epoch": 2478, "lr": 8.440656216807197e-06}
{"train_loss": 0.0967155322432518, "global_step": 220603, "epoch": 2478, "lr": 8.440333890023173e-06}
{"train_loss": 0.021178776398301125, "global_step": 220604, "epoch": 2478, "lr": 8.44001156882635e-06}
{"train_loss": 0.060900550335645676, "global_step": 220605, "epoch": 2478, "lr": 8.439689253216749e-06}
{"train_loss": 0.04234183952212334, "global_step": 220606, "epoch": 2478, "lr": 8.439366943194437e-06}
{"train_loss": 0.06962680071592331, "global_step": 220607, "epoch": 2478, "lr": 8.439044638759437e-06}
{"train_loss": 0.04061251878738403, "global_step": 220608, "epoch": 2478, "lr": 8.438722339911809e-06}
{"train_loss": 0.040574006736278534, "global_step": 220609, "epoch": 2478, "lr": 8.438400046651578e-06}
{"train_loss": 0.09082828462123871, "global_step": 220610, "epoch": 2478, "lr": 8.438077758978807e-06}
{"train_loss": 0.056160420179367065, "global_step": 220611, "epoch": 2478, "lr": 8.437755476893521e-06}
{"train_loss": 0.030696289613842964, "global_step": 220612, "epoch": 2478, "lr": 8.437433200395783e-06}
{"train_loss": 0.04938863590359688, "global_step": 220613, "epoch": 2478, "lr": 8.437110929485626e-06}
{"train_loss": 0.05816435068845749, "global_step": 220614, "epoch": 2478, "lr": 8.436788664163082e-06}
{"train_loss": 0.05610428377985954, "global_step": 220615, "epoch": 2478, "lr": 8.436466404428212e-06}
{"train_loss": 0.049056701362133026, "global_step": 220616, "epoch": 2478, "lr": 8.436144150281045e-06}
{"train_loss": 0.0341748371720314, "global_step": 220617, "epoch": 2478, "lr": 8.435821901721647e-06}
{"train_loss": 0.061304133385419846, "global_step": 220618, "epoch": 2478, "lr": 8.435499658750028e-06}
{"train_loss": 0.03729277849197388, "global_step": 220619, "epoch": 2478, "lr": 8.435177421366264e-06}
{"train_loss": 0.059173841029405594, "global_step": 220620, "epoch": 2478, "lr": 8.434855189570368e-06}
{"train_loss": 0.11767720431089401, "global_step": 220621, "epoch": 2478, "lr": 8.434532963362419e-06}
{"train_loss": 0.07077816128730774, "global_step": 220622, "epoch": 2478, "lr": 8.434210742742421e-06}
{"train_loss": 0.05336255952715874, "global_step": 220623, "epoch": 2478, "lr": 8.433888527710453e-06}
{"train_loss": 0.06073363497853279, "global_step": 220624, "epoch": 2478, "lr": 8.433566318266528e-06}
{"train_loss": 0.09686939418315887, "global_step": 220625, "epoch": 2478, "lr": 8.433244114410715e-06}
{"train_loss": 0.01989705115556717, "global_step": 220626, "epoch": 2478, "lr": 8.432921916143039e-06}
{"train_loss": 0.07675212621688843, "global_step": 220627, "epoch": 2478, "lr": 8.432599723463553e-06}
{"train_loss": 0.06674523651599884, "global_step": 220628, "epoch": 2478, "lr": 8.43227753637229e-06}
{"train_loss": 0.06430138647556305, "global_step": 220629, "epoch": 2478, "lr": 8.431955354869314e-06}
{"train_loss": 0.05939828129380607, "global_step": 220630, "epoch": 2478, "lr": 8.43163317895464e-06, "val_loss": 8.85239028930664}
{"train_loss": 0.06092590466141701, "global_step": 220631, "epoch": 2479, "lr": 8.431311008628328e-06}
{"train_loss": 0.05301232635974884, "global_step": 220632, "epoch": 2479, "lr": 8.430988843890436e-06}
{"train_loss": 0.07207781821489334, "global_step": 220633, "epoch": 2479, "lr": 8.430666684740973e-06}
{"train_loss": 0.04563957452774048, "global_step": 220634, "epoch": 2479, "lr": 8.430344531180013e-06}
{"train_loss": 0.04606941342353821, "global_step": 220635, "epoch": 2479, "lr": 8.430022383207576e-06}
{"train_loss": 0.06045645847916603, "global_step": 220636, "epoch": 2479, "lr": 8.429700240823723e-06}
{"train_loss": 0.04087642952799797, "global_step": 220637, "epoch": 2479, "lr": 8.429378104028484e-06}
{"train_loss": 0.02146766521036625, "global_step": 220638, "epoch": 2479, "lr": 8.42905597282192e-06}
{"train_loss": 0.07355637848377228, "global_step": 220639, "epoch": 2479, "lr": 8.428733847204051e-06}
{"train_loss": 0.061450861394405365, "global_step": 220640, "epoch": 2479, "lr": 8.42841172717494e-06}
{"train_loss": 0.05922304838895798, "global_step": 220641, "epoch": 2479, "lr": 8.428089612734613e-06}
{"train_loss": 0.047757770866155624, "global_step": 220642, "epoch": 2479, "lr": 8.427767503883133e-06}
{"train_loss": 0.026501720771193504, "global_step": 220643, "epoch": 2479, "lr": 8.427445400620526e-06}
{"train_loss": 0.0984310731291771, "global_step": 220644, "epoch": 2479, "lr": 8.427123302946848e-06}
{"train_loss": 0.03530314192175865, "global_step": 220645, "epoch": 2479, "lr": 8.426801210862128e-06}
{"train_loss": 0.06210250407457352, "global_step": 220646, "epoch": 2479, "lr": 8.426479124366431e-06}
{"train_loss": 0.011420448310673237, "global_step": 220647, "epoch": 2479, "lr": 8.42615704345977e-06}
{"train_loss": 0.07742635160684586, "global_step": 220648, "epoch": 2479, "lr": 8.42583496814222e-06}
{"train_loss": 0.048504799604415894, "global_step": 220649, "epoch": 2479, "lr": 8.4255128984138e-06}
{"train_loss": 0.05303049460053444, "global_step": 220650, "epoch": 2479, "lr": 8.425190834274577e-06}
{"train_loss": 0.06954799592494965, "global_step": 220651, "epoch": 2479, "lr": 8.424868775724564e-06}
{"train_loss": 0.03255905956029892, "global_step": 220652, "epoch": 2479, "lr": 8.424546722763832e-06}
{"train_loss": 0.0497659370303154, "global_step": 220653, "epoch": 2479, "lr": 8.424224675392417e-06}
{"train_loss": 0.08884920179843903, "global_step": 220654, "epoch": 2479, "lr": 8.423902633610343e-06}
{"train_loss": 0.04182768985629082, "global_step": 220655, "epoch": 2479, "lr": 8.42358059741768e-06}
{"train_loss": 0.03959779068827629, "global_step": 220656, "epoch": 2479, "lr": 8.423258566814451e-06}
{"train_loss": 0.07647676765918732, "global_step": 220657, "epoch": 2479, "lr": 8.422936541800719e-06}
{"train_loss": 0.061304669827222824, "global_step": 220658, "epoch": 2479, "lr": 8.422614522376499e-06}
{"train_loss": 0.037253960967063904, "global_step": 220659, "epoch": 2479, "lr": 8.422292508541873e-06}
{"train_loss": 0.035923153162002563, "global_step": 220660, "epoch": 2479, "lr": 8.421970500296844e-06}
{"train_loss": 0.13817135989665985, "global_step": 220661, "epoch": 2479, "lr": 8.421648497641488e-06}
{"train_loss": 0.06481096148490906, "global_step": 220662, "epoch": 2479, "lr": 8.421326500575828e-06}
{"train_loss": 0.05827418714761734, "global_step": 220663, "epoch": 2479, "lr": 8.421004509099917e-06}
{"train_loss": 0.03548368066549301, "global_step": 220664, "epoch": 2479, "lr": 8.420682523213791e-06}
{"train_loss": 0.04252954572439194, "global_step": 220665, "epoch": 2479, "lr": 8.420360542917505e-06}
{"train_loss": 0.046420153230428696, "global_step": 220666, "epoch": 2479, "lr": 8.42003856821108e-06}
{"train_loss": 0.03192256763577461, "global_step": 220667, "epoch": 2479, "lr": 8.41971659909458e-06}
{"train_loss": 0.02410132624208927, "global_step": 220668, "epoch": 2479, "lr": 8.419394635568051e-06}
{"train_loss": 0.038855984807014465, "global_step": 220669, "epoch": 2479, "lr": 8.419072677631518e-06}
{"train_loss": 0.05299815163016319, "global_step": 220670, "epoch": 2479, "lr": 8.41875072528504e-06}
{"train_loss": 0.04236532002687454, "global_step": 220671, "epoch": 2479, "lr": 8.418428778528648e-06}
{"train_loss": 0.03838779777288437, "global_step": 220672, "epoch": 2479, "lr": 8.418106837362406e-06}
{"train_loss": 0.09895920753479004, "global_step": 220673, "epoch": 2479, "lr": 8.417784901786325e-06}
{"train_loss": 0.034118689596652985, "global_step": 220674, "epoch": 2479, "lr": 8.417462971800477e-06}
{"train_loss": 0.05265393108129501, "global_step": 220675, "epoch": 2479, "lr": 8.417141047404887e-06}
{"train_loss": 0.05569278076291084, "global_step": 220676, "epoch": 2479, "lr": 8.416819128599618e-06}
{"train_loss": 0.07934922724962234, "global_step": 220677, "epoch": 2479, "lr": 8.416497215384683e-06}
{"train_loss": 0.050315652042627335, "global_step": 220678, "epoch": 2479, "lr": 8.41617530776016e-06}
{"train_loss": 0.05189639329910278, "global_step": 220679, "epoch": 2479, "lr": 8.41585340572606e-06}
{"train_loss": 0.05182290077209473, "global_step": 220680, "epoch": 2479, "lr": 8.41553150928246e-06}
{"train_loss": 0.03530111908912659, "global_step": 220681, "epoch": 2479, "lr": 8.415209618429365e-06}
{"train_loss": 0.029577232897281647, "global_step": 220682, "epoch": 2479, "lr": 8.414887733166854e-06}
{"train_loss": 0.029803363606333733, "global_step": 220683, "epoch": 2479, "lr": 8.414565853494943e-06}
{"train_loss": 0.06369844079017639, "global_step": 220684, "epoch": 2479, "lr": 8.4142439794137e-06}
{"train_loss": 0.020623473450541496, "global_step": 220685, "epoch": 2479, "lr": 8.41392211092314e-06}
{"train_loss": 0.049216121435165405, "global_step": 220686, "epoch": 2479, "lr": 8.413600248023334e-06}
{"train_loss": 0.08256794512271881, "global_step": 220687, "epoch": 2479, "lr": 8.413278390714302e-06}
{"train_loss": 0.02708638459444046, "global_step": 220688, "epoch": 2479, "lr": 8.412956538996108e-06}
{"train_loss": 0.058046791702508926, "global_step": 220689, "epoch": 2479, "lr": 8.412634692868776e-06}
{"train_loss": 0.09381026774644852, "global_step": 220690, "epoch": 2479, "lr": 8.412312852332365e-06}
{"train_loss": 0.0756501853466034, "global_step": 220691, "epoch": 2479, "lr": 8.411991017386917e-06}
{"train_loss": 0.10972315818071365, "global_step": 220692, "epoch": 2479, "lr": 8.411669188032456e-06}
{"train_loss": 0.026528706774115562, "global_step": 220693, "epoch": 2479, "lr": 8.41134736426905e-06}
{"train_loss": 0.029144609346985817, "global_step": 220694, "epoch": 2479, "lr": 8.411025546096723e-06}
{"train_loss": 0.05174438655376434, "global_step": 220695, "epoch": 2479, "lr": 8.410703733515535e-06}
{"train_loss": 0.04986652359366417, "global_step": 220696, "epoch": 2479, "lr": 8.410381926525507e-06}
{"train_loss": 0.06222279369831085, "global_step": 220697, "epoch": 2479, "lr": 8.410060125126712e-06}
{"train_loss": 0.017791936174035072, "global_step": 220698, "epoch": 2479, "lr": 8.409738329319167e-06}
{"train_loss": 0.03459944203495979, "global_step": 220699, "epoch": 2479, "lr": 8.409416539102939e-06}
{"train_loss": 0.05832861736416817, "global_step": 220700, "epoch": 2479, "lr": 8.409094754478042e-06}
{"train_loss": 0.04905904829502106, "global_step": 220701, "epoch": 2479, "lr": 8.408772975444545e-06}
{"train_loss": 0.05149362236261368, "global_step": 220702, "epoch": 2479, "lr": 8.40845120200247e-06}
{"train_loss": 0.023216430097818375, "global_step": 220703, "epoch": 2479, "lr": 8.40812943415188e-06}
{"train_loss": 0.04492191597819328, "global_step": 220704, "epoch": 2479, "lr": 8.407807671892814e-06}
{"train_loss": 0.06989441066980362, "global_step": 220705, "epoch": 2479, "lr": 8.407485915225299e-06}
{"train_loss": 0.0422988198697567, "global_step": 220706, "epoch": 2479, "lr": 8.407164164149406e-06}
{"train_loss": 0.03550500050187111, "global_step": 220707, "epoch": 2479, "lr": 8.40684241866515e-06}
{"train_loss": 0.06638778746128082, "global_step": 220708, "epoch": 2479, "lr": 8.4065206787726e-06}
{"train_loss": 0.05968455970287323, "global_step": 220709, "epoch": 2479, "lr": 8.406198944471778e-06}
{"train_loss": 0.0531051903963089, "global_step": 220710, "epoch": 2479, "lr": 8.405877215762742e-06}
{"train_loss": 0.056726302951574326, "global_step": 220711, "epoch": 2479, "lr": 8.405555492645517e-06}
{"train_loss": 0.023586921393871307, "global_step": 220712, "epoch": 2479, "lr": 8.40523377512017e-06}
{"train_loss": 0.050980545580387115, "global_step": 220713, "epoch": 2479, "lr": 8.40491206318672e-06}
{"train_loss": 0.09628096967935562, "global_step": 220714, "epoch": 2479, "lr": 8.404590356845238e-06}
{"train_loss": 0.03417707607150078, "global_step": 220715, "epoch": 2479, "lr": 8.404268656095737e-06}
{"train_loss": 0.020283523947000504, "global_step": 220716, "epoch": 2479, "lr": 8.40394696093829e-06}
{"train_loss": 0.041855014860630035, "global_step": 220717, "epoch": 2479, "lr": 8.40362527137291e-06}
{"train_loss": 0.04948394373059273, "global_step": 220718, "epoch": 2479, "lr": 8.40330358739967e-06}
{"train_loss": 0.05240483528651883, "global_step": 220719, "epoch": 2479, "lr": 8.402981909018587e-06, "val_loss": 8.888371467590332}
{"train_loss": 0.05518655478954315, "global_step": 220720, "epoch": 2480, "lr": 8.402660236229726e-06}
{"train_loss": 0.06677451729774475, "global_step": 220721, "epoch": 2480, "lr": 8.402338569033108e-06}
{"train_loss": 0.05700105428695679, "global_step": 220722, "epoch": 2480, "lr": 8.402016907428789e-06}
{"train_loss": 0.08802303671836853, "global_step": 220723, "epoch": 2480, "lr": 8.40169525141683e-06}
{"train_loss": 0.10668253153562546, "global_step": 220724, "epoch": 2480, "lr": 8.401373600997247e-06}
{"train_loss": 0.11296118050813675, "global_step": 220725, "epoch": 2480, "lr": 8.401051956170086e-06}
{"train_loss": 0.06244158372282982, "global_step": 220726, "epoch": 2480, "lr": 8.400730316935407e-06}
{"train_loss": 0.028109543025493622, "global_step": 220727, "epoch": 2480, "lr": 8.400408683293232e-06}
{"train_loss": 0.06639327853918076, "global_step": 220728, "epoch": 2480, "lr": 8.400087055243621e-06}
{"train_loss": 0.11346815526485443, "global_step": 220729, "epoch": 2480, "lr": 8.399765432786615e-06}
{"train_loss": 0.07238975167274475, "global_step": 220730, "epoch": 2480, "lr": 8.399443815922242e-06}
{"train_loss": 0.0335981622338295, "global_step": 220731, "epoch": 2480, "lr": 8.399122204650572e-06}
{"train_loss": 0.066824771463871, "global_step": 220732, "epoch": 2480, "lr": 8.398800598971619e-06}
{"train_loss": 0.06076819449663162, "global_step": 220733, "epoch": 2480, "lr": 8.39847899888545e-06}
{"train_loss": 0.01801186054944992, "global_step": 220734, "epoch": 2480, "lr": 8.398157404392087e-06}
{"train_loss": 0.0924132913351059, "global_step": 220735, "epoch": 2480, "lr": 8.397835815491595e-06}
{"train_loss": 0.03132592514157295, "global_step": 220736, "epoch": 2480, "lr": 8.397514232184001e-06}
{"train_loss": 0.03410612419247627, "global_step": 220737, "epoch": 2480, "lr": 8.397192654469361e-06}
{"train_loss": 0.04553665593266487, "global_step": 220738, "epoch": 2480, "lr": 8.396871082347701e-06}
{"train_loss": 0.033147700130939484, "global_step": 220739, "epoch": 2480, "lr": 8.396549515819074e-06}
{"train_loss": 0.06213061511516571, "global_step": 220740, "epoch": 2480, "lr": 8.396227954883535e-06}
{"train_loss": 0.02964947372674942, "global_step": 220741, "epoch": 2480, "lr": 8.395906399541103e-06}
{"train_loss": 0.03623548895120621, "global_step": 220742, "epoch": 2480, "lr": 8.395584849791855e-06}
{"train_loss": 0.03997831419110298, "global_step": 220743, "epoch": 2480, "lr": 8.395263305635792e-06}
{"train_loss": 0.046235084533691406, "global_step": 220744, "epoch": 2480, "lr": 8.394941767072995e-06}
{"train_loss": 0.057421907782554626, "global_step": 220745, "epoch": 2480, "lr": 8.394620234103478e-06}
{"train_loss": 0.0487784780561924, "global_step": 220746, "epoch": 2480, "lr": 8.394298706727304e-06}
{"train_loss": 0.03317810595035553, "global_step": 220747, "epoch": 2480, "lr": 8.393977184944508e-06}
{"train_loss": 0.06494444608688354, "global_step": 220748, "epoch": 2480, "lr": 8.39365566875514e-06}
{"train_loss": 0.044572148472070694, "global_step": 220749, "epoch": 2480, "lr": 8.393334158159228e-06}
{"train_loss": 0.05485181882977486, "global_step": 220750, "epoch": 2480, "lr": 8.393012653156835e-06}
{"train_loss": 0.025954248383641243, "global_step": 220751, "epoch": 2480, "lr": 8.392691153747983e-06}
{"train_loss": 0.05797676742076874, "global_step": 220752, "epoch": 2480, "lr": 8.392369659932735e-06}
{"train_loss": 0.10971445590257645, "global_step": 220753, "epoch": 2480, "lr": 8.39204817171112e-06}
{"train_loss": 0.020701028406620026, "global_step": 220754, "epoch": 2480, "lr": 8.391726689083196e-06}
{"train_loss": 0.02545539289712906, "global_step": 220755, "epoch": 2480, "lr": 8.391405212048985e-06}
{"train_loss": 0.10603365302085876, "global_step": 220756, "epoch": 2480, "lr": 8.391083740608557e-06}
{"train_loss": 0.05156689137220383, "global_step": 220757, "epoch": 2480, "lr": 8.390762274761927e-06}
{"train_loss": 0.034316450357437134, "global_step": 220758, "epoch": 2480, "lr": 8.390440814509153e-06}
{"train_loss": 0.05697362869977951, "global_step": 220759, "epoch": 2480, "lr": 8.39011935985029e-06}
{"train_loss": 0.08318503201007843, "global_step": 220760, "epoch": 2480, "lr": 8.389797910785352e-06}
{"train_loss": 0.053986139595508575, "global_step": 220761, "epoch": 2480, "lr": 8.38947646731441e-06}
{"train_loss": 0.05322626978158951, "global_step": 220762, "epoch": 2480, "lr": 8.38915502943749e-06}
{"train_loss": 0.046771567314863205, "global_step": 220763, "epoch": 2480, "lr": 8.38883359715465e-06}
{"train_loss": 0.017886152490973473, "global_step": 220764, "epoch": 2480, "lr": 8.388512170465923e-06}
{"train_loss": 0.13134661316871643, "global_step": 220765, "epoch": 2480, "lr": 8.388190749371344e-06}
{"train_loss": 0.028915654867887497, "global_step": 220766, "epoch": 2480, "lr": 8.387869333870974e-06}
{"train_loss": 0.0507209338247776, "global_step": 220767, "epoch": 2480, "lr": 8.387547923964834e-06}
{"train_loss": 0.03837153688073158, "global_step": 220768, "epoch": 2480, "lr": 8.387226519652996e-06}
{"train_loss": 0.060314569622278214, "global_step": 220769, "epoch": 2480, "lr": 8.38690512093549e-06}
{"train_loss": 0.04294160380959511, "global_step": 220770, "epoch": 2480, "lr": 8.386583727812347e-06}
{"train_loss": 0.06515739858150482, "global_step": 220771, "epoch": 2480, "lr": 8.386262340283629e-06}
{"train_loss": 0.0708574652671814, "global_step": 220772, "epoch": 2480, "lr": 8.385940958349358e-06}
{"train_loss": 0.01874767802655697, "global_step": 220773, "epoch": 2480, "lr": 8.385619582009601e-06}
{"train_loss": 0.03190068155527115, "global_step": 220774, "epoch": 2480, "lr": 8.385298211264381e-06}
{"train_loss": 0.08958236873149872, "global_step": 220775, "epoch": 2480, "lr": 8.384976846113762e-06}
{"train_loss": 0.06839480996131897, "global_step": 220776, "epoch": 2480, "lr": 8.384655486557768e-06}
{"train_loss": 0.03596813231706619, "global_step": 220777, "epoch": 2480, "lr": 8.384334132596443e-06}
{"train_loss": 0.044194694608449936, "global_step": 220778, "epoch": 2480, "lr": 8.38401278422985e-06}
{"train_loss": 0.035789936780929565, "global_step": 220779, "epoch": 2480, "lr": 8.383691441458013e-06}
{"train_loss": 0.014234469272196293, "global_step": 220780, "epoch": 2480, "lr": 8.383370104280985e-06}
{"train_loss": 0.07632957398891449, "global_step": 220781, "epoch": 2480, "lr": 8.383048772698798e-06}
{"train_loss": 0.09514010697603226, "global_step": 220782, "epoch": 2480, "lr": 8.38272744671152e-06}
{"train_loss": 0.081326924264431, "global_step": 220783, "epoch": 2480, "lr": 8.38240612631916e-06}
{"train_loss": 0.020283518359065056, "global_step": 220784, "epoch": 2480, "lr": 8.382084811521789e-06}
{"train_loss": 0.06536442786455154, "global_step": 220785, "epoch": 2480, "lr": 8.381763502319429e-06}
{"train_loss": 0.021913697943091393, "global_step": 220786, "epoch": 2480, "lr": 8.381442198712141e-06}
{"train_loss": 0.018532149493694305, "global_step": 220787, "epoch": 2480, "lr": 8.381120900699957e-06}
{"train_loss": 0.046159643679857254, "global_step": 220788, "epoch": 2480, "lr": 8.38079960828293e-06}
{"train_loss": 0.039822839200496674, "global_step": 220789, "epoch": 2480, "lr": 8.380478321461088e-06}
{"train_loss": 0.054477009922266006, "global_step": 220790, "epoch": 2480, "lr": 8.380157040234492e-06}
{"train_loss": 0.0438624732196331, "global_step": 220791, "epoch": 2480, "lr": 8.379835764603172e-06}
{"train_loss": 0.030431175604462624, "global_step": 220792, "epoch": 2480, "lr": 8.379514494567181e-06}
{"train_loss": 0.07201041281223297, "global_step": 220793, "epoch": 2480, "lr": 8.379193230126548e-06}
{"train_loss": 0.0499272495508194, "global_step": 220794, "epoch": 2480, "lr": 8.378871971281326e-06}
{"train_loss": 0.05854630842804909, "global_step": 220795, "epoch": 2480, "lr": 8.378550718031564e-06}
{"train_loss": 0.029441947117447853, "global_step": 220796, "epoch": 2480, "lr": 8.37822947037729e-06}
{"train_loss": 0.057946521788835526, "global_step": 220797, "epoch": 2480, "lr": 8.377908228318571e-06}
{"train_loss": 0.05831744894385338, "global_step": 220798, "epoch": 2480, "lr": 8.377586991855418e-06}
{"train_loss": 0.06752244383096695, "global_step": 220799, "epoch": 2480, "lr": 8.377265760987906e-06}
{"train_loss": 0.09054329246282578, "global_step": 220800, "epoch": 2480, "lr": 8.376944535716053e-06}
{"train_loss": 0.08135806769132614, "global_step": 220801, "epoch": 2480, "lr": 8.376623316039922e-06}
{"train_loss": 0.07813378423452377, "global_step": 220802, "epoch": 2480, "lr": 8.376302101959543e-06}
{"train_loss": 0.029266778379678726, "global_step": 220803, "epoch": 2480, "lr": 8.375980893474955e-06}
{"train_loss": 0.08099763840436935, "global_step": 220804, "epoch": 2480, "lr": 8.375659690586224e-06}
{"train_loss": 0.05276012420654297, "global_step": 220805, "epoch": 2480, "lr": 8.37533849329336e-06}
{"train_loss": 0.07495848089456558, "global_step": 220806, "epoch": 2480, "lr": 8.375017301596439e-06}
{"train_loss": 0.09528721123933792, "global_step": 220807, "epoch": 2480, "lr": 8.374696115495489e-06}
{"train_loss": 0.055779181368481576, "global_step": 220808, "epoch": 2480, "lr": 8.374374934990542e-06, "val_loss": 8.7824125289917, "train_action_mse_error": 6.598809242248535}
{"train_loss": 0.11076053231954575, "global_step": 220809, "epoch": 2481, "lr": 8.374053760081668e-06}
{"train_loss": 0.031828057020902634, "global_step": 220810, "epoch": 2481, "lr": 8.373732590768879e-06}
{"train_loss": 0.06468578428030014, "global_step": 220811, "epoch": 2481, "lr": 8.373411427052247e-06}
{"train_loss": 0.03724902868270874, "global_step": 220812, "epoch": 2481, "lr": 8.373090268931794e-06}
{"train_loss": 0.08530506491661072, "global_step": 220813, "epoch": 2481, "lr": 8.37276911640757e-06}
{"train_loss": 0.05672372877597809, "global_step": 220814, "epoch": 2481, "lr": 8.37244796947963e-06}
{"train_loss": 0.04319063946604729, "global_step": 220815, "epoch": 2481, "lr": 8.372126828147997e-06}
{"train_loss": 0.07108460366725922, "global_step": 220816, "epoch": 2481, "lr": 8.371805692412731e-06}
{"train_loss": 0.05862808972597122, "global_step": 220817, "epoch": 2481, "lr": 8.371484562273863e-06}
{"train_loss": 0.05141739174723625, "global_step": 220818, "epoch": 2481, "lr": 8.37116343773145e-06}
{"train_loss": 0.029537243768572807, "global_step": 220819, "epoch": 2481, "lr": 8.370842318785516e-06}
{"train_loss": 0.037475571036338806, "global_step": 220820, "epoch": 2481, "lr": 8.370521205436127e-06}
{"train_loss": 0.044730208814144135, "global_step": 220821, "epoch": 2481, "lr": 8.3702000976833e-06}
{"train_loss": 0.043398212641477585, "global_step": 220822, "epoch": 2481, "lr": 8.369878995527102e-06}
{"train_loss": 0.06848662346601486, "global_step": 220823, "epoch": 2481, "lr": 8.369557898967562e-06}
{"train_loss": 0.032994698733091354, "global_step": 220824, "epoch": 2481, "lr": 8.369236808004738e-06}
{"train_loss": 0.015615765005350113, "global_step": 220825, "epoch": 2481, "lr": 8.368915722638642e-06}
{"train_loss": 0.05760914832353592, "global_step": 220826, "epoch": 2481, "lr": 8.36859464286936e-06}
{"train_loss": 0.027714500203728676, "global_step": 220827, "epoch": 2481, "lr": 8.368273568696894e-06}
{"train_loss": 0.0643511414527893, "global_step": 220828, "epoch": 2481, "lr": 8.367952500121322e-06}
{"train_loss": 0.05957453325390816, "global_step": 220829, "epoch": 2481, "lr": 8.36763143714266e-06}
{"train_loss": 0.044357117265462875, "global_step": 220830, "epoch": 2481, "lr": 8.367310379760962e-06}
{"train_loss": 0.04430375620722771, "global_step": 220831, "epoch": 2481, "lr": 8.366989327976283e-06}
{"train_loss": 0.06630660593509674, "global_step": 220832, "epoch": 2481, "lr": 8.366668281788642e-06}
{"train_loss": 0.052454784512519836, "global_step": 220833, "epoch": 2481, "lr": 8.366347241198109e-06}
{"train_loss": 0.03243844583630562, "global_step": 220834, "epoch": 2481, "lr": 8.366026206204696e-06}
{"train_loss": 0.04266851767897606, "global_step": 220835, "epoch": 2481, "lr": 8.36570517680848e-06}
{"train_loss": 0.02322121150791645, "global_step": 220836, "epoch": 2481, "lr": 8.365384153009476e-06}
{"train_loss": 0.06354773789644241, "global_step": 220837, "epoch": 2481, "lr": 8.36506313480775e-06}
{"train_loss": 0.04789905995130539, "global_step": 220838, "epoch": 2481, "lr": 8.364742122203323e-06}
{"train_loss": 0.04710523411631584, "global_step": 220839, "epoch": 2481, "lr": 8.364421115196259e-06}
{"train_loss": 0.06823621690273285, "global_step": 220840, "epoch": 2481, "lr": 8.364100113786588e-06}
{"train_loss": 0.04520507529377937, "global_step": 220841, "epoch": 2481, "lr": 8.363779117974351e-06}
{"train_loss": 0.04744597151875496, "global_step": 220842, "epoch": 2481, "lr": 8.363458127759605e-06}
{"train_loss": 0.0294375978410244, "global_step": 220843, "epoch": 2481, "lr": 8.363137143142374e-06}
{"train_loss": 0.037303049117326736, "global_step": 220844, "epoch": 2481, "lr": 8.362816164122723e-06}
{"train_loss": 0.059855468571186066, "global_step": 220845, "epoch": 2481, "lr": 8.36249519070068e-06}
{"train_loss": 0.026615457609295845, "global_step": 220846, "epoch": 2481, "lr": 8.362174222876284e-06}
{"train_loss": 0.042235929518938065, "global_step": 220847, "epoch": 2481, "lr": 8.361853260649599e-06}
{"train_loss": 0.05541863292455673, "global_step": 220848, "epoch": 2481, "lr": 8.361532304020642e-06}
{"train_loss": 0.0457976758480072, "global_step": 220849, "epoch": 2481, "lr": 8.361211352989467e-06}
{"train_loss": 0.08551284670829773, "global_step": 220850, "epoch": 2481, "lr": 8.360890407556133e-06}
{"train_loss": 0.04278812184929848, "global_step": 220851, "epoch": 2481, "lr": 8.360569467720658e-06}
{"train_loss": 0.03349726274609566, "global_step": 220852, "epoch": 2481, "lr": 8.360248533483112e-06}
{"train_loss": 0.04251299798488617, "global_step": 220853, "epoch": 2481, "lr": 8.35992760484351e-06}
{"train_loss": 0.07409925758838654, "global_step": 220854, "epoch": 2481, "lr": 8.359606681801919e-06}
{"train_loss": 0.038669414818286896, "global_step": 220855, "epoch": 2481, "lr": 8.359285764358354e-06}
{"train_loss": 0.07157443463802338, "global_step": 220856, "epoch": 2481, "lr": 8.358964852512895e-06}
{"train_loss": 0.03572925180196762, "global_step": 220857, "epoch": 2481, "lr": 8.358643946265548e-06}
{"train_loss": 0.028524234890937805, "global_step": 220858, "epoch": 2481, "lr": 8.358323045616389e-06}
{"train_loss": 0.026244059205055237, "global_step": 220859, "epoch": 2481, "lr": 8.358002150565437e-06}
{"train_loss": 0.057280875742435455, "global_step": 220860, "epoch": 2481, "lr": 8.357681261112754e-06}
{"train_loss": 0.04491599649190903, "global_step": 220861, "epoch": 2481, "lr": 8.35736037725836e-06}
{"train_loss": 0.06300780177116394, "global_step": 220862, "epoch": 2481, "lr": 8.357039499002323e-06}
{"train_loss": 0.05853347107768059, "global_step": 220863, "epoch": 2481, "lr": 8.356718626344662e-06}
{"train_loss": 0.027614392340183258, "global_step": 220864, "epoch": 2481, "lr": 8.356397759285445e-06}
{"train_loss": 0.09472702443599701, "global_step": 220865, "epoch": 2481, "lr": 8.356076897824694e-06}
{"train_loss": 0.062049660831689835, "global_step": 220866, "epoch": 2481, "lr": 8.35575604196246e-06}
{"train_loss": 0.04366922006011009, "global_step": 220867, "epoch": 2481, "lr": 8.355435191698796e-06}
{"train_loss": 0.0901726558804512, "global_step": 220868, "epoch": 2481, "lr": 8.355114347033727e-06}
{"train_loss": 0.039569512009620667, "global_step": 220869, "epoch": 2481, "lr": 8.354793507967313e-06}
{"train_loss": 0.049345921725034714, "global_step": 220870, "epoch": 2481, "lr": 8.354472674499581e-06}
{"train_loss": 0.025939717888832092, "global_step": 220871, "epoch": 2481, "lr": 8.354151846630593e-06}
{"train_loss": 0.038157667964696884, "global_step": 220872, "epoch": 2481, "lr": 8.35383102436037e-06}
{"train_loss": 0.06440084427595139, "global_step": 220873, "epoch": 2481, "lr": 8.35351020768898e-06}
{"train_loss": 0.043649233877658844, "global_step": 220874, "epoch": 2481, "lr": 8.35318939661644e-06}
{"train_loss": 0.05612526088953018, "global_step": 220875, "epoch": 2481, "lr": 8.35286859114282e-06}
{"train_loss": 0.010750089772045612, "global_step": 220876, "epoch": 2481, "lr": 8.352547791268134e-06}
{"train_loss": 0.07921590656042099, "global_step": 220877, "epoch": 2481, "lr": 8.352226996992457e-06}
{"train_loss": 0.03276147320866585, "global_step": 220878, "epoch": 2481, "lr": 8.351906208315797e-06}
{"train_loss": 0.05913514271378517, "global_step": 220879, "epoch": 2481, "lr": 8.351585425238235e-06}
{"train_loss": 0.058946505188941956, "global_step": 220880, "epoch": 2481, "lr": 8.351264647759787e-06}
{"train_loss": 0.022839846089482307, "global_step": 220881, "epoch": 2481, "lr": 8.350943875880497e-06}
{"train_loss": 0.01935933344066143, "global_step": 220882, "epoch": 2481, "lr": 8.350623109600424e-06}
{"train_loss": 0.05098974332213402, "global_step": 220883, "epoch": 2481, "lr": 8.350302348919592e-06}
{"train_loss": 0.08907416462898254, "global_step": 220884, "epoch": 2481, "lr": 8.349981593838068e-06}
{"train_loss": 0.04598381370306015, "global_step": 220885, "epoch": 2481, "lr": 8.349660844355867e-06}
{"train_loss": 0.0620422437787056, "global_step": 220886, "epoch": 2481, "lr": 8.349340100473057e-06}
{"train_loss": 0.04500541836023331, "global_step": 220887, "epoch": 2481, "lr": 8.349019362189658e-06}
{"train_loss": 0.09814703464508057, "global_step": 220888, "epoch": 2481, "lr": 8.348698629505742e-06}
{"train_loss": 0.041114021092653275, "global_step": 220889, "epoch": 2481, "lr": 8.348377902421323e-06}
{"train_loss": 0.08476058393716812, "global_step": 220890, "epoch": 2481, "lr": 8.34805718093647e-06}
{"train_loss": 0.048981841653585434, "global_step": 220891, "epoch": 2481, "lr": 8.347736465051198e-06}
{"train_loss": 0.07364512234926224, "global_step": 220892, "epoch": 2481, "lr": 8.347415754765576e-06}
{"train_loss": 0.05516748130321503, "global_step": 220893, "epoch": 2481, "lr": 8.347095050079623e-06}
{"train_loss": 0.06928454339504242, "global_step": 220894, "epoch": 2481, "lr": 8.346774350993414e-06}
{"train_loss": 0.029851196333765984, "global_step": 220895, "epoch": 2481, "lr": 8.346453657506959e-06}
{"train_loss": 0.028461573645472527, "global_step": 220896, "epoch": 2481, "lr": 8.346132969620325e-06}
{"train_loss": 0.05076822806986865, "global_step": 220897, "epoch": 2481, "lr": 8.345812287333537e-06, "val_loss": 8.792740821838379}
{"train_loss": 0.03930658474564552, "global_step": 220898, "epoch": 2482, "lr": 8.345491610646655e-06}
{"train_loss": 0.028902066871523857, "global_step": 220899, "epoch": 2482, "lr": 8.345170939559705e-06}
{"train_loss": 0.06505268812179565, "global_step": 220900, "epoch": 2482, "lr": 8.344850274072752e-06}
{"train_loss": 0.03093511052429676, "global_step": 220901, "epoch": 2482, "lr": 8.344529614185808e-06}
{"train_loss": 0.10026057064533234, "global_step": 220902, "epoch": 2482, "lr": 8.344208959898953e-06}
{"train_loss": 0.05757494270801544, "global_step": 220903, "epoch": 2482, "lr": 8.343888311212194e-06}
{"train_loss": 0.013502337969839573, "global_step": 220904, "epoch": 2482, "lr": 8.343567668125596e-06}
{"train_loss": 0.0517720952630043, "global_step": 220905, "epoch": 2482, "lr": 8.343247030639211e-06}
{"train_loss": 0.09852135181427002, "global_step": 220906, "epoch": 2482, "lr": 8.342926398753054e-06}
{"train_loss": 0.023505162447690964, "global_step": 220907, "epoch": 2482, "lr": 8.342605772467194e-06}
{"train_loss": 0.07376286387443542, "global_step": 220908, "epoch": 2482, "lr": 8.342285151781649e-06}
{"train_loss": 0.04790647327899933, "global_step": 220909, "epoch": 2482, "lr": 8.341964536696495e-06}
{"train_loss": 0.059178002178668976, "global_step": 220910, "epoch": 2482, "lr": 8.34164392721174e-06}
{"train_loss": 0.05118729919195175, "global_step": 220911, "epoch": 2482, "lr": 8.341323323327455e-06}
{"train_loss": 0.05572591722011566, "global_step": 220912, "epoch": 2482, "lr": 8.341002725043656e-06}
{"train_loss": 0.04161630570888519, "global_step": 220913, "epoch": 2482, "lr": 8.340682132360416e-06}
{"train_loss": 0.04838736727833748, "global_step": 220914, "epoch": 2482, "lr": 8.340361545277753e-06}
{"train_loss": 0.053010016679763794, "global_step": 220915, "epoch": 2482, "lr": 8.340040963795732e-06}
{"train_loss": 0.02474949322640896, "global_step": 220916, "epoch": 2482, "lr": 8.339720387914373e-06}
{"train_loss": 0.025609927251935005, "global_step": 220917, "epoch": 2482, "lr": 8.339399817633742e-06}
{"train_loss": 0.044359978288412094, "global_step": 220918, "epoch": 2482, "lr": 8.339079252953874e-06}
{"train_loss": 0.053705234080553055, "global_step": 220919, "epoch": 2482, "lr": 8.338758693874793e-06}
{"train_loss": 0.048268258571624756, "global_step": 220920, "epoch": 2482, "lr": 8.33843814039657e-06}
{"train_loss": 0.07608936727046967, "global_step": 220921, "epoch": 2482, "lr": 8.338117592519229e-06}
{"train_loss": 0.045735571533441544, "global_step": 220922, "epoch": 2482, "lr": 8.33779705024283e-06}
{"train_loss": 0.04701864346861839, "global_step": 220923, "epoch": 2482, "lr": 8.337476513567388e-06}
{"train_loss": 0.0477566123008728, "global_step": 220924, "epoch": 2482, "lr": 8.337155982492983e-06}
{"train_loss": 0.033766262233257294, "global_step": 220925, "epoch": 2482, "lr": 8.336835457019625e-06}
{"train_loss": 0.02658916637301445, "global_step": 220926, "epoch": 2482, "lr": 8.336514937147389e-06}
{"train_loss": 0.09623436629772186, "global_step": 220927, "epoch": 2482, "lr": 8.336194422876286e-06}
{"train_loss": 0.053805142641067505, "global_step": 220928, "epoch": 2482, "lr": 8.335873914206382e-06}
{"train_loss": 0.046324800699949265, "global_step": 220929, "epoch": 2482, "lr": 8.335553411137704e-06}
{"train_loss": 0.056740447878837585, "global_step": 220930, "epoch": 2482, "lr": 8.335232913670315e-06}
{"train_loss": 0.047303952276706696, "global_step": 220931, "epoch": 2482, "lr": 8.33491242180423e-06}
{"train_loss": 0.00938734132796526, "global_step": 220932, "epoch": 2482, "lr": 8.334591935539527e-06}
{"train_loss": 0.0431521013379097, "global_step": 220933, "epoch": 2482, "lr": 8.334271454876213e-06}
{"train_loss": 0.019691620022058487, "global_step": 220934, "epoch": 2482, "lr": 8.333950979814359e-06}
{"train_loss": 0.04333583638072014, "global_step": 220935, "epoch": 2482, "lr": 8.33363051035399e-06}
{"train_loss": 0.03102990984916687, "global_step": 220936, "epoch": 2482, "lr": 8.33331004649517e-06}
{"train_loss": 0.030180448666214943, "global_step": 220937, "epoch": 2482, "lr": 8.33298958823791e-06}
{"train_loss": 0.1402159482240677, "global_step": 220938, "epoch": 2482, "lr": 8.332669135582288e-06}
{"train_loss": 0.05937560275197029, "global_step": 220939, "epoch": 2482, "lr": 8.332348688528318e-06}
{"train_loss": 0.061408981680870056, "global_step": 220940, "epoch": 2482, "lr": 8.332028247076057e-06}
{"train_loss": 0.03881920501589775, "global_step": 220941, "epoch": 2482, "lr": 8.331707811225554e-06}
{"train_loss": 0.051065992563962936, "global_step": 220942, "epoch": 2482, "lr": 8.331387380976835e-06}
{"train_loss": 0.05304914712905884, "global_step": 220943, "epoch": 2482, "lr": 8.33106695632997e-06}
{"train_loss": 0.09334025532007217, "global_step": 220944, "epoch": 2482, "lr": 8.330746537284968e-06}
{"train_loss": 0.0845780223608017, "global_step": 220945, "epoch": 2482, "lr": 8.330426123841906e-06}
{"train_loss": 0.053827784955501556, "global_step": 220946, "epoch": 2482, "lr": 8.330105716000796e-06}
{"train_loss": 0.028608743101358414, "global_step": 220947, "epoch": 2482, "lr": 8.329785313761707e-06}
{"train_loss": 0.059167563915252686, "global_step": 220948, "epoch": 2482, "lr": 8.329464917124658e-06}
{"train_loss": 0.039818037301301956, "global_step": 220949, "epoch": 2482, "lr": 8.329144526089716e-06}
{"train_loss": 0.03208913281559944, "global_step": 220950, "epoch": 2482, "lr": 8.328824140656898e-06}
{"train_loss": 0.046563901007175446, "global_step": 220951, "epoch": 2482, "lr": 8.328503760826279e-06}
{"train_loss": 0.05605991557240486, "global_step": 220952, "epoch": 2482, "lr": 8.328183386597871e-06}
{"train_loss": 0.053671564906835556, "global_step": 220953, "epoch": 2482, "lr": 8.327863017971743e-06}
{"train_loss": 0.05345830321311951, "global_step": 220954, "epoch": 2482, "lr": 8.327542654947912e-06}
{"train_loss": 0.03464145213365555, "global_step": 220955, "epoch": 2482, "lr": 8.327222297526449e-06}
{"train_loss": 0.11060135811567307, "global_step": 220956, "epoch": 2482, "lr": 8.326901945707382e-06}
{"train_loss": 0.07910497486591339, "global_step": 220957, "epoch": 2482, "lr": 8.326581599490745e-06}
{"train_loss": 0.07519719004631042, "global_step": 220958, "epoch": 2482, "lr": 8.326261258876605e-06}
{"train_loss": 0.04082528501749039, "global_step": 220959, "epoch": 2482, "lr": 8.325940923864977e-06}
{"train_loss": 0.06946947425603867, "global_step": 220960, "epoch": 2482, "lr": 8.32562059445593e-06}
{"train_loss": 0.07645465433597565, "global_step": 220961, "epoch": 2482, "lr": 8.325300270649483e-06}
{"train_loss": 0.04631757363677025, "global_step": 220962, "epoch": 2482, "lr": 8.3249799524457e-06}
{"train_loss": 0.02474041096866131, "global_step": 220963, "epoch": 2482, "lr": 8.324659639844607e-06}
{"train_loss": 0.024779006838798523, "global_step": 220964, "epoch": 2482, "lr": 8.32433933284627e-06}
{"train_loss": 0.047688648104667664, "global_step": 220965, "epoch": 2482, "lr": 8.324019031450703e-06}
{"train_loss": 0.03749196231365204, "global_step": 220966, "epoch": 2482, "lr": 8.323698735657975e-06}
{"train_loss": 0.08005130290985107, "global_step": 220967, "epoch": 2482, "lr": 8.323378445468105e-06}
{"train_loss": 0.045968711376190186, "global_step": 220968, "epoch": 2482, "lr": 8.323058160881164e-06}
{"train_loss": 0.06027050316333771, "global_step": 220969, "epoch": 2482, "lr": 8.322737881897164e-06}
{"train_loss": 0.07897508889436722, "global_step": 220970, "epoch": 2482, "lr": 8.32241760851618e-06}
{"train_loss": 0.06688027828931808, "global_step": 220971, "epoch": 2482, "lr": 8.322097340738222e-06}
{"train_loss": 0.03854936733841896, "global_step": 220972, "epoch": 2482, "lr": 8.321777078563369e-06}
{"train_loss": 0.07537122815847397, "global_step": 220973, "epoch": 2482, "lr": 8.32145682199163e-06}
{"train_loss": 0.024266904219985008, "global_step": 220974, "epoch": 2482, "lr": 8.321136571023074e-06}
{"train_loss": 0.047020405530929565, "global_step": 220975, "epoch": 2482, "lr": 8.320816325657721e-06}
{"train_loss": 0.03844211995601654, "global_step": 220976, "epoch": 2482, "lr": 8.320496085895624e-06}
{"train_loss": 0.08708040416240692, "global_step": 220977, "epoch": 2482, "lr": 8.320175851736845e-06}
{"train_loss": 0.07884105294942856, "global_step": 220978, "epoch": 2482, "lr": 8.319855623181394e-06}
{"train_loss": 0.0424857996404171, "global_step": 220979, "epoch": 2482, "lr": 8.319535400229345e-06}
{"train_loss": 0.03844974935054779, "global_step": 220980, "epoch": 2482, "lr": 8.319215182880713e-06}
{"train_loss": 0.04632001370191574, "global_step": 220981, "epoch": 2482, "lr": 8.31889497113557e-06}
{"train_loss": 0.05193757265806198, "global_step": 220982, "epoch": 2482, "lr": 8.318574764993931e-06}
{"train_loss": 0.026119524613022804, "global_step": 220983, "epoch": 2482, "lr": 8.318254564455863e-06}
{"train_loss": 0.046016301959753036, "global_step": 220984, "epoch": 2482, "lr": 8.31793436952138e-06}
{"train_loss": 0.05461619794368744, "global_step": 220985, "epoch": 2482, "lr": 8.317614180190564e-06}
{"train_loss": 0.05231732467001074, "global_step": 220986, "epoch": 2482, "lr": 8.317293996463415e-06, "val_loss": 8.968711853027344}
{"train_loss": 0.04459141194820404, "global_step": 220987, "epoch": 2483, "lr": 8.316973818340018e-06}
{"train_loss": 0.04997178539633751, "global_step": 220988, "epoch": 2483, "lr": 8.316653645820382e-06}
{"train_loss": 0.05698184296488762, "global_step": 220989, "epoch": 2483, "lr": 8.31633347890457e-06}
{"train_loss": 0.03669687360525131, "global_step": 220990, "epoch": 2483, "lr": 8.316013317592613e-06}
{"train_loss": 0.03740810975432396, "global_step": 220991, "epoch": 2483, "lr": 8.315693161884569e-06}
{"train_loss": 0.039985328912734985, "global_step": 220992, "epoch": 2483, "lr": 8.315373011780464e-06}
{"train_loss": 0.022299956530332565, "global_step": 220993, "epoch": 2483, "lr": 8.31505286728036e-06}
{"train_loss": 0.05850540101528168, "global_step": 220994, "epoch": 2483, "lr": 8.314732728384273e-06}
{"train_loss": 0.06882060319185257, "global_step": 220995, "epoch": 2483, "lr": 8.314412595092275e-06}
{"train_loss": 0.04763076454401016, "global_step": 220996, "epoch": 2483, "lr": 8.3140924674044e-06}
{"train_loss": 0.01623089797794819, "global_step": 220997, "epoch": 2483, "lr": 8.313772345320669e-06}
{"train_loss": 0.023833895102143288, "global_step": 220998, "epoch": 2483, "lr": 8.313452228841161e-06}
{"train_loss": 0.047034718096256256, "global_step": 220999, "epoch": 2483, "lr": 8.313132117965888e-06}
{"train_loss": 0.04548971727490425, "global_step": 221000, "epoch": 2483, "lr": 8.31281201269492e-06}
{"train_loss": 0.06540610641241074, "global_step": 221001, "epoch": 2483, "lr": 8.31249191302827e-06}
{"train_loss": 0.03353636711835861, "global_step": 221002, "epoch": 2483, "lr": 8.312171818966013e-06}
{"train_loss": 0.028094535693526268, "global_step": 221003, "epoch": 2483, "lr": 8.311851730508163e-06}
{"train_loss": 0.05785288289189339, "global_step": 221004, "epoch": 2483, "lr": 8.311531647654786e-06}
{"train_loss": 0.08425456285476685, "global_step": 221005, "epoch": 2483, "lr": 8.311211570405908e-06}
{"train_loss": 0.038975585252046585, "global_step": 221006, "epoch": 2483, "lr": 8.310891498761586e-06}
{"train_loss": 0.06563351303339005, "global_step": 221007, "epoch": 2483, "lr": 8.310571432721847e-06}
{"train_loss": 0.05238901823759079, "global_step": 221008, "epoch": 2483, "lr": 8.310251372286759e-06}
{"train_loss": 0.06128470227122307, "global_step": 221009, "epoch": 2483, "lr": 8.309931317456332e-06}
{"train_loss": 0.030522311106324196, "global_step": 221010, "epoch": 2483, "lr": 8.309611268230644e-06}
{"train_loss": 0.02089666575193405, "global_step": 221011, "epoch": 2483, "lr": 8.309291224609706e-06}
{"train_loss": 0.0408070869743824, "global_step": 221012, "epoch": 2483, "lr": 8.308971186593573e-06}
{"train_loss": 0.03835871443152428, "global_step": 221013, "epoch": 2483, "lr": 8.308651154182307e-06}
{"train_loss": 0.03835944086313248, "global_step": 221014, "epoch": 2483, "lr": 8.308331127375923e-06}
{"train_loss": 0.06418532133102417, "global_step": 221015, "epoch": 2483, "lr": 8.308011106174485e-06}
{"train_loss": 0.05022144690155983, "global_step": 221016, "epoch": 2483, "lr": 8.307691090578012e-06}
{"train_loss": 0.05217251926660538, "global_step": 221017, "epoch": 2483, "lr": 8.30737108058658e-06}
{"train_loss": 0.010599896311759949, "global_step": 221018, "epoch": 2483, "lr": 8.3070510762002e-06}
{"train_loss": 0.038394030183553696, "global_step": 221019, "epoch": 2483, "lr": 8.30673107741894e-06}
{"train_loss": 0.021245906129479408, "global_step": 221020, "epoch": 2483, "lr": 8.306411084242816e-06}
{"train_loss": 0.0493113175034523, "global_step": 221021, "epoch": 2483, "lr": 8.306091096671903e-06}
{"train_loss": 0.03303562104701996, "global_step": 221022, "epoch": 2483, "lr": 8.305771114706218e-06}
{"train_loss": 0.0437217615544796, "global_step": 221023, "epoch": 2483, "lr": 8.305451138345821e-06}
{"train_loss": 0.03478604927659035, "global_step": 221024, "epoch": 2483, "lr": 8.30513116759074e-06}
{"train_loss": 0.06774095445871353, "global_step": 221025, "epoch": 2483, "lr": 8.304811202441038e-06}
{"train_loss": 0.038672368973493576, "global_step": 221026, "epoch": 2483, "lr": 8.304491242896728e-06}
{"train_loss": 0.04223144054412842, "global_step": 221027, "epoch": 2483, "lr": 8.30417128895789e-06}
{"train_loss": 0.043544430285692215, "global_step": 221028, "epoch": 2483, "lr": 8.30385134062453e-06}
{"train_loss": 0.027588382363319397, "global_step": 221029, "epoch": 2483, "lr": 8.303531397896725e-06}
{"train_loss": 0.02919015847146511, "global_step": 221030, "epoch": 2483, "lr": 8.30321146077449e-06}
{"train_loss": 0.05730438977479935, "global_step": 221031, "epoch": 2483, "lr": 8.302891529257888e-06}
{"train_loss": 0.07231780141592026, "global_step": 221032, "epoch": 2483, "lr": 8.302571603346948e-06}
{"train_loss": 0.07948018610477448, "global_step": 221033, "epoch": 2483, "lr": 8.302251683041729e-06}
{"train_loss": 0.0560389943420887, "global_step": 221034, "epoch": 2483, "lr": 8.301931768342264e-06}
{"train_loss": 0.04073616489768028, "global_step": 221035, "epoch": 2483, "lr": 8.30161185924858e-06}
{"train_loss": 0.023752061650156975, "global_step": 221036, "epoch": 2483, "lr": 8.301291955760754e-06}
{"train_loss": 0.047498010098934174, "global_step": 221037, "epoch": 2483, "lr": 8.300972057878797e-06}
{"train_loss": 0.09571093320846558, "global_step": 221038, "epoch": 2483, "lr": 8.300652165602773e-06}
{"train_loss": 0.054996199905872345, "global_step": 221039, "epoch": 2483, "lr": 8.300332278932715e-06}
{"train_loss": 0.035702552646398544, "global_step": 221040, "epoch": 2483, "lr": 8.300012397868673e-06}
{"train_loss": 0.07494276016950607, "global_step": 221041, "epoch": 2483, "lr": 8.29969252241068e-06}
{"train_loss": 0.03437164053320885, "global_step": 221042, "epoch": 2483, "lr": 8.299372652558796e-06}
{"train_loss": 0.024692168459296227, "global_step": 221043, "epoch": 2483, "lr": 8.299052788313039e-06}
{"train_loss": 0.057563554495573044, "global_step": 221044, "epoch": 2483, "lr": 8.298732929673474e-06}
{"train_loss": 0.0867062360048294, "global_step": 221045, "epoch": 2483, "lr": 8.298413076640133e-06}
{"train_loss": 0.04646714776754379, "global_step": 221046, "epoch": 2483, "lr": 8.298093229213072e-06}
{"train_loss": 0.05109025910496712, "global_step": 221047, "epoch": 2483, "lr": 8.29777338739231e-06}
{"train_loss": 0.04189891740679741, "global_step": 221048, "epoch": 2483, "lr": 8.297453551177903e-06}
{"train_loss": 0.052004825323820114, "global_step": 221049, "epoch": 2483, "lr": 8.297133720569906e-06}
{"train_loss": 0.0639888122677803, "global_step": 221050, "epoch": 2483, "lr": 8.296813895568346e-06}
{"train_loss": 0.04728555306792259, "global_step": 221051, "epoch": 2483, "lr": 8.296494076173273e-06}
{"train_loss": 0.04736480861902237, "global_step": 221052, "epoch": 2483, "lr": 8.296174262384726e-06}
{"train_loss": 0.07609730213880539, "global_step": 221053, "epoch": 2483, "lr": 8.295854454202757e-06}
{"train_loss": 0.04249128699302673, "global_step": 221054, "epoch": 2483, "lr": 8.29553465162739e-06}
{"train_loss": 0.028071319684386253, "global_step": 221055, "epoch": 2483, "lr": 8.29521485465869e-06}
{"train_loss": 0.07697946578264236, "global_step": 221056, "epoch": 2483, "lr": 8.294895063296682e-06}
{"train_loss": 0.018132414668798447, "global_step": 221057, "epoch": 2483, "lr": 8.294575277541428e-06}
{"train_loss": 0.017500698566436768, "global_step": 221058, "epoch": 2483, "lr": 8.29425549739295e-06}
{"train_loss": 0.03732086718082428, "global_step": 221059, "epoch": 2483, "lr": 8.29393572285131e-06}
{"train_loss": 0.029686547815799713, "global_step": 221060, "epoch": 2483, "lr": 8.293615953916529e-06}
{"train_loss": 0.11143608391284943, "global_step": 221061, "epoch": 2483, "lr": 8.29329619058868e-06}
{"train_loss": 0.05074722319841385, "global_step": 221062, "epoch": 2483, "lr": 8.292976432867771e-06}
{"train_loss": 0.02404705062508583, "global_step": 221063, "epoch": 2483, "lr": 8.29265668075388e-06}
{"train_loss": 0.0512695237994194, "global_step": 221064, "epoch": 2483, "lr": 8.292336934247019e-06}
{"train_loss": 0.03647463768720627, "global_step": 221065, "epoch": 2483, "lr": 8.292017193347257e-06}
{"train_loss": 0.11750341206789017, "global_step": 221066, "epoch": 2483, "lr": 8.291697458054609e-06}
{"train_loss": 0.036779336631298065, "global_step": 221067, "epoch": 2483, "lr": 8.291377728369154e-06}
{"train_loss": 0.06595724821090698, "global_step": 221068, "epoch": 2483, "lr": 8.291058004290897e-06}
{"train_loss": 0.039750900119543076, "global_step": 221069, "epoch": 2483, "lr": 8.29073828581991e-06}
{"train_loss": 0.060417722910642624, "global_step": 221070, "epoch": 2483, "lr": 8.290418572956215e-06}
{"train_loss": 0.06325843185186386, "global_step": 221071, "epoch": 2483, "lr": 8.29009886569988e-06}
{"train_loss": 0.03682096302509308, "global_step": 221072, "epoch": 2483, "lr": 8.289779164050931e-06}
{"train_loss": 0.014984878711402416, "global_step": 221073, "epoch": 2483, "lr": 8.289459468009397e-06}
{"train_loss": 0.06422900408506393, "global_step": 221074, "epoch": 2483, "lr": 8.289139777575356e-06}
{"train_loss": 0.04778564751692367, "global_step": 221075, "epoch": 2483, "lr": 8.28882009274881e-06, "val_loss": 8.857232093811035}
{"train_loss": 0.02495536394417286, "global_step": 221076, "epoch": 2484, "lr": 8.288500413529842e-06}
{"train_loss": 0.02646172046661377, "global_step": 221077, "epoch": 2484, "lr": 8.288180739918467e-06}
{"train_loss": 0.03683597967028618, "global_step": 221078, "epoch": 2484, "lr": 8.287861071914743e-06}
{"train_loss": 0.04382510483264923, "global_step": 221079, "epoch": 2484, "lr": 8.2875414095187e-06}
{"train_loss": 0.03782062977552414, "global_step": 221080, "epoch": 2484, "lr": 8.287221752730401e-06}
{"train_loss": 0.06472200900316238, "global_step": 221081, "epoch": 2484, "lr": 8.286902101549864e-06}
{"train_loss": 0.04517766833305359, "global_step": 221082, "epoch": 2484, "lr": 8.286582455977155e-06}
{"train_loss": 0.025400016456842422, "global_step": 221083, "epoch": 2484, "lr": 8.286262816012297e-06}
{"train_loss": 0.06692083925008774, "global_step": 221084, "epoch": 2484, "lr": 8.285943181655342e-06}
{"train_loss": 0.09825482219457626, "global_step": 221085, "epoch": 2484, "lr": 8.28562355290634e-06}
{"train_loss": 0.049241337925195694, "global_step": 221086, "epoch": 2484, "lr": 8.285303929765325e-06}
{"train_loss": 0.06445812433958054, "global_step": 221087, "epoch": 2484, "lr": 8.284984312232347e-06}
{"train_loss": 0.02308298833668232, "global_step": 221088, "epoch": 2484, "lr": 8.284664700307437e-06}
{"train_loss": 0.049297865480184555, "global_step": 221089, "epoch": 2484, "lr": 8.284345093990653e-06}
{"train_loss": 0.03617413714528084, "global_step": 221090, "epoch": 2484, "lr": 8.284025493282021e-06}
{"train_loss": 0.05945682153105736, "global_step": 221091, "epoch": 2484, "lr": 8.28370589818161e-06}
{"train_loss": 0.05546553060412407, "global_step": 221092, "epoch": 2484, "lr": 8.283386308689427e-06}
{"train_loss": 0.06721001863479614, "global_step": 221093, "epoch": 2484, "lr": 8.28306672480555e-06}
{"train_loss": 0.07778770476579666, "global_step": 221094, "epoch": 2484, "lr": 8.28274714652999e-06}
{"train_loss": 0.029720934107899666, "global_step": 221095, "epoch": 2484, "lr": 8.282427573862822e-06}
{"train_loss": 0.0581757128238678, "global_step": 221096, "epoch": 2484, "lr": 8.282108006804062e-06}
{"train_loss": 0.045424435287714005, "global_step": 221097, "epoch": 2484, "lr": 8.281788445353772e-06}
{"train_loss": 0.02245664782822132, "global_step": 221098, "epoch": 2484, "lr": 8.28146888951198e-06}
{"train_loss": 0.0384228453040123, "global_step": 221099, "epoch": 2484, "lr": 8.281149339278744e-06}
{"train_loss": 0.09033914655447006, "global_step": 221100, "epoch": 2484, "lr": 8.280829794654088e-06}
{"train_loss": 0.02241804264485836, "global_step": 221101, "epoch": 2484, "lr": 8.28051025563808e-06}
{"train_loss": 0.07634742558002472, "global_step": 221102, "epoch": 2484, "lr": 8.280190722230735e-06}
{"train_loss": 0.05209344998002052, "global_step": 221103, "epoch": 2484, "lr": 8.27987119443211e-06}
{"train_loss": 0.06352770328521729, "global_step": 221104, "epoch": 2484, "lr": 8.279551672242258e-06}
{"train_loss": 0.04850909858942032, "global_step": 221105, "epoch": 2484, "lr": 8.279232155661204e-06}
{"train_loss": 0.013635316863656044, "global_step": 221106, "epoch": 2484, "lr": 8.278912644689008e-06}
{"train_loss": 0.043005745857954025, "global_step": 221107, "epoch": 2484, "lr": 8.278593139325708e-06}
{"train_loss": 0.04389133304357529, "global_step": 221108, "epoch": 2484, "lr": 8.278273639571322e-06}
{"train_loss": 0.07460834085941315, "global_step": 221109, "epoch": 2484, "lr": 8.277954145425932e-06}
{"train_loss": 0.07502398639917374, "global_step": 221110, "epoch": 2484, "lr": 8.27763465688955e-06}
{"train_loss": 0.020968196913599968, "global_step": 221111, "epoch": 2484, "lr": 8.277315173962242e-06}
{"train_loss": 0.06347859650850296, "global_step": 221112, "epoch": 2484, "lr": 8.276995696644042e-06}
{"train_loss": 0.10889054089784622, "global_step": 221113, "epoch": 2484, "lr": 8.276676224934975e-06}
{"train_loss": 0.02960742637515068, "global_step": 221114, "epoch": 2484, "lr": 8.276356758835119e-06}
{"train_loss": 0.07523075491189957, "global_step": 221115, "epoch": 2484, "lr": 8.27603729834448e-06}
{"train_loss": 0.07945188879966736, "global_step": 221116, "epoch": 2484, "lr": 8.275717843463138e-06}
{"train_loss": 0.04986786097288132, "global_step": 221117, "epoch": 2484, "lr": 8.275398394191097e-06}
{"train_loss": 0.05485263466835022, "global_step": 221118, "epoch": 2484, "lr": 8.275078950528442e-06}
{"train_loss": 0.04363853111863136, "global_step": 221119, "epoch": 2484, "lr": 8.274759512475172e-06}
{"train_loss": 0.06911411881446838, "global_step": 221120, "epoch": 2484, "lr": 8.274440080031358e-06}
{"train_loss": 0.06655925512313843, "global_step": 221121, "epoch": 2484, "lr": 8.274120653197049e-06}
{"train_loss": 0.021124808117747307, "global_step": 221122, "epoch": 2484, "lr": 8.273801231972261e-06}
{"train_loss": 0.06116926670074463, "global_step": 221123, "epoch": 2484, "lr": 8.273481816357066e-06}
{"train_loss": 0.0415012501180172, "global_step": 221124, "epoch": 2484, "lr": 8.273162406351481e-06}
{"train_loss": 0.05540229380130768, "global_step": 221125, "epoch": 2484, "lr": 8.272843001955571e-06}
{"train_loss": 0.05374917760491371, "global_step": 221126, "epoch": 2484, "lr": 8.272523603169357e-06}
{"train_loss": 0.03236258029937744, "global_step": 221127, "epoch": 2484, "lr": 8.272204209992906e-06}
{"train_loss": 0.019077399745583534, "global_step": 221128, "epoch": 2484, "lr": 8.27188482242624e-06}
{"train_loss": 0.07950856536626816, "global_step": 221129, "epoch": 2484, "lr": 8.27156544046942e-06}
{"train_loss": 0.05379556119441986, "global_step": 221130, "epoch": 2484, "lr": 8.271246064122461e-06}
{"train_loss": 0.052251994609832764, "global_step": 221131, "epoch": 2484, "lr": 8.270926693385445e-06}
{"train_loss": 0.07690747082233429, "global_step": 221132, "epoch": 2484, "lr": 8.270607328258378e-06}
{"train_loss": 0.0337512381374836, "global_step": 221133, "epoch": 2484, "lr": 8.270287968741337e-06}
{"train_loss": 0.05582347512245178, "global_step": 221134, "epoch": 2484, "lr": 8.269968614834328e-06}
{"train_loss": 0.05043473839759827, "global_step": 221135, "epoch": 2484, "lr": 8.269649266537426e-06}
{"train_loss": 0.042623862624168396, "global_step": 221136, "epoch": 2484, "lr": 8.269329923850655e-06}
{"train_loss": 0.055754560977220535, "global_step": 221137, "epoch": 2484, "lr": 8.269010586774067e-06}
{"train_loss": 0.04472247138619423, "global_step": 221138, "epoch": 2484, "lr": 8.268691255307699e-06}
{"train_loss": 0.035516515374183655, "global_step": 221139, "epoch": 2484, "lr": 8.26837192945159e-06}
{"train_loss": 0.03091098740696907, "global_step": 221140, "epoch": 2484, "lr": 8.268052609205806e-06}
{"train_loss": 0.034007664769887924, "global_step": 221141, "epoch": 2484, "lr": 8.267733294570357e-06}
{"train_loss": 0.09079276025295258, "global_step": 221142, "epoch": 2484, "lr": 8.26741398554532e-06}
{"train_loss": 0.06776434183120728, "global_step": 221143, "epoch": 2484, "lr": 8.267094682130706e-06}
{"train_loss": 0.051068712025880814, "global_step": 221144, "epoch": 2484, "lr": 8.266775384326585e-06}
{"train_loss": 0.028274599462747574, "global_step": 221145, "epoch": 2484, "lr": 8.266456092132984e-06}
{"train_loss": 0.04785749688744545, "global_step": 221146, "epoch": 2484, "lr": 8.266136805549946e-06}
{"train_loss": 0.032795581966638565, "global_step": 221147, "epoch": 2484, "lr": 8.265817524577524e-06}
{"train_loss": 0.016661280766129494, "global_step": 221148, "epoch": 2484, "lr": 8.265498249215737e-06}
{"train_loss": 0.028960686177015305, "global_step": 221149, "epoch": 2484, "lr": 8.265178979464666e-06}
{"train_loss": 0.10869492590427399, "global_step": 221150, "epoch": 2484, "lr": 8.264859715324324e-06}
{"train_loss": 0.036237865686416626, "global_step": 221151, "epoch": 2484, "lr": 8.264540456794756e-06}
{"train_loss": 0.10769034177064896, "global_step": 221152, "epoch": 2484, "lr": 8.264221203876026e-06}
{"train_loss": 0.0809965506196022, "global_step": 221153, "epoch": 2484, "lr": 8.263901956568148e-06}
{"train_loss": 0.1304631382226944, "global_step": 221154, "epoch": 2484, "lr": 8.263582714871193e-06}
{"train_loss": 0.04455173388123512, "global_step": 221155, "epoch": 2484, "lr": 8.263263478785177e-06}
{"train_loss": 0.051594749093055725, "global_step": 221156, "epoch": 2484, "lr": 8.262944248310172e-06}
{"train_loss": 0.05931901931762695, "global_step": 221157, "epoch": 2484, "lr": 8.262625023446192e-06}
{"train_loss": 0.05474228411912918, "global_step": 221158, "epoch": 2484, "lr": 8.262305804193288e-06}
{"train_loss": 0.06539111584424973, "global_step": 221159, "epoch": 2484, "lr": 8.261986590551523e-06}
{"train_loss": 0.06081760302186012, "global_step": 221160, "epoch": 2484, "lr": 8.261667382520916e-06}
{"train_loss": 0.09045607596635818, "global_step": 221161, "epoch": 2484, "lr": 8.26134818010153e-06}
{"train_loss": 0.1172635480761528, "global_step": 221162, "epoch": 2484, "lr": 8.261028983293383e-06}
{"train_loss": 0.09695839881896973, "global_step": 221163, "epoch": 2484, "lr": 8.260709792096549e-06}
{"train_loss": 0.05452723847178931, "global_step": 221164, "epoch": 2484, "lr": 8.260390606511038e-06, "val_loss": 8.850369453430176}
{"train_loss": 0.036141373217105865, "global_step": 221165, "epoch": 2485, "lr": 8.26007142653692e-06}
{"train_loss": 0.044830139726400375, "global_step": 221166, "epoch": 2485, "lr": 8.259752252174213e-06}
{"train_loss": 0.046732183545827866, "global_step": 221167, "epoch": 2485, "lr": 8.25943308342299e-06}
{"train_loss": 0.06794213503599167, "global_step": 221168, "epoch": 2485, "lr": 8.259113920283263e-06}
{"train_loss": 0.022493235766887665, "global_step": 221169, "epoch": 2485, "lr": 8.258794762755107e-06}
{"train_loss": 0.050940535962581635, "global_step": 221170, "epoch": 2485, "lr": 8.258475610838529e-06}
{"train_loss": 0.059454888105392456, "global_step": 221171, "epoch": 2485, "lr": 8.2581564645336e-06}
{"train_loss": 0.028553225100040436, "global_step": 221172, "epoch": 2485, "lr": 8.257837323840346e-06}
{"train_loss": 0.05744687840342522, "global_step": 221173, "epoch": 2485, "lr": 8.257518188758828e-06}
{"train_loss": 0.07042019814252853, "global_step": 221174, "epoch": 2485, "lr": 8.257199059289067e-06}
{"train_loss": 0.0881999060511589, "global_step": 221175, "epoch": 2485, "lr": 8.256879935431116e-06}
{"train_loss": 0.06520531326532364, "global_step": 221176, "epoch": 2485, "lr": 8.256560817185028e-06}
{"train_loss": 0.043824825435876846, "global_step": 221177, "epoch": 2485, "lr": 8.256241704550826e-06}
{"train_loss": 0.04899745434522629, "global_step": 221178, "epoch": 2485, "lr": 8.255922597528581e-06}
{"train_loss": 0.040139712393283844, "global_step": 221179, "epoch": 2485, "lr": 8.255603496118298e-06}
{"train_loss": 0.025508882477879524, "global_step": 221180, "epoch": 2485, "lr": 8.255284400320056e-06}
{"train_loss": 0.07153070718050003, "global_step": 221181, "epoch": 2485, "lr": 8.254965310133872e-06}
{"train_loss": 0.020557641983032227, "global_step": 221182, "epoch": 2485, "lr": 8.254646225559808e-06}
{"train_loss": 0.038420744240283966, "global_step": 221183, "epoch": 2485, "lr": 8.254327146597902e-06}
{"train_loss": 0.03739418461918831, "global_step": 221184, "epoch": 2485, "lr": 8.254008073248176e-06}
{"train_loss": 0.032506268471479416, "global_step": 221185, "epoch": 2485, "lr": 8.253689005510707e-06}
{"train_loss": 0.03499144688248634, "global_step": 221186, "epoch": 2485, "lr": 8.253369943385503e-06}
{"train_loss": 0.06502054631710052, "global_step": 221187, "epoch": 2485, "lr": 8.253050886872643e-06}
{"train_loss": 0.06153039634227753, "global_step": 221188, "epoch": 2485, "lr": 8.252731835972143e-06}
{"train_loss": 0.09044957160949707, "global_step": 221189, "epoch": 2485, "lr": 8.25241279068405e-06}
{"train_loss": 0.07884817570447922, "global_step": 221190, "epoch": 2485, "lr": 8.252093751008421e-06}
{"train_loss": 0.07342123985290527, "global_step": 221191, "epoch": 2485, "lr": 8.251774716945277e-06}
{"train_loss": 0.024608064442873, "global_step": 221192, "epoch": 2485, "lr": 8.251455688494686e-06}
{"train_loss": 0.04433966800570488, "global_step": 221193, "epoch": 2485, "lr": 8.25113666565666e-06}
{"train_loss": 0.02723010443150997, "global_step": 221194, "epoch": 2485, "lr": 8.250817648431269e-06}
{"train_loss": 0.06953099370002747, "global_step": 221195, "epoch": 2485, "lr": 8.250498636818555e-06}
{"train_loss": 0.0987594872713089, "global_step": 221196, "epoch": 2485, "lr": 8.250179630818538e-06}
{"train_loss": 0.0486849807202816, "global_step": 221197, "epoch": 2485, "lr": 8.249860630431289e-06}
{"train_loss": 0.10608480870723724, "global_step": 221198, "epoch": 2485, "lr": 8.249541635656827e-06}
{"train_loss": 0.013018532656133175, "global_step": 221199, "epoch": 2485, "lr": 8.249222646495214e-06}
{"train_loss": 0.05456184968352318, "global_step": 221200, "epoch": 2485, "lr": 8.248903662946477e-06}
{"train_loss": 0.0525868684053421, "global_step": 221201, "epoch": 2485, "lr": 8.248584685010675e-06}
{"train_loss": 0.04297298938035965, "global_step": 221202, "epoch": 2485, "lr": 8.248265712687831e-06}
{"train_loss": 0.04141616076231003, "global_step": 221203, "epoch": 2485, "lr": 8.247946745978009e-06}
{"train_loss": 0.06775274127721786, "global_step": 221204, "epoch": 2485, "lr": 8.247627784881228e-06}
{"train_loss": 0.11876501142978668, "global_step": 221205, "epoch": 2485, "lr": 8.24730882939756e-06}
{"train_loss": 0.045114051550626755, "global_step": 221206, "epoch": 2485, "lr": 8.246989879527022e-06}
{"train_loss": 0.01931612938642502, "global_step": 221207, "epoch": 2485, "lr": 8.246670935269674e-06}
{"train_loss": 0.030569061636924744, "global_step": 221208, "epoch": 2485, "lr": 8.246351996625546e-06}
{"train_loss": 0.03907281160354614, "global_step": 221209, "epoch": 2485, "lr": 8.246033063594694e-06}
{"train_loss": 0.015079102478921413, "global_step": 221210, "epoch": 2485, "lr": 8.245714136177146e-06}
{"train_loss": 0.0429806262254715, "global_step": 221211, "epoch": 2485, "lr": 8.245395214372947e-06}
{"train_loss": 0.07126883417367935, "global_step": 221212, "epoch": 2485, "lr": 8.245076298182163e-06}
{"train_loss": 0.04518129304051399, "global_step": 221213, "epoch": 2485, "lr": 8.244757387604806e-06}
{"train_loss": 0.04750398173928261, "global_step": 221214, "epoch": 2485, "lr": 8.244438482640948e-06}
{"train_loss": 0.05922802910208702, "global_step": 221215, "epoch": 2485, "lr": 8.2441195832906e-06}
{"train_loss": 0.07180243730545044, "global_step": 221216, "epoch": 2485, "lr": 8.243800689553833e-06}
{"train_loss": 0.062463052570819855, "global_step": 221217, "epoch": 2485, "lr": 8.24348180143067e-06}
{"train_loss": 0.08794883638620377, "global_step": 221218, "epoch": 2485, "lr": 8.243162918921166e-06}
{"train_loss": 0.07821106165647507, "global_step": 221219, "epoch": 2485, "lr": 8.242844042025355e-06}
{"train_loss": 0.062351107597351074, "global_step": 221220, "epoch": 2485, "lr": 8.242525170743293e-06}
{"train_loss": 0.11674981564283371, "global_step": 221221, "epoch": 2485, "lr": 8.242206305075006e-06}
{"train_loss": 0.039442334324121475, "global_step": 221222, "epoch": 2485, "lr": 8.241887445020557e-06}
{"train_loss": 0.07251516729593277, "global_step": 221223, "epoch": 2485, "lr": 8.241568590579973e-06}
{"train_loss": 0.10197807848453522, "global_step": 221224, "epoch": 2485, "lr": 8.241249741753293e-06}
{"train_loss": 0.03002658113837242, "global_step": 221225, "epoch": 2485, "lr": 8.240930898540578e-06}
{"train_loss": 0.03450296074151993, "global_step": 221226, "epoch": 2485, "lr": 8.240612060941849e-06}
{"train_loss": 0.04785900563001633, "global_step": 221227, "epoch": 2485, "lr": 8.240293228957174e-06}
{"train_loss": 0.03994449973106384, "global_step": 221228, "epoch": 2485, "lr": 8.23997440258658e-06}
{"train_loss": 0.06468069553375244, "global_step": 221229, "epoch": 2485, "lr": 8.239655581830108e-06}
{"train_loss": 0.04060446470975876, "global_step": 221230, "epoch": 2485, "lr": 8.239336766687799e-06}
{"train_loss": 0.05586845800280571, "global_step": 221231, "epoch": 2485, "lr": 8.23901795715971e-06}
{"train_loss": 0.04497820511460304, "global_step": 221232, "epoch": 2485, "lr": 8.238699153245872e-06}
{"train_loss": 0.013824797235429287, "global_step": 221233, "epoch": 2485, "lr": 8.23838035494634e-06}
{"train_loss": 0.06991030275821686, "global_step": 221234, "epoch": 2485, "lr": 8.23806156226114e-06}
{"train_loss": 0.06459631770849228, "global_step": 221235, "epoch": 2485, "lr": 8.237742775190332e-06}
{"train_loss": 0.03295865282416344, "global_step": 221236, "epoch": 2485, "lr": 8.237423993733939e-06}
{"train_loss": 0.04307520389556885, "global_step": 221237, "epoch": 2485, "lr": 8.237105217892032e-06}
{"train_loss": 0.02842574566602707, "global_step": 221238, "epoch": 2485, "lr": 8.236786447664618e-06}
{"train_loss": 0.0319625549018383, "global_step": 221239, "epoch": 2485, "lr": 8.236467683051774e-06}
{"train_loss": 0.08708030730485916, "global_step": 221240, "epoch": 2485, "lr": 8.236148924053516e-06}
{"train_loss": 0.057720962911844254, "global_step": 221241, "epoch": 2485, "lr": 8.23583017066991e-06}
{"train_loss": 0.05396324768662453, "global_step": 221242, "epoch": 2485, "lr": 8.235511422900982e-06}
{"train_loss": 0.053579699248075485, "global_step": 221243, "epoch": 2485, "lr": 8.235192680746783e-06}
{"train_loss": 0.05736441910266876, "global_step": 221244, "epoch": 2485, "lr": 8.23487394420735e-06}
{"train_loss": 0.04366721585392952, "global_step": 221245, "epoch": 2485, "lr": 8.234555213282734e-06}
{"train_loss": 0.04184770584106445, "global_step": 221246, "epoch": 2485, "lr": 8.234236487972962e-06}
{"train_loss": 0.03678519278764725, "global_step": 221247, "epoch": 2485, "lr": 8.233917768278093e-06}
{"train_loss": 0.06983211636543274, "global_step": 221248, "epoch": 2485, "lr": 8.233599054198171e-06}
{"train_loss": 0.029581909999251366, "global_step": 221249, "epoch": 2485, "lr": 8.233280345733225e-06}
{"train_loss": 0.05213446915149689, "global_step": 221250, "epoch": 2485, "lr": 8.232961642883313e-06}
{"train_loss": 0.06600506603717804, "global_step": 221251, "epoch": 2485, "lr": 8.232642945648461e-06}
{"train_loss": 0.023579727858304977, "global_step": 221252, "epoch": 2485, "lr": 8.232324254028733e-06}
{"train_loss": 0.052245797338278104, "global_step": 221253, "epoch": 2485, "lr": 8.232005568024149e-06, "val_loss": 8.962472915649414, "train_action_mse_error": 8.19517993927002}
{"train_loss": 0.05398879200220108, "global_step": 221254, "epoch": 2486, "lr": 8.231686887634771e-06}
{"train_loss": 0.033578768372535706, "global_step": 221255, "epoch": 2486, "lr": 8.231368212860624e-06}
{"train_loss": 0.03030303679406643, "global_step": 221256, "epoch": 2486, "lr": 8.231049543701769e-06}
{"train_loss": 0.043572306632995605, "global_step": 221257, "epoch": 2486, "lr": 8.230730880158233e-06}
{"train_loss": 0.03375813737511635, "global_step": 221258, "epoch": 2486, "lr": 8.230412222230078e-06}
{"train_loss": 0.03407479450106621, "global_step": 221259, "epoch": 2486, "lr": 8.230093569917318e-06}
{"train_loss": 0.04767877981066704, "global_step": 221260, "epoch": 2486, "lr": 8.22977492322003e-06}
{"train_loss": 0.04786965996026993, "global_step": 221261, "epoch": 2486, "lr": 8.229456282138238e-06}
{"train_loss": 0.05377732962369919, "global_step": 221262, "epoch": 2486, "lr": 8.229137646671974e-06}
{"train_loss": 0.03660571202635765, "global_step": 221263, "epoch": 2486, "lr": 8.22881901682131e-06}
{"train_loss": 0.031959254294633865, "global_step": 221264, "epoch": 2486, "lr": 8.228500392586253e-06}
{"train_loss": 0.06569399684667587, "global_step": 221265, "epoch": 2486, "lr": 8.22818177396688e-06}
{"train_loss": 0.0547165647149086, "global_step": 221266, "epoch": 2486, "lr": 8.227863160963206e-06}
{"train_loss": 0.03671761602163315, "global_step": 221267, "epoch": 2486, "lr": 8.2275445535753e-06}
{"train_loss": 0.07529744505882263, "global_step": 221268, "epoch": 2486, "lr": 8.22722595180318e-06}
{"train_loss": 0.027176758274435997, "global_step": 221269, "epoch": 2486, "lr": 8.226907355646912e-06}
{"train_loss": 0.077010877430439, "global_step": 221270, "epoch": 2486, "lr": 8.226588765106513e-06}
{"train_loss": 0.06447567045688629, "global_step": 221271, "epoch": 2486, "lr": 8.226270180182055e-06}
{"train_loss": 0.07201720029115677, "global_step": 221272, "epoch": 2486, "lr": 8.225951600873555e-06}
{"train_loss": 0.11735653877258301, "global_step": 221273, "epoch": 2486, "lr": 8.225633027181074e-06}
{"train_loss": 0.10630085319280624, "global_step": 221274, "epoch": 2486, "lr": 8.225314459104638e-06}
{"train_loss": 0.05278472602367401, "global_step": 221275, "epoch": 2486, "lr": 8.22499589664431e-06}
{"train_loss": 0.056430377066135406, "global_step": 221276, "epoch": 2486, "lr": 8.224677339800113e-06}
{"train_loss": 0.05884520336985588, "global_step": 221277, "epoch": 2486, "lr": 8.224358788572106e-06}
{"train_loss": 0.05214231461286545, "global_step": 221278, "epoch": 2486, "lr": 8.224040242960317e-06}
{"train_loss": 0.03890157863497734, "global_step": 221279, "epoch": 2486, "lr": 8.223721702964804e-06}
{"train_loss": 0.019076287746429443, "global_step": 221280, "epoch": 2486, "lr": 8.223403168585597e-06}
{"train_loss": 0.04509421065449715, "global_step": 221281, "epoch": 2486, "lr": 8.223084639822753e-06}
{"train_loss": 0.045719318091869354, "global_step": 221282, "epoch": 2486, "lr": 8.222766116676295e-06}
{"train_loss": 0.023032788187265396, "global_step": 221283, "epoch": 2486, "lr": 8.222447599146288e-06}
{"train_loss": 0.05115791782736778, "global_step": 221284, "epoch": 2486, "lr": 8.22212908723275e-06}
{"train_loss": 0.032736606895923615, "global_step": 221285, "epoch": 2486, "lr": 8.22181058093574e-06}
{"train_loss": 0.06632313132286072, "global_step": 221286, "epoch": 2486, "lr": 8.22149208025531e-06}
{"train_loss": 0.09892365336418152, "global_step": 221287, "epoch": 2486, "lr": 8.221173585191478e-06}
{"train_loss": 0.038550280034542084, "global_step": 221288, "epoch": 2486, "lr": 8.220855095744311e-06}
{"train_loss": 0.04103297367691994, "global_step": 221289, "epoch": 2486, "lr": 8.220536611913832e-06}
{"train_loss": 0.0644102394580841, "global_step": 221290, "epoch": 2486, "lr": 8.220218133700102e-06}
{"train_loss": 0.024151140823960304, "global_step": 221291, "epoch": 2486, "lr": 8.219899661103142e-06}
{"train_loss": 0.06561937928199768, "global_step": 221292, "epoch": 2486, "lr": 8.219581194123022e-06}
{"train_loss": 0.04562230780720711, "global_step": 221293, "epoch": 2486, "lr": 8.219262732759758e-06}
{"train_loss": 0.013523049652576447, "global_step": 221294, "epoch": 2486, "lr": 8.218944277013418e-06}
{"train_loss": 0.04450156167149544, "global_step": 221295, "epoch": 2486, "lr": 8.218625826884013e-06}
{"train_loss": 0.05822368711233139, "global_step": 221296, "epoch": 2486, "lr": 8.218307382371626e-06}
{"train_loss": 0.04943302273750305, "global_step": 221297, "epoch": 2486, "lr": 8.217988943476256e-06}
{"train_loss": 0.0358499214053154, "global_step": 221298, "epoch": 2486, "lr": 8.217670510197989e-06}
{"train_loss": 0.05171831697225571, "global_step": 221299, "epoch": 2486, "lr": 8.217352082536844e-06}
{"train_loss": 0.03906268626451492, "global_step": 221300, "epoch": 2486, "lr": 8.217033660492851e-06}
{"train_loss": 0.06307005137205124, "global_step": 221301, "epoch": 2486, "lr": 8.216715244066086e-06}
{"train_loss": 0.05631683021783829, "global_step": 221302, "epoch": 2486, "lr": 8.216396833256562e-06}
{"train_loss": 0.04776833951473236, "global_step": 221303, "epoch": 2486, "lr": 8.216078428064345e-06}
{"train_loss": 0.04494676738977432, "global_step": 221304, "epoch": 2486, "lr": 8.215760028489449e-06}
{"train_loss": 0.01908039301633835, "global_step": 221305, "epoch": 2486, "lr": 8.215441634531956e-06}
{"train_loss": 0.10546490550041199, "global_step": 221306, "epoch": 2486, "lr": 8.21512324619187e-06}
{"train_loss": 0.04739384353160858, "global_step": 221307, "epoch": 2486, "lr": 8.214804863469266e-06}
{"train_loss": 0.038186054676771164, "global_step": 221308, "epoch": 2486, "lr": 8.214486486364159e-06}
{"train_loss": 0.0421660840511322, "global_step": 221309, "epoch": 2486, "lr": 8.21416811487662e-06}
{"train_loss": 0.06032702326774597, "global_step": 221310, "epoch": 2486, "lr": 8.213849749006658e-06}
{"train_loss": 0.05783811956644058, "global_step": 221311, "epoch": 2486, "lr": 8.213531388754353e-06}
{"train_loss": 0.11439282447099686, "global_step": 221312, "epoch": 2486, "lr": 8.213213034119715e-06}
{"train_loss": 0.07612036913633347, "global_step": 221313, "epoch": 2486, "lr": 8.212894685102812e-06}
{"train_loss": 0.047825317829847336, "global_step": 221314, "epoch": 2486, "lr": 8.212576341703665e-06}
{"train_loss": 0.07601114362478256, "global_step": 221315, "epoch": 2486, "lr": 8.212258003922341e-06}
{"train_loss": 0.09027015417814255, "global_step": 221316, "epoch": 2486, "lr": 8.211939671758856e-06}
{"train_loss": 0.060886312276124954, "global_step": 221317, "epoch": 2486, "lr": 8.211621345213282e-06}
{"train_loss": 0.06665860116481781, "global_step": 221318, "epoch": 2486, "lr": 8.211303024285633e-06}
{"train_loss": 0.05481324717402458, "global_step": 221319, "epoch": 2486, "lr": 8.210984708975972e-06}
{"train_loss": 0.07194272428750992, "global_step": 221320, "epoch": 2486, "lr": 8.210666399284328e-06}
{"train_loss": 0.044122811406850815, "global_step": 221321, "epoch": 2486, "lr": 8.210348095210747e-06}
{"train_loss": 0.029646199196577072, "global_step": 221322, "epoch": 2486, "lr": 8.210029796755286e-06}
{"train_loss": 0.05181141942739487, "global_step": 221323, "epoch": 2486, "lr": 8.209711503917971e-06}
{"train_loss": 0.13066673278808594, "global_step": 221324, "epoch": 2486, "lr": 8.209393216698857e-06}
{"train_loss": 0.02538861706852913, "global_step": 221325, "epoch": 2486, "lr": 8.209074935097976e-06}
{"train_loss": 0.08383556455373764, "global_step": 221326, "epoch": 2486, "lr": 8.208756659115379e-06}
{"train_loss": 0.03900053724646568, "global_step": 221327, "epoch": 2486, "lr": 8.208438388751099e-06}
{"train_loss": 0.04696396738290787, "global_step": 221328, "epoch": 2486, "lr": 8.208120124005198e-06}
{"train_loss": 0.057371895760297775, "global_step": 221329, "epoch": 2486, "lr": 8.20780186487769e-06}
{"train_loss": 0.05317964032292366, "global_step": 221330, "epoch": 2486, "lr": 8.20748361136865e-06}
{"train_loss": 0.026115505024790764, "global_step": 221331, "epoch": 2486, "lr": 8.207165363478086e-06}
{"train_loss": 0.035556256771087646, "global_step": 221332, "epoch": 2486, "lr": 8.206847121206074e-06}
{"train_loss": 0.06147720664739609, "global_step": 221333, "epoch": 2486, "lr": 8.206528884552633e-06}
{"train_loss": 0.014164046384394169, "global_step": 221334, "epoch": 2486, "lr": 8.206210653517827e-06}
{"train_loss": 0.02759743481874466, "global_step": 221335, "epoch": 2486, "lr": 8.20589242810167e-06}
{"train_loss": 0.05702757090330124, "global_step": 221336, "epoch": 2486, "lr": 8.205574208304233e-06}
{"train_loss": 0.04776924103498459, "global_step": 221337, "epoch": 2486, "lr": 8.205255994125537e-06}
{"train_loss": 0.04569346085190773, "global_step": 221338, "epoch": 2486, "lr": 8.204937785565653e-06}
{"train_loss": 0.06690970063209534, "global_step": 221339, "epoch": 2486, "lr": 8.2046195826246e-06}
{"train_loss": 0.05585409328341484, "global_step": 221340, "epoch": 2486, "lr": 8.204301385302416e-06}
{"train_loss": 0.07351797819137573, "global_step": 221341, "epoch": 2486, "lr": 8.203983193599163e-06}
{"train_loss": 0.05257892824207129, "global_step": 221342, "epoch": 2486, "lr": 8.203665007514866e-06, "val_loss": 8.820051193237305}
{"train_loss": 0.022945785894989967, "global_step": 221343, "epoch": 2487, "lr": 8.203346827049586e-06}
{"train_loss": 0.08115413784980774, "global_step": 221344, "epoch": 2487, "lr": 8.203028652203349e-06}
{"train_loss": 0.0455753467977047, "global_step": 221345, "epoch": 2487, "lr": 8.202710482976212e-06}
{"train_loss": 0.09582481533288956, "global_step": 221346, "epoch": 2487, "lr": 8.202392319368202e-06}
{"train_loss": 0.06496479362249374, "global_step": 221347, "epoch": 2487, "lr": 8.202074161379386e-06}
{"train_loss": 0.02039111591875553, "global_step": 221348, "epoch": 2487, "lr": 8.201756009009776e-06}
{"train_loss": 0.03426269069314003, "global_step": 221349, "epoch": 2487, "lr": 8.201437862259442e-06}
{"train_loss": 0.04943346977233887, "global_step": 221350, "epoch": 2487, "lr": 8.201119721128403e-06}
{"train_loss": 0.03643351048231125, "global_step": 221351, "epoch": 2487, "lr": 8.20080158561673e-06}
{"train_loss": 0.04447918012738228, "global_step": 221352, "epoch": 2487, "lr": 8.200483455724433e-06}
{"train_loss": 0.040290072560310364, "global_step": 221353, "epoch": 2487, "lr": 8.200165331451588e-06}
{"train_loss": 0.06976636499166489, "global_step": 221354, "epoch": 2487, "lr": 8.199847212798206e-06}
{"train_loss": 0.050927069038152695, "global_step": 221355, "epoch": 2487, "lr": 8.199529099764358e-06}
{"train_loss": 0.04404766112565994, "global_step": 221356, "epoch": 2487, "lr": 8.199210992350059e-06}
{"train_loss": 0.04746769741177559, "global_step": 221357, "epoch": 2487, "lr": 8.19889289055537e-06}
{"train_loss": 0.05301264300942421, "global_step": 221358, "epoch": 2487, "lr": 8.198574794380343e-06}
{"train_loss": 0.0536988228559494, "global_step": 221359, "epoch": 2487, "lr": 8.198256703824997e-06}
{"train_loss": 0.030079590156674385, "global_step": 221360, "epoch": 2487, "lr": 8.197938618889394e-06}
{"train_loss": 0.03926718980073929, "global_step": 221361, "epoch": 2487, "lr": 8.19762053957356e-06}
{"train_loss": 0.029431819915771484, "global_step": 221362, "epoch": 2487, "lr": 8.197302465877554e-06}
{"train_loss": 0.03880498930811882, "global_step": 221363, "epoch": 2487, "lr": 8.196984397801405e-06}
{"train_loss": 0.0705113559961319, "global_step": 221364, "epoch": 2487, "lr": 8.196666335345171e-06}
{"train_loss": 0.0313134640455246, "global_step": 221365, "epoch": 2487, "lr": 8.196348278508876e-06}
{"train_loss": 0.036467596888542175, "global_step": 221366, "epoch": 2487, "lr": 8.196030227292578e-06}
{"train_loss": 0.02638845518231392, "global_step": 221367, "epoch": 2487, "lr": 8.195712181696307e-06}
{"train_loss": 0.06340160965919495, "global_step": 221368, "epoch": 2487, "lr": 8.195394141720125e-06}
{"train_loss": 0.042864590883255005, "global_step": 221369, "epoch": 2487, "lr": 8.195076107364052e-06}
{"train_loss": 0.04726671054959297, "global_step": 221370, "epoch": 2487, "lr": 8.19475807862815e-06}
{"train_loss": 0.0646182969212532, "global_step": 221371, "epoch": 2487, "lr": 8.194440055512442e-06}
{"train_loss": 0.05375559255480766, "global_step": 221372, "epoch": 2487, "lr": 8.194122038016994e-06}
{"train_loss": 0.05507589131593704, "global_step": 221373, "epoch": 2487, "lr": 8.193804026141827e-06}
{"train_loss": 0.06403379142284393, "global_step": 221374, "epoch": 2487, "lr": 8.193486019887004e-06}
{"train_loss": 0.05399986356496811, "global_step": 221375, "epoch": 2487, "lr": 8.193168019252544e-06}
{"train_loss": 0.08310064673423767, "global_step": 221376, "epoch": 2487, "lr": 8.19285002423852e-06}
{"train_loss": 0.04585476219654083, "global_step": 221377, "epoch": 2487, "lr": 8.192532034844952e-06}
{"train_loss": 0.04479006305336952, "global_step": 221378, "epoch": 2487, "lr": 8.192214051071878e-06}
{"train_loss": 0.07145493477582932, "global_step": 221379, "epoch": 2487, "lr": 8.191896072919363e-06}
{"train_loss": 0.032466039061546326, "global_step": 221380, "epoch": 2487, "lr": 8.191578100387425e-06}
{"train_loss": 0.04572649300098419, "global_step": 221381, "epoch": 2487, "lr": 8.191260133476136e-06}
{"train_loss": 0.020636428147554398, "global_step": 221382, "epoch": 2487, "lr": 8.190942172185511e-06}
{"train_loss": 0.0325738824903965, "global_step": 221383, "epoch": 2487, "lr": 8.190624216515614e-06}
{"train_loss": 0.07369011640548706, "global_step": 221384, "epoch": 2487, "lr": 8.190306266466463e-06}
{"train_loss": 0.04701998457312584, "global_step": 221385, "epoch": 2487, "lr": 8.189988322038128e-06}
{"train_loss": 0.03596073389053345, "global_step": 221386, "epoch": 2487, "lr": 8.18967038323063e-06}
{"train_loss": 0.03670487180352211, "global_step": 221387, "epoch": 2487, "lr": 8.189352450044035e-06}
{"train_loss": 0.07418636232614517, "global_step": 221388, "epoch": 2487, "lr": 8.189034522478355e-06}
{"train_loss": 0.03840462118387222, "global_step": 221389, "epoch": 2487, "lr": 8.18871660053367e-06}
{"train_loss": 0.053142137825489044, "global_step": 221390, "epoch": 2487, "lr": 8.18839868420998e-06}
{"train_loss": 0.03592817112803459, "global_step": 221391, "epoch": 2487, "lr": 8.188080773507368e-06}
{"train_loss": 0.05949397012591362, "global_step": 221392, "epoch": 2487, "lr": 8.187762868425846e-06}
{"train_loss": 0.06492243707180023, "global_step": 221393, "epoch": 2487, "lr": 8.187444968965469e-06}
{"train_loss": 0.04296159744262695, "global_step": 221394, "epoch": 2487, "lr": 8.187127075126293e-06}
{"train_loss": 0.034536272287368774, "global_step": 221395, "epoch": 2487, "lr": 8.186809186908339e-06}
{"train_loss": 0.05940791964530945, "global_step": 221396, "epoch": 2487, "lr": 8.186491304311666e-06}
{"train_loss": 0.04162834212183952, "global_step": 221397, "epoch": 2487, "lr": 8.186173427336302e-06}
{"train_loss": 0.033960260450839996, "global_step": 221398, "epoch": 2487, "lr": 8.185855555982309e-06}
{"train_loss": 0.03580315038561821, "global_step": 221399, "epoch": 2487, "lr": 8.185537690249701e-06}
{"train_loss": 0.06361322849988937, "global_step": 221400, "epoch": 2487, "lr": 8.185219830138553e-06}
{"train_loss": 0.0812106728553772, "global_step": 221401, "epoch": 2487, "lr": 8.18490197564888e-06}
{"train_loss": 0.035993792116642, "global_step": 221402, "epoch": 2487, "lr": 8.184584126780748e-06}
{"train_loss": 0.045711882412433624, "global_step": 221403, "epoch": 2487, "lr": 8.184266283534182e-06}
{"train_loss": 0.06625668704509735, "global_step": 221404, "epoch": 2487, "lr": 8.183948445909239e-06}
{"train_loss": 0.06838533282279968, "global_step": 221405, "epoch": 2487, "lr": 8.183630613905946e-06}
{"train_loss": 0.0819011926651001, "global_step": 221406, "epoch": 2487, "lr": 8.18331278752436e-06}
{"train_loss": 0.05327777564525604, "global_step": 221407, "epoch": 2487, "lr": 8.182994966764512e-06}
{"train_loss": 0.05965572968125343, "global_step": 221408, "epoch": 2487, "lr": 8.18267715162646e-06}
{"train_loss": 0.05637649446725845, "global_step": 221409, "epoch": 2487, "lr": 8.18235934211023e-06}
{"train_loss": 0.07905329018831253, "global_step": 221410, "epoch": 2487, "lr": 8.18204153821588e-06}
{"train_loss": 0.03972152993083, "global_step": 221411, "epoch": 2487, "lr": 8.181723739943432e-06}
{"train_loss": 0.04765480384230614, "global_step": 221412, "epoch": 2487, "lr": 8.181405947292952e-06}
{"train_loss": 0.03688769415020943, "global_step": 221413, "epoch": 2487, "lr": 8.181088160264465e-06}
{"train_loss": 0.03683523088693619, "global_step": 221414, "epoch": 2487, "lr": 8.180770378858026e-06}
{"train_loss": 0.0275218915194273, "global_step": 221415, "epoch": 2487, "lr": 8.18045260307368e-06}
{"train_loss": 0.05959293618798256, "global_step": 221416, "epoch": 2487, "lr": 8.18013483291145e-06}
{"train_loss": 0.09698452055454254, "global_step": 221417, "epoch": 2487, "lr": 8.179817068371397e-06}
{"train_loss": 0.027689628303050995, "global_step": 221418, "epoch": 2487, "lr": 8.179499309453554e-06}
{"train_loss": 0.08761084079742432, "global_step": 221419, "epoch": 2487, "lr": 8.179181556157978e-06}
{"train_loss": 0.0713706761598587, "global_step": 221420, "epoch": 2487, "lr": 8.178863808484682e-06}
{"train_loss": 0.03510577231645584, "global_step": 221421, "epoch": 2487, "lr": 8.178546066433745e-06}
{"train_loss": 0.04608985781669617, "global_step": 221422, "epoch": 2487, "lr": 8.178228330005184e-06}
{"train_loss": 0.05597253888845444, "global_step": 221423, "epoch": 2487, "lr": 8.177910599199056e-06}
{"train_loss": 0.08429207652807236, "global_step": 221424, "epoch": 2487, "lr": 8.177592874015394e-06}
{"train_loss": 0.07959451526403427, "global_step": 221425, "epoch": 2487, "lr": 8.177275154454251e-06}
{"train_loss": 0.023669477552175522, "global_step": 221426, "epoch": 2487, "lr": 8.176957440515653e-06}
{"train_loss": 0.03576800599694252, "global_step": 221427, "epoch": 2487, "lr": 8.176639732199665e-06}
{"train_loss": 0.03266425430774689, "global_step": 221428, "epoch": 2487, "lr": 8.176322029506306e-06}
{"train_loss": 0.0576285757124424, "global_step": 221429, "epoch": 2487, "lr": 8.176004332435628e-06}
{"train_loss": 0.06256476789712906, "global_step": 221430, "epoch": 2487, "lr": 8.175686640987696e-06}
{"train_loss": 0.05058281630110205, "global_step": 221431, "epoch": 2487, "lr": 8.175368955162516e-06, "val_loss": 9.012384414672852}
{"train_loss": 0.053887952119112015, "global_step": 221432, "epoch": 2488, "lr": 8.175051274960161e-06}
{"train_loss": 0.0810013860464096, "global_step": 221433, "epoch": 2488, "lr": 8.174733600380646e-06}
{"train_loss": 0.023319318890571594, "global_step": 221434, "epoch": 2488, "lr": 8.17441593142404e-06}
{"train_loss": 0.048052914440631866, "global_step": 221435, "epoch": 2488, "lr": 8.17409826809037e-06}
{"train_loss": 0.02893436700105667, "global_step": 221436, "epoch": 2488, "lr": 8.173780610379684e-06}
{"train_loss": 0.035971030592918396, "global_step": 221437, "epoch": 2488, "lr": 8.173462958292016e-06}
{"train_loss": 0.01918383501470089, "global_step": 221438, "epoch": 2488, "lr": 8.173145311827429e-06}
{"train_loss": 0.07185237109661102, "global_step": 221439, "epoch": 2488, "lr": 8.172827670985944e-06}
{"train_loss": 0.04005453363060951, "global_step": 221440, "epoch": 2488, "lr": 8.172510035767617e-06}
{"train_loss": 0.0431445837020874, "global_step": 221441, "epoch": 2488, "lr": 8.17219240617248e-06}
{"train_loss": 0.08678148686885834, "global_step": 221442, "epoch": 2488, "lr": 8.171874782200589e-06}
{"train_loss": 0.07212886959314346, "global_step": 221443, "epoch": 2488, "lr": 8.171557163851973e-06}
{"train_loss": 0.02679881453514099, "global_step": 221444, "epoch": 2488, "lr": 8.171239551126692e-06}
{"train_loss": 0.02944296784698963, "global_step": 221445, "epoch": 2488, "lr": 8.170921944024761e-06}
{"train_loss": 0.009597633965313435, "global_step": 221446, "epoch": 2488, "lr": 8.170604342546262e-06}
{"train_loss": 0.06271103769540787, "global_step": 221447, "epoch": 2488, "lr": 8.170286746691197e-06}
{"train_loss": 0.06315154582262039, "global_step": 221448, "epoch": 2488, "lr": 8.169969156459623e-06}
{"train_loss": 0.057078707963228226, "global_step": 221449, "epoch": 2488, "lr": 8.169651571851606e-06}
{"train_loss": 0.049386877566576004, "global_step": 221450, "epoch": 2488, "lr": 8.169333992867167e-06}
{"train_loss": 0.10267563909292221, "global_step": 221451, "epoch": 2488, "lr": 8.169016419506336e-06}
{"train_loss": 0.06232733651995659, "global_step": 221452, "epoch": 2488, "lr": 8.168698851769186e-06}
{"train_loss": 0.049407266080379486, "global_step": 221453, "epoch": 2488, "lr": 8.168381289655736e-06}
{"train_loss": 0.05076134577393532, "global_step": 221454, "epoch": 2488, "lr": 8.168063733166043e-06}
{"train_loss": 0.06071263179183006, "global_step": 221455, "epoch": 2488, "lr": 8.167746182300146e-06}
{"train_loss": 0.0418633408844471, "global_step": 221456, "epoch": 2488, "lr": 8.167428637058073e-06}
{"train_loss": 0.02053460292518139, "global_step": 221457, "epoch": 2488, "lr": 8.167111097439894e-06}
{"train_loss": 0.04754578694701195, "global_step": 221458, "epoch": 2488, "lr": 8.166793563445624e-06}
{"train_loss": 0.057393211871385574, "global_step": 221459, "epoch": 2488, "lr": 8.166476035075333e-06}
{"train_loss": 0.0684976577758789, "global_step": 221460, "epoch": 2488, "lr": 8.16615851232903e-06}
{"train_loss": 0.053538136184215546, "global_step": 221461, "epoch": 2488, "lr": 8.165840995206798e-06}
{"train_loss": 0.05674538016319275, "global_step": 221462, "epoch": 2488, "lr": 8.165523483708637e-06}
{"train_loss": 0.06857268512248993, "global_step": 221463, "epoch": 2488, "lr": 8.165205977834634e-06}
{"train_loss": 0.07420654594898224, "global_step": 221464, "epoch": 2488, "lr": 8.164888477584792e-06}
{"train_loss": 0.06935840845108032, "global_step": 221465, "epoch": 2488, "lr": 8.164570982959168e-06}
{"train_loss": 0.042694393545389175, "global_step": 221466, "epoch": 2488, "lr": 8.164253493957825e-06}
{"train_loss": 0.04125569388270378, "global_step": 221467, "epoch": 2488, "lr": 8.163936010580775e-06}
{"train_loss": 0.05459650978446007, "global_step": 221468, "epoch": 2488, "lr": 8.163618532828083e-06}
{"train_loss": 0.05375698208808899, "global_step": 221469, "epoch": 2488, "lr": 8.163301060699773e-06}
{"train_loss": 0.0533541664481163, "global_step": 221470, "epoch": 2488, "lr": 8.162983594195906e-06}
{"train_loss": 0.02887214533984661, "global_step": 221471, "epoch": 2488, "lr": 8.16266613331651e-06}
{"train_loss": 0.04145839065313339, "global_step": 221472, "epoch": 2488, "lr": 8.162348678061637e-06}
{"train_loss": 0.049638450145721436, "global_step": 221473, "epoch": 2488, "lr": 8.162031228431322e-06}
{"train_loss": 0.033123549073934555, "global_step": 221474, "epoch": 2488, "lr": 8.161713784425622e-06}
{"train_loss": 0.015768256038427353, "global_step": 221475, "epoch": 2488, "lr": 8.161396346044559e-06}
{"train_loss": 0.06594875454902649, "global_step": 221476, "epoch": 2488, "lr": 8.161078913288195e-06}
{"train_loss": 0.05220406502485275, "global_step": 221477, "epoch": 2488, "lr": 8.160761486156554e-06}
{"train_loss": 0.05511597916483879, "global_step": 221478, "epoch": 2488, "lr": 8.160444064649698e-06}
{"train_loss": 0.0699198991060257, "global_step": 221479, "epoch": 2488, "lr": 8.160126648767652e-06}
{"train_loss": 0.0321357287466526, "global_step": 221480, "epoch": 2488, "lr": 8.159809238510474e-06}
{"train_loss": 0.07100940495729446, "global_step": 221481, "epoch": 2488, "lr": 8.159491833878191e-06}
{"train_loss": 0.035050634294748306, "global_step": 221482, "epoch": 2488, "lr": 8.15917443487087e-06}
{"train_loss": 0.04299834370613098, "global_step": 221483, "epoch": 2488, "lr": 8.15885704148852e-06}
{"train_loss": 0.023544548079371452, "global_step": 221484, "epoch": 2488, "lr": 8.158539653731206e-06}
{"train_loss": 0.05274447053670883, "global_step": 221485, "epoch": 2488, "lr": 8.158222271598976e-06}
{"train_loss": 0.026925593614578247, "global_step": 221486, "epoch": 2488, "lr": 8.15790489509185e-06}
{"train_loss": 0.03888636454939842, "global_step": 221487, "epoch": 2488, "lr": 8.157587524209898e-06}
{"train_loss": 0.019053565338253975, "global_step": 221488, "epoch": 2488, "lr": 8.157270158953146e-06}
{"train_loss": 0.06502746790647507, "global_step": 221489, "epoch": 2488, "lr": 8.156952799321632e-06}
{"train_loss": 0.035626400262117386, "global_step": 221490, "epoch": 2488, "lr": 8.156635445315414e-06}
{"train_loss": 0.0660017654299736, "global_step": 221491, "epoch": 2488, "lr": 8.156318096934513e-06}
{"train_loss": 0.012089891359210014, "global_step": 221492, "epoch": 2488, "lr": 8.156000754179e-06}
{"train_loss": 0.051926348358392715, "global_step": 221493, "epoch": 2488, "lr": 8.155683417048898e-06}
{"train_loss": 0.06937270611524582, "global_step": 221494, "epoch": 2488, "lr": 8.155366085544248e-06}
{"train_loss": 0.07904381304979324, "global_step": 221495, "epoch": 2488, "lr": 8.155048759665107e-06}
{"train_loss": 0.041601475328207016, "global_step": 221496, "epoch": 2488, "lr": 8.1547314394115e-06}
{"train_loss": 0.028335116803646088, "global_step": 221497, "epoch": 2488, "lr": 8.154414124783494e-06}
{"train_loss": 0.058687590062618256, "global_step": 221498, "epoch": 2488, "lr": 8.154096815781099e-06}
{"train_loss": 0.050246309489011765, "global_step": 221499, "epoch": 2488, "lr": 8.153779512404391e-06}
{"train_loss": 0.04995255172252655, "global_step": 221500, "epoch": 2488, "lr": 8.153462214653384e-06}
{"train_loss": 0.054306238889694214, "global_step": 221501, "epoch": 2488, "lr": 8.153144922528134e-06}
{"train_loss": 0.07390322536230087, "global_step": 221502, "epoch": 2488, "lr": 8.152827636028698e-06}
{"train_loss": 0.09101773798465729, "global_step": 221503, "epoch": 2488, "lr": 8.152510355155092e-06}
{"train_loss": 0.03455151990056038, "global_step": 221504, "epoch": 2488, "lr": 8.152193079907378e-06}
{"train_loss": 0.0502142459154129, "global_step": 221505, "epoch": 2488, "lr": 8.151875810285586e-06}
{"train_loss": 0.06049877032637596, "global_step": 221506, "epoch": 2488, "lr": 8.151558546289773e-06}
{"train_loss": 0.06313708424568176, "global_step": 221507, "epoch": 2488, "lr": 8.151241287919958e-06}
{"train_loss": 0.06971750408411026, "global_step": 221508, "epoch": 2488, "lr": 8.150924035176217e-06}
{"train_loss": 0.05422632768750191, "global_step": 221509, "epoch": 2488, "lr": 8.150606788058557e-06}
{"train_loss": 0.09176604449748993, "global_step": 221510, "epoch": 2488, "lr": 8.150289546567053e-06}
{"train_loss": 0.05115589499473572, "global_step": 221511, "epoch": 2488, "lr": 8.14997231070172e-06}
{"train_loss": 0.013570445589721203, "global_step": 221512, "epoch": 2488, "lr": 8.149655080462625e-06}
{"train_loss": 0.06753388792276382, "global_step": 221513, "epoch": 2488, "lr": 8.149337855849787e-06}
{"train_loss": 0.08965475857257843, "global_step": 221514, "epoch": 2488, "lr": 8.149020636863269e-06}
{"train_loss": 0.021643459796905518, "global_step": 221515, "epoch": 2488, "lr": 8.148703423503096e-06}
{"train_loss": 0.0530778132379055, "global_step": 221516, "epoch": 2488, "lr": 8.148386215769332e-06}
{"train_loss": 0.030559414997696877, "global_step": 221517, "epoch": 2488, "lr": 8.148069013661997e-06}
{"train_loss": 0.09226470440626144, "global_step": 221518, "epoch": 2488, "lr": 8.147751817181154e-06}
{"train_loss": 0.032062772661447525, "global_step": 221519, "epoch": 2488, "lr": 8.147434626326823e-06}
{"train_loss": 0.050856978119759075, "global_step": 221520, "epoch": 2488, "lr": 8.147117441099062e-06, "val_loss": 9.09230899810791}
{"train_loss": 0.04053473472595215, "global_step": 221521, "epoch": 2489, "lr": 8.146800261497917e-06}
{"train_loss": 0.05835407227277756, "global_step": 221522, "epoch": 2489, "lr": 8.146483087523416e-06}
{"train_loss": 0.08597550541162491, "global_step": 221523, "epoch": 2489, "lr": 8.146165919175625e-06}
{"train_loss": 0.04019208624958992, "global_step": 221524, "epoch": 2489, "lr": 8.145848756454555e-06}
{"train_loss": 0.04308922961354256, "global_step": 221525, "epoch": 2489, "lr": 8.145531599360278e-06}
{"train_loss": 0.062244217842817307, "global_step": 221526, "epoch": 2489, "lr": 8.14521444789283e-06}
{"train_loss": 0.06168408319354057, "global_step": 221527, "epoch": 2489, "lr": 8.14489730205223e-06}
{"train_loss": 0.04814176261425018, "global_step": 221528, "epoch": 2489, "lr": 8.144580161838555e-06}
{"train_loss": 0.029609141871333122, "global_step": 221529, "epoch": 2489, "lr": 8.144263027251814e-06}
{"train_loss": 0.06676653772592545, "global_step": 221530, "epoch": 2489, "lr": 8.143945898292083e-06}
{"train_loss": 0.0818987488746643, "global_step": 221531, "epoch": 2489, "lr": 8.143628774959382e-06}
{"train_loss": 0.02540883794426918, "global_step": 221532, "epoch": 2489, "lr": 8.143311657253756e-06}
{"train_loss": 0.09132040292024612, "global_step": 221533, "epoch": 2489, "lr": 8.142994545175259e-06}
{"train_loss": 0.07078409194946289, "global_step": 221534, "epoch": 2489, "lr": 8.142677438723916e-06}
{"train_loss": 0.0441746711730957, "global_step": 221535, "epoch": 2489, "lr": 8.14236033789979e-06}
{"train_loss": 0.024837424978613853, "global_step": 221536, "epoch": 2489, "lr": 8.142043242702902e-06}
{"train_loss": 0.07933495193719864, "global_step": 221537, "epoch": 2489, "lr": 8.141726153133323e-06}
{"train_loss": 0.02815306931734085, "global_step": 221538, "epoch": 2489, "lr": 8.141409069191058e-06}
{"train_loss": 0.07269813120365143, "global_step": 221539, "epoch": 2489, "lr": 8.141091990876177e-06}
{"train_loss": 0.03899342566728592, "global_step": 221540, "epoch": 2489, "lr": 8.140774918188726e-06}
{"train_loss": 0.056984078139066696, "global_step": 221541, "epoch": 2489, "lr": 8.140457851128724e-06}
{"train_loss": 0.048253074288368225, "global_step": 221542, "epoch": 2489, "lr": 8.140140789696243e-06}
{"train_loss": 0.03072412870824337, "global_step": 221543, "epoch": 2489, "lr": 8.139823733891295e-06}
{"train_loss": 0.038972847163677216, "global_step": 221544, "epoch": 2489, "lr": 8.139506683713948e-06}
{"train_loss": 0.05689026415348053, "global_step": 221545, "epoch": 2489, "lr": 8.139189639164225e-06}
{"train_loss": 0.023001166060566902, "global_step": 221546, "epoch": 2489, "lr": 8.138872600242186e-06}
{"train_loss": 0.04369731992483139, "global_step": 221547, "epoch": 2489, "lr": 8.138555566947858e-06}
{"train_loss": 0.06381864845752716, "global_step": 221548, "epoch": 2489, "lr": 8.138238539281302e-06}
{"train_loss": 0.0583508163690567, "global_step": 221549, "epoch": 2489, "lr": 8.137921517242536e-06}
{"train_loss": 0.07999185472726822, "global_step": 221550, "epoch": 2489, "lr": 8.137604500831631e-06}
{"train_loss": 0.0779518112540245, "global_step": 221551, "epoch": 2489, "lr": 8.137287490048601e-06}
{"train_loss": 0.09252715855836868, "global_step": 221552, "epoch": 2489, "lr": 8.136970484893514e-06}
{"train_loss": 0.04742291569709778, "global_step": 221553, "epoch": 2489, "lr": 8.136653485366385e-06}
{"train_loss": 0.06123371422290802, "global_step": 221554, "epoch": 2489, "lr": 8.136336491467288e-06}
{"train_loss": 0.06509820371866226, "global_step": 221555, "epoch": 2489, "lr": 8.136019503196241e-06}
{"train_loss": 0.07675378769636154, "global_step": 221556, "epoch": 2489, "lr": 8.135702520553296e-06}
{"train_loss": 0.05353265255689621, "global_step": 221557, "epoch": 2489, "lr": 8.135385543538504e-06}
{"train_loss": 0.05501500144600868, "global_step": 221558, "epoch": 2489, "lr": 8.135068572151888e-06}
{"train_loss": 0.023913655430078506, "global_step": 221559, "epoch": 2489, "lr": 8.134751606393521e-06}
{"train_loss": 0.051053643226623535, "global_step": 221560, "epoch": 2489, "lr": 8.134434646263405e-06}
{"train_loss": 0.03186187520623207, "global_step": 221561, "epoch": 2489, "lr": 8.134117691761617e-06}
{"train_loss": 0.05227020010352135, "global_step": 221562, "epoch": 2489, "lr": 8.133800742888181e-06}
{"train_loss": 0.07517461478710175, "global_step": 221563, "epoch": 2489, "lr": 8.133483799643149e-06}
{"train_loss": 0.05111902207136154, "global_step": 221564, "epoch": 2489, "lr": 8.133166862026553e-06}
{"train_loss": 0.06017720326781273, "global_step": 221565, "epoch": 2489, "lr": 8.132849930038456e-06}
{"train_loss": 0.0759100541472435, "global_step": 221566, "epoch": 2489, "lr": 8.132533003678883e-06}
{"train_loss": 0.06502927094697952, "global_step": 221567, "epoch": 2489, "lr": 8.132216082947868e-06}
{"train_loss": 0.0531022809445858, "global_step": 221568, "epoch": 2489, "lr": 8.13189916784548e-06}
{"train_loss": 0.08829560875892639, "global_step": 221569, "epoch": 2489, "lr": 8.13158225837174e-06}
{"train_loss": 0.03266122564673424, "global_step": 221570, "epoch": 2489, "lr": 8.131265354526701e-06}
{"train_loss": 0.07612349838018417, "global_step": 221571, "epoch": 2489, "lr": 8.130948456310411e-06}
{"train_loss": 0.08750598877668381, "global_step": 221572, "epoch": 2489, "lr": 8.130631563722891e-06}
{"train_loss": 0.03597043827176094, "global_step": 221573, "epoch": 2489, "lr": 8.130314676764206e-06}
{"train_loss": 0.04383907467126846, "global_step": 221574, "epoch": 2489, "lr": 8.12999779543438e-06}
{"train_loss": 0.07552526146173477, "global_step": 221575, "epoch": 2489, "lr": 8.129680919733468e-06}
{"train_loss": 0.04437967389822006, "global_step": 221576, "epoch": 2489, "lr": 8.12936404966152e-06}
{"train_loss": 0.03787478059530258, "global_step": 221577, "epoch": 2489, "lr": 8.129047185218558e-06}
{"train_loss": 0.060115791857242584, "global_step": 221578, "epoch": 2489, "lr": 8.128730326404644e-06}
{"train_loss": 0.03712688758969307, "global_step": 221579, "epoch": 2489, "lr": 8.128413473219803e-06}
{"train_loss": 0.047173768281936646, "global_step": 221580, "epoch": 2489, "lr": 8.128096625664095e-06}
{"train_loss": 0.03903517499566078, "global_step": 221581, "epoch": 2489, "lr": 8.127779783737544e-06}
{"train_loss": 0.048531729727983475, "global_step": 221582, "epoch": 2489, "lr": 8.127462947440218e-06}
{"train_loss": 0.033289484679698944, "global_step": 221583, "epoch": 2489, "lr": 8.127146116772128e-06}
{"train_loss": 0.04651091247797012, "global_step": 221584, "epoch": 2489, "lr": 8.126829291733345e-06}
{"train_loss": 0.0853486955165863, "global_step": 221585, "epoch": 2489, "lr": 8.126512472323888e-06}
{"train_loss": 0.08485831320285797, "global_step": 221586, "epoch": 2489, "lr": 8.126195658543822e-06}
{"train_loss": 0.07484296709299088, "global_step": 221587, "epoch": 2489, "lr": 8.12587885039317e-06}
{"train_loss": 0.05586037039756775, "global_step": 221588, "epoch": 2489, "lr": 8.125562047871993e-06}
{"train_loss": 0.1346074938774109, "global_step": 221589, "epoch": 2489, "lr": 8.125245250980312e-06}
{"train_loss": 0.012613052502274513, "global_step": 221590, "epoch": 2489, "lr": 8.124928459718195e-06}
{"train_loss": 0.05468440800905228, "global_step": 221591, "epoch": 2489, "lr": 8.124611674085659e-06}
{"train_loss": 0.03418505936861038, "global_step": 221592, "epoch": 2489, "lr": 8.12429489408275e-06}
{"train_loss": 0.07743745297193527, "global_step": 221593, "epoch": 2489, "lr": 8.12397811970954e-06}
{"train_loss": 0.0631944015622139, "global_step": 221594, "epoch": 2489, "lr": 8.123661350966038e-06}
{"train_loss": 0.07612431049346924, "global_step": 221595, "epoch": 2489, "lr": 8.123344587852312e-06}
{"train_loss": 0.053688183426856995, "global_step": 221596, "epoch": 2489, "lr": 8.12302783036838e-06}
{"train_loss": 0.08230932801961899, "global_step": 221597, "epoch": 2489, "lr": 8.122711078514306e-06}
{"train_loss": 0.03561943769454956, "global_step": 221598, "epoch": 2489, "lr": 8.122394332290113e-06}
{"train_loss": 0.04434473440051079, "global_step": 221599, "epoch": 2489, "lr": 8.12207759169586e-06}
{"train_loss": 0.051774799823760986, "global_step": 221600, "epoch": 2489, "lr": 8.121760856731576e-06}
{"train_loss": 0.03469220548868179, "global_step": 221601, "epoch": 2489, "lr": 8.121444127397322e-06}
{"train_loss": 0.11360837519168854, "global_step": 221602, "epoch": 2489, "lr": 8.121127403693119e-06}
{"train_loss": 0.05936223641037941, "global_step": 221603, "epoch": 2489, "lr": 8.12081068561903e-06}
{"train_loss": 0.05802248418331146, "global_step": 221604, "epoch": 2489, "lr": 8.120493973175092e-06}
{"train_loss": 0.08398420363664627, "global_step": 221605, "epoch": 2489, "lr": 8.120177266361322e-06}
{"train_loss": 0.04273408278822899, "global_step": 221606, "epoch": 2489, "lr": 8.119860565177807e-06}
{"train_loss": 0.04148591682314873, "global_step": 221607, "epoch": 2489, "lr": 8.119543869624546e-06}
{"train_loss": 0.05662710219621658, "global_step": 221608, "epoch": 2489, "lr": 8.119227179701617e-06}
{"train_loss": 0.05696512195752578, "global_step": 221609, "epoch": 2489, "lr": 8.118910495409049e-06, "val_loss": 9.037129402160645}
{"train_loss": 0.034988630563020706, "global_step": 221610, "epoch": 2490, "lr": 8.118593816746867e-06}
{"train_loss": 0.04371515288949013, "global_step": 221611, "epoch": 2490, "lr": 8.11827714371513e-06}
{"train_loss": 0.02222873456776142, "global_step": 221612, "epoch": 2490, "lr": 8.117960476313895e-06}
{"train_loss": 0.04309946671128273, "global_step": 221613, "epoch": 2490, "lr": 8.117643814543175e-06}
{"train_loss": 0.0620158426463604, "global_step": 221614, "epoch": 2490, "lr": 8.117327158403044e-06}
{"train_loss": 0.07163001596927643, "global_step": 221615, "epoch": 2490, "lr": 8.117010507893513e-06}
{"train_loss": 0.04271433874964714, "global_step": 221616, "epoch": 2490, "lr": 8.116693863014652e-06}
{"train_loss": 0.008924150839447975, "global_step": 221617, "epoch": 2490, "lr": 8.116377223766475e-06}
{"train_loss": 0.059783320873975754, "global_step": 221618, "epoch": 2490, "lr": 8.116060590149055e-06}
{"train_loss": 0.06172611191868782, "global_step": 221619, "epoch": 2490, "lr": 8.11574396216241e-06}
{"train_loss": 0.037632621824741364, "global_step": 221620, "epoch": 2490, "lr": 8.115427339806602e-06}
{"train_loss": 0.08241583406925201, "global_step": 221621, "epoch": 2490, "lr": 8.115110723081654e-06}
{"train_loss": 0.041020095348358154, "global_step": 221622, "epoch": 2490, "lr": 8.114794111987633e-06}
{"train_loss": 0.04971836134791374, "global_step": 221623, "epoch": 2490, "lr": 8.114477506524554e-06}
{"train_loss": 0.08736694604158401, "global_step": 221624, "epoch": 2490, "lr": 8.114160906692481e-06}
{"train_loss": 0.06978867948055267, "global_step": 221625, "epoch": 2490, "lr": 8.113844312491443e-06}
{"train_loss": 0.03412679210305214, "global_step": 221626, "epoch": 2490, "lr": 8.113527723921494e-06}
{"train_loss": 0.07605821639299393, "global_step": 221627, "epoch": 2490, "lr": 8.113211140982663e-06}
{"train_loss": 0.04794084280729294, "global_step": 221628, "epoch": 2490, "lr": 8.112894563675e-06}
{"train_loss": 0.056268077343702316, "global_step": 221629, "epoch": 2490, "lr": 8.112577991998555e-06}
{"train_loss": 0.06048537418246269, "global_step": 221630, "epoch": 2490, "lr": 8.112261425953355e-06}
{"train_loss": 0.05055272579193115, "global_step": 221631, "epoch": 2490, "lr": 8.111944865539467e-06}
{"train_loss": 0.033121127635240555, "global_step": 221632, "epoch": 2490, "lr": 8.111628310756897e-06}
{"train_loss": 0.029429282993078232, "global_step": 221633, "epoch": 2490, "lr": 8.111311761605727e-06}
{"train_loss": 0.054438166320323944, "global_step": 221634, "epoch": 2490, "lr": 8.110995218085965e-06}
{"train_loss": 0.0893445611000061, "global_step": 221635, "epoch": 2490, "lr": 8.11067868019768e-06}
{"train_loss": 0.05321982130408287, "global_step": 221636, "epoch": 2490, "lr": 8.110362147940898e-06}
{"train_loss": 0.07043676823377609, "global_step": 221637, "epoch": 2490, "lr": 8.11004562131567e-06}
{"train_loss": 0.04839149862527847, "global_step": 221638, "epoch": 2490, "lr": 8.109729100322034e-06}
{"train_loss": 0.09320101886987686, "global_step": 221639, "epoch": 2490, "lr": 8.10941258496004e-06}
{"train_loss": 0.06727471202611923, "global_step": 221640, "epoch": 2490, "lr": 8.109096075229711e-06}
{"train_loss": 0.0672341138124466, "global_step": 221641, "epoch": 2490, "lr": 8.10877957113112e-06}
{"train_loss": 0.04591911658644676, "global_step": 221642, "epoch": 2490, "lr": 8.108463072664291e-06}
{"train_loss": 0.09629063308238983, "global_step": 221643, "epoch": 2490, "lr": 8.108146579829257e-06}
{"train_loss": 0.046579863876104355, "global_step": 221644, "epoch": 2490, "lr": 8.107830092626084e-06}
{"train_loss": 0.040560103952884674, "global_step": 221645, "epoch": 2490, "lr": 8.10751361105479e-06}
{"train_loss": 0.022806568071246147, "global_step": 221646, "epoch": 2490, "lr": 8.107197135115446e-06}
{"train_loss": 0.05455508083105087, "global_step": 221647, "epoch": 2490, "lr": 8.106880664808065e-06}
{"train_loss": 0.05325424298644066, "global_step": 221648, "epoch": 2490, "lr": 8.10656420013271e-06}
{"train_loss": 0.03913719579577446, "global_step": 221649, "epoch": 2490, "lr": 8.106247741089413e-06}
{"train_loss": 0.045763712376356125, "global_step": 221650, "epoch": 2490, "lr": 8.105931287678226e-06}
{"train_loss": 0.02376694604754448, "global_step": 221651, "epoch": 2490, "lr": 8.105614839899179e-06}
{"train_loss": 0.055770210921764374, "global_step": 221652, "epoch": 2490, "lr": 8.105298397752326e-06}
{"train_loss": 0.045902639627456665, "global_step": 221653, "epoch": 2490, "lr": 8.1049819612377e-06}
{"train_loss": 0.0694475919008255, "global_step": 221654, "epoch": 2490, "lr": 8.10466553035536e-06}
{"train_loss": 0.039004549384117126, "global_step": 221655, "epoch": 2490, "lr": 8.104349105105319e-06}
{"train_loss": 0.040722619742155075, "global_step": 221656, "epoch": 2490, "lr": 8.104032685487655e-06}
{"train_loss": 0.031214598566293716, "global_step": 221657, "epoch": 2490, "lr": 8.103716271502381e-06}
{"train_loss": 0.04326522350311279, "global_step": 221658, "epoch": 2490, "lr": 8.103399863149563e-06}
{"train_loss": 0.09529805183410645, "global_step": 221659, "epoch": 2490, "lr": 8.103083460429218e-06}
{"train_loss": 0.0669780969619751, "global_step": 221660, "epoch": 2490, "lr": 8.102767063341415e-06}
{"train_loss": 0.034319546073675156, "global_step": 221661, "epoch": 2490, "lr": 8.102450671886175e-06}
{"train_loss": 0.055363159626722336, "global_step": 221662, "epoch": 2490, "lr": 8.102134286063561e-06}
{"train_loss": 0.10900811105966568, "global_step": 221663, "epoch": 2490, "lr": 8.101817905873593e-06}
{"train_loss": 0.0322110541164875, "global_step": 221664, "epoch": 2490, "lr": 8.101501531316335e-06}
{"train_loss": 0.03834736347198486, "global_step": 221665, "epoch": 2490, "lr": 8.101185162391812e-06}
{"train_loss": 0.06002653017640114, "global_step": 221666, "epoch": 2490, "lr": 8.100868799100069e-06}
{"train_loss": 0.07316870987415314, "global_step": 221667, "epoch": 2490, "lr": 8.100552441441167e-06}
{"train_loss": 0.049898210912942886, "global_step": 221668, "epoch": 2490, "lr": 8.100236089415125e-06}
{"train_loss": 0.03424706310033798, "global_step": 221669, "epoch": 2490, "lr": 8.099919743022005e-06}
{"train_loss": 0.050955791026353836, "global_step": 221670, "epoch": 2490, "lr": 8.09960340226183e-06}
{"train_loss": 0.041479140520095825, "global_step": 221671, "epoch": 2490, "lr": 8.099287067134664e-06}
{"train_loss": 0.03762500733137131, "global_step": 221672, "epoch": 2490, "lr": 8.098970737640527e-06}
{"train_loss": 0.02802535891532898, "global_step": 221673, "epoch": 2490, "lr": 8.098654413779483e-06}
{"train_loss": 0.07191251218318939, "global_step": 221674, "epoch": 2490, "lr": 8.09833809555155e-06}
{"train_loss": 0.033128589391708374, "global_step": 221675, "epoch": 2490, "lr": 8.098021782956799e-06}
{"train_loss": 0.03335999324917793, "global_step": 221676, "epoch": 2490, "lr": 8.09770547599525e-06}
{"train_loss": 0.03789743036031723, "global_step": 221677, "epoch": 2490, "lr": 8.097389174666958e-06}
{"train_loss": 0.054414570331573486, "global_step": 221678, "epoch": 2490, "lr": 8.097072878971957e-06}
{"train_loss": 0.07245287299156189, "global_step": 221679, "epoch": 2490, "lr": 8.096756588910298e-06}
{"train_loss": 0.0636005774140358, "global_step": 221680, "epoch": 2490, "lr": 8.096440304482017e-06}
{"train_loss": 0.06495492160320282, "global_step": 221681, "epoch": 2490, "lr": 8.096124025687163e-06}
{"train_loss": 0.03936341777443886, "global_step": 221682, "epoch": 2490, "lr": 8.095807752525775e-06}
{"train_loss": 0.04620470851659775, "global_step": 221683, "epoch": 2490, "lr": 8.095491484997885e-06}
{"train_loss": 0.07386337965726852, "global_step": 221684, "epoch": 2490, "lr": 8.095175223103557e-06}
{"train_loss": 0.08910058438777924, "global_step": 221685, "epoch": 2490, "lr": 8.094858966842811e-06}
{"train_loss": 0.056809913367033005, "global_step": 221686, "epoch": 2490, "lr": 8.094542716215714e-06}
{"train_loss": 0.02051510661840439, "global_step": 221687, "epoch": 2490, "lr": 8.094226471222283e-06}
{"train_loss": 0.07406526058912277, "global_step": 221688, "epoch": 2490, "lr": 8.093910231862579e-06}
{"train_loss": 0.05475267022848129, "global_step": 221689, "epoch": 2490, "lr": 8.093593998136628e-06}
{"train_loss": 0.09292291849851608, "global_step": 221690, "epoch": 2490, "lr": 8.0932777700445e-06}
{"train_loss": 0.06407322734594345, "global_step": 221691, "epoch": 2490, "lr": 8.092961547586203e-06}
{"train_loss": 0.09124468266963959, "global_step": 221692, "epoch": 2490, "lr": 8.092645330761806e-06}
{"train_loss": 0.04500424861907959, "global_step": 221693, "epoch": 2490, "lr": 8.092329119571335e-06}
{"train_loss": 0.0349138006567955, "global_step": 221694, "epoch": 2490, "lr": 8.092012914014852e-06}
{"train_loss": 0.04531014710664749, "global_step": 221695, "epoch": 2490, "lr": 8.091696714092373e-06}
{"train_loss": 0.04598943144083023, "global_step": 221696, "epoch": 2490, "lr": 8.091380519803964e-06}
{"train_loss": 0.0212812889367342, "global_step": 221697, "epoch": 2490, "lr": 8.091064331149651e-06}
{"train_loss": 0.05278954418438874, "global_step": 221698, "epoch": 2490, "lr": 8.09074814812949e-06, "val_loss": 8.880134582519531, "train_action_mse_error": 35.934322357177734}
{"train_loss": 0.06810715049505234, "global_step": 221699, "epoch": 2491, "lr": 8.090431970743511e-06}
{"train_loss": 0.09266643971204758, "global_step": 221700, "epoch": 2491, "lr": 8.090115798991771e-06}
{"train_loss": 0.07010197639465332, "global_step": 221701, "epoch": 2491, "lr": 8.089799632874296e-06}
{"train_loss": 0.052968449890613556, "global_step": 221702, "epoch": 2491, "lr": 8.08948347239113e-06}
{"train_loss": 0.06331828236579895, "global_step": 221703, "epoch": 2491, "lr": 8.08916731754234e-06}
{"train_loss": 0.03210911527276039, "global_step": 221704, "epoch": 2491, "lr": 8.088851168327938e-06}
{"train_loss": 0.022294361144304276, "global_step": 221705, "epoch": 2491, "lr": 8.08853502474799e-06}
{"train_loss": 0.032261110842227936, "global_step": 221706, "epoch": 2491, "lr": 8.088218886802512e-06}
{"train_loss": 0.040876541286706924, "global_step": 221707, "epoch": 2491, "lr": 8.087902754491577e-06}
{"train_loss": 0.024340514093637466, "global_step": 221708, "epoch": 2491, "lr": 8.087586627815202e-06}
{"train_loss": 0.053456276655197144, "global_step": 221709, "epoch": 2491, "lr": 8.087270506773454e-06}
{"train_loss": 0.05524803325533867, "global_step": 221710, "epoch": 2491, "lr": 8.086954391366347e-06}
{"train_loss": 0.038240887224674225, "global_step": 221711, "epoch": 2491, "lr": 8.08663828159395e-06}
{"train_loss": 0.04976292699575424, "global_step": 221712, "epoch": 2491, "lr": 8.08632217745628e-06}
{"train_loss": 0.03028109110891819, "global_step": 221713, "epoch": 2491, "lr": 8.086006078953406e-06}
{"train_loss": 0.028606504201889038, "global_step": 221714, "epoch": 2491, "lr": 8.085689986085348e-06}
{"train_loss": 0.08744022995233536, "global_step": 221715, "epoch": 2491, "lr": 8.085373898852166e-06}
{"train_loss": 0.05262712016701698, "global_step": 221716, "epoch": 2491, "lr": 8.085057817253888e-06}
{"train_loss": 0.060881029814481735, "global_step": 221717, "epoch": 2491, "lr": 8.084741741290568e-06}
{"train_loss": 0.08454220741987228, "global_step": 221718, "epoch": 2491, "lr": 8.084425670962236e-06}
{"train_loss": 0.057132262736558914, "global_step": 221719, "epoch": 2491, "lr": 8.084109606268957e-06}
{"train_loss": 0.09158536046743393, "global_step": 221720, "epoch": 2491, "lr": 8.083793547210754e-06}
{"train_loss": 0.04443446174263954, "global_step": 221721, "epoch": 2491, "lr": 8.083477493787667e-06}
{"train_loss": 0.04538002982735634, "global_step": 221722, "epoch": 2491, "lr": 8.083161445999754e-06}
{"train_loss": 0.05755039304494858, "global_step": 221723, "epoch": 2491, "lr": 8.082845403847033e-06}
{"train_loss": 0.0950903668999672, "global_step": 221724, "epoch": 2491, "lr": 8.082529367329584e-06}
{"train_loss": 0.05522458627820015, "global_step": 221725, "epoch": 2491, "lr": 8.082213336447408e-06}
{"train_loss": 0.043778274208307266, "global_step": 221726, "epoch": 2491, "lr": 8.08189731120058e-06}
{"train_loss": 0.0636879950761795, "global_step": 221727, "epoch": 2491, "lr": 8.081581291589124e-06}
{"train_loss": 0.07028543204069138, "global_step": 221728, "epoch": 2491, "lr": 8.081265277613093e-06}
{"train_loss": 0.038477446883916855, "global_step": 221729, "epoch": 2491, "lr": 8.08094926927252e-06}
{"train_loss": 0.04460609331727028, "global_step": 221730, "epoch": 2491, "lr": 8.08063326656746e-06}
{"train_loss": 0.16013121604919434, "global_step": 221731, "epoch": 2491, "lr": 8.08031726949794e-06}
{"train_loss": 0.04495157673954964, "global_step": 221732, "epoch": 2491, "lr": 8.080001278064015e-06}
{"train_loss": 0.054201096296310425, "global_step": 221733, "epoch": 2491, "lr": 8.079685292265715e-06}
{"train_loss": 0.06933160126209259, "global_step": 221734, "epoch": 2491, "lr": 8.079369312103102e-06}
{"train_loss": 0.05457710474729538, "global_step": 221735, "epoch": 2491, "lr": 8.079053337576197e-06}
{"train_loss": 0.05122549086809158, "global_step": 221736, "epoch": 2491, "lr": 8.078737368685064e-06}
{"train_loss": 0.09808632731437683, "global_step": 221737, "epoch": 2491, "lr": 8.078421405429721e-06}
{"train_loss": 0.09977523982524872, "global_step": 221738, "epoch": 2491, "lr": 8.078105447810224e-06}
{"train_loss": 0.03927155211567879, "global_step": 221739, "epoch": 2491, "lr": 8.077789495826627e-06}
{"train_loss": 0.07295376807451248, "global_step": 221740, "epoch": 2491, "lr": 8.077473549478948e-06}
{"train_loss": 0.07126980274915695, "global_step": 221741, "epoch": 2491, "lr": 8.07715760876725e-06}
{"train_loss": 0.05124383047223091, "global_step": 221742, "epoch": 2491, "lr": 8.07684167369156e-06}
{"train_loss": 0.030569789931178093, "global_step": 221743, "epoch": 2491, "lr": 8.076525744251935e-06}
{"train_loss": 0.05065189674496651, "global_step": 221744, "epoch": 2491, "lr": 8.0762098204484e-06}
{"train_loss": 0.04999533295631409, "global_step": 221745, "epoch": 2491, "lr": 8.075893902281023e-06}
{"train_loss": 0.06568992137908936, "global_step": 221746, "epoch": 2491, "lr": 8.075577989749816e-06}
{"train_loss": 0.040900103747844696, "global_step": 221747, "epoch": 2491, "lr": 8.07526208285485e-06}
{"train_loss": 0.05100160092115402, "global_step": 221748, "epoch": 2491, "lr": 8.074946181596138e-06}
{"train_loss": 0.07480435818433762, "global_step": 221749, "epoch": 2491, "lr": 8.074630285973755e-06}
{"train_loss": 0.06291361898183823, "global_step": 221750, "epoch": 2491, "lr": 8.074314395987715e-06}
{"train_loss": 0.038991499692201614, "global_step": 221751, "epoch": 2491, "lr": 8.07399851163808e-06}
{"train_loss": 0.061002444475889206, "global_step": 221752, "epoch": 2491, "lr": 8.073682632924878e-06}
{"train_loss": 0.1351219266653061, "global_step": 221753, "epoch": 2491, "lr": 8.073366759848166e-06}
{"train_loss": 0.0560116246342659, "global_step": 221754, "epoch": 2491, "lr": 8.07305089240797e-06}
{"train_loss": 0.051963530480861664, "global_step": 221755, "epoch": 2491, "lr": 8.07273503060435e-06}
{"train_loss": 0.03417167440056801, "global_step": 221756, "epoch": 2491, "lr": 8.072419174437334e-06}
{"train_loss": 0.055625103414058685, "global_step": 221757, "epoch": 2491, "lr": 8.072103323906976e-06}
{"train_loss": 0.06323370337486267, "global_step": 221758, "epoch": 2491, "lr": 8.071787479013316e-06}
{"train_loss": 0.02937314473092556, "global_step": 221759, "epoch": 2491, "lr": 8.071471639756378e-06}
{"train_loss": 0.0777287483215332, "global_step": 221760, "epoch": 2491, "lr": 8.071155806136238e-06}
{"train_loss": 0.09842181205749512, "global_step": 221761, "epoch": 2491, "lr": 8.0708399781529e-06}
{"train_loss": 0.08133460581302643, "global_step": 221762, "epoch": 2491, "lr": 8.070524155806441e-06}
{"train_loss": 0.05080960690975189, "global_step": 221763, "epoch": 2491, "lr": 8.07020833909688e-06}
{"train_loss": 0.08799596130847931, "global_step": 221764, "epoch": 2491, "lr": 8.069892528024281e-06}
{"train_loss": 0.031931374222040176, "global_step": 221765, "epoch": 2491, "lr": 8.069576722588656e-06}
{"train_loss": 0.10692812502384186, "global_step": 221766, "epoch": 2491, "lr": 8.069260922790085e-06}
{"train_loss": 0.04485075920820236, "global_step": 221767, "epoch": 2491, "lr": 8.068945128628575e-06}
{"train_loss": 0.04662071168422699, "global_step": 221768, "epoch": 2491, "lr": 8.068629340104195e-06}
{"train_loss": 0.060845158994197845, "global_step": 221769, "epoch": 2491, "lr": 8.068313557216966e-06}
{"train_loss": 0.027793506160378456, "global_step": 221770, "epoch": 2491, "lr": 8.067997779966952e-06}
{"train_loss": 0.0867118313908577, "global_step": 221771, "epoch": 2491, "lr": 8.067682008354172e-06}
{"train_loss": 0.07139357179403305, "global_step": 221772, "epoch": 2491, "lr": 8.067366242378693e-06}
{"train_loss": 0.05083716660737991, "global_step": 221773, "epoch": 2491, "lr": 8.067050482040539e-06}
{"train_loss": 0.01704348810017109, "global_step": 221774, "epoch": 2491, "lr": 8.066734727339753e-06}
{"train_loss": 0.05170198902487755, "global_step": 221775, "epoch": 2491, "lr": 8.066418978276397e-06}
{"train_loss": 0.043376754969358444, "global_step": 221776, "epoch": 2491, "lr": 8.066103234850492e-06}
{"train_loss": 0.04087667912244797, "global_step": 221777, "epoch": 2491, "lr": 8.065787497062094e-06}
{"train_loss": 0.02572263590991497, "global_step": 221778, "epoch": 2491, "lr": 8.065471764911231e-06}
{"train_loss": 0.07201436907052994, "global_step": 221779, "epoch": 2491, "lr": 8.065156038397963e-06}
{"train_loss": 0.06428287923336029, "global_step": 221780, "epoch": 2491, "lr": 8.064840317522315e-06}
{"train_loss": 0.05391653627157211, "global_step": 221781, "epoch": 2491, "lr": 8.064524602284352e-06}
{"train_loss": 0.043771762400865555, "global_step": 221782, "epoch": 2491, "lr": 8.064208892684088e-06}
{"train_loss": 0.041397567838430405, "global_step": 221783, "epoch": 2491, "lr": 8.063893188721588e-06}
{"train_loss": 0.033291950821876526, "global_step": 221784, "epoch": 2491, "lr": 8.063577490396878e-06}
{"train_loss": 0.03734667971730232, "global_step": 221785, "epoch": 2491, "lr": 8.063261797710025e-06}
{"train_loss": 0.0392366386950016, "global_step": 221786, "epoch": 2491, "lr": 8.06294611066104e-06}
{"train_loss": 0.05731863896833377, "global_step": 221787, "epoch": 2491, "lr": 8.06263042924999e-06, "val_loss": 8.868890762329102}
{"train_loss": 0.07772809267044067, "global_step": 221788, "epoch": 2492, "lr": 8.062314753476902e-06}
{"train_loss": 0.0904545783996582, "global_step": 221789, "epoch": 2492, "lr": 8.06199908334183e-06}
{"train_loss": 0.025423159822821617, "global_step": 221790, "epoch": 2492, "lr": 8.0616834188448e-06}
{"train_loss": 0.032926201820373535, "global_step": 221791, "epoch": 2492, "lr": 8.061367759985882e-06}
{"train_loss": 0.03681861236691475, "global_step": 221792, "epoch": 2492, "lr": 8.061052106765088e-06}
{"train_loss": 0.06625889986753464, "global_step": 221793, "epoch": 2492, "lr": 8.060736459182489e-06}
{"train_loss": 0.038666293025016785, "global_step": 221794, "epoch": 2492, "lr": 8.060420817238095e-06}
{"train_loss": 0.07329797744750977, "global_step": 221795, "epoch": 2492, "lr": 8.060105180931982e-06}
{"train_loss": 0.03896009922027588, "global_step": 221796, "epoch": 2492, "lr": 8.059789550264163e-06}
{"train_loss": 0.12192709743976593, "global_step": 221797, "epoch": 2492, "lr": 8.059473925234706e-06}
{"train_loss": 0.05293788015842438, "global_step": 221798, "epoch": 2492, "lr": 8.059158305843645e-06}
{"train_loss": 0.018667563796043396, "global_step": 221799, "epoch": 2492, "lr": 8.058842692091e-06}
{"train_loss": 0.08586666733026505, "global_step": 221800, "epoch": 2492, "lr": 8.058527083976852e-06}
{"train_loss": 0.024214062839746475, "global_step": 221801, "epoch": 2492, "lr": 8.05821148150121e-06}
{"train_loss": 0.08148536086082458, "global_step": 221802, "epoch": 2492, "lr": 8.05789588466414e-06}
{"train_loss": 0.03723379597067833, "global_step": 221803, "epoch": 2492, "lr": 8.057580293465662e-06}
{"train_loss": 0.06792108714580536, "global_step": 221804, "epoch": 2492, "lr": 8.057264707905849e-06}
{"train_loss": 0.022444602102041245, "global_step": 221805, "epoch": 2492, "lr": 8.056949127984708e-06}
{"train_loss": 0.008835150860249996, "global_step": 221806, "epoch": 2492, "lr": 8.056633553702309e-06}
{"train_loss": 0.04048123583197594, "global_step": 221807, "epoch": 2492, "lr": 8.056317985058676e-06}
{"train_loss": 0.048206377774477005, "global_step": 221808, "epoch": 2492, "lr": 8.05600242205387e-06}
{"train_loss": 0.047848306596279144, "global_step": 221809, "epoch": 2492, "lr": 8.055686864687917e-06}
{"train_loss": 0.012793589383363724, "global_step": 221810, "epoch": 2492, "lr": 8.055371312960857e-06}
{"train_loss": 0.029382003471255302, "global_step": 221811, "epoch": 2492, "lr": 8.05505576687276e-06}
{"train_loss": 0.031089024618268013, "global_step": 221812, "epoch": 2492, "lr": 8.054740226423635e-06}
{"train_loss": 0.06617545336484909, "global_step": 221813, "epoch": 2492, "lr": 8.05442469161355e-06}
{"train_loss": 0.04585171490907669, "global_step": 221814, "epoch": 2492, "lr": 8.054109162442524e-06}
{"train_loss": 0.05795268341898918, "global_step": 221815, "epoch": 2492, "lr": 8.053793638910628e-06}
{"train_loss": 0.04434767737984657, "global_step": 221816, "epoch": 2492, "lr": 8.05347812101787e-06}
{"train_loss": 0.08053269982337952, "global_step": 221817, "epoch": 2492, "lr": 8.053162608764331e-06}
{"train_loss": 0.043050795793533325, "global_step": 221818, "epoch": 2492, "lr": 8.052847102150012e-06}
{"train_loss": 0.028405137360095978, "global_step": 221819, "epoch": 2492, "lr": 8.052531601174995e-06}
{"train_loss": 0.0850139930844307, "global_step": 221820, "epoch": 2492, "lr": 8.052216105839289e-06}
{"train_loss": 0.04215651750564575, "global_step": 221821, "epoch": 2492, "lr": 8.051900616142965e-06}
{"train_loss": 0.04922160878777504, "global_step": 221822, "epoch": 2492, "lr": 8.051585132086037e-06}
{"train_loss": 0.025790199637413025, "global_step": 221823, "epoch": 2492, "lr": 8.051269653668575e-06}
{"train_loss": 0.05629074573516846, "global_step": 221824, "epoch": 2492, "lr": 8.050954180890597e-06}
{"train_loss": 0.07513134926557541, "global_step": 221825, "epoch": 2492, "lr": 8.050638713752168e-06}
{"train_loss": 0.055781517177820206, "global_step": 221826, "epoch": 2492, "lr": 8.050323252253306e-06}
{"train_loss": 0.04044657200574875, "global_step": 221827, "epoch": 2492, "lr": 8.050007796394082e-06}
{"train_loss": 0.05092478543519974, "global_step": 221828, "epoch": 2492, "lr": 8.049692346174508e-06}
{"train_loss": 0.02429385483264923, "global_step": 221829, "epoch": 2492, "lr": 8.049376901594646e-06}
{"train_loss": 0.03890835866332054, "global_step": 221830, "epoch": 2492, "lr": 8.049061462654545e-06}
{"train_loss": 0.04712029919028282, "global_step": 221831, "epoch": 2492, "lr": 8.048746029354232e-06}
{"train_loss": 0.03884033486247063, "global_step": 221832, "epoch": 2492, "lr": 8.048430601693746e-06}
{"train_loss": 0.049843378365039825, "global_step": 221833, "epoch": 2492, "lr": 8.048115179673148e-06}
{"train_loss": 0.04870094731450081, "global_step": 221834, "epoch": 2492, "lr": 8.047799763292458e-06}
{"train_loss": 0.062320366501808167, "global_step": 221835, "epoch": 2492, "lr": 8.047484352551743e-06}
{"train_loss": 0.036502283066511154, "global_step": 221836, "epoch": 2492, "lr": 8.047168947451028e-06}
{"train_loss": 0.09083732217550278, "global_step": 221837, "epoch": 2492, "lr": 8.046853547990351e-06}
{"train_loss": 0.03796439990401268, "global_step": 221838, "epoch": 2492, "lr": 8.046538154169775e-06}
{"train_loss": 0.0688532292842865, "global_step": 221839, "epoch": 2492, "lr": 8.046222765989319e-06}
{"train_loss": 0.047158438712358475, "global_step": 221840, "epoch": 2492, "lr": 8.045907383449047e-06}
{"train_loss": 0.043662961572408676, "global_step": 221841, "epoch": 2492, "lr": 8.04559200654898e-06}
{"train_loss": 0.050444576889276505, "global_step": 221842, "epoch": 2492, "lr": 8.045276635289185e-06}
{"train_loss": 0.053728822618722916, "global_step": 221843, "epoch": 2492, "lr": 8.044961269669681e-06}
{"train_loss": 0.08577045798301697, "global_step": 221844, "epoch": 2492, "lr": 8.04464590969053e-06}
{"train_loss": 0.0258413627743721, "global_step": 221845, "epoch": 2492, "lr": 8.044330555351753e-06}
{"train_loss": 0.06436629593372345, "global_step": 221846, "epoch": 2492, "lr": 8.044015206653405e-06}
{"train_loss": 0.019921816885471344, "global_step": 221847, "epoch": 2492, "lr": 8.04369986359554e-06}
{"train_loss": 0.03999147191643715, "global_step": 221848, "epoch": 2492, "lr": 8.043384526178178e-06}
{"train_loss": 0.08395376801490784, "global_step": 221849, "epoch": 2492, "lr": 8.043069194401381e-06}
{"train_loss": 0.041006900370121, "global_step": 221850, "epoch": 2492, "lr": 8.042753868265168e-06}
{"train_loss": 0.04741385951638222, "global_step": 221851, "epoch": 2492, "lr": 8.04243854776961e-06}
{"train_loss": 0.051428135484457016, "global_step": 221852, "epoch": 2492, "lr": 8.042123232914723e-06}
{"train_loss": 0.042984865605831146, "global_step": 221853, "epoch": 2492, "lr": 8.041807923700573e-06}
{"train_loss": 0.07538297772407532, "global_step": 221854, "epoch": 2492, "lr": 8.041492620127179e-06}
{"train_loss": 0.055982377380132675, "global_step": 221855, "epoch": 2492, "lr": 8.041177322194605e-06}
{"train_loss": 0.042945243418216705, "global_step": 221856, "epoch": 2492, "lr": 8.040862029902874e-06}
{"train_loss": 0.053715452551841736, "global_step": 221857, "epoch": 2492, "lr": 8.040546743252048e-06}
{"train_loss": 0.04344067722558975, "global_step": 221858, "epoch": 2492, "lr": 8.040231462242142e-06}
{"train_loss": 0.046608299016952515, "global_step": 221859, "epoch": 2492, "lr": 8.039916186873237e-06}
{"train_loss": 0.0630493238568306, "global_step": 221860, "epoch": 2492, "lr": 8.039600917145334e-06}
{"train_loss": 0.02780485898256302, "global_step": 221861, "epoch": 2492, "lr": 8.03928565305851e-06}
{"train_loss": 0.056204307824373245, "global_step": 221862, "epoch": 2492, "lr": 8.038970394612782e-06}
{"train_loss": 0.03796302154660225, "global_step": 221863, "epoch": 2492, "lr": 8.038655141808216e-06}
{"train_loss": 0.04406972974538803, "global_step": 221864, "epoch": 2492, "lr": 8.038339894644826e-06}
{"train_loss": 0.06801530718803406, "global_step": 221865, "epoch": 2492, "lr": 8.038024653122673e-06}
{"train_loss": 0.05279270559549332, "global_step": 221866, "epoch": 2492, "lr": 8.037709417241807e-06}
{"train_loss": 0.03141848370432854, "global_step": 221867, "epoch": 2492, "lr": 8.037394187002244e-06}
{"train_loss": 0.013293677009642124, "global_step": 221868, "epoch": 2492, "lr": 8.03707896240406e-06}
{"train_loss": 0.054574768990278244, "global_step": 221869, "epoch": 2492, "lr": 8.036763743447273e-06}
{"train_loss": 0.08954360336065292, "global_step": 221870, "epoch": 2492, "lr": 8.036448530131924e-06}
{"train_loss": 0.044284332543611526, "global_step": 221871, "epoch": 2492, "lr": 8.036133322458073e-06}
{"train_loss": 0.08456948399543762, "global_step": 221872, "epoch": 2492, "lr": 8.035818120425742e-06}
{"train_loss": 0.09050736576318741, "global_step": 221873, "epoch": 2492, "lr": 8.035502924034993e-06}
{"train_loss": 0.08279890567064285, "global_step": 221874, "epoch": 2492, "lr": 8.035187733285865e-06}
{"train_loss": 0.02622622437775135, "global_step": 221875, "epoch": 2492, "lr": 8.03487254817838e-06}
{"train_loss": 0.050945986104145476, "global_step": 221876, "epoch": 2492, "lr": 8.034557368712603e-06, "val_loss": 9.118937492370605}
{"train_loss": 0.06457223743200302, "global_step": 221877, "epoch": 2493, "lr": 8.034242194888558e-06}
{"train_loss": 0.04046165570616722, "global_step": 221878, "epoch": 2493, "lr": 8.033927026706312e-06}
{"train_loss": 0.025625117123126984, "global_step": 221879, "epoch": 2493, "lr": 8.03361186416588e-06}
{"train_loss": 0.04249297082424164, "global_step": 221880, "epoch": 2493, "lr": 8.03329670726733e-06}
{"train_loss": 0.04112204536795616, "global_step": 221881, "epoch": 2493, "lr": 8.032981556010677e-06}
{"train_loss": 0.05007649213075638, "global_step": 221882, "epoch": 2493, "lr": 8.032666410395978e-06}
{"train_loss": 0.06862491369247437, "global_step": 221883, "epoch": 2493, "lr": 8.032351270423294e-06}
{"train_loss": 0.03531409800052643, "global_step": 221884, "epoch": 2493, "lr": 8.03203613609263e-06}
{"train_loss": 0.06289473921060562, "global_step": 221885, "epoch": 2493, "lr": 8.031721007404058e-06}
{"train_loss": 0.07460109144449234, "global_step": 221886, "epoch": 2493, "lr": 8.0314058843576e-06}
{"train_loss": 0.048637572675943375, "global_step": 221887, "epoch": 2493, "lr": 8.031090766953325e-06}
{"train_loss": 0.02460199035704136, "global_step": 221888, "epoch": 2493, "lr": 8.030775655191241e-06}
{"train_loss": 0.05296185240149498, "global_step": 221889, "epoch": 2493, "lr": 8.030460549071423e-06}
{"train_loss": 0.020402971655130386, "global_step": 221890, "epoch": 2493, "lr": 8.030145448593885e-06}
{"train_loss": 0.05981500819325447, "global_step": 221891, "epoch": 2493, "lr": 8.029830353758693e-06}
{"train_loss": 0.04158572107553482, "global_step": 221892, "epoch": 2493, "lr": 8.029515264565868e-06}
{"train_loss": 0.06350071728229523, "global_step": 221893, "epoch": 2493, "lr": 8.029200181015473e-06}
{"train_loss": 0.0685490220785141, "global_step": 221894, "epoch": 2493, "lr": 8.028885103107531e-06}
{"train_loss": 0.02516351453959942, "global_step": 221895, "epoch": 2493, "lr": 8.028570030842104e-06}
{"train_loss": 0.030710261315107346, "global_step": 221896, "epoch": 2493, "lr": 8.028254964219212e-06}
{"train_loss": 0.05508793890476227, "global_step": 221897, "epoch": 2493, "lr": 8.027939903238923e-06}
{"train_loss": 0.0758385956287384, "global_step": 221898, "epoch": 2493, "lr": 8.027624847901249e-06}
{"train_loss": 0.0837007388472557, "global_step": 221899, "epoch": 2493, "lr": 8.027309798206266e-06}
{"train_loss": 0.023775633424520493, "global_step": 221900, "epoch": 2493, "lr": 8.026994754153989e-06}
{"train_loss": 0.046349912881851196, "global_step": 221901, "epoch": 2493, "lr": 8.026679715744467e-06}
{"train_loss": 0.025243716314435005, "global_step": 221902, "epoch": 2493, "lr": 8.026364682977761e-06}
{"train_loss": 0.04867104813456535, "global_step": 221903, "epoch": 2493, "lr": 8.02604965585389e-06}
{"train_loss": 0.06428997218608856, "global_step": 221904, "epoch": 2493, "lr": 8.025734634372911e-06}
{"train_loss": 0.04467131942510605, "global_step": 221905, "epoch": 2493, "lr": 8.025419618534852e-06}
{"train_loss": 0.05945425480604172, "global_step": 221906, "epoch": 2493, "lr": 8.025104608339772e-06}
{"train_loss": 0.07507557421922684, "global_step": 221907, "epoch": 2493, "lr": 8.024789603787697e-06}
{"train_loss": 0.03979814052581787, "global_step": 221908, "epoch": 2493, "lr": 8.02447460487869e-06}
{"train_loss": 0.02427348494529724, "global_step": 221909, "epoch": 2493, "lr": 8.024159611612775e-06}
{"train_loss": 0.07488425821065903, "global_step": 221910, "epoch": 2493, "lr": 8.023844623989995e-06}
{"train_loss": 0.06967376917600632, "global_step": 221911, "epoch": 2493, "lr": 8.023529642010402e-06}
{"train_loss": 0.03351857513189316, "global_step": 221912, "epoch": 2493, "lr": 8.023214665674028e-06}
{"train_loss": 0.04937693849205971, "global_step": 221913, "epoch": 2493, "lr": 8.022899694980934e-06}
{"train_loss": 0.06304524838924408, "global_step": 221914, "epoch": 2493, "lr": 8.022584729931143e-06}
{"train_loss": 0.04053349047899246, "global_step": 221915, "epoch": 2493, "lr": 8.0222697705247e-06}
{"train_loss": 0.04895642399787903, "global_step": 221916, "epoch": 2493, "lr": 8.021954816761656e-06}
{"train_loss": 0.043480802327394485, "global_step": 221917, "epoch": 2493, "lr": 8.02163986864204e-06}
{"train_loss": 0.04085950180888176, "global_step": 221918, "epoch": 2493, "lr": 8.021324926165912e-06}
{"train_loss": 0.037004511803388596, "global_step": 221919, "epoch": 2493, "lr": 8.021009989333294e-06}
{"train_loss": 0.0813504233956337, "global_step": 221920, "epoch": 2493, "lr": 8.020695058144245e-06}
{"train_loss": 0.02322359010577202, "global_step": 221921, "epoch": 2493, "lr": 8.020380132598809e-06}
{"train_loss": 0.029339753091335297, "global_step": 221922, "epoch": 2493, "lr": 8.020065212697009e-06}
{"train_loss": 0.01913708634674549, "global_step": 221923, "epoch": 2493, "lr": 8.01975029843891e-06}
{"train_loss": 0.022640489041805267, "global_step": 221924, "epoch": 2493, "lr": 8.019435389824536e-06}
{"train_loss": 0.08802278339862823, "global_step": 221925, "epoch": 2493, "lr": 8.019120486853949e-06}
{"train_loss": 0.05017407238483429, "global_step": 221926, "epoch": 2493, "lr": 8.018805589527168e-06}
{"train_loss": 0.07409071177244186, "global_step": 221927, "epoch": 2493, "lr": 8.018490697844256e-06}
{"train_loss": 0.0652884766459465, "global_step": 221928, "epoch": 2493, "lr": 8.018175811805234e-06}
{"train_loss": 0.03649064153432846, "global_step": 221929, "epoch": 2493, "lr": 8.01786093141017e-06}
{"train_loss": 0.08035194873809814, "global_step": 221930, "epoch": 2493, "lr": 8.01754605665908e-06}
{"train_loss": 0.09831451624631882, "global_step": 221931, "epoch": 2493, "lr": 8.017231187552032e-06}
{"train_loss": 0.06045740470290184, "global_step": 221932, "epoch": 2493, "lr": 8.016916324089046e-06}
{"train_loss": 0.020113630220294, "global_step": 221933, "epoch": 2493, "lr": 8.016601466270185e-06}
{"train_loss": 0.033898837864398956, "global_step": 221934, "epoch": 2493, "lr": 8.016286614095465e-06}
{"train_loss": 0.028222395107150078, "global_step": 221935, "epoch": 2493, "lr": 8.015971767564956e-06}
{"train_loss": 0.024966957047581673, "global_step": 221936, "epoch": 2493, "lr": 8.015656926678678e-06}
{"train_loss": 0.051815636456012726, "global_step": 221937, "epoch": 2493, "lr": 8.015342091436684e-06}
{"train_loss": 0.012264206074178219, "global_step": 221938, "epoch": 2493, "lr": 8.015027261839025e-06}
{"train_loss": 0.05058316886425018, "global_step": 221939, "epoch": 2493, "lr": 8.014712437885724e-06}
{"train_loss": 0.04420732334256172, "global_step": 221940, "epoch": 2493, "lr": 8.014397619576846e-06}
{"train_loss": 0.035859812051057816, "global_step": 221941, "epoch": 2493, "lr": 8.014082806912404e-06}
{"train_loss": 0.04307562857866287, "global_step": 221942, "epoch": 2493, "lr": 8.013767999892474e-06}
{"train_loss": 0.058402471244335175, "global_step": 221943, "epoch": 2493, "lr": 8.013453198517062e-06}
{"train_loss": 0.02679751068353653, "global_step": 221944, "epoch": 2493, "lr": 8.013138402786247e-06}
{"train_loss": 0.0635572150349617, "global_step": 221945, "epoch": 2493, "lr": 8.012823612700043e-06}
{"train_loss": 0.07471007108688354, "global_step": 221946, "epoch": 2493, "lr": 8.012508828258513e-06}
{"train_loss": 0.06031155586242676, "global_step": 221947, "epoch": 2493, "lr": 8.01219404946169e-06}
{"train_loss": 0.07020875066518784, "global_step": 221948, "epoch": 2493, "lr": 8.011879276309603e-06}
{"train_loss": 0.07295581698417664, "global_step": 221949, "epoch": 2493, "lr": 8.011564508802316e-06}
{"train_loss": 0.031528279185295105, "global_step": 221950, "epoch": 2493, "lr": 8.011249746939858e-06}
{"train_loss": 0.030697762966156006, "global_step": 221951, "epoch": 2493, "lr": 8.01093499072228e-06}
{"train_loss": 0.0588376484811306, "global_step": 221952, "epoch": 2493, "lr": 8.010620240149625e-06}
{"train_loss": 0.04664553701877594, "global_step": 221953, "epoch": 2493, "lr": 8.010305495221914e-06}
{"train_loss": 0.10849819332361221, "global_step": 221954, "epoch": 2493, "lr": 8.009990755939223e-06}
{"train_loss": 0.055153001099824905, "global_step": 221955, "epoch": 2493, "lr": 8.00967602230156e-06}
{"train_loss": 0.046761270612478256, "global_step": 221956, "epoch": 2493, "lr": 8.009361294308987e-06}
{"train_loss": 0.05325620621442795, "global_step": 221957, "epoch": 2493, "lr": 8.009046571961559e-06}
{"train_loss": 0.05750415101647377, "global_step": 221958, "epoch": 2493, "lr": 8.008731855259282e-06}
{"train_loss": 0.03959858790040016, "global_step": 221959, "epoch": 2493, "lr": 8.00841714420224e-06}
{"train_loss": 0.049991581588983536, "global_step": 221960, "epoch": 2493, "lr": 8.008102438790439e-06}
{"train_loss": 0.03578838333487511, "global_step": 221961, "epoch": 2493, "lr": 8.007787739023947e-06}
{"train_loss": 0.043731939047575, "global_step": 221962, "epoch": 2493, "lr": 8.007473044902786e-06}
{"train_loss": 0.07291071116924286, "global_step": 221963, "epoch": 2493, "lr": 8.00715835642702e-06}
{"train_loss": 0.05177553743124008, "global_step": 221964, "epoch": 2493, "lr": 8.00684367359667e-06}
{"train_loss": 0.051013037279834236, "global_step": 221965, "epoch": 2493, "lr": 8.006528996411794e-06, "val_loss": 9.127938270568848}
{"train_loss": 0.06447737663984299, "global_step": 221966, "epoch": 2494, "lr": 8.006214324872424e-06}
{"train_loss": 0.025719115510582924, "global_step": 221967, "epoch": 2494, "lr": 8.00589965897861e-06}
{"train_loss": 0.06422560662031174, "global_step": 221968, "epoch": 2494, "lr": 8.005584998730386e-06}
{"train_loss": 0.09134823083877563, "global_step": 221969, "epoch": 2494, "lr": 8.005270344127808e-06}
{"train_loss": 0.0740450918674469, "global_step": 221970, "epoch": 2494, "lr": 8.004955695170901e-06}
{"train_loss": 0.011583233252167702, "global_step": 221971, "epoch": 2494, "lr": 8.004641051859724e-06}
{"train_loss": 0.07340505719184875, "global_step": 221972, "epoch": 2494, "lr": 8.004326414194302e-06}
{"train_loss": 0.033195868134498596, "global_step": 221973, "epoch": 2494, "lr": 8.004011782174686e-06}
{"train_loss": 0.043878428637981415, "global_step": 221974, "epoch": 2494, "lr": 8.003697155800927e-06}
{"train_loss": 0.04896850138902664, "global_step": 221975, "epoch": 2494, "lr": 8.00338253507305e-06}
{"train_loss": 0.01621421054005623, "global_step": 221976, "epoch": 2494, "lr": 8.00306791999112e-06}
{"train_loss": 0.07003390043973923, "global_step": 221977, "epoch": 2494, "lr": 8.00275331055515e-06}
{"train_loss": 0.09474143385887146, "global_step": 221978, "epoch": 2494, "lr": 8.002438706765213e-06}
{"train_loss": 0.04629005119204521, "global_step": 221979, "epoch": 2494, "lr": 8.002124108621323e-06}
{"train_loss": 0.0676870048046112, "global_step": 221980, "epoch": 2494, "lr": 8.001809516123548e-06}
{"train_loss": 0.09391549974679947, "global_step": 221981, "epoch": 2494, "lr": 8.001494929271908e-06}
{"train_loss": 0.036532316356897354, "global_step": 221982, "epoch": 2494, "lr": 8.00118034806646e-06}
{"train_loss": 0.045288365334272385, "global_step": 221983, "epoch": 2494, "lr": 8.000865772507237e-06}
{"train_loss": 0.05891543626785278, "global_step": 221984, "epoch": 2494, "lr": 8.000551202594293e-06}
{"train_loss": 0.0397794134914875, "global_step": 221985, "epoch": 2494, "lr": 8.000236638327668e-06}
{"train_loss": 0.048011548817157745, "global_step": 221986, "epoch": 2494, "lr": 7.999922079707384e-06}
{"train_loss": 0.0847606435418129, "global_step": 221987, "epoch": 2494, "lr": 7.99960752673351e-06}
{"train_loss": 0.04671143740415573, "global_step": 221988, "epoch": 2494, "lr": 7.999292979406065e-06}
{"train_loss": 0.059211574494838715, "global_step": 221989, "epoch": 2494, "lr": 7.998978437725113e-06}
{"train_loss": 0.11501877009868622, "global_step": 221990, "epoch": 2494, "lr": 7.99866390169069e-06}
{"train_loss": 0.04396388679742813, "global_step": 221991, "epoch": 2494, "lr": 7.998349371302822e-06}
{"train_loss": 0.035266999155282974, "global_step": 221992, "epoch": 2494, "lr": 7.998034846561563e-06}
{"train_loss": 0.09808210283517838, "global_step": 221993, "epoch": 2494, "lr": 7.997720327466967e-06}
{"train_loss": 0.047674357891082764, "global_step": 221994, "epoch": 2494, "lr": 7.99740581401906e-06}
{"train_loss": 0.05249867960810661, "global_step": 221995, "epoch": 2494, "lr": 7.997091306217896e-06}
{"train_loss": 0.04901294782757759, "global_step": 221996, "epoch": 2494, "lr": 7.9967768040635e-06}
{"train_loss": 0.08724727481603622, "global_step": 221997, "epoch": 2494, "lr": 7.996462307555935e-06}
{"train_loss": 0.08578938245773315, "global_step": 221998, "epoch": 2494, "lr": 7.996147816695227e-06}
{"train_loss": 0.04584319517016411, "global_step": 221999, "epoch": 2494, "lr": 7.995833331481434e-06}
{"train_loss": 0.027971278876066208, "global_step": 222000, "epoch": 2494, "lr": 7.99551885191458e-06}
{"train_loss": 0.03376640006899834, "global_step": 222001, "epoch": 2494, "lr": 7.995204377994724e-06}
{"train_loss": 0.03057076595723629, "global_step": 222002, "epoch": 2494, "lr": 7.994889909721891e-06}
{"train_loss": 0.023087559267878532, "global_step": 222003, "epoch": 2494, "lr": 7.994575447096148e-06}
{"train_loss": 0.03227738291025162, "global_step": 222004, "epoch": 2494, "lr": 7.994260990117508e-06}
{"train_loss": 0.02337450347840786, "global_step": 222005, "epoch": 2494, "lr": 7.993946538786035e-06}
{"train_loss": 0.06175415590405464, "global_step": 222006, "epoch": 2494, "lr": 7.99363209310176e-06}
{"train_loss": 0.05120597034692764, "global_step": 222007, "epoch": 2494, "lr": 7.993317653064735e-06}
{"train_loss": 0.06599968671798706, "global_step": 222008, "epoch": 2494, "lr": 7.993003218674989e-06}
{"train_loss": 0.06244824454188347, "global_step": 222009, "epoch": 2494, "lr": 7.992688789932567e-06}
{"train_loss": 0.0349239781498909, "global_step": 222010, "epoch": 2494, "lr": 7.992374366837529e-06}
{"train_loss": 0.02913743257522583, "global_step": 222011, "epoch": 2494, "lr": 7.992059949389896e-06}
{"train_loss": 0.05093640089035034, "global_step": 222012, "epoch": 2494, "lr": 7.99174553758973e-06}
{"train_loss": 0.024754231795668602, "global_step": 222013, "epoch": 2494, "lr": 7.991431131437044e-06}
{"train_loss": 0.04732362553477287, "global_step": 222014, "epoch": 2494, "lr": 7.991116730931919e-06}
{"train_loss": 0.0354052372276783, "global_step": 222015, "epoch": 2494, "lr": 7.990802336074354e-06}
{"train_loss": 0.02304118126630783, "global_step": 222016, "epoch": 2494, "lr": 7.990487946864434e-06}
{"train_loss": 0.043046608567237854, "global_step": 222017, "epoch": 2494, "lr": 7.990173563302166e-06}
{"train_loss": 0.0598815381526947, "global_step": 222018, "epoch": 2494, "lr": 7.989859185387622e-06}
{"train_loss": 0.03972640261054039, "global_step": 222019, "epoch": 2494, "lr": 7.989544813120813e-06}
{"train_loss": 0.057139839977025986, "global_step": 222020, "epoch": 2494, "lr": 7.98923044650181e-06}
{"train_loss": 0.051803186535835266, "global_step": 222021, "epoch": 2494, "lr": 7.988916085530634e-06}
{"train_loss": 0.0644732117652893, "global_step": 222022, "epoch": 2494, "lr": 7.988601730207345e-06}
{"train_loss": 0.07163462787866592, "global_step": 222023, "epoch": 2494, "lr": 7.988287380531966e-06}
{"train_loss": 0.033636029809713364, "global_step": 222024, "epoch": 2494, "lr": 7.987973036504558e-06}
{"train_loss": 0.0615023672580719, "global_step": 222025, "epoch": 2494, "lr": 7.987658698125161e-06}
{"train_loss": 0.054931674152612686, "global_step": 222026, "epoch": 2494, "lr": 7.987344365393795e-06}
{"train_loss": 0.04856225475668907, "global_step": 222027, "epoch": 2494, "lr": 7.987030038310533e-06}
{"train_loss": 0.023171821609139442, "global_step": 222028, "epoch": 2494, "lr": 7.986715716875393e-06}
{"train_loss": 0.04997318610548973, "global_step": 222029, "epoch": 2494, "lr": 7.986401401088433e-06}
{"train_loss": 0.05383515730500221, "global_step": 222030, "epoch": 2494, "lr": 7.986087090949679e-06}
{"train_loss": 0.024763118475675583, "global_step": 222031, "epoch": 2494, "lr": 7.9857727864592e-06}
{"train_loss": 0.03322215750813484, "global_step": 222032, "epoch": 2494, "lr": 7.985458487617003e-06}
{"train_loss": 0.08337397873401642, "global_step": 222033, "epoch": 2494, "lr": 7.985144194423167e-06}
{"train_loss": 0.039686914533376694, "global_step": 222034, "epoch": 2494, "lr": 7.9848299068777e-06}
{"train_loss": 0.1142951250076294, "global_step": 222035, "epoch": 2494, "lr": 7.984515624980676e-06}
{"train_loss": 0.04431409016251564, "global_step": 222036, "epoch": 2494, "lr": 7.984201348732106e-06}
{"train_loss": 0.045166630297899246, "global_step": 222037, "epoch": 2494, "lr": 7.983887078132063e-06}
{"train_loss": 0.08280143141746521, "global_step": 222038, "epoch": 2494, "lr": 7.983572813180562e-06}
{"train_loss": 0.054075535386800766, "global_step": 222039, "epoch": 2494, "lr": 7.98325855387767e-06}
{"train_loss": 0.06559864431619644, "global_step": 222040, "epoch": 2494, "lr": 7.9829443002234e-06}
{"train_loss": 0.12028612941503525, "global_step": 222041, "epoch": 2494, "lr": 7.982630052217827e-06}
{"train_loss": 0.079532690346241, "global_step": 222042, "epoch": 2494, "lr": 7.982315809860968e-06}
{"train_loss": 0.09308099746704102, "global_step": 222043, "epoch": 2494, "lr": 7.98200157315288e-06}
{"train_loss": 0.0332031324505806, "global_step": 222044, "epoch": 2494, "lr": 7.981687342093591e-06}
{"train_loss": 0.08687785267829895, "global_step": 222045, "epoch": 2494, "lr": 7.981373116683167e-06}
{"train_loss": 0.051678236573934555, "global_step": 222046, "epoch": 2494, "lr": 7.981058896921616e-06}
{"train_loss": 0.04558676853775978, "global_step": 222047, "epoch": 2494, "lr": 7.980744682809005e-06}
{"train_loss": 0.027164548635482788, "global_step": 222048, "epoch": 2494, "lr": 7.980430474345385e-06}
{"train_loss": 0.0656517893075943, "global_step": 222049, "epoch": 2494, "lr": 7.980116271530769e-06}
{"train_loss": 0.05005606636404991, "global_step": 222050, "epoch": 2494, "lr": 7.979802074365222e-06}
{"train_loss": 0.056877125054597855, "global_step": 222051, "epoch": 2494, "lr": 7.979487882848775e-06}
{"train_loss": 0.056002870202064514, "global_step": 222052, "epoch": 2494, "lr": 7.979173696981478e-06}
{"train_loss": 0.038616981357336044, "global_step": 222053, "epoch": 2494, "lr": 7.97885951676336e-06}
{"train_loss": 0.053933706699629845, "global_step": 222054, "epoch": 2494, "lr": 7.978545342194482e-06, "val_loss": 9.028227806091309}
{"train_loss": 0.035851720720529556, "global_step": 222055, "epoch": 2495, "lr": 7.97823117327487e-06}
{"train_loss": 0.06796880066394806, "global_step": 222056, "epoch": 2495, "lr": 7.977917010004582e-06}
{"train_loss": 0.027522575110197067, "global_step": 222057, "epoch": 2495, "lr": 7.977602852383636e-06}
{"train_loss": 0.030836166813969612, "global_step": 222058, "epoch": 2495, "lr": 7.977288700412106e-06}
{"train_loss": 0.041105616837739944, "global_step": 222059, "epoch": 2495, "lr": 7.97697455409e-06}
{"train_loss": 0.0404285229742527, "global_step": 222060, "epoch": 2495, "lr": 7.976660413417391e-06}
{"train_loss": 0.02995014190673828, "global_step": 222061, "epoch": 2495, "lr": 7.976346278394298e-06}
{"train_loss": 0.02259356714785099, "global_step": 222062, "epoch": 2495, "lr": 7.976032149020785e-06}
{"train_loss": 0.03427769988775253, "global_step": 222063, "epoch": 2495, "lr": 7.97571802529688e-06}
{"train_loss": 0.03628334403038025, "global_step": 222064, "epoch": 2495, "lr": 7.975403907222617e-06}
{"train_loss": 0.0440278984606266, "global_step": 222065, "epoch": 2495, "lr": 7.975089794798063e-06}
{"train_loss": 0.06698815524578094, "global_step": 222066, "epoch": 2495, "lr": 7.974775688023228e-06}
{"train_loss": 0.034327805042266846, "global_step": 222067, "epoch": 2495, "lr": 7.97446158689819e-06}
{"train_loss": 0.08157051354646683, "global_step": 222068, "epoch": 2495, "lr": 7.974147491422957e-06}
{"train_loss": 0.043065574020147324, "global_step": 222069, "epoch": 2495, "lr": 7.973833401597602e-06}
{"train_loss": 0.06715293228626251, "global_step": 222070, "epoch": 2495, "lr": 7.97351931742214e-06}
{"train_loss": 0.040924642235040665, "global_step": 222071, "epoch": 2495, "lr": 7.973205238896641e-06}
{"train_loss": 0.07584939897060394, "global_step": 222072, "epoch": 2495, "lr": 7.972891166021117e-06}
{"train_loss": 0.08456332236528397, "global_step": 222073, "epoch": 2495, "lr": 7.972577098795635e-06}
{"train_loss": 0.06026997044682503, "global_step": 222074, "epoch": 2495, "lr": 7.972263037220224e-06}
{"train_loss": 0.12182369828224182, "global_step": 222075, "epoch": 2495, "lr": 7.971948981294936e-06}
{"train_loss": 0.0696839690208435, "global_step": 222076, "epoch": 2495, "lr": 7.971634931019794e-06}
{"train_loss": 0.0664239376783371, "global_step": 222077, "epoch": 2495, "lr": 7.971320886394872e-06}
{"train_loss": 0.09786748886108398, "global_step": 222078, "epoch": 2495, "lr": 7.971006847420175e-06}
{"train_loss": 0.05446913093328476, "global_step": 222079, "epoch": 2495, "lr": 7.970692814095781e-06}
{"train_loss": 0.016527503728866577, "global_step": 222080, "epoch": 2495, "lr": 7.970378786421701e-06}
{"train_loss": 0.09034832566976547, "global_step": 222081, "epoch": 2495, "lr": 7.970064764398e-06}
{"train_loss": 0.05554845184087753, "global_step": 222082, "epoch": 2495, "lr": 7.969750748024702e-06}
{"train_loss": 0.0663599967956543, "global_step": 222083, "epoch": 2495, "lr": 7.96943673730186e-06}
{"train_loss": 0.097215935587883, "global_step": 222084, "epoch": 2495, "lr": 7.969122732229529e-06}
{"train_loss": 0.034781817346811295, "global_step": 222085, "epoch": 2495, "lr": 7.96880873280772e-06}
{"train_loss": 0.044624559581279755, "global_step": 222086, "epoch": 2495, "lr": 7.968494739036508e-06}
{"train_loss": 0.10243390500545502, "global_step": 222087, "epoch": 2495, "lr": 7.968180750915905e-06}
{"train_loss": 0.10104811936616898, "global_step": 222088, "epoch": 2495, "lr": 7.96786676844598e-06}
{"train_loss": 0.06957348436117172, "global_step": 222089, "epoch": 2495, "lr": 7.967552791626754e-06}
{"train_loss": 0.0363277830183506, "global_step": 222090, "epoch": 2495, "lr": 7.967238820458284e-06}
{"train_loss": 0.052280619740486145, "global_step": 222091, "epoch": 2495, "lr": 7.9669248549406e-06}
{"train_loss": 0.030602918937802315, "global_step": 222092, "epoch": 2495, "lr": 7.966610895073763e-06}
{"train_loss": 0.08444250375032425, "global_step": 222093, "epoch": 2495, "lr": 7.966296940857787e-06}
{"train_loss": 0.07533406466245651, "global_step": 222094, "epoch": 2495, "lr": 7.965982992292747e-06}
{"train_loss": 0.01176075916737318, "global_step": 222095, "epoch": 2495, "lr": 7.965669049378654e-06}
{"train_loss": 0.020934877917170525, "global_step": 222096, "epoch": 2495, "lr": 7.965355112115575e-06}
{"train_loss": 0.10155273973941803, "global_step": 222097, "epoch": 2495, "lr": 7.96504118050353e-06}
{"train_loss": 0.055057741701602936, "global_step": 222098, "epoch": 2495, "lr": 7.964727254542587e-06}
{"train_loss": 0.026901934295892715, "global_step": 222099, "epoch": 2495, "lr": 7.964413334232757e-06}
{"train_loss": 0.019878093153238297, "global_step": 222100, "epoch": 2495, "lr": 7.964099419574117e-06}
{"train_loss": 0.06165912002325058, "global_step": 222101, "epoch": 2495, "lr": 7.96378551056669e-06}
{"train_loss": 0.03884762153029442, "global_step": 222102, "epoch": 2495, "lr": 7.963471607210505e-06}
{"train_loss": 0.06167388707399368, "global_step": 222103, "epoch": 2495, "lr": 7.963157709505636e-06}
{"train_loss": 0.06913765519857407, "global_step": 222104, "epoch": 2495, "lr": 7.962843817452092e-06}
{"train_loss": 0.07527594268321991, "global_step": 222105, "epoch": 2495, "lr": 7.962529931049945e-06}
{"train_loss": 0.08341294527053833, "global_step": 222106, "epoch": 2495, "lr": 7.962216050299209e-06}
{"train_loss": 0.03168468549847603, "global_step": 222107, "epoch": 2495, "lr": 7.961902175199954e-06}
{"train_loss": 0.023451602086424828, "global_step": 222108, "epoch": 2495, "lr": 7.9615883057522e-06}
{"train_loss": 0.07141045480966568, "global_step": 222109, "epoch": 2495, "lr": 7.961274441956007e-06}
{"train_loss": 0.03944488614797592, "global_step": 222110, "epoch": 2495, "lr": 7.960960583811394e-06}
{"train_loss": 0.055912066251039505, "global_step": 222111, "epoch": 2495, "lr": 7.960646731318433e-06}
{"train_loss": 0.02407326176762581, "global_step": 222112, "epoch": 2495, "lr": 7.96033288447714e-06}
{"train_loss": 0.09716526418924332, "global_step": 222113, "epoch": 2495, "lr": 7.960019043287575e-06}
{"train_loss": 0.06426629424095154, "global_step": 222114, "epoch": 2495, "lr": 7.959705207749763e-06}
{"train_loss": 0.03306145220994949, "global_step": 222115, "epoch": 2495, "lr": 7.959391377863767e-06}
{"train_loss": 0.08037527650594711, "global_step": 222116, "epoch": 2495, "lr": 7.959077553629613e-06}
{"train_loss": 0.08305200189352036, "global_step": 222117, "epoch": 2495, "lr": 7.958763735047353e-06}
{"train_loss": 0.031183689832687378, "global_step": 222118, "epoch": 2495, "lr": 7.958449922117018e-06}
{"train_loss": 0.06783784180879593, "global_step": 222119, "epoch": 2495, "lr": 7.958136114838655e-06}
{"train_loss": 0.08493833243846893, "global_step": 222120, "epoch": 2495, "lr": 7.95782231321232e-06}
{"train_loss": 0.022568700835108757, "global_step": 222121, "epoch": 2495, "lr": 7.957508517238033e-06}
{"train_loss": 0.04171759635210037, "global_step": 222122, "epoch": 2495, "lr": 7.957194726915856e-06}
{"train_loss": 0.045201580971479416, "global_step": 222123, "epoch": 2495, "lr": 7.956880942245809e-06}
{"train_loss": 0.06362970918416977, "global_step": 222124, "epoch": 2495, "lr": 7.956567163227963e-06}
{"train_loss": 0.03273143246769905, "global_step": 222125, "epoch": 2495, "lr": 7.956253389862328e-06}
{"train_loss": 0.06298130750656128, "global_step": 222126, "epoch": 2495, "lr": 7.955939622148972e-06}
{"train_loss": 0.040570758283138275, "global_step": 222127, "epoch": 2495, "lr": 7.955625860087923e-06}
{"train_loss": 0.038918543606996536, "global_step": 222128, "epoch": 2495, "lr": 7.955312103679236e-06}
{"train_loss": 0.071779765188694, "global_step": 222129, "epoch": 2495, "lr": 7.954998352922932e-06}
{"train_loss": 0.024578362703323364, "global_step": 222130, "epoch": 2495, "lr": 7.95468460781908e-06}
{"train_loss": 0.051058243960142136, "global_step": 222131, "epoch": 2495, "lr": 7.954370868367694e-06}
{"train_loss": 0.12692083418369293, "global_step": 222132, "epoch": 2495, "lr": 7.954057134568842e-06}
{"train_loss": 0.08496645092964172, "global_step": 222133, "epoch": 2495, "lr": 7.953743406422548e-06}
{"train_loss": 0.04180552810430527, "global_step": 222134, "epoch": 2495, "lr": 7.953429683928864e-06}
{"train_loss": 0.05933474004268646, "global_step": 222135, "epoch": 2495, "lr": 7.953115967087826e-06}
{"train_loss": 0.07048586010932922, "global_step": 222136, "epoch": 2495, "lr": 7.952802255899483e-06}
{"train_loss": 0.068694107234478, "global_step": 222137, "epoch": 2495, "lr": 7.952488550363868e-06}
{"train_loss": 0.03668562322854996, "global_step": 222138, "epoch": 2495, "lr": 7.952174850481038e-06}
{"train_loss": 0.04517330229282379, "global_step": 222139, "epoch": 2495, "lr": 7.951861156251012e-06}
{"train_loss": 0.01722869463264942, "global_step": 222140, "epoch": 2495, "lr": 7.951547467673858e-06}
{"train_loss": 0.02157052606344223, "global_step": 222141, "epoch": 2495, "lr": 7.951233784749606e-06}
{"train_loss": 0.05485076084733009, "global_step": 222142, "epoch": 2495, "lr": 7.950920107478288e-06}
{"train_loss": 0.05512129277869891, "global_step": 222143, "epoch": 2495, "lr": 7.950606435859964e-06, "val_loss": 8.969278335571289, "train_action_mse_error": 11.01423454284668}
{"train_loss": 0.03145397827029228, "global_step": 222144, "epoch": 2496, "lr": 7.950292769894664e-06}
{"train_loss": 0.05722494050860405, "global_step": 222145, "epoch": 2496, "lr": 7.949979109582439e-06}
{"train_loss": 0.03532733768224716, "global_step": 222146, "epoch": 2496, "lr": 7.949665454923322e-06}
{"train_loss": 0.07358399778604507, "global_step": 222147, "epoch": 2496, "lr": 7.949351805917365e-06}
{"train_loss": 0.03266698122024536, "global_step": 222148, "epoch": 2496, "lr": 7.949038162564598e-06}
{"train_loss": 0.022021615877747536, "global_step": 222149, "epoch": 2496, "lr": 7.948724524865082e-06}
{"train_loss": 0.0966770201921463, "global_step": 222150, "epoch": 2496, "lr": 7.948410892818831e-06}
{"train_loss": 0.09549269080162048, "global_step": 222151, "epoch": 2496, "lr": 7.948097266425919e-06}
{"train_loss": 0.13064442574977875, "global_step": 222152, "epoch": 2496, "lr": 7.947783645686357e-06}
{"train_loss": 0.05834493041038513, "global_step": 222153, "epoch": 2496, "lr": 7.947470030600218e-06}
{"train_loss": 0.06174258142709732, "global_step": 222154, "epoch": 2496, "lr": 7.947156421167518e-06}
{"train_loss": 0.04769524931907654, "global_step": 222155, "epoch": 2496, "lr": 7.946842817388312e-06}
{"train_loss": 0.05841580033302307, "global_step": 222156, "epoch": 2496, "lr": 7.946529219262649e-06}
{"train_loss": 0.05284988880157471, "global_step": 222157, "epoch": 2496, "lr": 7.946215626790554e-06}
{"train_loss": 0.04972483962774277, "global_step": 222158, "epoch": 2496, "lr": 7.945902039972086e-06}
{"train_loss": 0.060541391372680664, "global_step": 222159, "epoch": 2496, "lr": 7.945588458807269e-06}
{"train_loss": 0.05183064565062523, "global_step": 222160, "epoch": 2496, "lr": 7.945274883296166e-06}
{"train_loss": 0.08655460178852081, "global_step": 222161, "epoch": 2496, "lr": 7.944961313438799e-06}
{"train_loss": 0.08244943618774414, "global_step": 222162, "epoch": 2496, "lr": 7.944647749235223e-06}
{"train_loss": 0.10043531656265259, "global_step": 222163, "epoch": 2496, "lr": 7.94433419068547e-06}
{"train_loss": 0.0436154268682003, "global_step": 222164, "epoch": 2496, "lr": 7.9440206377896e-06}
{"train_loss": 0.044504452496767044, "global_step": 222165, "epoch": 2496, "lr": 7.943707090547636e-06}
{"train_loss": 0.033781807869672775, "global_step": 222166, "epoch": 2496, "lr": 7.943393548959637e-06}
{"train_loss": 0.03215177729725838, "global_step": 222167, "epoch": 2496, "lr": 7.943080013025622e-06}
{"train_loss": 0.04703614488244057, "global_step": 222168, "epoch": 2496, "lr": 7.942766482745662e-06}
{"train_loss": 0.03757718950510025, "global_step": 222169, "epoch": 2496, "lr": 7.942452958119773e-06}
{"train_loss": 0.04847290366888046, "global_step": 222170, "epoch": 2496, "lr": 7.942139439148023e-06}
{"train_loss": 0.07001519203186035, "global_step": 222171, "epoch": 2496, "lr": 7.94182592583042e-06}
{"train_loss": 0.05614987015724182, "global_step": 222172, "epoch": 2496, "lr": 7.94151241816704e-06}
{"train_loss": 0.04113198071718216, "global_step": 222173, "epoch": 2496, "lr": 7.941198916157904e-06}
{"train_loss": 0.06707781553268433, "global_step": 222174, "epoch": 2496, "lr": 7.940885419803069e-06}
{"train_loss": 0.10194610804319382, "global_step": 222175, "epoch": 2496, "lr": 7.940571929102558e-06}
{"train_loss": 0.05072837322950363, "global_step": 222176, "epoch": 2496, "lr": 7.940258444056436e-06}
{"train_loss": 0.030385196208953857, "global_step": 222177, "epoch": 2496, "lr": 7.939944964664724e-06}
{"train_loss": 0.0485813282430172, "global_step": 222178, "epoch": 2496, "lr": 7.939631490927485e-06}
{"train_loss": 0.05678129196166992, "global_step": 222179, "epoch": 2496, "lr": 7.93931802284475e-06}
{"train_loss": 0.061749815940856934, "global_step": 222180, "epoch": 2496, "lr": 7.939004560416546e-06}
{"train_loss": 0.0968552976846695, "global_step": 222181, "epoch": 2496, "lr": 7.938691103642947e-06}
{"train_loss": 0.08026830106973648, "global_step": 222182, "epoch": 2496, "lr": 7.938377652523964e-06}
{"train_loss": 0.05309407040476799, "global_step": 222183, "epoch": 2496, "lr": 7.938064207059665e-06}
{"train_loss": 0.030981529504060745, "global_step": 222184, "epoch": 2496, "lr": 7.937750767250069e-06}
{"train_loss": 0.03500111773610115, "global_step": 222185, "epoch": 2496, "lr": 7.937437333095243e-06}
{"train_loss": 0.027438748627901077, "global_step": 222186, "epoch": 2496, "lr": 7.9371239045952e-06}
{"train_loss": 0.03661738336086273, "global_step": 222187, "epoch": 2496, "lr": 7.936810481750011e-06}
{"train_loss": 0.05932892858982086, "global_step": 222188, "epoch": 2496, "lr": 7.936497064559695e-06}
{"train_loss": 0.06268538534641266, "global_step": 222189, "epoch": 2496, "lr": 7.93618365302432e-06}
{"train_loss": 0.04360836371779442, "global_step": 222190, "epoch": 2496, "lr": 7.935870247143895e-06}
{"train_loss": 0.026379577815532684, "global_step": 222191, "epoch": 2496, "lr": 7.93555684691848e-06}
{"train_loss": 0.043317653238773346, "global_step": 222192, "epoch": 2496, "lr": 7.93524345234813e-06}
{"train_loss": 0.06567621976137161, "global_step": 222193, "epoch": 2496, "lr": 7.93493006343286e-06}
{"train_loss": 0.09893888235092163, "global_step": 222194, "epoch": 2496, "lr": 7.934616680172741e-06}
{"train_loss": 0.03670300170779228, "global_step": 222195, "epoch": 2496, "lr": 7.934303302567786e-06}
{"train_loss": 0.05016564577817917, "global_step": 222196, "epoch": 2496, "lr": 7.933989930618062e-06}
{"train_loss": 0.05526614189147949, "global_step": 222197, "epoch": 2496, "lr": 7.933676564323595e-06}
{"train_loss": 0.09079447388648987, "global_step": 222198, "epoch": 2496, "lr": 7.933363203684435e-06}
{"train_loss": 0.0314641147851944, "global_step": 222199, "epoch": 2496, "lr": 7.933049848700614e-06}
{"train_loss": 0.044805899262428284, "global_step": 222200, "epoch": 2496, "lr": 7.932736499372195e-06}
{"train_loss": 0.023062676191329956, "global_step": 222201, "epoch": 2496, "lr": 7.932423155699193e-06}
{"train_loss": 0.050473809242248535, "global_step": 222202, "epoch": 2496, "lr": 7.932109817681676e-06}
{"train_loss": 0.06681804358959198, "global_step": 222203, "epoch": 2496, "lr": 7.931796485319665e-06}
{"train_loss": 0.021402675658464432, "global_step": 222204, "epoch": 2496, "lr": 7.931483158613223e-06}
{"train_loss": 0.05622628703713417, "global_step": 222205, "epoch": 2496, "lr": 7.931169837562363e-06}
{"train_loss": 0.06110655143857002, "global_step": 222206, "epoch": 2496, "lr": 7.930856522167162e-06}
{"train_loss": 0.07693494856357574, "global_step": 222207, "epoch": 2496, "lr": 7.930543212427627e-06}
{"train_loss": 0.07407097518444061, "global_step": 222208, "epoch": 2496, "lr": 7.930229908343833e-06}
{"train_loss": 0.06091959774494171, "global_step": 222209, "epoch": 2496, "lr": 7.9299166099158e-06}
{"train_loss": 0.04771355539560318, "global_step": 222210, "epoch": 2496, "lr": 7.92960331714357e-06}
{"train_loss": 0.035969872027635574, "global_step": 222211, "epoch": 2496, "lr": 7.929290030027203e-06}
{"train_loss": 0.03902744874358177, "global_step": 222212, "epoch": 2496, "lr": 7.928976748566735e-06}
{"train_loss": 0.057156436145305634, "global_step": 222213, "epoch": 2496, "lr": 7.928663472762188e-06}
{"train_loss": 0.0409054197371006, "global_step": 222214, "epoch": 2496, "lr": 7.928350202613632e-06}
{"train_loss": 0.036607157438993454, "global_step": 222215, "epoch": 2496, "lr": 7.928036938121086e-06}
{"train_loss": 0.05216468870639801, "global_step": 222216, "epoch": 2496, "lr": 7.927723679284616e-06}
{"train_loss": 0.043851085007190704, "global_step": 222217, "epoch": 2496, "lr": 7.927410426104248e-06}
{"train_loss": 0.04742153361439705, "global_step": 222218, "epoch": 2496, "lr": 7.927097178580018e-06}
{"train_loss": 0.06293044239282608, "global_step": 222219, "epoch": 2496, "lr": 7.926783936711985e-06}
{"train_loss": 0.0673196017742157, "global_step": 222220, "epoch": 2496, "lr": 7.926470700500172e-06}
{"train_loss": 0.03236830607056618, "global_step": 222221, "epoch": 2496, "lr": 7.926157469944644e-06}
{"train_loss": 0.067696712911129, "global_step": 222222, "epoch": 2496, "lr": 7.925844245045422e-06}
{"train_loss": 0.09490800648927689, "global_step": 222223, "epoch": 2496, "lr": 7.925531025802569e-06}
{"train_loss": 0.044686198234558105, "global_step": 222224, "epoch": 2496, "lr": 7.925217812216102e-06}
{"train_loss": 0.03605031222105026, "global_step": 222225, "epoch": 2496, "lr": 7.924904604286088e-06}
{"train_loss": 0.05674697831273079, "global_step": 222226, "epoch": 2496, "lr": 7.924591402012544e-06}
{"train_loss": 0.05812491476535797, "global_step": 222227, "epoch": 2496, "lr": 7.92427820539553e-06}
{"train_loss": 0.06166824698448181, "global_step": 222228, "epoch": 2496, "lr": 7.923965014435098e-06}
{"train_loss": 0.02941455878317356, "global_step": 222229, "epoch": 2496, "lr": 7.923651829131256e-06}
{"train_loss": 0.043253082782030106, "global_step": 222230, "epoch": 2496, "lr": 7.923338649484086e-06}
{"train_loss": 0.10556169599294662, "global_step": 222231, "epoch": 2496, "lr": 7.923025475493595e-06}
{"train_loss": 0.05539763455143135, "global_step": 222232, "epoch": 2496, "lr": 7.922712307159851e-06, "val_loss": 9.001330375671387}
{"train_loss": 0.04342149198055267, "global_step": 222233, "epoch": 2497, "lr": 7.922399144482873e-06}
{"train_loss": 0.06718908995389938, "global_step": 222234, "epoch": 2497, "lr": 7.92208598746273e-06}
{"train_loss": 0.07064242660999298, "global_step": 222235, "epoch": 2497, "lr": 7.921772836099439e-06}
{"train_loss": 0.06933356076478958, "global_step": 222236, "epoch": 2497, "lr": 7.921459690393064e-06}
{"train_loss": 0.04901527985930443, "global_step": 222237, "epoch": 2497, "lr": 7.921146550343622e-06}
{"train_loss": 0.023880362510681152, "global_step": 222238, "epoch": 2497, "lr": 7.92083341595118e-06}
{"train_loss": 0.02543441765010357, "global_step": 222239, "epoch": 2497, "lr": 7.92052028721576e-06}
{"train_loss": 0.03464232385158539, "global_step": 222240, "epoch": 2497, "lr": 7.920207164137423e-06}
{"train_loss": 0.021857306361198425, "global_step": 222241, "epoch": 2497, "lr": 7.919894046716192e-06}
{"train_loss": 0.03557663783431053, "global_step": 222242, "epoch": 2497, "lr": 7.919580934952126e-06}
{"train_loss": 0.057584989815950394, "global_step": 222243, "epoch": 2497, "lr": 7.919267828845251e-06}
{"train_loss": 0.04087246209383011, "global_step": 222244, "epoch": 2497, "lr": 7.918954728395628e-06}
{"train_loss": 0.038682207465171814, "global_step": 222245, "epoch": 2497, "lr": 7.91864163360328e-06}
{"train_loss": 0.05690309405326843, "global_step": 222246, "epoch": 2497, "lr": 7.91832854446825e-06}
{"train_loss": 0.03311658278107643, "global_step": 222247, "epoch": 2497, "lr": 7.918015460990608e-06}
{"train_loss": 0.0443364754319191, "global_step": 222248, "epoch": 2497, "lr": 7.91770238317036e-06}
{"train_loss": 0.051022544503211975, "global_step": 222249, "epoch": 2497, "lr": 7.917389311007579e-06}
{"train_loss": 0.07333416491746902, "global_step": 222250, "epoch": 2497, "lr": 7.917076244502274e-06}
{"train_loss": 0.07340335100889206, "global_step": 222251, "epoch": 2497, "lr": 7.916763183654525e-06}
{"train_loss": 0.02646559290587902, "global_step": 222252, "epoch": 2497, "lr": 7.916450128464348e-06}
{"train_loss": 0.018314342945814133, "global_step": 222253, "epoch": 2497, "lr": 7.916137078931785e-06}
{"train_loss": 0.020700518041849136, "global_step": 222254, "epoch": 2497, "lr": 7.91582403505689e-06}
{"train_loss": 0.07652591168880463, "global_step": 222255, "epoch": 2497, "lr": 7.915510996839692e-06}
{"train_loss": 0.11230660229921341, "global_step": 222256, "epoch": 2497, "lr": 7.915197964280253e-06}
{"train_loss": 0.06271559000015259, "global_step": 222257, "epoch": 2497, "lr": 7.914884937378597e-06}
{"train_loss": 0.05167322978377342, "global_step": 222258, "epoch": 2497, "lr": 7.914571916134767e-06}
{"train_loss": 0.0688764899969101, "global_step": 222259, "epoch": 2497, "lr": 7.914258900548816e-06}
{"train_loss": 0.10044823586940765, "global_step": 222260, "epoch": 2497, "lr": 7.913945890620772e-06}
{"train_loss": 0.022533057257533073, "global_step": 222261, "epoch": 2497, "lr": 7.913632886350691e-06}
{"train_loss": 0.04263709858059883, "global_step": 222262, "epoch": 2497, "lr": 7.913319887738603e-06}
{"train_loss": 0.061629489064216614, "global_step": 222263, "epoch": 2497, "lr": 7.913006894784558e-06}
{"train_loss": 0.051976218819618225, "global_step": 222264, "epoch": 2497, "lr": 7.912693907488605e-06}
{"train_loss": 0.020678287371993065, "global_step": 222265, "epoch": 2497, "lr": 7.912380925850765e-06}
{"train_loss": 0.028833474963903427, "global_step": 222266, "epoch": 2497, "lr": 7.912067949871105e-06}
{"train_loss": 0.051024436950683594, "global_step": 222267, "epoch": 2497, "lr": 7.911754979549641e-06}
{"train_loss": 0.0347236730158329, "global_step": 222268, "epoch": 2497, "lr": 7.911442014886438e-06}
{"train_loss": 0.039638713002204895, "global_step": 222269, "epoch": 2497, "lr": 7.911129055881522e-06}
{"train_loss": 0.0799894779920578, "global_step": 222270, "epoch": 2497, "lr": 7.910816102534952e-06}
{"train_loss": 0.05504818633198738, "global_step": 222271, "epoch": 2497, "lr": 7.910503154846748e-06}
{"train_loss": 0.023620810359716415, "global_step": 222272, "epoch": 2497, "lr": 7.910190212816976e-06}
{"train_loss": 0.02728995308279991, "global_step": 222273, "epoch": 2497, "lr": 7.909877276445654e-06}
{"train_loss": 0.05987727642059326, "global_step": 222274, "epoch": 2497, "lr": 7.90956434573285e-06}
{"train_loss": 0.07616870105266571, "global_step": 222275, "epoch": 2497, "lr": 7.909251420678576e-06}
{"train_loss": 0.060321416705846786, "global_step": 222276, "epoch": 2497, "lr": 7.908938501282903e-06}
{"train_loss": 0.04571625590324402, "global_step": 222277, "epoch": 2497, "lr": 7.908625587545848e-06}
{"train_loss": 0.05684906616806984, "global_step": 222278, "epoch": 2497, "lr": 7.908312679467484e-06}
{"train_loss": 0.06411471962928772, "global_step": 222279, "epoch": 2497, "lr": 7.907999777047814e-06}
{"train_loss": 0.05799500644207001, "global_step": 222280, "epoch": 2497, "lr": 7.907686880286912e-06}
{"train_loss": 0.025231288745999336, "global_step": 222281, "epoch": 2497, "lr": 7.907373989184803e-06}
{"train_loss": 0.03446812555193901, "global_step": 222282, "epoch": 2497, "lr": 7.907061103741532e-06}
{"train_loss": 0.04288166016340256, "global_step": 222283, "epoch": 2497, "lr": 7.90674822395715e-06}
{"train_loss": 0.03327834606170654, "global_step": 222284, "epoch": 2497, "lr": 7.906435349831687e-06}
{"train_loss": 0.054099567234516144, "global_step": 222285, "epoch": 2497, "lr": 7.906122481365203e-06}
{"train_loss": 0.10521692037582397, "global_step": 222286, "epoch": 2497, "lr": 7.905809618557713e-06}
{"train_loss": 0.09626504778862, "global_step": 222287, "epoch": 2497, "lr": 7.905496761409287e-06}
{"train_loss": 0.07888133823871613, "global_step": 222288, "epoch": 2497, "lr": 7.905183909919938e-06}
{"train_loss": 0.050389159470796585, "global_step": 222289, "epoch": 2497, "lr": 7.904871064089742e-06}
{"train_loss": 0.05717924237251282, "global_step": 222290, "epoch": 2497, "lr": 7.904558223918724e-06}
{"train_loss": 0.056770727038383484, "global_step": 222291, "epoch": 2497, "lr": 7.904245389406905e-06}
{"train_loss": 0.04585176706314087, "global_step": 222292, "epoch": 2497, "lr": 7.903932560554366e-06}
{"train_loss": 0.08659982681274414, "global_step": 222293, "epoch": 2497, "lr": 7.903619737361117e-06}
{"train_loss": 0.024063769727945328, "global_step": 222294, "epoch": 2497, "lr": 7.903306919827226e-06}
{"train_loss": 0.05612412095069885, "global_step": 222295, "epoch": 2497, "lr": 7.902994107952717e-06}
{"train_loss": 0.05575529485940933, "global_step": 222296, "epoch": 2497, "lr": 7.90268130173763e-06}
{"train_loss": 0.06642124056816101, "global_step": 222297, "epoch": 2497, "lr": 7.902368501182022e-06}
{"train_loss": 0.05617908015847206, "global_step": 222298, "epoch": 2497, "lr": 7.902055706285921e-06}
{"train_loss": 0.03988698497414589, "global_step": 222299, "epoch": 2497, "lr": 7.901742917049388e-06}
{"train_loss": 0.04377971217036247, "global_step": 222300, "epoch": 2497, "lr": 7.901430133472437e-06}
{"train_loss": 0.028091242536902428, "global_step": 222301, "epoch": 2497, "lr": 7.901117355555127e-06}
{"train_loss": 0.08931111544370651, "global_step": 222302, "epoch": 2497, "lr": 7.900804583297506e-06}
{"train_loss": 0.04878884181380272, "global_step": 222303, "epoch": 2497, "lr": 7.900491816699602e-06}
{"train_loss": 0.05551034212112427, "global_step": 222304, "epoch": 2497, "lr": 7.900179055761475e-06}
{"train_loss": 0.04759220406413078, "global_step": 222305, "epoch": 2497, "lr": 7.899866300483145e-06}
{"train_loss": 0.06661580502986908, "global_step": 222306, "epoch": 2497, "lr": 7.899553550864675e-06}
{"train_loss": 0.08603174239397049, "global_step": 222307, "epoch": 2497, "lr": 7.899240806906083e-06}
{"train_loss": 0.036458734422922134, "global_step": 222308, "epoch": 2497, "lr": 7.898928068607436e-06}
{"train_loss": 0.02930944226682186, "global_step": 222309, "epoch": 2497, "lr": 7.898615335968756e-06}
{"train_loss": 0.047507286071777344, "global_step": 222310, "epoch": 2497, "lr": 7.89830260899011e-06}
{"train_loss": 0.01448844838887453, "global_step": 222311, "epoch": 2497, "lr": 7.897989887671509e-06}
{"train_loss": 0.09594344347715378, "global_step": 222312, "epoch": 2497, "lr": 7.897677172013018e-06}
{"train_loss": 0.0473061166703701, "global_step": 222313, "epoch": 2497, "lr": 7.897364462014661e-06}
{"train_loss": 0.06381607055664062, "global_step": 222314, "epoch": 2497, "lr": 7.897051757676504e-06}
{"train_loss": 0.040655553340911865, "global_step": 222315, "epoch": 2497, "lr": 7.896739058998564e-06}
{"train_loss": 0.05307049676775932, "global_step": 222316, "epoch": 2497, "lr": 7.896426365980902e-06}
{"train_loss": 0.0533866249024868, "global_step": 222317, "epoch": 2497, "lr": 7.896113678623545e-06}
{"train_loss": 0.03142879903316498, "global_step": 222318, "epoch": 2497, "lr": 7.895800996926539e-06}
{"train_loss": 0.045785315334796906, "global_step": 222319, "epoch": 2497, "lr": 7.895488320889943e-06}
{"train_loss": 0.058717306703329086, "global_step": 222320, "epoch": 2497, "lr": 7.895175650513775e-06}
{"train_loss": 0.05139014328924123, "global_step": 222321, "epoch": 2497, "lr": 7.894862985798102e-06, "val_loss": 8.823203086853027}
{"train_loss": 0.0365593396127224, "global_step": 222322, "epoch": 2498, "lr": 7.894550326742933e-06}
{"train_loss": 0.022609403356909752, "global_step": 222323, "epoch": 2498, "lr": 7.894237673348343e-06}
{"train_loss": 0.07127578556537628, "global_step": 222324, "epoch": 2498, "lr": 7.893925025614347e-06}
{"train_loss": 0.05555728077888489, "global_step": 222325, "epoch": 2498, "lr": 7.89361238354101e-06}
{"train_loss": 0.035770170390605927, "global_step": 222326, "epoch": 2498, "lr": 7.893299747128357e-06}
{"train_loss": 0.028851991519331932, "global_step": 222327, "epoch": 2498, "lr": 7.892987116376443e-06}
{"train_loss": 0.05241936817765236, "global_step": 222328, "epoch": 2498, "lr": 7.892674491285307e-06}
{"train_loss": 0.055735401809215546, "global_step": 222329, "epoch": 2498, "lr": 7.892361871854976e-06}
{"train_loss": 0.04099619761109352, "global_step": 222330, "epoch": 2498, "lr": 7.89204925808551e-06}
{"train_loss": 0.04922980070114136, "global_step": 222331, "epoch": 2498, "lr": 7.891736649976939e-06}
{"train_loss": 0.03284775838255882, "global_step": 222332, "epoch": 2498, "lr": 7.891424047529322e-06}
{"train_loss": 0.05403302237391472, "global_step": 222333, "epoch": 2498, "lr": 7.891111450742689e-06}
{"train_loss": 0.054129716008901596, "global_step": 222334, "epoch": 2498, "lr": 7.89079885961707e-06}
{"train_loss": 0.07973956316709518, "global_step": 222335, "epoch": 2498, "lr": 7.89048627415253e-06}
{"train_loss": 0.060031499713659286, "global_step": 222336, "epoch": 2498, "lr": 7.89017369434909e-06}
{"train_loss": 0.041070666164159775, "global_step": 222337, "epoch": 2498, "lr": 7.889861120206805e-06}
{"train_loss": 0.08125929534435272, "global_step": 222338, "epoch": 2498, "lr": 7.889548551725728e-06}
{"train_loss": 0.08202219754457474, "global_step": 222339, "epoch": 2498, "lr": 7.889235988905875e-06}
{"train_loss": 0.05972955375909805, "global_step": 222340, "epoch": 2498, "lr": 7.888923431747307e-06}
{"train_loss": 0.050458379089832306, "global_step": 222341, "epoch": 2498, "lr": 7.888610880250057e-06}
{"train_loss": 0.03829379007220268, "global_step": 222342, "epoch": 2498, "lr": 7.88829833441418e-06}
{"train_loss": 0.0492275096476078, "global_step": 222343, "epoch": 2498, "lr": 7.887985794239695e-06}
{"train_loss": 0.030361128970980644, "global_step": 222344, "epoch": 2498, "lr": 7.887673259726663e-06}
{"train_loss": 0.0442926362156868, "global_step": 222345, "epoch": 2498, "lr": 7.887360730875115e-06}
{"train_loss": 0.07255537807941437, "global_step": 222346, "epoch": 2498, "lr": 7.887048207685105e-06}
{"train_loss": 0.03349284827709198, "global_step": 222347, "epoch": 2498, "lr": 7.88673569015666e-06}
{"train_loss": 0.07764789462089539, "global_step": 222348, "epoch": 2498, "lr": 7.886423178289842e-06}
{"train_loss": 0.02746415324509144, "global_step": 222349, "epoch": 2498, "lr": 7.886110672084668e-06}
{"train_loss": 0.04159113019704819, "global_step": 222350, "epoch": 2498, "lr": 7.885798171541209e-06}
{"train_loss": 0.07775100320577621, "global_step": 222351, "epoch": 2498, "lr": 7.885485676659471e-06}
{"train_loss": 0.030810102820396423, "global_step": 222352, "epoch": 2498, "lr": 7.885173187439538e-06}
{"train_loss": 0.05021117627620697, "global_step": 222353, "epoch": 2498, "lr": 7.884860703881408e-06}
{"train_loss": 0.028533412143588066, "global_step": 222354, "epoch": 2498, "lr": 7.88454822598515e-06}
{"train_loss": 0.02711699716746807, "global_step": 222355, "epoch": 2498, "lr": 7.884235753750819e-06}
{"train_loss": 0.044230516999959946, "global_step": 222356, "epoch": 2498, "lr": 7.883923287178418e-06}
{"train_loss": 0.09355629235506058, "global_step": 222357, "epoch": 2498, "lr": 7.883610826268029e-06}
{"train_loss": 0.020189302042126656, "global_step": 222358, "epoch": 2498, "lr": 7.883298371019654e-06}
{"train_loss": 0.029767807573080063, "global_step": 222359, "epoch": 2498, "lr": 7.882985921433378e-06}
{"train_loss": 0.05494927614927292, "global_step": 222360, "epoch": 2498, "lr": 7.882673477509206e-06}
{"train_loss": 0.05528407543897629, "global_step": 222361, "epoch": 2498, "lr": 7.882361039247205e-06}
{"train_loss": 0.04330724477767944, "global_step": 222362, "epoch": 2498, "lr": 7.882048606647397e-06}
{"train_loss": 0.053351450711488724, "global_step": 222363, "epoch": 2498, "lr": 7.881736179709847e-06}
{"train_loss": 0.07576068490743637, "global_step": 222364, "epoch": 2498, "lr": 7.88142375843457e-06}
{"train_loss": 0.08611326664686203, "global_step": 222365, "epoch": 2498, "lr": 7.881111342821633e-06}
{"train_loss": 0.05635974183678627, "global_step": 222366, "epoch": 2498, "lr": 7.880798932871058e-06}
{"train_loss": 0.06321550160646439, "global_step": 222367, "epoch": 2498, "lr": 7.880486528582908e-06}
{"train_loss": 0.029907893389463425, "global_step": 222368, "epoch": 2498, "lr": 7.880174129957212e-06}
{"train_loss": 0.05340355634689331, "global_step": 222369, "epoch": 2498, "lr": 7.879861736994005e-06}
{"train_loss": 0.03655940294265747, "global_step": 222370, "epoch": 2498, "lr": 7.879549349693343e-06}
{"train_loss": 0.03343143314123154, "global_step": 222371, "epoch": 2498, "lr": 7.879236968055259e-06}
{"train_loss": 0.042359914630651474, "global_step": 222372, "epoch": 2498, "lr": 7.8789245920798e-06}
{"train_loss": 0.09250973910093307, "global_step": 222373, "epoch": 2498, "lr": 7.878612221767e-06}
{"train_loss": 0.10838843137025833, "global_step": 222374, "epoch": 2498, "lr": 7.878299857116921e-06}
{"train_loss": 0.028284423053264618, "global_step": 222375, "epoch": 2498, "lr": 7.877987498129574e-06}
{"train_loss": 0.04433630779385567, "global_step": 222376, "epoch": 2498, "lr": 7.877675144805035e-06}
{"train_loss": 0.03713307902216911, "global_step": 222377, "epoch": 2498, "lr": 7.877362797143311e-06}
{"train_loss": 0.061008308082818985, "global_step": 222378, "epoch": 2498, "lr": 7.87705045514448e-06}
{"train_loss": 0.042707886546850204, "global_step": 222379, "epoch": 2498, "lr": 7.876738118808552e-06}
{"train_loss": 0.042081329971551895, "global_step": 222380, "epoch": 2498, "lr": 7.876425788135593e-06}
{"train_loss": 0.03566668555140495, "global_step": 222381, "epoch": 2498, "lr": 7.876113463125628e-06}
{"train_loss": 0.05920105800032616, "global_step": 222382, "epoch": 2498, "lr": 7.875801143778716e-06}
{"train_loss": 0.029284602031111717, "global_step": 222383, "epoch": 2498, "lr": 7.875488830094878e-06}
{"train_loss": 0.034310486167669296, "global_step": 222384, "epoch": 2498, "lr": 7.87517652207418e-06}
{"train_loss": 0.06354844570159912, "global_step": 222385, "epoch": 2498, "lr": 7.874864219716633e-06}
{"train_loss": 0.06850621104240417, "global_step": 222386, "epoch": 2498, "lr": 7.874551923022316e-06}
{"train_loss": 0.025739232078194618, "global_step": 222387, "epoch": 2498, "lr": 7.874239631991236e-06}
{"train_loss": 0.04840456321835518, "global_step": 222388, "epoch": 2498, "lr": 7.873927346623461e-06}
{"train_loss": 0.03499344363808632, "global_step": 222389, "epoch": 2498, "lr": 7.873615066919016e-06}
{"train_loss": 0.04204725846648216, "global_step": 222390, "epoch": 2498, "lr": 7.873302792877945e-06}
{"train_loss": 0.06327829509973526, "global_step": 222391, "epoch": 2498, "lr": 7.872990524500312e-06}
{"train_loss": 0.06304767727851868, "global_step": 222392, "epoch": 2498, "lr": 7.872678261786132e-06}
{"train_loss": 0.03228158876299858, "global_step": 222393, "epoch": 2498, "lr": 7.87236600473546e-06}
{"train_loss": 0.03935060650110245, "global_step": 222394, "epoch": 2498, "lr": 7.87205375334833e-06}
{"train_loss": 0.055711790919303894, "global_step": 222395, "epoch": 2498, "lr": 7.871741507624797e-06}
{"train_loss": 0.04178931936621666, "global_step": 222396, "epoch": 2498, "lr": 7.871429267564889e-06}
{"train_loss": 0.07011394947767258, "global_step": 222397, "epoch": 2498, "lr": 7.871117033168656e-06}
{"train_loss": 0.06879188120365143, "global_step": 222398, "epoch": 2498, "lr": 7.870804804436133e-06}
{"train_loss": 0.07733769714832306, "global_step": 222399, "epoch": 2498, "lr": 7.870492581367377e-06}
{"train_loss": 0.02920021302998066, "global_step": 222400, "epoch": 2498, "lr": 7.870180363962409e-06}
{"train_loss": 0.1003381684422493, "global_step": 222401, "epoch": 2498, "lr": 7.869868152221294e-06}
{"train_loss": 0.056937478482723236, "global_step": 222402, "epoch": 2498, "lr": 7.869555946144047e-06}
{"train_loss": 0.03464852273464203, "global_step": 222403, "epoch": 2498, "lr": 7.869243745730743e-06}
{"train_loss": 0.05028941482305527, "global_step": 222404, "epoch": 2498, "lr": 7.868931550981385e-06}
{"train_loss": 0.049371249973773956, "global_step": 222405, "epoch": 2498, "lr": 7.868619361896051e-06}
{"train_loss": 0.070893794298172, "global_step": 222406, "epoch": 2498, "lr": 7.868307178474771e-06}
{"train_loss": 0.061072416603565216, "global_step": 222407, "epoch": 2498, "lr": 7.867995000717565e-06}
{"train_loss": 0.03179854527115822, "global_step": 222408, "epoch": 2498, "lr": 7.867682828624512e-06}
{"train_loss": 0.04727013409137726, "global_step": 222409, "epoch": 2498, "lr": 7.867370662195623e-06}
{"train_loss": 0.05125180083546746, "global_step": 222410, "epoch": 2498, "lr": 7.867058501430957e-06, "val_loss": 9.060918807983398}
{"train_loss": 0.08927261084318161, "global_step": 222411, "epoch": 2499, "lr": 7.866746346330545e-06}
{"train_loss": 0.10269108414649963, "global_step": 222412, "epoch": 2499, "lr": 7.866434196894451e-06}
{"train_loss": 0.028432121500372887, "global_step": 222413, "epoch": 2499, "lr": 7.866122053122682e-06}
{"train_loss": 0.1014733836054802, "global_step": 222414, "epoch": 2499, "lr": 7.865809915015315e-06}
{"train_loss": 0.03898949176073074, "global_step": 222415, "epoch": 2499, "lr": 7.865497782572367e-06}
{"train_loss": 0.06826207041740417, "global_step": 222416, "epoch": 2499, "lr": 7.8651856557939e-06}
{"train_loss": 0.04475231096148491, "global_step": 222417, "epoch": 2499, "lr": 7.864873534679928e-06}
{"train_loss": 0.08403655886650085, "global_step": 222418, "epoch": 2499, "lr": 7.864561419230527e-06}
{"train_loss": 0.046740543097257614, "global_step": 222419, "epoch": 2499, "lr": 7.864249309445709e-06}
{"train_loss": 0.05945061892271042, "global_step": 222420, "epoch": 2499, "lr": 7.863937205325539e-06}
{"train_loss": 0.02632983773946762, "global_step": 222421, "epoch": 2499, "lr": 7.863625106870042e-06}
{"train_loss": 0.022801510989665985, "global_step": 222422, "epoch": 2499, "lr": 7.863313014079277e-06}
{"train_loss": 0.0648035779595375, "global_step": 222423, "epoch": 2499, "lr": 7.863000926953263e-06}
{"train_loss": 0.05469244718551636, "global_step": 222424, "epoch": 2499, "lr": 7.862688845492066e-06}
{"train_loss": 0.03190983831882477, "global_step": 222425, "epoch": 2499, "lr": 7.862376769695706e-06}
{"train_loss": 0.03025228902697563, "global_step": 222426, "epoch": 2499, "lr": 7.862064699564247e-06}
{"train_loss": 0.053647127002477646, "global_step": 222427, "epoch": 2499, "lr": 7.861752635097708e-06}
{"train_loss": 0.061169352382421494, "global_step": 222428, "epoch": 2499, "lr": 7.861440576296147e-06}
{"train_loss": 0.02912908047437668, "global_step": 222429, "epoch": 2499, "lr": 7.861128523159612e-06}
{"train_loss": 0.042895976454019547, "global_step": 222430, "epoch": 2499, "lr": 7.860816475688121e-06}
{"train_loss": 0.049160074442625046, "global_step": 222431, "epoch": 2499, "lr": 7.860504433881737e-06}
{"train_loss": 0.052683062851428986, "global_step": 222432, "epoch": 2499, "lr": 7.86019239774049e-06}
{"train_loss": 0.13315877318382263, "global_step": 222433, "epoch": 2499, "lr": 7.859880367264432e-06}
{"train_loss": 0.04833933338522911, "global_step": 222434, "epoch": 2499, "lr": 7.859568342453594e-06}
{"train_loss": 0.06672804057598114, "global_step": 222435, "epoch": 2499, "lr": 7.859256323308035e-06}
{"train_loss": 0.06649944186210632, "global_step": 222436, "epoch": 2499, "lr": 7.85894430982777e-06}
{"train_loss": 0.03940294682979584, "global_step": 222437, "epoch": 2499, "lr": 7.858632302012869e-06}
{"train_loss": 0.10021903365850449, "global_step": 222438, "epoch": 2499, "lr": 7.858320299863348e-06}
{"train_loss": 0.042598579078912735, "global_step": 222439, "epoch": 2499, "lr": 7.858008303379277e-06}
{"train_loss": 0.07577050477266312, "global_step": 222440, "epoch": 2499, "lr": 7.85769631256067e-06}
{"train_loss": 0.05629243701696396, "global_step": 222441, "epoch": 2499, "lr": 7.85738432740759e-06}
{"train_loss": 0.06995005905628204, "global_step": 222442, "epoch": 2499, "lr": 7.857072347920064e-06}
{"train_loss": 0.043082911521196365, "global_step": 222443, "epoch": 2499, "lr": 7.856760374098154e-06}
{"train_loss": 0.0700177475810051, "global_step": 222444, "epoch": 2499, "lr": 7.856448405941885e-06}
{"train_loss": 0.017459196969866753, "global_step": 222445, "epoch": 2499, "lr": 7.856136443451295e-06}
{"train_loss": 0.02143023908138275, "global_step": 222446, "epoch": 2499, "lr": 7.855824486626445e-06}
{"train_loss": 0.03269259259104729, "global_step": 222447, "epoch": 2499, "lr": 7.855512535467352e-06}
{"train_loss": 0.02368752285838127, "global_step": 222448, "epoch": 2499, "lr": 7.855200589974083e-06}
{"train_loss": 0.07531871646642685, "global_step": 222449, "epoch": 2499, "lr": 7.854888650146658e-06}
{"train_loss": 0.06340482831001282, "global_step": 222450, "epoch": 2499, "lr": 7.85457671598514e-06}
{"train_loss": 0.04931766539812088, "global_step": 222451, "epoch": 2499, "lr": 7.854264787489552e-06}
{"train_loss": 0.05771889165043831, "global_step": 222452, "epoch": 2499, "lr": 7.853952864659952e-06}
{"train_loss": 0.030020732432603836, "global_step": 222453, "epoch": 2499, "lr": 7.853640947496366e-06}
{"train_loss": 0.07781493663787842, "global_step": 222454, "epoch": 2499, "lr": 7.853329035998852e-06}
{"train_loss": 0.026902083307504654, "global_step": 222455, "epoch": 2499, "lr": 7.853017130167433e-06}
{"train_loss": 0.04439177364110947, "global_step": 222456, "epoch": 2499, "lr": 7.852705230002177e-06}
{"train_loss": 0.0704229325056076, "global_step": 222457, "epoch": 2499, "lr": 7.8523933355031e-06}
{"train_loss": 0.06579435616731644, "global_step": 222458, "epoch": 2499, "lr": 7.852081446670262e-06}
{"train_loss": 0.0677519142627716, "global_step": 222459, "epoch": 2499, "lr": 7.851769563503691e-06}
{"train_loss": 0.020160630345344543, "global_step": 222460, "epoch": 2499, "lr": 7.851457686003444e-06}
{"train_loss": 0.10286989063024521, "global_step": 222461, "epoch": 2499, "lr": 7.851145814169548e-06}
{"train_loss": 0.045720696449279785, "global_step": 222462, "epoch": 2499, "lr": 7.850833948002056e-06}
{"train_loss": 0.05866749584674835, "global_step": 222463, "epoch": 2499, "lr": 7.850522087501e-06}
{"train_loss": 0.028045523911714554, "global_step": 222464, "epoch": 2499, "lr": 7.850210232666427e-06}
{"train_loss": 0.03129067271947861, "global_step": 222465, "epoch": 2499, "lr": 7.849898383498389e-06}
{"train_loss": 0.030812837183475494, "global_step": 222466, "epoch": 2499, "lr": 7.849586539996906e-06}
{"train_loss": 0.06036718189716339, "global_step": 222467, "epoch": 2499, "lr": 7.849274702162046e-06}
{"train_loss": 0.05454634875059128, "global_step": 222468, "epoch": 2499, "lr": 7.848962869993825e-06}
{"train_loss": 0.057196930050849915, "global_step": 222469, "epoch": 2499, "lr": 7.84865104349231e-06}
{"train_loss": 0.054619621485471725, "global_step": 222470, "epoch": 2499, "lr": 7.848339222657519e-06}
{"train_loss": 0.03673390671610832, "global_step": 222471, "epoch": 2499, "lr": 7.848027407489517e-06}
{"train_loss": 0.06634961068630219, "global_step": 222472, "epoch": 2499, "lr": 7.84771559798832e-06}
{"train_loss": 0.0712154433131218, "global_step": 222473, "epoch": 2499, "lr": 7.847403794153995e-06}
{"train_loss": 0.04706249758601189, "global_step": 222474, "epoch": 2499, "lr": 7.847091995986561e-06}
{"train_loss": 0.05181719735264778, "global_step": 222475, "epoch": 2499, "lr": 7.846780203486087e-06}
{"train_loss": 0.04384144768118858, "global_step": 222476, "epoch": 2499, "lr": 7.846468416652585e-06}
{"train_loss": 0.05765943229198456, "global_step": 222477, "epoch": 2499, "lr": 7.846156635486124e-06}
{"train_loss": 0.07629549503326416, "global_step": 222478, "epoch": 2499, "lr": 7.845844859986723e-06}
{"train_loss": 0.017701035365462303, "global_step": 222479, "epoch": 2499, "lr": 7.845533090154445e-06}
{"train_loss": 0.06683552265167236, "global_step": 222480, "epoch": 2499, "lr": 7.845221325989305e-06}
{"train_loss": 0.033065758645534515, "global_step": 222481, "epoch": 2499, "lr": 7.844909567491376e-06}
{"train_loss": 0.0834871381521225, "global_step": 222482, "epoch": 2499, "lr": 7.844597814660676e-06}
{"train_loss": 0.04534769058227539, "global_step": 222483, "epoch": 2499, "lr": 7.84428606749727e-06}
{"train_loss": 0.08696375787258148, "global_step": 222484, "epoch": 2499, "lr": 7.84397432600118e-06}
{"train_loss": 0.07660634815692902, "global_step": 222485, "epoch": 2499, "lr": 7.843662590172446e-06}
{"train_loss": 0.03306327015161514, "global_step": 222486, "epoch": 2499, "lr": 7.843350860011129e-06}
{"train_loss": 0.022108716890215874, "global_step": 222487, "epoch": 2499, "lr": 7.843039135517244e-06}
{"train_loss": 0.08364395052194595, "global_step": 222488, "epoch": 2499, "lr": 7.842727416690864e-06}
{"train_loss": 0.056367721408605576, "global_step": 222489, "epoch": 2499, "lr": 7.842415703532002e-06}
{"train_loss": 0.046586986631155014, "global_step": 222490, "epoch": 2499, "lr": 7.842103996040722e-06}
{"train_loss": 0.05273234844207764, "global_step": 222491, "epoch": 2499, "lr": 7.841792294217053e-06}
{"train_loss": 0.08195136487483978, "global_step": 222492, "epoch": 2499, "lr": 7.841480598061046e-06}
{"train_loss": 0.024452775716781616, "global_step": 222493, "epoch": 2499, "lr": 7.841168907572733e-06}
{"train_loss": 0.06647223979234695, "global_step": 222494, "epoch": 2499, "lr": 7.840857222752168e-06}
{"train_loss": 0.052249617874622345, "global_step": 222495, "epoch": 2499, "lr": 7.840545543599376e-06}
{"train_loss": 0.031714875251054764, "global_step": 222496, "epoch": 2499, "lr": 7.840233870114422e-06}
{"train_loss": 0.05880843847990036, "global_step": 222497, "epoch": 2499, "lr": 7.839922202297317e-06}
{"train_loss": 0.030323229730129242, "global_step": 222498, "epoch": 2499, "lr": 7.83961054014814e-06}
{"train_loss": 0.05507141694928823, "global_step": 222499, "epoch": 2499, "lr": 7.839298883666896e-06, "val_loss": 8.914031028747559}
{"train_loss": 0.03702416643500328, "global_step": 222500, "epoch": 2500, "lr": 7.838987232853645e-06}
{"train_loss": 0.04892214760184288, "global_step": 222501, "epoch": 2500, "lr": 7.838675587708444e-06}
{"train_loss": 0.02803107164800167, "global_step": 222502, "epoch": 2500, "lr": 7.838363948231304e-06}
{"train_loss": 0.04715944081544876, "global_step": 222503, "epoch": 2500, "lr": 7.838052314422295e-06}
{"train_loss": 0.07415878027677536, "global_step": 222504, "epoch": 2500, "lr": 7.837740686281436e-06}
{"train_loss": 0.06173742935061455, "global_step": 222505, "epoch": 2500, "lr": 7.837429063808787e-06}
{"train_loss": 0.03864115849137306, "global_step": 222506, "epoch": 2500, "lr": 7.837117447004372e-06}
{"train_loss": 0.042558591812849045, "global_step": 222507, "epoch": 2500, "lr": 7.836805835868255e-06}
{"train_loss": 0.04177764803171158, "global_step": 222508, "epoch": 2500, "lr": 7.836494230400453e-06}
{"train_loss": 0.03891289606690407, "global_step": 222509, "epoch": 2500, "lr": 7.836182630601035e-06}
{"train_loss": 0.047558460384607315, "global_step": 222510, "epoch": 2500, "lr": 7.835871036470011e-06}
{"train_loss": 0.06426268070936203, "global_step": 222511, "epoch": 2500, "lr": 7.835559448007457e-06}
{"train_loss": 0.05321526899933815, "global_step": 222512, "epoch": 2500, "lr": 7.83524786521338e-06}
{"train_loss": 0.043501757085323334, "global_step": 222513, "epoch": 2500, "lr": 7.834936288087857e-06}
{"train_loss": 0.06850320845842361, "global_step": 222514, "epoch": 2500, "lr": 7.8346247166309e-06}
{"train_loss": 0.07969902455806732, "global_step": 222515, "epoch": 2500, "lr": 7.834313150842576e-06}
{"train_loss": 0.05321624130010605, "global_step": 222516, "epoch": 2500, "lr": 7.834001590722906e-06}
{"train_loss": 0.0782506912946701, "global_step": 222517, "epoch": 2500, "lr": 7.833690036271945e-06}
{"train_loss": 0.059890687465667725, "global_step": 222518, "epoch": 2500, "lr": 7.833378487489723e-06}
{"train_loss": 0.024584395810961723, "global_step": 222519, "epoch": 2500, "lr": 7.8330669443763e-06}
{"train_loss": 0.047788310796022415, "global_step": 222520, "epoch": 2500, "lr": 7.832755406931696e-06}
{"train_loss": 0.06785274296998978, "global_step": 222521, "epoch": 2500, "lr": 7.832443875155977e-06}
{"train_loss": 0.09669298678636551, "global_step": 222522, "epoch": 2500, "lr": 7.83213234904917e-06}
{"train_loss": 0.10869094729423523, "global_step": 222523, "epoch": 2500, "lr": 7.831820828611309e-06}
{"train_loss": 0.05033210664987564, "global_step": 222524, "epoch": 2500, "lr": 7.831509313842455e-06}
{"train_loss": 0.029276050627231598, "global_step": 222525, "epoch": 2500, "lr": 7.831197804742634e-06}
{"train_loss": 0.04157446324825287, "global_step": 222526, "epoch": 2500, "lr": 7.830886301311901e-06}
{"train_loss": 0.06844036281108856, "global_step": 222527, "epoch": 2500, "lr": 7.830574803550284e-06}
{"train_loss": 0.048361148685216904, "global_step": 222528, "epoch": 2500, "lr": 7.830263311457836e-06}
{"train_loss": 0.06993741542100906, "global_step": 222529, "epoch": 2500, "lr": 7.829951825034592e-06}
{"train_loss": 0.03257026523351669, "global_step": 222530, "epoch": 2500, "lr": 7.829640344280604e-06}
{"train_loss": 0.049796491861343384, "global_step": 222531, "epoch": 2500, "lr": 7.829328869195896e-06}
{"train_loss": 0.06420686841011047, "global_step": 222532, "epoch": 2500, "lr": 7.829017399780536e-06}
{"train_loss": 0.10135744512081146, "global_step": 222533, "epoch": 2500, "lr": 7.82870593603453e-06}
{"train_loss": 0.04682771861553192, "global_step": 222534, "epoch": 2500, "lr": 7.828394477957963e-06}
{"train_loss": 0.0672449916601181, "global_step": 222535, "epoch": 2500, "lr": 7.828083025550836e-06}
{"train_loss": 0.06889545172452927, "global_step": 222536, "epoch": 2500, "lr": 7.827771578813209e-06}
{"train_loss": 0.07115024328231812, "global_step": 222537, "epoch": 2500, "lr": 7.827460137745141e-06}
{"train_loss": 0.04126785695552826, "global_step": 222538, "epoch": 2500, "lr": 7.827148702346638e-06}
{"train_loss": 0.06986550986766815, "global_step": 222539, "epoch": 2500, "lr": 7.826837272617777e-06}
{"train_loss": 0.04521868750452995, "global_step": 222540, "epoch": 2500, "lr": 7.826525848558574e-06}
{"train_loss": 0.07040931284427643, "global_step": 222541, "epoch": 2500, "lr": 7.826214430169087e-06}
{"train_loss": 0.029573116451501846, "global_step": 222542, "epoch": 2500, "lr": 7.825903017449344e-06}
{"train_loss": 0.03688300773501396, "global_step": 222543, "epoch": 2500, "lr": 7.825591610399408e-06}
{"train_loss": 0.11605178564786911, "global_step": 222544, "epoch": 2500, "lr": 7.825280209019286e-06}
{"train_loss": 0.04542570561170578, "global_step": 222545, "epoch": 2500, "lr": 7.824968813309064e-06}
{"train_loss": 0.0341251902282238, "global_step": 222546, "epoch": 2500, "lr": 7.824657423268744e-06}
{"train_loss": 0.06076934561133385, "global_step": 222547, "epoch": 2500, "lr": 7.824346038898395e-06}
{"train_loss": 0.06625019758939743, "global_step": 222548, "epoch": 2500, "lr": 7.824034660198038e-06}
{"train_loss": 0.0686734989285469, "global_step": 222549, "epoch": 2500, "lr": 7.82372328716774e-06}
{"train_loss": 0.04564004763960838, "global_step": 222550, "epoch": 2500, "lr": 7.823411919807511e-06}
{"train_loss": 0.05812881886959076, "global_step": 222551, "epoch": 2500, "lr": 7.823100558117424e-06}
{"train_loss": 0.01362429466098547, "global_step": 222552, "epoch": 2500, "lr": 7.822789202097496e-06}
{"train_loss": 0.07945837825536728, "global_step": 222553, "epoch": 2500, "lr": 7.822477851747794e-06}
{"train_loss": 0.05422202870249748, "global_step": 222554, "epoch": 2500, "lr": 7.822166507068328e-06}
{"train_loss": 0.058018364012241364, "global_step": 222555, "epoch": 2500, "lr": 7.821855168059178e-06}
{"train_loss": 0.023683004081249237, "global_step": 222556, "epoch": 2500, "lr": 7.821543834720346e-06}
{"train_loss": 0.042511213570833206, "global_step": 222557, "epoch": 2500, "lr": 7.821232507051911e-06}
{"train_loss": 0.044326696544885635, "global_step": 222558, "epoch": 2500, "lr": 7.82092118505388e-06}
{"train_loss": 0.03647886961698532, "global_step": 222559, "epoch": 2500, "lr": 7.82060986872633e-06}
{"train_loss": 0.021603716537356377, "global_step": 222560, "epoch": 2500, "lr": 7.820298558069283e-06}
{"train_loss": 0.06812143325805664, "global_step": 222561, "epoch": 2500, "lr": 7.819987253082766e-06}
{"train_loss": 0.09287174046039581, "global_step": 222562, "epoch": 2500, "lr": 7.819675953766853e-06}
{"train_loss": 0.04084460064768791, "global_step": 222563, "epoch": 2500, "lr": 7.819364660121558e-06}
{"train_loss": 0.02798989973962307, "global_step": 222564, "epoch": 2500, "lr": 7.81905337214695e-06}
{"train_loss": 0.05435215309262276, "global_step": 222565, "epoch": 2500, "lr": 7.81874208984304e-06}
{"train_loss": 0.03113732859492302, "global_step": 222566, "epoch": 2500, "lr": 7.818430813209903e-06}
{"train_loss": 0.024128923192620277, "global_step": 222567, "epoch": 2500, "lr": 7.818119542247548e-06}
{"train_loss": 0.0549006424844265, "global_step": 222568, "epoch": 2500, "lr": 7.817808276956045e-06}
{"train_loss": 0.051225434988737106, "global_step": 222569, "epoch": 2500, "lr": 7.817497017335413e-06}
{"train_loss": 0.06347887217998505, "global_step": 222570, "epoch": 2500, "lr": 7.817185763385716e-06}
{"train_loss": 0.04535385221242905, "global_step": 222571, "epoch": 2500, "lr": 7.81687451510697e-06}
{"train_loss": 0.028124988079071045, "global_step": 222572, "epoch": 2500, "lr": 7.816563272499233e-06}
{"train_loss": 0.04271354526281357, "global_step": 222573, "epoch": 2500, "lr": 7.816252035562555e-06}
{"train_loss": 0.08643031865358353, "global_step": 222574, "epoch": 2500, "lr": 7.815940804296956e-06}
{"train_loss": 0.018656086176633835, "global_step": 222575, "epoch": 2500, "lr": 7.815629578702504e-06}
{"train_loss": 0.04770469292998314, "global_step": 222576, "epoch": 2500, "lr": 7.815318358779216e-06}
{"train_loss": 0.045535456389188766, "global_step": 222577, "epoch": 2500, "lr": 7.815007144527153e-06}
{"train_loss": 0.11307269334793091, "global_step": 222578, "epoch": 2500, "lr": 7.814695935946337e-06}
{"train_loss": 0.05310889333486557, "global_step": 222579, "epoch": 2500, "lr": 7.814384733036833e-06}
{"train_loss": 0.027939241379499435, "global_step": 222580, "epoch": 2500, "lr": 7.814073535798661e-06}
{"train_loss": 0.018089180812239647, "global_step": 222581, "epoch": 2500, "lr": 7.813762344231878e-06}
{"train_loss": 0.07748718559741974, "global_step": 222582, "epoch": 2500, "lr": 7.813451158336516e-06}
{"train_loss": 0.04734553396701813, "global_step": 222583, "epoch": 2500, "lr": 7.813139978112627e-06}
{"train_loss": 0.04731839895248413, "global_step": 222584, "epoch": 2500, "lr": 7.81282880356024e-06}
{"train_loss": 0.06839260458946228, "global_step": 222585, "epoch": 2500, "lr": 7.812517634679417e-06}
{"train_loss": 0.027967996895313263, "global_step": 222586, "epoch": 2500, "lr": 7.812206471470167e-06}
{"train_loss": 0.0670296773314476, "global_step": 222587, "epoch": 2500, "lr": 7.81189531393257e-06}
{"train_loss": 0.053308980717334184, "global_step": 222588, "epoch": 2500, "lr": 7.811584162066637e-06, "train/sim_max_reward_0": 0.6609275603643794, "train/sim_max_reward_1": 0.9887174636313434, "train/sim_max_reward_2": 0.6460695056959477, "train/sim_max_reward_3": 0.9804733389759819, "train/sim_max_reward_4": 0.5959352014120609, "train/sim_max_reward_5": 0.9574336666579278, "test/sim_max_reward_4500000": 0.4570940777648876, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.8330829183504594, "test/sim_max_reward_4500003": 0.674145063474674, "test/sim_max_reward_4500004": 0.20304812676502998, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.9344603274244888, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.3178452350221833, "test/sim_max_reward_4500009": 0.05030204089235116, "test/sim_max_reward_4500010": 0.9829355763441746, "test/sim_max_reward_4500011": 0.9657847029461438, "test/sim_max_reward_4500012": 0.9663892552614095, "test/sim_max_reward_4500013": 0.5059697437057624, "test/sim_max_reward_4500014": 0.3667504764620676, "test/sim_max_reward_4500015": 1.0, "test/sim_max_reward_4500016": 1.0, "test/sim_max_reward_4500017": 0.7268269177006607, "test/sim_max_reward_4500018": 0.9973801715349123, "test/sim_max_reward_4500019": 0.4546080806346203, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9458687449613201, "test/sim_max_reward_4500022": 0.9503092655850033, "test/sim_max_reward_4500023": 0.5921217295490955, "test/sim_max_reward_4500024": 0.4825317781099384, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.41825860044372254, "test/sim_max_reward_4500029": 0.49033382540083115, "test/sim_max_reward_4500030": 0.969715183264624, "test/sim_max_reward_4500031": 0.5191541700374604, "test/sim_max_reward_4500032": 1.0, "test/sim_max_reward_4500033": 0.9703606728571732, "test/sim_max_reward_4500034": 0.6113071408809823, "test/sim_max_reward_4500035": 0.5173067377034538, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.16189257846853528, "test/sim_max_reward_4500039": 0.9784517451539309, "test/sim_max_reward_4500040": 0.9462833356044511, "test/sim_max_reward_4500041": 0.6912235515775398, "test/sim_max_reward_4500042": 0.31031165232153557, "test/sim_max_reward_4500043": 0.9635339447908973, "test/sim_max_reward_4500044": 0.18330465758363243, "test/sim_max_reward_4500045": 0.9938474304370353, "test/sim_max_reward_4500046": 0.39185241119077596, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9754961295317144, "test/sim_max_reward_4500049": 0.9359075122257804, "train/mean_score": 0.8049261227896068, "test/mean_score": 0.7087199102392652, "val_loss": 8.99531078338623, "train_action_mse_error": 33.74809646606445}
{"train_loss": 0.05918705835938454, "global_step": 222589, "epoch": 2501, "lr": 7.811273015872434e-06}
{"train_loss": 0.05918344855308533, "global_step": 222590, "epoch": 2501, "lr": 7.810961875349976e-06}
{"train_loss": 0.03561299666762352, "global_step": 222591, "epoch": 2501, "lr": 7.810650740499326e-06}
{"train_loss": 0.04622938483953476, "global_step": 222592, "epoch": 2501, "lr": 7.810339611320528e-06}
{"train_loss": 0.04747595265507698, "global_step": 222593, "epoch": 2501, "lr": 7.810028487813603e-06}
{"train_loss": 0.050514400005340576, "global_step": 222594, "epoch": 2501, "lr": 7.80971736997862e-06}
{"train_loss": 0.09311297535896301, "global_step": 222595, "epoch": 2501, "lr": 7.809406257815604e-06}
{"train_loss": 0.046791572123765945, "global_step": 222596, "epoch": 2501, "lr": 7.809095151324587e-06}
{"train_loss": 0.04732451215386391, "global_step": 222597, "epoch": 2501, "lr": 7.808784050505636e-06}
{"train_loss": 0.05534708499908447, "global_step": 222598, "epoch": 2501, "lr": 7.80847295535877e-06}
{"train_loss": 0.04543090611696243, "global_step": 222599, "epoch": 2501, "lr": 7.808161865884044e-06}
{"train_loss": 0.06232038512825966, "global_step": 222600, "epoch": 2501, "lr": 7.807850782081506e-06}
{"train_loss": 0.04676207900047302, "global_step": 222601, "epoch": 2501, "lr": 7.807539703951172e-06}
{"train_loss": 0.03569920361042023, "global_step": 222602, "epoch": 2501, "lr": 7.807228631493114e-06}
{"train_loss": 0.02542523853480816, "global_step": 222603, "epoch": 2501, "lr": 7.806917564707345e-06}
{"train_loss": 0.029051626101136208, "global_step": 222604, "epoch": 2501, "lr": 7.806606503593939e-06}
{"train_loss": 0.03348010033369064, "global_step": 222605, "epoch": 2501, "lr": 7.806295448152901e-06}
{"train_loss": 0.077732615172863, "global_step": 222606, "epoch": 2501, "lr": 7.805984398384309e-06}
{"train_loss": 0.045169245451688766, "global_step": 222607, "epoch": 2501, "lr": 7.805673354288173e-06}
{"train_loss": 0.03499127924442291, "global_step": 222608, "epoch": 2501, "lr": 7.805362315864557e-06}
{"train_loss": 0.03686639294028282, "global_step": 222609, "epoch": 2501, "lr": 7.805051283113502e-06}
{"train_loss": 0.040303125977516174, "global_step": 222610, "epoch": 2501, "lr": 7.804740256035032e-06}
{"train_loss": 0.033318955451250076, "global_step": 222611, "epoch": 2501, "lr": 7.804429234629213e-06}
{"train_loss": 0.04802945256233215, "global_step": 222612, "epoch": 2501, "lr": 7.804118218896061e-06}
{"train_loss": 0.03512553125619888, "global_step": 222613, "epoch": 2501, "lr": 7.803807208835645e-06}
{"train_loss": 0.07222660630941391, "global_step": 222614, "epoch": 2501, "lr": 7.80349620444798e-06}
{"train_loss": 0.05812884122133255, "global_step": 222615, "epoch": 2501, "lr": 7.803185205733133e-06}
{"train_loss": 0.04918881878256798, "global_step": 222616, "epoch": 2501, "lr": 7.80287421269112e-06}
{"train_loss": 0.027355775237083435, "global_step": 222617, "epoch": 2501, "lr": 7.802563225322007e-06}
{"train_loss": 0.03833361342549324, "global_step": 222618, "epoch": 2501, "lr": 7.802252243625812e-06}
{"train_loss": 0.05046998709440231, "global_step": 222619, "epoch": 2501, "lr": 7.801941267602609e-06}
{"train_loss": 0.07203690707683563, "global_step": 222620, "epoch": 2501, "lr": 7.801630297252405e-06}
{"train_loss": 0.031611427664756775, "global_step": 222621, "epoch": 2501, "lr": 7.801319332575269e-06}
{"train_loss": 0.02168162539601326, "global_step": 222622, "epoch": 2501, "lr": 7.801008373571217e-06}
{"train_loss": 0.05740716680884361, "global_step": 222623, "epoch": 2501, "lr": 7.800697420240321e-06}
{"train_loss": 0.041726816445589066, "global_step": 222624, "epoch": 2501, "lr": 7.800386472582594e-06}
{"train_loss": 0.037618160247802734, "global_step": 222625, "epoch": 2501, "lr": 7.8000755305981e-06}
{"train_loss": 0.077554851770401, "global_step": 222626, "epoch": 2501, "lr": 7.799764594286862e-06}
{"train_loss": 0.04072102904319763, "global_step": 222627, "epoch": 2501, "lr": 7.799453663648932e-06}
{"train_loss": 0.06718096882104874, "global_step": 222628, "epoch": 2501, "lr": 7.799142738684361e-06}
{"train_loss": 0.04791325330734253, "global_step": 222629, "epoch": 2501, "lr": 7.798831819393172e-06}
{"train_loss": 0.057502616196870804, "global_step": 222630, "epoch": 2501, "lr": 7.798520905775426e-06}
{"train_loss": 0.037542182952165604, "global_step": 222631, "epoch": 2501, "lr": 7.79820999783114e-06}
{"train_loss": 0.08668743073940277, "global_step": 222632, "epoch": 2501, "lr": 7.797899095560384e-06}
{"train_loss": 0.05421461910009384, "global_step": 222633, "epoch": 2501, "lr": 7.797588198963191e-06}
{"train_loss": 0.03141799196600914, "global_step": 222634, "epoch": 2501, "lr": 7.79727730803958e-06}
{"train_loss": 0.04516235366463661, "global_step": 222635, "epoch": 2501, "lr": 7.796966422789625e-06}
{"train_loss": 0.037704043090343475, "global_step": 222636, "epoch": 2501, "lr": 7.796655543213338e-06}
{"train_loss": 0.10052221268415451, "global_step": 222637, "epoch": 2501, "lr": 7.79634466931079e-06}
{"train_loss": 0.07056425511837006, "global_step": 222638, "epoch": 2501, "lr": 7.79603380108201e-06}
{"train_loss": 0.0181187242269516, "global_step": 222639, "epoch": 2501, "lr": 7.795722938527028e-06}
{"train_loss": 0.05950479954481125, "global_step": 222640, "epoch": 2501, "lr": 7.795412081645908e-06}
{"train_loss": 0.046409379690885544, "global_step": 222641, "epoch": 2501, "lr": 7.795101230438668e-06}
{"train_loss": 0.0637223869562149, "global_step": 222642, "epoch": 2501, "lr": 7.794790384905376e-06}
{"train_loss": 0.01720014214515686, "global_step": 222643, "epoch": 2501, "lr": 7.794479545046046e-06}
{"train_loss": 0.031289201229810715, "global_step": 222644, "epoch": 2501, "lr": 7.794168710860734e-06}
{"train_loss": 0.053318124264478683, "global_step": 222645, "epoch": 2501, "lr": 7.793857882349492e-06}
{"train_loss": 0.03852677345275879, "global_step": 222646, "epoch": 2501, "lr": 7.793547059512341e-06}
{"train_loss": 0.029654471203684807, "global_step": 222647, "epoch": 2501, "lr": 7.793236242349344e-06}
{"train_loss": 0.036025263369083405, "global_step": 222648, "epoch": 2501, "lr": 7.79292543086052e-06}
{"train_loss": 0.06311487406492233, "global_step": 222649, "epoch": 2501, "lr": 7.792614625045936e-06}
{"train_loss": 0.06703262776136398, "global_step": 222650, "epoch": 2501, "lr": 7.79230382490561e-06}
{"train_loss": 0.05581460893154144, "global_step": 222651, "epoch": 2501, "lr": 7.791993030439605e-06}
{"train_loss": 0.043174102902412415, "global_step": 222652, "epoch": 2501, "lr": 7.79168224164794e-06}
{"train_loss": 0.04034475237131119, "global_step": 222653, "epoch": 2501, "lr": 7.791371458530683e-06}
{"train_loss": 0.10207223892211914, "global_step": 222654, "epoch": 2501, "lr": 7.791060681087847e-06}
{"train_loss": 0.02251332253217697, "global_step": 222655, "epoch": 2501, "lr": 7.7907499093195e-06}
{"train_loss": 0.06627387553453445, "global_step": 222656, "epoch": 2501, "lr": 7.790439143225659e-06}
{"train_loss": 0.045550599694252014, "global_step": 222657, "epoch": 2501, "lr": 7.790128382806395e-06}
{"train_loss": 0.05967624485492706, "global_step": 222658, "epoch": 2501, "lr": 7.789817628061718e-06}
{"train_loss": 0.06380482763051987, "global_step": 222659, "epoch": 2501, "lr": 7.789506878991699e-06}
{"train_loss": 0.05812428146600723, "global_step": 222660, "epoch": 2501, "lr": 7.789196135596355e-06}
{"train_loss": 0.08199317008256912, "global_step": 222661, "epoch": 2501, "lr": 7.788885397875751e-06}
{"train_loss": 0.018170394003391266, "global_step": 222662, "epoch": 2501, "lr": 7.788574665829907e-06}
{"train_loss": 0.05298277735710144, "global_step": 222663, "epoch": 2501, "lr": 7.788263939458867e-06}
{"train_loss": 0.04231610521674156, "global_step": 222664, "epoch": 2501, "lr": 7.7879532187627e-06}
{"train_loss": 0.06368979811668396, "global_step": 222665, "epoch": 2501, "lr": 7.787642503741415e-06}
{"train_loss": 0.049793582409620285, "global_step": 222666, "epoch": 2501, "lr": 7.787331794395075e-06}
{"train_loss": 0.07213076204061508, "global_step": 222667, "epoch": 2501, "lr": 7.787021090723706e-06}
{"train_loss": 0.018143294379115105, "global_step": 222668, "epoch": 2501, "lr": 7.78671039272737e-06}
{"train_loss": 0.04799067601561546, "global_step": 222669, "epoch": 2501, "lr": 7.786399700406084e-06}
{"train_loss": 0.0464104562997818, "global_step": 222670, "epoch": 2501, "lr": 7.786089013759911e-06}
{"train_loss": 0.058510977774858475, "global_step": 222671, "epoch": 2501, "lr": 7.785778332788885e-06}
{"train_loss": 0.06165815889835358, "global_step": 222672, "epoch": 2501, "lr": 7.785467657493033e-06}
{"train_loss": 0.03830859065055847, "global_step": 222673, "epoch": 2501, "lr": 7.785156987872427e-06}
{"train_loss": 0.12223336100578308, "global_step": 222674, "epoch": 2501, "lr": 7.784846323927076e-06}
{"train_loss": 0.05848675221204758, "global_step": 222675, "epoch": 2501, "lr": 7.78453566565705e-06}
{"train_loss": 0.05351877212524414, "global_step": 222676, "epoch": 2501, "lr": 7.784225013062379e-06}
{"train_loss": 0.05050122223041031, "global_step": 222677, "epoch": 2501, "lr": 7.783914366143097e-06, "val_loss": 8.952738761901855}
{"train_loss": 0.08391957730054855, "global_step": 222678, "epoch": 2502, "lr": 7.783603724899257e-06}
{"train_loss": 0.044149000197649, "global_step": 222679, "epoch": 2502, "lr": 7.78329308933089e-06}
{"train_loss": 0.06832946836948395, "global_step": 222680, "epoch": 2502, "lr": 7.782982459438059e-06}
{"train_loss": 0.08791327476501465, "global_step": 222681, "epoch": 2502, "lr": 7.782671835220773e-06}
{"train_loss": 0.06020929291844368, "global_step": 222682, "epoch": 2502, "lr": 7.782361216679095e-06}
{"train_loss": 0.057664502412080765, "global_step": 222683, "epoch": 2502, "lr": 7.782050603813079e-06}
{"train_loss": 0.06156716123223305, "global_step": 222684, "epoch": 2502, "lr": 7.781739996622739e-06}
{"train_loss": 0.040537912398576736, "global_step": 222685, "epoch": 2502, "lr": 7.781429395108136e-06}
{"train_loss": 0.12447769939899445, "global_step": 222686, "epoch": 2502, "lr": 7.7811187992693e-06}
{"train_loss": 0.026712307706475258, "global_step": 222687, "epoch": 2502, "lr": 7.780808209106289e-06}
{"train_loss": 0.13212351500988007, "global_step": 222688, "epoch": 2502, "lr": 7.780497624619115e-06}
{"train_loss": 0.03322981297969818, "global_step": 222689, "epoch": 2502, "lr": 7.780187045807857e-06}
{"train_loss": 0.11637318879365921, "global_step": 222690, "epoch": 2502, "lr": 7.779876472672526e-06}
{"train_loss": 0.030240841209888458, "global_step": 222691, "epoch": 2502, "lr": 7.779565905213188e-06}
{"train_loss": 0.06344261765480042, "global_step": 222692, "epoch": 2502, "lr": 7.779255343429858e-06}
{"train_loss": 0.09489478170871735, "global_step": 222693, "epoch": 2502, "lr": 7.778944787322606e-06}
{"train_loss": 0.04403439164161682, "global_step": 222694, "epoch": 2502, "lr": 7.778634236891446e-06}
{"train_loss": 0.024187671020627022, "global_step": 222695, "epoch": 2502, "lr": 7.778323692136452e-06}
{"train_loss": 0.04240039363503456, "global_step": 222696, "epoch": 2502, "lr": 7.778013153057628e-06}
{"train_loss": 0.04769675061106682, "global_step": 222697, "epoch": 2502, "lr": 7.777702619655052e-06}
{"train_loss": 0.06023240461945534, "global_step": 222698, "epoch": 2502, "lr": 7.777392091928737e-06}
{"train_loss": 0.042598482221364975, "global_step": 222699, "epoch": 2502, "lr": 7.777081569878736e-06}
{"train_loss": 0.0543142706155777, "global_step": 222700, "epoch": 2502, "lr": 7.776771053505105e-06}
{"train_loss": 0.04537750035524368, "global_step": 222701, "epoch": 2502, "lr": 7.776460542807862e-06}
{"train_loss": 0.0870114266872406, "global_step": 222702, "epoch": 2502, "lr": 7.776150037787067e-06}
{"train_loss": 0.06172103434801102, "global_step": 222703, "epoch": 2502, "lr": 7.775839538442747e-06}
{"train_loss": 0.05752783641219139, "global_step": 222704, "epoch": 2502, "lr": 7.77552904477496e-06}
{"train_loss": 0.030440734699368477, "global_step": 222705, "epoch": 2502, "lr": 7.775218556783725e-06}
{"train_loss": 0.02538064308464527, "global_step": 222706, "epoch": 2502, "lr": 7.774908074469112e-06}
{"train_loss": 0.032747913151979446, "global_step": 222707, "epoch": 2502, "lr": 7.774597597831135e-06}
{"train_loss": 0.04427843913435936, "global_step": 222708, "epoch": 2502, "lr": 7.77428712686986e-06}
{"train_loss": 0.049763135612010956, "global_step": 222709, "epoch": 2502, "lr": 7.773976661585309e-06}
{"train_loss": 0.03367837518453598, "global_step": 222710, "epoch": 2502, "lr": 7.773666201977536e-06}
{"train_loss": 0.04869251325726509, "global_step": 222711, "epoch": 2502, "lr": 7.773355748046584e-06}
{"train_loss": 0.049226462841033936, "global_step": 222712, "epoch": 2502, "lr": 7.773045299792482e-06}
{"train_loss": 0.061888664960861206, "global_step": 222713, "epoch": 2502, "lr": 7.772734857215285e-06}
{"train_loss": 0.05884722247719765, "global_step": 222714, "epoch": 2502, "lr": 7.772424420315017e-06}
{"train_loss": 0.09347524493932724, "global_step": 222715, "epoch": 2502, "lr": 7.772113989091745e-06}
{"train_loss": 0.03238597512245178, "global_step": 222716, "epoch": 2502, "lr": 7.7718035635455e-06}
{"train_loss": 0.07452703267335892, "global_step": 222717, "epoch": 2502, "lr": 7.771493143676306e-06}
{"train_loss": 0.06660429388284683, "global_step": 222718, "epoch": 2502, "lr": 7.771182729484222e-06}
{"train_loss": 0.09422465413808823, "global_step": 222719, "epoch": 2502, "lr": 7.770872320969302e-06}
{"train_loss": 0.04496460035443306, "global_step": 222720, "epoch": 2502, "lr": 7.770561918131558e-06}
{"train_loss": 0.04316837340593338, "global_step": 222721, "epoch": 2502, "lr": 7.770251520971062e-06}
{"train_loss": 0.04148198291659355, "global_step": 222722, "epoch": 2502, "lr": 7.76994112948783e-06}
{"train_loss": 0.036765843629837036, "global_step": 222723, "epoch": 2502, "lr": 7.769630743681922e-06}
{"train_loss": 0.0514206625521183, "global_step": 222724, "epoch": 2502, "lr": 7.769320363553361e-06}
{"train_loss": 0.08365406095981598, "global_step": 222725, "epoch": 2502, "lr": 7.769009989102217e-06}
{"train_loss": 0.06632457673549652, "global_step": 222726, "epoch": 2502, "lr": 7.768699620328495e-06}
{"train_loss": 0.049210045486688614, "global_step": 222727, "epoch": 2502, "lr": 7.768389257232278e-06}
{"train_loss": 0.08891096711158752, "global_step": 222728, "epoch": 2502, "lr": 7.768078899813568e-06}
{"train_loss": 0.08546268194913864, "global_step": 222729, "epoch": 2502, "lr": 7.767768548072441e-06}
{"train_loss": 0.05182959511876106, "global_step": 222730, "epoch": 2502, "lr": 7.767458202008904e-06}
{"train_loss": 0.036381859332323074, "global_step": 222731, "epoch": 2502, "lr": 7.767147861623036e-06}
{"train_loss": 0.05524366721510887, "global_step": 222732, "epoch": 2502, "lr": 7.766837526914844e-06}
{"train_loss": 0.01988540031015873, "global_step": 222733, "epoch": 2502, "lr": 7.7665271978844e-06}
{"train_loss": 0.06340713053941727, "global_step": 222734, "epoch": 2502, "lr": 7.766216874531724e-06}
{"train_loss": 0.045052919536828995, "global_step": 222735, "epoch": 2502, "lr": 7.765906556856855e-06}
{"train_loss": 0.044840678572654724, "global_step": 222736, "epoch": 2502, "lr": 7.765596244859869e-06}
{"train_loss": 0.05791611969470978, "global_step": 222737, "epoch": 2502, "lr": 7.765285938540762e-06}
{"train_loss": 0.04425037279725075, "global_step": 222738, "epoch": 2502, "lr": 7.764975637899612e-06}
{"train_loss": 0.047512274235486984, "global_step": 222739, "epoch": 2502, "lr": 7.764665342936438e-06}
{"train_loss": 0.04336603358387947, "global_step": 222740, "epoch": 2502, "lr": 7.7643550536513e-06}
{"train_loss": 0.07757937163114548, "global_step": 222741, "epoch": 2502, "lr": 7.764044770044216e-06}
{"train_loss": 0.05414185672998428, "global_step": 222742, "epoch": 2502, "lr": 7.763734492115254e-06}
{"train_loss": 0.041752006858587265, "global_step": 222743, "epoch": 2502, "lr": 7.763424219864435e-06}
{"train_loss": 0.041989766061306, "global_step": 222744, "epoch": 2502, "lr": 7.763113953291817e-06}
{"train_loss": 0.06617303192615509, "global_step": 222745, "epoch": 2502, "lr": 7.762803692397424e-06}
{"train_loss": 0.06298615783452988, "global_step": 222746, "epoch": 2502, "lr": 7.762493437181317e-06}
{"train_loss": 0.030046790838241577, "global_step": 222747, "epoch": 2502, "lr": 7.762183187643517e-06}
{"train_loss": 0.07512419670820236, "global_step": 222748, "epoch": 2502, "lr": 7.761872943784087e-06}
{"train_loss": 0.09733473509550095, "global_step": 222749, "epoch": 2502, "lr": 7.76156270560306e-06}
{"train_loss": 0.08474384993314743, "global_step": 222750, "epoch": 2502, "lr": 7.761252473100461e-06}
{"train_loss": 0.03523831441998482, "global_step": 222751, "epoch": 2502, "lr": 7.760942246276365e-06}
{"train_loss": 0.07500739395618439, "global_step": 222752, "epoch": 2502, "lr": 7.760632025130775e-06}
{"train_loss": 0.03241802379488945, "global_step": 222753, "epoch": 2502, "lr": 7.760321809663773e-06}
{"train_loss": 0.05307735875248909, "global_step": 222754, "epoch": 2502, "lr": 7.760011599875366e-06}
{"train_loss": 0.05612846836447716, "global_step": 222755, "epoch": 2502, "lr": 7.759701395765623e-06}
{"train_loss": 0.019656097516417503, "global_step": 222756, "epoch": 2502, "lr": 7.75939119733456e-06}
{"train_loss": 0.038074880838394165, "global_step": 222757, "epoch": 2502, "lr": 7.759081004582246e-06}
{"train_loss": 0.036383625119924545, "global_step": 222758, "epoch": 2502, "lr": 7.758770817508698e-06}
{"train_loss": 0.03323869779706001, "global_step": 222759, "epoch": 2502, "lr": 7.758460636113974e-06}
{"train_loss": 0.01925414428114891, "global_step": 222760, "epoch": 2502, "lr": 7.758150460398105e-06}
{"train_loss": 0.05676324665546417, "global_step": 222761, "epoch": 2502, "lr": 7.757840290361146e-06}
{"train_loss": 0.037960294634103775, "global_step": 222762, "epoch": 2502, "lr": 7.75753012600312e-06}
{"train_loss": 0.019940828904509544, "global_step": 222763, "epoch": 2502, "lr": 7.757219967324092e-06}
{"train_loss": 0.029439281672239304, "global_step": 222764, "epoch": 2502, "lr": 7.756909814324075e-06}
{"train_loss": 0.04149562120437622, "global_step": 222765, "epoch": 2502, "lr": 7.75659966700314e-06}
{"train_loss": 0.05418490195709668, "global_step": 222766, "epoch": 2502, "lr": 7.756289525361305e-06, "val_loss": 9.106100082397461}
{"train_loss": 0.04173895716667175, "global_step": 222767, "epoch": 2503, "lr": 7.755979389398638e-06}
{"train_loss": 0.03229966387152672, "global_step": 222768, "epoch": 2503, "lr": 7.755669259115145e-06}
{"train_loss": 0.06813541054725647, "global_step": 222769, "epoch": 2503, "lr": 7.755359134510904e-06}
{"train_loss": 0.02604454755783081, "global_step": 222770, "epoch": 2503, "lr": 7.755049015585925e-06}
{"train_loss": 0.05297616496682167, "global_step": 222771, "epoch": 2503, "lr": 7.75473890234027e-06}
{"train_loss": 0.051659662276506424, "global_step": 222772, "epoch": 2503, "lr": 7.754428794773982e-06}
{"train_loss": 0.11344058066606522, "global_step": 222773, "epoch": 2503, "lr": 7.754118692887086e-06}
{"train_loss": 0.03680361434817314, "global_step": 222774, "epoch": 2503, "lr": 7.753808596679646e-06}
{"train_loss": 0.05164135619997978, "global_step": 222775, "epoch": 2503, "lr": 7.753498506151679e-06}
{"train_loss": 0.0552220493555069, "global_step": 222776, "epoch": 2503, "lr": 7.753188421303254e-06}
{"train_loss": 0.05295262113213539, "global_step": 222777, "epoch": 2503, "lr": 7.752878342134378e-06}
{"train_loss": 0.06445104628801346, "global_step": 222778, "epoch": 2503, "lr": 7.752568268645132e-06}
{"train_loss": 0.04365066438913345, "global_step": 222779, "epoch": 2503, "lr": 7.752258200835527e-06}
{"train_loss": 0.02278091013431549, "global_step": 222780, "epoch": 2503, "lr": 7.751948138705622e-06}
{"train_loss": 0.029691601172089577, "global_step": 222781, "epoch": 2503, "lr": 7.751638082255447e-06}
{"train_loss": 0.04504157975316048, "global_step": 222782, "epoch": 2503, "lr": 7.751328031485056e-06}
{"train_loss": 0.02909570373594761, "global_step": 222783, "epoch": 2503, "lr": 7.751017986394476e-06}
{"train_loss": 0.08162373304367065, "global_step": 222784, "epoch": 2503, "lr": 7.750707946983764e-06}
{"train_loss": 0.04362991824746132, "global_step": 222785, "epoch": 2503, "lr": 7.75039791325295e-06}
{"train_loss": 0.04448728263378143, "global_step": 222786, "epoch": 2503, "lr": 7.750087885202085e-06}
{"train_loss": 0.032585009932518005, "global_step": 222787, "epoch": 2503, "lr": 7.74977786283121e-06}
{"train_loss": 0.046151839196681976, "global_step": 222788, "epoch": 2503, "lr": 7.749467846140346e-06}
{"train_loss": 0.037814777344465256, "global_step": 222789, "epoch": 2503, "lr": 7.749157835129567e-06}
{"train_loss": 0.04551320523023605, "global_step": 222790, "epoch": 2503, "lr": 7.748847829798883e-06}
{"train_loss": 0.0739457830786705, "global_step": 222791, "epoch": 2503, "lr": 7.748537830148367e-06}
{"train_loss": 0.07207787036895752, "global_step": 222792, "epoch": 2503, "lr": 7.748227836178034e-06}
{"train_loss": 0.09597951173782349, "global_step": 222793, "epoch": 2503, "lr": 7.747917847887943e-06}
{"train_loss": 0.07669368386268616, "global_step": 222794, "epoch": 2503, "lr": 7.747607865278123e-06}
{"train_loss": 0.0650358647108078, "global_step": 222795, "epoch": 2503, "lr": 7.74729788834863e-06}
{"train_loss": 0.06724782288074493, "global_step": 222796, "epoch": 2503, "lr": 7.74698791709949e-06}
{"train_loss": 0.019343197345733643, "global_step": 222797, "epoch": 2503, "lr": 7.746677951530762e-06}
{"train_loss": 0.030765347182750702, "global_step": 222798, "epoch": 2503, "lr": 7.746367991642467e-06}
{"train_loss": 0.022279120981693268, "global_step": 222799, "epoch": 2503, "lr": 7.746058037434672e-06}
{"train_loss": 0.048265449702739716, "global_step": 222800, "epoch": 2503, "lr": 7.74574808890739e-06}
{"train_loss": 0.05372588336467743, "global_step": 222801, "epoch": 2503, "lr": 7.74543814606069e-06}
{"train_loss": 0.030760163441300392, "global_step": 222802, "epoch": 2503, "lr": 7.745128208894586e-06}
{"train_loss": 0.019270816817879677, "global_step": 222803, "epoch": 2503, "lr": 7.74481827740915e-06}
{"train_loss": 0.08766303211450577, "global_step": 222804, "epoch": 2503, "lr": 7.744508351604397e-06}
{"train_loss": 0.03097626566886902, "global_step": 222805, "epoch": 2503, "lr": 7.744198431480387e-06}
{"train_loss": 0.05420997738838196, "global_step": 222806, "epoch": 2503, "lr": 7.743888517037145e-06}
{"train_loss": 0.050217464566230774, "global_step": 222807, "epoch": 2503, "lr": 7.743578608274732e-06}
{"train_loss": 0.04994513839483261, "global_step": 222808, "epoch": 2503, "lr": 7.743268705193174e-06}
{"train_loss": 0.0400136336684227, "global_step": 222809, "epoch": 2503, "lr": 7.742958807792512e-06}
{"train_loss": 0.0520501583814621, "global_step": 222810, "epoch": 2503, "lr": 7.74264891607281e-06}
{"train_loss": 0.04660055786371231, "global_step": 222811, "epoch": 2503, "lr": 7.74233903003408e-06}
{"train_loss": 0.017299793660640717, "global_step": 222812, "epoch": 2503, "lr": 7.74202914967639e-06}
{"train_loss": 0.04361683502793312, "global_step": 222813, "epoch": 2503, "lr": 7.741719274999759e-06}
{"train_loss": 0.07420101016759872, "global_step": 222814, "epoch": 2503, "lr": 7.741409406004247e-06}
{"train_loss": 0.010667414404451847, "global_step": 222815, "epoch": 2503, "lr": 7.741099542689878e-06}
{"train_loss": 0.06032421067357063, "global_step": 222816, "epoch": 2503, "lr": 7.740789685056716e-06}
{"train_loss": 0.06685864925384521, "global_step": 222817, "epoch": 2503, "lr": 7.740479833104775e-06}
{"train_loss": 0.016550861299037933, "global_step": 222818, "epoch": 2503, "lr": 7.740169986834122e-06}
{"train_loss": 0.04838012903928757, "global_step": 222819, "epoch": 2503, "lr": 7.739860146244781e-06}
{"train_loss": 0.03698277845978737, "global_step": 222820, "epoch": 2503, "lr": 7.739550311336807e-06}
{"train_loss": 0.05407068133354187, "global_step": 222821, "epoch": 2503, "lr": 7.739240482110227e-06}
{"train_loss": 0.09750089794397354, "global_step": 222822, "epoch": 2503, "lr": 7.738930658565103e-06}
{"train_loss": 0.033530209213495255, "global_step": 222823, "epoch": 2503, "lr": 7.738620840701454e-06}
{"train_loss": 0.0573207251727581, "global_step": 222824, "epoch": 2503, "lr": 7.73831102851934e-06}
{"train_loss": 0.06236891448497772, "global_step": 222825, "epoch": 2503, "lr": 7.738001222018786e-06}
{"train_loss": 0.029403677210211754, "global_step": 222826, "epoch": 2503, "lr": 7.737691421199855e-06}
{"train_loss": 0.04338068887591362, "global_step": 222827, "epoch": 2503, "lr": 7.737381626062578e-06}
{"train_loss": 0.028504759073257446, "global_step": 222828, "epoch": 2503, "lr": 7.73707183660698e-06}
{"train_loss": 0.08447200804948807, "global_step": 222829, "epoch": 2503, "lr": 7.736762052833129e-06}
{"train_loss": 0.012909683398902416, "global_step": 222830, "epoch": 2503, "lr": 7.736452274741047e-06}
{"train_loss": 0.06087207794189453, "global_step": 222831, "epoch": 2503, "lr": 7.736142502330796e-06}
{"train_loss": 0.04275905713438988, "global_step": 222832, "epoch": 2503, "lr": 7.735832735602389e-06}
{"train_loss": 0.042172931134700775, "global_step": 222833, "epoch": 2503, "lr": 7.735522974555898e-06}
{"train_loss": 0.02655043452978134, "global_step": 222834, "epoch": 2503, "lr": 7.73521321919134e-06}
{"train_loss": 0.0609065443277359, "global_step": 222835, "epoch": 2503, "lr": 7.734903469508775e-06}
{"train_loss": 0.09040457010269165, "global_step": 222836, "epoch": 2503, "lr": 7.734593725508227e-06}
{"train_loss": 0.036753132939338684, "global_step": 222837, "epoch": 2503, "lr": 7.734283987189761e-06}
{"train_loss": 0.05212216451764107, "global_step": 222838, "epoch": 2503, "lr": 7.733974254553394e-06}
{"train_loss": 0.02827535569667816, "global_step": 222839, "epoch": 2503, "lr": 7.733664527599194e-06}
{"train_loss": 0.06877806782722473, "global_step": 222840, "epoch": 2503, "lr": 7.73335480632717e-06}
{"train_loss": 0.012651605531573296, "global_step": 222841, "epoch": 2503, "lr": 7.733045090737395e-06}
{"train_loss": 0.013235400430858135, "global_step": 222842, "epoch": 2503, "lr": 7.732735380829886e-06}
{"train_loss": 0.05138137936592102, "global_step": 222843, "epoch": 2503, "lr": 7.732425676604705e-06}
{"train_loss": 0.05161060392856598, "global_step": 222844, "epoch": 2503, "lr": 7.732115978061878e-06}
{"train_loss": 0.05798126757144928, "global_step": 222845, "epoch": 2503, "lr": 7.731806285201449e-06}
{"train_loss": 0.07713872939348221, "global_step": 222846, "epoch": 2503, "lr": 7.731496598023475e-06}
{"train_loss": 0.027705565094947815, "global_step": 222847, "epoch": 2503, "lr": 7.731186916527972e-06}
{"train_loss": 0.030571237206459045, "global_step": 222848, "epoch": 2503, "lr": 7.730877240715011e-06}
{"train_loss": 0.06642502546310425, "global_step": 222849, "epoch": 2503, "lr": 7.730567570584607e-06}
{"train_loss": 0.08699195086956024, "global_step": 222850, "epoch": 2503, "lr": 7.730257906136823e-06}
{"train_loss": 0.05691149830818176, "global_step": 222851, "epoch": 2503, "lr": 7.729948247371676e-06}
{"train_loss": 0.045146506279706955, "global_step": 222852, "epoch": 2503, "lr": 7.72963859428924e-06}
{"train_loss": 0.08308935910463333, "global_step": 222853, "epoch": 2503, "lr": 7.729328946889525e-06}
{"train_loss": 0.060204409062862396, "global_step": 222854, "epoch": 2503, "lr": 7.729019305172597e-06}
{"train_loss": 0.04968354376974735, "global_step": 222855, "epoch": 2503, "lr": 7.728709669138478e-06, "val_loss": 9.01842975616455}
{"train_loss": 0.05741368234157562, "global_step": 222856, "epoch": 2504, "lr": 7.728400038787226e-06}
{"train_loss": 0.05437115207314491, "global_step": 222857, "epoch": 2504, "lr": 7.728090414118866e-06}
{"train_loss": 0.035759106278419495, "global_step": 222858, "epoch": 2504, "lr": 7.727780795133461e-06}
{"train_loss": 0.033768199384212494, "global_step": 222859, "epoch": 2504, "lr": 7.72747118183103e-06}
{"train_loss": 0.019122086465358734, "global_step": 222860, "epoch": 2504, "lr": 7.727161574211639e-06}
{"train_loss": 0.03132349252700806, "global_step": 222861, "epoch": 2504, "lr": 7.726851972275301e-06}
{"train_loss": 0.03338225558400154, "global_step": 222862, "epoch": 2504, "lr": 7.726542376022084e-06}
{"train_loss": 0.06304454058408737, "global_step": 222863, "epoch": 2504, "lr": 7.726232785452004e-06}
{"train_loss": 0.03763497620820999, "global_step": 222864, "epoch": 2504, "lr": 7.725923200565133e-06}
{"train_loss": 0.07062724232673645, "global_step": 222865, "epoch": 2504, "lr": 7.725613621361495e-06}
{"train_loss": 0.06490162760019302, "global_step": 222866, "epoch": 2504, "lr": 7.72530404784112e-06}
{"train_loss": 0.1056709736585617, "global_step": 222867, "epoch": 2504, "lr": 7.724994480004078e-06}
{"train_loss": 0.07034768164157867, "global_step": 222868, "epoch": 2504, "lr": 7.724684917850377e-06}
{"train_loss": 0.03516499325633049, "global_step": 222869, "epoch": 2504, "lr": 7.724375361380093e-06}
{"train_loss": 0.042341165244579315, "global_step": 222870, "epoch": 2504, "lr": 7.72406581059324e-06}
{"train_loss": 0.031143303960561752, "global_step": 222871, "epoch": 2504, "lr": 7.723756265489884e-06}
{"train_loss": 0.03011680208146572, "global_step": 222872, "epoch": 2504, "lr": 7.723446726070038e-06}
{"train_loss": 0.05692062899470329, "global_step": 222873, "epoch": 2504, "lr": 7.723137192333774e-06}
{"train_loss": 0.06724494695663452, "global_step": 222874, "epoch": 2504, "lr": 7.722827664281107e-06}
{"train_loss": 0.05213480442762375, "global_step": 222875, "epoch": 2504, "lr": 7.7225181419121e-06}
{"train_loss": 0.04433154687285423, "global_step": 222876, "epoch": 2504, "lr": 7.722208625226773e-06}
{"train_loss": 0.06376712024211884, "global_step": 222877, "epoch": 2504, "lr": 7.721899114225195e-06}
{"train_loss": 0.06165456399321556, "global_step": 222878, "epoch": 2504, "lr": 7.721589608907375e-06}
{"train_loss": 0.02589195780456066, "global_step": 222879, "epoch": 2504, "lr": 7.721280109273393e-06}
{"train_loss": 0.093831866979599, "global_step": 222880, "epoch": 2504, "lr": 7.720970615323248e-06}
{"train_loss": 0.058522436767816544, "global_step": 222881, "epoch": 2504, "lr": 7.720661127057011e-06}
{"train_loss": 0.08413290977478027, "global_step": 222882, "epoch": 2504, "lr": 7.720351644474722e-06}
{"train_loss": 0.03618597611784935, "global_step": 222883, "epoch": 2504, "lr": 7.720042167576408e-06}
{"train_loss": 0.08502957969903946, "global_step": 222884, "epoch": 2504, "lr": 7.719732696362131e-06}
{"train_loss": 0.024347921833395958, "global_step": 222885, "epoch": 2504, "lr": 7.719423230831907e-06}
{"train_loss": 0.06123228743672371, "global_step": 222886, "epoch": 2504, "lr": 7.719113770985803e-06}
{"train_loss": 0.04248904436826706, "global_step": 222887, "epoch": 2504, "lr": 7.718804316823835e-06}
{"train_loss": 0.0726013109087944, "global_step": 222888, "epoch": 2504, "lr": 7.718494868346076e-06}
{"train_loss": 0.030956575646996498, "global_step": 222889, "epoch": 2504, "lr": 7.718185425552531e-06}
{"train_loss": 0.07886066287755966, "global_step": 222890, "epoch": 2504, "lr": 7.71787598844328e-06}
{"train_loss": 0.036327119916677475, "global_step": 222891, "epoch": 2504, "lr": 7.71756655701833e-06}
{"train_loss": 0.040562596172094345, "global_step": 222892, "epoch": 2504, "lr": 7.717257131277749e-06}
{"train_loss": 0.038498952984809875, "global_step": 222893, "epoch": 2504, "lr": 7.716947711221556e-06}
{"train_loss": 0.04909566789865494, "global_step": 222894, "epoch": 2504, "lr": 7.716638296849815e-06}
{"train_loss": 0.04829515516757965, "global_step": 222895, "epoch": 2504, "lr": 7.716328888162544e-06}
{"train_loss": 0.04294496774673462, "global_step": 222896, "epoch": 2504, "lr": 7.716019485159814e-06}
{"train_loss": 0.03499782085418701, "global_step": 222897, "epoch": 2504, "lr": 7.715710087841632e-06}
{"train_loss": 0.07605158537626266, "global_step": 222898, "epoch": 2504, "lr": 7.715400696208074e-06}
{"train_loss": 0.044108252972364426, "global_step": 222899, "epoch": 2504, "lr": 7.715091310259148e-06}
{"train_loss": 0.0699818879365921, "global_step": 222900, "epoch": 2504, "lr": 7.71478192999493e-06}
{"train_loss": 0.045518144965171814, "global_step": 222901, "epoch": 2504, "lr": 7.71447255541543e-06}
{"train_loss": 0.03276398405432701, "global_step": 222902, "epoch": 2504, "lr": 7.714163186520718e-06}
{"train_loss": 0.09993533790111542, "global_step": 222903, "epoch": 2504, "lr": 7.713853823310818e-06}
{"train_loss": 0.036087796092033386, "global_step": 222904, "epoch": 2504, "lr": 7.713544465785761e-06}
{"train_loss": 0.1254679560661316, "global_step": 222905, "epoch": 2504, "lr": 7.713235113945616e-06}
{"train_loss": 0.10505624860525131, "global_step": 222906, "epoch": 2504, "lr": 7.712925767790401e-06}
{"train_loss": 0.06671784073114395, "global_step": 222907, "epoch": 2504, "lr": 7.71261642732018e-06}
{"train_loss": 0.033547259867191315, "global_step": 222908, "epoch": 2504, "lr": 7.712307092534971e-06}
{"train_loss": 0.06356927752494812, "global_step": 222909, "epoch": 2504, "lr": 7.711997763434836e-06}
{"train_loss": 0.03674270212650299, "global_step": 222910, "epoch": 2504, "lr": 7.7116884400198e-06}
{"train_loss": 0.06377533078193665, "global_step": 222911, "epoch": 2504, "lr": 7.711379122289919e-06}
{"train_loss": 0.05428812652826309, "global_step": 222912, "epoch": 2504, "lr": 7.711069810245219e-06}
{"train_loss": 0.05630701035261154, "global_step": 222913, "epoch": 2504, "lr": 7.710760503885761e-06}
{"train_loss": 0.024260761216282845, "global_step": 222914, "epoch": 2504, "lr": 7.71045120321156e-06}
{"train_loss": 0.10134663432836533, "global_step": 222915, "epoch": 2504, "lr": 7.71014190822269e-06}
{"train_loss": 0.04282277822494507, "global_step": 222916, "epoch": 2504, "lr": 7.709832618919166e-06}
{"train_loss": 0.0364447645843029, "global_step": 222917, "epoch": 2504, "lr": 7.709523335301039e-06}
{"train_loss": 0.1451989710330963, "global_step": 222918, "epoch": 2504, "lr": 7.709214057368358e-06}
{"train_loss": 0.02800760604441166, "global_step": 222919, "epoch": 2504, "lr": 7.708904785121147e-06}
{"train_loss": 0.06115616112947464, "global_step": 222920, "epoch": 2504, "lr": 7.708595518559475e-06}
{"train_loss": 0.044964518398046494, "global_step": 222921, "epoch": 2504, "lr": 7.70828625768335e-06}
{"train_loss": 0.04511174559593201, "global_step": 222922, "epoch": 2504, "lr": 7.707977002492845e-06}
{"train_loss": 0.03863092139363289, "global_step": 222923, "epoch": 2504, "lr": 7.707667752987974e-06}
{"train_loss": 0.059349607676267624, "global_step": 222924, "epoch": 2504, "lr": 7.707358509168805e-06}
{"train_loss": 0.05105322226881981, "global_step": 222925, "epoch": 2504, "lr": 7.707049271035349e-06}
{"train_loss": 0.046574562788009644, "global_step": 222926, "epoch": 2504, "lr": 7.706740038587685e-06}
{"train_loss": 0.02214387245476246, "global_step": 222927, "epoch": 2504, "lr": 7.706430811825816e-06}
{"train_loss": 0.0400318019092083, "global_step": 222928, "epoch": 2504, "lr": 7.706121590749815e-06}
{"train_loss": 0.08762551844120026, "global_step": 222929, "epoch": 2504, "lr": 7.7058123753597e-06}
{"train_loss": 0.0415964275598526, "global_step": 222930, "epoch": 2504, "lr": 7.705503165655537e-06}
{"train_loss": 0.056609369814395905, "global_step": 222931, "epoch": 2504, "lr": 7.705193961637336e-06}
{"train_loss": 0.03434094414114952, "global_step": 222932, "epoch": 2504, "lr": 7.70488476330517e-06}
{"train_loss": 0.060105711221694946, "global_step": 222933, "epoch": 2504, "lr": 7.704575570659057e-06}
{"train_loss": 0.042331937700510025, "global_step": 222934, "epoch": 2504, "lr": 7.70426638369906e-06}
{"train_loss": 0.06443432718515396, "global_step": 222935, "epoch": 2504, "lr": 7.703957202425194e-06}
{"train_loss": 0.05983898416161537, "global_step": 222936, "epoch": 2504, "lr": 7.703648026837517e-06}
{"train_loss": 0.14461025595664978, "global_step": 222937, "epoch": 2504, "lr": 7.703338856936082e-06}
{"train_loss": 0.05079829320311546, "global_step": 222938, "epoch": 2504, "lr": 7.70302969272092e-06}
{"train_loss": 0.03532432019710541, "global_step": 222939, "epoch": 2504, "lr": 7.702720534192055e-06}
{"train_loss": 0.030935408547520638, "global_step": 222940, "epoch": 2504, "lr": 7.702411381349555e-06}
{"train_loss": 0.06396634876728058, "global_step": 222941, "epoch": 2504, "lr": 7.702102234193442e-06}
{"train_loss": 0.039114274084568024, "global_step": 222942, "epoch": 2504, "lr": 7.701793092723774e-06}
{"train_loss": 0.01938658207654953, "global_step": 222943, "epoch": 2504, "lr": 7.701483956940586e-06}
{"train_loss": 0.05422874083847142, "global_step": 222944, "epoch": 2504, "lr": 7.701174826843904e-06, "val_loss": 8.9419527053833}
{"train_loss": 0.03365733474493027, "global_step": 222945, "epoch": 2505, "lr": 7.700865702433797e-06}
{"train_loss": 0.06691917032003403, "global_step": 222946, "epoch": 2505, "lr": 7.700556583710277e-06}
{"train_loss": 0.030276259407401085, "global_step": 222947, "epoch": 2505, "lr": 7.700247470673422e-06}
{"train_loss": 0.02576456218957901, "global_step": 222948, "epoch": 2505, "lr": 7.699938363323234e-06}
{"train_loss": 0.06573314964771271, "global_step": 222949, "epoch": 2505, "lr": 7.699629261659791e-06}
{"train_loss": 0.04948034510016441, "global_step": 222950, "epoch": 2505, "lr": 7.6993201656831e-06}
{"train_loss": 0.05910487473011017, "global_step": 222951, "epoch": 2505, "lr": 7.699011075393236e-06}
{"train_loss": 0.07128305733203888, "global_step": 222952, "epoch": 2505, "lr": 7.698701990790208e-06}
{"train_loss": 0.033835217356681824, "global_step": 222953, "epoch": 2505, "lr": 7.69839291187408e-06}
{"train_loss": 0.040287140756845474, "global_step": 222954, "epoch": 2505, "lr": 7.698083838644899e-06}
{"train_loss": 0.012070877477526665, "global_step": 222955, "epoch": 2505, "lr": 7.697774771102678e-06}
{"train_loss": 0.05784623324871063, "global_step": 222956, "epoch": 2505, "lr": 7.697465709247487e-06}
{"train_loss": 0.044652700424194336, "global_step": 222957, "epoch": 2505, "lr": 7.697156653079351e-06}
{"train_loss": 0.06943058222532272, "global_step": 222958, "epoch": 2505, "lr": 7.696847602598323e-06}
{"train_loss": 0.05644610896706581, "global_step": 222959, "epoch": 2505, "lr": 7.696538557804429e-06}
{"train_loss": 0.06785773485898972, "global_step": 222960, "epoch": 2505, "lr": 7.696229518697729e-06}
{"train_loss": 0.06113963574171066, "global_step": 222961, "epoch": 2505, "lr": 7.695920485278247e-06}
{"train_loss": 0.07068952172994614, "global_step": 222962, "epoch": 2505, "lr": 7.695611457546042e-06}
{"train_loss": 0.035894736647605896, "global_step": 222963, "epoch": 2505, "lr": 7.695302435501134e-06}
{"train_loss": 0.0987347960472107, "global_step": 222964, "epoch": 2505, "lr": 7.694993419143592e-06}
{"train_loss": 0.03478596732020378, "global_step": 222965, "epoch": 2505, "lr": 7.69468440847343e-06}
{"train_loss": 0.06644447147846222, "global_step": 222966, "epoch": 2505, "lr": 7.694375403490711e-06}
{"train_loss": 0.027374349534511566, "global_step": 222967, "epoch": 2505, "lr": 7.69406640419546e-06}
{"train_loss": 0.11880747973918915, "global_step": 222968, "epoch": 2505, "lr": 7.693757410587732e-06}
{"train_loss": 0.06564151495695114, "global_step": 222969, "epoch": 2505, "lr": 7.693448422667555e-06}
{"train_loss": 0.017343057319521904, "global_step": 222970, "epoch": 2505, "lr": 7.693139440434987e-06}
{"train_loss": 0.055121637880802155, "global_step": 222971, "epoch": 2505, "lr": 7.692830463890056e-06}
{"train_loss": 0.05572903901338577, "global_step": 222972, "epoch": 2505, "lr": 7.6925214930328e-06}
{"train_loss": 0.08163813501596451, "global_step": 222973, "epoch": 2505, "lr": 7.692212527863286e-06}
{"train_loss": 0.05385841801762581, "global_step": 222974, "epoch": 2505, "lr": 7.691903568381525e-06}
{"train_loss": 0.09952607750892639, "global_step": 222975, "epoch": 2505, "lr": 7.691594614587582e-06}
{"train_loss": 0.035732999444007874, "global_step": 222976, "epoch": 2505, "lr": 7.69128566648149e-06}
{"train_loss": 0.05210200697183609, "global_step": 222977, "epoch": 2505, "lr": 7.690976724063281e-06}
{"train_loss": 0.05525454133749008, "global_step": 222978, "epoch": 2505, "lr": 7.690667787333012e-06}
{"train_loss": 0.034492917358875275, "global_step": 222979, "epoch": 2505, "lr": 7.690358856290702e-06}
{"train_loss": 0.0421452522277832, "global_step": 222980, "epoch": 2505, "lr": 7.690049930936428e-06}
{"train_loss": 0.027209363877773285, "global_step": 222981, "epoch": 2505, "lr": 7.689741011270201e-06}
{"train_loss": 0.04414111375808716, "global_step": 222982, "epoch": 2505, "lr": 7.68943209729207e-06}
{"train_loss": 0.035993386059999466, "global_step": 222983, "epoch": 2505, "lr": 7.689123189002084e-06}
{"train_loss": 0.051788464188575745, "global_step": 222984, "epoch": 2505, "lr": 7.688814286400269e-06}
{"train_loss": 0.05513372644782066, "global_step": 222985, "epoch": 2505, "lr": 7.688505389486694e-06}
{"train_loss": 0.05484601855278015, "global_step": 222986, "epoch": 2505, "lr": 7.688196498261369e-06}
{"train_loss": 0.04141725227236748, "global_step": 222987, "epoch": 2505, "lr": 7.687887612724365e-06}
{"train_loss": 0.03751553222537041, "global_step": 222988, "epoch": 2505, "lr": 7.687578732875688e-06}
{"train_loss": 0.06455211341381073, "global_step": 222989, "epoch": 2505, "lr": 7.687269858715407e-06}
{"train_loss": 0.053155045956373215, "global_step": 222990, "epoch": 2505, "lr": 7.686960990243569e-06}
{"train_loss": 0.04469731077551842, "global_step": 222991, "epoch": 2505, "lr": 7.686652127460192e-06}
{"train_loss": 0.03132301941514015, "global_step": 222992, "epoch": 2505, "lr": 7.686343270365336e-06}
{"train_loss": 0.05119471251964569, "global_step": 222993, "epoch": 2505, "lr": 7.686034418959032e-06}
{"train_loss": 0.07377072423696518, "global_step": 222994, "epoch": 2505, "lr": 7.685725573241331e-06}
{"train_loss": 0.06277991831302643, "global_step": 222995, "epoch": 2505, "lr": 7.685416733212253e-06}
{"train_loss": 0.06507831066846848, "global_step": 222996, "epoch": 2505, "lr": 7.685107898871875e-06}
{"train_loss": 0.15427038073539734, "global_step": 222997, "epoch": 2505, "lr": 7.6847990702202e-06}
{"train_loss": 0.038158632814884186, "global_step": 222998, "epoch": 2505, "lr": 7.684490247257303e-06}
{"train_loss": 0.05546519532799721, "global_step": 222999, "epoch": 2505, "lr": 7.6841814299832e-06}
{"train_loss": 0.02343299612402916, "global_step": 223000, "epoch": 2505, "lr": 7.683872618397953e-06}
{"train_loss": 0.0960119366645813, "global_step": 223001, "epoch": 2505, "lr": 7.683563812501582e-06}
{"train_loss": 0.04134649410843849, "global_step": 223002, "epoch": 2505, "lr": 7.68325501229415e-06}
{"train_loss": 0.05286646634340286, "global_step": 223003, "epoch": 2505, "lr": 7.682946217775677e-06}
{"train_loss": 0.06634698063135147, "global_step": 223004, "epoch": 2505, "lr": 7.682637428946232e-06}
{"train_loss": 0.05227302387356758, "global_step": 223005, "epoch": 2505, "lr": 7.682328645805825e-06}
{"train_loss": 0.05268357694149017, "global_step": 223006, "epoch": 2505, "lr": 7.682019868354529e-06}
{"train_loss": 0.023020392283797264, "global_step": 223007, "epoch": 2505, "lr": 7.681711096592354e-06}
{"train_loss": 0.053386714309453964, "global_step": 223008, "epoch": 2505, "lr": 7.681402330519355e-06}
{"train_loss": 0.07559671998023987, "global_step": 223009, "epoch": 2505, "lr": 7.681093570135594e-06}
{"train_loss": 0.03701505810022354, "global_step": 223010, "epoch": 2505, "lr": 7.680784815441077e-06}
{"train_loss": 0.051300421357154846, "global_step": 223011, "epoch": 2505, "lr": 7.680476066435877e-06}
{"train_loss": 0.04627874121069908, "global_step": 223012, "epoch": 2505, "lr": 7.680167323120008e-06}
{"train_loss": 0.04399017244577408, "global_step": 223013, "epoch": 2505, "lr": 7.679858585493538e-06}
{"train_loss": 0.0417296402156353, "global_step": 223014, "epoch": 2505, "lr": 7.679549853556494e-06}
{"train_loss": 0.13000966608524323, "global_step": 223015, "epoch": 2505, "lr": 7.679241127308906e-06}
{"train_loss": 0.050935715436935425, "global_step": 223016, "epoch": 2505, "lr": 7.678932406750843e-06}
{"train_loss": 0.0368366464972496, "global_step": 223017, "epoch": 2505, "lr": 7.67862369188232e-06}
{"train_loss": 0.06887633353471756, "global_step": 223018, "epoch": 2505, "lr": 7.6783149827034e-06}
{"train_loss": 0.019158201292157173, "global_step": 223019, "epoch": 2505, "lr": 7.678006279214111e-06}
{"train_loss": 0.043531984090805054, "global_step": 223020, "epoch": 2505, "lr": 7.677697581414495e-06}
{"train_loss": 0.10271773487329483, "global_step": 223021, "epoch": 2505, "lr": 7.677388889304603e-06}
{"train_loss": 0.05944685637950897, "global_step": 223022, "epoch": 2505, "lr": 7.677080202884457e-06}
{"train_loss": 0.1023322120308876, "global_step": 223023, "epoch": 2505, "lr": 7.676771522154131e-06}
{"train_loss": 0.042213551700115204, "global_step": 223024, "epoch": 2505, "lr": 7.67646284711363e-06}
{"train_loss": 0.06510668247938156, "global_step": 223025, "epoch": 2505, "lr": 7.676154177763012e-06}
{"train_loss": 0.0750151053071022, "global_step": 223026, "epoch": 2505, "lr": 7.675845514102331e-06}
{"train_loss": 0.059223879128694534, "global_step": 223027, "epoch": 2505, "lr": 7.675536856131604e-06}
{"train_loss": 0.08090148866176605, "global_step": 223028, "epoch": 2505, "lr": 7.675228203850899e-06}
{"train_loss": 0.06722942739725113, "global_step": 223029, "epoch": 2505, "lr": 7.674919557260235e-06}
{"train_loss": 0.03583736717700958, "global_step": 223030, "epoch": 2505, "lr": 7.67461091635967e-06}
{"train_loss": 0.07750537246465683, "global_step": 223031, "epoch": 2505, "lr": 7.674302281149227e-06}
{"train_loss": 0.04863264039158821, "global_step": 223032, "epoch": 2505, "lr": 7.673993651628969e-06}
{"train_loss": 0.055794696438680874, "global_step": 223033, "epoch": 2505, "lr": 7.673685027798916e-06, "val_loss": 9.024724960327148, "train_action_mse_error": 29.361492156982422}
{"train_loss": 0.04217781499028206, "global_step": 223034, "epoch": 2506, "lr": 7.673376409659133e-06}
{"train_loss": 0.02808096446096897, "global_step": 223035, "epoch": 2506, "lr": 7.673067797209632e-06}
{"train_loss": 0.0820438489317894, "global_step": 223036, "epoch": 2506, "lr": 7.672759190450491e-06}
{"train_loss": 0.06232854351401329, "global_step": 223037, "epoch": 2506, "lr": 7.672450589381713e-06}
{"train_loss": 0.04595586284995079, "global_step": 223038, "epoch": 2506, "lr": 7.672141994003373e-06}
{"train_loss": 0.048782434314489365, "global_step": 223039, "epoch": 2506, "lr": 7.671833404315487e-06}
{"train_loss": 0.06337091326713562, "global_step": 223040, "epoch": 2506, "lr": 7.671524820318116e-06}
{"train_loss": 0.08377858251333237, "global_step": 223041, "epoch": 2506, "lr": 7.671216242011281e-06}
{"train_loss": 0.025343913584947586, "global_step": 223042, "epoch": 2506, "lr": 7.670907669395045e-06}
{"train_loss": 0.05933922901749611, "global_step": 223043, "epoch": 2506, "lr": 7.670599102469433e-06}
{"train_loss": 0.04586685448884964, "global_step": 223044, "epoch": 2506, "lr": 7.670290541234488e-06}
{"train_loss": 0.038580358028411865, "global_step": 223045, "epoch": 2506, "lr": 7.669981985690271e-06}
{"train_loss": 0.057799287140369415, "global_step": 223046, "epoch": 2506, "lr": 7.6696734358368e-06}
{"train_loss": 0.04264027625322342, "global_step": 223047, "epoch": 2506, "lr": 7.66936489167413e-06}
{"train_loss": 0.03312312066555023, "global_step": 223048, "epoch": 2506, "lr": 7.669056353202292e-06}
{"train_loss": 0.033458586782217026, "global_step": 223049, "epoch": 2506, "lr": 7.668747820421346e-06}
{"train_loss": 0.06107601150870323, "global_step": 223050, "epoch": 2506, "lr": 7.668439293331304e-06}
{"train_loss": 0.061000704765319824, "global_step": 223051, "epoch": 2506, "lr": 7.668130771932242e-06}
{"train_loss": 0.06573823094367981, "global_step": 223052, "epoch": 2506, "lr": 7.667822256224166e-06}
{"train_loss": 0.04352489858865738, "global_step": 223053, "epoch": 2506, "lr": 7.66751374620715e-06}
{"train_loss": 0.11372638493776321, "global_step": 223054, "epoch": 2506, "lr": 7.66720524188122e-06}
{"train_loss": 0.05164501443505287, "global_step": 223055, "epoch": 2506, "lr": 7.66689674324641e-06}
{"train_loss": 0.01645897701382637, "global_step": 223056, "epoch": 2506, "lr": 7.666588250302775e-06}
{"train_loss": 0.07619860768318176, "global_step": 223057, "epoch": 2506, "lr": 7.666279763050344e-06}
{"train_loss": 0.042377788573503494, "global_step": 223058, "epoch": 2506, "lr": 7.665971281489175e-06}
{"train_loss": 0.03310002386569977, "global_step": 223059, "epoch": 2506, "lr": 7.665662805619305e-06}
{"train_loss": 0.023329507559537888, "global_step": 223060, "epoch": 2506, "lr": 7.665354335440755e-06}
{"train_loss": 0.036804478615522385, "global_step": 223061, "epoch": 2506, "lr": 7.665045870953597e-06}
{"train_loss": 0.0487937405705452, "global_step": 223062, "epoch": 2506, "lr": 7.66473741215784e-06}
{"train_loss": 0.06550894677639008, "global_step": 223063, "epoch": 2506, "lr": 7.66442895905355e-06}
{"train_loss": 0.047637373208999634, "global_step": 223064, "epoch": 2506, "lr": 7.664120511640772e-06}
{"train_loss": 0.09534183144569397, "global_step": 223065, "epoch": 2506, "lr": 7.663812069919518e-06}
{"train_loss": 0.0389576256275177, "global_step": 223066, "epoch": 2506, "lr": 7.663503633889868e-06}
{"train_loss": 0.0359007902443409, "global_step": 223067, "epoch": 2506, "lr": 7.663195203551832e-06}
{"train_loss": 0.02757320925593376, "global_step": 223068, "epoch": 2506, "lr": 7.66288677890547e-06}
{"train_loss": 0.06542140245437622, "global_step": 223069, "epoch": 2506, "lr": 7.662578359950806e-06}
{"train_loss": 0.08819480985403061, "global_step": 223070, "epoch": 2506, "lr": 7.662269946687906e-06}
{"train_loss": 0.0668470486998558, "global_step": 223071, "epoch": 2506, "lr": 7.661961539116785e-06}
{"train_loss": 0.04473523050546646, "global_step": 223072, "epoch": 2506, "lr": 7.661653137237513e-06}
{"train_loss": 0.022616490721702576, "global_step": 223073, "epoch": 2506, "lr": 7.661344741050102e-06}
{"train_loss": 0.06315749138593674, "global_step": 223074, "epoch": 2506, "lr": 7.661036350554618e-06}
{"train_loss": 0.03468858450651169, "global_step": 223075, "epoch": 2506, "lr": 7.66072796575108e-06}
{"train_loss": 0.04690103977918625, "global_step": 223076, "epoch": 2506, "lr": 7.66041958663955e-06}
{"train_loss": 0.036532290279865265, "global_step": 223077, "epoch": 2506, "lr": 7.66011121322005e-06}
{"train_loss": 0.024366345256567, "global_step": 223078, "epoch": 2506, "lr": 7.659802845492648e-06}
{"train_loss": 0.04277396202087402, "global_step": 223079, "epoch": 2506, "lr": 7.659494483457352e-06}
{"train_loss": 0.04323242977261543, "global_step": 223080, "epoch": 2506, "lr": 7.659186127114225e-06}
{"train_loss": 0.040164995938539505, "global_step": 223081, "epoch": 2506, "lr": 7.658877776463313e-06}
{"train_loss": 0.07901377975940704, "global_step": 223082, "epoch": 2506, "lr": 7.658569431504641e-06}
{"train_loss": 0.04485004022717476, "global_step": 223083, "epoch": 2506, "lr": 7.658261092238267e-06}
{"train_loss": 0.03696606308221817, "global_step": 223084, "epoch": 2506, "lr": 7.657952758664211e-06}
{"train_loss": 0.14025025069713593, "global_step": 223085, "epoch": 2506, "lr": 7.65764443078254e-06}
{"train_loss": 0.05789943411946297, "global_step": 223086, "epoch": 2506, "lr": 7.657336108593272e-06}
{"train_loss": 0.03973659127950668, "global_step": 223087, "epoch": 2506, "lr": 7.657027792096473e-06}
{"train_loss": 0.0858047753572464, "global_step": 223088, "epoch": 2506, "lr": 7.656719481292157e-06}
{"train_loss": 0.05717294663190842, "global_step": 223089, "epoch": 2506, "lr": 7.656411176180388e-06}
{"train_loss": 0.036223944276571274, "global_step": 223090, "epoch": 2506, "lr": 7.65610287676119e-06}
{"train_loss": 0.017285214737057686, "global_step": 223091, "epoch": 2506, "lr": 7.65579458303462e-06}
{"train_loss": 0.05528215691447258, "global_step": 223092, "epoch": 2506, "lr": 7.65548629500072e-06}
{"train_loss": 0.07287903875112534, "global_step": 223093, "epoch": 2506, "lr": 7.655178012659503e-06}
{"train_loss": 0.015280167572200298, "global_step": 223094, "epoch": 2506, "lr": 7.65486973601105e-06}
{"train_loss": 0.049150250852108, "global_step": 223095, "epoch": 2506, "lr": 7.65456146505537e-06}
{"train_loss": 0.06247865781188011, "global_step": 223096, "epoch": 2506, "lr": 7.654253199792526e-06}
{"train_loss": 0.09573833644390106, "global_step": 223097, "epoch": 2506, "lr": 7.653944940222557e-06}
{"train_loss": 0.028870727866888046, "global_step": 223098, "epoch": 2506, "lr": 7.653636686345484e-06}
{"train_loss": 0.06586875766515732, "global_step": 223099, "epoch": 2506, "lr": 7.653328438161367e-06}
{"train_loss": 0.061482056975364685, "global_step": 223100, "epoch": 2506, "lr": 7.653020195670251e-06}
{"train_loss": 0.03266840800642967, "global_step": 223101, "epoch": 2506, "lr": 7.652711958872161e-06}
{"train_loss": 0.0342332124710083, "global_step": 223102, "epoch": 2506, "lr": 7.65240372776716e-06}
{"train_loss": 0.02784513309597969, "global_step": 223103, "epoch": 2506, "lr": 7.652095502355262e-06}
{"train_loss": 0.07926225662231445, "global_step": 223104, "epoch": 2506, "lr": 7.65178728263654e-06}
{"train_loss": 0.043483857065439224, "global_step": 223105, "epoch": 2506, "lr": 7.651479068611006e-06}
{"train_loss": 0.08947653323411942, "global_step": 223106, "epoch": 2506, "lr": 7.651170860278723e-06}
{"train_loss": 0.0392192080616951, "global_step": 223107, "epoch": 2506, "lr": 7.650862657639719e-06}
{"train_loss": 0.06463706493377686, "global_step": 223108, "epoch": 2506, "lr": 7.650554460694042e-06}
{"train_loss": 0.029803525656461716, "global_step": 223109, "epoch": 2506, "lr": 7.65024626944173e-06}
{"train_loss": 0.05042519420385361, "global_step": 223110, "epoch": 2506, "lr": 7.64993808388283e-06}
{"train_loss": 0.028433984145522118, "global_step": 223111, "epoch": 2506, "lr": 7.649629904017375e-06}
{"train_loss": 0.05221915990114212, "global_step": 223112, "epoch": 2506, "lr": 7.649321729845416e-06}
{"train_loss": 0.013767672702670097, "global_step": 223113, "epoch": 2506, "lr": 7.64901356136698e-06}
{"train_loss": 0.0464681051671505, "global_step": 223114, "epoch": 2506, "lr": 7.648705398582135e-06}
{"train_loss": 0.05063154548406601, "global_step": 223115, "epoch": 2506, "lr": 7.64839724149089e-06}
{"train_loss": 0.027746595442295074, "global_step": 223116, "epoch": 2506, "lr": 7.648089090093296e-06}
{"train_loss": 0.03112562745809555, "global_step": 223117, "epoch": 2506, "lr": 7.647780944389422e-06}
{"train_loss": 0.05709529295563698, "global_step": 223118, "epoch": 2506, "lr": 7.64747280437927e-06}
{"train_loss": 0.052573319524526596, "global_step": 223119, "epoch": 2506, "lr": 7.647164670062911e-06}
{"train_loss": 0.05321647226810455, "global_step": 223120, "epoch": 2506, "lr": 7.64685654144036e-06}
{"train_loss": 0.03524494543671608, "global_step": 223121, "epoch": 2506, "lr": 7.64654841851169e-06}
{"train_loss": 0.05032780945426627, "global_step": 223122, "epoch": 2506, "lr": 7.646240301276913e-06, "val_loss": 8.89684772491455}
{"train_loss": 0.06705737113952637, "global_step": 223123, "epoch": 2507, "lr": 7.645932189736093e-06}
{"train_loss": 0.03283059224486351, "global_step": 223124, "epoch": 2507, "lr": 7.645624083889253e-06}
{"train_loss": 0.053975895047187805, "global_step": 223125, "epoch": 2507, "lr": 7.64531598373645e-06}
{"train_loss": 0.03547300398349762, "global_step": 223126, "epoch": 2507, "lr": 7.645007889277705e-06}
{"train_loss": 0.022788936272263527, "global_step": 223127, "epoch": 2507, "lr": 7.644699800513088e-06}
{"train_loss": 0.09431950002908707, "global_step": 223128, "epoch": 2507, "lr": 7.644391717442612e-06}
{"train_loss": 0.059303171932697296, "global_step": 223129, "epoch": 2507, "lr": 7.644083640066346e-06}
{"train_loss": 0.045212484896183014, "global_step": 223130, "epoch": 2507, "lr": 7.64377556838431e-06}
{"train_loss": 0.0919584333896637, "global_step": 223131, "epoch": 2507, "lr": 7.643467502396546e-06}
{"train_loss": 0.04701848700642586, "global_step": 223132, "epoch": 2507, "lr": 7.64315944210311e-06}
{"train_loss": 0.0651637464761734, "global_step": 223133, "epoch": 2507, "lr": 7.642851387504024e-06}
{"train_loss": 0.04815395921468735, "global_step": 223134, "epoch": 2507, "lr": 7.642543338599357e-06}
{"train_loss": 0.04540548473596573, "global_step": 223135, "epoch": 2507, "lr": 7.642235295389117e-06}
{"train_loss": 0.05102279782295227, "global_step": 223136, "epoch": 2507, "lr": 7.641927257873371e-06}
{"train_loss": 0.06035665422677994, "global_step": 223137, "epoch": 2507, "lr": 7.641619226052143e-06}
{"train_loss": 0.07222478091716766, "global_step": 223138, "epoch": 2507, "lr": 7.6413111999255e-06}
{"train_loss": 0.043531425297260284, "global_step": 223139, "epoch": 2507, "lr": 7.641003179493444e-06}
{"train_loss": 0.1087409257888794, "global_step": 223140, "epoch": 2507, "lr": 7.640695164756063e-06}
{"train_loss": 0.05023707076907158, "global_step": 223141, "epoch": 2507, "lr": 7.640387155713351e-06}
{"train_loss": 0.05251007899641991, "global_step": 223142, "epoch": 2507, "lr": 7.640079152365392e-06}
{"train_loss": 0.04193321615457535, "global_step": 223143, "epoch": 2507, "lr": 7.639771154712194e-06}
{"train_loss": 0.049112070351839066, "global_step": 223144, "epoch": 2507, "lr": 7.639463162753824e-06}
{"train_loss": 0.025761930271983147, "global_step": 223145, "epoch": 2507, "lr": 7.639155176490299e-06}
{"train_loss": 0.07328654825687408, "global_step": 223146, "epoch": 2507, "lr": 7.638847195921683e-06}
{"train_loss": 0.04316570609807968, "global_step": 223147, "epoch": 2507, "lr": 7.638539221048003e-06}
{"train_loss": 0.08958686143159866, "global_step": 223148, "epoch": 2507, "lr": 7.63823125186931e-06}
{"train_loss": 0.06467252224683762, "global_step": 223149, "epoch": 2507, "lr": 7.637923288385635e-06}
{"train_loss": 0.08007259666919708, "global_step": 223150, "epoch": 2507, "lr": 7.637615330597031e-06}
{"train_loss": 0.01594592072069645, "global_step": 223151, "epoch": 2507, "lr": 7.637307378503521e-06}
{"train_loss": 0.06928732991218567, "global_step": 223152, "epoch": 2507, "lr": 7.636999432105163e-06}
{"train_loss": 0.04255814850330353, "global_step": 223153, "epoch": 2507, "lr": 7.636691491402003e-06}
{"train_loss": 0.02857748046517372, "global_step": 223154, "epoch": 2507, "lr": 7.636383556394066e-06}
{"train_loss": 0.030967803671956062, "global_step": 223155, "epoch": 2507, "lr": 7.636075627081413e-06}
{"train_loss": 0.052699364721775055, "global_step": 223156, "epoch": 2507, "lr": 7.635767703464053e-06}
{"train_loss": 0.07933233678340912, "global_step": 223157, "epoch": 2507, "lr": 7.635459785542065e-06}
{"train_loss": 0.058008238673210144, "global_step": 223158, "epoch": 2507, "lr": 7.635151873315465e-06}
{"train_loss": 0.061268582940101624, "global_step": 223159, "epoch": 2507, "lr": 7.634843966784312e-06}
{"train_loss": 0.04286434128880501, "global_step": 223160, "epoch": 2507, "lr": 7.634536065948621e-06}
{"train_loss": 0.05104383826255798, "global_step": 223161, "epoch": 2507, "lr": 7.634228170808467e-06}
{"train_loss": 0.046804193407297134, "global_step": 223162, "epoch": 2507, "lr": 7.63392028136386e-06}
{"train_loss": 0.028448235243558884, "global_step": 223163, "epoch": 2507, "lr": 7.63361239761487e-06}
{"train_loss": 0.053993210196495056, "global_step": 223164, "epoch": 2507, "lr": 7.633304519561512e-06}
{"train_loss": 0.04631496220827103, "global_step": 223165, "epoch": 2507, "lr": 7.63299664720385e-06}
{"train_loss": 0.05500050261616707, "global_step": 223166, "epoch": 2507, "lr": 7.632688780541902e-06}
{"train_loss": 0.037040628492832184, "global_step": 223167, "epoch": 2507, "lr": 7.632380919575738e-06}
{"train_loss": 0.030092773959040642, "global_step": 223168, "epoch": 2507, "lr": 7.632073064305373e-06}
{"train_loss": 0.05843536928296089, "global_step": 223169, "epoch": 2507, "lr": 7.631765214730868e-06}
{"train_loss": 0.06684128195047379, "global_step": 223170, "epoch": 2507, "lr": 7.631457370852258e-06}
{"train_loss": 0.035141389816999435, "global_step": 223171, "epoch": 2507, "lr": 7.631149532669569e-06}
{"train_loss": 0.039026662707328796, "global_step": 223172, "epoch": 2507, "lr": 7.630841700182873e-06}
{"train_loss": 0.06697750836610794, "global_step": 223173, "epoch": 2507, "lr": 7.630533873392177e-06}
{"train_loss": 0.049832783639431, "global_step": 223174, "epoch": 2507, "lr": 7.63022605229755e-06}
{"train_loss": 0.0602068230509758, "global_step": 223175, "epoch": 2507, "lr": 7.629918236899009e-06}
{"train_loss": 0.079288050532341, "global_step": 223176, "epoch": 2507, "lr": 7.629610427196626e-06}
{"train_loss": 0.07182244211435318, "global_step": 223177, "epoch": 2507, "lr": 7.6293026231904094e-06}
{"train_loss": 0.054592035710811615, "global_step": 223178, "epoch": 2507, "lr": 7.6289948248804315e-06}
{"train_loss": 0.018467392772436142, "global_step": 223179, "epoch": 2507, "lr": 7.628687032266707e-06}
{"train_loss": 0.046841517090797424, "global_step": 223180, "epoch": 2507, "lr": 7.628379245349298e-06}
{"train_loss": 0.029845481738448143, "global_step": 223181, "epoch": 2507, "lr": 7.628071464128228e-06}
{"train_loss": 0.030022135004401207, "global_step": 223182, "epoch": 2507, "lr": 7.627763688603562e-06}
{"train_loss": 0.07041732221841812, "global_step": 223183, "epoch": 2507, "lr": 7.627455918775311e-06}
{"train_loss": 0.06423793733119965, "global_step": 223184, "epoch": 2507, "lr": 7.627148154643543e-06}
{"train_loss": 0.0755167156457901, "global_step": 223185, "epoch": 2507, "lr": 7.6268403962082794e-06}
{"train_loss": 0.03503654897212982, "global_step": 223186, "epoch": 2507, "lr": 7.62653264346958e-06}
{"train_loss": 0.05405771732330322, "global_step": 223187, "epoch": 2507, "lr": 7.626224896427464e-06}
{"train_loss": 0.04081220179796219, "global_step": 223188, "epoch": 2507, "lr": 7.625917155082002e-06}
{"train_loss": 0.07112332433462143, "global_step": 223189, "epoch": 2507, "lr": 7.625609419433205e-06}
{"train_loss": 0.06000378727912903, "global_step": 223190, "epoch": 2507, "lr": 7.625301689481129e-06}
{"train_loss": 0.04227931797504425, "global_step": 223191, "epoch": 2507, "lr": 7.624993965225824e-06}
{"train_loss": 0.07116655260324478, "global_step": 223192, "epoch": 2507, "lr": 7.6246862466673115e-06}
{"train_loss": 0.03195299953222275, "global_step": 223193, "epoch": 2507, "lr": 7.624378533805654e-06}
{"train_loss": 0.05940217152237892, "global_step": 223194, "epoch": 2507, "lr": 7.624070826640872e-06}
{"train_loss": 0.05739687383174896, "global_step": 223195, "epoch": 2507, "lr": 7.623763125173028e-06}
{"train_loss": 0.04664023965597153, "global_step": 223196, "epoch": 2507, "lr": 7.623455429402143e-06}
{"train_loss": 0.051433585584163666, "global_step": 223197, "epoch": 2507, "lr": 7.623147739328279e-06}
{"train_loss": 0.036068011075258255, "global_step": 223198, "epoch": 2507, "lr": 7.6228400549514525e-06}
{"train_loss": 0.05750944837927818, "global_step": 223199, "epoch": 2507, "lr": 7.622532376271735e-06}
{"train_loss": 0.015504855662584305, "global_step": 223200, "epoch": 2507, "lr": 7.622224703289138e-06}
{"train_loss": 0.05577928572893143, "global_step": 223201, "epoch": 2507, "lr": 7.621917036003723e-06}
{"train_loss": 0.1126549020409584, "global_step": 223202, "epoch": 2507, "lr": 7.6216093744155174e-06}
{"train_loss": 0.037636011838912964, "global_step": 223203, "epoch": 2507, "lr": 7.621301718524582e-06}
{"train_loss": 0.050405871123075485, "global_step": 223204, "epoch": 2507, "lr": 7.620994068330934e-06}
{"train_loss": 0.06596047431230545, "global_step": 223205, "epoch": 2507, "lr": 7.6206864238346384e-06}
{"train_loss": 0.02358001098036766, "global_step": 223206, "epoch": 2507, "lr": 7.620378785035709e-06}
{"train_loss": 0.0464010164141655, "global_step": 223207, "epoch": 2507, "lr": 7.620071151934222e-06}
{"train_loss": 0.03269580006599426, "global_step": 223208, "epoch": 2507, "lr": 7.619763524530193e-06}
{"train_loss": 0.04325225204229355, "global_step": 223209, "epoch": 2507, "lr": 7.619455902823663e-06}
{"train_loss": 0.07702074199914932, "global_step": 223210, "epoch": 2507, "lr": 7.619148286814692e-06}
{"train_loss": 0.052310599788521116, "global_step": 223211, "epoch": 2507, "lr": 7.618840676503297e-06, "val_loss": 8.879796981811523}
{"train_loss": 0.06671173125505447, "global_step": 223212, "epoch": 2508, "lr": 7.618533071889544e-06}
{"train_loss": 0.06706051528453827, "global_step": 223213, "epoch": 2508, "lr": 7.618225472973456e-06}
{"train_loss": 0.020229283720254898, "global_step": 223214, "epoch": 2508, "lr": 7.617917879755088e-06}
{"train_loss": 0.018334893509745598, "global_step": 223215, "epoch": 2508, "lr": 7.6176102922344615e-06}
{"train_loss": 0.06382491439580917, "global_step": 223216, "epoch": 2508, "lr": 7.617302710411644e-06}
{"train_loss": 0.036841198801994324, "global_step": 223217, "epoch": 2508, "lr": 7.616995134286653e-06}
{"train_loss": 0.07733574509620667, "global_step": 223218, "epoch": 2508, "lr": 7.616687563859548e-06}
{"train_loss": 0.07305272668600082, "global_step": 223219, "epoch": 2508, "lr": 7.616379999130357e-06}
{"train_loss": 0.026806149631738663, "global_step": 223220, "epoch": 2508, "lr": 7.6160724400991365e-06}
{"train_loss": 0.03996105492115021, "global_step": 223221, "epoch": 2508, "lr": 7.615764886765902e-06}
{"train_loss": 0.09663710743188858, "global_step": 223222, "epoch": 2508, "lr": 7.615457339130727e-06}
{"train_loss": 0.019785190001130104, "global_step": 223223, "epoch": 2508, "lr": 7.615149797193627e-06}
{"train_loss": 0.04938935115933418, "global_step": 223224, "epoch": 2508, "lr": 7.614842260954664e-06}
{"train_loss": 0.04575970396399498, "global_step": 223225, "epoch": 2508, "lr": 7.614534730413858e-06}
{"train_loss": 0.029502607882022858, "global_step": 223226, "epoch": 2508, "lr": 7.614227205571261e-06}
{"train_loss": 0.09628886729478836, "global_step": 223227, "epoch": 2508, "lr": 7.613919686426924e-06}
{"train_loss": 0.03176988288760185, "global_step": 223228, "epoch": 2508, "lr": 7.613612172980867e-06}
{"train_loss": 0.09791014343500137, "global_step": 223229, "epoch": 2508, "lr": 7.613304665233157e-06}
{"train_loss": 0.057880789041519165, "global_step": 223230, "epoch": 2508, "lr": 7.61299716318381e-06}
{"train_loss": 0.0717603787779808, "global_step": 223231, "epoch": 2508, "lr": 7.61268966683289e-06}
{"train_loss": 0.058218128979206085, "global_step": 223232, "epoch": 2508, "lr": 7.612382176180416e-06}
{"train_loss": 0.02931751310825348, "global_step": 223233, "epoch": 2508, "lr": 7.612074691226451e-06}
{"train_loss": 0.03184547275304794, "global_step": 223234, "epoch": 2508, "lr": 7.611767211971016e-06}
{"train_loss": 0.09179474413394928, "global_step": 223235, "epoch": 2508, "lr": 7.611459738414173e-06}
{"train_loss": 0.06409582495689392, "global_step": 223236, "epoch": 2508, "lr": 7.611152270555943e-06}
{"train_loss": 0.06067457050085068, "global_step": 223237, "epoch": 2508, "lr": 7.610844808396383e-06}
{"train_loss": 0.039125192910432816, "global_step": 223238, "epoch": 2508, "lr": 7.610537351935521e-06}
{"train_loss": 0.04339710250496864, "global_step": 223239, "epoch": 2508, "lr": 7.610229901173416e-06}
{"train_loss": 0.055851783603429794, "global_step": 223240, "epoch": 2508, "lr": 7.609922456110091e-06}
{"train_loss": 0.05829966440796852, "global_step": 223241, "epoch": 2508, "lr": 7.609615016745603e-06}
{"train_loss": 0.04956524819135666, "global_step": 223242, "epoch": 2508, "lr": 7.609307583079972e-06}
{"train_loss": 0.049436308443546295, "global_step": 223243, "epoch": 2508, "lr": 7.609000155113266e-06}
{"train_loss": 0.04704813286662102, "global_step": 223244, "epoch": 2508, "lr": 7.6086927328455074e-06}
{"train_loss": 0.019557859748601913, "global_step": 223245, "epoch": 2508, "lr": 7.608385316276745e-06}
{"train_loss": 0.08053149282932281, "global_step": 223246, "epoch": 2508, "lr": 7.608077905407024e-06}
{"train_loss": 0.030920907855033875, "global_step": 223247, "epoch": 2508, "lr": 7.607770500236372e-06}
{"train_loss": 0.04062548279762268, "global_step": 223248, "epoch": 2508, "lr": 7.6074631007648445e-06}
{"train_loss": 0.037277352064847946, "global_step": 223249, "epoch": 2508, "lr": 7.607155706992469e-06}
{"train_loss": 0.039844464510679245, "global_step": 223250, "epoch": 2508, "lr": 7.606848318919302e-06}
{"train_loss": 0.05672026053071022, "global_step": 223251, "epoch": 2508, "lr": 7.60654093654537e-06}
{"train_loss": 0.035201866179704666, "global_step": 223252, "epoch": 2508, "lr": 7.606233559870734e-06}
{"train_loss": 0.04371558129787445, "global_step": 223253, "epoch": 2508, "lr": 7.605926188895407e-06}
{"train_loss": 0.0816752165555954, "global_step": 223254, "epoch": 2508, "lr": 7.605618823619465e-06}
{"train_loss": 0.032246869057416916, "global_step": 223255, "epoch": 2508, "lr": 7.605311464042913e-06}
{"train_loss": 0.06775559484958649, "global_step": 223256, "epoch": 2508, "lr": 7.605004110165825e-06}
{"train_loss": 0.037066228687763214, "global_step": 223257, "epoch": 2508, "lr": 7.604696761988211e-06}
{"train_loss": 0.03641761466860771, "global_step": 223258, "epoch": 2508, "lr": 7.604389419510144e-06}
{"train_loss": 0.017358088865876198, "global_step": 223259, "epoch": 2508, "lr": 7.60408208273164e-06}
{"train_loss": 0.07397382706403732, "global_step": 223260, "epoch": 2508, "lr": 7.603774751652759e-06}
{"train_loss": 0.0962892398238182, "global_step": 223261, "epoch": 2508, "lr": 7.603467426273525e-06}
{"train_loss": 0.04509567469358444, "global_step": 223262, "epoch": 2508, "lr": 7.603160106593987e-06}
{"train_loss": 0.06209209933876991, "global_step": 223263, "epoch": 2508, "lr": 7.602852792614201e-06}
{"train_loss": 0.05099564790725708, "global_step": 223264, "epoch": 2508, "lr": 7.602545484334178e-06}
{"train_loss": 0.06745433062314987, "global_step": 223265, "epoch": 2508, "lr": 7.60223818175399e-06}
{"train_loss": 0.06283244490623474, "global_step": 223266, "epoch": 2508, "lr": 7.601930884873653e-06}
{"train_loss": 0.031479474157094955, "global_step": 223267, "epoch": 2508, "lr": 7.601623593693236e-06}
{"train_loss": 0.08725251257419586, "global_step": 223268, "epoch": 2508, "lr": 7.601316308212747e-06}
{"train_loss": 0.028697364032268524, "global_step": 223269, "epoch": 2508, "lr": 7.601009028432254e-06}
{"train_loss": 0.04164842516183853, "global_step": 223270, "epoch": 2508, "lr": 7.60070175435178e-06}
{"train_loss": 0.045380450785160065, "global_step": 223271, "epoch": 2508, "lr": 7.600394485971385e-06}
{"train_loss": 0.09942861646413803, "global_step": 223272, "epoch": 2508, "lr": 7.6000872232910916e-06}
{"train_loss": 0.07256291806697845, "global_step": 223273, "epoch": 2508, "lr": 7.59977996631096e-06}
{"train_loss": 0.07190318405628204, "global_step": 223274, "epoch": 2508, "lr": 7.599472715031008e-06}
{"train_loss": 0.05380045622587204, "global_step": 223275, "epoch": 2508, "lr": 7.5991654694513026e-06}
{"train_loss": 0.09400451928377151, "global_step": 223276, "epoch": 2508, "lr": 7.598858229571859e-06}
{"train_loss": 0.06247587874531746, "global_step": 223277, "epoch": 2508, "lr": 7.59855099539275e-06}
{"train_loss": 0.05240938067436218, "global_step": 223278, "epoch": 2508, "lr": 7.598243766913982e-06}
{"train_loss": 0.03119650110602379, "global_step": 223279, "epoch": 2508, "lr": 7.597936544135614e-06}
{"train_loss": 0.08005351573228836, "global_step": 223280, "epoch": 2508, "lr": 7.597629327057698e-06}
{"train_loss": 0.05535570904612541, "global_step": 223281, "epoch": 2508, "lr": 7.5973221156802665e-06}
{"train_loss": 0.07828240841627121, "global_step": 223282, "epoch": 2508, "lr": 7.597014910003342e-06}
{"train_loss": 0.030144572257995605, "global_step": 223283, "epoch": 2508, "lr": 7.5967077100270016e-06}
{"train_loss": 0.07279613614082336, "global_step": 223284, "epoch": 2508, "lr": 7.5964005157512455e-06}
{"train_loss": 0.08711747080087662, "global_step": 223285, "epoch": 2508, "lr": 7.596093327176157e-06}
{"train_loss": 0.020742453634738922, "global_step": 223286, "epoch": 2508, "lr": 7.595786144301753e-06}
{"train_loss": 0.07907550781965256, "global_step": 223287, "epoch": 2508, "lr": 7.595478967128067e-06}
{"train_loss": 0.033325184136629105, "global_step": 223288, "epoch": 2508, "lr": 7.59517179565516e-06}
{"train_loss": 0.0488334558904171, "global_step": 223289, "epoch": 2508, "lr": 7.594864629883058e-06}
{"train_loss": 0.06824406236410141, "global_step": 223290, "epoch": 2508, "lr": 7.594557469811819e-06}
{"train_loss": 0.02644682116806507, "global_step": 223291, "epoch": 2508, "lr": 7.594250315441465e-06}
{"train_loss": 0.06751934438943863, "global_step": 223292, "epoch": 2508, "lr": 7.593943166772061e-06}
{"train_loss": 0.044601920992136, "global_step": 223293, "epoch": 2508, "lr": 7.593636023803619e-06}
{"train_loss": 0.03655378893017769, "global_step": 223294, "epoch": 2508, "lr": 7.5933288865362106e-06}
{"train_loss": 0.042462460696697235, "global_step": 223295, "epoch": 2508, "lr": 7.5930217549698475e-06}
{"train_loss": 0.0313018262386322, "global_step": 223296, "epoch": 2508, "lr": 7.592714629104597e-06}
{"train_loss": 0.08576063811779022, "global_step": 223297, "epoch": 2508, "lr": 7.59240750894048e-06}
{"train_loss": 0.09371302276849747, "global_step": 223298, "epoch": 2508, "lr": 7.592100394477547e-06}
{"train_loss": 0.07833921909332275, "global_step": 223299, "epoch": 2508, "lr": 7.591793285715848e-06}
{"train_loss": 0.05448904101935665, "global_step": 223300, "epoch": 2508, "lr": 7.5914861826554105e-06, "val_loss": 8.883504867553711}
{"train_loss": 0.06850621104240417, "global_step": 223301, "epoch": 2509, "lr": 7.591179085296285e-06}
{"train_loss": 0.06802807748317719, "global_step": 223302, "epoch": 2509, "lr": 7.5908719936384985e-06}
{"train_loss": 0.04294893145561218, "global_step": 223303, "epoch": 2509, "lr": 7.590564907682119e-06}
{"train_loss": 0.03296249732375145, "global_step": 223304, "epoch": 2509, "lr": 7.590257827427155e-06}
{"train_loss": 0.047572195529937744, "global_step": 223305, "epoch": 2509, "lr": 7.589950752873676e-06}
{"train_loss": 0.029494015499949455, "global_step": 223306, "epoch": 2509, "lr": 7.5896436840217025e-06}
{"train_loss": 0.0557979978621006, "global_step": 223307, "epoch": 2509, "lr": 7.58933662087129e-06}
{"train_loss": 0.03441022336483002, "global_step": 223308, "epoch": 2509, "lr": 7.589029563422467e-06}
{"train_loss": 0.04077885299921036, "global_step": 223309, "epoch": 2509, "lr": 7.588722511675294e-06}
{"train_loss": 0.04749451205134392, "global_step": 223310, "epoch": 2509, "lr": 7.588415465629789e-06}
{"train_loss": 0.08542723953723907, "global_step": 223311, "epoch": 2509, "lr": 7.5881084252860115e-06}
{"train_loss": 0.04702078923583031, "global_step": 223312, "epoch": 2509, "lr": 7.58780139064399e-06}
{"train_loss": 0.06694497913122177, "global_step": 223313, "epoch": 2509, "lr": 7.5874943617037785e-06}
{"train_loss": 0.06800857931375504, "global_step": 223314, "epoch": 2509, "lr": 7.587187338465401e-06}
{"train_loss": 0.0738750696182251, "global_step": 223315, "epoch": 2509, "lr": 7.586880320928924e-06}
{"train_loss": 0.09056705981492996, "global_step": 223316, "epoch": 2509, "lr": 7.586573309094364e-06}
{"train_loss": 0.07250245660543442, "global_step": 223317, "epoch": 2509, "lr": 7.586266302961764e-06}
{"train_loss": 0.04202583432197571, "global_step": 223318, "epoch": 2509, "lr": 7.585959302531193e-06}
{"train_loss": 0.05242516100406647, "global_step": 223319, "epoch": 2509, "lr": 7.585652307802671e-06}
{"train_loss": 0.08488725125789642, "global_step": 223320, "epoch": 2509, "lr": 7.585345318776227e-06}
{"train_loss": 0.05759318917989731, "global_step": 223321, "epoch": 2509, "lr": 7.585038335451927e-06}
{"train_loss": 0.04538675770163536, "global_step": 223322, "epoch": 2509, "lr": 7.584731357829794e-06}
{"train_loss": 0.029690692201256752, "global_step": 223323, "epoch": 2509, "lr": 7.584424385909888e-06}
{"train_loss": 0.061156053096055984, "global_step": 223324, "epoch": 2509, "lr": 7.584117419692238e-06}
{"train_loss": 0.015403127297759056, "global_step": 223325, "epoch": 2509, "lr": 7.583810459176871e-06}
{"train_loss": 0.03542069345712662, "global_step": 223326, "epoch": 2509, "lr": 7.5835035043638594e-06}
{"train_loss": 0.044973816722631454, "global_step": 223327, "epoch": 2509, "lr": 7.583196555253219e-06}
{"train_loss": 0.049128998070955276, "global_step": 223328, "epoch": 2509, "lr": 7.582889611845006e-06}
{"train_loss": 0.02605881541967392, "global_step": 223329, "epoch": 2509, "lr": 7.582582674139254e-06}
{"train_loss": 0.06603056192398071, "global_step": 223330, "epoch": 2509, "lr": 7.582275742136014e-06}
{"train_loss": 0.028289970010519028, "global_step": 223331, "epoch": 2509, "lr": 7.581968815835305e-06}
{"train_loss": 0.030534598976373672, "global_step": 223332, "epoch": 2509, "lr": 7.5816618952371965e-06}
{"train_loss": 0.05023065209388733, "global_step": 223333, "epoch": 2509, "lr": 7.58135498034171e-06}
{"train_loss": 0.07759018242359161, "global_step": 223334, "epoch": 2509, "lr": 7.581048071148888e-06}
{"train_loss": 0.09842056035995483, "global_step": 223335, "epoch": 2509, "lr": 7.580741167658789e-06}
{"train_loss": 0.06466969847679138, "global_step": 223336, "epoch": 2509, "lr": 7.580434269871433e-06}
{"train_loss": 0.03430996090173721, "global_step": 223337, "epoch": 2509, "lr": 7.580127377786883e-06}
{"train_loss": 0.053848184645175934, "global_step": 223338, "epoch": 2509, "lr": 7.579820491405154e-06}
{"train_loss": 0.020029813051223755, "global_step": 223339, "epoch": 2509, "lr": 7.579513610726313e-06}
{"train_loss": 0.036844201385974884, "global_step": 223340, "epoch": 2509, "lr": 7.579206735750377e-06}
{"train_loss": 0.02828005701303482, "global_step": 223341, "epoch": 2509, "lr": 7.578899866477412e-06}
{"train_loss": 0.06253709644079208, "global_step": 223342, "epoch": 2509, "lr": 7.578593002907436e-06}
{"train_loss": 0.06860974431037903, "global_step": 223343, "epoch": 2509, "lr": 7.5782861450405085e-06}
{"train_loss": 0.04067835211753845, "global_step": 223344, "epoch": 2509, "lr": 7.577979292876658e-06}
{"train_loss": 0.05543900653719902, "global_step": 223345, "epoch": 2509, "lr": 7.57767244641594e-06}
{"train_loss": 0.05527009814977646, "global_step": 223346, "epoch": 2509, "lr": 7.577365605658376e-06}
{"train_loss": 0.06814607232809067, "global_step": 223347, "epoch": 2509, "lr": 7.5770587706040285e-06}
{"train_loss": 0.019030895084142685, "global_step": 223348, "epoch": 2509, "lr": 7.576751941252919e-06}
{"train_loss": 0.023644274100661278, "global_step": 223349, "epoch": 2509, "lr": 7.576445117605108e-06}
{"train_loss": 0.0349690206348896, "global_step": 223350, "epoch": 2509, "lr": 7.576138299660612e-06}
{"train_loss": 0.058823361992836, "global_step": 223351, "epoch": 2509, "lr": 7.575831487419505e-06}
{"train_loss": 0.05430544540286064, "global_step": 223352, "epoch": 2509, "lr": 7.5755246808817955e-06}
{"train_loss": 0.025370361283421516, "global_step": 223353, "epoch": 2509, "lr": 7.575217880047542e-06}
{"train_loss": 0.08211702108383179, "global_step": 223354, "epoch": 2509, "lr": 7.574911084916791e-06}
{"train_loss": 0.08234849572181702, "global_step": 223355, "epoch": 2509, "lr": 7.5746042954895676e-06}
{"train_loss": 0.028610721230506897, "global_step": 223356, "epoch": 2509, "lr": 7.574297511765932e-06}
{"train_loss": 0.022439423948526382, "global_step": 223357, "epoch": 2509, "lr": 7.573990733745912e-06}
{"train_loss": 0.03363044559955597, "global_step": 223358, "epoch": 2509, "lr": 7.573683961429545e-06}
{"train_loss": 0.07216935604810715, "global_step": 223359, "epoch": 2509, "lr": 7.573377194816888e-06}
{"train_loss": 0.041118670254945755, "global_step": 223360, "epoch": 2509, "lr": 7.573070433907964e-06}
{"train_loss": 0.050418954342603683, "global_step": 223361, "epoch": 2509, "lr": 7.572763678702832e-06}
{"train_loss": 0.11707475781440735, "global_step": 223362, "epoch": 2509, "lr": 7.572456929201527e-06}
{"train_loss": 0.020231477916240692, "global_step": 223363, "epoch": 2509, "lr": 7.5721501854040755e-06}
{"train_loss": 0.03242942690849304, "global_step": 223364, "epoch": 2509, "lr": 7.571843447310545e-06}
{"train_loss": 0.061160001903772354, "global_step": 223365, "epoch": 2509, "lr": 7.571536714920951e-06}
{"train_loss": 0.05356154590845108, "global_step": 223366, "epoch": 2509, "lr": 7.571229988235357e-06}
{"train_loss": 0.059162821620702744, "global_step": 223367, "epoch": 2509, "lr": 7.5709232672537825e-06}
{"train_loss": 0.05410625785589218, "global_step": 223368, "epoch": 2509, "lr": 7.570616551976295e-06}
{"train_loss": 0.053511738777160645, "global_step": 223369, "epoch": 2509, "lr": 7.5703098424029066e-06}
{"train_loss": 0.07508660107851028, "global_step": 223370, "epoch": 2509, "lr": 7.570003138533671e-06}
{"train_loss": 0.042069826275110245, "global_step": 223371, "epoch": 2509, "lr": 7.569696440368645e-06}
{"train_loss": 0.042694151401519775, "global_step": 223372, "epoch": 2509, "lr": 7.569389747907846e-06}
{"train_loss": 0.049233969300985336, "global_step": 223373, "epoch": 2509, "lr": 7.569083061151333e-06}
{"train_loss": 0.0699021965265274, "global_step": 223374, "epoch": 2509, "lr": 7.56877638009913e-06}
{"train_loss": 0.05523167923092842, "global_step": 223375, "epoch": 2509, "lr": 7.568469704751302e-06}
{"train_loss": 0.03756619244813919, "global_step": 223376, "epoch": 2509, "lr": 7.568163035107862e-06}
{"train_loss": 0.04424126818776131, "global_step": 223377, "epoch": 2509, "lr": 7.567856371168874e-06}
{"train_loss": 0.07034709304571152, "global_step": 223378, "epoch": 2509, "lr": 7.567549712934363e-06}
{"train_loss": 0.091111920773983, "global_step": 223379, "epoch": 2509, "lr": 7.567243060404389e-06}
{"train_loss": 0.05856591835618019, "global_step": 223380, "epoch": 2509, "lr": 7.5669364135789675e-06}
{"train_loss": 0.06535179913043976, "global_step": 223381, "epoch": 2509, "lr": 7.566629772458167e-06}
{"train_loss": 0.009986555203795433, "global_step": 223382, "epoch": 2509, "lr": 7.566323137042009e-06}
{"train_loss": 0.013304168358445168, "global_step": 223383, "epoch": 2509, "lr": 7.566016507330548e-06}
{"train_loss": 0.10097355395555496, "global_step": 223384, "epoch": 2509, "lr": 7.565709883323807e-06}
{"train_loss": 0.041956037282943726, "global_step": 223385, "epoch": 2509, "lr": 7.565403265021853e-06}
{"train_loss": 0.0457606166601181, "global_step": 223386, "epoch": 2509, "lr": 7.565096652424702e-06}
{"train_loss": 0.03583260998129845, "global_step": 223387, "epoch": 2509, "lr": 7.564790045532421e-06}
{"train_loss": 0.04908403754234314, "global_step": 223388, "epoch": 2509, "lr": 7.564483444345022e-06}
{"train_loss": 0.05192771137514141, "global_step": 223389, "epoch": 2509, "lr": 7.5641768488625584e-06, "val_loss": 8.929343223571777}
{"train_loss": 0.055710140615701675, "global_step": 223390, "epoch": 2510, "lr": 7.563870259085087e-06}
{"train_loss": 0.0475858710706234, "global_step": 223391, "epoch": 2510, "lr": 7.5635636750126245e-06}
{"train_loss": 0.13483332097530365, "global_step": 223392, "epoch": 2510, "lr": 7.563257096645238e-06}
{"train_loss": 0.04006124660372734, "global_step": 223393, "epoch": 2510, "lr": 7.562950523982942e-06}
{"train_loss": 0.11406104266643524, "global_step": 223394, "epoch": 2510, "lr": 7.5626439570258e-06}
{"train_loss": 0.10383157432079315, "global_step": 223395, "epoch": 2510, "lr": 7.562337395773833e-06}
{"train_loss": 0.04248281568288803, "global_step": 223396, "epoch": 2510, "lr": 7.562030840227102e-06}
{"train_loss": 0.07204843312501907, "global_step": 223397, "epoch": 2510, "lr": 7.561724290385641e-06}
{"train_loss": 0.04404469579458237, "global_step": 223398, "epoch": 2510, "lr": 7.5614177462494715e-06}
{"train_loss": 0.024341609328985214, "global_step": 223399, "epoch": 2510, "lr": 7.561111207818672e-06}
{"train_loss": 0.07424595206975937, "global_step": 223400, "epoch": 2510, "lr": 7.560804675093247e-06}
{"train_loss": 0.05642686039209366, "global_step": 223401, "epoch": 2510, "lr": 7.5604981480732695e-06}
{"train_loss": 0.05581482872366905, "global_step": 223402, "epoch": 2510, "lr": 7.560191626758761e-06}
{"train_loss": 0.03625604510307312, "global_step": 223403, "epoch": 2510, "lr": 7.5598851111497615e-06}
{"train_loss": 0.02206132374703884, "global_step": 223404, "epoch": 2510, "lr": 7.559578601246326e-06}
{"train_loss": 0.02804393321275711, "global_step": 223405, "epoch": 2510, "lr": 7.559272097048481e-06}
{"train_loss": 0.06817689538002014, "global_step": 223406, "epoch": 2510, "lr": 7.558965598556267e-06}
{"train_loss": 0.0528532974421978, "global_step": 223407, "epoch": 2510, "lr": 7.55865910576975e-06}
{"train_loss": 0.06321542710065842, "global_step": 223408, "epoch": 2510, "lr": 7.558352618688941e-06}
{"train_loss": 0.0259934701025486, "global_step": 223409, "epoch": 2510, "lr": 7.5580461373139065e-06}
{"train_loss": 0.027343910187482834, "global_step": 223410, "epoch": 2510, "lr": 7.5577396616446635e-06}
{"train_loss": 0.039136916399002075, "global_step": 223411, "epoch": 2510, "lr": 7.557433191681273e-06}
{"train_loss": 0.06965367496013641, "global_step": 223412, "epoch": 2510, "lr": 7.557126727423763e-06}
{"train_loss": 0.03273564949631691, "global_step": 223413, "epoch": 2510, "lr": 7.556820268872189e-06}
{"train_loss": 0.05334031209349632, "global_step": 223414, "epoch": 2510, "lr": 7.556513816026573e-06}
{"train_loss": 0.02586684189736843, "global_step": 223415, "epoch": 2510, "lr": 7.556207368886975e-06}
{"train_loss": 0.05906670540571213, "global_step": 223416, "epoch": 2510, "lr": 7.555900927453413e-06}
{"train_loss": 0.03997885808348656, "global_step": 223417, "epoch": 2510, "lr": 7.5555944917259595e-06}
{"train_loss": 0.03077000193297863, "global_step": 223418, "epoch": 2510, "lr": 7.55528806170463e-06}
{"train_loss": 0.06283001601696014, "global_step": 223419, "epoch": 2510, "lr": 7.55498163738948e-06}
{"train_loss": 0.05068989470601082, "global_step": 223420, "epoch": 2510, "lr": 7.554675218780538e-06}
{"train_loss": 0.04039236530661583, "global_step": 223421, "epoch": 2510, "lr": 7.554368805877865e-06}
{"train_loss": 0.051938496530056, "global_step": 223422, "epoch": 2510, "lr": 7.5540623986814715e-06}
{"train_loss": 0.03868372365832329, "global_step": 223423, "epoch": 2510, "lr": 7.553755997191436e-06}
{"train_loss": 0.04777505621314049, "global_step": 223424, "epoch": 2510, "lr": 7.5534496014077636e-06}
{"train_loss": 0.05278026685118675, "global_step": 223425, "epoch": 2510, "lr": 7.5531432113305155e-06}
{"train_loss": 0.033148396760225296, "global_step": 223426, "epoch": 2510, "lr": 7.552836826959742e-06}
{"train_loss": 0.05166275426745415, "global_step": 223427, "epoch": 2510, "lr": 7.552530448295458e-06}
{"train_loss": 0.04361153021454811, "global_step": 223428, "epoch": 2510, "lr": 7.552224075337733e-06}
{"train_loss": 0.06389676779508591, "global_step": 223429, "epoch": 2510, "lr": 7.551917708086581e-06}
{"train_loss": 0.034857120364904404, "global_step": 223430, "epoch": 2510, "lr": 7.55161134654207e-06}
{"train_loss": 0.07612074166536331, "global_step": 223431, "epoch": 2510, "lr": 7.551304990704217e-06}
{"train_loss": 0.05480509623885155, "global_step": 223432, "epoch": 2510, "lr": 7.550998640573082e-06}
{"train_loss": 0.06880562007427216, "global_step": 223433, "epoch": 2510, "lr": 7.550692296148687e-06}
{"train_loss": 0.09717438369989395, "global_step": 223434, "epoch": 2510, "lr": 7.550385957431094e-06}
{"train_loss": 0.0400933213531971, "global_step": 223435, "epoch": 2510, "lr": 7.550079624420336e-06}
{"train_loss": 0.04665002226829529, "global_step": 223436, "epoch": 2510, "lr": 7.549773297116442e-06}
{"train_loss": 0.06910388171672821, "global_step": 223437, "epoch": 2510, "lr": 7.5494669755194705e-06}
{"train_loss": 0.10292621701955795, "global_step": 223438, "epoch": 2510, "lr": 7.549160659629451e-06}
{"train_loss": 0.038198746740818024, "global_step": 223439, "epoch": 2510, "lr": 7.548854349446432e-06}
{"train_loss": 0.03808216750621796, "global_step": 223440, "epoch": 2510, "lr": 7.54854804497046e-06}
{"train_loss": 0.027846679091453552, "global_step": 223441, "epoch": 2510, "lr": 7.548241746201556e-06}
{"train_loss": 0.05435510352253914, "global_step": 223442, "epoch": 2510, "lr": 7.5479354531397816e-06}
{"train_loss": 0.04839189350605011, "global_step": 223443, "epoch": 2510, "lr": 7.547629165785158e-06}
{"train_loss": 0.04789484664797783, "global_step": 223444, "epoch": 2510, "lr": 7.547322884137742e-06}
{"train_loss": 0.02752266824245453, "global_step": 223445, "epoch": 2510, "lr": 7.547016608197582e-06}
{"train_loss": 0.10893769562244415, "global_step": 223446, "epoch": 2510, "lr": 7.546710337964697e-06}
{"train_loss": 0.05407879874110222, "global_step": 223447, "epoch": 2510, "lr": 7.54640407343915e-06}
{"train_loss": 0.027980832383036613, "global_step": 223448, "epoch": 2510, "lr": 7.546097814620956e-06}
{"train_loss": 0.04502885043621063, "global_step": 223449, "epoch": 2510, "lr": 7.54579156151019e-06}
{"train_loss": 0.053235720843076706, "global_step": 223450, "epoch": 2510, "lr": 7.545485314106859e-06}
{"train_loss": 0.07071532309055328, "global_step": 223451, "epoch": 2510, "lr": 7.545179072411029e-06}
{"train_loss": 0.05055951327085495, "global_step": 223452, "epoch": 2510, "lr": 7.544872836422728e-06}
{"train_loss": 0.08200521022081375, "global_step": 223453, "epoch": 2510, "lr": 7.544566606142006e-06}
{"train_loss": 0.028723370283842087, "global_step": 223454, "epoch": 2510, "lr": 7.544260381568896e-06}
{"train_loss": 0.0698409304022789, "global_step": 223455, "epoch": 2510, "lr": 7.543954162703449e-06}
{"train_loss": 0.06708956509828568, "global_step": 223456, "epoch": 2510, "lr": 7.543647949545684e-06}
{"train_loss": 0.10797513276338577, "global_step": 223457, "epoch": 2510, "lr": 7.543341742095678e-06}
{"train_loss": 0.07592033594846725, "global_step": 223458, "epoch": 2510, "lr": 7.543035540353438e-06}
{"train_loss": 0.04745199903845787, "global_step": 223459, "epoch": 2510, "lr": 7.542729344319027e-06}
{"train_loss": 0.08328309655189514, "global_step": 223460, "epoch": 2510, "lr": 7.542423153992473e-06}
{"train_loss": 0.042964834719896317, "global_step": 223461, "epoch": 2510, "lr": 7.5421169693738196e-06}
{"train_loss": 0.04518600180745125, "global_step": 223462, "epoch": 2510, "lr": 7.541810790463122e-06}
{"train_loss": 0.06617889553308487, "global_step": 223463, "epoch": 2510, "lr": 7.541504617260397e-06}
{"train_loss": 0.06726700067520142, "global_step": 223464, "epoch": 2510, "lr": 7.541198449765718e-06}
{"train_loss": 0.05281071737408638, "global_step": 223465, "epoch": 2510, "lr": 7.54089228797909e-06}
{"train_loss": 0.05097381770610809, "global_step": 223466, "epoch": 2510, "lr": 7.540586131900584e-06}
{"train_loss": 0.02554802969098091, "global_step": 223467, "epoch": 2510, "lr": 7.540279981530219e-06}
{"train_loss": 0.053874991834163666, "global_step": 223468, "epoch": 2510, "lr": 7.539973836868053e-06}
{"train_loss": 0.03273864835500717, "global_step": 223469, "epoch": 2510, "lr": 7.5396676979141154e-06}
{"train_loss": 0.09726902842521667, "global_step": 223470, "epoch": 2510, "lr": 7.5393615646684565e-06}
{"train_loss": 0.03436487168073654, "global_step": 223471, "epoch": 2510, "lr": 7.539055437131104e-06}
{"train_loss": 0.039550259709358215, "global_step": 223472, "epoch": 2510, "lr": 7.538749315302124e-06}
{"train_loss": 0.05499362573027611, "global_step": 223473, "epoch": 2510, "lr": 7.538443199181533e-06}
{"train_loss": 0.04566964879631996, "global_step": 223474, "epoch": 2510, "lr": 7.538137088769376e-06}
{"train_loss": 0.05260986089706421, "global_step": 223475, "epoch": 2510, "lr": 7.537830984065708e-06}
{"train_loss": 0.051832664757966995, "global_step": 223476, "epoch": 2510, "lr": 7.537524885070552e-06}
{"train_loss": 0.07778295874595642, "global_step": 223477, "epoch": 2510, "lr": 7.537218791783968e-06}
{"train_loss": 0.05531984195113182, "global_step": 223478, "epoch": 2510, "lr": 7.536912704205984e-06, "val_loss": 9.141097068786621, "train_action_mse_error": 16.385658264160156}
{"train_loss": 0.07879175245761871, "global_step": 223479, "epoch": 2511, "lr": 7.536606622336634e-06}
{"train_loss": 0.07175827026367188, "global_step": 223480, "epoch": 2511, "lr": 7.536300546175973e-06}
{"train_loss": 0.07495298981666565, "global_step": 223481, "epoch": 2511, "lr": 7.5359944757240504e-06}
{"train_loss": 0.0483078807592392, "global_step": 223482, "epoch": 2511, "lr": 7.535688410980884e-06}
{"train_loss": 0.04071873053908348, "global_step": 223483, "epoch": 2511, "lr": 7.535382351946535e-06}
{"train_loss": 0.0712766945362091, "global_step": 223484, "epoch": 2511, "lr": 7.5350762986210245e-06}
{"train_loss": 0.04095863923430443, "global_step": 223485, "epoch": 2511, "lr": 7.534770251004419e-06}
{"train_loss": 0.05564718693494797, "global_step": 223486, "epoch": 2511, "lr": 7.534464209096737e-06}
{"train_loss": 0.0689062550663948, "global_step": 223487, "epoch": 2511, "lr": 7.534158172898037e-06}
{"train_loss": 0.024641353636980057, "global_step": 223488, "epoch": 2511, "lr": 7.533852142408343e-06}
{"train_loss": 0.07244650274515152, "global_step": 223489, "epoch": 2511, "lr": 7.533546117627715e-06}
{"train_loss": 0.07522144168615341, "global_step": 223490, "epoch": 2511, "lr": 7.533240098556172e-06}
{"train_loss": 0.06550946831703186, "global_step": 223491, "epoch": 2511, "lr": 7.532934085193783e-06}
{"train_loss": 0.058972131460905075, "global_step": 223492, "epoch": 2511, "lr": 7.532628077540554e-06}
{"train_loss": 0.04519103094935417, "global_step": 223493, "epoch": 2511, "lr": 7.53232207559656e-06}
{"train_loss": 0.07355239987373352, "global_step": 223494, "epoch": 2511, "lr": 7.53201607936182e-06}
{"train_loss": 0.02518080361187458, "global_step": 223495, "epoch": 2511, "lr": 7.531710088836391e-06}
{"train_loss": 0.03781689703464508, "global_step": 223496, "epoch": 2511, "lr": 7.531404104020295e-06}
{"train_loss": 0.05405484884977341, "global_step": 223497, "epoch": 2511, "lr": 7.531098124913583e-06}
{"train_loss": 0.0771605595946312, "global_step": 223498, "epoch": 2511, "lr": 7.530792151516314e-06}
{"train_loss": 0.0693967342376709, "global_step": 223499, "epoch": 2511, "lr": 7.5304861838284945e-06}
{"train_loss": 0.07790390402078629, "global_step": 223500, "epoch": 2511, "lr": 7.530180221850197e-06}
{"train_loss": 0.07081384211778641, "global_step": 223501, "epoch": 2511, "lr": 7.529874265581438e-06}
{"train_loss": 0.03664172440767288, "global_step": 223502, "epoch": 2511, "lr": 7.529568315022284e-06}
{"train_loss": 0.07717057317495346, "global_step": 223503, "epoch": 2511, "lr": 7.529262370172746e-06}
{"train_loss": 0.062366094440221786, "global_step": 223504, "epoch": 2511, "lr": 7.5289564310328965e-06}
{"train_loss": 0.04267829656600952, "global_step": 223505, "epoch": 2511, "lr": 7.528650497602746e-06}
{"train_loss": 0.020924698561429977, "global_step": 223506, "epoch": 2511, "lr": 7.528344569882362e-06}
{"train_loss": 0.022729098796844482, "global_step": 223507, "epoch": 2511, "lr": 7.528038647871765e-06}
{"train_loss": 0.016245929524302483, "global_step": 223508, "epoch": 2511, "lr": 7.527732731571019e-06}
{"train_loss": 0.040669288486242294, "global_step": 223509, "epoch": 2511, "lr": 7.527426820980138e-06}
{"train_loss": 0.03333141654729843, "global_step": 223510, "epoch": 2511, "lr": 7.527120916099184e-06}
{"train_loss": 0.047649458050727844, "global_step": 223511, "epoch": 2511, "lr": 7.5268150169281846e-06}
{"train_loss": 0.04532264918088913, "global_step": 223512, "epoch": 2511, "lr": 7.5265091234671954e-06}
{"train_loss": 0.05769161880016327, "global_step": 223513, "epoch": 2511, "lr": 7.526203235716256e-06}
{"train_loss": 0.07461751997470856, "global_step": 223514, "epoch": 2511, "lr": 7.525897353675382e-06}
{"train_loss": 0.06745260208845139, "global_step": 223515, "epoch": 2511, "lr": 7.525591477344651e-06}
{"train_loss": 0.028508882969617844, "global_step": 223516, "epoch": 2511, "lr": 7.5252856067240694e-06}
{"train_loss": 0.05025207996368408, "global_step": 223517, "epoch": 2511, "lr": 7.524979741813709e-06}
{"train_loss": 0.036989402025938034, "global_step": 223518, "epoch": 2511, "lr": 7.5246738826135865e-06}
{"train_loss": 0.02283567003905773, "global_step": 223519, "epoch": 2511, "lr": 7.524368029123769e-06}
{"train_loss": 0.03569207713007927, "global_step": 223520, "epoch": 2511, "lr": 7.524062181344266e-06}
{"train_loss": 0.037136249244213104, "global_step": 223521, "epoch": 2511, "lr": 7.523756339275145e-06}
{"train_loss": 0.04201327636837959, "global_step": 223522, "epoch": 2511, "lr": 7.523450502916429e-06}
{"train_loss": 0.04041179269552231, "global_step": 223523, "epoch": 2511, "lr": 7.5231446722681785e-06}
{"train_loss": 0.04810424521565437, "global_step": 223524, "epoch": 2511, "lr": 7.522838847330416e-06}
{"train_loss": 0.03698963299393654, "global_step": 223525, "epoch": 2511, "lr": 7.522533028103196e-06}
{"train_loss": 0.035758599638938904, "global_step": 223526, "epoch": 2511, "lr": 7.522227214586541e-06}
{"train_loss": 0.08276069164276123, "global_step": 223527, "epoch": 2511, "lr": 7.521921406780525e-06}
{"train_loss": 0.05673356354236603, "global_step": 223528, "epoch": 2511, "lr": 7.52161560468515e-06}
{"train_loss": 0.07877060025930405, "global_step": 223529, "epoch": 2511, "lr": 7.521309808300486e-06}
{"train_loss": 0.03308951482176781, "global_step": 223530, "epoch": 2511, "lr": 7.52100401762656e-06}
{"train_loss": 0.07784091681241989, "global_step": 223531, "epoch": 2511, "lr": 7.52069823266342e-06}
{"train_loss": 0.057899244129657745, "global_step": 223532, "epoch": 2511, "lr": 7.520392453411101e-06}
{"train_loss": 0.038583021610975266, "global_step": 223533, "epoch": 2511, "lr": 7.520086679869643e-06}
{"train_loss": 0.0472291074693203, "global_step": 223534, "epoch": 2511, "lr": 7.519780912039104e-06}
{"train_loss": 0.08218280971050262, "global_step": 223535, "epoch": 2511, "lr": 7.519475149919503e-06}
{"train_loss": 0.05609036237001419, "global_step": 223536, "epoch": 2511, "lr": 7.5191693935109e-06}
{"train_loss": 0.055901896208524704, "global_step": 223537, "epoch": 2511, "lr": 7.518863642813312e-06}
{"train_loss": 0.04761610925197601, "global_step": 223538, "epoch": 2511, "lr": 7.518557897826811e-06}
{"train_loss": 0.021181583404541016, "global_step": 223539, "epoch": 2511, "lr": 7.518252158551409e-06}
{"train_loss": 0.03767715021967888, "global_step": 223540, "epoch": 2511, "lr": 7.5179464249871766e-06}
{"train_loss": 0.02968115545809269, "global_step": 223541, "epoch": 2511, "lr": 7.5176406971341204e-06}
{"train_loss": 0.044948186725378036, "global_step": 223542, "epoch": 2511, "lr": 7.517334974992313e-06}
{"train_loss": 0.042148977518081665, "global_step": 223543, "epoch": 2511, "lr": 7.517029258561775e-06}
{"train_loss": 0.06853093206882477, "global_step": 223544, "epoch": 2511, "lr": 7.5167235478425635e-06}
{"train_loss": 0.03700696676969528, "global_step": 223545, "epoch": 2511, "lr": 7.516417842834694e-06}
{"train_loss": 0.07361849397420883, "global_step": 223546, "epoch": 2511, "lr": 7.51611214353824e-06}
{"train_loss": 0.0558059960603714, "global_step": 223547, "epoch": 2511, "lr": 7.515806449953216e-06}
{"train_loss": 0.07888469845056534, "global_step": 223548, "epoch": 2511, "lr": 7.515500762079686e-06}
{"train_loss": 0.056715454906225204, "global_step": 223549, "epoch": 2511, "lr": 7.51519507991767e-06}
{"train_loss": 0.0717606469988823, "global_step": 223550, "epoch": 2511, "lr": 7.514889403467224e-06}
{"train_loss": 0.040861666202545166, "global_step": 223551, "epoch": 2511, "lr": 7.514583732728386e-06}
{"train_loss": 0.03815652057528496, "global_step": 223552, "epoch": 2511, "lr": 7.51427806770118e-06}
{"train_loss": 0.049410849809646606, "global_step": 223553, "epoch": 2511, "lr": 7.513972408385678e-06}
{"train_loss": 0.02547697350382805, "global_step": 223554, "epoch": 2511, "lr": 7.513666754781889e-06}
{"train_loss": 0.0427640900015831, "global_step": 223555, "epoch": 2511, "lr": 7.513361106889883e-06}
{"train_loss": 0.05456871911883354, "global_step": 223556, "epoch": 2511, "lr": 7.5130554647096784e-06}
{"train_loss": 0.06685115396976471, "global_step": 223557, "epoch": 2511, "lr": 7.512749828241333e-06}
{"train_loss": 0.054833974689245224, "global_step": 223558, "epoch": 2511, "lr": 7.512444197484869e-06}
{"train_loss": 0.06901528686285019, "global_step": 223559, "epoch": 2511, "lr": 7.512138572440352e-06}
{"train_loss": 0.05276590958237648, "global_step": 223560, "epoch": 2511, "lr": 7.5118329531078e-06}
{"train_loss": 0.0731758326292038, "global_step": 223561, "epoch": 2511, "lr": 7.511527339487279e-06}
{"train_loss": 0.06861336529254913, "global_step": 223562, "epoch": 2511, "lr": 7.511221731578799e-06}
{"train_loss": 0.06489133089780807, "global_step": 223563, "epoch": 2511, "lr": 7.5109161293824285e-06}
{"train_loss": 0.10822375118732452, "global_step": 223564, "epoch": 2511, "lr": 7.510610532898188e-06}
{"train_loss": 0.06566813588142395, "global_step": 223565, "epoch": 2511, "lr": 7.510304942126134e-06}
{"train_loss": 0.10309000313282013, "global_step": 223566, "epoch": 2511, "lr": 7.509999357066294e-06}
{"train_loss": 0.05353562539087588, "global_step": 223567, "epoch": 2511, "lr": 7.5096937777187295e-06, "val_loss": 9.03429889678955}
{"train_loss": 0.0538179837167263, "global_step": 223568, "epoch": 2512, "lr": 7.509388204083456e-06}
{"train_loss": 0.03862697631120682, "global_step": 223569, "epoch": 2512, "lr": 7.5090826361605415e-06}
{"train_loss": 0.1023363545536995, "global_step": 223570, "epoch": 2512, "lr": 7.508777073950002e-06}
{"train_loss": 0.03527367487549782, "global_step": 223571, "epoch": 2512, "lr": 7.508471517451882e-06}
{"train_loss": 0.045670285820961, "global_step": 223572, "epoch": 2512, "lr": 7.508165966666247e-06}
{"train_loss": 0.04670935869216919, "global_step": 223573, "epoch": 2512, "lr": 7.50786042159311e-06}
{"train_loss": 0.019835317507386208, "global_step": 223574, "epoch": 2512, "lr": 7.507554882232537e-06}
{"train_loss": 0.03315797448158264, "global_step": 223575, "epoch": 2512, "lr": 7.507249348584539e-06}
{"train_loss": 0.04909587651491165, "global_step": 223576, "epoch": 2512, "lr": 7.5069438206491874e-06}
{"train_loss": 0.043650638312101364, "global_step": 223577, "epoch": 2512, "lr": 7.506638298426494e-06}
{"train_loss": 0.04007672891020775, "global_step": 223578, "epoch": 2512, "lr": 7.506332781916531e-06}
{"train_loss": 0.0488317534327507, "global_step": 223579, "epoch": 2512, "lr": 7.506027271119309e-06}
{"train_loss": 0.08254330605268478, "global_step": 223580, "epoch": 2512, "lr": 7.505721766034901e-06}
{"train_loss": 0.08105714619159698, "global_step": 223581, "epoch": 2512, "lr": 7.505416266663312e-06}
{"train_loss": 0.0974675714969635, "global_step": 223582, "epoch": 2512, "lr": 7.50511077300462e-06}
{"train_loss": 0.06310788542032242, "global_step": 223583, "epoch": 2512, "lr": 7.50480528505883e-06}
{"train_loss": 0.05136922374367714, "global_step": 223584, "epoch": 2512, "lr": 7.504499802826015e-06}
{"train_loss": 0.03203419968485832, "global_step": 223585, "epoch": 2512, "lr": 7.504194326306191e-06}
{"train_loss": 0.023920563980937004, "global_step": 223586, "epoch": 2512, "lr": 7.503888855499424e-06}
{"train_loss": 0.03909438103437424, "global_step": 223587, "epoch": 2512, "lr": 7.503583390405728e-06}
{"train_loss": 0.05804676190018654, "global_step": 223588, "epoch": 2512, "lr": 7.503277931025166e-06}
{"train_loss": 0.06481078267097473, "global_step": 223589, "epoch": 2512, "lr": 7.502972477357772e-06}
{"train_loss": 0.04351197928190231, "global_step": 223590, "epoch": 2512, "lr": 7.502667029403576e-06}
{"train_loss": 0.020997820422053337, "global_step": 223591, "epoch": 2512, "lr": 7.502361587162637e-06}
{"train_loss": 0.06457779556512833, "global_step": 223592, "epoch": 2512, "lr": 7.502056150634978e-06}
{"train_loss": 0.04861234128475189, "global_step": 223593, "epoch": 2512, "lr": 7.50175071982066e-06}
{"train_loss": 0.07253701984882355, "global_step": 223594, "epoch": 2512, "lr": 7.501445294719705e-06}
{"train_loss": 0.05548526719212532, "global_step": 223595, "epoch": 2512, "lr": 7.501139875332169e-06}
{"train_loss": 0.04574281722307205, "global_step": 223596, "epoch": 2512, "lr": 7.500834461658079e-06}
{"train_loss": 0.0548967570066452, "global_step": 223597, "epoch": 2512, "lr": 7.500529053697497e-06}
{"train_loss": 0.05778682604432106, "global_step": 223598, "epoch": 2512, "lr": 7.500223651450433e-06}
{"train_loss": 0.051480863243341446, "global_step": 223599, "epoch": 2512, "lr": 7.49991825491696e-06}
{"train_loss": 0.038555730134248734, "global_step": 223600, "epoch": 2512, "lr": 7.499612864097094e-06}
{"train_loss": 0.05167311802506447, "global_step": 223601, "epoch": 2512, "lr": 7.499307478990902e-06}
{"train_loss": 0.02252284623682499, "global_step": 223602, "epoch": 2512, "lr": 7.4990020995983955e-06}
{"train_loss": 0.04259943589568138, "global_step": 223603, "epoch": 2512, "lr": 7.49869672591964e-06}
{"train_loss": 0.054035261273384094, "global_step": 223604, "epoch": 2512, "lr": 7.4983913579546594e-06}
{"train_loss": 0.026233961805701256, "global_step": 223605, "epoch": 2512, "lr": 7.498085995703513e-06}
{"train_loss": 0.034481119364500046, "global_step": 223606, "epoch": 2512, "lr": 7.4977806391662135e-06}
{"train_loss": 0.040346626192331314, "global_step": 223607, "epoch": 2512, "lr": 7.4974752883428255e-06}
{"train_loss": 0.06605613976716995, "global_step": 223608, "epoch": 2512, "lr": 7.497169943233396e-06}
{"train_loss": 0.05143685266375542, "global_step": 223609, "epoch": 2512, "lr": 7.496864603837939e-06}
{"train_loss": 0.08983823657035828, "global_step": 223610, "epoch": 2512, "lr": 7.496559270156523e-06}
{"train_loss": 0.029181767255067825, "global_step": 223611, "epoch": 2512, "lr": 7.496253942189163e-06}
{"train_loss": 0.05700401961803436, "global_step": 223612, "epoch": 2512, "lr": 7.495948619935927e-06}
{"train_loss": 0.055458951741456985, "global_step": 223613, "epoch": 2512, "lr": 7.495643303396832e-06}
{"train_loss": 0.0826183632016182, "global_step": 223614, "epoch": 2512, "lr": 7.495337992571938e-06}
{"train_loss": 0.10853774845600128, "global_step": 223615, "epoch": 2512, "lr": 7.495032687461273e-06}
{"train_loss": 0.0817696824669838, "global_step": 223616, "epoch": 2512, "lr": 7.494727388064887e-06}
{"train_loss": 0.08638853579759598, "global_step": 223617, "epoch": 2512, "lr": 7.494422094382808e-06}
{"train_loss": 0.04850492626428604, "global_step": 223618, "epoch": 2512, "lr": 7.494116806415097e-06}
{"train_loss": 0.04242157191038132, "global_step": 223619, "epoch": 2512, "lr": 7.493811524161776e-06}
{"train_loss": 0.04491790384054184, "global_step": 223620, "epoch": 2512, "lr": 7.4935062476229e-06}
{"train_loss": 0.046570997685194016, "global_step": 223621, "epoch": 2512, "lr": 7.493200976798498e-06}
{"train_loss": 0.0983051061630249, "global_step": 223622, "epoch": 2512, "lr": 7.492895711688624e-06}
{"train_loss": 0.05706404149532318, "global_step": 223623, "epoch": 2512, "lr": 7.492590452293302e-06}
{"train_loss": 0.0870177298784256, "global_step": 223624, "epoch": 2512, "lr": 7.492285198612597e-06}
{"train_loss": 0.052194274961948395, "global_step": 223625, "epoch": 2512, "lr": 7.4919799506465215e-06}
{"train_loss": 0.057838331907987595, "global_step": 223626, "epoch": 2512, "lr": 7.491674708395141e-06}
{"train_loss": 0.0943804532289505, "global_step": 223627, "epoch": 2512, "lr": 7.491369471858478e-06}
{"train_loss": 0.02791338972747326, "global_step": 223628, "epoch": 2512, "lr": 7.491064241036594e-06}
{"train_loss": 0.08376960456371307, "global_step": 223629, "epoch": 2512, "lr": 7.490759015929522e-06}
{"train_loss": 0.03959166258573532, "global_step": 223630, "epoch": 2512, "lr": 7.490453796537283e-06}
{"train_loss": 0.027822345495224, "global_step": 223631, "epoch": 2512, "lr": 7.490148582859946e-06}
{"train_loss": 0.053877636790275574, "global_step": 223632, "epoch": 2512, "lr": 7.489843374897526e-06}
{"train_loss": 0.05641850456595421, "global_step": 223633, "epoch": 2512, "lr": 7.489538172650096e-06}
{"train_loss": 0.05111971125006676, "global_step": 223634, "epoch": 2512, "lr": 7.489232976117666e-06}
{"train_loss": 0.06594786792993546, "global_step": 223635, "epoch": 2512, "lr": 7.4889277853003036e-06}
{"train_loss": 0.013943844474852085, "global_step": 223636, "epoch": 2512, "lr": 7.48862260019802e-06}
{"train_loss": 0.03927646949887276, "global_step": 223637, "epoch": 2512, "lr": 7.488317420810887e-06}
{"train_loss": 0.039150744676589966, "global_step": 223638, "epoch": 2512, "lr": 7.48801224713892e-06}
{"train_loss": 0.034384194761514664, "global_step": 223639, "epoch": 2512, "lr": 7.487707079182188e-06}
{"train_loss": 0.03455974534153938, "global_step": 223640, "epoch": 2512, "lr": 7.487401916940695e-06}
{"train_loss": 0.050114087760448456, "global_step": 223641, "epoch": 2512, "lr": 7.487096760414519e-06}
{"train_loss": 0.013600781559944153, "global_step": 223642, "epoch": 2512, "lr": 7.486791609603677e-06}
{"train_loss": 0.021523885428905487, "global_step": 223643, "epoch": 2512, "lr": 7.486486464508213e-06}
{"train_loss": 0.0318341925740242, "global_step": 223644, "epoch": 2512, "lr": 7.4861813251281835e-06}
{"train_loss": 0.06163576617836952, "global_step": 223645, "epoch": 2512, "lr": 7.4858761914636086e-06}
{"train_loss": 0.05922122672200203, "global_step": 223646, "epoch": 2512, "lr": 7.485571063514552e-06}
{"train_loss": 0.0527869276702404, "global_step": 223647, "epoch": 2512, "lr": 7.485265941281028e-06}
{"train_loss": 0.06198481470346451, "global_step": 223648, "epoch": 2512, "lr": 7.4849608247631094e-06}
{"train_loss": 0.034067023545503616, "global_step": 223649, "epoch": 2512, "lr": 7.484655713960803e-06}
{"train_loss": 0.04829515516757965, "global_step": 223650, "epoch": 2512, "lr": 7.48435060887418e-06}
{"train_loss": 0.09949987381696701, "global_step": 223651, "epoch": 2512, "lr": 7.4840455095032516e-06}
{"train_loss": 0.02165679633617401, "global_step": 223652, "epoch": 2512, "lr": 7.48374041584809e-06}
{"train_loss": 0.06651201099157333, "global_step": 223653, "epoch": 2512, "lr": 7.483435327908711e-06}
{"train_loss": 0.06259934604167938, "global_step": 223654, "epoch": 2512, "lr": 7.483130245685177e-06}
{"train_loss": 0.08550363779067993, "global_step": 223655, "epoch": 2512, "lr": 7.482825169177504e-06}
{"train_loss": 0.05269368655268061, "global_step": 223656, "epoch": 2512, "lr": 7.482520098385765e-06, "val_loss": 8.989909172058105}
{"train_loss": 0.06470642238855362, "global_step": 223657, "epoch": 2513, "lr": 7.482215033309964e-06}
{"train_loss": 0.009655570611357689, "global_step": 223658, "epoch": 2513, "lr": 7.481909973950174e-06}
{"train_loss": 0.037973418831825256, "global_step": 223659, "epoch": 2513, "lr": 7.481604920306417e-06}
{"train_loss": 0.05899697169661522, "global_step": 223660, "epoch": 2513, "lr": 7.481299872378733e-06}
{"train_loss": 0.05072064697742462, "global_step": 223661, "epoch": 2513, "lr": 7.4809948301671875e-06}
{"train_loss": 0.04995216801762581, "global_step": 223662, "epoch": 2513, "lr": 7.480689793671797e-06}
{"train_loss": 0.05124366283416748, "global_step": 223663, "epoch": 2513, "lr": 7.480384762892601e-06}
{"train_loss": 0.08314643800258636, "global_step": 223664, "epoch": 2513, "lr": 7.480079737829665e-06}
{"train_loss": 0.031398121267557144, "global_step": 223665, "epoch": 2513, "lr": 7.479774718482996e-06}
{"train_loss": 0.04041336476802826, "global_step": 223666, "epoch": 2513, "lr": 7.479469704852671e-06}
{"train_loss": 0.028228450566530228, "global_step": 223667, "epoch": 2513, "lr": 7.479164696938707e-06}
{"train_loss": 0.049540191888809204, "global_step": 223668, "epoch": 2513, "lr": 7.478859694741141e-06}
{"train_loss": 0.04853282868862152, "global_step": 223669, "epoch": 2513, "lr": 7.478554698260037e-06}
{"train_loss": 0.07084406912326813, "global_step": 223670, "epoch": 2513, "lr": 7.478249707495411e-06}
{"train_loss": 0.056632913649082184, "global_step": 223671, "epoch": 2513, "lr": 7.477944722447328e-06}
{"train_loss": 0.06665366142988205, "global_step": 223672, "epoch": 2513, "lr": 7.4776397431158056e-06}
{"train_loss": 0.062290191650390625, "global_step": 223673, "epoch": 2513, "lr": 7.47733476950091e-06}
{"train_loss": 0.06598097831010818, "global_step": 223674, "epoch": 2513, "lr": 7.477029801602653e-06}
{"train_loss": 0.05373461917042732, "global_step": 223675, "epoch": 2513, "lr": 7.476724839421101e-06}
{"train_loss": 0.030079152435064316, "global_step": 223676, "epoch": 2513, "lr": 7.476419882956276e-06}
{"train_loss": 0.02924577333033085, "global_step": 223677, "epoch": 2513, "lr": 7.476114932208245e-06}
{"train_loss": 0.021061912178993225, "global_step": 223678, "epoch": 2513, "lr": 7.4758099871770125e-06}
{"train_loss": 0.0279610026627779, "global_step": 223679, "epoch": 2513, "lr": 7.475505047862641e-06}
{"train_loss": 0.061221644282341, "global_step": 223680, "epoch": 2513, "lr": 7.475200114265185e-06}
{"train_loss": 0.021023286506533623, "global_step": 223681, "epoch": 2513, "lr": 7.474895186384656e-06}
{"train_loss": 0.043427783995866776, "global_step": 223682, "epoch": 2513, "lr": 7.474590264221121e-06}
{"train_loss": 0.0255192369222641, "global_step": 223683, "epoch": 2513, "lr": 7.474285347774596e-06}
{"train_loss": 0.030171625316143036, "global_step": 223684, "epoch": 2513, "lr": 7.473980437045147e-06}
{"train_loss": 0.061359040439128876, "global_step": 223685, "epoch": 2513, "lr": 7.473675532032792e-06}
{"train_loss": 0.018911678344011307, "global_step": 223686, "epoch": 2513, "lr": 7.473370632737592e-06}
{"train_loss": 0.04052870720624924, "global_step": 223687, "epoch": 2513, "lr": 7.473065739159568e-06}
{"train_loss": 0.021926341578364372, "global_step": 223688, "epoch": 2513, "lr": 7.472760851298788e-06}
{"train_loss": 0.02925405278801918, "global_step": 223689, "epoch": 2513, "lr": 7.472455969155262e-06}
{"train_loss": 0.08872810751199722, "global_step": 223690, "epoch": 2513, "lr": 7.472151092729058e-06}
{"train_loss": 0.06256534904241562, "global_step": 223691, "epoch": 2513, "lr": 7.471846222020195e-06}
{"train_loss": 0.02482365071773529, "global_step": 223692, "epoch": 2513, "lr": 7.4715413570287326e-06}
{"train_loss": 0.07022783905267715, "global_step": 223693, "epoch": 2513, "lr": 7.4712364977546965e-06}
{"train_loss": 0.06352361291646957, "global_step": 223694, "epoch": 2513, "lr": 7.470931644198142e-06}
{"train_loss": 0.05655013397336006, "global_step": 223695, "epoch": 2513, "lr": 7.470626796359087e-06}
{"train_loss": 0.04342501983046532, "global_step": 223696, "epoch": 2513, "lr": 7.470321954237608e-06}
{"train_loss": 0.09096311777830124, "global_step": 223697, "epoch": 2513, "lr": 7.470017117833711e-06}
{"train_loss": 0.08744627237319946, "global_step": 223698, "epoch": 2513, "lr": 7.469712287147446e-06}
{"train_loss": 0.05827150493860245, "global_step": 223699, "epoch": 2513, "lr": 7.469407462178879e-06}
{"train_loss": 0.035481665283441544, "global_step": 223700, "epoch": 2513, "lr": 7.469102642928027e-06}
{"train_loss": 0.03809882700443268, "global_step": 223701, "epoch": 2513, "lr": 7.468797829394925e-06}
{"train_loss": 0.05949581414461136, "global_step": 223702, "epoch": 2513, "lr": 7.468493021579642e-06}
{"train_loss": 0.07153934240341187, "global_step": 223703, "epoch": 2513, "lr": 7.468188219482181e-06}
{"train_loss": 0.035803865641355515, "global_step": 223704, "epoch": 2513, "lr": 7.467883423102623e-06}
{"train_loss": 0.05365193635225296, "global_step": 223705, "epoch": 2513, "lr": 7.4675786324409855e-06}
{"train_loss": 0.0762270838022232, "global_step": 223706, "epoch": 2513, "lr": 7.4672738474973025e-06}
{"train_loss": 0.051521144807338715, "global_step": 223707, "epoch": 2513, "lr": 7.46696906827164e-06}
{"train_loss": 0.07426731288433075, "global_step": 223708, "epoch": 2513, "lr": 7.466664294764009e-06}
{"train_loss": 0.07499584555625916, "global_step": 223709, "epoch": 2513, "lr": 7.466359526974481e-06}
{"train_loss": 0.0524563305079937, "global_step": 223710, "epoch": 2513, "lr": 7.4660547649030745e-06}
{"train_loss": 0.08330471068620682, "global_step": 223711, "epoch": 2513, "lr": 7.4657500085498435e-06}
{"train_loss": 0.06522898375988007, "global_step": 223712, "epoch": 2513, "lr": 7.465445257914816e-06}
{"train_loss": 0.03042958676815033, "global_step": 223713, "epoch": 2513, "lr": 7.465140512998053e-06}
{"train_loss": 0.07932208478450775, "global_step": 223714, "epoch": 2513, "lr": 7.464835773799572e-06}
{"train_loss": 0.05751412734389305, "global_step": 223715, "epoch": 2513, "lr": 7.464531040319422e-06}
{"train_loss": 0.03794116899371147, "global_step": 223716, "epoch": 2513, "lr": 7.464226312557665e-06}
{"train_loss": 0.0872064158320427, "global_step": 223717, "epoch": 2513, "lr": 7.463921590514305e-06}
{"train_loss": 0.08543144166469574, "global_step": 223718, "epoch": 2513, "lr": 7.463616874189422e-06}
{"train_loss": 0.05294935032725334, "global_step": 223719, "epoch": 2513, "lr": 7.463312163583019e-06}
{"train_loss": 0.07122601568698883, "global_step": 223720, "epoch": 2513, "lr": 7.46300745869517e-06}
{"train_loss": 0.032422274351119995, "global_step": 223721, "epoch": 2513, "lr": 7.4627027595258915e-06}
{"train_loss": 0.04162272810935974, "global_step": 223722, "epoch": 2513, "lr": 7.462398066075243e-06}
{"train_loss": 0.02319769375026226, "global_step": 223723, "epoch": 2513, "lr": 7.462093378343244e-06}
{"train_loss": 0.09351397305727005, "global_step": 223724, "epoch": 2513, "lr": 7.461788696329963e-06}
{"train_loss": 0.05339912325143814, "global_step": 223725, "epoch": 2513, "lr": 7.461484020035408e-06}
{"train_loss": 0.047231245785951614, "global_step": 223726, "epoch": 2513, "lr": 7.461179349459657e-06}
{"train_loss": 0.013884343206882477, "global_step": 223727, "epoch": 2513, "lr": 7.460874684602715e-06}
{"train_loss": 0.0453743152320385, "global_step": 223728, "epoch": 2513, "lr": 7.460570025464653e-06}
{"train_loss": 0.05220619589090347, "global_step": 223729, "epoch": 2513, "lr": 7.460265372045489e-06}
{"train_loss": 0.03657558560371399, "global_step": 223730, "epoch": 2513, "lr": 7.459960724345289e-06}
{"train_loss": 0.054428160190582275, "global_step": 223731, "epoch": 2513, "lr": 7.459656082364059e-06}
{"train_loss": 0.027168268337845802, "global_step": 223732, "epoch": 2513, "lr": 7.4593514461018756e-06}
{"train_loss": 0.10673055052757263, "global_step": 223733, "epoch": 2513, "lr": 7.459046815558751e-06}
{"train_loss": 0.02789829671382904, "global_step": 223734, "epoch": 2513, "lr": 7.458742190734741e-06}
{"train_loss": 0.02825896069407463, "global_step": 223735, "epoch": 2513, "lr": 7.458437571629895e-06}
{"train_loss": 0.054118912667036057, "global_step": 223736, "epoch": 2513, "lr": 7.458132958244235e-06}
{"train_loss": 0.062422387301921844, "global_step": 223737, "epoch": 2513, "lr": 7.457828350577817e-06}
{"train_loss": 0.11332538723945618, "global_step": 223738, "epoch": 2513, "lr": 7.457523748630668e-06}
{"train_loss": 0.07268141210079193, "global_step": 223739, "epoch": 2513, "lr": 7.457219152402844e-06}
{"train_loss": 0.06526142358779907, "global_step": 223740, "epoch": 2513, "lr": 7.456914561894379e-06}
{"train_loss": 0.06589074432849884, "global_step": 223741, "epoch": 2513, "lr": 7.456609977105306e-06}
{"train_loss": 0.03618231415748596, "global_step": 223742, "epoch": 2513, "lr": 7.456305398035679e-06}
{"train_loss": 0.04805457964539528, "global_step": 223743, "epoch": 2513, "lr": 7.456000824685527e-06}
{"train_loss": 0.07662862539291382, "global_step": 223744, "epoch": 2513, "lr": 7.455696257054906e-06}
{"train_loss": 0.05232102551570769, "global_step": 223745, "epoch": 2513, "lr": 7.455391695143849e-06, "val_loss": 9.037324905395508}
{"train_loss": 0.06456815451383591, "global_step": 223746, "epoch": 2514, "lr": 7.455087138952382e-06}
{"train_loss": 0.017147544771432877, "global_step": 223747, "epoch": 2514, "lr": 7.4547825884805744e-06}
{"train_loss": 0.05609213560819626, "global_step": 223748, "epoch": 2514, "lr": 7.454478043728441e-06}
{"train_loss": 0.03696008771657944, "global_step": 223749, "epoch": 2514, "lr": 7.4541735046960436e-06}
{"train_loss": 0.016107000410556793, "global_step": 223750, "epoch": 2514, "lr": 7.4538689713834045e-06}
{"train_loss": 0.05057968199253082, "global_step": 223751, "epoch": 2514, "lr": 7.453564443790573e-06}
{"train_loss": 0.04618639871478081, "global_step": 223752, "epoch": 2514, "lr": 7.453259921917604e-06}
{"train_loss": 0.09226245433092117, "global_step": 223753, "epoch": 2514, "lr": 7.452955405764511e-06}
{"train_loss": 0.09769884496927261, "global_step": 223754, "epoch": 2514, "lr": 7.4526508953313645e-06}
{"train_loss": 0.05359776318073273, "global_step": 223755, "epoch": 2514, "lr": 7.452346390618181e-06}
{"train_loss": 0.05982520803809166, "global_step": 223756, "epoch": 2514, "lr": 7.452041891625017e-06}
{"train_loss": 0.03889939934015274, "global_step": 223757, "epoch": 2514, "lr": 7.451737398351899e-06}
{"train_loss": 0.05031387507915497, "global_step": 223758, "epoch": 2514, "lr": 7.451432910798889e-06}
{"train_loss": 0.03412148728966713, "global_step": 223759, "epoch": 2514, "lr": 7.451128428966003e-06}
{"train_loss": 0.05079913139343262, "global_step": 223760, "epoch": 2514, "lr": 7.450823952853303e-06}
{"train_loss": 0.0262922216206789, "global_step": 223761, "epoch": 2514, "lr": 7.450519482460811e-06}
{"train_loss": 0.06470116227865219, "global_step": 223762, "epoch": 2514, "lr": 7.450215017788592e-06}
{"train_loss": 0.03476748988032341, "global_step": 223763, "epoch": 2514, "lr": 7.449910558836659e-06}
{"train_loss": 0.028384411707520485, "global_step": 223764, "epoch": 2514, "lr": 7.449606105605078e-06}
{"train_loss": 0.06651081889867783, "global_step": 223765, "epoch": 2514, "lr": 7.4493016580938655e-06}
{"train_loss": 0.1030132994055748, "global_step": 223766, "epoch": 2514, "lr": 7.448997216303094e-06}
{"train_loss": 0.0694601833820343, "global_step": 223767, "epoch": 2514, "lr": 7.448692780232769e-06}
{"train_loss": 0.02981646917760372, "global_step": 223768, "epoch": 2514, "lr": 7.448388349882962e-06}
{"train_loss": 0.06495703011751175, "global_step": 223769, "epoch": 2514, "lr": 7.4480839252536905e-06}
{"train_loss": 0.07379255443811417, "global_step": 223770, "epoch": 2514, "lr": 7.447779506345004e-06}
{"train_loss": 0.03770535811781883, "global_step": 223771, "epoch": 2514, "lr": 7.447475093156952e-06}
{"train_loss": 0.055730920284986496, "global_step": 223772, "epoch": 2514, "lr": 7.447170685689564e-06}
{"train_loss": 0.05160971358418465, "global_step": 223773, "epoch": 2514, "lr": 7.446866283942894e-06}
{"train_loss": 0.02872936800122261, "global_step": 223774, "epoch": 2514, "lr": 7.446561887916958e-06}
{"train_loss": 0.07349951565265656, "global_step": 223775, "epoch": 2514, "lr": 7.44625749761183e-06}
{"train_loss": 0.03319544717669487, "global_step": 223776, "epoch": 2514, "lr": 7.445953113027526e-06}
{"train_loss": 0.03687308356165886, "global_step": 223777, "epoch": 2514, "lr": 7.445648734164101e-06}
{"train_loss": 0.0469009168446064, "global_step": 223778, "epoch": 2514, "lr": 7.445344361021589e-06}
{"train_loss": 0.019787849858403206, "global_step": 223779, "epoch": 2514, "lr": 7.4450399936000225e-06}
{"train_loss": 0.054986730217933655, "global_step": 223780, "epoch": 2514, "lr": 7.444735631899463e-06}
{"train_loss": 0.02976997196674347, "global_step": 223781, "epoch": 2514, "lr": 7.444431275919927e-06}
{"train_loss": 0.0813731998205185, "global_step": 223782, "epoch": 2514, "lr": 7.444126925661482e-06}
{"train_loss": 0.015494673512876034, "global_step": 223783, "epoch": 2514, "lr": 7.443822581124155e-06}
{"train_loss": 0.046683695167303085, "global_step": 223784, "epoch": 2514, "lr": 7.4435182423079785e-06}
{"train_loss": 0.028016306459903717, "global_step": 223785, "epoch": 2514, "lr": 7.443213909213009e-06}
{"train_loss": 0.03408800810575485, "global_step": 223786, "epoch": 2514, "lr": 7.442909581839274e-06}
{"train_loss": 0.08820316195487976, "global_step": 223787, "epoch": 2514, "lr": 7.442605260186819e-06}
{"train_loss": 0.026821278035640717, "global_step": 223788, "epoch": 2514, "lr": 7.442300944255698e-06}
{"train_loss": 0.061346523463726044, "global_step": 223789, "epoch": 2514, "lr": 7.441996634045928e-06}
{"train_loss": 0.09969853609800339, "global_step": 223790, "epoch": 2514, "lr": 7.441692329557575e-06}
{"train_loss": 0.05482916906476021, "global_step": 223791, "epoch": 2514, "lr": 7.441388030790658e-06}
{"train_loss": 0.10915040224790573, "global_step": 223792, "epoch": 2514, "lr": 7.441083737745236e-06}
{"train_loss": 0.07419533282518387, "global_step": 223793, "epoch": 2514, "lr": 7.440779450421331e-06}
{"train_loss": 0.056021690368652344, "global_step": 223794, "epoch": 2514, "lr": 7.440475168819011e-06}
{"train_loss": 0.03649522736668587, "global_step": 223795, "epoch": 2514, "lr": 7.440170892938281e-06}
{"train_loss": 0.05745626613497734, "global_step": 223796, "epoch": 2514, "lr": 7.439866622779218e-06}
{"train_loss": 0.06864523887634277, "global_step": 223797, "epoch": 2514, "lr": 7.439562358341834e-06}
{"train_loss": 0.027010880410671234, "global_step": 223798, "epoch": 2514, "lr": 7.439258099626195e-06}
{"train_loss": 0.034263864159584045, "global_step": 223799, "epoch": 2514, "lr": 7.4389538466323126e-06}
{"train_loss": 0.07366980612277985, "global_step": 223800, "epoch": 2514, "lr": 7.438649599360259e-06}
{"train_loss": 0.062141116708517075, "global_step": 223801, "epoch": 2514, "lr": 7.43834535781005e-06}
{"train_loss": 0.08405845612287521, "global_step": 223802, "epoch": 2514, "lr": 7.438041121981748e-06}
{"train_loss": 0.050264548510313034, "global_step": 223803, "epoch": 2514, "lr": 7.437736891875369e-06}
{"train_loss": 0.03492823988199234, "global_step": 223804, "epoch": 2514, "lr": 7.437432667490979e-06}
{"train_loss": 0.04354667291045189, "global_step": 223805, "epoch": 2514, "lr": 7.437128448828595e-06}
{"train_loss": 0.056884996592998505, "global_step": 223806, "epoch": 2514, "lr": 7.436824235888274e-06}
{"train_loss": 0.03658498078584671, "global_step": 223807, "epoch": 2514, "lr": 7.436520028670063e-06}
{"train_loss": 0.07160245627164841, "global_step": 223808, "epoch": 2514, "lr": 7.436215827173981e-06}
{"train_loss": 0.05130273103713989, "global_step": 223809, "epoch": 2514, "lr": 7.435911631400094e-06}
{"train_loss": 0.06359988451004028, "global_step": 223810, "epoch": 2514, "lr": 7.435607441348413e-06}
{"train_loss": 0.05003819987177849, "global_step": 223811, "epoch": 2514, "lr": 7.435303257019016e-06}
{"train_loss": 0.03286983072757721, "global_step": 223812, "epoch": 2514, "lr": 7.434999078411903e-06}
{"train_loss": 0.03607865795493126, "global_step": 223813, "epoch": 2514, "lr": 7.434694905527157e-06}
{"train_loss": 0.04366431385278702, "global_step": 223814, "epoch": 2514, "lr": 7.434390738364777e-06}
{"train_loss": 0.04968886449933052, "global_step": 223815, "epoch": 2514, "lr": 7.434086576924842e-06}
{"train_loss": 0.051773954182863235, "global_step": 223816, "epoch": 2514, "lr": 7.433782421207369e-06}
{"train_loss": 0.042891670018434525, "global_step": 223817, "epoch": 2514, "lr": 7.433478271212402e-06}
{"train_loss": 0.0981791615486145, "global_step": 223818, "epoch": 2514, "lr": 7.43317412693999e-06}
{"train_loss": 0.044373467564582825, "global_step": 223819, "epoch": 2514, "lr": 7.4328699883901565e-06}
{"train_loss": 0.04464062303304672, "global_step": 223820, "epoch": 2514, "lr": 7.432565855562973e-06}
{"train_loss": 0.028981151059269905, "global_step": 223821, "epoch": 2514, "lr": 7.432261728458456e-06}
{"train_loss": 0.0388900563120842, "global_step": 223822, "epoch": 2514, "lr": 7.431957607076645e-06}
{"train_loss": 0.09421344101428986, "global_step": 223823, "epoch": 2514, "lr": 7.4316534914176e-06}
{"train_loss": 0.031431712210178375, "global_step": 223824, "epoch": 2514, "lr": 7.431349381481334e-06}
{"train_loss": 0.04574086517095566, "global_step": 223825, "epoch": 2514, "lr": 7.431045277267912e-06}
{"train_loss": 0.0355621799826622, "global_step": 223826, "epoch": 2514, "lr": 7.430741178777373e-06}
{"train_loss": 0.04737427085638046, "global_step": 223827, "epoch": 2514, "lr": 7.430437086009739e-06}
{"train_loss": 0.05709163472056389, "global_step": 223828, "epoch": 2514, "lr": 7.430132998965078e-06}
{"train_loss": 0.034832295030355453, "global_step": 223829, "epoch": 2514, "lr": 7.4298289176434056e-06}
{"train_loss": 0.05274422839283943, "global_step": 223830, "epoch": 2514, "lr": 7.4295248420447884e-06}
{"train_loss": 0.05013906955718994, "global_step": 223831, "epoch": 2514, "lr": 7.4292207721692375e-06}
{"train_loss": 0.056792501360177994, "global_step": 223832, "epoch": 2514, "lr": 7.42891670801682e-06}
{"train_loss": 0.06598706543445587, "global_step": 223833, "epoch": 2514, "lr": 7.428612649587552e-06}
{"train_loss": 0.051984686414930925, "global_step": 223834, "epoch": 2514, "lr": 7.4283085968815054e-06, "val_loss": 8.982701301574707}
{"train_loss": 0.07534956932067871, "global_step": 223835, "epoch": 2515, "lr": 7.428004549898687e-06}
{"train_loss": 0.057664476335048676, "global_step": 223836, "epoch": 2515, "lr": 7.427700508639173e-06}
{"train_loss": 0.04096531495451927, "global_step": 223837, "epoch": 2515, "lr": 7.427396473102966e-06}
{"train_loss": 0.06152572110295296, "global_step": 223838, "epoch": 2515, "lr": 7.4270924432901454e-06}
{"train_loss": 0.040368352085351944, "global_step": 223839, "epoch": 2515, "lr": 7.42678841920072e-06}
{"train_loss": 0.027379710227251053, "global_step": 223840, "epoch": 2515, "lr": 7.4264844008347486e-06}
{"train_loss": 0.03482760861515999, "global_step": 223841, "epoch": 2515, "lr": 7.426180388192266e-06}
{"train_loss": 0.06598098576068878, "global_step": 223842, "epoch": 2515, "lr": 7.425876381273311e-06}
{"train_loss": 0.04830729588866234, "global_step": 223843, "epoch": 2515, "lr": 7.425572380077939e-06}
{"train_loss": 0.03805295377969742, "global_step": 223844, "epoch": 2515, "lr": 7.425268384606165e-06}
{"train_loss": 0.04131182283163071, "global_step": 223845, "epoch": 2515, "lr": 7.424964394858058e-06}
{"train_loss": 0.01900874823331833, "global_step": 223846, "epoch": 2515, "lr": 7.424660410833639e-06}
{"train_loss": 0.028210705146193504, "global_step": 223847, "epoch": 2515, "lr": 7.424356432532958e-06}
{"train_loss": 0.03794708847999573, "global_step": 223848, "epoch": 2515, "lr": 7.424052459956049e-06}
{"train_loss": 0.07152367383241653, "global_step": 223849, "epoch": 2515, "lr": 7.423748493102966e-06}
{"train_loss": 0.02392972819507122, "global_step": 223850, "epoch": 2515, "lr": 7.423444531973734e-06}
{"train_loss": 0.05658847093582153, "global_step": 223851, "epoch": 2515, "lr": 7.423140576568405e-06}
{"train_loss": 0.042994286864995956, "global_step": 223852, "epoch": 2515, "lr": 7.422836626887009e-06}
{"train_loss": 0.07286836951971054, "global_step": 223853, "epoch": 2515, "lr": 7.422532682929606e-06}
{"train_loss": 0.044437699019908905, "global_step": 223854, "epoch": 2515, "lr": 7.422228744696214e-06}
{"train_loss": 0.06829604506492615, "global_step": 223855, "epoch": 2515, "lr": 7.4219248121868926e-06}
{"train_loss": 0.03912325203418732, "global_step": 223856, "epoch": 2515, "lr": 7.421620885401675e-06}
{"train_loss": 0.07117494195699692, "global_step": 223857, "epoch": 2515, "lr": 7.421316964340591e-06}
{"train_loss": 0.05344275012612343, "global_step": 223858, "epoch": 2515, "lr": 7.4210130490037e-06}
{"train_loss": 0.08010256290435791, "global_step": 223859, "epoch": 2515, "lr": 7.4207091393910245e-06}
{"train_loss": 0.02399190329015255, "global_step": 223860, "epoch": 2515, "lr": 7.420405235502631e-06}
{"train_loss": 0.027143800631165504, "global_step": 223861, "epoch": 2515, "lr": 7.420101337338531e-06}
{"train_loss": 0.02189481630921364, "global_step": 223862, "epoch": 2515, "lr": 7.41979744489879e-06}
{"train_loss": 0.043408337980508804, "global_step": 223863, "epoch": 2515, "lr": 7.419493558183427e-06}
{"train_loss": 0.036895737051963806, "global_step": 223864, "epoch": 2515, "lr": 7.419189677192507e-06}
{"train_loss": 0.07933296263217926, "global_step": 223865, "epoch": 2515, "lr": 7.418885801926046e-06}
{"train_loss": 0.023448364809155464, "global_step": 223866, "epoch": 2515, "lr": 7.418581932384111e-06}
{"train_loss": 0.036654967814683914, "global_step": 223867, "epoch": 2515, "lr": 7.4182780685667144e-06}
{"train_loss": 0.12699955701828003, "global_step": 223868, "epoch": 2515, "lr": 7.417974210473921e-06}
{"train_loss": 0.08193015307188034, "global_step": 223869, "epoch": 2515, "lr": 7.417670358105755e-06}
{"train_loss": 0.05509740859270096, "global_step": 223870, "epoch": 2515, "lr": 7.4173665114622704e-06}
{"train_loss": 0.04491815343499184, "global_step": 223871, "epoch": 2515, "lr": 7.417062670543495e-06}
{"train_loss": 0.06926531344652176, "global_step": 223872, "epoch": 2515, "lr": 7.4167588353494855e-06}
{"train_loss": 0.03441976010799408, "global_step": 223873, "epoch": 2515, "lr": 7.416455005880263e-06}
{"train_loss": 0.07323336601257324, "global_step": 223874, "epoch": 2515, "lr": 7.416151182135889e-06}
{"train_loss": 0.05730090290307999, "global_step": 223875, "epoch": 2515, "lr": 7.415847364116385e-06}
{"train_loss": 0.05828806757926941, "global_step": 223876, "epoch": 2515, "lr": 7.415543551821813e-06}
{"train_loss": 0.079609215259552, "global_step": 223877, "epoch": 2515, "lr": 7.415239745252189e-06}
{"train_loss": 0.03200926631689072, "global_step": 223878, "epoch": 2515, "lr": 7.414935944407569e-06}
{"train_loss": 0.05075966939330101, "global_step": 223879, "epoch": 2515, "lr": 7.414632149287998e-06}
{"train_loss": 0.06847310066223145, "global_step": 223880, "epoch": 2515, "lr": 7.4143283598935075e-06}
{"train_loss": 0.06783427298069, "global_step": 223881, "epoch": 2515, "lr": 7.4140245762241485e-06}
{"train_loss": 0.04309038445353508, "global_step": 223882, "epoch": 2515, "lr": 7.413720798279944e-06}
{"train_loss": 0.025878259912133217, "global_step": 223883, "epoch": 2515, "lr": 7.413417026060954e-06}
{"train_loss": 0.021979650482535362, "global_step": 223884, "epoch": 2515, "lr": 7.4131132595672005e-06}
{"train_loss": 0.04284648597240448, "global_step": 223885, "epoch": 2515, "lr": 7.41280949879875e-06}
{"train_loss": 0.032177191227674484, "global_step": 223886, "epoch": 2515, "lr": 7.412505743755615e-06}
{"train_loss": 0.042128581553697586, "global_step": 223887, "epoch": 2515, "lr": 7.412201994437862e-06}
{"train_loss": 0.039275430142879486, "global_step": 223888, "epoch": 2515, "lr": 7.411898250845506e-06}
{"train_loss": 0.026747576892375946, "global_step": 223889, "epoch": 2515, "lr": 7.411594512978614e-06}
{"train_loss": 0.029208136722445488, "global_step": 223890, "epoch": 2515, "lr": 7.4112907808371986e-06}
{"train_loss": 0.035875026136636734, "global_step": 223891, "epoch": 2515, "lr": 7.410987054421331e-06}
{"train_loss": 0.039421726018190384, "global_step": 223892, "epoch": 2515, "lr": 7.410683333731028e-06}
{"train_loss": 0.046948596835136414, "global_step": 223893, "epoch": 2515, "lr": 7.41037961876635e-06}
{"train_loss": 0.06364638358354568, "global_step": 223894, "epoch": 2515, "lr": 7.410075909527326e-06}
{"train_loss": 0.0882306918501854, "global_step": 223895, "epoch": 2515, "lr": 7.409772206013987e-06}
{"train_loss": 0.06574644148349762, "global_step": 223896, "epoch": 2515, "lr": 7.409468508226397e-06}
{"train_loss": 0.0722847506403923, "global_step": 223897, "epoch": 2515, "lr": 7.409164816164571e-06}
{"train_loss": 0.036024633795022964, "global_step": 223898, "epoch": 2515, "lr": 7.408861129828576e-06}
{"train_loss": 0.06065913662314415, "global_step": 223899, "epoch": 2515, "lr": 7.408557449218428e-06}
{"train_loss": 0.06222601979970932, "global_step": 223900, "epoch": 2515, "lr": 7.408253774334195e-06}
{"train_loss": 0.06933028250932693, "global_step": 223901, "epoch": 2515, "lr": 7.407950105175893e-06}
{"train_loss": 0.0838206559419632, "global_step": 223902, "epoch": 2515, "lr": 7.407646441743577e-06}
{"train_loss": 0.07231776416301727, "global_step": 223903, "epoch": 2515, "lr": 7.40734278403728e-06}
{"train_loss": 0.06802228838205338, "global_step": 223904, "epoch": 2515, "lr": 7.407039132057053e-06}
{"train_loss": 0.022356092929840088, "global_step": 223905, "epoch": 2515, "lr": 7.406735485802923e-06}
{"train_loss": 0.06186164543032646, "global_step": 223906, "epoch": 2515, "lr": 7.406431845274947e-06}
{"train_loss": 0.09844200313091278, "global_step": 223907, "epoch": 2515, "lr": 7.406128210473146e-06}
{"train_loss": 0.07441867142915726, "global_step": 223908, "epoch": 2515, "lr": 7.405824581397586e-06}
{"train_loss": 0.03315313905477524, "global_step": 223909, "epoch": 2515, "lr": 7.405520958048279e-06}
{"train_loss": 0.08551933616399765, "global_step": 223910, "epoch": 2515, "lr": 7.4052173404252915e-06}
{"train_loss": 0.03327289968729019, "global_step": 223911, "epoch": 2515, "lr": 7.404913728528645e-06}
{"train_loss": 0.02728893980383873, "global_step": 223912, "epoch": 2515, "lr": 7.404610122358396e-06}
{"train_loss": 0.03719433769583702, "global_step": 223913, "epoch": 2515, "lr": 7.404306521914573e-06}
{"train_loss": 0.0498649887740612, "global_step": 223914, "epoch": 2515, "lr": 7.404002927197218e-06}
{"train_loss": 0.010321121662855148, "global_step": 223915, "epoch": 2515, "lr": 7.403699338206388e-06}
{"train_loss": 0.05025692284107208, "global_step": 223916, "epoch": 2515, "lr": 7.403395754942094e-06}
{"train_loss": 0.0931171327829361, "global_step": 223917, "epoch": 2515, "lr": 7.4030921774044136e-06}
{"train_loss": 0.08652046322822571, "global_step": 223918, "epoch": 2515, "lr": 7.402788605593352e-06}
{"train_loss": 0.047127097845077515, "global_step": 223919, "epoch": 2515, "lr": 7.402485039508983e-06}
{"train_loss": 0.053574997931718826, "global_step": 223920, "epoch": 2515, "lr": 7.402181479151316e-06}
{"train_loss": 0.05796819180250168, "global_step": 223921, "epoch": 2515, "lr": 7.401877924520417e-06}
{"train_loss": 0.08049024641513824, "global_step": 223922, "epoch": 2515, "lr": 7.401574375616305e-06}
{"train_loss": 0.0521754938899801, "global_step": 223923, "epoch": 2515, "lr": 7.401270832439044e-06, "val_loss": 8.985773086547852, "train_action_mse_error": 24.83682632446289}
{"train_loss": 0.051550887525081635, "global_step": 223924, "epoch": 2516, "lr": 7.4009672949886536e-06}
{"train_loss": 0.06657523661851883, "global_step": 223925, "epoch": 2516, "lr": 7.400663763265192e-06}
{"train_loss": 0.046662475913763046, "global_step": 223926, "epoch": 2516, "lr": 7.400360237268683e-06}
{"train_loss": 0.03510192781686783, "global_step": 223927, "epoch": 2516, "lr": 7.400056716999188e-06}
{"train_loss": 0.03943582996726036, "global_step": 223928, "epoch": 2516, "lr": 7.399753202456722e-06}
{"train_loss": 0.037383001297712326, "global_step": 223929, "epoch": 2516, "lr": 7.399449693641353e-06}
{"train_loss": 0.0674481987953186, "global_step": 223930, "epoch": 2516, "lr": 7.3991461905530965e-06}
{"train_loss": 0.08565765619277954, "global_step": 223931, "epoch": 2516, "lr": 7.398842693192021e-06}
{"train_loss": 0.05562156066298485, "global_step": 223932, "epoch": 2516, "lr": 7.398539201558147e-06}
{"train_loss": 0.045822057873010635, "global_step": 223933, "epoch": 2516, "lr": 7.398235715651514e-06}
{"train_loss": 0.03271571546792984, "global_step": 223934, "epoch": 2516, "lr": 7.397932235472177e-06}
{"train_loss": 0.020698003470897675, "global_step": 223935, "epoch": 2516, "lr": 7.397628761020153e-06}
{"train_loss": 0.03365441411733627, "global_step": 223936, "epoch": 2516, "lr": 7.397325292295515e-06}
{"train_loss": 0.08116266876459122, "global_step": 223937, "epoch": 2516, "lr": 7.397021829298278e-06}
{"train_loss": 0.04360886663198471, "global_step": 223938, "epoch": 2516, "lr": 7.396718372028499e-06}
{"train_loss": 0.07793577015399933, "global_step": 223939, "epoch": 2516, "lr": 7.3964149204862e-06}
{"train_loss": 0.04393438994884491, "global_step": 223940, "epoch": 2516, "lr": 7.3961114746714525e-06}
{"train_loss": 0.07692963629961014, "global_step": 223941, "epoch": 2516, "lr": 7.395808034584262e-06}
{"train_loss": 0.030153347179293633, "global_step": 223942, "epoch": 2516, "lr": 7.395504600224695e-06}
{"train_loss": 0.023745162412524223, "global_step": 223943, "epoch": 2516, "lr": 7.395201171592775e-06}
{"train_loss": 0.0872795581817627, "global_step": 223944, "epoch": 2516, "lr": 7.3948977486885616e-06}
{"train_loss": 0.04617881402373314, "global_step": 223945, "epoch": 2516, "lr": 7.394594331512072e-06}
{"train_loss": 0.07551988214254379, "global_step": 223946, "epoch": 2516, "lr": 7.394290920063374e-06}
{"train_loss": 0.056000933051109314, "global_step": 223947, "epoch": 2516, "lr": 7.393987514342482e-06}
{"train_loss": 0.08655842393636703, "global_step": 223948, "epoch": 2516, "lr": 7.3936841143494585e-06}
{"train_loss": 0.10060456395149231, "global_step": 223949, "epoch": 2516, "lr": 7.393380720084325e-06}
{"train_loss": 0.044881802052259445, "global_step": 223950, "epoch": 2516, "lr": 7.393077331547144e-06}
{"train_loss": 0.06304328143596649, "global_step": 223951, "epoch": 2516, "lr": 7.39277394873793e-06}
{"train_loss": 0.023808028548955917, "global_step": 223952, "epoch": 2516, "lr": 7.39247057165674e-06}
{"train_loss": 0.02649291232228279, "global_step": 223953, "epoch": 2516, "lr": 7.392167200303629e-06}
{"train_loss": 0.08015409857034683, "global_step": 223954, "epoch": 2516, "lr": 7.391863834678603e-06}
{"train_loss": 0.05667899549007416, "global_step": 223955, "epoch": 2516, "lr": 7.3915604747817395e-06}
{"train_loss": 0.06436259299516678, "global_step": 223956, "epoch": 2516, "lr": 7.391257120613049e-06}
{"train_loss": 0.019688284024596214, "global_step": 223957, "epoch": 2516, "lr": 7.390953772172593e-06}
{"train_loss": 0.05601520091295242, "global_step": 223958, "epoch": 2516, "lr": 7.390650429460394e-06}
{"train_loss": 0.04558516666293144, "global_step": 223959, "epoch": 2516, "lr": 7.390347092476513e-06}
{"train_loss": 0.019323499873280525, "global_step": 223960, "epoch": 2516, "lr": 7.390043761220972e-06}
{"train_loss": 0.044396497309207916, "global_step": 223961, "epoch": 2516, "lr": 7.389740435693826e-06}
{"train_loss": 0.04668084904551506, "global_step": 223962, "epoch": 2516, "lr": 7.389437115895104e-06}
{"train_loss": 0.06770830601453781, "global_step": 223963, "epoch": 2516, "lr": 7.38913380182486e-06}
{"train_loss": 0.07326850295066833, "global_step": 223964, "epoch": 2516, "lr": 7.388830493483118e-06}
{"train_loss": 0.0581391341984272, "global_step": 223965, "epoch": 2516, "lr": 7.388527190869943e-06}
{"train_loss": 0.051774948835372925, "global_step": 223966, "epoch": 2516, "lr": 7.388223893985347e-06}
{"train_loss": 0.04839416965842247, "global_step": 223967, "epoch": 2516, "lr": 7.387920602829396e-06}
{"train_loss": 0.08828162401914597, "global_step": 223968, "epoch": 2516, "lr": 7.387617317402107e-06}
{"train_loss": 0.060859233140945435, "global_step": 223969, "epoch": 2516, "lr": 7.387314037703547e-06}
{"train_loss": 0.03903855010867119, "global_step": 223970, "epoch": 2516, "lr": 7.387010763733731e-06}
{"train_loss": 0.04854697361588478, "global_step": 223971, "epoch": 2516, "lr": 7.386707495492728e-06}
{"train_loss": 0.02948012761771679, "global_step": 223972, "epoch": 2516, "lr": 7.3864042329805595e-06}
{"train_loss": 0.042993828654289246, "global_step": 223973, "epoch": 2516, "lr": 7.3861009761972575e-06}
{"train_loss": 0.08840732276439667, "global_step": 223974, "epoch": 2516, "lr": 7.385797725142884e-06}
{"train_loss": 0.02791444957256317, "global_step": 223975, "epoch": 2516, "lr": 7.385494479817462e-06}
{"train_loss": 0.014096301048994064, "global_step": 223976, "epoch": 2516, "lr": 7.385191240221051e-06}
{"train_loss": 0.03733304515480995, "global_step": 223977, "epoch": 2516, "lr": 7.3848880063536686e-06}
{"train_loss": 0.023724844679236412, "global_step": 223978, "epoch": 2516, "lr": 7.384584778215386e-06}
{"train_loss": 0.04242275282740593, "global_step": 223979, "epoch": 2516, "lr": 7.384281555806211e-06}
{"train_loss": 0.046459343284368515, "global_step": 223980, "epoch": 2516, "lr": 7.383978339126213e-06}
{"train_loss": 0.047147925943136215, "global_step": 223981, "epoch": 2516, "lr": 7.383675128175405e-06}
{"train_loss": 0.07324935495853424, "global_step": 223982, "epoch": 2516, "lr": 7.383371922953858e-06}
{"train_loss": 0.046624183654785156, "global_step": 223983, "epoch": 2516, "lr": 7.383068723461584e-06}
{"train_loss": 0.02603246085345745, "global_step": 223984, "epoch": 2516, "lr": 7.3827655296986434e-06}
{"train_loss": 0.10319662094116211, "global_step": 223985, "epoch": 2516, "lr": 7.382462341665064e-06}
{"train_loss": 0.04398511350154877, "global_step": 223986, "epoch": 2516, "lr": 7.382159159360907e-06}
{"train_loss": 0.04369954392313957, "global_step": 223987, "epoch": 2516, "lr": 7.381855982786184e-06}
{"train_loss": 0.06112564355134964, "global_step": 223988, "epoch": 2516, "lr": 7.381552811940951e-06}
{"train_loss": 0.06824922561645508, "global_step": 223989, "epoch": 2516, "lr": 7.381249646825261e-06}
{"train_loss": 0.038158636540174484, "global_step": 223990, "epoch": 2516, "lr": 7.380946487439133e-06}
{"train_loss": 0.05069863423705101, "global_step": 223991, "epoch": 2516, "lr": 7.380643333782628e-06}
{"train_loss": 0.043062638491392136, "global_step": 223992, "epoch": 2516, "lr": 7.3803401858557666e-06}
{"train_loss": 0.06992548704147339, "global_step": 223993, "epoch": 2516, "lr": 7.380037043658606e-06}
{"train_loss": 0.025945929810404778, "global_step": 223994, "epoch": 2516, "lr": 7.379733907191172e-06}
{"train_loss": 0.06952044367790222, "global_step": 223995, "epoch": 2516, "lr": 7.379430776453522e-06}
{"train_loss": 0.01503423135727644, "global_step": 223996, "epoch": 2516, "lr": 7.379127651445677e-06}
{"train_loss": 0.059776972979307175, "global_step": 223997, "epoch": 2516, "lr": 7.3788245321677e-06}
{"train_loss": 0.07599926739931107, "global_step": 223998, "epoch": 2516, "lr": 7.378521418619611e-06}
{"train_loss": 0.05311083793640137, "global_step": 223999, "epoch": 2516, "lr": 7.3782183108014715e-06}
{"train_loss": 0.06287061423063278, "global_step": 224000, "epoch": 2516, "lr": 7.3779152087132985e-06}
{"train_loss": 0.020805317908525467, "global_step": 224001, "epoch": 2516, "lr": 7.3776121123551535e-06}
{"train_loss": 0.01867722161114216, "global_step": 224002, "epoch": 2516, "lr": 7.3773090217270635e-06}
{"train_loss": 0.07044970244169235, "global_step": 224003, "epoch": 2516, "lr": 7.3770059368290845e-06}
{"train_loss": 0.047550808638334274, "global_step": 224004, "epoch": 2516, "lr": 7.376702857661233e-06}
{"train_loss": 0.025200335308909416, "global_step": 224005, "epoch": 2516, "lr": 7.376399784223581e-06}
{"train_loss": 0.031061965972185135, "global_step": 224006, "epoch": 2516, "lr": 7.3760967165161395e-06}
{"train_loss": 0.06966991722583771, "global_step": 224007, "epoch": 2516, "lr": 7.37579365453897e-06}
{"train_loss": 0.08394861966371536, "global_step": 224008, "epoch": 2516, "lr": 7.375490598292095e-06}
{"train_loss": 0.022745149210095406, "global_step": 224009, "epoch": 2516, "lr": 7.375187547775581e-06}
{"train_loss": 0.07110361754894257, "global_step": 224010, "epoch": 2516, "lr": 7.374884502989454e-06}
{"train_loss": 0.07183276861906052, "global_step": 224011, "epoch": 2516, "lr": 7.3745814639337385e-06}
{"train_loss": 0.052173988129734325, "global_step": 224012, "epoch": 2516, "lr": 7.374278430608506e-06, "val_loss": 9.054049491882324}
{"train_loss": 0.058125920593738556, "global_step": 224013, "epoch": 2517, "lr": 7.373975403013767e-06}
{"train_loss": 0.04755481332540512, "global_step": 224014, "epoch": 2517, "lr": 7.3736723811495935e-06}
{"train_loss": 0.059205230325460434, "global_step": 224015, "epoch": 2517, "lr": 7.373369365015992e-06}
{"train_loss": 0.04319893568754196, "global_step": 224016, "epoch": 2517, "lr": 7.3730663546130396e-06}
{"train_loss": 0.0438913069665432, "global_step": 224017, "epoch": 2517, "lr": 7.372763349940743e-06}
{"train_loss": 0.04229782521724701, "global_step": 224018, "epoch": 2517, "lr": 7.372460350999172e-06}
{"train_loss": 0.029942786321043968, "global_step": 224019, "epoch": 2517, "lr": 7.372157357788345e-06}
{"train_loss": 0.05700269713997841, "global_step": 224020, "epoch": 2517, "lr": 7.371854370308323e-06}
{"train_loss": 0.07367202639579773, "global_step": 224021, "epoch": 2517, "lr": 7.371551388559122e-06}
{"train_loss": 0.049042534083127975, "global_step": 224022, "epoch": 2517, "lr": 7.371248412540804e-06}
{"train_loss": 0.08133039623498917, "global_step": 224023, "epoch": 2517, "lr": 7.370945442253396e-06}
{"train_loss": 0.030334411188960075, "global_step": 224024, "epoch": 2517, "lr": 7.370642477696943e-06}
{"train_loss": 0.07006888091564178, "global_step": 224025, "epoch": 2517, "lr": 7.3703395188714995e-06}
{"train_loss": 0.03138076886534691, "global_step": 224026, "epoch": 2517, "lr": 7.370036565777083e-06}
{"train_loss": 0.10913390666246414, "global_step": 224027, "epoch": 2517, "lr": 7.36973361841376e-06}
{"train_loss": 0.03179727867245674, "global_step": 224028, "epoch": 2517, "lr": 7.369430676781541e-06}
{"train_loss": 0.03888404741883278, "global_step": 224029, "epoch": 2517, "lr": 7.369127740880494e-06}
{"train_loss": 0.022259730845689774, "global_step": 224030, "epoch": 2517, "lr": 7.368824810710639e-06}
{"train_loss": 0.04017224162817001, "global_step": 224031, "epoch": 2517, "lr": 7.368521886272034e-06}
{"train_loss": 0.04774424433708191, "global_step": 224032, "epoch": 2517, "lr": 7.368218967564705e-06}
{"train_loss": 0.025339221581816673, "global_step": 224033, "epoch": 2517, "lr": 7.3679160545887084e-06}
{"train_loss": 0.0291017796844244, "global_step": 224034, "epoch": 2517, "lr": 7.3676131473440605e-06}
{"train_loss": 0.07252209633588791, "global_step": 224035, "epoch": 2517, "lr": 7.367310245830833e-06}
{"train_loss": 0.04117129370570183, "global_step": 224036, "epoch": 2517, "lr": 7.367007350049038e-06}
{"train_loss": 0.08023730665445328, "global_step": 224037, "epoch": 2517, "lr": 7.366704459998741e-06}
{"train_loss": 0.039520204067230225, "global_step": 224038, "epoch": 2517, "lr": 7.366401575679966e-06}
{"train_loss": 0.02952755242586136, "global_step": 224039, "epoch": 2517, "lr": 7.366098697092761e-06}
{"train_loss": 0.03699931129813194, "global_step": 224040, "epoch": 2517, "lr": 7.365795824237154e-06}
{"train_loss": 0.033509720116853714, "global_step": 224041, "epoch": 2517, "lr": 7.3654929571132015e-06}
{"train_loss": 0.06563645601272583, "global_step": 224042, "epoch": 2517, "lr": 7.365190095720947e-06}
{"train_loss": 0.03347582370042801, "global_step": 224043, "epoch": 2517, "lr": 7.3648872400604255e-06}
{"train_loss": 0.07664899528026581, "global_step": 224044, "epoch": 2517, "lr": 7.364584390131657e-06}
{"train_loss": 0.028830209746956825, "global_step": 224045, "epoch": 2517, "lr": 7.3642815459347206e-06}
{"train_loss": 0.045266032218933105, "global_step": 224046, "epoch": 2517, "lr": 7.3639787074696206e-06}
{"train_loss": 0.03848660737276077, "global_step": 224047, "epoch": 2517, "lr": 7.3636758747364255e-06}
{"train_loss": 0.07267981022596359, "global_step": 224048, "epoch": 2517, "lr": 7.363373047735167e-06}
{"train_loss": 0.014476323500275612, "global_step": 224049, "epoch": 2517, "lr": 7.363070226465868e-06}
{"train_loss": 0.026955928653478622, "global_step": 224050, "epoch": 2517, "lr": 7.362767410928595e-06}
{"train_loss": 0.04401089996099472, "global_step": 224051, "epoch": 2517, "lr": 7.36246460112337e-06}
{"train_loss": 0.0910206288099289, "global_step": 224052, "epoch": 2517, "lr": 7.362161797050254e-06}
{"train_loss": 0.030091729015111923, "global_step": 224053, "epoch": 2517, "lr": 7.361858998709264e-06}
{"train_loss": 0.053619422018527985, "global_step": 224054, "epoch": 2517, "lr": 7.361556206100462e-06}
{"train_loss": 0.05204286426305771, "global_step": 224055, "epoch": 2517, "lr": 7.361253419223873e-06}
{"train_loss": 0.05866074189543724, "global_step": 224056, "epoch": 2517, "lr": 7.36095063807955e-06}
{"train_loss": 0.04222356155514717, "global_step": 224057, "epoch": 2517, "lr": 7.360647862667519e-06}
{"train_loss": 0.04392281919717789, "global_step": 224058, "epoch": 2517, "lr": 7.360345092987836e-06}
{"train_loss": 0.02278793603181839, "global_step": 224059, "epoch": 2517, "lr": 7.360042329040528e-06}
{"train_loss": 0.06126116216182709, "global_step": 224060, "epoch": 2517, "lr": 7.359739570825641e-06}
{"train_loss": 0.02992970123887062, "global_step": 224061, "epoch": 2517, "lr": 7.359436818343229e-06}
{"train_loss": 0.03703918680548668, "global_step": 224062, "epoch": 2517, "lr": 7.359134071593304e-06}
{"train_loss": 0.057394109666347504, "global_step": 224063, "epoch": 2517, "lr": 7.358831330575944e-06}
{"train_loss": 0.034792207181453705, "global_step": 224064, "epoch": 2517, "lr": 7.358528595291153e-06}
{"train_loss": 0.05729394778609276, "global_step": 224065, "epoch": 2517, "lr": 7.358225865738999e-06}
{"train_loss": 0.03236273676156998, "global_step": 224066, "epoch": 2517, "lr": 7.357923141919504e-06}
{"train_loss": 0.05259568244218826, "global_step": 224067, "epoch": 2517, "lr": 7.357620423832729e-06}
{"train_loss": 0.0814489796757698, "global_step": 224068, "epoch": 2517, "lr": 7.357317711478684e-06}
{"train_loss": 0.03549247980117798, "global_step": 224069, "epoch": 2517, "lr": 7.357015004857443e-06}
{"train_loss": 0.06051217019557953, "global_step": 224070, "epoch": 2517, "lr": 7.3567123039690214e-06}
{"train_loss": 0.05081379786133766, "global_step": 224071, "epoch": 2517, "lr": 7.356409608813481e-06}
{"train_loss": 0.06667152792215347, "global_step": 224072, "epoch": 2517, "lr": 7.356106919390843e-06}
{"train_loss": 0.04736338183283806, "global_step": 224073, "epoch": 2517, "lr": 7.355804235701164e-06}
{"train_loss": 0.04690191522240639, "global_step": 224074, "epoch": 2517, "lr": 7.3555015577444655e-06}
{"train_loss": 0.035209327936172485, "global_step": 224075, "epoch": 2517, "lr": 7.355198885520814e-06}
{"train_loss": 0.02637823112308979, "global_step": 224076, "epoch": 2517, "lr": 7.354896219030222e-06}
{"train_loss": 0.05353837087750435, "global_step": 224077, "epoch": 2517, "lr": 7.354593558272754e-06}
{"train_loss": 0.028088614344596863, "global_step": 224078, "epoch": 2517, "lr": 7.3542909032484345e-06}
{"train_loss": 0.05807735398411751, "global_step": 224079, "epoch": 2517, "lr": 7.353988253957311e-06}
{"train_loss": 0.04395357519388199, "global_step": 224080, "epoch": 2517, "lr": 7.353685610399436e-06}
{"train_loss": 0.04117121547460556, "global_step": 224081, "epoch": 2517, "lr": 7.353382972574824e-06}
{"train_loss": 0.03669791296124458, "global_step": 224082, "epoch": 2517, "lr": 7.353080340483537e-06}
{"train_loss": 0.05060899630188942, "global_step": 224083, "epoch": 2517, "lr": 7.352777714125614e-06}
{"train_loss": 0.02864338643848896, "global_step": 224084, "epoch": 2517, "lr": 7.352475093501082e-06}
{"train_loss": 0.03359466791152954, "global_step": 224085, "epoch": 2517, "lr": 7.3521724786099976e-06}
{"train_loss": 0.0433315746486187, "global_step": 224086, "epoch": 2517, "lr": 7.351869869452382e-06}
{"train_loss": 0.029629971832036972, "global_step": 224087, "epoch": 2517, "lr": 7.351567266028297e-06}
{"train_loss": 0.02426268719136715, "global_step": 224088, "epoch": 2517, "lr": 7.351264668337776e-06}
{"train_loss": 0.06913286447525024, "global_step": 224089, "epoch": 2517, "lr": 7.350962076380846e-06}
{"train_loss": 0.04175937548279762, "global_step": 224090, "epoch": 2517, "lr": 7.350659490157574e-06}
{"train_loss": 0.034860141575336456, "global_step": 224091, "epoch": 2517, "lr": 7.350356909667971e-06}
{"train_loss": 0.02559839002788067, "global_step": 224092, "epoch": 2517, "lr": 7.3500543349121044e-06}
{"train_loss": 0.04732488468289375, "global_step": 224093, "epoch": 2517, "lr": 7.349751765889995e-06}
{"train_loss": 0.045495130121707916, "global_step": 224094, "epoch": 2517, "lr": 7.349449202601699e-06}
{"train_loss": 0.06301490217447281, "global_step": 224095, "epoch": 2517, "lr": 7.349146645047239e-06}
{"train_loss": 0.04164334014058113, "global_step": 224096, "epoch": 2517, "lr": 7.348844093226665e-06}
{"train_loss": 0.04666157066822052, "global_step": 224097, "epoch": 2517, "lr": 7.348541547140037e-06}
{"train_loss": 0.04349112883210182, "global_step": 224098, "epoch": 2517, "lr": 7.348239006787361e-06}
{"train_loss": 0.05431031435728073, "global_step": 224099, "epoch": 2517, "lr": 7.347936472168704e-06}
{"train_loss": 0.0952775627374649, "global_step": 224100, "epoch": 2517, "lr": 7.347633943284088e-06}
{"train_loss": 0.047289397358224634, "global_step": 224101, "epoch": 2517, "lr": 7.347331420133574e-06, "val_loss": 9.022857666015625}
{"train_loss": 0.0421607531607151, "global_step": 224102, "epoch": 2518, "lr": 7.347028902717185e-06}
{"train_loss": 0.04147270321846008, "global_step": 224103, "epoch": 2518, "lr": 7.346726391034975e-06}
{"train_loss": 0.03910677880048752, "global_step": 224104, "epoch": 2518, "lr": 7.346423885086967e-06}
{"train_loss": 0.04287787154316902, "global_step": 224105, "epoch": 2518, "lr": 7.346121384873223e-06}
{"train_loss": 0.03368667513132095, "global_step": 224106, "epoch": 2518, "lr": 7.345818890393757e-06}
{"train_loss": 0.06622931361198425, "global_step": 224107, "epoch": 2518, "lr": 7.345516401648644e-06}
{"train_loss": 0.07759303599596024, "global_step": 224108, "epoch": 2518, "lr": 7.345213918637894e-06}
{"train_loss": 0.07477477937936783, "global_step": 224109, "epoch": 2518, "lr": 7.344911441361574e-06}
{"train_loss": 0.03790447860956192, "global_step": 224110, "epoch": 2518, "lr": 7.344608969819694e-06}
{"train_loss": 0.09024427831172943, "global_step": 224111, "epoch": 2518, "lr": 7.344306504012327e-06}
{"train_loss": 0.051908206194639206, "global_step": 224112, "epoch": 2518, "lr": 7.344004043939484e-06}
{"train_loss": 0.04304329305887222, "global_step": 224113, "epoch": 2518, "lr": 7.343701589601232e-06}
{"train_loss": 0.03198585286736488, "global_step": 224114, "epoch": 2518, "lr": 7.343399140997587e-06}
{"train_loss": 0.06257546693086624, "global_step": 224115, "epoch": 2518, "lr": 7.343096698128599e-06}
{"train_loss": 0.047521114349365234, "global_step": 224116, "epoch": 2518, "lr": 7.342794260994329e-06}
{"train_loss": 0.07062780857086182, "global_step": 224117, "epoch": 2518, "lr": 7.3424918295947886e-06}
{"train_loss": 0.06585278362035751, "global_step": 224118, "epoch": 2518, "lr": 7.342189403930038e-06}
{"train_loss": 0.03284779563546181, "global_step": 224119, "epoch": 2518, "lr": 7.341886984000102e-06}
{"train_loss": 0.044617462903261185, "global_step": 224120, "epoch": 2518, "lr": 7.3415845698050374e-06}
{"train_loss": 0.031239433214068413, "global_step": 224121, "epoch": 2518, "lr": 7.341282161344882e-06}
{"train_loss": 0.0829579085111618, "global_step": 224122, "epoch": 2518, "lr": 7.340979758619654e-06}
{"train_loss": 0.05322134867310524, "global_step": 224123, "epoch": 2518, "lr": 7.340677361629428e-06}
{"train_loss": 0.03791235014796257, "global_step": 224124, "epoch": 2518, "lr": 7.340374970374214e-06}
{"train_loss": 0.05094866082072258, "global_step": 224125, "epoch": 2518, "lr": 7.340072584854079e-06}
{"train_loss": 0.07216473668813705, "global_step": 224126, "epoch": 2518, "lr": 7.3397702050690466e-06}
{"train_loss": 0.07367170602083206, "global_step": 224127, "epoch": 2518, "lr": 7.339467831019159e-06}
{"train_loss": 0.08014404773712158, "global_step": 224128, "epoch": 2518, "lr": 7.339165462704467e-06}
{"train_loss": 0.03388768807053566, "global_step": 224129, "epoch": 2518, "lr": 7.338863100124993e-06}
{"train_loss": 0.04568492993712425, "global_step": 224130, "epoch": 2518, "lr": 7.338560743280798e-06}
{"train_loss": 0.061073511838912964, "global_step": 224131, "epoch": 2518, "lr": 7.338258392171904e-06}
{"train_loss": 0.05394919216632843, "global_step": 224132, "epoch": 2518, "lr": 7.337956046798361e-06}
{"train_loss": 0.06312578171491623, "global_step": 224133, "epoch": 2518, "lr": 7.337653707160225e-06}
{"train_loss": 0.05945615842938423, "global_step": 224134, "epoch": 2518, "lr": 7.3373513732575064e-06}
{"train_loss": 0.021761896088719368, "global_step": 224135, "epoch": 2518, "lr": 7.337049045090272e-06}
{"train_loss": 0.06659752875566483, "global_step": 224136, "epoch": 2518, "lr": 7.336746722658539e-06}
{"train_loss": 0.07289065420627594, "global_step": 224137, "epoch": 2518, "lr": 7.336444405962372e-06}
{"train_loss": 0.08794566243886948, "global_step": 224138, "epoch": 2518, "lr": 7.336142095001791e-06}
{"train_loss": 0.02733965776860714, "global_step": 224139, "epoch": 2518, "lr": 7.335839789776855e-06}
{"train_loss": 0.04165574163198471, "global_step": 224140, "epoch": 2518, "lr": 7.335537490287586e-06}
{"train_loss": 0.037860509008169174, "global_step": 224141, "epoch": 2518, "lr": 7.335235196534046e-06}
{"train_loss": 0.023682506754994392, "global_step": 224142, "epoch": 2518, "lr": 7.334932908516251e-06}
{"train_loss": 0.03161581978201866, "global_step": 224143, "epoch": 2518, "lr": 7.334630626234262e-06}
{"train_loss": 0.04174123331904411, "global_step": 224144, "epoch": 2518, "lr": 7.334328349688108e-06}
{"train_loss": 0.021139847114682198, "global_step": 224145, "epoch": 2518, "lr": 7.334026078877837e-06}
{"train_loss": 0.0887625515460968, "global_step": 224146, "epoch": 2518, "lr": 7.333723813803478e-06}
{"train_loss": 0.0854988843202591, "global_step": 224147, "epoch": 2518, "lr": 7.333421554465092e-06}
{"train_loss": 0.05079968273639679, "global_step": 224148, "epoch": 2518, "lr": 7.333119300862701e-06}
{"train_loss": 0.05990707129240036, "global_step": 224149, "epoch": 2518, "lr": 7.332817052996355e-06}
{"train_loss": 0.028667867183685303, "global_step": 224150, "epoch": 2518, "lr": 7.332514810866087e-06}
{"train_loss": 0.09142275154590607, "global_step": 224151, "epoch": 2518, "lr": 7.332212574471936e-06}
{"train_loss": 0.023035261780023575, "global_step": 224152, "epoch": 2518, "lr": 7.331910343813964e-06}
{"train_loss": 0.0900062546133995, "global_step": 224153, "epoch": 2518, "lr": 7.331608118892186e-06}
{"train_loss": 0.029119186103343964, "global_step": 224154, "epoch": 2518, "lr": 7.331305899706658e-06}
{"train_loss": 0.04418673366308212, "global_step": 224155, "epoch": 2518, "lr": 7.331003686257415e-06}
{"train_loss": 0.0605303980410099, "global_step": 224156, "epoch": 2518, "lr": 7.3307014785445046e-06}
{"train_loss": 0.05053550377488136, "global_step": 224157, "epoch": 2518, "lr": 7.330399276567945e-06}
{"train_loss": 0.049704063683748245, "global_step": 224158, "epoch": 2518, "lr": 7.3300970803278135e-06}
{"train_loss": 0.05531516671180725, "global_step": 224159, "epoch": 2518, "lr": 7.329794889824127e-06}
{"train_loss": 0.07757292687892914, "global_step": 224160, "epoch": 2518, "lr": 7.329492705056917e-06}
{"train_loss": 0.0285805556923151, "global_step": 224161, "epoch": 2518, "lr": 7.329190526026248e-06}
{"train_loss": 0.028875211253762245, "global_step": 224162, "epoch": 2518, "lr": 7.328888352732138e-06}
{"train_loss": 0.016841866075992584, "global_step": 224163, "epoch": 2518, "lr": 7.328586185174652e-06}
{"train_loss": 0.06655076891183853, "global_step": 224164, "epoch": 2518, "lr": 7.328284023353815e-06}
{"train_loss": 0.07440169900655746, "global_step": 224165, "epoch": 2518, "lr": 7.327981867269657e-06}
{"train_loss": 0.023927129805088043, "global_step": 224166, "epoch": 2518, "lr": 7.327679716922248e-06}
{"train_loss": 0.06582801043987274, "global_step": 224167, "epoch": 2518, "lr": 7.3273775723116004e-06}
{"train_loss": 0.04373479261994362, "global_step": 224168, "epoch": 2518, "lr": 7.327075433437763e-06}
{"train_loss": 0.021391192451119423, "global_step": 224169, "epoch": 2518, "lr": 7.326773300300793e-06}
{"train_loss": 0.06819687783718109, "global_step": 224170, "epoch": 2518, "lr": 7.326471172900706e-06}
{"train_loss": 0.09362468868494034, "global_step": 224171, "epoch": 2518, "lr": 7.326169051237569e-06}
{"train_loss": 0.05052981153130531, "global_step": 224172, "epoch": 2518, "lr": 7.325866935311399e-06}
{"train_loss": 0.032469816505908966, "global_step": 224173, "epoch": 2518, "lr": 7.32556482512225e-06}
{"train_loss": 0.00814530998468399, "global_step": 224174, "epoch": 2518, "lr": 7.325262720670151e-06}
{"train_loss": 0.05617474764585495, "global_step": 224175, "epoch": 2518, "lr": 7.324960621955162e-06}
{"train_loss": 0.05805708095431328, "global_step": 224176, "epoch": 2518, "lr": 7.324658528977301e-06}
{"train_loss": 0.08282215893268585, "global_step": 224177, "epoch": 2518, "lr": 7.324356441736635e-06}
{"train_loss": 0.08622424304485321, "global_step": 224178, "epoch": 2518, "lr": 7.324054360233173e-06}
{"train_loss": 0.06590135395526886, "global_step": 224179, "epoch": 2518, "lr": 7.323752284466984e-06}
{"train_loss": 0.015776697546243668, "global_step": 224180, "epoch": 2518, "lr": 7.323450214438082e-06}
{"train_loss": 0.030429475009441376, "global_step": 224181, "epoch": 2518, "lr": 7.323148150146536e-06}
{"train_loss": 0.07449597120285034, "global_step": 224182, "epoch": 2518, "lr": 7.3228460915923615e-06}
{"train_loss": 0.055581234395504, "global_step": 224183, "epoch": 2518, "lr": 7.322544038775619e-06}
{"train_loss": 0.037292271852493286, "global_step": 224184, "epoch": 2518, "lr": 7.322241991696333e-06}
{"train_loss": 0.05513845756649971, "global_step": 224185, "epoch": 2518, "lr": 7.321939950354562e-06}
{"train_loss": 0.0312182679772377, "global_step": 224186, "epoch": 2518, "lr": 7.321637914750323e-06}
{"train_loss": 0.04663184657692909, "global_step": 224187, "epoch": 2518, "lr": 7.321335884883673e-06}
{"train_loss": 0.060883183032274246, "global_step": 224188, "epoch": 2518, "lr": 7.321033860754656e-06}
{"train_loss": 0.04683065041899681, "global_step": 224189, "epoch": 2518, "lr": 7.3207318423632934e-06}
{"train_loss": 0.052013398849227455, "global_step": 224190, "epoch": 2518, "lr": 7.320429829709652e-06, "val_loss": 9.064764022827148}
{"train_loss": 0.0769730731844902, "global_step": 224191, "epoch": 2519, "lr": 7.3201278227937486e-06}
{"train_loss": 0.030826076865196228, "global_step": 224192, "epoch": 2519, "lr": 7.319825821615645e-06}
{"train_loss": 0.035858217626810074, "global_step": 224193, "epoch": 2519, "lr": 7.319523826175356e-06}
{"train_loss": 0.017315806820988655, "global_step": 224194, "epoch": 2519, "lr": 7.319221836472956e-06}
{"train_loss": 0.05340546742081642, "global_step": 224195, "epoch": 2519, "lr": 7.318919852508449e-06}
{"train_loss": 0.01859425939619541, "global_step": 224196, "epoch": 2519, "lr": 7.318617874281903e-06}
{"train_loss": 0.03419332578778267, "global_step": 224197, "epoch": 2519, "lr": 7.318315901793338e-06}
{"train_loss": 0.1239701434969902, "global_step": 224198, "epoch": 2519, "lr": 7.318013935042817e-06}
{"train_loss": 0.023449672386050224, "global_step": 224199, "epoch": 2519, "lr": 7.317711974030372e-06}
{"train_loss": 0.04176962003111839, "global_step": 224200, "epoch": 2519, "lr": 7.317410018756027e-06}
{"train_loss": 0.07975395768880844, "global_step": 224201, "epoch": 2519, "lr": 7.317108069219847e-06}
{"train_loss": 0.03510815650224686, "global_step": 224202, "epoch": 2519, "lr": 7.316806125421849e-06}
{"train_loss": 0.06296443194150925, "global_step": 224203, "epoch": 2519, "lr": 7.3165041873621e-06}
{"train_loss": 0.07836907356977463, "global_step": 224204, "epoch": 2519, "lr": 7.316202255040628e-06}
{"train_loss": 0.04907522350549698, "global_step": 224205, "epoch": 2519, "lr": 7.315900328457459e-06}
{"train_loss": 0.04135408625006676, "global_step": 224206, "epoch": 2519, "lr": 7.315598407612645e-06}
{"train_loss": 0.07327210158109665, "global_step": 224207, "epoch": 2519, "lr": 7.315296492506241e-06}
{"train_loss": 0.08291992545127869, "global_step": 224208, "epoch": 2519, "lr": 7.314994583138263e-06}
{"train_loss": 0.05290606617927551, "global_step": 224209, "epoch": 2519, "lr": 7.314692679508778e-06}
{"train_loss": 0.11136443167924881, "global_step": 224210, "epoch": 2519, "lr": 7.314390781617797e-06}
{"train_loss": 0.043847646564245224, "global_step": 224211, "epoch": 2519, "lr": 7.314088889465393e-06}
{"train_loss": 0.031215226277709007, "global_step": 224212, "epoch": 2519, "lr": 7.31378700305157e-06}
{"train_loss": 0.028252238407731056, "global_step": 224213, "epoch": 2519, "lr": 7.313485122376407e-06}
{"train_loss": 0.05467686802148819, "global_step": 224214, "epoch": 2519, "lr": 7.313183247439909e-06}
{"train_loss": 0.0658237487077713, "global_step": 224215, "epoch": 2519, "lr": 7.312881378242148e-06}
{"train_loss": 0.05594610422849655, "global_step": 224216, "epoch": 2519, "lr": 7.3125795147831354e-06}
{"train_loss": 0.022125959396362305, "global_step": 224217, "epoch": 2519, "lr": 7.312277657062944e-06}
{"train_loss": 0.03935229405760765, "global_step": 224218, "epoch": 2519, "lr": 7.311975805081578e-06}
{"train_loss": 0.054100994020700455, "global_step": 224219, "epoch": 2519, "lr": 7.31167395883911e-06}
{"train_loss": 0.040087051689624786, "global_step": 224220, "epoch": 2519, "lr": 7.311372118335558e-06}
{"train_loss": 0.0183135736733675, "global_step": 224221, "epoch": 2519, "lr": 7.311070283570981e-06}
{"train_loss": 0.028904281556606293, "global_step": 224222, "epoch": 2519, "lr": 7.3107684545454025e-06}
{"train_loss": 0.05149627476930618, "global_step": 224223, "epoch": 2519, "lr": 7.310466631258866e-06}
{"train_loss": 0.0956161618232727, "global_step": 224224, "epoch": 2519, "lr": 7.310164813711429e-06}
{"train_loss": 0.018876101821660995, "global_step": 224225, "epoch": 2519, "lr": 7.309863001903111e-06}
{"train_loss": 0.05796465277671814, "global_step": 224226, "epoch": 2519, "lr": 7.309561195833969e-06}
{"train_loss": 0.048310860991477966, "global_step": 224227, "epoch": 2519, "lr": 7.309259395504031e-06}
{"train_loss": 0.044018372893333435, "global_step": 224228, "epoch": 2519, "lr": 7.308957600913352e-06}
{"train_loss": 0.030871300026774406, "global_step": 224229, "epoch": 2519, "lr": 7.308655812061949e-06}
{"train_loss": 0.07881053537130356, "global_step": 224230, "epoch": 2519, "lr": 7.308354028949887e-06}
{"train_loss": 0.030823394656181335, "global_step": 224231, "epoch": 2519, "lr": 7.308052251577191e-06}
{"train_loss": 0.04989711567759514, "global_step": 224232, "epoch": 2519, "lr": 7.307750479943914e-06}
{"train_loss": 0.06585513055324554, "global_step": 224233, "epoch": 2519, "lr": 7.30744871405008e-06}
{"train_loss": 0.03163512051105499, "global_step": 224234, "epoch": 2519, "lr": 7.3071469538957495e-06}
{"train_loss": 0.0920657217502594, "global_step": 224235, "epoch": 2519, "lr": 7.306845199480944e-06}
{"train_loss": 0.0575731098651886, "global_step": 224236, "epoch": 2519, "lr": 7.30654345080572e-06}
{"train_loss": 0.043547697365283966, "global_step": 224237, "epoch": 2519, "lr": 7.306241707870115e-06}
{"train_loss": 0.06474839150905609, "global_step": 224238, "epoch": 2519, "lr": 7.305939970674152e-06}
{"train_loss": 0.02643457055091858, "global_step": 224239, "epoch": 2519, "lr": 7.305638239217899e-06}
{"train_loss": 0.06908881664276123, "global_step": 224240, "epoch": 2519, "lr": 7.3053365135013655e-06}
{"train_loss": 0.022526416927576065, "global_step": 224241, "epoch": 2519, "lr": 7.305034793524623e-06}
{"train_loss": 0.02032735012471676, "global_step": 224242, "epoch": 2519, "lr": 7.30473307928769e-06}
{"train_loss": 0.0545145720243454, "global_step": 224243, "epoch": 2519, "lr": 7.3044313707906205e-06}
{"train_loss": 0.05553586035966873, "global_step": 224244, "epoch": 2519, "lr": 7.304129668033444e-06}
{"train_loss": 0.03373304754495621, "global_step": 224245, "epoch": 2519, "lr": 7.303827971016219e-06}
{"train_loss": 0.032122593373060226, "global_step": 224246, "epoch": 2519, "lr": 7.303526279738959e-06}
{"train_loss": 0.0987052172422409, "global_step": 224247, "epoch": 2519, "lr": 7.30322459420173e-06}
{"train_loss": 0.04545452445745468, "global_step": 224248, "epoch": 2519, "lr": 7.302922914404553e-06}
{"train_loss": 0.08860331028699875, "global_step": 224249, "epoch": 2519, "lr": 7.30262124034749e-06}
{"train_loss": 0.048786260187625885, "global_step": 224250, "epoch": 2519, "lr": 7.302319572030558e-06}
{"train_loss": 0.05328093096613884, "global_step": 224251, "epoch": 2519, "lr": 7.302017909453818e-06}
{"train_loss": 0.06714597344398499, "global_step": 224252, "epoch": 2519, "lr": 7.301716252617285e-06}
{"train_loss": 0.045416928827762604, "global_step": 224253, "epoch": 2519, "lr": 7.301414601521034e-06}
{"train_loss": 0.02321426384150982, "global_step": 224254, "epoch": 2519, "lr": 7.3011129561650734e-06}
{"train_loss": 0.028354419395327568, "global_step": 224255, "epoch": 2519, "lr": 7.3008113165494716e-06}
{"train_loss": 0.03478815034031868, "global_step": 224256, "epoch": 2519, "lr": 7.300509682674245e-06}
{"train_loss": 0.02795574441552162, "global_step": 224257, "epoch": 2519, "lr": 7.300208054539454e-06}
{"train_loss": 0.03766512870788574, "global_step": 224258, "epoch": 2519, "lr": 7.299906432145115e-06}
{"train_loss": 0.05220484361052513, "global_step": 224259, "epoch": 2519, "lr": 7.299604815491285e-06}
{"train_loss": 0.07280129194259644, "global_step": 224260, "epoch": 2519, "lr": 7.299303204578018e-06}
{"train_loss": 0.04151448234915733, "global_step": 224261, "epoch": 2519, "lr": 7.299001599405325e-06}
{"train_loss": 0.058107201009988785, "global_step": 224262, "epoch": 2519, "lr": 7.298699999973269e-06}
{"train_loss": 0.05617667734622955, "global_step": 224263, "epoch": 2519, "lr": 7.2983984062818754e-06}
{"train_loss": 0.02145451493561268, "global_step": 224264, "epoch": 2519, "lr": 7.2980968183312015e-06}
{"train_loss": 0.08369716256856918, "global_step": 224265, "epoch": 2519, "lr": 7.297795236121263e-06}
{"train_loss": 0.029144326224923134, "global_step": 224266, "epoch": 2519, "lr": 7.297493659652132e-06}
{"train_loss": 0.10311183333396912, "global_step": 224267, "epoch": 2519, "lr": 7.297192088923821e-06}
{"train_loss": 0.06177406385540962, "global_step": 224268, "epoch": 2519, "lr": 7.296890523936389e-06}
{"train_loss": 0.047101736068725586, "global_step": 224269, "epoch": 2519, "lr": 7.296588964689866e-06}
{"train_loss": 0.06993710994720459, "global_step": 224270, "epoch": 2519, "lr": 7.296287411184299e-06}
{"train_loss": 0.07921281456947327, "global_step": 224271, "epoch": 2519, "lr": 7.2959858634197186e-06}
{"train_loss": 0.05951650068163872, "global_step": 224272, "epoch": 2519, "lr": 7.295684321396185e-06}
{"train_loss": 0.06628954410552979, "global_step": 224273, "epoch": 2519, "lr": 7.2953827851137135e-06}
{"train_loss": 0.05212016776204109, "global_step": 224274, "epoch": 2519, "lr": 7.295081254572367e-06}
{"train_loss": 0.059099797159433365, "global_step": 224275, "epoch": 2519, "lr": 7.294779729772178e-06}
{"train_loss": 0.02958560362458229, "global_step": 224276, "epoch": 2519, "lr": 7.2944782107131735e-06}
{"train_loss": 0.03155034780502319, "global_step": 224277, "epoch": 2519, "lr": 7.294176697395416e-06}
{"train_loss": 0.023484082892537117, "global_step": 224278, "epoch": 2519, "lr": 7.293875189818927e-06}
{"train_loss": 0.05065290041770158, "global_step": 224279, "epoch": 2519, "lr": 7.293573687983763e-06, "val_loss": 9.001571655273438}
{"train_loss": 0.0535922534763813, "global_step": 224280, "epoch": 2520, "lr": 7.293272191889949e-06}
{"train_loss": 0.035062115639448166, "global_step": 224281, "epoch": 2520, "lr": 7.292970701537549e-06}
{"train_loss": 0.0856529027223587, "global_step": 224282, "epoch": 2520, "lr": 7.292669216926573e-06}
{"train_loss": 0.0657900869846344, "global_step": 224283, "epoch": 2520, "lr": 7.292367738057088e-06}
{"train_loss": 0.04118570312857628, "global_step": 224284, "epoch": 2520, "lr": 7.292066264929115e-06}
{"train_loss": 0.02748691849410534, "global_step": 224285, "epoch": 2520, "lr": 7.29176479754271e-06}
{"train_loss": 0.03152221068739891, "global_step": 224286, "epoch": 2520, "lr": 7.2914633358979024e-06}
{"train_loss": 0.06430409848690033, "global_step": 224287, "epoch": 2520, "lr": 7.29116187999474e-06}
{"train_loss": 0.04846885800361633, "global_step": 224288, "epoch": 2520, "lr": 7.290860429833257e-06}
{"train_loss": 0.0586581751704216, "global_step": 224289, "epoch": 2520, "lr": 7.290558985413504e-06}
{"train_loss": 0.04333021491765976, "global_step": 224290, "epoch": 2520, "lr": 7.290257546735502e-06}
{"train_loss": 0.051710568368434906, "global_step": 224291, "epoch": 2520, "lr": 7.289956113799318e-06}
{"train_loss": 0.07152300328016281, "global_step": 224292, "epoch": 2520, "lr": 7.289654686604969e-06}
{"train_loss": 0.0416751354932785, "global_step": 224293, "epoch": 2520, "lr": 7.289353265152515e-06}
{"train_loss": 0.031325433403253555, "global_step": 224294, "epoch": 2520, "lr": 7.2890518494419745e-06}
{"train_loss": 0.042558860033750534, "global_step": 224295, "epoch": 2520, "lr": 7.2887504394734015e-06}
{"train_loss": 0.044503405690193176, "global_step": 224296, "epoch": 2520, "lr": 7.288449035246847e-06}
{"train_loss": 0.03237724304199219, "global_step": 224297, "epoch": 2520, "lr": 7.288147636762327e-06}
{"train_loss": 0.043810244649648666, "global_step": 224298, "epoch": 2520, "lr": 7.287846244019908e-06}
{"train_loss": 0.02673950046300888, "global_step": 224299, "epoch": 2520, "lr": 7.287544857019607e-06}
{"train_loss": 0.07284047454595566, "global_step": 224300, "epoch": 2520, "lr": 7.287243475761485e-06}
{"train_loss": 0.05478566884994507, "global_step": 224301, "epoch": 2520, "lr": 7.286942100245564e-06}
{"train_loss": 0.035075340420007706, "global_step": 224302, "epoch": 2520, "lr": 7.2866407304719e-06}
{"train_loss": 0.026588041335344315, "global_step": 224303, "epoch": 2520, "lr": 7.28633936644052e-06}
{"train_loss": 0.040715187788009644, "global_step": 224304, "epoch": 2520, "lr": 7.28603800815148e-06}
{"train_loss": 0.08501975983381271, "global_step": 224305, "epoch": 2520, "lr": 7.285736655604797e-06}
{"train_loss": 0.035975679755210876, "global_step": 224306, "epoch": 2520, "lr": 7.285435308800542e-06}
{"train_loss": 0.03470151871442795, "global_step": 224307, "epoch": 2520, "lr": 7.285133967738722e-06}
{"train_loss": 0.022634321823716164, "global_step": 224308, "epoch": 2520, "lr": 7.284832632419414e-06}
{"train_loss": 0.023463647812604904, "global_step": 224309, "epoch": 2520, "lr": 7.284531302842623e-06}
{"train_loss": 0.06986954808235168, "global_step": 224310, "epoch": 2520, "lr": 7.284229979008422e-06}
{"train_loss": 0.03915391117334366, "global_step": 224311, "epoch": 2520, "lr": 7.2839286609168226e-06}
{"train_loss": 0.030222555622458458, "global_step": 224312, "epoch": 2520, "lr": 7.2836273485678896e-06}
{"train_loss": 0.03136134520173073, "global_step": 224313, "epoch": 2520, "lr": 7.283326041961641e-06}
{"train_loss": 0.049543239176273346, "global_step": 224314, "epoch": 2520, "lr": 7.2830247410981375e-06}
{"train_loss": 0.06175154075026512, "global_step": 224315, "epoch": 2520, "lr": 7.282723445977413e-06}
{"train_loss": 0.035997603088617325, "global_step": 224316, "epoch": 2520, "lr": 7.282422156599495e-06}
{"train_loss": 0.033603981137275696, "global_step": 224317, "epoch": 2520, "lr": 7.282120872964443e-06}
{"train_loss": 0.07372907549142838, "global_step": 224318, "epoch": 2520, "lr": 7.281819595072281e-06}
{"train_loss": 0.041665878146886826, "global_step": 224319, "epoch": 2520, "lr": 7.281518322923064e-06}
{"train_loss": 0.06352578103542328, "global_step": 224320, "epoch": 2520, "lr": 7.28121705651682e-06}
{"train_loss": 0.02386312745511532, "global_step": 224321, "epoch": 2520, "lr": 7.280915795853605e-06}
{"train_loss": 0.03905337676405907, "global_step": 224322, "epoch": 2520, "lr": 7.280614540933439e-06}
{"train_loss": 0.037602152675390244, "global_step": 224323, "epoch": 2520, "lr": 7.280313291756386e-06}
{"train_loss": 0.04064744710922241, "global_step": 224324, "epoch": 2520, "lr": 7.280012048322466e-06}
{"train_loss": 0.014940571039915085, "global_step": 224325, "epoch": 2520, "lr": 7.279710810631735e-06}
{"train_loss": 0.012938618659973145, "global_step": 224326, "epoch": 2520, "lr": 7.279409578684215e-06}
{"train_loss": 0.022937815636396408, "global_step": 224327, "epoch": 2520, "lr": 7.279108352479969e-06}
{"train_loss": 0.050817519426345825, "global_step": 224328, "epoch": 2520, "lr": 7.278807132019011e-06}
{"train_loss": 0.05934822931885719, "global_step": 224329, "epoch": 2520, "lr": 7.278505917301415e-06}
{"train_loss": 0.07748840004205704, "global_step": 224330, "epoch": 2520, "lr": 7.278204708327191e-06}
{"train_loss": 0.05683395266532898, "global_step": 224331, "epoch": 2520, "lr": 7.2779035050964005e-06}
{"train_loss": 0.05403340235352516, "global_step": 224332, "epoch": 2520, "lr": 7.27760230760906e-06}
{"train_loss": 0.057887788861989975, "global_step": 224333, "epoch": 2520, "lr": 7.277301115865232e-06}
{"train_loss": 0.05872971937060356, "global_step": 224334, "epoch": 2520, "lr": 7.276999929864958e-06}
{"train_loss": 0.04494557902216911, "global_step": 224335, "epoch": 2520, "lr": 7.2766987496082626e-06}
{"train_loss": 0.05467003583908081, "global_step": 224336, "epoch": 2520, "lr": 7.276397575095201e-06}
{"train_loss": 0.053974784910678864, "global_step": 224337, "epoch": 2520, "lr": 7.2760964063258e-06}
{"train_loss": 0.024701563641428947, "global_step": 224338, "epoch": 2520, "lr": 7.275795243300115e-06}
{"train_loss": 0.05668453499674797, "global_step": 224339, "epoch": 2520, "lr": 7.27549408601817e-06}
{"train_loss": 0.05210110545158386, "global_step": 224340, "epoch": 2520, "lr": 7.275192934480024e-06}
{"train_loss": 0.05646425113081932, "global_step": 224341, "epoch": 2520, "lr": 7.274891788685695e-06}
{"train_loss": 0.05791512131690979, "global_step": 224342, "epoch": 2520, "lr": 7.274590648635249e-06}
{"train_loss": 0.05607577785849571, "global_step": 224343, "epoch": 2520, "lr": 7.274289514328703e-06}
{"train_loss": 0.05311867594718933, "global_step": 224344, "epoch": 2520, "lr": 7.2739883857661174e-06}
{"train_loss": 0.04614454135298729, "global_step": 224345, "epoch": 2520, "lr": 7.273687262947515e-06}
{"train_loss": 0.015013111755251884, "global_step": 224346, "epoch": 2520, "lr": 7.273386145872951e-06}
{"train_loss": 0.03073129430413246, "global_step": 224347, "epoch": 2520, "lr": 7.273085034542454e-06}
{"train_loss": 0.03997988626360893, "global_step": 224348, "epoch": 2520, "lr": 7.2727839289560785e-06}
{"train_loss": 0.04520507901906967, "global_step": 224349, "epoch": 2520, "lr": 7.272482829113842e-06}
{"train_loss": 0.08803500235080719, "global_step": 224350, "epoch": 2520, "lr": 7.272181735015815e-06}
{"train_loss": 0.051395393908023834, "global_step": 224351, "epoch": 2520, "lr": 7.2718806466620105e-06}
{"train_loss": 0.038860712200403214, "global_step": 224352, "epoch": 2520, "lr": 7.271579564052489e-06}
{"train_loss": 0.03697580099105835, "global_step": 224353, "epoch": 2520, "lr": 7.271278487187283e-06}
{"train_loss": 0.025118330493569374, "global_step": 224354, "epoch": 2520, "lr": 7.270977416066427e-06}
{"train_loss": 0.05283359810709953, "global_step": 224355, "epoch": 2520, "lr": 7.270676350689976e-06}
{"train_loss": 0.041177839040756226, "global_step": 224356, "epoch": 2520, "lr": 7.270375291057946e-06}
{"train_loss": 0.0369100496172905, "global_step": 224357, "epoch": 2520, "lr": 7.27007423717041e-06}
{"train_loss": 0.03812912851572037, "global_step": 224358, "epoch": 2520, "lr": 7.269773189027379e-06}
{"train_loss": 0.07069990783929825, "global_step": 224359, "epoch": 2520, "lr": 7.269472146628919e-06}
{"train_loss": 0.03132938966155052, "global_step": 224360, "epoch": 2520, "lr": 7.269171109975042e-06}
{"train_loss": 0.07092411071062088, "global_step": 224361, "epoch": 2520, "lr": 7.268870079065821e-06}
{"train_loss": 0.036489732563495636, "global_step": 224362, "epoch": 2520, "lr": 7.268569053901264e-06}
{"train_loss": 0.036465901881456375, "global_step": 224363, "epoch": 2520, "lr": 7.26826803448144e-06}
{"train_loss": 0.06700249016284943, "global_step": 224364, "epoch": 2520, "lr": 7.267967020806365e-06}
{"train_loss": 0.0680059865117073, "global_step": 224365, "epoch": 2520, "lr": 7.267666012876101e-06}
{"train_loss": 0.09073898941278458, "global_step": 224366, "epoch": 2520, "lr": 7.267365010690669e-06}
{"train_loss": 0.042958248406648636, "global_step": 224367, "epoch": 2520, "lr": 7.26706401425013e-06}
{"train_loss": 0.04656061588713292, "global_step": 224368, "epoch": 2520, "lr": 7.2667630235545015e-06, "val_loss": 8.958407402038574, "train_action_mse_error": 33.42793273925781}
{"train_loss": 0.042860325425863266, "global_step": 224369, "epoch": 2521, "lr": 7.266462038603833e-06}
{"train_loss": 0.0688355565071106, "global_step": 224370, "epoch": 2521, "lr": 7.266161059398185e-06}
{"train_loss": 0.0948055163025856, "global_step": 224371, "epoch": 2521, "lr": 7.265860085937565e-06}
{"train_loss": 0.031038476154208183, "global_step": 224372, "epoch": 2521, "lr": 7.2655591182220426e-06}
{"train_loss": 0.03661905601620674, "global_step": 224373, "epoch": 2521, "lr": 7.265258156251631e-06}
{"train_loss": 0.06607598811388016, "global_step": 224374, "epoch": 2521, "lr": 7.264957200026401e-06}
{"train_loss": 0.040796298533678055, "global_step": 224375, "epoch": 2521, "lr": 7.264656249546359e-06}
{"train_loss": 0.0367935411632061, "global_step": 224376, "epoch": 2521, "lr": 7.264355304811582e-06}
{"train_loss": 0.04748048633337021, "global_step": 224377, "epoch": 2521, "lr": 7.264054365822076e-06}
{"train_loss": 0.03988070785999298, "global_step": 224378, "epoch": 2521, "lr": 7.263753432577908e-06}
{"train_loss": 0.0638972669839859, "global_step": 224379, "epoch": 2521, "lr": 7.263452505079093e-06}
{"train_loss": 0.062348201870918274, "global_step": 224380, "epoch": 2521, "lr": 7.2631515833257e-06}
{"train_loss": 0.05221593752503395, "global_step": 224381, "epoch": 2521, "lr": 7.26285066731775e-06}
{"train_loss": 0.03006976842880249, "global_step": 224382, "epoch": 2521, "lr": 7.262549757055292e-06}
{"train_loss": 0.046562835574150085, "global_step": 224383, "epoch": 2521, "lr": 7.262248852538356e-06}
{"train_loss": 0.039598144590854645, "global_step": 224384, "epoch": 2521, "lr": 7.261947953767001e-06}
{"train_loss": 0.024997711181640625, "global_step": 224385, "epoch": 2521, "lr": 7.2616470607412446e-06}
{"train_loss": 0.05786309763789177, "global_step": 224386, "epoch": 2521, "lr": 7.2613461734611475e-06}
{"train_loss": 0.011875584721565247, "global_step": 224387, "epoch": 2521, "lr": 7.2610452919267325e-06}
{"train_loss": 0.03992083668708801, "global_step": 224388, "epoch": 2521, "lr": 7.26074441613806e-06}
{"train_loss": 0.04846956208348274, "global_step": 224389, "epoch": 2521, "lr": 7.260443546095147e-06}
{"train_loss": 0.008879152126610279, "global_step": 224390, "epoch": 2521, "lr": 7.26014268179806e-06}
{"train_loss": 0.04881774261593819, "global_step": 224391, "epoch": 2521, "lr": 7.259841823246827e-06}
{"train_loss": 0.060138702392578125, "global_step": 224392, "epoch": 2521, "lr": 7.25954097044147e-06}
{"train_loss": 0.06339480727910995, "global_step": 224393, "epoch": 2521, "lr": 7.259240123382066e-06}
{"train_loss": 0.07001597434282303, "global_step": 224394, "epoch": 2521, "lr": 7.2589392820686165e-06}
{"train_loss": 0.03423096984624863, "global_step": 224395, "epoch": 2521, "lr": 7.258638446501198e-06}
{"train_loss": 0.1058339774608612, "global_step": 224396, "epoch": 2521, "lr": 7.258337616679822e-06}
{"train_loss": 0.053428515791893005, "global_step": 224397, "epoch": 2521, "lr": 7.258036792604556e-06}
{"train_loss": 0.07192811369895935, "global_step": 224398, "epoch": 2521, "lr": 7.257735974275409e-06}
{"train_loss": 0.1181417852640152, "global_step": 224399, "epoch": 2521, "lr": 7.2574351616924555e-06}
{"train_loss": 0.04906299710273743, "global_step": 224400, "epoch": 2521, "lr": 7.257134354855704e-06}
{"train_loss": 0.06915634870529175, "global_step": 224401, "epoch": 2521, "lr": 7.2568335537652246e-06}
{"train_loss": 0.06490152329206467, "global_step": 224402, "epoch": 2521, "lr": 7.256532758421025e-06}
{"train_loss": 0.04886046424508095, "global_step": 224403, "epoch": 2521, "lr": 7.25623196882318e-06}
{"train_loss": 0.04855632781982422, "global_step": 224404, "epoch": 2521, "lr": 7.255931184971703e-06}
{"train_loss": 0.05531412363052368, "global_step": 224405, "epoch": 2521, "lr": 7.255630406866643e-06}
{"train_loss": 0.03337762877345085, "global_step": 224406, "epoch": 2521, "lr": 7.255329634508057e-06}
{"train_loss": 0.04400799050927162, "global_step": 224407, "epoch": 2521, "lr": 7.255028867895957e-06}
{"train_loss": 0.057196490466594696, "global_step": 224408, "epoch": 2521, "lr": 7.2547281070304065e-06}
{"train_loss": 0.04457241669297218, "global_step": 224409, "epoch": 2521, "lr": 7.2544273519114305e-06}
{"train_loss": 0.027961211279034615, "global_step": 224410, "epoch": 2521, "lr": 7.25412660253908e-06}
{"train_loss": 0.05323963984847069, "global_step": 224411, "epoch": 2521, "lr": 7.253825858913388e-06}
{"train_loss": 0.05549764633178711, "global_step": 224412, "epoch": 2521, "lr": 7.253525121034405e-06}
{"train_loss": 0.057348545640707016, "global_step": 224413, "epoch": 2521, "lr": 7.253224388902152e-06}
{"train_loss": 0.04319267347455025, "global_step": 224414, "epoch": 2521, "lr": 7.252923662516697e-06}
{"train_loss": 0.04228607192635536, "global_step": 224415, "epoch": 2521, "lr": 7.252622941878051e-06}
{"train_loss": 0.0741581842303276, "global_step": 224416, "epoch": 2521, "lr": 7.25232222698628e-06}
{"train_loss": 0.0748622715473175, "global_step": 224417, "epoch": 2521, "lr": 7.252021517841406e-06}
{"train_loss": 0.06102249398827553, "global_step": 224418, "epoch": 2521, "lr": 7.251720814443486e-06}
{"train_loss": 0.04077979177236557, "global_step": 224419, "epoch": 2521, "lr": 7.251420116792534e-06}
{"train_loss": 0.035531483590602875, "global_step": 224420, "epoch": 2521, "lr": 7.251119424888625e-06}
{"train_loss": 0.07607365399599075, "global_step": 224421, "epoch": 2521, "lr": 7.250818738731768e-06}
{"train_loss": 0.07246466726064682, "global_step": 224422, "epoch": 2521, "lr": 7.25051805832202e-06}
{"train_loss": 0.04452949017286301, "global_step": 224423, "epoch": 2521, "lr": 7.25021738365943e-06}
{"train_loss": 0.06284185498952866, "global_step": 224424, "epoch": 2521, "lr": 7.249916714744026e-06}
{"train_loss": 0.05849261209368706, "global_step": 224425, "epoch": 2521, "lr": 7.249616051575842e-06}
{"train_loss": 0.08370152860879898, "global_step": 224426, "epoch": 2521, "lr": 7.2493153941549316e-06}
{"train_loss": 0.06389333307743073, "global_step": 224427, "epoch": 2521, "lr": 7.2490147424813196e-06}
{"train_loss": 0.05740954354405403, "global_step": 224428, "epoch": 2521, "lr": 7.24871409655507e-06}
{"train_loss": 0.04400194063782692, "global_step": 224429, "epoch": 2521, "lr": 7.248413456376197e-06}
{"train_loss": 0.049947671592235565, "global_step": 224430, "epoch": 2521, "lr": 7.2481128219447635e-06}
{"train_loss": 0.032100606709718704, "global_step": 224431, "epoch": 2521, "lr": 7.247812193260806e-06}
{"train_loss": 0.09161046147346497, "global_step": 224432, "epoch": 2521, "lr": 7.247511570324345e-06}
{"train_loss": 0.04444384574890137, "global_step": 224433, "epoch": 2521, "lr": 7.2472109531354415e-06}
{"train_loss": 0.08298435807228088, "global_step": 224434, "epoch": 2521, "lr": 7.246910341694124e-06}
{"train_loss": 0.07113667577505112, "global_step": 224435, "epoch": 2521, "lr": 7.246609736000448e-06}
{"train_loss": 0.03319214656949043, "global_step": 224436, "epoch": 2521, "lr": 7.24630913605443e-06}
{"train_loss": 0.052676767110824585, "global_step": 224437, "epoch": 2521, "lr": 7.246008541856136e-06}
{"train_loss": 0.07922038435935974, "global_step": 224438, "epoch": 2521, "lr": 7.245707953405589e-06}
{"train_loss": 0.10365008562803268, "global_step": 224439, "epoch": 2521, "lr": 7.245407370702845e-06}
{"train_loss": 0.051885105669498444, "global_step": 224440, "epoch": 2521, "lr": 7.245106793747919e-06}
{"train_loss": 0.09577639400959015, "global_step": 224441, "epoch": 2521, "lr": 7.244806222540873e-06}
{"train_loss": 0.05044412612915039, "global_step": 224442, "epoch": 2521, "lr": 7.244505657081752e-06}
{"train_loss": 0.0625058263540268, "global_step": 224443, "epoch": 2521, "lr": 7.244205097370571e-06}
{"train_loss": 0.0310019813477993, "global_step": 224444, "epoch": 2521, "lr": 7.2439045434073986e-06}
{"train_loss": 0.045909348875284195, "global_step": 224445, "epoch": 2521, "lr": 7.24360399519225e-06}
{"train_loss": 0.06182336434721947, "global_step": 224446, "epoch": 2521, "lr": 7.243303452725192e-06}
{"train_loss": 0.0618455708026886, "global_step": 224447, "epoch": 2521, "lr": 7.243002916006237e-06}
{"train_loss": 0.06113989278674126, "global_step": 224448, "epoch": 2521, "lr": 7.242702385035455e-06}
{"train_loss": 0.023103849962353706, "global_step": 224449, "epoch": 2521, "lr": 7.242401859812853e-06}
{"train_loss": 0.06760892271995544, "global_step": 224450, "epoch": 2521, "lr": 7.242101340338503e-06}
{"train_loss": 0.03575499355792999, "global_step": 224451, "epoch": 2521, "lr": 7.241800826612422e-06}
{"train_loss": 0.06772366911172867, "global_step": 224452, "epoch": 2521, "lr": 7.2415003186346646e-06}
{"train_loss": 0.04880829155445099, "global_step": 224453, "epoch": 2521, "lr": 7.241199816405258e-06}
{"train_loss": 0.023835590109229088, "global_step": 224454, "epoch": 2521, "lr": 7.240899319924266e-06}
{"train_loss": 0.041405729949474335, "global_step": 224455, "epoch": 2521, "lr": 7.240598829191703e-06}
{"train_loss": 0.023039212450385094, "global_step": 224456, "epoch": 2521, "lr": 7.240298344207625e-06}
{"train_loss": 0.05469126818345839, "global_step": 224457, "epoch": 2521, "lr": 7.239997864972059e-06, "val_loss": 9.095629692077637}
{"train_loss": 0.06836055964231491, "global_step": 224458, "epoch": 2522, "lr": 7.239697391485068e-06}
{"train_loss": 0.028310615569353104, "global_step": 224459, "epoch": 2522, "lr": 7.239396923746666e-06}
{"train_loss": 0.09095904976129532, "global_step": 224460, "epoch": 2522, "lr": 7.239096461756906e-06}
{"train_loss": 0.08869758248329163, "global_step": 224461, "epoch": 2522, "lr": 7.238796005515841e-06}
{"train_loss": 0.04164321348071098, "global_step": 224462, "epoch": 2522, "lr": 7.2384955550234835e-06}
{"train_loss": 0.020444391295313835, "global_step": 224463, "epoch": 2522, "lr": 7.238195110279905e-06}
{"train_loss": 0.06476622819900513, "global_step": 224464, "epoch": 2522, "lr": 7.237894671285123e-06}
{"train_loss": 0.06393270939588547, "global_step": 224465, "epoch": 2522, "lr": 7.23759423803918e-06}
{"train_loss": 0.11535384505987167, "global_step": 224466, "epoch": 2522, "lr": 7.2372938105421285e-06}
{"train_loss": 0.0986773818731308, "global_step": 224467, "epoch": 2522, "lr": 7.236993388793995e-06}
{"train_loss": 0.034861207008361816, "global_step": 224468, "epoch": 2522, "lr": 7.236692972794834e-06}
{"train_loss": 0.029518665745854378, "global_step": 224469, "epoch": 2522, "lr": 7.23639256254468e-06}
{"train_loss": 0.06912250816822052, "global_step": 224470, "epoch": 2522, "lr": 7.236092158043562e-06}
{"train_loss": 0.09169381111860275, "global_step": 224471, "epoch": 2522, "lr": 7.235791759291538e-06}
{"train_loss": 0.05069683864712715, "global_step": 224472, "epoch": 2522, "lr": 7.235491366288632e-06}
{"train_loss": 0.048794277012348175, "global_step": 224473, "epoch": 2522, "lr": 7.2351909790349005e-06}
{"train_loss": 0.04172375798225403, "global_step": 224474, "epoch": 2522, "lr": 7.234890597530369e-06}
{"train_loss": 0.07969115674495697, "global_step": 224475, "epoch": 2522, "lr": 7.234590221775095e-06}
{"train_loss": 0.07780833542346954, "global_step": 224476, "epoch": 2522, "lr": 7.234289851769099e-06}
{"train_loss": 0.053961724042892456, "global_step": 224477, "epoch": 2522, "lr": 7.233989487512432e-06}
{"train_loss": 0.03486476093530655, "global_step": 224478, "epoch": 2522, "lr": 7.233689129005145e-06}
{"train_loss": 0.0853164792060852, "global_step": 224479, "epoch": 2522, "lr": 7.2333887762472515e-06}
{"train_loss": 0.08480926603078842, "global_step": 224480, "epoch": 2522, "lr": 7.2330884292388215e-06}
{"train_loss": 0.0795045793056488, "global_step": 224481, "epoch": 2522, "lr": 7.232788087979875e-06}
{"train_loss": 0.07519479095935822, "global_step": 224482, "epoch": 2522, "lr": 7.232487752470463e-06}
{"train_loss": 0.07204397767782211, "global_step": 224483, "epoch": 2522, "lr": 7.232187422710613e-06}
{"train_loss": 0.031079187989234924, "global_step": 224484, "epoch": 2522, "lr": 7.231887098700385e-06}
{"train_loss": 0.028339894488453865, "global_step": 224485, "epoch": 2522, "lr": 7.2315867804397975e-06}
{"train_loss": 0.11539839953184128, "global_step": 224486, "epoch": 2522, "lr": 7.231286467928916e-06}
{"train_loss": 0.06281158328056335, "global_step": 224487, "epoch": 2522, "lr": 7.230986161167752e-06}
{"train_loss": 0.05828261002898216, "global_step": 224488, "epoch": 2522, "lr": 7.2306858601563725e-06}
{"train_loss": 0.05756949633359909, "global_step": 224489, "epoch": 2522, "lr": 7.230385564894798e-06}
{"train_loss": 0.04121498763561249, "global_step": 224490, "epoch": 2522, "lr": 7.230085275383086e-06}
{"train_loss": 0.04204616695642471, "global_step": 224491, "epoch": 2522, "lr": 7.229784991621252e-06}
{"train_loss": 0.03837144374847412, "global_step": 224492, "epoch": 2522, "lr": 7.229484713609369e-06}
{"train_loss": 0.04134508594870567, "global_step": 224493, "epoch": 2522, "lr": 7.229184441347448e-06}
{"train_loss": 0.04531855508685112, "global_step": 224494, "epoch": 2522, "lr": 7.228884174835554e-06}
{"train_loss": 0.02967109903693199, "global_step": 224495, "epoch": 2522, "lr": 7.2285839140737e-06}
{"train_loss": 0.05726325511932373, "global_step": 224496, "epoch": 2522, "lr": 7.228283659061946e-06}
{"train_loss": 0.04966946318745613, "global_step": 224497, "epoch": 2522, "lr": 7.227983409800337e-06}
{"train_loss": 0.07797373086214066, "global_step": 224498, "epoch": 2522, "lr": 7.227683166288895e-06}
{"train_loss": 0.03682525083422661, "global_step": 224499, "epoch": 2522, "lr": 7.2273829285276816e-06}
{"train_loss": 0.05398084968328476, "global_step": 224500, "epoch": 2522, "lr": 7.227082696516707e-06}
{"train_loss": 0.03737331181764603, "global_step": 224501, "epoch": 2522, "lr": 7.22678247025605e-06}
{"train_loss": 0.04493189975619316, "global_step": 224502, "epoch": 2522, "lr": 7.226482249745725e-06}
{"train_loss": 0.03929304704070091, "global_step": 224503, "epoch": 2522, "lr": 7.226182034985768e-06}
{"train_loss": 0.07976751774549484, "global_step": 224504, "epoch": 2522, "lr": 7.225881825976244e-06}
{"train_loss": 0.06375669687986374, "global_step": 224505, "epoch": 2522, "lr": 7.2255816227171645e-06}
{"train_loss": 0.07587865740060806, "global_step": 224506, "epoch": 2522, "lr": 7.225281425208596e-06}
{"train_loss": 0.0588802732527256, "global_step": 224507, "epoch": 2522, "lr": 7.224981233450568e-06}
{"train_loss": 0.07722347229719162, "global_step": 224508, "epoch": 2522, "lr": 7.224681047443105e-06}
{"train_loss": 0.04485468193888664, "global_step": 224509, "epoch": 2522, "lr": 7.224380867186275e-06}
{"train_loss": 0.034815698862075806, "global_step": 224510, "epoch": 2522, "lr": 7.224080692680096e-06}
{"train_loss": 0.09907622635364532, "global_step": 224511, "epoch": 2522, "lr": 7.223780523924628e-06}
{"train_loss": 0.03788300231099129, "global_step": 224512, "epoch": 2522, "lr": 7.223480360919893e-06}
{"train_loss": 0.05792267248034477, "global_step": 224513, "epoch": 2522, "lr": 7.223180203665936e-06}
{"train_loss": 0.03976498171687126, "global_step": 224514, "epoch": 2522, "lr": 7.2228800521628116e-06}
{"train_loss": 0.048389144241809845, "global_step": 224515, "epoch": 2522, "lr": 7.222579906410542e-06}
{"train_loss": 0.02085641212761402, "global_step": 224516, "epoch": 2522, "lr": 7.222279766409185e-06}
{"train_loss": 0.039103321731090546, "global_step": 224517, "epoch": 2522, "lr": 7.22197963215876e-06}
{"train_loss": 0.08870697021484375, "global_step": 224518, "epoch": 2522, "lr": 7.2216795036593284e-06}
{"train_loss": 0.038174692541360855, "global_step": 224519, "epoch": 2522, "lr": 7.221379380910915e-06}
{"train_loss": 0.05395713075995445, "global_step": 224520, "epoch": 2522, "lr": 7.221079263913572e-06}
{"train_loss": 0.05550530552864075, "global_step": 224521, "epoch": 2522, "lr": 7.220779152667317e-06}
{"train_loss": 0.06133323535323143, "global_step": 224522, "epoch": 2522, "lr": 7.220479047172224e-06}
{"train_loss": 0.07421489804983139, "global_step": 224523, "epoch": 2522, "lr": 7.220178947428308e-06}
{"train_loss": 0.06986681371927261, "global_step": 224524, "epoch": 2522, "lr": 7.219878853435624e-06}
{"train_loss": 0.030727064236998558, "global_step": 224525, "epoch": 2522, "lr": 7.219578765194196e-06}
{"train_loss": 0.042220015078783035, "global_step": 224526, "epoch": 2522, "lr": 7.2192786827040835e-06}
{"train_loss": 0.06647654622793198, "global_step": 224527, "epoch": 2522, "lr": 7.21897860596531e-06}
{"train_loss": 0.04913177713751793, "global_step": 224528, "epoch": 2522, "lr": 7.218678534977935e-06}
{"train_loss": 0.07122334092855453, "global_step": 224529, "epoch": 2522, "lr": 7.218378469741971e-06}
{"train_loss": 0.07208860665559769, "global_step": 224530, "epoch": 2522, "lr": 7.218078410257489e-06}
{"train_loss": 0.055493682622909546, "global_step": 224531, "epoch": 2522, "lr": 7.217778356524507e-06}
{"train_loss": 0.06785916537046432, "global_step": 224532, "epoch": 2522, "lr": 7.217478308543069e-06}
{"train_loss": 0.06110001355409622, "global_step": 224533, "epoch": 2522, "lr": 7.217178266313229e-06}
{"train_loss": 0.023441998288035393, "global_step": 224534, "epoch": 2522, "lr": 7.216878229835011e-06}
{"train_loss": 0.03037763573229313, "global_step": 224535, "epoch": 2522, "lr": 7.21657819910847e-06}
{"train_loss": 0.050504185259342194, "global_step": 224536, "epoch": 2522, "lr": 7.216278174133628e-06}
{"train_loss": 0.04020809009671211, "global_step": 224537, "epoch": 2522, "lr": 7.215978154910552e-06}
{"train_loss": 0.034348100423812866, "global_step": 224538, "epoch": 2522, "lr": 7.215678141439247e-06}
{"train_loss": 0.053542349487543106, "global_step": 224539, "epoch": 2522, "lr": 7.215378133719791e-06}
{"train_loss": 0.025081021711230278, "global_step": 224540, "epoch": 2522, "lr": 7.21507813175219e-06}
{"train_loss": 0.05069228261709213, "global_step": 224541, "epoch": 2522, "lr": 7.214778135536515e-06}
{"train_loss": 0.03337428718805313, "global_step": 224542, "epoch": 2522, "lr": 7.21447814507279e-06}
{"train_loss": 0.0326443575322628, "global_step": 224543, "epoch": 2522, "lr": 7.214178160361046e-06}
{"train_loss": 0.05713258683681488, "global_step": 224544, "epoch": 2522, "lr": 7.213878181401345e-06}
{"train_loss": 0.057417143136262894, "global_step": 224545, "epoch": 2522, "lr": 7.213578208193705e-06}
{"train_loss": 0.05567614453729619, "global_step": 224546, "epoch": 2522, "lr": 7.213278240738192e-06, "val_loss": 9.045245170593262}
{"train_loss": 0.06359441578388214, "global_step": 224547, "epoch": 2523, "lr": 7.2129782790348265e-06}
{"train_loss": 0.05309490114450455, "global_step": 224548, "epoch": 2523, "lr": 7.212678323083649e-06}
{"train_loss": 0.04539826884865761, "global_step": 224549, "epoch": 2523, "lr": 7.2123783728847035e-06}
{"train_loss": 0.05255582928657532, "global_step": 224550, "epoch": 2523, "lr": 7.212078428438041e-06}
{"train_loss": 0.06867696344852448, "global_step": 224551, "epoch": 2523, "lr": 7.211778489743687e-06}
{"train_loss": 0.017301367595791817, "global_step": 224552, "epoch": 2523, "lr": 7.211478556801693e-06}
{"train_loss": 0.06000277027487755, "global_step": 224553, "epoch": 2523, "lr": 7.211178629612086e-06}
{"train_loss": 0.039683595299720764, "global_step": 224554, "epoch": 2523, "lr": 7.2108787081749285e-06}
{"train_loss": 0.06314101815223694, "global_step": 224555, "epoch": 2523, "lr": 7.21057879249023e-06}
{"train_loss": 0.0553966723382473, "global_step": 224556, "epoch": 2523, "lr": 7.210278882558064e-06}
{"train_loss": 0.017011025920510292, "global_step": 224557, "epoch": 2523, "lr": 7.20997897837844e-06}
{"train_loss": 0.06186523288488388, "global_step": 224558, "epoch": 2523, "lr": 7.209679079951426e-06}
{"train_loss": 0.0691009983420372, "global_step": 224559, "epoch": 2523, "lr": 7.209379187277038e-06}
{"train_loss": 0.0889897346496582, "global_step": 224560, "epoch": 2523, "lr": 7.209079300355337e-06}
{"train_loss": 0.03464693948626518, "global_step": 224561, "epoch": 2523, "lr": 7.20877941918634e-06}
{"train_loss": 0.05441682040691376, "global_step": 224562, "epoch": 2523, "lr": 7.2084795437701195e-06}
{"train_loss": 0.02810819447040558, "global_step": 224563, "epoch": 2523, "lr": 7.20817967410668e-06}
{"train_loss": 0.042636651545763016, "global_step": 224564, "epoch": 2523, "lr": 7.207879810196094e-06}
{"train_loss": 0.05328639969229698, "global_step": 224565, "epoch": 2523, "lr": 7.2075799520383735e-06}
{"train_loss": 0.06793244183063507, "global_step": 224566, "epoch": 2523, "lr": 7.207280099633584e-06}
{"train_loss": 0.04353797063231468, "global_step": 224567, "epoch": 2523, "lr": 7.206980252981743e-06}
{"train_loss": 0.05369766056537628, "global_step": 224568, "epoch": 2523, "lr": 7.2066804120829004e-06}
{"train_loss": 0.04346771165728569, "global_step": 224569, "epoch": 2523, "lr": 7.206380576937111e-06}
{"train_loss": 0.03752504289150238, "global_step": 224570, "epoch": 2523, "lr": 7.2060807475443915e-06}
{"train_loss": 0.038210369646549225, "global_step": 224571, "epoch": 2523, "lr": 7.205780923904803e-06}
{"train_loss": 0.015696996822953224, "global_step": 224572, "epoch": 2523, "lr": 7.205481106018363e-06}
{"train_loss": 0.07392793148756027, "global_step": 224573, "epoch": 2523, "lr": 7.205181293885138e-06}
{"train_loss": 0.07632099092006683, "global_step": 224574, "epoch": 2523, "lr": 7.204881487505144e-06}
{"train_loss": 0.05708387494087219, "global_step": 224575, "epoch": 2523, "lr": 7.204581686878442e-06}
{"train_loss": 0.04681391641497612, "global_step": 224576, "epoch": 2523, "lr": 7.204281892005049e-06}
{"train_loss": 0.06695003062486649, "global_step": 224577, "epoch": 2523, "lr": 7.203982102885037e-06}
{"train_loss": 0.01649291254580021, "global_step": 224578, "epoch": 2523, "lr": 7.203682319518412e-06}
{"train_loss": 0.023528648540377617, "global_step": 224579, "epoch": 2523, "lr": 7.203382541905241e-06}
{"train_loss": 0.036073893308639526, "global_step": 224580, "epoch": 2523, "lr": 7.203082770045555e-06}
{"train_loss": 0.03741242364048958, "global_step": 224581, "epoch": 2523, "lr": 7.202783003939378e-06}
{"train_loss": 0.05608104541897774, "global_step": 224582, "epoch": 2523, "lr": 7.2024832435867825e-06}
{"train_loss": 0.06658414751291275, "global_step": 224583, "epoch": 2523, "lr": 7.202183488987774e-06}
{"train_loss": 0.08526802808046341, "global_step": 224584, "epoch": 2523, "lr": 7.201883740142429e-06}
{"train_loss": 0.030037609860301018, "global_step": 224585, "epoch": 2523, "lr": 7.2015839970507645e-06}
{"train_loss": 0.05582735687494278, "global_step": 224586, "epoch": 2523, "lr": 7.201284259712809e-06}
{"train_loss": 0.03896916285157204, "global_step": 224587, "epoch": 2523, "lr": 7.2009845281286295e-06}
{"train_loss": 0.06291110813617706, "global_step": 224588, "epoch": 2523, "lr": 7.200684802298263e-06}
{"train_loss": 0.05977993458509445, "global_step": 224589, "epoch": 2523, "lr": 7.200385082221727e-06}
{"train_loss": 0.06736012548208237, "global_step": 224590, "epoch": 2523, "lr": 7.200085367899096e-06}
{"train_loss": 0.06190359219908714, "global_step": 224591, "epoch": 2523, "lr": 7.1997856593303766e-06}
{"train_loss": 0.062348321080207825, "global_step": 224592, "epoch": 2523, "lr": 7.1994859565156336e-06}
{"train_loss": 0.06045093759894371, "global_step": 224593, "epoch": 2523, "lr": 7.199186259454893e-06}
{"train_loss": 0.06828270852565765, "global_step": 224594, "epoch": 2523, "lr": 7.1988865681482056e-06}
{"train_loss": 0.06832905858755112, "global_step": 224595, "epoch": 2523, "lr": 7.1985868825955985e-06}
{"train_loss": 0.03368029743432999, "global_step": 224596, "epoch": 2523, "lr": 7.198287202797127e-06}
{"train_loss": 0.051944658160209656, "global_step": 224597, "epoch": 2523, "lr": 7.19798752875282e-06}
{"train_loss": 0.06577961146831512, "global_step": 224598, "epoch": 2523, "lr": 7.197687860462727e-06}
{"train_loss": 0.023578254505991936, "global_step": 224599, "epoch": 2523, "lr": 7.1973881979268745e-06}
{"train_loss": 0.058061592280864716, "global_step": 224600, "epoch": 2523, "lr": 7.1970885411453194e-06}
{"train_loss": 0.03794810175895691, "global_step": 224601, "epoch": 2523, "lr": 7.196788890118089e-06}
{"train_loss": 0.031048355624079704, "global_step": 224602, "epoch": 2523, "lr": 7.196489244845234e-06}
{"train_loss": 0.04384993016719818, "global_step": 224603, "epoch": 2523, "lr": 7.196189605326781e-06}
{"train_loss": 0.04815462604165077, "global_step": 224604, "epoch": 2523, "lr": 7.195889971562775e-06}
{"train_loss": 0.06006944552063942, "global_step": 224605, "epoch": 2523, "lr": 7.195590343553276e-06}
{"train_loss": 0.029172591865062714, "global_step": 224606, "epoch": 2523, "lr": 7.195290721298292e-06}
{"train_loss": 0.04838463291525841, "global_step": 224607, "epoch": 2523, "lr": 7.194991104797899e-06}
{"train_loss": 0.05326858535408974, "global_step": 224608, "epoch": 2523, "lr": 7.194691494052097e-06}
{"train_loss": 0.0810476765036583, "global_step": 224609, "epoch": 2523, "lr": 7.194391889060964e-06}
{"train_loss": 0.04249119386076927, "global_step": 224610, "epoch": 2523, "lr": 7.194092289824511e-06}
{"train_loss": 0.0661933422088623, "global_step": 224611, "epoch": 2523, "lr": 7.1937926963427996e-06}
{"train_loss": 0.07556258141994476, "global_step": 224612, "epoch": 2523, "lr": 7.193493108615851e-06}
{"train_loss": 0.08901537209749222, "global_step": 224613, "epoch": 2523, "lr": 7.193193526643727e-06}
{"train_loss": 0.026163781061768532, "global_step": 224614, "epoch": 2523, "lr": 7.192893950426449e-06}
{"train_loss": 0.055987242609262466, "global_step": 224615, "epoch": 2523, "lr": 7.192594379964068e-06}
{"train_loss": 0.09425898641347885, "global_step": 224616, "epoch": 2523, "lr": 7.192294815256617e-06}
{"train_loss": 0.04649975150823593, "global_step": 224617, "epoch": 2523, "lr": 7.191995256304146e-06}
{"train_loss": 0.0791882574558258, "global_step": 224618, "epoch": 2523, "lr": 7.191695703106693e-06}
{"train_loss": 0.05076530948281288, "global_step": 224619, "epoch": 2523, "lr": 7.191396155664281e-06}
{"train_loss": 0.03762631490826607, "global_step": 224620, "epoch": 2523, "lr": 7.191096613976978e-06}
{"train_loss": 0.04583752527832985, "global_step": 224621, "epoch": 2523, "lr": 7.1907970780447975e-06}
{"train_loss": 0.028257649391889572, "global_step": 224622, "epoch": 2523, "lr": 7.190497547867803e-06}
{"train_loss": 0.05123945325613022, "global_step": 224623, "epoch": 2523, "lr": 7.190198023446016e-06}
{"train_loss": 0.04834363982081413, "global_step": 224624, "epoch": 2523, "lr": 7.1898985047794965e-06}
{"train_loss": 0.05962682515382767, "global_step": 224625, "epoch": 2523, "lr": 7.189598991868257e-06}
{"train_loss": 0.048073623329401016, "global_step": 224626, "epoch": 2523, "lr": 7.189299484712369e-06}
{"train_loss": 0.053711164742708206, "global_step": 224627, "epoch": 2523, "lr": 7.188999983311845e-06}
{"train_loss": 0.048075560480356216, "global_step": 224628, "epoch": 2523, "lr": 7.18870048766675e-06}
{"train_loss": 0.07985725998878479, "global_step": 224629, "epoch": 2523, "lr": 7.188400997777101e-06}
{"train_loss": 0.09730757772922516, "global_step": 224630, "epoch": 2523, "lr": 7.188101513642964e-06}
{"train_loss": 0.0736316367983818, "global_step": 224631, "epoch": 2523, "lr": 7.187802035264351e-06}
{"train_loss": 0.02368917129933834, "global_step": 224632, "epoch": 2523, "lr": 7.187502562641329e-06}
{"train_loss": 0.06365892291069031, "global_step": 224633, "epoch": 2523, "lr": 7.187203095773915e-06}
{"train_loss": 0.07225136458873749, "global_step": 224634, "epoch": 2523, "lr": 7.186903634662168e-06}
{"train_loss": 0.053110358545954306, "global_step": 224635, "epoch": 2523, "lr": 7.186604179306106e-06, "val_loss": 9.047457695007324}
{"train_loss": 0.08511324226856232, "global_step": 224636, "epoch": 2524, "lr": 7.1863047297058e-06}
{"train_loss": 0.05877649039030075, "global_step": 224637, "epoch": 2524, "lr": 7.186005285861258e-06}
{"train_loss": 0.042744919657707214, "global_step": 224638, "epoch": 2524, "lr": 7.18570584777255e-06}
{"train_loss": 0.03496430441737175, "global_step": 224639, "epoch": 2524, "lr": 7.185406415439694e-06}
{"train_loss": 0.07034008204936981, "global_step": 224640, "epoch": 2524, "lr": 7.185106988862733e-06}
{"train_loss": 0.06559118628501892, "global_step": 224641, "epoch": 2524, "lr": 7.184807568041724e-06}
{"train_loss": 0.06229954585433006, "global_step": 224642, "epoch": 2524, "lr": 7.184508152976688e-06}
{"train_loss": 0.042806461453437805, "global_step": 224643, "epoch": 2524, "lr": 7.184208743667681e-06}
{"train_loss": 0.022865470498800278, "global_step": 224644, "epoch": 2524, "lr": 7.183909340114725e-06}
{"train_loss": 0.03007948026061058, "global_step": 224645, "epoch": 2524, "lr": 7.183609942317881e-06}
{"train_loss": 0.06125311926007271, "global_step": 224646, "epoch": 2524, "lr": 7.183310550277167e-06}
{"train_loss": 0.013115004636347294, "global_step": 224647, "epoch": 2524, "lr": 7.183011163992648e-06}
{"train_loss": 0.06306536495685577, "global_step": 224648, "epoch": 2524, "lr": 7.182711783464341e-06}
{"train_loss": 0.08942656219005585, "global_step": 224649, "epoch": 2524, "lr": 7.182412408692307e-06}
{"train_loss": 0.04147233068943024, "global_step": 224650, "epoch": 2524, "lr": 7.182113039676563e-06}
{"train_loss": 0.03839583322405815, "global_step": 224651, "epoch": 2524, "lr": 7.181813676417176e-06}
{"train_loss": 0.03294594958424568, "global_step": 224652, "epoch": 2524, "lr": 7.181514318914156e-06}
{"train_loss": 0.05067824199795723, "global_step": 224653, "epoch": 2524, "lr": 7.181214967167577e-06}
{"train_loss": 0.04808714613318443, "global_step": 224654, "epoch": 2524, "lr": 7.1809156211774474e-06}
{"train_loss": 0.038832176476716995, "global_step": 224655, "epoch": 2524, "lr": 7.18061628094383e-06}
{"train_loss": 0.03994712606072426, "global_step": 224656, "epoch": 2524, "lr": 7.180316946466753e-06}
{"train_loss": 0.04531701281666756, "global_step": 224657, "epoch": 2524, "lr": 7.180017617746265e-06}
{"train_loss": 0.06060855835676193, "global_step": 224658, "epoch": 2524, "lr": 7.179718294782406e-06}
{"train_loss": 0.025294974446296692, "global_step": 224659, "epoch": 2524, "lr": 7.179418977575197e-06}
{"train_loss": 0.03895508497953415, "global_step": 224660, "epoch": 2524, "lr": 7.1791196661247065e-06}
{"train_loss": 0.08072549104690552, "global_step": 224661, "epoch": 2524, "lr": 7.178820360430949e-06}
{"train_loss": 0.02436637319624424, "global_step": 224662, "epoch": 2524, "lr": 7.178521060493993e-06}
{"train_loss": 0.047415878623723984, "global_step": 224663, "epoch": 2524, "lr": 7.1782217663138474e-06}
{"train_loss": 0.05593946948647499, "global_step": 224664, "epoch": 2524, "lr": 7.177922477890581e-06}
{"train_loss": 0.05022173002362251, "global_step": 224665, "epoch": 2524, "lr": 7.177623195224209e-06}
{"train_loss": 0.07684902101755142, "global_step": 224666, "epoch": 2524, "lr": 7.1773239183147935e-06}
{"train_loss": 0.054075807332992554, "global_step": 224667, "epoch": 2524, "lr": 7.177024647162356e-06}
{"train_loss": 0.02640491910278797, "global_step": 224668, "epoch": 2524, "lr": 7.176725381766952e-06}
{"train_loss": 0.059082552790641785, "global_step": 224669, "epoch": 2524, "lr": 7.17642612212861e-06}
{"train_loss": 0.04966479167342186, "global_step": 224670, "epoch": 2524, "lr": 7.176126868247385e-06}
{"train_loss": 0.06649000942707062, "global_step": 224671, "epoch": 2524, "lr": 7.175827620123293e-06}
{"train_loss": 0.036100834608078, "global_step": 224672, "epoch": 2524, "lr": 7.175528377756407e-06}
{"train_loss": 0.0817086324095726, "global_step": 224673, "epoch": 2524, "lr": 7.175229141146733e-06}
{"train_loss": 0.05271787941455841, "global_step": 224674, "epoch": 2524, "lr": 7.174929910294342e-06}
{"train_loss": 0.06369844824075699, "global_step": 224675, "epoch": 2524, "lr": 7.174630685199246e-06}
{"train_loss": 0.05109383538365364, "global_step": 224676, "epoch": 2524, "lr": 7.1743314658615004e-06}
{"train_loss": 0.04192357137799263, "global_step": 224677, "epoch": 2524, "lr": 7.174032252281154e-06}
{"train_loss": 0.009413806721568108, "global_step": 224678, "epoch": 2524, "lr": 7.173733044458225e-06}
{"train_loss": 0.09753454476594925, "global_step": 224679, "epoch": 2524, "lr": 7.17343384239278e-06}
{"train_loss": 0.05196283757686615, "global_step": 224680, "epoch": 2524, "lr": 7.173134646084828e-06}
{"train_loss": 0.04289982467889786, "global_step": 224681, "epoch": 2524, "lr": 7.172835455534444e-06}
{"train_loss": 0.07527121901512146, "global_step": 224682, "epoch": 2524, "lr": 7.172536270741631e-06}
{"train_loss": 0.024882785975933075, "global_step": 224683, "epoch": 2524, "lr": 7.172237091706468e-06}
{"train_loss": 0.02021271176636219, "global_step": 224684, "epoch": 2524, "lr": 7.171937918428962e-06}
{"train_loss": 0.03660842403769493, "global_step": 224685, "epoch": 2524, "lr": 7.171638750909177e-06}
{"train_loss": 0.04633857309818268, "global_step": 224686, "epoch": 2524, "lr": 7.17133958914713e-06}
{"train_loss": 0.09153179079294205, "global_step": 224687, "epoch": 2524, "lr": 7.1710404331428894e-06}
{"train_loss": 0.07968959957361221, "global_step": 224688, "epoch": 2524, "lr": 7.170741282896465e-06}
{"train_loss": 0.028693344444036484, "global_step": 224689, "epoch": 2524, "lr": 7.170442138407929e-06}
{"train_loss": 0.035612355917692184, "global_step": 224690, "epoch": 2524, "lr": 7.170142999677293e-06}
{"train_loss": 0.08545423299074173, "global_step": 224691, "epoch": 2524, "lr": 7.169843866704617e-06}
{"train_loss": 0.07315031439065933, "global_step": 224692, "epoch": 2524, "lr": 7.169544739489925e-06}
{"train_loss": 0.03204404190182686, "global_step": 224693, "epoch": 2524, "lr": 7.1692456180332715e-06}
{"train_loss": 0.046424053609371185, "global_step": 224694, "epoch": 2524, "lr": 7.1689465023346835e-06}
{"train_loss": 0.035873133689165115, "global_step": 224695, "epoch": 2524, "lr": 7.168647392394223e-06}
{"train_loss": 0.05367639660835266, "global_step": 224696, "epoch": 2524, "lr": 7.168348288211913e-06}
{"train_loss": 0.04387128725647926, "global_step": 224697, "epoch": 2524, "lr": 7.168049189787784e-06}
{"train_loss": 0.05583434924483299, "global_step": 224698, "epoch": 2524, "lr": 7.167750097121901e-06}
{"train_loss": 0.09763865172863007, "global_step": 224699, "epoch": 2524, "lr": 7.167451010214282e-06}
{"train_loss": 0.05425454303622246, "global_step": 224700, "epoch": 2524, "lr": 7.167151929064985e-06}
{"train_loss": 0.0938204973936081, "global_step": 224701, "epoch": 2524, "lr": 7.166852853674033e-06}
{"train_loss": 0.07964757084846497, "global_step": 224702, "epoch": 2524, "lr": 7.166553784041491e-06}
{"train_loss": 0.019771426916122437, "global_step": 224703, "epoch": 2524, "lr": 7.16625472016737e-06}
{"train_loss": 0.0166854839771986, "global_step": 224704, "epoch": 2524, "lr": 7.165955662051732e-06}
{"train_loss": 0.036979611963033676, "global_step": 224705, "epoch": 2524, "lr": 7.165656609694599e-06}
{"train_loss": 0.06452777981758118, "global_step": 224706, "epoch": 2524, "lr": 7.165357563096037e-06}
{"train_loss": 0.04622124508023262, "global_step": 224707, "epoch": 2524, "lr": 7.165058522256052e-06}
{"train_loss": 0.06151323765516281, "global_step": 224708, "epoch": 2524, "lr": 7.164759487174721e-06}
{"train_loss": 0.054457034915685654, "global_step": 224709, "epoch": 2524, "lr": 7.164460457852051e-06}
{"train_loss": 0.043820369988679886, "global_step": 224710, "epoch": 2524, "lr": 7.164161434288108e-06}
{"train_loss": 0.05934079736471176, "global_step": 224711, "epoch": 2524, "lr": 7.163862416482914e-06}
{"train_loss": 0.04914050176739693, "global_step": 224712, "epoch": 2524, "lr": 7.1635634044365245e-06}
{"train_loss": 0.10444960743188858, "global_step": 224713, "epoch": 2524, "lr": 7.163264398148961e-06}
{"train_loss": 0.06292524188756943, "global_step": 224714, "epoch": 2524, "lr": 7.162965397620275e-06}
{"train_loss": 0.04140349477529526, "global_step": 224715, "epoch": 2524, "lr": 7.162666402850515e-06}
{"train_loss": 0.07060248404741287, "global_step": 224716, "epoch": 2524, "lr": 7.162367413839705e-06}
{"train_loss": 0.014919459819793701, "global_step": 224717, "epoch": 2524, "lr": 7.162068430587904e-06}
{"train_loss": 0.008553318679332733, "global_step": 224718, "epoch": 2524, "lr": 7.161769453095124e-06}
{"train_loss": 0.02764434926211834, "global_step": 224719, "epoch": 2524, "lr": 7.161470481361437e-06}
{"train_loss": 0.04554259032011032, "global_step": 224720, "epoch": 2524, "lr": 7.1611715153868555e-06}
{"train_loss": 0.0693063735961914, "global_step": 224721, "epoch": 2524, "lr": 7.16087255517145e-06}
{"train_loss": 0.050049662590026855, "global_step": 224722, "epoch": 2524, "lr": 7.160573600715226e-06}
{"train_loss": 0.039464861154556274, "global_step": 224723, "epoch": 2524, "lr": 7.160274652018251e-06}
{"train_loss": 0.050845239342849576, "global_step": 224724, "epoch": 2524, "lr": 7.159975709080546e-06, "val_loss": 9.195058822631836}
{"train_loss": 0.08502570539712906, "global_step": 224725, "epoch": 2525, "lr": 7.159676771902169e-06}
{"train_loss": 0.03280537948012352, "global_step": 224726, "epoch": 2525, "lr": 7.159377840483145e-06}
{"train_loss": 0.06942937523126602, "global_step": 224727, "epoch": 2525, "lr": 7.1590789148235246e-06}
{"train_loss": 0.04633583128452301, "global_step": 224728, "epoch": 2525, "lr": 7.158779994923337e-06}
{"train_loss": 0.04911213368177414, "global_step": 224729, "epoch": 2525, "lr": 7.158481080782642e-06}
{"train_loss": 0.04236464947462082, "global_step": 224730, "epoch": 2525, "lr": 7.158182172401456e-06}
{"train_loss": 0.052753984928131104, "global_step": 224731, "epoch": 2525, "lr": 7.157883269779836e-06}
{"train_loss": 0.06856726855039597, "global_step": 224732, "epoch": 2525, "lr": 7.157584372917808e-06}
{"train_loss": 0.04726114869117737, "global_step": 224733, "epoch": 2525, "lr": 7.157285481815435e-06}
{"train_loss": 0.028575638309121132, "global_step": 224734, "epoch": 2525, "lr": 7.156986596472742e-06}
{"train_loss": 0.07448771595954895, "global_step": 224735, "epoch": 2525, "lr": 7.156687716889754e-06}
{"train_loss": 0.06376862525939941, "global_step": 224736, "epoch": 2525, "lr": 7.156388843066541e-06}
{"train_loss": 0.07596749067306519, "global_step": 224737, "epoch": 2525, "lr": 7.156089975003122e-06}
{"train_loss": 0.094776451587677, "global_step": 224738, "epoch": 2525, "lr": 7.15579111269955e-06}
{"train_loss": 0.03059873729944229, "global_step": 224739, "epoch": 2525, "lr": 7.155492256155849e-06}
{"train_loss": 0.03991846367716789, "global_step": 224740, "epoch": 2525, "lr": 7.155193405372085e-06}
{"train_loss": 0.02012595348060131, "global_step": 224741, "epoch": 2525, "lr": 7.154894560348268e-06}
{"train_loss": 0.05642813444137573, "global_step": 224742, "epoch": 2525, "lr": 7.154595721084467e-06}
{"train_loss": 0.05982199311256409, "global_step": 224743, "epoch": 2525, "lr": 7.154296887580697e-06}
{"train_loss": 0.06486019492149353, "global_step": 224744, "epoch": 2525, "lr": 7.15399805983702e-06}
{"train_loss": 0.051822762936353683, "global_step": 224745, "epoch": 2525, "lr": 7.15369923785345e-06}
{"train_loss": 0.06316862255334854, "global_step": 224746, "epoch": 2525, "lr": 7.1534004216300635e-06}
{"train_loss": 0.054475706070661545, "global_step": 224747, "epoch": 2525, "lr": 7.153101611166862e-06}
{"train_loss": 0.03452463448047638, "global_step": 224748, "epoch": 2525, "lr": 7.1528028064639155e-06}
{"train_loss": 0.04526740312576294, "global_step": 224749, "epoch": 2525, "lr": 7.152504007521243e-06}
{"train_loss": 0.035722311586141586, "global_step": 224750, "epoch": 2525, "lr": 7.152205214338897e-06}
{"train_loss": 0.027907725423574448, "global_step": 224751, "epoch": 2525, "lr": 7.151906426916921e-06}
{"train_loss": 0.056463681161403656, "global_step": 224752, "epoch": 2525, "lr": 7.151607645255337e-06}
{"train_loss": 0.10397934168577194, "global_step": 224753, "epoch": 2525, "lr": 7.151308869354212e-06}
{"train_loss": 0.04927326738834381, "global_step": 224754, "epoch": 2525, "lr": 7.151010099213562e-06}
{"train_loss": 0.04053092002868652, "global_step": 224755, "epoch": 2525, "lr": 7.150711334833443e-06}
{"train_loss": 0.02024013176560402, "global_step": 224756, "epoch": 2525, "lr": 7.1504125762138776e-06}
{"train_loss": 0.04944630339741707, "global_step": 224757, "epoch": 2525, "lr": 7.150113823354932e-06}
{"train_loss": 0.052276596426963806, "global_step": 224758, "epoch": 2525, "lr": 7.149815076256616e-06}
{"train_loss": 0.05131860077381134, "global_step": 224759, "epoch": 2525, "lr": 7.149516334918999e-06}
{"train_loss": 0.06878127157688141, "global_step": 224760, "epoch": 2525, "lr": 7.1492175993420955e-06}
{"train_loss": 0.050222594290971756, "global_step": 224761, "epoch": 2525, "lr": 7.148918869525967e-06}
{"train_loss": 0.05776292458176613, "global_step": 224762, "epoch": 2525, "lr": 7.148620145470636e-06}
{"train_loss": 0.07000257819890976, "global_step": 224763, "epoch": 2525, "lr": 7.148321427176158e-06}
{"train_loss": 0.04710318148136139, "global_step": 224764, "epoch": 2525, "lr": 7.148022714642555e-06}
{"train_loss": 0.01684744656085968, "global_step": 224765, "epoch": 2525, "lr": 7.147724007869894e-06}
{"train_loss": 0.024480201303958893, "global_step": 224766, "epoch": 2525, "lr": 7.1474253068581795e-06}
{"train_loss": 0.04754098877310753, "global_step": 224767, "epoch": 2525, "lr": 7.147126611607491e-06}
{"train_loss": 0.03608859330415726, "global_step": 224768, "epoch": 2525, "lr": 7.146827922117833e-06}
{"train_loss": 0.026041017845273018, "global_step": 224769, "epoch": 2525, "lr": 7.1465292383892765e-06}
{"train_loss": 0.046392183750867844, "global_step": 224770, "epoch": 2525, "lr": 7.146230560421835e-06}
{"train_loss": 0.045179400593042374, "global_step": 224771, "epoch": 2525, "lr": 7.145931888215568e-06}
{"train_loss": 0.08881103247404099, "global_step": 224772, "epoch": 2525, "lr": 7.145633221770498e-06}
{"train_loss": 0.05094967782497406, "global_step": 224773, "epoch": 2525, "lr": 7.145334561086692e-06}
{"train_loss": 0.06818144023418427, "global_step": 224774, "epoch": 2525, "lr": 7.145035906164166e-06}
{"train_loss": 0.026322945952415466, "global_step": 224775, "epoch": 2525, "lr": 7.144737257002959e-06}
{"train_loss": 0.03854149207472801, "global_step": 224776, "epoch": 2525, "lr": 7.144438613603133e-06}
{"train_loss": 0.07977813482284546, "global_step": 224777, "epoch": 2525, "lr": 7.144139975964703e-06}
{"train_loss": 0.020949721336364746, "global_step": 224778, "epoch": 2525, "lr": 7.143841344087732e-06}
{"train_loss": 0.05179525911808014, "global_step": 224779, "epoch": 2525, "lr": 7.143542717972235e-06}
{"train_loss": 0.06496776640415192, "global_step": 224780, "epoch": 2525, "lr": 7.143244097618285e-06}
{"train_loss": 0.06657690554857254, "global_step": 224781, "epoch": 2525, "lr": 7.142945483025887e-06}
{"train_loss": 0.05629442632198334, "global_step": 224782, "epoch": 2525, "lr": 7.142646874195108e-06}
{"train_loss": 0.08229771256446838, "global_step": 224783, "epoch": 2525, "lr": 7.1423482711259645e-06}
{"train_loss": 0.054360754787921906, "global_step": 224784, "epoch": 2525, "lr": 7.142049673818529e-06}
{"train_loss": 0.026595786213874817, "global_step": 224785, "epoch": 2525, "lr": 7.1417510822728015e-06}
{"train_loss": 0.05707676708698273, "global_step": 224786, "epoch": 2525, "lr": 7.141452496488849e-06}
{"train_loss": 0.08169054985046387, "global_step": 224787, "epoch": 2525, "lr": 7.14115391646672e-06}
{"train_loss": 0.020050613209605217, "global_step": 224788, "epoch": 2525, "lr": 7.140855342206426e-06}
{"train_loss": 0.07196402549743652, "global_step": 224789, "epoch": 2525, "lr": 7.140556773708029e-06}
{"train_loss": 0.0173029862344265, "global_step": 224790, "epoch": 2525, "lr": 7.140258210971557e-06}
{"train_loss": 0.09918313473463058, "global_step": 224791, "epoch": 2525, "lr": 7.139959653997064e-06}
{"train_loss": 0.09365233778953552, "global_step": 224792, "epoch": 2525, "lr": 7.139661102784567e-06}
{"train_loss": 0.030975280329585075, "global_step": 224793, "epoch": 2525, "lr": 7.1393625573341335e-06}
{"train_loss": 0.013854864053428173, "global_step": 224794, "epoch": 2525, "lr": 7.13906401764578e-06}
{"train_loss": 0.04083380103111267, "global_step": 224795, "epoch": 2525, "lr": 7.138765483719568e-06}
{"train_loss": 0.04168194532394409, "global_step": 224796, "epoch": 2525, "lr": 7.138466955555511e-06}
{"train_loss": 0.07655347883701324, "global_step": 224797, "epoch": 2525, "lr": 7.1381684331536804e-06}
{"train_loss": 0.03637814521789551, "global_step": 224798, "epoch": 2525, "lr": 7.137869916514089e-06}
{"train_loss": 0.03159581497311592, "global_step": 224799, "epoch": 2525, "lr": 7.137571405636801e-06}
{"train_loss": 0.07713565230369568, "global_step": 224800, "epoch": 2525, "lr": 7.13727290052183e-06}
{"train_loss": 0.05707467719912529, "global_step": 224801, "epoch": 2525, "lr": 7.136974401169244e-06}
{"train_loss": 0.057251881808042526, "global_step": 224802, "epoch": 2525, "lr": 7.136675907579054e-06}
{"train_loss": 0.0562584362924099, "global_step": 224803, "epoch": 2525, "lr": 7.136377419751322e-06}
{"train_loss": 0.06585519760847092, "global_step": 224804, "epoch": 2525, "lr": 7.136078937686086e-06}
{"train_loss": 0.04343532770872116, "global_step": 224805, "epoch": 2525, "lr": 7.1357804613833736e-06}
{"train_loss": 0.026458723470568657, "global_step": 224806, "epoch": 2525, "lr": 7.135481990843246e-06}
{"train_loss": 0.023972461000084877, "global_step": 224807, "epoch": 2525, "lr": 7.135183526065731e-06}
{"train_loss": 0.0693967342376709, "global_step": 224808, "epoch": 2525, "lr": 7.13488506705085e-06}
{"train_loss": 0.04985836520791054, "global_step": 224809, "epoch": 2525, "lr": 7.134586613798677e-06}
{"train_loss": 0.06513701379299164, "global_step": 224810, "epoch": 2525, "lr": 7.134288166309227e-06}
{"train_loss": 0.10326164960861206, "global_step": 224811, "epoch": 2525, "lr": 7.133989724582557e-06}
{"train_loss": 0.028043869882822037, "global_step": 224812, "epoch": 2525, "lr": 7.133691288618705e-06}
{"train_loss": 0.052477885885268784, "global_step": 224813, "epoch": 2525, "lr": 7.133392858417693e-06, "val_loss": 9.078961372375488, "train_action_mse_error": 21.507919311523438}
{"train_loss": 0.05994206666946411, "global_step": 224814, "epoch": 2526, "lr": 7.133094433979582e-06}
{"train_loss": 0.07814392447471619, "global_step": 224815, "epoch": 2526, "lr": 7.132796015304394e-06}
{"train_loss": 0.04122477024793625, "global_step": 224816, "epoch": 2526, "lr": 7.132497602392191e-06}
{"train_loss": 0.02898179367184639, "global_step": 224817, "epoch": 2526, "lr": 7.132199195242989e-06}
{"train_loss": 0.04778248071670532, "global_step": 224818, "epoch": 2526, "lr": 7.131900793856849e-06}
{"train_loss": 0.062369801104068756, "global_step": 224819, "epoch": 2526, "lr": 7.131602398233794e-06}
{"train_loss": 0.08584751933813095, "global_step": 224820, "epoch": 2526, "lr": 7.1313040083738845e-06}
{"train_loss": 0.048791926354169846, "global_step": 224821, "epoch": 2526, "lr": 7.131005624277137e-06}
{"train_loss": 0.044892605394124985, "global_step": 224822, "epoch": 2526, "lr": 7.130707245943602e-06}
{"train_loss": 0.047744475305080414, "global_step": 224823, "epoch": 2526, "lr": 7.130408873373335e-06}
{"train_loss": 0.06210920587182045, "global_step": 224824, "epoch": 2526, "lr": 7.130110506566345e-06}
{"train_loss": 0.03274223953485489, "global_step": 224825, "epoch": 2526, "lr": 7.129812145522708e-06}
{"train_loss": 0.15382800996303558, "global_step": 224826, "epoch": 2526, "lr": 7.129513790242431e-06}
{"train_loss": 0.05175846070051193, "global_step": 224827, "epoch": 2526, "lr": 7.129215440725578e-06}
{"train_loss": 0.10802831500768661, "global_step": 224828, "epoch": 2526, "lr": 7.128917096972171e-06}
{"train_loss": 0.05382981896400452, "global_step": 224829, "epoch": 2526, "lr": 7.128618758982264e-06}
{"train_loss": 0.0240954402834177, "global_step": 224830, "epoch": 2526, "lr": 7.128320426755886e-06}
{"train_loss": 0.018472326919436455, "global_step": 224831, "epoch": 2526, "lr": 7.1280221002930915e-06}
{"train_loss": 0.03602047637104988, "global_step": 224832, "epoch": 2526, "lr": 7.1277237795939034e-06}
{"train_loss": 0.10835811495780945, "global_step": 224833, "epoch": 2526, "lr": 7.127425464658377e-06}
{"train_loss": 0.06867528706789017, "global_step": 224834, "epoch": 2526, "lr": 7.12712715548654e-06}
{"train_loss": 0.023272404447197914, "global_step": 224835, "epoch": 2526, "lr": 7.126828852078449e-06}
{"train_loss": 0.09732925146818161, "global_step": 224836, "epoch": 2526, "lr": 7.126530554434119e-06}
{"train_loss": 0.041912250220775604, "global_step": 224837, "epoch": 2526, "lr": 7.126232262553611e-06}
{"train_loss": 0.032472457736730576, "global_step": 224838, "epoch": 2526, "lr": 7.125933976436955e-06}
{"train_loss": 0.09708770364522934, "global_step": 224839, "epoch": 2526, "lr": 7.125635696084204e-06}
{"train_loss": 0.0352984182536602, "global_step": 224840, "epoch": 2526, "lr": 7.125337421495376e-06}
{"train_loss": 0.05111246928572655, "global_step": 224841, "epoch": 2526, "lr": 7.125039152670527e-06}
{"train_loss": 0.07283768057823181, "global_step": 224842, "epoch": 2526, "lr": 7.124740889609705e-06}
{"train_loss": 0.037542205303907394, "global_step": 224843, "epoch": 2526, "lr": 7.124442632312922e-06}
{"train_loss": 0.029623551294207573, "global_step": 224844, "epoch": 2526, "lr": 7.124144380780251e-06}
{"train_loss": 0.029065387323498726, "global_step": 224845, "epoch": 2526, "lr": 7.123846135011713e-06}
{"train_loss": 0.04680364951491356, "global_step": 224846, "epoch": 2526, "lr": 7.123547895007343e-06}
{"train_loss": 0.07584081590175629, "global_step": 224847, "epoch": 2526, "lr": 7.123249660767201e-06}
{"train_loss": 0.05134289339184761, "global_step": 224848, "epoch": 2526, "lr": 7.1229514322913025e-06}
{"train_loss": 0.08201523125171661, "global_step": 224849, "epoch": 2526, "lr": 7.122653209579716e-06}
{"train_loss": 0.11497770994901657, "global_step": 224850, "epoch": 2526, "lr": 7.1223549926324625e-06}
{"train_loss": 0.023335600271821022, "global_step": 224851, "epoch": 2526, "lr": 7.122056781449571e-06}
{"train_loss": 0.04586353525519371, "global_step": 224852, "epoch": 2526, "lr": 7.121758576031112e-06}
{"train_loss": 0.06486097723245621, "global_step": 224853, "epoch": 2526, "lr": 7.121460376377098e-06}
{"train_loss": 0.03770316764712334, "global_step": 224854, "epoch": 2526, "lr": 7.121162182487595e-06}
{"train_loss": 0.059604521840810776, "global_step": 224855, "epoch": 2526, "lr": 7.120863994362614e-06}
{"train_loss": 0.026185283437371254, "global_step": 224856, "epoch": 2526, "lr": 7.120565812002222e-06}
{"train_loss": 0.05090939998626709, "global_step": 224857, "epoch": 2526, "lr": 7.120267635406436e-06}
{"train_loss": 0.03308742865920067, "global_step": 224858, "epoch": 2526, "lr": 7.1199694645753105e-06}
{"train_loss": 0.04677150771021843, "global_step": 224859, "epoch": 2526, "lr": 7.11967129950889e-06}
{"train_loss": 0.10482273250818253, "global_step": 224860, "epoch": 2526, "lr": 7.119373140207197e-06}
{"train_loss": 0.06458596140146255, "global_step": 224861, "epoch": 2526, "lr": 7.1190749866702934e-06}
{"train_loss": 0.04825354367494583, "global_step": 224862, "epoch": 2526, "lr": 7.118776838898195e-06}
{"train_loss": 0.03588221222162247, "global_step": 224863, "epoch": 2526, "lr": 7.118478696890968e-06}
{"train_loss": 0.03581370785832405, "global_step": 224864, "epoch": 2526, "lr": 7.1181805606486245e-06}
{"train_loss": 0.03317055478692055, "global_step": 224865, "epoch": 2526, "lr": 7.117882430171235e-06}
{"train_loss": 0.06381253153085709, "global_step": 224866, "epoch": 2526, "lr": 7.117584305458808e-06}
{"train_loss": 0.03955024853348732, "global_step": 224867, "epoch": 2526, "lr": 7.117286186511412e-06}
{"train_loss": 0.07619630545377731, "global_step": 224868, "epoch": 2526, "lr": 7.116988073329061e-06}
{"train_loss": 0.07340008020401001, "global_step": 224869, "epoch": 2526, "lr": 7.1166899659118215e-06}
{"train_loss": 0.036794621497392654, "global_step": 224870, "epoch": 2526, "lr": 7.116391864259708e-06}
{"train_loss": 0.016697892919182777, "global_step": 224871, "epoch": 2526, "lr": 7.116093768372783e-06}
{"train_loss": 0.04135817661881447, "global_step": 224872, "epoch": 2526, "lr": 7.115795678251069e-06}
{"train_loss": 0.09156010299921036, "global_step": 224873, "epoch": 2526, "lr": 7.115497593894621e-06}
{"train_loss": 0.09554801136255264, "global_step": 224874, "epoch": 2526, "lr": 7.115199515303467e-06}
{"train_loss": 0.032747674733400345, "global_step": 224875, "epoch": 2526, "lr": 7.1149014424776565e-06}
{"train_loss": 0.017960596829652786, "global_step": 224876, "epoch": 2526, "lr": 7.114603375417217e-06}
{"train_loss": 0.07225003093481064, "global_step": 224877, "epoch": 2526, "lr": 7.114305314122194e-06}
{"train_loss": 0.06960276514291763, "global_step": 224878, "epoch": 2526, "lr": 7.114007258592642e-06}
{"train_loss": 0.05829768255352974, "global_step": 224879, "epoch": 2526, "lr": 7.113709208828578e-06}
{"train_loss": 0.03249181807041168, "global_step": 224880, "epoch": 2526, "lr": 7.113411164830064e-06}
{"train_loss": 0.07980558276176453, "global_step": 224881, "epoch": 2526, "lr": 7.113113126597115e-06}
{"train_loss": 0.04071534052491188, "global_step": 224882, "epoch": 2526, "lr": 7.112815094129799e-06}
{"train_loss": 0.05740106850862503, "global_step": 224883, "epoch": 2526, "lr": 7.112517067428131e-06}
{"train_loss": 0.057006750255823135, "global_step": 224884, "epoch": 2526, "lr": 7.1122190464921745e-06}
{"train_loss": 0.057737477123737335, "global_step": 224885, "epoch": 2526, "lr": 7.111921031321961e-06}
{"train_loss": 0.045764777809381485, "global_step": 224886, "epoch": 2526, "lr": 7.111623021917513e-06}
{"train_loss": 0.03858228027820587, "global_step": 224887, "epoch": 2526, "lr": 7.1113250182788914e-06}
{"train_loss": 0.04729742184281349, "global_step": 224888, "epoch": 2526, "lr": 7.111027020406119e-06}
{"train_loss": 0.036750759929418564, "global_step": 224889, "epoch": 2526, "lr": 7.110729028299262e-06}
{"train_loss": 0.034436874091625214, "global_step": 224890, "epoch": 2526, "lr": 7.110431041958343e-06}
{"train_loss": 0.051590513437986374, "global_step": 224891, "epoch": 2526, "lr": 7.110133061383395e-06}
{"train_loss": 0.07966740429401398, "global_step": 224892, "epoch": 2526, "lr": 7.1098350865744735e-06}
{"train_loss": 0.054321739822626114, "global_step": 224893, "epoch": 2526, "lr": 7.1095371175316006e-06}
{"train_loss": 0.08959819376468658, "global_step": 224894, "epoch": 2526, "lr": 7.109239154254832e-06}
{"train_loss": 0.05742420256137848, "global_step": 224895, "epoch": 2526, "lr": 7.108941196744212e-06}
{"train_loss": 0.06030799448490143, "global_step": 224896, "epoch": 2526, "lr": 7.108643244999758e-06}
{"train_loss": 0.04485552757978439, "global_step": 224897, "epoch": 2526, "lr": 7.10834529902154e-06}
{"train_loss": 0.026792192831635475, "global_step": 224898, "epoch": 2526, "lr": 7.108047358809572e-06}
{"train_loss": 0.04451654106378555, "global_step": 224899, "epoch": 2526, "lr": 7.107749424363908e-06}
{"train_loss": 0.08512770384550095, "global_step": 224900, "epoch": 2526, "lr": 7.1074514956845805e-06}
{"train_loss": 0.05119398981332779, "global_step": 224901, "epoch": 2526, "lr": 7.107153572771641e-06}
{"train_loss": 0.054680253190689546, "global_step": 224902, "epoch": 2526, "lr": 7.106855655625111e-06, "val_loss": 8.960192680358887}
{"train_loss": 0.048922982066869736, "global_step": 224903, "epoch": 2527, "lr": 7.106557744245057e-06}
{"train_loss": 0.048552509397268295, "global_step": 224904, "epoch": 2527, "lr": 7.106259838631485e-06}
{"train_loss": 0.051974788308143616, "global_step": 224905, "epoch": 2527, "lr": 7.105961938784472e-06}
{"train_loss": 0.03383389487862587, "global_step": 224906, "epoch": 2527, "lr": 7.105664044704025e-06}
{"train_loss": 0.057582829147577286, "global_step": 224907, "epoch": 2527, "lr": 7.105366156390209e-06}
{"train_loss": 0.05264338105916977, "global_step": 224908, "epoch": 2527, "lr": 7.10506827384304e-06}
{"train_loss": 0.061646394431591034, "global_step": 224909, "epoch": 2527, "lr": 7.1047703970625875e-06}
{"train_loss": 0.02824006788432598, "global_step": 224910, "epoch": 2527, "lr": 7.104472526048861e-06}
{"train_loss": 0.043950315564870834, "global_step": 224911, "epoch": 2527, "lr": 7.104174660801932e-06}
{"train_loss": 0.06581708043813705, "global_step": 224912, "epoch": 2527, "lr": 7.103876801321807e-06}
{"train_loss": 0.05220666155219078, "global_step": 224913, "epoch": 2527, "lr": 7.103578947608547e-06}
{"train_loss": 0.05746928229928017, "global_step": 224914, "epoch": 2527, "lr": 7.1032810996621955e-06}
{"train_loss": 0.03702615201473236, "global_step": 224915, "epoch": 2527, "lr": 7.102983257482776e-06}
{"train_loss": 0.043924231082201004, "global_step": 224916, "epoch": 2527, "lr": 7.102685421070349e-06}
{"train_loss": 0.07428430765867233, "global_step": 224917, "epoch": 2527, "lr": 7.102387590424936e-06}
{"train_loss": 0.04431404545903206, "global_step": 224918, "epoch": 2527, "lr": 7.102089765546588e-06}
{"train_loss": 0.059339746832847595, "global_step": 224919, "epoch": 2527, "lr": 7.101791946435332e-06}
{"train_loss": 0.028858954086899757, "global_step": 224920, "epoch": 2527, "lr": 7.10149413309123e-06}
{"train_loss": 0.05806837975978851, "global_step": 224921, "epoch": 2527, "lr": 7.101196325514298e-06}
{"train_loss": 0.04404168948531151, "global_step": 224922, "epoch": 2527, "lr": 7.100898523704597e-06}
{"train_loss": 0.0695662647485733, "global_step": 224923, "epoch": 2527, "lr": 7.10060072766216e-06}
{"train_loss": 0.0459892675280571, "global_step": 224924, "epoch": 2527, "lr": 7.10030293738701e-06}
{"train_loss": 0.023200692608952522, "global_step": 224925, "epoch": 2527, "lr": 7.100005152879213e-06}
{"train_loss": 0.027292199432849884, "global_step": 224926, "epoch": 2527, "lr": 7.099707374138786e-06}
{"train_loss": 0.018553264439105988, "global_step": 224927, "epoch": 2527, "lr": 7.099409601165796e-06}
{"train_loss": 0.05291922762989998, "global_step": 224928, "epoch": 2527, "lr": 7.099111833960265e-06}
{"train_loss": 0.06131221354007721, "global_step": 224929, "epoch": 2527, "lr": 7.098814072522225e-06}
{"train_loss": 0.06073440611362457, "global_step": 224930, "epoch": 2527, "lr": 7.098516316851722e-06}
{"train_loss": 0.07454828917980194, "global_step": 224931, "epoch": 2527, "lr": 7.0982185669488165e-06}
{"train_loss": 0.024467801675200462, "global_step": 224932, "epoch": 2527, "lr": 7.09792082281352e-06}
{"train_loss": 0.02950129099190235, "global_step": 224933, "epoch": 2527, "lr": 7.097623084445898e-06}
{"train_loss": 0.05829353630542755, "global_step": 224934, "epoch": 2527, "lr": 7.097325351845962e-06}
{"train_loss": 0.06517025828361511, "global_step": 224935, "epoch": 2527, "lr": 7.09702762501378e-06}
{"train_loss": 0.05569859594106674, "global_step": 224936, "epoch": 2527, "lr": 7.096729903949373e-06}
{"train_loss": 0.055698711425065994, "global_step": 224937, "epoch": 2527, "lr": 7.096432188652796e-06}
{"train_loss": 0.08056370168924332, "global_step": 224938, "epoch": 2527, "lr": 7.0961344791240666e-06}
{"train_loss": 0.03961700201034546, "global_step": 224939, "epoch": 2527, "lr": 7.095836775363257e-06}
{"train_loss": 0.07557868957519531, "global_step": 224940, "epoch": 2527, "lr": 7.095539077370372e-06}
{"train_loss": 0.07593302428722382, "global_step": 224941, "epoch": 2527, "lr": 7.095241385145485e-06}
{"train_loss": 0.014548240229487419, "global_step": 224942, "epoch": 2527, "lr": 7.094943698688605e-06}
{"train_loss": 0.05009595677256584, "global_step": 224943, "epoch": 2527, "lr": 7.0946460179997955e-06}
{"train_loss": 0.0725979134440422, "global_step": 224944, "epoch": 2527, "lr": 7.094348343079082e-06}
{"train_loss": 0.0695265457034111, "global_step": 224945, "epoch": 2527, "lr": 7.094050673926522e-06}
{"train_loss": 0.06498786807060242, "global_step": 224946, "epoch": 2527, "lr": 7.093753010542131e-06}
{"train_loss": 0.05369487777352333, "global_step": 224947, "epoch": 2527, "lr": 7.09345535292597e-06}
{"train_loss": 0.07209204137325287, "global_step": 224948, "epoch": 2527, "lr": 7.0931577010780624e-06}
{"train_loss": 0.03190423548221588, "global_step": 224949, "epoch": 2527, "lr": 7.092860054998457e-06}
{"train_loss": 0.02454560436308384, "global_step": 224950, "epoch": 2527, "lr": 7.092562414687204e-06}
{"train_loss": 0.03175130859017372, "global_step": 224951, "epoch": 2527, "lr": 7.0922647801443255e-06}
{"train_loss": 0.055496904999017715, "global_step": 224952, "epoch": 2527, "lr": 7.091967151369877e-06}
{"train_loss": 0.05085470899939537, "global_step": 224953, "epoch": 2527, "lr": 7.091669528363881e-06}
{"train_loss": 0.0293369609862566, "global_step": 224954, "epoch": 2527, "lr": 7.091371911126393e-06}
{"train_loss": 0.021831531077623367, "global_step": 224955, "epoch": 2527, "lr": 7.0910742996574405e-06}
{"train_loss": 0.031095579266548157, "global_step": 224956, "epoch": 2527, "lr": 7.0907766939570795e-06}
{"train_loss": 0.06291715055704117, "global_step": 224957, "epoch": 2527, "lr": 7.090479094025332e-06}
{"train_loss": 0.06488938629627228, "global_step": 224958, "epoch": 2527, "lr": 7.090181499862259e-06}
{"train_loss": 0.022044429555535316, "global_step": 224959, "epoch": 2527, "lr": 7.089883911467871e-06}
{"train_loss": 0.029289092868566513, "global_step": 224960, "epoch": 2527, "lr": 7.089586328842241e-06}
{"train_loss": 0.046652939170598984, "global_step": 224961, "epoch": 2527, "lr": 7.08928875198539e-06}
{"train_loss": 0.04338432103395462, "global_step": 224962, "epoch": 2527, "lr": 7.088991180897353e-06}
{"train_loss": 0.04961572587490082, "global_step": 224963, "epoch": 2527, "lr": 7.0886936155781845e-06}
{"train_loss": 0.016733981668949127, "global_step": 224964, "epoch": 2527, "lr": 7.088396056027913e-06}
{"train_loss": 0.0905364453792572, "global_step": 224965, "epoch": 2527, "lr": 7.088098502246593e-06}
{"train_loss": 0.032209981232881546, "global_step": 224966, "epoch": 2527, "lr": 7.087800954234253e-06}
{"train_loss": 0.09301106631755829, "global_step": 224967, "epoch": 2527, "lr": 7.087503411990925e-06}
{"train_loss": 0.06313833594322205, "global_step": 224968, "epoch": 2527, "lr": 7.087205875516656e-06}
{"train_loss": 0.02745712362229824, "global_step": 224969, "epoch": 2527, "lr": 7.086908344811505e-06}
{"train_loss": 0.0757874995470047, "global_step": 224970, "epoch": 2527, "lr": 7.0866108198754835e-06}
{"train_loss": 0.03530363738536835, "global_step": 224971, "epoch": 2527, "lr": 7.086313300708658e-06}
{"train_loss": 0.01556486077606678, "global_step": 224972, "epoch": 2527, "lr": 7.08601578731104e-06}
{"train_loss": 0.048100173473358154, "global_step": 224973, "epoch": 2527, "lr": 7.085718279682696e-06}
{"train_loss": 0.024463875219225883, "global_step": 224974, "epoch": 2527, "lr": 7.085420777823642e-06}
{"train_loss": 0.013880367390811443, "global_step": 224975, "epoch": 2527, "lr": 7.085123281733941e-06}
{"train_loss": 0.05777771398425102, "global_step": 224976, "epoch": 2527, "lr": 7.0848257914136125e-06}
{"train_loss": 0.048011694103479385, "global_step": 224977, "epoch": 2527, "lr": 7.08452830686272e-06}
{"train_loss": 0.039950497448444366, "global_step": 224978, "epoch": 2527, "lr": 7.084230828081273e-06}
{"train_loss": 0.05054238438606262, "global_step": 224979, "epoch": 2527, "lr": 7.083933355069344e-06}
{"train_loss": 0.04709789901971817, "global_step": 224980, "epoch": 2527, "lr": 7.083635887826939e-06}
{"train_loss": 0.08268601447343826, "global_step": 224981, "epoch": 2527, "lr": 7.083338426354136e-06}
{"train_loss": 0.08378678560256958, "global_step": 224982, "epoch": 2527, "lr": 7.08304097065094e-06}
{"train_loss": 0.05973321571946144, "global_step": 224983, "epoch": 2527, "lr": 7.082743520717416e-06}
{"train_loss": 0.04847674071788788, "global_step": 224984, "epoch": 2527, "lr": 7.0824460765535835e-06}
{"train_loss": 0.06388981640338898, "global_step": 224985, "epoch": 2527, "lr": 7.082148638159491e-06}
{"train_loss": 0.02026911824941635, "global_step": 224986, "epoch": 2527, "lr": 7.081851205535195e-06}
{"train_loss": 0.05982015281915665, "global_step": 224987, "epoch": 2527, "lr": 7.08155377868071e-06}
{"train_loss": 0.018190857023000717, "global_step": 224988, "epoch": 2527, "lr": 7.0812563575961e-06}
{"train_loss": 0.03159602731466293, "global_step": 224989, "epoch": 2527, "lr": 7.080958942281374e-06}
{"train_loss": 0.04830854758620262, "global_step": 224990, "epoch": 2527, "lr": 7.080661532736604e-06}
{"train_loss": 0.04839116212482868, "global_step": 224991, "epoch": 2527, "lr": 7.080364128961808e-06, "val_loss": 9.099309921264648}
{"train_loss": 0.09328094869852066, "global_step": 224992, "epoch": 2528, "lr": 7.080066730957041e-06}
{"train_loss": 0.03180963173508644, "global_step": 224993, "epoch": 2528, "lr": 7.079769338722331e-06}
{"train_loss": 0.03676992282271385, "global_step": 224994, "epoch": 2528, "lr": 7.079471952257727e-06}
{"train_loss": 0.07283204793930054, "global_step": 224995, "epoch": 2528, "lr": 7.0791745715632575e-06}
{"train_loss": 0.043990183621644974, "global_step": 224996, "epoch": 2528, "lr": 7.078877196638978e-06}
{"train_loss": 0.04449518024921417, "global_step": 224997, "epoch": 2528, "lr": 7.07857982748491e-06}
{"train_loss": 0.14230696856975555, "global_step": 224998, "epoch": 2528, "lr": 7.078282464101116e-06}
{"train_loss": 0.060713764280080795, "global_step": 224999, "epoch": 2528, "lr": 7.077985106487617e-06}
{"train_loss": 0.07947045564651489, "global_step": 225000, "epoch": 2528, "lr": 7.077687754644463e-06}
{"train_loss": 0.058564599603414536, "global_step": 225001, "epoch": 2528, "lr": 7.077390408571693e-06}
{"train_loss": 0.02745414711534977, "global_step": 225002, "epoch": 2528, "lr": 7.0770930682693355e-06}
{"train_loss": 0.07676592469215393, "global_step": 225003, "epoch": 2528, "lr": 7.076795733737452e-06}
{"train_loss": 0.04811164364218712, "global_step": 225004, "epoch": 2528, "lr": 7.076498404976056e-06}
{"train_loss": 0.04700420796871185, "global_step": 225005, "epoch": 2528, "lr": 7.076201081985217e-06}
{"train_loss": 0.04374713450670242, "global_step": 225006, "epoch": 2528, "lr": 7.075903764764946e-06}
{"train_loss": 0.1278318613767624, "global_step": 225007, "epoch": 2528, "lr": 7.075606453315303e-06}
{"train_loss": 0.05077119544148445, "global_step": 225008, "epoch": 2528, "lr": 7.075309147636317e-06}
{"train_loss": 0.0685630664229393, "global_step": 225009, "epoch": 2528, "lr": 7.075011847728041e-06}
{"train_loss": 0.06387786567211151, "global_step": 225010, "epoch": 2528, "lr": 7.074714553590495e-06}
{"train_loss": 0.04381667077541351, "global_step": 225011, "epoch": 2528, "lr": 7.074417265223743e-06}
{"train_loss": 0.05108889564871788, "global_step": 225012, "epoch": 2528, "lr": 7.074119982627803e-06}
{"train_loss": 0.09626929461956024, "global_step": 225013, "epoch": 2528, "lr": 7.073822705802729e-06}
{"train_loss": 0.06298520416021347, "global_step": 225014, "epoch": 2528, "lr": 7.073525434748552e-06}
{"train_loss": 0.048526786267757416, "global_step": 225015, "epoch": 2528, "lr": 7.073228169465324e-06}
{"train_loss": 0.043841395527124405, "global_step": 225016, "epoch": 2528, "lr": 7.072930909953063e-06}
{"train_loss": 0.034292977303266525, "global_step": 225017, "epoch": 2528, "lr": 7.072633656211841e-06}
{"train_loss": 0.024950966238975525, "global_step": 225018, "epoch": 2528, "lr": 7.072336408241664e-06}
{"train_loss": 0.045924704521894455, "global_step": 225019, "epoch": 2528, "lr": 7.072039166042604e-06}
{"train_loss": 0.056083567440509796, "global_step": 225020, "epoch": 2528, "lr": 7.071741929614673e-06}
{"train_loss": 0.03552400693297386, "global_step": 225021, "epoch": 2528, "lr": 7.071444698957919e-06}
{"train_loss": 0.03371957689523697, "global_step": 225022, "epoch": 2528, "lr": 7.0711474740723984e-06}
{"train_loss": 0.040212709456682205, "global_step": 225023, "epoch": 2528, "lr": 7.070850254958128e-06}
{"train_loss": 0.045235853642225266, "global_step": 225024, "epoch": 2528, "lr": 7.0705530416151745e-06}
{"train_loss": 0.046935465186834335, "global_step": 225025, "epoch": 2528, "lr": 7.070255834043543e-06}
{"train_loss": 0.01995469443500042, "global_step": 225026, "epoch": 2528, "lr": 7.069958632243307e-06}
{"train_loss": 0.018370701000094414, "global_step": 225027, "epoch": 2528, "lr": 7.069661436214481e-06}
{"train_loss": 0.07581839710474014, "global_step": 225028, "epoch": 2528, "lr": 7.069364245957127e-06}
{"train_loss": 0.050950031727552414, "global_step": 225029, "epoch": 2528, "lr": 7.0690670614712625e-06}
{"train_loss": 0.06210591644048691, "global_step": 225030, "epoch": 2528, "lr": 7.068769882756953e-06}
{"train_loss": 0.05818236619234085, "global_step": 225031, "epoch": 2528, "lr": 7.0684727098142055e-06}
{"train_loss": 0.051875513046979904, "global_step": 225032, "epoch": 2528, "lr": 7.068175542643096e-06}
{"train_loss": 0.028806012123823166, "global_step": 225033, "epoch": 2528, "lr": 7.067878381243636e-06}
{"train_loss": 0.10437527298927307, "global_step": 225034, "epoch": 2528, "lr": 7.067581225615883e-06}
{"train_loss": 0.07222624868154526, "global_step": 225035, "epoch": 2528, "lr": 7.067284075759861e-06}
{"train_loss": 0.06414363533258438, "global_step": 225036, "epoch": 2528, "lr": 7.0669869316756345e-06}
{"train_loss": 0.027416229248046875, "global_step": 225037, "epoch": 2528, "lr": 7.066689793363213e-06}
{"train_loss": 0.04968078061938286, "global_step": 225038, "epoch": 2528, "lr": 7.066392660822663e-06}
{"train_loss": 0.054981742054224014, "global_step": 225039, "epoch": 2528, "lr": 7.066095534054018e-06}
{"train_loss": 0.07420454174280167, "global_step": 225040, "epoch": 2528, "lr": 7.065798413057295e-06}
{"train_loss": 0.04667791351675987, "global_step": 225041, "epoch": 2528, "lr": 7.065501297832566e-06}
{"train_loss": 0.03852953761816025, "global_step": 225042, "epoch": 2528, "lr": 7.065204188379848e-06}
{"train_loss": 0.057919349521398544, "global_step": 225043, "epoch": 2528, "lr": 7.064907084699196e-06}
{"train_loss": 0.04671046510338783, "global_step": 225044, "epoch": 2528, "lr": 7.064609986790638e-06}
{"train_loss": 0.052796103060245514, "global_step": 225045, "epoch": 2528, "lr": 7.0643128946542294e-06}
{"train_loss": 0.03201209753751755, "global_step": 225046, "epoch": 2528, "lr": 7.064015808289992e-06}
{"train_loss": 0.07794830203056335, "global_step": 225047, "epoch": 2528, "lr": 7.063718727697982e-06}
{"train_loss": 0.06987226009368896, "global_step": 225048, "epoch": 2528, "lr": 7.063421652878222e-06}
{"train_loss": 0.08546633273363113, "global_step": 225049, "epoch": 2528, "lr": 7.063124583830772e-06}
{"train_loss": 0.07922481745481491, "global_step": 225050, "epoch": 2528, "lr": 7.062827520555648e-06}
{"train_loss": 0.013721263967454433, "global_step": 225051, "epoch": 2528, "lr": 7.062530463052919e-06}
{"train_loss": 0.037960879504680634, "global_step": 225052, "epoch": 2528, "lr": 7.062233411322594e-06}
{"train_loss": 0.0429573692381382, "global_step": 225053, "epoch": 2528, "lr": 7.0619363653647466e-06}
{"train_loss": 0.059618543833494186, "global_step": 225054, "epoch": 2528, "lr": 7.061639325179381e-06}
{"train_loss": 0.02061217837035656, "global_step": 225055, "epoch": 2528, "lr": 7.061342290766565e-06}
{"train_loss": 0.06595304608345032, "global_step": 225056, "epoch": 2528, "lr": 7.0610452621263205e-06}
{"train_loss": 0.0749681368470192, "global_step": 225057, "epoch": 2528, "lr": 7.060748239258697e-06}
{"train_loss": 0.021505804732441902, "global_step": 225058, "epoch": 2528, "lr": 7.060451222163739e-06}
{"train_loss": 0.04297590255737305, "global_step": 225059, "epoch": 2528, "lr": 7.060154210841474e-06}
{"train_loss": 0.03176320344209671, "global_step": 225060, "epoch": 2528, "lr": 7.059857205291953e-06}
{"train_loss": 0.08677958697080612, "global_step": 225061, "epoch": 2528, "lr": 7.059560205515203e-06}
{"train_loss": 0.061919763684272766, "global_step": 225062, "epoch": 2528, "lr": 7.05926321151128e-06}
{"train_loss": 0.038775570690631866, "global_step": 225063, "epoch": 2528, "lr": 7.058966223280211e-06}
{"train_loss": 0.06057920679450035, "global_step": 225064, "epoch": 2528, "lr": 7.058669240822046e-06}
{"train_loss": 0.05155652016401291, "global_step": 225065, "epoch": 2528, "lr": 7.058372264136809e-06}
{"train_loss": 0.05270973965525627, "global_step": 225066, "epoch": 2528, "lr": 7.058075293224564e-06}
{"train_loss": 0.019012240692973137, "global_step": 225067, "epoch": 2528, "lr": 7.057778328085324e-06}
{"train_loss": 0.009849632158875465, "global_step": 225068, "epoch": 2528, "lr": 7.057481368719155e-06}
{"train_loss": 0.05871069058775902, "global_step": 225069, "epoch": 2528, "lr": 7.057184415126072e-06}
{"train_loss": 0.06384824961423874, "global_step": 225070, "epoch": 2528, "lr": 7.056887467306139e-06}
{"train_loss": 0.02988472208380699, "global_step": 225071, "epoch": 2528, "lr": 7.0565905252593765e-06}
{"train_loss": 0.08517703413963318, "global_step": 225072, "epoch": 2528, "lr": 7.056293588985835e-06}
{"train_loss": 0.03072272427380085, "global_step": 225073, "epoch": 2528, "lr": 7.0559966584855475e-06}
{"train_loss": 0.04076887667179108, "global_step": 225074, "epoch": 2528, "lr": 7.0556997337585636e-06}
{"train_loss": 0.01788116991519928, "global_step": 225075, "epoch": 2528, "lr": 7.055402814804912e-06}
{"train_loss": 0.02352115884423256, "global_step": 225076, "epoch": 2528, "lr": 7.0551059016246475e-06}
{"train_loss": 0.026325881481170654, "global_step": 225077, "epoch": 2528, "lr": 7.0548089942177984e-06}
{"train_loss": 0.06307164579629898, "global_step": 225078, "epoch": 2528, "lr": 7.054512092584398e-06}
{"train_loss": 0.052755746990442276, "global_step": 225079, "epoch": 2528, "lr": 7.054215196724501e-06}
{"train_loss": 0.05227372262710601, "global_step": 225080, "epoch": 2528, "lr": 7.053918306638136e-06, "val_loss": 8.99466323852539}
{"train_loss": 0.039307236671447754, "global_step": 225081, "epoch": 2529, "lr": 7.053621422325357e-06}
{"train_loss": 0.03426666185259819, "global_step": 225082, "epoch": 2529, "lr": 7.053324543786183e-06}
{"train_loss": 0.05326434224843979, "global_step": 225083, "epoch": 2529, "lr": 7.053027671020679e-06}
{"train_loss": 0.08853935450315475, "global_step": 225084, "epoch": 2529, "lr": 7.052730804028862e-06}
{"train_loss": 0.05562959238886833, "global_step": 225085, "epoch": 2529, "lr": 7.052433942810788e-06}
{"train_loss": 0.07923678308725357, "global_step": 225086, "epoch": 2529, "lr": 7.052137087366484e-06}
{"train_loss": 0.06171610951423645, "global_step": 225087, "epoch": 2529, "lr": 7.051840237696005e-06}
{"train_loss": 0.027077006176114082, "global_step": 225088, "epoch": 2529, "lr": 7.051543393799376e-06}
{"train_loss": 0.029177550226449966, "global_step": 225089, "epoch": 2529, "lr": 7.05124655567665e-06}
{"train_loss": 0.04765462502837181, "global_step": 225090, "epoch": 2529, "lr": 7.050949723327854e-06}
{"train_loss": 0.08450273424386978, "global_step": 225091, "epoch": 2529, "lr": 7.050652896753041e-06}
{"train_loss": 0.014053638093173504, "global_step": 225092, "epoch": 2529, "lr": 7.050356075952236e-06}
{"train_loss": 0.08385176956653595, "global_step": 225093, "epoch": 2529, "lr": 7.050059260925496e-06}
{"train_loss": 0.08039984107017517, "global_step": 225094, "epoch": 2529, "lr": 7.049762451672842e-06}
{"train_loss": 0.10701017081737518, "global_step": 225095, "epoch": 2529, "lr": 7.049465648194325e-06}
{"train_loss": 0.06030677258968353, "global_step": 225096, "epoch": 2529, "lr": 7.0491688504899896e-06}
{"train_loss": 0.09562085568904877, "global_step": 225097, "epoch": 2529, "lr": 7.048872058559863e-06}
{"train_loss": 0.03712991625070572, "global_step": 225098, "epoch": 2529, "lr": 7.048575272404007e-06}
{"train_loss": 0.05839453265070915, "global_step": 225099, "epoch": 2529, "lr": 7.0482784920224305e-06}
{"train_loss": 0.04050690308213234, "global_step": 225100, "epoch": 2529, "lr": 7.047981717415203e-06}
{"train_loss": 0.050350043922662735, "global_step": 225101, "epoch": 2529, "lr": 7.047684948582339e-06}
{"train_loss": 0.04760565236210823, "global_step": 225102, "epoch": 2529, "lr": 7.047388185523901e-06}
{"train_loss": 0.06460744887590408, "global_step": 225103, "epoch": 2529, "lr": 7.047091428239905e-06}
{"train_loss": 0.04756012558937073, "global_step": 225104, "epoch": 2529, "lr": 7.0467946767304175e-06}
{"train_loss": 0.042185597121715546, "global_step": 225105, "epoch": 2529, "lr": 7.046497930995455e-06}
{"train_loss": 0.05749158933758736, "global_step": 225106, "epoch": 2529, "lr": 7.046201191035073e-06}
{"train_loss": 0.043750423938035965, "global_step": 225107, "epoch": 2529, "lr": 7.0459044568493e-06}
{"train_loss": 0.06140966713428497, "global_step": 225108, "epoch": 2529, "lr": 7.045607728438191e-06}
{"train_loss": 0.07805265486240387, "global_step": 225109, "epoch": 2529, "lr": 7.045311005801769e-06}
{"train_loss": 0.09432592988014221, "global_step": 225110, "epoch": 2529, "lr": 7.0450142889400885e-06}
{"train_loss": 0.0574808306992054, "global_step": 225111, "epoch": 2529, "lr": 7.044717577853172e-06}
{"train_loss": 0.0756680965423584, "global_step": 225112, "epoch": 2529, "lr": 7.04442087254108e-06}
{"train_loss": 0.049554791301488876, "global_step": 225113, "epoch": 2529, "lr": 7.04412417300383e-06}
{"train_loss": 0.08613026142120361, "global_step": 225114, "epoch": 2529, "lr": 7.043827479241488e-06}
{"train_loss": 0.049703728407621384, "global_step": 225115, "epoch": 2529, "lr": 7.043530791254066e-06}
{"train_loss": 0.05953598767518997, "global_step": 225116, "epoch": 2529, "lr": 7.04323410904163e-06}
{"train_loss": 0.041448548436164856, "global_step": 225117, "epoch": 2529, "lr": 7.0429374326042074e-06}
{"train_loss": 0.07951518148183823, "global_step": 225118, "epoch": 2529, "lr": 7.042640761941827e-06}
{"train_loss": 0.06286001205444336, "global_step": 225119, "epoch": 2529, "lr": 7.042344097054554e-06}
{"train_loss": 0.06707330048084259, "global_step": 225120, "epoch": 2529, "lr": 7.042047437942401e-06}
{"train_loss": 0.09115154296159744, "global_step": 225121, "epoch": 2529, "lr": 7.041750784605427e-06}
{"train_loss": 0.05879436805844307, "global_step": 225122, "epoch": 2529, "lr": 7.041454137043662e-06}
{"train_loss": 0.04469088092446327, "global_step": 225123, "epoch": 2529, "lr": 7.041157495257161e-06}
{"train_loss": 0.030407244339585304, "global_step": 225124, "epoch": 2529, "lr": 7.0408608592459405e-06}
{"train_loss": 0.024318644776940346, "global_step": 225125, "epoch": 2529, "lr": 7.040564229010061e-06}
{"train_loss": 0.0815732330083847, "global_step": 225126, "epoch": 2529, "lr": 7.040267604549544e-06}
{"train_loss": 0.056192610412836075, "global_step": 225127, "epoch": 2529, "lr": 7.039970985864453e-06}
{"train_loss": 0.05258391797542572, "global_step": 225128, "epoch": 2529, "lr": 7.039674372954802e-06}
{"train_loss": 0.08095547556877136, "global_step": 225129, "epoch": 2529, "lr": 7.039377765820654e-06}
{"train_loss": 0.03081165812909603, "global_step": 225130, "epoch": 2529, "lr": 7.039081164462025e-06}
{"train_loss": 0.030620329082012177, "global_step": 225131, "epoch": 2529, "lr": 7.03878456887897e-06}
{"train_loss": 0.018368972465395927, "global_step": 225132, "epoch": 2529, "lr": 7.038487979071534e-06}
{"train_loss": 0.05189807340502739, "global_step": 225133, "epoch": 2529, "lr": 7.0381913950397445e-06}
{"train_loss": 0.05573611706495285, "global_step": 225134, "epoch": 2529, "lr": 7.037894816783653e-06}
{"train_loss": 0.08491488546133041, "global_step": 225135, "epoch": 2529, "lr": 7.037598244303284e-06}
{"train_loss": 0.038204386830329895, "global_step": 225136, "epoch": 2529, "lr": 7.037301677598701e-06}
{"train_loss": 0.03819277137517929, "global_step": 225137, "epoch": 2529, "lr": 7.037005116669909e-06}
{"train_loss": 0.028844667598605156, "global_step": 225138, "epoch": 2529, "lr": 7.036708561516986e-06}
{"train_loss": 0.03934408351778984, "global_step": 225139, "epoch": 2529, "lr": 7.036412012139943e-06}
{"train_loss": 0.05476456880569458, "global_step": 225140, "epoch": 2529, "lr": 7.03611546853884e-06}
{"train_loss": 0.06979409605264664, "global_step": 225141, "epoch": 2529, "lr": 7.035818930713694e-06}
{"train_loss": 0.06807959079742432, "global_step": 225142, "epoch": 2529, "lr": 7.0355223986645735e-06}
{"train_loss": 0.025788748636841774, "global_step": 225143, "epoch": 2529, "lr": 7.035225872391493e-06}
{"train_loss": 0.07797235995531082, "global_step": 225144, "epoch": 2529, "lr": 7.03492935189451e-06}
{"train_loss": 0.0765853300690651, "global_step": 225145, "epoch": 2529, "lr": 7.034632837173649e-06}
{"train_loss": 0.030225830152630806, "global_step": 225146, "epoch": 2529, "lr": 7.0343363282289685e-06}
{"train_loss": 0.03578554093837738, "global_step": 225147, "epoch": 2529, "lr": 7.034039825060485e-06}
{"train_loss": 0.028641965240240097, "global_step": 225148, "epoch": 2529, "lr": 7.033743327668252e-06}
{"train_loss": 0.010221181437373161, "global_step": 225149, "epoch": 2529, "lr": 7.033446836052321e-06}
{"train_loss": 0.049065105617046356, "global_step": 225150, "epoch": 2529, "lr": 7.033150350212719e-06}
{"train_loss": 0.04348628222942352, "global_step": 225151, "epoch": 2529, "lr": 7.032853870149475e-06}
{"train_loss": 0.0655711442232132, "global_step": 225152, "epoch": 2529, "lr": 7.032557395862649e-06}
{"train_loss": 0.023116841912269592, "global_step": 225153, "epoch": 2529, "lr": 7.032260927352263e-06}
{"train_loss": 0.024462470784783363, "global_step": 225154, "epoch": 2529, "lr": 7.0319644646183735e-06}
{"train_loss": 0.05016353726387024, "global_step": 225155, "epoch": 2529, "lr": 7.031668007661019e-06}
{"train_loss": 0.01869698241353035, "global_step": 225156, "epoch": 2529, "lr": 7.031371556480215e-06}
{"train_loss": 0.024359695613384247, "global_step": 225157, "epoch": 2529, "lr": 7.031075111076035e-06}
{"train_loss": 0.04946557804942131, "global_step": 225158, "epoch": 2529, "lr": 7.03077867144849e-06}
{"train_loss": 0.06653100252151489, "global_step": 225159, "epoch": 2529, "lr": 7.030482237597647e-06}
{"train_loss": 0.06423082202672958, "global_step": 225160, "epoch": 2529, "lr": 7.03018580952352e-06}
{"train_loss": 0.10278384387493134, "global_step": 225161, "epoch": 2529, "lr": 7.029889387226169e-06}
{"train_loss": 0.047430798411369324, "global_step": 225162, "epoch": 2529, "lr": 7.0295929707056175e-06}
{"train_loss": 0.03800585865974426, "global_step": 225163, "epoch": 2529, "lr": 7.029296559961923e-06}
{"train_loss": 0.06132930517196655, "global_step": 225164, "epoch": 2529, "lr": 7.029000154995102e-06}
{"train_loss": 0.021264497190713882, "global_step": 225165, "epoch": 2529, "lr": 7.028703755805227e-06}
{"train_loss": 0.0698137953877449, "global_step": 225166, "epoch": 2529, "lr": 7.028407362392303e-06}
{"train_loss": 0.039642274379730225, "global_step": 225167, "epoch": 2529, "lr": 7.028110974756386e-06}
{"train_loss": 0.05266118794679642, "global_step": 225168, "epoch": 2529, "lr": 7.027814592897525e-06}
{"train_loss": 0.05364285253532482, "global_step": 225169, "epoch": 2529, "lr": 7.027518216815743e-06, "val_loss": 9.038430213928223}
{"train_loss": 0.04039066284894943, "global_step": 225170, "epoch": 2530, "lr": 7.027221846511095e-06}
{"train_loss": 0.10945449024438858, "global_step": 225171, "epoch": 2530, "lr": 7.026925481983604e-06}
{"train_loss": 0.04769771173596382, "global_step": 225172, "epoch": 2530, "lr": 7.026629123233336e-06}
{"train_loss": 0.019909996539354324, "global_step": 225173, "epoch": 2530, "lr": 7.026332770260297e-06}
{"train_loss": 0.03661186248064041, "global_step": 225174, "epoch": 2530, "lr": 7.026036423064552e-06}
{"train_loss": 0.04802659526467323, "global_step": 225175, "epoch": 2530, "lr": 7.025740081646126e-06}
{"train_loss": 0.04341382533311844, "global_step": 225176, "epoch": 2530, "lr": 7.025443746005073e-06}
{"train_loss": 0.0686597228050232, "global_step": 225177, "epoch": 2530, "lr": 7.02514741614142e-06}
{"train_loss": 0.07044760882854462, "global_step": 225178, "epoch": 2530, "lr": 7.0248510920552176e-06}
{"train_loss": 0.047838494181632996, "global_step": 225179, "epoch": 2530, "lr": 7.024554773746494e-06}
{"train_loss": 0.0422220453619957, "global_step": 225180, "epoch": 2530, "lr": 7.0242584612152995e-06}
{"train_loss": 0.05672064796090126, "global_step": 225181, "epoch": 2530, "lr": 7.0239621544616666e-06}
{"train_loss": 0.04497288167476654, "global_step": 225182, "epoch": 2530, "lr": 7.0236658534856445e-06}
{"train_loss": 0.018076013773679733, "global_step": 225183, "epoch": 2530, "lr": 7.023369558287257e-06}
{"train_loss": 0.048450253903865814, "global_step": 225184, "epoch": 2530, "lr": 7.023073268866553e-06}
{"train_loss": 0.06509087979793549, "global_step": 225185, "epoch": 2530, "lr": 7.022776985223583e-06}
{"train_loss": 0.06743750721216202, "global_step": 225186, "epoch": 2530, "lr": 7.022480707358364e-06}
{"train_loss": 0.03521003946661949, "global_step": 225187, "epoch": 2530, "lr": 7.022184435270967e-06}
{"train_loss": 0.08730737119913101, "global_step": 225188, "epoch": 2530, "lr": 7.02188816896141e-06}
{"train_loss": 0.031500108540058136, "global_step": 225189, "epoch": 2530, "lr": 7.02159190842972e-06}
{"train_loss": 0.11082994937896729, "global_step": 225190, "epoch": 2530, "lr": 7.021295653675969e-06}
{"train_loss": 0.08395438641309738, "global_step": 225191, "epoch": 2530, "lr": 7.020999404700168e-06}
{"train_loss": 0.023231787607073784, "global_step": 225192, "epoch": 2530, "lr": 7.020703161502384e-06}
{"train_loss": 0.05543249472975731, "global_step": 225193, "epoch": 2530, "lr": 7.02040692408264e-06}
{"train_loss": 0.08189645409584045, "global_step": 225194, "epoch": 2530, "lr": 7.020110692440973e-06}
{"train_loss": 0.03144759312272072, "global_step": 225195, "epoch": 2530, "lr": 7.019814466577435e-06}
{"train_loss": 0.09724375605583191, "global_step": 225196, "epoch": 2530, "lr": 7.019518246492046e-06}
{"train_loss": 0.06502129882574081, "global_step": 225197, "epoch": 2530, "lr": 7.019222032184869e-06}
{"train_loss": 0.05477512627840042, "global_step": 225198, "epoch": 2530, "lr": 7.018925823655925e-06}
{"train_loss": 0.07528131455183029, "global_step": 225199, "epoch": 2530, "lr": 7.018629620905276e-06}
{"train_loss": 0.02852749265730381, "global_step": 225200, "epoch": 2530, "lr": 7.018333423932932e-06}
{"train_loss": 0.036916814744472504, "global_step": 225201, "epoch": 2530, "lr": 7.018037232738967e-06}
{"train_loss": 0.08364084362983704, "global_step": 225202, "epoch": 2530, "lr": 7.01774104732339e-06}
{"train_loss": 0.0645468607544899, "global_step": 225203, "epoch": 2530, "lr": 7.017444867686251e-06}
{"train_loss": 0.07411031424999237, "global_step": 225204, "epoch": 2530, "lr": 7.0171486938276025e-06}
{"train_loss": 0.016388682648539543, "global_step": 225205, "epoch": 2530, "lr": 7.0168525257474695e-06}
{"train_loss": 0.03274138644337654, "global_step": 225206, "epoch": 2530, "lr": 7.016556363445903e-06}
{"train_loss": 0.017161717638373375, "global_step": 225207, "epoch": 2530, "lr": 7.016260206922925e-06}
{"train_loss": 0.0376875065267086, "global_step": 225208, "epoch": 2530, "lr": 7.015964056178603e-06}
{"train_loss": 0.04855465516448021, "global_step": 225209, "epoch": 2530, "lr": 7.015667911212942e-06}
{"train_loss": 0.059451598674058914, "global_step": 225210, "epoch": 2530, "lr": 7.0153717720260195e-06}
{"train_loss": 0.023188818246126175, "global_step": 225211, "epoch": 2530, "lr": 7.015075638617841e-06}
{"train_loss": 0.08745361119508743, "global_step": 225212, "epoch": 2530, "lr": 7.014779510988473e-06}
{"train_loss": 0.03891971334815025, "global_step": 225213, "epoch": 2530, "lr": 7.014483389137933e-06}
{"train_loss": 0.05746123939752579, "global_step": 225214, "epoch": 2530, "lr": 7.014187273066286e-06}
{"train_loss": 0.03566524013876915, "global_step": 225215, "epoch": 2530, "lr": 7.013891162773539e-06}
{"train_loss": 0.05798441171646118, "global_step": 225216, "epoch": 2530, "lr": 7.0135950582597695e-06}
{"train_loss": 0.0964837446808815, "global_step": 225217, "epoch": 2530, "lr": 7.013298959524983e-06}
{"train_loss": 0.05242493748664856, "global_step": 225218, "epoch": 2530, "lr": 7.0130028665692456e-06}
{"train_loss": 0.04812605679035187, "global_step": 225219, "epoch": 2530, "lr": 7.012706779392575e-06}
{"train_loss": 0.03751426935195923, "global_step": 225220, "epoch": 2530, "lr": 7.012410697995037e-06}
{"train_loss": 0.11534839868545532, "global_step": 225221, "epoch": 2530, "lr": 7.012114622376642e-06}
{"train_loss": 0.07512331753969193, "global_step": 225222, "epoch": 2530, "lr": 7.0118185525374465e-06}
{"train_loss": 0.032634150236845016, "global_step": 225223, "epoch": 2530, "lr": 7.011522488477495e-06}
{"train_loss": 0.0647965520620346, "global_step": 225224, "epoch": 2530, "lr": 7.01122643019681e-06}
{"train_loss": 0.06607218086719513, "global_step": 225225, "epoch": 2530, "lr": 7.010930377695457e-06}
{"train_loss": 0.09115512669086456, "global_step": 225226, "epoch": 2530, "lr": 7.010634330973448e-06}
{"train_loss": 0.11609938740730286, "global_step": 225227, "epoch": 2530, "lr": 7.010338290030843e-06}
{"train_loss": 0.04901851341128349, "global_step": 225228, "epoch": 2530, "lr": 7.010042254867677e-06}
{"train_loss": 0.046051450073719025, "global_step": 225229, "epoch": 2530, "lr": 7.00974622548397e-06}
{"train_loss": 0.04594094306230545, "global_step": 225230, "epoch": 2530, "lr": 7.009450201879797e-06}
{"train_loss": 0.04802766069769859, "global_step": 225231, "epoch": 2530, "lr": 7.009154184055167e-06}
{"train_loss": 0.05338387191295624, "global_step": 225232, "epoch": 2530, "lr": 7.008858172010141e-06}
{"train_loss": 0.0497392937541008, "global_step": 225233, "epoch": 2530, "lr": 7.008562165744753e-06}
{"train_loss": 0.04317673295736313, "global_step": 225234, "epoch": 2530, "lr": 7.0082661652590255e-06}
{"train_loss": 0.06037089601159096, "global_step": 225235, "epoch": 2530, "lr": 7.007970170553024e-06}
{"train_loss": 0.03101063333451748, "global_step": 225236, "epoch": 2530, "lr": 7.007674181626766e-06}
{"train_loss": 0.0380006805062294, "global_step": 225237, "epoch": 2530, "lr": 7.0073781984803184e-06}
{"train_loss": 0.02877630665898323, "global_step": 225238, "epoch": 2530, "lr": 7.007082221113686e-06}
{"train_loss": 0.05245618149638176, "global_step": 225239, "epoch": 2530, "lr": 7.006786249526931e-06}
{"train_loss": 0.051489342004060745, "global_step": 225240, "epoch": 2530, "lr": 7.006490283720102e-06}
{"train_loss": 0.05364103615283966, "global_step": 225241, "epoch": 2530, "lr": 7.006194323693216e-06}
{"train_loss": 0.03882212936878204, "global_step": 225242, "epoch": 2530, "lr": 7.005898369446328e-06}
{"train_loss": 0.02892199158668518, "global_step": 225243, "epoch": 2530, "lr": 7.005602420979468e-06}
{"train_loss": 0.03918995335698128, "global_step": 225244, "epoch": 2530, "lr": 7.005306478292689e-06}
{"train_loss": 0.05943673476576805, "global_step": 225245, "epoch": 2530, "lr": 7.005010541386015e-06}
{"train_loss": 0.04255177453160286, "global_step": 225246, "epoch": 2530, "lr": 7.0047146102595e-06}
{"train_loss": 0.045183323323726654, "global_step": 225247, "epoch": 2530, "lr": 7.0044186849131666e-06}
{"train_loss": 0.041887782514095306, "global_step": 225248, "epoch": 2530, "lr": 7.004122765347077e-06}
{"train_loss": 0.06145321950316429, "global_step": 225249, "epoch": 2530, "lr": 7.0038268515612515e-06}
{"train_loss": 0.052690539509058, "global_step": 225250, "epoch": 2530, "lr": 7.003530943555742e-06}
{"train_loss": 0.058766789734363556, "global_step": 225251, "epoch": 2530, "lr": 7.003235041330575e-06}
{"train_loss": 0.04572328180074692, "global_step": 225252, "epoch": 2530, "lr": 7.002939144885812e-06}
{"train_loss": 0.044120654463768005, "global_step": 225253, "epoch": 2530, "lr": 7.0026432542214696e-06}
{"train_loss": 0.034568674862384796, "global_step": 225254, "epoch": 2530, "lr": 7.002347369337603e-06}
{"train_loss": 0.06776605546474457, "global_step": 225255, "epoch": 2530, "lr": 7.0020514902342405e-06}
{"train_loss": 0.08657418936491013, "global_step": 225256, "epoch": 2530, "lr": 7.0017556169114375e-06}
{"train_loss": 0.0951676145195961, "global_step": 225257, "epoch": 2530, "lr": 7.001459749369216e-06}
{"train_loss": 0.05454108066689432, "global_step": 225258, "epoch": 2530, "lr": 7.00116388760762e-06, "val_loss": 9.073210716247559, "train_action_mse_error": 22.68570899963379}
{"train_loss": 0.07475421577692032, "global_step": 225259, "epoch": 2531, "lr": 7.000868031626706e-06}
{"train_loss": 0.011198357678949833, "global_step": 225260, "epoch": 2531, "lr": 7.0005721814264905e-06}
{"train_loss": 0.06291355937719345, "global_step": 225261, "epoch": 2531, "lr": 7.000276337007039e-06}
{"train_loss": 0.03447847068309784, "global_step": 225262, "epoch": 2531, "lr": 6.999980498368358e-06}
{"train_loss": 0.050485365092754364, "global_step": 225263, "epoch": 2531, "lr": 6.999684665510519e-06}
{"train_loss": 0.028488125652074814, "global_step": 225264, "epoch": 2531, "lr": 6.99938883843354e-06}
{"train_loss": 0.06700696051120758, "global_step": 225265, "epoch": 2531, "lr": 6.999093017137481e-06}
{"train_loss": 0.023244740441441536, "global_step": 225266, "epoch": 2531, "lr": 6.998797201622364e-06}
{"train_loss": 0.02746107243001461, "global_step": 225267, "epoch": 2531, "lr": 6.998501391888229e-06}
{"train_loss": 0.07471892982721329, "global_step": 225268, "epoch": 2531, "lr": 6.9982055879351305e-06}
{"train_loss": 0.05456433445215225, "global_step": 225269, "epoch": 2531, "lr": 6.997909789763085e-06}
{"train_loss": 0.03380563110113144, "global_step": 225270, "epoch": 2531, "lr": 6.99761399737216e-06}
{"train_loss": 0.03809112310409546, "global_step": 225271, "epoch": 2531, "lr": 6.997318210762383e-06}
{"train_loss": 0.05548549443483353, "global_step": 225272, "epoch": 2531, "lr": 6.99702242993378e-06}
{"train_loss": 0.06388247013092041, "global_step": 225273, "epoch": 2531, "lr": 6.9967266548864155e-06}
{"train_loss": 0.03675513714551926, "global_step": 225274, "epoch": 2531, "lr": 6.996430885620303e-06}
{"train_loss": 0.06391707807779312, "global_step": 225275, "epoch": 2531, "lr": 6.9961351221355e-06}
{"train_loss": 0.05211682990193367, "global_step": 225276, "epoch": 2531, "lr": 6.9958393644320496e-06}
{"train_loss": 0.04058590531349182, "global_step": 225277, "epoch": 2531, "lr": 6.99554361250998e-06}
{"train_loss": 0.05959556624293327, "global_step": 225278, "epoch": 2531, "lr": 6.995247866369342e-06}
{"train_loss": 0.02710537239909172, "global_step": 225279, "epoch": 2531, "lr": 6.994952126010157e-06}
{"train_loss": 0.03835682198405266, "global_step": 225280, "epoch": 2531, "lr": 6.994656391432486e-06}
{"train_loss": 0.022062135860323906, "global_step": 225281, "epoch": 2531, "lr": 6.9943606626363514e-06}
{"train_loss": 0.09712164103984833, "global_step": 225282, "epoch": 2531, "lr": 6.994064939621808e-06}
{"train_loss": 0.030447108671069145, "global_step": 225283, "epoch": 2531, "lr": 6.993769222388885e-06}
{"train_loss": 0.04622895270586014, "global_step": 225284, "epoch": 2531, "lr": 6.9934735109376315e-06}
{"train_loss": 0.07075244188308716, "global_step": 225285, "epoch": 2531, "lr": 6.99317780526807e-06}
{"train_loss": 0.034868258982896805, "global_step": 225286, "epoch": 2531, "lr": 6.99288210538026e-06}
{"train_loss": 0.05528460070490837, "global_step": 225287, "epoch": 2531, "lr": 6.992586411274227e-06}
{"train_loss": 0.0702366977930069, "global_step": 225288, "epoch": 2531, "lr": 6.992290722950029e-06}
{"train_loss": 0.049627792090177536, "global_step": 225289, "epoch": 2531, "lr": 6.991995040407679e-06}
{"train_loss": 0.024550165981054306, "global_step": 225290, "epoch": 2531, "lr": 6.991699363647241e-06}
{"train_loss": 0.03196234628558159, "global_step": 225291, "epoch": 2531, "lr": 6.991403692668735e-06}
{"train_loss": 0.04148885980248451, "global_step": 225292, "epoch": 2531, "lr": 6.99110802747222e-06}
{"train_loss": 0.07984782010316849, "global_step": 225293, "epoch": 2531, "lr": 6.990812368057714e-06}
{"train_loss": 0.03655802831053734, "global_step": 225294, "epoch": 2531, "lr": 6.990516714425277e-06}
{"train_loss": 0.02636113576591015, "global_step": 225295, "epoch": 2531, "lr": 6.990221066574943e-06}
{"train_loss": 0.047344278544187546, "global_step": 225296, "epoch": 2531, "lr": 6.989925424506744e-06}
{"train_loss": 0.02133162133395672, "global_step": 225297, "epoch": 2531, "lr": 6.989629788220736e-06}
{"train_loss": 0.04871436208486557, "global_step": 225298, "epoch": 2531, "lr": 6.9893341577169315e-06}
{"train_loss": 0.03969062864780426, "global_step": 225299, "epoch": 2531, "lr": 6.989038532995407e-06}
{"train_loss": 0.02514246478676796, "global_step": 225300, "epoch": 2531, "lr": 6.988742914056162e-06}
{"train_loss": 0.04097235947847366, "global_step": 225301, "epoch": 2531, "lr": 6.98844730089927e-06}
{"train_loss": 0.05022645741701126, "global_step": 225302, "epoch": 2531, "lr": 6.988151693524747e-06}
{"train_loss": 0.021416297182440758, "global_step": 225303, "epoch": 2531, "lr": 6.987856091932654e-06}
{"train_loss": 0.03977982699871063, "global_step": 225304, "epoch": 2531, "lr": 6.987560496123024e-06}
{"train_loss": 0.04343286529183388, "global_step": 225305, "epoch": 2531, "lr": 6.987264906095875e-06}
{"train_loss": 0.040216114372015, "global_step": 225306, "epoch": 2531, "lr": 6.986969321851278e-06}
{"train_loss": 0.13690276443958282, "global_step": 225307, "epoch": 2531, "lr": 6.9866737433892435e-06}
{"train_loss": 0.034627366811037064, "global_step": 225308, "epoch": 2531, "lr": 6.98637817070984e-06}
{"train_loss": 0.04699820280075073, "global_step": 225309, "epoch": 2531, "lr": 6.986082603813099e-06}
{"train_loss": 0.034890636801719666, "global_step": 225310, "epoch": 2531, "lr": 6.985787042699038e-06}
{"train_loss": 0.036733560264110565, "global_step": 225311, "epoch": 2531, "lr": 6.985491487367713e-06}
{"train_loss": 0.07263477146625519, "global_step": 225312, "epoch": 2531, "lr": 6.985195937819178e-06}
{"train_loss": 0.08533091843128204, "global_step": 225313, "epoch": 2531, "lr": 6.9849003940534454e-06}
{"train_loss": 0.019662031903862953, "global_step": 225314, "epoch": 2531, "lr": 6.984604856070581e-06}
{"train_loss": 0.019255585968494415, "global_step": 225315, "epoch": 2531, "lr": 6.984309323870603e-06}
{"train_loss": 0.01409602165222168, "global_step": 225316, "epoch": 2531, "lr": 6.98401379745357e-06}
{"train_loss": 0.05691409111022949, "global_step": 225317, "epoch": 2531, "lr": 6.9837182768195005e-06}
{"train_loss": 0.029032284393906593, "global_step": 225318, "epoch": 2531, "lr": 6.983422761968456e-06}
{"train_loss": 0.06638512015342712, "global_step": 225319, "epoch": 2531, "lr": 6.9831272529004564e-06}
{"train_loss": 0.05804245546460152, "global_step": 225320, "epoch": 2531, "lr": 6.982831749615565e-06}
{"train_loss": 0.07377217710018158, "global_step": 225321, "epoch": 2531, "lr": 6.9825362521137915e-06}
{"train_loss": 0.046951472759246826, "global_step": 225322, "epoch": 2531, "lr": 6.982240760395203e-06}
{"train_loss": 0.05928594246506691, "global_step": 225323, "epoch": 2531, "lr": 6.9819452744598226e-06}
{"train_loss": 0.05840997397899628, "global_step": 225324, "epoch": 2531, "lr": 6.9816497943076985e-06}
{"train_loss": 0.047020670026540756, "global_step": 225325, "epoch": 2531, "lr": 6.9813543199388596e-06}
{"train_loss": 0.0987735390663147, "global_step": 225326, "epoch": 2531, "lr": 6.981058851353367e-06}
{"train_loss": 0.05840346962213516, "global_step": 225327, "epoch": 2531, "lr": 6.980763388551237e-06}
{"train_loss": 0.06405226141214371, "global_step": 225328, "epoch": 2531, "lr": 6.980467931532525e-06}
{"train_loss": 0.027817899361252785, "global_step": 225329, "epoch": 2531, "lr": 6.980172480297253e-06}
{"train_loss": 0.04215545207262039, "global_step": 225330, "epoch": 2531, "lr": 6.979877034845478e-06}
{"train_loss": 0.03422464057803154, "global_step": 225331, "epoch": 2531, "lr": 6.979581595177243e-06}
{"train_loss": 0.06430738419294357, "global_step": 225332, "epoch": 2531, "lr": 6.979286161292564e-06}
{"train_loss": 0.02791355550289154, "global_step": 225333, "epoch": 2531, "lr": 6.97899073319151e-06}
{"train_loss": 0.06559375673532486, "global_step": 225334, "epoch": 2531, "lr": 6.978695310874095e-06}
{"train_loss": 0.053234994411468506, "global_step": 225335, "epoch": 2531, "lr": 6.978399894340382e-06}
{"train_loss": 0.07048755139112473, "global_step": 225336, "epoch": 2531, "lr": 6.978104483590387e-06}
{"train_loss": 0.04184214770793915, "global_step": 225337, "epoch": 2531, "lr": 6.97780907862417e-06}
{"train_loss": 0.06410156935453415, "global_step": 225338, "epoch": 2531, "lr": 6.977513679441755e-06}
{"train_loss": 0.08909166604280472, "global_step": 225339, "epoch": 2531, "lr": 6.9772182860431965e-06}
{"train_loss": 0.06794150173664093, "global_step": 225340, "epoch": 2531, "lr": 6.976922898428517e-06}
{"train_loss": 0.04427666217088699, "global_step": 225341, "epoch": 2531, "lr": 6.976627516597778e-06}
{"train_loss": 0.07965242117643356, "global_step": 225342, "epoch": 2531, "lr": 6.9763321405509955e-06}
{"train_loss": 0.048793282359838486, "global_step": 225343, "epoch": 2531, "lr": 6.97603677028823e-06}
{"train_loss": 0.04002106189727783, "global_step": 225344, "epoch": 2531, "lr": 6.975741405809516e-06}
{"train_loss": 0.023806555196642876, "global_step": 225345, "epoch": 2531, "lr": 6.9754460471148755e-06}
{"train_loss": 0.05792541429400444, "global_step": 225346, "epoch": 2531, "lr": 6.9751506942043745e-06}
{"train_loss": 0.0495488800924648, "global_step": 225347, "epoch": 2531, "lr": 6.974855347078024e-06, "val_loss": 9.022970199584961}
{"train_loss": 0.04851243644952774, "global_step": 225348, "epoch": 2532, "lr": 6.974560005735897e-06}
{"train_loss": 0.05403049290180206, "global_step": 225349, "epoch": 2532, "lr": 6.974264670178004e-06}
{"train_loss": 0.09428567439317703, "global_step": 225350, "epoch": 2532, "lr": 6.973969340404407e-06}
{"train_loss": 0.033965688198804855, "global_step": 225351, "epoch": 2532, "lr": 6.973674016415121e-06}
{"train_loss": 0.05470246821641922, "global_step": 225352, "epoch": 2532, "lr": 6.9733786982102146e-06}
{"train_loss": 0.09013599157333374, "global_step": 225353, "epoch": 2532, "lr": 6.973083385789703e-06}
{"train_loss": 0.04772695153951645, "global_step": 225354, "epoch": 2532, "lr": 6.972788079153642e-06}
{"train_loss": 0.06411752849817276, "global_step": 225355, "epoch": 2532, "lr": 6.972492778302059e-06}
{"train_loss": 0.06127038970589638, "global_step": 225356, "epoch": 2532, "lr": 6.97219748323501e-06}
{"train_loss": 0.03288559243083, "global_step": 225357, "epoch": 2532, "lr": 6.971902193952512e-06}
{"train_loss": 0.05829736590385437, "global_step": 225358, "epoch": 2532, "lr": 6.971606910454626e-06}
{"train_loss": 0.07555780559778214, "global_step": 225359, "epoch": 2532, "lr": 6.971311632741379e-06}
{"train_loss": 0.11075867712497711, "global_step": 225360, "epoch": 2532, "lr": 6.971016360812821e-06}
{"train_loss": 0.04493191838264465, "global_step": 225361, "epoch": 2532, "lr": 6.970721094668975e-06}
{"train_loss": 0.07465989887714386, "global_step": 225362, "epoch": 2532, "lr": 6.970425834309902e-06}
{"train_loss": 0.034695032984018326, "global_step": 225363, "epoch": 2532, "lr": 6.970130579735617e-06}
{"train_loss": 0.05424089357256889, "global_step": 225364, "epoch": 2532, "lr": 6.96983533094619e-06}
{"train_loss": 0.07421515136957169, "global_step": 225365, "epoch": 2532, "lr": 6.969540087941628e-06}
{"train_loss": 0.03971577808260918, "global_step": 225366, "epoch": 2532, "lr": 6.96924485072199e-06}
{"train_loss": 0.11239365488290787, "global_step": 225367, "epoch": 2532, "lr": 6.968949619287324e-06}
{"train_loss": 0.02840040810406208, "global_step": 225368, "epoch": 2532, "lr": 6.968654393637647e-06}
{"train_loss": 0.04359813034534454, "global_step": 225369, "epoch": 2532, "lr": 6.96835917377302e-06}
{"train_loss": 0.026201147586107254, "global_step": 225370, "epoch": 2532, "lr": 6.9680639596934604e-06}
{"train_loss": 0.033258166164159775, "global_step": 225371, "epoch": 2532, "lr": 6.967768751399034e-06}
{"train_loss": 0.05982532352209091, "global_step": 225372, "epoch": 2532, "lr": 6.967473548889758e-06}
{"train_loss": 0.0410001203417778, "global_step": 225373, "epoch": 2532, "lr": 6.9671783521656875e-06}
{"train_loss": 0.06310905516147614, "global_step": 225374, "epoch": 2532, "lr": 6.966883161226845e-06}
{"train_loss": 0.07606784999370575, "global_step": 225375, "epoch": 2532, "lr": 6.966587976073291e-06}
{"train_loss": 0.021237613633275032, "global_step": 225376, "epoch": 2532, "lr": 6.9662927967050425e-06}
{"train_loss": 0.08364982157945633, "global_step": 225377, "epoch": 2532, "lr": 6.965997623122167e-06}
{"train_loss": 0.03883147984743118, "global_step": 225378, "epoch": 2532, "lr": 6.965702455324674e-06}
{"train_loss": 0.07040874660015106, "global_step": 225379, "epoch": 2532, "lr": 6.965407293312626e-06}
{"train_loss": 0.058189988136291504, "global_step": 225380, "epoch": 2532, "lr": 6.965112137086049e-06}
{"train_loss": 0.07643949240446091, "global_step": 225381, "epoch": 2532, "lr": 6.9648169866449955e-06}
{"train_loss": 0.02059885859489441, "global_step": 225382, "epoch": 2532, "lr": 6.964521841989502e-06}
{"train_loss": 0.042328957468271255, "global_step": 225383, "epoch": 2532, "lr": 6.964226703119592e-06}
{"train_loss": 0.03738464042544365, "global_step": 225384, "epoch": 2532, "lr": 6.963931570035326e-06}
{"train_loss": 0.055560849606990814, "global_step": 225385, "epoch": 2532, "lr": 6.963636442736721e-06}
{"train_loss": 0.03819849714636803, "global_step": 225386, "epoch": 2532, "lr": 6.963341321223848e-06}
{"train_loss": 0.03535959869623184, "global_step": 225387, "epoch": 2532, "lr": 6.963046205496715e-06}
{"train_loss": 0.035906147211790085, "global_step": 225388, "epoch": 2532, "lr": 6.962751095555387e-06}
{"train_loss": 0.05691387876868248, "global_step": 225389, "epoch": 2532, "lr": 6.96245599139988e-06}
{"train_loss": 0.0864754244685173, "global_step": 225390, "epoch": 2532, "lr": 6.9621608930302615e-06}
{"train_loss": 0.05321295186877251, "global_step": 225391, "epoch": 2532, "lr": 6.961865800446543e-06}
{"train_loss": 0.0777302160859108, "global_step": 225392, "epoch": 2532, "lr": 6.961570713648785e-06}
{"train_loss": 0.03688919544219971, "global_step": 225393, "epoch": 2532, "lr": 6.96127563263701e-06}
{"train_loss": 0.0647711455821991, "global_step": 225394, "epoch": 2532, "lr": 6.960980557411278e-06}
{"train_loss": 0.06664144992828369, "global_step": 225395, "epoch": 2532, "lr": 6.9606854879716074e-06}
{"train_loss": 0.04976717010140419, "global_step": 225396, "epoch": 2532, "lr": 6.960390424318059e-06}
{"train_loss": 0.0793689638376236, "global_step": 225397, "epoch": 2532, "lr": 6.960095366450653e-06}
{"train_loss": 0.041752371937036514, "global_step": 225398, "epoch": 2532, "lr": 6.959800314369447e-06}
{"train_loss": 0.020719271153211594, "global_step": 225399, "epoch": 2532, "lr": 6.959505268074457e-06}
{"train_loss": 0.05190841108560562, "global_step": 225400, "epoch": 2532, "lr": 6.959210227565749e-06}
{"train_loss": 0.07054916769266129, "global_step": 225401, "epoch": 2532, "lr": 6.95891519284334e-06}
{"train_loss": 0.06840485334396362, "global_step": 225402, "epoch": 2532, "lr": 6.958620163907281e-06}
{"train_loss": 0.02689821831882, "global_step": 225403, "epoch": 2532, "lr": 6.958325140757627e-06}
{"train_loss": 0.04222845658659935, "global_step": 225404, "epoch": 2532, "lr": 6.958030123394382e-06}
{"train_loss": 0.08234116435050964, "global_step": 225405, "epoch": 2532, "lr": 6.95773511181762e-06}
{"train_loss": 0.11048276722431183, "global_step": 225406, "epoch": 2532, "lr": 6.9574401060273576e-06}
{"train_loss": 0.05444404110312462, "global_step": 225407, "epoch": 2532, "lr": 6.957145106023649e-06}
{"train_loss": 0.03267195075750351, "global_step": 225408, "epoch": 2532, "lr": 6.956850111806518e-06}
{"train_loss": 0.034709010273218155, "global_step": 225409, "epoch": 2532, "lr": 6.956555123376024e-06}
{"train_loss": 0.045992229133844376, "global_step": 225410, "epoch": 2532, "lr": 6.95626014073219e-06}
{"train_loss": 0.030464868992567062, "global_step": 225411, "epoch": 2532, "lr": 6.955965163875066e-06}
{"train_loss": 0.03443608060479164, "global_step": 225412, "epoch": 2532, "lr": 6.955670192804681e-06}
{"train_loss": 0.03647904843091965, "global_step": 225413, "epoch": 2532, "lr": 6.955375227521093e-06}
{"train_loss": 0.02324913814663887, "global_step": 225414, "epoch": 2532, "lr": 6.955080268024322e-06}
{"train_loss": 0.03169171139597893, "global_step": 225415, "epoch": 2532, "lr": 6.954785314314422e-06}
{"train_loss": 0.0721697211265564, "global_step": 225416, "epoch": 2532, "lr": 6.954490366391414e-06}
{"train_loss": 0.08327074348926544, "global_step": 225417, "epoch": 2532, "lr": 6.954195424255367e-06}
{"train_loss": 0.02919432334601879, "global_step": 225418, "epoch": 2532, "lr": 6.95390048790629e-06}
{"train_loss": 0.06738092005252838, "global_step": 225419, "epoch": 2532, "lr": 6.953605557344245e-06}
{"train_loss": 0.03857393562793732, "global_step": 225420, "epoch": 2532, "lr": 6.953310632569265e-06}
{"train_loss": 0.023149140179157257, "global_step": 225421, "epoch": 2532, "lr": 6.953015713581379e-06}
{"train_loss": 0.07512207329273224, "global_step": 225422, "epoch": 2532, "lr": 6.952720800380647e-06}
{"train_loss": 0.0652930960059166, "global_step": 225423, "epoch": 2532, "lr": 6.9524258929670796e-06}
{"train_loss": 0.06576159596443176, "global_step": 225424, "epoch": 2532, "lr": 6.95213099134075e-06}
{"train_loss": 0.05213816463947296, "global_step": 225425, "epoch": 2532, "lr": 6.951836095501668e-06}
{"train_loss": 0.04731262847781181, "global_step": 225426, "epoch": 2532, "lr": 6.951541205449902e-06}
{"train_loss": 0.08294423669576645, "global_step": 225427, "epoch": 2532, "lr": 6.9512463211854615e-06}
{"train_loss": 0.03785919398069382, "global_step": 225428, "epoch": 2532, "lr": 6.950951442708414e-06}
{"train_loss": 0.045110512524843216, "global_step": 225429, "epoch": 2532, "lr": 6.950656570018776e-06}
{"train_loss": 0.04749077185988426, "global_step": 225430, "epoch": 2532, "lr": 6.9503617031166035e-06}
{"train_loss": 0.06247582286596298, "global_step": 225431, "epoch": 2532, "lr": 6.950066842001923e-06}
{"train_loss": 0.020855296403169632, "global_step": 225432, "epoch": 2532, "lr": 6.949771986674797e-06}
{"train_loss": 0.03287287801504135, "global_step": 225433, "epoch": 2532, "lr": 6.949477137135229e-06}
{"train_loss": 0.012853920459747314, "global_step": 225434, "epoch": 2532, "lr": 6.9491822933833e-06}
{"train_loss": 0.045440129935741425, "global_step": 225435, "epoch": 2532, "lr": 6.948887455419012e-06}
{"train_loss": 0.052593085663707066, "global_step": 225436, "epoch": 2532, "lr": 6.948592623242433e-06, "val_loss": 9.021804809570312}
{"train_loss": 0.05760244280099869, "global_step": 225437, "epoch": 2533, "lr": 6.9482977968535814e-06}
{"train_loss": 0.04164094105362892, "global_step": 225438, "epoch": 2533, "lr": 6.948002976252505e-06}
{"train_loss": 0.015117458999156952, "global_step": 225439, "epoch": 2533, "lr": 6.947708161439253e-06}
{"train_loss": 0.05169116333127022, "global_step": 225440, "epoch": 2533, "lr": 6.947413352413851e-06}
{"train_loss": 0.022145267575979233, "global_step": 225441, "epoch": 2533, "lr": 6.947118549176352e-06}
{"train_loss": 0.05142739787697792, "global_step": 225442, "epoch": 2533, "lr": 6.9468237517267776e-06}
{"train_loss": 0.034786634147167206, "global_step": 225443, "epoch": 2533, "lr": 6.9465289600651905e-06}
{"train_loss": 0.031659457832574844, "global_step": 225444, "epoch": 2533, "lr": 6.946234174191607e-06}
{"train_loss": 0.06054767966270447, "global_step": 225445, "epoch": 2533, "lr": 6.945939394106088e-06}
{"train_loss": 0.053005151450634, "global_step": 225446, "epoch": 2533, "lr": 6.94564461980865e-06}
{"train_loss": 0.07665830850601196, "global_step": 225447, "epoch": 2533, "lr": 6.94534985129936e-06}
{"train_loss": 0.0293755903840065, "global_step": 225448, "epoch": 2533, "lr": 6.945055088578228e-06}
{"train_loss": 0.029118282720446587, "global_step": 225449, "epoch": 2533, "lr": 6.944760331645322e-06}
{"train_loss": 0.05599100887775421, "global_step": 225450, "epoch": 2533, "lr": 6.9444655805006586e-06}
{"train_loss": 0.061270058155059814, "global_step": 225451, "epoch": 2533, "lr": 6.944170835144298e-06}
{"train_loss": 0.019486891105771065, "global_step": 225452, "epoch": 2533, "lr": 6.943876095576252e-06}
{"train_loss": 0.029236529022455215, "global_step": 225453, "epoch": 2533, "lr": 6.943581361796592e-06}
{"train_loss": 0.04860803857445717, "global_step": 225454, "epoch": 2533, "lr": 6.9432866338053346e-06}
{"train_loss": 0.04236342012882233, "global_step": 225455, "epoch": 2533, "lr": 6.942991911602536e-06}
{"train_loss": 0.03231959417462349, "global_step": 225456, "epoch": 2533, "lr": 6.942697195188214e-06}
{"train_loss": 0.04850315302610397, "global_step": 225457, "epoch": 2533, "lr": 6.942402484562432e-06}
{"train_loss": 0.1223151758313179, "global_step": 225458, "epoch": 2533, "lr": 6.9421077797252156e-06}
{"train_loss": 0.05364091321825981, "global_step": 225459, "epoch": 2533, "lr": 6.941813080676613e-06}
{"train_loss": 0.02817663550376892, "global_step": 225460, "epoch": 2533, "lr": 6.9415183874166625e-06}
{"train_loss": 0.07556315511465073, "global_step": 225461, "epoch": 2533, "lr": 6.941223699945381e-06}
{"train_loss": 0.0580175518989563, "global_step": 225462, "epoch": 2533, "lr": 6.940929018262849e-06}
{"train_loss": 0.0669989287853241, "global_step": 225463, "epoch": 2533, "lr": 6.940634342369068e-06}
{"train_loss": 0.04314614087343216, "global_step": 225464, "epoch": 2533, "lr": 6.940339672264101e-06}
{"train_loss": 0.09282680600881577, "global_step": 225465, "epoch": 2533, "lr": 6.940045007947976e-06}
{"train_loss": 0.05314147472381592, "global_step": 225466, "epoch": 2533, "lr": 6.939750349420743e-06}
{"train_loss": 0.0732431709766388, "global_step": 225467, "epoch": 2533, "lr": 6.939455696682429e-06}
{"train_loss": 0.04393797740340233, "global_step": 225468, "epoch": 2533, "lr": 6.939161049733089e-06}
{"train_loss": 0.03642958402633667, "global_step": 225469, "epoch": 2533, "lr": 6.9388664085727415e-06}
{"train_loss": 0.056597642600536346, "global_step": 225470, "epoch": 2533, "lr": 6.938571773201452e-06}
{"train_loss": 0.010177288204431534, "global_step": 225471, "epoch": 2533, "lr": 6.938277143619237e-06}
{"train_loss": 0.05137236788868904, "global_step": 225472, "epoch": 2533, "lr": 6.937982519826153e-06}
{"train_loss": 0.05035312846302986, "global_step": 225473, "epoch": 2533, "lr": 6.9376879018222265e-06}
{"train_loss": 0.035460274666547775, "global_step": 225474, "epoch": 2533, "lr": 6.937393289607508e-06}
{"train_loss": 0.029347969219088554, "global_step": 225475, "epoch": 2533, "lr": 6.937098683182025e-06}
{"train_loss": 0.07376857846975327, "global_step": 225476, "epoch": 2533, "lr": 6.9368040825458225e-06}
{"train_loss": 0.03737197071313858, "global_step": 225477, "epoch": 2533, "lr": 6.936509487698955e-06}
{"train_loss": 0.0419590063393116, "global_step": 225478, "epoch": 2533, "lr": 6.936214898641435e-06}
{"train_loss": 0.050711099058389664, "global_step": 225479, "epoch": 2533, "lr": 6.935920315373329e-06}
{"train_loss": 0.03912966325879097, "global_step": 225480, "epoch": 2533, "lr": 6.935625737894652e-06}
{"train_loss": 0.056809354573488235, "global_step": 225481, "epoch": 2533, "lr": 6.9353311662054666e-06}
{"train_loss": 0.0458870492875576, "global_step": 225482, "epoch": 2533, "lr": 6.9350366003057944e-06}
{"train_loss": 0.056849975138902664, "global_step": 225483, "epoch": 2533, "lr": 6.934742040195686e-06}
{"train_loss": 0.09034829586744308, "global_step": 225484, "epoch": 2533, "lr": 6.934447485875167e-06}
{"train_loss": 0.03364323452115059, "global_step": 225485, "epoch": 2533, "lr": 6.934152937344301e-06}
{"train_loss": 0.04078289121389389, "global_step": 225486, "epoch": 2533, "lr": 6.933858394603099e-06}
{"train_loss": 0.01142667606472969, "global_step": 225487, "epoch": 2533, "lr": 6.933563857651626e-06}
{"train_loss": 0.06488301604986191, "global_step": 225488, "epoch": 2533, "lr": 6.933269326489899e-06}
{"train_loss": 0.07465866208076477, "global_step": 225489, "epoch": 2533, "lr": 6.932974801117981e-06}
{"train_loss": 0.05791367590427399, "global_step": 225490, "epoch": 2533, "lr": 6.932680281535891e-06}
{"train_loss": 0.07550682872533798, "global_step": 225491, "epoch": 2533, "lr": 6.932385767743682e-06}
{"train_loss": 0.06808671355247498, "global_step": 225492, "epoch": 2533, "lr": 6.932091259741386e-06}
{"train_loss": 0.04806269332766533, "global_step": 225493, "epoch": 2533, "lr": 6.931796757529052e-06}
{"train_loss": 0.0494648739695549, "global_step": 225494, "epoch": 2533, "lr": 6.931502261106698e-06}
{"train_loss": 0.025328442454338074, "global_step": 225495, "epoch": 2533, "lr": 6.931207770474396e-06}
{"train_loss": 0.04070626571774483, "global_step": 225496, "epoch": 2533, "lr": 6.9309132856321515e-06}
{"train_loss": 0.07697026431560516, "global_step": 225497, "epoch": 2533, "lr": 6.930618806580036e-06}
{"train_loss": 0.06194951385259628, "global_step": 225498, "epoch": 2533, "lr": 6.930324333318072e-06}
{"train_loss": 0.023822050541639328, "global_step": 225499, "epoch": 2533, "lr": 6.9300298658462934e-06}
{"train_loss": 0.10498840361833572, "global_step": 225500, "epoch": 2533, "lr": 6.9297354041647555e-06}
{"train_loss": 0.04305832087993622, "global_step": 225501, "epoch": 2533, "lr": 6.92944094827348e-06}
{"train_loss": 0.04388851299881935, "global_step": 225502, "epoch": 2533, "lr": 6.929146498172528e-06}
{"train_loss": 0.040860243141651154, "global_step": 225503, "epoch": 2533, "lr": 6.928852053861911e-06}
{"train_loss": 0.060864709317684174, "global_step": 225504, "epoch": 2533, "lr": 6.928557615341702e-06}
{"train_loss": 0.05230389162898064, "global_step": 225505, "epoch": 2533, "lr": 6.92826318261191e-06}
{"train_loss": 0.04356985166668892, "global_step": 225506, "epoch": 2533, "lr": 6.927968755672598e-06}
{"train_loss": 0.020472824573516846, "global_step": 225507, "epoch": 2533, "lr": 6.9276743345237874e-06}
{"train_loss": 0.030624020844697952, "global_step": 225508, "epoch": 2533, "lr": 6.927379919165533e-06}
{"train_loss": 0.05187114328145981, "global_step": 225509, "epoch": 2533, "lr": 6.927085509597858e-06}
{"train_loss": 0.058617882430553436, "global_step": 225510, "epoch": 2533, "lr": 6.926791105820824e-06}
{"train_loss": 0.03509669750928879, "global_step": 225511, "epoch": 2533, "lr": 6.926496707834446e-06}
{"train_loss": 0.08141865581274033, "global_step": 225512, "epoch": 2533, "lr": 6.9262023156387745e-06}
{"train_loss": 0.05798095092177391, "global_step": 225513, "epoch": 2533, "lr": 6.92590792923386e-06}
{"train_loss": 0.08158468455076218, "global_step": 225514, "epoch": 2533, "lr": 6.925613548619725e-06}
{"train_loss": 0.03998890891671181, "global_step": 225515, "epoch": 2533, "lr": 6.925319173796424e-06}
{"train_loss": 0.05240632966160774, "global_step": 225516, "epoch": 2533, "lr": 6.92502480476398e-06}
{"train_loss": 0.06867247074842453, "global_step": 225517, "epoch": 2533, "lr": 6.924730441522453e-06}
{"train_loss": 0.06961063295602798, "global_step": 225518, "epoch": 2533, "lr": 6.924436084071856e-06}
{"train_loss": 0.052684079855680466, "global_step": 225519, "epoch": 2533, "lr": 6.924141732412259e-06}
{"train_loss": 0.08823283761739731, "global_step": 225520, "epoch": 2533, "lr": 6.923847386543675e-06}
{"train_loss": 0.06670228391885757, "global_step": 225521, "epoch": 2533, "lr": 6.923553046466164e-06}
{"train_loss": 0.01573900878429413, "global_step": 225522, "epoch": 2533, "lr": 6.9232587121797425e-06}
{"train_loss": 0.04114716127514839, "global_step": 225523, "epoch": 2533, "lr": 6.922964383684477e-06}
{"train_loss": 0.032109953463077545, "global_step": 225524, "epoch": 2533, "lr": 6.922670060980385e-06}
{"train_loss": 0.050254276629244345, "global_step": 225525, "epoch": 2533, "lr": 6.922375744067527e-06, "val_loss": 9.016663551330566}
{"train_loss": 0.03719686344265938, "global_step": 225526, "epoch": 2534, "lr": 6.922081432945915e-06}
{"train_loss": 0.04588763415813446, "global_step": 225527, "epoch": 2534, "lr": 6.92178712761562e-06}
{"train_loss": 0.057537686079740524, "global_step": 225528, "epoch": 2534, "lr": 6.9214928280766475e-06}
{"train_loss": 0.05229862406849861, "global_step": 225529, "epoch": 2534, "lr": 6.9211985343290595e-06}
{"train_loss": 0.10505009442567825, "global_step": 225530, "epoch": 2534, "lr": 6.920904246372906e-06}
{"train_loss": 0.04603183642029762, "global_step": 225531, "epoch": 2534, "lr": 6.920609964208208e-06}
{"train_loss": 0.04117245599627495, "global_step": 225532, "epoch": 2534, "lr": 6.920315687834994e-06}
{"train_loss": 0.033115267753601074, "global_step": 225533, "epoch": 2534, "lr": 6.9200214172533364e-06}
{"train_loss": 0.04197879135608673, "global_step": 225534, "epoch": 2534, "lr": 6.919727152463246e-06}
{"train_loss": 0.034478798508644104, "global_step": 225535, "epoch": 2534, "lr": 6.919432893464778e-06}
{"train_loss": 0.03659794107079506, "global_step": 225536, "epoch": 2534, "lr": 6.9191386402579714e-06}
{"train_loss": 0.035144589841365814, "global_step": 225537, "epoch": 2534, "lr": 6.918844392842849e-06}
{"train_loss": 0.027693822979927063, "global_step": 225538, "epoch": 2534, "lr": 6.9185501512194775e-06}
{"train_loss": 0.041120946407318115, "global_step": 225539, "epoch": 2534, "lr": 6.918255915387867e-06}
{"train_loss": 0.1000354066491127, "global_step": 225540, "epoch": 2534, "lr": 6.917961685348079e-06}
{"train_loss": 0.031058985739946365, "global_step": 225541, "epoch": 2534, "lr": 6.917667461100142e-06}
{"train_loss": 0.0322207473218441, "global_step": 225542, "epoch": 2534, "lr": 6.9173732426441094e-06}
{"train_loss": 0.07962770015001297, "global_step": 225543, "epoch": 2534, "lr": 6.9170790299799945e-06}
{"train_loss": 0.05955734848976135, "global_step": 225544, "epoch": 2534, "lr": 6.916784823107869e-06}
{"train_loss": 0.06566546112298965, "global_step": 225545, "epoch": 2534, "lr": 6.9164906220277495e-06}
{"train_loss": 0.0819430947303772, "global_step": 225546, "epoch": 2534, "lr": 6.916196426739685e-06}
{"train_loss": 0.023556983098387718, "global_step": 225547, "epoch": 2534, "lr": 6.91590223724371e-06}
{"train_loss": 0.04062625393271446, "global_step": 225548, "epoch": 2534, "lr": 6.915608053539863e-06}
{"train_loss": 0.09082671999931335, "global_step": 225549, "epoch": 2534, "lr": 6.915313875628199e-06}
{"train_loss": 0.07615995407104492, "global_step": 225550, "epoch": 2534, "lr": 6.9150197035087295e-06}
{"train_loss": 0.0555487759411335, "global_step": 225551, "epoch": 2534, "lr": 6.914725537181527e-06}
{"train_loss": 0.11769493669271469, "global_step": 225552, "epoch": 2534, "lr": 6.914431376646602e-06}
{"train_loss": 0.07258761674165726, "global_step": 225553, "epoch": 2534, "lr": 6.914137221904021e-06}
{"train_loss": 0.03497854247689247, "global_step": 225554, "epoch": 2534, "lr": 6.9138430729537965e-06}
{"train_loss": 0.017895961180329323, "global_step": 225555, "epoch": 2534, "lr": 6.913548929795988e-06}
{"train_loss": 0.09177642315626144, "global_step": 225556, "epoch": 2534, "lr": 6.913254792430618e-06}
{"train_loss": 0.07819099724292755, "global_step": 225557, "epoch": 2534, "lr": 6.912960660857748e-06}
{"train_loss": 0.027545884251594543, "global_step": 225558, "epoch": 2534, "lr": 6.912666535077394e-06}
{"train_loss": 0.04554414004087448, "global_step": 225559, "epoch": 2534, "lr": 6.912372415089618e-06}
{"train_loss": 0.046781934797763824, "global_step": 225560, "epoch": 2534, "lr": 6.912078300894437e-06}
{"train_loss": 0.05688123777508736, "global_step": 225561, "epoch": 2534, "lr": 6.911784192491916e-06}
{"train_loss": 0.07606002688407898, "global_step": 225562, "epoch": 2534, "lr": 6.911490089882067e-06}
{"train_loss": 0.06990190595388412, "global_step": 225563, "epoch": 2534, "lr": 6.911195993064951e-06}
{"train_loss": 0.0565672442317009, "global_step": 225564, "epoch": 2534, "lr": 6.9109019020405895e-06}
{"train_loss": 0.04862724989652634, "global_step": 225565, "epoch": 2534, "lr": 6.910607816809034e-06}
{"train_loss": 0.06198812648653984, "global_step": 225566, "epoch": 2534, "lr": 6.910313737370333e-06}
{"train_loss": 0.059590402990579605, "global_step": 225567, "epoch": 2534, "lr": 6.910019663724504e-06}
{"train_loss": 0.049331411719322205, "global_step": 225568, "epoch": 2534, "lr": 6.909725595871614e-06}
{"train_loss": 0.049370840191841125, "global_step": 225569, "epoch": 2534, "lr": 6.909431533811678e-06}
{"train_loss": 0.0772215947508812, "global_step": 225570, "epoch": 2534, "lr": 6.909137477544736e-06}
{"train_loss": 0.06332985311746597, "global_step": 225571, "epoch": 2534, "lr": 6.908843427070843e-06}
{"train_loss": 0.020777516067028046, "global_step": 225572, "epoch": 2534, "lr": 6.908549382390023e-06}
{"train_loss": 0.06554077565670013, "global_step": 225573, "epoch": 2534, "lr": 6.9082553435023345e-06}
{"train_loss": 0.024738546460866928, "global_step": 225574, "epoch": 2534, "lr": 6.907961310407796e-06}
{"train_loss": 0.08153814822435379, "global_step": 225575, "epoch": 2534, "lr": 6.907667283106467e-06}
{"train_loss": 0.055021412670612335, "global_step": 225576, "epoch": 2534, "lr": 6.907373261598377e-06}
{"train_loss": 0.057302359491586685, "global_step": 225577, "epoch": 2534, "lr": 6.907079245883552e-06}
{"train_loss": 0.06663493812084198, "global_step": 225578, "epoch": 2534, "lr": 6.9067852359620545e-06}
{"train_loss": 0.039189137518405914, "global_step": 225579, "epoch": 2534, "lr": 6.9064912318339115e-06}
{"train_loss": 0.08153509348630905, "global_step": 225580, "epoch": 2534, "lr": 6.906197233499167e-06}
{"train_loss": 0.015949690714478493, "global_step": 225581, "epoch": 2534, "lr": 6.905903240957856e-06}
{"train_loss": 0.03758152946829796, "global_step": 225582, "epoch": 2534, "lr": 6.905609254210027e-06}
{"train_loss": 0.09270162135362625, "global_step": 225583, "epoch": 2534, "lr": 6.905315273255703e-06}
{"train_loss": 0.07619000971317291, "global_step": 225584, "epoch": 2534, "lr": 6.905021298094938e-06}
{"train_loss": 0.036720890551805496, "global_step": 225585, "epoch": 2534, "lr": 6.904727328727773e-06}
{"train_loss": 0.04208415374159813, "global_step": 225586, "epoch": 2534, "lr": 6.904433365154234e-06}
{"train_loss": 0.029387766495347023, "global_step": 225587, "epoch": 2534, "lr": 6.904139407374383e-06}
{"train_loss": 0.07947298884391785, "global_step": 225588, "epoch": 2534, "lr": 6.90384545538823e-06}
{"train_loss": 0.053266942501068115, "global_step": 225589, "epoch": 2534, "lr": 6.903551509195844e-06}
{"train_loss": 0.10693266242742538, "global_step": 225590, "epoch": 2534, "lr": 6.903257568797233e-06}
{"train_loss": 0.04504336789250374, "global_step": 225591, "epoch": 2534, "lr": 6.902963634192472e-06}
{"train_loss": 0.05421491712331772, "global_step": 225592, "epoch": 2534, "lr": 6.902669705381565e-06}
{"train_loss": 0.04028499871492386, "global_step": 225593, "epoch": 2534, "lr": 6.902375782364584e-06}
{"train_loss": 0.06964658945798874, "global_step": 225594, "epoch": 2534, "lr": 6.902081865141541e-06}
{"train_loss": 0.05397439002990723, "global_step": 225595, "epoch": 2534, "lr": 6.9017879537124965e-06}
{"train_loss": 0.027864942327141762, "global_step": 225596, "epoch": 2534, "lr": 6.9014940480774725e-06}
{"train_loss": 0.01828577183187008, "global_step": 225597, "epoch": 2534, "lr": 6.901200148236531e-06}
{"train_loss": 0.08865512907505035, "global_step": 225598, "epoch": 2534, "lr": 6.9009062541896815e-06}
{"train_loss": 0.039232391864061356, "global_step": 225599, "epoch": 2534, "lr": 6.900612365936992e-06}
{"train_loss": 0.0949012041091919, "global_step": 225600, "epoch": 2534, "lr": 6.900318483478485e-06}
{"train_loss": 0.09327255189418793, "global_step": 225601, "epoch": 2534, "lr": 6.900024606814209e-06}
{"train_loss": 0.013469924218952656, "global_step": 225602, "epoch": 2534, "lr": 6.899730735944193e-06}
{"train_loss": 0.02576441876590252, "global_step": 225603, "epoch": 2534, "lr": 6.899436870868481e-06}
{"train_loss": 0.028016097843647003, "global_step": 225604, "epoch": 2534, "lr": 6.899143011587123e-06}
{"train_loss": 0.03309933841228485, "global_step": 225605, "epoch": 2534, "lr": 6.89884915810014e-06}
{"train_loss": 0.042651765048503876, "global_step": 225606, "epoch": 2534, "lr": 6.898555310407595e-06}
{"train_loss": 0.03781093284487724, "global_step": 225607, "epoch": 2534, "lr": 6.898261468509498e-06}
{"train_loss": 0.049432992935180664, "global_step": 225608, "epoch": 2534, "lr": 6.897967632405922e-06}
{"train_loss": 0.054333508014678955, "global_step": 225609, "epoch": 2534, "lr": 6.8976738020968824e-06}
{"train_loss": 0.04466036334633827, "global_step": 225610, "epoch": 2534, "lr": 6.897379977582419e-06}
{"train_loss": 0.06458631157875061, "global_step": 225611, "epoch": 2534, "lr": 6.8970861588625866e-06}
{"train_loss": 0.0319003164768219, "global_step": 225612, "epoch": 2534, "lr": 6.896792345937403e-06}
{"train_loss": 0.02617700584232807, "global_step": 225613, "epoch": 2534, "lr": 6.896498538806934e-06}
{"train_loss": 0.05333630750060416, "global_step": 225614, "epoch": 2534, "lr": 6.896204737471207e-06, "val_loss": 9.01926326751709}
{"train_loss": 0.054339051246643066, "global_step": 225615, "epoch": 2535, "lr": 6.895910941930245e-06}
{"train_loss": 0.05494541674852371, "global_step": 225616, "epoch": 2535, "lr": 6.895617152184114e-06}
{"train_loss": 0.027652665972709656, "global_step": 225617, "epoch": 2535, "lr": 6.8953233682328315e-06}
{"train_loss": 0.033903494477272034, "global_step": 225618, "epoch": 2535, "lr": 6.8950295900764625e-06}
{"train_loss": 0.06122494488954544, "global_step": 225619, "epoch": 2535, "lr": 6.8947358177150144e-06}
{"train_loss": 0.07071954756975174, "global_step": 225620, "epoch": 2535, "lr": 6.894442051148547e-06}
{"train_loss": 0.03328899294137955, "global_step": 225621, "epoch": 2535, "lr": 6.894148290377106e-06}
{"train_loss": 0.06078959256410599, "global_step": 225622, "epoch": 2535, "lr": 6.893854535400712e-06}
{"train_loss": 0.041037991642951965, "global_step": 225623, "epoch": 2535, "lr": 6.8935607862194216e-06}
{"train_loss": 0.04169901832938194, "global_step": 225624, "epoch": 2535, "lr": 6.893267042833257e-06}
{"train_loss": 0.031902287155389786, "global_step": 225625, "epoch": 2535, "lr": 6.8929733052422785e-06}
{"train_loss": 0.04011759161949158, "global_step": 225626, "epoch": 2535, "lr": 6.892679573446504e-06}
{"train_loss": 0.04275726154446602, "global_step": 225627, "epoch": 2535, "lr": 6.892385847445992e-06}
{"train_loss": 0.04703017696738243, "global_step": 225628, "epoch": 2535, "lr": 6.892092127240768e-06}
{"train_loss": 0.04026090353727341, "global_step": 225629, "epoch": 2535, "lr": 6.8917984128308855e-06}
{"train_loss": 0.06363730877637863, "global_step": 225630, "epoch": 2535, "lr": 6.891504704216362e-06}
{"train_loss": 0.034684401005506516, "global_step": 225631, "epoch": 2535, "lr": 6.891211001397263e-06}
{"train_loss": 0.04329852759838104, "global_step": 225632, "epoch": 2535, "lr": 6.8909173043736004e-06}
{"train_loss": 0.026728754863142967, "global_step": 225633, "epoch": 2535, "lr": 6.890623613145441e-06}
{"train_loss": 0.07103981077671051, "global_step": 225634, "epoch": 2535, "lr": 6.890329927712808e-06}
{"train_loss": 0.0402841754257679, "global_step": 225635, "epoch": 2535, "lr": 6.890036248075749e-06}
{"train_loss": 0.048628777265548706, "global_step": 225636, "epoch": 2535, "lr": 6.889742574234287e-06}
{"train_loss": 0.07653067260980606, "global_step": 225637, "epoch": 2535, "lr": 6.889448906188489e-06}
{"train_loss": 0.07212464511394501, "global_step": 225638, "epoch": 2535, "lr": 6.889155243938366e-06}
{"train_loss": 0.017199937254190445, "global_step": 225639, "epoch": 2535, "lr": 6.888861587483969e-06}
{"train_loss": 0.027265403419733047, "global_step": 225640, "epoch": 2535, "lr": 6.888567936825352e-06}
{"train_loss": 0.04205355793237686, "global_step": 225641, "epoch": 2535, "lr": 6.8882742919625324e-06}
{"train_loss": 0.08016989380121231, "global_step": 225642, "epoch": 2535, "lr": 6.887980652895565e-06}
{"train_loss": 0.06728578358888626, "global_step": 225643, "epoch": 2535, "lr": 6.887687019624478e-06}
{"train_loss": 0.09615166485309601, "global_step": 225644, "epoch": 2535, "lr": 6.887393392149322e-06}
{"train_loss": 0.08339393883943558, "global_step": 225645, "epoch": 2535, "lr": 6.887099770470123e-06}
{"train_loss": 0.04812714084982872, "global_step": 225646, "epoch": 2535, "lr": 6.886806154586934e-06}
{"train_loss": 0.013752756640315056, "global_step": 225647, "epoch": 2535, "lr": 6.886512544499796e-06}
{"train_loss": 0.08233513683080673, "global_step": 225648, "epoch": 2535, "lr": 6.8862189402087215e-06}
{"train_loss": 0.044012732803821564, "global_step": 225649, "epoch": 2535, "lr": 6.8859253417137895e-06}
{"train_loss": 0.044654957950115204, "global_step": 225650, "epoch": 2535, "lr": 6.885631749015003e-06}
{"train_loss": 0.05322440341114998, "global_step": 225651, "epoch": 2535, "lr": 6.885338162112426e-06}
{"train_loss": 0.06308069080114365, "global_step": 225652, "epoch": 2535, "lr": 6.885044581006095e-06}
{"train_loss": 0.05372881889343262, "global_step": 225653, "epoch": 2535, "lr": 6.884751005696033e-06}
{"train_loss": 0.02236533910036087, "global_step": 225654, "epoch": 2535, "lr": 6.884457436182301e-06}
{"train_loss": 0.07053565233945847, "global_step": 225655, "epoch": 2535, "lr": 6.884163872464916e-06}
{"train_loss": 0.04634932801127434, "global_step": 225656, "epoch": 2535, "lr": 6.883870314543927e-06}
{"train_loss": 0.0461367592215538, "global_step": 225657, "epoch": 2535, "lr": 6.883576762419391e-06}
{"train_loss": 0.05453585460782051, "global_step": 225658, "epoch": 2535, "lr": 6.8832832160913185e-06}
{"train_loss": 0.027177434414625168, "global_step": 225659, "epoch": 2535, "lr": 6.882989675559781e-06}
{"train_loss": 0.027949143201112747, "global_step": 225660, "epoch": 2535, "lr": 6.882696140824779e-06}
{"train_loss": 0.019013965502381325, "global_step": 225661, "epoch": 2535, "lr": 6.88240261188639e-06}
{"train_loss": 0.03385554254055023, "global_step": 225662, "epoch": 2535, "lr": 6.882109088744626e-06}
{"train_loss": 0.04634084925055504, "global_step": 225663, "epoch": 2535, "lr": 6.881815571399547e-06}
{"train_loss": 0.018608978018164635, "global_step": 225664, "epoch": 2535, "lr": 6.88152205985117e-06}
{"train_loss": 0.06280841678380966, "global_step": 225665, "epoch": 2535, "lr": 6.881228554099561e-06}
{"train_loss": 0.050131283700466156, "global_step": 225666, "epoch": 2535, "lr": 6.8809350541447325e-06}
{"train_loss": 0.028608838096261024, "global_step": 225667, "epoch": 2535, "lr": 6.880641559986745e-06}
{"train_loss": 0.03519846871495247, "global_step": 225668, "epoch": 2535, "lr": 6.88034807162562e-06}
{"train_loss": 0.02208588644862175, "global_step": 225669, "epoch": 2535, "lr": 6.880054589061419e-06}
{"train_loss": 0.01769053377211094, "global_step": 225670, "epoch": 2535, "lr": 6.879761112294159e-06}
{"train_loss": 0.03191670775413513, "global_step": 225671, "epoch": 2535, "lr": 6.8794676413239e-06}
{"train_loss": 0.04178110137581825, "global_step": 225672, "epoch": 2535, "lr": 6.87917417615066e-06}
{"train_loss": 0.10176167637109756, "global_step": 225673, "epoch": 2535, "lr": 6.878880716774499e-06}
{"train_loss": 0.021732492372393608, "global_step": 225674, "epoch": 2535, "lr": 6.878587263195435e-06}
{"train_loss": 0.023444795981049538, "global_step": 225675, "epoch": 2535, "lr": 6.878293815413522e-06}
{"train_loss": 0.028584115207195282, "global_step": 225676, "epoch": 2535, "lr": 6.878000373428806e-06}
{"train_loss": 0.04303562641143799, "global_step": 225677, "epoch": 2535, "lr": 6.877706937241307e-06}
{"train_loss": 0.047867219895124435, "global_step": 225678, "epoch": 2535, "lr": 6.877413506851088e-06}
{"train_loss": 0.07459771633148193, "global_step": 225679, "epoch": 2535, "lr": 6.877120082258159e-06}
{"train_loss": 0.036926958709955215, "global_step": 225680, "epoch": 2535, "lr": 6.876826663462593e-06}
{"train_loss": 0.039164941757917404, "global_step": 225681, "epoch": 2535, "lr": 6.876533250464395e-06}
{"train_loss": 0.05270247906446457, "global_step": 225682, "epoch": 2535, "lr": 6.8762398432636374e-06}
{"train_loss": 0.05185863375663757, "global_step": 225683, "epoch": 2535, "lr": 6.875946441860331e-06}
{"train_loss": 0.03388126567006111, "global_step": 225684, "epoch": 2535, "lr": 6.875653046254538e-06}
{"train_loss": 0.05532568693161011, "global_step": 225685, "epoch": 2535, "lr": 6.875359656446278e-06}
{"train_loss": 0.05271400138735771, "global_step": 225686, "epoch": 2535, "lr": 6.87506627243561e-06}
{"train_loss": 0.08657592535018921, "global_step": 225687, "epoch": 2535, "lr": 6.874772894222565e-06}
{"train_loss": 0.0479278489947319, "global_step": 225688, "epoch": 2535, "lr": 6.8744795218071704e-06}
{"train_loss": 0.07334434241056442, "global_step": 225689, "epoch": 2535, "lr": 6.874186155189483e-06}
{"train_loss": 0.06566943228244781, "global_step": 225690, "epoch": 2535, "lr": 6.873892794369529e-06}
{"train_loss": 0.046186503022909164, "global_step": 225691, "epoch": 2535, "lr": 6.873599439347366e-06}
{"train_loss": 0.07514749467372894, "global_step": 225692, "epoch": 2535, "lr": 6.873306090123011e-06}
{"train_loss": 0.025517262518405914, "global_step": 225693, "epoch": 2535, "lr": 6.873012746696522e-06}
{"train_loss": 0.022101733833551407, "global_step": 225694, "epoch": 2535, "lr": 6.8727194090679234e-06}
{"train_loss": 0.047877781093120575, "global_step": 225695, "epoch": 2535, "lr": 6.872426077237271e-06}
{"train_loss": 0.06252989917993546, "global_step": 225696, "epoch": 2535, "lr": 6.872132751204585e-06}
{"train_loss": 0.037523653358221054, "global_step": 225697, "epoch": 2535, "lr": 6.871839430969923e-06}
{"train_loss": 0.07543110102415085, "global_step": 225698, "epoch": 2535, "lr": 6.871546116533312e-06}
{"train_loss": 0.030297625809907913, "global_step": 225699, "epoch": 2535, "lr": 6.8712528078948016e-06}
{"train_loss": 0.04518171027302742, "global_step": 225700, "epoch": 2535, "lr": 6.870959505054414e-06}
{"train_loss": 0.03279289975762367, "global_step": 225701, "epoch": 2535, "lr": 6.870666208012216e-06}
{"train_loss": 0.029250405728816986, "global_step": 225702, "epoch": 2535, "lr": 6.870372916768214e-06}
{"train_loss": 0.046881890133692976, "global_step": 225703, "epoch": 2535, "lr": 6.870079631322479e-06, "val_loss": 8.992226600646973, "train_action_mse_error": 14.474310874938965}
{"train_loss": 0.06660180538892746, "global_step": 225704, "epoch": 2536, "lr": 6.8697863516750285e-06}
{"train_loss": 0.06383305788040161, "global_step": 225705, "epoch": 2536, "lr": 6.869493077825917e-06}
{"train_loss": 0.05052315071225166, "global_step": 225706, "epoch": 2536, "lr": 6.869199809775167e-06}
{"train_loss": 0.08066830784082413, "global_step": 225707, "epoch": 2536, "lr": 6.868906547522835e-06}
{"train_loss": 0.02041536197066307, "global_step": 225708, "epoch": 2536, "lr": 6.868613291068942e-06}
{"train_loss": 0.044282834976911545, "global_step": 225709, "epoch": 2536, "lr": 6.868320040413556e-06}
{"train_loss": 0.055714599788188934, "global_step": 225710, "epoch": 2536, "lr": 6.868026795556681e-06}
{"train_loss": 0.03344825655221939, "global_step": 225711, "epoch": 2536, "lr": 6.8677335564983794e-06}
{"train_loss": 0.015900567173957825, "global_step": 225712, "epoch": 2536, "lr": 6.867440323238688e-06}
{"train_loss": 0.032236240804195404, "global_step": 225713, "epoch": 2536, "lr": 6.867147095777643e-06}
{"train_loss": 0.035316284745931625, "global_step": 225714, "epoch": 2536, "lr": 6.866853874115287e-06}
{"train_loss": 0.04243209585547447, "global_step": 225715, "epoch": 2536, "lr": 6.866560658251647e-06}
{"train_loss": 0.0577506422996521, "global_step": 225716, "epoch": 2536, "lr": 6.866267448186786e-06}
{"train_loss": 0.03672686591744423, "global_step": 225717, "epoch": 2536, "lr": 6.8659742439207196e-06}
{"train_loss": 0.0332857221364975, "global_step": 225718, "epoch": 2536, "lr": 6.865681045453504e-06}
{"train_loss": 0.05830587446689606, "global_step": 225719, "epoch": 2536, "lr": 6.865387852785166e-06}
{"train_loss": 0.058753304183483124, "global_step": 225720, "epoch": 2536, "lr": 6.8650946659157554e-06}
{"train_loss": 0.12486916780471802, "global_step": 225721, "epoch": 2536, "lr": 6.864801484845301e-06}
{"train_loss": 0.034784916788339615, "global_step": 225722, "epoch": 2536, "lr": 6.8645083095738585e-06}
{"train_loss": 0.10441222041845322, "global_step": 225723, "epoch": 2536, "lr": 6.864215140101443e-06}
{"train_loss": 0.05341155454516411, "global_step": 225724, "epoch": 2536, "lr": 6.863921976428122e-06}
{"train_loss": 0.038480862975120544, "global_step": 225725, "epoch": 2536, "lr": 6.863628818553925e-06}
{"train_loss": 0.040152985602617264, "global_step": 225726, "epoch": 2536, "lr": 6.8633356664788705e-06}
{"train_loss": 0.025748999789357185, "global_step": 225727, "epoch": 2536, "lr": 6.863042520203028e-06}
{"train_loss": 0.026238825172185898, "global_step": 225728, "epoch": 2536, "lr": 6.862749379726408e-06}
{"train_loss": 0.04119669646024704, "global_step": 225729, "epoch": 2536, "lr": 6.862456245049082e-06}
{"train_loss": 0.061143651604652405, "global_step": 225730, "epoch": 2536, "lr": 6.862163116171061e-06}
{"train_loss": 0.046386364847421646, "global_step": 225731, "epoch": 2536, "lr": 6.861869993092407e-06}
{"train_loss": 0.04193875193595886, "global_step": 225732, "epoch": 2536, "lr": 6.861576875813136e-06}
{"train_loss": 0.10260557383298874, "global_step": 225733, "epoch": 2536, "lr": 6.861283764333315e-06}
{"train_loss": 0.06111643835902214, "global_step": 225734, "epoch": 2536, "lr": 6.860990658652955e-06}
{"train_loss": 0.036227546632289886, "global_step": 225735, "epoch": 2536, "lr": 6.860697558772116e-06}
{"train_loss": 0.03197363018989563, "global_step": 225736, "epoch": 2536, "lr": 6.860404464690829e-06}
{"train_loss": 0.05097656697034836, "global_step": 225737, "epoch": 2536, "lr": 6.860111376409139e-06}
{"train_loss": 0.054698601365089417, "global_step": 225738, "epoch": 2536, "lr": 6.859818293927073e-06}
{"train_loss": 0.03602874279022217, "global_step": 225739, "epoch": 2536, "lr": 6.859525217244689e-06}
{"train_loss": 0.06383026391267776, "global_step": 225740, "epoch": 2536, "lr": 6.859232146362005e-06}
{"train_loss": 0.0722450464963913, "global_step": 225741, "epoch": 2536, "lr": 6.858939081279081e-06}
{"train_loss": 0.044423408806324005, "global_step": 225742, "epoch": 2536, "lr": 6.858646021995934e-06}
{"train_loss": 0.021330825984477997, "global_step": 225743, "epoch": 2536, "lr": 6.858352968512633e-06}
{"train_loss": 0.06213543936610222, "global_step": 225744, "epoch": 2536, "lr": 6.858059920829185e-06}
{"train_loss": 0.058795586228370667, "global_step": 225745, "epoch": 2536, "lr": 6.85776687894566e-06}
{"train_loss": 0.03578932583332062, "global_step": 225746, "epoch": 2536, "lr": 6.857473842862067e-06}
{"train_loss": 0.08960837870836258, "global_step": 225747, "epoch": 2536, "lr": 6.857180812578462e-06}
{"train_loss": 0.07298825681209564, "global_step": 225748, "epoch": 2536, "lr": 6.856887788094895e-06}
{"train_loss": 0.04196464642882347, "global_step": 225749, "epoch": 2536, "lr": 6.856594769411384e-06}
{"train_loss": 0.1401638388633728, "global_step": 225750, "epoch": 2536, "lr": 6.856301756527989e-06}
{"train_loss": 0.0675341933965683, "global_step": 225751, "epoch": 2536, "lr": 6.85600874944472e-06}
{"train_loss": 0.03276469558477402, "global_step": 225752, "epoch": 2536, "lr": 6.8557157481616565e-06}
{"train_loss": 0.05051768571138382, "global_step": 225753, "epoch": 2536, "lr": 6.855422752678797e-06}
{"train_loss": 0.03575548529624939, "global_step": 225754, "epoch": 2536, "lr": 6.855129762996215e-06}
{"train_loss": 0.012258832342922688, "global_step": 225755, "epoch": 2536, "lr": 6.8548367791139265e-06}
{"train_loss": 0.06926038861274719, "global_step": 225756, "epoch": 2536, "lr": 6.8545438010319875e-06}
{"train_loss": 0.025938741862773895, "global_step": 225757, "epoch": 2536, "lr": 6.854250828750413e-06}
{"train_loss": 0.026177717372775078, "global_step": 225758, "epoch": 2536, "lr": 6.853957862269278e-06}
{"train_loss": 0.049334682524204254, "global_step": 225759, "epoch": 2536, "lr": 6.853664901588586e-06}
{"train_loss": 0.060262564569711685, "global_step": 225760, "epoch": 2536, "lr": 6.853371946708409e-06}
{"train_loss": 0.056507110595703125, "global_step": 225761, "epoch": 2536, "lr": 6.853078997628754e-06}
{"train_loss": 0.09834688901901245, "global_step": 225762, "epoch": 2536, "lr": 6.852786054349691e-06}
{"train_loss": 0.027824973687529564, "global_step": 225763, "epoch": 2536, "lr": 6.852493116871245e-06}
{"train_loss": 0.042861562222242355, "global_step": 225764, "epoch": 2536, "lr": 6.8522001851934415e-06}
{"train_loss": 0.044756997376680374, "global_step": 225765, "epoch": 2536, "lr": 6.851907259316348e-06}
{"train_loss": 0.023476392030715942, "global_step": 225766, "epoch": 2536, "lr": 6.851614339239976e-06}
{"train_loss": 0.04547175019979477, "global_step": 225767, "epoch": 2536, "lr": 6.851321424964391e-06}
{"train_loss": 0.04488830640912056, "global_step": 225768, "epoch": 2536, "lr": 6.851028516489605e-06}
{"train_loss": 0.03143260255455971, "global_step": 225769, "epoch": 2536, "lr": 6.85073561381569e-06}
{"train_loss": 0.08349265158176422, "global_step": 225770, "epoch": 2536, "lr": 6.850442716942651e-06}
{"train_loss": 0.042227763682603836, "global_step": 225771, "epoch": 2536, "lr": 6.850149825870561e-06}
{"train_loss": 0.05803612619638443, "global_step": 225772, "epoch": 2536, "lr": 6.849856940599425e-06}
{"train_loss": 0.09682200849056244, "global_step": 225773, "epoch": 2536, "lr": 6.849564061129315e-06}
{"train_loss": 0.035951413214206696, "global_step": 225774, "epoch": 2536, "lr": 6.849271187460243e-06}
{"train_loss": 0.03269975259900093, "global_step": 225775, "epoch": 2536, "lr": 6.848978319592275e-06}
{"train_loss": 0.029324878007173538, "global_step": 225776, "epoch": 2536, "lr": 6.8486854575254165e-06}
{"train_loss": 0.02743734046816826, "global_step": 225777, "epoch": 2536, "lr": 6.848392601259745e-06}
{"train_loss": 0.06818729639053345, "global_step": 225778, "epoch": 2536, "lr": 6.848099750795267e-06}
{"train_loss": 0.0783528983592987, "global_step": 225779, "epoch": 2536, "lr": 6.847806906132048e-06}
{"train_loss": 0.05136299505829811, "global_step": 225780, "epoch": 2536, "lr": 6.847514067270106e-06}
{"train_loss": 0.09499486535787582, "global_step": 225781, "epoch": 2536, "lr": 6.8472212342095e-06}
{"train_loss": 0.06356319785118103, "global_step": 225782, "epoch": 2536, "lr": 6.846928406950243e-06}
{"train_loss": 0.06092749908566475, "global_step": 225783, "epoch": 2536, "lr": 6.846635585492395e-06}
{"train_loss": 0.07180606573820114, "global_step": 225784, "epoch": 2536, "lr": 6.846342769836001e-06}
{"train_loss": 0.025790473446249962, "global_step": 225785, "epoch": 2536, "lr": 6.8460499599810825e-06}
{"train_loss": 0.04153301939368248, "global_step": 225786, "epoch": 2536, "lr": 6.84575715592769e-06}
{"train_loss": 0.047075048089027405, "global_step": 225787, "epoch": 2536, "lr": 6.845464357675857e-06}
{"train_loss": 0.03355070576071739, "global_step": 225788, "epoch": 2536, "lr": 6.845171565225633e-06}
{"train_loss": 0.04126698896288872, "global_step": 225789, "epoch": 2536, "lr": 6.8448787785770355e-06}
{"train_loss": 0.040457431226968765, "global_step": 225790, "epoch": 2536, "lr": 6.844585997730135e-06}
{"train_loss": 0.03170347213745117, "global_step": 225791, "epoch": 2536, "lr": 6.844293222684939e-06}
{"train_loss": 0.05117407107328096, "global_step": 225792, "epoch": 2536, "lr": 6.844000453441512e-06, "val_loss": 9.069646835327148}
{"train_loss": 0.14721444249153137, "global_step": 225793, "epoch": 2537, "lr": 6.843707689999873e-06}
{"train_loss": 0.08263576775789261, "global_step": 225794, "epoch": 2537, "lr": 6.843414932360081e-06}
{"train_loss": 0.05074422061443329, "global_step": 225795, "epoch": 2537, "lr": 6.84312218052216e-06}
{"train_loss": 0.08227230608463287, "global_step": 225796, "epoch": 2537, "lr": 6.8428294344861645e-06}
{"train_loss": 0.04517587646842003, "global_step": 225797, "epoch": 2537, "lr": 6.8425366942521105e-06}
{"train_loss": 0.05154919624328613, "global_step": 225798, "epoch": 2537, "lr": 6.842243959820066e-06}
{"train_loss": 0.041946377605199814, "global_step": 225799, "epoch": 2537, "lr": 6.841951231190042e-06}
{"train_loss": 0.028241366147994995, "global_step": 225800, "epoch": 2537, "lr": 6.841658508362103e-06}
{"train_loss": 0.08833593130111694, "global_step": 225801, "epoch": 2537, "lr": 6.841365791336268e-06}
{"train_loss": 0.09052158147096634, "global_step": 225802, "epoch": 2537, "lr": 6.841073080112592e-06}
{"train_loss": 0.04864395782351494, "global_step": 225803, "epoch": 2537, "lr": 6.840780374691114e-06}
{"train_loss": 0.07724233716726303, "global_step": 225804, "epoch": 2537, "lr": 6.840487675071849e-06}
{"train_loss": 0.053234875202178955, "global_step": 225805, "epoch": 2537, "lr": 6.840194981254872e-06}
{"train_loss": 0.046351078897714615, "global_step": 225806, "epoch": 2537, "lr": 6.8399022932401915e-06}
{"train_loss": 0.06853798776865005, "global_step": 225807, "epoch": 2537, "lr": 6.83960961102787e-06}
{"train_loss": 0.03461523354053497, "global_step": 225808, "epoch": 2537, "lr": 6.839316934617923e-06}
{"train_loss": 0.06767862290143967, "global_step": 225809, "epoch": 2537, "lr": 6.8390242640104185e-06}
{"train_loss": 0.02338571660220623, "global_step": 225810, "epoch": 2537, "lr": 6.838731599205373e-06}
{"train_loss": 0.07245441526174545, "global_step": 225811, "epoch": 2537, "lr": 6.8384389402028405e-06}
{"train_loss": 0.10200095921754837, "global_step": 225812, "epoch": 2537, "lr": 6.838146287002845e-06}
{"train_loss": 0.03905760869383812, "global_step": 225813, "epoch": 2537, "lr": 6.837853639605441e-06}
{"train_loss": 0.08884605765342712, "global_step": 225814, "epoch": 2537, "lr": 6.837560998010656e-06}
{"train_loss": 0.08523265272378922, "global_step": 225815, "epoch": 2537, "lr": 6.837268362218546e-06}
{"train_loss": 0.062258511781692505, "global_step": 225816, "epoch": 2537, "lr": 6.836975732229123e-06}
{"train_loss": 0.04138774797320366, "global_step": 225817, "epoch": 2537, "lr": 6.836683108042458e-06}
{"train_loss": 0.08134320378303528, "global_step": 225818, "epoch": 2537, "lr": 6.836390489658561e-06}
{"train_loss": 0.04414650425314903, "global_step": 225819, "epoch": 2537, "lr": 6.836097877077491e-06}
{"train_loss": 0.048862215131521225, "global_step": 225820, "epoch": 2537, "lr": 6.8358052702992835e-06}
{"train_loss": 0.055044449865818024, "global_step": 225821, "epoch": 2537, "lr": 6.835512669323973e-06}
{"train_loss": 0.04989079013466835, "global_step": 225822, "epoch": 2537, "lr": 6.83522007415161e-06}
{"train_loss": 0.04922611638903618, "global_step": 225823, "epoch": 2537, "lr": 6.834927484782211e-06}
{"train_loss": 0.04865622520446777, "global_step": 225824, "epoch": 2537, "lr": 6.834634901215847e-06}
{"train_loss": 0.02898062951862812, "global_step": 225825, "epoch": 2537, "lr": 6.834342323452525e-06}
{"train_loss": 0.026449978351593018, "global_step": 225826, "epoch": 2537, "lr": 6.834049751492311e-06}
{"train_loss": 0.07076296955347061, "global_step": 225827, "epoch": 2537, "lr": 6.8337571853352215e-06}
{"train_loss": 0.06502599269151688, "global_step": 225828, "epoch": 2537, "lr": 6.833464624981323e-06}
{"train_loss": 0.02947510965168476, "global_step": 225829, "epoch": 2537, "lr": 6.833172070430621e-06}
{"train_loss": 0.03829353675246239, "global_step": 225830, "epoch": 2537, "lr": 6.832879521683189e-06}
{"train_loss": 0.043629199266433716, "global_step": 225831, "epoch": 2537, "lr": 6.832586978739036e-06}
{"train_loss": 0.08737543225288391, "global_step": 225832, "epoch": 2537, "lr": 6.83229444159823e-06}
{"train_loss": 0.02597174420952797, "global_step": 225833, "epoch": 2537, "lr": 6.832001910260782e-06}
{"train_loss": 0.04806524142622948, "global_step": 225834, "epoch": 2537, "lr": 6.831709384726759e-06}
{"train_loss": 0.06924272328615189, "global_step": 225835, "epoch": 2537, "lr": 6.831416864996171e-06}
{"train_loss": 0.02812853828072548, "global_step": 225836, "epoch": 2537, "lr": 6.8311243510690845e-06}
{"train_loss": 0.11039330065250397, "global_step": 225837, "epoch": 2537, "lr": 6.830831842945523e-06}
{"train_loss": 0.06943338364362717, "global_step": 225838, "epoch": 2537, "lr": 6.830539340625536e-06}
{"train_loss": 0.01804288476705551, "global_step": 225839, "epoch": 2537, "lr": 6.830246844109145e-06}
{"train_loss": 0.04623746871948242, "global_step": 225840, "epoch": 2537, "lr": 6.829954353396412e-06}
{"train_loss": 0.0337325781583786, "global_step": 225841, "epoch": 2537, "lr": 6.829661868487369e-06}
{"train_loss": 0.057125601917505264, "global_step": 225842, "epoch": 2537, "lr": 6.829369389382034e-06}
{"train_loss": 0.06679006665945053, "global_step": 225843, "epoch": 2537, "lr": 6.829076916080479e-06}
{"train_loss": 0.03167479857802391, "global_step": 225844, "epoch": 2537, "lr": 6.82878444858272e-06}
{"train_loss": 0.037295565009117126, "global_step": 225845, "epoch": 2537, "lr": 6.828491986888813e-06}
{"train_loss": 0.018142390996217728, "global_step": 225846, "epoch": 2537, "lr": 6.8281995309987805e-06}
{"train_loss": 0.03926872834563255, "global_step": 225847, "epoch": 2537, "lr": 6.827907080912677e-06}
{"train_loss": 0.032496556639671326, "global_step": 225848, "epoch": 2537, "lr": 6.82761463663053e-06}
{"train_loss": 0.036349277943372726, "global_step": 225849, "epoch": 2537, "lr": 6.8273221981523915e-06}
{"train_loss": 0.06862477213144302, "global_step": 225850, "epoch": 2537, "lr": 6.8270297654782826e-06}
{"train_loss": 0.06916691362857819, "global_step": 225851, "epoch": 2537, "lr": 6.826737338608264e-06}
{"train_loss": 0.08514012396335602, "global_step": 225852, "epoch": 2537, "lr": 6.8264449175423516e-06}
{"train_loss": 0.04302576184272766, "global_step": 225853, "epoch": 2537, "lr": 6.826152502280614e-06}
{"train_loss": 0.049646779894828796, "global_step": 225854, "epoch": 2537, "lr": 6.825860092823056e-06}
{"train_loss": 0.08520510792732239, "global_step": 225855, "epoch": 2537, "lr": 6.825567689169748e-06}
{"train_loss": 0.04829312860965729, "global_step": 225856, "epoch": 2537, "lr": 6.8252752913207095e-06}
{"train_loss": 0.03721003979444504, "global_step": 225857, "epoch": 2537, "lr": 6.8249828992759834e-06}
{"train_loss": 0.055382657796144485, "global_step": 225858, "epoch": 2537, "lr": 6.824690513035625e-06}
{"train_loss": 0.052706170827150345, "global_step": 225859, "epoch": 2537, "lr": 6.824398132599647e-06}
{"train_loss": 0.050019271671772, "global_step": 225860, "epoch": 2537, "lr": 6.824105757968113e-06}
{"train_loss": 0.05840253457427025, "global_step": 225861, "epoch": 2537, "lr": 6.823813389141043e-06}
{"train_loss": 0.03926975652575493, "global_step": 225862, "epoch": 2537, "lr": 6.8235210261184955e-06}
{"train_loss": 0.05068836361169815, "global_step": 225863, "epoch": 2537, "lr": 6.823228668900489e-06}
{"train_loss": 0.025670623406767845, "global_step": 225864, "epoch": 2537, "lr": 6.822936317487083e-06}
{"train_loss": 0.1231042891740799, "global_step": 225865, "epoch": 2537, "lr": 6.822643971878295e-06}
{"train_loss": 0.08735765516757965, "global_step": 225866, "epoch": 2537, "lr": 6.8223516320741924e-06}
{"train_loss": 0.04509308561682701, "global_step": 225867, "epoch": 2537, "lr": 6.8220592980747784e-06}
{"train_loss": 0.030113620683550835, "global_step": 225868, "epoch": 2537, "lr": 6.821766969880134e-06}
{"train_loss": 0.12832553684711456, "global_step": 225869, "epoch": 2537, "lr": 6.821474647490256e-06}
{"train_loss": 0.046141572296619415, "global_step": 225870, "epoch": 2537, "lr": 6.821182330905223e-06}
{"train_loss": 0.045455463230609894, "global_step": 225871, "epoch": 2537, "lr": 6.820890020125037e-06}
{"train_loss": 0.11464010924100876, "global_step": 225872, "epoch": 2537, "lr": 6.820597715149774e-06}
{"train_loss": 0.09196572750806808, "global_step": 225873, "epoch": 2537, "lr": 6.820305415979444e-06}
{"train_loss": 0.09862729907035828, "global_step": 225874, "epoch": 2537, "lr": 6.820013122614105e-06}
{"train_loss": 0.058699555695056915, "global_step": 225875, "epoch": 2537, "lr": 6.819720835053778e-06}
{"train_loss": 0.07905411720275879, "global_step": 225876, "epoch": 2537, "lr": 6.819428553298529e-06}
{"train_loss": 0.0565071702003479, "global_step": 225877, "epoch": 2537, "lr": 6.819136277348365e-06}
{"train_loss": 0.059848468750715256, "global_step": 225878, "epoch": 2537, "lr": 6.8188440072033564e-06}
{"train_loss": 0.04520747810602188, "global_step": 225879, "epoch": 2537, "lr": 6.818551742863527e-06}
{"train_loss": 0.09775950759649277, "global_step": 225880, "epoch": 2537, "lr": 6.818259484328904e-06}
{"train_loss": 0.05885755544884151, "global_step": 225881, "epoch": 2537, "lr": 6.817967231599554e-06, "val_loss": 8.987988471984863}
{"train_loss": 0.06058512628078461, "global_step": 225882, "epoch": 2538, "lr": 6.8176749846754875e-06}
{"train_loss": 0.02347382716834545, "global_step": 225883, "epoch": 2538, "lr": 6.817382743556772e-06}
{"train_loss": 0.049355220049619675, "global_step": 225884, "epoch": 2538, "lr": 6.817090508243418e-06}
{"train_loss": 0.036916520446538925, "global_step": 225885, "epoch": 2538, "lr": 6.816798278735498e-06}
{"train_loss": 0.04107647389173508, "global_step": 225886, "epoch": 2538, "lr": 6.816506055033018e-06}
{"train_loss": 0.03474658355116844, "global_step": 225887, "epoch": 2538, "lr": 6.816213837136043e-06}
{"train_loss": 0.0318194217979908, "global_step": 225888, "epoch": 2538, "lr": 6.8159216250445914e-06}
{"train_loss": 0.03344738110899925, "global_step": 225889, "epoch": 2538, "lr": 6.815629418758723e-06}
{"train_loss": 0.07388103008270264, "global_step": 225890, "epoch": 2538, "lr": 6.815337218278456e-06}
{"train_loss": 0.027416609227657318, "global_step": 225891, "epoch": 2538, "lr": 6.81504502360385e-06}
{"train_loss": 0.04818233475089073, "global_step": 225892, "epoch": 2538, "lr": 6.8147528347349275e-06}
{"train_loss": 0.0562785305082798, "global_step": 225893, "epoch": 2538, "lr": 6.814460651671734e-06}
{"train_loss": 0.06627243012189865, "global_step": 225894, "epoch": 2538, "lr": 6.814168474414317e-06}
{"train_loss": 0.013713940978050232, "global_step": 225895, "epoch": 2538, "lr": 6.813876302962702e-06}
{"train_loss": 0.05086798965930939, "global_step": 225896, "epoch": 2538, "lr": 6.813584137316948e-06}
{"train_loss": 0.06829501688480377, "global_step": 225897, "epoch": 2538, "lr": 6.8132919774770655e-06}
{"train_loss": 0.06963147968053818, "global_step": 225898, "epoch": 2538, "lr": 6.8129998234431225e-06}
{"train_loss": 0.07151611894369125, "global_step": 225899, "epoch": 2538, "lr": 6.812707675215135e-06}
{"train_loss": 0.05568946525454521, "global_step": 225900, "epoch": 2538, "lr": 6.812415532793159e-06}
{"train_loss": 0.04183211177587509, "global_step": 225901, "epoch": 2538, "lr": 6.812123396177222e-06}
{"train_loss": 0.0510222390294075, "global_step": 225902, "epoch": 2538, "lr": 6.81183126536738e-06}
{"train_loss": 0.05090213194489479, "global_step": 225903, "epoch": 2538, "lr": 6.811539140363648e-06}
{"train_loss": 0.03636191412806511, "global_step": 225904, "epoch": 2538, "lr": 6.811247021166089e-06}
{"train_loss": 0.03517656773328781, "global_step": 225905, "epoch": 2538, "lr": 6.810954907774719e-06}
{"train_loss": 0.03663437440991402, "global_step": 225906, "epoch": 2538, "lr": 6.810662800189599e-06}
{"train_loss": 0.040411993861198425, "global_step": 225907, "epoch": 2538, "lr": 6.810370698410751e-06}
{"train_loss": 0.08276677876710892, "global_step": 225908, "epoch": 2538, "lr": 6.810078602438236e-06}
{"train_loss": 0.033345967531204224, "global_step": 225909, "epoch": 2538, "lr": 6.809786512272065e-06}
{"train_loss": 0.023793835192918777, "global_step": 225910, "epoch": 2538, "lr": 6.809494427912294e-06}
{"train_loss": 0.0759357139468193, "global_step": 225911, "epoch": 2538, "lr": 6.809202349358967e-06}
{"train_loss": 0.040757372975349426, "global_step": 225912, "epoch": 2538, "lr": 6.808910276612118e-06}
{"train_loss": 0.08435321599245071, "global_step": 225913, "epoch": 2538, "lr": 6.808618209671774e-06}
{"train_loss": 0.02620980702340603, "global_step": 225914, "epoch": 2538, "lr": 6.8083261485379965e-06}
{"train_loss": 0.021466806530952454, "global_step": 225915, "epoch": 2538, "lr": 6.8080340932108026e-06}
{"train_loss": 0.042277947068214417, "global_step": 225916, "epoch": 2538, "lr": 6.807742043690252e-06}
{"train_loss": 0.04385608062148094, "global_step": 225917, "epoch": 2538, "lr": 6.807449999976362e-06}
{"train_loss": 0.0985926166176796, "global_step": 225918, "epoch": 2538, "lr": 6.8071579620692e-06}
{"train_loss": 0.06563368439674377, "global_step": 225919, "epoch": 2538, "lr": 6.806865929968787e-06}
{"train_loss": 0.062427110970020294, "global_step": 225920, "epoch": 2538, "lr": 6.8065739036751506e-06}
{"train_loss": 0.06236555799841881, "global_step": 225921, "epoch": 2538, "lr": 6.806281883188354e-06}
{"train_loss": 0.11471377313137054, "global_step": 225922, "epoch": 2538, "lr": 6.805989868508417e-06}
{"train_loss": 0.05874878913164139, "global_step": 225923, "epoch": 2538, "lr": 6.805697859635402e-06}
{"train_loss": 0.04411628469824791, "global_step": 225924, "epoch": 2538, "lr": 6.805405856569319e-06}
{"train_loss": 0.034124501049518585, "global_step": 225925, "epoch": 2538, "lr": 6.805113859310236e-06}
{"train_loss": 0.05825026333332062, "global_step": 225926, "epoch": 2538, "lr": 6.804821867858169e-06}
{"train_loss": 0.022139504551887512, "global_step": 225927, "epoch": 2538, "lr": 6.804529882213178e-06}
{"train_loss": 0.05984467640519142, "global_step": 225928, "epoch": 2538, "lr": 6.804237902375277e-06}
{"train_loss": 0.06391902267932892, "global_step": 225929, "epoch": 2538, "lr": 6.803945928344524e-06}
{"train_loss": 0.05401402711868286, "global_step": 225930, "epoch": 2538, "lr": 6.803653960120964e-06}
{"train_loss": 0.04023777320981026, "global_step": 225931, "epoch": 2538, "lr": 6.803361997704616e-06}
{"train_loss": 0.04317327216267586, "global_step": 225932, "epoch": 2538, "lr": 6.803070041095539e-06}
{"train_loss": 0.0625295341014862, "global_step": 225933, "epoch": 2538, "lr": 6.802778090293749e-06}
{"train_loss": 0.07577022910118103, "global_step": 225934, "epoch": 2538, "lr": 6.802486145299314e-06}
{"train_loss": 0.05543314665555954, "global_step": 225935, "epoch": 2538, "lr": 6.802194206112245e-06}
{"train_loss": 0.055395446717739105, "global_step": 225936, "epoch": 2538, "lr": 6.801902272732608e-06}
{"train_loss": 0.05938767269253731, "global_step": 225937, "epoch": 2538, "lr": 6.8016103451604145e-06}
{"train_loss": 0.061088815331459045, "global_step": 225938, "epoch": 2538, "lr": 6.8013184233957304e-06}
{"train_loss": 0.024410495534539223, "global_step": 225939, "epoch": 2538, "lr": 6.801026507438574e-06}
{"train_loss": 0.07516343891620636, "global_step": 225940, "epoch": 2538, "lr": 6.800734597288999e-06}
{"train_loss": 0.03432789444923401, "global_step": 225941, "epoch": 2538, "lr": 6.800442692947034e-06}
{"train_loss": 0.06342373788356781, "global_step": 225942, "epoch": 2538, "lr": 6.800150794412729e-06}
{"train_loss": 0.07953940331935883, "global_step": 225943, "epoch": 2538, "lr": 6.799858901686107e-06}
{"train_loss": 0.04659963771700859, "global_step": 225944, "epoch": 2538, "lr": 6.799567014767227e-06}
{"train_loss": 0.056770555675029755, "global_step": 225945, "epoch": 2538, "lr": 6.799275133656113e-06}
{"train_loss": 0.02554793655872345, "global_step": 225946, "epoch": 2538, "lr": 6.798983258352803e-06}
{"train_loss": 0.0356656052172184, "global_step": 225947, "epoch": 2538, "lr": 6.798691388857359e-06}
{"train_loss": 0.0567886084318161, "global_step": 225948, "epoch": 2538, "lr": 6.798399525169796e-06}
{"train_loss": 0.08150731027126312, "global_step": 225949, "epoch": 2538, "lr": 6.798107667290166e-06}
{"train_loss": 0.043643105775117874, "global_step": 225950, "epoch": 2538, "lr": 6.7978158152184945e-06}
{"train_loss": 0.03566828742623329, "global_step": 225951, "epoch": 2538, "lr": 6.797523968954844e-06}
{"train_loss": 0.04961296170949936, "global_step": 225952, "epoch": 2538, "lr": 6.7972321284992365e-06}
{"train_loss": 0.0695473849773407, "global_step": 225953, "epoch": 2538, "lr": 6.796940293851706e-06}
{"train_loss": 0.031696829944849014, "global_step": 225954, "epoch": 2538, "lr": 6.796648465012312e-06}
{"train_loss": 0.07357237488031387, "global_step": 225955, "epoch": 2538, "lr": 6.796356641981072e-06}
{"train_loss": 0.011305464431643486, "global_step": 225956, "epoch": 2538, "lr": 6.7960648247580415e-06}
{"train_loss": 0.023242516443133354, "global_step": 225957, "epoch": 2538, "lr": 6.795773013343254e-06}
{"train_loss": 0.09637080878019333, "global_step": 225958, "epoch": 2538, "lr": 6.795481207736742e-06}
{"train_loss": 0.030174318701028824, "global_step": 225959, "epoch": 2538, "lr": 6.795189407938557e-06}
{"train_loss": 0.08172155171632767, "global_step": 225960, "epoch": 2538, "lr": 6.794897613948726e-06}
{"train_loss": 0.05505616217851639, "global_step": 225961, "epoch": 2538, "lr": 6.794605825767303e-06}
{"train_loss": 0.047104157507419586, "global_step": 225962, "epoch": 2538, "lr": 6.794314043394306e-06}
{"train_loss": 0.039038415998220444, "global_step": 225963, "epoch": 2538, "lr": 6.794022266829797e-06}
{"train_loss": 0.04858957231044769, "global_step": 225964, "epoch": 2538, "lr": 6.793730496073797e-06}
{"train_loss": 0.04326682910323143, "global_step": 225965, "epoch": 2538, "lr": 6.793438731126356e-06}
{"train_loss": 0.03684403374791145, "global_step": 225966, "epoch": 2538, "lr": 6.793146971987513e-06}
{"train_loss": 0.0409616157412529, "global_step": 225967, "epoch": 2538, "lr": 6.7928552186573016e-06}
{"train_loss": 0.10482758283615112, "global_step": 225968, "epoch": 2538, "lr": 6.792563471135771e-06}
{"train_loss": 0.054529301822185516, "global_step": 225969, "epoch": 2538, "lr": 6.792271729422944e-06}
{"train_loss": 0.051645248395840775, "global_step": 225970, "epoch": 2538, "lr": 6.791979993518882e-06, "val_loss": 8.988007545471191}
{"train_loss": 0.09499794989824295, "global_step": 225971, "epoch": 2539, "lr": 6.791688263423601e-06}
{"train_loss": 0.07191458344459534, "global_step": 225972, "epoch": 2539, "lr": 6.791396539137157e-06}
{"train_loss": 0.07367563247680664, "global_step": 225973, "epoch": 2539, "lr": 6.791104820659577e-06}
{"train_loss": 0.04886261746287346, "global_step": 225974, "epoch": 2539, "lr": 6.790813107990917e-06}
{"train_loss": 0.02748478390276432, "global_step": 225975, "epoch": 2539, "lr": 6.790521401131189e-06}
{"train_loss": 0.035292692482471466, "global_step": 225976, "epoch": 2539, "lr": 6.7902297000804634e-06}
{"train_loss": 0.07186063379049301, "global_step": 225977, "epoch": 2539, "lr": 6.7899380048387585e-06}
{"train_loss": 0.07331345230340958, "global_step": 225978, "epoch": 2539, "lr": 6.789646315406123e-06}
{"train_loss": 0.04175593703985214, "global_step": 225979, "epoch": 2539, "lr": 6.789354631782585e-06}
{"train_loss": 0.05483437702059746, "global_step": 225980, "epoch": 2539, "lr": 6.789062953968206e-06}
{"train_loss": 0.03880760818719864, "global_step": 225981, "epoch": 2539, "lr": 6.7887712819629964e-06}
{"train_loss": 0.06421604752540588, "global_step": 225982, "epoch": 2539, "lr": 6.7884796157670186e-06}
{"train_loss": 0.059780459851026535, "global_step": 225983, "epoch": 2539, "lr": 6.788187955380298e-06}
{"train_loss": 0.0343288779258728, "global_step": 225984, "epoch": 2539, "lr": 6.787896300802876e-06}
{"train_loss": 0.06078307330608368, "global_step": 225985, "epoch": 2539, "lr": 6.787604652034802e-06}
{"train_loss": 0.0103685911744833, "global_step": 225986, "epoch": 2539, "lr": 6.787313009076102e-06}
{"train_loss": 0.0411829799413681, "global_step": 225987, "epoch": 2539, "lr": 6.787021371926827e-06}
{"train_loss": 0.0655527338385582, "global_step": 225988, "epoch": 2539, "lr": 6.786729740587005e-06}
{"train_loss": 0.07842367887496948, "global_step": 225989, "epoch": 2539, "lr": 6.786438115056687e-06}
{"train_loss": 0.024461181834340096, "global_step": 225990, "epoch": 2539, "lr": 6.78614649533591e-06}
{"train_loss": 0.06344965845346451, "global_step": 225991, "epoch": 2539, "lr": 6.785854881424697e-06}
{"train_loss": 0.051332179456949234, "global_step": 225992, "epoch": 2539, "lr": 6.785563273323109e-06}
{"train_loss": 0.03216056898236275, "global_step": 225993, "epoch": 2539, "lr": 6.785271671031163e-06}
{"train_loss": 0.04264752194285393, "global_step": 225994, "epoch": 2539, "lr": 6.784980074548919e-06}
{"train_loss": 0.0469522625207901, "global_step": 225995, "epoch": 2539, "lr": 6.784688483876412e-06}
{"train_loss": 0.05966721847653389, "global_step": 225996, "epoch": 2539, "lr": 6.784396899013667e-06}
{"train_loss": 0.05098804086446762, "global_step": 225997, "epoch": 2539, "lr": 6.784105319960743e-06}
{"train_loss": 0.026314016431570053, "global_step": 225998, "epoch": 2539, "lr": 6.7838137467176544e-06}
{"train_loss": 0.03775620460510254, "global_step": 225999, "epoch": 2539, "lr": 6.783522179284468e-06}
{"train_loss": 0.029512016102671623, "global_step": 226000, "epoch": 2539, "lr": 6.783230617661202e-06}
{"train_loss": 0.059300150722265244, "global_step": 226001, "epoch": 2539, "lr": 6.782939061847904e-06}
{"train_loss": 0.05815298110246658, "global_step": 226002, "epoch": 2539, "lr": 6.782647511844625e-06}
{"train_loss": 0.0839875340461731, "global_step": 226003, "epoch": 2539, "lr": 6.782355967651377e-06}
{"train_loss": 0.07159852236509323, "global_step": 226004, "epoch": 2539, "lr": 6.782064429268226e-06}
{"train_loss": 0.04142623022198677, "global_step": 226005, "epoch": 2539, "lr": 6.781772896695193e-06}
{"train_loss": 0.09453044086694717, "global_step": 226006, "epoch": 2539, "lr": 6.78148136993233e-06}
{"train_loss": 0.061037275940179825, "global_step": 226007, "epoch": 2539, "lr": 6.781189848979663e-06}
{"train_loss": 0.11374030262231827, "global_step": 226008, "epoch": 2539, "lr": 6.780898333837243e-06}
{"train_loss": 0.060522228479385376, "global_step": 226009, "epoch": 2539, "lr": 6.780606824505098e-06}
{"train_loss": 0.036930788308382034, "global_step": 226010, "epoch": 2539, "lr": 6.780315320983288e-06}
{"train_loss": 0.04432972893118858, "global_step": 226011, "epoch": 2539, "lr": 6.780023823271825e-06}
{"train_loss": 0.06344185024499893, "global_step": 226012, "epoch": 2539, "lr": 6.779732331370769e-06}
{"train_loss": 0.038478486239910126, "global_step": 226013, "epoch": 2539, "lr": 6.779440845280138e-06}
{"train_loss": 0.05533374845981598, "global_step": 226014, "epoch": 2539, "lr": 6.779149365000004e-06}
{"train_loss": 0.0655837282538414, "global_step": 226015, "epoch": 2539, "lr": 6.778857890530365e-06}
{"train_loss": 0.06716399639844894, "global_step": 226016, "epoch": 2539, "lr": 6.778566421871302e-06}
{"train_loss": 0.036850325763225555, "global_step": 226017, "epoch": 2539, "lr": 6.778274959022818e-06}
{"train_loss": 0.06847977638244629, "global_step": 226018, "epoch": 2539, "lr": 6.77798350198498e-06}
{"train_loss": 0.06243671476840973, "global_step": 226019, "epoch": 2539, "lr": 6.777692050757806e-06}
{"train_loss": 0.04540745168924332, "global_step": 226020, "epoch": 2539, "lr": 6.777400605341344e-06}
{"train_loss": 0.1066826730966568, "global_step": 226021, "epoch": 2539, "lr": 6.77710916573564e-06}
{"train_loss": 0.03687530755996704, "global_step": 226022, "epoch": 2539, "lr": 6.776817731940721e-06}
{"train_loss": 0.015113966539502144, "global_step": 226023, "epoch": 2539, "lr": 6.776526303956643e-06}
{"train_loss": 0.038645870983600616, "global_step": 226024, "epoch": 2539, "lr": 6.776234881783422e-06}
{"train_loss": 0.0970345288515091, "global_step": 226025, "epoch": 2539, "lr": 6.77594346542112e-06}
{"train_loss": 0.04199381172657013, "global_step": 226026, "epoch": 2539, "lr": 6.775652054869752e-06}
{"train_loss": 0.044861312955617905, "global_step": 226027, "epoch": 2539, "lr": 6.7753606501293866e-06}
{"train_loss": 0.04134995862841606, "global_step": 226028, "epoch": 2539, "lr": 6.775069251200034e-06}
{"train_loss": 0.06290099024772644, "global_step": 226029, "epoch": 2539, "lr": 6.77477785808176e-06}
{"train_loss": 0.027954932302236557, "global_step": 226030, "epoch": 2539, "lr": 6.774486470774588e-06}
{"train_loss": 0.0235463697463274, "global_step": 226031, "epoch": 2539, "lr": 6.774195089278546e-06}
{"train_loss": 0.07086561620235443, "global_step": 226032, "epoch": 2539, "lr": 6.773903713593699e-06}
{"train_loss": 0.09650745242834091, "global_step": 226033, "epoch": 2539, "lr": 6.773612343720065e-06}
{"train_loss": 0.012140842154622078, "global_step": 226034, "epoch": 2539, "lr": 6.773320979657705e-06}
{"train_loss": 0.049974389374256134, "global_step": 226035, "epoch": 2539, "lr": 6.77302962140664e-06}
{"train_loss": 0.046906549483537674, "global_step": 226036, "epoch": 2539, "lr": 6.772738268966905e-06}
{"train_loss": 0.07647102326154709, "global_step": 226037, "epoch": 2539, "lr": 6.772446922338549e-06}
{"train_loss": 0.09231216460466385, "global_step": 226038, "epoch": 2539, "lr": 6.772155581521622e-06}
{"train_loss": 0.02421681024134159, "global_step": 226039, "epoch": 2539, "lr": 6.771864246516141e-06}
{"train_loss": 0.030135558918118477, "global_step": 226040, "epoch": 2539, "lr": 6.771572917322167e-06}
{"train_loss": 0.0442805252969265, "global_step": 226041, "epoch": 2539, "lr": 6.771281593939716e-06}
{"train_loss": 0.061187587678432465, "global_step": 226042, "epoch": 2539, "lr": 6.770990276368849e-06}
{"train_loss": 0.060073304921388626, "global_step": 226043, "epoch": 2539, "lr": 6.7706989646095845e-06}
{"train_loss": 0.04383176192641258, "global_step": 226044, "epoch": 2539, "lr": 6.770407658661987e-06}
{"train_loss": 0.04638106748461723, "global_step": 226045, "epoch": 2539, "lr": 6.7701163585260685e-06}
{"train_loss": 0.05202729254961014, "global_step": 226046, "epoch": 2539, "lr": 6.76982506420189e-06}
{"train_loss": 0.06622658669948578, "global_step": 226047, "epoch": 2539, "lr": 6.769533775689474e-06}
{"train_loss": 0.04913684353232384, "global_step": 226048, "epoch": 2539, "lr": 6.769242492988875e-06}
{"train_loss": 0.02279079519212246, "global_step": 226049, "epoch": 2539, "lr": 6.768951216100117e-06}
{"train_loss": 0.066050685942173, "global_step": 226050, "epoch": 2539, "lr": 6.7686599450232536e-06}
{"train_loss": 0.05834183469414711, "global_step": 226051, "epoch": 2539, "lr": 6.768368679758308e-06}
{"train_loss": 0.03959690034389496, "global_step": 226052, "epoch": 2539, "lr": 6.768077420305341e-06}
{"train_loss": 0.05754280462861061, "global_step": 226053, "epoch": 2539, "lr": 6.767786166664369e-06}
{"train_loss": 0.06553767621517181, "global_step": 226054, "epoch": 2539, "lr": 6.7674949188354485e-06}
{"train_loss": 0.03377678617835045, "global_step": 226055, "epoch": 2539, "lr": 6.767203676818601e-06}
{"train_loss": 0.049290429800748825, "global_step": 226056, "epoch": 2539, "lr": 6.766912440613876e-06}
{"train_loss": 0.04137568548321724, "global_step": 226057, "epoch": 2539, "lr": 6.766621210221325e-06}
{"train_loss": 0.04196970537304878, "global_step": 226058, "epoch": 2539, "lr": 6.766329985640962e-06}
{"train_loss": 0.053662559289610785, "global_step": 226059, "epoch": 2539, "lr": 6.766038766872851e-06, "val_loss": 8.910727500915527}
{"train_loss": 0.03816031292080879, "global_step": 226060, "epoch": 2540, "lr": 6.765747553917013e-06}
{"train_loss": 0.023527977988123894, "global_step": 226061, "epoch": 2540, "lr": 6.765456346773497e-06}
{"train_loss": 0.046626877039670944, "global_step": 226062, "epoch": 2540, "lr": 6.765165145442332e-06}
{"train_loss": 0.07000061124563217, "global_step": 226063, "epoch": 2540, "lr": 6.764873949923572e-06}
{"train_loss": 0.04335339739918709, "global_step": 226064, "epoch": 2540, "lr": 6.7645827602172366e-06}
{"train_loss": 0.05729786306619644, "global_step": 226065, "epoch": 2540, "lr": 6.764291576323389e-06}
{"train_loss": 0.03905520588159561, "global_step": 226066, "epoch": 2540, "lr": 6.764000398242043e-06}
{"train_loss": 0.06339064240455627, "global_step": 226067, "epoch": 2540, "lr": 6.763709225973258e-06}
{"train_loss": 0.04457159340381622, "global_step": 226068, "epoch": 2540, "lr": 6.763418059517069e-06}
{"train_loss": 0.0448979027569294, "global_step": 226069, "epoch": 2540, "lr": 6.763126898873501e-06}
{"train_loss": 0.06075584888458252, "global_step": 226070, "epoch": 2540, "lr": 6.762835744042617e-06}
{"train_loss": 0.08168162405490875, "global_step": 226071, "epoch": 2540, "lr": 6.7625445950244286e-06}
{"train_loss": 0.015610512346029282, "global_step": 226072, "epoch": 2540, "lr": 6.762253451819001e-06}
{"train_loss": 0.04883295297622681, "global_step": 226073, "epoch": 2540, "lr": 6.761962314426346e-06}
{"train_loss": 0.04017287492752075, "global_step": 226074, "epoch": 2540, "lr": 6.761671182846535e-06}
{"train_loss": 0.06809771806001663, "global_step": 226075, "epoch": 2540, "lr": 6.761380057079575e-06}
{"train_loss": 0.04423955827951431, "global_step": 226076, "epoch": 2540, "lr": 6.7610889371255315e-06}
{"train_loss": 0.07088376581668854, "global_step": 226077, "epoch": 2540, "lr": 6.760797822984427e-06}
{"train_loss": 0.10273467749357224, "global_step": 226078, "epoch": 2540, "lr": 6.760506714656312e-06}
{"train_loss": 0.05073515325784683, "global_step": 226079, "epoch": 2540, "lr": 6.760215612141213e-06}
{"train_loss": 0.04780770093202591, "global_step": 226080, "epoch": 2540, "lr": 6.759924515439181e-06}
{"train_loss": 0.05669024959206581, "global_step": 226081, "epoch": 2540, "lr": 6.759633424550243e-06}
{"train_loss": 0.05711644887924194, "global_step": 226082, "epoch": 2540, "lr": 6.7593423394744555e-06}
{"train_loss": 0.07799389958381653, "global_step": 226083, "epoch": 2540, "lr": 6.75905126021184e-06}
{"train_loss": 0.05675530433654785, "global_step": 226084, "epoch": 2540, "lr": 6.758760186762447e-06}
{"train_loss": 0.03590627759695053, "global_step": 226085, "epoch": 2540, "lr": 6.75846911912631e-06}
{"train_loss": 0.07029008120298386, "global_step": 226086, "epoch": 2540, "lr": 6.758178057303471e-06}
{"train_loss": 0.0484568327665329, "global_step": 226087, "epoch": 2540, "lr": 6.757887001293961e-06}
{"train_loss": 0.07885182648897171, "global_step": 226088, "epoch": 2540, "lr": 6.75759595109784e-06}
{"train_loss": 0.07936225086450577, "global_step": 226089, "epoch": 2540, "lr": 6.757304906715117e-06}
{"train_loss": 0.020661137998104095, "global_step": 226090, "epoch": 2540, "lr": 6.757013868145861e-06}
{"train_loss": 0.039575252681970596, "global_step": 226091, "epoch": 2540, "lr": 6.756722835390089e-06}
{"train_loss": 0.01906961388885975, "global_step": 226092, "epoch": 2540, "lr": 6.756431808447844e-06}
{"train_loss": 0.02360864169895649, "global_step": 226093, "epoch": 2540, "lr": 6.756140787319182e-06}
{"train_loss": 0.07117071747779846, "global_step": 226094, "epoch": 2540, "lr": 6.7558497720041194e-06}
{"train_loss": 0.05337463319301605, "global_step": 226095, "epoch": 2540, "lr": 6.755558762502717e-06}
{"train_loss": 0.05395613983273506, "global_step": 226096, "epoch": 2540, "lr": 6.755267758814993e-06}
{"train_loss": 0.058940403163433075, "global_step": 226097, "epoch": 2540, "lr": 6.754976760941006e-06}
{"train_loss": 0.07715331763029099, "global_step": 226098, "epoch": 2540, "lr": 6.754685768880769e-06}
{"train_loss": 0.09197601675987244, "global_step": 226099, "epoch": 2540, "lr": 6.754394782634355e-06}
{"train_loss": 0.049249809235334396, "global_step": 226100, "epoch": 2540, "lr": 6.754103802201772e-06}
{"train_loss": 0.05476444587111473, "global_step": 226101, "epoch": 2540, "lr": 6.753812827583084e-06}
{"train_loss": 0.03966865316033363, "global_step": 226102, "epoch": 2540, "lr": 6.753521858778311e-06}
{"train_loss": 0.07271763682365417, "global_step": 226103, "epoch": 2540, "lr": 6.75323089578751e-06}
{"train_loss": 0.055720601230859756, "global_step": 226104, "epoch": 2540, "lr": 6.752939938610692e-06}
{"train_loss": 0.09230396151542664, "global_step": 226105, "epoch": 2540, "lr": 6.752648987247934e-06}
{"train_loss": 0.03942669928073883, "global_step": 226106, "epoch": 2540, "lr": 6.752358041699247e-06}
{"train_loss": 0.036646775901317596, "global_step": 226107, "epoch": 2540, "lr": 6.752067101964677e-06}
{"train_loss": 0.024833105504512787, "global_step": 226108, "epoch": 2540, "lr": 6.751776168044266e-06}
{"train_loss": 0.04314956068992615, "global_step": 226109, "epoch": 2540, "lr": 6.751485239938043e-06}
{"train_loss": 0.06633824110031128, "global_step": 226110, "epoch": 2540, "lr": 6.7511943176460704e-06}
{"train_loss": 0.006647072266787291, "global_step": 226111, "epoch": 2540, "lr": 6.750903401168362e-06}
{"train_loss": 0.07674716413021088, "global_step": 226112, "epoch": 2540, "lr": 6.750612490504976e-06}
{"train_loss": 0.072560615837574, "global_step": 226113, "epoch": 2540, "lr": 6.750321585655933e-06}
{"train_loss": 0.06099230423569679, "global_step": 226114, "epoch": 2540, "lr": 6.750030686621289e-06}
{"train_loss": 0.050709281116724014, "global_step": 226115, "epoch": 2540, "lr": 6.749739793401072e-06}
{"train_loss": 0.0330340750515461, "global_step": 226116, "epoch": 2540, "lr": 6.7494489059953314e-06}
{"train_loss": 0.04793991148471832, "global_step": 226117, "epoch": 2540, "lr": 6.74915802440409e-06}
{"train_loss": 0.031153202056884766, "global_step": 226118, "epoch": 2540, "lr": 6.748867148627408e-06}
{"train_loss": 0.06255575269460678, "global_step": 226119, "epoch": 2540, "lr": 6.748576278665303e-06}
{"train_loss": 0.04798855632543564, "global_step": 226120, "epoch": 2540, "lr": 6.748285414517835e-06}
{"train_loss": 0.037416793406009674, "global_step": 226121, "epoch": 2540, "lr": 6.747994556185027e-06}
{"train_loss": 0.07061979174613953, "global_step": 226122, "epoch": 2540, "lr": 6.74770370366693e-06}
{"train_loss": 0.022282171994447708, "global_step": 226123, "epoch": 2540, "lr": 6.747412856963564e-06}
{"train_loss": 0.04593415558338165, "global_step": 226124, "epoch": 2540, "lr": 6.747122016074997e-06}
{"train_loss": 0.0428004264831543, "global_step": 226125, "epoch": 2540, "lr": 6.74683118100124e-06}
{"train_loss": 0.055660348385572433, "global_step": 226126, "epoch": 2540, "lr": 6.746540351742353e-06}
{"train_loss": 0.051594674587249756, "global_step": 226127, "epoch": 2540, "lr": 6.74624952829836e-06}
{"train_loss": 0.05530007556080818, "global_step": 226128, "epoch": 2540, "lr": 6.745958710669303e-06}
{"train_loss": 0.07923515141010284, "global_step": 226129, "epoch": 2540, "lr": 6.745667898855235e-06}
{"train_loss": 0.030974388122558594, "global_step": 226130, "epoch": 2540, "lr": 6.745377092856175e-06}
{"train_loss": 0.05775941535830498, "global_step": 226131, "epoch": 2540, "lr": 6.7450862926721866e-06}
{"train_loss": 0.04288383573293686, "global_step": 226132, "epoch": 2540, "lr": 6.74479549830328e-06}
{"train_loss": 0.08809405565261841, "global_step": 226133, "epoch": 2540, "lr": 6.744504709749516e-06}
{"train_loss": 0.036028917878866196, "global_step": 226134, "epoch": 2540, "lr": 6.744213927010917e-06}
{"train_loss": 0.013082781806588173, "global_step": 226135, "epoch": 2540, "lr": 6.7439231500875445e-06}
{"train_loss": 0.06367336958646774, "global_step": 226136, "epoch": 2540, "lr": 6.7436323789794145e-06}
{"train_loss": 0.07562632113695145, "global_step": 226137, "epoch": 2540, "lr": 6.743341613686582e-06}
{"train_loss": 0.055669356137514114, "global_step": 226138, "epoch": 2540, "lr": 6.743050854209076e-06}
{"train_loss": 0.04928107559680939, "global_step": 226139, "epoch": 2540, "lr": 6.742760100546947e-06}
{"train_loss": 0.09456775337457657, "global_step": 226140, "epoch": 2540, "lr": 6.742469352700209e-06}
{"train_loss": 0.1034468412399292, "global_step": 226141, "epoch": 2540, "lr": 6.742178610668936e-06}
{"train_loss": 0.05124262720346451, "global_step": 226142, "epoch": 2540, "lr": 6.74188787445314e-06}
{"train_loss": 0.05794106051325798, "global_step": 226143, "epoch": 2540, "lr": 6.74159714405288e-06}
{"train_loss": 0.034267496317625046, "global_step": 226144, "epoch": 2540, "lr": 6.741306419468174e-06}
{"train_loss": 0.029241103678941727, "global_step": 226145, "epoch": 2540, "lr": 6.741015700699083e-06}
{"train_loss": 0.04042257368564606, "global_step": 226146, "epoch": 2540, "lr": 6.740724987745634e-06}
{"train_loss": 0.04197821393609047, "global_step": 226147, "epoch": 2540, "lr": 6.74043428060786e-06}
{"train_loss": 0.0535743381315319, "global_step": 226148, "epoch": 2540, "lr": 6.7401435792858125e-06, "val_loss": 9.058633804321289, "train_action_mse_error": 8.256017684936523}
{"train_loss": 0.03918641060590744, "global_step": 226149, "epoch": 2541, "lr": 6.739852883779518e-06}
{"train_loss": 0.07530258595943451, "global_step": 226150, "epoch": 2541, "lr": 6.739562194089033e-06}
{"train_loss": 0.039968401193618774, "global_step": 226151, "epoch": 2541, "lr": 6.739271510214379e-06}
{"train_loss": 0.05278744921088219, "global_step": 226152, "epoch": 2541, "lr": 6.738980832155606e-06}
{"train_loss": 0.03769579529762268, "global_step": 226153, "epoch": 2541, "lr": 6.738690159912747e-06}
{"train_loss": 0.05895395576953888, "global_step": 226154, "epoch": 2541, "lr": 6.738399493485847e-06}
{"train_loss": 0.06821838021278381, "global_step": 226155, "epoch": 2541, "lr": 6.73810883287494e-06}
{"train_loss": 0.04099874943494797, "global_step": 226156, "epoch": 2541, "lr": 6.737818178080069e-06}
{"train_loss": 0.05701723322272301, "global_step": 226157, "epoch": 2541, "lr": 6.7375275291012675e-06}
{"train_loss": 0.07678882032632828, "global_step": 226158, "epoch": 2541, "lr": 6.737236885938586e-06}
{"train_loss": 0.0537504181265831, "global_step": 226159, "epoch": 2541, "lr": 6.736946248592041e-06}
{"train_loss": 0.04375528544187546, "global_step": 226160, "epoch": 2541, "lr": 6.736655617061704e-06}
{"train_loss": 0.03904943913221359, "global_step": 226161, "epoch": 2541, "lr": 6.736364991347583e-06}
{"train_loss": 0.04800084978342056, "global_step": 226162, "epoch": 2541, "lr": 6.736074371449741e-06}
{"train_loss": 0.08670752495527267, "global_step": 226163, "epoch": 2541, "lr": 6.735783757368191e-06}
{"train_loss": 0.06786932796239853, "global_step": 226164, "epoch": 2541, "lr": 6.735493149102995e-06}
{"train_loss": 0.07075652480125427, "global_step": 226165, "epoch": 2541, "lr": 6.73520254665419e-06}
{"train_loss": 0.03301152214407921, "global_step": 226166, "epoch": 2541, "lr": 6.734911950021799e-06}
{"train_loss": 0.10430370271205902, "global_step": 226167, "epoch": 2541, "lr": 6.734621359205889e-06}
{"train_loss": 0.05098162218928337, "global_step": 226168, "epoch": 2541, "lr": 6.734330774206465e-06}
{"train_loss": 0.03559665009379387, "global_step": 226169, "epoch": 2541, "lr": 6.7340401950235946e-06}
{"train_loss": 0.023398738354444504, "global_step": 226170, "epoch": 2541, "lr": 6.733749621657292e-06}
{"train_loss": 0.029157398268580437, "global_step": 226171, "epoch": 2541, "lr": 6.733459054107627e-06}
{"train_loss": 0.07158239185810089, "global_step": 226172, "epoch": 2541, "lr": 6.733168492374603e-06}
{"train_loss": 0.08704883605241776, "global_step": 226173, "epoch": 2541, "lr": 6.732877936458293e-06}
{"train_loss": 0.04569918289780617, "global_step": 226174, "epoch": 2541, "lr": 6.732587386358708e-06}
{"train_loss": 0.04938913881778717, "global_step": 226175, "epoch": 2541, "lr": 6.732296842075914e-06}
{"train_loss": 0.02068498358130455, "global_step": 226176, "epoch": 2541, "lr": 6.732006303609917e-06}
{"train_loss": 0.04562680423259735, "global_step": 226177, "epoch": 2541, "lr": 6.731715770960789e-06}
{"train_loss": 0.04323336109519005, "global_step": 226178, "epoch": 2541, "lr": 6.731425244128542e-06}
{"train_loss": 0.0406591072678566, "global_step": 226179, "epoch": 2541, "lr": 6.731134723113241e-06}
{"train_loss": 0.035496797412633896, "global_step": 226180, "epoch": 2541, "lr": 6.730844207914899e-06}
{"train_loss": 0.08550507575273514, "global_step": 226181, "epoch": 2541, "lr": 6.730553698533582e-06}
{"train_loss": 0.0866590142250061, "global_step": 226182, "epoch": 2541, "lr": 6.7302631949692995e-06}
{"train_loss": 0.07473070919513702, "global_step": 226183, "epoch": 2541, "lr": 6.729972697222114e-06}
{"train_loss": 0.0373668447136879, "global_step": 226184, "epoch": 2541, "lr": 6.729682205292065e-06}
{"train_loss": 0.045881714671850204, "global_step": 226185, "epoch": 2541, "lr": 6.729391719179162e-06}
{"train_loss": 0.04095577448606491, "global_step": 226186, "epoch": 2541, "lr": 6.729101238883484e-06}
{"train_loss": 0.09609783440828323, "global_step": 226187, "epoch": 2541, "lr": 6.728810764405035e-06}
{"train_loss": 0.033652566373348236, "global_step": 226188, "epoch": 2541, "lr": 6.728520295743884e-06}
{"train_loss": 0.08070012181997299, "global_step": 226189, "epoch": 2541, "lr": 6.728229832900046e-06}
{"train_loss": 0.08299872279167175, "global_step": 226190, "epoch": 2541, "lr": 6.727939375873576e-06}
{"train_loss": 0.02996532991528511, "global_step": 226191, "epoch": 2541, "lr": 6.727648924664498e-06}
{"train_loss": 0.07018901407718658, "global_step": 226192, "epoch": 2541, "lr": 6.727358479272872e-06}
{"train_loss": 0.026052920147776604, "global_step": 226193, "epoch": 2541, "lr": 6.727068039698714e-06}
{"train_loss": 0.04235489293932915, "global_step": 226194, "epoch": 2541, "lr": 6.726777605942086e-06}
{"train_loss": 0.049045659601688385, "global_step": 226195, "epoch": 2541, "lr": 6.726487178003005e-06}
{"train_loss": 0.04509352147579193, "global_step": 226196, "epoch": 2541, "lr": 6.726196755881531e-06}
{"train_loss": 0.08533109724521637, "global_step": 226197, "epoch": 2541, "lr": 6.725906339577681e-06}
{"train_loss": 0.06472023576498032, "global_step": 226198, "epoch": 2541, "lr": 6.725615929091517e-06}
{"train_loss": 0.05894649028778076, "global_step": 226199, "epoch": 2541, "lr": 6.725325524423054e-06}
{"train_loss": 0.0448002889752388, "global_step": 226200, "epoch": 2541, "lr": 6.725035125572343e-06}
{"train_loss": 0.05171116068959236, "global_step": 226201, "epoch": 2541, "lr": 6.72474473253944e-06}
{"train_loss": 0.0769558697938919, "global_step": 226202, "epoch": 2541, "lr": 6.724454345324349e-06}
{"train_loss": 0.05076254904270172, "global_step": 226203, "epoch": 2541, "lr": 6.724163963927143e-06}
{"train_loss": 0.05041303485631943, "global_step": 226204, "epoch": 2541, "lr": 6.7238735883478345e-06}
{"train_loss": 0.033128805458545685, "global_step": 226205, "epoch": 2541, "lr": 6.7235832185864824e-06}
{"train_loss": 0.039386384189128876, "global_step": 226206, "epoch": 2541, "lr": 6.72329285464311e-06}
{"train_loss": 0.019731596112251282, "global_step": 226207, "epoch": 2541, "lr": 6.723002496517772e-06}
{"train_loss": 0.05773656815290451, "global_step": 226208, "epoch": 2541, "lr": 6.722712144210486e-06}
{"train_loss": 0.05365041643381119, "global_step": 226209, "epoch": 2541, "lr": 6.7224217977213176e-06}
{"train_loss": 0.038950107991695404, "global_step": 226210, "epoch": 2541, "lr": 6.7221314570502795e-06}
{"train_loss": 0.036638982594013214, "global_step": 226211, "epoch": 2541, "lr": 6.721841122197436e-06}
{"train_loss": 0.04690255597233772, "global_step": 226212, "epoch": 2541, "lr": 6.7215507931628004e-06}
{"train_loss": 0.06697021424770355, "global_step": 226213, "epoch": 2541, "lr": 6.721260469946439e-06}
{"train_loss": 0.03890877217054367, "global_step": 226214, "epoch": 2541, "lr": 6.720970152548361e-06}
{"train_loss": 0.029031241312623024, "global_step": 226215, "epoch": 2541, "lr": 6.720679840968636e-06}
{"train_loss": 0.01865309476852417, "global_step": 226216, "epoch": 2541, "lr": 6.720389535207277e-06}
{"train_loss": 0.05685631185770035, "global_step": 226217, "epoch": 2541, "lr": 6.720099235264343e-06}
{"train_loss": 0.0660506933927536, "global_step": 226218, "epoch": 2541, "lr": 6.719808941139854e-06}
{"train_loss": 0.04349488392472267, "global_step": 226219, "epoch": 2541, "lr": 6.719518652833873e-06}
{"train_loss": 0.039103541523218155, "global_step": 226220, "epoch": 2541, "lr": 6.719228370346409e-06}
{"train_loss": 0.06206898018717766, "global_step": 226221, "epoch": 2541, "lr": 6.718938093677535e-06}
{"train_loss": 0.05468439683318138, "global_step": 226222, "epoch": 2541, "lr": 6.718647822827268e-06}
{"train_loss": 0.06003953516483307, "global_step": 226223, "epoch": 2541, "lr": 6.71835755779564e-06}
{"train_loss": 0.051501594483852386, "global_step": 226224, "epoch": 2541, "lr": 6.7180672985827085e-06}
{"train_loss": 0.01561777200549841, "global_step": 226225, "epoch": 2541, "lr": 6.7177770451885e-06}
{"train_loss": 0.04952273145318031, "global_step": 226226, "epoch": 2541, "lr": 6.7174867976130705e-06}
{"train_loss": 0.03845924139022827, "global_step": 226227, "epoch": 2541, "lr": 6.717196555856431e-06}
{"train_loss": 0.06529122591018677, "global_step": 226228, "epoch": 2541, "lr": 6.7169063199186536e-06}
{"train_loss": 0.03264598548412323, "global_step": 226229, "epoch": 2541, "lr": 6.7166160897997435e-06}
{"train_loss": 0.033624548465013504, "global_step": 226230, "epoch": 2541, "lr": 6.716325865499773e-06}
{"train_loss": 0.06358353048563004, "global_step": 226231, "epoch": 2541, "lr": 6.716035647018753e-06}
{"train_loss": 0.035336367785930634, "global_step": 226232, "epoch": 2541, "lr": 6.715745434356746e-06}
{"train_loss": 0.022870922461152077, "global_step": 226233, "epoch": 2541, "lr": 6.715455227513767e-06}
{"train_loss": 0.11464569717645645, "global_step": 226234, "epoch": 2541, "lr": 6.715165026489878e-06}
{"train_loss": 0.046950772404670715, "global_step": 226235, "epoch": 2541, "lr": 6.714874831285095e-06}
{"train_loss": 0.05376191437244415, "global_step": 226236, "epoch": 2541, "lr": 6.714584641899485e-06}
{"train_loss": 0.05202082740247584, "global_step": 226237, "epoch": 2541, "lr": 6.7142944583330595e-06, "val_loss": 9.041360855102539}
{"train_loss": 0.08026009052991867, "global_step": 226238, "epoch": 2542, "lr": 6.714004280585867e-06}
{"train_loss": 0.07441633939743042, "global_step": 226239, "epoch": 2542, "lr": 6.713714108657959e-06}
{"train_loss": 0.0437188595533371, "global_step": 226240, "epoch": 2542, "lr": 6.713423942549357e-06}
{"train_loss": 0.025436174124479294, "global_step": 226241, "epoch": 2542, "lr": 6.713133782260117e-06}
{"train_loss": 0.04748520627617836, "global_step": 226242, "epoch": 2542, "lr": 6.712843627790261e-06}
{"train_loss": 0.041684091091156006, "global_step": 226243, "epoch": 2542, "lr": 6.712553479139844e-06}
{"train_loss": 0.06863686442375183, "global_step": 226244, "epoch": 2542, "lr": 6.7122633363088825e-06}
{"train_loss": 0.026915293186903, "global_step": 226245, "epoch": 2542, "lr": 6.711973199297444e-06}
{"train_loss": 0.0755411684513092, "global_step": 226246, "epoch": 2542, "lr": 6.71168306810554e-06}
{"train_loss": 0.046534132212400436, "global_step": 226247, "epoch": 2542, "lr": 6.711392942733236e-06}
{"train_loss": 0.036852847784757614, "global_step": 226248, "epoch": 2542, "lr": 6.711102823180543e-06}
{"train_loss": 0.08254171907901764, "global_step": 226249, "epoch": 2542, "lr": 6.710812709447528e-06}
{"train_loss": 0.05264636129140854, "global_step": 226250, "epoch": 2542, "lr": 6.710522601534208e-06}
{"train_loss": 0.06310215592384338, "global_step": 226251, "epoch": 2542, "lr": 6.710232499440638e-06}
{"train_loss": 0.05258486419916153, "global_step": 226252, "epoch": 2542, "lr": 6.7099424031668414e-06}
{"train_loss": 0.07647742331027985, "global_step": 226253, "epoch": 2542, "lr": 6.7096523127128715e-06}
{"train_loss": 0.08331522345542908, "global_step": 226254, "epoch": 2542, "lr": 6.709362228078758e-06}
{"train_loss": 0.07641586661338806, "global_step": 226255, "epoch": 2542, "lr": 6.70907214926455e-06}
{"train_loss": 0.0511695072054863, "global_step": 226256, "epoch": 2542, "lr": 6.708782076270264e-06}
{"train_loss": 0.030310001224279404, "global_step": 226257, "epoch": 2542, "lr": 6.7084920090959734e-06}
{"train_loss": 0.03498780354857445, "global_step": 226258, "epoch": 2542, "lr": 6.708201947741682e-06}
{"train_loss": 0.029575621709227562, "global_step": 226259, "epoch": 2542, "lr": 6.707911892207458e-06}
{"train_loss": 0.03668809309601784, "global_step": 226260, "epoch": 2542, "lr": 6.707621842493316e-06}
{"train_loss": 0.06077079847455025, "global_step": 226261, "epoch": 2542, "lr": 6.70733179859932e-06}
{"train_loss": 0.038016125559806824, "global_step": 226262, "epoch": 2542, "lr": 6.707041760525495e-06}
{"train_loss": 0.08899438381195068, "global_step": 226263, "epoch": 2542, "lr": 6.706751728271871e-06}
{"train_loss": 0.03505701944231987, "global_step": 226264, "epoch": 2542, "lr": 6.706461701838502e-06}
{"train_loss": 0.05442902445793152, "global_step": 226265, "epoch": 2542, "lr": 6.706171681225415e-06}
{"train_loss": 0.03441630303859711, "global_step": 226266, "epoch": 2542, "lr": 6.705881666432662e-06}
{"train_loss": 0.11663772910833359, "global_step": 226267, "epoch": 2542, "lr": 6.70559165746027e-06}
{"train_loss": 0.04931841045618057, "global_step": 226268, "epoch": 2542, "lr": 6.705301654308294e-06}
{"train_loss": 0.019635630771517754, "global_step": 226269, "epoch": 2542, "lr": 6.705011656976745e-06}
{"train_loss": 0.05056107044219971, "global_step": 226270, "epoch": 2542, "lr": 6.704721665465697e-06}
{"train_loss": 0.07804492861032486, "global_step": 226271, "epoch": 2542, "lr": 6.704431679775158e-06}
{"train_loss": 0.055687759071588516, "global_step": 226272, "epoch": 2542, "lr": 6.704141699905192e-06}
{"train_loss": 0.02955879457294941, "global_step": 226273, "epoch": 2542, "lr": 6.7038517258558146e-06}
{"train_loss": 0.021973472088575363, "global_step": 226274, "epoch": 2542, "lr": 6.70356175762708e-06}
{"train_loss": 0.021237948909401894, "global_step": 226275, "epoch": 2542, "lr": 6.703271795219035e-06}
{"train_loss": 0.06756842881441116, "global_step": 226276, "epoch": 2542, "lr": 6.702981838631689e-06}
{"train_loss": 0.027644911780953407, "global_step": 226277, "epoch": 2542, "lr": 6.702691887865115e-06}
{"train_loss": 0.051926881074905396, "global_step": 226278, "epoch": 2542, "lr": 6.702401942919329e-06}
{"train_loss": 0.010974538512527943, "global_step": 226279, "epoch": 2542, "lr": 6.702112003794387e-06}
{"train_loss": 0.010231594555079937, "global_step": 226280, "epoch": 2542, "lr": 6.701822070490299e-06}
{"train_loss": 0.009898599237203598, "global_step": 226281, "epoch": 2542, "lr": 6.701532143007144e-06}
{"train_loss": 0.05268491804599762, "global_step": 226282, "epoch": 2542, "lr": 6.701242221344922e-06}
{"train_loss": 0.08122529834508896, "global_step": 226283, "epoch": 2542, "lr": 6.70095230550371e-06}
{"train_loss": 0.040890924632549286, "global_step": 226284, "epoch": 2542, "lr": 6.700662395483509e-06}
{"train_loss": 0.058379776775836945, "global_step": 226285, "epoch": 2542, "lr": 6.70037249128439e-06}
{"train_loss": 0.04734901711344719, "global_step": 226286, "epoch": 2542, "lr": 6.700082592906365e-06}
{"train_loss": 0.0374859981238842, "global_step": 226287, "epoch": 2542, "lr": 6.6997927003495e-06}
{"train_loss": 0.039542511105537415, "global_step": 226288, "epoch": 2542, "lr": 6.699502813613806e-06}
{"train_loss": 0.055755216628313065, "global_step": 226289, "epoch": 2542, "lr": 6.69921293269935e-06}
{"train_loss": 0.09270740300416946, "global_step": 226290, "epoch": 2542, "lr": 6.698923057606143e-06}
{"train_loss": 0.0541493222117424, "global_step": 226291, "epoch": 2542, "lr": 6.698633188334241e-06}
{"train_loss": 0.04845988377928734, "global_step": 226292, "epoch": 2542, "lr": 6.698343324883693e-06}
{"train_loss": 0.03974667191505432, "global_step": 226293, "epoch": 2542, "lr": 6.6980534672545115e-06}
{"train_loss": 0.052771590650081635, "global_step": 226294, "epoch": 2542, "lr": 6.697763615446756e-06}
{"train_loss": 0.03298622742295265, "global_step": 226295, "epoch": 2542, "lr": 6.6974737694604605e-06}
{"train_loss": 0.04355330020189285, "global_step": 226296, "epoch": 2542, "lr": 6.697183929295653e-06}
{"train_loss": 0.06833522766828537, "global_step": 226297, "epoch": 2542, "lr": 6.696894094952394e-06}
{"train_loss": 0.03890109434723854, "global_step": 226298, "epoch": 2542, "lr": 6.696604266430695e-06}
{"train_loss": 0.05547166243195534, "global_step": 226299, "epoch": 2542, "lr": 6.696314443730622e-06}
{"train_loss": 0.03668856993317604, "global_step": 226300, "epoch": 2542, "lr": 6.696024626852204e-06}
{"train_loss": 0.051997292786836624, "global_step": 226301, "epoch": 2542, "lr": 6.695734815795462e-06}
{"train_loss": 0.049707286059856415, "global_step": 226302, "epoch": 2542, "lr": 6.695445010560464e-06}
{"train_loss": 0.0352964773774147, "global_step": 226303, "epoch": 2542, "lr": 6.695155211147225e-06}
{"train_loss": 0.051174674183130264, "global_step": 226304, "epoch": 2542, "lr": 6.694865417555806e-06}
{"train_loss": 0.08275002241134644, "global_step": 226305, "epoch": 2542, "lr": 6.694575629786221e-06}
{"train_loss": 0.04595603421330452, "global_step": 226306, "epoch": 2542, "lr": 6.6942858478385385e-06}
{"train_loss": 0.03031325899064541, "global_step": 226307, "epoch": 2542, "lr": 6.693996071712766e-06}
{"train_loss": 0.03231082856655121, "global_step": 226308, "epoch": 2542, "lr": 6.693706301408969e-06}
{"train_loss": 0.04904340207576752, "global_step": 226309, "epoch": 2542, "lr": 6.6934165369271665e-06}
{"train_loss": 0.04187697544693947, "global_step": 226310, "epoch": 2542, "lr": 6.693126778267406e-06}
{"train_loss": 0.047818057239055634, "global_step": 226311, "epoch": 2542, "lr": 6.6928370254297376e-06}
{"train_loss": 0.060909003019332886, "global_step": 226312, "epoch": 2542, "lr": 6.692547278414179e-06}
{"train_loss": 0.05299960821866989, "global_step": 226313, "epoch": 2542, "lr": 6.692257537220787e-06}
{"train_loss": 0.05625765770673752, "global_step": 226314, "epoch": 2542, "lr": 6.691967801849586e-06}
{"train_loss": 0.0774398073554039, "global_step": 226315, "epoch": 2542, "lr": 6.691678072300633e-06}
{"train_loss": 0.044298671185970306, "global_step": 226316, "epoch": 2542, "lr": 6.691388348573941e-06}
{"train_loss": 0.09359993785619736, "global_step": 226317, "epoch": 2542, "lr": 6.691098630669579e-06}
{"train_loss": 0.058422647416591644, "global_step": 226318, "epoch": 2542, "lr": 6.6908089185875596e-06}
{"train_loss": 0.10518060624599457, "global_step": 226319, "epoch": 2542, "lr": 6.690519212327939e-06}
{"train_loss": 0.03801274299621582, "global_step": 226320, "epoch": 2542, "lr": 6.690229511890744e-06}
{"train_loss": 0.04513266310095787, "global_step": 226321, "epoch": 2542, "lr": 6.689939817276031e-06}
{"train_loss": 0.08096884191036224, "global_step": 226322, "epoch": 2542, "lr": 6.689650128483815e-06}
{"train_loss": 0.0565505288541317, "global_step": 226323, "epoch": 2542, "lr": 6.6893604455141586e-06}
{"train_loss": 0.059229400008916855, "global_step": 226324, "epoch": 2542, "lr": 6.689070768367084e-06}
{"train_loss": 0.038886770606040955, "global_step": 226325, "epoch": 2542, "lr": 6.68878109704264e-06}
{"train_loss": 0.05111847516525997, "global_step": 226326, "epoch": 2542, "lr": 6.6884914315408555e-06, "val_loss": 9.124563217163086}
{"train_loss": 0.07903790473937988, "global_step": 226327, "epoch": 2543, "lr": 6.688201771861774e-06}
{"train_loss": 0.04194357618689537, "global_step": 226328, "epoch": 2543, "lr": 6.687912118005446e-06}
{"train_loss": 0.055536095052957535, "global_step": 226329, "epoch": 2543, "lr": 6.687622469971888e-06}
{"train_loss": 0.02862510085105896, "global_step": 226330, "epoch": 2543, "lr": 6.687332827761162e-06}
{"train_loss": 0.06552917510271072, "global_step": 226331, "epoch": 2543, "lr": 6.687043191373288e-06}
{"train_loss": 0.08133158087730408, "global_step": 226332, "epoch": 2543, "lr": 6.686753560808323e-06}
{"train_loss": 0.07029255479574203, "global_step": 226333, "epoch": 2543, "lr": 6.6864639360662955e-06}
{"train_loss": 0.05713003873825073, "global_step": 226334, "epoch": 2543, "lr": 6.686174317147237e-06}
{"train_loss": 0.02470272406935692, "global_step": 226335, "epoch": 2543, "lr": 6.685884704051198e-06}
{"train_loss": 0.06372316181659698, "global_step": 226336, "epoch": 2543, "lr": 6.685595096778208e-06}
{"train_loss": 0.05572604760527611, "global_step": 226337, "epoch": 2543, "lr": 6.6853054953283255e-06}
{"train_loss": 0.026755932718515396, "global_step": 226338, "epoch": 2543, "lr": 6.685015899701574e-06}
{"train_loss": 0.02536848746240139, "global_step": 226339, "epoch": 2543, "lr": 6.684726309897982e-06}
{"train_loss": 0.036833636462688446, "global_step": 226340, "epoch": 2543, "lr": 6.684436725917609e-06}
{"train_loss": 0.05446586012840271, "global_step": 226341, "epoch": 2543, "lr": 6.684147147760478e-06}
{"train_loss": 0.06111564487218857, "global_step": 226342, "epoch": 2543, "lr": 6.683857575426644e-06}
{"train_loss": 0.07610388100147247, "global_step": 226343, "epoch": 2543, "lr": 6.68356800891613e-06}
{"train_loss": 0.0641014501452446, "global_step": 226344, "epoch": 2543, "lr": 6.683278448228991e-06}
{"train_loss": 0.06161714345216751, "global_step": 226345, "epoch": 2543, "lr": 6.682988893365244e-06}
{"train_loss": 0.045956019312143326, "global_step": 226346, "epoch": 2543, "lr": 6.6826993443249444e-06}
{"train_loss": 0.05979170650243759, "global_step": 226347, "epoch": 2543, "lr": 6.682409801108142e-06}
{"train_loss": 0.08867602795362473, "global_step": 226348, "epoch": 2543, "lr": 6.682120263714842e-06}
{"train_loss": 0.08015773445367813, "global_step": 226349, "epoch": 2543, "lr": 6.681830732145122e-06}
{"train_loss": 0.044102415442466736, "global_step": 226350, "epoch": 2543, "lr": 6.681541206398989e-06}
{"train_loss": 0.02056587114930153, "global_step": 226351, "epoch": 2543, "lr": 6.681251686476503e-06}
{"train_loss": 0.09937474876642227, "global_step": 226352, "epoch": 2543, "lr": 6.680962172377686e-06}
{"train_loss": 0.062422048300504684, "global_step": 226353, "epoch": 2543, "lr": 6.680672664102594e-06}
{"train_loss": 0.040054623037576675, "global_step": 226354, "epoch": 2543, "lr": 6.680383161651249e-06}
{"train_loss": 0.032597679644823074, "global_step": 226355, "epoch": 2543, "lr": 6.680093665023707e-06}
{"train_loss": 0.060452111065387726, "global_step": 226356, "epoch": 2543, "lr": 6.679804174219989e-06}
{"train_loss": 0.03651085123419762, "global_step": 226357, "epoch": 2543, "lr": 6.679514689240157e-06}
{"train_loss": 0.03653751313686371, "global_step": 226358, "epoch": 2543, "lr": 6.679225210084222e-06}
{"train_loss": 0.031063197180628777, "global_step": 226359, "epoch": 2543, "lr": 6.6789357367522505e-06}
{"train_loss": 0.035326968878507614, "global_step": 226360, "epoch": 2543, "lr": 6.6786462692442546e-06}
{"train_loss": 0.062059298157691956, "global_step": 226361, "epoch": 2543, "lr": 6.678356807560299e-06}
{"train_loss": 0.01846127212047577, "global_step": 226362, "epoch": 2543, "lr": 6.678067351700401e-06}
{"train_loss": 0.053089678287506104, "global_step": 226363, "epoch": 2543, "lr": 6.677777901664617e-06}
{"train_loss": 0.05487115681171417, "global_step": 226364, "epoch": 2543, "lr": 6.677488457452969e-06}
{"train_loss": 0.042467374354600906, "global_step": 226365, "epoch": 2543, "lr": 6.6771990190654995e-06}
{"train_loss": 0.08182810246944427, "global_step": 226366, "epoch": 2543, "lr": 6.676909586502272e-06}
{"train_loss": 0.05985560268163681, "global_step": 226367, "epoch": 2543, "lr": 6.676620159763286e-06}
{"train_loss": 0.08149157464504242, "global_step": 226368, "epoch": 2543, "lr": 6.676330738848619e-06}
{"train_loss": 0.04360760748386383, "global_step": 226369, "epoch": 2543, "lr": 6.676041323758275e-06}
{"train_loss": 0.05346711724996567, "global_step": 226370, "epoch": 2543, "lr": 6.675751914492323e-06}
{"train_loss": 0.03564011678099632, "global_step": 226371, "epoch": 2543, "lr": 6.67546251105079e-06}
{"train_loss": 0.03479072451591492, "global_step": 226372, "epoch": 2543, "lr": 6.675173113433697e-06}
{"train_loss": 0.014415941201150417, "global_step": 226373, "epoch": 2543, "lr": 6.6748837216411116e-06}
{"train_loss": 0.05881480500102043, "global_step": 226374, "epoch": 2543, "lr": 6.6745943356730514e-06}
{"train_loss": 0.02005409449338913, "global_step": 226375, "epoch": 2543, "lr": 6.67430495552957e-06}
{"train_loss": 0.04126341640949249, "global_step": 226376, "epoch": 2543, "lr": 6.6740155812106965e-06}
{"train_loss": 0.039691708981990814, "global_step": 226377, "epoch": 2543, "lr": 6.67372621271648e-06}
{"train_loss": 0.04177199676632881, "global_step": 226378, "epoch": 2543, "lr": 6.673436850046954e-06}
{"train_loss": 0.047075845301151276, "global_step": 226379, "epoch": 2543, "lr": 6.673147493202142e-06}
{"train_loss": 0.030916068702936172, "global_step": 226380, "epoch": 2543, "lr": 6.672858142182115e-06}
{"train_loss": 0.057216230779886246, "global_step": 226381, "epoch": 2543, "lr": 6.672568796986877e-06}
{"train_loss": 0.05408863723278046, "global_step": 226382, "epoch": 2543, "lr": 6.6722794576164875e-06}
{"train_loss": 0.04383016750216484, "global_step": 226383, "epoch": 2543, "lr": 6.671990124070987e-06}
{"train_loss": 0.04629147797822952, "global_step": 226384, "epoch": 2543, "lr": 6.671700796350406e-06}
{"train_loss": 0.03459784761071205, "global_step": 226385, "epoch": 2543, "lr": 6.6714114744547916e-06}
{"train_loss": 0.07414989173412323, "global_step": 226386, "epoch": 2543, "lr": 6.671122158384174e-06}
{"train_loss": 0.02929919958114624, "global_step": 226387, "epoch": 2543, "lr": 6.670832848138603e-06}
{"train_loss": 0.04546968266367912, "global_step": 226388, "epoch": 2543, "lr": 6.670543543718099e-06}
{"train_loss": 0.04133225232362747, "global_step": 226389, "epoch": 2543, "lr": 6.67025424512272e-06}
{"train_loss": 0.04604276269674301, "global_step": 226390, "epoch": 2543, "lr": 6.669964952352492e-06}
{"train_loss": 0.11194667965173721, "global_step": 226391, "epoch": 2543, "lr": 6.669675665407465e-06}
{"train_loss": 0.09011002629995346, "global_step": 226392, "epoch": 2543, "lr": 6.669386384287668e-06}
{"train_loss": 0.05490068718791008, "global_step": 226393, "epoch": 2543, "lr": 6.66909710899315e-06}
{"train_loss": 0.05217690020799637, "global_step": 226394, "epoch": 2543, "lr": 6.668807839523933e-06}
{"train_loss": 0.04721078276634216, "global_step": 226395, "epoch": 2543, "lr": 6.668518575880078e-06}
{"train_loss": 0.12772485613822937, "global_step": 226396, "epoch": 2543, "lr": 6.668229318061603e-06}
{"train_loss": 0.08563745766878128, "global_step": 226397, "epoch": 2543, "lr": 6.667940066068562e-06}
{"train_loss": 0.04595299810171127, "global_step": 226398, "epoch": 2543, "lr": 6.667650819900983e-06}
{"train_loss": 0.018179690465331078, "global_step": 226399, "epoch": 2543, "lr": 6.667361579558923e-06}
{"train_loss": 0.04491930827498436, "global_step": 226400, "epoch": 2543, "lr": 6.6670723450423915e-06}
{"train_loss": 0.04303717985749245, "global_step": 226401, "epoch": 2543, "lr": 6.666783116351444e-06}
{"train_loss": 0.03169575706124306, "global_step": 226402, "epoch": 2543, "lr": 6.666493893486136e-06}
{"train_loss": 0.039334412664175034, "global_step": 226403, "epoch": 2543, "lr": 6.666204676446475e-06}
{"train_loss": 0.06293565034866333, "global_step": 226404, "epoch": 2543, "lr": 6.6659154652325254e-06}
{"train_loss": 0.02983773499727249, "global_step": 226405, "epoch": 2543, "lr": 6.6656262598443045e-06}
{"train_loss": 0.04675045982003212, "global_step": 226406, "epoch": 2543, "lr": 6.665337060281874e-06}
{"train_loss": 0.06483781337738037, "global_step": 226407, "epoch": 2543, "lr": 6.665047866545249e-06}
{"train_loss": 0.026968996971845627, "global_step": 226408, "epoch": 2543, "lr": 6.6647586786344916e-06}
{"train_loss": 0.04109097272157669, "global_step": 226409, "epoch": 2543, "lr": 6.664469496549613e-06}
{"train_loss": 0.018529905006289482, "global_step": 226410, "epoch": 2543, "lr": 6.664180320290686e-06}
{"train_loss": 0.023838482797145844, "global_step": 226411, "epoch": 2543, "lr": 6.663891149857732e-06}
{"train_loss": 0.04259367287158966, "global_step": 226412, "epoch": 2543, "lr": 6.6636019852507784e-06}
{"train_loss": 0.04889722913503647, "global_step": 226413, "epoch": 2543, "lr": 6.6633128264698805e-06}
{"train_loss": 0.05775441601872444, "global_step": 226414, "epoch": 2543, "lr": 6.663023673515067e-06}
{"train_loss": 0.05015065803537878, "global_step": 226415, "epoch": 2543, "lr": 6.662734526386388e-06, "val_loss": 8.979647636413574}
{"train_loss": 0.051647622138261795, "global_step": 226416, "epoch": 2544, "lr": 6.662445385083882e-06}
{"train_loss": 0.04576689749956131, "global_step": 226417, "epoch": 2544, "lr": 6.662156249607565e-06}
{"train_loss": 0.028888637199997902, "global_step": 226418, "epoch": 2544, "lr": 6.661867119957499e-06}
{"train_loss": 0.08697383850812912, "global_step": 226419, "epoch": 2544, "lr": 6.661577996133722e-06}
{"train_loss": 0.023348191753029823, "global_step": 226420, "epoch": 2544, "lr": 6.661288878136257e-06}
{"train_loss": 0.04775376245379448, "global_step": 226421, "epoch": 2544, "lr": 6.6609997659651705e-06}
{"train_loss": 0.031573519110679626, "global_step": 226422, "epoch": 2544, "lr": 6.660710659620467e-06}
{"train_loss": 0.05911978706717491, "global_step": 226423, "epoch": 2544, "lr": 6.660421559102215e-06}
{"train_loss": 0.051707640290260315, "global_step": 226424, "epoch": 2544, "lr": 6.660132464410435e-06}
{"train_loss": 0.03127967193722725, "global_step": 226425, "epoch": 2544, "lr": 6.659843375545177e-06}
{"train_loss": 0.056204795837402344, "global_step": 226426, "epoch": 2544, "lr": 6.659554292506465e-06}
{"train_loss": 0.019021499902009964, "global_step": 226427, "epoch": 2544, "lr": 6.659265215294358e-06}
{"train_loss": 0.030916471034288406, "global_step": 226428, "epoch": 2544, "lr": 6.6589761439088796e-06}
{"train_loss": 0.04735875129699707, "global_step": 226429, "epoch": 2544, "lr": 6.658687078350079e-06}
{"train_loss": 0.07239148765802383, "global_step": 226430, "epoch": 2544, "lr": 6.658398018617978e-06}
{"train_loss": 0.056710001081228256, "global_step": 226431, "epoch": 2544, "lr": 6.658108964712639e-06}
{"train_loss": 0.049500446766614914, "global_step": 226432, "epoch": 2544, "lr": 6.6578199166340785e-06}
{"train_loss": 0.0556042343378067, "global_step": 226433, "epoch": 2544, "lr": 6.657530874382362e-06}
{"train_loss": 0.03641635924577713, "global_step": 226434, "epoch": 2544, "lr": 6.657241837957495e-06}
{"train_loss": 0.041756678372621536, "global_step": 226435, "epoch": 2544, "lr": 6.6569528073595456e-06}
{"train_loss": 0.05707647278904915, "global_step": 226436, "epoch": 2544, "lr": 6.656663782588529e-06}
{"train_loss": 0.05216103792190552, "global_step": 226437, "epoch": 2544, "lr": 6.656374763644496e-06}
{"train_loss": 0.07168763130903244, "global_step": 226438, "epoch": 2544, "lr": 6.656085750527496e-06}
{"train_loss": 0.061758335679769516, "global_step": 226439, "epoch": 2544, "lr": 6.655796743237547e-06}
{"train_loss": 0.05889301002025604, "global_step": 226440, "epoch": 2544, "lr": 6.655507741774708e-06}
{"train_loss": 0.046391818672418594, "global_step": 226441, "epoch": 2544, "lr": 6.655218746138997e-06}
{"train_loss": 0.07514353841543198, "global_step": 226442, "epoch": 2544, "lr": 6.6549297563304804e-06}
{"train_loss": 0.05645197257399559, "global_step": 226443, "epoch": 2544, "lr": 6.654640772349163e-06}
{"train_loss": 0.044544897973537445, "global_step": 226444, "epoch": 2544, "lr": 6.654351794195113e-06}
{"train_loss": 0.08399469405412674, "global_step": 226445, "epoch": 2544, "lr": 6.654062821868345e-06}
{"train_loss": 0.04254075512290001, "global_step": 226446, "epoch": 2544, "lr": 6.653773855368922e-06}
{"train_loss": 0.03128288313746452, "global_step": 226447, "epoch": 2544, "lr": 6.653484894696854e-06}
{"train_loss": 0.03417101874947548, "global_step": 226448, "epoch": 2544, "lr": 6.653195939852213e-06}
{"train_loss": 0.06386549770832062, "global_step": 226449, "epoch": 2544, "lr": 6.6529069908350225e-06}
{"train_loss": 0.026633914560079575, "global_step": 226450, "epoch": 2544, "lr": 6.652618047645309e-06}
{"train_loss": 0.08363424241542816, "global_step": 226451, "epoch": 2544, "lr": 6.652329110283134e-06}
{"train_loss": 0.11118941009044647, "global_step": 226452, "epoch": 2544, "lr": 6.652040178748514e-06}
{"train_loss": 0.034980013966560364, "global_step": 226453, "epoch": 2544, "lr": 6.65175125304151e-06}
{"train_loss": 0.04835398495197296, "global_step": 226454, "epoch": 2544, "lr": 6.651462333162134e-06}
{"train_loss": 0.04392024874687195, "global_step": 226455, "epoch": 2544, "lr": 6.651173419110457e-06}
{"train_loss": 0.0618634894490242, "global_step": 226456, "epoch": 2544, "lr": 6.65088451088649e-06}
{"train_loss": 0.05591471493244171, "global_step": 226457, "epoch": 2544, "lr": 6.65059560849029e-06}
{"train_loss": 0.07092592865228653, "global_step": 226458, "epoch": 2544, "lr": 6.6503067119218835e-06}
{"train_loss": 0.030793799087405205, "global_step": 226459, "epoch": 2544, "lr": 6.65001782118132e-06}
{"train_loss": 0.030239369720220566, "global_step": 226460, "epoch": 2544, "lr": 6.649728936268629e-06}
{"train_loss": 0.031222227960824966, "global_step": 226461, "epoch": 2544, "lr": 6.649440057183859e-06}
{"train_loss": 0.06450151652097702, "global_step": 226462, "epoch": 2544, "lr": 6.649151183927033e-06}
{"train_loss": 0.049639925360679626, "global_step": 226463, "epoch": 2544, "lr": 6.648862316498211e-06}
{"train_loss": 0.05599639564752579, "global_step": 226464, "epoch": 2544, "lr": 6.648573454897417e-06}
{"train_loss": 0.040467798709869385, "global_step": 226465, "epoch": 2544, "lr": 6.648284599124699e-06}
{"train_loss": 0.045737870037555695, "global_step": 226466, "epoch": 2544, "lr": 6.64799574918008e-06}
{"train_loss": 0.030971521511673927, "global_step": 226467, "epoch": 2544, "lr": 6.647706905063622e-06}
{"train_loss": 0.06033729761838913, "global_step": 226468, "epoch": 2544, "lr": 6.647418066775335e-06}
{"train_loss": 0.06130319461226463, "global_step": 226469, "epoch": 2544, "lr": 6.647129234315291e-06}
{"train_loss": 0.04854794219136238, "global_step": 226470, "epoch": 2544, "lr": 6.646840407683502e-06}
{"train_loss": 0.04335235804319382, "global_step": 226471, "epoch": 2544, "lr": 6.6465515868800236e-06}
{"train_loss": 0.056514106690883636, "global_step": 226472, "epoch": 2544, "lr": 6.646262771904882e-06}
{"train_loss": 0.03645249083638191, "global_step": 226473, "epoch": 2544, "lr": 6.645973962758117e-06}
{"train_loss": 0.03448063135147095, "global_step": 226474, "epoch": 2544, "lr": 6.645685159439785e-06}
{"train_loss": 0.03456762060523033, "global_step": 226475, "epoch": 2544, "lr": 6.6453963619499015e-06}
{"train_loss": 0.054749369621276855, "global_step": 226476, "epoch": 2544, "lr": 6.645107570288528e-06}
{"train_loss": 0.049741797149181366, "global_step": 226477, "epoch": 2544, "lr": 6.644818784455681e-06}
{"train_loss": 0.049318864941596985, "global_step": 226478, "epoch": 2544, "lr": 6.644530004451416e-06}
{"train_loss": 0.049333468079566956, "global_step": 226479, "epoch": 2544, "lr": 6.644241230275761e-06}
{"train_loss": 0.07416065037250519, "global_step": 226480, "epoch": 2544, "lr": 6.643952461928765e-06}
{"train_loss": 0.09436079114675522, "global_step": 226481, "epoch": 2544, "lr": 6.643663699410452e-06}
{"train_loss": 0.054366763681173325, "global_step": 226482, "epoch": 2544, "lr": 6.6433749427208815e-06}
{"train_loss": 0.061592090874910355, "global_step": 226483, "epoch": 2544, "lr": 6.643086191860071e-06}
{"train_loss": 0.0358290933072567, "global_step": 226484, "epoch": 2544, "lr": 6.642797446828075e-06}
{"train_loss": 0.04903694987297058, "global_step": 226485, "epoch": 2544, "lr": 6.642508707624917e-06}
{"train_loss": 0.02027539536356926, "global_step": 226486, "epoch": 2544, "lr": 6.642219974250663e-06}
{"train_loss": 0.08882930129766464, "global_step": 226487, "epoch": 2544, "lr": 6.641931246705318e-06}
{"train_loss": 0.10445605218410492, "global_step": 226488, "epoch": 2544, "lr": 6.641642524988945e-06}
{"train_loss": 0.04681333154439926, "global_step": 226489, "epoch": 2544, "lr": 6.6413538091015805e-06}
{"train_loss": 0.054152000695466995, "global_step": 226490, "epoch": 2544, "lr": 6.641065099043242e-06}
{"train_loss": 0.06049329787492752, "global_step": 226491, "epoch": 2544, "lr": 6.640776394813997e-06}
{"train_loss": 0.07263881713151932, "global_step": 226492, "epoch": 2544, "lr": 6.640487696413861e-06}
{"train_loss": 0.057583630084991455, "global_step": 226493, "epoch": 2544, "lr": 6.6401990038428895e-06}
{"train_loss": 0.058563753962516785, "global_step": 226494, "epoch": 2544, "lr": 6.639910317101111e-06}
{"train_loss": 0.017556725069880486, "global_step": 226495, "epoch": 2544, "lr": 6.639621636188576e-06}
{"train_loss": 0.055060070008039474, "global_step": 226496, "epoch": 2544, "lr": 6.639332961105299e-06}
{"train_loss": 0.06936188787221909, "global_step": 226497, "epoch": 2544, "lr": 6.639044291851354e-06}
{"train_loss": 0.07478578388690948, "global_step": 226498, "epoch": 2544, "lr": 6.638755628426746e-06}
{"train_loss": 0.06824901700019836, "global_step": 226499, "epoch": 2544, "lr": 6.638466970831542e-06}
{"train_loss": 0.06334074586629868, "global_step": 226500, "epoch": 2544, "lr": 6.638178319065752e-06}
{"train_loss": 0.04936264827847481, "global_step": 226501, "epoch": 2544, "lr": 6.637889673129444e-06}
{"train_loss": 0.05927879363298416, "global_step": 226502, "epoch": 2544, "lr": 6.637601033022633e-06}
{"train_loss": 0.052832335233688354, "global_step": 226503, "epoch": 2544, "lr": 6.6373123987453825e-06}
{"train_loss": 0.052713513060399654, "global_step": 226504, "epoch": 2544, "lr": 6.637023770297701e-06, "val_loss": 9.03674602508545}
{"train_loss": 0.045191362500190735, "global_step": 226505, "epoch": 2545, "lr": 6.636735147679657e-06}
{"train_loss": 0.045763690024614334, "global_step": 226506, "epoch": 2545, "lr": 6.6364465308912605e-06}
{"train_loss": 0.05028358846902847, "global_step": 226507, "epoch": 2545, "lr": 6.636157919932579e-06}
{"train_loss": 0.09389478713274002, "global_step": 226508, "epoch": 2545, "lr": 6.635869314803622e-06}
{"train_loss": 0.04325449839234352, "global_step": 226509, "epoch": 2545, "lr": 6.635580715504452e-06}
{"train_loss": 0.04069512337446213, "global_step": 226510, "epoch": 2545, "lr": 6.6352921220351026e-06}
{"train_loss": 0.030311180278658867, "global_step": 226511, "epoch": 2545, "lr": 6.635003534395606e-06}
{"train_loss": 0.041299644857645035, "global_step": 226512, "epoch": 2545, "lr": 6.634714952586013e-06}
{"train_loss": 0.025913793593645096, "global_step": 226513, "epoch": 2545, "lr": 6.6344263766063396e-06}
{"train_loss": 0.016645222902297974, "global_step": 226514, "epoch": 2545, "lr": 6.634137806456653e-06}
{"train_loss": 0.050274815410375595, "global_step": 226515, "epoch": 2545, "lr": 6.633849242136969e-06}
{"train_loss": 0.029163572937250137, "global_step": 226516, "epoch": 2545, "lr": 6.633560683647344e-06}
{"train_loss": 0.028042331337928772, "global_step": 226517, "epoch": 2545, "lr": 6.6332721309877946e-06}
{"train_loss": 0.07261363416910172, "global_step": 226518, "epoch": 2545, "lr": 6.632983584158386e-06}
{"train_loss": 0.05062512308359146, "global_step": 226519, "epoch": 2545, "lr": 6.632695043159137e-06}
{"train_loss": 0.0694405809044838, "global_step": 226520, "epoch": 2545, "lr": 6.632406507990096e-06}
{"train_loss": 0.07525354623794556, "global_step": 226521, "epoch": 2545, "lr": 6.632117978651297e-06}
{"train_loss": 0.04982870817184448, "global_step": 226522, "epoch": 2545, "lr": 6.631829455142791e-06}
{"train_loss": 0.08457348495721817, "global_step": 226523, "epoch": 2545, "lr": 6.631540937464592e-06}
{"train_loss": 0.025030873715877533, "global_step": 226524, "epoch": 2545, "lr": 6.631252425616769e-06}
{"train_loss": 0.042290862649679184, "global_step": 226525, "epoch": 2545, "lr": 6.6309639195993315e-06}
{"train_loss": 0.07296626269817352, "global_step": 226526, "epoch": 2545, "lr": 6.630675419412341e-06}
{"train_loss": 0.016160327941179276, "global_step": 226527, "epoch": 2545, "lr": 6.630386925055831e-06}
{"train_loss": 0.05393190681934357, "global_step": 226528, "epoch": 2545, "lr": 6.630098436529824e-06}
{"train_loss": 0.0388064980506897, "global_step": 226529, "epoch": 2545, "lr": 6.629809953834387e-06}
{"train_loss": 0.0405065156519413, "global_step": 226530, "epoch": 2545, "lr": 6.62952147696953e-06}
{"train_loss": 0.06252460926771164, "global_step": 226531, "epoch": 2545, "lr": 6.629233005935315e-06}
{"train_loss": 0.062066201120615005, "global_step": 226532, "epoch": 2545, "lr": 6.628944540731763e-06}
{"train_loss": 0.04396285489201546, "global_step": 226533, "epoch": 2545, "lr": 6.62865608135893e-06}
{"train_loss": 0.042987652122974396, "global_step": 226534, "epoch": 2545, "lr": 6.628367627816834e-06}
{"train_loss": 0.042920589447021484, "global_step": 226535, "epoch": 2545, "lr": 6.628079180105534e-06}
{"train_loss": 0.01791568659245968, "global_step": 226536, "epoch": 2545, "lr": 6.627790738225054e-06}
{"train_loss": 0.03999575227499008, "global_step": 226537, "epoch": 2545, "lr": 6.627502302175448e-06}
{"train_loss": 0.08319693803787231, "global_step": 226538, "epoch": 2545, "lr": 6.627213871956734e-06}
{"train_loss": 0.051127877086400986, "global_step": 226539, "epoch": 2545, "lr": 6.626925447568971e-06}
{"train_loss": 0.06518279761075974, "global_step": 226540, "epoch": 2545, "lr": 6.6266370290121835e-06}
{"train_loss": 0.08102374523878098, "global_step": 226541, "epoch": 2545, "lr": 6.6263486162864205e-06}
{"train_loss": 0.056330014020204544, "global_step": 226542, "epoch": 2545, "lr": 6.62606020939171e-06}
{"train_loss": 0.0420706644654274, "global_step": 226543, "epoch": 2545, "lr": 6.625771808328107e-06}
{"train_loss": 0.04107886552810669, "global_step": 226544, "epoch": 2545, "lr": 6.6254834130956345e-06}
{"train_loss": 0.03732836991548538, "global_step": 226545, "epoch": 2545, "lr": 6.625195023694331e-06}
{"train_loss": 0.022590452805161476, "global_step": 226546, "epoch": 2545, "lr": 6.624906640124251e-06}
{"train_loss": 0.030676353722810745, "global_step": 226547, "epoch": 2545, "lr": 6.624618262385418e-06}
{"train_loss": 0.04426148906350136, "global_step": 226548, "epoch": 2545, "lr": 6.624329890477882e-06}
{"train_loss": 0.09929684549570084, "global_step": 226549, "epoch": 2545, "lr": 6.62404152440167e-06}
{"train_loss": 0.035430971533060074, "global_step": 226550, "epoch": 2545, "lr": 6.6237531641568315e-06}
{"train_loss": 0.051272664219141006, "global_step": 226551, "epoch": 2545, "lr": 6.623464809743396e-06}
{"train_loss": 0.03212612867355347, "global_step": 226552, "epoch": 2545, "lr": 6.623176461161418e-06}
{"train_loss": 0.07687431573867798, "global_step": 226553, "epoch": 2545, "lr": 6.6228881184109084e-06}
{"train_loss": 0.052020132541656494, "global_step": 226554, "epoch": 2545, "lr": 6.62259978149194e-06}
{"train_loss": 0.023431453853845596, "global_step": 226555, "epoch": 2545, "lr": 6.6223114504045225e-06}
{"train_loss": 0.07103292644023895, "global_step": 226556, "epoch": 2545, "lr": 6.622023125148713e-06}
{"train_loss": 0.037429507821798325, "global_step": 226557, "epoch": 2545, "lr": 6.621734805724539e-06}
{"train_loss": 0.0604739636182785, "global_step": 226558, "epoch": 2545, "lr": 6.621446492132049e-06}
{"train_loss": 0.04519732668995857, "global_step": 226559, "epoch": 2545, "lr": 6.621158184371267e-06}
{"train_loss": 0.041460663080215454, "global_step": 226560, "epoch": 2545, "lr": 6.620869882442255e-06}
{"train_loss": 0.07108403742313385, "global_step": 226561, "epoch": 2545, "lr": 6.620581586345026e-06}
{"train_loss": 0.041220616549253464, "global_step": 226562, "epoch": 2545, "lr": 6.620293296079644e-06}
{"train_loss": 0.03714749589562416, "global_step": 226563, "epoch": 2545, "lr": 6.620005011646124e-06}
{"train_loss": 0.05296580493450165, "global_step": 226564, "epoch": 2545, "lr": 6.619716733044523e-06}
{"train_loss": 0.10011419653892517, "global_step": 226565, "epoch": 2545, "lr": 6.619428460274873e-06}
{"train_loss": 0.06542031466960907, "global_step": 226566, "epoch": 2545, "lr": 6.619140193337203e-06}
{"train_loss": 0.09642814844846725, "global_step": 226567, "epoch": 2545, "lr": 6.618851932231574e-06}
{"train_loss": 0.046915724873542786, "global_step": 226568, "epoch": 2545, "lr": 6.618563676957995e-06}
{"train_loss": 0.047456659376621246, "global_step": 226569, "epoch": 2545, "lr": 6.618275427516535e-06}
{"train_loss": 0.05713534355163574, "global_step": 226570, "epoch": 2545, "lr": 6.617987183907209e-06}
{"train_loss": 0.017539583146572113, "global_step": 226571, "epoch": 2545, "lr": 6.61769894613008e-06}
{"train_loss": 0.07955925166606903, "global_step": 226572, "epoch": 2545, "lr": 6.617410714185157e-06}
{"train_loss": 0.03831982985138893, "global_step": 226573, "epoch": 2545, "lr": 6.617122488072503e-06}
{"train_loss": 0.06449688225984573, "global_step": 226574, "epoch": 2545, "lr": 6.616834267792143e-06}
{"train_loss": 0.15227676928043365, "global_step": 226575, "epoch": 2545, "lr": 6.61654605334413e-06}
{"train_loss": 0.06887948513031006, "global_step": 226576, "epoch": 2545, "lr": 6.616257844728479e-06}
{"train_loss": 0.060513999313116074, "global_step": 226577, "epoch": 2545, "lr": 6.615969641945258e-06}
{"train_loss": 0.08024349808692932, "global_step": 226578, "epoch": 2545, "lr": 6.615681444994481e-06}
{"train_loss": 0.03947491571307182, "global_step": 226579, "epoch": 2545, "lr": 6.615393253876207e-06}
{"train_loss": 0.05355150252580643, "global_step": 226580, "epoch": 2545, "lr": 6.6151050685904495e-06}
{"train_loss": 0.039984043687582016, "global_step": 226581, "epoch": 2545, "lr": 6.6148168891372654e-06}
{"train_loss": 0.1299474686384201, "global_step": 226582, "epoch": 2545, "lr": 6.6145287155167045e-06}
{"train_loss": 0.058380674570798874, "global_step": 226583, "epoch": 2545, "lr": 6.614240547728779e-06}
{"train_loss": 0.04812055826187134, "global_step": 226584, "epoch": 2545, "lr": 6.613952385773548e-06}
{"train_loss": 0.0966712236404419, "global_step": 226585, "epoch": 2545, "lr": 6.613664229651029e-06}
{"train_loss": 0.030458616092801094, "global_step": 226586, "epoch": 2545, "lr": 6.613376079361289e-06}
{"train_loss": 0.05874910205602646, "global_step": 226587, "epoch": 2545, "lr": 6.613087934904344e-06}
{"train_loss": 0.06257075071334839, "global_step": 226588, "epoch": 2545, "lr": 6.612799796280244e-06}
{"train_loss": 0.03301273658871651, "global_step": 226589, "epoch": 2545, "lr": 6.612511663489019e-06}
{"train_loss": 0.06692812591791153, "global_step": 226590, "epoch": 2545, "lr": 6.612223536530726e-06}
{"train_loss": 0.09009965509176254, "global_step": 226591, "epoch": 2545, "lr": 6.6119354154053734e-06}
{"train_loss": 0.0380098819732666, "global_step": 226592, "epoch": 2545, "lr": 6.611647300113033e-06}
{"train_loss": 0.05423992361580388, "global_step": 226593, "epoch": 2545, "lr": 6.611359190653715e-06, "val_loss": 9.235587120056152, "train_action_mse_error": 21.34994125366211}
{"train_loss": 0.01672312244772911, "global_step": 226594, "epoch": 2546, "lr": 6.611071087027482e-06}
{"train_loss": 0.056294266134500504, "global_step": 226595, "epoch": 2546, "lr": 6.610782989234349e-06}
{"train_loss": 0.022278189659118652, "global_step": 226596, "epoch": 2546, "lr": 6.610494897274378e-06}
{"train_loss": 0.10270361602306366, "global_step": 226597, "epoch": 2546, "lr": 6.6102068111475905e-06}
{"train_loss": 0.02315286360681057, "global_step": 226598, "epoch": 2546, "lr": 6.609918730854037e-06}
{"train_loss": 0.0409267395734787, "global_step": 226599, "epoch": 2546, "lr": 6.6096306563937446e-06}
{"train_loss": 0.027217641472816467, "global_step": 226600, "epoch": 2546, "lr": 6.609342587766771e-06}
{"train_loss": 0.06940688192844391, "global_step": 226601, "epoch": 2546, "lr": 6.609054524973124e-06}
{"train_loss": 0.0476214773952961, "global_step": 226602, "epoch": 2546, "lr": 6.608766468012878e-06}
{"train_loss": 0.09129147231578827, "global_step": 226603, "epoch": 2546, "lr": 6.608478416886044e-06}
{"train_loss": 0.05571647733449936, "global_step": 226604, "epoch": 2546, "lr": 6.608190371592682e-06}
{"train_loss": 0.02408928982913494, "global_step": 226605, "epoch": 2546, "lr": 6.607902332132815e-06}
{"train_loss": 0.05931505933403969, "global_step": 226606, "epoch": 2546, "lr": 6.607614298506482e-06}
{"train_loss": 0.06366925686597824, "global_step": 226607, "epoch": 2546, "lr": 6.607326270713732e-06}
{"train_loss": 0.0790858045220375, "global_step": 226608, "epoch": 2546, "lr": 6.607038248754588e-06}
{"train_loss": 0.029195444658398628, "global_step": 226609, "epoch": 2546, "lr": 6.606750232629111e-06}
{"train_loss": 0.0887836217880249, "global_step": 226610, "epoch": 2546, "lr": 6.606462222337317e-06}
{"train_loss": 0.06192151829600334, "global_step": 226611, "epoch": 2546, "lr": 6.606174217879263e-06}
{"train_loss": 0.0368909016251564, "global_step": 226612, "epoch": 2546, "lr": 6.605886219254975e-06}
{"train_loss": 0.06834569573402405, "global_step": 226613, "epoch": 2546, "lr": 6.6055982264645044e-06}
{"train_loss": 0.045218512415885925, "global_step": 226614, "epoch": 2546, "lr": 6.605310239507872e-06}
{"train_loss": 0.03780652582645416, "global_step": 226615, "epoch": 2546, "lr": 6.605022258385135e-06}
{"train_loss": 0.03802378103137016, "global_step": 226616, "epoch": 2546, "lr": 6.604734283096314e-06}
{"train_loss": 0.04729965329170227, "global_step": 226617, "epoch": 2546, "lr": 6.604446313641471e-06}
{"train_loss": 0.033848222345113754, "global_step": 226618, "epoch": 2546, "lr": 6.604158350020617e-06}
{"train_loss": 0.020231500267982483, "global_step": 226619, "epoch": 2546, "lr": 6.603870392233807e-06}
{"train_loss": 0.08797303587198257, "global_step": 226620, "epoch": 2546, "lr": 6.603582440281092e-06}
{"train_loss": 0.0346330888569355, "global_step": 226621, "epoch": 2546, "lr": 6.603294494162477e-06}
{"train_loss": 0.049271147698163986, "global_step": 226622, "epoch": 2546, "lr": 6.603006553878038e-06}
{"train_loss": 0.04860617220401764, "global_step": 226623, "epoch": 2546, "lr": 6.602718619427783e-06}
{"train_loss": 0.04126963019371033, "global_step": 226624, "epoch": 2546, "lr": 6.6024306908117725e-06}
{"train_loss": 0.059278905391693115, "global_step": 226625, "epoch": 2546, "lr": 6.602142768030028e-06}
{"train_loss": 0.017243584617972374, "global_step": 226626, "epoch": 2546, "lr": 6.601854851082606e-06}
{"train_loss": 0.050597503781318665, "global_step": 226627, "epoch": 2546, "lr": 6.6015669399695275e-06}
{"train_loss": 0.012693691998720169, "global_step": 226628, "epoch": 2546, "lr": 6.601279034690849e-06}
{"train_loss": 0.011302730068564415, "global_step": 226629, "epoch": 2546, "lr": 6.600991135246587e-06}
{"train_loss": 0.06935424357652664, "global_step": 226630, "epoch": 2546, "lr": 6.600703241636802e-06}
{"train_loss": 0.056319378316402435, "global_step": 226631, "epoch": 2546, "lr": 6.600415353861517e-06}
{"train_loss": 0.04297435283660889, "global_step": 226632, "epoch": 2546, "lr": 6.600127471920787e-06}
{"train_loss": 0.0420948788523674, "global_step": 226633, "epoch": 2546, "lr": 6.599839595814628e-06}
{"train_loss": 0.03954552859067917, "global_step": 226634, "epoch": 2546, "lr": 6.599551725543102e-06}
{"train_loss": 0.039995379745960236, "global_step": 226635, "epoch": 2546, "lr": 6.59926386110623e-06}
{"train_loss": 0.03910507261753082, "global_step": 226636, "epoch": 2546, "lr": 6.598976002504054e-06}
{"train_loss": 0.024846848100423813, "global_step": 226637, "epoch": 2546, "lr": 6.598688149736632e-06}
{"train_loss": 0.050916362553834915, "global_step": 226638, "epoch": 2546, "lr": 6.598400302803981e-06}
{"train_loss": 0.033502548933029175, "global_step": 226639, "epoch": 2546, "lr": 6.598112461706141e-06}
{"train_loss": 0.04895533248782158, "global_step": 226640, "epoch": 2546, "lr": 6.597824626443167e-06}
{"train_loss": 0.0750860944390297, "global_step": 226641, "epoch": 2546, "lr": 6.597536797015069e-06}
{"train_loss": 0.02456746995449066, "global_step": 226642, "epoch": 2546, "lr": 6.597248973421921e-06}
{"train_loss": 0.026161495596170425, "global_step": 226643, "epoch": 2546, "lr": 6.596961155663739e-06}
{"train_loss": 0.03732307255268097, "global_step": 226644, "epoch": 2546, "lr": 6.596673343740561e-06}
{"train_loss": 0.0369134396314621, "global_step": 226645, "epoch": 2546, "lr": 6.5963855376524386e-06}
{"train_loss": 0.022157983854413033, "global_step": 226646, "epoch": 2546, "lr": 6.596097737399393e-06}
{"train_loss": 0.05107181519269943, "global_step": 226647, "epoch": 2546, "lr": 6.5958099429814845e-06}
{"train_loss": 0.008328679017722607, "global_step": 226648, "epoch": 2546, "lr": 6.595522154398726e-06}
{"train_loss": 0.02453939989209175, "global_step": 226649, "epoch": 2546, "lr": 6.595234371651182e-06}
{"train_loss": 0.059116948395967484, "global_step": 226650, "epoch": 2546, "lr": 6.5949465947388715e-06}
{"train_loss": 0.09448589384555817, "global_step": 226651, "epoch": 2546, "lr": 6.594658823661854e-06}
{"train_loss": 0.042670633643865585, "global_step": 226652, "epoch": 2546, "lr": 6.59437105842014e-06}
{"train_loss": 0.04848388954997063, "global_step": 226653, "epoch": 2546, "lr": 6.594083299013798e-06}
{"train_loss": 0.07776301354169846, "global_step": 226654, "epoch": 2546, "lr": 6.593795545442844e-06}
{"train_loss": 0.03342737630009651, "global_step": 226655, "epoch": 2546, "lr": 6.5935077977073215e-06}
{"train_loss": 0.06165027245879173, "global_step": 226656, "epoch": 2546, "lr": 6.593220055807281e-06}
{"train_loss": 0.0749417170882225, "global_step": 226657, "epoch": 2546, "lr": 6.5929323197427454e-06}
{"train_loss": 0.037880245596170425, "global_step": 226658, "epoch": 2546, "lr": 6.592644589513769e-06}
{"train_loss": 0.043732963502407074, "global_step": 226659, "epoch": 2546, "lr": 6.592356865120375e-06}
{"train_loss": 0.04939524456858635, "global_step": 226660, "epoch": 2546, "lr": 6.592069146562618e-06}
{"train_loss": 0.055826425552368164, "global_step": 226661, "epoch": 2546, "lr": 6.591781433840522e-06}
{"train_loss": 0.04678821936249733, "global_step": 226662, "epoch": 2546, "lr": 6.59149372695414e-06}
{"train_loss": 0.01937885209918022, "global_step": 226663, "epoch": 2546, "lr": 6.59120602590349e-06}
{"train_loss": 0.06018100306391716, "global_step": 226664, "epoch": 2546, "lr": 6.590918330688639e-06}
{"train_loss": 0.033683374524116516, "global_step": 226665, "epoch": 2546, "lr": 6.590630641309592e-06}
{"train_loss": 0.026482392102479935, "global_step": 226666, "epoch": 2546, "lr": 6.590342957766421e-06}
{"train_loss": 0.08055128157138824, "global_step": 226667, "epoch": 2546, "lr": 6.5900552800591375e-06}
{"train_loss": 0.057251326739788055, "global_step": 226668, "epoch": 2546, "lr": 6.589767608187803e-06}
{"train_loss": 0.06909817457199097, "global_step": 226669, "epoch": 2546, "lr": 6.5894799421524325e-06}
{"train_loss": 0.030730586498975754, "global_step": 226670, "epoch": 2546, "lr": 6.589192281953094e-06}
{"train_loss": 0.037799157202243805, "global_step": 226671, "epoch": 2546, "lr": 6.588904627589793e-06}
{"train_loss": 0.031107936054468155, "global_step": 226672, "epoch": 2546, "lr": 6.588616979062584e-06}
{"train_loss": 0.04928576573729515, "global_step": 226673, "epoch": 2546, "lr": 6.588329336371518e-06}
{"train_loss": 0.047540660947561264, "global_step": 226674, "epoch": 2546, "lr": 6.588041699516617e-06}
{"train_loss": 0.06625039130449295, "global_step": 226675, "epoch": 2546, "lr": 6.587754068497931e-06}
{"train_loss": 0.0694180428981781, "global_step": 226676, "epoch": 2546, "lr": 6.587466443315493e-06}
{"train_loss": 0.04034567251801491, "global_step": 226677, "epoch": 2546, "lr": 6.587178823969325e-06}
{"train_loss": 0.06634914875030518, "global_step": 226678, "epoch": 2546, "lr": 6.586891210459501e-06}
{"train_loss": 0.03761516883969307, "global_step": 226679, "epoch": 2546, "lr": 6.586603602786024e-06}
{"train_loss": 0.04257997125387192, "global_step": 226680, "epoch": 2546, "lr": 6.586316000948961e-06}
{"train_loss": 0.048354919999837875, "global_step": 226681, "epoch": 2546, "lr": 6.586028404948341e-06}
{"train_loss": 0.046917624175129985, "global_step": 226682, "epoch": 2546, "lr": 6.585740814784186e-06, "val_loss": 9.147650718688965}
{"train_loss": 0.04559444263577461, "global_step": 226683, "epoch": 2547, "lr": 6.585453230456562e-06}
{"train_loss": 0.02107621170580387, "global_step": 226684, "epoch": 2547, "lr": 6.5851656519654804e-06}
{"train_loss": 0.0794588103890419, "global_step": 226685, "epoch": 2547, "lr": 6.584878079311008e-06}
{"train_loss": 0.04601139947772026, "global_step": 226686, "epoch": 2547, "lr": 6.58459051249316e-06}
{"train_loss": 0.04048758000135422, "global_step": 226687, "epoch": 2547, "lr": 6.584302951511995e-06}
{"train_loss": 0.1002415344119072, "global_step": 226688, "epoch": 2547, "lr": 6.584015396367526e-06}
{"train_loss": 0.05271926149725914, "global_step": 226689, "epoch": 2547, "lr": 6.583727847059823e-06}
{"train_loss": 0.11516963690519333, "global_step": 226690, "epoch": 2547, "lr": 6.583440303588895e-06}
{"train_loss": 0.049569811671972275, "global_step": 226691, "epoch": 2547, "lr": 6.583152765954798e-06}
{"train_loss": 0.060799531638622284, "global_step": 226692, "epoch": 2547, "lr": 6.582865234157576e-06}
{"train_loss": 0.03965243697166443, "global_step": 226693, "epoch": 2547, "lr": 6.582577708197246e-06}
{"train_loss": 0.017281146720051765, "global_step": 226694, "epoch": 2547, "lr": 6.58229018807387e-06}
{"train_loss": 0.015726670622825623, "global_step": 226695, "epoch": 2547, "lr": 6.5820026737874684e-06}
{"train_loss": 0.04228179156780243, "global_step": 226696, "epoch": 2547, "lr": 6.581715165338092e-06}
{"train_loss": 0.04064970090985298, "global_step": 226697, "epoch": 2547, "lr": 6.58142766272577e-06}
{"train_loss": 0.07251633703708649, "global_step": 226698, "epoch": 2547, "lr": 6.581140165950556e-06}
{"train_loss": 0.05949320271611214, "global_step": 226699, "epoch": 2547, "lr": 6.580852675012467e-06}
{"train_loss": 0.04342947155237198, "global_step": 226700, "epoch": 2547, "lr": 6.580565189911558e-06}
{"train_loss": 0.02431267499923706, "global_step": 226701, "epoch": 2547, "lr": 6.580277710647858e-06}
{"train_loss": 0.031230341643095016, "global_step": 226702, "epoch": 2547, "lr": 6.579990237221423e-06}
{"train_loss": 0.013546555303037167, "global_step": 226703, "epoch": 2547, "lr": 6.579702769632262e-06}
{"train_loss": 0.04334045946598053, "global_step": 226704, "epoch": 2547, "lr": 6.579415307880449e-06}
{"train_loss": 0.04681317135691643, "global_step": 226705, "epoch": 2547, "lr": 6.579127851965988e-06}
{"train_loss": 0.034523315727710724, "global_step": 226706, "epoch": 2547, "lr": 6.5788404018889475e-06}
{"train_loss": 0.0473686158657074, "global_step": 226707, "epoch": 2547, "lr": 6.578552957649342e-06}
{"train_loss": 0.045070286840200424, "global_step": 226708, "epoch": 2547, "lr": 6.578265519247218e-06}
{"train_loss": 0.048955079168081284, "global_step": 226709, "epoch": 2547, "lr": 6.577978086682629e-06}
{"train_loss": 0.04632945358753204, "global_step": 226710, "epoch": 2547, "lr": 6.577690659955594e-06}
{"train_loss": 0.05288271605968475, "global_step": 226711, "epoch": 2547, "lr": 6.5774032390661655e-06}
{"train_loss": 0.02819998934864998, "global_step": 226712, "epoch": 2547, "lr": 6.577115824014368e-06}
{"train_loss": 0.09885511547327042, "global_step": 226713, "epoch": 2547, "lr": 6.576828414800257e-06}
{"train_loss": 0.036250900477170944, "global_step": 226714, "epoch": 2547, "lr": 6.576541011423865e-06}
{"train_loss": 0.04365110024809837, "global_step": 226715, "epoch": 2547, "lr": 6.576253613885214e-06}
{"train_loss": 0.0887359231710434, "global_step": 226716, "epoch": 2547, "lr": 6.575966222184365e-06}
{"train_loss": 0.0386577844619751, "global_step": 226717, "epoch": 2547, "lr": 6.575678836321342e-06}
{"train_loss": 0.02854565903544426, "global_step": 226718, "epoch": 2547, "lr": 6.575391456296198e-06}
{"train_loss": 0.07498359680175781, "global_step": 226719, "epoch": 2547, "lr": 6.575104082108957e-06}
{"train_loss": 0.022959114983677864, "global_step": 226720, "epoch": 2547, "lr": 6.574816713759668e-06}
{"train_loss": 0.05152256786823273, "global_step": 226721, "epoch": 2547, "lr": 6.574529351248371e-06}
{"train_loss": 0.02245115488767624, "global_step": 226722, "epoch": 2547, "lr": 6.574241994575086e-06}
{"train_loss": 0.024516914039850235, "global_step": 226723, "epoch": 2547, "lr": 6.573954643739877e-06}
{"train_loss": 0.05056137964129448, "global_step": 226724, "epoch": 2547, "lr": 6.573667298742758e-06}
{"train_loss": 0.07540948688983917, "global_step": 226725, "epoch": 2547, "lr": 6.573379959583792e-06}
{"train_loss": 0.04934261366724968, "global_step": 226726, "epoch": 2547, "lr": 6.573092626262994e-06}
{"train_loss": 0.06838982552289963, "global_step": 226727, "epoch": 2547, "lr": 6.572805298780416e-06}
{"train_loss": 0.04778596758842468, "global_step": 226728, "epoch": 2547, "lr": 6.572517977136106e-06}
{"train_loss": 0.03425094857811928, "global_step": 226729, "epoch": 2547, "lr": 6.572230661330081e-06}
{"train_loss": 0.023987527936697006, "global_step": 226730, "epoch": 2547, "lr": 6.571943351362397e-06}
{"train_loss": 0.09022986143827438, "global_step": 226731, "epoch": 2547, "lr": 6.571656047233077e-06}
{"train_loss": 0.052201371639966965, "global_step": 226732, "epoch": 2547, "lr": 6.57136874894218e-06}
{"train_loss": 0.06640087813138962, "global_step": 226733, "epoch": 2547, "lr": 6.571081456489725e-06}
{"train_loss": 0.03653787076473236, "global_step": 226734, "epoch": 2547, "lr": 6.570794169875766e-06}
{"train_loss": 0.06719064712524414, "global_step": 226735, "epoch": 2547, "lr": 6.570506889100325e-06}
{"train_loss": 0.037366725504398346, "global_step": 226736, "epoch": 2547, "lr": 6.570219614163464e-06}
{"train_loss": 0.051953721791505814, "global_step": 226737, "epoch": 2547, "lr": 6.569932345065194e-06}
{"train_loss": 0.07899511605501175, "global_step": 226738, "epoch": 2547, "lr": 6.569645081805576e-06}
{"train_loss": 0.04131628945469856, "global_step": 226739, "epoch": 2547, "lr": 6.569357824384631e-06}
{"train_loss": 0.06882396340370178, "global_step": 226740, "epoch": 2547, "lr": 6.569070572802416e-06}
{"train_loss": 0.04191277548670769, "global_step": 226741, "epoch": 2547, "lr": 6.5687833270589474e-06}
{"train_loss": 0.059196460992097855, "global_step": 226742, "epoch": 2547, "lr": 6.568496087154291e-06}
{"train_loss": 0.11311226338148117, "global_step": 226743, "epoch": 2547, "lr": 6.568208853088459e-06}
{"train_loss": 0.09277692437171936, "global_step": 226744, "epoch": 2547, "lr": 6.567921624861517e-06}
{"train_loss": 0.018427221104502678, "global_step": 226745, "epoch": 2547, "lr": 6.567634402473472e-06}
{"train_loss": 0.048637229949235916, "global_step": 226746, "epoch": 2547, "lr": 6.567347185924383e-06}
{"train_loss": 0.09326211363077164, "global_step": 226747, "epoch": 2547, "lr": 6.56705997521429e-06}
{"train_loss": 0.055628903210163116, "global_step": 226748, "epoch": 2547, "lr": 6.566772770343222e-06}
{"train_loss": 0.08828315883874893, "global_step": 226749, "epoch": 2547, "lr": 6.566485571311226e-06}
{"train_loss": 0.04880394786596298, "global_step": 226750, "epoch": 2547, "lr": 6.566198378118332e-06}
{"train_loss": 0.062319476157426834, "global_step": 226751, "epoch": 2547, "lr": 6.56591119076459e-06}
{"train_loss": 0.0486028753221035, "global_step": 226752, "epoch": 2547, "lr": 6.565624009250027e-06}
{"train_loss": 0.031233983114361763, "global_step": 226753, "epoch": 2547, "lr": 6.565336833574692e-06}
{"train_loss": 0.056962352246046066, "global_step": 226754, "epoch": 2547, "lr": 6.5650496637386204e-06}
{"train_loss": 0.04816287010908127, "global_step": 226755, "epoch": 2547, "lr": 6.564762499741833e-06}
{"train_loss": 0.02525145746767521, "global_step": 226756, "epoch": 2547, "lr": 6.5644753415844026e-06}
{"train_loss": 0.02044994756579399, "global_step": 226757, "epoch": 2547, "lr": 6.564188189266329e-06}
{"train_loss": 0.015500620938837528, "global_step": 226758, "epoch": 2547, "lr": 6.56390104278769e-06}
{"train_loss": 0.011132135055959225, "global_step": 226759, "epoch": 2547, "lr": 6.563613902148502e-06}
{"train_loss": 0.058104123920202255, "global_step": 226760, "epoch": 2547, "lr": 6.563326767348793e-06}
{"train_loss": 0.04876706749200821, "global_step": 226761, "epoch": 2547, "lr": 6.563039638388629e-06}
{"train_loss": 0.059981659054756165, "global_step": 226762, "epoch": 2547, "lr": 6.562752515268028e-06}
{"train_loss": 0.020566336810588837, "global_step": 226763, "epoch": 2547, "lr": 6.562465397987028e-06}
{"train_loss": 0.09304571896791458, "global_step": 226764, "epoch": 2547, "lr": 6.562178286545689e-06}
{"train_loss": 0.03414683789014816, "global_step": 226765, "epoch": 2547, "lr": 6.5618911809440246e-06}
{"train_loss": 0.031234359368681908, "global_step": 226766, "epoch": 2547, "lr": 6.561604081182099e-06}
{"train_loss": 0.052988603711128235, "global_step": 226767, "epoch": 2547, "lr": 6.561316987259919e-06}
{"train_loss": 0.05577448010444641, "global_step": 226768, "epoch": 2547, "lr": 6.561029899177556e-06}
{"train_loss": 0.03772634640336037, "global_step": 226769, "epoch": 2547, "lr": 6.560742816935023e-06}
{"train_loss": 0.03949963301420212, "global_step": 226770, "epoch": 2547, "lr": 6.560455740532379e-06}
{"train_loss": 0.04972416749537996, "global_step": 226771, "epoch": 2547, "lr": 6.560168669969635e-06, "val_loss": 9.083831787109375}
{"train_loss": 0.05791780352592468, "global_step": 226772, "epoch": 2548, "lr": 6.559881605246865e-06}
{"train_loss": 0.08285925537347794, "global_step": 226773, "epoch": 2548, "lr": 6.559594546364073e-06}
{"train_loss": 0.06157554313540459, "global_step": 226774, "epoch": 2548, "lr": 6.559307493321331e-06}
{"train_loss": 0.06719616055488586, "global_step": 226775, "epoch": 2548, "lr": 6.5590204461186465e-06}
{"train_loss": 0.04133010283112526, "global_step": 226776, "epoch": 2548, "lr": 6.5587334047560845e-06}
{"train_loss": 0.07672499865293503, "global_step": 226777, "epoch": 2548, "lr": 6.558446369233656e-06}
{"train_loss": 0.029021399095654488, "global_step": 226778, "epoch": 2548, "lr": 6.558159339551429e-06}
{"train_loss": 0.04914043843746185, "global_step": 226779, "epoch": 2548, "lr": 6.557872315709418e-06}
{"train_loss": 0.0320267379283905, "global_step": 226780, "epoch": 2548, "lr": 6.557585297707686e-06}
{"train_loss": 0.06975456327199936, "global_step": 226781, "epoch": 2548, "lr": 6.5572982855462375e-06}
{"train_loss": 0.027206288650631905, "global_step": 226782, "epoch": 2548, "lr": 6.557011279225139e-06}
{"train_loss": 0.06184772029519081, "global_step": 226783, "epoch": 2548, "lr": 6.556724278744425e-06}
{"train_loss": 0.0669270008802414, "global_step": 226784, "epoch": 2548, "lr": 6.556437284104122e-06}
{"train_loss": 0.054853443056344986, "global_step": 226785, "epoch": 2548, "lr": 6.556150295304286e-06}
{"train_loss": 0.07100324332714081, "global_step": 226786, "epoch": 2548, "lr": 6.55586331234494e-06}
{"train_loss": 0.04356797784566879, "global_step": 226787, "epoch": 2548, "lr": 6.555576335226132e-06}
{"train_loss": 0.020887600257992744, "global_step": 226788, "epoch": 2548, "lr": 6.555289363947892e-06}
{"train_loss": 0.050800397992134094, "global_step": 226789, "epoch": 2548, "lr": 6.555002398510273e-06}
{"train_loss": 0.07042239606380463, "global_step": 226790, "epoch": 2548, "lr": 6.554715438913295e-06}
{"train_loss": 0.07978835701942444, "global_step": 226791, "epoch": 2548, "lr": 6.5544284851570155e-06}
{"train_loss": 0.04571374133229256, "global_step": 226792, "epoch": 2548, "lr": 6.554141537241465e-06}
{"train_loss": 0.06797776371240616, "global_step": 226793, "epoch": 2548, "lr": 6.55385459516667e-06}
{"train_loss": 0.023057622835040092, "global_step": 226794, "epoch": 2548, "lr": 6.553567658932685e-06}
{"train_loss": 0.08338435739278793, "global_step": 226795, "epoch": 2548, "lr": 6.5532807285395405e-06}
{"train_loss": 0.03511728718876839, "global_step": 226796, "epoch": 2548, "lr": 6.552993803987284e-06}
{"train_loss": 0.03130355104804039, "global_step": 226797, "epoch": 2548, "lr": 6.55270688527595e-06}
{"train_loss": 0.06474894285202026, "global_step": 226798, "epoch": 2548, "lr": 6.55241997240556e-06}
{"train_loss": 0.04847673326730728, "global_step": 226799, "epoch": 2548, "lr": 6.55213306537617e-06}
{"train_loss": 0.04451291263103485, "global_step": 226800, "epoch": 2548, "lr": 6.55184616418783e-06}
{"train_loss": 0.043951041996479034, "global_step": 226801, "epoch": 2548, "lr": 6.551559268840552e-06}
{"train_loss": 0.09851647913455963, "global_step": 226802, "epoch": 2548, "lr": 6.551272379334394e-06}
{"train_loss": 0.06609007716178894, "global_step": 226803, "epoch": 2548, "lr": 6.550985495669381e-06}
{"train_loss": 0.05153002217411995, "global_step": 226804, "epoch": 2548, "lr": 6.550698617845569e-06}
{"train_loss": 0.0533609613776207, "global_step": 226805, "epoch": 2548, "lr": 6.550411745862972e-06}
{"train_loss": 0.04239271208643913, "global_step": 226806, "epoch": 2548, "lr": 6.550124879721659e-06}
{"train_loss": 0.0333566889166832, "global_step": 226807, "epoch": 2548, "lr": 6.549838019421639e-06}
{"train_loss": 0.04345502331852913, "global_step": 226808, "epoch": 2548, "lr": 6.549551164962975e-06}
{"train_loss": 0.034296661615371704, "global_step": 226809, "epoch": 2548, "lr": 6.5492643163456825e-06}
{"train_loss": 0.05674045532941818, "global_step": 226810, "epoch": 2548, "lr": 6.548977473569823e-06}
{"train_loss": 0.06000526621937752, "global_step": 226811, "epoch": 2548, "lr": 6.548690636635413e-06}
{"train_loss": 0.04670840501785278, "global_step": 226812, "epoch": 2548, "lr": 6.548403805542508e-06}
{"train_loss": 0.04910263046622276, "global_step": 226813, "epoch": 2548, "lr": 6.548116980291136e-06}
{"train_loss": 0.05993151664733887, "global_step": 226814, "epoch": 2548, "lr": 6.547830160881352e-06}
{"train_loss": 0.04388701543211937, "global_step": 226815, "epoch": 2548, "lr": 6.547543347313168e-06}
{"train_loss": 0.05464009940624237, "global_step": 226816, "epoch": 2548, "lr": 6.547256539586649e-06}
{"train_loss": 0.02006257139146328, "global_step": 226817, "epoch": 2548, "lr": 6.546969737701808e-06}
{"train_loss": 0.025298964232206345, "global_step": 226818, "epoch": 2548, "lr": 6.5466829416587e-06}
{"train_loss": 0.039624448865652084, "global_step": 226819, "epoch": 2548, "lr": 6.546396151457374e-06}
{"train_loss": 0.03694191575050354, "global_step": 226820, "epoch": 2548, "lr": 6.546109367097841e-06}
{"train_loss": 0.07053670287132263, "global_step": 226821, "epoch": 2548, "lr": 6.545822588580164e-06}
{"train_loss": 0.052044499665498734, "global_step": 226822, "epoch": 2548, "lr": 6.545535815904364e-06}
{"train_loss": 0.0387246310710907, "global_step": 226823, "epoch": 2548, "lr": 6.545249049070495e-06}
{"train_loss": 0.07456978410482407, "global_step": 226824, "epoch": 2548, "lr": 6.544962288078582e-06}
{"train_loss": 0.05223137512803078, "global_step": 226825, "epoch": 2548, "lr": 6.544675532928673e-06}
{"train_loss": 0.04577271640300751, "global_step": 226826, "epoch": 2548, "lr": 6.544388783620797e-06}
{"train_loss": 0.0862630233168602, "global_step": 226827, "epoch": 2548, "lr": 6.544102040155009e-06}
{"train_loss": 0.058167606592178345, "global_step": 226828, "epoch": 2548, "lr": 6.543815302531326e-06}
{"train_loss": 0.03845078870654106, "global_step": 226829, "epoch": 2548, "lr": 6.543528570749807e-06}
{"train_loss": 0.0798536166548729, "global_step": 226830, "epoch": 2548, "lr": 6.543241844810466e-06}
{"train_loss": 0.045180268585681915, "global_step": 226831, "epoch": 2548, "lr": 6.542955124713374e-06}
{"train_loss": 0.0546317994594574, "global_step": 226832, "epoch": 2548, "lr": 6.542668410458547e-06}
{"train_loss": 0.039729174226522446, "global_step": 226833, "epoch": 2548, "lr": 6.542381702046019e-06}
{"train_loss": 0.049228135496377945, "global_step": 226834, "epoch": 2548, "lr": 6.542094999475851e-06}
{"train_loss": 0.06850767135620117, "global_step": 226835, "epoch": 2548, "lr": 6.541808302748059e-06}
{"train_loss": 0.07465333491563797, "global_step": 226836, "epoch": 2548, "lr": 6.5415216118627e-06}
{"train_loss": 0.09404650330543518, "global_step": 226837, "epoch": 2548, "lr": 6.541234926819789e-06}
{"train_loss": 0.039585426449775696, "global_step": 226838, "epoch": 2548, "lr": 6.540948247619394e-06}
{"train_loss": 0.06647881865501404, "global_step": 226839, "epoch": 2548, "lr": 6.54066157426153e-06}
{"train_loss": 0.04550619795918465, "global_step": 226840, "epoch": 2548, "lr": 6.540374906746249e-06}
{"train_loss": 0.057798098772764206, "global_step": 226841, "epoch": 2548, "lr": 6.540088245073583e-06}
{"train_loss": 0.03609044849872589, "global_step": 226842, "epoch": 2548, "lr": 6.539801589243577e-06}
{"train_loss": 0.025215288624167442, "global_step": 226843, "epoch": 2548, "lr": 6.539514939256253e-06}
{"train_loss": 0.07121584564447403, "global_step": 226844, "epoch": 2548, "lr": 6.539228295111677e-06}
{"train_loss": 0.08582041412591934, "global_step": 226845, "epoch": 2548, "lr": 6.538941656809861e-06}
{"train_loss": 0.05808757245540619, "global_step": 226846, "epoch": 2548, "lr": 6.538655024350859e-06}
{"train_loss": 0.03863260895013809, "global_step": 226847, "epoch": 2548, "lr": 6.538368397734701e-06}
{"train_loss": 0.03696882724761963, "global_step": 226848, "epoch": 2548, "lr": 6.538081776961436e-06}
{"train_loss": 0.03245707228779793, "global_step": 226849, "epoch": 2548, "lr": 6.537795162031091e-06}
{"train_loss": 0.048298709094524384, "global_step": 226850, "epoch": 2548, "lr": 6.537508552943716e-06}
{"train_loss": 0.044925034046173096, "global_step": 226851, "epoch": 2548, "lr": 6.53722194969933e-06}
{"train_loss": 0.029102366417646408, "global_step": 226852, "epoch": 2548, "lr": 6.536935352298002e-06}
{"train_loss": 0.03800070285797119, "global_step": 226853, "epoch": 2548, "lr": 6.5366487607397385e-06}
{"train_loss": 0.029196713119745255, "global_step": 226854, "epoch": 2548, "lr": 6.5363621750245905e-06}
{"train_loss": 0.10668184608221054, "global_step": 226855, "epoch": 2548, "lr": 6.536075595152613e-06}
{"train_loss": 0.04144807532429695, "global_step": 226856, "epoch": 2548, "lr": 6.535789021123823e-06}
{"train_loss": 0.029901308938860893, "global_step": 226857, "epoch": 2548, "lr": 6.535502452938275e-06}
{"train_loss": 0.0483013316988945, "global_step": 226858, "epoch": 2548, "lr": 6.535215890595986e-06}
{"train_loss": 0.03995745629072189, "global_step": 226859, "epoch": 2548, "lr": 6.534929334097017e-06}
{"train_loss": 0.052018099935369544, "global_step": 226860, "epoch": 2548, "lr": 6.534642783441391e-06, "val_loss": 9.0698823928833}
{"train_loss": 0.040134355425834656, "global_step": 226861, "epoch": 2549, "lr": 6.534356238629158e-06}
{"train_loss": 0.06147974357008934, "global_step": 226862, "epoch": 2549, "lr": 6.5340696996603445e-06}
{"train_loss": 0.020027510821819305, "global_step": 226863, "epoch": 2549, "lr": 6.533783166535001e-06}
{"train_loss": 0.0679536685347557, "global_step": 226864, "epoch": 2549, "lr": 6.533496639253156e-06}
{"train_loss": 0.06320945918560028, "global_step": 226865, "epoch": 2549, "lr": 6.533210117814859e-06}
{"train_loss": 0.05007588863372803, "global_step": 226866, "epoch": 2549, "lr": 6.532923602220137e-06}
{"train_loss": 0.03698533773422241, "global_step": 226867, "epoch": 2549, "lr": 6.5326370924690405e-06}
{"train_loss": 0.030031507834792137, "global_step": 226868, "epoch": 2549, "lr": 6.5323505885615865e-06}
{"train_loss": 0.10645154118537903, "global_step": 226869, "epoch": 2549, "lr": 6.532064090497847e-06}
{"train_loss": 0.08716179430484772, "global_step": 226870, "epoch": 2549, "lr": 6.531777598277839e-06}
{"train_loss": 0.03977907821536064, "global_step": 226871, "epoch": 2549, "lr": 6.531491111901588e-06}
{"train_loss": 0.056641265749931335, "global_step": 226872, "epoch": 2549, "lr": 6.531204631369165e-06}
{"train_loss": 0.06891566514968872, "global_step": 226873, "epoch": 2549, "lr": 6.530918156680576e-06}
{"train_loss": 0.04684699326753616, "global_step": 226874, "epoch": 2549, "lr": 6.530631687835886e-06}
{"train_loss": 0.049992579966783524, "global_step": 226875, "epoch": 2549, "lr": 6.530345224835116e-06}
{"train_loss": 0.03492450341582298, "global_step": 226876, "epoch": 2549, "lr": 6.530058767678321e-06}
{"train_loss": 0.1089581698179245, "global_step": 226877, "epoch": 2549, "lr": 6.529772316365518e-06}
{"train_loss": 0.05278605967760086, "global_step": 226878, "epoch": 2549, "lr": 6.529485870896768e-06}
{"train_loss": 0.03891449794173241, "global_step": 226879, "epoch": 2549, "lr": 6.529199431272093e-06}
{"train_loss": 0.053072016686201096, "global_step": 226880, "epoch": 2549, "lr": 6.528912997491543e-06}
{"train_loss": 0.03130332753062248, "global_step": 226881, "epoch": 2549, "lr": 6.528626569555141e-06}
{"train_loss": 0.04996214061975479, "global_step": 226882, "epoch": 2549, "lr": 6.5283401474629415e-06}
{"train_loss": 0.04636187106370926, "global_step": 226883, "epoch": 2549, "lr": 6.528053731214973e-06}
{"train_loss": 0.046153292059898376, "global_step": 226884, "epoch": 2549, "lr": 6.527767320811284e-06}
{"train_loss": 0.036621447652578354, "global_step": 226885, "epoch": 2549, "lr": 6.527480916251899e-06}
{"train_loss": 0.039570800960063934, "global_step": 226886, "epoch": 2549, "lr": 6.527194517536872e-06}
{"train_loss": 0.013733538798987865, "global_step": 226887, "epoch": 2549, "lr": 6.526908124666226e-06}
{"train_loss": 0.05893804132938385, "global_step": 226888, "epoch": 2549, "lr": 6.5266217376400205e-06}
{"train_loss": 0.04310346394777298, "global_step": 226889, "epoch": 2549, "lr": 6.526335356458263e-06}
{"train_loss": 0.06482523679733276, "global_step": 226890, "epoch": 2549, "lr": 6.5260489811210135e-06}
{"train_loss": 0.06048555299639702, "global_step": 226891, "epoch": 2549, "lr": 6.5257626116283225e-06}
{"train_loss": 0.06487976014614105, "global_step": 226892, "epoch": 2549, "lr": 6.5254762479801945e-06}
{"train_loss": 0.040325943380594254, "global_step": 226893, "epoch": 2549, "lr": 6.525189890176703e-06}
{"train_loss": 0.08281659334897995, "global_step": 226894, "epoch": 2549, "lr": 6.5249035382178525e-06}
{"train_loss": 0.07026205956935883, "global_step": 226895, "epoch": 2549, "lr": 6.524617192103716e-06}
{"train_loss": 0.05991240218281746, "global_step": 226896, "epoch": 2549, "lr": 6.524330851834298e-06}
{"train_loss": 0.04160619527101517, "global_step": 226897, "epoch": 2549, "lr": 6.524044517409667e-06}
{"train_loss": 0.03377331420779228, "global_step": 226898, "epoch": 2549, "lr": 6.523758188829843e-06}
{"train_loss": 0.03994166851043701, "global_step": 226899, "epoch": 2549, "lr": 6.523471866094872e-06}
{"train_loss": 0.02175467647612095, "global_step": 226900, "epoch": 2549, "lr": 6.523185549204786e-06}
{"train_loss": 0.025499144569039345, "global_step": 226901, "epoch": 2549, "lr": 6.522899238159635e-06}
{"train_loss": 0.07559947669506073, "global_step": 226902, "epoch": 2549, "lr": 6.5226129329594375e-06}
{"train_loss": 0.04699156433343887, "global_step": 226903, "epoch": 2549, "lr": 6.522326633604259e-06}
{"train_loss": 0.05096884071826935, "global_step": 226904, "epoch": 2549, "lr": 6.522040340094115e-06}
{"train_loss": 0.05467076972126961, "global_step": 226905, "epoch": 2549, "lr": 6.5217540524290634e-06}
{"train_loss": 0.07842149585485458, "global_step": 226906, "epoch": 2549, "lr": 6.521467770609119e-06}
{"train_loss": 0.06563805043697357, "global_step": 226907, "epoch": 2549, "lr": 6.521181494634343e-06}
{"train_loss": 0.11131755262613297, "global_step": 226908, "epoch": 2549, "lr": 6.520895224504764e-06}
{"train_loss": 0.05042733997106552, "global_step": 226909, "epoch": 2549, "lr": 6.520608960220409e-06}
{"train_loss": 0.06504049152135849, "global_step": 226910, "epoch": 2549, "lr": 6.520322701781345e-06}
{"train_loss": 0.025549132376909256, "global_step": 226911, "epoch": 2549, "lr": 6.520036449187578e-06}
{"train_loss": 0.05684930086135864, "global_step": 226912, "epoch": 2549, "lr": 6.519750202439173e-06}
{"train_loss": 0.06088857352733612, "global_step": 226913, "epoch": 2549, "lr": 6.519463961536148e-06}
{"train_loss": 0.03911200538277626, "global_step": 226914, "epoch": 2549, "lr": 6.519177726478564e-06}
{"train_loss": 0.04376711696386337, "global_step": 226915, "epoch": 2549, "lr": 6.518891497266433e-06}
{"train_loss": 0.050860848277807236, "global_step": 226916, "epoch": 2549, "lr": 6.51860527389982e-06}
{"train_loss": 0.07429991662502289, "global_step": 226917, "epoch": 2549, "lr": 6.518319056378735e-06}
{"train_loss": 0.04912196844816208, "global_step": 226918, "epoch": 2549, "lr": 6.518032844703248e-06}
{"train_loss": 0.040228188037872314, "global_step": 226919, "epoch": 2549, "lr": 6.5177466388733675e-06}
{"train_loss": 0.02754204533994198, "global_step": 226920, "epoch": 2549, "lr": 6.517460438889156e-06}
{"train_loss": 0.09162262827157974, "global_step": 226921, "epoch": 2549, "lr": 6.517174244750635e-06}
{"train_loss": 0.029681839048862457, "global_step": 226922, "epoch": 2549, "lr": 6.5168880564578595e-06}
{"train_loss": 0.045663997530937195, "global_step": 226923, "epoch": 2549, "lr": 6.516601874010847e-06}
{"train_loss": 0.028066221624612808, "global_step": 226924, "epoch": 2549, "lr": 6.5163156974096584e-06}
{"train_loss": 0.09576334804296494, "global_step": 226925, "epoch": 2549, "lr": 6.516029526654311e-06}
{"train_loss": 0.04591640457510948, "global_step": 226926, "epoch": 2549, "lr": 6.515743361744853e-06}
{"train_loss": 0.06361868977546692, "global_step": 226927, "epoch": 2549, "lr": 6.51545720268133e-06}
{"train_loss": 0.07932274043560028, "global_step": 226928, "epoch": 2549, "lr": 6.515171049463769e-06}
{"train_loss": 0.060364700853824615, "global_step": 226929, "epoch": 2549, "lr": 6.514884902092222e-06}
{"train_loss": 0.05895865708589554, "global_step": 226930, "epoch": 2549, "lr": 6.514598760566709e-06}
{"train_loss": 0.046556003391742706, "global_step": 226931, "epoch": 2549, "lr": 6.514312624887286e-06}
{"train_loss": 0.0999641865491867, "global_step": 226932, "epoch": 2549, "lr": 6.514026495053976e-06}
{"train_loss": 0.0468410886824131, "global_step": 226933, "epoch": 2549, "lr": 6.513740371066834e-06}
{"train_loss": 0.07234175503253937, "global_step": 226934, "epoch": 2549, "lr": 6.5134542529258765e-06}
{"train_loss": 0.0711856260895729, "global_step": 226935, "epoch": 2549, "lr": 6.513168140631171e-06}
{"train_loss": 0.06143632158637047, "global_step": 226936, "epoch": 2549, "lr": 6.512882034182727e-06}
{"train_loss": 0.04681849107146263, "global_step": 226937, "epoch": 2549, "lr": 6.512595933580606e-06}
{"train_loss": 0.044017355889081955, "global_step": 226938, "epoch": 2549, "lr": 6.512309838824826e-06}
{"train_loss": 0.06931101530790329, "global_step": 226939, "epoch": 2549, "lr": 6.512023749915447e-06}
{"train_loss": 0.04779725894331932, "global_step": 226940, "epoch": 2549, "lr": 6.511737666852485e-06}
{"train_loss": 0.047135911881923676, "global_step": 226941, "epoch": 2549, "lr": 6.511451589636003e-06}
{"train_loss": 0.016081463545560837, "global_step": 226942, "epoch": 2549, "lr": 6.51116551826601e-06}
{"train_loss": 0.05896596238017082, "global_step": 226943, "epoch": 2549, "lr": 6.5108794527425745e-06}
{"train_loss": 0.03427112475037575, "global_step": 226944, "epoch": 2549, "lr": 6.510593393065711e-06}
{"train_loss": 0.020244963467121124, "global_step": 226945, "epoch": 2549, "lr": 6.5103073392354826e-06}
{"train_loss": 0.04752099886536598, "global_step": 226946, "epoch": 2549, "lr": 6.510021291251895e-06}
{"train_loss": 0.03583066910505295, "global_step": 226947, "epoch": 2549, "lr": 6.509735249115023e-06}
{"train_loss": 0.05424557998776436, "global_step": 226948, "epoch": 2549, "lr": 6.509449212824881e-06}
{"train_loss": 0.052889942861256305, "global_step": 226949, "epoch": 2549, "lr": 6.509163182381506e-06, "val_loss": 9.158158302307129}
{"train_loss": 0.1046130508184433, "global_step": 226950, "epoch": 2550, "lr": 6.5088771577849536e-06}
{"train_loss": 0.012076844461262226, "global_step": 226951, "epoch": 2550, "lr": 6.508591139035242e-06}
{"train_loss": 0.06488776952028275, "global_step": 226952, "epoch": 2550, "lr": 6.50830512613243e-06}
{"train_loss": 0.06167532876133919, "global_step": 226953, "epoch": 2550, "lr": 6.508019119076542e-06}
{"train_loss": 0.04900398105382919, "global_step": 226954, "epoch": 2550, "lr": 6.507733117867626e-06}
{"train_loss": 0.14576022326946259, "global_step": 226955, "epoch": 2550, "lr": 6.5074471225057e-06}
{"train_loss": 0.03695254400372505, "global_step": 226956, "epoch": 2550, "lr": 6.507161132990836e-06}
{"train_loss": 0.032645486295223236, "global_step": 226957, "epoch": 2550, "lr": 6.506875149323039e-06}
{"train_loss": 0.0716334655880928, "global_step": 226958, "epoch": 2550, "lr": 6.5065891715023755e-06}
{"train_loss": 0.07795310020446777, "global_step": 226959, "epoch": 2550, "lr": 6.506303199528857e-06}
{"train_loss": 0.03531615436077118, "global_step": 226960, "epoch": 2550, "lr": 6.50601723340255e-06}
{"train_loss": 0.018277054652571678, "global_step": 226961, "epoch": 2550, "lr": 6.5057312731234666e-06}
{"train_loss": 0.04775107279419899, "global_step": 226962, "epoch": 2550, "lr": 6.505445318691661e-06}
{"train_loss": 0.06821516156196594, "global_step": 226963, "epoch": 2550, "lr": 6.505159370107177e-06}
{"train_loss": 0.0603925883769989, "global_step": 226964, "epoch": 2550, "lr": 6.5048734273700285e-06}
{"train_loss": 0.05246590077877045, "global_step": 226965, "epoch": 2550, "lr": 6.504587490480285e-06}
{"train_loss": 0.05100839212536812, "global_step": 226966, "epoch": 2550, "lr": 6.504301559437958e-06}
{"train_loss": 0.024433910846710205, "global_step": 226967, "epoch": 2550, "lr": 6.50401563424311e-06}
{"train_loss": 0.07786332070827484, "global_step": 226968, "epoch": 2550, "lr": 6.503729714895757e-06}
{"train_loss": 0.06688836961984634, "global_step": 226969, "epoch": 2550, "lr": 6.503443801395953e-06}
{"train_loss": 0.012147191911935806, "global_step": 226970, "epoch": 2550, "lr": 6.503157893743722e-06}
{"train_loss": 0.032161369919776917, "global_step": 226971, "epoch": 2550, "lr": 6.502871991939125e-06}
{"train_loss": 0.05797435715794563, "global_step": 226972, "epoch": 2550, "lr": 6.502586095982171e-06}
{"train_loss": 0.019151300191879272, "global_step": 226973, "epoch": 2550, "lr": 6.5023002058729295e-06}
{"train_loss": 0.017557673156261444, "global_step": 226974, "epoch": 2550, "lr": 6.502014321611411e-06}
{"train_loss": 0.0978025272488594, "global_step": 226975, "epoch": 2550, "lr": 6.50172844319768e-06}
{"train_loss": 0.07275586575269699, "global_step": 226976, "epoch": 2550, "lr": 6.501442570631744e-06}
{"train_loss": 0.10083179175853729, "global_step": 226977, "epoch": 2550, "lr": 6.501156703913675e-06}
{"train_loss": 0.038929618895053864, "global_step": 226978, "epoch": 2550, "lr": 6.5008708430434825e-06}
{"train_loss": 0.07050133496522903, "global_step": 226979, "epoch": 2550, "lr": 6.50058498802123e-06}
{"train_loss": 0.08657477051019669, "global_step": 226980, "epoch": 2550, "lr": 6.5002991388469325e-06}
{"train_loss": 0.02617758698761463, "global_step": 226981, "epoch": 2550, "lr": 6.500013295520652e-06}
{"train_loss": 0.027402738109230995, "global_step": 226982, "epoch": 2550, "lr": 6.499727458042398e-06}
{"train_loss": 0.05691725015640259, "global_step": 226983, "epoch": 2550, "lr": 6.499441626412245e-06}
{"train_loss": 0.06447167694568634, "global_step": 226984, "epoch": 2550, "lr": 6.499155800630191e-06}
{"train_loss": 0.08324912190437317, "global_step": 226985, "epoch": 2550, "lr": 6.498869980696315e-06}
{"train_loss": 0.09446555376052856, "global_step": 226986, "epoch": 2550, "lr": 6.498584166610633e-06}
{"train_loss": 0.08357765525579453, "global_step": 226987, "epoch": 2550, "lr": 6.4982983583731725e-06}
{"train_loss": 0.17489701509475708, "global_step": 226988, "epoch": 2550, "lr": 6.498012555983996e-06}
{"train_loss": 0.06826720386743546, "global_step": 226989, "epoch": 2550, "lr": 6.497726759443123e-06}
{"train_loss": 0.05855082720518112, "global_step": 226990, "epoch": 2550, "lr": 6.497440968750612e-06}
{"train_loss": 0.03061078116297722, "global_step": 226991, "epoch": 2550, "lr": 6.4971551839064785e-06}
{"train_loss": 0.06995685398578644, "global_step": 226992, "epoch": 2550, "lr": 6.496869404910782e-06}
{"train_loss": 0.07200425863265991, "global_step": 226993, "epoch": 2550, "lr": 6.496583631763542e-06}
{"train_loss": 0.04944142699241638, "global_step": 226994, "epoch": 2550, "lr": 6.496297864464818e-06}
{"train_loss": 0.032941095530986786, "global_step": 226995, "epoch": 2550, "lr": 6.49601210301462e-06}
{"train_loss": 0.0831054300069809, "global_step": 226996, "epoch": 2550, "lr": 6.495726347413017e-06}
{"train_loss": 0.06003929302096367, "global_step": 226997, "epoch": 2550, "lr": 6.4954405976600245e-06}
{"train_loss": 0.054704733192920685, "global_step": 226998, "epoch": 2550, "lr": 6.495154853755697e-06}
{"train_loss": 0.027224348857998848, "global_step": 226999, "epoch": 2550, "lr": 6.4948691157000585e-06}
{"train_loss": 0.022463226690888405, "global_step": 227000, "epoch": 2550, "lr": 6.494583383493153e-06}
{"train_loss": 0.06090336665511131, "global_step": 227001, "epoch": 2550, "lr": 6.494297657135029e-06}
{"train_loss": 0.020144976675510406, "global_step": 227002, "epoch": 2550, "lr": 6.49401193662571e-06}
{"train_loss": 0.03086015395820141, "global_step": 227003, "epoch": 2550, "lr": 6.493726221965252e-06}
{"train_loss": 0.05285370722413063, "global_step": 227004, "epoch": 2550, "lr": 6.493440513153665e-06}
{"train_loss": 0.04498272016644478, "global_step": 227005, "epoch": 2550, "lr": 6.493154810191021e-06}
{"train_loss": 0.020068921148777008, "global_step": 227006, "epoch": 2550, "lr": 6.492869113077327e-06}
{"train_loss": 0.047425612807273865, "global_step": 227007, "epoch": 2550, "lr": 6.492583421812654e-06}
{"train_loss": 0.11120925098657608, "global_step": 227008, "epoch": 2550, "lr": 6.492297736397008e-06}
{"train_loss": 0.041829489171504974, "global_step": 227009, "epoch": 2550, "lr": 6.492012056830448e-06}
{"train_loss": 0.05230484530329704, "global_step": 227010, "epoch": 2550, "lr": 6.491726383113006e-06}
{"train_loss": 0.0833832249045372, "global_step": 227011, "epoch": 2550, "lr": 6.491440715244723e-06}
{"train_loss": 0.07304955273866653, "global_step": 227012, "epoch": 2550, "lr": 6.4911550532256285e-06}
{"train_loss": 0.07660596817731857, "global_step": 227013, "epoch": 2550, "lr": 6.490869397055782e-06}
{"train_loss": 0.057029448449611664, "global_step": 227014, "epoch": 2550, "lr": 6.49058374673519e-06}
{"train_loss": 0.06369076669216156, "global_step": 227015, "epoch": 2550, "lr": 6.490298102263925e-06}
{"train_loss": 0.05135778710246086, "global_step": 227016, "epoch": 2550, "lr": 6.490012463641998e-06}
{"train_loss": 0.05221378058195114, "global_step": 227017, "epoch": 2550, "lr": 6.489726830869458e-06}
{"train_loss": 0.04463135823607445, "global_step": 227018, "epoch": 2550, "lr": 6.489441203946351e-06}
{"train_loss": 0.034171413630247116, "global_step": 227019, "epoch": 2550, "lr": 6.489155582872714e-06}
{"train_loss": 0.05586414784193039, "global_step": 227020, "epoch": 2550, "lr": 6.4888699676485645e-06}
{"train_loss": 0.05038667470216751, "global_step": 227021, "epoch": 2550, "lr": 6.488584358273969e-06}
{"train_loss": 0.029137548059225082, "global_step": 227022, "epoch": 2550, "lr": 6.4882987547489395e-06}
{"train_loss": 0.047648631036281586, "global_step": 227023, "epoch": 2550, "lr": 6.488013157073542e-06}
{"train_loss": 0.03830160200595856, "global_step": 227024, "epoch": 2550, "lr": 6.4877275652477975e-06}
{"train_loss": 0.07143355906009674, "global_step": 227025, "epoch": 2550, "lr": 6.487441979271741e-06}
{"train_loss": 0.08353918790817261, "global_step": 227026, "epoch": 2550, "lr": 6.487156399145427e-06}
{"train_loss": 0.08368979394435883, "global_step": 227027, "epoch": 2550, "lr": 6.486870824868873e-06}
{"train_loss": 0.06975087523460388, "global_step": 227028, "epoch": 2550, "lr": 6.48658525644214e-06}
{"train_loss": 0.06233138218522072, "global_step": 227029, "epoch": 2550, "lr": 6.486299693865245e-06}
{"train_loss": 0.09215234965085983, "global_step": 227030, "epoch": 2550, "lr": 6.486014137138247e-06}
{"train_loss": 0.061311274766922, "global_step": 227031, "epoch": 2550, "lr": 6.485728586261164e-06}
{"train_loss": 0.05722189322113991, "global_step": 227032, "epoch": 2550, "lr": 6.485443041234052e-06}
{"train_loss": 0.024201057851314545, "global_step": 227033, "epoch": 2550, "lr": 6.485157502056932e-06}
{"train_loss": 0.03930014744400978, "global_step": 227034, "epoch": 2550, "lr": 6.484871968729867e-06}
{"train_loss": 0.030545448884367943, "global_step": 227035, "epoch": 2550, "lr": 6.4845864412528656e-06}
{"train_loss": 0.02986793965101242, "global_step": 227036, "epoch": 2550, "lr": 6.48430091962598e-06}
{"train_loss": 0.05579952523112297, "global_step": 227037, "epoch": 2550, "lr": 6.484015403849264e-06}
{"train_loss": 0.05678172492202413, "global_step": 227038, "epoch": 2550, "lr": 6.48372989392273e-06, "train/sim_max_reward_0": 0.5557678411370527, "train/sim_max_reward_1": 0.8076600647307502, "train/sim_max_reward_2": 0.6433553589731229, "train/sim_max_reward_3": 0.3977429065792172, "train/sim_max_reward_4": 0.9564206992710741, "train/sim_max_reward_5": 0.9775393178916003, "test/sim_max_reward_4500000": 0.9853836376502816, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.19617144988378335, "test/sim_max_reward_4500004": 0.5708704550805249, "test/sim_max_reward_4500005": 1.0, "test/sim_max_reward_4500006": 0.9841769553890081, "test/sim_max_reward_4500007": 0.9143951872215752, "test/sim_max_reward_4500008": 0.22608793794065735, "test/sim_max_reward_4500009": 0.04907154754677626, "test/sim_max_reward_4500010": 0.5471014839192988, "test/sim_max_reward_4500011": 0.9830794586594801, "test/sim_max_reward_4500012": 1.0, "test/sim_max_reward_4500013": 0.49395987880757924, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9865705725534133, "test/sim_max_reward_4500016": 0.5686526726183375, "test/sim_max_reward_4500017": 0.9465079184400101, "test/sim_max_reward_4500018": 0.9927426846234494, "test/sim_max_reward_4500019": 0.39496762743265895, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.937386929073143, "test/sim_max_reward_4500023": 0.5955367383070195, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.8475317419519773, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.5582263130595803, "test/sim_max_reward_4500029": 0.2743057213316826, "test/sim_max_reward_4500030": 0.9065515587471427, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.3931565779990177, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.4066141137437851, "test/sim_max_reward_4500035": 0.47250354595220645, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.162329845396134, "test/sim_max_reward_4500039": 0.9728543923246539, "test/sim_max_reward_4500040": 0.9934482960794859, "test/sim_max_reward_4500041": 0.5853138470438551, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9533657606424504, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.4207666361897011, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9968642949207879, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7230810314304694, "test/mean_score": 0.626329915610589, "val_loss": 9.130756378173828, "train_action_mse_error": 7.181906700134277}
{"train_loss": 0.0668368861079216, "global_step": 227039, "epoch": 2551, "lr": 6.483444389846438e-06}
{"train_loss": 0.047271065413951874, "global_step": 227040, "epoch": 2551, "lr": 6.483158891620411e-06}
{"train_loss": 0.010537609457969666, "global_step": 227041, "epoch": 2551, "lr": 6.4828733992446985e-06}
{"train_loss": 0.039271965622901917, "global_step": 227042, "epoch": 2551, "lr": 6.482587912719323e-06}
{"train_loss": 0.10598916560411453, "global_step": 227043, "epoch": 2551, "lr": 6.482302432044341e-06}
{"train_loss": 0.0508047491312027, "global_step": 227044, "epoch": 2551, "lr": 6.4820169572197775e-06}
{"train_loss": 0.10347399860620499, "global_step": 227045, "epoch": 2551, "lr": 6.4817314882456845e-06}
{"train_loss": 0.049102794378995895, "global_step": 227046, "epoch": 2551, "lr": 6.4814460251220844e-06}
{"train_loss": 0.03577869385480881, "global_step": 227047, "epoch": 2551, "lr": 6.481160567849032e-06}
{"train_loss": 0.01682952232658863, "global_step": 227048, "epoch": 2551, "lr": 6.480875116426544e-06}
{"train_loss": 0.019539110362529755, "global_step": 227049, "epoch": 2551, "lr": 6.480589670854686e-06}
{"train_loss": 0.019138706848025322, "global_step": 227050, "epoch": 2551, "lr": 6.4803042311334714e-06}
{"train_loss": 0.019382774829864502, "global_step": 227051, "epoch": 2551, "lr": 6.48001879726296e-06}
{"train_loss": 0.060349710285663605, "global_step": 227052, "epoch": 2551, "lr": 6.479733369243168e-06}
{"train_loss": 0.057293348014354706, "global_step": 227053, "epoch": 2551, "lr": 6.4794479470741464e-06}
{"train_loss": 0.04585614055395126, "global_step": 227054, "epoch": 2551, "lr": 6.479162530755939e-06}
{"train_loss": 0.099562868475914, "global_step": 227055, "epoch": 2551, "lr": 6.478877120288573e-06}
{"train_loss": 0.04734629765152931, "global_step": 227056, "epoch": 2551, "lr": 6.4785917156721e-06}
{"train_loss": 0.04215944930911064, "global_step": 227057, "epoch": 2551, "lr": 6.478306316906551e-06}
{"train_loss": 0.06881251186132431, "global_step": 227058, "epoch": 2551, "lr": 6.478020923991951e-06}
{"train_loss": 0.027368614450097084, "global_step": 227059, "epoch": 2551, "lr": 6.477735536928359e-06}
{"train_loss": 0.06122038885951042, "global_step": 227060, "epoch": 2551, "lr": 6.477450155715797e-06}
{"train_loss": 0.06437011063098907, "global_step": 227061, "epoch": 2551, "lr": 6.4771647803543225e-06}
{"train_loss": 0.030095698311924934, "global_step": 227062, "epoch": 2551, "lr": 6.47687941084395e-06}
{"train_loss": 0.07956373691558838, "global_step": 227063, "epoch": 2551, "lr": 6.476594047184747e-06}
{"train_loss": 0.04422083497047424, "global_step": 227064, "epoch": 2551, "lr": 6.476308689376731e-06}
{"train_loss": 0.10331831872463226, "global_step": 227065, "epoch": 2551, "lr": 6.476023337419934e-06}
{"train_loss": 0.033442266285419464, "global_step": 227066, "epoch": 2551, "lr": 6.475737991314418e-06}
{"train_loss": 0.05688222125172615, "global_step": 227067, "epoch": 2551, "lr": 6.475452651060193e-06}
{"train_loss": 0.052692148834466934, "global_step": 227068, "epoch": 2551, "lr": 6.475167316657327e-06}
{"train_loss": 0.046982791274785995, "global_step": 227069, "epoch": 2551, "lr": 6.474881988105835e-06}
{"train_loss": 0.04890771582722664, "global_step": 227070, "epoch": 2551, "lr": 6.474596665405774e-06}
{"train_loss": 0.03467395901679993, "global_step": 227071, "epoch": 2551, "lr": 6.4743113485571606e-06}
{"train_loss": 0.043827157467603683, "global_step": 227072, "epoch": 2551, "lr": 6.474026037560049e-06}
{"train_loss": 0.040136195719242096, "global_step": 227073, "epoch": 2551, "lr": 6.47374073241448e-06}
{"train_loss": 0.026894710958003998, "global_step": 227074, "epoch": 2551, "lr": 6.473455433120479e-06}
{"train_loss": 0.03433214873075485, "global_step": 227075, "epoch": 2551, "lr": 6.473170139678103e-06}
{"train_loss": 0.07763340324163437, "global_step": 227076, "epoch": 2551, "lr": 6.472884852087363e-06}
{"train_loss": 0.04943757504224777, "global_step": 227077, "epoch": 2551, "lr": 6.472599570348326e-06}
{"train_loss": 0.06043354049324989, "global_step": 227078, "epoch": 2551, "lr": 6.472314294461007e-06}
{"train_loss": 0.04344189167022705, "global_step": 227079, "epoch": 2551, "lr": 6.472029024425469e-06}
{"train_loss": 0.08583967387676239, "global_step": 227080, "epoch": 2551, "lr": 6.471743760241722e-06}
{"train_loss": 0.05534496158361435, "global_step": 227081, "epoch": 2551, "lr": 6.471458501909833e-06}
{"train_loss": 0.0457315556704998, "global_step": 227082, "epoch": 2551, "lr": 6.471173249429807e-06}
{"train_loss": 0.058664582669734955, "global_step": 227083, "epoch": 2551, "lr": 6.470888002801717e-06}
{"train_loss": 0.06696977466344833, "global_step": 227084, "epoch": 2551, "lr": 6.4706027620255795e-06}
{"train_loss": 0.06911933422088623, "global_step": 227085, "epoch": 2551, "lr": 6.4703175271014435e-06}
{"train_loss": 0.062211692333221436, "global_step": 227086, "epoch": 2551, "lr": 6.470032298029333e-06}
{"train_loss": 0.03899957239627838, "global_step": 227087, "epoch": 2551, "lr": 6.469747074809307e-06}
{"train_loss": 0.07271897792816162, "global_step": 227088, "epoch": 2551, "lr": 6.4694618574413834e-06}
{"train_loss": 0.06431499123573303, "global_step": 227089, "epoch": 2551, "lr": 6.469176645925607e-06}
{"train_loss": 0.09626314043998718, "global_step": 227090, "epoch": 2551, "lr": 6.468891440262032e-06}
{"train_loss": 0.03052535466849804, "global_step": 227091, "epoch": 2551, "lr": 6.468606240450676e-06}
{"train_loss": 0.05545695498585701, "global_step": 227092, "epoch": 2551, "lr": 6.468321046491593e-06}
{"train_loss": 0.08045157790184021, "global_step": 227093, "epoch": 2551, "lr": 6.468035858384802e-06}
{"train_loss": 0.05412988364696503, "global_step": 227094, "epoch": 2551, "lr": 6.467750676130369e-06}
{"train_loss": 0.0650443509221077, "global_step": 227095, "epoch": 2551, "lr": 6.467465499728298e-06}
{"train_loss": 0.06725704669952393, "global_step": 227096, "epoch": 2551, "lr": 6.467180329178663e-06}
{"train_loss": 0.04541589319705963, "global_step": 227097, "epoch": 2551, "lr": 6.466895164481479e-06}
{"train_loss": 0.03228723257780075, "global_step": 227098, "epoch": 2551, "lr": 6.4666100056367865e-06}
{"train_loss": 0.043432239443063736, "global_step": 227099, "epoch": 2551, "lr": 6.466324852644634e-06}
{"train_loss": 0.06394094973802567, "global_step": 227100, "epoch": 2551, "lr": 6.466039705505045e-06}
{"train_loss": 0.04410563409328461, "global_step": 227101, "epoch": 2551, "lr": 6.465754564218074e-06}
{"train_loss": 0.08537588268518448, "global_step": 227102, "epoch": 2551, "lr": 6.465469428783754e-06}
{"train_loss": 0.022546814754605293, "global_step": 227103, "epoch": 2551, "lr": 6.465184299202109e-06}
{"train_loss": 0.02211948111653328, "global_step": 227104, "epoch": 2551, "lr": 6.464899175473199e-06}
{"train_loss": 0.0656760111451149, "global_step": 227105, "epoch": 2551, "lr": 6.464614057597046e-06}
{"train_loss": 0.034743089228868484, "global_step": 227106, "epoch": 2551, "lr": 6.4643289455737e-06}
{"train_loss": 0.01690014638006687, "global_step": 227107, "epoch": 2551, "lr": 6.464043839403189e-06}
{"train_loss": 0.04357032850384712, "global_step": 227108, "epoch": 2551, "lr": 6.463758739085551e-06}
{"train_loss": 0.05386098474264145, "global_step": 227109, "epoch": 2551, "lr": 6.463473644620849e-06}
{"train_loss": 0.05732293799519539, "global_step": 227110, "epoch": 2551, "lr": 6.463188556009087e-06}
{"train_loss": 0.0958632081747055, "global_step": 227111, "epoch": 2551, "lr": 6.462903473250326e-06}
{"train_loss": 0.06933694332838058, "global_step": 227112, "epoch": 2551, "lr": 6.462618396344594e-06}
{"train_loss": 0.041942499577999115, "global_step": 227113, "epoch": 2551, "lr": 6.462333325291936e-06}
{"train_loss": 0.06292032450437546, "global_step": 227114, "epoch": 2551, "lr": 6.462048260092379e-06}
{"train_loss": 0.08944819122552872, "global_step": 227115, "epoch": 2551, "lr": 6.461763200745979e-06}
{"train_loss": 0.04311414435505867, "global_step": 227116, "epoch": 2551, "lr": 6.461478147252753e-06}
{"train_loss": 0.06290144473314285, "global_step": 227117, "epoch": 2551, "lr": 6.461193099612761e-06}
{"train_loss": 0.05030657351016998, "global_step": 227118, "epoch": 2551, "lr": 6.460908057826026e-06}
{"train_loss": 0.03796764090657234, "global_step": 227119, "epoch": 2551, "lr": 6.460623021892598e-06}
{"train_loss": 0.032254111021757126, "global_step": 227120, "epoch": 2551, "lr": 6.460337991812493e-06}
{"train_loss": 0.06191985309123993, "global_step": 227121, "epoch": 2551, "lr": 6.460052967585784e-06}
{"train_loss": 0.09685741364955902, "global_step": 227122, "epoch": 2551, "lr": 6.45976794921247e-06}
{"train_loss": 0.03995170444250107, "global_step": 227123, "epoch": 2551, "lr": 6.459482936692629e-06}
{"train_loss": 0.032497089356184006, "global_step": 227124, "epoch": 2551, "lr": 6.459197930026262e-06}
{"train_loss": 0.046217553317546844, "global_step": 227125, "epoch": 2551, "lr": 6.4589129292134395e-06}
{"train_loss": 0.05728764459490776, "global_step": 227126, "epoch": 2551, "lr": 6.45862793425418e-06}
{"train_loss": 0.0539035370803616, "global_step": 227127, "epoch": 2551, "lr": 6.458342945148521e-06, "val_loss": 9.212417602539062}
{"train_loss": 0.08724483102560043, "global_step": 227128, "epoch": 2552, "lr": 6.458057961896519e-06}
{"train_loss": 0.08411131799221039, "global_step": 227129, "epoch": 2552, "lr": 6.45777298449819e-06}
{"train_loss": 0.045416321605443954, "global_step": 227130, "epoch": 2552, "lr": 6.4574880129535886e-06}
{"train_loss": 0.08646328747272491, "global_step": 227131, "epoch": 2552, "lr": 6.457203047262744e-06}
{"train_loss": 0.04801023751497269, "global_step": 227132, "epoch": 2552, "lr": 6.456918087425706e-06}
{"train_loss": 0.047529544681310654, "global_step": 227133, "epoch": 2552, "lr": 6.456633133442491e-06}
{"train_loss": 0.015700077638030052, "global_step": 227134, "epoch": 2552, "lr": 6.456348185313166e-06}
{"train_loss": 0.03137148544192314, "global_step": 227135, "epoch": 2552, "lr": 6.456063243037752e-06}
{"train_loss": 0.03900311887264252, "global_step": 227136, "epoch": 2552, "lr": 6.455778306616278e-06}
{"train_loss": 0.02519080601632595, "global_step": 227137, "epoch": 2552, "lr": 6.455493376048804e-06}
{"train_loss": 0.03789135068655014, "global_step": 227138, "epoch": 2552, "lr": 6.455208451335349e-06}
{"train_loss": 0.04653428494930267, "global_step": 227139, "epoch": 2552, "lr": 6.454923532475971e-06}
{"train_loss": 0.08947951346635818, "global_step": 227140, "epoch": 2552, "lr": 6.454638619470698e-06}
{"train_loss": 0.045063503086566925, "global_step": 227141, "epoch": 2552, "lr": 6.454353712319561e-06}
{"train_loss": 0.06653071939945221, "global_step": 227142, "epoch": 2552, "lr": 6.454068811022612e-06}
{"train_loss": 0.045088376849889755, "global_step": 227143, "epoch": 2552, "lr": 6.45378391557987e-06}
{"train_loss": 0.06813272833824158, "global_step": 227144, "epoch": 2552, "lr": 6.453499025991389e-06}
{"train_loss": 0.041354477405548096, "global_step": 227145, "epoch": 2552, "lr": 6.4532141422572136e-06}
{"train_loss": 0.035339079797267914, "global_step": 227146, "epoch": 2552, "lr": 6.452929264377367e-06}
{"train_loss": 0.036132946610450745, "global_step": 227147, "epoch": 2552, "lr": 6.452644392351897e-06}
{"train_loss": 0.07104700803756714, "global_step": 227148, "epoch": 2552, "lr": 6.452359526180834e-06}
{"train_loss": 0.04564499855041504, "global_step": 227149, "epoch": 2552, "lr": 6.452074665864227e-06}
{"train_loss": 0.10286688804626465, "global_step": 227150, "epoch": 2552, "lr": 6.451789811402098e-06}
{"train_loss": 0.03680812940001488, "global_step": 227151, "epoch": 2552, "lr": 6.4515049627945025e-06}
{"train_loss": 0.07540294528007507, "global_step": 227152, "epoch": 2552, "lr": 6.451220120041468e-06}
{"train_loss": 0.047092776745557785, "global_step": 227153, "epoch": 2552, "lr": 6.450935283143039e-06}
{"train_loss": 0.07325135916471481, "global_step": 227154, "epoch": 2552, "lr": 6.450650452099244e-06}
{"train_loss": 0.06361208111047745, "global_step": 227155, "epoch": 2552, "lr": 6.4503656269101375e-06}
{"train_loss": 0.04618429020047188, "global_step": 227156, "epoch": 2552, "lr": 6.4500808075757425e-06}
{"train_loss": 0.03849488124251366, "global_step": 227157, "epoch": 2552, "lr": 6.449795994096109e-06}
{"train_loss": 0.046116236597299576, "global_step": 227158, "epoch": 2552, "lr": 6.449511186471263e-06}
{"train_loss": 0.05197572335600853, "global_step": 227159, "epoch": 2552, "lr": 6.449226384701257e-06}
{"train_loss": 0.04795118793845177, "global_step": 227160, "epoch": 2552, "lr": 6.448941588786112e-06}
{"train_loss": 0.07952500134706497, "global_step": 227161, "epoch": 2552, "lr": 6.4486567987258886e-06}
{"train_loss": 0.06174670532345772, "global_step": 227162, "epoch": 2552, "lr": 6.448372014520599e-06}
{"train_loss": 0.03206545114517212, "global_step": 227163, "epoch": 2552, "lr": 6.448087236170292e-06}
{"train_loss": 0.028134740889072418, "global_step": 227164, "epoch": 2552, "lr": 6.447802463675023e-06}
{"train_loss": 0.01905737817287445, "global_step": 227165, "epoch": 2552, "lr": 6.447517697034805e-06}
{"train_loss": 0.05238116905093193, "global_step": 227166, "epoch": 2552, "lr": 6.447232936249703e-06}
{"train_loss": 0.06873713433742523, "global_step": 227167, "epoch": 2552, "lr": 6.446948181319723e-06}
{"train_loss": 0.045207880437374115, "global_step": 227168, "epoch": 2552, "lr": 6.446663432244931e-06}
{"train_loss": 0.03698571398854256, "global_step": 227169, "epoch": 2552, "lr": 6.446378689025346e-06}
{"train_loss": 0.019922716543078423, "global_step": 227170, "epoch": 2552, "lr": 6.446093951661025e-06}
{"train_loss": 0.044273462146520615, "global_step": 227171, "epoch": 2552, "lr": 6.4458092201519824e-06}
{"train_loss": 0.04163737967610359, "global_step": 227172, "epoch": 2552, "lr": 6.445524494498284e-06}
{"train_loss": 0.02604656293988228, "global_step": 227173, "epoch": 2552, "lr": 6.445239774699952e-06}
{"train_loss": 0.07439839094877243, "global_step": 227174, "epoch": 2552, "lr": 6.444955060757019e-06}
{"train_loss": 0.045213911682367325, "global_step": 227175, "epoch": 2552, "lr": 6.444670352669535e-06}
{"train_loss": 0.034795667976140976, "global_step": 227176, "epoch": 2552, "lr": 6.444385650437529e-06}
{"train_loss": 0.06376981735229492, "global_step": 227177, "epoch": 2552, "lr": 6.4441009540610555e-06}
{"train_loss": 0.031069980934262276, "global_step": 227178, "epoch": 2552, "lr": 6.443816263540125e-06}
{"train_loss": 0.05979900434613228, "global_step": 227179, "epoch": 2552, "lr": 6.443531578874812e-06}
{"train_loss": 0.026700451970100403, "global_step": 227180, "epoch": 2552, "lr": 6.44324690006512e-06}
{"train_loss": 0.043577492237091064, "global_step": 227181, "epoch": 2552, "lr": 6.44296222711111e-06}
{"train_loss": 0.04146110638976097, "global_step": 227182, "epoch": 2552, "lr": 6.4426775600128055e-06}
{"train_loss": 0.05196397006511688, "global_step": 227183, "epoch": 2552, "lr": 6.442392898770266e-06}
{"train_loss": 0.0641041249036789, "global_step": 227184, "epoch": 2552, "lr": 6.442108243383499e-06}
{"train_loss": 0.07638981193304062, "global_step": 227185, "epoch": 2552, "lr": 6.441823593852575e-06}
{"train_loss": 0.06072591617703438, "global_step": 227186, "epoch": 2552, "lr": 6.441538950177506e-06}
{"train_loss": 0.06657229363918304, "global_step": 227187, "epoch": 2552, "lr": 6.4412543123583525e-06}
{"train_loss": 0.05822281911969185, "global_step": 227188, "epoch": 2552, "lr": 6.440969680395126e-06}
{"train_loss": 0.04938562959432602, "global_step": 227189, "epoch": 2552, "lr": 6.440685054287892e-06}
{"train_loss": 0.0281072985380888, "global_step": 227190, "epoch": 2552, "lr": 6.44040043403667e-06}
{"train_loss": 0.05749359354376793, "global_step": 227191, "epoch": 2552, "lr": 6.4401158196415125e-06}
{"train_loss": 0.038996536284685135, "global_step": 227192, "epoch": 2552, "lr": 6.439831211102443e-06}
{"train_loss": 0.03645170107483864, "global_step": 227193, "epoch": 2552, "lr": 6.439546608419516e-06}
{"train_loss": 0.050405584275722504, "global_step": 227194, "epoch": 2552, "lr": 6.43926201159275e-06}
{"train_loss": 0.08264387398958206, "global_step": 227195, "epoch": 2552, "lr": 6.438977420622205e-06}
{"train_loss": 0.09207213670015335, "global_step": 227196, "epoch": 2552, "lr": 6.438692835507898e-06}
{"train_loss": 0.025860482826828957, "global_step": 227197, "epoch": 2552, "lr": 6.4384082562498895e-06}
{"train_loss": 0.04362303763628006, "global_step": 227198, "epoch": 2552, "lr": 6.438123682848196e-06}
{"train_loss": 0.02002721279859543, "global_step": 227199, "epoch": 2552, "lr": 6.437839115302863e-06}
{"train_loss": 0.045514557510614395, "global_step": 227200, "epoch": 2552, "lr": 6.4375545536139445e-06}
{"train_loss": 0.07887104898691177, "global_step": 227201, "epoch": 2552, "lr": 6.437269997781454e-06}
{"train_loss": 0.055505674332380295, "global_step": 227202, "epoch": 2552, "lr": 6.436985447805454e-06}
{"train_loss": 0.06836656481027603, "global_step": 227203, "epoch": 2552, "lr": 6.4367009036859604e-06}
{"train_loss": 0.080796018242836, "global_step": 227204, "epoch": 2552, "lr": 6.436416365423026e-06}
{"train_loss": 0.05186517536640167, "global_step": 227205, "epoch": 2552, "lr": 6.436131833016679e-06}
{"train_loss": 0.03485925495624542, "global_step": 227206, "epoch": 2552, "lr": 6.435847306466974e-06}
{"train_loss": 0.03463326394557953, "global_step": 227207, "epoch": 2552, "lr": 6.4355627857739245e-06}
{"train_loss": 0.024360712617635727, "global_step": 227208, "epoch": 2552, "lr": 6.4352782709375945e-06}
{"train_loss": 0.10621176660060883, "global_step": 227209, "epoch": 2552, "lr": 6.434993761958003e-06}
{"train_loss": 0.07350830733776093, "global_step": 227210, "epoch": 2552, "lr": 6.434709258835203e-06}
{"train_loss": 0.03177989646792412, "global_step": 227211, "epoch": 2552, "lr": 6.434424761569213e-06}
{"train_loss": 0.00865801889449358, "global_step": 227212, "epoch": 2552, "lr": 6.434140270160099e-06}
{"train_loss": 0.08235511183738708, "global_step": 227213, "epoch": 2552, "lr": 6.433855784607878e-06}
{"train_loss": 0.030190283432602882, "global_step": 227214, "epoch": 2552, "lr": 6.433571304912589e-06}
{"train_loss": 0.05129724368453026, "global_step": 227215, "epoch": 2552, "lr": 6.4332868310742866e-06}
{"train_loss": 0.05104760493915737, "global_step": 227216, "epoch": 2552, "lr": 6.4330023630929825e-06, "val_loss": 9.216880798339844}
{"train_loss": 0.0799359530210495, "global_step": 227217, "epoch": 2553, "lr": 6.4327179009687435e-06}
{"train_loss": 0.02434036321938038, "global_step": 227218, "epoch": 2553, "lr": 6.43243344470158e-06}
{"train_loss": 0.07354728877544403, "global_step": 227219, "epoch": 2553, "lr": 6.432148994291559e-06}
{"train_loss": 0.05008900538086891, "global_step": 227220, "epoch": 2553, "lr": 6.431864549738698e-06}
{"train_loss": 0.011087882332503796, "global_step": 227221, "epoch": 2553, "lr": 6.431580111043045e-06}
{"train_loss": 0.03441796451807022, "global_step": 227222, "epoch": 2553, "lr": 6.43129567820463e-06}
{"train_loss": 0.04962869733572006, "global_step": 227223, "epoch": 2553, "lr": 6.431011251223501e-06}
{"train_loss": 0.06806521862745285, "global_step": 227224, "epoch": 2553, "lr": 6.430726830099687e-06}
{"train_loss": 0.0390864834189415, "global_step": 227225, "epoch": 2553, "lr": 6.4304424148332375e-06}
{"train_loss": 0.031060149893164635, "global_step": 227226, "epoch": 2553, "lr": 6.430158005424175e-06}
{"train_loss": 0.06334783136844635, "global_step": 227227, "epoch": 2553, "lr": 6.429873601872554e-06}
{"train_loss": 0.05887515842914581, "global_step": 227228, "epoch": 2553, "lr": 6.429589204178399e-06}
{"train_loss": 0.0697748139500618, "global_step": 227229, "epoch": 2553, "lr": 6.429304812341763e-06}
{"train_loss": 0.05719061195850372, "global_step": 227230, "epoch": 2553, "lr": 6.429020426362664e-06}
{"train_loss": 0.03544598072767258, "global_step": 227231, "epoch": 2553, "lr": 6.4287360462411686e-06}
{"train_loss": 0.05138307064771652, "global_step": 227232, "epoch": 2553, "lr": 6.428451671977281e-06}
{"train_loss": 0.03296634182333946, "global_step": 227233, "epoch": 2553, "lr": 6.428167303571076e-06}
{"train_loss": 0.06405123323202133, "global_step": 227234, "epoch": 2553, "lr": 6.427882941022556e-06}
{"train_loss": 0.02725129947066307, "global_step": 227235, "epoch": 2553, "lr": 6.4275985843317735e-06}
{"train_loss": 0.06603217124938965, "global_step": 227236, "epoch": 2553, "lr": 6.427314233498788e-06}
{"train_loss": 0.0748746246099472, "global_step": 227237, "epoch": 2553, "lr": 6.4270298885236e-06}
{"train_loss": 0.0715101882815361, "global_step": 227238, "epoch": 2553, "lr": 6.4267455494062825e-06}
{"train_loss": 0.058004312217235565, "global_step": 227239, "epoch": 2553, "lr": 6.426461216146845e-06}
{"train_loss": 0.08807581663131714, "global_step": 227240, "epoch": 2553, "lr": 6.42617688874535e-06}
{"train_loss": 0.05782032385468483, "global_step": 227241, "epoch": 2553, "lr": 6.425892567201819e-06}
{"train_loss": 0.028325999155640602, "global_step": 227242, "epoch": 2553, "lr": 6.4256082515162966e-06}
{"train_loss": 0.07771793007850647, "global_step": 227243, "epoch": 2553, "lr": 6.425323941688815e-06}
{"train_loss": 0.06132298707962036, "global_step": 227244, "epoch": 2553, "lr": 6.425039637719427e-06}
{"train_loss": 0.05860118195414543, "global_step": 227245, "epoch": 2553, "lr": 6.424755339608152e-06}
{"train_loss": 0.020350849255919456, "global_step": 227246, "epoch": 2553, "lr": 6.424471047355046e-06}
{"train_loss": 0.033348601311445236, "global_step": 227247, "epoch": 2553, "lr": 6.424186760960133e-06}
{"train_loss": 0.07448231428861618, "global_step": 227248, "epoch": 2553, "lr": 6.423902480423466e-06}
{"train_loss": 0.10608456283807755, "global_step": 227249, "epoch": 2553, "lr": 6.4236182057450575e-06}
{"train_loss": 0.03737286850810051, "global_step": 227250, "epoch": 2553, "lr": 6.42333393692498e-06}
{"train_loss": 0.05574044957756996, "global_step": 227251, "epoch": 2553, "lr": 6.4230496739632495e-06}
{"train_loss": 0.025102321058511734, "global_step": 227252, "epoch": 2553, "lr": 6.422765416859899e-06}
{"train_loss": 0.04659653455018997, "global_step": 227253, "epoch": 2553, "lr": 6.422481165614991e-06}
{"train_loss": 0.04680338874459267, "global_step": 227254, "epoch": 2553, "lr": 6.4221969202285345e-06}
{"train_loss": 0.0478036031126976, "global_step": 227255, "epoch": 2553, "lr": 6.4219126807005984e-06}
{"train_loss": 0.03238803148269653, "global_step": 227256, "epoch": 2553, "lr": 6.421628447031186e-06}
{"train_loss": 0.06279774010181427, "global_step": 227257, "epoch": 2553, "lr": 6.421344219220371e-06}
{"train_loss": 0.05908302217721939, "global_step": 227258, "epoch": 2553, "lr": 6.421059997268164e-06}
{"train_loss": 0.0745113268494606, "global_step": 227259, "epoch": 2553, "lr": 6.420775781174626e-06}
{"train_loss": 0.04976063221693039, "global_step": 227260, "epoch": 2553, "lr": 6.420491570939768e-06}
{"train_loss": 0.07358638197183609, "global_step": 227261, "epoch": 2553, "lr": 6.420207366563658e-06}
{"train_loss": 0.045056071132421494, "global_step": 227262, "epoch": 2553, "lr": 6.41992316804631e-06}
{"train_loss": 0.08725271373987198, "global_step": 227263, "epoch": 2553, "lr": 6.419638975387782e-06}
{"train_loss": 0.024808188900351524, "global_step": 227264, "epoch": 2553, "lr": 6.419354788588089e-06}
{"train_loss": 0.0401649996638298, "global_step": 227265, "epoch": 2553, "lr": 6.419070607647298e-06}
{"train_loss": 0.0934881865978241, "global_step": 227266, "epoch": 2553, "lr": 6.418786432565416e-06}
{"train_loss": 0.021309392526745796, "global_step": 227267, "epoch": 2553, "lr": 6.418502263342513e-06}
{"train_loss": 0.04258895665407181, "global_step": 227268, "epoch": 2553, "lr": 6.418218099978595e-06}
{"train_loss": 0.07481794059276581, "global_step": 227269, "epoch": 2553, "lr": 6.417933942473731e-06}
{"train_loss": 0.018049979582428932, "global_step": 227270, "epoch": 2553, "lr": 6.417649790827935e-06}
{"train_loss": 0.02259794995188713, "global_step": 227271, "epoch": 2553, "lr": 6.417365645041251e-06}
{"train_loss": 0.016635984182357788, "global_step": 227272, "epoch": 2553, "lr": 6.417081505113731e-06}
{"train_loss": 0.0401061549782753, "global_step": 227273, "epoch": 2553, "lr": 6.416797371045397e-06}
{"train_loss": 0.04077596217393875, "global_step": 227274, "epoch": 2553, "lr": 6.416513242836303e-06}
{"train_loss": 0.0683048889040947, "global_step": 227275, "epoch": 2553, "lr": 6.4162291204864614e-06}
{"train_loss": 0.013711811043322086, "global_step": 227276, "epoch": 2553, "lr": 6.415945003995944e-06}
{"train_loss": 0.06239543855190277, "global_step": 227277, "epoch": 2553, "lr": 6.415660893364756e-06}
{"train_loss": 0.07417844980955124, "global_step": 227278, "epoch": 2553, "lr": 6.415376788592964e-06}
{"train_loss": 0.050010763108730316, "global_step": 227279, "epoch": 2553, "lr": 6.415092689680585e-06}
{"train_loss": 0.032553255558013916, "global_step": 227280, "epoch": 2553, "lr": 6.414808596627675e-06}
{"train_loss": 0.07903774827718735, "global_step": 227281, "epoch": 2553, "lr": 6.41452450943425e-06}
{"train_loss": 0.0342932865023613, "global_step": 227282, "epoch": 2553, "lr": 6.414240428100371e-06}
{"train_loss": 0.04177195206284523, "global_step": 227283, "epoch": 2553, "lr": 6.413956352626061e-06}
{"train_loss": 0.026095902547240257, "global_step": 227284, "epoch": 2553, "lr": 6.4136722830113685e-06}
{"train_loss": 0.11599700152873993, "global_step": 227285, "epoch": 2553, "lr": 6.413388219256316e-06}
{"train_loss": 0.031023789197206497, "global_step": 227286, "epoch": 2553, "lr": 6.413104161360966e-06}
{"train_loss": 0.05522218719124794, "global_step": 227287, "epoch": 2553, "lr": 6.412820109325329e-06}
{"train_loss": 0.0735011100769043, "global_step": 227288, "epoch": 2553, "lr": 6.412536063149471e-06}
{"train_loss": 0.09676346927881241, "global_step": 227289, "epoch": 2553, "lr": 6.4122520228334036e-06}
{"train_loss": 0.05511977896094322, "global_step": 227290, "epoch": 2553, "lr": 6.411967988377193e-06}
{"train_loss": 0.03338905796408653, "global_step": 227291, "epoch": 2553, "lr": 6.411683959780857e-06}
{"train_loss": 0.05188372731208801, "global_step": 227292, "epoch": 2553, "lr": 6.411399937044426e-06}
{"train_loss": 0.09080077707767487, "global_step": 227293, "epoch": 2553, "lr": 6.4111159201679646e-06}
{"train_loss": 0.06465011090040207, "global_step": 227294, "epoch": 2553, "lr": 6.410831909151488e-06}
{"train_loss": 0.03391987457871437, "global_step": 227295, "epoch": 2553, "lr": 6.410547903995057e-06}
{"train_loss": 0.04821290075778961, "global_step": 227296, "epoch": 2553, "lr": 6.4102639046986825e-06}
{"train_loss": 0.05035638064146042, "global_step": 227297, "epoch": 2553, "lr": 6.409979911262426e-06}
{"train_loss": 0.037455055862665176, "global_step": 227298, "epoch": 2553, "lr": 6.409695923686304e-06}
{"train_loss": 0.023752599954605103, "global_step": 227299, "epoch": 2553, "lr": 6.409411941970384e-06}
{"train_loss": 0.02580145001411438, "global_step": 227300, "epoch": 2553, "lr": 6.40912796611467e-06}
{"train_loss": 0.060274139046669006, "global_step": 227301, "epoch": 2553, "lr": 6.408843996119235e-06}
{"train_loss": 0.03456598147749901, "global_step": 227302, "epoch": 2553, "lr": 6.408560031984084e-06}
{"train_loss": 0.037055302411317825, "global_step": 227303, "epoch": 2553, "lr": 6.408276073709279e-06}
{"train_loss": 0.06253883987665176, "global_step": 227304, "epoch": 2553, "lr": 6.407992121294848e-06}
{"train_loss": 0.05195052662257398, "global_step": 227305, "epoch": 2553, "lr": 6.407708174740834e-06, "val_loss": 9.14415454864502}
{"train_loss": 0.06271369755268097, "global_step": 227306, "epoch": 2554, "lr": 6.407424234047266e-06}
{"train_loss": 0.017450476065278053, "global_step": 227307, "epoch": 2554, "lr": 6.407140299214187e-06}
{"train_loss": 0.06475189328193665, "global_step": 227308, "epoch": 2554, "lr": 6.406856370241649e-06}
{"train_loss": 0.012005094438791275, "global_step": 227309, "epoch": 2554, "lr": 6.406572447129661e-06}
{"train_loss": 0.09175237268209457, "global_step": 227310, "epoch": 2554, "lr": 6.406288529878296e-06}
{"train_loss": 0.030580095946788788, "global_step": 227311, "epoch": 2554, "lr": 6.4060046184875595e-06}
{"train_loss": 0.025772657245397568, "global_step": 227312, "epoch": 2554, "lr": 6.405720712957519e-06}
{"train_loss": 0.04825659468770027, "global_step": 227313, "epoch": 2554, "lr": 6.4054368132881845e-06}
{"train_loss": 0.0510181188583374, "global_step": 227314, "epoch": 2554, "lr": 6.405152919479618e-06}
{"train_loss": 0.07568109780550003, "global_step": 227315, "epoch": 2554, "lr": 6.404869031531835e-06}
{"train_loss": 0.016504867002367973, "global_step": 227316, "epoch": 2554, "lr": 6.404585149444897e-06}
{"train_loss": 0.06944151222705841, "global_step": 227317, "epoch": 2554, "lr": 6.404301273218826e-06}
{"train_loss": 0.030269676819443703, "global_step": 227318, "epoch": 2554, "lr": 6.404017402853668e-06}
{"train_loss": 0.014175130054354668, "global_step": 227319, "epoch": 2554, "lr": 6.403733538349455e-06}
{"train_loss": 0.04672125726938248, "global_step": 227320, "epoch": 2554, "lr": 6.403449679706236e-06}
{"train_loss": 0.0736674815416336, "global_step": 227321, "epoch": 2554, "lr": 6.4031658269240345e-06}
{"train_loss": 0.03183398395776749, "global_step": 227322, "epoch": 2554, "lr": 6.402881980002901e-06}
{"train_loss": 0.047569639980793, "global_step": 227323, "epoch": 2554, "lr": 6.402598138942861e-06}
{"train_loss": 0.06542247533798218, "global_step": 227324, "epoch": 2554, "lr": 6.402314303743972e-06}
{"train_loss": 0.03513963520526886, "global_step": 227325, "epoch": 2554, "lr": 6.4020304744062456e-06}
{"train_loss": 0.03085232898592949, "global_step": 227326, "epoch": 2554, "lr": 6.401746650929752e-06}
{"train_loss": 0.034699223935604095, "global_step": 227327, "epoch": 2554, "lr": 6.401462833314498e-06}
{"train_loss": 0.1143876388669014, "global_step": 227328, "epoch": 2554, "lr": 6.401179021560544e-06}
{"train_loss": 0.02522888407111168, "global_step": 227329, "epoch": 2554, "lr": 6.400895215667924e-06}
{"train_loss": 0.07107046991586685, "global_step": 227330, "epoch": 2554, "lr": 6.400611415636659e-06}
{"train_loss": 0.02159246988594532, "global_step": 227331, "epoch": 2554, "lr": 6.400327621466811e-06}
{"train_loss": 0.028790641576051712, "global_step": 227332, "epoch": 2554, "lr": 6.4000438331583975e-06}
{"train_loss": 0.034812211990356445, "global_step": 227333, "epoch": 2554, "lr": 6.399760050711479e-06}
{"train_loss": 0.04947027191519737, "global_step": 227334, "epoch": 2554, "lr": 6.39947627412607e-06}
{"train_loss": 0.032020557671785355, "global_step": 227335, "epoch": 2554, "lr": 6.399192503402229e-06}
{"train_loss": 0.058798499405384064, "global_step": 227336, "epoch": 2554, "lr": 6.398908738539977e-06}
{"train_loss": 0.05398406833410263, "global_step": 227337, "epoch": 2554, "lr": 6.3986249795393695e-06}
{"train_loss": 0.03508232533931732, "global_step": 227338, "epoch": 2554, "lr": 6.3983412264004285e-06}
{"train_loss": 0.0948481410741806, "global_step": 227339, "epoch": 2554, "lr": 6.398057479123204e-06}
{"train_loss": 0.02926691807806492, "global_step": 227340, "epoch": 2554, "lr": 6.397773737707719e-06}
{"train_loss": 0.04590728506445885, "global_step": 227341, "epoch": 2554, "lr": 6.397490002154038e-06}
{"train_loss": 0.06234545633196831, "global_step": 227342, "epoch": 2554, "lr": 6.397206272462164e-06}
{"train_loss": 0.04257466644048691, "global_step": 227343, "epoch": 2554, "lr": 6.396922548632162e-06}
{"train_loss": 0.04679720103740692, "global_step": 227344, "epoch": 2554, "lr": 6.39663883066407e-06}
{"train_loss": 0.05161352455615997, "global_step": 227345, "epoch": 2554, "lr": 6.396355118557906e-06}
{"train_loss": 0.05442139878869057, "global_step": 227346, "epoch": 2554, "lr": 6.3960714123137365e-06}
{"train_loss": 0.0746280774474144, "global_step": 227347, "epoch": 2554, "lr": 6.395787711931567e-06}
{"train_loss": 0.07808645814657211, "global_step": 227348, "epoch": 2554, "lr": 6.395504017411464e-06}
{"train_loss": 0.04753590747714043, "global_step": 227349, "epoch": 2554, "lr": 6.395220328753449e-06}
{"train_loss": 0.05373954400420189, "global_step": 227350, "epoch": 2554, "lr": 6.394936645957572e-06}
{"train_loss": 0.05952637642621994, "global_step": 227351, "epoch": 2554, "lr": 6.394652969023851e-06}
{"train_loss": 0.04778193682432175, "global_step": 227352, "epoch": 2554, "lr": 6.394369297952351e-06}
{"train_loss": 0.09500822424888611, "global_step": 227353, "epoch": 2554, "lr": 6.394085632743086e-06}
{"train_loss": 0.06034209951758385, "global_step": 227354, "epoch": 2554, "lr": 6.393801973396113e-06}
{"train_loss": 0.025995030999183655, "global_step": 227355, "epoch": 2554, "lr": 6.393518319911451e-06}
{"train_loss": 0.0645923987030983, "global_step": 227356, "epoch": 2554, "lr": 6.3932346722891665e-06}
{"train_loss": 0.056410882622003555, "global_step": 227357, "epoch": 2554, "lr": 6.392951030529259e-06}
{"train_loss": 0.07452819496393204, "global_step": 227358, "epoch": 2554, "lr": 6.392667394631807e-06}
{"train_loss": 0.06425191462039948, "global_step": 227359, "epoch": 2554, "lr": 6.392383764596815e-06}
{"train_loss": 0.010189137421548367, "global_step": 227360, "epoch": 2554, "lr": 6.392100140424345e-06}
{"train_loss": 0.04171338677406311, "global_step": 227361, "epoch": 2554, "lr": 6.391816522114419e-06}
{"train_loss": 0.06253838539123535, "global_step": 227362, "epoch": 2554, "lr": 6.391532909667091e-06}
{"train_loss": 0.038620881736278534, "global_step": 227363, "epoch": 2554, "lr": 6.3912493030823804e-06}
{"train_loss": 0.051761481910943985, "global_step": 227364, "epoch": 2554, "lr": 6.390965702360341e-06}
{"train_loss": 0.04069634526968002, "global_step": 227365, "epoch": 2554, "lr": 6.390682107500995e-06}
{"train_loss": 0.05421718582510948, "global_step": 227366, "epoch": 2554, "lr": 6.390398518504404e-06}
{"train_loss": 0.07864763587713242, "global_step": 227367, "epoch": 2554, "lr": 6.390114935370589e-06}
{"train_loss": 0.09056319296360016, "global_step": 227368, "epoch": 2554, "lr": 6.389831358099585e-06}
{"train_loss": 0.06883851438760757, "global_step": 227369, "epoch": 2554, "lr": 6.3895477866914464e-06}
{"train_loss": 0.048139967024326324, "global_step": 227370, "epoch": 2554, "lr": 6.389264221146191e-06}
{"train_loss": 0.05986015498638153, "global_step": 227371, "epoch": 2554, "lr": 6.388980661463873e-06}
{"train_loss": 0.1321176141500473, "global_step": 227372, "epoch": 2554, "lr": 6.38869710764452e-06}
{"train_loss": 0.052074361592531204, "global_step": 227373, "epoch": 2554, "lr": 6.388413559688183e-06}
{"train_loss": 0.046165820211172104, "global_step": 227374, "epoch": 2554, "lr": 6.388130017594885e-06}
{"train_loss": 0.03821582347154617, "global_step": 227375, "epoch": 2554, "lr": 6.387846481364679e-06}
{"train_loss": 0.04930742457509041, "global_step": 227376, "epoch": 2554, "lr": 6.387562950997583e-06}
{"train_loss": 0.04911484569311142, "global_step": 227377, "epoch": 2554, "lr": 6.387279426493659e-06}
{"train_loss": 0.06361331045627594, "global_step": 227378, "epoch": 2554, "lr": 6.386995907852927e-06}
{"train_loss": 0.1267140656709671, "global_step": 227379, "epoch": 2554, "lr": 6.38671239507544e-06}
{"train_loss": 0.0730021595954895, "global_step": 227380, "epoch": 2554, "lr": 6.386428888161217e-06}
{"train_loss": 0.051031794399023056, "global_step": 227381, "epoch": 2554, "lr": 6.386145387110309e-06}
{"train_loss": 0.07102886587381363, "global_step": 227382, "epoch": 2554, "lr": 6.385861891922762e-06}
{"train_loss": 0.05678213760256767, "global_step": 227383, "epoch": 2554, "lr": 6.385578402598591e-06}
{"train_loss": 0.07041089981794357, "global_step": 227384, "epoch": 2554, "lr": 6.3852949191378576e-06}
{"train_loss": 0.04948192462325096, "global_step": 227385, "epoch": 2554, "lr": 6.385011441540583e-06}
{"train_loss": 0.07856301218271255, "global_step": 227386, "epoch": 2554, "lr": 6.384727969806819e-06}
{"train_loss": 0.05491853505373001, "global_step": 227387, "epoch": 2554, "lr": 6.384444503936588e-06}
{"train_loss": 0.07032361626625061, "global_step": 227388, "epoch": 2554, "lr": 6.384161043929948e-06}
{"train_loss": 0.046310458332300186, "global_step": 227389, "epoch": 2554, "lr": 6.383877589786919e-06}
{"train_loss": 0.048629239201545715, "global_step": 227390, "epoch": 2554, "lr": 6.383594141507548e-06}
{"train_loss": 0.07013465464115143, "global_step": 227391, "epoch": 2554, "lr": 6.383310699091865e-06}
{"train_loss": 0.02308053709566593, "global_step": 227392, "epoch": 2554, "lr": 6.383027262539926e-06}
{"train_loss": 0.06257862597703934, "global_step": 227393, "epoch": 2554, "lr": 6.382743831851745e-06}
{"train_loss": 0.053751331565671426, "global_step": 227394, "epoch": 2554, "lr": 6.382460407027385e-06, "val_loss": 9.169899940490723}
{"train_loss": 0.04941694065928459, "global_step": 227395, "epoch": 2555, "lr": 6.382176988066863e-06}
{"train_loss": 0.04426936060190201, "global_step": 227396, "epoch": 2555, "lr": 6.381893574970233e-06}
{"train_loss": 0.0343346931040287, "global_step": 227397, "epoch": 2555, "lr": 6.381610167737512e-06}
{"train_loss": 0.024228369817137718, "global_step": 227398, "epoch": 2555, "lr": 6.381326766368756e-06}
{"train_loss": 0.05648520588874817, "global_step": 227399, "epoch": 2555, "lr": 6.381043370864009e-06}
{"train_loss": 0.019505197182297707, "global_step": 227400, "epoch": 2555, "lr": 6.3807599812233e-06}
{"train_loss": 0.06135183200240135, "global_step": 227401, "epoch": 2555, "lr": 6.3804765974466556e-06}
{"train_loss": 0.04904378205537796, "global_step": 227402, "epoch": 2555, "lr": 6.380193219534131e-06}
{"train_loss": 0.042796842753887177, "global_step": 227403, "epoch": 2555, "lr": 6.379909847485749e-06}
{"train_loss": 0.044781044125556946, "global_step": 227404, "epoch": 2555, "lr": 6.3796264813015705e-06}
{"train_loss": 0.028537692502141, "global_step": 227405, "epoch": 2555, "lr": 6.379343120981607e-06}
{"train_loss": 0.04387292638421059, "global_step": 227406, "epoch": 2555, "lr": 6.379059766525919e-06}
{"train_loss": 0.021352659910917282, "global_step": 227407, "epoch": 2555, "lr": 6.378776417934534e-06}
{"train_loss": 0.050397831946611404, "global_step": 227408, "epoch": 2555, "lr": 6.378493075207481e-06}
{"train_loss": 0.07556316256523132, "global_step": 227409, "epoch": 2555, "lr": 6.37820973834482e-06}
{"train_loss": 0.056426532566547394, "global_step": 227410, "epoch": 2555, "lr": 6.377926407346563e-06}
{"train_loss": 0.029988326132297516, "global_step": 227411, "epoch": 2555, "lr": 6.3776430822127755e-06}
{"train_loss": 0.041079338639974594, "global_step": 227412, "epoch": 2555, "lr": 6.377359762943474e-06}
{"train_loss": 0.04855024442076683, "global_step": 227413, "epoch": 2555, "lr": 6.37707644953871e-06}
{"train_loss": 0.07396695762872696, "global_step": 227414, "epoch": 2555, "lr": 6.3767931419985105e-06}
{"train_loss": 0.06351752579212189, "global_step": 227415, "epoch": 2555, "lr": 6.376509840322925e-06}
{"train_loss": 0.05447094887495041, "global_step": 227416, "epoch": 2555, "lr": 6.376226544511982e-06}
{"train_loss": 0.05531291663646698, "global_step": 227417, "epoch": 2555, "lr": 6.3759432545657195e-06}
{"train_loss": 0.04006807133555412, "global_step": 227418, "epoch": 2555, "lr": 6.375659970484188e-06}
{"train_loss": 0.05215891823172569, "global_step": 227419, "epoch": 2555, "lr": 6.37537669226741e-06}
{"train_loss": 0.09661168605089188, "global_step": 227420, "epoch": 2555, "lr": 6.37509341991544e-06}
{"train_loss": 0.0548098161816597, "global_step": 227421, "epoch": 2555, "lr": 6.374810153428301e-06}
{"train_loss": 0.03974644094705582, "global_step": 227422, "epoch": 2555, "lr": 6.374526892806043e-06}
{"train_loss": 0.03551216050982475, "global_step": 227423, "epoch": 2555, "lr": 6.374243638048688e-06}
{"train_loss": 0.07181674987077713, "global_step": 227424, "epoch": 2555, "lr": 6.373960389156297e-06}
{"train_loss": 0.01373350527137518, "global_step": 227425, "epoch": 2555, "lr": 6.3736771461288815e-06}
{"train_loss": 0.04294479265809059, "global_step": 227426, "epoch": 2555, "lr": 6.373393908966507e-06}
{"train_loss": 0.08032487332820892, "global_step": 227427, "epoch": 2555, "lr": 6.373110677669186e-06}
{"train_loss": 0.09186310321092606, "global_step": 227428, "epoch": 2555, "lr": 6.372827452236979e-06}
{"train_loss": 0.05622642859816551, "global_step": 227429, "epoch": 2555, "lr": 6.372544232669902e-06}
{"train_loss": 0.03525698930025101, "global_step": 227430, "epoch": 2555, "lr": 6.372261018968018e-06}
{"train_loss": 0.053572747856378555, "global_step": 227431, "epoch": 2555, "lr": 6.371977811131335e-06}
{"train_loss": 0.026360508054494858, "global_step": 227432, "epoch": 2555, "lr": 6.371694609159928e-06}
{"train_loss": 0.03857571631669998, "global_step": 227433, "epoch": 2555, "lr": 6.371411413053796e-06}
{"train_loss": 0.05130024626851082, "global_step": 227434, "epoch": 2555, "lr": 6.371128222813e-06}
{"train_loss": 0.02995682694017887, "global_step": 227435, "epoch": 2555, "lr": 6.370845038437584e-06}
{"train_loss": 0.040033482015132904, "global_step": 227436, "epoch": 2555, "lr": 6.370561859927565e-06}
{"train_loss": 0.052198536694049835, "global_step": 227437, "epoch": 2555, "lr": 6.370278687283005e-06}
{"train_loss": 0.02881799452006817, "global_step": 227438, "epoch": 2555, "lr": 6.369995520503913e-06}
{"train_loss": 0.019202126190066338, "global_step": 227439, "epoch": 2555, "lr": 6.369712359590357e-06}
{"train_loss": 0.04972515255212784, "global_step": 227440, "epoch": 2555, "lr": 6.369429204542365e-06}
{"train_loss": 0.05540815740823746, "global_step": 227441, "epoch": 2555, "lr": 6.369146055359959e-06}
{"train_loss": 0.03028492070734501, "global_step": 227442, "epoch": 2555, "lr": 6.3688629120432e-06}
{"train_loss": 0.038488298654556274, "global_step": 227443, "epoch": 2555, "lr": 6.368579774592104e-06}
{"train_loss": 0.06866610050201416, "global_step": 227444, "epoch": 2555, "lr": 6.3682966430067325e-06}
{"train_loss": 0.007986356504261494, "global_step": 227445, "epoch": 2555, "lr": 6.368013517287108e-06}
{"train_loss": 0.037102751433849335, "global_step": 227446, "epoch": 2555, "lr": 6.367730397433263e-06}
{"train_loss": 0.021331118419766426, "global_step": 227447, "epoch": 2555, "lr": 6.3674472834452606e-06}
{"train_loss": 0.05535442382097244, "global_step": 227448, "epoch": 2555, "lr": 6.367164175323104e-06}
{"train_loss": 0.06921834498643875, "global_step": 227449, "epoch": 2555, "lr": 6.366881073066866e-06}
{"train_loss": 0.022011324763298035, "global_step": 227450, "epoch": 2555, "lr": 6.366597976676558e-06}
{"train_loss": 0.027806293219327927, "global_step": 227451, "epoch": 2555, "lr": 6.366314886152242e-06}
{"train_loss": 0.04400511085987091, "global_step": 227452, "epoch": 2555, "lr": 6.366031801493932e-06}
{"train_loss": 0.044736456125974655, "global_step": 227453, "epoch": 2555, "lr": 6.3657487227016755e-06}
{"train_loss": 0.06387543678283691, "global_step": 227454, "epoch": 2555, "lr": 6.36546564977552e-06}
{"train_loss": 0.03535373508930206, "global_step": 227455, "epoch": 2555, "lr": 6.36518258271549e-06}
{"train_loss": 0.06673324853181839, "global_step": 227456, "epoch": 2555, "lr": 6.3648995215216386e-06}
{"train_loss": 0.08430007100105286, "global_step": 227457, "epoch": 2555, "lr": 6.364616466193979e-06}
{"train_loss": 0.029833493754267693, "global_step": 227458, "epoch": 2555, "lr": 6.364333416732582e-06}
{"train_loss": 0.01980111002922058, "global_step": 227459, "epoch": 2555, "lr": 6.364050373137453e-06}
{"train_loss": 0.04006164148449898, "global_step": 227460, "epoch": 2555, "lr": 6.36376733540866e-06}
{"train_loss": 0.0530187226831913, "global_step": 227461, "epoch": 2555, "lr": 6.363484303546219e-06}
{"train_loss": 0.03832916542887688, "global_step": 227462, "epoch": 2555, "lr": 6.363201277550179e-06}
{"train_loss": 0.05109521746635437, "global_step": 227463, "epoch": 2555, "lr": 6.362918257420569e-06}
{"train_loss": 0.04861777275800705, "global_step": 227464, "epoch": 2555, "lr": 6.362635243157445e-06}
{"train_loss": 0.0963929072022438, "global_step": 227465, "epoch": 2555, "lr": 6.362352234760816e-06}
{"train_loss": 0.05015117675065994, "global_step": 227466, "epoch": 2555, "lr": 6.362069232230749e-06}
{"train_loss": 0.019763527438044548, "global_step": 227467, "epoch": 2555, "lr": 6.361786235567263e-06}
{"train_loss": 0.036075159907341, "global_step": 227468, "epoch": 2555, "lr": 6.3615032447704115e-06}
{"train_loss": 0.056193068623542786, "global_step": 227469, "epoch": 2555, "lr": 6.361220259840212e-06}
{"train_loss": 0.0638376995921135, "global_step": 227470, "epoch": 2555, "lr": 6.360937280776718e-06}
{"train_loss": 0.09256535768508911, "global_step": 227471, "epoch": 2555, "lr": 6.360654307579972e-06}
{"train_loss": 0.05242427438497543, "global_step": 227472, "epoch": 2555, "lr": 6.360371340249993e-06}
{"train_loss": 0.04102647304534912, "global_step": 227473, "epoch": 2555, "lr": 6.360088378786844e-06}
{"train_loss": 0.05902206897735596, "global_step": 227474, "epoch": 2555, "lr": 6.3598054231905345e-06}
{"train_loss": 0.0633145272731781, "global_step": 227475, "epoch": 2555, "lr": 6.359522473461133e-06}
{"train_loss": 0.040518537163734436, "global_step": 227476, "epoch": 2555, "lr": 6.359239529598643e-06}
{"train_loss": 0.04152047261595726, "global_step": 227477, "epoch": 2555, "lr": 6.358956591603138e-06}
{"train_loss": 0.04469522088766098, "global_step": 227478, "epoch": 2555, "lr": 6.35867365947464e-06}
{"train_loss": 0.05683356150984764, "global_step": 227479, "epoch": 2555, "lr": 6.358390733213176e-06}
{"train_loss": 0.012602644972503185, "global_step": 227480, "epoch": 2555, "lr": 6.358107812818803e-06}
{"train_loss": 0.01809481345117092, "global_step": 227481, "epoch": 2555, "lr": 6.357824898291537e-06}
{"train_loss": 0.0475437305867672, "global_step": 227482, "epoch": 2555, "lr": 6.357541989631444e-06}
{"train_loss": 0.04655380694593272, "global_step": 227483, "epoch": 2555, "lr": 6.357259086838546e-06, "val_loss": 9.2089204788208, "train_action_mse_error": 7.968039512634277}
{"train_loss": 0.04373841732740402, "global_step": 227484, "epoch": 2556, "lr": 6.356976189912872e-06}
{"train_loss": 0.0671810731291771, "global_step": 227485, "epoch": 2556, "lr": 6.356693298854482e-06}
{"train_loss": 0.02530129998922348, "global_step": 227486, "epoch": 2556, "lr": 6.356410413663394e-06}
{"train_loss": 0.052571482956409454, "global_step": 227487, "epoch": 2556, "lr": 6.356127534339662e-06}
{"train_loss": 0.06745727360248566, "global_step": 227488, "epoch": 2556, "lr": 6.355844660883309e-06}
{"train_loss": 0.03916402533650398, "global_step": 227489, "epoch": 2556, "lr": 6.355561793294379e-06}
{"train_loss": 0.08590222895145416, "global_step": 227490, "epoch": 2556, "lr": 6.355278931572922e-06}
{"train_loss": 0.07139493525028229, "global_step": 227491, "epoch": 2556, "lr": 6.354996075718955e-06}
{"train_loss": 0.03992646932601929, "global_step": 227492, "epoch": 2556, "lr": 6.354713225732539e-06}
{"train_loss": 0.04208141192793846, "global_step": 227493, "epoch": 2556, "lr": 6.354430381613691e-06}
{"train_loss": 0.04026052728295326, "global_step": 227494, "epoch": 2556, "lr": 6.354147543362465e-06}
{"train_loss": 0.06091873347759247, "global_step": 227495, "epoch": 2556, "lr": 6.35386471097888e-06}
{"train_loss": 0.03438844531774521, "global_step": 227496, "epoch": 2556, "lr": 6.3535818844630005e-06}
{"train_loss": 0.09465551376342773, "global_step": 227497, "epoch": 2556, "lr": 6.353299063814838e-06}
{"train_loss": 0.05657311528921127, "global_step": 227498, "epoch": 2556, "lr": 6.353016249034449e-06}
{"train_loss": 0.05616970732808113, "global_step": 227499, "epoch": 2556, "lr": 6.3527334401218604e-06}
{"train_loss": 0.057553768157958984, "global_step": 227500, "epoch": 2556, "lr": 6.352450637077123e-06}
{"train_loss": 0.05253127962350845, "global_step": 227501, "epoch": 2556, "lr": 6.352167839900258e-06}
{"train_loss": 0.022153140977025032, "global_step": 227502, "epoch": 2556, "lr": 6.351885048591316e-06}
{"train_loss": 0.0633283406496048, "global_step": 227503, "epoch": 2556, "lr": 6.35160226315033e-06}
{"train_loss": 0.06400005519390106, "global_step": 227504, "epoch": 2556, "lr": 6.3513194835773446e-06}
{"train_loss": 0.03532752767205238, "global_step": 227505, "epoch": 2556, "lr": 6.351036709872382e-06}
{"train_loss": 0.04021581634879112, "global_step": 227506, "epoch": 2556, "lr": 6.350753942035498e-06}
{"train_loss": 0.04167158901691437, "global_step": 227507, "epoch": 2556, "lr": 6.350471180066714e-06}
{"train_loss": 0.03242240101099014, "global_step": 227508, "epoch": 2556, "lr": 6.35018842396608e-06}
{"train_loss": 0.03990078717470169, "global_step": 227509, "epoch": 2556, "lr": 6.349905673733641e-06}
{"train_loss": 0.04871491715312004, "global_step": 227510, "epoch": 2556, "lr": 6.349622929369415e-06}
{"train_loss": 0.05109290033578873, "global_step": 227511, "epoch": 2556, "lr": 6.349340190873459e-06}
{"train_loss": 0.08079987019300461, "global_step": 227512, "epoch": 2556, "lr": 6.349057458245794e-06}
{"train_loss": 0.03190804645419121, "global_step": 227513, "epoch": 2556, "lr": 6.348774731486473e-06}
{"train_loss": 0.07035017013549805, "global_step": 227514, "epoch": 2556, "lr": 6.348492010595519e-06}
{"train_loss": 0.07724263519048691, "global_step": 227515, "epoch": 2556, "lr": 6.348209295572987e-06}
{"train_loss": 0.024203652516007423, "global_step": 227516, "epoch": 2556, "lr": 6.347926586418912e-06}
{"train_loss": 0.016363264992833138, "global_step": 227517, "epoch": 2556, "lr": 6.347643883133309e-06}
{"train_loss": 0.06101377308368683, "global_step": 227518, "epoch": 2556, "lr": 6.347361185716244e-06}
{"train_loss": 0.02583233080804348, "global_step": 227519, "epoch": 2556, "lr": 6.347078494167735e-06}
{"train_loss": 0.0936739519238472, "global_step": 227520, "epoch": 2556, "lr": 6.3467958084878435e-06}
{"train_loss": 0.026400193572044373, "global_step": 227521, "epoch": 2556, "lr": 6.346513128676579e-06}
{"train_loss": 0.03885144740343094, "global_step": 227522, "epoch": 2556, "lr": 6.346230454734009e-06}
{"train_loss": 0.0294368714094162, "global_step": 227523, "epoch": 2556, "lr": 6.34594778666015e-06}
{"train_loss": 0.09633594006299973, "global_step": 227524, "epoch": 2556, "lr": 6.345665124455036e-06}
{"train_loss": 0.04479505494236946, "global_step": 227525, "epoch": 2556, "lr": 6.345382468118722e-06}
{"train_loss": 0.05072887986898422, "global_step": 227526, "epoch": 2556, "lr": 6.34509981765124e-06}
{"train_loss": 0.07275174558162689, "global_step": 227527, "epoch": 2556, "lr": 6.3448171730526255e-06}
{"train_loss": 0.021745841950178146, "global_step": 227528, "epoch": 2556, "lr": 6.344534534322921e-06}
{"train_loss": 0.052998136729002, "global_step": 227529, "epoch": 2556, "lr": 6.344251901462156e-06}
{"train_loss": 0.05348675325512886, "global_step": 227530, "epoch": 2556, "lr": 6.343969274470385e-06}
{"train_loss": 0.057390306144952774, "global_step": 227531, "epoch": 2556, "lr": 6.3436866533476245e-06}
{"train_loss": 0.04449520632624626, "global_step": 227532, "epoch": 2556, "lr": 6.3434040380939365e-06}
{"train_loss": 0.07462609559297562, "global_step": 227533, "epoch": 2556, "lr": 6.343121428709331e-06}
{"train_loss": 0.05805503949522972, "global_step": 227534, "epoch": 2556, "lr": 6.34283882519387e-06}
{"train_loss": 0.032423682510852814, "global_step": 227535, "epoch": 2556, "lr": 6.342556227547574e-06}
{"train_loss": 0.04071332886815071, "global_step": 227536, "epoch": 2556, "lr": 6.342273635770501e-06}
{"train_loss": 0.08330903947353363, "global_step": 227537, "epoch": 2556, "lr": 6.341991049862661e-06}
{"train_loss": 0.03827764838933945, "global_step": 227538, "epoch": 2556, "lr": 6.341708469824126e-06}
{"train_loss": 0.04062870517373085, "global_step": 227539, "epoch": 2556, "lr": 6.341425895654901e-06}
{"train_loss": 0.03638594597578049, "global_step": 227540, "epoch": 2556, "lr": 6.341143327355054e-06}
{"train_loss": 0.05600971356034279, "global_step": 227541, "epoch": 2556, "lr": 6.340860764924594e-06}
{"train_loss": 0.08034385740756989, "global_step": 227542, "epoch": 2556, "lr": 6.340578208363584e-06}
{"train_loss": 0.03470749408006668, "global_step": 227543, "epoch": 2556, "lr": 6.340295657672041e-06}
{"train_loss": 0.04902900010347366, "global_step": 227544, "epoch": 2556, "lr": 6.340013112850013e-06}
{"train_loss": 0.06288168579339981, "global_step": 227545, "epoch": 2556, "lr": 6.339730573897551e-06}
{"train_loss": 0.06640180945396423, "global_step": 227546, "epoch": 2556, "lr": 6.3394480408146666e-06}
{"train_loss": 0.0335046760737896, "global_step": 227547, "epoch": 2556, "lr": 6.33916551360142e-06}
{"train_loss": 0.09208815544843674, "global_step": 227548, "epoch": 2556, "lr": 6.338882992257833e-06}
{"train_loss": 0.018105223774909973, "global_step": 227549, "epoch": 2556, "lr": 6.338600476783962e-06}
{"train_loss": 0.06188651919364929, "global_step": 227550, "epoch": 2556, "lr": 6.338317967179825e-06}
{"train_loss": 0.015033320523798466, "global_step": 227551, "epoch": 2556, "lr": 6.338035463445474e-06}
{"train_loss": 0.027930129319429398, "global_step": 227552, "epoch": 2556, "lr": 6.337752965580934e-06}
{"train_loss": 0.06729979068040848, "global_step": 227553, "epoch": 2556, "lr": 6.33747047358626e-06}
{"train_loss": 0.03653741255402565, "global_step": 227554, "epoch": 2556, "lr": 6.3371879874614735e-06}
{"train_loss": 0.08464015275239944, "global_step": 227555, "epoch": 2556, "lr": 6.336905507206631e-06}
{"train_loss": 0.05380859225988388, "global_step": 227556, "epoch": 2556, "lr": 6.336623032821759e-06}
{"train_loss": 0.030683238059282303, "global_step": 227557, "epoch": 2556, "lr": 6.336340564306881e-06}
{"train_loss": 0.06682512164115906, "global_step": 227558, "epoch": 2556, "lr": 6.336058101662063e-06}
{"train_loss": 0.06006767600774765, "global_step": 227559, "epoch": 2556, "lr": 6.335775644887321e-06}
{"train_loss": 0.05728122591972351, "global_step": 227560, "epoch": 2556, "lr": 6.3354931939827115e-06}
{"train_loss": 0.013312555849552155, "global_step": 227561, "epoch": 2556, "lr": 6.33521074894825e-06}
{"train_loss": 0.06418155133724213, "global_step": 227562, "epoch": 2556, "lr": 6.3349283097839995e-06}
{"train_loss": 0.027892254292964935, "global_step": 227563, "epoch": 2556, "lr": 6.334645876489975e-06}
{"train_loss": 0.06210826337337494, "global_step": 227564, "epoch": 2556, "lr": 6.3343634490662386e-06}
{"train_loss": 0.08096209168434143, "global_step": 227565, "epoch": 2556, "lr": 6.3340810275128005e-06}
{"train_loss": 0.12417567521333694, "global_step": 227566, "epoch": 2556, "lr": 6.3337986118297225e-06}
{"train_loss": 0.040087539702653885, "global_step": 227567, "epoch": 2556, "lr": 6.333516202017026e-06}
{"train_loss": 0.0366632454097271, "global_step": 227568, "epoch": 2556, "lr": 6.3332337980747665e-06}
{"train_loss": 0.04072965309023857, "global_step": 227569, "epoch": 2556, "lr": 6.332951400002956e-06}
{"train_loss": 0.04176466539502144, "global_step": 227570, "epoch": 2556, "lr": 6.332669007801662e-06}
{"train_loss": 0.044469866901636124, "global_step": 227571, "epoch": 2556, "lr": 6.332386621470898e-06}
{"train_loss": 0.05115456739963775, "global_step": 227572, "epoch": 2556, "lr": 6.332104241010722e-06, "val_loss": 9.07911491394043}
{"train_loss": 0.01671447418630123, "global_step": 227573, "epoch": 2557, "lr": 6.331821866421156e-06}
{"train_loss": 0.0403551310300827, "global_step": 227574, "epoch": 2557, "lr": 6.3315394977022495e-06}
{"train_loss": 0.014108477160334587, "global_step": 227575, "epoch": 2557, "lr": 6.33125713485403e-06}
{"train_loss": 0.0430021733045578, "global_step": 227576, "epoch": 2557, "lr": 6.330974777876547e-06}
{"train_loss": 0.05865205451846123, "global_step": 227577, "epoch": 2557, "lr": 6.330692426769819e-06}
{"train_loss": 0.03561440855264664, "global_step": 227578, "epoch": 2557, "lr": 6.33041008153391e-06}
{"train_loss": 0.05704507604241371, "global_step": 227579, "epoch": 2557, "lr": 6.33012774216884e-06}
{"train_loss": 0.0599348321557045, "global_step": 227580, "epoch": 2557, "lr": 6.329845408674645e-06}
{"train_loss": 0.03860975056886673, "global_step": 227581, "epoch": 2557, "lr": 6.3295630810513875e-06}
{"train_loss": 0.04029890149831772, "global_step": 227582, "epoch": 2557, "lr": 6.329280759299072e-06}
{"train_loss": 0.04740866273641586, "global_step": 227583, "epoch": 2557, "lr": 6.32899844341776e-06}
{"train_loss": 0.04841335862874985, "global_step": 227584, "epoch": 2557, "lr": 6.3287161334074805e-06}
{"train_loss": 0.0581502690911293, "global_step": 227585, "epoch": 2557, "lr": 6.3284338292682766e-06}
{"train_loss": 0.05330986529588699, "global_step": 227586, "epoch": 2557, "lr": 6.328151531000176e-06}
{"train_loss": 0.08679287880659103, "global_step": 227587, "epoch": 2557, "lr": 6.327869238603229e-06}
{"train_loss": 0.04493856430053711, "global_step": 227588, "epoch": 2557, "lr": 6.327586952077463e-06}
{"train_loss": 0.027769774198532104, "global_step": 227589, "epoch": 2557, "lr": 6.327304671422924e-06}
{"train_loss": 0.17325645685195923, "global_step": 227590, "epoch": 2557, "lr": 6.327022396639642e-06}
{"train_loss": 0.05871821939945221, "global_step": 227591, "epoch": 2557, "lr": 6.32674012772767e-06}
{"train_loss": 0.030274739488959312, "global_step": 227592, "epoch": 2557, "lr": 6.326457864687019e-06}
{"train_loss": 0.010498268529772758, "global_step": 227593, "epoch": 2557, "lr": 6.326175607517759e-06}
{"train_loss": 0.10786957293748856, "global_step": 227594, "epoch": 2557, "lr": 6.325893356219914e-06}
{"train_loss": 0.094986692070961, "global_step": 227595, "epoch": 2557, "lr": 6.3256111107935055e-06}
{"train_loss": 0.02655475214123726, "global_step": 227596, "epoch": 2557, "lr": 6.3253288712386e-06}
{"train_loss": 0.061313461512327194, "global_step": 227597, "epoch": 2557, "lr": 6.32504663755521e-06}
{"train_loss": 0.04553470388054848, "global_step": 227598, "epoch": 2557, "lr": 6.324764409743394e-06}
{"train_loss": 0.04928663745522499, "global_step": 227599, "epoch": 2557, "lr": 6.324482187803171e-06}
{"train_loss": 0.07833106815814972, "global_step": 227600, "epoch": 2557, "lr": 6.3241999717346016e-06}
{"train_loss": 0.04990607872605324, "global_step": 227601, "epoch": 2557, "lr": 6.323917761537701e-06}
{"train_loss": 0.04586464911699295, "global_step": 227602, "epoch": 2557, "lr": 6.323635557212526e-06}
{"train_loss": 0.08334347605705261, "global_step": 227603, "epoch": 2557, "lr": 6.323353358759093e-06}
{"train_loss": 0.014483127743005753, "global_step": 227604, "epoch": 2557, "lr": 6.323071166177469e-06}
{"train_loss": 0.07370799779891968, "global_step": 227605, "epoch": 2557, "lr": 6.3227889794676585e-06}
{"train_loss": 0.10668425261974335, "global_step": 227606, "epoch": 2557, "lr": 6.322506798629735e-06}
{"train_loss": 0.06639712303876877, "global_step": 227607, "epoch": 2557, "lr": 6.322224623663703e-06}
{"train_loss": 0.02530454285442829, "global_step": 227608, "epoch": 2557, "lr": 6.321942454569623e-06}
{"train_loss": 0.04950449988245964, "global_step": 227609, "epoch": 2557, "lr": 6.32166029134752e-06}
{"train_loss": 0.02864426001906395, "global_step": 227610, "epoch": 2557, "lr": 6.321378133997447e-06}
{"train_loss": 0.038957204669713974, "global_step": 227611, "epoch": 2557, "lr": 6.3210959825194205e-06}
{"train_loss": 0.028795072808861732, "global_step": 227612, "epoch": 2557, "lr": 6.3208138369135025e-06}
{"train_loss": 0.0252725500613451, "global_step": 227613, "epoch": 2557, "lr": 6.320531697179705e-06}
{"train_loss": 0.021790502592921257, "global_step": 227614, "epoch": 2557, "lr": 6.320249563318093e-06}
{"train_loss": 0.03985266759991646, "global_step": 227615, "epoch": 2557, "lr": 6.319967435328683e-06}
{"train_loss": 0.053734540939331055, "global_step": 227616, "epoch": 2557, "lr": 6.319685313211515e-06}
{"train_loss": 0.012459122575819492, "global_step": 227617, "epoch": 2557, "lr": 6.31940319696665e-06}
{"train_loss": 0.06314428150653839, "global_step": 227618, "epoch": 2557, "lr": 6.3191210865940935e-06}
{"train_loss": 0.04461519047617912, "global_step": 227619, "epoch": 2557, "lr": 6.318838982093911e-06}
{"train_loss": 0.06150973215699196, "global_step": 227620, "epoch": 2557, "lr": 6.318556883466115e-06}
{"train_loss": 0.01910833828151226, "global_step": 227621, "epoch": 2557, "lr": 6.318274790710771e-06}
{"train_loss": 0.028534043580293655, "global_step": 227622, "epoch": 2557, "lr": 6.317992703827891e-06}
{"train_loss": 0.02362077683210373, "global_step": 227623, "epoch": 2557, "lr": 6.317710622817535e-06}
{"train_loss": 0.0529741607606411, "global_step": 227624, "epoch": 2557, "lr": 6.317428547679715e-06}
{"train_loss": 0.02582636848092079, "global_step": 227625, "epoch": 2557, "lr": 6.317146478414498e-06}
{"train_loss": 0.09419122338294983, "global_step": 227626, "epoch": 2557, "lr": 6.316864415021901e-06}
{"train_loss": 0.04580190032720566, "global_step": 227627, "epoch": 2557, "lr": 6.316582357501976e-06}
{"train_loss": 0.06589289754629135, "global_step": 227628, "epoch": 2557, "lr": 6.31630030585475e-06}
{"train_loss": 0.0645720362663269, "global_step": 227629, "epoch": 2557, "lr": 6.31601826008027e-06}
{"train_loss": 0.05796371400356293, "global_step": 227630, "epoch": 2557, "lr": 6.315736220178559e-06}
{"train_loss": 0.04966466501355171, "global_step": 227631, "epoch": 2557, "lr": 6.315454186149672e-06}
{"train_loss": 0.07224185764789581, "global_step": 227632, "epoch": 2557, "lr": 6.315172157993632e-06}
{"train_loss": 0.03644658252596855, "global_step": 227633, "epoch": 2557, "lr": 6.3148901357104995e-06}
{"train_loss": 0.06319702416658401, "global_step": 227634, "epoch": 2557, "lr": 6.314608119300292e-06}
{"train_loss": 0.03892286494374275, "global_step": 227635, "epoch": 2557, "lr": 6.314326108763046e-06}
{"train_loss": 0.03251618146896362, "global_step": 227636, "epoch": 2557, "lr": 6.314044104098815e-06}
{"train_loss": 0.0655294731259346, "global_step": 227637, "epoch": 2557, "lr": 6.313762105307613e-06}
{"train_loss": 0.03492646664381027, "global_step": 227638, "epoch": 2557, "lr": 6.313480112389514e-06}
{"train_loss": 0.055426113307476044, "global_step": 227639, "epoch": 2557, "lr": 6.313198125344516e-06}
{"train_loss": 0.06307477504014969, "global_step": 227640, "epoch": 2557, "lr": 6.3129161441726925e-06}
{"train_loss": 0.060788046568632126, "global_step": 227641, "epoch": 2557, "lr": 6.312634168874049e-06}
{"train_loss": 0.047749847173690796, "global_step": 227642, "epoch": 2557, "lr": 6.312352199448657e-06}
{"train_loss": 0.04170401021838188, "global_step": 227643, "epoch": 2557, "lr": 6.312070235896516e-06}
{"train_loss": 0.07576737552881241, "global_step": 227644, "epoch": 2557, "lr": 6.3117882782177064e-06}
{"train_loss": 0.10736716538667679, "global_step": 227645, "epoch": 2557, "lr": 6.3115063264122254e-06}
{"train_loss": 0.07232353091239929, "global_step": 227646, "epoch": 2557, "lr": 6.311224380480146e-06}
{"train_loss": 0.050489313900470734, "global_step": 227647, "epoch": 2557, "lr": 6.310942440421475e-06}
{"train_loss": 0.05289217829704285, "global_step": 227648, "epoch": 2557, "lr": 6.3106605062362765e-06}
{"train_loss": 0.02582727186381817, "global_step": 227649, "epoch": 2557, "lr": 6.31037857792457e-06}
{"train_loss": 0.03936615213751793, "global_step": 227650, "epoch": 2557, "lr": 6.310096655486408e-06}
{"train_loss": 0.03426641225814819, "global_step": 227651, "epoch": 2557, "lr": 6.309814738921816e-06}
{"train_loss": 0.02705838903784752, "global_step": 227652, "epoch": 2557, "lr": 6.30953282823083e-06}
{"train_loss": 0.025460166856646538, "global_step": 227653, "epoch": 2557, "lr": 6.309250923413507e-06}
{"train_loss": 0.04516259953379631, "global_step": 227654, "epoch": 2557, "lr": 6.308969024469863e-06}
{"train_loss": 0.021112732589244843, "global_step": 227655, "epoch": 2557, "lr": 6.308687131399954e-06}
{"train_loss": 0.06806350499391556, "global_step": 227656, "epoch": 2557, "lr": 6.3084052442038025e-06}
{"train_loss": 0.039285074919462204, "global_step": 227657, "epoch": 2557, "lr": 6.308123362881463e-06}
{"train_loss": 0.05242283269762993, "global_step": 227658, "epoch": 2557, "lr": 6.307841487432953e-06}
{"train_loss": 0.043399419635534286, "global_step": 227659, "epoch": 2557, "lr": 6.307559617858333e-06}
{"train_loss": 0.034897662699222565, "global_step": 227660, "epoch": 2557, "lr": 6.307277754157614e-06}
{"train_loss": 0.04996174773766419, "global_step": 227661, "epoch": 2557, "lr": 6.306995896330864e-06, "val_loss": 9.047054290771484}
{"train_loss": 0.05631231889128685, "global_step": 227662, "epoch": 2558, "lr": 6.3067140443780924e-06}
{"train_loss": 0.05034136027097702, "global_step": 227663, "epoch": 2558, "lr": 6.3064321982993664e-06}
{"train_loss": 0.05887848138809204, "global_step": 227664, "epoch": 2558, "lr": 6.306150358094692e-06}
{"train_loss": 0.02492421120405197, "global_step": 227665, "epoch": 2558, "lr": 6.305868523764136e-06}
{"train_loss": 0.04005526378750801, "global_step": 227666, "epoch": 2558, "lr": 6.3055866953077135e-06}
{"train_loss": 0.10005933791399002, "global_step": 227667, "epoch": 2558, "lr": 6.3053048727254815e-06}
{"train_loss": 0.050207871943712234, "global_step": 227668, "epoch": 2558, "lr": 6.305023056017456e-06}
{"train_loss": 0.06311125308275223, "global_step": 227669, "epoch": 2558, "lr": 6.304741245183698e-06}
{"train_loss": 0.05708032846450806, "global_step": 227670, "epoch": 2558, "lr": 6.3044594402242306e-06}
{"train_loss": 0.037582192569971085, "global_step": 227671, "epoch": 2558, "lr": 6.3041776411391025e-06}
{"train_loss": 0.027452193200588226, "global_step": 227672, "epoch": 2558, "lr": 6.303895847928343e-06}
{"train_loss": 0.08272036164999008, "global_step": 227673, "epoch": 2558, "lr": 6.3036140605919836e-06}
{"train_loss": 0.04529651254415512, "global_step": 227674, "epoch": 2558, "lr": 6.303332279130087e-06}
{"train_loss": 0.08513131737709045, "global_step": 227675, "epoch": 2558, "lr": 6.303050503542657e-06}
{"train_loss": 0.062375009059906006, "global_step": 227676, "epoch": 2558, "lr": 6.302768733829761e-06}
{"train_loss": 0.014265335164964199, "global_step": 227677, "epoch": 2558, "lr": 6.302486969991417e-06}
{"train_loss": 0.03440789133310318, "global_step": 227678, "epoch": 2558, "lr": 6.302205212027679e-06}
{"train_loss": 0.05818953365087509, "global_step": 227679, "epoch": 2558, "lr": 6.301923459938564e-06}
{"train_loss": 0.04247254878282547, "global_step": 227680, "epoch": 2558, "lr": 6.301641713724138e-06}
{"train_loss": 0.023228516802191734, "global_step": 227681, "epoch": 2558, "lr": 6.301359973384413e-06}
{"train_loss": 0.07813237607479095, "global_step": 227682, "epoch": 2558, "lr": 6.301078238919445e-06}
{"train_loss": 0.026310540735721588, "global_step": 227683, "epoch": 2558, "lr": 6.3007965103292546e-06}
{"train_loss": 0.04967406764626503, "global_step": 227684, "epoch": 2558, "lr": 6.300514787613904e-06}
{"train_loss": 0.03605656698346138, "global_step": 227685, "epoch": 2558, "lr": 6.300233070773404e-06}
{"train_loss": 0.09302885830402374, "global_step": 227686, "epoch": 2558, "lr": 6.299951359807815e-06}
{"train_loss": 0.010000777430832386, "global_step": 227687, "epoch": 2558, "lr": 6.29966965471715e-06}
{"train_loss": 0.033945076167583466, "global_step": 227688, "epoch": 2558, "lr": 6.299387955501468e-06}
{"train_loss": 0.027980268001556396, "global_step": 227689, "epoch": 2558, "lr": 6.299106262160809e-06}
{"train_loss": 0.03366613760590553, "global_step": 227690, "epoch": 2558, "lr": 6.298824574695189e-06}
{"train_loss": 0.04571203142404556, "global_step": 227691, "epoch": 2558, "lr": 6.298542893104675e-06}
{"train_loss": 0.03143121674656868, "global_step": 227692, "epoch": 2558, "lr": 6.298261217389278e-06}
{"train_loss": 0.07120556384325027, "global_step": 227693, "epoch": 2558, "lr": 6.2979795475490545e-06}
{"train_loss": 0.051457084715366364, "global_step": 227694, "epoch": 2558, "lr": 6.297697883584025e-06}
{"train_loss": 0.05260256677865982, "global_step": 227695, "epoch": 2558, "lr": 6.297416225494251e-06}
{"train_loss": 0.046091385185718536, "global_step": 227696, "epoch": 2558, "lr": 6.297134573279745e-06}
{"train_loss": 0.03900124877691269, "global_step": 227697, "epoch": 2558, "lr": 6.296852926940566e-06}
{"train_loss": 0.04434340447187424, "global_step": 227698, "epoch": 2558, "lr": 6.296571286476732e-06}
{"train_loss": 0.05703119561076164, "global_step": 227699, "epoch": 2558, "lr": 6.296289651888304e-06}
{"train_loss": 0.04233897849917412, "global_step": 227700, "epoch": 2558, "lr": 6.296008023175293e-06}
{"train_loss": 0.03477458283305168, "global_step": 227701, "epoch": 2558, "lr": 6.295726400337764e-06}
{"train_loss": 0.06663482636213303, "global_step": 227702, "epoch": 2558, "lr": 6.295444783375731e-06}
{"train_loss": 0.046894997358322144, "global_step": 227703, "epoch": 2558, "lr": 6.295163172289253e-06}
{"train_loss": 0.04846961796283722, "global_step": 227704, "epoch": 2558, "lr": 6.294881567078348e-06}
{"train_loss": 0.09190274775028229, "global_step": 227705, "epoch": 2558, "lr": 6.294599967743076e-06}
{"train_loss": 0.039917368441820145, "global_step": 227706, "epoch": 2558, "lr": 6.294318374283448e-06}
{"train_loss": 0.0615093931555748, "global_step": 227707, "epoch": 2558, "lr": 6.294036786699525e-06}
{"train_loss": 0.0577666349709034, "global_step": 227708, "epoch": 2558, "lr": 6.2937552049913304e-06}
{"train_loss": 0.07101272791624069, "global_step": 227709, "epoch": 2558, "lr": 6.2934736291589135e-06}
{"train_loss": 0.04499322548508644, "global_step": 227710, "epoch": 2558, "lr": 6.293192059202313e-06}
{"train_loss": 0.045655250549316406, "global_step": 227711, "epoch": 2558, "lr": 6.292910495121546e-06}
{"train_loss": 0.037374068051576614, "global_step": 227712, "epoch": 2558, "lr": 6.292628936916672e-06}
{"train_loss": 0.06973360478878021, "global_step": 227713, "epoch": 2558, "lr": 6.292347384587715e-06}
{"train_loss": 0.04352714493870735, "global_step": 227714, "epoch": 2558, "lr": 6.292065838134725e-06}
{"train_loss": 0.029546502977609634, "global_step": 227715, "epoch": 2558, "lr": 6.2917842975577294e-06}
{"train_loss": 0.05970592051744461, "global_step": 227716, "epoch": 2558, "lr": 6.291502762856777e-06}
{"train_loss": 0.04336113855242729, "global_step": 227717, "epoch": 2558, "lr": 6.291221234031891e-06}
{"train_loss": 0.06804809719324112, "global_step": 227718, "epoch": 2558, "lr": 6.290939711083127e-06}
{"train_loss": 0.07821177691221237, "global_step": 227719, "epoch": 2558, "lr": 6.290658194010501e-06}
{"train_loss": 0.05984866991639137, "global_step": 227720, "epoch": 2558, "lr": 6.290376682814081e-06}
{"train_loss": 0.07313010096549988, "global_step": 227721, "epoch": 2558, "lr": 6.290095177493871e-06}
{"train_loss": 0.041669636964797974, "global_step": 227722, "epoch": 2558, "lr": 6.289813678049933e-06}
{"train_loss": 0.03958945721387863, "global_step": 227723, "epoch": 2558, "lr": 6.289532184482288e-06}
{"train_loss": 0.03153131529688835, "global_step": 227724, "epoch": 2558, "lr": 6.289250696790988e-06}
{"train_loss": 0.06011473387479782, "global_step": 227725, "epoch": 2558, "lr": 6.2889692149760706e-06}
{"train_loss": 0.04967542737722397, "global_step": 227726, "epoch": 2558, "lr": 6.2886877390375576e-06}
{"train_loss": 0.05583931878209114, "global_step": 227727, "epoch": 2558, "lr": 6.288406268975511e-06}
{"train_loss": 0.09175597876310349, "global_step": 227728, "epoch": 2558, "lr": 6.288124804789947e-06}
{"train_loss": 0.03083672560751438, "global_step": 227729, "epoch": 2558, "lr": 6.287843346480915e-06}
{"train_loss": 0.03230414167046547, "global_step": 227730, "epoch": 2558, "lr": 6.287561894048444e-06}
{"train_loss": 0.04849740490317345, "global_step": 227731, "epoch": 2558, "lr": 6.2872804474925885e-06}
{"train_loss": 0.0633050948381424, "global_step": 227732, "epoch": 2558, "lr": 6.286999006813365e-06}
{"train_loss": 0.04969142749905586, "global_step": 227733, "epoch": 2558, "lr": 6.28671757201083e-06}
{"train_loss": 0.03803513944149017, "global_step": 227734, "epoch": 2558, "lr": 6.286436143085006e-06}
{"train_loss": 0.04993092641234398, "global_step": 227735, "epoch": 2558, "lr": 6.2861547200359414e-06}
{"train_loss": 0.03917806223034859, "global_step": 227736, "epoch": 2558, "lr": 6.2858733028636655e-06}
{"train_loss": 0.06273398548364639, "global_step": 227737, "epoch": 2558, "lr": 6.285591891568232e-06}
{"train_loss": 0.06876654177904129, "global_step": 227738, "epoch": 2558, "lr": 6.28531048614966e-06}
{"train_loss": 0.06560882925987244, "global_step": 227739, "epoch": 2558, "lr": 6.285029086607997e-06}
{"train_loss": 0.12185036391019821, "global_step": 227740, "epoch": 2558, "lr": 6.284747692943277e-06}
{"train_loss": 0.04654289782047272, "global_step": 227741, "epoch": 2558, "lr": 6.2844663051555455e-06}
{"train_loss": 0.09741445630788803, "global_step": 227742, "epoch": 2558, "lr": 6.284184923244823e-06}
{"train_loss": 0.12011164426803589, "global_step": 227743, "epoch": 2558, "lr": 6.283903547211173e-06}
{"train_loss": 0.04512158781290054, "global_step": 227744, "epoch": 2558, "lr": 6.28362217705461e-06}
{"train_loss": 0.04247229918837547, "global_step": 227745, "epoch": 2558, "lr": 6.283340812775185e-06}
{"train_loss": 0.06323534995317459, "global_step": 227746, "epoch": 2558, "lr": 6.2830594543729305e-06}
{"train_loss": 0.04242333397269249, "global_step": 227747, "epoch": 2558, "lr": 6.282778101847891e-06}
{"train_loss": 0.03703947737812996, "global_step": 227748, "epoch": 2558, "lr": 6.28249675520009e-06}
{"train_loss": 0.05046071484684944, "global_step": 227749, "epoch": 2558, "lr": 6.282215414429582e-06}
{"train_loss": 0.0528469549471073, "global_step": 227750, "epoch": 2558, "lr": 6.2819340795364e-06, "val_loss": 9.21124267578125}
{"train_loss": 0.011767195537686348, "global_step": 227751, "epoch": 2559, "lr": 6.281652750520567e-06}
{"train_loss": 0.06834336370229721, "global_step": 227752, "epoch": 2559, "lr": 6.2813714273821435e-06}
{"train_loss": 0.04338151216506958, "global_step": 227753, "epoch": 2559, "lr": 6.281090110121152e-06}
{"train_loss": 0.05451958253979683, "global_step": 227754, "epoch": 2559, "lr": 6.280808798737636e-06}
{"train_loss": 0.03844282776117325, "global_step": 227755, "epoch": 2559, "lr": 6.28052749323163e-06}
{"train_loss": 0.037507958710193634, "global_step": 227756, "epoch": 2559, "lr": 6.280246193603179e-06}
{"train_loss": 0.024656493216753006, "global_step": 227757, "epoch": 2559, "lr": 6.279964899852309e-06}
{"train_loss": 0.019383611157536507, "global_step": 227758, "epoch": 2559, "lr": 6.279683611979076e-06}
{"train_loss": 0.016675090417265892, "global_step": 227759, "epoch": 2559, "lr": 6.2794023299834966e-06}
{"train_loss": 0.08088227361440659, "global_step": 227760, "epoch": 2559, "lr": 6.279121053865628e-06}
{"train_loss": 0.07520069926977158, "global_step": 227761, "epoch": 2559, "lr": 6.278839783625484e-06}
{"train_loss": 0.03868013620376587, "global_step": 227762, "epoch": 2559, "lr": 6.278558519263123e-06}
{"train_loss": 0.04751553758978844, "global_step": 227763, "epoch": 2559, "lr": 6.2782772607785814e-06}
{"train_loss": 0.0315038338303566, "global_step": 227764, "epoch": 2559, "lr": 6.277996008171888e-06}
{"train_loss": 0.06294906884431839, "global_step": 227765, "epoch": 2559, "lr": 6.277714761443093e-06}
{"train_loss": 0.07152167707681656, "global_step": 227766, "epoch": 2559, "lr": 6.277433520592213e-06}
{"train_loss": 0.053312886506319046, "global_step": 227767, "epoch": 2559, "lr": 6.277152285619314e-06}
{"train_loss": 0.06994455307722092, "global_step": 227768, "epoch": 2559, "lr": 6.276871056524408e-06}
{"train_loss": 0.06329915672540665, "global_step": 227769, "epoch": 2559, "lr": 6.276589833307556e-06}
{"train_loss": 0.06880614161491394, "global_step": 227770, "epoch": 2559, "lr": 6.276308615968768e-06}
{"train_loss": 0.06419718265533447, "global_step": 227771, "epoch": 2559, "lr": 6.276027404508111e-06}
{"train_loss": 0.04935067147016525, "global_step": 227772, "epoch": 2559, "lr": 6.275746198925597e-06}
{"train_loss": 0.0731777772307396, "global_step": 227773, "epoch": 2559, "lr": 6.275464999221287e-06}
{"train_loss": 0.026928458362817764, "global_step": 227774, "epoch": 2559, "lr": 6.275183805395202e-06}
{"train_loss": 0.01691804639995098, "global_step": 227775, "epoch": 2559, "lr": 6.274902617447387e-06}
{"train_loss": 0.11138668656349182, "global_step": 227776, "epoch": 2559, "lr": 6.274621435377876e-06}
{"train_loss": 0.04288201406598091, "global_step": 227777, "epoch": 2559, "lr": 6.274340259186717e-06}
{"train_loss": 0.03292342647910118, "global_step": 227778, "epoch": 2559, "lr": 6.274059088873929e-06}
{"train_loss": 0.08900870382785797, "global_step": 227779, "epoch": 2559, "lr": 6.273777924439561e-06}
{"train_loss": 0.009929024614393711, "global_step": 227780, "epoch": 2559, "lr": 6.273496765883657e-06}
{"train_loss": 0.060724638402462006, "global_step": 227781, "epoch": 2559, "lr": 6.273215613206246e-06}
{"train_loss": 0.05606983229517937, "global_step": 227782, "epoch": 2559, "lr": 6.272934466407376e-06}
{"train_loss": 0.0534101203083992, "global_step": 227783, "epoch": 2559, "lr": 6.272653325487071e-06}
{"train_loss": 0.06781815737485886, "global_step": 227784, "epoch": 2559, "lr": 6.2723721904453695e-06}
{"train_loss": 0.05111243948340416, "global_step": 227785, "epoch": 2559, "lr": 6.2720910612823255e-06}
{"train_loss": 0.048760123550891876, "global_step": 227786, "epoch": 2559, "lr": 6.271809937997958e-06}
{"train_loss": 0.05914264917373657, "global_step": 227787, "epoch": 2559, "lr": 6.271528820592315e-06}
{"train_loss": 0.04093194380402565, "global_step": 227788, "epoch": 2559, "lr": 6.271247709065436e-06}
{"train_loss": 0.0302041657269001, "global_step": 227789, "epoch": 2559, "lr": 6.27096660341735e-06}
{"train_loss": 0.131395161151886, "global_step": 227790, "epoch": 2559, "lr": 6.270685503648105e-06}
{"train_loss": 0.0343591645359993, "global_step": 227791, "epoch": 2559, "lr": 6.270404409757719e-06}
{"train_loss": 0.04569704458117485, "global_step": 227792, "epoch": 2559, "lr": 6.270123321746263e-06}
{"train_loss": 0.030990764498710632, "global_step": 227793, "epoch": 2559, "lr": 6.269842239613738e-06}
{"train_loss": 0.05045021325349808, "global_step": 227794, "epoch": 2559, "lr": 6.269561163360216e-06}
{"train_loss": 0.03766399621963501, "global_step": 227795, "epoch": 2559, "lr": 6.269280092985702e-06}
{"train_loss": 0.027505025267601013, "global_step": 227796, "epoch": 2559, "lr": 6.268999028490263e-06}
{"train_loss": 0.04667650908231735, "global_step": 227797, "epoch": 2559, "lr": 6.268717969873916e-06}
{"train_loss": 0.04744816944003105, "global_step": 227798, "epoch": 2559, "lr": 6.268436917136705e-06}
{"train_loss": 0.11486102640628815, "global_step": 227799, "epoch": 2559, "lr": 6.26815587027868e-06}
{"train_loss": 0.02527812123298645, "global_step": 227800, "epoch": 2559, "lr": 6.267874829299858e-06}
{"train_loss": 0.03227150812745094, "global_step": 227801, "epoch": 2559, "lr": 6.2675937942003e-06}
{"train_loss": 0.054909564554691315, "global_step": 227802, "epoch": 2559, "lr": 6.267312764980016e-06}
{"train_loss": 0.051903001964092255, "global_step": 227803, "epoch": 2559, "lr": 6.267031741639073e-06}
{"train_loss": 0.0645192414522171, "global_step": 227804, "epoch": 2559, "lr": 6.266750724177483e-06}
{"train_loss": 0.03729283809661865, "global_step": 227805, "epoch": 2559, "lr": 6.266469712595308e-06}
{"train_loss": 0.018404323607683182, "global_step": 227806, "epoch": 2559, "lr": 6.2661887068925565e-06}
{"train_loss": 0.04399707168340683, "global_step": 227807, "epoch": 2559, "lr": 6.265907707069301e-06}
{"train_loss": 0.07187675684690475, "global_step": 227808, "epoch": 2559, "lr": 6.265626713125544e-06}
{"train_loss": 0.03334967419505119, "global_step": 227809, "epoch": 2559, "lr": 6.265345725061355e-06}
{"train_loss": 0.031521715223789215, "global_step": 227810, "epoch": 2559, "lr": 6.2650647428767475e-06}
{"train_loss": 0.07005150616168976, "global_step": 227811, "epoch": 2559, "lr": 6.264783766571775e-06}
{"train_loss": 0.046116262674331665, "global_step": 227812, "epoch": 2559, "lr": 6.26450279614646e-06}
{"train_loss": 0.05972417816519737, "global_step": 227813, "epoch": 2559, "lr": 6.264221831600864e-06}
{"train_loss": 0.08532633632421494, "global_step": 227814, "epoch": 2559, "lr": 6.263940872934998e-06}
{"train_loss": 0.0335209034383297, "global_step": 227815, "epoch": 2559, "lr": 6.263659920148912e-06}
{"train_loss": 0.06677497178316116, "global_step": 227816, "epoch": 2559, "lr": 6.26337897324265e-06}
{"train_loss": 0.06906263530254364, "global_step": 227817, "epoch": 2559, "lr": 6.2630980322162355e-06}
{"train_loss": 0.05449935421347618, "global_step": 227818, "epoch": 2559, "lr": 6.262817097069729e-06}
{"train_loss": 0.0340765118598938, "global_step": 227819, "epoch": 2559, "lr": 6.26253616780314e-06}
{"train_loss": 0.047467444092035294, "global_step": 227820, "epoch": 2559, "lr": 6.262255244416532e-06}
{"train_loss": 0.040180791169404984, "global_step": 227821, "epoch": 2559, "lr": 6.261974326909931e-06}
{"train_loss": 0.04872475191950798, "global_step": 227822, "epoch": 2559, "lr": 6.26169341528336e-06}
{"train_loss": 0.03063332661986351, "global_step": 227823, "epoch": 2559, "lr": 6.2614125095368845e-06}
{"train_loss": 0.04607326164841652, "global_step": 227824, "epoch": 2559, "lr": 6.261131609670517e-06}
{"train_loss": 0.052089937031269073, "global_step": 227825, "epoch": 2559, "lr": 6.2608507156843185e-06}
{"train_loss": 0.03303516283631325, "global_step": 227826, "epoch": 2559, "lr": 6.2605698275783166e-06}
{"train_loss": 0.029151130467653275, "global_step": 227827, "epoch": 2559, "lr": 6.2602889453525385e-06}
{"train_loss": 0.034573085606098175, "global_step": 227828, "epoch": 2559, "lr": 6.260008069007039e-06}
{"train_loss": 0.08143307268619537, "global_step": 227829, "epoch": 2559, "lr": 6.259727198541837e-06}
{"train_loss": 0.026931921020150185, "global_step": 227830, "epoch": 2559, "lr": 6.259446333956997e-06}
{"train_loss": 0.044414810836315155, "global_step": 227831, "epoch": 2559, "lr": 6.259165475252526e-06}
{"train_loss": 0.06221771612763405, "global_step": 227832, "epoch": 2559, "lr": 6.25888462242849e-06}
{"train_loss": 0.025799274444580078, "global_step": 227833, "epoch": 2559, "lr": 6.2586037754849005e-06}
{"train_loss": 0.021685613319277763, "global_step": 227834, "epoch": 2559, "lr": 6.258322934421812e-06}
{"train_loss": 0.04910913482308388, "global_step": 227835, "epoch": 2559, "lr": 6.2580420992392694e-06}
{"train_loss": 0.033980585634708405, "global_step": 227836, "epoch": 2559, "lr": 6.257761269937285e-06}
{"train_loss": 0.047596585005521774, "global_step": 227837, "epoch": 2559, "lr": 6.257480446515929e-06}
{"train_loss": 0.03525999188423157, "global_step": 227838, "epoch": 2559, "lr": 6.257199628975208e-06}
{"train_loss": 0.049048939165188356, "global_step": 227839, "epoch": 2559, "lr": 6.256918817315183e-06, "val_loss": 9.052288055419922}
{"train_loss": 0.05025871843099594, "global_step": 227840, "epoch": 2560, "lr": 6.256638011535871e-06}
{"train_loss": 0.061547473073005676, "global_step": 227841, "epoch": 2560, "lr": 6.256357211637331e-06}
{"train_loss": 0.061322975903749466, "global_step": 227842, "epoch": 2560, "lr": 6.256076417619583e-06}
{"train_loss": 0.043657202273607254, "global_step": 227843, "epoch": 2560, "lr": 6.2557956294826805e-06}
{"train_loss": 0.05062379688024521, "global_step": 227844, "epoch": 2560, "lr": 6.2555148472266404e-06}
{"train_loss": 0.041180528700351715, "global_step": 227845, "epoch": 2560, "lr": 6.255234070851529e-06}
{"train_loss": 0.01933375746011734, "global_step": 227846, "epoch": 2560, "lr": 6.254953300357352e-06}
{"train_loss": 0.07247469574213028, "global_step": 227847, "epoch": 2560, "lr": 6.254672535744177e-06}
{"train_loss": 0.10852453857660294, "global_step": 227848, "epoch": 2560, "lr": 6.254391777012019e-06}
{"train_loss": 0.07162563502788544, "global_step": 227849, "epoch": 2560, "lr": 6.2541110241609356e-06}
{"train_loss": 0.0904834195971489, "global_step": 227850, "epoch": 2560, "lr": 6.253830277190942e-06}
{"train_loss": 0.055900778621435165, "global_step": 227851, "epoch": 2560, "lr": 6.253549536102082e-06}
{"train_loss": 0.037571512162685394, "global_step": 227852, "epoch": 2560, "lr": 6.253268800894419e-06}
{"train_loss": 0.028847195208072662, "global_step": 227853, "epoch": 2560, "lr": 6.252988071567956e-06}
{"train_loss": 0.04261188954114914, "global_step": 227854, "epoch": 2560, "lr": 6.252707348122755e-06}
{"train_loss": 0.059304505586624146, "global_step": 227855, "epoch": 2560, "lr": 6.252426630558833e-06}
{"train_loss": 0.04949978366494179, "global_step": 227856, "epoch": 2560, "lr": 6.252145918876256e-06}
{"train_loss": 0.04693473502993584, "global_step": 227857, "epoch": 2560, "lr": 6.251865213075025e-06}
{"train_loss": 0.04231206327676773, "global_step": 227858, "epoch": 2560, "lr": 6.2515845131552176e-06}
{"train_loss": 0.06542659550905228, "global_step": 227859, "epoch": 2560, "lr": 6.251303819116844e-06}
{"train_loss": 0.048304036259651184, "global_step": 227860, "epoch": 2560, "lr": 6.251023130959943e-06}
{"train_loss": 0.016062507405877113, "global_step": 227861, "epoch": 2560, "lr": 6.250742448684565e-06}
{"train_loss": 0.0491548627614975, "global_step": 227862, "epoch": 2560, "lr": 6.2504617722907385e-06}
{"train_loss": 0.047212157398462296, "global_step": 227863, "epoch": 2560, "lr": 6.250181101778507e-06}
{"train_loss": 0.10945907235145569, "global_step": 227864, "epoch": 2560, "lr": 6.249900437147899e-06}
{"train_loss": 0.02867763303220272, "global_step": 227865, "epoch": 2560, "lr": 6.249619778398968e-06}
{"train_loss": 0.0672367513179779, "global_step": 227866, "epoch": 2560, "lr": 6.249339125531744e-06}
{"train_loss": 0.07055812329053879, "global_step": 227867, "epoch": 2560, "lr": 6.249058478546249e-06}
{"train_loss": 0.034779179841279984, "global_step": 227868, "epoch": 2560, "lr": 6.248777837442549e-06}
{"train_loss": 0.051811859011650085, "global_step": 227869, "epoch": 2560, "lr": 6.248497202220655e-06}
{"train_loss": 0.02075687050819397, "global_step": 227870, "epoch": 2560, "lr": 6.248216572880616e-06}
{"train_loss": 0.08614809066057205, "global_step": 227871, "epoch": 2560, "lr": 6.2479359494224846e-06}
{"train_loss": 0.039041951298713684, "global_step": 227872, "epoch": 2560, "lr": 6.247655331846269e-06}
{"train_loss": 0.04988368600606918, "global_step": 227873, "epoch": 2560, "lr": 6.2473747201520385e-06}
{"train_loss": 0.07465264946222305, "global_step": 227874, "epoch": 2560, "lr": 6.247094114339808e-06}
{"train_loss": 0.05955972522497177, "global_step": 227875, "epoch": 2560, "lr": 6.246813514409627e-06}
{"train_loss": 0.026567690074443817, "global_step": 227876, "epoch": 2560, "lr": 6.24653292036152e-06}
{"train_loss": 0.038024500012397766, "global_step": 227877, "epoch": 2560, "lr": 6.24625233219554e-06}
{"train_loss": 0.0360853336751461, "global_step": 227878, "epoch": 2560, "lr": 6.245971749911711e-06}
{"train_loss": 0.02101922407746315, "global_step": 227879, "epoch": 2560, "lr": 6.245691173510093e-06}
{"train_loss": 0.07910826057195663, "global_step": 227880, "epoch": 2560, "lr": 6.245410602990692e-06}
{"train_loss": 0.060055386275053024, "global_step": 227881, "epoch": 2560, "lr": 6.245130038353575e-06}
{"train_loss": 0.017484072595834732, "global_step": 227882, "epoch": 2560, "lr": 6.244849479598752e-06}
{"train_loss": 0.03690206632018089, "global_step": 227883, "epoch": 2560, "lr": 6.2445689267262906e-06}
{"train_loss": 0.027781467884778976, "global_step": 227884, "epoch": 2560, "lr": 6.244288379736202e-06}
{"train_loss": 0.0811115950345993, "global_step": 227885, "epoch": 2560, "lr": 6.244007838628546e-06}
{"train_loss": 0.07972259819507599, "global_step": 227886, "epoch": 2560, "lr": 6.243727303403341e-06}
{"train_loss": 0.0350160151720047, "global_step": 227887, "epoch": 2560, "lr": 6.2434467740606466e-06}
{"train_loss": 0.04437384381890297, "global_step": 227888, "epoch": 2560, "lr": 6.243166250600468e-06}
{"train_loss": 0.043804313987493515, "global_step": 227889, "epoch": 2560, "lr": 6.242885733022869e-06}
{"train_loss": 0.06303635984659195, "global_step": 227890, "epoch": 2560, "lr": 6.242605221327891e-06}
{"train_loss": 0.012546205893158913, "global_step": 227891, "epoch": 2560, "lr": 6.242324715515552e-06}
{"train_loss": 0.05869923532009125, "global_step": 227892, "epoch": 2560, "lr": 6.242044215585907e-06}
{"train_loss": 0.06421300768852234, "global_step": 227893, "epoch": 2560, "lr": 6.241763721538973e-06}
{"train_loss": 0.027842499315738678, "global_step": 227894, "epoch": 2560, "lr": 6.241483233374812e-06}
{"train_loss": 0.020666804164648056, "global_step": 227895, "epoch": 2560, "lr": 6.241202751093445e-06}
{"train_loss": 0.03387754037976265, "global_step": 227896, "epoch": 2560, "lr": 6.240922274694921e-06}
{"train_loss": 0.04027768224477768, "global_step": 227897, "epoch": 2560, "lr": 6.240641804179259e-06}
{"train_loss": 0.07800283282995224, "global_step": 227898, "epoch": 2560, "lr": 6.240361339546524e-06}
{"train_loss": 0.0329655259847641, "global_step": 227899, "epoch": 2560, "lr": 6.240080880796739e-06}
{"train_loss": 0.08359172195196152, "global_step": 227900, "epoch": 2560, "lr": 6.239800427929932e-06}
{"train_loss": 0.07236532121896744, "global_step": 227901, "epoch": 2560, "lr": 6.239519980946157e-06}
{"train_loss": 0.0738539844751358, "global_step": 227902, "epoch": 2560, "lr": 6.239239539845437e-06}
{"train_loss": 0.0792342945933342, "global_step": 227903, "epoch": 2560, "lr": 6.238959104627828e-06}
{"train_loss": 0.046532806009054184, "global_step": 227904, "epoch": 2560, "lr": 6.238678675293358e-06}
{"train_loss": 0.0790133848786354, "global_step": 227905, "epoch": 2560, "lr": 6.238398251842053e-06}
{"train_loss": 0.027352014556527138, "global_step": 227906, "epoch": 2560, "lr": 6.23811783427396e-06}
{"train_loss": 0.05219702795147896, "global_step": 227907, "epoch": 2560, "lr": 6.237837422589132e-06}
{"train_loss": 0.038370728492736816, "global_step": 227908, "epoch": 2560, "lr": 6.2375570167875815e-06}
{"train_loss": 0.047447025775909424, "global_step": 227909, "epoch": 2560, "lr": 6.237276616869375e-06}
{"train_loss": 0.054401904344558716, "global_step": 227910, "epoch": 2560, "lr": 6.236996222834513e-06}
{"train_loss": 0.06296264380216599, "global_step": 227911, "epoch": 2560, "lr": 6.2367158346830715e-06}
{"train_loss": 0.08371944725513458, "global_step": 227912, "epoch": 2560, "lr": 6.236435452415057e-06}
{"train_loss": 0.05502379685640335, "global_step": 227913, "epoch": 2560, "lr": 6.236155076030531e-06}
{"train_loss": 0.07028234750032425, "global_step": 227914, "epoch": 2560, "lr": 6.23587470552951e-06}
{"train_loss": 0.015771042555570602, "global_step": 227915, "epoch": 2560, "lr": 6.235594340912049e-06}
{"train_loss": 0.036108266562223434, "global_step": 227916, "epoch": 2560, "lr": 6.23531398217817e-06}
{"train_loss": 0.08731667697429657, "global_step": 227917, "epoch": 2560, "lr": 6.235033629327935e-06}
{"train_loss": 0.07517979294061661, "global_step": 227918, "epoch": 2560, "lr": 6.23475328236135e-06}
{"train_loss": 0.05239880084991455, "global_step": 227919, "epoch": 2560, "lr": 6.234472941278485e-06}
{"train_loss": 0.0793568566441536, "global_step": 227920, "epoch": 2560, "lr": 6.2341926060793475e-06}
{"train_loss": 0.06204979494214058, "global_step": 227921, "epoch": 2560, "lr": 6.233912276763998e-06}
{"train_loss": 0.06309916824102402, "global_step": 227922, "epoch": 2560, "lr": 6.233631953332458e-06}
{"train_loss": 0.031160665675997734, "global_step": 227923, "epoch": 2560, "lr": 6.233351635784784e-06}
{"train_loss": 0.04797537624835968, "global_step": 227924, "epoch": 2560, "lr": 6.233071324120993e-06}
{"train_loss": 0.0690939649939537, "global_step": 227925, "epoch": 2560, "lr": 6.232791018341127e-06}
{"train_loss": 0.04586953669786453, "global_step": 227926, "epoch": 2560, "lr": 6.232510718445239e-06}
{"train_loss": 0.021971266716718674, "global_step": 227927, "epoch": 2560, "lr": 6.232230424433349e-06}
{"train_loss": 0.052580712404981085, "global_step": 227928, "epoch": 2560, "lr": 6.231950136305515e-06, "val_loss": 9.130982398986816, "train_action_mse_error": 7.398433208465576}
{"train_loss": 0.03659135103225708, "global_step": 227929, "epoch": 2561, "lr": 6.2316698540617505e-06}
{"train_loss": 0.02863655798137188, "global_step": 227930, "epoch": 2561, "lr": 6.231389577702112e-06}
{"train_loss": 0.0946563258767128, "global_step": 227931, "epoch": 2561, "lr": 6.231109307226623e-06}
{"train_loss": 0.06625140458345413, "global_step": 227932, "epoch": 2561, "lr": 6.230829042635333e-06}
{"train_loss": 0.02268984541296959, "global_step": 227933, "epoch": 2561, "lr": 6.23054878392827e-06}
{"train_loss": 0.04571326822042465, "global_step": 227934, "epoch": 2561, "lr": 6.230268531105482e-06}
{"train_loss": 0.049471259117126465, "global_step": 227935, "epoch": 2561, "lr": 6.229988284166993e-06}
{"train_loss": 0.022068582475185394, "global_step": 227936, "epoch": 2561, "lr": 6.229708043112858e-06}
{"train_loss": 0.023172715678811073, "global_step": 227937, "epoch": 2561, "lr": 6.229427807943106e-06}
{"train_loss": 0.0627058818936348, "global_step": 227938, "epoch": 2561, "lr": 6.229147578657762e-06}
{"train_loss": 0.02634689398109913, "global_step": 227939, "epoch": 2561, "lr": 6.22886735525689e-06}
{"train_loss": 0.08012547343969345, "global_step": 227940, "epoch": 2561, "lr": 6.228587137740499e-06}
{"train_loss": 0.024956777691841125, "global_step": 227941, "epoch": 2561, "lr": 6.228306926108657e-06}
{"train_loss": 0.03440763056278229, "global_step": 227942, "epoch": 2561, "lr": 6.228026720361369e-06}
{"train_loss": 0.04729493707418442, "global_step": 227943, "epoch": 2561, "lr": 6.227746520498701e-06}
{"train_loss": 0.04738500714302063, "global_step": 227944, "epoch": 2561, "lr": 6.2274663265206704e-06}
{"train_loss": 0.06782439351081848, "global_step": 227945, "epoch": 2561, "lr": 6.227186138427332e-06}
{"train_loss": 0.03958564251661301, "global_step": 227946, "epoch": 2561, "lr": 6.2269059562187035e-06}
{"train_loss": 0.060371268540620804, "global_step": 227947, "epoch": 2561, "lr": 6.226625779894846e-06}
{"train_loss": 0.029055533930659294, "global_step": 227948, "epoch": 2561, "lr": 6.226345609455775e-06}
{"train_loss": 0.022066710516810417, "global_step": 227949, "epoch": 2561, "lr": 6.226065444901546e-06}
{"train_loss": 0.06338634341955185, "global_step": 227950, "epoch": 2561, "lr": 6.225785286232183e-06}
{"train_loss": 0.01952606812119484, "global_step": 227951, "epoch": 2561, "lr": 6.22550513344774e-06}
{"train_loss": 0.02898464724421501, "global_step": 227952, "epoch": 2561, "lr": 6.2252249865482284e-06}
{"train_loss": 0.027154825627803802, "global_step": 227953, "epoch": 2561, "lr": 6.22494484553372e-06}
{"train_loss": 0.017297275364398956, "global_step": 227954, "epoch": 2561, "lr": 6.224664710404215e-06}
{"train_loss": 0.02135474979877472, "global_step": 227955, "epoch": 2561, "lr": 6.224384581159787e-06}
{"train_loss": 0.033916376531124115, "global_step": 227956, "epoch": 2561, "lr": 6.224104457800445e-06}
{"train_loss": 0.06991884112358093, "global_step": 227957, "epoch": 2561, "lr": 6.223824340326251e-06}
{"train_loss": 0.04442444071173668, "global_step": 227958, "epoch": 2561, "lr": 6.223544228737216e-06}
{"train_loss": 0.08041930198669434, "global_step": 227959, "epoch": 2561, "lr": 6.223264123033401e-06}
{"train_loss": 0.012373760342597961, "global_step": 227960, "epoch": 2561, "lr": 6.2229840232148284e-06}
{"train_loss": 0.06969383358955383, "global_step": 227961, "epoch": 2561, "lr": 6.222703929281543e-06}
{"train_loss": 0.06972937285900116, "global_step": 227962, "epoch": 2561, "lr": 6.2224238412335874e-06}
{"train_loss": 0.04497402533888817, "global_step": 227963, "epoch": 2561, "lr": 6.2221437590709866e-06}
{"train_loss": 0.07549741119146347, "global_step": 227964, "epoch": 2561, "lr": 6.221863682793793e-06}
{"train_loss": 0.047138817608356476, "global_step": 227965, "epoch": 2561, "lr": 6.2215836124020265e-06}
{"train_loss": 0.0329304002225399, "global_step": 227966, "epoch": 2561, "lr": 6.221303547895746e-06}
{"train_loss": 0.047045037150382996, "global_step": 227967, "epoch": 2561, "lr": 6.221023489274968e-06}
{"train_loss": 0.032829657196998596, "global_step": 227968, "epoch": 2561, "lr": 6.220743436539745e-06}
{"train_loss": 0.06398740410804749, "global_step": 227969, "epoch": 2561, "lr": 6.220463389690107e-06}
{"train_loss": 0.03172421082854271, "global_step": 227970, "epoch": 2561, "lr": 6.2201833487261e-06}
{"train_loss": 0.09227089583873749, "global_step": 227971, "epoch": 2561, "lr": 6.2199033136477405e-06}
{"train_loss": 0.07333525270223618, "global_step": 227972, "epoch": 2561, "lr": 6.2196232844551005e-06}
{"train_loss": 0.07824002206325531, "global_step": 227973, "epoch": 2561, "lr": 6.219343261148181e-06}
{"train_loss": 0.07403463125228882, "global_step": 227974, "epoch": 2561, "lr": 6.2190632437270525e-06}
{"train_loss": 0.07321056723594666, "global_step": 227975, "epoch": 2561, "lr": 6.218783232191727e-06}
{"train_loss": 0.03442234545946121, "global_step": 227976, "epoch": 2561, "lr": 6.218503226542261e-06}
{"train_loss": 0.04625341296195984, "global_step": 227977, "epoch": 2561, "lr": 6.218223226778685e-06}
{"train_loss": 0.10508308559656143, "global_step": 227978, "epoch": 2561, "lr": 6.217943232901025e-06}
{"train_loss": 0.04267524927854538, "global_step": 227979, "epoch": 2561, "lr": 6.217663244909333e-06}
{"train_loss": 0.039731524884700775, "global_step": 227980, "epoch": 2561, "lr": 6.217383262803639e-06}
{"train_loss": 0.07297709584236145, "global_step": 227981, "epoch": 2561, "lr": 6.2171032865839915e-06}
{"train_loss": 0.10316479206085205, "global_step": 227982, "epoch": 2561, "lr": 6.216823316250414e-06}
{"train_loss": 0.05157110467553139, "global_step": 227983, "epoch": 2561, "lr": 6.216543351802962e-06}
{"train_loss": 0.05795901268720627, "global_step": 227984, "epoch": 2561, "lr": 6.216263393241645e-06}
{"train_loss": 0.04439032822847366, "global_step": 227985, "epoch": 2561, "lr": 6.215983440566531e-06}
{"train_loss": 0.08335527777671814, "global_step": 227986, "epoch": 2561, "lr": 6.215703493777636e-06}
{"train_loss": 0.05088113993406296, "global_step": 227987, "epoch": 2561, "lr": 6.215423552875016e-06}
{"train_loss": 0.05172383412718773, "global_step": 227988, "epoch": 2561, "lr": 6.215143617858687e-06}
{"train_loss": 0.04376131296157837, "global_step": 227989, "epoch": 2561, "lr": 6.214863688728712e-06}
{"train_loss": 0.037728242576122284, "global_step": 227990, "epoch": 2561, "lr": 6.214583765485099e-06}
{"train_loss": 0.018545515835285187, "global_step": 227991, "epoch": 2561, "lr": 6.214303848127917e-06}
{"train_loss": 0.0769137591123581, "global_step": 227992, "epoch": 2561, "lr": 6.214023936657176e-06}
{"train_loss": 0.035179466009140015, "global_step": 227993, "epoch": 2561, "lr": 6.2137440310729326e-06}
{"train_loss": 0.045626815408468246, "global_step": 227994, "epoch": 2561, "lr": 6.213464131375213e-06}
{"train_loss": 0.03986641764640808, "global_step": 227995, "epoch": 2561, "lr": 6.2131842375640684e-06}
{"train_loss": 0.024424472823739052, "global_step": 227996, "epoch": 2561, "lr": 6.212904349639515e-06}
{"train_loss": 0.05176480486989021, "global_step": 227997, "epoch": 2561, "lr": 6.212624467601602e-06}
{"train_loss": 0.031050268560647964, "global_step": 227998, "epoch": 2561, "lr": 6.21234459145038e-06}
{"train_loss": 0.08646886050701141, "global_step": 227999, "epoch": 2561, "lr": 6.2120647211858674e-06}
{"train_loss": 0.10759222507476807, "global_step": 228000, "epoch": 2561, "lr": 6.2117848568081115e-06}
{"train_loss": 0.05135384947061539, "global_step": 228001, "epoch": 2561, "lr": 6.211504998317141e-06}
{"train_loss": 0.05586759373545647, "global_step": 228002, "epoch": 2561, "lr": 6.211225145713012e-06}
{"train_loss": 0.06143326684832573, "global_step": 228003, "epoch": 2561, "lr": 6.2109452989957405e-06}
{"train_loss": 0.054288700222969055, "global_step": 228004, "epoch": 2561, "lr": 6.210665458165382e-06}
{"train_loss": 0.049923162907361984, "global_step": 228005, "epoch": 2561, "lr": 6.210385623221959e-06}
{"train_loss": 0.019808100536465645, "global_step": 228006, "epoch": 2561, "lr": 6.210105794165522e-06}
{"train_loss": 0.040455855429172516, "global_step": 228007, "epoch": 2561, "lr": 6.2098259709960915e-06}
{"train_loss": 0.08575870096683502, "global_step": 228008, "epoch": 2561, "lr": 6.20954615371373e-06}
{"train_loss": 0.035043634474277496, "global_step": 228009, "epoch": 2561, "lr": 6.209266342318448e-06}
{"train_loss": 0.08430048078298569, "global_step": 228010, "epoch": 2561, "lr": 6.208986536810307e-06}
{"train_loss": 0.08965971320867538, "global_step": 228011, "epoch": 2561, "lr": 6.208706737189329e-06}
{"train_loss": 0.01204392034560442, "global_step": 228012, "epoch": 2561, "lr": 6.208426943455559e-06}
{"train_loss": 0.061784375458955765, "global_step": 228013, "epoch": 2561, "lr": 6.2081471556090285e-06}
{"train_loss": 0.03764748200774193, "global_step": 228014, "epoch": 2561, "lr": 6.207867373649784e-06}
{"train_loss": 0.06095757707953453, "global_step": 228015, "epoch": 2561, "lr": 6.207587597577863e-06}
{"train_loss": 0.04496259242296219, "global_step": 228016, "epoch": 2561, "lr": 6.207307827393283e-06}
{"train_loss": 0.050628267741354, "global_step": 228017, "epoch": 2561, "lr": 6.20702806309611e-06, "val_loss": 9.298966407775879}
{"train_loss": 0.03514853119850159, "global_step": 228018, "epoch": 2562, "lr": 6.206748304686355e-06}
{"train_loss": 0.05565408989787102, "global_step": 228019, "epoch": 2562, "lr": 6.206468552164086e-06}
{"train_loss": 0.026815125718712807, "global_step": 228020, "epoch": 2562, "lr": 6.206188805529306e-06}
{"train_loss": 0.10408449918031693, "global_step": 228021, "epoch": 2562, "lr": 6.205909064782084e-06}
{"train_loss": 0.031506430357694626, "global_step": 228022, "epoch": 2562, "lr": 6.20562932992243e-06}
{"train_loss": 0.031372424215078354, "global_step": 228023, "epoch": 2562, "lr": 6.205349600950411e-06}
{"train_loss": 0.05724921077489853, "global_step": 228024, "epoch": 2562, "lr": 6.205069877866038e-06}
{"train_loss": 0.05393177643418312, "global_step": 228025, "epoch": 2562, "lr": 6.204790160669366e-06}
{"train_loss": 0.05200910568237305, "global_step": 228026, "epoch": 2562, "lr": 6.204510449360418e-06}
{"train_loss": 0.07435674220323563, "global_step": 228027, "epoch": 2562, "lr": 6.204230743939249e-06}
{"train_loss": 0.031735196709632874, "global_step": 228028, "epoch": 2562, "lr": 6.203951044405875e-06}
{"train_loss": 0.0826234370470047, "global_step": 228029, "epoch": 2562, "lr": 6.2036713507603595e-06}
{"train_loss": 0.09266438335180283, "global_step": 228030, "epoch": 2562, "lr": 6.203391663002717e-06}
{"train_loss": 0.03090190887451172, "global_step": 228031, "epoch": 2562, "lr": 6.203111981133003e-06}
{"train_loss": 0.06820299476385117, "global_step": 228032, "epoch": 2562, "lr": 6.20283230515124e-06}
{"train_loss": 0.026703594252467155, "global_step": 228033, "epoch": 2562, "lr": 6.202552635057468e-06}
{"train_loss": 0.04522440955042839, "global_step": 228034, "epoch": 2562, "lr": 6.2022729708517415e-06}
{"train_loss": 0.09023134410381317, "global_step": 228035, "epoch": 2562, "lr": 6.201993312534072e-06}
{"train_loss": 0.08938855677843094, "global_step": 228036, "epoch": 2562, "lr": 6.201713660104524e-06}
{"train_loss": 0.055570606142282486, "global_step": 228037, "epoch": 2562, "lr": 6.201434013563112e-06}
{"train_loss": 0.07260958105325699, "global_step": 228038, "epoch": 2562, "lr": 6.201154372909896e-06}
{"train_loss": 0.05266578868031502, "global_step": 228039, "epoch": 2562, "lr": 6.200874738144885e-06}
{"train_loss": 0.06513935327529907, "global_step": 228040, "epoch": 2562, "lr": 6.200595109268148e-06}
{"train_loss": 0.07012391835451126, "global_step": 228041, "epoch": 2562, "lr": 6.20031548627969e-06}
{"train_loss": 0.045378755778074265, "global_step": 228042, "epoch": 2562, "lr": 6.200035869179582e-06}
{"train_loss": 0.02041800506412983, "global_step": 228043, "epoch": 2562, "lr": 6.199756257967831e-06}
{"train_loss": 0.043724630028009415, "global_step": 228044, "epoch": 2562, "lr": 6.199476652644503e-06}
{"train_loss": 0.07151485979557037, "global_step": 228045, "epoch": 2562, "lr": 6.19919705320961e-06}
{"train_loss": 0.06799428164958954, "global_step": 228046, "epoch": 2562, "lr": 6.1989174596632115e-06}
{"train_loss": 0.07332068681716919, "global_step": 228047, "epoch": 2562, "lr": 6.198637872005325e-06}
{"train_loss": 0.03933890536427498, "global_step": 228048, "epoch": 2562, "lr": 6.198358290236006e-06}
{"train_loss": 0.06150060519576073, "global_step": 228049, "epoch": 2562, "lr": 6.198078714355277e-06}
{"train_loss": 0.026027392596006393, "global_step": 228050, "epoch": 2562, "lr": 6.197799144363187e-06}
{"train_loss": 0.05889595299959183, "global_step": 228051, "epoch": 2562, "lr": 6.1975195802597645e-06}
{"train_loss": 0.026633482426404953, "global_step": 228052, "epoch": 2562, "lr": 6.197240022045059e-06}
{"train_loss": 0.03888441622257233, "global_step": 228053, "epoch": 2562, "lr": 6.1969604697190986e-06}
{"train_loss": 0.055127259343862534, "global_step": 228054, "epoch": 2562, "lr": 6.196680923281917e-06}
{"train_loss": 0.044418029487133026, "global_step": 228055, "epoch": 2562, "lr": 6.1964013827335685e-06}
{"train_loss": 0.03459533676505089, "global_step": 228056, "epoch": 2562, "lr": 6.196121848074066e-06}
{"train_loss": 0.07202963531017303, "global_step": 228057, "epoch": 2562, "lr": 6.195842319303474e-06}
{"train_loss": 0.0479293130338192, "global_step": 228058, "epoch": 2562, "lr": 6.195562796421805e-06}
{"train_loss": 0.03468077629804611, "global_step": 228059, "epoch": 2562, "lr": 6.19528327942912e-06}
{"train_loss": 0.02633965201675892, "global_step": 228060, "epoch": 2562, "lr": 6.195003768325436e-06}
{"train_loss": 0.04109001159667969, "global_step": 228061, "epoch": 2562, "lr": 6.194724263110807e-06}
{"train_loss": 0.05002310499548912, "global_step": 228062, "epoch": 2562, "lr": 6.1944447637852565e-06}
{"train_loss": 0.049010928720235825, "global_step": 228063, "epoch": 2562, "lr": 6.194165270348834e-06}
{"train_loss": 0.05769514665007591, "global_step": 228064, "epoch": 2562, "lr": 6.193885782801567e-06}
{"train_loss": 0.04448993131518364, "global_step": 228065, "epoch": 2562, "lr": 6.193606301143506e-06}
{"train_loss": 0.0565185546875, "global_step": 228066, "epoch": 2562, "lr": 6.193326825374674e-06}
{"train_loss": 0.05048082768917084, "global_step": 228067, "epoch": 2562, "lr": 6.1930473554951185e-06}
{"train_loss": 0.03555333614349365, "global_step": 228068, "epoch": 2562, "lr": 6.19276789150487e-06}
{"train_loss": 0.05975549668073654, "global_step": 228069, "epoch": 2562, "lr": 6.192488433403965e-06}
{"train_loss": 0.04857688397169113, "global_step": 228070, "epoch": 2562, "lr": 6.19220898119246e-06}
{"train_loss": 0.06282250583171844, "global_step": 228071, "epoch": 2562, "lr": 6.191929534870367e-06}
{"train_loss": 0.040572166442871094, "global_step": 228072, "epoch": 2562, "lr": 6.191650094437746e-06}
{"train_loss": 0.05670684948563576, "global_step": 228073, "epoch": 2562, "lr": 6.1913706598946134e-06}
{"train_loss": 0.0499892383813858, "global_step": 228074, "epoch": 2562, "lr": 6.191091231241031e-06}
{"train_loss": 0.056021273136138916, "global_step": 228075, "epoch": 2562, "lr": 6.190811808477004e-06}
{"train_loss": 0.06163797900080681, "global_step": 228076, "epoch": 2562, "lr": 6.190532391602604e-06}
{"train_loss": 0.03536372631788254, "global_step": 228077, "epoch": 2562, "lr": 6.190252980617844e-06}
{"train_loss": 0.031747929751873016, "global_step": 228078, "epoch": 2562, "lr": 6.189973575522778e-06}
{"train_loss": 0.036875419318675995, "global_step": 228079, "epoch": 2562, "lr": 6.189694176317423e-06}
{"train_loss": 0.05034829303622246, "global_step": 228080, "epoch": 2562, "lr": 6.189414783001846e-06}
{"train_loss": 0.058747999370098114, "global_step": 228081, "epoch": 2562, "lr": 6.189135395576051e-06}
{"train_loss": 0.05468077212572098, "global_step": 228082, "epoch": 2562, "lr": 6.188856014040112e-06}
{"train_loss": 0.044853899627923965, "global_step": 228083, "epoch": 2562, "lr": 6.188576638394028e-06}
{"train_loss": 0.028418175876140594, "global_step": 228084, "epoch": 2562, "lr": 6.188297268637871e-06}
{"train_loss": 0.0709916353225708, "global_step": 228085, "epoch": 2562, "lr": 6.188017904771653e-06}
{"train_loss": 0.02552911639213562, "global_step": 228086, "epoch": 2562, "lr": 6.18773854679543e-06}
{"train_loss": 0.016119133681058884, "global_step": 228087, "epoch": 2562, "lr": 6.187459194709228e-06}
{"train_loss": 0.043686896562576294, "global_step": 228088, "epoch": 2562, "lr": 6.1871798485130915e-06}
{"train_loss": 0.058580636978149414, "global_step": 228089, "epoch": 2562, "lr": 6.186900508207044e-06}
{"train_loss": 0.07578548043966293, "global_step": 228090, "epoch": 2562, "lr": 6.186621173791146e-06}
{"train_loss": 0.08458933234214783, "global_step": 228091, "epoch": 2562, "lr": 6.186341845265414e-06}
{"train_loss": 0.029135687276721, "global_step": 228092, "epoch": 2562, "lr": 6.186062522629904e-06}
{"train_loss": 0.07032277435064316, "global_step": 228093, "epoch": 2562, "lr": 6.185783205884643e-06}
{"train_loss": 0.04125843197107315, "global_step": 228094, "epoch": 2562, "lr": 6.1855038950296594e-06}
{"train_loss": 0.03629043698310852, "global_step": 228095, "epoch": 2562, "lr": 6.185224590065014e-06}
{"train_loss": 0.025219980627298355, "global_step": 228096, "epoch": 2562, "lr": 6.184945290990718e-06}
{"train_loss": 0.1202021911740303, "global_step": 228097, "epoch": 2562, "lr": 6.184665997806832e-06}
{"train_loss": 0.05937453731894493, "global_step": 228098, "epoch": 2562, "lr": 6.184386710513374e-06}
{"train_loss": 0.030881643295288086, "global_step": 228099, "epoch": 2562, "lr": 6.1841074291103976e-06}
{"train_loss": 0.02471882849931717, "global_step": 228100, "epoch": 2562, "lr": 6.1838281535979256e-06}
{"train_loss": 0.041762981563806534, "global_step": 228101, "epoch": 2562, "lr": 6.1835488839760205e-06}
{"train_loss": 0.04945076256990433, "global_step": 228102, "epoch": 2562, "lr": 6.183269620244686e-06}
{"train_loss": 0.0368383564054966, "global_step": 228103, "epoch": 2562, "lr": 6.18299036240399e-06}
{"train_loss": 0.06547709554433823, "global_step": 228104, "epoch": 2562, "lr": 6.182711110453943e-06}
{"train_loss": 0.03167129307985306, "global_step": 228105, "epoch": 2562, "lr": 6.1824318643946e-06}
{"train_loss": 0.05138460126067145, "global_step": 228106, "epoch": 2562, "lr": 6.182152624226006e-06, "val_loss": 9.205676078796387}
{"train_loss": 0.07052142173051834, "global_step": 228107, "epoch": 2563, "lr": 6.181873389948173e-06}
{"train_loss": 0.0411360077559948, "global_step": 228108, "epoch": 2563, "lr": 6.181594161561166e-06}
{"train_loss": 0.06775421649217606, "global_step": 228109, "epoch": 2563, "lr": 6.181314939065003e-06}
{"train_loss": 0.022540874779224396, "global_step": 228110, "epoch": 2563, "lr": 6.181035722459733e-06}
{"train_loss": 0.04621298238635063, "global_step": 228111, "epoch": 2563, "lr": 6.1807565117453785e-06}
{"train_loss": 0.02759994938969612, "global_step": 228112, "epoch": 2563, "lr": 6.1804773069220004e-06}
{"train_loss": 0.04992509260773659, "global_step": 228113, "epoch": 2563, "lr": 6.1801981079896095e-06}
{"train_loss": 0.0892815887928009, "global_step": 228114, "epoch": 2563, "lr": 6.179918914948274e-06}
{"train_loss": 0.1028393805027008, "global_step": 228115, "epoch": 2563, "lr": 6.179639727797998e-06}
{"train_loss": 0.0876769870519638, "global_step": 228116, "epoch": 2563, "lr": 6.179360546538848e-06}
{"train_loss": 0.031004462391138077, "global_step": 228117, "epoch": 2563, "lr": 6.1790813711708365e-06}
{"train_loss": 0.04734417423605919, "global_step": 228118, "epoch": 2563, "lr": 6.178802201694023e-06}
{"train_loss": 0.06592025607824326, "global_step": 228119, "epoch": 2563, "lr": 6.178523038108425e-06}
{"train_loss": 0.025355331599712372, "global_step": 228120, "epoch": 2563, "lr": 6.178243880414103e-06}
{"train_loss": 0.08608373254537582, "global_step": 228121, "epoch": 2563, "lr": 6.177964728611074e-06}
{"train_loss": 0.04490268975496292, "global_step": 228122, "epoch": 2563, "lr": 6.177685582699388e-06}
{"train_loss": 0.05439944937825203, "global_step": 228123, "epoch": 2563, "lr": 6.177406442679074e-06}
{"train_loss": 0.021982410922646523, "global_step": 228124, "epoch": 2563, "lr": 6.177127308550168e-06}
{"train_loss": 0.023143870756030083, "global_step": 228125, "epoch": 2563, "lr": 6.176848180312728e-06}
{"train_loss": 0.06532386690378189, "global_step": 228126, "epoch": 2563, "lr": 6.176569057966775e-06}
{"train_loss": 0.07029861211776733, "global_step": 228127, "epoch": 2563, "lr": 6.176289941512337e-06}
{"train_loss": 0.0184556245803833, "global_step": 228128, "epoch": 2563, "lr": 6.176010830949474e-06}
{"train_loss": 0.048037879168987274, "global_step": 228129, "epoch": 2563, "lr": 6.1757317262782e-06}
{"train_loss": 0.04720955342054367, "global_step": 228130, "epoch": 2563, "lr": 6.1754526274985735e-06}
{"train_loss": 0.05477961525321007, "global_step": 228131, "epoch": 2563, "lr": 6.175173534610629e-06}
{"train_loss": 0.048042673617601395, "global_step": 228132, "epoch": 2563, "lr": 6.1748944476143825e-06}
{"train_loss": 0.029749630019068718, "global_step": 228133, "epoch": 2563, "lr": 6.1746153665099005e-06}
{"train_loss": 0.01853792555630207, "global_step": 228134, "epoch": 2563, "lr": 6.174336291297195e-06}
{"train_loss": 0.07356192916631699, "global_step": 228135, "epoch": 2563, "lr": 6.174057221976331e-06}
{"train_loss": 0.06355840712785721, "global_step": 228136, "epoch": 2563, "lr": 6.173778158547316e-06}
{"train_loss": 0.03571730852127075, "global_step": 228137, "epoch": 2563, "lr": 6.173499101010216e-06}
{"train_loss": 0.041894301772117615, "global_step": 228138, "epoch": 2563, "lr": 6.173220049365041e-06}
{"train_loss": 0.08956972509622574, "global_step": 228139, "epoch": 2563, "lr": 6.172941003611854e-06}
{"train_loss": 0.05451833829283714, "global_step": 228140, "epoch": 2563, "lr": 6.17266196375067e-06}
{"train_loss": 0.07153269648551941, "global_step": 228141, "epoch": 2563, "lr": 6.17238292978155e-06}
{"train_loss": 0.05144878476858139, "global_step": 228142, "epoch": 2563, "lr": 6.172103901704507e-06}
{"train_loss": 0.018802130594849586, "global_step": 228143, "epoch": 2563, "lr": 6.171824879519589e-06}
{"train_loss": 0.11898763477802277, "global_step": 228144, "epoch": 2563, "lr": 6.171545863226847e-06}
{"train_loss": 0.06918110698461533, "global_step": 228145, "epoch": 2563, "lr": 6.171266852826296e-06}
{"train_loss": 0.06535135209560394, "global_step": 228146, "epoch": 2563, "lr": 6.170987848317994e-06}
{"train_loss": 0.06732477992773056, "global_step": 228147, "epoch": 2563, "lr": 6.170708849701956e-06}
{"train_loss": 0.06077297031879425, "global_step": 228148, "epoch": 2563, "lr": 6.1704298569782495e-06}
{"train_loss": 0.03436797484755516, "global_step": 228149, "epoch": 2563, "lr": 6.1701508701468794e-06}
{"train_loss": 0.04612935334444046, "global_step": 228150, "epoch": 2563, "lr": 6.169871889207907e-06}
{"train_loss": 0.04033342003822327, "global_step": 228151, "epoch": 2563, "lr": 6.169592914161354e-06}
{"train_loss": 0.07874389737844467, "global_step": 228152, "epoch": 2563, "lr": 6.169313945007271e-06}
{"train_loss": 0.04843878373503685, "global_step": 228153, "epoch": 2563, "lr": 6.169034981745681e-06}
{"train_loss": 0.05679983273148537, "global_step": 228154, "epoch": 2563, "lr": 6.168756024376643e-06}
{"train_loss": 0.05280561372637749, "global_step": 228155, "epoch": 2563, "lr": 6.168477072900169e-06}
{"train_loss": 0.03268587589263916, "global_step": 228156, "epoch": 2563, "lr": 6.168198127316321e-06}
{"train_loss": 0.04819326475262642, "global_step": 228157, "epoch": 2563, "lr": 6.1679191876251096e-06}
{"train_loss": 0.045174747705459595, "global_step": 228158, "epoch": 2563, "lr": 6.167640253826601e-06}
{"train_loss": 0.043457143008708954, "global_step": 228159, "epoch": 2563, "lr": 6.167361325920812e-06}
{"train_loss": 0.08697113394737244, "global_step": 228160, "epoch": 2563, "lr": 6.167082403907781e-06}
{"train_loss": 0.027846837416291237, "global_step": 228161, "epoch": 2563, "lr": 6.166803487787565e-06}
{"train_loss": 0.05001009628176689, "global_step": 228162, "epoch": 2563, "lr": 6.166524577560179e-06}
{"train_loss": 0.06755479425191879, "global_step": 228163, "epoch": 2563, "lr": 6.166245673225679e-06}
{"train_loss": 0.037303775548934937, "global_step": 228164, "epoch": 2563, "lr": 6.165966774784087e-06}
{"train_loss": 0.016651403158903122, "global_step": 228165, "epoch": 2563, "lr": 6.165687882235444e-06}
{"train_loss": 0.06083949655294418, "global_step": 228166, "epoch": 2563, "lr": 6.165408995579797e-06}
{"train_loss": 0.046215299516916275, "global_step": 228167, "epoch": 2563, "lr": 6.165130114817169e-06}
{"train_loss": 0.04465145617723465, "global_step": 228168, "epoch": 2563, "lr": 6.164851239947611e-06}
{"train_loss": 0.05902332440018654, "global_step": 228169, "epoch": 2563, "lr": 6.164572370971155e-06}
{"train_loss": 0.03297974169254303, "global_step": 228170, "epoch": 2563, "lr": 6.16429350788783e-06}
{"train_loss": 0.04855688288807869, "global_step": 228171, "epoch": 2563, "lr": 6.164014650697691e-06}
{"train_loss": 0.019810307770967484, "global_step": 228172, "epoch": 2563, "lr": 6.163735799400755e-06}
{"train_loss": 0.0752820074558258, "global_step": 228173, "epoch": 2563, "lr": 6.163456953997082e-06}
{"train_loss": 0.026008833199739456, "global_step": 228174, "epoch": 2563, "lr": 6.16317811448669e-06}
{"train_loss": 0.03749166429042816, "global_step": 228175, "epoch": 2563, "lr": 6.162899280869627e-06}
{"train_loss": 0.022215349599719048, "global_step": 228176, "epoch": 2563, "lr": 6.162620453145923e-06}
{"train_loss": 0.03060767613351345, "global_step": 228177, "epoch": 2563, "lr": 6.162341631315633e-06}
{"train_loss": 0.07522488385438919, "global_step": 228178, "epoch": 2563, "lr": 6.162062815378766e-06}
{"train_loss": 0.047113656997680664, "global_step": 228179, "epoch": 2563, "lr": 6.1617840053353746e-06}
{"train_loss": 0.013674171641469002, "global_step": 228180, "epoch": 2563, "lr": 6.161505201185514e-06}
{"train_loss": 0.06986259669065475, "global_step": 228181, "epoch": 2563, "lr": 6.161226402929188e-06}
{"train_loss": 0.04340372979640961, "global_step": 228182, "epoch": 2563, "lr": 6.160947610566465e-06}
{"train_loss": 0.06304595619440079, "global_step": 228183, "epoch": 2563, "lr": 6.160668824097359e-06}
{"train_loss": 0.019306421279907227, "global_step": 228184, "epoch": 2563, "lr": 6.160390043521924e-06}
{"train_loss": 0.11452139168977737, "global_step": 228185, "epoch": 2563, "lr": 6.16011126884018e-06}
{"train_loss": 0.024203162640333176, "global_step": 228186, "epoch": 2563, "lr": 6.159832500052187e-06}
{"train_loss": 0.032838549464941025, "global_step": 228187, "epoch": 2563, "lr": 6.159553737157958e-06}
{"train_loss": 0.0542648620903492, "global_step": 228188, "epoch": 2563, "lr": 6.159274980157553e-06}
{"train_loss": 0.08186127990484238, "global_step": 228189, "epoch": 2563, "lr": 6.158996229050989e-06}
{"train_loss": 0.027487127110362053, "global_step": 228190, "epoch": 2563, "lr": 6.158717483838328e-06}
{"train_loss": 0.027621470391750336, "global_step": 228191, "epoch": 2563, "lr": 6.158438744519579e-06}
{"train_loss": 0.09129229187965393, "global_step": 228192, "epoch": 2563, "lr": 6.158160011094805e-06}
{"train_loss": 0.048153750598430634, "global_step": 228193, "epoch": 2563, "lr": 6.1578812835640205e-06}
{"train_loss": 0.03843485191464424, "global_step": 228194, "epoch": 2563, "lr": 6.157602561927284e-06}
{"train_loss": 0.05110427653521634, "global_step": 228195, "epoch": 2563, "lr": 6.157323846184615e-06, "val_loss": 9.212163925170898}
{"train_loss": 0.06295423954725266, "global_step": 228196, "epoch": 2564, "lr": 6.157045136336059e-06}
{"train_loss": 0.03907056897878647, "global_step": 228197, "epoch": 2564, "lr": 6.156766432381661e-06}
{"train_loss": 0.047409538179636, "global_step": 228198, "epoch": 2564, "lr": 6.156487734321448e-06}
{"train_loss": 0.05522102490067482, "global_step": 228199, "epoch": 2564, "lr": 6.1562090421554644e-06}
{"train_loss": 0.11519494652748108, "global_step": 228200, "epoch": 2564, "lr": 6.155930355883738e-06}
{"train_loss": 0.054001208394765854, "global_step": 228201, "epoch": 2564, "lr": 6.1556516755063185e-06}
{"train_loss": 0.048305071890354156, "global_step": 228202, "epoch": 2564, "lr": 6.15537300102324e-06}
{"train_loss": 0.015781939029693604, "global_step": 228203, "epoch": 2564, "lr": 6.15509433243453e-06}
{"train_loss": 0.0218876451253891, "global_step": 228204, "epoch": 2564, "lr": 6.154815669740238e-06}
{"train_loss": 0.03469596058130264, "global_step": 228205, "epoch": 2564, "lr": 6.154537012940392e-06}
{"train_loss": 0.020777394995093346, "global_step": 228206, "epoch": 2564, "lr": 6.154258362035037e-06}
{"train_loss": 0.01832481659948826, "global_step": 228207, "epoch": 2564, "lr": 6.1539797170242055e-06}
{"train_loss": 0.05117025226354599, "global_step": 228208, "epoch": 2564, "lr": 6.153701077907942e-06}
{"train_loss": 0.05850556865334511, "global_step": 228209, "epoch": 2564, "lr": 6.1534224446862796e-06}
{"train_loss": 0.07581865042448044, "global_step": 228210, "epoch": 2564, "lr": 6.1531438173592475e-06}
{"train_loss": 0.07341765612363815, "global_step": 228211, "epoch": 2564, "lr": 6.152865195926899e-06}
{"train_loss": 0.05370672792196274, "global_step": 228212, "epoch": 2564, "lr": 6.1525865803892525e-06}
{"train_loss": 0.08258585631847382, "global_step": 228213, "epoch": 2564, "lr": 6.1523079707463684e-06}
{"train_loss": 0.038437969982624054, "global_step": 228214, "epoch": 2564, "lr": 6.152029366998258e-06}
{"train_loss": 0.05842055007815361, "global_step": 228215, "epoch": 2564, "lr": 6.151750769144976e-06}
{"train_loss": 0.039154451340436935, "global_step": 228216, "epoch": 2564, "lr": 6.151472177186568e-06}
{"train_loss": 0.041510265320539474, "global_step": 228217, "epoch": 2564, "lr": 6.151193591123045e-06}
{"train_loss": 0.09607799351215363, "global_step": 228218, "epoch": 2564, "lr": 6.150915010954472e-06}
{"train_loss": 0.02172096073627472, "global_step": 228219, "epoch": 2564, "lr": 6.150636436680868e-06}
{"train_loss": 0.04944496601819992, "global_step": 228220, "epoch": 2564, "lr": 6.150357868302281e-06}
{"train_loss": 0.04033414646983147, "global_step": 228221, "epoch": 2564, "lr": 6.15007930581874e-06}
{"train_loss": 0.06462362408638, "global_step": 228222, "epoch": 2564, "lr": 6.149800749230289e-06}
{"train_loss": 0.03648713231086731, "global_step": 228223, "epoch": 2564, "lr": 6.149522198536956e-06}
{"train_loss": 0.05187588557600975, "global_step": 228224, "epoch": 2564, "lr": 6.149243653738795e-06}
{"train_loss": 0.040455810725688934, "global_step": 228225, "epoch": 2564, "lr": 6.148965114835825e-06}
{"train_loss": 0.03667229041457176, "global_step": 228226, "epoch": 2564, "lr": 6.148686581828106e-06}
{"train_loss": 0.0274814460426569, "global_step": 228227, "epoch": 2564, "lr": 6.148408054715643e-06}
{"train_loss": 0.04935397580265999, "global_step": 228228, "epoch": 2564, "lr": 6.148129533498509e-06}
{"train_loss": 0.02022814378142357, "global_step": 228229, "epoch": 2564, "lr": 6.147851018176709e-06}
{"train_loss": 0.05605518817901611, "global_step": 228230, "epoch": 2564, "lr": 6.14757250875031e-06}
{"train_loss": 0.032890819013118744, "global_step": 228231, "epoch": 2564, "lr": 6.147294005219329e-06}
{"train_loss": 0.04257207363843918, "global_step": 228232, "epoch": 2564, "lr": 6.147015507583803e-06}
{"train_loss": 0.03219674527645111, "global_step": 228233, "epoch": 2564, "lr": 6.14673701584379e-06}
{"train_loss": 0.06153571978211403, "global_step": 228234, "epoch": 2564, "lr": 6.146458529999299e-06}
{"train_loss": 0.034215930849313736, "global_step": 228235, "epoch": 2564, "lr": 6.146180050050393e-06}
{"train_loss": 0.06742562353610992, "global_step": 228236, "epoch": 2564, "lr": 6.145901575997093e-06}
{"train_loss": 0.0414203479886055, "global_step": 228237, "epoch": 2564, "lr": 6.145623107839449e-06}
{"train_loss": 0.0567828044295311, "global_step": 228238, "epoch": 2564, "lr": 6.145344645577483e-06}
{"train_loss": 0.06470516324043274, "global_step": 228239, "epoch": 2564, "lr": 6.145066189211252e-06}
{"train_loss": 0.041760656982660294, "global_step": 228240, "epoch": 2564, "lr": 6.14478773874077e-06}
{"train_loss": 0.03929559886455536, "global_step": 228241, "epoch": 2564, "lr": 6.144509294166095e-06}
{"train_loss": 0.029079662635922432, "global_step": 228242, "epoch": 2564, "lr": 6.144230855487259e-06}
{"train_loss": 0.022273268550634384, "global_step": 228243, "epoch": 2564, "lr": 6.143952422704285e-06}
{"train_loss": 0.04437239095568657, "global_step": 228244, "epoch": 2564, "lr": 6.143673995817234e-06}
{"train_loss": 0.04424851015210152, "global_step": 228245, "epoch": 2564, "lr": 6.143395574826122e-06}
{"train_loss": 0.03214646503329277, "global_step": 228246, "epoch": 2564, "lr": 6.1431171597310045e-06}
{"train_loss": 0.04705766960978508, "global_step": 228247, "epoch": 2564, "lr": 6.14283875053191e-06}
{"train_loss": 0.05442134290933609, "global_step": 228248, "epoch": 2564, "lr": 6.1425603472288705e-06}
{"train_loss": 0.04135331138968468, "global_step": 228249, "epoch": 2564, "lr": 6.142281949821932e-06}
{"train_loss": 0.02774006314575672, "global_step": 228250, "epoch": 2564, "lr": 6.142003558311121e-06}
{"train_loss": 0.09534234553575516, "global_step": 228251, "epoch": 2564, "lr": 6.141725172696488e-06}
{"train_loss": 0.03542512282729149, "global_step": 228252, "epoch": 2564, "lr": 6.141446792978073e-06}
{"train_loss": 0.04658201336860657, "global_step": 228253, "epoch": 2564, "lr": 6.141168419155896e-06}
{"train_loss": 0.01807085983455181, "global_step": 228254, "epoch": 2564, "lr": 6.140890051230014e-06}
{"train_loss": 0.03450779616832733, "global_step": 228255, "epoch": 2564, "lr": 6.140611689200443e-06}
{"train_loss": 0.0521516278386116, "global_step": 228256, "epoch": 2564, "lr": 6.140333333067244e-06}
{"train_loss": 0.05296945199370384, "global_step": 228257, "epoch": 2564, "lr": 6.140054982830429e-06}
{"train_loss": 0.0229568500071764, "global_step": 228258, "epoch": 2564, "lr": 6.139776638490064e-06}
{"train_loss": 0.028993751853704453, "global_step": 228259, "epoch": 2564, "lr": 6.1394983000461595e-06}
{"train_loss": 0.08650203794240952, "global_step": 228260, "epoch": 2564, "lr": 6.139219967498772e-06}
{"train_loss": 0.025656579062342644, "global_step": 228261, "epoch": 2564, "lr": 6.138941640847928e-06}
{"train_loss": 0.04488183930516243, "global_step": 228262, "epoch": 2564, "lr": 6.1386633200936735e-06}
{"train_loss": 0.07645826786756516, "global_step": 228263, "epoch": 2564, "lr": 6.138385005236036e-06}
{"train_loss": 0.03718949481844902, "global_step": 228264, "epoch": 2564, "lr": 6.138106696275064e-06}
{"train_loss": 0.04203145205974579, "global_step": 228265, "epoch": 2564, "lr": 6.137828393210776e-06}
{"train_loss": 0.026679836213588715, "global_step": 228266, "epoch": 2564, "lr": 6.137550096043237e-06}
{"train_loss": 0.04368475824594498, "global_step": 228267, "epoch": 2564, "lr": 6.13727180477246e-06}
{"train_loss": 0.05906587839126587, "global_step": 228268, "epoch": 2564, "lr": 6.136993519398504e-06}
{"train_loss": 0.026533309370279312, "global_step": 228269, "epoch": 2564, "lr": 6.136715239921381e-06}
{"train_loss": 0.060285743325948715, "global_step": 228270, "epoch": 2564, "lr": 6.1364369663411415e-06}
{"train_loss": 0.05475887656211853, "global_step": 228271, "epoch": 2564, "lr": 6.136158698657834e-06}
{"train_loss": 0.04789001867175102, "global_step": 228272, "epoch": 2564, "lr": 6.135880436871477e-06}
{"train_loss": 0.03975191339850426, "global_step": 228273, "epoch": 2564, "lr": 6.135602180982125e-06}
{"train_loss": 0.06566814333200455, "global_step": 228274, "epoch": 2564, "lr": 6.1353239309898e-06}
{"train_loss": 0.04507073014974594, "global_step": 228275, "epoch": 2564, "lr": 6.135045686894553e-06}
{"train_loss": 0.051094550639390945, "global_step": 228276, "epoch": 2564, "lr": 6.134767448696405e-06}
{"train_loss": 0.04567820206284523, "global_step": 228277, "epoch": 2564, "lr": 6.134489216395411e-06}
{"train_loss": 0.029803065583109856, "global_step": 228278, "epoch": 2564, "lr": 6.13421098999159e-06}
{"train_loss": 0.05585392192006111, "global_step": 228279, "epoch": 2564, "lr": 6.133932769485001e-06}
{"train_loss": 0.033021584153175354, "global_step": 228280, "epoch": 2564, "lr": 6.133654554875673e-06}
{"train_loss": 0.02007140778005123, "global_step": 228281, "epoch": 2564, "lr": 6.133376346163627e-06}
{"train_loss": 0.029013749212026596, "global_step": 228282, "epoch": 2564, "lr": 6.133098143348925e-06}
{"train_loss": 0.05531196668744087, "global_step": 228283, "epoch": 2564, "lr": 6.132819946431584e-06}
{"train_loss": 0.04618358113959934, "global_step": 228284, "epoch": 2564, "lr": 6.132541755411658e-06, "val_loss": 9.106590270996094}
{"train_loss": 0.04759783297777176, "global_step": 228285, "epoch": 2565, "lr": 6.132263570289182e-06}
{"train_loss": 0.07007059454917908, "global_step": 228286, "epoch": 2565, "lr": 6.131985391064177e-06}
{"train_loss": 0.03128596022725105, "global_step": 228287, "epoch": 2565, "lr": 6.131707217736688e-06}
{"train_loss": 0.059458259493112564, "global_step": 228288, "epoch": 2565, "lr": 6.1314290503067695e-06}
{"train_loss": 0.02490353025496006, "global_step": 228289, "epoch": 2565, "lr": 6.131150888774434e-06}
{"train_loss": 0.08316358178853989, "global_step": 228290, "epoch": 2565, "lr": 6.130872733139742e-06}
{"train_loss": 0.09784480929374695, "global_step": 228291, "epoch": 2565, "lr": 6.13059458340271e-06}
{"train_loss": 0.04635605961084366, "global_step": 228292, "epoch": 2565, "lr": 6.130316439563394e-06}
{"train_loss": 0.04054497927427292, "global_step": 228293, "epoch": 2565, "lr": 6.13003830162181e-06}
{"train_loss": 0.04597483202815056, "global_step": 228294, "epoch": 2565, "lr": 6.1297601695780195e-06}
{"train_loss": 0.0572507418692112, "global_step": 228295, "epoch": 2565, "lr": 6.129482043432039e-06}
{"train_loss": 0.04286527633666992, "global_step": 228296, "epoch": 2565, "lr": 6.1292039231839235e-06}
{"train_loss": 0.03204715996980667, "global_step": 228297, "epoch": 2565, "lr": 6.128925808833691e-06}
{"train_loss": 0.012844614684581757, "global_step": 228298, "epoch": 2565, "lr": 6.128647700381401e-06}
{"train_loss": 0.044875722378492355, "global_step": 228299, "epoch": 2565, "lr": 6.128369597827072e-06}
{"train_loss": 0.10134846717119217, "global_step": 228300, "epoch": 2565, "lr": 6.128091501170757e-06}
{"train_loss": 0.04827774316072464, "global_step": 228301, "epoch": 2565, "lr": 6.1278134104124756e-06}
{"train_loss": 0.05122651159763336, "global_step": 228302, "epoch": 2565, "lr": 6.127535325552286e-06}
{"train_loss": 0.05476778745651245, "global_step": 228303, "epoch": 2565, "lr": 6.1272572465902014e-06}
{"train_loss": 0.052193839102983475, "global_step": 228304, "epoch": 2565, "lr": 6.126979173526282e-06}
{"train_loss": 0.054237544536590576, "global_step": 228305, "epoch": 2565, "lr": 6.1267011063605494e-06}
{"train_loss": 0.08330624550580978, "global_step": 228306, "epoch": 2565, "lr": 6.126423045093044e-06}
{"train_loss": 0.06432238966226578, "global_step": 228307, "epoch": 2565, "lr": 6.12614498972382e-06}
{"train_loss": 0.06046426296234131, "global_step": 228308, "epoch": 2565, "lr": 6.125866940252883e-06}
{"train_loss": 0.04818582534790039, "global_step": 228309, "epoch": 2565, "lr": 6.125588896680306e-06}
{"train_loss": 0.047860387712717056, "global_step": 228310, "epoch": 2565, "lr": 6.1253108590061005e-06}
{"train_loss": 0.022303888574242592, "global_step": 228311, "epoch": 2565, "lr": 6.125032827230315e-06}
{"train_loss": 0.04064086824655533, "global_step": 228312, "epoch": 2565, "lr": 6.124754801352977e-06}
{"train_loss": 0.023182427510619164, "global_step": 228313, "epoch": 2565, "lr": 6.1244767813741445e-06}
{"train_loss": 0.03207632154226303, "global_step": 228314, "epoch": 2565, "lr": 6.1241987672938264e-06}
{"train_loss": 0.06240830197930336, "global_step": 228315, "epoch": 2565, "lr": 6.123920759112089e-06}
{"train_loss": 0.09199932217597961, "global_step": 228316, "epoch": 2565, "lr": 6.12364275682894e-06}
{"train_loss": 0.06585061550140381, "global_step": 228317, "epoch": 2565, "lr": 6.12336476044445e-06}
{"train_loss": 0.03838726878166199, "global_step": 228318, "epoch": 2565, "lr": 6.123086769958636e-06}
{"train_loss": 0.049066171050071716, "global_step": 228319, "epoch": 2565, "lr": 6.122808785371526e-06}
{"train_loss": 0.06976427882909775, "global_step": 228320, "epoch": 2565, "lr": 6.12253080668318e-06}
{"train_loss": 0.0597020760178566, "global_step": 228321, "epoch": 2565, "lr": 6.122252833893616e-06}
{"train_loss": 0.0942816212773323, "global_step": 228322, "epoch": 2565, "lr": 6.121974867002889e-06}
{"train_loss": 0.07667536288499832, "global_step": 228323, "epoch": 2565, "lr": 6.1216969060110205e-06}
{"train_loss": 0.049025971442461014, "global_step": 228324, "epoch": 2565, "lr": 6.121418950918062e-06}
{"train_loss": 0.024866163730621338, "global_step": 228325, "epoch": 2565, "lr": 6.121141001724034e-06}
{"train_loss": 0.04637446254491806, "global_step": 228326, "epoch": 2565, "lr": 6.120863058428994e-06}
{"train_loss": 0.010177784599363804, "global_step": 228327, "epoch": 2565, "lr": 6.120585121032962e-06}
{"train_loss": 0.06085885316133499, "global_step": 228328, "epoch": 2565, "lr": 6.12030718953599e-06}
{"train_loss": 0.06332819163799286, "global_step": 228329, "epoch": 2565, "lr": 6.120029263938098e-06}
{"train_loss": 0.03870333358645439, "global_step": 228330, "epoch": 2565, "lr": 6.119751344239344e-06}
{"train_loss": 0.04740860313177109, "global_step": 228331, "epoch": 2565, "lr": 6.119473430439743e-06}
{"train_loss": 0.03317689895629883, "global_step": 228332, "epoch": 2565, "lr": 6.119195522539356e-06}
{"train_loss": 0.055726926773786545, "global_step": 228333, "epoch": 2565, "lr": 6.1189176205382004e-06}
{"train_loss": 0.05733398348093033, "global_step": 228334, "epoch": 2565, "lr": 6.118639724436326e-06}
{"train_loss": 0.0244306568056345, "global_step": 228335, "epoch": 2565, "lr": 6.11836183423376e-06}
{"train_loss": 0.007569110952317715, "global_step": 228336, "epoch": 2565, "lr": 6.118083949930553e-06}
{"train_loss": 0.1084943562746048, "global_step": 228337, "epoch": 2565, "lr": 6.117806071526727e-06}
{"train_loss": 0.09595990180969238, "global_step": 228338, "epoch": 2565, "lr": 6.117528199022332e-06}
{"train_loss": 0.05332474038004875, "global_step": 228339, "epoch": 2565, "lr": 6.1172503324173955e-06}
{"train_loss": 0.03935529664158821, "global_step": 228340, "epoch": 2565, "lr": 6.116972471711968e-06}
{"train_loss": 0.02577051892876625, "global_step": 228341, "epoch": 2565, "lr": 6.116694616906071e-06}
{"train_loss": 0.04742666333913803, "global_step": 228342, "epoch": 2565, "lr": 6.1164167679997445e-06}
{"train_loss": 0.06859840452671051, "global_step": 228343, "epoch": 2565, "lr": 6.116138924993048e-06}
{"train_loss": 0.0961582288146019, "global_step": 228344, "epoch": 2565, "lr": 6.115861087885988e-06}
{"train_loss": 0.04329754784703255, "global_step": 228345, "epoch": 2565, "lr": 6.115583256678626e-06}
{"train_loss": 0.027538979426026344, "global_step": 228346, "epoch": 2565, "lr": 6.115305431370977e-06}
{"train_loss": 0.05522748455405235, "global_step": 228347, "epoch": 2565, "lr": 6.115027611963098e-06}
{"train_loss": 0.0770232081413269, "global_step": 228348, "epoch": 2565, "lr": 6.1147497984550165e-06}
{"train_loss": 0.08398506045341492, "global_step": 228349, "epoch": 2565, "lr": 6.114471990846776e-06}
{"train_loss": 0.05563309043645859, "global_step": 228350, "epoch": 2565, "lr": 6.1141941891384e-06}
{"train_loss": 0.06317109614610672, "global_step": 228351, "epoch": 2565, "lr": 6.1139163933299495e-06}
{"train_loss": 0.07556023448705673, "global_step": 228352, "epoch": 2565, "lr": 6.113638603421434e-06}
{"train_loss": 0.06310496479272842, "global_step": 228353, "epoch": 2565, "lr": 6.113360819412917e-06}
{"train_loss": 0.050872448831796646, "global_step": 228354, "epoch": 2565, "lr": 6.113083041304412e-06}
{"train_loss": 0.02117752470076084, "global_step": 228355, "epoch": 2565, "lr": 6.112805269095984e-06}
{"train_loss": 0.03992564231157303, "global_step": 228356, "epoch": 2565, "lr": 6.11252750278764e-06}
{"train_loss": 0.08394104987382889, "global_step": 228357, "epoch": 2565, "lr": 6.112249742379439e-06}
{"train_loss": 0.049885720014572144, "global_step": 228358, "epoch": 2565, "lr": 6.1119719878714176e-06}
{"train_loss": 0.05592293292284012, "global_step": 228359, "epoch": 2565, "lr": 6.111694239263588e-06}
{"train_loss": 0.023512283340096474, "global_step": 228360, "epoch": 2565, "lr": 6.111416496556021e-06}
{"train_loss": 0.04042147845029831, "global_step": 228361, "epoch": 2565, "lr": 6.11113875974873e-06}
{"train_loss": 0.047079119831323624, "global_step": 228362, "epoch": 2565, "lr": 6.1108610288417745e-06}
{"train_loss": 0.07388179749250412, "global_step": 228363, "epoch": 2565, "lr": 6.110583303835166e-06}
{"train_loss": 0.043280474841594696, "global_step": 228364, "epoch": 2565, "lr": 6.110305584728959e-06}
{"train_loss": 0.023848332464694977, "global_step": 228365, "epoch": 2565, "lr": 6.1100278715231825e-06}
{"train_loss": 0.06282974779605865, "global_step": 228366, "epoch": 2565, "lr": 6.109750164217887e-06}
{"train_loss": 0.0794677734375, "global_step": 228367, "epoch": 2565, "lr": 6.109472462813093e-06}
{"train_loss": 0.026200253516435623, "global_step": 228368, "epoch": 2565, "lr": 6.109194767308851e-06}
{"train_loss": 0.033487580716609955, "global_step": 228369, "epoch": 2565, "lr": 6.108917077705184e-06}
{"train_loss": 0.05834152549505234, "global_step": 228370, "epoch": 2565, "lr": 6.108639394002147e-06}
{"train_loss": 0.029468368738889694, "global_step": 228371, "epoch": 2565, "lr": 6.108361716199762e-06}
{"train_loss": 0.06525241583585739, "global_step": 228372, "epoch": 2565, "lr": 6.1080840442980795e-06}
{"train_loss": 0.05256313987578569, "global_step": 228373, "epoch": 2565, "lr": 6.107806378297121e-06, "val_loss": 9.048611640930176, "train_action_mse_error": 10.437520027160645}
{"train_loss": 0.03547156974673271, "global_step": 228374, "epoch": 2566, "lr": 6.107528718196942e-06}
{"train_loss": 0.06194041296839714, "global_step": 228375, "epoch": 2566, "lr": 6.107251063997566e-06}
{"train_loss": 0.09392920136451721, "global_step": 228376, "epoch": 2566, "lr": 6.106973415699041e-06}
{"train_loss": 0.05035747215151787, "global_step": 228377, "epoch": 2566, "lr": 6.1066957733013855e-06}
{"train_loss": 0.08221757411956787, "global_step": 228378, "epoch": 2566, "lr": 6.106418136804654e-06}
{"train_loss": 0.037545595318078995, "global_step": 228379, "epoch": 2566, "lr": 6.10614050620889e-06}
{"train_loss": 0.07041066139936447, "global_step": 228380, "epoch": 2566, "lr": 6.105862881514107e-06}
{"train_loss": 0.07866093516349792, "global_step": 228381, "epoch": 2566, "lr": 6.10558526272037e-06}
{"train_loss": 0.08666941523551941, "global_step": 228382, "epoch": 2566, "lr": 6.105307649827691e-06}
{"train_loss": 0.07727382332086563, "global_step": 228383, "epoch": 2566, "lr": 6.105030042836124e-06}
{"train_loss": 0.0840090662240982, "global_step": 228384, "epoch": 2566, "lr": 6.104752441745692e-06}
{"train_loss": 0.056783996522426605, "global_step": 228385, "epoch": 2566, "lr": 6.104474846556457e-06}
{"train_loss": 0.06907003372907639, "global_step": 228386, "epoch": 2566, "lr": 6.1041972572684235e-06}
{"train_loss": 0.09210195392370224, "global_step": 228387, "epoch": 2566, "lr": 6.103919673881659e-06}
{"train_loss": 0.025159945711493492, "global_step": 228388, "epoch": 2566, "lr": 6.103642096396178e-06}
{"train_loss": 0.02305302396416664, "global_step": 228389, "epoch": 2566, "lr": 6.103364524812039e-06}
{"train_loss": 0.04955703765153885, "global_step": 228390, "epoch": 2566, "lr": 6.10308695912925e-06}
{"train_loss": 0.027820542454719543, "global_step": 228391, "epoch": 2566, "lr": 6.102809399347881e-06}
{"train_loss": 0.03573622554540634, "global_step": 228392, "epoch": 2566, "lr": 6.102531845467946e-06}
{"train_loss": 0.040631700307130814, "global_step": 228393, "epoch": 2566, "lr": 6.102254297489502e-06}
{"train_loss": 0.0476260706782341, "global_step": 228394, "epoch": 2566, "lr": 6.101976755412558e-06}
{"train_loss": 0.039084479212760925, "global_step": 228395, "epoch": 2566, "lr": 6.101699219237184e-06}
{"train_loss": 0.02677089162170887, "global_step": 228396, "epoch": 2566, "lr": 6.101421688963399e-06}
{"train_loss": 0.03655386343598366, "global_step": 228397, "epoch": 2566, "lr": 6.101144164591233e-06}
{"train_loss": 0.06381623446941376, "global_step": 228398, "epoch": 2566, "lr": 6.10086664612074e-06}
{"train_loss": 0.023846060037612915, "global_step": 228399, "epoch": 2566, "lr": 6.100589133551943e-06}
{"train_loss": 0.11820560693740845, "global_step": 228400, "epoch": 2566, "lr": 6.100311626884897e-06}
{"train_loss": 0.07866048067808151, "global_step": 228401, "epoch": 2566, "lr": 6.100034126119619e-06}
{"train_loss": 0.01693175546824932, "global_step": 228402, "epoch": 2566, "lr": 6.09975663125617e-06}
{"train_loss": 0.03442561253905296, "global_step": 228403, "epoch": 2566, "lr": 6.099479142294556e-06}
{"train_loss": 0.05953160673379898, "global_step": 228404, "epoch": 2566, "lr": 6.0992016592348485e-06}
{"train_loss": 0.0702117532491684, "global_step": 228405, "epoch": 2566, "lr": 6.09892418207706e-06}
{"train_loss": 0.041928090155124664, "global_step": 228406, "epoch": 2566, "lr": 6.098646710821237e-06}
{"train_loss": 0.041744813323020935, "global_step": 228407, "epoch": 2566, "lr": 6.0983692454674115e-06}
{"train_loss": 0.03606975078582764, "global_step": 228408, "epoch": 2566, "lr": 6.098091786015636e-06}
{"train_loss": 0.04550396278500557, "global_step": 228409, "epoch": 2566, "lr": 6.097814332465923e-06}
{"train_loss": 0.05985923111438751, "global_step": 228410, "epoch": 2566, "lr": 6.097536884818339e-06}
{"train_loss": 0.057868607342243195, "global_step": 228411, "epoch": 2566, "lr": 6.097259443072894e-06}
{"train_loss": 0.05219949781894684, "global_step": 228412, "epoch": 2566, "lr": 6.096982007229646e-06}
{"train_loss": 0.05673901364207268, "global_step": 228413, "epoch": 2566, "lr": 6.096704577288614e-06}
{"train_loss": 0.042995844036340714, "global_step": 228414, "epoch": 2566, "lr": 6.096427153249845e-06}
{"train_loss": 0.04705357179045677, "global_step": 228415, "epoch": 2566, "lr": 6.096149735113388e-06}
{"train_loss": 0.0629730373620987, "global_step": 228416, "epoch": 2566, "lr": 6.095872322879254e-06}
{"train_loss": 0.05910854786634445, "global_step": 228417, "epoch": 2566, "lr": 6.095594916547509e-06}
{"train_loss": 0.008323557674884796, "global_step": 228418, "epoch": 2566, "lr": 6.095317516118166e-06}
{"train_loss": 0.028269417583942413, "global_step": 228419, "epoch": 2566, "lr": 6.0950401215912835e-06}
{"train_loss": 0.023564225062727928, "global_step": 228420, "epoch": 2566, "lr": 6.094762732966874e-06}
{"train_loss": 0.020522745326161385, "global_step": 228421, "epoch": 2566, "lr": 6.0944853502449995e-06}
{"train_loss": 0.03188247233629227, "global_step": 228422, "epoch": 2566, "lr": 6.094207973425681e-06}
{"train_loss": 0.032683342695236206, "global_step": 228423, "epoch": 2566, "lr": 6.093930602508963e-06}
{"train_loss": 0.04867817461490631, "global_step": 228424, "epoch": 2566, "lr": 6.0936532374948784e-06}
{"train_loss": 0.05241590738296509, "global_step": 228425, "epoch": 2566, "lr": 6.093375878383473e-06}
{"train_loss": 0.08842261135578156, "global_step": 228426, "epoch": 2566, "lr": 6.093098525174767e-06}
{"train_loss": 0.026194563135504723, "global_step": 228427, "epoch": 2566, "lr": 6.092821177868824e-06}
{"train_loss": 0.01676732674241066, "global_step": 228428, "epoch": 2566, "lr": 6.092543836465653e-06}
{"train_loss": 0.12036273628473282, "global_step": 228429, "epoch": 2566, "lr": 6.09226650096531e-06}
{"train_loss": 0.027649756520986557, "global_step": 228430, "epoch": 2566, "lr": 6.091989171367823e-06}
{"train_loss": 0.05210735648870468, "global_step": 228431, "epoch": 2566, "lr": 6.091711847673237e-06}
{"train_loss": 0.04897617921233177, "global_step": 228432, "epoch": 2566, "lr": 6.091434529881579e-06}
{"train_loss": 0.07638493180274963, "global_step": 228433, "epoch": 2566, "lr": 6.091157217992904e-06}
{"train_loss": 0.0657988041639328, "global_step": 228434, "epoch": 2566, "lr": 6.090879912007225e-06}
{"train_loss": 0.053037453442811966, "global_step": 228435, "epoch": 2566, "lr": 6.090602611924606e-06}
{"train_loss": 0.032475363463163376, "global_step": 228436, "epoch": 2566, "lr": 6.090325317745066e-06}
{"train_loss": 0.07836735993623734, "global_step": 228437, "epoch": 2566, "lr": 6.090048029468637e-06}
{"train_loss": 0.036887019872665405, "global_step": 228438, "epoch": 2566, "lr": 6.089770747095375e-06}
{"train_loss": 0.06393857300281525, "global_step": 228439, "epoch": 2566, "lr": 6.089493470625302e-06}
{"train_loss": 0.028967538848519325, "global_step": 228440, "epoch": 2566, "lr": 6.089216200058468e-06}
{"train_loss": 0.05967763066291809, "global_step": 228441, "epoch": 2566, "lr": 6.088938935394894e-06}
{"train_loss": 0.025610288605093956, "global_step": 228442, "epoch": 2566, "lr": 6.088661676634638e-06}
{"train_loss": 0.05566777288913727, "global_step": 228443, "epoch": 2566, "lr": 6.088384423777715e-06}
{"train_loss": 0.056939225643873215, "global_step": 228444, "epoch": 2566, "lr": 6.088107176824187e-06}
{"train_loss": 0.027273114770650864, "global_step": 228445, "epoch": 2566, "lr": 6.087829935774064e-06}
{"train_loss": 0.08066033571958542, "global_step": 228446, "epoch": 2566, "lr": 6.087552700627408e-06}
{"train_loss": 0.04396957531571388, "global_step": 228447, "epoch": 2566, "lr": 6.087275471384235e-06}
{"train_loss": 0.04297436773777008, "global_step": 228448, "epoch": 2566, "lr": 6.086998248044606e-06}
{"train_loss": 0.024063609540462494, "global_step": 228449, "epoch": 2566, "lr": 6.086721030608533e-06}
{"train_loss": 0.04132143408060074, "global_step": 228450, "epoch": 2566, "lr": 6.08644381907606e-06}
{"train_loss": 0.08612718433141708, "global_step": 228451, "epoch": 2566, "lr": 6.086166613447247e-06}
{"train_loss": 0.06613034009933472, "global_step": 228452, "epoch": 2566, "lr": 6.085889413722102e-06}
{"train_loss": 0.07405144721269608, "global_step": 228453, "epoch": 2566, "lr": 6.0856122199006774e-06}
{"train_loss": 0.023401815444231033, "global_step": 228454, "epoch": 2566, "lr": 6.085335031983003e-06}
{"train_loss": 0.10203378647565842, "global_step": 228455, "epoch": 2566, "lr": 6.085057849969128e-06}
{"train_loss": 0.01749737188220024, "global_step": 228456, "epoch": 2566, "lr": 6.0847806738590755e-06}
{"train_loss": 0.03214835003018379, "global_step": 228457, "epoch": 2566, "lr": 6.084503503652894e-06}
{"train_loss": 0.016192704439163208, "global_step": 228458, "epoch": 2566, "lr": 6.084226339350607e-06}
{"train_loss": 0.060602352023124695, "global_step": 228459, "epoch": 2566, "lr": 6.0839491809522755e-06}
{"train_loss": 0.06814049184322357, "global_step": 228460, "epoch": 2566, "lr": 6.083672028457904e-06}
{"train_loss": 0.07564963400363922, "global_step": 228461, "epoch": 2566, "lr": 6.08339488186756e-06}
{"train_loss": 0.0516292822829793, "global_step": 228462, "epoch": 2566, "lr": 6.08311774118126e-06, "val_loss": 9.190741539001465}
{"train_loss": 0.03872612491250038, "global_step": 228463, "epoch": 2567, "lr": 6.08284060639906e-06}
{"train_loss": 0.010093249380588531, "global_step": 228464, "epoch": 2567, "lr": 6.082563477520975e-06}
{"train_loss": 0.06388243287801743, "global_step": 228465, "epoch": 2567, "lr": 6.082286354547068e-06}
{"train_loss": 0.0396810807287693, "global_step": 228466, "epoch": 2567, "lr": 6.082009237477349e-06}
{"train_loss": 0.07537531107664108, "global_step": 228467, "epoch": 2567, "lr": 6.081732126311879e-06}
{"train_loss": 0.047803446650505066, "global_step": 228468, "epoch": 2567, "lr": 6.081455021050675e-06}
{"train_loss": 0.04598139226436615, "global_step": 228469, "epoch": 2567, "lr": 6.081177921693792e-06}
{"train_loss": 0.03180413320660591, "global_step": 228470, "epoch": 2567, "lr": 6.080900828241248e-06}
{"train_loss": 0.036229223012924194, "global_step": 228471, "epoch": 2567, "lr": 6.0806237406931084e-06}
{"train_loss": 0.050762977451086044, "global_step": 228472, "epoch": 2567, "lr": 6.0803466590493795e-06}
{"train_loss": 0.030434470623731613, "global_step": 228473, "epoch": 2567, "lr": 6.0800695833101215e-06}
{"train_loss": 0.03187529742717743, "global_step": 228474, "epoch": 2567, "lr": 6.079792513475363e-06}
{"train_loss": 0.04197557270526886, "global_step": 228475, "epoch": 2567, "lr": 6.0795154495451364e-06}
{"train_loss": 0.04559243470430374, "global_step": 228476, "epoch": 2567, "lr": 6.079238391519487e-06}
{"train_loss": 0.030970945954322815, "global_step": 228477, "epoch": 2567, "lr": 6.078961339398442e-06}
{"train_loss": 0.048483990132808685, "global_step": 228478, "epoch": 2567, "lr": 6.078684293182052e-06}
{"train_loss": 0.029030421748757362, "global_step": 228479, "epoch": 2567, "lr": 6.078407252870344e-06}
{"train_loss": 0.06739575415849686, "global_step": 228480, "epoch": 2567, "lr": 6.078130218463363e-06}
{"train_loss": 0.033367324620485306, "global_step": 228481, "epoch": 2567, "lr": 6.077853189961136e-06}
{"train_loss": 0.04842016473412514, "global_step": 228482, "epoch": 2567, "lr": 6.077576167363713e-06}
{"train_loss": 0.07235075533390045, "global_step": 228483, "epoch": 2567, "lr": 6.077299150671118e-06}
{"train_loss": 0.037376031279563904, "global_step": 228484, "epoch": 2567, "lr": 6.077022139883398e-06}
{"train_loss": 0.03443802148103714, "global_step": 228485, "epoch": 2567, "lr": 6.076745135000583e-06}
{"train_loss": 0.07090341299772263, "global_step": 228486, "epoch": 2567, "lr": 6.076468136022711e-06}
{"train_loss": 0.03446910157799721, "global_step": 228487, "epoch": 2567, "lr": 6.076191142949839e-06}
{"train_loss": 0.03507983684539795, "global_step": 228488, "epoch": 2567, "lr": 6.07591415578197e-06}
{"train_loss": 0.03551005199551582, "global_step": 228489, "epoch": 2567, "lr": 6.075637174519172e-06}
{"train_loss": 0.06514922529459, "global_step": 228490, "epoch": 2567, "lr": 6.075360199161462e-06}
{"train_loss": 0.03385654464364052, "global_step": 228491, "epoch": 2567, "lr": 6.075083229708894e-06}
{"train_loss": 0.0639161691069603, "global_step": 228492, "epoch": 2567, "lr": 6.074806266161481e-06}
{"train_loss": 0.03728485852479935, "global_step": 228493, "epoch": 2567, "lr": 6.074529308519289e-06}
{"train_loss": 0.03701022267341614, "global_step": 228494, "epoch": 2567, "lr": 6.074252356782328e-06}
{"train_loss": 0.020256780087947845, "global_step": 228495, "epoch": 2567, "lr": 6.073975410950661e-06}
{"train_loss": 0.06413428485393524, "global_step": 228496, "epoch": 2567, "lr": 6.073698471024303e-06}
{"train_loss": 0.06292425841093063, "global_step": 228497, "epoch": 2567, "lr": 6.07342153700331e-06}
{"train_loss": 0.03917950764298439, "global_step": 228498, "epoch": 2567, "lr": 6.0731446088876985e-06}
{"train_loss": 0.05118090659379959, "global_step": 228499, "epoch": 2567, "lr": 6.0728676866775304e-06}
{"train_loss": 0.013286507688462734, "global_step": 228500, "epoch": 2567, "lr": 6.072590770372816e-06}
{"train_loss": 0.024110568687319756, "global_step": 228501, "epoch": 2567, "lr": 6.072313859973616e-06}
{"train_loss": 0.06091190502047539, "global_step": 228502, "epoch": 2567, "lr": 6.072036955479954e-06}
{"train_loss": 0.04098222777247429, "global_step": 228503, "epoch": 2567, "lr": 6.071760056891873e-06}
{"train_loss": 0.0487985759973526, "global_step": 228504, "epoch": 2567, "lr": 6.0714831642094074e-06}
{"train_loss": 0.03663519397377968, "global_step": 228505, "epoch": 2567, "lr": 6.071206277432589e-06}
{"train_loss": 0.06482744216918945, "global_step": 228506, "epoch": 2567, "lr": 6.070929396561475e-06}
{"train_loss": 0.05888025090098381, "global_step": 228507, "epoch": 2567, "lr": 6.070652521596088e-06}
{"train_loss": 0.06674209982156754, "global_step": 228508, "epoch": 2567, "lr": 6.070375652536453e-06}
{"train_loss": 0.04299590736627579, "global_step": 228509, "epoch": 2567, "lr": 6.070098789382633e-06}
{"train_loss": 0.05179406329989433, "global_step": 228510, "epoch": 2567, "lr": 6.069821932134645e-06}
{"train_loss": 0.0733739361166954, "global_step": 228511, "epoch": 2567, "lr": 6.0695450807925434e-06}
{"train_loss": 0.09849308431148529, "global_step": 228512, "epoch": 2567, "lr": 6.069268235356357e-06}
{"train_loss": 0.05436382815241814, "global_step": 228513, "epoch": 2567, "lr": 6.068991395826107e-06}
{"train_loss": 0.041892487555742264, "global_step": 228514, "epoch": 2567, "lr": 6.068714562201861e-06}
{"train_loss": 0.03870607540011406, "global_step": 228515, "epoch": 2567, "lr": 6.068437734483629e-06}
{"train_loss": 0.05269714817404747, "global_step": 228516, "epoch": 2567, "lr": 6.068160912671467e-06}
{"train_loss": 0.08089590072631836, "global_step": 228517, "epoch": 2567, "lr": 6.0678840967653975e-06}
{"train_loss": 0.07153797149658203, "global_step": 228518, "epoch": 2567, "lr": 6.067607286765475e-06}
{"train_loss": 0.061008814722299576, "global_step": 228519, "epoch": 2567, "lr": 6.067330482671718e-06}
{"train_loss": 0.09705878794193268, "global_step": 228520, "epoch": 2567, "lr": 6.067053684484186e-06}
{"train_loss": 0.021630652248859406, "global_step": 228521, "epoch": 2567, "lr": 6.066776892202891e-06}
{"train_loss": 0.033700130879879, "global_step": 228522, "epoch": 2567, "lr": 6.0665001058278884e-06}
{"train_loss": 0.04247284308075905, "global_step": 228523, "epoch": 2567, "lr": 6.066223325359205e-06}
{"train_loss": 0.016600849106907845, "global_step": 228524, "epoch": 2567, "lr": 6.0659465507968804e-06}
{"train_loss": 0.03794759884476662, "global_step": 228525, "epoch": 2567, "lr": 6.065669782140965e-06}
{"train_loss": 0.03945234790444374, "global_step": 228526, "epoch": 2567, "lr": 6.065393019391469e-06}
{"train_loss": 0.05381578952074051, "global_step": 228527, "epoch": 2567, "lr": 6.065116262548459e-06}
{"train_loss": 0.024081982672214508, "global_step": 228528, "epoch": 2567, "lr": 6.064839511611953e-06}
{"train_loss": 0.047584857791662216, "global_step": 228529, "epoch": 2567, "lr": 6.064562766581999e-06}
{"train_loss": 0.039023395627737045, "global_step": 228530, "epoch": 2567, "lr": 6.064286027458621e-06}
{"train_loss": 0.07565657049417496, "global_step": 228531, "epoch": 2567, "lr": 6.0640092942418736e-06}
{"train_loss": 0.061148084700107574, "global_step": 228532, "epoch": 2567, "lr": 6.063732566931773e-06}
{"train_loss": 0.02910156548023224, "global_step": 228533, "epoch": 2567, "lr": 6.063455845528382e-06}
{"train_loss": 0.06959376484155655, "global_step": 228534, "epoch": 2567, "lr": 6.06317913003171e-06}
{"train_loss": 0.06048482283949852, "global_step": 228535, "epoch": 2567, "lr": 6.062902420441818e-06}
{"train_loss": 0.0634620189666748, "global_step": 228536, "epoch": 2567, "lr": 6.0626257167587246e-06}
{"train_loss": 0.07542373985052109, "global_step": 228537, "epoch": 2567, "lr": 6.062349018982488e-06}
{"train_loss": 0.06506215035915375, "global_step": 228538, "epoch": 2567, "lr": 6.062072327113122e-06}
{"train_loss": 0.024169832468032837, "global_step": 228539, "epoch": 2567, "lr": 6.06179564115068e-06}
{"train_loss": 0.042829934507608414, "global_step": 228540, "epoch": 2567, "lr": 6.061518961095192e-06}
{"train_loss": 0.048202820122241974, "global_step": 228541, "epoch": 2567, "lr": 6.061242286946689e-06}
{"train_loss": 0.02622857689857483, "global_step": 228542, "epoch": 2567, "lr": 6.060965618705233e-06}
{"train_loss": 0.05675515532493591, "global_step": 228543, "epoch": 2567, "lr": 6.060688956370831e-06}
{"train_loss": 0.04205675050616264, "global_step": 228544, "epoch": 2567, "lr": 6.060412299943547e-06}
{"train_loss": 0.13812249898910522, "global_step": 228545, "epoch": 2567, "lr": 6.0601356494233986e-06}
{"train_loss": 0.0668070912361145, "global_step": 228546, "epoch": 2567, "lr": 6.059859004810425e-06}
{"train_loss": 0.0355512909591198, "global_step": 228547, "epoch": 2567, "lr": 6.0595823661046765e-06}
{"train_loss": 0.0765327587723732, "global_step": 228548, "epoch": 2567, "lr": 6.0593057333061695e-06}
{"train_loss": 0.08746816962957382, "global_step": 228549, "epoch": 2567, "lr": 6.059029106414971e-06}
{"train_loss": 0.07388871908187866, "global_step": 228550, "epoch": 2567, "lr": 6.0587524854310794e-06}
{"train_loss": 0.04980361688714684, "global_step": 228551, "epoch": 2567, "lr": 6.0584758703545684e-06, "val_loss": 9.128786087036133}
{"train_loss": 0.046955838799476624, "global_step": 228552, "epoch": 2568, "lr": 6.05819926118546e-06}
{"train_loss": 0.08452876657247543, "global_step": 228553, "epoch": 2568, "lr": 6.057922657923781e-06}
{"train_loss": 0.018063316121697426, "global_step": 228554, "epoch": 2568, "lr": 6.057646060569588e-06}
{"train_loss": 0.0676988735795021, "global_step": 228555, "epoch": 2568, "lr": 6.0573694691229035e-06}
{"train_loss": 0.040463875979185104, "global_step": 228556, "epoch": 2568, "lr": 6.057092883583776e-06}
{"train_loss": 0.028439927846193314, "global_step": 228557, "epoch": 2568, "lr": 6.056816303952223e-06}
{"train_loss": 0.029458917677402496, "global_step": 228558, "epoch": 2568, "lr": 6.056539730228311e-06}
{"train_loss": 0.03461333364248276, "global_step": 228559, "epoch": 2568, "lr": 6.056263162412051e-06}
{"train_loss": 0.07746417820453644, "global_step": 228560, "epoch": 2568, "lr": 6.055986600503488e-06}
{"train_loss": 0.07578752189874649, "global_step": 228561, "epoch": 2568, "lr": 6.055710044502677e-06}
{"train_loss": 0.08458709716796875, "global_step": 228562, "epoch": 2568, "lr": 6.055433494409624e-06}
{"train_loss": 0.09950000047683716, "global_step": 228563, "epoch": 2568, "lr": 6.055156950224394e-06}
{"train_loss": 0.08940638601779938, "global_step": 228564, "epoch": 2568, "lr": 6.054880411947006e-06}
{"train_loss": 0.1062643975019455, "global_step": 228565, "epoch": 2568, "lr": 6.054603879577514e-06}
{"train_loss": 0.029055949300527573, "global_step": 228566, "epoch": 2568, "lr": 6.054327353115929e-06}
{"train_loss": 0.04562194645404816, "global_step": 228567, "epoch": 2568, "lr": 6.0540508325623195e-06}
{"train_loss": 0.031251732259988785, "global_step": 228568, "epoch": 2568, "lr": 6.053774317916694e-06}
{"train_loss": 0.054456666111946106, "global_step": 228569, "epoch": 2568, "lr": 6.053497809179115e-06}
{"train_loss": 0.044500771909952164, "global_step": 228570, "epoch": 2568, "lr": 6.053221306349599e-06}
{"train_loss": 0.020968345925211906, "global_step": 228571, "epoch": 2568, "lr": 6.052944809428196e-06}
{"train_loss": 0.10926999896764755, "global_step": 228572, "epoch": 2568, "lr": 6.052668318414934e-06}
{"train_loss": 0.12033043801784515, "global_step": 228573, "epoch": 2568, "lr": 6.052391833309862e-06}
{"train_loss": 0.04961913824081421, "global_step": 228574, "epoch": 2568, "lr": 6.052115354113002e-06}
{"train_loss": 0.06887692213058472, "global_step": 228575, "epoch": 2568, "lr": 6.051838880824412e-06}
{"train_loss": 0.06058125197887421, "global_step": 228576, "epoch": 2568, "lr": 6.051562413444101e-06}
{"train_loss": 0.1041952446103096, "global_step": 228577, "epoch": 2568, "lr": 6.051285951972124e-06}
{"train_loss": 0.07412451505661011, "global_step": 228578, "epoch": 2568, "lr": 6.051009496408527e-06}
{"train_loss": 0.03903888165950775, "global_step": 228579, "epoch": 2568, "lr": 6.050733046753327e-06}
{"train_loss": 0.044643331319093704, "global_step": 228580, "epoch": 2568, "lr": 6.0504566030065765e-06}
{"train_loss": 0.022814547643065453, "global_step": 228581, "epoch": 2568, "lr": 6.050180165168301e-06}
{"train_loss": 0.05536315217614174, "global_step": 228582, "epoch": 2568, "lr": 6.049903733238549e-06}
{"train_loss": 0.0405542366206646, "global_step": 228583, "epoch": 2568, "lr": 6.049627307217348e-06}
{"train_loss": 0.07288247346878052, "global_step": 228584, "epoch": 2568, "lr": 6.0493508871047435e-06}
{"train_loss": 0.043721169233322144, "global_step": 228585, "epoch": 2568, "lr": 6.049074472900768e-06}
{"train_loss": 0.06012822687625885, "global_step": 228586, "epoch": 2568, "lr": 6.048798064605448e-06}
{"train_loss": 0.0661194920539856, "global_step": 228587, "epoch": 2568, "lr": 6.048521662218848e-06}
{"train_loss": 0.04106971248984337, "global_step": 228588, "epoch": 2568, "lr": 6.04824526574097e-06}
{"train_loss": 0.049804896116256714, "global_step": 228589, "epoch": 2568, "lr": 6.047968875171889e-06}
{"train_loss": 0.026552686467766762, "global_step": 228590, "epoch": 2568, "lr": 6.047692490511614e-06}
{"train_loss": 0.044673990458250046, "global_step": 228591, "epoch": 2568, "lr": 6.0474161117601845e-06}
{"train_loss": 0.058801159262657166, "global_step": 228592, "epoch": 2568, "lr": 6.047139738917657e-06}
{"train_loss": 0.013352221809327602, "global_step": 228593, "epoch": 2568, "lr": 6.04686337198404e-06}
{"train_loss": 0.04846478998661041, "global_step": 228594, "epoch": 2568, "lr": 6.046587010959404e-06}
{"train_loss": 0.04744137078523636, "global_step": 228595, "epoch": 2568, "lr": 6.046310655843751e-06}
{"train_loss": 0.03172726929187775, "global_step": 228596, "epoch": 2568, "lr": 6.046034306637138e-06}
{"train_loss": 0.06436622142791748, "global_step": 228597, "epoch": 2568, "lr": 6.045757963339615e-06}
{"train_loss": 0.04055952653288841, "global_step": 228598, "epoch": 2568, "lr": 6.045481625951188e-06}
{"train_loss": 0.08968605101108551, "global_step": 228599, "epoch": 2568, "lr": 6.045205294471923e-06}
{"train_loss": 0.0888194665312767, "global_step": 228600, "epoch": 2568, "lr": 6.044928968901836e-06}
{"train_loss": 0.06384412199258804, "global_step": 228601, "epoch": 2568, "lr": 6.044652649240978e-06}
{"train_loss": 0.055918917059898376, "global_step": 228602, "epoch": 2568, "lr": 6.044376335489371e-06}
{"train_loss": 0.043745554983615875, "global_step": 228603, "epoch": 2568, "lr": 6.044100027647076e-06}
{"train_loss": 0.051190443336963654, "global_step": 228604, "epoch": 2568, "lr": 6.043823725714104e-06}
{"train_loss": 0.036856651306152344, "global_step": 228605, "epoch": 2568, "lr": 6.043547429690516e-06}
{"train_loss": 0.04033006727695465, "global_step": 228606, "epoch": 2568, "lr": 6.043271139576323e-06}
{"train_loss": 0.018699420616030693, "global_step": 228607, "epoch": 2568, "lr": 6.042994855371592e-06}
{"train_loss": 0.04588107019662857, "global_step": 228608, "epoch": 2568, "lr": 6.042718577076328e-06}
{"train_loss": 0.05389983579516411, "global_step": 228609, "epoch": 2568, "lr": 6.042442304690599e-06}
{"train_loss": 0.03794734925031662, "global_step": 228610, "epoch": 2568, "lr": 6.04216603821442e-06}
{"train_loss": 0.03673572093248367, "global_step": 228611, "epoch": 2568, "lr": 6.0418897776478415e-06}
{"train_loss": 0.07719819247722626, "global_step": 228612, "epoch": 2568, "lr": 6.041613522990886e-06}
{"train_loss": 0.02996027283370495, "global_step": 228613, "epoch": 2568, "lr": 6.041337274243603e-06}
{"train_loss": 0.11458201706409454, "global_step": 228614, "epoch": 2568, "lr": 6.0410610314060314e-06}
{"train_loss": 0.08626139909029007, "global_step": 228615, "epoch": 2568, "lr": 6.0407847944782e-06}
{"train_loss": 0.06141179800033569, "global_step": 228616, "epoch": 2568, "lr": 6.040508563460151e-06}
{"train_loss": 0.01957135647535324, "global_step": 228617, "epoch": 2568, "lr": 6.040232338351915e-06}
{"train_loss": 0.0512976348400116, "global_step": 228618, "epoch": 2568, "lr": 6.0399561191535446e-06}
{"train_loss": 0.047893330454826355, "global_step": 228619, "epoch": 2568, "lr": 6.039679905865053e-06}
{"train_loss": 0.049754757434129715, "global_step": 228620, "epoch": 2568, "lr": 6.039403698486501e-06}
{"train_loss": 0.03731594979763031, "global_step": 228621, "epoch": 2568, "lr": 6.03912749701791e-06}
{"train_loss": 0.0419267900288105, "global_step": 228622, "epoch": 2568, "lr": 6.038851301459331e-06}
{"train_loss": 0.020358765497803688, "global_step": 228623, "epoch": 2568, "lr": 6.03857511181079e-06}
{"train_loss": 0.0637146532535553, "global_step": 228624, "epoch": 2568, "lr": 6.038298928072317e-06}
{"train_loss": 0.0622548833489418, "global_step": 228625, "epoch": 2568, "lr": 6.038022750243971e-06}
{"train_loss": 0.06660342216491699, "global_step": 228626, "epoch": 2568, "lr": 6.037746578325765e-06}
{"train_loss": 0.03630435839295387, "global_step": 228627, "epoch": 2568, "lr": 6.0374704123177585e-06}
{"train_loss": 0.02439560554921627, "global_step": 228628, "epoch": 2568, "lr": 6.03719425221998e-06}
{"train_loss": 0.0528537854552269, "global_step": 228629, "epoch": 2568, "lr": 6.0369180980324515e-06}
{"train_loss": 0.08644146472215652, "global_step": 228630, "epoch": 2568, "lr": 6.036641949755234e-06}
{"train_loss": 0.05972214415669441, "global_step": 228631, "epoch": 2568, "lr": 6.036365807388345e-06}
{"train_loss": 0.06574034690856934, "global_step": 228632, "epoch": 2568, "lr": 6.0360896709318325e-06}
{"train_loss": 0.055278170853853226, "global_step": 228633, "epoch": 2568, "lr": 6.035813540385743e-06}
{"train_loss": 0.02788432314991951, "global_step": 228634, "epoch": 2568, "lr": 6.035537415750087e-06}
{"train_loss": 0.02946629375219345, "global_step": 228635, "epoch": 2568, "lr": 6.03526129702493e-06}
{"train_loss": 0.028274795040488243, "global_step": 228636, "epoch": 2568, "lr": 6.03498518421029e-06}
{"train_loss": 0.06270605325698853, "global_step": 228637, "epoch": 2568, "lr": 6.034709077306217e-06}
{"train_loss": 0.03772566467523575, "global_step": 228638, "epoch": 2568, "lr": 6.034432976312732e-06}
{"train_loss": 0.0802469477057457, "global_step": 228639, "epoch": 2568, "lr": 6.034156881229891e-06}
{"train_loss": 0.054425755002943985, "global_step": 228640, "epoch": 2568, "lr": 6.0338807920577115e-06, "val_loss": 9.333816528320312}
{"train_loss": 0.09229946881532669, "global_step": 228641, "epoch": 2569, "lr": 6.033604708796248e-06}
{"train_loss": 0.05791926756501198, "global_step": 228642, "epoch": 2569, "lr": 6.033328631445523e-06}
{"train_loss": 0.050109341740608215, "global_step": 228643, "epoch": 2569, "lr": 6.033052560005592e-06}
{"train_loss": 0.045568984001874924, "global_step": 228644, "epoch": 2569, "lr": 6.032776494476472e-06}
{"train_loss": 0.04918422922492027, "global_step": 228645, "epoch": 2569, "lr": 6.032500434858218e-06}
{"train_loss": 0.0531003437936306, "global_step": 228646, "epoch": 2569, "lr": 6.032224381150847e-06}
{"train_loss": 0.06059674173593521, "global_step": 228647, "epoch": 2569, "lr": 6.03194833335442e-06}
{"train_loss": 0.03030303679406643, "global_step": 228648, "epoch": 2569, "lr": 6.031672291468954e-06}
{"train_loss": 0.06189185380935669, "global_step": 228649, "epoch": 2569, "lr": 6.031396255494498e-06}
{"train_loss": 0.015872325748205185, "global_step": 228650, "epoch": 2569, "lr": 6.03112022543108e-06}
{"train_loss": 0.041518401354551315, "global_step": 228651, "epoch": 2569, "lr": 6.03084420127874e-06}
{"train_loss": 0.07836675643920898, "global_step": 228652, "epoch": 2569, "lr": 6.030568183037527e-06}
{"train_loss": 0.06258413940668106, "global_step": 228653, "epoch": 2569, "lr": 6.030292170707458e-06}
{"train_loss": 0.027849683538079262, "global_step": 228654, "epoch": 2569, "lr": 6.030016164288588e-06}
{"train_loss": 0.05164669454097748, "global_step": 228655, "epoch": 2569, "lr": 6.029740163780939e-06}
{"train_loss": 0.0664953961968422, "global_step": 228656, "epoch": 2569, "lr": 6.029464169184568e-06}
{"train_loss": 0.046249497681856155, "global_step": 228657, "epoch": 2569, "lr": 6.029188180499484e-06}
{"train_loss": 0.03553241118788719, "global_step": 228658, "epoch": 2569, "lr": 6.02891219772575e-06}
{"train_loss": 0.0611068457365036, "global_step": 228659, "epoch": 2569, "lr": 6.028636220863387e-06}
{"train_loss": 0.06572036445140839, "global_step": 228660, "epoch": 2569, "lr": 6.028360249912446e-06}
{"train_loss": 0.014631198719143867, "global_step": 228661, "epoch": 2569, "lr": 6.028084284872959e-06}
{"train_loss": 0.10944696515798569, "global_step": 228662, "epoch": 2569, "lr": 6.027808325744943e-06}
{"train_loss": 0.03404172882437706, "global_step": 228663, "epoch": 2569, "lr": 6.027532372528472e-06}
{"train_loss": 0.05368366092443466, "global_step": 228664, "epoch": 2569, "lr": 6.027256425223543e-06}
{"train_loss": 0.06928595900535583, "global_step": 228665, "epoch": 2569, "lr": 6.02698048383023e-06}
{"train_loss": 0.06586149334907532, "global_step": 228666, "epoch": 2569, "lr": 6.026704548348544e-06}
{"train_loss": 0.03788632154464722, "global_step": 228667, "epoch": 2569, "lr": 6.02642861877854e-06}
{"train_loss": 0.009215342812240124, "global_step": 228668, "epoch": 2569, "lr": 6.0261526951202355e-06}
{"train_loss": 0.020296234637498856, "global_step": 228669, "epoch": 2569, "lr": 6.02587677737369e-06}
{"train_loss": 0.047442737966775894, "global_step": 228670, "epoch": 2569, "lr": 6.025600865538916e-06}
{"train_loss": 0.025212351232767105, "global_step": 228671, "epoch": 2569, "lr": 6.02532495961598e-06}
{"train_loss": 0.04177852347493172, "global_step": 228672, "epoch": 2569, "lr": 6.025049059604887e-06}
{"train_loss": 0.023687848821282387, "global_step": 228673, "epoch": 2569, "lr": 6.0247731655057035e-06}
{"train_loss": 0.10772763192653656, "global_step": 228674, "epoch": 2569, "lr": 6.024497277318442e-06}
{"train_loss": 0.07428936660289764, "global_step": 228675, "epoch": 2569, "lr": 6.024221395043161e-06}
{"train_loss": 0.056428153067827225, "global_step": 228676, "epoch": 2569, "lr": 6.02394551867988e-06}
{"train_loss": 0.031066009774804115, "global_step": 228677, "epoch": 2569, "lr": 6.023669648228653e-06}
{"train_loss": 0.11518440395593643, "global_step": 228678, "epoch": 2569, "lr": 6.023393783689496e-06}
{"train_loss": 0.0671018585562706, "global_step": 228679, "epoch": 2569, "lr": 6.023117925062466e-06}
{"train_loss": 0.021399129182100296, "global_step": 228680, "epoch": 2569, "lr": 6.0228420723475845e-06}
{"train_loss": 0.05556436628103256, "global_step": 228681, "epoch": 2569, "lr": 6.022566225544907e-06}
{"train_loss": 0.015150532126426697, "global_step": 228682, "epoch": 2569, "lr": 6.0222903846544445e-06}
{"train_loss": 0.11070138961076736, "global_step": 228683, "epoch": 2569, "lr": 6.022014549676264e-06}
{"train_loss": 0.028895726427435875, "global_step": 228684, "epoch": 2569, "lr": 6.021738720610376e-06}
{"train_loss": 0.057942211627960205, "global_step": 228685, "epoch": 2569, "lr": 6.0214628974568365e-06}
{"train_loss": 0.0785147175192833, "global_step": 228686, "epoch": 2569, "lr": 6.021187080215668e-06}
{"train_loss": 0.041988763958215714, "global_step": 228687, "epoch": 2569, "lr": 6.020911268886914e-06}
{"train_loss": 0.05047845095396042, "global_step": 228688, "epoch": 2569, "lr": 6.020635463470625e-06}
{"train_loss": 0.047045063227415085, "global_step": 228689, "epoch": 2569, "lr": 6.020359663966813e-06}
{"train_loss": 0.08199656009674072, "global_step": 228690, "epoch": 2569, "lr": 6.020083870375537e-06}
{"train_loss": 0.03327048569917679, "global_step": 228691, "epoch": 2569, "lr": 6.019808082696815e-06}
{"train_loss": 0.017655014991760254, "global_step": 228692, "epoch": 2569, "lr": 6.019532300930708e-06}
{"train_loss": 0.024619072675704956, "global_step": 228693, "epoch": 2569, "lr": 6.019256525077227e-06}
{"train_loss": 0.0369461327791214, "global_step": 228694, "epoch": 2569, "lr": 6.018980755136427e-06}
{"train_loss": 0.05322748422622681, "global_step": 228695, "epoch": 2569, "lr": 6.018704991108332e-06}
{"train_loss": 0.020249687135219574, "global_step": 228696, "epoch": 2569, "lr": 6.0184292329929945e-06}
{"train_loss": 0.024653367698192596, "global_step": 228697, "epoch": 2569, "lr": 6.018153480790434e-06}
{"train_loss": 0.05743919685482979, "global_step": 228698, "epoch": 2569, "lr": 6.0178777345007095e-06}
{"train_loss": 0.031191008165478706, "global_step": 228699, "epoch": 2569, "lr": 6.017601994123828e-06}
{"train_loss": 0.022227182984352112, "global_step": 228700, "epoch": 2569, "lr": 6.017326259659861e-06}
{"train_loss": 0.02707752026617527, "global_step": 228701, "epoch": 2569, "lr": 6.017050531108826e-06}
{"train_loss": 0.0854591578245163, "global_step": 228702, "epoch": 2569, "lr": 6.0167748084707545e-06}
{"train_loss": 0.02261882647871971, "global_step": 228703, "epoch": 2569, "lr": 6.016499091745697e-06}
{"train_loss": 0.04757045581936836, "global_step": 228704, "epoch": 2569, "lr": 6.0162233809336774e-06}
{"train_loss": 0.027848578989505768, "global_step": 228705, "epoch": 2569, "lr": 6.0159476760347495e-06}
{"train_loss": 0.02432398684322834, "global_step": 228706, "epoch": 2569, "lr": 6.015671977048937e-06}
{"train_loss": 0.10019794851541519, "global_step": 228707, "epoch": 2569, "lr": 6.015396283976282e-06}
{"train_loss": 0.060661107301712036, "global_step": 228708, "epoch": 2569, "lr": 6.01512059681682e-06}
{"train_loss": 0.034277528524398804, "global_step": 228709, "epoch": 2569, "lr": 6.014844915570594e-06}
{"train_loss": 0.024964308366179466, "global_step": 228710, "epoch": 2569, "lr": 6.014569240237622e-06}
{"train_loss": 0.04637773707509041, "global_step": 228711, "epoch": 2569, "lr": 6.014293570817975e-06}
{"train_loss": 0.013936684466898441, "global_step": 228712, "epoch": 2569, "lr": 6.014017907311653e-06}
{"train_loss": 0.06736104190349579, "global_step": 228713, "epoch": 2569, "lr": 6.0137422497187235e-06}
{"train_loss": 0.028142254799604416, "global_step": 228714, "epoch": 2569, "lr": 6.013466598039197e-06}
{"train_loss": 0.06522607058286667, "global_step": 228715, "epoch": 2569, "lr": 6.013190952273135e-06}
{"train_loss": 0.03308127447962761, "global_step": 228716, "epoch": 2569, "lr": 6.012915312420553e-06}
{"train_loss": 0.053861353546381, "global_step": 228717, "epoch": 2569, "lr": 6.012639678481508e-06}
{"train_loss": 0.03860953822731972, "global_step": 228718, "epoch": 2569, "lr": 6.0123640504560206e-06}
{"train_loss": 0.05627971887588501, "global_step": 228719, "epoch": 2569, "lr": 6.012088428344142e-06}
{"train_loss": 0.08919069916009903, "global_step": 228720, "epoch": 2569, "lr": 6.011812812145895e-06}
{"train_loss": 0.06818774342536926, "global_step": 228721, "epoch": 2569, "lr": 6.011537201861328e-06}
{"train_loss": 0.05430975183844566, "global_step": 228722, "epoch": 2569, "lr": 6.011261597490469e-06}
{"train_loss": 0.0027291192673146725, "global_step": 228723, "epoch": 2569, "lr": 6.010985999033358e-06}
{"train_loss": 0.04670032858848572, "global_step": 228724, "epoch": 2569, "lr": 6.010710406490045e-06}
{"train_loss": 0.039325494319200516, "global_step": 228725, "epoch": 2569, "lr": 6.010434819860544e-06}
{"train_loss": 0.06043964624404907, "global_step": 228726, "epoch": 2569, "lr": 6.0101592391449135e-06}
{"train_loss": 0.031024865806102753, "global_step": 228727, "epoch": 2569, "lr": 6.0098836643431685e-06}
{"train_loss": 0.05841359123587608, "global_step": 228728, "epoch": 2569, "lr": 6.009608095455371e-06}
{"train_loss": 0.048183299763274665, "global_step": 228729, "epoch": 2569, "lr": 6.0093325324815374e-06, "val_loss": 9.129804611206055}
{"train_loss": 0.032698191702365875, "global_step": 228730, "epoch": 2570, "lr": 6.0090569754217225e-06}
{"train_loss": 0.055762145668268204, "global_step": 228731, "epoch": 2570, "lr": 6.008781424275944e-06}
{"train_loss": 0.04719012603163719, "global_step": 228732, "epoch": 2570, "lr": 6.0085058790442575e-06}
{"train_loss": 0.052769772708415985, "global_step": 228733, "epoch": 2570, "lr": 6.0082303397266784e-06}
{"train_loss": 0.046349454671144485, "global_step": 228734, "epoch": 2570, "lr": 6.007954806323274e-06}
{"train_loss": 0.03727969527244568, "global_step": 228735, "epoch": 2570, "lr": 6.007679278834044e-06}
{"train_loss": 0.04096454009413719, "global_step": 228736, "epoch": 2570, "lr": 6.007403757259061e-06}
{"train_loss": 0.07153163105249405, "global_step": 228737, "epoch": 2570, "lr": 6.007128241598336e-06}
{"train_loss": 0.03799549117684364, "global_step": 228738, "epoch": 2570, "lr": 6.006852731851931e-06}
{"train_loss": 0.053960803896188736, "global_step": 228739, "epoch": 2570, "lr": 6.006577228019861e-06}
{"train_loss": 0.05569068714976311, "global_step": 228740, "epoch": 2570, "lr": 6.0063017301021654e-06}
{"train_loss": 0.06431882083415985, "global_step": 228741, "epoch": 2570, "lr": 6.006026238098894e-06}
{"train_loss": 0.013492299243807793, "global_step": 228742, "epoch": 2570, "lr": 6.005750752010064e-06}
{"train_loss": 0.07897190749645233, "global_step": 228743, "epoch": 2570, "lr": 6.005475271835736e-06}
{"train_loss": 0.031178878620266914, "global_step": 228744, "epoch": 2570, "lr": 6.005199797575928e-06}
{"train_loss": 0.056066133081912994, "global_step": 228745, "epoch": 2570, "lr": 6.004924329230693e-06}
{"train_loss": 0.057056453078985214, "global_step": 228746, "epoch": 2570, "lr": 6.00464886680005e-06}
{"train_loss": 0.058465562760829926, "global_step": 228747, "epoch": 2570, "lr": 6.004373410284059e-06}
{"train_loss": 0.04249458387494087, "global_step": 228748, "epoch": 2570, "lr": 6.00409795968273e-06}
{"train_loss": 0.05514580011367798, "global_step": 228749, "epoch": 2570, "lr": 6.003822514996121e-06}
{"train_loss": 0.09707451611757278, "global_step": 228750, "epoch": 2570, "lr": 6.003547076224258e-06}
{"train_loss": 0.06539618223905563, "global_step": 228751, "epoch": 2570, "lr": 6.003271643367187e-06}
{"train_loss": 0.033763039857149124, "global_step": 228752, "epoch": 2570, "lr": 6.002996216424933e-06}
{"train_loss": 0.04406968131661415, "global_step": 228753, "epoch": 2570, "lr": 6.0027207953975484e-06}
{"train_loss": 0.06115688756108284, "global_step": 228754, "epoch": 2570, "lr": 6.002445380285055e-06}
{"train_loss": 0.04276549816131592, "global_step": 228755, "epoch": 2570, "lr": 6.002169971087501e-06}
{"train_loss": 0.0631091445684433, "global_step": 228756, "epoch": 2570, "lr": 6.001894567804911e-06}
{"train_loss": 0.06861066073179245, "global_step": 228757, "epoch": 2570, "lr": 6.001619170437339e-06}
{"train_loss": 0.03684045746922493, "global_step": 228758, "epoch": 2570, "lr": 6.001343778984808e-06}
{"train_loss": 0.05509498342871666, "global_step": 228759, "epoch": 2570, "lr": 6.001068393447356e-06}
{"train_loss": 0.020635491237044334, "global_step": 228760, "epoch": 2570, "lr": 6.000793013825035e-06}
{"train_loss": 0.0487692654132843, "global_step": 228761, "epoch": 2570, "lr": 6.000517640117859e-06}
{"train_loss": 0.07102976739406586, "global_step": 228762, "epoch": 2570, "lr": 6.00024227232589e-06}
{"train_loss": 0.06720206886529922, "global_step": 228763, "epoch": 2570, "lr": 5.99996691044914e-06}
{"train_loss": 0.03978269174695015, "global_step": 228764, "epoch": 2570, "lr": 5.999691554487669e-06}
{"train_loss": 0.04678909480571747, "global_step": 228765, "epoch": 2570, "lr": 5.9994162044414935e-06}
{"train_loss": 0.03274547681212425, "global_step": 228766, "epoch": 2570, "lr": 5.99914086031067e-06}
{"train_loss": 0.06174489110708237, "global_step": 228767, "epoch": 2570, "lr": 5.99886552209522e-06}
{"train_loss": 0.013655707240104675, "global_step": 228768, "epoch": 2570, "lr": 5.998590189795189e-06}
{"train_loss": 0.039110295474529266, "global_step": 228769, "epoch": 2570, "lr": 5.9983148634106035e-06}
{"train_loss": 0.07951132953166962, "global_step": 228770, "epoch": 2570, "lr": 5.998039542941519e-06}
{"train_loss": 0.04927148297429085, "global_step": 228771, "epoch": 2570, "lr": 5.997764228387953e-06}
{"train_loss": 0.057282447814941406, "global_step": 228772, "epoch": 2570, "lr": 5.99748891974996e-06}
{"train_loss": 0.0240089762955904, "global_step": 228773, "epoch": 2570, "lr": 5.997213617027559e-06}
{"train_loss": 0.04466948285698891, "global_step": 228774, "epoch": 2570, "lr": 5.996938320220807e-06}
{"train_loss": 0.0177894439548254, "global_step": 228775, "epoch": 2570, "lr": 5.996663029329719e-06}
{"train_loss": 0.04052154719829559, "global_step": 228776, "epoch": 2570, "lr": 5.996387744354354e-06}
{"train_loss": 0.039932072162628174, "global_step": 228777, "epoch": 2570, "lr": 5.996112465294728e-06}
{"train_loss": 0.031904738396406174, "global_step": 228778, "epoch": 2570, "lr": 5.995837192150899e-06}
{"train_loss": 0.021015305072069168, "global_step": 228779, "epoch": 2570, "lr": 5.995561924922893e-06}
{"train_loss": 0.0883798822760582, "global_step": 228780, "epoch": 2570, "lr": 5.995286663610733e-06}
{"train_loss": 0.05708117038011551, "global_step": 228781, "epoch": 2570, "lr": 5.995011408214485e-06}
{"train_loss": 0.03664756938815117, "global_step": 228782, "epoch": 2570, "lr": 5.99473615873416e-06}
{"train_loss": 0.06654289364814758, "global_step": 228783, "epoch": 2570, "lr": 5.99446091516982e-06}
{"train_loss": 0.018347522243857384, "global_step": 228784, "epoch": 2570, "lr": 5.994185677521475e-06}
{"train_loss": 0.021146293729543686, "global_step": 228785, "epoch": 2570, "lr": 5.993910445789186e-06}
{"train_loss": 0.009184149093925953, "global_step": 228786, "epoch": 2570, "lr": 5.99363521997297e-06}
{"train_loss": 0.02380809374153614, "global_step": 228787, "epoch": 2570, "lr": 5.993360000072884e-06}
{"train_loss": 0.06326386332511902, "global_step": 228788, "epoch": 2570, "lr": 5.993084786088943e-06}
{"train_loss": 0.04593166708946228, "global_step": 228789, "epoch": 2570, "lr": 5.992809578021202e-06}
{"train_loss": 0.058834221214056015, "global_step": 228790, "epoch": 2570, "lr": 5.992534375869685e-06}
{"train_loss": 0.05781857296824455, "global_step": 228791, "epoch": 2570, "lr": 5.992259179634446e-06}
{"train_loss": 0.05701027810573578, "global_step": 228792, "epoch": 2570, "lr": 5.991983989315503e-06}
{"train_loss": 0.04415246471762657, "global_step": 228793, "epoch": 2570, "lr": 5.991708804912904e-06}
{"train_loss": 0.059508148580789566, "global_step": 228794, "epoch": 2570, "lr": 5.991433626426679e-06}
{"train_loss": 0.03554745763540268, "global_step": 228795, "epoch": 2570, "lr": 5.991158453856871e-06}
{"train_loss": 0.032283395528793335, "global_step": 228796, "epoch": 2570, "lr": 5.990883287203519e-06}
{"train_loss": 0.01610589772462845, "global_step": 228797, "epoch": 2570, "lr": 5.9906081264666515e-06}
{"train_loss": 0.09629867970943451, "global_step": 228798, "epoch": 2570, "lr": 5.990332971646323e-06}
{"train_loss": 0.07620057463645935, "global_step": 228799, "epoch": 2570, "lr": 5.99005782274254e-06}
{"train_loss": 0.046978794038295746, "global_step": 228800, "epoch": 2570, "lr": 5.989782679755373e-06}
{"train_loss": 0.06095237284898758, "global_step": 228801, "epoch": 2570, "lr": 5.98950754268483e-06}
{"train_loss": 0.05512967333197594, "global_step": 228802, "epoch": 2570, "lr": 5.989232411530976e-06}
{"train_loss": 0.08433182537555695, "global_step": 228803, "epoch": 2570, "lr": 5.988957286293817e-06}
{"train_loss": 0.032579194754362106, "global_step": 228804, "epoch": 2570, "lr": 5.988682166973419e-06}
{"train_loss": 0.04771874472498894, "global_step": 228805, "epoch": 2570, "lr": 5.9884070535698e-06}
{"train_loss": 0.03609074279665947, "global_step": 228806, "epoch": 2570, "lr": 5.988131946083009e-06}
{"train_loss": 0.0522461012005806, "global_step": 228807, "epoch": 2570, "lr": 5.987856844513068e-06}
{"train_loss": 0.039288561791181564, "global_step": 228808, "epoch": 2570, "lr": 5.987581748860033e-06}
{"train_loss": 0.037790533155202866, "global_step": 228809, "epoch": 2570, "lr": 5.98730665912392e-06}
{"train_loss": 0.07357548922300339, "global_step": 228810, "epoch": 2570, "lr": 5.9870315753047916e-06}
{"train_loss": 0.036750975996255875, "global_step": 228811, "epoch": 2570, "lr": 5.986756497402657e-06}
{"train_loss": 0.07268492877483368, "global_step": 228812, "epoch": 2570, "lr": 5.986481425417578e-06}
{"train_loss": 0.07211747020483017, "global_step": 228813, "epoch": 2570, "lr": 5.9862063593495665e-06}
{"train_loss": 0.10296902060508728, "global_step": 228814, "epoch": 2570, "lr": 5.985931299198688e-06}
{"train_loss": 0.060895998030900955, "global_step": 228815, "epoch": 2570, "lr": 5.985656244964949e-06}
{"train_loss": 0.05787099152803421, "global_step": 228816, "epoch": 2570, "lr": 5.985381196648415e-06}
{"train_loss": 0.03945435583591461, "global_step": 228817, "epoch": 2570, "lr": 5.985106154249115e-06}
{"train_loss": 0.05004498734023799, "global_step": 228818, "epoch": 2570, "lr": 5.9848311177670645e-06, "val_loss": 9.200427055358887, "train_action_mse_error": 25.384693145751953}
{"train_loss": 0.04226814955472946, "global_step": 228819, "epoch": 2571, "lr": 5.984556087202331e-06}
{"train_loss": 0.023581702262163162, "global_step": 228820, "epoch": 2571, "lr": 5.984281062554925e-06}
{"train_loss": 0.04350612685084343, "global_step": 228821, "epoch": 2571, "lr": 5.984006043824902e-06}
{"train_loss": 0.038227543234825134, "global_step": 228822, "epoch": 2571, "lr": 5.983731031012291e-06}
{"train_loss": 0.02874545194208622, "global_step": 228823, "epoch": 2571, "lr": 5.983456024117135e-06}
{"train_loss": 0.05965719744563103, "global_step": 228824, "epoch": 2571, "lr": 5.9831810231394615e-06}
{"train_loss": 0.06780939549207687, "global_step": 228825, "epoch": 2571, "lr": 5.9829060280793155e-06}
{"train_loss": 0.04707874357700348, "global_step": 228826, "epoch": 2571, "lr": 5.9826310389367315e-06}
{"train_loss": 0.049618612974882126, "global_step": 228827, "epoch": 2571, "lr": 5.982356055711747e-06}
{"train_loss": 0.0210468377918005, "global_step": 228828, "epoch": 2571, "lr": 5.982081078404395e-06}
{"train_loss": 0.08410798013210297, "global_step": 228829, "epoch": 2571, "lr": 5.98180610701472e-06}
{"train_loss": 0.06734780967235565, "global_step": 228830, "epoch": 2571, "lr": 5.981531141542751e-06}
{"train_loss": 0.043256524950265884, "global_step": 228831, "epoch": 2571, "lr": 5.9812561819885205e-06}
{"train_loss": 0.053850218653678894, "global_step": 228832, "epoch": 2571, "lr": 5.98098122835209e-06}
{"train_loss": 0.03782801330089569, "global_step": 228833, "epoch": 2571, "lr": 5.980706280633469e-06}
{"train_loss": 0.02173641324043274, "global_step": 228834, "epoch": 2571, "lr": 5.980431338832715e-06}
{"train_loss": 0.030855175107717514, "global_step": 228835, "epoch": 2571, "lr": 5.980156402949843e-06}
{"train_loss": 0.060861736536026, "global_step": 228836, "epoch": 2571, "lr": 5.979881472984916e-06}
{"train_loss": 0.07459903508424759, "global_step": 228837, "epoch": 2571, "lr": 5.979606548937944e-06}
{"train_loss": 0.027744947001338005, "global_step": 228838, "epoch": 2571, "lr": 5.979331630808993e-06}
{"train_loss": 0.04509669169783592, "global_step": 228839, "epoch": 2571, "lr": 5.97905671859807e-06}
{"train_loss": 0.09211701154708862, "global_step": 228840, "epoch": 2571, "lr": 5.978781812305234e-06}
{"train_loss": 0.045447856187820435, "global_step": 228841, "epoch": 2571, "lr": 5.9785069119305095e-06}
{"train_loss": 0.08013800531625748, "global_step": 228842, "epoch": 2571, "lr": 5.978232017473945e-06}
{"train_loss": 0.01421796903014183, "global_step": 228843, "epoch": 2571, "lr": 5.977957128935563e-06}
{"train_loss": 0.05245053023099899, "global_step": 228844, "epoch": 2571, "lr": 5.9776822463154195e-06}
{"train_loss": 0.026268314570188522, "global_step": 228845, "epoch": 2571, "lr": 5.977407369613524e-06}
{"train_loss": 0.03698591887950897, "global_step": 228846, "epoch": 2571, "lr": 5.9771324988299405e-06}
{"train_loss": 0.05170349404215813, "global_step": 228847, "epoch": 2571, "lr": 5.976857633964689e-06}
{"train_loss": 0.04635140299797058, "global_step": 228848, "epoch": 2571, "lr": 5.976582775017819e-06}
{"train_loss": 0.07677747309207916, "global_step": 228849, "epoch": 2571, "lr": 5.976307921989355e-06}
{"train_loss": 0.07467921078205109, "global_step": 228850, "epoch": 2571, "lr": 5.97603307487935e-06}
{"train_loss": 0.051985129714012146, "global_step": 228851, "epoch": 2571, "lr": 5.975758233687817e-06}
{"train_loss": 0.024038808420300484, "global_step": 228852, "epoch": 2571, "lr": 5.9754833984148165e-06}
{"train_loss": 0.05255647748708725, "global_step": 228853, "epoch": 2571, "lr": 5.97520856906037e-06}
{"train_loss": 0.03831300511956215, "global_step": 228854, "epoch": 2571, "lr": 5.9749337456245226e-06}
{"train_loss": 0.04800104349851608, "global_step": 228855, "epoch": 2571, "lr": 5.974658928107313e-06}
{"train_loss": 0.03959706053137779, "global_step": 228856, "epoch": 2571, "lr": 5.974384116508769e-06}
{"train_loss": 0.05950988084077835, "global_step": 228857, "epoch": 2571, "lr": 5.974109310828935e-06}
{"train_loss": 0.062064237892627716, "global_step": 228858, "epoch": 2571, "lr": 5.973834511067839e-06}
{"train_loss": 0.05296984314918518, "global_step": 228859, "epoch": 2571, "lr": 5.973559717225535e-06}
{"train_loss": 0.03590718284249306, "global_step": 228860, "epoch": 2571, "lr": 5.973284929302037e-06}
{"train_loss": 0.037946779280900955, "global_step": 228861, "epoch": 2571, "lr": 5.973010147297403e-06}
{"train_loss": 0.03541405498981476, "global_step": 228862, "epoch": 2571, "lr": 5.972735371211652e-06}
{"train_loss": 0.027355235069990158, "global_step": 228863, "epoch": 2571, "lr": 5.972460601044843e-06}
{"train_loss": 0.055893056094646454, "global_step": 228864, "epoch": 2571, "lr": 5.9721858367969894e-06}
{"train_loss": 0.04025980085134506, "global_step": 228865, "epoch": 2571, "lr": 5.971911078468145e-06}
{"train_loss": 0.06997411698102951, "global_step": 228866, "epoch": 2571, "lr": 5.971636326058333e-06}
{"train_loss": 0.04366718977689743, "global_step": 228867, "epoch": 2571, "lr": 5.9713615795676025e-06}
{"train_loss": 0.043835535645484924, "global_step": 228868, "epoch": 2571, "lr": 5.971086838995987e-06}
{"train_loss": 0.03567975014448166, "global_step": 228869, "epoch": 2571, "lr": 5.9708121043435205e-06}
{"train_loss": 0.07714333385229111, "global_step": 228870, "epoch": 2571, "lr": 5.970537375610247e-06}
{"train_loss": 0.0804026871919632, "global_step": 228871, "epoch": 2571, "lr": 5.970262652796188e-06}
{"train_loss": 0.020001396536827087, "global_step": 228872, "epoch": 2571, "lr": 5.9699879359014e-06}
{"train_loss": 0.02039155922830105, "global_step": 228873, "epoch": 2571, "lr": 5.969713224925905e-06}
{"train_loss": 0.060043707489967346, "global_step": 228874, "epoch": 2571, "lr": 5.9694385198697525e-06}
{"train_loss": 0.04159119725227356, "global_step": 228875, "epoch": 2571, "lr": 5.969163820732959e-06}
{"train_loss": 0.03818045184016228, "global_step": 228876, "epoch": 2571, "lr": 5.968889127515592e-06}
{"train_loss": 0.04010270908474922, "global_step": 228877, "epoch": 2571, "lr": 5.968614440217657e-06}
{"train_loss": 0.019563550129532814, "global_step": 228878, "epoch": 2571, "lr": 5.9683397588392145e-06}
{"train_loss": 0.025303605943918228, "global_step": 228879, "epoch": 2571, "lr": 5.968065083380281e-06}
{"train_loss": 0.01102373655885458, "global_step": 228880, "epoch": 2571, "lr": 5.967790413840918e-06}
{"train_loss": 0.03871719911694527, "global_step": 228881, "epoch": 2571, "lr": 5.967515750221136e-06}
{"train_loss": 0.061542026698589325, "global_step": 228882, "epoch": 2571, "lr": 5.967241092520998e-06}
{"train_loss": 0.0597115121781826, "global_step": 228883, "epoch": 2571, "lr": 5.966966440740513e-06}
{"train_loss": 0.04546649754047394, "global_step": 228884, "epoch": 2571, "lr": 5.9666917948797475e-06}
{"train_loss": 0.038572244346141815, "global_step": 228885, "epoch": 2571, "lr": 5.9664171549387085e-06}
{"train_loss": 0.062148503959178925, "global_step": 228886, "epoch": 2571, "lr": 5.966142520917451e-06}
{"train_loss": 0.04989218711853027, "global_step": 228887, "epoch": 2571, "lr": 5.965867892816024e-06}
{"train_loss": 0.05260172486305237, "global_step": 228888, "epoch": 2571, "lr": 5.96559327063444e-06}
{"train_loss": 0.08164526522159576, "global_step": 228889, "epoch": 2571, "lr": 5.965318654372742e-06}
{"train_loss": 0.06242214888334274, "global_step": 228890, "epoch": 2571, "lr": 5.965044044030977e-06}
{"train_loss": 0.04046005755662918, "global_step": 228891, "epoch": 2571, "lr": 5.964769439609164e-06}
{"train_loss": 0.04578270763158798, "global_step": 228892, "epoch": 2571, "lr": 5.964494841107365e-06}
{"train_loss": 0.0349150151014328, "global_step": 228893, "epoch": 2571, "lr": 5.964220248525593e-06}
{"train_loss": 0.07843447476625443, "global_step": 228894, "epoch": 2571, "lr": 5.963945661863901e-06}
{"train_loss": 0.0449955128133297, "global_step": 228895, "epoch": 2571, "lr": 5.963671081122324e-06}
{"train_loss": 0.055423982441425323, "global_step": 228896, "epoch": 2571, "lr": 5.963396506300878e-06}
{"train_loss": 0.060744501650333405, "global_step": 228897, "epoch": 2571, "lr": 5.963121937399635e-06}
{"train_loss": 0.09317252039909363, "global_step": 228898, "epoch": 2571, "lr": 5.962847374418595e-06}
{"train_loss": 0.06392531096935272, "global_step": 228899, "epoch": 2571, "lr": 5.962572817357831e-06}
{"train_loss": 0.008915847167372704, "global_step": 228900, "epoch": 2571, "lr": 5.962298266217348e-06}
{"train_loss": 0.033115316182374954, "global_step": 228901, "epoch": 2571, "lr": 5.962023720997212e-06}
{"train_loss": 0.023142436519265175, "global_step": 228902, "epoch": 2571, "lr": 5.961749181697429e-06}
{"train_loss": 0.047970354557037354, "global_step": 228903, "epoch": 2571, "lr": 5.961474648318066e-06}
{"train_loss": 0.10139339417219162, "global_step": 228904, "epoch": 2571, "lr": 5.961200120859134e-06}
{"train_loss": 0.026313541457057, "global_step": 228905, "epoch": 2571, "lr": 5.960925599320683e-06}
{"train_loss": 0.07678016275167465, "global_step": 228906, "epoch": 2571, "lr": 5.960651083702756e-06}
{"train_loss": 0.048103085823691966, "global_step": 228907, "epoch": 2571, "lr": 5.9603765740053775e-06, "val_loss": 9.133298873901367}
{"train_loss": 0.10325436294078827, "global_step": 228908, "epoch": 2572, "lr": 5.9601020702285956e-06}
{"train_loss": 0.047688115388154984, "global_step": 228909, "epoch": 2572, "lr": 5.959827572372434e-06}
{"train_loss": 0.06620461493730545, "global_step": 228910, "epoch": 2572, "lr": 5.959553080436947e-06}
{"train_loss": 0.06628197431564331, "global_step": 228911, "epoch": 2572, "lr": 5.959278594422147e-06}
{"train_loss": 0.04277396202087402, "global_step": 228912, "epoch": 2572, "lr": 5.9590041143280986e-06}
{"train_loss": 0.06404993683099747, "global_step": 228913, "epoch": 2572, "lr": 5.958729640154814e-06}
{"train_loss": 0.026235297322273254, "global_step": 228914, "epoch": 2572, "lr": 5.958455171902355e-06}
{"train_loss": 0.08798084408044815, "global_step": 228915, "epoch": 2572, "lr": 5.958180709570732e-06}
{"train_loss": 0.04505816847085953, "global_step": 228916, "epoch": 2572, "lr": 5.95790625316e-06}
{"train_loss": 0.0748099610209465, "global_step": 228917, "epoch": 2572, "lr": 5.957631802670188e-06}
{"train_loss": 0.03716456890106201, "global_step": 228918, "epoch": 2572, "lr": 5.957357358101345e-06}
{"train_loss": 0.027634989470243454, "global_step": 228919, "epoch": 2572, "lr": 5.9570829194534825e-06}
{"train_loss": 0.07305342704057693, "global_step": 228920, "epoch": 2572, "lr": 5.956808486726673e-06}
{"train_loss": 0.037640564143657684, "global_step": 228921, "epoch": 2572, "lr": 5.956534059920915e-06}
{"train_loss": 0.03604230284690857, "global_step": 228922, "epoch": 2572, "lr": 5.9562596390362656e-06}
{"train_loss": 0.007370417006313801, "global_step": 228923, "epoch": 2572, "lr": 5.9559852240727745e-06}
{"train_loss": 0.0485977828502655, "global_step": 228924, "epoch": 2572, "lr": 5.955710815030452e-06}
{"train_loss": 0.060896795243024826, "global_step": 228925, "epoch": 2572, "lr": 5.95543641190936e-06}
{"train_loss": 0.09418874233961105, "global_step": 228926, "epoch": 2572, "lr": 5.9551620147095085e-06}
{"train_loss": 0.03418527543544769, "global_step": 228927, "epoch": 2572, "lr": 5.95488762343096e-06}
{"train_loss": 0.03665637969970703, "global_step": 228928, "epoch": 2572, "lr": 5.9546132380737415e-06}
{"train_loss": 0.020867522805929184, "global_step": 228929, "epoch": 2572, "lr": 5.954338858637876e-06}
{"train_loss": 0.06034152954816818, "global_step": 228930, "epoch": 2572, "lr": 5.954064485123423e-06}
{"train_loss": 0.07867082953453064, "global_step": 228931, "epoch": 2572, "lr": 5.9537901175304e-06}
{"train_loss": 0.04513109475374222, "global_step": 228932, "epoch": 2572, "lr": 5.953515755858868e-06}
{"train_loss": 0.04098520427942276, "global_step": 228933, "epoch": 2572, "lr": 5.953241400108844e-06}
{"train_loss": 0.09708572179079056, "global_step": 228934, "epoch": 2572, "lr": 5.952967050280356e-06}
{"train_loss": 0.0438203290104866, "global_step": 228935, "epoch": 2572, "lr": 5.95269270637347e-06}
{"train_loss": 0.05393591150641441, "global_step": 228936, "epoch": 2572, "lr": 5.9524183683881966e-06}
{"train_loss": 0.05502551421523094, "global_step": 228937, "epoch": 2572, "lr": 5.952144036324598e-06}
{"train_loss": 0.04787849634885788, "global_step": 228938, "epoch": 2572, "lr": 5.951869710182678e-06}
{"train_loss": 0.09303192794322968, "global_step": 228939, "epoch": 2572, "lr": 5.9515953899625056e-06}
{"train_loss": 0.04741058498620987, "global_step": 228940, "epoch": 2572, "lr": 5.951321075664095e-06}
{"train_loss": 0.05754539370536804, "global_step": 228941, "epoch": 2572, "lr": 5.951046767287493e-06}
{"train_loss": 0.04018497094511986, "global_step": 228942, "epoch": 2572, "lr": 5.950772464832749e-06}
{"train_loss": 0.07443848252296448, "global_step": 228943, "epoch": 2572, "lr": 5.950498168299873e-06}
{"train_loss": 0.05546475201845169, "global_step": 228944, "epoch": 2572, "lr": 5.950223877688926e-06}
{"train_loss": 0.03922728821635246, "global_step": 228945, "epoch": 2572, "lr": 5.949949592999926e-06}
{"train_loss": 0.056056901812553406, "global_step": 228946, "epoch": 2572, "lr": 5.949675314232922e-06}
{"train_loss": 0.11537840217351913, "global_step": 228947, "epoch": 2572, "lr": 5.949401041387942e-06}
{"train_loss": 0.05742558091878891, "global_step": 228948, "epoch": 2572, "lr": 5.949126774465042e-06}
{"train_loss": 0.05901525542140007, "global_step": 228949, "epoch": 2572, "lr": 5.9488525134642316e-06}
{"train_loss": 0.021252362057566643, "global_step": 228950, "epoch": 2572, "lr": 5.9485782583855675e-06}
{"train_loss": 0.062447868287563324, "global_step": 228951, "epoch": 2572, "lr": 5.9483040092290774e-06}
{"train_loss": 0.05305932089686394, "global_step": 228952, "epoch": 2572, "lr": 5.948029765994806e-06}
{"train_loss": 0.07016942650079727, "global_step": 228953, "epoch": 2572, "lr": 5.947755528682774e-06}
{"train_loss": 0.01914234459400177, "global_step": 228954, "epoch": 2572, "lr": 5.947481297293039e-06}
{"train_loss": 0.04323773831129074, "global_step": 228955, "epoch": 2572, "lr": 5.947207071825622e-06}
{"train_loss": 0.050949860364198685, "global_step": 228956, "epoch": 2572, "lr": 5.946932852280579e-06}
{"train_loss": 0.04809755086898804, "global_step": 228957, "epoch": 2572, "lr": 5.9466586386579145e-06}
{"train_loss": 0.032029710710048676, "global_step": 228958, "epoch": 2572, "lr": 5.946384430957691e-06}
{"train_loss": 0.026833435520529747, "global_step": 228959, "epoch": 2572, "lr": 5.9461102291799465e-06}
{"train_loss": 0.027313269674777985, "global_step": 228960, "epoch": 2572, "lr": 5.945836033324703e-06}
{"train_loss": 0.04482397064566612, "global_step": 228961, "epoch": 2572, "lr": 5.945561843392017e-06}
{"train_loss": 0.04398699849843979, "global_step": 228962, "epoch": 2572, "lr": 5.945287659381898e-06}
{"train_loss": 0.021290138363838196, "global_step": 228963, "epoch": 2572, "lr": 5.945013481294409e-06}
{"train_loss": 0.029841938987374306, "global_step": 228964, "epoch": 2572, "lr": 5.944739309129571e-06}
{"train_loss": 0.029787376523017883, "global_step": 228965, "epoch": 2572, "lr": 5.944465142887428e-06}
{"train_loss": 0.03182341158390045, "global_step": 228966, "epoch": 2572, "lr": 5.944190982568021e-06}
{"train_loss": 0.011928930878639221, "global_step": 228967, "epoch": 2572, "lr": 5.94391682817137e-06}
{"train_loss": 0.06281258910894394, "global_step": 228968, "epoch": 2572, "lr": 5.943642679697531e-06}
{"train_loss": 0.03970656543970108, "global_step": 228969, "epoch": 2572, "lr": 5.943368537146521e-06}
{"train_loss": 0.07312552630901337, "global_step": 228970, "epoch": 2572, "lr": 5.943094400518401e-06}
{"train_loss": 0.03926568478345871, "global_step": 228971, "epoch": 2572, "lr": 5.942820269813193e-06}
{"train_loss": 0.057740580290555954, "global_step": 228972, "epoch": 2572, "lr": 5.942546145030925e-06}
{"train_loss": 0.02805665135383606, "global_step": 228973, "epoch": 2572, "lr": 5.942272026171658e-06}
{"train_loss": 0.07603322714567184, "global_step": 228974, "epoch": 2572, "lr": 5.941997913235403e-06}
{"train_loss": 0.055378153920173645, "global_step": 228975, "epoch": 2572, "lr": 5.941723806222216e-06}
{"train_loss": 0.057597000151872635, "global_step": 228976, "epoch": 2572, "lr": 5.941449705132124e-06}
{"train_loss": 0.048208240419626236, "global_step": 228977, "epoch": 2572, "lr": 5.941175609965161e-06}
{"train_loss": 0.03504016250371933, "global_step": 228978, "epoch": 2572, "lr": 5.940901520721387e-06}
{"train_loss": 0.05304316431283951, "global_step": 228979, "epoch": 2572, "lr": 5.940627437400803e-06}
{"train_loss": 0.0529940165579319, "global_step": 228980, "epoch": 2572, "lr": 5.9403533600034814e-06}
{"train_loss": 0.0513884574174881, "global_step": 228981, "epoch": 2572, "lr": 5.940079288529432e-06}
{"train_loss": 0.058944132179021835, "global_step": 228982, "epoch": 2572, "lr": 5.939805222978712e-06}
{"train_loss": 0.11800208687782288, "global_step": 228983, "epoch": 2572, "lr": 5.939531163351331e-06}
{"train_loss": 0.045291900634765625, "global_step": 228984, "epoch": 2572, "lr": 5.939257109647361e-06}
{"train_loss": 0.06462565064430237, "global_step": 228985, "epoch": 2572, "lr": 5.93898306186681e-06}
{"train_loss": 0.06854173541069031, "global_step": 228986, "epoch": 2572, "lr": 5.93870902000973e-06}
{"train_loss": 0.08541582524776459, "global_step": 228987, "epoch": 2572, "lr": 5.938434984076147e-06}
{"train_loss": 0.03758695349097252, "global_step": 228988, "epoch": 2572, "lr": 5.938160954066118e-06}
{"train_loss": 0.024983251467347145, "global_step": 228989, "epoch": 2572, "lr": 5.937886929979653e-06}
{"train_loss": 0.03937447816133499, "global_step": 228990, "epoch": 2572, "lr": 5.937612911816809e-06}
{"train_loss": 0.12074893712997437, "global_step": 228991, "epoch": 2572, "lr": 5.93733889957761e-06}
{"train_loss": 0.04101603478193283, "global_step": 228992, "epoch": 2572, "lr": 5.937064893262107e-06}
{"train_loss": 0.0395023375749588, "global_step": 228993, "epoch": 2572, "lr": 5.93679089287032e-06}
{"train_loss": 0.021362509578466415, "global_step": 228994, "epoch": 2572, "lr": 5.936516898402295e-06}
{"train_loss": 0.09499752521514893, "global_step": 228995, "epoch": 2572, "lr": 5.9362429098580754e-06}
{"train_loss": 0.05205445566078585, "global_step": 228996, "epoch": 2572, "lr": 5.935968927237684e-06, "val_loss": 9.123919486999512}
{"train_loss": 0.037673600018024445, "global_step": 228997, "epoch": 2573, "lr": 5.9356949505411705e-06}
{"train_loss": 0.022720864042639732, "global_step": 228998, "epoch": 2573, "lr": 5.935420979768558e-06}
{"train_loss": 0.024906184524297714, "global_step": 228999, "epoch": 2573, "lr": 5.935147014919906e-06}
{"train_loss": 0.029696807265281677, "global_step": 229000, "epoch": 2573, "lr": 5.934873055995221e-06}
{"train_loss": 0.06588423252105713, "global_step": 229001, "epoch": 2573, "lr": 5.934599102994564e-06}
{"train_loss": 0.03358136862516403, "global_step": 229002, "epoch": 2573, "lr": 5.9343251559179524e-06}
{"train_loss": 0.037979062646627426, "global_step": 229003, "epoch": 2573, "lr": 5.934051214765452e-06}
{"train_loss": 0.0691743865609169, "global_step": 229004, "epoch": 2573, "lr": 5.933777279537072e-06}
{"train_loss": 0.05137553811073303, "global_step": 229005, "epoch": 2573, "lr": 5.933503350232855e-06}
{"train_loss": 0.0508098378777504, "global_step": 229006, "epoch": 2573, "lr": 5.933229426852849e-06}
{"train_loss": 0.06083860248327255, "global_step": 229007, "epoch": 2573, "lr": 5.9329555093970705e-06}
{"train_loss": 0.016397839412093163, "global_step": 229008, "epoch": 2573, "lr": 5.9326815978655804e-06}
{"train_loss": 0.06348782777786255, "global_step": 229009, "epoch": 2573, "lr": 5.932407692258396e-06}
{"train_loss": 0.06187622249126434, "global_step": 229010, "epoch": 2573, "lr": 5.9321337925755725e-06}
{"train_loss": 0.0307854525744915, "global_step": 229011, "epoch": 2573, "lr": 5.931859898817133e-06}
{"train_loss": 0.08430074900388718, "global_step": 229012, "epoch": 2573, "lr": 5.931586010983109e-06}
{"train_loss": 0.04509362205862999, "global_step": 229013, "epoch": 2573, "lr": 5.931312129073552e-06}
{"train_loss": 0.04444100707769394, "global_step": 229014, "epoch": 2573, "lr": 5.931038253088495e-06}
{"train_loss": 0.04085150361061096, "global_step": 229015, "epoch": 2573, "lr": 5.930764383027965e-06}
{"train_loss": 0.0597422793507576, "global_step": 229016, "epoch": 2573, "lr": 5.930490518892018e-06}
{"train_loss": 0.0410362109541893, "global_step": 229017, "epoch": 2573, "lr": 5.930216660680671e-06}
{"train_loss": 0.03246580809354782, "global_step": 229018, "epoch": 2573, "lr": 5.929942808393979e-06}
{"train_loss": 0.06731529533863068, "global_step": 229019, "epoch": 2573, "lr": 5.9296689620319535e-06}
{"train_loss": 0.07727747410535812, "global_step": 229020, "epoch": 2573, "lr": 5.929395121594661e-06}
{"train_loss": 0.0280732624232769, "global_step": 229021, "epoch": 2573, "lr": 5.929121287082118e-06}
{"train_loss": 0.07770898938179016, "global_step": 229022, "epoch": 2573, "lr": 5.928847458494369e-06}
{"train_loss": 0.009240051731467247, "global_step": 229023, "epoch": 2573, "lr": 5.928573635831447e-06}
{"train_loss": 0.049445778131484985, "global_step": 229024, "epoch": 2573, "lr": 5.928299819093397e-06}
{"train_loss": 0.05316554754972458, "global_step": 229025, "epoch": 2573, "lr": 5.928026008280241e-06}
{"train_loss": 0.025272253900766373, "global_step": 229026, "epoch": 2573, "lr": 5.927752203392034e-06}
{"train_loss": 0.07004678249359131, "global_step": 229027, "epoch": 2573, "lr": 5.927478404428793e-06}
{"train_loss": 0.03728129714727402, "global_step": 229028, "epoch": 2573, "lr": 5.927204611390574e-06}
{"train_loss": 0.03692296892404556, "global_step": 229029, "epoch": 2573, "lr": 5.926930824277399e-06}
{"train_loss": 0.06042001023888588, "global_step": 229030, "epoch": 2573, "lr": 5.9266570430893235e-06}
{"train_loss": 0.037891075015068054, "global_step": 229031, "epoch": 2573, "lr": 5.926383267826358e-06}
{"train_loss": 0.08709225058555603, "global_step": 229032, "epoch": 2573, "lr": 5.926109498488553e-06}
{"train_loss": 0.06496953219175339, "global_step": 229033, "epoch": 2573, "lr": 5.9258357350759515e-06}
{"train_loss": 0.0376911461353302, "global_step": 229034, "epoch": 2573, "lr": 5.925561977588579e-06}
{"train_loss": 0.05138269439339638, "global_step": 229035, "epoch": 2573, "lr": 5.925288226026488e-06}
{"train_loss": 0.055958617478609085, "global_step": 229036, "epoch": 2573, "lr": 5.92501448038969e-06}
{"train_loss": 0.02912958152592182, "global_step": 229037, "epoch": 2573, "lr": 5.9247407406782525e-06}
{"train_loss": 0.05541562661528587, "global_step": 229038, "epoch": 2573, "lr": 5.924467006892181e-06}
{"train_loss": 0.032765209674835205, "global_step": 229039, "epoch": 2573, "lr": 5.9241932790315425e-06}
{"train_loss": 0.0632011815905571, "global_step": 229040, "epoch": 2573, "lr": 5.9239195570963464e-06}
{"train_loss": 0.05133117362856865, "global_step": 229041, "epoch": 2573, "lr": 5.92364584108665e-06}
{"train_loss": 0.024631790816783905, "global_step": 229042, "epoch": 2573, "lr": 5.923372131002481e-06}
{"train_loss": 0.034280069172382355, "global_step": 229043, "epoch": 2573, "lr": 5.923098426843876e-06}
{"train_loss": 0.03478115051984787, "global_step": 229044, "epoch": 2573, "lr": 5.922824728610882e-06}
{"train_loss": 0.020618677139282227, "global_step": 229045, "epoch": 2573, "lr": 5.9225510363035145e-06}
{"train_loss": 0.0535239651799202, "global_step": 229046, "epoch": 2573, "lr": 5.922277349921828e-06}
{"train_loss": 0.054085761308670044, "global_step": 229047, "epoch": 2573, "lr": 5.922003669465848e-06}
{"train_loss": 0.022260330617427826, "global_step": 229048, "epoch": 2573, "lr": 5.921729994935626e-06}
{"train_loss": 0.028806013986468315, "global_step": 229049, "epoch": 2573, "lr": 5.9214563263311816e-06}
{"train_loss": 0.01675296016037464, "global_step": 229050, "epoch": 2573, "lr": 5.921182663652575e-06}
{"train_loss": 0.05159558728337288, "global_step": 229051, "epoch": 2573, "lr": 5.920909006899811e-06}
{"train_loss": 0.08284218609333038, "global_step": 229052, "epoch": 2573, "lr": 5.920635356072956e-06}
{"train_loss": 0.02891130931675434, "global_step": 229053, "epoch": 2573, "lr": 5.920361711172024e-06}
{"train_loss": 0.05977265536785126, "global_step": 229054, "epoch": 2573, "lr": 5.920088072197072e-06}
{"train_loss": 0.033918753266334534, "global_step": 229055, "epoch": 2573, "lr": 5.919814439148114e-06}
{"train_loss": 0.07154916971921921, "global_step": 229056, "epoch": 2573, "lr": 5.9195408120252145e-06}
{"train_loss": 0.09518004208803177, "global_step": 229057, "epoch": 2573, "lr": 5.919267190828381e-06}
{"train_loss": 0.036389417946338654, "global_step": 229058, "epoch": 2573, "lr": 5.918993575557679e-06}
{"train_loss": 0.0518367737531662, "global_step": 229059, "epoch": 2573, "lr": 5.918719966213121e-06}
{"train_loss": 0.038153648376464844, "global_step": 229060, "epoch": 2573, "lr": 5.91844636279476e-06}
{"train_loss": 0.06365795433521271, "global_step": 229061, "epoch": 2573, "lr": 5.918172765302621e-06}
{"train_loss": 0.05892515182495117, "global_step": 229062, "epoch": 2573, "lr": 5.917899173736752e-06}
{"train_loss": 0.10259123146533966, "global_step": 229063, "epoch": 2573, "lr": 5.917625588097181e-06}
{"train_loss": 0.021060235798358917, "global_step": 229064, "epoch": 2573, "lr": 5.917352008383953e-06}
{"train_loss": 0.07923676818609238, "global_step": 229065, "epoch": 2573, "lr": 5.91707843459709e-06}
{"train_loss": 0.04791224002838135, "global_step": 229066, "epoch": 2573, "lr": 5.916804866736647e-06}
{"train_loss": 0.04325259476900101, "global_step": 229067, "epoch": 2573, "lr": 5.916531304802642e-06}
{"train_loss": 0.04795657470822334, "global_step": 229068, "epoch": 2573, "lr": 5.9162577487951234e-06}
{"train_loss": 0.047139473259449005, "global_step": 229069, "epoch": 2573, "lr": 5.915984198714141e-06}
{"train_loss": 0.039979953318834305, "global_step": 229070, "epoch": 2573, "lr": 5.915710654559703e-06}
{"train_loss": 0.055176734924316406, "global_step": 229071, "epoch": 2573, "lr": 5.915437116331868e-06}
{"train_loss": 0.032782912254333496, "global_step": 229072, "epoch": 2573, "lr": 5.915163584030658e-06}
{"train_loss": 0.03828926011919975, "global_step": 229073, "epoch": 2573, "lr": 5.914890057656131e-06}
{"train_loss": 0.06008090078830719, "global_step": 229074, "epoch": 2573, "lr": 5.914616537208295e-06}
{"train_loss": 0.08322043716907501, "global_step": 229075, "epoch": 2573, "lr": 5.914343022687208e-06}
{"train_loss": 0.02853061445057392, "global_step": 229076, "epoch": 2573, "lr": 5.914069514092896e-06}
{"train_loss": 0.07745806127786636, "global_step": 229077, "epoch": 2573, "lr": 5.91379601142541e-06}
{"train_loss": 0.0353381372988224, "global_step": 229078, "epoch": 2573, "lr": 5.913522514684766e-06}
{"train_loss": 0.02390372008085251, "global_step": 229079, "epoch": 2573, "lr": 5.91324902387102e-06}
{"train_loss": 0.034348905086517334, "global_step": 229080, "epoch": 2573, "lr": 5.912975538984189e-06}
{"train_loss": 0.05403430759906769, "global_step": 229081, "epoch": 2573, "lr": 5.912702060024333e-06}
{"train_loss": 0.050781745463609695, "global_step": 229082, "epoch": 2573, "lr": 5.9124285869914754e-06}
{"train_loss": 0.06778579205274582, "global_step": 229083, "epoch": 2573, "lr": 5.912155119885649e-06}
{"train_loss": 0.019189348444342613, "global_step": 229084, "epoch": 2573, "lr": 5.911881658706903e-06}
{"train_loss": 0.0485929733958472, "global_step": 229085, "epoch": 2573, "lr": 5.9116082034552545e-06, "val_loss": 9.181193351745605}
{"train_loss": 0.047596611082553864, "global_step": 229086, "epoch": 2574, "lr": 5.911334754130765e-06}
{"train_loss": 0.0973927304148674, "global_step": 229087, "epoch": 2574, "lr": 5.911061310733451e-06}
{"train_loss": 0.038133349269628525, "global_step": 229088, "epoch": 2574, "lr": 5.9107878732633686e-06}
{"train_loss": 0.07448741048574448, "global_step": 229089, "epoch": 2574, "lr": 5.9105144417205275e-06}
{"train_loss": 0.033532481640577316, "global_step": 229090, "epoch": 2574, "lr": 5.910241016104995e-06}
{"train_loss": 0.04746975004673004, "global_step": 229091, "epoch": 2574, "lr": 5.9099675964167825e-06}
{"train_loss": 0.041344307363033295, "global_step": 229092, "epoch": 2574, "lr": 5.9096941826559506e-06}
{"train_loss": 0.06465687602758408, "global_step": 229093, "epoch": 2574, "lr": 5.909420774822505e-06}
{"train_loss": 0.05834333226084709, "global_step": 229094, "epoch": 2574, "lr": 5.909147372916518e-06}
{"train_loss": 0.02593165636062622, "global_step": 229095, "epoch": 2574, "lr": 5.908873976937995e-06}
{"train_loss": 0.04360350966453552, "global_step": 229096, "epoch": 2574, "lr": 5.9086005868869975e-06}
{"train_loss": 0.04394029453396797, "global_step": 229097, "epoch": 2574, "lr": 5.908327202763542e-06}
{"train_loss": 0.07503876835107803, "global_step": 229098, "epoch": 2574, "lr": 5.908053824567683e-06}
{"train_loss": 0.041039109230041504, "global_step": 229099, "epoch": 2574, "lr": 5.907780452299438e-06}
{"train_loss": 0.05229545757174492, "global_step": 229100, "epoch": 2574, "lr": 5.907507085958863e-06}
{"train_loss": 0.03682103380560875, "global_step": 229101, "epoch": 2574, "lr": 5.90723372554598e-06}
{"train_loss": 0.06472600996494293, "global_step": 229102, "epoch": 2574, "lr": 5.906960371060838e-06}
{"train_loss": 0.029997266829013824, "global_step": 229103, "epoch": 2574, "lr": 5.906687022503466e-06}
{"train_loss": 0.037375546991825104, "global_step": 229104, "epoch": 2574, "lr": 5.906413679873896e-06}
{"train_loss": 0.09036778658628464, "global_step": 229105, "epoch": 2574, "lr": 5.9061403431721785e-06}
{"train_loss": 0.05394568666815758, "global_step": 229106, "epoch": 2574, "lr": 5.9058670123983366e-06}
{"train_loss": 0.01752639003098011, "global_step": 229107, "epoch": 2574, "lr": 5.905593687552424e-06}
{"train_loss": 0.04825534299015999, "global_step": 229108, "epoch": 2574, "lr": 5.905320368634459e-06}
{"train_loss": 0.05433667078614235, "global_step": 229109, "epoch": 2574, "lr": 5.905047055644492e-06}
{"train_loss": 0.058123745024204254, "global_step": 229110, "epoch": 2574, "lr": 5.9047737485825484e-06}
{"train_loss": 0.054886434227228165, "global_step": 229111, "epoch": 2574, "lr": 5.904500447448675e-06}
{"train_loss": 0.04118814691901207, "global_step": 229112, "epoch": 2574, "lr": 5.904227152242897e-06}
{"train_loss": 0.03253379464149475, "global_step": 229113, "epoch": 2574, "lr": 5.903953862965267e-06}
{"train_loss": 0.06209034100174904, "global_step": 229114, "epoch": 2574, "lr": 5.9036805796158055e-06}
{"train_loss": 0.059546101838350296, "global_step": 229115, "epoch": 2574, "lr": 5.903407302194569e-06}
{"train_loss": 0.05008578673005104, "global_step": 229116, "epoch": 2574, "lr": 5.903134030701568e-06}
{"train_loss": 0.03461259976029396, "global_step": 229117, "epoch": 2574, "lr": 5.902860765136858e-06}
{"train_loss": 0.08701623231172562, "global_step": 229118, "epoch": 2574, "lr": 5.902587505500468e-06}
{"train_loss": 0.01870908960700035, "global_step": 229119, "epoch": 2574, "lr": 5.902314251792445e-06}
{"train_loss": 0.025754524394869804, "global_step": 229120, "epoch": 2574, "lr": 5.90204100401282e-06}
{"train_loss": 0.06327027082443237, "global_step": 229121, "epoch": 2574, "lr": 5.9017677621616195e-06}
{"train_loss": 0.027228662744164467, "global_step": 229122, "epoch": 2574, "lr": 5.901494526238899e-06}
{"train_loss": 0.06368790566921234, "global_step": 229123, "epoch": 2574, "lr": 5.9012212962446685e-06}
{"train_loss": 0.10861315578222275, "global_step": 229124, "epoch": 2574, "lr": 5.900948072178996e-06}
{"train_loss": 0.0336763896048069, "global_step": 229125, "epoch": 2574, "lr": 5.900674854041893e-06}
{"train_loss": 0.11012745648622513, "global_step": 229126, "epoch": 2574, "lr": 5.900401641833419e-06}
{"train_loss": 0.10449090600013733, "global_step": 229127, "epoch": 2574, "lr": 5.900128435553587e-06}
{"train_loss": 0.05764301121234894, "global_step": 229128, "epoch": 2574, "lr": 5.899855235202456e-06}
{"train_loss": 0.042570725083351135, "global_step": 229129, "epoch": 2574, "lr": 5.899582040780038e-06}
{"train_loss": 0.03536589816212654, "global_step": 229130, "epoch": 2574, "lr": 5.8993088522863945e-06}
{"train_loss": 0.05108848214149475, "global_step": 229131, "epoch": 2574, "lr": 5.899035669721548e-06}
{"train_loss": 0.039029449224472046, "global_step": 229132, "epoch": 2574, "lr": 5.898762493085541e-06}
{"train_loss": 0.04623144492506981, "global_step": 229133, "epoch": 2574, "lr": 5.898489322378403e-06}
{"train_loss": 0.05497169867157936, "global_step": 229134, "epoch": 2574, "lr": 5.898216157600184e-06}
{"train_loss": 0.06820690631866455, "global_step": 229135, "epoch": 2574, "lr": 5.897942998750899e-06}
{"train_loss": 0.03961166739463806, "global_step": 229136, "epoch": 2574, "lr": 5.897669845830612e-06}
{"train_loss": 0.05760014429688454, "global_step": 229137, "epoch": 2574, "lr": 5.89739669883933e-06}
{"train_loss": 0.029386749491095543, "global_step": 229138, "epoch": 2574, "lr": 5.8971235577771236e-06}
{"train_loss": 0.03690178692340851, "global_step": 229139, "epoch": 2574, "lr": 5.896850422643996e-06}
{"train_loss": 0.032509997487068176, "global_step": 229140, "epoch": 2574, "lr": 5.896577293440003e-06}
{"train_loss": 0.03534350544214249, "global_step": 229141, "epoch": 2574, "lr": 5.896304170165184e-06}
{"train_loss": 0.059080787003040314, "global_step": 229142, "epoch": 2574, "lr": 5.8960310528195615e-06}
{"train_loss": 0.05275234580039978, "global_step": 229143, "epoch": 2574, "lr": 5.89575794140319e-06}
{"train_loss": 0.04559113085269928, "global_step": 229144, "epoch": 2574, "lr": 5.895484835916087e-06}
{"train_loss": 0.06253187358379364, "global_step": 229145, "epoch": 2574, "lr": 5.895211736358308e-06}
{"train_loss": 0.028707381337881088, "global_step": 229146, "epoch": 2574, "lr": 5.894938642729869e-06}
{"train_loss": 0.07978245615959167, "global_step": 229147, "epoch": 2574, "lr": 5.894665555030826e-06}
{"train_loss": 0.04762592166662216, "global_step": 229148, "epoch": 2574, "lr": 5.894392473261201e-06}
{"train_loss": 0.035235702991485596, "global_step": 229149, "epoch": 2574, "lr": 5.89411939742105e-06}
{"train_loss": 0.06689554452896118, "global_step": 229150, "epoch": 2574, "lr": 5.8938463275103825e-06}
{"train_loss": 0.019064169377088547, "global_step": 229151, "epoch": 2574, "lr": 5.8935732635292566e-06}
{"train_loss": 0.04030901938676834, "global_step": 229152, "epoch": 2574, "lr": 5.8933002054776975e-06}
{"train_loss": 0.0773133635520935, "global_step": 229153, "epoch": 2574, "lr": 5.893027153355757e-06}
{"train_loss": 0.044811345636844635, "global_step": 229154, "epoch": 2574, "lr": 5.8927541071634505e-06}
{"train_loss": 0.0457439161837101, "global_step": 229155, "epoch": 2574, "lr": 5.892481066900834e-06}
{"train_loss": 0.07409213483333588, "global_step": 229156, "epoch": 2574, "lr": 5.892208032567925e-06}
{"train_loss": 0.0581185445189476, "global_step": 229157, "epoch": 2574, "lr": 5.891935004164783e-06}
{"train_loss": 0.0463537722826004, "global_step": 229158, "epoch": 2574, "lr": 5.891661981691426e-06}
{"train_loss": 0.015383129939436913, "global_step": 229159, "epoch": 2574, "lr": 5.8913889651479035e-06}
{"train_loss": 0.019882842898368835, "global_step": 229160, "epoch": 2574, "lr": 5.8911159545342485e-06}
{"train_loss": 0.04694472998380661, "global_step": 229161, "epoch": 2574, "lr": 5.890842949850483e-06}
{"train_loss": 0.06738520413637161, "global_step": 229162, "epoch": 2574, "lr": 5.890569951096664e-06}
{"train_loss": 0.05651925504207611, "global_step": 229163, "epoch": 2574, "lr": 5.890296958272817e-06}
{"train_loss": 0.07549775391817093, "global_step": 229164, "epoch": 2574, "lr": 5.890023971378988e-06}
{"train_loss": 0.025656290352344513, "global_step": 229165, "epoch": 2574, "lr": 5.889750990415194e-06}
{"train_loss": 0.04512779414653778, "global_step": 229166, "epoch": 2574, "lr": 5.8894780153815e-06}
{"train_loss": 0.042097218334674835, "global_step": 229167, "epoch": 2574, "lr": 5.889205046277918e-06}
{"train_loss": 0.05717899650335312, "global_step": 229168, "epoch": 2574, "lr": 5.888932083104504e-06}
{"train_loss": 0.035717494785785675, "global_step": 229169, "epoch": 2574, "lr": 5.888659125861273e-06}
{"train_loss": 0.03914855420589447, "global_step": 229170, "epoch": 2574, "lr": 5.888386174548288e-06}
{"train_loss": 0.06622419506311417, "global_step": 229171, "epoch": 2574, "lr": 5.888113229165559e-06}
{"train_loss": 0.05884426087141037, "global_step": 229172, "epoch": 2574, "lr": 5.887840289713148e-06}
{"train_loss": 0.03932595252990723, "global_step": 229173, "epoch": 2574, "lr": 5.887567356191065e-06}
{"train_loss": 0.050518274830465906, "global_step": 229174, "epoch": 2574, "lr": 5.887294428599377e-06, "val_loss": 9.225638389587402}
{"train_loss": 0.030065732076764107, "global_step": 229175, "epoch": 2575, "lr": 5.887021506938089e-06}
{"train_loss": 0.018640996888279915, "global_step": 229176, "epoch": 2575, "lr": 5.8867485912072585e-06}
{"train_loss": 0.036882754415273666, "global_step": 229177, "epoch": 2575, "lr": 5.886475681406922e-06}
{"train_loss": 0.09977704286575317, "global_step": 229178, "epoch": 2575, "lr": 5.8862027775371035e-06}
{"train_loss": 0.04832509532570839, "global_step": 229179, "epoch": 2575, "lr": 5.8859298795978565e-06}
{"train_loss": 0.01730293221771717, "global_step": 229180, "epoch": 2575, "lr": 5.8856569875892e-06}
{"train_loss": 0.07340084761381149, "global_step": 229181, "epoch": 2575, "lr": 5.885384101511188e-06}
{"train_loss": 0.03898254409432411, "global_step": 229182, "epoch": 2575, "lr": 5.885111221363837e-06}
{"train_loss": 0.10907021164894104, "global_step": 229183, "epoch": 2575, "lr": 5.884838347147209e-06}
{"train_loss": 0.0388813316822052, "global_step": 229184, "epoch": 2575, "lr": 5.884565478861315e-06}
{"train_loss": 0.06534835696220398, "global_step": 229185, "epoch": 2575, "lr": 5.884292616506215e-06}
{"train_loss": 0.06077812984585762, "global_step": 229186, "epoch": 2575, "lr": 5.884019760081921e-06}
{"train_loss": 0.05021511763334274, "global_step": 229187, "epoch": 2575, "lr": 5.883746909588495e-06}
{"train_loss": 0.04014742374420166, "global_step": 229188, "epoch": 2575, "lr": 5.883474065025951e-06}
{"train_loss": 0.051181841641664505, "global_step": 229189, "epoch": 2575, "lr": 5.883201226394347e-06}
{"train_loss": 0.0331408716738224, "global_step": 229190, "epoch": 2575, "lr": 5.882928393693699e-06}
{"train_loss": 0.02233438938856125, "global_step": 229191, "epoch": 2575, "lr": 5.882655566924062e-06}
{"train_loss": 0.06646469980478287, "global_step": 229192, "epoch": 2575, "lr": 5.882382746085452e-06}
{"train_loss": 0.08016760647296906, "global_step": 229193, "epoch": 2575, "lr": 5.882109931177932e-06}
{"train_loss": 0.042213935405015945, "global_step": 229194, "epoch": 2575, "lr": 5.881837122201511e-06}
{"train_loss": 0.03359117731451988, "global_step": 229195, "epoch": 2575, "lr": 5.881564319156252e-06}
{"train_loss": 0.03455678001046181, "global_step": 229196, "epoch": 2575, "lr": 5.881291522042171e-06}
{"train_loss": 0.020965952426195145, "global_step": 229197, "epoch": 2575, "lr": 5.881018730859317e-06}
{"train_loss": 0.05541303753852844, "global_step": 229198, "epoch": 2575, "lr": 5.88074594560773e-06}
{"train_loss": 0.08613215386867523, "global_step": 229199, "epoch": 2575, "lr": 5.8804731662874204e-06}
{"train_loss": 0.031278204172849655, "global_step": 229200, "epoch": 2575, "lr": 5.88020039289846e-06}
{"train_loss": 0.08743900805711746, "global_step": 229201, "epoch": 2575, "lr": 5.879927625440856e-06}
{"train_loss": 0.05162797123193741, "global_step": 229202, "epoch": 2575, "lr": 5.879654863914669e-06}
{"train_loss": 0.059633396565914154, "global_step": 229203, "epoch": 2575, "lr": 5.879382108319914e-06}
{"train_loss": 0.044473081827163696, "global_step": 229204, "epoch": 2575, "lr": 5.879109358656648e-06}
{"train_loss": 0.10046310722827911, "global_step": 229205, "epoch": 2575, "lr": 5.8788366149248875e-06}
{"train_loss": 0.051054697483778, "global_step": 229206, "epoch": 2575, "lr": 5.8785638771246876e-06}
{"train_loss": 0.09120939671993256, "global_step": 229207, "epoch": 2575, "lr": 5.87829114525607e-06}
{"train_loss": 0.02760937251150608, "global_step": 229208, "epoch": 2575, "lr": 5.878018419319092e-06}
{"train_loss": 0.04816649854183197, "global_step": 229209, "epoch": 2575, "lr": 5.877745699313763e-06}
{"train_loss": 0.04272414371371269, "global_step": 229210, "epoch": 2575, "lr": 5.87747298524014e-06}
{"train_loss": 0.06094719469547272, "global_step": 229211, "epoch": 2575, "lr": 5.877200277098249e-06}
{"train_loss": 0.06304990500211716, "global_step": 229212, "epoch": 2575, "lr": 5.876927574888125e-06}
{"train_loss": 0.04796340689063072, "global_step": 229213, "epoch": 2575, "lr": 5.876654878609822e-06}
{"train_loss": 0.07384339720010757, "global_step": 229214, "epoch": 2575, "lr": 5.876382188263357e-06}
{"train_loss": 0.047071728855371475, "global_step": 229215, "epoch": 2575, "lr": 5.876109503848787e-06}
{"train_loss": 0.05026239901781082, "global_step": 229216, "epoch": 2575, "lr": 5.875836825366121e-06}
{"train_loss": 0.048408668488264084, "global_step": 229217, "epoch": 2575, "lr": 5.875564152815421e-06}
{"train_loss": 0.015146465040743351, "global_step": 229218, "epoch": 2575, "lr": 5.87529148619671e-06}
{"train_loss": 0.03102118708193302, "global_step": 229219, "epoch": 2575, "lr": 5.875018825510031e-06}
{"train_loss": 0.0908910408616066, "global_step": 229220, "epoch": 2575, "lr": 5.874746170755413e-06}
{"train_loss": 0.02102772146463394, "global_step": 229221, "epoch": 2575, "lr": 5.874473521932905e-06}
{"train_loss": 0.040958378463983536, "global_step": 229222, "epoch": 2575, "lr": 5.874200879042524e-06}
{"train_loss": 0.06385315209627151, "global_step": 229223, "epoch": 2575, "lr": 5.873928242084331e-06}
{"train_loss": 0.0636201798915863, "global_step": 229224, "epoch": 2575, "lr": 5.873655611058343e-06}
{"train_loss": 0.07057839632034302, "global_step": 229225, "epoch": 2575, "lr": 5.8733829859646096e-06}
{"train_loss": 0.05850548669695854, "global_step": 229226, "epoch": 2575, "lr": 5.873110366803159e-06}
{"train_loss": 0.0463063046336174, "global_step": 229227, "epoch": 2575, "lr": 5.872837753574034e-06}
{"train_loss": 0.06349573284387589, "global_step": 229228, "epoch": 2575, "lr": 5.872565146277259e-06}
{"train_loss": 0.03573736175894737, "global_step": 229229, "epoch": 2575, "lr": 5.872292544912894e-06}
{"train_loss": 0.08817996829748154, "global_step": 229230, "epoch": 2575, "lr": 5.8720199494809446e-06}
{"train_loss": 0.040565624833106995, "global_step": 229231, "epoch": 2575, "lr": 5.871747359981483e-06}
{"train_loss": 0.0399920716881752, "global_step": 229232, "epoch": 2575, "lr": 5.8714747764145096e-06}
{"train_loss": 0.027150223031640053, "global_step": 229233, "epoch": 2575, "lr": 5.871202198780091e-06}
{"train_loss": 0.04581121355295181, "global_step": 229234, "epoch": 2575, "lr": 5.870929627078242e-06}
{"train_loss": 0.05237620323896408, "global_step": 229235, "epoch": 2575, "lr": 5.870657061309021e-06}
{"train_loss": 0.0724167600274086, "global_step": 229236, "epoch": 2575, "lr": 5.870384501472437e-06}
{"train_loss": 0.05148206278681755, "global_step": 229237, "epoch": 2575, "lr": 5.8701119475685575e-06}
{"train_loss": 0.025781186297535896, "global_step": 229238, "epoch": 2575, "lr": 5.869839399597399e-06}
{"train_loss": 0.06848496198654175, "global_step": 229239, "epoch": 2575, "lr": 5.869566857558994e-06}
{"train_loss": 0.0272755715996027, "global_step": 229240, "epoch": 2575, "lr": 5.8692943214534006e-06}
{"train_loss": 0.024686699733138084, "global_step": 229241, "epoch": 2575, "lr": 5.869021791280632e-06}
{"train_loss": 0.011754552833735943, "global_step": 229242, "epoch": 2575, "lr": 5.868749267040741e-06}
{"train_loss": 0.057819753885269165, "global_step": 229243, "epoch": 2575, "lr": 5.868476748733753e-06}
{"train_loss": 0.05868091061711311, "global_step": 229244, "epoch": 2575, "lr": 5.86820423635972e-06}
{"train_loss": 0.030611541122198105, "global_step": 229245, "epoch": 2575, "lr": 5.867931729918658e-06}
{"train_loss": 0.06513114273548126, "global_step": 229246, "epoch": 2575, "lr": 5.867659229410627e-06}
{"train_loss": 0.04503660276532173, "global_step": 229247, "epoch": 2575, "lr": 5.867386734835639e-06}
{"train_loss": 0.08493614196777344, "global_step": 229248, "epoch": 2575, "lr": 5.86711424619375e-06}
{"train_loss": 0.053861103951931, "global_step": 229249, "epoch": 2575, "lr": 5.866841763484982e-06}
{"train_loss": 0.01886696182191372, "global_step": 229250, "epoch": 2575, "lr": 5.866569286709384e-06}
{"train_loss": 0.02862144261598587, "global_step": 229251, "epoch": 2575, "lr": 5.86629681586699e-06}
{"train_loss": 0.054376911371946335, "global_step": 229252, "epoch": 2575, "lr": 5.866024350957833e-06}
{"train_loss": 0.060547567903995514, "global_step": 229253, "epoch": 2575, "lr": 5.8657518919819525e-06}
{"train_loss": 0.07303395122289658, "global_step": 229254, "epoch": 2575, "lr": 5.865479438939381e-06}
{"train_loss": 0.011312434449791908, "global_step": 229255, "epoch": 2575, "lr": 5.865206991830163e-06}
{"train_loss": 0.09114351868629456, "global_step": 229256, "epoch": 2575, "lr": 5.864934550654322e-06}
{"train_loss": 0.028314882889389992, "global_step": 229257, "epoch": 2575, "lr": 5.864662115411912e-06}
{"train_loss": 0.03682994842529297, "global_step": 229258, "epoch": 2575, "lr": 5.86438968610295e-06}
{"train_loss": 0.04041978716850281, "global_step": 229259, "epoch": 2575, "lr": 5.864117262727492e-06}
{"train_loss": 0.03861100226640701, "global_step": 229260, "epoch": 2575, "lr": 5.86384484528556e-06}
{"train_loss": 0.03331400454044342, "global_step": 229261, "epoch": 2575, "lr": 5.863572433777204e-06}
{"train_loss": 0.04860452190041542, "global_step": 229262, "epoch": 2575, "lr": 5.86330002820244e-06}
{"train_loss": 0.0500033414095975, "global_step": 229263, "epoch": 2575, "lr": 5.86302762856133e-06, "val_loss": 9.348605155944824, "train_action_mse_error": 14.724099159240723}
{"train_loss": 0.020861759781837463, "global_step": 229264, "epoch": 2576, "lr": 5.862755234853884e-06}
{"train_loss": 0.05701685696840286, "global_step": 229265, "epoch": 2576, "lr": 5.862482847080164e-06}
{"train_loss": 0.038054656237363815, "global_step": 229266, "epoch": 2576, "lr": 5.862210465240187e-06}
{"train_loss": 0.03161732852458954, "global_step": 229267, "epoch": 2576, "lr": 5.861938089334001e-06}
{"train_loss": 0.06773020327091217, "global_step": 229268, "epoch": 2576, "lr": 5.8616657193616475e-06}
{"train_loss": 0.014712304808199406, "global_step": 229269, "epoch": 2576, "lr": 5.861393355323141e-06}
{"train_loss": 0.053333718329668045, "global_step": 229270, "epoch": 2576, "lr": 5.861120997218544e-06}
{"train_loss": 0.041573915630578995, "global_step": 229271, "epoch": 2576, "lr": 5.860848645047884e-06}
{"train_loss": 0.05679924041032791, "global_step": 229272, "epoch": 2576, "lr": 5.860576298811182e-06}
{"train_loss": 0.047148920595645905, "global_step": 229273, "epoch": 2576, "lr": 5.860303958508501e-06}
{"train_loss": 0.01269941870123148, "global_step": 229274, "epoch": 2576, "lr": 5.86003162413985e-06}
{"train_loss": 0.02700456604361534, "global_step": 229275, "epoch": 2576, "lr": 5.859759295705292e-06}
{"train_loss": 0.04901048541069031, "global_step": 229276, "epoch": 2576, "lr": 5.859486973204853e-06}
{"train_loss": 0.052282992750406265, "global_step": 229277, "epoch": 2576, "lr": 5.8592146566385506e-06}
{"train_loss": 0.07427652925252914, "global_step": 229278, "epoch": 2576, "lr": 5.858942346006458e-06}
{"train_loss": 0.059861790388822556, "global_step": 229279, "epoch": 2576, "lr": 5.858670041308578e-06}
{"train_loss": 0.028283627703785896, "global_step": 229280, "epoch": 2576, "lr": 5.858397742544974e-06}
{"train_loss": 0.08028951287269592, "global_step": 229281, "epoch": 2576, "lr": 5.858125449715657e-06}
{"train_loss": 0.07630220055580139, "global_step": 229282, "epoch": 2576, "lr": 5.857853162820686e-06}
{"train_loss": 0.07191921770572662, "global_step": 229283, "epoch": 2576, "lr": 5.857580881860087e-06}
{"train_loss": 0.05745788663625717, "global_step": 229284, "epoch": 2576, "lr": 5.8573086068339004e-06}
{"train_loss": 0.13212303817272186, "global_step": 229285, "epoch": 2576, "lr": 5.857036337742156e-06}
{"train_loss": 0.0586676150560379, "global_step": 229286, "epoch": 2576, "lr": 5.856764074584886e-06}
{"train_loss": 0.0808509811758995, "global_step": 229287, "epoch": 2576, "lr": 5.856491817362153e-06}
{"train_loss": 0.08827882260084152, "global_step": 229288, "epoch": 2576, "lr": 5.856219566073967e-06}
{"train_loss": 0.08567358553409576, "global_step": 229289, "epoch": 2576, "lr": 5.855947320720384e-06}
{"train_loss": 0.04486512020230293, "global_step": 229290, "epoch": 2576, "lr": 5.85567508130142e-06}
{"train_loss": 0.06535939872264862, "global_step": 229291, "epoch": 2576, "lr": 5.8554028478171316e-06}
{"train_loss": 0.03728774935007095, "global_step": 229292, "epoch": 2576, "lr": 5.855130620267535e-06}
{"train_loss": 0.07314840704202652, "global_step": 229293, "epoch": 2576, "lr": 5.85485839865269e-06}
{"train_loss": 0.03387666493654251, "global_step": 229294, "epoch": 2576, "lr": 5.854586182972616e-06}
{"train_loss": 0.03215961530804634, "global_step": 229295, "epoch": 2576, "lr": 5.85431397322736e-06}
{"train_loss": 0.04698159918189049, "global_step": 229296, "epoch": 2576, "lr": 5.8540417694169405e-06}
{"train_loss": 0.014492268674075603, "global_step": 229297, "epoch": 2576, "lr": 5.853769571541418e-06}
{"train_loss": 0.053631603717803955, "global_step": 229298, "epoch": 2576, "lr": 5.853497379600814e-06}
{"train_loss": 0.045513857156038284, "global_step": 229299, "epoch": 2576, "lr": 5.853225193595174e-06}
{"train_loss": 0.05086351931095123, "global_step": 229300, "epoch": 2576, "lr": 5.85295301352452e-06}
{"train_loss": 0.038582202047109604, "global_step": 229301, "epoch": 2576, "lr": 5.852680839388908e-06}
{"train_loss": 0.033234551548957825, "global_step": 229302, "epoch": 2576, "lr": 5.852408671188359e-06}
{"train_loss": 0.1219833567738533, "global_step": 229303, "epoch": 2576, "lr": 5.852136508922912e-06}
{"train_loss": 0.017276916652917862, "global_step": 229304, "epoch": 2576, "lr": 5.851864352592617e-06}
{"train_loss": 0.0540766566991806, "global_step": 229305, "epoch": 2576, "lr": 5.851592202197492e-06}
{"train_loss": 0.05492676794528961, "global_step": 229306, "epoch": 2576, "lr": 5.851320057737592e-06}
{"train_loss": 0.05163684859871864, "global_step": 229307, "epoch": 2576, "lr": 5.8510479192129376e-06}
{"train_loss": 0.05612091347575188, "global_step": 229308, "epoch": 2576, "lr": 5.850775786623575e-06}
{"train_loss": 0.1297333538532257, "global_step": 229309, "epoch": 2576, "lr": 5.850503659969542e-06}
{"train_loss": 0.03510815277695656, "global_step": 229310, "epoch": 2576, "lr": 5.850231539250861e-06}
{"train_loss": 0.07834487408399582, "global_step": 229311, "epoch": 2576, "lr": 5.8499594244675885e-06}
{"train_loss": 0.10341530293226242, "global_step": 229312, "epoch": 2576, "lr": 5.849687315619734e-06}
{"train_loss": 0.031109647825360298, "global_step": 229313, "epoch": 2576, "lr": 5.849415212707371e-06}
{"train_loss": 0.024041784927248955, "global_step": 229314, "epoch": 2576, "lr": 5.84914311573051e-06}
{"train_loss": 0.033841732889413834, "global_step": 229315, "epoch": 2576, "lr": 5.848871024689184e-06}
{"train_loss": 0.06264355033636093, "global_step": 229316, "epoch": 2576, "lr": 5.84859893958345e-06}
{"train_loss": 0.06354472786188126, "global_step": 229317, "epoch": 2576, "lr": 5.8483268604133224e-06}
{"train_loss": 0.06332305818796158, "global_step": 229318, "epoch": 2576, "lr": 5.848054787178858e-06}
{"train_loss": 0.04189978539943695, "global_step": 229319, "epoch": 2576, "lr": 5.84778271988008e-06}
{"train_loss": 0.07293500006198883, "global_step": 229320, "epoch": 2576, "lr": 5.8475106585170355e-06}
{"train_loss": 0.07994285970926285, "global_step": 229321, "epoch": 2576, "lr": 5.847238603089744e-06}
{"train_loss": 0.06452857702970505, "global_step": 229322, "epoch": 2576, "lr": 5.8469665535982545e-06}
{"train_loss": 0.046173159033060074, "global_step": 229323, "epoch": 2576, "lr": 5.846694510042616e-06}
{"train_loss": 0.06172274053096771, "global_step": 229324, "epoch": 2576, "lr": 5.846422472422835e-06}
{"train_loss": 0.045033037662506104, "global_step": 229325, "epoch": 2576, "lr": 5.8461504407389785e-06}
{"train_loss": 0.044440820813179016, "global_step": 229326, "epoch": 2576, "lr": 5.845878414991057e-06}
{"train_loss": 0.014737051911652088, "global_step": 229327, "epoch": 2576, "lr": 5.845606395179131e-06}
{"train_loss": 0.0579969622194767, "global_step": 229328, "epoch": 2576, "lr": 5.845334381303214e-06}
{"train_loss": 0.030815379694104195, "global_step": 229329, "epoch": 2576, "lr": 5.8450623733633635e-06}
{"train_loss": 0.03304310515522957, "global_step": 229330, "epoch": 2576, "lr": 5.844790371359593e-06}
{"train_loss": 0.05303207412362099, "global_step": 229331, "epoch": 2576, "lr": 5.844518375291963e-06}
{"train_loss": 0.042297814041376114, "global_step": 229332, "epoch": 2576, "lr": 5.844246385160496e-06}
{"train_loss": 0.029413266107439995, "global_step": 229333, "epoch": 2576, "lr": 5.843974400965235e-06}
{"train_loss": 0.08119521290063858, "global_step": 229334, "epoch": 2576, "lr": 5.8437024227062046e-06}
{"train_loss": 0.05242475867271423, "global_step": 229335, "epoch": 2576, "lr": 5.843430450383464e-06}
{"train_loss": 0.055503394454717636, "global_step": 229336, "epoch": 2576, "lr": 5.843158483997024e-06}
{"train_loss": 0.025621457025408745, "global_step": 229337, "epoch": 2576, "lr": 5.842886523546942e-06}
{"train_loss": 0.02584361657500267, "global_step": 229338, "epoch": 2576, "lr": 5.842614569033234e-06}
{"train_loss": 0.01645314134657383, "global_step": 229339, "epoch": 2576, "lr": 5.8423426204559536e-06}
{"train_loss": 0.06774239987134933, "global_step": 229340, "epoch": 2576, "lr": 5.842070677815137e-06}
{"train_loss": 0.08205195516347885, "global_step": 229341, "epoch": 2576, "lr": 5.84179874111081e-06}
{"train_loss": 0.045468103140592575, "global_step": 229342, "epoch": 2576, "lr": 5.841526810343023e-06}
{"train_loss": 0.021970758214592934, "global_step": 229343, "epoch": 2576, "lr": 5.841254885511793e-06}
{"train_loss": 0.03904756158590317, "global_step": 229344, "epoch": 2576, "lr": 5.840982966617175e-06}
{"train_loss": 0.03762497007846832, "global_step": 229345, "epoch": 2576, "lr": 5.840711053659198e-06}
{"train_loss": 0.09804992377758026, "global_step": 229346, "epoch": 2576, "lr": 5.840439146637899e-06}
{"train_loss": 0.039802875369787216, "global_step": 229347, "epoch": 2576, "lr": 5.840167245553324e-06}
{"train_loss": 0.02854175865650177, "global_step": 229348, "epoch": 2576, "lr": 5.839895350405483e-06}
{"train_loss": 0.07384520769119263, "global_step": 229349, "epoch": 2576, "lr": 5.8396234611944434e-06}
{"train_loss": 0.06095509976148605, "global_step": 229350, "epoch": 2576, "lr": 5.839351577920221e-06}
{"train_loss": 0.05777975171804428, "global_step": 229351, "epoch": 2576, "lr": 5.8390797005828664e-06}
{"train_loss": 0.053047432763020645, "global_step": 229352, "epoch": 2576, "lr": 5.838807829182397e-06, "val_loss": 9.316224098205566}
{"train_loss": 0.02844103053212166, "global_step": 229353, "epoch": 2577, "lr": 5.838535963718872e-06}
{"train_loss": 0.07294373959302902, "global_step": 229354, "epoch": 2577, "lr": 5.83826410419232e-06}
{"train_loss": 0.05750745162367821, "global_step": 229355, "epoch": 2577, "lr": 5.837992250602764e-06}
{"train_loss": 0.0660645067691803, "global_step": 229356, "epoch": 2577, "lr": 5.837720402950264e-06}
{"train_loss": 0.049082882702350616, "global_step": 229357, "epoch": 2577, "lr": 5.837448561234832e-06}
{"train_loss": 0.04401031881570816, "global_step": 229358, "epoch": 2577, "lr": 5.837176725456517e-06}
{"train_loss": 0.04937918856739998, "global_step": 229359, "epoch": 2577, "lr": 5.836904895615364e-06}
{"train_loss": 0.06455677002668381, "global_step": 229360, "epoch": 2577, "lr": 5.8366330717113945e-06}
{"train_loss": 0.04902934283018112, "global_step": 229361, "epoch": 2577, "lr": 5.83636125374466e-06}
{"train_loss": 0.03944157436490059, "global_step": 229362, "epoch": 2577, "lr": 5.836089441715175e-06}
{"train_loss": 0.08171555399894714, "global_step": 229363, "epoch": 2577, "lr": 5.8358176356230035e-06}
{"train_loss": 0.06881167739629745, "global_step": 229364, "epoch": 2577, "lr": 5.835545835468159e-06}
{"train_loss": 0.06531515717506409, "global_step": 229365, "epoch": 2577, "lr": 5.835274041250694e-06}
{"train_loss": 0.07564248889684677, "global_step": 229366, "epoch": 2577, "lr": 5.835002252970628e-06}
{"train_loss": 0.06518571823835373, "global_step": 229367, "epoch": 2577, "lr": 5.834730470628019e-06}
{"train_loss": 0.05533316731452942, "global_step": 229368, "epoch": 2577, "lr": 5.834458694222883e-06}
{"train_loss": 0.023410070687532425, "global_step": 229369, "epoch": 2577, "lr": 5.834186923755275e-06}
{"train_loss": 0.041838109493255615, "global_step": 229370, "epoch": 2577, "lr": 5.8339151592252114e-06}
{"train_loss": 0.05997326597571373, "global_step": 229371, "epoch": 2577, "lr": 5.833643400632749e-06}
{"train_loss": 0.012860775925219059, "global_step": 229372, "epoch": 2577, "lr": 5.833371647977903e-06}
{"train_loss": 0.05570247024297714, "global_step": 229373, "epoch": 2577, "lr": 5.833099901260736e-06}
{"train_loss": 0.07322397083044052, "global_step": 229374, "epoch": 2577, "lr": 5.832828160481257e-06}
{"train_loss": 0.06712263822555542, "global_step": 229375, "epoch": 2577, "lr": 5.832556425639529e-06}
{"train_loss": 0.060459185391664505, "global_step": 229376, "epoch": 2577, "lr": 5.832284696735568e-06}
{"train_loss": 0.047801073640584946, "global_step": 229377, "epoch": 2577, "lr": 5.832012973769413e-06}
{"train_loss": 0.03524439036846161, "global_step": 229378, "epoch": 2577, "lr": 5.831741256741119e-06}
{"train_loss": 0.04432402178645134, "global_step": 229379, "epoch": 2577, "lr": 5.8314695456506974e-06}
{"train_loss": 0.040587302297353745, "global_step": 229380, "epoch": 2577, "lr": 5.8311978404982036e-06}
{"train_loss": 0.061293136328458786, "global_step": 229381, "epoch": 2577, "lr": 5.8309261412836605e-06}
{"train_loss": 0.050363391637802124, "global_step": 229382, "epoch": 2577, "lr": 5.830654448007117e-06}
{"train_loss": 0.02745255082845688, "global_step": 229383, "epoch": 2577, "lr": 5.8303827606685956e-06}
{"train_loss": 0.06911606341600418, "global_step": 229384, "epoch": 2577, "lr": 5.830111079268152e-06}
{"train_loss": 0.05657917633652687, "global_step": 229385, "epoch": 2577, "lr": 5.829839403805803e-06}
{"train_loss": 0.08155319094657898, "global_step": 229386, "epoch": 2577, "lr": 5.829567734281605e-06}
{"train_loss": 0.055032189935445786, "global_step": 229387, "epoch": 2577, "lr": 5.829296070695578e-06}
{"train_loss": 0.07384596765041351, "global_step": 229388, "epoch": 2577, "lr": 5.8290244130477565e-06}
{"train_loss": 0.0690285786986351, "global_step": 229389, "epoch": 2577, "lr": 5.828752761338197e-06}
{"train_loss": 0.08671585470438004, "global_step": 229390, "epoch": 2577, "lr": 5.8284811155669085e-06}
{"train_loss": 0.03164947032928467, "global_step": 229391, "epoch": 2577, "lr": 5.828209475733953e-06}
{"train_loss": 0.0467892661690712, "global_step": 229392, "epoch": 2577, "lr": 5.827937841839359e-06}
{"train_loss": 0.07876191288232803, "global_step": 229393, "epoch": 2577, "lr": 5.827666213883148e-06}
{"train_loss": 0.051620181649923325, "global_step": 229394, "epoch": 2577, "lr": 5.82739459186537e-06}
{"train_loss": 0.07904024422168732, "global_step": 229395, "epoch": 2577, "lr": 5.82712297578607e-06}
{"train_loss": 0.07676999270915985, "global_step": 229396, "epoch": 2577, "lr": 5.8268513656452635e-06}
{"train_loss": 0.02323022112250328, "global_step": 229397, "epoch": 2577, "lr": 5.826579761443013e-06}
{"train_loss": 0.037602104246616364, "global_step": 229398, "epoch": 2577, "lr": 5.826308163179328e-06}
{"train_loss": 0.025734422728419304, "global_step": 229399, "epoch": 2577, "lr": 5.826036570854265e-06}
{"train_loss": 0.04826167970895767, "global_step": 229400, "epoch": 2577, "lr": 5.825764984467846e-06}
{"train_loss": 0.019164305180311203, "global_step": 229401, "epoch": 2577, "lr": 5.825493404020122e-06}
{"train_loss": 0.030397770926356316, "global_step": 229402, "epoch": 2577, "lr": 5.825221829511112e-06}
{"train_loss": 0.0527397058904171, "global_step": 229403, "epoch": 2577, "lr": 5.8249502609408754e-06}
{"train_loss": 0.031571466475725174, "global_step": 229404, "epoch": 2577, "lr": 5.824678698309427e-06}
{"train_loss": 0.07073451578617096, "global_step": 229405, "epoch": 2577, "lr": 5.824407141616817e-06}
{"train_loss": 0.041719116270542145, "global_step": 229406, "epoch": 2577, "lr": 5.824135590863073e-06}
{"train_loss": 0.06938309967517853, "global_step": 229407, "epoch": 2577, "lr": 5.823864046048238e-06}
{"train_loss": 0.07875669002532959, "global_step": 229408, "epoch": 2577, "lr": 5.823592507172343e-06}
{"train_loss": 0.021451938897371292, "global_step": 229409, "epoch": 2577, "lr": 5.823320974235435e-06}
{"train_loss": 0.02639044262468815, "global_step": 229410, "epoch": 2577, "lr": 5.823049447237533e-06}
{"train_loss": 0.03212662413716316, "global_step": 229411, "epoch": 2577, "lr": 5.822777926178691e-06}
{"train_loss": 0.0727958083152771, "global_step": 229412, "epoch": 2577, "lr": 5.822506411058931e-06}
{"train_loss": 0.0471159853041172, "global_step": 229413, "epoch": 2577, "lr": 5.822234901878299e-06}
{"train_loss": 0.08388611674308777, "global_step": 229414, "epoch": 2577, "lr": 5.8219633986368315e-06}
{"train_loss": 0.07013062387704849, "global_step": 229415, "epoch": 2577, "lr": 5.821691901334558e-06}
{"train_loss": 0.03680694103240967, "global_step": 229416, "epoch": 2577, "lr": 5.821420409971529e-06}
{"train_loss": 0.011737333610653877, "global_step": 229417, "epoch": 2577, "lr": 5.821148924547759e-06}
{"train_loss": 0.03493128716945648, "global_step": 229418, "epoch": 2577, "lr": 5.8208774450633105e-06}
{"train_loss": 0.07701914012432098, "global_step": 229419, "epoch": 2577, "lr": 5.820605971518195e-06}
{"train_loss": 0.037680283188819885, "global_step": 229420, "epoch": 2577, "lr": 5.8203345039124725e-06}
{"train_loss": 0.027493536472320557, "global_step": 229421, "epoch": 2577, "lr": 5.8200630422461545e-06}
{"train_loss": 0.07316854596138, "global_step": 229422, "epoch": 2577, "lr": 5.819791586519302e-06}
{"train_loss": 0.05634548142552376, "global_step": 229423, "epoch": 2577, "lr": 5.8195201367319315e-06}
{"train_loss": 0.04965453967452049, "global_step": 229424, "epoch": 2577, "lr": 5.819248692884099e-06}
{"train_loss": 0.050017692148685455, "global_step": 229425, "epoch": 2577, "lr": 5.818977254975827e-06}
{"train_loss": 0.055901121348142624, "global_step": 229426, "epoch": 2577, "lr": 5.8187058230071425e-06}
{"train_loss": 0.0345209576189518, "global_step": 229427, "epoch": 2577, "lr": 5.818434396978112e-06}
{"train_loss": 0.044559795409440994, "global_step": 229428, "epoch": 2577, "lr": 5.818162976888736e-06}
{"train_loss": 0.043275345116853714, "global_step": 229429, "epoch": 2577, "lr": 5.8178915627390866e-06}
{"train_loss": 0.05655086040496826, "global_step": 229430, "epoch": 2577, "lr": 5.817620154529169e-06}
{"train_loss": 0.04926275089383125, "global_step": 229431, "epoch": 2577, "lr": 5.817348752259044e-06}
{"train_loss": 0.046867139637470245, "global_step": 229432, "epoch": 2577, "lr": 5.817077355928729e-06}
{"train_loss": 0.07023239135742188, "global_step": 229433, "epoch": 2577, "lr": 5.81680596553828e-06}
{"train_loss": 0.04630172625184059, "global_step": 229434, "epoch": 2577, "lr": 5.816534581087713e-06}
{"train_loss": 0.10687900334596634, "global_step": 229435, "epoch": 2577, "lr": 5.8162632025770825e-06}
{"train_loss": 0.05251549184322357, "global_step": 229436, "epoch": 2577, "lr": 5.815991830006406e-06}
{"train_loss": 0.035628389567136765, "global_step": 229437, "epoch": 2577, "lr": 5.815720463375746e-06}
{"train_loss": 0.04456809535622597, "global_step": 229438, "epoch": 2577, "lr": 5.815449102685111e-06}
{"train_loss": 0.0430106483399868, "global_step": 229439, "epoch": 2577, "lr": 5.815177747934558e-06}
{"train_loss": 0.0292180348187685, "global_step": 229440, "epoch": 2577, "lr": 5.814906399124109e-06}
{"train_loss": 0.052188498323804206, "global_step": 229441, "epoch": 2577, "lr": 5.81463505625382e-06, "val_loss": 9.163763999938965}
{"train_loss": 0.06093774735927582, "global_step": 229442, "epoch": 2578, "lr": 5.814363719323701e-06}
{"train_loss": 0.10220973193645477, "global_step": 229443, "epoch": 2578, "lr": 5.814092388333808e-06}
{"train_loss": 0.05685538798570633, "global_step": 229444, "epoch": 2578, "lr": 5.813821063284169e-06}
{"train_loss": 0.0702345222234726, "global_step": 229445, "epoch": 2578, "lr": 5.813549744174829e-06}
{"train_loss": 0.03489579260349274, "global_step": 229446, "epoch": 2578, "lr": 5.813278431005808e-06}
{"train_loss": 0.06812868267297745, "global_step": 229447, "epoch": 2578, "lr": 5.81300712377717e-06}
{"train_loss": 0.04113847762346268, "global_step": 229448, "epoch": 2578, "lr": 5.812735822488918e-06}
{"train_loss": 0.06353580951690674, "global_step": 229449, "epoch": 2578, "lr": 5.812464527141103e-06}
{"train_loss": 0.018604740500450134, "global_step": 229450, "epoch": 2578, "lr": 5.812193237733782e-06}
{"train_loss": 0.05400240421295166, "global_step": 229451, "epoch": 2578, "lr": 5.811921954266958e-06}
{"train_loss": 0.03708040341734886, "global_step": 229452, "epoch": 2578, "lr": 5.811650676740693e-06}
{"train_loss": 0.0406043641269207, "global_step": 229453, "epoch": 2578, "lr": 5.811379405155004e-06}
{"train_loss": 0.05077818036079407, "global_step": 229454, "epoch": 2578, "lr": 5.811108139509941e-06}
{"train_loss": 0.03471381589770317, "global_step": 229455, "epoch": 2578, "lr": 5.810836879805531e-06}
{"train_loss": 0.06530693918466568, "global_step": 229456, "epoch": 2578, "lr": 5.810565626041825e-06}
{"train_loss": 0.05652143433690071, "global_step": 229457, "epoch": 2578, "lr": 5.810294378218839e-06}
{"train_loss": 0.0119605278596282, "global_step": 229458, "epoch": 2578, "lr": 5.810023136336634e-06}
{"train_loss": 0.018826697021722794, "global_step": 229459, "epoch": 2578, "lr": 5.8097519003952165e-06}
{"train_loss": 0.10629133135080338, "global_step": 229460, "epoch": 2578, "lr": 5.809480670394652e-06}
{"train_loss": 0.05453430116176605, "global_step": 229461, "epoch": 2578, "lr": 5.809209446334957e-06}
{"train_loss": 0.02333030290901661, "global_step": 229462, "epoch": 2578, "lr": 5.808938228216182e-06}
{"train_loss": 0.05077524110674858, "global_step": 229463, "epoch": 2578, "lr": 5.80866701603836e-06}
{"train_loss": 0.06310783326625824, "global_step": 229464, "epoch": 2578, "lr": 5.808395809801509e-06}
{"train_loss": 0.04559139534831047, "global_step": 229465, "epoch": 2578, "lr": 5.808124609505694e-06}
{"train_loss": 0.0712151750922203, "global_step": 229466, "epoch": 2578, "lr": 5.807853415150927e-06}
{"train_loss": 0.05982270836830139, "global_step": 229467, "epoch": 2578, "lr": 5.807582226737268e-06}
{"train_loss": 0.01685456372797489, "global_step": 229468, "epoch": 2578, "lr": 5.807311044264724e-06}
{"train_loss": 0.0509331077337265, "global_step": 229469, "epoch": 2578, "lr": 5.807039867733366e-06}
{"train_loss": 0.039070215076208115, "global_step": 229470, "epoch": 2578, "lr": 5.8067686971432e-06}
{"train_loss": 0.03337187319993973, "global_step": 229471, "epoch": 2578, "lr": 5.806497532494287e-06}
{"train_loss": 0.045816078782081604, "global_step": 229472, "epoch": 2578, "lr": 5.806226373786638e-06}
{"train_loss": 0.0413547083735466, "global_step": 229473, "epoch": 2578, "lr": 5.805955221020315e-06}
{"train_loss": 0.08729932457208633, "global_step": 229474, "epoch": 2578, "lr": 5.805684074195333e-06}
{"train_loss": 0.02484462596476078, "global_step": 229475, "epoch": 2578, "lr": 5.805412933311749e-06}
{"train_loss": 0.07330651581287384, "global_step": 229476, "epoch": 2578, "lr": 5.805141798369573e-06}
{"train_loss": 0.03167203813791275, "global_step": 229477, "epoch": 2578, "lr": 5.804870669368872e-06}
{"train_loss": 0.09107321500778198, "global_step": 229478, "epoch": 2578, "lr": 5.8045995463096584e-06}
{"train_loss": 0.04083831608295441, "global_step": 229479, "epoch": 2578, "lr": 5.804328429191985e-06}
{"train_loss": 0.02748667635023594, "global_step": 229480, "epoch": 2578, "lr": 5.804057318015876e-06}
{"train_loss": 0.031208546832203865, "global_step": 229481, "epoch": 2578, "lr": 5.803786212781376e-06}
{"train_loss": 0.03201710805296898, "global_step": 229482, "epoch": 2578, "lr": 5.8035151134885115e-06}
{"train_loss": 0.06233951076865196, "global_step": 229483, "epoch": 2578, "lr": 5.8032440201373386e-06}
{"train_loss": 0.013635321520268917, "global_step": 229484, "epoch": 2578, "lr": 5.802972932727868e-06}
{"train_loss": 0.022011233493685722, "global_step": 229485, "epoch": 2578, "lr": 5.802701851260145e-06}
{"train_loss": 0.04811485856771469, "global_step": 229486, "epoch": 2578, "lr": 5.802430775734224e-06}
{"train_loss": 0.0436570905148983, "global_step": 229487, "epoch": 2578, "lr": 5.802159706150117e-06}
{"train_loss": 0.04369283467531204, "global_step": 229488, "epoch": 2578, "lr": 5.801888642507886e-06}
{"train_loss": 0.07660706341266632, "global_step": 229489, "epoch": 2578, "lr": 5.801617584807534e-06}
{"train_loss": 0.04884978011250496, "global_step": 229490, "epoch": 2578, "lr": 5.801346533049129e-06}
{"train_loss": 0.04069358855485916, "global_step": 229491, "epoch": 2578, "lr": 5.8010754872326875e-06}
{"train_loss": 0.0746612548828125, "global_step": 229492, "epoch": 2578, "lr": 5.80080444735826e-06}
{"train_loss": 0.0708455741405487, "global_step": 229493, "epoch": 2578, "lr": 5.8005334134258625e-06}
{"train_loss": 0.05407562106847763, "global_step": 229494, "epoch": 2578, "lr": 5.800262385435562e-06}
{"train_loss": 0.1012687161564827, "global_step": 229495, "epoch": 2578, "lr": 5.799991363387358e-06}
{"train_loss": 0.035676851868629456, "global_step": 229496, "epoch": 2578, "lr": 5.799720347281323e-06}
{"train_loss": 0.039943329989910126, "global_step": 229497, "epoch": 2578, "lr": 5.799449337117469e-06}
{"train_loss": 0.04809403419494629, "global_step": 229498, "epoch": 2578, "lr": 5.799178332895844e-06}
{"train_loss": 0.03447187691926956, "global_step": 229499, "epoch": 2578, "lr": 5.798907334616472e-06}
{"train_loss": 0.01483524963259697, "global_step": 229500, "epoch": 2578, "lr": 5.798636342279412e-06}
{"train_loss": 0.05612460523843765, "global_step": 229501, "epoch": 2578, "lr": 5.798365355884672e-06}
{"train_loss": 0.056238021701574326, "global_step": 229502, "epoch": 2578, "lr": 5.798094375432317e-06}
{"train_loss": 0.035516560077667236, "global_step": 229503, "epoch": 2578, "lr": 5.79782340092237e-06}
{"train_loss": 0.032477427273988724, "global_step": 229504, "epoch": 2578, "lr": 5.797552432354852e-06}
{"train_loss": 0.03756234422326088, "global_step": 229505, "epoch": 2578, "lr": 5.7972814697298264e-06}
{"train_loss": 0.02382081001996994, "global_step": 229506, "epoch": 2578, "lr": 5.797010513047308e-06}
{"train_loss": 0.05260239541530609, "global_step": 229507, "epoch": 2578, "lr": 5.796739562307357e-06}
{"train_loss": 0.06189766526222229, "global_step": 229508, "epoch": 2578, "lr": 5.796468617509981e-06}
{"train_loss": 0.05863039195537567, "global_step": 229509, "epoch": 2578, "lr": 5.796197678655241e-06}
{"train_loss": 0.01905990019440651, "global_step": 229510, "epoch": 2578, "lr": 5.795926745743152e-06}
{"train_loss": 0.07238426804542542, "global_step": 229511, "epoch": 2578, "lr": 5.795655818773776e-06}
{"train_loss": 0.060238271951675415, "global_step": 229512, "epoch": 2578, "lr": 5.795384897747119e-06}
{"train_loss": 0.06122365593910217, "global_step": 229513, "epoch": 2578, "lr": 5.795113982663253e-06}
{"train_loss": 0.02213122323155403, "global_step": 229514, "epoch": 2578, "lr": 5.794843073522177e-06}
{"train_loss": 0.04532237350940704, "global_step": 229515, "epoch": 2578, "lr": 5.794572170323959e-06}
{"train_loss": 0.03933414816856384, "global_step": 229516, "epoch": 2578, "lr": 5.794301273068614e-06}
{"train_loss": 0.09494758397340775, "global_step": 229517, "epoch": 2578, "lr": 5.794030381756194e-06}
{"train_loss": 0.054760828614234924, "global_step": 229518, "epoch": 2578, "lr": 5.793759496386714e-06}
{"train_loss": 0.03290697559714317, "global_step": 229519, "epoch": 2578, "lr": 5.793488616960241e-06}
{"train_loss": 0.0525658093392849, "global_step": 229520, "epoch": 2578, "lr": 5.793217743476781e-06}
{"train_loss": 0.044500548392534256, "global_step": 229521, "epoch": 2578, "lr": 5.792946875936384e-06}
{"train_loss": 0.048618536442518234, "global_step": 229522, "epoch": 2578, "lr": 5.792676014339099e-06}
{"train_loss": 0.05504433810710907, "global_step": 229523, "epoch": 2578, "lr": 5.792405158684938e-06}
{"train_loss": 0.0270035769790411, "global_step": 229524, "epoch": 2578, "lr": 5.792134308973957e-06}
{"train_loss": 0.05967900529503822, "global_step": 229525, "epoch": 2578, "lr": 5.791863465206177e-06}
{"train_loss": 0.06180037558078766, "global_step": 229526, "epoch": 2578, "lr": 5.791592627381654e-06}
{"train_loss": 0.037740666419267654, "global_step": 229527, "epoch": 2578, "lr": 5.791321795500399e-06}
{"train_loss": 0.0381622277200222, "global_step": 229528, "epoch": 2578, "lr": 5.791050969562478e-06}
{"train_loss": 0.09433108568191528, "global_step": 229529, "epoch": 2578, "lr": 5.790780149567893e-06}
{"train_loss": 0.04900388695885626, "global_step": 229530, "epoch": 2578, "lr": 5.790509335516714e-06, "val_loss": 9.306843757629395}
{"train_loss": 0.03312387689948082, "global_step": 229531, "epoch": 2579, "lr": 5.790238527408953e-06}
{"train_loss": 0.03720434010028839, "global_step": 229532, "epoch": 2579, "lr": 5.7899677252446606e-06}
{"train_loss": 0.05477830767631531, "global_step": 229533, "epoch": 2579, "lr": 5.7896969290238635e-06}
{"train_loss": 0.05306847020983696, "global_step": 229534, "epoch": 2579, "lr": 5.789426138746612e-06}
{"train_loss": 0.04619424045085907, "global_step": 229535, "epoch": 2579, "lr": 5.789155354412923e-06}
{"train_loss": 0.04442949965596199, "global_step": 229536, "epoch": 2579, "lr": 5.788884576022852e-06}
{"train_loss": 0.06407910585403442, "global_step": 229537, "epoch": 2579, "lr": 5.7886138035764145e-06}
{"train_loss": 0.05023114010691643, "global_step": 229538, "epoch": 2579, "lr": 5.7883430370736735e-06}
{"train_loss": 0.02541344054043293, "global_step": 229539, "epoch": 2579, "lr": 5.788072276514639e-06}
{"train_loss": 0.040553018450737, "global_step": 229540, "epoch": 2579, "lr": 5.787801521899367e-06}
{"train_loss": 0.06689032912254333, "global_step": 229541, "epoch": 2579, "lr": 5.78753077322789e-06}
{"train_loss": 0.06670013815164566, "global_step": 229542, "epoch": 2579, "lr": 5.787260030500224e-06}
{"train_loss": 0.051172539591789246, "global_step": 229543, "epoch": 2579, "lr": 5.786989293716438e-06}
{"train_loss": 0.0642007440328598, "global_step": 229544, "epoch": 2579, "lr": 5.786718562876542e-06}
{"train_loss": 0.05964742973446846, "global_step": 229545, "epoch": 2579, "lr": 5.7864478379805905e-06}
{"train_loss": 0.06545969843864441, "global_step": 229546, "epoch": 2579, "lr": 5.786177119028607e-06}
{"train_loss": 0.036450792104005814, "global_step": 229547, "epoch": 2579, "lr": 5.785906406020641e-06}
{"train_loss": 0.04242097958922386, "global_step": 229548, "epoch": 2579, "lr": 5.785635698956709e-06}
{"train_loss": 0.04077617824077606, "global_step": 229549, "epoch": 2579, "lr": 5.785364997836867e-06}
{"train_loss": 0.03423810750246048, "global_step": 229550, "epoch": 2579, "lr": 5.785094302661137e-06}
{"train_loss": 0.042863331735134125, "global_step": 229551, "epoch": 2579, "lr": 5.7848236134295745e-06}
{"train_loss": 0.021723521873354912, "global_step": 229552, "epoch": 2579, "lr": 5.7845529301421895e-06}
{"train_loss": 0.06367332488298416, "global_step": 229553, "epoch": 2579, "lr": 5.7842822527990445e-06}
{"train_loss": 0.01997634395956993, "global_step": 229554, "epoch": 2579, "lr": 5.784011581400156e-06}
{"train_loss": 0.02672211453318596, "global_step": 229555, "epoch": 2579, "lr": 5.783740915945574e-06}
{"train_loss": 0.02662365511059761, "global_step": 229556, "epoch": 2579, "lr": 5.783470256435319e-06}
{"train_loss": 0.055033694952726364, "global_step": 229557, "epoch": 2579, "lr": 5.783199602869444e-06}
{"train_loss": 0.04190196841955185, "global_step": 229558, "epoch": 2579, "lr": 5.782928955247985e-06}
{"train_loss": 0.05247661843895912, "global_step": 229559, "epoch": 2579, "lr": 5.78265831357096e-06}
{"train_loss": 0.09043552726507187, "global_step": 229560, "epoch": 2579, "lr": 5.7823876778384305e-06}
{"train_loss": 0.028438355773687363, "global_step": 229561, "epoch": 2579, "lr": 5.782117048050406e-06}
{"train_loss": 0.028057456016540527, "global_step": 229562, "epoch": 2579, "lr": 5.781846424206949e-06}
{"train_loss": 0.04780524969100952, "global_step": 229563, "epoch": 2579, "lr": 5.7815758063080764e-06}
{"train_loss": 0.02358100563287735, "global_step": 229564, "epoch": 2579, "lr": 5.781305194353836e-06}
{"train_loss": 0.048875268548727036, "global_step": 229565, "epoch": 2579, "lr": 5.781034588344259e-06}
{"train_loss": 0.05810724198818207, "global_step": 229566, "epoch": 2579, "lr": 5.780763988279386e-06}
{"train_loss": 0.029367202892899513, "global_step": 229567, "epoch": 2579, "lr": 5.780493394159242e-06}
{"train_loss": 0.05719618871808052, "global_step": 229568, "epoch": 2579, "lr": 5.780222805983881e-06}
{"train_loss": 0.074845090508461, "global_step": 229569, "epoch": 2579, "lr": 5.779952223753321e-06}
{"train_loss": 0.0425039678812027, "global_step": 229570, "epoch": 2579, "lr": 5.779681647467622e-06}
{"train_loss": 0.027551298961043358, "global_step": 229571, "epoch": 2579, "lr": 5.779411077126789e-06}
{"train_loss": 0.014947056770324707, "global_step": 229572, "epoch": 2579, "lr": 5.779140512730885e-06}
{"train_loss": 0.027354005724191666, "global_step": 229573, "epoch": 2579, "lr": 5.778869954279931e-06}
{"train_loss": 0.04187976196408272, "global_step": 229574, "epoch": 2579, "lr": 5.778599401773976e-06}
{"train_loss": 0.03739088773727417, "global_step": 229575, "epoch": 2579, "lr": 5.7783288552130444e-06}
{"train_loss": 0.023730428889393806, "global_step": 229576, "epoch": 2579, "lr": 5.7780583145971845e-06}
{"train_loss": 0.0517871268093586, "global_step": 229577, "epoch": 2579, "lr": 5.777787779926413e-06}
{"train_loss": 0.06990499049425125, "global_step": 229578, "epoch": 2579, "lr": 5.777517251200793e-06}
{"train_loss": 0.054061658680438995, "global_step": 229579, "epoch": 2579, "lr": 5.777246728420332e-06}
{"train_loss": 0.05671851336956024, "global_step": 229580, "epoch": 2579, "lr": 5.776976211585095e-06}
{"train_loss": 0.048399418592453, "global_step": 229581, "epoch": 2579, "lr": 5.776705700695101e-06}
{"train_loss": 0.03473326936364174, "global_step": 229582, "epoch": 2579, "lr": 5.776435195750385e-06}
{"train_loss": 0.05003071576356888, "global_step": 229583, "epoch": 2579, "lr": 5.776164696750996e-06}
{"train_loss": 0.057155583053827286, "global_step": 229584, "epoch": 2579, "lr": 5.775894203696952e-06}
{"train_loss": 0.04750799387693405, "global_step": 229585, "epoch": 2579, "lr": 5.775623716588308e-06}
{"train_loss": 0.055644407868385315, "global_step": 229586, "epoch": 2579, "lr": 5.775353235425085e-06}
{"train_loss": 0.039131999015808105, "global_step": 229587, "epoch": 2579, "lr": 5.775082760207334e-06}
{"train_loss": 0.024434972554445267, "global_step": 229588, "epoch": 2579, "lr": 5.7748122909350774e-06}
{"train_loss": 0.044865965843200684, "global_step": 229589, "epoch": 2579, "lr": 5.774541827608371e-06}
{"train_loss": 0.03618938475847244, "global_step": 229590, "epoch": 2579, "lr": 5.774271370227224e-06}
{"train_loss": 0.10338844358921051, "global_step": 229591, "epoch": 2579, "lr": 5.774000918791694e-06}
{"train_loss": 0.028115084394812584, "global_step": 229592, "epoch": 2579, "lr": 5.773730473301808e-06}
{"train_loss": 0.08885639160871506, "global_step": 229593, "epoch": 2579, "lr": 5.7734600337576e-06}
{"train_loss": 0.06772801280021667, "global_step": 229594, "epoch": 2579, "lr": 5.773189600159123e-06}
{"train_loss": 0.06544119119644165, "global_step": 229595, "epoch": 2579, "lr": 5.772919172506392e-06}
{"train_loss": 0.021808667108416557, "global_step": 229596, "epoch": 2579, "lr": 5.772648750799459e-06}
{"train_loss": 0.04953378811478615, "global_step": 229597, "epoch": 2579, "lr": 5.772378335038348e-06}
{"train_loss": 0.04582054913043976, "global_step": 229598, "epoch": 2579, "lr": 5.772107925223114e-06}
{"train_loss": 0.060165874660015106, "global_step": 229599, "epoch": 2579, "lr": 5.771837521353768e-06}
{"train_loss": 0.08063110709190369, "global_step": 229600, "epoch": 2579, "lr": 5.771567123430366e-06}
{"train_loss": 0.033443838357925415, "global_step": 229601, "epoch": 2579, "lr": 5.771296731452935e-06}
{"train_loss": 0.030158838257193565, "global_step": 229602, "epoch": 2579, "lr": 5.7710263454215205e-06}
{"train_loss": 0.03658953681588173, "global_step": 229603, "epoch": 2579, "lr": 5.770755965336138e-06}
{"train_loss": 0.049499522894620895, "global_step": 229604, "epoch": 2579, "lr": 5.7704855911968555e-06}
{"train_loss": 0.059711650013923645, "global_step": 229605, "epoch": 2579, "lr": 5.770215223003678e-06}
{"train_loss": 0.04479340836405754, "global_step": 229606, "epoch": 2579, "lr": 5.769944860756666e-06}
{"train_loss": 0.02894669584929943, "global_step": 229607, "epoch": 2579, "lr": 5.769674504455836e-06}
{"train_loss": 0.06353040784597397, "global_step": 229608, "epoch": 2579, "lr": 5.769404154101249e-06}
{"train_loss": 0.04079108312726021, "global_step": 229609, "epoch": 2579, "lr": 5.769133809692912e-06}
{"train_loss": 0.06451048702001572, "global_step": 229610, "epoch": 2579, "lr": 5.768863471230884e-06}
{"train_loss": 0.05865432322025299, "global_step": 229611, "epoch": 2579, "lr": 5.768593138715189e-06}
{"train_loss": 0.10699483007192612, "global_step": 229612, "epoch": 2579, "lr": 5.7683228121458596e-06}
{"train_loss": 0.03604849427938461, "global_step": 229613, "epoch": 2579, "lr": 5.768052491522957e-06}
{"train_loss": 0.05196867883205414, "global_step": 229614, "epoch": 2579, "lr": 5.767782176846498e-06}
{"train_loss": 0.037324994802474976, "global_step": 229615, "epoch": 2579, "lr": 5.767511868116515e-06}
{"train_loss": 0.09630810469388962, "global_step": 229616, "epoch": 2579, "lr": 5.767241565333054e-06}
{"train_loss": 0.041658829897642136, "global_step": 229617, "epoch": 2579, "lr": 5.7669712684961466e-06}
{"train_loss": 0.11091692745685577, "global_step": 229618, "epoch": 2579, "lr": 5.7667009776058325e-06}
{"train_loss": 0.04886109953264842, "global_step": 229619, "epoch": 2579, "lr": 5.7664306926621506e-06, "val_loss": 9.217824935913086}
{"train_loss": 0.05156126618385315, "global_step": 229620, "epoch": 2580, "lr": 5.766160413665123e-06}
{"train_loss": 0.061928797513246536, "global_step": 229621, "epoch": 2580, "lr": 5.765890140614805e-06}
{"train_loss": 0.029477950185537338, "global_step": 229622, "epoch": 2580, "lr": 5.765619873511213e-06}
{"train_loss": 0.04611474648118019, "global_step": 229623, "epoch": 2580, "lr": 5.765349612354409e-06}
{"train_loss": 0.09249495714902878, "global_step": 229624, "epoch": 2580, "lr": 5.765079357144398e-06}
{"train_loss": 0.05123162642121315, "global_step": 229625, "epoch": 2580, "lr": 5.7648091078812475e-06}
{"train_loss": 0.03034258261322975, "global_step": 229626, "epoch": 2580, "lr": 5.764538864564967e-06}
{"train_loss": 0.028285641223192215, "global_step": 229627, "epoch": 2580, "lr": 5.764268627195618e-06}
{"train_loss": 0.03584088385105133, "global_step": 229628, "epoch": 2580, "lr": 5.763998395773213e-06}
{"train_loss": 0.02127693220973015, "global_step": 229629, "epoch": 2580, "lr": 5.763728170297805e-06}
{"train_loss": 0.10424327850341797, "global_step": 229630, "epoch": 2580, "lr": 5.7634579507694186e-06}
{"train_loss": 0.035601451992988586, "global_step": 229631, "epoch": 2580, "lr": 5.763187737188097e-06}
{"train_loss": 0.026528405025601387, "global_step": 229632, "epoch": 2580, "lr": 5.762917529553885e-06}
{"train_loss": 0.04301060736179352, "global_step": 229633, "epoch": 2580, "lr": 5.762647327866799e-06}
{"train_loss": 0.06558271497488022, "global_step": 229634, "epoch": 2580, "lr": 5.7623771321268946e-06}
{"train_loss": 0.023960480466485023, "global_step": 229635, "epoch": 2580, "lr": 5.762106942334189e-06}
{"train_loss": 0.0336599200963974, "global_step": 229636, "epoch": 2580, "lr": 5.7618367584887424e-06}
{"train_loss": 0.02791537344455719, "global_step": 229637, "epoch": 2580, "lr": 5.76156658059056e-06}
{"train_loss": 0.07404104620218277, "global_step": 229638, "epoch": 2580, "lr": 5.761296408639716e-06}
{"train_loss": 0.07112977653741837, "global_step": 229639, "epoch": 2580, "lr": 5.761026242636209e-06}
{"train_loss": 0.033334989100694656, "global_step": 229640, "epoch": 2580, "lr": 5.760756082580105e-06}
{"train_loss": 0.11592154949903488, "global_step": 229641, "epoch": 2580, "lr": 5.76048592847142e-06}
{"train_loss": 0.06627850234508514, "global_step": 229642, "epoch": 2580, "lr": 5.760215780310208e-06}
{"train_loss": 0.04975714534521103, "global_step": 229643, "epoch": 2580, "lr": 5.759945638096487e-06}
{"train_loss": 0.041850678622722626, "global_step": 229644, "epoch": 2580, "lr": 5.759675501830308e-06}
{"train_loss": 0.04236216098070145, "global_step": 229645, "epoch": 2580, "lr": 5.759405371511695e-06}
{"train_loss": 0.08461672067642212, "global_step": 229646, "epoch": 2580, "lr": 5.759135247140696e-06}
{"train_loss": 0.058447301387786865, "global_step": 229647, "epoch": 2580, "lr": 5.758865128717339e-06}
{"train_loss": 0.031221872195601463, "global_step": 229648, "epoch": 2580, "lr": 5.758595016241658e-06}
{"train_loss": 0.03806993365287781, "global_step": 229649, "epoch": 2580, "lr": 5.758324909713702e-06}
{"train_loss": 0.11947392672300339, "global_step": 229650, "epoch": 2580, "lr": 5.758054809133495e-06}
{"train_loss": 0.040951117873191833, "global_step": 229651, "epoch": 2580, "lr": 5.757784714501091e-06}
{"train_loss": 0.019958432763814926, "global_step": 229652, "epoch": 2580, "lr": 5.757514625816507e-06}
{"train_loss": 0.050827860832214355, "global_step": 229653, "epoch": 2580, "lr": 5.757244543079782e-06}
{"train_loss": 0.017642179504036903, "global_step": 229654, "epoch": 2580, "lr": 5.756974466290966e-06}
{"train_loss": 0.06396157294511795, "global_step": 229655, "epoch": 2580, "lr": 5.756704395450069e-06}
{"train_loss": 0.05974981188774109, "global_step": 229656, "epoch": 2580, "lr": 5.756434330557159e-06}
{"train_loss": 0.06548791378736496, "global_step": 229657, "epoch": 2580, "lr": 5.756164271612258e-06}
{"train_loss": 0.07818123698234558, "global_step": 229658, "epoch": 2580, "lr": 5.7558942186153874e-06}
{"train_loss": 0.03450334444642067, "global_step": 229659, "epoch": 2580, "lr": 5.755624171566609e-06}
{"train_loss": 0.013711212202906609, "global_step": 229660, "epoch": 2580, "lr": 5.7553541304659395e-06}
{"train_loss": 0.08548690378665924, "global_step": 229661, "epoch": 2580, "lr": 5.755084095313429e-06}
{"train_loss": 0.0795549675822258, "global_step": 229662, "epoch": 2580, "lr": 5.754814066109104e-06}
{"train_loss": 0.06839779764413834, "global_step": 229663, "epoch": 2580, "lr": 5.7545440428530105e-06}
{"train_loss": 0.031138740479946136, "global_step": 229664, "epoch": 2580, "lr": 5.75427402554517e-06}
{"train_loss": 0.06450910121202469, "global_step": 229665, "epoch": 2580, "lr": 5.754004014185638e-06}
{"train_loss": 0.04809080809354782, "global_step": 229666, "epoch": 2580, "lr": 5.7537340087744306e-06}
{"train_loss": 0.028918474912643433, "global_step": 229667, "epoch": 2580, "lr": 5.753464009311599e-06}
{"train_loss": 0.07477281987667084, "global_step": 229668, "epoch": 2580, "lr": 5.753194015797181e-06}
{"train_loss": 0.01559070497751236, "global_step": 229669, "epoch": 2580, "lr": 5.752924028231199e-06}
{"train_loss": 0.12179704755544662, "global_step": 229670, "epoch": 2580, "lr": 5.752654046613704e-06}
{"train_loss": 0.06697206199169159, "global_step": 229671, "epoch": 2580, "lr": 5.752384070944716e-06}
{"train_loss": 0.029602179303765297, "global_step": 229672, "epoch": 2580, "lr": 5.752114101224293e-06}
{"train_loss": 0.027078969404101372, "global_step": 229673, "epoch": 2580, "lr": 5.751844137452444e-06}
{"train_loss": 0.03349967673420906, "global_step": 229674, "epoch": 2580, "lr": 5.751574179629238e-06}
{"train_loss": 0.061719637364149094, "global_step": 229675, "epoch": 2580, "lr": 5.751304227754678e-06}
{"train_loss": 0.02869265154004097, "global_step": 229676, "epoch": 2580, "lr": 5.751034281828827e-06}
{"train_loss": 0.03633847087621689, "global_step": 229677, "epoch": 2580, "lr": 5.7507643418516995e-06}
{"train_loss": 0.04135515168309212, "global_step": 229678, "epoch": 2580, "lr": 5.750494407823353e-06}
{"train_loss": 0.04914001002907753, "global_step": 229679, "epoch": 2580, "lr": 5.750224479743804e-06}
{"train_loss": 0.052552271634340286, "global_step": 229680, "epoch": 2580, "lr": 5.749954557613107e-06}
{"train_loss": 0.08951792120933533, "global_step": 229681, "epoch": 2580, "lr": 5.74968464143128e-06}
{"train_loss": 0.08202029764652252, "global_step": 229682, "epoch": 2580, "lr": 5.749414731198377e-06}
{"train_loss": 0.04876161739230156, "global_step": 229683, "epoch": 2580, "lr": 5.749144826914416e-06}
{"train_loss": 0.06513869762420654, "global_step": 229684, "epoch": 2580, "lr": 5.748874928579445e-06}
{"train_loss": 0.05109236389398575, "global_step": 229685, "epoch": 2580, "lr": 5.748605036193505e-06}
{"train_loss": 0.04682360216975212, "global_step": 229686, "epoch": 2580, "lr": 5.748335149756617e-06}
{"train_loss": 0.07057791948318481, "global_step": 229687, "epoch": 2580, "lr": 5.7480652692688374e-06}
{"train_loss": 0.041179753839969635, "global_step": 229688, "epoch": 2580, "lr": 5.7477953947301825e-06}
{"train_loss": 0.05400051176548004, "global_step": 229689, "epoch": 2580, "lr": 5.747525526140707e-06}
{"train_loss": 0.055302344262599945, "global_step": 229690, "epoch": 2580, "lr": 5.747255663500434e-06}
{"train_loss": 0.05111753195524216, "global_step": 229691, "epoch": 2580, "lr": 5.7469858068093965e-06}
{"train_loss": 0.07143139839172363, "global_step": 229692, "epoch": 2580, "lr": 5.746715956067644e-06}
{"train_loss": 0.04865308105945587, "global_step": 229693, "epoch": 2580, "lr": 5.7464461112752e-06}
{"train_loss": 0.0676887184381485, "global_step": 229694, "epoch": 2580, "lr": 5.7461762724321135e-06}
{"train_loss": 0.04907091334462166, "global_step": 229695, "epoch": 2580, "lr": 5.745906439538407e-06}
{"train_loss": 0.07068052887916565, "global_step": 229696, "epoch": 2580, "lr": 5.74563661259413e-06}
{"train_loss": 0.06082523986697197, "global_step": 229697, "epoch": 2580, "lr": 5.745366791599316e-06}
{"train_loss": 0.022903064265847206, "global_step": 229698, "epoch": 2580, "lr": 5.745096976553988e-06}
{"train_loss": 0.030900798738002777, "global_step": 229699, "epoch": 2580, "lr": 5.7448271674581995e-06}
{"train_loss": 0.0233777966350317, "global_step": 229700, "epoch": 2580, "lr": 5.744557364311976e-06}
{"train_loss": 0.08826690167188644, "global_step": 229701, "epoch": 2580, "lr": 5.744287567115364e-06}
{"train_loss": 0.04566530883312225, "global_step": 229702, "epoch": 2580, "lr": 5.744017775868383e-06}
{"train_loss": 0.04639625921845436, "global_step": 229703, "epoch": 2580, "lr": 5.743747990571074e-06}
{"train_loss": 0.05169457197189331, "global_step": 229704, "epoch": 2580, "lr": 5.743478211223496e-06}
{"train_loss": 0.03736190125346184, "global_step": 229705, "epoch": 2580, "lr": 5.743208437825653e-06}
{"train_loss": 0.04467817023396492, "global_step": 229706, "epoch": 2580, "lr": 5.742938670377612e-06}
{"train_loss": 0.047572195529937744, "global_step": 229707, "epoch": 2580, "lr": 5.742668908879378e-06}
{"train_loss": 0.05233609799839807, "global_step": 229708, "epoch": 2580, "lr": 5.742399153331013e-06, "val_loss": 9.270480155944824, "train_action_mse_error": 20.592844009399414}
{"train_loss": 0.02726549468934536, "global_step": 229709, "epoch": 2581, "lr": 5.742129403732538e-06}
{"train_loss": 0.10408572107553482, "global_step": 229710, "epoch": 2581, "lr": 5.741859660083998e-06}
{"train_loss": 0.059014029800891876, "global_step": 229711, "epoch": 2581, "lr": 5.741589922385421e-06}
{"train_loss": 0.06827569752931595, "global_step": 229712, "epoch": 2581, "lr": 5.741320190636856e-06}
{"train_loss": 0.12365277111530304, "global_step": 229713, "epoch": 2581, "lr": 5.741050464838321e-06}
{"train_loss": 0.03434649854898453, "global_step": 229714, "epoch": 2581, "lr": 5.740780744989871e-06}
{"train_loss": 0.0348295196890831, "global_step": 229715, "epoch": 2581, "lr": 5.740511031091522e-06}
{"train_loss": 0.02890382707118988, "global_step": 229716, "epoch": 2581, "lr": 5.740241323143336e-06}
{"train_loss": 0.041052330285310745, "global_step": 229717, "epoch": 2581, "lr": 5.739971621145323e-06}
{"train_loss": 0.011885528452694416, "global_step": 229718, "epoch": 2581, "lr": 5.7397019250975405e-06}
{"train_loss": 0.042348314076662064, "global_step": 229719, "epoch": 2581, "lr": 5.739432235000008e-06}
{"train_loss": 0.0278897937387228, "global_step": 229720, "epoch": 2581, "lr": 5.739162550852767e-06}
{"train_loss": 0.04656762629747391, "global_step": 229721, "epoch": 2581, "lr": 5.738892872655866e-06}
{"train_loss": 0.05079561471939087, "global_step": 229722, "epoch": 2581, "lr": 5.738623200409327e-06}
{"train_loss": 0.06295398622751236, "global_step": 229723, "epoch": 2581, "lr": 5.738353534113194e-06}
{"train_loss": 0.050962187349796295, "global_step": 229724, "epoch": 2581, "lr": 5.738083873767491e-06}
{"train_loss": 0.03614327684044838, "global_step": 229725, "epoch": 2581, "lr": 5.737814219372273e-06}
{"train_loss": 0.04711553826928139, "global_step": 229726, "epoch": 2581, "lr": 5.7375445709275614e-06}
{"train_loss": 0.04749571532011032, "global_step": 229727, "epoch": 2581, "lr": 5.7372749284334e-06}
{"train_loss": 0.0241739172488451, "global_step": 229728, "epoch": 2581, "lr": 5.737005291889818e-06}
{"train_loss": 0.05099525675177574, "global_step": 229729, "epoch": 2581, "lr": 5.736735661296866e-06}
{"train_loss": 0.09846050292253494, "global_step": 229730, "epoch": 2581, "lr": 5.736466036654569e-06}
{"train_loss": 0.023163672536611557, "global_step": 229731, "epoch": 2581, "lr": 5.736196417962952e-06}
{"train_loss": 0.03327164053916931, "global_step": 229732, "epoch": 2581, "lr": 5.73592680522208e-06}
{"train_loss": 0.06504852324724197, "global_step": 229733, "epoch": 2581, "lr": 5.73565719843196e-06}
{"train_loss": 0.08838310092687607, "global_step": 229734, "epoch": 2581, "lr": 5.735387597592651e-06}
{"train_loss": 0.021586032584309578, "global_step": 229735, "epoch": 2581, "lr": 5.735118002704176e-06}
{"train_loss": 0.060303978621959686, "global_step": 229736, "epoch": 2581, "lr": 5.734848413766569e-06}
{"train_loss": 0.10329524427652359, "global_step": 229737, "epoch": 2581, "lr": 5.73457883077988e-06}
{"train_loss": 0.03959093242883682, "global_step": 229738, "epoch": 2581, "lr": 5.73430925374413e-06}
{"train_loss": 0.03898085281252861, "global_step": 229739, "epoch": 2581, "lr": 5.7340396826593646e-06}
{"train_loss": 0.08347228914499283, "global_step": 229740, "epoch": 2581, "lr": 5.733770117525622e-06}
{"train_loss": 0.08263909071683884, "global_step": 229741, "epoch": 2581, "lr": 5.7335005583429305e-06}
{"train_loss": 0.07115116715431213, "global_step": 229742, "epoch": 2581, "lr": 5.733231005111333e-06}
{"train_loss": 0.05114613100886345, "global_step": 229743, "epoch": 2581, "lr": 5.73296145783086e-06}
{"train_loss": 0.040057774633169174, "global_step": 229744, "epoch": 2581, "lr": 5.732691916501559e-06}
{"train_loss": 0.01912698522210121, "global_step": 229745, "epoch": 2581, "lr": 5.7324223811234425e-06}
{"train_loss": 0.029461894184350967, "global_step": 229746, "epoch": 2581, "lr": 5.732152851696581e-06}
{"train_loss": 0.07652635127305984, "global_step": 229747, "epoch": 2581, "lr": 5.731883328220977e-06}
{"train_loss": 0.04583696275949478, "global_step": 229748, "epoch": 2581, "lr": 5.73161381069669e-06}
{"train_loss": 0.04752872884273529, "global_step": 229749, "epoch": 2581, "lr": 5.731344299123742e-06}
{"train_loss": 0.04272657632827759, "global_step": 229750, "epoch": 2581, "lr": 5.731074793502183e-06}
{"train_loss": 0.01987014152109623, "global_step": 229751, "epoch": 2581, "lr": 5.730805293832031e-06}
{"train_loss": 0.042636554688215256, "global_step": 229752, "epoch": 2581, "lr": 5.730535800113346e-06}
{"train_loss": 0.05106651037931442, "global_step": 229753, "epoch": 2581, "lr": 5.730266312346139e-06}
{"train_loss": 0.029828932136297226, "global_step": 229754, "epoch": 2581, "lr": 5.729996830530465e-06}
{"train_loss": 0.04778728261590004, "global_step": 229755, "epoch": 2581, "lr": 5.7297273546663476e-06}
{"train_loss": 0.06213381141424179, "global_step": 229756, "epoch": 2581, "lr": 5.729457884753842e-06}
{"train_loss": 0.049806147813797, "global_step": 229757, "epoch": 2581, "lr": 5.7291884207929525e-06}
{"train_loss": 0.062020186334848404, "global_step": 229758, "epoch": 2581, "lr": 5.728918962783741e-06}
{"train_loss": 0.05461170896887779, "global_step": 229759, "epoch": 2581, "lr": 5.728649510726242e-06}
{"train_loss": 0.025293318554759026, "global_step": 229760, "epoch": 2581, "lr": 5.728380064620475e-06}
{"train_loss": 0.07663221657276154, "global_step": 229761, "epoch": 2581, "lr": 5.728110624466504e-06}
{"train_loss": 0.028244569897651672, "global_step": 229762, "epoch": 2581, "lr": 5.727841190264338e-06}
{"train_loss": 0.0572839081287384, "global_step": 229763, "epoch": 2581, "lr": 5.7275717620140276e-06}
{"train_loss": 0.03797208517789841, "global_step": 229764, "epoch": 2581, "lr": 5.727302339715601e-06}
{"train_loss": 0.06281821429729462, "global_step": 229765, "epoch": 2581, "lr": 5.727032923369108e-06}
{"train_loss": 0.018807299435138702, "global_step": 229766, "epoch": 2581, "lr": 5.7267635129745704e-06}
{"train_loss": 0.035309433937072754, "global_step": 229767, "epoch": 2581, "lr": 5.726494108532032e-06}
{"train_loss": 0.0783366858959198, "global_step": 229768, "epoch": 2581, "lr": 5.726224710041533e-06}
{"train_loss": 0.09278780966997147, "global_step": 229769, "epoch": 2581, "lr": 5.72595531750309e-06}
{"train_loss": 0.030542397871613503, "global_step": 229770, "epoch": 2581, "lr": 5.725685930916763e-06}
{"train_loss": 0.04725794494152069, "global_step": 229771, "epoch": 2581, "lr": 5.72541655028257e-06}
{"train_loss": 0.015973972156643867, "global_step": 229772, "epoch": 2581, "lr": 5.725147175600559e-06}
{"train_loss": 0.05783756822347641, "global_step": 229773, "epoch": 2581, "lr": 5.724877806870771e-06}
{"train_loss": 0.025073902681469917, "global_step": 229774, "epoch": 2581, "lr": 5.724608444093216e-06}
{"train_loss": 0.014127112925052643, "global_step": 229775, "epoch": 2581, "lr": 5.724339087267949e-06}
{"train_loss": 0.050592172890901566, "global_step": 229776, "epoch": 2581, "lr": 5.724069736395021e-06}
{"train_loss": 0.03913377225399017, "global_step": 229777, "epoch": 2581, "lr": 5.723800391474438e-06}
{"train_loss": 0.02053161710500717, "global_step": 229778, "epoch": 2581, "lr": 5.723531052506259e-06}
{"train_loss": 0.03976722061634064, "global_step": 229779, "epoch": 2581, "lr": 5.723261719490503e-06}
{"train_loss": 0.02836519293487072, "global_step": 229780, "epoch": 2581, "lr": 5.7229923924272235e-06}
{"train_loss": 0.07271084189414978, "global_step": 229781, "epoch": 2581, "lr": 5.722723071316444e-06}
{"train_loss": 0.012750418856739998, "global_step": 229782, "epoch": 2581, "lr": 5.722453756158208e-06}
{"train_loss": 0.028424685820937157, "global_step": 229783, "epoch": 2581, "lr": 5.722184446952539e-06}
{"train_loss": 0.034967340528964996, "global_step": 229784, "epoch": 2581, "lr": 5.7219151436994965e-06}
{"train_loss": 0.0378032922744751, "global_step": 229785, "epoch": 2581, "lr": 5.721645846399087e-06}
{"train_loss": 0.04517974704504013, "global_step": 229786, "epoch": 2581, "lr": 5.721376555051377e-06}
{"train_loss": 0.023673592135310173, "global_step": 229787, "epoch": 2581, "lr": 5.721107269656379e-06}
{"train_loss": 0.04943723976612091, "global_step": 229788, "epoch": 2581, "lr": 5.7208379902141455e-06}
{"train_loss": 0.04468823969364166, "global_step": 229789, "epoch": 2581, "lr": 5.7205687167247006e-06}
{"train_loss": 0.07637804746627808, "global_step": 229790, "epoch": 2581, "lr": 5.720299449188088e-06}
{"train_loss": 0.04143449664115906, "global_step": 229791, "epoch": 2581, "lr": 5.720030187604336e-06}
{"train_loss": 0.07484088838100433, "global_step": 229792, "epoch": 2581, "lr": 5.719760931973495e-06}
{"train_loss": 0.05901775509119034, "global_step": 229793, "epoch": 2581, "lr": 5.719491682295586e-06}
{"train_loss": 0.03905993327498436, "global_step": 229794, "epoch": 2581, "lr": 5.719222438570648e-06}
{"train_loss": 0.05439358204603195, "global_step": 229795, "epoch": 2581, "lr": 5.718953200798732e-06}
{"train_loss": 0.037354081869125366, "global_step": 229796, "epoch": 2581, "lr": 5.718683968979849e-06}
{"train_loss": 0.047705219294666576, "global_step": 229797, "epoch": 2581, "lr": 5.718414743114065e-06, "val_loss": 9.290639877319336}
{"train_loss": 0.05312202125787735, "global_step": 229798, "epoch": 2582, "lr": 5.718145523201385e-06}
{"train_loss": 0.028595631942152977, "global_step": 229799, "epoch": 2582, "lr": 5.7178763092418765e-06}
{"train_loss": 0.03642004355788231, "global_step": 229800, "epoch": 2582, "lr": 5.717607101235545e-06}
{"train_loss": 0.016021154820919037, "global_step": 229801, "epoch": 2582, "lr": 5.717337899182457e-06}
{"train_loss": 0.026061641052365303, "global_step": 229802, "epoch": 2582, "lr": 5.717068703082617e-06}
{"train_loss": 0.030195964500308037, "global_step": 229803, "epoch": 2582, "lr": 5.716799512936088e-06}
{"train_loss": 0.04360508173704147, "global_step": 229804, "epoch": 2582, "lr": 5.716530328742892e-06}
{"train_loss": 0.029101811349391937, "global_step": 229805, "epoch": 2582, "lr": 5.716261150503072e-06}
{"train_loss": 0.02579285204410553, "global_step": 229806, "epoch": 2582, "lr": 5.715991978216667e-06}
{"train_loss": 0.03495161607861519, "global_step": 229807, "epoch": 2582, "lr": 5.715722811883689e-06}
{"train_loss": 0.03414328768849373, "global_step": 229808, "epoch": 2582, "lr": 5.71545365150421e-06}
{"train_loss": 0.026215318590402603, "global_step": 229809, "epoch": 2582, "lr": 5.715184497078235e-06}
{"train_loss": 0.03194465488195419, "global_step": 229810, "epoch": 2582, "lr": 5.71491534860582e-06}
{"train_loss": 0.05666794627904892, "global_step": 229811, "epoch": 2582, "lr": 5.714646206086993e-06}
{"train_loss": 0.05559473857283592, "global_step": 229812, "epoch": 2582, "lr": 5.714377069521798e-06}
{"train_loss": 0.04533519968390465, "global_step": 229813, "epoch": 2582, "lr": 5.714107938910251e-06}
{"train_loss": 0.06169946864247322, "global_step": 229814, "epoch": 2582, "lr": 5.71383881425242e-06}
{"train_loss": 0.042035095393657684, "global_step": 229815, "epoch": 2582, "lr": 5.7135696955483095e-06}
{"train_loss": 0.03543328866362572, "global_step": 229816, "epoch": 2582, "lr": 5.713300582797981e-06}
{"train_loss": 0.01583482138812542, "global_step": 229817, "epoch": 2582, "lr": 5.713031476001451e-06}
{"train_loss": 0.02048310823738575, "global_step": 229818, "epoch": 2582, "lr": 5.71276237515877e-06}
{"train_loss": 0.05798172950744629, "global_step": 229819, "epoch": 2582, "lr": 5.71249328026996e-06}
{"train_loss": 0.030805880203843117, "global_step": 229820, "epoch": 2582, "lr": 5.712224191335075e-06}
{"train_loss": 0.047592759132385254, "global_step": 229821, "epoch": 2582, "lr": 5.711955108354133e-06}
{"train_loss": 0.01812070980668068, "global_step": 229822, "epoch": 2582, "lr": 5.71168603132719e-06}
{"train_loss": 0.06548481434583664, "global_step": 229823, "epoch": 2582, "lr": 5.711416960254257e-06}
{"train_loss": 0.03338169306516647, "global_step": 229824, "epoch": 2582, "lr": 5.711147895135399e-06}
{"train_loss": 0.02721303328871727, "global_step": 229825, "epoch": 2582, "lr": 5.710878835970624e-06}
{"train_loss": 0.032264593988657, "global_step": 229826, "epoch": 2582, "lr": 5.71060978275999e-06}
{"train_loss": 0.05657380446791649, "global_step": 229827, "epoch": 2582, "lr": 5.710340735503522e-06}
{"train_loss": 0.06087145954370499, "global_step": 229828, "epoch": 2582, "lr": 5.710071694201263e-06}
{"train_loss": 0.06877059489488602, "global_step": 229829, "epoch": 2582, "lr": 5.709802658853236e-06}
{"train_loss": 0.033453263342380524, "global_step": 229830, "epoch": 2582, "lr": 5.709533629459485e-06}
{"train_loss": 0.07456962764263153, "global_step": 229831, "epoch": 2582, "lr": 5.70926460602006e-06}
{"train_loss": 0.07444130629301071, "global_step": 229832, "epoch": 2582, "lr": 5.708995588534971e-06}
{"train_loss": 0.0570535846054554, "global_step": 229833, "epoch": 2582, "lr": 5.708726577004281e-06}
{"train_loss": 0.04914253205060959, "global_step": 229834, "epoch": 2582, "lr": 5.708457571428e-06}
{"train_loss": 0.028316009789705276, "global_step": 229835, "epoch": 2582, "lr": 5.708188571806189e-06}
{"train_loss": 0.02790960669517517, "global_step": 229836, "epoch": 2582, "lr": 5.707919578138865e-06}
{"train_loss": 0.046579375863075256, "global_step": 229837, "epoch": 2582, "lr": 5.707650590426078e-06}
{"train_loss": 0.04585005342960358, "global_step": 229838, "epoch": 2582, "lr": 5.707381608667845e-06}
{"train_loss": 0.058876268565654755, "global_step": 229839, "epoch": 2582, "lr": 5.707112632864231e-06}
{"train_loss": 0.05740775540471077, "global_step": 229840, "epoch": 2582, "lr": 5.706843663015243e-06}
{"train_loss": 0.061379075050354004, "global_step": 229841, "epoch": 2582, "lr": 5.706574699120937e-06}
{"train_loss": 0.023924611508846283, "global_step": 229842, "epoch": 2582, "lr": 5.706305741181334e-06}
{"train_loss": 0.04114067927002907, "global_step": 229843, "epoch": 2582, "lr": 5.706036789196489e-06}
{"train_loss": 0.04182582348585129, "global_step": 229844, "epoch": 2582, "lr": 5.705767843166421e-06}
{"train_loss": 0.06453545391559601, "global_step": 229845, "epoch": 2582, "lr": 5.705498903091178e-06}
{"train_loss": 0.040470242500305176, "global_step": 229846, "epoch": 2582, "lr": 5.705229968970793e-06}
{"train_loss": 0.023017914965748787, "global_step": 229847, "epoch": 2582, "lr": 5.704961040805284e-06}
{"train_loss": 0.046174559742212296, "global_step": 229848, "epoch": 2582, "lr": 5.704692118594723e-06}
{"train_loss": 0.07332435250282288, "global_step": 229849, "epoch": 2582, "lr": 5.70442320233911e-06}
{"train_loss": 0.023304134607315063, "global_step": 229850, "epoch": 2582, "lr": 5.7041542920385105e-06}
{"train_loss": 0.03865557163953781, "global_step": 229851, "epoch": 2582, "lr": 5.703885387692931e-06}
{"train_loss": 0.07139834016561508, "global_step": 229852, "epoch": 2582, "lr": 5.703616489302443e-06}
{"train_loss": 0.03584788367152214, "global_step": 229853, "epoch": 2582, "lr": 5.703347596867048e-06}
{"train_loss": 0.013032982125878334, "global_step": 229854, "epoch": 2582, "lr": 5.703078710386811e-06}
{"train_loss": 0.12296536564826965, "global_step": 229855, "epoch": 2582, "lr": 5.702809829861744e-06}
{"train_loss": 0.041830211877822876, "global_step": 229856, "epoch": 2582, "lr": 5.702540955291902e-06}
{"train_loss": 0.04653496295213699, "global_step": 229857, "epoch": 2582, "lr": 5.702272086677307e-06}
{"train_loss": 0.03346075117588043, "global_step": 229858, "epoch": 2582, "lr": 5.70200322401801e-06}
{"train_loss": 0.11270163953304291, "global_step": 229859, "epoch": 2582, "lr": 5.701734367314032e-06}
{"train_loss": 0.049914710223674774, "global_step": 229860, "epoch": 2582, "lr": 5.7014655165654184e-06}
{"train_loss": 0.02287900261580944, "global_step": 229861, "epoch": 2582, "lr": 5.701196671772202e-06}
{"train_loss": 0.024496635422110558, "global_step": 229862, "epoch": 2582, "lr": 5.700927832934421e-06}
{"train_loss": 0.034940414130687714, "global_step": 229863, "epoch": 2582, "lr": 5.700659000052105e-06}
{"train_loss": 0.02673734724521637, "global_step": 229864, "epoch": 2582, "lr": 5.700390173125309e-06}
{"train_loss": 0.09718163311481476, "global_step": 229865, "epoch": 2582, "lr": 5.7001213521540366e-06}
{"train_loss": 0.06297743320465088, "global_step": 229866, "epoch": 2582, "lr": 5.6998525371383505e-06}
{"train_loss": 0.05202409625053406, "global_step": 229867, "epoch": 2582, "lr": 5.6995837280782894e-06}
{"train_loss": 0.02904057502746582, "global_step": 229868, "epoch": 2582, "lr": 5.699314924973864e-06}
{"train_loss": 0.07549536973237991, "global_step": 229869, "epoch": 2582, "lr": 5.6990461278251364e-06}
{"train_loss": 0.019079208374023438, "global_step": 229870, "epoch": 2582, "lr": 5.6987773366321275e-06}
{"train_loss": 0.045595698058605194, "global_step": 229871, "epoch": 2582, "lr": 5.6985085513948825e-06}
{"train_loss": 0.05160875618457794, "global_step": 229872, "epoch": 2582, "lr": 5.6982397721134285e-06}
{"train_loss": 0.09701263159513474, "global_step": 229873, "epoch": 2582, "lr": 5.6979709987878105e-06}
{"train_loss": 0.04459988698363304, "global_step": 229874, "epoch": 2582, "lr": 5.69770223141805e-06}
{"train_loss": 0.06971058249473572, "global_step": 229875, "epoch": 2582, "lr": 5.69743347000421e-06}
{"train_loss": 0.03551343083381653, "global_step": 229876, "epoch": 2582, "lr": 5.697164714546294e-06}
{"train_loss": 0.06812633574008942, "global_step": 229877, "epoch": 2582, "lr": 5.6968959650443686e-06}
{"train_loss": 0.05732566490769386, "global_step": 229878, "epoch": 2582, "lr": 5.696627221498441e-06}
{"train_loss": 0.05634145438671112, "global_step": 229879, "epoch": 2582, "lr": 5.696358483908576e-06}
{"train_loss": 0.044360868632793427, "global_step": 229880, "epoch": 2582, "lr": 5.696089752274786e-06}
{"train_loss": 0.06199066340923309, "global_step": 229881, "epoch": 2582, "lr": 5.695821026597126e-06}
{"train_loss": 0.056634627282619476, "global_step": 229882, "epoch": 2582, "lr": 5.6955523068756065e-06}
{"train_loss": 0.10412374883890152, "global_step": 229883, "epoch": 2582, "lr": 5.695283593110295e-06}
{"train_loss": 0.0871318057179451, "global_step": 229884, "epoch": 2582, "lr": 5.6950148853012144e-06}
{"train_loss": 0.034933265298604965, "global_step": 229885, "epoch": 2582, "lr": 5.694746183448391e-06}
{"train_loss": 0.04735058407937543, "global_step": 229886, "epoch": 2582, "lr": 5.694477487551875e-06, "val_loss": 9.240694046020508}
{"train_loss": 0.07584796100854874, "global_step": 229887, "epoch": 2583, "lr": 5.694208797611683e-06}
{"train_loss": 0.05765447020530701, "global_step": 229888, "epoch": 2583, "lr": 5.693940113627883e-06}
{"train_loss": 0.0663985162973404, "global_step": 229889, "epoch": 2583, "lr": 5.693671435600473e-06}
{"train_loss": 0.03437857702374458, "global_step": 229890, "epoch": 2583, "lr": 5.693402763529526e-06}
{"train_loss": 0.03121749684214592, "global_step": 229891, "epoch": 2583, "lr": 5.693134097415048e-06}
{"train_loss": 0.062125224620103836, "global_step": 229892, "epoch": 2583, "lr": 5.6928654372571e-06}
{"train_loss": 0.03761601820588112, "global_step": 229893, "epoch": 2583, "lr": 5.692596783055698e-06}
{"train_loss": 0.11936958134174347, "global_step": 229894, "epoch": 2583, "lr": 5.692328134810893e-06}
{"train_loss": 0.049069564789533615, "global_step": 229895, "epoch": 2583, "lr": 5.692059492522706e-06}
{"train_loss": 0.02415342628955841, "global_step": 229896, "epoch": 2583, "lr": 5.691790856191187e-06}
{"train_loss": 0.04889962822198868, "global_step": 229897, "epoch": 2583, "lr": 5.6915222258163594e-06}
{"train_loss": 0.02094910852611065, "global_step": 229898, "epoch": 2583, "lr": 5.691253601398277e-06}
{"train_loss": 0.02675165981054306, "global_step": 229899, "epoch": 2583, "lr": 5.690984982936959e-06}
{"train_loss": 0.06117222085595131, "global_step": 229900, "epoch": 2583, "lr": 5.690716370432453e-06}
{"train_loss": 0.060263827443122864, "global_step": 229901, "epoch": 2583, "lr": 5.690447763884782e-06}
{"train_loss": 0.06904982775449753, "global_step": 229902, "epoch": 2583, "lr": 5.690179163293991e-06}
{"train_loss": 0.05065017566084862, "global_step": 229903, "epoch": 2583, "lr": 5.689910568660128e-06}
{"train_loss": 0.05741974711418152, "global_step": 229904, "epoch": 2583, "lr": 5.689641979983202e-06}
{"train_loss": 0.04702327027916908, "global_step": 229905, "epoch": 2583, "lr": 5.689373397263276e-06}
{"train_loss": 0.05979032441973686, "global_step": 229906, "epoch": 2583, "lr": 5.689104820500363e-06}
{"train_loss": 0.07681754976511002, "global_step": 229907, "epoch": 2583, "lr": 5.688836249694518e-06}
{"train_loss": 0.01984390802681446, "global_step": 229908, "epoch": 2583, "lr": 5.688567684845758e-06}
{"train_loss": 0.06546946614980698, "global_step": 229909, "epoch": 2583, "lr": 5.688299125954144e-06}
{"train_loss": 0.04278232157230377, "global_step": 229910, "epoch": 2583, "lr": 5.688030573019687e-06}
{"train_loss": 0.05597260966897011, "global_step": 229911, "epoch": 2583, "lr": 5.687762026042442e-06}
{"train_loss": 0.06460665911436081, "global_step": 229912, "epoch": 2583, "lr": 5.687493485022427e-06}
{"train_loss": 0.025340713560581207, "global_step": 229913, "epoch": 2583, "lr": 5.687224949959702e-06}
{"train_loss": 0.04386346787214279, "global_step": 229914, "epoch": 2583, "lr": 5.686956420854278e-06}
{"train_loss": 0.0542258583009243, "global_step": 229915, "epoch": 2583, "lr": 5.686687897706211e-06}
{"train_loss": 0.049515288323163986, "global_step": 229916, "epoch": 2583, "lr": 5.686419380515523e-06}
{"train_loss": 0.06983580440282822, "global_step": 229917, "epoch": 2583, "lr": 5.686150869282259e-06}
{"train_loss": 0.01846844144165516, "global_step": 229918, "epoch": 2583, "lr": 5.685882364006445e-06}
{"train_loss": 0.0746728777885437, "global_step": 229919, "epoch": 2583, "lr": 5.685613864688139e-06}
{"train_loss": 0.04168345779180527, "global_step": 229920, "epoch": 2583, "lr": 5.685345371327344e-06}
{"train_loss": 0.03336027264595032, "global_step": 229921, "epoch": 2583, "lr": 5.685076883924128e-06}
{"train_loss": 0.033767443150281906, "global_step": 229922, "epoch": 2583, "lr": 5.684808402478503e-06}
{"train_loss": 0.06497380882501602, "global_step": 229923, "epoch": 2583, "lr": 5.6845399269905285e-06}
{"train_loss": 0.027866583317518234, "global_step": 229924, "epoch": 2583, "lr": 5.684271457460228e-06}
{"train_loss": 0.04072432219982147, "global_step": 229925, "epoch": 2583, "lr": 5.684002993887621e-06}
{"train_loss": 0.05436822772026062, "global_step": 229926, "epoch": 2583, "lr": 5.683734536272772e-06}
{"train_loss": 0.09196111559867859, "global_step": 229927, "epoch": 2583, "lr": 5.6834660846156965e-06}
{"train_loss": 0.05501128360629082, "global_step": 229928, "epoch": 2583, "lr": 5.683197638916443e-06}
{"train_loss": 0.11281754821538925, "global_step": 229929, "epoch": 2583, "lr": 5.682929199175041e-06}
{"train_loss": 0.046302396804094315, "global_step": 229930, "epoch": 2583, "lr": 5.682660765391534e-06}
{"train_loss": 0.057226188480854034, "global_step": 229931, "epoch": 2583, "lr": 5.682392337565945e-06}
{"train_loss": 0.033816590905189514, "global_step": 229932, "epoch": 2583, "lr": 5.6821239156983285e-06}
{"train_loss": 0.0780562162399292, "global_step": 229933, "epoch": 2583, "lr": 5.681855499788702e-06}
{"train_loss": 0.06143879145383835, "global_step": 229934, "epoch": 2583, "lr": 5.681587089837115e-06}
{"train_loss": 0.07073547691106796, "global_step": 229935, "epoch": 2583, "lr": 5.681318685843595e-06}
{"train_loss": 0.02748178131878376, "global_step": 229936, "epoch": 2583, "lr": 5.6810502878081875e-06}
{"train_loss": 0.03881458938121796, "global_step": 229937, "epoch": 2583, "lr": 5.680781895730913e-06}
{"train_loss": 0.07034333050251007, "global_step": 229938, "epoch": 2583, "lr": 5.680513509611818e-06}
{"train_loss": 0.03816048800945282, "global_step": 229939, "epoch": 2583, "lr": 5.680245129450945e-06}
{"train_loss": 0.04892464727163315, "global_step": 229940, "epoch": 2583, "lr": 5.679976755248317e-06}
{"train_loss": 0.07367537170648575, "global_step": 229941, "epoch": 2583, "lr": 5.6797083870039845e-06}
{"train_loss": 0.03307412192225456, "global_step": 229942, "epoch": 2583, "lr": 5.6794400247179635e-06}
{"train_loss": 0.05715673044323921, "global_step": 229943, "epoch": 2583, "lr": 5.679171668390309e-06}
{"train_loss": 0.021221933886408806, "global_step": 229944, "epoch": 2583, "lr": 5.678903318021045e-06}
{"train_loss": 0.057787008583545685, "global_step": 229945, "epoch": 2583, "lr": 5.678634973610219e-06}
{"train_loss": 0.05339057371020317, "global_step": 229946, "epoch": 2583, "lr": 5.6783666351578504e-06}
{"train_loss": 0.038973335176706314, "global_step": 229947, "epoch": 2583, "lr": 5.6780983026639975e-06}
{"train_loss": 0.07788455486297607, "global_step": 229948, "epoch": 2583, "lr": 5.677829976128673e-06}
{"train_loss": 0.10872397571802139, "global_step": 229949, "epoch": 2583, "lr": 5.677561655551933e-06}
{"train_loss": 0.03474009782075882, "global_step": 229950, "epoch": 2583, "lr": 5.677293340933798e-06}
{"train_loss": 0.07249074429273605, "global_step": 229951, "epoch": 2583, "lr": 5.6770250322743194e-06}
{"train_loss": 0.0510486401617527, "global_step": 229952, "epoch": 2583, "lr": 5.676756729573512e-06}
{"train_loss": 0.031877681612968445, "global_step": 229953, "epoch": 2583, "lr": 5.676488432831434e-06}
{"train_loss": 0.07960380613803864, "global_step": 229954, "epoch": 2583, "lr": 5.676220142048105e-06}
{"train_loss": 0.052454933524131775, "global_step": 229955, "epoch": 2583, "lr": 5.6759518572235774e-06}
{"train_loss": 0.08341232687234879, "global_step": 229956, "epoch": 2583, "lr": 5.6756835783578656e-06}
{"train_loss": 0.03241930902004242, "global_step": 229957, "epoch": 2583, "lr": 5.675415305451031e-06}
{"train_loss": 0.05169384554028511, "global_step": 229958, "epoch": 2583, "lr": 5.675147038503087e-06}
{"train_loss": 0.05659227818250656, "global_step": 229959, "epoch": 2583, "lr": 5.674878777514086e-06}
{"train_loss": 0.06572449952363968, "global_step": 229960, "epoch": 2583, "lr": 5.674610522484048e-06}
{"train_loss": 0.04773029685020447, "global_step": 229961, "epoch": 2583, "lr": 5.674342273413031e-06}
{"train_loss": 0.0290505550801754, "global_step": 229962, "epoch": 2583, "lr": 5.674074030301057e-06}
{"train_loss": 0.05095449090003967, "global_step": 229963, "epoch": 2583, "lr": 5.673805793148151e-06}
{"train_loss": 0.036230068653821945, "global_step": 229964, "epoch": 2583, "lr": 5.6735375619543725e-06}
{"train_loss": 0.05161856487393379, "global_step": 229965, "epoch": 2583, "lr": 5.673269336719738e-06}
{"train_loss": 0.055128611624240875, "global_step": 229966, "epoch": 2583, "lr": 5.673001117444305e-06}
{"train_loss": 0.033331383019685745, "global_step": 229967, "epoch": 2583, "lr": 5.6727329041280805e-06}
{"train_loss": 0.07865854352712631, "global_step": 229968, "epoch": 2583, "lr": 5.67246469677113e-06}
{"train_loss": 0.05826372280716896, "global_step": 229969, "epoch": 2583, "lr": 5.672196495373466e-06}
{"train_loss": 0.04090826213359833, "global_step": 229970, "epoch": 2583, "lr": 5.671928299935147e-06}
{"train_loss": 0.051736149936914444, "global_step": 229971, "epoch": 2583, "lr": 5.671660110456184e-06}
{"train_loss": 0.037150051444768906, "global_step": 229972, "epoch": 2583, "lr": 5.6713919269366365e-06}
{"train_loss": 0.08366333693265915, "global_step": 229973, "epoch": 2583, "lr": 5.6711237493765225e-06}
{"train_loss": 0.05724678561091423, "global_step": 229974, "epoch": 2583, "lr": 5.670855577775885e-06}
{"train_loss": 0.05318756658960594, "global_step": 229975, "epoch": 2583, "lr": 5.670587412134765e-06, "val_loss": 9.2174711227417}
{"train_loss": 0.05084097012877464, "global_step": 229976, "epoch": 2584, "lr": 5.670319252453194e-06}
{"train_loss": 0.041659336537122726, "global_step": 229977, "epoch": 2584, "lr": 5.670051098731211e-06}
{"train_loss": 0.03480354696512222, "global_step": 229978, "epoch": 2584, "lr": 5.6697829509688385e-06}
{"train_loss": 0.06125952675938606, "global_step": 229979, "epoch": 2584, "lr": 5.669514809166137e-06}
{"train_loss": 0.038141828030347824, "global_step": 229980, "epoch": 2584, "lr": 5.6692466733231194e-06}
{"train_loss": 0.040502823889255524, "global_step": 229981, "epoch": 2584, "lr": 5.6689785434398405e-06}
{"train_loss": 0.10857707262039185, "global_step": 229982, "epoch": 2584, "lr": 5.668710419516315e-06}
{"train_loss": 0.059387192130088806, "global_step": 229983, "epoch": 2584, "lr": 5.668442301552601e-06}
{"train_loss": 0.04668314382433891, "global_step": 229984, "epoch": 2584, "lr": 5.6681741895487195e-06}
{"train_loss": 0.05345669761300087, "global_step": 229985, "epoch": 2584, "lr": 5.6679060835047144e-06}
{"train_loss": 0.033195704221725464, "global_step": 229986, "epoch": 2584, "lr": 5.667637983420615e-06}
{"train_loss": 0.06034727394580841, "global_step": 229987, "epoch": 2584, "lr": 5.66736988929647e-06}
{"train_loss": 0.039854951202869415, "global_step": 229988, "epoch": 2584, "lr": 5.667101801132296e-06}
{"train_loss": 0.05061378702521324, "global_step": 229989, "epoch": 2584, "lr": 5.666833718928149e-06}
{"train_loss": 0.06108333542943001, "global_step": 229990, "epoch": 2584, "lr": 5.666565642684047e-06}
{"train_loss": 0.04826189577579498, "global_step": 229991, "epoch": 2584, "lr": 5.666297572400042e-06}
{"train_loss": 0.040437255054712296, "global_step": 229992, "epoch": 2584, "lr": 5.66602950807616e-06}
{"train_loss": 0.049086544662714005, "global_step": 229993, "epoch": 2584, "lr": 5.665761449712431e-06}
{"train_loss": 0.07987775653600693, "global_step": 229994, "epoch": 2584, "lr": 5.665493397308919e-06}
{"train_loss": 0.06857678294181824, "global_step": 229995, "epoch": 2584, "lr": 5.665225350865633e-06}
{"train_loss": 0.041818730533123016, "global_step": 229996, "epoch": 2584, "lr": 5.664957310382613e-06}
{"train_loss": 0.04588936269283295, "global_step": 229997, "epoch": 2584, "lr": 5.664689275859908e-06}
{"train_loss": 0.0502108670771122, "global_step": 229998, "epoch": 2584, "lr": 5.66442124729753e-06}
{"train_loss": 0.023389294743537903, "global_step": 229999, "epoch": 2584, "lr": 5.664153224695545e-06}
{"train_loss": 0.03807312995195389, "global_step": 230000, "epoch": 2584, "lr": 5.663885208053976e-06}
{"train_loss": 0.05764831230044365, "global_step": 230001, "epoch": 2584, "lr": 5.663617197372844e-06}
{"train_loss": 0.024925410747528076, "global_step": 230002, "epoch": 2584, "lr": 5.663349192652206e-06}
{"train_loss": 0.02589178830385208, "global_step": 230003, "epoch": 2584, "lr": 5.663081193892084e-06}
{"train_loss": 0.04373742267489433, "global_step": 230004, "epoch": 2584, "lr": 5.662813201092532e-06}
{"train_loss": 0.07002740353345871, "global_step": 230005, "epoch": 2584, "lr": 5.662545214253562e-06}
{"train_loss": 0.0818752720952034, "global_step": 230006, "epoch": 2584, "lr": 5.6622772333752295e-06}
{"train_loss": 0.04222814738750458, "global_step": 230007, "epoch": 2584, "lr": 5.662009258457557e-06}
{"train_loss": 0.07980494201183319, "global_step": 230008, "epoch": 2584, "lr": 5.6617412895005995e-06}
{"train_loss": 0.05490613356232643, "global_step": 230009, "epoch": 2584, "lr": 5.661473326504363e-06}
{"train_loss": 0.020548351109027863, "global_step": 230010, "epoch": 2584, "lr": 5.6612053694689196e-06}
{"train_loss": 0.028202546760439873, "global_step": 230011, "epoch": 2584, "lr": 5.660937418394269e-06}
{"train_loss": 0.06688431650400162, "global_step": 230012, "epoch": 2584, "lr": 5.6606694732804666e-06}
{"train_loss": 0.07595393806695938, "global_step": 230013, "epoch": 2584, "lr": 5.660401534127563e-06}
{"train_loss": 0.05303839594125748, "global_step": 230014, "epoch": 2584, "lr": 5.660133600935563e-06}
{"train_loss": 0.035470131784677505, "global_step": 230015, "epoch": 2584, "lr": 5.659865673704528e-06}
{"train_loss": 0.04213850945234299, "global_step": 230016, "epoch": 2584, "lr": 5.65959775243447e-06}
{"train_loss": 0.07306111603975296, "global_step": 230017, "epoch": 2584, "lr": 5.659329837125455e-06}
{"train_loss": 0.04098324477672577, "global_step": 230018, "epoch": 2584, "lr": 5.659061927777487e-06}
{"train_loss": 0.08446040749549866, "global_step": 230019, "epoch": 2584, "lr": 5.658794024390635e-06}
{"train_loss": 0.06182752549648285, "global_step": 230020, "epoch": 2584, "lr": 5.658526126964903e-06}
{"train_loss": 0.13565793633460999, "global_step": 230021, "epoch": 2584, "lr": 5.6582582355003524e-06}
{"train_loss": 0.08166419714689255, "global_step": 230022, "epoch": 2584, "lr": 5.6579903499969945e-06}
{"train_loss": 0.055190425366163254, "global_step": 230023, "epoch": 2584, "lr": 5.657722470454896e-06}
{"train_loss": 0.03472154214978218, "global_step": 230024, "epoch": 2584, "lr": 5.657454596874062e-06}
{"train_loss": 0.048787541687488556, "global_step": 230025, "epoch": 2584, "lr": 5.657186729254554e-06}
{"train_loss": 0.06405490636825562, "global_step": 230026, "epoch": 2584, "lr": 5.656918867596389e-06}
{"train_loss": 0.042694929987192154, "global_step": 230027, "epoch": 2584, "lr": 5.656651011899617e-06}
{"train_loss": 0.03516065329313278, "global_step": 230028, "epoch": 2584, "lr": 5.656383162164259e-06}
{"train_loss": 0.10401954501867294, "global_step": 230029, "epoch": 2584, "lr": 5.656115318390359e-06}
{"train_loss": 0.030134284868836403, "global_step": 230030, "epoch": 2584, "lr": 5.655847480577964e-06}
{"train_loss": 0.059891477227211, "global_step": 230031, "epoch": 2584, "lr": 5.655579648727094e-06}
{"train_loss": 0.02451923117041588, "global_step": 230032, "epoch": 2584, "lr": 5.655311822837794e-06}
{"train_loss": 0.05297866091132164, "global_step": 230033, "epoch": 2584, "lr": 5.655044002910104e-06}
{"train_loss": 0.08339686691761017, "global_step": 230034, "epoch": 2584, "lr": 5.654776188944039e-06}
{"train_loss": 0.06117166578769684, "global_step": 230035, "epoch": 2584, "lr": 5.65450838093966e-06}
{"train_loss": 0.0959024429321289, "global_step": 230036, "epoch": 2584, "lr": 5.6542405788969795e-06}
{"train_loss": 0.025501318275928497, "global_step": 230037, "epoch": 2584, "lr": 5.653972782816058e-06}
{"train_loss": 0.020596040412783623, "global_step": 230038, "epoch": 2584, "lr": 5.653704992696907e-06}
{"train_loss": 0.0656580701470375, "global_step": 230039, "epoch": 2584, "lr": 5.653437208539586e-06}
{"train_loss": 0.017501210793852806, "global_step": 230040, "epoch": 2584, "lr": 5.653169430344124e-06}
{"train_loss": 0.06959829479455948, "global_step": 230041, "epoch": 2584, "lr": 5.652901658110537e-06}
{"train_loss": 0.029550909996032715, "global_step": 230042, "epoch": 2584, "lr": 5.652633891838893e-06}
{"train_loss": 0.06105450913310051, "global_step": 230043, "epoch": 2584, "lr": 5.652366131529196e-06}
{"train_loss": 0.027077993378043175, "global_step": 230044, "epoch": 2584, "lr": 5.652098377181508e-06}
{"train_loss": 0.0706675574183464, "global_step": 230045, "epoch": 2584, "lr": 5.6518306287958454e-06}
{"train_loss": 0.04113837704062462, "global_step": 230046, "epoch": 2584, "lr": 5.651562886372269e-06}
{"train_loss": 0.039087098091840744, "global_step": 230047, "epoch": 2584, "lr": 5.6512951499107845e-06}
{"train_loss": 0.062472160905599594, "global_step": 230048, "epoch": 2584, "lr": 5.651027419411442e-06}
{"train_loss": 0.08020845800638199, "global_step": 230049, "epoch": 2584, "lr": 5.650759694874291e-06}
{"train_loss": 0.10027797520160675, "global_step": 230050, "epoch": 2584, "lr": 5.650491976299343e-06}
{"train_loss": 0.06330044567584991, "global_step": 230051, "epoch": 2584, "lr": 5.65022426368666e-06}
{"train_loss": 0.053016938269138336, "global_step": 230052, "epoch": 2584, "lr": 5.649956557036251e-06}
{"train_loss": 0.02441619522869587, "global_step": 230053, "epoch": 2584, "lr": 5.6496888563481734e-06}
{"train_loss": 0.07551348209381104, "global_step": 230054, "epoch": 2584, "lr": 5.649421161622443e-06}
{"train_loss": 0.03974726423621178, "global_step": 230055, "epoch": 2584, "lr": 5.64915347285912e-06}
{"train_loss": 0.043455835431814194, "global_step": 230056, "epoch": 2584, "lr": 5.648885790058217e-06}
{"train_loss": 0.10086602717638016, "global_step": 230057, "epoch": 2584, "lr": 5.6486181132197944e-06}
{"train_loss": 0.0124352453276515, "global_step": 230058, "epoch": 2584, "lr": 5.6483504423438585e-06}
{"train_loss": 0.038916099816560745, "global_step": 230059, "epoch": 2584, "lr": 5.648082777430475e-06}
{"train_loss": 0.06272035092115402, "global_step": 230060, "epoch": 2584, "lr": 5.647815118479655e-06}
{"train_loss": 0.05541292205452919, "global_step": 230061, "epoch": 2584, "lr": 5.64754746549146e-06}
{"train_loss": 0.025812724605202675, "global_step": 230062, "epoch": 2584, "lr": 5.647279818465895e-06}
{"train_loss": 0.07852943986654282, "global_step": 230063, "epoch": 2584, "lr": 5.6470121774030275e-06}
{"train_loss": 0.05372048727121581, "global_step": 230064, "epoch": 2584, "lr": 5.646744542302867e-06, "val_loss": 9.243377685546875}
{"train_loss": 0.06154376640915871, "global_step": 230065, "epoch": 2585, "lr": 5.646476913165461e-06}
{"train_loss": 0.07800185680389404, "global_step": 230066, "epoch": 2585, "lr": 5.646209289990856e-06}
{"train_loss": 0.1024562194943428, "global_step": 230067, "epoch": 2585, "lr": 5.6459416727790645e-06}
{"train_loss": 0.0839892327785492, "global_step": 230068, "epoch": 2585, "lr": 5.6456740615301484e-06}
{"train_loss": 0.0628761574625969, "global_step": 230069, "epoch": 2585, "lr": 5.645406456244118e-06}
{"train_loss": 0.08639463782310486, "global_step": 230070, "epoch": 2585, "lr": 5.645138856921034e-06}
{"train_loss": 0.07995707541704178, "global_step": 230071, "epoch": 2585, "lr": 5.644871263560913e-06}
{"train_loss": 0.041461504995822906, "global_step": 230072, "epoch": 2585, "lr": 5.644603676163806e-06}
{"train_loss": 0.029364733025431633, "global_step": 230073, "epoch": 2585, "lr": 5.644336094729741e-06}
{"train_loss": 0.044880203902721405, "global_step": 230074, "epoch": 2585, "lr": 5.64406851925875e-06}
{"train_loss": 0.038197558373212814, "global_step": 230075, "epoch": 2585, "lr": 5.643800949750877e-06}
{"train_loss": 0.0636281743645668, "global_step": 230076, "epoch": 2585, "lr": 5.643533386206146e-06}
{"train_loss": 0.06459159404039383, "global_step": 230077, "epoch": 2585, "lr": 5.643265828624611e-06}
{"train_loss": 0.024101397022604942, "global_step": 230078, "epoch": 2585, "lr": 5.642998277006295e-06}
{"train_loss": 0.031153906136751175, "global_step": 230079, "epoch": 2585, "lr": 5.6427307313512315e-06}
{"train_loss": 0.056112684309482574, "global_step": 230080, "epoch": 2585, "lr": 5.642463191659475e-06}
{"train_loss": 0.04766632616519928, "global_step": 230081, "epoch": 2585, "lr": 5.642195657931032e-06}
{"train_loss": 0.0325741209089756, "global_step": 230082, "epoch": 2585, "lr": 5.641928130165969e-06}
{"train_loss": 0.07026195526123047, "global_step": 230083, "epoch": 2585, "lr": 5.641660608364296e-06}
{"train_loss": 0.049603916704654694, "global_step": 230084, "epoch": 2585, "lr": 5.641393092526065e-06}
{"train_loss": 0.049984924495220184, "global_step": 230085, "epoch": 2585, "lr": 5.641125582651313e-06}
{"train_loss": 0.021335672587156296, "global_step": 230086, "epoch": 2585, "lr": 5.640858078740063e-06}
{"train_loss": 0.05777636542916298, "global_step": 230087, "epoch": 2585, "lr": 5.6405905807923705e-06}
{"train_loss": 0.0627087876200676, "global_step": 230088, "epoch": 2585, "lr": 5.640323088808247e-06}
{"train_loss": 0.09078248590230942, "global_step": 230089, "epoch": 2585, "lr": 5.6400556027877585e-06}
{"train_loss": 0.027343938127160072, "global_step": 230090, "epoch": 2585, "lr": 5.639788122730904e-06}
{"train_loss": 0.047311119735240936, "global_step": 230091, "epoch": 2585, "lr": 5.6395206486377585e-06}
{"train_loss": 0.023691173642873764, "global_step": 230092, "epoch": 2585, "lr": 5.639253180508319e-06}
{"train_loss": 0.02734377048909664, "global_step": 230093, "epoch": 2585, "lr": 5.638985718342659e-06}
{"train_loss": 0.060328323394060135, "global_step": 230094, "epoch": 2585, "lr": 5.638718262140786e-06}
{"train_loss": 0.049140460789203644, "global_step": 230095, "epoch": 2585, "lr": 5.6384508119027515e-06}
{"train_loss": 0.02543012611567974, "global_step": 230096, "epoch": 2585, "lr": 5.63818336762858e-06}
{"train_loss": 0.034478168934583664, "global_step": 230097, "epoch": 2585, "lr": 5.637915929318327e-06}
{"train_loss": 0.03887840732932091, "global_step": 230098, "epoch": 2585, "lr": 5.637648496972003e-06}
{"train_loss": 0.06915302574634552, "global_step": 230099, "epoch": 2585, "lr": 5.637381070589664e-06}
{"train_loss": 0.046850718557834625, "global_step": 230100, "epoch": 2585, "lr": 5.637113650171328e-06}
{"train_loss": 0.03151940554380417, "global_step": 230101, "epoch": 2585, "lr": 5.636846235717047e-06}
{"train_loss": 0.06805048882961273, "global_step": 230102, "epoch": 2585, "lr": 5.636578827226857e-06}
{"train_loss": 0.04266056790947914, "global_step": 230103, "epoch": 2585, "lr": 5.63631142470078e-06}
{"train_loss": 0.0362577885389328, "global_step": 230104, "epoch": 2585, "lr": 5.63604402813887e-06}
{"train_loss": 0.015842363238334656, "global_step": 230105, "epoch": 2585, "lr": 5.63577663754114e-06}
{"train_loss": 0.022474976256489754, "global_step": 230106, "epoch": 2585, "lr": 5.635509252907651e-06}
{"train_loss": 0.049369119107723236, "global_step": 230107, "epoch": 2585, "lr": 5.635241874238417e-06}
{"train_loss": 0.10114393383264542, "global_step": 230108, "epoch": 2585, "lr": 5.634974501533497e-06}
{"train_loss": 0.0973174050450325, "global_step": 230109, "epoch": 2585, "lr": 5.634707134792899e-06}
{"train_loss": 0.019854390993714333, "global_step": 230110, "epoch": 2585, "lr": 5.634439774016692e-06}
{"train_loss": 0.03693336248397827, "global_step": 230111, "epoch": 2585, "lr": 5.6341724192048865e-06}
{"train_loss": 0.07902855426073074, "global_step": 230112, "epoch": 2585, "lr": 5.633905070357514e-06}
{"train_loss": 0.017057687044143677, "global_step": 230113, "epoch": 2585, "lr": 5.633637727474639e-06}
{"train_loss": 0.05400752276182175, "global_step": 230114, "epoch": 2585, "lr": 5.63337039055627e-06}
{"train_loss": 0.05760142207145691, "global_step": 230115, "epoch": 2585, "lr": 5.633103059602457e-06}
{"train_loss": 0.030287161469459534, "global_step": 230116, "epoch": 2585, "lr": 5.63283573461324e-06}
{"train_loss": 0.02361270599067211, "global_step": 230117, "epoch": 2585, "lr": 5.632568415588635e-06}
{"train_loss": 0.08077026903629303, "global_step": 230118, "epoch": 2585, "lr": 5.632301102528698e-06}
{"train_loss": 0.07340414077043533, "global_step": 230119, "epoch": 2585, "lr": 5.63203379543345e-06}
{"train_loss": 0.06818054616451263, "global_step": 230120, "epoch": 2585, "lr": 5.631766494302931e-06}
{"train_loss": 0.059305768460035324, "global_step": 230121, "epoch": 2585, "lr": 5.631499199137191e-06}
{"train_loss": 0.06829702109098434, "global_step": 230122, "epoch": 2585, "lr": 5.631231909936247e-06}
{"train_loss": 0.06993662565946579, "global_step": 230123, "epoch": 2585, "lr": 5.6309646267001525e-06}
{"train_loss": 0.061545174568891525, "global_step": 230124, "epoch": 2585, "lr": 5.630697349428926e-06}
{"train_loss": 0.02837979793548584, "global_step": 230125, "epoch": 2585, "lr": 5.630430078122617e-06}
{"train_loss": 0.06387987732887268, "global_step": 230126, "epoch": 2585, "lr": 5.630162812781253e-06}
{"train_loss": 0.05032084509730339, "global_step": 230127, "epoch": 2585, "lr": 5.6298955534048796e-06}
{"train_loss": 0.06599949300289154, "global_step": 230128, "epoch": 2585, "lr": 5.629628299993511e-06}
{"train_loss": 0.00715976394712925, "global_step": 230129, "epoch": 2585, "lr": 5.629361052547211e-06}
{"train_loss": 0.025315698236227036, "global_step": 230130, "epoch": 2585, "lr": 5.629093811065994e-06}
{"train_loss": 0.03637140244245529, "global_step": 230131, "epoch": 2585, "lr": 5.628826575549917e-06}
{"train_loss": 0.06627976894378662, "global_step": 230132, "epoch": 2585, "lr": 5.628559345998991e-06}
{"train_loss": 0.028822679072618484, "global_step": 230133, "epoch": 2585, "lr": 5.62829212241327e-06}
{"train_loss": 0.035848889499902725, "global_step": 230134, "epoch": 2585, "lr": 5.628024904792778e-06}
{"train_loss": 0.014813610352575779, "global_step": 230135, "epoch": 2585, "lr": 5.6277576931375695e-06}
{"train_loss": 0.06950005143880844, "global_step": 230136, "epoch": 2585, "lr": 5.627490487447657e-06}
{"train_loss": 0.10354474186897278, "global_step": 230137, "epoch": 2585, "lr": 5.627223287723094e-06}
{"train_loss": 0.04073086753487587, "global_step": 230138, "epoch": 2585, "lr": 5.6269560939639035e-06}
{"train_loss": 0.04396549239754677, "global_step": 230139, "epoch": 2585, "lr": 5.626688906170124e-06}
{"train_loss": 0.06067022681236267, "global_step": 230140, "epoch": 2585, "lr": 5.626421724341813e-06}
{"train_loss": 0.06800586730241776, "global_step": 230141, "epoch": 2585, "lr": 5.626154548478974e-06}
{"train_loss": 0.03599231317639351, "global_step": 230142, "epoch": 2585, "lr": 5.625887378581668e-06}
{"train_loss": 0.08779381960630417, "global_step": 230143, "epoch": 2585, "lr": 5.625620214649907e-06}
{"train_loss": 0.03350386396050453, "global_step": 230144, "epoch": 2585, "lr": 5.6253530566837575e-06}
{"train_loss": 0.0236817579716444, "global_step": 230145, "epoch": 2585, "lr": 5.625085904683225e-06}
{"train_loss": 0.05007475987076759, "global_step": 230146, "epoch": 2585, "lr": 5.624818758648364e-06}
{"train_loss": 0.06123054772615433, "global_step": 230147, "epoch": 2585, "lr": 5.6245516185792045e-06}
{"train_loss": 0.058157164603471756, "global_step": 230148, "epoch": 2585, "lr": 5.624284484475783e-06}
{"train_loss": 0.05762951448559761, "global_step": 230149, "epoch": 2585, "lr": 5.624017356338146e-06}
{"train_loss": 0.05540633574128151, "global_step": 230150, "epoch": 2585, "lr": 5.623750234166303e-06}
{"train_loss": 0.082255519926548, "global_step": 230151, "epoch": 2585, "lr": 5.623483117960315e-06}
{"train_loss": 0.058946721255779266, "global_step": 230152, "epoch": 2585, "lr": 5.6232160077202e-06}
{"train_loss": 0.0517327379581801, "global_step": 230153, "epoch": 2585, "lr": 5.622948903446018e-06, "val_loss": 9.313859939575195, "train_action_mse_error": 20.969314575195312}
{"train_loss": 0.07120673358440399, "global_step": 230154, "epoch": 2586, "lr": 5.622681805137775e-06}
{"train_loss": 0.07900901883840561, "global_step": 230155, "epoch": 2586, "lr": 5.622414712795531e-06}
{"train_loss": 0.03587697818875313, "global_step": 230156, "epoch": 2586, "lr": 5.6221476264193056e-06}
{"train_loss": 0.029807161539793015, "global_step": 230157, "epoch": 2586, "lr": 5.621880546009145e-06}
{"train_loss": 0.05912806838750839, "global_step": 230158, "epoch": 2586, "lr": 5.621613471565079e-06}
{"train_loss": 0.014540345408022404, "global_step": 230159, "epoch": 2586, "lr": 5.621346403087152e-06}
{"train_loss": 0.06429049372673035, "global_step": 230160, "epoch": 2586, "lr": 5.6210793405753855e-06}
{"train_loss": 0.09382256865501404, "global_step": 230161, "epoch": 2586, "lr": 5.6208122840298305e-06}
{"train_loss": 0.06852111220359802, "global_step": 230162, "epoch": 2586, "lr": 5.620545233450514e-06}
{"train_loss": 0.0363173671066761, "global_step": 230163, "epoch": 2586, "lr": 5.62027818883748e-06}
{"train_loss": 0.0643778145313263, "global_step": 230164, "epoch": 2586, "lr": 5.620011150190746e-06}
{"train_loss": 0.06806302070617676, "global_step": 230165, "epoch": 2586, "lr": 5.6197441175103725e-06}
{"train_loss": 0.052208274602890015, "global_step": 230166, "epoch": 2586, "lr": 5.619477090796377e-06}
{"train_loss": 0.05223240330815315, "global_step": 230167, "epoch": 2586, "lr": 5.619210070048808e-06}
{"train_loss": 0.055676255375146866, "global_step": 230168, "epoch": 2586, "lr": 5.618943055267683e-06}
{"train_loss": 0.022920070216059685, "global_step": 230169, "epoch": 2586, "lr": 5.618676046453064e-06}
{"train_loss": 0.06587620824575424, "global_step": 230170, "epoch": 2586, "lr": 5.61840904360496e-06}
{"train_loss": 0.023287136107683182, "global_step": 230171, "epoch": 2586, "lr": 5.618142046723435e-06}
{"train_loss": 0.057109568268060684, "global_step": 230172, "epoch": 2586, "lr": 5.617875055808497e-06}
{"train_loss": 0.03217056766152382, "global_step": 230173, "epoch": 2586, "lr": 5.617608070860209e-06}
{"train_loss": 0.07375115901231766, "global_step": 230174, "epoch": 2586, "lr": 5.617341091878575e-06}
{"train_loss": 0.0354023203253746, "global_step": 230175, "epoch": 2586, "lr": 5.617074118863652e-06}
{"train_loss": 0.06216244399547577, "global_step": 230176, "epoch": 2586, "lr": 5.616807151815484e-06}
{"train_loss": 0.030062904581427574, "global_step": 230177, "epoch": 2586, "lr": 5.616540190734082e-06}
{"train_loss": 0.11938019841909409, "global_step": 230178, "epoch": 2586, "lr": 5.616273235619507e-06}
{"train_loss": 0.051655109971761703, "global_step": 230179, "epoch": 2586, "lr": 5.6160062864717756e-06}
{"train_loss": 0.03107539378106594, "global_step": 230180, "epoch": 2586, "lr": 5.6157393432909376e-06}
{"train_loss": 0.02247018553316593, "global_step": 230181, "epoch": 2586, "lr": 5.6154724060770155e-06}
{"train_loss": 0.059832438826560974, "global_step": 230182, "epoch": 2586, "lr": 5.6152054748300595e-06}
{"train_loss": 0.0445699505507946, "global_step": 230183, "epoch": 2586, "lr": 5.614938549550091e-06}
{"train_loss": 0.0667998343706131, "global_step": 230184, "epoch": 2586, "lr": 5.614671630237161e-06}
{"train_loss": 0.04641210660338402, "global_step": 230185, "epoch": 2586, "lr": 5.614404716891286e-06}
{"train_loss": 0.020293261855840683, "global_step": 230186, "epoch": 2586, "lr": 5.6141378095125265e-06}
{"train_loss": 0.0510692372918129, "global_step": 230187, "epoch": 2586, "lr": 5.613870908100893e-06}
{"train_loss": 0.026641299948096275, "global_step": 230188, "epoch": 2586, "lr": 5.6136040126564426e-06}
{"train_loss": 0.06761794537305832, "global_step": 230189, "epoch": 2586, "lr": 5.613337123179202e-06}
{"train_loss": 0.056506894528865814, "global_step": 230190, "epoch": 2586, "lr": 5.613070239669199e-06}
{"train_loss": 0.05135679244995117, "global_step": 230191, "epoch": 2586, "lr": 5.61280336212649e-06}
{"train_loss": 0.018292004242539406, "global_step": 230192, "epoch": 2586, "lr": 5.612536490551085e-06}
{"train_loss": 0.009625091217458248, "global_step": 230193, "epoch": 2586, "lr": 5.612269624943045e-06}
{"train_loss": 0.04248154163360596, "global_step": 230194, "epoch": 2586, "lr": 5.612002765302382e-06}
{"train_loss": 0.0448782742023468, "global_step": 230195, "epoch": 2586, "lr": 5.611735911629157e-06}
{"train_loss": 0.09674077481031418, "global_step": 230196, "epoch": 2586, "lr": 5.611469063923386e-06}
{"train_loss": 0.07060596346855164, "global_step": 230197, "epoch": 2586, "lr": 5.6112022221851135e-06}
{"train_loss": 0.06684087216854095, "global_step": 230198, "epoch": 2586, "lr": 5.610935386414373e-06}
{"train_loss": 0.022398008033633232, "global_step": 230199, "epoch": 2586, "lr": 5.610668556611204e-06}
{"train_loss": 0.05473894253373146, "global_step": 230200, "epoch": 2586, "lr": 5.610401732775633e-06}
{"train_loss": 0.03726765513420105, "global_step": 230201, "epoch": 2586, "lr": 5.610134914907711e-06}
{"train_loss": 0.048250265419483185, "global_step": 230202, "epoch": 2586, "lr": 5.609868103007459e-06}
{"train_loss": 0.03299108147621155, "global_step": 230203, "epoch": 2586, "lr": 5.609601297074923e-06}
{"train_loss": 0.038123372942209244, "global_step": 230204, "epoch": 2586, "lr": 5.60933449711013e-06}
{"train_loss": 0.09676118195056915, "global_step": 230205, "epoch": 2586, "lr": 5.60906770311313e-06}
{"train_loss": 0.050178833305835724, "global_step": 230206, "epoch": 2586, "lr": 5.60880091508394e-06}
{"train_loss": 0.05131901800632477, "global_step": 230207, "epoch": 2586, "lr": 5.608534133022614e-06}
{"train_loss": 0.04148116707801819, "global_step": 230208, "epoch": 2586, "lr": 5.60826735692917e-06}
{"train_loss": 0.02348768152296543, "global_step": 230209, "epoch": 2586, "lr": 5.608000586803664e-06}
{"train_loss": 0.0564386248588562, "global_step": 230210, "epoch": 2586, "lr": 5.6077338226461106e-06}
{"train_loss": 0.04257674515247345, "global_step": 230211, "epoch": 2586, "lr": 5.607467064456557e-06}
{"train_loss": 0.05508078634738922, "global_step": 230212, "epoch": 2586, "lr": 5.607200312235045e-06}
{"train_loss": 0.03776085749268532, "global_step": 230213, "epoch": 2586, "lr": 5.606933565981598e-06}
{"train_loss": 0.07665785402059555, "global_step": 230214, "epoch": 2586, "lr": 5.6066668256962665e-06}
{"train_loss": 0.047596823424100876, "global_step": 230215, "epoch": 2586, "lr": 5.606400091379072e-06}
{"train_loss": 0.04270847514271736, "global_step": 230216, "epoch": 2586, "lr": 5.606133363030058e-06}
{"train_loss": 0.035273194313049316, "global_step": 230217, "epoch": 2586, "lr": 5.605866640649254e-06}
{"train_loss": 0.05525287613272667, "global_step": 230218, "epoch": 2586, "lr": 5.6055999242367045e-06}
{"train_loss": 0.10063432902097702, "global_step": 230219, "epoch": 2586, "lr": 5.605333213792435e-06}
{"train_loss": 0.09162182360887527, "global_step": 230220, "epoch": 2586, "lr": 5.6050665093164975e-06}
{"train_loss": 0.13792000710964203, "global_step": 230221, "epoch": 2586, "lr": 5.604799810808908e-06}
{"train_loss": 0.05821753665804863, "global_step": 230222, "epoch": 2586, "lr": 5.604533118269723e-06}
{"train_loss": 0.050396692007780075, "global_step": 230223, "epoch": 2586, "lr": 5.604266431698951e-06}
{"train_loss": 0.029157157987356186, "global_step": 230224, "epoch": 2586, "lr": 5.603999751096661e-06}
{"train_loss": 0.0605098232626915, "global_step": 230225, "epoch": 2586, "lr": 5.603733076462858e-06}
{"train_loss": 0.08671706914901733, "global_step": 230226, "epoch": 2586, "lr": 5.603466407797608e-06}
{"train_loss": 0.05780575051903725, "global_step": 230227, "epoch": 2586, "lr": 5.603199745100923e-06}
{"train_loss": 0.04416441544890404, "global_step": 230228, "epoch": 2586, "lr": 5.602933088372842e-06}
{"train_loss": 0.03458070009946823, "global_step": 230229, "epoch": 2586, "lr": 5.602666437613413e-06}
{"train_loss": 0.06418132781982422, "global_step": 230230, "epoch": 2586, "lr": 5.6023997928226545e-06}
{"train_loss": 0.04550623521208763, "global_step": 230231, "epoch": 2586, "lr": 5.6021331540006275e-06}
{"train_loss": 0.0513097420334816, "global_step": 230232, "epoch": 2586, "lr": 5.601866521147337e-06}
{"train_loss": 0.05743063986301422, "global_step": 230233, "epoch": 2586, "lr": 5.601599894262843e-06}
{"train_loss": 0.03011190891265869, "global_step": 230234, "epoch": 2586, "lr": 5.601333273347165e-06}
{"train_loss": 0.02815527282655239, "global_step": 230235, "epoch": 2586, "lr": 5.6010666584003616e-06}
{"train_loss": 0.02395438216626644, "global_step": 230236, "epoch": 2586, "lr": 5.6008000494224385e-06}
{"train_loss": 0.0831681340932846, "global_step": 230237, "epoch": 2586, "lr": 5.600533446413453e-06}
{"train_loss": 0.05351206660270691, "global_step": 230238, "epoch": 2586, "lr": 5.600266849373431e-06}
{"train_loss": 0.07780960202217102, "global_step": 230239, "epoch": 2586, "lr": 5.600000258302418e-06}
{"train_loss": 0.06829793751239777, "global_step": 230240, "epoch": 2586, "lr": 5.5997336732004355e-06}
{"train_loss": 0.03049590438604355, "global_step": 230241, "epoch": 2586, "lr": 5.59946709406754e-06}
{"train_loss": 0.052400848922434816, "global_step": 230242, "epoch": 2586, "lr": 5.599200520903741e-06, "val_loss": 9.246743202209473}
{"train_loss": 0.046461205929517746, "global_step": 230243, "epoch": 2587, "lr": 5.598933953709101e-06}
{"train_loss": 0.06659141182899475, "global_step": 230244, "epoch": 2587, "lr": 5.598667392483631e-06}
{"train_loss": 0.029693787917494774, "global_step": 230245, "epoch": 2587, "lr": 5.598400837227391e-06}
{"train_loss": 0.0607149600982666, "global_step": 230246, "epoch": 2587, "lr": 5.598134287940393e-06}
{"train_loss": 0.06774162501096725, "global_step": 230247, "epoch": 2587, "lr": 5.597867744622687e-06}
{"train_loss": 0.03917102888226509, "global_step": 230248, "epoch": 2587, "lr": 5.597601207274317e-06}
{"train_loss": 0.06905962526798248, "global_step": 230249, "epoch": 2587, "lr": 5.597334675895294e-06}
{"train_loss": 0.1072879433631897, "global_step": 230250, "epoch": 2587, "lr": 5.597068150485679e-06}
{"train_loss": 0.02427939884364605, "global_step": 230251, "epoch": 2587, "lr": 5.59680163104549e-06}
{"train_loss": 0.04114794358611107, "global_step": 230252, "epoch": 2587, "lr": 5.59653511757478e-06}
{"train_loss": 0.08086193352937698, "global_step": 230253, "epoch": 2587, "lr": 5.596268610073563e-06}
{"train_loss": 0.07168909907341003, "global_step": 230254, "epoch": 2587, "lr": 5.596002108541893e-06}
{"train_loss": 0.029400696977972984, "global_step": 230255, "epoch": 2587, "lr": 5.5957356129797965e-06}
{"train_loss": 0.04178530350327492, "global_step": 230256, "epoch": 2587, "lr": 5.595469123387314e-06}
{"train_loss": 0.023844026029109955, "global_step": 230257, "epoch": 2587, "lr": 5.5952026397644745e-06}
{"train_loss": 0.02670559287071228, "global_step": 230258, "epoch": 2587, "lr": 5.594936162111331e-06}
{"train_loss": 0.07030724734067917, "global_step": 230259, "epoch": 2587, "lr": 5.594669690427895e-06}
{"train_loss": 0.07267720997333527, "global_step": 230260, "epoch": 2587, "lr": 5.594403224714223e-06}
{"train_loss": 0.09404315054416656, "global_step": 230261, "epoch": 2587, "lr": 5.594136764970332e-06}
{"train_loss": 0.07358995079994202, "global_step": 230262, "epoch": 2587, "lr": 5.593870311196281e-06}
{"train_loss": 0.0352201871573925, "global_step": 230263, "epoch": 2587, "lr": 5.593603863392077e-06}
{"train_loss": 0.04526103660464287, "global_step": 230264, "epoch": 2587, "lr": 5.593337421557787e-06}
{"train_loss": 0.051706843078136444, "global_step": 230265, "epoch": 2587, "lr": 5.593070985693433e-06}
{"train_loss": 0.04620968922972679, "global_step": 230266, "epoch": 2587, "lr": 5.592804555799036e-06}
{"train_loss": 0.06194914132356644, "global_step": 230267, "epoch": 2587, "lr": 5.592538131874653e-06}
{"train_loss": 0.05751457065343857, "global_step": 230268, "epoch": 2587, "lr": 5.5922717139203055e-06}
{"train_loss": 0.05590662360191345, "global_step": 230269, "epoch": 2587, "lr": 5.592005301936049e-06}
{"train_loss": 0.04227762296795845, "global_step": 230270, "epoch": 2587, "lr": 5.591738895921889e-06}
{"train_loss": 0.04970007389783859, "global_step": 230271, "epoch": 2587, "lr": 5.591472495877892e-06}
{"train_loss": 0.03229355067014694, "global_step": 230272, "epoch": 2587, "lr": 5.59120610180407e-06}
{"train_loss": 0.04101663827896118, "global_step": 230273, "epoch": 2587, "lr": 5.590939713700477e-06}
{"train_loss": 0.03846803680062294, "global_step": 230274, "epoch": 2587, "lr": 5.5906733315671364e-06}
{"train_loss": 0.06091560423374176, "global_step": 230275, "epoch": 2587, "lr": 5.590406955404093e-06}
{"train_loss": 0.060863103717565536, "global_step": 230276, "epoch": 2587, "lr": 5.590140585211373e-06}
{"train_loss": 0.026641154661774635, "global_step": 230277, "epoch": 2587, "lr": 5.589874220989022e-06}
{"train_loss": 0.0683051198720932, "global_step": 230278, "epoch": 2587, "lr": 5.5896078627370616e-06}
{"train_loss": 0.08346649259328842, "global_step": 230279, "epoch": 2587, "lr": 5.589341510455554e-06}
{"train_loss": 0.07012089341878891, "global_step": 230280, "epoch": 2587, "lr": 5.589075164144503e-06}
{"train_loss": 0.045641470700502396, "global_step": 230281, "epoch": 2587, "lr": 5.588808823803965e-06}
{"train_loss": 0.02492791786789894, "global_step": 230282, "epoch": 2587, "lr": 5.588542489433967e-06}
{"train_loss": 0.07531824707984924, "global_step": 230283, "epoch": 2587, "lr": 5.588276161034544e-06}
{"train_loss": 0.042079173028469086, "global_step": 230284, "epoch": 2587, "lr": 5.588009838605751e-06}
{"train_loss": 0.059409402310848236, "global_step": 230285, "epoch": 2587, "lr": 5.587743522147598e-06}
{"train_loss": 0.03462337329983711, "global_step": 230286, "epoch": 2587, "lr": 5.587477211660136e-06}
{"train_loss": 0.03126184269785881, "global_step": 230287, "epoch": 2587, "lr": 5.587210907143392e-06}
{"train_loss": 0.03320486098527908, "global_step": 230288, "epoch": 2587, "lr": 5.5869446085974105e-06}
{"train_loss": 0.05949284881353378, "global_step": 230289, "epoch": 2587, "lr": 5.58667831602222e-06}
{"train_loss": 0.06569535285234451, "global_step": 230290, "epoch": 2587, "lr": 5.586412029417865e-06}
{"train_loss": 0.08386009186506271, "global_step": 230291, "epoch": 2587, "lr": 5.586145748784366e-06}
{"train_loss": 0.05235584080219269, "global_step": 230292, "epoch": 2587, "lr": 5.5858794741217805e-06}
{"train_loss": 0.04615121707320213, "global_step": 230293, "epoch": 2587, "lr": 5.585613205430118e-06}
{"train_loss": 0.0175784919410944, "global_step": 230294, "epoch": 2587, "lr": 5.585346942709441e-06}
{"train_loss": 0.01779353618621826, "global_step": 230295, "epoch": 2587, "lr": 5.585080685959765e-06}
{"train_loss": 0.03440554440021515, "global_step": 230296, "epoch": 2587, "lr": 5.584814435181141e-06}
{"train_loss": 0.08028274774551392, "global_step": 230297, "epoch": 2587, "lr": 5.58454819037359e-06}
{"train_loss": 0.02636568993330002, "global_step": 230298, "epoch": 2587, "lr": 5.584281951537162e-06}
{"train_loss": 0.04606889560818672, "global_step": 230299, "epoch": 2587, "lr": 5.584015718671876e-06}
{"train_loss": 0.05586491525173187, "global_step": 230300, "epoch": 2587, "lr": 5.58374949177779e-06}
{"train_loss": 0.04153602570295334, "global_step": 230301, "epoch": 2587, "lr": 5.583483270854911e-06}
{"train_loss": 0.05410512164235115, "global_step": 230302, "epoch": 2587, "lr": 5.583217055903311e-06}
{"train_loss": 0.06557586044073105, "global_step": 230303, "epoch": 2587, "lr": 5.582950846922991e-06}
{"train_loss": 0.06380940973758698, "global_step": 230304, "epoch": 2587, "lr": 5.58268464391401e-06}
{"train_loss": 0.0756225511431694, "global_step": 230305, "epoch": 2587, "lr": 5.582418446876397e-06}
{"train_loss": 0.041441019624471664, "global_step": 230306, "epoch": 2587, "lr": 5.58215225581018e-06}
{"train_loss": 0.04508396238088608, "global_step": 230307, "epoch": 2587, "lr": 5.581886070715409e-06}
{"train_loss": 0.07386558502912521, "global_step": 230308, "epoch": 2587, "lr": 5.581619891592099e-06}
{"train_loss": 0.01817937381565571, "global_step": 230309, "epoch": 2587, "lr": 5.581353718440313e-06}
{"train_loss": 0.042493000626564026, "global_step": 230310, "epoch": 2587, "lr": 5.581087551260061e-06}
{"train_loss": 0.012313270941376686, "global_step": 230311, "epoch": 2587, "lr": 5.5808213900513995e-06}
{"train_loss": 0.04600691795349121, "global_step": 230312, "epoch": 2587, "lr": 5.58055523481435e-06}
{"train_loss": 0.04003746435046196, "global_step": 230313, "epoch": 2587, "lr": 5.580289085548957e-06}
{"train_loss": 0.0688566118478775, "global_step": 230314, "epoch": 2587, "lr": 5.580022942255248e-06}
{"train_loss": 0.055838074535131454, "global_step": 230315, "epoch": 2587, "lr": 5.579756804933267e-06}
{"train_loss": 0.028039706870913506, "global_step": 230316, "epoch": 2587, "lr": 5.579490673583043e-06}
{"train_loss": 0.039313774555921555, "global_step": 230317, "epoch": 2587, "lr": 5.57922454820462e-06}
{"train_loss": 0.06183065101504326, "global_step": 230318, "epoch": 2587, "lr": 5.57895842879802e-06}
{"train_loss": 0.027511125430464745, "global_step": 230319, "epoch": 2587, "lr": 5.578692315363293e-06}
{"train_loss": 0.03886282444000244, "global_step": 230320, "epoch": 2587, "lr": 5.578426207900472e-06}
{"train_loss": 0.1102423369884491, "global_step": 230321, "epoch": 2587, "lr": 5.5781601064095855e-06}
{"train_loss": 0.03345910832285881, "global_step": 230322, "epoch": 2587, "lr": 5.577894010890683e-06}
{"train_loss": 0.06277510523796082, "global_step": 230323, "epoch": 2587, "lr": 5.577627921343776e-06}
{"train_loss": 0.05384989455342293, "global_step": 230324, "epoch": 2587, "lr": 5.5773618377689294e-06}
{"train_loss": 0.045942097902297974, "global_step": 230325, "epoch": 2587, "lr": 5.577095760166157e-06}
{"train_loss": 0.025507351383566856, "global_step": 230326, "epoch": 2587, "lr": 5.576829688535512e-06}
{"train_loss": 0.0371079295873642, "global_step": 230327, "epoch": 2587, "lr": 5.5765636228770065e-06}
{"train_loss": 0.056782882660627365, "global_step": 230328, "epoch": 2587, "lr": 5.576297563190708e-06}
{"train_loss": 0.02983887679874897, "global_step": 230329, "epoch": 2587, "lr": 5.57603150947662e-06}
{"train_loss": 0.05867994576692581, "global_step": 230330, "epoch": 2587, "lr": 5.5757654617348e-06}
{"train_loss": 0.05059544166570969, "global_step": 230331, "epoch": 2587, "lr": 5.5754994199652744e-06, "val_loss": 9.3030424118042}
{"train_loss": 0.04901162162423134, "global_step": 230332, "epoch": 2588, "lr": 5.575233384168088e-06}
{"train_loss": 0.07061997801065445, "global_step": 230333, "epoch": 2588, "lr": 5.5749673543432575e-06}
{"train_loss": 0.10393214970827103, "global_step": 230334, "epoch": 2588, "lr": 5.574701330490844e-06}
{"train_loss": 0.03583015874028206, "global_step": 230335, "epoch": 2588, "lr": 5.5744353126108595e-06}
{"train_loss": 0.05483630672097206, "global_step": 230336, "epoch": 2588, "lr": 5.574169300703358e-06}
{"train_loss": 0.04436305910348892, "global_step": 230337, "epoch": 2588, "lr": 5.573903294768363e-06}
{"train_loss": 0.06521130353212357, "global_step": 230338, "epoch": 2588, "lr": 5.5736372948059225e-06}
{"train_loss": 0.03899684175848961, "global_step": 230339, "epoch": 2588, "lr": 5.573371300816055e-06}
{"train_loss": 0.030786743387579918, "global_step": 230340, "epoch": 2588, "lr": 5.5731053127988154e-06}
{"train_loss": 0.0629325732588768, "global_step": 230341, "epoch": 2588, "lr": 5.572839330754226e-06}
{"train_loss": 0.0400552861392498, "global_step": 230342, "epoch": 2588, "lr": 5.572573354682331e-06}
{"train_loss": 0.05243758112192154, "global_step": 230343, "epoch": 2588, "lr": 5.572307384583164e-06}
{"train_loss": 0.03906899318099022, "global_step": 230344, "epoch": 2588, "lr": 5.572041420456747e-06}
{"train_loss": 0.05282839387655258, "global_step": 230345, "epoch": 2588, "lr": 5.571775462303141e-06}
{"train_loss": 0.023416627198457718, "global_step": 230346, "epoch": 2588, "lr": 5.571509510122352e-06}
{"train_loss": 0.05264260247349739, "global_step": 230347, "epoch": 2588, "lr": 5.5712435639144465e-06}
{"train_loss": 0.0668463259935379, "global_step": 230348, "epoch": 2588, "lr": 5.570977623679435e-06}
{"train_loss": 0.036049678921699524, "global_step": 230349, "epoch": 2588, "lr": 5.570711689417379e-06}
{"train_loss": 0.05563388392329216, "global_step": 230350, "epoch": 2588, "lr": 5.570445761128284e-06}
{"train_loss": 0.06650986522436142, "global_step": 230351, "epoch": 2588, "lr": 5.570179838812212e-06}
{"train_loss": 0.0597788542509079, "global_step": 230352, "epoch": 2588, "lr": 5.569913922469178e-06}
{"train_loss": 0.04653928801417351, "global_step": 230353, "epoch": 2588, "lr": 5.569648012099238e-06}
{"train_loss": 0.03710861876606941, "global_step": 230354, "epoch": 2588, "lr": 5.569382107702403e-06}
{"train_loss": 0.024314701557159424, "global_step": 230355, "epoch": 2588, "lr": 5.569116209278729e-06}
{"train_loss": 0.05001768469810486, "global_step": 230356, "epoch": 2588, "lr": 5.56885031682825e-06}
{"train_loss": 0.061861902475357056, "global_step": 230357, "epoch": 2588, "lr": 5.568584430350993e-06}
{"train_loss": 0.05147477239370346, "global_step": 230358, "epoch": 2588, "lr": 5.568318549847007e-06}
{"train_loss": 0.04816717281937599, "global_step": 230359, "epoch": 2588, "lr": 5.56805267531631e-06}
{"train_loss": 0.020411187782883644, "global_step": 230360, "epoch": 2588, "lr": 5.567786806758951e-06}
{"train_loss": 0.04114704579114914, "global_step": 230361, "epoch": 2588, "lr": 5.567520944174959e-06}
{"train_loss": 0.0618840716779232, "global_step": 230362, "epoch": 2588, "lr": 5.567255087564377e-06}
{"train_loss": 0.05288952216506004, "global_step": 230363, "epoch": 2588, "lr": 5.566989236927228e-06}
{"train_loss": 0.031671907752752304, "global_step": 230364, "epoch": 2588, "lr": 5.566723392263568e-06}
{"train_loss": 0.11495890468358994, "global_step": 230365, "epoch": 2588, "lr": 5.566457553573406e-06}
{"train_loss": 0.056294821202754974, "global_step": 230366, "epoch": 2588, "lr": 5.566191720856806e-06}
{"train_loss": 0.057014524936676025, "global_step": 230367, "epoch": 2588, "lr": 5.565925894113777e-06}
{"train_loss": 0.03972558677196503, "global_step": 230368, "epoch": 2588, "lr": 5.565660073344375e-06}
{"train_loss": 0.06108299270272255, "global_step": 230369, "epoch": 2588, "lr": 5.565394258548623e-06}
{"train_loss": 0.05718244984745979, "global_step": 230370, "epoch": 2588, "lr": 5.565128449726575e-06}
{"train_loss": 0.03792791813611984, "global_step": 230371, "epoch": 2588, "lr": 5.564862646878239e-06}
{"train_loss": 0.1085236594080925, "global_step": 230372, "epoch": 2588, "lr": 5.564596850003678e-06}
{"train_loss": 0.025859486311674118, "global_step": 230373, "epoch": 2588, "lr": 5.5643310591029005e-06}
{"train_loss": 0.02488785982131958, "global_step": 230374, "epoch": 2588, "lr": 5.564065274175962e-06}
{"train_loss": 0.03756323829293251, "global_step": 230375, "epoch": 2588, "lr": 5.563799495222905e-06}
{"train_loss": 0.036569248884916306, "global_step": 230376, "epoch": 2588, "lr": 5.563533722243747e-06}
{"train_loss": 0.041669875383377075, "global_step": 230377, "epoch": 2588, "lr": 5.563267955238527e-06}
{"train_loss": 0.023032544180750847, "global_step": 230378, "epoch": 2588, "lr": 5.563002194207289e-06}
{"train_loss": 0.03720533102750778, "global_step": 230379, "epoch": 2588, "lr": 5.562736439150057e-06}
{"train_loss": 0.0409635454416275, "global_step": 230380, "epoch": 2588, "lr": 5.562470690066879e-06}
{"train_loss": 0.08238521218299866, "global_step": 230381, "epoch": 2588, "lr": 5.562204946957783e-06}
{"train_loss": 0.08568491041660309, "global_step": 230382, "epoch": 2588, "lr": 5.561939209822808e-06}
{"train_loss": 0.05903516337275505, "global_step": 230383, "epoch": 2588, "lr": 5.561673478661994e-06}
{"train_loss": 0.033317483961582184, "global_step": 230384, "epoch": 2588, "lr": 5.561407753475362e-06}
{"train_loss": 0.060970671474933624, "global_step": 230385, "epoch": 2588, "lr": 5.561142034262967e-06}
{"train_loss": 0.08340218663215637, "global_step": 230386, "epoch": 2588, "lr": 5.560876321024822e-06}
{"train_loss": 0.02783004194498062, "global_step": 230387, "epoch": 2588, "lr": 5.560610613760986e-06}
{"train_loss": 0.08123980462551117, "global_step": 230388, "epoch": 2588, "lr": 5.5603449124714766e-06}
{"train_loss": 0.07209768891334534, "global_step": 230389, "epoch": 2588, "lr": 5.560079217156344e-06}
{"train_loss": 0.01456356793642044, "global_step": 230390, "epoch": 2588, "lr": 5.559813527815611e-06}
{"train_loss": 0.047970160841941833, "global_step": 230391, "epoch": 2588, "lr": 5.559547844449326e-06}
{"train_loss": 0.049901362508535385, "global_step": 230392, "epoch": 2588, "lr": 5.5592821670575115e-06}
{"train_loss": 0.05082910135388374, "global_step": 230393, "epoch": 2588, "lr": 5.559016495640207e-06}
{"train_loss": 0.02547106146812439, "global_step": 230394, "epoch": 2588, "lr": 5.558750830197463e-06}
{"train_loss": 0.05944472551345825, "global_step": 230395, "epoch": 2588, "lr": 5.558485170729288e-06}
{"train_loss": 0.0531507171690464, "global_step": 230396, "epoch": 2588, "lr": 5.558219517235746e-06}
{"train_loss": 0.04915596917271614, "global_step": 230397, "epoch": 2588, "lr": 5.5579538697168535e-06}
{"train_loss": 0.03553984686732292, "global_step": 230398, "epoch": 2588, "lr": 5.557688228172659e-06}
{"train_loss": 0.0242384634912014, "global_step": 230399, "epoch": 2588, "lr": 5.557422592603184e-06}
{"train_loss": 0.03274921700358391, "global_step": 230400, "epoch": 2588, "lr": 5.557156963008481e-06}
{"train_loss": 0.06997323781251907, "global_step": 230401, "epoch": 2588, "lr": 5.556891339388565e-06}
{"train_loss": 0.018906977027654648, "global_step": 230402, "epoch": 2588, "lr": 5.5566257217434915e-06}
{"train_loss": 0.07341097295284271, "global_step": 230403, "epoch": 2588, "lr": 5.556360110073283e-06}
{"train_loss": 0.04321866109967232, "global_step": 230404, "epoch": 2588, "lr": 5.556094504377985e-06}
{"train_loss": 0.046446770429611206, "global_step": 230405, "epoch": 2588, "lr": 5.555828904657623e-06}
{"train_loss": 0.06278149038553238, "global_step": 230406, "epoch": 2588, "lr": 5.5555633109122425e-06}
{"train_loss": 0.06506039947271347, "global_step": 230407, "epoch": 2588, "lr": 5.555297723141872e-06}
{"train_loss": 0.07115284353494644, "global_step": 230408, "epoch": 2588, "lr": 5.555032141346556e-06}
{"train_loss": 0.08381319791078568, "global_step": 230409, "epoch": 2588, "lr": 5.554766565526314e-06}
{"train_loss": 0.0644015371799469, "global_step": 230410, "epoch": 2588, "lr": 5.554500995681194e-06}
{"train_loss": 0.03666900843381882, "global_step": 230411, "epoch": 2588, "lr": 5.554235431811239e-06}
{"train_loss": 0.10195121169090271, "global_step": 230412, "epoch": 2588, "lr": 5.553969873916465e-06}
{"train_loss": 0.027766311541199684, "global_step": 230413, "epoch": 2588, "lr": 5.553704321996928e-06}
{"train_loss": 0.028345033526420593, "global_step": 230414, "epoch": 2588, "lr": 5.553438776052644e-06}
{"train_loss": 0.03381015360355377, "global_step": 230415, "epoch": 2588, "lr": 5.553173236083664e-06}
{"train_loss": 0.06370735168457031, "global_step": 230416, "epoch": 2588, "lr": 5.552907702090021e-06}
{"train_loss": 0.05201897770166397, "global_step": 230417, "epoch": 2588, "lr": 5.552642174071743e-06}
{"train_loss": 0.04776757210493088, "global_step": 230418, "epoch": 2588, "lr": 5.552376652028874e-06}
{"train_loss": 0.02856465056538582, "global_step": 230419, "epoch": 2588, "lr": 5.552111135961441e-06}
{"train_loss": 0.05081096873357055, "global_step": 230420, "epoch": 2588, "lr": 5.551845625869489e-06, "val_loss": 9.35296630859375}
{"train_loss": 0.06000012904405594, "global_step": 230421, "epoch": 2589, "lr": 5.551580121753053e-06}
{"train_loss": 0.04904952272772789, "global_step": 230422, "epoch": 2589, "lr": 5.551314623612158e-06}
{"train_loss": 0.052073828876018524, "global_step": 230423, "epoch": 2589, "lr": 5.551049131446856e-06}
{"train_loss": 0.0704948678612709, "global_step": 230424, "epoch": 2589, "lr": 5.550783645257157e-06}
{"train_loss": 0.07183844596147537, "global_step": 230425, "epoch": 2589, "lr": 5.550518165043134e-06}
{"train_loss": 0.03788819909095764, "global_step": 230426, "epoch": 2589, "lr": 5.550252690804786e-06}
{"train_loss": 0.056702569127082825, "global_step": 230427, "epoch": 2589, "lr": 5.549987222542175e-06}
{"train_loss": 0.11033539474010468, "global_step": 230428, "epoch": 2589, "lr": 5.549721760255317e-06}
{"train_loss": 0.038961123675107956, "global_step": 230429, "epoch": 2589, "lr": 5.549456303944256e-06}
{"train_loss": 0.07047422975301743, "global_step": 230430, "epoch": 2589, "lr": 5.549190853609038e-06}
{"train_loss": 0.04151514172554016, "global_step": 230431, "epoch": 2589, "lr": 5.548925409249684e-06}
{"train_loss": 0.049820296466350555, "global_step": 230432, "epoch": 2589, "lr": 5.548659970866238e-06}
{"train_loss": 0.06993847340345383, "global_step": 230433, "epoch": 2589, "lr": 5.548394538458729e-06}
{"train_loss": 0.04042193293571472, "global_step": 230434, "epoch": 2589, "lr": 5.548129112027206e-06}
{"train_loss": 0.09426423907279968, "global_step": 230435, "epoch": 2589, "lr": 5.547863691571681e-06}
{"train_loss": 0.045694008469581604, "global_step": 230436, "epoch": 2589, "lr": 5.5475982770922195e-06}
{"train_loss": 0.04256853088736534, "global_step": 230437, "epoch": 2589, "lr": 5.547332868588828e-06}
{"train_loss": 0.07946518063545227, "global_step": 230438, "epoch": 2589, "lr": 5.547067466061568e-06}
{"train_loss": 0.06431909650564194, "global_step": 230439, "epoch": 2589, "lr": 5.546802069510448e-06}
{"train_loss": 0.06568529456853867, "global_step": 230440, "epoch": 2589, "lr": 5.546536678935537e-06}
{"train_loss": 0.05524730309844017, "global_step": 230441, "epoch": 2589, "lr": 5.546271294336835e-06}
{"train_loss": 0.04193105921149254, "global_step": 230442, "epoch": 2589, "lr": 5.546005915714408e-06}
{"train_loss": 0.04712257161736488, "global_step": 230443, "epoch": 2589, "lr": 5.545740543068268e-06}
{"train_loss": 0.04402293637394905, "global_step": 230444, "epoch": 2589, "lr": 5.545475176398473e-06}
{"train_loss": 0.04598582535982132, "global_step": 230445, "epoch": 2589, "lr": 5.5452098157050385e-06}
{"train_loss": 0.055110033601522446, "global_step": 230446, "epoch": 2589, "lr": 5.5449444609880055e-06}
{"train_loss": 0.08983981609344482, "global_step": 230447, "epoch": 2589, "lr": 5.544679112247419e-06}
{"train_loss": 0.0181167870759964, "global_step": 230448, "epoch": 2589, "lr": 5.544413769483309e-06}
{"train_loss": 0.05912364274263382, "global_step": 230449, "epoch": 2589, "lr": 5.544148432695712e-06}
{"train_loss": 0.06425955146551132, "global_step": 230450, "epoch": 2589, "lr": 5.543883101884656e-06}
{"train_loss": 0.01049294788390398, "global_step": 230451, "epoch": 2589, "lr": 5.543617777050192e-06}
{"train_loss": 0.029888300225138664, "global_step": 230452, "epoch": 2589, "lr": 5.543352458192336e-06}
{"train_loss": 0.050505656749010086, "global_step": 230453, "epoch": 2589, "lr": 5.543087145311149e-06}
{"train_loss": 0.05293610692024231, "global_step": 230454, "epoch": 2589, "lr": 5.542821838406647e-06}
{"train_loss": 0.05588355287909508, "global_step": 230455, "epoch": 2589, "lr": 5.54255653747886e-06}
{"train_loss": 0.02494570054113865, "global_step": 230456, "epoch": 2589, "lr": 5.542291242527847e-06}
{"train_loss": 0.06406040489673615, "global_step": 230457, "epoch": 2589, "lr": 5.54202595355362e-06}
{"train_loss": 0.023840976879000664, "global_step": 230458, "epoch": 2589, "lr": 5.54176067055624e-06}
{"train_loss": 0.05033162981271744, "global_step": 230459, "epoch": 2589, "lr": 5.541495393535723e-06}
{"train_loss": 0.06147151067852974, "global_step": 230460, "epoch": 2589, "lr": 5.541230122492103e-06}
{"train_loss": 0.06567566841840744, "global_step": 230461, "epoch": 2589, "lr": 5.540964857425429e-06}
{"train_loss": 0.08641157299280167, "global_step": 230462, "epoch": 2589, "lr": 5.540699598335725e-06}
{"train_loss": 0.05512291193008423, "global_step": 230463, "epoch": 2589, "lr": 5.540434345223039e-06}
{"train_loss": 0.0409354530274868, "global_step": 230464, "epoch": 2589, "lr": 5.540169098087395e-06}
{"train_loss": 0.04323907569050789, "global_step": 230465, "epoch": 2589, "lr": 5.539903856928824e-06}
{"train_loss": 0.048964787274599075, "global_step": 230466, "epoch": 2589, "lr": 5.539638621747389e-06}
{"train_loss": 0.04120325669646263, "global_step": 230467, "epoch": 2589, "lr": 5.539373392543095e-06}
{"train_loss": 0.03015834279358387, "global_step": 230468, "epoch": 2589, "lr": 5.539108169315999e-06}
{"train_loss": 0.07736793160438538, "global_step": 230469, "epoch": 2589, "lr": 5.5388429520661204e-06}
{"train_loss": 0.031209414824843407, "global_step": 230470, "epoch": 2589, "lr": 5.5385777407935106e-06}
{"train_loss": 0.05983379855751991, "global_step": 230471, "epoch": 2589, "lr": 5.538312535498191e-06}
{"train_loss": 0.06095077097415924, "global_step": 230472, "epoch": 2589, "lr": 5.538047336180208e-06}
{"train_loss": 0.050142884254455566, "global_step": 230473, "epoch": 2589, "lr": 5.537782142839587e-06}
{"train_loss": 0.04944249987602234, "global_step": 230474, "epoch": 2589, "lr": 5.5375169554763736e-06}
{"train_loss": 0.06322543323040009, "global_step": 230475, "epoch": 2589, "lr": 5.537251774090596e-06}
{"train_loss": 0.044267550110816956, "global_step": 230476, "epoch": 2589, "lr": 5.536986598682298e-06}
{"train_loss": 0.028248395770788193, "global_step": 230477, "epoch": 2589, "lr": 5.5367214292515014e-06}
{"train_loss": 0.03483578562736511, "global_step": 230478, "epoch": 2589, "lr": 5.536456265798262e-06}
{"train_loss": 0.04007093608379364, "global_step": 230479, "epoch": 2589, "lr": 5.536191108322592e-06}
{"train_loss": 0.05865294858813286, "global_step": 230480, "epoch": 2589, "lr": 5.5359259568245505e-06}
{"train_loss": 0.030975332483649254, "global_step": 230481, "epoch": 2589, "lr": 5.535660811304156e-06}
{"train_loss": 0.09849347174167633, "global_step": 230482, "epoch": 2589, "lr": 5.535395671761445e-06}
{"train_loss": 0.06247366964817047, "global_step": 230483, "epoch": 2589, "lr": 5.535130538196464e-06}
{"train_loss": 0.034407150000333786, "global_step": 230484, "epoch": 2589, "lr": 5.534865410609241e-06}
{"train_loss": 0.04224296286702156, "global_step": 230485, "epoch": 2589, "lr": 5.53460028899982e-06}
{"train_loss": 0.027898933738470078, "global_step": 230486, "epoch": 2589, "lr": 5.5343351733682214e-06}
{"train_loss": 0.035729970782995224, "global_step": 230487, "epoch": 2589, "lr": 5.534070063714497e-06}
{"train_loss": 0.04914281144738197, "global_step": 230488, "epoch": 2589, "lr": 5.533804960038663e-06}
{"train_loss": 0.023724226281046867, "global_step": 230489, "epoch": 2589, "lr": 5.5335398623407804e-06}
{"train_loss": 0.06112249195575714, "global_step": 230490, "epoch": 2589, "lr": 5.533274770620861e-06}
{"train_loss": 0.0628276988863945, "global_step": 230491, "epoch": 2589, "lr": 5.533009684878965e-06}
{"train_loss": 0.051026735454797745, "global_step": 230492, "epoch": 2589, "lr": 5.532744605115109e-06}
{"train_loss": 0.06230810284614563, "global_step": 230493, "epoch": 2589, "lr": 5.532479531329321e-06}
{"train_loss": 0.03234514966607094, "global_step": 230494, "epoch": 2589, "lr": 5.532214463521662e-06}
{"train_loss": 0.054895348846912384, "global_step": 230495, "epoch": 2589, "lr": 5.531949401692149e-06}
{"train_loss": 0.05430658161640167, "global_step": 230496, "epoch": 2589, "lr": 5.531684345840826e-06}
{"train_loss": 0.03644099459052086, "global_step": 230497, "epoch": 2589, "lr": 5.531419295967721e-06}
{"train_loss": 0.04163772612810135, "global_step": 230498, "epoch": 2589, "lr": 5.531154252072884e-06}
{"train_loss": 0.06296814233064651, "global_step": 230499, "epoch": 2589, "lr": 5.530889214156338e-06}
{"train_loss": 0.05598926171660423, "global_step": 230500, "epoch": 2589, "lr": 5.530624182218113e-06}
{"train_loss": 0.06468338519334793, "global_step": 230501, "epoch": 2589, "lr": 5.530359156258258e-06}
{"train_loss": 0.051096200942993164, "global_step": 230502, "epoch": 2589, "lr": 5.530094136276809e-06}
{"train_loss": 0.01914328709244728, "global_step": 230503, "epoch": 2589, "lr": 5.5298291222737885e-06}
{"train_loss": 0.04511455073952675, "global_step": 230504, "epoch": 2589, "lr": 5.529564114249253e-06}
{"train_loss": 0.04651876166462898, "global_step": 230505, "epoch": 2589, "lr": 5.529299112203212e-06}
{"train_loss": 0.038272302597761154, "global_step": 230506, "epoch": 2589, "lr": 5.529034116135728e-06}
{"train_loss": 0.05194146931171417, "global_step": 230507, "epoch": 2589, "lr": 5.5287691260468125e-06}
{"train_loss": 0.04971199110150337, "global_step": 230508, "epoch": 2589, "lr": 5.528504141936519e-06}
{"train_loss": 0.052037043340001886, "global_step": 230509, "epoch": 2589, "lr": 5.528239163804872e-06, "val_loss": 9.065885543823242}
{"train_loss": 0.06332478672266006, "global_step": 230510, "epoch": 2590, "lr": 5.527974191651913e-06}
{"train_loss": 0.03751743957400322, "global_step": 230511, "epoch": 2590, "lr": 5.527709225477673e-06}
{"train_loss": 0.03776078671216965, "global_step": 230512, "epoch": 2590, "lr": 5.5274442652822e-06}
{"train_loss": 0.04300651326775551, "global_step": 230513, "epoch": 2590, "lr": 5.5271793110655055e-06}
{"train_loss": 0.03729286789894104, "global_step": 230514, "epoch": 2590, "lr": 5.526914362827651e-06}
{"train_loss": 0.050627198070287704, "global_step": 230515, "epoch": 2590, "lr": 5.526649420568653e-06}
{"train_loss": 0.06733284145593643, "global_step": 230516, "epoch": 2590, "lr": 5.526384484288566e-06}
{"train_loss": 0.03373735025525093, "global_step": 230517, "epoch": 2590, "lr": 5.526119553987402e-06}
{"train_loss": 0.0706951767206192, "global_step": 230518, "epoch": 2590, "lr": 5.525854629665217e-06}
{"train_loss": 0.0695636123418808, "global_step": 230519, "epoch": 2590, "lr": 5.525589711322032e-06}
{"train_loss": 0.039929576218128204, "global_step": 230520, "epoch": 2590, "lr": 5.525324798957887e-06}
{"train_loss": 0.040273480117321014, "global_step": 230521, "epoch": 2590, "lr": 5.525059892572831e-06}
{"train_loss": 0.050102751702070236, "global_step": 230522, "epoch": 2590, "lr": 5.524794992166882e-06}
{"train_loss": 0.057761430740356445, "global_step": 230523, "epoch": 2590, "lr": 5.524530097740088e-06}
{"train_loss": 0.09080776572227478, "global_step": 230524, "epoch": 2590, "lr": 5.524265209292473e-06}
{"train_loss": 0.05991920456290245, "global_step": 230525, "epoch": 2590, "lr": 5.524000326824086e-06}
{"train_loss": 0.015251802280545235, "global_step": 230526, "epoch": 2590, "lr": 5.523735450334944e-06}
{"train_loss": 0.056081634014844894, "global_step": 230527, "epoch": 2590, "lr": 5.523470579825102e-06}
{"train_loss": 0.05475233495235443, "global_step": 230528, "epoch": 2590, "lr": 5.523205715294582e-06}
{"train_loss": 0.025450222194194794, "global_step": 230529, "epoch": 2590, "lr": 5.522940856743431e-06}
{"train_loss": 0.06278207153081894, "global_step": 230530, "epoch": 2590, "lr": 5.522676004171673e-06}
{"train_loss": 0.0582265704870224, "global_step": 230531, "epoch": 2590, "lr": 5.522411157579355e-06}
{"train_loss": 0.07988841831684113, "global_step": 230532, "epoch": 2590, "lr": 5.5221463169665085e-06}
{"train_loss": 0.061730023473501205, "global_step": 230533, "epoch": 2590, "lr": 5.5218814823331575e-06}
{"train_loss": 0.05340650677680969, "global_step": 230534, "epoch": 2590, "lr": 5.521616653679357e-06}
{"train_loss": 0.04574362933635712, "global_step": 230535, "epoch": 2590, "lr": 5.5213518310051225e-06}
{"train_loss": 0.022744152694940567, "global_step": 230536, "epoch": 2590, "lr": 5.5210870143105105e-06}
{"train_loss": 0.022992363199591637, "global_step": 230537, "epoch": 2590, "lr": 5.520822203595538e-06}
{"train_loss": 0.060000140219926834, "global_step": 230538, "epoch": 2590, "lr": 5.520557398860254e-06}
{"train_loss": 0.028860852122306824, "global_step": 230539, "epoch": 2590, "lr": 5.520292600104682e-06}
{"train_loss": 0.014483421109616756, "global_step": 230540, "epoch": 2590, "lr": 5.520027807328876e-06}
{"train_loss": 0.06264311820268631, "global_step": 230541, "epoch": 2590, "lr": 5.5197630205328474e-06}
{"train_loss": 0.07868972420692444, "global_step": 230542, "epoch": 2590, "lr": 5.519498239716658e-06}
{"train_loss": 0.04396034777164459, "global_step": 230543, "epoch": 2590, "lr": 5.519233464880314e-06}
{"train_loss": 0.03715400770306587, "global_step": 230544, "epoch": 2590, "lr": 5.5189686960238795e-06}
{"train_loss": 0.08281183987855911, "global_step": 230545, "epoch": 2590, "lr": 5.5187039331473735e-06}
{"train_loss": 0.02947855554521084, "global_step": 230546, "epoch": 2590, "lr": 5.518439176250839e-06}
{"train_loss": 0.05125788226723671, "global_step": 230547, "epoch": 2590, "lr": 5.518174425334299e-06}
{"train_loss": 0.02377704717218876, "global_step": 230548, "epoch": 2590, "lr": 5.5179096803978096e-06}
{"train_loss": 0.07401219755411148, "global_step": 230549, "epoch": 2590, "lr": 5.517644941441385e-06}
{"train_loss": 0.04530090093612671, "global_step": 230550, "epoch": 2590, "lr": 5.517380208465078e-06}
{"train_loss": 0.043773990124464035, "global_step": 230551, "epoch": 2590, "lr": 5.51711548146891e-06}
{"train_loss": 0.030651869252324104, "global_step": 230552, "epoch": 2590, "lr": 5.516850760452935e-06}
{"train_loss": 0.026088248938322067, "global_step": 230553, "epoch": 2590, "lr": 5.516586045417166e-06}
{"train_loss": 0.042974796146154404, "global_step": 230554, "epoch": 2590, "lr": 5.516321336361657e-06}
{"train_loss": 0.03949926048517227, "global_step": 230555, "epoch": 2590, "lr": 5.516056633286432e-06}
{"train_loss": 0.05546824261546135, "global_step": 230556, "epoch": 2590, "lr": 5.515791936191528e-06}
{"train_loss": 0.048021331429481506, "global_step": 230557, "epoch": 2590, "lr": 5.51552724507699e-06}
{"train_loss": 0.050430722534656525, "global_step": 230558, "epoch": 2590, "lr": 5.515262559942841e-06}
{"train_loss": 0.038484904915094376, "global_step": 230559, "epoch": 2590, "lr": 5.514997880789135e-06}
{"train_loss": 0.06256379187107086, "global_step": 230560, "epoch": 2590, "lr": 5.51473320761588e-06}
{"train_loss": 0.05162883922457695, "global_step": 230561, "epoch": 2590, "lr": 5.51446854042314e-06}
{"train_loss": 0.0883270651102066, "global_step": 230562, "epoch": 2590, "lr": 5.514203879210927e-06}
{"train_loss": 0.044694047421216965, "global_step": 230563, "epoch": 2590, "lr": 5.5139392239792975e-06}
{"train_loss": 0.023346275091171265, "global_step": 230564, "epoch": 2590, "lr": 5.513674574728267e-06}
{"train_loss": 0.05724528804421425, "global_step": 230565, "epoch": 2590, "lr": 5.513409931457891e-06}
{"train_loss": 0.016620058566331863, "global_step": 230566, "epoch": 2590, "lr": 5.513145294168182e-06}
{"train_loss": 0.017019234597682953, "global_step": 230567, "epoch": 2590, "lr": 5.512880662859199e-06}
{"train_loss": 0.03316561132669449, "global_step": 230568, "epoch": 2590, "lr": 5.51261603753096e-06}
{"train_loss": 0.05184948071837425, "global_step": 230569, "epoch": 2590, "lr": 5.5123514181835145e-06}
{"train_loss": 0.08638407289981842, "global_step": 230570, "epoch": 2590, "lr": 5.5120868048168965e-06}
{"train_loss": 0.025954831391572952, "global_step": 230571, "epoch": 2590, "lr": 5.511822197431121e-06}
{"train_loss": 0.028006168082356453, "global_step": 230572, "epoch": 2590, "lr": 5.511557596026246e-06}
{"train_loss": 0.07026373594999313, "global_step": 230573, "epoch": 2590, "lr": 5.511293000602296e-06}
{"train_loss": 0.024319913238286972, "global_step": 230574, "epoch": 2590, "lr": 5.511028411159319e-06}
{"train_loss": 0.10613663494586945, "global_step": 230575, "epoch": 2590, "lr": 5.510763827697329e-06}
{"train_loss": 0.08143114298582077, "global_step": 230576, "epoch": 2590, "lr": 5.510499250216389e-06}
{"train_loss": 0.056475527584552765, "global_step": 230577, "epoch": 2590, "lr": 5.5102346787165086e-06}
{"train_loss": 0.04876626655459404, "global_step": 230578, "epoch": 2590, "lr": 5.509970113197743e-06}
{"train_loss": 0.04554157704114914, "global_step": 230579, "epoch": 2590, "lr": 5.509705553660111e-06}
{"train_loss": 0.07594701647758484, "global_step": 230580, "epoch": 2590, "lr": 5.509441000103666e-06}
{"train_loss": 0.03939417004585266, "global_step": 230581, "epoch": 2590, "lr": 5.509176452528425e-06}
{"train_loss": 0.03169311583042145, "global_step": 230582, "epoch": 2590, "lr": 5.508911910934444e-06}
{"train_loss": 0.030831873416900635, "global_step": 230583, "epoch": 2590, "lr": 5.508647375321735e-06}
{"train_loss": 0.03366881608963013, "global_step": 230584, "epoch": 2590, "lr": 5.508382845690357e-06}
{"train_loss": 0.03837623819708824, "global_step": 230585, "epoch": 2590, "lr": 5.508118322040323e-06}
{"train_loss": 0.044259730726480484, "global_step": 230586, "epoch": 2590, "lr": 5.507853804371693e-06}
{"train_loss": 0.07910149544477463, "global_step": 230587, "epoch": 2590, "lr": 5.507589292684478e-06}
{"train_loss": 0.025328245013952255, "global_step": 230588, "epoch": 2590, "lr": 5.50732478697874e-06}
{"train_loss": 0.05440301448106766, "global_step": 230589, "epoch": 2590, "lr": 5.507060287254484e-06}
{"train_loss": 0.016919303685426712, "global_step": 230590, "epoch": 2590, "lr": 5.506795793511776e-06}
{"train_loss": 0.03359321132302284, "global_step": 230591, "epoch": 2590, "lr": 5.506531305750623e-06}
{"train_loss": 0.052918072789907455, "global_step": 230592, "epoch": 2590, "lr": 5.506266823971079e-06}
{"train_loss": 0.048970531672239304, "global_step": 230593, "epoch": 2590, "lr": 5.506002348173178e-06}
{"train_loss": 0.058411117643117905, "global_step": 230594, "epoch": 2590, "lr": 5.505737878356948e-06}
{"train_loss": 0.04400603100657463, "global_step": 230595, "epoch": 2590, "lr": 5.5054734145224385e-06}
{"train_loss": 0.04350559785962105, "global_step": 230596, "epoch": 2590, "lr": 5.505208956669666e-06}
{"train_loss": 0.05811057239770889, "global_step": 230597, "epoch": 2590, "lr": 5.504944504798687e-06}
{"train_loss": 0.048214164123022825, "global_step": 230598, "epoch": 2590, "lr": 5.504680058909517e-06, "val_loss": 9.147789001464844, "train_action_mse_error": 14.36018180847168}
{"train_loss": 0.034512631595134735, "global_step": 230599, "epoch": 2591, "lr": 5.504415619002207e-06}
{"train_loss": 0.07208933681249619, "global_step": 230600, "epoch": 2591, "lr": 5.504151185076778e-06}
{"train_loss": 0.025311777368187904, "global_step": 230601, "epoch": 2591, "lr": 5.503886757133281e-06}
{"train_loss": 0.046236298978328705, "global_step": 230602, "epoch": 2591, "lr": 5.503622335171732e-06}
{"train_loss": 0.024810662493109703, "global_step": 230603, "epoch": 2591, "lr": 5.503357919192193e-06}
{"train_loss": 0.046334754675626755, "global_step": 230604, "epoch": 2591, "lr": 5.5030935091946735e-06}
{"train_loss": 0.06571487337350845, "global_step": 230605, "epoch": 2591, "lr": 5.502829105179236e-06}
{"train_loss": 0.08186176419258118, "global_step": 230606, "epoch": 2591, "lr": 5.502564707145885e-06}
{"train_loss": 0.05154780298471451, "global_step": 230607, "epoch": 2591, "lr": 5.502300315094682e-06}
{"train_loss": 0.08868535608053207, "global_step": 230608, "epoch": 2591, "lr": 5.502035929025656e-06}
{"train_loss": 0.04739910364151001, "global_step": 230609, "epoch": 2591, "lr": 5.501771548938822e-06}
{"train_loss": 0.02569190040230751, "global_step": 230610, "epoch": 2591, "lr": 5.5015071748342476e-06}
{"train_loss": 0.07880879193544388, "global_step": 230611, "epoch": 2591, "lr": 5.501242806711943e-06}
{"train_loss": 0.06073769927024841, "global_step": 230612, "epoch": 2591, "lr": 5.500978444571964e-06}
{"train_loss": 0.08436502516269684, "global_step": 230613, "epoch": 2591, "lr": 5.500714088414327e-06}
{"train_loss": 0.07514118403196335, "global_step": 230614, "epoch": 2591, "lr": 5.500449738239083e-06}
{"train_loss": 0.04562186449766159, "global_step": 230615, "epoch": 2591, "lr": 5.5001853940462535e-06}
{"train_loss": 0.04989801347255707, "global_step": 230616, "epoch": 2591, "lr": 5.499921055835894e-06}
{"train_loss": 0.07033372670412064, "global_step": 230617, "epoch": 2591, "lr": 5.499656723608015e-06}
{"train_loss": 0.044300377368927, "global_step": 230618, "epoch": 2591, "lr": 5.499392397362674e-06}
{"train_loss": 0.06730987876653671, "global_step": 230619, "epoch": 2591, "lr": 5.499128077099891e-06}
{"train_loss": 0.045985233038663864, "global_step": 230620, "epoch": 2591, "lr": 5.4988637628197115e-06}
{"train_loss": 0.026943858712911606, "global_step": 230621, "epoch": 2591, "lr": 5.498599454522163e-06}
{"train_loss": 0.03745711222290993, "global_step": 230622, "epoch": 2591, "lr": 5.498335152207296e-06}
{"train_loss": 0.12218734622001648, "global_step": 230623, "epoch": 2591, "lr": 5.49807085587512e-06}
{"train_loss": 0.053246404975652695, "global_step": 230624, "epoch": 2591, "lr": 5.497806565525698e-06}
{"train_loss": 0.12447331100702286, "global_step": 230625, "epoch": 2591, "lr": 5.497542281159046e-06}
{"train_loss": 0.032634198665618896, "global_step": 230626, "epoch": 2591, "lr": 5.497278002775213e-06}
{"train_loss": 0.10275726765394211, "global_step": 230627, "epoch": 2591, "lr": 5.497013730374223e-06}
{"train_loss": 0.03777911514043808, "global_step": 230628, "epoch": 2591, "lr": 5.496749463956114e-06}
{"train_loss": 0.016119400039315224, "global_step": 230629, "epoch": 2591, "lr": 5.496485203520935e-06}
{"train_loss": 0.03496626019477844, "global_step": 230630, "epoch": 2591, "lr": 5.496220949068703e-06}
{"train_loss": 0.028678175061941147, "global_step": 230631, "epoch": 2591, "lr": 5.495956700599469e-06}
{"train_loss": 0.041991930454969406, "global_step": 230632, "epoch": 2591, "lr": 5.495692458113255e-06}
{"train_loss": 0.05212464928627014, "global_step": 230633, "epoch": 2591, "lr": 5.495428221610111e-06}
{"train_loss": 0.04333692044019699, "global_step": 230634, "epoch": 2591, "lr": 5.495163991090052e-06}
{"train_loss": 0.050060901790857315, "global_step": 230635, "epoch": 2591, "lr": 5.494899766553135e-06}
{"train_loss": 0.043764740228652954, "global_step": 230636, "epoch": 2591, "lr": 5.494635547999383e-06}
{"train_loss": 0.0651254653930664, "global_step": 230637, "epoch": 2591, "lr": 5.4943713354288386e-06}
{"train_loss": 0.03477449715137482, "global_step": 230638, "epoch": 2591, "lr": 5.494107128841525e-06}
{"train_loss": 0.03250875324010849, "global_step": 230639, "epoch": 2591, "lr": 5.493842928237497e-06}
{"train_loss": 0.049032069742679596, "global_step": 230640, "epoch": 2591, "lr": 5.493578733616772e-06}
{"train_loss": 0.07185689359903336, "global_step": 230641, "epoch": 2591, "lr": 5.4933145449794e-06}
{"train_loss": 0.048808224499225616, "global_step": 230642, "epoch": 2591, "lr": 5.493050362325397e-06}
{"train_loss": 0.04515000060200691, "global_step": 230643, "epoch": 2591, "lr": 5.4927861856548246e-06}
{"train_loss": 0.012794732116162777, "global_step": 230644, "epoch": 2591, "lr": 5.492522014967694e-06}
{"train_loss": 0.06658093631267548, "global_step": 230645, "epoch": 2591, "lr": 5.49225785026406e-06}
{"train_loss": 0.038680218160152435, "global_step": 230646, "epoch": 2591, "lr": 5.491993691543939e-06}
{"train_loss": 0.0587991438806057, "global_step": 230647, "epoch": 2591, "lr": 5.491729538807389e-06}
{"train_loss": 0.0821494311094284, "global_step": 230648, "epoch": 2591, "lr": 5.491465392054429e-06}
{"train_loss": 0.06001686677336693, "global_step": 230649, "epoch": 2591, "lr": 5.491201251285094e-06}
{"train_loss": 0.05345156416296959, "global_step": 230650, "epoch": 2591, "lr": 5.490937116499434e-06}
{"train_loss": 0.08226470649242401, "global_step": 230651, "epoch": 2591, "lr": 5.490672987697465e-06}
{"train_loss": 0.0778009369969368, "global_step": 230652, "epoch": 2591, "lr": 5.490408864879243e-06}
{"train_loss": 0.04933909699320793, "global_step": 230653, "epoch": 2591, "lr": 5.490144748044779e-06}
{"train_loss": 0.034377194941043854, "global_step": 230654, "epoch": 2591, "lr": 5.489880637194134e-06}
{"train_loss": 0.03543700650334358, "global_step": 230655, "epoch": 2591, "lr": 5.489616532327324e-06}
{"train_loss": 0.05538484454154968, "global_step": 230656, "epoch": 2591, "lr": 5.4893524334444e-06}
{"train_loss": 0.02719222940504551, "global_step": 230657, "epoch": 2591, "lr": 5.489088340545384e-06}
{"train_loss": 0.044189099222421646, "global_step": 230658, "epoch": 2591, "lr": 5.488824253630326e-06}
{"train_loss": 0.020906006917357445, "global_step": 230659, "epoch": 2591, "lr": 5.488560172699242e-06}
{"train_loss": 0.036519601941108704, "global_step": 230660, "epoch": 2591, "lr": 5.488296097752188e-06}
{"train_loss": 0.019187087193131447, "global_step": 230661, "epoch": 2591, "lr": 5.488032028789181e-06}
{"train_loss": 0.03609517216682434, "global_step": 230662, "epoch": 2591, "lr": 5.487767965810275e-06}
{"train_loss": 0.049129050225019455, "global_step": 230663, "epoch": 2591, "lr": 5.487503908815489e-06}
{"train_loss": 0.04408036544919014, "global_step": 230664, "epoch": 2591, "lr": 5.48723985780486e-06}
{"train_loss": 0.017871271818876266, "global_step": 230665, "epoch": 2591, "lr": 5.4869758127784445e-06}
{"train_loss": 0.03082660399377346, "global_step": 230666, "epoch": 2591, "lr": 5.486711773736253e-06}
{"train_loss": 0.041430920362472534, "global_step": 230667, "epoch": 2591, "lr": 5.486447740678336e-06}
{"train_loss": 0.03697289153933525, "global_step": 230668, "epoch": 2591, "lr": 5.486183713604714e-06}
{"train_loss": 0.04828977957367897, "global_step": 230669, "epoch": 2591, "lr": 5.485919692515445e-06}
{"train_loss": 0.039928797632455826, "global_step": 230670, "epoch": 2591, "lr": 5.48565567741054e-06}
{"train_loss": 0.04221521317958832, "global_step": 230671, "epoch": 2591, "lr": 5.485391668290058e-06}
{"train_loss": 0.0329841747879982, "global_step": 230672, "epoch": 2591, "lr": 5.485127665154011e-06}
{"train_loss": 0.04610263556241989, "global_step": 230673, "epoch": 2591, "lr": 5.484863668002454e-06}
{"train_loss": 0.09617640823125839, "global_step": 230674, "epoch": 2591, "lr": 5.484599676835406e-06}
{"train_loss": 0.028041694313287735, "global_step": 230675, "epoch": 2591, "lr": 5.48433569165292e-06}
{"train_loss": 0.023978354409337044, "global_step": 230676, "epoch": 2591, "lr": 5.484071712455013e-06}
{"train_loss": 0.0363108366727829, "global_step": 230677, "epoch": 2591, "lr": 5.483807739241742e-06}
{"train_loss": 0.03611218184232712, "global_step": 230678, "epoch": 2591, "lr": 5.483543772013116e-06}
{"train_loss": 0.09339214861392975, "global_step": 230679, "epoch": 2591, "lr": 5.483279810769204e-06}
{"train_loss": 0.04541055113077164, "global_step": 230680, "epoch": 2591, "lr": 5.483015855510004e-06}
{"train_loss": 0.04260234534740448, "global_step": 230681, "epoch": 2591, "lr": 5.482751906235584e-06}
{"train_loss": 0.07281278073787689, "global_step": 230682, "epoch": 2591, "lr": 5.4824879629459535e-06}
{"train_loss": 0.07361277937889099, "global_step": 230683, "epoch": 2591, "lr": 5.482224025641169e-06}
{"train_loss": 0.08381035178899765, "global_step": 230684, "epoch": 2591, "lr": 5.481960094321248e-06}
{"train_loss": 0.03149835020303726, "global_step": 230685, "epoch": 2591, "lr": 5.4816961689862445e-06}
{"train_loss": 0.03097296506166458, "global_step": 230686, "epoch": 2591, "lr": 5.481432249636187e-06}
{"train_loss": 0.05022052933953786, "global_step": 230687, "epoch": 2591, "lr": 5.481168336271098e-06, "val_loss": 9.025066375732422}
{"train_loss": 0.07664983719587326, "global_step": 230688, "epoch": 2592, "lr": 5.480904428891026e-06}
{"train_loss": 0.06456904113292694, "global_step": 230689, "epoch": 2592, "lr": 5.480640527496e-06}
{"train_loss": 0.02788795903325081, "global_step": 230690, "epoch": 2592, "lr": 5.48037663208607e-06}
{"train_loss": 0.037839993834495544, "global_step": 230691, "epoch": 2592, "lr": 5.480112742661247e-06}
{"train_loss": 0.04236854985356331, "global_step": 230692, "epoch": 2592, "lr": 5.479848859221598e-06}
{"train_loss": 0.10166674852371216, "global_step": 230693, "epoch": 2592, "lr": 5.479584981767122e-06}
{"train_loss": 0.04571715369820595, "global_step": 230694, "epoch": 2592, "lr": 5.479321110297886e-06}
{"train_loss": 0.07102193683385849, "global_step": 230695, "epoch": 2592, "lr": 5.4790572448139064e-06}
{"train_loss": 0.03562682867050171, "global_step": 230696, "epoch": 2592, "lr": 5.478793385315228e-06}
{"train_loss": 0.05775997415184975, "global_step": 230697, "epoch": 2592, "lr": 5.478529531801879e-06}
{"train_loss": 0.031243745237588882, "global_step": 230698, "epoch": 2592, "lr": 5.478265684273909e-06}
{"train_loss": 0.046292319893836975, "global_step": 230699, "epoch": 2592, "lr": 5.478001842731329e-06}
{"train_loss": 0.06726589053869247, "global_step": 230700, "epoch": 2592, "lr": 5.477738007174194e-06}
{"train_loss": 0.029992371797561646, "global_step": 230701, "epoch": 2592, "lr": 5.477474177602537e-06}
{"train_loss": 0.03705739974975586, "global_step": 230702, "epoch": 2592, "lr": 5.4772103540163875e-06}
{"train_loss": 0.047813721001148224, "global_step": 230703, "epoch": 2592, "lr": 5.476946536415794e-06}
{"train_loss": 0.03592338413000107, "global_step": 230704, "epoch": 2592, "lr": 5.476682724800769e-06}
{"train_loss": 0.04703964665532112, "global_step": 230705, "epoch": 2592, "lr": 5.476418919171372e-06}
{"train_loss": 0.042569927871227264, "global_step": 230706, "epoch": 2592, "lr": 5.476155119527621e-06}
{"train_loss": 0.07442724704742432, "global_step": 230707, "epoch": 2592, "lr": 5.47589132586957e-06}
{"train_loss": 0.053210847079753876, "global_step": 230708, "epoch": 2592, "lr": 5.475627538197225e-06}
{"train_loss": 0.051120441406965256, "global_step": 230709, "epoch": 2592, "lr": 5.4753637565106545e-06}
{"train_loss": 0.09386197477579117, "global_step": 230710, "epoch": 2592, "lr": 5.475099980809867e-06}
{"train_loss": 0.05223900452256203, "global_step": 230711, "epoch": 2592, "lr": 5.474836211094925e-06}
{"train_loss": 0.05637376755475998, "global_step": 230712, "epoch": 2592, "lr": 5.474572447365833e-06}
{"train_loss": 0.06272660195827484, "global_step": 230713, "epoch": 2592, "lr": 5.474308689622648e-06}
{"train_loss": 0.012377441860735416, "global_step": 230714, "epoch": 2592, "lr": 5.474044937865397e-06}
{"train_loss": 0.02794746123254299, "global_step": 230715, "epoch": 2592, "lr": 5.473781192094124e-06}
{"train_loss": 0.07430823147296906, "global_step": 230716, "epoch": 2592, "lr": 5.4735174523088515e-06}
{"train_loss": 0.02146131545305252, "global_step": 230717, "epoch": 2592, "lr": 5.47325371850963e-06}
{"train_loss": 0.040347516536712646, "global_step": 230718, "epoch": 2592, "lr": 5.472989990696476e-06}
{"train_loss": 0.03313199430704117, "global_step": 230719, "epoch": 2592, "lr": 5.472726268869449e-06}
{"train_loss": 0.05207454413175583, "global_step": 230720, "epoch": 2592, "lr": 5.4724625530285566e-06}
{"train_loss": 0.05163713917136192, "global_step": 230721, "epoch": 2592, "lr": 5.4721988431738646e-06}
{"train_loss": 0.004139839671552181, "global_step": 230722, "epoch": 2592, "lr": 5.471935139305379e-06}
{"train_loss": 0.01949496380984783, "global_step": 230723, "epoch": 2592, "lr": 5.4716714414231596e-06}
{"train_loss": 0.03799479082226753, "global_step": 230724, "epoch": 2592, "lr": 5.471407749527219e-06}
{"train_loss": 0.04449162259697914, "global_step": 230725, "epoch": 2592, "lr": 5.471144063617617e-06}
{"train_loss": 0.025140050798654556, "global_step": 230726, "epoch": 2592, "lr": 5.4708803836943765e-06}
{"train_loss": 0.06944719702005386, "global_step": 230727, "epoch": 2592, "lr": 5.470616709757525e-06}
{"train_loss": 0.0655122920870781, "global_step": 230728, "epoch": 2592, "lr": 5.4703530418071135e-06}
{"train_loss": 0.029446976259350777, "global_step": 230729, "epoch": 2592, "lr": 5.470089379843163e-06}
{"train_loss": 0.053675420582294464, "global_step": 230730, "epoch": 2592, "lr": 5.4698257238657235e-06}
{"train_loss": 0.04490148276090622, "global_step": 230731, "epoch": 2592, "lr": 5.469562073874817e-06}
{"train_loss": 0.03074430301785469, "global_step": 230732, "epoch": 2592, "lr": 5.469298429870495e-06}
{"train_loss": 0.0295546967536211, "global_step": 230733, "epoch": 2592, "lr": 5.4690347918527675e-06}
{"train_loss": 0.030956245958805084, "global_step": 230734, "epoch": 2592, "lr": 5.468771159821701e-06}
{"train_loss": 0.06174169108271599, "global_step": 230735, "epoch": 2592, "lr": 5.468507533777306e-06}
{"train_loss": 0.09452643990516663, "global_step": 230736, "epoch": 2592, "lr": 5.468243913719623e-06}
{"train_loss": 0.04119027405977249, "global_step": 230737, "epoch": 2592, "lr": 5.467980299648706e-06}
{"train_loss": 0.03474542498588562, "global_step": 230738, "epoch": 2592, "lr": 5.467716691564562e-06}
{"train_loss": 0.04785250499844551, "global_step": 230739, "epoch": 2592, "lr": 5.467453089467251e-06}
{"train_loss": 0.06377226859331131, "global_step": 230740, "epoch": 2592, "lr": 5.467189493356789e-06}
{"train_loss": 0.014888095669448376, "global_step": 230741, "epoch": 2592, "lr": 5.466925903233233e-06}
{"train_loss": 0.027308423072099686, "global_step": 230742, "epoch": 2592, "lr": 5.466662319096594e-06}
{"train_loss": 0.026668425649404526, "global_step": 230743, "epoch": 2592, "lr": 5.466398740946932e-06}
{"train_loss": 0.044917311519384384, "global_step": 230744, "epoch": 2592, "lr": 5.466135168784259e-06}
{"train_loss": 0.041004180908203125, "global_step": 230745, "epoch": 2592, "lr": 5.46587160260863e-06}
{"train_loss": 0.019698135554790497, "global_step": 230746, "epoch": 2592, "lr": 5.465608042420061e-06}
{"train_loss": 0.11953304708003998, "global_step": 230747, "epoch": 2592, "lr": 5.46534448821861e-06}
{"train_loss": 0.04232979565858841, "global_step": 230748, "epoch": 2592, "lr": 5.465080940004291e-06}
{"train_loss": 0.08351524919271469, "global_step": 230749, "epoch": 2592, "lr": 5.464817397777155e-06}
{"train_loss": 0.0640006810426712, "global_step": 230750, "epoch": 2592, "lr": 5.464553861537225e-06}
{"train_loss": 0.051835909485816956, "global_step": 230751, "epoch": 2592, "lr": 5.464290331284555e-06}
{"train_loss": 0.025638118386268616, "global_step": 230752, "epoch": 2592, "lr": 5.4640268070191566e-06}
{"train_loss": 0.04848352074623108, "global_step": 230753, "epoch": 2592, "lr": 5.463763288741086e-06}
{"train_loss": 0.03249155357480049, "global_step": 230754, "epoch": 2592, "lr": 5.463499776450359e-06}
{"train_loss": 0.03507503122091293, "global_step": 230755, "epoch": 2592, "lr": 5.4632362701470205e-06}
{"train_loss": 0.020028384402394295, "global_step": 230756, "epoch": 2592, "lr": 5.46297276983112e-06}
{"train_loss": 0.08137974143028259, "global_step": 230757, "epoch": 2592, "lr": 5.46270927550267e-06}
{"train_loss": 0.05253042280673981, "global_step": 230758, "epoch": 2592, "lr": 5.462445787161724e-06}
{"train_loss": 0.13808010518550873, "global_step": 230759, "epoch": 2592, "lr": 5.462182304808311e-06}
{"train_loss": 0.05868319422006607, "global_step": 230760, "epoch": 2592, "lr": 5.461918828442453e-06}
{"train_loss": 0.05174660310149193, "global_step": 230761, "epoch": 2592, "lr": 5.461655358064205e-06}
{"train_loss": 0.08851148188114166, "global_step": 230762, "epoch": 2592, "lr": 5.4613918936735895e-06}
{"train_loss": 0.05883412063121796, "global_step": 230763, "epoch": 2592, "lr": 5.461128435270657e-06}
{"train_loss": 0.04230165109038353, "global_step": 230764, "epoch": 2592, "lr": 5.4608649828554295e-06}
{"train_loss": 0.12377120554447174, "global_step": 230765, "epoch": 2592, "lr": 5.460601536427939e-06}
{"train_loss": 0.046132393181324005, "global_step": 230766, "epoch": 2592, "lr": 5.4603380959882374e-06}
{"train_loss": 0.057064708322286606, "global_step": 230767, "epoch": 2592, "lr": 5.460074661536335e-06}
{"train_loss": 0.053010713309049606, "global_step": 230768, "epoch": 2592, "lr": 5.459811233072299e-06}
{"train_loss": 0.08632750064134598, "global_step": 230769, "epoch": 2592, "lr": 5.459547810596139e-06}
{"train_loss": 0.08847099542617798, "global_step": 230770, "epoch": 2592, "lr": 5.4592843941079055e-06}
{"train_loss": 0.04294675961136818, "global_step": 230771, "epoch": 2592, "lr": 5.459020983607621e-06}
{"train_loss": 0.07617901265621185, "global_step": 230772, "epoch": 2592, "lr": 5.458757579095336e-06}
{"train_loss": 0.033499833196401596, "global_step": 230773, "epoch": 2592, "lr": 5.458494180571072e-06}
{"train_loss": 0.07668394595384598, "global_step": 230774, "epoch": 2592, "lr": 5.458230788034868e-06}
{"train_loss": 0.03705025836825371, "global_step": 230775, "epoch": 2592, "lr": 5.4579674014867686e-06}
{"train_loss": 0.05132401312879297, "global_step": 230776, "epoch": 2592, "lr": 5.457704020926796e-06, "val_loss": 9.109003067016602}
{"train_loss": 0.046254970133304596, "global_step": 230777, "epoch": 2593, "lr": 5.457440646355006e-06}
{"train_loss": 0.09145735949277878, "global_step": 230778, "epoch": 2593, "lr": 5.457177277771403e-06}
{"train_loss": 0.05192531645298004, "global_step": 230779, "epoch": 2593, "lr": 5.456913915176054e-06}
{"train_loss": 0.03537120297551155, "global_step": 230780, "epoch": 2593, "lr": 5.456650558568971e-06}
{"train_loss": 0.03513723239302635, "global_step": 230781, "epoch": 2593, "lr": 5.456387207950209e-06}
{"train_loss": 0.07495718449354172, "global_step": 230782, "epoch": 2593, "lr": 5.456123863319779e-06}
{"train_loss": 0.03300024941563606, "global_step": 230783, "epoch": 2593, "lr": 5.455860524677742e-06}
{"train_loss": 0.07788190245628357, "global_step": 230784, "epoch": 2593, "lr": 5.455597192024109e-06}
{"train_loss": 0.06339262425899506, "global_step": 230785, "epoch": 2593, "lr": 5.455333865358942e-06}
{"train_loss": 0.038428764790296555, "global_step": 230786, "epoch": 2593, "lr": 5.455070544682251e-06}
{"train_loss": 0.04929443821310997, "global_step": 230787, "epoch": 2593, "lr": 5.454807229994091e-06}
{"train_loss": 0.10352402925491333, "global_step": 230788, "epoch": 2593, "lr": 5.454543921294481e-06}
{"train_loss": 0.01924876868724823, "global_step": 230789, "epoch": 2593, "lr": 5.454280618583474e-06}
{"train_loss": 0.027376806363463402, "global_step": 230790, "epoch": 2593, "lr": 5.454017321861088e-06}
{"train_loss": 0.07349246740341187, "global_step": 230791, "epoch": 2593, "lr": 5.4537540311273675e-06}
{"train_loss": 0.048984043300151825, "global_step": 230792, "epoch": 2593, "lr": 5.45349074638235e-06}
{"train_loss": 0.06642769277095795, "global_step": 230793, "epoch": 2593, "lr": 5.453227467626065e-06}
{"train_loss": 0.05017950385808945, "global_step": 230794, "epoch": 2593, "lr": 5.4529641948585565e-06}
{"train_loss": 0.057040102779865265, "global_step": 230795, "epoch": 2593, "lr": 5.452700928079846e-06}
{"train_loss": 0.045079007744789124, "global_step": 230796, "epoch": 2593, "lr": 5.4524376672899835e-06}
{"train_loss": 0.035767536610364914, "global_step": 230797, "epoch": 2593, "lr": 5.452174412489003e-06}
{"train_loss": 0.061176691204309464, "global_step": 230798, "epoch": 2593, "lr": 5.45191116367692e-06}
{"train_loss": 0.017336146906018257, "global_step": 230799, "epoch": 2593, "lr": 5.451647920853797e-06}
{"train_loss": 0.028561491519212723, "global_step": 230800, "epoch": 2593, "lr": 5.45138468401965e-06}
{"train_loss": 0.0841410905122757, "global_step": 230801, "epoch": 2593, "lr": 5.4511214531745235e-06}
{"train_loss": 0.045990485697984695, "global_step": 230802, "epoch": 2593, "lr": 5.450858228318456e-06}
{"train_loss": 0.042155634611845016, "global_step": 230803, "epoch": 2593, "lr": 5.45059500945147e-06}
{"train_loss": 0.049722619354724884, "global_step": 230804, "epoch": 2593, "lr": 5.450331796573616e-06}
{"train_loss": 0.07378719002008438, "global_step": 230805, "epoch": 2593, "lr": 5.450068589684909e-06}
{"train_loss": 0.02368495799601078, "global_step": 230806, "epoch": 2593, "lr": 5.449805388785412e-06}
{"train_loss": 0.0670938715338707, "global_step": 230807, "epoch": 2593, "lr": 5.449542193875129e-06}
{"train_loss": 0.059577811509370804, "global_step": 230808, "epoch": 2593, "lr": 5.449279004954128e-06}
{"train_loss": 0.07816413044929504, "global_step": 230809, "epoch": 2593, "lr": 5.449015822022418e-06}
{"train_loss": 0.009563867934048176, "global_step": 230810, "epoch": 2593, "lr": 5.448752645080041e-06}
{"train_loss": 0.0330287404358387, "global_step": 230811, "epoch": 2593, "lr": 5.44848947412705e-06}
{"train_loss": 0.08766240626573563, "global_step": 230812, "epoch": 2593, "lr": 5.448226309163456e-06}
{"train_loss": 0.02543780952692032, "global_step": 230813, "epoch": 2593, "lr": 5.447963150189317e-06}
{"train_loss": 0.04040645435452461, "global_step": 230814, "epoch": 2593, "lr": 5.447699997204642e-06}
{"train_loss": 0.0627109631896019, "global_step": 230815, "epoch": 2593, "lr": 5.447436850209492e-06}
{"train_loss": 0.05445335805416107, "global_step": 230816, "epoch": 2593, "lr": 5.447173709203879e-06}
{"train_loss": 0.017929116263985634, "global_step": 230817, "epoch": 2593, "lr": 5.446910574187863e-06}
{"train_loss": 0.05583575367927551, "global_step": 230818, "epoch": 2593, "lr": 5.446647445161457e-06}
{"train_loss": 0.017143940553069115, "global_step": 230819, "epoch": 2593, "lr": 5.44638432212472e-06}
{"train_loss": 0.034814029932022095, "global_step": 230820, "epoch": 2593, "lr": 5.4461212050776585e-06}
{"train_loss": 0.05149444192647934, "global_step": 230821, "epoch": 2593, "lr": 5.445858094020334e-06}
{"train_loss": 0.04384942725300789, "global_step": 230822, "epoch": 2593, "lr": 5.445594988952762e-06}
{"train_loss": 0.12920330464839935, "global_step": 230823, "epoch": 2593, "lr": 5.445331889874994e-06}
{"train_loss": 0.032921887934207916, "global_step": 230824, "epoch": 2593, "lr": 5.445068796787051e-06}
{"train_loss": 0.03524012118577957, "global_step": 230825, "epoch": 2593, "lr": 5.444805709688983e-06}
{"train_loss": 0.03526797890663147, "global_step": 230826, "epoch": 2593, "lr": 5.4445426285808064e-06}
{"train_loss": 0.029828641563653946, "global_step": 230827, "epoch": 2593, "lr": 5.444279553462573e-06}
{"train_loss": 0.03352174162864685, "global_step": 230828, "epoch": 2593, "lr": 5.444016484334319e-06}
{"train_loss": 0.044913798570632935, "global_step": 230829, "epoch": 2593, "lr": 5.4437534211960684e-06}
{"train_loss": 0.029109790921211243, "global_step": 230830, "epoch": 2593, "lr": 5.443490364047871e-06}
{"train_loss": 0.04252508655190468, "global_step": 230831, "epoch": 2593, "lr": 5.443227312889737e-06}
{"train_loss": 0.0623260959982872, "global_step": 230832, "epoch": 2593, "lr": 5.442964267721734e-06}
{"train_loss": 0.044871050864458084, "global_step": 230833, "epoch": 2593, "lr": 5.4427012285438725e-06}
{"train_loss": 0.02103864960372448, "global_step": 230834, "epoch": 2593, "lr": 5.442438195356203e-06}
{"train_loss": 0.06771355122327805, "global_step": 230835, "epoch": 2593, "lr": 5.442175168158753e-06}
{"train_loss": 0.038570038974285126, "global_step": 230836, "epoch": 2593, "lr": 5.441912146951555e-06}
{"train_loss": 0.012614908628165722, "global_step": 230837, "epoch": 2593, "lr": 5.4416491317346555e-06}
{"train_loss": 0.024516845121979713, "global_step": 230838, "epoch": 2593, "lr": 5.4413861225080745e-06}
{"train_loss": 0.034402262419462204, "global_step": 230839, "epoch": 2593, "lr": 5.441123119271863e-06}
{"train_loss": 0.04153440147638321, "global_step": 230840, "epoch": 2593, "lr": 5.440860122026043e-06}
{"train_loss": 0.06302674859762192, "global_step": 230841, "epoch": 2593, "lr": 5.440597130770664e-06}
{"train_loss": 0.04461466521024704, "global_step": 230842, "epoch": 2593, "lr": 5.440334145505754e-06}
{"train_loss": 0.053241368383169174, "global_step": 230843, "epoch": 2593, "lr": 5.440071166231342e-06}
{"train_loss": 0.016560597345232964, "global_step": 230844, "epoch": 2593, "lr": 5.439808192947477e-06}
{"train_loss": 0.05902041494846344, "global_step": 230845, "epoch": 2593, "lr": 5.439545225654175e-06}
{"train_loss": 0.02546818181872368, "global_step": 230846, "epoch": 2593, "lr": 5.43928226435148e-06}
{"train_loss": 0.04886869713664055, "global_step": 230847, "epoch": 2593, "lr": 5.439019309039445e-06}
{"train_loss": 0.05878753960132599, "global_step": 230848, "epoch": 2593, "lr": 5.438756359718083e-06}
{"train_loss": 0.06639040261507034, "global_step": 230849, "epoch": 2593, "lr": 5.438493416387441e-06}
{"train_loss": 0.04101065173745155, "global_step": 230850, "epoch": 2593, "lr": 5.438230479047546e-06}
{"train_loss": 0.04164866358041763, "global_step": 230851, "epoch": 2593, "lr": 5.437967547698442e-06}
{"train_loss": 0.07553720474243164, "global_step": 230852, "epoch": 2593, "lr": 5.437704622340156e-06}
{"train_loss": 0.0483386367559433, "global_step": 230853, "epoch": 2593, "lr": 5.437441702972734e-06}
{"train_loss": 0.07696519047021866, "global_step": 230854, "epoch": 2593, "lr": 5.437178789596192e-06}
{"train_loss": 0.06405390053987503, "global_step": 230855, "epoch": 2593, "lr": 5.436915882210597e-06}
{"train_loss": 0.0697445496916771, "global_step": 230856, "epoch": 2593, "lr": 5.436652980815949e-06}
{"train_loss": 0.01775416173040867, "global_step": 230857, "epoch": 2593, "lr": 5.436390085412307e-06}
{"train_loss": 0.022632168605923653, "global_step": 230858, "epoch": 2593, "lr": 5.436127195999696e-06}
{"train_loss": 0.04209589213132858, "global_step": 230859, "epoch": 2593, "lr": 5.435864312578159e-06}
{"train_loss": 0.09133195132017136, "global_step": 230860, "epoch": 2593, "lr": 5.4356014351477246e-06}
{"train_loss": 0.052674222737550735, "global_step": 230861, "epoch": 2593, "lr": 5.435338563708431e-06}
{"train_loss": 0.030431726947426796, "global_step": 230862, "epoch": 2593, "lr": 5.4350756982603114e-06}
{"train_loss": 0.08236877620220184, "global_step": 230863, "epoch": 2593, "lr": 5.434812838803399e-06}
{"train_loss": 0.06917791813611984, "global_step": 230864, "epoch": 2593, "lr": 5.434549985337739e-06}
{"train_loss": 0.049014005077521454, "global_step": 230865, "epoch": 2593, "lr": 5.434287137863353e-06, "val_loss": 9.165234565734863}
{"train_loss": 0.026159608736634254, "global_step": 230866, "epoch": 2594, "lr": 5.434024296380297e-06}
{"train_loss": 0.028638504445552826, "global_step": 230867, "epoch": 2594, "lr": 5.4337614608885815e-06}
{"train_loss": 0.031062856316566467, "global_step": 230868, "epoch": 2594, "lr": 5.433498631388262e-06}
{"train_loss": 0.031147425994277, "global_step": 230869, "epoch": 2594, "lr": 5.433235807879361e-06}
{"train_loss": 0.06282303482294083, "global_step": 230870, "epoch": 2594, "lr": 5.432972990361923e-06}
{"train_loss": 0.014782645739614964, "global_step": 230871, "epoch": 2594, "lr": 5.43271017883597e-06}
{"train_loss": 0.08401252329349518, "global_step": 230872, "epoch": 2594, "lr": 5.432447373301553e-06}
{"train_loss": 0.0632331371307373, "global_step": 230873, "epoch": 2594, "lr": 5.432184573758692e-06}
{"train_loss": 0.075424425303936, "global_step": 230874, "epoch": 2594, "lr": 5.431921780207444e-06}
{"train_loss": 0.04819963872432709, "global_step": 230875, "epoch": 2594, "lr": 5.4316589926478255e-06}
{"train_loss": 0.07461598515510559, "global_step": 230876, "epoch": 2594, "lr": 5.43139621107987e-06}
{"train_loss": 0.04642302915453911, "global_step": 230877, "epoch": 2594, "lr": 5.431133435503633e-06}
{"train_loss": 0.04564040154218674, "global_step": 230878, "epoch": 2594, "lr": 5.4308706659191245e-06}
{"train_loss": 0.03900222107768059, "global_step": 230879, "epoch": 2594, "lr": 5.430607902326401e-06}
{"train_loss": 0.045033279806375504, "global_step": 230880, "epoch": 2594, "lr": 5.43034514472549e-06}
{"train_loss": 0.10263983905315399, "global_step": 230881, "epoch": 2594, "lr": 5.430082393116414e-06}
{"train_loss": 0.032216548919677734, "global_step": 230882, "epoch": 2594, "lr": 5.429819647499223e-06}
{"train_loss": 0.02583533339202404, "global_step": 230883, "epoch": 2594, "lr": 5.429556907873956e-06}
{"train_loss": 0.0844532921910286, "global_step": 230884, "epoch": 2594, "lr": 5.4292941742406345e-06}
{"train_loss": 0.053014714270830154, "global_step": 230885, "epoch": 2594, "lr": 5.429031446599314e-06}
{"train_loss": 0.038763564079999924, "global_step": 230886, "epoch": 2594, "lr": 5.428768724950001e-06}
{"train_loss": 0.0443451851606369, "global_step": 230887, "epoch": 2594, "lr": 5.428506009292761e-06}
{"train_loss": 0.04322086274623871, "global_step": 230888, "epoch": 2594, "lr": 5.428243299627605e-06}
{"train_loss": 0.0516817644238472, "global_step": 230889, "epoch": 2594, "lr": 5.4279805959545845e-06}
{"train_loss": 0.07389988750219345, "global_step": 230890, "epoch": 2594, "lr": 5.427717898273721e-06}
{"train_loss": 0.04560791701078415, "global_step": 230891, "epoch": 2594, "lr": 5.427455206585069e-06}
{"train_loss": 0.06479766219854355, "global_step": 230892, "epoch": 2594, "lr": 5.427192520888641e-06}
{"train_loss": 0.050181057304143906, "global_step": 230893, "epoch": 2594, "lr": 5.426929841184497e-06}
{"train_loss": 0.03320261090993881, "global_step": 230894, "epoch": 2594, "lr": 5.426667167472644e-06}
{"train_loss": 0.035467423498630524, "global_step": 230895, "epoch": 2594, "lr": 5.426404499753146e-06}
{"train_loss": 0.03665608540177345, "global_step": 230896, "epoch": 2594, "lr": 5.426141838026011e-06}
{"train_loss": 0.04663833975791931, "global_step": 230897, "epoch": 2594, "lr": 5.425879182291299e-06}
{"train_loss": 0.06517250835895538, "global_step": 230898, "epoch": 2594, "lr": 5.425616532549027e-06}
{"train_loss": 0.043118949979543686, "global_step": 230899, "epoch": 2594, "lr": 5.425353888799245e-06}
{"train_loss": 0.039101555943489075, "global_step": 230900, "epoch": 2594, "lr": 5.425091251041975e-06}
{"train_loss": 0.06543449312448502, "global_step": 230901, "epoch": 2594, "lr": 5.42482861927725e-06}
{"train_loss": 0.04327484592795372, "global_step": 230902, "epoch": 2594, "lr": 5.424565993505132e-06}
{"train_loss": 0.046101201325654984, "global_step": 230903, "epoch": 2594, "lr": 5.42430337372562e-06}
{"train_loss": 0.04230635240674019, "global_step": 230904, "epoch": 2594, "lr": 5.424040759938786e-06}
{"train_loss": 0.059100255370140076, "global_step": 230905, "epoch": 2594, "lr": 5.423778152144632e-06}
{"train_loss": 0.06536298245191574, "global_step": 230906, "epoch": 2594, "lr": 5.423515550343217e-06}
{"train_loss": 0.044794801622629166, "global_step": 230907, "epoch": 2594, "lr": 5.423252954534558e-06}
{"train_loss": 0.04845735430717468, "global_step": 230908, "epoch": 2594, "lr": 5.4229903647187055e-06}
{"train_loss": 0.05255335941910744, "global_step": 230909, "epoch": 2594, "lr": 5.422727780895681e-06}
{"train_loss": 0.08606944978237152, "global_step": 230910, "epoch": 2594, "lr": 5.422465203065541e-06}
{"train_loss": 0.03549765795469284, "global_step": 230911, "epoch": 2594, "lr": 5.422202631228296e-06}
{"train_loss": 0.03905379772186279, "global_step": 230912, "epoch": 2594, "lr": 5.421940065384001e-06}
{"train_loss": 0.038802437484264374, "global_step": 230913, "epoch": 2594, "lr": 5.42167750553268e-06}
{"train_loss": 0.07008720934391022, "global_step": 230914, "epoch": 2594, "lr": 5.421414951674364e-06}
{"train_loss": 0.05038650333881378, "global_step": 230915, "epoch": 2594, "lr": 5.421152403809105e-06}
{"train_loss": 0.042520008981227875, "global_step": 230916, "epoch": 2594, "lr": 5.420889861936923e-06}
{"train_loss": 0.05200846865773201, "global_step": 230917, "epoch": 2594, "lr": 5.420627326057865e-06}
{"train_loss": 0.06374246627092361, "global_step": 230918, "epoch": 2594, "lr": 5.420364796171951e-06}
{"train_loss": 0.04318031296133995, "global_step": 230919, "epoch": 2594, "lr": 5.420102272279232e-06}
{"train_loss": 0.05230777710676193, "global_step": 230920, "epoch": 2594, "lr": 5.419839754379729e-06}
{"train_loss": 0.06376136094331741, "global_step": 230921, "epoch": 2594, "lr": 5.419577242473495e-06}
{"train_loss": 0.06877201050519943, "global_step": 230922, "epoch": 2594, "lr": 5.419314736560549e-06}
{"train_loss": 0.06346864998340607, "global_step": 230923, "epoch": 2594, "lr": 5.4190522366409425e-06}
{"train_loss": 0.07205241173505783, "global_step": 230924, "epoch": 2594, "lr": 5.418789742714686e-06}
{"train_loss": 0.06655898690223694, "global_step": 230925, "epoch": 2594, "lr": 5.418527254781847e-06}
{"train_loss": 0.019811350852251053, "global_step": 230926, "epoch": 2594, "lr": 5.418264772842424e-06}
{"train_loss": 0.02528909593820572, "global_step": 230927, "epoch": 2594, "lr": 5.418002296896491e-06}
{"train_loss": 0.0213925763964653, "global_step": 230928, "epoch": 2594, "lr": 5.417739826944051e-06}
{"train_loss": 0.036843255162239075, "global_step": 230929, "epoch": 2594, "lr": 5.417477362985163e-06}
{"train_loss": 0.0866982713341713, "global_step": 230930, "epoch": 2594, "lr": 5.417214905019841e-06}
{"train_loss": 0.02186770550906658, "global_step": 230931, "epoch": 2594, "lr": 5.4169524530481405e-06}
{"train_loss": 0.04494613781571388, "global_step": 230932, "epoch": 2594, "lr": 5.416690007070074e-06}
{"train_loss": 0.05258795619010925, "global_step": 230933, "epoch": 2594, "lr": 5.416427567085708e-06}
{"train_loss": 0.10386031121015549, "global_step": 230934, "epoch": 2594, "lr": 5.416165133095042e-06}
{"train_loss": 0.0452532060444355, "global_step": 230935, "epoch": 2594, "lr": 5.415902705098147e-06}
{"train_loss": 0.06097949296236038, "global_step": 230936, "epoch": 2594, "lr": 5.415640283095024e-06}
{"train_loss": 0.04689786955714226, "global_step": 230937, "epoch": 2594, "lr": 5.41537786708573e-06}
{"train_loss": 0.06154985353350639, "global_step": 230938, "epoch": 2594, "lr": 5.4151154570703026e-06}
{"train_loss": 0.04591934382915497, "global_step": 230939, "epoch": 2594, "lr": 5.414853053048757e-06}
{"train_loss": 0.02856582961976528, "global_step": 230940, "epoch": 2594, "lr": 5.4145906550211516e-06}
{"train_loss": 0.03844977542757988, "global_step": 230941, "epoch": 2594, "lr": 5.414328262987506e-06}
{"train_loss": 0.023311572149395943, "global_step": 230942, "epoch": 2594, "lr": 5.414065876947866e-06}
{"train_loss": 0.03764518350362778, "global_step": 230943, "epoch": 2594, "lr": 5.413803496902253e-06}
{"train_loss": 0.026682432740926743, "global_step": 230944, "epoch": 2594, "lr": 5.413541122850718e-06}
{"train_loss": 0.09765565395355225, "global_step": 230945, "epoch": 2594, "lr": 5.413278754793283e-06}
{"train_loss": 0.07713214308023453, "global_step": 230946, "epoch": 2594, "lr": 5.413016392729997e-06}
{"train_loss": 0.027462955564260483, "global_step": 230947, "epoch": 2594, "lr": 5.412754036660878e-06}
{"train_loss": 0.0536961629986763, "global_step": 230948, "epoch": 2594, "lr": 5.41249168658598e-06}
{"train_loss": 0.026797108352184296, "global_step": 230949, "epoch": 2594, "lr": 5.412229342505321e-06}
{"train_loss": 0.058587171137332916, "global_step": 230950, "epoch": 2594, "lr": 5.411967004418955e-06}
{"train_loss": 0.04690483957529068, "global_step": 230951, "epoch": 2594, "lr": 5.411704672326906e-06}
{"train_loss": 0.08316818624734879, "global_step": 230952, "epoch": 2594, "lr": 5.411442346229195e-06}
{"train_loss": 0.029491156339645386, "global_step": 230953, "epoch": 2594, "lr": 5.4111800261258896e-06}
{"train_loss": 0.05029408785441283, "global_step": 230954, "epoch": 2594, "lr": 5.410917712016988e-06, "val_loss": 9.14384651184082}
{"train_loss": 0.03479944169521332, "global_step": 230955, "epoch": 2595, "lr": 5.410655403902565e-06}
{"train_loss": 0.023265814408659935, "global_step": 230956, "epoch": 2595, "lr": 5.410393101782618e-06}
{"train_loss": 0.05607433244585991, "global_step": 230957, "epoch": 2595, "lr": 5.410130805657215e-06}
{"train_loss": 0.03300115466117859, "global_step": 230958, "epoch": 2595, "lr": 5.409868515526367e-06}
{"train_loss": 0.03913262486457825, "global_step": 230959, "epoch": 2595, "lr": 5.409606231390124e-06}
{"train_loss": 0.05380084365606308, "global_step": 230960, "epoch": 2595, "lr": 5.409343953248513e-06}
{"train_loss": 0.02387845516204834, "global_step": 230961, "epoch": 2595, "lr": 5.40908168110158e-06}
{"train_loss": 0.06677502393722534, "global_step": 230962, "epoch": 2595, "lr": 5.40881941494934e-06}
{"train_loss": 0.049595486372709274, "global_step": 230963, "epoch": 2595, "lr": 5.408557154791849e-06}
{"train_loss": 0.04251823574304581, "global_step": 230964, "epoch": 2595, "lr": 5.408294900629124e-06}
{"train_loss": 0.03069876879453659, "global_step": 230965, "epoch": 2595, "lr": 5.4080326524612255e-06}
{"train_loss": 0.04458755627274513, "global_step": 230966, "epoch": 2595, "lr": 5.40777041028816e-06}
{"train_loss": 0.09717360883951187, "global_step": 230967, "epoch": 2595, "lr": 5.407508174109988e-06}
{"train_loss": 0.07425164431333542, "global_step": 230968, "epoch": 2595, "lr": 5.40724594392672e-06}
{"train_loss": 0.05086264759302139, "global_step": 230969, "epoch": 2595, "lr": 5.406983719738418e-06}
{"train_loss": 0.03021526150405407, "global_step": 230970, "epoch": 2595, "lr": 5.406721501545092e-06}
{"train_loss": 0.035155899822711945, "global_step": 230971, "epoch": 2595, "lr": 5.406459289346799e-06}
{"train_loss": 0.09816936403512955, "global_step": 230972, "epoch": 2595, "lr": 5.406197083143555e-06}
{"train_loss": 0.051669470965862274, "global_step": 230973, "epoch": 2595, "lr": 5.405934882935398e-06}
{"train_loss": 0.047917794436216354, "global_step": 230974, "epoch": 2595, "lr": 5.405672688722385e-06}
{"train_loss": 0.10388968139886856, "global_step": 230975, "epoch": 2595, "lr": 5.405410500504526e-06}
{"train_loss": 0.04993365332484245, "global_step": 230976, "epoch": 2595, "lr": 5.405148318281878e-06}
{"train_loss": 0.026300761848688126, "global_step": 230977, "epoch": 2595, "lr": 5.4048861420544494e-06}
{"train_loss": 0.039698049426078796, "global_step": 230978, "epoch": 2595, "lr": 5.4046239718223035e-06}
{"train_loss": 0.08126702904701233, "global_step": 230979, "epoch": 2595, "lr": 5.404361807585451e-06}
{"train_loss": 0.05288999155163765, "global_step": 230980, "epoch": 2595, "lr": 5.404099649343947e-06}
{"train_loss": 0.046553634107112885, "global_step": 230981, "epoch": 2595, "lr": 5.4038374970978135e-06}
{"train_loss": 0.020357400178909302, "global_step": 230982, "epoch": 2595, "lr": 5.403575350847096e-06}
{"train_loss": 0.03399181738495827, "global_step": 230983, "epoch": 2595, "lr": 5.403313210591815e-06}
{"train_loss": 0.0667588859796524, "global_step": 230984, "epoch": 2595, "lr": 5.403051076332022e-06}
{"train_loss": 0.06522107124328613, "global_step": 230985, "epoch": 2595, "lr": 5.402788948067738e-06}
{"train_loss": 0.03813505545258522, "global_step": 230986, "epoch": 2595, "lr": 5.40252682579902e-06}
{"train_loss": 0.06524762511253357, "global_step": 230987, "epoch": 2595, "lr": 5.4022647095258725e-06}
{"train_loss": 0.02989048883318901, "global_step": 230988, "epoch": 2595, "lr": 5.402002599248363e-06}
{"train_loss": 0.08776383846998215, "global_step": 230989, "epoch": 2595, "lr": 5.401740494966495e-06}
{"train_loss": 0.032866448163986206, "global_step": 230990, "epoch": 2595, "lr": 5.401478396680332e-06}
{"train_loss": 0.040426552295684814, "global_step": 230991, "epoch": 2595, "lr": 5.401216304389894e-06}
{"train_loss": 0.0577748566865921, "global_step": 230992, "epoch": 2595, "lr": 5.400954218095211e-06}
{"train_loss": 0.08815641701221466, "global_step": 230993, "epoch": 2595, "lr": 5.400692137796337e-06}
{"train_loss": 0.058800313621759415, "global_step": 230994, "epoch": 2595, "lr": 5.4004300634932835e-06}
{"train_loss": 0.04341317340731621, "global_step": 230995, "epoch": 2595, "lr": 5.400167995186111e-06}
{"train_loss": 0.07563083618879318, "global_step": 230996, "epoch": 2595, "lr": 5.399905932874832e-06}
{"train_loss": 0.08742218464612961, "global_step": 230997, "epoch": 2595, "lr": 5.399643876559501e-06}
{"train_loss": 0.025920795276761055, "global_step": 230998, "epoch": 2595, "lr": 5.399381826240135e-06}
{"train_loss": 0.02899193949997425, "global_step": 230999, "epoch": 2595, "lr": 5.39911978191679e-06}
{"train_loss": 0.0686337873339653, "global_step": 231000, "epoch": 2595, "lr": 5.398857743589475e-06}
{"train_loss": 0.04257594794034958, "global_step": 231001, "epoch": 2595, "lr": 5.398595711258253e-06}
{"train_loss": 0.03836753964424133, "global_step": 231002, "epoch": 2595, "lr": 5.39833368492314e-06}
{"train_loss": 0.07539679110050201, "global_step": 231003, "epoch": 2595, "lr": 5.398071664584181e-06}
{"train_loss": 0.03130994737148285, "global_step": 231004, "epoch": 2595, "lr": 5.397809650241403e-06}
{"train_loss": 0.0544334352016449, "global_step": 231005, "epoch": 2595, "lr": 5.397547641894851e-06}
{"train_loss": 0.02989162504673004, "global_step": 231006, "epoch": 2595, "lr": 5.397285639544547e-06}
{"train_loss": 0.03046455979347229, "global_step": 231007, "epoch": 2595, "lr": 5.3970236431905465e-06}
{"train_loss": 0.025119313970208168, "global_step": 231008, "epoch": 2595, "lr": 5.39676165283286e-06}
{"train_loss": 0.08514020591974258, "global_step": 231009, "epoch": 2595, "lr": 5.396499668471533e-06}
{"train_loss": 0.07509779930114746, "global_step": 231010, "epoch": 2595, "lr": 5.3962376901066205e-06}
{"train_loss": 0.05505947023630142, "global_step": 231011, "epoch": 2595, "lr": 5.395975717738122e-06}
{"train_loss": 0.07647477090358734, "global_step": 231012, "epoch": 2595, "lr": 5.395713751366105e-06}
{"train_loss": 0.03393079340457916, "global_step": 231013, "epoch": 2595, "lr": 5.39545179099058e-06}
{"train_loss": 0.023589594289660454, "global_step": 231014, "epoch": 2595, "lr": 5.395189836611603e-06}
{"train_loss": 0.032146770507097244, "global_step": 231015, "epoch": 2595, "lr": 5.3949278882291895e-06}
{"train_loss": 0.07114161550998688, "global_step": 231016, "epoch": 2595, "lr": 5.3946659458433964e-06}
{"train_loss": 0.03171071037650108, "global_step": 231017, "epoch": 2595, "lr": 5.394404009454235e-06}
{"train_loss": 0.026590008288621902, "global_step": 231018, "epoch": 2595, "lr": 5.394142079061759e-06}
{"train_loss": 0.015744715929031372, "global_step": 231019, "epoch": 2595, "lr": 5.393880154665986e-06}
{"train_loss": 0.012470629066228867, "global_step": 231020, "epoch": 2595, "lr": 5.393618236266979e-06}
{"train_loss": 0.058635782450437546, "global_step": 231021, "epoch": 2595, "lr": 5.393356323864745e-06}
{"train_loss": 0.04995975270867348, "global_step": 231022, "epoch": 2595, "lr": 5.393094417459338e-06}
{"train_loss": 0.028245791792869568, "global_step": 231023, "epoch": 2595, "lr": 5.392832517050772e-06}
{"train_loss": 0.02737104706466198, "global_step": 231024, "epoch": 2595, "lr": 5.3925706226391145e-06}
{"train_loss": 0.03121257945895195, "global_step": 231025, "epoch": 2595, "lr": 5.392308734224366e-06}
{"train_loss": 0.030552133917808533, "global_step": 231026, "epoch": 2595, "lr": 5.392046851806592e-06}
{"train_loss": 0.06200111284852028, "global_step": 231027, "epoch": 2595, "lr": 5.3917849753858e-06}
{"train_loss": 0.039577655494213104, "global_step": 231028, "epoch": 2595, "lr": 5.391523104962048e-06}
{"train_loss": 0.03131434693932533, "global_step": 231029, "epoch": 2595, "lr": 5.391261240535367e-06}
{"train_loss": 0.08512069284915924, "global_step": 231030, "epoch": 2595, "lr": 5.390999382105777e-06}
{"train_loss": 0.02816213108599186, "global_step": 231031, "epoch": 2595, "lr": 5.3907375296733285e-06}
{"train_loss": 0.13501101732254028, "global_step": 231032, "epoch": 2595, "lr": 5.3904756832380445e-06}
{"train_loss": 0.03737183287739754, "global_step": 231033, "epoch": 2595, "lr": 5.390213842799974e-06}
{"train_loss": 0.0633745864033699, "global_step": 231034, "epoch": 2595, "lr": 5.389952008359139e-06}
{"train_loss": 0.07159557193517685, "global_step": 231035, "epoch": 2595, "lr": 5.38969017991559e-06}
{"train_loss": 0.03577813506126404, "global_step": 231036, "epoch": 2595, "lr": 5.389428357469345e-06}
{"train_loss": 0.03206648305058479, "global_step": 231037, "epoch": 2595, "lr": 5.389166541020457e-06}
{"train_loss": 0.039343807846307755, "global_step": 231038, "epoch": 2595, "lr": 5.388904730568945e-06}
{"train_loss": 0.043452855199575424, "global_step": 231039, "epoch": 2595, "lr": 5.3886429261148564e-06}
{"train_loss": 0.08990459144115448, "global_step": 231040, "epoch": 2595, "lr": 5.38838112765821e-06}
{"train_loss": 0.05199144780635834, "global_step": 231041, "epoch": 2595, "lr": 5.388119335199066e-06}
{"train_loss": 0.030407747253775597, "global_step": 231042, "epoch": 2595, "lr": 5.387857548737435e-06}
{"train_loss": 0.04962317538813928, "global_step": 231043, "epoch": 2595, "lr": 5.387595768273374e-06, "val_loss": 9.070757865905762, "train_action_mse_error": 22.383991241455078}
{"train_loss": 0.07583288848400116, "global_step": 231044, "epoch": 2596, "lr": 5.387333993806892e-06}
{"train_loss": 0.04744962602853775, "global_step": 231045, "epoch": 2596, "lr": 5.3870722253380415e-06}
{"train_loss": 0.07187347859144211, "global_step": 231046, "epoch": 2596, "lr": 5.386810462866864e-06}
{"train_loss": 0.06221044063568115, "global_step": 231047, "epoch": 2596, "lr": 5.386548706393379e-06}
{"train_loss": 0.07373128831386566, "global_step": 231048, "epoch": 2596, "lr": 5.3862869559176355e-06}
{"train_loss": 0.06557215750217438, "global_step": 231049, "epoch": 2596, "lr": 5.386025211439655e-06}
{"train_loss": 0.10864926129579544, "global_step": 231050, "epoch": 2596, "lr": 5.385763472959493e-06}
{"train_loss": 0.018419388681650162, "global_step": 231051, "epoch": 2596, "lr": 5.385501740477156e-06}
{"train_loss": 0.04117066413164139, "global_step": 231052, "epoch": 2596, "lr": 5.385240013992704e-06}
{"train_loss": 0.04736562818288803, "global_step": 231053, "epoch": 2596, "lr": 5.384978293506155e-06}
{"train_loss": 0.0233604833483696, "global_step": 231054, "epoch": 2596, "lr": 5.384716579017563e-06}
{"train_loss": 0.05062440410256386, "global_step": 231055, "epoch": 2596, "lr": 5.384454870526939e-06}
{"train_loss": 0.0532321035861969, "global_step": 231056, "epoch": 2596, "lr": 5.38419316803434e-06}
{"train_loss": 0.07358156144618988, "global_step": 231057, "epoch": 2596, "lr": 5.383931471539788e-06}
{"train_loss": 0.0647977888584137, "global_step": 231058, "epoch": 2596, "lr": 5.383669781043327e-06}
{"train_loss": 0.04053252935409546, "global_step": 231059, "epoch": 2596, "lr": 5.383408096544978e-06}
{"train_loss": 0.04973038285970688, "global_step": 231060, "epoch": 2596, "lr": 5.3831464180447985e-06}
{"train_loss": 0.04736631363630295, "global_step": 231061, "epoch": 2596, "lr": 5.382884745542805e-06}
{"train_loss": 0.0388292595744133, "global_step": 231062, "epoch": 2596, "lr": 5.382623079039045e-06}
{"train_loss": 0.06525541841983795, "global_step": 231063, "epoch": 2596, "lr": 5.382361418533538e-06}
{"train_loss": 0.04790699481964111, "global_step": 231064, "epoch": 2596, "lr": 5.382099764026339e-06}
{"train_loss": 0.07647315412759781, "global_step": 231065, "epoch": 2596, "lr": 5.381838115517457e-06}
{"train_loss": 0.05382329598069191, "global_step": 231066, "epoch": 2596, "lr": 5.381576473006961e-06}
{"train_loss": 0.0687098503112793, "global_step": 231067, "epoch": 2596, "lr": 5.381314836494866e-06}
{"train_loss": 0.028716860339045525, "global_step": 231068, "epoch": 2596, "lr": 5.381053205981196e-06}
{"train_loss": 0.03229013457894325, "global_step": 231069, "epoch": 2596, "lr": 5.3807915814660105e-06}
{"train_loss": 0.021046118810772896, "global_step": 231070, "epoch": 2596, "lr": 5.380529962949327e-06}
{"train_loss": 0.06359606981277466, "global_step": 231071, "epoch": 2596, "lr": 5.3802683504311945e-06}
{"train_loss": 0.05954593047499657, "global_step": 231072, "epoch": 2596, "lr": 5.380006743911637e-06}
{"train_loss": 0.021613167598843575, "global_step": 231073, "epoch": 2596, "lr": 5.379745143390697e-06}
{"train_loss": 0.03330185264348984, "global_step": 231074, "epoch": 2596, "lr": 5.379483548868397e-06}
{"train_loss": 0.04752068966627121, "global_step": 231075, "epoch": 2596, "lr": 5.379221960344793e-06}
{"train_loss": 0.043712910264730453, "global_step": 231076, "epoch": 2596, "lr": 5.378960377819903e-06}
{"train_loss": 0.10146034508943558, "global_step": 231077, "epoch": 2596, "lr": 5.378698801293769e-06}
{"train_loss": 0.05707261338829994, "global_step": 231078, "epoch": 2596, "lr": 5.37843723076642e-06}
{"train_loss": 0.06090901046991348, "global_step": 231079, "epoch": 2596, "lr": 5.378175666237905e-06}
{"train_loss": 0.04519467428326607, "global_step": 231080, "epoch": 2596, "lr": 5.377914107708243e-06}
{"train_loss": 0.029754159972071648, "global_step": 231081, "epoch": 2596, "lr": 5.3776525551774756e-06}
{"train_loss": 0.08908295631408691, "global_step": 231082, "epoch": 2596, "lr": 5.377391008645649e-06}
{"train_loss": 0.08895032852888107, "global_step": 231083, "epoch": 2596, "lr": 5.377129468112779e-06}
{"train_loss": 0.04205663874745369, "global_step": 231084, "epoch": 2596, "lr": 5.376867933578916e-06}
{"train_loss": 0.04927443340420723, "global_step": 231085, "epoch": 2596, "lr": 5.376606405044083e-06}
{"train_loss": 0.023983096703886986, "global_step": 231086, "epoch": 2596, "lr": 5.376344882508333e-06}
{"train_loss": 0.04225733131170273, "global_step": 231087, "epoch": 2596, "lr": 5.376083365971674e-06}
{"train_loss": 0.05982132628560066, "global_step": 231088, "epoch": 2596, "lr": 5.375821855434171e-06}
{"train_loss": 0.02713709883391857, "global_step": 231089, "epoch": 2596, "lr": 5.375560350895837e-06}
{"train_loss": 0.05799068883061409, "global_step": 231090, "epoch": 2596, "lr": 5.375298852356719e-06}
{"train_loss": 0.0436384379863739, "global_step": 231091, "epoch": 2596, "lr": 5.375037359816842e-06}
{"train_loss": 0.11272558569908142, "global_step": 231092, "epoch": 2596, "lr": 5.374775873276255e-06}
{"train_loss": 0.011321377009153366, "global_step": 231093, "epoch": 2596, "lr": 5.374514392734975e-06}
{"train_loss": 0.07965339720249176, "global_step": 231094, "epoch": 2596, "lr": 5.3742529181930626e-06}
{"train_loss": 0.05049682781100273, "global_step": 231095, "epoch": 2596, "lr": 5.3739914496505225e-06}
{"train_loss": 0.057066768407821655, "global_step": 231096, "epoch": 2596, "lr": 5.373729987107418e-06}
{"train_loss": 0.06973788142204285, "global_step": 231097, "epoch": 2596, "lr": 5.3734685305637575e-06}
{"train_loss": 0.0471607968211174, "global_step": 231098, "epoch": 2596, "lr": 5.3732070800196045e-06}
{"train_loss": 0.01647874154150486, "global_step": 231099, "epoch": 2596, "lr": 5.372945635474969e-06}
{"train_loss": 0.028597967699170113, "global_step": 231100, "epoch": 2596, "lr": 5.3726841969299e-06}
{"train_loss": 0.06142609193921089, "global_step": 231101, "epoch": 2596, "lr": 5.372422764384438e-06}
{"train_loss": 0.07202452421188354, "global_step": 231102, "epoch": 2596, "lr": 5.372161337838605e-06}
{"train_loss": 0.04249853268265724, "global_step": 231103, "epoch": 2596, "lr": 5.371899917292439e-06}
{"train_loss": 0.07250816375017166, "global_step": 231104, "epoch": 2596, "lr": 5.371638502745979e-06}
{"train_loss": 0.038171783089637756, "global_step": 231105, "epoch": 2596, "lr": 5.3713770941992536e-06}
{"train_loss": 0.07618528604507446, "global_step": 231106, "epoch": 2596, "lr": 5.371115691652312e-06}
{"train_loss": 0.04137586057186127, "global_step": 231107, "epoch": 2596, "lr": 5.3708542951051765e-06}
{"train_loss": 0.07970229536294937, "global_step": 231108, "epoch": 2596, "lr": 5.370592904557881e-06}
{"train_loss": 0.05503581091761589, "global_step": 231109, "epoch": 2596, "lr": 5.3703315200104685e-06}
{"train_loss": 0.058307796716690063, "global_step": 231110, "epoch": 2596, "lr": 5.370070141462968e-06}
{"train_loss": 0.07484051585197449, "global_step": 231111, "epoch": 2596, "lr": 5.369808768915424e-06}
{"train_loss": 0.013464973308146, "global_step": 231112, "epoch": 2596, "lr": 5.369547402367853e-06}
{"train_loss": 0.08416977524757385, "global_step": 231113, "epoch": 2596, "lr": 5.369286041820315e-06}
{"train_loss": 0.06785718351602554, "global_step": 231114, "epoch": 2596, "lr": 5.369024687272828e-06}
{"train_loss": 0.04436943307518959, "global_step": 231115, "epoch": 2596, "lr": 5.368763338725435e-06}
{"train_loss": 0.0422581322491169, "global_step": 231116, "epoch": 2596, "lr": 5.36850199617816e-06}
{"train_loss": 0.03664946183562279, "global_step": 231117, "epoch": 2596, "lr": 5.368240659631041e-06}
{"train_loss": 0.07776719331741333, "global_step": 231118, "epoch": 2596, "lr": 5.367979329084133e-06}
{"train_loss": 0.04973074048757553, "global_step": 231119, "epoch": 2596, "lr": 5.3677180045374474e-06}
{"train_loss": 0.06204504892230034, "global_step": 231120, "epoch": 2596, "lr": 5.367456685991035e-06}
{"train_loss": 0.026130467653274536, "global_step": 231121, "epoch": 2596, "lr": 5.367195373444917e-06}
{"train_loss": 0.042911674827337265, "global_step": 231122, "epoch": 2596, "lr": 5.366934066899143e-06}
{"train_loss": 0.014925666153430939, "global_step": 231123, "epoch": 2596, "lr": 5.366672766353731e-06}
{"train_loss": 0.0861116573214531, "global_step": 231124, "epoch": 2596, "lr": 5.366411471808736e-06}
{"train_loss": 0.05826599895954132, "global_step": 231125, "epoch": 2596, "lr": 5.366150183264173e-06}
{"train_loss": 0.08744087815284729, "global_step": 231126, "epoch": 2596, "lr": 5.3658889007200954e-06}
{"train_loss": 0.013556956313550472, "global_step": 231127, "epoch": 2596, "lr": 5.365627624176523e-06}
{"train_loss": 0.03410043939948082, "global_step": 231128, "epoch": 2596, "lr": 5.365366353633505e-06}
{"train_loss": 0.058346088975667953, "global_step": 231129, "epoch": 2596, "lr": 5.365105089091066e-06}
{"train_loss": 0.08646555989980698, "global_step": 231130, "epoch": 2596, "lr": 5.364843830549249e-06}
{"train_loss": 0.06311861425638199, "global_step": 231131, "epoch": 2596, "lr": 5.364582578008076e-06}
{"train_loss": 0.053779448100020374, "global_step": 231132, "epoch": 2596, "lr": 5.364321331467598e-06, "val_loss": 9.20911693572998}
{"train_loss": 0.037330880761146545, "global_step": 231133, "epoch": 2597, "lr": 5.364060090927836e-06}
{"train_loss": 0.04601955786347389, "global_step": 231134, "epoch": 2597, "lr": 5.363798856388841e-06}
{"train_loss": 0.030407261103391647, "global_step": 231135, "epoch": 2597, "lr": 5.363537627850629e-06}
{"train_loss": 0.041618503630161285, "global_step": 231136, "epoch": 2597, "lr": 5.36327640531325e-06}
{"train_loss": 0.07768619060516357, "global_step": 231137, "epoch": 2597, "lr": 5.3630151887767374e-06}
{"train_loss": 0.06626574695110321, "global_step": 231138, "epoch": 2597, "lr": 5.362753978241119e-06}
{"train_loss": 0.04840124398469925, "global_step": 231139, "epoch": 2597, "lr": 5.36249277370644e-06}
{"train_loss": 0.0297060776501894, "global_step": 231140, "epoch": 2597, "lr": 5.362231575172733e-06}
{"train_loss": 0.05335329473018646, "global_step": 231141, "epoch": 2597, "lr": 5.361970382640019e-06}
{"train_loss": 0.109767846763134, "global_step": 231142, "epoch": 2597, "lr": 5.36170919610835e-06}
{"train_loss": 0.07936885207891464, "global_step": 231143, "epoch": 2597, "lr": 5.361448015577752e-06}
{"train_loss": 0.0719386488199234, "global_step": 231144, "epoch": 2597, "lr": 5.361186841048266e-06}
{"train_loss": 0.09018420428037643, "global_step": 231145, "epoch": 2597, "lr": 5.360925672519929e-06}
{"train_loss": 0.05157005041837692, "global_step": 231146, "epoch": 2597, "lr": 5.360664509992763e-06}
{"train_loss": 0.02502451278269291, "global_step": 231147, "epoch": 2597, "lr": 5.3604033534668195e-06}
{"train_loss": 0.04820745065808296, "global_step": 231148, "epoch": 2597, "lr": 5.360142202942114e-06}
{"train_loss": 0.06007303670048714, "global_step": 231149, "epoch": 2597, "lr": 5.3598810584187034e-06}
{"train_loss": 0.02720109559595585, "global_step": 231150, "epoch": 2597, "lr": 5.359619919896608e-06}
{"train_loss": 0.014132446609437466, "global_step": 231151, "epoch": 2597, "lr": 5.359358787375873e-06}
{"train_loss": 0.05720489099621773, "global_step": 231152, "epoch": 2597, "lr": 5.359097660856516e-06}
{"train_loss": 0.07269434630870819, "global_step": 231153, "epoch": 2597, "lr": 5.358836540338597e-06}
{"train_loss": 0.05026448145508766, "global_step": 231154, "epoch": 2597, "lr": 5.358575425822127e-06}
{"train_loss": 0.013385391794145107, "global_step": 231155, "epoch": 2597, "lr": 5.358314317307156e-06}
{"train_loss": 0.029711147770285606, "global_step": 231156, "epoch": 2597, "lr": 5.3580532147937236e-06}
{"train_loss": 0.026094410568475723, "global_step": 231157, "epoch": 2597, "lr": 5.357792118281846e-06}
{"train_loss": 0.05468665435910225, "global_step": 231158, "epoch": 2597, "lr": 5.357531027771578e-06}
{"train_loss": 0.08533176779747009, "global_step": 231159, "epoch": 2597, "lr": 5.357269943262938e-06}
{"train_loss": 0.0381455235183239, "global_step": 231160, "epoch": 2597, "lr": 5.35700886475598e-06}
{"train_loss": 0.0634177178144455, "global_step": 231161, "epoch": 2597, "lr": 5.356747792250716e-06}
{"train_loss": 0.050159115344285965, "global_step": 231162, "epoch": 2597, "lr": 5.356486725747206e-06}
{"train_loss": 0.058400776237249374, "global_step": 231163, "epoch": 2597, "lr": 5.3562256652454565e-06}
{"train_loss": 0.02931034378707409, "global_step": 231164, "epoch": 2597, "lr": 5.355964610745534e-06}
{"train_loss": 0.0348675437271595, "global_step": 231165, "epoch": 2597, "lr": 5.355703562247444e-06}
{"train_loss": 0.10237312316894531, "global_step": 231166, "epoch": 2597, "lr": 5.355442519751247e-06}
{"train_loss": 0.09709606319665909, "global_step": 231167, "epoch": 2597, "lr": 5.355181483256955e-06}
{"train_loss": 0.025042975321412086, "global_step": 231168, "epoch": 2597, "lr": 5.354920452764628e-06}
{"train_loss": 0.01821066625416279, "global_step": 231169, "epoch": 2597, "lr": 5.354659428274278e-06}
{"train_loss": 0.03382720425724983, "global_step": 231170, "epoch": 2597, "lr": 5.354398409785955e-06}
{"train_loss": 0.05458299443125725, "global_step": 231171, "epoch": 2597, "lr": 5.3541373972996855e-06}
{"train_loss": 0.04817933961749077, "global_step": 231172, "epoch": 2597, "lr": 5.3538763908155045e-06}
{"train_loss": 0.1042599231004715, "global_step": 231173, "epoch": 2597, "lr": 5.353615390333461e-06}
{"train_loss": 0.028228256851434708, "global_step": 231174, "epoch": 2597, "lr": 5.353354395853566e-06}
{"train_loss": 0.06860063970088959, "global_step": 231175, "epoch": 2597, "lr": 5.353093407375887e-06}
{"train_loss": 0.09637954831123352, "global_step": 231176, "epoch": 2597, "lr": 5.352832424900423e-06}
{"train_loss": 0.04383070394396782, "global_step": 231177, "epoch": 2597, "lr": 5.352571448427241e-06}
{"train_loss": 0.09268591552972794, "global_step": 231178, "epoch": 2597, "lr": 5.352310477956357e-06}
{"train_loss": 0.037023354321718216, "global_step": 231179, "epoch": 2597, "lr": 5.3520495134878055e-06}
{"train_loss": 0.0529358834028244, "global_step": 231180, "epoch": 2597, "lr": 5.351788555021631e-06}
{"train_loss": 0.04782735928893089, "global_step": 231181, "epoch": 2597, "lr": 5.35152760255786e-06}
{"train_loss": 0.0473320409655571, "global_step": 231182, "epoch": 2597, "lr": 5.351266656096537e-06}
{"train_loss": 0.0254365187138319, "global_step": 231183, "epoch": 2597, "lr": 5.351005715637686e-06}
{"train_loss": 0.021862391382455826, "global_step": 231184, "epoch": 2597, "lr": 5.350744781181355e-06}
{"train_loss": 0.02329578623175621, "global_step": 231185, "epoch": 2597, "lr": 5.350483852727578e-06}
{"train_loss": 0.04498439282178879, "global_step": 231186, "epoch": 2597, "lr": 5.3502229302763665e-06}
{"train_loss": 0.0331333763897419, "global_step": 231187, "epoch": 2597, "lr": 5.3499620138277864e-06}
{"train_loss": 0.07039134204387665, "global_step": 231188, "epoch": 2597, "lr": 5.349701103381854e-06}
{"train_loss": 0.09148228913545609, "global_step": 231189, "epoch": 2597, "lr": 5.349440198938615e-06}
{"train_loss": 0.05006054416298866, "global_step": 231190, "epoch": 2597, "lr": 5.349179300498092e-06}
{"train_loss": 0.038582269102334976, "global_step": 231191, "epoch": 2597, "lr": 5.348918408060327e-06}
{"train_loss": 0.019462689757347107, "global_step": 231192, "epoch": 2597, "lr": 5.348657521625361e-06}
{"train_loss": 0.08494260907173157, "global_step": 231193, "epoch": 2597, "lr": 5.34839664119322e-06}
{"train_loss": 0.060631025582551956, "global_step": 231194, "epoch": 2597, "lr": 5.348135766763951e-06}
{"train_loss": 0.02685726061463356, "global_step": 231195, "epoch": 2597, "lr": 5.347874898337568e-06}
{"train_loss": 0.05105391517281532, "global_step": 231196, "epoch": 2597, "lr": 5.347614035914134e-06}
{"train_loss": 0.027294235303997993, "global_step": 231197, "epoch": 2597, "lr": 5.347353179493658e-06}
{"train_loss": 0.05327468737959862, "global_step": 231198, "epoch": 2597, "lr": 5.347092329076192e-06}
{"train_loss": 0.04907176271080971, "global_step": 231199, "epoch": 2597, "lr": 5.346831484661763e-06}
{"train_loss": 0.029239658266305923, "global_step": 231200, "epoch": 2597, "lr": 5.346570646250409e-06}
{"train_loss": 0.053285688161849976, "global_step": 231201, "epoch": 2597, "lr": 5.34630981384216e-06}
{"train_loss": 0.03351520374417305, "global_step": 231202, "epoch": 2597, "lr": 5.346048987437063e-06}
{"train_loss": 0.03202527016401291, "global_step": 231203, "epoch": 2597, "lr": 5.345788167035137e-06}
{"train_loss": 0.08343303948640823, "global_step": 231204, "epoch": 2597, "lr": 5.345527352636437e-06}
{"train_loss": 0.042115580290555954, "global_step": 231205, "epoch": 2597, "lr": 5.345266544240973e-06}
{"train_loss": 0.03894244506955147, "global_step": 231206, "epoch": 2597, "lr": 5.3450057418488075e-06}
{"train_loss": 0.04962010681629181, "global_step": 231207, "epoch": 2597, "lr": 5.3447449454599505e-06}
{"train_loss": 0.0740765854716301, "global_step": 231208, "epoch": 2597, "lr": 5.344484155074447e-06}
{"train_loss": 0.047332972288131714, "global_step": 231209, "epoch": 2597, "lr": 5.344223370692341e-06}
{"train_loss": 0.07640477269887924, "global_step": 231210, "epoch": 2597, "lr": 5.34396259231365e-06}
{"train_loss": 0.05241701751947403, "global_step": 231211, "epoch": 2597, "lr": 5.343701819938435e-06}
{"train_loss": 0.06370346248149872, "global_step": 231212, "epoch": 2597, "lr": 5.3434410535667e-06}
{"train_loss": 0.052085500210523605, "global_step": 231213, "epoch": 2597, "lr": 5.343180293198508e-06}
{"train_loss": 0.0865795686841011, "global_step": 231214, "epoch": 2597, "lr": 5.342919538833868e-06}
{"train_loss": 0.05712878704071045, "global_step": 231215, "epoch": 2597, "lr": 5.342658790472843e-06}
{"train_loss": 0.04330639913678169, "global_step": 231216, "epoch": 2597, "lr": 5.3423980481154375e-06}
{"train_loss": 0.06440548598766327, "global_step": 231217, "epoch": 2597, "lr": 5.3421373117617134e-06}
{"train_loss": 0.055556874722242355, "global_step": 231218, "epoch": 2597, "lr": 5.341876581411698e-06}
{"train_loss": 0.037130676209926605, "global_step": 231219, "epoch": 2597, "lr": 5.341615857065413e-06}
{"train_loss": 0.09511204063892365, "global_step": 231220, "epoch": 2597, "lr": 5.341355138722909e-06}
{"train_loss": 0.052183177866292806, "global_step": 231221, "epoch": 2597, "lr": 5.341094426384208e-06, "val_loss": 9.28018856048584}
{"train_loss": 0.01920187845826149, "global_step": 231222, "epoch": 2598, "lr": 5.340833720049365e-06}
{"train_loss": 0.02379048615694046, "global_step": 231223, "epoch": 2598, "lr": 5.340573019718398e-06}
{"train_loss": 0.05971372500061989, "global_step": 231224, "epoch": 2598, "lr": 5.340312325391339e-06}
{"train_loss": 0.05446014553308487, "global_step": 231225, "epoch": 2598, "lr": 5.340051637068238e-06}
{"train_loss": 0.0633496344089508, "global_step": 231226, "epoch": 2598, "lr": 5.339790954749119e-06}
{"train_loss": 0.07288769632577896, "global_step": 231227, "epoch": 2598, "lr": 5.339530278434013e-06}
{"train_loss": 0.042462386190891266, "global_step": 231228, "epoch": 2598, "lr": 5.339269608122982e-06}
{"train_loss": 0.11069270968437195, "global_step": 231229, "epoch": 2598, "lr": 5.3390089438160264e-06}
{"train_loss": 0.031831804662942886, "global_step": 231230, "epoch": 2598, "lr": 5.338748285513206e-06}
{"train_loss": 0.04134301096200943, "global_step": 231231, "epoch": 2598, "lr": 5.338487633214534e-06}
{"train_loss": 0.016134178265929222, "global_step": 231232, "epoch": 2598, "lr": 5.338226986920075e-06}
{"train_loss": 0.03173542022705078, "global_step": 231233, "epoch": 2598, "lr": 5.33796634662983e-06}
{"train_loss": 0.0490945465862751, "global_step": 231234, "epoch": 2598, "lr": 5.337705712343866e-06}
{"train_loss": 0.024720020592212677, "global_step": 231235, "epoch": 2598, "lr": 5.337445084062193e-06}
{"train_loss": 0.014209678396582603, "global_step": 231236, "epoch": 2598, "lr": 5.337184461784866e-06}
{"train_loss": 0.05564218387007713, "global_step": 231237, "epoch": 2598, "lr": 5.336923845511899e-06}
{"train_loss": 0.017088593915104866, "global_step": 231238, "epoch": 2598, "lr": 5.336663235243345e-06}
{"train_loss": 0.054136987775564194, "global_step": 231239, "epoch": 2598, "lr": 5.336402630979226e-06}
{"train_loss": 0.07214991748332977, "global_step": 231240, "epoch": 2598, "lr": 5.336142032719593e-06}
{"train_loss": 0.04583766311407089, "global_step": 231241, "epoch": 2598, "lr": 5.335881440464463e-06}
{"train_loss": 0.02495790831744671, "global_step": 231242, "epoch": 2598, "lr": 5.335620854213885e-06}
{"train_loss": 0.038788992911577225, "global_step": 231243, "epoch": 2598, "lr": 5.335360273967882e-06}
{"train_loss": 0.015720829367637634, "global_step": 231244, "epoch": 2598, "lr": 5.335099699726492e-06}
{"train_loss": 0.101649709045887, "global_step": 231245, "epoch": 2598, "lr": 5.334839131489766e-06}
{"train_loss": 0.05884391441941261, "global_step": 231246, "epoch": 2598, "lr": 5.33457856925772e-06}
{"train_loss": 0.03592073544859886, "global_step": 231247, "epoch": 2598, "lr": 5.334318013030398e-06}
{"train_loss": 0.06464924663305283, "global_step": 231248, "epoch": 2598, "lr": 5.334057462807828e-06}
{"train_loss": 0.060967396944761276, "global_step": 231249, "epoch": 2598, "lr": 5.333796918590056e-06}
{"train_loss": 0.03248224034905434, "global_step": 231250, "epoch": 2598, "lr": 5.333536380377102e-06}
{"train_loss": 0.06565159559249878, "global_step": 231251, "epoch": 2598, "lr": 5.333275848169023e-06}
{"train_loss": 0.01602502539753914, "global_step": 231252, "epoch": 2598, "lr": 5.333015321965829e-06}
{"train_loss": 0.036596134305000305, "global_step": 231253, "epoch": 2598, "lr": 5.332754801767575e-06}
{"train_loss": 0.055644549429416656, "global_step": 231254, "epoch": 2598, "lr": 5.332494287574274e-06}
{"train_loss": 0.04680297523736954, "global_step": 231255, "epoch": 2598, "lr": 5.332233779385992e-06}
{"train_loss": 0.048264920711517334, "global_step": 231256, "epoch": 2598, "lr": 5.331973277202745e-06}
{"train_loss": 0.05208586901426315, "global_step": 231257, "epoch": 2598, "lr": 5.331712781024556e-06}
{"train_loss": 0.08689980208873749, "global_step": 231258, "epoch": 2598, "lr": 5.331452290851485e-06}
{"train_loss": 0.029964705929160118, "global_step": 231259, "epoch": 2598, "lr": 5.33119180668355e-06}
{"train_loss": 0.03740162029862404, "global_step": 231260, "epoch": 2598, "lr": 5.330931328520799e-06}
{"train_loss": 0.0661756619811058, "global_step": 231261, "epoch": 2598, "lr": 5.330670856363257e-06}
{"train_loss": 0.03365790843963623, "global_step": 231262, "epoch": 2598, "lr": 5.330410390210955e-06}
{"train_loss": 0.05018797889351845, "global_step": 231263, "epoch": 2598, "lr": 5.330149930063938e-06}
{"train_loss": 0.03810660168528557, "global_step": 231264, "epoch": 2598, "lr": 5.32988947592224e-06}
{"train_loss": 0.05888834223151207, "global_step": 231265, "epoch": 2598, "lr": 5.3296290277858875e-06}
{"train_loss": 0.042920470237731934, "global_step": 231266, "epoch": 2598, "lr": 5.329368585654937e-06}
{"train_loss": 0.03679916262626648, "global_step": 231267, "epoch": 2598, "lr": 5.329108149529394e-06}
{"train_loss": 0.07104957848787308, "global_step": 231268, "epoch": 2598, "lr": 5.328847719409313e-06}
{"train_loss": 0.0421655997633934, "global_step": 231269, "epoch": 2598, "lr": 5.328587295294724e-06}
{"train_loss": 0.03607456013560295, "global_step": 231270, "epoch": 2598, "lr": 5.328326877185663e-06}
{"train_loss": 0.04065970331430435, "global_step": 231271, "epoch": 2598, "lr": 5.32806646508216e-06}
{"train_loss": 0.11438105255365372, "global_step": 231272, "epoch": 2598, "lr": 5.327806058984264e-06}
{"train_loss": 0.04376200959086418, "global_step": 231273, "epoch": 2598, "lr": 5.3275456588919855e-06}
{"train_loss": 0.03743404522538185, "global_step": 231274, "epoch": 2598, "lr": 5.3272852648053874e-06}
{"train_loss": 0.05311288312077522, "global_step": 231275, "epoch": 2598, "lr": 5.327024876724479e-06}
{"train_loss": 0.05332096293568611, "global_step": 231276, "epoch": 2598, "lr": 5.3267644946493165e-06}
{"train_loss": 0.06404910236597061, "global_step": 231277, "epoch": 2598, "lr": 5.3265041185799224e-06}
{"train_loss": 0.027992499992251396, "global_step": 231278, "epoch": 2598, "lr": 5.326243748516341e-06}
{"train_loss": 0.0711938738822937, "global_step": 231279, "epoch": 2598, "lr": 5.325983384458588e-06}
{"train_loss": 0.05880053713917732, "global_step": 231280, "epoch": 2598, "lr": 5.325723026406726e-06}
{"train_loss": 0.025321003049612045, "global_step": 231281, "epoch": 2598, "lr": 5.325462674360771e-06}
{"train_loss": 0.023913713172078133, "global_step": 231282, "epoch": 2598, "lr": 5.3252023283207555e-06}
{"train_loss": 0.06939143687486649, "global_step": 231283, "epoch": 2598, "lr": 5.324941988286736e-06}
{"train_loss": 0.04147130995988846, "global_step": 231284, "epoch": 2598, "lr": 5.324681654258723e-06}
{"train_loss": 0.05335365608334541, "global_step": 231285, "epoch": 2598, "lr": 5.324421326236767e-06}
{"train_loss": 0.08206232637166977, "global_step": 231286, "epoch": 2598, "lr": 5.324161004220895e-06}
{"train_loss": 0.06682765483856201, "global_step": 231287, "epoch": 2598, "lr": 5.323900688211153e-06}
{"train_loss": 0.03876971825957298, "global_step": 231288, "epoch": 2598, "lr": 5.323640378207556e-06}
{"train_loss": 0.046586524695158005, "global_step": 231289, "epoch": 2598, "lr": 5.32338007421016e-06}
{"train_loss": 0.055539991706609726, "global_step": 231290, "epoch": 2598, "lr": 5.323119776218987e-06}
{"train_loss": 0.054935749620199203, "global_step": 231291, "epoch": 2598, "lr": 5.322859484234083e-06}
{"train_loss": 0.05996505916118622, "global_step": 231292, "epoch": 2598, "lr": 5.322599198255462e-06}
{"train_loss": 0.06114863604307175, "global_step": 231293, "epoch": 2598, "lr": 5.3223389182831875e-06}
{"train_loss": 0.0522608608007431, "global_step": 231294, "epoch": 2598, "lr": 5.32207864431728e-06}
{"train_loss": 0.027016064152121544, "global_step": 231295, "epoch": 2598, "lr": 5.321818376357762e-06}
{"train_loss": 0.025698594748973846, "global_step": 231296, "epoch": 2598, "lr": 5.32155811440469e-06}
{"train_loss": 0.044911980628967285, "global_step": 231297, "epoch": 2598, "lr": 5.321297858458085e-06}
{"train_loss": 0.01671699807047844, "global_step": 231298, "epoch": 2598, "lr": 5.321037608517992e-06}
{"train_loss": 0.061939533799886703, "global_step": 231299, "epoch": 2598, "lr": 5.320777364584434e-06}
{"train_loss": 0.028470776975154877, "global_step": 231300, "epoch": 2598, "lr": 5.320517126657465e-06}
{"train_loss": 0.03687380254268646, "global_step": 231301, "epoch": 2598, "lr": 5.320256894737091e-06}
{"train_loss": 0.0682200938463211, "global_step": 231302, "epoch": 2598, "lr": 5.31999666882338e-06}
{"train_loss": 0.037029676139354706, "global_step": 231303, "epoch": 2598, "lr": 5.319736448916341e-06}
{"train_loss": 0.07117915153503418, "global_step": 231304, "epoch": 2598, "lr": 5.319476235016024e-06}
{"train_loss": 0.030499791726469994, "global_step": 231305, "epoch": 2598, "lr": 5.3192160271224535e-06}
{"train_loss": 0.07598425447940826, "global_step": 231306, "epoch": 2598, "lr": 5.318955825235677e-06}
{"train_loss": 0.014940249733626842, "global_step": 231307, "epoch": 2598, "lr": 5.318695629355713e-06}
{"train_loss": 0.04618972912430763, "global_step": 231308, "epoch": 2598, "lr": 5.318435439482617e-06}
{"train_loss": 0.018950242549180984, "global_step": 231309, "epoch": 2598, "lr": 5.318175255616398e-06}
{"train_loss": 0.04739370884562141, "global_step": 231310, "epoch": 2598, "lr": 5.317915077757119e-06, "val_loss": 9.279478073120117}
{"train_loss": 0.1194252073764801, "global_step": 231311, "epoch": 2599, "lr": 5.317654905904796e-06}
{"train_loss": 0.054677337408065796, "global_step": 231312, "epoch": 2599, "lr": 5.317394740059473e-06}
{"train_loss": 0.04544110596179962, "global_step": 231313, "epoch": 2599, "lr": 5.317134580221172e-06}
{"train_loss": 0.029971839860081673, "global_step": 231314, "epoch": 2599, "lr": 5.31687442638995e-06}
{"train_loss": 0.0658508837223053, "global_step": 231315, "epoch": 2599, "lr": 5.3166142785658235e-06}
{"train_loss": 0.02648754045367241, "global_step": 231316, "epoch": 2599, "lr": 5.316354136748836e-06}
{"train_loss": 0.01597198098897934, "global_step": 231317, "epoch": 2599, "lr": 5.316094000939015e-06}
{"train_loss": 0.0671347975730896, "global_step": 231318, "epoch": 2599, "lr": 5.3158338711364e-06}
{"train_loss": 0.05101026967167854, "global_step": 231319, "epoch": 2599, "lr": 5.315573747341035e-06}
{"train_loss": 0.044367868453264236, "global_step": 231320, "epoch": 2599, "lr": 5.315313629552937e-06}
{"train_loss": 0.07736850529909134, "global_step": 231321, "epoch": 2599, "lr": 5.315053517772162e-06}
{"train_loss": 0.027943268418312073, "global_step": 231322, "epoch": 2599, "lr": 5.314793411998725e-06}
{"train_loss": 0.04946974664926529, "global_step": 231323, "epoch": 2599, "lr": 5.3145333122326726e-06}
{"train_loss": 0.040534600615501404, "global_step": 231324, "epoch": 2599, "lr": 5.314273218474031e-06}
{"train_loss": 0.03259339928627014, "global_step": 231325, "epoch": 2599, "lr": 5.3140131307228506e-06}
{"train_loss": 0.05774063989520073, "global_step": 231326, "epoch": 2599, "lr": 5.313753048979148e-06}
{"train_loss": 0.09855768084526062, "global_step": 231327, "epoch": 2599, "lr": 5.313492973242973e-06}
{"train_loss": 0.035055723041296005, "global_step": 231328, "epoch": 2599, "lr": 5.313232903514348e-06}
{"train_loss": 0.060102637857198715, "global_step": 231329, "epoch": 2599, "lr": 5.3129728397933235e-06}
{"train_loss": 0.08512228727340698, "global_step": 231330, "epoch": 2599, "lr": 5.312712782079909e-06}
{"train_loss": 0.05032674968242645, "global_step": 231331, "epoch": 2599, "lr": 5.312452730374173e-06}
{"train_loss": 0.030139438807964325, "global_step": 231332, "epoch": 2599, "lr": 5.31219268467612e-06}
{"train_loss": 0.06701043993234634, "global_step": 231333, "epoch": 2599, "lr": 5.311932644985812e-06}
{"train_loss": 0.09590248018503189, "global_step": 231334, "epoch": 2599, "lr": 5.311672611303264e-06}
{"train_loss": 0.045200247317552567, "global_step": 231335, "epoch": 2599, "lr": 5.311412583628511e-06}
{"train_loss": 0.06332084536552429, "global_step": 231336, "epoch": 2599, "lr": 5.311152561961602e-06}
{"train_loss": 0.05417952686548233, "global_step": 231337, "epoch": 2599, "lr": 5.3108925463025535e-06}
{"train_loss": 0.08046510070562363, "global_step": 231338, "epoch": 2599, "lr": 5.3106325366514224e-06}
{"train_loss": 0.028751082718372345, "global_step": 231339, "epoch": 2599, "lr": 5.310372533008217e-06}
{"train_loss": 0.05873379856348038, "global_step": 231340, "epoch": 2599, "lr": 5.3101125353730075e-06}
{"train_loss": 0.05011102184653282, "global_step": 231341, "epoch": 2599, "lr": 5.3098525437457914e-06}
{"train_loss": 0.041491515934467316, "global_step": 231342, "epoch": 2599, "lr": 5.30959255812663e-06}
{"train_loss": 0.03946203365921974, "global_step": 231343, "epoch": 2599, "lr": 5.309332578515541e-06}
{"train_loss": 0.08765154331922531, "global_step": 231344, "epoch": 2599, "lr": 5.30907260491258e-06}
{"train_loss": 0.06241174787282944, "global_step": 231345, "epoch": 2599, "lr": 5.308812637317756e-06}
{"train_loss": 0.05740118771791458, "global_step": 231346, "epoch": 2599, "lr": 5.308552675731132e-06}
{"train_loss": 0.02989371307194233, "global_step": 231347, "epoch": 2599, "lr": 5.308292720152714e-06}
{"train_loss": 0.03803005442023277, "global_step": 231348, "epoch": 2599, "lr": 5.30803277058256e-06}
{"train_loss": 0.04219375178217888, "global_step": 231349, "epoch": 2599, "lr": 5.3077728270206905e-06}
{"train_loss": 0.03708431497216225, "global_step": 231350, "epoch": 2599, "lr": 5.307512889467153e-06}
{"train_loss": 0.05341624468564987, "global_step": 231351, "epoch": 2599, "lr": 5.307252957921971e-06}
{"train_loss": 0.07806748896837234, "global_step": 231352, "epoch": 2599, "lr": 5.306993032385188e-06}
{"train_loss": 0.0342346727848053, "global_step": 231353, "epoch": 2599, "lr": 5.3067331128568265e-06}
{"train_loss": 0.06871114671230316, "global_step": 231354, "epoch": 2599, "lr": 5.306473199336931e-06}
{"train_loss": 0.05565646290779114, "global_step": 231355, "epoch": 2599, "lr": 5.3062132918255456e-06}
{"train_loss": 0.08247614651918411, "global_step": 231356, "epoch": 2599, "lr": 5.305953390322688e-06}
{"train_loss": 0.10631585866212845, "global_step": 231357, "epoch": 2599, "lr": 5.305693494828407e-06}
{"train_loss": 0.04676755517721176, "global_step": 231358, "epoch": 2599, "lr": 5.305433605342719e-06}
{"train_loss": 0.04690997302532196, "global_step": 231359, "epoch": 2599, "lr": 5.305173721865681e-06}
{"train_loss": 0.02498549409210682, "global_step": 231360, "epoch": 2599, "lr": 5.304913844397313e-06}
{"train_loss": 0.06181672587990761, "global_step": 231361, "epoch": 2599, "lr": 5.304653972937657e-06}
{"train_loss": 0.05912503972649574, "global_step": 231362, "epoch": 2599, "lr": 5.3043941074867436e-06}
{"train_loss": 0.05306074023246765, "global_step": 231363, "epoch": 2599, "lr": 5.304134248044612e-06}
{"train_loss": 0.031485941261053085, "global_step": 231364, "epoch": 2599, "lr": 5.303874394611291e-06}
{"train_loss": 0.12302597612142563, "global_step": 231365, "epoch": 2599, "lr": 5.303614547186825e-06}
{"train_loss": 0.04981295391917229, "global_step": 231366, "epoch": 2599, "lr": 5.303354705771235e-06}
{"train_loss": 0.039820265024900436, "global_step": 231367, "epoch": 2599, "lr": 5.303094870364572e-06}
{"train_loss": 0.07285293191671371, "global_step": 231368, "epoch": 2599, "lr": 5.302835040966858e-06}
{"train_loss": 0.08689969033002853, "global_step": 231369, "epoch": 2599, "lr": 5.302575217578137e-06}
{"train_loss": 0.06616557389497757, "global_step": 231370, "epoch": 2599, "lr": 5.302315400198432e-06}
{"train_loss": 0.0696866512298584, "global_step": 231371, "epoch": 2599, "lr": 5.3020555888277975e-06}
{"train_loss": 0.03666933998465538, "global_step": 231372, "epoch": 2599, "lr": 5.301795783466257e-06}
{"train_loss": 0.01593627594411373, "global_step": 231373, "epoch": 2599, "lr": 5.301535984113837e-06}
{"train_loss": 0.021390026435256004, "global_step": 231374, "epoch": 2599, "lr": 5.301276190770588e-06}
{"train_loss": 0.06704431027173996, "global_step": 231375, "epoch": 2599, "lr": 5.301016403436526e-06}
{"train_loss": 0.0604638010263443, "global_step": 231376, "epoch": 2599, "lr": 5.300756622111708e-06}
{"train_loss": 0.05068068206310272, "global_step": 231377, "epoch": 2599, "lr": 5.300496846796149e-06}
{"train_loss": 0.06070482358336449, "global_step": 231378, "epoch": 2599, "lr": 5.3002370774899055e-06}
{"train_loss": 0.042425282299518585, "global_step": 231379, "epoch": 2599, "lr": 5.2999773141929886e-06}
{"train_loss": 0.06426035612821579, "global_step": 231380, "epoch": 2599, "lr": 5.299717556905459e-06}
{"train_loss": 0.05805250257253647, "global_step": 231381, "epoch": 2599, "lr": 5.299457805627323e-06}
{"train_loss": 0.04741668701171875, "global_step": 231382, "epoch": 2599, "lr": 5.29919806035864e-06}
{"train_loss": 0.021830877289175987, "global_step": 231383, "epoch": 2599, "lr": 5.298938321099428e-06}
{"train_loss": 0.025261297821998596, "global_step": 231384, "epoch": 2599, "lr": 5.298678587849737e-06}
{"train_loss": 0.08004634082317352, "global_step": 231385, "epoch": 2599, "lr": 5.298418860609583e-06}
{"train_loss": 0.02822881005704403, "global_step": 231386, "epoch": 2599, "lr": 5.298159139379022e-06}
{"train_loss": 0.06252612173557281, "global_step": 231387, "epoch": 2599, "lr": 5.297899424158065e-06}
{"train_loss": 0.1011415496468544, "global_step": 231388, "epoch": 2599, "lr": 5.2976397149467785e-06}
{"train_loss": 0.03308944031596184, "global_step": 231389, "epoch": 2599, "lr": 5.297380011745162e-06}
{"train_loss": 0.01847747340798378, "global_step": 231390, "epoch": 2599, "lr": 5.297120314553272e-06}
{"train_loss": 0.02038971334695816, "global_step": 231391, "epoch": 2599, "lr": 5.296860623371153e-06}
{"train_loss": 0.048740558326244354, "global_step": 231392, "epoch": 2599, "lr": 5.296600938198809e-06}
{"train_loss": 0.03966310992836952, "global_step": 231393, "epoch": 2599, "lr": 5.296341259036308e-06}
{"train_loss": 0.048337336629629135, "global_step": 231394, "epoch": 2599, "lr": 5.296081585883656e-06}
{"train_loss": 0.045944903045892715, "global_step": 231395, "epoch": 2599, "lr": 5.295821918740912e-06}
{"train_loss": 0.04211993142962456, "global_step": 231396, "epoch": 2599, "lr": 5.295562257608089e-06}
{"train_loss": 0.03067426010966301, "global_step": 231397, "epoch": 2599, "lr": 5.295302602485241e-06}
{"train_loss": 0.053080491721630096, "global_step": 231398, "epoch": 2599, "lr": 5.295042953372387e-06}
{"train_loss": 0.05354333139453711, "global_step": 231399, "epoch": 2599, "lr": 5.29478331026958e-06, "val_loss": 9.105286598205566}
{"train_loss": 0.08256221562623978, "global_step": 231400, "epoch": 2600, "lr": 5.294523673176832e-06}
{"train_loss": 0.051739275455474854, "global_step": 231401, "epoch": 2600, "lr": 5.2942640420942045e-06}
{"train_loss": 0.03305090218782425, "global_step": 231402, "epoch": 2600, "lr": 5.2940044170217026e-06}
{"train_loss": 0.034322891384363174, "global_step": 231403, "epoch": 2600, "lr": 5.293744797959393e-06}
{"train_loss": 0.040179114788770676, "global_step": 231404, "epoch": 2600, "lr": 5.293485184907282e-06}
{"train_loss": 0.03594342619180679, "global_step": 231405, "epoch": 2600, "lr": 5.293225577865424e-06}
{"train_loss": 0.04402438551187515, "global_step": 231406, "epoch": 2600, "lr": 5.2929659768338415e-06}
{"train_loss": 0.0479898676276207, "global_step": 231407, "epoch": 2600, "lr": 5.2927063818125845e-06}
{"train_loss": 0.06633761525154114, "global_step": 231408, "epoch": 2600, "lr": 5.292446792801664e-06}
{"train_loss": 0.03418782725930214, "global_step": 231409, "epoch": 2600, "lr": 5.292187209801142e-06}
{"train_loss": 0.07188836485147476, "global_step": 231410, "epoch": 2600, "lr": 5.291927632811045e-06}
{"train_loss": 0.03287153318524361, "global_step": 231411, "epoch": 2600, "lr": 5.291668061831384e-06}
{"train_loss": 0.03873531147837639, "global_step": 231412, "epoch": 2600, "lr": 5.291408496862232e-06}
{"train_loss": 0.018483491614460945, "global_step": 231413, "epoch": 2600, "lr": 5.291148937903589e-06}
{"train_loss": 0.03468018025159836, "global_step": 231414, "epoch": 2600, "lr": 5.290889384955522e-06}
{"train_loss": 0.0843253955245018, "global_step": 231415, "epoch": 2600, "lr": 5.290629838018035e-06}
{"train_loss": 0.05726214125752449, "global_step": 231416, "epoch": 2600, "lr": 5.29037029709119e-06}
{"train_loss": 0.07681673765182495, "global_step": 231417, "epoch": 2600, "lr": 5.290110762174999e-06}
{"train_loss": 0.04201686382293701, "global_step": 231418, "epoch": 2600, "lr": 5.289851233269516e-06}
{"train_loss": 0.02319623902440071, "global_step": 231419, "epoch": 2600, "lr": 5.289591710374759e-06}
{"train_loss": 0.04084465280175209, "global_step": 231420, "epoch": 2600, "lr": 5.289332193490782e-06}
{"train_loss": 0.10684554278850555, "global_step": 231421, "epoch": 2600, "lr": 5.289072682617602e-06}
{"train_loss": 0.07752224802970886, "global_step": 231422, "epoch": 2600, "lr": 5.2888131777552645e-06}
{"train_loss": 0.04991839826107025, "global_step": 231423, "epoch": 2600, "lr": 5.2885536789037955e-06}
{"train_loss": 0.053523559123277664, "global_step": 231424, "epoch": 2600, "lr": 5.288294186063248e-06}
{"train_loss": 0.05317050591111183, "global_step": 231425, "epoch": 2600, "lr": 5.288034699233629e-06}
{"train_loss": 0.06547807157039642, "global_step": 231426, "epoch": 2600, "lr": 5.287775218414992e-06}
{"train_loss": 0.065595343708992, "global_step": 231427, "epoch": 2600, "lr": 5.28751574360738e-06}
{"train_loss": 0.03866049647331238, "global_step": 231428, "epoch": 2600, "lr": 5.287256274810803e-06}
{"train_loss": 0.03938715159893036, "global_step": 231429, "epoch": 2600, "lr": 5.286996812025319e-06}
{"train_loss": 0.035760972648859024, "global_step": 231430, "epoch": 2600, "lr": 5.286737355250948e-06}
{"train_loss": 0.05913648381829262, "global_step": 231431, "epoch": 2600, "lr": 5.2864779044877356e-06}
{"train_loss": 0.027396732941269875, "global_step": 231432, "epoch": 2600, "lr": 5.286218459735704e-06}
{"train_loss": 0.05865102261304855, "global_step": 231433, "epoch": 2600, "lr": 5.285959020994902e-06}
{"train_loss": 0.0693214014172554, "global_step": 231434, "epoch": 2600, "lr": 5.285699588265352e-06}
{"train_loss": 0.07212832570075989, "global_step": 231435, "epoch": 2600, "lr": 5.285440161547101e-06}
{"train_loss": 0.02696961909532547, "global_step": 231436, "epoch": 2600, "lr": 5.2851807408401735e-06}
{"train_loss": 0.030470918864011765, "global_step": 231437, "epoch": 2600, "lr": 5.284921326144609e-06}
{"train_loss": 0.019133776426315308, "global_step": 231438, "epoch": 2600, "lr": 5.284661917460437e-06}
{"train_loss": 0.04816846549510956, "global_step": 231439, "epoch": 2600, "lr": 5.284402514787712e-06}
{"train_loss": 0.016507666558027267, "global_step": 231440, "epoch": 2600, "lr": 5.284143118126439e-06}
{"train_loss": 0.03921304643154144, "global_step": 231441, "epoch": 2600, "lr": 5.283883727476679e-06}
{"train_loss": 0.06551342457532883, "global_step": 231442, "epoch": 2600, "lr": 5.283624342838445e-06}
{"train_loss": 0.0519697368144989, "global_step": 231443, "epoch": 2600, "lr": 5.28336496421179e-06}
{"train_loss": 0.0562107227742672, "global_step": 231444, "epoch": 2600, "lr": 5.283105591596737e-06}
{"train_loss": 0.04563156142830849, "global_step": 231445, "epoch": 2600, "lr": 5.282846224993332e-06}
{"train_loss": 0.04987300932407379, "global_step": 231446, "epoch": 2600, "lr": 5.282586864401595e-06}
{"train_loss": 0.09636159986257553, "global_step": 231447, "epoch": 2600, "lr": 5.282327509821577e-06}
{"train_loss": 0.03636316955089569, "global_step": 231448, "epoch": 2600, "lr": 5.282068161253295e-06}
{"train_loss": 0.06240726262331009, "global_step": 231449, "epoch": 2600, "lr": 5.2818088186968095e-06}
{"train_loss": 0.07718517631292343, "global_step": 231450, "epoch": 2600, "lr": 5.281549482152131e-06}
{"train_loss": 0.05440828204154968, "global_step": 231451, "epoch": 2600, "lr": 5.281290151619301e-06}
{"train_loss": 0.051211174577474594, "global_step": 231452, "epoch": 2600, "lr": 5.28103082709836e-06}
{"train_loss": 0.10081285238265991, "global_step": 231453, "epoch": 2600, "lr": 5.280771508589332e-06}
{"train_loss": 0.057483166456222534, "global_step": 231454, "epoch": 2600, "lr": 5.280512196092274e-06}
{"train_loss": 0.028560124337673187, "global_step": 231455, "epoch": 2600, "lr": 5.28025288960719e-06}
{"train_loss": 0.1057107076048851, "global_step": 231456, "epoch": 2600, "lr": 5.279993589134141e-06}
{"train_loss": 0.07970859855413437, "global_step": 231457, "epoch": 2600, "lr": 5.279734294673144e-06}
{"train_loss": 0.04778938367962837, "global_step": 231458, "epoch": 2600, "lr": 5.279475006224255e-06}
{"train_loss": 0.06523525714874268, "global_step": 231459, "epoch": 2600, "lr": 5.279215723787479e-06}
{"train_loss": 0.047589533030986786, "global_step": 231460, "epoch": 2600, "lr": 5.2789564473628815e-06}
{"train_loss": 0.06619758158922195, "global_step": 231461, "epoch": 2600, "lr": 5.27869717695047e-06}
{"train_loss": 0.056738581508398056, "global_step": 231462, "epoch": 2600, "lr": 5.278437912550294e-06}
{"train_loss": 0.08966369181871414, "global_step": 231463, "epoch": 2600, "lr": 5.278178654162397e-06}
{"train_loss": 0.04701007530093193, "global_step": 231464, "epoch": 2600, "lr": 5.2779194017867965e-06}
{"train_loss": 0.04867476969957352, "global_step": 231465, "epoch": 2600, "lr": 5.277660155423542e-06}
{"train_loss": 0.04509760066866875, "global_step": 231466, "epoch": 2600, "lr": 5.27740091507265e-06}
{"train_loss": 0.03597212955355644, "global_step": 231467, "epoch": 2600, "lr": 5.277141680734177e-06}
{"train_loss": 0.021832961589097977, "global_step": 231468, "epoch": 2600, "lr": 5.276882452408138e-06}
{"train_loss": 0.02723904885351658, "global_step": 231469, "epoch": 2600, "lr": 5.276623230094591e-06}
{"train_loss": 0.033688440918922424, "global_step": 231470, "epoch": 2600, "lr": 5.276364013793539e-06}
{"train_loss": 0.05531006306409836, "global_step": 231471, "epoch": 2600, "lr": 5.27610480350505e-06}
{"train_loss": 0.06798192858695984, "global_step": 231472, "epoch": 2600, "lr": 5.2758455992291346e-06}
{"train_loss": 0.05680137500166893, "global_step": 231473, "epoch": 2600, "lr": 5.275586400965843e-06}
{"train_loss": 0.07138639688491821, "global_step": 231474, "epoch": 2600, "lr": 5.275327208715197e-06}
{"train_loss": 0.0703371912240982, "global_step": 231475, "epoch": 2600, "lr": 5.275068022477242e-06}
{"train_loss": 0.043418038636446, "global_step": 231476, "epoch": 2600, "lr": 5.2748088422520046e-06}
{"train_loss": 0.09608151763677597, "global_step": 231477, "epoch": 2600, "lr": 5.2745496680395355e-06}
{"train_loss": 0.08869375288486481, "global_step": 231478, "epoch": 2600, "lr": 5.27429049983984e-06}
{"train_loss": 0.04575272649526596, "global_step": 231479, "epoch": 2600, "lr": 5.274031337652991e-06}
{"train_loss": 0.023189570754766464, "global_step": 231480, "epoch": 2600, "lr": 5.273772181478986e-06}
{"train_loss": 0.03720100224018097, "global_step": 231481, "epoch": 2600, "lr": 5.273513031317878e-06}
{"train_loss": 0.0068824319168925285, "global_step": 231482, "epoch": 2600, "lr": 5.273253887169716e-06}
{"train_loss": 0.04176807031035423, "global_step": 231483, "epoch": 2600, "lr": 5.272994749034515e-06}
{"train_loss": 0.06553343683481216, "global_step": 231484, "epoch": 2600, "lr": 5.272735616912305e-06}
{"train_loss": 0.09154915809631348, "global_step": 231485, "epoch": 2600, "lr": 5.272476490803141e-06}
{"train_loss": 0.04187848046422005, "global_step": 231486, "epoch": 2600, "lr": 5.272217370707039e-06}
{"train_loss": 0.04626716300845146, "global_step": 231487, "epoch": 2600, "lr": 5.271958256624055e-06}
{"train_loss": 0.052578636612533855, "global_step": 231488, "epoch": 2600, "lr": 5.271699148554204e-06, "train/sim_max_reward_0": 0.48850434310865115, "train/sim_max_reward_1": 0.48119484705033855, "train/sim_max_reward_2": 0.5648821090371995, "train/sim_max_reward_3": 0.9119442599007158, "train/sim_max_reward_4": 0.5910078446821484, "train/sim_max_reward_5": 0.5855359538575904, "test/sim_max_reward_4500000": 0.4174942252256205, "test/sim_max_reward_4500001": 0.8671530902762846, "test/sim_max_reward_4500002": 0.9162958474703244, "test/sim_max_reward_4500003": 0.47574076228950063, "test/sim_max_reward_4500004": 0.9985964827061822, "test/sim_max_reward_4500005": 0.3973241730154948, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.3205028859831928, "test/sim_max_reward_4500009": 0.04947027001929232, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.417798764110746, "test/sim_max_reward_4500014": 0.962055665027847, "test/sim_max_reward_4500015": 0.9525558335675353, "test/sim_max_reward_4500016": 0.5802905273419865, "test/sim_max_reward_4500017": 0.6825857554299108, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.5472983930511919, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9873567901875913, "test/sim_max_reward_4500023": 0.6168655969203353, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.7401054755301973, "test/sim_max_reward_4500026": 0.23270194252749113, "test/sim_max_reward_4500027": 0.9965209523854744, "test/sim_max_reward_4500028": 0.525196586611833, "test/sim_max_reward_4500029": 0.67757611125813, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.7233054713114581, "test/sim_max_reward_4500033": 0.916476517137561, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.49139863881443524, "test/sim_max_reward_4500036": 0.9518670995284363, "test/sim_max_reward_4500037": 0.9838281875023225, "test/sim_max_reward_4500038": 0.1625854804463737, "test/sim_max_reward_4500039": 0.9867500453128742, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.5523196105071687, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.6678490624578525, "test/sim_max_reward_4500045": 0.9731350123096533, "test/sim_max_reward_4500046": 0.4061890354483626, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9881269226111588, "test/sim_max_reward_4500049": 0.978352754444363, "train/mean_score": 0.6038448929394407, "test/mean_score": 0.6428733993753637, "val_loss": 9.177106857299805, "train_action_mse_error": 14.46121883392334}
{"train_loss": 0.05324454978108406, "global_step": 231489, "epoch": 2601, "lr": 5.2714400464975224e-06}
{"train_loss": 0.04865755885839462, "global_step": 231490, "epoch": 2601, "lr": 5.271180950454057e-06}
{"train_loss": 0.08244074136018753, "global_step": 231491, "epoch": 2601, "lr": 5.270921860423833e-06}
{"train_loss": 0.0916791632771492, "global_step": 231492, "epoch": 2601, "lr": 5.2706627764068925e-06}
{"train_loss": 0.04067603871226311, "global_step": 231493, "epoch": 2601, "lr": 5.270403698403259e-06}
{"train_loss": 0.0388166606426239, "global_step": 231494, "epoch": 2601, "lr": 5.2701446264129815e-06}
{"train_loss": 0.024707403033971786, "global_step": 231495, "epoch": 2601, "lr": 5.269885560436083e-06}
{"train_loss": 0.05422977730631828, "global_step": 231496, "epoch": 2601, "lr": 5.2696265004726074e-06}
{"train_loss": 0.017445411533117294, "global_step": 231497, "epoch": 2601, "lr": 5.269367446522578e-06}
{"train_loss": 0.05264903977513313, "global_step": 231498, "epoch": 2601, "lr": 5.269108398586037e-06}
{"train_loss": 0.10201849043369293, "global_step": 231499, "epoch": 2601, "lr": 5.2688493566630305e-06}
{"train_loss": 0.06264495104551315, "global_step": 231500, "epoch": 2601, "lr": 5.268590320753569e-06}
{"train_loss": 0.04847041144967079, "global_step": 231501, "epoch": 2601, "lr": 5.268331290857709e-06}
{"train_loss": 0.04524838924407959, "global_step": 231502, "epoch": 2601, "lr": 5.2680722669754715e-06}
{"train_loss": 0.03589983657002449, "global_step": 231503, "epoch": 2601, "lr": 5.2678132491069064e-06}
{"train_loss": 0.020460039377212524, "global_step": 231504, "epoch": 2601, "lr": 5.267554237252026e-06}
{"train_loss": 0.03806639462709427, "global_step": 231505, "epoch": 2601, "lr": 5.267295231410884e-06}
{"train_loss": 0.05054434388875961, "global_step": 231506, "epoch": 2601, "lr": 5.267036231583505e-06}
{"train_loss": 0.02205984853208065, "global_step": 231507, "epoch": 2601, "lr": 5.266777237769937e-06}
{"train_loss": 0.06643086671829224, "global_step": 231508, "epoch": 2601, "lr": 5.266518249970192e-06}
{"train_loss": 0.047152645885944366, "global_step": 231509, "epoch": 2601, "lr": 5.266259268184331e-06}
{"train_loss": 0.1009148508310318, "global_step": 231510, "epoch": 2601, "lr": 5.266000292412371e-06}
{"train_loss": 0.08303067088127136, "global_step": 231511, "epoch": 2601, "lr": 5.265741322654355e-06}
{"train_loss": 0.04209328070282936, "global_step": 231512, "epoch": 2601, "lr": 5.265482358910307e-06}
{"train_loss": 0.07185652107000351, "global_step": 231513, "epoch": 2601, "lr": 5.2652234011802805e-06}
{"train_loss": 0.02258605696260929, "global_step": 231514, "epoch": 2601, "lr": 5.264964449464288e-06}
{"train_loss": 0.03622337058186531, "global_step": 231515, "epoch": 2601, "lr": 5.264705503762385e-06}
{"train_loss": 0.0555557943880558, "global_step": 231516, "epoch": 2601, "lr": 5.264446564074593e-06}
{"train_loss": 0.059405118227005005, "global_step": 231517, "epoch": 2601, "lr": 5.264187630400946e-06}
{"train_loss": 0.04110551252961159, "global_step": 231518, "epoch": 2601, "lr": 5.263928702741494e-06}
{"train_loss": 0.03768373653292656, "global_step": 231519, "epoch": 2601, "lr": 5.263669781096248e-06}
{"train_loss": 0.04703468829393387, "global_step": 231520, "epoch": 2601, "lr": 5.263410865465274e-06}
{"train_loss": 0.06387322396039963, "global_step": 231521, "epoch": 2601, "lr": 5.263151955848583e-06}
{"train_loss": 0.01854829490184784, "global_step": 231522, "epoch": 2601, "lr": 5.2628930522462094e-06}
{"train_loss": 0.07202079147100449, "global_step": 231523, "epoch": 2601, "lr": 5.2626341546582024e-06}
{"train_loss": 0.02018587850034237, "global_step": 231524, "epoch": 2601, "lr": 5.262375263084579e-06}
{"train_loss": 0.05844685435295105, "global_step": 231525, "epoch": 2601, "lr": 5.262116377525395e-06}
{"train_loss": 0.06739936023950577, "global_step": 231526, "epoch": 2601, "lr": 5.261857497980666e-06}
{"train_loss": 0.07392486184835434, "global_step": 231527, "epoch": 2601, "lr": 5.261598624450443e-06}
{"train_loss": 0.04294315725564957, "global_step": 231528, "epoch": 2601, "lr": 5.261339756934752e-06}
{"train_loss": 0.06340043991804123, "global_step": 231529, "epoch": 2601, "lr": 5.2610808954336175e-06}
{"train_loss": 0.039431728422641754, "global_step": 231530, "epoch": 2601, "lr": 5.260822039947099e-06}
{"train_loss": 0.03997889161109924, "global_step": 231531, "epoch": 2601, "lr": 5.260563190475204e-06}
{"train_loss": 0.056571267545223236, "global_step": 231532, "epoch": 2601, "lr": 5.260304347017997e-06}
{"train_loss": 0.04062740132212639, "global_step": 231533, "epoch": 2601, "lr": 5.260045509575484e-06}
{"train_loss": 0.06434160470962524, "global_step": 231534, "epoch": 2601, "lr": 5.259786678147721e-06}
{"train_loss": 0.0686022937297821, "global_step": 231535, "epoch": 2601, "lr": 5.259527852734725e-06}
{"train_loss": 0.06220829114317894, "global_step": 231536, "epoch": 2601, "lr": 5.259269033336539e-06}
{"train_loss": 0.04994780570268631, "global_step": 231537, "epoch": 2601, "lr": 5.259010219953208e-06}
{"train_loss": 0.045160673558712006, "global_step": 231538, "epoch": 2601, "lr": 5.258751412584751e-06}
{"train_loss": 0.09943559020757675, "global_step": 231539, "epoch": 2601, "lr": 5.25849261123122e-06}
{"train_loss": 0.05425726994872093, "global_step": 231540, "epoch": 2601, "lr": 5.258233815892627e-06}
{"train_loss": 0.005562312435358763, "global_step": 231541, "epoch": 2601, "lr": 5.257975026569028e-06}
{"train_loss": 0.07273225486278534, "global_step": 231542, "epoch": 2601, "lr": 5.257716243260441e-06}
{"train_loss": 0.03417966142296791, "global_step": 231543, "epoch": 2601, "lr": 5.25745746596692e-06}
{"train_loss": 0.0773288682103157, "global_step": 231544, "epoch": 2601, "lr": 5.257198694688475e-06}
{"train_loss": 0.05765413120388985, "global_step": 231545, "epoch": 2601, "lr": 5.256939929425164e-06}
{"train_loss": 0.08443930000066757, "global_step": 231546, "epoch": 2601, "lr": 5.256681170177003e-06}
{"train_loss": 0.047030068933963776, "global_step": 231547, "epoch": 2601, "lr": 5.2564224169440514e-06}
{"train_loss": 0.07349769026041031, "global_step": 231548, "epoch": 2601, "lr": 5.25616366972631e-06}
{"train_loss": 0.0314340703189373, "global_step": 231549, "epoch": 2601, "lr": 5.255904928523847e-06}
{"train_loss": 0.05512220785021782, "global_step": 231550, "epoch": 2601, "lr": 5.255646193336672e-06}
{"train_loss": 0.05077473446726799, "global_step": 231551, "epoch": 2601, "lr": 5.25538746416484e-06}
{"train_loss": 0.07060413807630539, "global_step": 231552, "epoch": 2601, "lr": 5.2551287410083635e-06}
{"train_loss": 0.05492010340094566, "global_step": 231553, "epoch": 2601, "lr": 5.254870023867292e-06}
{"train_loss": 0.04417937248945236, "global_step": 231554, "epoch": 2601, "lr": 5.25461131274167e-06}
{"train_loss": 0.061793550848960876, "global_step": 231555, "epoch": 2601, "lr": 5.254352607631508e-06}
{"train_loss": 0.02999819628894329, "global_step": 231556, "epoch": 2601, "lr": 5.254093908536861e-06}
{"train_loss": 0.03337832912802696, "global_step": 231557, "epoch": 2601, "lr": 5.2538352154577476e-06}
{"train_loss": 0.017296284437179565, "global_step": 231558, "epoch": 2601, "lr": 5.253576528394216e-06}
{"train_loss": 0.035540226846933365, "global_step": 231559, "epoch": 2601, "lr": 5.253317847346289e-06}
{"train_loss": 0.0537281297147274, "global_step": 231560, "epoch": 2601, "lr": 5.253059172314023e-06}
{"train_loss": 0.039351340383291245, "global_step": 231561, "epoch": 2601, "lr": 5.252800503297434e-06}
{"train_loss": 0.06465090066194534, "global_step": 231562, "epoch": 2601, "lr": 5.252541840296549e-06}
{"train_loss": 0.06247960403561592, "global_step": 231563, "epoch": 2601, "lr": 5.252283183311424e-06}
{"train_loss": 0.015909390524029732, "global_step": 231564, "epoch": 2601, "lr": 5.2520245323420754e-06}
{"train_loss": 0.012318684719502926, "global_step": 231565, "epoch": 2601, "lr": 5.251765887388554e-06}
{"train_loss": 0.02793113701045513, "global_step": 231566, "epoch": 2601, "lr": 5.251507248450893e-06}
{"train_loss": 0.05494910851120949, "global_step": 231567, "epoch": 2601, "lr": 5.251248615529108e-06}
{"train_loss": 0.10417088866233826, "global_step": 231568, "epoch": 2601, "lr": 5.250989988623256e-06}
{"train_loss": 0.06052176281809807, "global_step": 231569, "epoch": 2601, "lr": 5.250731367733358e-06}
{"train_loss": 0.032590530812740326, "global_step": 231570, "epoch": 2601, "lr": 5.250472752859459e-06}
{"train_loss": 0.0718018189072609, "global_step": 231571, "epoch": 2601, "lr": 5.25021414400158e-06}
{"train_loss": 0.030350621789693832, "global_step": 231572, "epoch": 2601, "lr": 5.249955541159762e-06}
{"train_loss": 0.06270322948694229, "global_step": 231573, "epoch": 2601, "lr": 5.2496969443340595e-06}
{"train_loss": 0.14434359967708588, "global_step": 231574, "epoch": 2601, "lr": 5.2494383535244716e-06}
{"train_loss": 0.020423589274287224, "global_step": 231575, "epoch": 2601, "lr": 5.2491797687310665e-06}
{"train_loss": 0.04495217278599739, "global_step": 231576, "epoch": 2601, "lr": 5.248921189953848e-06}
{"train_loss": 0.052467159555408725, "global_step": 231577, "epoch": 2601, "lr": 5.248662617192884e-06, "val_loss": 9.198868751525879}
{"train_loss": 0.04085724428296089, "global_step": 231578, "epoch": 2602, "lr": 5.248404050448175e-06}
{"train_loss": 0.03770710527896881, "global_step": 231579, "epoch": 2602, "lr": 5.248145489719786e-06}
{"train_loss": 0.05198964476585388, "global_step": 231580, "epoch": 2602, "lr": 5.247886935007728e-06}
{"train_loss": 0.027480052784085274, "global_step": 231581, "epoch": 2602, "lr": 5.247628386312059e-06}
{"train_loss": 0.04859153553843498, "global_step": 231582, "epoch": 2602, "lr": 5.247369843632788e-06}
{"train_loss": 0.029899049550294876, "global_step": 231583, "epoch": 2602, "lr": 5.247111306969971e-06}
{"train_loss": 0.04794534668326378, "global_step": 231584, "epoch": 2602, "lr": 5.2468527763236245e-06}
{"train_loss": 0.06341324001550674, "global_step": 231585, "epoch": 2602, "lr": 5.246594251693804e-06}
{"train_loss": 0.027459751814603806, "global_step": 231586, "epoch": 2602, "lr": 5.2463357330805266e-06}
{"train_loss": 0.02624090015888214, "global_step": 231587, "epoch": 2602, "lr": 5.246077220483842e-06}
{"train_loss": 0.05256390944123268, "global_step": 231588, "epoch": 2602, "lr": 5.245818713903767e-06}
{"train_loss": 0.07443048804998398, "global_step": 231589, "epoch": 2602, "lr": 5.245560213340344e-06}
{"train_loss": 0.0817892998456955, "global_step": 231590, "epoch": 2602, "lr": 5.245301718793622e-06}
{"train_loss": 0.0139818349853158, "global_step": 231591, "epoch": 2602, "lr": 5.245043230263613e-06}
{"train_loss": 0.026910295709967613, "global_step": 231592, "epoch": 2602, "lr": 5.2447847477503755e-06}
{"train_loss": 0.04774788022041321, "global_step": 231593, "epoch": 2602, "lr": 5.244526271253919e-06}
{"train_loss": 0.06283832341432571, "global_step": 231594, "epoch": 2602, "lr": 5.2442678007743e-06}
{"train_loss": 0.04196804389357567, "global_step": 231595, "epoch": 2602, "lr": 5.244009336311529e-06}
{"train_loss": 0.029236648231744766, "global_step": 231596, "epoch": 2602, "lr": 5.243750877865672e-06}
{"train_loss": 0.05926664173603058, "global_step": 231597, "epoch": 2602, "lr": 5.243492425436736e-06}
{"train_loss": 0.021184194833040237, "global_step": 231598, "epoch": 2602, "lr": 5.2432339790247755e-06}
{"train_loss": 0.06862297654151917, "global_step": 231599, "epoch": 2602, "lr": 5.2429755386298186e-06}
{"train_loss": 0.025158241391181946, "global_step": 231600, "epoch": 2602, "lr": 5.242717104251887e-06}
{"train_loss": 0.06433017551898956, "global_step": 231601, "epoch": 2602, "lr": 5.242458675891032e-06}
{"train_loss": 0.05780160054564476, "global_step": 231602, "epoch": 2602, "lr": 5.24220025354728e-06}
{"train_loss": 0.04546045884490013, "global_step": 231603, "epoch": 2602, "lr": 5.241941837220676e-06}
{"train_loss": 0.04783345013856888, "global_step": 231604, "epoch": 2602, "lr": 5.241683426911248e-06}
{"train_loss": 0.05989978834986687, "global_step": 231605, "epoch": 2602, "lr": 5.2414250226190174e-06}
{"train_loss": 0.06287232041358948, "global_step": 231606, "epoch": 2602, "lr": 5.241166624344046e-06}
{"train_loss": 0.08521756529808044, "global_step": 231607, "epoch": 2602, "lr": 5.240908232086339e-06}
{"train_loss": 0.030749531462788582, "global_step": 231608, "epoch": 2602, "lr": 5.240649845845952e-06}
{"train_loss": 0.07735713571310043, "global_step": 231609, "epoch": 2602, "lr": 5.240391465622924e-06}
{"train_loss": 0.06045517697930336, "global_step": 231610, "epoch": 2602, "lr": 5.240133091417265e-06}
{"train_loss": 0.07875339686870575, "global_step": 231611, "epoch": 2602, "lr": 5.239874723229038e-06}
{"train_loss": 0.04035618156194687, "global_step": 231612, "epoch": 2602, "lr": 5.239616361058252e-06}
{"train_loss": 0.052796870470047, "global_step": 231613, "epoch": 2602, "lr": 5.239358004904965e-06}
{"train_loss": 0.07023902237415314, "global_step": 231614, "epoch": 2602, "lr": 5.239099654769192e-06}
{"train_loss": 0.018909789621829987, "global_step": 231615, "epoch": 2602, "lr": 5.238841310650983e-06}
{"train_loss": 0.01833566278219223, "global_step": 231616, "epoch": 2602, "lr": 5.23858297255036e-06}
{"train_loss": 0.03948759287595749, "global_step": 231617, "epoch": 2602, "lr": 5.238324640467373e-06}
{"train_loss": 0.04354637488722801, "global_step": 231618, "epoch": 2602, "lr": 5.238066314402035e-06}
{"train_loss": 0.04966767132282257, "global_step": 231619, "epoch": 2602, "lr": 5.23780799435441e-06}
{"train_loss": 0.06039450317621231, "global_step": 231620, "epoch": 2602, "lr": 5.237549680324499e-06}
{"train_loss": 0.04342273995280266, "global_step": 231621, "epoch": 2602, "lr": 5.23729137231237e-06}
{"train_loss": 0.058713410049676895, "global_step": 231622, "epoch": 2602, "lr": 5.237033070318026e-06}
{"train_loss": 0.044768963009119034, "global_step": 231623, "epoch": 2602, "lr": 5.236774774341529e-06}
{"train_loss": 0.04410674050450325, "global_step": 231624, "epoch": 2602, "lr": 5.236516484382892e-06}
{"train_loss": 0.052521198987960815, "global_step": 231625, "epoch": 2602, "lr": 5.236258200442162e-06}
{"train_loss": 0.08495667576789856, "global_step": 231626, "epoch": 2602, "lr": 5.23599992251938e-06}
{"train_loss": 0.07258892059326172, "global_step": 231627, "epoch": 2602, "lr": 5.235741650614562e-06}
{"train_loss": 0.06192796677350998, "global_step": 231628, "epoch": 2602, "lr": 5.235483384727763e-06}
{"train_loss": 0.06761327385902405, "global_step": 231629, "epoch": 2602, "lr": 5.235225124859e-06}
{"train_loss": 0.045965395867824554, "global_step": 231630, "epoch": 2602, "lr": 5.2349668710083234e-06}
{"train_loss": 0.048383504152297974, "global_step": 231631, "epoch": 2602, "lr": 5.234708623175749e-06}
{"train_loss": 0.04347120225429535, "global_step": 231632, "epoch": 2602, "lr": 5.234450381361333e-06}
{"train_loss": 0.05623628944158554, "global_step": 231633, "epoch": 2602, "lr": 5.234192145565087e-06}
{"train_loss": 0.05403056740760803, "global_step": 231634, "epoch": 2602, "lr": 5.23393391578707e-06}
{"train_loss": 0.04682980105280876, "global_step": 231635, "epoch": 2602, "lr": 5.2336756920272955e-06}
{"train_loss": 0.010597304441034794, "global_step": 231636, "epoch": 2602, "lr": 5.233417474285818e-06}
{"train_loss": 0.05036448314785957, "global_step": 231637, "epoch": 2602, "lr": 5.2331592625626645e-06}
{"train_loss": 0.02597733400762081, "global_step": 231638, "epoch": 2602, "lr": 5.232901056857853e-06}
{"train_loss": 0.05314943566918373, "global_step": 231639, "epoch": 2602, "lr": 5.232642857171443e-06}
{"train_loss": 0.074729323387146, "global_step": 231640, "epoch": 2602, "lr": 5.232384663503448e-06}
{"train_loss": 0.03943256288766861, "global_step": 231641, "epoch": 2602, "lr": 5.232126475853927e-06}
{"train_loss": 0.058077000081539154, "global_step": 231642, "epoch": 2602, "lr": 5.231868294222886e-06}
{"train_loss": 0.03498521074652672, "global_step": 231643, "epoch": 2602, "lr": 5.231610118610392e-06}
{"train_loss": 0.062236856669187546, "global_step": 231644, "epoch": 2602, "lr": 5.231351949016444e-06}
{"train_loss": 0.053342726081609726, "global_step": 231645, "epoch": 2602, "lr": 5.23109378544111e-06}
{"train_loss": 0.03486272320151329, "global_step": 231646, "epoch": 2602, "lr": 5.2308356278844e-06}
{"train_loss": 0.03668390214443207, "global_step": 231647, "epoch": 2602, "lr": 5.230577476346366e-06}
{"train_loss": 0.07330426573753357, "global_step": 231648, "epoch": 2602, "lr": 5.230319330827027e-06}
{"train_loss": 0.05257686227560043, "global_step": 231649, "epoch": 2602, "lr": 5.230061191326435e-06}
{"train_loss": 0.07329808175563812, "global_step": 231650, "epoch": 2602, "lr": 5.229803057844601e-06}
{"train_loss": 0.04670577496290207, "global_step": 231651, "epoch": 2602, "lr": 5.2295449303815915e-06}
{"train_loss": 0.036573030054569244, "global_step": 231652, "epoch": 2602, "lr": 5.229286808937411e-06}
{"train_loss": 0.041565902531147, "global_step": 231653, "epoch": 2602, "lr": 5.229028693512117e-06}
{"train_loss": 0.023239806294441223, "global_step": 231654, "epoch": 2602, "lr": 5.228770584105725e-06}
{"train_loss": 0.04470308497548103, "global_step": 231655, "epoch": 2602, "lr": 5.228512480718289e-06}
{"train_loss": 0.057047173380851746, "global_step": 231656, "epoch": 2602, "lr": 5.2282543833498235e-06}
{"train_loss": 0.07756948471069336, "global_step": 231657, "epoch": 2602, "lr": 5.2279962920003865e-06}
{"train_loss": 0.053709711879491806, "global_step": 231658, "epoch": 2602, "lr": 5.227738206669985e-06}
{"train_loss": 0.039364978671073914, "global_step": 231659, "epoch": 2602, "lr": 5.22748012735868e-06}
{"train_loss": 0.04966257885098457, "global_step": 231660, "epoch": 2602, "lr": 5.227222054066488e-06}
{"train_loss": 0.03445959836244583, "global_step": 231661, "epoch": 2602, "lr": 5.2269639867934536e-06}
{"train_loss": 0.03008940815925598, "global_step": 231662, "epoch": 2602, "lr": 5.2267059255396045e-06}
{"train_loss": 0.05788138881325722, "global_step": 231663, "epoch": 2602, "lr": 5.226447870304979e-06}
{"train_loss": 0.03170802816748619, "global_step": 231664, "epoch": 2602, "lr": 5.226189821089616e-06}
{"train_loss": 0.07463211566209793, "global_step": 231665, "epoch": 2602, "lr": 5.225931777893539e-06}
{"train_loss": 0.048989400455958385, "global_step": 231666, "epoch": 2602, "lr": 5.225673740716802e-06, "val_loss": 9.184126853942871}
{"train_loss": 0.034396227449178696, "global_step": 231667, "epoch": 2603, "lr": 5.225415709559417e-06}
{"train_loss": 0.04499443247914314, "global_step": 231668, "epoch": 2603, "lr": 5.225157684421439e-06}
{"train_loss": 0.05516235902905464, "global_step": 231669, "epoch": 2603, "lr": 5.224899665302879e-06}
{"train_loss": 0.04036886990070343, "global_step": 231670, "epoch": 2603, "lr": 5.2246416522037985e-06}
{"train_loss": 0.044135820120573044, "global_step": 231671, "epoch": 2603, "lr": 5.224383645124209e-06}
{"train_loss": 0.05432280898094177, "global_step": 231672, "epoch": 2603, "lr": 5.224125644064165e-06}
{"train_loss": 0.033904653042554855, "global_step": 231673, "epoch": 2603, "lr": 5.223867649023684e-06}
{"train_loss": 0.05704338476061821, "global_step": 231674, "epoch": 2603, "lr": 5.223609660002815e-06}
{"train_loss": 0.02946803905069828, "global_step": 231675, "epoch": 2603, "lr": 5.223351677001575e-06}
{"train_loss": 0.03744987025856972, "global_step": 231676, "epoch": 2603, "lr": 5.223093700020021e-06}
{"train_loss": 0.02746066451072693, "global_step": 231677, "epoch": 2603, "lr": 5.222835729058179e-06}
{"train_loss": 0.06582954525947571, "global_step": 231678, "epoch": 2603, "lr": 5.2225777641160655e-06}
{"train_loss": 0.046531349420547485, "global_step": 231679, "epoch": 2603, "lr": 5.222319805193748e-06}
{"train_loss": 0.023497123271226883, "global_step": 231680, "epoch": 2603, "lr": 5.2220618522912265e-06}
{"train_loss": 0.012230402790009975, "global_step": 231681, "epoch": 2603, "lr": 5.221803905408568e-06}
{"train_loss": 0.009186962619423866, "global_step": 231682, "epoch": 2603, "lr": 5.221545964545782e-06}
{"train_loss": 0.08568798750638962, "global_step": 231683, "epoch": 2603, "lr": 5.22128802970292e-06}
{"train_loss": 0.04329364001750946, "global_step": 231684, "epoch": 2603, "lr": 5.221030100880004e-06}
{"train_loss": 0.044405099004507065, "global_step": 231685, "epoch": 2603, "lr": 5.220772178077082e-06}
{"train_loss": 0.06426317244768143, "global_step": 231686, "epoch": 2603, "lr": 5.220514261294174e-06}
{"train_loss": 0.04435530677437782, "global_step": 231687, "epoch": 2603, "lr": 5.220256350531328e-06}
{"train_loss": 0.03645290434360504, "global_step": 231688, "epoch": 2603, "lr": 5.219998445788565e-06}
{"train_loss": 0.03866462782025337, "global_step": 231689, "epoch": 2603, "lr": 5.219740547065938e-06}
{"train_loss": 0.013269725255668163, "global_step": 231690, "epoch": 2603, "lr": 5.2194826543634616e-06}
{"train_loss": 0.03213980048894882, "global_step": 231691, "epoch": 2603, "lr": 5.219224767681185e-06}
{"train_loss": 0.06423921883106232, "global_step": 231692, "epoch": 2603, "lr": 5.218966887019133e-06}
{"train_loss": 0.010024173185229301, "global_step": 231693, "epoch": 2603, "lr": 5.218709012377354e-06}
{"train_loss": 0.062129367142915726, "global_step": 231694, "epoch": 2603, "lr": 5.218451143755865e-06}
{"train_loss": 0.05847641080617905, "global_step": 231695, "epoch": 2603, "lr": 5.218193281154716e-06}
{"train_loss": 0.046020474284887314, "global_step": 231696, "epoch": 2603, "lr": 5.217935424573928e-06}
{"train_loss": 0.04570840671658516, "global_step": 231697, "epoch": 2603, "lr": 5.217677574013552e-06}
{"train_loss": 0.022805768996477127, "global_step": 231698, "epoch": 2603, "lr": 5.217419729473605e-06}
{"train_loss": 0.04817648604512215, "global_step": 231699, "epoch": 2603, "lr": 5.217161890954125e-06}
{"train_loss": 0.06721509993076324, "global_step": 231700, "epoch": 2603, "lr": 5.2169040584551685e-06}
{"train_loss": 0.02825584076344967, "global_step": 231701, "epoch": 2603, "lr": 5.216646231976741e-06}
{"train_loss": 0.0708899199962616, "global_step": 231702, "epoch": 2603, "lr": 5.216388411518897e-06}
{"train_loss": 0.02296091429889202, "global_step": 231703, "epoch": 2603, "lr": 5.216130597081653e-06}
{"train_loss": 0.05973929166793823, "global_step": 231704, "epoch": 2603, "lr": 5.215872788665066e-06}
{"train_loss": 0.061865050345659256, "global_step": 231705, "epoch": 2603, "lr": 5.215614986269152e-06}
{"train_loss": 0.05962027981877327, "global_step": 231706, "epoch": 2603, "lr": 5.21535718989396e-06}
{"train_loss": 0.029841439798474312, "global_step": 231707, "epoch": 2603, "lr": 5.215099399539508e-06}
{"train_loss": 0.084988072514534, "global_step": 231708, "epoch": 2603, "lr": 5.214841615205851e-06}
{"train_loss": 0.027670733630657196, "global_step": 231709, "epoch": 2603, "lr": 5.214583836893e-06}
{"train_loss": 0.03245509788393974, "global_step": 231710, "epoch": 2603, "lr": 5.214326064601016e-06}
{"train_loss": 0.04722560942173004, "global_step": 231711, "epoch": 2603, "lr": 5.214068298329905e-06}
{"train_loss": 0.016597406938672066, "global_step": 231712, "epoch": 2603, "lr": 5.213810538079733e-06}
{"train_loss": 0.046293988823890686, "global_step": 231713, "epoch": 2603, "lr": 5.213552783850506e-06}
{"train_loss": 0.01869870349764824, "global_step": 231714, "epoch": 2603, "lr": 5.213295035642279e-06}
{"train_loss": 0.06898629665374756, "global_step": 231715, "epoch": 2603, "lr": 5.213037293455076e-06}
{"train_loss": 0.08488316833972931, "global_step": 231716, "epoch": 2603, "lr": 5.212779557288927e-06}
{"train_loss": 0.08625493943691254, "global_step": 231717, "epoch": 2603, "lr": 5.212521827143885e-06}
{"train_loss": 0.02954210713505745, "global_step": 231718, "epoch": 2603, "lr": 5.2122641030199645e-06}
{"train_loss": 0.11614596843719482, "global_step": 231719, "epoch": 2603, "lr": 5.212006384917218e-06}
{"train_loss": 0.05018705874681473, "global_step": 231720, "epoch": 2603, "lr": 5.211748672835665e-06}
{"train_loss": 0.048358235508203506, "global_step": 231721, "epoch": 2603, "lr": 5.211490966775351e-06}
{"train_loss": 0.06267271190881729, "global_step": 231722, "epoch": 2603, "lr": 5.211233266736298e-06}
{"train_loss": 0.038466304540634155, "global_step": 231723, "epoch": 2603, "lr": 5.210975572718557e-06}
{"train_loss": 0.020580271258950233, "global_step": 231724, "epoch": 2603, "lr": 5.21071788472215e-06}
{"train_loss": 0.03960356488823891, "global_step": 231725, "epoch": 2603, "lr": 5.210460202747119e-06}
{"train_loss": 0.08610153943300247, "global_step": 231726, "epoch": 2603, "lr": 5.210202526793489e-06}
{"train_loss": 0.060514409095048904, "global_step": 231727, "epoch": 2603, "lr": 5.20994485686131e-06}
{"train_loss": 0.06408034265041351, "global_step": 231728, "epoch": 2603, "lr": 5.209687192950602e-06}
{"train_loss": 0.061226993799209595, "global_step": 231729, "epoch": 2603, "lr": 5.2094295350614105e-06}
{"train_loss": 0.032456137239933014, "global_step": 231730, "epoch": 2603, "lr": 5.2091718831937586e-06}
{"train_loss": 0.029465755447745323, "global_step": 231731, "epoch": 2603, "lr": 5.208914237347695e-06}
{"train_loss": 0.021990148350596428, "global_step": 231732, "epoch": 2603, "lr": 5.2086565975232425e-06}
{"train_loss": 0.018839208409190178, "global_step": 231733, "epoch": 2603, "lr": 5.208398963720445e-06}
{"train_loss": 0.02100454457104206, "global_step": 231734, "epoch": 2603, "lr": 5.2081413359393204e-06}
{"train_loss": 0.07441644370555878, "global_step": 231735, "epoch": 2603, "lr": 5.207883714179923e-06}
{"train_loss": 0.0746719017624855, "global_step": 231736, "epoch": 2603, "lr": 5.207626098442281e-06}
{"train_loss": 0.053527429699897766, "global_step": 231737, "epoch": 2603, "lr": 5.207368488726422e-06}
{"train_loss": 0.0158254262059927, "global_step": 231738, "epoch": 2603, "lr": 5.207110885032402e-06}
{"train_loss": 0.04633454233407974, "global_step": 231739, "epoch": 2603, "lr": 5.206853287360219e-06}
{"train_loss": 0.0702490359544754, "global_step": 231740, "epoch": 2603, "lr": 5.206595695709948e-06}
{"train_loss": 0.09146473556756973, "global_step": 231741, "epoch": 2603, "lr": 5.206338110081588e-06}
{"train_loss": 0.03739317134022713, "global_step": 231742, "epoch": 2603, "lr": 5.2060805304752044e-06}
{"train_loss": 0.07356291264295578, "global_step": 231743, "epoch": 2603, "lr": 5.205822956890804e-06}
{"train_loss": 0.09068043529987335, "global_step": 231744, "epoch": 2603, "lr": 5.205565389328442e-06}
{"train_loss": 0.10622198134660721, "global_step": 231745, "epoch": 2603, "lr": 5.205307827788142e-06}
{"train_loss": 0.04584135115146637, "global_step": 231746, "epoch": 2603, "lr": 5.2050502722699515e-06}
{"train_loss": 0.05689419433474541, "global_step": 231747, "epoch": 2603, "lr": 5.2047927227738825e-06}
{"train_loss": 0.03187883272767067, "global_step": 231748, "epoch": 2603, "lr": 5.204535179299997e-06}
{"train_loss": 0.024890191853046417, "global_step": 231749, "epoch": 2603, "lr": 5.204277641848304e-06}
{"train_loss": 0.06846453249454498, "global_step": 231750, "epoch": 2603, "lr": 5.204020110418856e-06}
{"train_loss": 0.07761389762163162, "global_step": 231751, "epoch": 2603, "lr": 5.203762585011679e-06}
{"train_loss": 0.09505247324705124, "global_step": 231752, "epoch": 2603, "lr": 5.203505065626812e-06}
{"train_loss": 0.05673894286155701, "global_step": 231753, "epoch": 2603, "lr": 5.203247552264296e-06}
{"train_loss": 0.033377472311258316, "global_step": 231754, "epoch": 2603, "lr": 5.202990044924144e-06}
{"train_loss": 0.04934218165914664, "global_step": 231755, "epoch": 2603, "lr": 5.202732543606415e-06, "val_loss": 9.164484977722168}
{"train_loss": 0.035450529307127, "global_step": 231756, "epoch": 2604, "lr": 5.202475048311118e-06}
{"train_loss": 0.03495806083083153, "global_step": 231757, "epoch": 2604, "lr": 5.202217559038314e-06}
{"train_loss": 0.03138607367873192, "global_step": 231758, "epoch": 2604, "lr": 5.201960075788021e-06}
{"train_loss": 0.044083718210458755, "global_step": 231759, "epoch": 2604, "lr": 5.201702598560282e-06}
{"train_loss": 0.028471942991018295, "global_step": 231760, "epoch": 2604, "lr": 5.20144512735512e-06}
{"train_loss": 0.06263597309589386, "global_step": 231761, "epoch": 2604, "lr": 5.201187662172591e-06}
{"train_loss": 0.023790527135133743, "global_step": 231762, "epoch": 2604, "lr": 5.2009302030127e-06}
{"train_loss": 0.02248295396566391, "global_step": 231763, "epoch": 2604, "lr": 5.200672749875513e-06}
{"train_loss": 0.029835589230060577, "global_step": 231764, "epoch": 2604, "lr": 5.200415302761035e-06}
{"train_loss": 0.034572623670101166, "global_step": 231765, "epoch": 2604, "lr": 5.200157861669331e-06}
{"train_loss": 0.050836604088544846, "global_step": 231766, "epoch": 2604, "lr": 5.199900426600407e-06}
{"train_loss": 0.03943156078457832, "global_step": 231767, "epoch": 2604, "lr": 5.199642997554321e-06}
{"train_loss": 0.05618459731340408, "global_step": 231768, "epoch": 2604, "lr": 5.199385574531085e-06}
{"train_loss": 0.054382070899009705, "global_step": 231769, "epoch": 2604, "lr": 5.199128157530758e-06}
{"train_loss": 0.0326402522623539, "global_step": 231770, "epoch": 2604, "lr": 5.198870746553353e-06}
{"train_loss": 0.01263399701565504, "global_step": 231771, "epoch": 2604, "lr": 5.198613341598912e-06}
{"train_loss": 0.04281553998589516, "global_step": 231772, "epoch": 2604, "lr": 5.198355942667477e-06}
{"train_loss": 0.04703393951058388, "global_step": 231773, "epoch": 2604, "lr": 5.198098549759073e-06}
{"train_loss": 0.10591080039739609, "global_step": 231774, "epoch": 2604, "lr": 5.1978411628737455e-06}
{"train_loss": 0.0262496005743742, "global_step": 231775, "epoch": 2604, "lr": 5.197583782011517e-06}
{"train_loss": 0.0424371063709259, "global_step": 231776, "epoch": 2604, "lr": 5.197326407172431e-06}
{"train_loss": 0.039850205183029175, "global_step": 231777, "epoch": 2604, "lr": 5.197069038356511e-06}
{"train_loss": 0.08741813153028488, "global_step": 231778, "epoch": 2604, "lr": 5.196811675563812e-06}
{"train_loss": 0.038610346615314484, "global_step": 231779, "epoch": 2604, "lr": 5.196554318794339e-06}
{"train_loss": 0.07210145890712738, "global_step": 231780, "epoch": 2604, "lr": 5.196296968048159e-06}
{"train_loss": 0.027241425588726997, "global_step": 231781, "epoch": 2604, "lr": 5.196039623325277e-06}
{"train_loss": 0.05201815068721771, "global_step": 231782, "epoch": 2604, "lr": 5.195782284625756e-06}
{"train_loss": 0.02958480268716812, "global_step": 231783, "epoch": 2604, "lr": 5.1955249519496044e-06}
{"train_loss": 0.0469253733754158, "global_step": 231784, "epoch": 2604, "lr": 5.195267625296874e-06}
{"train_loss": 0.06716431677341461, "global_step": 231785, "epoch": 2604, "lr": 5.195010304667586e-06}
{"train_loss": 0.024183396250009537, "global_step": 231786, "epoch": 2604, "lr": 5.194752990061797e-06}
{"train_loss": 0.05118810385465622, "global_step": 231787, "epoch": 2604, "lr": 5.194495681479511e-06}
{"train_loss": 0.054786860942840576, "global_step": 231788, "epoch": 2604, "lr": 5.19423837892079e-06}
{"train_loss": 0.06465394049882889, "global_step": 231789, "epoch": 2604, "lr": 5.193981082385652e-06}
{"train_loss": 0.029789108783006668, "global_step": 231790, "epoch": 2604, "lr": 5.193723791874144e-06}
{"train_loss": 0.0329640693962574, "global_step": 231791, "epoch": 2604, "lr": 5.193466507386285e-06}
{"train_loss": 0.018511541187763214, "global_step": 231792, "epoch": 2604, "lr": 5.19320922892213e-06}
{"train_loss": 0.05358795449137688, "global_step": 231793, "epoch": 2604, "lr": 5.192951956481696e-06}
{"train_loss": 0.0459928922355175, "global_step": 231794, "epoch": 2604, "lr": 5.192694690065019e-06}
{"train_loss": 0.08720456808805466, "global_step": 231795, "epoch": 2604, "lr": 5.1924374296721475e-06}
{"train_loss": 0.06028936430811882, "global_step": 231796, "epoch": 2604, "lr": 5.192180175303091e-06}
{"train_loss": 0.03631934896111488, "global_step": 231797, "epoch": 2604, "lr": 5.191922926957915e-06}
{"train_loss": 0.031803689897060394, "global_step": 231798, "epoch": 2604, "lr": 5.191665684636626e-06}
{"train_loss": 0.08510922640562057, "global_step": 231799, "epoch": 2604, "lr": 5.191408448339286e-06}
{"train_loss": 0.0641295537352562, "global_step": 231800, "epoch": 2604, "lr": 5.191151218065904e-06}
{"train_loss": 0.03965965658426285, "global_step": 231801, "epoch": 2604, "lr": 5.1908939938165316e-06}
{"train_loss": 0.04566629230976105, "global_step": 231802, "epoch": 2604, "lr": 5.19063677559119e-06}
{"train_loss": 0.02894466184079647, "global_step": 231803, "epoch": 2604, "lr": 5.19037956338993e-06}
{"train_loss": 0.050818685442209244, "global_step": 231804, "epoch": 2604, "lr": 5.190122357212768e-06}
{"train_loss": 0.04975425824522972, "global_step": 231805, "epoch": 2604, "lr": 5.189865157059759e-06}
{"train_loss": 0.053717780858278275, "global_step": 231806, "epoch": 2604, "lr": 5.189607962930915e-06}
{"train_loss": 0.0588877908885479, "global_step": 231807, "epoch": 2604, "lr": 5.189350774826285e-06}
{"train_loss": 0.0629025399684906, "global_step": 231808, "epoch": 2604, "lr": 5.189093592745909e-06}
{"train_loss": 0.04277448356151581, "global_step": 231809, "epoch": 2604, "lr": 5.188836416689796e-06}
{"train_loss": 0.0616891123354435, "global_step": 231810, "epoch": 2604, "lr": 5.188579246658015e-06}
{"train_loss": 0.021233435720205307, "global_step": 231811, "epoch": 2604, "lr": 5.1883220826505766e-06}
{"train_loss": 0.06775541603565216, "global_step": 231812, "epoch": 2604, "lr": 5.1880649246675235e-06}
{"train_loss": 0.05048740282654762, "global_step": 231813, "epoch": 2604, "lr": 5.187807772708886e-06}
{"train_loss": 0.023555491119623184, "global_step": 231814, "epoch": 2604, "lr": 5.187550626774707e-06}
{"train_loss": 0.05068438872694969, "global_step": 231815, "epoch": 2604, "lr": 5.1872934868650084e-06}
{"train_loss": 0.014836796559393406, "global_step": 231816, "epoch": 2604, "lr": 5.18703635297984e-06}
{"train_loss": 0.04897255823016167, "global_step": 231817, "epoch": 2604, "lr": 5.18677922511922e-06}
{"train_loss": 0.029084473848342896, "global_step": 231818, "epoch": 2604, "lr": 5.186522103283203e-06}
{"train_loss": 0.04492669180035591, "global_step": 231819, "epoch": 2604, "lr": 5.186264987471801e-06}
{"train_loss": 0.047622378915548325, "global_step": 231820, "epoch": 2604, "lr": 5.186007877685067e-06}
{"train_loss": 0.06107030436396599, "global_step": 231821, "epoch": 2604, "lr": 5.185750773923021e-06}
{"train_loss": 0.04145217686891556, "global_step": 231822, "epoch": 2604, "lr": 5.1854936761857095e-06}
{"train_loss": 0.021612180396914482, "global_step": 231823, "epoch": 2604, "lr": 5.185236584473157e-06}
{"train_loss": 0.05269058421254158, "global_step": 231824, "epoch": 2604, "lr": 5.184979498785408e-06}
{"train_loss": 0.05698990449309349, "global_step": 231825, "epoch": 2604, "lr": 5.18472241912249e-06}
{"train_loss": 0.02643798291683197, "global_step": 231826, "epoch": 2604, "lr": 5.184465345484446e-06}
{"train_loss": 0.05375358834862709, "global_step": 231827, "epoch": 2604, "lr": 5.184208277871294e-06}
{"train_loss": 0.05209169536828995, "global_step": 231828, "epoch": 2604, "lr": 5.183951216283089e-06}
{"train_loss": 0.04917840287089348, "global_step": 231829, "epoch": 2604, "lr": 5.183694160719843e-06}
{"train_loss": 0.07063619792461395, "global_step": 231830, "epoch": 2604, "lr": 5.183437111181616e-06}
{"train_loss": 0.05534364655613899, "global_step": 231831, "epoch": 2604, "lr": 5.1831800676684305e-06}
{"train_loss": 0.049188774079084396, "global_step": 231832, "epoch": 2604, "lr": 5.182923030180309e-06}
{"train_loss": 0.05210500955581665, "global_step": 231833, "epoch": 2604, "lr": 5.1826659987173065e-06}
{"train_loss": 0.049029067158699036, "global_step": 231834, "epoch": 2604, "lr": 5.18240897327944e-06}
{"train_loss": 0.05560263246297836, "global_step": 231835, "epoch": 2604, "lr": 5.182151953866759e-06}
{"train_loss": 0.035853683948516846, "global_step": 231836, "epoch": 2604, "lr": 5.181894940479287e-06}
{"train_loss": 0.05515303462743759, "global_step": 231837, "epoch": 2604, "lr": 5.1816379331170675e-06}
{"train_loss": 0.04608311876654625, "global_step": 231838, "epoch": 2604, "lr": 5.181380931780128e-06}
{"train_loss": 0.06468404084444046, "global_step": 231839, "epoch": 2604, "lr": 5.181123936468507e-06}
{"train_loss": 0.04694552719593048, "global_step": 231840, "epoch": 2604, "lr": 5.1808669471822335e-06}
{"train_loss": 0.03786250203847885, "global_step": 231841, "epoch": 2604, "lr": 5.1806099639213514e-06}
{"train_loss": 0.0424184575676918, "global_step": 231842, "epoch": 2604, "lr": 5.180352986685882e-06}
{"train_loss": 0.04359854385256767, "global_step": 231843, "epoch": 2604, "lr": 5.180096015475872e-06}
{"train_loss": 0.046202653746926385, "global_step": 231844, "epoch": 2604, "lr": 5.179839050291357e-06, "val_loss": 9.142716407775879}
{"train_loss": 0.03613949194550514, "global_step": 231845, "epoch": 2605, "lr": 5.179582091132356e-06}
{"train_loss": 0.027642730623483658, "global_step": 231846, "epoch": 2605, "lr": 5.179325137998925e-06}
{"train_loss": 0.06472362577915192, "global_step": 231847, "epoch": 2605, "lr": 5.179068190891079e-06}
{"train_loss": 0.060102302581071854, "global_step": 231848, "epoch": 2605, "lr": 5.178811249808868e-06}
{"train_loss": 0.04968256503343582, "global_step": 231849, "epoch": 2605, "lr": 5.178554314752315e-06}
{"train_loss": 0.046059828251600266, "global_step": 231850, "epoch": 2605, "lr": 5.178297385721465e-06}
{"train_loss": 0.04695482924580574, "global_step": 231851, "epoch": 2605, "lr": 5.178040462716332e-06}
{"train_loss": 0.026029953733086586, "global_step": 231852, "epoch": 2605, "lr": 5.177783545736986e-06}
{"train_loss": 0.09192212671041489, "global_step": 231853, "epoch": 2605, "lr": 5.177526634783425e-06}
{"train_loss": 0.047831375151872635, "global_step": 231854, "epoch": 2605, "lr": 5.177269729855705e-06}
{"train_loss": 0.06798206269741058, "global_step": 231855, "epoch": 2605, "lr": 5.177012830953848e-06}
{"train_loss": 0.051729049533605576, "global_step": 231856, "epoch": 2605, "lr": 5.1767559380779094e-06}
{"train_loss": 0.02296118624508381, "global_step": 231857, "epoch": 2605, "lr": 5.1764990512278954e-06}
{"train_loss": 0.04842717945575714, "global_step": 231858, "epoch": 2605, "lr": 5.176242170403867e-06}
{"train_loss": 0.036879245191812515, "global_step": 231859, "epoch": 2605, "lr": 5.175985295605834e-06}
{"train_loss": 0.043336160480976105, "global_step": 231860, "epoch": 2605, "lr": 5.175728426833854e-06}
{"train_loss": 0.03756412863731384, "global_step": 231861, "epoch": 2605, "lr": 5.175471564087941e-06}
{"train_loss": 0.03848913684487343, "global_step": 231862, "epoch": 2605, "lr": 5.175214707368143e-06}
{"train_loss": 0.07319385558366776, "global_step": 231863, "epoch": 2605, "lr": 5.174957856674501e-06}
{"train_loss": 0.04756869003176689, "global_step": 231864, "epoch": 2605, "lr": 5.1747010120070325e-06}
{"train_loss": 0.04593926668167114, "global_step": 231865, "epoch": 2605, "lr": 5.174444173365778e-06}
{"train_loss": 0.06656122952699661, "global_step": 231866, "epoch": 2605, "lr": 5.174187340750774e-06}
{"train_loss": 0.03908950462937355, "global_step": 231867, "epoch": 2605, "lr": 5.173930514162051e-06}
{"train_loss": 0.060163963586091995, "global_step": 231868, "epoch": 2605, "lr": 5.1736736935996564e-06}
{"train_loss": 0.04752245172858238, "global_step": 231869, "epoch": 2605, "lr": 5.173416879063608e-06}
{"train_loss": 0.0596734844148159, "global_step": 231870, "epoch": 2605, "lr": 5.173160070553945e-06}
{"train_loss": 0.04717554152011871, "global_step": 231871, "epoch": 2605, "lr": 5.172903268070712e-06}
{"train_loss": 0.08562453091144562, "global_step": 231872, "epoch": 2605, "lr": 5.1726464716139235e-06}
{"train_loss": 0.08677849173545837, "global_step": 231873, "epoch": 2605, "lr": 5.172389681183643e-06}
{"train_loss": 0.041761673986911774, "global_step": 231874, "epoch": 2605, "lr": 5.172132896779869e-06}
{"train_loss": 0.029851563274860382, "global_step": 231875, "epoch": 2605, "lr": 5.171876118402674e-06}
{"train_loss": 0.024557046592235565, "global_step": 231876, "epoch": 2605, "lr": 5.171619346052059e-06}
{"train_loss": 0.04524161294102669, "global_step": 231877, "epoch": 2605, "lr": 5.171362579728084e-06}
{"train_loss": 0.061835166066884995, "global_step": 231878, "epoch": 2605, "lr": 5.171105819430761e-06}
{"train_loss": 0.058874763548374176, "global_step": 231879, "epoch": 2605, "lr": 5.1708490651601385e-06}
{"train_loss": 0.06698587536811829, "global_step": 231880, "epoch": 2605, "lr": 5.1705923169162615e-06}
{"train_loss": 0.09390276670455933, "global_step": 231881, "epoch": 2605, "lr": 5.170335574699137e-06}
{"train_loss": 0.06278659403324127, "global_step": 231882, "epoch": 2605, "lr": 5.170078838508829e-06}
{"train_loss": 0.03907693549990654, "global_step": 231883, "epoch": 2605, "lr": 5.169822108345346e-06}
{"train_loss": 0.04161161929368973, "global_step": 231884, "epoch": 2605, "lr": 5.169565384208741e-06}
{"train_loss": 0.029835432767868042, "global_step": 231885, "epoch": 2605, "lr": 5.169308666099037e-06}
{"train_loss": 0.034598980098962784, "global_step": 231886, "epoch": 2605, "lr": 5.169051954016279e-06}
{"train_loss": 0.07525691390037537, "global_step": 231887, "epoch": 2605, "lr": 5.168795247960489e-06}
{"train_loss": 0.015863388776779175, "global_step": 231888, "epoch": 2605, "lr": 5.168538547931712e-06}
{"train_loss": 0.06114568933844566, "global_step": 231889, "epoch": 2605, "lr": 5.168281853929974e-06}
{"train_loss": 0.02976434864103794, "global_step": 231890, "epoch": 2605, "lr": 5.16802516595532e-06}
{"train_loss": 0.024612654000520706, "global_step": 231891, "epoch": 2605, "lr": 5.1677684840077736e-06}
{"train_loss": 0.05783824250102043, "global_step": 231892, "epoch": 2605, "lr": 5.167511808087383e-06}
{"train_loss": 0.021262524649500847, "global_step": 231893, "epoch": 2605, "lr": 5.1672551381941604e-06}
{"train_loss": 0.045754771679639816, "global_step": 231894, "epoch": 2605, "lr": 5.166998474328166e-06}
{"train_loss": 0.06725791841745377, "global_step": 231895, "epoch": 2605, "lr": 5.166741816489412e-06}
{"train_loss": 0.0566018745303154, "global_step": 231896, "epoch": 2605, "lr": 5.166485164677954e-06}
{"train_loss": 0.07489695399999619, "global_step": 231897, "epoch": 2605, "lr": 5.166228518893806e-06}
{"train_loss": 0.041365981101989746, "global_step": 231898, "epoch": 2605, "lr": 5.1659718791370105e-06}
{"train_loss": 0.05577798932790756, "global_step": 231899, "epoch": 2605, "lr": 5.165715245407615e-06}
{"train_loss": 0.03020988032221794, "global_step": 231900, "epoch": 2605, "lr": 5.165458617705632e-06}
{"train_loss": 0.03261692449450493, "global_step": 231901, "epoch": 2605, "lr": 5.165201996031116e-06}
{"train_loss": 0.03888572007417679, "global_step": 231902, "epoch": 2605, "lr": 5.164945380384084e-06}
{"train_loss": 0.07870212942361832, "global_step": 231903, "epoch": 2605, "lr": 5.1646887707645866e-06}
{"train_loss": 0.024073226377367973, "global_step": 231904, "epoch": 2605, "lr": 5.16443216717265e-06}
{"train_loss": 0.06973782926797867, "global_step": 231905, "epoch": 2605, "lr": 5.164175569608304e-06}
{"train_loss": 0.05199860408902168, "global_step": 231906, "epoch": 2605, "lr": 5.163918978071591e-06}
{"train_loss": 0.042224619537591934, "global_step": 231907, "epoch": 2605, "lr": 5.16366239256254e-06}
{"train_loss": 0.04405731335282326, "global_step": 231908, "epoch": 2605, "lr": 5.16340581308119e-06}
{"train_loss": 0.07322986423969269, "global_step": 231909, "epoch": 2605, "lr": 5.16314923962758e-06}
{"train_loss": 0.03085988201200962, "global_step": 231910, "epoch": 2605, "lr": 5.162892672201725e-06}
{"train_loss": 0.05404626950621605, "global_step": 231911, "epoch": 2605, "lr": 5.162636110803687e-06}
{"train_loss": 0.0791349709033966, "global_step": 231912, "epoch": 2605, "lr": 5.162379555433472e-06}
{"train_loss": 0.047475460916757584, "global_step": 231913, "epoch": 2605, "lr": 5.1621230060911356e-06}
{"train_loss": 0.06123022362589836, "global_step": 231914, "epoch": 2605, "lr": 5.161866462776699e-06}
{"train_loss": 0.0359925776720047, "global_step": 231915, "epoch": 2605, "lr": 5.161609925490213e-06}
{"train_loss": 0.020219065248966217, "global_step": 231916, "epoch": 2605, "lr": 5.1613533942316935e-06}
{"train_loss": 0.058772001415491104, "global_step": 231917, "epoch": 2605, "lr": 5.161096869001181e-06}
{"train_loss": 0.07505364716053009, "global_step": 231918, "epoch": 2605, "lr": 5.160840349798724e-06}
{"train_loss": 0.045943256467580795, "global_step": 231919, "epoch": 2605, "lr": 5.160583836624333e-06}
{"train_loss": 0.029267527163028717, "global_step": 231920, "epoch": 2605, "lr": 5.160327329478065e-06}
{"train_loss": 0.02751288190484047, "global_step": 231921, "epoch": 2605, "lr": 5.160070828359942e-06}
{"train_loss": 0.07391159236431122, "global_step": 231922, "epoch": 2605, "lr": 5.159814333270002e-06}
{"train_loss": 0.03920292854309082, "global_step": 231923, "epoch": 2605, "lr": 5.159557844208274e-06}
{"train_loss": 0.050387702882289886, "global_step": 231924, "epoch": 2605, "lr": 5.1593013611748006e-06}
{"train_loss": 0.04544514790177345, "global_step": 231925, "epoch": 2605, "lr": 5.159044884169606e-06}
{"train_loss": 0.0680341124534607, "global_step": 231926, "epoch": 2605, "lr": 5.158788413192745e-06}
{"train_loss": 0.06289979070425034, "global_step": 231927, "epoch": 2605, "lr": 5.158531948244222e-06}
{"train_loss": 0.08524927496910095, "global_step": 231928, "epoch": 2605, "lr": 5.158275489324105e-06}
{"train_loss": 0.042015716433525085, "global_step": 231929, "epoch": 2605, "lr": 5.158019036432393e-06}
{"train_loss": 0.015587463974952698, "global_step": 231930, "epoch": 2605, "lr": 5.157762589569154e-06}
{"train_loss": 0.0793728455901146, "global_step": 231931, "epoch": 2605, "lr": 5.157506148734398e-06}
{"train_loss": 0.05232634395360947, "global_step": 231932, "epoch": 2605, "lr": 5.157249713928175e-06}
{"train_loss": 0.05028522562947166, "global_step": 231933, "epoch": 2605, "lr": 5.156993285150508e-06, "val_loss": 9.223692893981934, "train_action_mse_error": 16.929147720336914}
{"train_loss": 0.03833966329693794, "global_step": 231934, "epoch": 2606, "lr": 5.1567368624014345e-06}
{"train_loss": 0.030078329145908356, "global_step": 231935, "epoch": 2606, "lr": 5.156480445681e-06}
{"train_loss": 0.034840669482946396, "global_step": 231936, "epoch": 2606, "lr": 5.156224034989221e-06}
{"train_loss": 0.048129335045814514, "global_step": 231937, "epoch": 2606, "lr": 5.1559676303261525e-06}
{"train_loss": 0.058700934052467346, "global_step": 231938, "epoch": 2606, "lr": 5.155711231691807e-06}
{"train_loss": 0.04273964464664459, "global_step": 231939, "epoch": 2606, "lr": 5.1554548390862375e-06}
{"train_loss": 0.014759952202439308, "global_step": 231940, "epoch": 2606, "lr": 5.155198452509457e-06}
{"train_loss": 0.027254987508058548, "global_step": 231941, "epoch": 2606, "lr": 5.1549420719615325e-06}
{"train_loss": 0.0486917644739151, "global_step": 231942, "epoch": 2606, "lr": 5.154685697442474e-06}
{"train_loss": 0.036183979362249374, "global_step": 231943, "epoch": 2606, "lr": 5.154429328952315e-06}
{"train_loss": 0.031278274953365326, "global_step": 231944, "epoch": 2606, "lr": 5.1541729664911e-06}
{"train_loss": 0.06214944273233414, "global_step": 231945, "epoch": 2606, "lr": 5.153916610058857e-06}
{"train_loss": 0.0506497323513031, "global_step": 231946, "epoch": 2606, "lr": 5.153660259655629e-06}
{"train_loss": 0.05011604726314545, "global_step": 231947, "epoch": 2606, "lr": 5.153403915281447e-06}
{"train_loss": 0.02631354331970215, "global_step": 231948, "epoch": 2606, "lr": 5.153147576936329e-06}
{"train_loss": 0.02137262560427189, "global_step": 231949, "epoch": 2606, "lr": 5.152891244620339e-06}
{"train_loss": 0.0362185463309288, "global_step": 231950, "epoch": 2606, "lr": 5.152634918333482e-06}
{"train_loss": 0.026975655928254128, "global_step": 231951, "epoch": 2606, "lr": 5.152378598075818e-06}
{"train_loss": 0.02753106690943241, "global_step": 231952, "epoch": 2606, "lr": 5.1521222838473606e-06}
{"train_loss": 0.08782078325748444, "global_step": 231953, "epoch": 2606, "lr": 5.151865975648157e-06}
{"train_loss": 0.02821861393749714, "global_step": 231954, "epoch": 2606, "lr": 5.151609673478241e-06}
{"train_loss": 0.06160998344421387, "global_step": 231955, "epoch": 2606, "lr": 5.151353377337642e-06}
{"train_loss": 0.0838901624083519, "global_step": 231956, "epoch": 2606, "lr": 5.151097087226403e-06}
{"train_loss": 0.07339148223400116, "global_step": 231957, "epoch": 2606, "lr": 5.150840803144541e-06}
{"train_loss": 0.04494740813970566, "global_step": 231958, "epoch": 2606, "lr": 5.150584525092117e-06}
{"train_loss": 0.027564477175474167, "global_step": 231959, "epoch": 2606, "lr": 5.150328253069137e-06}
{"train_loss": 0.02061769738793373, "global_step": 231960, "epoch": 2606, "lr": 5.1500719870756555e-06}
{"train_loss": 0.03810662403702736, "global_step": 231961, "epoch": 2606, "lr": 5.149815727111695e-06}
{"train_loss": 0.06639651954174042, "global_step": 231962, "epoch": 2606, "lr": 5.149559473177301e-06}
{"train_loss": 0.023114822804927826, "global_step": 231963, "epoch": 2606, "lr": 5.149303225272495e-06}
{"train_loss": 0.07624364644289017, "global_step": 231964, "epoch": 2606, "lr": 5.149046983397332e-06}
{"train_loss": 0.019336778670549393, "global_step": 231965, "epoch": 2606, "lr": 5.148790747551818e-06}
{"train_loss": 0.029136506840586662, "global_step": 231966, "epoch": 2606, "lr": 5.1485345177360145e-06}
{"train_loss": 0.036969445645809174, "global_step": 231967, "epoch": 2606, "lr": 5.148278293949932e-06}
{"train_loss": 0.03259333595633507, "global_step": 231968, "epoch": 2606, "lr": 5.148022076193626e-06}
{"train_loss": 0.0516720712184906, "global_step": 231969, "epoch": 2606, "lr": 5.147765864467113e-06}
{"train_loss": 0.030366605147719383, "global_step": 231970, "epoch": 2606, "lr": 5.147509658770439e-06}
{"train_loss": 0.04149066284298897, "global_step": 231971, "epoch": 2606, "lr": 5.14725345910364e-06}
{"train_loss": 0.040776822715997696, "global_step": 231972, "epoch": 2606, "lr": 5.146997265466741e-06}
{"train_loss": 0.023226546123623848, "global_step": 231973, "epoch": 2606, "lr": 5.14674107785979e-06}
{"train_loss": 0.044770196080207825, "global_step": 231974, "epoch": 2606, "lr": 5.146484896282805e-06}
{"train_loss": 0.03592770919203758, "global_step": 231975, "epoch": 2606, "lr": 5.146228720735835e-06}
{"train_loss": 0.042665138840675354, "global_step": 231976, "epoch": 2606, "lr": 5.1459725512188975e-06}
{"train_loss": 0.059588607400655746, "global_step": 231977, "epoch": 2606, "lr": 5.1457163877320524e-06}
{"train_loss": 0.04487347602844238, "global_step": 231978, "epoch": 2606, "lr": 5.145460230275301e-06}
{"train_loss": 0.028207307681441307, "global_step": 231979, "epoch": 2606, "lr": 5.145204078848714e-06}
{"train_loss": 0.05773217976093292, "global_step": 231980, "epoch": 2606, "lr": 5.144947933452304e-06}
{"train_loss": 0.033486660569906235, "global_step": 231981, "epoch": 2606, "lr": 5.144691794086098e-06}
{"train_loss": 0.022200625389814377, "global_step": 231982, "epoch": 2606, "lr": 5.144435660750152e-06}
{"train_loss": 0.05627191811800003, "global_step": 231983, "epoch": 2606, "lr": 5.144179533444482e-06}
{"train_loss": 0.03690353035926819, "global_step": 231984, "epoch": 2606, "lr": 5.143923412169138e-06}
{"train_loss": 0.029044557362794876, "global_step": 231985, "epoch": 2606, "lr": 5.143667296924137e-06}
{"train_loss": 0.02573450841009617, "global_step": 231986, "epoch": 2606, "lr": 5.143411187709535e-06}
{"train_loss": 0.03186318278312683, "global_step": 231987, "epoch": 2606, "lr": 5.143155084525353e-06}
{"train_loss": 0.09275703132152557, "global_step": 231988, "epoch": 2606, "lr": 5.142898987371619e-06}
{"train_loss": 0.04542757570743561, "global_step": 231989, "epoch": 2606, "lr": 5.142642896248373e-06}
{"train_loss": 0.03361492231488228, "global_step": 231990, "epoch": 2606, "lr": 5.142386811155664e-06}
{"train_loss": 0.04473603144288063, "global_step": 231991, "epoch": 2606, "lr": 5.142130732093503e-06}
{"train_loss": 0.029916606843471527, "global_step": 231992, "epoch": 2606, "lr": 5.1418746590619456e-06}
{"train_loss": 0.034362807869911194, "global_step": 231993, "epoch": 2606, "lr": 5.141618592061004e-06}
{"train_loss": 0.04701684042811394, "global_step": 231994, "epoch": 2606, "lr": 5.141362531090738e-06}
{"train_loss": 0.05847165733575821, "global_step": 231995, "epoch": 2606, "lr": 5.141106476151158e-06}
{"train_loss": 0.042177021503448486, "global_step": 231996, "epoch": 2606, "lr": 5.140850427242322e-06}
{"train_loss": 0.07011929154396057, "global_step": 231997, "epoch": 2606, "lr": 5.140594384364239e-06}
{"train_loss": 0.05310463905334473, "global_step": 231998, "epoch": 2606, "lr": 5.1403383475169705e-06}
{"train_loss": 0.06554968655109406, "global_step": 231999, "epoch": 2606, "lr": 5.140082316700523e-06}
{"train_loss": 0.07635913044214249, "global_step": 232000, "epoch": 2606, "lr": 5.139826291914957e-06}
{"train_loss": 0.04112562537193298, "global_step": 232001, "epoch": 2606, "lr": 5.139570273160282e-06}
{"train_loss": 0.062354713678359985, "global_step": 232002, "epoch": 2606, "lr": 5.139314260436556e-06}
{"train_loss": 0.0659010037779808, "global_step": 232003, "epoch": 2606, "lr": 5.139058253743795e-06}
{"train_loss": 0.013363425619900227, "global_step": 232004, "epoch": 2606, "lr": 5.138802253082048e-06}
{"train_loss": 0.04095027595758438, "global_step": 232005, "epoch": 2606, "lr": 5.138546258451332e-06}
{"train_loss": 0.032030925154685974, "global_step": 232006, "epoch": 2606, "lr": 5.138290269851692e-06}
{"train_loss": 0.02755053900182247, "global_step": 232007, "epoch": 2606, "lr": 5.138034287283172e-06}
{"train_loss": 0.012524154037237167, "global_step": 232008, "epoch": 2606, "lr": 5.137778310745789e-06}
{"train_loss": 0.019361868500709534, "global_step": 232009, "epoch": 2606, "lr": 5.137522340239592e-06}
{"train_loss": 0.050369635224342346, "global_step": 232010, "epoch": 2606, "lr": 5.137266375764599e-06}
{"train_loss": 0.057684075087308884, "global_step": 232011, "epoch": 2606, "lr": 5.137010417320865e-06}
{"train_loss": 0.024984486401081085, "global_step": 232012, "epoch": 2606, "lr": 5.1367544649084e-06}
{"train_loss": 0.04025665298104286, "global_step": 232013, "epoch": 2606, "lr": 5.136498518527261e-06}
{"train_loss": 0.029186123982071877, "global_step": 232014, "epoch": 2606, "lr": 5.13624257817747e-06}
{"train_loss": 0.018097778782248497, "global_step": 232015, "epoch": 2606, "lr": 5.13598664385907e-06}
{"train_loss": 0.05723198130726814, "global_step": 232016, "epoch": 2606, "lr": 5.135730715572079e-06}
{"train_loss": 0.06448127329349518, "global_step": 232017, "epoch": 2606, "lr": 5.135474793316553e-06}
{"train_loss": 0.055735960602760315, "global_step": 232018, "epoch": 2606, "lr": 5.135218877092507e-06}
{"train_loss": 0.037438225001096725, "global_step": 232019, "epoch": 2606, "lr": 5.1349629668999925e-06}
{"train_loss": 0.038679394870996475, "global_step": 232020, "epoch": 2606, "lr": 5.13470706273903e-06}
{"train_loss": 0.03398188203573227, "global_step": 232021, "epoch": 2606, "lr": 5.1344511646096545e-06}
{"train_loss": 0.042147584583903296, "global_step": 232022, "epoch": 2606, "lr": 5.134195272511916e-06, "val_loss": 9.303192138671875}
{"train_loss": 0.053837746381759644, "global_step": 232023, "epoch": 2607, "lr": 5.1339393864458295e-06}
{"train_loss": 0.05461202561855316, "global_step": 232024, "epoch": 2607, "lr": 5.133683506411446e-06}
{"train_loss": 0.032691437751054764, "global_step": 232025, "epoch": 2607, "lr": 5.1334276324087825e-06}
{"train_loss": 0.026231475174427032, "global_step": 232026, "epoch": 2607, "lr": 5.133171764437894e-06}
{"train_loss": 0.021128635853528976, "global_step": 232027, "epoch": 2607, "lr": 5.132915902498791e-06}
{"train_loss": 0.015655694529414177, "global_step": 232028, "epoch": 2607, "lr": 5.132660046591531e-06}
{"train_loss": 0.06974226981401443, "global_step": 232029, "epoch": 2607, "lr": 5.132404196716129e-06}
{"train_loss": 0.045813072472810745, "global_step": 232030, "epoch": 2607, "lr": 5.132148352872634e-06}
{"train_loss": 0.05194016546010971, "global_step": 232031, "epoch": 2607, "lr": 5.131892515061071e-06}
{"train_loss": 0.1162266954779625, "global_step": 232032, "epoch": 2607, "lr": 5.131636683281488e-06}
{"train_loss": 0.03654139116406441, "global_step": 232033, "epoch": 2607, "lr": 5.131380857533896e-06}
{"train_loss": 0.0727938860654831, "global_step": 232034, "epoch": 2607, "lr": 5.1311250378183575e-06}
{"train_loss": 0.035330891609191895, "global_step": 232035, "epoch": 2607, "lr": 5.130869224134877e-06}
{"train_loss": 0.0563519224524498, "global_step": 232036, "epoch": 2607, "lr": 5.130613416483515e-06}
{"train_loss": 0.07583114504814148, "global_step": 232037, "epoch": 2607, "lr": 5.130357614864289e-06}
{"train_loss": 0.021112795919179916, "global_step": 232038, "epoch": 2607, "lr": 5.130101819277244e-06}
{"train_loss": 0.034860461950302124, "global_step": 232039, "epoch": 2607, "lr": 5.129846029722407e-06}
{"train_loss": 0.025566954165697098, "global_step": 232040, "epoch": 2607, "lr": 5.129590246199822e-06}
{"train_loss": 0.05336163192987442, "global_step": 232041, "epoch": 2607, "lr": 5.129334468709507e-06}
{"train_loss": 0.044564440846443176, "global_step": 232042, "epoch": 2607, "lr": 5.12907869725151e-06}
{"train_loss": 0.06901025772094727, "global_step": 232043, "epoch": 2607, "lr": 5.128822931825861e-06}
{"train_loss": 0.05605442449450493, "global_step": 232044, "epoch": 2607, "lr": 5.128567172432591e-06}
{"train_loss": 0.04011666402220726, "global_step": 232045, "epoch": 2607, "lr": 5.128311419071746e-06}
{"train_loss": 0.08778724819421768, "global_step": 232046, "epoch": 2607, "lr": 5.1280556717433425e-06}
{"train_loss": 0.06638508290052414, "global_step": 232047, "epoch": 2607, "lr": 5.127799930447441e-06}
{"train_loss": 0.07085345685482025, "global_step": 232048, "epoch": 2607, "lr": 5.1275441951840416e-06}
{"train_loss": 0.038662854582071304, "global_step": 232049, "epoch": 2607, "lr": 5.127288465953212e-06}
{"train_loss": 0.03654760122299194, "global_step": 232050, "epoch": 2607, "lr": 5.127032742754961e-06}
{"train_loss": 0.041101835668087006, "global_step": 232051, "epoch": 2607, "lr": 5.126777025589347e-06}
{"train_loss": 0.07613718509674072, "global_step": 232052, "epoch": 2607, "lr": 5.126521314456373e-06}
{"train_loss": 0.06551919877529144, "global_step": 232053, "epoch": 2607, "lr": 5.126265609356107e-06}
{"train_loss": 0.037995994091033936, "global_step": 232054, "epoch": 2607, "lr": 5.126009910288554e-06}
{"train_loss": 0.03597775474190712, "global_step": 232055, "epoch": 2607, "lr": 5.125754217253776e-06}
{"train_loss": 0.057358916848897934, "global_step": 232056, "epoch": 2607, "lr": 5.125498530251782e-06}
{"train_loss": 0.05168267339468002, "global_step": 232057, "epoch": 2607, "lr": 5.12524284928263e-06}
{"train_loss": 0.05106543004512787, "global_step": 232058, "epoch": 2607, "lr": 5.124987174346335e-06}
{"train_loss": 0.025728752836585045, "global_step": 232059, "epoch": 2607, "lr": 5.124731505442937e-06}
{"train_loss": 0.03802954778075218, "global_step": 232060, "epoch": 2607, "lr": 5.124475842572474e-06}
{"train_loss": 0.03414919227361679, "global_step": 232061, "epoch": 2607, "lr": 5.124220185734974e-06}
{"train_loss": 0.05240354686975479, "global_step": 232062, "epoch": 2607, "lr": 5.123964534930487e-06}
{"train_loss": 0.027302611619234085, "global_step": 232063, "epoch": 2607, "lr": 5.123708890159018e-06}
{"train_loss": 0.05866583064198494, "global_step": 232064, "epoch": 2607, "lr": 5.123453251420635e-06}
{"train_loss": 0.05673300474882126, "global_step": 232065, "epoch": 2607, "lr": 5.123197618715347e-06}
{"train_loss": 0.06598352640867233, "global_step": 232066, "epoch": 2607, "lr": 5.122941992043207e-06}
{"train_loss": 0.05969761312007904, "global_step": 232067, "epoch": 2607, "lr": 5.122686371404229e-06}
{"train_loss": 0.04516841471195221, "global_step": 232068, "epoch": 2607, "lr": 5.122430756798474e-06}
{"train_loss": 0.0829024612903595, "global_step": 232069, "epoch": 2607, "lr": 5.122175148225944e-06}
{"train_loss": 0.06186117231845856, "global_step": 232070, "epoch": 2607, "lr": 5.121919545686704e-06}
{"train_loss": 0.022844014689326286, "global_step": 232071, "epoch": 2607, "lr": 5.121663949180766e-06}
{"train_loss": 0.0646544024348259, "global_step": 232072, "epoch": 2607, "lr": 5.1214083587081854e-06}
{"train_loss": 0.06140310317277908, "global_step": 232073, "epoch": 2607, "lr": 5.121152774268967e-06}
{"train_loss": 0.060652684420347214, "global_step": 232074, "epoch": 2607, "lr": 5.120897195863178e-06}
{"train_loss": 0.0751926600933075, "global_step": 232075, "epoch": 2607, "lr": 5.12064162349083e-06}
{"train_loss": 0.04439891502261162, "global_step": 232076, "epoch": 2607, "lr": 5.120386057151971e-06}
{"train_loss": 0.05172479525208473, "global_step": 232077, "epoch": 2607, "lr": 5.1201304968466145e-06}
{"train_loss": 0.0577576719224453, "global_step": 232078, "epoch": 2607, "lr": 5.119874942574826e-06}
{"train_loss": 0.021967822685837746, "global_step": 232079, "epoch": 2607, "lr": 5.119619394336611e-06}
{"train_loss": 0.04254576191306114, "global_step": 232080, "epoch": 2607, "lr": 5.11936385213202e-06}
{"train_loss": 0.04342779889702797, "global_step": 232081, "epoch": 2607, "lr": 5.1191083159610855e-06}
{"train_loss": 0.018589742481708527, "global_step": 232082, "epoch": 2607, "lr": 5.118852785823836e-06}
{"train_loss": 0.029037943109869957, "global_step": 232083, "epoch": 2607, "lr": 5.118597261720321e-06}
{"train_loss": 0.060844700783491135, "global_step": 232084, "epoch": 2607, "lr": 5.118341743650551e-06}
{"train_loss": 0.031869594007730484, "global_step": 232085, "epoch": 2607, "lr": 5.1180862316145824e-06}
{"train_loss": 0.04334639385342598, "global_step": 232086, "epoch": 2607, "lr": 5.117830725612427e-06}
{"train_loss": 0.0944058895111084, "global_step": 232087, "epoch": 2607, "lr": 5.117575225644144e-06}
{"train_loss": 0.06979339569807053, "global_step": 232088, "epoch": 2607, "lr": 5.117319731709752e-06}
{"train_loss": 0.022535599768161774, "global_step": 232089, "epoch": 2607, "lr": 5.117064243809294e-06}
{"train_loss": 0.03967773914337158, "global_step": 232090, "epoch": 2607, "lr": 5.116808761942793e-06}
{"train_loss": 0.06339626759290695, "global_step": 232091, "epoch": 2607, "lr": 5.116553286110293e-06}
{"train_loss": 0.03230995312333107, "global_step": 232092, "epoch": 2607, "lr": 5.116297816311821e-06}
{"train_loss": 0.05659160763025284, "global_step": 232093, "epoch": 2607, "lr": 5.116042352547423e-06}
{"train_loss": 0.029668694362044334, "global_step": 232094, "epoch": 2607, "lr": 5.11578689481712e-06}
{"train_loss": 0.0342082679271698, "global_step": 232095, "epoch": 2607, "lr": 5.115531443120957e-06}
{"train_loss": 0.04129231721162796, "global_step": 232096, "epoch": 2607, "lr": 5.115275997458968e-06}
{"train_loss": 0.06745373457670212, "global_step": 232097, "epoch": 2607, "lr": 5.115020557831174e-06}
{"train_loss": 0.020701205357909203, "global_step": 232098, "epoch": 2607, "lr": 5.114765124237625e-06}
{"train_loss": 0.05996686592698097, "global_step": 232099, "epoch": 2607, "lr": 5.114509696678343e-06}
{"train_loss": 0.050392646342515945, "global_step": 232100, "epoch": 2607, "lr": 5.114254275153374e-06}
{"train_loss": 0.06642033159732819, "global_step": 232101, "epoch": 2607, "lr": 5.113998859662739e-06}
{"train_loss": 0.03293464332818985, "global_step": 232102, "epoch": 2607, "lr": 5.113743450206488e-06}
{"train_loss": 0.045918501913547516, "global_step": 232103, "epoch": 2607, "lr": 5.1134880467846366e-06}
{"train_loss": 0.05063067749142647, "global_step": 232104, "epoch": 2607, "lr": 5.113232649397237e-06}
{"train_loss": 0.0492740198969841, "global_step": 232105, "epoch": 2607, "lr": 5.11297725804431e-06}
{"train_loss": 0.0537998229265213, "global_step": 232106, "epoch": 2607, "lr": 5.112721872725906e-06}
{"train_loss": 0.046520039439201355, "global_step": 232107, "epoch": 2607, "lr": 5.112466493442042e-06}
{"train_loss": 0.04564628005027771, "global_step": 232108, "epoch": 2607, "lr": 5.1122111201927665e-06}
{"train_loss": 0.033850401639938354, "global_step": 232109, "epoch": 2607, "lr": 5.1119557529780985e-06}
{"train_loss": 0.05060739070177078, "global_step": 232110, "epoch": 2607, "lr": 5.111700391798091e-06}
{"train_loss": 0.04926954228640272, "global_step": 232111, "epoch": 2607, "lr": 5.111445036652757e-06, "val_loss": 9.354485511779785}
{"train_loss": 0.03422071784734726, "global_step": 232112, "epoch": 2608, "lr": 5.111189687542151e-06}
{"train_loss": 0.07204661518335342, "global_step": 232113, "epoch": 2608, "lr": 5.11093434446629e-06}
{"train_loss": 0.07879780977964401, "global_step": 232114, "epoch": 2608, "lr": 5.110679007425229e-06}
{"train_loss": 0.09721028059720993, "global_step": 232115, "epoch": 2608, "lr": 5.110423676418974e-06}
{"train_loss": 0.031420402228832245, "global_step": 232116, "epoch": 2608, "lr": 5.1101683514475805e-06}
{"train_loss": 0.03033137507736683, "global_step": 232117, "epoch": 2608, "lr": 5.109913032511088e-06}
{"train_loss": 0.009364007040858269, "global_step": 232118, "epoch": 2608, "lr": 5.109657719609506e-06}
{"train_loss": 0.07486221194267273, "global_step": 232119, "epoch": 2608, "lr": 5.1094024127428975e-06}
{"train_loss": 0.022430414333939552, "global_step": 232120, "epoch": 2608, "lr": 5.109147111911272e-06}
{"train_loss": 0.10869744420051575, "global_step": 232121, "epoch": 2608, "lr": 5.108891817114686e-06}
{"train_loss": 0.03994682431221008, "global_step": 232122, "epoch": 2608, "lr": 5.10863652835315e-06}
{"train_loss": 0.09624943137168884, "global_step": 232123, "epoch": 2608, "lr": 5.10838124562672e-06}
{"train_loss": 0.05238764360547066, "global_step": 232124, "epoch": 2608, "lr": 5.108125968935412e-06}
{"train_loss": 0.04830661788582802, "global_step": 232125, "epoch": 2608, "lr": 5.107870698279277e-06}
{"train_loss": 0.08514144271612167, "global_step": 232126, "epoch": 2608, "lr": 5.107615433658336e-06}
{"train_loss": 0.043105002492666245, "global_step": 232127, "epoch": 2608, "lr": 5.107360175072634e-06}
{"train_loss": 0.05010212957859039, "global_step": 232128, "epoch": 2608, "lr": 5.107104922522194e-06}
{"train_loss": 0.036647532135248184, "global_step": 232129, "epoch": 2608, "lr": 5.1068496760070695e-06}
{"train_loss": 0.04409816488623619, "global_step": 232130, "epoch": 2608, "lr": 5.106594435527268e-06}
{"train_loss": 0.05711470544338226, "global_step": 232131, "epoch": 2608, "lr": 5.106339201082844e-06}
{"train_loss": 0.040862467139959335, "global_step": 232132, "epoch": 2608, "lr": 5.10608397267382e-06}
{"train_loss": 0.044296205043792725, "global_step": 232133, "epoch": 2608, "lr": 5.105828750300246e-06}
{"train_loss": 0.031163252890110016, "global_step": 232134, "epoch": 2608, "lr": 5.105573533962138e-06}
{"train_loss": 0.02527378313243389, "global_step": 232135, "epoch": 2608, "lr": 5.105318323659541e-06}
{"train_loss": 0.07685139030218124, "global_step": 232136, "epoch": 2608, "lr": 5.105063119392495e-06}
{"train_loss": 0.01975233480334282, "global_step": 232137, "epoch": 2608, "lr": 5.104807921161009e-06}
{"train_loss": 0.06885068863630295, "global_step": 232138, "epoch": 2608, "lr": 5.104552728965151e-06}
{"train_loss": 0.014629289507865906, "global_step": 232139, "epoch": 2608, "lr": 5.104297542804925e-06}
{"train_loss": 0.03380639851093292, "global_step": 232140, "epoch": 2608, "lr": 5.104042362680389e-06}
{"train_loss": 0.1053377017378807, "global_step": 232141, "epoch": 2608, "lr": 5.1037871885915575e-06}
{"train_loss": 0.06341008841991425, "global_step": 232142, "epoch": 2608, "lr": 5.103532020538488e-06}
{"train_loss": 0.02765650302171707, "global_step": 232143, "epoch": 2608, "lr": 5.103276858521189e-06}
{"train_loss": 0.029682472348213196, "global_step": 232144, "epoch": 2608, "lr": 5.103021702539712e-06}
{"train_loss": 0.055190496146678925, "global_step": 232145, "epoch": 2608, "lr": 5.102766552594085e-06}
{"train_loss": 0.10492556542158127, "global_step": 232146, "epoch": 2608, "lr": 5.1025114086843464e-06}
{"train_loss": 0.06191704049706459, "global_step": 232147, "epoch": 2608, "lr": 5.1022562708105235e-06}
{"train_loss": 0.045700933784246445, "global_step": 232148, "epoch": 2608, "lr": 5.102001138972662e-06}
{"train_loss": 0.06299097836017609, "global_step": 232149, "epoch": 2608, "lr": 5.101746013170783e-06}
{"train_loss": 0.03493515029549599, "global_step": 232150, "epoch": 2608, "lr": 5.101490893404931e-06}
{"train_loss": 0.03317447751760483, "global_step": 232151, "epoch": 2608, "lr": 5.101235779675128e-06}
{"train_loss": 0.04357762634754181, "global_step": 232152, "epoch": 2608, "lr": 5.1009806719814205e-06}
{"train_loss": 0.05084918811917305, "global_step": 232153, "epoch": 2608, "lr": 5.10072557032385e-06}
{"train_loss": 0.04997110739350319, "global_step": 232154, "epoch": 2608, "lr": 5.100470474702423e-06}
{"train_loss": 0.04365258663892746, "global_step": 232155, "epoch": 2608, "lr": 5.1002153851172074e-06}
{"train_loss": 0.011831386014819145, "global_step": 232156, "epoch": 2608, "lr": 5.099960301568202e-06}
{"train_loss": 0.06264383345842361, "global_step": 232157, "epoch": 2608, "lr": 5.099705224055479e-06}
{"train_loss": 0.041286032646894455, "global_step": 232158, "epoch": 2608, "lr": 5.099450152579038e-06}
{"train_loss": 0.034621432423591614, "global_step": 232159, "epoch": 2608, "lr": 5.099195087138941e-06}
{"train_loss": 0.04709995537996292, "global_step": 232160, "epoch": 2608, "lr": 5.098940027735194e-06}
{"train_loss": 0.031191714107990265, "global_step": 232161, "epoch": 2608, "lr": 5.098684974367862e-06}
{"train_loss": 0.05103272572159767, "global_step": 232162, "epoch": 2608, "lr": 5.098429927036958e-06}
{"train_loss": 0.023041892796754837, "global_step": 232163, "epoch": 2608, "lr": 5.098174885742529e-06}
{"train_loss": 0.047332361340522766, "global_step": 232164, "epoch": 2608, "lr": 5.097919850484595e-06}
{"train_loss": 0.05789691582322121, "global_step": 232165, "epoch": 2608, "lr": 5.097664821263204e-06}
{"train_loss": 0.03424061834812164, "global_step": 232166, "epoch": 2608, "lr": 5.097409798078378e-06}
{"train_loss": 0.08428090810775757, "global_step": 232167, "epoch": 2608, "lr": 5.097154780930169e-06}
{"train_loss": 0.06829416751861572, "global_step": 232168, "epoch": 2608, "lr": 5.096899769818586e-06}
{"train_loss": 0.1067974641919136, "global_step": 232169, "epoch": 2608, "lr": 5.096644764743691e-06}
{"train_loss": 0.03612862527370453, "global_step": 232170, "epoch": 2608, "lr": 5.096389765705495e-06}
{"train_loss": 0.07326781749725342, "global_step": 232171, "epoch": 2608, "lr": 5.0961347727040535e-06}
{"train_loss": 0.06556510180234909, "global_step": 232172, "epoch": 2608, "lr": 5.095879785739371e-06}
{"train_loss": 0.06288612633943558, "global_step": 232173, "epoch": 2608, "lr": 5.095624804811522e-06}
{"train_loss": 0.0551796555519104, "global_step": 232174, "epoch": 2608, "lr": 5.09536982992051e-06}
{"train_loss": 0.06021427735686302, "global_step": 232175, "epoch": 2608, "lr": 5.0951148610663694e-06}
{"train_loss": 0.03954175114631653, "global_step": 232176, "epoch": 2608, "lr": 5.094859898249155e-06}
{"train_loss": 0.016581200063228607, "global_step": 232177, "epoch": 2608, "lr": 5.094604941468883e-06}
{"train_loss": 0.04031768813729286, "global_step": 232178, "epoch": 2608, "lr": 5.0943499907256e-06}
{"train_loss": 0.06476649641990662, "global_step": 232179, "epoch": 2608, "lr": 5.094095046019326e-06}
{"train_loss": 0.026911310851573944, "global_step": 232180, "epoch": 2608, "lr": 5.093840107350112e-06}
{"train_loss": 0.07415985316038132, "global_step": 232181, "epoch": 2608, "lr": 5.093585174717974e-06}
{"train_loss": 0.0467093400657177, "global_step": 232182, "epoch": 2608, "lr": 5.0933302481229685e-06}
{"train_loss": 0.04336561635136604, "global_step": 232183, "epoch": 2608, "lr": 5.093075327565105e-06}
{"train_loss": 0.046777788549661636, "global_step": 232184, "epoch": 2608, "lr": 5.092820413044441e-06}
{"train_loss": 0.04962439090013504, "global_step": 232185, "epoch": 2608, "lr": 5.092565504560987e-06}
{"train_loss": 0.026569662615656853, "global_step": 232186, "epoch": 2608, "lr": 5.092310602114802e-06}
{"train_loss": 0.07410377264022827, "global_step": 232187, "epoch": 2608, "lr": 5.092055705705901e-06}
{"train_loss": 0.05971326678991318, "global_step": 232188, "epoch": 2608, "lr": 5.091800815334324e-06}
{"train_loss": 0.03225966542959213, "global_step": 232189, "epoch": 2608, "lr": 5.091545931000119e-06}
{"train_loss": 0.03483991324901581, "global_step": 232190, "epoch": 2608, "lr": 5.091291052703296e-06}
{"train_loss": 0.03630732372403145, "global_step": 232191, "epoch": 2608, "lr": 5.091036180443909e-06}
{"train_loss": 0.07855472713708878, "global_step": 232192, "epoch": 2608, "lr": 5.090781314221982e-06}
{"train_loss": 0.022506143897771835, "global_step": 232193, "epoch": 2608, "lr": 5.090526454037553e-06}
{"train_loss": 0.041978269815444946, "global_step": 232194, "epoch": 2608, "lr": 5.09027159989065e-06}
{"train_loss": 0.040215637534856796, "global_step": 232195, "epoch": 2608, "lr": 5.090016751781323e-06}
{"train_loss": 0.02807583101093769, "global_step": 232196, "epoch": 2608, "lr": 5.08976190970959e-06}
{"train_loss": 0.05577930808067322, "global_step": 232197, "epoch": 2608, "lr": 5.0895070736754925e-06}
{"train_loss": 0.04089789092540741, "global_step": 232198, "epoch": 2608, "lr": 5.08925224367906e-06}
{"train_loss": 0.10691691935062408, "global_step": 232199, "epoch": 2608, "lr": 5.088997419720337e-06}
{"train_loss": 0.050381048697601544, "global_step": 232200, "epoch": 2608, "lr": 5.08874260179934e-06, "val_loss": 9.036310195922852}
{"train_loss": 0.03143569454550743, "global_step": 232201, "epoch": 2609, "lr": 5.08848778991613e-06}
{"train_loss": 0.034424372017383575, "global_step": 232202, "epoch": 2609, "lr": 5.088232984070712e-06}
{"train_loss": 0.09648966044187546, "global_step": 232203, "epoch": 2609, "lr": 5.0879781842631425e-06}
{"train_loss": 0.04144307225942612, "global_step": 232204, "epoch": 2609, "lr": 5.087723390493437e-06}
{"train_loss": 0.07297972589731216, "global_step": 232205, "epoch": 2609, "lr": 5.087468602761647e-06}
{"train_loss": 0.04358329623937607, "global_step": 232206, "epoch": 2609, "lr": 5.087213821067793e-06}
{"train_loss": 0.03952251747250557, "global_step": 232207, "epoch": 2609, "lr": 5.086959045411927e-06}
{"train_loss": 0.06761659681797028, "global_step": 232208, "epoch": 2609, "lr": 5.086704275794057e-06}
{"train_loss": 0.05323256179690361, "global_step": 232209, "epoch": 2609, "lr": 5.086449512214248e-06}
{"train_loss": 0.028673268854618073, "global_step": 232210, "epoch": 2609, "lr": 5.086194754672508e-06}
{"train_loss": 0.056515827775001526, "global_step": 232211, "epoch": 2609, "lr": 5.085940003168888e-06}
{"train_loss": 0.050422731786966324, "global_step": 232212, "epoch": 2609, "lr": 5.085685257703416e-06}
{"train_loss": 0.04995550960302353, "global_step": 232213, "epoch": 2609, "lr": 5.085430518276119e-06}
{"train_loss": 0.05764370039105415, "global_step": 232214, "epoch": 2609, "lr": 5.085175784887047e-06}
{"train_loss": 0.0508921816945076, "global_step": 232215, "epoch": 2609, "lr": 5.084921057536218e-06}
{"train_loss": 0.06686869263648987, "global_step": 232216, "epoch": 2609, "lr": 5.0846663362236816e-06}
{"train_loss": 0.04585245996713638, "global_step": 232217, "epoch": 2609, "lr": 5.084411620949453e-06}
{"train_loss": 0.0391499362885952, "global_step": 232218, "epoch": 2609, "lr": 5.084156911713589e-06}
{"train_loss": 0.04090909659862518, "global_step": 232219, "epoch": 2609, "lr": 5.083902208516101e-06}
{"train_loss": 0.0603022500872612, "global_step": 232220, "epoch": 2609, "lr": 5.083647511357048e-06}
{"train_loss": 0.07884997874498367, "global_step": 232221, "epoch": 2609, "lr": 5.0833928202364434e-06}
{"train_loss": 0.07994554936885834, "global_step": 232222, "epoch": 2609, "lr": 5.083138135154331e-06}
{"train_loss": 0.0345759242773056, "global_step": 232223, "epoch": 2609, "lr": 5.082883456110743e-06}
{"train_loss": 0.05832073837518692, "global_step": 232224, "epoch": 2609, "lr": 5.082628783105708e-06}
{"train_loss": 0.05706918239593506, "global_step": 232225, "epoch": 2609, "lr": 5.082374116139277e-06}
{"train_loss": 0.06432344764471054, "global_step": 232226, "epoch": 2609, "lr": 5.082119455211459e-06}
{"train_loss": 0.02373051829636097, "global_step": 232227, "epoch": 2609, "lr": 5.081864800322317e-06}
{"train_loss": 0.021257100626826286, "global_step": 232228, "epoch": 2609, "lr": 5.081610151471861e-06}
{"train_loss": 0.03960422798991203, "global_step": 232229, "epoch": 2609, "lr": 5.081355508660146e-06}
{"train_loss": 0.03069581650197506, "global_step": 232230, "epoch": 2609, "lr": 5.081100871887185e-06}
{"train_loss": 0.05501246452331543, "global_step": 232231, "epoch": 2609, "lr": 5.080846241153031e-06}
{"train_loss": 0.041634172201156616, "global_step": 232232, "epoch": 2609, "lr": 5.080591616457703e-06}
{"train_loss": 0.08723340183496475, "global_step": 232233, "epoch": 2609, "lr": 5.080336997801244e-06}
{"train_loss": 0.048065800219774246, "global_step": 232234, "epoch": 2609, "lr": 5.080082385183682e-06}
{"train_loss": 0.10269850492477417, "global_step": 232235, "epoch": 2609, "lr": 5.079827778605068e-06}
{"train_loss": 0.04603220522403717, "global_step": 232236, "epoch": 2609, "lr": 5.079573178065411e-06}
{"train_loss": 0.048014044761657715, "global_step": 232237, "epoch": 2609, "lr": 5.0793185835647685e-06}
{"train_loss": 0.09376765787601471, "global_step": 232238, "epoch": 2609, "lr": 5.07906399510315e-06}
{"train_loss": 0.050360020250082016, "global_step": 232239, "epoch": 2609, "lr": 5.078809412680618e-06}
{"train_loss": 0.04931800067424774, "global_step": 232240, "epoch": 2609, "lr": 5.078554836297178e-06}
{"train_loss": 0.06793831288814545, "global_step": 232241, "epoch": 2609, "lr": 5.078300265952896e-06}
{"train_loss": 0.011710639111697674, "global_step": 232242, "epoch": 2609, "lr": 5.078045701647771e-06}
{"train_loss": 0.06530477851629257, "global_step": 232243, "epoch": 2609, "lr": 5.077791143381861e-06}
{"train_loss": 0.05185059830546379, "global_step": 232244, "epoch": 2609, "lr": 5.077536591155202e-06}
{"train_loss": 0.03662027418613434, "global_step": 232245, "epoch": 2609, "lr": 5.077282044967812e-06}
{"train_loss": 0.050099316984415054, "global_step": 232246, "epoch": 2609, "lr": 5.077027504819742e-06}
{"train_loss": 0.06735212355852127, "global_step": 232247, "epoch": 2609, "lr": 5.076772970711024e-06}
{"train_loss": 0.04139949381351471, "global_step": 232248, "epoch": 2609, "lr": 5.07651844264167e-06}
{"train_loss": 0.07711327075958252, "global_step": 232249, "epoch": 2609, "lr": 5.07626392061174e-06}
{"train_loss": 0.05868162587285042, "global_step": 232250, "epoch": 2609, "lr": 5.076009404621251e-06}
{"train_loss": 0.05417431518435478, "global_step": 232251, "epoch": 2609, "lr": 5.0757548946702595e-06}
{"train_loss": 0.06962236016988754, "global_step": 232252, "epoch": 2609, "lr": 5.0755003907587805e-06}
{"train_loss": 0.03069816529750824, "global_step": 232253, "epoch": 2609, "lr": 5.075245892886843e-06}
{"train_loss": 0.043143387883901596, "global_step": 232254, "epoch": 2609, "lr": 5.074991401054502e-06}
{"train_loss": 0.026324553415179253, "global_step": 232255, "epoch": 2609, "lr": 5.074736915261769e-06}
{"train_loss": 0.045328643172979355, "global_step": 232256, "epoch": 2609, "lr": 5.0744824355086985e-06}
{"train_loss": 0.04921742156147957, "global_step": 232257, "epoch": 2609, "lr": 5.074227961795314e-06}
{"train_loss": 0.07386709004640579, "global_step": 232258, "epoch": 2609, "lr": 5.073973494121653e-06}
{"train_loss": 0.012371519580483437, "global_step": 232259, "epoch": 2609, "lr": 5.073719032487745e-06}
{"train_loss": 0.041454315185546875, "global_step": 232260, "epoch": 2609, "lr": 5.073464576893627e-06}
{"train_loss": 0.061844129115343094, "global_step": 232261, "epoch": 2609, "lr": 5.073210127339339e-06}
{"train_loss": 0.06492411345243454, "global_step": 232262, "epoch": 2609, "lr": 5.072955683824904e-06}
{"train_loss": 0.04648055136203766, "global_step": 232263, "epoch": 2609, "lr": 5.072701246350375e-06}
{"train_loss": 0.03693997487425804, "global_step": 232264, "epoch": 2609, "lr": 5.07244681491576e-06}
{"train_loss": 0.041733309626579285, "global_step": 232265, "epoch": 2609, "lr": 5.072192389521119e-06}
{"train_loss": 0.061599962413311005, "global_step": 232266, "epoch": 2609, "lr": 5.071937970166462e-06}
{"train_loss": 0.02079705148935318, "global_step": 232267, "epoch": 2609, "lr": 5.071683556851848e-06}
{"train_loss": 0.03418852388858795, "global_step": 232268, "epoch": 2609, "lr": 5.071429149577289e-06}
{"train_loss": 0.016463829204440117, "global_step": 232269, "epoch": 2609, "lr": 5.071174748342838e-06}
{"train_loss": 0.04394056275486946, "global_step": 232270, "epoch": 2609, "lr": 5.0709203531485065e-06}
{"train_loss": 0.050087571144104004, "global_step": 232271, "epoch": 2609, "lr": 5.070665963994353e-06}
{"train_loss": 0.053193364292383194, "global_step": 232272, "epoch": 2609, "lr": 5.0704115808803966e-06}
{"train_loss": 0.03645370900630951, "global_step": 232273, "epoch": 2609, "lr": 5.07015720380668e-06}
{"train_loss": 0.07061157375574112, "global_step": 232274, "epoch": 2609, "lr": 5.069902832773226e-06}
{"train_loss": 0.10251719504594803, "global_step": 232275, "epoch": 2609, "lr": 5.069648467780086e-06}
{"train_loss": 0.031127482652664185, "global_step": 232276, "epoch": 2609, "lr": 5.069394108827269e-06}
{"train_loss": 0.024011513218283653, "global_step": 232277, "epoch": 2609, "lr": 5.069139755914842e-06}
{"train_loss": 0.04449767619371414, "global_step": 232278, "epoch": 2609, "lr": 5.068885409042806e-06}
{"train_loss": 0.015111131593585014, "global_step": 232279, "epoch": 2609, "lr": 5.068631068211211e-06}
{"train_loss": 0.03551993891596794, "global_step": 232280, "epoch": 2609, "lr": 5.068376733420105e-06}
{"train_loss": 0.07125619053840637, "global_step": 232281, "epoch": 2609, "lr": 5.068122404669495e-06}
{"train_loss": 0.06639929860830307, "global_step": 232282, "epoch": 2609, "lr": 5.067868081959437e-06}
{"train_loss": 0.060359325259923935, "global_step": 232283, "epoch": 2609, "lr": 5.067613765289947e-06}
{"train_loss": 0.06391353160142899, "global_step": 232284, "epoch": 2609, "lr": 5.06735945466108e-06}
{"train_loss": 0.03997836261987686, "global_step": 232285, "epoch": 2609, "lr": 5.06710515007286e-06}
{"train_loss": 0.06706105172634125, "global_step": 232286, "epoch": 2609, "lr": 5.066850851525312e-06}
{"train_loss": 0.03465088829398155, "global_step": 232287, "epoch": 2609, "lr": 5.066596559018483e-06}
{"train_loss": 0.03963109105825424, "global_step": 232288, "epoch": 2609, "lr": 5.0663422725523935e-06}
{"train_loss": 0.05032306014863628, "global_step": 232289, "epoch": 2609, "lr": 5.0660879921271e-06, "val_loss": 9.209671020507812}
{"train_loss": 0.061118464916944504, "global_step": 232290, "epoch": 2610, "lr": 5.0658337177426184e-06}
{"train_loss": 0.04621401056647301, "global_step": 232291, "epoch": 2610, "lr": 5.065579449398983e-06}
{"train_loss": 0.02483179047703743, "global_step": 232292, "epoch": 2610, "lr": 5.065325187096237e-06}
{"train_loss": 0.07103750854730606, "global_step": 232293, "epoch": 2610, "lr": 5.065070930834404e-06}
{"train_loss": 0.07551389187574387, "global_step": 232294, "epoch": 2610, "lr": 5.064816680613538e-06}
{"train_loss": 0.07241598516702652, "global_step": 232295, "epoch": 2610, "lr": 5.064562436433646e-06}
{"train_loss": 0.03675071522593498, "global_step": 232296, "epoch": 2610, "lr": 5.064308198294781e-06}
{"train_loss": 0.05648818612098694, "global_step": 232297, "epoch": 2610, "lr": 5.064053966196969e-06}
{"train_loss": 0.049676068127155304, "global_step": 232298, "epoch": 2610, "lr": 5.063799740140246e-06}
{"train_loss": 0.09486784785985947, "global_step": 232299, "epoch": 2610, "lr": 5.063545520124658e-06}
{"train_loss": 0.0447104349732399, "global_step": 232300, "epoch": 2610, "lr": 5.063291306150214e-06}
{"train_loss": 0.05783465877175331, "global_step": 232301, "epoch": 2610, "lr": 5.063037098216977e-06}
{"train_loss": 0.03612256795167923, "global_step": 232302, "epoch": 2610, "lr": 5.062782896324952e-06}
{"train_loss": 0.1163829118013382, "global_step": 232303, "epoch": 2610, "lr": 5.062528700474206e-06}
{"train_loss": 0.042731475085020065, "global_step": 232304, "epoch": 2610, "lr": 5.062274510664739e-06}
{"train_loss": 0.04376934841275215, "global_step": 232305, "epoch": 2610, "lr": 5.062020326896616e-06}
{"train_loss": 0.033997125923633575, "global_step": 232306, "epoch": 2610, "lr": 5.061766149169844e-06}
{"train_loss": 0.02297534979879856, "global_step": 232307, "epoch": 2610, "lr": 5.061511977484479e-06}
{"train_loss": 0.04913348704576492, "global_step": 232308, "epoch": 2610, "lr": 5.061257811840536e-06}
{"train_loss": 0.04351796954870224, "global_step": 232309, "epoch": 2610, "lr": 5.061003652238072e-06}
{"train_loss": 0.05396411940455437, "global_step": 232310, "epoch": 2610, "lr": 5.060749498677097e-06}
{"train_loss": 0.08676810562610626, "global_step": 232311, "epoch": 2610, "lr": 5.060495351157662e-06}
{"train_loss": 0.03971558436751366, "global_step": 232312, "epoch": 2610, "lr": 5.060241209679795e-06}
{"train_loss": 0.06958074122667313, "global_step": 232313, "epoch": 2610, "lr": 5.059987074243533e-06}
{"train_loss": 0.050919562578201294, "global_step": 232314, "epoch": 2610, "lr": 5.0597329448489e-06}
{"train_loss": 0.055392827838659286, "global_step": 232315, "epoch": 2610, "lr": 5.059478821495938e-06}
{"train_loss": 0.060916636139154434, "global_step": 232316, "epoch": 2610, "lr": 5.059224704184695e-06}
{"train_loss": 0.05910862237215042, "global_step": 232317, "epoch": 2610, "lr": 5.0589705929151795e-06}
{"train_loss": 0.03818752244114876, "global_step": 232318, "epoch": 2610, "lr": 5.058716487687442e-06}
{"train_loss": 0.030026648193597794, "global_step": 232319, "epoch": 2610, "lr": 5.0584623885015116e-06}
{"train_loss": 0.010550481267273426, "global_step": 232320, "epoch": 2610, "lr": 5.058208295357425e-06}
{"train_loss": 0.0372520387172699, "global_step": 232321, "epoch": 2610, "lr": 5.0579542082552115e-06}
{"train_loss": 0.03421558439731598, "global_step": 232322, "epoch": 2610, "lr": 5.057700127194914e-06}
{"train_loss": 0.026881489902734756, "global_step": 232323, "epoch": 2610, "lr": 5.057446052176562e-06}
{"train_loss": 0.09402523189783096, "global_step": 232324, "epoch": 2610, "lr": 5.057191983200177e-06}
{"train_loss": 0.07821701467037201, "global_step": 232325, "epoch": 2610, "lr": 5.056937920265819e-06}
{"train_loss": 0.03853144869208336, "global_step": 232326, "epoch": 2610, "lr": 5.056683863373496e-06}
{"train_loss": 0.08646151423454285, "global_step": 232327, "epoch": 2610, "lr": 5.056429812523261e-06}
{"train_loss": 0.09189686179161072, "global_step": 232328, "epoch": 2610, "lr": 5.056175767715132e-06}
{"train_loss": 0.03467739745974541, "global_step": 232329, "epoch": 2610, "lr": 5.055921728949164e-06}
{"train_loss": 0.04383490979671478, "global_step": 232330, "epoch": 2610, "lr": 5.0556676962253805e-06}
{"train_loss": 0.03672301769256592, "global_step": 232331, "epoch": 2610, "lr": 5.055413669543801e-06}
{"train_loss": 0.04515150934457779, "global_step": 232332, "epoch": 2610, "lr": 5.055159648904489e-06}
{"train_loss": 0.03347642719745636, "global_step": 232333, "epoch": 2610, "lr": 5.05490563430745e-06}
{"train_loss": 0.04664556682109833, "global_step": 232334, "epoch": 2610, "lr": 5.054651625752732e-06}
{"train_loss": 0.03192931041121483, "global_step": 232335, "epoch": 2610, "lr": 5.054397623240376e-06}
{"train_loss": 0.03303903341293335, "global_step": 232336, "epoch": 2610, "lr": 5.054143626770397e-06}
{"train_loss": 0.0559210330247879, "global_step": 232337, "epoch": 2610, "lr": 5.053889636342857e-06}
{"train_loss": 0.06949672102928162, "global_step": 232338, "epoch": 2610, "lr": 5.053635651957761e-06}
{"train_loss": 0.05564047396183014, "global_step": 232339, "epoch": 2610, "lr": 5.05338167361516e-06}
{"train_loss": 0.052348531782627106, "global_step": 232340, "epoch": 2610, "lr": 5.0531277013150805e-06}
{"train_loss": 0.042340658605098724, "global_step": 232341, "epoch": 2610, "lr": 5.052873735057567e-06}
{"train_loss": 0.03359449654817581, "global_step": 232342, "epoch": 2610, "lr": 5.052619774842643e-06}
{"train_loss": 0.029325157403945923, "global_step": 232343, "epoch": 2610, "lr": 5.052365820670351e-06}
{"train_loss": 0.01911250688135624, "global_step": 232344, "epoch": 2610, "lr": 5.0521118725407104e-06}
{"train_loss": 0.07963597774505615, "global_step": 232345, "epoch": 2610, "lr": 5.051857930453774e-06}
{"train_loss": 0.04742510989308357, "global_step": 232346, "epoch": 2610, "lr": 5.05160399440956e-06}
{"train_loss": 0.056438084691762924, "global_step": 232347, "epoch": 2610, "lr": 5.051350064408122e-06}
{"train_loss": 0.038204893469810486, "global_step": 232348, "epoch": 2610, "lr": 5.0510961404494685e-06}
{"train_loss": 0.035079531371593475, "global_step": 232349, "epoch": 2610, "lr": 5.050842222533658e-06}
{"train_loss": 0.02539401315152645, "global_step": 232350, "epoch": 2610, "lr": 5.050588310660709e-06}
{"train_loss": 0.06975366920232773, "global_step": 232351, "epoch": 2610, "lr": 5.0503344048306535e-06}
{"train_loss": 0.07698673009872437, "global_step": 232352, "epoch": 2610, "lr": 5.050080505043548e-06}
{"train_loss": 0.06102479621767998, "global_step": 232353, "epoch": 2610, "lr": 5.049826611299396e-06}
{"train_loss": 0.023184116929769516, "global_step": 232354, "epoch": 2610, "lr": 5.049572723598261e-06}
{"train_loss": 0.02204521745443344, "global_step": 232355, "epoch": 2610, "lr": 5.049318841940154e-06}
{"train_loss": 0.047627002000808716, "global_step": 232356, "epoch": 2610, "lr": 5.049064966325123e-06}
{"train_loss": 0.05807263404130936, "global_step": 232357, "epoch": 2610, "lr": 5.048811096753192e-06}
{"train_loss": 0.02810666337609291, "global_step": 232358, "epoch": 2610, "lr": 5.04855723322441e-06}
{"train_loss": 0.020231127738952637, "global_step": 232359, "epoch": 2610, "lr": 5.0483033757387886e-06}
{"train_loss": 0.04728402942419052, "global_step": 232360, "epoch": 2610, "lr": 5.048049524296389e-06}
{"train_loss": 0.03608599677681923, "global_step": 232361, "epoch": 2610, "lr": 5.047795678897216e-06}
{"train_loss": 0.08541584759950638, "global_step": 232362, "epoch": 2610, "lr": 5.047541839541331e-06}
{"train_loss": 0.03197687491774559, "global_step": 232363, "epoch": 2610, "lr": 5.047288006228756e-06}
{"train_loss": 0.06411600112915039, "global_step": 232364, "epoch": 2610, "lr": 5.04703417895952e-06}
{"train_loss": 0.0364384688436985, "global_step": 232365, "epoch": 2610, "lr": 5.0467803577336655e-06}
{"train_loss": 0.08589395135641098, "global_step": 232366, "epoch": 2610, "lr": 5.046526542551217e-06}
{"train_loss": 0.026832709088921547, "global_step": 232367, "epoch": 2610, "lr": 5.046272733412228e-06}
{"train_loss": 0.054284464567899704, "global_step": 232368, "epoch": 2610, "lr": 5.046018930316715e-06}
{"train_loss": 0.08950449526309967, "global_step": 232369, "epoch": 2610, "lr": 5.045765133264707e-06}
{"train_loss": 0.08557640016078949, "global_step": 232370, "epoch": 2610, "lr": 5.045511342256249e-06}
{"train_loss": 0.07405626773834229, "global_step": 232371, "epoch": 2610, "lr": 5.0452575572913885e-06}
{"train_loss": 0.038344789296388626, "global_step": 232372, "epoch": 2610, "lr": 5.045003778370128e-06}
{"train_loss": 0.061033282428979874, "global_step": 232373, "epoch": 2610, "lr": 5.044750005492532e-06}
{"train_loss": 0.08534681797027588, "global_step": 232374, "epoch": 2610, "lr": 5.044496238658614e-06}
{"train_loss": 0.06343835592269897, "global_step": 232375, "epoch": 2610, "lr": 5.044242477868422e-06}
{"train_loss": 0.057261306792497635, "global_step": 232376, "epoch": 2610, "lr": 5.043988723121973e-06}
{"train_loss": 0.030056240037083626, "global_step": 232377, "epoch": 2610, "lr": 5.043734974419323e-06}
{"train_loss": 0.051805405291529855, "global_step": 232378, "epoch": 2610, "lr": 5.0434812317604885e-06, "val_loss": 9.238523483276367, "train_action_mse_error": 11.753840446472168}
{"train_loss": 0.026528606191277504, "global_step": 232379, "epoch": 2611, "lr": 5.043227495145514e-06}
{"train_loss": 0.05916140228509903, "global_step": 232380, "epoch": 2611, "lr": 5.042973764574421e-06}
{"train_loss": 0.05346601828932762, "global_step": 232381, "epoch": 2611, "lr": 5.042720040047266e-06}
{"train_loss": 0.05207150802016258, "global_step": 232382, "epoch": 2611, "lr": 5.042466321564054e-06}
{"train_loss": 0.06490596383810043, "global_step": 232383, "epoch": 2611, "lr": 5.042212609124852e-06}
{"train_loss": 0.04813147708773613, "global_step": 232384, "epoch": 2611, "lr": 5.04195890272966e-06}
{"train_loss": 0.017812402918934822, "global_step": 232385, "epoch": 2611, "lr": 5.0417052023785386e-06}
{"train_loss": 0.0398256778717041, "global_step": 232386, "epoch": 2611, "lr": 5.041451508071504e-06}
{"train_loss": 0.06623199582099915, "global_step": 232387, "epoch": 2611, "lr": 5.041197819808602e-06}
{"train_loss": 0.08368165791034698, "global_step": 232388, "epoch": 2611, "lr": 5.040944137589865e-06}
{"train_loss": 0.029928622767329216, "global_step": 232389, "epoch": 2611, "lr": 5.0406904614153205e-06}
{"train_loss": 0.012279707007110119, "global_step": 232390, "epoch": 2611, "lr": 5.040436791285019e-06}
{"train_loss": 0.02424999326467514, "global_step": 232391, "epoch": 2611, "lr": 5.040183127198966e-06}
{"train_loss": 0.0490698367357254, "global_step": 232392, "epoch": 2611, "lr": 5.039929469157229e-06}
{"train_loss": 0.06717751920223236, "global_step": 232393, "epoch": 2611, "lr": 5.0396758171598115e-06}
{"train_loss": 0.023209886625409126, "global_step": 232394, "epoch": 2611, "lr": 5.0394221712067705e-06}
{"train_loss": 0.039334218949079514, "global_step": 232395, "epoch": 2611, "lr": 5.039168531298127e-06}
{"train_loss": 0.03367827460169792, "global_step": 232396, "epoch": 2611, "lr": 5.038914897433928e-06}
{"train_loss": 0.07333981245756149, "global_step": 232397, "epoch": 2611, "lr": 5.038661269614181e-06}
{"train_loss": 0.012728548608720303, "global_step": 232398, "epoch": 2611, "lr": 5.038407647838955e-06}
{"train_loss": 0.0587010458111763, "global_step": 232399, "epoch": 2611, "lr": 5.0381540321082555e-06}
{"train_loss": 0.08462464064359665, "global_step": 232400, "epoch": 2611, "lr": 5.037900422422137e-06}
{"train_loss": 0.0448814257979393, "global_step": 232401, "epoch": 2611, "lr": 5.037646818780628e-06}
{"train_loss": 0.05350867286324501, "global_step": 232402, "epoch": 2611, "lr": 5.03739322118375e-06}
{"train_loss": 0.016203198581933975, "global_step": 232403, "epoch": 2611, "lr": 5.037139629631555e-06}
{"train_loss": 0.0458415187895298, "global_step": 232404, "epoch": 2611, "lr": 5.036886044124056e-06}
{"train_loss": 0.040068428963422775, "global_step": 232405, "epoch": 2611, "lr": 5.036632464661312e-06}
{"train_loss": 0.03927453234791756, "global_step": 232406, "epoch": 2611, "lr": 5.036378891243332e-06}
{"train_loss": 0.04145234078168869, "global_step": 232407, "epoch": 2611, "lr": 5.036125323870173e-06}
{"train_loss": 0.047007329761981964, "global_step": 232408, "epoch": 2611, "lr": 5.035871762541855e-06}
{"train_loss": 0.036129266023635864, "global_step": 232409, "epoch": 2611, "lr": 5.035618207258419e-06}
{"train_loss": 0.009277602657675743, "global_step": 232410, "epoch": 2611, "lr": 5.035364658019886e-06}
{"train_loss": 0.07207425683736801, "global_step": 232411, "epoch": 2611, "lr": 5.0351111148263185e-06}
{"train_loss": 0.06025892123579979, "global_step": 232412, "epoch": 2611, "lr": 5.034857577677715e-06}
{"train_loss": 0.040649257600307465, "global_step": 232413, "epoch": 2611, "lr": 5.034604046574137e-06}
{"train_loss": 0.0351637601852417, "global_step": 232414, "epoch": 2611, "lr": 5.034350521515602e-06}
{"train_loss": 0.04456021264195442, "global_step": 232415, "epoch": 2611, "lr": 5.034097002502158e-06}
{"train_loss": 0.03120960108935833, "global_step": 232416, "epoch": 2611, "lr": 5.0338434895338235e-06}
{"train_loss": 0.03391497954726219, "global_step": 232417, "epoch": 2611, "lr": 5.033589982610648e-06}
{"train_loss": 0.07525801658630371, "global_step": 232418, "epoch": 2611, "lr": 5.033336481732648e-06}
{"train_loss": 0.08129729330539703, "global_step": 232419, "epoch": 2611, "lr": 5.0330829868998785e-06}
{"train_loss": 0.021237105131149292, "global_step": 232420, "epoch": 2611, "lr": 5.032829498112357e-06}
{"train_loss": 0.026664189994335175, "global_step": 232421, "epoch": 2611, "lr": 5.032576015370127e-06}
{"train_loss": 0.017550846561789513, "global_step": 232422, "epoch": 2611, "lr": 5.032322538673218e-06}
{"train_loss": 0.040331948548555374, "global_step": 232423, "epoch": 2611, "lr": 5.032069068021667e-06}
{"train_loss": 0.015304923988878727, "global_step": 232424, "epoch": 2611, "lr": 5.0318156034154966e-06}
{"train_loss": 0.04579022526741028, "global_step": 232425, "epoch": 2611, "lr": 5.031562144854756e-06}
{"train_loss": 0.07861790806055069, "global_step": 232426, "epoch": 2611, "lr": 5.031308692339482e-06}
{"train_loss": 0.06329214572906494, "global_step": 232427, "epoch": 2611, "lr": 5.031055245869687e-06}
{"train_loss": 0.03861767053604126, "global_step": 232428, "epoch": 2611, "lr": 5.030801805445429e-06}
{"train_loss": 0.03682101145386696, "global_step": 232429, "epoch": 2611, "lr": 5.030548371066723e-06}
{"train_loss": 0.07109631597995758, "global_step": 232430, "epoch": 2611, "lr": 5.030294942733621e-06}
{"train_loss": 0.05232749879360199, "global_step": 232431, "epoch": 2611, "lr": 5.030041520446138e-06}
{"train_loss": 0.04158005490899086, "global_step": 232432, "epoch": 2611, "lr": 5.02978810420433e-06}
{"train_loss": 0.06256692111492157, "global_step": 232433, "epoch": 2611, "lr": 5.029534694008209e-06}
{"train_loss": 0.04502690210938454, "global_step": 232434, "epoch": 2611, "lr": 5.029281289857824e-06}
{"train_loss": 0.03462991863489151, "global_step": 232435, "epoch": 2611, "lr": 5.029027891753202e-06}
{"train_loss": 0.10253585129976273, "global_step": 232436, "epoch": 2611, "lr": 5.028774499694383e-06}
{"train_loss": 0.05321643501520157, "global_step": 232437, "epoch": 2611, "lr": 5.028521113681389e-06}
{"train_loss": 0.02899111434817314, "global_step": 232438, "epoch": 2611, "lr": 5.02826773371427e-06}
{"train_loss": 0.06919728219509125, "global_step": 232439, "epoch": 2611, "lr": 5.028014359793054e-06}
{"train_loss": 0.06349347531795502, "global_step": 232440, "epoch": 2611, "lr": 5.027760991917763e-06}
{"train_loss": 0.035725392401218414, "global_step": 232441, "epoch": 2611, "lr": 5.027507630088452e-06}
{"train_loss": 0.043718256056308746, "global_step": 232442, "epoch": 2611, "lr": 5.027254274305132e-06}
{"train_loss": 0.022589532658457756, "global_step": 232443, "epoch": 2611, "lr": 5.027000924567865e-06}
{"train_loss": 0.03309350088238716, "global_step": 232444, "epoch": 2611, "lr": 5.026747580876656e-06}
{"train_loss": 0.12340990453958511, "global_step": 232445, "epoch": 2611, "lr": 5.026494243231561e-06}
{"train_loss": 0.062245000153779984, "global_step": 232446, "epoch": 2611, "lr": 5.026240911632596e-06}
{"train_loss": 0.04372270405292511, "global_step": 232447, "epoch": 2611, "lr": 5.025987586079817e-06}
{"train_loss": 0.029710091650485992, "global_step": 232448, "epoch": 2611, "lr": 5.025734266573234e-06}
{"train_loss": 0.05409683659672737, "global_step": 232449, "epoch": 2611, "lr": 5.025480953112905e-06}
{"train_loss": 0.06868651509284973, "global_step": 232450, "epoch": 2611, "lr": 5.025227645698844e-06}
{"train_loss": 0.054206836968660355, "global_step": 232451, "epoch": 2611, "lr": 5.024974344331096e-06}
{"train_loss": 0.047423992305994034, "global_step": 232452, "epoch": 2611, "lr": 5.02472104900969e-06}
{"train_loss": 0.049345843493938446, "global_step": 232453, "epoch": 2611, "lr": 5.024467759734663e-06}
{"train_loss": 0.0374123677611351, "global_step": 232454, "epoch": 2611, "lr": 5.024214476506045e-06}
{"train_loss": 0.05841668322682381, "global_step": 232455, "epoch": 2611, "lr": 5.023961199323879e-06}
{"train_loss": 0.06879545748233795, "global_step": 232456, "epoch": 2611, "lr": 5.023707928188187e-06}
{"train_loss": 0.05539863556623459, "global_step": 232457, "epoch": 2611, "lr": 5.02345466309902e-06}
{"train_loss": 0.07558159530162811, "global_step": 232458, "epoch": 2611, "lr": 5.023201404056388e-06}
{"train_loss": 0.03361435607075691, "global_step": 232459, "epoch": 2611, "lr": 5.022948151060347e-06}
{"train_loss": 0.030555980280041695, "global_step": 232460, "epoch": 2611, "lr": 5.022694904110919e-06}
{"train_loss": 0.058192670345306396, "global_step": 232461, "epoch": 2611, "lr": 5.0224416632081385e-06}
{"train_loss": 0.07248493283987045, "global_step": 232462, "epoch": 2611, "lr": 5.022188428352048e-06}
{"train_loss": 0.046997588127851486, "global_step": 232463, "epoch": 2611, "lr": 5.021935199542671e-06}
{"train_loss": 0.08931025862693787, "global_step": 232464, "epoch": 2611, "lr": 5.021681976780057e-06}
{"train_loss": 0.05137860029935837, "global_step": 232465, "epoch": 2611, "lr": 5.021428760064217e-06}
{"train_loss": 0.04376937448978424, "global_step": 232466, "epoch": 2611, "lr": 5.021175549395207e-06}
{"train_loss": 0.048404747693475046, "global_step": 232467, "epoch": 2611, "lr": 5.020922344773043e-06, "val_loss": 9.25911808013916}
{"train_loss": 0.007708821911364794, "global_step": 232468, "epoch": 2612, "lr": 5.020669146197776e-06}
{"train_loss": 0.0770963802933693, "global_step": 232469, "epoch": 2612, "lr": 5.020415953669427e-06}
{"train_loss": 0.03834528475999832, "global_step": 232470, "epoch": 2612, "lr": 5.020162767188041e-06}
{"train_loss": 0.06997555494308472, "global_step": 232471, "epoch": 2612, "lr": 5.019909586753635e-06}
{"train_loss": 0.030341435223817825, "global_step": 232472, "epoch": 2612, "lr": 5.019656412366269e-06}
{"train_loss": 0.04156192019581795, "global_step": 232473, "epoch": 2612, "lr": 5.019403244025944e-06}
{"train_loss": 0.12901093065738678, "global_step": 232474, "epoch": 2612, "lr": 5.0191500817327265e-06}
{"train_loss": 0.03650844842195511, "global_step": 232475, "epoch": 2612, "lr": 5.018896925486627e-06}
{"train_loss": 0.057524967938661575, "global_step": 232476, "epoch": 2612, "lr": 5.018643775287696e-06}
{"train_loss": 0.013661970384418964, "global_step": 232477, "epoch": 2612, "lr": 5.018390631135955e-06}
{"train_loss": 0.11786776036024094, "global_step": 232478, "epoch": 2612, "lr": 5.018137493031449e-06}
{"train_loss": 0.05927426740527153, "global_step": 232479, "epoch": 2612, "lr": 5.017884360974206e-06}
{"train_loss": 0.051461897790431976, "global_step": 232480, "epoch": 2612, "lr": 5.017631234964254e-06}
{"train_loss": 0.028142953291535378, "global_step": 232481, "epoch": 2612, "lr": 5.017378115001636e-06}
{"train_loss": 0.028944503515958786, "global_step": 232482, "epoch": 2612, "lr": 5.0171250010863814e-06}
{"train_loss": 0.030749406665563583, "global_step": 232483, "epoch": 2612, "lr": 5.0168718932185275e-06}
{"train_loss": 0.09827470779418945, "global_step": 232484, "epoch": 2612, "lr": 5.016618791398103e-06}
{"train_loss": 0.06583806127309799, "global_step": 232485, "epoch": 2612, "lr": 5.016365695625158e-06}
{"train_loss": 0.04642040655016899, "global_step": 232486, "epoch": 2612, "lr": 5.016112605899698e-06}
{"train_loss": 0.039395082741975784, "global_step": 232487, "epoch": 2612, "lr": 5.015859522221789e-06}
{"train_loss": 0.08593110740184784, "global_step": 232488, "epoch": 2612, "lr": 5.0156064445914365e-06}
{"train_loss": 0.02651950716972351, "global_step": 232489, "epoch": 2612, "lr": 5.015353373008697e-06}
{"train_loss": 0.05453915148973465, "global_step": 232490, "epoch": 2612, "lr": 5.0151003074735854e-06}
{"train_loss": 0.04339475929737091, "global_step": 232491, "epoch": 2612, "lr": 5.0148472479861545e-06}
{"train_loss": 0.040812648832798004, "global_step": 232492, "epoch": 2612, "lr": 5.014594194546418e-06}
{"train_loss": 0.03219769895076752, "global_step": 232493, "epoch": 2612, "lr": 5.0143411471544335e-06}
{"train_loss": 0.02678264118731022, "global_step": 232494, "epoch": 2612, "lr": 5.014088105810211e-06}
{"train_loss": 0.0752677395939827, "global_step": 232495, "epoch": 2612, "lr": 5.013835070513812e-06}
{"train_loss": 0.060191843658685684, "global_step": 232496, "epoch": 2612, "lr": 5.013582041265236e-06}
{"train_loss": 0.03285239636898041, "global_step": 232497, "epoch": 2612, "lr": 5.013329018064539e-06}
{"train_loss": 0.039750467985868454, "global_step": 232498, "epoch": 2612, "lr": 5.0130760009117646e-06}
{"train_loss": 0.09284966439008713, "global_step": 232499, "epoch": 2612, "lr": 5.01282298980692e-06}
{"train_loss": 0.046112462878227234, "global_step": 232500, "epoch": 2612, "lr": 5.012569984750065e-06}
{"train_loss": 0.024794161319732666, "global_step": 232501, "epoch": 2612, "lr": 5.012316985741211e-06}
{"train_loss": 0.02837330661714077, "global_step": 232502, "epoch": 2612, "lr": 5.012063992780408e-06}
{"train_loss": 0.028247453272342682, "global_step": 232503, "epoch": 2612, "lr": 5.011811005867684e-06}
{"train_loss": 0.05164070427417755, "global_step": 232504, "epoch": 2612, "lr": 5.0115580250030765e-06}
{"train_loss": 0.0561225526034832, "global_step": 232505, "epoch": 2612, "lr": 5.01130505018661e-06}
{"train_loss": 0.09733451902866364, "global_step": 232506, "epoch": 2612, "lr": 5.011052081418333e-06}
{"train_loss": 0.03544146195054054, "global_step": 232507, "epoch": 2612, "lr": 5.010799118698262e-06}
{"train_loss": 0.01451508142054081, "global_step": 232508, "epoch": 2612, "lr": 5.010546162026453e-06}
{"train_loss": 0.05972785875201225, "global_step": 232509, "epoch": 2612, "lr": 5.010293211402917e-06}
{"train_loss": 0.028655027970671654, "global_step": 232510, "epoch": 2612, "lr": 5.010040266827709e-06}
{"train_loss": 0.0325612872838974, "global_step": 232511, "epoch": 2612, "lr": 5.009787328300841e-06}
{"train_loss": 0.07013296335935593, "global_step": 232512, "epoch": 2612, "lr": 5.009534395822369e-06}
{"train_loss": 0.016249964013695717, "global_step": 232513, "epoch": 2612, "lr": 5.009281469392308e-06}
{"train_loss": 0.06464381515979767, "global_step": 232514, "epoch": 2612, "lr": 5.009028549010708e-06}
{"train_loss": 0.011868397705256939, "global_step": 232515, "epoch": 2612, "lr": 5.008775634677587e-06}
{"train_loss": 0.03609234094619751, "global_step": 232516, "epoch": 2612, "lr": 5.0085227263930005e-06}
{"train_loss": 0.09465498477220535, "global_step": 232517, "epoch": 2612, "lr": 5.008269824156969e-06}
{"train_loss": 0.0855979174375534, "global_step": 232518, "epoch": 2612, "lr": 5.008016927969511e-06}
{"train_loss": 0.03291470184922218, "global_step": 232519, "epoch": 2612, "lr": 5.0077640378306924e-06}
{"train_loss": 0.0661408007144928, "global_step": 232520, "epoch": 2612, "lr": 5.007511153740524e-06}
{"train_loss": 0.02807706966996193, "global_step": 232521, "epoch": 2612, "lr": 5.00725827569905e-06}
{"train_loss": 0.0379377044737339, "global_step": 232522, "epoch": 2612, "lr": 5.0070054037063e-06}
{"train_loss": 0.027443526312708855, "global_step": 232523, "epoch": 2612, "lr": 5.006752537762316e-06}
{"train_loss": 0.03905113786458969, "global_step": 232524, "epoch": 2612, "lr": 5.006499677867116e-06}
{"train_loss": 0.03344683721661568, "global_step": 232525, "epoch": 2612, "lr": 5.00624682402075e-06}
{"train_loss": 0.02571176365017891, "global_step": 232526, "epoch": 2612, "lr": 5.005993976223239e-06}
{"train_loss": 0.048740699887275696, "global_step": 232527, "epoch": 2612, "lr": 5.005741134474634e-06}
{"train_loss": 0.05593641474843025, "global_step": 232528, "epoch": 2612, "lr": 5.005488298774952e-06}
{"train_loss": 0.05195734649896622, "global_step": 232529, "epoch": 2612, "lr": 5.005235469124237e-06}
{"train_loss": 0.03930129110813141, "global_step": 232530, "epoch": 2612, "lr": 5.00498264552251e-06}
{"train_loss": 0.05482988804578781, "global_step": 232531, "epoch": 2612, "lr": 5.004729827969829e-06}
{"train_loss": 0.04726533591747284, "global_step": 232532, "epoch": 2612, "lr": 5.0044770164662024e-06}
{"train_loss": 0.019545821473002434, "global_step": 232533, "epoch": 2612, "lr": 5.0042242110116765e-06}
{"train_loss": 0.025560662150382996, "global_step": 232534, "epoch": 2612, "lr": 5.00397141160629e-06}
{"train_loss": 0.040889840573072433, "global_step": 232535, "epoch": 2612, "lr": 5.003718618250064e-06}
{"train_loss": 0.08183334767818451, "global_step": 232536, "epoch": 2612, "lr": 5.00346583094305e-06}
{"train_loss": 0.06083664670586586, "global_step": 232537, "epoch": 2612, "lr": 5.003213049685262e-06}
{"train_loss": 0.09524909406900406, "global_step": 232538, "epoch": 2612, "lr": 5.002960274476753e-06}
{"train_loss": 0.06358557939529419, "global_step": 232539, "epoch": 2612, "lr": 5.002707505317539e-06}
{"train_loss": 0.044696733355522156, "global_step": 232540, "epoch": 2612, "lr": 5.002454742207668e-06}
{"train_loss": 0.031733520328998566, "global_step": 232541, "epoch": 2612, "lr": 5.002201985147164e-06}
{"train_loss": 0.06051533296704292, "global_step": 232542, "epoch": 2612, "lr": 5.001949234136072e-06}
{"train_loss": 0.0380919985473156, "global_step": 232543, "epoch": 2612, "lr": 5.001696489174407e-06}
{"train_loss": 0.09338311851024628, "global_step": 232544, "epoch": 2612, "lr": 5.00144375026223e-06}
{"train_loss": 0.011568681336939335, "global_step": 232545, "epoch": 2612, "lr": 5.001191017399548e-06}
{"train_loss": 0.023488475009799004, "global_step": 232546, "epoch": 2612, "lr": 5.000938290586421e-06}
{"train_loss": 0.02576824091374874, "global_step": 232547, "epoch": 2612, "lr": 5.000685569822855e-06}
{"train_loss": 0.045130934566259384, "global_step": 232548, "epoch": 2612, "lr": 5.0004328551089105e-06}
{"train_loss": 0.05663818493485451, "global_step": 232549, "epoch": 2612, "lr": 5.0001801464446e-06}
{"train_loss": 0.07596968859434128, "global_step": 232550, "epoch": 2612, "lr": 4.999927443829977e-06}
{"train_loss": 0.04767478257417679, "global_step": 232551, "epoch": 2612, "lr": 4.999674747265054e-06}
{"train_loss": 0.04851176217198372, "global_step": 232552, "epoch": 2612, "lr": 4.999422056749886e-06}
{"train_loss": 0.02307073213160038, "global_step": 232553, "epoch": 2612, "lr": 4.99916937228449e-06}
{"train_loss": 0.034243661910295486, "global_step": 232554, "epoch": 2612, "lr": 4.99891669386891e-06}
{"train_loss": 0.04887838289141655, "global_step": 232555, "epoch": 2612, "lr": 4.998664021503186e-06}
{"train_loss": 0.048353739872867806, "global_step": 232556, "epoch": 2612, "lr": 4.9984113551873265e-06, "val_loss": 9.270271301269531}
{"train_loss": 0.06815433502197266, "global_step": 232557, "epoch": 2613, "lr": 4.998158694921396e-06}
{"train_loss": 0.042942360043525696, "global_step": 232558, "epoch": 2613, "lr": 4.997906040705402e-06}
{"train_loss": 0.10839751362800598, "global_step": 232559, "epoch": 2613, "lr": 4.997653392539403e-06}
{"train_loss": 0.029620064422488213, "global_step": 232560, "epoch": 2613, "lr": 4.997400750423409e-06}
{"train_loss": 0.022509468719363213, "global_step": 232561, "epoch": 2613, "lr": 4.997148114357481e-06}
{"train_loss": 0.05729660019278526, "global_step": 232562, "epoch": 2613, "lr": 4.996895484341624e-06}
{"train_loss": 0.059658367186784744, "global_step": 232563, "epoch": 2613, "lr": 4.996642860375894e-06}
{"train_loss": 0.10353615134954453, "global_step": 232564, "epoch": 2613, "lr": 4.996390242460308e-06}
{"train_loss": 0.06183492764830589, "global_step": 232565, "epoch": 2613, "lr": 4.996137630594921e-06}
{"train_loss": 0.027971267700195312, "global_step": 232566, "epoch": 2613, "lr": 4.995885024779745e-06}
{"train_loss": 0.018116075545549393, "global_step": 232567, "epoch": 2613, "lr": 4.995632425014829e-06}
{"train_loss": 0.028844613581895828, "global_step": 232568, "epoch": 2613, "lr": 4.995379831300195e-06}
{"train_loss": 0.06797052174806595, "global_step": 232569, "epoch": 2613, "lr": 4.9951272436358885e-06}
{"train_loss": 0.07504349946975708, "global_step": 232570, "epoch": 2613, "lr": 4.994874662021942e-06}
{"train_loss": 0.03669526055455208, "global_step": 232571, "epoch": 2613, "lr": 4.994622086458378e-06}
{"train_loss": 0.035960257053375244, "global_step": 232572, "epoch": 2613, "lr": 4.994369516945246e-06}
{"train_loss": 0.047586020082235336, "global_step": 232573, "epoch": 2613, "lr": 4.994116953482564e-06}
{"train_loss": 0.034175533801317215, "global_step": 232574, "epoch": 2613, "lr": 4.993864396070386e-06}
{"train_loss": 0.05392146855592728, "global_step": 232575, "epoch": 2613, "lr": 4.993611844708724e-06}
{"train_loss": 0.06022924184799194, "global_step": 232576, "epoch": 2613, "lr": 4.993359299397632e-06}
{"train_loss": 0.040530622005462646, "global_step": 232577, "epoch": 2613, "lr": 4.993106760137128e-06}
{"train_loss": 0.026665473356842995, "global_step": 232578, "epoch": 2613, "lr": 4.992854226927257e-06}
{"train_loss": 0.022518126294016838, "global_step": 232579, "epoch": 2613, "lr": 4.99260169976804e-06}
{"train_loss": 0.04096861183643341, "global_step": 232580, "epoch": 2613, "lr": 4.992349178659528e-06}
{"train_loss": 0.030616533011198044, "global_step": 232581, "epoch": 2613, "lr": 4.992096663601737e-06}
{"train_loss": 0.047765567898750305, "global_step": 232582, "epoch": 2613, "lr": 4.991844154594721e-06}
{"train_loss": 0.09215134382247925, "global_step": 232583, "epoch": 2613, "lr": 4.9915916516384884e-06}
{"train_loss": 0.03295421972870827, "global_step": 232584, "epoch": 2613, "lr": 4.991339154733105e-06}
{"train_loss": 0.06649765372276306, "global_step": 232585, "epoch": 2613, "lr": 4.991086663878575e-06}
{"train_loss": 0.04939251020550728, "global_step": 232586, "epoch": 2613, "lr": 4.990834179074949e-06}
{"train_loss": 0.05748415365815163, "global_step": 232587, "epoch": 2613, "lr": 4.990581700322256e-06}
{"train_loss": 0.046618469059467316, "global_step": 232588, "epoch": 2613, "lr": 4.990329227620522e-06}
{"train_loss": 0.02505706250667572, "global_step": 232589, "epoch": 2613, "lr": 4.990076760969803e-06}
{"train_loss": 0.06498809158802032, "global_step": 232590, "epoch": 2613, "lr": 4.989824300370122e-06}
{"train_loss": 0.022938616573810577, "global_step": 232591, "epoch": 2613, "lr": 4.9895718458215e-06}
{"train_loss": 0.07010579854249954, "global_step": 232592, "epoch": 2613, "lr": 4.989319397323988e-06}
{"train_loss": 0.046468041837215424, "global_step": 232593, "epoch": 2613, "lr": 4.989066954877608e-06}
{"train_loss": 0.04407868906855583, "global_step": 232594, "epoch": 2613, "lr": 4.988814518482404e-06}
{"train_loss": 0.036542437970638275, "global_step": 232595, "epoch": 2613, "lr": 4.98856208813841e-06}
{"train_loss": 0.044974830001592636, "global_step": 232596, "epoch": 2613, "lr": 4.9883096638456414e-06}
{"train_loss": 0.026289677247405052, "global_step": 232597, "epoch": 2613, "lr": 4.988057245604161e-06}
{"train_loss": 0.06899861991405487, "global_step": 232598, "epoch": 2613, "lr": 4.987804833413973e-06}
{"train_loss": 0.057321809232234955, "global_step": 232599, "epoch": 2613, "lr": 4.987552427275138e-06}
{"train_loss": 0.05333443358540535, "global_step": 232600, "epoch": 2613, "lr": 4.987300027187669e-06}
{"train_loss": 0.04547013342380524, "global_step": 232601, "epoch": 2613, "lr": 4.987047633151615e-06}
{"train_loss": 0.04212972894310951, "global_step": 232602, "epoch": 2613, "lr": 4.9867952451669975e-06}
{"train_loss": 0.0377366840839386, "global_step": 232603, "epoch": 2613, "lr": 4.986542863233867e-06}
{"train_loss": 0.03298936411738396, "global_step": 232604, "epoch": 2613, "lr": 4.986290487352235e-06}
{"train_loss": 0.04511411115527153, "global_step": 232605, "epoch": 2613, "lr": 4.986038117522146e-06}
{"train_loss": 0.06775988638401031, "global_step": 232606, "epoch": 2613, "lr": 4.985785753743649e-06}
{"train_loss": 0.09413396567106247, "global_step": 232607, "epoch": 2613, "lr": 4.985533396016756e-06}
{"train_loss": 0.035216134041547775, "global_step": 232608, "epoch": 2613, "lr": 4.985281044341517e-06}
{"train_loss": 0.026585081592202187, "global_step": 232609, "epoch": 2613, "lr": 4.985028698717947e-06}
{"train_loss": 0.03398684412240982, "global_step": 232610, "epoch": 2613, "lr": 4.9847763591461035e-06}
{"train_loss": 0.0396774597465992, "global_step": 232611, "epoch": 2613, "lr": 4.984524025625998e-06}
{"train_loss": 0.046846527606248856, "global_step": 232612, "epoch": 2613, "lr": 4.984271698157683e-06}
{"train_loss": 0.10108323395252228, "global_step": 232613, "epoch": 2613, "lr": 4.9840193767411725e-06}
{"train_loss": 0.031758155673742294, "global_step": 232614, "epoch": 2613, "lr": 4.983767061376527e-06}
{"train_loss": 0.037091393023729324, "global_step": 232615, "epoch": 2613, "lr": 4.98351475206375e-06}
{"train_loss": 0.03294198215007782, "global_step": 232616, "epoch": 2613, "lr": 4.9832624488029e-06}
{"train_loss": 0.04994630068540573, "global_step": 232617, "epoch": 2613, "lr": 4.983010151593998e-06}
{"train_loss": 0.05234549939632416, "global_step": 232618, "epoch": 2613, "lr": 4.982757860437088e-06}
{"train_loss": 0.024087652564048767, "global_step": 232619, "epoch": 2613, "lr": 4.982505575332186e-06}
{"train_loss": 0.03943682089447975, "global_step": 232620, "epoch": 2613, "lr": 4.982253296279349e-06}
{"train_loss": 0.05329166725277901, "global_step": 232621, "epoch": 2613, "lr": 4.982001023278593e-06}
{"train_loss": 0.052343424409627914, "global_step": 232622, "epoch": 2613, "lr": 4.981748756329963e-06}
{"train_loss": 0.09509805589914322, "global_step": 232623, "epoch": 2613, "lr": 4.9814964954334795e-06}
{"train_loss": 0.05384121835231781, "global_step": 232624, "epoch": 2613, "lr": 4.981244240589189e-06}
{"train_loss": 0.047210559248924255, "global_step": 232625, "epoch": 2613, "lr": 4.98099199179713e-06}
{"train_loss": 0.045600272715091705, "global_step": 232626, "epoch": 2613, "lr": 4.980739749057312e-06}
{"train_loss": 0.059784404933452606, "global_step": 232627, "epoch": 2613, "lr": 4.980487512369802e-06}
{"train_loss": 0.07148617506027222, "global_step": 232628, "epoch": 2613, "lr": 4.980235281734613e-06}
{"train_loss": 0.04661182314157486, "global_step": 232629, "epoch": 2613, "lr": 4.979983057151771e-06}
{"train_loss": 0.08872433006763458, "global_step": 232630, "epoch": 2613, "lr": 4.979730838621333e-06}
{"train_loss": 0.028181560337543488, "global_step": 232631, "epoch": 2613, "lr": 4.9794786261433134e-06}
{"train_loss": 0.0463976226747036, "global_step": 232632, "epoch": 2613, "lr": 4.979226419717764e-06}
{"train_loss": 0.04057096317410469, "global_step": 232633, "epoch": 2613, "lr": 4.978974219344707e-06}
{"train_loss": 0.044083282351493835, "global_step": 232634, "epoch": 2613, "lr": 4.9787220250241684e-06}
{"train_loss": 0.026130352169275284, "global_step": 232635, "epoch": 2613, "lr": 4.978469836756206e-06}
{"train_loss": 0.058598317205905914, "global_step": 232636, "epoch": 2613, "lr": 4.978217654540823e-06}
{"train_loss": 0.03900393098592758, "global_step": 232637, "epoch": 2613, "lr": 4.977965478378088e-06}
{"train_loss": 0.08526858687400818, "global_step": 232638, "epoch": 2613, "lr": 4.977713308268001e-06}
{"train_loss": 0.03980591148138046, "global_step": 232639, "epoch": 2613, "lr": 4.977461144210621e-06}
{"train_loss": 0.04249601066112518, "global_step": 232640, "epoch": 2613, "lr": 4.977208986205967e-06}
{"train_loss": 0.023767633363604546, "global_step": 232641, "epoch": 2613, "lr": 4.976956834254076e-06}
{"train_loss": 0.03718720003962517, "global_step": 232642, "epoch": 2613, "lr": 4.976704688354994e-06}
{"train_loss": 0.03323814272880554, "global_step": 232643, "epoch": 2613, "lr": 4.976452548508736e-06}
{"train_loss": 0.04642949253320694, "global_step": 232644, "epoch": 2613, "lr": 4.976200414715354e-06}
{"train_loss": 0.0487611134944672, "global_step": 232645, "epoch": 2613, "lr": 4.975948286974863e-06, "val_loss": 9.374983787536621}
{"train_loss": 0.05254695564508438, "global_step": 232646, "epoch": 2614, "lr": 4.975696165287319e-06}
{"train_loss": 0.03088124841451645, "global_step": 232647, "epoch": 2614, "lr": 4.975444049652734e-06}
{"train_loss": 0.03967868909239769, "global_step": 232648, "epoch": 2614, "lr": 4.975191940071155e-06}
{"train_loss": 0.054010216146707535, "global_step": 232649, "epoch": 2614, "lr": 4.974939836542614e-06}
{"train_loss": 0.050192613154649734, "global_step": 232650, "epoch": 2614, "lr": 4.9746877390671465e-06}
{"train_loss": 0.12857581675052643, "global_step": 232651, "epoch": 2614, "lr": 4.974435647644776e-06}
{"train_loss": 0.04168695956468582, "global_step": 232652, "epoch": 2614, "lr": 4.9741835622755475e-06}
{"train_loss": 0.07028243690729141, "global_step": 232653, "epoch": 2614, "lr": 4.973931482959487e-06}
{"train_loss": 0.05857102572917938, "global_step": 232654, "epoch": 2614, "lr": 4.973679409696641e-06}
{"train_loss": 0.04685280844569206, "global_step": 232655, "epoch": 2614, "lr": 4.97342734248703e-06}
{"train_loss": 0.0786844864487648, "global_step": 232656, "epoch": 2614, "lr": 4.9731752813307e-06}
{"train_loss": 0.03507101908326149, "global_step": 232657, "epoch": 2614, "lr": 4.972923226227666e-06}
{"train_loss": 0.024108223617076874, "global_step": 232658, "epoch": 2614, "lr": 4.972671177177984e-06}
{"train_loss": 0.0411192923784256, "global_step": 232659, "epoch": 2614, "lr": 4.972419134181666e-06}
{"train_loss": 0.059408098459243774, "global_step": 232660, "epoch": 2614, "lr": 4.972167097238762e-06}
{"train_loss": 0.058058962225914, "global_step": 232661, "epoch": 2614, "lr": 4.971915066349309e-06}
{"train_loss": 0.06109314039349556, "global_step": 232662, "epoch": 2614, "lr": 4.971663041513319e-06}
{"train_loss": 0.061429642140865326, "global_step": 232663, "epoch": 2614, "lr": 4.971411022730854e-06}
{"train_loss": 0.011987910605967045, "global_step": 232664, "epoch": 2614, "lr": 4.9711590100019246e-06}
{"train_loss": 0.03960133716464043, "global_step": 232665, "epoch": 2614, "lr": 4.970907003326586e-06}
{"train_loss": 0.03474481403827667, "global_step": 232666, "epoch": 2614, "lr": 4.970655002704855e-06}
{"train_loss": 0.0628204494714737, "global_step": 232667, "epoch": 2614, "lr": 4.970403008136765e-06}
{"train_loss": 0.04474594444036484, "global_step": 232668, "epoch": 2614, "lr": 4.97015101962236e-06}
{"train_loss": 0.026033908128738403, "global_step": 232669, "epoch": 2614, "lr": 4.9698990371616624e-06}
{"train_loss": 0.03374490141868591, "global_step": 232670, "epoch": 2614, "lr": 4.969647060754723e-06}
{"train_loss": 0.07625595480203629, "global_step": 232671, "epoch": 2614, "lr": 4.969395090401557e-06}
{"train_loss": 0.05217432230710983, "global_step": 232672, "epoch": 2614, "lr": 4.969143126102216e-06}
{"train_loss": 0.029126819223165512, "global_step": 232673, "epoch": 2614, "lr": 4.968891167856726e-06}
{"train_loss": 0.04670022800564766, "global_step": 232674, "epoch": 2614, "lr": 4.96863921566511e-06}
{"train_loss": 0.044375333935022354, "global_step": 232675, "epoch": 2614, "lr": 4.9683872695274195e-06}
{"train_loss": 0.04849889501929283, "global_step": 232676, "epoch": 2614, "lr": 4.968135329443674e-06}
{"train_loss": 0.04123198986053467, "global_step": 232677, "epoch": 2614, "lr": 4.967883395413919e-06}
{"train_loss": 0.028687549754977226, "global_step": 232678, "epoch": 2614, "lr": 4.967631467438177e-06}
{"train_loss": 0.04247267171740532, "global_step": 232679, "epoch": 2614, "lr": 4.967379545516482e-06}
{"train_loss": 0.036013051867485046, "global_step": 232680, "epoch": 2614, "lr": 4.967127629648894e-06}
{"train_loss": 0.03916585072875023, "global_step": 232681, "epoch": 2614, "lr": 4.966875719835407e-06}
{"train_loss": 0.10612686723470688, "global_step": 232682, "epoch": 2614, "lr": 4.966623816076088e-06}
{"train_loss": 0.030498307198286057, "global_step": 232683, "epoch": 2614, "lr": 4.966371918370949e-06}
{"train_loss": 0.06020426005125046, "global_step": 232684, "epoch": 2614, "lr": 4.966120026720045e-06}
{"train_loss": 0.062409672886133194, "global_step": 232685, "epoch": 2614, "lr": 4.965868141123381e-06}
{"train_loss": 0.050675224512815475, "global_step": 232686, "epoch": 2614, "lr": 4.965616261581024e-06}
{"train_loss": 0.076891228556633, "global_step": 232687, "epoch": 2614, "lr": 4.9653643880929745e-06}
{"train_loss": 0.12013420462608337, "global_step": 232688, "epoch": 2614, "lr": 4.965112520659299e-06}
{"train_loss": 0.015172576531767845, "global_step": 232689, "epoch": 2614, "lr": 4.964860659280001e-06}
{"train_loss": 0.06492143124341965, "global_step": 232690, "epoch": 2614, "lr": 4.9646088039551395e-06}
{"train_loss": 0.05453164502978325, "global_step": 232691, "epoch": 2614, "lr": 4.964356954684723e-06}
{"train_loss": 0.0910503938794136, "global_step": 232692, "epoch": 2614, "lr": 4.964105111468814e-06}
{"train_loss": 0.06351780891418457, "global_step": 232693, "epoch": 2614, "lr": 4.963853274307423e-06}
{"train_loss": 0.08828697353601456, "global_step": 232694, "epoch": 2614, "lr": 4.963601443200605e-06}
{"train_loss": 0.06426456570625305, "global_step": 232695, "epoch": 2614, "lr": 4.963349618148366e-06}
{"train_loss": 0.059190649539232254, "global_step": 232696, "epoch": 2614, "lr": 4.963097799150762e-06}
{"train_loss": 0.08013241738080978, "global_step": 232697, "epoch": 2614, "lr": 4.962845986207826e-06}
{"train_loss": 0.03278335928916931, "global_step": 232698, "epoch": 2614, "lr": 4.9625941793195795e-06}
{"train_loss": 0.014797016978263855, "global_step": 232699, "epoch": 2614, "lr": 4.962342378486073e-06}
{"train_loss": 0.05437720566987991, "global_step": 232700, "epoch": 2614, "lr": 4.962090583707318e-06}
{"train_loss": 0.03697122260928154, "global_step": 232701, "epoch": 2614, "lr": 4.96183879498337e-06}
{"train_loss": 0.04869354888796806, "global_step": 232702, "epoch": 2614, "lr": 4.961587012314251e-06}
{"train_loss": 0.06441548466682434, "global_step": 232703, "epoch": 2614, "lr": 4.9613352356999985e-06}
{"train_loss": 0.07075652480125427, "global_step": 232704, "epoch": 2614, "lr": 4.9610834651406425e-06}
{"train_loss": 0.0547715499997139, "global_step": 232705, "epoch": 2614, "lr": 4.960831700636226e-06}
{"train_loss": 0.03202115371823311, "global_step": 232706, "epoch": 2614, "lr": 4.960579942186777e-06}
{"train_loss": 0.03896210715174675, "global_step": 232707, "epoch": 2614, "lr": 4.960328189792323e-06}
{"train_loss": 0.05739235505461693, "global_step": 232708, "epoch": 2614, "lr": 4.96007644345291e-06}
{"train_loss": 0.06527666747570038, "global_step": 232709, "epoch": 2614, "lr": 4.9598247031685586e-06}
{"train_loss": 0.1257704794406891, "global_step": 232710, "epoch": 2614, "lr": 4.959572968939313e-06}
{"train_loss": 0.04966269060969353, "global_step": 232711, "epoch": 2614, "lr": 4.959321240765213e-06}
{"train_loss": 0.019277220591902733, "global_step": 232712, "epoch": 2614, "lr": 4.95906951864627e-06}
{"train_loss": 0.016717003658413887, "global_step": 232713, "epoch": 2614, "lr": 4.958817802582538e-06}
{"train_loss": 0.06243395805358887, "global_step": 232714, "epoch": 2614, "lr": 4.958566092574035e-06}
{"train_loss": 0.05567945912480354, "global_step": 232715, "epoch": 2614, "lr": 4.958314388620805e-06}
{"train_loss": 0.05681430548429489, "global_step": 232716, "epoch": 2614, "lr": 4.958062690722892e-06}
{"train_loss": 0.01973223127424717, "global_step": 232717, "epoch": 2614, "lr": 4.957810998880308e-06}
{"train_loss": 0.037489213049411774, "global_step": 232718, "epoch": 2614, "lr": 4.9575593130931074e-06}
{"train_loss": 0.06728392094373703, "global_step": 232719, "epoch": 2614, "lr": 4.957307633361302e-06}
{"train_loss": 0.07727181911468506, "global_step": 232720, "epoch": 2614, "lr": 4.957055959684948e-06}
{"train_loss": 0.03565206006169319, "global_step": 232721, "epoch": 2614, "lr": 4.95680429206406e-06}
{"train_loss": 0.04578176885843277, "global_step": 232722, "epoch": 2614, "lr": 4.956552630498689e-06}
{"train_loss": 0.0757240280508995, "global_step": 232723, "epoch": 2614, "lr": 4.956300974988853e-06}
{"train_loss": 0.03835352137684822, "global_step": 232724, "epoch": 2614, "lr": 4.9560493255346e-06}
{"train_loss": 0.052952200174331665, "global_step": 232725, "epoch": 2614, "lr": 4.955797682135949e-06}
{"train_loss": 0.060663290321826935, "global_step": 232726, "epoch": 2614, "lr": 4.955546044792952e-06}
{"train_loss": 0.03629332780838013, "global_step": 232727, "epoch": 2614, "lr": 4.955294413505623e-06}
{"train_loss": 0.027412237599492073, "global_step": 232728, "epoch": 2614, "lr": 4.955042788274017e-06}
{"train_loss": 0.01172922644764185, "global_step": 232729, "epoch": 2614, "lr": 4.954791169098144e-06}
{"train_loss": 0.05044584348797798, "global_step": 232730, "epoch": 2614, "lr": 4.954539555978061e-06}
{"train_loss": 0.04848737269639969, "global_step": 232731, "epoch": 2614, "lr": 4.954287948913783e-06}
{"train_loss": 0.05054553225636482, "global_step": 232732, "epoch": 2614, "lr": 4.95403634790535e-06}
{"train_loss": 0.05801462009549141, "global_step": 232733, "epoch": 2614, "lr": 4.953784752952812e-06}
{"train_loss": 0.052100008144304995, "global_step": 232734, "epoch": 2614, "lr": 4.953533164056173e-06, "val_loss": 9.204154014587402}
{"train_loss": 0.08017268031835556, "global_step": 232735, "epoch": 2615, "lr": 4.953281581215497e-06}
{"train_loss": 0.04461441561579704, "global_step": 232736, "epoch": 2615, "lr": 4.953030004430786e-06}
{"train_loss": 0.024579215794801712, "global_step": 232737, "epoch": 2615, "lr": 4.952778433702104e-06}
{"train_loss": 0.03936516493558884, "global_step": 232738, "epoch": 2615, "lr": 4.952526869029467e-06}
{"train_loss": 0.03746875748038292, "global_step": 232739, "epoch": 2615, "lr": 4.952275310412918e-06}
{"train_loss": 0.04210623353719711, "global_step": 232740, "epoch": 2615, "lr": 4.952023757852481e-06}
{"train_loss": 0.054182667285203934, "global_step": 232741, "epoch": 2615, "lr": 4.951772211348205e-06}
{"train_loss": 0.014712762087583542, "global_step": 232742, "epoch": 2615, "lr": 4.9515206709001e-06}
{"train_loss": 0.02919214591383934, "global_step": 232743, "epoch": 2615, "lr": 4.951269136508224e-06}
{"train_loss": 0.026844428852200508, "global_step": 232744, "epoch": 2615, "lr": 4.951017608172603e-06}
{"train_loss": 0.07783925533294678, "global_step": 232745, "epoch": 2615, "lr": 4.95076608589326e-06}
{"train_loss": 0.01215655729174614, "global_step": 232746, "epoch": 2615, "lr": 4.9505145696702446e-06}
{"train_loss": 0.11325766891241074, "global_step": 232747, "epoch": 2615, "lr": 4.9502630595035795e-06}
{"train_loss": 0.027720632031559944, "global_step": 232748, "epoch": 2615, "lr": 4.9500115553933025e-06}
{"train_loss": 0.031719010323286057, "global_step": 232749, "epoch": 2615, "lr": 4.949760057339453e-06}
{"train_loss": 0.07183675467967987, "global_step": 232750, "epoch": 2615, "lr": 4.949508565342053e-06}
{"train_loss": 0.04751588776707649, "global_step": 232751, "epoch": 2615, "lr": 4.949257079401137e-06}
{"train_loss": 0.0513780415058136, "global_step": 232752, "epoch": 2615, "lr": 4.949005599516754e-06}
{"train_loss": 0.031131863594055176, "global_step": 232753, "epoch": 2615, "lr": 4.94875412568892e-06}
{"train_loss": 0.03332659974694252, "global_step": 232754, "epoch": 2615, "lr": 4.9485026579176864e-06}
{"train_loss": 0.012033793143928051, "global_step": 232755, "epoch": 2615, "lr": 4.948251196203069e-06}
{"train_loss": 0.030716780573129654, "global_step": 232756, "epoch": 2615, "lr": 4.947999740545118e-06}
{"train_loss": 0.10855944454669952, "global_step": 232757, "epoch": 2615, "lr": 4.9477482909438485e-06}
{"train_loss": 0.04752941429615021, "global_step": 232758, "epoch": 2615, "lr": 4.947496847399319e-06}
{"train_loss": 0.02851821295917034, "global_step": 232759, "epoch": 2615, "lr": 4.9472454099115384e-06}
{"train_loss": 0.06605571508407593, "global_step": 232760, "epoch": 2615, "lr": 4.9469939784805576e-06}
{"train_loss": 0.05390138924121857, "global_step": 232761, "epoch": 2615, "lr": 4.9467425531063984e-06}
{"train_loss": 0.030094074085354805, "global_step": 232762, "epoch": 2615, "lr": 4.946491133789111e-06}
{"train_loss": 0.06486178189516068, "global_step": 232763, "epoch": 2615, "lr": 4.9462397205287055e-06}
{"train_loss": 0.033291954547166824, "global_step": 232764, "epoch": 2615, "lr": 4.9459883133252395e-06}
{"train_loss": 0.03524135425686836, "global_step": 232765, "epoch": 2615, "lr": 4.945736912178728e-06}
{"train_loss": 0.03635722026228905, "global_step": 232766, "epoch": 2615, "lr": 4.945485517089221e-06}
{"train_loss": 0.047543127089738846, "global_step": 232767, "epoch": 2615, "lr": 4.945234128056736e-06}
{"train_loss": 0.032629214227199554, "global_step": 232768, "epoch": 2615, "lr": 4.9449827450813165e-06}
{"train_loss": 0.05099765211343765, "global_step": 232769, "epoch": 2615, "lr": 4.944731368163003e-06}
{"train_loss": 0.07966553419828415, "global_step": 232770, "epoch": 2615, "lr": 4.944479997301815e-06}
{"train_loss": 0.01676255278289318, "global_step": 232771, "epoch": 2615, "lr": 4.944228632497794e-06}
{"train_loss": 0.02329982817173004, "global_step": 232772, "epoch": 2615, "lr": 4.943977273750971e-06}
{"train_loss": 0.05162465572357178, "global_step": 232773, "epoch": 2615, "lr": 4.943725921061387e-06}
{"train_loss": 0.06558280438184738, "global_step": 232774, "epoch": 2615, "lr": 4.943474574429057e-06}
{"train_loss": 0.03697643056511879, "global_step": 232775, "epoch": 2615, "lr": 4.943223233854044e-06}
{"train_loss": 0.06751307100057602, "global_step": 232776, "epoch": 2615, "lr": 4.9429718993363515e-06}
{"train_loss": 0.049519605934619904, "global_step": 232777, "epoch": 2615, "lr": 4.942720570876036e-06}
{"train_loss": 0.049650054425001144, "global_step": 232778, "epoch": 2615, "lr": 4.9424692484731195e-06}
{"train_loss": 0.032172128558158875, "global_step": 232779, "epoch": 2615, "lr": 4.94221793212764e-06}
{"train_loss": 0.05025535821914673, "global_step": 232780, "epoch": 2615, "lr": 4.9419666218396276e-06}
{"train_loss": 0.025000428780913353, "global_step": 232781, "epoch": 2615, "lr": 4.941715317609124e-06}
{"train_loss": 0.06850191950798035, "global_step": 232782, "epoch": 2615, "lr": 4.941464019436159e-06}
{"train_loss": 0.05364000052213669, "global_step": 232783, "epoch": 2615, "lr": 4.941212727320754e-06}
{"train_loss": 0.09090583771467209, "global_step": 232784, "epoch": 2615, "lr": 4.940961441262965e-06}
{"train_loss": 0.04415198042988777, "global_step": 232785, "epoch": 2615, "lr": 4.940710161262801e-06}
{"train_loss": 0.05796051770448685, "global_step": 232786, "epoch": 2615, "lr": 4.940458887320326e-06}
{"train_loss": 0.03194905444979668, "global_step": 232787, "epoch": 2615, "lr": 4.940207619435544e-06}
{"train_loss": 0.08276922255754471, "global_step": 232788, "epoch": 2615, "lr": 4.93995635760851e-06}
{"train_loss": 0.02936503104865551, "global_step": 232789, "epoch": 2615, "lr": 4.939705101839243e-06}
{"train_loss": 0.03592432662844658, "global_step": 232790, "epoch": 2615, "lr": 4.93945385212779e-06}
{"train_loss": 0.05080372467637062, "global_step": 232791, "epoch": 2615, "lr": 4.939202608474169e-06}
{"train_loss": 0.060091111809015274, "global_step": 232792, "epoch": 2615, "lr": 4.938951370878431e-06}
{"train_loss": 0.05642440915107727, "global_step": 232793, "epoch": 2615, "lr": 4.938700139340597e-06}
{"train_loss": 0.043764203786849976, "global_step": 232794, "epoch": 2615, "lr": 4.938448913860711e-06}
{"train_loss": 0.07264535874128342, "global_step": 232795, "epoch": 2615, "lr": 4.938197694438795e-06}
{"train_loss": 0.08952935039997101, "global_step": 232796, "epoch": 2615, "lr": 4.937946481074896e-06}
{"train_loss": 0.039656586945056915, "global_step": 232797, "epoch": 2615, "lr": 4.937695273769033e-06}
{"train_loss": 0.04735342785716057, "global_step": 232798, "epoch": 2615, "lr": 4.9374440725212574e-06}
{"train_loss": 0.01987110637128353, "global_step": 232799, "epoch": 2615, "lr": 4.9371928773315815e-06}
{"train_loss": 0.061932407319545746, "global_step": 232800, "epoch": 2615, "lr": 4.936941688200064e-06}
{"train_loss": 0.05947447568178177, "global_step": 232801, "epoch": 2615, "lr": 4.936690505126712e-06}
{"train_loss": 0.06750324368476868, "global_step": 232802, "epoch": 2615, "lr": 4.936439328111586e-06}
{"train_loss": 0.055175792425870895, "global_step": 232803, "epoch": 2615, "lr": 4.936188157154692e-06}
{"train_loss": 0.09837391972541809, "global_step": 232804, "epoch": 2615, "lr": 4.93593699225609e-06}
{"train_loss": 0.02522308938205242, "global_step": 232805, "epoch": 2615, "lr": 4.935685833415793e-06}
{"train_loss": 0.04748876765370369, "global_step": 232806, "epoch": 2615, "lr": 4.935434680633844e-06}
{"train_loss": 0.007288468535989523, "global_step": 232807, "epoch": 2615, "lr": 4.935183533910287e-06}
{"train_loss": 0.06287573277950287, "global_step": 232808, "epoch": 2615, "lr": 4.934932393245134e-06}
{"train_loss": 0.03745878115296364, "global_step": 232809, "epoch": 2615, "lr": 4.934681258638441e-06}
{"train_loss": 0.06472623348236084, "global_step": 232810, "epoch": 2615, "lr": 4.934430130090223e-06}
{"train_loss": 0.06983373314142227, "global_step": 232811, "epoch": 2615, "lr": 4.934179007600526e-06}
{"train_loss": 0.06584129482507706, "global_step": 232812, "epoch": 2615, "lr": 4.933927891169377e-06}
{"train_loss": 0.03250692039728165, "global_step": 232813, "epoch": 2615, "lr": 4.933676780796814e-06}
{"train_loss": 0.039592161774635315, "global_step": 232814, "epoch": 2615, "lr": 4.933425676482861e-06}
{"train_loss": 0.041279301047325134, "global_step": 232815, "epoch": 2615, "lr": 4.933174578227573e-06}
{"train_loss": 0.07306913286447525, "global_step": 232816, "epoch": 2615, "lr": 4.93292348603096e-06}
{"train_loss": 0.06478777527809143, "global_step": 232817, "epoch": 2615, "lr": 4.932672399893073e-06}
{"train_loss": 0.05931829661130905, "global_step": 232818, "epoch": 2615, "lr": 4.932421319813929e-06}
{"train_loss": 0.05108480155467987, "global_step": 232819, "epoch": 2615, "lr": 4.932170245793582e-06}
{"train_loss": 0.024329114705324173, "global_step": 232820, "epoch": 2615, "lr": 4.93191917783205e-06}
{"train_loss": 0.05750647187232971, "global_step": 232821, "epoch": 2615, "lr": 4.9316681159293775e-06}
{"train_loss": 0.06290539354085922, "global_step": 232822, "epoch": 2615, "lr": 4.931417060085591e-06}
{"train_loss": 0.049142586281837035, "global_step": 232823, "epoch": 2615, "lr": 4.931166010300725e-06, "val_loss": 9.359817504882812, "train_action_mse_error": 19.953170776367188}
{"train_loss": 0.07891610264778137, "global_step": 232824, "epoch": 2616, "lr": 4.930914966574818e-06}
{"train_loss": 0.03383973240852356, "global_step": 232825, "epoch": 2616, "lr": 4.930663928907892e-06}
{"train_loss": 0.050775062292814255, "global_step": 232826, "epoch": 2616, "lr": 4.930412897299996e-06}
{"train_loss": 0.04555708169937134, "global_step": 232827, "epoch": 2616, "lr": 4.930161871751149e-06}
{"train_loss": 0.06419549882411957, "global_step": 232828, "epoch": 2616, "lr": 4.9299108522613994e-06}
{"train_loss": 0.05740395933389664, "global_step": 232829, "epoch": 2616, "lr": 4.929659838830769e-06}
{"train_loss": 0.05446837469935417, "global_step": 232830, "epoch": 2616, "lr": 4.929408831459304e-06}
{"train_loss": 0.048732537776231766, "global_step": 232831, "epoch": 2616, "lr": 4.9291578301470185e-06}
{"train_loss": 0.04430306330323219, "global_step": 232832, "epoch": 2616, "lr": 4.928906834893971e-06}
{"train_loss": 0.04516005888581276, "global_step": 232833, "epoch": 2616, "lr": 4.92865584570017e-06}
{"train_loss": 0.044631391763687134, "global_step": 232834, "epoch": 2616, "lr": 4.9284048625656776e-06}
{"train_loss": 0.03573428466916084, "global_step": 232835, "epoch": 2616, "lr": 4.928153885490494e-06}
{"train_loss": 0.06023789197206497, "global_step": 232836, "epoch": 2616, "lr": 4.927902914474686e-06}
{"train_loss": 0.025719204917550087, "global_step": 232837, "epoch": 2616, "lr": 4.927651949518263e-06}
{"train_loss": 0.08087069541215897, "global_step": 232838, "epoch": 2616, "lr": 4.9274009906212724e-06}
{"train_loss": 0.08325807750225067, "global_step": 232839, "epoch": 2616, "lr": 4.92715003778374e-06}
{"train_loss": 0.032645177096128464, "global_step": 232840, "epoch": 2616, "lr": 4.92689909100571e-06}
{"train_loss": 0.040902674198150635, "global_step": 232841, "epoch": 2616, "lr": 4.9266481502872e-06}
{"train_loss": 0.031131180003285408, "global_step": 232842, "epoch": 2616, "lr": 4.9263972156282474e-06}
{"train_loss": 0.042376559227705, "global_step": 232843, "epoch": 2616, "lr": 4.92614628702891e-06}
{"train_loss": 0.12752462923526764, "global_step": 232844, "epoch": 2616, "lr": 4.9258953644891855e-06}
{"train_loss": 0.05336165800690651, "global_step": 232845, "epoch": 2616, "lr": 4.925644448009137e-06}
{"train_loss": 0.05862119793891907, "global_step": 232846, "epoch": 2616, "lr": 4.9253935375887794e-06}
{"train_loss": 0.042416755110025406, "global_step": 232847, "epoch": 2616, "lr": 4.925142633228158e-06}
{"train_loss": 0.05139796808362007, "global_step": 232848, "epoch": 2616, "lr": 4.924891734927295e-06}
{"train_loss": 0.032402969896793365, "global_step": 232849, "epoch": 2616, "lr": 4.92464084268624e-06}
{"train_loss": 0.03014574572443962, "global_step": 232850, "epoch": 2616, "lr": 4.92438995650501e-06}
{"train_loss": 0.05649679899215698, "global_step": 232851, "epoch": 2616, "lr": 4.924139076383655e-06}
{"train_loss": 0.05558802932500839, "global_step": 232852, "epoch": 2616, "lr": 4.923888202322191e-06}
{"train_loss": 0.04602387547492981, "global_step": 232853, "epoch": 2616, "lr": 4.923637334320669e-06}
{"train_loss": 0.05864870920777321, "global_step": 232854, "epoch": 2616, "lr": 4.923386472379104e-06}
{"train_loss": 0.07654157280921936, "global_step": 232855, "epoch": 2616, "lr": 4.923135616497554e-06}
{"train_loss": 0.06281314790248871, "global_step": 232856, "epoch": 2616, "lr": 4.922884766676028e-06}
{"train_loss": 0.0558503158390522, "global_step": 232857, "epoch": 2616, "lr": 4.922633922914577e-06}
{"train_loss": 0.044381510466337204, "global_step": 232858, "epoch": 2616, "lr": 4.922383085213223e-06}
{"train_loss": 0.05445554107427597, "global_step": 232859, "epoch": 2616, "lr": 4.92213225357201e-06}
{"train_loss": 0.07161961495876312, "global_step": 232860, "epoch": 2616, "lr": 4.921881427990971e-06}
{"train_loss": 0.07514356821775436, "global_step": 232861, "epoch": 2616, "lr": 4.921630608470123e-06}
{"train_loss": 0.016769176349043846, "global_step": 232862, "epoch": 2616, "lr": 4.921379795009523e-06}
{"train_loss": 0.031489647924900055, "global_step": 232863, "epoch": 2616, "lr": 4.921128987609186e-06}
{"train_loss": 0.0489489920437336, "global_step": 232864, "epoch": 2616, "lr": 4.9208781862691625e-06}
{"train_loss": 0.06137765571475029, "global_step": 232865, "epoch": 2616, "lr": 4.920627390989469e-06}
{"train_loss": 0.051307618618011475, "global_step": 232866, "epoch": 2616, "lr": 4.920376601770155e-06}
{"train_loss": 0.06273161619901657, "global_step": 232867, "epoch": 2616, "lr": 4.920125818611238e-06}
{"train_loss": 0.059407688677310944, "global_step": 232868, "epoch": 2616, "lr": 4.919875041512772e-06}
{"train_loss": 0.07503893226385117, "global_step": 232869, "epoch": 2616, "lr": 4.919624270474765e-06}
{"train_loss": 0.1160932183265686, "global_step": 232870, "epoch": 2616, "lr": 4.919373505497283e-06}
{"train_loss": 0.02425120584666729, "global_step": 232871, "epoch": 2616, "lr": 4.919122746580324e-06}
{"train_loss": 0.031736936420202255, "global_step": 232872, "epoch": 2616, "lr": 4.9188719937239506e-06}
{"train_loss": 0.04313378036022186, "global_step": 232873, "epoch": 2616, "lr": 4.918621246928179e-06}
{"train_loss": 0.05813954398036003, "global_step": 232874, "epoch": 2616, "lr": 4.91837050619306e-06}
{"train_loss": 0.019491367042064667, "global_step": 232875, "epoch": 2616, "lr": 4.918119771518604e-06}
{"train_loss": 0.039232440292835236, "global_step": 232876, "epoch": 2616, "lr": 4.917869042904866e-06}
{"train_loss": 0.0873558446764946, "global_step": 232877, "epoch": 2616, "lr": 4.917618320351863e-06}
{"train_loss": 0.02940838783979416, "global_step": 232878, "epoch": 2616, "lr": 4.917367603859635e-06}
{"train_loss": 0.10098736733198166, "global_step": 232879, "epoch": 2616, "lr": 4.917116893428231e-06}
{"train_loss": 0.062155794352293015, "global_step": 232880, "epoch": 2616, "lr": 4.916866189057656e-06}
{"train_loss": 0.06359264999628067, "global_step": 232881, "epoch": 2616, "lr": 4.916615490747972e-06}
{"train_loss": 0.05591217800974846, "global_step": 232882, "epoch": 2616, "lr": 4.91636479849919e-06}
{"train_loss": 0.045873284339904785, "global_step": 232883, "epoch": 2616, "lr": 4.91611411231136e-06}
{"train_loss": 0.0546182319521904, "global_step": 232884, "epoch": 2616, "lr": 4.915863432184503e-06}
{"train_loss": 0.045089200139045715, "global_step": 232885, "epoch": 2616, "lr": 4.915612758118671e-06}
{"train_loss": 0.0403946228325367, "global_step": 232886, "epoch": 2616, "lr": 4.915362090113867e-06}
{"train_loss": 0.024614978581666946, "global_step": 232887, "epoch": 2616, "lr": 4.9151114281701606e-06}
{"train_loss": 0.0603509396314621, "global_step": 232888, "epoch": 2616, "lr": 4.914860772287555e-06}
{"train_loss": 0.08860620111227036, "global_step": 232889, "epoch": 2616, "lr": 4.9146101224661066e-06}
{"train_loss": 0.04121330380439758, "global_step": 232890, "epoch": 2616, "lr": 4.914359478705826e-06}
{"train_loss": 0.037105727940797806, "global_step": 232891, "epoch": 2616, "lr": 4.914108841006776e-06}
{"train_loss": 0.045755885541439056, "global_step": 232892, "epoch": 2616, "lr": 4.91385820936896e-06}
{"train_loss": 0.028959987685084343, "global_step": 232893, "epoch": 2616, "lr": 4.91360758379244e-06}
{"train_loss": 0.0438498817384243, "global_step": 232894, "epoch": 2616, "lr": 4.913356964277227e-06}
{"train_loss": 0.012007245793938637, "global_step": 232895, "epoch": 2616, "lr": 4.913106350823371e-06}
{"train_loss": 0.030542388558387756, "global_step": 232896, "epoch": 2616, "lr": 4.912855743430888e-06}
{"train_loss": 0.0419769287109375, "global_step": 232897, "epoch": 2616, "lr": 4.91260514209983e-06}
{"train_loss": 0.06775451451539993, "global_step": 232898, "epoch": 2616, "lr": 4.912354546830228e-06}
{"train_loss": 0.04480791836977005, "global_step": 232899, "epoch": 2616, "lr": 4.912103957622094e-06}
{"train_loss": 0.07552377134561539, "global_step": 232900, "epoch": 2616, "lr": 4.911853374475495e-06}
{"train_loss": 0.057087529450654984, "global_step": 232901, "epoch": 2616, "lr": 4.911602797390435e-06}
{"train_loss": 0.04513343423604965, "global_step": 232902, "epoch": 2616, "lr": 4.911352226366966e-06}
{"train_loss": 0.08828391879796982, "global_step": 232903, "epoch": 2616, "lr": 4.911101661405115e-06}
{"train_loss": 0.046985872089862823, "global_step": 232904, "epoch": 2616, "lr": 4.910851102504921e-06}
{"train_loss": 0.03390028327703476, "global_step": 232905, "epoch": 2616, "lr": 4.910600549666405e-06}
{"train_loss": 0.0544198676943779, "global_step": 232906, "epoch": 2616, "lr": 4.9103500028896185e-06}
{"train_loss": 0.08556937426328659, "global_step": 232907, "epoch": 2616, "lr": 4.910099462174572e-06}
{"train_loss": 0.08699960261583328, "global_step": 232908, "epoch": 2616, "lr": 4.909848927521327e-06}
{"train_loss": 0.06548457592725754, "global_step": 232909, "epoch": 2616, "lr": 4.909598398929893e-06}
{"train_loss": 0.03491728752851486, "global_step": 232910, "epoch": 2616, "lr": 4.9093478764003274e-06}
{"train_loss": 0.0389745868742466, "global_step": 232911, "epoch": 2616, "lr": 4.909097359932635e-06}
{"train_loss": 0.05332292051295216, "global_step": 232912, "epoch": 2616, "lr": 4.908846849526877e-06, "val_loss": 9.252870559692383}
{"train_loss": 0.031343214213848114, "global_step": 232913, "epoch": 2617, "lr": 4.908596345183064e-06}
{"train_loss": 0.013957157731056213, "global_step": 232914, "epoch": 2617, "lr": 4.908345846901241e-06}
{"train_loss": 0.07089751213788986, "global_step": 232915, "epoch": 2617, "lr": 4.908095354681452e-06}
{"train_loss": 0.09165708720684052, "global_step": 232916, "epoch": 2617, "lr": 4.907844868523714e-06}
{"train_loss": 0.04831911623477936, "global_step": 232917, "epoch": 2617, "lr": 4.907594388428072e-06}
{"train_loss": 0.06969153881072998, "global_step": 232918, "epoch": 2617, "lr": 4.9073439143945465e-06}
{"train_loss": 0.05596555396914482, "global_step": 232919, "epoch": 2617, "lr": 4.907093446423183e-06}
{"train_loss": 0.025853004306554794, "global_step": 232920, "epoch": 2617, "lr": 4.9068429845140095e-06}
{"train_loss": 0.025242745876312256, "global_step": 232921, "epoch": 2617, "lr": 4.9065925286670644e-06}
{"train_loss": 0.0380457304418087, "global_step": 232922, "epoch": 2617, "lr": 4.90634207888237e-06}
{"train_loss": 0.04275289922952652, "global_step": 232923, "epoch": 2617, "lr": 4.906091635159981e-06}
{"train_loss": 0.03989206254482269, "global_step": 232924, "epoch": 2617, "lr": 4.905841197499911e-06}
{"train_loss": 0.019190577790141106, "global_step": 232925, "epoch": 2617, "lr": 4.905590765902201e-06}
{"train_loss": 0.05213114246726036, "global_step": 232926, "epoch": 2617, "lr": 4.905340340366882e-06}
{"train_loss": 0.08233381062746048, "global_step": 232927, "epoch": 2617, "lr": 4.9050899208940015e-06}
{"train_loss": 0.034247465431690216, "global_step": 232928, "epoch": 2617, "lr": 4.904839507483571e-06}
{"train_loss": 0.03640221804380417, "global_step": 232929, "epoch": 2617, "lr": 4.9045891001356425e-06}
{"train_loss": 0.06472192704677582, "global_step": 232930, "epoch": 2617, "lr": 4.904338698850236e-06}
{"train_loss": 0.03904994949698448, "global_step": 232931, "epoch": 2617, "lr": 4.904088303627396e-06}
{"train_loss": 0.045706938952207565, "global_step": 232932, "epoch": 2617, "lr": 4.903837914467147e-06}
{"train_loss": 0.02968544326722622, "global_step": 232933, "epoch": 2617, "lr": 4.903587531369536e-06}
{"train_loss": 0.07050569355487823, "global_step": 232934, "epoch": 2617, "lr": 4.9033371543345755e-06}
{"train_loss": 0.027476148679852486, "global_step": 232935, "epoch": 2617, "lr": 4.903086783362321e-06}
{"train_loss": 0.04571269080042839, "global_step": 232936, "epoch": 2617, "lr": 4.902836418452789e-06}
{"train_loss": 0.08550231158733368, "global_step": 232937, "epoch": 2617, "lr": 4.902586059606035e-06}
{"train_loss": 0.0660516619682312, "global_step": 232938, "epoch": 2617, "lr": 4.902335706822075e-06}
{"train_loss": 0.07911732792854309, "global_step": 232939, "epoch": 2617, "lr": 4.902085360100933e-06}
{"train_loss": 0.07229583710432053, "global_step": 232940, "epoch": 2617, "lr": 4.901835019442669e-06}
{"train_loss": 0.034843623638153076, "global_step": 232941, "epoch": 2617, "lr": 4.9015846848472935e-06}
{"train_loss": 0.03786478564143181, "global_step": 232942, "epoch": 2617, "lr": 4.901334356314863e-06}
{"train_loss": 0.08702845126390457, "global_step": 232943, "epoch": 2617, "lr": 4.9010840338453875e-06}
{"train_loss": 0.04539724811911583, "global_step": 232944, "epoch": 2617, "lr": 4.900833717438919e-06}
{"train_loss": 0.02795889787375927, "global_step": 232945, "epoch": 2617, "lr": 4.900583407095471e-06}
{"train_loss": 0.043439898639917374, "global_step": 232946, "epoch": 2617, "lr": 4.900333102815108e-06}
{"train_loss": 0.043909668922424316, "global_step": 232947, "epoch": 2617, "lr": 4.900082804597827e-06}
{"train_loss": 0.06631004065275192, "global_step": 232948, "epoch": 2617, "lr": 4.899832512443697e-06}
{"train_loss": 0.0421760156750679, "global_step": 232949, "epoch": 2617, "lr": 4.8995822263527225e-06}
{"train_loss": 0.08808205276727676, "global_step": 232950, "epoch": 2617, "lr": 4.8993319463249476e-06}
{"train_loss": 0.05126381665468216, "global_step": 232951, "epoch": 2617, "lr": 4.8990816723604225e-06}
{"train_loss": 0.034554533660411835, "global_step": 232952, "epoch": 2617, "lr": 4.898831404459153e-06}
{"train_loss": 0.060437288135290146, "global_step": 232953, "epoch": 2617, "lr": 4.898581142621195e-06}
{"train_loss": 0.04836416244506836, "global_step": 232954, "epoch": 2617, "lr": 4.8983308868465635e-06}
{"train_loss": 0.044070519506931305, "global_step": 232955, "epoch": 2617, "lr": 4.898080637135311e-06}
{"train_loss": 0.0702449157834053, "global_step": 232956, "epoch": 2617, "lr": 4.897830393487452e-06}
{"train_loss": 0.0615929439663887, "global_step": 232957, "epoch": 2617, "lr": 4.897580155903042e-06}
{"train_loss": 0.04130173102021217, "global_step": 232958, "epoch": 2617, "lr": 4.897329924382093e-06}
{"train_loss": 0.02940213307738304, "global_step": 232959, "epoch": 2617, "lr": 4.897079698924656e-06}
{"train_loss": 0.032500483095645905, "global_step": 232960, "epoch": 2617, "lr": 4.896829479530746e-06}
{"train_loss": 0.051700614392757416, "global_step": 232961, "epoch": 2617, "lr": 4.896579266200424e-06}
{"train_loss": 0.06103129684925079, "global_step": 232962, "epoch": 2617, "lr": 4.8963290589336905e-06}
{"train_loss": 0.06395915895700455, "global_step": 232963, "epoch": 2617, "lr": 4.896078857730607e-06}
{"train_loss": 0.04889324679970741, "global_step": 232964, "epoch": 2617, "lr": 4.89582866259119e-06}
{"train_loss": 0.06723303347826004, "global_step": 232965, "epoch": 2617, "lr": 4.895578473515483e-06}
{"train_loss": 0.031144598498940468, "global_step": 232966, "epoch": 2617, "lr": 4.895328290503509e-06}
{"train_loss": 0.019092630594968796, "global_step": 232967, "epoch": 2617, "lr": 4.895078113555318e-06}
{"train_loss": 0.036976661533117294, "global_step": 232968, "epoch": 2617, "lr": 4.894827942670926e-06}
{"train_loss": 0.039423730224370956, "global_step": 232969, "epoch": 2617, "lr": 4.8945777778503735e-06}
{"train_loss": 0.07815169543027878, "global_step": 232970, "epoch": 2617, "lr": 4.894327619093702e-06}
{"train_loss": 0.07225752621889114, "global_step": 232971, "epoch": 2617, "lr": 4.894077466400943e-06}
{"train_loss": 0.0437811017036438, "global_step": 232972, "epoch": 2617, "lr": 4.89382731977211e-06}
{"train_loss": 0.0379689559340477, "global_step": 232973, "epoch": 2617, "lr": 4.893577179207265e-06}
{"train_loss": 0.035309087485075, "global_step": 232974, "epoch": 2617, "lr": 4.893327044706419e-06}
{"train_loss": 0.13594289124011993, "global_step": 232975, "epoch": 2617, "lr": 4.893076916269629e-06}
{"train_loss": 0.03332831710577011, "global_step": 232976, "epoch": 2617, "lr": 4.89282679389691e-06}
{"train_loss": 0.034221962094306946, "global_step": 232977, "epoch": 2617, "lr": 4.892576677588289e-06}
{"train_loss": 0.06964022666215897, "global_step": 232978, "epoch": 2617, "lr": 4.892326567343824e-06}
{"train_loss": 0.04408803954720497, "global_step": 232979, "epoch": 2617, "lr": 4.892076463163531e-06}
{"train_loss": 0.021730715408921242, "global_step": 232980, "epoch": 2617, "lr": 4.891826365047453e-06}
{"train_loss": 0.0377826951444149, "global_step": 232981, "epoch": 2617, "lr": 4.891576272995607e-06}
{"train_loss": 0.04532136768102646, "global_step": 232982, "epoch": 2617, "lr": 4.891326187008055e-06}
{"train_loss": 0.04671340808272362, "global_step": 232983, "epoch": 2617, "lr": 4.891076107084802e-06}
{"train_loss": 0.06252949684858322, "global_step": 232984, "epoch": 2617, "lr": 4.890826033225903e-06}
{"train_loss": 0.04740672558546066, "global_step": 232985, "epoch": 2617, "lr": 4.890575965431371e-06}
{"train_loss": 0.025859102606773376, "global_step": 232986, "epoch": 2617, "lr": 4.8903259037012535e-06}
{"train_loss": 0.0741744413971901, "global_step": 232987, "epoch": 2617, "lr": 4.890075848035591e-06}
{"train_loss": 0.07585654407739639, "global_step": 232988, "epoch": 2617, "lr": 4.8898257984344e-06}
{"train_loss": 0.038989607244729996, "global_step": 232989, "epoch": 2617, "lr": 4.889575754897729e-06}
{"train_loss": 0.024724319577217102, "global_step": 232990, "epoch": 2617, "lr": 4.889325717425597e-06}
{"train_loss": 0.026793241500854492, "global_step": 232991, "epoch": 2617, "lr": 4.8890756860180575e-06}
{"train_loss": 0.019549354910850525, "global_step": 232992, "epoch": 2617, "lr": 4.8888256606751175e-06}
{"train_loss": 0.03494120389223099, "global_step": 232993, "epoch": 2617, "lr": 4.888575641396837e-06}
{"train_loss": 0.05969202518463135, "global_step": 232994, "epoch": 2617, "lr": 4.888325628183226e-06}
{"train_loss": 0.06888680905103683, "global_step": 232995, "epoch": 2617, "lr": 4.888075621034344e-06}
{"train_loss": 0.07954001426696777, "global_step": 232996, "epoch": 2617, "lr": 4.887825619950198e-06}
{"train_loss": 0.05723242834210396, "global_step": 232997, "epoch": 2617, "lr": 4.88757562493084e-06}
{"train_loss": 0.09462350606918335, "global_step": 232998, "epoch": 2617, "lr": 4.887325635976292e-06}
{"train_loss": 0.061799056828022, "global_step": 232999, "epoch": 2617, "lr": 4.887075653086604e-06}
{"train_loss": 0.06199498474597931, "global_step": 233000, "epoch": 2617, "lr": 4.886825676261786e-06}
{"train_loss": 0.05083618455388573, "global_step": 233001, "epoch": 2617, "lr": 4.8865757055019e-06, "val_loss": 9.245367050170898}
{"train_loss": 0.026065435260534286, "global_step": 233002, "epoch": 2618, "lr": 4.886325740806946e-06}
{"train_loss": 0.030370071530342102, "global_step": 233003, "epoch": 2618, "lr": 4.886075782176991e-06}
{"train_loss": 0.04891221225261688, "global_step": 233004, "epoch": 2618, "lr": 4.885825829612039e-06}
{"train_loss": 0.032473452389240265, "global_step": 233005, "epoch": 2618, "lr": 4.885575883112142e-06}
{"train_loss": 0.03605335205793381, "global_step": 233006, "epoch": 2618, "lr": 4.885325942677338e-06}
{"train_loss": 0.06644556671380997, "global_step": 233007, "epoch": 2618, "lr": 4.885076008307643e-06}
{"train_loss": 0.024697471410036087, "global_step": 233008, "epoch": 2618, "lr": 4.884826080003108e-06}
{"train_loss": 0.04287300258874893, "global_step": 233009, "epoch": 2618, "lr": 4.884576157763759e-06}
{"train_loss": 0.035413388162851334, "global_step": 233010, "epoch": 2618, "lr": 4.884326241589621e-06}
{"train_loss": 0.07679235935211182, "global_step": 233011, "epoch": 2618, "lr": 4.884076331480741e-06}
{"train_loss": 0.062412649393081665, "global_step": 233012, "epoch": 2618, "lr": 4.8838264274371384e-06}
{"train_loss": 0.03126676380634308, "global_step": 233013, "epoch": 2618, "lr": 4.8835765294588666e-06}
{"train_loss": 0.0356053002178669, "global_step": 233014, "epoch": 2618, "lr": 4.883326637545949e-06}
{"train_loss": 0.05217332765460014, "global_step": 233015, "epoch": 2618, "lr": 4.8830767516984066e-06}
{"train_loss": 0.03620649501681328, "global_step": 233016, "epoch": 2618, "lr": 4.882826871916296e-06}
{"train_loss": 0.06953858584165573, "global_step": 233017, "epoch": 2618, "lr": 4.882576998199628e-06}
{"train_loss": 0.06337494403123856, "global_step": 233018, "epoch": 2618, "lr": 4.882327130548459e-06}
{"train_loss": 0.022538673132658005, "global_step": 233019, "epoch": 2618, "lr": 4.882077268962798e-06}
{"train_loss": 0.03268377110362053, "global_step": 233020, "epoch": 2618, "lr": 4.8818274134427075e-06}
{"train_loss": 0.043473947793245316, "global_step": 233021, "epoch": 2618, "lr": 4.8815775639881935e-06}
{"train_loss": 0.024999653920531273, "global_step": 233022, "epoch": 2618, "lr": 4.8813277205992995e-06}
{"train_loss": 0.02786645106971264, "global_step": 233023, "epoch": 2618, "lr": 4.881077883276069e-06}
{"train_loss": 0.05060109496116638, "global_step": 233024, "epoch": 2618, "lr": 4.880828052018521e-06}
{"train_loss": 0.07495669275522232, "global_step": 233025, "epoch": 2618, "lr": 4.880578226826704e-06}
{"train_loss": 0.06981310993432999, "global_step": 233026, "epoch": 2618, "lr": 4.880328407700635e-06}
{"train_loss": 0.062037393450737, "global_step": 233027, "epoch": 2618, "lr": 4.880078594640364e-06}
{"train_loss": 0.036574989557266235, "global_step": 233028, "epoch": 2618, "lr": 4.879828787645907e-06}
{"train_loss": 0.06044236943125725, "global_step": 233029, "epoch": 2618, "lr": 4.8795789867173205e-06}
{"train_loss": 0.028651157394051552, "global_step": 233030, "epoch": 2618, "lr": 4.879329191854609e-06}
{"train_loss": 0.031522952020168304, "global_step": 233031, "epoch": 2618, "lr": 4.879079403057835e-06}
{"train_loss": 0.0734458938241005, "global_step": 233032, "epoch": 2618, "lr": 4.878829620327013e-06}
{"train_loss": 0.03747159615159035, "global_step": 233033, "epoch": 2618, "lr": 4.878579843662184e-06}
{"train_loss": 0.06983164697885513, "global_step": 233034, "epoch": 2618, "lr": 4.8783300730633755e-06}
{"train_loss": 0.024415047839283943, "global_step": 233035, "epoch": 2618, "lr": 4.878080308530636e-06}
{"train_loss": 0.04505833983421326, "global_step": 233036, "epoch": 2618, "lr": 4.8778305500639774e-06}
{"train_loss": 0.042722444981336594, "global_step": 233037, "epoch": 2618, "lr": 4.877580797663451e-06}
{"train_loss": 0.0348191000521183, "global_step": 233038, "epoch": 2618, "lr": 4.877331051329076e-06}
{"train_loss": 0.05668911710381508, "global_step": 233039, "epoch": 2618, "lr": 4.877081311060905e-06}
{"train_loss": 0.08279698342084885, "global_step": 233040, "epoch": 2618, "lr": 4.876831576858948e-06}
{"train_loss": 0.06646678596735, "global_step": 233041, "epoch": 2618, "lr": 4.8765818487232554e-06}
{"train_loss": 0.04556584358215332, "global_step": 233042, "epoch": 2618, "lr": 4.8763321266538655e-06}
{"train_loss": 0.05341906473040581, "global_step": 233043, "epoch": 2618, "lr": 4.87608241065079e-06}
{"train_loss": 0.03916017338633537, "global_step": 233044, "epoch": 2618, "lr": 4.875832700714089e-06}
{"train_loss": 0.03557993844151497, "global_step": 233045, "epoch": 2618, "lr": 4.8755829968437694e-06}
{"train_loss": 0.038222167640924454, "global_step": 233046, "epoch": 2618, "lr": 4.875333299039886e-06}
{"train_loss": 0.022889157757163048, "global_step": 233047, "epoch": 2618, "lr": 4.875083607302456e-06}
{"train_loss": 0.017208952456712723, "global_step": 233048, "epoch": 2618, "lr": 4.874833921631533e-06}
{"train_loss": 0.03959546983242035, "global_step": 233049, "epoch": 2618, "lr": 4.874584242027136e-06}
{"train_loss": 0.04627954959869385, "global_step": 233050, "epoch": 2618, "lr": 4.874334568489291e-06}
{"train_loss": 0.08653827756643295, "global_step": 233051, "epoch": 2618, "lr": 4.87408490101805e-06}
{"train_loss": 0.03449613228440285, "global_step": 233052, "epoch": 2618, "lr": 4.873835239613433e-06}
{"train_loss": 0.02023867703974247, "global_step": 233053, "epoch": 2618, "lr": 4.873585584275486e-06}
{"train_loss": 0.036409199237823486, "global_step": 233054, "epoch": 2618, "lr": 4.873335935004236e-06}
{"train_loss": 0.04122339189052582, "global_step": 233055, "epoch": 2618, "lr": 4.873086291799705e-06}
{"train_loss": 0.11564315855503082, "global_step": 233056, "epoch": 2618, "lr": 4.87283665466195e-06}
{"train_loss": 0.05240878835320473, "global_step": 233057, "epoch": 2618, "lr": 4.87258702359098e-06}
{"train_loss": 0.056293606758117676, "global_step": 233058, "epoch": 2618, "lr": 4.872337398586852e-06}
{"train_loss": 0.045456305146217346, "global_step": 233059, "epoch": 2618, "lr": 4.872087779649575e-06}
{"train_loss": 0.07331857830286026, "global_step": 233060, "epoch": 2618, "lr": 4.871838166779202e-06}
{"train_loss": 0.03786826878786087, "global_step": 233061, "epoch": 2618, "lr": 4.871588559975765e-06}
{"train_loss": 0.0603938028216362, "global_step": 233062, "epoch": 2618, "lr": 4.871338959239286e-06}
{"train_loss": 0.03126651048660278, "global_step": 233063, "epoch": 2618, "lr": 4.871089364569814e-06}
{"train_loss": 0.06367750465869904, "global_step": 233064, "epoch": 2618, "lr": 4.870839775967362e-06}
{"train_loss": 0.03463272005319595, "global_step": 233065, "epoch": 2618, "lr": 4.87059019343199e-06}
{"train_loss": 0.03709404170513153, "global_step": 233066, "epoch": 2618, "lr": 4.870340616963703e-06}
{"train_loss": 0.08174475282430649, "global_step": 233067, "epoch": 2618, "lr": 4.870091046562558e-06}
{"train_loss": 0.0602978877723217, "global_step": 233068, "epoch": 2618, "lr": 4.869841482228577e-06}
{"train_loss": 0.02307751215994358, "global_step": 233069, "epoch": 2618, "lr": 4.869591923961797e-06}
{"train_loss": 0.02017565444111824, "global_step": 233070, "epoch": 2618, "lr": 4.869342371762242e-06}
{"train_loss": 0.031063392758369446, "global_step": 233071, "epoch": 2618, "lr": 4.8690928256299674e-06}
{"train_loss": 0.04848475381731987, "global_step": 233072, "epoch": 2618, "lr": 4.868843285564984e-06}
{"train_loss": 0.02185792289674282, "global_step": 233073, "epoch": 2618, "lr": 4.868593751567341e-06}
{"train_loss": 0.024504786357283592, "global_step": 233074, "epoch": 2618, "lr": 4.868344223637056e-06}
{"train_loss": 0.06862808018922806, "global_step": 233075, "epoch": 2618, "lr": 4.868094701774178e-06}
{"train_loss": 0.044634632766246796, "global_step": 233076, "epoch": 2618, "lr": 4.8678451859787314e-06}
{"train_loss": 0.05031689628958702, "global_step": 233077, "epoch": 2618, "lr": 4.867595676250752e-06}
{"train_loss": 0.05012967810034752, "global_step": 233078, "epoch": 2618, "lr": 4.867346172590281e-06}
{"train_loss": 0.031092556193470955, "global_step": 233079, "epoch": 2618, "lr": 4.867096674997335e-06}
{"train_loss": 0.041101209819316864, "global_step": 233080, "epoch": 2618, "lr": 4.866847183471973e-06}
{"train_loss": 0.06474539637565613, "global_step": 233081, "epoch": 2618, "lr": 4.866597698014197e-06}
{"train_loss": 0.10196714103221893, "global_step": 233082, "epoch": 2618, "lr": 4.866348218624072e-06}
{"train_loss": 0.02804907225072384, "global_step": 233083, "epoch": 2618, "lr": 4.866098745301606e-06}
{"train_loss": 0.027043480426073074, "global_step": 233084, "epoch": 2618, "lr": 4.865849278046847e-06}
{"train_loss": 0.06882305443286896, "global_step": 233085, "epoch": 2618, "lr": 4.865599816859823e-06}
{"train_loss": 0.05137529596686363, "global_step": 233086, "epoch": 2618, "lr": 4.865350361740573e-06}
{"train_loss": 0.04324515908956528, "global_step": 233087, "epoch": 2618, "lr": 4.865100912689124e-06}
{"train_loss": 0.03175156190991402, "global_step": 233088, "epoch": 2618, "lr": 4.864851469705506e-06}
{"train_loss": 0.029957836493849754, "global_step": 233089, "epoch": 2618, "lr": 4.864602032789772e-06}
{"train_loss": 0.04658902721123749, "global_step": 233090, "epoch": 2618, "lr": 4.864352601941929e-06, "val_loss": 9.268240928649902}
{"train_loss": 0.01468188688158989, "global_step": 233091, "epoch": 2619, "lr": 4.864103177162033e-06}
{"train_loss": 0.05611968785524368, "global_step": 233092, "epoch": 2619, "lr": 4.86385375845011e-06}
{"train_loss": 0.026402346789836884, "global_step": 233093, "epoch": 2619, "lr": 4.8636043458061775e-06}
{"train_loss": 0.04923756793141365, "global_step": 233094, "epoch": 2619, "lr": 4.863354939230297e-06}
{"train_loss": 0.05259257182478905, "global_step": 233095, "epoch": 2619, "lr": 4.863105538722479e-06}
{"train_loss": 0.0625668615102768, "global_step": 233096, "epoch": 2619, "lr": 4.862856144282762e-06}
{"train_loss": 0.043676115572452545, "global_step": 233097, "epoch": 2619, "lr": 4.862606755911197e-06}
{"train_loss": 0.053506530821323395, "global_step": 233098, "epoch": 2619, "lr": 4.862357373607796e-06}
{"train_loss": 0.02317299135029316, "global_step": 233099, "epoch": 2619, "lr": 4.862107997372606e-06}
{"train_loss": 0.059305619448423386, "global_step": 233100, "epoch": 2619, "lr": 4.861858627205651e-06}
{"train_loss": 0.04571952670812607, "global_step": 233101, "epoch": 2619, "lr": 4.861609263106976e-06}
{"train_loss": 0.03548422083258629, "global_step": 233102, "epoch": 2619, "lr": 4.861359905076596e-06}
{"train_loss": 0.09986115247011185, "global_step": 233103, "epoch": 2619, "lr": 4.8611105531145675e-06}
{"train_loss": 0.09471960365772247, "global_step": 233104, "epoch": 2619, "lr": 4.860861207220907e-06}
{"train_loss": 0.03280555456876755, "global_step": 233105, "epoch": 2619, "lr": 4.860611867395659e-06}
{"train_loss": 0.048385363072156906, "global_step": 233106, "epoch": 2619, "lr": 4.86036253363884e-06}
{"train_loss": 0.05860385298728943, "global_step": 233107, "epoch": 2619, "lr": 4.860113205950506e-06}
{"train_loss": 0.02916296012699604, "global_step": 233108, "epoch": 2619, "lr": 4.859863884330673e-06}
{"train_loss": 0.030215755105018616, "global_step": 233109, "epoch": 2619, "lr": 4.859614568779386e-06}
{"train_loss": 0.06206531822681427, "global_step": 233110, "epoch": 2619, "lr": 4.859365259296666e-06}
{"train_loss": 0.030311385169625282, "global_step": 233111, "epoch": 2619, "lr": 4.859115955882565e-06}
{"train_loss": 0.03384377807378769, "global_step": 233112, "epoch": 2619, "lr": 4.858866658537098e-06}
{"train_loss": 0.057914361357688904, "global_step": 233113, "epoch": 2619, "lr": 4.858617367260299e-06}
{"train_loss": 0.0751931294798851, "global_step": 233114, "epoch": 2619, "lr": 4.858368082052223e-06}
{"train_loss": 0.0541500449180603, "global_step": 233115, "epoch": 2619, "lr": 4.858118802912881e-06}
{"train_loss": 0.041840195655822754, "global_step": 233116, "epoch": 2619, "lr": 4.857869529842324e-06}
{"train_loss": 0.07116995751857758, "global_step": 233117, "epoch": 2619, "lr": 4.857620262840562e-06}
{"train_loss": 0.039940331131219864, "global_step": 233118, "epoch": 2619, "lr": 4.857371001907657e-06}
{"train_loss": 0.04569442942738533, "global_step": 233119, "epoch": 2619, "lr": 4.8571217470436195e-06}
{"train_loss": 0.07754410058259964, "global_step": 233120, "epoch": 2619, "lr": 4.856872498248499e-06}
{"train_loss": 0.0759798213839531, "global_step": 233121, "epoch": 2619, "lr": 4.856623255522313e-06}
{"train_loss": 0.060512762516736984, "global_step": 233122, "epoch": 2619, "lr": 4.856374018865112e-06}
{"train_loss": 0.020420456305146217, "global_step": 233123, "epoch": 2619, "lr": 4.856124788276911e-06}
{"train_loss": 0.10907869786024094, "global_step": 233124, "epoch": 2619, "lr": 4.855875563757767e-06}
{"train_loss": 0.04115617275238037, "global_step": 233125, "epoch": 2619, "lr": 4.855626345307696e-06}
{"train_loss": 0.04236273095011711, "global_step": 233126, "epoch": 2619, "lr": 4.855377132926731e-06}
{"train_loss": 0.05588994175195694, "global_step": 233127, "epoch": 2619, "lr": 4.8551279266149165e-06}
{"train_loss": 0.07384929060935974, "global_step": 233128, "epoch": 2619, "lr": 4.854878726372269e-06}
{"train_loss": 0.051408831030130386, "global_step": 233129, "epoch": 2619, "lr": 4.854629532198845e-06}
{"train_loss": 0.025982262566685677, "global_step": 233130, "epoch": 2619, "lr": 4.854380344094655e-06}
{"train_loss": 0.0614733025431633, "global_step": 233131, "epoch": 2619, "lr": 4.854131162059755e-06}
{"train_loss": 0.03651905432343483, "global_step": 233132, "epoch": 2619, "lr": 4.853881986094155e-06}
{"train_loss": 0.08189376443624496, "global_step": 233133, "epoch": 2619, "lr": 4.853632816197912e-06}
{"train_loss": 0.04740804061293602, "global_step": 233134, "epoch": 2619, "lr": 4.8533836523710356e-06}
{"train_loss": 0.03044067695736885, "global_step": 233135, "epoch": 2619, "lr": 4.8531344946135815e-06}
{"train_loss": 0.05534658953547478, "global_step": 233136, "epoch": 2619, "lr": 4.8528853429255674e-06}
{"train_loss": 0.050746794790029526, "global_step": 233137, "epoch": 2619, "lr": 4.8526361973070366e-06}
{"train_loss": 0.04708794876933098, "global_step": 233138, "epoch": 2619, "lr": 4.852387057758012e-06}
{"train_loss": 0.09977735579013824, "global_step": 233139, "epoch": 2619, "lr": 4.852137924278544e-06}
{"train_loss": 0.0634787380695343, "global_step": 233140, "epoch": 2619, "lr": 4.851888796868648e-06}
{"train_loss": 0.041411738842725754, "global_step": 233141, "epoch": 2619, "lr": 4.8516396755283684e-06}
{"train_loss": 0.06160666048526764, "global_step": 233142, "epoch": 2619, "lr": 4.8513905602577296e-06}
{"train_loss": 0.06503252685070038, "global_step": 233143, "epoch": 2619, "lr": 4.851141451056784e-06}
{"train_loss": 0.042127810418605804, "global_step": 233144, "epoch": 2619, "lr": 4.85089234792554e-06}
{"train_loss": 0.06528352946043015, "global_step": 233145, "epoch": 2619, "lr": 4.850643250864051e-06}
{"train_loss": 0.02661547251045704, "global_step": 233146, "epoch": 2619, "lr": 4.8503941598723345e-06}
{"train_loss": 0.04165273532271385, "global_step": 233147, "epoch": 2619, "lr": 4.850145074950441e-06}
{"train_loss": 0.058277953416109085, "global_step": 233148, "epoch": 2619, "lr": 4.849895996098386e-06}
{"train_loss": 0.052072662860155106, "global_step": 233149, "epoch": 2619, "lr": 4.849646923316209e-06}
{"train_loss": 0.03479619324207306, "global_step": 233150, "epoch": 2619, "lr": 4.84939785660396e-06}
{"train_loss": 0.06125235557556152, "global_step": 233151, "epoch": 2619, "lr": 4.84914879596165e-06}
{"train_loss": 0.037478215992450714, "global_step": 233152, "epoch": 2619, "lr": 4.848899741389329e-06}
{"train_loss": 0.031008785590529442, "global_step": 233153, "epoch": 2619, "lr": 4.848650692887019e-06}
{"train_loss": 0.058227475732564926, "global_step": 233154, "epoch": 2619, "lr": 4.848401650454759e-06}
{"train_loss": 0.13661538064479828, "global_step": 233155, "epoch": 2619, "lr": 4.848152614092577e-06}
{"train_loss": 0.043569520115852356, "global_step": 233156, "epoch": 2619, "lr": 4.847903583800517e-06}
{"train_loss": 0.07445208728313446, "global_step": 233157, "epoch": 2619, "lr": 4.847654559578602e-06}
{"train_loss": 0.09902548044919968, "global_step": 233158, "epoch": 2619, "lr": 4.847405541426875e-06}
{"train_loss": 0.05609387531876564, "global_step": 233159, "epoch": 2619, "lr": 4.847156529345353e-06}
{"train_loss": 0.03611219301819801, "global_step": 233160, "epoch": 2619, "lr": 4.846907523334093e-06}
{"train_loss": 0.051777251064777374, "global_step": 233161, "epoch": 2619, "lr": 4.846658523393105e-06}
{"train_loss": 0.07029060274362564, "global_step": 233162, "epoch": 2619, "lr": 4.846409529522444e-06}
{"train_loss": 0.07566934823989868, "global_step": 233163, "epoch": 2619, "lr": 4.846160541722122e-06}
{"train_loss": 0.04733049124479294, "global_step": 233164, "epoch": 2619, "lr": 4.845911559992189e-06}
{"train_loss": 0.08517836779356003, "global_step": 233165, "epoch": 2619, "lr": 4.845662584332677e-06}
{"train_loss": 0.09624919295310974, "global_step": 233166, "epoch": 2619, "lr": 4.84541361474361e-06}
{"train_loss": 0.030247393995523453, "global_step": 233167, "epoch": 2619, "lr": 4.845164651225032e-06}
{"train_loss": 0.06988564133644104, "global_step": 233168, "epoch": 2619, "lr": 4.844915693776958e-06}
{"train_loss": 0.04958769306540489, "global_step": 233169, "epoch": 2619, "lr": 4.844666742399451e-06}
{"train_loss": 0.05533188581466675, "global_step": 233170, "epoch": 2619, "lr": 4.844417797092515e-06}
{"train_loss": 0.05488009378314018, "global_step": 233171, "epoch": 2619, "lr": 4.844168857856207e-06}
{"train_loss": 0.06891943514347076, "global_step": 233172, "epoch": 2619, "lr": 4.843919924690543e-06}
{"train_loss": 0.019460070878267288, "global_step": 233173, "epoch": 2619, "lr": 4.843670997595567e-06}
{"train_loss": 0.09320304542779922, "global_step": 233174, "epoch": 2619, "lr": 4.843422076571303e-06}
{"train_loss": 0.04214245826005936, "global_step": 233175, "epoch": 2619, "lr": 4.843173161617803e-06}
{"train_loss": 0.016835523769259453, "global_step": 233176, "epoch": 2619, "lr": 4.842924252735076e-06}
{"train_loss": 0.08227001875638962, "global_step": 233177, "epoch": 2619, "lr": 4.842675349923176e-06}
{"train_loss": 0.0308238435536623, "global_step": 233178, "epoch": 2619, "lr": 4.8424264531821194e-06}
{"train_loss": 0.05417955725380544, "global_step": 233179, "epoch": 2619, "lr": 4.8421775625119566e-06, "val_loss": 9.263692855834961}
{"train_loss": 0.05161746218800545, "global_step": 233180, "epoch": 2620, "lr": 4.841928677912705e-06}
{"train_loss": 0.06836500018835068, "global_step": 233181, "epoch": 2620, "lr": 4.841679799384413e-06}
{"train_loss": 0.03565439209342003, "global_step": 233182, "epoch": 2620, "lr": 4.841430926927098e-06}
{"train_loss": 0.052837882190942764, "global_step": 233183, "epoch": 2620, "lr": 4.841182060540811e-06}
{"train_loss": 0.041592493653297424, "global_step": 233184, "epoch": 2620, "lr": 4.840933200225572e-06}
{"train_loss": 0.08378501981496811, "global_step": 233185, "epoch": 2620, "lr": 4.840684345981422e-06}
{"train_loss": 0.04274579882621765, "global_step": 233186, "epoch": 2620, "lr": 4.840435497808388e-06}
{"train_loss": 0.04720153287053108, "global_step": 233187, "epoch": 2620, "lr": 4.840186655706502e-06}
{"train_loss": 0.02703799493610859, "global_step": 233188, "epoch": 2620, "lr": 4.839937819675816e-06}
{"train_loss": 0.038401126861572266, "global_step": 233189, "epoch": 2620, "lr": 4.839688989716334e-06}
{"train_loss": 0.03350459784269333, "global_step": 233190, "epoch": 2620, "lr": 4.839440165828118e-06}
{"train_loss": 0.06071018800139427, "global_step": 233191, "epoch": 2620, "lr": 4.839191348011185e-06}
{"train_loss": 0.07211107760667801, "global_step": 233192, "epoch": 2620, "lr": 4.838942536265578e-06}
{"train_loss": 0.036106690764427185, "global_step": 233193, "epoch": 2620, "lr": 4.838693730591315e-06}
{"train_loss": 0.03917525336146355, "global_step": 233194, "epoch": 2620, "lr": 4.838444930988445e-06}
{"train_loss": 0.07061733305454254, "global_step": 233195, "epoch": 2620, "lr": 4.838196137456991e-06}
{"train_loss": 0.05779782682657242, "global_step": 233196, "epoch": 2620, "lr": 4.837947349996996e-06}
{"train_loss": 0.05065581202507019, "global_step": 233197, "epoch": 2620, "lr": 4.837698568608484e-06}
{"train_loss": 0.033075328916311264, "global_step": 233198, "epoch": 2620, "lr": 4.837449793291499e-06}
{"train_loss": 0.03274521231651306, "global_step": 233199, "epoch": 2620, "lr": 4.837201024046062e-06}
{"train_loss": 0.11068616807460785, "global_step": 233200, "epoch": 2620, "lr": 4.836952260872224e-06}
{"train_loss": 0.04502476379275322, "global_step": 233201, "epoch": 2620, "lr": 4.8367035037699905e-06}
{"train_loss": 0.04321436211466789, "global_step": 233202, "epoch": 2620, "lr": 4.836454752739428e-06}
{"train_loss": 0.028150634840130806, "global_step": 233203, "epoch": 2620, "lr": 4.8362060077805526e-06}
{"train_loss": 0.024153366684913635, "global_step": 233204, "epoch": 2620, "lr": 4.835957268893387e-06}
{"train_loss": 0.037853874266147614, "global_step": 233205, "epoch": 2620, "lr": 4.835708536077988e-06}
{"train_loss": 0.035134971141815186, "global_step": 233206, "epoch": 2620, "lr": 4.835459809334364e-06}
{"train_loss": 0.024663174524903297, "global_step": 233207, "epoch": 2620, "lr": 4.835211088662572e-06}
{"train_loss": 0.04662594571709633, "global_step": 233208, "epoch": 2620, "lr": 4.834962374062629e-06}
{"train_loss": 0.02073335088789463, "global_step": 233209, "epoch": 2620, "lr": 4.834713665534585e-06}
{"train_loss": 0.02780032344162464, "global_step": 233210, "epoch": 2620, "lr": 4.83446496307845e-06}
{"train_loss": 0.07921620458364487, "global_step": 233211, "epoch": 2620, "lr": 4.8342162666942855e-06}
{"train_loss": 0.025912992656230927, "global_step": 233212, "epoch": 2620, "lr": 4.833967576382098e-06}
{"train_loss": 0.08700568228960037, "global_step": 233213, "epoch": 2620, "lr": 4.833718892141942e-06}
{"train_loss": 0.0815608948469162, "global_step": 233214, "epoch": 2620, "lr": 4.8334702139738285e-06}
{"train_loss": 0.03366273641586304, "global_step": 233215, "epoch": 2620, "lr": 4.833221541877819e-06}
{"train_loss": 0.06611013412475586, "global_step": 233216, "epoch": 2620, "lr": 4.832972875853925e-06}
{"train_loss": 0.05273095518350601, "global_step": 233217, "epoch": 2620, "lr": 4.83272421590219e-06}
{"train_loss": 0.04401489347219467, "global_step": 233218, "epoch": 2620, "lr": 4.832475562022643e-06}
{"train_loss": 0.050826799124479294, "global_step": 233219, "epoch": 2620, "lr": 4.832226914215321e-06}
{"train_loss": 0.021077493205666542, "global_step": 233220, "epoch": 2620, "lr": 4.831978272480253e-06}
{"train_loss": 0.06830234080553055, "global_step": 233221, "epoch": 2620, "lr": 4.831729636817478e-06}
{"train_loss": 0.09580916911363602, "global_step": 233222, "epoch": 2620, "lr": 4.831481007227023e-06}
{"train_loss": 0.03981582075357437, "global_step": 233223, "epoch": 2620, "lr": 4.831232383708922e-06}
{"train_loss": 0.06309754401445389, "global_step": 233224, "epoch": 2620, "lr": 4.8309837662632195e-06}
{"train_loss": 0.018710702657699585, "global_step": 233225, "epoch": 2620, "lr": 4.830735154889931e-06}
{"train_loss": 0.02037489041686058, "global_step": 233226, "epoch": 2620, "lr": 4.830486549589114e-06}
{"train_loss": 0.07556423544883728, "global_step": 233227, "epoch": 2620, "lr": 4.830237950360772e-06}
{"train_loss": 0.06658095121383667, "global_step": 233228, "epoch": 2620, "lr": 4.829989357204967e-06}
{"train_loss": 0.04537340998649597, "global_step": 233229, "epoch": 2620, "lr": 4.82974077012171e-06}
{"train_loss": 0.0368247851729393, "global_step": 233230, "epoch": 2620, "lr": 4.8294921891110516e-06}
{"train_loss": 0.01959337666630745, "global_step": 233231, "epoch": 2620, "lr": 4.829243614173007e-06}
{"train_loss": 0.04070641100406647, "global_step": 233232, "epoch": 2620, "lr": 4.828995045307633e-06}
{"train_loss": 0.028316175565123558, "global_step": 233233, "epoch": 2620, "lr": 4.828746482514934e-06}
{"train_loss": 0.026224510744214058, "global_step": 233234, "epoch": 2620, "lr": 4.828497925794978e-06}
{"train_loss": 0.1041967049241066, "global_step": 233235, "epoch": 2620, "lr": 4.828249375147764e-06}
{"train_loss": 0.026442712172865868, "global_step": 233236, "epoch": 2620, "lr": 4.828000830573354e-06}
{"train_loss": 0.03983603045344353, "global_step": 233237, "epoch": 2620, "lr": 4.827752292071758e-06}
{"train_loss": 0.04085961729288101, "global_step": 233238, "epoch": 2620, "lr": 4.8275037596430265e-06}
{"train_loss": 0.02477027103304863, "global_step": 233239, "epoch": 2620, "lr": 4.8272552332871815e-06}
{"train_loss": 0.020715627819299698, "global_step": 233240, "epoch": 2620, "lr": 4.827006713004273e-06}
{"train_loss": 0.07025374472141266, "global_step": 233241, "epoch": 2620, "lr": 4.826758198794318e-06}
{"train_loss": 0.055916789919137955, "global_step": 233242, "epoch": 2620, "lr": 4.826509690657344e-06}
{"train_loss": 0.06421753764152527, "global_step": 233243, "epoch": 2620, "lr": 4.826261188593412e-06}
{"train_loss": 0.04876777157187462, "global_step": 233244, "epoch": 2620, "lr": 4.8260126926025216e-06}
{"train_loss": 0.013276771642267704, "global_step": 233245, "epoch": 2620, "lr": 4.825764202684735e-06}
{"train_loss": 0.07206900417804718, "global_step": 233246, "epoch": 2620, "lr": 4.825515718840068e-06}
{"train_loss": 0.043236538767814636, "global_step": 233247, "epoch": 2620, "lr": 4.825267241068565e-06}
{"train_loss": 0.07746361941099167, "global_step": 233248, "epoch": 2620, "lr": 4.825018769370243e-06}
{"train_loss": 0.04670901596546173, "global_step": 233249, "epoch": 2620, "lr": 4.824770303745163e-06}
{"train_loss": 0.08432795852422714, "global_step": 233250, "epoch": 2620, "lr": 4.824521844193325e-06}
{"train_loss": 0.06524713337421417, "global_step": 233251, "epoch": 2620, "lr": 4.824273390714795e-06}
{"train_loss": 0.02670321613550186, "global_step": 233252, "epoch": 2620, "lr": 4.82402494330958e-06}
{"train_loss": 0.09588629007339478, "global_step": 233253, "epoch": 2620, "lr": 4.823776501977728e-06}
{"train_loss": 0.055440742522478104, "global_step": 233254, "epoch": 2620, "lr": 4.823528066719269e-06}
{"train_loss": 0.07388053834438324, "global_step": 233255, "epoch": 2620, "lr": 4.8232796375342395e-06}
{"train_loss": 0.06711538136005402, "global_step": 233256, "epoch": 2620, "lr": 4.823031214422658e-06}
{"train_loss": 0.04272211343050003, "global_step": 233257, "epoch": 2620, "lr": 4.822782797384584e-06}
{"train_loss": 0.07584917545318604, "global_step": 233258, "epoch": 2620, "lr": 4.822534386420024e-06}
{"train_loss": 0.03123110719025135, "global_step": 233259, "epoch": 2620, "lr": 4.822285981529023e-06}
{"train_loss": 0.05130964517593384, "global_step": 233260, "epoch": 2620, "lr": 4.8220375827116245e-06}
{"train_loss": 0.016508229076862335, "global_step": 233261, "epoch": 2620, "lr": 4.821789189967846e-06}
{"train_loss": 0.015402746386826038, "global_step": 233262, "epoch": 2620, "lr": 4.821540803297731e-06}
{"train_loss": 0.046371787786483765, "global_step": 233263, "epoch": 2620, "lr": 4.821292422701307e-06}
{"train_loss": 0.05400886386632919, "global_step": 233264, "epoch": 2620, "lr": 4.821044048178614e-06}
{"train_loss": 0.04864618182182312, "global_step": 233265, "epoch": 2620, "lr": 4.820795679729673e-06}
{"train_loss": 0.02823001518845558, "global_step": 233266, "epoch": 2620, "lr": 4.820547317354534e-06}
{"train_loss": 0.05610431730747223, "global_step": 233267, "epoch": 2620, "lr": 4.820298961053216e-06}
{"train_loss": 0.048643627005179275, "global_step": 233268, "epoch": 2620, "lr": 4.8200506108257606e-06, "val_loss": 9.256757736206055, "train_action_mse_error": 7.138984680175781}
{"train_loss": 0.0764206200838089, "global_step": 233269, "epoch": 2621, "lr": 4.819802266672196e-06}
{"train_loss": 0.023318171501159668, "global_step": 233270, "epoch": 2621, "lr": 4.819553928592563e-06}
{"train_loss": 0.01535400003194809, "global_step": 233271, "epoch": 2621, "lr": 4.819305596586881e-06}
{"train_loss": 0.041398998349905014, "global_step": 233272, "epoch": 2621, "lr": 4.819057270655208e-06}
{"train_loss": 0.05360996723175049, "global_step": 233273, "epoch": 2621, "lr": 4.8188089507975484e-06}
{"train_loss": 0.05145145580172539, "global_step": 233274, "epoch": 2621, "lr": 4.8185606370139576e-06}
{"train_loss": 0.048490334302186966, "global_step": 233275, "epoch": 2621, "lr": 4.818312329304453e-06}
{"train_loss": 0.02217094786465168, "global_step": 233276, "epoch": 2621, "lr": 4.818064027669084e-06}
{"train_loss": 0.07856883853673935, "global_step": 233277, "epoch": 2621, "lr": 4.817815732107867e-06}
{"train_loss": 0.08775989711284637, "global_step": 233278, "epoch": 2621, "lr": 4.817567442620852e-06}
{"train_loss": 0.02340984158217907, "global_step": 233279, "epoch": 2621, "lr": 4.817319159208056e-06}
{"train_loss": 0.04803824797272682, "global_step": 233280, "epoch": 2621, "lr": 4.817070881869529e-06}
{"train_loss": 0.05408501625061035, "global_step": 233281, "epoch": 2621, "lr": 4.816822610605298e-06}
{"train_loss": 0.023803401738405228, "global_step": 233282, "epoch": 2621, "lr": 4.816574345415387e-06}
{"train_loss": 0.06020249053835869, "global_step": 233283, "epoch": 2621, "lr": 4.816326086299844e-06}
{"train_loss": 0.07365911453962326, "global_step": 233284, "epoch": 2621, "lr": 4.816077833258681e-06}
{"train_loss": 0.04105336219072342, "global_step": 233285, "epoch": 2621, "lr": 4.815829586291965e-06}
{"train_loss": 0.032274432480335236, "global_step": 233286, "epoch": 2621, "lr": 4.815581345399694e-06}
{"train_loss": 0.044064171612262726, "global_step": 233287, "epoch": 2621, "lr": 4.815333110581926e-06}
{"train_loss": 0.04497133567929268, "global_step": 233288, "epoch": 2621, "lr": 4.815084881838677e-06}
{"train_loss": 0.04285335913300514, "global_step": 233289, "epoch": 2621, "lr": 4.814836659170002e-06}
{"train_loss": 0.03932574391365051, "global_step": 233290, "epoch": 2621, "lr": 4.814588442575907e-06}
{"train_loss": 0.041998736560344696, "global_step": 233291, "epoch": 2621, "lr": 4.814340232056452e-06}
{"train_loss": 0.031191494315862656, "global_step": 233292, "epoch": 2621, "lr": 4.8140920276116495e-06}
{"train_loss": 0.06302936375141144, "global_step": 233293, "epoch": 2621, "lr": 4.813843829241554e-06}
{"train_loss": 0.03154497221112251, "global_step": 233294, "epoch": 2621, "lr": 4.813595636946172e-06}
{"train_loss": 0.047647614032030106, "global_step": 233295, "epoch": 2621, "lr": 4.8133474507255524e-06}
{"train_loss": 0.05231814831495285, "global_step": 233296, "epoch": 2621, "lr": 4.813099270579735e-06}
{"train_loss": 0.029040493071079254, "global_step": 233297, "epoch": 2621, "lr": 4.812851096508741e-06}
{"train_loss": 0.027365904301404953, "global_step": 233298, "epoch": 2621, "lr": 4.812602928512616e-06}
{"train_loss": 0.026349207386374474, "global_step": 233299, "epoch": 2621, "lr": 4.812354766591376e-06}
{"train_loss": 0.06550106406211853, "global_step": 233300, "epoch": 2621, "lr": 4.812106610745071e-06}
{"train_loss": 0.036232415586709976, "global_step": 233301, "epoch": 2621, "lr": 4.811858460973723e-06}
{"train_loss": 0.052602995187044144, "global_step": 233302, "epoch": 2621, "lr": 4.811610317277376e-06}
{"train_loss": 0.06770050525665283, "global_step": 233303, "epoch": 2621, "lr": 4.8113621796560485e-06}
{"train_loss": 0.025247685611248016, "global_step": 233304, "epoch": 2621, "lr": 4.811114048109794e-06}
{"train_loss": 0.0353245884180069, "global_step": 233305, "epoch": 2621, "lr": 4.810865922638619e-06}
{"train_loss": 0.06608007848262787, "global_step": 233306, "epoch": 2621, "lr": 4.810617803242584e-06}
{"train_loss": 0.012471705675125122, "global_step": 233307, "epoch": 2621, "lr": 4.810369689921707e-06}
{"train_loss": 0.04041546955704689, "global_step": 233308, "epoch": 2621, "lr": 4.81012158267603e-06}
{"train_loss": 0.03329817205667496, "global_step": 233309, "epoch": 2621, "lr": 4.809873481505572e-06}
{"train_loss": 0.03969700261950493, "global_step": 233310, "epoch": 2621, "lr": 4.8096253864103824e-06}
{"train_loss": 0.026293322443962097, "global_step": 233311, "epoch": 2621, "lr": 4.809377297390482e-06}
{"train_loss": 0.05463170260190964, "global_step": 233312, "epoch": 2621, "lr": 4.809129214445924e-06}
{"train_loss": 0.027284523472189903, "global_step": 233313, "epoch": 2621, "lr": 4.8088811375767095e-06}
{"train_loss": 0.04317580908536911, "global_step": 233314, "epoch": 2621, "lr": 4.808633066782903e-06}
{"train_loss": 0.025948720052838326, "global_step": 233315, "epoch": 2621, "lr": 4.808385002064519e-06}
{"train_loss": 0.033273935317993164, "global_step": 233316, "epoch": 2621, "lr": 4.808136943421598e-06}
{"train_loss": 0.07610327750444412, "global_step": 233317, "epoch": 2621, "lr": 4.807888890854173e-06}
{"train_loss": 0.03359227627515793, "global_step": 233318, "epoch": 2621, "lr": 4.807640844362277e-06}
{"train_loss": 0.02236591838300228, "global_step": 233319, "epoch": 2621, "lr": 4.807392803945948e-06}
{"train_loss": 0.015391969121992588, "global_step": 233320, "epoch": 2621, "lr": 4.807144769605204e-06}
{"train_loss": 0.0835503339767456, "global_step": 233321, "epoch": 2621, "lr": 4.8068967413401e-06}
{"train_loss": 0.04947651922702789, "global_step": 233322, "epoch": 2621, "lr": 4.8066487191506415e-06}
{"train_loss": 0.032078567892313004, "global_step": 233323, "epoch": 2621, "lr": 4.806400703036895e-06}
{"train_loss": 0.02940935268998146, "global_step": 233324, "epoch": 2621, "lr": 4.806152692998867e-06}
{"train_loss": 0.05260314792394638, "global_step": 233325, "epoch": 2621, "lr": 4.805904689036606e-06}
{"train_loss": 0.042299602180719376, "global_step": 233326, "epoch": 2621, "lr": 4.805656691150134e-06}
{"train_loss": 0.06671051681041718, "global_step": 233327, "epoch": 2621, "lr": 4.805408699339498e-06}
{"train_loss": 0.03619292378425598, "global_step": 233328, "epoch": 2621, "lr": 4.8051607136047174e-06}
{"train_loss": 0.045749638229608536, "global_step": 233329, "epoch": 2621, "lr": 4.8049127339458375e-06}
{"train_loss": 0.07424946874380112, "global_step": 233330, "epoch": 2621, "lr": 4.804664760362882e-06}
{"train_loss": 0.03375440835952759, "global_step": 233331, "epoch": 2621, "lr": 4.804416792855881e-06}
{"train_loss": 0.06482265144586563, "global_step": 233332, "epoch": 2621, "lr": 4.804168831424888e-06}
{"train_loss": 0.045953962951898575, "global_step": 233333, "epoch": 2621, "lr": 4.803920876069917e-06}
{"train_loss": 0.030349021777510643, "global_step": 233334, "epoch": 2621, "lr": 4.80367292679102e-06}
{"train_loss": 0.060377683490514755, "global_step": 233335, "epoch": 2621, "lr": 4.803424983588201e-06}
{"train_loss": 0.02317952737212181, "global_step": 233336, "epoch": 2621, "lr": 4.803177046461527e-06}
{"train_loss": 0.05809894949197769, "global_step": 233337, "epoch": 2621, "lr": 4.802929115410998e-06}
{"train_loss": 0.03316862881183624, "global_step": 233338, "epoch": 2621, "lr": 4.8026811904366806e-06}
{"train_loss": 0.021934011951088905, "global_step": 233339, "epoch": 2621, "lr": 4.802433271538581e-06}
{"train_loss": 0.07528171688318253, "global_step": 233340, "epoch": 2621, "lr": 4.802185358716749e-06}
{"train_loss": 0.05253063887357712, "global_step": 233341, "epoch": 2621, "lr": 4.801937451971206e-06}
{"train_loss": 0.04526010528206825, "global_step": 233342, "epoch": 2621, "lr": 4.801689551302002e-06}
{"train_loss": 0.007975892163813114, "global_step": 233343, "epoch": 2621, "lr": 4.801441656709149e-06}
{"train_loss": 0.07408393919467926, "global_step": 233344, "epoch": 2621, "lr": 4.801193768192702e-06}
{"train_loss": 0.06369265168905258, "global_step": 233345, "epoch": 2621, "lr": 4.800945885752672e-06}
{"train_loss": 0.03008466772735119, "global_step": 233346, "epoch": 2621, "lr": 4.800698009389115e-06}
{"train_loss": 0.053701769560575485, "global_step": 233347, "epoch": 2621, "lr": 4.800450139102042e-06}
{"train_loss": 0.07950260490179062, "global_step": 233348, "epoch": 2621, "lr": 4.800202274891508e-06}
{"train_loss": 0.045170288532972336, "global_step": 233349, "epoch": 2621, "lr": 4.799954416757524e-06}
{"train_loss": 0.020444020628929138, "global_step": 233350, "epoch": 2621, "lr": 4.79970656470014e-06}
{"train_loss": 0.07021045684814453, "global_step": 233351, "epoch": 2621, "lr": 4.799458718719391e-06}
{"train_loss": 0.10795804113149643, "global_step": 233352, "epoch": 2621, "lr": 4.799210878815308e-06}
{"train_loss": 0.039655864238739014, "global_step": 233353, "epoch": 2621, "lr": 4.798963044987903e-06}
{"train_loss": 0.05126625671982765, "global_step": 233354, "epoch": 2621, "lr": 4.798715217237243e-06}
{"train_loss": 0.06610246002674103, "global_step": 233355, "epoch": 2621, "lr": 4.798467395563333e-06}
{"train_loss": 0.07584919035434723, "global_step": 233356, "epoch": 2621, "lr": 4.798219579966229e-06}
{"train_loss": 0.0459725046425723, "global_step": 233357, "epoch": 2621, "lr": 4.797971770445947e-06, "val_loss": 9.283768653869629}
{"train_loss": 0.07157176733016968, "global_step": 233358, "epoch": 2622, "lr": 4.797723967002527e-06}
{"train_loss": 0.038252733647823334, "global_step": 233359, "epoch": 2622, "lr": 4.797476169636006e-06}
{"train_loss": 0.03272810950875282, "global_step": 233360, "epoch": 2622, "lr": 4.797228378346402e-06}
{"train_loss": 0.022942209616303444, "global_step": 233361, "epoch": 2622, "lr": 4.79698059313377e-06}
{"train_loss": 0.04764295741915703, "global_step": 233362, "epoch": 2622, "lr": 4.796732813998128e-06}
{"train_loss": 0.06495451927185059, "global_step": 233363, "epoch": 2622, "lr": 4.796485040939519e-06}
{"train_loss": 0.02700822427868843, "global_step": 233364, "epoch": 2622, "lr": 4.796237273957965e-06}
{"train_loss": 0.04127814620733261, "global_step": 233365, "epoch": 2622, "lr": 4.795989513053517e-06}
{"train_loss": 0.024490199983119965, "global_step": 233366, "epoch": 2622, "lr": 4.7957417582261855e-06}
{"train_loss": 0.06703347712755203, "global_step": 233367, "epoch": 2622, "lr": 4.7954940094760146e-06}
{"train_loss": 0.02758483774960041, "global_step": 233368, "epoch": 2622, "lr": 4.795246266803049e-06}
{"train_loss": 0.04544684663414955, "global_step": 233369, "epoch": 2622, "lr": 4.7949985302073005e-06}
{"train_loss": 0.05916180834174156, "global_step": 233370, "epoch": 2622, "lr": 4.794750799688824e-06}
{"train_loss": 0.02949858084321022, "global_step": 233371, "epoch": 2622, "lr": 4.794503075247636e-06}
{"train_loss": 0.06173224374651909, "global_step": 233372, "epoch": 2622, "lr": 4.794255356883781e-06}
{"train_loss": 0.01711413823068142, "global_step": 233373, "epoch": 2622, "lr": 4.7940076445972814e-06}
{"train_loss": 0.054712798446416855, "global_step": 233374, "epoch": 2622, "lr": 4.7937599383881815e-06}
{"train_loss": 0.06478367000818253, "global_step": 233375, "epoch": 2622, "lr": 4.793512238256503e-06}
{"train_loss": 0.0920056477189064, "global_step": 233376, "epoch": 2622, "lr": 4.793264544202297e-06}
{"train_loss": 0.034585047513246536, "global_step": 233377, "epoch": 2622, "lr": 4.793016856225574e-06}
{"train_loss": 0.04499368742108345, "global_step": 233378, "epoch": 2622, "lr": 4.792769174326389e-06}
{"train_loss": 0.03369782492518425, "global_step": 233379, "epoch": 2622, "lr": 4.792521498504754e-06}
{"train_loss": 0.04052463918924332, "global_step": 233380, "epoch": 2622, "lr": 4.792273828760724e-06}
{"train_loss": 0.05576912686228752, "global_step": 233381, "epoch": 2622, "lr": 4.792026165094315e-06}
{"train_loss": 0.02264283411204815, "global_step": 233382, "epoch": 2622, "lr": 4.791778507505573e-06}
{"train_loss": 0.029600292444229126, "global_step": 233383, "epoch": 2622, "lr": 4.791530855994514e-06}
{"train_loss": 0.04540224000811577, "global_step": 233384, "epoch": 2622, "lr": 4.791283210561198e-06}
{"train_loss": 0.048031650483608246, "global_step": 233385, "epoch": 2622, "lr": 4.791035571205632e-06}
{"train_loss": 0.03783796727657318, "global_step": 233386, "epoch": 2622, "lr": 4.790787937927854e-06}
{"train_loss": 0.07730194926261902, "global_step": 233387, "epoch": 2622, "lr": 4.79054031072792e-06}
{"train_loss": 0.02869819849729538, "global_step": 233388, "epoch": 2622, "lr": 4.790292689605835e-06}
{"train_loss": 0.04177780821919441, "global_step": 233389, "epoch": 2622, "lr": 4.790045074561655e-06}
{"train_loss": 0.07687637209892273, "global_step": 233390, "epoch": 2622, "lr": 4.789797465595391e-06}
{"train_loss": 0.07584784179925919, "global_step": 233391, "epoch": 2622, "lr": 4.789549862707099e-06}
{"train_loss": 0.04262952134013176, "global_step": 233392, "epoch": 2622, "lr": 4.7893022658968005e-06}
{"train_loss": 0.04129474610090256, "global_step": 233393, "epoch": 2622, "lr": 4.789054675164517e-06}
{"train_loss": 0.04312394931912422, "global_step": 233394, "epoch": 2622, "lr": 4.788807090510306e-06}
{"train_loss": 0.03414079546928406, "global_step": 233395, "epoch": 2622, "lr": 4.788559511934182e-06}
{"train_loss": 0.05185917019844055, "global_step": 233396, "epoch": 2622, "lr": 4.788311939436191e-06}
{"train_loss": 0.05532830208539963, "global_step": 233397, "epoch": 2622, "lr": 4.78806437301636e-06}
{"train_loss": 0.03175793215632439, "global_step": 233398, "epoch": 2622, "lr": 4.7878168126747165e-06}
{"train_loss": 0.021442437544465065, "global_step": 233399, "epoch": 2622, "lr": 4.787569258411306e-06}
{"train_loss": 0.03320254012942314, "global_step": 233400, "epoch": 2622, "lr": 4.787321710226144e-06}
{"train_loss": 0.026981176808476448, "global_step": 233401, "epoch": 2622, "lr": 4.787074168119293e-06}
{"train_loss": 0.040883831679821014, "global_step": 233402, "epoch": 2622, "lr": 4.786826632090752e-06}
{"train_loss": 0.22539636492729187, "global_step": 233403, "epoch": 2622, "lr": 4.786579102140576e-06}
{"train_loss": 0.053733762353658676, "global_step": 233404, "epoch": 2622, "lr": 4.786331578268799e-06}
{"train_loss": 0.06112558767199516, "global_step": 233405, "epoch": 2622, "lr": 4.786084060475443e-06}
{"train_loss": 0.03829524293541908, "global_step": 233406, "epoch": 2622, "lr": 4.785836548760553e-06}
{"train_loss": 0.06885866820812225, "global_step": 233407, "epoch": 2622, "lr": 4.78558904312415e-06}
{"train_loss": 0.05227947607636452, "global_step": 233408, "epoch": 2622, "lr": 4.78534154356628e-06}
{"train_loss": 0.09351645410060883, "global_step": 233409, "epoch": 2622, "lr": 4.7850940500869636e-06}
{"train_loss": 0.030723698437213898, "global_step": 233410, "epoch": 2622, "lr": 4.784846562686246e-06}
{"train_loss": 0.032983310520648956, "global_step": 233411, "epoch": 2622, "lr": 4.784599081364144e-06}
{"train_loss": 0.06208399310708046, "global_step": 233412, "epoch": 2622, "lr": 4.784351606120713e-06}
{"train_loss": 0.060870520770549774, "global_step": 233413, "epoch": 2622, "lr": 4.784104136955969e-06}
{"train_loss": 0.061802007257938385, "global_step": 233414, "epoch": 2622, "lr": 4.783856673869958e-06}
{"train_loss": 0.07484076172113419, "global_step": 233415, "epoch": 2622, "lr": 4.783609216862694e-06}
{"train_loss": 0.07148054242134094, "global_step": 233416, "epoch": 2622, "lr": 4.783361765934236e-06}
{"train_loss": 0.05437867343425751, "global_step": 233417, "epoch": 2622, "lr": 4.783114321084592e-06}
{"train_loss": 0.03150617331266403, "global_step": 233418, "epoch": 2622, "lr": 4.782866882313819e-06}
{"train_loss": 0.04434387758374214, "global_step": 233419, "epoch": 2622, "lr": 4.7826194496219275e-06}
{"train_loss": 0.012158586643636227, "global_step": 233420, "epoch": 2622, "lr": 4.782372023008968e-06}
{"train_loss": 0.05038553848862648, "global_step": 233421, "epoch": 2622, "lr": 4.782124602474963e-06}
{"train_loss": 0.07272420078516006, "global_step": 233422, "epoch": 2622, "lr": 4.781877188019951e-06}
{"train_loss": 0.05689726024866104, "global_step": 233423, "epoch": 2622, "lr": 4.78162977964397e-06}
{"train_loss": 0.028375454246997833, "global_step": 233424, "epoch": 2622, "lr": 4.781382377347038e-06}
{"train_loss": 0.023745056241750717, "global_step": 233425, "epoch": 2622, "lr": 4.781134981129209e-06}
{"train_loss": 0.046875275671482086, "global_step": 233426, "epoch": 2622, "lr": 4.780887590990496e-06}
{"train_loss": 0.02102448046207428, "global_step": 233427, "epoch": 2622, "lr": 4.780640206930953e-06}
{"train_loss": 0.055211324244737625, "global_step": 233428, "epoch": 2622, "lr": 4.780392828950592e-06}
{"train_loss": 0.06727434694766998, "global_step": 233429, "epoch": 2622, "lr": 4.780145457049462e-06}
{"train_loss": 0.05514154210686684, "global_step": 233430, "epoch": 2622, "lr": 4.7798980912275915e-06}
{"train_loss": 0.040987778455019, "global_step": 233431, "epoch": 2622, "lr": 4.779650731485003e-06}
{"train_loss": 0.044197190552949905, "global_step": 233432, "epoch": 2622, "lr": 4.779403377821751e-06}
{"train_loss": 0.0378931388258934, "global_step": 233433, "epoch": 2622, "lr": 4.779156030237847e-06}
{"train_loss": 0.07173480093479156, "global_step": 233434, "epoch": 2622, "lr": 4.778908688733341e-06}
{"train_loss": 0.057126473635435104, "global_step": 233435, "epoch": 2622, "lr": 4.778661353308261e-06}
{"train_loss": 0.06582488864660263, "global_step": 233436, "epoch": 2622, "lr": 4.77841402396263e-06}
{"train_loss": 0.03219045698642731, "global_step": 233437, "epoch": 2622, "lr": 4.778166700696496e-06}
{"train_loss": 0.07502897828817368, "global_step": 233438, "epoch": 2622, "lr": 4.777919383509883e-06}
{"train_loss": 0.023816660046577454, "global_step": 233439, "epoch": 2622, "lr": 4.777672072402833e-06}
{"train_loss": 0.02573845349252224, "global_step": 233440, "epoch": 2622, "lr": 4.777424767375366e-06}
{"train_loss": 0.0925258994102478, "global_step": 233441, "epoch": 2622, "lr": 4.7771774684275236e-06}
{"train_loss": 0.02633131667971611, "global_step": 233442, "epoch": 2622, "lr": 4.7769301755593465e-06}
{"train_loss": 0.025565333664417267, "global_step": 233443, "epoch": 2622, "lr": 4.77668288877085e-06}
{"train_loss": 0.038674309849739075, "global_step": 233444, "epoch": 2622, "lr": 4.776435608062085e-06}
{"train_loss": 0.06258654594421387, "global_step": 233445, "epoch": 2622, "lr": 4.776188333433074e-06}
{"train_loss": 0.04897294608938895, "global_step": 233446, "epoch": 2622, "lr": 4.7759410648838545e-06, "val_loss": 9.178452491760254}
{"train_loss": 0.05583428591489792, "global_step": 233447, "epoch": 2623, "lr": 4.775693802414455e-06}
{"train_loss": 0.04514677822589874, "global_step": 233448, "epoch": 2623, "lr": 4.77544654602492e-06}
{"train_loss": 0.0477890707552433, "global_step": 233449, "epoch": 2623, "lr": 4.775199295715266e-06}
{"train_loss": 0.067148856818676, "global_step": 233450, "epoch": 2623, "lr": 4.774952051485543e-06}
{"train_loss": 0.0746484324336052, "global_step": 233451, "epoch": 2623, "lr": 4.774704813335773e-06}
{"train_loss": 0.0403284952044487, "global_step": 233452, "epoch": 2623, "lr": 4.774457581265995e-06}
{"train_loss": 0.08494943380355835, "global_step": 233453, "epoch": 2623, "lr": 4.774210355276232e-06}
{"train_loss": 0.0714397206902504, "global_step": 233454, "epoch": 2623, "lr": 4.773963135366538e-06}
{"train_loss": 0.045753657817840576, "global_step": 233455, "epoch": 2623, "lr": 4.77371592153692e-06}
{"train_loss": 0.05367375165224075, "global_step": 233456, "epoch": 2623, "lr": 4.773468713787438e-06}
{"train_loss": 0.06614969670772552, "global_step": 233457, "epoch": 2623, "lr": 4.773221512118098e-06}
{"train_loss": 0.05839584395289421, "global_step": 233458, "epoch": 2623, "lr": 4.7729743165289495e-06}
{"train_loss": 0.04899747297167778, "global_step": 233459, "epoch": 2623, "lr": 4.772727127020032e-06}
{"train_loss": 0.0231329258531332, "global_step": 233460, "epoch": 2623, "lr": 4.772479943591362e-06}
{"train_loss": 0.08367801457643509, "global_step": 233461, "epoch": 2623, "lr": 4.77223276624299e-06}
{"train_loss": 0.037933409214019775, "global_step": 233462, "epoch": 2623, "lr": 4.771985594974932e-06}
{"train_loss": 0.03943108394742012, "global_step": 233463, "epoch": 2623, "lr": 4.771738429787237e-06}
{"train_loss": 0.017125234007835388, "global_step": 233464, "epoch": 2623, "lr": 4.7714912706799174e-06}
{"train_loss": 0.025527071207761765, "global_step": 233465, "epoch": 2623, "lr": 4.7712441176530344e-06}
{"train_loss": 0.028030429035425186, "global_step": 233466, "epoch": 2623, "lr": 4.770996970706593e-06}
{"train_loss": 0.03681321442127228, "global_step": 233467, "epoch": 2623, "lr": 4.770749829840654e-06}
{"train_loss": 0.046362146735191345, "global_step": 233468, "epoch": 2623, "lr": 4.770502695055235e-06}
{"train_loss": 0.04508208855986595, "global_step": 233469, "epoch": 2623, "lr": 4.770255566350356e-06}
{"train_loss": 0.06876000761985779, "global_step": 233470, "epoch": 2623, "lr": 4.770008443726082e-06}
{"train_loss": 0.04987134411931038, "global_step": 233471, "epoch": 2623, "lr": 4.769761327182415e-06}
{"train_loss": 0.018968818709254265, "global_step": 233472, "epoch": 2623, "lr": 4.769514216719412e-06}
{"train_loss": 0.040359728038311005, "global_step": 233473, "epoch": 2623, "lr": 4.769267112337089e-06}
{"train_loss": 0.022770360112190247, "global_step": 233474, "epoch": 2623, "lr": 4.769020014035491e-06}
{"train_loss": 0.03982476145029068, "global_step": 233475, "epoch": 2623, "lr": 4.768772921814651e-06}
{"train_loss": 0.07085121423006058, "global_step": 233476, "epoch": 2623, "lr": 4.768525835674592e-06}
{"train_loss": 0.051791734993457794, "global_step": 233477, "epoch": 2623, "lr": 4.7682787556153515e-06}
{"train_loss": 0.07744896411895752, "global_step": 233478, "epoch": 2623, "lr": 4.76803168163697e-06}
{"train_loss": 0.03932062163949013, "global_step": 233479, "epoch": 2623, "lr": 4.767784613739468e-06}
{"train_loss": 0.06405133008956909, "global_step": 233480, "epoch": 2623, "lr": 4.767537551922896e-06}
{"train_loss": 0.053389422595500946, "global_step": 233481, "epoch": 2623, "lr": 4.767290496187271e-06}
{"train_loss": 0.05986153706908226, "global_step": 233482, "epoch": 2623, "lr": 4.767043446532638e-06}
{"train_loss": 0.044106625020504, "global_step": 233483, "epoch": 2623, "lr": 4.766796402959018e-06}
{"train_loss": 0.029817746952176094, "global_step": 233484, "epoch": 2623, "lr": 4.766549365466455e-06}
{"train_loss": 0.05606153607368469, "global_step": 233485, "epoch": 2623, "lr": 4.766302334054973e-06}
{"train_loss": 0.03724164888262749, "global_step": 233486, "epoch": 2623, "lr": 4.766055308724621e-06}
{"train_loss": 0.03897390514612198, "global_step": 233487, "epoch": 2623, "lr": 4.765808289475404e-06}
{"train_loss": 0.05051148682832718, "global_step": 233488, "epoch": 2623, "lr": 4.765561276307389e-06}
{"train_loss": 0.06518980115652084, "global_step": 233489, "epoch": 2623, "lr": 4.765314269220583e-06}
{"train_loss": 0.06052383780479431, "global_step": 233490, "epoch": 2623, "lr": 4.76506726821504e-06}
{"train_loss": 0.04405093193054199, "global_step": 233491, "epoch": 2623, "lr": 4.764820273290771e-06}
{"train_loss": 0.03092682734131813, "global_step": 233492, "epoch": 2623, "lr": 4.7645732844478316e-06}
{"train_loss": 0.017435096204280853, "global_step": 233493, "epoch": 2623, "lr": 4.764326301686234e-06}
{"train_loss": 0.04257809370756149, "global_step": 233494, "epoch": 2623, "lr": 4.764079325006016e-06}
{"train_loss": 0.042200855910778046, "global_step": 233495, "epoch": 2623, "lr": 4.7638323544072326e-06}
{"train_loss": 0.047130096703767776, "global_step": 233496, "epoch": 2623, "lr": 4.763585389889891e-06}
{"train_loss": 0.0380803681910038, "global_step": 233497, "epoch": 2623, "lr": 4.763338431454039e-06}
{"train_loss": 0.06005306541919708, "global_step": 233498, "epoch": 2623, "lr": 4.763091479099696e-06}
{"train_loss": 0.04272470250725746, "global_step": 233499, "epoch": 2623, "lr": 4.762844532826916e-06}
{"train_loss": 0.02091085910797119, "global_step": 233500, "epoch": 2623, "lr": 4.762597592635709e-06}
{"train_loss": 0.04550820216536522, "global_step": 233501, "epoch": 2623, "lr": 4.762350658526133e-06}
{"train_loss": 0.04527289420366287, "global_step": 233502, "epoch": 2623, "lr": 4.762103730498191e-06}
{"train_loss": 0.05051657557487488, "global_step": 233503, "epoch": 2623, "lr": 4.761856808551945e-06}
{"train_loss": 0.03681899607181549, "global_step": 233504, "epoch": 2623, "lr": 4.761609892687407e-06}
{"train_loss": 0.047572605311870575, "global_step": 233505, "epoch": 2623, "lr": 4.7613629829046325e-06}
{"train_loss": 0.03631626442074776, "global_step": 233506, "epoch": 2623, "lr": 4.7611160792036255e-06}
{"train_loss": 0.04950328543782234, "global_step": 233507, "epoch": 2623, "lr": 4.760869181584449e-06}
{"train_loss": 0.0970618799328804, "global_step": 233508, "epoch": 2623, "lr": 4.760622290047123e-06}
{"train_loss": 0.04348473995923996, "global_step": 233509, "epoch": 2623, "lr": 4.7603754045916656e-06}
{"train_loss": 0.039413101971149445, "global_step": 233510, "epoch": 2623, "lr": 4.7601285252181385e-06}
{"train_loss": 0.094846710562706, "global_step": 233511, "epoch": 2623, "lr": 4.759881651926545e-06}
{"train_loss": 0.03777303174138069, "global_step": 233512, "epoch": 2623, "lr": 4.759634784716948e-06}
{"train_loss": 0.08256032317876816, "global_step": 233513, "epoch": 2623, "lr": 4.759387923589359e-06}
{"train_loss": 0.031127644702792168, "global_step": 233514, "epoch": 2623, "lr": 4.759141068543827e-06}
{"train_loss": 0.020344959571957588, "global_step": 233515, "epoch": 2623, "lr": 4.758894219580367e-06}
{"train_loss": 0.09357200562953949, "global_step": 233516, "epoch": 2623, "lr": 4.758647376699032e-06}
{"train_loss": 0.034525804221630096, "global_step": 233517, "epoch": 2623, "lr": 4.758400539899832e-06}
{"train_loss": 0.0440584272146225, "global_step": 233518, "epoch": 2623, "lr": 4.758153709182828e-06}
{"train_loss": 0.041627563536167145, "global_step": 233519, "epoch": 2623, "lr": 4.75790688454803e-06}
{"train_loss": 0.02786974236369133, "global_step": 233520, "epoch": 2623, "lr": 4.757660065995484e-06}
{"train_loss": 0.05113353580236435, "global_step": 233521, "epoch": 2623, "lr": 4.757413253525217e-06}
{"train_loss": 0.01711694523692131, "global_step": 233522, "epoch": 2623, "lr": 4.7571664471372675e-06}
{"train_loss": 0.06213754415512085, "global_step": 233523, "epoch": 2623, "lr": 4.756919646831659e-06}
{"train_loss": 0.0157974474132061, "global_step": 233524, "epoch": 2623, "lr": 4.756672852608441e-06}
{"train_loss": 0.05180563032627106, "global_step": 233525, "epoch": 2623, "lr": 4.756426064467623e-06}
{"train_loss": 0.026233715936541557, "global_step": 233526, "epoch": 2623, "lr": 4.756179282409268e-06}
{"train_loss": 0.01281778421252966, "global_step": 233527, "epoch": 2623, "lr": 4.755932506433381e-06}
{"train_loss": 0.03583189472556114, "global_step": 233528, "epoch": 2623, "lr": 4.7556857365400165e-06}
{"train_loss": 0.0422062873840332, "global_step": 233529, "epoch": 2623, "lr": 4.755438972729187e-06}
{"train_loss": 0.026705898344516754, "global_step": 233530, "epoch": 2623, "lr": 4.755192215000942e-06}
{"train_loss": 0.026412395760416985, "global_step": 233531, "epoch": 2623, "lr": 4.7549454633553145e-06}
{"train_loss": 0.023463401943445206, "global_step": 233532, "epoch": 2623, "lr": 4.754698717792327e-06}
{"train_loss": 0.030285799875855446, "global_step": 233533, "epoch": 2623, "lr": 4.75445197831203e-06}
{"train_loss": 0.01583722233772278, "global_step": 233534, "epoch": 2623, "lr": 4.754205244914434e-06}
{"train_loss": 0.045933561996044064, "global_step": 233535, "epoch": 2623, "lr": 4.753958517599593e-06, "val_loss": 9.323126792907715}
{"train_loss": 0.03997713327407837, "global_step": 233536, "epoch": 2624, "lr": 4.753711796367522e-06}
{"train_loss": 0.0264875628054142, "global_step": 233537, "epoch": 2624, "lr": 4.753465081218267e-06}
{"train_loss": 0.02144133485853672, "global_step": 233538, "epoch": 2624, "lr": 4.753218372151852e-06}
{"train_loss": 0.07711655646562576, "global_step": 233539, "epoch": 2624, "lr": 4.752971669168327e-06}
{"train_loss": 0.050161268562078476, "global_step": 233540, "epoch": 2624, "lr": 4.752724972267703e-06}
{"train_loss": 0.020350562408566475, "global_step": 233541, "epoch": 2624, "lr": 4.752478281450029e-06}
{"train_loss": 0.009320889599621296, "global_step": 233542, "epoch": 2624, "lr": 4.7522315967153285e-06}
{"train_loss": 0.05385177582502365, "global_step": 233543, "epoch": 2624, "lr": 4.751984918063645e-06}
{"train_loss": 0.06919880211353302, "global_step": 233544, "epoch": 2624, "lr": 4.751738245494996e-06}
{"train_loss": 0.0710054337978363, "global_step": 233545, "epoch": 2624, "lr": 4.751491579009437e-06}
{"train_loss": 0.030995327979326248, "global_step": 233546, "epoch": 2624, "lr": 4.75124491860699e-06}
{"train_loss": 0.03188768029212952, "global_step": 233547, "epoch": 2624, "lr": 4.7509982642876706e-06}
{"train_loss": 0.0370185486972332, "global_step": 233548, "epoch": 2624, "lr": 4.750751616051541e-06}
{"train_loss": 0.043230339884757996, "global_step": 233549, "epoch": 2624, "lr": 4.750504973898612e-06}
{"train_loss": 0.059711191803216934, "global_step": 233550, "epoch": 2624, "lr": 4.750258337828934e-06}
{"train_loss": 0.03147002309560776, "global_step": 233551, "epoch": 2624, "lr": 4.750011707842522e-06}
{"train_loss": 0.024721337482333183, "global_step": 233552, "epoch": 2624, "lr": 4.7497650839394336e-06}
{"train_loss": 0.048008278012275696, "global_step": 233553, "epoch": 2624, "lr": 4.749518466119674e-06}
{"train_loss": 0.08574787527322769, "global_step": 233554, "epoch": 2624, "lr": 4.749271854383303e-06}
{"train_loss": 0.03432193398475647, "global_step": 233555, "epoch": 2624, "lr": 4.749025248730327e-06}
{"train_loss": 0.02924158051609993, "global_step": 233556, "epoch": 2624, "lr": 4.748778649160807e-06}
{"train_loss": 0.08127304911613464, "global_step": 233557, "epoch": 2624, "lr": 4.748532055674748e-06}
{"train_loss": 0.07858613133430481, "global_step": 233558, "epoch": 2624, "lr": 4.748285468272212e-06}
{"train_loss": 0.050449568778276443, "global_step": 233559, "epoch": 2624, "lr": 4.748038886953204e-06}
{"train_loss": 0.04017703980207443, "global_step": 233560, "epoch": 2624, "lr": 4.74779231171778e-06}
{"train_loss": 0.08883564174175262, "global_step": 233561, "epoch": 2624, "lr": 4.747545742565956e-06}
{"train_loss": 0.06302253901958466, "global_step": 233562, "epoch": 2624, "lr": 4.7472991794977826e-06}
{"train_loss": 0.06900157034397125, "global_step": 233563, "epoch": 2624, "lr": 4.747052622513276e-06}
{"train_loss": 0.028813427314162254, "global_step": 233564, "epoch": 2624, "lr": 4.746806071612481e-06}
{"train_loss": 0.033694371581077576, "global_step": 233565, "epoch": 2624, "lr": 4.7465595267954244e-06}
{"train_loss": 0.08800595253705978, "global_step": 233566, "epoch": 2624, "lr": 4.746312988062141e-06}
{"train_loss": 0.04707762971520424, "global_step": 233567, "epoch": 2624, "lr": 4.746066455412662e-06}
{"train_loss": 0.07263988256454468, "global_step": 233568, "epoch": 2624, "lr": 4.7458199288470175e-06}
{"train_loss": 0.022545645013451576, "global_step": 233569, "epoch": 2624, "lr": 4.745573408365262e-06}
{"train_loss": 0.06137099117040634, "global_step": 233570, "epoch": 2624, "lr": 4.7453268939674e-06}
{"train_loss": 0.05285588279366493, "global_step": 233571, "epoch": 2624, "lr": 4.745080385653483e-06}
{"train_loss": 0.08291351050138474, "global_step": 233572, "epoch": 2624, "lr": 4.744833883423533e-06}
{"train_loss": 0.05284063145518303, "global_step": 233573, "epoch": 2624, "lr": 4.744587387277599e-06}
{"train_loss": 0.0382576547563076, "global_step": 233574, "epoch": 2624, "lr": 4.744340897215688e-06}
{"train_loss": 0.064626544713974, "global_step": 233575, "epoch": 2624, "lr": 4.7440944132378655e-06}
{"train_loss": 0.03662465885281563, "global_step": 233576, "epoch": 2624, "lr": 4.743847935344137e-06}
{"train_loss": 0.06663883477449417, "global_step": 233577, "epoch": 2624, "lr": 4.743601463534553e-06}
{"train_loss": 0.04207215830683708, "global_step": 233578, "epoch": 2624, "lr": 4.743354997809129e-06}
{"train_loss": 0.10548651963472366, "global_step": 233579, "epoch": 2624, "lr": 4.743108538167923e-06}
{"train_loss": 0.04680947959423065, "global_step": 233580, "epoch": 2624, "lr": 4.742862084610944e-06}
{"train_loss": 0.0404534637928009, "global_step": 233581, "epoch": 2624, "lr": 4.742615637138248e-06}
{"train_loss": 0.056796956807374954, "global_step": 233582, "epoch": 2624, "lr": 4.742369195749841e-06}
{"train_loss": 0.07437902688980103, "global_step": 233583, "epoch": 2624, "lr": 4.742122760445783e-06}
{"train_loss": 0.06004960462450981, "global_step": 233584, "epoch": 2624, "lr": 4.741876331226092e-06}
{"train_loss": 0.035573966801166534, "global_step": 233585, "epoch": 2624, "lr": 4.7416299080908e-06}
{"train_loss": 0.05173976719379425, "global_step": 233586, "epoch": 2624, "lr": 4.7413834910399465e-06}
{"train_loss": 0.029013995081186295, "global_step": 233587, "epoch": 2624, "lr": 4.74113708007356e-06}
{"train_loss": 0.04152098298072815, "global_step": 233588, "epoch": 2624, "lr": 4.7408906751916835e-06}
{"train_loss": 0.0323428176343441, "global_step": 233589, "epoch": 2624, "lr": 4.7406442763943295e-06}
{"train_loss": 0.13604886829853058, "global_step": 233590, "epoch": 2624, "lr": 4.740397883681558e-06}
{"train_loss": 0.033323369920253754, "global_step": 233591, "epoch": 2624, "lr": 4.74015149705338e-06}
{"train_loss": 0.024952294304966927, "global_step": 233592, "epoch": 2624, "lr": 4.739905116509841e-06}
{"train_loss": 0.07913646101951599, "global_step": 233593, "epoch": 2624, "lr": 4.739658742050962e-06}
{"train_loss": 0.03509777411818504, "global_step": 233594, "epoch": 2624, "lr": 4.7394123736767994e-06}
{"train_loss": 0.06460195779800415, "global_step": 233595, "epoch": 2624, "lr": 4.739166011387352e-06}
{"train_loss": 0.0548187755048275, "global_step": 233596, "epoch": 2624, "lr": 4.738919655182688e-06}
{"train_loss": 0.051021758466959, "global_step": 233597, "epoch": 2624, "lr": 4.738673305062819e-06}
{"train_loss": 0.07307940721511841, "global_step": 233598, "epoch": 2624, "lr": 4.738426961027786e-06}
{"train_loss": 0.055488064885139465, "global_step": 233599, "epoch": 2624, "lr": 4.738180623077615e-06}
{"train_loss": 0.011947611346840858, "global_step": 233600, "epoch": 2624, "lr": 4.737934291212348e-06}
{"train_loss": 0.017882341518998146, "global_step": 233601, "epoch": 2624, "lr": 4.737687965432009e-06}
{"train_loss": 0.03290607035160065, "global_step": 233602, "epoch": 2624, "lr": 4.737441645736646e-06}
{"train_loss": 0.01585342176258564, "global_step": 233603, "epoch": 2624, "lr": 4.737195332126271e-06}
{"train_loss": 0.05176986753940582, "global_step": 233604, "epoch": 2624, "lr": 4.736949024600929e-06}
{"train_loss": 0.05114222317934036, "global_step": 233605, "epoch": 2624, "lr": 4.736702723160657e-06}
{"train_loss": 0.04664979502558708, "global_step": 233606, "epoch": 2624, "lr": 4.73645642780548e-06}
{"train_loss": 0.010934500023722649, "global_step": 233607, "epoch": 2624, "lr": 4.736210138535447e-06}
{"train_loss": 0.05317974090576172, "global_step": 233608, "epoch": 2624, "lr": 4.7359638553505614e-06}
{"train_loss": 0.03839314356446266, "global_step": 233609, "epoch": 2624, "lr": 4.735717578250886e-06}
{"train_loss": 0.028436558321118355, "global_step": 233610, "epoch": 2624, "lr": 4.735471307236433e-06}
{"train_loss": 0.026637883856892586, "global_step": 233611, "epoch": 2624, "lr": 4.735225042307256e-06}
{"train_loss": 0.040964409708976746, "global_step": 233612, "epoch": 2624, "lr": 4.734978783463367e-06}
{"train_loss": 0.05833250656723976, "global_step": 233613, "epoch": 2624, "lr": 4.734732530704816e-06}
{"train_loss": 0.024426475167274475, "global_step": 233614, "epoch": 2624, "lr": 4.73448628403162e-06}
{"train_loss": 0.07708551734685898, "global_step": 233615, "epoch": 2624, "lr": 4.7342400434438275e-06}
{"train_loss": 0.03473886474967003, "global_step": 233616, "epoch": 2624, "lr": 4.733993808941456e-06}
{"train_loss": 0.07941737771034241, "global_step": 233617, "epoch": 2624, "lr": 4.733747580524556e-06}
{"train_loss": 0.019610153511166573, "global_step": 233618, "epoch": 2624, "lr": 4.733501358193143e-06}
{"train_loss": 0.021390793845057487, "global_step": 233619, "epoch": 2624, "lr": 4.733255141947268e-06}
{"train_loss": 0.05654270574450493, "global_step": 233620, "epoch": 2624, "lr": 4.7330089317869475e-06}
{"train_loss": 0.07601159811019897, "global_step": 233621, "epoch": 2624, "lr": 4.732762727712231e-06}
{"train_loss": 0.06772458553314209, "global_step": 233622, "epoch": 2624, "lr": 4.732516529723136e-06}
{"train_loss": 0.07198724150657654, "global_step": 233623, "epoch": 2624, "lr": 4.732270337819711e-06}
{"train_loss": 0.049450666015821224, "global_step": 233624, "epoch": 2624, "lr": 4.73202415200198e-06, "val_loss": 9.396622657775879}
{"train_loss": 0.028936607763171196, "global_step": 233625, "epoch": 2625, "lr": 4.731777972269963e-06}
{"train_loss": 0.027119025588035583, "global_step": 233626, "epoch": 2625, "lr": 4.731531798623718e-06}
{"train_loss": 0.03956938162446022, "global_step": 233627, "epoch": 2625, "lr": 4.731285631063259e-06}
{"train_loss": 0.043269623070955276, "global_step": 233628, "epoch": 2625, "lr": 4.731039469588639e-06}
{"train_loss": 0.05762224271893501, "global_step": 233629, "epoch": 2625, "lr": 4.7307933141998665e-06}
{"train_loss": 0.03832549601793289, "global_step": 233630, "epoch": 2625, "lr": 4.730547164896992e-06}
{"train_loss": 0.04635593667626381, "global_step": 233631, "epoch": 2625, "lr": 4.730301021680039e-06}
{"train_loss": 0.05475710704922676, "global_step": 233632, "epoch": 2625, "lr": 4.730054884549057e-06}
{"train_loss": 0.03785218298435211, "global_step": 233633, "epoch": 2625, "lr": 4.729808753504056e-06}
{"train_loss": 0.06958772987127304, "global_step": 233634, "epoch": 2625, "lr": 4.729562628545087e-06}
{"train_loss": 0.07821616530418396, "global_step": 233635, "epoch": 2625, "lr": 4.729316509672172e-06}
{"train_loss": 0.012052313424646854, "global_step": 233636, "epoch": 2625, "lr": 4.729070396885354e-06}
{"train_loss": 0.07118509709835052, "global_step": 233637, "epoch": 2625, "lr": 4.7288242901846525e-06}
{"train_loss": 0.06358002126216888, "global_step": 233638, "epoch": 2625, "lr": 4.728578189570121e-06}
{"train_loss": 0.06545380502939224, "global_step": 233639, "epoch": 2625, "lr": 4.728332095041766e-06}
{"train_loss": 0.09168471395969391, "global_step": 233640, "epoch": 2625, "lr": 4.728086006599636e-06}
{"train_loss": 0.017235955223441124, "global_step": 233641, "epoch": 2625, "lr": 4.727839924243776e-06}
{"train_loss": 0.03542729839682579, "global_step": 233642, "epoch": 2625, "lr": 4.7275938479741934e-06}
{"train_loss": 0.024044999852776527, "global_step": 233643, "epoch": 2625, "lr": 4.727347777790947e-06}
{"train_loss": 0.021111490204930305, "global_step": 233644, "epoch": 2625, "lr": 4.727101713694043e-06}
{"train_loss": 0.044490449130535126, "global_step": 233645, "epoch": 2625, "lr": 4.726855655683543e-06}
{"train_loss": 0.03963084518909454, "global_step": 233646, "epoch": 2625, "lr": 4.726609603759452e-06}
{"train_loss": 0.08441438525915146, "global_step": 233647, "epoch": 2625, "lr": 4.726363557921826e-06}
{"train_loss": 0.09430546313524246, "global_step": 233648, "epoch": 2625, "lr": 4.726117518170681e-06}
{"train_loss": 0.05818644538521767, "global_step": 233649, "epoch": 2625, "lr": 4.725871484506067e-06}
{"train_loss": 0.05294999107718468, "global_step": 233650, "epoch": 2625, "lr": 4.725625456927996e-06}
{"train_loss": 0.07305256277322769, "global_step": 233651, "epoch": 2625, "lr": 4.725379435436522e-06}
{"train_loss": 0.04646340757608414, "global_step": 233652, "epoch": 2625, "lr": 4.725133420031663e-06}
{"train_loss": 0.05770057812333107, "global_step": 233653, "epoch": 2625, "lr": 4.724887410713464e-06}
{"train_loss": 0.017257386818528175, "global_step": 233654, "epoch": 2625, "lr": 4.724641407481944e-06}
{"train_loss": 0.03262260556221008, "global_step": 233655, "epoch": 2625, "lr": 4.7243954103371565e-06}
{"train_loss": 0.01617077738046646, "global_step": 233656, "epoch": 2625, "lr": 4.724149419279106e-06}
{"train_loss": 0.028400223702192307, "global_step": 233657, "epoch": 2625, "lr": 4.723903434307858e-06}
{"train_loss": 0.02938045933842659, "global_step": 233658, "epoch": 2625, "lr": 4.72365745542342e-06}
{"train_loss": 0.05986686423420906, "global_step": 233659, "epoch": 2625, "lr": 4.723411482625839e-06}
{"train_loss": 0.04052210599184036, "global_step": 233660, "epoch": 2625, "lr": 4.723165515915135e-06}
{"train_loss": 0.013302251696586609, "global_step": 233661, "epoch": 2625, "lr": 4.722919555291361e-06}
{"train_loss": 0.060240037739276886, "global_step": 233662, "epoch": 2625, "lr": 4.722673600754535e-06}
{"train_loss": 0.018469274044036865, "global_step": 233663, "epoch": 2625, "lr": 4.7224276523046894e-06}
{"train_loss": 0.05372786894440651, "global_step": 233664, "epoch": 2625, "lr": 4.72218170994187e-06}
{"train_loss": 0.0693107545375824, "global_step": 233665, "epoch": 2625, "lr": 4.721935773666086e-06}
{"train_loss": 0.04891104996204376, "global_step": 233666, "epoch": 2625, "lr": 4.7216898434774e-06}
{"train_loss": 0.04717984050512314, "global_step": 233667, "epoch": 2625, "lr": 4.721443919375823e-06}
{"train_loss": 0.05668317526578903, "global_step": 233668, "epoch": 2625, "lr": 4.721198001361404e-06}
{"train_loss": 0.08850331604480743, "global_step": 233669, "epoch": 2625, "lr": 4.720952089434161e-06}
{"train_loss": 0.00726698525249958, "global_step": 233670, "epoch": 2625, "lr": 4.7207061835941365e-06}
{"train_loss": 0.03338170424103737, "global_step": 233671, "epoch": 2625, "lr": 4.72046028384136e-06}
{"train_loss": 0.08759493380784988, "global_step": 233672, "epoch": 2625, "lr": 4.72021439017587e-06}
{"train_loss": 0.028330665081739426, "global_step": 233673, "epoch": 2625, "lr": 4.7199685025976885e-06}
{"train_loss": 0.07475095242261887, "global_step": 233674, "epoch": 2625, "lr": 4.719722621106865e-06}
{"train_loss": 0.04709526151418686, "global_step": 233675, "epoch": 2625, "lr": 4.719476745703411e-06}
{"train_loss": 0.0636371523141861, "global_step": 233676, "epoch": 2625, "lr": 4.7192308763873704e-06}
{"train_loss": 0.04812473803758621, "global_step": 233677, "epoch": 2625, "lr": 4.718985013158789e-06}
{"train_loss": 0.035895995795726776, "global_step": 233678, "epoch": 2625, "lr": 4.718739156017676e-06}
{"train_loss": 0.06927845627069473, "global_step": 233679, "epoch": 2625, "lr": 4.718493304964089e-06}
{"train_loss": 0.07810503244400024, "global_step": 233680, "epoch": 2625, "lr": 4.718247459998043e-06}
{"train_loss": 0.031714968383312225, "global_step": 233681, "epoch": 2625, "lr": 4.718001621119578e-06}
{"train_loss": 0.03028390184044838, "global_step": 233682, "epoch": 2625, "lr": 4.71775578832872e-06}
{"train_loss": 0.02054520510137081, "global_step": 233683, "epoch": 2625, "lr": 4.717509961625521e-06}
{"train_loss": 0.08712810277938843, "global_step": 233684, "epoch": 2625, "lr": 4.717264141009986e-06}
{"train_loss": 0.05027757212519646, "global_step": 233685, "epoch": 2625, "lr": 4.717018326482176e-06}
{"train_loss": 0.07162163406610489, "global_step": 233686, "epoch": 2625, "lr": 4.716772518042101e-06}
{"train_loss": 0.10029637068510056, "global_step": 233687, "epoch": 2625, "lr": 4.716526715689812e-06}
{"train_loss": 0.04152553528547287, "global_step": 233688, "epoch": 2625, "lr": 4.716280919425326e-06}
{"train_loss": 0.07433057576417923, "global_step": 233689, "epoch": 2625, "lr": 4.7160351292486915e-06}
{"train_loss": 0.0350145660340786, "global_step": 233690, "epoch": 2625, "lr": 4.715789345159921e-06}
{"train_loss": 0.0753297284245491, "global_step": 233691, "epoch": 2625, "lr": 4.7155435671590755e-06}
{"train_loss": 0.04999814182519913, "global_step": 233692, "epoch": 2625, "lr": 4.715297795246165e-06}
{"train_loss": 0.045377787202596664, "global_step": 233693, "epoch": 2625, "lr": 4.715052029421241e-06}
{"train_loss": 0.05181286484003067, "global_step": 233694, "epoch": 2625, "lr": 4.714806269684313e-06}
{"train_loss": 0.06545891612768173, "global_step": 233695, "epoch": 2625, "lr": 4.714560516035437e-06}
{"train_loss": 0.0083006601780653, "global_step": 233696, "epoch": 2625, "lr": 4.714314768474626e-06}
{"train_loss": 0.060589227825403214, "global_step": 233697, "epoch": 2625, "lr": 4.714069027001938e-06}
{"train_loss": 0.09452934563159943, "global_step": 233698, "epoch": 2625, "lr": 4.713823291617376e-06}
{"train_loss": 0.016944503411650658, "global_step": 233699, "epoch": 2625, "lr": 4.713577562321003e-06}
{"train_loss": 0.039131589233875275, "global_step": 233700, "epoch": 2625, "lr": 4.7133318391128335e-06}
{"train_loss": 0.03430844098329544, "global_step": 233701, "epoch": 2625, "lr": 4.713086121992899e-06}
{"train_loss": 0.06104706600308418, "global_step": 233702, "epoch": 2625, "lr": 4.712840410961244e-06}
{"train_loss": 0.07925516366958618, "global_step": 233703, "epoch": 2625, "lr": 4.7125947060178864e-06}
{"train_loss": 0.06752228736877441, "global_step": 233704, "epoch": 2625, "lr": 4.71234900716288e-06}
{"train_loss": 0.025687111541628838, "global_step": 233705, "epoch": 2625, "lr": 4.712103314396238e-06}
{"train_loss": 0.0688532292842865, "global_step": 233706, "epoch": 2625, "lr": 4.711857627718008e-06}
{"train_loss": 0.03915456309914589, "global_step": 233707, "epoch": 2625, "lr": 4.711611947128208e-06}
{"train_loss": 0.05294327810406685, "global_step": 233708, "epoch": 2625, "lr": 4.711366272626888e-06}
{"train_loss": 0.03191743418574333, "global_step": 233709, "epoch": 2625, "lr": 4.711120604214064e-06}
{"train_loss": 0.03968394175171852, "global_step": 233710, "epoch": 2625, "lr": 4.7108749418897926e-06}
{"train_loss": 0.07533743977546692, "global_step": 233711, "epoch": 2625, "lr": 4.710629285654078e-06}
{"train_loss": 0.09271550178527832, "global_step": 233712, "epoch": 2625, "lr": 4.710383635506965e-06}
{"train_loss": 0.05035951596506861, "global_step": 233713, "epoch": 2625, "lr": 4.710137991448504e-06, "val_loss": 9.382539749145508, "train_action_mse_error": 17.494661331176758}
{"train_loss": 0.08176067471504211, "global_step": 233714, "epoch": 2626, "lr": 4.709892353478701e-06}
{"train_loss": 0.07220812141895294, "global_step": 233715, "epoch": 2626, "lr": 4.709646721597611e-06}
{"train_loss": 0.022270705550909042, "global_step": 233716, "epoch": 2626, "lr": 4.709401095805243e-06}
{"train_loss": 0.03845374658703804, "global_step": 233717, "epoch": 2626, "lr": 4.709155476101656e-06}
{"train_loss": 0.02659844420850277, "global_step": 233718, "epoch": 2626, "lr": 4.7089098624868656e-06}
{"train_loss": 0.08283903449773788, "global_step": 233719, "epoch": 2626, "lr": 4.708664254960915e-06}
{"train_loss": 0.031131897121667862, "global_step": 233720, "epoch": 2626, "lr": 4.708418653523822e-06}
{"train_loss": 0.040413279086351395, "global_step": 233721, "epoch": 2626, "lr": 4.708173058175647e-06}
{"train_loss": 0.10413076728582382, "global_step": 233722, "epoch": 2626, "lr": 4.70792746891639e-06}
{"train_loss": 0.02879449352622032, "global_step": 233723, "epoch": 2626, "lr": 4.7076818857461124e-06}
{"train_loss": 0.0738784670829773, "global_step": 233724, "epoch": 2626, "lr": 4.70743630866482e-06}
{"train_loss": 0.038160540163517, "global_step": 233725, "epoch": 2626, "lr": 4.707190737672579e-06}
{"train_loss": 0.04576319828629494, "global_step": 233726, "epoch": 2626, "lr": 4.706945172769389e-06}
{"train_loss": 0.04745752364397049, "global_step": 233727, "epoch": 2626, "lr": 4.706699613955306e-06}
{"train_loss": 0.03650222718715668, "global_step": 233728, "epoch": 2626, "lr": 4.706454061230353e-06}
{"train_loss": 0.03532255440950394, "global_step": 233729, "epoch": 2626, "lr": 4.706208514594568e-06}
{"train_loss": 0.06870749592781067, "global_step": 233730, "epoch": 2626, "lr": 4.705962974047973e-06}
{"train_loss": 0.03017200157046318, "global_step": 233731, "epoch": 2626, "lr": 4.705717439590612e-06}
{"train_loss": 0.026597870513796806, "global_step": 233732, "epoch": 2626, "lr": 4.705471911222525e-06}
{"train_loss": 0.05928891897201538, "global_step": 233733, "epoch": 2626, "lr": 4.705226388943723e-06}
{"train_loss": 0.04416933283209801, "global_step": 233734, "epoch": 2626, "lr": 4.704980872754261e-06}
{"train_loss": 0.06640654057264328, "global_step": 233735, "epoch": 2626, "lr": 4.704735362654161e-06}
{"train_loss": 0.028020966798067093, "global_step": 233736, "epoch": 2626, "lr": 4.704489858643452e-06}
{"train_loss": 0.061855804175138474, "global_step": 233737, "epoch": 2626, "lr": 4.704244360722177e-06}
{"train_loss": 0.04660949483513832, "global_step": 233738, "epoch": 2626, "lr": 4.703998868890358e-06}
{"train_loss": 0.06796903163194656, "global_step": 233739, "epoch": 2626, "lr": 4.703753383148041e-06}
{"train_loss": 0.027306294068694115, "global_step": 233740, "epoch": 2626, "lr": 4.703507903495258e-06}
{"train_loss": 0.04334471374750137, "global_step": 233741, "epoch": 2626, "lr": 4.70326242993202e-06}
{"train_loss": 0.04526347294449806, "global_step": 233742, "epoch": 2626, "lr": 4.70301696245839e-06}
{"train_loss": 0.03195205703377724, "global_step": 233743, "epoch": 2626, "lr": 4.702771501074377e-06}
{"train_loss": 0.05808502808213234, "global_step": 233744, "epoch": 2626, "lr": 4.702526045780032e-06}
{"train_loss": 0.10540508478879929, "global_step": 233745, "epoch": 2626, "lr": 4.70228059657537e-06}
{"train_loss": 0.05535447224974632, "global_step": 233746, "epoch": 2626, "lr": 4.702035153460449e-06}
{"train_loss": 0.04470130428671837, "global_step": 233747, "epoch": 2626, "lr": 4.701789716435273e-06}
{"train_loss": 0.018965477123856544, "global_step": 233748, "epoch": 2626, "lr": 4.701544285499893e-06}
{"train_loss": 0.0612647719681263, "global_step": 233749, "epoch": 2626, "lr": 4.701298860654346e-06}
{"train_loss": 0.04309836030006409, "global_step": 233750, "epoch": 2626, "lr": 4.701053441898645e-06}
{"train_loss": 0.030101358890533447, "global_step": 233751, "epoch": 2626, "lr": 4.70080802923285e-06}
{"train_loss": 0.019472798332571983, "global_step": 233752, "epoch": 2626, "lr": 4.700562622656962e-06}
{"train_loss": 0.06414998322725296, "global_step": 233753, "epoch": 2626, "lr": 4.700317222171047e-06}
{"train_loss": 0.06044875830411911, "global_step": 233754, "epoch": 2626, "lr": 4.70007182777511e-06}
{"train_loss": 0.013462900184094906, "global_step": 233755, "epoch": 2626, "lr": 4.699826439469207e-06}
{"train_loss": 0.059066906571388245, "global_step": 233756, "epoch": 2626, "lr": 4.699581057253355e-06}
{"train_loss": 0.08232886344194412, "global_step": 233757, "epoch": 2626, "lr": 4.699335681127598e-06}
{"train_loss": 0.033791813999414444, "global_step": 233758, "epoch": 2626, "lr": 4.699090311091953e-06}
{"train_loss": 0.055182307958602905, "global_step": 233759, "epoch": 2626, "lr": 4.698844947146469e-06}
{"train_loss": 0.02864731289446354, "global_step": 233760, "epoch": 2626, "lr": 4.69859958929117e-06}
{"train_loss": 0.02936239168047905, "global_step": 233761, "epoch": 2626, "lr": 4.698354237526098e-06}
{"train_loss": 0.051107726991176605, "global_step": 233762, "epoch": 2626, "lr": 4.698108891851272e-06}
{"train_loss": 0.022868718951940536, "global_step": 233763, "epoch": 2626, "lr": 4.69786355226674e-06}
{"train_loss": 0.054357659071683884, "global_step": 233764, "epoch": 2626, "lr": 4.697618218772521e-06}
{"train_loss": 0.029525086283683777, "global_step": 233765, "epoch": 2626, "lr": 4.697372891368668e-06}
{"train_loss": 0.026026548817753792, "global_step": 233766, "epoch": 2626, "lr": 4.697127570055188e-06}
{"train_loss": 0.04635276645421982, "global_step": 233767, "epoch": 2626, "lr": 4.696882254832124e-06}
{"train_loss": 0.017954032868146896, "global_step": 233768, "epoch": 2626, "lr": 4.696636945699523e-06}
{"train_loss": 0.09212888032197952, "global_step": 233769, "epoch": 2626, "lr": 4.6963916426574e-06}
{"train_loss": 0.031879957765340805, "global_step": 233770, "epoch": 2626, "lr": 4.696146345705804e-06}
{"train_loss": 0.07600346207618713, "global_step": 233771, "epoch": 2626, "lr": 4.695901054844748e-06}
{"train_loss": 0.039096418768167496, "global_step": 233772, "epoch": 2626, "lr": 4.6956557700742865e-06}
{"train_loss": 0.04092632234096527, "global_step": 233773, "epoch": 2626, "lr": 4.695410491394442e-06}
{"train_loss": 0.04452583193778992, "global_step": 233774, "epoch": 2626, "lr": 4.695165218805236e-06}
{"train_loss": 0.030471613630652428, "global_step": 233775, "epoch": 2626, "lr": 4.694919952306725e-06}
{"train_loss": 0.05505973845720291, "global_step": 233776, "epoch": 2626, "lr": 4.69467469189892e-06}
{"train_loss": 0.02370161935687065, "global_step": 233777, "epoch": 2626, "lr": 4.694429437581876e-06}
{"train_loss": 0.08635653555393219, "global_step": 233778, "epoch": 2626, "lr": 4.694184189355611e-06}
{"train_loss": 0.06330748647451401, "global_step": 233779, "epoch": 2626, "lr": 4.69393894722015e-06}
{"train_loss": 0.052873872220516205, "global_step": 233780, "epoch": 2626, "lr": 4.693693711175545e-06}
{"train_loss": 0.07325144857168198, "global_step": 233781, "epoch": 2626, "lr": 4.693448481221812e-06}
{"train_loss": 0.06089594215154648, "global_step": 233782, "epoch": 2626, "lr": 4.693203257359003e-06}
{"train_loss": 0.06523218005895615, "global_step": 233783, "epoch": 2626, "lr": 4.692958039587131e-06}
{"train_loss": 0.03258592262864113, "global_step": 233784, "epoch": 2626, "lr": 4.692712827906237e-06}
{"train_loss": 0.05696406215429306, "global_step": 233785, "epoch": 2626, "lr": 4.692467622316371e-06}
{"train_loss": 0.03623545542359352, "global_step": 233786, "epoch": 2626, "lr": 4.692222422817538e-06}
{"train_loss": 0.038674257695674896, "global_step": 233787, "epoch": 2626, "lr": 4.691977229409788e-06}
{"train_loss": 0.08875375986099243, "global_step": 233788, "epoch": 2626, "lr": 4.691732042093144e-06}
{"train_loss": 0.06578133255243301, "global_step": 233789, "epoch": 2626, "lr": 4.691486860867656e-06}
{"train_loss": 0.08090681582689285, "global_step": 233790, "epoch": 2626, "lr": 4.691241685733333e-06}
{"train_loss": 0.02428983710706234, "global_step": 233791, "epoch": 2626, "lr": 4.690996516690227e-06}
{"train_loss": 0.045465800911188126, "global_step": 233792, "epoch": 2626, "lr": 4.6907513537383595e-06}
{"train_loss": 0.059249505400657654, "global_step": 233793, "epoch": 2626, "lr": 4.6905061968777754e-06}
{"train_loss": 0.05647578090429306, "global_step": 233794, "epoch": 2626, "lr": 4.690261046108496e-06}
{"train_loss": 0.044069331139326096, "global_step": 233795, "epoch": 2626, "lr": 4.690015901430561e-06}
{"train_loss": 0.050877418369054794, "global_step": 233796, "epoch": 2626, "lr": 4.689770762843998e-06}
{"train_loss": 0.09778732061386108, "global_step": 233797, "epoch": 2626, "lr": 4.689525630348845e-06}
{"train_loss": 0.034839797765016556, "global_step": 233798, "epoch": 2626, "lr": 4.6892805039451315e-06}
{"train_loss": 0.019031964242458344, "global_step": 233799, "epoch": 2626, "lr": 4.689035383632895e-06}
{"train_loss": 0.032432008534669876, "global_step": 233800, "epoch": 2626, "lr": 4.688790269412158e-06}
{"train_loss": 0.08931806683540344, "global_step": 233801, "epoch": 2626, "lr": 4.688545161282976e-06}
{"train_loss": 0.049205472305668206, "global_step": 233802, "epoch": 2626, "lr": 4.688300059245348e-06, "val_loss": 9.272358894348145}
{"train_loss": 0.02451453171670437, "global_step": 233803, "epoch": 2627, "lr": 4.688054963299332e-06}
{"train_loss": 0.04129326343536377, "global_step": 233804, "epoch": 2627, "lr": 4.687809873444965e-06}
{"train_loss": 0.03778570145368576, "global_step": 233805, "epoch": 2627, "lr": 4.687564789682259e-06}
{"train_loss": 0.01935366354882717, "global_step": 233806, "epoch": 2627, "lr": 4.6873197120112625e-06}
{"train_loss": 0.0236643236130476, "global_step": 233807, "epoch": 2627, "lr": 4.6870746404319986e-06}
{"train_loss": 0.036008257418870926, "global_step": 233808, "epoch": 2627, "lr": 4.6868295749445126e-06}
{"train_loss": 0.03344492241740227, "global_step": 233809, "epoch": 2627, "lr": 4.68658451554882e-06}
{"train_loss": 0.03857499733567238, "global_step": 233810, "epoch": 2627, "lr": 4.686339462244976e-06}
{"train_loss": 0.07404178380966187, "global_step": 233811, "epoch": 2627, "lr": 4.686094415032999e-06}
{"train_loss": 0.01632840745151043, "global_step": 233812, "epoch": 2627, "lr": 4.685849373912915e-06}
{"train_loss": 0.059610120952129364, "global_step": 233813, "epoch": 2627, "lr": 4.68560433888478e-06}
{"train_loss": 0.06048940494656563, "global_step": 233814, "epoch": 2627, "lr": 4.6853593099486e-06}
{"train_loss": 0.045105889439582825, "global_step": 233815, "epoch": 2627, "lr": 4.685114287104431e-06}
{"train_loss": 0.03444892540574074, "global_step": 233816, "epoch": 2627, "lr": 4.684869270352293e-06}
{"train_loss": 0.022059954702854156, "global_step": 233817, "epoch": 2627, "lr": 4.6846242596922165e-06}
{"train_loss": 0.03679512441158295, "global_step": 233818, "epoch": 2627, "lr": 4.684379255124249e-06}
{"train_loss": 0.04575708135962486, "global_step": 233819, "epoch": 2627, "lr": 4.684134256648404e-06}
{"train_loss": 0.044896211475133896, "global_step": 233820, "epoch": 2627, "lr": 4.683889264264734e-06}
{"train_loss": 0.055900998413562775, "global_step": 233821, "epoch": 2627, "lr": 4.683644277973259e-06}
{"train_loss": 0.05051141604781151, "global_step": 233822, "epoch": 2627, "lr": 4.6833992977740105e-06}
{"train_loss": 0.041996072977781296, "global_step": 233823, "epoch": 2627, "lr": 4.6831543236670324e-06}
{"train_loss": 0.050924286246299744, "global_step": 233824, "epoch": 2627, "lr": 4.682909355652348e-06}
{"train_loss": 0.04615074023604393, "global_step": 233825, "epoch": 2627, "lr": 4.682664393730002e-06}
{"train_loss": 0.06687166541814804, "global_step": 233826, "epoch": 2627, "lr": 4.682419437900009e-06}
{"train_loss": 0.04974082484841347, "global_step": 233827, "epoch": 2627, "lr": 4.6821744881624216e-06}
{"train_loss": 0.06391928344964981, "global_step": 233828, "epoch": 2627, "lr": 4.681929544517261e-06}
{"train_loss": 0.04084157943725586, "global_step": 233829, "epoch": 2627, "lr": 4.6816846069645645e-06}
{"train_loss": 0.03456857427954674, "global_step": 233830, "epoch": 2627, "lr": 4.681439675504356e-06}
{"train_loss": 0.05668359622359276, "global_step": 233831, "epoch": 2627, "lr": 4.681194750136681e-06}
{"train_loss": 0.04719257354736328, "global_step": 233832, "epoch": 2627, "lr": 4.680949830861564e-06}
{"train_loss": 0.03190306946635246, "global_step": 233833, "epoch": 2627, "lr": 4.680704917679047e-06}
{"train_loss": 0.04863307625055313, "global_step": 233834, "epoch": 2627, "lr": 4.6804600105891494e-06}
{"train_loss": 0.03390396758913994, "global_step": 233835, "epoch": 2627, "lr": 4.680215109591918e-06}
{"train_loss": 0.01845615729689598, "global_step": 233836, "epoch": 2627, "lr": 4.679970214687368e-06}
{"train_loss": 0.033521369099617004, "global_step": 233837, "epoch": 2627, "lr": 4.6797253258755616e-06}
{"train_loss": 0.056087642908096313, "global_step": 233838, "epoch": 2627, "lr": 4.679480443156498e-06}
{"train_loss": 0.08367890864610672, "global_step": 233839, "epoch": 2627, "lr": 4.679235566530227e-06}
{"train_loss": 0.04816483333706856, "global_step": 233840, "epoch": 2627, "lr": 4.678990695996788e-06}
{"train_loss": 0.06262295693159103, "global_step": 233841, "epoch": 2627, "lr": 4.678745831556197e-06}
{"train_loss": 0.04605884104967117, "global_step": 233842, "epoch": 2627, "lr": 4.678500973208511e-06}
{"train_loss": 0.04352324455976486, "global_step": 233843, "epoch": 2627, "lr": 4.6782561209537335e-06}
{"train_loss": 0.05600139498710632, "global_step": 233844, "epoch": 2627, "lr": 4.6780112747919216e-06}
{"train_loss": 0.05159692466259003, "global_step": 233845, "epoch": 2627, "lr": 4.677766434723091e-06}
{"train_loss": 0.08091451227664948, "global_step": 233846, "epoch": 2627, "lr": 4.677521600747292e-06}
{"train_loss": 0.031008988618850708, "global_step": 233847, "epoch": 2627, "lr": 4.677276772864542e-06}
{"train_loss": 0.0765315517783165, "global_step": 233848, "epoch": 2627, "lr": 4.677031951074884e-06}
{"train_loss": 0.061778485774993896, "global_step": 233849, "epoch": 2627, "lr": 4.676787135378335e-06}
{"train_loss": 0.03610831871628761, "global_step": 233850, "epoch": 2627, "lr": 4.676542325774952e-06}
{"train_loss": 0.0317949540913105, "global_step": 233851, "epoch": 2627, "lr": 4.676297522264755e-06}
{"train_loss": 0.020879531279206276, "global_step": 233852, "epoch": 2627, "lr": 4.676052724847773e-06}
{"train_loss": 0.06671778112649918, "global_step": 233853, "epoch": 2627, "lr": 4.675807933524046e-06}
{"train_loss": 0.05311625450849533, "global_step": 233854, "epoch": 2627, "lr": 4.675563148293599e-06}
{"train_loss": 0.02367376536130905, "global_step": 233855, "epoch": 2627, "lr": 4.675318369156478e-06}
{"train_loss": 0.028393063694238663, "global_step": 233856, "epoch": 2627, "lr": 4.675073596112706e-06}
{"train_loss": 0.03561238944530487, "global_step": 233857, "epoch": 2627, "lr": 4.6748288291623084e-06}
{"train_loss": 0.05346048250794411, "global_step": 233858, "epoch": 2627, "lr": 4.674584068305332e-06}
{"train_loss": 0.019712837412953377, "global_step": 233859, "epoch": 2627, "lr": 4.674339313541809e-06}
{"train_loss": 0.08422631025314331, "global_step": 233860, "epoch": 2627, "lr": 4.674094564871767e-06}
{"train_loss": 0.018161695450544357, "global_step": 233861, "epoch": 2627, "lr": 4.673849822295246e-06}
{"train_loss": 0.09127025306224823, "global_step": 233862, "epoch": 2627, "lr": 4.673605085812261e-06}
{"train_loss": 0.03945412114262581, "global_step": 233863, "epoch": 2627, "lr": 4.673360355422868e-06}
{"train_loss": 0.06305785477161407, "global_step": 233864, "epoch": 2627, "lr": 4.673115631127084e-06}
{"train_loss": 0.04989367350935936, "global_step": 233865, "epoch": 2627, "lr": 4.672870912924953e-06}
{"train_loss": 0.0882863849401474, "global_step": 233866, "epoch": 2627, "lr": 4.672626200816493e-06}
{"train_loss": 0.060798708349466324, "global_step": 233867, "epoch": 2627, "lr": 4.672381494801753e-06}
{"train_loss": 0.05094043165445328, "global_step": 233868, "epoch": 2627, "lr": 4.672136794880755e-06}
{"train_loss": 0.06483159959316254, "global_step": 233869, "epoch": 2627, "lr": 4.671892101053538e-06}
{"train_loss": 0.05120100826025009, "global_step": 233870, "epoch": 2627, "lr": 4.6716474133201296e-06}
{"train_loss": 0.056967318058013916, "global_step": 233871, "epoch": 2627, "lr": 4.671402731680575e-06}
{"train_loss": 0.06601128727197647, "global_step": 233872, "epoch": 2627, "lr": 4.671158056134883e-06}
{"train_loss": 0.01614844985306263, "global_step": 233873, "epoch": 2627, "lr": 4.670913386683118e-06}
{"train_loss": 0.059426382184028625, "global_step": 233874, "epoch": 2627, "lr": 4.670668723325278e-06}
{"train_loss": 0.05804559588432312, "global_step": 233875, "epoch": 2627, "lr": 4.670424066061418e-06}
{"train_loss": 0.03861682862043381, "global_step": 233876, "epoch": 2627, "lr": 4.670179414891579e-06}
{"train_loss": 0.05295008048415184, "global_step": 233877, "epoch": 2627, "lr": 4.66993476981577e-06}
{"train_loss": 0.0601290725171566, "global_step": 233878, "epoch": 2627, "lr": 4.669690130834048e-06}
{"train_loss": 0.057869575917720795, "global_step": 233879, "epoch": 2627, "lr": 4.669445497946423e-06}
{"train_loss": 0.05006031692028046, "global_step": 233880, "epoch": 2627, "lr": 4.669200871152946e-06}
{"train_loss": 0.054045140743255615, "global_step": 233881, "epoch": 2627, "lr": 4.668956250453632e-06}
{"train_loss": 0.038937851786613464, "global_step": 233882, "epoch": 2627, "lr": 4.668711635848538e-06}
{"train_loss": 0.05673334002494812, "global_step": 233883, "epoch": 2627, "lr": 4.668467027337675e-06}
{"train_loss": 0.017348799854516983, "global_step": 233884, "epoch": 2627, "lr": 4.668222424921087e-06}
{"train_loss": 0.02284921519458294, "global_step": 233885, "epoch": 2627, "lr": 4.667977828598796e-06}
{"train_loss": 0.03468283265829086, "global_step": 233886, "epoch": 2627, "lr": 4.667733238370858e-06}
{"train_loss": 0.021322468295693398, "global_step": 233887, "epoch": 2627, "lr": 4.667488654237279e-06}
{"train_loss": 0.03205497935414314, "global_step": 233888, "epoch": 2627, "lr": 4.667244076198108e-06}
{"train_loss": 0.0328064002096653, "global_step": 233889, "epoch": 2627, "lr": 4.6669995042533775e-06}
{"train_loss": 0.07072293758392334, "global_step": 233890, "epoch": 2627, "lr": 4.666754938403106e-06}
{"train_loss": 0.04745525060995911, "global_step": 233891, "epoch": 2627, "lr": 4.666510378647349e-06, "val_loss": 9.152883529663086}
{"train_loss": 0.03117508813738823, "global_step": 233892, "epoch": 2628, "lr": 4.666265824986116e-06}
{"train_loss": 0.08048024773597717, "global_step": 233893, "epoch": 2628, "lr": 4.666021277419458e-06}
{"train_loss": 0.028412193059921265, "global_step": 233894, "epoch": 2628, "lr": 4.665776735947391e-06}
{"train_loss": 0.06788761168718338, "global_step": 233895, "epoch": 2628, "lr": 4.665532200569972e-06}
{"train_loss": 0.0554337203502655, "global_step": 233896, "epoch": 2628, "lr": 4.665287671287205e-06}
{"train_loss": 0.029489729553461075, "global_step": 233897, "epoch": 2628, "lr": 4.665043148099152e-06}
{"train_loss": 0.046077124774456024, "global_step": 233898, "epoch": 2628, "lr": 4.664798631005818e-06}
{"train_loss": 0.0418989397585392, "global_step": 233899, "epoch": 2628, "lr": 4.664554120007264e-06}
{"train_loss": 0.03411567583680153, "global_step": 233900, "epoch": 2628, "lr": 4.664309615103491e-06}
{"train_loss": 0.06411413848400116, "global_step": 233901, "epoch": 2628, "lr": 4.664065116294563e-06}
{"train_loss": 0.06693229079246521, "global_step": 233902, "epoch": 2628, "lr": 4.6638206235804896e-06}
{"train_loss": 0.07286155968904495, "global_step": 233903, "epoch": 2628, "lr": 4.663576136961323e-06}
{"train_loss": 0.033495817333459854, "global_step": 233904, "epoch": 2628, "lr": 4.663331656437075e-06}
{"train_loss": 0.038792964071035385, "global_step": 233905, "epoch": 2628, "lr": 4.663087182007797e-06}
{"train_loss": 0.06605315953493118, "global_step": 233906, "epoch": 2628, "lr": 4.66284271367351e-06}
{"train_loss": 0.04498486965894699, "global_step": 233907, "epoch": 2628, "lr": 4.662598251434258e-06}
{"train_loss": 0.061611905694007874, "global_step": 233908, "epoch": 2628, "lr": 4.662353795290053e-06}
{"train_loss": 0.03631938248872757, "global_step": 233909, "epoch": 2628, "lr": 4.662109345240956e-06}
{"train_loss": 0.05331297963857651, "global_step": 233910, "epoch": 2628, "lr": 4.661864901286977e-06}
{"train_loss": 0.03551904857158661, "global_step": 233911, "epoch": 2628, "lr": 4.661620463428157e-06}
{"train_loss": 0.03840785101056099, "global_step": 233912, "epoch": 2628, "lr": 4.661376031664538e-06}
{"train_loss": 0.08435562252998352, "global_step": 233913, "epoch": 2628, "lr": 4.661131605996138e-06}
{"train_loss": 0.05736026540398598, "global_step": 233914, "epoch": 2628, "lr": 4.660887186423002e-06}
{"train_loss": 0.039967961609363556, "global_step": 233915, "epoch": 2628, "lr": 4.66064277294515e-06}
{"train_loss": 0.025102313607931137, "global_step": 233916, "epoch": 2628, "lr": 4.660398365562629e-06}
{"train_loss": 0.038202155381441116, "global_step": 233917, "epoch": 2628, "lr": 4.6601539642754595e-06}
{"train_loss": 0.03556644916534424, "global_step": 233918, "epoch": 2628, "lr": 4.659909569083687e-06}
{"train_loss": 0.05453761667013168, "global_step": 233919, "epoch": 2628, "lr": 4.659665179987327e-06}
{"train_loss": 0.029741421341896057, "global_step": 233920, "epoch": 2628, "lr": 4.65942079698643e-06}
{"train_loss": 0.04464978724718094, "global_step": 233921, "epoch": 2628, "lr": 4.659176420081013e-06}
{"train_loss": 0.018915681168437004, "global_step": 233922, "epoch": 2628, "lr": 4.658932049271131e-06}
{"train_loss": 0.03043275885283947, "global_step": 233923, "epoch": 2628, "lr": 4.658687684556789e-06}
{"train_loss": 0.051672641187906265, "global_step": 233924, "epoch": 2628, "lr": 4.658443325938045e-06}
{"train_loss": 0.03320779278874397, "global_step": 233925, "epoch": 2628, "lr": 4.6581989734149065e-06}
{"train_loss": 0.04548339173197746, "global_step": 233926, "epoch": 2628, "lr": 4.657954626987432e-06}
{"train_loss": 0.027657678350806236, "global_step": 233927, "epoch": 2628, "lr": 4.657710286655647e-06}
{"train_loss": 0.022130895406007767, "global_step": 233928, "epoch": 2628, "lr": 4.657465952419571e-06}
{"train_loss": 0.07588597387075424, "global_step": 233929, "epoch": 2628, "lr": 4.657221624279251e-06}
{"train_loss": 0.056686677038669586, "global_step": 233930, "epoch": 2628, "lr": 4.656977302234705e-06}
{"train_loss": 0.07021325826644897, "global_step": 233931, "epoch": 2628, "lr": 4.656732986285989e-06}
{"train_loss": 0.030212143436074257, "global_step": 233932, "epoch": 2628, "lr": 4.656488676433113e-06}
{"train_loss": 0.05115344375371933, "global_step": 233933, "epoch": 2628, "lr": 4.656244372676127e-06}
{"train_loss": 0.04627307504415512, "global_step": 233934, "epoch": 2628, "lr": 4.656000075015049e-06}
{"train_loss": 0.03375634178519249, "global_step": 233935, "epoch": 2628, "lr": 4.6557557834499225e-06}
{"train_loss": 0.053875990211963654, "global_step": 233936, "epoch": 2628, "lr": 4.655511497980775e-06}
{"train_loss": 0.037487614899873734, "global_step": 233937, "epoch": 2628, "lr": 4.655267218607645e-06}
{"train_loss": 0.05683485418558121, "global_step": 233938, "epoch": 2628, "lr": 4.655022945330556e-06}
{"train_loss": 0.04868133366107941, "global_step": 233939, "epoch": 2628, "lr": 4.654778678149557e-06}
{"train_loss": 0.08677893131971359, "global_step": 233940, "epoch": 2628, "lr": 4.65453441706466e-06}
{"train_loss": 0.06702302396297455, "global_step": 233941, "epoch": 2628, "lr": 4.654290162075914e-06}
{"train_loss": 0.03498917818069458, "global_step": 233942, "epoch": 2628, "lr": 4.654045913183336e-06}
{"train_loss": 0.04423423111438751, "global_step": 233943, "epoch": 2628, "lr": 4.653801670386981e-06}
{"train_loss": 0.057395488023757935, "global_step": 233944, "epoch": 2628, "lr": 4.653557433686861e-06}
{"train_loss": 0.0618993379175663, "global_step": 233945, "epoch": 2628, "lr": 4.653313203083026e-06}
{"train_loss": 0.03703618794679642, "global_step": 233946, "epoch": 2628, "lr": 4.653068978575492e-06}
{"train_loss": 0.06502925604581833, "global_step": 233947, "epoch": 2628, "lr": 4.65282476016431e-06}
{"train_loss": 0.04292971268296242, "global_step": 233948, "epoch": 2628, "lr": 4.652580547849494e-06}
{"train_loss": 0.02172040194272995, "global_step": 233949, "epoch": 2628, "lr": 4.652336341631081e-06}
{"train_loss": 0.0515817329287529, "global_step": 233950, "epoch": 2628, "lr": 4.652092141509124e-06}
{"train_loss": 0.03606384992599487, "global_step": 233951, "epoch": 2628, "lr": 4.651847947483628e-06}
{"train_loss": 0.0579342395067215, "global_step": 233952, "epoch": 2628, "lr": 4.651603759554646e-06}
{"train_loss": 0.031749527901411057, "global_step": 233953, "epoch": 2628, "lr": 4.6513595777221975e-06}
{"train_loss": 0.08058696240186691, "global_step": 233954, "epoch": 2628, "lr": 4.651115401986328e-06}
{"train_loss": 0.05254286900162697, "global_step": 233955, "epoch": 2628, "lr": 4.650871232347054e-06}
{"train_loss": 0.05382360890507698, "global_step": 233956, "epoch": 2628, "lr": 4.650627068804431e-06}
{"train_loss": 0.01004414539784193, "global_step": 233957, "epoch": 2628, "lr": 4.650382911358464e-06}
{"train_loss": 0.08294042944908142, "global_step": 233958, "epoch": 2628, "lr": 4.6501387600092095e-06}
{"train_loss": 0.0645756945014, "global_step": 233959, "epoch": 2628, "lr": 4.649894614756684e-06}
{"train_loss": 0.012019753456115723, "global_step": 233960, "epoch": 2628, "lr": 4.649650475600937e-06}
{"train_loss": 0.03506530821323395, "global_step": 233961, "epoch": 2628, "lr": 4.6494063425419844e-06}
{"train_loss": 0.03734761103987694, "global_step": 233962, "epoch": 2628, "lr": 4.649162215579872e-06}
{"train_loss": 0.03871310129761696, "global_step": 233963, "epoch": 2628, "lr": 4.648918094714622e-06}
{"train_loss": 0.054762646555900574, "global_step": 233964, "epoch": 2628, "lr": 4.648673979946278e-06}
{"train_loss": 0.04354938492178917, "global_step": 233965, "epoch": 2628, "lr": 4.648429871274856e-06}
{"train_loss": 0.06148982420563698, "global_step": 233966, "epoch": 2628, "lr": 4.648185768700414e-06}
{"train_loss": 0.021388011053204536, "global_step": 233967, "epoch": 2628, "lr": 4.6479416722229666e-06}
{"train_loss": 0.16663077473640442, "global_step": 233968, "epoch": 2628, "lr": 4.647697581842542e-06}
{"train_loss": 0.05840964615345001, "global_step": 233969, "epoch": 2628, "lr": 4.64745349755919e-06}
{"train_loss": 0.07145712524652481, "global_step": 233970, "epoch": 2628, "lr": 4.647209419372928e-06}
{"train_loss": 0.06057645007967949, "global_step": 233971, "epoch": 2628, "lr": 4.646965347283805e-06}
{"train_loss": 0.040931813418865204, "global_step": 233972, "epoch": 2628, "lr": 4.646721281291833e-06}
{"train_loss": 0.0324687622487545, "global_step": 233973, "epoch": 2628, "lr": 4.646477221397072e-06}
{"train_loss": 0.09577195346355438, "global_step": 233974, "epoch": 2628, "lr": 4.646233167599523e-06}
{"train_loss": 0.04609328880906105, "global_step": 233975, "epoch": 2628, "lr": 4.645989119899247e-06}
{"train_loss": 0.02760932967066765, "global_step": 233976, "epoch": 2628, "lr": 4.64574507829626e-06}
{"train_loss": 0.044220052659511566, "global_step": 233977, "epoch": 2628, "lr": 4.645501042790601e-06}
{"train_loss": 0.05625121295452118, "global_step": 233978, "epoch": 2628, "lr": 4.645257013382298e-06}
{"train_loss": 0.05081180855631828, "global_step": 233979, "epoch": 2628, "lr": 4.645012990071396e-06}
{"train_loss": 0.04893478308542726, "global_step": 233980, "epoch": 2628, "lr": 4.64476897285791e-06, "val_loss": 9.267863273620605}
{"train_loss": 0.08389513194561005, "global_step": 233981, "epoch": 2629, "lr": 4.644524961741892e-06}
{"train_loss": 0.04303348809480667, "global_step": 233982, "epoch": 2629, "lr": 4.644280956723351e-06}
{"train_loss": 0.041825905442237854, "global_step": 233983, "epoch": 2629, "lr": 4.644036957802344e-06}
{"train_loss": 0.04967350512742996, "global_step": 233984, "epoch": 2629, "lr": 4.643792964978888e-06}
{"train_loss": 0.06617331504821777, "global_step": 233985, "epoch": 2629, "lr": 4.643548978253021e-06}
{"train_loss": 0.029774703085422516, "global_step": 233986, "epoch": 2629, "lr": 4.643304997624781e-06}
{"train_loss": 0.021135030314326286, "global_step": 233987, "epoch": 2629, "lr": 4.643061023094192e-06}
{"train_loss": 0.047118235379457474, "global_step": 233988, "epoch": 2629, "lr": 4.642817054661297e-06}
{"train_loss": 0.02921299636363983, "global_step": 233989, "epoch": 2629, "lr": 4.642573092326113e-06}
{"train_loss": 0.021757671609520912, "global_step": 233990, "epoch": 2629, "lr": 4.642329136088691e-06}
{"train_loss": 0.09285573661327362, "global_step": 233991, "epoch": 2629, "lr": 4.642085185949047e-06}
{"train_loss": 0.05750574916601181, "global_step": 233992, "epoch": 2629, "lr": 4.64184124190723e-06}
{"train_loss": 0.03418879956007004, "global_step": 233993, "epoch": 2629, "lr": 4.6415973039632575e-06}
{"train_loss": 0.04084025323390961, "global_step": 233994, "epoch": 2629, "lr": 4.641353372117174e-06}
{"train_loss": 0.06920266896486282, "global_step": 233995, "epoch": 2629, "lr": 4.6411094463690015e-06}
{"train_loss": 0.049226175993680954, "global_step": 233996, "epoch": 2629, "lr": 4.640865526718791e-06}
{"train_loss": 0.035435158759355545, "global_step": 233997, "epoch": 2629, "lr": 4.640621613166552e-06}
{"train_loss": 0.02457350492477417, "global_step": 233998, "epoch": 2629, "lr": 4.640377705712334e-06}
{"train_loss": 0.03979113698005676, "global_step": 233999, "epoch": 2629, "lr": 4.640133804356156e-06}
{"train_loss": 0.03228340297937393, "global_step": 234000, "epoch": 2629, "lr": 4.639889909098072e-06}
{"train_loss": 0.020554671064019203, "global_step": 234001, "epoch": 2629, "lr": 4.639646019938093e-06}
{"train_loss": 0.06163211911916733, "global_step": 234002, "epoch": 2629, "lr": 4.639402136876269e-06}
{"train_loss": 0.034657254815101624, "global_step": 234003, "epoch": 2629, "lr": 4.639158259912613e-06}
{"train_loss": 0.06269189715385437, "global_step": 234004, "epoch": 2629, "lr": 4.638914389047177e-06}
{"train_loss": 0.08035766333341599, "global_step": 234005, "epoch": 2629, "lr": 4.638670524279992e-06}
{"train_loss": 0.04444526135921478, "global_step": 234006, "epoch": 2629, "lr": 4.638426665611073e-06}
{"train_loss": 0.03677491471171379, "global_step": 234007, "epoch": 2629, "lr": 4.6381828130404705e-06}
{"train_loss": 0.0669831857085228, "global_step": 234008, "epoch": 2629, "lr": 4.637938966568206e-06}
{"train_loss": 0.09226123243570328, "global_step": 234009, "epoch": 2629, "lr": 4.6376951261943305e-06}
{"train_loss": 0.038911089301109314, "global_step": 234010, "epoch": 2629, "lr": 4.637451291918848e-06}
{"train_loss": 0.036860425025224686, "global_step": 234011, "epoch": 2629, "lr": 4.637207463741822e-06}
{"train_loss": 0.05634721741080284, "global_step": 234012, "epoch": 2629, "lr": 4.636963641663256e-06}
{"train_loss": 0.05308399349451065, "global_step": 234013, "epoch": 2629, "lr": 4.636719825683206e-06}
{"train_loss": 0.043202005326747894, "global_step": 234014, "epoch": 2629, "lr": 4.636476015801694e-06}
{"train_loss": 0.017206504940986633, "global_step": 234015, "epoch": 2629, "lr": 4.636232212018759e-06}
{"train_loss": 0.028799930587410927, "global_step": 234016, "epoch": 2629, "lr": 4.635988414334419e-06}
{"train_loss": 0.0700133740901947, "global_step": 234017, "epoch": 2629, "lr": 4.635744622748734e-06}
{"train_loss": 0.033962614834308624, "global_step": 234018, "epoch": 2629, "lr": 4.635500837261703e-06}
{"train_loss": 0.06117253005504608, "global_step": 234019, "epoch": 2629, "lr": 4.635257057873387e-06}
{"train_loss": 0.01851460337638855, "global_step": 234020, "epoch": 2629, "lr": 4.6350132845838005e-06}
{"train_loss": 0.047407202422618866, "global_step": 234021, "epoch": 2629, "lr": 4.634769517392984e-06}
{"train_loss": 0.03877468407154083, "global_step": 234022, "epoch": 2629, "lr": 4.634525756300978e-06}
{"train_loss": 0.03510751947760582, "global_step": 234023, "epoch": 2629, "lr": 4.634282001307799e-06}
{"train_loss": 0.0220413226634264, "global_step": 234024, "epoch": 2629, "lr": 4.634038252413492e-06}
{"train_loss": 0.07915931940078735, "global_step": 234025, "epoch": 2629, "lr": 4.633794509618083e-06}
{"train_loss": 0.10767275094985962, "global_step": 234026, "epoch": 2629, "lr": 4.633550772921613e-06}
{"train_loss": 0.03238092362880707, "global_step": 234027, "epoch": 2629, "lr": 4.633307042324103e-06}
{"train_loss": 0.08196491748094559, "global_step": 234028, "epoch": 2629, "lr": 4.6330633178255975e-06}
{"train_loss": 0.05167021229863167, "global_step": 234029, "epoch": 2629, "lr": 4.632819599426114e-06}
{"train_loss": 0.03366014361381531, "global_step": 234030, "epoch": 2629, "lr": 4.632575887125706e-06}
{"train_loss": 0.05882064998149872, "global_step": 234031, "epoch": 2629, "lr": 4.632332180924387e-06}
{"train_loss": 0.04745527729392052, "global_step": 234032, "epoch": 2629, "lr": 4.632088480822205e-06}
{"train_loss": 0.07326582819223404, "global_step": 234033, "epoch": 2629, "lr": 4.631844786819178e-06}
{"train_loss": 0.05069785565137863, "global_step": 234034, "epoch": 2629, "lr": 4.631601098915356e-06}
{"train_loss": 0.0349726602435112, "global_step": 234035, "epoch": 2629, "lr": 4.631357417110754e-06}
{"train_loss": 0.09988735616207123, "global_step": 234036, "epoch": 2629, "lr": 4.631113741405418e-06}
{"train_loss": 0.06940009444952011, "global_step": 234037, "epoch": 2629, "lr": 4.63087007179937e-06}
{"train_loss": 0.03214822709560394, "global_step": 234038, "epoch": 2629, "lr": 4.6306264082926595e-06}
{"train_loss": 0.045030903071165085, "global_step": 234039, "epoch": 2629, "lr": 4.6303827508852986e-06}
{"train_loss": 0.04594955965876579, "global_step": 234040, "epoch": 2629, "lr": 4.6301390995773355e-06}
{"train_loss": 0.050927869975566864, "global_step": 234041, "epoch": 2629, "lr": 4.629895454368788e-06}
{"train_loss": 0.0775323137640953, "global_step": 234042, "epoch": 2629, "lr": 4.6296518152597114e-06}
{"train_loss": 0.015302647836506367, "global_step": 234043, "epoch": 2629, "lr": 4.629408182250122e-06}
{"train_loss": 0.037318456918001175, "global_step": 234044, "epoch": 2629, "lr": 4.629164555340049e-06}
{"train_loss": 0.033128995448350906, "global_step": 234045, "epoch": 2629, "lr": 4.628920934529535e-06}
{"train_loss": 0.04062597081065178, "global_step": 234046, "epoch": 2629, "lr": 4.628677319818608e-06}
{"train_loss": 0.04078575223684311, "global_step": 234047, "epoch": 2629, "lr": 4.628433711207308e-06}
{"train_loss": 0.040838032960891724, "global_step": 234048, "epoch": 2629, "lr": 4.628190108695651e-06}
{"train_loss": 0.05833829566836357, "global_step": 234049, "epoch": 2629, "lr": 4.627946512283693e-06}
{"train_loss": 0.057995960116386414, "global_step": 234050, "epoch": 2629, "lr": 4.627702921971444e-06}
{"train_loss": 0.009646251797676086, "global_step": 234051, "epoch": 2629, "lr": 4.627459337758955e-06}
{"train_loss": 0.02218843251466751, "global_step": 234052, "epoch": 2629, "lr": 4.627215759646248e-06}
{"train_loss": 0.046623215079307556, "global_step": 234053, "epoch": 2629, "lr": 4.62697218763336e-06}
{"train_loss": 0.038071777671575546, "global_step": 234054, "epoch": 2629, "lr": 4.6267286217203165e-06}
{"train_loss": 0.05176002159714699, "global_step": 234055, "epoch": 2629, "lr": 4.626485061907165e-06}
{"train_loss": 0.04096313565969467, "global_step": 234056, "epoch": 2629, "lr": 4.626241508193924e-06}
{"train_loss": 0.05179300904273987, "global_step": 234057, "epoch": 2629, "lr": 4.6259979605806306e-06}
{"train_loss": 0.05327369272708893, "global_step": 234058, "epoch": 2629, "lr": 4.6257544190673245e-06}
{"train_loss": 0.05225555971264839, "global_step": 234059, "epoch": 2629, "lr": 4.625510883654027e-06}
{"train_loss": 0.011239277198910713, "global_step": 234060, "epoch": 2629, "lr": 4.625267354340785e-06}
{"train_loss": 0.10007988661527634, "global_step": 234061, "epoch": 2629, "lr": 4.625023831127612e-06}
{"train_loss": 0.033974483609199524, "global_step": 234062, "epoch": 2629, "lr": 4.624780314014565e-06}
{"train_loss": 0.05026564374566078, "global_step": 234063, "epoch": 2629, "lr": 4.62453680300165e-06}
{"train_loss": 0.07044434547424316, "global_step": 234064, "epoch": 2629, "lr": 4.624293298088922e-06}
{"train_loss": 0.035020556300878525, "global_step": 234065, "epoch": 2629, "lr": 4.624049799276398e-06}
{"train_loss": 0.03931168466806412, "global_step": 234066, "epoch": 2629, "lr": 4.623806306564121e-06}
{"train_loss": 0.07159708440303802, "global_step": 234067, "epoch": 2629, "lr": 4.623562819952115e-06}
{"train_loss": 0.07037805765867233, "global_step": 234068, "epoch": 2629, "lr": 4.623319339440429e-06}
{"train_loss": 0.048270961175557604, "global_step": 234069, "epoch": 2629, "lr": 4.623075865029075e-06, "val_loss": 9.310434341430664}
{"train_loss": 0.03522152081131935, "global_step": 234070, "epoch": 2630, "lr": 4.622832396718102e-06}
{"train_loss": 0.02901490032672882, "global_step": 234071, "epoch": 2630, "lr": 4.622588934507527e-06}
{"train_loss": 0.0328228585422039, "global_step": 234072, "epoch": 2630, "lr": 4.622345478397405e-06}
{"train_loss": 0.054518502205610275, "global_step": 234073, "epoch": 2630, "lr": 4.622102028387742e-06}
{"train_loss": 0.028911316767334938, "global_step": 234074, "epoch": 2630, "lr": 4.621858584478594e-06}
{"train_loss": 0.0453498438000679, "global_step": 234075, "epoch": 2630, "lr": 4.621615146669977e-06}
{"train_loss": 0.02644660323858261, "global_step": 234076, "epoch": 2630, "lr": 4.6213717149619304e-06}
{"train_loss": 0.05229348689317703, "global_step": 234077, "epoch": 2630, "lr": 4.621128289354493e-06}
{"train_loss": 0.06173813343048096, "global_step": 234078, "epoch": 2630, "lr": 4.620884869847697e-06}
{"train_loss": 0.04164133220911026, "global_step": 234079, "epoch": 2630, "lr": 4.620641456441555e-06}
{"train_loss": 0.02106359601020813, "global_step": 234080, "epoch": 2630, "lr": 4.6203980491361265e-06}
{"train_loss": 0.02411169372498989, "global_step": 234081, "epoch": 2630, "lr": 4.620154647931424e-06}
{"train_loss": 0.06524309515953064, "global_step": 234082, "epoch": 2630, "lr": 4.619911252827502e-06}
{"train_loss": 0.049959588795900345, "global_step": 234083, "epoch": 2630, "lr": 4.619667863824373e-06}
{"train_loss": 0.03551845997571945, "global_step": 234084, "epoch": 2630, "lr": 4.619424480922069e-06}
{"train_loss": 0.04601658880710602, "global_step": 234085, "epoch": 2630, "lr": 4.61918110412064e-06}
{"train_loss": 0.041695043444633484, "global_step": 234086, "epoch": 2630, "lr": 4.618937733420103e-06}
{"train_loss": 0.034937866032123566, "global_step": 234087, "epoch": 2630, "lr": 4.618694368820503e-06}
{"train_loss": 0.05239172652363777, "global_step": 234088, "epoch": 2630, "lr": 4.618451010321856e-06}
{"train_loss": 0.017540713772177696, "global_step": 234089, "epoch": 2630, "lr": 4.618207657924217e-06}
{"train_loss": 0.041855279356241226, "global_step": 234090, "epoch": 2630, "lr": 4.6179643116275924e-06}
{"train_loss": 0.06121270731091499, "global_step": 234091, "epoch": 2630, "lr": 4.617720971432043e-06}
{"train_loss": 0.03555643931031227, "global_step": 234092, "epoch": 2630, "lr": 4.61747763733758e-06}
{"train_loss": 0.022210795432329178, "global_step": 234093, "epoch": 2630, "lr": 4.617234309344243e-06}
{"train_loss": 0.06673029810190201, "global_step": 234094, "epoch": 2630, "lr": 4.616990987452075e-06}
{"train_loss": 0.07451379299163818, "global_step": 234095, "epoch": 2630, "lr": 4.616747671661087e-06}
{"train_loss": 0.05957415699958801, "global_step": 234096, "epoch": 2630, "lr": 4.616504361971336e-06}
{"train_loss": 0.03952381759881973, "global_step": 234097, "epoch": 2630, "lr": 4.616261058382831e-06}
{"train_loss": 0.058474794030189514, "global_step": 234098, "epoch": 2630, "lr": 4.61601776089563e-06}
{"train_loss": 0.042527928948402405, "global_step": 234099, "epoch": 2630, "lr": 4.615774469509737e-06}
{"train_loss": 0.052348047494888306, "global_step": 234100, "epoch": 2630, "lr": 4.615531184225214e-06}
{"train_loss": 0.06895759701728821, "global_step": 234101, "epoch": 2630, "lr": 4.615287905042071e-06}
{"train_loss": 0.0667828842997551, "global_step": 234102, "epoch": 2630, "lr": 4.615044631960358e-06}
{"train_loss": 0.040239959955215454, "global_step": 234103, "epoch": 2630, "lr": 4.614801364980087e-06}
{"train_loss": 0.022079413756728172, "global_step": 234104, "epoch": 2630, "lr": 4.614558104101313e-06}
{"train_loss": 0.021628959104418755, "global_step": 234105, "epoch": 2630, "lr": 4.614314849324053e-06}
{"train_loss": 0.028488026931881905, "global_step": 234106, "epoch": 2630, "lr": 4.614071600648351e-06}
{"train_loss": 0.054857976734638214, "global_step": 234107, "epoch": 2630, "lr": 4.613828358074223e-06}
{"train_loss": 0.030474426224827766, "global_step": 234108, "epoch": 2630, "lr": 4.613585121601727e-06}
{"train_loss": 0.04038073495030403, "global_step": 234109, "epoch": 2630, "lr": 4.6133418912308715e-06}
{"train_loss": 0.0659814178943634, "global_step": 234110, "epoch": 2630, "lr": 4.613098666961702e-06}
{"train_loss": 0.05120781436562538, "global_step": 234111, "epoch": 2630, "lr": 4.6128554487942455e-06}
{"train_loss": 0.03534761816263199, "global_step": 234112, "epoch": 2630, "lr": 4.612612236728536e-06}
{"train_loss": 0.03419824317097664, "global_step": 234113, "epoch": 2630, "lr": 4.612369030764613e-06}
{"train_loss": 0.05778474360704422, "global_step": 234114, "epoch": 2630, "lr": 4.612125830902498e-06}
{"train_loss": 0.08851132541894913, "global_step": 234115, "epoch": 2630, "lr": 4.61188263714224e-06}
{"train_loss": 0.05900224298238754, "global_step": 234116, "epoch": 2630, "lr": 4.611639449483857e-06}
{"train_loss": 0.026722779497504234, "global_step": 234117, "epoch": 2630, "lr": 4.611396267927381e-06}
{"train_loss": 0.06984765082597733, "global_step": 234118, "epoch": 2630, "lr": 4.611153092472853e-06}
{"train_loss": 0.041810985654592514, "global_step": 234119, "epoch": 2630, "lr": 4.610909923120299e-06}
{"train_loss": 0.04244409129023552, "global_step": 234120, "epoch": 2630, "lr": 4.610666759869758e-06}
{"train_loss": 0.08149677515029907, "global_step": 234121, "epoch": 2630, "lr": 4.6104236027212644e-06}
{"train_loss": 0.06498479843139648, "global_step": 234122, "epoch": 2630, "lr": 4.6101804516748395e-06}
{"train_loss": 0.09871536493301392, "global_step": 234123, "epoch": 2630, "lr": 4.609937306730527e-06}
{"train_loss": 0.047266796231269836, "global_step": 234124, "epoch": 2630, "lr": 4.609694167888345e-06}
{"train_loss": 0.06196398288011551, "global_step": 234125, "epoch": 2630, "lr": 4.6094510351483485e-06}
{"train_loss": 0.07155583798885345, "global_step": 234126, "epoch": 2630, "lr": 4.609207908510549e-06}
{"train_loss": 0.06027887389063835, "global_step": 234127, "epoch": 2630, "lr": 4.608964787974996e-06}
{"train_loss": 0.061993204057216644, "global_step": 234128, "epoch": 2630, "lr": 4.608721673541705e-06}
{"train_loss": 0.03699880093336105, "global_step": 234129, "epoch": 2630, "lr": 4.608478565210722e-06}
{"train_loss": 0.01211679819971323, "global_step": 234130, "epoch": 2630, "lr": 4.60823546298208e-06}
{"train_loss": 0.06384928524494171, "global_step": 234131, "epoch": 2630, "lr": 4.607992366855801e-06}
{"train_loss": 0.040483780205249786, "global_step": 234132, "epoch": 2630, "lr": 4.607749276831935e-06}
{"train_loss": 0.08290500193834305, "global_step": 234133, "epoch": 2630, "lr": 4.6075061929104875e-06}
{"train_loss": 0.04239422082901001, "global_step": 234134, "epoch": 2630, "lr": 4.607263115091525e-06}
{"train_loss": 0.06134345009922981, "global_step": 234135, "epoch": 2630, "lr": 4.6070200433750485e-06}
{"train_loss": 0.025138117372989655, "global_step": 234136, "epoch": 2630, "lr": 4.606776977761118e-06}
{"train_loss": 0.018402811139822006, "global_step": 234137, "epoch": 2630, "lr": 4.6065339182497395e-06}
{"train_loss": 0.05149725079536438, "global_step": 234138, "epoch": 2630, "lr": 4.6062908648409734e-06}
{"train_loss": 0.026328949257731438, "global_step": 234139, "epoch": 2630, "lr": 4.606047817534826e-06}
{"train_loss": 0.06960050016641617, "global_step": 234140, "epoch": 2630, "lr": 4.605804776331352e-06}
{"train_loss": 0.03781312704086304, "global_step": 234141, "epoch": 2630, "lr": 4.605561741230563e-06}
{"train_loss": 0.05275624245405197, "global_step": 234142, "epoch": 2630, "lr": 4.605318712232515e-06}
{"train_loss": 0.08033432066440582, "global_step": 234143, "epoch": 2630, "lr": 4.605075689337213e-06}
{"train_loss": 0.038115017116069794, "global_step": 234144, "epoch": 2630, "lr": 4.604832672544723e-06}
{"train_loss": 0.048131063580513, "global_step": 234145, "epoch": 2630, "lr": 4.604589661855046e-06}
{"train_loss": 0.0269753560423851, "global_step": 234146, "epoch": 2630, "lr": 4.604346657268238e-06}
{"train_loss": 0.0337059423327446, "global_step": 234147, "epoch": 2630, "lr": 4.604103658784314e-06}
{"train_loss": 0.05872469022870064, "global_step": 234148, "epoch": 2630, "lr": 4.603860666403314e-06}
{"train_loss": 0.07193881273269653, "global_step": 234149, "epoch": 2630, "lr": 4.603617680125283e-06}
{"train_loss": 0.034227970987558365, "global_step": 234150, "epoch": 2630, "lr": 4.603374699950236e-06}
{"train_loss": 0.015081284567713737, "global_step": 234151, "epoch": 2630, "lr": 4.603131725878218e-06}
{"train_loss": 0.039027780294418335, "global_step": 234152, "epoch": 2630, "lr": 4.602888757909246e-06}
{"train_loss": 0.0400955006480217, "global_step": 234153, "epoch": 2630, "lr": 4.60264579604337e-06}
{"train_loss": 0.05141802132129669, "global_step": 234154, "epoch": 2630, "lr": 4.602402840280617e-06}
{"train_loss": 0.052020683884620667, "global_step": 234155, "epoch": 2630, "lr": 4.602159890621005e-06}
{"train_loss": 0.040179844945669174, "global_step": 234156, "epoch": 2630, "lr": 4.6019169470645936e-06}
{"train_loss": 0.043808914721012115, "global_step": 234157, "epoch": 2630, "lr": 4.60167400961139e-06}
{"train_loss": 0.04703954742237758, "global_step": 234158, "epoch": 2630, "lr": 4.601431078261443e-06, "val_loss": 9.283440589904785, "train_action_mse_error": 13.850539207458496}
{"train_loss": 0.014330249279737473, "global_step": 234159, "epoch": 2631, "lr": 4.601188153014785e-06}
{"train_loss": 0.04631654918193817, "global_step": 234160, "epoch": 2631, "lr": 4.600945233871435e-06}
{"train_loss": 0.04471304267644882, "global_step": 234161, "epoch": 2631, "lr": 4.6007023208314406e-06}
{"train_loss": 0.024228554219007492, "global_step": 234162, "epoch": 2631, "lr": 4.6004594138948195e-06}
{"train_loss": 0.10204075276851654, "global_step": 234163, "epoch": 2631, "lr": 4.600216513061628e-06}
{"train_loss": 0.050814855843782425, "global_step": 234164, "epoch": 2631, "lr": 4.599973618331871e-06}
{"train_loss": 0.042106784880161285, "global_step": 234165, "epoch": 2631, "lr": 4.599730729705592e-06}
{"train_loss": 0.049301330000162125, "global_step": 234166, "epoch": 2631, "lr": 4.599487847182837e-06}
{"train_loss": 0.04205867275595665, "global_step": 234167, "epoch": 2631, "lr": 4.5992449707636225e-06}
{"train_loss": 0.041894927620887756, "global_step": 234168, "epoch": 2631, "lr": 4.599002100447991e-06}
{"train_loss": 0.04448440670967102, "global_step": 234169, "epoch": 2631, "lr": 4.598759236235961e-06}
{"train_loss": 0.08315736800432205, "global_step": 234170, "epoch": 2631, "lr": 4.598516378127582e-06}
{"train_loss": 0.0315435416996479, "global_step": 234171, "epoch": 2631, "lr": 4.59827352612287e-06}
{"train_loss": 0.04700179398059845, "global_step": 234172, "epoch": 2631, "lr": 4.598030680221882e-06}
{"train_loss": 0.029998822137713432, "global_step": 234173, "epoch": 2631, "lr": 4.597787840424622e-06}
{"train_loss": 0.06152857840061188, "global_step": 234174, "epoch": 2631, "lr": 4.597545006731141e-06}
{"train_loss": 0.037182170897722244, "global_step": 234175, "epoch": 2631, "lr": 4.597302179141466e-06}
{"train_loss": 0.04671596363186836, "global_step": 234176, "epoch": 2631, "lr": 4.597059357655631e-06}
{"train_loss": 0.02186824008822441, "global_step": 234177, "epoch": 2631, "lr": 4.5968165422736635e-06}
{"train_loss": 0.027668628841638565, "global_step": 234178, "epoch": 2631, "lr": 4.596573732995613e-06}
{"train_loss": 0.029841536656022072, "global_step": 234179, "epoch": 2631, "lr": 4.596330929821485e-06}
{"train_loss": 0.027305057272315025, "global_step": 234180, "epoch": 2631, "lr": 4.596088132751342e-06}
{"train_loss": 0.011900616809725761, "global_step": 234181, "epoch": 2631, "lr": 4.595845341785188e-06}
{"train_loss": 0.02294924110174179, "global_step": 234182, "epoch": 2631, "lr": 4.595602556923079e-06}
{"train_loss": 0.05501798167824745, "global_step": 234183, "epoch": 2631, "lr": 4.595359778165026e-06}
{"train_loss": 0.09481105208396912, "global_step": 234184, "epoch": 2631, "lr": 4.5951170055110734e-06}
{"train_loss": 0.04108107089996338, "global_step": 234185, "epoch": 2631, "lr": 4.594874238961266e-06}
{"train_loss": 0.07807538658380508, "global_step": 234186, "epoch": 2631, "lr": 4.594631478515615e-06}
{"train_loss": 0.0294604804366827, "global_step": 234187, "epoch": 2631, "lr": 4.594388724174176e-06}
{"train_loss": 0.0588064081966877, "global_step": 234188, "epoch": 2631, "lr": 4.594145975936953e-06}
{"train_loss": 0.037683963775634766, "global_step": 234189, "epoch": 2631, "lr": 4.593903233804003e-06}
{"train_loss": 0.024207904934883118, "global_step": 234190, "epoch": 2631, "lr": 4.5936604977753425e-06}
{"train_loss": 0.05522274598479271, "global_step": 234191, "epoch": 2631, "lr": 4.593417767851016e-06}
{"train_loss": 0.08300416171550751, "global_step": 234192, "epoch": 2631, "lr": 4.593175044031045e-06}
{"train_loss": 0.04103843867778778, "global_step": 234193, "epoch": 2631, "lr": 4.59293232631548e-06}
{"train_loss": 0.04816783219575882, "global_step": 234194, "epoch": 2631, "lr": 4.5926896147043374e-06}
{"train_loss": 0.05288807675242424, "global_step": 234195, "epoch": 2631, "lr": 4.592446909197645e-06}
{"train_loss": 0.00882682204246521, "global_step": 234196, "epoch": 2631, "lr": 4.592204209795459e-06}
{"train_loss": 0.03397019952535629, "global_step": 234197, "epoch": 2631, "lr": 4.5919615164977835e-06}
{"train_loss": 0.034430988132953644, "global_step": 234198, "epoch": 2631, "lr": 4.591718829304675e-06}
{"train_loss": 0.07054171711206436, "global_step": 234199, "epoch": 2631, "lr": 4.591476148216161e-06}
{"train_loss": 0.030914220958948135, "global_step": 234200, "epoch": 2631, "lr": 4.591233473232254e-06}
{"train_loss": 0.06477011740207672, "global_step": 234201, "epoch": 2631, "lr": 4.590990804353018e-06}
{"train_loss": 0.0407533273100853, "global_step": 234202, "epoch": 2631, "lr": 4.590748141578455e-06}
{"train_loss": 0.020780961960554123, "global_step": 234203, "epoch": 2631, "lr": 4.5905054849086134e-06}
{"train_loss": 0.038901425898075104, "global_step": 234204, "epoch": 2631, "lr": 4.590262834343534e-06}
{"train_loss": 0.02556607313454151, "global_step": 234205, "epoch": 2631, "lr": 4.590020189883232e-06}
{"train_loss": 0.06764866411685944, "global_step": 234206, "epoch": 2631, "lr": 4.5897775515277575e-06}
{"train_loss": 0.05268889665603638, "global_step": 234207, "epoch": 2631, "lr": 4.589534919277127e-06}
{"train_loss": 0.03741617128252983, "global_step": 234208, "epoch": 2631, "lr": 4.589292293131387e-06}
{"train_loss": 0.021166741847991943, "global_step": 234209, "epoch": 2631, "lr": 4.589049673090551e-06}
{"train_loss": 0.0372546911239624, "global_step": 234210, "epoch": 2631, "lr": 4.588807059154676e-06}
{"train_loss": 0.05728445202112198, "global_step": 234211, "epoch": 2631, "lr": 4.588564451323774e-06}
{"train_loss": 0.058537159115076065, "global_step": 234212, "epoch": 2631, "lr": 4.588321849597898e-06}
{"train_loss": 0.052108362317085266, "global_step": 234213, "epoch": 2631, "lr": 4.588079253977057e-06}
{"train_loss": 0.01702025718986988, "global_step": 234214, "epoch": 2631, "lr": 4.587836664461298e-06}
{"train_loss": 0.02759825624525547, "global_step": 234215, "epoch": 2631, "lr": 4.58759408105065e-06}
{"train_loss": 0.03999298810958862, "global_step": 234216, "epoch": 2631, "lr": 4.587351503745152e-06}
{"train_loss": 0.05451458692550659, "global_step": 234217, "epoch": 2631, "lr": 4.58710893254482e-06}
{"train_loss": 0.05598720535635948, "global_step": 234218, "epoch": 2631, "lr": 4.586866367449711e-06}
{"train_loss": 0.07895661890506744, "global_step": 234219, "epoch": 2631, "lr": 4.586623808459833e-06}
{"train_loss": 0.07693828642368317, "global_step": 234220, "epoch": 2631, "lr": 4.586381255575234e-06}
{"train_loss": 0.06176172196865082, "global_step": 234221, "epoch": 2631, "lr": 4.586138708795945e-06}
{"train_loss": 0.039113618433475494, "global_step": 234222, "epoch": 2631, "lr": 4.585896168121989e-06}
{"train_loss": 0.05260617285966873, "global_step": 234223, "epoch": 2631, "lr": 4.5856536335534215e-06}
{"train_loss": 0.047658875584602356, "global_step": 234224, "epoch": 2631, "lr": 4.585411105090243e-06}
{"train_loss": 0.04231572151184082, "global_step": 234225, "epoch": 2631, "lr": 4.585168582732513e-06}
{"train_loss": 0.027952654287219048, "global_step": 234226, "epoch": 2631, "lr": 4.584926066480244e-06}
{"train_loss": 0.04654194042086601, "global_step": 234227, "epoch": 2631, "lr": 4.584683556333491e-06}
{"train_loss": 0.030492577701807022, "global_step": 234228, "epoch": 2631, "lr": 4.58444105229226e-06}
{"train_loss": 0.04362523928284645, "global_step": 234229, "epoch": 2631, "lr": 4.584198554356606e-06}
{"train_loss": 0.03220057860016823, "global_step": 234230, "epoch": 2631, "lr": 4.583956062526545e-06}
{"train_loss": 0.029156353324651718, "global_step": 234231, "epoch": 2631, "lr": 4.583713576802129e-06}
{"train_loss": 0.034596044570207596, "global_step": 234232, "epoch": 2631, "lr": 4.583471097183378e-06}
{"train_loss": 0.03195765241980553, "global_step": 234233, "epoch": 2631, "lr": 4.583228623670316e-06}
{"train_loss": 0.08619587123394012, "global_step": 234234, "epoch": 2631, "lr": 4.582986156262997e-06}
{"train_loss": 0.063616082072258, "global_step": 234235, "epoch": 2631, "lr": 4.582743694961434e-06}
{"train_loss": 0.03858882561326027, "global_step": 234236, "epoch": 2631, "lr": 4.582501239765668e-06}
{"train_loss": 0.04154162108898163, "global_step": 234237, "epoch": 2631, "lr": 4.582258790675737e-06}
{"train_loss": 0.0227708350867033, "global_step": 234238, "epoch": 2631, "lr": 4.582016347691659e-06}
{"train_loss": 0.02186404913663864, "global_step": 234239, "epoch": 2631, "lr": 4.581773910813475e-06}
{"train_loss": 0.03597363084554672, "global_step": 234240, "epoch": 2631, "lr": 4.581531480041229e-06}
{"train_loss": 0.07905644178390503, "global_step": 234241, "epoch": 2631, "lr": 4.581289055374932e-06}
{"train_loss": 0.0192672498524189, "global_step": 234242, "epoch": 2631, "lr": 4.581046636814634e-06}
{"train_loss": 0.07364434748888016, "global_step": 234243, "epoch": 2631, "lr": 4.580804224360352e-06}
{"train_loss": 0.07189210504293442, "global_step": 234244, "epoch": 2631, "lr": 4.5805618180121355e-06}
{"train_loss": 0.07097114622592926, "global_step": 234245, "epoch": 2631, "lr": 4.580319417770002e-06}
{"train_loss": 0.04930596053600311, "global_step": 234246, "epoch": 2631, "lr": 4.580077023634e-06}
{"train_loss": 0.04504454709338338, "global_step": 234247, "epoch": 2631, "lr": 4.579834635604146e-06, "val_loss": 9.27955436706543}
{"train_loss": 0.03888968750834465, "global_step": 234248, "epoch": 2632, "lr": 4.579592253680487e-06}
{"train_loss": 0.04923584312200546, "global_step": 234249, "epoch": 2632, "lr": 4.5793498778630375e-06}
{"train_loss": 0.03969903290271759, "global_step": 234250, "epoch": 2632, "lr": 4.579107508151847e-06}
{"train_loss": 0.01124211773276329, "global_step": 234251, "epoch": 2632, "lr": 4.57886514454694e-06}
{"train_loss": 0.03259967267513275, "global_step": 234252, "epoch": 2632, "lr": 4.578622787048353e-06}
{"train_loss": 0.08590014278888702, "global_step": 234253, "epoch": 2632, "lr": 4.5783804356561155e-06}
{"train_loss": 0.0669289082288742, "global_step": 234254, "epoch": 2632, "lr": 4.578138090370265e-06}
{"train_loss": 0.0535321943461895, "global_step": 234255, "epoch": 2632, "lr": 4.57789575119082e-06}
{"train_loss": 0.03596387803554535, "global_step": 234256, "epoch": 2632, "lr": 4.577653418117828e-06}
{"train_loss": 0.07212658226490021, "global_step": 234257, "epoch": 2632, "lr": 4.577411091151323e-06}
{"train_loss": 0.06265035271644592, "global_step": 234258, "epoch": 2632, "lr": 4.577168770291323e-06}
{"train_loss": 0.0634467676281929, "global_step": 234259, "epoch": 2632, "lr": 4.576926455537878e-06}
{"train_loss": 0.04435935989022255, "global_step": 234260, "epoch": 2632, "lr": 4.576684146891003e-06}
{"train_loss": 0.0639878511428833, "global_step": 234261, "epoch": 2632, "lr": 4.5764418443507425e-06}
{"train_loss": 0.0611715093255043, "global_step": 234262, "epoch": 2632, "lr": 4.576199547917126e-06}
{"train_loss": 0.05786256864666939, "global_step": 234263, "epoch": 2632, "lr": 4.575957257590191e-06}
{"train_loss": 0.05592026934027672, "global_step": 234264, "epoch": 2632, "lr": 4.5757149733699544e-06}
{"train_loss": 0.06472120434045792, "global_step": 234265, "epoch": 2632, "lr": 4.5754726952564666e-06}
{"train_loss": 0.053908683359622955, "global_step": 234266, "epoch": 2632, "lr": 4.575230423249749e-06}
{"train_loss": 0.07051655650138855, "global_step": 234267, "epoch": 2632, "lr": 4.574988157349846e-06}
{"train_loss": 0.023538697510957718, "global_step": 234268, "epoch": 2632, "lr": 4.57474589755677e-06}
{"train_loss": 0.026562392711639404, "global_step": 234269, "epoch": 2632, "lr": 4.574503643870576e-06}
{"train_loss": 0.03431318700313568, "global_step": 234270, "epoch": 2632, "lr": 4.574261396291291e-06}
{"train_loss": 0.06678204983472824, "global_step": 234271, "epoch": 2632, "lr": 4.574019154818926e-06}
{"train_loss": 0.010033232159912586, "global_step": 234272, "epoch": 2632, "lr": 4.573776919453543e-06}
{"train_loss": 0.06390325725078583, "global_step": 234273, "epoch": 2632, "lr": 4.573534690195153e-06}
{"train_loss": 0.04351726174354553, "global_step": 234274, "epoch": 2632, "lr": 4.57329246704381e-06}
{"train_loss": 0.01680661179125309, "global_step": 234275, "epoch": 2632, "lr": 4.573050249999522e-06}
{"train_loss": 0.045112304389476776, "global_step": 234276, "epoch": 2632, "lr": 4.572808039062343e-06}
{"train_loss": 0.08015348762273788, "global_step": 234277, "epoch": 2632, "lr": 4.572565834232284e-06}
{"train_loss": 0.052572157233953476, "global_step": 234278, "epoch": 2632, "lr": 4.572323635509401e-06}
{"train_loss": 0.051393575966358185, "global_step": 234279, "epoch": 2632, "lr": 4.572081442893711e-06}
{"train_loss": 0.04201542213559151, "global_step": 234280, "epoch": 2632, "lr": 4.571839256385258e-06}
{"train_loss": 0.041333265602588654, "global_step": 234281, "epoch": 2632, "lr": 4.5715970759840525e-06}
{"train_loss": 0.049139607697725296, "global_step": 234282, "epoch": 2632, "lr": 4.571354901690155e-06}
{"train_loss": 0.029520981013774872, "global_step": 234283, "epoch": 2632, "lr": 4.571112733503579e-06}
{"train_loss": 0.035278018563985825, "global_step": 234284, "epoch": 2632, "lr": 4.570870571424368e-06}
{"train_loss": 0.02992122247815132, "global_step": 234285, "epoch": 2632, "lr": 4.570628415452544e-06}
{"train_loss": 0.025057462975382805, "global_step": 234286, "epoch": 2632, "lr": 4.57038626558815e-06}
{"train_loss": 0.02752470038831234, "global_step": 234287, "epoch": 2632, "lr": 4.57014412183121e-06}
{"train_loss": 0.031316302716732025, "global_step": 234288, "epoch": 2632, "lr": 4.569901984181768e-06}
{"train_loss": 0.05187496542930603, "global_step": 234289, "epoch": 2632, "lr": 4.56965985263984e-06}
{"train_loss": 0.03744706138968468, "global_step": 234290, "epoch": 2632, "lr": 4.569417727205472e-06}
{"train_loss": 0.030047373846173286, "global_step": 234291, "epoch": 2632, "lr": 4.5691756078786895e-06}
{"train_loss": 0.045069579035043716, "global_step": 234292, "epoch": 2632, "lr": 4.568933494659522e-06}
{"train_loss": 0.06269824504852295, "global_step": 234293, "epoch": 2632, "lr": 4.568691387548019e-06}
{"train_loss": 0.05097144842147827, "global_step": 234294, "epoch": 2632, "lr": 4.568449286544196e-06}
{"train_loss": 0.03637068718671799, "global_step": 234295, "epoch": 2632, "lr": 4.568207191648094e-06}
{"train_loss": 0.0768376961350441, "global_step": 234296, "epoch": 2632, "lr": 4.567965102859739e-06}
{"train_loss": 0.021385585889220238, "global_step": 234297, "epoch": 2632, "lr": 4.567723020179177e-06}
{"train_loss": 0.05263914167881012, "global_step": 234298, "epoch": 2632, "lr": 4.567480943606417e-06}
{"train_loss": 0.023439357057213783, "global_step": 234299, "epoch": 2632, "lr": 4.567238873141516e-06}
{"train_loss": 0.05414377525448799, "global_step": 234300, "epoch": 2632, "lr": 4.5669968087844905e-06}
{"train_loss": 0.05326154828071594, "global_step": 234301, "epoch": 2632, "lr": 4.5667547505353846e-06}
{"train_loss": 0.04251732677221298, "global_step": 234302, "epoch": 2632, "lr": 4.566512698394215e-06}
{"train_loss": 0.13375085592269897, "global_step": 234303, "epoch": 2632, "lr": 4.566270652361032e-06}
{"train_loss": 0.0417393259704113, "global_step": 234304, "epoch": 2632, "lr": 4.566028612435852e-06}
{"train_loss": 0.07677917927503586, "global_step": 234305, "epoch": 2632, "lr": 4.56578657861873e-06}
{"train_loss": 0.07039087265729904, "global_step": 234306, "epoch": 2632, "lr": 4.565544550909667e-06}
{"train_loss": 0.02828967012465, "global_step": 234307, "epoch": 2632, "lr": 4.5653025293087294e-06}
{"train_loss": 0.01816495880484581, "global_step": 234308, "epoch": 2632, "lr": 4.565060513815922e-06}
{"train_loss": 0.06838453561067581, "global_step": 234309, "epoch": 2632, "lr": 4.564818504431295e-06}
{"train_loss": 0.02781849540770054, "global_step": 234310, "epoch": 2632, "lr": 4.564576501154871e-06}
{"train_loss": 0.05358818545937538, "global_step": 234311, "epoch": 2632, "lr": 4.564334503986683e-06}
{"train_loss": 0.03955870494246483, "global_step": 234312, "epoch": 2632, "lr": 4.564092512926776e-06}
{"train_loss": 0.05098835378885269, "global_step": 234313, "epoch": 2632, "lr": 4.5638505279751605e-06}
{"train_loss": 0.028993235900998116, "global_step": 234314, "epoch": 2632, "lr": 4.5636085491318926e-06}
{"train_loss": 0.036420632153749466, "global_step": 234315, "epoch": 2632, "lr": 4.563366576396982e-06}
{"train_loss": 0.06536785513162613, "global_step": 234316, "epoch": 2632, "lr": 4.563124609770486e-06}
{"train_loss": 0.03659283369779587, "global_step": 234317, "epoch": 2632, "lr": 4.562882649252409e-06}
{"train_loss": 0.06459955126047134, "global_step": 234318, "epoch": 2632, "lr": 4.562640694842813e-06}
{"train_loss": 0.06645897775888443, "global_step": 234319, "epoch": 2632, "lr": 4.562398746541707e-06}
{"train_loss": 0.06695516407489777, "global_step": 234320, "epoch": 2632, "lr": 4.562156804349138e-06}
{"train_loss": 0.03385142236948013, "global_step": 234321, "epoch": 2632, "lr": 4.561914868265127e-06}
{"train_loss": 0.03102615475654602, "global_step": 234322, "epoch": 2632, "lr": 4.561672938289724e-06}
{"train_loss": 0.007470687385648489, "global_step": 234323, "epoch": 2632, "lr": 4.56143101442294e-06}
{"train_loss": 0.06699386984109879, "global_step": 234324, "epoch": 2632, "lr": 4.561189096664825e-06}
{"train_loss": 0.049870725721120834, "global_step": 234325, "epoch": 2632, "lr": 4.560947185015396e-06}
{"train_loss": 0.06554091721773148, "global_step": 234326, "epoch": 2632, "lr": 4.560705279474703e-06}
{"train_loss": 0.05993880331516266, "global_step": 234327, "epoch": 2632, "lr": 4.560463380042757e-06}
{"train_loss": 0.1086149588227272, "global_step": 234328, "epoch": 2632, "lr": 4.560221486719612e-06}
{"train_loss": 0.05563963204622269, "global_step": 234329, "epoch": 2632, "lr": 4.559979599505287e-06}
{"train_loss": 0.06241096183657646, "global_step": 234330, "epoch": 2632, "lr": 4.55973771839982e-06}
{"train_loss": 0.08255168050527573, "global_step": 234331, "epoch": 2632, "lr": 4.559495843403244e-06}
{"train_loss": 0.035948313772678375, "global_step": 234332, "epoch": 2632, "lr": 4.559253974515587e-06}
{"train_loss": 0.04242816939949989, "global_step": 234333, "epoch": 2632, "lr": 4.5590121117368936e-06}
{"train_loss": 0.034388937056064606, "global_step": 234334, "epoch": 2632, "lr": 4.5587702550671744e-06}
{"train_loss": 0.08865893632173538, "global_step": 234335, "epoch": 2632, "lr": 4.558528404506485e-06}
{"train_loss": 0.049611757692535605, "global_step": 234336, "epoch": 2632, "lr": 4.5582865600548416e-06, "val_loss": 9.3605318069458}
{"train_loss": 0.03910365328192711, "global_step": 234337, "epoch": 2633, "lr": 4.5580447217122905e-06}
{"train_loss": 0.042826227843761444, "global_step": 234338, "epoch": 2633, "lr": 4.557802889478846e-06}
{"train_loss": 0.03550746664404869, "global_step": 234339, "epoch": 2633, "lr": 4.55756106335456e-06}
{"train_loss": 0.022146930918097496, "global_step": 234340, "epoch": 2633, "lr": 4.557319243339448e-06}
{"train_loss": 0.03613097593188286, "global_step": 234341, "epoch": 2633, "lr": 4.557077429433559e-06}
{"train_loss": 0.019367339089512825, "global_step": 234342, "epoch": 2633, "lr": 4.556835621636907e-06}
{"train_loss": 0.042438969016075134, "global_step": 234343, "epoch": 2633, "lr": 4.556593819949545e-06}
{"train_loss": 0.029961179941892624, "global_step": 234344, "epoch": 2633, "lr": 4.5563520243714855e-06}
{"train_loss": 0.03218471258878708, "global_step": 234345, "epoch": 2633, "lr": 4.556110234902777e-06}
{"train_loss": 0.01607426442205906, "global_step": 234346, "epoch": 2633, "lr": 4.5558684515434435e-06}
{"train_loss": 0.04497385770082474, "global_step": 234347, "epoch": 2633, "lr": 4.555626674293522e-06}
{"train_loss": 0.05060049518942833, "global_step": 234348, "epoch": 2633, "lr": 4.555384903153048e-06}
{"train_loss": 0.056789420545101166, "global_step": 234349, "epoch": 2633, "lr": 4.5551431381220365e-06}
{"train_loss": 0.05167020857334137, "global_step": 234350, "epoch": 2633, "lr": 4.554901379200538e-06}
{"train_loss": 0.04709889739751816, "global_step": 234351, "epoch": 2633, "lr": 4.554659626388574e-06}
{"train_loss": 0.02820153348147869, "global_step": 234352, "epoch": 2633, "lr": 4.55441787968619e-06}
{"train_loss": 0.051116522401571274, "global_step": 234353, "epoch": 2633, "lr": 4.554176139093403e-06}
{"train_loss": 0.0648665577173233, "global_step": 234354, "epoch": 2633, "lr": 4.5539344046102606e-06}
{"train_loss": 0.045333098620176315, "global_step": 234355, "epoch": 2633, "lr": 4.5536926762367815e-06}
{"train_loss": 0.07270147651433945, "global_step": 234356, "epoch": 2633, "lr": 4.55345095397301e-06}
{"train_loss": 0.05237164720892906, "global_step": 234357, "epoch": 2633, "lr": 4.553209237818968e-06}
{"train_loss": 0.02137138321995735, "global_step": 234358, "epoch": 2633, "lr": 4.552967527774699e-06}
{"train_loss": 0.049399081617593765, "global_step": 234359, "epoch": 2633, "lr": 4.55272582384022e-06}
{"train_loss": 0.06389458477497101, "global_step": 234360, "epoch": 2633, "lr": 4.552484126015582e-06}
{"train_loss": 0.046954188495874405, "global_step": 234361, "epoch": 2633, "lr": 4.5522424343008015e-06}
{"train_loss": 0.06745988875627518, "global_step": 234362, "epoch": 2633, "lr": 4.552000748695929e-06}
{"train_loss": 0.09359528124332428, "global_step": 234363, "epoch": 2633, "lr": 4.551759069200973e-06}
{"train_loss": 0.031071800738573074, "global_step": 234364, "epoch": 2633, "lr": 4.551517395815991e-06}
{"train_loss": 0.044276706874370575, "global_step": 234365, "epoch": 2633, "lr": 4.551275728540993e-06}
{"train_loss": 0.07054939866065979, "global_step": 234366, "epoch": 2633, "lr": 4.55103406737602e-06}
{"train_loss": 0.04255630820989609, "global_step": 234367, "epoch": 2633, "lr": 4.55079241232112e-06}
{"train_loss": 0.02390456758439541, "global_step": 234368, "epoch": 2633, "lr": 4.550550763376299e-06}
{"train_loss": 0.029674777761101723, "global_step": 234369, "epoch": 2633, "lr": 4.550309120541613e-06}
{"train_loss": 0.021255314350128174, "global_step": 234370, "epoch": 2633, "lr": 4.550067483817077e-06}
{"train_loss": 0.02992408350110054, "global_step": 234371, "epoch": 2633, "lr": 4.549825853202738e-06}
{"train_loss": 0.09210552275180817, "global_step": 234372, "epoch": 2633, "lr": 4.549584228698611e-06}
{"train_loss": 0.04375571012496948, "global_step": 234373, "epoch": 2633, "lr": 4.549342610304746e-06}
{"train_loss": 0.022290309891104698, "global_step": 234374, "epoch": 2633, "lr": 4.54910099802116e-06}
{"train_loss": 0.02576347254216671, "global_step": 234375, "epoch": 2633, "lr": 4.548859391847904e-06}
{"train_loss": 0.032536059617996216, "global_step": 234376, "epoch": 2633, "lr": 4.548617791784993e-06}
{"train_loss": 0.03333813697099686, "global_step": 234377, "epoch": 2633, "lr": 4.548376197832471e-06}
{"train_loss": 0.05000702664256096, "global_step": 234378, "epoch": 2633, "lr": 4.548134609990356e-06}
{"train_loss": 0.024543985724449158, "global_step": 234379, "epoch": 2633, "lr": 4.547893028258704e-06}
{"train_loss": 0.09393982589244843, "global_step": 234380, "epoch": 2633, "lr": 4.547651452637525e-06}
{"train_loss": 0.016557766124606133, "global_step": 234381, "epoch": 2633, "lr": 4.547409883126863e-06}
{"train_loss": 0.07325486093759537, "global_step": 234382, "epoch": 2633, "lr": 4.547168319726747e-06}
{"train_loss": 0.04637183994054794, "global_step": 234383, "epoch": 2633, "lr": 4.546926762437215e-06}
{"train_loss": 0.021739685907959938, "global_step": 234384, "epoch": 2633, "lr": 4.546685211258284e-06}
{"train_loss": 0.05041869357228279, "global_step": 234385, "epoch": 2633, "lr": 4.546443666190009e-06}
{"train_loss": 0.036498356610536575, "global_step": 234386, "epoch": 2633, "lr": 4.546202127232408e-06}
{"train_loss": 0.09973149001598358, "global_step": 234387, "epoch": 2633, "lr": 4.545960594385507e-06}
{"train_loss": 0.04754789546132088, "global_step": 234388, "epoch": 2633, "lr": 4.545719067649357e-06}
{"train_loss": 0.048453159630298615, "global_step": 234389, "epoch": 2633, "lr": 4.545477547023974e-06}
{"train_loss": 0.04549063369631767, "global_step": 234390, "epoch": 2633, "lr": 4.545236032509409e-06}
{"train_loss": 0.09856105595827103, "global_step": 234391, "epoch": 2633, "lr": 4.544994524105667e-06}
{"train_loss": 0.05431320145726204, "global_step": 234392, "epoch": 2633, "lr": 4.544753021812808e-06}
{"train_loss": 0.07325045019388199, "global_step": 234393, "epoch": 2633, "lr": 4.544511525630846e-06}
{"train_loss": 0.07045648247003555, "global_step": 234394, "epoch": 2633, "lr": 4.544270035559833e-06}
{"train_loss": 0.08061914891004562, "global_step": 234395, "epoch": 2633, "lr": 4.544028551599771e-06}
{"train_loss": 0.04609251022338867, "global_step": 234396, "epoch": 2633, "lr": 4.5437870737507274e-06}
{"train_loss": 0.0382913202047348, "global_step": 234397, "epoch": 2633, "lr": 4.543545602012705e-06}
{"train_loss": 0.03745448589324951, "global_step": 234398, "epoch": 2633, "lr": 4.5433041363857565e-06}
{"train_loss": 0.029478533193469048, "global_step": 234399, "epoch": 2633, "lr": 4.543062676869897e-06}
{"train_loss": 0.0683390200138092, "global_step": 234400, "epoch": 2633, "lr": 4.542821223465182e-06}
{"train_loss": 0.05412137880921364, "global_step": 234401, "epoch": 2633, "lr": 4.542579776171618e-06}
{"train_loss": 0.041352011263370514, "global_step": 234402, "epoch": 2633, "lr": 4.542338334989249e-06}
{"train_loss": 0.06144963204860687, "global_step": 234403, "epoch": 2633, "lr": 4.542096899918124e-06}
{"train_loss": 0.06463193148374557, "global_step": 234404, "epoch": 2633, "lr": 4.541855470958245e-06}
{"train_loss": 0.09736217558383942, "global_step": 234405, "epoch": 2633, "lr": 4.541614048109671e-06}
{"train_loss": 0.05086682736873627, "global_step": 234406, "epoch": 2633, "lr": 4.541372631372415e-06}
{"train_loss": 0.034589823335409164, "global_step": 234407, "epoch": 2633, "lr": 4.5411312207465255e-06}
{"train_loss": 0.04724377021193504, "global_step": 234408, "epoch": 2633, "lr": 4.540889816232014e-06}
{"train_loss": 0.037451133131980896, "global_step": 234409, "epoch": 2633, "lr": 4.540648417828941e-06}
{"train_loss": 0.0470859669148922, "global_step": 234410, "epoch": 2633, "lr": 4.540407025537308e-06}
{"train_loss": 0.01998668909072876, "global_step": 234411, "epoch": 2633, "lr": 4.54016563935718e-06}
{"train_loss": 0.05349636450409889, "global_step": 234412, "epoch": 2633, "lr": 4.539924259288558e-06}
{"train_loss": 0.051298271864652634, "global_step": 234413, "epoch": 2633, "lr": 4.539682885331498e-06}
{"train_loss": 0.0312928631901741, "global_step": 234414, "epoch": 2633, "lr": 4.539441517486015e-06}
{"train_loss": 0.034272268414497375, "global_step": 234415, "epoch": 2633, "lr": 4.5392001557521605e-06}
{"train_loss": 0.03510245308279991, "global_step": 234416, "epoch": 2633, "lr": 4.53895880012995e-06}
{"train_loss": 0.04227260872721672, "global_step": 234417, "epoch": 2633, "lr": 4.538717450619428e-06}
{"train_loss": 0.026757994666695595, "global_step": 234418, "epoch": 2633, "lr": 4.538476107220613e-06}
{"train_loss": 0.0673656165599823, "global_step": 234419, "epoch": 2633, "lr": 4.538234769933547e-06}
{"train_loss": 0.0794977992773056, "global_step": 234420, "epoch": 2633, "lr": 4.5379934387582644e-06}
{"train_loss": 0.02579846978187561, "global_step": 234421, "epoch": 2633, "lr": 4.5377521136947985e-06}
{"train_loss": 0.061806295067071915, "global_step": 234422, "epoch": 2633, "lr": 4.537510794743172e-06}
{"train_loss": 0.051641833037137985, "global_step": 234423, "epoch": 2633, "lr": 4.537269481903428e-06}
{"train_loss": 0.06817913800477982, "global_step": 234424, "epoch": 2633, "lr": 4.537028175175589e-06}
{"train_loss": 0.04783098001995783, "global_step": 234425, "epoch": 2633, "lr": 4.536786874559696e-06, "val_loss": 9.31890869140625}
{"train_loss": 0.024695388972759247, "global_step": 234426, "epoch": 2634, "lr": 4.536545580055779e-06}
{"train_loss": 0.05005830526351929, "global_step": 234427, "epoch": 2634, "lr": 4.536304291663862e-06}
{"train_loss": 0.035434357821941376, "global_step": 234428, "epoch": 2634, "lr": 4.536063009383995e-06}
{"train_loss": 0.041958022862672806, "global_step": 234429, "epoch": 2634, "lr": 4.535821733216189e-06}
{"train_loss": 0.10828647017478943, "global_step": 234430, "epoch": 2634, "lr": 4.535580463160499e-06}
{"train_loss": 0.05036945268511772, "global_step": 234431, "epoch": 2634, "lr": 4.535339199216937e-06}
{"train_loss": 0.02450060471892357, "global_step": 234432, "epoch": 2634, "lr": 4.5350979413855525e-06}
{"train_loss": 0.08035903424024582, "global_step": 234433, "epoch": 2634, "lr": 4.534856689666361e-06}
{"train_loss": 0.032328516244888306, "global_step": 234434, "epoch": 2634, "lr": 4.5346154440594095e-06}
{"train_loss": 0.11649635434150696, "global_step": 234435, "epoch": 2634, "lr": 4.534374204564718e-06}
{"train_loss": 0.020166194066405296, "global_step": 234436, "epoch": 2634, "lr": 4.534132971182336e-06}
{"train_loss": 0.06004374101758003, "global_step": 234437, "epoch": 2634, "lr": 4.5338917439122775e-06}
{"train_loss": 0.049435149878263474, "global_step": 234438, "epoch": 2634, "lr": 4.533650522754579e-06}
{"train_loss": 0.07242880761623383, "global_step": 234439, "epoch": 2634, "lr": 4.533409307709291e-06}
{"train_loss": 0.04306791350245476, "global_step": 234440, "epoch": 2634, "lr": 4.533168098776419e-06}
{"train_loss": 0.03444990515708923, "global_step": 234441, "epoch": 2634, "lr": 4.532926895956019e-06}
{"train_loss": 0.024932431057095528, "global_step": 234442, "epoch": 2634, "lr": 4.532685699248107e-06}
{"train_loss": 0.05142407491803169, "global_step": 234443, "epoch": 2634, "lr": 4.532444508652728e-06}
{"train_loss": 0.04489215835928917, "global_step": 234444, "epoch": 2634, "lr": 4.532203324169898e-06}
{"train_loss": 0.029387572780251503, "global_step": 234445, "epoch": 2634, "lr": 4.531962145799668e-06}
{"train_loss": 0.054809972643852234, "global_step": 234446, "epoch": 2634, "lr": 4.531720973542048e-06}
{"train_loss": 0.04702211916446686, "global_step": 234447, "epoch": 2634, "lr": 4.531479807397099e-06}
{"train_loss": 0.013239052146673203, "global_step": 234448, "epoch": 2634, "lr": 4.531238647364827e-06}
{"train_loss": 0.047630105167627335, "global_step": 234449, "epoch": 2634, "lr": 4.530997493445288e-06}
{"train_loss": 0.08206359297037125, "global_step": 234450, "epoch": 2634, "lr": 4.530756345638487e-06}
{"train_loss": 0.059924785047769547, "global_step": 234451, "epoch": 2634, "lr": 4.530515203944485e-06}
{"train_loss": 0.08525564521551132, "global_step": 234452, "epoch": 2634, "lr": 4.530274068363294e-06}
{"train_loss": 0.048887282609939575, "global_step": 234453, "epoch": 2634, "lr": 4.530032938894957e-06}
{"train_loss": 0.047109052538871765, "global_step": 234454, "epoch": 2634, "lr": 4.529791815539497e-06}
{"train_loss": 0.040589071810245514, "global_step": 234455, "epoch": 2634, "lr": 4.529550698296964e-06}
{"train_loss": 0.03532527759671211, "global_step": 234456, "epoch": 2634, "lr": 4.529309587167363e-06}
{"train_loss": 0.03375180810689926, "global_step": 234457, "epoch": 2634, "lr": 4.52906848215075e-06}
{"train_loss": 0.01766739971935749, "global_step": 234458, "epoch": 2634, "lr": 4.528827383247153e-06}
{"train_loss": 0.02544810064136982, "global_step": 234459, "epoch": 2634, "lr": 4.528586290456599e-06}
{"train_loss": 0.08253379911184311, "global_step": 234460, "epoch": 2634, "lr": 4.528345203779116e-06}
{"train_loss": 0.09454364329576492, "global_step": 234461, "epoch": 2634, "lr": 4.528104123214755e-06}
{"train_loss": 0.030514098703861237, "global_step": 234462, "epoch": 2634, "lr": 4.527863048763526e-06}
{"train_loss": 0.02912941202521324, "global_step": 234463, "epoch": 2634, "lr": 4.527621980425478e-06}
{"train_loss": 0.053222618997097015, "global_step": 234464, "epoch": 2634, "lr": 4.5273809182006355e-06}
{"train_loss": 0.05606050789356232, "global_step": 234465, "epoch": 2634, "lr": 4.52713986208903e-06}
{"train_loss": 0.04423259198665619, "global_step": 234466, "epoch": 2634, "lr": 4.526898812090702e-06}
{"train_loss": 0.07137661427259445, "global_step": 234467, "epoch": 2634, "lr": 4.526657768205667e-06}
{"train_loss": 0.058902304619550705, "global_step": 234468, "epoch": 2634, "lr": 4.52641673043398e-06}
{"train_loss": 0.04497643932700157, "global_step": 234469, "epoch": 2634, "lr": 4.526175698775653e-06}
{"train_loss": 0.02864500693976879, "global_step": 234470, "epoch": 2634, "lr": 4.525934673230731e-06}
{"train_loss": 0.048439327627420425, "global_step": 234471, "epoch": 2634, "lr": 4.52569365379924e-06}
{"train_loss": 0.02952679619193077, "global_step": 234472, "epoch": 2634, "lr": 4.525452640481226e-06}
{"train_loss": 0.06792504340410233, "global_step": 234473, "epoch": 2634, "lr": 4.525211633276694e-06}
{"train_loss": 0.0781967043876648, "global_step": 234474, "epoch": 2634, "lr": 4.5249706321856985e-06}
{"train_loss": 0.07396528124809265, "global_step": 234475, "epoch": 2634, "lr": 4.524729637208275e-06}
{"train_loss": 0.029805513098835945, "global_step": 234476, "epoch": 2634, "lr": 4.524488648344438e-06}
{"train_loss": 0.10239744186401367, "global_step": 234477, "epoch": 2634, "lr": 4.524247665594239e-06}
{"train_loss": 0.04400159418582916, "global_step": 234478, "epoch": 2634, "lr": 4.524006688957688e-06}
{"train_loss": 0.0521315298974514, "global_step": 234479, "epoch": 2634, "lr": 4.5237657184348415e-06}
{"train_loss": 0.05376807972788811, "global_step": 234480, "epoch": 2634, "lr": 4.52352475402571e-06}
{"train_loss": 0.04222976788878441, "global_step": 234481, "epoch": 2634, "lr": 4.523283795730348e-06}
{"train_loss": 0.019263360649347305, "global_step": 234482, "epoch": 2634, "lr": 4.523042843548769e-06}
{"train_loss": 0.06409066915512085, "global_step": 234483, "epoch": 2634, "lr": 4.522801897481016e-06}
{"train_loss": 0.018061958253383636, "global_step": 234484, "epoch": 2634, "lr": 4.522560957527111e-06}
{"train_loss": 0.022120758891105652, "global_step": 234485, "epoch": 2634, "lr": 4.522320023687104e-06}
{"train_loss": 0.05271212384104729, "global_step": 234486, "epoch": 2634, "lr": 4.522079095961007e-06}
{"train_loss": 0.0349779836833477, "global_step": 234487, "epoch": 2634, "lr": 4.521838174348869e-06}
{"train_loss": 0.04987822100520134, "global_step": 234488, "epoch": 2634, "lr": 4.5215972588507074e-06}
{"train_loss": 0.02140348218381405, "global_step": 234489, "epoch": 2634, "lr": 4.521356349466577e-06}
{"train_loss": 0.039796605706214905, "global_step": 234490, "epoch": 2634, "lr": 4.521115446196483e-06}
{"train_loss": 0.013102498836815357, "global_step": 234491, "epoch": 2634, "lr": 4.520874549040477e-06}
{"train_loss": 0.032317738980054855, "global_step": 234492, "epoch": 2634, "lr": 4.52063365799858e-06}
{"train_loss": 0.04644235223531723, "global_step": 234493, "epoch": 2634, "lr": 4.520392773070825e-06}
{"train_loss": 0.03432648256421089, "global_step": 234494, "epoch": 2634, "lr": 4.520151894257263e-06}
{"train_loss": 0.02767135761678219, "global_step": 234495, "epoch": 2634, "lr": 4.519911021557899e-06}
{"train_loss": 0.0389224998652935, "global_step": 234496, "epoch": 2634, "lr": 4.519670154972794e-06}
{"train_loss": 0.0834946483373642, "global_step": 234497, "epoch": 2634, "lr": 4.5194292945019586e-06}
{"train_loss": 0.0500820055603981, "global_step": 234498, "epoch": 2634, "lr": 4.519188440145422e-06}
{"train_loss": 0.06323609501123428, "global_step": 234499, "epoch": 2634, "lr": 4.518947591903239e-06}
{"train_loss": 0.06397786736488342, "global_step": 234500, "epoch": 2634, "lr": 4.51870674977542e-06}
{"train_loss": 0.043323613703250885, "global_step": 234501, "epoch": 2634, "lr": 4.518465913762016e-06}
{"train_loss": 0.051895927637815475, "global_step": 234502, "epoch": 2634, "lr": 4.518225083863048e-06}
{"train_loss": 0.04423104226589203, "global_step": 234503, "epoch": 2634, "lr": 4.517984260078539e-06}
{"train_loss": 0.03275540471076965, "global_step": 234504, "epoch": 2634, "lr": 4.517743442408545e-06}
{"train_loss": 0.051929011940956116, "global_step": 234505, "epoch": 2634, "lr": 4.5175026308530765e-06}
{"train_loss": 0.031857311725616455, "global_step": 234506, "epoch": 2634, "lr": 4.517261825412189e-06}
{"train_loss": 0.025404518470168114, "global_step": 234507, "epoch": 2634, "lr": 4.517021026085888e-06}
{"train_loss": 0.07297763228416443, "global_step": 234508, "epoch": 2634, "lr": 4.5167802328742235e-06}
{"train_loss": 0.10035824030637741, "global_step": 234509, "epoch": 2634, "lr": 4.516539445777224e-06}
{"train_loss": 0.09281258285045624, "global_step": 234510, "epoch": 2634, "lr": 4.5162986647949155e-06}
{"train_loss": 0.021419107913970947, "global_step": 234511, "epoch": 2634, "lr": 4.516057889927344e-06}
{"train_loss": 0.05245055630803108, "global_step": 234512, "epoch": 2634, "lr": 4.515817121174526e-06}
{"train_loss": 0.03826889023184776, "global_step": 234513, "epoch": 2634, "lr": 4.515576358536516e-06}
{"train_loss": 0.04882031542071131, "global_step": 234514, "epoch": 2634, "lr": 4.515335602013321e-06, "val_loss": 9.23102855682373}
{"train_loss": 0.03329847380518913, "global_step": 234515, "epoch": 2635, "lr": 4.515094851604989e-06}
{"train_loss": 0.07732833176851273, "global_step": 234516, "epoch": 2635, "lr": 4.514854107311545e-06}
{"train_loss": 0.05252615362405777, "global_step": 234517, "epoch": 2635, "lr": 4.51461336913303e-06}
{"train_loss": 0.04680386930704117, "global_step": 234518, "epoch": 2635, "lr": 4.5143726370694636e-06}
{"train_loss": 0.02872517704963684, "global_step": 234519, "epoch": 2635, "lr": 4.514131911120894e-06}
{"train_loss": 0.055979467928409576, "global_step": 234520, "epoch": 2635, "lr": 4.513891191287334e-06}
{"train_loss": 0.05141063034534454, "global_step": 234521, "epoch": 2635, "lr": 4.513650477568837e-06}
{"train_loss": 0.032856542617082596, "global_step": 234522, "epoch": 2635, "lr": 4.513409769965415e-06}
{"train_loss": 0.08829478919506073, "global_step": 234523, "epoch": 2635, "lr": 4.513169068477125e-06}
{"train_loss": 0.05835780128836632, "global_step": 234524, "epoch": 2635, "lr": 4.51292837310397e-06}
{"train_loss": 0.06221606582403183, "global_step": 234525, "epoch": 2635, "lr": 4.512687683846012e-06}
{"train_loss": 0.03894258663058281, "global_step": 234526, "epoch": 2635, "lr": 4.512447000703257e-06}
{"train_loss": 0.05929121747612953, "global_step": 234527, "epoch": 2635, "lr": 4.512206323675755e-06}
{"train_loss": 0.08729133009910583, "global_step": 234528, "epoch": 2635, "lr": 4.511965652763528e-06}
{"train_loss": 0.05673035979270935, "global_step": 234529, "epoch": 2635, "lr": 4.511724987966609e-06}
{"train_loss": 0.041114792227745056, "global_step": 234530, "epoch": 2635, "lr": 4.511484329285049e-06}
{"train_loss": 0.055157966911792755, "global_step": 234531, "epoch": 2635, "lr": 4.511243676718851e-06}
{"train_loss": 0.05362628027796745, "global_step": 234532, "epoch": 2635, "lr": 4.511003030268074e-06}
{"train_loss": 0.0824589729309082, "global_step": 234533, "epoch": 2635, "lr": 4.510762389932726e-06}
{"train_loss": 0.057077862322330475, "global_step": 234534, "epoch": 2635, "lr": 4.51052175571286e-06}
{"train_loss": 0.07358738780021667, "global_step": 234535, "epoch": 2635, "lr": 4.510281127608495e-06}
{"train_loss": 0.09283275157213211, "global_step": 234536, "epoch": 2635, "lr": 4.510040505619672e-06}
{"train_loss": 0.040154509246349335, "global_step": 234537, "epoch": 2635, "lr": 4.509799889746424e-06}
{"train_loss": 0.01550260465592146, "global_step": 234538, "epoch": 2635, "lr": 4.509559279988773e-06}
{"train_loss": 0.05300140008330345, "global_step": 234539, "epoch": 2635, "lr": 4.509318676346763e-06}
{"train_loss": 0.030779492110013962, "global_step": 234540, "epoch": 2635, "lr": 4.509078078820412e-06}
{"train_loss": 0.03508426994085312, "global_step": 234541, "epoch": 2635, "lr": 4.5088374874097675e-06}
{"train_loss": 0.044914063066244125, "global_step": 234542, "epoch": 2635, "lr": 4.508596902114853e-06}
{"train_loss": 0.06604094058275223, "global_step": 234543, "epoch": 2635, "lr": 4.5083563229357025e-06}
{"train_loss": 0.060684725642204285, "global_step": 234544, "epoch": 2635, "lr": 4.5081157498723494e-06}
{"train_loss": 0.03109661117196083, "global_step": 234545, "epoch": 2635, "lr": 4.50787518292482e-06}
{"train_loss": 0.021938979625701904, "global_step": 234546, "epoch": 2635, "lr": 4.507634622093154e-06}
{"train_loss": 0.09310819953680038, "global_step": 234547, "epoch": 2635, "lr": 4.5073940673773905e-06}
{"train_loss": 0.030128033831715584, "global_step": 234548, "epoch": 2635, "lr": 4.507153518777546e-06}
{"train_loss": 0.03222987428307533, "global_step": 234549, "epoch": 2635, "lr": 4.506912976293664e-06}
{"train_loss": 0.03917178511619568, "global_step": 234550, "epoch": 2635, "lr": 4.506672439925769e-06}
{"train_loss": 0.035640593618154526, "global_step": 234551, "epoch": 2635, "lr": 4.506431909673908e-06}
{"train_loss": 0.048673827201128006, "global_step": 234552, "epoch": 2635, "lr": 4.506191385538089e-06}
{"train_loss": 0.04902907460927963, "global_step": 234553, "epoch": 2635, "lr": 4.505950867518366e-06}
{"train_loss": 0.09313368052244186, "global_step": 234554, "epoch": 2635, "lr": 4.5057103556147565e-06}
{"train_loss": 0.11179722845554352, "global_step": 234555, "epoch": 2635, "lr": 4.5054698498273095e-06}
{"train_loss": 0.019454188644886017, "global_step": 234556, "epoch": 2635, "lr": 4.5052293501560426e-06}
{"train_loss": 0.015831222757697105, "global_step": 234557, "epoch": 2635, "lr": 4.504988856600995e-06}
{"train_loss": 0.06922101229429245, "global_step": 234558, "epoch": 2635, "lr": 4.504748369162193e-06}
{"train_loss": 0.049801431596279144, "global_step": 234559, "epoch": 2635, "lr": 4.504507887839677e-06}
{"train_loss": 0.015388903208076954, "global_step": 234560, "epoch": 2635, "lr": 4.504267412633472e-06}
{"train_loss": 0.05120563507080078, "global_step": 234561, "epoch": 2635, "lr": 4.504026943543621e-06}
{"train_loss": 0.023700160905718803, "global_step": 234562, "epoch": 2635, "lr": 4.503786480570138e-06}
{"train_loss": 0.10676461458206177, "global_step": 234563, "epoch": 2635, "lr": 4.503546023713073e-06}
{"train_loss": 0.04117491841316223, "global_step": 234564, "epoch": 2635, "lr": 4.50330557297245e-06}
{"train_loss": 0.02425362356007099, "global_step": 234565, "epoch": 2635, "lr": 4.5030651283483e-06}
{"train_loss": 0.06994873285293579, "global_step": 234566, "epoch": 2635, "lr": 4.502824689840668e-06}
{"train_loss": 0.0355815663933754, "global_step": 234567, "epoch": 2635, "lr": 4.502584257449566e-06}
{"train_loss": 0.054276395589113235, "global_step": 234568, "epoch": 2635, "lr": 4.502343831175043e-06}
{"train_loss": 0.0531243197619915, "global_step": 234569, "epoch": 2635, "lr": 4.502103411017122e-06}
{"train_loss": 0.05696959048509598, "global_step": 234570, "epoch": 2635, "lr": 4.501862996975842e-06}
{"train_loss": 0.08150886744260788, "global_step": 234571, "epoch": 2635, "lr": 4.50162258905123e-06}
{"train_loss": 0.021893752738833427, "global_step": 234572, "epoch": 2635, "lr": 4.501382187243325e-06}
{"train_loss": 0.02887096256017685, "global_step": 234573, "epoch": 2635, "lr": 4.501141791552144e-06}
{"train_loss": 0.03118717297911644, "global_step": 234574, "epoch": 2635, "lr": 4.500901401977742e-06}
{"train_loss": 0.049193840473890305, "global_step": 234575, "epoch": 2635, "lr": 4.5006610185201415e-06}
{"train_loss": 0.05427537113428116, "global_step": 234576, "epoch": 2635, "lr": 4.500420641179359e-06}
{"train_loss": 0.08919614553451538, "global_step": 234577, "epoch": 2635, "lr": 4.500180269955451e-06}
{"train_loss": 0.03008008375763893, "global_step": 234578, "epoch": 2635, "lr": 4.499939904848427e-06}
{"train_loss": 0.011767680756747723, "global_step": 234579, "epoch": 2635, "lr": 4.499699545858343e-06}
{"train_loss": 0.07376379519701004, "global_step": 234580, "epoch": 2635, "lr": 4.499459192985222e-06}
{"train_loss": 0.0316733680665493, "global_step": 234581, "epoch": 2635, "lr": 4.499218846229086e-06}
{"train_loss": 0.05742327496409416, "global_step": 234582, "epoch": 2635, "lr": 4.4989785055899795e-06}
{"train_loss": 0.09012163430452347, "global_step": 234583, "epoch": 2635, "lr": 4.498738171067923e-06}
{"train_loss": 0.07102543115615845, "global_step": 234584, "epoch": 2635, "lr": 4.498497842662963e-06}
{"train_loss": 0.018016545102000237, "global_step": 234585, "epoch": 2635, "lr": 4.498257520375126e-06}
{"train_loss": 0.040694501250982285, "global_step": 234586, "epoch": 2635, "lr": 4.498017204204441e-06}
{"train_loss": 0.04311042279005051, "global_step": 234587, "epoch": 2635, "lr": 4.497776894150946e-06}
{"train_loss": 0.072736456990242, "global_step": 234588, "epoch": 2635, "lr": 4.497536590214662e-06}
{"train_loss": 0.07636281102895737, "global_step": 234589, "epoch": 2635, "lr": 4.497296292395642e-06}
{"train_loss": 0.061348918825387955, "global_step": 234590, "epoch": 2635, "lr": 4.4970560006939e-06}
{"train_loss": 0.029079105705022812, "global_step": 234591, "epoch": 2635, "lr": 4.496815715109476e-06}
{"train_loss": 0.06985186040401459, "global_step": 234592, "epoch": 2635, "lr": 4.496575435642398e-06}
{"train_loss": 0.08110834658145905, "global_step": 234593, "epoch": 2635, "lr": 4.496335162292703e-06}
{"train_loss": 0.07770568877458572, "global_step": 234594, "epoch": 2635, "lr": 4.496094895060415e-06}
{"train_loss": 0.015238989144563675, "global_step": 234595, "epoch": 2635, "lr": 4.495854633945584e-06}
{"train_loss": 0.040563154965639114, "global_step": 234596, "epoch": 2635, "lr": 4.49561437894822e-06}
{"train_loss": 0.09595721960067749, "global_step": 234597, "epoch": 2635, "lr": 4.495374130068375e-06}
{"train_loss": 0.018891237676143646, "global_step": 234598, "epoch": 2635, "lr": 4.495133887306064e-06}
{"train_loss": 0.05442238971590996, "global_step": 234599, "epoch": 2635, "lr": 4.494893650661331e-06}
{"train_loss": 0.019839709624648094, "global_step": 234600, "epoch": 2635, "lr": 4.494653420134204e-06}
{"train_loss": 0.02376140095293522, "global_step": 234601, "epoch": 2635, "lr": 4.4944131957247115e-06}
{"train_loss": 0.03593195602297783, "global_step": 234602, "epoch": 2635, "lr": 4.494172977432898e-06}
{"train_loss": 0.05102455614950885, "global_step": 234603, "epoch": 2635, "lr": 4.493932765258785e-06, "val_loss": 9.274175643920898, "train_action_mse_error": 15.991203308105469}
{"train_loss": 0.06085577234625816, "global_step": 234604, "epoch": 2636, "lr": 4.493692559202412e-06}
{"train_loss": 0.03736924007534981, "global_step": 234605, "epoch": 2636, "lr": 4.4934523592638e-06}
{"train_loss": 0.07545672357082367, "global_step": 234606, "epoch": 2636, "lr": 4.493212165443e-06}
{"train_loss": 0.04926837235689163, "global_step": 234607, "epoch": 2636, "lr": 4.492971977740023e-06}
{"train_loss": 0.05894578993320465, "global_step": 234608, "epoch": 2636, "lr": 4.49273179615492e-06}
{"train_loss": 0.022052546963095665, "global_step": 234609, "epoch": 2636, "lr": 4.4924916206877055e-06}
{"train_loss": 0.024753587320446968, "global_step": 234610, "epoch": 2636, "lr": 4.492251451338425e-06}
{"train_loss": 0.055052436888217926, "global_step": 234611, "epoch": 2636, "lr": 4.492011288107106e-06}
{"train_loss": 0.022694656625390053, "global_step": 234612, "epoch": 2636, "lr": 4.491771130993782e-06}
{"train_loss": 0.07697171717882156, "global_step": 234613, "epoch": 2636, "lr": 4.4915309799984925e-06}
{"train_loss": 0.049777284264564514, "global_step": 234614, "epoch": 2636, "lr": 4.491290835121248e-06}
{"train_loss": 0.033933572471141815, "global_step": 234615, "epoch": 2636, "lr": 4.491050696362103e-06}
{"train_loss": 0.053994640707969666, "global_step": 234616, "epoch": 2636, "lr": 4.490810563721076e-06}
{"train_loss": 0.02897053025662899, "global_step": 234617, "epoch": 2636, "lr": 4.490570437198216e-06}
{"train_loss": 0.0331389494240284, "global_step": 234618, "epoch": 2636, "lr": 4.490330316793529e-06}
{"train_loss": 0.062078334391117096, "global_step": 234619, "epoch": 2636, "lr": 4.490090202507075e-06}
{"train_loss": 0.033490486443042755, "global_step": 234620, "epoch": 2636, "lr": 4.489850094338865e-06}
{"train_loss": 0.05784633383154869, "global_step": 234621, "epoch": 2636, "lr": 4.489609992288946e-06}
{"train_loss": 0.060408007353544235, "global_step": 234622, "epoch": 2636, "lr": 4.489369896357338e-06}
{"train_loss": 0.06824473291635513, "global_step": 234623, "epoch": 2636, "lr": 4.48912980654409e-06}
{"train_loss": 0.04599898308515549, "global_step": 234624, "epoch": 2636, "lr": 4.48888972284921e-06}
{"train_loss": 0.04106590524315834, "global_step": 234625, "epoch": 2636, "lr": 4.488649645272757e-06}
{"train_loss": 0.05576068162918091, "global_step": 234626, "epoch": 2636, "lr": 4.488409573814739e-06}
{"train_loss": 0.02553451620042324, "global_step": 234627, "epoch": 2636, "lr": 4.48816950847521e-06}
{"train_loss": 0.028602613136172295, "global_step": 234628, "epoch": 2636, "lr": 4.4879294492541804e-06}
{"train_loss": 0.07970284670591354, "global_step": 234629, "epoch": 2636, "lr": 4.4876893961517065e-06}
{"train_loss": 0.03249267116189003, "global_step": 234630, "epoch": 2636, "lr": 4.4874493491677995e-06}
{"train_loss": 0.057004865258932114, "global_step": 234631, "epoch": 2636, "lr": 4.487209308302504e-06}
{"train_loss": 0.016806453466415405, "global_step": 234632, "epoch": 2636, "lr": 4.486969273555847e-06}
{"train_loss": 0.0692133903503418, "global_step": 234633, "epoch": 2636, "lr": 4.486729244927862e-06}
{"train_loss": 0.04019599407911301, "global_step": 234634, "epoch": 2636, "lr": 4.486489222418578e-06}
{"train_loss": 0.027111820876598358, "global_step": 234635, "epoch": 2636, "lr": 4.486249206028042e-06}
{"train_loss": 0.04473242536187172, "global_step": 234636, "epoch": 2636, "lr": 4.4860091957562635e-06}
{"train_loss": 0.02443394996225834, "global_step": 234637, "epoch": 2636, "lr": 4.485769191603284e-06}
{"train_loss": 0.058132950216531754, "global_step": 234638, "epoch": 2636, "lr": 4.485529193569149e-06}
{"train_loss": 0.06941286474466324, "global_step": 234639, "epoch": 2636, "lr": 4.485289201653875e-06}
{"train_loss": 0.04455915838479996, "global_step": 234640, "epoch": 2636, "lr": 4.485049215857501e-06}
{"train_loss": 0.0270178634673357, "global_step": 234641, "epoch": 2636, "lr": 4.4848092361800545e-06}
{"train_loss": 0.04085739701986313, "global_step": 234642, "epoch": 2636, "lr": 4.484569262621574e-06}
{"train_loss": 0.015519723296165466, "global_step": 234643, "epoch": 2636, "lr": 4.484329295182083e-06}
{"train_loss": 0.031014008447527885, "global_step": 234644, "epoch": 2636, "lr": 4.48408933386163e-06}
{"train_loss": 0.07538487017154694, "global_step": 234645, "epoch": 2636, "lr": 4.483849378660221e-06}
{"train_loss": 0.05229781195521355, "global_step": 234646, "epoch": 2636, "lr": 4.483609429577918e-06}
{"train_loss": 0.046490173786878586, "global_step": 234647, "epoch": 2636, "lr": 4.4833694866147305e-06}
{"train_loss": 0.037142783403396606, "global_step": 234648, "epoch": 2636, "lr": 4.483129549770704e-06}
{"train_loss": 0.026762520894408226, "global_step": 234649, "epoch": 2636, "lr": 4.4828896190458605e-06}
{"train_loss": 0.02090088464319706, "global_step": 234650, "epoch": 2636, "lr": 4.482649694440244e-06}
{"train_loss": 0.03633685037493706, "global_step": 234651, "epoch": 2636, "lr": 4.482409775953872e-06}
{"train_loss": 0.03234240412712097, "global_step": 234652, "epoch": 2636, "lr": 4.482169863586794e-06}
{"train_loss": 0.059645455330610275, "global_step": 234653, "epoch": 2636, "lr": 4.481929957339037e-06}
{"train_loss": 0.08546927571296692, "global_step": 234654, "epoch": 2636, "lr": 4.4816900572106194e-06}
{"train_loss": 0.022088833153247833, "global_step": 234655, "epoch": 2636, "lr": 4.4814501632015896e-06}
{"train_loss": 0.03583782538771629, "global_step": 234656, "epoch": 2636, "lr": 4.4812102753119644e-06}
{"train_loss": 0.045934583991765976, "global_step": 234657, "epoch": 2636, "lr": 4.4809703935418e-06}
{"train_loss": 0.03190826624631882, "global_step": 234658, "epoch": 2636, "lr": 4.480730517891102e-06}
{"train_loss": 0.03585762158036232, "global_step": 234659, "epoch": 2636, "lr": 4.480490648359925e-06}
{"train_loss": 0.0385696180164814, "global_step": 234660, "epoch": 2636, "lr": 4.480250784948281e-06}
{"train_loss": 0.0385340116918087, "global_step": 234661, "epoch": 2636, "lr": 4.480010927656225e-06}
{"train_loss": 0.02345142886042595, "global_step": 234662, "epoch": 2636, "lr": 4.479771076483763e-06}
{"train_loss": 0.0296052023768425, "global_step": 234663, "epoch": 2636, "lr": 4.47953123143095e-06}
{"train_loss": 0.05902520939707756, "global_step": 234664, "epoch": 2636, "lr": 4.479291392497803e-06}
{"train_loss": 0.05420182645320892, "global_step": 234665, "epoch": 2636, "lr": 4.4790515596843715e-06}
{"train_loss": 0.134500652551651, "global_step": 234666, "epoch": 2636, "lr": 4.478811732990662e-06}
{"train_loss": 0.027008825913071632, "global_step": 234667, "epoch": 2636, "lr": 4.478571912416735e-06}
{"train_loss": 0.051998864859342575, "global_step": 234668, "epoch": 2636, "lr": 4.478332097962595e-06}
{"train_loss": 0.039301712065935135, "global_step": 234669, "epoch": 2636, "lr": 4.478092289628305e-06}
{"train_loss": 0.0362548790872097, "global_step": 234670, "epoch": 2636, "lr": 4.477852487413864e-06}
{"train_loss": 0.037807825952768326, "global_step": 234671, "epoch": 2636, "lr": 4.477612691319339e-06}
{"train_loss": 0.019639508798718452, "global_step": 234672, "epoch": 2636, "lr": 4.477372901344728e-06}
{"train_loss": 0.04643998667597771, "global_step": 234673, "epoch": 2636, "lr": 4.4771331174900785e-06}
{"train_loss": 0.0554221011698246, "global_step": 234674, "epoch": 2636, "lr": 4.476893339755434e-06}
{"train_loss": 0.04717061296105385, "global_step": 234675, "epoch": 2636, "lr": 4.47665356814081e-06}
{"train_loss": 0.039822593331336975, "global_step": 234676, "epoch": 2636, "lr": 4.476413802646251e-06}
{"train_loss": 0.07126010209321976, "global_step": 234677, "epoch": 2636, "lr": 4.47617404327178e-06}
{"train_loss": 0.0713053047657013, "global_step": 234678, "epoch": 2636, "lr": 4.4759342900174365e-06}
{"train_loss": 0.06283299624919891, "global_step": 234679, "epoch": 2636, "lr": 4.475694542883241e-06}
{"train_loss": 0.058855388313531876, "global_step": 234680, "epoch": 2636, "lr": 4.475454801869239e-06}
{"train_loss": 0.052963268011808395, "global_step": 234681, "epoch": 2636, "lr": 4.475215066975453e-06}
{"train_loss": 0.0692664086818695, "global_step": 234682, "epoch": 2636, "lr": 4.4749753382019255e-06}
{"train_loss": 0.03372060880064964, "global_step": 234683, "epoch": 2636, "lr": 4.474735615548676e-06}
{"train_loss": 0.06372059881687164, "global_step": 234684, "epoch": 2636, "lr": 4.474495899015751e-06}
{"train_loss": 0.012029626406729221, "global_step": 234685, "epoch": 2636, "lr": 4.474256188603165e-06}
{"train_loss": 0.06582019478082657, "global_step": 234686, "epoch": 2636, "lr": 4.474016484310972e-06}
{"train_loss": 0.057701632380485535, "global_step": 234687, "epoch": 2636, "lr": 4.473776786139183e-06}
{"train_loss": 0.043158143758773804, "global_step": 234688, "epoch": 2636, "lr": 4.473537094087849e-06}
{"train_loss": 0.04014876112341881, "global_step": 234689, "epoch": 2636, "lr": 4.473297408156985e-06}
{"train_loss": 0.040571365505456924, "global_step": 234690, "epoch": 2636, "lr": 4.473057728346636e-06}
{"train_loss": 0.05017704516649246, "global_step": 234691, "epoch": 2636, "lr": 4.472818054656835e-06}
{"train_loss": 0.04645679471491093, "global_step": 234692, "epoch": 2636, "lr": 4.472578387087595e-06, "val_loss": 9.301690101623535}
{"train_loss": 0.05112648010253906, "global_step": 234693, "epoch": 2637, "lr": 4.472338725638975e-06}
{"train_loss": 0.04324145242571831, "global_step": 234694, "epoch": 2637, "lr": 4.472099070310981e-06}
{"train_loss": 0.0396474190056324, "global_step": 234695, "epoch": 2637, "lr": 4.471859421103669e-06}
{"train_loss": 0.1177787110209465, "global_step": 234696, "epoch": 2637, "lr": 4.47161977801705e-06}
{"train_loss": 0.05063915625214577, "global_step": 234697, "epoch": 2637, "lr": 4.471380141051179e-06}
{"train_loss": 0.013646350242197514, "global_step": 234698, "epoch": 2637, "lr": 4.471140510206068e-06}
{"train_loss": 0.07658275961875916, "global_step": 234699, "epoch": 2637, "lr": 4.4709008854817606e-06}
{"train_loss": 0.06041142717003822, "global_step": 234700, "epoch": 2637, "lr": 4.470661266878284e-06}
{"train_loss": 0.054166100919246674, "global_step": 234701, "epoch": 2637, "lr": 4.470421654395673e-06}
{"train_loss": 0.027145227417349815, "global_step": 234702, "epoch": 2637, "lr": 4.470182048033955e-06}
{"train_loss": 0.03242703899741173, "global_step": 234703, "epoch": 2637, "lr": 4.469942447793174e-06}
{"train_loss": 0.045042891055345535, "global_step": 234704, "epoch": 2637, "lr": 4.469702853673346e-06}
{"train_loss": 0.0705372542142868, "global_step": 234705, "epoch": 2637, "lr": 4.469463265674523e-06}
{"train_loss": 0.028719253838062286, "global_step": 234706, "epoch": 2637, "lr": 4.469223683796708e-06}
{"train_loss": 0.08240924030542374, "global_step": 234707, "epoch": 2637, "lr": 4.46898410803997e-06}
{"train_loss": 0.07148932665586472, "global_step": 234708, "epoch": 2637, "lr": 4.468744538404307e-06}
{"train_loss": 0.017922893166542053, "global_step": 234709, "epoch": 2637, "lr": 4.468504974889775e-06}
{"train_loss": 0.02819272130727768, "global_step": 234710, "epoch": 2637, "lr": 4.468265417496392e-06}
{"train_loss": 0.09201773256063461, "global_step": 234711, "epoch": 2637, "lr": 4.468025866224196e-06}
{"train_loss": 0.02294868603348732, "global_step": 234712, "epoch": 2637, "lr": 4.467786321073225e-06}
{"train_loss": 0.0425403006374836, "global_step": 234713, "epoch": 2637, "lr": 4.4675467820434965e-06}
{"train_loss": 0.06827492266893387, "global_step": 234714, "epoch": 2637, "lr": 4.467307249135061e-06}
{"train_loss": 0.024883203208446503, "global_step": 234715, "epoch": 2637, "lr": 4.4670677223479344e-06}
{"train_loss": 0.030079049989581108, "global_step": 234716, "epoch": 2637, "lr": 4.466828201682161e-06}
{"train_loss": 0.026266053318977356, "global_step": 234717, "epoch": 2637, "lr": 4.466588687137763e-06}
{"train_loss": 0.06764821708202362, "global_step": 234718, "epoch": 2637, "lr": 4.466349178714785e-06}
{"train_loss": 0.04616454243659973, "global_step": 234719, "epoch": 2637, "lr": 4.466109676413238e-06}
{"train_loss": 0.019811594858765602, "global_step": 234720, "epoch": 2637, "lr": 4.4658701802331786e-06}
{"train_loss": 0.03903401270508766, "global_step": 234721, "epoch": 2637, "lr": 4.465630690174621e-06}
{"train_loss": 0.02810482122004032, "global_step": 234722, "epoch": 2637, "lr": 4.465391206237612e-06}
{"train_loss": 0.02098807319998741, "global_step": 234723, "epoch": 2637, "lr": 4.465151728422167e-06}
{"train_loss": 0.08971711248159409, "global_step": 234724, "epoch": 2637, "lr": 4.464912256728337e-06}
{"train_loss": 0.029473548755049706, "global_step": 234725, "epoch": 2637, "lr": 4.464672791156133e-06}
{"train_loss": 0.042464859783649445, "global_step": 234726, "epoch": 2637, "lr": 4.464433331705614e-06}
{"train_loss": 0.04418320953845978, "global_step": 234727, "epoch": 2637, "lr": 4.464193878376782e-06}
{"train_loss": 0.04184781014919281, "global_step": 234728, "epoch": 2637, "lr": 4.463954431169692e-06}
{"train_loss": 0.07477046549320221, "global_step": 234729, "epoch": 2637, "lr": 4.463714990084373e-06}
{"train_loss": 0.027313651517033577, "global_step": 234730, "epoch": 2637, "lr": 4.463475555120844e-06}
{"train_loss": 0.07913729548454285, "global_step": 234731, "epoch": 2637, "lr": 4.463236126279152e-06}
{"train_loss": 0.04469366371631622, "global_step": 234732, "epoch": 2637, "lr": 4.4629967035593135e-06}
{"train_loss": 0.049918707460165024, "global_step": 234733, "epoch": 2637, "lr": 4.4627572869613774e-06}
{"train_loss": 0.03584516420960426, "global_step": 234734, "epoch": 2637, "lr": 4.462517876485367e-06}
{"train_loss": 0.07327822595834732, "global_step": 234735, "epoch": 2637, "lr": 4.46227847213132e-06}
{"train_loss": 0.1128644272685051, "global_step": 234736, "epoch": 2637, "lr": 4.462039073899255e-06}
{"train_loss": 0.042591121047735214, "global_step": 234737, "epoch": 2637, "lr": 4.461799681789225e-06}
{"train_loss": 0.037191376090049744, "global_step": 234738, "epoch": 2637, "lr": 4.461560295801237e-06}
{"train_loss": 0.05608951300382614, "global_step": 234739, "epoch": 2637, "lr": 4.461320915935352e-06}
{"train_loss": 0.08453849703073502, "global_step": 234740, "epoch": 2637, "lr": 4.461081542191576e-06}
{"train_loss": 0.049463365226984024, "global_step": 234741, "epoch": 2637, "lr": 4.460842174569962e-06}
{"train_loss": 0.031607650220394135, "global_step": 234742, "epoch": 2637, "lr": 4.460602813070519e-06}
{"train_loss": 0.05854492262005806, "global_step": 234743, "epoch": 2637, "lr": 4.460363457693312e-06}
{"train_loss": 0.05581365525722504, "global_step": 234744, "epoch": 2637, "lr": 4.460124108438335e-06}
{"train_loss": 0.016233228147029877, "global_step": 234745, "epoch": 2637, "lr": 4.459884765305655e-06}
{"train_loss": 0.030420193448662758, "global_step": 234746, "epoch": 2637, "lr": 4.459645428295278e-06}
{"train_loss": 0.030150994658470154, "global_step": 234747, "epoch": 2637, "lr": 4.459406097407243e-06}
{"train_loss": 0.05285647511482239, "global_step": 234748, "epoch": 2637, "lr": 4.459166772641598e-06}
{"train_loss": 0.03179413452744484, "global_step": 234749, "epoch": 2637, "lr": 4.458927453998351e-06}
{"train_loss": 0.023990914225578308, "global_step": 234750, "epoch": 2637, "lr": 4.4586881414775625e-06}
{"train_loss": 0.06762250512838364, "global_step": 234751, "epoch": 2637, "lr": 4.458448835079232e-06}
{"train_loss": 0.03689880296587944, "global_step": 234752, "epoch": 2637, "lr": 4.45820953480342e-06}
{"train_loss": 0.05741504207253456, "global_step": 234753, "epoch": 2637, "lr": 4.457970240650139e-06}
{"train_loss": 0.03640609607100487, "global_step": 234754, "epoch": 2637, "lr": 4.457730952619432e-06}
{"train_loss": 0.028474215418100357, "global_step": 234755, "epoch": 2637, "lr": 4.457491670711323e-06}
{"train_loss": 0.023942505940794945, "global_step": 234756, "epoch": 2637, "lr": 4.457252394925859e-06}
{"train_loss": 0.028364483267068863, "global_step": 234757, "epoch": 2637, "lr": 4.4570131252630545e-06}
{"train_loss": 0.055192627012729645, "global_step": 234758, "epoch": 2637, "lr": 4.456773861722957e-06}
{"train_loss": 0.019442113116383553, "global_step": 234759, "epoch": 2637, "lr": 4.456534604305579e-06}
{"train_loss": 0.03652273491024971, "global_step": 234760, "epoch": 2637, "lr": 4.456295353010975e-06}
{"train_loss": 0.04703088477253914, "global_step": 234761, "epoch": 2637, "lr": 4.456056107839163e-06}
{"train_loss": 0.036840829998254776, "global_step": 234762, "epoch": 2637, "lr": 4.4558168687901805e-06}
{"train_loss": 0.07673336565494537, "global_step": 234763, "epoch": 2637, "lr": 4.455577635864055e-06}
{"train_loss": 0.019817978143692017, "global_step": 234764, "epoch": 2637, "lr": 4.455338409060827e-06}
{"train_loss": 0.01326519064605236, "global_step": 234765, "epoch": 2637, "lr": 4.4550991883805185e-06}
{"train_loss": 0.03713756054639816, "global_step": 234766, "epoch": 2637, "lr": 4.454859973823178e-06}
{"train_loss": 0.04188236594200134, "global_step": 234767, "epoch": 2637, "lr": 4.454620765388811e-06}
{"train_loss": 0.018920566886663437, "global_step": 234768, "epoch": 2637, "lr": 4.454381563077476e-06}
{"train_loss": 0.05303993076086044, "global_step": 234769, "epoch": 2637, "lr": 4.4541423668891966e-06}
{"train_loss": 0.06561467051506042, "global_step": 234770, "epoch": 2637, "lr": 4.453903176823992e-06}
{"train_loss": 0.035722021013498306, "global_step": 234771, "epoch": 2637, "lr": 4.453663992881912e-06}
{"train_loss": 0.04156649485230446, "global_step": 234772, "epoch": 2637, "lr": 4.453424815062979e-06}
{"train_loss": 0.01999310590326786, "global_step": 234773, "epoch": 2637, "lr": 4.45318564336723e-06}
{"train_loss": 0.0307178795337677, "global_step": 234774, "epoch": 2637, "lr": 4.45294647779469e-06}
{"train_loss": 0.0736655667424202, "global_step": 234775, "epoch": 2637, "lr": 4.452707318345401e-06}
{"train_loss": 0.022815583273768425, "global_step": 234776, "epoch": 2637, "lr": 4.452468165019386e-06}
{"train_loss": 0.03972630947828293, "global_step": 234777, "epoch": 2637, "lr": 4.452229017816689e-06}
{"train_loss": 0.03619372099637985, "global_step": 234778, "epoch": 2637, "lr": 4.451989876737328e-06}
{"train_loss": 0.04085677117109299, "global_step": 234779, "epoch": 2637, "lr": 4.451750741781347e-06}
{"train_loss": 0.04307027906179428, "global_step": 234780, "epoch": 2637, "lr": 4.4515116129487665e-06}
{"train_loss": 0.044858199355893594, "global_step": 234781, "epoch": 2637, "lr": 4.451272490239633e-06, "val_loss": 9.290053367614746}
{"train_loss": 0.03692696988582611, "global_step": 234782, "epoch": 2638, "lr": 4.451033373653957e-06}
{"train_loss": 0.02547634206712246, "global_step": 234783, "epoch": 2638, "lr": 4.4507942631917935e-06}
{"train_loss": 0.036119796335697174, "global_step": 234784, "epoch": 2638, "lr": 4.45055515885317e-06}
{"train_loss": 0.05108407512307167, "global_step": 234785, "epoch": 2638, "lr": 4.4503160606381044e-06}
{"train_loss": 0.06090565025806427, "global_step": 234786, "epoch": 2638, "lr": 4.450076968546646e-06}
{"train_loss": 0.0581209771335125, "global_step": 234787, "epoch": 2638, "lr": 4.449837882578817e-06}
{"train_loss": 0.027376241981983185, "global_step": 234788, "epoch": 2638, "lr": 4.449598802734656e-06}
{"train_loss": 0.053519073873758316, "global_step": 234789, "epoch": 2638, "lr": 4.449359729014179e-06}
{"train_loss": 0.04908393323421478, "global_step": 234790, "epoch": 2638, "lr": 4.449120661417444e-06}
{"train_loss": 0.047883059829473495, "global_step": 234791, "epoch": 2638, "lr": 4.448881599944465e-06}
{"train_loss": 0.03525347262620926, "global_step": 234792, "epoch": 2638, "lr": 4.448642544595283e-06}
{"train_loss": 0.04980389401316643, "global_step": 234793, "epoch": 2638, "lr": 4.448403495369913e-06}
{"train_loss": 0.08377132564783096, "global_step": 234794, "epoch": 2638, "lr": 4.4481644522684165e-06}
{"train_loss": 0.032063327729701996, "global_step": 234795, "epoch": 2638, "lr": 4.447925415290794e-06}
{"train_loss": 0.06327646225690842, "global_step": 234796, "epoch": 2638, "lr": 4.447686384437105e-06}
{"train_loss": 0.028217356652021408, "global_step": 234797, "epoch": 2638, "lr": 4.4474473597073576e-06}
{"train_loss": 0.06310564279556274, "global_step": 234798, "epoch": 2638, "lr": 4.447208341101611e-06}
{"train_loss": 0.046720851212739944, "global_step": 234799, "epoch": 2638, "lr": 4.446969328619865e-06}
{"train_loss": 0.042366474866867065, "global_step": 234800, "epoch": 2638, "lr": 4.446730322262172e-06}
{"train_loss": 0.033422037959098816, "global_step": 234801, "epoch": 2638, "lr": 4.446491322028573e-06}
{"train_loss": 0.03936106711626053, "global_step": 234802, "epoch": 2638, "lr": 4.446252327919081e-06}
{"train_loss": 0.07422739267349243, "global_step": 234803, "epoch": 2638, "lr": 4.446013339933736e-06}
{"train_loss": 0.030057106167078018, "global_step": 234804, "epoch": 2638, "lr": 4.445774358072569e-06}
{"train_loss": 0.03338097035884857, "global_step": 234805, "epoch": 2638, "lr": 4.445535382335609e-06}
{"train_loss": 0.018313037231564522, "global_step": 234806, "epoch": 2638, "lr": 4.445296412722899e-06}
{"train_loss": 0.06482043117284775, "global_step": 234807, "epoch": 2638, "lr": 4.445057449234463e-06}
{"train_loss": 0.06544138491153717, "global_step": 234808, "epoch": 2638, "lr": 4.444818491870323e-06}
{"train_loss": 0.04736844077706337, "global_step": 234809, "epoch": 2638, "lr": 4.4445795406305335e-06}
{"train_loss": 0.04808182269334793, "global_step": 234810, "epoch": 2638, "lr": 4.444340595515106e-06}
{"train_loss": 0.07537341862916946, "global_step": 234811, "epoch": 2638, "lr": 4.444101656524091e-06}
{"train_loss": 0.06747160851955414, "global_step": 234812, "epoch": 2638, "lr": 4.4438627236574995e-06}
{"train_loss": 0.055836863815784454, "global_step": 234813, "epoch": 2638, "lr": 4.4436237969153856e-06}
{"train_loss": 0.046309944242239, "global_step": 234814, "epoch": 2638, "lr": 4.443384876297768e-06}
{"train_loss": 0.10364948213100433, "global_step": 234815, "epoch": 2638, "lr": 4.443145961804685e-06}
{"train_loss": 0.0424010343849659, "global_step": 234816, "epoch": 2638, "lr": 4.442907053436157e-06}
{"train_loss": 0.06536200642585754, "global_step": 234817, "epoch": 2638, "lr": 4.442668151192237e-06}
{"train_loss": 0.028501834720373154, "global_step": 234818, "epoch": 2638, "lr": 4.442429255072933e-06}
{"train_loss": 0.029077963903546333, "global_step": 234819, "epoch": 2638, "lr": 4.442190365078286e-06}
{"train_loss": 0.0865878090262413, "global_step": 234820, "epoch": 2638, "lr": 4.441951481208345e-06}
{"train_loss": 0.06414365768432617, "global_step": 234821, "epoch": 2638, "lr": 4.4417126034631215e-06}
{"train_loss": 0.0507621243596077, "global_step": 234822, "epoch": 2638, "lr": 4.44147373184266e-06}
{"train_loss": 0.0193166583776474, "global_step": 234823, "epoch": 2638, "lr": 4.441234866346977e-06}
{"train_loss": 0.017048636451363564, "global_step": 234824, "epoch": 2638, "lr": 4.440996006976128e-06}
{"train_loss": 0.03545292466878891, "global_step": 234825, "epoch": 2638, "lr": 4.4407571537301185e-06}
{"train_loss": 0.05768149346113205, "global_step": 234826, "epoch": 2638, "lr": 4.440518306609004e-06}
{"train_loss": 0.05518389120697975, "global_step": 234827, "epoch": 2638, "lr": 4.440279465612795e-06}
{"train_loss": 0.11351529508829117, "global_step": 234828, "epoch": 2638, "lr": 4.440040630741549e-06}
{"train_loss": 0.06663664430379868, "global_step": 234829, "epoch": 2638, "lr": 4.439801801995275e-06}
{"train_loss": 0.056380078196525574, "global_step": 234830, "epoch": 2638, "lr": 4.439562979374018e-06}
{"train_loss": 0.052516937255859375, "global_step": 234831, "epoch": 2638, "lr": 4.439324162877806e-06}
{"train_loss": 0.04256834834814072, "global_step": 234832, "epoch": 2638, "lr": 4.439085352506672e-06}
{"train_loss": 0.022547448053956032, "global_step": 234833, "epoch": 2638, "lr": 4.438846548260644e-06}
{"train_loss": 0.037441566586494446, "global_step": 234834, "epoch": 2638, "lr": 4.438607750139767e-06}
{"train_loss": 0.08158063888549805, "global_step": 234835, "epoch": 2638, "lr": 4.4383689581440515e-06}
{"train_loss": 0.07026973366737366, "global_step": 234836, "epoch": 2638, "lr": 4.438130172273552e-06}
{"train_loss": 0.04035414010286331, "global_step": 234837, "epoch": 2638, "lr": 4.437891392528281e-06}
{"train_loss": 0.08033443242311478, "global_step": 234838, "epoch": 2638, "lr": 4.437652618908283e-06}
{"train_loss": 0.07472021132707596, "global_step": 234839, "epoch": 2638, "lr": 4.437413851413591e-06}
{"train_loss": 0.03460368886590004, "global_step": 234840, "epoch": 2638, "lr": 4.437175090044238e-06}
{"train_loss": 0.03590495511889458, "global_step": 234841, "epoch": 2638, "lr": 4.43693633480024e-06}
{"train_loss": 0.02533467300236225, "global_step": 234842, "epoch": 2638, "lr": 4.436697585681654e-06}
{"train_loss": 0.041555337607860565, "global_step": 234843, "epoch": 2638, "lr": 4.436458842688485e-06}
{"train_loss": 0.03548840433359146, "global_step": 234844, "epoch": 2638, "lr": 4.436220105820787e-06}
{"train_loss": 0.05680864304304123, "global_step": 234845, "epoch": 2638, "lr": 4.4359813750785905e-06}
{"train_loss": 0.04959965869784355, "global_step": 234846, "epoch": 2638, "lr": 4.435742650461905e-06}
{"train_loss": 0.09909310936927795, "global_step": 234847, "epoch": 2638, "lr": 4.435503931970791e-06}
{"train_loss": 0.05141815170645714, "global_step": 234848, "epoch": 2638, "lr": 4.435265219605262e-06}
{"train_loss": 0.05544440820813179, "global_step": 234849, "epoch": 2638, "lr": 4.435026513365359e-06}
{"train_loss": 0.07455124706029892, "global_step": 234850, "epoch": 2638, "lr": 4.434787813251106e-06}
{"train_loss": 0.04562589153647423, "global_step": 234851, "epoch": 2638, "lr": 4.4345491192625485e-06}
{"train_loss": 0.037500739097595215, "global_step": 234852, "epoch": 2638, "lr": 4.434310431399702e-06}
{"train_loss": 0.026187920942902565, "global_step": 234853, "epoch": 2638, "lr": 4.4340717496626105e-06}
{"train_loss": 0.07824130356311798, "global_step": 234854, "epoch": 2638, "lr": 4.433833074051302e-06}
{"train_loss": 0.04047509655356407, "global_step": 234855, "epoch": 2638, "lr": 4.433594404565805e-06}
{"train_loss": 0.028458932414650917, "global_step": 234856, "epoch": 2638, "lr": 4.433355741206164e-06}
{"train_loss": 0.06053399294614792, "global_step": 234857, "epoch": 2638, "lr": 4.433117083972399e-06}
{"train_loss": 0.020056502893567085, "global_step": 234858, "epoch": 2638, "lr": 4.432878432864551e-06}
{"train_loss": 0.04737885668873787, "global_step": 234859, "epoch": 2638, "lr": 4.4326397878826364e-06}
{"train_loss": 0.07441632449626923, "global_step": 234860, "epoch": 2638, "lr": 4.43240114902671e-06}
{"train_loss": 0.03418660908937454, "global_step": 234861, "epoch": 2638, "lr": 4.432162516296784e-06}
{"train_loss": 0.006681105121970177, "global_step": 234862, "epoch": 2638, "lr": 4.431923889692907e-06}
{"train_loss": 0.047487225383520126, "global_step": 234863, "epoch": 2638, "lr": 4.431685269215091e-06}
{"train_loss": 0.02839890867471695, "global_step": 234864, "epoch": 2638, "lr": 4.431446654863386e-06}
{"train_loss": 0.04198120906949043, "global_step": 234865, "epoch": 2638, "lr": 4.431208046637814e-06}
{"train_loss": 0.019801581278443336, "global_step": 234866, "epoch": 2638, "lr": 4.430969444538418e-06}
{"train_loss": 0.016914747655391693, "global_step": 234867, "epoch": 2638, "lr": 4.430730848565212e-06}
{"train_loss": 0.03281552717089653, "global_step": 234868, "epoch": 2638, "lr": 4.43049225871825e-06}
{"train_loss": 0.11314456164836884, "global_step": 234869, "epoch": 2638, "lr": 4.430253674997547e-06}
{"train_loss": 0.05001210310402211, "global_step": 234870, "epoch": 2638, "lr": 4.430015097403145e-06, "val_loss": 9.25001335144043}
{"train_loss": 0.05367964506149292, "global_step": 234871, "epoch": 2639, "lr": 4.429776525935064e-06}
{"train_loss": 0.03231266513466835, "global_step": 234872, "epoch": 2639, "lr": 4.429537960593355e-06}
{"train_loss": 0.050394922494888306, "global_step": 234873, "epoch": 2639, "lr": 4.429299401378029e-06}
{"train_loss": 0.06226512789726257, "global_step": 234874, "epoch": 2639, "lr": 4.429060848289124e-06}
{"train_loss": 0.052626125514507294, "global_step": 234875, "epoch": 2639, "lr": 4.428822301326691e-06}
{"train_loss": 0.059728048741817474, "global_step": 234876, "epoch": 2639, "lr": 4.428583760490734e-06}
{"train_loss": 0.09158604592084885, "global_step": 234877, "epoch": 2639, "lr": 4.4283452257813105e-06}
{"train_loss": 0.1260785162448883, "global_step": 234878, "epoch": 2639, "lr": 4.428106697198431e-06}
{"train_loss": 0.04482397437095642, "global_step": 234879, "epoch": 2639, "lr": 4.42786817474215e-06}
{"train_loss": 0.022176815196871758, "global_step": 234880, "epoch": 2639, "lr": 4.42762965841248e-06}
{"train_loss": 0.036966681480407715, "global_step": 234881, "epoch": 2639, "lr": 4.427391148209453e-06}
{"train_loss": 0.028315164148807526, "global_step": 234882, "epoch": 2639, "lr": 4.42715264413312e-06}
{"train_loss": 0.06094104424118996, "global_step": 234883, "epoch": 2639, "lr": 4.426914146183492e-06}
{"train_loss": 0.044450271874666214, "global_step": 234884, "epoch": 2639, "lr": 4.426675654360613e-06}
{"train_loss": 0.05527534335851669, "global_step": 234885, "epoch": 2639, "lr": 4.426437168664516e-06}
{"train_loss": 0.03513164818286896, "global_step": 234886, "epoch": 2639, "lr": 4.426198689095218e-06}
{"train_loss": 0.039307016879320145, "global_step": 234887, "epoch": 2639, "lr": 4.425960215652774e-06}
{"train_loss": 0.03200811147689819, "global_step": 234888, "epoch": 2639, "lr": 4.4257217483371915e-06}
{"train_loss": 0.01873116008937359, "global_step": 234889, "epoch": 2639, "lr": 4.42548328714853e-06}
{"train_loss": 0.08362795412540436, "global_step": 234890, "epoch": 2639, "lr": 4.425244832086794e-06}
{"train_loss": 0.03737202659249306, "global_step": 234891, "epoch": 2639, "lr": 4.425006383152025e-06}
{"train_loss": 0.01928219571709633, "global_step": 234892, "epoch": 2639, "lr": 4.424767940344271e-06}
{"train_loss": 0.04753442853689194, "global_step": 234893, "epoch": 2639, "lr": 4.424529503663543e-06}
{"train_loss": 0.03596271947026253, "global_step": 234894, "epoch": 2639, "lr": 4.424291073109893e-06}
{"train_loss": 0.03365286812186241, "global_step": 234895, "epoch": 2639, "lr": 4.42405264868333e-06}
{"train_loss": 0.02875617891550064, "global_step": 234896, "epoch": 2639, "lr": 4.423814230383905e-06}
{"train_loss": 0.044858451932668686, "global_step": 234897, "epoch": 2639, "lr": 4.423575818211634e-06}
{"train_loss": 0.03421279788017273, "global_step": 234898, "epoch": 2639, "lr": 4.423337412166567e-06}
{"train_loss": 0.021154353395104408, "global_step": 234899, "epoch": 2639, "lr": 4.423099012248721e-06}
{"train_loss": 0.12920962274074554, "global_step": 234900, "epoch": 2639, "lr": 4.422860618458141e-06}
{"train_loss": 0.05556821823120117, "global_step": 234901, "epoch": 2639, "lr": 4.422622230794837e-06}
{"train_loss": 0.058224163949489594, "global_step": 234902, "epoch": 2639, "lr": 4.422383849258871e-06}
{"train_loss": 0.04013347625732422, "global_step": 234903, "epoch": 2639, "lr": 4.422145473850253e-06}
{"train_loss": 0.06037522107362747, "global_step": 234904, "epoch": 2639, "lr": 4.421907104569029e-06}
{"train_loss": 0.07381587475538254, "global_step": 234905, "epoch": 2639, "lr": 4.421668741415213e-06}
{"train_loss": 0.03539534658193588, "global_step": 234906, "epoch": 2639, "lr": 4.421430384388858e-06}
{"train_loss": 0.04012094810605049, "global_step": 234907, "epoch": 2639, "lr": 4.421192033489979e-06}
{"train_loss": 0.05357576906681061, "global_step": 234908, "epoch": 2639, "lr": 4.420953688718621e-06}
{"train_loss": 0.05637311562895775, "global_step": 234909, "epoch": 2639, "lr": 4.4207153500748e-06}
{"train_loss": 0.027207007631659508, "global_step": 234910, "epoch": 2639, "lr": 4.420477017558561e-06}
{"train_loss": 0.05867576226592064, "global_step": 234911, "epoch": 2639, "lr": 4.420238691169942e-06}
{"train_loss": 0.06668568402528763, "global_step": 234912, "epoch": 2639, "lr": 4.420000370908961e-06}
{"train_loss": 0.039700862020254135, "global_step": 234913, "epoch": 2639, "lr": 4.419762056775661e-06}
{"train_loss": 0.07423177361488342, "global_step": 234914, "epoch": 2639, "lr": 4.4195237487700605e-06}
{"train_loss": 0.07221642136573792, "global_step": 234915, "epoch": 2639, "lr": 4.419285446892207e-06}
{"train_loss": 0.04790759086608887, "global_step": 234916, "epoch": 2639, "lr": 4.4190471511421135e-06}
{"train_loss": 0.06516385078430176, "global_step": 234917, "epoch": 2639, "lr": 4.418808861519841e-06}
{"train_loss": 0.023323703557252884, "global_step": 234918, "epoch": 2639, "lr": 4.418570578025394e-06}
{"train_loss": 0.02017788589000702, "global_step": 234919, "epoch": 2639, "lr": 4.418332300658812e-06}
{"train_loss": 0.015431039035320282, "global_step": 234920, "epoch": 2639, "lr": 4.418094029420139e-06}
{"train_loss": 0.023627066984772682, "global_step": 234921, "epoch": 2639, "lr": 4.417855764309386e-06}
{"train_loss": 0.05712967738509178, "global_step": 234922, "epoch": 2639, "lr": 4.417617505326604e-06}
{"train_loss": 0.04266301915049553, "global_step": 234923, "epoch": 2639, "lr": 4.41737925247182e-06}
{"train_loss": 0.02988765947520733, "global_step": 234924, "epoch": 2639, "lr": 4.417141005745057e-06}
{"train_loss": 0.033216141164302826, "global_step": 234925, "epoch": 2639, "lr": 4.416902765146364e-06}
{"train_loss": 0.03988881781697273, "global_step": 234926, "epoch": 2639, "lr": 4.416664530675746e-06}
{"train_loss": 0.036512672901153564, "global_step": 234927, "epoch": 2639, "lr": 4.41642630233326e-06}
{"train_loss": 0.044357944279909134, "global_step": 234928, "epoch": 2639, "lr": 4.416188080118933e-06}
{"train_loss": 0.04607904329895973, "global_step": 234929, "epoch": 2639, "lr": 4.415949864032787e-06}
{"train_loss": 0.03005836345255375, "global_step": 234930, "epoch": 2639, "lr": 4.415711654074872e-06}
{"train_loss": 0.0427609458565712, "global_step": 234931, "epoch": 2639, "lr": 4.4154734502452e-06}
{"train_loss": 0.10647036135196686, "global_step": 234932, "epoch": 2639, "lr": 4.41523525254382e-06}
{"train_loss": 0.04272205010056496, "global_step": 234933, "epoch": 2639, "lr": 4.414997060970744e-06}
{"train_loss": 0.07274160534143448, "global_step": 234934, "epoch": 2639, "lr": 4.414758875526026e-06}
{"train_loss": 0.022831818088889122, "global_step": 234935, "epoch": 2639, "lr": 4.414520696209684e-06}
{"train_loss": 0.04065247252583504, "global_step": 234936, "epoch": 2639, "lr": 4.4142825230217565e-06}
{"train_loss": 0.03873966634273529, "global_step": 234937, "epoch": 2639, "lr": 4.4140443559622655e-06}
{"train_loss": 0.0457136295735836, "global_step": 234938, "epoch": 2639, "lr": 4.413806195031261e-06}
{"train_loss": 0.1292390525341034, "global_step": 234939, "epoch": 2639, "lr": 4.41356804022876e-06}
{"train_loss": 0.017692867666482925, "global_step": 234940, "epoch": 2639, "lr": 4.413329891554801e-06}
{"train_loss": 0.0569993257522583, "global_step": 234941, "epoch": 2639, "lr": 4.413091749009407e-06}
{"train_loss": 0.039007674902677536, "global_step": 234942, "epoch": 2639, "lr": 4.412853612592632e-06}
{"train_loss": 0.04275807365775108, "global_step": 234943, "epoch": 2639, "lr": 4.412615482304478e-06}
{"train_loss": 0.05775712803006172, "global_step": 234944, "epoch": 2639, "lr": 4.412377358145003e-06}
{"train_loss": 0.06110071390867233, "global_step": 234945, "epoch": 2639, "lr": 4.412139240114221e-06}
{"train_loss": 0.03949655219912529, "global_step": 234946, "epoch": 2639, "lr": 4.411901128212165e-06}
{"train_loss": 0.0414077527821064, "global_step": 234947, "epoch": 2639, "lr": 4.411663022438889e-06}
{"train_loss": 0.036041926592588425, "global_step": 234948, "epoch": 2639, "lr": 4.411424922794399e-06}
{"train_loss": 0.08444155007600784, "global_step": 234949, "epoch": 2639, "lr": 4.4111868292787465e-06}
{"train_loss": 0.06512436270713806, "global_step": 234950, "epoch": 2639, "lr": 4.41094874189194e-06}
{"train_loss": 0.03908855840563774, "global_step": 234951, "epoch": 2639, "lr": 4.410710660634037e-06}
{"train_loss": 0.09612789005041122, "global_step": 234952, "epoch": 2639, "lr": 4.410472585505055e-06}
{"train_loss": 0.059304799884557724, "global_step": 234953, "epoch": 2639, "lr": 4.410234516505035e-06}
{"train_loss": 0.04150965064764023, "global_step": 234954, "epoch": 2639, "lr": 4.409996453633991e-06}
{"train_loss": 0.03400685638189316, "global_step": 234955, "epoch": 2639, "lr": 4.409758396891977e-06}
{"train_loss": 0.03479732573032379, "global_step": 234956, "epoch": 2639, "lr": 4.409520346279017e-06}
{"train_loss": 0.04789295047521591, "global_step": 234957, "epoch": 2639, "lr": 4.4092823017951364e-06}
{"train_loss": 0.053560130298137665, "global_step": 234958, "epoch": 2639, "lr": 4.4090442634403765e-06}
{"train_loss": 0.04942992555626323, "global_step": 234959, "epoch": 2639, "lr": 4.408806231214757e-06, "val_loss": 9.392586708068848}
{"train_loss": 0.02821047231554985, "global_step": 234960, "epoch": 2640, "lr": 4.408568205118324e-06}
{"train_loss": 0.05646931752562523, "global_step": 234961, "epoch": 2640, "lr": 4.408330185151105e-06}
{"train_loss": 0.046861931681632996, "global_step": 234962, "epoch": 2640, "lr": 4.408092171313122e-06}
{"train_loss": 0.03855350986123085, "global_step": 234963, "epoch": 2640, "lr": 4.4078541636044245e-06}
{"train_loss": 0.04088561609387398, "global_step": 234964, "epoch": 2640, "lr": 4.40761616202503e-06}
{"train_loss": 0.06035115197300911, "global_step": 234965, "epoch": 2640, "lr": 4.407378166574971e-06}
{"train_loss": 0.01739349216222763, "global_step": 234966, "epoch": 2640, "lr": 4.407140177254293e-06}
{"train_loss": 0.04885731637477875, "global_step": 234967, "epoch": 2640, "lr": 4.406902194063012e-06}
{"train_loss": 0.0510249026119709, "global_step": 234968, "epoch": 2640, "lr": 4.406664217001177e-06}
{"train_loss": 0.07739423215389252, "global_step": 234969, "epoch": 2640, "lr": 4.406426246068801e-06}
{"train_loss": 0.05387529730796814, "global_step": 234970, "epoch": 2640, "lr": 4.406188281265932e-06}
{"train_loss": 0.03473697602748871, "global_step": 234971, "epoch": 2640, "lr": 4.405950322592589e-06}
{"train_loss": 0.06490299105644226, "global_step": 234972, "epoch": 2640, "lr": 4.40571237004882e-06}
{"train_loss": 0.03526592627167702, "global_step": 234973, "epoch": 2640, "lr": 4.4054744236346315e-06}
{"train_loss": 0.025530481711030006, "global_step": 234974, "epoch": 2640, "lr": 4.405236483350089e-06}
{"train_loss": 0.040354907512664795, "global_step": 234975, "epoch": 2640, "lr": 4.4049985491951935e-06}
{"train_loss": 0.056830957531929016, "global_step": 234976, "epoch": 2640, "lr": 4.404760621170001e-06}
{"train_loss": 0.035419318825006485, "global_step": 234977, "epoch": 2640, "lr": 4.404522699274521e-06}
{"train_loss": 0.04683653637766838, "global_step": 234978, "epoch": 2640, "lr": 4.404284783508805e-06}
{"train_loss": 0.05595703423023224, "global_step": 234979, "epoch": 2640, "lr": 4.404046873872869e-06}
{"train_loss": 0.0560610368847847, "global_step": 234980, "epoch": 2640, "lr": 4.403808970366768e-06}
{"train_loss": 0.06156885251402855, "global_step": 234981, "epoch": 2640, "lr": 4.403571072990503e-06}
{"train_loss": 0.02845054119825363, "global_step": 234982, "epoch": 2640, "lr": 4.403333181744129e-06}
{"train_loss": 0.045747559517621994, "global_step": 234983, "epoch": 2640, "lr": 4.4030952966276736e-06}
{"train_loss": 0.007796627469360828, "global_step": 234984, "epoch": 2640, "lr": 4.40285741764116e-06}
{"train_loss": 0.04884753376245499, "global_step": 234985, "epoch": 2640, "lr": 4.402619544784636e-06}
{"train_loss": 0.06004032492637634, "global_step": 234986, "epoch": 2640, "lr": 4.402381678058115e-06}
{"train_loss": 0.055026743561029434, "global_step": 234987, "epoch": 2640, "lr": 4.4021438174616454e-06}
{"train_loss": 0.00493102241307497, "global_step": 234988, "epoch": 2640, "lr": 4.4019059629952455e-06}
{"train_loss": 0.03215488791465759, "global_step": 234989, "epoch": 2640, "lr": 4.401668114658958e-06}
{"train_loss": 0.05608208850026131, "global_step": 234990, "epoch": 2640, "lr": 4.401430272452806e-06}
{"train_loss": 0.03656258434057236, "global_step": 234991, "epoch": 2640, "lr": 4.4011924363768335e-06}
{"train_loss": 0.030917352065443993, "global_step": 234992, "epoch": 2640, "lr": 4.400954606431057e-06}
{"train_loss": 0.032698966562747955, "global_step": 234993, "epoch": 2640, "lr": 4.400716782615521e-06}
{"train_loss": 0.011333396658301353, "global_step": 234994, "epoch": 2640, "lr": 4.400478964930249e-06}
{"train_loss": 0.0545409694314003, "global_step": 234995, "epoch": 2640, "lr": 4.400241153375284e-06}
{"train_loss": 0.06612243503332138, "global_step": 234996, "epoch": 2640, "lr": 4.4000033479506486e-06}
{"train_loss": 0.05462625250220299, "global_step": 234997, "epoch": 2640, "lr": 4.399765548656371e-06}
{"train_loss": 0.04173216596245766, "global_step": 234998, "epoch": 2640, "lr": 4.399527755492494e-06}
{"train_loss": 0.09130751341581345, "global_step": 234999, "epoch": 2640, "lr": 4.3992899684590415e-06}
{"train_loss": 0.04474715143442154, "global_step": 235000, "epoch": 2640, "lr": 4.399052187556052e-06}
{"train_loss": 0.06066692993044853, "global_step": 235001, "epoch": 2640, "lr": 4.398814412783553e-06}
{"train_loss": 0.015851743519306183, "global_step": 235002, "epoch": 2640, "lr": 4.398576644141577e-06}
{"train_loss": 0.07178295403718948, "global_step": 235003, "epoch": 2640, "lr": 4.398338881630154e-06}
{"train_loss": 0.05040294677019119, "global_step": 235004, "epoch": 2640, "lr": 4.398101125249327e-06}
{"train_loss": 0.09784749895334244, "global_step": 235005, "epoch": 2640, "lr": 4.397863374999106e-06}
{"train_loss": 0.028127115219831467, "global_step": 235006, "epoch": 2640, "lr": 4.397625630879548e-06}
{"train_loss": 0.074005126953125, "global_step": 235007, "epoch": 2640, "lr": 4.39738789289067e-06}
{"train_loss": 0.034322429448366165, "global_step": 235008, "epoch": 2640, "lr": 4.397150161032509e-06}
{"train_loss": 0.10220427811145782, "global_step": 235009, "epoch": 2640, "lr": 4.396912435305089e-06}
{"train_loss": 0.029016228392720222, "global_step": 235010, "epoch": 2640, "lr": 4.39667471570846e-06}
{"train_loss": 0.05938367173075676, "global_step": 235011, "epoch": 2640, "lr": 4.396437002242626e-06}
{"train_loss": 0.03979387506842613, "global_step": 235012, "epoch": 2640, "lr": 4.396199294907649e-06}
{"train_loss": 0.04229757562279701, "global_step": 235013, "epoch": 2640, "lr": 4.39596159370354e-06}
{"train_loss": 0.04520861431956291, "global_step": 235014, "epoch": 2640, "lr": 4.395723898630344e-06}
{"train_loss": 0.046301040798425674, "global_step": 235015, "epoch": 2640, "lr": 4.395486209688077e-06}
{"train_loss": 0.036084290593862534, "global_step": 235016, "epoch": 2640, "lr": 4.395248526876794e-06}
{"train_loss": 0.05420362576842308, "global_step": 235017, "epoch": 2640, "lr": 4.395010850196501e-06}
{"train_loss": 0.04403986036777496, "global_step": 235018, "epoch": 2640, "lr": 4.394773179647249e-06}
{"train_loss": 0.04733496904373169, "global_step": 235019, "epoch": 2640, "lr": 4.394535515229064e-06}
{"train_loss": 0.027477504685521126, "global_step": 235020, "epoch": 2640, "lr": 4.3942978569419755e-06}
{"train_loss": 0.026889462023973465, "global_step": 235021, "epoch": 2640, "lr": 4.394060204786027e-06}
{"train_loss": 0.03795638307929039, "global_step": 235022, "epoch": 2640, "lr": 4.393822558761229e-06}
{"train_loss": 0.06261628121137619, "global_step": 235023, "epoch": 2640, "lr": 4.393584918867638e-06}
{"train_loss": 0.028588049113750458, "global_step": 235024, "epoch": 2640, "lr": 4.393347285105259e-06}
{"train_loss": 0.06023888289928436, "global_step": 235025, "epoch": 2640, "lr": 4.393109657474154e-06}
{"train_loss": 0.019034141674637794, "global_step": 235026, "epoch": 2640, "lr": 4.392872035974327e-06}
{"train_loss": 0.06296397745609283, "global_step": 235027, "epoch": 2640, "lr": 4.392634420605834e-06}
{"train_loss": 0.02196694351732731, "global_step": 235028, "epoch": 2640, "lr": 4.392396811368682e-06}
{"train_loss": 0.04729459434747696, "global_step": 235029, "epoch": 2640, "lr": 4.39215920826293e-06}
{"train_loss": 0.03984476253390312, "global_step": 235030, "epoch": 2640, "lr": 4.391921611288585e-06}
{"train_loss": 0.020374979823827744, "global_step": 235031, "epoch": 2640, "lr": 4.391684020445697e-06}
{"train_loss": 0.0642252117395401, "global_step": 235032, "epoch": 2640, "lr": 4.391446435734287e-06}
{"train_loss": 0.04122015833854675, "global_step": 235033, "epoch": 2640, "lr": 4.3912088571543995e-06}
{"train_loss": 0.04029097035527229, "global_step": 235034, "epoch": 2640, "lr": 4.390971284706052e-06}
{"train_loss": 0.039777450263500214, "global_step": 235035, "epoch": 2640, "lr": 4.390733718389278e-06}
{"train_loss": 0.027959687635302544, "global_step": 235036, "epoch": 2640, "lr": 4.390496158204127e-06}
{"train_loss": 0.05404960736632347, "global_step": 235037, "epoch": 2640, "lr": 4.3902586041506045e-06}
{"train_loss": 0.04374336823821068, "global_step": 235038, "epoch": 2640, "lr": 4.3900210562287654e-06}
{"train_loss": 0.021971629932522774, "global_step": 235039, "epoch": 2640, "lr": 4.3897835144386225e-06}
{"train_loss": 0.04727659374475479, "global_step": 235040, "epoch": 2640, "lr": 4.389545978780224e-06}
{"train_loss": 0.04141534864902496, "global_step": 235041, "epoch": 2640, "lr": 4.389308449253587e-06}
{"train_loss": 0.0235580001026392, "global_step": 235042, "epoch": 2640, "lr": 4.389070925858762e-06}
{"train_loss": 0.030342238023877144, "global_step": 235043, "epoch": 2640, "lr": 4.388833408595766e-06}
{"train_loss": 0.03501073643565178, "global_step": 235044, "epoch": 2640, "lr": 4.388595897464637e-06}
{"train_loss": 0.06175466999411583, "global_step": 235045, "epoch": 2640, "lr": 4.388358392465402e-06}
{"train_loss": 0.0873592272400856, "global_step": 235046, "epoch": 2640, "lr": 4.3881208935981025e-06}
{"train_loss": 0.05114324763417244, "global_step": 235047, "epoch": 2640, "lr": 4.387883400862758e-06}
{"train_loss": 0.04521006818735198, "global_step": 235048, "epoch": 2640, "lr": 4.3876459142594085e-06, "val_loss": 9.355164527893066, "train_action_mse_error": 20.577281951904297}
{"train_loss": 0.05656858906149864, "global_step": 235049, "epoch": 2641, "lr": 4.387408433788081e-06}
{"train_loss": 0.06815410405397415, "global_step": 235050, "epoch": 2641, "lr": 4.387170959448816e-06}
{"train_loss": 0.04753834009170532, "global_step": 235051, "epoch": 2641, "lr": 4.386933491241635e-06}
{"train_loss": 0.0548279769718647, "global_step": 235052, "epoch": 2641, "lr": 4.3866960291665815e-06}
{"train_loss": 0.04210376366972923, "global_step": 235053, "epoch": 2641, "lr": 4.386458573223673e-06}
{"train_loss": 0.0348643884062767, "global_step": 235054, "epoch": 2641, "lr": 4.386221123412948e-06}
{"train_loss": 0.024052467197179794, "global_step": 235055, "epoch": 2641, "lr": 4.385983679734451e-06}
{"train_loss": 0.054291050881147385, "global_step": 235056, "epoch": 2641, "lr": 4.385746242188194e-06}
{"train_loss": 0.12662562727928162, "global_step": 235057, "epoch": 2641, "lr": 4.38550881077422e-06}
{"train_loss": 0.10144300013780594, "global_step": 235058, "epoch": 2641, "lr": 4.3852713854925565e-06}
{"train_loss": 0.024258330464363098, "global_step": 235059, "epoch": 2641, "lr": 4.385033966343244e-06}
{"train_loss": 0.034729234874248505, "global_step": 235060, "epoch": 2641, "lr": 4.384796553326298e-06}
{"train_loss": 0.03152768313884735, "global_step": 235061, "epoch": 2641, "lr": 4.384559146441769e-06}
{"train_loss": 0.05147910118103027, "global_step": 235062, "epoch": 2641, "lr": 4.384321745689668e-06}
{"train_loss": 0.03262259438633919, "global_step": 235063, "epoch": 2641, "lr": 4.38408435107005e-06}
{"train_loss": 0.04068790376186371, "global_step": 235064, "epoch": 2641, "lr": 4.3838469625829325e-06}
{"train_loss": 0.0478573739528656, "global_step": 235065, "epoch": 2641, "lr": 4.383609580228354e-06}
{"train_loss": 0.05040973052382469, "global_step": 235066, "epoch": 2641, "lr": 4.3833722040063364e-06}
{"train_loss": 0.09314058721065521, "global_step": 235067, "epoch": 2641, "lr": 4.383134833916924e-06}
{"train_loss": 0.07068423181772232, "global_step": 235068, "epoch": 2641, "lr": 4.38289746996014e-06}
{"train_loss": 0.038056906312704086, "global_step": 235069, "epoch": 2641, "lr": 4.382660112136022e-06}
{"train_loss": 0.0378754697740078, "global_step": 235070, "epoch": 2641, "lr": 4.382422760444593e-06}
{"train_loss": 0.07323386520147324, "global_step": 235071, "epoch": 2641, "lr": 4.382185414885903e-06}
{"train_loss": 0.05044954642653465, "global_step": 235072, "epoch": 2641, "lr": 4.381948075459968e-06}
{"train_loss": 0.05185863748192787, "global_step": 235073, "epoch": 2641, "lr": 4.3817107421668165e-06}
{"train_loss": 0.08411602675914764, "global_step": 235074, "epoch": 2641, "lr": 4.3814734150064986e-06}
{"train_loss": 0.014831913635134697, "global_step": 235075, "epoch": 2641, "lr": 4.3812360939790244e-06}
{"train_loss": 0.03578993305563927, "global_step": 235076, "epoch": 2641, "lr": 4.38099877908445e-06}
{"train_loss": 0.040202170610427856, "global_step": 235077, "epoch": 2641, "lr": 4.38076147032278e-06}
{"train_loss": 0.05820512771606445, "global_step": 235078, "epoch": 2641, "lr": 4.380524167694072e-06}
{"train_loss": 0.04282974824309349, "global_step": 235079, "epoch": 2641, "lr": 4.380286871198336e-06}
{"train_loss": 0.06815128028392792, "global_step": 235080, "epoch": 2641, "lr": 4.380049580835627e-06}
{"train_loss": 0.10292395949363708, "global_step": 235081, "epoch": 2641, "lr": 4.379812296605951e-06}
{"train_loss": 0.06372333317995071, "global_step": 235082, "epoch": 2641, "lr": 4.3795750185093685e-06}
{"train_loss": 0.07333490997552872, "global_step": 235083, "epoch": 2641, "lr": 4.379337746545881e-06}
{"train_loss": 0.06396564096212387, "global_step": 235084, "epoch": 2641, "lr": 4.379100480715548e-06}
{"train_loss": 0.05163944512605667, "global_step": 235085, "epoch": 2641, "lr": 4.3788632210183755e-06}
{"train_loss": 0.060036949813365936, "global_step": 235086, "epoch": 2641, "lr": 4.378625967454419e-06}
{"train_loss": 0.032735079526901245, "global_step": 235087, "epoch": 2641, "lr": 4.378388720023696e-06}
{"train_loss": 0.020095406100153923, "global_step": 235088, "epoch": 2641, "lr": 4.37815147872625e-06}
{"train_loss": 0.03083171136677265, "global_step": 235089, "epoch": 2641, "lr": 4.377914243562092e-06}
{"train_loss": 0.05110824108123779, "global_step": 235090, "epoch": 2641, "lr": 4.377677014531279e-06}
{"train_loss": 0.034864623099565506, "global_step": 235091, "epoch": 2641, "lr": 4.377439791633825e-06}
{"train_loss": 0.06356362253427505, "global_step": 235092, "epoch": 2641, "lr": 4.377202574869765e-06}
{"train_loss": 0.03106478601694107, "global_step": 235093, "epoch": 2641, "lr": 4.3769653642391436e-06}
{"train_loss": 0.027556797489523888, "global_step": 235094, "epoch": 2641, "lr": 4.376728159741972e-06}
{"train_loss": 0.05982036516070366, "global_step": 235095, "epoch": 2641, "lr": 4.3764909613783045e-06}
{"train_loss": 0.03755100071430206, "global_step": 235096, "epoch": 2641, "lr": 4.376253769148153e-06}
{"train_loss": 0.03290332853794098, "global_step": 235097, "epoch": 2641, "lr": 4.376016583051567e-06}
{"train_loss": 0.02135312370955944, "global_step": 235098, "epoch": 2641, "lr": 4.3757794030885594e-06}
{"train_loss": 0.034069426357746124, "global_step": 235099, "epoch": 2641, "lr": 4.375542229259183e-06}
{"train_loss": 0.04717954993247986, "global_step": 235100, "epoch": 2641, "lr": 4.3753050615634504e-06}
{"train_loss": 0.045044347643852234, "global_step": 235101, "epoch": 2641, "lr": 4.375067900001412e-06}
{"train_loss": 0.01403596717864275, "global_step": 235102, "epoch": 2641, "lr": 4.374830744573078e-06}
{"train_loss": 0.030385691672563553, "global_step": 235103, "epoch": 2641, "lr": 4.374593595278498e-06}
{"train_loss": 0.06041104719042778, "global_step": 235104, "epoch": 2641, "lr": 4.374356452117695e-06}
{"train_loss": 0.06221437826752663, "global_step": 235105, "epoch": 2641, "lr": 4.374119315090708e-06}
{"train_loss": 0.032452069222927094, "global_step": 235106, "epoch": 2641, "lr": 4.373882184197559e-06}
{"train_loss": 0.025122029706835747, "global_step": 235107, "epoch": 2641, "lr": 4.373645059438291e-06}
{"train_loss": 0.05241260305047035, "global_step": 235108, "epoch": 2641, "lr": 4.373407940812924e-06}
{"train_loss": 0.08304288238286972, "global_step": 235109, "epoch": 2641, "lr": 4.3731708283215044e-06}
{"train_loss": 0.05200690031051636, "global_step": 235110, "epoch": 2641, "lr": 4.372933721964045e-06}
{"train_loss": 0.0229820404201746, "global_step": 235111, "epoch": 2641, "lr": 4.3726966217405965e-06}
{"train_loss": 0.017919808626174927, "global_step": 235112, "epoch": 2641, "lr": 4.372459527651185e-06}
{"train_loss": 0.06900060921907425, "global_step": 235113, "epoch": 2641, "lr": 4.372222439695833e-06}
{"train_loss": 0.041414469480514526, "global_step": 235114, "epoch": 2641, "lr": 4.371985357874586e-06}
{"train_loss": 0.038265012204647064, "global_step": 235115, "epoch": 2641, "lr": 4.371748282187466e-06}
{"train_loss": 0.04168589413166046, "global_step": 235116, "epoch": 2641, "lr": 4.371511212634511e-06}
{"train_loss": 0.03216122090816498, "global_step": 235117, "epoch": 2641, "lr": 4.371274149215743e-06}
{"train_loss": 0.049524419009685516, "global_step": 235118, "epoch": 2641, "lr": 4.371037091931207e-06}
{"train_loss": 0.043533049523830414, "global_step": 235119, "epoch": 2641, "lr": 4.370800040780926e-06}
{"train_loss": 0.04303558170795441, "global_step": 235120, "epoch": 2641, "lr": 4.370562995764943e-06}
{"train_loss": 0.028908710926771164, "global_step": 235121, "epoch": 2641, "lr": 4.37032595688327e-06}
{"train_loss": 0.020282423123717308, "global_step": 235122, "epoch": 2641, "lr": 4.370088924135956e-06}
{"train_loss": 0.03823159262537956, "global_step": 235123, "epoch": 2641, "lr": 4.369851897523025e-06}
{"train_loss": 0.025255650281906128, "global_step": 235124, "epoch": 2641, "lr": 4.369614877044514e-06}
{"train_loss": 0.059012677520513535, "global_step": 235125, "epoch": 2641, "lr": 4.369377862700453e-06}
{"train_loss": 0.06704182922840118, "global_step": 235126, "epoch": 2641, "lr": 4.369140854490872e-06}
{"train_loss": 0.04226379096508026, "global_step": 235127, "epoch": 2641, "lr": 4.368903852415801e-06}
{"train_loss": 0.0416153147816658, "global_step": 235128, "epoch": 2641, "lr": 4.368666856475273e-06}
{"train_loss": 0.07499682903289795, "global_step": 235129, "epoch": 2641, "lr": 4.368429866669327e-06}
{"train_loss": 0.04320318624377251, "global_step": 235130, "epoch": 2641, "lr": 4.3681928829979846e-06}
{"train_loss": 0.035441406071186066, "global_step": 235131, "epoch": 2641, "lr": 4.367955905461291e-06}
{"train_loss": 0.027830898761749268, "global_step": 235132, "epoch": 2641, "lr": 4.367718934059256e-06}
{"train_loss": 0.05729887634515762, "global_step": 235133, "epoch": 2641, "lr": 4.367481968791942e-06}
{"train_loss": 0.019418638199567795, "global_step": 235134, "epoch": 2641, "lr": 4.367245009659349e-06}
{"train_loss": 0.04776942729949951, "global_step": 235135, "epoch": 2641, "lr": 4.367008056661536e-06}
{"train_loss": 0.03485746681690216, "global_step": 235136, "epoch": 2641, "lr": 4.366771109798512e-06}
{"train_loss": 0.04718484692903382, "global_step": 235137, "epoch": 2641, "lr": 4.366534169070324e-06, "val_loss": 9.474076271057129}
{"train_loss": 0.04484639689326286, "global_step": 235138, "epoch": 2642, "lr": 4.3662972344769955e-06}
{"train_loss": 0.0559377484023571, "global_step": 235139, "epoch": 2642, "lr": 4.366060306018571e-06}
{"train_loss": 0.05093437433242798, "global_step": 235140, "epoch": 2642, "lr": 4.365823383695067e-06}
{"train_loss": 0.025035807862877846, "global_step": 235141, "epoch": 2642, "lr": 4.365586467506527e-06}
{"train_loss": 0.04448854178190231, "global_step": 235142, "epoch": 2642, "lr": 4.365349557452969e-06}
{"train_loss": 0.03379154950380325, "global_step": 235143, "epoch": 2642, "lr": 4.365112653534442e-06}
{"train_loss": 0.028858093544840813, "global_step": 235144, "epoch": 2642, "lr": 4.364875755750963e-06}
{"train_loss": 0.027527382597327232, "global_step": 235145, "epoch": 2642, "lr": 4.3646388641025774e-06}
{"train_loss": 0.048096202313899994, "global_step": 235146, "epoch": 2642, "lr": 4.3644019785893e-06}
{"train_loss": 0.033729467540979385, "global_step": 235147, "epoch": 2642, "lr": 4.364165099211182e-06}
{"train_loss": 0.06105060875415802, "global_step": 235148, "epoch": 2642, "lr": 4.363928225968239e-06}
{"train_loss": 0.05767570808529854, "global_step": 235149, "epoch": 2642, "lr": 4.3636913588605165e-06}
{"train_loss": 0.044254861772060394, "global_step": 235150, "epoch": 2642, "lr": 4.363454497888036e-06}
{"train_loss": 0.02766023389995098, "global_step": 235151, "epoch": 2642, "lr": 4.363217643050832e-06}
{"train_loss": 0.03249519690871239, "global_step": 235152, "epoch": 2642, "lr": 4.362980794348942e-06}
{"train_loss": 0.06629366427659988, "global_step": 235153, "epoch": 2642, "lr": 4.362743951782383e-06}
{"train_loss": 0.0371023491024971, "global_step": 235154, "epoch": 2642, "lr": 4.362507115351205e-06}
{"train_loss": 0.039164330810308456, "global_step": 235155, "epoch": 2642, "lr": 4.362270285055425e-06}
{"train_loss": 0.04172937572002411, "global_step": 235156, "epoch": 2642, "lr": 4.362033460895093e-06}
{"train_loss": 0.021975994110107422, "global_step": 235157, "epoch": 2642, "lr": 4.361796642870214e-06}
{"train_loss": 0.05157966911792755, "global_step": 235158, "epoch": 2642, "lr": 4.361559830980849e-06}
{"train_loss": 0.052482202649116516, "global_step": 235159, "epoch": 2642, "lr": 4.361323025227004e-06}
{"train_loss": 0.033190157264471054, "global_step": 235160, "epoch": 2642, "lr": 4.361086225608735e-06}
{"train_loss": 0.06186237558722496, "global_step": 235161, "epoch": 2642, "lr": 4.360849432126052e-06}
{"train_loss": 0.06613555550575256, "global_step": 235162, "epoch": 2642, "lr": 4.360612644779005e-06}
{"train_loss": 0.06397900730371475, "global_step": 235163, "epoch": 2642, "lr": 4.360375863567606e-06}
{"train_loss": 0.03281604126095772, "global_step": 235164, "epoch": 2642, "lr": 4.3601390884918995e-06}
{"train_loss": 0.05600697919726372, "global_step": 235165, "epoch": 2642, "lr": 4.359902319551928e-06}
{"train_loss": 0.04154249653220177, "global_step": 235166, "epoch": 2642, "lr": 4.3596655567477e-06}
{"train_loss": 0.05158289894461632, "global_step": 235167, "epoch": 2642, "lr": 4.359428800079268e-06}
{"train_loss": 0.05394743010401726, "global_step": 235168, "epoch": 2642, "lr": 4.359192049546645e-06}
{"train_loss": 0.04719929397106171, "global_step": 235169, "epoch": 2642, "lr": 4.358955305149881e-06}
{"train_loss": 0.06849461048841476, "global_step": 235170, "epoch": 2642, "lr": 4.358718566888992e-06}
{"train_loss": 0.046091604977846146, "global_step": 235171, "epoch": 2642, "lr": 4.358481834764022e-06}
{"train_loss": 0.05070878565311432, "global_step": 235172, "epoch": 2642, "lr": 4.358245108774994e-06}
{"train_loss": 0.058670856058597565, "global_step": 235173, "epoch": 2642, "lr": 4.358008388921947e-06}
{"train_loss": 0.08155372738838196, "global_step": 235174, "epoch": 2642, "lr": 4.357771675204902e-06}
{"train_loss": 0.0882527232170105, "global_step": 235175, "epoch": 2642, "lr": 4.357534967623911e-06}
{"train_loss": 0.013481528498232365, "global_step": 235176, "epoch": 2642, "lr": 4.357298266178983e-06}
{"train_loss": 0.04254816472530365, "global_step": 235177, "epoch": 2642, "lr": 4.357061570870169e-06}
{"train_loss": 0.046639397740364075, "global_step": 235178, "epoch": 2642, "lr": 4.35682488169748e-06}
{"train_loss": 0.046696554869413376, "global_step": 235179, "epoch": 2642, "lr": 4.356588198660972e-06}
{"train_loss": 0.0299922376871109, "global_step": 235180, "epoch": 2642, "lr": 4.356351521760654e-06}
{"train_loss": 0.029120055958628654, "global_step": 235181, "epoch": 2642, "lr": 4.356114850996567e-06}
{"train_loss": 0.029445968568325043, "global_step": 235182, "epoch": 2642, "lr": 4.355878186368756e-06}
{"train_loss": 0.06106221675872803, "global_step": 235183, "epoch": 2642, "lr": 4.3556415278772354e-06}
{"train_loss": 0.015532963909208775, "global_step": 235184, "epoch": 2642, "lr": 4.35540487552204e-06}
{"train_loss": 0.05867648869752884, "global_step": 235185, "epoch": 2642, "lr": 4.355168229303208e-06}
{"train_loss": 0.04821787402033806, "global_step": 235186, "epoch": 2642, "lr": 4.354931589220762e-06}
{"train_loss": 0.04144531115889549, "global_step": 235187, "epoch": 2642, "lr": 4.354694955274747e-06}
{"train_loss": 0.028147153556346893, "global_step": 235188, "epoch": 2642, "lr": 4.3544583274651826e-06}
{"train_loss": 0.012701817788183689, "global_step": 235189, "epoch": 2642, "lr": 4.3542217057920995e-06}
{"train_loss": 0.04585955664515495, "global_step": 235190, "epoch": 2642, "lr": 4.3539850902555465e-06}
{"train_loss": 0.08409886062145233, "global_step": 235191, "epoch": 2642, "lr": 4.3537484808555355e-06}
{"train_loss": 0.07830992341041565, "global_step": 235192, "epoch": 2642, "lr": 4.353511877592109e-06}
{"train_loss": 0.04158348590135574, "global_step": 235193, "epoch": 2642, "lr": 4.353275280465291e-06}
{"train_loss": 0.0529368594288826, "global_step": 235194, "epoch": 2642, "lr": 4.353038689475131e-06}
{"train_loss": 0.04934235289692879, "global_step": 235195, "epoch": 2642, "lr": 4.352802104621634e-06}
{"train_loss": 0.0694088265299797, "global_step": 235196, "epoch": 2642, "lr": 4.352565525904861e-06}
{"train_loss": 0.017412027344107628, "global_step": 235197, "epoch": 2642, "lr": 4.3523289533248135e-06}
{"train_loss": 0.0357414074242115, "global_step": 235198, "epoch": 2642, "lr": 4.352092386881551e-06}
{"train_loss": 0.04184095934033394, "global_step": 235199, "epoch": 2642, "lr": 4.351855826575086e-06}
{"train_loss": 0.03733532130718231, "global_step": 235200, "epoch": 2642, "lr": 4.351619272405461e-06}
{"train_loss": 0.05730707198381424, "global_step": 235201, "epoch": 2642, "lr": 4.351382724372705e-06}
{"train_loss": 0.07760769873857498, "global_step": 235202, "epoch": 2642, "lr": 4.351146182476845e-06}
{"train_loss": 0.035698942840099335, "global_step": 235203, "epoch": 2642, "lr": 4.350909646717926e-06}
{"train_loss": 0.05956023558974266, "global_step": 235204, "epoch": 2642, "lr": 4.35067311709596e-06}
{"train_loss": 0.024351609870791435, "global_step": 235205, "epoch": 2642, "lr": 4.350436593611001e-06}
{"train_loss": 0.05615212023258209, "global_step": 235206, "epoch": 2642, "lr": 4.3502000762630555e-06}
{"train_loss": 0.04657753184437752, "global_step": 235207, "epoch": 2642, "lr": 4.3499635650521846e-06}
{"train_loss": 0.06941636651754379, "global_step": 235208, "epoch": 2642, "lr": 4.3497270599783926e-06}
{"train_loss": 0.1097305566072464, "global_step": 235209, "epoch": 2642, "lr": 4.34949056104173e-06}
{"train_loss": 0.02565770596265793, "global_step": 235210, "epoch": 2642, "lr": 4.34925406824222e-06}
{"train_loss": 0.041244566440582275, "global_step": 235211, "epoch": 2642, "lr": 4.3490175815799e-06}
{"train_loss": 0.039499279111623764, "global_step": 235212, "epoch": 2642, "lr": 4.348781101054788e-06}
{"train_loss": 0.05882382392883301, "global_step": 235213, "epoch": 2642, "lr": 4.348544626666939e-06}
{"train_loss": 0.0452902615070343, "global_step": 235214, "epoch": 2642, "lr": 4.348308158416359e-06}
{"train_loss": 0.02528683841228485, "global_step": 235215, "epoch": 2642, "lr": 4.348071696303108e-06}
{"train_loss": 0.05802803859114647, "global_step": 235216, "epoch": 2642, "lr": 4.347835240327186e-06}
{"train_loss": 0.031912557780742645, "global_step": 235217, "epoch": 2642, "lr": 4.347598790488655e-06}
{"train_loss": 0.08257852494716644, "global_step": 235218, "epoch": 2642, "lr": 4.34736234678752e-06}
{"train_loss": 0.061001524329185486, "global_step": 235219, "epoch": 2642, "lr": 4.347125909223826e-06}
{"train_loss": 0.03977098688483238, "global_step": 235220, "epoch": 2642, "lr": 4.346889477797617e-06}
{"train_loss": 0.02553088776767254, "global_step": 235221, "epoch": 2642, "lr": 4.346653052508903e-06}
{"train_loss": 0.07471219450235367, "global_step": 235222, "epoch": 2642, "lr": 4.346416633357731e-06}
{"train_loss": 0.015775198116898537, "global_step": 235223, "epoch": 2642, "lr": 4.346180220344126e-06}
{"train_loss": 0.0482788048684597, "global_step": 235224, "epoch": 2642, "lr": 4.345943813468118e-06}
{"train_loss": 0.06537298113107681, "global_step": 235225, "epoch": 2642, "lr": 4.345707412729743e-06}
{"train_loss": 0.04764376904062006, "global_step": 235226, "epoch": 2642, "lr": 4.345471018129027e-06, "val_loss": 9.440858840942383}
{"train_loss": 0.04867686703801155, "global_step": 235227, "epoch": 2643, "lr": 4.3452346296660115e-06}
{"train_loss": 0.03773031383752823, "global_step": 235228, "epoch": 2643, "lr": 4.344998247340726e-06}
{"train_loss": 0.05453957989811897, "global_step": 235229, "epoch": 2643, "lr": 4.344761871153192e-06}
{"train_loss": 0.055502887815237045, "global_step": 235230, "epoch": 2643, "lr": 4.344525501103452e-06}
{"train_loss": 0.046376846730709076, "global_step": 235231, "epoch": 2643, "lr": 4.344289137191526e-06}
{"train_loss": 0.06772171705961227, "global_step": 235232, "epoch": 2643, "lr": 4.344052779417462e-06}
{"train_loss": 0.04402697831392288, "global_step": 235233, "epoch": 2643, "lr": 4.343816427781278e-06}
{"train_loss": 0.04672776162624359, "global_step": 235234, "epoch": 2643, "lr": 4.343580082283022e-06}
{"train_loss": 0.053072866052389145, "global_step": 235235, "epoch": 2643, "lr": 4.343343742922701e-06}
{"train_loss": 0.009115973487496376, "global_step": 235236, "epoch": 2643, "lr": 4.343107409700364e-06}
{"train_loss": 0.06502126902341843, "global_step": 235237, "epoch": 2643, "lr": 4.342871082616046e-06}
{"train_loss": 0.030523104593157768, "global_step": 235238, "epoch": 2643, "lr": 4.3426347616697675e-06}
{"train_loss": 0.09882063418626785, "global_step": 235239, "epoch": 2643, "lr": 4.342398446861573e-06}
{"train_loss": 0.025691557675600052, "global_step": 235240, "epoch": 2643, "lr": 4.3421621381914745e-06}
{"train_loss": 0.01377891842275858, "global_step": 235241, "epoch": 2643, "lr": 4.3419258356595275e-06}
{"train_loss": 0.06727555394172668, "global_step": 235242, "epoch": 2643, "lr": 4.341689539265742e-06}
{"train_loss": 0.060404952615499496, "global_step": 235243, "epoch": 2643, "lr": 4.341453249010169e-06}
{"train_loss": 0.03157635033130646, "global_step": 235244, "epoch": 2643, "lr": 4.3412169648928246e-06}
{"train_loss": 0.04483707994222641, "global_step": 235245, "epoch": 2643, "lr": 4.340980686913754e-06}
{"train_loss": 0.0854533314704895, "global_step": 235246, "epoch": 2643, "lr": 4.340744415072972e-06}
{"train_loss": 0.010684261098504066, "global_step": 235247, "epoch": 2643, "lr": 4.340508149370531e-06}
{"train_loss": 0.030317645519971848, "global_step": 235248, "epoch": 2643, "lr": 4.340271889806441e-06}
{"train_loss": 0.028393760323524475, "global_step": 235249, "epoch": 2643, "lr": 4.3400356363807575e-06}
{"train_loss": 0.02088233269751072, "global_step": 235250, "epoch": 2643, "lr": 4.339799389093491e-06}
{"train_loss": 0.040878117084503174, "global_step": 235251, "epoch": 2643, "lr": 4.339563147944687e-06}
{"train_loss": 0.04174204543232918, "global_step": 235252, "epoch": 2643, "lr": 4.339326912934366e-06}
{"train_loss": 0.04078003019094467, "global_step": 235253, "epoch": 2643, "lr": 4.339090684062575e-06}
{"train_loss": 0.04012920707464218, "global_step": 235254, "epoch": 2643, "lr": 4.338854461329322e-06}
{"train_loss": 0.0829136073589325, "global_step": 235255, "epoch": 2643, "lr": 4.3386182447346604e-06}
{"train_loss": 0.024377740919589996, "global_step": 235256, "epoch": 2643, "lr": 4.338382034278621e-06}
{"train_loss": 0.07439479976892471, "global_step": 235257, "epoch": 2643, "lr": 4.338145829961221e-06}
{"train_loss": 0.02496633306145668, "global_step": 235258, "epoch": 2643, "lr": 4.3379096317825105e-06}
{"train_loss": 0.02826603688299656, "global_step": 235259, "epoch": 2643, "lr": 4.337673439742501e-06}
{"train_loss": 0.015983371064066887, "global_step": 235260, "epoch": 2643, "lr": 4.337437253841247e-06}
{"train_loss": 0.05547085404396057, "global_step": 235261, "epoch": 2643, "lr": 4.337201074078767e-06}
{"train_loss": 0.03039952740073204, "global_step": 235262, "epoch": 2643, "lr": 4.3369649004550815e-06}
{"train_loss": 0.018880192190408707, "global_step": 235263, "epoch": 2643, "lr": 4.3367287329702466e-06}
{"train_loss": 0.04504340514540672, "global_step": 235264, "epoch": 2643, "lr": 4.336492571624274e-06}
{"train_loss": 0.036176953464746475, "global_step": 235265, "epoch": 2643, "lr": 4.336256416417206e-06}
{"train_loss": 0.00918680988252163, "global_step": 235266, "epoch": 2643, "lr": 4.336020267349078e-06}
{"train_loss": 0.05271247774362564, "global_step": 235267, "epoch": 2643, "lr": 4.335784124419906e-06}
{"train_loss": 0.039411354809999466, "global_step": 235268, "epoch": 2643, "lr": 4.33554798762974e-06}
{"train_loss": 0.05012986809015274, "global_step": 235269, "epoch": 2643, "lr": 4.335311856978591e-06}
{"train_loss": 0.05947357788681984, "global_step": 235270, "epoch": 2643, "lr": 4.335075732466515e-06}
{"train_loss": 0.03532934561371803, "global_step": 235271, "epoch": 2643, "lr": 4.334839614093522e-06}
{"train_loss": 0.05486961826682091, "global_step": 235272, "epoch": 2643, "lr": 4.334603501859652e-06}
{"train_loss": 0.030488504096865654, "global_step": 235273, "epoch": 2643, "lr": 4.334367395764949e-06}
{"train_loss": 0.031607262790203094, "global_step": 235274, "epoch": 2643, "lr": 4.334131295809418e-06}
{"train_loss": 0.052468497306108475, "global_step": 235275, "epoch": 2643, "lr": 4.3338952019931215e-06}
{"train_loss": 0.04919144883751869, "global_step": 235276, "epoch": 2643, "lr": 4.333659114316069e-06}
{"train_loss": 0.04009831324219704, "global_step": 235277, "epoch": 2643, "lr": 4.333423032778305e-06}
{"train_loss": 0.03352401778101921, "global_step": 235278, "epoch": 2643, "lr": 4.333186957379848e-06}
{"train_loss": 0.053173430263996124, "global_step": 235279, "epoch": 2643, "lr": 4.332950888120746e-06}
{"train_loss": 0.02571525052189827, "global_step": 235280, "epoch": 2643, "lr": 4.3327148250010104e-06}
{"train_loss": 0.01603301800787449, "global_step": 235281, "epoch": 2643, "lr": 4.332478768020698e-06}
{"train_loss": 0.06063966453075409, "global_step": 235282, "epoch": 2643, "lr": 4.332242717179819e-06}
{"train_loss": 0.07134659588336945, "global_step": 235283, "epoch": 2643, "lr": 4.332006672478417e-06}
{"train_loss": 0.07218766957521439, "global_step": 235284, "epoch": 2643, "lr": 4.331770633916515e-06}
{"train_loss": 0.05572527274489403, "global_step": 235285, "epoch": 2643, "lr": 4.331534601494159e-06}
{"train_loss": 0.04670550674200058, "global_step": 235286, "epoch": 2643, "lr": 4.331298575211357e-06}
{"train_loss": 0.06022120639681816, "global_step": 235287, "epoch": 2643, "lr": 4.3310625550681715e-06}
{"train_loss": 0.04441903904080391, "global_step": 235288, "epoch": 2643, "lr": 4.330826541064603e-06}
{"train_loss": 0.049978919327259064, "global_step": 235289, "epoch": 2643, "lr": 4.330590533200712e-06}
{"train_loss": 0.04506034776568413, "global_step": 235290, "epoch": 2643, "lr": 4.330354531476505e-06}
{"train_loss": 0.0599183663725853, "global_step": 235291, "epoch": 2643, "lr": 4.33011853589203e-06}
{"train_loss": 0.03589579090476036, "global_step": 235292, "epoch": 2643, "lr": 4.329882546447317e-06}
{"train_loss": 0.045046113431453705, "global_step": 235293, "epoch": 2643, "lr": 4.329646563142387e-06}
{"train_loss": 0.04983043670654297, "global_step": 235294, "epoch": 2643, "lr": 4.329410585977289e-06}
{"train_loss": 0.04106331616640091, "global_step": 235295, "epoch": 2643, "lr": 4.329174614952036e-06}
{"train_loss": 0.02957400493323803, "global_step": 235296, "epoch": 2643, "lr": 4.328938650066677e-06}
{"train_loss": 0.057105276733636856, "global_step": 235297, "epoch": 2643, "lr": 4.328702691321229e-06}
{"train_loss": 0.03816382214426994, "global_step": 235298, "epoch": 2643, "lr": 4.3284667387157365e-06}
{"train_loss": 0.10579746216535568, "global_step": 235299, "epoch": 2643, "lr": 4.3282307922502265e-06}
{"train_loss": 0.06768477708101273, "global_step": 235300, "epoch": 2643, "lr": 4.327994851924722e-06}
{"train_loss": 0.034367673099040985, "global_step": 235301, "epoch": 2643, "lr": 4.327758917739266e-06}
{"train_loss": 0.06550249457359314, "global_step": 235302, "epoch": 2643, "lr": 4.3275229896938825e-06}
{"train_loss": 0.055120475590229034, "global_step": 235303, "epoch": 2643, "lr": 4.32728706778861e-06}
{"train_loss": 0.056571751832962036, "global_step": 235304, "epoch": 2643, "lr": 4.327051152023482e-06}
{"train_loss": 0.03404249623417854, "global_step": 235305, "epoch": 2643, "lr": 4.326815242398513e-06}
{"train_loss": 0.09275304526090622, "global_step": 235306, "epoch": 2643, "lr": 4.326579338913761e-06}
{"train_loss": 0.05392710119485855, "global_step": 235307, "epoch": 2643, "lr": 4.326343441569231e-06}
{"train_loss": 0.021604860201478004, "global_step": 235308, "epoch": 2643, "lr": 4.326107550364966e-06}
{"train_loss": 0.04784780740737915, "global_step": 235309, "epoch": 2643, "lr": 4.325871665301012e-06}
{"train_loss": 0.028976982459425926, "global_step": 235310, "epoch": 2643, "lr": 4.325635786377374e-06}
{"train_loss": 0.08028779178857803, "global_step": 235311, "epoch": 2643, "lr": 4.325399913594113e-06}
{"train_loss": 0.03810305520892143, "global_step": 235312, "epoch": 2643, "lr": 4.32516404695123e-06}
{"train_loss": 0.023010065779089928, "global_step": 235313, "epoch": 2643, "lr": 4.324928186448785e-06}
{"train_loss": 0.07605046778917313, "global_step": 235314, "epoch": 2643, "lr": 4.324692332086788e-06}
{"train_loss": 0.04583152248576451, "global_step": 235315, "epoch": 2643, "lr": 4.324456483865285e-06, "val_loss": 9.356439590454102}
{"train_loss": 0.056361258029937744, "global_step": 235316, "epoch": 2644, "lr": 4.324220641784293e-06}
{"train_loss": 0.03378855064511299, "global_step": 235317, "epoch": 2644, "lr": 4.323984805843867e-06}
{"train_loss": 0.07081872969865799, "global_step": 235318, "epoch": 2644, "lr": 4.323748976044012e-06}
{"train_loss": 0.08380609005689621, "global_step": 235319, "epoch": 2644, "lr": 4.323513152384784e-06}
{"train_loss": 0.027115952223539352, "global_step": 235320, "epoch": 2644, "lr": 4.3232773348661885e-06}
{"train_loss": 0.050031889230012894, "global_step": 235321, "epoch": 2644, "lr": 4.323041523488286e-06}
{"train_loss": 0.03169538825750351, "global_step": 235322, "epoch": 2644, "lr": 4.322805718251083e-06}
{"train_loss": 0.00913260132074356, "global_step": 235323, "epoch": 2644, "lr": 4.322569919154629e-06}
{"train_loss": 0.059528350830078125, "global_step": 235324, "epoch": 2644, "lr": 4.3223341261989405e-06}
{"train_loss": 0.07442917674779892, "global_step": 235325, "epoch": 2644, "lr": 4.322098339384067e-06}
{"train_loss": 0.030881710350513458, "global_step": 235326, "epoch": 2644, "lr": 4.321862558710021e-06}
{"train_loss": 0.044614922255277634, "global_step": 235327, "epoch": 2644, "lr": 4.321626784176847e-06}
{"train_loss": 0.016699377447366714, "global_step": 235328, "epoch": 2644, "lr": 4.321391015784582e-06}
{"train_loss": 0.013613353483378887, "global_step": 235329, "epoch": 2644, "lr": 4.321155253533238e-06}
{"train_loss": 0.04307852312922478, "global_step": 235330, "epoch": 2644, "lr": 4.320919497422865e-06}
{"train_loss": 0.07094789296388626, "global_step": 235331, "epoch": 2644, "lr": 4.32068374745348e-06}
{"train_loss": 0.07448979467153549, "global_step": 235332, "epoch": 2644, "lr": 4.320448003625133e-06}
{"train_loss": 0.01842854730784893, "global_step": 235333, "epoch": 2644, "lr": 4.3202122659378345e-06}
{"train_loss": 0.06369225680828094, "global_step": 235334, "epoch": 2644, "lr": 4.31997653439164e-06}
{"train_loss": 0.03879701718688011, "global_step": 235335, "epoch": 2644, "lr": 4.3197408089865555e-06}
{"train_loss": 0.05403389781713486, "global_step": 235336, "epoch": 2644, "lr": 4.319505089722631e-06}
{"train_loss": 0.035137612372636795, "global_step": 235337, "epoch": 2644, "lr": 4.319269376599888e-06}
{"train_loss": 0.054343000054359436, "global_step": 235338, "epoch": 2644, "lr": 4.319033669618372e-06}
{"train_loss": 0.04497351124882698, "global_step": 235339, "epoch": 2644, "lr": 4.318797968778099e-06}
{"train_loss": 0.03293795883655548, "global_step": 235340, "epoch": 2644, "lr": 4.3185622740791025e-06}
{"train_loss": 0.03897862508893013, "global_step": 235341, "epoch": 2644, "lr": 4.318326585521426e-06}
{"train_loss": 0.034587230533361435, "global_step": 235342, "epoch": 2644, "lr": 4.318090903105088e-06}
{"train_loss": 0.037091951817274094, "global_step": 235343, "epoch": 2644, "lr": 4.317855226830131e-06}
{"train_loss": 0.02251940220594406, "global_step": 235344, "epoch": 2644, "lr": 4.317619556696584e-06}
{"train_loss": 0.04015533998608589, "global_step": 235345, "epoch": 2644, "lr": 4.317383892704469e-06}
{"train_loss": 0.03171849995851517, "global_step": 235346, "epoch": 2644, "lr": 4.3171482348538185e-06}
{"train_loss": 0.05874437838792801, "global_step": 235347, "epoch": 2644, "lr": 4.316912583144683e-06}
{"train_loss": 0.05745472386479378, "global_step": 235348, "epoch": 2644, "lr": 4.3166769375770744e-06}
{"train_loss": 0.04106341674923897, "global_step": 235349, "epoch": 2644, "lr": 4.316441298151042e-06}
{"train_loss": 0.048408426344394684, "global_step": 235350, "epoch": 2644, "lr": 4.316205664866596e-06}
{"train_loss": 0.034008514136075974, "global_step": 235351, "epoch": 2644, "lr": 4.3159700377237874e-06}
{"train_loss": 0.04645950719714165, "global_step": 235352, "epoch": 2644, "lr": 4.315734416722628e-06}
{"train_loss": 0.04193820804357529, "global_step": 235353, "epoch": 2644, "lr": 4.315498801863177e-06}
{"train_loss": 0.031333617866039276, "global_step": 235354, "epoch": 2644, "lr": 4.315263193145435e-06}
{"train_loss": 0.028383970260620117, "global_step": 235355, "epoch": 2644, "lr": 4.315027590569465e-06}
{"train_loss": 0.02669423632323742, "global_step": 235356, "epoch": 2644, "lr": 4.31479199413527e-06}
{"train_loss": 0.04186833277344704, "global_step": 235357, "epoch": 2644, "lr": 4.314556403842901e-06}
{"train_loss": 0.05383950471878052, "global_step": 235358, "epoch": 2644, "lr": 4.314320819692375e-06}
{"train_loss": 0.050104446709156036, "global_step": 235359, "epoch": 2644, "lr": 4.314085241683746e-06}
{"train_loss": 0.04107804596424103, "global_step": 235360, "epoch": 2644, "lr": 4.313849669817016e-06}
{"train_loss": 0.0387837253510952, "global_step": 235361, "epoch": 2644, "lr": 4.313614104092245e-06}
{"train_loss": 0.0675434097647667, "global_step": 235362, "epoch": 2644, "lr": 4.313378544509439e-06}
{"train_loss": 0.014606567099690437, "global_step": 235363, "epoch": 2644, "lr": 4.313142991068647e-06}
{"train_loss": 0.047161124646663666, "global_step": 235364, "epoch": 2644, "lr": 4.312907443769904e-06}
{"train_loss": 0.04040239751338959, "global_step": 235365, "epoch": 2644, "lr": 4.312671902613225e-06}
{"train_loss": 0.0429222397506237, "global_step": 235366, "epoch": 2644, "lr": 4.312436367598655e-06}
{"train_loss": 0.05714816600084305, "global_step": 235367, "epoch": 2644, "lr": 4.312200838726216e-06}
{"train_loss": 0.04335053265094757, "global_step": 235368, "epoch": 2644, "lr": 4.311965315995952e-06}
{"train_loss": 0.08918707072734833, "global_step": 235369, "epoch": 2644, "lr": 4.311729799407882e-06}
{"train_loss": 0.029763681814074516, "global_step": 235370, "epoch": 2644, "lr": 4.311494288962048e-06}
{"train_loss": 0.0384148471057415, "global_step": 235371, "epoch": 2644, "lr": 4.3112587846584674e-06}
{"train_loss": 0.03800397738814354, "global_step": 235372, "epoch": 2644, "lr": 4.31102328649719e-06}
{"train_loss": 0.016054775565862656, "global_step": 235373, "epoch": 2644, "lr": 4.310787794478233e-06}
{"train_loss": 0.06325650215148926, "global_step": 235374, "epoch": 2644, "lr": 4.31055230860164e-06}
{"train_loss": 0.029390841722488403, "global_step": 235375, "epoch": 2644, "lr": 4.3103168288674275e-06}
{"train_loss": 0.022821497172117233, "global_step": 235376, "epoch": 2644, "lr": 4.310081355275647e-06}
{"train_loss": 0.05426916852593422, "global_step": 235377, "epoch": 2644, "lr": 4.3098458878263195e-06}
{"train_loss": 0.05111165717244148, "global_step": 235378, "epoch": 2644, "lr": 4.309610426519467e-06}
{"train_loss": 0.08832034468650818, "global_step": 235379, "epoch": 2644, "lr": 4.30937497135514e-06}
{"train_loss": 0.061571504920721054, "global_step": 235380, "epoch": 2644, "lr": 4.30913952233335e-06}
{"train_loss": 0.07897713780403137, "global_step": 235381, "epoch": 2644, "lr": 4.308904079454146e-06}
{"train_loss": 0.03949161246418953, "global_step": 235382, "epoch": 2644, "lr": 4.308668642717551e-06}
{"train_loss": 0.0687168762087822, "global_step": 235383, "epoch": 2644, "lr": 4.308433212123602e-06}
{"train_loss": 0.07446662336587906, "global_step": 235384, "epoch": 2644, "lr": 4.308197787672319e-06}
{"train_loss": 0.10409358143806458, "global_step": 235385, "epoch": 2644, "lr": 4.307962369363749e-06}
{"train_loss": 0.05732235684990883, "global_step": 235386, "epoch": 2644, "lr": 4.307726957197911e-06}
{"train_loss": 0.08094650506973267, "global_step": 235387, "epoch": 2644, "lr": 4.307491551174847e-06}
{"train_loss": 0.07965087890625, "global_step": 235388, "epoch": 2644, "lr": 4.307256151294581e-06}
{"train_loss": 0.022502964362502098, "global_step": 235389, "epoch": 2644, "lr": 4.3070207575571506e-06}
{"train_loss": 0.03160727024078369, "global_step": 235390, "epoch": 2644, "lr": 4.306785369962579e-06}
{"train_loss": 0.037236135452985764, "global_step": 235391, "epoch": 2644, "lr": 4.306549988510911e-06}
{"train_loss": 0.043951816856861115, "global_step": 235392, "epoch": 2644, "lr": 4.3063146132021615e-06}
{"train_loss": 0.06820660829544067, "global_step": 235393, "epoch": 2644, "lr": 4.306079244036382e-06}
{"train_loss": 0.03270230069756508, "global_step": 235394, "epoch": 2644, "lr": 4.305843881013583e-06}
{"train_loss": 0.08296967297792435, "global_step": 235395, "epoch": 2644, "lr": 4.305608524133814e-06}
{"train_loss": 0.06095286086201668, "global_step": 235396, "epoch": 2644, "lr": 4.305373173397093e-06}
{"train_loss": 0.032498691231012344, "global_step": 235397, "epoch": 2644, "lr": 4.3051378288034625e-06}
{"train_loss": 0.028873510658740997, "global_step": 235398, "epoch": 2644, "lr": 4.304902490352941e-06}
{"train_loss": 0.03793727234005928, "global_step": 235399, "epoch": 2644, "lr": 4.304667158045572e-06}
{"train_loss": 0.07711688429117203, "global_step": 235400, "epoch": 2644, "lr": 4.304431831881389e-06}
{"train_loss": 0.03236762434244156, "global_step": 235401, "epoch": 2644, "lr": 4.304196511860409e-06}
{"train_loss": 0.07310070097446442, "global_step": 235402, "epoch": 2644, "lr": 4.3039611979826865e-06}
{"train_loss": 0.05211010202765465, "global_step": 235403, "epoch": 2644, "lr": 4.303725890248228e-06}
{"train_loss": 0.04736421350389719, "global_step": 235404, "epoch": 2644, "lr": 4.303490588657083e-06, "val_loss": 9.37842082977295}
{"train_loss": 0.033693328499794006, "global_step": 235405, "epoch": 2645, "lr": 4.303255293209268e-06}
{"train_loss": 0.06650079786777496, "global_step": 235406, "epoch": 2645, "lr": 4.3030200039048344e-06}
{"train_loss": 0.05190729349851608, "global_step": 235407, "epoch": 2645, "lr": 4.302784720743791e-06}
{"train_loss": 0.032544322311878204, "global_step": 235408, "epoch": 2645, "lr": 4.302549443726195e-06}
{"train_loss": 0.04966903105378151, "global_step": 235409, "epoch": 2645, "lr": 4.302314172852051e-06}
{"train_loss": 0.04135416820645332, "global_step": 235410, "epoch": 2645, "lr": 4.302078908121415e-06}
{"train_loss": 0.05610818415880203, "global_step": 235411, "epoch": 2645, "lr": 4.3018436495342985e-06}
{"train_loss": 0.03948276862502098, "global_step": 235412, "epoch": 2645, "lr": 4.301608397090751e-06}
{"train_loss": 0.028167366981506348, "global_step": 235413, "epoch": 2645, "lr": 4.3013731507907885e-06}
{"train_loss": 0.03973108530044556, "global_step": 235414, "epoch": 2645, "lr": 4.301137910634456e-06}
{"train_loss": 0.043619390577077866, "global_step": 235415, "epoch": 2645, "lr": 4.300902676621776e-06}
{"train_loss": 0.030417265370488167, "global_step": 235416, "epoch": 2645, "lr": 4.300667448752777e-06}
{"train_loss": 0.04499315097928047, "global_step": 235417, "epoch": 2645, "lr": 4.300432227027506e-06}
{"train_loss": 0.05802563577890396, "global_step": 235418, "epoch": 2645, "lr": 4.300197011445972e-06}
{"train_loss": 0.04201406240463257, "global_step": 235419, "epoch": 2645, "lr": 4.2999618020082286e-06}
{"train_loss": 0.055250104516744614, "global_step": 235420, "epoch": 2645, "lr": 4.299726598714293e-06}
{"train_loss": 0.03583614528179169, "global_step": 235421, "epoch": 2645, "lr": 4.299491401564209e-06}
{"train_loss": 0.06472361832857132, "global_step": 235422, "epoch": 2645, "lr": 4.299256210557995e-06}
{"train_loss": 0.045531950891017914, "global_step": 235423, "epoch": 2645, "lr": 4.299021025695693e-06}
{"train_loss": 0.015271276235580444, "global_step": 235424, "epoch": 2645, "lr": 4.298785846977327e-06}
{"train_loss": 0.05152599886059761, "global_step": 235425, "epoch": 2645, "lr": 4.298550674402941e-06}
{"train_loss": 0.043836429715156555, "global_step": 235426, "epoch": 2645, "lr": 4.298315507972544e-06}
{"train_loss": 0.04774927347898483, "global_step": 235427, "epoch": 2645, "lr": 4.298080347686195e-06}
{"train_loss": 0.026613306254148483, "global_step": 235428, "epoch": 2645, "lr": 4.297845193543898e-06}
{"train_loss": 0.022501129657030106, "global_step": 235429, "epoch": 2645, "lr": 4.297610045545714e-06}
{"train_loss": 0.027085253968834877, "global_step": 235430, "epoch": 2645, "lr": 4.297374903691648e-06}
{"train_loss": 0.05177106708288193, "global_step": 235431, "epoch": 2645, "lr": 4.297139767981751e-06}
{"train_loss": 0.015922611579298973, "global_step": 235432, "epoch": 2645, "lr": 4.2969046384160384e-06}
{"train_loss": 0.035816989839076996, "global_step": 235433, "epoch": 2645, "lr": 4.296669514994556e-06}
{"train_loss": 0.034700218588113785, "global_step": 235434, "epoch": 2645, "lr": 4.296434397717319e-06}
{"train_loss": 0.09393627196550369, "global_step": 235435, "epoch": 2645, "lr": 4.296199286584374e-06}
{"train_loss": 0.043596163392066956, "global_step": 235436, "epoch": 2645, "lr": 4.295964181595758e-06}
{"train_loss": 0.07282064110040665, "global_step": 235437, "epoch": 2645, "lr": 4.295729082751482e-06}
{"train_loss": 0.09255436807870865, "global_step": 235438, "epoch": 2645, "lr": 4.295493990051596e-06}
{"train_loss": 0.0170401893556118, "global_step": 235439, "epoch": 2645, "lr": 4.295258903496113e-06}
{"train_loss": 0.08280725032091141, "global_step": 235440, "epoch": 2645, "lr": 4.295023823085087e-06}
{"train_loss": 0.02819947712123394, "global_step": 235441, "epoch": 2645, "lr": 4.294788748818529e-06}
{"train_loss": 0.05639408901333809, "global_step": 235442, "epoch": 2645, "lr": 4.294553680696489e-06}
{"train_loss": 0.03184281662106514, "global_step": 235443, "epoch": 2645, "lr": 4.294318618718979e-06}
{"train_loss": 0.04048146307468414, "global_step": 235444, "epoch": 2645, "lr": 4.294083562886053e-06}
{"train_loss": 0.07281653583049774, "global_step": 235445, "epoch": 2645, "lr": 4.293848513197718e-06}
{"train_loss": 0.03309892863035202, "global_step": 235446, "epoch": 2645, "lr": 4.293613469654023e-06}
{"train_loss": 0.029193036258220673, "global_step": 235447, "epoch": 2645, "lr": 4.293378432254991e-06}
{"train_loss": 0.0388711653649807, "global_step": 235448, "epoch": 2645, "lr": 4.293143401000666e-06}
{"train_loss": 0.03232072293758392, "global_step": 235449, "epoch": 2645, "lr": 4.292908375891064e-06}
{"train_loss": 0.03948323056101799, "global_step": 235450, "epoch": 2645, "lr": 4.2926733569262304e-06}
{"train_loss": 0.08149430900812149, "global_step": 235451, "epoch": 2645, "lr": 4.2924383441061755e-06}
{"train_loss": 0.03379169479012489, "global_step": 235452, "epoch": 2645, "lr": 4.292203337430961e-06}
{"train_loss": 0.046706415712833405, "global_step": 235453, "epoch": 2645, "lr": 4.291968336900593e-06}
{"train_loss": 0.038994379341602325, "global_step": 235454, "epoch": 2645, "lr": 4.29173334251512e-06}
{"train_loss": 0.027359263971447945, "global_step": 235455, "epoch": 2645, "lr": 4.291498354274565e-06}
{"train_loss": 0.06699065864086151, "global_step": 235456, "epoch": 2645, "lr": 4.29126337217895e-06}
{"train_loss": 0.044522229582071304, "global_step": 235457, "epoch": 2645, "lr": 4.291028396228331e-06}
{"train_loss": 0.021958358585834503, "global_step": 235458, "epoch": 2645, "lr": 4.290793426422718e-06}
{"train_loss": 0.04705023765563965, "global_step": 235459, "epoch": 2645, "lr": 4.290558462762156e-06}
{"train_loss": 0.06604155153036118, "global_step": 235460, "epoch": 2645, "lr": 4.290323505246663e-06}
{"train_loss": 0.037412628531455994, "global_step": 235461, "epoch": 2645, "lr": 4.290088553876287e-06}
{"train_loss": 0.05191396176815033, "global_step": 235462, "epoch": 2645, "lr": 4.289853608651045e-06}
{"train_loss": 0.010609441436827183, "global_step": 235463, "epoch": 2645, "lr": 4.289618669570983e-06}
{"train_loss": 0.0430256724357605, "global_step": 235464, "epoch": 2645, "lr": 4.289383736636115e-06}
{"train_loss": 0.030819665640592575, "global_step": 235465, "epoch": 2645, "lr": 4.289148809846494e-06}
{"train_loss": 0.041092649102211, "global_step": 235466, "epoch": 2645, "lr": 4.288913889202129e-06}
{"train_loss": 0.026980198919773102, "global_step": 235467, "epoch": 2645, "lr": 4.28867897470307e-06}
{"train_loss": 0.043900709599256516, "global_step": 235468, "epoch": 2645, "lr": 4.288444066349334e-06}
{"train_loss": 0.0555921271443367, "global_step": 235469, "epoch": 2645, "lr": 4.288209164140971e-06}
{"train_loss": 0.02244754508137703, "global_step": 235470, "epoch": 2645, "lr": 4.287974268077988e-06}
{"train_loss": 0.08124051988124847, "global_step": 235471, "epoch": 2645, "lr": 4.287739378160438e-06}
{"train_loss": 0.05220795422792435, "global_step": 235472, "epoch": 2645, "lr": 4.287504494388339e-06}
{"train_loss": 0.07473668456077576, "global_step": 235473, "epoch": 2645, "lr": 4.287269616761724e-06}
{"train_loss": 0.024915996938943863, "global_step": 235474, "epoch": 2645, "lr": 4.287034745280638e-06}
{"train_loss": 0.045500364154577255, "global_step": 235475, "epoch": 2645, "lr": 4.286799879945097e-06}
{"train_loss": 0.018556557595729828, "global_step": 235476, "epoch": 2645, "lr": 4.2865650207551455e-06}
{"train_loss": 0.03464038297533989, "global_step": 235477, "epoch": 2645, "lr": 4.2863301677108e-06}
{"train_loss": 0.0243599321693182, "global_step": 235478, "epoch": 2645, "lr": 4.286095320812111e-06}
{"train_loss": 0.01734774373471737, "global_step": 235479, "epoch": 2645, "lr": 4.285860480059084e-06}
{"train_loss": 0.08447965234518051, "global_step": 235480, "epoch": 2645, "lr": 4.285625645451779e-06}
{"train_loss": 0.051619578152894974, "global_step": 235481, "epoch": 2645, "lr": 4.285390816990209e-06}
{"train_loss": 0.05057903379201889, "global_step": 235482, "epoch": 2645, "lr": 4.285155994674417e-06}
{"train_loss": 0.07290460914373398, "global_step": 235483, "epoch": 2645, "lr": 4.28492117850442e-06}
{"train_loss": 0.046739138662815094, "global_step": 235484, "epoch": 2645, "lr": 4.284686368480262e-06}
{"train_loss": 0.023247718811035156, "global_step": 235485, "epoch": 2645, "lr": 4.2844515646019654e-06}
{"train_loss": 0.021273726597428322, "global_step": 235486, "epoch": 2645, "lr": 4.284216766869581e-06}
{"train_loss": 0.03919797018170357, "global_step": 235487, "epoch": 2645, "lr": 4.2839819752831076e-06}
{"train_loss": 0.0261443629860878, "global_step": 235488, "epoch": 2645, "lr": 4.283747189842613e-06}
{"train_loss": 0.04530542716383934, "global_step": 235489, "epoch": 2645, "lr": 4.283512410548102e-06}
{"train_loss": 0.04004279524087906, "global_step": 235490, "epoch": 2645, "lr": 4.2832776373996185e-06}
{"train_loss": 0.03204904496669769, "global_step": 235491, "epoch": 2645, "lr": 4.283042870397186e-06}
{"train_loss": 0.03148597478866577, "global_step": 235492, "epoch": 2645, "lr": 4.282808109540848e-06}
{"train_loss": 0.04390254383383507, "global_step": 235493, "epoch": 2645, "lr": 4.282573354830633e-06, "val_loss": 9.475335121154785, "train_action_mse_error": 11.436264991760254}
{"train_loss": 0.03453734144568443, "global_step": 235494, "epoch": 2646, "lr": 4.282338606266556e-06}
{"train_loss": 0.04952513799071312, "global_step": 235495, "epoch": 2646, "lr": 4.282103863848675e-06}
{"train_loss": 0.03792167454957962, "global_step": 235496, "epoch": 2646, "lr": 4.281869127576993e-06}
{"train_loss": 0.05572899803519249, "global_step": 235497, "epoch": 2646, "lr": 4.281634397451573e-06}
{"train_loss": 0.027618562802672386, "global_step": 235498, "epoch": 2646, "lr": 4.281399673472414e-06}
{"train_loss": 0.02558496594429016, "global_step": 235499, "epoch": 2646, "lr": 4.2811649556395774e-06}
{"train_loss": 0.04405700042843819, "global_step": 235500, "epoch": 2646, "lr": 4.280930243953068e-06}
{"train_loss": 0.072540283203125, "global_step": 235501, "epoch": 2646, "lr": 4.280695538412943e-06}
{"train_loss": 0.05718515068292618, "global_step": 235502, "epoch": 2646, "lr": 4.280460839019207e-06}
{"train_loss": 0.05434776842594147, "global_step": 235503, "epoch": 2646, "lr": 4.280226145771921e-06}
{"train_loss": 0.02231973595917225, "global_step": 235504, "epoch": 2646, "lr": 4.27999145867109e-06}
{"train_loss": 0.03970959410071373, "global_step": 235505, "epoch": 2646, "lr": 4.279756777716765e-06}
{"train_loss": 0.05442005395889282, "global_step": 235506, "epoch": 2646, "lr": 4.279522102908961e-06}
{"train_loss": 0.07495623081922531, "global_step": 235507, "epoch": 2646, "lr": 4.279287434247726e-06}
{"train_loss": 0.04868984594941139, "global_step": 235508, "epoch": 2646, "lr": 4.2790527717330716e-06}
{"train_loss": 0.029331272467970848, "global_step": 235509, "epoch": 2646, "lr": 4.278818115365046e-06}
{"train_loss": 0.05441965535283089, "global_step": 235510, "epoch": 2646, "lr": 4.278583465143688e-06}
{"train_loss": 0.03332550823688507, "global_step": 235511, "epoch": 2646, "lr": 4.278348821069e-06}
{"train_loss": 0.03844811022281647, "global_step": 235512, "epoch": 2646, "lr": 4.278114183141046e-06}
{"train_loss": 0.03862590342760086, "global_step": 235513, "epoch": 2646, "lr": 4.277879551359832e-06}
{"train_loss": 0.028698813170194626, "global_step": 235514, "epoch": 2646, "lr": 4.277644925725405e-06}
{"train_loss": 0.0479329489171505, "global_step": 235515, "epoch": 2646, "lr": 4.277410306237789e-06}
{"train_loss": 0.039108216762542725, "global_step": 235516, "epoch": 2646, "lr": 4.277175692897023e-06}
{"train_loss": 0.02974446304142475, "global_step": 235517, "epoch": 2646, "lr": 4.276941085703123e-06}
{"train_loss": 0.03238821029663086, "global_step": 235518, "epoch": 2646, "lr": 4.27670648465614e-06}
{"train_loss": 0.01951112225651741, "global_step": 235519, "epoch": 2646, "lr": 4.276471889756089e-06}
{"train_loss": 0.03560261055827141, "global_step": 235520, "epoch": 2646, "lr": 4.276237301003022e-06}
{"train_loss": 0.03778015449643135, "global_step": 235521, "epoch": 2646, "lr": 4.2760027183969435e-06}
{"train_loss": 0.054654836654663086, "global_step": 235522, "epoch": 2646, "lr": 4.275768141937908e-06}
{"train_loss": 0.06969384849071503, "global_step": 235523, "epoch": 2646, "lr": 4.2755335716259335e-06}
{"train_loss": 0.04916058108210564, "global_step": 235524, "epoch": 2646, "lr": 4.2752990074610645e-06}
{"train_loss": 0.045425619930028915, "global_step": 235525, "epoch": 2646, "lr": 4.275064449443311e-06}
{"train_loss": 0.0811411663889885, "global_step": 235526, "epoch": 2646, "lr": 4.2748298975727294e-06}
{"train_loss": 0.05258220061659813, "global_step": 235527, "epoch": 2646, "lr": 4.274595351849331e-06}
{"train_loss": 0.03532436490058899, "global_step": 235528, "epoch": 2646, "lr": 4.2743608122731646e-06}
{"train_loss": 0.027840660884976387, "global_step": 235529, "epoch": 2646, "lr": 4.274126278844243e-06}
{"train_loss": 0.02514236606657505, "global_step": 235530, "epoch": 2646, "lr": 4.273891751562615e-06}
{"train_loss": 0.01698244921863079, "global_step": 235531, "epoch": 2646, "lr": 4.273657230428308e-06}
{"train_loss": 0.06509667634963989, "global_step": 235532, "epoch": 2646, "lr": 4.273422715441339e-06}
{"train_loss": 0.04682045429944992, "global_step": 235533, "epoch": 2646, "lr": 4.273188206601763e-06}
{"train_loss": 0.03508978337049484, "global_step": 235534, "epoch": 2646, "lr": 4.272953703909588e-06}
{"train_loss": 0.046160563826560974, "global_step": 235535, "epoch": 2646, "lr": 4.272719207364867e-06}
{"train_loss": 0.034756794571876526, "global_step": 235536, "epoch": 2646, "lr": 4.272484716967618e-06}
{"train_loss": 0.05612926185131073, "global_step": 235537, "epoch": 2646, "lr": 4.272250232717878e-06}
{"train_loss": 0.047966450452804565, "global_step": 235538, "epoch": 2646, "lr": 4.272015754615666e-06}
{"train_loss": 0.039953164756298065, "global_step": 235539, "epoch": 2646, "lr": 4.2717812826610414e-06}
{"train_loss": 0.03466979041695595, "global_step": 235540, "epoch": 2646, "lr": 4.271546816854005e-06}
{"train_loss": 0.02502773329615593, "global_step": 235541, "epoch": 2646, "lr": 4.271312357194607e-06}
{"train_loss": 0.05255987122654915, "global_step": 235542, "epoch": 2646, "lr": 4.271077903682868e-06}
{"train_loss": 0.03544256091117859, "global_step": 235543, "epoch": 2646, "lr": 4.270843456318835e-06}
{"train_loss": 0.022163262590765953, "global_step": 235544, "epoch": 2646, "lr": 4.270609015102523e-06}
{"train_loss": 0.05307967960834503, "global_step": 235545, "epoch": 2646, "lr": 4.270374580033965e-06}
{"train_loss": 0.054601460695266724, "global_step": 235546, "epoch": 2646, "lr": 4.270140151113211e-06}
{"train_loss": 0.09273982048034668, "global_step": 235547, "epoch": 2646, "lr": 4.2699057283402675e-06}
{"train_loss": 0.07166698575019836, "global_step": 235548, "epoch": 2646, "lr": 4.2696713117151895e-06}
{"train_loss": 0.03545806556940079, "global_step": 235549, "epoch": 2646, "lr": 4.269436901237983e-06}
{"train_loss": 0.07363370060920715, "global_step": 235550, "epoch": 2646, "lr": 4.269202496908708e-06}
{"train_loss": 0.07299468666315079, "global_step": 235551, "epoch": 2646, "lr": 4.268968098727366e-06}
{"train_loss": 0.05164887756109238, "global_step": 235552, "epoch": 2646, "lr": 4.268733706694017e-06}
{"train_loss": 0.03059043549001217, "global_step": 235553, "epoch": 2646, "lr": 4.2684993208086725e-06}
{"train_loss": 0.047152239829301834, "global_step": 235554, "epoch": 2646, "lr": 4.268264941071376e-06}
{"train_loss": 0.020993836224079132, "global_step": 235555, "epoch": 2646, "lr": 4.268030567482146e-06}
{"train_loss": 0.05475914850831032, "global_step": 235556, "epoch": 2646, "lr": 4.267796200041035e-06}
{"train_loss": 0.035014547407627106, "global_step": 235557, "epoch": 2646, "lr": 4.267561838748046e-06}
{"train_loss": 0.03710075840353966, "global_step": 235558, "epoch": 2646, "lr": 4.2673274836032396e-06}
{"train_loss": 0.06458812952041626, "global_step": 235559, "epoch": 2646, "lr": 4.2670931346066205e-06}
{"train_loss": 0.05203314125537872, "global_step": 235560, "epoch": 2646, "lr": 4.2668587917582434e-06}
{"train_loss": 0.034370794892311096, "global_step": 235561, "epoch": 2646, "lr": 4.266624455058121e-06}
{"train_loss": 0.10786323994398117, "global_step": 235562, "epoch": 2646, "lr": 4.2663901245062975e-06}
{"train_loss": 0.04277841001749039, "global_step": 235563, "epoch": 2646, "lr": 4.266155800102806e-06}
{"train_loss": 0.021457023918628693, "global_step": 235564, "epoch": 2646, "lr": 4.265921481847662e-06}
{"train_loss": 0.05366404727101326, "global_step": 235565, "epoch": 2646, "lr": 4.2656871697409175e-06}
{"train_loss": 0.058301232755184174, "global_step": 235566, "epoch": 2646, "lr": 4.265452863782593e-06}
{"train_loss": 0.049679234623909, "global_step": 235567, "epoch": 2646, "lr": 4.265218563972717e-06}
{"train_loss": 0.02656863071024418, "global_step": 235568, "epoch": 2646, "lr": 4.264984270311328e-06}
{"train_loss": 0.03478753939270973, "global_step": 235569, "epoch": 2646, "lr": 4.264749982798449e-06}
{"train_loss": 0.06277656555175781, "global_step": 235570, "epoch": 2646, "lr": 4.264515701434124e-06}
{"train_loss": 0.05897494778037071, "global_step": 235571, "epoch": 2646, "lr": 4.2642814262183805e-06}
{"train_loss": 0.07372584939002991, "global_step": 235572, "epoch": 2646, "lr": 4.264047157151235e-06}
{"train_loss": 0.07175832241773605, "global_step": 235573, "epoch": 2646, "lr": 4.2638128942327385e-06}
{"train_loss": 0.11860281229019165, "global_step": 235574, "epoch": 2646, "lr": 4.263578637462906e-06}
{"train_loss": 0.03916460648179054, "global_step": 235575, "epoch": 2646, "lr": 4.263344386841794e-06}
{"train_loss": 0.02466450445353985, "global_step": 235576, "epoch": 2646, "lr": 4.263110142369403e-06}
{"train_loss": 0.03362990543246269, "global_step": 235577, "epoch": 2646, "lr": 4.262875904045787e-06}
{"train_loss": 0.0469178780913353, "global_step": 235578, "epoch": 2646, "lr": 4.262641671870965e-06}
{"train_loss": 0.03839309141039848, "global_step": 235579, "epoch": 2646, "lr": 4.26240744584498e-06}
{"train_loss": 0.05045221373438835, "global_step": 235580, "epoch": 2646, "lr": 4.262173225967847e-06}
{"train_loss": 0.06331775337457657, "global_step": 235581, "epoch": 2646, "lr": 4.261939012239607e-06}
{"train_loss": 0.04662649253948351, "global_step": 235582, "epoch": 2646, "lr": 4.261704804660305e-06, "val_loss": 9.346506118774414}
{"train_loss": 0.06601639837026596, "global_step": 235583, "epoch": 2647, "lr": 4.261470603229945e-06}
{"train_loss": 0.07781793177127838, "global_step": 235584, "epoch": 2647, "lr": 4.2612364079485875e-06}
{"train_loss": 0.06404227018356323, "global_step": 235585, "epoch": 2647, "lr": 4.2610022188162346e-06}
{"train_loss": 0.044139254838228226, "global_step": 235586, "epoch": 2647, "lr": 4.260768035832946e-06}
{"train_loss": 0.09670407325029373, "global_step": 235587, "epoch": 2647, "lr": 4.260533858998728e-06}
{"train_loss": 0.06612498313188553, "global_step": 235588, "epoch": 2647, "lr": 4.2602996883136345e-06}
{"train_loss": 0.04376409575343132, "global_step": 235589, "epoch": 2647, "lr": 4.260065523777673e-06}
{"train_loss": 0.04281405359506607, "global_step": 235590, "epoch": 2647, "lr": 4.259831365390904e-06}
{"train_loss": 0.08229364454746246, "global_step": 235591, "epoch": 2647, "lr": 4.259597213153327e-06}
{"train_loss": 0.09521226584911346, "global_step": 235592, "epoch": 2647, "lr": 4.259363067065003e-06}
{"train_loss": 0.11941342800855637, "global_step": 235593, "epoch": 2647, "lr": 4.259128927125944e-06}
{"train_loss": 0.03571203723549843, "global_step": 235594, "epoch": 2647, "lr": 4.258894793336193e-06}
{"train_loss": 0.041507523506879807, "global_step": 235595, "epoch": 2647, "lr": 4.2586606656957625e-06}
{"train_loss": 0.025310948491096497, "global_step": 235596, "epoch": 2647, "lr": 4.258426544204713e-06}
{"train_loss": 0.03918393328785896, "global_step": 235597, "epoch": 2647, "lr": 4.2581924288630506e-06}
{"train_loss": 0.038231249898672104, "global_step": 235598, "epoch": 2647, "lr": 4.257958319670824e-06}
{"train_loss": 0.04900407791137695, "global_step": 235599, "epoch": 2647, "lr": 4.257724216628051e-06}
{"train_loss": 0.03508928418159485, "global_step": 235600, "epoch": 2647, "lr": 4.2574901197347635e-06}
{"train_loss": 0.01733667403459549, "global_step": 235601, "epoch": 2647, "lr": 4.257256028991013e-06}
{"train_loss": 0.0786098837852478, "global_step": 235602, "epoch": 2647, "lr": 4.25702194439681e-06}
{"train_loss": 0.04402090236544609, "global_step": 235603, "epoch": 2647, "lr": 4.2567878659522e-06}
{"train_loss": 0.01918349228799343, "global_step": 235604, "epoch": 2647, "lr": 4.256553793657203e-06}
{"train_loss": 0.07225020229816437, "global_step": 235605, "epoch": 2647, "lr": 4.256319727511848e-06}
{"train_loss": 0.046250615268945694, "global_step": 235606, "epoch": 2647, "lr": 4.256085667516186e-06}
{"train_loss": 0.03418637439608574, "global_step": 235607, "epoch": 2647, "lr": 4.255851613670226e-06}
{"train_loss": 0.08596350252628326, "global_step": 235608, "epoch": 2647, "lr": 4.255617565974013e-06}
{"train_loss": 0.05643082037568092, "global_step": 235609, "epoch": 2647, "lr": 4.2553835244275804e-06}
{"train_loss": 0.07608576864004135, "global_step": 235610, "epoch": 2647, "lr": 4.255149489030941e-06}
{"train_loss": 0.016152190044522285, "global_step": 235611, "epoch": 2647, "lr": 4.2549154597841475e-06}
{"train_loss": 0.040562696754932404, "global_step": 235612, "epoch": 2647, "lr": 4.254681436687219e-06}
{"train_loss": 0.0408504456281662, "global_step": 235613, "epoch": 2647, "lr": 4.254447419740199e-06}
{"train_loss": 0.06717616319656372, "global_step": 235614, "epoch": 2647, "lr": 4.254213408943098e-06}
{"train_loss": 0.0643228068947792, "global_step": 235615, "epoch": 2647, "lr": 4.253979404295972e-06}
{"train_loss": 0.0176096074283123, "global_step": 235616, "epoch": 2647, "lr": 4.2537454057988316e-06}
{"train_loss": 0.05131998285651207, "global_step": 235617, "epoch": 2647, "lr": 4.2535114134517165e-06}
{"train_loss": 0.07239626348018646, "global_step": 235618, "epoch": 2647, "lr": 4.253277427254671e-06}
{"train_loss": 0.04742593690752983, "global_step": 235619, "epoch": 2647, "lr": 4.2530434472077e-06}
{"train_loss": 0.01951375976204872, "global_step": 235620, "epoch": 2647, "lr": 4.252809473310865e-06}
{"train_loss": 0.022410772740840912, "global_step": 235621, "epoch": 2647, "lr": 4.2525755055641715e-06}
{"train_loss": 0.05531406030058861, "global_step": 235622, "epoch": 2647, "lr": 4.2523415439676695e-06}
{"train_loss": 0.03171137720346451, "global_step": 235623, "epoch": 2647, "lr": 4.252107588521375e-06}
{"train_loss": 0.02173473872244358, "global_step": 235624, "epoch": 2647, "lr": 4.251873639225334e-06}
{"train_loss": 0.058072756975889206, "global_step": 235625, "epoch": 2647, "lr": 4.251639696079563e-06}
{"train_loss": 0.0463881716132164, "global_step": 235626, "epoch": 2647, "lr": 4.251405759084109e-06}
{"train_loss": 0.04664463922381401, "global_step": 235627, "epoch": 2647, "lr": 4.251171828238992e-06}
{"train_loss": 0.03289375454187393, "global_step": 235628, "epoch": 2647, "lr": 4.25093790354425e-06}
{"train_loss": 0.05702695623040199, "global_step": 235629, "epoch": 2647, "lr": 4.2507039849999095e-06}
{"train_loss": 0.02934572659432888, "global_step": 235630, "epoch": 2647, "lr": 4.2504700726060056e-06}
{"train_loss": 0.06426765769720078, "global_step": 235631, "epoch": 2647, "lr": 4.2502361663625645e-06}
{"train_loss": 0.04148406907916069, "global_step": 235632, "epoch": 2647, "lr": 4.250002266269631e-06}
{"train_loss": 0.03341718018054962, "global_step": 235633, "epoch": 2647, "lr": 4.249768372327212e-06}
{"train_loss": 0.043497856706380844, "global_step": 235634, "epoch": 2647, "lr": 4.249534484535372e-06}
{"train_loss": 0.038540393114089966, "global_step": 235635, "epoch": 2647, "lr": 4.249300602894107e-06}
{"train_loss": 0.03270898386836052, "global_step": 235636, "epoch": 2647, "lr": 4.249066727403472e-06}
{"train_loss": 0.012073776684701443, "global_step": 235637, "epoch": 2647, "lr": 4.248832858063501e-06}
{"train_loss": 0.03155441954731941, "global_step": 235638, "epoch": 2647, "lr": 4.248598994874203e-06}
{"train_loss": 0.14393675327301025, "global_step": 235639, "epoch": 2647, "lr": 4.248365137835636e-06}
{"train_loss": 0.05067870765924454, "global_step": 235640, "epoch": 2647, "lr": 4.24813128694781e-06}
{"train_loss": 0.055390506982803345, "global_step": 235641, "epoch": 2647, "lr": 4.247897442210774e-06}
{"train_loss": 0.04611915722489357, "global_step": 235642, "epoch": 2647, "lr": 4.247663603624546e-06}
{"train_loss": 0.041643254458904266, "global_step": 235643, "epoch": 2647, "lr": 4.24742977118916e-06}
{"train_loss": 0.039822302758693695, "global_step": 235644, "epoch": 2647, "lr": 4.247195944904653e-06}
{"train_loss": 0.050796955823898315, "global_step": 235645, "epoch": 2647, "lr": 4.246962124771043e-06}
{"train_loss": 0.04105772078037262, "global_step": 235646, "epoch": 2647, "lr": 4.2467283107883835e-06}
{"train_loss": 0.08477132022380829, "global_step": 235647, "epoch": 2647, "lr": 4.246494502956694e-06}
{"train_loss": 0.044857315719127655, "global_step": 235648, "epoch": 2647, "lr": 4.246260701275995e-06}
{"train_loss": 0.07350032031536102, "global_step": 235649, "epoch": 2647, "lr": 4.246026905746342e-06}
{"train_loss": 0.06975457072257996, "global_step": 235650, "epoch": 2647, "lr": 4.245793116367741e-06}
{"train_loss": 0.03670576214790344, "global_step": 235651, "epoch": 2647, "lr": 4.245559333140242e-06}
{"train_loss": 0.023471560329198837, "global_step": 235652, "epoch": 2647, "lr": 4.245325556063862e-06}
{"train_loss": 0.026713205501437187, "global_step": 235653, "epoch": 2647, "lr": 4.2450917851386444e-06}
{"train_loss": 0.07172229140996933, "global_step": 235654, "epoch": 2647, "lr": 4.244858020364617e-06}
{"train_loss": 0.06226405128836632, "global_step": 235655, "epoch": 2647, "lr": 4.244624261741808e-06}
{"train_loss": 0.04424683004617691, "global_step": 235656, "epoch": 2647, "lr": 4.2443905092702625e-06}
{"train_loss": 0.030719410628080368, "global_step": 235657, "epoch": 2647, "lr": 4.244156762949986e-06}
{"train_loss": 0.05168242007493973, "global_step": 235658, "epoch": 2647, "lr": 4.243923022781038e-06}
{"train_loss": 0.02040906809270382, "global_step": 235659, "epoch": 2647, "lr": 4.243689288763431e-06}
{"train_loss": 0.05525176227092743, "global_step": 235660, "epoch": 2647, "lr": 4.2434555608972026e-06}
{"train_loss": 0.0430661216378212, "global_step": 235661, "epoch": 2647, "lr": 4.243221839182382e-06}
{"train_loss": 0.033401716500520706, "global_step": 235662, "epoch": 2647, "lr": 4.2429881236190075e-06}
{"train_loss": 0.02613656036555767, "global_step": 235663, "epoch": 2647, "lr": 4.242754414207101e-06}
{"train_loss": 0.03812142834067345, "global_step": 235664, "epoch": 2647, "lr": 4.242520710946701e-06}
{"train_loss": 0.05913197249174118, "global_step": 235665, "epoch": 2647, "lr": 4.242287013837831e-06}
{"train_loss": 0.04071729630231857, "global_step": 235666, "epoch": 2647, "lr": 4.24205332288054e-06}
{"train_loss": 0.047449786216020584, "global_step": 235667, "epoch": 2647, "lr": 4.241819638074835e-06}
{"train_loss": 0.058691516518592834, "global_step": 235668, "epoch": 2647, "lr": 4.2415859594207685e-06}
{"train_loss": 0.05459412559866905, "global_step": 235669, "epoch": 2647, "lr": 4.2413522869183545e-06}
{"train_loss": 0.020459525287151337, "global_step": 235670, "epoch": 2647, "lr": 4.241118620567641e-06}
{"train_loss": 0.049304296388133856, "global_step": 235671, "epoch": 2647, "lr": 4.240884960368641e-06, "val_loss": 9.38446044921875}
{"train_loss": 0.04556506499648094, "global_step": 235672, "epoch": 2648, "lr": 4.240651306321397e-06}
{"train_loss": 0.023381298407912254, "global_step": 235673, "epoch": 2648, "lr": 4.2404176584259496e-06}
{"train_loss": 0.03937245532870293, "global_step": 235674, "epoch": 2648, "lr": 4.240184016682314e-06}
{"train_loss": 0.052948419004678726, "global_step": 235675, "epoch": 2648, "lr": 4.239950381090535e-06}
{"train_loss": 0.04921311140060425, "global_step": 235676, "epoch": 2648, "lr": 4.239716751650624e-06}
{"train_loss": 0.0687691941857338, "global_step": 235677, "epoch": 2648, "lr": 4.239483128362642e-06}
{"train_loss": 0.04890673607587814, "global_step": 235678, "epoch": 2648, "lr": 4.2392495112265895e-06}
{"train_loss": 0.051105424761772156, "global_step": 235679, "epoch": 2648, "lr": 4.239015900242521e-06}
{"train_loss": 0.05832060053944588, "global_step": 235680, "epoch": 2648, "lr": 4.238782295410454e-06}
{"train_loss": 0.11390091478824615, "global_step": 235681, "epoch": 2648, "lr": 4.238548696730432e-06}
{"train_loss": 0.04958202317357063, "global_step": 235682, "epoch": 2648, "lr": 4.2383151042024784e-06}
{"train_loss": 0.06983885169029236, "global_step": 235683, "epoch": 2648, "lr": 4.238081517826614e-06}
{"train_loss": 0.0636061280965805, "global_step": 235684, "epoch": 2648, "lr": 4.237847937602896e-06}
{"train_loss": 0.04727356508374214, "global_step": 235685, "epoch": 2648, "lr": 4.237614363531328e-06}
{"train_loss": 0.05023609846830368, "global_step": 235686, "epoch": 2648, "lr": 4.237380795611967e-06}
{"train_loss": 0.027827970683574677, "global_step": 235687, "epoch": 2648, "lr": 4.237147233844835e-06}
{"train_loss": 0.028383878991007805, "global_step": 235688, "epoch": 2648, "lr": 4.236913678229948e-06}
{"train_loss": 0.0868094265460968, "global_step": 235689, "epoch": 2648, "lr": 4.23668012876735e-06}
{"train_loss": 0.05230008810758591, "global_step": 235690, "epoch": 2648, "lr": 4.236446585457082e-06}
{"train_loss": 0.06259488314390182, "global_step": 235691, "epoch": 2648, "lr": 4.2362130482991535e-06}
{"train_loss": 0.03126402944326401, "global_step": 235692, "epoch": 2648, "lr": 4.2359795172936256e-06}
{"train_loss": 0.04586925357580185, "global_step": 235693, "epoch": 2648, "lr": 4.2357459924404986e-06}
{"train_loss": 0.08483663946390152, "global_step": 235694, "epoch": 2648, "lr": 4.2355124737398225e-06}
{"train_loss": 0.019758587703108788, "global_step": 235695, "epoch": 2648, "lr": 4.235278961191624e-06}
{"train_loss": 0.04011359065771103, "global_step": 235696, "epoch": 2648, "lr": 4.235045454795938e-06}
{"train_loss": 0.09185429662466049, "global_step": 235697, "epoch": 2648, "lr": 4.234811954552781e-06}
{"train_loss": 0.041588857769966125, "global_step": 235698, "epoch": 2648, "lr": 4.234578460462213e-06}
{"train_loss": 0.03521047532558441, "global_step": 235699, "epoch": 2648, "lr": 4.234344972524229e-06}
{"train_loss": 0.06082276627421379, "global_step": 235700, "epoch": 2648, "lr": 4.2341114907388954e-06}
{"train_loss": 0.17937254905700684, "global_step": 235701, "epoch": 2648, "lr": 4.2338780151062186e-06}
{"train_loss": 0.030857140198349953, "global_step": 235702, "epoch": 2648, "lr": 4.233644545626248e-06}
{"train_loss": 0.041721757501363754, "global_step": 235703, "epoch": 2648, "lr": 4.2334110822989935e-06}
{"train_loss": 0.021903662011027336, "global_step": 235704, "epoch": 2648, "lr": 4.233177625124512e-06}
{"train_loss": 0.04201941192150116, "global_step": 235705, "epoch": 2648, "lr": 4.2329441741028096e-06}
{"train_loss": 0.015890276059508324, "global_step": 235706, "epoch": 2648, "lr": 4.232710729233941e-06}
{"train_loss": 0.04817134886980057, "global_step": 235707, "epoch": 2648, "lr": 4.232477290517917e-06}
{"train_loss": 0.05885987728834152, "global_step": 235708, "epoch": 2648, "lr": 4.2322438579547765e-06}
{"train_loss": 0.038247887045145035, "global_step": 235709, "epoch": 2648, "lr": 4.232010431544564e-06}
{"train_loss": 0.0723985880613327, "global_step": 235710, "epoch": 2648, "lr": 4.231777011287291e-06}
{"train_loss": 0.07031252235174179, "global_step": 235711, "epoch": 2648, "lr": 4.231543597183007e-06}
{"train_loss": 0.05148068442940712, "global_step": 235712, "epoch": 2648, "lr": 4.2313101892317246e-06}
{"train_loss": 0.041589830070734024, "global_step": 235713, "epoch": 2648, "lr": 4.231076787433497e-06}
{"train_loss": 0.044491011649370193, "global_step": 235714, "epoch": 2648, "lr": 4.230843391788331e-06}
{"train_loss": 0.05173832178115845, "global_step": 235715, "epoch": 2648, "lr": 4.2306100022962755e-06}
{"train_loss": 0.05226224660873413, "global_step": 235716, "epoch": 2648, "lr": 4.230376618957355e-06}
{"train_loss": 0.04338303208351135, "global_step": 235717, "epoch": 2648, "lr": 4.2301432417716055e-06}
{"train_loss": 0.08878114074468613, "global_step": 235718, "epoch": 2648, "lr": 4.229909870739052e-06}
{"train_loss": 0.05169607326388359, "global_step": 235719, "epoch": 2648, "lr": 4.229676505859737e-06}
{"train_loss": 0.07207373529672623, "global_step": 235720, "epoch": 2648, "lr": 4.229443147133683e-06}
{"train_loss": 0.021666327491402626, "global_step": 235721, "epoch": 2648, "lr": 4.229209794560912e-06}
{"train_loss": 0.06009401008486748, "global_step": 235722, "epoch": 2648, "lr": 4.2289764481414755e-06}
{"train_loss": 0.045431435108184814, "global_step": 235723, "epoch": 2648, "lr": 4.228743107875388e-06}
{"train_loss": 0.06792815029621124, "global_step": 235724, "epoch": 2648, "lr": 4.228509773762696e-06}
{"train_loss": 0.04512575641274452, "global_step": 235725, "epoch": 2648, "lr": 4.22827644580342e-06}
{"train_loss": 0.04179913178086281, "global_step": 235726, "epoch": 2648, "lr": 4.228043123997588e-06}
{"train_loss": 0.062475744634866714, "global_step": 235727, "epoch": 2648, "lr": 4.22780980834524e-06}
{"train_loss": 0.05469776317477226, "global_step": 235728, "epoch": 2648, "lr": 4.227576498846409e-06}
{"train_loss": 0.07866153120994568, "global_step": 235729, "epoch": 2648, "lr": 4.227343195501116e-06}
{"train_loss": 0.05914271995425224, "global_step": 235730, "epoch": 2648, "lr": 4.227109898309412e-06}
{"train_loss": 0.05670323595404625, "global_step": 235731, "epoch": 2648, "lr": 4.226876607271302e-06}
{"train_loss": 0.04158424586057663, "global_step": 235732, "epoch": 2648, "lr": 4.226643322386836e-06}
{"train_loss": 0.08374915271997452, "global_step": 235733, "epoch": 2648, "lr": 4.226410043656037e-06}
{"train_loss": 0.035576384514570236, "global_step": 235734, "epoch": 2648, "lr": 4.2261767710789435e-06}
{"train_loss": 0.04618099331855774, "global_step": 235735, "epoch": 2648, "lr": 4.225943504655577e-06}
{"train_loss": 0.08511640131473541, "global_step": 235736, "epoch": 2648, "lr": 4.225710244385983e-06}
{"train_loss": 0.058847445994615555, "global_step": 235737, "epoch": 2648, "lr": 4.2254769902701775e-06}
{"train_loss": 0.06453549861907959, "global_step": 235738, "epoch": 2648, "lr": 4.225243742308205e-06}
{"train_loss": 0.05471420660614967, "global_step": 235739, "epoch": 2648, "lr": 4.225010500500082e-06}
{"train_loss": 0.057426225394010544, "global_step": 235740, "epoch": 2648, "lr": 4.2247772648458525e-06}
{"train_loss": 0.03612259402871132, "global_step": 235741, "epoch": 2648, "lr": 4.22454403534554e-06}
{"train_loss": 0.0903974249958992, "global_step": 235742, "epoch": 2648, "lr": 4.224310811999188e-06}
{"train_loss": 0.0974230170249939, "global_step": 235743, "epoch": 2648, "lr": 4.224077594806813e-06}
{"train_loss": 0.044326506555080414, "global_step": 235744, "epoch": 2648, "lr": 4.223844383768449e-06}
{"train_loss": 0.05659530684351921, "global_step": 235745, "epoch": 2648, "lr": 4.22361117888414e-06}
{"train_loss": 0.0682242214679718, "global_step": 235746, "epoch": 2648, "lr": 4.223377980153903e-06}
{"train_loss": 0.07281441986560822, "global_step": 235747, "epoch": 2648, "lr": 4.223144787577782e-06}
{"train_loss": 0.0940631702542305, "global_step": 235748, "epoch": 2648, "lr": 4.222911601155794e-06}
{"train_loss": 0.032399315387010574, "global_step": 235749, "epoch": 2648, "lr": 4.222678420887982e-06}
{"train_loss": 0.03092203661799431, "global_step": 235750, "epoch": 2648, "lr": 4.222445246774371e-06}
{"train_loss": 0.06116199120879173, "global_step": 235751, "epoch": 2648, "lr": 4.2222120788149964e-06}
{"train_loss": 0.07930609583854675, "global_step": 235752, "epoch": 2648, "lr": 4.221978917009883e-06}
{"train_loss": 0.030743254348635674, "global_step": 235753, "epoch": 2648, "lr": 4.221745761359075e-06}
{"train_loss": 0.021709542721509933, "global_step": 235754, "epoch": 2648, "lr": 4.221512611862582e-06}
{"train_loss": 0.04876329004764557, "global_step": 235755, "epoch": 2648, "lr": 4.221279468520461e-06}
{"train_loss": 0.07644975930452347, "global_step": 235756, "epoch": 2648, "lr": 4.221046331332723e-06}
{"train_loss": 0.06889230757951736, "global_step": 235757, "epoch": 2648, "lr": 4.220813200299412e-06}
{"train_loss": 0.04243902489542961, "global_step": 235758, "epoch": 2648, "lr": 4.22058007542056e-06}
{"train_loss": 0.019838696345686913, "global_step": 235759, "epoch": 2648, "lr": 4.22034695669618e-06}
{"train_loss": 0.055108339787366685, "global_step": 235760, "epoch": 2648, "lr": 4.220113844126328e-06, "val_loss": 9.246903419494629}
{"train_loss": 0.06933483481407166, "global_step": 235761, "epoch": 2649, "lr": 4.219880737711013e-06}
{"train_loss": 0.05915972962975502, "global_step": 235762, "epoch": 2649, "lr": 4.219647637450291e-06}
{"train_loss": 0.030858460813760757, "global_step": 235763, "epoch": 2649, "lr": 4.219414543344164e-06}
{"train_loss": 0.04657190293073654, "global_step": 235764, "epoch": 2649, "lr": 4.21918145539269e-06}
{"train_loss": 0.04761156067252159, "global_step": 235765, "epoch": 2649, "lr": 4.218948373595882e-06}
{"train_loss": 0.06351806968450546, "global_step": 235766, "epoch": 2649, "lr": 4.218715297953784e-06}
{"train_loss": 0.02229246497154236, "global_step": 235767, "epoch": 2649, "lr": 4.2184822284664134e-06}
{"train_loss": 0.08702506870031357, "global_step": 235768, "epoch": 2649, "lr": 4.218249165133819e-06}
{"train_loss": 0.02886057272553444, "global_step": 235769, "epoch": 2649, "lr": 4.218016107956019e-06}
{"train_loss": 0.006651001051068306, "global_step": 235770, "epoch": 2649, "lr": 4.21778305693305e-06}
{"train_loss": 0.013175282627344131, "global_step": 235771, "epoch": 2649, "lr": 4.217550012064936e-06}
{"train_loss": 0.03072463534772396, "global_step": 235772, "epoch": 2649, "lr": 4.217316973351726e-06}
{"train_loss": 0.015729345381259918, "global_step": 235773, "epoch": 2649, "lr": 4.217083940793426e-06}
{"train_loss": 0.024964910000562668, "global_step": 235774, "epoch": 2649, "lr": 4.216850914390097e-06}
{"train_loss": 0.0573505200445652, "global_step": 235775, "epoch": 2649, "lr": 4.216617894141739e-06}
{"train_loss": 0.01767336204648018, "global_step": 235776, "epoch": 2649, "lr": 4.216384880048407e-06}
{"train_loss": 0.04047224298119545, "global_step": 235777, "epoch": 2649, "lr": 4.216151872110119e-06}
{"train_loss": 0.03216009959578514, "global_step": 235778, "epoch": 2649, "lr": 4.215918870326918e-06}
{"train_loss": 0.05788002535700798, "global_step": 235779, "epoch": 2649, "lr": 4.215685874698821e-06}
{"train_loss": 0.0434139259159565, "global_step": 235780, "epoch": 2649, "lr": 4.215452885225868e-06}
{"train_loss": 0.06089749559760094, "global_step": 235781, "epoch": 2649, "lr": 4.215219901908096e-06}
{"train_loss": 0.02007196471095085, "global_step": 235782, "epoch": 2649, "lr": 4.2149869247455236e-06}
{"train_loss": 0.07491762191057205, "global_step": 235783, "epoch": 2649, "lr": 4.2147539537381935e-06}
{"train_loss": 0.02304910682141781, "global_step": 235784, "epoch": 2649, "lr": 4.214520988886128e-06}
{"train_loss": 0.028590990230441093, "global_step": 235785, "epoch": 2649, "lr": 4.214288030189367e-06}
{"train_loss": 0.042360588908195496, "global_step": 235786, "epoch": 2649, "lr": 4.214055077647927e-06}
{"train_loss": 0.05499051511287689, "global_step": 235787, "epoch": 2649, "lr": 4.213822131261863e-06}
{"train_loss": 0.05562175065279007, "global_step": 235788, "epoch": 2649, "lr": 4.213589191031175e-06}
{"train_loss": 0.0470520555973053, "global_step": 235789, "epoch": 2649, "lr": 4.213356256955931e-06}
{"train_loss": 0.03585774824023247, "global_step": 235790, "epoch": 2649, "lr": 4.213123329036128e-06}
{"train_loss": 0.039037976413965225, "global_step": 235791, "epoch": 2649, "lr": 4.212890407271824e-06}
{"train_loss": 0.02415255643427372, "global_step": 235792, "epoch": 2649, "lr": 4.212657491663025e-06}
{"train_loss": 0.012894988991320133, "global_step": 235793, "epoch": 2649, "lr": 4.21242458220979e-06}
{"train_loss": 0.03029085323214531, "global_step": 235794, "epoch": 2649, "lr": 4.2121916789121255e-06}
{"train_loss": 0.03598536178469658, "global_step": 235795, "epoch": 2649, "lr": 4.211958781770087e-06}
{"train_loss": 0.05953444540500641, "global_step": 235796, "epoch": 2649, "lr": 4.2117258907836795e-06}
{"train_loss": 0.049503110349178314, "global_step": 235797, "epoch": 2649, "lr": 4.211493005952955e-06}
{"train_loss": 0.06561578065156937, "global_step": 235798, "epoch": 2649, "lr": 4.211260127277938e-06}
{"train_loss": 0.02621927298605442, "global_step": 235799, "epoch": 2649, "lr": 4.2110272547586535e-06}
{"train_loss": 0.04758056253194809, "global_step": 235800, "epoch": 2649, "lr": 4.210794388395145e-06}
{"train_loss": 0.04699496552348137, "global_step": 235801, "epoch": 2649, "lr": 4.210561528187429e-06}
{"train_loss": 0.04558773338794708, "global_step": 235802, "epoch": 2649, "lr": 4.21032867413555e-06}
{"train_loss": 0.06578465551137924, "global_step": 235803, "epoch": 2649, "lr": 4.2100958262395295e-06}
{"train_loss": 0.0534248873591423, "global_step": 235804, "epoch": 2649, "lr": 4.209862984499413e-06}
{"train_loss": 0.02400624193251133, "global_step": 235805, "epoch": 2649, "lr": 4.2096301489152115e-06}
{"train_loss": 0.07216550409793854, "global_step": 235806, "epoch": 2649, "lr": 4.209397319486974e-06}
{"train_loss": 0.041388433426618576, "global_step": 235807, "epoch": 2649, "lr": 4.209164496214718e-06}
{"train_loss": 0.06622112542390823, "global_step": 235808, "epoch": 2649, "lr": 4.208931679098494e-06}
{"train_loss": 0.02891635149717331, "global_step": 235809, "epoch": 2649, "lr": 4.208698868138306e-06}
{"train_loss": 0.055159371346235275, "global_step": 235810, "epoch": 2649, "lr": 4.208466063334215e-06}
{"train_loss": 0.02544882334768772, "global_step": 235811, "epoch": 2649, "lr": 4.208233264686223e-06}
{"train_loss": 0.03424294665455818, "global_step": 235812, "epoch": 2649, "lr": 4.20800047219439e-06}
{"train_loss": 0.029375294223427773, "global_step": 235813, "epoch": 2649, "lr": 4.207767685858721e-06}
{"train_loss": 0.04768488183617592, "global_step": 235814, "epoch": 2649, "lr": 4.207534905679267e-06}
{"train_loss": 0.027552524581551552, "global_step": 235815, "epoch": 2649, "lr": 4.207302131656049e-06}
{"train_loss": 0.08648078143596649, "global_step": 235816, "epoch": 2649, "lr": 4.207069363789096e-06}
{"train_loss": 0.04620347544550896, "global_step": 235817, "epoch": 2649, "lr": 4.206836602078457e-06}
{"train_loss": 0.044295553117990494, "global_step": 235818, "epoch": 2649, "lr": 4.2066038465241375e-06}
{"train_loss": 0.042268432676792145, "global_step": 235819, "epoch": 2649, "lr": 4.206371097126194e-06}
{"train_loss": 0.09945806115865707, "global_step": 235820, "epoch": 2649, "lr": 4.206138353884631e-06}
{"train_loss": 0.02244699001312256, "global_step": 235821, "epoch": 2649, "lr": 4.20590561679951e-06}
{"train_loss": 0.05907260999083519, "global_step": 235822, "epoch": 2649, "lr": 4.205672885870837e-06}
{"train_loss": 0.023058688268065453, "global_step": 235823, "epoch": 2649, "lr": 4.2054401610986614e-06}
{"train_loss": 0.02484908141195774, "global_step": 235824, "epoch": 2649, "lr": 4.205207442482995e-06}
{"train_loss": 0.030907446518540382, "global_step": 235825, "epoch": 2649, "lr": 4.204974730023891e-06}
{"train_loss": 0.031314607709646225, "global_step": 235826, "epoch": 2649, "lr": 4.204742023721364e-06}
{"train_loss": 0.03406358137726784, "global_step": 235827, "epoch": 2649, "lr": 4.204509323575457e-06}
{"train_loss": 0.046677783131599426, "global_step": 235828, "epoch": 2649, "lr": 4.2042766295861844e-06}
{"train_loss": 0.03385123237967491, "global_step": 235829, "epoch": 2649, "lr": 4.204043941753599e-06}
{"train_loss": 0.04313630610704422, "global_step": 235830, "epoch": 2649, "lr": 4.203811260077717e-06}
{"train_loss": 0.09172497689723969, "global_step": 235831, "epoch": 2649, "lr": 4.203578584558582e-06}
{"train_loss": 0.05685478821396828, "global_step": 235832, "epoch": 2649, "lr": 4.203345915196205e-06}
{"train_loss": 0.03209979459643364, "global_step": 235833, "epoch": 2649, "lr": 4.203113251990643e-06}
{"train_loss": 0.04518185555934906, "global_step": 235834, "epoch": 2649, "lr": 4.2028805949418995e-06}
{"train_loss": 0.036469925194978714, "global_step": 235835, "epoch": 2649, "lr": 4.202647944050036e-06}
{"train_loss": 0.058973874896764755, "global_step": 235836, "epoch": 2649, "lr": 4.202415299315065e-06}
{"train_loss": 0.06809528917074203, "global_step": 235837, "epoch": 2649, "lr": 4.202182660737014e-06}
{"train_loss": 0.05731874331831932, "global_step": 235838, "epoch": 2649, "lr": 4.2019500283159265e-06}
{"train_loss": 0.019735731184482574, "global_step": 235839, "epoch": 2649, "lr": 4.201717402051825e-06}
{"train_loss": 0.02790858782827854, "global_step": 235840, "epoch": 2649, "lr": 4.201484781944748e-06}
{"train_loss": 0.02505849488079548, "global_step": 235841, "epoch": 2649, "lr": 4.201252167994718e-06}
{"train_loss": 0.06974200904369354, "global_step": 235842, "epoch": 2649, "lr": 4.20101956020178e-06}
{"train_loss": 0.11714747548103333, "global_step": 235843, "epoch": 2649, "lr": 4.200786958565944e-06}
{"train_loss": 0.06290996819734573, "global_step": 235844, "epoch": 2649, "lr": 4.200554363087267e-06}
{"train_loss": 0.05119575560092926, "global_step": 235845, "epoch": 2649, "lr": 4.200321773765759e-06}
{"train_loss": 0.014601231552660465, "global_step": 235846, "epoch": 2649, "lr": 4.2000891906014635e-06}
{"train_loss": 0.0463339239358902, "global_step": 235847, "epoch": 2649, "lr": 4.199856613594405e-06}
{"train_loss": 0.08890728652477264, "global_step": 235848, "epoch": 2649, "lr": 4.199624042744621e-06}
{"train_loss": 0.044531023946036113, "global_step": 235849, "epoch": 2649, "lr": 4.199391478052134e-06, "val_loss": 9.314286231994629}
{"train_loss": 0.04972003400325775, "global_step": 235850, "epoch": 2650, "lr": 4.1991589195169875e-06}
{"train_loss": 0.021350817754864693, "global_step": 235851, "epoch": 2650, "lr": 4.1989263671392e-06}
{"train_loss": 0.024500606581568718, "global_step": 235852, "epoch": 2650, "lr": 4.198693820918814e-06}
{"train_loss": 0.008221182972192764, "global_step": 235853, "epoch": 2650, "lr": 4.198461280855848e-06}
{"train_loss": 0.05133230239152908, "global_step": 235854, "epoch": 2650, "lr": 4.1982287469503385e-06}
{"train_loss": 0.023420874029397964, "global_step": 235855, "epoch": 2650, "lr": 4.197996219202327e-06}
{"train_loss": 0.03853611275553703, "global_step": 235856, "epoch": 2650, "lr": 4.197763697611834e-06}
{"train_loss": 0.041305478662252426, "global_step": 235857, "epoch": 2650, "lr": 4.197531182178899e-06}
{"train_loss": 0.07569878548383713, "global_step": 235858, "epoch": 2650, "lr": 4.1972986729035375e-06}
{"train_loss": 0.07318542152643204, "global_step": 235859, "epoch": 2650, "lr": 4.197066169785801e-06}
{"train_loss": 0.038621436804533005, "global_step": 235860, "epoch": 2650, "lr": 4.196833672825701e-06}
{"train_loss": 0.05213989317417145, "global_step": 235861, "epoch": 2650, "lr": 4.196601182023285e-06}
{"train_loss": 0.07806883752346039, "global_step": 235862, "epoch": 2650, "lr": 4.196368697378572e-06}
{"train_loss": 0.03665182366967201, "global_step": 235863, "epoch": 2650, "lr": 4.196136218891611e-06}
{"train_loss": 0.09096210449934006, "global_step": 235864, "epoch": 2650, "lr": 4.195903746562408e-06}
{"train_loss": 0.08502829074859619, "global_step": 235865, "epoch": 2650, "lr": 4.195671280391017e-06}
{"train_loss": 0.03466824069619179, "global_step": 235866, "epoch": 2650, "lr": 4.195438820377451e-06}
{"train_loss": 0.030456317588686943, "global_step": 235867, "epoch": 2650, "lr": 4.195206366521759e-06}
{"train_loss": 0.0406639464199543, "global_step": 235868, "epoch": 2650, "lr": 4.194973918823952e-06}
{"train_loss": 0.049183815717697144, "global_step": 235869, "epoch": 2650, "lr": 4.194741477284087e-06}
{"train_loss": 0.06782218813896179, "global_step": 235870, "epoch": 2650, "lr": 4.1945090419021675e-06}
{"train_loss": 0.043851934373378754, "global_step": 235871, "epoch": 2650, "lr": 4.19427661267825e-06}
{"train_loss": 0.04700538143515587, "global_step": 235872, "epoch": 2650, "lr": 4.19404418961234e-06}
{"train_loss": 0.07018115371465683, "global_step": 235873, "epoch": 2650, "lr": 4.193811772704492e-06}
{"train_loss": 0.04823014885187149, "global_step": 235874, "epoch": 2650, "lr": 4.193579361954731e-06}
{"train_loss": 0.05590164288878441, "global_step": 235875, "epoch": 2650, "lr": 4.193346957363076e-06}
{"train_loss": 0.10920094698667526, "global_step": 235876, "epoch": 2650, "lr": 4.193114558929573e-06}
{"train_loss": 0.05503815412521362, "global_step": 235877, "epoch": 2650, "lr": 4.192882166654244e-06}
{"train_loss": 0.02397536113858223, "global_step": 235878, "epoch": 2650, "lr": 4.192649780537128e-06}
{"train_loss": 0.037669435143470764, "global_step": 235879, "epoch": 2650, "lr": 4.192417400578241e-06}
{"train_loss": 0.01747109740972519, "global_step": 235880, "epoch": 2650, "lr": 4.19218502677764e-06}
{"train_loss": 0.05877750739455223, "global_step": 235881, "epoch": 2650, "lr": 4.191952659135329e-06}
{"train_loss": 0.03665488585829735, "global_step": 235882, "epoch": 2650, "lr": 4.1917202976513635e-06}
{"train_loss": 0.03103233315050602, "global_step": 235883, "epoch": 2650, "lr": 4.19148794232575e-06}
{"train_loss": 0.05879861116409302, "global_step": 235884, "epoch": 2650, "lr": 4.191255593158544e-06}
{"train_loss": 0.01940625160932541, "global_step": 235885, "epoch": 2650, "lr": 4.191023250149756e-06}
{"train_loss": 0.05126975104212761, "global_step": 235886, "epoch": 2650, "lr": 4.1907909132994365e-06}
{"train_loss": 0.04498719796538353, "global_step": 235887, "epoch": 2650, "lr": 4.190558582607595e-06}
{"train_loss": 0.05992085486650467, "global_step": 235888, "epoch": 2650, "lr": 4.1903262580742845e-06}
{"train_loss": 0.06773477047681808, "global_step": 235889, "epoch": 2650, "lr": 4.190093939699519e-06}
{"train_loss": 0.03980956971645355, "global_step": 235890, "epoch": 2650, "lr": 4.189861627483332e-06}
{"train_loss": 0.05963313579559326, "global_step": 235891, "epoch": 2650, "lr": 4.189629321425775e-06}
{"train_loss": 0.06542206555604935, "global_step": 235892, "epoch": 2650, "lr": 4.189397021526853e-06}
{"train_loss": 0.05089898779988289, "global_step": 235893, "epoch": 2650, "lr": 4.189164727786615e-06}
{"train_loss": 0.05660116672515869, "global_step": 235894, "epoch": 2650, "lr": 4.188932440205079e-06}
{"train_loss": 0.049505025148391724, "global_step": 235895, "epoch": 2650, "lr": 4.1887001587822885e-06}
{"train_loss": 0.1244177520275116, "global_step": 235896, "epoch": 2650, "lr": 4.188467883518266e-06}
{"train_loss": 0.03885788843035698, "global_step": 235897, "epoch": 2650, "lr": 4.1882356144130495e-06}
{"train_loss": 0.01943434774875641, "global_step": 235898, "epoch": 2650, "lr": 4.188003351466657e-06}
{"train_loss": 0.07259337604045868, "global_step": 235899, "epoch": 2650, "lr": 4.187771094679144e-06}
{"train_loss": 0.04552096500992775, "global_step": 235900, "epoch": 2650, "lr": 4.1875388440505094e-06}
{"train_loss": 0.04339417442679405, "global_step": 235901, "epoch": 2650, "lr": 4.187306599580814e-06}
{"train_loss": 0.05853627622127533, "global_step": 235902, "epoch": 2650, "lr": 4.187074361270071e-06}
{"train_loss": 0.04142799228429794, "global_step": 235903, "epoch": 2650, "lr": 4.186842129118324e-06}
{"train_loss": 0.052385155111551285, "global_step": 235904, "epoch": 2650, "lr": 4.186609903125594e-06}
{"train_loss": 0.04000615328550339, "global_step": 235905, "epoch": 2650, "lr": 4.1863776832919145e-06}
{"train_loss": 0.0396503210067749, "global_step": 235906, "epoch": 2650, "lr": 4.1861454696173194e-06}
{"train_loss": 0.017867792397737503, "global_step": 235907, "epoch": 2650, "lr": 4.185913262101831e-06}
{"train_loss": 0.07987390458583832, "global_step": 235908, "epoch": 2650, "lr": 4.1856810607455e-06}
{"train_loss": 0.056491669267416, "global_step": 235909, "epoch": 2650, "lr": 4.1854488655483474e-06}
{"train_loss": 0.02666729874908924, "global_step": 235910, "epoch": 2650, "lr": 4.185216676510389e-06}
{"train_loss": 0.08762308955192566, "global_step": 235911, "epoch": 2650, "lr": 4.184984493631683e-06}
{"train_loss": 0.04506794735789299, "global_step": 235912, "epoch": 2650, "lr": 4.184752316912238e-06}
{"train_loss": 0.05175047367811203, "global_step": 235913, "epoch": 2650, "lr": 4.184520146352105e-06}
{"train_loss": 0.06878636032342911, "global_step": 235914, "epoch": 2650, "lr": 4.184287981951301e-06}
{"train_loss": 0.05655577406287193, "global_step": 235915, "epoch": 2650, "lr": 4.184055823709854e-06}
{"train_loss": 0.017852364107966423, "global_step": 235916, "epoch": 2650, "lr": 4.183823671627812e-06}
{"train_loss": 0.04259496554732323, "global_step": 235917, "epoch": 2650, "lr": 4.1835915257051885e-06}
{"train_loss": 0.018398314714431763, "global_step": 235918, "epoch": 2650, "lr": 4.183359385942032e-06}
{"train_loss": 0.03264888748526573, "global_step": 235919, "epoch": 2650, "lr": 4.183127252338348e-06}
{"train_loss": 0.06039366126060486, "global_step": 235920, "epoch": 2650, "lr": 4.182895124894204e-06}
{"train_loss": 0.044727928936481476, "global_step": 235921, "epoch": 2650, "lr": 4.182663003609594e-06}
{"train_loss": 0.031887203454971313, "global_step": 235922, "epoch": 2650, "lr": 4.1824308884845784e-06}
{"train_loss": 0.05559651926159859, "global_step": 235923, "epoch": 2650, "lr": 4.18219877951917e-06}
{"train_loss": 0.04874241352081299, "global_step": 235924, "epoch": 2650, "lr": 4.1819666767134116e-06}
{"train_loss": 0.057841964066028595, "global_step": 235925, "epoch": 2650, "lr": 4.18173458006732e-06}
{"train_loss": 0.030976150184869766, "global_step": 235926, "epoch": 2650, "lr": 4.181502489580941e-06}
{"train_loss": 0.06834186613559723, "global_step": 235927, "epoch": 2650, "lr": 4.181270405254306e-06}
{"train_loss": 0.03796977922320366, "global_step": 235928, "epoch": 2650, "lr": 4.1810383270874274e-06}
{"train_loss": 0.0363909937441349, "global_step": 235929, "epoch": 2650, "lr": 4.180806255080366e-06}
{"train_loss": 0.07026185095310211, "global_step": 235930, "epoch": 2650, "lr": 4.180574189233122e-06}
{"train_loss": 0.013099473901093006, "global_step": 235931, "epoch": 2650, "lr": 4.1803421295457555e-06}
{"train_loss": 0.015064150094985962, "global_step": 235932, "epoch": 2650, "lr": 4.180110076018274e-06}
{"train_loss": 0.032577794045209885, "global_step": 235933, "epoch": 2650, "lr": 4.1798780286507245e-06}
{"train_loss": 0.018774185329675674, "global_step": 235934, "epoch": 2650, "lr": 4.17964598744312e-06}
{"train_loss": 0.040597062557935715, "global_step": 235935, "epoch": 2650, "lr": 4.179413952395522e-06}
{"train_loss": 0.10037226229906082, "global_step": 235936, "epoch": 2650, "lr": 4.1791819235079245e-06}
{"train_loss": 0.03418215736746788, "global_step": 235937, "epoch": 2650, "lr": 4.1789499007803936e-06}
{"train_loss": 0.04834069664265667, "global_step": 235938, "epoch": 2650, "lr": 4.178717884212929e-06, "train/sim_max_reward_0": 0.5614450055667549, "train/sim_max_reward_1": 0.9546070565161039, "train/sim_max_reward_2": 0.6627432860610686, "train/sim_max_reward_3": 0.48923975484318527, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9686081676362341, "test/sim_max_reward_4500000": 0.8226027933636235, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.9555863106770716, "test/sim_max_reward_4500003": 0.1780425002049271, "test/sim_max_reward_4500004": 1.0, "test/sim_max_reward_4500005": 0.7419079442789112, "test/sim_max_reward_4500006": 0.9881648294274614, "test/sim_max_reward_4500007": 1.0, "test/sim_max_reward_4500008": 0.5012005040465286, "test/sim_max_reward_4500009": 0.05056864188741636, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.4454276979727087, "test/sim_max_reward_4500014": 0.9807271941199166, "test/sim_max_reward_4500015": 0.9215640451764768, "test/sim_max_reward_4500016": 0.8920617595462766, "test/sim_max_reward_4500017": 0.2026796600762533, "test/sim_max_reward_4500018": 0.988247616934976, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9341366140345742, "test/sim_max_reward_4500023": 0.9694762018608513, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.43882748962131224, "test/sim_max_reward_4500027": 0.9882860154157345, "test/sim_max_reward_4500028": 0.595222850035422, "test/sim_max_reward_4500029": 1.0, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 1.0, "test/sim_max_reward_4500033": 0.9937356300849075, "test/sim_max_reward_4500034": 0.9883161861269458, "test/sim_max_reward_4500035": 0.4660307433547296, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.052408087166270954, "test/sim_max_reward_4500038": 0.1625449087842638, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.508878026064296, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.972576361900594, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.9699112673745178, "test/sim_max_reward_4500046": 0.40422102974525376, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9930928556138839, "test/sim_max_reward_4500049": 0.9807313199399142, "train/mean_score": 0.7727738784372246, "test/mean_score": 0.6417435416967204, "val_loss": 9.480320930480957, "train_action_mse_error": 47.21753692626953}
{"train_loss": 0.03837316110730171, "global_step": 235939, "epoch": 2651, "lr": 4.178485873805593e-06}
{"train_loss": 0.07009480148553848, "global_step": 235940, "epoch": 2651, "lr": 4.178253869558391e-06}
{"train_loss": 0.038394745439291, "global_step": 235941, "epoch": 2651, "lr": 4.178021871471371e-06}
{"train_loss": 0.0595565065741539, "global_step": 235942, "epoch": 2651, "lr": 4.177789879544547e-06}
{"train_loss": 0.07074472308158875, "global_step": 235943, "epoch": 2651, "lr": 4.177557893777962e-06}
{"train_loss": 0.04296349361538887, "global_step": 235944, "epoch": 2651, "lr": 4.177325914171653e-06}
{"train_loss": 0.04183902218937874, "global_step": 235945, "epoch": 2651, "lr": 4.17709394072564e-06}
{"train_loss": 0.01360790990293026, "global_step": 235946, "epoch": 2651, "lr": 4.1768619734399665e-06}
{"train_loss": 0.06873216480016708, "global_step": 235947, "epoch": 2651, "lr": 4.176630012314653e-06}
{"train_loss": 0.013672935776412487, "global_step": 235948, "epoch": 2651, "lr": 4.176398057349723e-06}
{"train_loss": 0.06397399306297302, "global_step": 235949, "epoch": 2651, "lr": 4.176166108545226e-06}
{"train_loss": 0.038535621017217636, "global_step": 235950, "epoch": 2651, "lr": 4.17593416590118e-06}
{"train_loss": 0.03209421783685684, "global_step": 235951, "epoch": 2651, "lr": 4.175702229417627e-06}
{"train_loss": 0.07214874029159546, "global_step": 235952, "epoch": 2651, "lr": 4.175470299094597e-06}
{"train_loss": 0.04463786259293556, "global_step": 235953, "epoch": 2651, "lr": 4.1752383749321055e-06}
{"train_loss": 0.037061769515275955, "global_step": 235954, "epoch": 2651, "lr": 4.175006456930203e-06}
{"train_loss": 0.0626746341586113, "global_step": 235955, "epoch": 2651, "lr": 4.1747745450889e-06}
{"train_loss": 0.06170981749892235, "global_step": 235956, "epoch": 2651, "lr": 4.174542639408252e-06}
{"train_loss": 0.032844845205545425, "global_step": 235957, "epoch": 2651, "lr": 4.174310739888271e-06}
{"train_loss": 0.03147566318511963, "global_step": 235958, "epoch": 2651, "lr": 4.174078846529e-06}
{"train_loss": 0.08329969644546509, "global_step": 235959, "epoch": 2651, "lr": 4.1738469593304576e-06}
{"train_loss": 0.02391008473932743, "global_step": 235960, "epoch": 2651, "lr": 4.173615078292697e-06}
{"train_loss": 0.06221364438533783, "global_step": 235961, "epoch": 2651, "lr": 4.17338320341572e-06}
{"train_loss": 0.031989943236112595, "global_step": 235962, "epoch": 2651, "lr": 4.173151334699576e-06}
{"train_loss": 0.022605812177062035, "global_step": 235963, "epoch": 2651, "lr": 4.1729194721443046e-06}
{"train_loss": 0.030804594978690147, "global_step": 235964, "epoch": 2651, "lr": 4.172687615749909e-06}
{"train_loss": 0.062419746071100235, "global_step": 235965, "epoch": 2651, "lr": 4.172455765516453e-06}
{"train_loss": 0.1130027025938034, "global_step": 235966, "epoch": 2651, "lr": 4.1722239214439405e-06}
{"train_loss": 0.03247813135385513, "global_step": 235967, "epoch": 2651, "lr": 4.171992083532422e-06}
{"train_loss": 0.04207374155521393, "global_step": 235968, "epoch": 2651, "lr": 4.171760251781914e-06}
{"train_loss": 0.028157413005828857, "global_step": 235969, "epoch": 2651, "lr": 4.171528426192461e-06}
{"train_loss": 0.031098870560526848, "global_step": 235970, "epoch": 2651, "lr": 4.17129660676408e-06}
{"train_loss": 0.03657102957367897, "global_step": 235971, "epoch": 2651, "lr": 4.171064793496821e-06}
{"train_loss": 0.06364099681377411, "global_step": 235972, "epoch": 2651, "lr": 4.170832986390688e-06}
{"train_loss": 0.06619084626436234, "global_step": 235973, "epoch": 2651, "lr": 4.170601185445744e-06}
{"train_loss": 0.04652722179889679, "global_step": 235974, "epoch": 2651, "lr": 4.1703693906619936e-06}
{"train_loss": 0.04984657093882561, "global_step": 235975, "epoch": 2651, "lr": 4.170137602039481e-06}
{"train_loss": 0.05716025084257126, "global_step": 235976, "epoch": 2651, "lr": 4.169905819578234e-06}
{"train_loss": 0.01403983123600483, "global_step": 235977, "epoch": 2651, "lr": 4.169674043278288e-06}
{"train_loss": 0.08798665553331375, "global_step": 235978, "epoch": 2651, "lr": 4.169442273139662e-06}
{"train_loss": 0.060329511761665344, "global_step": 235979, "epoch": 2651, "lr": 4.169210509162408e-06}
{"train_loss": 0.022631298750638962, "global_step": 235980, "epoch": 2651, "lr": 4.168978751346536e-06}
{"train_loss": 0.04766734316945076, "global_step": 235981, "epoch": 2651, "lr": 4.1687469996920856e-06}
{"train_loss": 0.06551241874694824, "global_step": 235982, "epoch": 2651, "lr": 4.1685152541990956e-06}
{"train_loss": 0.033952388912439346, "global_step": 235983, "epoch": 2651, "lr": 4.1682835148675825e-06}
{"train_loss": 0.0903492271900177, "global_step": 235984, "epoch": 2651, "lr": 4.168051781697596e-06}
{"train_loss": 0.06624245643615723, "global_step": 235985, "epoch": 2651, "lr": 4.167820054689153e-06}
{"train_loss": 0.09570035338401794, "global_step": 235986, "epoch": 2651, "lr": 4.1675883338422816e-06}
{"train_loss": 0.026786349713802338, "global_step": 235987, "epoch": 2651, "lr": 4.167356619157026e-06}
{"train_loss": 0.031760357320308685, "global_step": 235988, "epoch": 2651, "lr": 4.167124910633402e-06}
{"train_loss": 0.07281333208084106, "global_step": 235989, "epoch": 2651, "lr": 4.16689320827146e-06}
{"train_loss": 0.07850301265716553, "global_step": 235990, "epoch": 2651, "lr": 4.166661512071218e-06}
{"train_loss": 0.031657177954912186, "global_step": 235991, "epoch": 2651, "lr": 4.166429822032703e-06}
{"train_loss": 0.043222732841968536, "global_step": 235992, "epoch": 2651, "lr": 4.166198138155963e-06}
{"train_loss": 0.0524318590760231, "global_step": 235993, "epoch": 2651, "lr": 4.165966460441006e-06}
{"train_loss": 0.0810270681977272, "global_step": 235994, "epoch": 2651, "lr": 4.165734788887887e-06}
{"train_loss": 0.025184785947203636, "global_step": 235995, "epoch": 2651, "lr": 4.165503123496617e-06}
{"train_loss": 0.07944285124540329, "global_step": 235996, "epoch": 2651, "lr": 4.165271464267245e-06}
{"train_loss": 0.04854081571102142, "global_step": 235997, "epoch": 2651, "lr": 4.165039811199789e-06}
{"train_loss": 0.07200560718774796, "global_step": 235998, "epoch": 2651, "lr": 4.164808164294281e-06}
{"train_loss": 0.05379737168550491, "global_step": 235999, "epoch": 2651, "lr": 4.1645765235507675e-06}
{"train_loss": 0.05195816978812218, "global_step": 236000, "epoch": 2651, "lr": 4.164344888969257e-06}
{"train_loss": 0.05147866904735565, "global_step": 236001, "epoch": 2651, "lr": 4.164113260549801e-06}
{"train_loss": 0.05183310806751251, "global_step": 236002, "epoch": 2651, "lr": 4.163881638292411e-06}
{"train_loss": 0.07516156136989594, "global_step": 236003, "epoch": 2651, "lr": 4.163650022197136e-06}
{"train_loss": 0.03323030099272728, "global_step": 236004, "epoch": 2651, "lr": 4.1634184122639915e-06}
{"train_loss": 0.06628201901912689, "global_step": 236005, "epoch": 2651, "lr": 4.1631868084930306e-06}
{"train_loss": 0.03213801980018616, "global_step": 236006, "epoch": 2651, "lr": 4.162955210884256e-06}
{"train_loss": 0.06489254534244537, "global_step": 236007, "epoch": 2651, "lr": 4.162723619437725e-06}
{"train_loss": 0.008787920698523521, "global_step": 236008, "epoch": 2651, "lr": 4.162492034153448e-06}
{"train_loss": 0.06987462937831879, "global_step": 236009, "epoch": 2651, "lr": 4.162260455031475e-06}
{"train_loss": 0.05202473700046539, "global_step": 236010, "epoch": 2651, "lr": 4.162028882071817e-06}
{"train_loss": 0.006478975061327219, "global_step": 236011, "epoch": 2651, "lr": 4.161797315274524e-06}
{"train_loss": 0.050270795822143555, "global_step": 236012, "epoch": 2651, "lr": 4.1615657546396125e-06}
{"train_loss": 0.055568188428878784, "global_step": 236013, "epoch": 2651, "lr": 4.161334200167127e-06}
{"train_loss": 0.0778692215681076, "global_step": 236014, "epoch": 2651, "lr": 4.161102651857085e-06}
{"train_loss": 0.04928833246231079, "global_step": 236015, "epoch": 2651, "lr": 4.16087110970953e-06}
{"train_loss": 0.01378557551652193, "global_step": 236016, "epoch": 2651, "lr": 4.160639573724478e-06}
{"train_loss": 0.01606198213994503, "global_step": 236017, "epoch": 2651, "lr": 4.16040804390197e-06}
{"train_loss": 0.03188030421733856, "global_step": 236018, "epoch": 2651, "lr": 4.160176520242043e-06}
{"train_loss": 0.04655374959111214, "global_step": 236019, "epoch": 2651, "lr": 4.159945002744714e-06}
{"train_loss": 0.06366772949695587, "global_step": 236020, "epoch": 2651, "lr": 4.1597134914100335e-06}
{"train_loss": 0.08992500603199005, "global_step": 236021, "epoch": 2651, "lr": 4.159481986238012e-06}
{"train_loss": 0.013952079229056835, "global_step": 236022, "epoch": 2651, "lr": 4.1592504872286944e-06}
{"train_loss": 0.02021710015833378, "global_step": 236023, "epoch": 2651, "lr": 4.159018994382097e-06}
{"train_loss": 0.0191787276417017, "global_step": 236024, "epoch": 2651, "lr": 4.158787507698275e-06}
{"train_loss": 0.04586564004421234, "global_step": 236025, "epoch": 2651, "lr": 4.158556027177241e-06}
{"train_loss": 0.03242131695151329, "global_step": 236026, "epoch": 2651, "lr": 4.158324552819026e-06}
{"train_loss": 0.04871037158215146, "global_step": 236027, "epoch": 2651, "lr": 4.158093084623671e-06, "val_loss": 9.36767292022705}
{"train_loss": 0.035509008914232254, "global_step": 236028, "epoch": 2652, "lr": 4.157861622591191e-06}
{"train_loss": 0.01281541958451271, "global_step": 236029, "epoch": 2652, "lr": 4.1576301667216434e-06}
{"train_loss": 0.04729105532169342, "global_step": 236030, "epoch": 2652, "lr": 4.157398717015037e-06}
{"train_loss": 0.012804189696907997, "global_step": 236031, "epoch": 2652, "lr": 4.157167273471402e-06}
{"train_loss": 0.020656557753682137, "global_step": 236032, "epoch": 2652, "lr": 4.156935836090786e-06}
{"train_loss": 0.033509187400341034, "global_step": 236033, "epoch": 2652, "lr": 4.156704404873207e-06}
{"train_loss": 0.03613563999533653, "global_step": 236034, "epoch": 2652, "lr": 4.156472979818693e-06}
{"train_loss": 0.05117331072688103, "global_step": 236035, "epoch": 2652, "lr": 4.156241560927299e-06}
{"train_loss": 0.056197959929704666, "global_step": 236036, "epoch": 2652, "lr": 4.1560101481990244e-06}
{"train_loss": 0.033121686428785324, "global_step": 236037, "epoch": 2652, "lr": 4.155778741633925e-06}
{"train_loss": 0.04891860485076904, "global_step": 236038, "epoch": 2652, "lr": 4.155547341232019e-06}
{"train_loss": 0.062428705394268036, "global_step": 236039, "epoch": 2652, "lr": 4.155315946993343e-06}
{"train_loss": 0.03247927874326706, "global_step": 236040, "epoch": 2652, "lr": 4.155084558917921e-06}
{"train_loss": 0.03930027037858963, "global_step": 236041, "epoch": 2652, "lr": 4.154853177005796e-06}
{"train_loss": 0.04569895192980766, "global_step": 236042, "epoch": 2652, "lr": 4.154621801256986e-06}
{"train_loss": 0.08197028189897537, "global_step": 236043, "epoch": 2652, "lr": 4.154390431671535e-06}
{"train_loss": 0.04037007689476013, "global_step": 236044, "epoch": 2652, "lr": 4.154159068249458e-06}
{"train_loss": 0.04559404030442238, "global_step": 236045, "epoch": 2652, "lr": 4.153927710990807e-06}
{"train_loss": 0.019612014293670654, "global_step": 236046, "epoch": 2652, "lr": 4.153696359895592e-06}
{"train_loss": 0.03572694957256317, "global_step": 236047, "epoch": 2652, "lr": 4.153465014963865e-06}
{"train_loss": 0.044426772743463516, "global_step": 236048, "epoch": 2652, "lr": 4.153233676195633e-06}
{"train_loss": 0.05781562253832817, "global_step": 236049, "epoch": 2652, "lr": 4.153002343590951e-06}
{"train_loss": 0.07832398265600204, "global_step": 236050, "epoch": 2652, "lr": 4.152771017149825e-06}
{"train_loss": 0.03868069872260094, "global_step": 236051, "epoch": 2652, "lr": 4.152539696872315e-06}
{"train_loss": 0.052881333976984024, "global_step": 236052, "epoch": 2652, "lr": 4.152308382758424e-06}
{"train_loss": 0.058711569756269455, "global_step": 236053, "epoch": 2652, "lr": 4.1520770748082025e-06}
{"train_loss": 0.04128431901335716, "global_step": 236054, "epoch": 2652, "lr": 4.151845773021679e-06}
{"train_loss": 0.021630041301250458, "global_step": 236055, "epoch": 2652, "lr": 4.151614477398874e-06}
{"train_loss": 0.08128675073385239, "global_step": 236056, "epoch": 2652, "lr": 4.151383187939834e-06}
{"train_loss": 0.06107621639966965, "global_step": 236057, "epoch": 2652, "lr": 4.151151904644574e-06}
{"train_loss": 0.0939536988735199, "global_step": 236058, "epoch": 2652, "lr": 4.150920627513138e-06}
{"train_loss": 0.02757788449525833, "global_step": 236059, "epoch": 2652, "lr": 4.150689356545545e-06}
{"train_loss": 0.013246831484138966, "global_step": 236060, "epoch": 2652, "lr": 4.150458091741843e-06}
{"train_loss": 0.05253100395202637, "global_step": 236061, "epoch": 2652, "lr": 4.1502268331020445e-06}
{"train_loss": 0.05157303437590599, "global_step": 236062, "epoch": 2652, "lr": 4.1499955806261985e-06}
{"train_loss": 0.03269641846418381, "global_step": 236063, "epoch": 2652, "lr": 4.149764334314326e-06}
{"train_loss": 0.045445699244737625, "global_step": 236064, "epoch": 2652, "lr": 4.149533094166447e-06}
{"train_loss": 0.02892991527915001, "global_step": 236065, "epoch": 2652, "lr": 4.149301860182614e-06}
{"train_loss": 0.08025011420249939, "global_step": 236066, "epoch": 2652, "lr": 4.149070632362845e-06}
{"train_loss": 0.06967256218194962, "global_step": 236067, "epoch": 2652, "lr": 4.148839410707178e-06}
{"train_loss": 0.04446805641055107, "global_step": 236068, "epoch": 2652, "lr": 4.148608195215642e-06}
{"train_loss": 0.04746858775615692, "global_step": 236069, "epoch": 2652, "lr": 4.148376985888258e-06}
{"train_loss": 0.07238982617855072, "global_step": 236070, "epoch": 2652, "lr": 4.148145782725072e-06}
{"train_loss": 0.06638841331005096, "global_step": 236071, "epoch": 2652, "lr": 4.147914585726109e-06}
{"train_loss": 0.055108893662691116, "global_step": 236072, "epoch": 2652, "lr": 4.147683394891394e-06}
{"train_loss": 0.03166727349162102, "global_step": 236073, "epoch": 2652, "lr": 4.147452210220975e-06}
{"train_loss": 0.0504801906645298, "global_step": 236074, "epoch": 2652, "lr": 4.14722103171486e-06}
{"train_loss": 0.07372022420167923, "global_step": 236075, "epoch": 2652, "lr": 4.146989859373102e-06}
{"train_loss": 0.10319755226373672, "global_step": 236076, "epoch": 2652, "lr": 4.146758693195718e-06}
{"train_loss": 0.028384603559970856, "global_step": 236077, "epoch": 2652, "lr": 4.146527533182748e-06}
{"train_loss": 0.08182371407747269, "global_step": 236078, "epoch": 2652, "lr": 4.146296379334208e-06}
{"train_loss": 0.03725402429699898, "global_step": 236079, "epoch": 2652, "lr": 4.1460652316501535e-06}
{"train_loss": 0.06099701672792435, "global_step": 236080, "epoch": 2652, "lr": 4.1458340901305845e-06}
{"train_loss": 0.0715366005897522, "global_step": 236081, "epoch": 2652, "lr": 4.145602954775568e-06}
{"train_loss": 0.07422895729541779, "global_step": 236082, "epoch": 2652, "lr": 4.145371825585104e-06}
{"train_loss": 0.05391726642847061, "global_step": 236083, "epoch": 2652, "lr": 4.145140702559242e-06}
{"train_loss": 0.10109788179397583, "global_step": 236084, "epoch": 2652, "lr": 4.144909585697998e-06}
{"train_loss": 0.04473418742418289, "global_step": 236085, "epoch": 2652, "lr": 4.144678475001423e-06}
{"train_loss": 0.03976091742515564, "global_step": 236086, "epoch": 2652, "lr": 4.144447370469528e-06}
{"train_loss": 0.03808760643005371, "global_step": 236087, "epoch": 2652, "lr": 4.144216272102364e-06}
{"train_loss": 0.04753674939274788, "global_step": 236088, "epoch": 2652, "lr": 4.14398517989994e-06}
{"train_loss": 0.03573503717780113, "global_step": 236089, "epoch": 2652, "lr": 4.143754093862295e-06}
{"train_loss": 0.04554060101509094, "global_step": 236090, "epoch": 2652, "lr": 4.143523013989476e-06}
{"train_loss": 0.07415232062339783, "global_step": 236091, "epoch": 2652, "lr": 4.143291940281496e-06}
{"train_loss": 0.05040090158581734, "global_step": 236092, "epoch": 2652, "lr": 4.1430608727383964e-06}
{"train_loss": 0.05540850758552551, "global_step": 236093, "epoch": 2652, "lr": 4.142829811360193e-06}
{"train_loss": 0.03670158237218857, "global_step": 236094, "epoch": 2652, "lr": 4.1425987561469374e-06}
{"train_loss": 0.03797731548547745, "global_step": 236095, "epoch": 2652, "lr": 4.142367707098643e-06}
{"train_loss": 0.04849453642964363, "global_step": 236096, "epoch": 2652, "lr": 4.142136664215357e-06}
{"train_loss": 0.0620347335934639, "global_step": 236097, "epoch": 2652, "lr": 4.141905627497089e-06}
{"train_loss": 0.0908607691526413, "global_step": 236098, "epoch": 2652, "lr": 4.141674596943896e-06}
{"train_loss": 0.04947048798203468, "global_step": 236099, "epoch": 2652, "lr": 4.1414435725557865e-06}
{"train_loss": 0.0655745193362236, "global_step": 236100, "epoch": 2652, "lr": 4.1412125543328076e-06}
{"train_loss": 0.06990106403827667, "global_step": 236101, "epoch": 2652, "lr": 4.140981542274985e-06}
{"train_loss": 0.07238908857107162, "global_step": 236102, "epoch": 2652, "lr": 4.1407505363823426e-06}
{"train_loss": 0.029901640489697456, "global_step": 236103, "epoch": 2652, "lr": 4.140519536654924e-06}
{"train_loss": 0.04491664096713066, "global_step": 236104, "epoch": 2652, "lr": 4.14028854309274e-06}
{"train_loss": 0.11414043605327606, "global_step": 236105, "epoch": 2652, "lr": 4.1400575556958535e-06}
{"train_loss": 0.05045178905129433, "global_step": 236106, "epoch": 2652, "lr": 4.139826574464273e-06}
{"train_loss": 0.04808058962225914, "global_step": 236107, "epoch": 2652, "lr": 4.139595599398022e-06}
{"train_loss": 0.052619460970163345, "global_step": 236108, "epoch": 2652, "lr": 4.139364630497144e-06}
{"train_loss": 0.05647379904985428, "global_step": 236109, "epoch": 2652, "lr": 4.13913366776168e-06}
{"train_loss": 0.023200582712888718, "global_step": 236110, "epoch": 2652, "lr": 4.138902711191644e-06}
{"train_loss": 0.045903973281383514, "global_step": 236111, "epoch": 2652, "lr": 4.138671760787077e-06}
{"train_loss": 0.027339745312929153, "global_step": 236112, "epoch": 2652, "lr": 4.1384408165480006e-06}
{"train_loss": 0.03933118283748627, "global_step": 236113, "epoch": 2652, "lr": 4.138209878474464e-06}
{"train_loss": 0.006398248951882124, "global_step": 236114, "epoch": 2652, "lr": 4.137978946566473e-06}
{"train_loss": 0.048162613064050674, "global_step": 236115, "epoch": 2652, "lr": 4.137748020824084e-06}
{"train_loss": 0.050122173424559026, "global_step": 236116, "epoch": 2652, "lr": 4.137517101247301e-06, "val_loss": 9.362302780151367}
{"train_loss": 0.028951136395335197, "global_step": 236117, "epoch": 2653, "lr": 4.137286187836187e-06}
{"train_loss": 0.029296543449163437, "global_step": 236118, "epoch": 2653, "lr": 4.13705528059074e-06}
{"train_loss": 0.0513741672039032, "global_step": 236119, "epoch": 2653, "lr": 4.136824379511017e-06}
{"train_loss": 0.012385142035782337, "global_step": 236120, "epoch": 2653, "lr": 4.1365934845970346e-06}
{"train_loss": 0.03789573162794113, "global_step": 236121, "epoch": 2653, "lr": 4.13636259584883e-06}
{"train_loss": 0.07232017070055008, "global_step": 236122, "epoch": 2653, "lr": 4.136131713266428e-06}
{"train_loss": 0.0701429694890976, "global_step": 236123, "epoch": 2653, "lr": 4.135900836849871e-06}
{"train_loss": 0.0268096923828125, "global_step": 236124, "epoch": 2653, "lr": 4.1356699665991765e-06}
{"train_loss": 0.03030364215373993, "global_step": 236125, "epoch": 2653, "lr": 4.135439102514383e-06}
{"train_loss": 0.031941913068294525, "global_step": 236126, "epoch": 2653, "lr": 4.135208244595529e-06}
{"train_loss": 0.06171303987503052, "global_step": 236127, "epoch": 2653, "lr": 4.134977392842626e-06}
{"train_loss": 0.05769851803779602, "global_step": 236128, "epoch": 2653, "lr": 4.134746547255724e-06}
{"train_loss": 0.0436319001019001, "global_step": 236129, "epoch": 2653, "lr": 4.134515707834841e-06}
{"train_loss": 0.02623402141034603, "global_step": 236130, "epoch": 2653, "lr": 4.134284874580025e-06}
{"train_loss": 0.06209453567862511, "global_step": 236131, "epoch": 2653, "lr": 4.1340540474912814e-06}
{"train_loss": 0.01672665774822235, "global_step": 236132, "epoch": 2653, "lr": 4.133823226568662e-06}
{"train_loss": 0.029920056462287903, "global_step": 236133, "epoch": 2653, "lr": 4.133592411812187e-06}
{"train_loss": 0.02429494634270668, "global_step": 236134, "epoch": 2653, "lr": 4.133361603221897e-06}
{"train_loss": 0.039708029478788376, "global_step": 236135, "epoch": 2653, "lr": 4.133130800797813e-06}
{"train_loss": 0.04775803163647652, "global_step": 236136, "epoch": 2653, "lr": 4.1329000045399745e-06}
{"train_loss": 0.05973475053906441, "global_step": 236137, "epoch": 2653, "lr": 4.132669214448404e-06}
{"train_loss": 0.022966479882597923, "global_step": 236138, "epoch": 2653, "lr": 4.132438430523145e-06}
{"train_loss": 0.0940970778465271, "global_step": 236139, "epoch": 2653, "lr": 4.132207652764209e-06}
{"train_loss": 0.09338808804750443, "global_step": 236140, "epoch": 2653, "lr": 4.131976881171651e-06}
{"train_loss": 0.029149657115340233, "global_step": 236141, "epoch": 2653, "lr": 4.1317461157454825e-06}
{"train_loss": 0.03997569531202316, "global_step": 236142, "epoch": 2653, "lr": 4.131515356485743e-06}
{"train_loss": 0.03428303450345993, "global_step": 236143, "epoch": 2653, "lr": 4.131284603392466e-06}
{"train_loss": 0.026279296725988388, "global_step": 236144, "epoch": 2653, "lr": 4.131053856465666e-06}
{"train_loss": 0.05039709806442261, "global_step": 236145, "epoch": 2653, "lr": 4.130823115705401e-06}
{"train_loss": 0.07042739540338516, "global_step": 236146, "epoch": 2653, "lr": 4.1305923811116755e-06}
{"train_loss": 0.08212710171937943, "global_step": 236147, "epoch": 2653, "lr": 4.130361652684545e-06}
{"train_loss": 0.041363559663295746, "global_step": 236148, "epoch": 2653, "lr": 4.13013093042402e-06}
{"train_loss": 0.07040955126285553, "global_step": 236149, "epoch": 2653, "lr": 4.129900214330146e-06}
{"train_loss": 0.07783439010381699, "global_step": 236150, "epoch": 2653, "lr": 4.129669504402939e-06}
{"train_loss": 0.028911888599395752, "global_step": 236151, "epoch": 2653, "lr": 4.129438800642449e-06}
{"train_loss": 0.04179766774177551, "global_step": 236152, "epoch": 2653, "lr": 4.129208103048687e-06}
{"train_loss": 0.03230205550789833, "global_step": 236153, "epoch": 2653, "lr": 4.128977411621704e-06}
{"train_loss": 0.057702235877513885, "global_step": 236154, "epoch": 2653, "lr": 4.12874672636151e-06}
{"train_loss": 0.023603305220603943, "global_step": 236155, "epoch": 2653, "lr": 4.12851604726816e-06}
{"train_loss": 0.048499904572963715, "global_step": 236156, "epoch": 2653, "lr": 4.128285374341656e-06}
{"train_loss": 0.024722298607230186, "global_step": 236157, "epoch": 2653, "lr": 4.128054707582058e-06}
{"train_loss": 0.04380370303988457, "global_step": 236158, "epoch": 2653, "lr": 4.127824046989376e-06}
{"train_loss": 0.03554192930459976, "global_step": 236159, "epoch": 2653, "lr": 4.127593392563655e-06}
{"train_loss": 0.026039721444249153, "global_step": 236160, "epoch": 2653, "lr": 4.127362744304913e-06}
{"train_loss": 0.037718068808317184, "global_step": 236161, "epoch": 2653, "lr": 4.1271321022131885e-06}
{"train_loss": 0.08000319451093674, "global_step": 236162, "epoch": 2653, "lr": 4.126901466288524e-06}
{"train_loss": 0.043580908328294754, "global_step": 236163, "epoch": 2653, "lr": 4.126670836530921e-06}
{"train_loss": 0.0144579466432333, "global_step": 236164, "epoch": 2653, "lr": 4.126440212940447e-06}
{"train_loss": 0.04296838492155075, "global_step": 236165, "epoch": 2653, "lr": 4.1262095955171e-06}
{"train_loss": 0.07059378921985626, "global_step": 236166, "epoch": 2653, "lr": 4.125978984260931e-06}
{"train_loss": 0.039553217589855194, "global_step": 236167, "epoch": 2653, "lr": 4.125748379171957e-06}
{"train_loss": 0.035969022661447525, "global_step": 236168, "epoch": 2653, "lr": 4.1255177802502324e-06}
{"train_loss": 0.03165309876203537, "global_step": 236169, "epoch": 2653, "lr": 4.125287187495758e-06}
{"train_loss": 0.038723837584257126, "global_step": 236170, "epoch": 2653, "lr": 4.125056600908589e-06}
{"train_loss": 0.023810070008039474, "global_step": 236171, "epoch": 2653, "lr": 4.124826020488737e-06}
{"train_loss": 0.03648030385375023, "global_step": 236172, "epoch": 2653, "lr": 4.124595446236257e-06}
{"train_loss": 0.06851532310247421, "global_step": 236173, "epoch": 2653, "lr": 4.124364878151149e-06}
{"train_loss": 0.05726909637451172, "global_step": 236174, "epoch": 2653, "lr": 4.124134316233474e-06}
{"train_loss": 0.037133485078811646, "global_step": 236175, "epoch": 2653, "lr": 4.123903760483244e-06}
{"train_loss": 0.08120738714933395, "global_step": 236176, "epoch": 2653, "lr": 4.123673210900503e-06}
{"train_loss": 0.04090167582035065, "global_step": 236177, "epoch": 2653, "lr": 4.123442667485267e-06}
{"train_loss": 0.03428710252046585, "global_step": 236178, "epoch": 2653, "lr": 4.12321213023758e-06}
{"train_loss": 0.06444086134433746, "global_step": 236179, "epoch": 2653, "lr": 4.122981599157472e-06}
{"train_loss": 0.03947161138057709, "global_step": 236180, "epoch": 2653, "lr": 4.122751074244957e-06}
{"train_loss": 0.038739945739507675, "global_step": 236181, "epoch": 2653, "lr": 4.1225205555000925e-06}
{"train_loss": 0.050381965935230255, "global_step": 236182, "epoch": 2653, "lr": 4.1222900429228885e-06}
{"train_loss": 0.07062194496393204, "global_step": 236183, "epoch": 2653, "lr": 4.122059536513384e-06}
{"train_loss": 0.0702591985464096, "global_step": 236184, "epoch": 2653, "lr": 4.121829036271607e-06}
{"train_loss": 0.11305058002471924, "global_step": 236185, "epoch": 2653, "lr": 4.121598542197596e-06}
{"train_loss": 0.08309297263622284, "global_step": 236186, "epoch": 2653, "lr": 4.121368054291375e-06}
{"train_loss": 0.04953480139374733, "global_step": 236187, "epoch": 2653, "lr": 4.12113757255298e-06}
{"train_loss": 0.047395702451467514, "global_step": 236188, "epoch": 2653, "lr": 4.12090709698243e-06}
{"train_loss": 0.05924544110894203, "global_step": 236189, "epoch": 2653, "lr": 4.120676627579772e-06}
{"train_loss": 0.05463118106126785, "global_step": 236190, "epoch": 2653, "lr": 4.120446164345027e-06}
{"train_loss": 0.057280249893665314, "global_step": 236191, "epoch": 2653, "lr": 4.120215707278235e-06}
{"train_loss": 0.052062131464481354, "global_step": 236192, "epoch": 2653, "lr": 4.119985256379411e-06}
{"train_loss": 0.038260459899902344, "global_step": 236193, "epoch": 2653, "lr": 4.1197548116486015e-06}
{"train_loss": 0.06113675981760025, "global_step": 236194, "epoch": 2653, "lr": 4.119524373085826e-06}
{"train_loss": 0.07051043212413788, "global_step": 236195, "epoch": 2653, "lr": 4.1192939406911324e-06}
{"train_loss": 0.02563195861876011, "global_step": 236196, "epoch": 2653, "lr": 4.119063514464533e-06}
{"train_loss": 0.04641406238079071, "global_step": 236197, "epoch": 2653, "lr": 4.11883309440606e-06}
{"train_loss": 0.03734529763460159, "global_step": 236198, "epoch": 2653, "lr": 4.118602680515765e-06}
{"train_loss": 0.040831226855516434, "global_step": 236199, "epoch": 2653, "lr": 4.1183722727936525e-06}
{"train_loss": 0.08174624294042587, "global_step": 236200, "epoch": 2653, "lr": 4.1181418712397725e-06}
{"train_loss": 0.0614822618663311, "global_step": 236201, "epoch": 2653, "lr": 4.117911475854147e-06}
{"train_loss": 0.04831130430102348, "global_step": 236202, "epoch": 2653, "lr": 4.11768108663681e-06}
{"train_loss": 0.04854746162891388, "global_step": 236203, "epoch": 2653, "lr": 4.117450703587789e-06}
{"train_loss": 0.01054108515381813, "global_step": 236204, "epoch": 2653, "lr": 4.117220326707122e-06}
{"train_loss": 0.04775591396590632, "global_step": 236205, "epoch": 2653, "lr": 4.116989955994832e-06, "val_loss": 9.45087718963623}
{"train_loss": 0.03700415417551994, "global_step": 236206, "epoch": 2654, "lr": 4.116759591450958e-06}
{"train_loss": 0.049451135098934174, "global_step": 236207, "epoch": 2654, "lr": 4.116529233075516e-06}
{"train_loss": 0.034491900354623795, "global_step": 236208, "epoch": 2654, "lr": 4.116298880868558e-06}
{"train_loss": 0.06944116950035095, "global_step": 236209, "epoch": 2654, "lr": 4.116068534830097e-06}
{"train_loss": 0.08392553776502609, "global_step": 236210, "epoch": 2654, "lr": 4.115838194960181e-06}
{"train_loss": 0.04524349421262741, "global_step": 236211, "epoch": 2654, "lr": 4.115607861258819e-06}
{"train_loss": 0.03295767679810524, "global_step": 236212, "epoch": 2654, "lr": 4.115377533726067e-06}
{"train_loss": 0.04249429702758789, "global_step": 236213, "epoch": 2654, "lr": 4.115147212361931e-06}
{"train_loss": 0.018926935270428658, "global_step": 236214, "epoch": 2654, "lr": 4.114916897166466e-06}
{"train_loss": 0.017770569771528244, "global_step": 236215, "epoch": 2654, "lr": 4.114686588139682e-06}
{"train_loss": 0.03674668073654175, "global_step": 236216, "epoch": 2654, "lr": 4.114456285281626e-06}
{"train_loss": 0.04457896947860718, "global_step": 236217, "epoch": 2654, "lr": 4.114225988592324e-06}
{"train_loss": 0.042327117174863815, "global_step": 236218, "epoch": 2654, "lr": 4.113995698071793e-06}
{"train_loss": 0.07268737256526947, "global_step": 236219, "epoch": 2654, "lr": 4.113765413720089e-06}
{"train_loss": 0.036953896284103394, "global_step": 236220, "epoch": 2654, "lr": 4.113535135537217e-06}
{"train_loss": 0.04734371602535248, "global_step": 236221, "epoch": 2654, "lr": 4.113304863523232e-06}
{"train_loss": 0.03549179807305336, "global_step": 236222, "epoch": 2654, "lr": 4.113074597678146e-06}
{"train_loss": 0.04890504106879234, "global_step": 236223, "epoch": 2654, "lr": 4.112844338002004e-06}
{"train_loss": 0.03072449564933777, "global_step": 236224, "epoch": 2654, "lr": 4.112614084494826e-06}
{"train_loss": 0.027223585173487663, "global_step": 236225, "epoch": 2654, "lr": 4.112383837156652e-06}
{"train_loss": 0.06828821450471878, "global_step": 236226, "epoch": 2654, "lr": 4.1121535959875054e-06}
{"train_loss": 0.05147680267691612, "global_step": 236227, "epoch": 2654, "lr": 4.111923360987424e-06}
{"train_loss": 0.09205134958028793, "global_step": 236228, "epoch": 2654, "lr": 4.11169313215643e-06}
{"train_loss": 0.041549261659383774, "global_step": 236229, "epoch": 2654, "lr": 4.111462909494568e-06}
{"train_loss": 0.03827542066574097, "global_step": 236230, "epoch": 2654, "lr": 4.111232693001849e-06}
{"train_loss": 0.02464687079191208, "global_step": 236231, "epoch": 2654, "lr": 4.111002482678322e-06}
{"train_loss": 0.03740953281521797, "global_step": 236232, "epoch": 2654, "lr": 4.110772278524011e-06}
{"train_loss": 0.04655858874320984, "global_step": 236233, "epoch": 2654, "lr": 4.1105420805389485e-06}
{"train_loss": 0.03461860865354538, "global_step": 236234, "epoch": 2654, "lr": 4.110311888723156e-06}
{"train_loss": 0.023458480834960938, "global_step": 236235, "epoch": 2654, "lr": 4.110081703076679e-06}
{"train_loss": 0.04047231003642082, "global_step": 236236, "epoch": 2654, "lr": 4.109851523599545e-06}
{"train_loss": 0.06370917707681656, "global_step": 236237, "epoch": 2654, "lr": 4.109621350291776e-06}
{"train_loss": 0.018903300166130066, "global_step": 236238, "epoch": 2654, "lr": 4.109391183153416e-06}
{"train_loss": 0.07351544499397278, "global_step": 236239, "epoch": 2654, "lr": 4.109161022184482e-06}
{"train_loss": 0.07775664329528809, "global_step": 236240, "epoch": 2654, "lr": 4.108930867385019e-06}
{"train_loss": 0.05262299254536629, "global_step": 236241, "epoch": 2654, "lr": 4.108700718755049e-06}
{"train_loss": 0.016915811225771904, "global_step": 236242, "epoch": 2654, "lr": 4.108470576294604e-06}
{"train_loss": 0.041213370859622955, "global_step": 236243, "epoch": 2654, "lr": 4.1082404400037135e-06}
{"train_loss": 0.029289910569787025, "global_step": 236244, "epoch": 2654, "lr": 4.1080103098824155e-06}
{"train_loss": 0.06317023187875748, "global_step": 236245, "epoch": 2654, "lr": 4.107780185930732e-06}
{"train_loss": 0.058576297014951706, "global_step": 236246, "epoch": 2654, "lr": 4.1075500681487026e-06}
{"train_loss": 0.04610658064484596, "global_step": 236247, "epoch": 2654, "lr": 4.1073199565363425e-06}
{"train_loss": 0.047933854162693024, "global_step": 236248, "epoch": 2654, "lr": 4.1070898510937095e-06}
{"train_loss": 0.059733130037784576, "global_step": 236249, "epoch": 2654, "lr": 4.106859751820808e-06}
{"train_loss": 0.031031321734189987, "global_step": 236250, "epoch": 2654, "lr": 4.106629658717686e-06}
{"train_loss": 0.05700783431529999, "global_step": 236251, "epoch": 2654, "lr": 4.106399571784364e-06}
{"train_loss": 0.0789402574300766, "global_step": 236252, "epoch": 2654, "lr": 4.106169491020878e-06}
{"train_loss": 0.02870197221636772, "global_step": 236253, "epoch": 2654, "lr": 4.105939416427257e-06}
{"train_loss": 0.031192529946565628, "global_step": 236254, "epoch": 2654, "lr": 4.105709348003539e-06}
{"train_loss": 0.04745472967624664, "global_step": 236255, "epoch": 2654, "lr": 4.105479285749741e-06}
{"train_loss": 0.04973512887954712, "global_step": 236256, "epoch": 2654, "lr": 4.105249229665908e-06}
{"train_loss": 0.03556118160486221, "global_step": 236257, "epoch": 2654, "lr": 4.105019179752068e-06}
{"train_loss": 0.03989339619874954, "global_step": 236258, "epoch": 2654, "lr": 4.104789136008241e-06}
{"train_loss": 0.035702526569366455, "global_step": 236259, "epoch": 2654, "lr": 4.104559098434474e-06}
{"train_loss": 0.03494458273053169, "global_step": 236260, "epoch": 2654, "lr": 4.1043290670307764e-06}
{"train_loss": 0.04387689009308815, "global_step": 236261, "epoch": 2654, "lr": 4.104099041797205e-06}
{"train_loss": 0.055364832282066345, "global_step": 236262, "epoch": 2654, "lr": 4.10386902273377e-06}
{"train_loss": 0.047576867043972015, "global_step": 236263, "epoch": 2654, "lr": 4.103639009840515e-06}
{"train_loss": 0.1006212905049324, "global_step": 236264, "epoch": 2654, "lr": 4.103409003117464e-06}
{"train_loss": 0.04082171991467476, "global_step": 236265, "epoch": 2654, "lr": 4.103179002564656e-06}
{"train_loss": 0.047043804079294205, "global_step": 236266, "epoch": 2654, "lr": 4.102949008182105e-06}
{"train_loss": 0.055185671895742416, "global_step": 236267, "epoch": 2654, "lr": 4.102719019969864e-06}
{"train_loss": 0.02714052051305771, "global_step": 236268, "epoch": 2654, "lr": 4.1024890379279424e-06}
{"train_loss": 0.048814259469509125, "global_step": 236269, "epoch": 2654, "lr": 4.102259062056391e-06}
{"train_loss": 0.033908434212207794, "global_step": 236270, "epoch": 2654, "lr": 4.102029092355225e-06}
{"train_loss": 0.05199938267469406, "global_step": 236271, "epoch": 2654, "lr": 4.1017991288244795e-06}
{"train_loss": 0.06850709021091461, "global_step": 236272, "epoch": 2654, "lr": 4.101569171464198e-06}
{"train_loss": 0.04358987882733345, "global_step": 236273, "epoch": 2654, "lr": 4.101339220274392e-06}
{"train_loss": 0.019401364028453827, "global_step": 236274, "epoch": 2654, "lr": 4.1011092752551105e-06}
{"train_loss": 0.027287574484944344, "global_step": 236275, "epoch": 2654, "lr": 4.100879336406366e-06}
{"train_loss": 0.025399891659617424, "global_step": 236276, "epoch": 2654, "lr": 4.100649403728207e-06}
{"train_loss": 0.0709371343255043, "global_step": 236277, "epoch": 2654, "lr": 4.100419477220652e-06}
{"train_loss": 0.07028516381978989, "global_step": 236278, "epoch": 2654, "lr": 4.100189556883738e-06}
{"train_loss": 0.06261789053678513, "global_step": 236279, "epoch": 2654, "lr": 4.099959642717488e-06}
{"train_loss": 0.0791817158460617, "global_step": 236280, "epoch": 2654, "lr": 4.099729734721952e-06}
{"train_loss": 0.03413371369242668, "global_step": 236281, "epoch": 2654, "lr": 4.0994998328971355e-06}
{"train_loss": 0.030849823728203773, "global_step": 236282, "epoch": 2654, "lr": 4.099269937243089e-06}
{"train_loss": 0.06070316210389137, "global_step": 236283, "epoch": 2654, "lr": 4.099040047759828e-06}
{"train_loss": 0.03920688480138779, "global_step": 236284, "epoch": 2654, "lr": 4.098810164447403e-06}
{"train_loss": 0.06605713814496994, "global_step": 236285, "epoch": 2654, "lr": 4.09858028730582e-06}
{"train_loss": 0.09116093069314957, "global_step": 236286, "epoch": 2654, "lr": 4.09835041633514e-06}
{"train_loss": 0.06690222769975662, "global_step": 236287, "epoch": 2654, "lr": 4.0981205515353616e-06}
{"train_loss": 0.02290211245417595, "global_step": 236288, "epoch": 2654, "lr": 4.097890692906531e-06}
{"train_loss": 0.06539705395698547, "global_step": 236289, "epoch": 2654, "lr": 4.097660840448692e-06}
{"train_loss": 0.07448309659957886, "global_step": 236290, "epoch": 2654, "lr": 4.097430994161861e-06}
{"train_loss": 0.06736456602811813, "global_step": 236291, "epoch": 2654, "lr": 4.097201154046065e-06}
{"train_loss": 0.022598043084144592, "global_step": 236292, "epoch": 2654, "lr": 4.09697132010135e-06}
{"train_loss": 0.056698013097047806, "global_step": 236293, "epoch": 2654, "lr": 4.096741492327727e-06}
{"train_loss": 0.04825147038346596, "global_step": 236294, "epoch": 2654, "lr": 4.096511670725245e-06, "val_loss": 9.402351379394531}
{"train_loss": 0.087635338306427, "global_step": 236295, "epoch": 2655, "lr": 4.096281855293926e-06}
{"train_loss": 0.03697189316153526, "global_step": 236296, "epoch": 2655, "lr": 4.0960520460337995e-06}
{"train_loss": 0.03794289752840996, "global_step": 236297, "epoch": 2655, "lr": 4.095822242944902e-06}
{"train_loss": 0.06383121758699417, "global_step": 236298, "epoch": 2655, "lr": 4.095592446027258e-06}
{"train_loss": 0.03141834959387779, "global_step": 236299, "epoch": 2655, "lr": 4.0953626552809044e-06}
{"train_loss": 0.036057647317647934, "global_step": 236300, "epoch": 2655, "lr": 4.09513287070587e-06}
{"train_loss": 0.061316050589084625, "global_step": 236301, "epoch": 2655, "lr": 4.094903092302188e-06}
{"train_loss": 0.05183658376336098, "global_step": 236302, "epoch": 2655, "lr": 4.094673320069881e-06}
{"train_loss": 0.05930701643228531, "global_step": 236303, "epoch": 2655, "lr": 4.094443554008992e-06}
{"train_loss": 0.08027809113264084, "global_step": 236304, "epoch": 2655, "lr": 4.094213794119539e-06}
{"train_loss": 0.06897523254156113, "global_step": 236305, "epoch": 2655, "lr": 4.093984040401566e-06}
{"train_loss": 0.02667076326906681, "global_step": 236306, "epoch": 2655, "lr": 4.09375429285509e-06}
{"train_loss": 0.06050591915845871, "global_step": 236307, "epoch": 2655, "lr": 4.093524551480149e-06}
{"train_loss": 0.04074205830693245, "global_step": 236308, "epoch": 2655, "lr": 4.0932948162767825e-06}
{"train_loss": 0.09039918333292007, "global_step": 236309, "epoch": 2655, "lr": 4.093065087245002e-06}
{"train_loss": 0.04199330136179924, "global_step": 236310, "epoch": 2655, "lr": 4.092835364384861e-06}
{"train_loss": 0.01624240353703499, "global_step": 236311, "epoch": 2655, "lr": 4.092605647696368e-06}
{"train_loss": 0.021076543256640434, "global_step": 236312, "epoch": 2655, "lr": 4.092375937179576e-06}
{"train_loss": 0.05142895132303238, "global_step": 236313, "epoch": 2655, "lr": 4.092146232834493e-06}
{"train_loss": 0.04108399897813797, "global_step": 236314, "epoch": 2655, "lr": 4.091916534661172e-06}
{"train_loss": 0.011816468089818954, "global_step": 236315, "epoch": 2655, "lr": 4.091686842659626e-06}
{"train_loss": 0.100712850689888, "global_step": 236316, "epoch": 2655, "lr": 4.091457156829897e-06}
{"train_loss": 0.09206779301166534, "global_step": 236317, "epoch": 2655, "lr": 4.091227477172011e-06}
{"train_loss": 0.053899042308330536, "global_step": 236318, "epoch": 2655, "lr": 4.090997803686003e-06}
{"train_loss": 0.05894368141889572, "global_step": 236319, "epoch": 2655, "lr": 4.090768136371892e-06}
{"train_loss": 0.044273145496845245, "global_step": 236320, "epoch": 2655, "lr": 4.090538475229727e-06}
{"train_loss": 0.08218533545732498, "global_step": 236321, "epoch": 2655, "lr": 4.09030882025952e-06}
{"train_loss": 0.04006741940975189, "global_step": 236322, "epoch": 2655, "lr": 4.0900791714613254e-06}
{"train_loss": 0.08163204789161682, "global_step": 236323, "epoch": 2655, "lr": 4.0898495288351435e-06}
{"train_loss": 0.06081194803118706, "global_step": 236324, "epoch": 2655, "lr": 4.0896198923810305e-06}
{"train_loss": 0.05682804808020592, "global_step": 236325, "epoch": 2655, "lr": 4.089390262099008e-06}
{"train_loss": 0.10908044874668121, "global_step": 236326, "epoch": 2655, "lr": 4.089160637989103e-06}
{"train_loss": 0.06074262782931328, "global_step": 236327, "epoch": 2655, "lr": 4.088931020051362e-06}
{"train_loss": 0.08781538158655167, "global_step": 236328, "epoch": 2655, "lr": 4.0887014082858e-06}
{"train_loss": 0.05813165754079819, "global_step": 236329, "epoch": 2655, "lr": 4.088471802692451e-06}
{"train_loss": 0.022746454924345016, "global_step": 236330, "epoch": 2655, "lr": 4.088242203271348e-06}
{"train_loss": 0.020602034404873848, "global_step": 236331, "epoch": 2655, "lr": 4.088012610022518e-06}
{"train_loss": 0.0309637188911438, "global_step": 236332, "epoch": 2655, "lr": 4.087783022946002e-06}
{"train_loss": 0.02296062372624874, "global_step": 236333, "epoch": 2655, "lr": 4.087553442041825e-06}
{"train_loss": 0.02324426732957363, "global_step": 236334, "epoch": 2655, "lr": 4.087323867310006e-06}
{"train_loss": 0.022796915844082832, "global_step": 236335, "epoch": 2655, "lr": 4.087094298750594e-06}
{"train_loss": 0.04270075261592865, "global_step": 236336, "epoch": 2655, "lr": 4.086864736363605e-06}
{"train_loss": 0.0737224742770195, "global_step": 236337, "epoch": 2655, "lr": 4.08663518014909e-06}
{"train_loss": 0.02437305636703968, "global_step": 236338, "epoch": 2655, "lr": 4.086405630107054e-06}
{"train_loss": 0.033560339361429214, "global_step": 236339, "epoch": 2655, "lr": 4.086176086237553e-06}
{"train_loss": 0.06471031904220581, "global_step": 236340, "epoch": 2655, "lr": 4.0859465485405925e-06}
{"train_loss": 0.06967013329267502, "global_step": 236341, "epoch": 2655, "lr": 4.085717017016233e-06}
{"train_loss": 0.06399434059858322, "global_step": 236342, "epoch": 2655, "lr": 4.085487491664475e-06}
{"train_loss": 0.02755868248641491, "global_step": 236343, "epoch": 2655, "lr": 4.085257972485363e-06}
{"train_loss": 0.018463648855686188, "global_step": 236344, "epoch": 2655, "lr": 4.08502845947894e-06}
{"train_loss": 0.0162394717335701, "global_step": 236345, "epoch": 2655, "lr": 4.084798952645219e-06}
{"train_loss": 0.03222652152180672, "global_step": 236346, "epoch": 2655, "lr": 4.084569451984238e-06}
{"train_loss": 0.07521389424800873, "global_step": 236347, "epoch": 2655, "lr": 4.084339957496025e-06}
{"train_loss": 0.055002763867378235, "global_step": 236348, "epoch": 2655, "lr": 4.084110469180619e-06}
{"train_loss": 0.14210867881774902, "global_step": 236349, "epoch": 2655, "lr": 4.083880987038036e-06}
{"train_loss": 0.034992851316928864, "global_step": 236350, "epoch": 2655, "lr": 4.083651511068326e-06}
{"train_loss": 0.022313963621854782, "global_step": 236351, "epoch": 2655, "lr": 4.083422041271501e-06}
{"train_loss": 0.027820702642202377, "global_step": 236352, "epoch": 2655, "lr": 4.083192577647604e-06}
{"train_loss": 0.033720098435878754, "global_step": 236353, "epoch": 2655, "lr": 4.082963120196659e-06}
{"train_loss": 0.029416540637612343, "global_step": 236354, "epoch": 2655, "lr": 4.082733668918709e-06}
{"train_loss": 0.060708917677402496, "global_step": 236355, "epoch": 2655, "lr": 4.08250422381376e-06}
{"train_loss": 0.07580597698688507, "global_step": 236356, "epoch": 2655, "lr": 4.082274784881873e-06}
{"train_loss": 0.074343241751194, "global_step": 236357, "epoch": 2655, "lr": 4.08204535212306e-06}
{"train_loss": 0.020242691040039062, "global_step": 236358, "epoch": 2655, "lr": 4.081815925537358e-06}
{"train_loss": 0.062103524804115295, "global_step": 236359, "epoch": 2655, "lr": 4.0815865051247905e-06}
{"train_loss": 0.07449212670326233, "global_step": 236360, "epoch": 2655, "lr": 4.081357090885401e-06}
{"train_loss": 0.11099393665790558, "global_step": 236361, "epoch": 2655, "lr": 4.081127682819203e-06}
{"train_loss": 0.04267016053199768, "global_step": 236362, "epoch": 2655, "lr": 4.080898280926243e-06}
{"train_loss": 0.03888942673802376, "global_step": 236363, "epoch": 2655, "lr": 4.080668885206551e-06}
{"train_loss": 0.039919208735227585, "global_step": 236364, "epoch": 2655, "lr": 4.080439495660143e-06}
{"train_loss": 0.046727504581213, "global_step": 236365, "epoch": 2655, "lr": 4.080210112287075e-06}
{"train_loss": 0.025886017829179764, "global_step": 236366, "epoch": 2655, "lr": 4.079980735087352e-06}
{"train_loss": 0.055679965764284134, "global_step": 236367, "epoch": 2655, "lr": 4.079751364061024e-06}
{"train_loss": 0.03928451985120773, "global_step": 236368, "epoch": 2655, "lr": 4.079521999208114e-06}
{"train_loss": 0.04410858452320099, "global_step": 236369, "epoch": 2655, "lr": 4.079292640528642e-06}
{"train_loss": 0.060496509075164795, "global_step": 236370, "epoch": 2655, "lr": 4.079063288022655e-06}
{"train_loss": 0.02858917787671089, "global_step": 236371, "epoch": 2655, "lr": 4.078833941690174e-06}
{"train_loss": 0.027897028252482414, "global_step": 236372, "epoch": 2655, "lr": 4.078604601531244e-06}
{"train_loss": 0.06331676244735718, "global_step": 236373, "epoch": 2655, "lr": 4.0783752675458864e-06}
{"train_loss": 0.053279805928468704, "global_step": 236374, "epoch": 2655, "lr": 4.078145939734124e-06}
{"train_loss": 0.04535749554634094, "global_step": 236375, "epoch": 2655, "lr": 4.0779166180960005e-06}
{"train_loss": 0.05722435936331749, "global_step": 236376, "epoch": 2655, "lr": 4.077687302631533e-06}
{"train_loss": 0.04043065011501312, "global_step": 236377, "epoch": 2655, "lr": 4.077457993340766e-06}
{"train_loss": 0.02981884405016899, "global_step": 236378, "epoch": 2655, "lr": 4.077228690223722e-06}
{"train_loss": 0.03872030973434448, "global_step": 236379, "epoch": 2655, "lr": 4.076999393280434e-06}
{"train_loss": 0.035216692835092545, "global_step": 236380, "epoch": 2655, "lr": 4.07677010251094e-06}
{"train_loss": 0.04185207188129425, "global_step": 236381, "epoch": 2655, "lr": 4.076540817915259e-06}
{"train_loss": 0.019297778606414795, "global_step": 236382, "epoch": 2655, "lr": 4.076311539493433e-06}
{"train_loss": 0.049823038782296554, "global_step": 236383, "epoch": 2655, "lr": 4.0760822672454794e-06, "val_loss": 9.238173484802246, "train_action_mse_error": 14.173896789550781}
{"train_loss": 0.029773622751235962, "global_step": 236384, "epoch": 2656, "lr": 4.075853001171448e-06}
{"train_loss": 0.09329083561897278, "global_step": 236385, "epoch": 2656, "lr": 4.075623741271345e-06}
{"train_loss": 0.03291104733943939, "global_step": 236386, "epoch": 2656, "lr": 4.075394487545226e-06}
{"train_loss": 0.03552443906664848, "global_step": 236387, "epoch": 2656, "lr": 4.0751652399931006e-06}
{"train_loss": 0.03996929153800011, "global_step": 236388, "epoch": 2656, "lr": 4.0749359986150205e-06}
{"train_loss": 0.0690443366765976, "global_step": 236389, "epoch": 2656, "lr": 4.074706763410996e-06}
{"train_loss": 0.040032822638750076, "global_step": 236390, "epoch": 2656, "lr": 4.074477534381077e-06}
{"train_loss": 0.09641237556934357, "global_step": 236391, "epoch": 2656, "lr": 4.074248311525275e-06}
{"train_loss": 0.04456726461648941, "global_step": 236392, "epoch": 2656, "lr": 4.07401909484364e-06}
{"train_loss": 0.08698999881744385, "global_step": 236393, "epoch": 2656, "lr": 4.073789884336182e-06}
{"train_loss": 0.025374731048941612, "global_step": 236394, "epoch": 2656, "lr": 4.073560680002953e-06}
{"train_loss": 0.013791626319289207, "global_step": 236395, "epoch": 2656, "lr": 4.073331481843967e-06}
{"train_loss": 0.05529516562819481, "global_step": 236396, "epoch": 2656, "lr": 4.073102289859265e-06}
{"train_loss": 0.039756011217832565, "global_step": 236397, "epoch": 2656, "lr": 4.072873104048874e-06}
{"train_loss": 0.036207232624292374, "global_step": 236398, "epoch": 2656, "lr": 4.072643924412822e-06}
{"train_loss": 0.03614374250173569, "global_step": 236399, "epoch": 2656, "lr": 4.072414750951148e-06}
{"train_loss": 0.029370540753006935, "global_step": 236400, "epoch": 2656, "lr": 4.0721855836638736e-06}
{"train_loss": 0.05073573440313339, "global_step": 236401, "epoch": 2656, "lr": 4.071956422551043e-06}
{"train_loss": 0.01522055920213461, "global_step": 236402, "epoch": 2656, "lr": 4.0717272676126685e-06}
{"train_loss": 0.04655439779162407, "global_step": 236403, "epoch": 2656, "lr": 4.071498118848799e-06}
{"train_loss": 0.029524261131882668, "global_step": 236404, "epoch": 2656, "lr": 4.071268976259452e-06}
{"train_loss": 0.05903077498078346, "global_step": 236405, "epoch": 2656, "lr": 4.0710398398446645e-06}
{"train_loss": 0.051954347640275955, "global_step": 236406, "epoch": 2656, "lr": 4.070810709604472e-06}
{"train_loss": 0.029188573360443115, "global_step": 236407, "epoch": 2656, "lr": 4.070581585538885e-06}
{"train_loss": 0.050248466432094574, "global_step": 236408, "epoch": 2656, "lr": 4.070352467647965e-06}
{"train_loss": 0.03997863084077835, "global_step": 236409, "epoch": 2656, "lr": 4.0701233559317095e-06}
{"train_loss": 0.01791551522910595, "global_step": 236410, "epoch": 2656, "lr": 4.069894250390177e-06}
{"train_loss": 0.053063929080963135, "global_step": 236411, "epoch": 2656, "lr": 4.069665151023389e-06}
{"train_loss": 0.024354148656129837, "global_step": 236412, "epoch": 2656, "lr": 4.0694360578313664e-06}
{"train_loss": 0.0459204725921154, "global_step": 236413, "epoch": 2656, "lr": 4.069206970814154e-06}
{"train_loss": 0.05041879788041115, "global_step": 236414, "epoch": 2656, "lr": 4.0689778899717756e-06}
{"train_loss": 0.08653374761343002, "global_step": 236415, "epoch": 2656, "lr": 4.068748815304257e-06}
{"train_loss": 0.034733694046735764, "global_step": 236416, "epoch": 2656, "lr": 4.068519746811644e-06}
{"train_loss": 0.07186345756053925, "global_step": 236417, "epoch": 2656, "lr": 4.0682906844939525e-06}
{"train_loss": 0.02213490940630436, "global_step": 236418, "epoch": 2656, "lr": 4.0680616283512275e-06}
{"train_loss": 0.038723066449165344, "global_step": 236419, "epoch": 2656, "lr": 4.067832578383479e-06}
{"train_loss": 0.04886865243315697, "global_step": 236420, "epoch": 2656, "lr": 4.067603534590764e-06}
{"train_loss": 0.018235238268971443, "global_step": 236421, "epoch": 2656, "lr": 4.067374496973092e-06}
{"train_loss": 0.0624820701777935, "global_step": 236422, "epoch": 2656, "lr": 4.067145465530503e-06}
{"train_loss": 0.037594933062791824, "global_step": 236423, "epoch": 2656, "lr": 4.066916440263024e-06}
{"train_loss": 0.026956742629408836, "global_step": 236424, "epoch": 2656, "lr": 4.066687421170695e-06}
{"train_loss": 0.06356078386306763, "global_step": 236425, "epoch": 2656, "lr": 4.066458408253537e-06}
{"train_loss": 0.13183936476707458, "global_step": 236426, "epoch": 2656, "lr": 4.0662294015115846e-06}
{"train_loss": 0.03256651386618614, "global_step": 236427, "epoch": 2656, "lr": 4.066000400944864e-06}
{"train_loss": 0.042765673249959946, "global_step": 236428, "epoch": 2656, "lr": 4.065771406553415e-06}
{"train_loss": 0.04657205939292908, "global_step": 236429, "epoch": 2656, "lr": 4.0655424183372595e-06}
{"train_loss": 0.05594300106167793, "global_step": 236430, "epoch": 2656, "lr": 4.065313436296436e-06}
{"train_loss": 0.038366127759218216, "global_step": 236431, "epoch": 2656, "lr": 4.065084460430962e-06}
{"train_loss": 0.07897341251373291, "global_step": 236432, "epoch": 2656, "lr": 4.064855490740888e-06}
{"train_loss": 0.017526822164654732, "global_step": 236433, "epoch": 2656, "lr": 4.064626527226228e-06}
{"train_loss": 0.05194342881441116, "global_step": 236434, "epoch": 2656, "lr": 4.064397569887013e-06}
{"train_loss": 0.10859565436840057, "global_step": 236435, "epoch": 2656, "lr": 4.064168618723296e-06}
{"train_loss": 0.05882792919874191, "global_step": 236436, "epoch": 2656, "lr": 4.0639396737350785e-06}
{"train_loss": 0.0653989166021347, "global_step": 236437, "epoch": 2656, "lr": 4.063710734922415e-06}
{"train_loss": 0.05023699626326561, "global_step": 236438, "epoch": 2656, "lr": 4.063481802285313e-06}
{"train_loss": 0.027911709621548653, "global_step": 236439, "epoch": 2656, "lr": 4.063252875823831e-06}
{"train_loss": 0.0992448478937149, "global_step": 236440, "epoch": 2656, "lr": 4.06302395553797e-06}
{"train_loss": 0.0713585764169693, "global_step": 236441, "epoch": 2656, "lr": 4.062795041427786e-06}
{"train_loss": 0.04755254462361336, "global_step": 236442, "epoch": 2656, "lr": 4.062566133493295e-06}
{"train_loss": 0.057697512209415436, "global_step": 236443, "epoch": 2656, "lr": 4.062337231734537e-06}
{"train_loss": 0.03301497548818588, "global_step": 236444, "epoch": 2656, "lr": 4.062108336151538e-06}
{"train_loss": 0.055347904562950134, "global_step": 236445, "epoch": 2656, "lr": 4.061879446744321e-06}
{"train_loss": 0.03783692792057991, "global_step": 236446, "epoch": 2656, "lr": 4.061650563512931e-06}
{"train_loss": 0.05406695604324341, "global_step": 236447, "epoch": 2656, "lr": 4.061421686457384e-06}
{"train_loss": 0.06372634321451187, "global_step": 236448, "epoch": 2656, "lr": 4.06119281557773e-06}
{"train_loss": 0.07700546830892563, "global_step": 236449, "epoch": 2656, "lr": 4.060963950873986e-06}
{"train_loss": 0.0686420351266861, "global_step": 236450, "epoch": 2656, "lr": 4.060735092346174e-06}
{"train_loss": 0.08222958445549011, "global_step": 236451, "epoch": 2656, "lr": 4.060506239994344e-06}
{"train_loss": 0.014561763033270836, "global_step": 236452, "epoch": 2656, "lr": 4.060277393818523e-06}
{"train_loss": 0.10648619383573532, "global_step": 236453, "epoch": 2656, "lr": 4.06004855381873e-06}
{"train_loss": 0.021278226748108864, "global_step": 236454, "epoch": 2656, "lr": 4.059819719995012e-06}
{"train_loss": 0.026296352967619896, "global_step": 236455, "epoch": 2656, "lr": 4.059590892347381e-06}
{"train_loss": 0.06882354617118835, "global_step": 236456, "epoch": 2656, "lr": 4.059362070875888e-06}
{"train_loss": 0.07971639931201935, "global_step": 236457, "epoch": 2656, "lr": 4.0591332555805495e-06}
{"train_loss": 0.02628832682967186, "global_step": 236458, "epoch": 2656, "lr": 4.058904446461403e-06}
{"train_loss": 0.09401297569274902, "global_step": 236459, "epoch": 2656, "lr": 4.058675643518473e-06}
{"train_loss": 0.04934576898813248, "global_step": 236460, "epoch": 2656, "lr": 4.058446846751796e-06}
{"train_loss": 0.029492279514670372, "global_step": 236461, "epoch": 2656, "lr": 4.0582180561614e-06}
{"train_loss": 0.026292894035577774, "global_step": 236462, "epoch": 2656, "lr": 4.05798927174732e-06}
{"train_loss": 0.07509069889783859, "global_step": 236463, "epoch": 2656, "lr": 4.057760493509577e-06}
{"train_loss": 0.0429055318236351, "global_step": 236464, "epoch": 2656, "lr": 4.057531721448216e-06}
{"train_loss": 0.045548371970653534, "global_step": 236465, "epoch": 2656, "lr": 4.057302955563252e-06}
{"train_loss": 0.03793283924460411, "global_step": 236466, "epoch": 2656, "lr": 4.057074195854732e-06}
{"train_loss": 0.02615618146955967, "global_step": 236467, "epoch": 2656, "lr": 4.056845442322671e-06}
{"train_loss": 0.03488069400191307, "global_step": 236468, "epoch": 2656, "lr": 4.0566166949671135e-06}
{"train_loss": 0.0471939742565155, "global_step": 236469, "epoch": 2656, "lr": 4.0563879537880766e-06}
{"train_loss": 0.03219563886523247, "global_step": 236470, "epoch": 2656, "lr": 4.056159218785599e-06}
{"train_loss": 0.08255749195814133, "global_step": 236471, "epoch": 2656, "lr": 4.055930489959714e-06}
{"train_loss": 0.05019057593360711, "global_step": 236472, "epoch": 2656, "lr": 4.055701767310449e-06, "val_loss": 9.314949989318848}
{"train_loss": 0.020348811522126198, "global_step": 236473, "epoch": 2657, "lr": 4.055473050837838e-06}
{"train_loss": 0.04922637715935707, "global_step": 236474, "epoch": 2657, "lr": 4.055244340541903e-06}
{"train_loss": 0.043468624353408813, "global_step": 236475, "epoch": 2657, "lr": 4.055015636422688e-06}
{"train_loss": 0.01557171531021595, "global_step": 236476, "epoch": 2657, "lr": 4.054786938480204e-06}
{"train_loss": 0.03555511683225632, "global_step": 236477, "epoch": 2657, "lr": 4.0545582467145075e-06}
{"train_loss": 0.05814535543322563, "global_step": 236478, "epoch": 2657, "lr": 4.0543295611256026e-06}
{"train_loss": 0.021970469504594803, "global_step": 236479, "epoch": 2657, "lr": 4.0541008817135465e-06}
{"train_loss": 0.01595809869468212, "global_step": 236480, "epoch": 2657, "lr": 4.0538722084783435e-06}
{"train_loss": 0.0423903651535511, "global_step": 236481, "epoch": 2657, "lr": 4.0536435414200495e-06}
{"train_loss": 0.045435741543769836, "global_step": 236482, "epoch": 2657, "lr": 4.0534148805386755e-06}
{"train_loss": 0.041197340935468674, "global_step": 236483, "epoch": 2657, "lr": 4.0531862258342655e-06}
{"train_loss": 0.05963269621133804, "global_step": 236484, "epoch": 2657, "lr": 4.0529575773068425e-06}
{"train_loss": 0.06648128479719162, "global_step": 236485, "epoch": 2657, "lr": 4.052728934956434e-06}
{"train_loss": 0.0646250769495964, "global_step": 236486, "epoch": 2657, "lr": 4.052500298783085e-06}
{"train_loss": 0.04233809933066368, "global_step": 236487, "epoch": 2657, "lr": 4.052271668786811e-06}
{"train_loss": 0.06406601518392563, "global_step": 236488, "epoch": 2657, "lr": 4.0520430449676515e-06}
{"train_loss": 0.01814458705484867, "global_step": 236489, "epoch": 2657, "lr": 4.051814427325629e-06}
{"train_loss": 0.03406521677970886, "global_step": 236490, "epoch": 2657, "lr": 4.051585815860792e-06}
{"train_loss": 0.033574607223272324, "global_step": 236491, "epoch": 2657, "lr": 4.051357210573148e-06}
{"train_loss": 0.08534786850214005, "global_step": 236492, "epoch": 2657, "lr": 4.051128611462746e-06}
{"train_loss": 0.023601898923516273, "global_step": 236493, "epoch": 2657, "lr": 4.050900018529602e-06}
{"train_loss": 0.034192800521850586, "global_step": 236494, "epoch": 2657, "lr": 4.0506714317737615e-06}
{"train_loss": 0.07963606715202332, "global_step": 236495, "epoch": 2657, "lr": 4.050442851195241e-06}
{"train_loss": 0.0622418150305748, "global_step": 236496, "epoch": 2657, "lr": 4.0502142767940895e-06}
{"train_loss": 0.04253537580370903, "global_step": 236497, "epoch": 2657, "lr": 4.0499857085703145e-06}
{"train_loss": 0.0175090990960598, "global_step": 236498, "epoch": 2657, "lr": 4.0497571465239694e-06}
{"train_loss": 0.046781040728092194, "global_step": 236499, "epoch": 2657, "lr": 4.049528590655066e-06}
{"train_loss": 0.017590167000889778, "global_step": 236500, "epoch": 2657, "lr": 4.0493000409636495e-06}
{"train_loss": 0.03541853651404381, "global_step": 236501, "epoch": 2657, "lr": 4.049071497449742e-06}
{"train_loss": 0.06038329005241394, "global_step": 236502, "epoch": 2657, "lr": 4.0488429601133805e-06}
{"train_loss": 0.04332725703716278, "global_step": 236503, "epoch": 2657, "lr": 4.048614428954583e-06}
{"train_loss": 0.03905655816197395, "global_step": 236504, "epoch": 2657, "lr": 4.048385903973395e-06}
{"train_loss": 0.12406027317047119, "global_step": 236505, "epoch": 2657, "lr": 4.048157385169837e-06}
{"train_loss": 0.062359027564525604, "global_step": 236506, "epoch": 2657, "lr": 4.047928872543944e-06}
{"train_loss": 0.044872842729091644, "global_step": 236507, "epoch": 2657, "lr": 4.047700366095758e-06}
{"train_loss": 0.02930881828069687, "global_step": 236508, "epoch": 2657, "lr": 4.047471865825287e-06}
{"train_loss": 0.05796597898006439, "global_step": 236509, "epoch": 2657, "lr": 4.047243371732584e-06}
{"train_loss": 0.0364459753036499, "global_step": 236510, "epoch": 2657, "lr": 4.047014883817657e-06}
{"train_loss": 0.059853941202163696, "global_step": 236511, "epoch": 2657, "lr": 4.04678640208056e-06}
{"train_loss": 0.03844795748591423, "global_step": 236512, "epoch": 2657, "lr": 4.046557926521299e-06}
{"train_loss": 0.030563686043024063, "global_step": 236513, "epoch": 2657, "lr": 4.046329457139936e-06}
{"train_loss": 0.04450958967208862, "global_step": 236514, "epoch": 2657, "lr": 4.046100993936469e-06}
{"train_loss": 0.07107148319482803, "global_step": 236515, "epoch": 2657, "lr": 4.045872536910949e-06}
{"train_loss": 0.10520699620246887, "global_step": 236516, "epoch": 2657, "lr": 4.0456440860633995e-06}
{"train_loss": 0.061074960976839066, "global_step": 236517, "epoch": 2657, "lr": 4.045415641393857e-06}
{"train_loss": 0.04140470176935196, "global_step": 236518, "epoch": 2657, "lr": 4.045187202902345e-06}
{"train_loss": 0.01988411322236061, "global_step": 236519, "epoch": 2657, "lr": 4.044958770588903e-06}
{"train_loss": 0.07217220216989517, "global_step": 236520, "epoch": 2657, "lr": 4.044730344453546e-06}
{"train_loss": 0.01382855512201786, "global_step": 236521, "epoch": 2657, "lr": 4.044501924496325e-06}
{"train_loss": 0.02951209247112274, "global_step": 236522, "epoch": 2657, "lr": 4.044273510717261e-06}
{"train_loss": 0.05356420949101448, "global_step": 236523, "epoch": 2657, "lr": 4.0440451031163726e-06}
{"train_loss": 0.057203102856874466, "global_step": 236524, "epoch": 2657, "lr": 4.0438167016937145e-06}
{"train_loss": 0.029123947024345398, "global_step": 236525, "epoch": 2657, "lr": 4.043588306449297e-06}
{"train_loss": 0.04895840957760811, "global_step": 236526, "epoch": 2657, "lr": 4.043359917383166e-06}
{"train_loss": 0.044909775257110596, "global_step": 236527, "epoch": 2657, "lr": 4.043131534495337e-06}
{"train_loss": 0.04338331148028374, "global_step": 236528, "epoch": 2657, "lr": 4.0429031577858545e-06}
{"train_loss": 0.05869077891111374, "global_step": 236529, "epoch": 2657, "lr": 4.042674787254741e-06}
{"train_loss": 0.060261670500040054, "global_step": 236530, "epoch": 2657, "lr": 4.0424464229020355e-06}
{"train_loss": 0.04711532965302467, "global_step": 236531, "epoch": 2657, "lr": 4.042218064727748e-06}
{"train_loss": 0.06347935646772385, "global_step": 236532, "epoch": 2657, "lr": 4.041989712731942e-06}
{"train_loss": 0.07783467322587967, "global_step": 236533, "epoch": 2657, "lr": 4.04176136691462e-06}
{"train_loss": 0.024968981742858887, "global_step": 236534, "epoch": 2657, "lr": 4.041533027275829e-06}
{"train_loss": 0.07894202321767807, "global_step": 236535, "epoch": 2657, "lr": 4.041304693815584e-06}
{"train_loss": 0.04857344925403595, "global_step": 236536, "epoch": 2657, "lr": 4.041076366533936e-06}
{"train_loss": 0.048881031572818756, "global_step": 236537, "epoch": 2657, "lr": 4.040848045430895e-06}
{"train_loss": 0.03914300352334976, "global_step": 236538, "epoch": 2657, "lr": 4.040619730506512e-06}
{"train_loss": 0.06171939894556999, "global_step": 236539, "epoch": 2657, "lr": 4.040391421760798e-06}
{"train_loss": 0.06612341105937958, "global_step": 236540, "epoch": 2657, "lr": 4.040163119193802e-06}
{"train_loss": 0.03266601264476776, "global_step": 236541, "epoch": 2657, "lr": 4.039934822805541e-06}
{"train_loss": 0.03896548971533775, "global_step": 236542, "epoch": 2657, "lr": 4.0397065325960436e-06}
{"train_loss": 0.05926140770316124, "global_step": 236543, "epoch": 2657, "lr": 4.039478248565365e-06}
{"train_loss": 0.027006473392248154, "global_step": 236544, "epoch": 2657, "lr": 4.039249970713504e-06}
{"train_loss": 0.029736975207924843, "global_step": 236545, "epoch": 2657, "lr": 4.0390216990405126e-06}
{"train_loss": 0.06011968106031418, "global_step": 236546, "epoch": 2657, "lr": 4.038793433546406e-06}
{"train_loss": 0.08052898943424225, "global_step": 236547, "epoch": 2657, "lr": 4.038565174231235e-06}
{"train_loss": 0.03253508359193802, "global_step": 236548, "epoch": 2657, "lr": 4.0383369210950095e-06}
{"train_loss": 0.06728437542915344, "global_step": 236549, "epoch": 2657, "lr": 4.03810867413778e-06}
{"train_loss": 0.06385412067174911, "global_step": 236550, "epoch": 2657, "lr": 4.037880433359553e-06}
{"train_loss": 0.03523321822285652, "global_step": 236551, "epoch": 2657, "lr": 4.037652198760389e-06}
{"train_loss": 0.06025105342268944, "global_step": 236552, "epoch": 2657, "lr": 4.037423970340287e-06}
{"train_loss": 0.08828775584697723, "global_step": 236553, "epoch": 2657, "lr": 4.037195748099304e-06}
{"train_loss": 0.03123590350151062, "global_step": 236554, "epoch": 2657, "lr": 4.03696753203745e-06}
{"train_loss": 0.019938113167881966, "global_step": 236555, "epoch": 2657, "lr": 4.036739322154776e-06}
{"train_loss": 0.022799229249358177, "global_step": 236556, "epoch": 2657, "lr": 4.036511118451292e-06}
{"train_loss": 0.051180023699998856, "global_step": 236557, "epoch": 2657, "lr": 4.0362829209270484e-06}
{"train_loss": 0.04260231554508209, "global_step": 236558, "epoch": 2657, "lr": 4.036054729582056e-06}
{"train_loss": 0.044634681195020676, "global_step": 236559, "epoch": 2657, "lr": 4.035826544416366e-06}
{"train_loss": 0.015061245299875736, "global_step": 236560, "epoch": 2657, "lr": 4.035598365429999e-06}
{"train_loss": 0.04719329278999835, "global_step": 236561, "epoch": 2657, "lr": 4.035370192622978e-06, "val_loss": 9.25932788848877}
{"train_loss": 0.046340785920619965, "global_step": 236562, "epoch": 2658, "lr": 4.0351420259953476e-06}
{"train_loss": 0.05610254406929016, "global_step": 236563, "epoch": 2658, "lr": 4.034913865547124e-06}
{"train_loss": 0.03479078784584999, "global_step": 236564, "epoch": 2658, "lr": 4.034685711278358e-06}
{"train_loss": 0.1115863099694252, "global_step": 236565, "epoch": 2658, "lr": 4.034457563189059e-06}
{"train_loss": 0.05435669422149658, "global_step": 236566, "epoch": 2658, "lr": 4.034229421279273e-06}
{"train_loss": 0.07475937902927399, "global_step": 236567, "epoch": 2658, "lr": 4.034001285549016e-06}
{"train_loss": 0.035812024027109146, "global_step": 236568, "epoch": 2658, "lr": 4.033773155998333e-06}
{"train_loss": 0.039531536400318146, "global_step": 236569, "epoch": 2658, "lr": 4.033545032627245e-06}
{"train_loss": 0.0287362989038229, "global_step": 236570, "epoch": 2658, "lr": 4.033316915435792e-06}
{"train_loss": 0.027097823098301888, "global_step": 236571, "epoch": 2658, "lr": 4.033088804423996e-06}
{"train_loss": 0.0531916469335556, "global_step": 236572, "epoch": 2658, "lr": 4.0328606995918946e-06}
{"train_loss": 0.036984726786613464, "global_step": 236573, "epoch": 2658, "lr": 4.032632600939507e-06}
{"train_loss": 0.036860886961221695, "global_step": 236574, "epoch": 2658, "lr": 4.032404508466881e-06}
{"train_loss": 0.04780973494052887, "global_step": 236575, "epoch": 2658, "lr": 4.0321764221740334e-06}
{"train_loss": 0.011939223855733871, "global_step": 236576, "epoch": 2658, "lr": 4.031948342061004e-06}
{"train_loss": 0.03451037406921387, "global_step": 236577, "epoch": 2658, "lr": 4.031720268127809e-06}
{"train_loss": 0.019277695566415787, "global_step": 236578, "epoch": 2658, "lr": 4.031492200374487e-06}
{"train_loss": 0.05295996367931366, "global_step": 236579, "epoch": 2658, "lr": 4.031264138801083e-06}
{"train_loss": 0.038529254496097565, "global_step": 236580, "epoch": 2658, "lr": 4.031036083407608e-06}
{"train_loss": 0.02897709049284458, "global_step": 236581, "epoch": 2658, "lr": 4.030808034194106e-06}
{"train_loss": 0.027030756697058678, "global_step": 236582, "epoch": 2658, "lr": 4.030579991160593e-06}
{"train_loss": 0.07807742059230804, "global_step": 236583, "epoch": 2658, "lr": 4.030351954307116e-06}
{"train_loss": 0.03709876164793968, "global_step": 236584, "epoch": 2658, "lr": 4.030123923633694e-06}
{"train_loss": 0.05127805843949318, "global_step": 236585, "epoch": 2658, "lr": 4.0298958991403625e-06}
{"train_loss": 0.0282851941883564, "global_step": 236586, "epoch": 2658, "lr": 4.029667880827148e-06}
{"train_loss": 0.060612473636865616, "global_step": 236587, "epoch": 2658, "lr": 4.02943986869409e-06}
{"train_loss": 0.04893268644809723, "global_step": 236588, "epoch": 2658, "lr": 4.029211862741211e-06}
{"train_loss": 0.07901722192764282, "global_step": 236589, "epoch": 2658, "lr": 4.028983862968549e-06}
{"train_loss": 0.051937468349933624, "global_step": 236590, "epoch": 2658, "lr": 4.028755869376116e-06}
{"train_loss": 0.011774449609220028, "global_step": 236591, "epoch": 2658, "lr": 4.028527881963973e-06}
{"train_loss": 0.08794968575239182, "global_step": 236592, "epoch": 2658, "lr": 4.0282999007321246e-06}
{"train_loss": 0.05084804445505142, "global_step": 236593, "epoch": 2658, "lr": 4.028071925680616e-06}
{"train_loss": 0.06950819492340088, "global_step": 236594, "epoch": 2658, "lr": 4.027843956809463e-06}
{"train_loss": 0.03466809168457985, "global_step": 236595, "epoch": 2658, "lr": 4.027615994118716e-06}
{"train_loss": 0.05376967787742615, "global_step": 236596, "epoch": 2658, "lr": 4.027388037608393e-06}
{"train_loss": 0.042919427156448364, "global_step": 236597, "epoch": 2658, "lr": 4.027160087278531e-06}
{"train_loss": 0.0208840761333704, "global_step": 236598, "epoch": 2658, "lr": 4.026932143129148e-06}
{"train_loss": 0.04410558566451073, "global_step": 236599, "epoch": 2658, "lr": 4.026704205160292e-06}
{"train_loss": 0.05221789330244064, "global_step": 236600, "epoch": 2658, "lr": 4.026476273371987e-06}
{"train_loss": 0.08282754570245743, "global_step": 236601, "epoch": 2658, "lr": 4.02624834776425e-06}
{"train_loss": 0.05526149272918701, "global_step": 236602, "epoch": 2658, "lr": 4.026020428337135e-06}
{"train_loss": 0.020567985251545906, "global_step": 236603, "epoch": 2658, "lr": 4.025792515090654e-06}
{"train_loss": 0.031218376010656357, "global_step": 236604, "epoch": 2658, "lr": 4.025564608024851e-06}
{"train_loss": 0.037069059908390045, "global_step": 236605, "epoch": 2658, "lr": 4.025336707139743e-06}
{"train_loss": 0.043210893869400024, "global_step": 236606, "epoch": 2658, "lr": 4.025108812435374e-06}
{"train_loss": 0.03885127976536751, "global_step": 236607, "epoch": 2658, "lr": 4.024880923911761e-06}
{"train_loss": 0.09301389753818512, "global_step": 236608, "epoch": 2658, "lr": 4.0246530415689546e-06}
{"train_loss": 0.08412273228168488, "global_step": 236609, "epoch": 2658, "lr": 4.024425165406964e-06}
{"train_loss": 0.04918661341071129, "global_step": 236610, "epoch": 2658, "lr": 4.024197295425835e-06}
{"train_loss": 0.05058848857879639, "global_step": 236611, "epoch": 2658, "lr": 4.023969431625585e-06}
{"train_loss": 0.03699738532304764, "global_step": 236612, "epoch": 2658, "lr": 4.0237415740062625e-06}
{"train_loss": 0.03313269466161728, "global_step": 236613, "epoch": 2658, "lr": 4.023513722567873e-06}
{"train_loss": 0.02576279640197754, "global_step": 236614, "epoch": 2658, "lr": 4.023285877310473e-06}
{"train_loss": 0.03715118393301964, "global_step": 236615, "epoch": 2658, "lr": 4.0230580382340785e-06}
{"train_loss": 0.017821375280618668, "global_step": 236616, "epoch": 2658, "lr": 4.0228302053387175e-06}
{"train_loss": 0.028336863964796066, "global_step": 236617, "epoch": 2658, "lr": 4.0226023786244346e-06}
{"train_loss": 0.03770151734352112, "global_step": 236618, "epoch": 2658, "lr": 4.022374558091252e-06}
{"train_loss": 0.06349342316389084, "global_step": 236619, "epoch": 2658, "lr": 4.022146743739202e-06}
{"train_loss": 0.05258013308048248, "global_step": 236620, "epoch": 2658, "lr": 4.0219189355683085e-06}
{"train_loss": 0.04534561187028885, "global_step": 236621, "epoch": 2658, "lr": 4.021691133578615e-06}
{"train_loss": 0.05563891679048538, "global_step": 236622, "epoch": 2658, "lr": 4.0214633377701315e-06}
{"train_loss": 0.04661038517951965, "global_step": 236623, "epoch": 2658, "lr": 4.021235548142915e-06}
{"train_loss": 0.03591432049870491, "global_step": 236624, "epoch": 2658, "lr": 4.021007764696977e-06}
{"train_loss": 0.05533241108059883, "global_step": 236625, "epoch": 2658, "lr": 4.02077998743236e-06}
{"train_loss": 0.03827740624547005, "global_step": 236626, "epoch": 2658, "lr": 4.020552216349077e-06}
{"train_loss": 0.04893691837787628, "global_step": 236627, "epoch": 2658, "lr": 4.020324451447183e-06}
{"train_loss": 0.05898058041930199, "global_step": 236628, "epoch": 2658, "lr": 4.020096692726688e-06}
{"train_loss": 0.06465347111225128, "global_step": 236629, "epoch": 2658, "lr": 4.019868940187638e-06}
{"train_loss": 0.026320703327655792, "global_step": 236630, "epoch": 2658, "lr": 4.019641193830043e-06}
{"train_loss": 0.0832776203751564, "global_step": 236631, "epoch": 2658, "lr": 4.019413453653964e-06}
{"train_loss": 0.040415018796920776, "global_step": 236632, "epoch": 2658, "lr": 4.0191857196594016e-06}
{"train_loss": 0.03890758752822876, "global_step": 236633, "epoch": 2658, "lr": 4.018957991846406e-06}
{"train_loss": 0.10450755059719086, "global_step": 236634, "epoch": 2658, "lr": 4.018730270214993e-06}
{"train_loss": 0.015146533958613873, "global_step": 236635, "epoch": 2658, "lr": 4.018502554765213e-06}
{"train_loss": 0.027941370382905006, "global_step": 236636, "epoch": 2658, "lr": 4.018274845497072e-06}
{"train_loss": 0.06067582964897156, "global_step": 236637, "epoch": 2658, "lr": 4.018047142410625e-06}
{"train_loss": 0.10505355894565582, "global_step": 236638, "epoch": 2658, "lr": 4.017819445505888e-06}
{"train_loss": 0.05498656630516052, "global_step": 236639, "epoch": 2658, "lr": 4.01759175478289e-06}
{"train_loss": 0.05216372385621071, "global_step": 236640, "epoch": 2658, "lr": 4.017364070241675e-06}
{"train_loss": 0.046532902866601944, "global_step": 236641, "epoch": 2658, "lr": 4.017136391882254e-06}
{"train_loss": 0.06994806230068207, "global_step": 236642, "epoch": 2658, "lr": 4.016908719704676e-06}
{"train_loss": 0.05015142634510994, "global_step": 236643, "epoch": 2658, "lr": 4.01668105370896e-06}
{"train_loss": 0.06469010561704636, "global_step": 236644, "epoch": 2658, "lr": 4.016453393895148e-06}
{"train_loss": 0.05500705540180206, "global_step": 236645, "epoch": 2658, "lr": 4.016225740263252e-06}
{"train_loss": 0.029075298458337784, "global_step": 236646, "epoch": 2658, "lr": 4.015998092813327e-06}
{"train_loss": 0.06053541973233223, "global_step": 236647, "epoch": 2658, "lr": 4.015770451545375e-06}
{"train_loss": 0.04972020164132118, "global_step": 236648, "epoch": 2658, "lr": 4.0155428164594555e-06}
{"train_loss": 0.065336674451828, "global_step": 236649, "epoch": 2658, "lr": 4.0153151875555795e-06}
{"train_loss": 0.048575588910097486, "global_step": 236650, "epoch": 2658, "lr": 4.015087564833791e-06, "val_loss": 9.242584228515625}
{"train_loss": 0.05570879578590393, "global_step": 236651, "epoch": 2659, "lr": 4.014859948294103e-06}
{"train_loss": 0.0700274333357811, "global_step": 236652, "epoch": 2659, "lr": 4.014632337936558e-06}
{"train_loss": 0.06032014265656471, "global_step": 236653, "epoch": 2659, "lr": 4.01440473376119e-06}
{"train_loss": 0.08123791962862015, "global_step": 236654, "epoch": 2659, "lr": 4.014177135768021e-06}
{"train_loss": 0.055445391684770584, "global_step": 236655, "epoch": 2659, "lr": 4.013949543957091e-06}
{"train_loss": 0.020937426015734673, "global_step": 236656, "epoch": 2659, "lr": 4.0137219583284205e-06}
{"train_loss": 0.006187859922647476, "global_step": 236657, "epoch": 2659, "lr": 4.013494378882049e-06}
{"train_loss": 0.03810514509677887, "global_step": 236658, "epoch": 2659, "lr": 4.0132668056179935e-06}
{"train_loss": 0.017964616417884827, "global_step": 236659, "epoch": 2659, "lr": 4.013039238536309e-06}
{"train_loss": 0.04035061225295067, "global_step": 236660, "epoch": 2659, "lr": 4.0128116776369966e-06}
{"train_loss": 0.04090305417776108, "global_step": 236661, "epoch": 2659, "lr": 4.01258412292011e-06}
{"train_loss": 0.04375841096043587, "global_step": 236662, "epoch": 2659, "lr": 4.012356574385662e-06}
{"train_loss": 0.024367081001400948, "global_step": 236663, "epoch": 2659, "lr": 4.012129032033701e-06}
{"train_loss": 0.031127169728279114, "global_step": 236664, "epoch": 2659, "lr": 4.011901495864245e-06}
{"train_loss": 0.07542464882135391, "global_step": 236665, "epoch": 2659, "lr": 4.0116739658773325e-06}
{"train_loss": 0.054732806980609894, "global_step": 236666, "epoch": 2659, "lr": 4.01144644207298e-06}
{"train_loss": 0.01572614163160324, "global_step": 236667, "epoch": 2659, "lr": 4.011218924451238e-06}
{"train_loss": 0.025369793176651, "global_step": 236668, "epoch": 2659, "lr": 4.010991413012122e-06}
{"train_loss": 0.0530131533741951, "global_step": 236669, "epoch": 2659, "lr": 4.01076390775566e-06}
{"train_loss": 0.060972001403570175, "global_step": 236670, "epoch": 2659, "lr": 4.010536408681903e-06}
{"train_loss": 0.013748941011726856, "global_step": 236671, "epoch": 2659, "lr": 4.010308915790873e-06}
{"train_loss": 0.06991102546453476, "global_step": 236672, "epoch": 2659, "lr": 4.010081429082585e-06}
{"train_loss": 0.03830087184906006, "global_step": 236673, "epoch": 2659, "lr": 4.009853948557085e-06}
{"train_loss": 0.04124839976429939, "global_step": 236674, "epoch": 2659, "lr": 4.009626474214395e-06}
{"train_loss": 0.05146121606230736, "global_step": 236675, "epoch": 2659, "lr": 4.009399006054559e-06}
{"train_loss": 0.0454321913421154, "global_step": 236676, "epoch": 2659, "lr": 4.0091715440776e-06}
{"train_loss": 0.025538943707942963, "global_step": 236677, "epoch": 2659, "lr": 4.008944088283534e-06}
{"train_loss": 0.049249567091464996, "global_step": 236678, "epoch": 2659, "lr": 4.008716638672416e-06}
{"train_loss": 0.05122843757271767, "global_step": 236679, "epoch": 2659, "lr": 4.008489195244259e-06}
{"train_loss": 0.016952497884631157, "global_step": 236680, "epoch": 2659, "lr": 4.0082617579991096e-06}
{"train_loss": 0.05354469269514084, "global_step": 236681, "epoch": 2659, "lr": 4.008034326936977e-06}
{"train_loss": 0.03891788423061371, "global_step": 236682, "epoch": 2659, "lr": 4.007806902057915e-06}
{"train_loss": 0.04729568585753441, "global_step": 236683, "epoch": 2659, "lr": 4.0075794833619285e-06}
{"train_loss": 0.07396624982357025, "global_step": 236684, "epoch": 2659, "lr": 4.007352070849074e-06}
{"train_loss": 0.04528283327817917, "global_step": 236685, "epoch": 2659, "lr": 4.007124664519363e-06}
{"train_loss": 0.035045262426137924, "global_step": 236686, "epoch": 2659, "lr": 4.006897264372844e-06}
{"train_loss": 0.0672709196805954, "global_step": 236687, "epoch": 2659, "lr": 4.006669870409524e-06}
{"train_loss": 0.06973189860582352, "global_step": 236688, "epoch": 2659, "lr": 4.006442482629446e-06}
{"train_loss": 0.0720483809709549, "global_step": 236689, "epoch": 2659, "lr": 4.006215101032657e-06}
{"train_loss": 0.01157399546355009, "global_step": 236690, "epoch": 2659, "lr": 4.005987725619159e-06}
{"train_loss": 0.039056289941072464, "global_step": 236691, "epoch": 2659, "lr": 4.0057603563890046e-06}
{"train_loss": 0.015021201223134995, "global_step": 236692, "epoch": 2659, "lr": 4.005532993342204e-06}
{"train_loss": 0.051316116005182266, "global_step": 236693, "epoch": 2659, "lr": 4.005305636478807e-06}
{"train_loss": 0.032483067363500595, "global_step": 236694, "epoch": 2659, "lr": 4.00507828579883e-06}
{"train_loss": 0.03247235715389252, "global_step": 236695, "epoch": 2659, "lr": 4.004850941302318e-06}
{"train_loss": 0.026133254170417786, "global_step": 236696, "epoch": 2659, "lr": 4.004623602989282e-06}
{"train_loss": 0.02290702424943447, "global_step": 236697, "epoch": 2659, "lr": 4.004396270859778e-06}
{"train_loss": 0.033298976719379425, "global_step": 236698, "epoch": 2659, "lr": 4.004168944913811e-06}
{"train_loss": 0.02807498350739479, "global_step": 236699, "epoch": 2659, "lr": 4.0039416251514304e-06}
{"train_loss": 0.07498908787965775, "global_step": 236700, "epoch": 2659, "lr": 4.003714311572649e-06}
{"train_loss": 0.06573338061571121, "global_step": 236701, "epoch": 2659, "lr": 4.00348700417752e-06}
{"train_loss": 0.023962926119565964, "global_step": 236702, "epoch": 2659, "lr": 4.00325970296605e-06}
{"train_loss": 0.0433901883661747, "global_step": 236703, "epoch": 2659, "lr": 4.003032407938296e-06}
{"train_loss": 0.028086837381124496, "global_step": 236704, "epoch": 2659, "lr": 4.002805119094255e-06}
{"train_loss": 0.07977109402418137, "global_step": 236705, "epoch": 2659, "lr": 4.002577836433985e-06}
{"train_loss": 0.05987539514899254, "global_step": 236706, "epoch": 2659, "lr": 4.002350559957513e-06}
{"train_loss": 0.04953603446483612, "global_step": 236707, "epoch": 2659, "lr": 4.002123289664855e-06}
{"train_loss": 0.04924418032169342, "global_step": 236708, "epoch": 2659, "lr": 4.001896025556057e-06}
{"train_loss": 0.02867085672914982, "global_step": 236709, "epoch": 2659, "lr": 4.0016687676311395e-06}
{"train_loss": 0.026269150897860527, "global_step": 236710, "epoch": 2659, "lr": 4.001441515890142e-06}
{"train_loss": 0.04416308179497719, "global_step": 236711, "epoch": 2659, "lr": 4.0012142703330925e-06}
{"train_loss": 0.05275795981287956, "global_step": 236712, "epoch": 2659, "lr": 4.000987030960013e-06}
{"train_loss": 0.024374959990382195, "global_step": 236713, "epoch": 2659, "lr": 4.000759797770942e-06}
{"train_loss": 0.0321379229426384, "global_step": 236714, "epoch": 2659, "lr": 4.000532570765903e-06}
{"train_loss": 0.07103525102138519, "global_step": 236715, "epoch": 2659, "lr": 4.0003053499449435e-06}
{"train_loss": 0.010585335083305836, "global_step": 236716, "epoch": 2659, "lr": 4.000078135308078e-06}
{"train_loss": 0.08827976137399673, "global_step": 236717, "epoch": 2659, "lr": 3.999850926855336e-06}
{"train_loss": 0.055005401372909546, "global_step": 236718, "epoch": 2659, "lr": 3.99962372458676e-06}
{"train_loss": 0.03057835064828396, "global_step": 236719, "epoch": 2659, "lr": 3.999396528502364e-06}
{"train_loss": 0.041056398302316666, "global_step": 236720, "epoch": 2659, "lr": 3.999169338602199e-06}
{"train_loss": 0.039563167840242386, "global_step": 236721, "epoch": 2659, "lr": 3.998942154886276e-06}
{"train_loss": 0.02669588476419449, "global_step": 236722, "epoch": 2659, "lr": 3.99871497735464e-06}
{"train_loss": 0.03602561727166176, "global_step": 236723, "epoch": 2659, "lr": 3.9984878060073115e-06}
{"train_loss": 0.01828489638864994, "global_step": 236724, "epoch": 2659, "lr": 3.998260640844326e-06}
{"train_loss": 0.0208390224725008, "global_step": 236725, "epoch": 2659, "lr": 3.9980334818657205e-06}
{"train_loss": 0.07566580176353455, "global_step": 236726, "epoch": 2659, "lr": 3.997806329071513e-06}
{"train_loss": 0.06135370954871178, "global_step": 236727, "epoch": 2659, "lr": 3.997579182461741e-06}
{"train_loss": 0.03160705789923668, "global_step": 236728, "epoch": 2659, "lr": 3.997352042036434e-06}
{"train_loss": 0.039099935442209244, "global_step": 236729, "epoch": 2659, "lr": 3.997124907795624e-06}
{"train_loss": 0.03924296051263809, "global_step": 236730, "epoch": 2659, "lr": 3.996897779739334e-06}
{"train_loss": 0.04306129738688469, "global_step": 236731, "epoch": 2659, "lr": 3.996670657867612e-06}
{"train_loss": 0.04663749784231186, "global_step": 236732, "epoch": 2659, "lr": 3.996443542180467e-06}
{"train_loss": 0.05054841563105583, "global_step": 236733, "epoch": 2659, "lr": 3.996216432677946e-06}
{"train_loss": 0.08145193755626678, "global_step": 236734, "epoch": 2659, "lr": 3.995989329360061e-06}
{"train_loss": 0.035063959658145905, "global_step": 236735, "epoch": 2659, "lr": 3.995762232226869e-06}
{"train_loss": 0.0428069531917572, "global_step": 236736, "epoch": 2659, "lr": 3.995535141278378e-06}
{"train_loss": 0.02754601277410984, "global_step": 236737, "epoch": 2659, "lr": 3.995308056514629e-06}
{"train_loss": 0.03523842990398407, "global_step": 236738, "epoch": 2659, "lr": 3.9950809779356445e-06}
{"train_loss": 0.043283682892078096, "global_step": 236739, "epoch": 2659, "lr": 3.994853905541474e-06, "val_loss": 9.264639854431152}
{"train_loss": 0.05610009655356407, "global_step": 236740, "epoch": 2660, "lr": 3.994626839332122e-06}
{"train_loss": 0.030389517545700073, "global_step": 236741, "epoch": 2660, "lr": 3.994399779307639e-06}
{"train_loss": 0.06648488342761993, "global_step": 236742, "epoch": 2660, "lr": 3.994172725468043e-06}
{"train_loss": 0.038203880190849304, "global_step": 236743, "epoch": 2660, "lr": 3.9939456778133655e-06}
{"train_loss": 0.05639725923538208, "global_step": 236744, "epoch": 2660, "lr": 3.993718636343652e-06}
{"train_loss": 0.0642509013414383, "global_step": 236745, "epoch": 2660, "lr": 3.993491601058913e-06}
{"train_loss": 0.05458523705601692, "global_step": 236746, "epoch": 2660, "lr": 3.993264571959199e-06}
{"train_loss": 0.044511910527944565, "global_step": 236747, "epoch": 2660, "lr": 3.99303754904452e-06}
{"train_loss": 0.04806298762559891, "global_step": 236748, "epoch": 2660, "lr": 3.9928105323149275e-06}
{"train_loss": 0.055869799107313156, "global_step": 236749, "epoch": 2660, "lr": 3.992583521770438e-06}
{"train_loss": 0.03080316260457039, "global_step": 236750, "epoch": 2660, "lr": 3.9923565174110725e-06}
{"train_loss": 0.03648042678833008, "global_step": 236751, "epoch": 2660, "lr": 3.992129519236887e-06}
{"train_loss": 0.03931238502264023, "global_step": 236752, "epoch": 2660, "lr": 3.991902527247893e-06}
{"train_loss": 0.05407370999455452, "global_step": 236753, "epoch": 2660, "lr": 3.9916755414441285e-06}
{"train_loss": 0.07523615658283234, "global_step": 236754, "epoch": 2660, "lr": 3.991448561825628e-06}
{"train_loss": 0.05292188376188278, "global_step": 236755, "epoch": 2660, "lr": 3.9912215883924075e-06}
{"train_loss": 0.07154075801372528, "global_step": 236756, "epoch": 2660, "lr": 3.990994621144512e-06}
{"train_loss": 0.045338526368141174, "global_step": 236757, "epoch": 2660, "lr": 3.990767660081962e-06}
{"train_loss": 0.014828388579189777, "global_step": 236758, "epoch": 2660, "lr": 3.990540705204798e-06}
{"train_loss": 0.030127346515655518, "global_step": 236759, "epoch": 2660, "lr": 3.990313756513037e-06}
{"train_loss": 0.04483328387141228, "global_step": 236760, "epoch": 2660, "lr": 3.990086814006721e-06}
{"train_loss": 0.02995726838707924, "global_step": 236761, "epoch": 2660, "lr": 3.989859877685881e-06}
{"train_loss": 0.09947995841503143, "global_step": 236762, "epoch": 2660, "lr": 3.989632947550537e-06}
{"train_loss": 0.05055613815784454, "global_step": 236763, "epoch": 2660, "lr": 3.989406023600733e-06}
{"train_loss": 0.019255606457591057, "global_step": 236764, "epoch": 2660, "lr": 3.989179105836488e-06}
{"train_loss": 0.06290680170059204, "global_step": 236765, "epoch": 2660, "lr": 3.988952194257839e-06}
{"train_loss": 0.06571297347545624, "global_step": 236766, "epoch": 2660, "lr": 3.988725288864809e-06}
{"train_loss": 0.030128031969070435, "global_step": 236767, "epoch": 2660, "lr": 3.988498389657447e-06}
{"train_loss": 0.1110113114118576, "global_step": 236768, "epoch": 2660, "lr": 3.9882714966357596e-06}
{"train_loss": 0.07013288140296936, "global_step": 236769, "epoch": 2660, "lr": 3.988044609799791e-06}
{"train_loss": 0.042088087648153305, "global_step": 236770, "epoch": 2660, "lr": 3.987817729149568e-06}
{"train_loss": 0.028647685423493385, "global_step": 236771, "epoch": 2660, "lr": 3.987590854685125e-06}
{"train_loss": 0.06217684969305992, "global_step": 236772, "epoch": 2660, "lr": 3.987363986406484e-06}
{"train_loss": 0.050676535815000534, "global_step": 236773, "epoch": 2660, "lr": 3.987137124313689e-06}
{"train_loss": 0.03834427520632744, "global_step": 236774, "epoch": 2660, "lr": 3.986910268406757e-06}
{"train_loss": 0.04220493137836456, "global_step": 236775, "epoch": 2660, "lr": 3.9866834186857325e-06}
{"train_loss": 0.04178039729595184, "global_step": 236776, "epoch": 2660, "lr": 3.986456575150627e-06}
{"train_loss": 0.03420243039727211, "global_step": 236777, "epoch": 2660, "lr": 3.986229737801489e-06}
{"train_loss": 0.038271378725767136, "global_step": 236778, "epoch": 2660, "lr": 3.986002906638331e-06}
{"train_loss": 0.012245438061654568, "global_step": 236779, "epoch": 2660, "lr": 3.9857760816611964e-06}
{"train_loss": 0.05272311717271805, "global_step": 236780, "epoch": 2660, "lr": 3.985549262870125e-06}
{"train_loss": 0.03447984158992767, "global_step": 236781, "epoch": 2660, "lr": 3.985322450265122e-06}
{"train_loss": 0.04391862452030182, "global_step": 236782, "epoch": 2660, "lr": 3.985095643846242e-06}
{"train_loss": 0.07284310460090637, "global_step": 236783, "epoch": 2660, "lr": 3.984868843613499e-06}
{"train_loss": 0.016133353114128113, "global_step": 236784, "epoch": 2660, "lr": 3.984642049566939e-06}
{"train_loss": 0.05889212712645531, "global_step": 236785, "epoch": 2660, "lr": 3.98441526170657e-06}
{"train_loss": 0.036292895674705505, "global_step": 236786, "epoch": 2660, "lr": 3.984188480032447e-06}
{"train_loss": 0.04809505492448807, "global_step": 236787, "epoch": 2660, "lr": 3.983961704544587e-06}
{"train_loss": 0.049100302159786224, "global_step": 236788, "epoch": 2660, "lr": 3.983734935243016e-06}
{"train_loss": 0.03559909388422966, "global_step": 236789, "epoch": 2660, "lr": 3.983508172127775e-06}
{"train_loss": 0.036772023886442184, "global_step": 236790, "epoch": 2660, "lr": 3.983281415198886e-06}
{"train_loss": 0.021439481526613235, "global_step": 236791, "epoch": 2660, "lr": 3.983054664456392e-06}
{"train_loss": 0.04927254468202591, "global_step": 236792, "epoch": 2660, "lr": 3.982827919900317e-06}
{"train_loss": 0.08574321866035461, "global_step": 236793, "epoch": 2660, "lr": 3.982601181530676e-06}
{"train_loss": 0.04614315927028656, "global_step": 236794, "epoch": 2660, "lr": 3.982374449347526e-06}
{"train_loss": 0.04367269203066826, "global_step": 236795, "epoch": 2660, "lr": 3.982147723350877e-06}
{"train_loss": 0.07585267722606659, "global_step": 236796, "epoch": 2660, "lr": 3.981921003540767e-06}
{"train_loss": 0.04001595452427864, "global_step": 236797, "epoch": 2660, "lr": 3.981694289917237e-06}
{"train_loss": 0.0617893747985363, "global_step": 236798, "epoch": 2660, "lr": 3.981467582480292e-06}
{"train_loss": 0.08515353500843048, "global_step": 236799, "epoch": 2660, "lr": 3.981240881229992e-06}
{"train_loss": 0.04615689069032669, "global_step": 236800, "epoch": 2660, "lr": 3.981014186166343e-06}
{"train_loss": 0.051098696887493134, "global_step": 236801, "epoch": 2660, "lr": 3.9807874972893955e-06}
{"train_loss": 0.028636876493692398, "global_step": 236802, "epoch": 2660, "lr": 3.980560814599161e-06}
{"train_loss": 0.04470430687069893, "global_step": 236803, "epoch": 2660, "lr": 3.980334138095687e-06}
{"train_loss": 0.06424890458583832, "global_step": 236804, "epoch": 2660, "lr": 3.980107467778988e-06}
{"train_loss": 0.03547661751508713, "global_step": 236805, "epoch": 2660, "lr": 3.979880803649111e-06}
{"train_loss": 0.046421442180871964, "global_step": 236806, "epoch": 2660, "lr": 3.979654145706074e-06}
{"train_loss": 0.09932832419872284, "global_step": 236807, "epoch": 2660, "lr": 3.979427493949916e-06}
{"train_loss": 0.06630939990282059, "global_step": 236808, "epoch": 2660, "lr": 3.9792008483806546e-06}
{"train_loss": 0.0567694716155529, "global_step": 236809, "epoch": 2660, "lr": 3.9789742089983376e-06}
{"train_loss": 0.06957293301820755, "global_step": 236810, "epoch": 2660, "lr": 3.978747575802977e-06}
{"train_loss": 0.0561053641140461, "global_step": 236811, "epoch": 2660, "lr": 3.978520948794623e-06}
{"train_loss": 0.034791577607393265, "global_step": 236812, "epoch": 2660, "lr": 3.9782943279732865e-06}
{"train_loss": 0.02860196679830551, "global_step": 236813, "epoch": 2660, "lr": 3.978067713339012e-06}
{"train_loss": 0.05013962462544441, "global_step": 236814, "epoch": 2660, "lr": 3.977841104891822e-06}
{"train_loss": 0.04076904430985451, "global_step": 236815, "epoch": 2660, "lr": 3.977614502631749e-06}
{"train_loss": 0.01639501564204693, "global_step": 236816, "epoch": 2660, "lr": 3.977387906558833e-06}
{"train_loss": 0.06110769137740135, "global_step": 236817, "epoch": 2660, "lr": 3.977161316673089e-06}
{"train_loss": 0.056489475071430206, "global_step": 236818, "epoch": 2660, "lr": 3.976934732974563e-06}
{"train_loss": 0.043865710496902466, "global_step": 236819, "epoch": 2660, "lr": 3.9767081554632716e-06}
{"train_loss": 0.03384830057621002, "global_step": 236820, "epoch": 2660, "lr": 3.976481584139252e-06}
{"train_loss": 0.07774602621793747, "global_step": 236821, "epoch": 2660, "lr": 3.976255019002528e-06}
{"train_loss": 0.008038737811148167, "global_step": 236822, "epoch": 2660, "lr": 3.97602846005315e-06}
{"train_loss": 0.04287921264767647, "global_step": 236823, "epoch": 2660, "lr": 3.9758019072911154e-06}
{"train_loss": 0.03339356556534767, "global_step": 236824, "epoch": 2660, "lr": 3.975575360716488e-06}
{"train_loss": 0.04991008713841438, "global_step": 236825, "epoch": 2660, "lr": 3.975348820329278e-06}
{"train_loss": 0.06073684245347977, "global_step": 236826, "epoch": 2660, "lr": 3.975122286129523e-06}
{"train_loss": 0.040661800652742386, "global_step": 236827, "epoch": 2660, "lr": 3.974895758117253e-06}
{"train_loss": 0.04872284158855007, "global_step": 236828, "epoch": 2660, "lr": 3.974669236292494e-06, "val_loss": 9.364997863769531, "train_action_mse_error": 19.967891693115234}
{"train_loss": 0.04918732866644859, "global_step": 236829, "epoch": 2661, "lr": 3.97444272065528e-06}
{"train_loss": 0.035386450588703156, "global_step": 236830, "epoch": 2661, "lr": 3.974216211205639e-06}
{"train_loss": 0.016456229612231255, "global_step": 236831, "epoch": 2661, "lr": 3.9739897079436095e-06}
{"train_loss": 0.04081445559859276, "global_step": 236832, "epoch": 2661, "lr": 3.9737632108692135e-06}
{"train_loss": 0.0629824548959732, "global_step": 236833, "epoch": 2661, "lr": 3.973536719982485e-06}
{"train_loss": 0.05156310275197029, "global_step": 236834, "epoch": 2661, "lr": 3.973310235283451e-06}
{"train_loss": 0.049035001546144485, "global_step": 236835, "epoch": 2661, "lr": 3.973083756772144e-06}
{"train_loss": 0.040412355214357376, "global_step": 236836, "epoch": 2661, "lr": 3.972857284448595e-06}
{"train_loss": 0.05759841948747635, "global_step": 236837, "epoch": 2661, "lr": 3.972630818312839e-06}
{"train_loss": 0.051565345376729965, "global_step": 236838, "epoch": 2661, "lr": 3.972404358364895e-06}
{"train_loss": 0.027758851647377014, "global_step": 236839, "epoch": 2661, "lr": 3.972177904604807e-06}
{"train_loss": 0.07561641186475754, "global_step": 236840, "epoch": 2661, "lr": 3.971951457032591e-06}
{"train_loss": 0.08605673909187317, "global_step": 236841, "epoch": 2661, "lr": 3.971725015648298e-06}
{"train_loss": 0.03882089629769325, "global_step": 236842, "epoch": 2661, "lr": 3.971498580451932e-06}
{"train_loss": 0.057688504457473755, "global_step": 236843, "epoch": 2661, "lr": 3.9712721514435444e-06}
{"train_loss": 0.05389817804098129, "global_step": 236844, "epoch": 2661, "lr": 3.971045728623152e-06}
{"train_loss": 0.0317283570766449, "global_step": 236845, "epoch": 2661, "lr": 3.970819311990798e-06}
{"train_loss": 0.02984917163848877, "global_step": 236846, "epoch": 2661, "lr": 3.970592901546499e-06}
{"train_loss": 0.035971127450466156, "global_step": 236847, "epoch": 2661, "lr": 3.970366497290301e-06}
{"train_loss": 0.08387293666601181, "global_step": 236848, "epoch": 2661, "lr": 3.970140099222219e-06}
{"train_loss": 0.00912158191204071, "global_step": 236849, "epoch": 2661, "lr": 3.9699137073422985e-06}
{"train_loss": 0.047922197729349136, "global_step": 236850, "epoch": 2661, "lr": 3.9696873216505555e-06}
{"train_loss": 0.051542073488235474, "global_step": 236851, "epoch": 2661, "lr": 3.9694609421470244e-06}
{"train_loss": 0.03672054782509804, "global_step": 236852, "epoch": 2661, "lr": 3.9692345688317475e-06}
{"train_loss": 0.01509235892444849, "global_step": 236853, "epoch": 2661, "lr": 3.969008201704738e-06}
{"train_loss": 0.033918000757694244, "global_step": 236854, "epoch": 2661, "lr": 3.968781840766045e-06}
{"train_loss": 0.040473200380802155, "global_step": 236855, "epoch": 2661, "lr": 3.968555486015674e-06}
{"train_loss": 0.09015540033578873, "global_step": 236856, "epoch": 2661, "lr": 3.96832913745368e-06}
{"train_loss": 0.09961805492639542, "global_step": 236857, "epoch": 2661, "lr": 3.968102795080075e-06}
{"train_loss": 0.06442226469516754, "global_step": 236858, "epoch": 2661, "lr": 3.9678764588949095e-06}
{"train_loss": 0.03851155936717987, "global_step": 236859, "epoch": 2661, "lr": 3.9676501288981925e-06}
{"train_loss": 0.053244419395923615, "global_step": 236860, "epoch": 2661, "lr": 3.96742380508997e-06}
{"train_loss": 0.04731695353984833, "global_step": 236861, "epoch": 2661, "lr": 3.967197487470259e-06}
{"train_loss": 0.04424944147467613, "global_step": 236862, "epoch": 2661, "lr": 3.966971176039102e-06}
{"train_loss": 0.04870982468128204, "global_step": 236863, "epoch": 2661, "lr": 3.9667448707965235e-06}
{"train_loss": 0.10379057377576828, "global_step": 236864, "epoch": 2661, "lr": 3.9665185717425554e-06}
{"train_loss": 0.027661990374326706, "global_step": 236865, "epoch": 2661, "lr": 3.966292278877232e-06}
{"train_loss": 0.07699557393789291, "global_step": 236866, "epoch": 2661, "lr": 3.9660659922005735e-06}
{"train_loss": 0.04825989902019501, "global_step": 236867, "epoch": 2661, "lr": 3.9658397117126215e-06}
{"train_loss": 0.06071281433105469, "global_step": 236868, "epoch": 2661, "lr": 3.965613437413395e-06}
{"train_loss": 0.061760734766721725, "global_step": 236869, "epoch": 2661, "lr": 3.965387169302937e-06}
{"train_loss": 0.022036155685782433, "global_step": 236870, "epoch": 2661, "lr": 3.965160907381266e-06}
{"train_loss": 0.04523176699876785, "global_step": 236871, "epoch": 2661, "lr": 3.964934651648422e-06}
{"train_loss": 0.05744769424200058, "global_step": 236872, "epoch": 2661, "lr": 3.964708402104422e-06}
{"train_loss": 0.045574307441711426, "global_step": 236873, "epoch": 2661, "lr": 3.9644821587493215e-06}
{"train_loss": 0.02125493250787258, "global_step": 236874, "epoch": 2661, "lr": 3.964255921583121e-06}
{"train_loss": 0.01709786243736744, "global_step": 236875, "epoch": 2661, "lr": 3.964029690605875e-06}
{"train_loss": 0.029596256092190742, "global_step": 236876, "epoch": 2661, "lr": 3.9638034658176005e-06}
{"train_loss": 0.07748398929834366, "global_step": 236877, "epoch": 2661, "lr": 3.963577247218331e-06}
{"train_loss": 0.06730734556913376, "global_step": 236878, "epoch": 2661, "lr": 3.963351034808094e-06}
{"train_loss": 0.08098280429840088, "global_step": 236879, "epoch": 2661, "lr": 3.963124828586934e-06}
{"train_loss": 0.03393357992172241, "global_step": 236880, "epoch": 2661, "lr": 3.962898628554862e-06}
{"train_loss": 0.08557089418172836, "global_step": 236881, "epoch": 2661, "lr": 3.962672434711923e-06}
{"train_loss": 0.10852311551570892, "global_step": 236882, "epoch": 2661, "lr": 3.962446247058133e-06}
{"train_loss": 0.010017601773142815, "global_step": 236883, "epoch": 2661, "lr": 3.962220065593542e-06}
{"train_loss": 0.0799163207411766, "global_step": 236884, "epoch": 2661, "lr": 3.961993890318155e-06}
{"train_loss": 0.034100256860256195, "global_step": 236885, "epoch": 2661, "lr": 3.9617677212320295e-06}
{"train_loss": 0.06085245683789253, "global_step": 236886, "epoch": 2661, "lr": 3.961541558335175e-06}
{"train_loss": 0.0782831683754921, "global_step": 236887, "epoch": 2661, "lr": 3.96131540162763e-06}
{"train_loss": 0.05295327305793762, "global_step": 236888, "epoch": 2661, "lr": 3.961089251109434e-06}
{"train_loss": 0.07275326550006866, "global_step": 236889, "epoch": 2661, "lr": 3.960863106780599e-06}
{"train_loss": 0.03241771087050438, "global_step": 236890, "epoch": 2661, "lr": 3.9606369686411735e-06}
{"train_loss": 0.03749339282512665, "global_step": 236891, "epoch": 2661, "lr": 3.96041083669117e-06}
{"train_loss": 0.05062104016542435, "global_step": 236892, "epoch": 2661, "lr": 3.960184710930636e-06}
{"train_loss": 0.034175798296928406, "global_step": 236893, "epoch": 2661, "lr": 3.959958591359586e-06}
{"train_loss": 0.06372294574975967, "global_step": 236894, "epoch": 2661, "lr": 3.959732477978068e-06}
{"train_loss": 0.035197142511606216, "global_step": 236895, "epoch": 2661, "lr": 3.959506370786092e-06}
{"train_loss": 0.02484661154448986, "global_step": 236896, "epoch": 2661, "lr": 3.95928026978371e-06}
{"train_loss": 0.02754046954214573, "global_step": 236897, "epoch": 2661, "lr": 3.9590541749709335e-06}
{"train_loss": 0.02664533257484436, "global_step": 236898, "epoch": 2661, "lr": 3.95882808634781e-06}
{"train_loss": 0.058439578860998154, "global_step": 236899, "epoch": 2661, "lr": 3.958602003914352e-06}
{"train_loss": 0.031973738223314285, "global_step": 236900, "epoch": 2661, "lr": 3.958375927670604e-06}
{"train_loss": 0.052505653351545334, "global_step": 236901, "epoch": 2661, "lr": 3.9581498576165885e-06}
{"train_loss": 0.07380128651857376, "global_step": 236902, "epoch": 2661, "lr": 3.957923793752344e-06}
{"train_loss": 0.0313042476773262, "global_step": 236903, "epoch": 2661, "lr": 3.957697736077892e-06}
{"train_loss": 0.017941271886229515, "global_step": 236904, "epoch": 2661, "lr": 3.957471684593261e-06}
{"train_loss": 0.08192741870880127, "global_step": 236905, "epoch": 2661, "lr": 3.957245639298496e-06}
{"train_loss": 0.08950196951627731, "global_step": 236906, "epoch": 2661, "lr": 3.957019600193607e-06}
{"train_loss": 0.046667903661727905, "global_step": 236907, "epoch": 2661, "lr": 3.956793567278649e-06}
{"train_loss": 0.07123032957315445, "global_step": 236908, "epoch": 2661, "lr": 3.956567540553624e-06}
{"train_loss": 0.027763614431023598, "global_step": 236909, "epoch": 2661, "lr": 3.956341520018592e-06}
{"train_loss": 0.06647863984107971, "global_step": 236910, "epoch": 2661, "lr": 3.956115505673558e-06}
{"train_loss": 0.0573556125164032, "global_step": 236911, "epoch": 2661, "lr": 3.955889497518572e-06}
{"train_loss": 0.024786202237010002, "global_step": 236912, "epoch": 2661, "lr": 3.9556634955536406e-06}
{"train_loss": 0.032697893679142, "global_step": 236913, "epoch": 2661, "lr": 3.955437499778825e-06}
{"train_loss": 0.03469526022672653, "global_step": 236914, "epoch": 2661, "lr": 3.955211510194129e-06}
{"train_loss": 0.07554656267166138, "global_step": 236915, "epoch": 2661, "lr": 3.954985526799604e-06}
{"train_loss": 0.0405285507440567, "global_step": 236916, "epoch": 2661, "lr": 3.95475954959526e-06}
{"train_loss": 0.04974369236969211, "global_step": 236917, "epoch": 2661, "lr": 3.954533578581143e-06, "val_loss": 9.269462585449219}
{"train_loss": 0.037087708711624146, "global_step": 236918, "epoch": 2662, "lr": 3.9543076137572745e-06}
{"train_loss": 0.05560413748025894, "global_step": 236919, "epoch": 2662, "lr": 3.954081655123692e-06}
{"train_loss": 0.05512332543730736, "global_step": 236920, "epoch": 2662, "lr": 3.953855702680415e-06}
{"train_loss": 0.06875711679458618, "global_step": 236921, "epoch": 2662, "lr": 3.95362975642749e-06}
{"train_loss": 0.07847322523593903, "global_step": 236922, "epoch": 2662, "lr": 3.953403816364931e-06}
{"train_loss": 0.05391599237918854, "global_step": 236923, "epoch": 2662, "lr": 3.9531778824927704e-06}
{"train_loss": 0.05311577022075653, "global_step": 236924, "epoch": 2662, "lr": 3.952951954811057e-06}
{"train_loss": 0.0223955437541008, "global_step": 236925, "epoch": 2662, "lr": 3.952726033319798e-06}
{"train_loss": 0.04106194153428078, "global_step": 236926, "epoch": 2662, "lr": 3.952500118019043e-06}
{"train_loss": 0.06474923342466354, "global_step": 236927, "epoch": 2662, "lr": 3.952274208908807e-06}
{"train_loss": 0.03629551827907562, "global_step": 236928, "epoch": 2662, "lr": 3.952048305989131e-06}
{"train_loss": 0.05353303998708725, "global_step": 236929, "epoch": 2662, "lr": 3.951822409260031e-06}
{"train_loss": 0.04679947718977928, "global_step": 236930, "epoch": 2662, "lr": 3.951596518721562e-06}
{"train_loss": 0.041544701904058456, "global_step": 236931, "epoch": 2662, "lr": 3.9513706343737245e-06}
{"train_loss": 0.08083964884281158, "global_step": 236932, "epoch": 2662, "lr": 3.951144756216574e-06}
{"train_loss": 0.06285679340362549, "global_step": 236933, "epoch": 2662, "lr": 3.950918884250126e-06}
{"train_loss": 0.0969884842634201, "global_step": 236934, "epoch": 2662, "lr": 3.9506930184744215e-06}
{"train_loss": 0.02677021361887455, "global_step": 236935, "epoch": 2662, "lr": 3.950467158889476e-06}
{"train_loss": 0.09651902318000793, "global_step": 236936, "epoch": 2662, "lr": 3.950241305495339e-06}
{"train_loss": 0.03372170403599739, "global_step": 236937, "epoch": 2662, "lr": 3.9500154582920225e-06}
{"train_loss": 0.02597636729478836, "global_step": 236938, "epoch": 2662, "lr": 3.949789617279576e-06}
{"train_loss": 0.05138207599520683, "global_step": 236939, "epoch": 2662, "lr": 3.949563782458005e-06}
{"train_loss": 0.014137367717921734, "global_step": 236940, "epoch": 2662, "lr": 3.949337953827364e-06}
{"train_loss": 0.05516071245074272, "global_step": 236941, "epoch": 2662, "lr": 3.949112131387667e-06}
{"train_loss": 0.05077771469950676, "global_step": 236942, "epoch": 2662, "lr": 3.948886315138956e-06}
{"train_loss": 0.03374176472425461, "global_step": 236943, "epoch": 2662, "lr": 3.948660505081259e-06}
{"train_loss": 0.06662124395370483, "global_step": 236944, "epoch": 2662, "lr": 3.948434701214593e-06}
{"train_loss": 0.04003644362092018, "global_step": 236945, "epoch": 2662, "lr": 3.948208903539008e-06}
{"train_loss": 0.05151592567563057, "global_step": 236946, "epoch": 2662, "lr": 3.947983112054515e-06}
{"train_loss": 0.0771532878279686, "global_step": 236947, "epoch": 2662, "lr": 3.947757326761164e-06}
{"train_loss": 0.02837659977376461, "global_step": 236948, "epoch": 2662, "lr": 3.947531547658967e-06}
{"train_loss": 0.06577353924512863, "global_step": 236949, "epoch": 2662, "lr": 3.947305774747973e-06}
{"train_loss": 0.02533630095422268, "global_step": 236950, "epoch": 2662, "lr": 3.947080008028192e-06}
{"train_loss": 0.020714858546853065, "global_step": 236951, "epoch": 2662, "lr": 3.946854247499671e-06}
{"train_loss": 0.0344957672059536, "global_step": 236952, "epoch": 2662, "lr": 3.94662849316243e-06}
{"train_loss": 0.046986792236566544, "global_step": 236953, "epoch": 2662, "lr": 3.946402745016509e-06}
{"train_loss": 0.02512657456099987, "global_step": 236954, "epoch": 2662, "lr": 3.946177003061929e-06}
{"train_loss": 0.04235789552330971, "global_step": 236955, "epoch": 2662, "lr": 3.945951267298731e-06}
{"train_loss": 0.027848701924085617, "global_step": 236956, "epoch": 2662, "lr": 3.9457255377269245e-06}
{"train_loss": 0.025596728548407555, "global_step": 236957, "epoch": 2662, "lr": 3.9454998143465704e-06}
{"train_loss": 0.09312868863344193, "global_step": 236958, "epoch": 2662, "lr": 3.94527409715767e-06}
{"train_loss": 0.04081013426184654, "global_step": 236959, "epoch": 2662, "lr": 3.945048386160266e-06}
{"train_loss": 0.031508903950452805, "global_step": 236960, "epoch": 2662, "lr": 3.944822681354393e-06}
{"train_loss": 0.04900062829256058, "global_step": 236961, "epoch": 2662, "lr": 3.944596982740079e-06}
{"train_loss": 0.03415967524051666, "global_step": 236962, "epoch": 2662, "lr": 3.94437129031735e-06}
{"train_loss": 0.05225870758295059, "global_step": 236963, "epoch": 2662, "lr": 3.944145604086241e-06}
{"train_loss": 0.028942083939909935, "global_step": 236964, "epoch": 2662, "lr": 3.943919924046779e-06}
{"train_loss": 0.0492449514567852, "global_step": 236965, "epoch": 2662, "lr": 3.943694250198993e-06}
{"train_loss": 0.06195845454931259, "global_step": 236966, "epoch": 2662, "lr": 3.9434685825429254e-06}
{"train_loss": 0.05473862215876579, "global_step": 236967, "epoch": 2662, "lr": 3.9432429210785885e-06}
{"train_loss": 0.06008940935134888, "global_step": 236968, "epoch": 2662, "lr": 3.943017265806026e-06}
{"train_loss": 0.029923342168331146, "global_step": 236969, "epoch": 2662, "lr": 3.942791616725256e-06}
{"train_loss": 0.011280790902674198, "global_step": 236970, "epoch": 2662, "lr": 3.942565973836326e-06}
{"train_loss": 0.04237896203994751, "global_step": 236971, "epoch": 2662, "lr": 3.94234033713925e-06}
{"train_loss": 0.057974036782979965, "global_step": 236972, "epoch": 2662, "lr": 3.94211470663407e-06}
{"train_loss": 0.03616918995976448, "global_step": 236973, "epoch": 2662, "lr": 3.941889082320805e-06}
{"train_loss": 0.03870537132024765, "global_step": 236974, "epoch": 2662, "lr": 3.941663464199497e-06}
{"train_loss": 0.0279763862490654, "global_step": 236975, "epoch": 2662, "lr": 3.9414378522701634e-06}
{"train_loss": 0.041835226118564606, "global_step": 236976, "epoch": 2662, "lr": 3.941212246532855e-06}
{"train_loss": 0.06847142428159714, "global_step": 236977, "epoch": 2662, "lr": 3.940986646987577e-06}
{"train_loss": 0.029939189553260803, "global_step": 236978, "epoch": 2662, "lr": 3.940761053634378e-06}
{"train_loss": 0.04475425183773041, "global_step": 236979, "epoch": 2662, "lr": 3.9405354664732775e-06}
{"train_loss": 0.0193087887018919, "global_step": 236980, "epoch": 2662, "lr": 3.940309885504318e-06}
{"train_loss": 0.06740572303533554, "global_step": 236981, "epoch": 2662, "lr": 3.9400843107275266e-06}
{"train_loss": 0.032968152314424515, "global_step": 236982, "epoch": 2662, "lr": 3.9398587421429154e-06}
{"train_loss": 0.026297910138964653, "global_step": 236983, "epoch": 2662, "lr": 3.93963317975054e-06}
{"train_loss": 0.0381508469581604, "global_step": 236984, "epoch": 2662, "lr": 3.939407623550406e-06}
{"train_loss": 0.04459403455257416, "global_step": 236985, "epoch": 2662, "lr": 3.9391820735425746e-06}
{"train_loss": 0.016689710319042206, "global_step": 236986, "epoch": 2662, "lr": 3.938956529727045e-06}
{"train_loss": 0.018453437834978104, "global_step": 236987, "epoch": 2662, "lr": 3.938730992103867e-06}
{"train_loss": 0.02566712722182274, "global_step": 236988, "epoch": 2662, "lr": 3.938505460673065e-06}
{"train_loss": 0.06809257715940475, "global_step": 236989, "epoch": 2662, "lr": 3.938279935434669e-06}
{"train_loss": 0.06670350581407547, "global_step": 236990, "epoch": 2662, "lr": 3.938054416388709e-06}
{"train_loss": 0.09390988200902939, "global_step": 236991, "epoch": 2662, "lr": 3.937828903535218e-06}
{"train_loss": 0.07657293230295181, "global_step": 236992, "epoch": 2662, "lr": 3.937603396874218e-06}
{"train_loss": 0.0354669913649559, "global_step": 236993, "epoch": 2662, "lr": 3.9373778964057585e-06}
{"train_loss": 0.030047643929719925, "global_step": 236994, "epoch": 2662, "lr": 3.937152402129845e-06}
{"train_loss": 0.03301694244146347, "global_step": 236995, "epoch": 2662, "lr": 3.93692691404653e-06}
{"train_loss": 0.03328270837664604, "global_step": 236996, "epoch": 2662, "lr": 3.93670143215582e-06}
{"train_loss": 0.07382781058549881, "global_step": 236997, "epoch": 2662, "lr": 3.936475956457764e-06}
{"train_loss": 0.03659849986433983, "global_step": 236998, "epoch": 2662, "lr": 3.936250486952397e-06}
{"train_loss": 0.03643171489238739, "global_step": 236999, "epoch": 2662, "lr": 3.9360250236397275e-06}
{"train_loss": 0.020860768854618073, "global_step": 237000, "epoch": 2662, "lr": 3.93579956651981e-06}
{"train_loss": 0.1005527451634407, "global_step": 237001, "epoch": 2662, "lr": 3.93557411559265e-06}
{"train_loss": 0.0845613181591034, "global_step": 237002, "epoch": 2662, "lr": 3.935348670858302e-06}
{"train_loss": 0.07515261322259903, "global_step": 237003, "epoch": 2662, "lr": 3.935123232316779e-06}
{"train_loss": 0.029431259259581566, "global_step": 237004, "epoch": 2662, "lr": 3.9348977999681235e-06}
{"train_loss": 0.0368715338408947, "global_step": 237005, "epoch": 2662, "lr": 3.934672373812348e-06}
{"train_loss": 0.04657851193058357, "global_step": 237006, "epoch": 2662, "lr": 3.9344469538495075e-06, "val_loss": 9.197510719299316}
{"train_loss": 0.08185632526874542, "global_step": 237007, "epoch": 2663, "lr": 3.934221540079608e-06}
{"train_loss": 0.04416649416089058, "global_step": 237008, "epoch": 2663, "lr": 3.933996132502699e-06}
{"train_loss": 0.041403647512197495, "global_step": 237009, "epoch": 2663, "lr": 3.9337707311187965e-06}
{"train_loss": 0.08177490532398224, "global_step": 237010, "epoch": 2663, "lr": 3.93354533592794e-06}
{"train_loss": 0.037983380258083344, "global_step": 237011, "epoch": 2663, "lr": 3.9333199469301525e-06}
{"train_loss": 0.020542366430163383, "global_step": 237012, "epoch": 2663, "lr": 3.933094564125478e-06}
{"train_loss": 0.04768081381917, "global_step": 237013, "epoch": 2663, "lr": 3.932869187513932e-06}
{"train_loss": 0.07356657832860947, "global_step": 237014, "epoch": 2663, "lr": 3.932643817095555e-06}
{"train_loss": 0.050022803246974945, "global_step": 237015, "epoch": 2663, "lr": 3.932418452870362e-06}
{"train_loss": 0.024333642795681953, "global_step": 237016, "epoch": 2663, "lr": 3.932193094838404e-06}
{"train_loss": 0.02426757849752903, "global_step": 237017, "epoch": 2663, "lr": 3.931967742999693e-06}
{"train_loss": 0.05044608190655708, "global_step": 237018, "epoch": 2663, "lr": 3.931742397354277e-06}
{"train_loss": 0.050540924072265625, "global_step": 237019, "epoch": 2663, "lr": 3.931517057902173e-06}
{"train_loss": 0.05007198080420494, "global_step": 237020, "epoch": 2663, "lr": 3.93129172464341e-06}
{"train_loss": 0.041267234832048416, "global_step": 237021, "epoch": 2663, "lr": 3.931066397578032e-06}
{"train_loss": 0.03029565140604973, "global_step": 237022, "epoch": 2663, "lr": 3.930841076706049e-06}
{"train_loss": 0.052546177059412, "global_step": 237023, "epoch": 2663, "lr": 3.930615762027518e-06}
{"train_loss": 0.01944630965590477, "global_step": 237024, "epoch": 2663, "lr": 3.930390453542437e-06}
{"train_loss": 0.017692305147647858, "global_step": 237025, "epoch": 2663, "lr": 3.9301651512508695e-06}
{"train_loss": 0.02760966308414936, "global_step": 237026, "epoch": 2663, "lr": 3.929939855152815e-06}
{"train_loss": 0.04463828355073929, "global_step": 237027, "epoch": 2663, "lr": 3.929714565248332e-06}
{"train_loss": 0.07969177514314651, "global_step": 237028, "epoch": 2663, "lr": 3.9294892815374295e-06}
{"train_loss": 0.06240881606936455, "global_step": 237029, "epoch": 2663, "lr": 3.92926400402015e-06}
{"train_loss": 0.10722609609365463, "global_step": 237030, "epoch": 2663, "lr": 3.92903873269651e-06}
{"train_loss": 0.05325700715184212, "global_step": 237031, "epoch": 2663, "lr": 3.928813467566561e-06}
{"train_loss": 0.01649506948888302, "global_step": 237032, "epoch": 2663, "lr": 3.928588208630312e-06}
{"train_loss": 0.043774936348199844, "global_step": 237033, "epoch": 2663, "lr": 3.928362955887804e-06}
{"train_loss": 0.026628030464053154, "global_step": 237034, "epoch": 2663, "lr": 3.928137709339075e-06}
{"train_loss": 0.026316436007618904, "global_step": 237035, "epoch": 2663, "lr": 3.9279124689841354e-06}
{"train_loss": 0.05006268247961998, "global_step": 237036, "epoch": 2663, "lr": 3.927687234823035e-06}
{"train_loss": 0.022524410858750343, "global_step": 237037, "epoch": 2663, "lr": 3.927462006855787e-06}
{"train_loss": 0.02807699143886566, "global_step": 237038, "epoch": 2663, "lr": 3.9272367850824385e-06}
{"train_loss": 0.06811491400003433, "global_step": 237039, "epoch": 2663, "lr": 3.927011569503009e-06}
{"train_loss": 0.02434544637799263, "global_step": 237040, "epoch": 2663, "lr": 3.926786360117535e-06}
{"train_loss": 0.08314087241888046, "global_step": 237041, "epoch": 2663, "lr": 3.926561156926034e-06}
{"train_loss": 0.0457865446805954, "global_step": 237042, "epoch": 2663, "lr": 3.926335959928551e-06}
{"train_loss": 0.04943729564547539, "global_step": 237043, "epoch": 2663, "lr": 3.926110769125108e-06}
{"train_loss": 0.029569974169135094, "global_step": 237044, "epoch": 2663, "lr": 3.9258855845157425e-06}
{"train_loss": 0.062169138342142105, "global_step": 237045, "epoch": 2663, "lr": 3.9256604061004735e-06}
{"train_loss": 0.04063718765974045, "global_step": 237046, "epoch": 2663, "lr": 3.925435233879343e-06}
{"train_loss": 0.03139253333210945, "global_step": 237047, "epoch": 2663, "lr": 3.92521006785237e-06}
{"train_loss": 0.020373448729515076, "global_step": 237048, "epoch": 2663, "lr": 3.9249849080196025e-06}
{"train_loss": 0.04987568408250809, "global_step": 237049, "epoch": 2663, "lr": 3.924759754381047e-06}
{"train_loss": 0.05138049274682999, "global_step": 237050, "epoch": 2663, "lr": 3.924534606936747e-06}
{"train_loss": 0.07041013985872269, "global_step": 237051, "epoch": 2663, "lr": 3.924309465686743e-06}
{"train_loss": 0.0643535777926445, "global_step": 237052, "epoch": 2663, "lr": 3.9240843306310395e-06}
{"train_loss": 0.031206099316477776, "global_step": 237053, "epoch": 2663, "lr": 3.923859201769692e-06}
{"train_loss": 0.08293937146663666, "global_step": 237054, "epoch": 2663, "lr": 3.923634079102723e-06}
{"train_loss": 0.040700167417526245, "global_step": 237055, "epoch": 2663, "lr": 3.923408962630148e-06}
{"train_loss": 0.048160992562770844, "global_step": 237056, "epoch": 2663, "lr": 3.9231838523520195e-06}
{"train_loss": 0.05906561389565468, "global_step": 237057, "epoch": 2663, "lr": 3.922958748268346e-06}
{"train_loss": 0.07312033325433731, "global_step": 237058, "epoch": 2663, "lr": 3.922733650379185e-06}
{"train_loss": 0.05267973244190216, "global_step": 237059, "epoch": 2663, "lr": 3.92250855868454e-06}
{"train_loss": 0.062222085893154144, "global_step": 237060, "epoch": 2663, "lr": 3.922283473184452e-06}
{"train_loss": 0.035054102540016174, "global_step": 237061, "epoch": 2663, "lr": 3.922058393878958e-06}
{"train_loss": 0.03586119785904884, "global_step": 237062, "epoch": 2663, "lr": 3.92183332076807e-06}
{"train_loss": 0.044346123933792114, "global_step": 237063, "epoch": 2663, "lr": 3.921608253851844e-06}
{"train_loss": 0.0754520520567894, "global_step": 237064, "epoch": 2663, "lr": 3.9213831931302844e-06}
{"train_loss": 0.017266834154725075, "global_step": 237065, "epoch": 2663, "lr": 3.921158138603437e-06}
{"train_loss": 0.03330489248037338, "global_step": 237066, "epoch": 2663, "lr": 3.920933090271328e-06}
{"train_loss": 0.07715532928705215, "global_step": 237067, "epoch": 2663, "lr": 3.920708048133992e-06}
{"train_loss": 0.05934092402458191, "global_step": 237068, "epoch": 2663, "lr": 3.9204830121914445e-06}
{"train_loss": 0.09436247497797012, "global_step": 237069, "epoch": 2663, "lr": 3.9202579824437316e-06}
{"train_loss": 0.014798179268836975, "global_step": 237070, "epoch": 2663, "lr": 3.920032958890885e-06}
{"train_loss": 0.03515736758708954, "global_step": 237071, "epoch": 2663, "lr": 3.919807941532921e-06}
{"train_loss": 0.05071616172790527, "global_step": 237072, "epoch": 2663, "lr": 3.919582930369881e-06}
{"train_loss": 0.03610896319150925, "global_step": 237073, "epoch": 2663, "lr": 3.91935792540179e-06}
{"train_loss": 0.033736035227775574, "global_step": 237074, "epoch": 2663, "lr": 3.919132926628683e-06}
{"train_loss": 0.0640002191066742, "global_step": 237075, "epoch": 2663, "lr": 3.918907934050581e-06}
{"train_loss": 0.013878444209694862, "global_step": 237076, "epoch": 2663, "lr": 3.918682947667523e-06}
{"train_loss": 0.05910349264740944, "global_step": 237077, "epoch": 2663, "lr": 3.918457967479533e-06}
{"train_loss": 0.043107934296131134, "global_step": 237078, "epoch": 2663, "lr": 3.918232993486653e-06}
{"train_loss": 0.04166676104068756, "global_step": 237079, "epoch": 2663, "lr": 3.918008025688896e-06}
{"train_loss": 0.023980580270290375, "global_step": 237080, "epoch": 2663, "lr": 3.917783064086311e-06}
{"train_loss": 0.05847444757819176, "global_step": 237081, "epoch": 2663, "lr": 3.917558108678909e-06}
{"train_loss": 0.045837804675102234, "global_step": 237082, "epoch": 2663, "lr": 3.917333159466735e-06}
{"train_loss": 0.029477223753929138, "global_step": 237083, "epoch": 2663, "lr": 3.91710821644981e-06}
{"train_loss": 0.022392572835087776, "global_step": 237084, "epoch": 2663, "lr": 3.9168832796281755e-06}
{"train_loss": 0.0185349490493536, "global_step": 237085, "epoch": 2663, "lr": 3.9166583490018445e-06}
{"train_loss": 0.036476366221904755, "global_step": 237086, "epoch": 2663, "lr": 3.916433424570859e-06}
{"train_loss": 0.057587433606386185, "global_step": 237087, "epoch": 2663, "lr": 3.916208506335251e-06}
{"train_loss": 0.031709108501672745, "global_step": 237088, "epoch": 2663, "lr": 3.915983594295042e-06}
{"train_loss": 0.05428064987063408, "global_step": 237089, "epoch": 2663, "lr": 3.9157586884502775e-06}
{"train_loss": 0.02832735702395439, "global_step": 237090, "epoch": 2663, "lr": 3.915533788800968e-06}
{"train_loss": 0.06497680395841599, "global_step": 237091, "epoch": 2663, "lr": 3.915308895347158e-06}
{"train_loss": 0.039456557482481, "global_step": 237092, "epoch": 2663, "lr": 3.915084008088882e-06}
{"train_loss": 0.04300858825445175, "global_step": 237093, "epoch": 2663, "lr": 3.914859127026144e-06}
{"train_loss": 0.014463759958744049, "global_step": 237094, "epoch": 2663, "lr": 3.914634252159005e-06}
{"train_loss": 0.04563724412844422, "global_step": 237095, "epoch": 2663, "lr": 3.914409383487472e-06, "val_loss": 9.448904991149902}
{"train_loss": 0.018749777227640152, "global_step": 237096, "epoch": 2664, "lr": 3.914184521011588e-06}
{"train_loss": 0.04914150759577751, "global_step": 237097, "epoch": 2664, "lr": 3.913959664731387e-06}
{"train_loss": 0.07045479118824005, "global_step": 237098, "epoch": 2664, "lr": 3.91373481464688e-06}
{"train_loss": 0.04039128124713898, "global_step": 237099, "epoch": 2664, "lr": 3.913509970758123e-06}
{"train_loss": 0.025790538638830185, "global_step": 237100, "epoch": 2664, "lr": 3.9132851330651195e-06}
{"train_loss": 0.03226310387253761, "global_step": 237101, "epoch": 2664, "lr": 3.913060301567923e-06}
{"train_loss": 0.02419670671224594, "global_step": 237102, "epoch": 2664, "lr": 3.9128354762665464e-06}
{"train_loss": 0.06671272963285446, "global_step": 237103, "epoch": 2664, "lr": 3.912610657161037e-06}
{"train_loss": 0.04705643281340599, "global_step": 237104, "epoch": 2664, "lr": 3.9123858442514035e-06}
{"train_loss": 0.06333048641681671, "global_step": 237105, "epoch": 2664, "lr": 3.912161037537687e-06}
{"train_loss": 0.03362835943698883, "global_step": 237106, "epoch": 2664, "lr": 3.9119362370199305e-06}
{"train_loss": 0.06277349591255188, "global_step": 237107, "epoch": 2664, "lr": 3.911711442698146e-06}
{"train_loss": 0.04042869433760643, "global_step": 237108, "epoch": 2664, "lr": 3.911486654572377e-06}
{"train_loss": 0.016961487010121346, "global_step": 237109, "epoch": 2664, "lr": 3.911261872642635e-06}
{"train_loss": 0.05484176427125931, "global_step": 237110, "epoch": 2664, "lr": 3.911037096908971e-06}
{"train_loss": 0.0470554381608963, "global_step": 237111, "epoch": 2664, "lr": 3.9108123273714e-06}
{"train_loss": 0.04230879992246628, "global_step": 237112, "epoch": 2664, "lr": 3.910587564029966e-06}
{"train_loss": 0.048839349299669266, "global_step": 237113, "epoch": 2664, "lr": 3.910362806884688e-06}
{"train_loss": 0.047346845269203186, "global_step": 237114, "epoch": 2664, "lr": 3.910138055935603e-06}
{"train_loss": 0.044298965483903885, "global_step": 237115, "epoch": 2664, "lr": 3.909913311182728e-06}
{"train_loss": 0.048549480736255646, "global_step": 237116, "epoch": 2664, "lr": 3.909688572626119e-06}
{"train_loss": 0.027390895411372185, "global_step": 237117, "epoch": 2664, "lr": 3.909463840265776e-06}
{"train_loss": 0.026541277766227722, "global_step": 237118, "epoch": 2664, "lr": 3.909239114101754e-06}
{"train_loss": 0.04803527891635895, "global_step": 237119, "epoch": 2664, "lr": 3.909014394134064e-06}
{"train_loss": 0.03665754571557045, "global_step": 237120, "epoch": 2664, "lr": 3.908789680362757e-06}
{"train_loss": 0.03948276489973068, "global_step": 237121, "epoch": 2664, "lr": 3.908564972787843e-06}
{"train_loss": 0.044994622468948364, "global_step": 237122, "epoch": 2664, "lr": 3.908340271409366e-06}
{"train_loss": 0.023867031559348106, "global_step": 237123, "epoch": 2664, "lr": 3.908115576227339e-06}
{"train_loss": 0.07222198694944382, "global_step": 237124, "epoch": 2664, "lr": 3.907890887241811e-06}
{"train_loss": 0.04349113255739212, "global_step": 237125, "epoch": 2664, "lr": 3.907666204452809e-06}
{"train_loss": 0.037707872688770294, "global_step": 237126, "epoch": 2664, "lr": 3.907441527860356e-06}
{"train_loss": 0.0790570005774498, "global_step": 237127, "epoch": 2664, "lr": 3.907216857464491e-06}
{"train_loss": 0.040019042789936066, "global_step": 237128, "epoch": 2664, "lr": 3.90699219326523e-06}
{"train_loss": 0.06240016222000122, "global_step": 237129, "epoch": 2664, "lr": 3.9067675352626235e-06}
{"train_loss": 0.0639127567410469, "global_step": 237130, "epoch": 2664, "lr": 3.9065428834566876e-06}
{"train_loss": 0.019664470106363297, "global_step": 237131, "epoch": 2664, "lr": 3.906318237847445e-06}
{"train_loss": 0.03509124740958214, "global_step": 237132, "epoch": 2664, "lr": 3.906093598434946e-06}
{"train_loss": 0.051800332963466644, "global_step": 237133, "epoch": 2664, "lr": 3.905868965219206e-06}
{"train_loss": 0.03379886597394943, "global_step": 237134, "epoch": 2664, "lr": 3.905644338200265e-06}
{"train_loss": 0.03293207287788391, "global_step": 237135, "epoch": 2664, "lr": 3.9054197173781445e-06}
{"train_loss": 0.04662029445171356, "global_step": 237136, "epoch": 2664, "lr": 3.905195102752879e-06}
{"train_loss": 0.05400066077709198, "global_step": 237137, "epoch": 2664, "lr": 3.904970494324501e-06}
{"train_loss": 0.04266109690070152, "global_step": 237138, "epoch": 2664, "lr": 3.9047458920930314e-06}
{"train_loss": 0.04901200160384178, "global_step": 237139, "epoch": 2664, "lr": 3.904521296058516e-06}
{"train_loss": 0.028095362707972527, "global_step": 237140, "epoch": 2664, "lr": 3.904296706220961e-06}
{"train_loss": 0.024519935250282288, "global_step": 237141, "epoch": 2664, "lr": 3.904072122580421e-06}
{"train_loss": 0.09885572642087936, "global_step": 237142, "epoch": 2664, "lr": 3.9038475451369184e-06}
{"train_loss": 0.03596121445298195, "global_step": 237143, "epoch": 2664, "lr": 3.903622973890475e-06}
{"train_loss": 0.019453350454568863, "global_step": 237144, "epoch": 2664, "lr": 3.903398408841136e-06}
{"train_loss": 0.03584999591112137, "global_step": 237145, "epoch": 2664, "lr": 3.903173849988917e-06}
{"train_loss": 0.04786689952015877, "global_step": 237146, "epoch": 2664, "lr": 3.902949297333863e-06}
{"train_loss": 0.0654706358909607, "global_step": 237147, "epoch": 2664, "lr": 3.9027247508759855e-06}
{"train_loss": 0.018919141963124275, "global_step": 237148, "epoch": 2664, "lr": 3.902500210615334e-06}
{"train_loss": 0.02391979657113552, "global_step": 237149, "epoch": 2664, "lr": 3.90227567655192e-06}
{"train_loss": 0.040521152317523956, "global_step": 237150, "epoch": 2664, "lr": 3.902051148685792e-06}
{"train_loss": 0.06553468853235245, "global_step": 237151, "epoch": 2664, "lr": 3.901826627016958e-06}
{"train_loss": 0.044993381947278976, "global_step": 237152, "epoch": 2664, "lr": 3.901602111545477e-06}
{"train_loss": 0.019703932106494904, "global_step": 237153, "epoch": 2664, "lr": 3.901377602271356e-06}
{"train_loss": 0.021034056320786476, "global_step": 237154, "epoch": 2664, "lr": 3.901153099194637e-06}
{"train_loss": 0.05506400391459465, "global_step": 237155, "epoch": 2664, "lr": 3.90092860231534e-06}
{"train_loss": 0.05597122013568878, "global_step": 237156, "epoch": 2664, "lr": 3.900704111633508e-06}
{"train_loss": 0.050834521651268005, "global_step": 237157, "epoch": 2664, "lr": 3.900479627149156e-06}
{"train_loss": 0.05210500210523605, "global_step": 237158, "epoch": 2664, "lr": 3.9002551488623305e-06}
{"train_loss": 0.032155536115169525, "global_step": 237159, "epoch": 2664, "lr": 3.900030676773048e-06}
{"train_loss": 0.10633404552936554, "global_step": 237160, "epoch": 2664, "lr": 3.89980621088134e-06}
{"train_loss": 0.061110302805900574, "global_step": 237161, "epoch": 2664, "lr": 3.899581751187253e-06}
{"train_loss": 0.018709005787968636, "global_step": 237162, "epoch": 2664, "lr": 3.899357297690798e-06}
{"train_loss": 0.06923706084489822, "global_step": 237163, "epoch": 2664, "lr": 3.899132850392023e-06}
{"train_loss": 0.06104864552617073, "global_step": 237164, "epoch": 2664, "lr": 3.898908409290936e-06}
{"train_loss": 0.033109474927186966, "global_step": 237165, "epoch": 2664, "lr": 3.898683974387585e-06}
{"train_loss": 0.016209885478019714, "global_step": 237166, "epoch": 2664, "lr": 3.898459545681987e-06}
{"train_loss": 0.03895466774702072, "global_step": 237167, "epoch": 2664, "lr": 3.898235123174193e-06}
{"train_loss": 0.039652660489082336, "global_step": 237168, "epoch": 2664, "lr": 3.898010706864203e-06}
{"train_loss": 0.023116614669561386, "global_step": 237169, "epoch": 2664, "lr": 3.897786296752076e-06}
{"train_loss": 0.06815493106842041, "global_step": 237170, "epoch": 2664, "lr": 3.897561892837831e-06}
{"train_loss": 0.04573384299874306, "global_step": 237171, "epoch": 2664, "lr": 3.897337495121483e-06}
{"train_loss": 0.06445275992155075, "global_step": 237172, "epoch": 2664, "lr": 3.897113103603089e-06}
{"train_loss": 0.02266726642847061, "global_step": 237173, "epoch": 2664, "lr": 3.89688871828266e-06}
{"train_loss": 0.060250960290431976, "global_step": 237174, "epoch": 2664, "lr": 3.896664339160239e-06}
{"train_loss": 0.08106394112110138, "global_step": 237175, "epoch": 2664, "lr": 3.896439966235849e-06}
{"train_loss": 0.07467461377382278, "global_step": 237176, "epoch": 2664, "lr": 3.896215599509512e-06}
{"train_loss": 0.06907179951667786, "global_step": 237177, "epoch": 2664, "lr": 3.895991238981262e-06}
{"train_loss": 0.008517352864146233, "global_step": 237178, "epoch": 2664, "lr": 3.895766884651148e-06}
{"train_loss": 0.062129221856594086, "global_step": 237179, "epoch": 2664, "lr": 3.895542536519176e-06}
{"train_loss": 0.047297026962041855, "global_step": 237180, "epoch": 2664, "lr": 3.895318194585396e-06}
{"train_loss": 0.02231963910162449, "global_step": 237181, "epoch": 2664, "lr": 3.8950938588498186e-06}
{"train_loss": 0.06547766178846359, "global_step": 237182, "epoch": 2664, "lr": 3.894869529312495e-06}
{"train_loss": 0.09739447385072708, "global_step": 237183, "epoch": 2664, "lr": 3.894645205973435e-06}
{"train_loss": 0.04569647059346853, "global_step": 237184, "epoch": 2664, "lr": 3.894420888832684e-06, "val_loss": 9.401070594787598}
{"train_loss": 0.01931656338274479, "global_step": 237185, "epoch": 2665, "lr": 3.8941965778902635e-06}
{"train_loss": 0.045657433569431305, "global_step": 237186, "epoch": 2665, "lr": 3.893972273146207e-06}
{"train_loss": 0.10034514218568802, "global_step": 237187, "epoch": 2665, "lr": 3.893747974600542e-06}
{"train_loss": 0.022063815966248512, "global_step": 237188, "epoch": 2665, "lr": 3.893523682253303e-06}
{"train_loss": 0.05883433297276497, "global_step": 237189, "epoch": 2665, "lr": 3.893299396104516e-06}
{"train_loss": 0.08130458742380142, "global_step": 237190, "epoch": 2665, "lr": 3.893075116154216e-06}
{"train_loss": 0.037673480808734894, "global_step": 237191, "epoch": 2665, "lr": 3.8928508424024235e-06}
{"train_loss": 0.0737306997179985, "global_step": 237192, "epoch": 2665, "lr": 3.892626574849184e-06}
{"train_loss": 0.03901008516550064, "global_step": 237193, "epoch": 2665, "lr": 3.892402313494514e-06}
{"train_loss": 0.039018530398607254, "global_step": 237194, "epoch": 2665, "lr": 3.892178058338453e-06}
{"train_loss": 0.09957537800073624, "global_step": 237195, "epoch": 2665, "lr": 3.8919538093810155e-06}
{"train_loss": 0.07116776704788208, "global_step": 237196, "epoch": 2665, "lr": 3.8917295666222484e-06}
{"train_loss": 0.04758994281291962, "global_step": 237197, "epoch": 2665, "lr": 3.8915053300621786e-06}
{"train_loss": 0.05608460679650307, "global_step": 237198, "epoch": 2665, "lr": 3.891281099700833e-06}
{"train_loss": 0.07291452586650848, "global_step": 237199, "epoch": 2665, "lr": 3.891056875538246e-06}
{"train_loss": 0.04999494180083275, "global_step": 237200, "epoch": 2665, "lr": 3.8908326575744335e-06}
{"train_loss": 0.05419181287288666, "global_step": 237201, "epoch": 2665, "lr": 3.8906084458094515e-06}
{"train_loss": 0.0390656441450119, "global_step": 237202, "epoch": 2665, "lr": 3.890384240243306e-06}
{"train_loss": 0.055641017854213715, "global_step": 237203, "epoch": 2665, "lr": 3.89016004087604e-06}
{"train_loss": 0.036164913326501846, "global_step": 237204, "epoch": 2665, "lr": 3.8899358477076765e-06}
{"train_loss": 0.06982868164777756, "global_step": 237205, "epoch": 2665, "lr": 3.889711660738254e-06}
{"train_loss": 0.07521288841962814, "global_step": 237206, "epoch": 2665, "lr": 3.889487479967791e-06}
{"train_loss": 0.07139750570058823, "global_step": 237207, "epoch": 2665, "lr": 3.889263305396329e-06}
{"train_loss": 0.06530584394931793, "global_step": 237208, "epoch": 2665, "lr": 3.889039137023898e-06}
{"train_loss": 0.0338471494615078, "global_step": 237209, "epoch": 2665, "lr": 3.888814974850513e-06}
{"train_loss": 0.07568008452653885, "global_step": 237210, "epoch": 2665, "lr": 3.888590818876226e-06}
{"train_loss": 0.08215434104204178, "global_step": 237211, "epoch": 2665, "lr": 3.888366669101046e-06}
{"train_loss": 0.01839287020266056, "global_step": 237212, "epoch": 2665, "lr": 3.888142525525023e-06}
{"train_loss": 0.03862027823925018, "global_step": 237213, "epoch": 2665, "lr": 3.887918388148165e-06}
{"train_loss": 0.01180682610720396, "global_step": 237214, "epoch": 2665, "lr": 3.887694256970526e-06}
{"train_loss": 0.04191204160451889, "global_step": 237215, "epoch": 2665, "lr": 3.887470131992116e-06}
{"train_loss": 0.04228540509939194, "global_step": 237216, "epoch": 2665, "lr": 3.88724601321298e-06}
{"train_loss": 0.05364660173654556, "global_step": 237217, "epoch": 2665, "lr": 3.887021900633136e-06}
{"train_loss": 0.03991570696234703, "global_step": 237218, "epoch": 2665, "lr": 3.886797794252628e-06}
{"train_loss": 0.06967458873987198, "global_step": 237219, "epoch": 2665, "lr": 3.886573694071466e-06}
{"train_loss": 0.03565134108066559, "global_step": 237220, "epoch": 2665, "lr": 3.886349600089706e-06}
{"train_loss": 0.037659671157598495, "global_step": 237221, "epoch": 2665, "lr": 3.886125512307348e-06}
{"train_loss": 0.023648476228117943, "global_step": 237222, "epoch": 2665, "lr": 3.8859014307244535e-06}
{"train_loss": 0.052843689918518066, "global_step": 237223, "epoch": 2665, "lr": 3.885677355341027e-06}
{"train_loss": 0.03265148401260376, "global_step": 237224, "epoch": 2665, "lr": 3.88545328615712e-06}
{"train_loss": 0.047266487032175064, "global_step": 237225, "epoch": 2665, "lr": 3.8852292231727374e-06}
{"train_loss": 0.0296871829777956, "global_step": 237226, "epoch": 2665, "lr": 3.88500516638794e-06}
{"train_loss": 0.023676184937357903, "global_step": 237227, "epoch": 2665, "lr": 3.884781115802727e-06}
{"train_loss": 0.046541403979063034, "global_step": 237228, "epoch": 2665, "lr": 3.884557071417155e-06}
{"train_loss": 0.04644199088215828, "global_step": 237229, "epoch": 2665, "lr": 3.88433303323123e-06}
{"train_loss": 0.03724316880106926, "global_step": 237230, "epoch": 2665, "lr": 3.8841090012450065e-06}
{"train_loss": 0.05212686210870743, "global_step": 237231, "epoch": 2665, "lr": 3.883884975458491e-06}
{"train_loss": 0.02898658625781536, "global_step": 237232, "epoch": 2665, "lr": 3.883660955871726e-06}
{"train_loss": 0.0903937891125679, "global_step": 237233, "epoch": 2665, "lr": 3.883436942484747e-06}
{"train_loss": 0.06952343881130219, "global_step": 237234, "epoch": 2665, "lr": 3.883212935297576e-06}
{"train_loss": 0.05088955909013748, "global_step": 237235, "epoch": 2665, "lr": 3.88298893431025e-06}
{"train_loss": 0.025968024507164955, "global_step": 237236, "epoch": 2665, "lr": 3.882764939522781e-06}
{"train_loss": 0.07976716011762619, "global_step": 237237, "epoch": 2665, "lr": 3.882540950935226e-06}
{"train_loss": 0.09569254517555237, "global_step": 237238, "epoch": 2665, "lr": 3.882316968547595e-06}
{"train_loss": 0.019768139347434044, "global_step": 237239, "epoch": 2665, "lr": 3.882092992359926e-06}
{"train_loss": 0.05413597449660301, "global_step": 237240, "epoch": 2665, "lr": 3.881869022372242e-06}
{"train_loss": 0.03499244526028633, "global_step": 237241, "epoch": 2665, "lr": 3.881645058584587e-06}
{"train_loss": 0.05862237885594368, "global_step": 237242, "epoch": 2665, "lr": 3.8814211009969736e-06}
{"train_loss": 0.05425214394927025, "global_step": 237243, "epoch": 2665, "lr": 3.881197149609455e-06}
{"train_loss": 0.023658614605665207, "global_step": 237244, "epoch": 2665, "lr": 3.880973204422034e-06}
{"train_loss": 0.031027166172862053, "global_step": 237245, "epoch": 2665, "lr": 3.880749265434764e-06}
{"train_loss": 0.03874996677041054, "global_step": 237246, "epoch": 2665, "lr": 3.880525332647661e-06}
{"train_loss": 0.048304419964551926, "global_step": 237247, "epoch": 2665, "lr": 3.8803014060607544e-06}
{"train_loss": 0.08045843243598938, "global_step": 237248, "epoch": 2665, "lr": 3.880077485674088e-06}
{"train_loss": 0.05523167550563812, "global_step": 237249, "epoch": 2665, "lr": 3.8798535714876734e-06}
{"train_loss": 0.06523893028497696, "global_step": 237250, "epoch": 2665, "lr": 3.87962966350156e-06}
{"train_loss": 0.021650098264217377, "global_step": 237251, "epoch": 2665, "lr": 3.8794057617157595e-06}
{"train_loss": 0.0734725371003151, "global_step": 237252, "epoch": 2665, "lr": 3.879181866130316e-06}
{"train_loss": 0.04646407812833786, "global_step": 237253, "epoch": 2665, "lr": 3.878957976745251e-06}
{"train_loss": 0.0439925491809845, "global_step": 237254, "epoch": 2665, "lr": 3.878734093560604e-06}
{"train_loss": 0.039183422923088074, "global_step": 237255, "epoch": 2665, "lr": 3.8785102165763864e-06}
{"train_loss": 0.06864269077777863, "global_step": 237256, "epoch": 2665, "lr": 3.878286345792659e-06}
{"train_loss": 0.03961225226521492, "global_step": 237257, "epoch": 2665, "lr": 3.878062481209416e-06}
{"train_loss": 0.07189252972602844, "global_step": 237258, "epoch": 2665, "lr": 3.877838622826719e-06}
{"train_loss": 0.025679972022771835, "global_step": 237259, "epoch": 2665, "lr": 3.877614770644578e-06}
{"train_loss": 0.04889403656125069, "global_step": 237260, "epoch": 2665, "lr": 3.877390924663033e-06}
{"train_loss": 0.026732739061117172, "global_step": 237261, "epoch": 2665, "lr": 3.877167084882105e-06}
{"train_loss": 0.04019088298082352, "global_step": 237262, "epoch": 2665, "lr": 3.876943251301835e-06}
{"train_loss": 0.03618929535150528, "global_step": 237263, "epoch": 2665, "lr": 3.876719423922243e-06}
{"train_loss": 0.013904673047363758, "global_step": 237264, "epoch": 2665, "lr": 3.876495602743369e-06}
{"train_loss": 0.008381904102861881, "global_step": 237265, "epoch": 2665, "lr": 3.876271787765234e-06}
{"train_loss": 0.04395955428481102, "global_step": 237266, "epoch": 2665, "lr": 3.876047978987873e-06}
{"train_loss": 0.07791262865066528, "global_step": 237267, "epoch": 2665, "lr": 3.875824176411313e-06}
{"train_loss": 0.045311424881219864, "global_step": 237268, "epoch": 2665, "lr": 3.875600380035588e-06}
{"train_loss": 0.044171229004859924, "global_step": 237269, "epoch": 2665, "lr": 3.875376589860729e-06}
{"train_loss": 0.03887222334742546, "global_step": 237270, "epoch": 2665, "lr": 3.875152805886762e-06}
{"train_loss": 0.01006205752491951, "global_step": 237271, "epoch": 2665, "lr": 3.874929028113722e-06}
{"train_loss": 0.04940148815512657, "global_step": 237272, "epoch": 2665, "lr": 3.8747052565416275e-06}
{"train_loss": 0.048451180214041406, "global_step": 237273, "epoch": 2665, "lr": 3.874481491170523e-06, "val_loss": 9.447664260864258, "train_action_mse_error": 29.937044143676758}
{"train_loss": 0.04383038356900215, "global_step": 237274, "epoch": 2666, "lr": 3.874257732000425e-06}
{"train_loss": 0.040483761578798294, "global_step": 237275, "epoch": 2666, "lr": 3.874033979031383e-06}
{"train_loss": 0.04380464926362038, "global_step": 237276, "epoch": 2666, "lr": 3.873810232263403e-06}
{"train_loss": 0.05160611867904663, "global_step": 237277, "epoch": 2666, "lr": 3.8735864916965345e-06}
{"train_loss": 0.03291311860084534, "global_step": 237278, "epoch": 2666, "lr": 3.8733627573307955e-06}
{"train_loss": 0.05377109721302986, "global_step": 237279, "epoch": 2666, "lr": 3.873139029166229e-06}
{"train_loss": 0.06592101603746414, "global_step": 237280, "epoch": 2666, "lr": 3.872915307202845e-06}
{"train_loss": 0.021908974274992943, "global_step": 237281, "epoch": 2666, "lr": 3.872691591440697e-06}
{"train_loss": 0.0521559938788414, "global_step": 237282, "epoch": 2666, "lr": 3.872467881879793e-06}
{"train_loss": 0.02432353049516678, "global_step": 237283, "epoch": 2666, "lr": 3.872244178520185e-06}
{"train_loss": 0.05567379668354988, "global_step": 237284, "epoch": 2666, "lr": 3.872020481361877e-06}
{"train_loss": 0.03738199546933174, "global_step": 237285, "epoch": 2666, "lr": 3.8717967904049245e-06}
{"train_loss": 0.047221481800079346, "global_step": 237286, "epoch": 2666, "lr": 3.871573105649351e-06}
{"train_loss": 0.040975604206323624, "global_step": 237287, "epoch": 2666, "lr": 3.871349427095167e-06}
{"train_loss": 0.029679730534553528, "global_step": 237288, "epoch": 2666, "lr": 3.871125754742433e-06}
{"train_loss": 0.06475920975208282, "global_step": 237289, "epoch": 2666, "lr": 3.87090208859115e-06}
{"train_loss": 0.05693463981151581, "global_step": 237290, "epoch": 2666, "lr": 3.870678428641372e-06}
{"train_loss": 0.13748277723789215, "global_step": 237291, "epoch": 2666, "lr": 3.870454774893112e-06}
{"train_loss": 0.019908810034394264, "global_step": 237292, "epoch": 2666, "lr": 3.8702311273464184e-06}
{"train_loss": 0.03197367861866951, "global_step": 237293, "epoch": 2666, "lr": 3.870007486001298e-06}
{"train_loss": 0.02963467873632908, "global_step": 237294, "epoch": 2666, "lr": 3.8697838508578e-06}
{"train_loss": 0.05530589073896408, "global_step": 237295, "epoch": 2666, "lr": 3.869560221915941e-06}
{"train_loss": 0.06449267268180847, "global_step": 237296, "epoch": 2666, "lr": 3.869336599175766e-06}
{"train_loss": 0.023104676976799965, "global_step": 237297, "epoch": 2666, "lr": 3.869112982637285e-06}
{"train_loss": 0.028731342405080795, "global_step": 237298, "epoch": 2666, "lr": 3.86888937230055e-06}
{"train_loss": 0.06952926516532898, "global_step": 237299, "epoch": 2666, "lr": 3.868665768165569e-06}
{"train_loss": 0.11357180029153824, "global_step": 237300, "epoch": 2666, "lr": 3.868442170232395e-06}
{"train_loss": 0.022813402116298676, "global_step": 237301, "epoch": 2666, "lr": 3.868218578501043e-06}
{"train_loss": 0.08147307485342026, "global_step": 237302, "epoch": 2666, "lr": 3.8679949929715465e-06}
{"train_loss": 0.04904517903923988, "global_step": 237303, "epoch": 2666, "lr": 3.867771413643934e-06}
{"train_loss": 0.06382680684328079, "global_step": 237304, "epoch": 2666, "lr": 3.867547840518232e-06}
{"train_loss": 0.07180899381637573, "global_step": 237305, "epoch": 2666, "lr": 3.867324273594491e-06}
{"train_loss": 0.040812477469444275, "global_step": 237306, "epoch": 2666, "lr": 3.867100712872712e-06}
{"train_loss": 0.019512291997671127, "global_step": 237307, "epoch": 2666, "lr": 3.86687715835295e-06}
{"train_loss": 0.024194786325097084, "global_step": 237308, "epoch": 2666, "lr": 3.866653610035214e-06}
{"train_loss": 0.036187730729579926, "global_step": 237309, "epoch": 2666, "lr": 3.866430067919552e-06}
{"train_loss": 0.019192658364772797, "global_step": 237310, "epoch": 2666, "lr": 3.866206532005978e-06}
{"train_loss": 0.10172772407531738, "global_step": 237311, "epoch": 2666, "lr": 3.865983002294543e-06}
{"train_loss": 0.04229560121893883, "global_step": 237312, "epoch": 2666, "lr": 3.865759478785252e-06}
{"train_loss": 0.02403353340923786, "global_step": 237313, "epoch": 2666, "lr": 3.865535961478156e-06}
{"train_loss": 0.07062165439128876, "global_step": 237314, "epoch": 2666, "lr": 3.8653124503732655e-06}
{"train_loss": 0.007308955304324627, "global_step": 237315, "epoch": 2666, "lr": 3.865088945470635e-06}
{"train_loss": 0.026923073455691338, "global_step": 237316, "epoch": 2666, "lr": 3.864865446770272e-06}
{"train_loss": 0.049589354544878006, "global_step": 237317, "epoch": 2666, "lr": 3.864641954272219e-06}
{"train_loss": 0.04031343013048172, "global_step": 237318, "epoch": 2666, "lr": 3.864418467976499e-06}
{"train_loss": 0.05648556351661682, "global_step": 237319, "epoch": 2666, "lr": 3.864194987883152e-06}
{"train_loss": 0.05118325725197792, "global_step": 237320, "epoch": 2666, "lr": 3.8639715139921925e-06}
{"train_loss": 0.06682059913873672, "global_step": 237321, "epoch": 2666, "lr": 3.863748046303672e-06}
{"train_loss": 0.056679002940654755, "global_step": 237322, "epoch": 2666, "lr": 3.863524584817596e-06}
{"train_loss": 0.041967589408159256, "global_step": 237323, "epoch": 2666, "lr": 3.863301129534014e-06}
{"train_loss": 0.01836290769279003, "global_step": 237324, "epoch": 2666, "lr": 3.863077680452953e-06}
{"train_loss": 0.03167807683348656, "global_step": 237325, "epoch": 2666, "lr": 3.862854237574431e-06}
{"train_loss": 0.07063170522451401, "global_step": 237326, "epoch": 2666, "lr": 3.862630800898493e-06}
{"train_loss": 0.08095134794712067, "global_step": 237327, "epoch": 2666, "lr": 3.862407370425147e-06}
{"train_loss": 0.07252935320138931, "global_step": 237328, "epoch": 2666, "lr": 3.862183946154457e-06}
{"train_loss": 0.021432770416140556, "global_step": 237329, "epoch": 2666, "lr": 3.861960528086416e-06}
{"train_loss": 0.0384678952395916, "global_step": 237330, "epoch": 2666, "lr": 3.861737116221087e-06}
{"train_loss": 0.03185040131211281, "global_step": 237331, "epoch": 2666, "lr": 3.861513710558473e-06}
{"train_loss": 0.026546621695160866, "global_step": 237332, "epoch": 2666, "lr": 3.861290311098625e-06}
{"train_loss": 0.03489452973008156, "global_step": 237333, "epoch": 2666, "lr": 3.861066917841561e-06}
{"train_loss": 0.03013550117611885, "global_step": 237334, "epoch": 2666, "lr": 3.860843530787317e-06}
{"train_loss": 0.06913276016712189, "global_step": 237335, "epoch": 2666, "lr": 3.860620149935912e-06}
{"train_loss": 0.0928943008184433, "global_step": 237336, "epoch": 2666, "lr": 3.860396775287395e-06}
{"train_loss": 0.037485942244529724, "global_step": 237337, "epoch": 2666, "lr": 3.860173406841777e-06}
{"train_loss": 0.0223862174898386, "global_step": 237338, "epoch": 2666, "lr": 3.859950044599103e-06}
{"train_loss": 0.04304744675755501, "global_step": 237339, "epoch": 2666, "lr": 3.85972668855939e-06}
{"train_loss": 0.049334149807691574, "global_step": 237340, "epoch": 2666, "lr": 3.85950333872267e-06}
{"train_loss": 0.04874388501048088, "global_step": 237341, "epoch": 2666, "lr": 3.8592799950889935e-06}
{"train_loss": 0.012728546746075153, "global_step": 237342, "epoch": 2666, "lr": 3.859056657658361e-06}
{"train_loss": 0.061986926943063736, "global_step": 237343, "epoch": 2666, "lr": 3.8588333264308215e-06}
{"train_loss": 0.05816154181957245, "global_step": 237344, "epoch": 2666, "lr": 3.858610001406393e-06}
{"train_loss": 0.04565158486366272, "global_step": 237345, "epoch": 2666, "lr": 3.858386682585124e-06}
{"train_loss": 0.023570585995912552, "global_step": 237346, "epoch": 2666, "lr": 3.858163369967021e-06}
{"train_loss": 0.030321089550852776, "global_step": 237347, "epoch": 2666, "lr": 3.857940063552134e-06}
{"train_loss": 0.04616232588887215, "global_step": 237348, "epoch": 2666, "lr": 3.8577167633404796e-06}
{"train_loss": 0.034096330404281616, "global_step": 237349, "epoch": 2666, "lr": 3.857493469332096e-06}
{"train_loss": 0.040950238704681396, "global_step": 237350, "epoch": 2666, "lr": 3.857270181527007e-06}
{"train_loss": 0.019616970792412758, "global_step": 237351, "epoch": 2666, "lr": 3.8570468999252485e-06}
{"train_loss": 0.06714248657226562, "global_step": 237352, "epoch": 2666, "lr": 3.856823624526845e-06}
{"train_loss": 0.02289712056517601, "global_step": 237353, "epoch": 2666, "lr": 3.8566003553318355e-06}
{"train_loss": 0.044918254017829895, "global_step": 237354, "epoch": 2666, "lr": 3.85637709234023e-06}
{"train_loss": 0.07868021726608276, "global_step": 237355, "epoch": 2666, "lr": 3.856153835552084e-06}
{"train_loss": 0.02325953170657158, "global_step": 237356, "epoch": 2666, "lr": 3.85593058496741e-06}
{"train_loss": 0.04204002022743225, "global_step": 237357, "epoch": 2666, "lr": 3.855707340586251e-06}
{"train_loss": 0.07363276183605194, "global_step": 237358, "epoch": 2666, "lr": 3.855484102408619e-06}
{"train_loss": 0.04241403937339783, "global_step": 237359, "epoch": 2666, "lr": 3.8552608704345685e-06}
{"train_loss": 0.02501031570136547, "global_step": 237360, "epoch": 2666, "lr": 3.8550376446641e-06}
{"train_loss": 0.02232241816818714, "global_step": 237361, "epoch": 2666, "lr": 3.85481442509727e-06}
{"train_loss": 0.04543593127196759, "global_step": 237362, "epoch": 2666, "lr": 3.854591211734099e-06, "val_loss": 9.341322898864746}
{"train_loss": 0.049106914550065994, "global_step": 237363, "epoch": 2667, "lr": 3.85436800457461e-06}
{"train_loss": 0.02037670835852623, "global_step": 237364, "epoch": 2667, "lr": 3.854144803618842e-06}
{"train_loss": 0.051185429096221924, "global_step": 237365, "epoch": 2667, "lr": 3.853921608866817e-06}
{"train_loss": 0.06566949933767319, "global_step": 237366, "epoch": 2667, "lr": 3.85369842031858e-06}
{"train_loss": 0.034219466149806976, "global_step": 237367, "epoch": 2667, "lr": 3.853475237974136e-06}
{"train_loss": 0.06424587965011597, "global_step": 237368, "epoch": 2667, "lr": 3.853252061833545e-06}
{"train_loss": 0.04442204535007477, "global_step": 237369, "epoch": 2667, "lr": 3.85302889189681e-06}
{"train_loss": 0.064353808760643, "global_step": 237370, "epoch": 2667, "lr": 3.852805728163977e-06}
{"train_loss": 0.07420340180397034, "global_step": 237371, "epoch": 2667, "lr": 3.852582570635072e-06}
{"train_loss": 0.024161141365766525, "global_step": 237372, "epoch": 2667, "lr": 3.852359419310125e-06}
{"train_loss": 0.028813673183321953, "global_step": 237373, "epoch": 2667, "lr": 3.852136274189161e-06}
{"train_loss": 0.04277481511235237, "global_step": 237374, "epoch": 2667, "lr": 3.851913135272228e-06}
{"train_loss": 0.016725080087780952, "global_step": 237375, "epoch": 2667, "lr": 3.851690002559327e-06}
{"train_loss": 0.03557564690709114, "global_step": 237376, "epoch": 2667, "lr": 3.851466876050519e-06}
{"train_loss": 0.021528106182813644, "global_step": 237377, "epoch": 2667, "lr": 3.851243755745803e-06}
{"train_loss": 0.05617727339267731, "global_step": 237378, "epoch": 2667, "lr": 3.851020641645231e-06}
{"train_loss": 0.01847446709871292, "global_step": 237379, "epoch": 2667, "lr": 3.850797533748829e-06}
{"train_loss": 0.05045868828892708, "global_step": 237380, "epoch": 2667, "lr": 3.8505744320566195e-06}
{"train_loss": 0.07081843912601471, "global_step": 237381, "epoch": 2667, "lr": 3.850351336568647e-06}
{"train_loss": 0.08347602188587189, "global_step": 237382, "epoch": 2667, "lr": 3.8501282472849235e-06}
{"train_loss": 0.08032386004924774, "global_step": 237383, "epoch": 2667, "lr": 3.849905164205498e-06}
{"train_loss": 0.046301137655973434, "global_step": 237384, "epoch": 2667, "lr": 3.849682087330381e-06}
{"train_loss": 0.07273922860622406, "global_step": 237385, "epoch": 2667, "lr": 3.849459016659618e-06}
{"train_loss": 0.06362608820199966, "global_step": 237386, "epoch": 2667, "lr": 3.8492359521932255e-06}
{"train_loss": 0.06485112011432648, "global_step": 237387, "epoch": 2667, "lr": 3.849012893931253e-06}
{"train_loss": 0.02598501555621624, "global_step": 237388, "epoch": 2667, "lr": 3.848789841873701e-06}
{"train_loss": 0.08457224071025848, "global_step": 237389, "epoch": 2667, "lr": 3.8485667960206305e-06}
{"train_loss": 0.08518568426370621, "global_step": 237390, "epoch": 2667, "lr": 3.848343756372052e-06}
{"train_loss": 0.0550333596765995, "global_step": 237391, "epoch": 2667, "lr": 3.848120722928006e-06}
{"train_loss": 0.02260972000658512, "global_step": 237392, "epoch": 2667, "lr": 3.847897695688507e-06}
{"train_loss": 0.09391758590936661, "global_step": 237393, "epoch": 2667, "lr": 3.847674674653606e-06}
{"train_loss": 0.055471889674663544, "global_step": 237394, "epoch": 2667, "lr": 3.84745165982332e-06}
{"train_loss": 0.08844423294067383, "global_step": 237395, "epoch": 2667, "lr": 3.847228651197676e-06}
{"train_loss": 0.0636758953332901, "global_step": 237396, "epoch": 2667, "lr": 3.847005648776719e-06}
{"train_loss": 0.05801717936992645, "global_step": 237397, "epoch": 2667, "lr": 3.846782652560466e-06}
{"train_loss": 0.038498032838106155, "global_step": 237398, "epoch": 2667, "lr": 3.846559662548949e-06}
{"train_loss": 0.05093934386968613, "global_step": 237399, "epoch": 2667, "lr": 3.846336678742202e-06}
{"train_loss": 0.02654113806784153, "global_step": 237400, "epoch": 2667, "lr": 3.8461137011402474e-06}
{"train_loss": 0.026834022253751755, "global_step": 237401, "epoch": 2667, "lr": 3.845890729743129e-06}
{"train_loss": 0.027808813378214836, "global_step": 237402, "epoch": 2667, "lr": 3.845667764550864e-06}
{"train_loss": 0.04208140820264816, "global_step": 237403, "epoch": 2667, "lr": 3.8454448055634805e-06}
{"train_loss": 0.051629941910505295, "global_step": 237404, "epoch": 2667, "lr": 3.845221852781022e-06}
{"train_loss": 0.0655030757188797, "global_step": 237405, "epoch": 2667, "lr": 3.844998906203506e-06}
{"train_loss": 0.07087941467761993, "global_step": 237406, "epoch": 2667, "lr": 3.8447759658309715e-06}
{"train_loss": 0.027755141258239746, "global_step": 237407, "epoch": 2667, "lr": 3.844553031663439e-06}
{"train_loss": 0.08403847366571426, "global_step": 237408, "epoch": 2667, "lr": 3.84433010370095e-06}
{"train_loss": 0.11960276961326599, "global_step": 237409, "epoch": 2667, "lr": 3.844107181943518e-06}
{"train_loss": 0.0532420352101326, "global_step": 237410, "epoch": 2667, "lr": 3.8438842663911954e-06}
{"train_loss": 0.050211142748594284, "global_step": 237411, "epoch": 2667, "lr": 3.843661357043993e-06}
{"train_loss": 0.04636606201529503, "global_step": 237412, "epoch": 2667, "lr": 3.8434384539019545e-06}
{"train_loss": 0.08150982856750488, "global_step": 237413, "epoch": 2667, "lr": 3.843215556965091e-06}
{"train_loss": 0.0707053542137146, "global_step": 237414, "epoch": 2667, "lr": 3.842992666233447e-06}
{"train_loss": 0.05181707814335823, "global_step": 237415, "epoch": 2667, "lr": 3.842769781707062e-06}
{"train_loss": 0.04451334848999977, "global_step": 237416, "epoch": 2667, "lr": 3.842546903385941e-06}
{"train_loss": 0.06148034706711769, "global_step": 237417, "epoch": 2667, "lr": 3.842324031270139e-06}
{"train_loss": 0.055377162992954254, "global_step": 237418, "epoch": 2667, "lr": 3.842101165359663e-06}
{"train_loss": 0.0344957672059536, "global_step": 237419, "epoch": 2667, "lr": 3.841878305654562e-06}
{"train_loss": 0.0864081010222435, "global_step": 237420, "epoch": 2667, "lr": 3.841655452154852e-06}
{"train_loss": 0.05368085578083992, "global_step": 237421, "epoch": 2667, "lr": 3.841432604860573e-06}
{"train_loss": 0.04414445534348488, "global_step": 237422, "epoch": 2667, "lr": 3.841209763771747e-06}
{"train_loss": 0.03753955289721489, "global_step": 237423, "epoch": 2667, "lr": 3.840986928888412e-06}
{"train_loss": 0.01649284176528454, "global_step": 237424, "epoch": 2667, "lr": 3.840764100210586e-06}
{"train_loss": 0.02851175144314766, "global_step": 237425, "epoch": 2667, "lr": 3.840541277738318e-06}
{"train_loss": 0.053781282156705856, "global_step": 237426, "epoch": 2667, "lr": 3.840318461471615e-06}
{"train_loss": 0.10502862930297852, "global_step": 237427, "epoch": 2667, "lr": 3.84009565141053e-06}
{"train_loss": 0.016446519643068314, "global_step": 237428, "epoch": 2667, "lr": 3.839872847555071e-06}
{"train_loss": 0.12930698692798615, "global_step": 237429, "epoch": 2667, "lr": 3.839650049905291e-06}
{"train_loss": 0.04943706467747688, "global_step": 237430, "epoch": 2667, "lr": 3.839427258461192e-06}
{"train_loss": 0.07001032680273056, "global_step": 237431, "epoch": 2667, "lr": 3.839204473222824e-06}
{"train_loss": 0.07790400832891464, "global_step": 237432, "epoch": 2667, "lr": 3.838981694190219e-06}
{"train_loss": 0.03308042883872986, "global_step": 237433, "epoch": 2667, "lr": 3.838758921363395e-06}
{"train_loss": 0.02056431770324707, "global_step": 237434, "epoch": 2667, "lr": 3.838536154742395e-06}
{"train_loss": 0.01825939118862152, "global_step": 237435, "epoch": 2667, "lr": 3.838313394327236e-06}
{"train_loss": 0.030334260314702988, "global_step": 237436, "epoch": 2667, "lr": 3.838090640117947e-06}
{"train_loss": 0.02799859829246998, "global_step": 237437, "epoch": 2667, "lr": 3.8378678921145774e-06}
{"train_loss": 0.057954370975494385, "global_step": 237438, "epoch": 2667, "lr": 3.837645150317132e-06}
{"train_loss": 0.0464862659573555, "global_step": 237439, "epoch": 2667, "lr": 3.837422414725661e-06}
{"train_loss": 0.02583075501024723, "global_step": 237440, "epoch": 2667, "lr": 3.837199685340182e-06}
{"train_loss": 0.03867984935641289, "global_step": 237441, "epoch": 2667, "lr": 3.836976962160727e-06}
{"train_loss": 0.04171261563897133, "global_step": 237442, "epoch": 2667, "lr": 3.836754245187329e-06}
{"train_loss": 0.0506695993244648, "global_step": 237443, "epoch": 2667, "lr": 3.836531534420018e-06}
{"train_loss": 0.057700611650943756, "global_step": 237444, "epoch": 2667, "lr": 3.836308829858826e-06}
{"train_loss": 0.02994462288916111, "global_step": 237445, "epoch": 2667, "lr": 3.836086131503769e-06}
{"train_loss": 0.07942231744527817, "global_step": 237446, "epoch": 2667, "lr": 3.8358634393548975e-06}
{"train_loss": 0.06945393234491348, "global_step": 237447, "epoch": 2667, "lr": 3.835640753412223e-06}
{"train_loss": 0.04393116384744644, "global_step": 237448, "epoch": 2667, "lr": 3.835418073675795e-06}
{"train_loss": 0.05390212684869766, "global_step": 237449, "epoch": 2667, "lr": 3.835195400145619e-06}
{"train_loss": 0.026094328612089157, "global_step": 237450, "epoch": 2667, "lr": 3.83497273282174e-06}
{"train_loss": 0.05178641285119432, "global_step": 237451, "epoch": 2667, "lr": 3.834750071704196e-06, "val_loss": 9.391558647155762}
{"train_loss": 0.051425833255052567, "global_step": 237452, "epoch": 2668, "lr": 3.834527416792999e-06}
{"train_loss": 0.06274570524692535, "global_step": 237453, "epoch": 2668, "lr": 3.834304768088198e-06}
{"train_loss": 0.027423936873674393, "global_step": 237454, "epoch": 2668, "lr": 3.8340821255898e-06}
{"train_loss": 0.024280399084091187, "global_step": 237455, "epoch": 2668, "lr": 3.833859489297853e-06}
{"train_loss": 0.005569214466959238, "global_step": 237456, "epoch": 2668, "lr": 3.833636859212375e-06}
{"train_loss": 0.06509558111429214, "global_step": 237457, "epoch": 2668, "lr": 3.8334142353334155e-06}
{"train_loss": 0.023480653762817383, "global_step": 237458, "epoch": 2668, "lr": 3.833191617660975e-06}
{"train_loss": 0.011468983255326748, "global_step": 237459, "epoch": 2668, "lr": 3.832969006195108e-06}
{"train_loss": 0.054499778896570206, "global_step": 237460, "epoch": 2668, "lr": 3.832746400935833e-06}
{"train_loss": 0.03448541462421417, "global_step": 237461, "epoch": 2668, "lr": 3.832523801883186e-06}
{"train_loss": 0.022689009085297585, "global_step": 237462, "epoch": 2668, "lr": 3.832301209037187e-06}
{"train_loss": 0.035019420087337494, "global_step": 237463, "epoch": 2668, "lr": 3.8320786223978834e-06}
{"train_loss": 0.034406520426273346, "global_step": 237464, "epoch": 2668, "lr": 3.831856041965282e-06}
{"train_loss": 0.03354029357433319, "global_step": 237465, "epoch": 2668, "lr": 3.831633467739432e-06}
{"train_loss": 0.0337214320898056, "global_step": 237466, "epoch": 2668, "lr": 3.83141089972035e-06}
{"train_loss": 0.03769398480653763, "global_step": 237467, "epoch": 2668, "lr": 3.83118833790807e-06}
{"train_loss": 0.05236002430319786, "global_step": 237468, "epoch": 2668, "lr": 3.830965782302636e-06}
{"train_loss": 0.022631801664829254, "global_step": 237469, "epoch": 2668, "lr": 3.830743232904055e-06}
{"train_loss": 0.08716770261526108, "global_step": 237470, "epoch": 2668, "lr": 3.830520689712375e-06}
{"train_loss": 0.07560626417398453, "global_step": 237471, "epoch": 2668, "lr": 3.830298152727613e-06}
{"train_loss": 0.027600187808275223, "global_step": 237472, "epoch": 2668, "lr": 3.830075621949813e-06}
{"train_loss": 0.06230888143181801, "global_step": 237473, "epoch": 2668, "lr": 3.829853097378994e-06}
{"train_loss": 0.07325790077447891, "global_step": 237474, "epoch": 2668, "lr": 3.829630579015186e-06}
{"train_loss": 0.06191381439566612, "global_step": 237475, "epoch": 2668, "lr": 3.829408066858425e-06}
{"train_loss": 0.0357317291200161, "global_step": 237476, "epoch": 2668, "lr": 3.829185560908726e-06}
{"train_loss": 0.05456232652068138, "global_step": 237477, "epoch": 2668, "lr": 3.828963061166146e-06}
{"train_loss": 0.02422783337533474, "global_step": 237478, "epoch": 2668, "lr": 3.828740567630695e-06}
{"train_loss": 0.0220039002597332, "global_step": 237479, "epoch": 2668, "lr": 3.828518080302401e-06}
{"train_loss": 0.05033721402287483, "global_step": 237480, "epoch": 2668, "lr": 3.828295599181303e-06}
{"train_loss": 0.025418641045689583, "global_step": 237481, "epoch": 2668, "lr": 3.828073124267423e-06}
{"train_loss": 0.015089355409145355, "global_step": 237482, "epoch": 2668, "lr": 3.827850655560805e-06}
{"train_loss": 0.022322818636894226, "global_step": 237483, "epoch": 2668, "lr": 3.827628193061461e-06}
{"train_loss": 0.04379591718316078, "global_step": 237484, "epoch": 2668, "lr": 3.82740573676944e-06}
{"train_loss": 0.05479420721530914, "global_step": 237485, "epoch": 2668, "lr": 3.827183286684749e-06}
{"train_loss": 0.058997344225645065, "global_step": 237486, "epoch": 2668, "lr": 3.826960842807437e-06}
{"train_loss": 0.08692151308059692, "global_step": 237487, "epoch": 2668, "lr": 3.8267384051375256e-06}
{"train_loss": 0.010058879852294922, "global_step": 237488, "epoch": 2668, "lr": 3.826515973675044e-06}
{"train_loss": 0.05060264840722084, "global_step": 237489, "epoch": 2668, "lr": 3.826293548420034e-06}
{"train_loss": 0.049818091094493866, "global_step": 237490, "epoch": 2668, "lr": 3.826071129372505e-06}
{"train_loss": 0.06255102902650833, "global_step": 237491, "epoch": 2668, "lr": 3.825848716532509e-06}
{"train_loss": 0.025314975529909134, "global_step": 237492, "epoch": 2668, "lr": 3.8256263099000546e-06}
{"train_loss": 0.06842850893735886, "global_step": 237493, "epoch": 2668, "lr": 3.825403909475189e-06}
{"train_loss": 0.02888529561460018, "global_step": 237494, "epoch": 2668, "lr": 3.825181515257931e-06}
{"train_loss": 0.030717499554157257, "global_step": 237495, "epoch": 2668, "lr": 3.824959127248317e-06}
{"train_loss": 0.018086016178131104, "global_step": 237496, "epoch": 2668, "lr": 3.824736745446373e-06}
{"train_loss": 0.05766833573579788, "global_step": 237497, "epoch": 2668, "lr": 3.824514369852133e-06}
{"train_loss": 0.034357137978076935, "global_step": 237498, "epoch": 2668, "lr": 3.824292000465618e-06}
{"train_loss": 0.0328393280506134, "global_step": 237499, "epoch": 2668, "lr": 3.824069637286876e-06}
{"train_loss": 0.04609978199005127, "global_step": 237500, "epoch": 2668, "lr": 3.823847280315912e-06}
{"train_loss": 0.03799184411764145, "global_step": 237501, "epoch": 2668, "lr": 3.823624929552777e-06}
{"train_loss": 0.05631425976753235, "global_step": 237502, "epoch": 2668, "lr": 3.823402584997488e-06}
{"train_loss": 0.0300280898809433, "global_step": 237503, "epoch": 2668, "lr": 3.823180246650082e-06}
{"train_loss": 0.024881599470973015, "global_step": 237504, "epoch": 2668, "lr": 3.822957914510583e-06}
{"train_loss": 0.04676319658756256, "global_step": 237505, "epoch": 2668, "lr": 3.822735588579024e-06}
{"train_loss": 0.04952859506011009, "global_step": 237506, "epoch": 2668, "lr": 3.822513268855444e-06}
{"train_loss": 0.05925244465470314, "global_step": 237507, "epoch": 2668, "lr": 3.822290955339858e-06}
{"train_loss": 0.05070134252309799, "global_step": 237508, "epoch": 2668, "lr": 3.822068648032307e-06}
{"train_loss": 0.04076424613595009, "global_step": 237509, "epoch": 2668, "lr": 3.821846346932806e-06}
{"train_loss": 0.057603929191827774, "global_step": 237510, "epoch": 2668, "lr": 3.821624052041406e-06}
{"train_loss": 0.07589717209339142, "global_step": 237511, "epoch": 2668, "lr": 3.821401763358118e-06}
{"train_loss": 0.07571583241224289, "global_step": 237512, "epoch": 2668, "lr": 3.821179480882986e-06}
{"train_loss": 0.03440818190574646, "global_step": 237513, "epoch": 2668, "lr": 3.820957204616038e-06}
{"train_loss": 0.033590637147426605, "global_step": 237514, "epoch": 2668, "lr": 3.820734934557285e-06}
{"train_loss": 0.03321212902665138, "global_step": 237515, "epoch": 2668, "lr": 3.820512670706783e-06}
{"train_loss": 0.04857894778251648, "global_step": 237516, "epoch": 2668, "lr": 3.8202904130645426e-06}
{"train_loss": 0.046338241547346115, "global_step": 237517, "epoch": 2668, "lr": 3.820068161630608e-06}
{"train_loss": 0.033997636288404465, "global_step": 237518, "epoch": 2668, "lr": 3.8198459164050075e-06}
{"train_loss": 0.03920914977788925, "global_step": 237519, "epoch": 2668, "lr": 3.819623677387751e-06}
{"train_loss": 0.0470733605325222, "global_step": 237520, "epoch": 2668, "lr": 3.81940144457889e-06}
{"train_loss": 0.03624136000871658, "global_step": 237521, "epoch": 2668, "lr": 3.819179217978447e-06}
{"train_loss": 0.017444806173443794, "global_step": 237522, "epoch": 2668, "lr": 3.818956997586448e-06}
{"train_loss": 0.02432258427143097, "global_step": 237523, "epoch": 2668, "lr": 3.818734783402933e-06}
{"train_loss": 0.04302896559238434, "global_step": 237524, "epoch": 2668, "lr": 3.818512575427924e-06}
{"train_loss": 0.05462241172790527, "global_step": 237525, "epoch": 2668, "lr": 3.818290373661459e-06}
{"train_loss": 0.04438207671046257, "global_step": 237526, "epoch": 2668, "lr": 3.818068178103551e-06}
{"train_loss": 0.021226637065410614, "global_step": 237527, "epoch": 2668, "lr": 3.817845988754254e-06}
{"train_loss": 0.06317716836929321, "global_step": 237528, "epoch": 2668, "lr": 3.817623805613574e-06}
{"train_loss": 0.018965033814311028, "global_step": 237529, "epoch": 2668, "lr": 3.817401628681555e-06}
{"train_loss": 0.06326019018888474, "global_step": 237530, "epoch": 2668, "lr": 3.8171794579582195e-06}
{"train_loss": 0.058792103081941605, "global_step": 237531, "epoch": 2668, "lr": 3.816957293443612e-06}
{"train_loss": 0.04985661059617996, "global_step": 237532, "epoch": 2668, "lr": 3.816735135137739e-06}
{"train_loss": 0.06629807502031326, "global_step": 237533, "epoch": 2668, "lr": 3.816512983040654e-06}
{"train_loss": 0.06734032928943634, "global_step": 237534, "epoch": 2668, "lr": 3.81629083715237e-06}
{"train_loss": 0.03526528552174568, "global_step": 237535, "epoch": 2668, "lr": 3.81606869747293e-06}
{"train_loss": 0.09866750985383987, "global_step": 237536, "epoch": 2668, "lr": 3.815846564002345e-06}
{"train_loss": 0.024735024198889732, "global_step": 237537, "epoch": 2668, "lr": 3.8156244367406664e-06}
{"train_loss": 0.0686526894569397, "global_step": 237538, "epoch": 2668, "lr": 3.815402315687905e-06}
{"train_loss": 0.031857408583164215, "global_step": 237539, "epoch": 2668, "lr": 3.8151802008441095e-06}
{"train_loss": 0.043406815001259695, "global_step": 237540, "epoch": 2668, "lr": 3.814958092209292e-06, "val_loss": 9.289671897888184}
{"train_loss": 0.016872463747859, "global_step": 237541, "epoch": 2669, "lr": 3.814735989783491e-06}
{"train_loss": 0.027550019323825836, "global_step": 237542, "epoch": 2669, "lr": 3.81451389356674e-06}
{"train_loss": 0.06186902895569801, "global_step": 237543, "epoch": 2669, "lr": 3.8142918035590613e-06}
{"train_loss": 0.04964549094438553, "global_step": 237544, "epoch": 2669, "lr": 3.8140697197604937e-06}
{"train_loss": 0.054399244487285614, "global_step": 237545, "epoch": 2669, "lr": 3.8138476421710533e-06}
{"train_loss": 0.044937312602996826, "global_step": 237546, "epoch": 2669, "lr": 3.813625570790791e-06}
{"train_loss": 0.03693331405520439, "global_step": 237547, "epoch": 2669, "lr": 3.8134035056197114e-06}
{"train_loss": 0.06413377821445465, "global_step": 237548, "epoch": 2669, "lr": 3.813181446657865e-06}
{"train_loss": 0.03810344263911247, "global_step": 237549, "epoch": 2669, "lr": 3.8129593939052632e-06}
{"train_loss": 0.06875994056463242, "global_step": 237550, "epoch": 2669, "lr": 3.8127373473619554e-06}
{"train_loss": 0.04320163279771805, "global_step": 237551, "epoch": 2669, "lr": 3.812515307027964e-06}
{"train_loss": 0.052909381687641144, "global_step": 237552, "epoch": 2669, "lr": 3.812293272903311e-06}
{"train_loss": 0.02297798916697502, "global_step": 237553, "epoch": 2669, "lr": 3.8120712449880357e-06}
{"train_loss": 0.051899440586566925, "global_step": 237554, "epoch": 2669, "lr": 3.8118492232821602e-06}
{"train_loss": 0.055475667119026184, "global_step": 237555, "epoch": 2669, "lr": 3.8116272077857228e-06}
{"train_loss": 0.03563480079174042, "global_step": 237556, "epoch": 2669, "lr": 3.811405198498752e-06}
{"train_loss": 0.022202590480446815, "global_step": 237557, "epoch": 2669, "lr": 3.811183195421264e-06}
{"train_loss": 0.03325207531452179, "global_step": 237558, "epoch": 2669, "lr": 3.810961198553303e-06}
{"train_loss": 0.03824159502983093, "global_step": 237559, "epoch": 2669, "lr": 3.8107392078949033e-06}
{"train_loss": 0.026812341064214706, "global_step": 237560, "epoch": 2669, "lr": 3.810517223446075e-06}
{"train_loss": 0.036900028586387634, "global_step": 237561, "epoch": 2669, "lr": 3.8102952452068685e-06}
{"train_loss": 0.06187231466174126, "global_step": 237562, "epoch": 2669, "lr": 3.8100732731773e-06}
{"train_loss": 0.026841402053833008, "global_step": 237563, "epoch": 2669, "lr": 3.809851307357409e-06}
{"train_loss": 0.02605540119111538, "global_step": 237564, "epoch": 2669, "lr": 3.8096293477472124e-06}
{"train_loss": 0.0538993775844574, "global_step": 237565, "epoch": 2669, "lr": 3.809407394346759e-06}
{"train_loss": 0.08446420729160309, "global_step": 237566, "epoch": 2669, "lr": 3.8091854471560605e-06}
{"train_loss": 0.06390368193387985, "global_step": 237567, "epoch": 2669, "lr": 3.808963506175156e-06}
{"train_loss": 0.053460098803043365, "global_step": 237568, "epoch": 2669, "lr": 3.808741571404073e-06}
{"train_loss": 0.06927122920751572, "global_step": 237569, "epoch": 2669, "lr": 3.8085196428428447e-06}
{"train_loss": 0.05966217815876007, "global_step": 237570, "epoch": 2669, "lr": 3.808297720491488e-06}
{"train_loss": 0.09454843401908875, "global_step": 237571, "epoch": 2669, "lr": 3.8080758043500587e-06}
{"train_loss": 0.06397861242294312, "global_step": 237572, "epoch": 2669, "lr": 3.807853894418556e-06}
{"train_loss": 0.023359034210443497, "global_step": 237573, "epoch": 2669, "lr": 3.807631990697036e-06}
{"train_loss": 0.04629819095134735, "global_step": 237574, "epoch": 2669, "lr": 3.807410093185504e-06}
{"train_loss": 0.039277736097574234, "global_step": 237575, "epoch": 2669, "lr": 3.8071882018840153e-06}
{"train_loss": 0.0462288074195385, "global_step": 237576, "epoch": 2669, "lr": 3.8069663167925763e-06}
{"train_loss": 0.058908555656671524, "global_step": 237577, "epoch": 2669, "lr": 3.8067444379112305e-06}
{"train_loss": 0.020672384649515152, "global_step": 237578, "epoch": 2669, "lr": 3.8065225652400118e-06}
{"train_loss": 0.04562068730592728, "global_step": 237579, "epoch": 2669, "lr": 3.8063006987789307e-06}
{"train_loss": 0.06429354846477509, "global_step": 237580, "epoch": 2669, "lr": 3.8060788385280433e-06}
{"train_loss": 0.03945254907011986, "global_step": 237581, "epoch": 2669, "lr": 3.805856984487355e-06}
{"train_loss": 0.061217181384563446, "global_step": 237582, "epoch": 2669, "lr": 3.8056351366569156e-06}
{"train_loss": 0.08104699850082397, "global_step": 237583, "epoch": 2669, "lr": 3.805413295036736e-06}
{"train_loss": 0.07672146707773209, "global_step": 237584, "epoch": 2669, "lr": 3.805191459626861e-06}
{"train_loss": 0.06573712825775146, "global_step": 237585, "epoch": 2669, "lr": 3.804969630427313e-06}
{"train_loss": 0.0641852393746376, "global_step": 237586, "epoch": 2669, "lr": 3.8047478074381304e-06}
{"train_loss": 0.04697185754776001, "global_step": 237587, "epoch": 2669, "lr": 3.8045259906593243e-06}
{"train_loss": 0.047809116542339325, "global_step": 237588, "epoch": 2669, "lr": 3.8043041800909507e-06}
{"train_loss": 0.04369167983531952, "global_step": 237589, "epoch": 2669, "lr": 3.80408237573302e-06}
{"train_loss": 0.03996703773736954, "global_step": 237590, "epoch": 2669, "lr": 3.803860577585561e-06}
{"train_loss": 0.02774070017039776, "global_step": 237591, "epoch": 2669, "lr": 3.8036387856486166e-06}
{"train_loss": 0.07376737892627716, "global_step": 237592, "epoch": 2669, "lr": 3.803416999922205e-06}
{"train_loss": 0.052721694111824036, "global_step": 237593, "epoch": 2669, "lr": 3.803195220406369e-06}
{"train_loss": 0.05681118369102478, "global_step": 237594, "epoch": 2669, "lr": 3.8029734471011215e-06}
{"train_loss": 0.057992540299892426, "global_step": 237595, "epoch": 2669, "lr": 3.8027516800065055e-06}
{"train_loss": 0.0476250983774662, "global_step": 237596, "epoch": 2669, "lr": 3.802529919122544e-06}
{"train_loss": 0.048210278153419495, "global_step": 237597, "epoch": 2669, "lr": 3.802308164449275e-06}
{"train_loss": 0.04123673960566521, "global_step": 237598, "epoch": 2669, "lr": 3.8020864159867107e-06}
{"train_loss": 0.04560592770576477, "global_step": 237599, "epoch": 2669, "lr": 3.801864673734906e-06}
{"train_loss": 0.0400402769446373, "global_step": 237600, "epoch": 2669, "lr": 3.801642937693867e-06}
{"train_loss": 0.079017274081707, "global_step": 237601, "epoch": 2669, "lr": 3.801421207863648e-06}
{"train_loss": 0.04466840997338295, "global_step": 237602, "epoch": 2669, "lr": 3.80119948424425e-06}
{"train_loss": 0.044024620205163956, "global_step": 237603, "epoch": 2669, "lr": 3.800977766835728e-06}
{"train_loss": 0.029512090608477592, "global_step": 237604, "epoch": 2669, "lr": 3.800756055638094e-06}
{"train_loss": 0.043611783534288406, "global_step": 237605, "epoch": 2669, "lr": 3.8005343506513914e-06}
{"train_loss": 0.08179101347923279, "global_step": 237606, "epoch": 2669, "lr": 3.800312651875637e-06}
{"train_loss": 0.03201979026198387, "global_step": 237607, "epoch": 2669, "lr": 3.8000909593108757e-06}
{"train_loss": 0.027938202023506165, "global_step": 237608, "epoch": 2669, "lr": 3.799869272957118e-06}
{"train_loss": 0.05446772277355194, "global_step": 237609, "epoch": 2669, "lr": 3.799647592814415e-06}
{"train_loss": 0.051663320511579514, "global_step": 237610, "epoch": 2669, "lr": 3.7994259188827762e-06}
{"train_loss": 0.01156670693308115, "global_step": 237611, "epoch": 2669, "lr": 3.799204251162253e-06}
{"train_loss": 0.07959803193807602, "global_step": 237612, "epoch": 2669, "lr": 3.798982589652855e-06}
{"train_loss": 0.033419083803892136, "global_step": 237613, "epoch": 2669, "lr": 3.7987609343546225e-06}
{"train_loss": 0.07655074447393417, "global_step": 237614, "epoch": 2669, "lr": 3.798539285267588e-06}
{"train_loss": 0.06501516699790955, "global_step": 237615, "epoch": 2669, "lr": 3.798317642391769e-06}
{"train_loss": 0.052823055535554886, "global_step": 237616, "epoch": 2669, "lr": 3.798096005727214e-06}
{"train_loss": 0.027974454686045647, "global_step": 237617, "epoch": 2669, "lr": 3.79787437527393e-06}
{"train_loss": 0.07601955533027649, "global_step": 237618, "epoch": 2669, "lr": 3.7976527510319716e-06}
{"train_loss": 0.02628282830119133, "global_step": 237619, "epoch": 2669, "lr": 3.7974311330013447e-06}
{"train_loss": 0.06495119631290436, "global_step": 237620, "epoch": 2669, "lr": 3.7972095211820934e-06}
{"train_loss": 0.08157362043857574, "global_step": 237621, "epoch": 2669, "lr": 3.7969879155742406e-06}
{"train_loss": 0.04687909409403801, "global_step": 237622, "epoch": 2669, "lr": 3.79676631617783e-06}
{"train_loss": 0.061451684683561325, "global_step": 237623, "epoch": 2669, "lr": 3.796544722992873e-06}
{"train_loss": 0.0240860003978014, "global_step": 237624, "epoch": 2669, "lr": 3.7963231360194085e-06}
{"train_loss": 0.03789122402667999, "global_step": 237625, "epoch": 2669, "lr": 3.7961015552574643e-06}
{"train_loss": 0.029506025835871696, "global_step": 237626, "epoch": 2669, "lr": 3.7958799807070787e-06}
{"train_loss": 0.04419226944446564, "global_step": 237627, "epoch": 2669, "lr": 3.7956584123682635e-06}
{"train_loss": 0.06062455102801323, "global_step": 237628, "epoch": 2669, "lr": 3.7954368502410687e-06}
{"train_loss": 0.04885674739947145, "global_step": 237629, "epoch": 2669, "lr": 3.7952152943255104e-06, "val_loss": 9.296049118041992}
{"train_loss": 0.023529227823019028, "global_step": 237630, "epoch": 2670, "lr": 3.7949937446216166e-06}
{"train_loss": 0.037665627896785736, "global_step": 237631, "epoch": 2670, "lr": 3.7947722011294316e-06}
{"train_loss": 0.013598941266536713, "global_step": 237632, "epoch": 2670, "lr": 3.794550663848967e-06}
{"train_loss": 0.030483050271868706, "global_step": 237633, "epoch": 2670, "lr": 3.794329132780272e-06}
{"train_loss": 0.049815498292446136, "global_step": 237634, "epoch": 2670, "lr": 3.7941076079233527e-06}
{"train_loss": 0.07115264981985092, "global_step": 237635, "epoch": 2670, "lr": 3.7938860892782648e-06}
{"train_loss": 0.07645933330059052, "global_step": 237636, "epoch": 2670, "lr": 3.7936645768450187e-06}
{"train_loss": 0.044970955699682236, "global_step": 237637, "epoch": 2670, "lr": 3.793443070623659e-06}
{"train_loss": 0.0954292044043541, "global_step": 237638, "epoch": 2670, "lr": 3.793221570614197e-06}
{"train_loss": 0.024417053908109665, "global_step": 237639, "epoch": 2670, "lr": 3.7930000768166827e-06}
{"train_loss": 0.03713313490152359, "global_step": 237640, "epoch": 2670, "lr": 3.7927785892311274e-06}
{"train_loss": 0.0573728010058403, "global_step": 237641, "epoch": 2670, "lr": 3.7925571078575805e-06}
{"train_loss": 0.07092097401618958, "global_step": 237642, "epoch": 2670, "lr": 3.7923356326960533e-06}
{"train_loss": 0.030277598649263382, "global_step": 237643, "epoch": 2670, "lr": 3.792114163746591e-06}
{"train_loss": 0.05858853831887245, "global_step": 237644, "epoch": 2670, "lr": 3.791892701009203e-06}
{"train_loss": 0.039997585117816925, "global_step": 237645, "epoch": 2670, "lr": 3.7916712444839463e-06}
{"train_loss": 0.06434211879968643, "global_step": 237646, "epoch": 2670, "lr": 3.7914497941708205e-06}
{"train_loss": 0.045590803027153015, "global_step": 237647, "epoch": 2670, "lr": 3.7912283500698865e-06}
{"train_loss": 0.05549275502562523, "global_step": 237648, "epoch": 2670, "lr": 3.7910069121811497e-06}
{"train_loss": 0.05052953213453293, "global_step": 237649, "epoch": 2670, "lr": 3.790785480504644e-06}
{"train_loss": 0.03105362318456173, "global_step": 237650, "epoch": 2670, "lr": 3.790564055040413e-06}
{"train_loss": 0.027448486536741257, "global_step": 237651, "epoch": 2670, "lr": 3.7903426357884743e-06}
{"train_loss": 0.009473971091210842, "global_step": 237652, "epoch": 2670, "lr": 3.7901212227488657e-06}
{"train_loss": 0.03393346443772316, "global_step": 237653, "epoch": 2670, "lr": 3.7898998159216047e-06}
{"train_loss": 0.039645835757255554, "global_step": 237654, "epoch": 2670, "lr": 3.7896784153067357e-06}
{"train_loss": 0.03090827539563179, "global_step": 237655, "epoch": 2670, "lr": 3.7894570209042744e-06}
{"train_loss": 0.02397954650223255, "global_step": 237656, "epoch": 2670, "lr": 3.7892356327142607e-06}
{"train_loss": 0.05658672749996185, "global_step": 237657, "epoch": 2670, "lr": 3.7890142507367165e-06}
{"train_loss": 0.03227207809686661, "global_step": 237658, "epoch": 2670, "lr": 3.788792874971686e-06}
{"train_loss": 0.06950284540653229, "global_step": 237659, "epoch": 2670, "lr": 3.7885715054191804e-06}
{"train_loss": 0.03978308290243149, "global_step": 237660, "epoch": 2670, "lr": 3.7883501420792444e-06}
{"train_loss": 0.04717059060931206, "global_step": 237661, "epoch": 2670, "lr": 3.7881287849518944e-06}
{"train_loss": 0.050472550094127655, "global_step": 237662, "epoch": 2670, "lr": 3.7879074340371744e-06}
{"train_loss": 0.04523129388689995, "global_step": 237663, "epoch": 2670, "lr": 3.787686089335102e-06}
{"train_loss": 0.026211244985461235, "global_step": 237664, "epoch": 2670, "lr": 3.787464750845715e-06}
{"train_loss": 0.06110161542892456, "global_step": 237665, "epoch": 2670, "lr": 3.7872434185690364e-06}
{"train_loss": 0.02523055113852024, "global_step": 237666, "epoch": 2670, "lr": 3.787022092505105e-06}
{"train_loss": 0.11923587322235107, "global_step": 237667, "epoch": 2670, "lr": 3.7868007726539477e-06}
{"train_loss": 0.06417573243379593, "global_step": 237668, "epoch": 2670, "lr": 3.7865794590155825e-06}
{"train_loss": 0.04840976372361183, "global_step": 237669, "epoch": 2670, "lr": 3.786358151590058e-06}
{"train_loss": 0.10861286520957947, "global_step": 237670, "epoch": 2670, "lr": 3.786136850377381e-06}
{"train_loss": 0.06121780723333359, "global_step": 237671, "epoch": 2670, "lr": 3.785915555377606e-06}
{"train_loss": 0.045743025839328766, "global_step": 237672, "epoch": 2670, "lr": 3.7856942665907447e-06}
{"train_loss": 0.07772190123796463, "global_step": 237673, "epoch": 2670, "lr": 3.7854729840168414e-06}
{"train_loss": 0.02579253725707531, "global_step": 237674, "epoch": 2670, "lr": 3.785251707655907e-06}
{"train_loss": 0.039722055196762085, "global_step": 237675, "epoch": 2670, "lr": 3.785030437507997e-06}
{"train_loss": 0.0410805270075798, "global_step": 237676, "epoch": 2670, "lr": 3.7848091735731117e-06}
{"train_loss": 0.026242610067129135, "global_step": 237677, "epoch": 2670, "lr": 3.7845879158513063e-06}
{"train_loss": 0.042298655956983566, "global_step": 237678, "epoch": 2670, "lr": 3.7843666643425926e-06}
{"train_loss": 0.04568000137805939, "global_step": 237679, "epoch": 2670, "lr": 3.7841454190470137e-06}
{"train_loss": 0.04123101755976677, "global_step": 237680, "epoch": 2670, "lr": 3.783924179964582e-06}
{"train_loss": 0.06785671412944794, "global_step": 237681, "epoch": 2670, "lr": 3.7837029470953523e-06}
{"train_loss": 0.0782616063952446, "global_step": 237682, "epoch": 2670, "lr": 3.78348172043933e-06}
{"train_loss": 0.03656913340091705, "global_step": 237683, "epoch": 2670, "lr": 3.7832604999965603e-06}
{"train_loss": 0.01594396121799946, "global_step": 237684, "epoch": 2670, "lr": 3.7830392857670647e-06}
{"train_loss": 0.05022666975855827, "global_step": 237685, "epoch": 2670, "lr": 3.7828180777508717e-06}
{"train_loss": 0.10216428339481354, "global_step": 237686, "epoch": 2670, "lr": 3.78259687594803e-06}
{"train_loss": 0.04738258197903633, "global_step": 237687, "epoch": 2670, "lr": 3.782375680358541e-06}
{"train_loss": 0.07692193984985352, "global_step": 237688, "epoch": 2670, "lr": 3.782154490982459e-06}
{"train_loss": 0.05909878388047218, "global_step": 237689, "epoch": 2670, "lr": 3.7819333078197906e-06}
{"train_loss": 0.05986912548542023, "global_step": 237690, "epoch": 2670, "lr": 3.7817121308705905e-06}
{"train_loss": 0.05974644795060158, "global_step": 237691, "epoch": 2670, "lr": 3.78149096013487e-06}
{"train_loss": 0.051037903875112534, "global_step": 237692, "epoch": 2670, "lr": 3.7812697956126686e-06}
{"train_loss": 0.08971740305423737, "global_step": 237693, "epoch": 2670, "lr": 3.7810486373040023e-06}
{"train_loss": 0.04410365968942642, "global_step": 237694, "epoch": 2670, "lr": 3.7808274852089265e-06}
{"train_loss": 0.06365910917520523, "global_step": 237695, "epoch": 2670, "lr": 3.7806063393274414e-06}
{"train_loss": 0.05567511171102524, "global_step": 237696, "epoch": 2670, "lr": 3.7803851996595973e-06}
{"train_loss": 0.033335618674755096, "global_step": 237697, "epoch": 2670, "lr": 3.7801640662054107e-06}
{"train_loss": 0.05609895661473274, "global_step": 237698, "epoch": 2670, "lr": 3.7799429389649257e-06}
{"train_loss": 0.039068739861249924, "global_step": 237699, "epoch": 2670, "lr": 3.7797218179381587e-06}
{"train_loss": 0.05227503180503845, "global_step": 237700, "epoch": 2670, "lr": 3.7795007031251496e-06}
{"train_loss": 0.037713002413511276, "global_step": 237701, "epoch": 2670, "lr": 3.779279594525914e-06}
{"train_loss": 0.04418335109949112, "global_step": 237702, "epoch": 2670, "lr": 3.7790584921405026e-06}
{"train_loss": 0.09867257624864578, "global_step": 237703, "epoch": 2670, "lr": 3.7788373959689205e-06}
{"train_loss": 0.03119785338640213, "global_step": 237704, "epoch": 2670, "lr": 3.7786163060112234e-06}
{"train_loss": 0.035992782562971115, "global_step": 237705, "epoch": 2670, "lr": 3.7783952222674277e-06}
{"train_loss": 0.0707530677318573, "global_step": 237706, "epoch": 2670, "lr": 3.7781741447375564e-06}
{"train_loss": 0.02753450907766819, "global_step": 237707, "epoch": 2670, "lr": 3.777953073421647e-06}
{"train_loss": 0.061290763318538666, "global_step": 237708, "epoch": 2670, "lr": 3.7777320083197286e-06}
{"train_loss": 0.08064518868923187, "global_step": 237709, "epoch": 2670, "lr": 3.777510949431834e-06}
{"train_loss": 0.03778241202235222, "global_step": 237710, "epoch": 2670, "lr": 3.7772898967579852e-06}
{"train_loss": 0.025445332750678062, "global_step": 237711, "epoch": 2670, "lr": 3.777068850298221e-06}
{"train_loss": 0.048637259751558304, "global_step": 237712, "epoch": 2670, "lr": 3.7768478100525584e-06}
{"train_loss": 0.06552809476852417, "global_step": 237713, "epoch": 2670, "lr": 3.776626776021047e-06}
{"train_loss": 0.04627780243754387, "global_step": 237714, "epoch": 2670, "lr": 3.776405748203693e-06}
{"train_loss": 0.08854640275239944, "global_step": 237715, "epoch": 2670, "lr": 3.7761847266005457e-06}
{"train_loss": 0.05875619128346443, "global_step": 237716, "epoch": 2670, "lr": 3.775963711211622e-06}
{"train_loss": 0.05248057469725609, "global_step": 237717, "epoch": 2670, "lr": 3.7757427020369605e-06}
{"train_loss": 0.0509505606873819, "global_step": 237718, "epoch": 2670, "lr": 3.775521699076584e-06, "val_loss": 9.408333778381348, "train_action_mse_error": 15.451334953308105}
{"train_loss": 0.050071269273757935, "global_step": 237719, "epoch": 2671, "lr": 3.7753007023305308e-06}
{"train_loss": 0.035409264266490936, "global_step": 237720, "epoch": 2671, "lr": 3.775079711798818e-06}
{"train_loss": 0.04920344799757004, "global_step": 237721, "epoch": 2671, "lr": 3.7748587274814785e-06}
{"train_loss": 0.04840031638741493, "global_step": 237722, "epoch": 2671, "lr": 3.7746377493785567e-06}
{"train_loss": 0.04527011513710022, "global_step": 237723, "epoch": 2671, "lr": 3.774416777490064e-06}
{"train_loss": 0.04008777439594269, "global_step": 237724, "epoch": 2671, "lr": 3.7741958118160446e-06}
{"train_loss": 0.046556517481803894, "global_step": 237725, "epoch": 2671, "lr": 3.7739748523565154e-06}
{"train_loss": 0.035391420125961304, "global_step": 237726, "epoch": 2671, "lr": 3.773753899111515e-06}
{"train_loss": 0.03414634242653847, "global_step": 237727, "epoch": 2671, "lr": 3.7735329520810658e-06}
{"train_loss": 0.05280635878443718, "global_step": 237728, "epoch": 2671, "lr": 3.773312011265212e-06}
{"train_loss": 0.019077599048614502, "global_step": 237729, "epoch": 2671, "lr": 3.7730910766639592e-06}
{"train_loss": 0.06053319573402405, "global_step": 237730, "epoch": 2671, "lr": 3.772870148277363e-06}
{"train_loss": 0.057229604572057724, "global_step": 237731, "epoch": 2671, "lr": 3.772649226105429e-06}
{"train_loss": 0.06077716499567032, "global_step": 237732, "epoch": 2671, "lr": 3.7724283101482126e-06}
{"train_loss": 0.04274024814367294, "global_step": 237733, "epoch": 2671, "lr": 3.7722074004057138e-06}
{"train_loss": 0.04705177620053291, "global_step": 237734, "epoch": 2671, "lr": 3.771986496877994e-06}
{"train_loss": 0.01885043829679489, "global_step": 237735, "epoch": 2671, "lr": 3.771765599565058e-06}
{"train_loss": 0.02091451734304428, "global_step": 237736, "epoch": 2671, "lr": 3.771544708466951e-06}
{"train_loss": 0.054495763033628464, "global_step": 237737, "epoch": 2671, "lr": 3.7713238235836835e-06}
{"train_loss": 0.028151268139481544, "global_step": 237738, "epoch": 2671, "lr": 3.7711029449153113e-06}
{"train_loss": 0.028728503733873367, "global_step": 237739, "epoch": 2671, "lr": 3.7708820724618455e-06}
{"train_loss": 0.05832775682210922, "global_step": 237740, "epoch": 2671, "lr": 3.7706612062233247e-06}
{"train_loss": 0.04281303659081459, "global_step": 237741, "epoch": 2671, "lr": 3.7704403461997663e-06}
{"train_loss": 0.03065168298780918, "global_step": 237742, "epoch": 2671, "lr": 3.7702194923912194e-06}
{"train_loss": 0.014361763373017311, "global_step": 237743, "epoch": 2671, "lr": 3.7699986447976955e-06}
{"train_loss": 0.0299842469394207, "global_step": 237744, "epoch": 2671, "lr": 3.769777803419239e-06}
{"train_loss": 0.07421848177909851, "global_step": 237745, "epoch": 2671, "lr": 3.7695569682558718e-06}
{"train_loss": 0.06995619833469391, "global_step": 237746, "epoch": 2671, "lr": 3.7693361393076165e-06}
{"train_loss": 0.05383589491248131, "global_step": 237747, "epoch": 2671, "lr": 3.7691153165745176e-06}
{"train_loss": 0.02372533641755581, "global_step": 237748, "epoch": 2671, "lr": 3.768894500056591e-06}
{"train_loss": 0.011235520243644714, "global_step": 237749, "epoch": 2671, "lr": 3.768673689753882e-06}
{"train_loss": 0.028354326263070107, "global_step": 237750, "epoch": 2671, "lr": 3.7684528856664016e-06}
{"train_loss": 0.046204157173633575, "global_step": 237751, "epoch": 2671, "lr": 3.768232087794199e-06}
{"train_loss": 0.03751569613814354, "global_step": 237752, "epoch": 2671, "lr": 3.7680112961372804e-06}
{"train_loss": 0.04595538601279259, "global_step": 237753, "epoch": 2671, "lr": 3.767790510695701e-06}
{"train_loss": 0.06009410694241524, "global_step": 237754, "epoch": 2671, "lr": 3.7675697314694724e-06}
{"train_loss": 0.03545743599534035, "global_step": 237755, "epoch": 2671, "lr": 3.7673489584586385e-06}
{"train_loss": 0.045136041939258575, "global_step": 237756, "epoch": 2671, "lr": 3.7671281916632105e-06}
{"train_loss": 0.04191700741648674, "global_step": 237757, "epoch": 2671, "lr": 3.7669074310832386e-06}
{"train_loss": 0.066505566239357, "global_step": 237758, "epoch": 2671, "lr": 3.7666866767187335e-06}
{"train_loss": 0.05621080473065376, "global_step": 237759, "epoch": 2671, "lr": 3.7664659285697346e-06}
{"train_loss": 0.030390990898013115, "global_step": 237760, "epoch": 2671, "lr": 3.7662451866362804e-06}
{"train_loss": 0.06153818592429161, "global_step": 237761, "epoch": 2671, "lr": 3.766024450918382e-06}
{"train_loss": 0.07853098213672638, "global_step": 237762, "epoch": 2671, "lr": 3.7658037214160836e-06}
{"train_loss": 0.03107721358537674, "global_step": 237763, "epoch": 2671, "lr": 3.765582998129402e-06}
{"train_loss": 0.06425198167562485, "global_step": 237764, "epoch": 2671, "lr": 3.765362281058382e-06}
{"train_loss": 0.0615960992872715, "global_step": 237765, "epoch": 2671, "lr": 3.76514157020304e-06}
{"train_loss": 0.008323420770466328, "global_step": 237766, "epoch": 2671, "lr": 3.7649208655634204e-06}
{"train_loss": 0.02587265335023403, "global_step": 237767, "epoch": 2671, "lr": 3.7647001671395288e-06}
{"train_loss": 0.03051699697971344, "global_step": 237768, "epoch": 2671, "lr": 3.7644794749314262e-06}
{"train_loss": 0.06962642818689346, "global_step": 237769, "epoch": 2671, "lr": 3.7642587889391123e-06}
{"train_loss": 0.01690293289721012, "global_step": 237770, "epoch": 2671, "lr": 3.7640381091626375e-06}
{"train_loss": 0.04687190428376198, "global_step": 237771, "epoch": 2671, "lr": 3.763817435602018e-06}
{"train_loss": 0.023655923083424568, "global_step": 237772, "epoch": 2671, "lr": 3.7635967682572994e-06}
{"train_loss": 0.04986507073044777, "global_step": 237773, "epoch": 2671, "lr": 3.763376107128491e-06}
{"train_loss": 0.0532417930662632, "global_step": 237774, "epoch": 2671, "lr": 3.7631554522156444e-06}
{"train_loss": 0.037411246448755264, "global_step": 237775, "epoch": 2671, "lr": 3.7629348035187693e-06}
{"train_loss": 0.025633497163653374, "global_step": 237776, "epoch": 2671, "lr": 3.7627141610379e-06}
{"train_loss": 0.08364220708608627, "global_step": 237777, "epoch": 2671, "lr": 3.762493524773081e-06}
{"train_loss": 0.031138112768530846, "global_step": 237778, "epoch": 2671, "lr": 3.7622728947243335e-06}
{"train_loss": 0.09539488703012466, "global_step": 237779, "epoch": 2671, "lr": 3.762052270891675e-06}
{"train_loss": 0.04961739480495453, "global_step": 237780, "epoch": 2671, "lr": 3.7618316532751553e-06}
{"train_loss": 0.09025654196739197, "global_step": 237781, "epoch": 2671, "lr": 3.7616110418747795e-06}
{"train_loss": 0.06451547145843506, "global_step": 237782, "epoch": 2671, "lr": 3.761390436690604e-06}
{"train_loss": 0.050751861184835434, "global_step": 237783, "epoch": 2671, "lr": 3.7611698377226447e-06}
{"train_loss": 0.06620669364929199, "global_step": 237784, "epoch": 2671, "lr": 3.760949244970924e-06}
{"train_loss": 0.0182989239692688, "global_step": 237785, "epoch": 2671, "lr": 3.7607286584354916e-06}
{"train_loss": 0.07355771213769913, "global_step": 237786, "epoch": 2671, "lr": 3.7605080781163536e-06}
{"train_loss": 0.05500153824687004, "global_step": 237787, "epoch": 2671, "lr": 3.76028750401356e-06}
{"train_loss": 0.03234259411692619, "global_step": 237788, "epoch": 2671, "lr": 3.7600669361271267e-06}
{"train_loss": 0.02769983559846878, "global_step": 237789, "epoch": 2671, "lr": 3.759846374457099e-06}
{"train_loss": 0.11895852535963058, "global_step": 237790, "epoch": 2671, "lr": 3.7596258190034814e-06}
{"train_loss": 0.07224872708320618, "global_step": 237791, "epoch": 2671, "lr": 3.7594052697663305e-06}
{"train_loss": 0.049057524651288986, "global_step": 237792, "epoch": 2671, "lr": 3.759184726745657e-06}
{"train_loss": 0.05292246863245964, "global_step": 237793, "epoch": 2671, "lr": 3.758964189941505e-06}
{"train_loss": 0.05544121935963631, "global_step": 237794, "epoch": 2671, "lr": 3.758743659353886e-06}
{"train_loss": 0.055128760635852814, "global_step": 237795, "epoch": 2671, "lr": 3.7585231349828385e-06}
{"train_loss": 0.03288362920284271, "global_step": 237796, "epoch": 2671, "lr": 3.7583026168284074e-06}
{"train_loss": 0.04242021590471268, "global_step": 237797, "epoch": 2671, "lr": 3.758082104890598e-06}
{"train_loss": 0.05739336460828781, "global_step": 237798, "epoch": 2671, "lr": 3.7578615991694597e-06}
{"train_loss": 0.06843650341033936, "global_step": 237799, "epoch": 2671, "lr": 3.7576410996650047e-06}
{"train_loss": 0.05072057619690895, "global_step": 237800, "epoch": 2671, "lr": 3.7574206063772823e-06}
{"train_loss": 0.0569089837372303, "global_step": 237801, "epoch": 2671, "lr": 3.7572001193063035e-06}
{"train_loss": 0.04923468083143234, "global_step": 237802, "epoch": 2671, "lr": 3.756979638452107e-06}
{"train_loss": 0.031811926513910294, "global_step": 237803, "epoch": 2671, "lr": 3.756759163814716e-06}
{"train_loss": 0.04483766108751297, "global_step": 237804, "epoch": 2671, "lr": 3.756538695394174e-06}
{"train_loss": 0.023192940279841423, "global_step": 237805, "epoch": 2671, "lr": 3.7563182331904976e-06}
{"train_loss": 0.061363134533166885, "global_step": 237806, "epoch": 2671, "lr": 3.756097777203721e-06}
{"train_loss": 0.0470195110112931, "global_step": 237807, "epoch": 2671, "lr": 3.755877327433871e-06, "val_loss": 9.342399597167969}
{"train_loss": 0.05366300791501999, "global_step": 237808, "epoch": 2672, "lr": 3.755656883880987e-06}
{"train_loss": 0.03399934619665146, "global_step": 237809, "epoch": 2672, "lr": 3.7554364465450796e-06}
{"train_loss": 0.013057684525847435, "global_step": 237810, "epoch": 2672, "lr": 3.7552160154261997e-06}
{"train_loss": 0.043188538402318954, "global_step": 237811, "epoch": 2672, "lr": 3.754995590524363e-06}
{"train_loss": 0.013134845532476902, "global_step": 237812, "epoch": 2672, "lr": 3.7547751718395975e-06}
{"train_loss": 0.052959978580474854, "global_step": 237813, "epoch": 2672, "lr": 3.7545547593719533e-06}
{"train_loss": 0.055320173501968384, "global_step": 237814, "epoch": 2672, "lr": 3.7543343531214304e-06}
{"train_loss": 0.03512021526694298, "global_step": 237815, "epoch": 2672, "lr": 3.754113953088084e-06}
{"train_loss": 0.047119442373514175, "global_step": 237816, "epoch": 2672, "lr": 3.753893559271937e-06}
{"train_loss": 0.0644952580332756, "global_step": 237817, "epoch": 2672, "lr": 3.7536731716730057e-06}
{"train_loss": 0.030511699616909027, "global_step": 237818, "epoch": 2672, "lr": 3.7534527902913342e-06}
{"train_loss": 0.0188046395778656, "global_step": 237819, "epoch": 2672, "lr": 3.7532324151269447e-06}
{"train_loss": 0.029103856533765793, "global_step": 237820, "epoch": 2672, "lr": 3.753012046179877e-06}
{"train_loss": 0.05137815698981285, "global_step": 237821, "epoch": 2672, "lr": 3.7527916834501462e-06}
{"train_loss": 0.042735323309898376, "global_step": 237822, "epoch": 2672, "lr": 3.7525713269377873e-06}
{"train_loss": 0.05149883031845093, "global_step": 237823, "epoch": 2672, "lr": 3.7523509766428378e-06}
{"train_loss": 0.06675539165735245, "global_step": 237824, "epoch": 2672, "lr": 3.752130632565315e-06}
{"train_loss": 0.032711975276470184, "global_step": 237825, "epoch": 2672, "lr": 3.7519102947052577e-06}
{"train_loss": 0.05150196701288223, "global_step": 237826, "epoch": 2672, "lr": 3.7516899630626878e-06}
{"train_loss": 0.05703748017549515, "global_step": 237827, "epoch": 2672, "lr": 3.7514696376376446e-06}
{"train_loss": 0.0599396750330925, "global_step": 237828, "epoch": 2672, "lr": 3.75124931843015e-06}
{"train_loss": 0.011012580245733261, "global_step": 237829, "epoch": 2672, "lr": 3.7510290054402376e-06}
{"train_loss": 0.10053440183401108, "global_step": 237830, "epoch": 2672, "lr": 3.7508086986679348e-06}
{"train_loss": 0.056654512882232666, "global_step": 237831, "epoch": 2672, "lr": 3.7505883981132637e-06}
{"train_loss": 0.01722039468586445, "global_step": 237832, "epoch": 2672, "lr": 3.750368103776275e-06}
{"train_loss": 0.016671692952513695, "global_step": 237833, "epoch": 2672, "lr": 3.7501478156569793e-06}
{"train_loss": 0.04387109354138374, "global_step": 237834, "epoch": 2672, "lr": 3.7499275337554205e-06}
{"train_loss": 0.031185900792479515, "global_step": 237835, "epoch": 2672, "lr": 3.749707258071611e-06}
{"train_loss": 0.020478706806898117, "global_step": 237836, "epoch": 2672, "lr": 3.749486988605594e-06}
{"train_loss": 0.03089541755616665, "global_step": 237837, "epoch": 2672, "lr": 3.749266725357392e-06}
{"train_loss": 0.033155281096696854, "global_step": 237838, "epoch": 2672, "lr": 3.7490464683270445e-06}
{"train_loss": 0.049035657197237015, "global_step": 237839, "epoch": 2672, "lr": 3.7488262175145616e-06}
{"train_loss": 0.05084146186709404, "global_step": 237840, "epoch": 2672, "lr": 3.74860597292e-06}
{"train_loss": 0.039414238184690475, "global_step": 237841, "epoch": 2672, "lr": 3.748385734543358e-06}
{"train_loss": 0.02907443605363369, "global_step": 237842, "epoch": 2672, "lr": 3.7481655023846986e-06}
{"train_loss": 0.041005488485097885, "global_step": 237843, "epoch": 2672, "lr": 3.7479452764440202e-06}
{"train_loss": 0.020113511011004448, "global_step": 237844, "epoch": 2672, "lr": 3.7477250567213794e-06}
{"train_loss": 0.021935734897851944, "global_step": 237845, "epoch": 2672, "lr": 3.7475048432167816e-06}
{"train_loss": 0.05935068801045418, "global_step": 237846, "epoch": 2672, "lr": 3.7472846359302814e-06}
{"train_loss": 0.02950596809387207, "global_step": 237847, "epoch": 2672, "lr": 3.7470644348618798e-06}
{"train_loss": 0.031794339418411255, "global_step": 237848, "epoch": 2672, "lr": 3.7468442400116266e-06}
{"train_loss": 0.05619289353489876, "global_step": 237849, "epoch": 2672, "lr": 3.7466240513795546e-06}
{"train_loss": 0.02484760619699955, "global_step": 237850, "epoch": 2672, "lr": 3.7464038689656755e-06}
{"train_loss": 0.046882543712854385, "global_step": 237851, "epoch": 2672, "lr": 3.7461836927700334e-06}
{"train_loss": 0.025272827595472336, "global_step": 237852, "epoch": 2672, "lr": 3.7459635227926503e-06}
{"train_loss": 0.028991607949137688, "global_step": 237853, "epoch": 2672, "lr": 3.7457433590335656e-06}
{"train_loss": 0.046574026346206665, "global_step": 237854, "epoch": 2672, "lr": 3.7455232014927955e-06}
{"train_loss": 0.02993527241051197, "global_step": 237855, "epoch": 2672, "lr": 3.7453030501703788e-06}
{"train_loss": 0.042183179408311844, "global_step": 237856, "epoch": 2672, "lr": 3.745082905066344e-06}
{"train_loss": 0.024483153596520424, "global_step": 237857, "epoch": 2672, "lr": 3.744862766180712e-06}
{"train_loss": 0.04662039130926132, "global_step": 237858, "epoch": 2672, "lr": 3.7446426335135286e-06}
{"train_loss": 0.05995463952422142, "global_step": 237859, "epoch": 2672, "lr": 3.74442250706481e-06}
{"train_loss": 0.04310091957449913, "global_step": 237860, "epoch": 2672, "lr": 3.744202386834589e-06}
{"train_loss": 0.0521790087223053, "global_step": 237861, "epoch": 2672, "lr": 3.7439822728229045e-06}
{"train_loss": 0.031630463898181915, "global_step": 237862, "epoch": 2672, "lr": 3.7437621650297625e-06}
{"train_loss": 0.07110608369112015, "global_step": 237863, "epoch": 2672, "lr": 3.7435420634552243e-06}
{"train_loss": 0.02125599980354309, "global_step": 237864, "epoch": 2672, "lr": 3.743321968099289e-06}
{"train_loss": 0.04248970001935959, "global_step": 237865, "epoch": 2672, "lr": 3.7431018789620075e-06}
{"train_loss": 0.047369539737701416, "global_step": 237866, "epoch": 2672, "lr": 3.742881796043396e-06}
{"train_loss": 0.05185356363654137, "global_step": 237867, "epoch": 2672, "lr": 3.7426617193434934e-06}
{"train_loss": 0.0824432447552681, "global_step": 237868, "epoch": 2672, "lr": 3.7424416488623272e-06}
{"train_loss": 0.05581505969166756, "global_step": 237869, "epoch": 2672, "lr": 3.7422215845999254e-06}
{"train_loss": 0.050656478852033615, "global_step": 237870, "epoch": 2672, "lr": 3.7420015265563212e-06}
{"train_loss": 0.03381815552711487, "global_step": 237871, "epoch": 2672, "lr": 3.7417814747315318e-06}
{"train_loss": 0.03812649846076965, "global_step": 237872, "epoch": 2672, "lr": 3.741561429125606e-06}
{"train_loss": 0.06723303347826004, "global_step": 237873, "epoch": 2672, "lr": 3.741341389738556e-06}
{"train_loss": 0.024433156475424767, "global_step": 237874, "epoch": 2672, "lr": 3.741121356570432e-06}
{"train_loss": 0.053469981998205185, "global_step": 237875, "epoch": 2672, "lr": 3.7409013296212327e-06}
{"train_loss": 0.07263831049203873, "global_step": 237876, "epoch": 2672, "lr": 3.7406813088910198e-06}
{"train_loss": 0.0375678725540638, "global_step": 237877, "epoch": 2672, "lr": 3.740461294379799e-06}
{"train_loss": 0.08282824605703354, "global_step": 237878, "epoch": 2672, "lr": 3.7402412860876203e-06}
{"train_loss": 0.07496543228626251, "global_step": 237879, "epoch": 2672, "lr": 3.740021284014489e-06}
{"train_loss": 0.10468866676092148, "global_step": 237880, "epoch": 2672, "lr": 3.739801288160455e-06}
{"train_loss": 0.0686526671051979, "global_step": 237881, "epoch": 2672, "lr": 3.7395812985255406e-06}
{"train_loss": 0.027335435152053833, "global_step": 237882, "epoch": 2672, "lr": 3.7393613151097794e-06}
{"train_loss": 0.047086201608181, "global_step": 237883, "epoch": 2672, "lr": 3.7391413379131877e-06}
{"train_loss": 0.02080601267516613, "global_step": 237884, "epoch": 2672, "lr": 3.73892136693581e-06}
{"train_loss": 0.08817038685083389, "global_step": 237885, "epoch": 2672, "lr": 3.738701402177669e-06}
{"train_loss": 0.028942983597517014, "global_step": 237886, "epoch": 2672, "lr": 3.7384814436387916e-06}
{"train_loss": 0.05313938111066818, "global_step": 237887, "epoch": 2672, "lr": 3.7382614913192226e-06}
{"train_loss": 0.02181553654372692, "global_step": 237888, "epoch": 2672, "lr": 3.7380415452189676e-06}
{"train_loss": 0.03229431435465813, "global_step": 237889, "epoch": 2672, "lr": 3.737821605338082e-06}
{"train_loss": 0.043699245899915695, "global_step": 237890, "epoch": 2672, "lr": 3.7376016716765716e-06}
{"train_loss": 0.08043357729911804, "global_step": 237891, "epoch": 2672, "lr": 3.737381744234486e-06}
{"train_loss": 0.04940620809793472, "global_step": 237892, "epoch": 2672, "lr": 3.737161823011842e-06}
{"train_loss": 0.025274228304624557, "global_step": 237893, "epoch": 2672, "lr": 3.7369419080086733e-06}
{"train_loss": 0.045807186514139175, "global_step": 237894, "epoch": 2672, "lr": 3.7367219992250126e-06}
{"train_loss": 0.07375235855579376, "global_step": 237895, "epoch": 2672, "lr": 3.7365020966608764e-06}
{"train_loss": 0.04465260973011845, "global_step": 237896, "epoch": 2672, "lr": 3.7362822003163155e-06, "val_loss": 9.442244529724121}
{"train_loss": 0.08189761638641357, "global_step": 237897, "epoch": 2673, "lr": 3.736062310191335e-06}
{"train_loss": 0.031137904152274132, "global_step": 237898, "epoch": 2673, "lr": 3.735842426285985e-06}
{"train_loss": 0.032188206911087036, "global_step": 237899, "epoch": 2673, "lr": 3.735622548600287e-06}
{"train_loss": 0.023456331342458725, "global_step": 237900, "epoch": 2673, "lr": 3.735402677134264e-06}
{"train_loss": 0.05987321957945824, "global_step": 237901, "epoch": 2673, "lr": 3.7351828118879606e-06}
{"train_loss": 0.05194113776087761, "global_step": 237902, "epoch": 2673, "lr": 3.734962952861393e-06}
{"train_loss": 0.07235611975193024, "global_step": 237903, "epoch": 2673, "lr": 3.7347431000545884e-06}
{"train_loss": 0.014056527987122536, "global_step": 237904, "epoch": 2673, "lr": 3.7345232534675977e-06}
{"train_loss": 0.04261166602373123, "global_step": 237905, "epoch": 2673, "lr": 3.7343034131004263e-06}
{"train_loss": 0.032232530415058136, "global_step": 237906, "epoch": 2673, "lr": 3.7340835789531237e-06}
{"train_loss": 0.024574365466833115, "global_step": 237907, "epoch": 2673, "lr": 3.7338637510257014e-06}
{"train_loss": 0.04917509853839874, "global_step": 237908, "epoch": 2673, "lr": 3.7336439293182033e-06}
{"train_loss": 0.032907795161008835, "global_step": 237909, "epoch": 2673, "lr": 3.7334241138306414e-06}
{"train_loss": 0.01413560938090086, "global_step": 237910, "epoch": 2673, "lr": 3.7332043045630703e-06}
{"train_loss": 0.058757007122039795, "global_step": 237911, "epoch": 2673, "lr": 3.7329845015154963e-06}
{"train_loss": 0.050124216824769974, "global_step": 237912, "epoch": 2673, "lr": 3.732764704687969e-06}
{"train_loss": 0.03787226974964142, "global_step": 237913, "epoch": 2673, "lr": 3.7325449140804935e-06}
{"train_loss": 0.03368925675749779, "global_step": 237914, "epoch": 2673, "lr": 3.732325129693126e-06}
{"train_loss": 0.048168033361434937, "global_step": 237915, "epoch": 2673, "lr": 3.7321053515258773e-06}
{"train_loss": 0.05244468152523041, "global_step": 237916, "epoch": 2673, "lr": 3.7318855795787866e-06}
{"train_loss": 0.013852723874151707, "global_step": 237917, "epoch": 2673, "lr": 3.7316658138518756e-06}
{"train_loss": 0.04477770999073982, "global_step": 237918, "epoch": 2673, "lr": 3.7314460543451837e-06}
{"train_loss": 0.009330145083367825, "global_step": 237919, "epoch": 2673, "lr": 3.731226301058727e-06}
{"train_loss": 0.05153721943497658, "global_step": 237920, "epoch": 2673, "lr": 3.731006553992555e-06}
{"train_loss": 0.023705214262008667, "global_step": 237921, "epoch": 2673, "lr": 3.7307868131466695e-06}
{"train_loss": 0.05659884959459305, "global_step": 237922, "epoch": 2673, "lr": 3.7305670785211244e-06}
{"train_loss": 0.06473743170499802, "global_step": 237923, "epoch": 2673, "lr": 3.7303473501159424e-06}
{"train_loss": 0.0490693673491478, "global_step": 237924, "epoch": 2673, "lr": 3.730127627931146e-06}
{"train_loss": 0.09945198893547058, "global_step": 237925, "epoch": 2673, "lr": 3.729907911966779e-06}
{"train_loss": 0.03795501962304115, "global_step": 237926, "epoch": 2673, "lr": 3.7296882022228474e-06}
{"train_loss": 0.03309302404522896, "global_step": 237927, "epoch": 2673, "lr": 3.7294684986994124e-06}
{"train_loss": 0.07639826089143753, "global_step": 237928, "epoch": 2673, "lr": 3.7292488013964734e-06}
{"train_loss": 0.03466064855456352, "global_step": 237929, "epoch": 2673, "lr": 3.729029110314081e-06}
{"train_loss": 0.07052106410264969, "global_step": 237930, "epoch": 2673, "lr": 3.7288094254522508e-06}
{"train_loss": 0.03003990463912487, "global_step": 237931, "epoch": 2673, "lr": 3.728589746811023e-06}
{"train_loss": 0.06390471011400223, "global_step": 237932, "epoch": 2673, "lr": 3.728370074390425e-06}
{"train_loss": 0.08043201267719269, "global_step": 237933, "epoch": 2673, "lr": 3.728150408190473e-06}
{"train_loss": 0.07184483855962753, "global_step": 237934, "epoch": 2673, "lr": 3.727930748211217e-06}
{"train_loss": 0.020543061196804047, "global_step": 237935, "epoch": 2673, "lr": 3.7277110944526685e-06}
{"train_loss": 0.06457024067640305, "global_step": 237936, "epoch": 2673, "lr": 3.7274914469148714e-06}
{"train_loss": 0.04520265385508537, "global_step": 237937, "epoch": 2673, "lr": 3.727271805597854e-06}
{"train_loss": 0.08621363341808319, "global_step": 237938, "epoch": 2673, "lr": 3.727052170501627e-06}
{"train_loss": 0.03062283992767334, "global_step": 237939, "epoch": 2673, "lr": 3.7268325416262405e-06}
{"train_loss": 0.05462111160159111, "global_step": 237940, "epoch": 2673, "lr": 3.7266129189717226e-06}
{"train_loss": 0.053699709475040436, "global_step": 237941, "epoch": 2673, "lr": 3.7263933025380837e-06}
{"train_loss": 0.032990533858537674, "global_step": 237942, "epoch": 2673, "lr": 3.7261736923253854e-06}
{"train_loss": 0.034866683185100555, "global_step": 237943, "epoch": 2673, "lr": 3.725954088333622e-06}
{"train_loss": 0.04135733097791672, "global_step": 237944, "epoch": 2673, "lr": 3.7257344905628544e-06}
{"train_loss": 0.028873253613710403, "global_step": 237945, "epoch": 2673, "lr": 3.7255148990130885e-06}
{"train_loss": 0.058805085718631744, "global_step": 237946, "epoch": 2673, "lr": 3.7252953136843683e-06}
{"train_loss": 0.062959223985672, "global_step": 237947, "epoch": 2673, "lr": 3.725075734576711e-06}
{"train_loss": 0.020961707457900047, "global_step": 237948, "epoch": 2673, "lr": 3.72485616169016e-06}
{"train_loss": 0.05514254793524742, "global_step": 237949, "epoch": 2673, "lr": 3.724636595024733e-06}
{"train_loss": 0.06362185627222061, "global_step": 237950, "epoch": 2673, "lr": 3.7244170345804743e-06}
{"train_loss": 0.010300942696630955, "global_step": 237951, "epoch": 2673, "lr": 3.724197480357394e-06}
{"train_loss": 0.02981116995215416, "global_step": 237952, "epoch": 2673, "lr": 3.723977932355538e-06}
{"train_loss": 0.03357807919383049, "global_step": 237953, "epoch": 2673, "lr": 3.7237583905749218e-06}
{"train_loss": 0.06230214238166809, "global_step": 237954, "epoch": 2673, "lr": 3.72353885501559e-06}
{"train_loss": 0.03821178153157234, "global_step": 237955, "epoch": 2673, "lr": 3.7233193256775545e-06}
{"train_loss": 0.0872005820274353, "global_step": 237956, "epoch": 2673, "lr": 3.7230998025608645e-06}
{"train_loss": 0.05165301263332367, "global_step": 237957, "epoch": 2673, "lr": 3.722880285665531e-06}
{"train_loss": 0.05337277054786682, "global_step": 237958, "epoch": 2673, "lr": 3.7226607749915933e-06}
{"train_loss": 0.03625434637069702, "global_step": 237959, "epoch": 2673, "lr": 3.72244127053909e-06}
{"train_loss": 0.056192122399806976, "global_step": 237960, "epoch": 2673, "lr": 3.7222217723080264e-06}
{"train_loss": 0.03867071121931076, "global_step": 237961, "epoch": 2673, "lr": 3.7220022802984587e-06}
{"train_loss": 0.04386864975094795, "global_step": 237962, "epoch": 2673, "lr": 3.721782794510392e-06}
{"train_loss": 0.0638539269566536, "global_step": 237963, "epoch": 2673, "lr": 3.721563314943882e-06}
{"train_loss": 0.03538154438138008, "global_step": 237964, "epoch": 2673, "lr": 3.721343841598929e-06}
{"train_loss": 0.037660371512174606, "global_step": 237965, "epoch": 2673, "lr": 3.721124374475593e-06}
{"train_loss": 0.058688439428806305, "global_step": 237966, "epoch": 2673, "lr": 3.7209049135738747e-06}
{"train_loss": 0.05333741754293442, "global_step": 237967, "epoch": 2673, "lr": 3.7206854588938245e-06}
{"train_loss": 0.06492546200752258, "global_step": 237968, "epoch": 2673, "lr": 3.720466010435453e-06}
{"train_loss": 0.015154484659433365, "global_step": 237969, "epoch": 2673, "lr": 3.720246568198815e-06}
{"train_loss": 0.04090554639697075, "global_step": 237970, "epoch": 2673, "lr": 3.720027132183912e-06}
{"train_loss": 0.07010471820831299, "global_step": 237971, "epoch": 2673, "lr": 3.7198077023907983e-06}
{"train_loss": 0.03567304089665413, "global_step": 237972, "epoch": 2673, "lr": 3.7195882788194967e-06}
{"train_loss": 0.047170355916023254, "global_step": 237973, "epoch": 2673, "lr": 3.7193688614700185e-06}
{"train_loss": 0.02902042493224144, "global_step": 237974, "epoch": 2673, "lr": 3.7191494503424183e-06}
{"train_loss": 0.05391976609826088, "global_step": 237975, "epoch": 2673, "lr": 3.7189300454367028e-06}
{"train_loss": 0.03881858289241791, "global_step": 237976, "epoch": 2673, "lr": 3.7187106467529207e-06}
{"train_loss": 0.04930393397808075, "global_step": 237977, "epoch": 2673, "lr": 3.71849125429109e-06}
{"train_loss": 0.05311044305562973, "global_step": 237978, "epoch": 2673, "lr": 3.7182718680512485e-06}
{"train_loss": 0.04974906146526337, "global_step": 237979, "epoch": 2673, "lr": 3.718052488033419e-06}
{"train_loss": 0.011136941611766815, "global_step": 237980, "epoch": 2673, "lr": 3.7178331142376345e-06}
{"train_loss": 0.016889194026589394, "global_step": 237981, "epoch": 2673, "lr": 3.717613746663917e-06}
{"train_loss": 0.023782454431056976, "global_step": 237982, "epoch": 2673, "lr": 3.7173943853123117e-06}
{"train_loss": 0.037247199565172195, "global_step": 237983, "epoch": 2673, "lr": 3.717175030182829e-06}
{"train_loss": 0.03218713775277138, "global_step": 237984, "epoch": 2673, "lr": 3.7169556812755193e-06}
{"train_loss": 0.045137763337305424, "global_step": 237985, "epoch": 2673, "lr": 3.7167363385903932e-06, "val_loss": 9.373842239379883}
{"train_loss": 0.032905999571084976, "global_step": 237986, "epoch": 2674, "lr": 3.71651700212749e-06}
{"train_loss": 0.03303542360663414, "global_step": 237987, "epoch": 2674, "lr": 3.7162976718868312e-06}
{"train_loss": 0.06979231536388397, "global_step": 237988, "epoch": 2674, "lr": 3.7160783478684622e-06}
{"train_loss": 0.03833966329693794, "global_step": 237989, "epoch": 2674, "lr": 3.715859030072394e-06}
{"train_loss": 0.05096295103430748, "global_step": 237990, "epoch": 2674, "lr": 3.7156397184986702e-06}
{"train_loss": 0.04178626835346222, "global_step": 237991, "epoch": 2674, "lr": 3.715420413147308e-06}
{"train_loss": 0.025527413934469223, "global_step": 237992, "epoch": 2674, "lr": 3.715201114018352e-06}
{"train_loss": 0.024280034005641937, "global_step": 237993, "epoch": 2674, "lr": 3.7149818211118126e-06}
{"train_loss": 0.03164549916982651, "global_step": 237994, "epoch": 2674, "lr": 3.714762534427735e-06}
{"train_loss": 0.026092102751135826, "global_step": 237995, "epoch": 2674, "lr": 3.7145432539661464e-06}
{"train_loss": 0.04618469253182411, "global_step": 237996, "epoch": 2674, "lr": 3.714323979727069e-06}
{"train_loss": 0.024854876101017, "global_step": 237997, "epoch": 2674, "lr": 3.714104711710542e-06}
{"train_loss": 0.11546607315540314, "global_step": 237998, "epoch": 2674, "lr": 3.713885449916582e-06}
{"train_loss": 0.01686697266995907, "global_step": 237999, "epoch": 2674, "lr": 3.7136661943452334e-06}
{"train_loss": 0.033771395683288574, "global_step": 238000, "epoch": 2674, "lr": 3.713446944996507e-06}
{"train_loss": 0.04543868079781532, "global_step": 238001, "epoch": 2674, "lr": 3.713227701870459e-06}
{"train_loss": 0.04374171048402786, "global_step": 238002, "epoch": 2674, "lr": 3.713008464967088e-06}
{"train_loss": 0.02081315964460373, "global_step": 238003, "epoch": 2674, "lr": 3.7127892342864513e-06}
{"train_loss": 0.020627610385417938, "global_step": 238004, "epoch": 2674, "lr": 3.712570009828559e-06}
{"train_loss": 0.06090712547302246, "global_step": 238005, "epoch": 2674, "lr": 3.712350791593455e-06}
{"train_loss": 0.029317662119865417, "global_step": 238006, "epoch": 2674, "lr": 3.7121315795811517e-06}
{"train_loss": 0.03870241343975067, "global_step": 238007, "epoch": 2674, "lr": 3.7119123737916928e-06}
{"train_loss": 0.059997521340847015, "global_step": 238008, "epoch": 2674, "lr": 3.7116931742251003e-06}
{"train_loss": 0.04085025191307068, "global_step": 238009, "epoch": 2674, "lr": 3.7114739808814136e-06}
{"train_loss": 0.05908101424574852, "global_step": 238010, "epoch": 2674, "lr": 3.7112547937606544e-06}
{"train_loss": 0.05643641576170921, "global_step": 238011, "epoch": 2674, "lr": 3.7110356128628455e-06}
{"train_loss": 0.060018666088581085, "global_step": 238012, "epoch": 2674, "lr": 3.7108164381880305e-06}
{"train_loss": 0.06356094032526016, "global_step": 238013, "epoch": 2674, "lr": 3.7105972697362214e-06}
{"train_loss": 0.02594912052154541, "global_step": 238014, "epoch": 2674, "lr": 3.7103781075074727e-06}
{"train_loss": 0.054920557886362076, "global_step": 238015, "epoch": 2674, "lr": 3.710158951501785e-06}
{"train_loss": 0.052477046847343445, "global_step": 238016, "epoch": 2674, "lr": 3.7099398017192145e-06}
{"train_loss": 0.04768810793757439, "global_step": 238017, "epoch": 2674, "lr": 3.709720658159771e-06}
{"train_loss": 0.03793654218316078, "global_step": 238018, "epoch": 2674, "lr": 3.7095015208234996e-06}
{"train_loss": 0.0548635870218277, "global_step": 238019, "epoch": 2674, "lr": 3.7092823897104114e-06}
{"train_loss": 0.09466531872749329, "global_step": 238020, "epoch": 2674, "lr": 3.7090632648205504e-06}
{"train_loss": 0.045122336596250534, "global_step": 238021, "epoch": 2674, "lr": 3.708844146153939e-06}
{"train_loss": 0.0334368571639061, "global_step": 238022, "epoch": 2674, "lr": 3.7086250337106166e-06}
{"train_loss": 0.030567560344934464, "global_step": 238023, "epoch": 2674, "lr": 3.7084059274905934e-06}
{"train_loss": 0.02920442633330822, "global_step": 238024, "epoch": 2674, "lr": 3.7081868274939258e-06}
{"train_loss": 0.04212300851941109, "global_step": 238025, "epoch": 2674, "lr": 3.707967733720613e-06}
{"train_loss": 0.021286576986312866, "global_step": 238026, "epoch": 2674, "lr": 3.707748646170711e-06}
{"train_loss": 0.03421426936984062, "global_step": 238027, "epoch": 2674, "lr": 3.707529564844231e-06}
{"train_loss": 0.11379622668027878, "global_step": 238028, "epoch": 2674, "lr": 3.707310489741217e-06}
{"train_loss": 0.02005232870578766, "global_step": 238029, "epoch": 2674, "lr": 3.70709142086168e-06}
{"train_loss": 0.0481543093919754, "global_step": 238030, "epoch": 2674, "lr": 3.7068723582056655e-06}
{"train_loss": 0.056987155228853226, "global_step": 238031, "epoch": 2674, "lr": 3.7066533017731995e-06}
{"train_loss": 0.047782644629478455, "global_step": 238032, "epoch": 2674, "lr": 3.7064342515643057e-06}
{"train_loss": 0.05382057651877403, "global_step": 238033, "epoch": 2674, "lr": 3.706215207579028e-06}
{"train_loss": 0.03000018559396267, "global_step": 238034, "epoch": 2674, "lr": 3.7059961698173717e-06}
{"train_loss": 0.07067346572875977, "global_step": 238035, "epoch": 2674, "lr": 3.7057771382793926e-06}
{"train_loss": 0.027614757418632507, "global_step": 238036, "epoch": 2674, "lr": 3.7055581129650965e-06}
{"train_loss": 0.07253342866897583, "global_step": 238037, "epoch": 2674, "lr": 3.7053390938745325e-06}
{"train_loss": 0.0317610464990139, "global_step": 238038, "epoch": 2674, "lr": 3.7051200810077124e-06}
{"train_loss": 0.027214763686060905, "global_step": 238039, "epoch": 2674, "lr": 3.704901074364686e-06}
{"train_loss": 0.06447628885507584, "global_step": 238040, "epoch": 2674, "lr": 3.704682073945459e-06}
{"train_loss": 0.05666687339544296, "global_step": 238041, "epoch": 2674, "lr": 3.704463079750087e-06}
{"train_loss": 0.0565633550286293, "global_step": 238042, "epoch": 2674, "lr": 3.7042440917785694e-06}
{"train_loss": 0.04621633142232895, "global_step": 238043, "epoch": 2674, "lr": 3.7040251100309674e-06}
{"train_loss": 0.06468342989683151, "global_step": 238044, "epoch": 2674, "lr": 3.7038061345072872e-06}
{"train_loss": 0.024641703814268112, "global_step": 238045, "epoch": 2674, "lr": 3.703587165207567e-06}
{"train_loss": 0.0378655381500721, "global_step": 238046, "epoch": 2674, "lr": 3.703368202131835e-06}
{"train_loss": 0.012008271180093288, "global_step": 238047, "epoch": 2674, "lr": 3.7031492452801244e-06}
{"train_loss": 0.06265193969011307, "global_step": 238048, "epoch": 2674, "lr": 3.7029302946524624e-06}
{"train_loss": 0.07955151051282883, "global_step": 238049, "epoch": 2674, "lr": 3.7027113502488663e-06}
{"train_loss": 0.028424907475709915, "global_step": 238050, "epoch": 2674, "lr": 3.702492412069386e-06}
{"train_loss": 0.04251130297780037, "global_step": 238051, "epoch": 2674, "lr": 3.7022734801140378e-06}
{"train_loss": 0.05507993698120117, "global_step": 238052, "epoch": 2674, "lr": 3.7020545543828557e-06}
{"train_loss": 0.048593632876873016, "global_step": 238053, "epoch": 2674, "lr": 3.7018356348758665e-06}
{"train_loss": 0.08246491849422455, "global_step": 238054, "epoch": 2674, "lr": 3.701616721593104e-06}
{"train_loss": 0.04261742904782295, "global_step": 238055, "epoch": 2674, "lr": 3.7013978145345906e-06}
{"train_loss": 0.11056765913963318, "global_step": 238056, "epoch": 2674, "lr": 3.701178913700365e-06}
{"train_loss": 0.02066764608025551, "global_step": 238057, "epoch": 2674, "lr": 3.7009600190904438e-06}
{"train_loss": 0.038008518517017365, "global_step": 238058, "epoch": 2674, "lr": 3.7007411307048766e-06}
{"train_loss": 0.04954635351896286, "global_step": 238059, "epoch": 2674, "lr": 3.700522248543664e-06}
{"train_loss": 0.04710671305656433, "global_step": 238060, "epoch": 2674, "lr": 3.700303372606867e-06}
{"train_loss": 0.03452395275235176, "global_step": 238061, "epoch": 2674, "lr": 3.7000845028944908e-06}
{"train_loss": 0.011510665528476238, "global_step": 238062, "epoch": 2674, "lr": 3.69986563940658e-06}
{"train_loss": 0.03523597866296768, "global_step": 238063, "epoch": 2674, "lr": 3.6996467821431513e-06}
{"train_loss": 0.05648581311106682, "global_step": 238064, "epoch": 2674, "lr": 3.699427931104249e-06}
{"train_loss": 0.06743729114532471, "global_step": 238065, "epoch": 2674, "lr": 3.699209086289884e-06}
{"train_loss": 0.043484434485435486, "global_step": 238066, "epoch": 2674, "lr": 3.6989902477001015e-06}
{"train_loss": 0.03858385607600212, "global_step": 238067, "epoch": 2674, "lr": 3.6987714153349285e-06}
{"train_loss": 0.04927621781826019, "global_step": 238068, "epoch": 2674, "lr": 3.6985525891943817e-06}
{"train_loss": 0.014582633040845394, "global_step": 238069, "epoch": 2674, "lr": 3.6983337692785114e-06}
{"train_loss": 0.0678829625248909, "global_step": 238070, "epoch": 2674, "lr": 3.6981149555873284e-06}
{"train_loss": 0.03565378859639168, "global_step": 238071, "epoch": 2674, "lr": 3.6978961481208773e-06}
{"train_loss": 0.046910133212804794, "global_step": 238072, "epoch": 2674, "lr": 3.6976773468791693e-06}
{"train_loss": 0.05680466443300247, "global_step": 238073, "epoch": 2674, "lr": 3.697458551862254e-06}
{"train_loss": 0.04611335873729392, "global_step": 238074, "epoch": 2674, "lr": 3.6972397630701427e-06, "val_loss": 9.3618803024292}
{"train_loss": 0.03695264086127281, "global_step": 238075, "epoch": 2675, "lr": 3.69702098050288e-06}
{"train_loss": 0.049587544053792953, "global_step": 238076, "epoch": 2675, "lr": 3.696802204160482e-06}
{"train_loss": 0.04732627049088478, "global_step": 238077, "epoch": 2675, "lr": 3.696583434042994e-06}
{"train_loss": 0.06557967513799667, "global_step": 238078, "epoch": 2675, "lr": 3.696364670150426e-06}
{"train_loss": 0.03643321618437767, "global_step": 238079, "epoch": 2675, "lr": 3.696145912482829e-06}
{"train_loss": 0.03979804739356041, "global_step": 238080, "epoch": 2675, "lr": 3.695927161040208e-06}
{"train_loss": 0.04757603630423546, "global_step": 238081, "epoch": 2675, "lr": 3.695708415822613e-06}
{"train_loss": 0.03958451747894287, "global_step": 238082, "epoch": 2675, "lr": 3.695489676830061e-06}
{"train_loss": 0.042217329144477844, "global_step": 238083, "epoch": 2675, "lr": 3.69527094406259e-06}
{"train_loss": 0.05083637684583664, "global_step": 238084, "epoch": 2675, "lr": 3.695052217520223e-06}
{"train_loss": 0.03721579909324646, "global_step": 238085, "epoch": 2675, "lr": 3.694833497202993e-06}
{"train_loss": 0.03995738923549652, "global_step": 238086, "epoch": 2675, "lr": 3.694614783110928e-06}
{"train_loss": 0.058750297874212265, "global_step": 238087, "epoch": 2675, "lr": 3.694396075244061e-06}
{"train_loss": 0.059152599424123764, "global_step": 238088, "epoch": 2675, "lr": 3.6941773736024197e-06}
{"train_loss": 0.042806267738342285, "global_step": 238089, "epoch": 2675, "lr": 3.6939586781860215e-06}
{"train_loss": 0.04944882169365883, "global_step": 238090, "epoch": 2675, "lr": 3.6937399889949153e-06}
{"train_loss": 0.08754727989435196, "global_step": 238091, "epoch": 2675, "lr": 3.693521306029113e-06}
{"train_loss": 0.04476248845458031, "global_step": 238092, "epoch": 2675, "lr": 3.693302629288664e-06}
{"train_loss": 0.03800225257873535, "global_step": 238093, "epoch": 2675, "lr": 3.6930839587735743e-06}
{"train_loss": 0.02949433960020542, "global_step": 238094, "epoch": 2675, "lr": 3.6928652944838938e-06}
{"train_loss": 0.03555073216557503, "global_step": 238095, "epoch": 2675, "lr": 3.6926466364196335e-06}
{"train_loss": 0.015260614454746246, "global_step": 238096, "epoch": 2675, "lr": 3.692427984580843e-06}
{"train_loss": 0.08665178716182709, "global_step": 238097, "epoch": 2675, "lr": 3.6922093389675337e-06}
{"train_loss": 0.042579542845487595, "global_step": 238098, "epoch": 2675, "lr": 3.6919906995797447e-06}
{"train_loss": 0.04311751946806908, "global_step": 238099, "epoch": 2675, "lr": 3.691772066417504e-06}
{"train_loss": 0.04072188585996628, "global_step": 238100, "epoch": 2675, "lr": 3.691553439480844e-06}
{"train_loss": 0.01627466082572937, "global_step": 238101, "epoch": 2675, "lr": 3.6913348187697814e-06}
{"train_loss": 0.04714075103402138, "global_step": 238102, "epoch": 2675, "lr": 3.6911162042843507e-06}
{"train_loss": 0.09276966005563736, "global_step": 238103, "epoch": 2675, "lr": 3.6908975960246007e-06}
{"train_loss": 0.05844224616885185, "global_step": 238104, "epoch": 2675, "lr": 3.6906789939905318e-06}
{"train_loss": 0.04836658015847206, "global_step": 238105, "epoch": 2675, "lr": 3.690460398182194e-06}
{"train_loss": 0.04224937781691551, "global_step": 238106, "epoch": 2675, "lr": 3.690241808599604e-06}
{"train_loss": 0.036464937031269073, "global_step": 238107, "epoch": 2675, "lr": 3.690023225242806e-06}
{"train_loss": 0.01804194226861, "global_step": 238108, "epoch": 2675, "lr": 3.689804648111811e-06}
{"train_loss": 0.0307321697473526, "global_step": 238109, "epoch": 2675, "lr": 3.6895860772066638e-06}
{"train_loss": 0.024661442264914513, "global_step": 238110, "epoch": 2675, "lr": 3.6893675125273806e-06}
{"train_loss": 0.03228985518217087, "global_step": 238111, "epoch": 2675, "lr": 3.6891489540740064e-06}
{"train_loss": 0.020082848146557808, "global_step": 238112, "epoch": 2675, "lr": 3.688930401846552e-06}
{"train_loss": 0.037898674607276917, "global_step": 238113, "epoch": 2675, "lr": 3.6887118558450617e-06}
{"train_loss": 0.09843103587627411, "global_step": 238114, "epoch": 2675, "lr": 3.6884933160695577e-06}
{"train_loss": 0.04264155030250549, "global_step": 238115, "epoch": 2675, "lr": 3.6882747825200737e-06}
{"train_loss": 0.0811622366309166, "global_step": 238116, "epoch": 2675, "lr": 3.688056255196631e-06}
{"train_loss": 0.05561748519539833, "global_step": 238117, "epoch": 2675, "lr": 3.687837734099275e-06}
{"train_loss": 0.07945304363965988, "global_step": 238118, "epoch": 2675, "lr": 3.687619219228011e-06}
{"train_loss": 0.025150585919618607, "global_step": 238119, "epoch": 2675, "lr": 3.6874007105828945e-06}
{"train_loss": 0.03988174349069595, "global_step": 238120, "epoch": 2675, "lr": 3.6871822081639363e-06}
{"train_loss": 0.08833522349596024, "global_step": 238121, "epoch": 2675, "lr": 3.686963711971175e-06}
{"train_loss": 0.06273149698972702, "global_step": 238122, "epoch": 2675, "lr": 3.686745222004634e-06}
{"train_loss": 0.049165721982717514, "global_step": 238123, "epoch": 2675, "lr": 3.686526738264351e-06}
{"train_loss": 0.029618309810757637, "global_step": 238124, "epoch": 2675, "lr": 3.686308260750343e-06}
{"train_loss": 0.03424639254808426, "global_step": 238125, "epoch": 2675, "lr": 3.6860897894626543e-06}
{"train_loss": 0.05565609410405159, "global_step": 238126, "epoch": 2675, "lr": 3.685871324401302e-06}
{"train_loss": 0.05540977418422699, "global_step": 238127, "epoch": 2675, "lr": 3.685652865566319e-06}
{"train_loss": 0.08267908543348312, "global_step": 238128, "epoch": 2675, "lr": 3.685434412957739e-06}
{"train_loss": 0.03202163800597191, "global_step": 238129, "epoch": 2675, "lr": 3.685215966575578e-06}
{"train_loss": 0.017275305464863777, "global_step": 238130, "epoch": 2675, "lr": 3.6849975264198865e-06}
{"train_loss": 0.05273556709289551, "global_step": 238131, "epoch": 2675, "lr": 3.6847790924906754e-06}
{"train_loss": 0.04020606726408005, "global_step": 238132, "epoch": 2675, "lr": 3.6845606647879895e-06}
{"train_loss": 0.059313736855983734, "global_step": 238133, "epoch": 2675, "lr": 3.684342243311839e-06}
{"train_loss": 0.06340838223695755, "global_step": 238134, "epoch": 2675, "lr": 3.6841238280622693e-06}
{"train_loss": 0.010937768965959549, "global_step": 238135, "epoch": 2675, "lr": 3.6839054190393018e-06}
{"train_loss": 0.0312545970082283, "global_step": 238136, "epoch": 2675, "lr": 3.683687016242976e-06}
{"train_loss": 0.022477779537439346, "global_step": 238137, "epoch": 2675, "lr": 3.6834686196733084e-06}
{"train_loss": 0.0662967711687088, "global_step": 238138, "epoch": 2675, "lr": 3.683250229330337e-06}
{"train_loss": 0.0546535886824131, "global_step": 238139, "epoch": 2675, "lr": 3.68303184521408e-06}
{"train_loss": 0.05727114528417587, "global_step": 238140, "epoch": 2675, "lr": 3.682813467324581e-06}
{"train_loss": 0.06872117519378662, "global_step": 238141, "epoch": 2675, "lr": 3.682595095661867e-06}
{"train_loss": 0.02939184010028839, "global_step": 238142, "epoch": 2675, "lr": 3.682376730225956e-06}
{"train_loss": 0.04249666631221771, "global_step": 238143, "epoch": 2675, "lr": 3.6821583710168915e-06}
{"train_loss": 0.030896728858351707, "global_step": 238144, "epoch": 2675, "lr": 3.6819400180346906e-06}
{"train_loss": 0.058573413640260696, "global_step": 238145, "epoch": 2675, "lr": 3.6817216712793923e-06}
{"train_loss": 0.03797198086977005, "global_step": 238146, "epoch": 2675, "lr": 3.681503330751018e-06}
{"train_loss": 0.08788108080625534, "global_step": 238147, "epoch": 2675, "lr": 3.6812849964496076e-06}
{"train_loss": 0.052444279193878174, "global_step": 238148, "epoch": 2675, "lr": 3.681066668375177e-06}
{"train_loss": 0.0627049058675766, "global_step": 238149, "epoch": 2675, "lr": 3.680848346527771e-06}
{"train_loss": 0.07505448162555695, "global_step": 238150, "epoch": 2675, "lr": 3.6806300309074004e-06}
{"train_loss": 0.023021992295980453, "global_step": 238151, "epoch": 2675, "lr": 3.68041172151411e-06}
{"train_loss": 0.020991502329707146, "global_step": 238152, "epoch": 2675, "lr": 3.6801934183479213e-06}
{"train_loss": 0.028470421209931374, "global_step": 238153, "epoch": 2675, "lr": 3.6799751214088685e-06}
{"train_loss": 0.06796260178089142, "global_step": 238154, "epoch": 2675, "lr": 3.6797568306969732e-06}
{"train_loss": 0.02585524506866932, "global_step": 238155, "epoch": 2675, "lr": 3.67953854621228e-06}
{"train_loss": 0.05072302371263504, "global_step": 238156, "epoch": 2675, "lr": 3.6793202679548e-06}
{"train_loss": 0.054425131529569626, "global_step": 238157, "epoch": 2675, "lr": 3.6791019959245663e-06}
{"train_loss": 0.020936716347932816, "global_step": 238158, "epoch": 2675, "lr": 3.678883730121624e-06}
{"train_loss": 0.050714705139398575, "global_step": 238159, "epoch": 2675, "lr": 3.6786654705459944e-06}
{"train_loss": 0.02577579766511917, "global_step": 238160, "epoch": 2675, "lr": 3.6784472171976893e-06}
{"train_loss": 0.07066052407026291, "global_step": 238161, "epoch": 2675, "lr": 3.6782289700767637e-06}
{"train_loss": 0.030941767618060112, "global_step": 238162, "epoch": 2675, "lr": 3.678010729183229e-06}
{"train_loss": 0.046744173706582424, "global_step": 238163, "epoch": 2675, "lr": 3.6777924945171295e-06, "val_loss": 9.429739952087402, "train_action_mse_error": 10.332440376281738}
{"train_loss": 0.013844598084688187, "global_step": 238164, "epoch": 2676, "lr": 3.6775742660784816e-06}
{"train_loss": 0.0440385602414608, "global_step": 238165, "epoch": 2676, "lr": 3.677356043867314e-06}
{"train_loss": 0.03978150710463524, "global_step": 238166, "epoch": 2676, "lr": 3.67713782788367e-06}
{"train_loss": 0.035691998898983, "global_step": 238167, "epoch": 2676, "lr": 3.676919618127561e-06}
{"train_loss": 0.030956976115703583, "global_step": 238168, "epoch": 2676, "lr": 3.6767014145990375e-06}
{"train_loss": 0.030922655016183853, "global_step": 238169, "epoch": 2676, "lr": 3.6764832172981046e-06}
{"train_loss": 0.03886823728680611, "global_step": 238170, "epoch": 2676, "lr": 3.676265026224812e-06}
{"train_loss": 0.06565480679273605, "global_step": 238171, "epoch": 2676, "lr": 3.6760468413791714e-06}
{"train_loss": 0.03432676941156387, "global_step": 238172, "epoch": 2676, "lr": 3.6758286627612328e-06}
{"train_loss": 0.03424442932009697, "global_step": 238173, "epoch": 2676, "lr": 3.6756104903710065e-06}
{"train_loss": 0.04715978354215622, "global_step": 238174, "epoch": 2676, "lr": 3.6753923242085375e-06}
{"train_loss": 0.0536951944231987, "global_step": 238175, "epoch": 2676, "lr": 3.6751741642738367e-06}
{"train_loss": 0.035724982619285583, "global_step": 238176, "epoch": 2676, "lr": 3.674956010566949e-06}
{"train_loss": 0.02530355006456375, "global_step": 238177, "epoch": 2676, "lr": 3.6747378630879017e-06}
{"train_loss": 0.06678013503551483, "global_step": 238178, "epoch": 2676, "lr": 3.6745197218367165e-06}
{"train_loss": 0.023137111216783524, "global_step": 238179, "epoch": 2676, "lr": 3.6743015868134334e-06}
{"train_loss": 0.11142397671937943, "global_step": 238180, "epoch": 2676, "lr": 3.674083458018068e-06}
{"train_loss": 0.07188217341899872, "global_step": 238181, "epoch": 2676, "lr": 3.673865335450666e-06}
{"train_loss": 0.056589823216199875, "global_step": 238182, "epoch": 2676, "lr": 3.6736472191112427e-06}
{"train_loss": 0.04022722318768501, "global_step": 238183, "epoch": 2676, "lr": 3.6734291089998372e-06}
{"train_loss": 0.026422938331961632, "global_step": 238184, "epoch": 2676, "lr": 3.673211005116467e-06}
{"train_loss": 0.05666777864098549, "global_step": 238185, "epoch": 2676, "lr": 3.672992907461181e-06}
{"train_loss": 0.047938644886016846, "global_step": 238186, "epoch": 2676, "lr": 3.672774816033986e-06}
{"train_loss": 0.07112372666597366, "global_step": 238187, "epoch": 2676, "lr": 3.6725567308349307e-06}
{"train_loss": 0.07240178436040878, "global_step": 238188, "epoch": 2676, "lr": 3.672338651864021e-06}
{"train_loss": 0.06867072731256485, "global_step": 238189, "epoch": 2676, "lr": 3.6721205791213186e-06}
{"train_loss": 0.04116338491439819, "global_step": 238190, "epoch": 2676, "lr": 3.6719025126068175e-06}
{"train_loss": 0.04135115072131157, "global_step": 238191, "epoch": 2676, "lr": 3.6716844523205785e-06}
{"train_loss": 0.026976022869348526, "global_step": 238192, "epoch": 2676, "lr": 3.6714663982626075e-06}
{"train_loss": 0.021461429074406624, "global_step": 238193, "epoch": 2676, "lr": 3.671248350432943e-06}
{"train_loss": 0.10353999584913254, "global_step": 238194, "epoch": 2676, "lr": 3.6710303088316245e-06}
{"train_loss": 0.07513513416051865, "global_step": 238195, "epoch": 2676, "lr": 3.6708122734586624e-06}
{"train_loss": 0.04415915906429291, "global_step": 238196, "epoch": 2676, "lr": 3.6705942443141018e-06}
{"train_loss": 0.02231545001268387, "global_step": 238197, "epoch": 2676, "lr": 3.670376221397953e-06}
{"train_loss": 0.05490221828222275, "global_step": 238198, "epoch": 2676, "lr": 3.6701582047102723e-06}
{"train_loss": 0.06007678806781769, "global_step": 238199, "epoch": 2676, "lr": 3.6699401942510704e-06}
{"train_loss": 0.060199074447155, "global_step": 238200, "epoch": 2676, "lr": 3.669722190020375e-06}
{"train_loss": 0.02565961703658104, "global_step": 238201, "epoch": 2676, "lr": 3.669504192018225e-06}
{"train_loss": 0.04590268060564995, "global_step": 238202, "epoch": 2676, "lr": 3.6692862002446427e-06}
{"train_loss": 0.02398851327598095, "global_step": 238203, "epoch": 2676, "lr": 3.6690682146996613e-06}
{"train_loss": 0.03791388124227524, "global_step": 238204, "epoch": 2676, "lr": 3.6688502353833143e-06}
{"train_loss": 0.053977955132722855, "global_step": 238205, "epoch": 2676, "lr": 3.668632262295618e-06}
{"train_loss": 0.0442807711660862, "global_step": 238206, "epoch": 2676, "lr": 3.6684142954366173e-06}
{"train_loss": 0.04634392634034157, "global_step": 238207, "epoch": 2676, "lr": 3.668196334806323e-06}
{"train_loss": 0.06433220952749252, "global_step": 238208, "epoch": 2676, "lr": 3.6679783804047795e-06}
{"train_loss": 0.05331497639417648, "global_step": 238209, "epoch": 2676, "lr": 3.667760432232009e-06}
{"train_loss": 0.028981104493141174, "global_step": 238210, "epoch": 2676, "lr": 3.6675424902880507e-06}
{"train_loss": 0.07004418224096298, "global_step": 238211, "epoch": 2676, "lr": 3.6673245545729207e-06}
{"train_loss": 0.040761783719062805, "global_step": 238212, "epoch": 2676, "lr": 3.6671066250866525e-06}
{"train_loss": 0.07998710870742798, "global_step": 238213, "epoch": 2676, "lr": 3.666888701829285e-06}
{"train_loss": 0.05162686109542847, "global_step": 238214, "epoch": 2676, "lr": 3.6666707848008297e-06}
{"train_loss": 0.08104094862937927, "global_step": 238215, "epoch": 2676, "lr": 3.666452874001336e-06}
{"train_loss": 0.06784506142139435, "global_step": 238216, "epoch": 2676, "lr": 3.6662349694308097e-06}
{"train_loss": 0.031239112839102745, "global_step": 238217, "epoch": 2676, "lr": 3.666017071089306e-06}
{"train_loss": 0.05478666350245476, "global_step": 238218, "epoch": 2676, "lr": 3.6657991789768364e-06}
{"train_loss": 0.06537410616874695, "global_step": 238219, "epoch": 2676, "lr": 3.6655812930934395e-06}
{"train_loss": 0.0407400019466877, "global_step": 238220, "epoch": 2676, "lr": 3.665363413439132e-06}
{"train_loss": 0.026805110275745392, "global_step": 238221, "epoch": 2676, "lr": 3.6651455400139645e-06}
{"train_loss": 0.04224429652094841, "global_step": 238222, "epoch": 2676, "lr": 3.6649276728179414e-06}
{"train_loss": 0.03272424265742302, "global_step": 238223, "epoch": 2676, "lr": 3.664709811851108e-06}
{"train_loss": 0.10667706280946732, "global_step": 238224, "epoch": 2676, "lr": 3.664491957113486e-06}
{"train_loss": 0.049781642854213715, "global_step": 238225, "epoch": 2676, "lr": 3.6642741086051202e-06}
{"train_loss": 0.06755769997835159, "global_step": 238226, "epoch": 2676, "lr": 3.664056266326016e-06}
{"train_loss": 0.036965396255254745, "global_step": 238227, "epoch": 2676, "lr": 3.6638384302762228e-06}
{"train_loss": 0.08358718454837799, "global_step": 238228, "epoch": 2676, "lr": 3.6636206004557526e-06}
{"train_loss": 0.0612298846244812, "global_step": 238229, "epoch": 2676, "lr": 3.663402776864644e-06}
{"train_loss": 0.046312522143125534, "global_step": 238230, "epoch": 2676, "lr": 3.6631849595029357e-06}
{"train_loss": 0.05195128545165062, "global_step": 238231, "epoch": 2676, "lr": 3.6629671483706385e-06}
{"train_loss": 0.04601529985666275, "global_step": 238232, "epoch": 2676, "lr": 3.662749343467803e-06}
{"train_loss": 0.026154642924666405, "global_step": 238233, "epoch": 2676, "lr": 3.662531544794434e-06}
{"train_loss": 0.09365860372781754, "global_step": 238234, "epoch": 2676, "lr": 3.6623137523505768e-06}
{"train_loss": 0.04936235025525093, "global_step": 238235, "epoch": 2676, "lr": 3.662095966136253e-06}
{"train_loss": 0.022119322791695595, "global_step": 238236, "epoch": 2676, "lr": 3.6618781861515074e-06}
{"train_loss": 0.06657993048429489, "global_step": 238237, "epoch": 2676, "lr": 3.6616604123963504e-06}
{"train_loss": 0.028175080195069313, "global_step": 238238, "epoch": 2676, "lr": 3.6614426448708163e-06}
{"train_loss": 0.0067305900156497955, "global_step": 238239, "epoch": 2676, "lr": 3.661224883574943e-06}
{"train_loss": 0.015240402892231941, "global_step": 238240, "epoch": 2676, "lr": 3.661007128508742e-06}
{"train_loss": 0.040151987224817276, "global_step": 238241, "epoch": 2676, "lr": 3.6607893796722637e-06}
{"train_loss": 0.04031063988804817, "global_step": 238242, "epoch": 2676, "lr": 3.6605716370655295e-06}
{"train_loss": 0.05279022827744484, "global_step": 238243, "epoch": 2676, "lr": 3.6603539006885567e-06}
{"train_loss": 0.05238580331206322, "global_step": 238244, "epoch": 2676, "lr": 3.6601361705413896e-06}
{"train_loss": 0.04356475919485092, "global_step": 238245, "epoch": 2676, "lr": 3.65991844662405e-06}
{"train_loss": 0.05359881371259689, "global_step": 238246, "epoch": 2676, "lr": 3.6597007289365715e-06}
{"train_loss": 0.03709644451737404, "global_step": 238247, "epoch": 2676, "lr": 3.6594830174789764e-06}
{"train_loss": 0.03148368000984192, "global_step": 238248, "epoch": 2676, "lr": 3.659265312251303e-06}
{"train_loss": 0.03910361975431442, "global_step": 238249, "epoch": 2676, "lr": 3.65904761325358e-06}
{"train_loss": 0.03666713461279869, "global_step": 238250, "epoch": 2676, "lr": 3.6588299204858235e-06}
{"train_loss": 0.04341517761349678, "global_step": 238251, "epoch": 2676, "lr": 3.6586122339480834e-06}
{"train_loss": 0.04815741820951526, "global_step": 238252, "epoch": 2676, "lr": 3.658394553640371e-06, "val_loss": 9.46153450012207}
{"train_loss": 0.05394190549850464, "global_step": 238253, "epoch": 2677, "lr": 3.6581768795627302e-06}
{"train_loss": 0.06292755156755447, "global_step": 238254, "epoch": 2677, "lr": 3.6579592117151727e-06}
{"train_loss": 0.036336980760097504, "global_step": 238255, "epoch": 2677, "lr": 3.6577415500977488e-06}
{"train_loss": 0.04810170456767082, "global_step": 238256, "epoch": 2677, "lr": 3.657523894710463e-06}
{"train_loss": 0.046663738787174225, "global_step": 238257, "epoch": 2677, "lr": 3.6573062455533715e-06}
{"train_loss": 0.05935103818774223, "global_step": 238258, "epoch": 2677, "lr": 3.6570886026264796e-06}
{"train_loss": 0.07671453058719635, "global_step": 238259, "epoch": 2677, "lr": 3.6568709659298373e-06}
{"train_loss": 0.1003880649805069, "global_step": 238260, "epoch": 2677, "lr": 3.6566533354634557e-06}
{"train_loss": 0.06346574425697327, "global_step": 238261, "epoch": 2677, "lr": 3.6564357112273795e-06}
{"train_loss": 0.006105114705860615, "global_step": 238262, "epoch": 2677, "lr": 3.656218093221625e-06}
{"train_loss": 0.0498424768447876, "global_step": 238263, "epoch": 2677, "lr": 3.656000481446231e-06}
{"train_loss": 0.02491641975939274, "global_step": 238264, "epoch": 2677, "lr": 3.6557828759012202e-06}
{"train_loss": 0.04284990206360817, "global_step": 238265, "epoch": 2677, "lr": 3.6555652765866254e-06}
{"train_loss": 0.06462565809488297, "global_step": 238266, "epoch": 2677, "lr": 3.6553476835024746e-06}
{"train_loss": 0.042310453951358795, "global_step": 238267, "epoch": 2677, "lr": 3.65513009664879e-06}
{"train_loss": 0.03003644570708275, "global_step": 238268, "epoch": 2677, "lr": 3.6549125160256215e-06}
{"train_loss": 0.05280351638793945, "global_step": 238269, "epoch": 2677, "lr": 3.6546949416329747e-06}
{"train_loss": 0.04174310341477394, "global_step": 238270, "epoch": 2677, "lr": 3.6544773734708994e-06}
{"train_loss": 0.06256178766489029, "global_step": 238271, "epoch": 2677, "lr": 3.6542598115394066e-06}
{"train_loss": 0.08020725846290588, "global_step": 238272, "epoch": 2677, "lr": 3.654042255838541e-06}
{"train_loss": 0.0525832436978817, "global_step": 238273, "epoch": 2677, "lr": 3.653824706368314e-06}
{"train_loss": 0.011575804091989994, "global_step": 238274, "epoch": 2677, "lr": 3.6536071631287747e-06}
{"train_loss": 0.03121211938560009, "global_step": 238275, "epoch": 2677, "lr": 3.653389626119946e-06}
{"train_loss": 0.036016158759593964, "global_step": 238276, "epoch": 2677, "lr": 3.653172095341845e-06}
{"train_loss": 0.08091721683740616, "global_step": 238277, "epoch": 2677, "lr": 3.65295457079452e-06}
{"train_loss": 0.03505241498351097, "global_step": 238278, "epoch": 2677, "lr": 3.652737052477978e-06}
{"train_loss": 0.04884520173072815, "global_step": 238279, "epoch": 2677, "lr": 3.6525195403922685e-06}
{"train_loss": 0.0458698496222496, "global_step": 238280, "epoch": 2677, "lr": 3.6523020345374137e-06}
{"train_loss": 0.06454094499349594, "global_step": 238281, "epoch": 2677, "lr": 3.652084534913436e-06}
{"train_loss": 0.07338738441467285, "global_step": 238282, "epoch": 2677, "lr": 3.6518670415203793e-06}
{"train_loss": 0.027604490518569946, "global_step": 238283, "epoch": 2677, "lr": 3.651649554358255e-06}
{"train_loss": 0.027745826169848442, "global_step": 238284, "epoch": 2677, "lr": 3.651432073427097e-06}
{"train_loss": 0.03810001164674759, "global_step": 238285, "epoch": 2677, "lr": 3.6512145987269543e-06}
{"train_loss": 0.048272229731082916, "global_step": 238286, "epoch": 2677, "lr": 3.6509971302578273e-06}
{"train_loss": 0.035215046256780624, "global_step": 238287, "epoch": 2677, "lr": 3.6507796680197716e-06}
{"train_loss": 0.05622674524784088, "global_step": 238288, "epoch": 2677, "lr": 3.6505622120127923e-06}
{"train_loss": 0.08743610233068466, "global_step": 238289, "epoch": 2677, "lr": 3.65034476223694e-06}
{"train_loss": 0.06342873722314835, "global_step": 238290, "epoch": 2677, "lr": 3.65012731869222e-06}
{"train_loss": 0.04161195084452629, "global_step": 238291, "epoch": 2677, "lr": 3.6499098813786938e-06}
{"train_loss": 0.022079486399888992, "global_step": 238292, "epoch": 2677, "lr": 3.6496924502963547e-06}
{"train_loss": 0.08040229976177216, "global_step": 238293, "epoch": 2677, "lr": 3.6494750254452648e-06}
{"train_loss": 0.039180900901556015, "global_step": 238294, "epoch": 2677, "lr": 3.6492576068254235e-06}
{"train_loss": 0.0495338961482048, "global_step": 238295, "epoch": 2677, "lr": 3.649040194436887e-06}
{"train_loss": 0.03879254683852196, "global_step": 238296, "epoch": 2677, "lr": 3.64882278827966e-06}
{"train_loss": 0.05878186225891113, "global_step": 238297, "epoch": 2677, "lr": 3.6486053883537986e-06}
{"train_loss": 0.05798172578215599, "global_step": 238298, "epoch": 2677, "lr": 3.6483879946593024e-06}
{"train_loss": 0.03185886889696121, "global_step": 238299, "epoch": 2677, "lr": 3.6481706071962273e-06}
{"train_loss": 0.03144321218132973, "global_step": 238300, "epoch": 2677, "lr": 3.6479532259645843e-06}
{"train_loss": 0.07648604363203049, "global_step": 238301, "epoch": 2677, "lr": 3.6477358509644123e-06}
{"train_loss": 0.05263761430978775, "global_step": 238302, "epoch": 2677, "lr": 3.647518482195733e-06}
{"train_loss": 0.06144599989056587, "global_step": 238303, "epoch": 2677, "lr": 3.6473011196585805e-06}
{"train_loss": 0.06703855097293854, "global_step": 238304, "epoch": 2677, "lr": 3.6470837633529877e-06}
{"train_loss": 0.05236435681581497, "global_step": 238305, "epoch": 2677, "lr": 3.646866413278971e-06}
{"train_loss": 0.025652823969721794, "global_step": 238306, "epoch": 2677, "lr": 3.6466490694365807e-06}
{"train_loss": 0.018649999052286148, "global_step": 238307, "epoch": 2677, "lr": 3.6464317318258223e-06}
{"train_loss": 0.026887066662311554, "global_step": 238308, "epoch": 2677, "lr": 3.646214400446746e-06}
{"train_loss": 0.053767699748277664, "global_step": 238309, "epoch": 2677, "lr": 3.645997075299362e-06}
{"train_loss": 0.06510855257511139, "global_step": 238310, "epoch": 2677, "lr": 3.645779756383716e-06}
{"train_loss": 0.02186385542154312, "global_step": 238311, "epoch": 2677, "lr": 3.645562443699824e-06}
{"train_loss": 0.0657542422413826, "global_step": 238312, "epoch": 2677, "lr": 3.64534513724773e-06}
{"train_loss": 0.09014423936605453, "global_step": 238313, "epoch": 2677, "lr": 3.64512783702744e-06}
{"train_loss": 0.020936286076903343, "global_step": 238314, "epoch": 2677, "lr": 3.6449105430390095e-06}
{"train_loss": 0.01834721490740776, "global_step": 238315, "epoch": 2677, "lr": 3.6446932552824607e-06}
{"train_loss": 0.07961823791265488, "global_step": 238316, "epoch": 2677, "lr": 3.6444759737578048e-06}
{"train_loss": 0.09064153581857681, "global_step": 238317, "epoch": 2677, "lr": 3.6442586984650917e-06}
{"train_loss": 0.03576318919658661, "global_step": 238318, "epoch": 2677, "lr": 3.644041429404338e-06}
{"train_loss": 0.06520318984985352, "global_step": 238319, "epoch": 2677, "lr": 3.6438241665755826e-06}
{"train_loss": 0.040823742747306824, "global_step": 238320, "epoch": 2677, "lr": 3.6436069099788416e-06}
{"train_loss": 0.03965625911951065, "global_step": 238321, "epoch": 2677, "lr": 3.643389659614166e-06}
{"train_loss": 0.053099315613508224, "global_step": 238322, "epoch": 2677, "lr": 3.643172415481555e-06}
{"train_loss": 0.07428153604269028, "global_step": 238323, "epoch": 2677, "lr": 3.64295517758107e-06}
{"train_loss": 0.03624572604894638, "global_step": 238324, "epoch": 2677, "lr": 3.6427379459127163e-06}
{"train_loss": 0.022058557718992233, "global_step": 238325, "epoch": 2677, "lr": 3.642520720476533e-06}
{"train_loss": 0.057926859706640244, "global_step": 238326, "epoch": 2677, "lr": 3.6423035012725482e-06}
{"train_loss": 0.07094980776309967, "global_step": 238327, "epoch": 2677, "lr": 3.6420862883007946e-06}
{"train_loss": 0.07131681591272354, "global_step": 238328, "epoch": 2677, "lr": 3.6418690815612887e-06}
{"train_loss": 0.027516460046172142, "global_step": 238329, "epoch": 2677, "lr": 3.6416518810540755e-06}
{"train_loss": 0.0719127431511879, "global_step": 238330, "epoch": 2677, "lr": 3.6414346867791714e-06}
{"train_loss": 0.07000574469566345, "global_step": 238331, "epoch": 2677, "lr": 3.641217498736621e-06}
{"train_loss": 0.06910157203674316, "global_step": 238332, "epoch": 2677, "lr": 3.641000316926435e-06}
{"train_loss": 0.03447875380516052, "global_step": 238333, "epoch": 2677, "lr": 3.6407831413486583e-06}
{"train_loss": 0.039732154458761215, "global_step": 238334, "epoch": 2677, "lr": 3.6405659720033015e-06}
{"train_loss": 0.040283236652612686, "global_step": 238335, "epoch": 2677, "lr": 3.6403488088904202e-06}
{"train_loss": 0.05653008446097374, "global_step": 238336, "epoch": 2677, "lr": 3.640131652010015e-06}
{"train_loss": 0.07396064698696136, "global_step": 238337, "epoch": 2677, "lr": 3.6399145013621405e-06}
{"train_loss": 0.03121783211827278, "global_step": 238338, "epoch": 2677, "lr": 3.6396973569468086e-06}
{"train_loss": 0.04303031787276268, "global_step": 238339, "epoch": 2677, "lr": 3.639480218764052e-06}
{"train_loss": 0.06470054388046265, "global_step": 238340, "epoch": 2677, "lr": 3.63926308681391e-06}
{"train_loss": 0.05005729997928223, "global_step": 238341, "epoch": 2677, "lr": 3.6390459610963936e-06, "val_loss": 9.443272590637207}
{"train_loss": 0.05220269784331322, "global_step": 238342, "epoch": 2678, "lr": 3.6388288416115524e-06}
{"train_loss": 0.04847193881869316, "global_step": 238343, "epoch": 2678, "lr": 3.6386117283593977e-06}
{"train_loss": 0.038208261132240295, "global_step": 238344, "epoch": 2678, "lr": 3.6383946213399745e-06}
{"train_loss": 0.059792984277009964, "global_step": 238345, "epoch": 2678, "lr": 3.638177520553299e-06}
{"train_loss": 0.02943388558924198, "global_step": 238346, "epoch": 2678, "lr": 3.637960425999409e-06}
{"train_loss": 0.04842550307512283, "global_step": 238347, "epoch": 2678, "lr": 3.6377433376783234e-06}
{"train_loss": 0.03783972188830376, "global_step": 238348, "epoch": 2678, "lr": 3.63752625559009e-06}
{"train_loss": 0.037757519632577896, "global_step": 238349, "epoch": 2678, "lr": 3.6373091797347158e-06}
{"train_loss": 0.035426393151283264, "global_step": 238350, "epoch": 2678, "lr": 3.6370921101122444e-06}
{"train_loss": 0.07427554577589035, "global_step": 238351, "epoch": 2678, "lr": 3.6368750467226984e-06}
{"train_loss": 0.03750044107437134, "global_step": 238352, "epoch": 2678, "lr": 3.636657989566117e-06}
{"train_loss": 0.03723958507180214, "global_step": 238353, "epoch": 2678, "lr": 3.6364409386425157e-06}
{"train_loss": 0.06173906475305557, "global_step": 238354, "epoch": 2678, "lr": 3.636223893951929e-06}
{"train_loss": 0.03996337205171585, "global_step": 238355, "epoch": 2678, "lr": 3.6360068554943894e-06}
{"train_loss": 0.06857050210237503, "global_step": 238356, "epoch": 2678, "lr": 3.6357898232699194e-06}
{"train_loss": 0.03677411004900932, "global_step": 238357, "epoch": 2678, "lr": 3.635572797278558e-06}
{"train_loss": 0.07558316737413406, "global_step": 238358, "epoch": 2678, "lr": 3.6353557775203217e-06}
{"train_loss": 0.0494597926735878, "global_step": 238359, "epoch": 2678, "lr": 3.635138763995255e-06}
{"train_loss": 0.0937536284327507, "global_step": 238360, "epoch": 2678, "lr": 3.634921756703369e-06}
{"train_loss": 0.04178106412291527, "global_step": 238361, "epoch": 2678, "lr": 3.6347047556447133e-06}
{"train_loss": 0.04203652963042259, "global_step": 238362, "epoch": 2678, "lr": 3.6344877608192996e-06}
{"train_loss": 0.0717974305152893, "global_step": 238363, "epoch": 2678, "lr": 3.6342707722271716e-06}
{"train_loss": 0.025970010086894035, "global_step": 238364, "epoch": 2678, "lr": 3.634053789868336e-06}
{"train_loss": 0.02795581892132759, "global_step": 238365, "epoch": 2678, "lr": 3.6338368137428524e-06}
{"train_loss": 0.05682557448744774, "global_step": 238366, "epoch": 2678, "lr": 3.6336198438507217e-06}
{"train_loss": 0.057280343025922775, "global_step": 238367, "epoch": 2678, "lr": 3.6334028801919996e-06}
{"train_loss": 0.01798676885664463, "global_step": 238368, "epoch": 2678, "lr": 3.6331859227666853e-06}
{"train_loss": 0.03498394787311554, "global_step": 238369, "epoch": 2678, "lr": 3.632968971574835e-06}
{"train_loss": 0.04143429175019264, "global_step": 238370, "epoch": 2678, "lr": 3.6327520266164596e-06}
{"train_loss": 0.06817293167114258, "global_step": 238371, "epoch": 2678, "lr": 3.6325350878916088e-06}
{"train_loss": 0.03790165111422539, "global_step": 238372, "epoch": 2678, "lr": 3.6323181554002826e-06}
{"train_loss": 0.010157765820622444, "global_step": 238373, "epoch": 2678, "lr": 3.632101229142537e-06}
{"train_loss": 0.02934688702225685, "global_step": 238374, "epoch": 2678, "lr": 3.631884309118383e-06}
{"train_loss": 0.05614771693944931, "global_step": 238375, "epoch": 2678, "lr": 3.6316673953278533e-06}
{"train_loss": 0.08950915187597275, "global_step": 238376, "epoch": 2678, "lr": 3.631450487770993e-06}
{"train_loss": 0.04358939081430435, "global_step": 238377, "epoch": 2678, "lr": 3.631233586447813e-06}
{"train_loss": 0.08874305337667465, "global_step": 238378, "epoch": 2678, "lr": 3.6310166913583522e-06}
{"train_loss": 0.056086890399456024, "global_step": 238379, "epoch": 2678, "lr": 3.6307998025026268e-06}
{"train_loss": 0.060540635138750076, "global_step": 238380, "epoch": 2678, "lr": 3.6305829198806873e-06}
{"train_loss": 0.07611974328756332, "global_step": 238381, "epoch": 2678, "lr": 3.6303660434925392e-06}
{"train_loss": 0.047935158014297485, "global_step": 238382, "epoch": 2678, "lr": 3.630149173338232e-06}
{"train_loss": 0.05798686668276787, "global_step": 238383, "epoch": 2678, "lr": 3.629932309417783e-06}
{"train_loss": 0.016016891226172447, "global_step": 238384, "epoch": 2678, "lr": 3.629715451731225e-06}
{"train_loss": 0.07537145167589188, "global_step": 238385, "epoch": 2678, "lr": 3.629498600278586e-06}
{"train_loss": 0.034252431243658066, "global_step": 238386, "epoch": 2678, "lr": 3.629281755059899e-06}
{"train_loss": 0.024992596358060837, "global_step": 238387, "epoch": 2678, "lr": 3.629064916075181e-06}
{"train_loss": 0.026878757402300835, "global_step": 238388, "epoch": 2678, "lr": 3.628848083324482e-06}
{"train_loss": 0.02300884947180748, "global_step": 238389, "epoch": 2678, "lr": 3.6286312568078074e-06}
{"train_loss": 0.054041873663663864, "global_step": 238390, "epoch": 2678, "lr": 3.628414436525207e-06}
{"train_loss": 0.05115355923771858, "global_step": 238391, "epoch": 2678, "lr": 3.6281976224767033e-06}
{"train_loss": 0.03234262764453888, "global_step": 238392, "epoch": 2678, "lr": 3.6279808146623127e-06}
{"train_loss": 0.05366082116961479, "global_step": 238393, "epoch": 2678, "lr": 3.6277640130820856e-06}
{"train_loss": 0.0462711825966835, "global_step": 238394, "epoch": 2678, "lr": 3.627547217736027e-06}
{"train_loss": 0.028806062415242195, "global_step": 238395, "epoch": 2678, "lr": 3.627330428624193e-06}
{"train_loss": 0.029154565185308456, "global_step": 238396, "epoch": 2678, "lr": 3.627113645746588e-06}
{"train_loss": 0.055841848254203796, "global_step": 238397, "epoch": 2678, "lr": 3.6268968691032634e-06}
{"train_loss": 0.09577751159667969, "global_step": 238398, "epoch": 2678, "lr": 3.6266800986942294e-06}
{"train_loss": 0.0428214967250824, "global_step": 238399, "epoch": 2678, "lr": 3.6264633345195252e-06}
{"train_loss": 0.008419888094067574, "global_step": 238400, "epoch": 2678, "lr": 3.6262465765791788e-06}
{"train_loss": 0.14785097539424896, "global_step": 238401, "epoch": 2678, "lr": 3.626029824873217e-06}
{"train_loss": 0.023634936660528183, "global_step": 238402, "epoch": 2678, "lr": 3.6258130794016687e-06}
{"train_loss": 0.04656291380524635, "global_step": 238403, "epoch": 2678, "lr": 3.625596340164572e-06}
{"train_loss": 0.04787936061620712, "global_step": 238404, "epoch": 2678, "lr": 3.625379607161938e-06}
{"train_loss": 0.027127932757139206, "global_step": 238405, "epoch": 2678, "lr": 3.6251628803938174e-06}
{"train_loss": 0.06083499640226364, "global_step": 238406, "epoch": 2678, "lr": 3.6249461598602154e-06}
{"train_loss": 0.033596016466617584, "global_step": 238407, "epoch": 2678, "lr": 3.624729445561187e-06}
{"train_loss": 0.038687773048877716, "global_step": 238408, "epoch": 2678, "lr": 3.624512737496738e-06}
{"train_loss": 0.05960656329989433, "global_step": 238409, "epoch": 2678, "lr": 3.6242960356669185e-06}
{"train_loss": 0.0335838757455349, "global_step": 238410, "epoch": 2678, "lr": 3.62407934007174e-06}
{"train_loss": 0.05931776762008667, "global_step": 238411, "epoch": 2678, "lr": 3.623862650711235e-06}
{"train_loss": 0.08819740265607834, "global_step": 238412, "epoch": 2678, "lr": 3.623645967585443e-06}
{"train_loss": 0.02327699400484562, "global_step": 238413, "epoch": 2678, "lr": 3.623429290694386e-06}
{"train_loss": 0.06730663776397705, "global_step": 238414, "epoch": 2678, "lr": 3.6232126200380965e-06}
{"train_loss": 0.06092046573758125, "global_step": 238415, "epoch": 2678, "lr": 3.6229959556165927e-06}
{"train_loss": 0.018846560269594193, "global_step": 238416, "epoch": 2678, "lr": 3.622779297429918e-06}
{"train_loss": 0.04906570166349411, "global_step": 238417, "epoch": 2678, "lr": 3.6225626454780893e-06}
{"train_loss": 0.0861208364367485, "global_step": 238418, "epoch": 2678, "lr": 3.6223459997611507e-06}
{"train_loss": 0.06378055363893509, "global_step": 238419, "epoch": 2678, "lr": 3.622129360279114e-06}
{"train_loss": 0.013785145245492458, "global_step": 238420, "epoch": 2678, "lr": 3.6219127270320285e-06}
{"train_loss": 0.023030079901218414, "global_step": 238421, "epoch": 2678, "lr": 3.6216961000199e-06}
{"train_loss": 0.029855644330382347, "global_step": 238422, "epoch": 2678, "lr": 3.621479479242773e-06}
{"train_loss": 0.026158642023801804, "global_step": 238423, "epoch": 2678, "lr": 3.62126286470067e-06}
{"train_loss": 0.0511641725897789, "global_step": 238424, "epoch": 2678, "lr": 3.6210462563936287e-06}
{"train_loss": 0.05556192621588707, "global_step": 238425, "epoch": 2678, "lr": 3.620829654321667e-06}
{"train_loss": 0.01811373047530651, "global_step": 238426, "epoch": 2678, "lr": 3.6206130584848284e-06}
{"train_loss": 0.04745497182011604, "global_step": 238427, "epoch": 2678, "lr": 3.6203964688831195e-06}
{"train_loss": 0.06653270870447159, "global_step": 238428, "epoch": 2678, "lr": 3.6201798855165947e-06}
{"train_loss": 0.027735820040106773, "global_step": 238429, "epoch": 2678, "lr": 3.6199633083852657e-06}
{"train_loss": 0.04745069998996646, "global_step": 238430, "epoch": 2678, "lr": 3.6197467374891716e-06, "val_loss": 9.579444885253906}
{"train_loss": 0.0480404756963253, "global_step": 238431, "epoch": 2679, "lr": 3.6195301728283394e-06}
{"train_loss": 0.02163877710700035, "global_step": 238432, "epoch": 2679, "lr": 3.619313614402786e-06}
{"train_loss": 0.026731863617897034, "global_step": 238433, "epoch": 2679, "lr": 3.619097062212562e-06}
{"train_loss": 0.02655906043946743, "global_step": 238434, "epoch": 2679, "lr": 3.6188805162576723e-06}
{"train_loss": 0.02593599073588848, "global_step": 238435, "epoch": 2679, "lr": 3.618663976538167e-06}
{"train_loss": 0.046257782727479935, "global_step": 238436, "epoch": 2679, "lr": 3.6184474430540625e-06}
{"train_loss": 0.036334820091724396, "global_step": 238437, "epoch": 2679, "lr": 3.6182309158053983e-06}
{"train_loss": 0.037190452218055725, "global_step": 238438, "epoch": 2679, "lr": 3.6180143947921908e-06}
{"train_loss": 0.04620184376835823, "global_step": 238439, "epoch": 2679, "lr": 3.617797880014484e-06}
{"train_loss": 0.045140817761421204, "global_step": 238440, "epoch": 2679, "lr": 3.6175813714722897e-06}
{"train_loss": 0.08004293590784073, "global_step": 238441, "epoch": 2679, "lr": 3.617364869165657e-06}
{"train_loss": 0.06692654639482498, "global_step": 238442, "epoch": 2679, "lr": 3.617148373094592e-06}
{"train_loss": 0.050422169268131256, "global_step": 238443, "epoch": 2679, "lr": 3.6169318832591503e-06}
{"train_loss": 0.033990487456321716, "global_step": 238444, "epoch": 2679, "lr": 3.616715399659332e-06}
{"train_loss": 0.048896241933107376, "global_step": 238445, "epoch": 2679, "lr": 3.6164989222951915e-06}
{"train_loss": 0.04751985892653465, "global_step": 238446, "epoch": 2679, "lr": 3.6162824511667414e-06}
{"train_loss": 0.0203762948513031, "global_step": 238447, "epoch": 2679, "lr": 3.6160659862740144e-06}
{"train_loss": 0.04557778686285019, "global_step": 238448, "epoch": 2679, "lr": 3.6158495276170546e-06}
{"train_loss": 0.042626481503248215, "global_step": 238449, "epoch": 2679, "lr": 3.6156330751958624e-06}
{"train_loss": 0.036819297820329666, "global_step": 238450, "epoch": 2679, "lr": 3.6154166290104983e-06}
{"train_loss": 0.05699419230222702, "global_step": 238451, "epoch": 2679, "lr": 3.6152001890609634e-06}
{"train_loss": 0.07838000357151031, "global_step": 238452, "epoch": 2679, "lr": 3.6149837553473065e-06}
{"train_loss": 0.06036479398608208, "global_step": 238453, "epoch": 2679, "lr": 3.614767327869545e-06}
{"train_loss": 0.05426853895187378, "global_step": 238454, "epoch": 2679, "lr": 3.6145509066277227e-06}
{"train_loss": 0.019851302728056908, "global_step": 238455, "epoch": 2679, "lr": 3.614334491621846e-06}
{"train_loss": 0.06363839656114578, "global_step": 238456, "epoch": 2679, "lr": 3.6141180828519695e-06}
{"train_loss": 0.06614166498184204, "global_step": 238457, "epoch": 2679, "lr": 3.613901680318099e-06}
{"train_loss": 0.03359406813979149, "global_step": 238458, "epoch": 2679, "lr": 3.6136852840202795e-06}
{"train_loss": 0.04301491379737854, "global_step": 238459, "epoch": 2679, "lr": 3.613468893958527e-06}
{"train_loss": 0.047387685626745224, "global_step": 238460, "epoch": 2679, "lr": 3.613252510132892e-06}
{"train_loss": 0.02519276924431324, "global_step": 238461, "epoch": 2679, "lr": 3.6130361325433735e-06}
{"train_loss": 0.054116666316986084, "global_step": 238462, "epoch": 2679, "lr": 3.612819761190034e-06}
{"train_loss": 0.02102135308086872, "global_step": 238463, "epoch": 2679, "lr": 3.6126033960728723e-06}
{"train_loss": 0.07665795087814331, "global_step": 238464, "epoch": 2679, "lr": 3.6123870371919387e-06}
{"train_loss": 0.0687267854809761, "global_step": 238465, "epoch": 2679, "lr": 3.6121706845472446e-06}
{"train_loss": 0.08753316104412079, "global_step": 238466, "epoch": 2679, "lr": 3.6119543381388398e-06}
{"train_loss": 0.07839374989271164, "global_step": 238467, "epoch": 2679, "lr": 3.6117379979667354e-06}
{"train_loss": 0.04224776476621628, "global_step": 238468, "epoch": 2679, "lr": 3.61152166403097e-06}
{"train_loss": 0.0992363914847374, "global_step": 238469, "epoch": 2679, "lr": 3.6113053363315773e-06}
{"train_loss": 0.03051820397377014, "global_step": 238470, "epoch": 2679, "lr": 3.6110890148685626e-06}
{"train_loss": 0.10473119467496872, "global_step": 238471, "epoch": 2679, "lr": 3.6108726996419872e-06}
{"train_loss": 0.06650279462337494, "global_step": 238472, "epoch": 2679, "lr": 3.6106563906518507e-06}
{"train_loss": 0.04586678743362427, "global_step": 238473, "epoch": 2679, "lr": 3.610440087898209e-06}
{"train_loss": 0.03950142860412598, "global_step": 238474, "epoch": 2679, "lr": 3.6102237913810676e-06}
{"train_loss": 0.03684423118829727, "global_step": 238475, "epoch": 2679, "lr": 3.6100075011004765e-06}
{"train_loss": 0.011391136795282364, "global_step": 238476, "epoch": 2679, "lr": 3.6097912170564462e-06}
{"train_loss": 0.04268215596675873, "global_step": 238477, "epoch": 2679, "lr": 3.6095749392490217e-06}
{"train_loss": 0.057685934007167816, "global_step": 238478, "epoch": 2679, "lr": 3.6093586676782144e-06}
{"train_loss": 0.08299615979194641, "global_step": 238479, "epoch": 2679, "lr": 3.6091424023440734e-06}
{"train_loss": 0.03403523564338684, "global_step": 238480, "epoch": 2679, "lr": 3.6089261432466047e-06}
{"train_loss": 0.023438433185219765, "global_step": 238481, "epoch": 2679, "lr": 3.608709890385864e-06}
{"train_loss": 0.03321864455938339, "global_step": 238482, "epoch": 2679, "lr": 3.6084936437618568e-06}
{"train_loss": 0.02221483737230301, "global_step": 238483, "epoch": 2679, "lr": 3.608277403374627e-06}
{"train_loss": 0.05684782564640045, "global_step": 238484, "epoch": 2679, "lr": 3.6080611692242027e-06}
{"train_loss": 0.06324480473995209, "global_step": 238485, "epoch": 2679, "lr": 3.607844941310601e-06}
{"train_loss": 0.024681681767106056, "global_step": 238486, "epoch": 2679, "lr": 3.6076287196338653e-06}
{"train_loss": 0.030190134420990944, "global_step": 238487, "epoch": 2679, "lr": 3.6074125041940133e-06}
{"train_loss": 0.05138606205582619, "global_step": 238488, "epoch": 2679, "lr": 3.607196294991083e-06}
{"train_loss": 0.023062782362103462, "global_step": 238489, "epoch": 2679, "lr": 3.6069800920250973e-06}
{"train_loss": 0.04584578424692154, "global_step": 238490, "epoch": 2679, "lr": 3.6067638952960947e-06}
{"train_loss": 0.06465134769678116, "global_step": 238491, "epoch": 2679, "lr": 3.606547704804086e-06}
{"train_loss": 0.07495208829641342, "global_step": 238492, "epoch": 2679, "lr": 3.6063315205491222e-06}
{"train_loss": 0.08655252307653427, "global_step": 238493, "epoch": 2679, "lr": 3.6061153425312134e-06}
{"train_loss": 0.032609935849905014, "global_step": 238494, "epoch": 2679, "lr": 3.605899170750404e-06}
{"train_loss": 0.059280820190906525, "global_step": 238495, "epoch": 2679, "lr": 3.6056830052067113e-06}
{"train_loss": 0.03355154022574425, "global_step": 238496, "epoch": 2679, "lr": 3.605466845900174e-06}
{"train_loss": 0.023446373641490936, "global_step": 238497, "epoch": 2679, "lr": 3.6052506928308083e-06}
{"train_loss": 0.01412417832762003, "global_step": 238498, "epoch": 2679, "lr": 3.605034545998659e-06}
{"train_loss": 0.04314207658171654, "global_step": 238499, "epoch": 2679, "lr": 3.604818405403737e-06}
{"train_loss": 0.02718185819685459, "global_step": 238500, "epoch": 2679, "lr": 3.6046022710460926e-06}
{"train_loss": 0.0412851981818676, "global_step": 238501, "epoch": 2679, "lr": 3.6043861429257364e-06}
{"train_loss": 0.04066793620586395, "global_step": 238502, "epoch": 2679, "lr": 3.604170021042713e-06}
{"train_loss": 0.078044593334198, "global_step": 238503, "epoch": 2679, "lr": 3.603953905397034e-06}
{"train_loss": 0.04895375296473503, "global_step": 238504, "epoch": 2679, "lr": 3.6037377959887487e-06}
{"train_loss": 0.053547557443380356, "global_step": 238505, "epoch": 2679, "lr": 3.6035216928178626e-06}
{"train_loss": 0.04496409744024277, "global_step": 238506, "epoch": 2679, "lr": 3.603305595884432e-06}
{"train_loss": 0.033947549760341644, "global_step": 238507, "epoch": 2679, "lr": 3.603089505188467e-06}
{"train_loss": 0.03891093283891678, "global_step": 238508, "epoch": 2679, "lr": 3.6028734207299908e-06}
{"train_loss": 0.039882101118564606, "global_step": 238509, "epoch": 2679, "lr": 3.6026573425090582e-06}
{"train_loss": 0.028217367827892303, "global_step": 238510, "epoch": 2679, "lr": 3.6024412705256695e-06}
{"train_loss": 0.053921353071928024, "global_step": 238511, "epoch": 2679, "lr": 3.602225204779874e-06}
{"train_loss": 0.047725483775138855, "global_step": 238512, "epoch": 2679, "lr": 3.6020091452716897e-06}
{"train_loss": 0.046940166503190994, "global_step": 238513, "epoch": 2679, "lr": 3.6017930920011544e-06}
{"train_loss": 0.025127043947577477, "global_step": 238514, "epoch": 2679, "lr": 3.601577044968285e-06}
{"train_loss": 0.025130638852715492, "global_step": 238515, "epoch": 2679, "lr": 3.601361004173126e-06}
{"train_loss": 0.05454790219664574, "global_step": 238516, "epoch": 2679, "lr": 3.601144969615694e-06}
{"train_loss": 0.07041344791650772, "global_step": 238517, "epoch": 2679, "lr": 3.600928941296028e-06}
{"train_loss": 0.031542085111141205, "global_step": 238518, "epoch": 2679, "lr": 3.6007129192141444e-06}
{"train_loss": 0.04657378873337856, "global_step": 238519, "epoch": 2679, "lr": 3.6004969033700873e-06, "val_loss": 9.45759391784668}
{"train_loss": 0.04527851194143295, "global_step": 238520, "epoch": 2680, "lr": 3.600280893763869e-06}
{"train_loss": 0.04158341512084007, "global_step": 238521, "epoch": 2680, "lr": 3.600064890395527e-06}
{"train_loss": 0.0451805405318737, "global_step": 238522, "epoch": 2680, "lr": 3.599848893265101e-06}
{"train_loss": 0.053330741822719574, "global_step": 238523, "epoch": 2680, "lr": 3.599632902372596e-06}
{"train_loss": 0.013463628478348255, "global_step": 238524, "epoch": 2680, "lr": 3.599416917718068e-06}
{"train_loss": 0.0717012956738472, "global_step": 238525, "epoch": 2680, "lr": 3.5992009393015225e-06}
{"train_loss": 0.04892133176326752, "global_step": 238526, "epoch": 2680, "lr": 3.598984967123009e-06}
{"train_loss": 0.03460242599248886, "global_step": 238527, "epoch": 2680, "lr": 3.5987690011825393e-06}
{"train_loss": 0.05237517133355141, "global_step": 238528, "epoch": 2680, "lr": 3.598553041480157e-06}
{"train_loss": 0.05661335214972496, "global_step": 238529, "epoch": 2680, "lr": 3.598337088015874e-06}
{"train_loss": 0.03100394271314144, "global_step": 238530, "epoch": 2680, "lr": 3.598121140789734e-06}
{"train_loss": 0.04203382879495621, "global_step": 238531, "epoch": 2680, "lr": 3.5979051998017597e-06}
{"train_loss": 0.053987812250852585, "global_step": 238532, "epoch": 2680, "lr": 3.5976892650519843e-06}
{"train_loss": 0.057341475039720535, "global_step": 238533, "epoch": 2680, "lr": 3.59747333654043e-06}
{"train_loss": 0.06158928945660591, "global_step": 238534, "epoch": 2680, "lr": 3.5972574142671357e-06}
{"train_loss": 0.060121480375528336, "global_step": 238535, "epoch": 2680, "lr": 3.5970414982321175e-06}
{"train_loss": 0.06041192263364792, "global_step": 238536, "epoch": 2680, "lr": 3.5968255884354208e-06}
{"train_loss": 0.032634034752845764, "global_step": 238537, "epoch": 2680, "lr": 3.5966096848770505e-06}
{"train_loss": 0.033242445439100266, "global_step": 238538, "epoch": 2680, "lr": 3.5963937875570563e-06}
{"train_loss": 0.046234335750341415, "global_step": 238539, "epoch": 2680, "lr": 3.596177896475472e-06}
{"train_loss": 0.06910295784473419, "global_step": 238540, "epoch": 2680, "lr": 3.5959620116323032e-06}
{"train_loss": 0.052783384919166565, "global_step": 238541, "epoch": 2680, "lr": 3.5957461330276e-06}
{"train_loss": 0.05652225762605667, "global_step": 238542, "epoch": 2680, "lr": 3.595530260661384e-06}
{"train_loss": 0.07567055523395538, "global_step": 238543, "epoch": 2680, "lr": 3.5953143945336774e-06}
{"train_loss": 0.05281927064061165, "global_step": 238544, "epoch": 2680, "lr": 3.5950985346445197e-06}
{"train_loss": 0.04053675755858421, "global_step": 238545, "epoch": 2680, "lr": 3.5948826809939275e-06}
{"train_loss": 0.04691743850708008, "global_step": 238546, "epoch": 2680, "lr": 3.59466683358195e-06}
{"train_loss": 0.05876757204532623, "global_step": 238547, "epoch": 2680, "lr": 3.594450992408599e-06}
{"train_loss": 0.03716914728283882, "global_step": 238548, "epoch": 2680, "lr": 3.594235157473902e-06}
{"train_loss": 0.08803890645503998, "global_step": 238549, "epoch": 2680, "lr": 3.5940193287779032e-06}
{"train_loss": 0.04366942495107651, "global_step": 238550, "epoch": 2680, "lr": 3.5938035063206143e-06}
{"train_loss": 0.05589766427874565, "global_step": 238551, "epoch": 2680, "lr": 3.5935876901020846e-06}
{"train_loss": 0.016107788309454918, "global_step": 238552, "epoch": 2680, "lr": 3.5933718801223203e-06}
{"train_loss": 0.05254712700843811, "global_step": 238553, "epoch": 2680, "lr": 3.5931560763813652e-06}
{"train_loss": 0.03219133988022804, "global_step": 238554, "epoch": 2680, "lr": 3.592940278879242e-06}
{"train_loss": 0.027360383421182632, "global_step": 238555, "epoch": 2680, "lr": 3.5927244876159894e-06}
{"train_loss": 0.016413060948252678, "global_step": 238556, "epoch": 2680, "lr": 3.5925087025916236e-06}
{"train_loss": 0.040815096348524094, "global_step": 238557, "epoch": 2680, "lr": 3.592292923806173e-06}
{"train_loss": 0.018281642347574234, "global_step": 238558, "epoch": 2680, "lr": 3.5920771512596873e-06}
{"train_loss": 0.03923157975077629, "global_step": 238559, "epoch": 2680, "lr": 3.591861384952172e-06}
{"train_loss": 0.04419105872511864, "global_step": 238560, "epoch": 2680, "lr": 3.5916456248836717e-06}
{"train_loss": 0.023726141080260277, "global_step": 238561, "epoch": 2680, "lr": 3.591429871054197e-06}
{"train_loss": 0.04691918566823006, "global_step": 238562, "epoch": 2680, "lr": 3.591214123463804e-06}
{"train_loss": 0.05345197767019272, "global_step": 238563, "epoch": 2680, "lr": 3.590998382112498e-06}
{"train_loss": 0.021762264892458916, "global_step": 238564, "epoch": 2680, "lr": 3.590782647000318e-06}
{"train_loss": 0.07248973101377487, "global_step": 238565, "epoch": 2680, "lr": 3.5905669181272917e-06}
{"train_loss": 0.02115965075790882, "global_step": 238566, "epoch": 2680, "lr": 3.5903511954934522e-06}
{"train_loss": 0.06567580252885818, "global_step": 238567, "epoch": 2680, "lr": 3.5901354790988163e-06}
{"train_loss": 0.07426097244024277, "global_step": 238568, "epoch": 2680, "lr": 3.589919768943428e-06}
{"train_loss": 0.031977634876966476, "global_step": 238569, "epoch": 2680, "lr": 3.589704065027305e-06}
{"train_loss": 0.03538989648222923, "global_step": 238570, "epoch": 2680, "lr": 3.5894883673504852e-06}
{"train_loss": 0.02697819285094738, "global_step": 238571, "epoch": 2680, "lr": 3.5892726759129857e-06}
{"train_loss": 0.04381754621863365, "global_step": 238572, "epoch": 2680, "lr": 3.5890569907148506e-06}
{"train_loss": 0.020809616893529892, "global_step": 238573, "epoch": 2680, "lr": 3.588841311756097e-06}
{"train_loss": 0.03558029606938362, "global_step": 238574, "epoch": 2680, "lr": 3.588625639036752e-06}
{"train_loss": 0.05676558241248131, "global_step": 238575, "epoch": 2680, "lr": 3.588409972556861e-06}
{"train_loss": 0.021619286388158798, "global_step": 238576, "epoch": 2680, "lr": 3.588194312316434e-06}
{"train_loss": 0.05985969305038452, "global_step": 238577, "epoch": 2680, "lr": 3.5879786583155215e-06}
{"train_loss": 0.022651318460702896, "global_step": 238578, "epoch": 2680, "lr": 3.587763010554124e-06}
{"train_loss": 0.046643439680337906, "global_step": 238579, "epoch": 2680, "lr": 3.5875473690323013e-06}
{"train_loss": 0.04502036049962044, "global_step": 238580, "epoch": 2680, "lr": 3.58733173375006e-06}
{"train_loss": 0.019706612452864647, "global_step": 238581, "epoch": 2680, "lr": 3.587116104707433e-06}
{"train_loss": 0.04945465549826622, "global_step": 238582, "epoch": 2680, "lr": 3.5869004819044595e-06}
{"train_loss": 0.03448312357068062, "global_step": 238583, "epoch": 2680, "lr": 3.5866848653411556e-06}
{"train_loss": 0.03860120102763176, "global_step": 238584, "epoch": 2680, "lr": 3.5864692550175605e-06}
{"train_loss": 0.02708003856241703, "global_step": 238585, "epoch": 2680, "lr": 3.586253650933702e-06}
{"train_loss": 0.04758340120315552, "global_step": 238586, "epoch": 2680, "lr": 3.5860380530895965e-06}
{"train_loss": 0.01820167526602745, "global_step": 238587, "epoch": 2680, "lr": 3.5858224614852886e-06}
{"train_loss": 0.05102849006652832, "global_step": 238588, "epoch": 2680, "lr": 3.585606876120795e-06}
{"train_loss": 0.05785384774208069, "global_step": 238589, "epoch": 2680, "lr": 3.58539129699616e-06}
{"train_loss": 0.04969439283013344, "global_step": 238590, "epoch": 2680, "lr": 3.5851757241113946e-06}
{"train_loss": 0.06434398889541626, "global_step": 238591, "epoch": 2680, "lr": 3.584960157466544e-06}
{"train_loss": 0.03733083978295326, "global_step": 238592, "epoch": 2680, "lr": 3.5847445970616234e-06}
{"train_loss": 0.0258796289563179, "global_step": 238593, "epoch": 2680, "lr": 3.5845290428966672e-06}
{"train_loss": 0.07480006664991379, "global_step": 238594, "epoch": 2680, "lr": 3.5843134949717083e-06}
{"train_loss": 0.06460163742303848, "global_step": 238595, "epoch": 2680, "lr": 3.584097953286769e-06}
{"train_loss": 0.04613437131047249, "global_step": 238596, "epoch": 2680, "lr": 3.5838824178418937e-06}
{"train_loss": 0.05362248420715332, "global_step": 238597, "epoch": 2680, "lr": 3.5836668886370885e-06}
{"train_loss": 0.040139779448509216, "global_step": 238598, "epoch": 2680, "lr": 3.5834513656724024e-06}
{"train_loss": 0.03587948530912399, "global_step": 238599, "epoch": 2680, "lr": 3.5832358489478467e-06}
{"train_loss": 0.03405377268791199, "global_step": 238600, "epoch": 2680, "lr": 3.5830203384634665e-06}
{"train_loss": 0.008906564675271511, "global_step": 238601, "epoch": 2680, "lr": 3.582804834219272e-06}
{"train_loss": 0.04447386413812637, "global_step": 238602, "epoch": 2680, "lr": 3.5825893362153197e-06}
{"train_loss": 0.046055104583501816, "global_step": 238603, "epoch": 2680, "lr": 3.5823738444516085e-06}
{"train_loss": 0.07236168533563614, "global_step": 238604, "epoch": 2680, "lr": 3.5821583589281894e-06}
{"train_loss": 0.04543168842792511, "global_step": 238605, "epoch": 2680, "lr": 3.5819428796450784e-06}
{"train_loss": 0.028136348351836205, "global_step": 238606, "epoch": 2680, "lr": 3.5817274066023197e-06}
{"train_loss": 0.01769690215587616, "global_step": 238607, "epoch": 2680, "lr": 3.58151193979992e-06}
{"train_loss": 0.04457411927621016, "global_step": 238608, "epoch": 2680, "lr": 3.5812964792379277e-06, "val_loss": 9.522462844848633, "train_action_mse_error": 36.81511306762695}
{"train_loss": 0.040573328733444214, "global_step": 238609, "epoch": 2681, "lr": 3.581081024916355e-06}
{"train_loss": 0.03704394772648811, "global_step": 238610, "epoch": 2681, "lr": 3.580865576835246e-06}
{"train_loss": 0.030497917905449867, "global_step": 238611, "epoch": 2681, "lr": 3.5806501349946285e-06}
{"train_loss": 0.024362554773688316, "global_step": 238612, "epoch": 2681, "lr": 3.5804346993945194e-06}
{"train_loss": 0.03606361895799637, "global_step": 238613, "epoch": 2681, "lr": 3.5802192700349624e-06}
{"train_loss": 0.033693280071020126, "global_step": 238614, "epoch": 2681, "lr": 3.580003846915969e-06}
{"train_loss": 0.03806836158037186, "global_step": 238615, "epoch": 2681, "lr": 3.5797884300375894e-06}
{"train_loss": 0.04049041494727135, "global_step": 238616, "epoch": 2681, "lr": 3.5795730193998343e-06}
{"train_loss": 0.05215970799326897, "global_step": 238617, "epoch": 2681, "lr": 3.579357615002743e-06}
{"train_loss": 0.051397550851106644, "global_step": 238618, "epoch": 2681, "lr": 3.5791422168463485e-06}
{"train_loss": 0.03588375821709633, "global_step": 238619, "epoch": 2681, "lr": 3.5789268249306563e-06}
{"train_loss": 0.0166949313133955, "global_step": 238620, "epoch": 2681, "lr": 3.5787114392557275e-06}
{"train_loss": 0.08088626712560654, "global_step": 238621, "epoch": 2681, "lr": 3.578496059821562e-06}
{"train_loss": 0.032906342297792435, "global_step": 238622, "epoch": 2681, "lr": 3.57828068662821e-06}
{"train_loss": 0.09339764714241028, "global_step": 238623, "epoch": 2681, "lr": 3.578065319675694e-06}
{"train_loss": 0.03950922191143036, "global_step": 238624, "epoch": 2681, "lr": 3.5778499589640358e-06}
{"train_loss": 0.06862873584032059, "global_step": 238625, "epoch": 2681, "lr": 3.577634604493274e-06}
{"train_loss": 0.05399361625313759, "global_step": 238626, "epoch": 2681, "lr": 3.5774192562634258e-06}
{"train_loss": 0.0708046704530716, "global_step": 238627, "epoch": 2681, "lr": 3.577203914274535e-06}
{"train_loss": 0.05147208273410797, "global_step": 238628, "epoch": 2681, "lr": 3.5769885785266187e-06}
{"train_loss": 0.037742432206869125, "global_step": 238629, "epoch": 2681, "lr": 3.5767732490197105e-06}
{"train_loss": 0.05073314160108566, "global_step": 238630, "epoch": 2681, "lr": 3.5765579257538427e-06}
{"train_loss": 0.027552885934710503, "global_step": 238631, "epoch": 2681, "lr": 3.5763426087290387e-06}
{"train_loss": 0.06490053236484528, "global_step": 238632, "epoch": 2681, "lr": 3.576127297945331e-06}
{"train_loss": 0.03851736709475517, "global_step": 238633, "epoch": 2681, "lr": 3.575911993402742e-06}
{"train_loss": 0.06234500929713249, "global_step": 238634, "epoch": 2681, "lr": 3.575696695101316e-06}
{"train_loss": 0.0397762805223465, "global_step": 238635, "epoch": 2681, "lr": 3.5754814030410645e-06}
{"train_loss": 0.06944911926984787, "global_step": 238636, "epoch": 2681, "lr": 3.5752661172220257e-06}
{"train_loss": 0.005357715766876936, "global_step": 238637, "epoch": 2681, "lr": 3.5750508376442225e-06}
{"train_loss": 0.05593986064195633, "global_step": 238638, "epoch": 2681, "lr": 3.5748355643076935e-06}
{"train_loss": 0.03670356050133705, "global_step": 238639, "epoch": 2681, "lr": 3.574620297212461e-06}
{"train_loss": 0.04555705934762955, "global_step": 238640, "epoch": 2681, "lr": 3.574405036358558e-06}
{"train_loss": 0.042179547250270844, "global_step": 238641, "epoch": 2681, "lr": 3.574189781746001e-06}
{"train_loss": 0.023138828575611115, "global_step": 238642, "epoch": 2681, "lr": 3.5739745333748352e-06}
{"train_loss": 0.06642061471939087, "global_step": 238643, "epoch": 2681, "lr": 3.573759291245077e-06}
{"train_loss": 0.029629619792103767, "global_step": 238644, "epoch": 2681, "lr": 3.57354405535677e-06}
{"train_loss": 0.05355161055922508, "global_step": 238645, "epoch": 2681, "lr": 3.5733288257099263e-06}
{"train_loss": 0.0382627509534359, "global_step": 238646, "epoch": 2681, "lr": 3.57311360230459e-06}
{"train_loss": 0.0234970860183239, "global_step": 238647, "epoch": 2681, "lr": 3.572898385140777e-06}
{"train_loss": 0.03503560647368431, "global_step": 238648, "epoch": 2681, "lr": 3.572683174218516e-06}
{"train_loss": 0.008965705521404743, "global_step": 238649, "epoch": 2681, "lr": 3.5724679695378573e-06}
{"train_loss": 0.05462239682674408, "global_step": 238650, "epoch": 2681, "lr": 3.5722527710987996e-06}
{"train_loss": 0.03957296535372734, "global_step": 238651, "epoch": 2681, "lr": 3.5720375789013993e-06}
{"train_loss": 0.0426911823451519, "global_step": 238652, "epoch": 2681, "lr": 3.5718223929456616e-06}
{"train_loss": 0.06274305284023285, "global_step": 238653, "epoch": 2681, "lr": 3.571607213231637e-06}
{"train_loss": 0.030477408319711685, "global_step": 238654, "epoch": 2681, "lr": 3.57139203975933e-06}
{"train_loss": 0.056643277406692505, "global_step": 238655, "epoch": 2681, "lr": 3.5711768725287973e-06}
{"train_loss": 0.010901659727096558, "global_step": 238656, "epoch": 2681, "lr": 3.570961711540044e-06}
{"train_loss": 0.010188066400587559, "global_step": 238657, "epoch": 2681, "lr": 3.5707465567931196e-06}
{"train_loss": 0.056824974715709686, "global_step": 238658, "epoch": 2681, "lr": 3.5705314082880416e-06}
{"train_loss": 0.05427948758006096, "global_step": 238659, "epoch": 2681, "lr": 3.5703162660248315e-06}
{"train_loss": 0.023444080725312233, "global_step": 238660, "epoch": 2681, "lr": 3.570101130003534e-06}
{"train_loss": 0.048899877816438675, "global_step": 238661, "epoch": 2681, "lr": 3.5698860002241596e-06}
{"train_loss": 0.029416630044579506, "global_step": 238662, "epoch": 2681, "lr": 3.569670876686759e-06}
{"train_loss": 0.04960807412862778, "global_step": 238663, "epoch": 2681, "lr": 3.569455759391349e-06}
{"train_loss": 0.08731495589017868, "global_step": 238664, "epoch": 2681, "lr": 3.569240648337957e-06}
{"train_loss": 0.04122716188430786, "global_step": 238665, "epoch": 2681, "lr": 3.5690255435266106e-06}
{"train_loss": 0.025039633736014366, "global_step": 238666, "epoch": 2681, "lr": 3.568810444957349e-06}
{"train_loss": 0.03162723407149315, "global_step": 238667, "epoch": 2681, "lr": 3.5685953526301884e-06}
{"train_loss": 0.060192469507455826, "global_step": 238668, "epoch": 2681, "lr": 3.568380266545174e-06}
{"train_loss": 0.015102003701031208, "global_step": 238669, "epoch": 2681, "lr": 3.568165186702316e-06}
{"train_loss": 0.021632559597492218, "global_step": 238670, "epoch": 2681, "lr": 3.567950113101659e-06}
{"train_loss": 0.03992108628153801, "global_step": 238671, "epoch": 2681, "lr": 3.5677350457432148e-06}
{"train_loss": 0.06344147771596909, "global_step": 238672, "epoch": 2681, "lr": 3.5675199846270325e-06}
{"train_loss": 0.037328459322452545, "global_step": 238673, "epoch": 2681, "lr": 3.5673049297531293e-06}
{"train_loss": 0.040157996118068695, "global_step": 238674, "epoch": 2681, "lr": 3.567089881121538e-06}
{"train_loss": 0.0333685502409935, "global_step": 238675, "epoch": 2681, "lr": 3.5668748387322757e-06}
{"train_loss": 0.02685891091823578, "global_step": 238676, "epoch": 2681, "lr": 3.5666598025853927e-06}
{"train_loss": 0.0661194697022438, "global_step": 238677, "epoch": 2681, "lr": 3.5664447726808994e-06}
{"train_loss": 0.05201559141278267, "global_step": 238678, "epoch": 2681, "lr": 3.5662297490188347e-06}
{"train_loss": 0.028462637215852737, "global_step": 238679, "epoch": 2681, "lr": 3.5660147315992208e-06}
{"train_loss": 0.03833816573023796, "global_step": 238680, "epoch": 2681, "lr": 3.5657997204220973e-06}
{"train_loss": 0.049048278480768204, "global_step": 238681, "epoch": 2681, "lr": 3.56558471548748e-06}
{"train_loss": 0.07546813040971756, "global_step": 238682, "epoch": 2681, "lr": 3.565369716795408e-06}
{"train_loss": 0.03614963963627815, "global_step": 238683, "epoch": 2681, "lr": 3.5651547243458985e-06}
{"train_loss": 0.07437866926193237, "global_step": 238684, "epoch": 2681, "lr": 3.564939738138989e-06}
{"train_loss": 0.0398380272090435, "global_step": 238685, "epoch": 2681, "lr": 3.56472475817472e-06}
{"train_loss": 0.05190304294228554, "global_step": 238686, "epoch": 2681, "lr": 3.564509784453096e-06}
{"train_loss": 0.019748469814658165, "global_step": 238687, "epoch": 2681, "lr": 3.5642948169741617e-06}
{"train_loss": 0.036077629774808884, "global_step": 238688, "epoch": 2681, "lr": 3.5640798557379395e-06}
{"train_loss": 0.025513507425785065, "global_step": 238689, "epoch": 2681, "lr": 3.563864900744468e-06}
{"train_loss": 0.04582635685801506, "global_step": 238690, "epoch": 2681, "lr": 3.5636499519937582e-06}
{"train_loss": 0.05860423669219017, "global_step": 238691, "epoch": 2681, "lr": 3.5634350094858605e-06}
{"train_loss": 0.03226207569241524, "global_step": 238692, "epoch": 2681, "lr": 3.5632200732207854e-06}
{"train_loss": 0.03267966955900192, "global_step": 238693, "epoch": 2681, "lr": 3.563005143198572e-06}
{"train_loss": 0.08737459778785706, "global_step": 238694, "epoch": 2681, "lr": 3.562790219419243e-06}
{"train_loss": 0.09163254499435425, "global_step": 238695, "epoch": 2681, "lr": 3.5625753018828367e-06}
{"train_loss": 0.03983321040868759, "global_step": 238696, "epoch": 2681, "lr": 3.562360390589381e-06}
{"train_loss": 0.044159428546142375, "global_step": 238697, "epoch": 2681, "lr": 3.562145485538887e-06, "val_loss": 9.569714546203613}
{"train_loss": 0.09028881043195724, "global_step": 238698, "epoch": 2682, "lr": 3.5619305867314044e-06}
{"train_loss": 0.02343469299376011, "global_step": 238699, "epoch": 2682, "lr": 3.5617156941669503e-06}
{"train_loss": 0.04183511808514595, "global_step": 238700, "epoch": 2682, "lr": 3.5615008078455635e-06}
{"train_loss": 0.017526086419820786, "global_step": 238701, "epoch": 2682, "lr": 3.5612859277672606e-06}
{"train_loss": 0.04484578222036362, "global_step": 238702, "epoch": 2682, "lr": 3.56107105393208e-06}
{"train_loss": 0.04167737439274788, "global_step": 238703, "epoch": 2682, "lr": 3.5608561863400445e-06}
{"train_loss": 0.030621688812971115, "global_step": 238704, "epoch": 2682, "lr": 3.5606413249911876e-06}
{"train_loss": 0.048974595963954926, "global_step": 238705, "epoch": 2682, "lr": 3.560426469885536e-06}
{"train_loss": 0.06464354693889618, "global_step": 238706, "epoch": 2682, "lr": 3.560211621023124e-06}
{"train_loss": 0.03630353510379791, "global_step": 238707, "epoch": 2682, "lr": 3.5599967784039625e-06}
{"train_loss": 0.03379024937748909, "global_step": 238708, "epoch": 2682, "lr": 3.5597819420281064e-06}
{"train_loss": 0.05249569192528725, "global_step": 238709, "epoch": 2682, "lr": 3.559567111895562e-06}
{"train_loss": 0.10713472217321396, "global_step": 238710, "epoch": 2682, "lr": 3.559352288006379e-06}
{"train_loss": 0.05053054913878441, "global_step": 238711, "epoch": 2682, "lr": 3.559137470360563e-06}
{"train_loss": 0.04096590727567673, "global_step": 238712, "epoch": 2682, "lr": 3.5589226589581693e-06}
{"train_loss": 0.040233537554740906, "global_step": 238713, "epoch": 2682, "lr": 3.558707853799198e-06}
{"train_loss": 0.10367024689912796, "global_step": 238714, "epoch": 2682, "lr": 3.5584930548837047e-06}
{"train_loss": 0.0419980064034462, "global_step": 238715, "epoch": 2682, "lr": 3.558278262211695e-06}
{"train_loss": 0.04194677621126175, "global_step": 238716, "epoch": 2682, "lr": 3.5580634757832133e-06}
{"train_loss": 0.05988525599241257, "global_step": 238717, "epoch": 2682, "lr": 3.5578486955982816e-06}
{"train_loss": 0.13325028121471405, "global_step": 238718, "epoch": 2682, "lr": 3.5576339216569386e-06}
{"train_loss": 0.03600723668932915, "global_step": 238719, "epoch": 2682, "lr": 3.557419153959196e-06}
{"train_loss": 0.038588155061006546, "global_step": 238720, "epoch": 2682, "lr": 3.5572043925050924e-06}
{"train_loss": 0.05109458789229393, "global_step": 238721, "epoch": 2682, "lr": 3.5569896372946667e-06}
{"train_loss": 0.03904739022254944, "global_step": 238722, "epoch": 2682, "lr": 3.556774888327924e-06}
{"train_loss": 0.06361333280801773, "global_step": 238723, "epoch": 2682, "lr": 3.55656014560492e-06}
{"train_loss": 0.0507381409406662, "global_step": 238724, "epoch": 2682, "lr": 3.5563454091256607e-06}
{"train_loss": 0.05999915301799774, "global_step": 238725, "epoch": 2682, "lr": 3.556130678890196e-06}
{"train_loss": 0.04981657862663269, "global_step": 238726, "epoch": 2682, "lr": 3.555915954898531e-06}
{"train_loss": 0.08988068252801895, "global_step": 238727, "epoch": 2682, "lr": 3.555701237150716e-06}
{"train_loss": 0.0411638543009758, "global_step": 238728, "epoch": 2682, "lr": 3.5554865256467675e-06}
{"train_loss": 0.03188064694404602, "global_step": 238729, "epoch": 2682, "lr": 3.5552718203867187e-06}
{"train_loss": 0.04934140294790268, "global_step": 238730, "epoch": 2682, "lr": 3.555057121370592e-06}
{"train_loss": 0.0590989924967289, "global_step": 238731, "epoch": 2682, "lr": 3.5548424285984317e-06}
{"train_loss": 0.02759641781449318, "global_step": 238732, "epoch": 2682, "lr": 3.554627742070249e-06}
{"train_loss": 0.0342048816382885, "global_step": 238733, "epoch": 2682, "lr": 3.554413061786088e-06}
{"train_loss": 0.026135725900530815, "global_step": 238734, "epoch": 2682, "lr": 3.5541983877459662e-06}
{"train_loss": 0.053782135248184204, "global_step": 238735, "epoch": 2682, "lr": 3.5539837199499103e-06}
{"train_loss": 0.04772951453924179, "global_step": 238736, "epoch": 2682, "lr": 3.5537690583979657e-06}
{"train_loss": 0.024364307522773743, "global_step": 238737, "epoch": 2682, "lr": 3.5535544030901426e-06}
{"train_loss": 0.03410417214035988, "global_step": 238738, "epoch": 2682, "lr": 3.553339754026486e-06}
{"train_loss": 0.06554116308689117, "global_step": 238739, "epoch": 2682, "lr": 3.5531251112070064e-06}
{"train_loss": 0.0491165965795517, "global_step": 238740, "epoch": 2682, "lr": 3.552910474631749e-06}
{"train_loss": 0.04188142716884613, "global_step": 238741, "epoch": 2682, "lr": 3.5526958443007353e-06}
{"train_loss": 0.07647033035755157, "global_step": 238742, "epoch": 2682, "lr": 3.552481220213999e-06}
{"train_loss": 0.09345807135105133, "global_step": 238743, "epoch": 2682, "lr": 3.5522666023715565e-06}
{"train_loss": 0.029206056147813797, "global_step": 238744, "epoch": 2682, "lr": 3.5520519907734584e-06}
{"train_loss": 0.025914965197443962, "global_step": 238745, "epoch": 2682, "lr": 3.5518373854197095e-06}
{"train_loss": 0.06029930338263512, "global_step": 238746, "epoch": 2682, "lr": 3.5516227863103604e-06}
{"train_loss": 0.05131935328245163, "global_step": 238747, "epoch": 2682, "lr": 3.5514081934454156e-06}
{"train_loss": 0.05698205530643463, "global_step": 238748, "epoch": 2682, "lr": 3.551193606824932e-06}
{"train_loss": 0.06177455186843872, "global_step": 238749, "epoch": 2682, "lr": 3.550979026448914e-06}
{"train_loss": 0.05664956569671631, "global_step": 238750, "epoch": 2682, "lr": 3.5507644523174067e-06}
{"train_loss": 0.11330373585224152, "global_step": 238751, "epoch": 2682, "lr": 3.5505498844304264e-06}
{"train_loss": 0.07917042821645737, "global_step": 238752, "epoch": 2682, "lr": 3.5503353227880178e-06}
{"train_loss": 0.033112820237874985, "global_step": 238753, "epoch": 2682, "lr": 3.5501207673901914e-06}
{"train_loss": 0.03195064142346382, "global_step": 238754, "epoch": 2682, "lr": 3.5499062182369926e-06}
{"train_loss": 0.07359733432531357, "global_step": 238755, "epoch": 2682, "lr": 3.549691675328437e-06}
{"train_loss": 0.05440959706902504, "global_step": 238756, "epoch": 2682, "lr": 3.5494771386645587e-06}
{"train_loss": 0.015409804880619049, "global_step": 238757, "epoch": 2682, "lr": 3.549262608245396e-06}
{"train_loss": 0.01594509743154049, "global_step": 238758, "epoch": 2682, "lr": 3.549048084070955e-06}
{"train_loss": 0.05163666978478432, "global_step": 238759, "epoch": 2682, "lr": 3.548833566141291e-06}
{"train_loss": 0.06741929054260254, "global_step": 238760, "epoch": 2682, "lr": 3.5486190544564145e-06}
{"train_loss": 0.028663350269198418, "global_step": 238761, "epoch": 2682, "lr": 3.548404549016365e-06}
{"train_loss": 0.04692326486110687, "global_step": 238762, "epoch": 2682, "lr": 3.5481900498211595e-06}
{"train_loss": 0.03146897628903389, "global_step": 238763, "epoch": 2682, "lr": 3.5479755568708363e-06}
{"train_loss": 0.02958497405052185, "global_step": 238764, "epoch": 2682, "lr": 3.5477610701654172e-06}
{"train_loss": 0.03734786808490753, "global_step": 238765, "epoch": 2682, "lr": 3.5475465897049477e-06}
{"train_loss": 0.02883937396109104, "global_step": 238766, "epoch": 2682, "lr": 3.547332115489432e-06}
{"train_loss": 0.023821420967578888, "global_step": 238767, "epoch": 2682, "lr": 3.5471176475189217e-06}
{"train_loss": 0.06182722747325897, "global_step": 238768, "epoch": 2682, "lr": 3.5469031857934267e-06}
{"train_loss": 0.03292812779545784, "global_step": 238769, "epoch": 2682, "lr": 3.5466887303129857e-06}
{"train_loss": 0.06025495007634163, "global_step": 238770, "epoch": 2682, "lr": 3.5464742810776276e-06}
{"train_loss": 0.03632007911801338, "global_step": 238771, "epoch": 2682, "lr": 3.5462598380873846e-06}
{"train_loss": 0.04902936890721321, "global_step": 238772, "epoch": 2682, "lr": 3.5460454013422685e-06}
{"train_loss": 0.058679014444351196, "global_step": 238773, "epoch": 2682, "lr": 3.5458309708423344e-06}
{"train_loss": 0.0157316867262125, "global_step": 238774, "epoch": 2682, "lr": 3.5456165465875936e-06}
{"train_loss": 0.04148529842495918, "global_step": 238775, "epoch": 2682, "lr": 3.545402128578074e-06}
{"train_loss": 0.052745893597602844, "global_step": 238776, "epoch": 2682, "lr": 3.545187716813814e-06}
{"train_loss": 0.05251894146203995, "global_step": 238777, "epoch": 2682, "lr": 3.5449733112948304e-06}
{"train_loss": 0.026932664215564728, "global_step": 238778, "epoch": 2682, "lr": 3.5447589120211677e-06}
{"train_loss": 0.08025521785020828, "global_step": 238779, "epoch": 2682, "lr": 3.5445445189928374e-06}
{"train_loss": 0.043396372348070145, "global_step": 238780, "epoch": 2682, "lr": 3.544330132209883e-06}
{"train_loss": 0.08277063071727753, "global_step": 238781, "epoch": 2682, "lr": 3.544115751672317e-06}
{"train_loss": 0.04243997111916542, "global_step": 238782, "epoch": 2682, "lr": 3.543901377380193e-06}
{"train_loss": 0.06320343911647797, "global_step": 238783, "epoch": 2682, "lr": 3.5436870093335127e-06}
{"train_loss": 0.06921198219060898, "global_step": 238784, "epoch": 2682, "lr": 3.5434726475323255e-06}
{"train_loss": 0.040347304195165634, "global_step": 238785, "epoch": 2682, "lr": 3.5432582919766476e-06}
{"train_loss": 0.049631475293067065, "global_step": 238786, "epoch": 2682, "lr": 3.5430439426665185e-06, "val_loss": 9.441460609436035}
{"train_loss": 0.04259933903813362, "global_step": 238787, "epoch": 2683, "lr": 3.5428295996019547e-06}
{"train_loss": 0.03758612275123596, "global_step": 238788, "epoch": 2683, "lr": 3.5426152627829946e-06}
{"train_loss": 0.05828249827027321, "global_step": 238789, "epoch": 2683, "lr": 3.5424009322096607e-06}
{"train_loss": 0.03335057199001312, "global_step": 238790, "epoch": 2683, "lr": 3.5421866078819866e-06}
{"train_loss": 0.06432869285345078, "global_step": 238791, "epoch": 2683, "lr": 3.541972289799994e-06}
{"train_loss": 0.029039857909083366, "global_step": 238792, "epoch": 2683, "lr": 3.5417579779637224e-06}
{"train_loss": 0.027837766334414482, "global_step": 238793, "epoch": 2683, "lr": 3.541543672373199e-06}
{"train_loss": 0.0875689834356308, "global_step": 238794, "epoch": 2683, "lr": 3.5413293730284404e-06}
{"train_loss": 0.043138645589351654, "global_step": 238795, "epoch": 2683, "lr": 3.541115079929491e-06}
{"train_loss": 0.04969111084938049, "global_step": 238796, "epoch": 2683, "lr": 3.5409007930763626e-06}
{"train_loss": 0.03560283035039902, "global_step": 238797, "epoch": 2683, "lr": 3.5406865124691047e-06}
{"train_loss": 0.04923579841852188, "global_step": 238798, "epoch": 2683, "lr": 3.5404722381077283e-06}
{"train_loss": 0.048082705587148666, "global_step": 238799, "epoch": 2683, "lr": 3.5402579699922776e-06}
{"train_loss": 0.041016194969415665, "global_step": 238800, "epoch": 2683, "lr": 3.5400437081227644e-06}
{"train_loss": 0.07710113376379013, "global_step": 238801, "epoch": 2683, "lr": 3.5398294524992324e-06}
{"train_loss": 0.055596355348825455, "global_step": 238802, "epoch": 2683, "lr": 3.5396152031216934e-06}
{"train_loss": 0.05234522372484207, "global_step": 238803, "epoch": 2683, "lr": 3.5394009599902024e-06}
{"train_loss": 0.06506584584712982, "global_step": 238804, "epoch": 2683, "lr": 3.5391867231047595e-06}
{"train_loss": 0.04854711890220642, "global_step": 238805, "epoch": 2683, "lr": 3.5389724924654146e-06}
{"train_loss": 0.04006747901439667, "global_step": 238806, "epoch": 2683, "lr": 3.5387582680721843e-06}
{"train_loss": 0.060279641300439835, "global_step": 238807, "epoch": 2683, "lr": 3.538544049925102e-06}
{"train_loss": 0.043144889175891876, "global_step": 238808, "epoch": 2683, "lr": 3.538329838024196e-06}
{"train_loss": 0.05017685145139694, "global_step": 238809, "epoch": 2683, "lr": 3.5381156323694987e-06}
{"train_loss": 0.08447173237800598, "global_step": 238810, "epoch": 2683, "lr": 3.5379014329610327e-06}
{"train_loss": 0.011706366203725338, "global_step": 238811, "epoch": 2683, "lr": 3.5376872397988313e-06}
{"train_loss": 0.014556367881596088, "global_step": 238812, "epoch": 2683, "lr": 3.5374730528829227e-06}
{"train_loss": 0.060572974383831024, "global_step": 238813, "epoch": 2683, "lr": 3.537258872213328e-06}
{"train_loss": 0.05619317293167114, "global_step": 238814, "epoch": 2683, "lr": 3.537044697790093e-06}
{"train_loss": 0.038777049630880356, "global_step": 238815, "epoch": 2683, "lr": 3.5368305296132276e-06}
{"train_loss": 0.04283687472343445, "global_step": 238816, "epoch": 2683, "lr": 3.536616367682771e-06}
{"train_loss": 0.028408920392394066, "global_step": 238817, "epoch": 2683, "lr": 3.536402211998746e-06}
{"train_loss": 0.037925947457551956, "global_step": 238818, "epoch": 2683, "lr": 3.5361880625611964e-06}
{"train_loss": 0.03463929519057274, "global_step": 238819, "epoch": 2683, "lr": 3.535973919370128e-06}
{"train_loss": 0.04203531891107559, "global_step": 238820, "epoch": 2683, "lr": 3.5357597824255905e-06}
{"train_loss": 0.058785781264305115, "global_step": 238821, "epoch": 2683, "lr": 3.5355456517275953e-06}
{"train_loss": 0.0527019239962101, "global_step": 238822, "epoch": 2683, "lr": 3.5353315272761924e-06}
{"train_loss": 0.03462638333439827, "global_step": 238823, "epoch": 2683, "lr": 3.5351174090713814e-06}
{"train_loss": 0.03787526115775108, "global_step": 238824, "epoch": 2683, "lr": 3.5349032971132236e-06}
{"train_loss": 0.019167741760611534, "global_step": 238825, "epoch": 2683, "lr": 3.534689191401719e-06}
{"train_loss": 0.04159636050462723, "global_step": 238826, "epoch": 2683, "lr": 3.5344750919369173e-06}
{"train_loss": 0.03716154024004936, "global_step": 238827, "epoch": 2683, "lr": 3.5342609987188302e-06}
{"train_loss": 0.04620873183012009, "global_step": 238828, "epoch": 2683, "lr": 3.5340469117475015e-06}
{"train_loss": 0.07246459275484085, "global_step": 238829, "epoch": 2683, "lr": 3.533832831022954e-06}
{"train_loss": 0.05062522366642952, "global_step": 238830, "epoch": 2683, "lr": 3.533618756545215e-06}
{"train_loss": 0.026214342564344406, "global_step": 238831, "epoch": 2683, "lr": 3.533404688314318e-06}
{"train_loss": 0.05522603541612625, "global_step": 238832, "epoch": 2683, "lr": 3.5331906263302793e-06}
{"train_loss": 0.01896088570356369, "global_step": 238833, "epoch": 2683, "lr": 3.5329765705931495e-06}
{"train_loss": 0.062151700258255005, "global_step": 238834, "epoch": 2683, "lr": 3.5327625211029337e-06}
{"train_loss": 0.025750314816832542, "global_step": 238835, "epoch": 2683, "lr": 3.5325484778596817e-06}
{"train_loss": 0.09340217709541321, "global_step": 238836, "epoch": 2683, "lr": 3.5323344408633998e-06}
{"train_loss": 0.04505685716867447, "global_step": 238837, "epoch": 2683, "lr": 3.5321204101141425e-06}
{"train_loss": 0.059298768639564514, "global_step": 238838, "epoch": 2683, "lr": 3.5319063856119163e-06}
{"train_loss": 0.025283992290496826, "global_step": 238839, "epoch": 2683, "lr": 3.5316923673567647e-06}
{"train_loss": 0.0689988061785698, "global_step": 238840, "epoch": 2683, "lr": 3.531478355348705e-06}
{"train_loss": 0.06754539906978607, "global_step": 238841, "epoch": 2683, "lr": 3.531264349587776e-06}
{"train_loss": 0.05972648039460182, "global_step": 238842, "epoch": 2683, "lr": 3.5310503500739943e-06}
{"train_loss": 0.06489219516515732, "global_step": 238843, "epoch": 2683, "lr": 3.53083635680741e-06}
{"train_loss": 0.012533669359982014, "global_step": 238844, "epoch": 2683, "lr": 3.530622369788028e-06}
{"train_loss": 0.07274658977985382, "global_step": 238845, "epoch": 2683, "lr": 3.5304083890158932e-06}
{"train_loss": 0.08035580068826675, "global_step": 238846, "epoch": 2683, "lr": 3.5301944144910224e-06}
{"train_loss": 0.034638144075870514, "global_step": 238847, "epoch": 2683, "lr": 3.5299804462134545e-06}
{"train_loss": 0.05431628227233887, "global_step": 238848, "epoch": 2683, "lr": 3.529766484183211e-06}
{"train_loss": 0.016345495358109474, "global_step": 238849, "epoch": 2683, "lr": 3.5295525284003317e-06}
{"train_loss": 0.03066434897482395, "global_step": 238850, "epoch": 2683, "lr": 3.5293385788648382e-06}
{"train_loss": 0.04737759381532669, "global_step": 238851, "epoch": 2683, "lr": 3.529124635576747e-06}
{"train_loss": 0.026629474014043808, "global_step": 238852, "epoch": 2683, "lr": 3.5289106985361086e-06}
{"train_loss": 0.026819605380296707, "global_step": 238853, "epoch": 2683, "lr": 3.528696767742934e-06}
{"train_loss": 0.045801449567079544, "global_step": 238854, "epoch": 2683, "lr": 3.528482843197267e-06}
{"train_loss": 0.01840810850262642, "global_step": 238855, "epoch": 2683, "lr": 3.528268924899125e-06}
{"train_loss": 0.033166565001010895, "global_step": 238856, "epoch": 2683, "lr": 3.5280550128485466e-06}
{"train_loss": 0.02341497875750065, "global_step": 238857, "epoch": 2683, "lr": 3.527841107045543e-06}
{"train_loss": 0.050035323947668076, "global_step": 238858, "epoch": 2683, "lr": 3.5276272074901695e-06}
{"train_loss": 0.045098934322595596, "global_step": 238859, "epoch": 2683, "lr": 3.527413314182426e-06}
{"train_loss": 0.023184482008218765, "global_step": 238860, "epoch": 2683, "lr": 3.527199427122363e-06}
{"train_loss": 0.05971556901931763, "global_step": 238861, "epoch": 2683, "lr": 3.5269855463099964e-06}
{"train_loss": 0.007692770101130009, "global_step": 238862, "epoch": 2683, "lr": 3.5267716717453713e-06}
{"train_loss": 0.0784539207816124, "global_step": 238863, "epoch": 2683, "lr": 3.526557803428493e-06}
{"train_loss": 0.05283777043223381, "global_step": 238864, "epoch": 2683, "lr": 3.526343941359406e-06}
{"train_loss": 0.05113071948289871, "global_step": 238865, "epoch": 2683, "lr": 3.5261300855381376e-06}
{"train_loss": 0.03127899393439293, "global_step": 238866, "epoch": 2683, "lr": 3.52591623596471e-06}
{"train_loss": 0.018266450613737106, "global_step": 238867, "epoch": 2683, "lr": 3.5257023926391687e-06}
{"train_loss": 0.04508558660745621, "global_step": 238868, "epoch": 2683, "lr": 3.525488555561518e-06}
{"train_loss": 0.046415992081165314, "global_step": 238869, "epoch": 2683, "lr": 3.525274724731803e-06}
{"train_loss": 0.07430781424045563, "global_step": 238870, "epoch": 2683, "lr": 3.5250609001500455e-06}
{"train_loss": 0.04562291502952576, "global_step": 238871, "epoch": 2683, "lr": 3.524847081816285e-06}
{"train_loss": 0.06971961259841919, "global_step": 238872, "epoch": 2683, "lr": 3.5246332697305316e-06}
{"train_loss": 0.034356337040662766, "global_step": 238873, "epoch": 2683, "lr": 3.5244194638928363e-06}
{"train_loss": 0.018096376210451126, "global_step": 238874, "epoch": 2683, "lr": 3.5242056643032097e-06}
{"train_loss": 0.04601959329558892, "global_step": 238875, "epoch": 2683, "lr": 3.5239918709616903e-06, "val_loss": 9.507493019104004}
{"train_loss": 0.12183290719985962, "global_step": 238876, "epoch": 2684, "lr": 3.5237780838682954e-06}
{"train_loss": 0.06569083034992218, "global_step": 238877, "epoch": 2684, "lr": 3.5235643030230747e-06}
{"train_loss": 0.05168275535106659, "global_step": 238878, "epoch": 2684, "lr": 3.5233505284260336e-06}
{"train_loss": 0.04723590612411499, "global_step": 238879, "epoch": 2684, "lr": 3.523136760077217e-06}
{"train_loss": 0.09413903951644897, "global_step": 238880, "epoch": 2684, "lr": 3.5229229979766464e-06}
{"train_loss": 0.06264413893222809, "global_step": 238881, "epoch": 2684, "lr": 3.522709242124356e-06}
{"train_loss": 0.04702639579772949, "global_step": 238882, "epoch": 2684, "lr": 3.5224954925203615e-06}
{"train_loss": 0.06095873564481735, "global_step": 238883, "epoch": 2684, "lr": 3.5222817491647076e-06}
{"train_loss": 0.06788396835327148, "global_step": 238884, "epoch": 2684, "lr": 3.5220680120574167e-06}
{"train_loss": 0.025950821116566658, "global_step": 238885, "epoch": 2684, "lr": 3.5218542811985223e-06}
{"train_loss": 0.053641177713871, "global_step": 238886, "epoch": 2684, "lr": 3.521640556588041e-06}
{"train_loss": 0.0240058321505785, "global_step": 238887, "epoch": 2684, "lr": 3.521426838226011e-06}
{"train_loss": 0.04455991089344025, "global_step": 238888, "epoch": 2684, "lr": 3.521213126112455e-06}
{"train_loss": 0.0384705550968647, "global_step": 238889, "epoch": 2684, "lr": 3.520999420247412e-06}
{"train_loss": 0.046142976731061935, "global_step": 238890, "epoch": 2684, "lr": 3.5207857206309037e-06}
{"train_loss": 0.062107305973768234, "global_step": 238891, "epoch": 2684, "lr": 3.5205720272629526e-06}
{"train_loss": 0.06237239018082619, "global_step": 238892, "epoch": 2684, "lr": 3.5203583401436037e-06}
{"train_loss": 0.022793130949139595, "global_step": 238893, "epoch": 2684, "lr": 3.5201446592728614e-06}
{"train_loss": 0.07309984415769577, "global_step": 238894, "epoch": 2684, "lr": 3.5199309846507823e-06}
{"train_loss": 0.03336204215884209, "global_step": 238895, "epoch": 2684, "lr": 3.5197173162773767e-06}
{"train_loss": 0.03505475074052811, "global_step": 238896, "epoch": 2684, "lr": 3.519503654152684e-06}
{"train_loss": 0.04527013748884201, "global_step": 238897, "epoch": 2684, "lr": 3.5192899982767145e-06}
{"train_loss": 0.02749982848763466, "global_step": 238898, "epoch": 2684, "lr": 3.519076348649525e-06}
{"train_loss": 0.035210832953453064, "global_step": 238899, "epoch": 2684, "lr": 3.51886270527112e-06}
{"train_loss": 0.08176832646131516, "global_step": 238900, "epoch": 2684, "lr": 3.5186490681415385e-06}
{"train_loss": 0.04071076959371567, "global_step": 238901, "epoch": 2684, "lr": 3.5184354372608084e-06}
{"train_loss": 0.05617339164018631, "global_step": 238902, "epoch": 2684, "lr": 3.5182218126289524e-06}
{"train_loss": 0.06829465925693512, "global_step": 238903, "epoch": 2684, "lr": 3.5180081942460198e-06}
{"train_loss": 0.059270892292261124, "global_step": 238904, "epoch": 2684, "lr": 3.5177945821120104e-06}
{"train_loss": 0.012510855682194233, "global_step": 238905, "epoch": 2684, "lr": 3.517580976226975e-06}
{"train_loss": 0.02712256647646427, "global_step": 238906, "epoch": 2684, "lr": 3.51736737659093e-06}
{"train_loss": 0.06079037860035896, "global_step": 238907, "epoch": 2684, "lr": 3.5171537832039136e-06}
{"train_loss": 0.08249063044786453, "global_step": 238908, "epoch": 2684, "lr": 3.516940196065943e-06}
{"train_loss": 0.09114447981119156, "global_step": 238909, "epoch": 2684, "lr": 3.5167266151770627e-06}
{"train_loss": 0.03164059296250343, "global_step": 238910, "epoch": 2684, "lr": 3.516513040537278e-06}
{"train_loss": 0.036354053765535355, "global_step": 238911, "epoch": 2684, "lr": 3.5162994721466446e-06}
{"train_loss": 0.07375264167785645, "global_step": 238912, "epoch": 2684, "lr": 3.5160859100051678e-06}
{"train_loss": 0.0433112233877182, "global_step": 238913, "epoch": 2684, "lr": 3.515872354112898e-06}
{"train_loss": 0.0740402340888977, "global_step": 238914, "epoch": 2684, "lr": 3.51565880446984e-06}
{"train_loss": 0.05261652171611786, "global_step": 238915, "epoch": 2684, "lr": 3.5154452610760445e-06}
{"train_loss": 0.03375660628080368, "global_step": 238916, "epoch": 2684, "lr": 3.515231723931528e-06}
{"train_loss": 0.07319334149360657, "global_step": 238917, "epoch": 2684, "lr": 3.5150181930363234e-06}
{"train_loss": 0.03974861651659012, "global_step": 238918, "epoch": 2684, "lr": 3.5148046683904536e-06}
{"train_loss": 0.012831330299377441, "global_step": 238919, "epoch": 2684, "lr": 3.514591149993951e-06}
{"train_loss": 0.028823794797062874, "global_step": 238920, "epoch": 2684, "lr": 3.5143776378468494e-06}
{"train_loss": 0.04298226162791252, "global_step": 238921, "epoch": 2684, "lr": 3.5141641319491713e-06}
{"train_loss": 0.030903860926628113, "global_step": 238922, "epoch": 2684, "lr": 3.5139506323009498e-06}
{"train_loss": 0.03437913954257965, "global_step": 238923, "epoch": 2684, "lr": 3.5137371389022177e-06}
{"train_loss": 0.055125825107097626, "global_step": 238924, "epoch": 2684, "lr": 3.513523651752981e-06}
{"train_loss": 0.022746825590729713, "global_step": 238925, "epoch": 2684, "lr": 3.5133101708532956e-06}
{"train_loss": 0.05876949056982994, "global_step": 238926, "epoch": 2684, "lr": 3.513096696203172e-06}
{"train_loss": 0.05778037756681442, "global_step": 238927, "epoch": 2684, "lr": 3.5128832278026545e-06}
{"train_loss": 0.07008545845746994, "global_step": 238928, "epoch": 2684, "lr": 3.5126697656517658e-06}
{"train_loss": 0.03101777285337448, "global_step": 238929, "epoch": 2684, "lr": 3.512456309750517e-06}
{"train_loss": 0.037361349910497665, "global_step": 238930, "epoch": 2684, "lr": 3.512242860098963e-06}
{"train_loss": 0.03837455436587334, "global_step": 238931, "epoch": 2684, "lr": 3.5120294166971157e-06}
{"train_loss": 0.02738572098314762, "global_step": 238932, "epoch": 2684, "lr": 3.511815979545019e-06}
{"train_loss": 0.09248866140842438, "global_step": 238933, "epoch": 2684, "lr": 3.5116025486426785e-06}
{"train_loss": 0.0847838819026947, "global_step": 238934, "epoch": 2684, "lr": 3.5113891239901442e-06}
{"train_loss": 0.018154270946979523, "global_step": 238935, "epoch": 2684, "lr": 3.511175705587433e-06}
{"train_loss": 0.05362710356712341, "global_step": 238936, "epoch": 2684, "lr": 3.5109622934345832e-06}
{"train_loss": 0.040567416697740555, "global_step": 238937, "epoch": 2684, "lr": 3.510748887531612e-06}
{"train_loss": 0.04160788282752037, "global_step": 238938, "epoch": 2684, "lr": 3.5105354878785525e-06}
{"train_loss": 0.07669365406036377, "global_step": 238939, "epoch": 2684, "lr": 3.5103220944754434e-06}
{"train_loss": 0.05046219378709793, "global_step": 238940, "epoch": 2684, "lr": 3.510108707322296e-06}
{"train_loss": 0.09428241848945618, "global_step": 238941, "epoch": 2684, "lr": 3.509895326419155e-06}
{"train_loss": 0.05978076532483101, "global_step": 238942, "epoch": 2684, "lr": 3.509681951766036e-06}
{"train_loss": 0.07301392406225204, "global_step": 238943, "epoch": 2684, "lr": 3.5094685833629846e-06}
{"train_loss": 0.04221149533987045, "global_step": 238944, "epoch": 2684, "lr": 3.5092552212100062e-06}
{"train_loss": 0.07064957916736603, "global_step": 238945, "epoch": 2684, "lr": 3.50904186530715e-06}
{"train_loss": 0.04417033493518829, "global_step": 238946, "epoch": 2684, "lr": 3.5088285156544278e-06}
{"train_loss": 0.05167142301797867, "global_step": 238947, "epoch": 2684, "lr": 3.508615172251889e-06}
{"train_loss": 0.04697379469871521, "global_step": 238948, "epoch": 2684, "lr": 3.5084018350995396e-06}
{"train_loss": 0.06206821650266647, "global_step": 238949, "epoch": 2684, "lr": 3.508188504197424e-06}
{"train_loss": 0.05281250178813934, "global_step": 238950, "epoch": 2684, "lr": 3.5079751795455582e-06}
{"train_loss": 0.09081164747476578, "global_step": 238951, "epoch": 2684, "lr": 3.507761861143993e-06}
{"train_loss": 0.0896768793463707, "global_step": 238952, "epoch": 2684, "lr": 3.5075485489927283e-06}
{"train_loss": 0.07452995330095291, "global_step": 238953, "epoch": 2684, "lr": 3.5073352430918196e-06}
{"train_loss": 0.0395955927670002, "global_step": 238954, "epoch": 2684, "lr": 3.5071219434412715e-06}
{"train_loss": 0.0813911184668541, "global_step": 238955, "epoch": 2684, "lr": 3.506908650041124e-06}
{"train_loss": 0.05760185420513153, "global_step": 238956, "epoch": 2684, "lr": 3.5066953628914155e-06}
{"train_loss": 0.011379878036677837, "global_step": 238957, "epoch": 2684, "lr": 3.506482081992157e-06}
{"train_loss": 0.047954708337783813, "global_step": 238958, "epoch": 2684, "lr": 3.5062688073433938e-06}
{"train_loss": 0.05378738418221474, "global_step": 238959, "epoch": 2684, "lr": 3.5060555389451355e-06}
{"train_loss": 0.09606029838323593, "global_step": 238960, "epoch": 2684, "lr": 3.5058422767974332e-06}
{"train_loss": 0.011661065742373466, "global_step": 238961, "epoch": 2684, "lr": 3.505629020900303e-06}
{"train_loss": 0.06229499727487564, "global_step": 238962, "epoch": 2684, "lr": 3.505415771253762e-06}
{"train_loss": 0.02315203845500946, "global_step": 238963, "epoch": 2684, "lr": 3.5052025278578647e-06}
{"train_loss": 0.0522401144079278, "global_step": 238964, "epoch": 2684, "lr": 3.5049892907126124e-06, "val_loss": 9.307555198669434}
{"train_loss": 0.01521613821387291, "global_step": 238965, "epoch": 2685, "lr": 3.50477605981806e-06}
{"train_loss": 0.034918881952762604, "global_step": 238966, "epoch": 2685, "lr": 3.5045628351742244e-06}
{"train_loss": 0.025726724416017532, "global_step": 238967, "epoch": 2685, "lr": 3.504349616781122e-06}
{"train_loss": 0.030642230063676834, "global_step": 238968, "epoch": 2685, "lr": 3.5041364046388026e-06}
{"train_loss": 0.06281734257936478, "global_step": 238969, "epoch": 2685, "lr": 3.5039231987472776e-06}
{"train_loss": 0.040830135345458984, "global_step": 238970, "epoch": 2685, "lr": 3.5037099991065914e-06}
{"train_loss": 0.0398731604218483, "global_step": 238971, "epoch": 2685, "lr": 3.5034968057167607e-06}
{"train_loss": 0.052915748208761215, "global_step": 238972, "epoch": 2685, "lr": 3.5032836185778184e-06}
{"train_loss": 0.03535262495279312, "global_step": 238973, "epoch": 2685, "lr": 3.503070437689793e-06}
{"train_loss": 0.031910039484500885, "global_step": 238974, "epoch": 2685, "lr": 3.502857263052706e-06}
{"train_loss": 0.04960333928465843, "global_step": 238975, "epoch": 2685, "lr": 3.5026440946666016e-06}
{"train_loss": 0.017932552844285965, "global_step": 238976, "epoch": 2685, "lr": 3.5024309325314976e-06}
{"train_loss": 0.02752114087343216, "global_step": 238977, "epoch": 2685, "lr": 3.502217776647426e-06}
{"train_loss": 0.03935006633400917, "global_step": 238978, "epoch": 2685, "lr": 3.5020046270144157e-06}
{"train_loss": 0.044002506881952286, "global_step": 238979, "epoch": 2685, "lr": 3.5017914836324937e-06}
{"train_loss": 0.06752318888902664, "global_step": 238980, "epoch": 2685, "lr": 3.501578346501683e-06}
{"train_loss": 0.08067142963409424, "global_step": 238981, "epoch": 2685, "lr": 3.501365215622027e-06}
{"train_loss": 0.06717058271169662, "global_step": 238982, "epoch": 2685, "lr": 3.5011520909935426e-06}
{"train_loss": 0.0289162490516901, "global_step": 238983, "epoch": 2685, "lr": 3.500938972616263e-06}
{"train_loss": 0.025226959958672523, "global_step": 238984, "epoch": 2685, "lr": 3.5007258604902114e-06}
{"train_loss": 0.1087324321269989, "global_step": 238985, "epoch": 2685, "lr": 3.500512754615426e-06}
{"train_loss": 0.04577554017305374, "global_step": 238986, "epoch": 2685, "lr": 3.500299654991923e-06}
{"train_loss": 0.042349882423877716, "global_step": 238987, "epoch": 2685, "lr": 3.500086561619742e-06}
{"train_loss": 0.05620559677481651, "global_step": 238988, "epoch": 2685, "lr": 3.499873474498905e-06}
{"train_loss": 0.030597344040870667, "global_step": 238989, "epoch": 2685, "lr": 3.4996603936294504e-06}
{"train_loss": 0.04826321825385094, "global_step": 238990, "epoch": 2685, "lr": 3.4994473190113898e-06}
{"train_loss": 0.04124724119901657, "global_step": 238991, "epoch": 2685, "lr": 3.4992342506447617e-06}
{"train_loss": 0.02648496814072132, "global_step": 238992, "epoch": 2685, "lr": 3.4990211885296053e-06}
{"train_loss": 0.05506552755832672, "global_step": 238993, "epoch": 2685, "lr": 3.498808132665926e-06}
{"train_loss": 0.028067979961633682, "global_step": 238994, "epoch": 2685, "lr": 3.4985950830537793e-06}
{"train_loss": 0.03802570700645447, "global_step": 238995, "epoch": 2685, "lr": 3.4983820396931656e-06}
{"train_loss": 0.10635166615247726, "global_step": 238996, "epoch": 2685, "lr": 3.4981690025841396e-06}
{"train_loss": 0.06008128821849823, "global_step": 238997, "epoch": 2685, "lr": 3.497955971726713e-06}
{"train_loss": 0.05991613119840622, "global_step": 238998, "epoch": 2685, "lr": 3.497742947120919e-06}
{"train_loss": 0.036638177931308746, "global_step": 238999, "epoch": 2685, "lr": 3.497529928766785e-06}
{"train_loss": 0.03355532884597778, "global_step": 239000, "epoch": 2685, "lr": 3.4973169166643503e-06}
{"train_loss": 0.04005151614546776, "global_step": 239001, "epoch": 2685, "lr": 3.497103910813632e-06}
{"train_loss": 0.019914142787456512, "global_step": 239002, "epoch": 2685, "lr": 3.496890911214651e-06}
{"train_loss": 0.03564629703760147, "global_step": 239003, "epoch": 2685, "lr": 3.496677917867458e-06}
{"train_loss": 0.06118284910917282, "global_step": 239004, "epoch": 2685, "lr": 3.4964649307720586e-06}
{"train_loss": 0.04567509517073631, "global_step": 239005, "epoch": 2685, "lr": 3.496251949928503e-06}
{"train_loss": 0.046144213527441025, "global_step": 239006, "epoch": 2685, "lr": 3.496038975336807e-06}
{"train_loss": 0.07088898867368698, "global_step": 239007, "epoch": 2685, "lr": 3.495826006996994e-06}
{"train_loss": 0.04001360386610031, "global_step": 239008, "epoch": 2685, "lr": 3.4956130449091128e-06}
{"train_loss": 0.044381290674209595, "global_step": 239009, "epoch": 2685, "lr": 3.49540008907317e-06}
{"train_loss": 0.05173178389668465, "global_step": 239010, "epoch": 2685, "lr": 3.495187139489203e-06}
{"train_loss": 0.07431027293205261, "global_step": 239011, "epoch": 2685, "lr": 3.494974196157247e-06}
{"train_loss": 0.03368620201945305, "global_step": 239012, "epoch": 2685, "lr": 3.494761259077317e-06}
{"train_loss": 0.034629032015800476, "global_step": 239013, "epoch": 2685, "lr": 3.4945483282494585e-06}
{"train_loss": 0.04718506336212158, "global_step": 239014, "epoch": 2685, "lr": 3.4943354036736876e-06}
{"train_loss": 0.04512801021337509, "global_step": 239015, "epoch": 2685, "lr": 3.4941224853500375e-06}
{"train_loss": 0.03400418907403946, "global_step": 239016, "epoch": 2685, "lr": 3.493909573278531e-06}
{"train_loss": 0.08614196628332138, "global_step": 239017, "epoch": 2685, "lr": 3.493696667459212e-06}
{"train_loss": 0.045672960579395294, "global_step": 239018, "epoch": 2685, "lr": 3.4934837678920864e-06}
{"train_loss": 0.07754850387573242, "global_step": 239019, "epoch": 2685, "lr": 3.493270874577209e-06}
{"train_loss": 0.056310784071683884, "global_step": 239020, "epoch": 2685, "lr": 3.493057987514581e-06}
{"train_loss": 0.026687897741794586, "global_step": 239021, "epoch": 2685, "lr": 3.492845106704257e-06}
{"train_loss": 0.08315098285675049, "global_step": 239022, "epoch": 2685, "lr": 3.4926322321462367e-06}
{"train_loss": 0.057459402829408646, "global_step": 239023, "epoch": 2685, "lr": 3.492419363840582e-06}
{"train_loss": 0.035078711807727814, "global_step": 239024, "epoch": 2685, "lr": 3.492206501787293e-06}
{"train_loss": 0.017419729381799698, "global_step": 239025, "epoch": 2685, "lr": 3.491993645986419e-06}
{"train_loss": 0.11975519359111786, "global_step": 239026, "epoch": 2685, "lr": 3.491780796437971e-06}
{"train_loss": 0.010091365315020084, "global_step": 239027, "epoch": 2685, "lr": 3.4915679531419943e-06}
{"train_loss": 0.05485750362277031, "global_step": 239028, "epoch": 2685, "lr": 3.4913551160984993e-06}
{"train_loss": 0.051133494824171066, "global_step": 239029, "epoch": 2685, "lr": 3.491142285307525e-06}
{"train_loss": 0.02458982542157173, "global_step": 239030, "epoch": 2685, "lr": 3.49092946076911e-06}
{"train_loss": 0.09061327576637268, "global_step": 239031, "epoch": 2685, "lr": 3.490716642483266e-06}
{"train_loss": 0.06296870857477188, "global_step": 239032, "epoch": 2685, "lr": 3.4905038304500316e-06}
{"train_loss": 0.018504425883293152, "global_step": 239033, "epoch": 2685, "lr": 3.4902910246694287e-06}
{"train_loss": 0.04662003368139267, "global_step": 239034, "epoch": 2685, "lr": 3.4900782251414964e-06}
{"train_loss": 0.03603485971689224, "global_step": 239035, "epoch": 2685, "lr": 3.489865431866246e-06}
{"train_loss": 0.03980182856321335, "global_step": 239036, "epoch": 2685, "lr": 3.4896526448437274e-06}
{"train_loss": 0.03992448002099991, "global_step": 239037, "epoch": 2685, "lr": 3.4894398640739457e-06}
{"train_loss": 0.025690997019410133, "global_step": 239038, "epoch": 2685, "lr": 3.4892270895569567e-06}
{"train_loss": 0.037007544189691544, "global_step": 239039, "epoch": 2685, "lr": 3.489014321292766e-06}
{"train_loss": 0.029238522052764893, "global_step": 239040, "epoch": 2685, "lr": 3.488801559281407e-06}
{"train_loss": 0.04147737845778465, "global_step": 239041, "epoch": 2685, "lr": 3.4885888035229184e-06}
{"train_loss": 0.041258472949266434, "global_step": 239042, "epoch": 2685, "lr": 3.4883760540173173e-06}
{"train_loss": 0.04499281570315361, "global_step": 239043, "epoch": 2685, "lr": 3.4881633107646416e-06}
{"train_loss": 0.04126027598977089, "global_step": 239044, "epoch": 2685, "lr": 3.48795057376492e-06}
{"train_loss": 0.049068745225667953, "global_step": 239045, "epoch": 2685, "lr": 3.4877378430181627e-06}
{"train_loss": 0.04068349301815033, "global_step": 239046, "epoch": 2685, "lr": 3.487525118524415e-06}
{"train_loss": 0.06085262820124626, "global_step": 239047, "epoch": 2685, "lr": 3.48731240028371e-06}
{"train_loss": 0.08565143495798111, "global_step": 239048, "epoch": 2685, "lr": 3.487099688296064e-06}
{"train_loss": 0.03463294357061386, "global_step": 239049, "epoch": 2685, "lr": 3.486886982561516e-06}
{"train_loss": 0.04841243848204613, "global_step": 239050, "epoch": 2685, "lr": 3.486674283080077e-06}
{"train_loss": 0.023670118302106857, "global_step": 239051, "epoch": 2685, "lr": 3.486461589851803e-06}
{"train_loss": 0.06639136373996735, "global_step": 239052, "epoch": 2685, "lr": 3.486248902876693e-06}
{"train_loss": 0.0468719613455822, "global_step": 239053, "epoch": 2685, "lr": 3.486036222154804e-06, "val_loss": 9.453457832336426, "train_action_mse_error": 15.180220603942871}
{"train_loss": 0.02052357606589794, "global_step": 239054, "epoch": 2686, "lr": 3.4858235476861345e-06}
{"train_loss": 0.046364009380340576, "global_step": 239055, "epoch": 2686, "lr": 3.4856108794707465e-06}
{"train_loss": 0.02812791056931019, "global_step": 239056, "epoch": 2686, "lr": 3.4853982175086343e-06}
{"train_loss": 0.03998162969946861, "global_step": 239057, "epoch": 2686, "lr": 3.4851855617998584e-06}
{"train_loss": 0.03834867477416992, "global_step": 239058, "epoch": 2686, "lr": 3.4849729123444198e-06}
{"train_loss": 0.05735564976930618, "global_step": 239059, "epoch": 2686, "lr": 3.484760269142373e-06}
{"train_loss": 0.047643110156059265, "global_step": 239060, "epoch": 2686, "lr": 3.484547632193719e-06}
{"train_loss": 0.04728937894105911, "global_step": 239061, "epoch": 2686, "lr": 3.484335001498512e-06}
{"train_loss": 0.03922136873006821, "global_step": 239062, "epoch": 2686, "lr": 3.484122377056759e-06}
{"train_loss": 0.10322711616754532, "global_step": 239063, "epoch": 2686, "lr": 3.483909758868509e-06}
{"train_loss": 0.051955826580524445, "global_step": 239064, "epoch": 2686, "lr": 3.483697146933773e-06}
{"train_loss": 0.04700932279229164, "global_step": 239065, "epoch": 2686, "lr": 3.483484541252585e-06}
{"train_loss": 0.020549634471535683, "global_step": 239066, "epoch": 2686, "lr": 3.4832719418249838e-06}
{"train_loss": 0.030048226937651634, "global_step": 239067, "epoch": 2686, "lr": 3.4830593486509856e-06}
{"train_loss": 0.0358906090259552, "global_step": 239068, "epoch": 2686, "lr": 3.4828467617306236e-06}
{"train_loss": 0.04018921032547951, "global_step": 239069, "epoch": 2686, "lr": 3.482634181063926e-06}
{"train_loss": 0.04075905680656433, "global_step": 239070, "epoch": 2686, "lr": 3.482421606650921e-06}
{"train_loss": 0.06881345808506012, "global_step": 239071, "epoch": 2686, "lr": 3.482209038491635e-06}
{"train_loss": 0.07117929309606552, "global_step": 239072, "epoch": 2686, "lr": 3.481996476586108e-06}
{"train_loss": 0.03898918628692627, "global_step": 239073, "epoch": 2686, "lr": 3.4817839209343506e-06}
{"train_loss": 0.02570599690079689, "global_step": 239074, "epoch": 2686, "lr": 3.4815713715364017e-06}
{"train_loss": 0.05034218728542328, "global_step": 239075, "epoch": 2686, "lr": 3.4813588283922893e-06}
{"train_loss": 0.052959948778152466, "global_step": 239076, "epoch": 2686, "lr": 3.4811462915020465e-06}
{"train_loss": 0.09707601368427277, "global_step": 239077, "epoch": 2686, "lr": 3.4809337608656957e-06}
{"train_loss": 0.09315463155508041, "global_step": 239078, "epoch": 2686, "lr": 3.4807212364832585e-06}
{"train_loss": 0.037230320274829865, "global_step": 239079, "epoch": 2686, "lr": 3.48050871835478e-06}
{"train_loss": 0.03473550081253052, "global_step": 239080, "epoch": 2686, "lr": 3.4802962064802714e-06}
{"train_loss": 0.06682652980089188, "global_step": 239081, "epoch": 2686, "lr": 3.4800837008597762e-06}
{"train_loss": 0.04491062089800835, "global_step": 239082, "epoch": 2686, "lr": 3.479871201493312e-06}
{"train_loss": 0.03622899204492569, "global_step": 239083, "epoch": 2686, "lr": 3.4796587083809175e-06}
{"train_loss": 0.04723210260272026, "global_step": 239084, "epoch": 2686, "lr": 3.4794462215226087e-06}
{"train_loss": 0.028579218313097954, "global_step": 239085, "epoch": 2686, "lr": 3.4792337409184307e-06}
{"train_loss": 0.07321856915950775, "global_step": 239086, "epoch": 2686, "lr": 3.479021266568394e-06}
{"train_loss": 0.03879225254058838, "global_step": 239087, "epoch": 2686, "lr": 3.4788087984725436e-06}
{"train_loss": 0.05954568460583687, "global_step": 239088, "epoch": 2686, "lr": 3.478596336630896e-06}
{"train_loss": 0.04089226946234703, "global_step": 239089, "epoch": 2686, "lr": 3.4783838810434844e-06}
{"train_loss": 0.06701228767633438, "global_step": 239090, "epoch": 2686, "lr": 3.4781714317103365e-06}
{"train_loss": 0.05860564857721329, "global_step": 239091, "epoch": 2686, "lr": 3.477958988631486e-06}
{"train_loss": 0.03123132884502411, "global_step": 239092, "epoch": 2686, "lr": 3.4777465518069485e-06}
{"train_loss": 0.03630247339606285, "global_step": 239093, "epoch": 2686, "lr": 3.4775341212367696e-06}
{"train_loss": 0.014049510471522808, "global_step": 239094, "epoch": 2686, "lr": 3.47732169692096e-06}
{"train_loss": 0.05148950219154358, "global_step": 239095, "epoch": 2686, "lr": 3.4771092788595696e-06}
{"train_loss": 0.036705899983644485, "global_step": 239096, "epoch": 2686, "lr": 3.4768968670526036e-06}
{"train_loss": 0.02758694253861904, "global_step": 239097, "epoch": 2686, "lr": 3.4766844615001073e-06}
{"train_loss": 0.088603176176548, "global_step": 239098, "epoch": 2686, "lr": 3.4764720622020963e-06}
{"train_loss": 0.05238433927297592, "global_step": 239099, "epoch": 2686, "lr": 3.4762596691586215e-06}
{"train_loss": 0.029032111167907715, "global_step": 239100, "epoch": 2686, "lr": 3.476047282369682e-06}
{"train_loss": 0.08546747267246246, "global_step": 239101, "epoch": 2686, "lr": 3.4758349018353233e-06}
{"train_loss": 0.0306050144135952, "global_step": 239102, "epoch": 2686, "lr": 3.4756225275555777e-06}
{"train_loss": 0.06804469972848892, "global_step": 239103, "epoch": 2686, "lr": 3.4754101595304623e-06}
{"train_loss": 0.039163921028375626, "global_step": 239104, "epoch": 2686, "lr": 3.4751977977600157e-06}
{"train_loss": 0.05261484161019325, "global_step": 239105, "epoch": 2686, "lr": 3.474985442244255e-06}
{"train_loss": 0.027220701798796654, "global_step": 239106, "epoch": 2686, "lr": 3.4747730929832244e-06}
{"train_loss": 0.040621791034936905, "global_step": 239107, "epoch": 2686, "lr": 3.474560749976935e-06}
{"train_loss": 0.02410164289176464, "global_step": 239108, "epoch": 2686, "lr": 3.474348413225431e-06}
{"train_loss": 0.0717364028096199, "global_step": 239109, "epoch": 2686, "lr": 3.4741360827287297e-06}
{"train_loss": 0.07909362763166428, "global_step": 239110, "epoch": 2686, "lr": 3.473923758486869e-06}
{"train_loss": 0.03042825683951378, "global_step": 239111, "epoch": 2686, "lr": 3.4737114404998607e-06}
{"train_loss": 0.0804482251405716, "global_step": 239112, "epoch": 2686, "lr": 3.4734991287677596e-06}
{"train_loss": 0.04143694043159485, "global_step": 239113, "epoch": 2686, "lr": 3.473286823290567e-06}
{"train_loss": 0.01879286952316761, "global_step": 239114, "epoch": 2686, "lr": 3.4730745240683316e-06}
{"train_loss": 0.012707183137536049, "global_step": 239115, "epoch": 2686, "lr": 3.472862231101065e-06}
{"train_loss": 0.05449382960796356, "global_step": 239116, "epoch": 2686, "lr": 3.4726499443888173e-06}
{"train_loss": 0.04401368647813797, "global_step": 239117, "epoch": 2686, "lr": 3.472437663931605e-06}
{"train_loss": 0.04632002115249634, "global_step": 239118, "epoch": 2686, "lr": 3.4722253897294445e-06}
{"train_loss": 0.055157389491796494, "global_step": 239119, "epoch": 2686, "lr": 3.4720131217823857e-06}
{"train_loss": 0.0312297735363245, "global_step": 239120, "epoch": 2686, "lr": 3.4718008600904408e-06}
{"train_loss": 0.04478190839290619, "global_step": 239121, "epoch": 2686, "lr": 3.471588604653653e-06}
{"train_loss": 0.019957423210144043, "global_step": 239122, "epoch": 2686, "lr": 3.4713763554720335e-06}
{"train_loss": 0.017700424417853355, "global_step": 239123, "epoch": 2686, "lr": 3.4711641125456274e-06}
{"train_loss": 0.061042916029691696, "global_step": 239124, "epoch": 2686, "lr": 3.470951875874451e-06}
{"train_loss": 0.028841180726885796, "global_step": 239125, "epoch": 2686, "lr": 3.4707396454585428e-06}
{"train_loss": 0.049144476652145386, "global_step": 239126, "epoch": 2686, "lr": 3.47052742129792e-06}
{"train_loss": 0.030273405835032463, "global_step": 239127, "epoch": 2686, "lr": 3.4703152033926267e-06}
{"train_loss": 0.04504008963704109, "global_step": 239128, "epoch": 2686, "lr": 3.4701029917426742e-06}
{"train_loss": 0.05533896014094353, "global_step": 239129, "epoch": 2686, "lr": 3.4698907863481067e-06}
{"train_loss": 0.05192796513438225, "global_step": 239130, "epoch": 2686, "lr": 3.469678587208941e-06}
{"train_loss": 0.05097031593322754, "global_step": 239131, "epoch": 2686, "lr": 3.469466394325216e-06}
{"train_loss": 0.04395279288291931, "global_step": 239132, "epoch": 2686, "lr": 3.4692542076969425e-06}
{"train_loss": 0.03389650583267212, "global_step": 239133, "epoch": 2686, "lr": 3.4690420273241708e-06}
{"train_loss": 0.03661467880010605, "global_step": 239134, "epoch": 2686, "lr": 3.468829853206912e-06}
{"train_loss": 0.09562327712774277, "global_step": 239135, "epoch": 2686, "lr": 3.46861768534521e-06}
{"train_loss": 0.028407203033566475, "global_step": 239136, "epoch": 2686, "lr": 3.4684055237390713e-06}
{"train_loss": 0.04035019874572754, "global_step": 239137, "epoch": 2686, "lr": 3.4681933683885448e-06}
{"train_loss": 0.022510036826133728, "global_step": 239138, "epoch": 2686, "lr": 3.467981219293659e-06}
{"train_loss": 0.037285491824150085, "global_step": 239139, "epoch": 2686, "lr": 3.4677690764544303e-06}
{"train_loss": 0.0593729205429554, "global_step": 239140, "epoch": 2686, "lr": 3.4675569398708975e-06}
{"train_loss": 0.05605452135205269, "global_step": 239141, "epoch": 2686, "lr": 3.4673448095430773e-06}
{"train_loss": 0.04602045273889652, "global_step": 239142, "epoch": 2686, "lr": 3.467132685471014e-06, "val_loss": 9.518882751464844}
{"train_loss": 0.05336310341954231, "global_step": 239143, "epoch": 2687, "lr": 3.466920567654719e-06}
{"train_loss": 0.09103185683488846, "global_step": 239144, "epoch": 2687, "lr": 3.4667084560942366e-06}
{"train_loss": 0.0713462233543396, "global_step": 239145, "epoch": 2687, "lr": 3.4664963507895832e-06}
{"train_loss": 0.03899722173810005, "global_step": 239146, "epoch": 2687, "lr": 3.466284251740798e-06}
{"train_loss": 0.018317310139536858, "global_step": 239147, "epoch": 2687, "lr": 3.4660721589478974e-06}
{"train_loss": 0.033151645213365555, "global_step": 239148, "epoch": 2687, "lr": 3.4658600724109204e-06}
{"train_loss": 0.030622512102127075, "global_step": 239149, "epoch": 2687, "lr": 3.465647992129889e-06}
{"train_loss": 0.061359260231256485, "global_step": 239150, "epoch": 2687, "lr": 3.465435918104837e-06}
{"train_loss": 0.05676412954926491, "global_step": 239151, "epoch": 2687, "lr": 3.46522385033578e-06}
{"train_loss": 0.08214949071407318, "global_step": 239152, "epoch": 2687, "lr": 3.4650117888227695e-06}
{"train_loss": 0.05007651075720787, "global_step": 239153, "epoch": 2687, "lr": 3.4647997335658155e-06}
{"train_loss": 0.07271188497543335, "global_step": 239154, "epoch": 2687, "lr": 3.4645876845649515e-06}
{"train_loss": 0.029042543843388557, "global_step": 239155, "epoch": 2687, "lr": 3.464375641820211e-06}
{"train_loss": 0.04979603365063667, "global_step": 239156, "epoch": 2687, "lr": 3.4641636053316107e-06}
{"train_loss": 0.033929355442523956, "global_step": 239157, "epoch": 2687, "lr": 3.4639515750991947e-06}
{"train_loss": 0.06465210765600204, "global_step": 239158, "epoch": 2687, "lr": 3.4637395511229743e-06}
{"train_loss": 0.04358290135860443, "global_step": 239159, "epoch": 2687, "lr": 3.4635275334029994e-06}
{"train_loss": 0.03398749977350235, "global_step": 239160, "epoch": 2687, "lr": 3.46331552193927e-06}
{"train_loss": 0.06720077246427536, "global_step": 239161, "epoch": 2687, "lr": 3.4631035167318415e-06}
{"train_loss": 0.027819814160466194, "global_step": 239162, "epoch": 2687, "lr": 3.4628915177807255e-06}
{"train_loss": 0.029439769685268402, "global_step": 239163, "epoch": 2687, "lr": 3.4626795250859654e-06}
{"train_loss": 0.04158417135477066, "global_step": 239164, "epoch": 2687, "lr": 3.462467538647568e-06}
{"train_loss": 0.04832451045513153, "global_step": 239165, "epoch": 2687, "lr": 3.462255558465588e-06}
{"train_loss": 0.028287256136536598, "global_step": 239166, "epoch": 2687, "lr": 3.4620435845400256e-06}
{"train_loss": 0.03131682798266411, "global_step": 239167, "epoch": 2687, "lr": 3.461831616870942e-06}
{"train_loss": 0.03284063935279846, "global_step": 239168, "epoch": 2687, "lr": 3.4616196554583315e-06}
{"train_loss": 0.020071642473340034, "global_step": 239169, "epoch": 2687, "lr": 3.4614077003022493e-06}
{"train_loss": 0.05763985961675644, "global_step": 239170, "epoch": 2687, "lr": 3.461195751402707e-06}
{"train_loss": 0.05119314417243004, "global_step": 239171, "epoch": 2687, "lr": 3.460983808759749e-06}
{"train_loss": 0.018550066277384758, "global_step": 239172, "epoch": 2687, "lr": 3.4607718723733806e-06}
{"train_loss": 0.03477262705564499, "global_step": 239173, "epoch": 2687, "lr": 3.4605599422436517e-06}
{"train_loss": 0.031910572201013565, "global_step": 239174, "epoch": 2687, "lr": 3.460348018370585e-06}
{"train_loss": 0.04330367222428322, "global_step": 239175, "epoch": 2687, "lr": 3.460136100754202e-06}
{"train_loss": 0.036113373935222626, "global_step": 239176, "epoch": 2687, "lr": 3.4599241893945476e-06}
{"train_loss": 0.06850052624940872, "global_step": 239177, "epoch": 2687, "lr": 3.4597122842916273e-06}
{"train_loss": 0.06058865040540695, "global_step": 239178, "epoch": 2687, "lr": 3.459500385445491e-06}
{"train_loss": 0.03253417834639549, "global_step": 239179, "epoch": 2687, "lr": 3.4592884928561443e-06}
{"train_loss": 0.035827092826366425, "global_step": 239180, "epoch": 2687, "lr": 3.4590766065236422e-06}
{"train_loss": 0.01967465691268444, "global_step": 239181, "epoch": 2687, "lr": 3.4588647264479913e-06}
{"train_loss": 0.042592503130435944, "global_step": 239182, "epoch": 2687, "lr": 3.458652852629235e-06}
{"train_loss": 0.02331705577671528, "global_step": 239183, "epoch": 2687, "lr": 3.4584409850673905e-06}
{"train_loss": 0.04712780565023422, "global_step": 239184, "epoch": 2687, "lr": 3.458229123762502e-06}
{"train_loss": 0.03642701730132103, "global_step": 239185, "epoch": 2687, "lr": 3.458017268714575e-06}
{"train_loss": 0.046463705599308014, "global_step": 239186, "epoch": 2687, "lr": 3.4578054199236543e-06}
{"train_loss": 0.028582079336047173, "global_step": 239187, "epoch": 2687, "lr": 3.457593577389762e-06}
{"train_loss": 0.024592800065875053, "global_step": 239188, "epoch": 2687, "lr": 3.457381741112936e-06}
{"train_loss": 0.018413184210658073, "global_step": 239189, "epoch": 2687, "lr": 3.457169911093189e-06}
{"train_loss": 0.03146840259432793, "global_step": 239190, "epoch": 2687, "lr": 3.4569580873305696e-06}
{"train_loss": 0.1283566802740097, "global_step": 239191, "epoch": 2687, "lr": 3.456746269825084e-06}
{"train_loss": 0.07765529304742813, "global_step": 239192, "epoch": 2687, "lr": 3.4565344585767766e-06}
{"train_loss": 0.048826709389686584, "global_step": 239193, "epoch": 2687, "lr": 3.456322653585675e-06}
{"train_loss": 0.05670357495546341, "global_step": 239194, "epoch": 2687, "lr": 3.4561108548517963e-06}
{"train_loss": 0.04874959588050842, "global_step": 239195, "epoch": 2687, "lr": 3.4558990623751784e-06}
{"train_loss": 0.07317659258842468, "global_step": 239196, "epoch": 2687, "lr": 3.4556872761558445e-06}
{"train_loss": 0.05397285521030426, "global_step": 239197, "epoch": 2687, "lr": 3.4554754961938327e-06}
{"train_loss": 0.04419715330004692, "global_step": 239198, "epoch": 2687, "lr": 3.4552637224891603e-06}
{"train_loss": 0.02367904596030712, "global_step": 239199, "epoch": 2687, "lr": 3.455051955041866e-06}
{"train_loss": 0.019230255857110023, "global_step": 239200, "epoch": 2687, "lr": 3.4548401938519605e-06}
{"train_loss": 0.023640397936105728, "global_step": 239201, "epoch": 2687, "lr": 3.4546284389194995e-06}
{"train_loss": 0.026765309274196625, "global_step": 239202, "epoch": 2687, "lr": 3.454416690244483e-06}
{"train_loss": 0.0676310658454895, "global_step": 239203, "epoch": 2687, "lr": 3.454204947826961e-06}
{"train_loss": 0.05841798707842827, "global_step": 239204, "epoch": 2687, "lr": 3.4539932116669503e-06}
{"train_loss": 0.04714420437812805, "global_step": 239205, "epoch": 2687, "lr": 3.45378148176449e-06}
{"train_loss": 0.09966310113668442, "global_step": 239206, "epoch": 2687, "lr": 3.45356975811959e-06}
{"train_loss": 0.07712829858064651, "global_step": 239207, "epoch": 2687, "lr": 3.453358040732302e-06}
{"train_loss": 0.034361548721790314, "global_step": 239208, "epoch": 2687, "lr": 3.45314632960263e-06}
{"train_loss": 0.0079185850918293, "global_step": 239209, "epoch": 2687, "lr": 3.4529346247306195e-06}
{"train_loss": 0.05904123932123184, "global_step": 239210, "epoch": 2687, "lr": 3.4527229261162976e-06}
{"train_loss": 0.03603190928697586, "global_step": 239211, "epoch": 2687, "lr": 3.4525112337596864e-06}
{"train_loss": 0.028516631573438644, "global_step": 239212, "epoch": 2687, "lr": 3.4522995476608256e-06}
{"train_loss": 0.061566539108753204, "global_step": 239213, "epoch": 2687, "lr": 3.4520878678197255e-06}
{"train_loss": 0.0627945214509964, "global_step": 239214, "epoch": 2687, "lr": 3.4518761942364363e-06}
{"train_loss": 0.07022135704755783, "global_step": 239215, "epoch": 2687, "lr": 3.451664526910964e-06}
{"train_loss": 0.07915440201759338, "global_step": 239216, "epoch": 2687, "lr": 3.4514528658433576e-06}
{"train_loss": 0.0583738312125206, "global_step": 239217, "epoch": 2687, "lr": 3.4512412110336236e-06}
{"train_loss": 0.08701223880052567, "global_step": 239218, "epoch": 2687, "lr": 3.451029562481817e-06}
{"train_loss": 0.03956518694758415, "global_step": 239219, "epoch": 2687, "lr": 3.4508179201879374e-06}
{"train_loss": 0.023266460746526718, "global_step": 239220, "epoch": 2687, "lr": 3.4506062841520415e-06}
{"train_loss": 0.08746877312660217, "global_step": 239221, "epoch": 2687, "lr": 3.450394654374134e-06}
{"train_loss": 0.031086374074220657, "global_step": 239222, "epoch": 2687, "lr": 3.450183030854265e-06}
{"train_loss": 0.0411674827337265, "global_step": 239223, "epoch": 2687, "lr": 3.44997141359244e-06}
{"train_loss": 0.05831526219844818, "global_step": 239224, "epoch": 2687, "lr": 3.449759802588709e-06}
{"train_loss": 0.06829833984375, "global_step": 239225, "epoch": 2687, "lr": 3.4495481978430832e-06}
{"train_loss": 0.048163894563913345, "global_step": 239226, "epoch": 2687, "lr": 3.4493365993556014e-06}
{"train_loss": 0.07273012399673462, "global_step": 239227, "epoch": 2687, "lr": 3.4491250071262857e-06}
{"train_loss": 0.0769583135843277, "global_step": 239228, "epoch": 2687, "lr": 3.44891342115517e-06}
{"train_loss": 0.045248135924339294, "global_step": 239229, "epoch": 2687, "lr": 3.448701841442281e-06}
{"train_loss": 0.05431405454874039, "global_step": 239230, "epoch": 2687, "lr": 3.4484902679876472e-06}
{"train_loss": 0.0483710447096088, "global_step": 239231, "epoch": 2687, "lr": 3.448278700791291e-06, "val_loss": 9.448782920837402}
{"train_loss": 0.040290214121341705, "global_step": 239232, "epoch": 2688, "lr": 3.448067139853256e-06}
{"train_loss": 0.030649080872535706, "global_step": 239233, "epoch": 2688, "lr": 3.4478555851735594e-06}
{"train_loss": 0.02669598162174225, "global_step": 239234, "epoch": 2688, "lr": 3.447644036752223e-06}
{"train_loss": 0.07403361052274704, "global_step": 239235, "epoch": 2688, "lr": 3.447432494589292e-06}
{"train_loss": 0.0343019999563694, "global_step": 239236, "epoch": 2688, "lr": 3.447220958684777e-06}
{"train_loss": 0.027012860402464867, "global_step": 239237, "epoch": 2688, "lr": 3.4470094290387276e-06}
{"train_loss": 0.06125425547361374, "global_step": 239238, "epoch": 2688, "lr": 3.4467979056511503e-06}
{"train_loss": 0.04265663027763367, "global_step": 239239, "epoch": 2688, "lr": 3.4465863885220886e-06}
{"train_loss": 0.030064111575484276, "global_step": 239240, "epoch": 2688, "lr": 3.446374877651565e-06}
{"train_loss": 0.08082519471645355, "global_step": 239241, "epoch": 2688, "lr": 3.4461633730396125e-06}
{"train_loss": 0.03970973566174507, "global_step": 239242, "epoch": 2688, "lr": 3.4459518746862486e-06}
{"train_loss": 0.07201513648033142, "global_step": 239243, "epoch": 2688, "lr": 3.4457403825915115e-06}
{"train_loss": 0.0522749125957489, "global_step": 239244, "epoch": 2688, "lr": 3.4455288967554235e-06}
{"train_loss": 0.05159716308116913, "global_step": 239245, "epoch": 2688, "lr": 3.4453174171780178e-06}
{"train_loss": 0.08123163878917694, "global_step": 239246, "epoch": 2688, "lr": 3.4451059438593335e-06}
{"train_loss": 0.06875835359096527, "global_step": 239247, "epoch": 2688, "lr": 3.4448944767993705e-06}
{"train_loss": 0.01777711696922779, "global_step": 239248, "epoch": 2688, "lr": 3.4446830159981904e-06}
{"train_loss": 0.05713178962469101, "global_step": 239249, "epoch": 2688, "lr": 3.444471561455792e-06}
{"train_loss": 0.03308575227856636, "global_step": 239250, "epoch": 2688, "lr": 3.444260113172226e-06}
{"train_loss": 0.04861001670360565, "global_step": 239251, "epoch": 2688, "lr": 3.444048671147504e-06}
{"train_loss": 0.044102657586336136, "global_step": 239252, "epoch": 2688, "lr": 3.4438372353816693e-06}
{"train_loss": 0.03940078988671303, "global_step": 239253, "epoch": 2688, "lr": 3.4436258058747397e-06}
{"train_loss": 0.040691398084163666, "global_step": 239254, "epoch": 2688, "lr": 3.4434143826267474e-06}
{"train_loss": 0.02406042069196701, "global_step": 239255, "epoch": 2688, "lr": 3.443202965637721e-06}
{"train_loss": 0.014599149115383625, "global_step": 239256, "epoch": 2688, "lr": 3.4429915549076872e-06}
{"train_loss": 0.0836656391620636, "global_step": 239257, "epoch": 2688, "lr": 3.4427801504366752e-06}
{"train_loss": 0.030430972576141357, "global_step": 239258, "epoch": 2688, "lr": 3.442568752224723e-06}
{"train_loss": 0.06577620655298233, "global_step": 239259, "epoch": 2688, "lr": 3.4423573602718364e-06}
{"train_loss": 0.02167571522295475, "global_step": 239260, "epoch": 2688, "lr": 3.442145974578065e-06}
{"train_loss": 0.02418568544089794, "global_step": 239261, "epoch": 2688, "lr": 3.441934595143426e-06}
{"train_loss": 0.04807993769645691, "global_step": 239262, "epoch": 2688, "lr": 3.4417232219679574e-06}
{"train_loss": 0.016454286873340607, "global_step": 239263, "epoch": 2688, "lr": 3.4415118550516765e-06}
{"train_loss": 0.051376644521951675, "global_step": 239264, "epoch": 2688, "lr": 3.4413004943946113e-06}
{"train_loss": 0.02314317598938942, "global_step": 239265, "epoch": 2688, "lr": 3.441089139996806e-06}
{"train_loss": 0.06677775084972382, "global_step": 239266, "epoch": 2688, "lr": 3.4408777918582824e-06}
{"train_loss": 0.07741991430521011, "global_step": 239267, "epoch": 2688, "lr": 3.440666449979052e-06}
{"train_loss": 0.04517599567770958, "global_step": 239268, "epoch": 2688, "lr": 3.4404551143591703e-06}
{"train_loss": 0.05286051705479622, "global_step": 239269, "epoch": 2688, "lr": 3.4402437849986367e-06}
{"train_loss": 0.041230589151382446, "global_step": 239270, "epoch": 2688, "lr": 3.4400324618975076e-06}
{"train_loss": 0.03451308235526085, "global_step": 239271, "epoch": 2688, "lr": 3.439821145055799e-06}
{"train_loss": 0.013023348525166512, "global_step": 239272, "epoch": 2688, "lr": 3.4396098344735282e-06}
{"train_loss": 0.08195993304252625, "global_step": 239273, "epoch": 2688, "lr": 3.4393985301507448e-06}
{"train_loss": 0.008804078213870525, "global_step": 239274, "epoch": 2688, "lr": 3.4391872320874594e-06}
{"train_loss": 0.06503556668758392, "global_step": 239275, "epoch": 2688, "lr": 3.438975940283712e-06}
{"train_loss": 0.07809991389513016, "global_step": 239276, "epoch": 2688, "lr": 3.4387646547395236e-06}
{"train_loss": 0.05301523953676224, "global_step": 239277, "epoch": 2688, "lr": 3.4385533754549282e-06}
{"train_loss": 0.06891673058271408, "global_step": 239278, "epoch": 2688, "lr": 3.438342102429948e-06}
{"train_loss": 0.08137515187263489, "global_step": 239279, "epoch": 2688, "lr": 3.4381308356646213e-06}
{"train_loss": 0.040251556783914566, "global_step": 239280, "epoch": 2688, "lr": 3.4379195751589653e-06}
{"train_loss": 0.04648739844560623, "global_step": 239281, "epoch": 2688, "lr": 3.437708320913019e-06}
{"train_loss": 0.06996090710163116, "global_step": 239282, "epoch": 2688, "lr": 3.4374970729267986e-06}
{"train_loss": 0.08232349902391434, "global_step": 239283, "epoch": 2688, "lr": 3.4372858312003373e-06}
{"train_loss": 0.008379795588552952, "global_step": 239284, "epoch": 2688, "lr": 3.437074595733675e-06}
{"train_loss": 0.031725116074085236, "global_step": 239285, "epoch": 2688, "lr": 3.4368633665268214e-06}
{"train_loss": 0.05062435194849968, "global_step": 239286, "epoch": 2688, "lr": 3.4366521435798214e-06}
{"train_loss": 0.03309568390250206, "global_step": 239287, "epoch": 2688, "lr": 3.436440926892692e-06}
{"train_loss": 0.006389951799064875, "global_step": 239288, "epoch": 2688, "lr": 3.4362297164654667e-06}
{"train_loss": 0.022560106590390205, "global_step": 239289, "epoch": 2688, "lr": 3.4360185122981726e-06}
{"train_loss": 0.023793471977114677, "global_step": 239290, "epoch": 2688, "lr": 3.4358073143908433e-06}
{"train_loss": 0.011109215207397938, "global_step": 239291, "epoch": 2688, "lr": 3.43559612274349e-06}
{"train_loss": 0.011643793433904648, "global_step": 239292, "epoch": 2688, "lr": 3.4353849373561677e-06}
{"train_loss": 0.04480215162038803, "global_step": 239293, "epoch": 2688, "lr": 3.435173758228877e-06}
{"train_loss": 0.029951514676213264, "global_step": 239294, "epoch": 2688, "lr": 3.4349625853616672e-06}
{"train_loss": 0.02424251101911068, "global_step": 239295, "epoch": 2688, "lr": 3.434751418754556e-06}
{"train_loss": 0.02576991729438305, "global_step": 239296, "epoch": 2688, "lr": 3.4345402584075813e-06}
{"train_loss": 0.039873816072940826, "global_step": 239297, "epoch": 2688, "lr": 3.434329104320755e-06}
{"train_loss": 0.01898913085460663, "global_step": 239298, "epoch": 2688, "lr": 3.434117956494126e-06}
{"train_loss": 0.01942489482462406, "global_step": 239299, "epoch": 2688, "lr": 3.4339068149277008e-06}
{"train_loss": 0.04965947940945625, "global_step": 239300, "epoch": 2688, "lr": 3.4336956796215235e-06}
{"train_loss": 0.03631896153092384, "global_step": 239301, "epoch": 2688, "lr": 3.4334845505756274e-06}
{"train_loss": 0.03184840828180313, "global_step": 239302, "epoch": 2688, "lr": 3.433273427790018e-06}
{"train_loss": 0.03791411593556404, "global_step": 239303, "epoch": 2688, "lr": 3.4330623112647507e-06}
{"train_loss": 0.0752476304769516, "global_step": 239304, "epoch": 2688, "lr": 3.432851200999837e-06}
{"train_loss": 0.038521915674209595, "global_step": 239305, "epoch": 2688, "lr": 3.4326400969953045e-06}
{"train_loss": 0.0928046777844429, "global_step": 239306, "epoch": 2688, "lr": 3.4324289992511917e-06}
{"train_loss": 0.054143570363521576, "global_step": 239307, "epoch": 2688, "lr": 3.43221790776751e-06}
{"train_loss": 0.03136416897177696, "global_step": 239308, "epoch": 2688, "lr": 3.432006822544309e-06}
{"train_loss": 0.027067452669143677, "global_step": 239309, "epoch": 2688, "lr": 3.4317957435816116e-06}
{"train_loss": 0.07835523039102554, "global_step": 239310, "epoch": 2688, "lr": 3.4315846708794285e-06}
{"train_loss": 0.06387209892272949, "global_step": 239311, "epoch": 2688, "lr": 3.4313736044378096e-06}
{"train_loss": 0.03766297921538353, "global_step": 239312, "epoch": 2688, "lr": 3.431162544256772e-06}
{"train_loss": 0.0316455140709877, "global_step": 239313, "epoch": 2688, "lr": 3.430951490336348e-06}
{"train_loss": 0.04025935009121895, "global_step": 239314, "epoch": 2688, "lr": 3.430740442676561e-06}
{"train_loss": 0.036349810659885406, "global_step": 239315, "epoch": 2688, "lr": 3.430529401277455e-06}
{"train_loss": 0.08388520032167435, "global_step": 239316, "epoch": 2688, "lr": 3.4303183661390294e-06}
{"train_loss": 0.028101490810513496, "global_step": 239317, "epoch": 2688, "lr": 3.430107337261346e-06}
{"train_loss": 0.06008591875433922, "global_step": 239318, "epoch": 2688, "lr": 3.429896314644404e-06}
{"train_loss": 0.031339697539806366, "global_step": 239319, "epoch": 2688, "lr": 3.429685298288249e-06}
{"train_loss": 0.04391767518343718, "global_step": 239320, "epoch": 2688, "lr": 3.429474288192908e-06, "val_loss": 9.429243087768555}
{"train_loss": 0.031195316463708878, "global_step": 239321, "epoch": 2689, "lr": 3.4292632843584027e-06}
{"train_loss": 0.05272012948989868, "global_step": 239322, "epoch": 2689, "lr": 3.429052286784773e-06}
{"train_loss": 0.03405597433447838, "global_step": 239323, "epoch": 2689, "lr": 3.4288412954720295e-06}
{"train_loss": 0.038460489362478256, "global_step": 239324, "epoch": 2689, "lr": 3.428630310420222e-06}
{"train_loss": 0.041219063103199005, "global_step": 239325, "epoch": 2689, "lr": 3.4284193316293565e-06}
{"train_loss": 0.040056996047496796, "global_step": 239326, "epoch": 2689, "lr": 3.4282083590994827e-06}
{"train_loss": 0.1270589679479599, "global_step": 239327, "epoch": 2689, "lr": 3.4279973928306063e-06}
{"train_loss": 0.062564916908741, "global_step": 239328, "epoch": 2689, "lr": 3.4277864328227828e-06}
{"train_loss": 0.04968235269188881, "global_step": 239329, "epoch": 2689, "lr": 3.4275754790760116e-06}
{"train_loss": 0.03417857736349106, "global_step": 239330, "epoch": 2689, "lr": 3.427364531590349e-06}
{"train_loss": 0.025587886571884155, "global_step": 239331, "epoch": 2689, "lr": 3.4271535903657945e-06}
{"train_loss": 0.045518141239881516, "global_step": 239332, "epoch": 2689, "lr": 3.426942655402404e-06}
{"train_loss": 0.034454215317964554, "global_step": 239333, "epoch": 2689, "lr": 3.426731726700183e-06}
{"train_loss": 0.043080590665340424, "global_step": 239334, "epoch": 2689, "lr": 3.4265208042591813e-06}
{"train_loss": 0.041306182742118835, "global_step": 239335, "epoch": 2689, "lr": 3.4263098880794098e-06}
{"train_loss": 0.07163672149181366, "global_step": 239336, "epoch": 2689, "lr": 3.4260989781608965e-06}
{"train_loss": 0.06283501535654068, "global_step": 239337, "epoch": 2689, "lr": 3.4258880745036914e-06}
{"train_loss": 0.023497845977544785, "global_step": 239338, "epoch": 2689, "lr": 3.4256771771077945e-06}
{"train_loss": 0.04892812669277191, "global_step": 239339, "epoch": 2689, "lr": 3.4254662859732612e-06}
{"train_loss": 0.04095030203461647, "global_step": 239340, "epoch": 2689, "lr": 3.4252554011000972e-06}
{"train_loss": 0.04964069277048111, "global_step": 239341, "epoch": 2689, "lr": 3.425044522488341e-06}
{"train_loss": 0.02398136630654335, "global_step": 239342, "epoch": 2689, "lr": 3.4248336501380207e-06}
{"train_loss": 0.03681689500808716, "global_step": 239343, "epoch": 2689, "lr": 3.424622784049164e-06}
{"train_loss": 0.022797321900725365, "global_step": 239344, "epoch": 2689, "lr": 3.4244119242218043e-06}
{"train_loss": 0.021631484851241112, "global_step": 239345, "epoch": 2689, "lr": 3.424201070655958e-06}
{"train_loss": 0.022442759945988655, "global_step": 239346, "epoch": 2689, "lr": 3.423990223351664e-06}
{"train_loss": 0.06767524778842926, "global_step": 239347, "epoch": 2689, "lr": 3.4237793823089393e-06}
{"train_loss": 0.058739036321640015, "global_step": 239348, "epoch": 2689, "lr": 3.4235685475278333e-06}
{"train_loss": 0.0491032712161541, "global_step": 239349, "epoch": 2689, "lr": 3.423357719008352e-06}
{"train_loss": 0.0559234619140625, "global_step": 239350, "epoch": 2689, "lr": 3.423146896750534e-06}
{"train_loss": 0.038905609399080276, "global_step": 239351, "epoch": 2689, "lr": 3.4229360807544077e-06}
{"train_loss": 0.04025711864233017, "global_step": 239352, "epoch": 2689, "lr": 3.4227252710199944e-06}
{"train_loss": 0.044403236359357834, "global_step": 239353, "epoch": 2689, "lr": 3.422514467547333e-06}
{"train_loss": 0.04451252520084381, "global_step": 239354, "epoch": 2689, "lr": 3.422303670336441e-06}
{"train_loss": 0.03080911934375763, "global_step": 239355, "epoch": 2689, "lr": 3.422092879387356e-06}
{"train_loss": 0.04979145899415016, "global_step": 239356, "epoch": 2689, "lr": 3.4218820947001073e-06}
{"train_loss": 0.025538429617881775, "global_step": 239357, "epoch": 2689, "lr": 3.4216713162747105e-06}
{"train_loss": 0.0391731932759285, "global_step": 239358, "epoch": 2689, "lr": 3.42146054411121e-06}
{"train_loss": 0.03154541552066803, "global_step": 239359, "epoch": 2689, "lr": 3.4212497782096177e-06}
{"train_loss": 0.06786540895700455, "global_step": 239360, "epoch": 2689, "lr": 3.4210390185699827e-06}
{"train_loss": 0.016788162291049957, "global_step": 239361, "epoch": 2689, "lr": 3.4208282651923106e-06}
{"train_loss": 0.025487860664725304, "global_step": 239362, "epoch": 2689, "lr": 3.4206175180766465e-06}
{"train_loss": 0.0884217694401741, "global_step": 239363, "epoch": 2689, "lr": 3.4204067772230062e-06}
{"train_loss": 0.030228108167648315, "global_step": 239364, "epoch": 2689, "lr": 3.420196042631435e-06}
{"train_loss": 0.05984149128198624, "global_step": 239365, "epoch": 2689, "lr": 3.4199853143019433e-06}
{"train_loss": 0.035507697612047195, "global_step": 239366, "epoch": 2689, "lr": 3.41977459223457e-06}
{"train_loss": 0.03416241332888603, "global_step": 239367, "epoch": 2689, "lr": 3.4195638764293327e-06}
{"train_loss": 0.04975546523928642, "global_step": 239368, "epoch": 2689, "lr": 3.4193531668862798e-06}
{"train_loss": 0.056070126593112946, "global_step": 239369, "epoch": 2689, "lr": 3.419142463605418e-06}
{"train_loss": 0.09358301013708115, "global_step": 239370, "epoch": 2689, "lr": 3.4189317665867916e-06}
{"train_loss": 0.051455941051244736, "global_step": 239371, "epoch": 2689, "lr": 3.418721075830411e-06}
{"train_loss": 0.04582497477531433, "global_step": 239372, "epoch": 2689, "lr": 3.4185103913363158e-06}
{"train_loss": 0.0702681839466095, "global_step": 239373, "epoch": 2689, "lr": 3.41829971310455e-06}
{"train_loss": 0.03452679514884949, "global_step": 239374, "epoch": 2689, "lr": 3.4180890411351084e-06}
{"train_loss": 0.06134425848722458, "global_step": 239375, "epoch": 2689, "lr": 3.417878375428052e-06}
{"train_loss": 0.01759222336113453, "global_step": 239376, "epoch": 2689, "lr": 3.4176677159833805e-06}
{"train_loss": 0.08045164495706558, "global_step": 239377, "epoch": 2689, "lr": 3.4174570628011494e-06}
{"train_loss": 0.029809201136231422, "global_step": 239378, "epoch": 2689, "lr": 3.4172464158813645e-06}
{"train_loss": 0.04323643073439598, "global_step": 239379, "epoch": 2689, "lr": 3.41703577522407e-06}
{"train_loss": 0.015700960531830788, "global_step": 239380, "epoch": 2689, "lr": 3.416825140829277e-06}
{"train_loss": 0.03788981959223747, "global_step": 239381, "epoch": 2689, "lr": 3.4166145126970308e-06}
{"train_loss": 0.052815526723861694, "global_step": 239382, "epoch": 2689, "lr": 3.4164038908273576e-06}
{"train_loss": 0.02115563489496708, "global_step": 239383, "epoch": 2689, "lr": 3.4161932752202695e-06}
{"train_loss": 0.03565880283713341, "global_step": 239384, "epoch": 2689, "lr": 3.4159826658758165e-06}
{"train_loss": 0.10344354808330536, "global_step": 239385, "epoch": 2689, "lr": 3.415772062794009e-06}
{"train_loss": 0.03750542551279068, "global_step": 239386, "epoch": 2689, "lr": 3.4155614659748924e-06}
{"train_loss": 0.06444744765758514, "global_step": 239387, "epoch": 2689, "lr": 3.4153508754184827e-06}
{"train_loss": 0.03367840126156807, "global_step": 239388, "epoch": 2689, "lr": 3.4151402911248076e-06}
{"train_loss": 0.03789101913571358, "global_step": 239389, "epoch": 2689, "lr": 3.414929713093906e-06}
{"train_loss": 0.06324749439954758, "global_step": 239390, "epoch": 2689, "lr": 3.4147191413257896e-06}
{"train_loss": 0.06770174950361252, "global_step": 239391, "epoch": 2689, "lr": 3.4145085758204966e-06}
{"train_loss": 0.06260643154382706, "global_step": 239392, "epoch": 2689, "lr": 3.4142980165780603e-06}
{"train_loss": 0.023416792973876, "global_step": 239393, "epoch": 2689, "lr": 3.414087463598498e-06}
{"train_loss": 0.08225473761558533, "global_step": 239394, "epoch": 2689, "lr": 3.413876916881853e-06}
{"train_loss": 0.0622776634991169, "global_step": 239395, "epoch": 2689, "lr": 3.4136663764281375e-06}
{"train_loss": 0.06472743302583694, "global_step": 239396, "epoch": 2689, "lr": 3.4134558422373896e-06}
{"train_loss": 0.058407336473464966, "global_step": 239397, "epoch": 2689, "lr": 3.413245314309632e-06}
{"train_loss": 0.029329033568501472, "global_step": 239398, "epoch": 2689, "lr": 3.4130347926449035e-06}
{"train_loss": 0.07160326093435287, "global_step": 239399, "epoch": 2689, "lr": 3.412824277243215e-06}
{"train_loss": 0.07911412417888641, "global_step": 239400, "epoch": 2689, "lr": 3.4126137681046112e-06}
{"train_loss": 0.08714017271995544, "global_step": 239401, "epoch": 2689, "lr": 3.4124032652291085e-06}
{"train_loss": 0.021799901500344276, "global_step": 239402, "epoch": 2689, "lr": 3.4121927686167456e-06}
{"train_loss": 0.07189393043518066, "global_step": 239403, "epoch": 2689, "lr": 3.411982278267539e-06}
{"train_loss": 0.050629157572984695, "global_step": 239404, "epoch": 2689, "lr": 3.4117717941815285e-06}
{"train_loss": 0.01763194426894188, "global_step": 239405, "epoch": 2689, "lr": 3.4115613163587358e-06}
{"train_loss": 0.037296220660209656, "global_step": 239406, "epoch": 2689, "lr": 3.411350844799194e-06}
{"train_loss": 0.020103678107261658, "global_step": 239407, "epoch": 2689, "lr": 3.4111403795029253e-06}
{"train_loss": 0.018689710646867752, "global_step": 239408, "epoch": 2689, "lr": 3.410929920469963e-06}
{"train_loss": 0.04646588379538126, "global_step": 239409, "epoch": 2689, "lr": 3.4107194677003296e-06, "val_loss": 9.42625904083252}
{"train_loss": 0.060912903398275375, "global_step": 239410, "epoch": 2690, "lr": 3.4105090211940525e-06}
{"train_loss": 0.026730388402938843, "global_step": 239411, "epoch": 2690, "lr": 3.4102985809511766e-06}
{"train_loss": 0.01941671222448349, "global_step": 239412, "epoch": 2690, "lr": 3.410088146971707e-06}
{"train_loss": 0.017940128222107887, "global_step": 239413, "epoch": 2690, "lr": 3.409877719255694e-06}
{"train_loss": 0.04874772951006889, "global_step": 239414, "epoch": 2690, "lr": 3.4096672978031486e-06}
{"train_loss": 0.035186778753995895, "global_step": 239415, "epoch": 2690, "lr": 3.4094568826141093e-06}
{"train_loss": 0.0741705521941185, "global_step": 239416, "epoch": 2690, "lr": 3.409246473688593e-06}
{"train_loss": 0.03378770500421524, "global_step": 239417, "epoch": 2690, "lr": 3.4090360710266447e-06}
{"train_loss": 0.024416066706180573, "global_step": 239418, "epoch": 2690, "lr": 3.40882567462828e-06}
{"train_loss": 0.0287656057626009, "global_step": 239419, "epoch": 2690, "lr": 3.4086152844935325e-06}
{"train_loss": 0.045895930379629135, "global_step": 239420, "epoch": 2690, "lr": 3.4084049006224306e-06}
{"train_loss": 0.042623091489076614, "global_step": 239421, "epoch": 2690, "lr": 3.408194523014996e-06}
{"train_loss": 0.06721008569002151, "global_step": 239422, "epoch": 2690, "lr": 3.4079841516712673e-06}
{"train_loss": 0.05595989525318146, "global_step": 239423, "epoch": 2690, "lr": 3.4077737865912563e-06}
{"train_loss": 0.029973404482007027, "global_step": 239424, "epoch": 2690, "lr": 3.4075634277750125e-06}
{"train_loss": 0.057660460472106934, "global_step": 239425, "epoch": 2690, "lr": 3.407353075222558e-06}
{"train_loss": 0.06135384365916252, "global_step": 239426, "epoch": 2690, "lr": 3.407142728933904e-06}
{"train_loss": 0.03207048773765564, "global_step": 239427, "epoch": 2690, "lr": 3.4069323889090897e-06}
{"train_loss": 0.043011054396629333, "global_step": 239428, "epoch": 2690, "lr": 3.406722055148159e-06}
{"train_loss": 0.02366703376173973, "global_step": 239429, "epoch": 2690, "lr": 3.4065117276511182e-06}
{"train_loss": 0.029413655400276184, "global_step": 239430, "epoch": 2690, "lr": 3.406301406418011e-06}
{"train_loss": 0.05422724783420563, "global_step": 239431, "epoch": 2690, "lr": 3.4060910914488485e-06}
{"train_loss": 0.016924535855650902, "global_step": 239432, "epoch": 2690, "lr": 3.405880782743681e-06}
{"train_loss": 0.06331140547990799, "global_step": 239433, "epoch": 2690, "lr": 3.405670480302514e-06}
{"train_loss": 0.05507804453372955, "global_step": 239434, "epoch": 2690, "lr": 3.405460184125392e-06}
{"train_loss": 0.07697110623121262, "global_step": 239435, "epoch": 2690, "lr": 3.4052498942123367e-06}
{"train_loss": 0.03631196916103363, "global_step": 239436, "epoch": 2690, "lr": 3.405039610563382e-06}
{"train_loss": 0.061819836497306824, "global_step": 239437, "epoch": 2690, "lr": 3.404829333178544e-06}
{"train_loss": 0.035538289695978165, "global_step": 239438, "epoch": 2690, "lr": 3.404619062057868e-06}
{"train_loss": 0.03774874657392502, "global_step": 239439, "epoch": 2690, "lr": 3.404408797201364e-06}
{"train_loss": 0.04755678027868271, "global_step": 239440, "epoch": 2690, "lr": 3.4041985386090768e-06}
{"train_loss": 0.03877086564898491, "global_step": 239441, "epoch": 2690, "lr": 3.403988286281018e-06}
{"train_loss": 0.04938772693276405, "global_step": 239442, "epoch": 2690, "lr": 3.4037780402172316e-06}
{"train_loss": 0.1128891259431839, "global_step": 239443, "epoch": 2690, "lr": 3.4035678004177397e-06}
{"train_loss": 0.05096505582332611, "global_step": 239444, "epoch": 2690, "lr": 3.40335756688257e-06}
{"train_loss": 0.020390795543789864, "global_step": 239445, "epoch": 2690, "lr": 3.4031473396117454e-06}
{"train_loss": 0.014541471377015114, "global_step": 239446, "epoch": 2690, "lr": 3.4029371186052985e-06}
{"train_loss": 0.04757751524448395, "global_step": 239447, "epoch": 2690, "lr": 3.402726903863268e-06}
{"train_loss": 0.02920306846499443, "global_step": 239448, "epoch": 2690, "lr": 3.4025166953856657e-06}
{"train_loss": 0.038585882633924484, "global_step": 239449, "epoch": 2690, "lr": 3.4023064931725357e-06}
{"train_loss": 0.047066643834114075, "global_step": 239450, "epoch": 2690, "lr": 3.402096297223889e-06}
{"train_loss": 0.03367972746491432, "global_step": 239451, "epoch": 2690, "lr": 3.40188610753977e-06}
{"train_loss": 0.06958414614200592, "global_step": 239452, "epoch": 2690, "lr": 3.40167592412019e-06}
{"train_loss": 0.01888711005449295, "global_step": 239453, "epoch": 2690, "lr": 3.401465746965199e-06}
{"train_loss": 0.06245484575629234, "global_step": 239454, "epoch": 2690, "lr": 3.401255576074802e-06}
{"train_loss": 0.03775506839156151, "global_step": 239455, "epoch": 2690, "lr": 3.401045411449044e-06}
{"train_loss": 0.05306987836956978, "global_step": 239456, "epoch": 2690, "lr": 3.4008352530879474e-06}
{"train_loss": 0.03056860901415348, "global_step": 239457, "epoch": 2690, "lr": 3.400625100991539e-06}
{"train_loss": 0.09986411035060883, "global_step": 239458, "epoch": 2690, "lr": 3.4004149551598476e-06}
{"train_loss": 0.027627304196357727, "global_step": 239459, "epoch": 2690, "lr": 3.4002048155929056e-06}
{"train_loss": 0.033281583338975906, "global_step": 239460, "epoch": 2690, "lr": 3.3999946822907416e-06}
{"train_loss": 0.030082859098911285, "global_step": 239461, "epoch": 2690, "lr": 3.3997845552533713e-06}
{"train_loss": 0.06285887956619263, "global_step": 239462, "epoch": 2690, "lr": 3.39957443448084e-06}
{"train_loss": 0.04179776832461357, "global_step": 239463, "epoch": 2690, "lr": 3.3993643199731586e-06}
{"train_loss": 0.03908905014395714, "global_step": 239464, "epoch": 2690, "lr": 3.399154211730371e-06}
{"train_loss": 0.07110637426376343, "global_step": 239465, "epoch": 2690, "lr": 3.3989441097524943e-06}
{"train_loss": 0.021899905055761337, "global_step": 239466, "epoch": 2690, "lr": 3.3987340140395676e-06}
{"train_loss": 0.0539138987660408, "global_step": 239467, "epoch": 2690, "lr": 3.398523924591607e-06}
{"train_loss": 0.03802500665187836, "global_step": 239468, "epoch": 2690, "lr": 3.398313841408657e-06}
{"train_loss": 0.0484854057431221, "global_step": 239469, "epoch": 2690, "lr": 3.3981037644907233e-06}
{"train_loss": 0.05153626203536987, "global_step": 239470, "epoch": 2690, "lr": 3.3978936938378558e-06}
{"train_loss": 0.04666756093502045, "global_step": 239471, "epoch": 2690, "lr": 3.3976836294500658e-06}
{"train_loss": 0.03834684565663338, "global_step": 239472, "epoch": 2690, "lr": 3.3974735713273973e-06}
{"train_loss": 0.049527499824762344, "global_step": 239473, "epoch": 2690, "lr": 3.3972635194698623e-06}
{"train_loss": 0.038719795644283295, "global_step": 239474, "epoch": 2690, "lr": 3.3970534738775093e-06}
{"train_loss": 0.06076303869485855, "global_step": 239475, "epoch": 2690, "lr": 3.39684343455034e-06}
{"train_loss": 0.04382413253188133, "global_step": 239476, "epoch": 2690, "lr": 3.3966334014884082e-06}
{"train_loss": 0.03217707574367523, "global_step": 239477, "epoch": 2690, "lr": 3.3964233746917206e-06}
{"train_loss": 0.03832131624221802, "global_step": 239478, "epoch": 2690, "lr": 3.396213354160327e-06}
{"train_loss": 0.02936771884560585, "global_step": 239479, "epoch": 2690, "lr": 3.396003339894238e-06}
{"train_loss": 0.04009135067462921, "global_step": 239480, "epoch": 2690, "lr": 3.395793331893493e-06}
{"train_loss": 0.04233634099364281, "global_step": 239481, "epoch": 2690, "lr": 3.3955833301581085e-06}
{"train_loss": 0.05639100819826126, "global_step": 239482, "epoch": 2690, "lr": 3.395373334688118e-06}
{"train_loss": 0.03184913843870163, "global_step": 239483, "epoch": 2690, "lr": 3.3951633454835597e-06}
{"train_loss": 0.031650058925151825, "global_step": 239484, "epoch": 2690, "lr": 3.394953362544451e-06}
{"train_loss": 0.035574764013290405, "global_step": 239485, "epoch": 2690, "lr": 3.3947433858708244e-06}
{"train_loss": 0.03713289275765419, "global_step": 239486, "epoch": 2690, "lr": 3.394533415462703e-06}
{"train_loss": 0.05036606267094612, "global_step": 239487, "epoch": 2690, "lr": 3.3943234513201195e-06}
{"train_loss": 0.03212450072169304, "global_step": 239488, "epoch": 2690, "lr": 3.394113493443102e-06}
{"train_loss": 0.036628831177949905, "global_step": 239489, "epoch": 2690, "lr": 3.3939035418316778e-06}
{"train_loss": 0.018975036218762398, "global_step": 239490, "epoch": 2690, "lr": 3.3936935964858753e-06}
{"train_loss": 0.022056084126234055, "global_step": 239491, "epoch": 2690, "lr": 3.393483657405727e-06}
{"train_loss": 0.04401588812470436, "global_step": 239492, "epoch": 2690, "lr": 3.393273724591245e-06}
{"train_loss": 0.027954982593655586, "global_step": 239493, "epoch": 2690, "lr": 3.393063798042484e-06}
{"train_loss": 0.03582439571619034, "global_step": 239494, "epoch": 2690, "lr": 3.3928538777594443e-06}
{"train_loss": 0.07978616654872894, "global_step": 239495, "epoch": 2690, "lr": 3.392643963742176e-06}
{"train_loss": 0.039515409618616104, "global_step": 239496, "epoch": 2690, "lr": 3.392434055990695e-06}
{"train_loss": 0.08209309726953506, "global_step": 239497, "epoch": 2690, "lr": 3.392224154505036e-06}
{"train_loss": 0.04345485065760237, "global_step": 239498, "epoch": 2690, "lr": 3.3920142592852257e-06, "val_loss": 9.381895065307617, "train_action_mse_error": 28.02939224243164}
{"train_loss": 0.05293811485171318, "global_step": 239499, "epoch": 2691, "lr": 3.3918043703312863e-06}
{"train_loss": 0.03393140435218811, "global_step": 239500, "epoch": 2691, "lr": 3.391594487643257e-06}
{"train_loss": 0.028477052226662636, "global_step": 239501, "epoch": 2691, "lr": 3.391384611221149e-06}
{"train_loss": 0.029450660571455956, "global_step": 239502, "epoch": 2691, "lr": 3.391174741065012e-06}
{"train_loss": 0.04011565074324608, "global_step": 239503, "epoch": 2691, "lr": 3.390964877174857e-06}
{"train_loss": 0.023071087896823883, "global_step": 239504, "epoch": 2691, "lr": 3.3907550195507232e-06}
{"train_loss": 0.07196281850337982, "global_step": 239505, "epoch": 2691, "lr": 3.390545168192627e-06}
{"train_loss": 0.029204804450273514, "global_step": 239506, "epoch": 2691, "lr": 3.390335323100613e-06}
{"train_loss": 0.027284685522317886, "global_step": 239507, "epoch": 2691, "lr": 3.390125484274692e-06}
{"train_loss": 0.05276038497686386, "global_step": 239508, "epoch": 2691, "lr": 3.389915651714909e-06}
{"train_loss": 0.04068312048912048, "global_step": 239509, "epoch": 2691, "lr": 3.389705825421274e-06}
{"train_loss": 0.05756683647632599, "global_step": 239510, "epoch": 2691, "lr": 3.3894960053938328e-06}
{"train_loss": 0.0647985115647316, "global_step": 239511, "epoch": 2691, "lr": 3.3892861916326014e-06}
{"train_loss": 0.04374599829316139, "global_step": 239512, "epoch": 2691, "lr": 3.3890763841376185e-06}
{"train_loss": 0.035878803580999374, "global_step": 239513, "epoch": 2691, "lr": 3.388866582908895e-06}
{"train_loss": 0.01957893744111061, "global_step": 239514, "epoch": 2691, "lr": 3.3886567879464813e-06}
{"train_loss": 0.031459975987672806, "global_step": 239515, "epoch": 2691, "lr": 3.388446999250389e-06}
{"train_loss": 0.01073289755731821, "global_step": 239516, "epoch": 2691, "lr": 3.3882372168206554e-06}
{"train_loss": 0.027914859354496002, "global_step": 239517, "epoch": 2691, "lr": 3.388027440657299e-06}
{"train_loss": 0.03523334115743637, "global_step": 239518, "epoch": 2691, "lr": 3.3878176707603513e-06}
{"train_loss": 0.041735753417015076, "global_step": 239519, "epoch": 2691, "lr": 3.3876079071298527e-06}
{"train_loss": 0.03241672366857529, "global_step": 239520, "epoch": 2691, "lr": 3.3873981497658193e-06}
{"train_loss": 0.045001689344644547, "global_step": 239521, "epoch": 2691, "lr": 3.387188398668284e-06}
{"train_loss": 0.0487435944378376, "global_step": 239522, "epoch": 2691, "lr": 3.3869786538372695e-06}
{"train_loss": 0.02454128861427307, "global_step": 239523, "epoch": 2691, "lr": 3.3867689152728144e-06}
{"train_loss": 0.041527748107910156, "global_step": 239524, "epoch": 2691, "lr": 3.38655918297493e-06}
{"train_loss": 0.08104682713747025, "global_step": 239525, "epoch": 2691, "lr": 3.3863494569436603e-06}
{"train_loss": 0.030357392504811287, "global_step": 239526, "epoch": 2691, "lr": 3.386139737179028e-06}
{"train_loss": 0.015864357352256775, "global_step": 239527, "epoch": 2691, "lr": 3.385930023681061e-06}
{"train_loss": 0.03404656797647476, "global_step": 239528, "epoch": 2691, "lr": 3.3857203164497864e-06}
{"train_loss": 0.018293257802724838, "global_step": 239529, "epoch": 2691, "lr": 3.3855106154852377e-06}
{"train_loss": 0.03309042006731033, "global_step": 239530, "epoch": 2691, "lr": 3.385300920787432e-06}
{"train_loss": 0.030093153938651085, "global_step": 239531, "epoch": 2691, "lr": 3.3850912323564077e-06}
{"train_loss": 0.03585077077150345, "global_step": 239532, "epoch": 2691, "lr": 3.3848815501921873e-06}
{"train_loss": 0.05256137251853943, "global_step": 239533, "epoch": 2691, "lr": 3.384671874294809e-06}
{"train_loss": 0.05604224652051926, "global_step": 239534, "epoch": 2691, "lr": 3.384462204664285e-06}
{"train_loss": 0.042306430637836456, "global_step": 239535, "epoch": 2691, "lr": 3.3842525413006586e-06}
{"train_loss": 0.06962539255619049, "global_step": 239536, "epoch": 2691, "lr": 3.384042884203953e-06}
{"train_loss": 0.07128328084945679, "global_step": 239537, "epoch": 2691, "lr": 3.383833233374184e-06}
{"train_loss": 0.052657533437013626, "global_step": 239538, "epoch": 2691, "lr": 3.3836235888113967e-06}
{"train_loss": 0.025665737688541412, "global_step": 239539, "epoch": 2691, "lr": 3.3834139505156127e-06}
{"train_loss": 0.05948653817176819, "global_step": 239540, "epoch": 2691, "lr": 3.38320431848686e-06}
{"train_loss": 0.04307951778173447, "global_step": 239541, "epoch": 2691, "lr": 3.3829946927251664e-06}
{"train_loss": 0.04179186746478081, "global_step": 239542, "epoch": 2691, "lr": 3.3827850732305653e-06}
{"train_loss": 0.02704397775232792, "global_step": 239543, "epoch": 2691, "lr": 3.3825754600030734e-06}
{"train_loss": 0.07602615654468536, "global_step": 239544, "epoch": 2691, "lr": 3.382365853042735e-06}
{"train_loss": 0.037248145788908005, "global_step": 239545, "epoch": 2691, "lr": 3.382156252349561e-06}
{"train_loss": 0.025254612788558006, "global_step": 239546, "epoch": 2691, "lr": 3.3819466579235902e-06}
{"train_loss": 0.04356933757662773, "global_step": 239547, "epoch": 2691, "lr": 3.381737069764845e-06}
{"train_loss": 0.03275623917579651, "global_step": 239548, "epoch": 2691, "lr": 3.3815274878733706e-06}
{"train_loss": 0.057812001556158066, "global_step": 239549, "epoch": 2691, "lr": 3.3813179122491656e-06}
{"train_loss": 0.01869906298816204, "global_step": 239550, "epoch": 2691, "lr": 3.381108342892286e-06}
{"train_loss": 0.08307491987943649, "global_step": 239551, "epoch": 2691, "lr": 3.3808987798027382e-06}
{"train_loss": 0.06686804443597794, "global_step": 239552, "epoch": 2691, "lr": 3.380689222980571e-06}
{"train_loss": 0.04618774354457855, "global_step": 239553, "epoch": 2691, "lr": 3.3804796724257904e-06}
{"train_loss": 0.055247291922569275, "global_step": 239554, "epoch": 2691, "lr": 3.380270128138441e-06}
{"train_loss": 0.028508076444268227, "global_step": 239555, "epoch": 2691, "lr": 3.38006059011855e-06}
{"train_loss": 0.051663778722286224, "global_step": 239556, "epoch": 2691, "lr": 3.379851058366135e-06}
{"train_loss": 0.03436646983027458, "global_step": 239557, "epoch": 2691, "lr": 3.379641532881239e-06}
{"train_loss": 0.03710436820983887, "global_step": 239558, "epoch": 2691, "lr": 3.379432013663869e-06}
{"train_loss": 0.023190505802631378, "global_step": 239559, "epoch": 2691, "lr": 3.3792225007140798e-06}
{"train_loss": 0.043592873960733414, "global_step": 239560, "epoch": 2691, "lr": 3.379012994031877e-06}
{"train_loss": 0.03203437104821205, "global_step": 239561, "epoch": 2691, "lr": 3.3788034936173053e-06}
{"train_loss": 0.0807412713766098, "global_step": 239562, "epoch": 2691, "lr": 3.3785939994703808e-06}
{"train_loss": 0.04870323836803436, "global_step": 239563, "epoch": 2691, "lr": 3.378384511591137e-06}
{"train_loss": 0.052843473851680756, "global_step": 239564, "epoch": 2691, "lr": 3.3781750299795966e-06}
{"train_loss": 0.03447364270687103, "global_step": 239565, "epoch": 2691, "lr": 3.3779655546358036e-06}
{"train_loss": 0.032808464020490646, "global_step": 239566, "epoch": 2691, "lr": 3.3777560855597633e-06}
{"train_loss": 0.048336781561374664, "global_step": 239567, "epoch": 2691, "lr": 3.377546622751521e-06}
{"train_loss": 0.05354728177189827, "global_step": 239568, "epoch": 2691, "lr": 3.3773371662110973e-06}
{"train_loss": 0.026572156697511673, "global_step": 239569, "epoch": 2691, "lr": 3.3771277159385274e-06}
{"train_loss": 0.08077067881822586, "global_step": 239570, "epoch": 2691, "lr": 3.3769182719338264e-06}
{"train_loss": 0.02969346009194851, "global_step": 239571, "epoch": 2691, "lr": 3.37670883419704e-06}
{"train_loss": 0.039474084973335266, "global_step": 239572, "epoch": 2691, "lr": 3.3764994027281726e-06}
{"train_loss": 0.06567290425300598, "global_step": 239573, "epoch": 2691, "lr": 3.3762899775272807e-06}
{"train_loss": 0.07434727251529694, "global_step": 239574, "epoch": 2691, "lr": 3.3760805585943687e-06}
{"train_loss": 0.023642338812351227, "global_step": 239575, "epoch": 2691, "lr": 3.3758711459294823e-06}
{"train_loss": 0.02991577424108982, "global_step": 239576, "epoch": 2691, "lr": 3.375661739532643e-06}
{"train_loss": 0.04562925547361374, "global_step": 239577, "epoch": 2691, "lr": 3.375452339403867e-06}
{"train_loss": 0.06945019960403442, "global_step": 239578, "epoch": 2691, "lr": 3.3752429455432053e-06}
{"train_loss": 0.0723172128200531, "global_step": 239579, "epoch": 2691, "lr": 3.375033557950663e-06}
{"train_loss": 0.03687959536910057, "global_step": 239580, "epoch": 2691, "lr": 3.37482417662629e-06}
{"train_loss": 0.020143035799264908, "global_step": 239581, "epoch": 2691, "lr": 3.3746148015700918e-06}
{"train_loss": 0.024486899375915527, "global_step": 239582, "epoch": 2691, "lr": 3.3744054327821185e-06}
{"train_loss": 0.05314258486032486, "global_step": 239583, "epoch": 2691, "lr": 3.374196070262381e-06}
{"train_loss": 0.07177800685167313, "global_step": 239584, "epoch": 2691, "lr": 3.3739867140109184e-06}
{"train_loss": 0.03991919383406639, "global_step": 239585, "epoch": 2691, "lr": 3.3737773640277524e-06}
{"train_loss": 0.05812413990497589, "global_step": 239586, "epoch": 2691, "lr": 3.373568020312917e-06}
{"train_loss": 0.04386631454853864, "global_step": 239587, "epoch": 2691, "lr": 3.3733586828664287e-06, "val_loss": 9.394392013549805}
{"train_loss": 0.05180307477712631, "global_step": 239588, "epoch": 2692, "lr": 3.3731493516883374e-06}
{"train_loss": 0.051368363201618195, "global_step": 239589, "epoch": 2692, "lr": 3.3729400267786482e-06}
{"train_loss": 0.08597426116466522, "global_step": 239590, "epoch": 2692, "lr": 3.372730708137395e-06}
{"train_loss": 0.03610723093152046, "global_step": 239591, "epoch": 2692, "lr": 3.372521395764622e-06}
{"train_loss": 0.037771351635456085, "global_step": 239592, "epoch": 2692, "lr": 3.3723120896603346e-06}
{"train_loss": 0.04837947338819504, "global_step": 239593, "epoch": 2692, "lr": 3.372102789824583e-06}
{"train_loss": 0.06871363520622253, "global_step": 239594, "epoch": 2692, "lr": 3.371893496257372e-06}
{"train_loss": 0.04424767196178436, "global_step": 239595, "epoch": 2692, "lr": 3.371684208958753e-06}
{"train_loss": 0.030493032187223434, "global_step": 239596, "epoch": 2692, "lr": 3.3714749279287306e-06}
{"train_loss": 0.04696046933531761, "global_step": 239597, "epoch": 2692, "lr": 3.3712656531673547e-06}
{"train_loss": 0.05812449753284454, "global_step": 239598, "epoch": 2692, "lr": 3.3710563846746367e-06}
{"train_loss": 0.032386891543865204, "global_step": 239599, "epoch": 2692, "lr": 3.370847122450621e-06}
{"train_loss": 0.06773246079683304, "global_step": 239600, "epoch": 2692, "lr": 3.370637866495313e-06}
{"train_loss": 0.033827099949121475, "global_step": 239601, "epoch": 2692, "lr": 3.3704286168087682e-06}
{"train_loss": 0.0767662525177002, "global_step": 239602, "epoch": 2692, "lr": 3.3702193733909928e-06}
{"train_loss": 0.046719226986169815, "global_step": 239603, "epoch": 2692, "lr": 3.37001013624203e-06}
{"train_loss": 0.04235129803419113, "global_step": 239604, "epoch": 2692, "lr": 3.3698009053618916e-06}
{"train_loss": 0.05133819952607155, "global_step": 239605, "epoch": 2692, "lr": 3.369591680750622e-06}
{"train_loss": 0.043367933481931686, "global_step": 239606, "epoch": 2692, "lr": 3.369382462408238e-06}
{"train_loss": 0.06473800539970398, "global_step": 239607, "epoch": 2692, "lr": 3.3691732503347784e-06}
{"train_loss": 0.03137368708848953, "global_step": 239608, "epoch": 2692, "lr": 3.3689640445302594e-06}
{"train_loss": 0.09409817308187485, "global_step": 239609, "epoch": 2692, "lr": 3.3687548449947203e-06}
{"train_loss": 0.07550157606601715, "global_step": 239610, "epoch": 2692, "lr": 3.3685456517281723e-06}
{"train_loss": 0.05574231967329979, "global_step": 239611, "epoch": 2692, "lr": 3.3683364647306702e-06}
{"train_loss": 0.020757261663675308, "global_step": 239612, "epoch": 2692, "lr": 3.368127284002215e-06}
{"train_loss": 0.013492744415998459, "global_step": 239613, "epoch": 2692, "lr": 3.3679181095428557e-06}
{"train_loss": 0.04125598445534706, "global_step": 239614, "epoch": 2692, "lr": 3.3677089413526098e-06}
{"train_loss": 0.07042261213064194, "global_step": 239615, "epoch": 2692, "lr": 3.367499779431499e-06}
{"train_loss": 0.01809036359190941, "global_step": 239616, "epoch": 2692, "lr": 3.367290623779573e-06}
{"train_loss": 0.03286264091730118, "global_step": 239617, "epoch": 2692, "lr": 3.3670814743968327e-06}
{"train_loss": 0.05601834878325462, "global_step": 239618, "epoch": 2692, "lr": 3.3668723312833274e-06}
{"train_loss": 0.04721015691757202, "global_step": 239619, "epoch": 2692, "lr": 3.3666631944390735e-06}
{"train_loss": 0.05848158150911331, "global_step": 239620, "epoch": 2692, "lr": 3.3664540638641106e-06}
{"train_loss": 0.03553015738725662, "global_step": 239621, "epoch": 2692, "lr": 3.3662449395584494e-06}
{"train_loss": 0.035303011536598206, "global_step": 239622, "epoch": 2692, "lr": 3.3660358215221343e-06}
{"train_loss": 0.027649274095892906, "global_step": 239623, "epoch": 2692, "lr": 3.3658267097551877e-06}
{"train_loss": 0.03647056594491005, "global_step": 239624, "epoch": 2692, "lr": 3.365617604257637e-06}
{"train_loss": 0.06441958248615265, "global_step": 239625, "epoch": 2692, "lr": 3.365408505029505e-06}
{"train_loss": 0.04239387437701225, "global_step": 239626, "epoch": 2692, "lr": 3.3651994120708297e-06}
{"train_loss": 0.030202969908714294, "global_step": 239627, "epoch": 2692, "lr": 3.364990325381634e-06}
{"train_loss": 0.02443801797926426, "global_step": 239628, "epoch": 2692, "lr": 3.3647812449619454e-06}
{"train_loss": 0.005941880866885185, "global_step": 239629, "epoch": 2692, "lr": 3.3645721708118027e-06}
{"train_loss": 0.047664690762758255, "global_step": 239630, "epoch": 2692, "lr": 3.3643631029312116e-06}
{"train_loss": 0.048092909157276154, "global_step": 239631, "epoch": 2692, "lr": 3.3641540413202275e-06}
{"train_loss": 0.052258823066949844, "global_step": 239632, "epoch": 2692, "lr": 3.363944985978851e-06}
{"train_loss": 0.02645723521709442, "global_step": 239633, "epoch": 2692, "lr": 3.3637359369071363e-06}
{"train_loss": 0.02770131081342697, "global_step": 239634, "epoch": 2692, "lr": 3.36352689410509e-06}
{"train_loss": 0.04444068670272827, "global_step": 239635, "epoch": 2692, "lr": 3.3633178575727564e-06}
{"train_loss": 0.03619186952710152, "global_step": 239636, "epoch": 2692, "lr": 3.363108827310146e-06}
{"train_loss": 0.054673075675964355, "global_step": 239637, "epoch": 2692, "lr": 3.3628998033173043e-06}
{"train_loss": 0.060929205268621445, "global_step": 239638, "epoch": 2692, "lr": 3.362690785594247e-06}
{"train_loss": 0.05937587097287178, "global_step": 239639, "epoch": 2692, "lr": 3.3624817741410184e-06}
{"train_loss": 0.026437630876898766, "global_step": 239640, "epoch": 2692, "lr": 3.362272768957625e-06}
{"train_loss": 0.03344327211380005, "global_step": 239641, "epoch": 2692, "lr": 3.3620637700441106e-06}
{"train_loss": 0.06364195793867111, "global_step": 239642, "epoch": 2692, "lr": 3.3618547774004917e-06}
{"train_loss": 0.05827150493860245, "global_step": 239643, "epoch": 2692, "lr": 3.3616457910268127e-06}
{"train_loss": 0.03330758213996887, "global_step": 239644, "epoch": 2692, "lr": 3.3614368109230855e-06}
{"train_loss": 0.03313993662595749, "global_step": 239645, "epoch": 2692, "lr": 3.3612278370893423e-06}
{"train_loss": 0.04276241362094879, "global_step": 239646, "epoch": 2692, "lr": 3.3610188695256227e-06}
{"train_loss": 0.053753096610307693, "global_step": 239647, "epoch": 2692, "lr": 3.3608099082319435e-06}
{"train_loss": 0.01767120510339737, "global_step": 239648, "epoch": 2692, "lr": 3.360600953208326e-06}
{"train_loss": 0.04108346626162529, "global_step": 239649, "epoch": 2692, "lr": 3.360392004454821e-06}
{"train_loss": 0.024486104026436806, "global_step": 239650, "epoch": 2692, "lr": 3.360183061971428e-06}
{"train_loss": 0.024927042424678802, "global_step": 239651, "epoch": 2692, "lr": 3.359974125758203e-06}
{"train_loss": 0.03718003258109093, "global_step": 239652, "epoch": 2692, "lr": 3.359765195815162e-06}
{"train_loss": 0.07257790863513947, "global_step": 239653, "epoch": 2692, "lr": 3.3595562721423225e-06}
{"train_loss": 0.038406360894441605, "global_step": 239654, "epoch": 2692, "lr": 3.359347354739728e-06}
{"train_loss": 0.009962517768144608, "global_step": 239655, "epoch": 2692, "lr": 3.3591384436073903e-06}
{"train_loss": 0.07713472098112106, "global_step": 239656, "epoch": 2692, "lr": 3.358929538745359e-06}
{"train_loss": 0.03011256828904152, "global_step": 239657, "epoch": 2692, "lr": 3.3587206401536455e-06}
{"train_loss": 0.09797148406505585, "global_step": 239658, "epoch": 2692, "lr": 3.358511747832288e-06}
{"train_loss": 0.11104650795459747, "global_step": 239659, "epoch": 2692, "lr": 3.358302861781304e-06}
{"train_loss": 0.05697048082947731, "global_step": 239660, "epoch": 2692, "lr": 3.3580939820007374e-06}
{"train_loss": 0.06275443732738495, "global_step": 239661, "epoch": 2692, "lr": 3.357885108490599e-06}
{"train_loss": 0.032830942422151566, "global_step": 239662, "epoch": 2692, "lr": 3.357676241250929e-06}
{"train_loss": 0.08254925906658173, "global_step": 239663, "epoch": 2692, "lr": 3.3574673802817425e-06}
{"train_loss": 0.03804892301559448, "global_step": 239664, "epoch": 2692, "lr": 3.357258525583079e-06}
{"train_loss": 0.03695615381002426, "global_step": 239665, "epoch": 2692, "lr": 3.3570496771549718e-06}
{"train_loss": 0.05076214671134949, "global_step": 239666, "epoch": 2692, "lr": 3.3568408349974324e-06}
{"train_loss": 0.05545864626765251, "global_step": 239667, "epoch": 2692, "lr": 3.3566319991105042e-06}
{"train_loss": 0.09264540672302246, "global_step": 239668, "epoch": 2692, "lr": 3.3564231694941995e-06}
{"train_loss": 0.02844414673745632, "global_step": 239669, "epoch": 2692, "lr": 3.356214346148562e-06}
{"train_loss": 0.01756303757429123, "global_step": 239670, "epoch": 2692, "lr": 3.356005529073608e-06}
{"train_loss": 0.05150681361556053, "global_step": 239671, "epoch": 2692, "lr": 3.355796718269377e-06}
{"train_loss": 0.05968201532959938, "global_step": 239672, "epoch": 2692, "lr": 3.3555879137358804e-06}
{"train_loss": 0.02773340977728367, "global_step": 239673, "epoch": 2692, "lr": 3.3553791154731673e-06}
{"train_loss": 0.052758701145648956, "global_step": 239674, "epoch": 2692, "lr": 3.355170323481249e-06}
{"train_loss": 0.05599348992109299, "global_step": 239675, "epoch": 2692, "lr": 3.3549615377601653e-06}
{"train_loss": 0.04711297148147996, "global_step": 239676, "epoch": 2692, "lr": 3.3547527583099315e-06, "val_loss": 9.473647117614746}
{"train_loss": 0.06691095232963562, "global_step": 239677, "epoch": 2693, "lr": 3.354543985130587e-06}
{"train_loss": 0.04011337831616402, "global_step": 239678, "epoch": 2693, "lr": 3.354335218222154e-06}
{"train_loss": 0.05600366368889809, "global_step": 239679, "epoch": 2693, "lr": 3.3541264575846665e-06}
{"train_loss": 0.020916465669870377, "global_step": 239680, "epoch": 2693, "lr": 3.3539177032181403e-06}
{"train_loss": 0.0275541003793478, "global_step": 239681, "epoch": 2693, "lr": 3.3537089551226086e-06}
{"train_loss": 0.053865429013967514, "global_step": 239682, "epoch": 2693, "lr": 3.3535002132981107e-06}
{"train_loss": 0.053072698414325714, "global_step": 239683, "epoch": 2693, "lr": 3.3532914777446636e-06}
{"train_loss": 0.05409671738743782, "global_step": 239684, "epoch": 2693, "lr": 3.3530827484622994e-06}
{"train_loss": 0.033269304782152176, "global_step": 239685, "epoch": 2693, "lr": 3.3528740254510415e-06}
{"train_loss": 0.041525937616825104, "global_step": 239686, "epoch": 2693, "lr": 3.3526653087109283e-06}
{"train_loss": 0.14593955874443054, "global_step": 239687, "epoch": 2693, "lr": 3.3524565982419765e-06}
{"train_loss": 0.046138983219861984, "global_step": 239688, "epoch": 2693, "lr": 3.3522478940442136e-06}
{"train_loss": 0.039690952748060226, "global_step": 239689, "epoch": 2693, "lr": 3.3520391961176844e-06}
{"train_loss": 0.036814119666814804, "global_step": 239690, "epoch": 2693, "lr": 3.351830504462389e-06}
{"train_loss": 0.03313380107283592, "global_step": 239691, "epoch": 2693, "lr": 3.3516218190783877e-06}
{"train_loss": 0.027335844933986664, "global_step": 239692, "epoch": 2693, "lr": 3.3514131399656867e-06}
{"train_loss": 0.07310964167118073, "global_step": 239693, "epoch": 2693, "lr": 3.3512044671243135e-06}
{"train_loss": 0.03317226469516754, "global_step": 239694, "epoch": 2693, "lr": 3.3509958005543075e-06}
{"train_loss": 0.013225260190665722, "global_step": 239695, "epoch": 2693, "lr": 3.3507871402556846e-06}
{"train_loss": 0.034865908324718475, "global_step": 239696, "epoch": 2693, "lr": 3.3505784862284895e-06}
{"train_loss": 0.036537665873765945, "global_step": 239697, "epoch": 2693, "lr": 3.3503698384727335e-06}
{"train_loss": 0.09019410610198975, "global_step": 239698, "epoch": 2693, "lr": 3.350161196988455e-06}
{"train_loss": 0.06074705719947815, "global_step": 239699, "epoch": 2693, "lr": 3.3499525617756766e-06}
{"train_loss": 0.05861900746822357, "global_step": 239700, "epoch": 2693, "lr": 3.349743932834426e-06}
{"train_loss": 0.03546783700585365, "global_step": 239701, "epoch": 2693, "lr": 3.349535310164742e-06}
{"train_loss": 0.0639931857585907, "global_step": 239702, "epoch": 2693, "lr": 3.3493266937666356e-06}
{"train_loss": 0.04353540018200874, "global_step": 239703, "epoch": 2693, "lr": 3.3491180836401514e-06}
{"train_loss": 0.04271949082612991, "global_step": 239704, "epoch": 2693, "lr": 3.3489094797853003e-06}
{"train_loss": 0.08568811416625977, "global_step": 239705, "epoch": 2693, "lr": 3.348700882202133e-06}
{"train_loss": 0.0736781656742096, "global_step": 239706, "epoch": 2693, "lr": 3.3484922908906534e-06}
{"train_loss": 0.021960211917757988, "global_step": 239707, "epoch": 2693, "lr": 3.3482837058509075e-06}
{"train_loss": 0.04923747107386589, "global_step": 239708, "epoch": 2693, "lr": 3.3480751270829057e-06}
{"train_loss": 0.04109426215291023, "global_step": 239709, "epoch": 2693, "lr": 3.3478665545866984e-06}
{"train_loss": 0.07529892772436142, "global_step": 239710, "epoch": 2693, "lr": 3.347657988362296e-06}
{"train_loss": 0.05198703333735466, "global_step": 239711, "epoch": 2693, "lr": 3.347449428409738e-06}
{"train_loss": 0.03346902132034302, "global_step": 239712, "epoch": 2693, "lr": 3.347240874729035e-06}
{"train_loss": 0.01897660456597805, "global_step": 239713, "epoch": 2693, "lr": 3.3470323273202377e-06}
{"train_loss": 0.0998641848564148, "global_step": 239714, "epoch": 2693, "lr": 3.3468237861833563e-06}
{"train_loss": 0.0350155271589756, "global_step": 239715, "epoch": 2693, "lr": 3.3466152513184355e-06}
{"train_loss": 0.07085593789815903, "global_step": 239716, "epoch": 2693, "lr": 3.3464067227254815e-06}
{"train_loss": 0.0350249744951725, "global_step": 239717, "epoch": 2693, "lr": 3.346198200404538e-06}
{"train_loss": 0.06069842725992203, "global_step": 239718, "epoch": 2693, "lr": 3.345989684355633e-06}
{"train_loss": 0.05166846141219139, "global_step": 239719, "epoch": 2693, "lr": 3.345781174578788e-06}
{"train_loss": 0.06619864702224731, "global_step": 239720, "epoch": 2693, "lr": 3.3455726710740377e-06}
{"train_loss": 0.09500572830438614, "global_step": 239721, "epoch": 2693, "lr": 3.3453641738414033e-06}
{"train_loss": 0.024519307538866997, "global_step": 239722, "epoch": 2693, "lr": 3.3451556828809182e-06}
{"train_loss": 0.03857985511422157, "global_step": 239723, "epoch": 2693, "lr": 3.344947198192605e-06}
{"train_loss": 0.10056056082248688, "global_step": 239724, "epoch": 2693, "lr": 3.3447387197765025e-06}
{"train_loss": 0.05673792585730553, "global_step": 239725, "epoch": 2693, "lr": 3.3445302476326324e-06}
{"train_loss": 0.03712189942598343, "global_step": 239726, "epoch": 2693, "lr": 3.3443217817610117e-06}
{"train_loss": 0.03425002470612526, "global_step": 239727, "epoch": 2693, "lr": 3.3441133221616848e-06}
{"train_loss": 0.024102676659822464, "global_step": 239728, "epoch": 2693, "lr": 3.343904868834663e-06}
{"train_loss": 0.02822178602218628, "global_step": 239729, "epoch": 2693, "lr": 3.343696421779996e-06}
{"train_loss": 0.036210305988788605, "global_step": 239730, "epoch": 2693, "lr": 3.3434879809977005e-06}
{"train_loss": 0.05501939356327057, "global_step": 239731, "epoch": 2693, "lr": 3.3432795464877986e-06}
{"train_loss": 0.09193124622106552, "global_step": 239732, "epoch": 2693, "lr": 3.3430711182503294e-06}
{"train_loss": 0.029466519132256508, "global_step": 239733, "epoch": 2693, "lr": 3.3428626962853095e-06}
{"train_loss": 0.061411891132593155, "global_step": 239734, "epoch": 2693, "lr": 3.3426542805927773e-06}
{"train_loss": 0.08459059149026871, "global_step": 239735, "epoch": 2693, "lr": 3.3424458711727503e-06}
{"train_loss": 0.07701552659273148, "global_step": 239736, "epoch": 2693, "lr": 3.3422374680252665e-06}
{"train_loss": 0.0340263657271862, "global_step": 239737, "epoch": 2693, "lr": 3.3420290711503543e-06}
{"train_loss": 0.00991899985820055, "global_step": 239738, "epoch": 2693, "lr": 3.3418206805480303e-06}
{"train_loss": 0.021652093157172203, "global_step": 239739, "epoch": 2693, "lr": 3.3416122962183385e-06}
{"train_loss": 0.047851573675870895, "global_step": 239740, "epoch": 2693, "lr": 3.34140391816129e-06}
{"train_loss": 0.046508897095918655, "global_step": 239741, "epoch": 2693, "lr": 3.341195546376924e-06}
{"train_loss": 0.016381530091166496, "global_step": 239742, "epoch": 2693, "lr": 3.3409871808652627e-06}
{"train_loss": 0.059803809970617294, "global_step": 239743, "epoch": 2693, "lr": 3.340778821626345e-06}
{"train_loss": 0.07563253492116928, "global_step": 239744, "epoch": 2693, "lr": 3.340570468660181e-06}
{"train_loss": 0.07421573251485825, "global_step": 239745, "epoch": 2693, "lr": 3.3403621219668224e-06}
{"train_loss": 0.03141549974679947, "global_step": 239746, "epoch": 2693, "lr": 3.340153781546268e-06}
{"train_loss": 0.05145275965332985, "global_step": 239747, "epoch": 2693, "lr": 3.3399454473985737e-06}
{"train_loss": 0.03545553982257843, "global_step": 239748, "epoch": 2693, "lr": 3.3397371195237447e-06}
{"train_loss": 0.0583529956638813, "global_step": 239749, "epoch": 2693, "lr": 3.339528797921826e-06}
{"train_loss": 0.043873243033885956, "global_step": 239750, "epoch": 2693, "lr": 3.3393204825928336e-06}
{"train_loss": 0.019701378419995308, "global_step": 239751, "epoch": 2693, "lr": 3.3391121735368126e-06}
{"train_loss": 0.010100842453539371, "global_step": 239752, "epoch": 2693, "lr": 3.3389038707537625e-06}
{"train_loss": 0.057654958218336105, "global_step": 239753, "epoch": 2693, "lr": 3.338695574243733e-06}
{"train_loss": 0.018046073615550995, "global_step": 239754, "epoch": 2693, "lr": 3.3384872840067584e-06}
{"train_loss": 0.037121035158634186, "global_step": 239755, "epoch": 2693, "lr": 3.3382790000428433e-06}
{"train_loss": 0.0233546644449234, "global_step": 239756, "epoch": 2693, "lr": 3.3380707223520326e-06}
{"train_loss": 0.06928841024637222, "global_step": 239757, "epoch": 2693, "lr": 3.3378624509343483e-06}
{"train_loss": 0.02259894087910652, "global_step": 239758, "epoch": 2693, "lr": 3.3376541857898235e-06}
{"train_loss": 0.04421306028962135, "global_step": 239759, "epoch": 2693, "lr": 3.3374459269184755e-06}
{"train_loss": 0.06173032149672508, "global_step": 239760, "epoch": 2693, "lr": 3.3372376743203484e-06}
{"train_loss": 0.0474364310503006, "global_step": 239761, "epoch": 2693, "lr": 3.3370294279954527e-06}
{"train_loss": 0.031765166670084, "global_step": 239762, "epoch": 2693, "lr": 3.3368211879438283e-06}
{"train_loss": 0.04912802577018738, "global_step": 239763, "epoch": 2693, "lr": 3.3366129541655023e-06}
{"train_loss": 0.08703959733247757, "global_step": 239764, "epoch": 2693, "lr": 3.3364047266604912e-06}
{"train_loss": 0.0485525851273972, "global_step": 239765, "epoch": 2693, "lr": 3.336196505428846e-06, "val_loss": 9.509515762329102}
{"train_loss": 0.04243772476911545, "global_step": 239766, "epoch": 2694, "lr": 3.3359882904705653e-06}
{"train_loss": 0.024107052013278008, "global_step": 239767, "epoch": 2694, "lr": 3.3357800817857e-06}
{"train_loss": 0.05848313495516777, "global_step": 239768, "epoch": 2694, "lr": 3.3355718793742772e-06}
{"train_loss": 0.0323752760887146, "global_step": 239769, "epoch": 2694, "lr": 3.3353636832363035e-06}
{"train_loss": 0.016532858833670616, "global_step": 239770, "epoch": 2694, "lr": 3.335155493371833e-06}
{"train_loss": 0.06269357353448868, "global_step": 239771, "epoch": 2694, "lr": 3.3349473097808725e-06}
{"train_loss": 0.039153121411800385, "global_step": 239772, "epoch": 2694, "lr": 3.33473913246346e-06}
{"train_loss": 0.05509670823812485, "global_step": 239773, "epoch": 2694, "lr": 3.3345309614196294e-06}
{"train_loss": 0.016237875446677208, "global_step": 239774, "epoch": 2694, "lr": 3.3343227966493974e-06}
{"train_loss": 0.05025942996144295, "global_step": 239775, "epoch": 2694, "lr": 3.334114638152802e-06}
{"train_loss": 0.04386357590556145, "global_step": 239776, "epoch": 2694, "lr": 3.333906485929855e-06}
{"train_loss": 0.03454510495066643, "global_step": 239777, "epoch": 2694, "lr": 3.333698339980612e-06}
{"train_loss": 0.049340032041072845, "global_step": 239778, "epoch": 2694, "lr": 3.333490200305073e-06}
{"train_loss": 0.04203154146671295, "global_step": 239779, "epoch": 2694, "lr": 3.3332820669032817e-06}
{"train_loss": 0.07408058643341064, "global_step": 239780, "epoch": 2694, "lr": 3.3330739397752553e-06}
{"train_loss": 0.05473115295171738, "global_step": 239781, "epoch": 2694, "lr": 3.3328658189210384e-06}
{"train_loss": 0.06768278032541275, "global_step": 239782, "epoch": 2694, "lr": 3.3326577043406417e-06}
{"train_loss": 0.03941652923822403, "global_step": 239783, "epoch": 2694, "lr": 3.332449596034104e-06}
{"train_loss": 0.031860075891017914, "global_step": 239784, "epoch": 2694, "lr": 3.332241494001448e-06}
{"train_loss": 0.0520455576479435, "global_step": 239785, "epoch": 2694, "lr": 3.3320333982427065e-06}
{"train_loss": 0.06487191468477249, "global_step": 239786, "epoch": 2694, "lr": 3.3318253087578966e-06}
{"train_loss": 0.05014531686902046, "global_step": 239787, "epoch": 2694, "lr": 3.3316172255470623e-06}
{"train_loss": 0.010103391483426094, "global_step": 239788, "epoch": 2694, "lr": 3.3314091486102205e-06}
{"train_loss": 0.07572883367538452, "global_step": 239789, "epoch": 2694, "lr": 3.3312010779474046e-06}
{"train_loss": 0.05045464634895325, "global_step": 239790, "epoch": 2694, "lr": 3.330993013558631e-06}
{"train_loss": 0.0352536179125309, "global_step": 239791, "epoch": 2694, "lr": 3.3307849554439385e-06}
{"train_loss": 0.04500911384820938, "global_step": 239792, "epoch": 2694, "lr": 3.3305769036033607e-06}
{"train_loss": 0.07567592710256577, "global_step": 239793, "epoch": 2694, "lr": 3.330368858036914e-06}
{"train_loss": 0.07161334156990051, "global_step": 239794, "epoch": 2694, "lr": 3.330160818744632e-06}
{"train_loss": 0.025198161602020264, "global_step": 239795, "epoch": 2694, "lr": 3.3299527857265366e-06}
{"train_loss": 0.03594269976019859, "global_step": 239796, "epoch": 2694, "lr": 3.329744758982667e-06}
{"train_loss": 0.047472722828388214, "global_step": 239797, "epoch": 2694, "lr": 3.329536738513034e-06}
{"train_loss": 0.019389133900403976, "global_step": 239798, "epoch": 2694, "lr": 3.3293287243176874e-06}
{"train_loss": 0.035179365426301956, "global_step": 239799, "epoch": 2694, "lr": 3.3291207163966386e-06}
{"train_loss": 0.0669247955083847, "global_step": 239800, "epoch": 2694, "lr": 3.3289127147499212e-06}
{"train_loss": 0.07026058435440063, "global_step": 239801, "epoch": 2694, "lr": 3.3287047193775623e-06}
{"train_loss": 0.04237576574087143, "global_step": 239802, "epoch": 2694, "lr": 3.3284967302795956e-06}
{"train_loss": 0.07482381910085678, "global_step": 239803, "epoch": 2694, "lr": 3.328288747456038e-06}
{"train_loss": 0.037377920001745224, "global_step": 239804, "epoch": 2694, "lr": 3.328080770906922e-06}
{"train_loss": 0.03484194725751877, "global_step": 239805, "epoch": 2694, "lr": 3.3278728006322812e-06}
{"train_loss": 0.05042477324604988, "global_step": 239806, "epoch": 2694, "lr": 3.327664836632133e-06}
{"train_loss": 0.03897550329566002, "global_step": 239807, "epoch": 2694, "lr": 3.327456878906521e-06}
{"train_loss": 0.04948136955499649, "global_step": 239808, "epoch": 2694, "lr": 3.3272489274554507e-06}
{"train_loss": 0.03643391281366348, "global_step": 239809, "epoch": 2694, "lr": 3.3270409822789727e-06}
{"train_loss": 0.06427738815546036, "global_step": 239810, "epoch": 2694, "lr": 3.326833043377098e-06}
{"train_loss": 0.09877421706914902, "global_step": 239811, "epoch": 2694, "lr": 3.32662511074987e-06}
{"train_loss": 0.05187352001667023, "global_step": 239812, "epoch": 2694, "lr": 3.3264171843973013e-06}
{"train_loss": 0.07603133469820023, "global_step": 239813, "epoch": 2694, "lr": 3.3262092643194353e-06}
{"train_loss": 0.017498593777418137, "global_step": 239814, "epoch": 2694, "lr": 3.326001350516278e-06}
{"train_loss": 0.06625746190547943, "global_step": 239815, "epoch": 2694, "lr": 3.3257934429878845e-06}
{"train_loss": 0.051727838814258575, "global_step": 239816, "epoch": 2694, "lr": 3.3255855417342606e-06}
{"train_loss": 0.04855860397219658, "global_step": 239817, "epoch": 2694, "lr": 3.325377646755451e-06}
{"train_loss": 0.026864275336265564, "global_step": 239818, "epoch": 2694, "lr": 3.3251697580514663e-06}
{"train_loss": 0.03421279415488243, "global_step": 239819, "epoch": 2694, "lr": 3.324961875622351e-06}
{"train_loss": 0.06205087527632713, "global_step": 239820, "epoch": 2694, "lr": 3.3247539994681166e-06}
{"train_loss": 0.029628053307533264, "global_step": 239821, "epoch": 2694, "lr": 3.324546129588807e-06}
{"train_loss": 0.02186966873705387, "global_step": 239822, "epoch": 2694, "lr": 3.324338265984439e-06}
{"train_loss": 0.0378369502723217, "global_step": 239823, "epoch": 2694, "lr": 3.324130408655052e-06}
{"train_loss": 0.042846132069826126, "global_step": 239824, "epoch": 2694, "lr": 3.3239225576006616e-06}
{"train_loss": 0.03066052496433258, "global_step": 239825, "epoch": 2694, "lr": 3.3237147128213073e-06}
{"train_loss": 0.051997069269418716, "global_step": 239826, "epoch": 2694, "lr": 3.3235068743170006e-06}
{"train_loss": 0.09298957884311676, "global_step": 239827, "epoch": 2694, "lr": 3.32329904208778e-06}
{"train_loss": 0.06726136803627014, "global_step": 239828, "epoch": 2694, "lr": 3.3230912161336782e-06}
{"train_loss": 0.0238505806773901, "global_step": 239829, "epoch": 2694, "lr": 3.322883396454712e-06}
{"train_loss": 0.07569971680641174, "global_step": 239830, "epoch": 2694, "lr": 3.322675583050927e-06}
{"train_loss": 0.048071157187223434, "global_step": 239831, "epoch": 2694, "lr": 3.322467775922328e-06}
{"train_loss": 0.03095671534538269, "global_step": 239832, "epoch": 2694, "lr": 3.3222599750689643e-06}
{"train_loss": 0.07016909122467041, "global_step": 239833, "epoch": 2694, "lr": 3.322052180490842e-06}
{"train_loss": 0.04701681435108185, "global_step": 239834, "epoch": 2694, "lr": 3.3218443921880115e-06}
{"train_loss": 0.07414750009775162, "global_step": 239835, "epoch": 2694, "lr": 3.3216366101604836e-06}
{"train_loss": 0.05804716795682907, "global_step": 239836, "epoch": 2694, "lr": 3.321428834408297e-06}
{"train_loss": 0.05856424942612648, "global_step": 239837, "epoch": 2694, "lr": 3.321221064931468e-06}
{"train_loss": 0.07985703647136688, "global_step": 239838, "epoch": 2694, "lr": 3.3210133017300415e-06}
{"train_loss": 0.021938370540738106, "global_step": 239839, "epoch": 2694, "lr": 3.320805544804029e-06}
{"train_loss": 0.0632527619600296, "global_step": 239840, "epoch": 2694, "lr": 3.3205977941534738e-06}
{"train_loss": 0.034532513469457626, "global_step": 239841, "epoch": 2694, "lr": 3.3203900497783934e-06}
{"train_loss": 0.022056100890040398, "global_step": 239842, "epoch": 2694, "lr": 3.3201823116788045e-06}
{"train_loss": 0.047863271087408066, "global_step": 239843, "epoch": 2694, "lr": 3.3199745798547623e-06}
{"train_loss": 0.061586201190948486, "global_step": 239844, "epoch": 2694, "lr": 3.319766854306272e-06}
{"train_loss": 0.02244657091796398, "global_step": 239845, "epoch": 2694, "lr": 3.3195591350333786e-06}
{"train_loss": 0.041688624769449234, "global_step": 239846, "epoch": 2694, "lr": 3.3193514220360935e-06}
{"train_loss": 0.022587686777114868, "global_step": 239847, "epoch": 2694, "lr": 3.31914371531446e-06}
{"train_loss": 0.046853192150592804, "global_step": 239848, "epoch": 2694, "lr": 3.31893601486849e-06}
{"train_loss": 0.086415596306324, "global_step": 239849, "epoch": 2694, "lr": 3.318728320698228e-06}
{"train_loss": 0.06098124012351036, "global_step": 239850, "epoch": 2694, "lr": 3.3185206328036843e-06}
{"train_loss": 0.0524725615978241, "global_step": 239851, "epoch": 2694, "lr": 3.3183129511849045e-06}
{"train_loss": 0.07279492169618607, "global_step": 239852, "epoch": 2694, "lr": 3.318105275841904e-06}
{"train_loss": 0.07202109694480896, "global_step": 239853, "epoch": 2694, "lr": 3.3178976067747226e-06}
{"train_loss": 0.04884753403452675, "global_step": 239854, "epoch": 2694, "lr": 3.317689943983371e-06, "val_loss": 9.412345886230469}
{"train_loss": 0.055838633328676224, "global_step": 239855, "epoch": 2695, "lr": 3.317482287467899e-06}
{"train_loss": 0.03673916310071945, "global_step": 239856, "epoch": 2695, "lr": 3.317274637228307e-06}
{"train_loss": 0.026343800127506256, "global_step": 239857, "epoch": 2695, "lr": 3.3170669932646503e-06}
{"train_loss": 0.07324089109897614, "global_step": 239858, "epoch": 2695, "lr": 3.31685935557694e-06}
{"train_loss": 0.04083079472184181, "global_step": 239859, "epoch": 2695, "lr": 3.316651724165215e-06}
{"train_loss": 0.019730739295482635, "global_step": 239860, "epoch": 2695, "lr": 3.3164440990294866e-06}
{"train_loss": 0.03185401111841202, "global_step": 239861, "epoch": 2695, "lr": 3.316236480169804e-06}
{"train_loss": 0.06807109713554382, "global_step": 239862, "epoch": 2695, "lr": 3.3160288675861793e-06}
{"train_loss": 0.05478626489639282, "global_step": 239863, "epoch": 2695, "lr": 3.31582126127864e-06}
{"train_loss": 0.031239798292517662, "global_step": 239864, "epoch": 2695, "lr": 3.31561366124723e-06}
{"train_loss": 0.0369197316467762, "global_step": 239865, "epoch": 2695, "lr": 3.3154060674919607e-06}
{"train_loss": 0.041187841445207596, "global_step": 239866, "epoch": 2695, "lr": 3.315198480012871e-06}
{"train_loss": 0.05750337988138199, "global_step": 239867, "epoch": 2695, "lr": 3.3149908988099776e-06}
{"train_loss": 0.033231765031814575, "global_step": 239868, "epoch": 2695, "lr": 3.314783323883319e-06}
{"train_loss": 0.0417349711060524, "global_step": 239869, "epoch": 2695, "lr": 3.3145757552329127e-06}
{"train_loss": 0.0415082685649395, "global_step": 239870, "epoch": 2695, "lr": 3.3143681928588023e-06}
{"train_loss": 0.038155216723680496, "global_step": 239871, "epoch": 2695, "lr": 3.314160636760999e-06}
{"train_loss": 0.020939255133271217, "global_step": 239872, "epoch": 2695, "lr": 3.313953086939542e-06}
{"train_loss": 0.07098656892776489, "global_step": 239873, "epoch": 2695, "lr": 3.313745543394453e-06}
{"train_loss": 0.05229227989912033, "global_step": 239874, "epoch": 2695, "lr": 3.313538006125766e-06}
{"train_loss": 0.03772842884063721, "global_step": 239875, "epoch": 2695, "lr": 3.3133304751334913e-06}
{"train_loss": 0.03756211698055267, "global_step": 239876, "epoch": 2695, "lr": 3.313122950417685e-06}
{"train_loss": 0.055615950375795364, "global_step": 239877, "epoch": 2695, "lr": 3.312915431978353e-06}
{"train_loss": 0.08578302711248398, "global_step": 239878, "epoch": 2695, "lr": 3.312707919815533e-06}
{"train_loss": 0.061732660979032516, "global_step": 239879, "epoch": 2695, "lr": 3.3125004139292593e-06}
{"train_loss": 0.07380928844213486, "global_step": 239880, "epoch": 2695, "lr": 3.3122929143195368e-06}
{"train_loss": 0.03142102435231209, "global_step": 239881, "epoch": 2695, "lr": 3.312085420986416e-06}
{"train_loss": 0.03882332518696785, "global_step": 239882, "epoch": 2695, "lr": 3.311877933929908e-06}
{"train_loss": 0.024593498557806015, "global_step": 239883, "epoch": 2695, "lr": 3.3116704531500564e-06}
{"train_loss": 0.051330484449863434, "global_step": 239884, "epoch": 2695, "lr": 3.3114629786468733e-06}
{"train_loss": 0.10112059861421585, "global_step": 239885, "epoch": 2695, "lr": 3.3112555104204025e-06}
{"train_loss": 0.025569114834070206, "global_step": 239886, "epoch": 2695, "lr": 3.3110480484706606e-06}
{"train_loss": 0.04687803238630295, "global_step": 239887, "epoch": 2695, "lr": 3.3108405927976814e-06}
{"train_loss": 0.05649580806493759, "global_step": 239888, "epoch": 2695, "lr": 3.310633143401487e-06}
{"train_loss": 0.03606104105710983, "global_step": 239889, "epoch": 2695, "lr": 3.310425700282116e-06}
{"train_loss": 0.07116054743528366, "global_step": 239890, "epoch": 2695, "lr": 3.310218263439585e-06}
{"train_loss": 0.03920842707157135, "global_step": 239891, "epoch": 2695, "lr": 3.3100108328739275e-06}
{"train_loss": 0.07892297208309174, "global_step": 239892, "epoch": 2695, "lr": 3.309803408585166e-06}
{"train_loss": 0.03421055153012276, "global_step": 239893, "epoch": 2695, "lr": 3.3095959905733385e-06}
{"train_loss": 0.02934955805540085, "global_step": 239894, "epoch": 2695, "lr": 3.309388578838457e-06}
{"train_loss": 0.020750636234879494, "global_step": 239895, "epoch": 2695, "lr": 3.3091811733805713e-06}
{"train_loss": 0.03034805878996849, "global_step": 239896, "epoch": 2695, "lr": 3.3089737741996863e-06}
{"train_loss": 0.07807867974042892, "global_step": 239897, "epoch": 2695, "lr": 3.3087663812958524e-06}
{"train_loss": 0.03222647309303284, "global_step": 239898, "epoch": 2695, "lr": 3.3085589946690697e-06}
{"train_loss": 0.015448683872818947, "global_step": 239899, "epoch": 2695, "lr": 3.308351614319388e-06}
{"train_loss": 0.03854062408208847, "global_step": 239900, "epoch": 2695, "lr": 3.308144240246841e-06}
{"train_loss": 0.03893224895000458, "global_step": 239901, "epoch": 2695, "lr": 3.307936872451428e-06}
{"train_loss": 0.070297010242939, "global_step": 239902, "epoch": 2695, "lr": 3.3077295109332107e-06}
{"train_loss": 0.08099959790706635, "global_step": 239903, "epoch": 2695, "lr": 3.307522155692183e-06}
{"train_loss": 0.043662335723638535, "global_step": 239904, "epoch": 2695, "lr": 3.3073148067284066e-06}
{"train_loss": 0.013504942879080772, "global_step": 239905, "epoch": 2695, "lr": 3.3071074640418807e-06}
{"train_loss": 0.03354163095355034, "global_step": 239906, "epoch": 2695, "lr": 3.3069001276326504e-06}
{"train_loss": 0.050073977559804916, "global_step": 239907, "epoch": 2695, "lr": 3.306692797500732e-06}
{"train_loss": 0.03958599641919136, "global_step": 239908, "epoch": 2695, "lr": 3.3064854736461703e-06}
{"train_loss": 0.048219598829746246, "global_step": 239909, "epoch": 2695, "lr": 3.3062781560689705e-06}
{"train_loss": 0.015623422339558601, "global_step": 239910, "epoch": 2695, "lr": 3.3060708447691878e-06}
{"train_loss": 0.06314241141080856, "global_step": 239911, "epoch": 2695, "lr": 3.3058635397468175e-06}
{"train_loss": 0.08293434232473373, "global_step": 239912, "epoch": 2695, "lr": 3.3056562410019197e-06}
{"train_loss": 0.04831409454345703, "global_step": 239913, "epoch": 2695, "lr": 3.3054489485344954e-06}
{"train_loss": 0.0640684962272644, "global_step": 239914, "epoch": 2695, "lr": 3.3052416623445936e-06}
{"train_loss": 0.06498520076274872, "global_step": 239915, "epoch": 2695, "lr": 3.305034382432226e-06}
{"train_loss": 0.043739527463912964, "global_step": 239916, "epoch": 2695, "lr": 3.3048271087974313e-06}
{"train_loss": 0.037269968539476395, "global_step": 239917, "epoch": 2695, "lr": 3.3046198414402318e-06}
{"train_loss": 0.0371534489095211, "global_step": 239918, "epoch": 2695, "lr": 3.3044125803606607e-06}
{"train_loss": 0.04467965289950371, "global_step": 239919, "epoch": 2695, "lr": 3.30420532555874e-06}
{"train_loss": 0.03619856759905815, "global_step": 239920, "epoch": 2695, "lr": 3.303998077034498e-06}
{"train_loss": 0.03300139307975769, "global_step": 239921, "epoch": 2695, "lr": 3.3037908347879677e-06}
{"train_loss": 0.027870342135429382, "global_step": 239922, "epoch": 2695, "lr": 3.3035835988191656e-06}
{"train_loss": 0.07410840690135956, "global_step": 239923, "epoch": 2695, "lr": 3.303376369128136e-06}
{"train_loss": 0.00681134220212698, "global_step": 239924, "epoch": 2695, "lr": 3.303169145714896e-06}
{"train_loss": 0.025353392586112022, "global_step": 239925, "epoch": 2695, "lr": 3.302961928579479e-06}
{"train_loss": 0.056058309972286224, "global_step": 239926, "epoch": 2695, "lr": 3.302754717721901e-06}
{"train_loss": 0.056010037660598755, "global_step": 239927, "epoch": 2695, "lr": 3.3025475131422067e-06}
{"train_loss": 0.013759941793978214, "global_step": 239928, "epoch": 2695, "lr": 3.3023403148404075e-06}
{"train_loss": 0.04649416729807854, "global_step": 239929, "epoch": 2695, "lr": 3.3021331228165476e-06}
{"train_loss": 0.06734826415777206, "global_step": 239930, "epoch": 2695, "lr": 3.301925937070638e-06}
{"train_loss": 0.08312719315290451, "global_step": 239931, "epoch": 2695, "lr": 3.301718757602723e-06}
{"train_loss": 0.05073132738471031, "global_step": 239932, "epoch": 2695, "lr": 3.30151158441282e-06}
{"train_loss": 0.015261671505868435, "global_step": 239933, "epoch": 2695, "lr": 3.301304417500961e-06}
{"train_loss": 0.06336989998817444, "global_step": 239934, "epoch": 2695, "lr": 3.3010972568671693e-06}
{"train_loss": 0.07053892314434052, "global_step": 239935, "epoch": 2695, "lr": 3.3008901025114723e-06}
{"train_loss": 0.018419727683067322, "global_step": 239936, "epoch": 2695, "lr": 3.300682954433909e-06}
{"train_loss": 0.055712491273880005, "global_step": 239937, "epoch": 2695, "lr": 3.3004758126344958e-06}
{"train_loss": 0.04170585796236992, "global_step": 239938, "epoch": 2695, "lr": 3.3002686771132718e-06}
{"train_loss": 0.05430278554558754, "global_step": 239939, "epoch": 2695, "lr": 3.300061547870248e-06}
{"train_loss": 0.03270997479557991, "global_step": 239940, "epoch": 2695, "lr": 3.2998544249054687e-06}
{"train_loss": 0.036539528518915176, "global_step": 239941, "epoch": 2695, "lr": 3.2996473082189506e-06}
{"train_loss": 0.04096987843513489, "global_step": 239942, "epoch": 2695, "lr": 3.299440197810727e-06}
{"train_loss": 0.045734171630040316, "global_step": 239943, "epoch": 2695, "lr": 3.2992330936808203e-06, "val_loss": 9.45447826385498, "train_action_mse_error": 10.601327896118164}
{"train_loss": 0.044616181403398514, "global_step": 239944, "epoch": 2696, "lr": 3.2990259958292745e-06}
{"train_loss": 0.057387325912714005, "global_step": 239945, "epoch": 2696, "lr": 3.2988189042560903e-06}
{"train_loss": 0.06020182743668556, "global_step": 239946, "epoch": 2696, "lr": 3.2986118189613226e-06}
{"train_loss": 0.03964151069521904, "global_step": 239947, "epoch": 2696, "lr": 3.298404739944977e-06}
{"train_loss": 0.023655172437429428, "global_step": 239948, "epoch": 2696, "lr": 3.2981976672071035e-06}
{"train_loss": 0.055558864027261734, "global_step": 239949, "epoch": 2696, "lr": 3.297990600747708e-06}
{"train_loss": 0.036066021770238876, "global_step": 239950, "epoch": 2696, "lr": 3.2977835405668345e-06}
{"train_loss": 0.058661989867687225, "global_step": 239951, "epoch": 2696, "lr": 3.2975764866645e-06}
{"train_loss": 0.03427252918481827, "global_step": 239952, "epoch": 2696, "lr": 3.297369439040743e-06}
{"train_loss": 0.05122249945998192, "global_step": 239953, "epoch": 2696, "lr": 3.2971623976955803e-06}
{"train_loss": 0.0343460775911808, "global_step": 239954, "epoch": 2696, "lr": 3.2969553626290507e-06}
{"train_loss": 0.039996013045310974, "global_step": 239955, "epoch": 2696, "lr": 3.296748333841165e-06}
{"train_loss": 0.0704021081328392, "global_step": 239956, "epoch": 2696, "lr": 3.2965413113319744e-06}
{"train_loss": 0.06475293636322021, "global_step": 239957, "epoch": 2696, "lr": 3.296334295101494e-06}
{"train_loss": 0.06210216134786606, "global_step": 239958, "epoch": 2696, "lr": 3.2961272851497472e-06}
{"train_loss": 0.014995278790593147, "global_step": 239959, "epoch": 2696, "lr": 3.295920281476772e-06}
{"train_loss": 0.04027961939573288, "global_step": 239960, "epoch": 2696, "lr": 3.2957132840825797e-06}
{"train_loss": 0.05043080449104309, "global_step": 239961, "epoch": 2696, "lr": 3.295506292967221e-06}
{"train_loss": 0.043761491775512695, "global_step": 239962, "epoch": 2696, "lr": 3.295299308130706e-06}
{"train_loss": 0.09057272225618362, "global_step": 239963, "epoch": 2696, "lr": 3.295092329573074e-06}
{"train_loss": 0.06510218977928162, "global_step": 239964, "epoch": 2696, "lr": 3.2948853572943415e-06}
{"train_loss": 0.10154541581869125, "global_step": 239965, "epoch": 2696, "lr": 3.2946783912945477e-06}
{"train_loss": 0.05480958893895149, "global_step": 239966, "epoch": 2696, "lr": 3.294471431573709e-06}
{"train_loss": 0.03530534356832504, "global_step": 239967, "epoch": 2696, "lr": 3.2942644781318645e-06}
{"train_loss": 0.012445763684809208, "global_step": 239968, "epoch": 2696, "lr": 3.29405753096903e-06}
{"train_loss": 0.02119000442326069, "global_step": 239969, "epoch": 2696, "lr": 3.2938505900852514e-06}
{"train_loss": 0.07038258016109467, "global_step": 239970, "epoch": 2696, "lr": 3.2936436554805385e-06}
{"train_loss": 0.030393235385417938, "global_step": 239971, "epoch": 2696, "lr": 3.2934367271549194e-06}
{"train_loss": 0.08022188395261765, "global_step": 239972, "epoch": 2696, "lr": 3.2932298051084385e-06}
{"train_loss": 0.05273057147860527, "global_step": 239973, "epoch": 2696, "lr": 3.2930228893411073e-06}
{"train_loss": 0.04302480071783066, "global_step": 239974, "epoch": 2696, "lr": 3.29281597985297e-06}
{"train_loss": 0.033269938081502914, "global_step": 239975, "epoch": 2696, "lr": 3.292609076644032e-06}
{"train_loss": 0.049694765359163284, "global_step": 239976, "epoch": 2696, "lr": 3.2924021797143435e-06}
{"train_loss": 0.07164280861616135, "global_step": 239977, "epoch": 2696, "lr": 3.292195289063915e-06}
{"train_loss": 0.02052929438650608, "global_step": 239978, "epoch": 2696, "lr": 3.2919884046927864e-06}
{"train_loss": 0.037865523248910904, "global_step": 239979, "epoch": 2696, "lr": 3.291781526600979e-06}
{"train_loss": 0.023997165262699127, "global_step": 239980, "epoch": 2696, "lr": 3.2915746547885205e-06}
{"train_loss": 0.06238207593560219, "global_step": 239981, "epoch": 2696, "lr": 3.2913677892554396e-06}
{"train_loss": 0.034803155809640884, "global_step": 239982, "epoch": 2696, "lr": 3.2911609300017744e-06}
{"train_loss": 0.06831321865320206, "global_step": 239983, "epoch": 2696, "lr": 3.2909540770275304e-06}
{"train_loss": 0.054224416613578796, "global_step": 239984, "epoch": 2696, "lr": 3.290747230332758e-06}
{"train_loss": 0.07239237427711487, "global_step": 239985, "epoch": 2696, "lr": 3.290540389917468e-06}
{"train_loss": 0.08611618727445602, "global_step": 239986, "epoch": 2696, "lr": 3.290333555781705e-06}
{"train_loss": 0.030749913305044174, "global_step": 239987, "epoch": 2696, "lr": 3.2901267279254743e-06}
{"train_loss": 0.01858757995069027, "global_step": 239988, "epoch": 2696, "lr": 3.289919906348832e-06}
{"train_loss": 0.05627783015370369, "global_step": 239989, "epoch": 2696, "lr": 3.289713091051777e-06}
{"train_loss": 0.031222449615597725, "global_step": 239990, "epoch": 2696, "lr": 3.2895062820343605e-06}
{"train_loss": 0.0411098450422287, "global_step": 239991, "epoch": 2696, "lr": 3.2892994792965927e-06}
{"train_loss": 0.06419822573661804, "global_step": 239992, "epoch": 2696, "lr": 3.289092682838518e-06}
{"train_loss": 0.09198511391878128, "global_step": 239993, "epoch": 2696, "lr": 3.288885892660143e-06}
{"train_loss": 0.06205805763602257, "global_step": 239994, "epoch": 2696, "lr": 3.288679108761522e-06}
{"train_loss": 0.05024503171443939, "global_step": 239995, "epoch": 2696, "lr": 3.2884723311426667e-06}
{"train_loss": 0.03924905136227608, "global_step": 239996, "epoch": 2696, "lr": 3.2882655598035937e-06}
{"train_loss": 0.032282017171382904, "global_step": 239997, "epoch": 2696, "lr": 3.2880587947443587e-06}
{"train_loss": 0.018169958144426346, "global_step": 239998, "epoch": 2696, "lr": 3.2878520359649667e-06}
{"train_loss": 0.019852785393595695, "global_step": 239999, "epoch": 2696, "lr": 3.287645283465457e-06}
{"train_loss": 0.07709236443042755, "global_step": 240000, "epoch": 2696, "lr": 3.2874385372458517e-06}
{"train_loss": 0.08079422265291214, "global_step": 240001, "epoch": 2696, "lr": 3.287231797306184e-06}
{"train_loss": 0.02427670545876026, "global_step": 240002, "epoch": 2696, "lr": 3.2870250636464763e-06}
{"train_loss": 0.04448065161705017, "global_step": 240003, "epoch": 2696, "lr": 3.2868183362667616e-06}
{"train_loss": 0.07070580869913101, "global_step": 240004, "epoch": 2696, "lr": 3.286611615167057e-06}
{"train_loss": 0.031542617827653885, "global_step": 240005, "epoch": 2696, "lr": 3.2864049003474063e-06}
{"train_loss": 0.039628129452466965, "global_step": 240006, "epoch": 2696, "lr": 3.2861981918078266e-06}
{"train_loss": 0.03209904953837395, "global_step": 240007, "epoch": 2696, "lr": 3.2859914895483402e-06}
{"train_loss": 0.07590217888355255, "global_step": 240008, "epoch": 2696, "lr": 3.2857847935689967e-06}
{"train_loss": 0.041789401322603226, "global_step": 240009, "epoch": 2696, "lr": 3.285578103869802e-06}
{"train_loss": 0.033706698566675186, "global_step": 240010, "epoch": 2696, "lr": 3.2853714204507948e-06}
{"train_loss": 0.04850620776414871, "global_step": 240011, "epoch": 2696, "lr": 3.285164743311997e-06}
{"train_loss": 0.02708524838089943, "global_step": 240012, "epoch": 2696, "lr": 3.284958072453448e-06}
{"train_loss": 0.07681431621313095, "global_step": 240013, "epoch": 2696, "lr": 3.284751407875153e-06}
{"train_loss": 0.05700894817709923, "global_step": 240014, "epoch": 2696, "lr": 3.2845447495771675e-06}
{"train_loss": 0.06798738241195679, "global_step": 240015, "epoch": 2696, "lr": 3.2843380975594916e-06}
{"train_loss": 0.040371254086494446, "global_step": 240016, "epoch": 2696, "lr": 3.284131451822181e-06}
{"train_loss": 0.014301965944468975, "global_step": 240017, "epoch": 2696, "lr": 3.2839248123652413e-06}
{"train_loss": 0.07046148926019669, "global_step": 240018, "epoch": 2696, "lr": 3.283718179188716e-06}
{"train_loss": 0.047696277499198914, "global_step": 240019, "epoch": 2696, "lr": 3.2835115522926117e-06}
{"train_loss": 0.03743554651737213, "global_step": 240020, "epoch": 2696, "lr": 3.2833049316769836e-06}
{"train_loss": 0.038355451077222824, "global_step": 240021, "epoch": 2696, "lr": 3.283098317341837e-06}
{"train_loss": 0.02423367090523243, "global_step": 240022, "epoch": 2696, "lr": 3.282891709287217e-06}
{"train_loss": 0.04260767996311188, "global_step": 240023, "epoch": 2696, "lr": 3.2826851075131336e-06}
{"train_loss": 0.09143012762069702, "global_step": 240024, "epoch": 2696, "lr": 3.282478512019632e-06}
{"train_loss": 0.06152036413550377, "global_step": 240025, "epoch": 2696, "lr": 3.2822719228067234e-06}
{"train_loss": 0.04148784279823303, "global_step": 240026, "epoch": 2696, "lr": 3.282065339874446e-06}
{"train_loss": 0.02480483613908291, "global_step": 240027, "epoch": 2696, "lr": 3.281858763222828e-06}
{"train_loss": 0.07971605658531189, "global_step": 240028, "epoch": 2696, "lr": 3.2816521928518916e-06}
{"train_loss": 0.013444843702018261, "global_step": 240029, "epoch": 2696, "lr": 3.2814456287616757e-06}
{"train_loss": 0.03079916164278984, "global_step": 240030, "epoch": 2696, "lr": 3.281239070952197e-06}
{"train_loss": 0.026643462479114532, "global_step": 240031, "epoch": 2696, "lr": 3.2810325194234825e-06}
{"train_loss": 0.04817750967327464, "global_step": 240032, "epoch": 2696, "lr": 3.2808259741755666e-06, "val_loss": 9.422589302062988}
{"train_loss": 0.0526408776640892, "global_step": 240033, "epoch": 2697, "lr": 3.2806194352084708e-06}
{"train_loss": 0.06267312914133072, "global_step": 240034, "epoch": 2697, "lr": 3.280412902522234e-06}
{"train_loss": 0.04048660397529602, "global_step": 240035, "epoch": 2697, "lr": 3.2802063761168734e-06}
{"train_loss": 0.010938365012407303, "global_step": 240036, "epoch": 2697, "lr": 3.2799998559924104e-06}
{"train_loss": 0.06022421643137932, "global_step": 240037, "epoch": 2697, "lr": 3.279793342148896e-06}
{"train_loss": 0.046056099236011505, "global_step": 240038, "epoch": 2697, "lr": 3.279586834586329e-06}
{"train_loss": 0.03901948034763336, "global_step": 240039, "epoch": 2697, "lr": 3.279380333304766e-06}
{"train_loss": 0.06186777353286743, "global_step": 240040, "epoch": 2697, "lr": 3.2791738383042115e-06}
{"train_loss": 0.05355720594525337, "global_step": 240041, "epoch": 2697, "lr": 3.278967349584705e-06}
{"train_loss": 0.047320686280727386, "global_step": 240042, "epoch": 2697, "lr": 3.278760867146269e-06}
{"train_loss": 0.0561029277741909, "global_step": 240043, "epoch": 2697, "lr": 3.2785543909889416e-06}
{"train_loss": 0.03697866573929787, "global_step": 240044, "epoch": 2697, "lr": 3.2783479211127343e-06}
{"train_loss": 0.04750153794884682, "global_step": 240045, "epoch": 2697, "lr": 3.278141457517686e-06}
{"train_loss": 0.020580878481268883, "global_step": 240046, "epoch": 2697, "lr": 3.27793500020383e-06}
{"train_loss": 0.09359192848205566, "global_step": 240047, "epoch": 2697, "lr": 3.2777285491711774e-06}
{"train_loss": 0.051735635846853256, "global_step": 240048, "epoch": 2697, "lr": 3.2775221044197723e-06}
{"train_loss": 0.026814758777618408, "global_step": 240049, "epoch": 2697, "lr": 3.2773156659496263e-06}
{"train_loss": 0.04984302818775177, "global_step": 240050, "epoch": 2697, "lr": 3.2771092337607833e-06}
{"train_loss": 0.054211799055337906, "global_step": 240051, "epoch": 2697, "lr": 3.276902807853255e-06}
{"train_loss": 0.05832791328430176, "global_step": 240052, "epoch": 2697, "lr": 3.2766963882270852e-06}
{"train_loss": 0.0445602685213089, "global_step": 240053, "epoch": 2697, "lr": 3.276489974882291e-06}
{"train_loss": 0.12989795207977295, "global_step": 240054, "epoch": 2697, "lr": 3.276283567818905e-06}
{"train_loss": 0.03937201946973801, "global_step": 240055, "epoch": 2697, "lr": 3.2760771670369507e-06}
{"train_loss": 0.03616931289434433, "global_step": 240056, "epoch": 2697, "lr": 3.27587077253646e-06}
{"train_loss": 0.053223468363285065, "global_step": 240057, "epoch": 2697, "lr": 3.2756643843174563e-06}
{"train_loss": 0.08390333503484726, "global_step": 240058, "epoch": 2697, "lr": 3.2754580023799774e-06}
{"train_loss": 0.10116085410118103, "global_step": 240059, "epoch": 2697, "lr": 3.275251626724035e-06}
{"train_loss": 0.024923749268054962, "global_step": 240060, "epoch": 2697, "lr": 3.2750452573496736e-06}
{"train_loss": 0.044793158769607544, "global_step": 240061, "epoch": 2697, "lr": 3.274838894256904e-06}
{"train_loss": 0.02285182476043701, "global_step": 240062, "epoch": 2697, "lr": 3.274632537445765e-06}
{"train_loss": 0.08116740733385086, "global_step": 240063, "epoch": 2697, "lr": 3.274426186916291e-06}
{"train_loss": 0.04126667603850365, "global_step": 240064, "epoch": 2697, "lr": 3.2742198426684857e-06}
{"train_loss": 0.06850127130746841, "global_step": 240065, "epoch": 2697, "lr": 3.2740135047024056e-06}
{"train_loss": 0.059198249131441116, "global_step": 240066, "epoch": 2697, "lr": 3.2738071730180566e-06}
{"train_loss": 0.03930897265672684, "global_step": 240067, "epoch": 2697, "lr": 3.273600847615482e-06}
{"train_loss": 0.04465646669268608, "global_step": 240068, "epoch": 2697, "lr": 3.2733945284947053e-06}
{"train_loss": 0.06553909182548523, "global_step": 240069, "epoch": 2697, "lr": 3.273188215655737e-06}
{"train_loss": 0.05706826224923134, "global_step": 240070, "epoch": 2697, "lr": 3.2729819090986325e-06}
{"train_loss": 0.03500470519065857, "global_step": 240071, "epoch": 2697, "lr": 3.272775608823397e-06}
{"train_loss": 0.02559792622923851, "global_step": 240072, "epoch": 2697, "lr": 3.2725693148300708e-06}
{"train_loss": 0.037011243402957916, "global_step": 240073, "epoch": 2697, "lr": 3.2723630271186854e-06}
{"train_loss": 0.042978134006261826, "global_step": 240074, "epoch": 2697, "lr": 3.2721567456892475e-06}
{"train_loss": 0.04481648653745651, "global_step": 240075, "epoch": 2697, "lr": 3.271950470541807e-06}
{"train_loss": 0.04969114810228348, "global_step": 240076, "epoch": 2697, "lr": 3.27174420167638e-06}
{"train_loss": 0.07276066392660141, "global_step": 240077, "epoch": 2697, "lr": 3.271537939093e-06}
{"train_loss": 0.024924691766500473, "global_step": 240078, "epoch": 2697, "lr": 3.271331682791684e-06}
{"train_loss": 0.042573291808366776, "global_step": 240079, "epoch": 2697, "lr": 3.2711254327724816e-06}
{"train_loss": 0.07809765636920929, "global_step": 240080, "epoch": 2697, "lr": 3.2709191890353928e-06}
{"train_loss": 0.0317104198038578, "global_step": 240081, "epoch": 2697, "lr": 3.2707129515804623e-06}
{"train_loss": 0.07079585641622543, "global_step": 240082, "epoch": 2697, "lr": 3.2705067204077232e-06}
{"train_loss": 0.047246769070625305, "global_step": 240083, "epoch": 2697, "lr": 3.270300495517187e-06}
{"train_loss": 0.0409022755920887, "global_step": 240084, "epoch": 2697, "lr": 3.270094276908897e-06}
{"train_loss": 0.016480734571814537, "global_step": 240085, "epoch": 2697, "lr": 3.2698880645828655e-06}
{"train_loss": 0.08043058961629868, "global_step": 240086, "epoch": 2697, "lr": 3.2696818585391363e-06}
{"train_loss": 0.02642236091196537, "global_step": 240087, "epoch": 2697, "lr": 3.269475658777721e-06}
{"train_loss": 0.04507729038596153, "global_step": 240088, "epoch": 2697, "lr": 3.2692694652986633e-06}
{"train_loss": 0.09311274439096451, "global_step": 240089, "epoch": 2697, "lr": 3.2690632781019694e-06}
{"train_loss": 0.03681768849492073, "global_step": 240090, "epoch": 2697, "lr": 3.2688570971876943e-06}
{"train_loss": 0.030508875846862793, "global_step": 240091, "epoch": 2697, "lr": 3.2686509225558436e-06}
{"train_loss": 0.053870152682065964, "global_step": 240092, "epoch": 2697, "lr": 3.2684447542064565e-06}
{"train_loss": 0.01949363760650158, "global_step": 240093, "epoch": 2697, "lr": 3.268238592139555e-06}
{"train_loss": 0.08031707257032394, "global_step": 240094, "epoch": 2697, "lr": 3.268032436355173e-06}
{"train_loss": 0.06035223230719566, "global_step": 240095, "epoch": 2697, "lr": 3.2678262868533316e-06}
{"train_loss": 0.06466694921255112, "global_step": 240096, "epoch": 2697, "lr": 3.267620143634065e-06}
{"train_loss": 0.0446254163980484, "global_step": 240097, "epoch": 2697, "lr": 3.267414006697389e-06}
{"train_loss": 0.04348643124103546, "global_step": 240098, "epoch": 2697, "lr": 3.2672078760433435e-06}
{"train_loss": 0.041100043803453445, "global_step": 240099, "epoch": 2697, "lr": 3.2670017516719554e-06}
{"train_loss": 0.050994087010622025, "global_step": 240100, "epoch": 2697, "lr": 3.266795633583242e-06}
{"train_loss": 0.04566996172070503, "global_step": 240101, "epoch": 2697, "lr": 3.266589521777247e-06}
{"train_loss": 0.03012824058532715, "global_step": 240102, "epoch": 2697, "lr": 3.2663834162539877e-06}
{"train_loss": 0.050416525453329086, "global_step": 240103, "epoch": 2697, "lr": 3.2661773170134915e-06}
{"train_loss": 0.05574270337820053, "global_step": 240104, "epoch": 2697, "lr": 3.265971224055786e-06}
{"train_loss": 0.03462269902229309, "global_step": 240105, "epoch": 2697, "lr": 3.2657651373809105e-06}
{"train_loss": 0.03779605031013489, "global_step": 240106, "epoch": 2697, "lr": 3.2655590569888816e-06}
{"train_loss": 0.06550481915473938, "global_step": 240107, "epoch": 2697, "lr": 3.2653529828797157e-06}
{"train_loss": 0.020513435825705528, "global_step": 240108, "epoch": 2697, "lr": 3.2651469150534684e-06}
{"train_loss": 0.01082745473831892, "global_step": 240109, "epoch": 2697, "lr": 3.26494085351014e-06}
{"train_loss": 0.05519159883260727, "global_step": 240110, "epoch": 2697, "lr": 3.26473479824978e-06}
{"train_loss": 0.019002841785550117, "global_step": 240111, "epoch": 2697, "lr": 3.2645287492724107e-06}
{"train_loss": 0.03760968893766403, "global_step": 240112, "epoch": 2697, "lr": 3.2643227065780435e-06}
{"train_loss": 0.05127571523189545, "global_step": 240113, "epoch": 2697, "lr": 3.264116670166728e-06}
{"train_loss": 0.04209332913160324, "global_step": 240114, "epoch": 2697, "lr": 3.26391064003847e-06}
{"train_loss": 0.040503498166799545, "global_step": 240115, "epoch": 2697, "lr": 3.263704616193325e-06}
{"train_loss": 0.04099184647202492, "global_step": 240116, "epoch": 2697, "lr": 3.2634985986312928e-06}
{"train_loss": 0.026256628334522247, "global_step": 240117, "epoch": 2697, "lr": 3.2632925873524124e-06}
{"train_loss": 0.04035339504480362, "global_step": 240118, "epoch": 2697, "lr": 3.2630865823567224e-06}
{"train_loss": 0.08374509960412979, "global_step": 240119, "epoch": 2697, "lr": 3.2628805836442343e-06}
{"train_loss": 0.031311698257923126, "global_step": 240120, "epoch": 2697, "lr": 3.262674591214987e-06}
{"train_loss": 0.048461501926099985, "global_step": 240121, "epoch": 2697, "lr": 3.262468605068997e-06, "val_loss": 9.383078575134277}
{"train_loss": 0.05072067305445671, "global_step": 240122, "epoch": 2698, "lr": 3.2622626252063084e-06}
{"train_loss": 0.07693162560462952, "global_step": 240123, "epoch": 2698, "lr": 3.2620566516269323e-06}
{"train_loss": 0.06157476827502251, "global_step": 240124, "epoch": 2698, "lr": 3.261850684330908e-06}
{"train_loss": 0.05645434185862541, "global_step": 240125, "epoch": 2698, "lr": 3.2616447233182468e-06}
{"train_loss": 0.05635704845190048, "global_step": 240126, "epoch": 2698, "lr": 3.261438768588998e-06}
{"train_loss": 0.053829044103622437, "global_step": 240127, "epoch": 2698, "lr": 3.261232820143173e-06}
{"train_loss": 0.05111560598015785, "global_step": 240128, "epoch": 2698, "lr": 3.261026877980816e-06}
{"train_loss": 0.03521398454904556, "global_step": 240129, "epoch": 2698, "lr": 3.260820942101933e-06}
{"train_loss": 0.05979696288704872, "global_step": 240130, "epoch": 2698, "lr": 3.260615012506568e-06}
{"train_loss": 0.05712350085377693, "global_step": 240131, "epoch": 2698, "lr": 3.2604090891947435e-06}
{"train_loss": 0.047081734985113144, "global_step": 240132, "epoch": 2698, "lr": 3.260203172166487e-06}
{"train_loss": 0.05406029894948006, "global_step": 240133, "epoch": 2698, "lr": 3.2599972614218265e-06}
{"train_loss": 0.055749889463186264, "global_step": 240134, "epoch": 2698, "lr": 3.259791356960784e-06}
{"train_loss": 0.024773815646767616, "global_step": 240135, "epoch": 2698, "lr": 3.2595854587833984e-06}
{"train_loss": 0.034555111080408096, "global_step": 240136, "epoch": 2698, "lr": 3.259379566889692e-06}
{"train_loss": 0.040801677852869034, "global_step": 240137, "epoch": 2698, "lr": 3.2591736812796926e-06}
{"train_loss": 0.08244733512401581, "global_step": 240138, "epoch": 2698, "lr": 3.2589678019534275e-06}
{"train_loss": 0.04644748568534851, "global_step": 240139, "epoch": 2698, "lr": 3.258761928910925e-06}
{"train_loss": 0.03747653588652611, "global_step": 240140, "epoch": 2698, "lr": 3.2585560621522073e-06}
{"train_loss": 0.04668181389570236, "global_step": 240141, "epoch": 2698, "lr": 3.258350201677318e-06}
{"train_loss": 0.04340735822916031, "global_step": 240142, "epoch": 2698, "lr": 3.258144347486258e-06}
{"train_loss": 0.037213284522295, "global_step": 240143, "epoch": 2698, "lr": 3.2579384995790887e-06}
{"train_loss": 0.02918282151222229, "global_step": 240144, "epoch": 2698, "lr": 3.2577326579558034e-06}
{"train_loss": 0.04529108107089996, "global_step": 240145, "epoch": 2698, "lr": 3.257526822616458e-06}
{"train_loss": 0.052103593945503235, "global_step": 240146, "epoch": 2698, "lr": 3.2573209935610695e-06}
{"train_loss": 0.03370247036218643, "global_step": 240147, "epoch": 2698, "lr": 3.257115170789654e-06}
{"train_loss": 0.04147185757756233, "global_step": 240148, "epoch": 2698, "lr": 3.2569093543022565e-06}
{"train_loss": 0.05314632132649422, "global_step": 240149, "epoch": 2698, "lr": 3.2567035440988935e-06}
{"train_loss": 0.01678529754281044, "global_step": 240150, "epoch": 2698, "lr": 3.256497740179604e-06}
{"train_loss": 0.061311352998018265, "global_step": 240151, "epoch": 2698, "lr": 3.2562919425444095e-06}
{"train_loss": 0.05025769770145416, "global_step": 240152, "epoch": 2698, "lr": 3.256086151193327e-06}
{"train_loss": 0.05116602033376694, "global_step": 240153, "epoch": 2698, "lr": 3.2558803661263957e-06}
{"train_loss": 0.025154033675789833, "global_step": 240154, "epoch": 2698, "lr": 3.2556745873436432e-06}
{"train_loss": 0.03904560208320618, "global_step": 240155, "epoch": 2698, "lr": 3.2554688148450972e-06}
{"train_loss": 0.033314913511276245, "global_step": 240156, "epoch": 2698, "lr": 3.2552630486307856e-06}
{"train_loss": 0.03151224926114082, "global_step": 240157, "epoch": 2698, "lr": 3.25505728870073e-06}
{"train_loss": 0.07328884303569794, "global_step": 240158, "epoch": 2698, "lr": 3.254851535054965e-06}
{"train_loss": 0.057939350605010986, "global_step": 240159, "epoch": 2698, "lr": 3.254645787693511e-06}
{"train_loss": 0.040205128490924835, "global_step": 240160, "epoch": 2698, "lr": 3.2544400466164082e-06}
{"train_loss": 0.03470929339528084, "global_step": 240161, "epoch": 2698, "lr": 3.2542343118236673e-06}
{"train_loss": 0.05990633741021156, "global_step": 240162, "epoch": 2698, "lr": 3.2540285833153327e-06}
{"train_loss": 0.07477150112390518, "global_step": 240163, "epoch": 2698, "lr": 3.2538228610914155e-06}
{"train_loss": 0.06246098503470421, "global_step": 240164, "epoch": 2698, "lr": 3.2536171451519606e-06}
{"train_loss": 0.019415223971009254, "global_step": 240165, "epoch": 2698, "lr": 3.253411435496978e-06}
{"train_loss": 0.03776373341679573, "global_step": 240166, "epoch": 2698, "lr": 3.2532057321265184e-06}
{"train_loss": 0.08395212888717651, "global_step": 240167, "epoch": 2698, "lr": 3.253000035040582e-06}
{"train_loss": 0.03208504244685173, "global_step": 240168, "epoch": 2698, "lr": 3.2527943442392184e-06}
{"train_loss": 0.035583965480327606, "global_step": 240169, "epoch": 2698, "lr": 3.2525886597224387e-06}
{"train_loss": 0.0517551563680172, "global_step": 240170, "epoch": 2698, "lr": 3.252382981490287e-06}
{"train_loss": 0.04582645371556282, "global_step": 240171, "epoch": 2698, "lr": 3.2521773095427755e-06}
{"train_loss": 0.05471085384488106, "global_step": 240172, "epoch": 2698, "lr": 3.251971643879942e-06}
{"train_loss": 0.04900520667433739, "global_step": 240173, "epoch": 2698, "lr": 3.2517659845018143e-06}
{"train_loss": 0.04465128481388092, "global_step": 240174, "epoch": 2698, "lr": 3.2515603314084098e-06}
{"train_loss": 0.05785210430622101, "global_step": 240175, "epoch": 2698, "lr": 3.251354684599772e-06}
{"train_loss": 0.04028471186757088, "global_step": 240176, "epoch": 2698, "lr": 3.251149044075913e-06}
{"train_loss": 0.05561809241771698, "global_step": 240177, "epoch": 2698, "lr": 3.2509434098368706e-06}
{"train_loss": 0.053169138729572296, "global_step": 240178, "epoch": 2698, "lr": 3.250737781882668e-06}
{"train_loss": 0.02939477562904358, "global_step": 240179, "epoch": 2698, "lr": 3.2505321602133378e-06}
{"train_loss": 0.04111083224415779, "global_step": 240180, "epoch": 2698, "lr": 3.2503265448288965e-06}
{"train_loss": 0.030084094032645226, "global_step": 240181, "epoch": 2698, "lr": 3.2501209357293837e-06}
{"train_loss": 0.11871535331010818, "global_step": 240182, "epoch": 2698, "lr": 3.2499153329148156e-06}
{"train_loss": 0.030017944052815437, "global_step": 240183, "epoch": 2698, "lr": 3.249709736385237e-06}
{"train_loss": 0.04216717928647995, "global_step": 240184, "epoch": 2698, "lr": 3.2495041461406638e-06}
{"train_loss": 0.03313903883099556, "global_step": 240185, "epoch": 2698, "lr": 3.249298562181119e-06}
{"train_loss": 0.022935327142477036, "global_step": 240186, "epoch": 2698, "lr": 3.249092984506641e-06}
{"train_loss": 0.04712352901697159, "global_step": 240187, "epoch": 2698, "lr": 3.2488874131172466e-06}
{"train_loss": 0.03882868215441704, "global_step": 240188, "epoch": 2698, "lr": 3.24868184801298e-06}
{"train_loss": 0.04574921727180481, "global_step": 240189, "epoch": 2698, "lr": 3.248476289193847e-06}
{"train_loss": 0.027303116396069527, "global_step": 240190, "epoch": 2698, "lr": 3.2482707366598976e-06}
{"train_loss": 0.03596794977784157, "global_step": 240191, "epoch": 2698, "lr": 3.248065190411137e-06}
{"train_loss": 0.035578690469264984, "global_step": 240192, "epoch": 2698, "lr": 3.24785965044761e-06}
{"train_loss": 0.07897596061229706, "global_step": 240193, "epoch": 2698, "lr": 3.247654116769333e-06}
{"train_loss": 0.05226743966341019, "global_step": 240194, "epoch": 2698, "lr": 3.2474485893763507e-06}
{"train_loss": 0.04970524087548256, "global_step": 240195, "epoch": 2698, "lr": 3.2472430682686682e-06}
{"train_loss": 0.04428757354617119, "global_step": 240196, "epoch": 2698, "lr": 3.2470375534463306e-06}
{"train_loss": 0.08188921213150024, "global_step": 240197, "epoch": 2698, "lr": 3.2468320449093536e-06}
{"train_loss": 0.03068184107542038, "global_step": 240198, "epoch": 2698, "lr": 3.246626542657777e-06}
{"train_loss": 0.03045291267335415, "global_step": 240199, "epoch": 2698, "lr": 3.246421046691611e-06}
{"train_loss": 0.05699114874005318, "global_step": 240200, "epoch": 2698, "lr": 3.2462155570109063e-06}
{"train_loss": 0.06463591754436493, "global_step": 240201, "epoch": 2698, "lr": 3.2460100736156684e-06}
{"train_loss": 0.04047594219446182, "global_step": 240202, "epoch": 2698, "lr": 3.2458045965059413e-06}
{"train_loss": 0.0317864865064621, "global_step": 240203, "epoch": 2698, "lr": 3.2455991256817366e-06}
{"train_loss": 0.02300557680428028, "global_step": 240204, "epoch": 2698, "lr": 3.2453936611431033e-06}
{"train_loss": 0.03685377910733223, "global_step": 240205, "epoch": 2698, "lr": 3.2451882028900483e-06}
{"train_loss": 0.06531000882387161, "global_step": 240206, "epoch": 2698, "lr": 3.244982750922615e-06}
{"train_loss": 0.03074578382074833, "global_step": 240207, "epoch": 2698, "lr": 3.244777305240815e-06}
{"train_loss": 0.05948571860790253, "global_step": 240208, "epoch": 2698, "lr": 3.244571865844687e-06}
{"train_loss": 0.02621951699256897, "global_step": 240209, "epoch": 2698, "lr": 3.244366432734258e-06}
{"train_loss": 0.04660447582267643, "global_step": 240210, "epoch": 2698, "lr": 3.2441610059095517e-06, "val_loss": 9.47705078125}
{"train_loss": 0.012171116657555103, "global_step": 240211, "epoch": 2699, "lr": 3.243955585370606e-06}
{"train_loss": 0.05941762030124664, "global_step": 240212, "epoch": 2699, "lr": 3.243750171117432e-06}
{"train_loss": 0.05546318739652634, "global_step": 240213, "epoch": 2699, "lr": 3.243544763150075e-06}
{"train_loss": 0.03742993623018265, "global_step": 240214, "epoch": 2699, "lr": 3.2433393614685446e-06}
{"train_loss": 0.04759139567613602, "global_step": 240215, "epoch": 2699, "lr": 3.2431339660728864e-06}
{"train_loss": 0.029983045533299446, "global_step": 240216, "epoch": 2699, "lr": 3.242928576963106e-06}
{"train_loss": 0.0378839485347271, "global_step": 240217, "epoch": 2699, "lr": 3.2427231941392577e-06}
{"train_loss": 0.007342654746025801, "global_step": 240218, "epoch": 2699, "lr": 3.2425178176013426e-06}
{"train_loss": 0.0704689621925354, "global_step": 240219, "epoch": 2699, "lr": 3.242312447349416e-06}
{"train_loss": 0.07706347107887268, "global_step": 240220, "epoch": 2699, "lr": 3.242107083383478e-06}
{"train_loss": 0.04064461588859558, "global_step": 240221, "epoch": 2699, "lr": 3.241901725703578e-06}
{"train_loss": 0.02589412033557892, "global_step": 240222, "epoch": 2699, "lr": 3.241696374309733e-06}
{"train_loss": 0.02056911028921604, "global_step": 240223, "epoch": 2699, "lr": 3.241491029201965e-06}
{"train_loss": 0.0683525875210762, "global_step": 240224, "epoch": 2699, "lr": 3.2412856903803133e-06}
{"train_loss": 0.033681128174066544, "global_step": 240225, "epoch": 2699, "lr": 3.2410803578447946e-06}
{"train_loss": 0.0472869947552681, "global_step": 240226, "epoch": 2699, "lr": 3.2408750315954526e-06}
{"train_loss": 0.03913335129618645, "global_step": 240227, "epoch": 2699, "lr": 3.240669711632299e-06}
{"train_loss": 0.050654444843530655, "global_step": 240228, "epoch": 2699, "lr": 3.2404643979553728e-06}
{"train_loss": 0.04625246301293373, "global_step": 240229, "epoch": 2699, "lr": 3.2402590905646844e-06}
{"train_loss": 0.06721298396587372, "global_step": 240230, "epoch": 2699, "lr": 3.2400537894602846e-06}
{"train_loss": 0.045433517545461655, "global_step": 240231, "epoch": 2699, "lr": 3.2398484946421838e-06}
{"train_loss": 0.04177560657262802, "global_step": 240232, "epoch": 2699, "lr": 3.239643206110421e-06}
{"train_loss": 0.03925313428044319, "global_step": 240233, "epoch": 2699, "lr": 3.239437923865013e-06}
{"train_loss": 0.076276995241642, "global_step": 240234, "epoch": 2699, "lr": 3.2392326479059985e-06}
{"train_loss": 0.06291017681360245, "global_step": 240235, "epoch": 2699, "lr": 3.239027378233389e-06}
{"train_loss": 0.03756535425782204, "global_step": 240236, "epoch": 2699, "lr": 3.2388221148472343e-06}
{"train_loss": 0.03664081543684006, "global_step": 240237, "epoch": 2699, "lr": 3.2386168577475395e-06}
{"train_loss": 0.07688412815332413, "global_step": 240238, "epoch": 2699, "lr": 3.238411606934355e-06}
{"train_loss": 0.06345362961292267, "global_step": 240239, "epoch": 2699, "lr": 3.238206362407681e-06}
{"train_loss": 0.05754309520125389, "global_step": 240240, "epoch": 2699, "lr": 3.2380011241675724e-06}
{"train_loss": 0.056700240820646286, "global_step": 240241, "epoch": 2699, "lr": 3.237795892214035e-06}
{"train_loss": 0.028716038912534714, "global_step": 240242, "epoch": 2699, "lr": 3.237590666547119e-06}
{"train_loss": 0.0497032105922699, "global_step": 240243, "epoch": 2699, "lr": 3.237385447166824e-06}
{"train_loss": 0.05745778977870941, "global_step": 240244, "epoch": 2699, "lr": 3.2371802340732004e-06}
{"train_loss": 0.11167102307081223, "global_step": 240245, "epoch": 2699, "lr": 3.23697502726627e-06}
{"train_loss": 0.0712534710764885, "global_step": 240246, "epoch": 2699, "lr": 3.23676982674605e-06}
{"train_loss": 0.01711714081466198, "global_step": 240247, "epoch": 2699, "lr": 3.236564632512584e-06}
{"train_loss": 0.02122032828629017, "global_step": 240248, "epoch": 2699, "lr": 3.2363594445658895e-06}
{"train_loss": 0.06823496520519257, "global_step": 240249, "epoch": 2699, "lr": 3.2361542629059995e-06}
{"train_loss": 0.03320860490202904, "global_step": 240250, "epoch": 2699, "lr": 3.23594908753293e-06}
{"train_loss": 0.09821721166372299, "global_step": 240251, "epoch": 2699, "lr": 3.2357439184467265e-06}
{"train_loss": 0.09092237055301666, "global_step": 240252, "epoch": 2699, "lr": 3.235538755647405e-06}
{"train_loss": 0.0584627240896225, "global_step": 240253, "epoch": 2699, "lr": 3.235333599134993e-06}
{"train_loss": 0.018249724060297012, "global_step": 240254, "epoch": 2699, "lr": 3.235128448909519e-06}
{"train_loss": 0.05330176651477814, "global_step": 240255, "epoch": 2699, "lr": 3.2349233049710216e-06}
{"train_loss": 0.022534742951393127, "global_step": 240256, "epoch": 2699, "lr": 3.234718167319506e-06}
{"train_loss": 0.020771581679582596, "global_step": 240257, "epoch": 2699, "lr": 3.2345130359550225e-06}
{"train_loss": 0.041199397295713425, "global_step": 240258, "epoch": 2699, "lr": 3.234307910877582e-06}
{"train_loss": 0.02980121038854122, "global_step": 240259, "epoch": 2699, "lr": 3.2341027920872237e-06}
{"train_loss": 0.053213972598314285, "global_step": 240260, "epoch": 2699, "lr": 3.233897679583964e-06}
{"train_loss": 0.08385568857192993, "global_step": 240261, "epoch": 2699, "lr": 3.233692573367847e-06}
{"train_loss": 0.026892056688666344, "global_step": 240262, "epoch": 2699, "lr": 3.2334874734388844e-06}
{"train_loss": 0.0393492728471756, "global_step": 240263, "epoch": 2699, "lr": 3.2332823797971035e-06}
{"train_loss": 0.07881469279527664, "global_step": 240264, "epoch": 2699, "lr": 3.233077292442549e-06}
{"train_loss": 0.018406972289085388, "global_step": 240265, "epoch": 2699, "lr": 3.2328722113752264e-06}
{"train_loss": 0.055854469537734985, "global_step": 240266, "epoch": 2699, "lr": 3.2326671365951857e-06}
{"train_loss": 0.056238025426864624, "global_step": 240267, "epoch": 2699, "lr": 3.232462068102432e-06}
{"train_loss": 0.041792403906583786, "global_step": 240268, "epoch": 2699, "lr": 3.23225700589701e-06}
{"train_loss": 0.04583534225821495, "global_step": 240269, "epoch": 2699, "lr": 3.232051949978937e-06}
{"train_loss": 0.04472828656435013, "global_step": 240270, "epoch": 2699, "lr": 3.231846900348251e-06}
{"train_loss": 0.027882009744644165, "global_step": 240271, "epoch": 2699, "lr": 3.2316418570049634e-06}
{"train_loss": 0.03794402256608009, "global_step": 240272, "epoch": 2699, "lr": 3.2314368199491186e-06}
{"train_loss": 0.0675218254327774, "global_step": 240273, "epoch": 2699, "lr": 3.231231789180733e-06}
{"train_loss": 0.02769814431667328, "global_step": 240274, "epoch": 2699, "lr": 3.231026764699846e-06}
{"train_loss": 0.058316249400377274, "global_step": 240275, "epoch": 2699, "lr": 3.2308217465064685e-06}
{"train_loss": 0.02655627951025963, "global_step": 240276, "epoch": 2699, "lr": 3.2306167346006446e-06}
{"train_loss": 0.029213091358542442, "global_step": 240277, "epoch": 2699, "lr": 3.2304117289823852e-06}
{"train_loss": 0.03819660097360611, "global_step": 240278, "epoch": 2699, "lr": 3.2302067296517357e-06}
{"train_loss": 0.05468758940696716, "global_step": 240279, "epoch": 2699, "lr": 3.2300017366087066e-06}
{"train_loss": 0.036585282534360886, "global_step": 240280, "epoch": 2699, "lr": 3.2297967498533365e-06}
{"train_loss": 0.06680372357368469, "global_step": 240281, "epoch": 2699, "lr": 3.2295917693856537e-06}
{"train_loss": 0.028818093240261078, "global_step": 240282, "epoch": 2699, "lr": 3.2293867952056744e-06}
{"train_loss": 0.022549357265233994, "global_step": 240283, "epoch": 2699, "lr": 3.2291818273134435e-06}
{"train_loss": 0.02692704275250435, "global_step": 240284, "epoch": 2699, "lr": 3.2289768657089714e-06}
{"train_loss": 0.051413536071777344, "global_step": 240285, "epoch": 2699, "lr": 3.228771910392303e-06}
{"train_loss": 0.03339949995279312, "global_step": 240286, "epoch": 2699, "lr": 3.2285669613634493e-06}
{"train_loss": 0.047893643379211426, "global_step": 240287, "epoch": 2699, "lr": 3.2283620186224493e-06}
{"train_loss": 0.03764847666025162, "global_step": 240288, "epoch": 2699, "lr": 3.2281570821693196e-06}
{"train_loss": 0.044530875980854034, "global_step": 240289, "epoch": 2699, "lr": 3.227952152004099e-06}
{"train_loss": 0.02711506187915802, "global_step": 240290, "epoch": 2699, "lr": 3.227747228126804e-06}
{"train_loss": 0.08875579386949539, "global_step": 240291, "epoch": 2699, "lr": 3.2275423105374792e-06}
{"train_loss": 0.041652802377939224, "global_step": 240292, "epoch": 2699, "lr": 3.22733739923613e-06}
{"train_loss": 0.06801043450832367, "global_step": 240293, "epoch": 2699, "lr": 3.227132494222801e-06}
{"train_loss": 0.029142364859580994, "global_step": 240294, "epoch": 2699, "lr": 3.226927595497509e-06}
{"train_loss": 0.03706669062376022, "global_step": 240295, "epoch": 2699, "lr": 3.2267227030602976e-06}
{"train_loss": 0.045327022671699524, "global_step": 240296, "epoch": 2699, "lr": 3.2265178169111675e-06}
{"train_loss": 0.014056584797799587, "global_step": 240297, "epoch": 2699, "lr": 3.226312937050174e-06}
{"train_loss": 0.014023075811564922, "global_step": 240298, "epoch": 2699, "lr": 3.226108063477329e-06}
{"train_loss": 0.04627508172514231, "global_step": 240299, "epoch": 2699, "lr": 3.2259031961926645e-06, "val_loss": 9.493438720703125}
{"train_loss": 0.0354502908885479, "global_step": 240300, "epoch": 2700, "lr": 3.2256983351962146e-06}
{"train_loss": 0.036327071487903595, "global_step": 240301, "epoch": 2700, "lr": 3.2254934804879845e-06}
{"train_loss": 0.02745368890464306, "global_step": 240302, "epoch": 2700, "lr": 3.22528863206803e-06}
{"train_loss": 0.07353706657886505, "global_step": 240303, "epoch": 2700, "lr": 3.225083789936356e-06}
{"train_loss": 0.011096618138253689, "global_step": 240304, "epoch": 2700, "lr": 3.2248789540930024e-06}
{"train_loss": 0.014888310804963112, "global_step": 240305, "epoch": 2700, "lr": 3.224674124537991e-06}
{"train_loss": 0.036349281668663025, "global_step": 240306, "epoch": 2700, "lr": 3.22446930127136e-06}
{"train_loss": 0.05750207230448723, "global_step": 240307, "epoch": 2700, "lr": 3.224264484293116e-06}
{"train_loss": 0.024036426097154617, "global_step": 240308, "epoch": 2700, "lr": 3.2240596736033135e-06}
{"train_loss": 0.08012498915195465, "global_step": 240309, "epoch": 2700, "lr": 3.2238548692019532e-06}
{"train_loss": 0.0631449893116951, "global_step": 240310, "epoch": 2700, "lr": 3.2236500710890905e-06}
{"train_loss": 0.0365154892206192, "global_step": 240311, "epoch": 2700, "lr": 3.2234452792647253e-06}
{"train_loss": 0.035310953855514526, "global_step": 240312, "epoch": 2700, "lr": 3.2232404937289016e-06}
{"train_loss": 0.059485722333192825, "global_step": 240313, "epoch": 2700, "lr": 3.2230357144816425e-06}
{"train_loss": 0.024589911103248596, "global_step": 240314, "epoch": 2700, "lr": 3.2228309415229808e-06}
{"train_loss": 0.03587838262319565, "global_step": 240315, "epoch": 2700, "lr": 3.222626174852933e-06}
{"train_loss": 0.05623363330960274, "global_step": 240316, "epoch": 2700, "lr": 3.222421414471527e-06}
{"train_loss": 0.015016253106296062, "global_step": 240317, "epoch": 2700, "lr": 3.2222166603788127e-06}
{"train_loss": 0.05127521976828575, "global_step": 240318, "epoch": 2700, "lr": 3.2220119125747905e-06}
{"train_loss": 0.03736409917473793, "global_step": 240319, "epoch": 2700, "lr": 3.221807171059504e-06}
{"train_loss": 0.046949807554483414, "global_step": 240320, "epoch": 2700, "lr": 3.2216024358329657e-06}
{"train_loss": 0.06729394197463989, "global_step": 240321, "epoch": 2700, "lr": 3.2213977068952237e-06}
{"train_loss": 0.0685080960392952, "global_step": 240322, "epoch": 2700, "lr": 3.2211929842462906e-06}
{"train_loss": 0.036318641155958176, "global_step": 240323, "epoch": 2700, "lr": 3.220988267886199e-06}
{"train_loss": 0.03881356492638588, "global_step": 240324, "epoch": 2700, "lr": 3.2207835578149716e-06}
{"train_loss": 0.0321083590388298, "global_step": 240325, "epoch": 2700, "lr": 3.2205788540326466e-06}
{"train_loss": 0.039572373032569885, "global_step": 240326, "epoch": 2700, "lr": 3.2203741565392355e-06}
{"train_loss": 0.03814244642853737, "global_step": 240327, "epoch": 2700, "lr": 3.220169465334788e-06}
{"train_loss": 0.046968746930360794, "global_step": 240328, "epoch": 2700, "lr": 3.2199647804193044e-06}
{"train_loss": 0.05362319573760033, "global_step": 240329, "epoch": 2700, "lr": 3.2197601017928348e-06}
{"train_loss": 0.04381897300481796, "global_step": 240330, "epoch": 2700, "lr": 3.2195554294553955e-06}
{"train_loss": 0.05476321652531624, "global_step": 240331, "epoch": 2700, "lr": 3.2193507634070195e-06}
{"train_loss": 0.035272348672151566, "global_step": 240332, "epoch": 2700, "lr": 3.2191461036477243e-06}
{"train_loss": 0.07086548209190369, "global_step": 240333, "epoch": 2700, "lr": 3.218941450177554e-06}
{"train_loss": 0.06780754029750824, "global_step": 240334, "epoch": 2700, "lr": 3.218736802996519e-06}
{"train_loss": 0.026986902579665184, "global_step": 240335, "epoch": 2700, "lr": 3.218532162104665e-06}
{"train_loss": 0.017039304599165916, "global_step": 240336, "epoch": 2700, "lr": 3.2183275275019964e-06}
{"train_loss": 0.11226015537977219, "global_step": 240337, "epoch": 2700, "lr": 3.218122899188564e-06}
{"train_loss": 0.03020492196083069, "global_step": 240338, "epoch": 2700, "lr": 3.217918277164389e-06}
{"train_loss": 0.03814025968313217, "global_step": 240339, "epoch": 2700, "lr": 3.217713661429478e-06}
{"train_loss": 0.040954336524009705, "global_step": 240340, "epoch": 2700, "lr": 3.2175090519838915e-06}
{"train_loss": 0.042499393224716187, "global_step": 240341, "epoch": 2700, "lr": 3.2173044488276237e-06}
{"train_loss": 0.047907158732414246, "global_step": 240342, "epoch": 2700, "lr": 3.2170998519607366e-06}
{"train_loss": 0.07095140963792801, "global_step": 240343, "epoch": 2700, "lr": 3.2168952613832237e-06}
{"train_loss": 0.027125902473926544, "global_step": 240344, "epoch": 2700, "lr": 3.216690677095141e-06}
{"train_loss": 0.030061719939112663, "global_step": 240345, "epoch": 2700, "lr": 3.2164860990964996e-06}
{"train_loss": 0.02020237222313881, "global_step": 240346, "epoch": 2700, "lr": 3.2162815273873325e-06}
{"train_loss": 0.02226630225777626, "global_step": 240347, "epoch": 2700, "lr": 3.2160769619676625e-06}
{"train_loss": 0.039602622389793396, "global_step": 240348, "epoch": 2700, "lr": 3.2158724028375276e-06}
{"train_loss": 0.06882985681295395, "global_step": 240349, "epoch": 2700, "lr": 3.2156678499969393e-06}
{"train_loss": 0.029111841693520546, "global_step": 240350, "epoch": 2700, "lr": 3.215463303445948e-06}
{"train_loss": 0.09463085979223251, "global_step": 240351, "epoch": 2700, "lr": 3.215258763184553e-06}
{"train_loss": 0.06552839279174805, "global_step": 240352, "epoch": 2700, "lr": 3.2150542292127995e-06}
{"train_loss": 0.023325925692915916, "global_step": 240353, "epoch": 2700, "lr": 3.2148497015307143e-06}
{"train_loss": 0.07372022420167923, "global_step": 240354, "epoch": 2700, "lr": 3.2146451801383206e-06}
{"train_loss": 0.02413848787546158, "global_step": 240355, "epoch": 2700, "lr": 3.2144406650356505e-06}
{"train_loss": 0.04163121059536934, "global_step": 240356, "epoch": 2700, "lr": 3.2142361562227273e-06}
{"train_loss": 0.033898353576660156, "global_step": 240357, "epoch": 2700, "lr": 3.214031653699584e-06}
{"train_loss": 0.04170655831694603, "global_step": 240358, "epoch": 2700, "lr": 3.213827157466237e-06}
{"train_loss": 0.03397892788052559, "global_step": 240359, "epoch": 2700, "lr": 3.2136226675227254e-06}
{"train_loss": 0.01202863734215498, "global_step": 240360, "epoch": 2700, "lr": 3.213418183869066e-06}
{"train_loss": 0.017371956259012222, "global_step": 240361, "epoch": 2700, "lr": 3.2132137065053025e-06}
{"train_loss": 0.049325041472911835, "global_step": 240362, "epoch": 2700, "lr": 3.213009235431441e-06}
{"train_loss": 0.035017989575862885, "global_step": 240363, "epoch": 2700, "lr": 3.212804770647532e-06}
{"train_loss": 0.03175249695777893, "global_step": 240364, "epoch": 2700, "lr": 3.2126003121535796e-06}
{"train_loss": 0.03391674906015396, "global_step": 240365, "epoch": 2700, "lr": 3.2123958599496297e-06}
{"train_loss": 0.0630798190832138, "global_step": 240366, "epoch": 2700, "lr": 3.2121914140356977e-06}
{"train_loss": 0.02445600926876068, "global_step": 240367, "epoch": 2700, "lr": 3.2119869744118235e-06}
{"train_loss": 0.07727088034152985, "global_step": 240368, "epoch": 2700, "lr": 3.2117825410780176e-06}
{"train_loss": 0.052741535007953644, "global_step": 240369, "epoch": 2700, "lr": 3.2115781140343306e-06}
{"train_loss": 0.04704451188445091, "global_step": 240370, "epoch": 2700, "lr": 3.2113736932807614e-06}
{"train_loss": 0.01771116442978382, "global_step": 240371, "epoch": 2700, "lr": 3.211169278817361e-06}
{"train_loss": 0.056621573865413666, "global_step": 240372, "epoch": 2700, "lr": 3.210964870644151e-06}
{"train_loss": 0.03745405748486519, "global_step": 240373, "epoch": 2700, "lr": 3.2107604687611535e-06}
{"train_loss": 0.04235095903277397, "global_step": 240374, "epoch": 2700, "lr": 3.210556073168397e-06}
{"train_loss": 0.031265947967767715, "global_step": 240375, "epoch": 2700, "lr": 3.2103516838659196e-06}
{"train_loss": 0.09548486769199371, "global_step": 240376, "epoch": 2700, "lr": 3.2101473008537275e-06}
{"train_loss": 0.052403293550014496, "global_step": 240377, "epoch": 2700, "lr": 3.20994292413187e-06}
{"train_loss": 0.07992872595787048, "global_step": 240378, "epoch": 2700, "lr": 3.2097385537003645e-06}
{"train_loss": 0.04324331134557724, "global_step": 240379, "epoch": 2700, "lr": 3.209534189559238e-06}
{"train_loss": 0.05850066617131233, "global_step": 240380, "epoch": 2700, "lr": 3.2093298317085187e-06}
{"train_loss": 0.04073600098490715, "global_step": 240381, "epoch": 2700, "lr": 3.209125480148234e-06}
{"train_loss": 0.0156533382833004, "global_step": 240382, "epoch": 2700, "lr": 3.2089211348784122e-06}
{"train_loss": 0.03821555897593498, "global_step": 240383, "epoch": 2700, "lr": 3.2087167958990805e-06}
{"train_loss": 0.012576407752931118, "global_step": 240384, "epoch": 2700, "lr": 3.208512463210267e-06}
{"train_loss": 0.0556689091026783, "global_step": 240385, "epoch": 2700, "lr": 3.2083081368119994e-06}
{"train_loss": 0.0689365565776825, "global_step": 240386, "epoch": 2700, "lr": 3.2081038167043055e-06}
{"train_loss": 0.01961062103509903, "global_step": 240387, "epoch": 2700, "lr": 3.207899502887207e-06}
{"train_loss": 0.04395031987616185, "global_step": 240388, "epoch": 2700, "lr": 3.2076951953607325e-06, "train/sim_max_reward_0": 0.9816279640549643, "train/sim_max_reward_1": 0.5227777394341908, "train/sim_max_reward_2": 0.9866617018732134, "train/sim_max_reward_3": 0.6908645892937205, "train/sim_max_reward_4": 1.0, "train/sim_max_reward_5": 0.9052371095317163, "test/sim_max_reward_4500000": 0.9503757391049178, "test/sim_max_reward_4500001": 0.9242707520060854, "test/sim_max_reward_4500002": 0.944428338151996, "test/sim_max_reward_4500003": 0.5904364039590525, "test/sim_max_reward_4500004": 0.8519797773442328, "test/sim_max_reward_4500005": 0.45750746178772767, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9948889858213049, "test/sim_max_reward_4500008": 0.31873589654526113, "test/sim_max_reward_4500009": 0.04931653441887075, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.07225207058211514, "test/sim_max_reward_4500013": 0.4583769225280085, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9766544430347401, "test/sim_max_reward_4500016": 0.9867383274274112, "test/sim_max_reward_4500017": 0.9969113697529872, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.3949729306195369, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9842994619203689, "test/sim_max_reward_4500022": 0.9888782841873138, "test/sim_max_reward_4500023": 0.7056634777453048, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.8721955092889443, "test/sim_max_reward_4500026": 0.396603680074735, "test/sim_max_reward_4500027": 0.9925606199353181, "test/sim_max_reward_4500028": 0.6499259969566102, "test/sim_max_reward_4500029": 0.9830317469449635, "test/sim_max_reward_4500030": 0.9920965797562853, "test/sim_max_reward_4500031": 0.29505121928301425, "test/sim_max_reward_4500032": 0.587076715610832, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.9953098686856803, "test/sim_max_reward_4500035": 0.49749751116240354, "test/sim_max_reward_4500036": 0.9324894361414975, "test/sim_max_reward_4500037": 0.9707283189861865, "test/sim_max_reward_4500038": 0.16211077655825226, "test/sim_max_reward_4500039": 0.9580625845327655, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.6087134177898514, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.01034476199353797, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.4656846777994308, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.9990869100038535, "train/mean_score": 0.8478615173646343, "test/mean_score": 0.6803051501688279, "val_loss": 9.433526039123535, "train_action_mse_error": 17.29024887084961}
{"train_loss": 0.09478631615638733, "global_step": 240389, "epoch": 2701, "lr": 3.207490894124926e-06}
{"train_loss": 0.035382963716983795, "global_step": 240390, "epoch": 2701, "lr": 3.207286599179793e-06}
{"train_loss": 0.02352031134068966, "global_step": 240391, "epoch": 2701, "lr": 3.2070823105253724e-06}
{"train_loss": 0.042243730276823044, "global_step": 240392, "epoch": 2701, "lr": 3.206878028161686e-06}
{"train_loss": 0.02781912125647068, "global_step": 240393, "epoch": 2701, "lr": 3.2066737520887735e-06}
{"train_loss": 0.05234137177467346, "global_step": 240394, "epoch": 2701, "lr": 3.2064694823066456e-06}
{"train_loss": 0.05036061257123947, "global_step": 240395, "epoch": 2701, "lr": 3.2062652188153407e-06}
{"train_loss": 0.046820301562547684, "global_step": 240396, "epoch": 2701, "lr": 3.206060961614882e-06}
{"train_loss": 0.0398658812046051, "global_step": 240397, "epoch": 2701, "lr": 3.205856710705302e-06}
{"train_loss": 0.033061131834983826, "global_step": 240398, "epoch": 2701, "lr": 3.205652466086617e-06}
{"train_loss": 0.04603319242596626, "global_step": 240399, "epoch": 2701, "lr": 3.205448227758867e-06}
{"train_loss": 0.049074213951826096, "global_step": 240400, "epoch": 2701, "lr": 3.2052439957220683e-06}
{"train_loss": 0.054336391389369965, "global_step": 240401, "epoch": 2701, "lr": 3.2050397699762592e-06}
{"train_loss": 0.04931046813726425, "global_step": 240402, "epoch": 2701, "lr": 3.204835550521462e-06}
{"train_loss": 0.06770772486925125, "global_step": 240403, "epoch": 2701, "lr": 3.2046313373577052e-06}
{"train_loss": 0.05088871717453003, "global_step": 240404, "epoch": 2701, "lr": 3.2044271304850104e-06}
{"train_loss": 0.036724675446748734, "global_step": 240405, "epoch": 2701, "lr": 3.2042229299034163e-06}
{"train_loss": 0.05973389744758606, "global_step": 240406, "epoch": 2701, "lr": 3.20401873561294e-06}
{"train_loss": 0.05297757685184479, "global_step": 240407, "epoch": 2701, "lr": 3.2038145476136094e-06}
{"train_loss": 0.046594392508268356, "global_step": 240408, "epoch": 2701, "lr": 3.203610365905463e-06}
{"train_loss": 0.040656331926584244, "global_step": 240409, "epoch": 2701, "lr": 3.2034061904885113e-06}
{"train_loss": 0.0554465726017952, "global_step": 240410, "epoch": 2701, "lr": 3.2032020213628055e-06}
{"train_loss": 0.03932866081595421, "global_step": 240411, "epoch": 2701, "lr": 3.202997858528356e-06}
{"train_loss": 0.036020003259181976, "global_step": 240412, "epoch": 2701, "lr": 3.202793701985185e-06}
{"train_loss": 0.02272901125252247, "global_step": 240413, "epoch": 2701, "lr": 3.2025895517333314e-06}
{"train_loss": 0.028705744072794914, "global_step": 240414, "epoch": 2701, "lr": 3.2023854077728178e-06}
{"train_loss": 0.07534129172563553, "global_step": 240415, "epoch": 2701, "lr": 3.202181270103677e-06}
{"train_loss": 0.039154499769210815, "global_step": 240416, "epoch": 2701, "lr": 3.2019771387259368e-06}
{"train_loss": 0.03081832081079483, "global_step": 240417, "epoch": 2701, "lr": 3.2017730136396084e-06}
{"train_loss": 0.020166579633951187, "global_step": 240418, "epoch": 2701, "lr": 3.201568894844742e-06}
{"train_loss": 0.02497114986181259, "global_step": 240419, "epoch": 2701, "lr": 3.2013647823413427e-06}
{"train_loss": 0.03876778483390808, "global_step": 240420, "epoch": 2701, "lr": 3.201160676129461e-06}
{"train_loss": 0.06860539317131042, "global_step": 240421, "epoch": 2701, "lr": 3.200956576209102e-06}
{"train_loss": 0.050293873995542526, "global_step": 240422, "epoch": 2701, "lr": 3.200752482580316e-06}
{"train_loss": 0.030263712629675865, "global_step": 240423, "epoch": 2701, "lr": 3.2005483952431027e-06}
{"train_loss": 0.053241975605487823, "global_step": 240424, "epoch": 2701, "lr": 3.200344314197523e-06}
{"train_loss": 0.02986104227602482, "global_step": 240425, "epoch": 2701, "lr": 3.200140239443572e-06}
{"train_loss": 0.02758835442364216, "global_step": 240426, "epoch": 2701, "lr": 3.1999361709812938e-06}
{"train_loss": 0.038845110684633255, "global_step": 240427, "epoch": 2701, "lr": 3.199732108810721e-06}
{"train_loss": 0.04041493311524391, "global_step": 240428, "epoch": 2701, "lr": 3.199528052931866e-06}
{"train_loss": 0.029140060767531395, "global_step": 240429, "epoch": 2701, "lr": 3.1993240033447724e-06}
{"train_loss": 0.04023129120469093, "global_step": 240430, "epoch": 2701, "lr": 3.1991199600494515e-06}
{"train_loss": 0.017157530412077904, "global_step": 240431, "epoch": 2701, "lr": 3.1989159230459476e-06}
{"train_loss": 0.06060231849551201, "global_step": 240432, "epoch": 2701, "lr": 3.1987118923342717e-06}
{"train_loss": 0.03237159922719002, "global_step": 240433, "epoch": 2701, "lr": 3.198507867914463e-06}
{"train_loss": 0.07062305510044098, "global_step": 240434, "epoch": 2701, "lr": 3.198303849786538e-06}
{"train_loss": 0.08069179952144623, "global_step": 240435, "epoch": 2701, "lr": 3.1980998379505413e-06}
{"train_loss": 0.02680053561925888, "global_step": 240436, "epoch": 2701, "lr": 3.197895832406478e-06}
{"train_loss": 0.03844544664025307, "global_step": 240437, "epoch": 2701, "lr": 3.1976918331543927e-06}
{"train_loss": 0.02581048384308815, "global_step": 240438, "epoch": 2701, "lr": 3.1974878401943075e-06}
{"train_loss": 0.01375341322273016, "global_step": 240439, "epoch": 2701, "lr": 3.1972838535262506e-06}
{"train_loss": 0.0696309506893158, "global_step": 240440, "epoch": 2701, "lr": 3.1970798731502437e-06}
{"train_loss": 0.023194463923573494, "global_step": 240441, "epoch": 2701, "lr": 3.1968758990663262e-06}
{"train_loss": 0.052019696682691574, "global_step": 240442, "epoch": 2701, "lr": 3.1966719312745087e-06}
{"train_loss": 0.02899211458861828, "global_step": 240443, "epoch": 2701, "lr": 3.19646796977483e-06}
{"train_loss": 0.05312465503811836, "global_step": 240444, "epoch": 2701, "lr": 3.196264014567324e-06}
{"train_loss": 0.024111753329634666, "global_step": 240445, "epoch": 2701, "lr": 3.196060065652001e-06}
{"train_loss": 0.015177328139543533, "global_step": 240446, "epoch": 2701, "lr": 3.1958561230289062e-06}
{"train_loss": 0.07722799479961395, "global_step": 240447, "epoch": 2701, "lr": 3.1956521866980506e-06}
{"train_loss": 0.059009719640016556, "global_step": 240448, "epoch": 2701, "lr": 3.1954482566594724e-06}
{"train_loss": 0.037071507424116135, "global_step": 240449, "epoch": 2701, "lr": 3.1952443329132e-06}
{"train_loss": 0.03532106429338455, "global_step": 240450, "epoch": 2701, "lr": 3.1950404154592495e-06}
{"train_loss": 0.02375713735818863, "global_step": 240451, "epoch": 2701, "lr": 3.1948365042976603e-06}
{"train_loss": 0.046444326639175415, "global_step": 240452, "epoch": 2701, "lr": 3.194632599428449e-06}
{"train_loss": 0.032410502433776855, "global_step": 240453, "epoch": 2701, "lr": 3.1944287008516592e-06}
{"train_loss": 0.05241112411022186, "global_step": 240454, "epoch": 2701, "lr": 3.1942248085673033e-06}
{"train_loss": 0.01984710805118084, "global_step": 240455, "epoch": 2701, "lr": 3.1940209225754082e-06}
{"train_loss": 0.027282465249300003, "global_step": 240456, "epoch": 2701, "lr": 3.193817042876013e-06}
{"train_loss": 0.035349372774362564, "global_step": 240457, "epoch": 2701, "lr": 3.1936131694691284e-06}
{"train_loss": 0.0244266539812088, "global_step": 240458, "epoch": 2701, "lr": 3.193409302354805e-06}
{"train_loss": 0.03212348371744156, "global_step": 240459, "epoch": 2701, "lr": 3.1932054415330424e-06}
{"train_loss": 0.08213649690151215, "global_step": 240460, "epoch": 2701, "lr": 3.1930015870038965e-06}
{"train_loss": 0.024474795907735825, "global_step": 240461, "epoch": 2701, "lr": 3.1927977387673723e-06}
{"train_loss": 0.05943534895777702, "global_step": 240462, "epoch": 2701, "lr": 3.1925938968235037e-06}
{"train_loss": 0.025116045027971268, "global_step": 240463, "epoch": 2701, "lr": 3.1923900611723343e-06}
{"train_loss": 0.05479823797941208, "global_step": 240464, "epoch": 2701, "lr": 3.192186231813865e-06}
{"train_loss": 0.029802631586790085, "global_step": 240465, "epoch": 2701, "lr": 3.19198240874814e-06}
{"train_loss": 0.019241362810134888, "global_step": 240466, "epoch": 2701, "lr": 3.191778591975181e-06}
{"train_loss": 0.03223470225930214, "global_step": 240467, "epoch": 2701, "lr": 3.1915747814950213e-06}
{"train_loss": 0.015485036186873913, "global_step": 240468, "epoch": 2701, "lr": 3.1913709773076785e-06}
{"train_loss": 0.057715293020009995, "global_step": 240469, "epoch": 2701, "lr": 3.1911671794131904e-06}
{"train_loss": 0.04435162618756294, "global_step": 240470, "epoch": 2701, "lr": 3.190963387811574e-06}
{"train_loss": 0.046546537429094315, "global_step": 240471, "epoch": 2701, "lr": 3.1907596025028684e-06}
{"train_loss": 0.07748236507177353, "global_step": 240472, "epoch": 2701, "lr": 3.1905558234870845e-06}
{"train_loss": 0.0740167647600174, "global_step": 240473, "epoch": 2701, "lr": 3.1903520507642726e-06}
{"train_loss": 0.06408636271953583, "global_step": 240474, "epoch": 2701, "lr": 3.190148284334438e-06}
{"train_loss": 0.05753461271524429, "global_step": 240475, "epoch": 2701, "lr": 3.1899445241976245e-06}
{"train_loss": 0.02192033641040325, "global_step": 240476, "epoch": 2701, "lr": 3.189740770353844e-06}
{"train_loss": 0.04311336503688539, "global_step": 240477, "epoch": 2701, "lr": 3.1895370228031405e-06, "val_loss": 9.332623481750488}
{"train_loss": 0.0324753075838089, "global_step": 240478, "epoch": 2702, "lr": 3.1893332815455257e-06}
{"train_loss": 0.05680049955844879, "global_step": 240479, "epoch": 2702, "lr": 3.1891295465810323e-06}
{"train_loss": 0.050225503742694855, "global_step": 240480, "epoch": 2702, "lr": 3.1889258179096994e-06}
{"train_loss": 0.0521114282310009, "global_step": 240481, "epoch": 2702, "lr": 3.1887220955315376e-06}
{"train_loss": 0.042346544563770294, "global_step": 240482, "epoch": 2702, "lr": 3.1885183794465866e-06}
{"train_loss": 0.017507391050457954, "global_step": 240483, "epoch": 2702, "lr": 3.188314669654868e-06}
{"train_loss": 0.02450679801404476, "global_step": 240484, "epoch": 2702, "lr": 3.1881109661564103e-06}
{"train_loss": 0.03218255192041397, "global_step": 240485, "epoch": 2702, "lr": 3.1879072689512346e-06}
{"train_loss": 0.04654306918382645, "global_step": 240486, "epoch": 2702, "lr": 3.187703578039386e-06}
{"train_loss": 0.017931777983903885, "global_step": 240487, "epoch": 2702, "lr": 3.18749989342087e-06}
{"train_loss": 0.052590642124414444, "global_step": 240488, "epoch": 2702, "lr": 3.1872962150957307e-06}
{"train_loss": 0.024036860093474388, "global_step": 240489, "epoch": 2702, "lr": 3.187092543063991e-06}
{"train_loss": 0.039222441613674164, "global_step": 240490, "epoch": 2702, "lr": 3.1868888773256665e-06}
{"train_loss": 0.03546050935983658, "global_step": 240491, "epoch": 2702, "lr": 3.186685217880803e-06}
{"train_loss": 0.023575980216264725, "global_step": 240492, "epoch": 2702, "lr": 3.18648156472941e-06}
{"train_loss": 0.026873521506786346, "global_step": 240493, "epoch": 2702, "lr": 3.1862779178715276e-06}
{"train_loss": 0.01790875568985939, "global_step": 240494, "epoch": 2702, "lr": 3.1860742773071884e-06}
{"train_loss": 0.016640828922390938, "global_step": 240495, "epoch": 2702, "lr": 3.1858706430363985e-06}
{"train_loss": 0.054395727813243866, "global_step": 240496, "epoch": 2702, "lr": 3.1856670150592073e-06}
{"train_loss": 0.046619612723588943, "global_step": 240497, "epoch": 2702, "lr": 3.1854633933756205e-06}
{"train_loss": 0.04997136816382408, "global_step": 240498, "epoch": 2702, "lr": 3.185259777985683e-06}
{"train_loss": 0.028824711218476295, "global_step": 240499, "epoch": 2702, "lr": 3.185056168889422e-06}
{"train_loss": 0.03451057896018028, "global_step": 240500, "epoch": 2702, "lr": 3.1848525660868487e-06}
{"train_loss": 0.06561409682035446, "global_step": 240501, "epoch": 2702, "lr": 3.1846489695780136e-06}
{"train_loss": 0.022555148229002953, "global_step": 240502, "epoch": 2702, "lr": 3.1844453793629214e-06}
{"train_loss": 0.027528339996933937, "global_step": 240503, "epoch": 2702, "lr": 3.1842417954416227e-06}
{"train_loss": 0.04512482509016991, "global_step": 240504, "epoch": 2702, "lr": 3.184038217814117e-06}
{"train_loss": 0.02266126126050949, "global_step": 240505, "epoch": 2702, "lr": 3.1838346464804604e-06}
{"train_loss": 0.019853245466947556, "global_step": 240506, "epoch": 2702, "lr": 3.1836310814406578e-06}
{"train_loss": 0.04023315757513046, "global_step": 240507, "epoch": 2702, "lr": 3.1834275226947487e-06}
{"train_loss": 0.041207604110240936, "global_step": 240508, "epoch": 2702, "lr": 3.1832239702427548e-06}
{"train_loss": 0.07247477769851685, "global_step": 240509, "epoch": 2702, "lr": 3.1830204240847095e-06}
{"train_loss": 0.10395042598247528, "global_step": 240510, "epoch": 2702, "lr": 3.1828168842206297e-06}
{"train_loss": 0.05852436274290085, "global_step": 240511, "epoch": 2702, "lr": 3.1826133506505596e-06}
{"train_loss": 0.01824103854596615, "global_step": 240512, "epoch": 2702, "lr": 3.1824098233745105e-06}
{"train_loss": 0.040672458708286285, "global_step": 240513, "epoch": 2702, "lr": 3.182206302392521e-06}
{"train_loss": 0.05237385630607605, "global_step": 240514, "epoch": 2702, "lr": 3.182002787704602e-06}
{"train_loss": 0.03502536937594414, "global_step": 240515, "epoch": 2702, "lr": 3.181799279310793e-06}
{"train_loss": 0.015329156070947647, "global_step": 240516, "epoch": 2702, "lr": 3.1815957772111327e-06}
{"train_loss": 0.04744330421090126, "global_step": 240517, "epoch": 2702, "lr": 3.1813922814056263e-06}
{"train_loss": 0.049919839948415756, "global_step": 240518, "epoch": 2702, "lr": 3.1811887918943184e-06}
{"train_loss": 0.05746688321232796, "global_step": 240519, "epoch": 2702, "lr": 3.18098530867722e-06}
{"train_loss": 0.011646303348243237, "global_step": 240520, "epoch": 2702, "lr": 3.1807818317543815e-06}
{"train_loss": 0.03152395039796829, "global_step": 240521, "epoch": 2702, "lr": 3.180578361125802e-06}
{"train_loss": 0.05060342326760292, "global_step": 240522, "epoch": 2702, "lr": 3.1803748967915325e-06}
{"train_loss": 0.03147648274898529, "global_step": 240523, "epoch": 2702, "lr": 3.1801714387515833e-06}
{"train_loss": 0.028852850198745728, "global_step": 240524, "epoch": 2702, "lr": 3.179967987005994e-06}
{"train_loss": 0.07872503250837326, "global_step": 240525, "epoch": 2702, "lr": 3.1797645415547863e-06}
{"train_loss": 0.03729032352566719, "global_step": 240526, "epoch": 2702, "lr": 3.1795611023979934e-06}
{"train_loss": 0.0655025988817215, "global_step": 240527, "epoch": 2702, "lr": 3.1793576695356375e-06}
{"train_loss": 0.05216013640165329, "global_step": 240528, "epoch": 2702, "lr": 3.179154242967736e-06}
{"train_loss": 0.014991060830652714, "global_step": 240529, "epoch": 2702, "lr": 3.1789508226943376e-06}
{"train_loss": 0.07463514804840088, "global_step": 240530, "epoch": 2702, "lr": 3.178747408715449e-06}
{"train_loss": 0.0664060041308403, "global_step": 240531, "epoch": 2702, "lr": 3.178544001031114e-06}
{"train_loss": 0.049380190670490265, "global_step": 240532, "epoch": 2702, "lr": 3.1783405996413553e-06}
{"train_loss": 0.03214716911315918, "global_step": 240533, "epoch": 2702, "lr": 3.178137204546189e-06}
{"train_loss": 0.03607282415032387, "global_step": 240534, "epoch": 2702, "lr": 3.177933815745654e-06}
{"train_loss": 0.03882206976413727, "global_step": 240535, "epoch": 2702, "lr": 3.1777304332397783e-06}
{"train_loss": 0.05057112127542496, "global_step": 240536, "epoch": 2702, "lr": 3.1775270570285843e-06}
{"train_loss": 0.05364061892032623, "global_step": 240537, "epoch": 2702, "lr": 3.177323687112105e-06}
{"train_loss": 0.042121194303035736, "global_step": 240538, "epoch": 2702, "lr": 3.177120323490357e-06}
{"train_loss": 0.052717749029397964, "global_step": 240539, "epoch": 2702, "lr": 3.176916966163379e-06}
{"train_loss": 0.05520489066839218, "global_step": 240540, "epoch": 2702, "lr": 3.1767136151311884e-06}
{"train_loss": 0.039258282631635666, "global_step": 240541, "epoch": 2702, "lr": 3.1765102703938286e-06}
{"train_loss": 0.05070273205637932, "global_step": 240542, "epoch": 2702, "lr": 3.1763069319513063e-06}
{"train_loss": 0.038492169231176376, "global_step": 240543, "epoch": 2702, "lr": 3.176103599803665e-06}
{"train_loss": 0.022334294393658638, "global_step": 240544, "epoch": 2702, "lr": 3.1759002739509215e-06}
{"train_loss": 0.036685761064291, "global_step": 240545, "epoch": 2702, "lr": 3.175696954393115e-06}
{"train_loss": 0.09157532453536987, "global_step": 240546, "epoch": 2702, "lr": 3.175493641130256e-06}
{"train_loss": 0.013609781861305237, "global_step": 240547, "epoch": 2702, "lr": 3.1752903341623896e-06}
{"train_loss": 0.06279439479112625, "global_step": 240548, "epoch": 2702, "lr": 3.1750870334895268e-06}
{"train_loss": 0.060912907123565674, "global_step": 240549, "epoch": 2702, "lr": 3.1748837391117057e-06}
{"train_loss": 0.043479692190885544, "global_step": 240550, "epoch": 2702, "lr": 3.1746804510289497e-06}
{"train_loss": 0.024620642885565758, "global_step": 240551, "epoch": 2702, "lr": 3.1744771692412912e-06}
{"train_loss": 0.015172314830124378, "global_step": 240552, "epoch": 2702, "lr": 3.174273893748747e-06}
{"train_loss": 0.0925653725862503, "global_step": 240553, "epoch": 2702, "lr": 3.1740706245513507e-06}
{"train_loss": 0.047594837844371796, "global_step": 240554, "epoch": 2702, "lr": 3.173867361649141e-06}
{"train_loss": 0.021299494430422783, "global_step": 240555, "epoch": 2702, "lr": 3.1736641050421233e-06}
{"train_loss": 0.14535054564476013, "global_step": 240556, "epoch": 2702, "lr": 3.1734608547303426e-06}
{"train_loss": 0.04635288193821907, "global_step": 240557, "epoch": 2702, "lr": 3.1732576107138145e-06}
{"train_loss": 0.07010184228420258, "global_step": 240558, "epoch": 2702, "lr": 3.1730543729925734e-06}
{"train_loss": 0.03796689584851265, "global_step": 240559, "epoch": 2702, "lr": 3.1728511415666407e-06}
{"train_loss": 0.05008291080594063, "global_step": 240560, "epoch": 2702, "lr": 3.1726479164360556e-06}
{"train_loss": 0.08627009391784668, "global_step": 240561, "epoch": 2702, "lr": 3.1724446976008294e-06}
{"train_loss": 0.06609716266393661, "global_step": 240562, "epoch": 2702, "lr": 3.1722414850610007e-06}
{"train_loss": 0.026534922420978546, "global_step": 240563, "epoch": 2702, "lr": 3.1720382788165915e-06}
{"train_loss": 0.039533454924821854, "global_step": 240564, "epoch": 2702, "lr": 3.1718350788676356e-06}
{"train_loss": 0.013710230588912964, "global_step": 240565, "epoch": 2702, "lr": 3.17163188521416e-06}
{"train_loss": 0.043633933953438585, "global_step": 240566, "epoch": 2702, "lr": 3.1714286978561712e-06, "val_loss": 9.411521911621094}
{"train_loss": 0.02154054120182991, "global_step": 240567, "epoch": 2703, "lr": 3.17122551679373e-06}
{"train_loss": 0.0108652887865901, "global_step": 240568, "epoch": 2703, "lr": 3.171022342026836e-06}
{"train_loss": 0.014436598867177963, "global_step": 240569, "epoch": 2703, "lr": 3.1708191735555335e-06}
{"train_loss": 0.07043097913265228, "global_step": 240570, "epoch": 2703, "lr": 3.1706160113798345e-06}
{"train_loss": 0.034664563834667206, "global_step": 240571, "epoch": 2703, "lr": 3.170412855499788e-06}
{"train_loss": 0.07710244506597519, "global_step": 240572, "epoch": 2703, "lr": 3.1702097059153945e-06}
{"train_loss": 0.0386933907866478, "global_step": 240573, "epoch": 2703, "lr": 3.1700065626267095e-06}
{"train_loss": 0.112720787525177, "global_step": 240574, "epoch": 2703, "lr": 3.169803425633733e-06}
{"train_loss": 0.03047706000506878, "global_step": 240575, "epoch": 2703, "lr": 3.16960029493652e-06}
{"train_loss": 0.05665132775902748, "global_step": 240576, "epoch": 2703, "lr": 3.169397170535071e-06}
{"train_loss": 0.029943572357296944, "global_step": 240577, "epoch": 2703, "lr": 3.169194052429436e-06}
{"train_loss": 0.0439196452498436, "global_step": 240578, "epoch": 2703, "lr": 3.1689909406196205e-06}
{"train_loss": 0.057252585887908936, "global_step": 240579, "epoch": 2703, "lr": 3.1687878351056742e-06}
{"train_loss": 0.08624598383903503, "global_step": 240580, "epoch": 2703, "lr": 3.168584735887603e-06}
{"train_loss": 0.018650107085704803, "global_step": 240581, "epoch": 2703, "lr": 3.1683816429654568e-06}
{"train_loss": 0.029599860310554504, "global_step": 240582, "epoch": 2703, "lr": 3.1681785563392407e-06}
{"train_loss": 0.03963402286171913, "global_step": 240583, "epoch": 2703, "lr": 3.1679754760089997e-06}
{"train_loss": 0.0184997096657753, "global_step": 240584, "epoch": 2703, "lr": 3.16777240197475e-06}
{"train_loss": 0.03677472099661827, "global_step": 240585, "epoch": 2703, "lr": 3.1675693342365254e-06}
{"train_loss": 0.04854847490787506, "global_step": 240586, "epoch": 2703, "lr": 3.167366272794342e-06}
{"train_loss": 0.03808130696415901, "global_step": 240587, "epoch": 2703, "lr": 3.1671632176482447e-06}
{"train_loss": 0.04822448268532753, "global_step": 240588, "epoch": 2703, "lr": 3.166960168798244e-06}
{"train_loss": 0.01380947232246399, "global_step": 240589, "epoch": 2703, "lr": 3.166757126244374e-06}
{"train_loss": 0.06511866301298141, "global_step": 240590, "epoch": 2703, "lr": 3.166554089986673e-06}
{"train_loss": 0.03521765023469925, "global_step": 240591, "epoch": 2703, "lr": 3.1663510600251465e-06}
{"train_loss": 0.06209945306181908, "global_step": 240592, "epoch": 2703, "lr": 3.166148036359845e-06}
{"train_loss": 0.05012407526373863, "global_step": 240593, "epoch": 2703, "lr": 3.1659450189907737e-06}
{"train_loss": 0.04764481261372566, "global_step": 240594, "epoch": 2703, "lr": 3.1657420079179767e-06}
{"train_loss": 0.02022486738860607, "global_step": 240595, "epoch": 2703, "lr": 3.165539003141471e-06}
{"train_loss": 0.03590797260403633, "global_step": 240596, "epoch": 2703, "lr": 3.1653360046612957e-06}
{"train_loss": 0.03325650468468666, "global_step": 240597, "epoch": 2703, "lr": 3.165133012477456e-06}
{"train_loss": 0.019731413573026657, "global_step": 240598, "epoch": 2703, "lr": 3.1649300265900074e-06}
{"train_loss": 0.022300658747553825, "global_step": 240599, "epoch": 2703, "lr": 3.1647270469989555e-06}
{"train_loss": 0.057337138801813126, "global_step": 240600, "epoch": 2703, "lr": 3.164524073704339e-06}
{"train_loss": 0.03647085651755333, "global_step": 240601, "epoch": 2703, "lr": 3.1643211067061752e-06}
{"train_loss": 0.047390926629304886, "global_step": 240602, "epoch": 2703, "lr": 3.1641181460045078e-06}
{"train_loss": 0.11160384118556976, "global_step": 240603, "epoch": 2703, "lr": 3.1639151915993428e-06}
{"train_loss": 0.024908293038606644, "global_step": 240604, "epoch": 2703, "lr": 3.1637122434907242e-06}
{"train_loss": 0.02087142877280712, "global_step": 240605, "epoch": 2703, "lr": 3.1635093016786797e-06}
{"train_loss": 0.09603094309568405, "global_step": 240606, "epoch": 2703, "lr": 3.163306366163221e-06}
{"train_loss": 0.06426788121461868, "global_step": 240607, "epoch": 2703, "lr": 3.1631034369443867e-06}
{"train_loss": 0.03312760591506958, "global_step": 240608, "epoch": 2703, "lr": 3.162900514022199e-06}
{"train_loss": 0.01370153110474348, "global_step": 240609, "epoch": 2703, "lr": 3.1626975973966967e-06}
{"train_loss": 0.022280098870396614, "global_step": 240610, "epoch": 2703, "lr": 3.1624946870678906e-06}
{"train_loss": 0.04338029772043228, "global_step": 240611, "epoch": 2703, "lr": 3.162291783035826e-06}
{"train_loss": 0.05317968502640724, "global_step": 240612, "epoch": 2703, "lr": 3.1620888853005125e-06}
{"train_loss": 0.03851601481437683, "global_step": 240613, "epoch": 2703, "lr": 3.1618859938619904e-06}
{"train_loss": 0.06136523187160492, "global_step": 240614, "epoch": 2703, "lr": 3.161683108720276e-06}
{"train_loss": 0.041849277913570404, "global_step": 240615, "epoch": 2703, "lr": 3.1614802298754076e-06}
{"train_loss": 0.039154745638370514, "global_step": 240616, "epoch": 2703, "lr": 3.1612773573274024e-06}
{"train_loss": 0.02883673459291458, "global_step": 240617, "epoch": 2703, "lr": 3.161074491076299e-06}
{"train_loss": 0.01983405090868473, "global_step": 240618, "epoch": 2703, "lr": 3.160871631122109e-06}
{"train_loss": 0.029623163864016533, "global_step": 240619, "epoch": 2703, "lr": 3.160668777464881e-06}
{"train_loss": 0.06225007399916649, "global_step": 240620, "epoch": 2703, "lr": 3.160465930104617e-06}
{"train_loss": 0.03415938466787338, "global_step": 240621, "epoch": 2703, "lr": 3.160263089041371e-06}
{"train_loss": 0.053855814039707184, "global_step": 240622, "epoch": 2703, "lr": 3.1600602542751435e-06}
{"train_loss": 0.033795952796936035, "global_step": 240623, "epoch": 2703, "lr": 3.1598574258059845e-06}
{"train_loss": 0.008385365828871727, "global_step": 240624, "epoch": 2703, "lr": 3.159654603633905e-06}
{"train_loss": 0.0902184471487999, "global_step": 240625, "epoch": 2703, "lr": 3.159451787758938e-06}
{"train_loss": 0.04324601963162422, "global_step": 240626, "epoch": 2703, "lr": 3.159248978181123e-06}
{"train_loss": 0.025280339643359184, "global_step": 240627, "epoch": 2703, "lr": 3.1590461749004652e-06}
{"train_loss": 0.05749441683292389, "global_step": 240628, "epoch": 2703, "lr": 3.158843377917009e-06}
{"train_loss": 0.06550363451242447, "global_step": 240629, "epoch": 2703, "lr": 3.1586405872307712e-06}
{"train_loss": 0.09893587231636047, "global_step": 240630, "epoch": 2703, "lr": 3.1584378028417905e-06}
{"train_loss": 0.024617021903395653, "global_step": 240631, "epoch": 2703, "lr": 3.158235024750078e-06}
{"train_loss": 0.07860372960567474, "global_step": 240632, "epoch": 2703, "lr": 3.1580322529556782e-06}
{"train_loss": 0.060390979051589966, "global_step": 240633, "epoch": 2703, "lr": 3.1578294874586023e-06}
{"train_loss": 0.01935776136815548, "global_step": 240634, "epoch": 2703, "lr": 3.157626728258889e-06}
{"train_loss": 0.059103675186634064, "global_step": 240635, "epoch": 2703, "lr": 3.1574239753565605e-06}
{"train_loss": 0.06914074718952179, "global_step": 240636, "epoch": 2703, "lr": 3.1572212287516502e-06}
{"train_loss": 0.04649399593472481, "global_step": 240637, "epoch": 2703, "lr": 3.1570184884441745e-06}
{"train_loss": 0.09994665533304214, "global_step": 240638, "epoch": 2703, "lr": 3.1568157544341728e-06}
{"train_loss": 0.02169877663254738, "global_step": 240639, "epoch": 2703, "lr": 3.1566130267216556e-06}
{"train_loss": 0.044803302735090256, "global_step": 240640, "epoch": 2703, "lr": 3.156410305306673e-06}
{"train_loss": 0.04362796992063522, "global_step": 240641, "epoch": 2703, "lr": 3.1562075901892307e-06}
{"train_loss": 0.023654799908399582, "global_step": 240642, "epoch": 2703, "lr": 3.156004881369373e-06}
{"train_loss": 0.044622752815485, "global_step": 240643, "epoch": 2703, "lr": 3.1558021788471172e-06}
{"train_loss": 0.04249167814850807, "global_step": 240644, "epoch": 2703, "lr": 3.15559948262249e-06}
{"train_loss": 0.04324662312865257, "global_step": 240645, "epoch": 2703, "lr": 3.155396792695525e-06}
{"train_loss": 0.041616372764110565, "global_step": 240646, "epoch": 2703, "lr": 3.1551941090662397e-06}
{"train_loss": 0.03136678412556648, "global_step": 240647, "epoch": 2703, "lr": 3.1549914317346772e-06}
{"train_loss": 0.05894642323255539, "global_step": 240648, "epoch": 2703, "lr": 3.1547887607008496e-06}
{"train_loss": 0.0824730172753334, "global_step": 240649, "epoch": 2703, "lr": 3.15458609596479e-06}
{"train_loss": 0.08692512661218643, "global_step": 240650, "epoch": 2703, "lr": 3.15438343752652e-06}
{"train_loss": 0.03831927105784416, "global_step": 240651, "epoch": 2703, "lr": 3.154180785386085e-06}
{"train_loss": 0.029871588572859764, "global_step": 240652, "epoch": 2703, "lr": 3.1539781395434896e-06}
{"train_loss": 0.10856082290410995, "global_step": 240653, "epoch": 2703, "lr": 3.1537754999987733e-06}
{"train_loss": 0.03493036329746246, "global_step": 240654, "epoch": 2703, "lr": 3.1535728667519584e-06}
{"train_loss": 0.04584498362343633, "global_step": 240655, "epoch": 2703, "lr": 3.153370239803083e-06, "val_loss": 9.565488815307617}
{"train_loss": 0.04349008575081825, "global_step": 240656, "epoch": 2704, "lr": 3.153167619152153e-06}
{"train_loss": 0.06756911426782608, "global_step": 240657, "epoch": 2704, "lr": 3.1529650047992186e-06}
{"train_loss": 0.05938256159424782, "global_step": 240658, "epoch": 2704, "lr": 3.152762396744291e-06}
{"train_loss": 0.028935372829437256, "global_step": 240659, "epoch": 2704, "lr": 3.152559794987414e-06}
{"train_loss": 0.027050670236349106, "global_step": 240660, "epoch": 2704, "lr": 3.1523571995285883e-06}
{"train_loss": 0.03341040387749672, "global_step": 240661, "epoch": 2704, "lr": 3.1521546103678634e-06}
{"train_loss": 0.03969069942831993, "global_step": 240662, "epoch": 2704, "lr": 3.1519520275052672e-06}
{"train_loss": 0.056173499673604965, "global_step": 240663, "epoch": 2704, "lr": 3.1517494509408108e-06}
{"train_loss": 0.05392511561512947, "global_step": 240664, "epoch": 2704, "lr": 3.1515468806745385e-06}
{"train_loss": 0.054782092571258545, "global_step": 240665, "epoch": 2704, "lr": 3.151344316706467e-06}
{"train_loss": 0.053936269134283066, "global_step": 240666, "epoch": 2704, "lr": 3.1511417590366297e-06}
{"train_loss": 0.09496653825044632, "global_step": 240667, "epoch": 2704, "lr": 3.1509392076650435e-06}
{"train_loss": 0.04218249022960663, "global_step": 240668, "epoch": 2704, "lr": 3.1507366625917465e-06}
{"train_loss": 0.0351182259619236, "global_step": 240669, "epoch": 2704, "lr": 3.150534123816756e-06}
{"train_loss": 0.09870480746030807, "global_step": 240670, "epoch": 2704, "lr": 3.1503315913401164e-06}
{"train_loss": 0.03183000162243843, "global_step": 240671, "epoch": 2704, "lr": 3.1501290651618333e-06}
{"train_loss": 0.07719839364290237, "global_step": 240672, "epoch": 2704, "lr": 3.1499265452819503e-06}
{"train_loss": 0.04316059872508049, "global_step": 240673, "epoch": 2704, "lr": 3.149724031700485e-06}
{"train_loss": 0.013122431933879852, "global_step": 240674, "epoch": 2704, "lr": 3.1495215244174757e-06}
{"train_loss": 0.025809314101934433, "global_step": 240675, "epoch": 2704, "lr": 3.149319023432934e-06}
{"train_loss": 0.03155286982655525, "global_step": 240676, "epoch": 2704, "lr": 3.1491165287468983e-06}
{"train_loss": 0.056792303919792175, "global_step": 240677, "epoch": 2704, "lr": 3.1489140403593907e-06}
{"train_loss": 0.06401175260543823, "global_step": 240678, "epoch": 2704, "lr": 3.148711558270445e-06}
{"train_loss": 0.06118068844079971, "global_step": 240679, "epoch": 2704, "lr": 3.1485090824800777e-06}
{"train_loss": 0.010036400519311428, "global_step": 240680, "epoch": 2704, "lr": 3.148306612988333e-06}
{"train_loss": 0.03984799236059189, "global_step": 240681, "epoch": 2704, "lr": 3.1481041497952226e-06}
{"train_loss": 0.04030323028564453, "global_step": 240682, "epoch": 2704, "lr": 3.147901692900773e-06}
{"train_loss": 0.05292104184627533, "global_step": 240683, "epoch": 2704, "lr": 3.1476992423050246e-06}
{"train_loss": 0.04272361099720001, "global_step": 240684, "epoch": 2704, "lr": 3.147496798007987e-06}
{"train_loss": 0.05364164710044861, "global_step": 240685, "epoch": 2704, "lr": 3.147294360009706e-06}
{"train_loss": 0.06445176899433136, "global_step": 240686, "epoch": 2704, "lr": 3.1470919283101973e-06}
{"train_loss": 0.07067224383354187, "global_step": 240687, "epoch": 2704, "lr": 3.1468895029094946e-06}
{"train_loss": 0.04045730084180832, "global_step": 240688, "epoch": 2704, "lr": 3.1466870838076147e-06}
{"train_loss": 0.007461335510015488, "global_step": 240689, "epoch": 2704, "lr": 3.146484671004596e-06}
{"train_loss": 0.05430033802986145, "global_step": 240690, "epoch": 2704, "lr": 3.1462822645004553e-06}
{"train_loss": 0.10938400775194168, "global_step": 240691, "epoch": 2704, "lr": 3.146079864295237e-06}
{"train_loss": 0.021828671917319298, "global_step": 240692, "epoch": 2704, "lr": 3.145877470388947e-06}
{"train_loss": 0.01758149079978466, "global_step": 240693, "epoch": 2704, "lr": 3.1456750827816296e-06}
{"train_loss": 0.036844540387392044, "global_step": 240694, "epoch": 2704, "lr": 3.145472701473301e-06}
{"train_loss": 0.06377292424440384, "global_step": 240695, "epoch": 2704, "lr": 3.145270326463995e-06}
{"train_loss": 0.05601407214999199, "global_step": 240696, "epoch": 2704, "lr": 3.1450679577537334e-06}
{"train_loss": 0.03124598227441311, "global_step": 240697, "epoch": 2704, "lr": 3.1448655953425443e-06}
{"train_loss": 0.041409872472286224, "global_step": 240698, "epoch": 2704, "lr": 3.1446632392304665e-06}
{"train_loss": 0.016310948878526688, "global_step": 240699, "epoch": 2704, "lr": 3.1444608894175055e-06}
{"train_loss": 0.05521639436483383, "global_step": 240700, "epoch": 2704, "lr": 3.144258545903711e-06}
{"train_loss": 0.03939821571111679, "global_step": 240701, "epoch": 2704, "lr": 3.1440562086890944e-06}
{"train_loss": 0.062408242374658585, "global_step": 240702, "epoch": 2704, "lr": 3.1438538777736947e-06}
{"train_loss": 0.042845431715250015, "global_step": 240703, "epoch": 2704, "lr": 3.1436515531575227e-06}
{"train_loss": 0.05065956711769104, "global_step": 240704, "epoch": 2704, "lr": 3.143449234840623e-06}
{"train_loss": 0.04656760394573212, "global_step": 240705, "epoch": 2704, "lr": 3.143246922823012e-06}
{"train_loss": 0.054226119071245193, "global_step": 240706, "epoch": 2704, "lr": 3.1430446171047235e-06}
{"train_loss": 0.05836569517850876, "global_step": 240707, "epoch": 2704, "lr": 3.142842317685779e-06}
{"train_loss": 0.03395304083824158, "global_step": 240708, "epoch": 2704, "lr": 3.1426400245662124e-06}
{"train_loss": 0.0682506263256073, "global_step": 240709, "epoch": 2704, "lr": 3.14243773774604e-06}
{"train_loss": 0.07480785250663757, "global_step": 240710, "epoch": 2704, "lr": 3.142235457225301e-06}
{"train_loss": 0.03186454996466637, "global_step": 240711, "epoch": 2704, "lr": 3.142033183004012e-06}
{"train_loss": 0.07858218252658844, "global_step": 240712, "epoch": 2704, "lr": 3.141830915082211e-06}
{"train_loss": 0.0688677653670311, "global_step": 240713, "epoch": 2704, "lr": 3.141628653459916e-06}
{"train_loss": 0.04950612038373947, "global_step": 240714, "epoch": 2704, "lr": 3.141426398137165e-06}
{"train_loss": 0.05162116140127182, "global_step": 240715, "epoch": 2704, "lr": 3.1412241491139695e-06}
{"train_loss": 0.07156646251678467, "global_step": 240716, "epoch": 2704, "lr": 3.141021906390368e-06}
{"train_loss": 0.022097257897257805, "global_step": 240717, "epoch": 2704, "lr": 3.1408196699663826e-06}
{"train_loss": 0.029659714549779892, "global_step": 240718, "epoch": 2704, "lr": 3.140617439842053e-06}
{"train_loss": 0.06553330272436142, "global_step": 240719, "epoch": 2704, "lr": 3.140415216017384e-06}
{"train_loss": 0.05494753271341324, "global_step": 240720, "epoch": 2704, "lr": 3.14021299849242e-06}
{"train_loss": 0.04536891356110573, "global_step": 240721, "epoch": 2704, "lr": 3.140010787267189e-06}
{"train_loss": 0.05020565167069435, "global_step": 240722, "epoch": 2704, "lr": 3.1398085823417024e-06}
{"train_loss": 0.03613976016640663, "global_step": 240723, "epoch": 2704, "lr": 3.139606383716004e-06}
{"train_loss": 0.025968460366129875, "global_step": 240724, "epoch": 2704, "lr": 3.1394041913901108e-06}
{"train_loss": 0.06887831538915634, "global_step": 240725, "epoch": 2704, "lr": 3.139202005364056e-06}
{"train_loss": 0.034644484519958496, "global_step": 240726, "epoch": 2704, "lr": 3.1389998256378616e-06}
{"train_loss": 0.044732462614774704, "global_step": 240727, "epoch": 2704, "lr": 3.1387976522115616e-06}
{"train_loss": 0.032025907188653946, "global_step": 240728, "epoch": 2704, "lr": 3.1385954850851716e-06}
{"train_loss": 0.036650124937295914, "global_step": 240729, "epoch": 2704, "lr": 3.138393324258737e-06}
{"train_loss": 0.0472230464220047, "global_step": 240730, "epoch": 2704, "lr": 3.138191169732263e-06}
{"train_loss": 0.05438980460166931, "global_step": 240731, "epoch": 2704, "lr": 3.1379890215057995e-06}
{"train_loss": 0.057225462049245834, "global_step": 240732, "epoch": 2704, "lr": 3.137786879579352e-06}
{"train_loss": 0.0465531162917614, "global_step": 240733, "epoch": 2704, "lr": 3.137584743952959e-06}
{"train_loss": 0.0510135143995285, "global_step": 240734, "epoch": 2704, "lr": 3.137382614626655e-06}
{"train_loss": 0.044260770082473755, "global_step": 240735, "epoch": 2704, "lr": 3.137180491600444e-06}
{"train_loss": 0.031005242839455605, "global_step": 240736, "epoch": 2704, "lr": 3.136978374874383e-06}
{"train_loss": 0.027465345337986946, "global_step": 240737, "epoch": 2704, "lr": 3.1367762644484765e-06}
{"train_loss": 0.023364227265119553, "global_step": 240738, "epoch": 2704, "lr": 3.136574160322764e-06}
{"train_loss": 0.015420036390423775, "global_step": 240739, "epoch": 2704, "lr": 3.1363720624972615e-06}
{"train_loss": 0.03392864018678665, "global_step": 240740, "epoch": 2704, "lr": 3.1361699709720083e-06}
{"train_loss": 0.0465535931289196, "global_step": 240741, "epoch": 2704, "lr": 3.135967885747021e-06}
{"train_loss": 0.03771368786692619, "global_step": 240742, "epoch": 2704, "lr": 3.1357658068223385e-06}
{"train_loss": 0.038823794573545456, "global_step": 240743, "epoch": 2704, "lr": 3.1355637341979716e-06}
{"train_loss": 0.047261915630085416, "global_step": 240744, "epoch": 2704, "lr": 3.135361667873965e-06, "val_loss": 9.564651489257812}
{"train_loss": 0.0319814570248127, "global_step": 240745, "epoch": 2705, "lr": 3.1351596078503353e-06}
{"train_loss": 0.06304266303777695, "global_step": 240746, "epoch": 2705, "lr": 3.134957554127116e-06}
{"train_loss": 0.04841563105583191, "global_step": 240747, "epoch": 2705, "lr": 3.134755506704318e-06}
{"train_loss": 0.061807047575712204, "global_step": 240748, "epoch": 2705, "lr": 3.1345534655819964e-06}
{"train_loss": 0.049711182713508606, "global_step": 240749, "epoch": 2705, "lr": 3.1343514307601517e-06}
{"train_loss": 0.01562829129397869, "global_step": 240750, "epoch": 2705, "lr": 3.1341494022388286e-06}
{"train_loss": 0.09950271248817444, "global_step": 240751, "epoch": 2705, "lr": 3.133947380018043e-06}
{"train_loss": 0.06420499086380005, "global_step": 240752, "epoch": 2705, "lr": 3.133745364097823e-06}
{"train_loss": 0.041291143745183945, "global_step": 240753, "epoch": 2705, "lr": 3.1335433544782133e-06}
{"train_loss": 0.03968580812215805, "global_step": 240754, "epoch": 2705, "lr": 3.1333413511592247e-06}
{"train_loss": 0.09014628827571869, "global_step": 240755, "epoch": 2705, "lr": 3.1331393541408793e-06}
{"train_loss": 0.060959894210100174, "global_step": 240756, "epoch": 2705, "lr": 3.132937363423216e-06}
{"train_loss": 0.05329796299338341, "global_step": 240757, "epoch": 2705, "lr": 3.1327353790062577e-06}
{"train_loss": 0.06693509221076965, "global_step": 240758, "epoch": 2705, "lr": 3.132533400890031e-06}
{"train_loss": 0.035951029509305954, "global_step": 240759, "epoch": 2705, "lr": 3.13233142907457e-06}
{"train_loss": 0.03320880979299545, "global_step": 240760, "epoch": 2705, "lr": 3.1321294635598853e-06}
{"train_loss": 0.02570478431880474, "global_step": 240761, "epoch": 2705, "lr": 3.1319275043460272e-06}
{"train_loss": 0.02058669924736023, "global_step": 240762, "epoch": 2705, "lr": 3.1317255514329957e-06}
{"train_loss": 0.08093702793121338, "global_step": 240763, "epoch": 2705, "lr": 3.131523604820846e-06}
{"train_loss": 0.03874044120311737, "global_step": 240764, "epoch": 2705, "lr": 3.1313216645095846e-06}
{"train_loss": 0.0417187474668026, "global_step": 240765, "epoch": 2705, "lr": 3.131119730499249e-06}
{"train_loss": 0.059452228248119354, "global_step": 240766, "epoch": 2705, "lr": 3.130917802789857e-06}
{"train_loss": 0.04955023154616356, "global_step": 240767, "epoch": 2705, "lr": 3.1307158813814463e-06}
{"train_loss": 0.06481137126684189, "global_step": 240768, "epoch": 2705, "lr": 3.1305139662740346e-06}
{"train_loss": 0.06400705128908157, "global_step": 240769, "epoch": 2705, "lr": 3.130312057467655e-06}
{"train_loss": 0.06533604860305786, "global_step": 240770, "epoch": 2705, "lr": 3.13011015496234e-06}
{"train_loss": 0.0329962857067585, "global_step": 240771, "epoch": 2705, "lr": 3.1299082587581076e-06}
{"train_loss": 0.030648112297058105, "global_step": 240772, "epoch": 2705, "lr": 3.12970636885499e-06}
{"train_loss": 0.022716311737895012, "global_step": 240773, "epoch": 2705, "lr": 3.129504485253004e-06}
{"train_loss": 0.04144164174795151, "global_step": 240774, "epoch": 2705, "lr": 3.1293026079521947e-06}
{"train_loss": 0.02699667029082775, "global_step": 240775, "epoch": 2705, "lr": 3.1291007369525728e-06}
{"train_loss": 0.051088765263557434, "global_step": 240776, "epoch": 2705, "lr": 3.1288988722541768e-06}
{"train_loss": 0.012164543382823467, "global_step": 240777, "epoch": 2705, "lr": 3.1286970138570236e-06}
{"train_loss": 0.05280432105064392, "global_step": 240778, "epoch": 2705, "lr": 3.1284951617611526e-06}
{"train_loss": 0.06356025487184525, "global_step": 240779, "epoch": 2705, "lr": 3.128293315966574e-06}
{"train_loss": 0.06647998839616776, "global_step": 240780, "epoch": 2705, "lr": 3.1280914764733383e-06}
{"train_loss": 0.04841603711247444, "global_step": 240781, "epoch": 2705, "lr": 3.1278896432814453e-06}
{"train_loss": 0.05029718205332756, "global_step": 240782, "epoch": 2705, "lr": 3.1276878163909506e-06}
{"train_loss": 0.07831177115440369, "global_step": 240783, "epoch": 2705, "lr": 3.127485995801854e-06}
{"train_loss": 0.052717410027980804, "global_step": 240784, "epoch": 2705, "lr": 3.1272841815142064e-06}
{"train_loss": 0.0632583275437355, "global_step": 240785, "epoch": 2705, "lr": 3.1270823735280175e-06}
{"train_loss": 0.0504472516477108, "global_step": 240786, "epoch": 2705, "lr": 3.1268805718433268e-06}
{"train_loss": 0.05752299353480339, "global_step": 240787, "epoch": 2705, "lr": 3.1266787764601457e-06}
{"train_loss": 0.0393323190510273, "global_step": 240788, "epoch": 2705, "lr": 3.126476987378513e-06}
{"train_loss": 0.12259022146463394, "global_step": 240789, "epoch": 2705, "lr": 3.1262752045984613e-06}
{"train_loss": 0.029272107407450676, "global_step": 240790, "epoch": 2705, "lr": 3.126073428120008e-06}
{"train_loss": 0.046431805938482285, "global_step": 240791, "epoch": 2705, "lr": 3.1258716579431856e-06}
{"train_loss": 0.06650032848119736, "global_step": 240792, "epoch": 2705, "lr": 3.1256698940680174e-06}
{"train_loss": 0.025688467547297478, "global_step": 240793, "epoch": 2705, "lr": 3.125468136494525e-06}
{"train_loss": 0.03274822235107422, "global_step": 240794, "epoch": 2705, "lr": 3.1252663852227526e-06}
{"train_loss": 0.04110930487513542, "global_step": 240795, "epoch": 2705, "lr": 3.1250646402527064e-06}
{"train_loss": 0.04460381716489792, "global_step": 240796, "epoch": 2705, "lr": 3.12486290158443e-06}
{"train_loss": 0.10061537474393845, "global_step": 240797, "epoch": 2705, "lr": 3.1246611692179462e-06}
{"train_loss": 0.026710422709584236, "global_step": 240798, "epoch": 2705, "lr": 3.124459443153277e-06}
{"train_loss": 0.009721433743834496, "global_step": 240799, "epoch": 2705, "lr": 3.124257723390456e-06}
{"train_loss": 0.04330216720700264, "global_step": 240800, "epoch": 2705, "lr": 3.124056009929499e-06}
{"train_loss": 0.028321746736764908, "global_step": 240801, "epoch": 2705, "lr": 3.1238543027704513e-06}
{"train_loss": 0.05113498121500015, "global_step": 240802, "epoch": 2705, "lr": 3.123652601913324e-06}
{"train_loss": 0.03017931804060936, "global_step": 240803, "epoch": 2705, "lr": 3.1234509073581553e-06}
{"train_loss": 0.03883913531899452, "global_step": 240804, "epoch": 2705, "lr": 3.1232492191049566e-06}
{"train_loss": 0.06785169988870621, "global_step": 240805, "epoch": 2705, "lr": 3.1230475371537784e-06}
{"train_loss": 0.045816440135240555, "global_step": 240806, "epoch": 2705, "lr": 3.122845861504625e-06}
{"train_loss": 0.057999275624752045, "global_step": 240807, "epoch": 2705, "lr": 3.122644192157531e-06}
{"train_loss": 0.04222278669476509, "global_step": 240808, "epoch": 2705, "lr": 3.12244252911254e-06}
{"train_loss": 0.036990486085414886, "global_step": 240809, "epoch": 2705, "lr": 3.1222408723696528e-06}
{"train_loss": 0.05275915563106537, "global_step": 240810, "epoch": 2705, "lr": 3.1220392219289184e-06}
{"train_loss": 0.01670406199991703, "global_step": 240811, "epoch": 2705, "lr": 3.1218375777903484e-06}
{"train_loss": 0.03235414996743202, "global_step": 240812, "epoch": 2705, "lr": 3.121635939953982e-06}
{"train_loss": 0.056772630661726, "global_step": 240813, "epoch": 2705, "lr": 3.1214343084198296e-06}
{"train_loss": 0.052527159452438354, "global_step": 240814, "epoch": 2705, "lr": 3.1212326831879413e-06}
{"train_loss": 0.05341557040810585, "global_step": 240815, "epoch": 2705, "lr": 3.121031064258323e-06}
{"train_loss": 0.018719801679253578, "global_step": 240816, "epoch": 2705, "lr": 3.120829451631019e-06}
{"train_loss": 0.022522198036313057, "global_step": 240817, "epoch": 2705, "lr": 3.12062784530604e-06}
{"train_loss": 0.010712038725614548, "global_step": 240818, "epoch": 2705, "lr": 3.1204262452834253e-06}
{"train_loss": 0.05197940021753311, "global_step": 240819, "epoch": 2705, "lr": 3.1202246515631917e-06}
{"train_loss": 0.008319897577166557, "global_step": 240820, "epoch": 2705, "lr": 3.120023064145383e-06}
{"train_loss": 0.03535046800971031, "global_step": 240821, "epoch": 2705, "lr": 3.1198214830300056e-06}
{"train_loss": 0.02427932620048523, "global_step": 240822, "epoch": 2705, "lr": 3.1196199082171085e-06}
{"train_loss": 0.0729055404663086, "global_step": 240823, "epoch": 2705, "lr": 3.1194183397066977e-06}
{"train_loss": 0.02482611872255802, "global_step": 240824, "epoch": 2705, "lr": 3.119216777498807e-06}
{"train_loss": 0.03492353484034538, "global_step": 240825, "epoch": 2705, "lr": 3.1190152215934742e-06}
{"train_loss": 0.02084244228899479, "global_step": 240826, "epoch": 2705, "lr": 3.118813671990711e-06}
{"train_loss": 0.0506080687046051, "global_step": 240827, "epoch": 2705, "lr": 3.1186121286905623e-06}
{"train_loss": 0.01603636145591736, "global_step": 240828, "epoch": 2705, "lr": 3.1184105916930386e-06}
{"train_loss": 0.05007975921034813, "global_step": 240829, "epoch": 2705, "lr": 3.1182090609981727e-06}
{"train_loss": 0.04653191566467285, "global_step": 240830, "epoch": 2705, "lr": 3.1180075366059935e-06}
{"train_loss": 0.05906756967306137, "global_step": 240831, "epoch": 2705, "lr": 3.117806018516528e-06}
{"train_loss": 0.03410065919160843, "global_step": 240832, "epoch": 2705, "lr": 3.1176045067298045e-06}
{"train_loss": 0.04577885119223528, "global_step": 240833, "epoch": 2705, "lr": 3.117403001245839e-06, "val_loss": 9.453033447265625, "train_action_mse_error": 1.542856216430664}
{"train_loss": 0.06710328906774521, "global_step": 240834, "epoch": 2706, "lr": 3.117201502064676e-06}
{"train_loss": 0.029665058478713036, "global_step": 240835, "epoch": 2706, "lr": 3.1170000091863272e-06}
{"train_loss": 0.05863746628165245, "global_step": 240836, "epoch": 2706, "lr": 3.1167985226108364e-06}
{"train_loss": 0.06731905788183212, "global_step": 240837, "epoch": 2706, "lr": 3.116597042338215e-06}
{"train_loss": 0.034715183079242706, "global_step": 240838, "epoch": 2706, "lr": 3.116395568368491e-06}
{"train_loss": 0.0932062566280365, "global_step": 240839, "epoch": 2706, "lr": 3.1161941007017025e-06}
{"train_loss": 0.04675199091434479, "global_step": 240840, "epoch": 2706, "lr": 3.1159926393378615e-06}
{"train_loss": 0.02598973549902439, "global_step": 240841, "epoch": 2706, "lr": 3.1157911842770116e-06}
{"train_loss": 0.08581122010946274, "global_step": 240842, "epoch": 2706, "lr": 3.11558973551917e-06}
{"train_loss": 0.05328840762376785, "global_step": 240843, "epoch": 2706, "lr": 3.1153882930643585e-06}
{"train_loss": 0.04624855890870094, "global_step": 240844, "epoch": 2706, "lr": 3.115186856912622e-06}
{"train_loss": 0.04127946123480797, "global_step": 240845, "epoch": 2706, "lr": 3.1149854270639713e-06}
{"train_loss": 0.07555685937404633, "global_step": 240846, "epoch": 2706, "lr": 3.114784003518445e-06}
{"train_loss": 0.036326441913843155, "global_step": 240847, "epoch": 2706, "lr": 3.1145825862760603e-06}
{"train_loss": 0.07219427078962326, "global_step": 240848, "epoch": 2706, "lr": 3.114381175336856e-06}
{"train_loss": 0.056738611310720444, "global_step": 240849, "epoch": 2706, "lr": 3.114179770700837e-06}
{"train_loss": 0.01878443919122219, "global_step": 240850, "epoch": 2706, "lr": 3.113978372368059e-06}
{"train_loss": 0.058440230786800385, "global_step": 240851, "epoch": 2706, "lr": 3.1137769803385287e-06}
{"train_loss": 0.04287073761224747, "global_step": 240852, "epoch": 2706, "lr": 3.1135755946122836e-06}
{"train_loss": 0.04611120745539665, "global_step": 240853, "epoch": 2706, "lr": 3.113374215189341e-06}
{"train_loss": 0.04623878002166748, "global_step": 240854, "epoch": 2706, "lr": 3.1131728420697392e-06}
{"train_loss": 0.029174629598855972, "global_step": 240855, "epoch": 2706, "lr": 3.11297147525349e-06}
{"train_loss": 0.04578197002410889, "global_step": 240856, "epoch": 2706, "lr": 3.112770114740643e-06}
{"train_loss": 0.01303982362151146, "global_step": 240857, "epoch": 2706, "lr": 3.112568760531204e-06}
{"train_loss": 0.021167553961277008, "global_step": 240858, "epoch": 2706, "lr": 3.1123674126252166e-06}
{"train_loss": 0.03805357217788696, "global_step": 240859, "epoch": 2706, "lr": 3.1121660710226872e-06}
{"train_loss": 0.09537822753190994, "global_step": 240860, "epoch": 2706, "lr": 3.11196473572366e-06}
{"train_loss": 0.02882426045835018, "global_step": 240861, "epoch": 2706, "lr": 3.111763406728169e-06}
{"train_loss": 0.06022340804338455, "global_step": 240862, "epoch": 2706, "lr": 3.111562084036218e-06}
{"train_loss": 0.0646311342716217, "global_step": 240863, "epoch": 2706, "lr": 3.111360767647853e-06}
{"train_loss": 0.08120747655630112, "global_step": 240864, "epoch": 2706, "lr": 3.1111594575630844e-06}
{"train_loss": 0.028076112270355225, "global_step": 240865, "epoch": 2706, "lr": 3.1109581537819623e-06}
{"train_loss": 0.04296952113509178, "global_step": 240866, "epoch": 2706, "lr": 3.110756856304492e-06}
{"train_loss": 0.0292595736682415, "global_step": 240867, "epoch": 2706, "lr": 3.110555565130713e-06}
{"train_loss": 0.06638195365667343, "global_step": 240868, "epoch": 2706, "lr": 3.110354280260641e-06}
{"train_loss": 0.011023067869246006, "global_step": 240869, "epoch": 2706, "lr": 3.1101530016943213e-06}
{"train_loss": 0.10594628006219864, "global_step": 240870, "epoch": 2706, "lr": 3.109951729431765e-06}
{"train_loss": 0.023738831281661987, "global_step": 240871, "epoch": 2706, "lr": 3.109750463472999e-06}
{"train_loss": 0.07224955409765244, "global_step": 240872, "epoch": 2706, "lr": 3.109549203818063e-06}
{"train_loss": 0.04681684076786041, "global_step": 240873, "epoch": 2706, "lr": 3.109347950466973e-06}
{"train_loss": 0.035377103835344315, "global_step": 240874, "epoch": 2706, "lr": 3.109146703419763e-06}
{"train_loss": 0.0509556345641613, "global_step": 240875, "epoch": 2706, "lr": 3.108945462676455e-06}
{"train_loss": 0.044652171432971954, "global_step": 240876, "epoch": 2706, "lr": 3.108744228237076e-06}
{"train_loss": 0.06781503558158875, "global_step": 240877, "epoch": 2706, "lr": 3.10854300010166e-06}
{"train_loss": 0.05760757997632027, "global_step": 240878, "epoch": 2706, "lr": 3.1083417782702184e-06}
{"train_loss": 0.04725469648838043, "global_step": 240879, "epoch": 2706, "lr": 3.1081405627427894e-06}
{"train_loss": 0.012870519421994686, "global_step": 240880, "epoch": 2706, "lr": 3.107939353519407e-06}
{"train_loss": 0.025704726576805115, "global_step": 240881, "epoch": 2706, "lr": 3.107738150600087e-06}
{"train_loss": 0.024292439222335815, "global_step": 240882, "epoch": 2706, "lr": 3.107536953984863e-06}
{"train_loss": 0.08725886791944504, "global_step": 240883, "epoch": 2706, "lr": 3.1073357636737577e-06}
{"train_loss": 0.05265356972813606, "global_step": 240884, "epoch": 2706, "lr": 3.1071345796667984e-06}
{"train_loss": 0.043385304510593414, "global_step": 240885, "epoch": 2706, "lr": 3.1069334019640126e-06}
{"train_loss": 0.03559735417366028, "global_step": 240886, "epoch": 2706, "lr": 3.1067322305654344e-06}
{"train_loss": 0.06744743138551712, "global_step": 240887, "epoch": 2706, "lr": 3.1065310654710746e-06}
{"train_loss": 0.03717311844229698, "global_step": 240888, "epoch": 2706, "lr": 3.106329906680977e-06}
{"train_loss": 0.028606824576854706, "global_step": 240889, "epoch": 2706, "lr": 3.106128754195159e-06}
{"train_loss": 0.04565848037600517, "global_step": 240890, "epoch": 2706, "lr": 3.1059276080136533e-06}
{"train_loss": 0.0537436418235302, "global_step": 240891, "epoch": 2706, "lr": 3.1057264681364774e-06}
{"train_loss": 0.05295354127883911, "global_step": 240892, "epoch": 2706, "lr": 3.105525334563675e-06}
{"train_loss": 0.024169832468032837, "global_step": 240893, "epoch": 2706, "lr": 3.1053242072952515e-06}
{"train_loss": 0.01874428428709507, "global_step": 240894, "epoch": 2706, "lr": 3.1051230863312575e-06}
{"train_loss": 0.04034274443984032, "global_step": 240895, "epoch": 2706, "lr": 3.1049219716716983e-06}
{"train_loss": 0.047912437468767166, "global_step": 240896, "epoch": 2706, "lr": 3.1047208633166125e-06}
{"train_loss": 0.03525708615779877, "global_step": 240897, "epoch": 2706, "lr": 3.1045197612660337e-06}
{"train_loss": 0.04130435734987259, "global_step": 240898, "epoch": 2706, "lr": 3.1043186655199676e-06}
{"train_loss": 0.15773141384124756, "global_step": 240899, "epoch": 2706, "lr": 3.104117576078469e-06}
{"train_loss": 0.04956994578242302, "global_step": 240900, "epoch": 2706, "lr": 3.1039164929415387e-06}
{"train_loss": 0.04368160292506218, "global_step": 240901, "epoch": 2706, "lr": 3.1037154161092207e-06}
{"train_loss": 0.03994186595082283, "global_step": 240902, "epoch": 2706, "lr": 3.103514345581532e-06}
{"train_loss": 0.06379413604736328, "global_step": 240903, "epoch": 2706, "lr": 3.1033132813585107e-06}
{"train_loss": 0.1070941612124443, "global_step": 240904, "epoch": 2706, "lr": 3.103112223440174e-06}
{"train_loss": 0.056519169360399246, "global_step": 240905, "epoch": 2706, "lr": 3.102911171826556e-06}
{"train_loss": 0.030015332624316216, "global_step": 240906, "epoch": 2706, "lr": 3.102710126517672e-06}
{"train_loss": 0.05341824144124985, "global_step": 240907, "epoch": 2706, "lr": 3.1025090875135666e-06}
{"train_loss": 0.04350033029913902, "global_step": 240908, "epoch": 2706, "lr": 3.102308054814251e-06}
{"train_loss": 0.0755215436220169, "global_step": 240909, "epoch": 2706, "lr": 3.1021070284197595e-06}
{"train_loss": 0.018721824511885643, "global_step": 240910, "epoch": 2706, "lr": 3.101906008330119e-06}
{"train_loss": 0.0180558729916811, "global_step": 240911, "epoch": 2706, "lr": 3.101704994545351e-06}
{"train_loss": 0.034806691110134125, "global_step": 240912, "epoch": 2706, "lr": 3.101503987065496e-06}
{"train_loss": 0.03188582509756088, "global_step": 240913, "epoch": 2706, "lr": 3.1013029858905696e-06}
{"train_loss": 0.038522202521562576, "global_step": 240914, "epoch": 2706, "lr": 3.101101991020594e-06}
{"train_loss": 0.07353829592466354, "global_step": 240915, "epoch": 2706, "lr": 3.1009010024556083e-06}
{"train_loss": 0.06935470551252365, "global_step": 240916, "epoch": 2706, "lr": 3.1007000201956403e-06}
{"train_loss": 0.039714790880680084, "global_step": 240917, "epoch": 2706, "lr": 3.100499044240701e-06}
{"train_loss": 0.027321822941303253, "global_step": 240918, "epoch": 2706, "lr": 3.100298074590835e-06}
{"train_loss": 0.035337675362825394, "global_step": 240919, "epoch": 2706, "lr": 3.100097111246059e-06}
{"train_loss": 0.025525329634547234, "global_step": 240920, "epoch": 2706, "lr": 3.0998961542064054e-06}
{"train_loss": 0.05119406804442406, "global_step": 240921, "epoch": 2706, "lr": 3.0996952034718975e-06}
{"train_loss": 0.048370084882284815, "global_step": 240922, "epoch": 2706, "lr": 3.099494259042568e-06, "val_loss": 9.437434196472168}
{"train_loss": 0.07510530203580856, "global_step": 240923, "epoch": 2707, "lr": 3.0992933209184336e-06}
{"train_loss": 0.035006530582904816, "global_step": 240924, "epoch": 2707, "lr": 3.099092389099534e-06}
{"train_loss": 0.03570374846458435, "global_step": 240925, "epoch": 2707, "lr": 3.0988914635858844e-06}
{"train_loss": 0.022876998409628868, "global_step": 240926, "epoch": 2707, "lr": 3.0986905443775248e-06}
{"train_loss": 0.035218313336372375, "global_step": 240927, "epoch": 2707, "lr": 3.0984896314744607e-06}
{"train_loss": 0.043503496795892715, "global_step": 240928, "epoch": 2707, "lr": 3.0982887248767467e-06}
{"train_loss": 0.0512322373688221, "global_step": 240929, "epoch": 2707, "lr": 3.098087824584389e-06}
{"train_loss": 0.017229199409484863, "global_step": 240930, "epoch": 2707, "lr": 3.0978869305974266e-06}
{"train_loss": 0.04686961695551872, "global_step": 240931, "epoch": 2707, "lr": 3.097686042915876e-06}
{"train_loss": 0.07008513808250427, "global_step": 240932, "epoch": 2707, "lr": 3.097485161539776e-06}
{"train_loss": 0.06663337349891663, "global_step": 240933, "epoch": 2707, "lr": 3.097284286469143e-06}
{"train_loss": 0.056947123259305954, "global_step": 240934, "epoch": 2707, "lr": 3.0970834177040053e-06}
{"train_loss": 0.10572147369384766, "global_step": 240935, "epoch": 2707, "lr": 3.0968825552444015e-06}
{"train_loss": 0.06246563047170639, "global_step": 240936, "epoch": 2707, "lr": 3.0966816990903426e-06}
{"train_loss": 0.060286376625299454, "global_step": 240937, "epoch": 2707, "lr": 3.0964808492418675e-06}
{"train_loss": 0.04387368634343147, "global_step": 240938, "epoch": 2707, "lr": 3.096280005698993e-06}
{"train_loss": 0.057619109749794006, "global_step": 240939, "epoch": 2707, "lr": 3.096079168461763e-06}
{"train_loss": 0.044049303978681564, "global_step": 240940, "epoch": 2707, "lr": 3.0958783375301837e-06}
{"train_loss": 0.023949850350618362, "global_step": 240941, "epoch": 2707, "lr": 3.095677512904299e-06}
{"train_loss": 0.01863463595509529, "global_step": 240942, "epoch": 2707, "lr": 3.095476694584121e-06}
{"train_loss": 0.09568248689174652, "global_step": 240943, "epoch": 2707, "lr": 3.0952758825696926e-06}
{"train_loss": 0.054054535925388336, "global_step": 240944, "epoch": 2707, "lr": 3.095075076861026e-06}
{"train_loss": 0.09421800822019577, "global_step": 240945, "epoch": 2707, "lr": 3.0948742774581597e-06}
{"train_loss": 0.05205514654517174, "global_step": 240946, "epoch": 2707, "lr": 3.09467348436111e-06}
{"train_loss": 0.05856848508119583, "global_step": 240947, "epoch": 2707, "lr": 3.094472697569917e-06}
{"train_loss": 0.09014128148555756, "global_step": 240948, "epoch": 2707, "lr": 3.0942719170846015e-06}
{"train_loss": 0.03995954245328903, "global_step": 240949, "epoch": 2707, "lr": 3.0940711429051804e-06}
{"train_loss": 0.05229796841740608, "global_step": 240950, "epoch": 2707, "lr": 3.093870375031699e-06}
{"train_loss": 0.06741216033697128, "global_step": 240951, "epoch": 2707, "lr": 3.093669613464167e-06}
{"train_loss": 0.013591505587100983, "global_step": 240952, "epoch": 2707, "lr": 3.0934688582026306e-06}
{"train_loss": 0.05346236377954483, "global_step": 240953, "epoch": 2707, "lr": 3.093268109247094e-06}
{"train_loss": 0.038506049662828445, "global_step": 240954, "epoch": 2707, "lr": 3.0930673665976017e-06}
{"train_loss": 0.031334832310676575, "global_step": 240955, "epoch": 2707, "lr": 3.092866630254171e-06}
{"train_loss": 0.029302261769771576, "global_step": 240956, "epoch": 2707, "lr": 3.0926659002168345e-06}
{"train_loss": 0.04639262333512306, "global_step": 240957, "epoch": 2707, "lr": 3.092465176485615e-06}
{"train_loss": 0.024981070309877396, "global_step": 240958, "epoch": 2707, "lr": 3.0922644590605454e-06}
{"train_loss": 0.026364650577306747, "global_step": 240959, "epoch": 2707, "lr": 3.092063747941648e-06}
{"train_loss": 0.08907593786716461, "global_step": 240960, "epoch": 2707, "lr": 3.091863043128951e-06}
{"train_loss": 0.055383894592523575, "global_step": 240961, "epoch": 2707, "lr": 3.0916623446224813e-06}
{"train_loss": 0.04603177681565285, "global_step": 240962, "epoch": 2707, "lr": 3.0914616524222673e-06}
{"train_loss": 0.04440409690141678, "global_step": 240963, "epoch": 2707, "lr": 3.091260966528331e-06}
{"train_loss": 0.0551045686006546, "global_step": 240964, "epoch": 2707, "lr": 3.0910602869407056e-06}
{"train_loss": 0.07792980968952179, "global_step": 240965, "epoch": 2707, "lr": 3.0908596136594137e-06}
{"train_loss": 0.054594431072473526, "global_step": 240966, "epoch": 2707, "lr": 3.090658946684488e-06}
{"train_loss": 0.06732475757598877, "global_step": 240967, "epoch": 2707, "lr": 3.0904582860159403e-06}
{"train_loss": 0.06784631311893463, "global_step": 240968, "epoch": 2707, "lr": 3.0902576316538256e-06}
{"train_loss": 0.04001970589160919, "global_step": 240969, "epoch": 2707, "lr": 3.0900569835981384e-06}
{"train_loss": 0.0457281619310379, "global_step": 240970, "epoch": 2707, "lr": 3.0898563418489233e-06}
{"train_loss": 0.05315233767032623, "global_step": 240971, "epoch": 2707, "lr": 3.0896557064062136e-06}
{"train_loss": 0.0812181904911995, "global_step": 240972, "epoch": 2707, "lr": 3.0894550772700204e-06}
{"train_loss": 0.06743187457323074, "global_step": 240973, "epoch": 2707, "lr": 3.089254454440388e-06}
{"train_loss": 0.06394494324922562, "global_step": 240974, "epoch": 2707, "lr": 3.089053837917322e-06}
{"train_loss": 0.04899725317955017, "global_step": 240975, "epoch": 2707, "lr": 3.088853227700872e-06}
{"train_loss": 0.04007897526025772, "global_step": 240976, "epoch": 2707, "lr": 3.088652623791044e-06}
{"train_loss": 0.06498890370130539, "global_step": 240977, "epoch": 2707, "lr": 3.0884520261878825e-06}
{"train_loss": 0.05527426302433014, "global_step": 240978, "epoch": 2707, "lr": 3.088251434891404e-06}
{"train_loss": 0.0528128482401371, "global_step": 240979, "epoch": 2707, "lr": 3.088050849901641e-06}
{"train_loss": 0.049807943403720856, "global_step": 240980, "epoch": 2707, "lr": 3.087850271218612e-06}
{"train_loss": 0.05209929496049881, "global_step": 240981, "epoch": 2707, "lr": 3.0876496988423543e-06}
{"train_loss": 0.016573064029216766, "global_step": 240982, "epoch": 2707, "lr": 3.0874491327728847e-06}
{"train_loss": 0.03169040381908417, "global_step": 240983, "epoch": 2707, "lr": 3.087248573010243e-06}
{"train_loss": 0.037779659032821655, "global_step": 240984, "epoch": 2707, "lr": 3.087048019554445e-06}
{"train_loss": 0.024590078741312027, "global_step": 240985, "epoch": 2707, "lr": 3.086847472405524e-06}
{"train_loss": 0.019274678081274033, "global_step": 240986, "epoch": 2707, "lr": 3.086646931563508e-06}
{"train_loss": 0.05055305361747742, "global_step": 240987, "epoch": 2707, "lr": 3.0864463970284085e-06}
{"train_loss": 0.0320095457136631, "global_step": 240988, "epoch": 2707, "lr": 3.0862458688002748e-06}
{"train_loss": 0.023910539224743843, "global_step": 240989, "epoch": 2707, "lr": 3.0860453468791183e-06}
{"train_loss": 0.0308134276419878, "global_step": 240990, "epoch": 2707, "lr": 3.085844831264978e-06}
{"train_loss": 0.0570819266140461, "global_step": 240991, "epoch": 2707, "lr": 3.0856443219578646e-06}
{"train_loss": 0.06317710131406784, "global_step": 240992, "epoch": 2707, "lr": 3.085443818957823e-06}
{"train_loss": 0.04412153735756874, "global_step": 240993, "epoch": 2707, "lr": 3.0852433222648638e-06}
{"train_loss": 0.07611575722694397, "global_step": 240994, "epoch": 2707, "lr": 3.0850428318790315e-06}
{"train_loss": 0.03638996183872223, "global_step": 240995, "epoch": 2707, "lr": 3.084842347800332e-06}
{"train_loss": 0.03775183856487274, "global_step": 240996, "epoch": 2707, "lr": 3.084641870028815e-06}
{"train_loss": 0.014457696117460728, "global_step": 240997, "epoch": 2707, "lr": 3.084441398564486e-06}
{"train_loss": 0.07417619973421097, "global_step": 240998, "epoch": 2707, "lr": 3.0842409334073896e-06}
{"train_loss": 0.009929797612130642, "global_step": 240999, "epoch": 2707, "lr": 3.0840404745575367e-06}
{"train_loss": 0.022442016750574112, "global_step": 241000, "epoch": 2707, "lr": 3.0838400220149723e-06}
{"train_loss": 0.04722606763243675, "global_step": 241001, "epoch": 2707, "lr": 3.0836395757797066e-06}
{"train_loss": 0.09704441577196121, "global_step": 241002, "epoch": 2707, "lr": 3.083439135851779e-06}
{"train_loss": 0.03289434686303139, "global_step": 241003, "epoch": 2707, "lr": 3.083238702231206e-06}
{"train_loss": 0.02803877741098404, "global_step": 241004, "epoch": 2707, "lr": 3.0830382749180263e-06}
{"train_loss": 0.0405646376311779, "global_step": 241005, "epoch": 2707, "lr": 3.0828378539122514e-06}
{"train_loss": 0.04892471060156822, "global_step": 241006, "epoch": 2707, "lr": 3.08263743921392e-06}
{"train_loss": 0.120200015604496, "global_step": 241007, "epoch": 2707, "lr": 3.08243703082306e-06}
{"train_loss": 0.03902337700128555, "global_step": 241008, "epoch": 2707, "lr": 3.0822366287396874e-06}
{"train_loss": 0.08981035649776459, "global_step": 241009, "epoch": 2707, "lr": 3.0820362329638473e-06}
{"train_loss": 0.04480521380901337, "global_step": 241010, "epoch": 2707, "lr": 3.081835843495545e-06}
{"train_loss": 0.049698400405350694, "global_step": 241011, "epoch": 2707, "lr": 3.081635460334825e-06, "val_loss": 9.454684257507324}
{"train_loss": 0.03967050090432167, "global_step": 241012, "epoch": 2708, "lr": 3.081435083481704e-06}
{"train_loss": 0.05796874687075615, "global_step": 241013, "epoch": 2708, "lr": 3.0812347129362153e-06}
{"train_loss": 0.03378761187195778, "global_step": 241014, "epoch": 2708, "lr": 3.0810343486983807e-06}
{"train_loss": 0.05486486107110977, "global_step": 241015, "epoch": 2708, "lr": 3.080833990768234e-06}
{"train_loss": 0.04006946086883545, "global_step": 241016, "epoch": 2708, "lr": 3.0806336391457857e-06}
{"train_loss": 0.05714081600308418, "global_step": 241017, "epoch": 2708, "lr": 3.0804332938310864e-06}
{"train_loss": 0.026812704280018806, "global_step": 241018, "epoch": 2708, "lr": 3.0802329548241415e-06}
{"train_loss": 0.039180170744657516, "global_step": 241019, "epoch": 2708, "lr": 3.0800326221249955e-06}
{"train_loss": 0.03843846172094345, "global_step": 241020, "epoch": 2708, "lr": 3.079832295733659e-06}
{"train_loss": 0.016966436058282852, "global_step": 241021, "epoch": 2708, "lr": 3.0796319756501767e-06}
{"train_loss": 0.06898026168346405, "global_step": 241022, "epoch": 2708, "lr": 3.0794316618745544e-06}
{"train_loss": 0.03878495842218399, "global_step": 241023, "epoch": 2708, "lr": 3.079231354406842e-06}
{"train_loss": 0.06741979718208313, "global_step": 241024, "epoch": 2708, "lr": 3.0790310532470558e-06}
{"train_loss": 0.042826369404792786, "global_step": 241025, "epoch": 2708, "lr": 3.0788307583952125e-06}
{"train_loss": 0.05583437532186508, "global_step": 241026, "epoch": 2708, "lr": 3.078630469851357e-06}
{"train_loss": 0.06649970263242722, "global_step": 241027, "epoch": 2708, "lr": 3.0784301876155e-06}
{"train_loss": 0.023265518248081207, "global_step": 241028, "epoch": 2708, "lr": 3.07822991168768e-06}
{"train_loss": 0.02596370503306389, "global_step": 241029, "epoch": 2708, "lr": 3.0780296420679144e-06}
{"train_loss": 0.03920421004295349, "global_step": 241030, "epoch": 2708, "lr": 3.077829378756247e-06}
{"train_loss": 0.11991912871599197, "global_step": 241031, "epoch": 2708, "lr": 3.077629121752684e-06}
{"train_loss": 0.057988569140434265, "global_step": 241032, "epoch": 2708, "lr": 3.0774288710572698e-06}
{"train_loss": 0.016104066744446754, "global_step": 241033, "epoch": 2708, "lr": 3.0772286266700147e-06}
{"train_loss": 0.03305429965257645, "global_step": 241034, "epoch": 2708, "lr": 3.077028388590958e-06}
{"train_loss": 0.09095294773578644, "global_step": 241035, "epoch": 2708, "lr": 3.076828156820122e-06}
{"train_loss": 0.03681183233857155, "global_step": 241036, "epoch": 2708, "lr": 3.0766279313575395e-06}
{"train_loss": 0.0202662143856287, "global_step": 241037, "epoch": 2708, "lr": 3.0764277122032227e-06}
{"train_loss": 0.06194228678941727, "global_step": 241038, "epoch": 2708, "lr": 3.0762274993572203e-06}
{"train_loss": 0.050786688923835754, "global_step": 241039, "epoch": 2708, "lr": 3.076027292819539e-06}
{"train_loss": 0.03909943252801895, "global_step": 241040, "epoch": 2708, "lr": 3.0758270925902167e-06}
{"train_loss": 0.08851256221532822, "global_step": 241041, "epoch": 2708, "lr": 3.075626898669276e-06}
{"train_loss": 0.06861524283885956, "global_step": 241042, "epoch": 2708, "lr": 3.075426711056739e-06}
{"train_loss": 0.046801283955574036, "global_step": 241043, "epoch": 2708, "lr": 3.0752265297526504e-06}
{"train_loss": 0.017771849408745766, "global_step": 241044, "epoch": 2708, "lr": 3.075026354757021e-06}
{"train_loss": 0.03659151867032051, "global_step": 241045, "epoch": 2708, "lr": 3.0748261860698846e-06}
{"train_loss": 0.05845888331532478, "global_step": 241046, "epoch": 2708, "lr": 3.0746260236912626e-06}
{"train_loss": 0.0408923476934433, "global_step": 241047, "epoch": 2708, "lr": 3.074425867621189e-06}
{"train_loss": 0.053575702011585236, "global_step": 241048, "epoch": 2708, "lr": 3.07422571785968e-06}
{"train_loss": 0.042779725044965744, "global_step": 241049, "epoch": 2708, "lr": 3.0740255744067803e-06}
{"train_loss": 0.03936675190925598, "global_step": 241050, "epoch": 2708, "lr": 3.0738254372624955e-06}
{"train_loss": 0.02438654750585556, "global_step": 241051, "epoch": 2708, "lr": 3.0736253064268692e-06}
{"train_loss": 0.044530414044857025, "global_step": 241052, "epoch": 2708, "lr": 3.0734251818999194e-06}
{"train_loss": 0.032657474279403687, "global_step": 241053, "epoch": 2708, "lr": 3.0732250636816786e-06}
{"train_loss": 0.03549790382385254, "global_step": 241054, "epoch": 2708, "lr": 3.0730249517721633e-06}
{"train_loss": 0.02443009242415428, "global_step": 241055, "epoch": 2708, "lr": 3.072824846171418e-06}
{"train_loss": 0.05338720232248306, "global_step": 241056, "epoch": 2708, "lr": 3.0726247468794487e-06}
{"train_loss": 0.07114095985889435, "global_step": 241057, "epoch": 2708, "lr": 3.0724246538963052e-06}
{"train_loss": 0.05320197716355324, "global_step": 241058, "epoch": 2708, "lr": 3.072224567221993e-06}
{"train_loss": 0.014740352518856525, "global_step": 241059, "epoch": 2708, "lr": 3.0720244868565507e-06}
{"train_loss": 0.024625932797789574, "global_step": 241060, "epoch": 2708, "lr": 3.0718244128000006e-06}
{"train_loss": 0.01960386335849762, "global_step": 241061, "epoch": 2708, "lr": 3.0716243450523763e-06}
{"train_loss": 0.031022340059280396, "global_step": 241062, "epoch": 2708, "lr": 3.071424283613694e-06}
{"train_loss": 0.04611753672361374, "global_step": 241063, "epoch": 2708, "lr": 3.0712242284839986e-06}
{"train_loss": 0.06720394641160965, "global_step": 241064, "epoch": 2708, "lr": 3.071024179663301e-06}
{"train_loss": 0.058737512677907944, "global_step": 241065, "epoch": 2708, "lr": 3.0708241371516232e-06}
{"train_loss": 0.01652941294014454, "global_step": 241066, "epoch": 2708, "lr": 3.07062410094901e-06}
{"train_loss": 0.0373227596282959, "global_step": 241067, "epoch": 2708, "lr": 3.0704240710554722e-06}
{"train_loss": 0.0078998152166605, "global_step": 241068, "epoch": 2708, "lr": 3.0702240474710485e-06}
{"train_loss": 0.04223748296499252, "global_step": 241069, "epoch": 2708, "lr": 3.070024030195756e-06}
{"train_loss": 0.018590513616800308, "global_step": 241070, "epoch": 2708, "lr": 3.069824019229639e-06}
{"train_loss": 0.0662088617682457, "global_step": 241071, "epoch": 2708, "lr": 3.0696240145726975e-06}
{"train_loss": 0.03344561904668808, "global_step": 241072, "epoch": 2708, "lr": 3.0694240162249864e-06}
{"train_loss": 0.049692679196596146, "global_step": 241073, "epoch": 2708, "lr": 3.0692240241865066e-06}
{"train_loss": 0.05120086297392845, "global_step": 241074, "epoch": 2708, "lr": 3.0690240384573075e-06}
{"train_loss": 0.04071144014596939, "global_step": 241075, "epoch": 2708, "lr": 3.068824059037395e-06}
{"train_loss": 0.024192266166210175, "global_step": 241076, "epoch": 2708, "lr": 3.068624085926819e-06}
{"train_loss": 0.0914253368973732, "global_step": 241077, "epoch": 2708, "lr": 3.0684241191255846e-06}
{"train_loss": 0.05377267673611641, "global_step": 241078, "epoch": 2708, "lr": 3.0682241586337313e-06}
{"train_loss": 0.05064902454614639, "global_step": 241079, "epoch": 2708, "lr": 3.0680242044512918e-06}
{"train_loss": 0.056261565536260605, "global_step": 241080, "epoch": 2708, "lr": 3.0678242565782723e-06}
{"train_loss": 0.08803018927574158, "global_step": 241081, "epoch": 2708, "lr": 3.067624315014722e-06}
{"train_loss": 0.060974784195423126, "global_step": 241082, "epoch": 2708, "lr": 3.067424379760647e-06}
{"train_loss": 0.07374417781829834, "global_step": 241083, "epoch": 2708, "lr": 3.0672244508160976e-06}
{"train_loss": 0.051726989448070526, "global_step": 241084, "epoch": 2708, "lr": 3.0670245281810783e-06}
{"train_loss": 0.05059905722737312, "global_step": 241085, "epoch": 2708, "lr": 3.0668246118556287e-06}
{"train_loss": 0.024716708809137344, "global_step": 241086, "epoch": 2708, "lr": 3.066624701839771e-06}
{"train_loss": 0.025102922692894936, "global_step": 241087, "epoch": 2708, "lr": 3.066424798133538e-06}
{"train_loss": 0.040034156292676926, "global_step": 241088, "epoch": 2708, "lr": 3.066224900736947e-06}
{"train_loss": 0.02647557482123375, "global_step": 241089, "epoch": 2708, "lr": 3.0660250096500363e-06}
{"train_loss": 0.044099386781454086, "global_step": 241090, "epoch": 2708, "lr": 3.0658251248728175e-06}
{"train_loss": 0.0381598174571991, "global_step": 241091, "epoch": 2708, "lr": 3.0656252464053347e-06}
{"train_loss": 0.06074105203151703, "global_step": 241092, "epoch": 2708, "lr": 3.065425374247599e-06}
{"train_loss": 0.030796140432357788, "global_step": 241093, "epoch": 2708, "lr": 3.0652255083996493e-06}
{"train_loss": 0.042477935552597046, "global_step": 241094, "epoch": 2708, "lr": 3.065025648861508e-06}
{"train_loss": 0.06341749429702759, "global_step": 241095, "epoch": 2708, "lr": 3.064825795633203e-06}
{"train_loss": 0.05071762949228287, "global_step": 241096, "epoch": 2708, "lr": 3.0646259487147556e-06}
{"train_loss": 0.05560002475976944, "global_step": 241097, "epoch": 2708, "lr": 3.0644261081062053e-06}
{"train_loss": 0.06732168793678284, "global_step": 241098, "epoch": 2708, "lr": 3.064226273807558e-06}
{"train_loss": 0.08011744171380997, "global_step": 241099, "epoch": 2708, "lr": 3.064026445818868e-06}
{"train_loss": 0.046251055951952266, "global_step": 241100, "epoch": 2708, "lr": 3.063826624140137e-06, "val_loss": 9.592804908752441}
{"train_loss": 0.014080461114645004, "global_step": 241101, "epoch": 2709, "lr": 3.063626808771408e-06}
{"train_loss": 0.048287130892276764, "global_step": 241102, "epoch": 2709, "lr": 3.063426999712704e-06}
{"train_loss": 0.03504660353064537, "global_step": 241103, "epoch": 2709, "lr": 3.0632271969640414e-06}
{"train_loss": 0.01224953681230545, "global_step": 241104, "epoch": 2709, "lr": 3.0630274005254645e-06}
{"train_loss": 0.01939302682876587, "global_step": 241105, "epoch": 2709, "lr": 3.0628276103969846e-06}
{"train_loss": 0.033133700489997864, "global_step": 241106, "epoch": 2709, "lr": 3.0626278265786455e-06}
{"train_loss": 0.028926091268658638, "global_step": 241107, "epoch": 2709, "lr": 3.0624280490704538e-06}
{"train_loss": 0.07609295099973679, "global_step": 241108, "epoch": 2709, "lr": 3.0622282778724533e-06}
{"train_loss": 0.07508803904056549, "global_step": 241109, "epoch": 2709, "lr": 3.062028512984655e-06}
{"train_loss": 0.07003340870141983, "global_step": 241110, "epoch": 2709, "lr": 3.0618287544071035e-06}
{"train_loss": 0.026515712961554527, "global_step": 241111, "epoch": 2709, "lr": 3.0616290021398155e-06}
{"train_loss": 0.029761405661702156, "global_step": 241112, "epoch": 2709, "lr": 3.0614292561828184e-06}
{"train_loss": 0.016816219314932823, "global_step": 241113, "epoch": 2709, "lr": 3.061229516536135e-06}
{"train_loss": 0.029530230909585953, "global_step": 241114, "epoch": 2709, "lr": 3.061029783199798e-06}
{"train_loss": 0.041386205703020096, "global_step": 241115, "epoch": 2709, "lr": 3.0608300561738414e-06}
{"train_loss": 0.11618753522634506, "global_step": 241116, "epoch": 2709, "lr": 3.0606303354582753e-06}
{"train_loss": 0.030110206454992294, "global_step": 241117, "epoch": 2709, "lr": 3.0604306210531454e-06}
{"train_loss": 0.034786973148584366, "global_step": 241118, "epoch": 2709, "lr": 3.0602309129584617e-06}
{"train_loss": 0.05232398957014084, "global_step": 241119, "epoch": 2709, "lr": 3.0600312111742636e-06}
{"train_loss": 0.04839947819709778, "global_step": 241120, "epoch": 2709, "lr": 3.059831515700562e-06}
{"train_loss": 0.05949486047029495, "global_step": 241121, "epoch": 2709, "lr": 3.0596318265374068e-06}
{"train_loss": 0.0841115415096283, "global_step": 241122, "epoch": 2709, "lr": 3.0594321436847983e-06}
{"train_loss": 0.06496325880289078, "global_step": 241123, "epoch": 2709, "lr": 3.0592324671427865e-06}
{"train_loss": 0.02058144100010395, "global_step": 241124, "epoch": 2709, "lr": 3.059032796911382e-06}
{"train_loss": 0.06311100721359253, "global_step": 241125, "epoch": 2709, "lr": 3.0588331329906293e-06}
{"train_loss": 0.01637335866689682, "global_step": 241126, "epoch": 2709, "lr": 3.0586334753805344e-06}
{"train_loss": 0.06801167875528336, "global_step": 241127, "epoch": 2709, "lr": 3.0584338240811416e-06}
{"train_loss": 0.05621480941772461, "global_step": 241128, "epoch": 2709, "lr": 3.058234179092462e-06}
{"train_loss": 0.1056700125336647, "global_step": 241129, "epoch": 2709, "lr": 3.058034540414534e-06}
{"train_loss": 0.03744450956583023, "global_step": 241130, "epoch": 2709, "lr": 3.0578349080473802e-06}
{"train_loss": 0.12853533029556274, "global_step": 241131, "epoch": 2709, "lr": 3.0576352819910338e-06}
{"train_loss": 0.053790491074323654, "global_step": 241132, "epoch": 2709, "lr": 3.057435662245506e-06}
{"train_loss": 0.03395800665020943, "global_step": 241133, "epoch": 2709, "lr": 3.0572360488108355e-06}
{"train_loss": 0.028578780591487885, "global_step": 241134, "epoch": 2709, "lr": 3.0570364416870556e-06}
{"train_loss": 0.034433089196681976, "global_step": 241135, "epoch": 2709, "lr": 3.0568368408741887e-06}
{"train_loss": 0.019731611013412476, "global_step": 241136, "epoch": 2709, "lr": 3.056637246372246e-06}
{"train_loss": 0.031233690679073334, "global_step": 241137, "epoch": 2709, "lr": 3.056437658181277e-06}
{"train_loss": 0.01562654972076416, "global_step": 241138, "epoch": 2709, "lr": 3.0562380763012876e-06}
{"train_loss": 0.10415292531251907, "global_step": 241139, "epoch": 2709, "lr": 3.056038500732322e-06}
{"train_loss": 0.00600225618109107, "global_step": 241140, "epoch": 2709, "lr": 3.0558389314743972e-06}
{"train_loss": 0.04919804632663727, "global_step": 241141, "epoch": 2709, "lr": 3.0556393685275407e-06}
{"train_loss": 0.05360922962427139, "global_step": 241142, "epoch": 2709, "lr": 3.055439811891786e-06}
{"train_loss": 0.0838470309972763, "global_step": 241143, "epoch": 2709, "lr": 3.0552402615671494e-06}
{"train_loss": 0.03539763391017914, "global_step": 241144, "epoch": 2709, "lr": 3.05504071755367e-06}
{"train_loss": 0.05548776686191559, "global_step": 241145, "epoch": 2709, "lr": 3.0548411798513587e-06}
{"train_loss": 0.07887281477451324, "global_step": 241146, "epoch": 2709, "lr": 3.05464164846026e-06}
{"train_loss": 0.0271390900015831, "global_step": 241147, "epoch": 2709, "lr": 3.054442123380391e-06}
{"train_loss": 0.04226702079176903, "global_step": 241148, "epoch": 2709, "lr": 3.0542426046117787e-06}
{"train_loss": 0.03843522071838379, "global_step": 241149, "epoch": 2709, "lr": 3.0540430921544517e-06}
{"train_loss": 0.02648881822824478, "global_step": 241150, "epoch": 2709, "lr": 3.053843586008431e-06}
{"train_loss": 0.05975114554166794, "global_step": 241151, "epoch": 2709, "lr": 3.053644086173757e-06}
{"train_loss": 0.041196491569280624, "global_step": 241152, "epoch": 2709, "lr": 3.0534445926504453e-06}
{"train_loss": 0.10614937543869019, "global_step": 241153, "epoch": 2709, "lr": 3.053245105438529e-06}
{"train_loss": 0.060870494693517685, "global_step": 241154, "epoch": 2709, "lr": 3.0530456245380257e-06}
{"train_loss": 0.06798163801431656, "global_step": 241155, "epoch": 2709, "lr": 3.0528461499489736e-06}
{"train_loss": 0.03463602438569069, "global_step": 241156, "epoch": 2709, "lr": 3.0526466816713893e-06}
{"train_loss": 0.09526834636926651, "global_step": 241157, "epoch": 2709, "lr": 3.0524472197053066e-06}
{"train_loss": 0.04216887429356575, "global_step": 241158, "epoch": 2709, "lr": 3.0522477640507472e-06}
{"train_loss": 0.07016676664352417, "global_step": 241159, "epoch": 2709, "lr": 3.0520483147077505e-06}
{"train_loss": 0.03861786425113678, "global_step": 241160, "epoch": 2709, "lr": 3.0518488716763216e-06}
{"train_loss": 0.037632476538419724, "global_step": 241161, "epoch": 2709, "lr": 3.0516494349565104e-06}
{"train_loss": 0.02700088918209076, "global_step": 241162, "epoch": 2709, "lr": 3.051450004548323e-06}
{"train_loss": 0.06647582352161407, "global_step": 241163, "epoch": 2709, "lr": 3.0512505804518033e-06}
{"train_loss": 0.029491323977708817, "global_step": 241164, "epoch": 2709, "lr": 3.051051162666968e-06}
{"train_loss": 0.08513697981834412, "global_step": 241165, "epoch": 2709, "lr": 3.0508517511938506e-06}
{"train_loss": 0.05591744929552078, "global_step": 241166, "epoch": 2709, "lr": 3.050652346032462e-06}
{"train_loss": 0.06760597229003906, "global_step": 241167, "epoch": 2709, "lr": 3.0504529471828523e-06}
{"train_loss": 0.05324322730302811, "global_step": 241168, "epoch": 2709, "lr": 3.0502535546450327e-06}
{"train_loss": 0.08697490394115448, "global_step": 241169, "epoch": 2709, "lr": 3.050054168419031e-06}
{"train_loss": 0.04745856299996376, "global_step": 241170, "epoch": 2709, "lr": 3.0498547885048855e-06}
{"train_loss": 0.031054088845849037, "global_step": 241171, "epoch": 2709, "lr": 3.049655414902608e-06}
{"train_loss": 0.039317067712545395, "global_step": 241172, "epoch": 2709, "lr": 3.049456047612237e-06}
{"train_loss": 0.06519943475723267, "global_step": 241173, "epoch": 2709, "lr": 3.049256686633789e-06}
{"train_loss": 0.055694807320833206, "global_step": 241174, "epoch": 2709, "lr": 3.0490573319673032e-06}
{"train_loss": 0.03662790358066559, "global_step": 241175, "epoch": 2709, "lr": 3.048857983612796e-06}
{"train_loss": 0.029709402471780777, "global_step": 241176, "epoch": 2709, "lr": 3.0486586415702958e-06}
{"train_loss": 0.03460993617773056, "global_step": 241177, "epoch": 2709, "lr": 3.0484593058398346e-06}
{"train_loss": 0.044842589646577835, "global_step": 241178, "epoch": 2709, "lr": 3.0482599764214304e-06}
{"train_loss": 0.025512877851724625, "global_step": 241179, "epoch": 2709, "lr": 3.0480606533151212e-06}
{"train_loss": 0.027782103046774864, "global_step": 241180, "epoch": 2709, "lr": 3.0478613365209295e-06}
{"train_loss": 0.05700623616576195, "global_step": 241181, "epoch": 2709, "lr": 3.0476620260388777e-06}
{"train_loss": 0.06616350263357162, "global_step": 241182, "epoch": 2709, "lr": 3.0474627218689934e-06}
{"train_loss": 0.08865711838006973, "global_step": 241183, "epoch": 2709, "lr": 3.047263424011304e-06}
{"train_loss": 0.037325453013181686, "global_step": 241184, "epoch": 2709, "lr": 3.0470641324658435e-06}
{"train_loss": 0.0673031359910965, "global_step": 241185, "epoch": 2709, "lr": 3.046864847232628e-06}
{"train_loss": 0.01571117900311947, "global_step": 241186, "epoch": 2709, "lr": 3.0466655683116963e-06}
{"train_loss": 0.07302147895097733, "global_step": 241187, "epoch": 2709, "lr": 3.0464662957030545e-06}
{"train_loss": 0.04254394769668579, "global_step": 241188, "epoch": 2709, "lr": 3.046267029406746e-06}
{"train_loss": 0.04927833102580704, "global_step": 241189, "epoch": 2709, "lr": 3.0460677694228056e-06, "val_loss": 9.493352890014648}
{"train_loss": 0.09122854471206665, "global_step": 241190, "epoch": 2710, "lr": 3.045868515751238e-06}
{"train_loss": 0.02188807725906372, "global_step": 241191, "epoch": 2710, "lr": 3.045669268392093e-06}
{"train_loss": 0.05843102186918259, "global_step": 241192, "epoch": 2710, "lr": 3.045470027345376e-06}
{"train_loss": 0.041881073266267776, "global_step": 241193, "epoch": 2710, "lr": 3.045270792611127e-06}
{"train_loss": 0.05108845233917236, "global_step": 241194, "epoch": 2710, "lr": 3.045071564189367e-06}
{"train_loss": 0.04631095007061958, "global_step": 241195, "epoch": 2710, "lr": 3.0448723420801294e-06}
{"train_loss": 0.02547115460038185, "global_step": 241196, "epoch": 2710, "lr": 3.044673126283426e-06}
{"train_loss": 0.05991147831082344, "global_step": 241197, "epoch": 2710, "lr": 3.0444739167993065e-06}
{"train_loss": 0.04373272508382797, "global_step": 241198, "epoch": 2710, "lr": 3.0442747136277762e-06}
{"train_loss": 0.03613729774951935, "global_step": 241199, "epoch": 2710, "lr": 3.0440755167688797e-06}
{"train_loss": 0.0321602039039135, "global_step": 241200, "epoch": 2710, "lr": 3.0438763262226224e-06}
{"train_loss": 0.045081790536642075, "global_step": 241201, "epoch": 2710, "lr": 3.0436771419890543e-06}
{"train_loss": 0.0705951526761055, "global_step": 241202, "epoch": 2710, "lr": 3.0434779640681864e-06}
{"train_loss": 0.0452837310731411, "global_step": 241203, "epoch": 2710, "lr": 3.043278792460058e-06}
{"train_loss": 0.04291091859340668, "global_step": 241204, "epoch": 2710, "lr": 3.0430796271646742e-06}
{"train_loss": 0.04367233067750931, "global_step": 241205, "epoch": 2710, "lr": 3.0428804681820854e-06}
{"train_loss": 0.025374814867973328, "global_step": 241206, "epoch": 2710, "lr": 3.0426813155123077e-06}
{"train_loss": 0.08367448300123215, "global_step": 241207, "epoch": 2710, "lr": 3.0424821691553697e-06}
{"train_loss": 0.0507749542593956, "global_step": 241208, "epoch": 2710, "lr": 3.042283029111298e-06}
{"train_loss": 0.05352052301168442, "global_step": 241209, "epoch": 2710, "lr": 3.0420838953801155e-06}
{"train_loss": 0.017469026148319244, "global_step": 241210, "epoch": 2710, "lr": 3.0418847679618613e-06}
{"train_loss": 0.09481044858694077, "global_step": 241211, "epoch": 2710, "lr": 3.0416856468565403e-06}
{"train_loss": 0.03023911640048027, "global_step": 241212, "epoch": 2710, "lr": 3.041486532064203e-06}
{"train_loss": 0.05947249382734299, "global_step": 241213, "epoch": 2710, "lr": 3.0412874235848655e-06}
{"train_loss": 0.03495725244283676, "global_step": 241214, "epoch": 2710, "lr": 3.04108832141855e-06}
{"train_loss": 0.04538143798708916, "global_step": 241215, "epoch": 2710, "lr": 3.040889225565291e-06}
{"train_loss": 0.07133013755083084, "global_step": 241216, "epoch": 2710, "lr": 3.0406901360251096e-06}
{"train_loss": 0.07080584019422531, "global_step": 241217, "epoch": 2710, "lr": 3.040491052798039e-06}
{"train_loss": 0.025511838495731354, "global_step": 241218, "epoch": 2710, "lr": 3.040291975884102e-06}
{"train_loss": 0.041548535227775574, "global_step": 241219, "epoch": 2710, "lr": 3.040092905283315e-06}
{"train_loss": 0.012340040877461433, "global_step": 241220, "epoch": 2710, "lr": 3.039893840995728e-06}
{"train_loss": 0.09456285089254379, "global_step": 241221, "epoch": 2710, "lr": 3.039694783021346e-06}
{"train_loss": 0.05225233733654022, "global_step": 241222, "epoch": 2710, "lr": 3.039495731360209e-06}
{"train_loss": 0.056774064898490906, "global_step": 241223, "epoch": 2710, "lr": 3.0392966860123327e-06}
{"train_loss": 0.04312225431203842, "global_step": 241224, "epoch": 2710, "lr": 3.039097646977751e-06}
{"train_loss": 0.03040137328207493, "global_step": 241225, "epoch": 2710, "lr": 3.0388986142565024e-06}
{"train_loss": 0.06430508941411972, "global_step": 241226, "epoch": 2710, "lr": 3.038699587848587e-06}
{"train_loss": 0.032432541251182556, "global_step": 241227, "epoch": 2710, "lr": 3.03850056775406e-06}
{"train_loss": 0.06420374661684036, "global_step": 241228, "epoch": 2710, "lr": 3.038301553972922e-06}
{"train_loss": 0.038553573191165924, "global_step": 241229, "epoch": 2710, "lr": 3.0381025465052224e-06}
{"train_loss": 0.01639377512037754, "global_step": 241230, "epoch": 2710, "lr": 3.0379035453509675e-06}
{"train_loss": 0.039000529795885086, "global_step": 241231, "epoch": 2710, "lr": 3.0377045505102064e-06}
{"train_loss": 0.08027052879333496, "global_step": 241232, "epoch": 2710, "lr": 3.0375055619829397e-06}
{"train_loss": 0.03956107422709465, "global_step": 241233, "epoch": 2710, "lr": 3.037306579769217e-06}
{"train_loss": 0.04607623443007469, "global_step": 241234, "epoch": 2710, "lr": 3.037107603869055e-06}
{"train_loss": 0.07666727155447006, "global_step": 241235, "epoch": 2710, "lr": 3.0369086342824815e-06}
{"train_loss": 0.07439251989126205, "global_step": 241236, "epoch": 2710, "lr": 3.036709671009519e-06}
{"train_loss": 0.047691117972135544, "global_step": 241237, "epoch": 2710, "lr": 3.036510714050206e-06}
{"train_loss": 0.036681775003671646, "global_step": 241238, "epoch": 2710, "lr": 3.0363117634045536e-06}
{"train_loss": 0.048782579600811005, "global_step": 241239, "epoch": 2710, "lr": 3.0361128190726005e-06}
{"train_loss": 0.04606271907687187, "global_step": 241240, "epoch": 2710, "lr": 3.035913881054364e-06}
{"train_loss": 0.0253162644803524, "global_step": 241241, "epoch": 2710, "lr": 3.0357149493498825e-06}
{"train_loss": 0.06827283650636673, "global_step": 241242, "epoch": 2710, "lr": 3.0355160239591786e-06}
{"train_loss": 0.018214967101812363, "global_step": 241243, "epoch": 2710, "lr": 3.035317104882268e-06}
{"train_loss": 0.03581780567765236, "global_step": 241244, "epoch": 2710, "lr": 3.035118192119196e-06}
{"train_loss": 0.04535503685474396, "global_step": 241245, "epoch": 2710, "lr": 3.0349192856699738e-06}
{"train_loss": 0.057972606271505356, "global_step": 241246, "epoch": 2710, "lr": 3.034720385534645e-06}
{"train_loss": 0.03075605444610119, "global_step": 241247, "epoch": 2710, "lr": 3.0345214917132103e-06}
{"train_loss": 0.039266958832740784, "global_step": 241248, "epoch": 2710, "lr": 3.034322604205725e-06}
{"train_loss": 0.04562242701649666, "global_step": 241249, "epoch": 2710, "lr": 3.0341237230121944e-06}
{"train_loss": 0.05967208743095398, "global_step": 241250, "epoch": 2710, "lr": 3.0339248481326574e-06}
{"train_loss": 0.0918126255273819, "global_step": 241251, "epoch": 2710, "lr": 3.0337259795671425e-06}
{"train_loss": 0.05291789025068283, "global_step": 241252, "epoch": 2710, "lr": 3.03352711731566e-06}
{"train_loss": 0.061598338186740875, "global_step": 241253, "epoch": 2710, "lr": 3.033328261378254e-06}
{"train_loss": 0.06540118902921677, "global_step": 241254, "epoch": 2710, "lr": 3.033129411754937e-06}
{"train_loss": 0.0688260942697525, "global_step": 241255, "epoch": 2710, "lr": 3.032930568445752e-06}
{"train_loss": 0.022628920152783394, "global_step": 241256, "epoch": 2710, "lr": 3.0327317314507166e-06}
{"train_loss": 0.08122505247592926, "global_step": 241257, "epoch": 2710, "lr": 3.0325329007698523e-06}
{"train_loss": 0.040516290813684464, "global_step": 241258, "epoch": 2710, "lr": 3.0323340764031984e-06}
{"train_loss": 0.08979380875825882, "global_step": 241259, "epoch": 2710, "lr": 3.0321352583507656e-06}
{"train_loss": 0.037358276546001434, "global_step": 241260, "epoch": 2710, "lr": 3.0319364466125878e-06}
{"train_loss": 0.04875162988901138, "global_step": 241261, "epoch": 2710, "lr": 3.0317376411887035e-06}
{"train_loss": 0.027714332565665245, "global_step": 241262, "epoch": 2710, "lr": 3.0315388420791237e-06}
{"train_loss": 0.08924425393342972, "global_step": 241263, "epoch": 2710, "lr": 3.0313400492838873e-06}
{"train_loss": 0.049580711871385574, "global_step": 241264, "epoch": 2710, "lr": 3.0311412628030055e-06}
{"train_loss": 0.04659295454621315, "global_step": 241265, "epoch": 2710, "lr": 3.030942482636523e-06}
{"train_loss": 0.04938490688800812, "global_step": 241266, "epoch": 2710, "lr": 3.0307437087844505e-06}
{"train_loss": 0.03333139419555664, "global_step": 241267, "epoch": 2710, "lr": 3.0305449412468325e-06}
{"train_loss": 0.04994579032063484, "global_step": 241268, "epoch": 2710, "lr": 3.0303461800236744e-06}
{"train_loss": 0.062454357743263245, "global_step": 241269, "epoch": 2710, "lr": 3.0301474251150207e-06}
{"train_loss": 0.05597057193517685, "global_step": 241270, "epoch": 2710, "lr": 3.029948676520883e-06}
{"train_loss": 0.05934302136301994, "global_step": 241271, "epoch": 2710, "lr": 3.0297499342413104e-06}
{"train_loss": 0.040995728224515915, "global_step": 241272, "epoch": 2710, "lr": 3.029551198276304e-06}
{"train_loss": 0.04790453985333443, "global_step": 241273, "epoch": 2710, "lr": 3.0293524686259066e-06}
{"train_loss": 0.013819889165461063, "global_step": 241274, "epoch": 2710, "lr": 3.0291537452901308e-06}
{"train_loss": 0.04650294780731201, "global_step": 241275, "epoch": 2710, "lr": 3.028955028269026e-06}
{"train_loss": 0.030407441779971123, "global_step": 241276, "epoch": 2710, "lr": 3.0287563175625975e-06}
{"train_loss": 0.010762465186417103, "global_step": 241277, "epoch": 2710, "lr": 3.028557613170879e-06}
{"train_loss": 0.04903675264187073, "global_step": 241278, "epoch": 2710, "lr": 3.028358915093904e-06, "val_loss": 9.468709945678711, "train_action_mse_error": 20.421215057373047}
{"train_loss": 0.04012861102819443, "global_step": 241279, "epoch": 2711, "lr": 3.0281602233316887e-06}
{"train_loss": 0.035848937928676605, "global_step": 241280, "epoch": 2711, "lr": 3.0279615378842717e-06}
{"train_loss": 0.06056366115808487, "global_step": 241281, "epoch": 2711, "lr": 3.027762858751665e-06}
{"train_loss": 0.03100363351404667, "global_step": 241282, "epoch": 2711, "lr": 3.0275641859339123e-06}
{"train_loss": 0.04716300591826439, "global_step": 241283, "epoch": 2711, "lr": 3.0273655194310246e-06}
{"train_loss": 0.04379906505346298, "global_step": 241284, "epoch": 2711, "lr": 3.027166859243036e-06}
{"train_loss": 0.044042378664016724, "global_step": 241285, "epoch": 2711, "lr": 3.026968205369973e-06}
{"train_loss": 0.05252921208739281, "global_step": 241286, "epoch": 2711, "lr": 3.026769557811865e-06}
{"train_loss": 0.036644019186496735, "global_step": 241287, "epoch": 2711, "lr": 3.0265709165687275e-06}
{"train_loss": 0.05686810240149498, "global_step": 241288, "epoch": 2711, "lr": 3.0263722816405992e-06}
{"train_loss": 0.10157797485589981, "global_step": 241289, "epoch": 2711, "lr": 3.026173653027503e-06}
{"train_loss": 0.06152788922190666, "global_step": 241290, "epoch": 2711, "lr": 3.0259750307294666e-06}
{"train_loss": 0.07089313864707947, "global_step": 241291, "epoch": 2711, "lr": 3.0257764147465173e-06}
{"train_loss": 0.043245457112789154, "global_step": 241292, "epoch": 2711, "lr": 3.025577805078672e-06}
{"train_loss": 0.03207893297076225, "global_step": 241293, "epoch": 2711, "lr": 3.025379201725975e-06}
{"train_loss": 0.018316002562642097, "global_step": 241294, "epoch": 2711, "lr": 3.025180604688432e-06}
{"train_loss": 0.051342226564884186, "global_step": 241295, "epoch": 2711, "lr": 3.024982013966088e-06}
{"train_loss": 0.07876385748386383, "global_step": 241296, "epoch": 2711, "lr": 3.0247834295589585e-06}
{"train_loss": 0.0313115119934082, "global_step": 241297, "epoch": 2711, "lr": 3.0245848514670828e-06}
{"train_loss": 0.02076849155128002, "global_step": 241298, "epoch": 2711, "lr": 3.0243862796904666e-06}
{"train_loss": 0.061899978667497635, "global_step": 241299, "epoch": 2711, "lr": 3.0241877142291596e-06}
{"train_loss": 0.06483612209558487, "global_step": 241300, "epoch": 2711, "lr": 3.0239891550831735e-06}
{"train_loss": 0.024019110947847366, "global_step": 241301, "epoch": 2711, "lr": 3.023790602252541e-06}
{"train_loss": 0.028232784941792488, "global_step": 241302, "epoch": 2711, "lr": 3.0235920557372844e-06}
{"train_loss": 0.07171060889959335, "global_step": 241303, "epoch": 2711, "lr": 3.023393515537437e-06}
{"train_loss": 0.06575353443622589, "global_step": 241304, "epoch": 2711, "lr": 3.023194981653016e-06}
{"train_loss": 0.05137195065617561, "global_step": 241305, "epoch": 2711, "lr": 3.022996454084065e-06}
{"train_loss": 0.04699757695198059, "global_step": 241306, "epoch": 2711, "lr": 3.0227979328305846e-06}
{"train_loss": 0.056458450853824615, "global_step": 241307, "epoch": 2711, "lr": 3.02259941789263e-06}
{"train_loss": 0.011047495529055595, "global_step": 241308, "epoch": 2711, "lr": 3.0224009092702068e-06}
{"train_loss": 0.052102938294410706, "global_step": 241309, "epoch": 2711, "lr": 3.0222024069633535e-06}
{"train_loss": 0.03124314174056053, "global_step": 241310, "epoch": 2711, "lr": 3.0220039109720877e-06}
{"train_loss": 0.042389608919620514, "global_step": 241311, "epoch": 2711, "lr": 3.0218054212964477e-06}
{"train_loss": 0.05879385396838188, "global_step": 241312, "epoch": 2711, "lr": 3.021606937936444e-06}
{"train_loss": 0.06297536194324493, "global_step": 241313, "epoch": 2711, "lr": 3.0214084608921224e-06}
{"train_loss": 0.02101089246571064, "global_step": 241314, "epoch": 2711, "lr": 3.021209990163487e-06}
{"train_loss": 0.021585017442703247, "global_step": 241315, "epoch": 2711, "lr": 3.021011525750583e-06}
{"train_loss": 0.03813199698925018, "global_step": 241316, "epoch": 2711, "lr": 3.0208130676534385e-06}
{"train_loss": 0.06008391082286835, "global_step": 241317, "epoch": 2711, "lr": 3.0206146158720637e-06}
{"train_loss": 0.08592873811721802, "global_step": 241318, "epoch": 2711, "lr": 3.0204161704065037e-06}
{"train_loss": 0.05204998329281807, "global_step": 241319, "epoch": 2711, "lr": 3.020217731256769e-06}
{"train_loss": 0.04379413276910782, "global_step": 241320, "epoch": 2711, "lr": 3.020019298422899e-06}
{"train_loss": 0.028508029878139496, "global_step": 241321, "epoch": 2711, "lr": 3.01982087190491e-06}
{"train_loss": 0.039230190217494965, "global_step": 241322, "epoch": 2711, "lr": 3.019622451702836e-06}
{"train_loss": 0.06395949423313141, "global_step": 241323, "epoch": 2711, "lr": 3.0194240378166983e-06}
{"train_loss": 0.05768290534615517, "global_step": 241324, "epoch": 2711, "lr": 3.0192256302465305e-06}
{"train_loss": 0.016943635419011116, "global_step": 241325, "epoch": 2711, "lr": 3.0190272289923494e-06}
{"train_loss": 0.041135963052511215, "global_step": 241326, "epoch": 2711, "lr": 3.018828834054199e-06}
{"train_loss": 0.04620407894253731, "global_step": 241327, "epoch": 2711, "lr": 3.01863044543208e-06}
{"train_loss": 0.029269520193338394, "global_step": 241328, "epoch": 2711, "lr": 3.0184320631260475e-06}
{"train_loss": 0.04687684029340744, "global_step": 241329, "epoch": 2711, "lr": 3.0182336871361073e-06}
{"train_loss": 0.045080650597810745, "global_step": 241330, "epoch": 2711, "lr": 3.018035317462292e-06}
{"train_loss": 0.041406068950891495, "global_step": 241331, "epoch": 2711, "lr": 3.017836954104636e-06}
{"train_loss": 0.02095792256295681, "global_step": 241332, "epoch": 2711, "lr": 3.017638597063149e-06}
{"train_loss": 0.05122392997145653, "global_step": 241333, "epoch": 2711, "lr": 3.0174402463378715e-06}
{"train_loss": 0.049974225461483, "global_step": 241334, "epoch": 2711, "lr": 3.0172419019288245e-06}
{"train_loss": 0.0742037370800972, "global_step": 241335, "epoch": 2711, "lr": 3.0170435638360417e-06}
{"train_loss": 0.04183211177587509, "global_step": 241336, "epoch": 2711, "lr": 3.01684523205954e-06}
{"train_loss": 0.05857041850686073, "global_step": 241337, "epoch": 2711, "lr": 3.0166469065993573e-06}
{"train_loss": 0.04624854028224945, "global_step": 241338, "epoch": 2711, "lr": 3.016448587455506e-06}
{"train_loss": 0.025327546522021294, "global_step": 241339, "epoch": 2711, "lr": 3.0162502746280297e-06}
{"train_loss": 0.05365217477083206, "global_step": 241340, "epoch": 2711, "lr": 3.0160519681169342e-06}
{"train_loss": 0.08604421466588974, "global_step": 241341, "epoch": 2711, "lr": 3.0158536679222693e-06}
{"train_loss": 0.055757660418748856, "global_step": 241342, "epoch": 2711, "lr": 3.015655374044041e-06}
{"train_loss": 0.04460269585251808, "global_step": 241343, "epoch": 2711, "lr": 3.0154570864822927e-06}
{"train_loss": 0.05206410959362984, "global_step": 241344, "epoch": 2711, "lr": 3.015258805237037e-06}
{"train_loss": 0.058359164744615555, "global_step": 241345, "epoch": 2711, "lr": 3.0150605303083114e-06}
{"train_loss": 0.030923059210181236, "global_step": 241346, "epoch": 2711, "lr": 3.014862261696133e-06}
{"train_loss": 0.04704248532652855, "global_step": 241347, "epoch": 2711, "lr": 3.014663999400541e-06}
{"train_loss": 0.05445079505443573, "global_step": 241348, "epoch": 2711, "lr": 3.014465743421552e-06}
{"train_loss": 0.052627380937337875, "global_step": 241349, "epoch": 2711, "lr": 3.0142674937591984e-06}
{"train_loss": 0.040891241282224655, "global_step": 241350, "epoch": 2711, "lr": 3.014069250413493e-06}
{"train_loss": 0.03433706611394882, "global_step": 241351, "epoch": 2711, "lr": 3.0138710133844783e-06}
{"train_loss": 0.031191153451800346, "global_step": 241352, "epoch": 2711, "lr": 3.0136727826721835e-06}
{"train_loss": 0.0617317333817482, "global_step": 241353, "epoch": 2711, "lr": 3.013474558276619e-06}
{"train_loss": 0.06511209160089493, "global_step": 241354, "epoch": 2711, "lr": 3.013276340197824e-06}
{"train_loss": 0.015119381248950958, "global_step": 241355, "epoch": 2711, "lr": 3.01307812843582e-06}
{"train_loss": 0.02584117464721203, "global_step": 241356, "epoch": 2711, "lr": 3.0128799229906357e-06}
{"train_loss": 0.01532696932554245, "global_step": 241357, "epoch": 2711, "lr": 3.012681723862293e-06}
{"train_loss": 0.06593766808509827, "global_step": 241358, "epoch": 2711, "lr": 3.0124835310508305e-06}
{"train_loss": 0.06483536958694458, "global_step": 241359, "epoch": 2711, "lr": 3.0122853445562594e-06}
{"train_loss": 0.010939310304820538, "global_step": 241360, "epoch": 2711, "lr": 3.012087164378624e-06}
{"train_loss": 0.017758198082447052, "global_step": 241361, "epoch": 2711, "lr": 3.0118889905179306e-06}
{"train_loss": 0.08439353853464127, "global_step": 241362, "epoch": 2711, "lr": 3.0116908229742224e-06}
{"train_loss": 0.03978241607546806, "global_step": 241363, "epoch": 2711, "lr": 3.0114926617475115e-06}
{"train_loss": 0.025469888001680374, "global_step": 241364, "epoch": 2711, "lr": 3.011294506837842e-06}
{"train_loss": 0.03167175501585007, "global_step": 241365, "epoch": 2711, "lr": 3.0110963582452245e-06}
{"train_loss": 0.057490844279527664, "global_step": 241366, "epoch": 2711, "lr": 3.0108982159696984e-06}
{"train_loss": 0.046189801610503976, "global_step": 241367, "epoch": 2711, "lr": 3.010700080011286e-06, "val_loss": 9.454194068908691}
{"train_loss": 0.06308609992265701, "global_step": 241368, "epoch": 2712, "lr": 3.010501950370004e-06}
{"train_loss": 0.08274132013320923, "global_step": 241369, "epoch": 2712, "lr": 3.0103038270458963e-06}
{"train_loss": 0.04235757142305374, "global_step": 241370, "epoch": 2712, "lr": 3.0101057100389685e-06}
{"train_loss": 0.043876778334379196, "global_step": 241371, "epoch": 2712, "lr": 3.009907599349271e-06}
{"train_loss": 0.017789777368307114, "global_step": 241372, "epoch": 2712, "lr": 3.009709494976809e-06}
{"train_loss": 0.12289659678936005, "global_step": 241373, "epoch": 2712, "lr": 3.009511396921627e-06}
{"train_loss": 0.12495099753141403, "global_step": 241374, "epoch": 2712, "lr": 3.0093133051837363e-06}
{"train_loss": 0.044150009751319885, "global_step": 241375, "epoch": 2712, "lr": 3.0091152197631754e-06}
{"train_loss": 0.025471579283475876, "global_step": 241376, "epoch": 2712, "lr": 3.0089171406599615e-06}
{"train_loss": 0.09221881628036499, "global_step": 241377, "epoch": 2712, "lr": 3.008719067874133e-06}
{"train_loss": 0.08648333698511124, "global_step": 241378, "epoch": 2712, "lr": 3.0085210014057063e-06}
{"train_loss": 0.10205581784248352, "global_step": 241379, "epoch": 2712, "lr": 3.0083229412547153e-06}
{"train_loss": 0.08951449394226074, "global_step": 241380, "epoch": 2712, "lr": 3.008124887421171e-06}
{"train_loss": 0.03696895018219948, "global_step": 241381, "epoch": 2712, "lr": 3.007926839905123e-06}
{"train_loss": 0.04047783836722374, "global_step": 241382, "epoch": 2712, "lr": 3.007728798706577e-06}
{"train_loss": 0.052944615483284, "global_step": 241383, "epoch": 2712, "lr": 3.007530763825578e-06}
{"train_loss": 0.024673091247677803, "global_step": 241384, "epoch": 2712, "lr": 3.007332735262136e-06}
{"train_loss": 0.02351895160973072, "global_step": 241385, "epoch": 2712, "lr": 3.0071347130162965e-06}
{"train_loss": 0.07949856668710709, "global_step": 241386, "epoch": 2712, "lr": 3.0069366970880643e-06}
{"train_loss": 0.02455305866897106, "global_step": 241387, "epoch": 2712, "lr": 3.0067386874774727e-06}
{"train_loss": 0.08619898557662964, "global_step": 241388, "epoch": 2712, "lr": 3.0065406841845665e-06}
{"train_loss": 0.06038879603147507, "global_step": 241389, "epoch": 2712, "lr": 3.0063426872093457e-06}
{"train_loss": 0.04284106194972992, "global_step": 241390, "epoch": 2712, "lr": 3.0061446965518604e-06}
{"train_loss": 0.034954044967889786, "global_step": 241391, "epoch": 2712, "lr": 3.0059467122121156e-06}
{"train_loss": 0.02089841663837433, "global_step": 241392, "epoch": 2712, "lr": 3.0057487341901558e-06}
{"train_loss": 0.06588371098041534, "global_step": 241393, "epoch": 2712, "lr": 3.005550762485998e-06}
{"train_loss": 0.026453180238604546, "global_step": 241394, "epoch": 2712, "lr": 3.0053527970996698e-06}
{"train_loss": 0.03606317937374115, "global_step": 241395, "epoch": 2712, "lr": 3.005154838031199e-06}
{"train_loss": 0.05434265360236168, "global_step": 241396, "epoch": 2712, "lr": 3.004956885280619e-06}
{"train_loss": 0.03992289677262306, "global_step": 241397, "epoch": 2712, "lr": 3.00475893884794e-06}
{"train_loss": 0.0418003648519516, "global_step": 241398, "epoch": 2712, "lr": 3.004560998733208e-06}
{"train_loss": 0.09814491122961044, "global_step": 241399, "epoch": 2712, "lr": 3.004363064936433e-06}
{"train_loss": 0.04680661857128143, "global_step": 241400, "epoch": 2712, "lr": 3.004165137457654e-06}
{"train_loss": 0.028481382876634598, "global_step": 241401, "epoch": 2712, "lr": 3.0039672162968825e-06}
{"train_loss": 0.01938985101878643, "global_step": 241402, "epoch": 2712, "lr": 3.003769301454168e-06}
{"train_loss": 0.041093431413173676, "global_step": 241403, "epoch": 2712, "lr": 3.0035713929295106e-06}
{"train_loss": 0.027049964293837547, "global_step": 241404, "epoch": 2712, "lr": 3.0033734907229605e-06}
{"train_loss": 0.017705366015434265, "global_step": 241405, "epoch": 2712, "lr": 3.0031755948345287e-06}
{"train_loss": 0.0640486404299736, "global_step": 241406, "epoch": 2712, "lr": 3.0029777052642482e-06}
{"train_loss": 0.0424935556948185, "global_step": 241407, "epoch": 2712, "lr": 3.002779822012153e-06}
{"train_loss": 0.05426903814077377, "global_step": 241408, "epoch": 2712, "lr": 3.002581945078248e-06}
{"train_loss": 0.06524067372083664, "global_step": 241409, "epoch": 2712, "lr": 3.002384074462583e-06}
{"train_loss": 0.044767316430807114, "global_step": 241410, "epoch": 2712, "lr": 3.002186210165164e-06}
{"train_loss": 0.044164396822452545, "global_step": 241411, "epoch": 2712, "lr": 3.001988352186036e-06}
{"train_loss": 0.033000629395246506, "global_step": 241412, "epoch": 2712, "lr": 3.0017905005252147e-06}
{"train_loss": 0.03764579817652702, "global_step": 241413, "epoch": 2712, "lr": 3.0015926551827333e-06}
{"train_loss": 0.040716949850320816, "global_step": 241414, "epoch": 2712, "lr": 3.001394816158609e-06}
{"train_loss": 0.042793046683073044, "global_step": 241415, "epoch": 2712, "lr": 3.001196983452881e-06}
{"train_loss": 0.044590190052986145, "global_step": 241416, "epoch": 2712, "lr": 3.000999157065565e-06}
{"train_loss": 0.07002253830432892, "global_step": 241417, "epoch": 2712, "lr": 3.0008013369966947e-06}
{"train_loss": 0.04236787185072899, "global_step": 241418, "epoch": 2712, "lr": 3.0006035232462925e-06}
{"train_loss": 0.03285406529903412, "global_step": 241419, "epoch": 2712, "lr": 3.000405715814386e-06}
{"train_loss": 0.07957947999238968, "global_step": 241420, "epoch": 2712, "lr": 3.0002079147009975e-06}
{"train_loss": 0.03487909212708473, "global_step": 241421, "epoch": 2712, "lr": 3.0000101199061716e-06}
{"train_loss": 0.05362923443317413, "global_step": 241422, "epoch": 2712, "lr": 2.9998123314299078e-06}
{"train_loss": 0.0633620023727417, "global_step": 241423, "epoch": 2712, "lr": 2.999614549272245e-06}
{"train_loss": 0.021558893844485283, "global_step": 241424, "epoch": 2712, "lr": 2.9994167734332224e-06}
{"train_loss": 0.03863700106739998, "global_step": 241425, "epoch": 2712, "lr": 2.9992190039128453e-06}
{"train_loss": 0.06624044477939606, "global_step": 241426, "epoch": 2712, "lr": 2.9990212407111582e-06}
{"train_loss": 0.02040739357471466, "global_step": 241427, "epoch": 2712, "lr": 2.998823483828178e-06}
{"train_loss": 0.039760101586580276, "global_step": 241428, "epoch": 2712, "lr": 2.998625733263932e-06}
{"train_loss": 0.056110963225364685, "global_step": 241429, "epoch": 2712, "lr": 2.9984279890184485e-06}
{"train_loss": 0.05839456990361214, "global_step": 241430, "epoch": 2712, "lr": 2.998230251091755e-06}
{"train_loss": 0.029672307893633842, "global_step": 241431, "epoch": 2712, "lr": 2.9980325194838733e-06}
{"train_loss": 0.014484388753771782, "global_step": 241432, "epoch": 2712, "lr": 2.997834794194837e-06}
{"train_loss": 0.07275256514549255, "global_step": 241433, "epoch": 2712, "lr": 2.997637075224663e-06}
{"train_loss": 0.07227034121751785, "global_step": 241434, "epoch": 2712, "lr": 2.99743936257339e-06}
{"train_loss": 0.023023240268230438, "global_step": 241435, "epoch": 2712, "lr": 2.9972416562410344e-06}
{"train_loss": 0.046057481318712234, "global_step": 241436, "epoch": 2712, "lr": 2.99704395622763e-06}
{"train_loss": 0.015925491228699684, "global_step": 241437, "epoch": 2712, "lr": 2.996846262533193e-06}
{"train_loss": 0.05903780087828636, "global_step": 241438, "epoch": 2712, "lr": 2.996648575157768e-06}
{"train_loss": 0.022885730490088463, "global_step": 241439, "epoch": 2712, "lr": 2.99645089410136e-06}
{"train_loss": 0.051201045513153076, "global_step": 241440, "epoch": 2712, "lr": 2.9962532193640148e-06}
{"train_loss": 0.02899167686700821, "global_step": 241441, "epoch": 2712, "lr": 2.9960555509457423e-06}
{"train_loss": 0.03942250832915306, "global_step": 241442, "epoch": 2712, "lr": 2.995857888846587e-06}
{"train_loss": 0.059297699481248856, "global_step": 241443, "epoch": 2712, "lr": 2.9956602330665607e-06}
{"train_loss": 0.026568034663796425, "global_step": 241444, "epoch": 2712, "lr": 2.995462583605696e-06}
{"train_loss": 0.03966782987117767, "global_step": 241445, "epoch": 2712, "lr": 2.995264940464021e-06}
{"train_loss": 0.0428386889398098, "global_step": 241446, "epoch": 2712, "lr": 2.995067303641552e-06}
{"train_loss": 0.034574929624795914, "global_step": 241447, "epoch": 2712, "lr": 2.994869673138334e-06}
{"train_loss": 0.03213346377015114, "global_step": 241448, "epoch": 2712, "lr": 2.994672048954372e-06}
{"train_loss": 0.02104092761874199, "global_step": 241449, "epoch": 2712, "lr": 2.994474431089711e-06}
{"train_loss": 0.03373539820313454, "global_step": 241450, "epoch": 2712, "lr": 2.9942768195443615e-06}
{"train_loss": 0.05985024571418762, "global_step": 241451, "epoch": 2712, "lr": 2.9940792143183684e-06}
{"train_loss": 0.030646512284874916, "global_step": 241452, "epoch": 2712, "lr": 2.993881615411742e-06}
{"train_loss": 0.03517771139740944, "global_step": 241453, "epoch": 2712, "lr": 2.9936840228245223e-06}
{"train_loss": 0.04614141583442688, "global_step": 241454, "epoch": 2712, "lr": 2.9934864365567193e-06}
{"train_loss": 0.07819820195436478, "global_step": 241455, "epoch": 2712, "lr": 2.993288856608373e-06}
{"train_loss": 0.04784193009221822, "global_step": 241456, "epoch": 2712, "lr": 2.9930912829795044e-06, "val_loss": 9.517867088317871}
{"train_loss": 0.04831249639391899, "global_step": 241457, "epoch": 2713, "lr": 2.9928937156701474e-06}
{"train_loss": 0.007802871055901051, "global_step": 241458, "epoch": 2713, "lr": 2.992696154680319e-06}
{"train_loss": 0.05994085222482681, "global_step": 241459, "epoch": 2713, "lr": 2.9924986000100464e-06}
{"train_loss": 0.02522248588502407, "global_step": 241460, "epoch": 2713, "lr": 2.992301051659363e-06}
{"train_loss": 0.027936045080423355, "global_step": 241461, "epoch": 2713, "lr": 2.9921035096282914e-06}
{"train_loss": 0.07091529667377472, "global_step": 241462, "epoch": 2713, "lr": 2.991905973916864e-06}
{"train_loss": 0.02671547792851925, "global_step": 241463, "epoch": 2713, "lr": 2.9917084445250933e-06}
{"train_loss": 0.04430530592799187, "global_step": 241464, "epoch": 2713, "lr": 2.9915109214530167e-06}
{"train_loss": 0.08511242270469666, "global_step": 241465, "epoch": 2713, "lr": 2.9913134047006576e-06}
{"train_loss": 0.05420365184545517, "global_step": 241466, "epoch": 2713, "lr": 2.9911158942680485e-06}
{"train_loss": 0.06465957313776016, "global_step": 241467, "epoch": 2713, "lr": 2.990918390155206e-06}
{"train_loss": 0.06013081595301628, "global_step": 241468, "epoch": 2713, "lr": 2.99072089236217e-06}
{"train_loss": 0.03917640075087547, "global_step": 241469, "epoch": 2713, "lr": 2.9905234008889447e-06}
{"train_loss": 0.045828986912965775, "global_step": 241470, "epoch": 2713, "lr": 2.9903259157355868e-06}
{"train_loss": 0.048729658126831055, "global_step": 241471, "epoch": 2713, "lr": 2.99012843690209e-06}
{"train_loss": 0.05778747424483299, "global_step": 241472, "epoch": 2713, "lr": 2.98993096438851e-06}
{"train_loss": 0.08221635967493057, "global_step": 241473, "epoch": 2713, "lr": 2.989733498194852e-06}
{"train_loss": 0.044154442846775055, "global_step": 241474, "epoch": 2713, "lr": 2.9895360383211614e-06}
{"train_loss": 0.012538810260593891, "global_step": 241475, "epoch": 2713, "lr": 2.9893385847674425e-06}
{"train_loss": 0.023181425407528877, "global_step": 241476, "epoch": 2713, "lr": 2.9891411375337463e-06}
{"train_loss": 0.046601682901382446, "global_step": 241477, "epoch": 2713, "lr": 2.988943696620078e-06}
{"train_loss": 0.04611250013113022, "global_step": 241478, "epoch": 2713, "lr": 2.988746262026482e-06}
{"train_loss": 0.06342997401952744, "global_step": 241479, "epoch": 2713, "lr": 2.9885488337529634e-06}
{"train_loss": 0.07115820050239563, "global_step": 241480, "epoch": 2713, "lr": 2.9883514117995725e-06}
{"train_loss": 0.10581991076469421, "global_step": 241481, "epoch": 2713, "lr": 2.988153996166321e-06}
{"train_loss": 0.026804693043231964, "global_step": 241482, "epoch": 2713, "lr": 2.987956586853241e-06}
{"train_loss": 0.07236101478338242, "global_step": 241483, "epoch": 2713, "lr": 2.9877591838603613e-06}
{"train_loss": 0.037810005247592926, "global_step": 241484, "epoch": 2713, "lr": 2.9875617871876927e-06}
{"train_loss": 0.02037566900253296, "global_step": 241485, "epoch": 2713, "lr": 2.987364396835285e-06}
{"train_loss": 0.021656228229403496, "global_step": 241486, "epoch": 2713, "lr": 2.9871670128031437e-06}
{"train_loss": 0.07661430537700653, "global_step": 241487, "epoch": 2713, "lr": 2.986969635091308e-06}
{"train_loss": 0.05798700824379921, "global_step": 241488, "epoch": 2713, "lr": 2.9867722636997998e-06}
{"train_loss": 0.09072840213775635, "global_step": 241489, "epoch": 2713, "lr": 2.9865748986286523e-06}
{"train_loss": 0.06544148176908493, "global_step": 241490, "epoch": 2713, "lr": 2.986377539877877e-06}
{"train_loss": 0.031240742653608322, "global_step": 241491, "epoch": 2713, "lr": 2.9861801874475236e-06}
{"train_loss": 0.042497165501117706, "global_step": 241492, "epoch": 2713, "lr": 2.9859828413375924e-06}
{"train_loss": 0.0451483279466629, "global_step": 241493, "epoch": 2713, "lr": 2.985785501548133e-06}
{"train_loss": 0.07161650061607361, "global_step": 241494, "epoch": 2713, "lr": 2.9855881680791566e-06}
{"train_loss": 0.05617891252040863, "global_step": 241495, "epoch": 2713, "lr": 2.9853908409306906e-06}
{"train_loss": 0.05985552817583084, "global_step": 241496, "epoch": 2713, "lr": 2.9851935201027748e-06}
{"train_loss": 0.03745459392666817, "global_step": 241497, "epoch": 2713, "lr": 2.9849962055954196e-06}
{"train_loss": 0.06638889014720917, "global_step": 241498, "epoch": 2713, "lr": 2.9847988974086637e-06}
{"train_loss": 0.038345493376255035, "global_step": 241499, "epoch": 2713, "lr": 2.9846015955425243e-06}
{"train_loss": 0.08246168494224548, "global_step": 241500, "epoch": 2713, "lr": 2.9844042999970346e-06}
{"train_loss": 0.04139366000890732, "global_step": 241501, "epoch": 2713, "lr": 2.9842070107722166e-06}
{"train_loss": 0.019501667469739914, "global_step": 241502, "epoch": 2713, "lr": 2.9840097278681033e-06}
{"train_loss": 0.03209257125854492, "global_step": 241503, "epoch": 2713, "lr": 2.983812451284712e-06}
{"train_loss": 0.060400914400815964, "global_step": 241504, "epoch": 2713, "lr": 2.9836151810220813e-06}
{"train_loss": 0.04831671714782715, "global_step": 241505, "epoch": 2713, "lr": 2.9834179170802223e-06}
{"train_loss": 0.018099457025527954, "global_step": 241506, "epoch": 2713, "lr": 2.9832206594591738e-06}
{"train_loss": 0.03153090924024582, "global_step": 241507, "epoch": 2713, "lr": 2.9830234081589583e-06}
{"train_loss": 0.051439832895994186, "global_step": 241508, "epoch": 2713, "lr": 2.982826163179603e-06}
{"train_loss": 0.021616138517856598, "global_step": 241509, "epoch": 2713, "lr": 2.98262892452113e-06}
{"train_loss": 0.04152554273605347, "global_step": 241510, "epoch": 2713, "lr": 2.982431692183574e-06}
{"train_loss": 0.009669429622590542, "global_step": 241511, "epoch": 2713, "lr": 2.98223446616695e-06}
{"train_loss": 0.02016998454928398, "global_step": 241512, "epoch": 2713, "lr": 2.9820372464712974e-06}
{"train_loss": 0.039584577083587646, "global_step": 241513, "epoch": 2713, "lr": 2.9818400330966336e-06}
{"train_loss": 0.03536521643400192, "global_step": 241514, "epoch": 2713, "lr": 2.9816428260429853e-06}
{"train_loss": 0.03581102192401886, "global_step": 241515, "epoch": 2713, "lr": 2.9814456253103916e-06}
{"train_loss": 0.014902069233357906, "global_step": 241516, "epoch": 2713, "lr": 2.9812484308988697e-06}
{"train_loss": 0.02146470546722412, "global_step": 241517, "epoch": 2713, "lr": 2.981051242808436e-06}
{"train_loss": 0.046356428414583206, "global_step": 241518, "epoch": 2713, "lr": 2.9808540610391344e-06}
{"train_loss": 0.025550493970513344, "global_step": 241519, "epoch": 2713, "lr": 2.9806568855909824e-06}
{"train_loss": 0.014889383688569069, "global_step": 241520, "epoch": 2713, "lr": 2.980459716464007e-06}
{"train_loss": 0.0198559258133173, "global_step": 241521, "epoch": 2713, "lr": 2.980262553658236e-06}
{"train_loss": 0.028620710596442223, "global_step": 241522, "epoch": 2713, "lr": 2.980065397173698e-06}
{"train_loss": 0.02961120940744877, "global_step": 241523, "epoch": 2713, "lr": 2.9798682470104144e-06}
{"train_loss": 0.03903766721487045, "global_step": 241524, "epoch": 2713, "lr": 2.9796711031684134e-06}
{"train_loss": 0.07082577049732208, "global_step": 241525, "epoch": 2713, "lr": 2.9794739656477277e-06}
{"train_loss": 0.053960852324962616, "global_step": 241526, "epoch": 2713, "lr": 2.979276834448369e-06}
{"train_loss": 0.08582542091608047, "global_step": 241527, "epoch": 2713, "lr": 2.9790797095703814e-06}
{"train_loss": 0.043346285820007324, "global_step": 241528, "epoch": 2713, "lr": 2.9788825910137817e-06}
{"train_loss": 0.04975370690226555, "global_step": 241529, "epoch": 2713, "lr": 2.9786854787785978e-06}
{"train_loss": 0.05175027623772621, "global_step": 241530, "epoch": 2713, "lr": 2.9784883728648517e-06}
{"train_loss": 0.05768265202641487, "global_step": 241531, "epoch": 2713, "lr": 2.9782912732725764e-06}
{"train_loss": 0.08263349533081055, "global_step": 241532, "epoch": 2713, "lr": 2.978094180001806e-06}
{"train_loss": 0.03897211700677872, "global_step": 241533, "epoch": 2713, "lr": 2.977897093052545e-06}
{"train_loss": 0.06063029542565346, "global_step": 241534, "epoch": 2713, "lr": 2.9777000124248445e-06}
{"train_loss": 0.028175801038742065, "global_step": 241535, "epoch": 2713, "lr": 2.9775029381187093e-06}
{"train_loss": 0.057655662298202515, "global_step": 241536, "epoch": 2713, "lr": 2.9773058701341838e-06}
{"train_loss": 0.05788131803274155, "global_step": 241537, "epoch": 2713, "lr": 2.9771088084712794e-06}
{"train_loss": 0.042446088045835495, "global_step": 241538, "epoch": 2713, "lr": 2.976911753130035e-06}
{"train_loss": 0.05429784953594208, "global_step": 241539, "epoch": 2713, "lr": 2.976714704110467e-06}
{"train_loss": 0.04546548053622246, "global_step": 241540, "epoch": 2713, "lr": 2.976517661412609e-06}
{"train_loss": 0.13550832867622375, "global_step": 241541, "epoch": 2713, "lr": 2.976320625036483e-06}
{"train_loss": 0.056571997702121735, "global_step": 241542, "epoch": 2713, "lr": 2.976123594982122e-06}
{"train_loss": 0.06714218109846115, "global_step": 241543, "epoch": 2713, "lr": 2.975926571249543e-06}
{"train_loss": 0.06857845187187195, "global_step": 241544, "epoch": 2713, "lr": 2.975729553838785e-06}
{"train_loss": 0.04779934525238664, "global_step": 241545, "epoch": 2713, "lr": 2.9755325427498594e-06, "val_loss": 9.489156723022461}
{"train_loss": 0.05071476846933365, "global_step": 241546, "epoch": 2714, "lr": 2.975335537982804e-06}
{"train_loss": 0.04495611414313316, "global_step": 241547, "epoch": 2714, "lr": 2.975138539537636e-06}
{"train_loss": 0.040600407868623734, "global_step": 241548, "epoch": 2714, "lr": 2.974941547414395e-06}
{"train_loss": 0.029049308970570564, "global_step": 241549, "epoch": 2714, "lr": 2.9747445616130963e-06}
{"train_loss": 0.060689810663461685, "global_step": 241550, "epoch": 2714, "lr": 2.974547582133763e-06}
{"train_loss": 0.04902214556932449, "global_step": 241551, "epoch": 2714, "lr": 2.974350608976445e-06}
{"train_loss": 0.04562576115131378, "global_step": 241552, "epoch": 2714, "lr": 2.9741536421411365e-06}
{"train_loss": 0.044444866478443146, "global_step": 241553, "epoch": 2714, "lr": 2.973956681627893e-06}
{"train_loss": 0.07553113251924515, "global_step": 241554, "epoch": 2714, "lr": 2.97375972743672e-06}
{"train_loss": 0.04354066029191017, "global_step": 241555, "epoch": 2714, "lr": 2.973562779567657e-06}
{"train_loss": 0.03484964370727539, "global_step": 241556, "epoch": 2714, "lr": 2.9733658380207254e-06}
{"train_loss": 0.04647698998451233, "global_step": 241557, "epoch": 2714, "lr": 2.973168902795942e-06}
{"train_loss": 0.013094881549477577, "global_step": 241558, "epoch": 2714, "lr": 2.9729719738933513e-06}
{"train_loss": 0.03661591187119484, "global_step": 241559, "epoch": 2714, "lr": 2.9727750513129706e-06}
{"train_loss": 0.022333504632115364, "global_step": 241560, "epoch": 2714, "lr": 2.9725781350548264e-06}
{"train_loss": 0.02347898855805397, "global_step": 241561, "epoch": 2714, "lr": 2.9723812251189477e-06}
{"train_loss": 0.030620133504271507, "global_step": 241562, "epoch": 2714, "lr": 2.972184321505356e-06}
{"train_loss": 0.08090676367282867, "global_step": 241563, "epoch": 2714, "lr": 2.9719874242140844e-06}
{"train_loss": 0.04245707765221596, "global_step": 241564, "epoch": 2714, "lr": 2.97179053324515e-06}
{"train_loss": 0.015347146429121494, "global_step": 241565, "epoch": 2714, "lr": 2.9715936485985917e-06}
{"train_loss": 0.03927481174468994, "global_step": 241566, "epoch": 2714, "lr": 2.9713967702744204e-06}
{"train_loss": 0.09456882625818253, "global_step": 241567, "epoch": 2714, "lr": 2.971199898272681e-06}
{"train_loss": 0.03983562812209129, "global_step": 241568, "epoch": 2714, "lr": 2.9710030325933834e-06}
{"train_loss": 0.06345351785421371, "global_step": 241569, "epoch": 2714, "lr": 2.970806173236562e-06}
{"train_loss": 0.043975308537483215, "global_step": 241570, "epoch": 2714, "lr": 2.97060932020225e-06}
{"train_loss": 0.06081695109605789, "global_step": 241571, "epoch": 2714, "lr": 2.970412473490458e-06}
{"train_loss": 0.0640067532658577, "global_step": 241572, "epoch": 2714, "lr": 2.9702156331012253e-06}
{"train_loss": 0.04395173862576485, "global_step": 241573, "epoch": 2714, "lr": 2.9700187990345683e-06}
{"train_loss": 0.04070621728897095, "global_step": 241574, "epoch": 2714, "lr": 2.9698219712905316e-06}
{"train_loss": 0.060894377529621124, "global_step": 241575, "epoch": 2714, "lr": 2.969625149869115e-06}
{"train_loss": 0.040412452071905136, "global_step": 241576, "epoch": 2714, "lr": 2.969428334770369e-06}
{"train_loss": 0.052381403744220734, "global_step": 241577, "epoch": 2714, "lr": 2.969231525994304e-06}
{"train_loss": 0.051374468952417374, "global_step": 241578, "epoch": 2714, "lr": 2.969034723540959e-06}
{"train_loss": 0.01567941904067993, "global_step": 241579, "epoch": 2714, "lr": 2.968837927410345e-06}
{"train_loss": 0.06531106680631638, "global_step": 241580, "epoch": 2714, "lr": 2.968641137602507e-06}
{"train_loss": 0.06492707878351212, "global_step": 241581, "epoch": 2714, "lr": 2.9684443541174555e-06}
{"train_loss": 0.05420093610882759, "global_step": 241582, "epoch": 2714, "lr": 2.968247576955224e-06}
{"train_loss": 0.029217159375548363, "global_step": 241583, "epoch": 2714, "lr": 2.9680508061158407e-06}
{"train_loss": 0.08095353841781616, "global_step": 241584, "epoch": 2714, "lr": 2.9678540415993326e-06}
{"train_loss": 0.03230679780244827, "global_step": 241585, "epoch": 2714, "lr": 2.9676572834057116e-06}
{"train_loss": 0.062064867466688156, "global_step": 241586, "epoch": 2714, "lr": 2.967460531535021e-06}
{"train_loss": 0.09117812663316727, "global_step": 241587, "epoch": 2714, "lr": 2.96726378598729e-06}
{"train_loss": 0.04701479151844978, "global_step": 241588, "epoch": 2714, "lr": 2.9670670467625283e-06}
{"train_loss": 0.03224876895546913, "global_step": 241589, "epoch": 2714, "lr": 2.9668703138607813e-06}
{"train_loss": 0.032940223813056946, "global_step": 241590, "epoch": 2714, "lr": 2.966673587282054e-06}
{"train_loss": 0.06315053999423981, "global_step": 241591, "epoch": 2714, "lr": 2.966476867026391e-06}
{"train_loss": 0.06457092612981796, "global_step": 241592, "epoch": 2714, "lr": 2.966280153093809e-06}
{"train_loss": 0.033083245158195496, "global_step": 241593, "epoch": 2714, "lr": 2.9660834454843413e-06}
{"train_loss": 0.059382542967796326, "global_step": 241594, "epoch": 2714, "lr": 2.9658867441980097e-06}
{"train_loss": 0.08237599581480026, "global_step": 241595, "epoch": 2714, "lr": 2.965690049234837e-06}
{"train_loss": 0.027543406933546066, "global_step": 241596, "epoch": 2714, "lr": 2.9654933605948567e-06}
{"train_loss": 0.09380650520324707, "global_step": 241597, "epoch": 2714, "lr": 2.96529667827809e-06}
{"train_loss": 0.034550201147794724, "global_step": 241598, "epoch": 2714, "lr": 2.965100002284571e-06}
{"train_loss": 0.06110473722219467, "global_step": 241599, "epoch": 2714, "lr": 2.9649033326143217e-06}
{"train_loss": 0.037104640156030655, "global_step": 241600, "epoch": 2714, "lr": 2.9647066692673644e-06}
{"train_loss": 0.025945790112018585, "global_step": 241601, "epoch": 2714, "lr": 2.964510012243732e-06}
{"train_loss": 0.057122137397527695, "global_step": 241602, "epoch": 2714, "lr": 2.964313361543436e-06}
{"train_loss": 0.035320717841386795, "global_step": 241603, "epoch": 2714, "lr": 2.964116717166532e-06}
{"train_loss": 0.0354076623916626, "global_step": 241604, "epoch": 2714, "lr": 2.963920079113014e-06}
{"train_loss": 0.03333968296647072, "global_step": 241605, "epoch": 2714, "lr": 2.963723447382927e-06}
{"train_loss": 0.03500932455062866, "global_step": 241606, "epoch": 2714, "lr": 2.9635268219763036e-06}
{"train_loss": 0.025429844856262207, "global_step": 241607, "epoch": 2714, "lr": 2.96333020289315e-06}
{"train_loss": 0.057924289256334305, "global_step": 241608, "epoch": 2714, "lr": 2.96313359013351e-06}
{"train_loss": 0.06687638908624649, "global_step": 241609, "epoch": 2714, "lr": 2.9629369836973953e-06}
{"train_loss": 0.028957506641745567, "global_step": 241610, "epoch": 2714, "lr": 2.96274038358485e-06}
{"train_loss": 0.04006271809339523, "global_step": 241611, "epoch": 2714, "lr": 2.9625437897958852e-06}
{"train_loss": 0.0655154138803482, "global_step": 241612, "epoch": 2714, "lr": 2.9623472023305397e-06}
{"train_loss": 0.02490200661122799, "global_step": 241613, "epoch": 2714, "lr": 2.962150621188825e-06}
{"train_loss": 0.06148069351911545, "global_step": 241614, "epoch": 2714, "lr": 2.961954046370785e-06}
{"train_loss": 0.055168695747852325, "global_step": 241615, "epoch": 2714, "lr": 2.961757477876426e-06}
{"train_loss": 0.04660441726446152, "global_step": 241616, "epoch": 2714, "lr": 2.9615609157057965e-06}
{"train_loss": 0.030863389372825623, "global_step": 241617, "epoch": 2714, "lr": 2.9613643598589034e-06}
{"train_loss": 0.04093122109770775, "global_step": 241618, "epoch": 2714, "lr": 2.9611678103357855e-06}
{"train_loss": 0.05117882788181305, "global_step": 241619, "epoch": 2714, "lr": 2.9609712671364644e-06}
{"train_loss": 0.04802515730261803, "global_step": 241620, "epoch": 2714, "lr": 2.960774730260968e-06}
{"train_loss": 0.0403134785592556, "global_step": 241621, "epoch": 2714, "lr": 2.9605781997093184e-06}
{"train_loss": 0.05036243796348572, "global_step": 241622, "epoch": 2714, "lr": 2.960381675481544e-06}
{"train_loss": 0.0945359617471695, "global_step": 241623, "epoch": 2714, "lr": 2.9601851575776827e-06}
{"train_loss": 0.048767443746328354, "global_step": 241624, "epoch": 2714, "lr": 2.959988645997741e-06}
{"train_loss": 0.05600732937455177, "global_step": 241625, "epoch": 2714, "lr": 2.9597921407417627e-06}
{"train_loss": 0.05694253370165825, "global_step": 241626, "epoch": 2714, "lr": 2.959595641809765e-06}
{"train_loss": 0.04269733652472496, "global_step": 241627, "epoch": 2714, "lr": 2.9593991492017804e-06}
{"train_loss": 0.03350890427827835, "global_step": 241628, "epoch": 2714, "lr": 2.9592026629178204e-06}
{"train_loss": 0.07099646329879761, "global_step": 241629, "epoch": 2714, "lr": 2.959006182957935e-06}
{"train_loss": 0.03382223844528198, "global_step": 241630, "epoch": 2714, "lr": 2.9588097093221244e-06}
{"train_loss": 0.038359060883522034, "global_step": 241631, "epoch": 2714, "lr": 2.958613242010444e-06}
{"train_loss": 0.053034551441669464, "global_step": 241632, "epoch": 2714, "lr": 2.9584167810228937e-06}
{"train_loss": 0.024131903424859047, "global_step": 241633, "epoch": 2714, "lr": 2.9582203263595178e-06}
{"train_loss": 0.04853049143520969, "global_step": 241634, "epoch": 2714, "lr": 2.958023878020333e-06, "val_loss": 9.50917911529541}
{"train_loss": 0.03678808733820915, "global_step": 241635, "epoch": 2715, "lr": 2.9578274360053616e-06}
{"train_loss": 0.031727153807878494, "global_step": 241636, "epoch": 2715, "lr": 2.957631000314648e-06}
{"train_loss": 0.026922205463051796, "global_step": 241637, "epoch": 2715, "lr": 2.9574345709481976e-06}
{"train_loss": 0.10857315361499786, "global_step": 241638, "epoch": 2715, "lr": 2.9572381479060553e-06}
{"train_loss": 0.06283801048994064, "global_step": 241639, "epoch": 2715, "lr": 2.957041731188237e-06}
{"train_loss": 0.022663936018943787, "global_step": 241640, "epoch": 2715, "lr": 2.9568453207947654e-06}
{"train_loss": 0.061814237385988235, "global_step": 241641, "epoch": 2715, "lr": 2.956648916725674e-06}
{"train_loss": 0.036091890186071396, "global_step": 241642, "epoch": 2715, "lr": 2.9564525189809953e-06}
{"train_loss": 0.04154153913259506, "global_step": 241643, "epoch": 2715, "lr": 2.9562561275607415e-06}
{"train_loss": 0.06362679600715637, "global_step": 241644, "epoch": 2715, "lr": 2.9560597424649506e-06}
{"train_loss": 0.02557113580405712, "global_step": 241645, "epoch": 2715, "lr": 2.9558633636936337e-06}
{"train_loss": 0.0265213530510664, "global_step": 241646, "epoch": 2715, "lr": 2.9556669912468416e-06}
{"train_loss": 0.06529109925031662, "global_step": 241647, "epoch": 2715, "lr": 2.9554706251245735e-06}
{"train_loss": 0.0660901591181755, "global_step": 241648, "epoch": 2715, "lr": 2.9552742653268794e-06}
{"train_loss": 0.05479304865002632, "global_step": 241649, "epoch": 2715, "lr": 2.955077911853765e-06}
{"train_loss": 0.04472912847995758, "global_step": 241650, "epoch": 2715, "lr": 2.954881564705275e-06}
{"train_loss": 0.03372113034129143, "global_step": 241651, "epoch": 2715, "lr": 2.954685223881426e-06}
{"train_loss": 0.04727514088153839, "global_step": 241652, "epoch": 2715, "lr": 2.954488889382251e-06}
{"train_loss": 0.10152694582939148, "global_step": 241653, "epoch": 2715, "lr": 2.954292561207761e-06}
{"train_loss": 0.019557621330022812, "global_step": 241654, "epoch": 2715, "lr": 2.954096239358001e-06}
{"train_loss": 0.02622019127011299, "global_step": 241655, "epoch": 2715, "lr": 2.953899923832981e-06}
{"train_loss": 0.047837842255830765, "global_step": 241656, "epoch": 2715, "lr": 2.9537036146327467e-06}
{"train_loss": 0.04201987758278847, "global_step": 241657, "epoch": 2715, "lr": 2.953507311757303e-06}
{"train_loss": 0.0634470134973526, "global_step": 241658, "epoch": 2715, "lr": 2.953311015206689e-06}
{"train_loss": 0.048900723457336426, "global_step": 241659, "epoch": 2715, "lr": 2.9531147249809376e-06}
{"train_loss": 0.05110640823841095, "global_step": 241660, "epoch": 2715, "lr": 2.952918441080055e-06}
{"train_loss": 0.021387603133916855, "global_step": 241661, "epoch": 2715, "lr": 2.9527221635040902e-06}
{"train_loss": 0.06669162958860397, "global_step": 241662, "epoch": 2715, "lr": 2.95252589225305e-06}
{"train_loss": 0.013615759089589119, "global_step": 241663, "epoch": 2715, "lr": 2.9523296273269773e-06}
{"train_loss": 0.055196188390254974, "global_step": 241664, "epoch": 2715, "lr": 2.952133368725879e-06}
{"train_loss": 0.07687293738126755, "global_step": 241665, "epoch": 2715, "lr": 2.9519371164498046e-06}
{"train_loss": 0.06899351626634598, "global_step": 241666, "epoch": 2715, "lr": 2.951740870498759e-06}
{"train_loss": 0.044969938695430756, "global_step": 241667, "epoch": 2715, "lr": 2.9515446308727878e-06}
{"train_loss": 0.05640162527561188, "global_step": 241668, "epoch": 2715, "lr": 2.9513483975719015e-06}
{"train_loss": 0.041766706854104996, "global_step": 241669, "epoch": 2715, "lr": 2.9511521705961385e-06}
{"train_loss": 0.020923521369695663, "global_step": 241670, "epoch": 2715, "lr": 2.9509559499455107e-06}
{"train_loss": 0.11258798837661743, "global_step": 241671, "epoch": 2715, "lr": 2.9507597356200623e-06}
{"train_loss": 0.04177658632397652, "global_step": 241672, "epoch": 2715, "lr": 2.950563527619815e-06}
{"train_loss": 0.025863556191325188, "global_step": 241673, "epoch": 2715, "lr": 2.950367325944775e-06}
{"train_loss": 0.0668003112077713, "global_step": 241674, "epoch": 2715, "lr": 2.9501711305949977e-06}
{"train_loss": 0.04193609580397606, "global_step": 241675, "epoch": 2715, "lr": 2.949974941570488e-06}
{"train_loss": 0.02493361197412014, "global_step": 241676, "epoch": 2715, "lr": 2.949778758871291e-06}
{"train_loss": 0.04027830809354782, "global_step": 241677, "epoch": 2715, "lr": 2.9495825824974123e-06}
{"train_loss": 0.06795709580183029, "global_step": 241678, "epoch": 2715, "lr": 2.9493864124488956e-06}
{"train_loss": 0.04501548036932945, "global_step": 241679, "epoch": 2715, "lr": 2.949190248725753e-06}
{"train_loss": 0.03182341530919075, "global_step": 241680, "epoch": 2715, "lr": 2.948994091328028e-06}
{"train_loss": 0.020143263041973114, "global_step": 241681, "epoch": 2715, "lr": 2.9487979402557262e-06}
{"train_loss": 0.04624786600470543, "global_step": 241682, "epoch": 2715, "lr": 2.948601795508893e-06}
{"train_loss": 0.025941699743270874, "global_step": 241683, "epoch": 2715, "lr": 2.948405657087544e-06}
{"train_loss": 0.03159557655453682, "global_step": 241684, "epoch": 2715, "lr": 2.948209524991713e-06}
{"train_loss": 0.05415884032845497, "global_step": 241685, "epoch": 2715, "lr": 2.948013399221411e-06}
{"train_loss": 0.0503452867269516, "global_step": 241686, "epoch": 2715, "lr": 2.9478172797766878e-06}
{"train_loss": 0.04764845594763756, "global_step": 241687, "epoch": 2715, "lr": 2.9476211666575494e-06}
{"train_loss": 0.05779112130403519, "global_step": 241688, "epoch": 2715, "lr": 2.947425059864034e-06}
{"train_loss": 0.03235840052366257, "global_step": 241689, "epoch": 2715, "lr": 2.947228959396159e-06}
{"train_loss": 0.061825208365917206, "global_step": 241690, "epoch": 2715, "lr": 2.9470328652539625e-06}
{"train_loss": 0.03218087926506996, "global_step": 241691, "epoch": 2715, "lr": 2.9468367774374563e-06}
{"train_loss": 0.027223339304327965, "global_step": 241692, "epoch": 2715, "lr": 2.9466406959466785e-06}
{"train_loss": 0.046669647097587585, "global_step": 241693, "epoch": 2715, "lr": 2.9464446207816466e-06}
{"train_loss": 0.08447033911943436, "global_step": 241694, "epoch": 2715, "lr": 2.946248551942399e-06}
{"train_loss": 0.03427690640091896, "global_step": 241695, "epoch": 2715, "lr": 2.9460524894289463e-06}
{"train_loss": 0.0772179365158081, "global_step": 241696, "epoch": 2715, "lr": 2.9458564332413284e-06}
{"train_loss": 0.12814082205295563, "global_step": 241697, "epoch": 2715, "lr": 2.945660383379567e-06}
{"train_loss": 0.03049023076891899, "global_step": 241698, "epoch": 2715, "lr": 2.945464339843684e-06}
{"train_loss": 0.07730233669281006, "global_step": 241699, "epoch": 2715, "lr": 2.945268302633719e-06}
{"train_loss": 0.06421009451150894, "global_step": 241700, "epoch": 2715, "lr": 2.945072271749677e-06}
{"train_loss": 0.059785742312669754, "global_step": 241701, "epoch": 2715, "lr": 2.944876247191608e-06}
{"train_loss": 0.028371382504701614, "global_step": 241702, "epoch": 2715, "lr": 2.944680228959518e-06}
{"train_loss": 0.023888081312179565, "global_step": 241703, "epoch": 2715, "lr": 2.9444842170534505e-06}
{"train_loss": 0.04412404075264931, "global_step": 241704, "epoch": 2715, "lr": 2.9442882114734173e-06}
{"train_loss": 0.032779667526483536, "global_step": 241705, "epoch": 2715, "lr": 2.944092212219457e-06}
{"train_loss": 0.03770073130726814, "global_step": 241706, "epoch": 2715, "lr": 2.9438962192915808e-06}
{"train_loss": 0.037236426025629044, "global_step": 241707, "epoch": 2715, "lr": 2.9437002326898332e-06}
{"train_loss": 0.04870053753256798, "global_step": 241708, "epoch": 2715, "lr": 2.9435042524142254e-06}
{"train_loss": 0.08488429337739944, "global_step": 241709, "epoch": 2715, "lr": 2.9433082784647958e-06}
{"train_loss": 0.015619037672877312, "global_step": 241710, "epoch": 2715, "lr": 2.9431123108415613e-06}
{"train_loss": 0.012406088411808014, "global_step": 241711, "epoch": 2715, "lr": 2.94291634954455e-06}
{"train_loss": 0.03828102722764015, "global_step": 241712, "epoch": 2715, "lr": 2.942720394573795e-06}
{"train_loss": 0.0642913207411766, "global_step": 241713, "epoch": 2715, "lr": 2.9425244459293123e-06}
{"train_loss": 0.09381438046693802, "global_step": 241714, "epoch": 2715, "lr": 2.9423285036111414e-06}
{"train_loss": 0.031574420630931854, "global_step": 241715, "epoch": 2715, "lr": 2.942132567619288e-06}
{"train_loss": 0.04691005125641823, "global_step": 241716, "epoch": 2715, "lr": 2.9419366379538014e-06}
{"train_loss": 0.014348536729812622, "global_step": 241717, "epoch": 2715, "lr": 2.9417407146146937e-06}
{"train_loss": 0.025548607110977173, "global_step": 241718, "epoch": 2715, "lr": 2.9415447976020027e-06}
{"train_loss": 0.0666268914937973, "global_step": 241719, "epoch": 2715, "lr": 2.9413488869157347e-06}
{"train_loss": 0.027000052854418755, "global_step": 241720, "epoch": 2715, "lr": 2.9411529825559393e-06}
{"train_loss": 0.04926498979330063, "global_step": 241721, "epoch": 2715, "lr": 2.9409570845226274e-06}
{"train_loss": 0.05510652810335159, "global_step": 241722, "epoch": 2715, "lr": 2.940761192815833e-06}
{"train_loss": 0.04811190146157581, "global_step": 241723, "epoch": 2715, "lr": 2.9405653074355777e-06, "val_loss": 9.404027938842773, "train_action_mse_error": 14.679418563842773}
{"train_loss": 0.05939098447561264, "global_step": 241724, "epoch": 2716, "lr": 2.940369428381895e-06}
{"train_loss": 0.03333065286278725, "global_step": 241725, "epoch": 2716, "lr": 2.940173555654796e-06}
{"train_loss": 0.04284079000353813, "global_step": 241726, "epoch": 2716, "lr": 2.939977689254325e-06}
{"train_loss": 0.032192427664995193, "global_step": 241727, "epoch": 2716, "lr": 2.9397818291804937e-06}
{"train_loss": 0.08128746598958969, "global_step": 241728, "epoch": 2716, "lr": 2.9395859754333453e-06}
{"train_loss": 0.02791403792798519, "global_step": 241729, "epoch": 2716, "lr": 2.9393901280128867e-06}
{"train_loss": 0.03243548050522804, "global_step": 241730, "epoch": 2716, "lr": 2.939194286919161e-06}
{"train_loss": 0.03839269280433655, "global_step": 241731, "epoch": 2716, "lr": 2.938998452152175e-06}
{"train_loss": 0.02533671073615551, "global_step": 241732, "epoch": 2716, "lr": 2.9388026237119727e-06}
{"train_loss": 0.04447724670171738, "global_step": 241733, "epoch": 2716, "lr": 2.9386068015985814e-06}
{"train_loss": 0.0350438617169857, "global_step": 241734, "epoch": 2716, "lr": 2.9384109858120124e-06}
{"train_loss": 0.07653135061264038, "global_step": 241735, "epoch": 2716, "lr": 2.9382151763523103e-06}
{"train_loss": 0.0543820895254612, "global_step": 241736, "epoch": 2716, "lr": 2.9380193732194806e-06}
{"train_loss": 0.06971535086631775, "global_step": 241737, "epoch": 2716, "lr": 2.9378235764135677e-06}
{"train_loss": 0.06761038303375244, "global_step": 241738, "epoch": 2716, "lr": 2.937627785934588e-06}
{"train_loss": 0.04057551175355911, "global_step": 241739, "epoch": 2716, "lr": 2.937432001782575e-06}
{"train_loss": 0.09021815657615662, "global_step": 241740, "epoch": 2716, "lr": 2.93723622395754e-06}
{"train_loss": 0.058433160185813904, "global_step": 241741, "epoch": 2716, "lr": 2.937040452459533e-06}
{"train_loss": 0.040799580514431, "global_step": 241742, "epoch": 2716, "lr": 2.936844687288559e-06}
{"train_loss": 0.060964006930589676, "global_step": 241743, "epoch": 2716, "lr": 2.9366489284446576e-06}
{"train_loss": 0.06051389500498772, "global_step": 241744, "epoch": 2716, "lr": 2.936453175927845e-06}
{"train_loss": 0.06169374659657478, "global_step": 241745, "epoch": 2716, "lr": 2.9362574297381594e-06}
{"train_loss": 0.04103253036737442, "global_step": 241746, "epoch": 2716, "lr": 2.936061689875613e-06}
{"train_loss": 0.0648098960518837, "global_step": 241747, "epoch": 2716, "lr": 2.9358659563402445e-06}
{"train_loss": 0.03377524018287659, "global_step": 241748, "epoch": 2716, "lr": 2.9356702291320702e-06}
{"train_loss": 0.03417719155550003, "global_step": 241749, "epoch": 2716, "lr": 2.935474508251129e-06}
{"train_loss": 0.03899620845913887, "global_step": 241750, "epoch": 2716, "lr": 2.935278793697438e-06}
{"train_loss": 0.05325140058994293, "global_step": 241751, "epoch": 2716, "lr": 2.935083085471019e-06}
{"train_loss": 0.04537352919578552, "global_step": 241752, "epoch": 2716, "lr": 2.9348873835719104e-06}
{"train_loss": 0.046177979558706284, "global_step": 241753, "epoch": 2716, "lr": 2.934691688000124e-06}
{"train_loss": 0.04740869998931885, "global_step": 241754, "epoch": 2716, "lr": 2.9344959987557043e-06}
{"train_loss": 0.042027611285448074, "global_step": 241755, "epoch": 2716, "lr": 2.9343003158386617e-06}
{"train_loss": 0.057232726365327835, "global_step": 241756, "epoch": 2716, "lr": 2.9341046392490357e-06}
{"train_loss": 0.0425080768764019, "global_step": 241757, "epoch": 2716, "lr": 2.933908968986837e-06}
{"train_loss": 0.01768309436738491, "global_step": 241758, "epoch": 2716, "lr": 2.9337133050521106e-06}
{"train_loss": 0.0566500648856163, "global_step": 241759, "epoch": 2716, "lr": 2.933517647444861e-06}
{"train_loss": 0.04313850402832031, "global_step": 241760, "epoch": 2716, "lr": 2.9333219961651393e-06}
{"train_loss": 0.05480070039629936, "global_step": 241761, "epoch": 2716, "lr": 2.933126351212945e-06}
{"train_loss": 0.04969963803887367, "global_step": 241762, "epoch": 2716, "lr": 2.932930712588328e-06}
{"train_loss": 0.059985335916280746, "global_step": 241763, "epoch": 2716, "lr": 2.932735080291299e-06}
{"train_loss": 0.018312720581889153, "global_step": 241764, "epoch": 2716, "lr": 2.9325394543218976e-06}
{"train_loss": 0.041084982454776764, "global_step": 241765, "epoch": 2716, "lr": 2.9323438346801347e-06}
{"train_loss": 0.021861271932721138, "global_step": 241766, "epoch": 2716, "lr": 2.932148221366049e-06}
{"train_loss": 0.03410341218113899, "global_step": 241767, "epoch": 2716, "lr": 2.931952614379657e-06}
{"train_loss": 0.05407094955444336, "global_step": 241768, "epoch": 2716, "lr": 2.9317570137209926e-06}
{"train_loss": 0.038887932896614075, "global_step": 241769, "epoch": 2716, "lr": 2.931561419390083e-06}
{"train_loss": 0.06702248007059097, "global_step": 241770, "epoch": 2716, "lr": 2.9313658313869452e-06}
{"train_loss": 0.042424269020557404, "global_step": 241771, "epoch": 2716, "lr": 2.9311702497116233e-06}
{"train_loss": 0.10368107259273529, "global_step": 241772, "epoch": 2716, "lr": 2.9309746743641177e-06}
{"train_loss": 0.044887084513902664, "global_step": 241773, "epoch": 2716, "lr": 2.9307791053444835e-06}
{"train_loss": 0.14073047041893005, "global_step": 241774, "epoch": 2716, "lr": 2.930583542652721e-06}
{"train_loss": 0.019515668973326683, "global_step": 241775, "epoch": 2716, "lr": 2.930387986288874e-06}
{"train_loss": 0.026780977845191956, "global_step": 241776, "epoch": 2716, "lr": 2.930192436252954e-06}
{"train_loss": 0.04727267846465111, "global_step": 241777, "epoch": 2716, "lr": 2.929996892545006e-06}
{"train_loss": 0.050285276025533676, "global_step": 241778, "epoch": 2716, "lr": 2.9298013551650404e-06}
{"train_loss": 0.039933040738105774, "global_step": 241779, "epoch": 2716, "lr": 2.9296058241130965e-06}
{"train_loss": 0.04228627309203148, "global_step": 241780, "epoch": 2716, "lr": 2.9294102993891846e-06}
{"train_loss": 0.038813166320323944, "global_step": 241781, "epoch": 2716, "lr": 2.9292147809933444e-06}
{"train_loss": 0.027764156460762024, "global_step": 241782, "epoch": 2716, "lr": 2.9290192689255926e-06}
{"train_loss": 0.031255342066287994, "global_step": 241783, "epoch": 2716, "lr": 2.928823763185967e-06}
{"train_loss": 0.10345837473869324, "global_step": 241784, "epoch": 2716, "lr": 2.9286282637744854e-06}
{"train_loss": 0.06819210201501846, "global_step": 241785, "epoch": 2716, "lr": 2.9284327706911754e-06}
{"train_loss": 0.05370275676250458, "global_step": 241786, "epoch": 2716, "lr": 2.9282372839360637e-06}
{"train_loss": 0.009947506710886955, "global_step": 241787, "epoch": 2716, "lr": 2.9280418035091793e-06}
{"train_loss": 0.06931412220001221, "global_step": 241788, "epoch": 2716, "lr": 2.927846329410544e-06}
{"train_loss": 0.05861353129148483, "global_step": 241789, "epoch": 2716, "lr": 2.9276508616401853e-06}
{"train_loss": 0.06133788451552391, "global_step": 241790, "epoch": 2716, "lr": 2.9274554001981315e-06}
{"train_loss": 0.054784972220659256, "global_step": 241791, "epoch": 2716, "lr": 2.927259945084404e-06}
{"train_loss": 0.06762592494487762, "global_step": 241792, "epoch": 2716, "lr": 2.9270644962990368e-06}
{"train_loss": 0.030545756220817566, "global_step": 241793, "epoch": 2716, "lr": 2.9268690538420517e-06}
{"train_loss": 0.032996971160173416, "global_step": 241794, "epoch": 2716, "lr": 2.926673617713477e-06}
{"train_loss": 0.08138521760702133, "global_step": 241795, "epoch": 2716, "lr": 2.9264781879133285e-06}
{"train_loss": 0.04836740344762802, "global_step": 241796, "epoch": 2716, "lr": 2.9262827644416514e-06}
{"train_loss": 0.06719156354665756, "global_step": 241797, "epoch": 2716, "lr": 2.926087347298456e-06}
{"train_loss": 0.05277476832270622, "global_step": 241798, "epoch": 2716, "lr": 2.9258919364837767e-06}
{"train_loss": 0.03482334315776825, "global_step": 241799, "epoch": 2716, "lr": 2.925696531997635e-06}
{"train_loss": 0.043778322637081146, "global_step": 241800, "epoch": 2716, "lr": 2.925501133840064e-06}
{"train_loss": 0.05147085711359978, "global_step": 241801, "epoch": 2716, "lr": 2.925305742011081e-06}
{"train_loss": 0.020409226417541504, "global_step": 241802, "epoch": 2716, "lr": 2.9251103565107186e-06}
{"train_loss": 0.036070678383111954, "global_step": 241803, "epoch": 2716, "lr": 2.9249149773389993e-06}
{"train_loss": 0.05166470631957054, "global_step": 241804, "epoch": 2716, "lr": 2.9247196044959514e-06}
{"train_loss": 0.0914648175239563, "global_step": 241805, "epoch": 2716, "lr": 2.9245242379816073e-06}
{"train_loss": 0.05705612152814865, "global_step": 241806, "epoch": 2716, "lr": 2.9243288777959786e-06}
{"train_loss": 0.08006901293992996, "global_step": 241807, "epoch": 2716, "lr": 2.92413352393911e-06}
{"train_loss": 0.030362719669938087, "global_step": 241808, "epoch": 2716, "lr": 2.9239381764110063e-06}
{"train_loss": 0.025000885128974915, "global_step": 241809, "epoch": 2716, "lr": 2.923742835211718e-06}
{"train_loss": 0.06797807663679123, "global_step": 241810, "epoch": 2716, "lr": 2.923547500341245e-06}
{"train_loss": 0.017648708075284958, "global_step": 241811, "epoch": 2716, "lr": 2.923352171799643e-06}
{"train_loss": 0.04936017432993048, "global_step": 241812, "epoch": 2716, "lr": 2.923156849586911e-06, "val_loss": 9.45810604095459}
{"train_loss": 0.05602812021970749, "global_step": 241813, "epoch": 2717, "lr": 2.922961533703089e-06}
{"train_loss": 0.060468509793281555, "global_step": 241814, "epoch": 2717, "lr": 2.922766224148199e-06}
{"train_loss": 0.03974917531013489, "global_step": 241815, "epoch": 2717, "lr": 2.9225709209222797e-06}
{"train_loss": 0.0849204957485199, "global_step": 241816, "epoch": 2717, "lr": 2.9223756240253308e-06}
{"train_loss": 0.03731060028076172, "global_step": 241817, "epoch": 2717, "lr": 2.9221803334574082e-06}
{"train_loss": 0.04959994554519653, "global_step": 241818, "epoch": 2717, "lr": 2.9219850492185176e-06}
{"train_loss": 0.030732903629541397, "global_step": 241819, "epoch": 2717, "lr": 2.9217897713086973e-06}
{"train_loss": 0.08110213279724121, "global_step": 241820, "epoch": 2717, "lr": 2.9215944997279644e-06}
{"train_loss": 0.06604798138141632, "global_step": 241821, "epoch": 2717, "lr": 2.921399234476352e-06}
{"train_loss": 0.02532590739428997, "global_step": 241822, "epoch": 2717, "lr": 2.9212039755538823e-06}
{"train_loss": 0.08840621262788773, "global_step": 241823, "epoch": 2717, "lr": 2.921008722960583e-06}
{"train_loss": 0.0352267287671566, "global_step": 241824, "epoch": 2717, "lr": 2.920813476696477e-06}
{"train_loss": 0.03540302813053131, "global_step": 241825, "epoch": 2717, "lr": 2.9206182367616018e-06}
{"train_loss": 0.04707947000861168, "global_step": 241826, "epoch": 2717, "lr": 2.9204230031559753e-06}
{"train_loss": 0.02559400163590908, "global_step": 241827, "epoch": 2717, "lr": 2.9202277758796137e-06}
{"train_loss": 0.020333899185061455, "global_step": 241828, "epoch": 2717, "lr": 2.920032554932567e-06}
{"train_loss": 0.019689852371811867, "global_step": 241829, "epoch": 2717, "lr": 2.919837340314835e-06}
{"train_loss": 0.03336043283343315, "global_step": 241830, "epoch": 2717, "lr": 2.9196421320264676e-06}
{"train_loss": 0.06405654549598694, "global_step": 241831, "epoch": 2717, "lr": 2.919446930067471e-06}
{"train_loss": 0.024219926446676254, "global_step": 241832, "epoch": 2717, "lr": 2.9192517344378887e-06}
{"train_loss": 0.06150514632463455, "global_step": 241833, "epoch": 2717, "lr": 2.9190565451377326e-06}
{"train_loss": 0.04297984391450882, "global_step": 241834, "epoch": 2717, "lr": 2.9188613621670467e-06}
{"train_loss": 0.025354063138365746, "global_step": 241835, "epoch": 2717, "lr": 2.9186661855258314e-06}
{"train_loss": 0.06277994066476822, "global_step": 241836, "epoch": 2717, "lr": 2.9184710152141415e-06}
{"train_loss": 0.03924950212240219, "global_step": 241837, "epoch": 2717, "lr": 2.9182758512319776e-06}
{"train_loss": 0.0603007972240448, "global_step": 241838, "epoch": 2717, "lr": 2.9180806935793835e-06}
{"train_loss": 0.050088752061128616, "global_step": 241839, "epoch": 2717, "lr": 2.917885542256377e-06}
{"train_loss": 0.042011894285678864, "global_step": 241840, "epoch": 2717, "lr": 2.9176903972629845e-06}
{"train_loss": 0.04761398211121559, "global_step": 241841, "epoch": 2717, "lr": 2.9174952585992453e-06}
{"train_loss": 0.019772253930568695, "global_step": 241842, "epoch": 2717, "lr": 2.91730012626516e-06}
{"train_loss": 0.042945846915245056, "global_step": 241843, "epoch": 2717, "lr": 2.9171050002607836e-06}
{"train_loss": 0.03853163123130798, "global_step": 241844, "epoch": 2717, "lr": 2.9169098805861216e-06}
{"train_loss": 0.02863110601902008, "global_step": 241845, "epoch": 2717, "lr": 2.9167147672412132e-06}
{"train_loss": 0.08621825277805328, "global_step": 241846, "epoch": 2717, "lr": 2.916519660226069e-06}
{"train_loss": 0.013867367058992386, "global_step": 241847, "epoch": 2717, "lr": 2.916324559540734e-06}
{"train_loss": 0.04864884540438652, "global_step": 241848, "epoch": 2717, "lr": 2.916129465185219e-06}
{"train_loss": 0.02551255188882351, "global_step": 241849, "epoch": 2717, "lr": 2.9159343771595628e-06}
{"train_loss": 0.030260156840085983, "global_step": 241850, "epoch": 2717, "lr": 2.915739295463776e-06}
{"train_loss": 0.01393984816968441, "global_step": 241851, "epoch": 2717, "lr": 2.9155442200979045e-06}
{"train_loss": 0.0581139400601387, "global_step": 241852, "epoch": 2717, "lr": 2.915349151061958e-06}
{"train_loss": 0.034191351383924484, "global_step": 241853, "epoch": 2717, "lr": 2.91515408835597e-06}
{"train_loss": 0.10672955214977264, "global_step": 241854, "epoch": 2717, "lr": 2.9149590319799635e-06}
{"train_loss": 0.06768085062503815, "global_step": 241855, "epoch": 2717, "lr": 2.9147639819339713e-06}
{"train_loss": 0.06645843386650085, "global_step": 241856, "epoch": 2717, "lr": 2.9145689382180095e-06}
{"train_loss": 0.04639064893126488, "global_step": 241857, "epoch": 2717, "lr": 2.914373900832118e-06}
{"train_loss": 0.03173849359154701, "global_step": 241858, "epoch": 2717, "lr": 2.9141788697763073e-06}
{"train_loss": 0.09002175182104111, "global_step": 241859, "epoch": 2717, "lr": 2.9139838450506218e-06}
{"train_loss": 0.023699752986431122, "global_step": 241860, "epoch": 2717, "lr": 2.913788826655067e-06}
{"train_loss": 0.07312650233507156, "global_step": 241861, "epoch": 2717, "lr": 2.9135938145896825e-06}
{"train_loss": 0.07143896818161011, "global_step": 241862, "epoch": 2717, "lr": 2.9133988088544896e-06}
{"train_loss": 0.021083323284983635, "global_step": 241863, "epoch": 2717, "lr": 2.9132038094495217e-06}
{"train_loss": 0.021685903891921043, "global_step": 241864, "epoch": 2717, "lr": 2.913008816374796e-06}
{"train_loss": 0.03350984305143356, "global_step": 241865, "epoch": 2717, "lr": 2.9128138296303454e-06}
{"train_loss": 0.08371508121490479, "global_step": 241866, "epoch": 2717, "lr": 2.912618849216192e-06}
{"train_loss": 0.08189669251441956, "global_step": 241867, "epoch": 2717, "lr": 2.9124238751323584e-06}
{"train_loss": 0.06431373953819275, "global_step": 241868, "epoch": 2717, "lr": 2.912228907378883e-06}
{"train_loss": 0.05558633804321289, "global_step": 241869, "epoch": 2717, "lr": 2.9120339459557776e-06}
{"train_loss": 0.07149656862020493, "global_step": 241870, "epoch": 2717, "lr": 2.9118389908630804e-06}
{"train_loss": 0.09232447296380997, "global_step": 241871, "epoch": 2717, "lr": 2.911644042100803e-06}
{"train_loss": 0.043041061609983444, "global_step": 241872, "epoch": 2717, "lr": 2.9114490996689948e-06}
{"train_loss": 0.04990215599536896, "global_step": 241873, "epoch": 2717, "lr": 2.911254163567656e-06}
{"train_loss": 0.024465564638376236, "global_step": 241874, "epoch": 2717, "lr": 2.9110592337968367e-06}
{"train_loss": 0.019092006608843803, "global_step": 241875, "epoch": 2717, "lr": 2.9108643103565423e-06}
{"train_loss": 0.0588545985519886, "global_step": 241876, "epoch": 2717, "lr": 2.9106693932468065e-06}
{"train_loss": 0.02853964827954769, "global_step": 241877, "epoch": 2717, "lr": 2.9104744824676675e-06}
{"train_loss": 0.05685656517744064, "global_step": 241878, "epoch": 2717, "lr": 2.9102795780191317e-06}
{"train_loss": 0.055460117757320404, "global_step": 241879, "epoch": 2717, "lr": 2.9100846799012425e-06}
{"train_loss": 0.06254575401544571, "global_step": 241880, "epoch": 2717, "lr": 2.909889788114012e-06}
{"train_loss": 0.024517610669136047, "global_step": 241881, "epoch": 2717, "lr": 2.909694902657478e-06}
{"train_loss": 0.06620311737060547, "global_step": 241882, "epoch": 2717, "lr": 2.9095000235316585e-06}
{"train_loss": 0.025968460366129875, "global_step": 241883, "epoch": 2717, "lr": 2.9093051507365854e-06}
{"train_loss": 0.024004727602005005, "global_step": 241884, "epoch": 2717, "lr": 2.909110284272276e-06}
{"train_loss": 0.029009830206632614, "global_step": 241885, "epoch": 2717, "lr": 2.9089154241387697e-06}
{"train_loss": 0.030107785016298294, "global_step": 241886, "epoch": 2717, "lr": 2.908720570336082e-06}
{"train_loss": 0.07318788021802902, "global_step": 241887, "epoch": 2717, "lr": 2.908525722864247e-06}
{"train_loss": 0.04573303088545799, "global_step": 241888, "epoch": 2717, "lr": 2.9083308817232813e-06}
{"train_loss": 0.025694964453577995, "global_step": 241889, "epoch": 2717, "lr": 2.9081360469132233e-06}
{"train_loss": 0.06153254956007004, "global_step": 241890, "epoch": 2717, "lr": 2.907941218434085e-06}
{"train_loss": 0.06650429964065552, "global_step": 241891, "epoch": 2717, "lr": 2.907746396285904e-06}
{"train_loss": 0.040961503982543945, "global_step": 241892, "epoch": 2717, "lr": 2.9075515804687036e-06}
{"train_loss": 0.06363905966281891, "global_step": 241893, "epoch": 2717, "lr": 2.907356770982511e-06}
{"train_loss": 0.04895365238189697, "global_step": 241894, "epoch": 2717, "lr": 2.9071619678273433e-06}
{"train_loss": 0.027755005285143852, "global_step": 241895, "epoch": 2717, "lr": 2.906967171003233e-06}
{"train_loss": 0.06459880620241165, "global_step": 241896, "epoch": 2717, "lr": 2.906772380510214e-06}
{"train_loss": 0.06162127107381821, "global_step": 241897, "epoch": 2717, "lr": 2.9065775963482977e-06}
{"train_loss": 0.026493370532989502, "global_step": 241898, "epoch": 2717, "lr": 2.9063828185175278e-06}
{"train_loss": 0.023025115951895714, "global_step": 241899, "epoch": 2717, "lr": 2.9061880470179158e-06}
{"train_loss": 0.07318049669265747, "global_step": 241900, "epoch": 2717, "lr": 2.9059932818494896e-06}
{"train_loss": 0.04772983158656051, "global_step": 241901, "epoch": 2717, "lr": 2.9057985230122875e-06, "val_loss": 9.401240348815918}
{"train_loss": 0.03685404360294342, "global_step": 241902, "epoch": 2718, "lr": 2.9056037705063153e-06}
{"train_loss": 0.030203072354197502, "global_step": 241903, "epoch": 2718, "lr": 2.9054090243316234e-06}
{"train_loss": 0.022438064217567444, "global_step": 241904, "epoch": 2718, "lr": 2.9052142844882225e-06}
{"train_loss": 0.053365688771009445, "global_step": 241905, "epoch": 2718, "lr": 2.905019550976135e-06}
{"train_loss": 0.05344712361693382, "global_step": 241906, "epoch": 2718, "lr": 2.904824823795399e-06}
{"train_loss": 0.0548822246491909, "global_step": 241907, "epoch": 2718, "lr": 2.9046301029460264e-06}
{"train_loss": 0.05420530214905739, "global_step": 241908, "epoch": 2718, "lr": 2.904435388428062e-06}
{"train_loss": 0.045198116451501846, "global_step": 241909, "epoch": 2718, "lr": 2.9042406802415156e-06}
{"train_loss": 0.07710962742567062, "global_step": 241910, "epoch": 2718, "lr": 2.9040459783864272e-06}
{"train_loss": 0.03936641290783882, "global_step": 241911, "epoch": 2718, "lr": 2.903851282862807e-06}
{"train_loss": 0.0419340580701828, "global_step": 241912, "epoch": 2718, "lr": 2.9036565936706894e-06}
{"train_loss": 0.02320771850645542, "global_step": 241913, "epoch": 2718, "lr": 2.9034619108101124e-06}
{"train_loss": 0.043336935341358185, "global_step": 241914, "epoch": 2718, "lr": 2.9032672342810817e-06}
{"train_loss": 0.027203472331166267, "global_step": 241915, "epoch": 2718, "lr": 2.9030725640836367e-06}
{"train_loss": 0.048502057790756226, "global_step": 241916, "epoch": 2718, "lr": 2.902877900217793e-06}
{"train_loss": 0.07619251310825348, "global_step": 241917, "epoch": 2718, "lr": 2.902683242683596e-06}
{"train_loss": 0.029682938009500504, "global_step": 241918, "epoch": 2718, "lr": 2.9024885914810452e-06}
{"train_loss": 0.028602194041013718, "global_step": 241919, "epoch": 2718, "lr": 2.902293946610196e-06}
{"train_loss": 0.08449064940214157, "global_step": 241920, "epoch": 2718, "lr": 2.9020993080710434e-06}
{"train_loss": 0.07410215586423874, "global_step": 241921, "epoch": 2718, "lr": 2.9019046758636427e-06}
{"train_loss": 0.043082114309072495, "global_step": 241922, "epoch": 2718, "lr": 2.901710049987999e-06}
{"train_loss": 0.04031835496425629, "global_step": 241923, "epoch": 2718, "lr": 2.901515430444152e-06}
{"train_loss": 0.04265489801764488, "global_step": 241924, "epoch": 2718, "lr": 2.9013208172321117e-06}
{"train_loss": 0.04395533725619316, "global_step": 241925, "epoch": 2718, "lr": 2.9011262103519287e-06}
{"train_loss": 0.02875894121825695, "global_step": 241926, "epoch": 2718, "lr": 2.9009316098036033e-06}
{"train_loss": 0.046946410089731216, "global_step": 241927, "epoch": 2718, "lr": 2.900737015587185e-06}
{"train_loss": 0.04295941814780235, "global_step": 241928, "epoch": 2718, "lr": 2.900542427702674e-06}
{"train_loss": 0.04670407623052597, "global_step": 241929, "epoch": 2718, "lr": 2.9003478461501253e-06}
{"train_loss": 0.06573422998189926, "global_step": 241930, "epoch": 2718, "lr": 2.9001532709295398e-06}
{"train_loss": 0.037427835166454315, "global_step": 241931, "epoch": 2718, "lr": 2.899958702040956e-06}
{"train_loss": 0.025748299434781075, "global_step": 241932, "epoch": 2718, "lr": 2.8997641394844065e-06}
{"train_loss": 0.014721369370818138, "global_step": 241933, "epoch": 2718, "lr": 2.8995695832599036e-06}
{"train_loss": 0.04578028991818428, "global_step": 241934, "epoch": 2718, "lr": 2.8993750333674853e-06}
{"train_loss": 0.03629834204912186, "global_step": 241935, "epoch": 2718, "lr": 2.8991804898071627e-06}
{"train_loss": 0.03520405665040016, "global_step": 241936, "epoch": 2718, "lr": 2.898985952578981e-06}
{"train_loss": 0.024373210966587067, "global_step": 241937, "epoch": 2718, "lr": 2.898791421682956e-06}
{"train_loss": 0.04454532265663147, "global_step": 241938, "epoch": 2718, "lr": 2.898596897119105e-06}
{"train_loss": 0.03305980935692787, "global_step": 241939, "epoch": 2718, "lr": 2.8984023788874716e-06}
{"train_loss": 0.08587882667779922, "global_step": 241940, "epoch": 2718, "lr": 2.8982078669880673e-06}
{"train_loss": 0.06747422367334366, "global_step": 241941, "epoch": 2718, "lr": 2.898013361420937e-06}
{"train_loss": 0.013876514509320259, "global_step": 241942, "epoch": 2718, "lr": 2.897818862186086e-06}
{"train_loss": 0.03734687343239784, "global_step": 241943, "epoch": 2718, "lr": 2.897624369283547e-06}
{"train_loss": 0.020251652225852013, "global_step": 241944, "epoch": 2718, "lr": 2.897429882713354e-06}
{"train_loss": 0.017125001177191734, "global_step": 241945, "epoch": 2718, "lr": 2.8972354024755233e-06}
{"train_loss": 0.06383858621120453, "global_step": 241946, "epoch": 2718, "lr": 2.8970409285700885e-06}
{"train_loss": 0.042238544672727585, "global_step": 241947, "epoch": 2718, "lr": 2.8968464609970657e-06}
{"train_loss": 0.045684438198804855, "global_step": 241948, "epoch": 2718, "lr": 2.8966519997565e-06}
{"train_loss": 0.031824078410863876, "global_step": 241949, "epoch": 2718, "lr": 2.896457544848391e-06}
{"train_loss": 0.040847938507795334, "global_step": 241950, "epoch": 2718, "lr": 2.8962630962727832e-06}
{"train_loss": 0.04592837765812874, "global_step": 241951, "epoch": 2718, "lr": 2.8960686540297043e-06}
{"train_loss": 0.030992377549409866, "global_step": 241952, "epoch": 2718, "lr": 2.895874218119171e-06}
{"train_loss": 0.03136603161692619, "global_step": 241953, "epoch": 2718, "lr": 2.895679788541217e-06}
{"train_loss": 0.06197841838002205, "global_step": 241954, "epoch": 2718, "lr": 2.895485365295858e-06}
{"train_loss": 0.05030829459428787, "global_step": 241955, "epoch": 2718, "lr": 2.8952909483831336e-06}
{"train_loss": 0.05014762282371521, "global_step": 241956, "epoch": 2718, "lr": 2.8950965378030546e-06}
{"train_loss": 0.02727944776415825, "global_step": 241957, "epoch": 2718, "lr": 2.894902133555666e-06}
{"train_loss": 0.06763624399900436, "global_step": 241958, "epoch": 2718, "lr": 2.894707735640978e-06}
{"train_loss": 0.04398110508918762, "global_step": 241959, "epoch": 2718, "lr": 2.8945133440590243e-06}
{"train_loss": 0.057191699743270874, "global_step": 241960, "epoch": 2718, "lr": 2.8943189588098274e-06}
{"train_loss": 0.049135949462652206, "global_step": 241961, "epoch": 2718, "lr": 2.89412457989342e-06}
{"train_loss": 0.0732082948088646, "global_step": 241962, "epoch": 2718, "lr": 2.893930207309814e-06}
{"train_loss": 0.08464344590902328, "global_step": 241963, "epoch": 2718, "lr": 2.8937358410590533e-06}
{"train_loss": 0.03845037892460823, "global_step": 241964, "epoch": 2718, "lr": 2.893541481141154e-06}
{"train_loss": 0.04252224415540695, "global_step": 241965, "epoch": 2718, "lr": 2.893347127556145e-06}
{"train_loss": 0.05395081639289856, "global_step": 241966, "epoch": 2718, "lr": 2.8931527803040424e-06}
{"train_loss": 0.02225428633391857, "global_step": 241967, "epoch": 2718, "lr": 2.892958439384885e-06}
{"train_loss": 0.04708901047706604, "global_step": 241968, "epoch": 2718, "lr": 2.892764104798706e-06}
{"train_loss": 0.09053163975477219, "global_step": 241969, "epoch": 2718, "lr": 2.8925697765455063e-06}
{"train_loss": 0.05228569731116295, "global_step": 241970, "epoch": 2718, "lr": 2.89237545462534e-06}
{"train_loss": 0.029993612319231033, "global_step": 241971, "epoch": 2718, "lr": 2.8921811390382083e-06}
{"train_loss": 0.03517817333340645, "global_step": 241972, "epoch": 2718, "lr": 2.8919868297841547e-06}
{"train_loss": 0.04575558006763458, "global_step": 241973, "epoch": 2718, "lr": 2.8917925268631962e-06}
{"train_loss": 0.05474444478750229, "global_step": 241974, "epoch": 2718, "lr": 2.8915982302753663e-06}
{"train_loss": 0.032298654317855835, "global_step": 241975, "epoch": 2718, "lr": 2.8914039400206814e-06}
{"train_loss": 0.07018620520830154, "global_step": 241976, "epoch": 2718, "lr": 2.891209656099181e-06}
{"train_loss": 0.04028921574354172, "global_step": 241977, "epoch": 2718, "lr": 2.8910153785108807e-06}
{"train_loss": 0.03561475872993469, "global_step": 241978, "epoch": 2718, "lr": 2.8908211072558035e-06}
{"train_loss": 0.03568233922123909, "global_step": 241979, "epoch": 2718, "lr": 2.8906268423339877e-06}
{"train_loss": 0.03042389079928398, "global_step": 241980, "epoch": 2718, "lr": 2.890432583745445e-06}
{"train_loss": 0.02046903409063816, "global_step": 241981, "epoch": 2718, "lr": 2.890238331490219e-06}
{"train_loss": 0.03167424723505974, "global_step": 241982, "epoch": 2718, "lr": 2.890044085568322e-06}
{"train_loss": 0.08685558289289474, "global_step": 241983, "epoch": 2718, "lr": 2.8898498459797805e-06}
{"train_loss": 0.024510789662599564, "global_step": 241984, "epoch": 2718, "lr": 2.8896556127246345e-06}
{"train_loss": 0.030678605660796165, "global_step": 241985, "epoch": 2718, "lr": 2.8894613858028886e-06}
{"train_loss": 0.07010010629892349, "global_step": 241986, "epoch": 2718, "lr": 2.889267165214582e-06}
{"train_loss": 0.039889927953481674, "global_step": 241987, "epoch": 2718, "lr": 2.8890729509597425e-06}
{"train_loss": 0.0580575168132782, "global_step": 241988, "epoch": 2718, "lr": 2.8888787430383923e-06}
{"train_loss": 0.0809582769870758, "global_step": 241989, "epoch": 2718, "lr": 2.8886845414505594e-06}
{"train_loss": 0.045401156534639636, "global_step": 241990, "epoch": 2718, "lr": 2.8884903461962655e-06, "val_loss": 9.385469436645508}
{"train_loss": 0.08415332436561584, "global_step": 241991, "epoch": 2719, "lr": 2.888296157275544e-06}
{"train_loss": 0.025466393679380417, "global_step": 241992, "epoch": 2719, "lr": 2.8881019746884117e-06}
{"train_loss": 0.0473211333155632, "global_step": 241993, "epoch": 2719, "lr": 2.887907798434908e-06}
{"train_loss": 0.06409325450658798, "global_step": 241994, "epoch": 2719, "lr": 2.887713628515043e-06}
{"train_loss": 0.021792327985167503, "global_step": 241995, "epoch": 2719, "lr": 2.887519464928856e-06}
{"train_loss": 0.0360146164894104, "global_step": 241996, "epoch": 2719, "lr": 2.887325307676364e-06}
{"train_loss": 0.04773162677884102, "global_step": 241997, "epoch": 2719, "lr": 2.8871311567575996e-06}
{"train_loss": 0.04912354052066803, "global_step": 241998, "epoch": 2719, "lr": 2.8869370121725858e-06}
{"train_loss": 0.01808340661227703, "global_step": 241999, "epoch": 2719, "lr": 2.8867428739213497e-06}
{"train_loss": 0.037199608981609344, "global_step": 242000, "epoch": 2719, "lr": 2.886548742003914e-06}
{"train_loss": 0.020369838923215866, "global_step": 242001, "epoch": 2719, "lr": 2.8863546164203114e-06}
{"train_loss": 0.03805327042937279, "global_step": 242002, "epoch": 2719, "lr": 2.886160497170559e-06}
{"train_loss": 0.021295147016644478, "global_step": 242003, "epoch": 2719, "lr": 2.8859663842546846e-06}
{"train_loss": 0.08292707800865173, "global_step": 242004, "epoch": 2719, "lr": 2.885772277672727e-06}
{"train_loss": 0.05762434005737305, "global_step": 242005, "epoch": 2719, "lr": 2.885578177424697e-06}
{"train_loss": 0.03612803295254707, "global_step": 242006, "epoch": 2719, "lr": 2.8853840835106337e-06}
{"train_loss": 0.036615047603845596, "global_step": 242007, "epoch": 2719, "lr": 2.8851899959305484e-06}
{"train_loss": 0.02228628657758236, "global_step": 242008, "epoch": 2719, "lr": 2.8849959146844853e-06}
{"train_loss": 0.02790190279483795, "global_step": 242009, "epoch": 2719, "lr": 2.8848018397724495e-06}
{"train_loss": 0.03294325992465019, "global_step": 242010, "epoch": 2719, "lr": 2.884607771194486e-06}
{"train_loss": 0.04908563196659088, "global_step": 242011, "epoch": 2719, "lr": 2.884413708950606e-06}
{"train_loss": 0.08303829282522202, "global_step": 242012, "epoch": 2719, "lr": 2.8842196530408485e-06}
{"train_loss": 0.047503240406513214, "global_step": 242013, "epoch": 2719, "lr": 2.884025603465229e-06}
{"train_loss": 0.0723566934466362, "global_step": 242014, "epoch": 2719, "lr": 2.883831560223782e-06}
{"train_loss": 0.04815312847495079, "global_step": 242015, "epoch": 2719, "lr": 2.883637523316529e-06}
{"train_loss": 0.035920415073633194, "global_step": 242016, "epoch": 2719, "lr": 2.883443492743493e-06}
{"train_loss": 0.029042286798357964, "global_step": 242017, "epoch": 2719, "lr": 2.883249468504712e-06}
{"train_loss": 0.02300332300364971, "global_step": 242018, "epoch": 2719, "lr": 2.883055450600192e-06}
{"train_loss": 0.025283291935920715, "global_step": 242019, "epoch": 2719, "lr": 2.882861439029977e-06}
{"train_loss": 0.023342709988355637, "global_step": 242020, "epoch": 2719, "lr": 2.8826674337940894e-06}
{"train_loss": 0.03925658017396927, "global_step": 242021, "epoch": 2719, "lr": 2.8824734348925465e-06}
{"train_loss": 0.032723840326070786, "global_step": 242022, "epoch": 2719, "lr": 2.8822794423253806e-06}
{"train_loss": 0.06007334217429161, "global_step": 242023, "epoch": 2719, "lr": 2.8820854560926257e-06}
{"train_loss": 0.03401535376906395, "global_step": 242024, "epoch": 2719, "lr": 2.8818914761942927e-06}
{"train_loss": 0.0728827491402626, "global_step": 242025, "epoch": 2719, "lr": 2.8816975026304203e-06}
{"train_loss": 0.027520908042788506, "global_step": 242026, "epoch": 2719, "lr": 2.8815035354010257e-06}
{"train_loss": 0.02771732769906521, "global_step": 242027, "epoch": 2719, "lr": 2.881309574506141e-06}
{"train_loss": 0.0723295584321022, "global_step": 242028, "epoch": 2719, "lr": 2.88111561994579e-06}
{"train_loss": 0.04103238135576248, "global_step": 242029, "epoch": 2719, "lr": 2.8809216717199994e-06}
{"train_loss": 0.03501760959625244, "global_step": 242030, "epoch": 2719, "lr": 2.8807277298287916e-06}
{"train_loss": 0.036173559725284576, "global_step": 242031, "epoch": 2719, "lr": 2.8805337942721997e-06}
{"train_loss": 0.05566766485571861, "global_step": 242032, "epoch": 2719, "lr": 2.8803398650502357e-06}
{"train_loss": 0.04283026233315468, "global_step": 242033, "epoch": 2719, "lr": 2.8801459421629486e-06}
{"train_loss": 0.042074836790561676, "global_step": 242034, "epoch": 2719, "lr": 2.879952025610344e-06}
{"train_loss": 0.06426556408405304, "global_step": 242035, "epoch": 2719, "lr": 2.8797581153924614e-06}
{"train_loss": 0.03728680685162544, "global_step": 242036, "epoch": 2719, "lr": 2.8795642115093114e-06}
{"train_loss": 0.04260444641113281, "global_step": 242037, "epoch": 2719, "lr": 2.8793703139609387e-06}
{"train_loss": 0.05807732045650482, "global_step": 242038, "epoch": 2719, "lr": 2.879176422747354e-06}
{"train_loss": 0.048019468784332275, "global_step": 242039, "epoch": 2719, "lr": 2.8789825378685907e-06}
{"train_loss": 0.05620872601866722, "global_step": 242040, "epoch": 2719, "lr": 2.878788659324677e-06}
{"train_loss": 0.04280347377061844, "global_step": 242041, "epoch": 2719, "lr": 2.878594787115635e-06}
{"train_loss": 0.04717820882797241, "global_step": 242042, "epoch": 2719, "lr": 2.878400921241492e-06}
{"train_loss": 0.055731769651174545, "global_step": 242043, "epoch": 2719, "lr": 2.8782070617022707e-06}
{"train_loss": 0.023068781942129135, "global_step": 242044, "epoch": 2719, "lr": 2.878013208498004e-06}
{"train_loss": 0.04202606528997421, "global_step": 242045, "epoch": 2719, "lr": 2.8778193616287085e-06}
{"train_loss": 0.03383397310972214, "global_step": 242046, "epoch": 2719, "lr": 2.8776255210944238e-06}
{"train_loss": 0.0607694573700428, "global_step": 242047, "epoch": 2719, "lr": 2.8774316868951602e-06}
{"train_loss": 0.09048175066709518, "global_step": 242048, "epoch": 2719, "lr": 2.8772378590309624e-06}
{"train_loss": 0.047123294323682785, "global_step": 242049, "epoch": 2719, "lr": 2.877044037501836e-06}
{"train_loss": 0.08243072032928467, "global_step": 242050, "epoch": 2719, "lr": 2.87685022230782e-06}
{"train_loss": 0.03196142241358757, "global_step": 242051, "epoch": 2719, "lr": 2.876656413448936e-06}
{"train_loss": 0.014337025582790375, "global_step": 242052, "epoch": 2719, "lr": 2.8764626109252124e-06}
{"train_loss": 0.023717623203992844, "global_step": 242053, "epoch": 2719, "lr": 2.876268814736677e-06}
{"train_loss": 0.050066057592630386, "global_step": 242054, "epoch": 2719, "lr": 2.8760750248833456e-06}
{"train_loss": 0.032726552337408066, "global_step": 242055, "epoch": 2719, "lr": 2.875881241365258e-06}
{"train_loss": 0.028489504009485245, "global_step": 242056, "epoch": 2719, "lr": 2.8756874641824307e-06}
{"train_loss": 0.04484403878450394, "global_step": 242057, "epoch": 2719, "lr": 2.8754936933348962e-06}
{"train_loss": 0.05483240261673927, "global_step": 242058, "epoch": 2719, "lr": 2.8752999288226667e-06}
{"train_loss": 0.06064506247639656, "global_step": 242059, "epoch": 2719, "lr": 2.8751061706457914e-06}
{"train_loss": 0.07720614224672318, "global_step": 242060, "epoch": 2719, "lr": 2.8749124188042766e-06}
{"train_loss": 0.03798556327819824, "global_step": 242061, "epoch": 2719, "lr": 2.8747186732981603e-06}
{"train_loss": 0.020848728716373444, "global_step": 242062, "epoch": 2719, "lr": 2.874524934127459e-06}
{"train_loss": 0.045059606432914734, "global_step": 242063, "epoch": 2719, "lr": 2.8743312012922075e-06}
{"train_loss": 0.02204822562634945, "global_step": 242064, "epoch": 2719, "lr": 2.874137474792421e-06}
{"train_loss": 0.02051057107746601, "global_step": 242065, "epoch": 2719, "lr": 2.873943754628139e-06}
{"train_loss": 0.031044235453009605, "global_step": 242066, "epoch": 2719, "lr": 2.873750040799378e-06}
{"train_loss": 0.047786079347133636, "global_step": 242067, "epoch": 2719, "lr": 2.873556333306171e-06}
{"train_loss": 0.042368628084659576, "global_step": 242068, "epoch": 2719, "lr": 2.873362632148535e-06}
{"train_loss": 0.05381816625595093, "global_step": 242069, "epoch": 2719, "lr": 2.8731689373265034e-06}
{"train_loss": 0.0201326422393322, "global_step": 242070, "epoch": 2719, "lr": 2.8729752488400986e-06}
{"train_loss": 0.06643766909837723, "global_step": 242071, "epoch": 2719, "lr": 2.8727815666893477e-06}
{"train_loss": 0.08743660897016525, "global_step": 242072, "epoch": 2719, "lr": 2.8725878908742733e-06}
{"train_loss": 0.06725386530160904, "global_step": 242073, "epoch": 2719, "lr": 2.8723942213949138e-06}
{"train_loss": 0.03997458517551422, "global_step": 242074, "epoch": 2719, "lr": 2.872200558251281e-06}
{"train_loss": 0.03313882276415825, "global_step": 242075, "epoch": 2719, "lr": 2.872006901443408e-06}
{"train_loss": 0.09636034816503525, "global_step": 242076, "epoch": 2719, "lr": 2.871813250971317e-06}
{"train_loss": 0.07351668924093246, "global_step": 242077, "epoch": 2719, "lr": 2.8716196068350355e-06}
{"train_loss": 0.036588091403245926, "global_step": 242078, "epoch": 2719, "lr": 2.871425969034597e-06}
{"train_loss": 0.044915976002812386, "global_step": 242079, "epoch": 2719, "lr": 2.871232337570012e-06, "val_loss": 9.515725135803223}
{"train_loss": 0.03914622962474823, "global_step": 242080, "epoch": 2720, "lr": 2.8710387124413206e-06}
{"train_loss": 0.029115520417690277, "global_step": 242081, "epoch": 2720, "lr": 2.8708450936485444e-06}
{"train_loss": 0.023353751748800278, "global_step": 242082, "epoch": 2720, "lr": 2.8706514811917106e-06}
{"train_loss": 0.06792155653238297, "global_step": 242083, "epoch": 2720, "lr": 2.8704578750708366e-06}
{"train_loss": 0.07444573938846588, "global_step": 242084, "epoch": 2720, "lr": 2.870264275285961e-06}
{"train_loss": 0.10089287161827087, "global_step": 242085, "epoch": 2720, "lr": 2.870070681837095e-06}
{"train_loss": 0.03097844310104847, "global_step": 242086, "epoch": 2720, "lr": 2.869877094724288e-06}
{"train_loss": 0.03670351952314377, "global_step": 242087, "epoch": 2720, "lr": 2.869683513947535e-06}
{"train_loss": 0.08479373157024384, "global_step": 242088, "epoch": 2720, "lr": 2.869489939506892e-06}
{"train_loss": 0.07587161660194397, "global_step": 242089, "epoch": 2720, "lr": 2.8692963714023634e-06}
{"train_loss": 0.08167888969182968, "global_step": 242090, "epoch": 2720, "lr": 2.8691028096339945e-06}
{"train_loss": 0.0746036171913147, "global_step": 242091, "epoch": 2720, "lr": 2.868909254201785e-06}
{"train_loss": 0.0347556434571743, "global_step": 242092, "epoch": 2720, "lr": 2.86871570510579e-06}
{"train_loss": 0.032261449843645096, "global_step": 242093, "epoch": 2720, "lr": 2.8685221623460158e-06}
{"train_loss": 0.05599123239517212, "global_step": 242094, "epoch": 2720, "lr": 2.8683286259224952e-06}
{"train_loss": 0.056266091763973236, "global_step": 242095, "epoch": 2720, "lr": 2.868135095835256e-06}
{"train_loss": 0.011833969503641129, "global_step": 242096, "epoch": 2720, "lr": 2.8679415720843096e-06}
{"train_loss": 0.04259984940290451, "global_step": 242097, "epoch": 2720, "lr": 2.867748054669711e-06}
{"train_loss": 0.04814590886235237, "global_step": 242098, "epoch": 2720, "lr": 2.8675545435914554e-06}
{"train_loss": 0.028250260278582573, "global_step": 242099, "epoch": 2720, "lr": 2.867361038849592e-06}
{"train_loss": 0.02880525216460228, "global_step": 242100, "epoch": 2720, "lr": 2.8671675404441267e-06}
{"train_loss": 0.04487394914031029, "global_step": 242101, "epoch": 2720, "lr": 2.8669740483751095e-06}
{"train_loss": 0.047844067215919495, "global_step": 242102, "epoch": 2720, "lr": 2.8667805626425404e-06}
{"train_loss": 0.09023630619049072, "global_step": 242103, "epoch": 2720, "lr": 2.8665870832464693e-06}
{"train_loss": 0.051184605807065964, "global_step": 242104, "epoch": 2720, "lr": 2.8663936101869017e-06}
{"train_loss": 0.05439917743206024, "global_step": 242105, "epoch": 2720, "lr": 2.866200143463882e-06}
{"train_loss": 0.07114097476005554, "global_step": 242106, "epoch": 2720, "lr": 2.866006683077416e-06}
{"train_loss": 0.07153227180242538, "global_step": 242107, "epoch": 2720, "lr": 2.865813229027553e-06}
{"train_loss": 0.04811292514204979, "global_step": 242108, "epoch": 2720, "lr": 2.8656197813142994e-06}
{"train_loss": 0.04573790356516838, "global_step": 242109, "epoch": 2720, "lr": 2.8654263399376935e-06}
{"train_loss": 0.061644501984119415, "global_step": 242110, "epoch": 2720, "lr": 2.8652329048977465e-06}
{"train_loss": 0.019482992589473724, "global_step": 242111, "epoch": 2720, "lr": 2.8650394761945086e-06}
{"train_loss": 0.031563613563776016, "global_step": 242112, "epoch": 2720, "lr": 2.86484605382798e-06}
{"train_loss": 0.05078578367829323, "global_step": 242113, "epoch": 2720, "lr": 2.8646526377981984e-06}
{"train_loss": 0.04560374468564987, "global_step": 242114, "epoch": 2720, "lr": 2.8644592281051987e-06}
{"train_loss": 0.030987141653895378, "global_step": 242115, "epoch": 2720, "lr": 2.8642658247489905e-06}
{"train_loss": 0.025230932980775833, "global_step": 242116, "epoch": 2720, "lr": 2.8640724277296137e-06}
{"train_loss": 0.026923656463623047, "global_step": 242117, "epoch": 2720, "lr": 2.8638790370470792e-06}
{"train_loss": 0.08323032408952713, "global_step": 242118, "epoch": 2720, "lr": 2.863685652701431e-06}
{"train_loss": 0.056154653429985046, "global_step": 242119, "epoch": 2720, "lr": 2.8634922746926806e-06}
{"train_loss": 0.0341201052069664, "global_step": 242120, "epoch": 2720, "lr": 2.8632989030208613e-06}
{"train_loss": 0.029389547184109688, "global_step": 242121, "epoch": 2720, "lr": 2.863105537685995e-06}
{"train_loss": 0.04747560992836952, "global_step": 242122, "epoch": 2720, "lr": 2.8629121786881152e-06}
{"train_loss": 0.05503968149423599, "global_step": 242123, "epoch": 2720, "lr": 2.862718826027233e-06}
{"train_loss": 0.05675651878118515, "global_step": 242124, "epoch": 2720, "lr": 2.862525479703393e-06}
{"train_loss": 0.07172136753797531, "global_step": 242125, "epoch": 2720, "lr": 2.862332139716606e-06}
{"train_loss": 0.05560184270143509, "global_step": 242126, "epoch": 2720, "lr": 2.862138806066905e-06}
{"train_loss": 0.06649333238601685, "global_step": 242127, "epoch": 2720, "lr": 2.861945478754313e-06}
{"train_loss": 0.05079417675733566, "global_step": 242128, "epoch": 2720, "lr": 2.861752157778863e-06}
{"train_loss": 0.04812689498066902, "global_step": 242129, "epoch": 2720, "lr": 2.8615588431405716e-06}
{"train_loss": 0.06923814862966537, "global_step": 242130, "epoch": 2720, "lr": 2.861365534839472e-06}
{"train_loss": 0.02433878183364868, "global_step": 242131, "epoch": 2720, "lr": 2.861172232875592e-06}
{"train_loss": 0.03938601538538933, "global_step": 242132, "epoch": 2720, "lr": 2.8609789372489425e-06}
{"train_loss": 0.03801484405994415, "global_step": 242133, "epoch": 2720, "lr": 2.8607856479595684e-06}
{"train_loss": 0.010741098783910275, "global_step": 242134, "epoch": 2720, "lr": 2.860592365007475e-06}
{"train_loss": 0.020154442638158798, "global_step": 242135, "epoch": 2720, "lr": 2.8603990883927123e-06}
{"train_loss": 0.0660238116979599, "global_step": 242136, "epoch": 2720, "lr": 2.8602058181152857e-06}
{"train_loss": 0.029378188773989677, "global_step": 242137, "epoch": 2720, "lr": 2.86001255417524e-06}
{"train_loss": 0.0627579540014267, "global_step": 242138, "epoch": 2720, "lr": 2.85981929657258e-06}
{"train_loss": 0.03968944028019905, "global_step": 242139, "epoch": 2720, "lr": 2.8596260453073565e-06}
{"train_loss": 0.03210555016994476, "global_step": 242140, "epoch": 2720, "lr": 2.859432800379569e-06}
{"train_loss": 0.07010484486818314, "global_step": 242141, "epoch": 2720, "lr": 2.859239561789262e-06}
{"train_loss": 0.0539110004901886, "global_step": 242142, "epoch": 2720, "lr": 2.859046329536452e-06}
{"train_loss": 0.06170910596847534, "global_step": 242143, "epoch": 2720, "lr": 2.858853103621173e-06}
{"train_loss": 0.0384279303252697, "global_step": 242144, "epoch": 2720, "lr": 2.858659884043441e-06}
{"train_loss": 0.027080588042736053, "global_step": 242145, "epoch": 2720, "lr": 2.858466670803295e-06}
{"train_loss": 0.04449405521154404, "global_step": 242146, "epoch": 2720, "lr": 2.8582734639007456e-06}
{"train_loss": 0.026234721764922142, "global_step": 242147, "epoch": 2720, "lr": 2.8580802633358328e-06}
{"train_loss": 0.0620451420545578, "global_step": 242148, "epoch": 2720, "lr": 2.8578870691085725e-06}
{"train_loss": 0.025258464738726616, "global_step": 242149, "epoch": 2720, "lr": 2.8576938812189922e-06}
{"train_loss": 0.05048995092511177, "global_step": 242150, "epoch": 2720, "lr": 2.8575006996671315e-06}
{"train_loss": 0.029744088649749756, "global_step": 242151, "epoch": 2720, "lr": 2.8573075244529957e-06}
{"train_loss": 0.06999385356903076, "global_step": 242152, "epoch": 2720, "lr": 2.8571143555766234e-06}
{"train_loss": 0.05570214241743088, "global_step": 242153, "epoch": 2720, "lr": 2.8569211930380367e-06}
{"train_loss": 0.023385345935821533, "global_step": 242154, "epoch": 2720, "lr": 2.856728036837264e-06}
{"train_loss": 0.08768337219953537, "global_step": 242155, "epoch": 2720, "lr": 2.856534886974327e-06}
{"train_loss": 0.1181592270731926, "global_step": 242156, "epoch": 2720, "lr": 2.856341743449259e-06}
{"train_loss": 0.021176878362894058, "global_step": 242157, "epoch": 2720, "lr": 2.8561486062620766e-06}
{"train_loss": 0.050684306770563126, "global_step": 242158, "epoch": 2720, "lr": 2.8559554754128136e-06}
{"train_loss": 0.05622761696577072, "global_step": 242159, "epoch": 2720, "lr": 2.8557623509014864e-06}
{"train_loss": 0.018412966281175613, "global_step": 242160, "epoch": 2720, "lr": 2.855569232728139e-06}
{"train_loss": 0.044553011655807495, "global_step": 242161, "epoch": 2720, "lr": 2.855376120892772e-06}
{"train_loss": 0.06465714424848557, "global_step": 242162, "epoch": 2720, "lr": 2.855183015395435e-06}
{"train_loss": 0.05228079482913017, "global_step": 242163, "epoch": 2720, "lr": 2.8549899162361393e-06}
{"train_loss": 0.04502847045660019, "global_step": 242164, "epoch": 2720, "lr": 2.8547968234149235e-06}
{"train_loss": 0.058704961091279984, "global_step": 242165, "epoch": 2720, "lr": 2.8546037369317936e-06}
{"train_loss": 0.034572407603263855, "global_step": 242166, "epoch": 2720, "lr": 2.854410656786799e-06}
{"train_loss": 0.03879595175385475, "global_step": 242167, "epoch": 2720, "lr": 2.8542175829799457e-06}
{"train_loss": 0.04903981936153736, "global_step": 242168, "epoch": 2720, "lr": 2.8540245155112723e-06, "val_loss": 9.416893005371094, "train_action_mse_error": 35.08244323730469}
{"train_loss": 0.03559501841664314, "global_step": 242169, "epoch": 2721, "lr": 2.853831454380801e-06}
{"train_loss": 0.05762983486056328, "global_step": 242170, "epoch": 2721, "lr": 2.8536383995885542e-06}
{"train_loss": 0.01788598671555519, "global_step": 242171, "epoch": 2721, "lr": 2.853445351134565e-06}
{"train_loss": 0.018924204632639885, "global_step": 242172, "epoch": 2721, "lr": 2.8532523090188503e-06}
{"train_loss": 0.0761820524930954, "global_step": 242173, "epoch": 2721, "lr": 2.8530592732414487e-06}
{"train_loss": 0.08046294003725052, "global_step": 242174, "epoch": 2721, "lr": 2.8528662438023713e-06}
{"train_loss": 0.05233660712838173, "global_step": 242175, "epoch": 2721, "lr": 2.852673220701657e-06}
{"train_loss": 0.05050121620297432, "global_step": 242176, "epoch": 2721, "lr": 2.8524802039393228e-06}
{"train_loss": 0.03920137137174606, "global_step": 242177, "epoch": 2721, "lr": 2.8522871935154016e-06}
{"train_loss": 0.043758831918239594, "global_step": 242178, "epoch": 2721, "lr": 2.85209418942991e-06}
{"train_loss": 0.06055654585361481, "global_step": 242179, "epoch": 2721, "lr": 2.851901191682882e-06}
{"train_loss": 0.05687480792403221, "global_step": 242180, "epoch": 2721, "lr": 2.851708200274339e-06}
{"train_loss": 0.07815462350845337, "global_step": 242181, "epoch": 2721, "lr": 2.8515152152043145e-06}
{"train_loss": 0.05397230014204979, "global_step": 242182, "epoch": 2721, "lr": 2.8513222364728255e-06}
{"train_loss": 0.07146279513835907, "global_step": 242183, "epoch": 2721, "lr": 2.851129264079905e-06}
{"train_loss": 0.05696961656212807, "global_step": 242184, "epoch": 2721, "lr": 2.8509362980255692e-06}
{"train_loss": 0.08064471185207367, "global_step": 242185, "epoch": 2721, "lr": 2.8507433383098527e-06}
{"train_loss": 0.05783873423933983, "global_step": 242186, "epoch": 2721, "lr": 2.8505503849327818e-06}
{"train_loss": 0.10231628268957138, "global_step": 242187, "epoch": 2721, "lr": 2.8503574378943744e-06}
{"train_loss": 0.044563405215740204, "global_step": 242188, "epoch": 2721, "lr": 2.850164497194668e-06}
{"train_loss": 0.06473725289106369, "global_step": 242189, "epoch": 2721, "lr": 2.849971562833681e-06}
{"train_loss": 0.03889603167772293, "global_step": 242190, "epoch": 2721, "lr": 2.84977863481144e-06}
{"train_loss": 0.04834287613630295, "global_step": 242191, "epoch": 2721, "lr": 2.8495857131279667e-06}
{"train_loss": 0.05531829968094826, "global_step": 242192, "epoch": 2721, "lr": 2.849392797783301e-06}
{"train_loss": 0.060939617455005646, "global_step": 242193, "epoch": 2721, "lr": 2.849199888777454e-06}
{"train_loss": 0.09319745004177094, "global_step": 242194, "epoch": 2721, "lr": 2.849006986110464e-06}
{"train_loss": 0.0493176244199276, "global_step": 242195, "epoch": 2721, "lr": 2.848814089782337e-06}
{"train_loss": 0.06836843490600586, "global_step": 242196, "epoch": 2721, "lr": 2.848621199793128e-06}
{"train_loss": 0.046825431287288666, "global_step": 242197, "epoch": 2721, "lr": 2.8484283161428317e-06}
{"train_loss": 0.029026255011558533, "global_step": 242198, "epoch": 2721, "lr": 2.8482354388315036e-06}
{"train_loss": 0.0298893004655838, "global_step": 242199, "epoch": 2721, "lr": 2.848042567859144e-06}
{"train_loss": 0.06582176685333252, "global_step": 242200, "epoch": 2721, "lr": 2.847849703225802e-06}
{"train_loss": 0.06670977920293808, "global_step": 242201, "epoch": 2721, "lr": 2.8476568449314788e-06}
{"train_loss": 0.047112029045820236, "global_step": 242202, "epoch": 2721, "lr": 2.8474639929762238e-06}
{"train_loss": 0.08883678168058395, "global_step": 242203, "epoch": 2721, "lr": 2.8472711473600477e-06}
{"train_loss": 0.05658871680498123, "global_step": 242204, "epoch": 2721, "lr": 2.8470783080829846e-06}
{"train_loss": 0.08447413891553879, "global_step": 242205, "epoch": 2721, "lr": 2.8468854751450503e-06}
{"train_loss": 0.07717211544513702, "global_step": 242206, "epoch": 2721, "lr": 2.84669264854629e-06}
{"train_loss": 0.036491844803094864, "global_step": 242207, "epoch": 2721, "lr": 2.846499828286703e-06}
{"train_loss": 0.034414466470479965, "global_step": 242208, "epoch": 2721, "lr": 2.8463070143663396e-06}
{"train_loss": 0.04117869958281517, "global_step": 242209, "epoch": 2721, "lr": 2.8461142067852163e-06}
{"train_loss": 0.03292591869831085, "global_step": 242210, "epoch": 2721, "lr": 2.8459214055433502e-06}
{"train_loss": 0.015492797829210758, "global_step": 242211, "epoch": 2721, "lr": 2.84572861064078e-06}
{"train_loss": 0.024975815787911415, "global_step": 242212, "epoch": 2721, "lr": 2.845535822077522e-06}
{"train_loss": 0.04318707808852196, "global_step": 242213, "epoch": 2721, "lr": 2.8453430398536153e-06}
{"train_loss": 0.09799535572528839, "global_step": 242214, "epoch": 2721, "lr": 2.845150263969071e-06}
{"train_loss": 0.03538599982857704, "global_step": 242215, "epoch": 2721, "lr": 2.8449574944239222e-06}
{"train_loss": 0.057787343859672546, "global_step": 242216, "epoch": 2721, "lr": 2.8447647312181914e-06}
{"train_loss": 0.02887125127017498, "global_step": 242217, "epoch": 2721, "lr": 2.8445719743519172e-06}
{"train_loss": 0.03856304660439491, "global_step": 242218, "epoch": 2721, "lr": 2.8443792238251053e-06}
{"train_loss": 0.08518139272928238, "global_step": 242219, "epoch": 2721, "lr": 2.8441864796378006e-06}
{"train_loss": 0.05324108153581619, "global_step": 242220, "epoch": 2721, "lr": 2.843993741790013e-06}
{"train_loss": 0.07435315102338791, "global_step": 242221, "epoch": 2721, "lr": 2.8438010102817713e-06}
{"train_loss": 0.058807797729969025, "global_step": 242222, "epoch": 2721, "lr": 2.8436082851131196e-06}
{"train_loss": 0.029343845322728157, "global_step": 242223, "epoch": 2721, "lr": 2.8434155662840577e-06}
{"train_loss": 0.0599515475332737, "global_step": 242224, "epoch": 2721, "lr": 2.8432228537946303e-06}
{"train_loss": 0.09642008692026138, "global_step": 242225, "epoch": 2721, "lr": 2.843030147644854e-06}
{"train_loss": 0.03911849856376648, "global_step": 242226, "epoch": 2721, "lr": 2.842837447834762e-06}
{"train_loss": 0.09227132797241211, "global_step": 242227, "epoch": 2721, "lr": 2.8426447543643707e-06}
{"train_loss": 0.030976027250289917, "global_step": 242228, "epoch": 2721, "lr": 2.8424520672337194e-06}
{"train_loss": 0.0353146567940712, "global_step": 242229, "epoch": 2721, "lr": 2.8422593864428137e-06}
{"train_loss": 0.02774963341653347, "global_step": 242230, "epoch": 2721, "lr": 2.842066711991703e-06}
{"train_loss": 0.05711042508482933, "global_step": 242231, "epoch": 2721, "lr": 2.841874043880388e-06}
{"train_loss": 0.0554455891251564, "global_step": 242232, "epoch": 2721, "lr": 2.841681382108924e-06}
{"train_loss": 0.020663803443312645, "global_step": 242233, "epoch": 2721, "lr": 2.8414887266773106e-06}
{"train_loss": 0.07954967021942139, "global_step": 242234, "epoch": 2721, "lr": 2.8412960775855868e-06}
{"train_loss": 0.04375775530934334, "global_step": 242235, "epoch": 2721, "lr": 2.841103434833775e-06}
{"train_loss": 0.05147626996040344, "global_step": 242236, "epoch": 2721, "lr": 2.8409107984219086e-06}
{"train_loss": 0.035158995538949966, "global_step": 242237, "epoch": 2721, "lr": 2.8407181683499983e-06}
{"train_loss": 0.04203600063920021, "global_step": 242238, "epoch": 2721, "lr": 2.8405255446180834e-06}
{"train_loss": 0.02780877985060215, "global_step": 242239, "epoch": 2721, "lr": 2.84033292722618e-06}
{"train_loss": 0.06515506654977798, "global_step": 242240, "epoch": 2721, "lr": 2.8401403161743167e-06}
{"train_loss": 0.047159187495708466, "global_step": 242241, "epoch": 2721, "lr": 2.8399477114625317e-06}
{"train_loss": 0.05108703672885895, "global_step": 242242, "epoch": 2721, "lr": 2.8397551130908418e-06}
{"train_loss": 0.04224369674921036, "global_step": 242243, "epoch": 2721, "lr": 2.8395625210592637e-06}
{"train_loss": 0.09320537000894547, "global_step": 242244, "epoch": 2721, "lr": 2.8393699353678416e-06}
{"train_loss": 0.07807192951440811, "global_step": 242245, "epoch": 2721, "lr": 2.8391773560165814e-06}
{"train_loss": 0.0336332768201828, "global_step": 242246, "epoch": 2721, "lr": 2.8389847830055216e-06}
{"train_loss": 0.05452757701277733, "global_step": 242247, "epoch": 2721, "lr": 2.83879221633469e-06}
{"train_loss": 0.03369731828570366, "global_step": 242248, "epoch": 2721, "lr": 2.838599656004104e-06}
{"train_loss": 0.034274596720933914, "global_step": 242249, "epoch": 2721, "lr": 2.838407102013796e-06}
{"train_loss": 0.07809283584356308, "global_step": 242250, "epoch": 2721, "lr": 2.8382145543637827e-06}
{"train_loss": 0.04391491785645485, "global_step": 242251, "epoch": 2721, "lr": 2.838022013054109e-06}
{"train_loss": 0.036164164543151855, "global_step": 242252, "epoch": 2721, "lr": 2.8378294780847746e-06}
{"train_loss": 0.07031818479299545, "global_step": 242253, "epoch": 2721, "lr": 2.8376369494558298e-06}
{"train_loss": 0.04004652798175812, "global_step": 242254, "epoch": 2721, "lr": 2.8374444271672797e-06}
{"train_loss": 0.062476690858602524, "global_step": 242255, "epoch": 2721, "lr": 2.8372519112191686e-06}
{"train_loss": 0.05273512005805969, "global_step": 242256, "epoch": 2721, "lr": 2.837059401611508e-06}
{"train_loss": 0.05334524620910374, "global_step": 242257, "epoch": 2721, "lr": 2.8368668983443315e-06, "val_loss": 9.384512901306152}
{"train_loss": 0.015277677215635777, "global_step": 242258, "epoch": 2722, "lr": 2.836674401417666e-06}
{"train_loss": 0.07023938000202179, "global_step": 242259, "epoch": 2722, "lr": 2.8364819108315345e-06}
{"train_loss": 0.03337610512971878, "global_step": 242260, "epoch": 2722, "lr": 2.836289426585964e-06}
{"train_loss": 0.049894459545612335, "global_step": 242261, "epoch": 2722, "lr": 2.8360969486809774e-06}
{"train_loss": 0.03735185042023659, "global_step": 242262, "epoch": 2722, "lr": 2.8359044771166075e-06}
{"train_loss": 0.030340394005179405, "global_step": 242263, "epoch": 2722, "lr": 2.8357120118928658e-06}
{"train_loss": 0.030980676412582397, "global_step": 242264, "epoch": 2722, "lr": 2.8355195530097967e-06}
{"train_loss": 0.06863018870353699, "global_step": 242265, "epoch": 2722, "lr": 2.835327100467411e-06}
{"train_loss": 0.027694305405020714, "global_step": 242266, "epoch": 2722, "lr": 2.8351346542657476e-06}
{"train_loss": 0.07235515117645264, "global_step": 242267, "epoch": 2722, "lr": 2.8349422144048176e-06}
{"train_loss": 0.04242390766739845, "global_step": 242268, "epoch": 2722, "lr": 2.83474978088466e-06}
{"train_loss": 0.06523447483778, "global_step": 242269, "epoch": 2722, "lr": 2.8345573537052914e-06}
{"train_loss": 0.055489033460617065, "global_step": 242270, "epoch": 2722, "lr": 2.834364932866751e-06}
{"train_loss": 0.02822311967611313, "global_step": 242271, "epoch": 2722, "lr": 2.8341725183690437e-06}
{"train_loss": 0.02582366019487381, "global_step": 242272, "epoch": 2722, "lr": 2.8339801102122143e-06}
{"train_loss": 0.012199063785374165, "global_step": 242273, "epoch": 2722, "lr": 2.833787708396274e-06}
{"train_loss": 0.04824689030647278, "global_step": 242274, "epoch": 2722, "lr": 2.8335953129212667e-06}
{"train_loss": 0.03195137903094292, "global_step": 242275, "epoch": 2722, "lr": 2.8334029237871985e-06}
{"train_loss": 0.04350217431783676, "global_step": 242276, "epoch": 2722, "lr": 2.8332105409941024e-06}
{"train_loss": 0.018519343808293343, "global_step": 242277, "epoch": 2722, "lr": 2.8330181645420173e-06}
{"train_loss": 0.04704734683036804, "global_step": 242278, "epoch": 2722, "lr": 2.8328257944309487e-06}
{"train_loss": 0.04184093326330185, "global_step": 242279, "epoch": 2722, "lr": 2.8326334306609416e-06}
{"train_loss": 0.0692119374871254, "global_step": 242280, "epoch": 2722, "lr": 2.832441073232006e-06}
{"train_loss": 0.043184325098991394, "global_step": 242281, "epoch": 2722, "lr": 2.8322487221441706e-06}
{"train_loss": 0.059633467346429825, "global_step": 242282, "epoch": 2722, "lr": 2.832056377397474e-06}
{"train_loss": 0.04449924826622009, "global_step": 242283, "epoch": 2722, "lr": 2.8318640389919215e-06}
{"train_loss": 0.029857557266950607, "global_step": 242284, "epoch": 2722, "lr": 2.831671706927558e-06}
{"train_loss": 0.03304905816912651, "global_step": 242285, "epoch": 2722, "lr": 2.831479381204405e-06}
{"train_loss": 0.027500834316015244, "global_step": 242286, "epoch": 2722, "lr": 2.831287061822474e-06}
{"train_loss": 0.016590381041169167, "global_step": 242287, "epoch": 2722, "lr": 2.8310947487818096e-06}
{"train_loss": 0.02825612761080265, "global_step": 242288, "epoch": 2722, "lr": 2.8309024420824227e-06}
{"train_loss": 0.0500146858394146, "global_step": 242289, "epoch": 2722, "lr": 2.830710141724352e-06}
{"train_loss": 0.05959019437432289, "global_step": 242290, "epoch": 2722, "lr": 2.8305178477076087e-06}
{"train_loss": 0.04091382771730423, "global_step": 242291, "epoch": 2722, "lr": 2.8303255600322374e-06}
{"train_loss": 0.04752591997385025, "global_step": 242292, "epoch": 2722, "lr": 2.8301332786982436e-06}
{"train_loss": 0.048077329993247986, "global_step": 242293, "epoch": 2722, "lr": 2.829941003705666e-06}
{"train_loss": 0.04239514470100403, "global_step": 242294, "epoch": 2722, "lr": 2.8297487350545325e-06}
{"train_loss": 0.0465460903942585, "global_step": 242295, "epoch": 2722, "lr": 2.8295564727448597e-06}
{"train_loss": 0.06672536581754684, "global_step": 242296, "epoch": 2722, "lr": 2.8293642167766866e-06}
{"train_loss": 0.06477241218090057, "global_step": 242297, "epoch": 2722, "lr": 2.829171967150024e-06}
{"train_loss": 0.06078776344656944, "global_step": 242298, "epoch": 2722, "lr": 2.828979723864905e-06}
{"train_loss": 0.041355330497026443, "global_step": 242299, "epoch": 2722, "lr": 2.8287874869213525e-06}
{"train_loss": 0.07144523411989212, "global_step": 242300, "epoch": 2722, "lr": 2.828595256319405e-06}
{"train_loss": 0.03144310787320137, "global_step": 242301, "epoch": 2722, "lr": 2.8284030320590626e-06}
{"train_loss": 0.05967913568019867, "global_step": 242302, "epoch": 2722, "lr": 2.828210814140381e-06}
{"train_loss": 0.03519423305988312, "global_step": 242303, "epoch": 2722, "lr": 2.8280186025633593e-06}
{"train_loss": 0.06168397143483162, "global_step": 242304, "epoch": 2722, "lr": 2.8278263973280484e-06}
{"train_loss": 0.08364808559417725, "global_step": 242305, "epoch": 2722, "lr": 2.827634198434448e-06}
{"train_loss": 0.03218725323677063, "global_step": 242306, "epoch": 2722, "lr": 2.827442005882608e-06}
{"train_loss": 0.08917967230081558, "global_step": 242307, "epoch": 2722, "lr": 2.8272498196725338e-06}
{"train_loss": 0.05845540389418602, "global_step": 242308, "epoch": 2722, "lr": 2.8270576398042702e-06}
{"train_loss": 0.07850858569145203, "global_step": 242309, "epoch": 2722, "lr": 2.826865466277823e-06}
{"train_loss": 0.0320599302649498, "global_step": 242310, "epoch": 2722, "lr": 2.826673299093241e-06}
{"train_loss": 0.06703530997037888, "global_step": 242311, "epoch": 2722, "lr": 2.826481138250525e-06}
{"train_loss": 0.044289276003837585, "global_step": 242312, "epoch": 2722, "lr": 2.8262889837497198e-06}
{"train_loss": 0.06827928870916367, "global_step": 242313, "epoch": 2722, "lr": 2.826096835590847e-06}
{"train_loss": 0.0333029106259346, "global_step": 242314, "epoch": 2722, "lr": 2.825904693773929e-06}
{"train_loss": 0.060655657202005386, "global_step": 242315, "epoch": 2722, "lr": 2.8257125582989983e-06}
{"train_loss": 0.03397522121667862, "global_step": 242316, "epoch": 2722, "lr": 2.8255204291660674e-06}
{"train_loss": 0.060927703976631165, "global_step": 242317, "epoch": 2722, "lr": 2.8253283063751745e-06}
{"train_loss": 0.04293664172291756, "global_step": 242318, "epoch": 2722, "lr": 2.825136189926336e-06}
{"train_loss": 0.04738622158765793, "global_step": 242319, "epoch": 2722, "lr": 2.8249440798195914e-06}
{"train_loss": 0.024073703214526176, "global_step": 242320, "epoch": 2722, "lr": 2.8247519760549567e-06}
{"train_loss": 0.05106041952967644, "global_step": 242321, "epoch": 2722, "lr": 2.8245598786324545e-06}
{"train_loss": 0.06461530923843384, "global_step": 242322, "epoch": 2722, "lr": 2.824367787552118e-06}
{"train_loss": 0.0499795563519001, "global_step": 242323, "epoch": 2722, "lr": 2.824175702813969e-06}
{"train_loss": 0.04180988669395447, "global_step": 242324, "epoch": 2722, "lr": 2.823983624418036e-06}
{"train_loss": 0.024022506549954414, "global_step": 242325, "epoch": 2722, "lr": 2.8237915523643464e-06}
{"train_loss": 0.07736703008413315, "global_step": 242326, "epoch": 2722, "lr": 2.823599486652917e-06}
{"train_loss": 0.04374295473098755, "global_step": 242327, "epoch": 2722, "lr": 2.8234074272837864e-06}
{"train_loss": 0.01814896985888481, "global_step": 242328, "epoch": 2722, "lr": 2.8232153742569657e-06}
{"train_loss": 0.08248458057641983, "global_step": 242329, "epoch": 2722, "lr": 2.823023327572494e-06}
{"train_loss": 0.030720049515366554, "global_step": 242330, "epoch": 2722, "lr": 2.8228312872303874e-06}
{"train_loss": 0.06070538982748985, "global_step": 242331, "epoch": 2722, "lr": 2.8226392532306745e-06}
{"train_loss": 0.03914351016283035, "global_step": 242332, "epoch": 2722, "lr": 2.822447225573388e-06}
{"train_loss": 0.02835226058959961, "global_step": 242333, "epoch": 2722, "lr": 2.8222552042585447e-06}
{"train_loss": 0.04317718371748924, "global_step": 242334, "epoch": 2722, "lr": 2.8220631892861784e-06}
{"train_loss": 0.049645550549030304, "global_step": 242335, "epoch": 2722, "lr": 2.8218711806563105e-06}
{"train_loss": 0.05597926303744316, "global_step": 242336, "epoch": 2722, "lr": 2.821679178368969e-06}
{"train_loss": 0.07339230924844742, "global_step": 242337, "epoch": 2722, "lr": 2.821487182424165e-06}
{"train_loss": 0.04593415930867195, "global_step": 242338, "epoch": 2722, "lr": 2.8212951928219543e-06}
{"train_loss": 0.049125924706459045, "global_step": 242339, "epoch": 2722, "lr": 2.8211032095623313e-06}
{"train_loss": 0.034287672489881516, "global_step": 242340, "epoch": 2722, "lr": 2.8209112326453456e-06}
{"train_loss": 0.061979807913303375, "global_step": 242341, "epoch": 2722, "lr": 2.8207192620710086e-06}
{"train_loss": 0.055900536477565765, "global_step": 242342, "epoch": 2722, "lr": 2.820527297839354e-06}
{"train_loss": 0.0685163363814354, "global_step": 242343, "epoch": 2722, "lr": 2.8203353399503975e-06}
{"train_loss": 0.04117036610841751, "global_step": 242344, "epoch": 2722, "lr": 2.820143388404184e-06}
{"train_loss": 0.03519701212644577, "global_step": 242345, "epoch": 2722, "lr": 2.8199514432007136e-06}
{"train_loss": 0.0469834916712193, "global_step": 242346, "epoch": 2722, "lr": 2.819759504340036e-06, "val_loss": 9.441819190979004}
{"train_loss": 0.016567029058933258, "global_step": 242347, "epoch": 2723, "lr": 2.819567571822157e-06}
{"train_loss": 0.05810119956731796, "global_step": 242348, "epoch": 2723, "lr": 2.8193756456471155e-06}
{"train_loss": 0.012678309343755245, "global_step": 242349, "epoch": 2723, "lr": 2.819183725814939e-06}
{"train_loss": 0.07406444847583771, "global_step": 242350, "epoch": 2723, "lr": 2.8189918123256443e-06}
{"train_loss": 0.05364573374390602, "global_step": 242351, "epoch": 2723, "lr": 2.8187999051792648e-06}
{"train_loss": 0.061112843453884125, "global_step": 242352, "epoch": 2723, "lr": 2.818608004375817e-06}
{"train_loss": 0.07437245547771454, "global_step": 242353, "epoch": 2723, "lr": 2.8184161099153392e-06}
{"train_loss": 0.04725268855690956, "global_step": 242354, "epoch": 2723, "lr": 2.8182242217978437e-06}
{"train_loss": 0.0956561416387558, "global_step": 242355, "epoch": 2723, "lr": 2.8180323400233687e-06}
{"train_loss": 0.022225016728043556, "global_step": 242356, "epoch": 2723, "lr": 2.8178404645919254e-06}
{"train_loss": 0.0524066723883152, "global_step": 242357, "epoch": 2723, "lr": 2.8176485955035582e-06}
{"train_loss": 0.04056685045361519, "global_step": 242358, "epoch": 2723, "lr": 2.8174567327582837e-06}
{"train_loss": 0.023705117404460907, "global_step": 242359, "epoch": 2723, "lr": 2.8172648763561182e-06}
{"train_loss": 0.036883920431137085, "global_step": 242360, "epoch": 2723, "lr": 2.8170730262971013e-06}
{"train_loss": 0.045082397758960724, "global_step": 242361, "epoch": 2723, "lr": 2.816881182581249e-06}
{"train_loss": 0.02291799709200859, "global_step": 242362, "epoch": 2723, "lr": 2.8166893452086008e-06}
{"train_loss": 0.018313754349946976, "global_step": 242363, "epoch": 2723, "lr": 2.8164975141791726e-06}
{"train_loss": 0.032588642090559006, "global_step": 242364, "epoch": 2723, "lr": 2.8163056894929874e-06}
{"train_loss": 0.0689082071185112, "global_step": 242365, "epoch": 2723, "lr": 2.8161138711500778e-06}
{"train_loss": 0.05231364443898201, "global_step": 242366, "epoch": 2723, "lr": 2.815922059150461e-06}
{"train_loss": 0.042338885366916656, "global_step": 242367, "epoch": 2723, "lr": 2.8157302534941645e-06}
{"train_loss": 0.0464307926595211, "global_step": 242368, "epoch": 2723, "lr": 2.815538454181227e-06}
{"train_loss": 0.06414458900690079, "global_step": 242369, "epoch": 2723, "lr": 2.8153466612116597e-06}
{"train_loss": 0.05274566635489464, "global_step": 242370, "epoch": 2723, "lr": 2.8151548745855015e-06}
{"train_loss": 0.06962428987026215, "global_step": 242371, "epoch": 2723, "lr": 2.8149630943027638e-06}
{"train_loss": 0.06540684401988983, "global_step": 242372, "epoch": 2723, "lr": 2.814771320363485e-06}
{"train_loss": 0.033383652567863464, "global_step": 242373, "epoch": 2723, "lr": 2.8145795527676766e-06}
{"train_loss": 0.04089422896504402, "global_step": 242374, "epoch": 2723, "lr": 2.8143877915153827e-06}
{"train_loss": 0.10216835141181946, "global_step": 242375, "epoch": 2723, "lr": 2.814196036606609e-06}
{"train_loss": 0.03490620478987694, "global_step": 242376, "epoch": 2723, "lr": 2.8140042880413995e-06}
{"train_loss": 0.03675834462046623, "global_step": 242377, "epoch": 2723, "lr": 2.8138125458197663e-06}
{"train_loss": 0.02910570055246353, "global_step": 242378, "epoch": 2723, "lr": 2.813620809941747e-06}
{"train_loss": 0.031434327363967896, "global_step": 242379, "epoch": 2723, "lr": 2.8134290804073594e-06}
{"train_loss": 0.028376005589962006, "global_step": 242380, "epoch": 2723, "lr": 2.8132373572166303e-06}
{"train_loss": 0.058164916932582855, "global_step": 242381, "epoch": 2723, "lr": 2.8130456403695825e-06}
{"train_loss": 0.033385880291461945, "global_step": 242382, "epoch": 2723, "lr": 2.812853929866255e-06}
{"train_loss": 0.03976434841752052, "global_step": 242383, "epoch": 2723, "lr": 2.8126622257066528e-06}
{"train_loss": 0.05419044569134712, "global_step": 242384, "epoch": 2723, "lr": 2.812470527890815e-06}
{"train_loss": 0.01146661676466465, "global_step": 242385, "epoch": 2723, "lr": 2.812278836418775e-06}
{"train_loss": 0.04900132492184639, "global_step": 242386, "epoch": 2723, "lr": 2.8120871512905376e-06}
{"train_loss": 0.019020894542336464, "global_step": 242387, "epoch": 2723, "lr": 2.8118954725061486e-06}
{"train_loss": 0.05719197541475296, "global_step": 242388, "epoch": 2723, "lr": 2.8117038000656183e-06}
{"train_loss": 0.07697305828332901, "global_step": 242389, "epoch": 2723, "lr": 2.8115121339689853e-06}
{"train_loss": 0.04912225529551506, "global_step": 242390, "epoch": 2723, "lr": 2.8113204742162615e-06}
{"train_loss": 0.05007019639015198, "global_step": 242391, "epoch": 2723, "lr": 2.8111288208074905e-06}
{"train_loss": 0.024641307070851326, "global_step": 242392, "epoch": 2723, "lr": 2.810937173742678e-06}
{"train_loss": 0.02636081352829933, "global_step": 242393, "epoch": 2723, "lr": 2.810745533021869e-06}
{"train_loss": 0.03023627959191799, "global_step": 242394, "epoch": 2723, "lr": 2.810553898645074e-06}
{"train_loss": 0.10061110556125641, "global_step": 242395, "epoch": 2723, "lr": 2.8103622706123323e-06}
{"train_loss": 0.03302045911550522, "global_step": 242396, "epoch": 2723, "lr": 2.8101706489236603e-06}
{"train_loss": 0.05893328785896301, "global_step": 242397, "epoch": 2723, "lr": 2.80997903357908e-06}
{"train_loss": 0.09209053218364716, "global_step": 242398, "epoch": 2723, "lr": 2.8097874245786305e-06}
{"train_loss": 0.043748509138822556, "global_step": 242399, "epoch": 2723, "lr": 2.8095958219223174e-06}
{"train_loss": 0.026727396994829178, "global_step": 242400, "epoch": 2723, "lr": 2.8094042256101904e-06}
{"train_loss": 0.031631384044885635, "global_step": 242401, "epoch": 2723, "lr": 2.8092126356422665e-06}
{"train_loss": 0.025258691981434822, "global_step": 242402, "epoch": 2723, "lr": 2.8090210520185566e-06}
{"train_loss": 0.04078846797347069, "global_step": 242403, "epoch": 2723, "lr": 2.8088294747391e-06}
{"train_loss": 0.06181608512997627, "global_step": 242404, "epoch": 2723, "lr": 2.808637903803929e-06}
{"train_loss": 0.06270887702703476, "global_step": 242405, "epoch": 2723, "lr": 2.8084463392130556e-06}
{"train_loss": 0.03738631680607796, "global_step": 242406, "epoch": 2723, "lr": 2.8082547809665126e-06}
{"train_loss": 0.0770028755068779, "global_step": 242407, "epoch": 2723, "lr": 2.8080632290643226e-06}
{"train_loss": 0.056039538234472275, "global_step": 242408, "epoch": 2723, "lr": 2.8078716835065246e-06}
{"train_loss": 0.04727032035589218, "global_step": 242409, "epoch": 2723, "lr": 2.8076801442931177e-06}
{"train_loss": 0.05095909535884857, "global_step": 242410, "epoch": 2723, "lr": 2.8074886114241527e-06}
{"train_loss": 0.038155313581228256, "global_step": 242411, "epoch": 2723, "lr": 2.8072970848996407e-06}
{"train_loss": 0.10310402512550354, "global_step": 242412, "epoch": 2723, "lr": 2.8071055647196198e-06}
{"train_loss": 0.04603271931409836, "global_step": 242413, "epoch": 2723, "lr": 2.8069140508840963e-06}
{"train_loss": 0.06328999251127243, "global_step": 242414, "epoch": 2723, "lr": 2.8067225433931197e-06}
{"train_loss": 0.04045141860842705, "global_step": 242415, "epoch": 2723, "lr": 2.806531042246696e-06}
{"train_loss": 0.08860083669424057, "global_step": 242416, "epoch": 2723, "lr": 2.806339547444864e-06}
{"train_loss": 0.06859421730041504, "global_step": 242417, "epoch": 2723, "lr": 2.8061480589876397e-06}
{"train_loss": 0.05010656267404556, "global_step": 242418, "epoch": 2723, "lr": 2.805956576875063e-06}
{"train_loss": 0.033882129937410355, "global_step": 242419, "epoch": 2723, "lr": 2.8057651011071385e-06}
{"train_loss": 0.07888522744178772, "global_step": 242420, "epoch": 2723, "lr": 2.8055736316839055e-06}
{"train_loss": 0.027492249384522438, "global_step": 242421, "epoch": 2723, "lr": 2.8053821686053918e-06}
{"train_loss": 0.04976571723818779, "global_step": 242422, "epoch": 2723, "lr": 2.805190711871619e-06}
{"train_loss": 0.05945553258061409, "global_step": 242423, "epoch": 2723, "lr": 2.804999261482616e-06}
{"train_loss": 0.042942654341459274, "global_step": 242424, "epoch": 2723, "lr": 2.8048078174383985e-06}
{"train_loss": 0.07428918778896332, "global_step": 242425, "epoch": 2723, "lr": 2.8046163797390058e-06}
{"train_loss": 0.036153022199869156, "global_step": 242426, "epoch": 2723, "lr": 2.804424948384449e-06}
{"train_loss": 0.06970673054456711, "global_step": 242427, "epoch": 2723, "lr": 2.8042335233747718e-06}
{"train_loss": 0.05764852836728096, "global_step": 242428, "epoch": 2723, "lr": 2.8040421047099864e-06}
{"train_loss": 0.04689783230423927, "global_step": 242429, "epoch": 2723, "lr": 2.8038506923901254e-06}
{"train_loss": 0.05004178360104561, "global_step": 242430, "epoch": 2723, "lr": 2.8036592864152e-06}
{"train_loss": 0.10097703337669373, "global_step": 242431, "epoch": 2723, "lr": 2.8034678867852602e-06}
{"train_loss": 0.03174465149641037, "global_step": 242432, "epoch": 2723, "lr": 2.803276493500312e-06}
{"train_loss": 0.022240977734327316, "global_step": 242433, "epoch": 2723, "lr": 2.8030851065603935e-06}
{"train_loss": 0.024922000244259834, "global_step": 242434, "epoch": 2723, "lr": 2.8028937259655162e-06}
{"train_loss": 0.04876733427930079, "global_step": 242435, "epoch": 2723, "lr": 2.8027023517157247e-06, "val_loss": 9.43420124053955}
{"train_loss": 0.13000880181789398, "global_step": 242436, "epoch": 2724, "lr": 2.8025109838110353e-06}
{"train_loss": 0.038412462919950485, "global_step": 242437, "epoch": 2724, "lr": 2.8023196222514647e-06}
{"train_loss": 0.04590887576341629, "global_step": 242438, "epoch": 2724, "lr": 2.8021282670370517e-06}
{"train_loss": 0.035466622561216354, "global_step": 242439, "epoch": 2724, "lr": 2.8019369181678134e-06}
{"train_loss": 0.05367865785956383, "global_step": 242440, "epoch": 2724, "lr": 2.8017455756437828e-06}
{"train_loss": 0.06988169997930527, "global_step": 242441, "epoch": 2724, "lr": 2.8015542394649763e-06}
{"train_loss": 0.06570127606391907, "global_step": 242442, "epoch": 2724, "lr": 2.801362909631433e-06}
{"train_loss": 0.053505029529333115, "global_step": 242443, "epoch": 2724, "lr": 2.8011715861431698e-06}
{"train_loss": 0.02775046043097973, "global_step": 242444, "epoch": 2724, "lr": 2.800980269000214e-06}
{"train_loss": 0.108788400888443, "global_step": 242445, "epoch": 2724, "lr": 2.8007889582025827e-06}
{"train_loss": 0.012151632457971573, "global_step": 242446, "epoch": 2724, "lr": 2.8005976537503253e-06}
{"train_loss": 0.01766868680715561, "global_step": 242447, "epoch": 2724, "lr": 2.800406355643437e-06}
{"train_loss": 0.0714658573269844, "global_step": 242448, "epoch": 2724, "lr": 2.800215063881972e-06}
{"train_loss": 0.06205874681472778, "global_step": 242449, "epoch": 2724, "lr": 2.800023778465932e-06}
{"train_loss": 0.04736277088522911, "global_step": 242450, "epoch": 2724, "lr": 2.799832499395361e-06}
{"train_loss": 0.03267589211463928, "global_step": 242451, "epoch": 2724, "lr": 2.7996412266702687e-06}
{"train_loss": 0.08503970503807068, "global_step": 242452, "epoch": 2724, "lr": 2.799449960290701e-06}
{"train_loss": 0.060014478862285614, "global_step": 242453, "epoch": 2724, "lr": 2.799258700256663e-06}
{"train_loss": 0.02103392593562603, "global_step": 242454, "epoch": 2724, "lr": 2.7990674465681934e-06}
{"train_loss": 0.04188983887434006, "global_step": 242455, "epoch": 2724, "lr": 2.798876199225309e-06}
{"train_loss": 0.0792577862739563, "global_step": 242456, "epoch": 2724, "lr": 2.7986849582280482e-06}
{"train_loss": 0.06038246676325798, "global_step": 242457, "epoch": 2724, "lr": 2.798493723576423e-06}
{"train_loss": 0.03584417328238487, "global_step": 242458, "epoch": 2724, "lr": 2.798302495270461e-06}
{"train_loss": 0.09209427237510681, "global_step": 242459, "epoch": 2724, "lr": 2.7981112733102e-06}
{"train_loss": 0.03577503189444542, "global_step": 242460, "epoch": 2724, "lr": 2.7979200576956522e-06}
{"train_loss": 0.040432270616292953, "global_step": 242461, "epoch": 2724, "lr": 2.797728848426856e-06}
{"train_loss": 0.032157812267541885, "global_step": 242462, "epoch": 2724, "lr": 2.797537645503823e-06}
{"train_loss": 0.07370805740356445, "global_step": 242463, "epoch": 2724, "lr": 2.797346448926591e-06}
{"train_loss": 0.07921898365020752, "global_step": 242464, "epoch": 2724, "lr": 2.797155258695172e-06}
{"train_loss": 0.06064693629741669, "global_step": 242465, "epoch": 2724, "lr": 2.7969640748096106e-06}
{"train_loss": 0.07900331169366837, "global_step": 242466, "epoch": 2724, "lr": 2.796772897269917e-06}
{"train_loss": 0.04860883206129074, "global_step": 242467, "epoch": 2724, "lr": 2.796581726076125e-06}
{"train_loss": 0.020758898928761482, "global_step": 242468, "epoch": 2724, "lr": 2.7963905612282516e-06}
{"train_loss": 0.04248020052909851, "global_step": 242469, "epoch": 2724, "lr": 2.796199402726335e-06}
{"train_loss": 0.048548225313425064, "global_step": 242470, "epoch": 2724, "lr": 2.7960082505703865e-06}
{"train_loss": 0.07003311812877655, "global_step": 242471, "epoch": 2724, "lr": 2.795817104760445e-06}
{"train_loss": 0.0627027377486229, "global_step": 242472, "epoch": 2724, "lr": 2.7956259652965276e-06}
{"train_loss": 0.0583394393324852, "global_step": 242473, "epoch": 2724, "lr": 2.7954348321786673e-06}
{"train_loss": 0.04169118404388428, "global_step": 242474, "epoch": 2724, "lr": 2.795243705406886e-06}
{"train_loss": 0.03882940113544464, "global_step": 242475, "epoch": 2724, "lr": 2.7950525849812005e-06}
{"train_loss": 0.050751686096191406, "global_step": 242476, "epoch": 2724, "lr": 2.7948614709016495e-06}
{"train_loss": 0.05193686485290527, "global_step": 242477, "epoch": 2724, "lr": 2.7946703631682502e-06}
{"train_loss": 0.05814177915453911, "global_step": 242478, "epoch": 2724, "lr": 2.794479261781041e-06}
{"train_loss": 0.07278653979301453, "global_step": 242479, "epoch": 2724, "lr": 2.7942881667400333e-06}
{"train_loss": 0.10578975081443787, "global_step": 242480, "epoch": 2724, "lr": 2.79409707804526e-06}
{"train_loss": 0.030713895335793495, "global_step": 242481, "epoch": 2724, "lr": 2.7939059956967383e-06}
{"train_loss": 0.03458922728896141, "global_step": 242482, "epoch": 2724, "lr": 2.7937149196945124e-06}
{"train_loss": 0.08065671473741531, "global_step": 242483, "epoch": 2724, "lr": 2.793523850038582e-06}
{"train_loss": 0.037009771913290024, "global_step": 242484, "epoch": 2724, "lr": 2.793332786728997e-06}
{"train_loss": 0.08442533016204834, "global_step": 242485, "epoch": 2724, "lr": 2.793141729765769e-06}
{"train_loss": 0.05256419628858566, "global_step": 242486, "epoch": 2724, "lr": 2.7929506791489314e-06}
{"train_loss": 0.039331451058387756, "global_step": 242487, "epoch": 2724, "lr": 2.7927596348785003e-06}
{"train_loss": 0.12362179160118103, "global_step": 242488, "epoch": 2724, "lr": 2.7925685969545144e-06}
{"train_loss": 0.05918039754033089, "global_step": 242489, "epoch": 2724, "lr": 2.7923775653769855e-06}
{"train_loss": 0.04450363293290138, "global_step": 242490, "epoch": 2724, "lr": 2.7921865401459523e-06}
{"train_loss": 0.06879740208387375, "global_step": 242491, "epoch": 2724, "lr": 2.791995521261431e-06}
{"train_loss": 0.03413219749927521, "global_step": 242492, "epoch": 2724, "lr": 2.791804508723456e-06}
{"train_loss": 0.0260574072599411, "global_step": 242493, "epoch": 2724, "lr": 2.791613502532037e-06}
{"train_loss": 0.038592346012592316, "global_step": 242494, "epoch": 2724, "lr": 2.7914225026872132e-06}
{"train_loss": 0.03620479628443718, "global_step": 242495, "epoch": 2724, "lr": 2.7912315091890128e-06}
{"train_loss": 0.03770078346133232, "global_step": 242496, "epoch": 2724, "lr": 2.7910405220374468e-06}
{"train_loss": 0.05734388157725334, "global_step": 242497, "epoch": 2724, "lr": 2.7908495412325597e-06}
{"train_loss": 0.03126392140984535, "global_step": 242498, "epoch": 2724, "lr": 2.7906585667743622e-06}
{"train_loss": 0.05722542852163315, "global_step": 242499, "epoch": 2724, "lr": 2.790467598662888e-06}
{"train_loss": 0.09413320571184158, "global_step": 242500, "epoch": 2724, "lr": 2.7902766368981537e-06}
{"train_loss": 0.049228716641664505, "global_step": 242501, "epoch": 2724, "lr": 2.7900856814802035e-06}
{"train_loss": 0.05626314878463745, "global_step": 242502, "epoch": 2724, "lr": 2.7898947324090374e-06}
{"train_loss": 0.05569476634263992, "global_step": 242503, "epoch": 2724, "lr": 2.7897037896847057e-06}
{"train_loss": 0.03686167299747467, "global_step": 242504, "epoch": 2724, "lr": 2.7895128533072134e-06}
{"train_loss": 0.0840558409690857, "global_step": 242505, "epoch": 2724, "lr": 2.7893219232766055e-06}
{"train_loss": 0.03844202682375908, "global_step": 242506, "epoch": 2724, "lr": 2.789130999592887e-06}
{"train_loss": 0.0275146272033453, "global_step": 242507, "epoch": 2724, "lr": 2.788940082256103e-06}
{"train_loss": 0.055020228028297424, "global_step": 242508, "epoch": 2724, "lr": 2.788749171266264e-06}
{"train_loss": 0.04214715585112572, "global_step": 242509, "epoch": 2724, "lr": 2.7885582666234144e-06}
{"train_loss": 0.024489304050803185, "global_step": 242510, "epoch": 2724, "lr": 2.7883673683275545e-06}
{"train_loss": 0.029459645971655846, "global_step": 242511, "epoch": 2724, "lr": 2.788176476378729e-06}
{"train_loss": 0.05859767273068428, "global_step": 242512, "epoch": 2724, "lr": 2.787985590776959e-06}
{"train_loss": 0.09699132293462753, "global_step": 242513, "epoch": 2724, "lr": 2.787794711522268e-06}
{"train_loss": 0.02158966287970543, "global_step": 242514, "epoch": 2724, "lr": 2.787603838614683e-06}
{"train_loss": 0.05692695081233978, "global_step": 242515, "epoch": 2724, "lr": 2.787412972054221e-06}
{"train_loss": 0.05086985230445862, "global_step": 242516, "epoch": 2724, "lr": 2.787222111840926e-06}
{"train_loss": 0.036125726997852325, "global_step": 242517, "epoch": 2724, "lr": 2.787031257974809e-06}
{"train_loss": 0.01671905256807804, "global_step": 242518, "epoch": 2724, "lr": 2.7868404104559045e-06}
{"train_loss": 0.05554423853754997, "global_step": 242519, "epoch": 2724, "lr": 2.7866495692842277e-06}
{"train_loss": 0.04912764951586723, "global_step": 242520, "epoch": 2724, "lr": 2.786458734459818e-06}
{"train_loss": 0.02444857358932495, "global_step": 242521, "epoch": 2724, "lr": 2.7862679059826813e-06}
{"train_loss": 0.0796600952744484, "global_step": 242522, "epoch": 2724, "lr": 2.786077083852867e-06}
{"train_loss": 0.05883723869919777, "global_step": 242523, "epoch": 2724, "lr": 2.7858862680703813e-06}
{"train_loss": 0.05316028149610155, "global_step": 242524, "epoch": 2724, "lr": 2.785695458635268e-06, "val_loss": 9.517878532409668}
{"train_loss": 0.01687915250658989, "global_step": 242525, "epoch": 2725, "lr": 2.785504655547527e-06}
{"train_loss": 0.041294001042842865, "global_step": 242526, "epoch": 2725, "lr": 2.7853138588072147e-06}
{"train_loss": 0.057962022721767426, "global_step": 242527, "epoch": 2725, "lr": 2.7851230684143305e-06}
{"train_loss": 0.018618065863847733, "global_step": 242528, "epoch": 2725, "lr": 2.7849322843689184e-06}
{"train_loss": 0.04582541063427925, "global_step": 242529, "epoch": 2725, "lr": 2.7847415066709905e-06}
{"train_loss": 0.04123787581920624, "global_step": 242530, "epoch": 2725, "lr": 2.784550735320579e-06}
{"train_loss": 0.05126095190644264, "global_step": 242531, "epoch": 2725, "lr": 2.784359970317718e-06}
{"train_loss": 0.07268363237380981, "global_step": 242532, "epoch": 2725, "lr": 2.784169211662413e-06}
{"train_loss": 0.06957698613405228, "global_step": 242533, "epoch": 2725, "lr": 2.783978459354708e-06}
{"train_loss": 0.08016956597566605, "global_step": 242534, "epoch": 2725, "lr": 2.78378771339462e-06}
{"train_loss": 0.09832698851823807, "global_step": 242535, "epoch": 2725, "lr": 2.7835969737821767e-06}
{"train_loss": 0.02048601396381855, "global_step": 242536, "epoch": 2725, "lr": 2.7834062405174e-06}
{"train_loss": 0.06246145814657211, "global_step": 242537, "epoch": 2725, "lr": 2.7832155136003233e-06}
{"train_loss": 0.06269227713346481, "global_step": 242538, "epoch": 2725, "lr": 2.783024793030964e-06}
{"train_loss": 0.05065973475575447, "global_step": 242539, "epoch": 2725, "lr": 2.78283407880936e-06}
{"train_loss": 0.04650615155696869, "global_step": 242540, "epoch": 2725, "lr": 2.7826433709355172e-06}
{"train_loss": 0.016199925914406776, "global_step": 242541, "epoch": 2725, "lr": 2.78245266940948e-06}
{"train_loss": 0.022777574136853218, "global_step": 242542, "epoch": 2725, "lr": 2.7822619742312594e-06}
{"train_loss": 0.07058104127645493, "global_step": 242543, "epoch": 2725, "lr": 2.7820712854009e-06}
{"train_loss": 0.025180811062455177, "global_step": 242544, "epoch": 2725, "lr": 2.781880602918402e-06}
{"train_loss": 0.05834266543388367, "global_step": 242545, "epoch": 2725, "lr": 2.7816899267838147e-06}
{"train_loss": 0.06359610706567764, "global_step": 242546, "epoch": 2725, "lr": 2.78149925699715e-06}
{"train_loss": 0.024278096854686737, "global_step": 242547, "epoch": 2725, "lr": 2.7813085935584405e-06}
{"train_loss": 0.037127744406461716, "global_step": 242548, "epoch": 2725, "lr": 2.7811179364676975e-06}
{"train_loss": 0.059035107493400574, "global_step": 242549, "epoch": 2725, "lr": 2.7809272857249714e-06}
{"train_loss": 0.05472990497946739, "global_step": 242550, "epoch": 2725, "lr": 2.780736641330267e-06}
{"train_loss": 0.06081467121839523, "global_step": 242551, "epoch": 2725, "lr": 2.7805460032836184e-06}
{"train_loss": 0.0296531580388546, "global_step": 242552, "epoch": 2725, "lr": 2.780355371585053e-06}
{"train_loss": 0.039370086044073105, "global_step": 242553, "epoch": 2725, "lr": 2.780164746234587e-06}
{"train_loss": 0.09143731743097305, "global_step": 242554, "epoch": 2725, "lr": 2.7799741272322597e-06}
{"train_loss": 0.06413193047046661, "global_step": 242555, "epoch": 2725, "lr": 2.779783514578083e-06}
{"train_loss": 0.024843014776706696, "global_step": 242556, "epoch": 2725, "lr": 2.779592908272094e-06}
{"train_loss": 0.044066231697797775, "global_step": 242557, "epoch": 2725, "lr": 2.779402308314305e-06}
{"train_loss": 0.03496036306023598, "global_step": 242558, "epoch": 2725, "lr": 2.77921171470476e-06}
{"train_loss": 0.045383796095848083, "global_step": 242559, "epoch": 2725, "lr": 2.779021127443465e-06}
{"train_loss": 0.04707121476531029, "global_step": 242560, "epoch": 2725, "lr": 2.778830546530464e-06}
{"train_loss": 0.03474796563386917, "global_step": 242561, "epoch": 2725, "lr": 2.778639971965763e-06}
{"train_loss": 0.07743280380964279, "global_step": 242562, "epoch": 2725, "lr": 2.7784494037494112e-06}
{"train_loss": 0.05321573466062546, "global_step": 242563, "epoch": 2725, "lr": 2.778258841881409e-06}
{"train_loss": 0.07346966862678528, "global_step": 242564, "epoch": 2725, "lr": 2.7780682863618013e-06}
{"train_loss": 0.01807643286883831, "global_step": 242565, "epoch": 2725, "lr": 2.777877737190604e-06}
{"train_loss": 0.06080768629908562, "global_step": 242566, "epoch": 2725, "lr": 2.7776871943678453e-06}
{"train_loss": 0.10321298986673355, "global_step": 242567, "epoch": 2725, "lr": 2.7774966578935526e-06}
{"train_loss": 0.03447055071592331, "global_step": 242568, "epoch": 2725, "lr": 2.777306127767748e-06}
{"train_loss": 0.01920936070382595, "global_step": 242569, "epoch": 2725, "lr": 2.7771156039904657e-06}
{"train_loss": 0.04946327954530716, "global_step": 242570, "epoch": 2725, "lr": 2.7769250865617156e-06}
{"train_loss": 0.03350326418876648, "global_step": 242571, "epoch": 2725, "lr": 2.7767345754815376e-06}
{"train_loss": 0.10131000727415085, "global_step": 242572, "epoch": 2725, "lr": 2.7765440707499477e-06}
{"train_loss": 0.06800654530525208, "global_step": 242573, "epoch": 2725, "lr": 2.7763535723669854e-06}
{"train_loss": 0.03892722353339195, "global_step": 242574, "epoch": 2725, "lr": 2.7761630803326554e-06}
{"train_loss": 0.020184170454740524, "global_step": 242575, "epoch": 2725, "lr": 2.775972594647003e-06}
{"train_loss": 0.039729587733745575, "global_step": 242576, "epoch": 2725, "lr": 2.7757821153100384e-06}
{"train_loss": 0.07729358971118927, "global_step": 242577, "epoch": 2725, "lr": 2.775591642321801e-06}
{"train_loss": 0.057053886353969574, "global_step": 242578, "epoch": 2725, "lr": 2.775401175682302e-06}
{"train_loss": 0.029244735836982727, "global_step": 242579, "epoch": 2725, "lr": 2.775210715391585e-06}
{"train_loss": 0.0457150898873806, "global_step": 242580, "epoch": 2725, "lr": 2.7750202614496566e-06}
{"train_loss": 0.043527752161026, "global_step": 242581, "epoch": 2725, "lr": 2.774829813856561e-06}
{"train_loss": 0.02905634604394436, "global_step": 242582, "epoch": 2725, "lr": 2.7746393726123034e-06}
{"train_loss": 0.06159781664609909, "global_step": 242583, "epoch": 2725, "lr": 2.774448937716928e-06}
{"train_loss": 0.07184845209121704, "global_step": 242584, "epoch": 2725, "lr": 2.774258509170441e-06}
{"train_loss": 0.055356401950120926, "global_step": 242585, "epoch": 2725, "lr": 2.7740680869728918e-06}
{"train_loss": 0.03766503930091858, "global_step": 242586, "epoch": 2725, "lr": 2.7738776711242862e-06}
{"train_loss": 0.06937748938798904, "global_step": 242587, "epoch": 2725, "lr": 2.7736872616246636e-06}
{"train_loss": 0.06335846334695816, "global_step": 242588, "epoch": 2725, "lr": 2.773496858474034e-06}
{"train_loss": 0.031127918511629105, "global_step": 242589, "epoch": 2725, "lr": 2.773306461672437e-06}
{"train_loss": 0.06080007553100586, "global_step": 242590, "epoch": 2725, "lr": 2.7731160712199e-06}
{"train_loss": 0.04863041639328003, "global_step": 242591, "epoch": 2725, "lr": 2.7729256871164288e-06}
{"train_loss": 0.05110896751284599, "global_step": 242592, "epoch": 2725, "lr": 2.7727353093620734e-06}
{"train_loss": 0.012654340825974941, "global_step": 242593, "epoch": 2725, "lr": 2.7725449379568393e-06}
{"train_loss": 0.043946005403995514, "global_step": 242594, "epoch": 2725, "lr": 2.7723545729007703e-06}
{"train_loss": 0.021067921072244644, "global_step": 242595, "epoch": 2725, "lr": 2.772164214193873e-06}
{"train_loss": 0.036697179079055786, "global_step": 242596, "epoch": 2725, "lr": 2.771973861836191e-06}
{"train_loss": 0.029916707426309586, "global_step": 242597, "epoch": 2725, "lr": 2.771783515827736e-06}
{"train_loss": 0.04982520639896393, "global_step": 242598, "epoch": 2725, "lr": 2.771593176168541e-06}
{"train_loss": 0.03989158570766449, "global_step": 242599, "epoch": 2725, "lr": 2.771402842858628e-06}
{"train_loss": 0.0696219652891159, "global_step": 242600, "epoch": 2725, "lr": 2.771212515898025e-06}
{"train_loss": 0.02492310106754303, "global_step": 242601, "epoch": 2725, "lr": 2.7710221952867545e-06}
{"train_loss": 0.013511011376976967, "global_step": 242602, "epoch": 2725, "lr": 2.7708318810248436e-06}
{"train_loss": 0.02640474960207939, "global_step": 242603, "epoch": 2725, "lr": 2.770641573112326e-06}
{"train_loss": 0.04840553179383278, "global_step": 242604, "epoch": 2725, "lr": 2.7704512715492127e-06}
{"train_loss": 0.04166678339242935, "global_step": 242605, "epoch": 2725, "lr": 2.7702609763355427e-06}
{"train_loss": 0.04489562287926674, "global_step": 242606, "epoch": 2725, "lr": 2.7700706874713324e-06}
{"train_loss": 0.018767312169075012, "global_step": 242607, "epoch": 2725, "lr": 2.7698804049566153e-06}
{"train_loss": 0.0422525592148304, "global_step": 242608, "epoch": 2725, "lr": 2.7696901287914025e-06}
{"train_loss": 0.061861906200647354, "global_step": 242609, "epoch": 2725, "lr": 2.7694998589757384e-06}
{"train_loss": 0.033547643572092056, "global_step": 242610, "epoch": 2725, "lr": 2.769309595509634e-06}
{"train_loss": 0.023339983075857162, "global_step": 242611, "epoch": 2725, "lr": 2.7691193383931223e-06}
{"train_loss": 0.05796331539750099, "global_step": 242612, "epoch": 2725, "lr": 2.768929087626221e-06}
{"train_loss": 0.04684655285660136, "global_step": 242613, "epoch": 2725, "lr": 2.7687388432089734e-06, "val_loss": 9.480722427368164, "train_action_mse_error": 1.5388245582580566}
{"train_loss": 0.04912024363875389, "global_step": 242614, "epoch": 2726, "lr": 2.76854860514138e-06}
{"train_loss": 0.05069923773407936, "global_step": 242615, "epoch": 2726, "lr": 2.7683583734234908e-06}
{"train_loss": 0.06605149060487747, "global_step": 242616, "epoch": 2726, "lr": 2.7681681480553114e-06}
{"train_loss": 0.09804932773113251, "global_step": 242617, "epoch": 2726, "lr": 2.7679779290368856e-06}
{"train_loss": 0.032426852732896805, "global_step": 242618, "epoch": 2726, "lr": 2.76778771636822e-06}
{"train_loss": 0.025949781760573387, "global_step": 242619, "epoch": 2726, "lr": 2.767597510049358e-06}
{"train_loss": 0.029303448274731636, "global_step": 242620, "epoch": 2726, "lr": 2.7674073100803055e-06}
{"train_loss": 0.06661245971918106, "global_step": 242621, "epoch": 2726, "lr": 2.767217116461107e-06}
{"train_loss": 0.0388447530567646, "global_step": 242622, "epoch": 2726, "lr": 2.7670269291917794e-06}
{"train_loss": 0.036316126585006714, "global_step": 242623, "epoch": 2726, "lr": 2.7668367482723556e-06}
{"train_loss": 0.050603900104761124, "global_step": 242624, "epoch": 2726, "lr": 2.766646573702847e-06}
{"train_loss": 0.008432958275079727, "global_step": 242625, "epoch": 2726, "lr": 2.766456405483292e-06}
{"train_loss": 0.038219697773456573, "global_step": 242626, "epoch": 2726, "lr": 2.7662662436137077e-06}
{"train_loss": 0.08915309607982635, "global_step": 242627, "epoch": 2726, "lr": 2.766076088094127e-06}
{"train_loss": 0.1438380926847458, "global_step": 242628, "epoch": 2726, "lr": 2.765885938924573e-06}
{"train_loss": 0.02074887976050377, "global_step": 242629, "epoch": 2726, "lr": 2.765695796105061e-06}
{"train_loss": 0.015214609913527966, "global_step": 242630, "epoch": 2726, "lr": 2.7655056596356312e-06}
{"train_loss": 0.061936017125844955, "global_step": 242631, "epoch": 2726, "lr": 2.765315529516299e-06}
{"train_loss": 0.019541431218385696, "global_step": 242632, "epoch": 2726, "lr": 2.7651254057471044e-06}
{"train_loss": 0.03294622153043747, "global_step": 242633, "epoch": 2726, "lr": 2.764935288328052e-06}
{"train_loss": 0.05965859070420265, "global_step": 242634, "epoch": 2726, "lr": 2.764745177259187e-06}
{"train_loss": 0.033623721450567245, "global_step": 242635, "epoch": 2726, "lr": 2.7645550725405144e-06}
{"train_loss": 0.01557803712785244, "global_step": 242636, "epoch": 2726, "lr": 2.7643649741720844e-06}
{"train_loss": 0.0323779359459877, "global_step": 242637, "epoch": 2726, "lr": 2.764174882153903e-06}
{"train_loss": 0.06358401477336884, "global_step": 242638, "epoch": 2726, "lr": 2.7639847964859966e-06}
{"train_loss": 0.07652737200260162, "global_step": 242639, "epoch": 2726, "lr": 2.763794717168411e-06}
{"train_loss": 0.05225541070103645, "global_step": 242640, "epoch": 2726, "lr": 2.7636046442011453e-06}
{"train_loss": 0.0695868730545044, "global_step": 242641, "epoch": 2726, "lr": 2.7634145775842447e-06}
{"train_loss": 0.03560035675764084, "global_step": 242642, "epoch": 2726, "lr": 2.763224517317725e-06}
{"train_loss": 0.04142766818404198, "global_step": 242643, "epoch": 2726, "lr": 2.7630344634016146e-06}
{"train_loss": 0.0551578663289547, "global_step": 242644, "epoch": 2726, "lr": 2.762844415835936e-06}
{"train_loss": 0.0670037567615509, "global_step": 242645, "epoch": 2726, "lr": 2.762654374620721e-06}
{"train_loss": 0.05843149870634079, "global_step": 242646, "epoch": 2726, "lr": 2.762464339755988e-06}
{"train_loss": 0.06437813490629196, "global_step": 242647, "epoch": 2726, "lr": 2.762274311241775e-06}
{"train_loss": 0.04797713831067085, "global_step": 242648, "epoch": 2726, "lr": 2.7620842890780874e-06}
{"train_loss": 0.03792745620012283, "global_step": 242649, "epoch": 2726, "lr": 2.7618942732649642e-06}
{"train_loss": 0.023810386657714844, "global_step": 242650, "epoch": 2726, "lr": 2.7617042638024283e-06}
{"train_loss": 0.04236583784222603, "global_step": 242651, "epoch": 2726, "lr": 2.761514260690512e-06}
{"train_loss": 0.045343052595853806, "global_step": 242652, "epoch": 2726, "lr": 2.761324263929227e-06}
{"train_loss": 0.04928603768348694, "global_step": 242653, "epoch": 2726, "lr": 2.761134273518612e-06}
{"train_loss": 0.031161194667220116, "global_step": 242654, "epoch": 2726, "lr": 2.760944289458678e-06}
{"train_loss": 0.09935746341943741, "global_step": 242655, "epoch": 2726, "lr": 2.7607543117494694e-06}
{"train_loss": 0.02805178612470627, "global_step": 242656, "epoch": 2726, "lr": 2.760564340390992e-06}
{"train_loss": 0.05163084715604782, "global_step": 242657, "epoch": 2726, "lr": 2.7603743753832843e-06}
{"train_loss": 0.05318290367722511, "global_step": 242658, "epoch": 2726, "lr": 2.7601844167263744e-06}
{"train_loss": 0.013240154832601547, "global_step": 242659, "epoch": 2726, "lr": 2.7599944644202737e-06}
{"train_loss": 0.06253495067358017, "global_step": 242660, "epoch": 2726, "lr": 2.75980451846502e-06}
{"train_loss": 0.02785959467291832, "global_step": 242661, "epoch": 2726, "lr": 2.759614578860642e-06}
{"train_loss": 0.05120743438601494, "global_step": 242662, "epoch": 2726, "lr": 2.759424645607145e-06}
{"train_loss": 0.05622362717986107, "global_step": 242663, "epoch": 2726, "lr": 2.7592347187045787e-06}
{"train_loss": 0.044756509363651276, "global_step": 242664, "epoch": 2726, "lr": 2.759044798152949e-06}
{"train_loss": 0.022645680233836174, "global_step": 242665, "epoch": 2726, "lr": 2.7588548839522945e-06}
{"train_loss": 0.04516040161252022, "global_step": 242666, "epoch": 2726, "lr": 2.7586649761026315e-06}
{"train_loss": 0.04577692225575447, "global_step": 242667, "epoch": 2726, "lr": 2.7584750746039946e-06}
{"train_loss": 0.014365620911121368, "global_step": 242668, "epoch": 2726, "lr": 2.75828517945641e-06}
{"train_loss": 0.025754490867257118, "global_step": 242669, "epoch": 2726, "lr": 2.75809529065989e-06}
{"train_loss": 0.04520575702190399, "global_step": 242670, "epoch": 2726, "lr": 2.757905408214473e-06}
{"train_loss": 0.033784978091716766, "global_step": 242671, "epoch": 2726, "lr": 2.7577155321201697e-06}
{"train_loss": 0.04066750034689903, "global_step": 242672, "epoch": 2726, "lr": 2.757525662377031e-06}
{"train_loss": 0.0313393697142601, "global_step": 242673, "epoch": 2726, "lr": 2.7573357989850556e-06}
{"train_loss": 0.039320532232522964, "global_step": 242674, "epoch": 2726, "lr": 2.757145941944278e-06}
{"train_loss": 0.010878432542085648, "global_step": 242675, "epoch": 2726, "lr": 2.7569560912547364e-06}
{"train_loss": 0.08256638795137405, "global_step": 242676, "epoch": 2726, "lr": 2.7567662469164424e-06}
{"train_loss": 0.042319413274526596, "global_step": 242677, "epoch": 2726, "lr": 2.756576408929429e-06}
{"train_loss": 0.049137089401483536, "global_step": 242678, "epoch": 2726, "lr": 2.756386577293707e-06}
{"train_loss": 0.03918478265404701, "global_step": 242679, "epoch": 2726, "lr": 2.7561967520093267e-06}
{"train_loss": 0.029123736545443535, "global_step": 242680, "epoch": 2726, "lr": 2.7560069330762885e-06}
{"train_loss": 0.03383677452802658, "global_step": 242681, "epoch": 2726, "lr": 2.755817120494636e-06}
{"train_loss": 0.03531501442193985, "global_step": 242682, "epoch": 2726, "lr": 2.755627314264386e-06}
{"train_loss": 0.05944555625319481, "global_step": 242683, "epoch": 2726, "lr": 2.755437514385567e-06}
{"train_loss": 0.04174579679965973, "global_step": 242684, "epoch": 2726, "lr": 2.7552477208582004e-06}
{"train_loss": 0.019328802824020386, "global_step": 242685, "epoch": 2726, "lr": 2.7550579336823256e-06}
{"train_loss": 0.052840858697891235, "global_step": 242686, "epoch": 2726, "lr": 2.7548681528579425e-06}
{"train_loss": 0.03141411393880844, "global_step": 242687, "epoch": 2726, "lr": 2.754678378385106e-06}
{"train_loss": 0.017162783071398735, "global_step": 242688, "epoch": 2726, "lr": 2.7544886102638114e-06}
{"train_loss": 0.044974517077207565, "global_step": 242689, "epoch": 2726, "lr": 2.7542988484941134e-06}
{"train_loss": 0.06915228813886642, "global_step": 242690, "epoch": 2726, "lr": 2.754109093076018e-06}
{"train_loss": 0.05364707484841347, "global_step": 242691, "epoch": 2726, "lr": 2.753919344009559e-06}
{"train_loss": 0.06674753874540329, "global_step": 242692, "epoch": 2726, "lr": 2.753729601294758e-06}
{"train_loss": 0.030133450403809547, "global_step": 242693, "epoch": 2726, "lr": 2.753539864931637e-06}
{"train_loss": 0.03600769117474556, "global_step": 242694, "epoch": 2726, "lr": 2.753350134920235e-06}
{"train_loss": 0.057405877858400345, "global_step": 242695, "epoch": 2726, "lr": 2.7531604112605637e-06}
{"train_loss": 0.04645008221268654, "global_step": 242696, "epoch": 2726, "lr": 2.752970693952661e-06}
{"train_loss": 0.013400949537754059, "global_step": 242697, "epoch": 2726, "lr": 2.752780982996539e-06}
{"train_loss": 0.08727877587080002, "global_step": 242698, "epoch": 2726, "lr": 2.752591278392236e-06}
{"train_loss": 0.01470435131341219, "global_step": 242699, "epoch": 2726, "lr": 2.752401580139763e-06}
{"train_loss": 0.04590437933802605, "global_step": 242700, "epoch": 2726, "lr": 2.752211888239159e-06}
{"train_loss": 0.0525028370320797, "global_step": 242701, "epoch": 2726, "lr": 2.752022202690452e-06}
{"train_loss": 0.04515139304436325, "global_step": 242702, "epoch": 2726, "lr": 2.751832523493647e-06, "val_loss": 9.523077964782715}
{"train_loss": 0.06031771004199982, "global_step": 242703, "epoch": 2727, "lr": 2.751642850648789e-06}
{"train_loss": 0.025907494127750397, "global_step": 242704, "epoch": 2727, "lr": 2.7514531841558886e-06}
{"train_loss": 0.06117767095565796, "global_step": 242705, "epoch": 2727, "lr": 2.751263524014991e-06}
{"train_loss": 0.03633575141429901, "global_step": 242706, "epoch": 2727, "lr": 2.7510738702261062e-06}
{"train_loss": 0.049588605761528015, "global_step": 242707, "epoch": 2727, "lr": 2.750884222789257e-06}
{"train_loss": 0.009685450233519077, "global_step": 242708, "epoch": 2727, "lr": 2.7506945817044882e-06}
{"train_loss": 0.05995975434780121, "global_step": 242709, "epoch": 2727, "lr": 2.7505049469717993e-06}
{"train_loss": 0.07028000056743622, "global_step": 242710, "epoch": 2727, "lr": 2.75031531859124e-06}
{"train_loss": 0.08260001987218857, "global_step": 242711, "epoch": 2727, "lr": 2.7501256965628163e-06}
{"train_loss": 0.07706248760223389, "global_step": 242712, "epoch": 2727, "lr": 2.7499360808865616e-06}
{"train_loss": 0.03442023694515228, "global_step": 242713, "epoch": 2727, "lr": 2.749746471562509e-06}
{"train_loss": 0.006665442138910294, "global_step": 242714, "epoch": 2727, "lr": 2.7495568685906692e-06}
{"train_loss": 0.06729953736066818, "global_step": 242715, "epoch": 2727, "lr": 2.7493672719710816e-06}
{"train_loss": 0.047896288335323334, "global_step": 242716, "epoch": 2727, "lr": 2.749177681703763e-06}
{"train_loss": 0.07249705493450165, "global_step": 242717, "epoch": 2727, "lr": 2.7489880977887463e-06}
{"train_loss": 0.020585766062140465, "global_step": 242718, "epoch": 2727, "lr": 2.7487985202260426e-06}
{"train_loss": 0.029167797416448593, "global_step": 242719, "epoch": 2727, "lr": 2.748608949015702e-06}
{"train_loss": 0.03665614128112793, "global_step": 242720, "epoch": 2727, "lr": 2.7484193841577186e-06}
{"train_loss": 0.09709621220827103, "global_step": 242721, "epoch": 2727, "lr": 2.7482298256521488e-06}
{"train_loss": 0.0496046282351017, "global_step": 242722, "epoch": 2727, "lr": 2.748040273498992e-06}
{"train_loss": 0.029831862077116966, "global_step": 242723, "epoch": 2727, "lr": 2.7478507276982923e-06}
{"train_loss": 0.03084077686071396, "global_step": 242724, "epoch": 2727, "lr": 2.7476611882500613e-06}
{"train_loss": 0.05939871445298195, "global_step": 242725, "epoch": 2727, "lr": 2.7474716551543435e-06}
{"train_loss": 0.015530052594840527, "global_step": 242726, "epoch": 2727, "lr": 2.7472821284111383e-06}
{"train_loss": 0.02896483801305294, "global_step": 242727, "epoch": 2727, "lr": 2.7470926080204963e-06}
{"train_loss": 0.0368509404361248, "global_step": 242728, "epoch": 2727, "lr": 2.7469030939824227e-06}
{"train_loss": 0.06422717124223709, "global_step": 242729, "epoch": 2727, "lr": 2.7467135862969505e-06}
{"train_loss": 0.044224273413419724, "global_step": 242730, "epoch": 2727, "lr": 2.7465240849641196e-06}
{"train_loss": 0.024541305378079414, "global_step": 242731, "epoch": 2727, "lr": 2.7463345899839286e-06}
{"train_loss": 0.028777634724974632, "global_step": 242732, "epoch": 2727, "lr": 2.746145101356429e-06}
{"train_loss": 0.04828691482543945, "global_step": 242733, "epoch": 2727, "lr": 2.745955619081625e-06}
{"train_loss": 0.026061885058879852, "global_step": 242734, "epoch": 2727, "lr": 2.7457661431595617e-06}
{"train_loss": 0.017657989636063576, "global_step": 242735, "epoch": 2727, "lr": 2.7455766735902443e-06}
{"train_loss": 0.05373425409197807, "global_step": 242736, "epoch": 2727, "lr": 2.745387210373712e-06}
{"train_loss": 0.04177263751626015, "global_step": 242737, "epoch": 2727, "lr": 2.745197753509987e-06}
{"train_loss": 0.04391941428184509, "global_step": 242738, "epoch": 2727, "lr": 2.7450083029990968e-06}
{"train_loss": 0.04492449387907982, "global_step": 242739, "epoch": 2727, "lr": 2.744818858841064e-06}
{"train_loss": 0.03671487420797348, "global_step": 242740, "epoch": 2727, "lr": 2.7446294210359047e-06}
{"train_loss": 0.029657116159796715, "global_step": 242741, "epoch": 2727, "lr": 2.7444399895836638e-06}
{"train_loss": 0.03494828939437866, "global_step": 242742, "epoch": 2727, "lr": 2.744250564484352e-06}
{"train_loss": 0.05754489451646805, "global_step": 242743, "epoch": 2727, "lr": 2.744061145738003e-06}
{"train_loss": 0.03245116397738457, "global_step": 242744, "epoch": 2727, "lr": 2.743871733344644e-06}
{"train_loss": 0.03764093294739723, "global_step": 242745, "epoch": 2727, "lr": 2.7436823273042865e-06}
{"train_loss": 0.022108426317572594, "global_step": 242746, "epoch": 2727, "lr": 2.74349292761697e-06}
{"train_loss": 0.03982425481081009, "global_step": 242747, "epoch": 2727, "lr": 2.7433035342827095e-06}
{"train_loss": 0.023173917084932327, "global_step": 242748, "epoch": 2727, "lr": 2.7431141473015342e-06}
{"train_loss": 0.032353319227695465, "global_step": 242749, "epoch": 2727, "lr": 2.742924766673477e-06}
{"train_loss": 0.048083700239658356, "global_step": 242750, "epoch": 2727, "lr": 2.7427353923985545e-06}
{"train_loss": 0.06552907824516296, "global_step": 242751, "epoch": 2727, "lr": 2.7425460244767996e-06}
{"train_loss": 0.0372246615588665, "global_step": 242752, "epoch": 2727, "lr": 2.7423566629082297e-06}
{"train_loss": 0.06591887772083282, "global_step": 242753, "epoch": 2727, "lr": 2.742167307692878e-06}
{"train_loss": 0.008483047597110271, "global_step": 242754, "epoch": 2727, "lr": 2.7419779588307602e-06}
{"train_loss": 0.020777486264705658, "global_step": 242755, "epoch": 2727, "lr": 2.741788616321911e-06}
{"train_loss": 0.10392871499061584, "global_step": 242756, "epoch": 2727, "lr": 2.7415992801663514e-06}
{"train_loss": 0.042265187948942184, "global_step": 242757, "epoch": 2727, "lr": 2.74140995036411e-06}
{"train_loss": 0.06352154910564423, "global_step": 242758, "epoch": 2727, "lr": 2.741220626915203e-06}
{"train_loss": 0.04207725077867508, "global_step": 242759, "epoch": 2727, "lr": 2.741031309819675e-06}
{"train_loss": 0.05482164025306702, "global_step": 242760, "epoch": 2727, "lr": 2.7408419990775256e-06}
{"train_loss": 0.04188871011137962, "global_step": 242761, "epoch": 2727, "lr": 2.7406526946888055e-06}
{"train_loss": 0.047957103699445724, "global_step": 242762, "epoch": 2727, "lr": 2.74046339665352e-06}
{"train_loss": 0.04224984347820282, "global_step": 242763, "epoch": 2727, "lr": 2.7402741049717072e-06}
{"train_loss": 0.03195947781205177, "global_step": 242764, "epoch": 2727, "lr": 2.740084819643385e-06}
{"train_loss": 0.0239584781229496, "global_step": 242765, "epoch": 2727, "lr": 2.7398955406685857e-06}
{"train_loss": 0.020284803584218025, "global_step": 242766, "epoch": 2727, "lr": 2.739706268047332e-06}
{"train_loss": 0.04320881515741348, "global_step": 242767, "epoch": 2727, "lr": 2.7395170017796455e-06}
{"train_loss": 0.047147028148174286, "global_step": 242768, "epoch": 2727, "lr": 2.7393277418655604e-06}
{"train_loss": 0.02755594067275524, "global_step": 242769, "epoch": 2727, "lr": 2.739138488305093e-06}
{"train_loss": 0.05936211720108986, "global_step": 242770, "epoch": 2727, "lr": 2.7389492410982765e-06}
{"train_loss": 0.05298848822712898, "global_step": 242771, "epoch": 2727, "lr": 2.738760000245122e-06}
{"train_loss": 0.07706805318593979, "global_step": 242772, "epoch": 2727, "lr": 2.7385707657456796e-06}
{"train_loss": 0.04746226593852043, "global_step": 242773, "epoch": 2727, "lr": 2.738381537599949e-06}
{"train_loss": 0.06895426660776138, "global_step": 242774, "epoch": 2727, "lr": 2.7381923158079746e-06}
{"train_loss": 0.04741987958550453, "global_step": 242775, "epoch": 2727, "lr": 2.7380031003697626e-06}
{"train_loss": 0.037017595022916794, "global_step": 242776, "epoch": 2727, "lr": 2.7378138912853624e-06}
{"train_loss": 0.036704834550619125, "global_step": 242777, "epoch": 2727, "lr": 2.7376246885547795e-06}
{"train_loss": 0.029912885278463364, "global_step": 242778, "epoch": 2727, "lr": 2.7374354921780534e-06}
{"train_loss": 0.05854104086756706, "global_step": 242779, "epoch": 2727, "lr": 2.737246302155205e-06}
{"train_loss": 0.02365824580192566, "global_step": 242780, "epoch": 2727, "lr": 2.737057118486247e-06}
{"train_loss": 0.02451101690530777, "global_step": 242781, "epoch": 2727, "lr": 2.736867941171228e-06}
{"train_loss": 0.06377917528152466, "global_step": 242782, "epoch": 2727, "lr": 2.736678770210149e-06}
{"train_loss": 0.049216821789741516, "global_step": 242783, "epoch": 2727, "lr": 2.7364896056030543e-06}
{"train_loss": 0.03763006255030632, "global_step": 242784, "epoch": 2727, "lr": 2.7363004473499597e-06}
{"train_loss": 0.03133855015039444, "global_step": 242785, "epoch": 2727, "lr": 2.736111295450894e-06}
{"train_loss": 0.05743744596838951, "global_step": 242786, "epoch": 2727, "lr": 2.735922149905884e-06}
{"train_loss": 0.01526353508234024, "global_step": 242787, "epoch": 2727, "lr": 2.7357330107149526e-06}
{"train_loss": 0.03672274947166443, "global_step": 242788, "epoch": 2727, "lr": 2.7355438778781217e-06}
{"train_loss": 0.03001878596842289, "global_step": 242789, "epoch": 2727, "lr": 2.7353547513954246e-06}
{"train_loss": 0.05987917631864548, "global_step": 242790, "epoch": 2727, "lr": 2.735165631266884e-06}
{"train_loss": 0.04325746524074439, "global_step": 242791, "epoch": 2727, "lr": 2.7349765174925267e-06, "val_loss": 9.45761775970459}
{"train_loss": 0.01801641657948494, "global_step": 242792, "epoch": 2728, "lr": 2.734787410072365e-06}
{"train_loss": 0.011106876656413078, "global_step": 242793, "epoch": 2728, "lr": 2.7345983090064473e-06}
{"train_loss": 0.00870700366795063, "global_step": 242794, "epoch": 2728, "lr": 2.7344092142947753e-06}
{"train_loss": 0.10726293921470642, "global_step": 242795, "epoch": 2728, "lr": 2.734220125937398e-06}
{"train_loss": 0.016235098242759705, "global_step": 242796, "epoch": 2728, "lr": 2.7340310439343155e-06}
{"train_loss": 0.0639054924249649, "global_step": 242797, "epoch": 2728, "lr": 2.7338419682855778e-06}
{"train_loss": 0.05020516365766525, "global_step": 242798, "epoch": 2728, "lr": 2.73365289899119e-06}
{"train_loss": 0.0537135973572731, "global_step": 242799, "epoch": 2728, "lr": 2.733463836051198e-06}
{"train_loss": 0.044555772095918655, "global_step": 242800, "epoch": 2728, "lr": 2.7332747794656054e-06}
{"train_loss": 0.04467940703034401, "global_step": 242801, "epoch": 2728, "lr": 2.733085729234447e-06}
{"train_loss": 0.04931126907467842, "global_step": 242802, "epoch": 2728, "lr": 2.7328966853577552e-06}
{"train_loss": 0.060973070561885834, "global_step": 242803, "epoch": 2728, "lr": 2.732707647835547e-06}
{"train_loss": 0.04132768511772156, "global_step": 242804, "epoch": 2728, "lr": 2.732518616667856e-06}
{"train_loss": 0.08872392028570175, "global_step": 242805, "epoch": 2728, "lr": 2.732329591854693e-06}
{"train_loss": 0.03905763104557991, "global_step": 242806, "epoch": 2728, "lr": 2.732140573396097e-06}
{"train_loss": 0.020100269466638565, "global_step": 242807, "epoch": 2728, "lr": 2.7319515612920843e-06}
{"train_loss": 0.08437198400497437, "global_step": 242808, "epoch": 2728, "lr": 2.7317625555426886e-06}
{"train_loss": 0.054593246430158615, "global_step": 242809, "epoch": 2728, "lr": 2.731573556147932e-06}
{"train_loss": 0.047187887132167816, "global_step": 242810, "epoch": 2728, "lr": 2.731384563107836e-06}
{"train_loss": 0.06481494754552841, "global_step": 242811, "epoch": 2728, "lr": 2.7311955764224296e-06}
{"train_loss": 0.03658382594585419, "global_step": 242812, "epoch": 2728, "lr": 2.731006596091745e-06}
{"train_loss": 0.029345033690333366, "global_step": 242813, "epoch": 2728, "lr": 2.730817622115789e-06}
{"train_loss": 0.037435322999954224, "global_step": 242814, "epoch": 2728, "lr": 2.7306286544946047e-06}
{"train_loss": 0.03365830332040787, "global_step": 242815, "epoch": 2728, "lr": 2.7304396932282093e-06}
{"train_loss": 0.056723494082689285, "global_step": 242816, "epoch": 2728, "lr": 2.7302507383166363e-06}
{"train_loss": 0.09850651770830154, "global_step": 242817, "epoch": 2728, "lr": 2.730061789759902e-06}
{"train_loss": 0.022601505741477013, "global_step": 242818, "epoch": 2728, "lr": 2.7298728475580293e-06}
{"train_loss": 0.03177987411618233, "global_step": 242819, "epoch": 2728, "lr": 2.7296839117110563e-06}
{"train_loss": 0.06344449520111084, "global_step": 242820, "epoch": 2728, "lr": 2.729494982218994e-06}
{"train_loss": 0.04105787351727486, "global_step": 242821, "epoch": 2728, "lr": 2.729306059081882e-06}
{"train_loss": 0.10018512606620789, "global_step": 242822, "epoch": 2728, "lr": 2.7291171422997306e-06}
{"train_loss": 0.05193977802991867, "global_step": 242823, "epoch": 2728, "lr": 2.7289282318725793e-06}
{"train_loss": 0.08436287939548492, "global_step": 242824, "epoch": 2728, "lr": 2.7287393278004447e-06}
{"train_loss": 0.03673315793275833, "global_step": 242825, "epoch": 2728, "lr": 2.72855043008336e-06}
{"train_loss": 0.1163816973567009, "global_step": 242826, "epoch": 2728, "lr": 2.728361538721336e-06}
{"train_loss": 0.0924663096666336, "global_step": 242827, "epoch": 2728, "lr": 2.728172653714417e-06}
{"train_loss": 0.06592200696468353, "global_step": 242828, "epoch": 2728, "lr": 2.727983775062609e-06}
{"train_loss": 0.04011312872171402, "global_step": 242829, "epoch": 2728, "lr": 2.727794902765962e-06}
{"train_loss": 0.029179519042372704, "global_step": 242830, "epoch": 2728, "lr": 2.7276060368244703e-06}
{"train_loss": 0.05390606448054314, "global_step": 242831, "epoch": 2728, "lr": 2.7274171772381898e-06}
{"train_loss": 0.027445102110505104, "global_step": 242832, "epoch": 2728, "lr": 2.7272283240071195e-06}
{"train_loss": 0.06287837028503418, "global_step": 242833, "epoch": 2728, "lr": 2.7270394771313103e-06}
{"train_loss": 0.03563150763511658, "global_step": 242834, "epoch": 2728, "lr": 2.7268506366107618e-06}
{"train_loss": 0.07093162089586258, "global_step": 242835, "epoch": 2728, "lr": 2.726661802445524e-06}
{"train_loss": 0.042430851608514786, "global_step": 242836, "epoch": 2728, "lr": 2.726472974635602e-06}
{"train_loss": 0.06577602028846741, "global_step": 242837, "epoch": 2728, "lr": 2.7262841531810356e-06}
{"train_loss": 0.07630957663059235, "global_step": 242838, "epoch": 2728, "lr": 2.7260953380818355e-06}
{"train_loss": 0.07702870666980743, "global_step": 242839, "epoch": 2728, "lr": 2.7259065293380346e-06}
{"train_loss": 0.0472167506814003, "global_step": 242840, "epoch": 2728, "lr": 2.7257177269496727e-06}
{"train_loss": 0.07250386476516724, "global_step": 242841, "epoch": 2728, "lr": 2.7255289309167485e-06}
{"train_loss": 0.07730389386415482, "global_step": 242842, "epoch": 2728, "lr": 2.7253401412393133e-06}
{"train_loss": 0.05671613663434982, "global_step": 242843, "epoch": 2728, "lr": 2.725151357917366e-06}
{"train_loss": 0.054391033947467804, "global_step": 242844, "epoch": 2728, "lr": 2.724962580950957e-06}
{"train_loss": 0.04087727889418602, "global_step": 242845, "epoch": 2728, "lr": 2.724773810340092e-06}
{"train_loss": 0.041773971170186996, "global_step": 242846, "epoch": 2728, "lr": 2.7245850460848154e-06}
{"train_loss": 0.09029532223939896, "global_step": 242847, "epoch": 2728, "lr": 2.7243962881851324e-06}
{"train_loss": 0.03899112343788147, "global_step": 242848, "epoch": 2728, "lr": 2.724207536641088e-06}
{"train_loss": 0.05478505417704582, "global_step": 242849, "epoch": 2728, "lr": 2.724018791452687e-06}
{"train_loss": 0.04556789994239807, "global_step": 242850, "epoch": 2728, "lr": 2.7238300526199746e-06}
{"train_loss": 0.04103642702102661, "global_step": 242851, "epoch": 2728, "lr": 2.723641320142961e-06}
{"train_loss": 0.037357017397880554, "global_step": 242852, "epoch": 2728, "lr": 2.7234525940216803e-06}
{"train_loss": 0.056026384234428406, "global_step": 242853, "epoch": 2728, "lr": 2.723263874256149e-06}
{"train_loss": 0.04881134629249573, "global_step": 242854, "epoch": 2728, "lr": 2.723075160846411e-06}
{"train_loss": 0.06957781314849854, "global_step": 242855, "epoch": 2728, "lr": 2.7228864537924727e-06}
{"train_loss": 0.05426802113652229, "global_step": 242856, "epoch": 2728, "lr": 2.7226977530943667e-06}
{"train_loss": 0.027306612581014633, "global_step": 242857, "epoch": 2728, "lr": 2.722509058752115e-06}
{"train_loss": 0.07269416749477386, "global_step": 242858, "epoch": 2728, "lr": 2.7223203707657463e-06}
{"train_loss": 0.06955767422914505, "global_step": 242859, "epoch": 2728, "lr": 2.7221316891352934e-06}
{"train_loss": 0.04721000790596008, "global_step": 242860, "epoch": 2728, "lr": 2.7219430138607614e-06}
{"train_loss": 0.040738869458436966, "global_step": 242861, "epoch": 2728, "lr": 2.7217543449421957e-06}
{"train_loss": 0.04472668468952179, "global_step": 242862, "epoch": 2728, "lr": 2.7215656823796064e-06}
{"train_loss": 0.040572769939899445, "global_step": 242863, "epoch": 2728, "lr": 2.721377026173039e-06}
{"train_loss": 0.07088324427604675, "global_step": 242864, "epoch": 2728, "lr": 2.721188376322492e-06}
{"train_loss": 0.02287554368376732, "global_step": 242865, "epoch": 2728, "lr": 2.720999732828017e-06}
{"train_loss": 0.046523526310920715, "global_step": 242866, "epoch": 2728, "lr": 2.720811095689618e-06}
{"train_loss": 0.03621114790439606, "global_step": 242867, "epoch": 2728, "lr": 2.7206224649073354e-06}
{"train_loss": 0.031006505712866783, "global_step": 242868, "epoch": 2728, "lr": 2.7204338404811848e-06}
{"train_loss": 0.050543732941150665, "global_step": 242869, "epoch": 2728, "lr": 2.7202452224111995e-06}
{"train_loss": 0.06334232538938522, "global_step": 242870, "epoch": 2728, "lr": 2.720056610697397e-06}
{"train_loss": 0.014587411656975746, "global_step": 242871, "epoch": 2728, "lr": 2.719868005339815e-06}
{"train_loss": 0.03142040967941284, "global_step": 242872, "epoch": 2728, "lr": 2.71967940633846e-06}
{"train_loss": 0.04393141344189644, "global_step": 242873, "epoch": 2728, "lr": 2.719490813693376e-06}
{"train_loss": 0.05979831516742706, "global_step": 242874, "epoch": 2728, "lr": 2.7193022274045744e-06}
{"train_loss": 0.04055646061897278, "global_step": 242875, "epoch": 2728, "lr": 2.7191136474720824e-06}
{"train_loss": 0.04984750971198082, "global_step": 242876, "epoch": 2728, "lr": 2.7189250738959395e-06}
{"train_loss": 0.025059383362531662, "global_step": 242877, "epoch": 2728, "lr": 2.718736506676156e-06}
{"train_loss": 0.062277212738990784, "global_step": 242878, "epoch": 2728, "lr": 2.718547945812766e-06}
{"train_loss": 0.031566161662340164, "global_step": 242879, "epoch": 2728, "lr": 2.7183593913057804e-06}
{"train_loss": 0.051130586293306246, "global_step": 242880, "epoch": 2728, "lr": 2.7181708431552487e-06, "val_loss": 9.492279052734375}
{"train_loss": 0.05601687729358673, "global_step": 242881, "epoch": 2729, "lr": 2.7179823013611716e-06}
{"train_loss": 0.03771279379725456, "global_step": 242882, "epoch": 2729, "lr": 2.717793765923593e-06}
{"train_loss": 0.06177753955125809, "global_step": 242883, "epoch": 2729, "lr": 2.717605236842524e-06}
{"train_loss": 0.03723793104290962, "global_step": 242884, "epoch": 2729, "lr": 2.717416714118004e-06}
{"train_loss": 0.036067526787519455, "global_step": 242885, "epoch": 2729, "lr": 2.717228197750049e-06}
{"train_loss": 0.0801350399851799, "global_step": 242886, "epoch": 2729, "lr": 2.7170396877386873e-06}
{"train_loss": 0.06457245349884033, "global_step": 242887, "epoch": 2729, "lr": 2.716851184083935e-06}
{"train_loss": 0.004206642042845488, "global_step": 242888, "epoch": 2729, "lr": 2.716662686785837e-06}
{"train_loss": 0.0431433767080307, "global_step": 242889, "epoch": 2729, "lr": 2.7164741958443985e-06}
{"train_loss": 0.06565775722265244, "global_step": 242890, "epoch": 2729, "lr": 2.716285711259664e-06}
{"train_loss": 0.04569535329937935, "global_step": 242891, "epoch": 2729, "lr": 2.7160972330316334e-06}
{"train_loss": 0.019845277070999146, "global_step": 242892, "epoch": 2729, "lr": 2.715908761160363e-06}
{"train_loss": 0.04077098146080971, "global_step": 242893, "epoch": 2729, "lr": 2.7157202956458517e-06}
{"train_loss": 0.05720255896449089, "global_step": 242894, "epoch": 2729, "lr": 2.715531836488144e-06}
{"train_loss": 0.0641261488199234, "global_step": 242895, "epoch": 2729, "lr": 2.715343383687252e-06}
{"train_loss": 0.07838035374879837, "global_step": 242896, "epoch": 2729, "lr": 2.715154937243203e-06}
{"train_loss": 0.02475123293697834, "global_step": 242897, "epoch": 2729, "lr": 2.7149664971560296e-06}
{"train_loss": 0.057420697063207626, "global_step": 242898, "epoch": 2729, "lr": 2.7147780634257492e-06}
{"train_loss": 0.04042918607592583, "global_step": 242899, "epoch": 2729, "lr": 2.714589636052395e-06}
{"train_loss": 0.03217293694615364, "global_step": 242900, "epoch": 2729, "lr": 2.7144012150359833e-06}
{"train_loss": 0.04867219924926758, "global_step": 242901, "epoch": 2729, "lr": 2.7142128003765477e-06}
{"train_loss": 0.05396087467670441, "global_step": 242902, "epoch": 2729, "lr": 2.714024392074105e-06}
{"train_loss": 0.019147608429193497, "global_step": 242903, "epoch": 2729, "lr": 2.7138359901286936e-06}
{"train_loss": 0.07219012826681137, "global_step": 242904, "epoch": 2729, "lr": 2.7136475945403196e-06}
{"train_loss": 0.039138853549957275, "global_step": 242905, "epoch": 2729, "lr": 2.713459205309027e-06}
{"train_loss": 0.05511121079325676, "global_step": 242906, "epoch": 2729, "lr": 2.7132708224348326e-06}
{"train_loss": 0.053717371076345444, "global_step": 242907, "epoch": 2729, "lr": 2.713082445917764e-06}
{"train_loss": 0.022685693576931953, "global_step": 242908, "epoch": 2729, "lr": 2.712894075757838e-06}
{"train_loss": 0.054289862513542175, "global_step": 242909, "epoch": 2729, "lr": 2.7127057119550935e-06}
{"train_loss": 0.053086623549461365, "global_step": 242910, "epoch": 2729, "lr": 2.7125173545095474e-06}
{"train_loss": 0.038222216069698334, "global_step": 242911, "epoch": 2729, "lr": 2.712329003421221e-06}
{"train_loss": 0.053078699856996536, "global_step": 242912, "epoch": 2729, "lr": 2.712140658690154e-06}
{"train_loss": 0.09017637372016907, "global_step": 242913, "epoch": 2729, "lr": 2.7119523203163578e-06}
{"train_loss": 0.08484356850385666, "global_step": 242914, "epoch": 2729, "lr": 2.7117639882998702e-06}
{"train_loss": 0.05332522839307785, "global_step": 242915, "epoch": 2729, "lr": 2.711575662640703e-06}
{"train_loss": 0.08357582241296768, "global_step": 242916, "epoch": 2729, "lr": 2.711387343338895e-06}
{"train_loss": 0.034606073051691055, "global_step": 242917, "epoch": 2729, "lr": 2.711199030394457e-06}
{"train_loss": 0.042604971677064896, "global_step": 242918, "epoch": 2729, "lr": 2.711010723807428e-06}
{"train_loss": 0.06639524549245834, "global_step": 242919, "epoch": 2729, "lr": 2.7108224235778244e-06}
{"train_loss": 0.05865418165922165, "global_step": 242920, "epoch": 2729, "lr": 2.710634129705675e-06}
{"train_loss": 0.06610944122076035, "global_step": 242921, "epoch": 2729, "lr": 2.710445842191001e-06}
{"train_loss": 0.03155050799250603, "global_step": 242922, "epoch": 2729, "lr": 2.710257561033841e-06}
{"train_loss": 0.043106961995363235, "global_step": 242923, "epoch": 2729, "lr": 2.710069286234196e-06}
{"train_loss": 0.039556946605443954, "global_step": 242924, "epoch": 2729, "lr": 2.709881017792121e-06}
{"train_loss": 0.053605299443006516, "global_step": 242925, "epoch": 2729, "lr": 2.7096927557076102e-06}
{"train_loss": 0.07843545824289322, "global_step": 242926, "epoch": 2729, "lr": 2.7095044999807197e-06}
{"train_loss": 0.05666261538863182, "global_step": 242927, "epoch": 2729, "lr": 2.709316250611449e-06}
{"train_loss": 0.02960962802171707, "global_step": 242928, "epoch": 2729, "lr": 2.7091280075998426e-06}
{"train_loss": 0.01867237128317356, "global_step": 242929, "epoch": 2729, "lr": 2.708939770945912e-06}
{"train_loss": 0.036496467888355255, "global_step": 242930, "epoch": 2729, "lr": 2.70875154064969e-06}
{"train_loss": 0.03872694820165634, "global_step": 242931, "epoch": 2729, "lr": 2.708563316711199e-06}
{"train_loss": 0.010027844458818436, "global_step": 242932, "epoch": 2729, "lr": 2.708375099130467e-06}
{"train_loss": 0.06576874107122421, "global_step": 242933, "epoch": 2729, "lr": 2.708186887907521e-06}
{"train_loss": 0.043835125863552094, "global_step": 242934, "epoch": 2729, "lr": 2.707998683042379e-06}
{"train_loss": 0.05329417809844017, "global_step": 242935, "epoch": 2729, "lr": 2.707810484535073e-06}
{"train_loss": 0.05909208580851555, "global_step": 242936, "epoch": 2729, "lr": 2.7076222923856144e-06}
{"train_loss": 0.075310118496418, "global_step": 242937, "epoch": 2729, "lr": 2.7074341065940534e-06}
{"train_loss": 0.0355256088078022, "global_step": 242938, "epoch": 2729, "lr": 2.70724592716039e-06}
{"train_loss": 0.044996265321969986, "global_step": 242939, "epoch": 2729, "lr": 2.7070577540846687e-06}
{"train_loss": 0.03826279193162918, "global_step": 242940, "epoch": 2729, "lr": 2.706869587366906e-06}
{"train_loss": 0.07233890146017075, "global_step": 242941, "epoch": 2729, "lr": 2.7066814270071294e-06}
{"train_loss": 0.06756225973367691, "global_step": 242942, "epoch": 2729, "lr": 2.706493273005356e-06}
{"train_loss": 0.0977967381477356, "global_step": 242943, "epoch": 2729, "lr": 2.7063051253616243e-06}
{"train_loss": 0.025783605873584747, "global_step": 242944, "epoch": 2729, "lr": 2.7061169840759513e-06}
{"train_loss": 0.054307326674461365, "global_step": 242945, "epoch": 2729, "lr": 2.7059288491483703e-06}
{"train_loss": 0.09768989682197571, "global_step": 242946, "epoch": 2729, "lr": 2.705740720578892e-06}
{"train_loss": 0.049160391092300415, "global_step": 242947, "epoch": 2729, "lr": 2.7055525983675502e-06}
{"train_loss": 0.08052631467580795, "global_step": 242948, "epoch": 2729, "lr": 2.7053644825143778e-06}
{"train_loss": 0.045861922204494476, "global_step": 242949, "epoch": 2729, "lr": 2.7051763730193857e-06}
{"train_loss": 0.07916463911533356, "global_step": 242950, "epoch": 2729, "lr": 2.704988269882608e-06}
{"train_loss": 0.06924157589673996, "global_step": 242951, "epoch": 2729, "lr": 2.704800173104066e-06}
{"train_loss": 0.05763731524348259, "global_step": 242952, "epoch": 2729, "lr": 2.7046120826837994e-06}
{"train_loss": 0.05335918813943863, "global_step": 242953, "epoch": 2729, "lr": 2.7044239986218076e-06}
{"train_loss": 0.033135849982500076, "global_step": 242954, "epoch": 2729, "lr": 2.7042359209181356e-06}
{"train_loss": 0.0321572907269001, "global_step": 242955, "epoch": 2729, "lr": 2.7040478495727994e-06}
{"train_loss": 0.06014779582619667, "global_step": 242956, "epoch": 2729, "lr": 2.703859784585833e-06}
{"train_loss": 0.03535895049571991, "global_step": 242957, "epoch": 2729, "lr": 2.703671725957252e-06}
{"train_loss": 0.04690604284405708, "global_step": 242958, "epoch": 2729, "lr": 2.7034836736870907e-06}
{"train_loss": 0.040434204041957855, "global_step": 242959, "epoch": 2729, "lr": 2.7032956277753595e-06}
{"train_loss": 0.031028708443045616, "global_step": 242960, "epoch": 2729, "lr": 2.7031075882221035e-06}
{"train_loss": 0.023859577253460884, "global_step": 242961, "epoch": 2729, "lr": 2.702919555027333e-06}
{"train_loss": 0.029773609712719917, "global_step": 242962, "epoch": 2729, "lr": 2.7027315281910826e-06}
{"train_loss": 0.04778565466403961, "global_step": 242963, "epoch": 2729, "lr": 2.7025435077133677e-06}
{"train_loss": 0.050096869468688965, "global_step": 242964, "epoch": 2729, "lr": 2.7023554935942275e-06}
{"train_loss": 0.037103306502103806, "global_step": 242965, "epoch": 2729, "lr": 2.7021674858336674e-06}
{"train_loss": 0.015297295525670052, "global_step": 242966, "epoch": 2729, "lr": 2.701979484431738e-06}
{"train_loss": 0.02112014964222908, "global_step": 242967, "epoch": 2729, "lr": 2.7017914893884388e-06}
{"train_loss": 0.056464795023202896, "global_step": 242968, "epoch": 2729, "lr": 2.7016035007038197e-06}
{"train_loss": 0.04940291198514653, "global_step": 242969, "epoch": 2729, "lr": 2.701415518377881e-06, "val_loss": 9.421290397644043}
{"train_loss": 0.06480474770069122, "global_step": 242970, "epoch": 2730, "lr": 2.7012275424106727e-06}
{"train_loss": 0.09798357635736465, "global_step": 242971, "epoch": 2730, "lr": 2.7010395728022e-06}
{"train_loss": 0.042295534163713455, "global_step": 242972, "epoch": 2730, "lr": 2.7008516095524963e-06}
{"train_loss": 0.0060513317584991455, "global_step": 242973, "epoch": 2730, "lr": 2.700663652661589e-06}
{"train_loss": 0.02682841196656227, "global_step": 242974, "epoch": 2730, "lr": 2.7004757021294956e-06}
{"train_loss": 0.04919900372624397, "global_step": 242975, "epoch": 2730, "lr": 2.7002877579562546e-06}
{"train_loss": 0.06587307900190353, "global_step": 242976, "epoch": 2730, "lr": 2.700099820141877e-06}
{"train_loss": 0.017528817057609558, "global_step": 242977, "epoch": 2730, "lr": 2.6999118886864017e-06}
{"train_loss": 0.03594973683357239, "global_step": 242978, "epoch": 2730, "lr": 2.69972396358984e-06}
{"train_loss": 0.026808787137269974, "global_step": 242979, "epoch": 2730, "lr": 2.69953604485223e-06}
{"train_loss": 0.03622626140713692, "global_step": 242980, "epoch": 2730, "lr": 2.6993481324735836e-06}
{"train_loss": 0.09462311863899231, "global_step": 242981, "epoch": 2730, "lr": 2.6991602264539396e-06}
{"train_loss": 0.07712797075510025, "global_step": 242982, "epoch": 2730, "lr": 2.6989723267933087e-06}
{"train_loss": 0.0568402074277401, "global_step": 242983, "epoch": 2730, "lr": 2.6987844334917243e-06}
{"train_loss": 0.06608564406633377, "global_step": 242984, "epoch": 2730, "lr": 2.6985965465492202e-06}
{"train_loss": 0.06543987989425659, "global_step": 242985, "epoch": 2730, "lr": 2.698408665965807e-06}
{"train_loss": 0.03482060506939888, "global_step": 242986, "epoch": 2730, "lr": 2.6982207917415235e-06}
{"train_loss": 0.06494131684303284, "global_step": 242987, "epoch": 2730, "lr": 2.6980329238763813e-06}
{"train_loss": 0.03024217113852501, "global_step": 242988, "epoch": 2730, "lr": 2.697845062370419e-06}
{"train_loss": 0.07280874252319336, "global_step": 242989, "epoch": 2730, "lr": 2.697657207223647e-06}
{"train_loss": 0.0316590815782547, "global_step": 242990, "epoch": 2730, "lr": 2.6974693584361055e-06}
{"train_loss": 0.04249761626124382, "global_step": 242991, "epoch": 2730, "lr": 2.697281516007805e-06}
{"train_loss": 0.04424573853611946, "global_step": 242992, "epoch": 2730, "lr": 2.6970936799387836e-06}
{"train_loss": 0.05533678084611893, "global_step": 242993, "epoch": 2730, "lr": 2.696905850229059e-06}
{"train_loss": 0.07692509144544601, "global_step": 242994, "epoch": 2730, "lr": 2.696718026878664e-06}
{"train_loss": 0.051131051033735275, "global_step": 242995, "epoch": 2730, "lr": 2.6965302098876098e-06}
{"train_loss": 0.04205462336540222, "global_step": 242996, "epoch": 2730, "lr": 2.6963423992559413e-06}
{"train_loss": 0.024389512836933136, "global_step": 242997, "epoch": 2730, "lr": 2.696154594983663e-06}
{"train_loss": 0.03228649124503136, "global_step": 242998, "epoch": 2730, "lr": 2.6959667970708148e-06}
{"train_loss": 0.03480546176433563, "global_step": 242999, "epoch": 2730, "lr": 2.695779005517418e-06}
{"train_loss": 0.060415588319301605, "global_step": 243000, "epoch": 2730, "lr": 2.695591220323496e-06}
{"train_loss": 0.07735506445169449, "global_step": 243001, "epoch": 2730, "lr": 2.695403441489075e-06}
{"train_loss": 0.08410738408565521, "global_step": 243002, "epoch": 2730, "lr": 2.6952156690141785e-06}
{"train_loss": 0.017420383170247078, "global_step": 243003, "epoch": 2730, "lr": 2.695027902898839e-06}
{"train_loss": 0.03927776217460632, "global_step": 243004, "epoch": 2730, "lr": 2.6948401431430737e-06}
{"train_loss": 0.04752467945218086, "global_step": 243005, "epoch": 2730, "lr": 2.69465238974691e-06}
{"train_loss": 0.023335693404078484, "global_step": 243006, "epoch": 2730, "lr": 2.694464642710376e-06}
{"train_loss": 0.027151847258210182, "global_step": 243007, "epoch": 2730, "lr": 2.694276902033488e-06}
{"train_loss": 0.01820835843682289, "global_step": 243008, "epoch": 2730, "lr": 2.694089167716285e-06}
{"train_loss": 0.06669549643993378, "global_step": 243009, "epoch": 2730, "lr": 2.693901439758778e-06}
{"train_loss": 0.021175729110836983, "global_step": 243010, "epoch": 2730, "lr": 2.693713718161012e-06}
{"train_loss": 0.07042142748832703, "global_step": 243011, "epoch": 2730, "lr": 2.6935260029229913e-06}
{"train_loss": 0.03922777622938156, "global_step": 243012, "epoch": 2730, "lr": 2.6933382940447503e-06}
{"train_loss": 0.04410590976476669, "global_step": 243013, "epoch": 2730, "lr": 2.6931505915263165e-06}
{"train_loss": 0.02912004292011261, "global_step": 243014, "epoch": 2730, "lr": 2.692962895367701e-06}
{"train_loss": 0.036856647580862045, "global_step": 243015, "epoch": 2730, "lr": 2.6927752055689538e-06}
{"train_loss": 0.04795657470822334, "global_step": 243016, "epoch": 2730, "lr": 2.6925875221300743e-06}
{"train_loss": 0.022333983331918716, "global_step": 243017, "epoch": 2730, "lr": 2.6923998450511134e-06}
{"train_loss": 0.0367242693901062, "global_step": 243018, "epoch": 2730, "lr": 2.6922121743320705e-06}
{"train_loss": 0.059773124754428864, "global_step": 243019, "epoch": 2730, "lr": 2.6920245099729847e-06}
{"train_loss": 0.08906389772891998, "global_step": 243020, "epoch": 2730, "lr": 2.6918368519738835e-06}
{"train_loss": 0.06321126967668533, "global_step": 243021, "epoch": 2730, "lr": 2.691649200334784e-06}
{"train_loss": 0.036628056317567825, "global_step": 243022, "epoch": 2730, "lr": 2.691461555055724e-06}
{"train_loss": 0.04317229986190796, "global_step": 243023, "epoch": 2730, "lr": 2.6912739161367105e-06}
{"train_loss": 0.01816108636558056, "global_step": 243024, "epoch": 2730, "lr": 2.6910862835777873e-06}
{"train_loss": 0.040872856974601746, "global_step": 243025, "epoch": 2730, "lr": 2.6908986573789596e-06}
{"train_loss": 0.021735306829214096, "global_step": 243026, "epoch": 2730, "lr": 2.690711037540278e-06}
{"train_loss": 0.0344773605465889, "global_step": 243027, "epoch": 2730, "lr": 2.6905234240617417e-06}
{"train_loss": 0.03723807632923126, "global_step": 243028, "epoch": 2730, "lr": 2.6903358169433954e-06}
{"train_loss": 0.06661422550678253, "global_step": 243029, "epoch": 2730, "lr": 2.690148216185251e-06}
{"train_loss": 0.051485396921634674, "global_step": 243030, "epoch": 2730, "lr": 2.689960621787346e-06}
{"train_loss": 0.02444935031235218, "global_step": 243031, "epoch": 2730, "lr": 2.6897730337496927e-06}
{"train_loss": 0.045449718832969666, "global_step": 243032, "epoch": 2730, "lr": 2.6895854520723296e-06}
{"train_loss": 0.11648255586624146, "global_step": 243033, "epoch": 2730, "lr": 2.689397876755273e-06}
{"train_loss": 0.050083376467227936, "global_step": 243034, "epoch": 2730, "lr": 2.6892103077985507e-06}
{"train_loss": 0.08694775402545929, "global_step": 243035, "epoch": 2730, "lr": 2.68902274520218e-06}
{"train_loss": 0.10390298813581467, "global_step": 243036, "epoch": 2730, "lr": 2.6888351889662043e-06}
{"train_loss": 0.11109047383069992, "global_step": 243037, "epoch": 2730, "lr": 2.68864763909063e-06}
{"train_loss": 0.03235350921750069, "global_step": 243038, "epoch": 2730, "lr": 2.6884600955754902e-06}
{"train_loss": 0.016848258674144745, "global_step": 243039, "epoch": 2730, "lr": 2.6882725584208178e-06}
{"train_loss": 0.025719445198774338, "global_step": 243040, "epoch": 2730, "lr": 2.6880850276266245e-06}
{"train_loss": 0.03843151777982712, "global_step": 243041, "epoch": 2730, "lr": 2.687897503192943e-06}
{"train_loss": 0.05381155014038086, "global_step": 243042, "epoch": 2730, "lr": 2.687709985119796e-06}
{"train_loss": 0.025822971016168594, "global_step": 243043, "epoch": 2730, "lr": 2.6875224734072113e-06}
{"train_loss": 0.04349315166473389, "global_step": 243044, "epoch": 2730, "lr": 2.6873349680552163e-06}
{"train_loss": 0.04076189547777176, "global_step": 243045, "epoch": 2730, "lr": 2.6871474690638275e-06}
{"train_loss": 0.062389977276325226, "global_step": 243046, "epoch": 2730, "lr": 2.686959976433079e-06}
{"train_loss": 0.04808322340250015, "global_step": 243047, "epoch": 2730, "lr": 2.6867724901629863e-06}
{"train_loss": 0.04304364696145058, "global_step": 243048, "epoch": 2730, "lr": 2.6865850102535837e-06}
{"train_loss": 0.05927601084113121, "global_step": 243049, "epoch": 2730, "lr": 2.686397536704899e-06}
{"train_loss": 0.024155793711543083, "global_step": 243050, "epoch": 2730, "lr": 2.6862100695169423e-06}
{"train_loss": 0.02239079400897026, "global_step": 243051, "epoch": 2730, "lr": 2.6860226086897534e-06}
{"train_loss": 0.07255445420742035, "global_step": 243052, "epoch": 2730, "lr": 2.6858351542233427e-06}
{"train_loss": 0.06695125252008438, "global_step": 243053, "epoch": 2730, "lr": 2.6856477061177556e-06}
{"train_loss": 0.05852722004055977, "global_step": 243054, "epoch": 2730, "lr": 2.685460264373002e-06}
{"train_loss": 0.07330620288848877, "global_step": 243055, "epoch": 2730, "lr": 2.685272828989105e-06}
{"train_loss": 0.07303474843502045, "global_step": 243056, "epoch": 2730, "lr": 2.6850853999661084e-06}
{"train_loss": 0.03675997629761696, "global_step": 243057, "epoch": 2730, "lr": 2.6848979773040184e-06}
{"train_loss": 0.048422577602558595, "global_step": 243058, "epoch": 2730, "lr": 2.684710561002873e-06, "val_loss": 9.443533897399902, "train_action_mse_error": 1.5716619491577148}
{"train_loss": 0.03136134520173073, "global_step": 243059, "epoch": 2731, "lr": 2.6845231510626844e-06}
{"train_loss": 0.024992166087031364, "global_step": 243060, "epoch": 2731, "lr": 2.6843357474834907e-06}
{"train_loss": 0.023351669311523438, "global_step": 243061, "epoch": 2731, "lr": 2.684148350265303e-06}
{"train_loss": 0.030295949429273605, "global_step": 243062, "epoch": 2731, "lr": 2.683960959408166e-06}
{"train_loss": 0.04701422527432442, "global_step": 243063, "epoch": 2731, "lr": 2.683773574912085e-06}
{"train_loss": 0.04415634647011757, "global_step": 243064, "epoch": 2731, "lr": 2.6835861967770993e-06}
{"train_loss": 0.03698861971497536, "global_step": 243065, "epoch": 2731, "lr": 2.683398825003225e-06}
{"train_loss": 0.03625962510704994, "global_step": 243066, "epoch": 2731, "lr": 2.6832114595904956e-06}
{"train_loss": 0.03138432279229164, "global_step": 243067, "epoch": 2731, "lr": 2.683024100538928e-06}
{"train_loss": 0.03348599001765251, "global_step": 243068, "epoch": 2731, "lr": 2.682836747848555e-06}
{"train_loss": 0.029278099536895752, "global_step": 243069, "epoch": 2731, "lr": 2.6826494015193882e-06}
{"train_loss": 0.07531807571649551, "global_step": 243070, "epoch": 2731, "lr": 2.6824620615514774e-06}
{"train_loss": 0.05148401856422424, "global_step": 243071, "epoch": 2731, "lr": 2.682274727944817e-06}
{"train_loss": 0.025698957964777946, "global_step": 243072, "epoch": 2731, "lr": 2.682087400699462e-06}
{"train_loss": 0.05018649250268936, "global_step": 243073, "epoch": 2731, "lr": 2.6819000798154137e-06}
{"train_loss": 0.10357055813074112, "global_step": 243074, "epoch": 2731, "lr": 2.6817127652927042e-06}
{"train_loss": 0.0494673028588295, "global_step": 243075, "epoch": 2731, "lr": 2.681525457131373e-06}
{"train_loss": 0.05347316712141037, "global_step": 243076, "epoch": 2731, "lr": 2.6813381553314254e-06}
{"train_loss": 0.05803167074918747, "global_step": 243077, "epoch": 2731, "lr": 2.6811508598929057e-06}
{"train_loss": 0.07042116671800613, "global_step": 243078, "epoch": 2731, "lr": 2.6809635708158143e-06}
{"train_loss": 0.052056312561035156, "global_step": 243079, "epoch": 2731, "lr": 2.6807762881002063e-06}
{"train_loss": 0.05971873924136162, "global_step": 243080, "epoch": 2731, "lr": 2.680589011746076e-06}
{"train_loss": 0.047803279012441635, "global_step": 243081, "epoch": 2731, "lr": 2.6804017417534744e-06}
{"train_loss": 0.032946620136499405, "global_step": 243082, "epoch": 2731, "lr": 2.680214478122417e-06}
{"train_loss": 0.02858850173652172, "global_step": 243083, "epoch": 2731, "lr": 2.680027220852921e-06}
{"train_loss": 0.04723898693919182, "global_step": 243084, "epoch": 2731, "lr": 2.679839969945025e-06}
{"train_loss": 0.04424896091222763, "global_step": 243085, "epoch": 2731, "lr": 2.67965272539874e-06}
{"train_loss": 0.06888847053050995, "global_step": 243086, "epoch": 2731, "lr": 2.6794654872141056e-06}
{"train_loss": 0.0501534640789032, "global_step": 243087, "epoch": 2731, "lr": 2.679278255391143e-06}
{"train_loss": 0.03333977237343788, "global_step": 243088, "epoch": 2731, "lr": 2.6790910299298646e-06}
{"train_loss": 0.056895993649959564, "global_step": 243089, "epoch": 2731, "lr": 2.6789038108303133e-06}
{"train_loss": 0.023958900943398476, "global_step": 243090, "epoch": 2731, "lr": 2.6787165980925012e-06}
{"train_loss": 0.031738147139549255, "global_step": 243091, "epoch": 2731, "lr": 2.678529391716461e-06}
{"train_loss": 0.06819433718919754, "global_step": 243092, "epoch": 2731, "lr": 2.6783421917022154e-06}
{"train_loss": 0.02584560215473175, "global_step": 243093, "epoch": 2731, "lr": 2.6781549980497865e-06}
{"train_loss": 0.053794365376234055, "global_step": 243094, "epoch": 2731, "lr": 2.6779678107592134e-06}
{"train_loss": 0.07037331163883209, "global_step": 243095, "epoch": 2731, "lr": 2.6777806298304954e-06}
{"train_loss": 0.035275306552648544, "global_step": 243096, "epoch": 2731, "lr": 2.677593455263688e-06}
{"train_loss": 0.06222881004214287, "global_step": 243097, "epoch": 2731, "lr": 2.6774062870587926e-06}
{"train_loss": 0.053851418197155, "global_step": 243098, "epoch": 2731, "lr": 2.6772191252158462e-06}
{"train_loss": 0.031554948538541794, "global_step": 243099, "epoch": 2731, "lr": 2.6770319697348665e-06}
{"train_loss": 0.03190326690673828, "global_step": 243100, "epoch": 2731, "lr": 2.6768448206158924e-06}
{"train_loss": 0.0428810715675354, "global_step": 243101, "epoch": 2731, "lr": 2.6766576778589293e-06}
{"train_loss": 0.028985504060983658, "global_step": 243102, "epoch": 2731, "lr": 2.6764705414640266e-06}
{"train_loss": 0.046990443021059036, "global_step": 243103, "epoch": 2731, "lr": 2.6762834114311798e-06}
{"train_loss": 0.05169086903333664, "global_step": 243104, "epoch": 2731, "lr": 2.6760962877604436e-06}
{"train_loss": 0.03277774900197983, "global_step": 243105, "epoch": 2731, "lr": 2.675909170451818e-06}
{"train_loss": 0.10329203307628632, "global_step": 243106, "epoch": 2731, "lr": 2.6757220595053476e-06}
{"train_loss": 0.04303470626473427, "global_step": 243107, "epoch": 2731, "lr": 2.675534954921044e-06}
{"train_loss": 0.04029757156968117, "global_step": 243108, "epoch": 2731, "lr": 2.6753478566989453e-06}
{"train_loss": 0.013566762208938599, "global_step": 243109, "epoch": 2731, "lr": 2.6751607648390575e-06}
{"train_loss": 0.04627383127808571, "global_step": 243110, "epoch": 2731, "lr": 2.6749736793414192e-06}
{"train_loss": 0.04259675741195679, "global_step": 243111, "epoch": 2731, "lr": 2.674786600206064e-06}
{"train_loss": 0.015742113813757896, "global_step": 243112, "epoch": 2731, "lr": 2.674599527432997e-06}
{"train_loss": 0.06613147258758545, "global_step": 243113, "epoch": 2731, "lr": 2.6744124610222634e-06}
{"train_loss": 0.08363976329565048, "global_step": 243114, "epoch": 2731, "lr": 2.674225400973873e-06}
{"train_loss": 0.04208356887102127, "global_step": 243115, "epoch": 2731, "lr": 2.674038347287855e-06}
{"train_loss": 0.08580554276704788, "global_step": 243116, "epoch": 2731, "lr": 2.673851299964236e-06}
{"train_loss": 0.02901841513812542, "global_step": 243117, "epoch": 2731, "lr": 2.6736642590030446e-06}
{"train_loss": 0.02582663856446743, "global_step": 243118, "epoch": 2731, "lr": 2.673477224404297e-06}
{"train_loss": 0.08879247307777405, "global_step": 243119, "epoch": 2731, "lr": 2.6732901961680267e-06}
{"train_loss": 0.08800749480724335, "global_step": 243120, "epoch": 2731, "lr": 2.673103174294256e-06}
{"train_loss": 0.030680250376462936, "global_step": 243121, "epoch": 2731, "lr": 2.672916158783012e-06}
{"train_loss": 0.06056015565991402, "global_step": 243122, "epoch": 2731, "lr": 2.6727291496343176e-06}
{"train_loss": 0.056600045412778854, "global_step": 243123, "epoch": 2731, "lr": 2.672542146848189e-06}
{"train_loss": 0.06390930712223053, "global_step": 243124, "epoch": 2731, "lr": 2.672355150424671e-06}
{"train_loss": 0.05262395739555359, "global_step": 243125, "epoch": 2731, "lr": 2.6721681603637694e-06}
{"train_loss": 0.0462072379887104, "global_step": 243126, "epoch": 2731, "lr": 2.671981176665522e-06}
{"train_loss": 0.05754198506474495, "global_step": 243127, "epoch": 2731, "lr": 2.6717941993299577e-06}
{"train_loss": 0.028103843331336975, "global_step": 243128, "epoch": 2731, "lr": 2.6716072283570815e-06}
{"train_loss": 0.06965534389019012, "global_step": 243129, "epoch": 2731, "lr": 2.671420263746932e-06}
{"train_loss": 0.0846354067325592, "global_step": 243130, "epoch": 2731, "lr": 2.6712333054995376e-06}
{"train_loss": 0.070010244846344, "global_step": 243131, "epoch": 2731, "lr": 2.6710463536149144e-06}
{"train_loss": 0.07664743810892105, "global_step": 243132, "epoch": 2731, "lr": 2.670859408093096e-06}
{"train_loss": 0.015338780358433723, "global_step": 243133, "epoch": 2731, "lr": 2.670672468934099e-06}
{"train_loss": 0.047860655933618546, "global_step": 243134, "epoch": 2731, "lr": 2.6704855361379623e-06}
{"train_loss": 0.05894027277827263, "global_step": 243135, "epoch": 2731, "lr": 2.670298609704691e-06}
{"train_loss": 0.0637524425983429, "global_step": 243136, "epoch": 2731, "lr": 2.6701116896343358e-06}
{"train_loss": 0.03311215341091156, "global_step": 243137, "epoch": 2731, "lr": 2.669924775926891e-06}
{"train_loss": 0.030406560748815536, "global_step": 243138, "epoch": 2731, "lr": 2.6697378685824116e-06}
{"train_loss": 0.03156645596027374, "global_step": 243139, "epoch": 2731, "lr": 2.669550967600898e-06}
{"train_loss": 0.03206408768892288, "global_step": 243140, "epoch": 2731, "lr": 2.6693640729823942e-06}
{"train_loss": 0.03374958783388138, "global_step": 243141, "epoch": 2731, "lr": 2.669177184726912e-06}
{"train_loss": 0.04812800511717796, "global_step": 243142, "epoch": 2731, "lr": 2.6689903028344896e-06}
{"train_loss": 0.04937943443655968, "global_step": 243143, "epoch": 2731, "lr": 2.668803427305133e-06}
{"train_loss": 0.03155549615621567, "global_step": 243144, "epoch": 2731, "lr": 2.668616558138892e-06}
{"train_loss": 0.02479730360209942, "global_step": 243145, "epoch": 2731, "lr": 2.668429695335767e-06}
{"train_loss": 0.05599046126008034, "global_step": 243146, "epoch": 2731, "lr": 2.6682428388957957e-06}
{"train_loss": 0.04774875254527237, "global_step": 243147, "epoch": 2731, "lr": 2.6680559888190126e-06, "val_loss": 9.541905403137207}
{"train_loss": 0.037113677710294724, "global_step": 243148, "epoch": 2732, "lr": 2.6678691451054227e-06}
{"train_loss": 0.038523536175489426, "global_step": 243149, "epoch": 2732, "lr": 2.667682307755065e-06}
{"train_loss": 0.030509449541568756, "global_step": 243150, "epoch": 2732, "lr": 2.667495476767956e-06}
{"train_loss": 0.053611863404512405, "global_step": 243151, "epoch": 2732, "lr": 2.667308652144135e-06}
{"train_loss": 0.05119389668107033, "global_step": 243152, "epoch": 2732, "lr": 2.6671218338836067e-06}
{"train_loss": 0.06759592890739441, "global_step": 243153, "epoch": 2732, "lr": 2.6669350219864166e-06}
{"train_loss": 0.02734963223338127, "global_step": 243154, "epoch": 2732, "lr": 2.666748216452569e-06}
{"train_loss": 0.037417784333229065, "global_step": 243155, "epoch": 2732, "lr": 2.666561417282115e-06}
{"train_loss": 0.03590461239218712, "global_step": 243156, "epoch": 2732, "lr": 2.6663746244750485e-06}
{"train_loss": 0.038050975650548935, "global_step": 243157, "epoch": 2732, "lr": 2.6661878380314255e-06}
{"train_loss": 0.05316459760069847, "global_step": 243158, "epoch": 2732, "lr": 2.666001057951245e-06}
{"train_loss": 0.06687499582767487, "global_step": 243159, "epoch": 2732, "lr": 2.6658142842345524e-06}
{"train_loss": 0.028138073161244392, "global_step": 243160, "epoch": 2732, "lr": 2.665627516881364e-06}
{"train_loss": 0.05495069548487663, "global_step": 243161, "epoch": 2732, "lr": 2.6654407558917017e-06}
{"train_loss": 0.037538185715675354, "global_step": 243162, "epoch": 2732, "lr": 2.6652540012655936e-06}
{"train_loss": 0.060331881046295166, "global_step": 243163, "epoch": 2732, "lr": 2.665067253003062e-06}
{"train_loss": 0.02817528508603573, "global_step": 243164, "epoch": 2732, "lr": 2.6648805111041455e-06}
{"train_loss": 0.03960328549146652, "global_step": 243165, "epoch": 2732, "lr": 2.6646937755688496e-06}
{"train_loss": 0.030196648091077805, "global_step": 243166, "epoch": 2732, "lr": 2.664507046397213e-06}
{"train_loss": 0.017918817698955536, "global_step": 243167, "epoch": 2732, "lr": 2.6643203235892535e-06}
{"train_loss": 0.0344180166721344, "global_step": 243168, "epoch": 2732, "lr": 2.664133607145003e-06}
{"train_loss": 0.021486496552824974, "global_step": 243169, "epoch": 2732, "lr": 2.6639468970644734e-06}
{"train_loss": 0.05068173259496689, "global_step": 243170, "epoch": 2732, "lr": 2.663760193347714e-06}
{"train_loss": 0.04001464694738388, "global_step": 243171, "epoch": 2732, "lr": 2.6635734959947202e-06}
{"train_loss": 0.04471933841705322, "global_step": 243172, "epoch": 2732, "lr": 2.6633868050055465e-06}
{"train_loss": 0.03306155279278755, "global_step": 243173, "epoch": 2732, "lr": 2.663200120380188e-06}
{"train_loss": 0.04014424979686737, "global_step": 243174, "epoch": 2732, "lr": 2.6630134421187002e-06}
{"train_loss": 0.027779458090662956, "global_step": 243175, "epoch": 2732, "lr": 2.662826770221083e-06}
{"train_loss": 0.020299546420574188, "global_step": 243176, "epoch": 2732, "lr": 2.662640104687375e-06}
{"train_loss": 0.044427916407585144, "global_step": 243177, "epoch": 2732, "lr": 2.662453445517599e-06}
{"train_loss": 0.03934721276164055, "global_step": 243178, "epoch": 2732, "lr": 2.662266792711782e-06}
{"train_loss": 0.040084999054670334, "global_step": 243179, "epoch": 2732, "lr": 2.662080146269935e-06}
{"train_loss": 0.01768503151834011, "global_step": 243180, "epoch": 2732, "lr": 2.661893506192109e-06}
{"train_loss": 0.0400678776204586, "global_step": 243181, "epoch": 2732, "lr": 2.661706872478303e-06}
{"train_loss": 0.05513571575284004, "global_step": 243182, "epoch": 2732, "lr": 2.6615202451285516e-06}
{"train_loss": 0.05298464372754097, "global_step": 243183, "epoch": 2732, "lr": 2.661333624142892e-06}
{"train_loss": 0.0492376871407032, "global_step": 243184, "epoch": 2732, "lr": 2.661147009521331e-06}
{"train_loss": 0.05729836970567703, "global_step": 243185, "epoch": 2732, "lr": 2.660960401263912e-06}
{"train_loss": 0.009000223129987717, "global_step": 243186, "epoch": 2732, "lr": 2.6607737993706416e-06}
{"train_loss": 0.1042303740978241, "global_step": 243187, "epoch": 2732, "lr": 2.660587203841558e-06}
{"train_loss": 0.04575364664196968, "global_step": 243188, "epoch": 2732, "lr": 2.660400614676678e-06}
{"train_loss": 0.018921563401818275, "global_step": 243189, "epoch": 2732, "lr": 2.660214031876035e-06}
{"train_loss": 0.05677631497383118, "global_step": 243190, "epoch": 2732, "lr": 2.66002745543964e-06}
{"train_loss": 0.06670103967189789, "global_step": 243191, "epoch": 2732, "lr": 2.6598408853675372e-06}
{"train_loss": 0.05814291536808014, "global_step": 243192, "epoch": 2732, "lr": 2.6596543216597323e-06}
{"train_loss": 0.04010305926203728, "global_step": 243193, "epoch": 2732, "lr": 2.6594677643162703e-06}
{"train_loss": 0.030017094686627388, "global_step": 243194, "epoch": 2732, "lr": 2.659281213337156e-06}
{"train_loss": 0.01702127419412136, "global_step": 243195, "epoch": 2732, "lr": 2.6590946687224338e-06}
{"train_loss": 0.10698434710502625, "global_step": 243196, "epoch": 2732, "lr": 2.6589081304721096e-06}
{"train_loss": 0.021594546735286713, "global_step": 243197, "epoch": 2732, "lr": 2.658721598586228e-06}
{"train_loss": 0.030124634504318237, "global_step": 243198, "epoch": 2732, "lr": 2.6585350730647994e-06}
{"train_loss": 0.04791141673922539, "global_step": 243199, "epoch": 2732, "lr": 2.6583485539078523e-06}
{"train_loss": 0.0387788861989975, "global_step": 243200, "epoch": 2732, "lr": 2.658162041115414e-06}
{"train_loss": 0.029778126627206802, "global_step": 243201, "epoch": 2732, "lr": 2.6579755346875068e-06}
{"train_loss": 0.08821135759353638, "global_step": 243202, "epoch": 2732, "lr": 2.657789034624164e-06}
{"train_loss": 0.03853636234998703, "global_step": 243203, "epoch": 2732, "lr": 2.6576025409253967e-06}
{"train_loss": 0.05537246912717819, "global_step": 243204, "epoch": 2732, "lr": 2.657416053591244e-06}
{"train_loss": 0.040121398866176605, "global_step": 243205, "epoch": 2732, "lr": 2.657229572621722e-06}
{"train_loss": 0.05586802214384079, "global_step": 243206, "epoch": 2732, "lr": 2.6570430980168593e-06}
{"train_loss": 0.05370355769991875, "global_step": 243207, "epoch": 2732, "lr": 2.656856629776677e-06}
{"train_loss": 0.045503728091716766, "global_step": 243208, "epoch": 2732, "lr": 2.6566701679012094e-06}
{"train_loss": 0.056945446878671646, "global_step": 243209, "epoch": 2732, "lr": 2.656483712390473e-06}
{"train_loss": 0.08430911600589752, "global_step": 243210, "epoch": 2732, "lr": 2.656297263244495e-06}
{"train_loss": 0.029427973553538322, "global_step": 243211, "epoch": 2732, "lr": 2.656110820463298e-06}
{"train_loss": 0.052164897322654724, "global_step": 243212, "epoch": 2732, "lr": 2.655924384046915e-06}
{"train_loss": 0.029511630535125732, "global_step": 243213, "epoch": 2732, "lr": 2.6557379539953575e-06}
{"train_loss": 0.058782223612070084, "global_step": 243214, "epoch": 2732, "lr": 2.6555515303086696e-06}
{"train_loss": 0.05570143088698387, "global_step": 243215, "epoch": 2732, "lr": 2.6553651129868574e-06}
{"train_loss": 0.04326189309358597, "global_step": 243216, "epoch": 2732, "lr": 2.655178702029959e-06}
{"train_loss": 0.0244322307407856, "global_step": 243217, "epoch": 2732, "lr": 2.6549922974379913e-06}
{"train_loss": 0.02348499558866024, "global_step": 243218, "epoch": 2732, "lr": 2.654805899210988e-06}
{"train_loss": 0.021521171554923058, "global_step": 243219, "epoch": 2732, "lr": 2.65461950734896e-06}
{"train_loss": 0.08682131767272949, "global_step": 243220, "epoch": 2732, "lr": 2.654433121851946e-06}
{"train_loss": 0.04241343215107918, "global_step": 243221, "epoch": 2732, "lr": 2.6542467427199736e-06}
{"train_loss": 0.0496455579996109, "global_step": 243222, "epoch": 2732, "lr": 2.6540603699530487e-06}
{"train_loss": 0.0546397864818573, "global_step": 243223, "epoch": 2732, "lr": 2.6538740035512155e-06}
{"train_loss": 0.06093323230743408, "global_step": 243224, "epoch": 2732, "lr": 2.6536876435144906e-06}
{"train_loss": 0.04522198438644409, "global_step": 243225, "epoch": 2732, "lr": 2.6535012898429024e-06}
{"train_loss": 0.05495723336935043, "global_step": 243226, "epoch": 2732, "lr": 2.653314942536467e-06}
{"train_loss": 0.062377240508794785, "global_step": 243227, "epoch": 2732, "lr": 2.653128601595228e-06}
{"train_loss": 0.051206305623054504, "global_step": 243228, "epoch": 2732, "lr": 2.652942267019187e-06}
{"train_loss": 0.03585834428668022, "global_step": 243229, "epoch": 2732, "lr": 2.6527559388083878e-06}
{"train_loss": 0.03560492396354675, "global_step": 243230, "epoch": 2732, "lr": 2.652569616962841e-06}
{"train_loss": 0.0392332561314106, "global_step": 243231, "epoch": 2732, "lr": 2.6523833014825916e-06}
{"train_loss": 0.13711729645729065, "global_step": 243232, "epoch": 2732, "lr": 2.6521969923676393e-06}
{"train_loss": 0.021581018343567848, "global_step": 243233, "epoch": 2732, "lr": 2.652010689618034e-06}
{"train_loss": 0.058991413563489914, "global_step": 243234, "epoch": 2732, "lr": 2.651824393233776e-06}
{"train_loss": 0.08487347513437271, "global_step": 243235, "epoch": 2732, "lr": 2.651638103214915e-06}
{"train_loss": 0.046056368880057603, "global_step": 243236, "epoch": 2732, "lr": 2.651451819561457e-06, "val_loss": 9.477639198303223}
{"train_loss": 0.048647116869688034, "global_step": 243237, "epoch": 2733, "lr": 2.65126554227344e-06}
{"train_loss": 0.024882517755031586, "global_step": 243238, "epoch": 2733, "lr": 2.6510792713508815e-06}
{"train_loss": 0.04089689627289772, "global_step": 243239, "epoch": 2733, "lr": 2.650893006793803e-06}
{"train_loss": 0.05567629635334015, "global_step": 243240, "epoch": 2733, "lr": 2.6507067486022385e-06}
{"train_loss": 0.02070956490933895, "global_step": 243241, "epoch": 2733, "lr": 2.650520496776204e-06}
{"train_loss": 0.05675220862030983, "global_step": 243242, "epoch": 2733, "lr": 2.6503342513157393e-06}
{"train_loss": 0.05124974995851517, "global_step": 243243, "epoch": 2733, "lr": 2.6501480122208545e-06}
{"train_loss": 0.05000841245055199, "global_step": 243244, "epoch": 2733, "lr": 2.649961779491583e-06}
{"train_loss": 0.016232240945100784, "global_step": 243245, "epoch": 2733, "lr": 2.6497755531279424e-06}
{"train_loss": 0.05537779629230499, "global_step": 243246, "epoch": 2733, "lr": 2.64958933312997e-06}
{"train_loss": 0.06009918451309204, "global_step": 243247, "epoch": 2733, "lr": 2.6494031194976733e-06}
{"train_loss": 0.03797294571995735, "global_step": 243248, "epoch": 2733, "lr": 2.649216912231095e-06}
{"train_loss": 0.06236465647816658, "global_step": 243249, "epoch": 2733, "lr": 2.6490307113302472e-06}
{"train_loss": 0.04476385936141014, "global_step": 243250, "epoch": 2733, "lr": 2.648844516795168e-06}
{"train_loss": 0.05829281732439995, "global_step": 243251, "epoch": 2733, "lr": 2.6486583286258635e-06}
{"train_loss": 0.04546958953142166, "global_step": 243252, "epoch": 2733, "lr": 2.6484721468223785e-06}
{"train_loss": 0.027367595583200455, "global_step": 243253, "epoch": 2733, "lr": 2.648285971384723e-06}
{"train_loss": 0.01797117292881012, "global_step": 243254, "epoch": 2733, "lr": 2.6480998023129368e-06}
{"train_loss": 0.046079121530056, "global_step": 243255, "epoch": 2733, "lr": 2.6479136396070246e-06}
{"train_loss": 0.040310345590114594, "global_step": 243256, "epoch": 2733, "lr": 2.6477274832670262e-06}
{"train_loss": 0.022263003513216972, "global_step": 243257, "epoch": 2733, "lr": 2.6475413332929743e-06}
{"train_loss": 0.018794627860188484, "global_step": 243258, "epoch": 2733, "lr": 2.647355189684869e-06}
{"train_loss": 0.03942394256591797, "global_step": 243259, "epoch": 2733, "lr": 2.6471690524427604e-06}
{"train_loss": 0.042741890996694565, "global_step": 243260, "epoch": 2733, "lr": 2.6469829215666595e-06}
{"train_loss": 0.07023593038320541, "global_step": 243261, "epoch": 2733, "lr": 2.6467967970565944e-06}
{"train_loss": 0.03612680360674858, "global_step": 243262, "epoch": 2733, "lr": 2.6466106789125867e-06}
{"train_loss": 0.049160830676555634, "global_step": 243263, "epoch": 2733, "lr": 2.6464245671346753e-06}
{"train_loss": 0.04989545792341232, "global_step": 243264, "epoch": 2733, "lr": 2.6462384617228607e-06}
{"train_loss": 0.051967889070510864, "global_step": 243265, "epoch": 2733, "lr": 2.6460523626771983e-06}
{"train_loss": 0.03390413895249367, "global_step": 243266, "epoch": 2733, "lr": 2.6458662699976823e-06}
{"train_loss": 0.04609740152955055, "global_step": 243267, "epoch": 2733, "lr": 2.6456801836843625e-06}
{"train_loss": 0.04057472571730614, "global_step": 243268, "epoch": 2733, "lr": 2.645494103737245e-06}
{"train_loss": 0.07311559468507767, "global_step": 243269, "epoch": 2733, "lr": 2.6453080301563737e-06}
{"train_loss": 0.03334570676088333, "global_step": 243270, "epoch": 2733, "lr": 2.6451219629417547e-06}
{"train_loss": 0.02788802608847618, "global_step": 243271, "epoch": 2733, "lr": 2.644935902093426e-06}
{"train_loss": 0.04056226462125778, "global_step": 243272, "epoch": 2733, "lr": 2.6447498476114053e-06}
{"train_loss": 0.03214750066399574, "global_step": 243273, "epoch": 2733, "lr": 2.644563799495731e-06}
{"train_loss": 0.05525484308600426, "global_step": 243274, "epoch": 2733, "lr": 2.644377757746402e-06}
{"train_loss": 0.10842519998550415, "global_step": 243275, "epoch": 2733, "lr": 2.644191722363476e-06}
{"train_loss": 0.04278048500418663, "global_step": 243276, "epoch": 2733, "lr": 2.644005693346957e-06}
{"train_loss": 0.01757574826478958, "global_step": 243277, "epoch": 2733, "lr": 2.643819670696862e-06}
{"train_loss": 0.06396450102329254, "global_step": 243278, "epoch": 2733, "lr": 2.643633654413241e-06}
{"train_loss": 0.048893485218286514, "global_step": 243279, "epoch": 2733, "lr": 2.6434476444961e-06}
{"train_loss": 0.03626686707139015, "global_step": 243280, "epoch": 2733, "lr": 2.6432616409454767e-06}
{"train_loss": 0.06132505461573601, "global_step": 243281, "epoch": 2733, "lr": 2.643075643761378e-06}
{"train_loss": 0.10123658180236816, "global_step": 243282, "epoch": 2733, "lr": 2.642889652943853e-06}
{"train_loss": 0.0188002847135067, "global_step": 243283, "epoch": 2733, "lr": 2.6427036684929074e-06}
{"train_loss": 0.06596294045448303, "global_step": 243284, "epoch": 2733, "lr": 2.6425176904085745e-06}
{"train_loss": 0.0760948583483696, "global_step": 243285, "epoch": 2733, "lr": 2.642331718690877e-06}
{"train_loss": 0.031069237738847733, "global_step": 243286, "epoch": 2733, "lr": 2.642145753339842e-06}
{"train_loss": 0.020350869745016098, "global_step": 243287, "epoch": 2733, "lr": 2.641959794355492e-06}
{"train_loss": 0.051612187176942825, "global_step": 243288, "epoch": 2733, "lr": 2.6417738417378548e-06}
{"train_loss": 0.07158800959587097, "global_step": 243289, "epoch": 2733, "lr": 2.641587895486952e-06}
{"train_loss": 0.02011900022625923, "global_step": 243290, "epoch": 2733, "lr": 2.6414019556028124e-06}
{"train_loss": 0.02054426446557045, "global_step": 243291, "epoch": 2733, "lr": 2.6412160220854577e-06}
{"train_loss": 0.04177731275558472, "global_step": 243292, "epoch": 2733, "lr": 2.64103009493491e-06}
{"train_loss": 0.0351078100502491, "global_step": 243293, "epoch": 2733, "lr": 2.640844174151208e-06}
{"train_loss": 0.01669563725590706, "global_step": 243294, "epoch": 2733, "lr": 2.6406582597343575e-06}
{"train_loss": 0.03999767079949379, "global_step": 243295, "epoch": 2733, "lr": 2.640472351684403e-06}
{"train_loss": 0.0321379192173481, "global_step": 243296, "epoch": 2733, "lr": 2.640286450001356e-06}
{"train_loss": 0.05559443682432175, "global_step": 243297, "epoch": 2733, "lr": 2.640100554685249e-06}
{"train_loss": 0.026337387040257454, "global_step": 243298, "epoch": 2733, "lr": 2.639914665736093e-06}
{"train_loss": 0.0521731935441494, "global_step": 243299, "epoch": 2733, "lr": 2.6397287831539332e-06}
{"train_loss": 0.032061152160167694, "global_step": 243300, "epoch": 2733, "lr": 2.6395429069387745e-06}
{"train_loss": 0.04073682054877281, "global_step": 243301, "epoch": 2733, "lr": 2.6393570370906617e-06}
{"train_loss": 0.07084975391626358, "global_step": 243302, "epoch": 2733, "lr": 2.639171173609606e-06}
{"train_loss": 0.057804785668849945, "global_step": 243303, "epoch": 2733, "lr": 2.6389853164956402e-06}
{"train_loss": 0.08624784648418427, "global_step": 243304, "epoch": 2733, "lr": 2.638799465748776e-06}
{"train_loss": 0.031569674611091614, "global_step": 243305, "epoch": 2733, "lr": 2.6386136213690626e-06}
{"train_loss": 0.061782509088516235, "global_step": 243306, "epoch": 2733, "lr": 2.638427783356495e-06}
{"train_loss": 0.07638303190469742, "global_step": 243307, "epoch": 2733, "lr": 2.6382419517111234e-06}
{"train_loss": 0.028920281678438187, "global_step": 243308, "epoch": 2733, "lr": 2.6380561264329584e-06}
{"train_loss": 0.06236453354358673, "global_step": 243309, "epoch": 2733, "lr": 2.6378703075220333e-06}
{"train_loss": 0.02854485996067524, "global_step": 243310, "epoch": 2733, "lr": 2.637684494978365e-06}
{"train_loss": 0.03028968535363674, "global_step": 243311, "epoch": 2733, "lr": 2.637498688801987e-06}
{"train_loss": 0.047728531062603, "global_step": 243312, "epoch": 2733, "lr": 2.6373128889929156e-06}
{"train_loss": 0.023622117936611176, "global_step": 243313, "epoch": 2733, "lr": 2.6371270955511896e-06}
{"train_loss": 0.02815655805170536, "global_step": 243314, "epoch": 2733, "lr": 2.6369413084768203e-06}
{"train_loss": 0.04777498170733452, "global_step": 243315, "epoch": 2733, "lr": 2.6367555277698297e-06}
{"train_loss": 0.059518374502658844, "global_step": 243316, "epoch": 2733, "lr": 2.636569753430257e-06}
{"train_loss": 0.0310490895062685, "global_step": 243317, "epoch": 2733, "lr": 2.636383985458113e-06}
{"train_loss": 0.02323160693049431, "global_step": 243318, "epoch": 2733, "lr": 2.6361982238534423e-06}
{"train_loss": 0.049655329436063766, "global_step": 243319, "epoch": 2733, "lr": 2.6360124686162446e-06}
{"train_loss": 0.041107580065727234, "global_step": 243320, "epoch": 2733, "lr": 2.6358267197465645e-06}
{"train_loss": 0.019689910113811493, "global_step": 243321, "epoch": 2733, "lr": 2.635640977244419e-06}
{"train_loss": 0.05675361678004265, "global_step": 243322, "epoch": 2733, "lr": 2.6354552411098354e-06}
{"train_loss": 0.02944815158843994, "global_step": 243323, "epoch": 2733, "lr": 2.6352695113428307e-06}
{"train_loss": 0.043269719928503036, "global_step": 243324, "epoch": 2733, "lr": 2.6350837879434487e-06}
{"train_loss": 0.043721096815143745, "global_step": 243325, "epoch": 2733, "lr": 2.63489807091169e-06, "val_loss": 9.41600227355957}
{"train_loss": 0.07102697342634201, "global_step": 243326, "epoch": 2734, "lr": 2.634712360247604e-06}
{"train_loss": 0.0460309162735939, "global_step": 243327, "epoch": 2734, "lr": 2.6345266559511915e-06}
{"train_loss": 0.06280041486024857, "global_step": 243328, "epoch": 2734, "lr": 2.634340958022491e-06}
{"train_loss": 0.02493021823465824, "global_step": 243329, "epoch": 2734, "lr": 2.6341552664615354e-06}
{"train_loss": 0.09289807081222534, "global_step": 243330, "epoch": 2734, "lr": 2.633969581268331e-06}
{"train_loss": 0.08173272013664246, "global_step": 243331, "epoch": 2734, "lr": 2.633783902442921e-06}
{"train_loss": 0.09873299300670624, "global_step": 243332, "epoch": 2734, "lr": 2.6335982299853122e-06}
{"train_loss": 0.039729852229356766, "global_step": 243333, "epoch": 2734, "lr": 2.633412563895549e-06}
{"train_loss": 0.06841682642698288, "global_step": 243334, "epoch": 2734, "lr": 2.6332269041736356e-06}
{"train_loss": 0.06426401436328888, "global_step": 243335, "epoch": 2734, "lr": 2.633041250819618e-06}
{"train_loss": 0.027133435010910034, "global_step": 243336, "epoch": 2734, "lr": 2.6328556038335006e-06}
{"train_loss": 0.02015298418700695, "global_step": 243337, "epoch": 2734, "lr": 2.6326699632153283e-06}
{"train_loss": 0.02971380203962326, "global_step": 243338, "epoch": 2734, "lr": 2.632484328965107e-06}
{"train_loss": 0.09558850526809692, "global_step": 243339, "epoch": 2734, "lr": 2.6322987010828805e-06}
{"train_loss": 0.0433478020131588, "global_step": 243340, "epoch": 2734, "lr": 2.6321130795686546e-06}
{"train_loss": 0.03425191342830658, "global_step": 243341, "epoch": 2734, "lr": 2.6319274644224677e-06}
{"train_loss": 0.09807117283344269, "global_step": 243342, "epoch": 2734, "lr": 2.6317418556443376e-06}
{"train_loss": 0.014372631907463074, "global_step": 243343, "epoch": 2734, "lr": 2.631556253234302e-06}
{"train_loss": 0.016950441524386406, "global_step": 243344, "epoch": 2734, "lr": 2.6313706571923668e-06}
{"train_loss": 0.10676021873950958, "global_step": 243345, "epoch": 2734, "lr": 2.631185067518571e-06}
{"train_loss": 0.034979283809661865, "global_step": 243346, "epoch": 2734, "lr": 2.6309994842129315e-06}
{"train_loss": 0.07218509167432785, "global_step": 243347, "epoch": 2734, "lr": 2.630813907275481e-06}
{"train_loss": 0.10156693309545517, "global_step": 243348, "epoch": 2734, "lr": 2.6306283367062367e-06}
{"train_loss": 0.0889759436249733, "global_step": 243349, "epoch": 2734, "lr": 2.630442772505232e-06}
{"train_loss": 0.06437170505523682, "global_step": 243350, "epoch": 2734, "lr": 2.6302572146724824e-06}
{"train_loss": 0.04952578619122505, "global_step": 243351, "epoch": 2734, "lr": 2.6300716632080227e-06}
{"train_loss": 0.05641772225499153, "global_step": 243352, "epoch": 2734, "lr": 2.629886118111863e-06}
{"train_loss": 0.04619866982102394, "global_step": 243353, "epoch": 2734, "lr": 2.6297005793840478e-06}
{"train_loss": 0.05048821121454239, "global_step": 243354, "epoch": 2734, "lr": 2.629515047024589e-06}
{"train_loss": 0.027725325897336006, "global_step": 243355, "epoch": 2734, "lr": 2.629329521033508e-06}
{"train_loss": 0.017724068835377693, "global_step": 243356, "epoch": 2734, "lr": 2.629144001410844e-06}
{"train_loss": 0.03316594287753105, "global_step": 243357, "epoch": 2734, "lr": 2.6289584881566074e-06}
{"train_loss": 0.016551680862903595, "global_step": 243358, "epoch": 2734, "lr": 2.6287729812708382e-06}
{"train_loss": 0.054326120764017105, "global_step": 243359, "epoch": 2734, "lr": 2.6285874807535472e-06}
{"train_loss": 0.01782422885298729, "global_step": 243360, "epoch": 2734, "lr": 2.628401986604767e-06}
{"train_loss": 0.06635336577892303, "global_step": 243361, "epoch": 2734, "lr": 2.6282164988245205e-06}
{"train_loss": 0.053291887044906616, "global_step": 243362, "epoch": 2734, "lr": 2.6280310174128352e-06}
{"train_loss": 0.020141858607530594, "global_step": 243363, "epoch": 2734, "lr": 2.627845542369728e-06}
{"train_loss": 0.08751267939805984, "global_step": 243364, "epoch": 2734, "lr": 2.6276600736952317e-06}
{"train_loss": 0.023883258923888206, "global_step": 243365, "epoch": 2734, "lr": 2.6274746113893744e-06}
{"train_loss": 0.03268462419509888, "global_step": 243366, "epoch": 2734, "lr": 2.6272891554521673e-06}
{"train_loss": 0.024442259222269058, "global_step": 243367, "epoch": 2734, "lr": 2.627103705883649e-06}
{"train_loss": 0.008845250122249126, "global_step": 243368, "epoch": 2734, "lr": 2.626918262683836e-06}
{"train_loss": 0.0681166797876358, "global_step": 243369, "epoch": 2734, "lr": 2.6267328258527624e-06}
{"train_loss": 0.05790944769978523, "global_step": 243370, "epoch": 2734, "lr": 2.6265473953904384e-06}
{"train_loss": 0.037345316261053085, "global_step": 243371, "epoch": 2734, "lr": 2.626361971296909e-06}
{"train_loss": 0.03161855787038803, "global_step": 243372, "epoch": 2734, "lr": 2.6261765535721796e-06}
{"train_loss": 0.05011838674545288, "global_step": 243373, "epoch": 2734, "lr": 2.625991142216289e-06}
{"train_loss": 0.032028499990701675, "global_step": 243374, "epoch": 2734, "lr": 2.6258057372292476e-06}
{"train_loss": 0.016680995002388954, "global_step": 243375, "epoch": 2734, "lr": 2.625620338611101e-06}
{"train_loss": 0.05772528424859047, "global_step": 243376, "epoch": 2734, "lr": 2.625434946361849e-06}
{"train_loss": 0.03736548498272896, "global_step": 243377, "epoch": 2734, "lr": 2.625249560481541e-06}
{"train_loss": 0.038515739142894745, "global_step": 243378, "epoch": 2734, "lr": 2.6250641809701824e-06}
{"train_loss": 0.039911262691020966, "global_step": 243379, "epoch": 2734, "lr": 2.624878807827819e-06}
{"train_loss": 0.04733424633741379, "global_step": 243380, "epoch": 2734, "lr": 2.624693441054449e-06}
{"train_loss": 0.09636228531599045, "global_step": 243381, "epoch": 2734, "lr": 2.6245080806501233e-06}
{"train_loss": 0.04333801567554474, "global_step": 243382, "epoch": 2734, "lr": 2.624322726614842e-06}
{"train_loss": 0.046568214893341064, "global_step": 243383, "epoch": 2734, "lr": 2.624137378948649e-06}
{"train_loss": 0.07253506779670715, "global_step": 243384, "epoch": 2734, "lr": 2.623952037651567e-06}
{"train_loss": 0.06332516670227051, "global_step": 243385, "epoch": 2734, "lr": 2.6237667027236125e-06}
{"train_loss": 0.047201063483953476, "global_step": 243386, "epoch": 2734, "lr": 2.6235813741648184e-06}
{"train_loss": 0.04178968071937561, "global_step": 243387, "epoch": 2734, "lr": 2.623396051975213e-06}
{"train_loss": 0.03657210245728493, "global_step": 243388, "epoch": 2734, "lr": 2.6232107361548017e-06}
{"train_loss": 0.05539555102586746, "global_step": 243389, "epoch": 2734, "lr": 2.6230254267036292e-06}
{"train_loss": 0.03317711129784584, "global_step": 243390, "epoch": 2734, "lr": 2.6228401236217117e-06}
{"train_loss": 0.07261838018894196, "global_step": 243391, "epoch": 2734, "lr": 2.6226548269090766e-06}
{"train_loss": 0.0663219466805458, "global_step": 243392, "epoch": 2734, "lr": 2.6224695365657527e-06}
{"train_loss": 0.012256674468517303, "global_step": 243393, "epoch": 2734, "lr": 2.6222842525917503e-06}
{"train_loss": 0.042796019464731216, "global_step": 243394, "epoch": 2734, "lr": 2.6220989749871137e-06}
{"train_loss": 0.031545933336019516, "global_step": 243395, "epoch": 2734, "lr": 2.6219137037518494e-06}
{"train_loss": 0.04205026105046272, "global_step": 243396, "epoch": 2734, "lr": 2.621728438886001e-06}
{"train_loss": 0.040232643485069275, "global_step": 243397, "epoch": 2734, "lr": 2.621543180389574e-06}
{"train_loss": 0.08222631365060806, "global_step": 243398, "epoch": 2734, "lr": 2.621357928262613e-06}
{"train_loss": 0.06916989386081696, "global_step": 243399, "epoch": 2734, "lr": 2.621172682505124e-06}
{"train_loss": 0.016633570194244385, "global_step": 243400, "epoch": 2734, "lr": 2.62098744311714e-06}
{"train_loss": 0.07036297768354416, "global_step": 243401, "epoch": 2734, "lr": 2.620802210098694e-06}
{"train_loss": 0.04677681252360344, "global_step": 243402, "epoch": 2734, "lr": 2.620616983449797e-06}
{"train_loss": 0.031623475253582, "global_step": 243403, "epoch": 2734, "lr": 2.6204317631704887e-06}
{"train_loss": 0.025929270312190056, "global_step": 243404, "epoch": 2734, "lr": 2.6202465492607795e-06}
{"train_loss": 0.03754105418920517, "global_step": 243405, "epoch": 2734, "lr": 2.6200613417207085e-06}
{"train_loss": 0.07417835295200348, "global_step": 243406, "epoch": 2734, "lr": 2.6198761405502813e-06}
{"train_loss": 0.018858950585126877, "global_step": 243407, "epoch": 2734, "lr": 2.6196909457495477e-06}
{"train_loss": 0.08101995289325714, "global_step": 243408, "epoch": 2734, "lr": 2.6195057573185078e-06}
{"train_loss": 0.03388486057519913, "global_step": 243409, "epoch": 2734, "lr": 2.619320575257206e-06}
{"train_loss": 0.061946045607328415, "global_step": 243410, "epoch": 2734, "lr": 2.6191353995656476e-06}
{"train_loss": 0.09641619771718979, "global_step": 243411, "epoch": 2734, "lr": 2.6189502302438827e-06}
{"train_loss": 0.02445697970688343, "global_step": 243412, "epoch": 2734, "lr": 2.6187650672919115e-06}
{"train_loss": 0.07024647295475006, "global_step": 243413, "epoch": 2734, "lr": 2.6185799107097784e-06}
{"train_loss": 0.049442682624532934, "global_step": 243414, "epoch": 2734, "lr": 2.618394760497489e-06, "val_loss": 9.494489669799805}
{"train_loss": 0.01483531016856432, "global_step": 243415, "epoch": 2735, "lr": 2.618209616655093e-06}
{"train_loss": 0.014314155094325542, "global_step": 243416, "epoch": 2735, "lr": 2.618024479182585e-06}
{"train_loss": 0.04509589076042175, "global_step": 243417, "epoch": 2735, "lr": 2.6178393480800202e-06}
{"train_loss": 0.044724833220243454, "global_step": 243418, "epoch": 2735, "lr": 2.6176542233473987e-06}
{"train_loss": 0.04008738696575165, "global_step": 243419, "epoch": 2735, "lr": 2.6174691049847597e-06}
{"train_loss": 0.03374730795621872, "global_step": 243420, "epoch": 2735, "lr": 2.6172839929921312e-06}
{"train_loss": 0.044308554381132126, "global_step": 243421, "epoch": 2735, "lr": 2.6170988873695234e-06}
{"train_loss": 0.057143375277519226, "global_step": 243422, "epoch": 2735, "lr": 2.616913788116976e-06}
{"train_loss": 0.044152528047561646, "global_step": 243423, "epoch": 2735, "lr": 2.6167286952344993e-06}
{"train_loss": 0.05319530889391899, "global_step": 243424, "epoch": 2735, "lr": 2.616543608722133e-06}
{"train_loss": 0.024663429707288742, "global_step": 243425, "epoch": 2735, "lr": 2.616358528579893e-06}
{"train_loss": 0.058667488396167755, "global_step": 243426, "epoch": 2735, "lr": 2.616173454807802e-06}
{"train_loss": 0.06702971458435059, "global_step": 243427, "epoch": 2735, "lr": 2.615988387405893e-06}
{"train_loss": 0.07240678369998932, "global_step": 243428, "epoch": 2735, "lr": 2.615803326374183e-06}
{"train_loss": 0.04188193753361702, "global_step": 243429, "epoch": 2735, "lr": 2.6156182717127053e-06}
{"train_loss": 0.07636471092700958, "global_step": 243430, "epoch": 2735, "lr": 2.615433223421482e-06}
{"train_loss": 0.02091352827847004, "global_step": 243431, "epoch": 2735, "lr": 2.6152481815005293e-06}
{"train_loss": 0.051754508167505264, "global_step": 243432, "epoch": 2735, "lr": 2.6150631459498862e-06}
{"train_loss": 0.025073973461985588, "global_step": 243433, "epoch": 2735, "lr": 2.6148781167695646e-06}
{"train_loss": 0.02856229990720749, "global_step": 243434, "epoch": 2735, "lr": 2.6146930939595972e-06}
{"train_loss": 0.022656919434666634, "global_step": 243435, "epoch": 2735, "lr": 2.6145080775200004e-06}
{"train_loss": 0.060222044587135315, "global_step": 243436, "epoch": 2735, "lr": 2.614323067450808e-06}
{"train_loss": 0.004516459070146084, "global_step": 243437, "epoch": 2735, "lr": 2.614138063752053e-06}
{"train_loss": 0.08612377196550369, "global_step": 243438, "epoch": 2735, "lr": 2.6139530664237355e-06}
{"train_loss": 0.026607930660247803, "global_step": 243439, "epoch": 2735, "lr": 2.6137680754659056e-06}
{"train_loss": 0.047097593545913696, "global_step": 243440, "epoch": 2735, "lr": 2.6135830908785685e-06}
{"train_loss": 0.01990126445889473, "global_step": 243441, "epoch": 2735, "lr": 2.6133981126617633e-06}
{"train_loss": 0.038524940609931946, "global_step": 243442, "epoch": 2735, "lr": 2.613213140815507e-06}
{"train_loss": 0.01638355292379856, "global_step": 243443, "epoch": 2735, "lr": 2.6130281753398267e-06}
{"train_loss": 0.06385787576436996, "global_step": 243444, "epoch": 2735, "lr": 2.612843216234745e-06}
{"train_loss": 0.025367382913827896, "global_step": 243445, "epoch": 2735, "lr": 2.612658263500295e-06}
{"train_loss": 0.031649667769670486, "global_step": 243446, "epoch": 2735, "lr": 2.612473317136488e-06}
{"train_loss": 0.050497181713581085, "global_step": 243447, "epoch": 2735, "lr": 2.6122883771433625e-06}
{"train_loss": 0.03893497586250305, "global_step": 243448, "epoch": 2735, "lr": 2.6121034435209357e-06}
{"train_loss": 0.04682706296443939, "global_step": 243449, "epoch": 2735, "lr": 2.611918516269235e-06}
{"train_loss": 0.033289436250925064, "global_step": 243450, "epoch": 2735, "lr": 2.6117335953882773e-06}
{"train_loss": 0.07238343358039856, "global_step": 243451, "epoch": 2735, "lr": 2.6115486808781067e-06}
{"train_loss": 0.04467296972870827, "global_step": 243452, "epoch": 2735, "lr": 2.611363772738723e-06}
{"train_loss": 0.024832941591739655, "global_step": 243453, "epoch": 2735, "lr": 2.6111788709701766e-06}
{"train_loss": 0.049003809690475464, "global_step": 243454, "epoch": 2735, "lr": 2.6109939755724676e-06}
{"train_loss": 0.04335933178663254, "global_step": 243455, "epoch": 2735, "lr": 2.6108090865456347e-06}
{"train_loss": 0.056248318403959274, "global_step": 243456, "epoch": 2735, "lr": 2.6106242038897055e-06}
{"train_loss": 0.08234349638223648, "global_step": 243457, "epoch": 2735, "lr": 2.610439327604697e-06}
{"train_loss": 0.03942355886101723, "global_step": 243458, "epoch": 2735, "lr": 2.610254457690642e-06}
{"train_loss": 0.0581011101603508, "global_step": 243459, "epoch": 2735, "lr": 2.610069594147557e-06}
{"train_loss": 0.029330773279070854, "global_step": 243460, "epoch": 2735, "lr": 2.609884736975471e-06}
{"train_loss": 0.023336969316005707, "global_step": 243461, "epoch": 2735, "lr": 2.609699886174405e-06}
{"train_loss": 0.039334896951913834, "global_step": 243462, "epoch": 2735, "lr": 2.609515041744398e-06}
{"train_loss": 0.03195047378540039, "global_step": 243463, "epoch": 2735, "lr": 2.6093302036854615e-06}
{"train_loss": 0.0699169710278511, "global_step": 243464, "epoch": 2735, "lr": 2.609145371997612e-06}
{"train_loss": 0.03209817409515381, "global_step": 243465, "epoch": 2735, "lr": 2.6089605466809e-06}
{"train_loss": 0.019357791170477867, "global_step": 243466, "epoch": 2735, "lr": 2.608775727735324e-06}
{"train_loss": 0.06325434893369675, "global_step": 243467, "epoch": 2735, "lr": 2.6085909151609245e-06}
{"train_loss": 0.033779747784137726, "global_step": 243468, "epoch": 2735, "lr": 2.6084061089577173e-06}
{"train_loss": 0.034646183252334595, "global_step": 243469, "epoch": 2735, "lr": 2.608221309125741e-06}
{"train_loss": 0.024911217391490936, "global_step": 243470, "epoch": 2735, "lr": 2.6080365156650135e-06}
{"train_loss": 0.04963189363479614, "global_step": 243471, "epoch": 2735, "lr": 2.60785172857555e-06}
{"train_loss": 0.03971624746918678, "global_step": 243472, "epoch": 2735, "lr": 2.6076669478573846e-06}
{"train_loss": 0.07399626821279526, "global_step": 243473, "epoch": 2735, "lr": 2.6074821735105394e-06}
{"train_loss": 0.07188218086957932, "global_step": 243474, "epoch": 2735, "lr": 2.6072974055350417e-06}
{"train_loss": 0.010985992848873138, "global_step": 243475, "epoch": 2735, "lr": 2.6071126439309144e-06}
{"train_loss": 0.023271912708878517, "global_step": 243476, "epoch": 2735, "lr": 2.606927888698185e-06}
{"train_loss": 0.03366736322641373, "global_step": 243477, "epoch": 2735, "lr": 2.6067431398368815e-06}
{"train_loss": 0.04250404238700867, "global_step": 243478, "epoch": 2735, "lr": 2.606558397347014e-06}
{"train_loss": 0.04300116375088692, "global_step": 243479, "epoch": 2735, "lr": 2.606373661228623e-06}
{"train_loss": 0.04546863213181496, "global_step": 243480, "epoch": 2735, "lr": 2.606188931481723e-06}
{"train_loss": 0.06890802830457687, "global_step": 243481, "epoch": 2735, "lr": 2.6060042081063498e-06}
{"train_loss": 0.045389048755168915, "global_step": 243482, "epoch": 2735, "lr": 2.6058194911025124e-06}
{"train_loss": 0.04812522232532501, "global_step": 243483, "epoch": 2735, "lr": 2.6056347804702562e-06}
{"train_loss": 0.044879354536533356, "global_step": 243484, "epoch": 2735, "lr": 2.605450076209587e-06}
{"train_loss": 0.039194099605083466, "global_step": 243485, "epoch": 2735, "lr": 2.605265378320543e-06}
{"train_loss": 0.04803336411714554, "global_step": 243486, "epoch": 2735, "lr": 2.6050806868031353e-06}
{"train_loss": 0.03669719770550728, "global_step": 243487, "epoch": 2735, "lr": 2.6048960016574086e-06}
{"train_loss": 0.03528643772006035, "global_step": 243488, "epoch": 2735, "lr": 2.604711322883363e-06}
{"train_loss": 0.048886362463235855, "global_step": 243489, "epoch": 2735, "lr": 2.6045266504810483e-06}
{"train_loss": 0.03304621949791908, "global_step": 243490, "epoch": 2735, "lr": 2.6043419844504644e-06}
{"train_loss": 0.05432847514748573, "global_step": 243491, "epoch": 2735, "lr": 2.6041573247916507e-06}
{"train_loss": 0.06540035456418991, "global_step": 243492, "epoch": 2735, "lr": 2.60397267150464e-06}
{"train_loss": 0.04843593388795853, "global_step": 243493, "epoch": 2735, "lr": 2.603788024589443e-06}
{"train_loss": 0.022247740998864174, "global_step": 243494, "epoch": 2735, "lr": 2.6036033840460938e-06}
{"train_loss": 0.023863526061177254, "global_step": 243495, "epoch": 2735, "lr": 2.603418749874603e-06}
{"train_loss": 0.0481736995279789, "global_step": 243496, "epoch": 2735, "lr": 2.60323412207501e-06}
{"train_loss": 0.03696535900235176, "global_step": 243497, "epoch": 2735, "lr": 2.603049500647331e-06}
{"train_loss": 0.050215255469083786, "global_step": 243498, "epoch": 2735, "lr": 2.6028648855916047e-06}
{"train_loss": 0.07217574119567871, "global_step": 243499, "epoch": 2735, "lr": 2.602680276907832e-06}
{"train_loss": 0.0379483737051487, "global_step": 243500, "epoch": 2735, "lr": 2.6024956745960616e-06}
{"train_loss": 0.06330665946006775, "global_step": 243501, "epoch": 2735, "lr": 2.6023110786563e-06}
{"train_loss": 0.05904477834701538, "global_step": 243502, "epoch": 2735, "lr": 2.6021264890885856e-06}
{"train_loss": 0.043380340676378, "global_step": 243503, "epoch": 2735, "lr": 2.6019419058929407e-06, "val_loss": 9.451042175292969, "train_action_mse_error": 15.676292419433594}
{"train_loss": 0.05876663327217102, "global_step": 243504, "epoch": 2736, "lr": 2.601757329069382e-06}
{"train_loss": 0.08305863291025162, "global_step": 243505, "epoch": 2736, "lr": 2.601572758617943e-06}
{"train_loss": 0.051813483238220215, "global_step": 243506, "epoch": 2736, "lr": 2.601388194538634e-06}
{"train_loss": 0.015036395750939846, "global_step": 243507, "epoch": 2736, "lr": 2.6012036368315007e-06}
{"train_loss": 0.0328017920255661, "global_step": 243508, "epoch": 2736, "lr": 2.601019085496559e-06}
{"train_loss": 0.08396788686513901, "global_step": 243509, "epoch": 2736, "lr": 2.6008345405338252e-06}
{"train_loss": 0.03661586344242096, "global_step": 243510, "epoch": 2736, "lr": 2.600650001943333e-06}
{"train_loss": 0.029053300619125366, "global_step": 243511, "epoch": 2736, "lr": 2.600465469725111e-06}
{"train_loss": 0.01899878866970539, "global_step": 243512, "epoch": 2736, "lr": 2.6002809438791744e-06}
{"train_loss": 0.020602239295840263, "global_step": 243513, "epoch": 2736, "lr": 2.600096424405557e-06}
{"train_loss": 0.03185851499438286, "global_step": 243514, "epoch": 2736, "lr": 2.5999119113042703e-06}
{"train_loss": 0.017347581684589386, "global_step": 243515, "epoch": 2736, "lr": 2.599727404575353e-06}
{"train_loss": 0.030040286481380463, "global_step": 243516, "epoch": 2736, "lr": 2.5995429042188213e-06}
{"train_loss": 0.019527681171894073, "global_step": 243517, "epoch": 2736, "lr": 2.599358410234709e-06}
{"train_loss": 0.022563517093658447, "global_step": 243518, "epoch": 2736, "lr": 2.5991739226230326e-06}
{"train_loss": 0.042047664523124695, "global_step": 243519, "epoch": 2736, "lr": 2.59898944138382e-06}
{"train_loss": 0.05076916143298149, "global_step": 243520, "epoch": 2736, "lr": 2.5988049665170933e-06}
{"train_loss": 0.038566652685403824, "global_step": 243521, "epoch": 2736, "lr": 2.5986204980228856e-06}
{"train_loss": 0.07279487699270248, "global_step": 243522, "epoch": 2736, "lr": 2.5984360359012083e-06}
{"train_loss": 0.040558334439992905, "global_step": 243523, "epoch": 2736, "lr": 2.5982515801521003e-06}
{"train_loss": 0.07318645715713501, "global_step": 243524, "epoch": 2736, "lr": 2.5980671307755723e-06}
{"train_loss": 0.025914398953318596, "global_step": 243525, "epoch": 2736, "lr": 2.5978826877716632e-06}
{"train_loss": 0.0748315304517746, "global_step": 243526, "epoch": 2736, "lr": 2.5976982511403845e-06}
{"train_loss": 0.0631784126162529, "global_step": 243527, "epoch": 2736, "lr": 2.597513820881764e-06}
{"train_loss": 0.04260760545730591, "global_step": 243528, "epoch": 2736, "lr": 2.59732939699584e-06}
{"train_loss": 0.04682237654924393, "global_step": 243529, "epoch": 2736, "lr": 2.597144979482624e-06}
{"train_loss": 0.03066685050725937, "global_step": 243530, "epoch": 2736, "lr": 2.5969605683421437e-06}
{"train_loss": 0.06988124549388885, "global_step": 243531, "epoch": 2736, "lr": 2.5967761635744216e-06}
{"train_loss": 0.052526891231536865, "global_step": 243532, "epoch": 2736, "lr": 2.596591765179496e-06}
{"train_loss": 0.017105069011449814, "global_step": 243533, "epoch": 2736, "lr": 2.5964073731573666e-06}
{"train_loss": 0.03061668761074543, "global_step": 243534, "epoch": 2736, "lr": 2.5962229875080845e-06}
{"train_loss": 0.02157294750213623, "global_step": 243535, "epoch": 2736, "lr": 2.5960386082316547e-06}
{"train_loss": 0.03891322761774063, "global_step": 243536, "epoch": 2736, "lr": 2.59585423532811e-06}
{"train_loss": 0.058026574552059174, "global_step": 243537, "epoch": 2736, "lr": 2.5956698687974734e-06}
{"train_loss": 0.06442634761333466, "global_step": 243538, "epoch": 2736, "lr": 2.5954855086397834e-06}
{"train_loss": 0.031861841678619385, "global_step": 243539, "epoch": 2736, "lr": 2.59530115485504e-06}
{"train_loss": 0.052861761301755905, "global_step": 243540, "epoch": 2736, "lr": 2.5951168074432875e-06}
{"train_loss": 0.03195566684007645, "global_step": 243541, "epoch": 2736, "lr": 2.5949324664045427e-06}
{"train_loss": 0.020775025710463524, "global_step": 243542, "epoch": 2736, "lr": 2.594748131738828e-06}
{"train_loss": 0.02139187976717949, "global_step": 243543, "epoch": 2736, "lr": 2.594563803446176e-06}
{"train_loss": 0.053355589509010315, "global_step": 243544, "epoch": 2736, "lr": 2.5943794815265988e-06}
{"train_loss": 0.04294200986623764, "global_step": 243545, "epoch": 2736, "lr": 2.59419516598014e-06}
{"train_loss": 0.032481931149959564, "global_step": 243546, "epoch": 2736, "lr": 2.5940108568068057e-06}
{"train_loss": 0.02349235676229, "global_step": 243547, "epoch": 2736, "lr": 2.5938265540066344e-06}
{"train_loss": 0.039688777178525925, "global_step": 243548, "epoch": 2736, "lr": 2.5936422575796373e-06}
{"train_loss": 0.046605199575424194, "global_step": 243549, "epoch": 2736, "lr": 2.5934579675258587e-06}
{"train_loss": 0.05132279545068741, "global_step": 243550, "epoch": 2736, "lr": 2.5932736838453043e-06}
{"train_loss": 0.014535671100020409, "global_step": 243551, "epoch": 2736, "lr": 2.5930894065380075e-06}
{"train_loss": 0.025778431445360184, "global_step": 243552, "epoch": 2736, "lr": 2.5929051356039903e-06}
{"train_loss": 0.035802990198135376, "global_step": 243553, "epoch": 2736, "lr": 2.592720871043286e-06}
{"train_loss": 0.07262472808361053, "global_step": 243554, "epoch": 2736, "lr": 2.592536612855906e-06}
{"train_loss": 0.06819351017475128, "global_step": 243555, "epoch": 2736, "lr": 2.592352361041883e-06}
{"train_loss": 0.04050377383828163, "global_step": 243556, "epoch": 2736, "lr": 2.59216811560124e-06}
{"train_loss": 0.052331168204545975, "global_step": 243557, "epoch": 2736, "lr": 2.591983876534004e-06}
{"train_loss": 0.028341500088572502, "global_step": 243558, "epoch": 2736, "lr": 2.5917996438401982e-06}
{"train_loss": 0.05203435570001602, "global_step": 243559, "epoch": 2736, "lr": 2.5916154175198493e-06}
{"train_loss": 0.06276480108499527, "global_step": 243560, "epoch": 2736, "lr": 2.5914311975729744e-06}
{"train_loss": 0.021060427650809288, "global_step": 243561, "epoch": 2736, "lr": 2.5912469839996066e-06}
{"train_loss": 0.051014065742492676, "global_step": 243562, "epoch": 2736, "lr": 2.591062776799763e-06}
{"train_loss": 0.04832691699266434, "global_step": 243563, "epoch": 2736, "lr": 2.5908785759734767e-06}
{"train_loss": 0.03276129066944122, "global_step": 243564, "epoch": 2736, "lr": 2.5906943815207697e-06}
{"train_loss": 0.10709493607282639, "global_step": 243565, "epoch": 2736, "lr": 2.590510193441664e-06}
{"train_loss": 0.03771848604083061, "global_step": 243566, "epoch": 2736, "lr": 2.5903260117361936e-06}
{"train_loss": 0.053427353501319885, "global_step": 243567, "epoch": 2736, "lr": 2.590141836404364e-06}
{"train_loss": 0.05502074211835861, "global_step": 243568, "epoch": 2736, "lr": 2.5899576674462246e-06}
{"train_loss": 0.016774173825979233, "global_step": 243569, "epoch": 2736, "lr": 2.5897735048617813e-06}
{"train_loss": 0.03444903716444969, "global_step": 243570, "epoch": 2736, "lr": 2.589589348651067e-06}
{"train_loss": 0.04507145658135414, "global_step": 243571, "epoch": 2736, "lr": 2.5894051988140988e-06}
{"train_loss": 0.021743901073932648, "global_step": 243572, "epoch": 2736, "lr": 2.5892210553509155e-06}
{"train_loss": 0.03766009956598282, "global_step": 243573, "epoch": 2736, "lr": 2.589036918261528e-06}
{"train_loss": 0.04184231907129288, "global_step": 243574, "epoch": 2736, "lr": 2.58885278754597e-06}
{"train_loss": 0.028506020084023476, "global_step": 243575, "epoch": 2736, "lr": 2.588668663204258e-06}
{"train_loss": 0.08503683656454086, "global_step": 243576, "epoch": 2736, "lr": 2.5884845452364247e-06}
{"train_loss": 0.09083759784698486, "global_step": 243577, "epoch": 2736, "lr": 2.588300433642493e-06}
{"train_loss": 0.06778567284345627, "global_step": 243578, "epoch": 2736, "lr": 2.588116328422485e-06}
{"train_loss": 0.05340733379125595, "global_step": 243579, "epoch": 2736, "lr": 2.587932229576423e-06}
{"train_loss": 0.04521534591913223, "global_step": 243580, "epoch": 2736, "lr": 2.5877481371043454e-06}
{"train_loss": 0.036899372935295105, "global_step": 243581, "epoch": 2736, "lr": 2.587564051006264e-06}
{"train_loss": 0.03724830970168114, "global_step": 243582, "epoch": 2736, "lr": 2.5873799712822e-06}
{"train_loss": 0.03897830471396446, "global_step": 243583, "epoch": 2736, "lr": 2.587195897932193e-06}
{"train_loss": 0.04848584160208702, "global_step": 243584, "epoch": 2736, "lr": 2.587011830956254e-06}
{"train_loss": 0.036717187613248825, "global_step": 243585, "epoch": 2736, "lr": 2.5868277703544165e-06}
{"train_loss": 0.09220913052558899, "global_step": 243586, "epoch": 2736, "lr": 2.5866437161266963e-06}
{"train_loss": 0.059194955974817276, "global_step": 243587, "epoch": 2736, "lr": 2.5864596682731333e-06}
{"train_loss": 0.0741180032491684, "global_step": 243588, "epoch": 2736, "lr": 2.586275626793738e-06}
{"train_loss": 0.027125049382448196, "global_step": 243589, "epoch": 2736, "lr": 2.5860915916885444e-06}
{"train_loss": 0.07216669619083405, "global_step": 243590, "epoch": 2736, "lr": 2.5859075629575625e-06}
{"train_loss": 0.02201865054666996, "global_step": 243591, "epoch": 2736, "lr": 2.5857235406008372e-06}
{"train_loss": 0.04471763965328423, "global_step": 243592, "epoch": 2736, "lr": 2.5855395246183746e-06, "val_loss": 9.529459953308105}
{"train_loss": 0.05545342341065407, "global_step": 243593, "epoch": 2737, "lr": 2.585355515010218e-06}
{"train_loss": 0.043360497802495956, "global_step": 243594, "epoch": 2737, "lr": 2.5851715117763742e-06}
{"train_loss": 0.04661654680967331, "global_step": 243595, "epoch": 2737, "lr": 2.5849875149168865e-06}
{"train_loss": 0.05238481983542442, "global_step": 243596, "epoch": 2737, "lr": 2.584803524431756e-06}
{"train_loss": 0.07783298194408417, "global_step": 243597, "epoch": 2737, "lr": 2.5846195403210318e-06}
{"train_loss": 0.05279053375124931, "global_step": 243598, "epoch": 2737, "lr": 2.5844355625847194e-06}
{"train_loss": 0.04599285498261452, "global_step": 243599, "epoch": 2737, "lr": 2.5842515912228584e-06}
{"train_loss": 0.05619049072265625, "global_step": 243600, "epoch": 2737, "lr": 2.5840676262354646e-06}
{"train_loss": 0.03595179691910744, "global_step": 243601, "epoch": 2737, "lr": 2.583883667622561e-06}
{"train_loss": 0.05510047823190689, "global_step": 243602, "epoch": 2737, "lr": 2.58369971538418e-06}
{"train_loss": 0.05365164950489998, "global_step": 243603, "epoch": 2737, "lr": 2.5835157695203394e-06}
{"train_loss": 0.04512341693043709, "global_step": 243604, "epoch": 2737, "lr": 2.5833318300310717e-06}
{"train_loss": 0.05528584122657776, "global_step": 243605, "epoch": 2737, "lr": 2.5831478969163935e-06}
{"train_loss": 0.044681672006845474, "global_step": 243606, "epoch": 2737, "lr": 2.5829639701763388e-06}
{"train_loss": 0.009328334592282772, "global_step": 243607, "epoch": 2737, "lr": 2.5827800498109235e-06}
{"train_loss": 0.03805878385901451, "global_step": 243608, "epoch": 2737, "lr": 2.5825961358201758e-06}
{"train_loss": 0.03364977985620499, "global_step": 243609, "epoch": 2737, "lr": 2.5824122282041175e-06}
{"train_loss": 0.04935275390744209, "global_step": 243610, "epoch": 2737, "lr": 2.582228326962782e-06}
{"train_loss": 0.026807818561792374, "global_step": 243611, "epoch": 2737, "lr": 2.5820444320961813e-06}
{"train_loss": 0.04428247734904289, "global_step": 243612, "epoch": 2737, "lr": 2.5818605436043473e-06}
{"train_loss": 0.041273925453424454, "global_step": 243613, "epoch": 2737, "lr": 2.5816766614873033e-06}
{"train_loss": 0.04317294433712959, "global_step": 243614, "epoch": 2737, "lr": 2.581492785745082e-06}
{"train_loss": 0.046322450041770935, "global_step": 243615, "epoch": 2737, "lr": 2.581308916377695e-06}
{"train_loss": 0.03360629454255104, "global_step": 243616, "epoch": 2737, "lr": 2.5811250533851805e-06}
{"train_loss": 0.019943485036492348, "global_step": 243617, "epoch": 2737, "lr": 2.5809411967675445e-06}
{"train_loss": 0.04858546704053879, "global_step": 243618, "epoch": 2737, "lr": 2.5807573465248313e-06}
{"train_loss": 0.045033954083919525, "global_step": 243619, "epoch": 2737, "lr": 2.5805735026570576e-06}
{"train_loss": 0.05966796353459358, "global_step": 243620, "epoch": 2737, "lr": 2.5803896651642456e-06}
{"train_loss": 0.030668698251247406, "global_step": 243621, "epoch": 2737, "lr": 2.580205834046423e-06}
{"train_loss": 0.06987550109624863, "global_step": 243622, "epoch": 2737, "lr": 2.5800220093036067e-06}
{"train_loss": 0.04175436496734619, "global_step": 243623, "epoch": 2737, "lr": 2.5798381909358406e-06}
{"train_loss": 0.07228504121303558, "global_step": 243624, "epoch": 2737, "lr": 2.579654378943125e-06}
{"train_loss": 0.06189221143722534, "global_step": 243625, "epoch": 2737, "lr": 2.5794705733255096e-06}
{"train_loss": 0.03593143820762634, "global_step": 243626, "epoch": 2737, "lr": 2.579286774082995e-06}
{"train_loss": 0.06443783640861511, "global_step": 243627, "epoch": 2737, "lr": 2.5791029812156255e-06}
{"train_loss": 0.04954993352293968, "global_step": 243628, "epoch": 2737, "lr": 2.5789191947234114e-06}
{"train_loss": 0.05357073247432709, "global_step": 243629, "epoch": 2737, "lr": 2.578735414606387e-06}
{"train_loss": 0.039238907396793365, "global_step": 243630, "epoch": 2737, "lr": 2.5785516408645683e-06}
{"train_loss": 0.03939106687903404, "global_step": 243631, "epoch": 2737, "lr": 2.5783678734979944e-06}
{"train_loss": 0.07629281282424927, "global_step": 243632, "epoch": 2737, "lr": 2.578184112506676e-06}
{"train_loss": 0.03466767445206642, "global_step": 243633, "epoch": 2737, "lr": 2.578000357890642e-06}
{"train_loss": 0.07783879339694977, "global_step": 243634, "epoch": 2737, "lr": 2.577816609649919e-06}
{"train_loss": 0.043635282665491104, "global_step": 243635, "epoch": 2737, "lr": 2.577632867784535e-06}
{"train_loss": 0.047544244676828384, "global_step": 243636, "epoch": 2737, "lr": 2.5774491322945015e-06}
{"train_loss": 0.0651160404086113, "global_step": 243637, "epoch": 2737, "lr": 2.5772654031798516e-06}
{"train_loss": 0.03293126821517944, "global_step": 243638, "epoch": 2737, "lr": 2.5770816804406184e-06}
{"train_loss": 0.05059805139899254, "global_step": 243639, "epoch": 2737, "lr": 2.5768979640768133e-06}
{"train_loss": 0.0547478161752224, "global_step": 243640, "epoch": 2737, "lr": 2.5767142540884693e-06}
{"train_loss": 0.015443992801010609, "global_step": 243641, "epoch": 2737, "lr": 2.5765305504756034e-06}
{"train_loss": 0.051566027104854584, "global_step": 243642, "epoch": 2737, "lr": 2.576346853238254e-06}
{"train_loss": 0.03977012634277344, "global_step": 243643, "epoch": 2737, "lr": 2.576163162376427e-06}
{"train_loss": 0.056218575686216354, "global_step": 243644, "epoch": 2737, "lr": 2.575979477890167e-06}
{"train_loss": 0.034872133284807205, "global_step": 243645, "epoch": 2737, "lr": 2.575795799779479e-06}
{"train_loss": 0.09241046756505966, "global_step": 243646, "epoch": 2737, "lr": 2.5756121280444022e-06}
{"train_loss": 0.03901899233460426, "global_step": 243647, "epoch": 2737, "lr": 2.5754284626849533e-06}
{"train_loss": 0.04526233300566673, "global_step": 243648, "epoch": 2737, "lr": 2.5752448037011658e-06}
{"train_loss": 0.047019559890031815, "global_step": 243649, "epoch": 2737, "lr": 2.57506115109305e-06}
{"train_loss": 0.047305844724178314, "global_step": 243650, "epoch": 2737, "lr": 2.5748775048606454e-06}
{"train_loss": 0.022714082151651382, "global_step": 243651, "epoch": 2737, "lr": 2.5746938650039685e-06}
{"train_loss": 0.05264151468873024, "global_step": 243652, "epoch": 2737, "lr": 2.5745102315230475e-06}
{"train_loss": 0.04022686555981636, "global_step": 243653, "epoch": 2737, "lr": 2.5743266044179036e-06}
{"train_loss": 0.04816347360610962, "global_step": 243654, "epoch": 2737, "lr": 2.5741429836885656e-06}
{"train_loss": 0.052220191806554794, "global_step": 243655, "epoch": 2737, "lr": 2.5739593693350495e-06}
{"train_loss": 0.045840974897146225, "global_step": 243656, "epoch": 2737, "lr": 2.5737757613574e-06}
{"train_loss": 0.046643562614917755, "global_step": 243657, "epoch": 2737, "lr": 2.5735921597556223e-06}
{"train_loss": 0.056246962398290634, "global_step": 243658, "epoch": 2737, "lr": 2.573408564529739e-06}
{"train_loss": 0.04666905850172043, "global_step": 243659, "epoch": 2737, "lr": 2.573224975679794e-06}
{"train_loss": 0.0717957392334938, "global_step": 243660, "epoch": 2737, "lr": 2.573041393205794e-06}
{"train_loss": 0.07236921787261963, "global_step": 243661, "epoch": 2737, "lr": 2.572857817107771e-06}
{"train_loss": 0.027879714965820312, "global_step": 243662, "epoch": 2737, "lr": 2.572674247385748e-06}
{"train_loss": 0.04879399761557579, "global_step": 243663, "epoch": 2737, "lr": 2.5724906840397576e-06}
{"train_loss": 0.03524304926395416, "global_step": 243664, "epoch": 2737, "lr": 2.5723071270698117e-06}
{"train_loss": 0.04228230565786362, "global_step": 243665, "epoch": 2737, "lr": 2.5721235764759435e-06}
{"train_loss": 0.02498442679643631, "global_step": 243666, "epoch": 2737, "lr": 2.5719400322581687e-06}
{"train_loss": 0.10974420607089996, "global_step": 243667, "epoch": 2737, "lr": 2.5717564944165327e-06}
{"train_loss": 0.04394969716668129, "global_step": 243668, "epoch": 2737, "lr": 2.57157296295103e-06}
{"train_loss": 0.02045438066124916, "global_step": 243669, "epoch": 2737, "lr": 2.571389437861715e-06}
{"train_loss": 0.03246086463332176, "global_step": 243670, "epoch": 2737, "lr": 2.5712059191485895e-06}
{"train_loss": 0.06871435046195984, "global_step": 243671, "epoch": 2737, "lr": 2.5710224068116907e-06}
{"train_loss": 0.05798393115401268, "global_step": 243672, "epoch": 2737, "lr": 2.5708389008510357e-06}
{"train_loss": 0.04739592596888542, "global_step": 243673, "epoch": 2737, "lr": 2.570655401266653e-06}
{"train_loss": 0.035158876329660416, "global_step": 243674, "epoch": 2737, "lr": 2.570471908058575e-06}
{"train_loss": 0.031432997435331345, "global_step": 243675, "epoch": 2737, "lr": 2.570288421226813e-06}
{"train_loss": 0.07408566027879715, "global_step": 243676, "epoch": 2737, "lr": 2.5701049407714063e-06}
{"train_loss": 0.04397360980510712, "global_step": 243677, "epoch": 2737, "lr": 2.56992146669236e-06}
{"train_loss": 0.07430266588926315, "global_step": 243678, "epoch": 2737, "lr": 2.5697379989897187e-06}
{"train_loss": 0.03872581571340561, "global_step": 243679, "epoch": 2737, "lr": 2.5695545376634878e-06}
{"train_loss": 0.01616375707089901, "global_step": 243680, "epoch": 2737, "lr": 2.569371082713712e-06}
{"train_loss": 0.04749275884182935, "global_step": 243681, "epoch": 2737, "lr": 2.5691876341404018e-06, "val_loss": 9.529365539550781}
{"train_loss": 0.023971285670995712, "global_step": 243682, "epoch": 2738, "lr": 2.5690041919435858e-06}
{"train_loss": 0.041046518832445145, "global_step": 243683, "epoch": 2738, "lr": 2.568820756123286e-06}
{"train_loss": 0.053666289895772934, "global_step": 243684, "epoch": 2738, "lr": 2.5686373266795406e-06}
{"train_loss": 0.060650479048490524, "global_step": 243685, "epoch": 2738, "lr": 2.5684539036123555e-06}
{"train_loss": 0.05885453149676323, "global_step": 243686, "epoch": 2738, "lr": 2.5682704869217645e-06}
{"train_loss": 0.027282854542136192, "global_step": 243687, "epoch": 2738, "lr": 2.568087076607789e-06}
{"train_loss": 0.015471389517188072, "global_step": 243688, "epoch": 2738, "lr": 2.567903672670463e-06}
{"train_loss": 0.039077553898096085, "global_step": 243689, "epoch": 2738, "lr": 2.5677202751097975e-06}
{"train_loss": 0.03171827644109726, "global_step": 243690, "epoch": 2738, "lr": 2.5675368839258308e-06}
{"train_loss": 0.0815804973244667, "global_step": 243691, "epoch": 2738, "lr": 2.567353499118574e-06}
{"train_loss": 0.03668370097875595, "global_step": 243692, "epoch": 2738, "lr": 2.567170120688067e-06}
{"train_loss": 0.03458164259791374, "global_step": 243693, "epoch": 2738, "lr": 2.5669867486343145e-06}
{"train_loss": 0.04651249572634697, "global_step": 243694, "epoch": 2738, "lr": 2.566803382957361e-06}
{"train_loss": 0.058615248650312424, "global_step": 243695, "epoch": 2738, "lr": 2.566620023657218e-06}
{"train_loss": 0.03920293226838112, "global_step": 243696, "epoch": 2738, "lr": 2.5664366707339184e-06}
{"train_loss": 0.06358414143323898, "global_step": 243697, "epoch": 2738, "lr": 2.56625332418749e-06}
{"train_loss": 0.04153011366724968, "global_step": 243698, "epoch": 2738, "lr": 2.566069984017938e-06}
{"train_loss": 0.03829116374254227, "global_step": 243699, "epoch": 2738, "lr": 2.5658866502253077e-06}
{"train_loss": 0.057716138660907745, "global_step": 243700, "epoch": 2738, "lr": 2.565703322809604e-06}
{"train_loss": 0.036912672221660614, "global_step": 243701, "epoch": 2738, "lr": 2.565520001770877e-06}
{"train_loss": 0.044169384986162186, "global_step": 243702, "epoch": 2738, "lr": 2.5653366871091267e-06}
{"train_loss": 0.05719888582825661, "global_step": 243703, "epoch": 2738, "lr": 2.5651533788243974e-06}
{"train_loss": 0.06768232583999634, "global_step": 243704, "epoch": 2738, "lr": 2.564970076916701e-06}
{"train_loss": 0.08368007093667984, "global_step": 243705, "epoch": 2738, "lr": 2.5647867813860694e-06}
{"train_loss": 0.04183642193675041, "global_step": 243706, "epoch": 2738, "lr": 2.5646034922325202e-06}
{"train_loss": 0.04245247319340706, "global_step": 243707, "epoch": 2738, "lr": 2.5644202094560865e-06}
{"train_loss": 0.02844267338514328, "global_step": 243708, "epoch": 2738, "lr": 2.5642369330567796e-06}
{"train_loss": 0.06388049572706223, "global_step": 243709, "epoch": 2738, "lr": 2.564053663034638e-06}
{"train_loss": 0.03495878726243973, "global_step": 243710, "epoch": 2738, "lr": 2.563870399389684e-06}
{"train_loss": 0.043926190584897995, "global_step": 243711, "epoch": 2738, "lr": 2.5636871421219345e-06}
{"train_loss": 0.01702197827398777, "global_step": 243712, "epoch": 2738, "lr": 2.5635038912314225e-06}
{"train_loss": 0.036676425486803055, "global_step": 243713, "epoch": 2738, "lr": 2.5633206467181647e-06}
{"train_loss": 0.0573655404150486, "global_step": 243714, "epoch": 2738, "lr": 2.5631374085822003e-06}
{"train_loss": 0.04438285529613495, "global_step": 243715, "epoch": 2738, "lr": 2.562954176823534e-06}
{"train_loss": 0.03591858223080635, "global_step": 243716, "epoch": 2738, "lr": 2.562770951442206e-06}
{"train_loss": 0.0668676495552063, "global_step": 243717, "epoch": 2738, "lr": 2.5625877324382318e-06}
{"train_loss": 0.04927702248096466, "global_step": 243718, "epoch": 2738, "lr": 2.5624045198116454e-06}
{"train_loss": 0.03686748072504997, "global_step": 243719, "epoch": 2738, "lr": 2.5622213135624573e-06}
{"train_loss": 0.028754526749253273, "global_step": 243720, "epoch": 2738, "lr": 2.562038113690707e-06}
{"train_loss": 0.08811938017606735, "global_step": 243721, "epoch": 2738, "lr": 2.561854920196405e-06}
{"train_loss": 0.03635866194963455, "global_step": 243722, "epoch": 2738, "lr": 2.5616717330795904e-06}
{"train_loss": 0.04651401564478874, "global_step": 243723, "epoch": 2738, "lr": 2.5614885523402742e-06}
{"train_loss": 0.061545103788375854, "global_step": 243724, "epoch": 2738, "lr": 2.5613053779784957e-06}
{"train_loss": 0.01843569241464138, "global_step": 243725, "epoch": 2738, "lr": 2.5611222099942654e-06}
{"train_loss": 0.01573527231812477, "global_step": 243726, "epoch": 2738, "lr": 2.5609390483876227e-06}
{"train_loss": 0.05831875652074814, "global_step": 243727, "epoch": 2738, "lr": 2.5607558931585727e-06}
{"train_loss": 0.07438977062702179, "global_step": 243728, "epoch": 2738, "lr": 2.5605727443071493e-06}
{"train_loss": 0.05761284381151199, "global_step": 243729, "epoch": 2738, "lr": 2.5603896018333906e-06}
{"train_loss": 0.04252304136753082, "global_step": 243730, "epoch": 2738, "lr": 2.560206465737308e-06}
{"train_loss": 0.07293678820133209, "global_step": 243731, "epoch": 2738, "lr": 2.5600233360189185e-06}
{"train_loss": 0.03339396417140961, "global_step": 243732, "epoch": 2738, "lr": 2.559840212678266e-06}
{"train_loss": 0.0726083517074585, "global_step": 243733, "epoch": 2738, "lr": 2.559657095715351e-06}
{"train_loss": 0.049355048686265945, "global_step": 243734, "epoch": 2738, "lr": 2.559473985130223e-06}
{"train_loss": 0.03430022671818733, "global_step": 243735, "epoch": 2738, "lr": 2.559290880922899e-06}
{"train_loss": 0.08829930424690247, "global_step": 243736, "epoch": 2738, "lr": 2.5591077830933898e-06}
{"train_loss": 0.03860307112336159, "global_step": 243737, "epoch": 2738, "lr": 2.5589246916417398e-06}
{"train_loss": 0.0520600862801075, "global_step": 243738, "epoch": 2738, "lr": 2.5587416065679493e-06}
{"train_loss": 0.04760919138789177, "global_step": 243739, "epoch": 2738, "lr": 2.5585585278720737e-06}
{"train_loss": 0.06421048939228058, "global_step": 243740, "epoch": 2738, "lr": 2.558375455554113e-06}
{"train_loss": 0.009042425081133842, "global_step": 243741, "epoch": 2738, "lr": 2.558192389614106e-06}
{"train_loss": 0.011149774305522442, "global_step": 243742, "epoch": 2738, "lr": 2.5580093300520634e-06}
{"train_loss": 0.03808603808283806, "global_step": 243743, "epoch": 2738, "lr": 2.557826276868025e-06}
{"train_loss": 0.07717768847942352, "global_step": 243744, "epoch": 2738, "lr": 2.5576432300620013e-06}
{"train_loss": 0.033374372869729996, "global_step": 243745, "epoch": 2738, "lr": 2.5574601896340257e-06}
{"train_loss": 0.04824412614107132, "global_step": 243746, "epoch": 2738, "lr": 2.557277155584131e-06}
{"train_loss": 0.04344163462519646, "global_step": 243747, "epoch": 2738, "lr": 2.557094127912324e-06}
{"train_loss": 0.03314794972538948, "global_step": 243748, "epoch": 2738, "lr": 2.5569111066186422e-06}
{"train_loss": 0.03000537119805813, "global_step": 243749, "epoch": 2738, "lr": 2.556728091703098e-06}
{"train_loss": 0.07112118601799011, "global_step": 243750, "epoch": 2738, "lr": 2.5565450831657344e-06}
{"train_loss": 0.060423512011766434, "global_step": 243751, "epoch": 2738, "lr": 2.5563620810065524e-06}
{"train_loss": 0.032236576080322266, "global_step": 243752, "epoch": 2738, "lr": 2.5561790852256017e-06}
{"train_loss": 0.03713017329573631, "global_step": 243753, "epoch": 2738, "lr": 2.5559960958228824e-06}
{"train_loss": 0.010626474395394325, "global_step": 243754, "epoch": 2738, "lr": 2.555813112798444e-06}
{"train_loss": 0.019114553928375244, "global_step": 243755, "epoch": 2738, "lr": 2.555630136152287e-06}
{"train_loss": 0.025226401165127754, "global_step": 243756, "epoch": 2738, "lr": 2.5554471658844557e-06}
{"train_loss": 0.04209275543689728, "global_step": 243757, "epoch": 2738, "lr": 2.5552642019949558e-06}
{"train_loss": 0.02462121844291687, "global_step": 243758, "epoch": 2738, "lr": 2.5550812444838367e-06}
{"train_loss": 0.05075518414378166, "global_step": 243759, "epoch": 2738, "lr": 2.5548982933510934e-06}
{"train_loss": 0.055720243602991104, "global_step": 243760, "epoch": 2738, "lr": 2.5547153485967754e-06}
{"train_loss": 0.04269753023982048, "global_step": 243761, "epoch": 2738, "lr": 2.5545324102208947e-06}
{"train_loss": 0.076014444231987, "global_step": 243762, "epoch": 2738, "lr": 2.5543494782234832e-06}
{"train_loss": 0.05244917422533035, "global_step": 243763, "epoch": 2738, "lr": 2.554166552604553e-06}
{"train_loss": 0.019427893683314323, "global_step": 243764, "epoch": 2738, "lr": 2.5539836333641376e-06}
{"train_loss": 0.030510276556015015, "global_step": 243765, "epoch": 2738, "lr": 2.5538007205022694e-06}
{"train_loss": 0.04084959998726845, "global_step": 243766, "epoch": 2738, "lr": 2.553617814018955e-06}
{"train_loss": 0.046315185725688934, "global_step": 243767, "epoch": 2738, "lr": 2.5534349139142377e-06}
{"train_loss": 0.09674955904483795, "global_step": 243768, "epoch": 2738, "lr": 2.553252020188129e-06}
{"train_loss": 0.061723265796899796, "global_step": 243769, "epoch": 2738, "lr": 2.5530691328406573e-06}
{"train_loss": 0.04588777654584539, "global_step": 243770, "epoch": 2738, "lr": 2.55288625187185e-06, "val_loss": 9.45740795135498}
{"train_loss": 0.028565246611833572, "global_step": 243771, "epoch": 2739, "lr": 2.5527033772817177e-06}
{"train_loss": 0.03940819948911667, "global_step": 243772, "epoch": 2739, "lr": 2.552520509070311e-06}
{"train_loss": 0.053665775805711746, "global_step": 243773, "epoch": 2739, "lr": 2.5523376472376347e-06}
{"train_loss": 0.0434604175388813, "global_step": 243774, "epoch": 2739, "lr": 2.552154791783712e-06}
{"train_loss": 0.06514586508274078, "global_step": 243775, "epoch": 2739, "lr": 2.5519719427085807e-06}
{"train_loss": 0.03898629918694496, "global_step": 243776, "epoch": 2739, "lr": 2.551789100012253e-06}
{"train_loss": 0.07179734110832214, "global_step": 243777, "epoch": 2739, "lr": 2.5516062636947612e-06}
{"train_loss": 0.06624346226453781, "global_step": 243778, "epoch": 2739, "lr": 2.551423433756128e-06}
{"train_loss": 0.03062383271753788, "global_step": 243779, "epoch": 2739, "lr": 2.5512406101963816e-06}
{"train_loss": 0.03075088933110237, "global_step": 243780, "epoch": 2739, "lr": 2.551057793015532e-06}
{"train_loss": 0.03299645334482193, "global_step": 243781, "epoch": 2739, "lr": 2.5508749822136193e-06}
{"train_loss": 0.06213167682290077, "global_step": 243782, "epoch": 2739, "lr": 2.5506921777906645e-06}
{"train_loss": 0.0664682611823082, "global_step": 243783, "epoch": 2739, "lr": 2.5505093797466907e-06}
{"train_loss": 0.02986612357199192, "global_step": 243784, "epoch": 2739, "lr": 2.550326588081725e-06}
{"train_loss": 0.03223326429724693, "global_step": 243785, "epoch": 2739, "lr": 2.5501438027957847e-06}
{"train_loss": 0.0856727585196495, "global_step": 243786, "epoch": 2739, "lr": 2.5499610238889025e-06}
{"train_loss": 0.028017506003379822, "global_step": 243787, "epoch": 2739, "lr": 2.5497782513610958e-06}
{"train_loss": 0.024274148046970367, "global_step": 243788, "epoch": 2739, "lr": 2.5495954852124026e-06}
{"train_loss": 0.03916743025183678, "global_step": 243789, "epoch": 2739, "lr": 2.5494127254428233e-06}
{"train_loss": 0.030325310304760933, "global_step": 243790, "epoch": 2739, "lr": 2.549229972052414e-06}
{"train_loss": 0.04131586477160454, "global_step": 243791, "epoch": 2739, "lr": 2.549047225041168e-06}
{"train_loss": 0.03844596818089485, "global_step": 243792, "epoch": 2739, "lr": 2.5488644844091302e-06}
{"train_loss": 0.10145748406648636, "global_step": 243793, "epoch": 2739, "lr": 2.5486817501563177e-06}
{"train_loss": 0.033401187509298325, "global_step": 243794, "epoch": 2739, "lr": 2.548499022282763e-06}
{"train_loss": 0.036674607545137405, "global_step": 243795, "epoch": 2739, "lr": 2.5483163007884725e-06}
{"train_loss": 0.025319263339042664, "global_step": 243796, "epoch": 2739, "lr": 2.5481335856734958e-06}
{"train_loss": 0.04850451275706291, "global_step": 243797, "epoch": 2739, "lr": 2.5479508769378324e-06}
{"train_loss": 0.05522537976503372, "global_step": 243798, "epoch": 2739, "lr": 2.5477681745815274e-06}
{"train_loss": 0.06544768065214157, "global_step": 243799, "epoch": 2739, "lr": 2.547585478604592e-06}
{"train_loss": 0.043819259852170944, "global_step": 243800, "epoch": 2739, "lr": 2.547402789007053e-06}
{"train_loss": 0.03617045655846596, "global_step": 243801, "epoch": 2739, "lr": 2.5472201057889446e-06}
{"train_loss": 0.04153140261769295, "global_step": 243802, "epoch": 2739, "lr": 2.547037428950272e-06}
{"train_loss": 0.037985894829034805, "global_step": 243803, "epoch": 2739, "lr": 2.5468547584910853e-06}
{"train_loss": 0.04602552950382233, "global_step": 243804, "epoch": 2739, "lr": 2.54667209441139e-06}
{"train_loss": 0.05495920404791832, "global_step": 243805, "epoch": 2739, "lr": 2.546489436711219e-06}
{"train_loss": 0.046767495572566986, "global_step": 243806, "epoch": 2739, "lr": 2.5463067853905953e-06}
{"train_loss": 0.03451620042324066, "global_step": 243807, "epoch": 2739, "lr": 2.546124140449535e-06}
{"train_loss": 0.019678298383951187, "global_step": 243808, "epoch": 2739, "lr": 2.5459415018880773e-06}
{"train_loss": 0.05953104794025421, "global_step": 243809, "epoch": 2739, "lr": 2.545758869706233e-06}
{"train_loss": 0.04353615269064903, "global_step": 243810, "epoch": 2739, "lr": 2.5455762439040354e-06}
{"train_loss": 0.037418946623802185, "global_step": 243811, "epoch": 2739, "lr": 2.545393624481507e-06}
{"train_loss": 0.042711321264505386, "global_step": 243812, "epoch": 2739, "lr": 2.545211011438675e-06}
{"train_loss": 0.06598809361457825, "global_step": 243813, "epoch": 2739, "lr": 2.5450284047755623e-06}
{"train_loss": 0.06199392303824425, "global_step": 243814, "epoch": 2739, "lr": 2.5448458044921853e-06}
{"train_loss": 0.060180697590112686, "global_step": 243815, "epoch": 2739, "lr": 2.544663210588577e-06}
{"train_loss": 0.04079342260956764, "global_step": 243816, "epoch": 2739, "lr": 2.54448062306476e-06}
{"train_loss": 0.03485384210944176, "global_step": 243817, "epoch": 2739, "lr": 2.544298041920756e-06}
{"train_loss": 0.053817786276340485, "global_step": 243818, "epoch": 2739, "lr": 2.5441154671565993e-06}
{"train_loss": 0.05302455648779869, "global_step": 243819, "epoch": 2739, "lr": 2.5439328987723056e-06}
{"train_loss": 0.0684114322066307, "global_step": 243820, "epoch": 2739, "lr": 2.5437503367679027e-06}
{"train_loss": 0.05270518362522125, "global_step": 243821, "epoch": 2739, "lr": 2.5435677811434133e-06}
{"train_loss": 0.09654901921749115, "global_step": 243822, "epoch": 2739, "lr": 2.543385231898865e-06}
{"train_loss": 0.040769804269075394, "global_step": 243823, "epoch": 2739, "lr": 2.5432026890342743e-06}
{"train_loss": 0.067156121134758, "global_step": 243824, "epoch": 2739, "lr": 2.54302015254968e-06}
{"train_loss": 0.01283496618270874, "global_step": 243825, "epoch": 2739, "lr": 2.5428376224450878e-06}
{"train_loss": 0.02259569801390171, "global_step": 243826, "epoch": 2739, "lr": 2.5426550987205423e-06}
{"train_loss": 0.099835105240345, "global_step": 243827, "epoch": 2739, "lr": 2.5424725813760485e-06}
{"train_loss": 0.051216207444667816, "global_step": 243828, "epoch": 2739, "lr": 2.5422900704116516e-06}
{"train_loss": 0.04816586524248123, "global_step": 243829, "epoch": 2739, "lr": 2.5421075658273564e-06}
{"train_loss": 0.02593543566763401, "global_step": 243830, "epoch": 2739, "lr": 2.5419250676232076e-06}
{"train_loss": 0.039419807493686676, "global_step": 243831, "epoch": 2739, "lr": 2.5417425757992054e-06}
{"train_loss": 0.05600493401288986, "global_step": 243832, "epoch": 2739, "lr": 2.541560090355399e-06}
{"train_loss": 0.06543004512786865, "global_step": 243833, "epoch": 2739, "lr": 2.54137761129179e-06}
{"train_loss": 0.07588808983564377, "global_step": 243834, "epoch": 2739, "lr": 2.5411951386084266e-06}
{"train_loss": 0.039611756801605225, "global_step": 243835, "epoch": 2739, "lr": 2.5410126723053098e-06}
{"train_loss": 0.05528712272644043, "global_step": 243836, "epoch": 2739, "lr": 2.540830212382478e-06}
{"train_loss": 0.0697890967130661, "global_step": 243837, "epoch": 2739, "lr": 2.540647758839959e-06}
{"train_loss": 0.04261218011379242, "global_step": 243838, "epoch": 2739, "lr": 2.5404653116777646e-06}
{"train_loss": 0.04359846189618111, "global_step": 243839, "epoch": 2739, "lr": 2.5402828708959325e-06}
{"train_loss": 0.06640983372926712, "global_step": 243840, "epoch": 2739, "lr": 2.540100436494475e-06}
{"train_loss": 0.014662125147879124, "global_step": 243841, "epoch": 2739, "lr": 2.5399180084734297e-06}
{"train_loss": 0.03611711412668228, "global_step": 243842, "epoch": 2739, "lr": 2.5397355868328034e-06}
{"train_loss": 0.046356718987226486, "global_step": 243843, "epoch": 2739, "lr": 2.539553171572645e-06}
{"train_loss": 0.03133583813905716, "global_step": 243844, "epoch": 2739, "lr": 2.53937076269295e-06}
{"train_loss": 0.02393238991498947, "global_step": 243845, "epoch": 2739, "lr": 2.5391883601937727e-06}
{"train_loss": 0.04550787806510925, "global_step": 243846, "epoch": 2739, "lr": 2.5390059640751197e-06}
{"train_loss": 0.02473047375679016, "global_step": 243847, "epoch": 2739, "lr": 2.5388235743370127e-06}
{"train_loss": 0.03735902160406113, "global_step": 243848, "epoch": 2739, "lr": 2.538641190979485e-06}
{"train_loss": 0.053098585456609726, "global_step": 243849, "epoch": 2739, "lr": 2.538458814002559e-06}
{"train_loss": 0.06121904030442238, "global_step": 243850, "epoch": 2739, "lr": 2.538276443406262e-06}
{"train_loss": 0.031026827171444893, "global_step": 243851, "epoch": 2739, "lr": 2.5380940791906118e-06}
{"train_loss": 0.046625252813100815, "global_step": 243852, "epoch": 2739, "lr": 2.5379117213556348e-06}
{"train_loss": 0.03723207488656044, "global_step": 243853, "epoch": 2739, "lr": 2.5377293699013595e-06}
{"train_loss": 0.021809052675962448, "global_step": 243854, "epoch": 2739, "lr": 2.5375470248278076e-06}
{"train_loss": 0.07227413356304169, "global_step": 243855, "epoch": 2739, "lr": 2.5373646861349964e-06}
{"train_loss": 0.0371972881257534, "global_step": 243856, "epoch": 2739, "lr": 2.53718235382297e-06}
{"train_loss": 0.05952008068561554, "global_step": 243857, "epoch": 2739, "lr": 2.537000027891734e-06}
{"train_loss": 0.020704716444015503, "global_step": 243858, "epoch": 2739, "lr": 2.5368177083413215e-06}
{"train_loss": 0.047072542192979475, "global_step": 243859, "epoch": 2739, "lr": 2.5366353951717547e-06, "val_loss": 9.378416061401367}
{"train_loss": 0.013810551725327969, "global_step": 243860, "epoch": 2740, "lr": 2.5364530883830616e-06}
{"train_loss": 0.04850002005696297, "global_step": 243861, "epoch": 2740, "lr": 2.536270787975259e-06}
{"train_loss": 0.024964097887277603, "global_step": 243862, "epoch": 2740, "lr": 2.5360884939483853e-06}
{"train_loss": 0.03998308628797531, "global_step": 243863, "epoch": 2740, "lr": 2.5359062063024462e-06}
{"train_loss": 0.022243762388825417, "global_step": 243864, "epoch": 2740, "lr": 2.5357239250374808e-06}
{"train_loss": 0.06013508141040802, "global_step": 243865, "epoch": 2740, "lr": 2.5355416501535055e-06}
{"train_loss": 0.057932812720537186, "global_step": 243866, "epoch": 2740, "lr": 2.5353593816505537e-06}
{"train_loss": 0.058301009237766266, "global_step": 243867, "epoch": 2740, "lr": 2.5351771195286366e-06}
{"train_loss": 0.05424841493368149, "global_step": 243868, "epoch": 2740, "lr": 2.5349948637877986e-06}
{"train_loss": 0.028522908687591553, "global_step": 243869, "epoch": 2740, "lr": 2.5348126144280393e-06}
{"train_loss": 0.04659329727292061, "global_step": 243870, "epoch": 2740, "lr": 2.5346303714494035e-06}
{"train_loss": 0.03673581779003143, "global_step": 243871, "epoch": 2740, "lr": 2.5344481348519023e-06}
{"train_loss": 0.03997877612709999, "global_step": 243872, "epoch": 2740, "lr": 2.5342659046355686e-06}
{"train_loss": 0.031021056696772575, "global_step": 243873, "epoch": 2740, "lr": 2.5340836808004255e-06}
{"train_loss": 0.009206447750329971, "global_step": 243874, "epoch": 2740, "lr": 2.533901463346494e-06}
{"train_loss": 0.024452244862914085, "global_step": 243875, "epoch": 2740, "lr": 2.5337192522738084e-06}
{"train_loss": 0.03018723428249359, "global_step": 243876, "epoch": 2740, "lr": 2.533537047582374e-06}
{"train_loss": 0.057064611464738846, "global_step": 243877, "epoch": 2740, "lr": 2.533354849272235e-06}
{"train_loss": 0.03132394328713417, "global_step": 243878, "epoch": 2740, "lr": 2.5331726573434024e-06}
{"train_loss": 0.05308530107140541, "global_step": 243879, "epoch": 2740, "lr": 2.5329904717959153e-06}
{"train_loss": 0.020186128094792366, "global_step": 243880, "epoch": 2740, "lr": 2.5328082926297793e-06}
{"train_loss": 0.07558466494083405, "global_step": 243881, "epoch": 2740, "lr": 2.5326261198450387e-06}
{"train_loss": 0.021848974749445915, "global_step": 243882, "epoch": 2740, "lr": 2.5324439534416986e-06}
{"train_loss": 0.03353353589773178, "global_step": 243883, "epoch": 2740, "lr": 2.532261793419799e-06}
{"train_loss": 0.05347427353262901, "global_step": 243884, "epoch": 2740, "lr": 2.532079639779361e-06}
{"train_loss": 0.07011151313781738, "global_step": 243885, "epoch": 2740, "lr": 2.5318974925203964e-06}
{"train_loss": 0.042850472033023834, "global_step": 243886, "epoch": 2740, "lr": 2.5317153516429494e-06}
{"train_loss": 0.02898913249373436, "global_step": 243887, "epoch": 2740, "lr": 2.5315332171470254e-06}
{"train_loss": 0.02543998509645462, "global_step": 243888, "epoch": 2740, "lr": 2.5313510890326632e-06}
{"train_loss": 0.06840462237596512, "global_step": 243889, "epoch": 2740, "lr": 2.531168967299885e-06}
{"train_loss": 0.02607339806854725, "global_step": 243890, "epoch": 2740, "lr": 2.530986851948708e-06}
{"train_loss": 0.07673443853855133, "global_step": 243891, "epoch": 2740, "lr": 2.530804742979159e-06}
{"train_loss": 0.009528596885502338, "global_step": 243892, "epoch": 2740, "lr": 2.5306226403912724e-06}
{"train_loss": 0.045227956026792526, "global_step": 243893, "epoch": 2740, "lr": 2.5304405441850585e-06}
{"train_loss": 0.0316726416349411, "global_step": 243894, "epoch": 2740, "lr": 2.530258454360551e-06}
{"train_loss": 0.07093680649995804, "global_step": 243895, "epoch": 2740, "lr": 2.5300763709177667e-06}
{"train_loss": 0.07002010941505432, "global_step": 243896, "epoch": 2740, "lr": 2.529894293856744e-06}
{"train_loss": 0.06479718536138535, "global_step": 243897, "epoch": 2740, "lr": 2.5297122231774937e-06}
{"train_loss": 0.01871776208281517, "global_step": 243898, "epoch": 2740, "lr": 2.529530158880045e-06}
{"train_loss": 0.06271816790103912, "global_step": 243899, "epoch": 2740, "lr": 2.5293481009644182e-06}
{"train_loss": 0.02028627321124077, "global_step": 243900, "epoch": 2740, "lr": 2.5291660494306534e-06}
{"train_loss": 0.054299257695674896, "global_step": 243901, "epoch": 2740, "lr": 2.5289840042787503e-06}
{"train_loss": 0.04755876958370209, "global_step": 243902, "epoch": 2740, "lr": 2.528801965508759e-06}
{"train_loss": 0.021112030372023582, "global_step": 243903, "epoch": 2740, "lr": 2.5286199331206793e-06}
{"train_loss": 0.0403556153178215, "global_step": 243904, "epoch": 2740, "lr": 2.5284379071145614e-06}
{"train_loss": 0.04671473801136017, "global_step": 243905, "epoch": 2740, "lr": 2.528255887490405e-06}
{"train_loss": 0.07017658650875092, "global_step": 243906, "epoch": 2740, "lr": 2.5280738742482546e-06}
{"train_loss": 0.04295315966010094, "global_step": 243907, "epoch": 2740, "lr": 2.527891867388116e-06}
{"train_loss": 0.059030648320913315, "global_step": 243908, "epoch": 2740, "lr": 2.5277098669100276e-06}
{"train_loss": 0.049401041120290756, "global_step": 243909, "epoch": 2740, "lr": 2.527527872814017e-06}
{"train_loss": 0.11356081813573837, "global_step": 243910, "epoch": 2740, "lr": 2.5273458851000907e-06}
{"train_loss": 0.04524589702486992, "global_step": 243911, "epoch": 2740, "lr": 2.527163903768298e-06}
{"train_loss": 0.03378734365105629, "global_step": 243912, "epoch": 2740, "lr": 2.5269819288186334e-06}
{"train_loss": 0.03601088747382164, "global_step": 243913, "epoch": 2740, "lr": 2.526799960251153e-06}
{"train_loss": 0.06195763126015663, "global_step": 243914, "epoch": 2740, "lr": 2.5266179980658555e-06}
{"train_loss": 0.04252369701862335, "global_step": 243915, "epoch": 2740, "lr": 2.5264360422627807e-06}
{"train_loss": 0.024465642869472504, "global_step": 243916, "epoch": 2740, "lr": 2.5262540928419454e-06}
{"train_loss": 0.06619232147932053, "global_step": 243917, "epoch": 2740, "lr": 2.526072149803377e-06}
{"train_loss": 0.034440916031599045, "global_step": 243918, "epoch": 2740, "lr": 2.5258902131470972e-06}
{"train_loss": 0.08824292570352554, "global_step": 243919, "epoch": 2740, "lr": 2.5257082828731406e-06}
{"train_loss": 0.060911625623703, "global_step": 243920, "epoch": 2740, "lr": 2.5255263589815116e-06}
{"train_loss": 0.016513388603925705, "global_step": 243921, "epoch": 2740, "lr": 2.5253444414722606e-06}
{"train_loss": 0.02412223629653454, "global_step": 243922, "epoch": 2740, "lr": 2.5251625303453872e-06}
{"train_loss": 0.05693822726607323, "global_step": 243923, "epoch": 2740, "lr": 2.5249806256009367e-06}
{"train_loss": 0.03667473793029785, "global_step": 243924, "epoch": 2740, "lr": 2.5247987272389197e-06}
{"train_loss": 0.10313184559345245, "global_step": 243925, "epoch": 2740, "lr": 2.5246168352593637e-06}
{"train_loss": 0.047951649874448776, "global_step": 243926, "epoch": 2740, "lr": 2.5244349496622967e-06}
{"train_loss": 0.05124617740511894, "global_step": 243927, "epoch": 2740, "lr": 2.524253070447735e-06}
{"train_loss": 0.0748782604932785, "global_step": 243928, "epoch": 2740, "lr": 2.5240711976157126e-06}
{"train_loss": 0.03270256891846657, "global_step": 243929, "epoch": 2740, "lr": 2.5238893311662514e-06}
{"train_loss": 0.04286952689290047, "global_step": 243930, "epoch": 2740, "lr": 2.5237074710993737e-06}
{"train_loss": 0.03357063606381416, "global_step": 243931, "epoch": 2740, "lr": 2.5235256174151013e-06}
{"train_loss": 0.04325093701481819, "global_step": 243932, "epoch": 2740, "lr": 2.5233437701134675e-06}
{"train_loss": 0.060786835849285126, "global_step": 243933, "epoch": 2740, "lr": 2.52316192919449e-06}
{"train_loss": 0.0474611297249794, "global_step": 243934, "epoch": 2740, "lr": 2.522980094658195e-06}
{"train_loss": 0.038462650030851364, "global_step": 243935, "epoch": 2740, "lr": 2.5227982665046056e-06}
{"train_loss": 0.051349200308322906, "global_step": 243936, "epoch": 2740, "lr": 2.5226164447337496e-06}
{"train_loss": 0.07114281505346298, "global_step": 243937, "epoch": 2740, "lr": 2.5224346293456436e-06}
{"train_loss": 0.029305199161171913, "global_step": 243938, "epoch": 2740, "lr": 2.5222528203403263e-06}
{"train_loss": 0.049527160823345184, "global_step": 243939, "epoch": 2740, "lr": 2.522071017717803e-06}
{"train_loss": 0.02992965653538704, "global_step": 243940, "epoch": 2740, "lr": 2.521889221478119e-06}
{"train_loss": 0.05122014880180359, "global_step": 243941, "epoch": 2740, "lr": 2.5217074316212786e-06}
{"train_loss": 0.010669201612472534, "global_step": 243942, "epoch": 2740, "lr": 2.5215256481473217e-06}
{"train_loss": 0.1034502238035202, "global_step": 243943, "epoch": 2740, "lr": 2.521343871056264e-06}
{"train_loss": 0.04450516402721405, "global_step": 243944, "epoch": 2740, "lr": 2.521162100348129e-06}
{"train_loss": 0.04764765501022339, "global_step": 243945, "epoch": 2740, "lr": 2.5209803360229545e-06}
{"train_loss": 0.06206592917442322, "global_step": 243946, "epoch": 2740, "lr": 2.520798578080746e-06}
{"train_loss": 0.04755769670009613, "global_step": 243947, "epoch": 2740, "lr": 2.520616826521549e-06}
{"train_loss": 0.045789701801337554, "global_step": 243948, "epoch": 2740, "lr": 2.520435081345368e-06, "val_loss": 9.462430000305176, "train_action_mse_error": 15.639691352844238}
{"train_loss": 0.05540752410888672, "global_step": 243949, "epoch": 2741, "lr": 2.5202533425522424e-06}
{"train_loss": 0.06400644034147263, "global_step": 243950, "epoch": 2741, "lr": 2.520071610142183e-06}
{"train_loss": 0.03385883942246437, "global_step": 243951, "epoch": 2741, "lr": 2.5198898841152286e-06}
{"train_loss": 0.02283681556582451, "global_step": 243952, "epoch": 2741, "lr": 2.519708164471385e-06}
{"train_loss": 0.06629504263401031, "global_step": 243953, "epoch": 2741, "lr": 2.519526451210702e-06}
{"train_loss": 0.0338541641831398, "global_step": 243954, "epoch": 2741, "lr": 2.51934474433318e-06}
{"train_loss": 0.0742652416229248, "global_step": 243955, "epoch": 2741, "lr": 2.519163043838857e-06}
{"train_loss": 0.054421015083789825, "global_step": 243956, "epoch": 2741, "lr": 2.5189813497277503e-06}
{"train_loss": 0.017608866095542908, "global_step": 243957, "epoch": 2741, "lr": 2.5187996619998935e-06}
{"train_loss": 0.032244324684143066, "global_step": 243958, "epoch": 2741, "lr": 2.5186179806552968e-06}
{"train_loss": 0.05428135395050049, "global_step": 243959, "epoch": 2741, "lr": 2.518436305694005e-06}
{"train_loss": 0.031754858791828156, "global_step": 243960, "epoch": 2741, "lr": 2.5182546371160187e-06}
{"train_loss": 0.03789105266332626, "global_step": 243961, "epoch": 2741, "lr": 2.5180729749213873e-06}
{"train_loss": 0.04730323702096939, "global_step": 243962, "epoch": 2741, "lr": 2.517891319110116e-06}
{"train_loss": 0.03438744693994522, "global_step": 243963, "epoch": 2741, "lr": 2.5177096696822332e-06}
{"train_loss": 0.028653603047132492, "global_step": 243964, "epoch": 2741, "lr": 2.517528026637772e-06}
{"train_loss": 0.11574558168649673, "global_step": 243965, "epoch": 2741, "lr": 2.517346389976738e-06}
{"train_loss": 0.03609493374824524, "global_step": 243966, "epoch": 2741, "lr": 2.5171647596991808e-06}
{"train_loss": 0.024768248200416565, "global_step": 243967, "epoch": 2741, "lr": 2.516983135805101e-06}
{"train_loss": 0.020662197843194008, "global_step": 243968, "epoch": 2741, "lr": 2.516801518294548e-06}
{"train_loss": 0.06580882519483566, "global_step": 243969, "epoch": 2741, "lr": 2.5166199071675165e-06}
{"train_loss": 0.042624954134225845, "global_step": 243970, "epoch": 2741, "lr": 2.5164383024240624e-06}
{"train_loss": 0.09803661704063416, "global_step": 243971, "epoch": 2741, "lr": 2.5162567040641793e-06}
{"train_loss": 0.03981255739927292, "global_step": 243972, "epoch": 2741, "lr": 2.5160751120879176e-06}
{"train_loss": 0.020789824426174164, "global_step": 243973, "epoch": 2741, "lr": 2.5158935264952834e-06}
{"train_loss": 0.06141204759478569, "global_step": 243974, "epoch": 2741, "lr": 2.5157119472863142e-06}
{"train_loss": 0.026237014681100845, "global_step": 243975, "epoch": 2741, "lr": 2.515530374461028e-06}
{"train_loss": 0.028332969173789024, "global_step": 243976, "epoch": 2741, "lr": 2.515348808019452e-06}
{"train_loss": 0.029111873358488083, "global_step": 243977, "epoch": 2741, "lr": 2.515167247961603e-06}
{"train_loss": 0.043487027287483215, "global_step": 243978, "epoch": 2741, "lr": 2.514985694287514e-06}
{"train_loss": 0.044000912457704544, "global_step": 243979, "epoch": 2741, "lr": 2.5148041469972018e-06}
{"train_loss": 0.06416688114404678, "global_step": 243980, "epoch": 2741, "lr": 2.514622606090705e-06}
{"train_loss": 0.025523681193590164, "global_step": 243981, "epoch": 2741, "lr": 2.51444107156803e-06}
{"train_loss": 0.05117405205965042, "global_step": 243982, "epoch": 2741, "lr": 2.5142595434292095e-06}
{"train_loss": 0.043958891183137894, "global_step": 243983, "epoch": 2741, "lr": 2.514078021674271e-06}
{"train_loss": 0.03948148712515831, "global_step": 243984, "epoch": 2741, "lr": 2.513896506303237e-06}
{"train_loss": 0.02064097672700882, "global_step": 243985, "epoch": 2741, "lr": 2.51371499731613e-06}
{"train_loss": 0.017278047278523445, "global_step": 243986, "epoch": 2741, "lr": 2.5135334947129717e-06}
{"train_loss": 0.062345653772354126, "global_step": 243987, "epoch": 2741, "lr": 2.5133519984937957e-06}
{"train_loss": 0.028988948091864586, "global_step": 243988, "epoch": 2741, "lr": 2.5131705086586186e-06}
{"train_loss": 0.06809579581022263, "global_step": 243989, "epoch": 2741, "lr": 2.5129890252074682e-06}
{"train_loss": 0.02213911898434162, "global_step": 243990, "epoch": 2741, "lr": 2.512807548140361e-06}
{"train_loss": 0.06297637522220612, "global_step": 243991, "epoch": 2741, "lr": 2.512626077457342e-06}
{"train_loss": 0.05206989124417305, "global_step": 243992, "epoch": 2741, "lr": 2.5124446131584047e-06}
{"train_loss": 0.045570679008960724, "global_step": 243993, "epoch": 2741, "lr": 2.512263155243605e-06}
{"train_loss": 0.047295521944761276, "global_step": 243994, "epoch": 2741, "lr": 2.512081703712943e-06}
{"train_loss": 0.055566925555467606, "global_step": 243995, "epoch": 2741, "lr": 2.5119002585664574e-06}
{"train_loss": 0.06582023203372955, "global_step": 243996, "epoch": 2741, "lr": 2.5117188198041652e-06}
{"train_loss": 0.046067975461483, "global_step": 243997, "epoch": 2741, "lr": 2.5115373874260993e-06}
{"train_loss": 0.04553668200969696, "global_step": 243998, "epoch": 2741, "lr": 2.5113559614322714e-06}
{"train_loss": 0.03310415893793106, "global_step": 243999, "epoch": 2741, "lr": 2.5111745418227194e-06}
{"train_loss": 0.037926074117422104, "global_step": 244000, "epoch": 2741, "lr": 2.510993128597461e-06}
{"train_loss": 0.05075779929757118, "global_step": 244001, "epoch": 2741, "lr": 2.510811721756512e-06}
{"train_loss": 0.027250347658991814, "global_step": 244002, "epoch": 2741, "lr": 2.5106303212999173e-06}
{"train_loss": 0.02752542681992054, "global_step": 244003, "epoch": 2741, "lr": 2.5104489272276764e-06}
{"train_loss": 0.05103949084877968, "global_step": 244004, "epoch": 2741, "lr": 2.51026753953984e-06}
{"train_loss": 0.05980762839317322, "global_step": 244005, "epoch": 2741, "lr": 2.5100861582364076e-06}
{"train_loss": 0.03474687039852142, "global_step": 244006, "epoch": 2741, "lr": 2.509904783317424e-06}
{"train_loss": 0.03252570331096649, "global_step": 244007, "epoch": 2741, "lr": 2.5097234147828997e-06}
{"train_loss": 0.049288492649793625, "global_step": 244008, "epoch": 2741, "lr": 2.5095420526328682e-06}
{"train_loss": 0.03710610046982765, "global_step": 244009, "epoch": 2741, "lr": 2.5093606968673467e-06}
{"train_loss": 0.043959762901067734, "global_step": 244010, "epoch": 2741, "lr": 2.5091793474863623e-06}
{"train_loss": 0.030560772866010666, "global_step": 244011, "epoch": 2741, "lr": 2.508998004489943e-06}
{"train_loss": 0.1059875637292862, "global_step": 244012, "epoch": 2741, "lr": 2.508816667878111e-06}
{"train_loss": 0.02089621312916279, "global_step": 244013, "epoch": 2741, "lr": 2.508635337650883e-06}
{"train_loss": 0.03626643493771553, "global_step": 244014, "epoch": 2741, "lr": 2.508454013808298e-06}
{"train_loss": 0.01739930361509323, "global_step": 244015, "epoch": 2741, "lr": 2.5082726963503665e-06}
{"train_loss": 0.03236493840813637, "global_step": 244016, "epoch": 2741, "lr": 2.5080913852771227e-06}
{"train_loss": 0.07333408296108246, "global_step": 244017, "epoch": 2741, "lr": 2.5079100805885823e-06}
{"train_loss": 0.06982419639825821, "global_step": 244018, "epoch": 2741, "lr": 2.507728782284774e-06}
{"train_loss": 0.02022348716855049, "global_step": 244019, "epoch": 2741, "lr": 2.5075474903657304e-06}
{"train_loss": 0.023933149874210358, "global_step": 244020, "epoch": 2741, "lr": 2.507366204831463e-06}
{"train_loss": 0.04619625583291054, "global_step": 244021, "epoch": 2741, "lr": 2.5071849256820044e-06}
{"train_loss": 0.05634009838104248, "global_step": 244022, "epoch": 2741, "lr": 2.5070036529173723e-06}
{"train_loss": 0.06344731152057648, "global_step": 244023, "epoch": 2741, "lr": 2.5068223865375995e-06}
{"train_loss": 0.05692869797348976, "global_step": 244024, "epoch": 2741, "lr": 2.5066411265427027e-06}
{"train_loss": 0.031490013003349304, "global_step": 244025, "epoch": 2741, "lr": 2.5064598729327092e-06}
{"train_loss": 0.073396697640419, "global_step": 244026, "epoch": 2741, "lr": 2.506278625707642e-06}
{"train_loss": 0.07385194301605225, "global_step": 244027, "epoch": 2741, "lr": 2.506097384867534e-06}
{"train_loss": 0.023432349786162376, "global_step": 244028, "epoch": 2741, "lr": 2.5059161504123906e-06}
{"train_loss": 0.05430719256401062, "global_step": 244029, "epoch": 2741, "lr": 2.5057349223422566e-06}
{"train_loss": 0.06050840765237808, "global_step": 244030, "epoch": 2741, "lr": 2.505553700657143e-06}
{"train_loss": 0.07287369668483734, "global_step": 244031, "epoch": 2741, "lr": 2.5053724853570826e-06}
{"train_loss": 0.05043511465191841, "global_step": 244032, "epoch": 2741, "lr": 2.5051912764420926e-06}
{"train_loss": 0.0342743918299675, "global_step": 244033, "epoch": 2741, "lr": 2.5050100739122064e-06}
{"train_loss": 0.022843455895781517, "global_step": 244034, "epoch": 2741, "lr": 2.5048288777674345e-06}
{"train_loss": 0.05459848791360855, "global_step": 244035, "epoch": 2741, "lr": 2.5046476880078163e-06}
{"train_loss": 0.07511203736066818, "global_step": 244036, "epoch": 2741, "lr": 2.5044665046333625e-06}
{"train_loss": 0.04547872216430273, "global_step": 244037, "epoch": 2741, "lr": 2.5042853276441124e-06, "val_loss": 9.512957572937012}
{"train_loss": 0.04315276816487312, "global_step": 244038, "epoch": 2742, "lr": 2.504104157040077e-06}
{"train_loss": 0.027677791193127632, "global_step": 244039, "epoch": 2742, "lr": 2.5039229928212893e-06}
{"train_loss": 0.020262697711586952, "global_step": 244040, "epoch": 2742, "lr": 2.503741834987772e-06}
{"train_loss": 0.03991386294364929, "global_step": 244041, "epoch": 2742, "lr": 2.5035606835395408e-06}
{"train_loss": 0.045841000974178314, "global_step": 244042, "epoch": 2742, "lr": 2.5033795384766302e-06}
{"train_loss": 0.04433869943022728, "global_step": 244043, "epoch": 2742, "lr": 2.5031983997990615e-06}
{"train_loss": 0.07024330645799637, "global_step": 244044, "epoch": 2742, "lr": 2.503017267506863e-06}
{"train_loss": 0.05079588294029236, "global_step": 244045, "epoch": 2742, "lr": 2.5028361416000455e-06}
{"train_loss": 0.053928833454847336, "global_step": 244046, "epoch": 2742, "lr": 2.502655022078648e-06}
{"train_loss": 0.05529193952679634, "global_step": 244047, "epoch": 2742, "lr": 2.5024739089426875e-06}
{"train_loss": 0.06655983626842499, "global_step": 244048, "epoch": 2742, "lr": 2.5022928021921965e-06}
{"train_loss": 0.04774356633424759, "global_step": 244049, "epoch": 2742, "lr": 2.5021117018271867e-06}
{"train_loss": 0.025904854759573936, "global_step": 244050, "epoch": 2742, "lr": 2.501930607847697e-06}
{"train_loss": 0.08740384876728058, "global_step": 244051, "epoch": 2742, "lr": 2.5017495202537323e-06}
{"train_loss": 0.02765429951250553, "global_step": 244052, "epoch": 2742, "lr": 2.5015684390453377e-06}
{"train_loss": 0.04611457884311676, "global_step": 244053, "epoch": 2742, "lr": 2.501387364222524e-06}
{"train_loss": 0.02429124154150486, "global_step": 244054, "epoch": 2742, "lr": 2.5012062957853134e-06}
{"train_loss": 0.07278304547071457, "global_step": 244055, "epoch": 2742, "lr": 2.50102523373375e-06}
{"train_loss": 0.06998706609010696, "global_step": 244056, "epoch": 2742, "lr": 2.500844178067835e-06}
{"train_loss": 0.05771041288971901, "global_step": 244057, "epoch": 2742, "lr": 2.5006631287876116e-06}
{"train_loss": 0.04656556621193886, "global_step": 244058, "epoch": 2742, "lr": 2.5004820858930854e-06}
{"train_loss": 0.0695236399769783, "global_step": 244059, "epoch": 2742, "lr": 2.5003010493842958e-06}
{"train_loss": 0.04479001834988594, "global_step": 244060, "epoch": 2742, "lr": 2.5001200192612596e-06}
{"train_loss": 0.05495132505893707, "global_step": 244061, "epoch": 2742, "lr": 2.4999389955240094e-06}
{"train_loss": 0.08185415714979172, "global_step": 244062, "epoch": 2742, "lr": 2.4997579781725567e-06}
{"train_loss": 0.06733068823814392, "global_step": 244063, "epoch": 2742, "lr": 2.4995769672069345e-06}
{"train_loss": 0.024711722508072853, "global_step": 244064, "epoch": 2742, "lr": 2.4993959626271658e-06}
{"train_loss": 0.035982996225357056, "global_step": 244065, "epoch": 2742, "lr": 2.4992149644332775e-06}
{"train_loss": 0.011417640373110771, "global_step": 244066, "epoch": 2742, "lr": 2.499033972625281e-06}
{"train_loss": 0.05430826172232628, "global_step": 244067, "epoch": 2742, "lr": 2.4988529872032207e-06}
{"train_loss": 0.05480935797095299, "global_step": 244068, "epoch": 2742, "lr": 2.4986720081671022e-06}
{"train_loss": 0.039171937853097916, "global_step": 244069, "epoch": 2742, "lr": 2.49849103551697e-06}
{"train_loss": 0.046529076993465424, "global_step": 244070, "epoch": 2742, "lr": 2.498310069252824e-06}
{"train_loss": 0.03893324360251427, "global_step": 244071, "epoch": 2742, "lr": 2.4981291093747027e-06}
{"train_loss": 0.06894747912883759, "global_step": 244072, "epoch": 2742, "lr": 2.4979481558826402e-06}
{"train_loss": 0.02445683255791664, "global_step": 244073, "epoch": 2742, "lr": 2.4977672087766412e-06}
{"train_loss": 0.057513363659381866, "global_step": 244074, "epoch": 2742, "lr": 2.4975862680567395e-06}
{"train_loss": 0.041088562458753586, "global_step": 244075, "epoch": 2742, "lr": 2.497405333722963e-06}
{"train_loss": 0.025018712505698204, "global_step": 244076, "epoch": 2742, "lr": 2.4972244057753226e-06}
{"train_loss": 0.03705751895904541, "global_step": 244077, "epoch": 2742, "lr": 2.497043484213857e-06}
{"train_loss": 0.038593634963035583, "global_step": 244078, "epoch": 2742, "lr": 2.496862569038588e-06}
{"train_loss": 0.042039256542921066, "global_step": 244079, "epoch": 2742, "lr": 2.496681660249528e-06}
{"train_loss": 0.037977613508701324, "global_step": 244080, "epoch": 2742, "lr": 2.4965007578467202e-06}
{"train_loss": 0.0381915308535099, "global_step": 244081, "epoch": 2742, "lr": 2.4963198618301707e-06}
{"train_loss": 0.037356071174144745, "global_step": 244082, "epoch": 2742, "lr": 2.4961389721999184e-06}
{"train_loss": 0.04231030493974686, "global_step": 244083, "epoch": 2742, "lr": 2.495958088955974e-06}
{"train_loss": 0.030917910858988762, "global_step": 244084, "epoch": 2742, "lr": 2.4957772120983768e-06}
{"train_loss": 0.030661631375551224, "global_step": 244085, "epoch": 2742, "lr": 2.4955963416271322e-06}
{"train_loss": 0.031126772984862328, "global_step": 244086, "epoch": 2742, "lr": 2.4954154775422845e-06}
{"train_loss": 0.013083255849778652, "global_step": 244087, "epoch": 2742, "lr": 2.495234619843845e-06}
{"train_loss": 0.02842622622847557, "global_step": 244088, "epoch": 2742, "lr": 2.4950537685318465e-06}
{"train_loss": 0.03827817365527153, "global_step": 244089, "epoch": 2742, "lr": 2.4948729236063064e-06}
{"train_loss": 0.04705139994621277, "global_step": 244090, "epoch": 2742, "lr": 2.4946920850672464e-06}
{"train_loss": 0.06300432980060577, "global_step": 244091, "epoch": 2742, "lr": 2.494511252914705e-06}
{"train_loss": 0.061144862323999405, "global_step": 244092, "epoch": 2742, "lr": 2.4943304271486944e-06}
{"train_loss": 0.07692685723304749, "global_step": 244093, "epoch": 2742, "lr": 2.4941496077692474e-06}
{"train_loss": 0.06776809692382812, "global_step": 244094, "epoch": 2742, "lr": 2.493968794776369e-06}
{"train_loss": 0.011680861003696918, "global_step": 244095, "epoch": 2742, "lr": 2.49378798817011e-06}
{"train_loss": 0.05644119158387184, "global_step": 244096, "epoch": 2742, "lr": 2.4936071879504752e-06}
{"train_loss": 0.05733434855937958, "global_step": 244097, "epoch": 2742, "lr": 2.493426394117504e-06}
{"train_loss": 0.026262208819389343, "global_step": 244098, "epoch": 2742, "lr": 2.4932456066712018e-06}
{"train_loss": 0.0660485327243805, "global_step": 244099, "epoch": 2742, "lr": 2.493064825611613e-06}
{"train_loss": 0.06233007460832596, "global_step": 244100, "epoch": 2742, "lr": 2.4928840509387487e-06}
{"train_loss": 0.03831116482615471, "global_step": 244101, "epoch": 2742, "lr": 2.492703282652642e-06}
{"train_loss": 0.07827810198068619, "global_step": 244102, "epoch": 2742, "lr": 2.4925225207533043e-06}
{"train_loss": 0.05502469465136528, "global_step": 244103, "epoch": 2742, "lr": 2.4923417652407742e-06}
{"train_loss": 0.005828470923006535, "global_step": 244104, "epoch": 2742, "lr": 2.492161016115063e-06}
{"train_loss": 0.06183420866727829, "global_step": 244105, "epoch": 2742, "lr": 2.49198027337621e-06}
{"train_loss": 0.03463254123926163, "global_step": 244106, "epoch": 2742, "lr": 2.4917995370242196e-06}
{"train_loss": 0.04455556720495224, "global_step": 244107, "epoch": 2742, "lr": 2.491618807059143e-06}
{"train_loss": 0.042285896837711334, "global_step": 244108, "epoch": 2742, "lr": 2.4914380834809734e-06}
{"train_loss": 0.10204481333494186, "global_step": 244109, "epoch": 2742, "lr": 2.4912573662897563e-06}
{"train_loss": 0.026064850389957428, "global_step": 244110, "epoch": 2742, "lr": 2.491076655485519e-06}
{"train_loss": 0.07303376495838165, "global_step": 244111, "epoch": 2742, "lr": 2.490895951068273e-06}
{"train_loss": 0.040773905813694, "global_step": 244112, "epoch": 2742, "lr": 2.4907152530380452e-06}
{"train_loss": 0.039588332176208496, "global_step": 244113, "epoch": 2742, "lr": 2.490534561394864e-06}
{"train_loss": 0.0507747083902359, "global_step": 244114, "epoch": 2742, "lr": 2.4903538761387456e-06}
{"train_loss": 0.029524123296141624, "global_step": 244115, "epoch": 2742, "lr": 2.490173197269724e-06}
{"train_loss": 0.022613506764173508, "global_step": 244116, "epoch": 2742, "lr": 2.4899925247878265e-06}
{"train_loss": 0.050297852605581284, "global_step": 244117, "epoch": 2742, "lr": 2.4898118586930583e-06}
{"train_loss": 0.07921241968870163, "global_step": 244118, "epoch": 2742, "lr": 2.489631198985465e-06}
{"train_loss": 0.06230079382658005, "global_step": 244119, "epoch": 2742, "lr": 2.4894505456650506e-06}
{"train_loss": 0.063105508685112, "global_step": 244120, "epoch": 2742, "lr": 2.489269898731861e-06}
{"train_loss": 0.08653394877910614, "global_step": 244121, "epoch": 2742, "lr": 2.489089258185906e-06}
{"train_loss": 0.07113814353942871, "global_step": 244122, "epoch": 2742, "lr": 2.4889086240272143e-06}
{"train_loss": 0.06710225343704224, "global_step": 244123, "epoch": 2742, "lr": 2.4887279962558072e-06}
{"train_loss": 0.09165144711732864, "global_step": 244124, "epoch": 2742, "lr": 2.4885473748717194e-06}
{"train_loss": 0.0495435930788517, "global_step": 244125, "epoch": 2742, "lr": 2.488366759874955e-06}
{"train_loss": 0.04862442561372947, "global_step": 244126, "epoch": 2742, "lr": 2.4881861512655537e-06, "val_loss": 9.519043922424316}
{"train_loss": 0.04762803018093109, "global_step": 244127, "epoch": 2743, "lr": 2.4880055490435483e-06}
{"train_loss": 0.034588489681482315, "global_step": 244128, "epoch": 2743, "lr": 2.4878249532089392e-06}
{"train_loss": 0.06693185865879059, "global_step": 244129, "epoch": 2743, "lr": 2.487644363761771e-06}
{"train_loss": 0.03249574452638626, "global_step": 244130, "epoch": 2743, "lr": 2.487463780702054e-06}
{"train_loss": 0.07396965473890305, "global_step": 244131, "epoch": 2743, "lr": 2.4872832040298223e-06}
{"train_loss": 0.06053885817527771, "global_step": 244132, "epoch": 2743, "lr": 2.487102633745092e-06}
{"train_loss": 0.05615953356027603, "global_step": 244133, "epoch": 2743, "lr": 2.4869220698478967e-06}
{"train_loss": 0.022465026006102562, "global_step": 244134, "epoch": 2743, "lr": 2.4867415123382474e-06}
{"train_loss": 0.05247843638062477, "global_step": 244135, "epoch": 2743, "lr": 2.486560961216189e-06}
{"train_loss": 0.02070240117609501, "global_step": 244136, "epoch": 2743, "lr": 2.4863804164817205e-06}
{"train_loss": 0.05781688913702965, "global_step": 244137, "epoch": 2743, "lr": 2.4861998781348927e-06}
{"train_loss": 0.04423515126109123, "global_step": 244138, "epoch": 2743, "lr": 2.486019346175705e-06}
{"train_loss": 0.021469654515385628, "global_step": 244139, "epoch": 2743, "lr": 2.4858388206041973e-06}
{"train_loss": 0.03928394243121147, "global_step": 244140, "epoch": 2743, "lr": 2.485658301420385e-06}
{"train_loss": 0.041612278670072556, "global_step": 244141, "epoch": 2743, "lr": 2.4854777886243074e-06}
{"train_loss": 0.013917607255280018, "global_step": 244142, "epoch": 2743, "lr": 2.485297282215965e-06}
{"train_loss": 0.033608417958021164, "global_step": 244143, "epoch": 2743, "lr": 2.485116782195407e-06}
{"train_loss": 0.026230765506625175, "global_step": 244144, "epoch": 2743, "lr": 2.484936288562639e-06}
{"train_loss": 0.07152718305587769, "global_step": 244145, "epoch": 2743, "lr": 2.48475580131769e-06}
{"train_loss": 0.03909195959568024, "global_step": 244146, "epoch": 2743, "lr": 2.4845753204605917e-06}
{"train_loss": 0.017035821452736855, "global_step": 244147, "epoch": 2743, "lr": 2.484394845991361e-06}
{"train_loss": 0.048133689910173416, "global_step": 244148, "epoch": 2743, "lr": 2.484214377910027e-06}
{"train_loss": 0.08117624372243881, "global_step": 244149, "epoch": 2743, "lr": 2.4840339162166105e-06}
{"train_loss": 0.060841936618089676, "global_step": 244150, "epoch": 2743, "lr": 2.483853460911134e-06}
{"train_loss": 0.035316940397024155, "global_step": 244151, "epoch": 2743, "lr": 2.483673011993631e-06}
{"train_loss": 0.026625994592905045, "global_step": 244152, "epoch": 2743, "lr": 2.4834925694641076e-06}
{"train_loss": 0.011588371358811855, "global_step": 244153, "epoch": 2743, "lr": 2.483312133322607e-06}
{"train_loss": 0.08533816784620285, "global_step": 244154, "epoch": 2743, "lr": 2.4831317035691417e-06}
{"train_loss": 0.0384562648832798, "global_step": 244155, "epoch": 2743, "lr": 2.482951280203749e-06}
{"train_loss": 0.06385942548513412, "global_step": 244156, "epoch": 2743, "lr": 2.482770863226441e-06}
{"train_loss": 0.02699730172753334, "global_step": 244157, "epoch": 2743, "lr": 2.48259045263724e-06}
{"train_loss": 0.0654064193367958, "global_step": 244158, "epoch": 2743, "lr": 2.482410048436179e-06}
{"train_loss": 0.04858418554067612, "global_step": 244159, "epoch": 2743, "lr": 2.4822296506232747e-06}
{"train_loss": 0.06930240988731384, "global_step": 244160, "epoch": 2743, "lr": 2.48204925919856e-06}
{"train_loss": 0.04560545086860657, "global_step": 244161, "epoch": 2743, "lr": 2.481868874162052e-06}
{"train_loss": 0.034998178482055664, "global_step": 244162, "epoch": 2743, "lr": 2.481688495513773e-06}
{"train_loss": 0.036692917346954346, "global_step": 244163, "epoch": 2743, "lr": 2.481508123253762e-06}
{"train_loss": 0.04163739085197449, "global_step": 244164, "epoch": 2743, "lr": 2.4813277573820293e-06}
{"train_loss": 0.040859583765268326, "global_step": 244165, "epoch": 2743, "lr": 2.4811473978986034e-06}
{"train_loss": 0.03657007962465286, "global_step": 244166, "epoch": 2743, "lr": 2.480967044803506e-06}
{"train_loss": 0.04618655517697334, "global_step": 244167, "epoch": 2743, "lr": 2.4807866980967653e-06}
{"train_loss": 0.013859488070011139, "global_step": 244168, "epoch": 2743, "lr": 2.480606357778398e-06}
{"train_loss": 0.03780318796634674, "global_step": 244169, "epoch": 2743, "lr": 2.480426023848442e-06}
{"train_loss": 0.05726635828614235, "global_step": 244170, "epoch": 2743, "lr": 2.48024569630691e-06}
{"train_loss": 0.04397062584757805, "global_step": 244171, "epoch": 2743, "lr": 2.480065375153834e-06}
{"train_loss": 0.035196591168642044, "global_step": 244172, "epoch": 2743, "lr": 2.4798850603892254e-06}
{"train_loss": 0.02724464051425457, "global_step": 244173, "epoch": 2743, "lr": 2.4797047520131235e-06}
{"train_loss": 0.06048809364438057, "global_step": 244174, "epoch": 2743, "lr": 2.4795244500255445e-06}
{"train_loss": 0.023182908073067665, "global_step": 244175, "epoch": 2743, "lr": 2.4793441544265162e-06}
{"train_loss": 0.031328972429037094, "global_step": 244176, "epoch": 2743, "lr": 2.4791638652160553e-06}
{"train_loss": 0.06929964572191238, "global_step": 244177, "epoch": 2743, "lr": 2.4789835823941954e-06}
{"train_loss": 0.016604457050561905, "global_step": 244178, "epoch": 2743, "lr": 2.478803305960953e-06}
{"train_loss": 0.047699738293886185, "global_step": 244179, "epoch": 2743, "lr": 2.4786230359163665e-06}
{"train_loss": 0.02009090967476368, "global_step": 244180, "epoch": 2743, "lr": 2.4784427722604364e-06}
{"train_loss": 0.03388042747974396, "global_step": 244181, "epoch": 2743, "lr": 2.478262514993207e-06}
{"train_loss": 0.08586272597312927, "global_step": 244182, "epoch": 2743, "lr": 2.4780822641147006e-06}
{"train_loss": 0.05423031374812126, "global_step": 244183, "epoch": 2743, "lr": 2.477902019624928e-06}
{"train_loss": 0.0533602200448513, "global_step": 244184, "epoch": 2743, "lr": 2.4777217815239286e-06}
{"train_loss": 0.037348151206970215, "global_step": 244185, "epoch": 2743, "lr": 2.4775415498117183e-06}
{"train_loss": 0.029879366979002953, "global_step": 244186, "epoch": 2743, "lr": 2.477361324488325e-06}
{"train_loss": 0.008793150074779987, "global_step": 244187, "epoch": 2743, "lr": 2.4771811055537666e-06}
{"train_loss": 0.15445417165756226, "global_step": 244188, "epoch": 2743, "lr": 2.47700089300808e-06}
{"train_loss": 0.02470875158905983, "global_step": 244189, "epoch": 2743, "lr": 2.476820686851278e-06}
{"train_loss": 0.024458926171064377, "global_step": 244190, "epoch": 2743, "lr": 2.476640487083387e-06}
{"train_loss": 0.029458103701472282, "global_step": 244191, "epoch": 2743, "lr": 2.4764602937044356e-06}
{"train_loss": 0.035672031342983246, "global_step": 244192, "epoch": 2743, "lr": 2.476280106714435e-06}
{"train_loss": 0.034461863338947296, "global_step": 244193, "epoch": 2743, "lr": 2.476099926113429e-06}
{"train_loss": 0.05135903134942055, "global_step": 244194, "epoch": 2743, "lr": 2.475919751901434e-06}
{"train_loss": 0.017338955774903297, "global_step": 244195, "epoch": 2743, "lr": 2.4757395840784627e-06}
{"train_loss": 0.07126022130250931, "global_step": 244196, "epoch": 2743, "lr": 2.475559422644558e-06}
{"train_loss": 0.04913964122533798, "global_step": 244197, "epoch": 2743, "lr": 2.4753792675997255e-06}
{"train_loss": 0.07698988169431686, "global_step": 244198, "epoch": 2743, "lr": 2.475199118944005e-06}
{"train_loss": 0.024625543504953384, "global_step": 244199, "epoch": 2743, "lr": 2.475018976677418e-06}
{"train_loss": 0.04515618085861206, "global_step": 244200, "epoch": 2743, "lr": 2.4748388407999755e-06}
{"train_loss": 0.03139229491353035, "global_step": 244201, "epoch": 2743, "lr": 2.4746587113117226e-06}
{"train_loss": 0.10313915461301804, "global_step": 244202, "epoch": 2743, "lr": 2.474478588212664e-06}
{"train_loss": 0.06271842867136002, "global_step": 244203, "epoch": 2743, "lr": 2.4742984715028392e-06}
{"train_loss": 0.0505424439907074, "global_step": 244204, "epoch": 2743, "lr": 2.4741183611822592e-06}
{"train_loss": 0.07518474757671356, "global_step": 244205, "epoch": 2743, "lr": 2.4739382572509626e-06}
{"train_loss": 0.04948616027832031, "global_step": 244206, "epoch": 2743, "lr": 2.4737581597089554e-06}
{"train_loss": 0.0585322342813015, "global_step": 244207, "epoch": 2743, "lr": 2.4735780685562817e-06}
{"train_loss": 0.014171918854117393, "global_step": 244208, "epoch": 2743, "lr": 2.4733979837929523e-06}
{"train_loss": 0.03671646863222122, "global_step": 244209, "epoch": 2743, "lr": 2.4732179054189954e-06}
{"train_loss": 0.038956694304943085, "global_step": 244210, "epoch": 2743, "lr": 2.473037833434433e-06}
{"train_loss": 0.052493609488010406, "global_step": 244211, "epoch": 2743, "lr": 2.472857767839298e-06}
{"train_loss": 0.018408184871077538, "global_step": 244212, "epoch": 2743, "lr": 2.4726777086335972e-06}
{"train_loss": 0.07344777882099152, "global_step": 244213, "epoch": 2743, "lr": 2.4724976558173797e-06}
{"train_loss": 0.023866964504122734, "global_step": 244214, "epoch": 2743, "lr": 2.4723176093906396e-06}
{"train_loss": 0.04453594115133701, "global_step": 244215, "epoch": 2743, "lr": 2.472137569353433e-06, "val_loss": 9.449995040893555}
{"train_loss": 0.05551619455218315, "global_step": 244216, "epoch": 2744, "lr": 2.4719575357057546e-06}
{"train_loss": 0.03344258666038513, "global_step": 244217, "epoch": 2744, "lr": 2.471777508447648e-06}
{"train_loss": 0.027183210477232933, "global_step": 244218, "epoch": 2744, "lr": 2.4715974875791305e-06}
{"train_loss": 0.042538128793239594, "global_step": 244219, "epoch": 2744, "lr": 2.4714174731002293e-06}
{"train_loss": 0.03613700717687607, "global_step": 244220, "epoch": 2744, "lr": 2.471237465010967e-06}
{"train_loss": 0.08583144843578339, "global_step": 244221, "epoch": 2744, "lr": 2.471057463311366e-06}
{"train_loss": 0.03790786862373352, "global_step": 244222, "epoch": 2744, "lr": 2.4708774680014534e-06}
{"train_loss": 0.06623028218746185, "global_step": 244223, "epoch": 2744, "lr": 2.4706974790812463e-06}
{"train_loss": 0.05450106039643288, "global_step": 244224, "epoch": 2744, "lr": 2.470517496550784e-06}
{"train_loss": 0.05588391795754433, "global_step": 244225, "epoch": 2744, "lr": 2.4703375204100766e-06}
{"train_loss": 0.05222693085670471, "global_step": 244226, "epoch": 2744, "lr": 2.4701575506591577e-06}
{"train_loss": 0.02708638459444046, "global_step": 244227, "epoch": 2744, "lr": 2.469977587298045e-06}
{"train_loss": 0.019422229379415512, "global_step": 244228, "epoch": 2744, "lr": 2.4697976303267588e-06}
{"train_loss": 0.03722897171974182, "global_step": 244229, "epoch": 2744, "lr": 2.4696176797453343e-06}
{"train_loss": 0.051416512578725815, "global_step": 244230, "epoch": 2744, "lr": 2.469437735553787e-06}
{"train_loss": 0.022733792662620544, "global_step": 244231, "epoch": 2744, "lr": 2.4692577977521504e-06}
{"train_loss": 0.019942479208111763, "global_step": 244232, "epoch": 2744, "lr": 2.4690778663404414e-06}
{"train_loss": 0.07615726441144943, "global_step": 244233, "epoch": 2744, "lr": 2.4688979413186765e-06}
{"train_loss": 0.03377203270792961, "global_step": 244234, "epoch": 2744, "lr": 2.4687180226869e-06}
{"train_loss": 0.032839030027389526, "global_step": 244235, "epoch": 2744, "lr": 2.468538110445112e-06}
{"train_loss": 0.062286704778671265, "global_step": 244236, "epoch": 2744, "lr": 2.468358204593357e-06}
{"train_loss": 0.03952912986278534, "global_step": 244237, "epoch": 2744, "lr": 2.4681783051316576e-06}
{"train_loss": 0.035087261348962784, "global_step": 244238, "epoch": 2744, "lr": 2.4679984120600242e-06}
{"train_loss": 0.044052671641111374, "global_step": 244239, "epoch": 2744, "lr": 2.467818525378496e-06}
{"train_loss": 0.06282325088977814, "global_step": 244240, "epoch": 2744, "lr": 2.467638645087078e-06}
{"train_loss": 0.04564476013183594, "global_step": 244241, "epoch": 2744, "lr": 2.467458771185821e-06}
{"train_loss": 0.07054021954536438, "global_step": 244242, "epoch": 2744, "lr": 2.4672789036747247e-06}
{"train_loss": 0.01627972722053528, "global_step": 244243, "epoch": 2744, "lr": 2.467099042553833e-06}
{"train_loss": 0.021275993436574936, "global_step": 244244, "epoch": 2744, "lr": 2.4669191878231524e-06}
{"train_loss": 0.038071975111961365, "global_step": 244245, "epoch": 2744, "lr": 2.466739339482721e-06}
{"train_loss": 0.02818882092833519, "global_step": 244246, "epoch": 2744, "lr": 2.46655949753255e-06}
{"train_loss": 0.028819674625992775, "global_step": 244247, "epoch": 2744, "lr": 2.466379661972679e-06}
{"train_loss": 0.050082672387361526, "global_step": 244248, "epoch": 2744, "lr": 2.466199832803112e-06}
{"train_loss": 0.0789932832121849, "global_step": 244249, "epoch": 2744, "lr": 2.466020010023895e-06}
{"train_loss": 0.049914538860321045, "global_step": 244250, "epoch": 2744, "lr": 2.4658401936350386e-06}
{"train_loss": 0.06217645853757858, "global_step": 244251, "epoch": 2744, "lr": 2.4656603836365757e-06}
{"train_loss": 0.06983281672000885, "global_step": 244252, "epoch": 2744, "lr": 2.4654805800285173e-06}
{"train_loss": 0.048852212727069855, "global_step": 244253, "epoch": 2744, "lr": 2.4653007828108977e-06}
{"train_loss": 0.022986575961112976, "global_step": 244254, "epoch": 2744, "lr": 2.4651209919837436e-06}
{"train_loss": 0.024155044928193092, "global_step": 244255, "epoch": 2744, "lr": 2.4649412075470723e-06}
{"train_loss": 0.019257044419646263, "global_step": 244256, "epoch": 2744, "lr": 2.4647614295009114e-06}
{"train_loss": 0.046519845724105835, "global_step": 244257, "epoch": 2744, "lr": 2.4645816578452774e-06}
{"train_loss": 0.03542855754494667, "global_step": 244258, "epoch": 2744, "lr": 2.4644018925802093e-06}
{"train_loss": 0.07529834657907486, "global_step": 244259, "epoch": 2744, "lr": 2.464222133705718e-06}
{"train_loss": 0.03960157930850983, "global_step": 244260, "epoch": 2744, "lr": 2.464042381221837e-06}
{"train_loss": 0.04940129444003105, "global_step": 244261, "epoch": 2744, "lr": 2.4638626351285833e-06}
{"train_loss": 0.027320604771375656, "global_step": 244262, "epoch": 2744, "lr": 2.4636828954259895e-06}
{"train_loss": 0.05298411101102829, "global_step": 244263, "epoch": 2744, "lr": 2.4635031621140615e-06}
{"train_loss": 0.049053438007831573, "global_step": 244264, "epoch": 2744, "lr": 2.4633234351928493e-06}
{"train_loss": 0.04963827505707741, "global_step": 244265, "epoch": 2744, "lr": 2.463143714662353e-06}
{"train_loss": 0.02163599245250225, "global_step": 244266, "epoch": 2744, "lr": 2.462964000522616e-06}
{"train_loss": 0.04641563817858696, "global_step": 244267, "epoch": 2744, "lr": 2.4627842927736567e-06}
{"train_loss": 0.05434388667345047, "global_step": 244268, "epoch": 2744, "lr": 2.4626045914154906e-06}
{"train_loss": 0.02568524330854416, "global_step": 244269, "epoch": 2744, "lr": 2.462424896448151e-06}
{"train_loss": 0.043821029365062714, "global_step": 244270, "epoch": 2744, "lr": 2.4622452078716495e-06}
{"train_loss": 0.0795396938920021, "global_step": 244271, "epoch": 2744, "lr": 2.4620655256860305e-06}
{"train_loss": 0.06869232654571533, "global_step": 244272, "epoch": 2744, "lr": 2.461885849891299e-06}
{"train_loss": 0.042301807552576065, "global_step": 244273, "epoch": 2744, "lr": 2.4617061804874997e-06}
{"train_loss": 0.043294116854667664, "global_step": 244274, "epoch": 2744, "lr": 2.461526517474633e-06}
{"train_loss": 0.05110161378979683, "global_step": 244275, "epoch": 2744, "lr": 2.461346860852742e-06}
{"train_loss": 0.033869657665491104, "global_step": 244276, "epoch": 2744, "lr": 2.461167210621834e-06}
{"train_loss": 0.027389045804739, "global_step": 244277, "epoch": 2744, "lr": 2.4609875667819527e-06}
{"train_loss": 0.031241105869412422, "global_step": 244278, "epoch": 2744, "lr": 2.4608079293331034e-06}
{"train_loss": 0.009015536867082119, "global_step": 244279, "epoch": 2744, "lr": 2.4606282982753303e-06}
{"train_loss": 0.02974053844809532, "global_step": 244280, "epoch": 2744, "lr": 2.460448673608634e-06}
{"train_loss": 0.03666390851140022, "global_step": 244281, "epoch": 2744, "lr": 2.460269055333059e-06}
{"train_loss": 0.027898814529180527, "global_step": 244282, "epoch": 2744, "lr": 2.4600894434486156e-06}
{"train_loss": 0.03707459196448326, "global_step": 244283, "epoch": 2744, "lr": 2.459909837955343e-06}
{"train_loss": 0.016646740958094597, "global_step": 244284, "epoch": 2744, "lr": 2.4597302388532473e-06}
{"train_loss": 0.07139579951763153, "global_step": 244285, "epoch": 2744, "lr": 2.459550646142367e-06}
{"train_loss": 0.05546313896775246, "global_step": 244286, "epoch": 2744, "lr": 2.4593710598227183e-06}
{"train_loss": 0.03351147472858429, "global_step": 244287, "epoch": 2744, "lr": 2.4591914798943294e-06}
{"train_loss": 0.025718901306390762, "global_step": 244288, "epoch": 2744, "lr": 2.459011906357217e-06}
{"train_loss": 0.03556714579463005, "global_step": 244289, "epoch": 2744, "lr": 2.4588323392114086e-06}
{"train_loss": 0.07810766249895096, "global_step": 244290, "epoch": 2744, "lr": 2.4586527784569437e-06}
{"train_loss": 0.024582048878073692, "global_step": 244291, "epoch": 2744, "lr": 2.4584732240938213e-06}
{"train_loss": 0.059357285499572754, "global_step": 244292, "epoch": 2744, "lr": 2.4582936761220864e-06}
{"train_loss": 0.022785043343901634, "global_step": 244293, "epoch": 2744, "lr": 2.45811413454175e-06}
{"train_loss": 0.07337165623903275, "global_step": 244294, "epoch": 2744, "lr": 2.4579345993528456e-06}
{"train_loss": 0.03315801918506622, "global_step": 244295, "epoch": 2744, "lr": 2.457755070555384e-06}
{"train_loss": 0.05027363821864128, "global_step": 244296, "epoch": 2744, "lr": 2.45757554814941e-06}
{"train_loss": 0.11845815926790237, "global_step": 244297, "epoch": 2744, "lr": 2.457396032134923e-06}
{"train_loss": 0.048615023493766785, "global_step": 244298, "epoch": 2744, "lr": 2.4572165225119683e-06}
{"train_loss": 0.06371083110570908, "global_step": 244299, "epoch": 2744, "lr": 2.457037019280556e-06}
{"train_loss": 0.04933987185359001, "global_step": 244300, "epoch": 2744, "lr": 2.4568575224407197e-06}
{"train_loss": 0.06072014570236206, "global_step": 244301, "epoch": 2744, "lr": 2.4566780319924708e-06}
{"train_loss": 0.06152181699872017, "global_step": 244302, "epoch": 2744, "lr": 2.4564985479358537e-06}
{"train_loss": 0.07469075918197632, "global_step": 244303, "epoch": 2744, "lr": 2.456319070270874e-06}
{"train_loss": 0.04511944538379988, "global_step": 244304, "epoch": 2744, "lr": 2.456139598997564e-06, "val_loss": 9.447527885437012}
{"train_loss": 0.0319075845181942, "global_step": 244305, "epoch": 2745, "lr": 2.455960134115953e-06}
{"train_loss": 0.038868971168994904, "global_step": 244306, "epoch": 2745, "lr": 2.4557806756260515e-06}
{"train_loss": 0.06254376471042633, "global_step": 244307, "epoch": 2745, "lr": 2.4556012235278922e-06}
{"train_loss": 0.04610998556017876, "global_step": 244308, "epoch": 2745, "lr": 2.4554217778214982e-06}
{"train_loss": 0.06839705258607864, "global_step": 244309, "epoch": 2745, "lr": 2.455242338506891e-06}
{"train_loss": 0.016400188207626343, "global_step": 244310, "epoch": 2745, "lr": 2.455062905584099e-06}
{"train_loss": 0.010266447439789772, "global_step": 244311, "epoch": 2745, "lr": 2.454883479053144e-06}
{"train_loss": 0.06168389692902565, "global_step": 244312, "epoch": 2745, "lr": 2.454704058914048e-06}
{"train_loss": 0.07370910048484802, "global_step": 244313, "epoch": 2745, "lr": 2.4545246451668446e-06}
{"train_loss": 0.0578954741358757, "global_step": 244314, "epoch": 2745, "lr": 2.454345237811545e-06}
{"train_loss": 0.06208660081028938, "global_step": 244315, "epoch": 2745, "lr": 2.4541658368481823e-06}
{"train_loss": 0.033804524689912796, "global_step": 244316, "epoch": 2745, "lr": 2.4539864422767734e-06}
{"train_loss": 0.029509274289011955, "global_step": 244317, "epoch": 2745, "lr": 2.4538070540973513e-06}
{"train_loss": 0.03651151433587074, "global_step": 244318, "epoch": 2745, "lr": 2.4536276723099274e-06}
{"train_loss": 0.060500048100948334, "global_step": 244319, "epoch": 2745, "lr": 2.4534482969145457e-06}
{"train_loss": 0.021194128319621086, "global_step": 244320, "epoch": 2745, "lr": 2.4532689279112066e-06}
{"train_loss": 0.039435017853975296, "global_step": 244321, "epoch": 2745, "lr": 2.4530895652999543e-06}
{"train_loss": 0.023551611229777336, "global_step": 244322, "epoch": 2745, "lr": 2.4529102090807997e-06}
{"train_loss": 0.03036613017320633, "global_step": 244323, "epoch": 2745, "lr": 2.4527308592537766e-06}
{"train_loss": 0.05580854415893555, "global_step": 244324, "epoch": 2745, "lr": 2.4525515158188962e-06}
{"train_loss": 0.11789575964212418, "global_step": 244325, "epoch": 2745, "lr": 2.452372178776191e-06}
{"train_loss": 0.036750677973032, "global_step": 244326, "epoch": 2745, "lr": 2.452192848125695e-06}
{"train_loss": 0.04972558096051216, "global_step": 244327, "epoch": 2745, "lr": 2.4520135238674134e-06}
{"train_loss": 0.03818833827972412, "global_step": 244328, "epoch": 2745, "lr": 2.4518342060013855e-06}
{"train_loss": 0.06018882617354393, "global_step": 244329, "epoch": 2745, "lr": 2.4516548945276217e-06}
{"train_loss": 0.05567037686705589, "global_step": 244330, "epoch": 2745, "lr": 2.4514755894461615e-06}
{"train_loss": 0.06058801710605621, "global_step": 244331, "epoch": 2745, "lr": 2.451296290757016e-06}
{"train_loss": 0.07889270782470703, "global_step": 244332, "epoch": 2745, "lr": 2.4511169984602177e-06}
{"train_loss": 0.019829601049423218, "global_step": 244333, "epoch": 2745, "lr": 2.4509377125557787e-06}
{"train_loss": 0.0584866963326931, "global_step": 244334, "epoch": 2745, "lr": 2.4507584330437427e-06}
{"train_loss": 0.03493731468915939, "global_step": 244335, "epoch": 2745, "lr": 2.4505791599241157e-06}
{"train_loss": 0.04398186504840851, "global_step": 244336, "epoch": 2745, "lr": 2.4503998931969307e-06}
{"train_loss": 0.03751177713274956, "global_step": 244337, "epoch": 2745, "lr": 2.4502206328622047e-06}
{"train_loss": 0.053900882601737976, "global_step": 244338, "epoch": 2745, "lr": 2.4500413789199763e-06}
{"train_loss": 0.03930383175611496, "global_step": 244339, "epoch": 2745, "lr": 2.4498621313702563e-06}
{"train_loss": 0.04047951474785805, "global_step": 244340, "epoch": 2745, "lr": 2.4496828902130733e-06}
{"train_loss": 0.05872983857989311, "global_step": 244341, "epoch": 2745, "lr": 2.4495036554484487e-06}
{"train_loss": 0.06888057291507721, "global_step": 244342, "epoch": 2745, "lr": 2.4493244270764106e-06}
{"train_loss": 0.02285805158317089, "global_step": 244343, "epoch": 2745, "lr": 2.4491452050969865e-06}
{"train_loss": 0.03775288164615631, "global_step": 244344, "epoch": 2745, "lr": 2.4489659895101823e-06}
{"train_loss": 0.03316232189536095, "global_step": 244345, "epoch": 2745, "lr": 2.448786780316048e-06}
{"train_loss": 0.05089837685227394, "global_step": 244346, "epoch": 2745, "lr": 2.4486075775145835e-06}
{"train_loss": 0.05518293008208275, "global_step": 244347, "epoch": 2745, "lr": 2.4484283811058385e-06}
{"train_loss": 0.060101721435785294, "global_step": 244348, "epoch": 2745, "lr": 2.4482491910898075e-06}
{"train_loss": 0.008102485910058022, "global_step": 244349, "epoch": 2745, "lr": 2.4480700074665407e-06}
{"train_loss": 0.028542717918753624, "global_step": 244350, "epoch": 2745, "lr": 2.447890830236044e-06}
{"train_loss": 0.04646025598049164, "global_step": 244351, "epoch": 2745, "lr": 2.4477116593983553e-06}
{"train_loss": 0.02513428032398224, "global_step": 244352, "epoch": 2745, "lr": 2.447532494953486e-06}
{"train_loss": 0.04335314407944679, "global_step": 244353, "epoch": 2745, "lr": 2.4473533369014703e-06}
{"train_loss": 0.041223637759685516, "global_step": 244354, "epoch": 2745, "lr": 2.4471741852423237e-06}
{"train_loss": 0.06216774135828018, "global_step": 244355, "epoch": 2745, "lr": 2.44699503997608e-06}
{"train_loss": 0.013757477514445782, "global_step": 244356, "epoch": 2745, "lr": 2.4468159011027557e-06}
{"train_loss": 0.07881145179271698, "global_step": 244357, "epoch": 2745, "lr": 2.446636768622379e-06}
{"train_loss": 0.049673471599817276, "global_step": 244358, "epoch": 2745, "lr": 2.4464576425349717e-06}
{"train_loss": 0.03677418455481529, "global_step": 244359, "epoch": 2745, "lr": 2.4462785228405616e-06}
{"train_loss": 0.06916840374469757, "global_step": 244360, "epoch": 2745, "lr": 2.44609940953916e-06}
{"train_loss": 0.05427950248122215, "global_step": 244361, "epoch": 2745, "lr": 2.4459203026308108e-06}
{"train_loss": 0.06079431623220444, "global_step": 244362, "epoch": 2745, "lr": 2.4457412021155204e-06}
{"train_loss": 0.04974506422877312, "global_step": 244363, "epoch": 2745, "lr": 2.4455621079933213e-06}
{"train_loss": 0.04991304874420166, "global_step": 244364, "epoch": 2745, "lr": 2.445383020264247e-06}
{"train_loss": 0.05965245142579079, "global_step": 244365, "epoch": 2745, "lr": 2.4452039389282977e-06}
{"train_loss": 0.08282474428415298, "global_step": 244366, "epoch": 2745, "lr": 2.4450248639855232e-06}
{"train_loss": 0.05210619792342186, "global_step": 244367, "epoch": 2745, "lr": 2.444845795435924e-06}
{"train_loss": 0.023471953347325325, "global_step": 244368, "epoch": 2745, "lr": 2.4446667332795492e-06}
{"train_loss": 0.03802413493394852, "global_step": 244369, "epoch": 2745, "lr": 2.4444876775163938e-06}
{"train_loss": 0.10861317813396454, "global_step": 244370, "epoch": 2745, "lr": 2.4443086281465134e-06}
{"train_loss": 0.025421207770705223, "global_step": 244371, "epoch": 2745, "lr": 2.4441295851699018e-06}
{"train_loss": 0.07423725724220276, "global_step": 244372, "epoch": 2745, "lr": 2.4439505485866097e-06}
{"train_loss": 0.021285802125930786, "global_step": 244373, "epoch": 2745, "lr": 2.443771518396637e-06}
{"train_loss": 0.03363353759050369, "global_step": 244374, "epoch": 2745, "lr": 2.443592494600028e-06}
{"train_loss": 0.028111491352319717, "global_step": 244375, "epoch": 2745, "lr": 2.443413477196793e-06}
{"train_loss": 0.047071926295757294, "global_step": 244376, "epoch": 2745, "lr": 2.443234466186972e-06}
{"train_loss": 0.038321562111377716, "global_step": 244377, "epoch": 2745, "lr": 2.443055461570565e-06}
{"train_loss": 0.057686854153871536, "global_step": 244378, "epoch": 2745, "lr": 2.442876463347621e-06}
{"train_loss": 0.037144985049963, "global_step": 244379, "epoch": 2745, "lr": 2.4426974715181404e-06}
{"train_loss": 0.04614380747079849, "global_step": 244380, "epoch": 2745, "lr": 2.442518486082168e-06}
{"train_loss": 0.053164102137088776, "global_step": 244381, "epoch": 2745, "lr": 2.442339507039715e-06}
{"train_loss": 0.035368144512176514, "global_step": 244382, "epoch": 2745, "lr": 2.442160534390819e-06}
{"train_loss": 0.007936522364616394, "global_step": 244383, "epoch": 2745, "lr": 2.4419815681354873e-06}
{"train_loss": 0.04558933898806572, "global_step": 244384, "epoch": 2745, "lr": 2.4418026082737524e-06}
{"train_loss": 0.03716637194156647, "global_step": 244385, "epoch": 2745, "lr": 2.4416236548056415e-06}
{"train_loss": 0.03758399188518524, "global_step": 244386, "epoch": 2745, "lr": 2.4414447077311666e-06}
{"train_loss": 0.048703406006097794, "global_step": 244387, "epoch": 2745, "lr": 2.4412657670503657e-06}
{"train_loss": 0.025686409324407578, "global_step": 244388, "epoch": 2745, "lr": 2.4410868327632507e-06}
{"train_loss": 0.03957272693514824, "global_step": 244389, "epoch": 2745, "lr": 2.44090790486986e-06}
{"train_loss": 0.02879413776099682, "global_step": 244390, "epoch": 2745, "lr": 2.4407289833701995e-06}
{"train_loss": 0.03705781325697899, "global_step": 244391, "epoch": 2745, "lr": 2.4405500682643134e-06}
{"train_loss": 0.009603471495211124, "global_step": 244392, "epoch": 2745, "lr": 2.4403711595522126e-06}
{"train_loss": 0.045168860491072195, "global_step": 244393, "epoch": 2745, "lr": 2.4401922572339252e-06, "val_loss": 9.426984786987305, "train_action_mse_error": 33.41790771484375}
{"train_loss": 0.04269818961620331, "global_step": 244394, "epoch": 2746, "lr": 2.4400133613094677e-06}
{"train_loss": 0.028822259977459908, "global_step": 244395, "epoch": 2746, "lr": 2.439834471778879e-06}
{"train_loss": 0.05201619863510132, "global_step": 244396, "epoch": 2746, "lr": 2.43965558864217e-06}
{"train_loss": 0.05398882180452347, "global_step": 244397, "epoch": 2746, "lr": 2.4394767118993745e-06}
{"train_loss": 0.048670995980501175, "global_step": 244398, "epoch": 2746, "lr": 2.439297841550503e-06}
{"train_loss": 0.06551142036914825, "global_step": 244399, "epoch": 2746, "lr": 2.4391189775955893e-06}
{"train_loss": 0.016491996124386787, "global_step": 244400, "epoch": 2746, "lr": 2.438940120034666e-06}
{"train_loss": 0.05700582265853882, "global_step": 244401, "epoch": 2746, "lr": 2.43876126886774e-06}
{"train_loss": 0.034574490040540695, "global_step": 244402, "epoch": 2746, "lr": 2.4385824240948487e-06}
{"train_loss": 0.059223536401987076, "global_step": 244403, "epoch": 2746, "lr": 2.4384035857160037e-06}
{"train_loss": 0.08223047852516174, "global_step": 244404, "epoch": 2746, "lr": 2.4382247537312386e-06}
{"train_loss": 0.06495489180088043, "global_step": 244405, "epoch": 2746, "lr": 2.4380459281405697e-06}
{"train_loss": 0.048270080238580704, "global_step": 244406, "epoch": 2746, "lr": 2.4378671089440365e-06}
{"train_loss": 0.058925729244947433, "global_step": 244407, "epoch": 2746, "lr": 2.4376882961416437e-06}
{"train_loss": 0.046826571226119995, "global_step": 244408, "epoch": 2746, "lr": 2.4375094897334305e-06}
{"train_loss": 0.04798465967178345, "global_step": 244409, "epoch": 2746, "lr": 2.437330689719408e-06}
{"train_loss": 0.05144882574677467, "global_step": 244410, "epoch": 2746, "lr": 2.43715189609961e-06}
{"train_loss": 0.04091794043779373, "global_step": 244411, "epoch": 2746, "lr": 2.436973108874052e-06}
{"train_loss": 0.016316542401909828, "global_step": 244412, "epoch": 2746, "lr": 2.436794328042774e-06}
{"train_loss": 0.06173461303114891, "global_step": 244413, "epoch": 2746, "lr": 2.4366155536057756e-06}
{"train_loss": 0.0431007519364357, "global_step": 244414, "epoch": 2746, "lr": 2.4364367855631067e-06}
{"train_loss": 0.051385894417762756, "global_step": 244415, "epoch": 2746, "lr": 2.4362580239147725e-06}
{"train_loss": 0.08929583430290222, "global_step": 244416, "epoch": 2746, "lr": 2.436079268660807e-06}
{"train_loss": 0.03636090084910393, "global_step": 244417, "epoch": 2746, "lr": 2.4359005198012264e-06}
{"train_loss": 0.04917226359248161, "global_step": 244418, "epoch": 2746, "lr": 2.4357217773360696e-06}
{"train_loss": 0.05042878910899162, "global_step": 244419, "epoch": 2746, "lr": 2.435543041265337e-06}
{"train_loss": 0.06845144182443619, "global_step": 244420, "epoch": 2746, "lr": 2.4353643115890778e-06}
{"train_loss": 0.050311129540205, "global_step": 244421, "epoch": 2746, "lr": 2.4351855883072983e-06}
{"train_loss": 0.05441926419734955, "global_step": 244422, "epoch": 2746, "lr": 2.435006871420026e-06}
{"train_loss": 0.051913321018218994, "global_step": 244423, "epoch": 2746, "lr": 2.434828160927294e-06}
{"train_loss": 0.08088520914316177, "global_step": 244424, "epoch": 2746, "lr": 2.4346494568291133e-06}
{"train_loss": 0.05131177231669426, "global_step": 244425, "epoch": 2746, "lr": 2.4344707591255176e-06}
{"train_loss": 0.033897969871759415, "global_step": 244426, "epoch": 2746, "lr": 2.434292067816524e-06}
{"train_loss": 0.028194120153784752, "global_step": 244427, "epoch": 2746, "lr": 2.4341133829021644e-06}
{"train_loss": 0.0672013983130455, "global_step": 244428, "epoch": 2746, "lr": 2.4339347043824513e-06}
{"train_loss": 0.04196344316005707, "global_step": 244429, "epoch": 2746, "lr": 2.433756032257428e-06}
{"train_loss": 0.0726805180311203, "global_step": 244430, "epoch": 2746, "lr": 2.4335773665270956e-06}
{"train_loss": 0.00947654154151678, "global_step": 244431, "epoch": 2746, "lr": 2.433398707191492e-06}
{"train_loss": 0.016227180138230324, "global_step": 244432, "epoch": 2746, "lr": 2.43322005425064e-06}
{"train_loss": 0.05391468107700348, "global_step": 244433, "epoch": 2746, "lr": 2.4330414077045614e-06}
{"train_loss": 0.07536669075489044, "global_step": 244434, "epoch": 2746, "lr": 2.4328627675532787e-06}
{"train_loss": 0.11100537329912186, "global_step": 244435, "epoch": 2746, "lr": 2.4326841337968144e-06}
{"train_loss": 0.09442146122455597, "global_step": 244436, "epoch": 2746, "lr": 2.4325055064352067e-06}
{"train_loss": 0.02928648144006729, "global_step": 244437, "epoch": 2746, "lr": 2.4323268854684554e-06}
{"train_loss": 0.06182391941547394, "global_step": 244438, "epoch": 2746, "lr": 2.4321482708966114e-06}
{"train_loss": 0.024369273334741592, "global_step": 244439, "epoch": 2746, "lr": 2.4319696627196744e-06}
{"train_loss": 0.029907332733273506, "global_step": 244440, "epoch": 2746, "lr": 2.4317910609376883e-06}
{"train_loss": 0.06450296193361282, "global_step": 244441, "epoch": 2746, "lr": 2.431612465550659e-06}
{"train_loss": 0.028394483029842377, "global_step": 244442, "epoch": 2746, "lr": 2.4314338765586307e-06}
{"train_loss": 0.034435465931892395, "global_step": 244443, "epoch": 2746, "lr": 2.431255293961604e-06}
{"train_loss": 0.07657457143068314, "global_step": 244444, "epoch": 2746, "lr": 2.4310767177596285e-06}
{"train_loss": 0.02958705648779869, "global_step": 244445, "epoch": 2746, "lr": 2.4308981479527037e-06}
{"train_loss": 0.018425125628709793, "global_step": 244446, "epoch": 2746, "lr": 2.4307195845408693e-06}
{"train_loss": 0.038390569388866425, "global_step": 244447, "epoch": 2746, "lr": 2.4305410275241416e-06}
{"train_loss": 0.08022661507129669, "global_step": 244448, "epoch": 2746, "lr": 2.4303624769025592e-06}
{"train_loss": 0.02217770367860794, "global_step": 244449, "epoch": 2746, "lr": 2.4301839326761222e-06}
{"train_loss": 0.033732812851667404, "global_step": 244450, "epoch": 2746, "lr": 2.4300053948448755e-06}
{"train_loss": 0.07456623017787933, "global_step": 244451, "epoch": 2746, "lr": 2.429826863408824e-06}
{"train_loss": 0.04149492084980011, "global_step": 244452, "epoch": 2746, "lr": 2.429648338368007e-06}
{"train_loss": 0.01424506027251482, "global_step": 244453, "epoch": 2746, "lr": 2.429469819722452e-06}
{"train_loss": 0.08694484829902649, "global_step": 244454, "epoch": 2746, "lr": 2.42929130747217e-06}
{"train_loss": 0.043341297656297684, "global_step": 244455, "epoch": 2746, "lr": 2.429112801617189e-06}
{"train_loss": 0.06724293529987335, "global_step": 244456, "epoch": 2746, "lr": 2.4289343021575374e-06}
{"train_loss": 0.04393669217824936, "global_step": 244457, "epoch": 2746, "lr": 2.428755809093225e-06}
{"train_loss": 0.045810021460056305, "global_step": 244458, "epoch": 2746, "lr": 2.4285773224243026e-06}
{"train_loss": 0.008025390096008778, "global_step": 244459, "epoch": 2746, "lr": 2.42839884215077e-06}
{"train_loss": 0.050338905304670334, "global_step": 244460, "epoch": 2746, "lr": 2.4282203682726545e-06}
{"train_loss": 0.03001425415277481, "global_step": 244461, "epoch": 2746, "lr": 2.4280419007899957e-06}
{"train_loss": 0.025868073105812073, "global_step": 244462, "epoch": 2746, "lr": 2.427863439702799e-06}
{"train_loss": 0.04674503579735756, "global_step": 244463, "epoch": 2746, "lr": 2.427684985011097e-06}
{"train_loss": 0.02016829326748848, "global_step": 244464, "epoch": 2746, "lr": 2.4275065367149132e-06}
{"train_loss": 0.05934367701411247, "global_step": 244465, "epoch": 2746, "lr": 2.427328094814274e-06}
{"train_loss": 0.07801056653261185, "global_step": 244466, "epoch": 2746, "lr": 2.427149659309197e-06}
{"train_loss": 0.023873886093497276, "global_step": 244467, "epoch": 2746, "lr": 2.426971230199715e-06}
{"train_loss": 0.041276805102825165, "global_step": 244468, "epoch": 2746, "lr": 2.426792807485839e-06}
{"train_loss": 0.06370961666107178, "global_step": 244469, "epoch": 2746, "lr": 2.4266143911676088e-06}
{"train_loss": 0.06614950299263, "global_step": 244470, "epoch": 2746, "lr": 2.4264359812450345e-06}
{"train_loss": 0.07181177288293839, "global_step": 244471, "epoch": 2746, "lr": 2.426257577718144e-06}
{"train_loss": 0.035313479602336884, "global_step": 244472, "epoch": 2746, "lr": 2.4260791805869708e-06}
{"train_loss": 0.031677328050136566, "global_step": 244473, "epoch": 2746, "lr": 2.425900789851526e-06}
{"train_loss": 0.01107497327029705, "global_step": 244474, "epoch": 2746, "lr": 2.425722405511843e-06}
{"train_loss": 0.0508672297000885, "global_step": 244475, "epoch": 2746, "lr": 2.4255440275679386e-06}
{"train_loss": 0.03962527588009834, "global_step": 244476, "epoch": 2746, "lr": 2.4253656560198456e-06}
{"train_loss": 0.017678838223218918, "global_step": 244477, "epoch": 2746, "lr": 2.4251872908675756e-06}
{"train_loss": 0.0333525687456131, "global_step": 244478, "epoch": 2746, "lr": 2.4250089321111615e-06}
{"train_loss": 0.03415459766983986, "global_step": 244479, "epoch": 2746, "lr": 2.4248305797506255e-06}
{"train_loss": 0.03547698259353638, "global_step": 244480, "epoch": 2746, "lr": 2.4246522337859954e-06}
{"train_loss": 0.017384007573127747, "global_step": 244481, "epoch": 2746, "lr": 2.424473894217283e-06}
{"train_loss": 0.046597613251945946, "global_step": 244482, "epoch": 2746, "lr": 2.424295561044526e-06, "val_loss": 9.483036994934082}
{"train_loss": 0.07700342684984207, "global_step": 244483, "epoch": 2747, "lr": 2.4241172342677364e-06}
{"train_loss": 0.051545027643442154, "global_step": 244484, "epoch": 2747, "lr": 2.423938913886953e-06}
{"train_loss": 0.034725774079561234, "global_step": 244485, "epoch": 2747, "lr": 2.423760599902186e-06}
{"train_loss": 0.03269476071000099, "global_step": 244486, "epoch": 2747, "lr": 2.423582292313464e-06}
{"train_loss": 0.04561147466301918, "global_step": 244487, "epoch": 2747, "lr": 2.4234039911208095e-06}
{"train_loss": 0.04592667147517204, "global_step": 244488, "epoch": 2747, "lr": 2.423225696324255e-06}
{"train_loss": 0.0726625993847847, "global_step": 244489, "epoch": 2747, "lr": 2.423047407923812e-06}
{"train_loss": 0.04450394958257675, "global_step": 244490, "epoch": 2747, "lr": 2.4228691259195134e-06}
{"train_loss": 0.0628010705113411, "global_step": 244491, "epoch": 2747, "lr": 2.422690850311382e-06}
{"train_loss": 0.039486926048994064, "global_step": 244492, "epoch": 2747, "lr": 2.42251258109944e-06}
{"train_loss": 0.019756097346544266, "global_step": 244493, "epoch": 2747, "lr": 2.4223343182837034e-06}
{"train_loss": 0.06792636960744858, "global_step": 244494, "epoch": 2747, "lr": 2.4221560618642167e-06}
{"train_loss": 0.03845236077904701, "global_step": 244495, "epoch": 2747, "lr": 2.421977811840981e-06}
{"train_loss": 0.014781690202653408, "global_step": 244496, "epoch": 2747, "lr": 2.4217995682140338e-06}
{"train_loss": 0.06370469927787781, "global_step": 244497, "epoch": 2747, "lr": 2.4216213309833924e-06}
{"train_loss": 0.0730162113904953, "global_step": 244498, "epoch": 2747, "lr": 2.42144310014909e-06}
{"train_loss": 0.044430211186409, "global_step": 244499, "epoch": 2747, "lr": 2.4212648757111487e-06}
{"train_loss": 0.009915024973452091, "global_step": 244500, "epoch": 2747, "lr": 2.42108665766958e-06}
{"train_loss": 0.040116630494594574, "global_step": 244501, "epoch": 2747, "lr": 2.420908446024417e-06}
{"train_loss": 0.07533001154661179, "global_step": 244502, "epoch": 2747, "lr": 2.4207302407756814e-06}
{"train_loss": 0.054653801023960114, "global_step": 244503, "epoch": 2747, "lr": 2.4205520419234075e-06}
{"train_loss": 0.04719395190477371, "global_step": 244504, "epoch": 2747, "lr": 2.4203738494676e-06}
{"train_loss": 0.02269297093153, "global_step": 244505, "epoch": 2747, "lr": 2.4201956634083035e-06}
{"train_loss": 0.049478739500045776, "global_step": 244506, "epoch": 2747, "lr": 2.420017483745518e-06}
{"train_loss": 0.048762984573841095, "global_step": 244507, "epoch": 2747, "lr": 2.4198393104792883e-06}
{"train_loss": 0.046650439500808716, "global_step": 244508, "epoch": 2747, "lr": 2.419661143609636e-06}
{"train_loss": 0.02354666031897068, "global_step": 244509, "epoch": 2747, "lr": 2.4194829831365727e-06}
{"train_loss": 0.05892369523644447, "global_step": 244510, "epoch": 2747, "lr": 2.419304829060137e-06}
{"train_loss": 0.03189496323466301, "global_step": 244511, "epoch": 2747, "lr": 2.41912668138034e-06}
{"train_loss": 0.023902079090476036, "global_step": 244512, "epoch": 2747, "lr": 2.418948540097216e-06}
{"train_loss": 0.042285673320293427, "global_step": 244513, "epoch": 2747, "lr": 2.4187704052107796e-06}
{"train_loss": 0.013646560721099377, "global_step": 244514, "epoch": 2747, "lr": 2.4185922767210655e-06}
{"train_loss": 0.020839715376496315, "global_step": 244515, "epoch": 2747, "lr": 2.4184141546280846e-06}
{"train_loss": 0.05341382697224617, "global_step": 244516, "epoch": 2747, "lr": 2.418236038931876e-06}
{"train_loss": 0.03886996954679489, "global_step": 244517, "epoch": 2747, "lr": 2.4180579296324502e-06}
{"train_loss": 0.02842576615512371, "global_step": 244518, "epoch": 2747, "lr": 2.417879826729841e-06}
{"train_loss": 0.08203604817390442, "global_step": 244519, "epoch": 2747, "lr": 2.417701730224059e-06}
{"train_loss": 0.08034087717533112, "global_step": 244520, "epoch": 2747, "lr": 2.417523640115149e-06}
{"train_loss": 0.08705230802297592, "global_step": 244521, "epoch": 2747, "lr": 2.4173455564031112e-06}
{"train_loss": 0.07270238548517227, "global_step": 244522, "epoch": 2747, "lr": 2.4171674790879896e-06}
{"train_loss": 0.023355955258011818, "global_step": 244523, "epoch": 2747, "lr": 2.4169894081697954e-06}
{"train_loss": 0.06938136368989944, "global_step": 244524, "epoch": 2747, "lr": 2.4168113436485617e-06}
{"train_loss": 0.04724080115556717, "global_step": 244525, "epoch": 2747, "lr": 2.4166332855243056e-06}
{"train_loss": 0.013197286054491997, "global_step": 244526, "epoch": 2747, "lr": 2.4164552337970493e-06}
{"train_loss": 0.04276111721992493, "global_step": 244527, "epoch": 2747, "lr": 2.4162771884668256e-06}
{"train_loss": 0.0332631915807724, "global_step": 244528, "epoch": 2747, "lr": 2.416099149533646e-06}
{"train_loss": 0.0388140007853508, "global_step": 244529, "epoch": 2747, "lr": 2.415921116997555e-06}
{"train_loss": 0.050285495817661285, "global_step": 244530, "epoch": 2747, "lr": 2.415743090858552e-06}
{"train_loss": 0.05734211206436157, "global_step": 244531, "epoch": 2747, "lr": 2.415565071116682e-06}
{"train_loss": 0.058668673038482666, "global_step": 244532, "epoch": 2747, "lr": 2.4153870577719563e-06}
{"train_loss": 0.04502315819263458, "global_step": 244533, "epoch": 2747, "lr": 2.415209050824396e-06}
{"train_loss": 0.04226972162723541, "global_step": 244534, "epoch": 2747, "lr": 2.415031050274036e-06}
{"train_loss": 0.037253282964229584, "global_step": 244535, "epoch": 2747, "lr": 2.4148530561208914e-06}
{"train_loss": 0.09464573860168457, "global_step": 244536, "epoch": 2747, "lr": 2.4146750683649964e-06}
{"train_loss": 0.03824712708592415, "global_step": 244537, "epoch": 2747, "lr": 2.4144970870063676e-06}
{"train_loss": 0.03321065381169319, "global_step": 244538, "epoch": 2747, "lr": 2.4143191120450213e-06}
{"train_loss": 0.045493412762880325, "global_step": 244539, "epoch": 2747, "lr": 2.4141411434809967e-06}
{"train_loss": 0.020805608481168747, "global_step": 244540, "epoch": 2747, "lr": 2.413963181314305e-06}
{"train_loss": 0.06681876629590988, "global_step": 244541, "epoch": 2747, "lr": 2.4137852255449846e-06}
{"train_loss": 0.07699834555387497, "global_step": 244542, "epoch": 2747, "lr": 2.413607276173041e-06}
{"train_loss": 0.04017599672079086, "global_step": 244543, "epoch": 2747, "lr": 2.413429333198508e-06}
{"train_loss": 0.013905737549066544, "global_step": 244544, "epoch": 2747, "lr": 2.4132513966214186e-06}
{"train_loss": 0.030238961800932884, "global_step": 244545, "epoch": 2747, "lr": 2.413073466441779e-06}
{"train_loss": 0.09124782681465149, "global_step": 244546, "epoch": 2747, "lr": 2.412895542659627e-06}
{"train_loss": 0.02536269649863243, "global_step": 244547, "epoch": 2747, "lr": 2.4127176252749796e-06}
{"train_loss": 0.03636645898222923, "global_step": 244548, "epoch": 2747, "lr": 2.4125397142878646e-06}
{"train_loss": 0.03674714267253876, "global_step": 244549, "epoch": 2747, "lr": 2.4123618096982993e-06}
{"train_loss": 0.053318560123443604, "global_step": 244550, "epoch": 2747, "lr": 2.412183911506316e-06}
{"train_loss": 0.06803112477064133, "global_step": 244551, "epoch": 2747, "lr": 2.412006019711932e-06}
{"train_loss": 0.020326465368270874, "global_step": 244552, "epoch": 2747, "lr": 2.411828134315175e-06}
{"train_loss": 0.09775189310312271, "global_step": 244553, "epoch": 2747, "lr": 2.4116502553160667e-06}
{"train_loss": 0.07378784567117691, "global_step": 244554, "epoch": 2747, "lr": 2.411472382714636e-06}
{"train_loss": 0.032194703817367554, "global_step": 244555, "epoch": 2747, "lr": 2.4112945165108923e-06}
{"train_loss": 0.026887642219662666, "global_step": 244556, "epoch": 2747, "lr": 2.4111166567048817e-06}
{"train_loss": 0.030332311987876892, "global_step": 244557, "epoch": 2747, "lr": 2.4109388032966084e-06}
{"train_loss": 0.05583716183900833, "global_step": 244558, "epoch": 2747, "lr": 2.410760956286112e-06}
{"train_loss": 0.019273970276117325, "global_step": 244559, "epoch": 2747, "lr": 2.4105831156734036e-06}
{"train_loss": 0.0692882239818573, "global_step": 244560, "epoch": 2747, "lr": 2.4104052814585165e-06}
{"train_loss": 0.10001236200332642, "global_step": 244561, "epoch": 2747, "lr": 2.4102274536414616e-06}
{"train_loss": 0.04933144897222519, "global_step": 244562, "epoch": 2747, "lr": 2.4100496322222723e-06}
{"train_loss": 0.041630033403635025, "global_step": 244563, "epoch": 2747, "lr": 2.4098718172009816e-06}
{"train_loss": 0.03939078375697136, "global_step": 244564, "epoch": 2747, "lr": 2.4096940085775954e-06}
{"train_loss": 0.03490757569670677, "global_step": 244565, "epoch": 2747, "lr": 2.4095162063521528e-06}
{"train_loss": 0.04833120107650757, "global_step": 244566, "epoch": 2747, "lr": 2.4093384105246642e-06}
{"train_loss": 0.051361579447984695, "global_step": 244567, "epoch": 2747, "lr": 2.4091606210951632e-06}
{"train_loss": 0.05421329662203789, "global_step": 244568, "epoch": 2747, "lr": 2.408982838063667e-06}
{"train_loss": 0.026706637814641, "global_step": 244569, "epoch": 2747, "lr": 2.408805061430208e-06}
{"train_loss": 0.04207281768321991, "global_step": 244570, "epoch": 2747, "lr": 2.4086272911948092e-06}
{"train_loss": 0.046848309519334455, "global_step": 244571, "epoch": 2747, "lr": 2.4084495273574812e-06, "val_loss": 9.516913414001465}
{"train_loss": 0.07066407054662704, "global_step": 244572, "epoch": 2748, "lr": 2.408271769918263e-06}
{"train_loss": 0.03631138801574707, "global_step": 244573, "epoch": 2748, "lr": 2.408094018877166e-06}
{"train_loss": 0.044380899518728256, "global_step": 244574, "epoch": 2748, "lr": 2.4079162742342286e-06}
{"train_loss": 0.032873641699552536, "global_step": 244575, "epoch": 2748, "lr": 2.4077385359894623e-06}
{"train_loss": 0.05151127651333809, "global_step": 244576, "epoch": 2748, "lr": 2.407560804142894e-06}
{"train_loss": 0.06644804775714874, "global_step": 244577, "epoch": 2748, "lr": 2.407383078694553e-06}
{"train_loss": 0.0500827357172966, "global_step": 244578, "epoch": 2748, "lr": 2.4072053596444545e-06}
{"train_loss": 0.04162417724728584, "global_step": 244579, "epoch": 2748, "lr": 2.4070276469926266e-06}
{"train_loss": 0.05445863679051399, "global_step": 244580, "epoch": 2748, "lr": 2.406849940739103e-06}
{"train_loss": 0.04466092959046364, "global_step": 244581, "epoch": 2748, "lr": 2.4066722408838893e-06}
{"train_loss": 0.010028477758169174, "global_step": 244582, "epoch": 2748, "lr": 2.406494547427024e-06}
{"train_loss": 0.06378211826086044, "global_step": 244583, "epoch": 2748, "lr": 2.406316860368518e-06}
{"train_loss": 0.05242356285452843, "global_step": 244584, "epoch": 2748, "lr": 2.406139179708411e-06}
{"train_loss": 0.025399550795555115, "global_step": 244585, "epoch": 2748, "lr": 2.4059615054467133e-06}
{"train_loss": 0.016520051285624504, "global_step": 244586, "epoch": 2748, "lr": 2.4057838375834584e-06}
{"train_loss": 0.11232607066631317, "global_step": 244587, "epoch": 2748, "lr": 2.4056061761186577e-06}
{"train_loss": 0.05589433014392853, "global_step": 244588, "epoch": 2748, "lr": 2.4054285210523554e-06}
{"train_loss": 0.03953586146235466, "global_step": 244589, "epoch": 2748, "lr": 2.4052508723845514e-06}
{"train_loss": 0.0450681634247303, "global_step": 244590, "epoch": 2748, "lr": 2.4050732301152902e-06}
{"train_loss": 0.045158859342336655, "global_step": 244591, "epoch": 2748, "lr": 2.4048955942445773e-06}
{"train_loss": 0.027186334133148193, "global_step": 244592, "epoch": 2748, "lr": 2.404717964772457e-06}
{"train_loss": 0.053079087287187576, "global_step": 244593, "epoch": 2748, "lr": 2.40454034169893e-06}
{"train_loss": 0.0705265998840332, "global_step": 244594, "epoch": 2748, "lr": 2.404362725024045e-06}
{"train_loss": 0.04369690269231796, "global_step": 244595, "epoch": 2748, "lr": 2.4041851147478032e-06}
{"train_loss": 0.05345025658607483, "global_step": 244596, "epoch": 2748, "lr": 2.4040075108702485e-06}
{"train_loss": 0.09289167821407318, "global_step": 244597, "epoch": 2748, "lr": 2.4038299133913866e-06}
{"train_loss": 0.03493790328502655, "global_step": 244598, "epoch": 2748, "lr": 2.4036523223112452e-06}
{"train_loss": 0.045689087361097336, "global_step": 244599, "epoch": 2748, "lr": 2.4034747376298627e-06}
{"train_loss": 0.054930396378040314, "global_step": 244600, "epoch": 2748, "lr": 2.403297159347251e-06}
{"train_loss": 0.032717011868953705, "global_step": 244601, "epoch": 2748, "lr": 2.403119587463437e-06}
{"train_loss": 0.04668590426445007, "global_step": 244602, "epoch": 2748, "lr": 2.402942021978438e-06}
{"train_loss": 0.033305615186691284, "global_step": 244603, "epoch": 2748, "lr": 2.4027644628922875e-06}
{"train_loss": 0.025006741285324097, "global_step": 244604, "epoch": 2748, "lr": 2.4025869102050013e-06}
{"train_loss": 0.10564030706882477, "global_step": 244605, "epoch": 2748, "lr": 2.4024093639166136e-06}
{"train_loss": 0.0351320281624794, "global_step": 244606, "epoch": 2748, "lr": 2.4022318240271346e-06}
{"train_loss": 0.014171034097671509, "global_step": 244607, "epoch": 2748, "lr": 2.402054290536604e-06}
{"train_loss": 0.05067922919988632, "global_step": 244608, "epoch": 2748, "lr": 2.401876763445038e-06}
{"train_loss": 0.07279128581285477, "global_step": 244609, "epoch": 2748, "lr": 2.4016992427524476e-06}
{"train_loss": 0.027308931574225426, "global_step": 244610, "epoch": 2748, "lr": 2.4015217284588774e-06}
{"train_loss": 0.04778618738055229, "global_step": 244611, "epoch": 2748, "lr": 2.4013442205643388e-06}
{"train_loss": 0.016694029793143272, "global_step": 244612, "epoch": 2748, "lr": 2.4011667190688647e-06}
{"train_loss": 0.06938224285840988, "global_step": 244613, "epoch": 2748, "lr": 2.400989223972466e-06}
{"train_loss": 0.011355566792190075, "global_step": 244614, "epoch": 2748, "lr": 2.4008117352751767e-06}
{"train_loss": 0.03385472297668457, "global_step": 244615, "epoch": 2748, "lr": 2.4006342529770242e-06}
{"train_loss": 0.11475440859794617, "global_step": 244616, "epoch": 2748, "lr": 2.4004567770780197e-06}
{"train_loss": 0.07679249346256256, "global_step": 244617, "epoch": 2748, "lr": 2.4002793075781903e-06}
{"train_loss": 0.0575663186609745, "global_step": 244618, "epoch": 2748, "lr": 2.4001018444775703e-06}
{"train_loss": 0.04640449583530426, "global_step": 244619, "epoch": 2748, "lr": 2.3999243877761703e-06}
{"train_loss": 0.06145621836185455, "global_step": 244620, "epoch": 2748, "lr": 2.399746937474029e-06}
{"train_loss": 0.04462349787354469, "global_step": 244621, "epoch": 2748, "lr": 2.3995694935711523e-06}
{"train_loss": 0.020021801814436913, "global_step": 244622, "epoch": 2748, "lr": 2.3993920560675787e-06}
{"train_loss": 0.02241380140185356, "global_step": 244623, "epoch": 2748, "lr": 2.399214624963325e-06}
{"train_loss": 0.023029150441288948, "global_step": 244624, "epoch": 2748, "lr": 2.3990372002584192e-06}
{"train_loss": 0.040491070598363876, "global_step": 244625, "epoch": 2748, "lr": 2.398859781952878e-06}
{"train_loss": 0.05858025699853897, "global_step": 244626, "epoch": 2748, "lr": 2.3986823700467344e-06}
{"train_loss": 0.028922993689775467, "global_step": 244627, "epoch": 2748, "lr": 2.398504964540005e-06}
{"train_loss": 0.05061415582895279, "global_step": 244628, "epoch": 2748, "lr": 2.3983275654327177e-06}
{"train_loss": 0.04718391224741936, "global_step": 244629, "epoch": 2748, "lr": 2.3981501727248945e-06}
{"train_loss": 0.06292714923620224, "global_step": 244630, "epoch": 2748, "lr": 2.3979727864165635e-06}
{"train_loss": 0.050921693444252014, "global_step": 244631, "epoch": 2748, "lr": 2.397795406507736e-06}
{"train_loss": 0.0319775752723217, "global_step": 244632, "epoch": 2748, "lr": 2.3976180329984552e-06}
{"train_loss": 0.036792606115341187, "global_step": 244633, "epoch": 2748, "lr": 2.3974406658887228e-06}
{"train_loss": 0.07405585050582886, "global_step": 244634, "epoch": 2748, "lr": 2.397263305178582e-06}
{"train_loss": 0.05554914474487305, "global_step": 244635, "epoch": 2748, "lr": 2.39708595086805e-06}
{"train_loss": 0.04361036792397499, "global_step": 244636, "epoch": 2748, "lr": 2.396908602957143e-06}
{"train_loss": 0.04498538747429848, "global_step": 244637, "epoch": 2748, "lr": 2.3967312614459002e-06}
{"train_loss": 0.03416065499186516, "global_step": 244638, "epoch": 2748, "lr": 2.3965539263343273e-06}
{"train_loss": 0.045173950493335724, "global_step": 244639, "epoch": 2748, "lr": 2.396376597622468e-06}
{"train_loss": 0.04071284085512161, "global_step": 244640, "epoch": 2748, "lr": 2.3961992753103237e-06}
{"train_loss": 0.03147460147738457, "global_step": 244641, "epoch": 2748, "lr": 2.3960219593979426e-06}
{"train_loss": 0.07245954126119614, "global_step": 244642, "epoch": 2748, "lr": 2.395844649885326e-06}
{"train_loss": 0.0371103398501873, "global_step": 244643, "epoch": 2748, "lr": 2.395667346772518e-06}
{"train_loss": 0.06397159397602081, "global_step": 244644, "epoch": 2748, "lr": 2.3954900500595234e-06}
{"train_loss": 0.025941263884305954, "global_step": 244645, "epoch": 2748, "lr": 2.395312759746382e-06}
{"train_loss": 0.03171095624566078, "global_step": 244646, "epoch": 2748, "lr": 2.395135475833099e-06}
{"train_loss": 0.05825210362672806, "global_step": 244647, "epoch": 2748, "lr": 2.3949581983197246e-06}
{"train_loss": 0.04311368614435196, "global_step": 244648, "epoch": 2748, "lr": 2.394780927206264e-06}
{"train_loss": 0.07738803327083588, "global_step": 244649, "epoch": 2748, "lr": 2.3946036624927337e-06}
{"train_loss": 0.014416195452213287, "global_step": 244650, "epoch": 2748, "lr": 2.3944264041791843e-06}
{"train_loss": 0.06656090170145035, "global_step": 244651, "epoch": 2748, "lr": 2.3942491522656095e-06}
{"train_loss": 0.02459115907549858, "global_step": 244652, "epoch": 2748, "lr": 2.3940719067520603e-06}
{"train_loss": 0.037855274975299835, "global_step": 244653, "epoch": 2748, "lr": 2.3938946676385353e-06}
{"train_loss": 0.03361867740750313, "global_step": 244654, "epoch": 2748, "lr": 2.3937174349250857e-06}
{"train_loss": 0.03644416481256485, "global_step": 244655, "epoch": 2748, "lr": 2.3935402086117054e-06}
{"train_loss": 0.0552406832575798, "global_step": 244656, "epoch": 2748, "lr": 2.393362988698444e-06}
{"train_loss": 0.06837381422519684, "global_step": 244657, "epoch": 2748, "lr": 2.3931857751853083e-06}
{"train_loss": 0.045642830431461334, "global_step": 244658, "epoch": 2748, "lr": 2.3930085680723357e-06}
{"train_loss": 0.024749455973505974, "global_step": 244659, "epoch": 2748, "lr": 2.3928313673595326e-06}
{"train_loss": 0.04729995101146149, "global_step": 244660, "epoch": 2748, "lr": 2.392654173046943e-06, "val_loss": 9.39398193359375}
{"train_loss": 0.03626646474003792, "global_step": 244661, "epoch": 2749, "lr": 2.392476985134573e-06}
{"train_loss": 0.03827603533864021, "global_step": 244662, "epoch": 2749, "lr": 2.3922998036224665e-06}
{"train_loss": 0.08867192268371582, "global_step": 244663, "epoch": 2749, "lr": 2.3921226285106235e-06}
{"train_loss": 0.03584649786353111, "global_step": 244664, "epoch": 2749, "lr": 2.391945459799083e-06}
{"train_loss": 0.02143930457532406, "global_step": 244665, "epoch": 2749, "lr": 2.391768297487862e-06}
{"train_loss": 0.07105927169322968, "global_step": 244666, "epoch": 2749, "lr": 2.391591141576993e-06}
{"train_loss": 0.06068377569317818, "global_step": 244667, "epoch": 2749, "lr": 2.391413992066488e-06}
{"train_loss": 0.07546072453260422, "global_step": 244668, "epoch": 2749, "lr": 2.3912368489563852e-06}
{"train_loss": 0.027564410120248795, "global_step": 244669, "epoch": 2749, "lr": 2.3910597122466905e-06}
{"train_loss": 0.06655698269605637, "global_step": 244670, "epoch": 2749, "lr": 2.3908825819374424e-06}
{"train_loss": 0.09574691951274872, "global_step": 244671, "epoch": 2749, "lr": 2.3907054580286635e-06}
{"train_loss": 0.051787011325359344, "global_step": 244672, "epoch": 2749, "lr": 2.390528340520365e-06}
{"train_loss": 0.03157874196767807, "global_step": 244673, "epoch": 2749, "lr": 2.3903512294125906e-06}
{"train_loss": 0.09032096713781357, "global_step": 244674, "epoch": 2749, "lr": 2.390174124705341e-06}
{"train_loss": 0.03783731162548065, "global_step": 244675, "epoch": 2749, "lr": 2.389997026398666e-06}
{"train_loss": 0.066494420170784, "global_step": 244676, "epoch": 2749, "lr": 2.3898199344925654e-06}
{"train_loss": 0.049296021461486816, "global_step": 244677, "epoch": 2749, "lr": 2.389642848987078e-06}
{"train_loss": 0.07922414690256119, "global_step": 244678, "epoch": 2749, "lr": 2.3894657698822208e-06}
{"train_loss": 0.03798220306634903, "global_step": 244679, "epoch": 2749, "lr": 2.3892886971780215e-06}
{"train_loss": 0.03510619327425957, "global_step": 244680, "epoch": 2749, "lr": 2.3891116308744965e-06}
{"train_loss": 0.06753239035606384, "global_step": 244681, "epoch": 2749, "lr": 2.388934570971685e-06}
{"train_loss": 0.035745032131671906, "global_step": 244682, "epoch": 2749, "lr": 2.388757517469592e-06}
{"train_loss": 0.03157638758420944, "global_step": 244683, "epoch": 2749, "lr": 2.388580470368257e-06}
{"train_loss": 0.03978617489337921, "global_step": 244684, "epoch": 2749, "lr": 2.3884034296676906e-06}
{"train_loss": 0.03837033733725548, "global_step": 244685, "epoch": 2749, "lr": 2.388226395367926e-06}
{"train_loss": 0.028511416167020798, "global_step": 244686, "epoch": 2749, "lr": 2.388049367468992e-06}
{"train_loss": 0.05311410129070282, "global_step": 244687, "epoch": 2749, "lr": 2.387872345970893e-06}
{"train_loss": 0.051642630249261856, "global_step": 244688, "epoch": 2749, "lr": 2.387695330873668e-06}
{"train_loss": 0.052673663944005966, "global_step": 244689, "epoch": 2749, "lr": 2.3875183221773346e-06}
{"train_loss": 0.0489376075565815, "global_step": 244690, "epoch": 2749, "lr": 2.387341319881925e-06}
{"train_loss": 0.057789456099271774, "global_step": 244691, "epoch": 2749, "lr": 2.3871643239874507e-06}
{"train_loss": 0.028977613896131516, "global_step": 244692, "epoch": 2749, "lr": 2.3869873344939453e-06}
{"train_loss": 0.029504766687750816, "global_step": 244693, "epoch": 2749, "lr": 2.386810351401425e-06}
{"train_loss": 0.04494994133710861, "global_step": 244694, "epoch": 2749, "lr": 2.386633374709929e-06}
{"train_loss": 0.057952094823122025, "global_step": 244695, "epoch": 2749, "lr": 2.3864564044194573e-06}
{"train_loss": 0.040608715265989304, "global_step": 244696, "epoch": 2749, "lr": 2.3862794405300536e-06}
{"train_loss": 0.05808105319738388, "global_step": 244697, "epoch": 2749, "lr": 2.38610248304173e-06}
{"train_loss": 0.06562062352895737, "global_step": 244698, "epoch": 2749, "lr": 2.385925531954519e-06}
{"train_loss": 0.04607211425900459, "global_step": 244699, "epoch": 2749, "lr": 2.385748587268433e-06}
{"train_loss": 0.05942605435848236, "global_step": 244700, "epoch": 2749, "lr": 2.3855716489835087e-06}
{"train_loss": 0.08231008797883987, "global_step": 244701, "epoch": 2749, "lr": 2.385394717099759e-06}
{"train_loss": 0.012987206690013409, "global_step": 244702, "epoch": 2749, "lr": 2.385217791617217e-06}
{"train_loss": 0.0401788093149662, "global_step": 244703, "epoch": 2749, "lr": 2.3850408725358988e-06}
{"train_loss": 0.042684245854616165, "global_step": 244704, "epoch": 2749, "lr": 2.3848639598558375e-06}
{"train_loss": 0.02188638225197792, "global_step": 244705, "epoch": 2749, "lr": 2.3846870535770447e-06}
{"train_loss": 0.050968896597623825, "global_step": 244706, "epoch": 2749, "lr": 2.3845101536995485e-06}
{"train_loss": 0.018892640247941017, "global_step": 244707, "epoch": 2749, "lr": 2.3843332602233814e-06}
{"train_loss": 0.05861242860555649, "global_step": 244708, "epoch": 2749, "lr": 2.384156373148555e-06}
{"train_loss": 0.0955410897731781, "global_step": 244709, "epoch": 2749, "lr": 2.383979492475108e-06}
{"train_loss": 0.03343025594949722, "global_step": 244710, "epoch": 2749, "lr": 2.3838026182030405e-06}
{"train_loss": 0.07100246846675873, "global_step": 244711, "epoch": 2749, "lr": 2.3836257503324024e-06}
{"train_loss": 0.06286793947219849, "global_step": 244712, "epoch": 2749, "lr": 2.383448888863199e-06}
{"train_loss": 0.05324380099773407, "global_step": 244713, "epoch": 2749, "lr": 2.383272033795464e-06}
{"train_loss": 0.0454113744199276, "global_step": 244714, "epoch": 2749, "lr": 2.383095185129214e-06}
{"train_loss": 0.05041174963116646, "global_step": 244715, "epoch": 2749, "lr": 2.3829183428644817e-06}
{"train_loss": 0.03514701873064041, "global_step": 244716, "epoch": 2749, "lr": 2.3827415070012794e-06}
{"train_loss": 0.07916891574859619, "global_step": 244717, "epoch": 2749, "lr": 2.382564677539639e-06}
{"train_loss": 0.04071081429719925, "global_step": 244718, "epoch": 2749, "lr": 2.382387854479584e-06}
{"train_loss": 0.036644790321588516, "global_step": 244719, "epoch": 2749, "lr": 2.382211037821136e-06}
{"train_loss": 0.05578373745083809, "global_step": 244720, "epoch": 2749, "lr": 2.382034227564317e-06}
{"train_loss": 0.051800936460494995, "global_step": 244721, "epoch": 2749, "lr": 2.381857423709161e-06}
{"train_loss": 0.01829603873193264, "global_step": 244722, "epoch": 2749, "lr": 2.3816806262556725e-06}
{"train_loss": 0.027899732813239098, "global_step": 244723, "epoch": 2749, "lr": 2.3815038352038965e-06}
{"train_loss": 0.04923347383737564, "global_step": 244724, "epoch": 2749, "lr": 2.381327050553839e-06}
{"train_loss": 0.04640987515449524, "global_step": 244725, "epoch": 2749, "lr": 2.3811502723055377e-06}
{"train_loss": 0.038585517555475235, "global_step": 244726, "epoch": 2749, "lr": 2.3809735004590162e-06}
{"train_loss": 0.04042087495326996, "global_step": 244727, "epoch": 2749, "lr": 2.380796735014279e-06}
{"train_loss": 0.03520575538277626, "global_step": 244728, "epoch": 2749, "lr": 2.380619975971371e-06}
{"train_loss": 0.031336743384599686, "global_step": 244729, "epoch": 2749, "lr": 2.380443223330303e-06}
{"train_loss": 0.011501460336148739, "global_step": 244730, "epoch": 2749, "lr": 2.3802664770911142e-06}
{"train_loss": 0.04639941453933716, "global_step": 244731, "epoch": 2749, "lr": 2.3800897372538043e-06}
{"train_loss": 0.033865511417388916, "global_step": 244732, "epoch": 2749, "lr": 2.3799130038184237e-06}
{"train_loss": 0.04410287365317345, "global_step": 244733, "epoch": 2749, "lr": 2.379736276784972e-06}
{"train_loss": 0.040846649557352066, "global_step": 244734, "epoch": 2749, "lr": 2.3795595561534933e-06}
{"train_loss": 0.04997503384947777, "global_step": 244735, "epoch": 2749, "lr": 2.379382841923994e-06}
{"train_loss": 0.047490742057561874, "global_step": 244736, "epoch": 2749, "lr": 2.3792061340965177e-06}
{"train_loss": 0.040806058794260025, "global_step": 244737, "epoch": 2749, "lr": 2.379029432671065e-06}
{"train_loss": 0.022883722558617592, "global_step": 244738, "epoch": 2749, "lr": 2.37885273764768e-06}
{"train_loss": 0.03703812509775162, "global_step": 244739, "epoch": 2749, "lr": 2.378676049026368e-06}
{"train_loss": 0.0360126867890358, "global_step": 244740, "epoch": 2749, "lr": 2.3784993668071743e-06}
{"train_loss": 0.022081483155488968, "global_step": 244741, "epoch": 2749, "lr": 2.3783226909901037e-06}
{"train_loss": 0.051084425300359726, "global_step": 244742, "epoch": 2749, "lr": 2.3781460215751893e-06}
{"train_loss": 0.0570061169564724, "global_step": 244743, "epoch": 2749, "lr": 2.377969358562454e-06}
{"train_loss": 0.04785013571381569, "global_step": 244744, "epoch": 2749, "lr": 2.3777927019519143e-06}
{"train_loss": 0.031001118943095207, "global_step": 244745, "epoch": 2749, "lr": 2.3776160517436087e-06}
{"train_loss": 0.02767818234860897, "global_step": 244746, "epoch": 2749, "lr": 2.377439407937543e-06}
{"train_loss": 0.06375601142644882, "global_step": 244747, "epoch": 2749, "lr": 2.3772627705337556e-06}
{"train_loss": 0.014535820111632347, "global_step": 244748, "epoch": 2749, "lr": 2.377086139532264e-06}
{"train_loss": 0.04622243181540725, "global_step": 244749, "epoch": 2749, "lr": 2.3769095149330954e-06, "val_loss": 9.397216796875}
{"train_loss": 0.05133068561553955, "global_step": 244750, "epoch": 2750, "lr": 2.376732896736267e-06}
{"train_loss": 0.04692407697439194, "global_step": 244751, "epoch": 2750, "lr": 2.376556284941811e-06}
{"train_loss": 0.019710835069417953, "global_step": 244752, "epoch": 2750, "lr": 2.3763796795497395e-06}
{"train_loss": 0.01435029599815607, "global_step": 244753, "epoch": 2750, "lr": 2.376203080560091e-06}
{"train_loss": 0.04727557674050331, "global_step": 244754, "epoch": 2750, "lr": 2.376026487972871e-06}
{"train_loss": 0.06416258215904236, "global_step": 244755, "epoch": 2750, "lr": 2.375849901788124e-06}
{"train_loss": 0.015151169151067734, "global_step": 244756, "epoch": 2750, "lr": 2.375673322005856e-06}
{"train_loss": 0.023787179961800575, "global_step": 244757, "epoch": 2750, "lr": 2.3754967486261048e-06}
{"train_loss": 0.04027874767780304, "global_step": 244758, "epoch": 2750, "lr": 2.3753201816488825e-06}
{"train_loss": 0.06563127785921097, "global_step": 244759, "epoch": 2750, "lr": 2.3751436210742273e-06}
{"train_loss": 0.007541906554251909, "global_step": 244760, "epoch": 2750, "lr": 2.3749670669021395e-06}
{"train_loss": 0.020939527079463005, "global_step": 244761, "epoch": 2750, "lr": 2.374790519132669e-06}
{"train_loss": 0.06846842914819717, "global_step": 244762, "epoch": 2750, "lr": 2.3746139777658214e-06}
{"train_loss": 0.030876291915774345, "global_step": 244763, "epoch": 2750, "lr": 2.3744374428016305e-06}
{"train_loss": 0.06695199757814407, "global_step": 244764, "epoch": 2750, "lr": 2.3742609142401175e-06}
{"train_loss": 0.05685190111398697, "global_step": 244765, "epoch": 2750, "lr": 2.374084392081294e-06}
{"train_loss": 0.03723736107349396, "global_step": 244766, "epoch": 2750, "lr": 2.3739078763252044e-06}
{"train_loss": 0.04060434177517891, "global_step": 244767, "epoch": 2750, "lr": 2.373731366971854e-06}
{"train_loss": 0.05624658241868019, "global_step": 244768, "epoch": 2750, "lr": 2.373554864021288e-06}
{"train_loss": 0.046567950397729874, "global_step": 244769, "epoch": 2750, "lr": 2.3733783674735e-06}
{"train_loss": 0.021302662789821625, "global_step": 244770, "epoch": 2750, "lr": 2.373201877328546e-06}
{"train_loss": 0.038874365389347076, "global_step": 244771, "epoch": 2750, "lr": 2.3730253935864256e-06}
{"train_loss": 0.05926224961876869, "global_step": 244772, "epoch": 2750, "lr": 2.3728489162471777e-06}
{"train_loss": 0.03340696915984154, "global_step": 244773, "epoch": 2750, "lr": 2.3726724453108086e-06}
{"train_loss": 0.01635780930519104, "global_step": 244774, "epoch": 2750, "lr": 2.3724959807773674e-06}
{"train_loss": 0.044923797249794006, "global_step": 244775, "epoch": 2750, "lr": 2.372319522646854e-06}
{"train_loss": 0.03632930666208267, "global_step": 244776, "epoch": 2750, "lr": 2.372143070919308e-06}
{"train_loss": 0.03793498873710632, "global_step": 244777, "epoch": 2750, "lr": 2.3719666255947402e-06}
{"train_loss": 0.033340923488140106, "global_step": 244778, "epoch": 2750, "lr": 2.371790186673184e-06}
{"train_loss": 0.043962910771369934, "global_step": 244779, "epoch": 2750, "lr": 2.3716137541546555e-06}
{"train_loss": 0.056777145713567734, "global_step": 244780, "epoch": 2750, "lr": 2.371437328039189e-06}
{"train_loss": 0.0540924035012722, "global_step": 244781, "epoch": 2750, "lr": 2.3712609083268e-06}
{"train_loss": 0.036111798137426376, "global_step": 244782, "epoch": 2750, "lr": 2.371084495017517e-06}
{"train_loss": 0.030858702957630157, "global_step": 244783, "epoch": 2750, "lr": 2.370908088111362e-06}
{"train_loss": 0.06244417652487755, "global_step": 244784, "epoch": 2750, "lr": 2.3707316876083516e-06}
{"train_loss": 0.07589978724718094, "global_step": 244785, "epoch": 2750, "lr": 2.3705552935085253e-06}
{"train_loss": 0.03732801601290703, "global_step": 244786, "epoch": 2750, "lr": 2.3703789058118877e-06}
{"train_loss": 0.07450450211763382, "global_step": 244787, "epoch": 2750, "lr": 2.370202524518478e-06}
{"train_loss": 0.04102536663413048, "global_step": 244788, "epoch": 2750, "lr": 2.370026149628307e-06}
{"train_loss": 0.02637428790330887, "global_step": 244789, "epoch": 2750, "lr": 2.3698497811414145e-06}
{"train_loss": 0.02159266173839569, "global_step": 244790, "epoch": 2750, "lr": 2.3696734190578106e-06}
{"train_loss": 0.06191106140613556, "global_step": 244791, "epoch": 2750, "lr": 2.369497063377524e-06}
{"train_loss": 0.0389864519238472, "global_step": 244792, "epoch": 2750, "lr": 2.3693207141005757e-06}
{"train_loss": 0.05231538787484169, "global_step": 244793, "epoch": 2750, "lr": 2.369144371227e-06}
{"train_loss": 0.08310762792825699, "global_step": 244794, "epoch": 2750, "lr": 2.3689680347568023e-06}
{"train_loss": 0.015498507767915726, "global_step": 244795, "epoch": 2750, "lr": 2.3687917046900266e-06}
{"train_loss": 0.027871496975421906, "global_step": 244796, "epoch": 2750, "lr": 2.368615381026673e-06}
{"train_loss": 0.028880564495921135, "global_step": 244797, "epoch": 2750, "lr": 2.368439063766792e-06}
{"train_loss": 0.02981823869049549, "global_step": 244798, "epoch": 2750, "lr": 2.368262752910383e-06}
{"train_loss": 0.036419112235307693, "global_step": 244799, "epoch": 2750, "lr": 2.3680864484574903e-06}
{"train_loss": 0.01853850856423378, "global_step": 244800, "epoch": 2750, "lr": 2.3679101504081204e-06}
{"train_loss": 0.07095281034708023, "global_step": 244801, "epoch": 2750, "lr": 2.3677338587623055e-06}
{"train_loss": 0.043936993926763535, "global_step": 244802, "epoch": 2750, "lr": 2.3675575735200737e-06}
{"train_loss": 0.04341503232717514, "global_step": 244803, "epoch": 2750, "lr": 2.367381294681437e-06}
{"train_loss": 0.0484447106719017, "global_step": 244804, "epoch": 2750, "lr": 2.3672050222464326e-06}
{"train_loss": 0.04552585259079933, "global_step": 244805, "epoch": 2750, "lr": 2.3670287562150673e-06}
{"train_loss": 0.04302283003926277, "global_step": 244806, "epoch": 2750, "lr": 2.3668524965873794e-06}
{"train_loss": 0.03929682821035385, "global_step": 244807, "epoch": 2750, "lr": 2.366676243363386e-06}
{"train_loss": 0.04657354578375816, "global_step": 244808, "epoch": 2750, "lr": 2.3664999965431147e-06}
{"train_loss": 0.0670393705368042, "global_step": 244809, "epoch": 2750, "lr": 2.3663237561265818e-06}
{"train_loss": 0.06567802280187607, "global_step": 244810, "epoch": 2750, "lr": 2.3661475221138263e-06}
{"train_loss": 0.07059867680072784, "global_step": 244811, "epoch": 2750, "lr": 2.3659712945048483e-06}
{"train_loss": 0.03286093845963478, "global_step": 244812, "epoch": 2750, "lr": 2.365795073299698e-06}
{"train_loss": 0.03706402704119682, "global_step": 244813, "epoch": 2750, "lr": 2.365618858498375e-06}
{"train_loss": 0.02784964069724083, "global_step": 244814, "epoch": 2750, "lr": 2.365442650100924e-06}
{"train_loss": 0.034640394151210785, "global_step": 244815, "epoch": 2750, "lr": 2.3652664481073507e-06}
{"train_loss": 0.016532564535737038, "global_step": 244816, "epoch": 2750, "lr": 2.365090252517682e-06}
{"train_loss": 0.034852828830480576, "global_step": 244817, "epoch": 2750, "lr": 2.3649140633319576e-06}
{"train_loss": 0.0798221006989479, "global_step": 244818, "epoch": 2750, "lr": 2.364737880550183e-06}
{"train_loss": 0.022431615740060806, "global_step": 244819, "epoch": 2750, "lr": 2.364561704172397e-06}
{"train_loss": 0.037543509155511856, "global_step": 244820, "epoch": 2750, "lr": 2.3643855341986098e-06}
{"train_loss": 0.06477219611406326, "global_step": 244821, "epoch": 2750, "lr": 2.3642093706288505e-06}
{"train_loss": 0.05180265009403229, "global_step": 244822, "epoch": 2750, "lr": 2.3640332134631405e-06}
{"train_loss": 0.07977570593357086, "global_step": 244823, "epoch": 2750, "lr": 2.3638570627015134e-06}
{"train_loss": 0.05135175958275795, "global_step": 244824, "epoch": 2750, "lr": 2.3636809183439745e-06}
{"train_loss": 0.034635428339242935, "global_step": 244825, "epoch": 2750, "lr": 2.3635047803905685e-06}
{"train_loss": 0.029100382700562477, "global_step": 244826, "epoch": 2750, "lr": 2.363328648841295e-06}
{"train_loss": 0.0698426365852356, "global_step": 244827, "epoch": 2750, "lr": 2.3631525236962047e-06}
{"train_loss": 0.03990067541599274, "global_step": 244828, "epoch": 2750, "lr": 2.362976404955303e-06}
{"train_loss": 0.053415048867464066, "global_step": 244829, "epoch": 2750, "lr": 2.3628002926186164e-06}
{"train_loss": 0.059096939861774445, "global_step": 244830, "epoch": 2750, "lr": 2.362624186686174e-06}
{"train_loss": 0.027798501774668694, "global_step": 244831, "epoch": 2750, "lr": 2.362448087157998e-06}
{"train_loss": 0.05269549414515495, "global_step": 244832, "epoch": 2750, "lr": 2.362271994034104e-06}
{"train_loss": 0.11855258047580719, "global_step": 244833, "epoch": 2750, "lr": 2.362095907314521e-06}
{"train_loss": 0.025340547785162926, "global_step": 244834, "epoch": 2750, "lr": 2.3619198269992815e-06}
{"train_loss": 0.07138209044933319, "global_step": 244835, "epoch": 2750, "lr": 2.3617437530883968e-06}
{"train_loss": 0.08268667757511139, "global_step": 244836, "epoch": 2750, "lr": 2.3615676855818947e-06}
{"train_loss": 0.025666287168860435, "global_step": 244837, "epoch": 2750, "lr": 2.3613916244798027e-06}
{"train_loss": 0.044910658920037275, "global_step": 244838, "epoch": 2750, "lr": 2.361215569782138e-06, "train/sim_max_reward_0": 0.6147694826164947, "train/sim_max_reward_1": 0.7269724618093156, "train/sim_max_reward_2": 0.6499763281699381, "train/sim_max_reward_3": 0.968564534797431, "train/sim_max_reward_4": 0.9836033563862857, "train/sim_max_reward_5": 0.9542672969916813, "test/sim_max_reward_4500000": 0.46005637301901126, "test/sim_max_reward_4500001": 0.9688591322156633, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.565417634515524, "test/sim_max_reward_4500004": 0.9963923601839156, "test/sim_max_reward_4500005": 0.40324718983648855, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.4725043575685413, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.049359339869625186, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9877050904144056, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 1.0, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9572473363329123, "test/sim_max_reward_4500016": 0.9689359924537814, "test/sim_max_reward_4500017": 0.15137678310035874, "test/sim_max_reward_4500018": 0.9902843077979646, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.983348141607862, "test/sim_max_reward_4500022": 0.9402855488168002, "test/sim_max_reward_4500023": 0.9875350488258203, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 1.0, "test/sim_max_reward_4500027": 0.9444162917182694, "test/sim_max_reward_4500028": 0.6180249979201495, "test/sim_max_reward_4500029": 0.690375377430786, "test/sim_max_reward_4500030": 0.9328482287933176, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.9963314165618983, "test/sim_max_reward_4500033": 0.9272072767266131, "test/sim_max_reward_4500034": 0.4413628620982575, "test/sim_max_reward_4500035": 0.34337548487638336, "test/sim_max_reward_4500036": 0.15459647055837644, "test/sim_max_reward_4500037": 0.9721730047120117, "test/sim_max_reward_4500038": 0.16273009227595167, "test/sim_max_reward_4500039": 0.9966393059262307, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.7027882857936166, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.6222304932124297, "test/sim_max_reward_4500045": 0.9738783660164958, "test/sim_max_reward_4500046": 0.4257373573399312, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.5405105297758048, "train/mean_score": 0.8163589101285244, "test/mean_score": 0.6665556095659039, "val_loss": 9.481996536254883, "train_action_mse_error": 26.857269287109375}
{"train_loss": 0.022433267906308174, "global_step": 244839, "epoch": 2751, "lr": 2.361039521488928e-06}
{"train_loss": 0.04970750957727432, "global_step": 244840, "epoch": 2751, "lr": 2.3608634796001947e-06}
{"train_loss": 0.08226232975721359, "global_step": 244841, "epoch": 2751, "lr": 2.3606874441159667e-06}
{"train_loss": 0.057741135358810425, "global_step": 244842, "epoch": 2751, "lr": 2.3605114150362595e-06}
{"train_loss": 0.05347394570708275, "global_step": 244843, "epoch": 2751, "lr": 2.3603353923611017e-06}
{"train_loss": 0.05412231385707855, "global_step": 244844, "epoch": 2751, "lr": 2.360159376090515e-06}
{"train_loss": 0.05400707945227623, "global_step": 244845, "epoch": 2751, "lr": 2.359983366224522e-06}
{"train_loss": 0.020433461293578148, "global_step": 244846, "epoch": 2751, "lr": 2.3598073627631556e-06}
{"train_loss": 0.042171940207481384, "global_step": 244847, "epoch": 2751, "lr": 2.359631365706427e-06}
{"train_loss": 0.03607404604554176, "global_step": 244848, "epoch": 2751, "lr": 2.3594553750543702e-06}
{"train_loss": 0.05174975469708443, "global_step": 244849, "epoch": 2751, "lr": 2.3592793908069954e-06}
{"train_loss": 0.019212983548641205, "global_step": 244850, "epoch": 2751, "lr": 2.3591034129643475e-06}
{"train_loss": 0.05370912700891495, "global_step": 244851, "epoch": 2751, "lr": 2.358927441526426e-06}
{"train_loss": 0.01872587390244007, "global_step": 244852, "epoch": 2751, "lr": 2.358751476493265e-06}
{"train_loss": 0.036039017140865326, "global_step": 244853, "epoch": 2751, "lr": 2.358575517864897e-06}
{"train_loss": 0.05687025561928749, "global_step": 244854, "epoch": 2751, "lr": 2.3583995656413337e-06}
{"train_loss": 0.051830582320690155, "global_step": 244855, "epoch": 2751, "lr": 2.3582236198226083e-06}
{"train_loss": 0.032359614968299866, "global_step": 244856, "epoch": 2751, "lr": 2.3580476804087314e-06}
{"train_loss": 0.04256264865398407, "global_step": 244857, "epoch": 2751, "lr": 2.3578717473997424e-06}
{"train_loss": 0.08838651329278946, "global_step": 244858, "epoch": 2751, "lr": 2.357695820795647e-06}
{"train_loss": 0.027393799275159836, "global_step": 244859, "epoch": 2751, "lr": 2.3575199005964887e-06}
{"train_loss": 0.05447135493159294, "global_step": 244860, "epoch": 2751, "lr": 2.357343986802274e-06}
{"train_loss": 0.0581086128950119, "global_step": 244861, "epoch": 2751, "lr": 2.3571680794130357e-06}
{"train_loss": 0.14416265487670898, "global_step": 244862, "epoch": 2751, "lr": 2.3569921784287963e-06}
{"train_loss": 0.013420626521110535, "global_step": 244863, "epoch": 2751, "lr": 2.3568162838495833e-06}
{"train_loss": 0.04596216231584549, "global_step": 244864, "epoch": 2751, "lr": 2.356640395675408e-06}
{"train_loss": 0.04789924994111061, "global_step": 244865, "epoch": 2751, "lr": 2.3564645139063034e-06}
{"train_loss": 0.017221735790371895, "global_step": 244866, "epoch": 2751, "lr": 2.3562886385422924e-06}
{"train_loss": 0.06607306748628616, "global_step": 244867, "epoch": 2751, "lr": 2.356112769583402e-06}
{"train_loss": 0.09167724847793579, "global_step": 244868, "epoch": 2751, "lr": 2.3559369070296435e-06}
{"train_loss": 0.030007900670170784, "global_step": 244869, "epoch": 2751, "lr": 2.3557610508810556e-06}
{"train_loss": 0.027533888816833496, "global_step": 244870, "epoch": 2751, "lr": 2.35558520113765e-06}
{"train_loss": 0.03279832750558853, "global_step": 244871, "epoch": 2751, "lr": 2.3554093577994486e-06}
{"train_loss": 0.04971425235271454, "global_step": 244872, "epoch": 2751, "lr": 2.3552335208664956e-06}
{"train_loss": 0.04283710941672325, "global_step": 244873, "epoch": 2751, "lr": 2.355057690338791e-06}
{"train_loss": 0.04301021620631218, "global_step": 244874, "epoch": 2751, "lr": 2.35488186621638e-06}
{"train_loss": 0.04017059504985809, "global_step": 244875, "epoch": 2751, "lr": 2.3547060484992666e-06}
{"train_loss": 0.03302689269185066, "global_step": 244876, "epoch": 2751, "lr": 2.35453023718748e-06}
{"train_loss": 0.02349874936044216, "global_step": 244877, "epoch": 2751, "lr": 2.354354432281053e-06}
{"train_loss": 0.037411488592624664, "global_step": 244878, "epoch": 2751, "lr": 2.3541786337799964e-06}
{"train_loss": 0.04614567384123802, "global_step": 244879, "epoch": 2751, "lr": 2.354002841684344e-06}
{"train_loss": 0.025714406743645668, "global_step": 244880, "epoch": 2751, "lr": 2.353827055994118e-06}
{"train_loss": 0.04356207698583603, "global_step": 244881, "epoch": 2751, "lr": 2.353651276709329e-06}
{"train_loss": 0.0295159500092268, "global_step": 244882, "epoch": 2751, "lr": 2.3534755038300162e-06}
{"train_loss": 0.07508882135152817, "global_step": 244883, "epoch": 2751, "lr": 2.353299737356196e-06}
{"train_loss": 0.024999426677823067, "global_step": 244884, "epoch": 2751, "lr": 2.3531239772878968e-06}
{"train_loss": 0.011747230775654316, "global_step": 244885, "epoch": 2751, "lr": 2.3529482236251345e-06}
{"train_loss": 0.050633952021598816, "global_step": 244886, "epoch": 2751, "lr": 2.352772476367948e-06}
{"train_loss": 0.04781173914670944, "global_step": 244887, "epoch": 2751, "lr": 2.352596735516338e-06}
{"train_loss": 0.06923534721136093, "global_step": 244888, "epoch": 2751, "lr": 2.352421001070343e-06}
{"train_loss": 0.05245232954621315, "global_step": 244889, "epoch": 2751, "lr": 2.3522452730299904e-06}
{"train_loss": 0.06329528987407684, "global_step": 244890, "epoch": 2751, "lr": 2.3520695513952917e-06}
{"train_loss": 0.02978828363120556, "global_step": 244891, "epoch": 2751, "lr": 2.35189383616628e-06}
{"train_loss": 0.025825349614024162, "global_step": 244892, "epoch": 2751, "lr": 2.351718127342972e-06}
{"train_loss": 0.02544018067419529, "global_step": 244893, "epoch": 2751, "lr": 2.3515424249254013e-06}
{"train_loss": 0.03656306117773056, "global_step": 244894, "epoch": 2751, "lr": 2.351366728913579e-06}
{"train_loss": 0.03868513181805611, "global_step": 244895, "epoch": 2751, "lr": 2.351191039307543e-06}
{"train_loss": 0.023907599970698357, "global_step": 244896, "epoch": 2751, "lr": 2.3510153561072943e-06}
{"train_loss": 0.043915700167417526, "global_step": 244897, "epoch": 2751, "lr": 2.350839679312883e-06}
{"train_loss": 0.040202945470809937, "global_step": 244898, "epoch": 2751, "lr": 2.3506640089243138e-06}
{"train_loss": 0.056199707090854645, "global_step": 244899, "epoch": 2751, "lr": 2.350488344941626e-06}
{"train_loss": 0.03354393690824509, "global_step": 244900, "epoch": 2751, "lr": 2.3503126873648253e-06}
{"train_loss": 0.040893372148275375, "global_step": 244901, "epoch": 2751, "lr": 2.35013703619395e-06}
{"train_loss": 0.038837142288684845, "global_step": 244902, "epoch": 2751, "lr": 2.3499613914290118e-06}
{"train_loss": 0.038964271545410156, "global_step": 244903, "epoch": 2751, "lr": 2.349785753070044e-06}
{"train_loss": 0.06535545736551285, "global_step": 244904, "epoch": 2751, "lr": 2.349610121117063e-06}
{"train_loss": 0.03493611514568329, "global_step": 244905, "epoch": 2751, "lr": 2.3494344955701076e-06}
{"train_loss": 0.041444163769483566, "global_step": 244906, "epoch": 2751, "lr": 2.349258876429178e-06}
{"train_loss": 0.039972443133592606, "global_step": 244907, "epoch": 2751, "lr": 2.349083263694307e-06}
{"train_loss": 0.040618203580379486, "global_step": 244908, "epoch": 2751, "lr": 2.3489076573655343e-06}
{"train_loss": 0.02991143800318241, "global_step": 244909, "epoch": 2751, "lr": 2.3487320574428595e-06}
{"train_loss": 0.07152854651212692, "global_step": 244910, "epoch": 2751, "lr": 2.348556463926327e-06}
{"train_loss": 0.03856028988957405, "global_step": 244911, "epoch": 2751, "lr": 2.3483808768159366e-06}
{"train_loss": 0.054062213748693466, "global_step": 244912, "epoch": 2751, "lr": 2.3482052961117384e-06}
{"train_loss": 0.06680496782064438, "global_step": 244913, "epoch": 2751, "lr": 2.3480297218137436e-06}
{"train_loss": 0.03799916058778763, "global_step": 244914, "epoch": 2751, "lr": 2.347854153921963e-06}
{"train_loss": 0.008402890525758266, "global_step": 244915, "epoch": 2751, "lr": 2.347678592436442e-06}
{"train_loss": 0.05497303232550621, "global_step": 244916, "epoch": 2751, "lr": 2.3475030373571903e-06}
{"train_loss": 0.05837836489081383, "global_step": 244917, "epoch": 2751, "lr": 2.3473274886842423e-06}
{"train_loss": 0.04096926748752594, "global_step": 244918, "epoch": 2751, "lr": 2.347151946417614e-06}
{"train_loss": 0.026842163875699043, "global_step": 244919, "epoch": 2751, "lr": 2.3469764105573277e-06}
{"train_loss": 0.036376554518938065, "global_step": 244920, "epoch": 2751, "lr": 2.3468008811034115e-06}
{"train_loss": 0.006612440105527639, "global_step": 244921, "epoch": 2751, "lr": 2.3466253580558817e-06}
{"train_loss": 0.057639557868242264, "global_step": 244922, "epoch": 2751, "lr": 2.3464498414147717e-06}
{"train_loss": 0.08671475201845169, "global_step": 244923, "epoch": 2751, "lr": 2.3462743311800983e-06}
{"train_loss": 0.02483469806611538, "global_step": 244924, "epoch": 2751, "lr": 2.3460988273518837e-06}
{"train_loss": 0.044442642480134964, "global_step": 244925, "epoch": 2751, "lr": 2.3459233299301664e-06}
{"train_loss": 0.030785316601395607, "global_step": 244926, "epoch": 2751, "lr": 2.3457478389149467e-06}
{"train_loss": 0.043764574008501984, "global_step": 244927, "epoch": 2751, "lr": 2.3455723543062692e-06, "val_loss": 9.491292953491211}
{"train_loss": 0.07564771175384521, "global_step": 244928, "epoch": 2752, "lr": 2.3453968761041445e-06}
{"train_loss": 0.044193435460329056, "global_step": 244929, "epoch": 2752, "lr": 2.3452214043086064e-06}
{"train_loss": 0.06586191803216934, "global_step": 244930, "epoch": 2752, "lr": 2.3450459389196656e-06}
{"train_loss": 0.036327656358480453, "global_step": 244931, "epoch": 2752, "lr": 2.3448704799373554e-06}
{"train_loss": 0.05385246500372887, "global_step": 244932, "epoch": 2752, "lr": 2.344695027361693e-06}
{"train_loss": 0.029276417568325996, "global_step": 244933, "epoch": 2752, "lr": 2.3445195811927114e-06}
{"train_loss": 0.053419824689626694, "global_step": 244934, "epoch": 2752, "lr": 2.344344141430421e-06}
{"train_loss": 0.059350814670324326, "global_step": 244935, "epoch": 2752, "lr": 2.344168708074862e-06}
{"train_loss": 0.059319186955690384, "global_step": 244936, "epoch": 2752, "lr": 2.343993281126039e-06}
{"train_loss": 0.024648107588291168, "global_step": 244937, "epoch": 2752, "lr": 2.3438178605839965e-06}
{"train_loss": 0.05516764521598816, "global_step": 244938, "epoch": 2752, "lr": 2.343642446448735e-06}
{"train_loss": 0.061803337186574936, "global_step": 244939, "epoch": 2752, "lr": 2.3434670387202984e-06}
{"train_loss": 0.021715613082051277, "global_step": 244940, "epoch": 2752, "lr": 2.343291637398698e-06}
{"train_loss": 0.018393579870462418, "global_step": 244941, "epoch": 2752, "lr": 2.3431162424839615e-06}
{"train_loss": 0.04888145998120308, "global_step": 244942, "epoch": 2752, "lr": 2.342940853976111e-06}
{"train_loss": 0.06445088982582092, "global_step": 244943, "epoch": 2752, "lr": 2.342765471875169e-06}
{"train_loss": 0.02179388515651226, "global_step": 244944, "epoch": 2752, "lr": 2.342590096181169e-06}
{"train_loss": 0.03847662732005119, "global_step": 244945, "epoch": 2752, "lr": 2.342414726894121e-06}
{"train_loss": 0.03912176564335823, "global_step": 244946, "epoch": 2752, "lr": 2.3422393640140593e-06}
{"train_loss": 0.08125346899032593, "global_step": 244947, "epoch": 2752, "lr": 2.342064007540995e-06}
{"train_loss": 0.04143228754401207, "global_step": 244948, "epoch": 2752, "lr": 2.3418886574749668e-06}
{"train_loss": 0.05261906236410141, "global_step": 244949, "epoch": 2752, "lr": 2.3417133138159852e-06}
{"train_loss": 0.042130015790462494, "global_step": 244950, "epoch": 2752, "lr": 2.34153797656409e-06}
{"train_loss": 0.05180629342794418, "global_step": 244951, "epoch": 2752, "lr": 2.341362645719286e-06}
{"train_loss": 0.044614776968955994, "global_step": 244952, "epoch": 2752, "lr": 2.3411873212816014e-06}
{"train_loss": 0.05404180660843849, "global_step": 244953, "epoch": 2752, "lr": 2.3410120032510696e-06}
{"train_loss": 0.03216078132390976, "global_step": 244954, "epoch": 2752, "lr": 2.340836691627707e-06}
{"train_loss": 0.04717711731791496, "global_step": 244955, "epoch": 2752, "lr": 2.340661386411536e-06}
{"train_loss": 0.09466680884361267, "global_step": 244956, "epoch": 2752, "lr": 2.3404860876025834e-06}
{"train_loss": 0.03366512060165405, "global_step": 244957, "epoch": 2752, "lr": 2.3403107952008727e-06}
{"train_loss": 0.01921524666249752, "global_step": 244958, "epoch": 2752, "lr": 2.340135509206426e-06}
{"train_loss": 0.06918894499540329, "global_step": 244959, "epoch": 2752, "lr": 2.3399602296192646e-06}
{"train_loss": 0.04010246694087982, "global_step": 244960, "epoch": 2752, "lr": 2.3397849564394116e-06}
{"train_loss": 0.05015357583761215, "global_step": 244961, "epoch": 2752, "lr": 2.3396096896668994e-06}
{"train_loss": 0.08142583817243576, "global_step": 244962, "epoch": 2752, "lr": 2.33943442930174e-06}
{"train_loss": 0.05050015449523926, "global_step": 244963, "epoch": 2752, "lr": 2.339259175343972e-06}
{"train_loss": 0.04626239836215973, "global_step": 244964, "epoch": 2752, "lr": 2.3390839277936004e-06}
{"train_loss": 0.030478622764348984, "global_step": 244965, "epoch": 2752, "lr": 2.338908686650665e-06}
{"train_loss": 0.05904300883412361, "global_step": 244966, "epoch": 2752, "lr": 2.3387334519151815e-06}
{"train_loss": 0.03622724488377571, "global_step": 244967, "epoch": 2752, "lr": 2.3385582235871727e-06}
{"train_loss": 0.054120536893606186, "global_step": 244968, "epoch": 2752, "lr": 2.3383830016666607e-06}
{"train_loss": 0.03025488369166851, "global_step": 244969, "epoch": 2752, "lr": 2.338207786153679e-06}
{"train_loss": 0.09239742159843445, "global_step": 244970, "epoch": 2752, "lr": 2.338032577048238e-06}
{"train_loss": 0.03439892828464508, "global_step": 244971, "epoch": 2752, "lr": 2.3378573743503716e-06}
{"train_loss": 0.053598709404468536, "global_step": 244972, "epoch": 2752, "lr": 2.3376821780600966e-06}
{"train_loss": 0.09215162694454193, "global_step": 244973, "epoch": 2752, "lr": 2.3375069881774458e-06}
{"train_loss": 0.051566142588853836, "global_step": 244974, "epoch": 2752, "lr": 2.337331804702425e-06}
{"train_loss": 0.029412202537059784, "global_step": 244975, "epoch": 2752, "lr": 2.3371566276350788e-06}
{"train_loss": 0.04497436434030533, "global_step": 244976, "epoch": 2752, "lr": 2.336981456975412e-06}
{"train_loss": 0.03488311544060707, "global_step": 244977, "epoch": 2752, "lr": 2.3368062927234646e-06}
{"train_loss": 0.03274915739893913, "global_step": 244978, "epoch": 2752, "lr": 2.3366311348792522e-06}
{"train_loss": 0.07931970059871674, "global_step": 244979, "epoch": 2752, "lr": 2.3364559834427925e-06}
{"train_loss": 0.021877329796552658, "global_step": 244980, "epoch": 2752, "lr": 2.3362808384141234e-06}
{"train_loss": 0.04079104959964752, "global_step": 244981, "epoch": 2752, "lr": 2.3361056997932506e-06}
{"train_loss": 0.06381506472826004, "global_step": 244982, "epoch": 2752, "lr": 2.335930567580219e-06}
{"train_loss": 0.05242626368999481, "global_step": 244983, "epoch": 2752, "lr": 2.3357554417750336e-06}
{"train_loss": 0.05099891498684883, "global_step": 244984, "epoch": 2752, "lr": 2.3355803223777285e-06}
{"train_loss": 0.07425104081630707, "global_step": 244985, "epoch": 2752, "lr": 2.3354052093883196e-06}
{"train_loss": 0.04602835327386856, "global_step": 244986, "epoch": 2752, "lr": 2.33523010280684e-06}
{"train_loss": 0.04009837284684181, "global_step": 244987, "epoch": 2752, "lr": 2.335055002633302e-06}
{"train_loss": 0.031522344797849655, "global_step": 244988, "epoch": 2752, "lr": 2.3348799088677377e-06}
{"train_loss": 0.02259971760213375, "global_step": 244989, "epoch": 2752, "lr": 2.334704821510164e-06}
{"train_loss": 0.0077130114659667015, "global_step": 244990, "epoch": 2752, "lr": 2.334529740560615e-06}
{"train_loss": 0.06569366157054901, "global_step": 244991, "epoch": 2752, "lr": 2.3343546660191117e-06}
{"train_loss": 0.06385082751512527, "global_step": 244992, "epoch": 2752, "lr": 2.3341795978856606e-06}
{"train_loss": 0.04547044262290001, "global_step": 244993, "epoch": 2752, "lr": 2.3340045361603056e-06}
{"train_loss": 0.0763784795999527, "global_step": 244994, "epoch": 2752, "lr": 2.3338294808430585e-06}
{"train_loss": 0.023852625861763954, "global_step": 244995, "epoch": 2752, "lr": 2.3336544319339514e-06}
{"train_loss": 0.07116498053073883, "global_step": 244996, "epoch": 2752, "lr": 2.3334793894330078e-06}
{"train_loss": 0.03957477584481239, "global_step": 244997, "epoch": 2752, "lr": 2.3333043533402377e-06}
{"train_loss": 0.07929407060146332, "global_step": 244998, "epoch": 2752, "lr": 2.3331293236556695e-06}
{"train_loss": 0.033912841230630875, "global_step": 244999, "epoch": 2752, "lr": 2.3329543003793418e-06}
{"train_loss": 0.028553863987326622, "global_step": 245000, "epoch": 2752, "lr": 2.3327792835112605e-06}
{"train_loss": 0.013261543586850166, "global_step": 245001, "epoch": 2752, "lr": 2.332604273051464e-06}
{"train_loss": 0.04755764082074165, "global_step": 245002, "epoch": 2752, "lr": 2.332429268999958e-06}
{"train_loss": 0.03346856310963631, "global_step": 245003, "epoch": 2752, "lr": 2.3322542713567876e-06}
{"train_loss": 0.03259069472551346, "global_step": 245004, "epoch": 2752, "lr": 2.3320792801219515e-06}
{"train_loss": 0.05088755860924721, "global_step": 245005, "epoch": 2752, "lr": 2.3319042952955007e-06}
{"train_loss": 0.020652731880545616, "global_step": 245006, "epoch": 2752, "lr": 2.331729316877429e-06}
{"train_loss": 0.037548236548900604, "global_step": 245007, "epoch": 2752, "lr": 2.3315543448677868e-06}
{"train_loss": 0.008096795529127121, "global_step": 245008, "epoch": 2752, "lr": 2.3313793792665796e-06}
{"train_loss": 0.050359100103378296, "global_step": 245009, "epoch": 2752, "lr": 2.3312044200738458e-06}
{"train_loss": 0.029530208557844162, "global_step": 245010, "epoch": 2752, "lr": 2.3310294672895917e-06}
{"train_loss": 0.031979288905858994, "global_step": 245011, "epoch": 2752, "lr": 2.3308545209138556e-06}
{"train_loss": 0.023283259943127632, "global_step": 245012, "epoch": 2752, "lr": 2.3306795809466487e-06}
{"train_loss": 0.02887340635061264, "global_step": 245013, "epoch": 2752, "lr": 2.3305046473880097e-06}
{"train_loss": 0.024508338421583176, "global_step": 245014, "epoch": 2752, "lr": 2.330329720237945e-06}
{"train_loss": 0.05024002119898796, "global_step": 245015, "epoch": 2752, "lr": 2.330154799496487e-06}
{"train_loss": 0.04573858609904399, "global_step": 245016, "epoch": 2752, "lr": 2.329979885163663e-06, "val_loss": 9.477470397949219}
{"train_loss": 0.030041057616472244, "global_step": 245017, "epoch": 2753, "lr": 2.3298049772394857e-06}
{"train_loss": 0.020612254738807678, "global_step": 245018, "epoch": 2753, "lr": 2.329630075723993e-06}
{"train_loss": 0.06332869827747345, "global_step": 245019, "epoch": 2753, "lr": 2.329455180617196e-06}
{"train_loss": 0.05184435844421387, "global_step": 245020, "epoch": 2753, "lr": 2.3292802919191282e-06}
{"train_loss": 0.014123612083494663, "global_step": 245021, "epoch": 2753, "lr": 2.3291054096298002e-06}
{"train_loss": 0.060843512415885925, "global_step": 245022, "epoch": 2753, "lr": 2.328930533749252e-06}
{"train_loss": 0.014777890406548977, "global_step": 245023, "epoch": 2753, "lr": 2.3287556642774877e-06}
{"train_loss": 0.05030125752091408, "global_step": 245024, "epoch": 2753, "lr": 2.3285808012145473e-06}
{"train_loss": 0.040383242070674896, "global_step": 245025, "epoch": 2753, "lr": 2.3284059445604465e-06}
{"train_loss": 0.08216673135757446, "global_step": 245026, "epoch": 2753, "lr": 2.328231094315214e-06}
{"train_loss": 0.024539180099964142, "global_step": 245027, "epoch": 2753, "lr": 2.3280562504788604e-06}
{"train_loss": 0.04287167266011238, "global_step": 245028, "epoch": 2753, "lr": 2.3278814130514306e-06}
{"train_loss": 0.04974592849612236, "global_step": 245029, "epoch": 2753, "lr": 2.327706582032935e-06}
{"train_loss": 0.03798069804906845, "global_step": 245030, "epoch": 2753, "lr": 2.3275317574233844e-06}
{"train_loss": 0.03809177130460739, "global_step": 245031, "epoch": 2753, "lr": 2.32735693922283e-06}
{"train_loss": 0.09583809226751328, "global_step": 245032, "epoch": 2753, "lr": 2.3271821274312766e-06}
{"train_loss": 0.04304905980825424, "global_step": 245033, "epoch": 2753, "lr": 2.3270073220487523e-06}
{"train_loss": 0.05225156620144844, "global_step": 245034, "epoch": 2753, "lr": 2.3268325230752786e-06}
{"train_loss": 0.12214772403240204, "global_step": 245035, "epoch": 2753, "lr": 2.3266577305108895e-06}
{"train_loss": 0.051456011831760406, "global_step": 245036, "epoch": 2753, "lr": 2.3264829443555904e-06}
{"train_loss": 0.06640157103538513, "global_step": 245037, "epoch": 2753, "lr": 2.32630816460942e-06}
{"train_loss": 0.03328407555818558, "global_step": 245038, "epoch": 2753, "lr": 2.326133391272395e-06}
{"train_loss": 0.031564537435770035, "global_step": 245039, "epoch": 2753, "lr": 2.325958624344543e-06}
{"train_loss": 0.021424531936645508, "global_step": 245040, "epoch": 2753, "lr": 2.3257838638258756e-06}
{"train_loss": 0.0527924969792366, "global_step": 245041, "epoch": 2753, "lr": 2.3256091097164367e-06}
{"train_loss": 0.01831309124827385, "global_step": 245042, "epoch": 2753, "lr": 2.3254343620162323e-06}
{"train_loss": 0.032923661172389984, "global_step": 245043, "epoch": 2753, "lr": 2.3252596207252953e-06}
{"train_loss": 0.07591350376605988, "global_step": 245044, "epoch": 2753, "lr": 2.3250848858436424e-06}
{"train_loss": 0.037246257066726685, "global_step": 245045, "epoch": 2753, "lr": 2.3249101573713073e-06}
{"train_loss": 0.09348735958337784, "global_step": 245046, "epoch": 2753, "lr": 2.3247354353082953e-06}
{"train_loss": 0.06525661796331406, "global_step": 245047, "epoch": 2753, "lr": 2.3245607196546557e-06}
{"train_loss": 0.056349195539951324, "global_step": 245048, "epoch": 2753, "lr": 2.324386010410384e-06}
{"train_loss": 0.03700311854481697, "global_step": 245049, "epoch": 2753, "lr": 2.3242113075755303e-06}
{"train_loss": 0.02333398163318634, "global_step": 245050, "epoch": 2753, "lr": 2.324036611150099e-06}
{"train_loss": 0.04458438232541084, "global_step": 245051, "epoch": 2753, "lr": 2.323861921134113e-06}
{"train_loss": 0.07338754832744598, "global_step": 245052, "epoch": 2753, "lr": 2.3236872375276165e-06}
{"train_loss": 0.05675484985113144, "global_step": 245053, "epoch": 2753, "lr": 2.3235125603306098e-06}
{"train_loss": 0.06765714287757874, "global_step": 245054, "epoch": 2753, "lr": 2.323337889543131e-06}
{"train_loss": 0.062445029616355896, "global_step": 245055, "epoch": 2753, "lr": 2.3231632251651926e-06}
{"train_loss": 0.01637228950858116, "global_step": 245056, "epoch": 2753, "lr": 2.3229885671968265e-06}
{"train_loss": 0.046929605305194855, "global_step": 245057, "epoch": 2753, "lr": 2.3228139156380557e-06}
{"train_loss": 0.04388648644089699, "global_step": 245058, "epoch": 2753, "lr": 2.3226392704889023e-06}
{"train_loss": 0.058983348309993744, "global_step": 245059, "epoch": 2753, "lr": 2.3224646317493827e-06}
{"train_loss": 0.05042893812060356, "global_step": 245060, "epoch": 2753, "lr": 2.322289999419536e-06}
{"train_loss": 0.0375019870698452, "global_step": 245061, "epoch": 2753, "lr": 2.322115373499362e-06}
{"train_loss": 0.012409420683979988, "global_step": 245062, "epoch": 2753, "lr": 2.3219407539889106e-06}
{"train_loss": 0.03692479059100151, "global_step": 245063, "epoch": 2753, "lr": 2.3217661408881873e-06}
{"train_loss": 0.059365347027778625, "global_step": 245064, "epoch": 2753, "lr": 2.3215915341972315e-06}
{"train_loss": 0.0209711454808712, "global_step": 245065, "epoch": 2753, "lr": 2.321416933916043e-06}
{"train_loss": 0.013773199170827866, "global_step": 245066, "epoch": 2753, "lr": 2.321242340044666e-06}
{"train_loss": 0.057850904762744904, "global_step": 245067, "epoch": 2753, "lr": 2.321067752583117e-06}
{"train_loss": 0.05692235380411148, "global_step": 245068, "epoch": 2753, "lr": 2.3208931715314187e-06}
{"train_loss": 0.026340976357460022, "global_step": 245069, "epoch": 2753, "lr": 2.3207185968895982e-06}
{"train_loss": 0.03558854013681412, "global_step": 245070, "epoch": 2753, "lr": 2.3205440286576674e-06}
{"train_loss": 0.022915173321962357, "global_step": 245071, "epoch": 2753, "lr": 2.3203694668356703e-06}
{"train_loss": 0.021548617631196976, "global_step": 245072, "epoch": 2753, "lr": 2.3201949114236064e-06}
{"train_loss": 0.02819974534213543, "global_step": 245073, "epoch": 2753, "lr": 2.320020362421521e-06}
{"train_loss": 0.024326294660568237, "global_step": 245074, "epoch": 2753, "lr": 2.3198458198294194e-06}
{"train_loss": 0.05567682534456253, "global_step": 245075, "epoch": 2753, "lr": 2.31967128364734e-06}
{"train_loss": 0.04508157819509506, "global_step": 245076, "epoch": 2753, "lr": 2.3194967538752887e-06}
{"train_loss": 0.05293026566505432, "global_step": 245077, "epoch": 2753, "lr": 2.3193222305133154e-06}
{"train_loss": 0.033893901854753494, "global_step": 245078, "epoch": 2753, "lr": 2.319147713561415e-06}
{"train_loss": 0.051756154745817184, "global_step": 245079, "epoch": 2753, "lr": 2.318973203019631e-06}
{"train_loss": 0.02297946624457836, "global_step": 245080, "epoch": 2753, "lr": 2.318798698887975e-06}
{"train_loss": 0.03278809413313866, "global_step": 245081, "epoch": 2753, "lr": 2.318624201166486e-06}
{"train_loss": 0.018306240439414978, "global_step": 245082, "epoch": 2753, "lr": 2.318449709855164e-06}
{"train_loss": 0.03634122759103775, "global_step": 245083, "epoch": 2753, "lr": 2.318275224954053e-06}
{"train_loss": 0.05510581284761429, "global_step": 245084, "epoch": 2753, "lr": 2.3181007464631642e-06}
{"train_loss": 0.05086156353354454, "global_step": 245085, "epoch": 2753, "lr": 2.3179262743825316e-06}
{"train_loss": 0.05435828119516373, "global_step": 245086, "epoch": 2753, "lr": 2.317751808712165e-06}
{"train_loss": 0.042826712131500244, "global_step": 245087, "epoch": 2753, "lr": 2.317577349452099e-06}
{"train_loss": 0.03196648880839348, "global_step": 245088, "epoch": 2753, "lr": 2.317402896602355e-06}
{"train_loss": 0.02936847321689129, "global_step": 245089, "epoch": 2753, "lr": 2.3172284501629506e-06}
{"train_loss": 0.041782502084970474, "global_step": 245090, "epoch": 2753, "lr": 2.3170540101339177e-06}
{"train_loss": 0.02013557404279709, "global_step": 245091, "epoch": 2753, "lr": 2.316879576515274e-06}
{"train_loss": 0.05136658996343613, "global_step": 245092, "epoch": 2753, "lr": 2.316705149307047e-06}
{"train_loss": 0.044478803873062134, "global_step": 245093, "epoch": 2753, "lr": 2.316530728509253e-06}
{"train_loss": 0.027958355844020844, "global_step": 245094, "epoch": 2753, "lr": 2.3163563141219313e-06}
{"train_loss": 0.03497544676065445, "global_step": 245095, "epoch": 2753, "lr": 2.3161819061450817e-06}
{"train_loss": 0.04049715772271156, "global_step": 245096, "epoch": 2753, "lr": 2.316007504578749e-06}
{"train_loss": 0.050316840410232544, "global_step": 245097, "epoch": 2753, "lr": 2.3158331094229435e-06}
{"train_loss": 0.05814312398433685, "global_step": 245098, "epoch": 2753, "lr": 2.3156587206776936e-06}
{"train_loss": 0.022917615249753, "global_step": 245099, "epoch": 2753, "lr": 2.3154843383430215e-06}
{"train_loss": 0.03160368651151657, "global_step": 245100, "epoch": 2753, "lr": 2.3153099624189544e-06}
{"train_loss": 0.02771644853055477, "global_step": 245101, "epoch": 2753, "lr": 2.3151355929055096e-06}
{"train_loss": 0.04375576600432396, "global_step": 245102, "epoch": 2753, "lr": 2.3149612298027146e-06}
{"train_loss": 0.058238670229911804, "global_step": 245103, "epoch": 2753, "lr": 2.3147868731105916e-06}
{"train_loss": 0.040286097675561905, "global_step": 245104, "epoch": 2753, "lr": 2.3146125228291683e-06}
{"train_loss": 0.04368901780147231, "global_step": 245105, "epoch": 2753, "lr": 2.314438178958456e-06, "val_loss": 9.448843002319336}
{"train_loss": 0.028800254687666893, "global_step": 245106, "epoch": 2754, "lr": 2.314263841498493e-06}
{"train_loss": 0.026061033830046654, "global_step": 245107, "epoch": 2754, "lr": 2.3140895104492966e-06}
{"train_loss": 0.07560577243566513, "global_step": 245108, "epoch": 2754, "lr": 2.313915185810883e-06}
{"train_loss": 0.02459143102169037, "global_step": 245109, "epoch": 2754, "lr": 2.3137408675832915e-06}
{"train_loss": 0.02974814362823963, "global_step": 245110, "epoch": 2754, "lr": 2.3135665557665276e-06}
{"train_loss": 0.09329110383987427, "global_step": 245111, "epoch": 2754, "lr": 2.31339225036063e-06}
{"train_loss": 0.040038466453552246, "global_step": 245112, "epoch": 2754, "lr": 2.3132179513656092e-06}
{"train_loss": 0.07714065164327621, "global_step": 245113, "epoch": 2754, "lr": 2.3130436587814996e-06}
{"train_loss": 0.026436658576130867, "global_step": 245114, "epoch": 2754, "lr": 2.3128693726083172e-06}
{"train_loss": 0.058786071836948395, "global_step": 245115, "epoch": 2754, "lr": 2.31269509284609e-06}
{"train_loss": 0.04385976865887642, "global_step": 245116, "epoch": 2754, "lr": 2.3125208194948347e-06}
{"train_loss": 0.017098911106586456, "global_step": 245117, "epoch": 2754, "lr": 2.3123465525545895e-06}
{"train_loss": 0.03316359594464302, "global_step": 245118, "epoch": 2754, "lr": 2.312172292025361e-06}
{"train_loss": 0.05220894142985344, "global_step": 245119, "epoch": 2754, "lr": 2.311998037907187e-06}
{"train_loss": 0.03658631816506386, "global_step": 245120, "epoch": 2754, "lr": 2.311823790200074e-06}
{"train_loss": 0.03276238590478897, "global_step": 245121, "epoch": 2754, "lr": 2.31164954890406e-06}
{"train_loss": 0.059816114604473114, "global_step": 245122, "epoch": 2754, "lr": 2.3114753140191627e-06}
{"train_loss": 0.04802185669541359, "global_step": 245123, "epoch": 2754, "lr": 2.3113010855454086e-06}
{"train_loss": 0.023702621459960938, "global_step": 245124, "epoch": 2754, "lr": 2.3111268634828155e-06}
{"train_loss": 0.025541173294186592, "global_step": 245125, "epoch": 2754, "lr": 2.3109526478314046e-06}
{"train_loss": 0.017215672880411148, "global_step": 245126, "epoch": 2754, "lr": 2.3107784385912158e-06}
{"train_loss": 0.051476821303367615, "global_step": 245127, "epoch": 2754, "lr": 2.3106042357622536e-06}
{"train_loss": 0.03384135663509369, "global_step": 245128, "epoch": 2754, "lr": 2.3104300393445575e-06}
{"train_loss": 0.06253369152545929, "global_step": 245129, "epoch": 2754, "lr": 2.3102558493381334e-06}
{"train_loss": 0.0537821501493454, "global_step": 245130, "epoch": 2754, "lr": 2.3100816657430245e-06}
{"train_loss": 0.03038773499429226, "global_step": 245131, "epoch": 2754, "lr": 2.3099074885592376e-06}
{"train_loss": 0.012081039138138294, "global_step": 245132, "epoch": 2754, "lr": 2.3097333177868052e-06}
{"train_loss": 0.03259333223104477, "global_step": 245133, "epoch": 2754, "lr": 2.3095591534257444e-06}
{"train_loss": 0.060337189584970474, "global_step": 245134, "epoch": 2754, "lr": 2.3093849954760827e-06}
{"train_loss": 0.025557484477758408, "global_step": 245135, "epoch": 2754, "lr": 2.3092108439378424e-06}
{"train_loss": 0.11756536364555359, "global_step": 245136, "epoch": 2754, "lr": 2.309036698811051e-06}
{"train_loss": 0.02600795403122902, "global_step": 245137, "epoch": 2754, "lr": 2.308862560095726e-06}
{"train_loss": 0.09603092819452286, "global_step": 245138, "epoch": 2754, "lr": 2.308688427791894e-06}
{"train_loss": 0.03670446574687958, "global_step": 245139, "epoch": 2754, "lr": 2.308514301899578e-06}
{"train_loss": 0.05740800127387047, "global_step": 245140, "epoch": 2754, "lr": 2.3083401824188e-06}
{"train_loss": 0.027339279651641846, "global_step": 245141, "epoch": 2754, "lr": 2.308166069349582e-06}
{"train_loss": 0.06631585210561752, "global_step": 245142, "epoch": 2754, "lr": 2.3079919626919578e-06}
{"train_loss": 0.04980132356286049, "global_step": 245143, "epoch": 2754, "lr": 2.3078178624459323e-06}
{"train_loss": 0.016564255580306053, "global_step": 245144, "epoch": 2754, "lr": 2.3076437686115448e-06}
{"train_loss": 0.074913389980793, "global_step": 245145, "epoch": 2754, "lr": 2.3074696811888176e-06}
{"train_loss": 0.06325659900903702, "global_step": 245146, "epoch": 2754, "lr": 2.3072956001777613e-06}
{"train_loss": 0.028174499049782753, "global_step": 245147, "epoch": 2754, "lr": 2.307121525578415e-06}
{"train_loss": 0.03956415876746178, "global_step": 245148, "epoch": 2754, "lr": 2.30694745739079e-06}
{"train_loss": 0.056051768362522125, "global_step": 245149, "epoch": 2754, "lr": 2.3067733956149195e-06}
{"train_loss": 0.048445940017700195, "global_step": 245150, "epoch": 2754, "lr": 2.3065993402508147e-06}
{"train_loss": 0.0636163055896759, "global_step": 245151, "epoch": 2754, "lr": 2.306425291298514e-06}
{"train_loss": 0.06232699751853943, "global_step": 245152, "epoch": 2754, "lr": 2.306251248758029e-06}
{"train_loss": 0.03235868737101555, "global_step": 245153, "epoch": 2754, "lr": 2.306077212629387e-06}
{"train_loss": 0.02638772502541542, "global_step": 245154, "epoch": 2754, "lr": 2.305903182912611e-06}
{"train_loss": 0.021045586094260216, "global_step": 245155, "epoch": 2754, "lr": 2.305729159607728e-06}
{"train_loss": 0.03446488454937935, "global_step": 245156, "epoch": 2754, "lr": 2.305555142714755e-06}
{"train_loss": 0.03398754447698593, "global_step": 245157, "epoch": 2754, "lr": 2.305381132233725e-06}
{"train_loss": 0.04534025117754936, "global_step": 245158, "epoch": 2754, "lr": 2.3052071281646437e-06}
{"train_loss": 0.0596906840801239, "global_step": 245159, "epoch": 2754, "lr": 2.3050331305075558e-06}
{"train_loss": 0.04931730404496193, "global_step": 245160, "epoch": 2754, "lr": 2.304859139262472e-06}
{"train_loss": 0.06518496572971344, "global_step": 245161, "epoch": 2754, "lr": 2.3046851544294147e-06}
{"train_loss": 0.0597054585814476, "global_step": 245162, "epoch": 2754, "lr": 2.304511176008417e-06}
{"train_loss": 0.04887128248810768, "global_step": 245163, "epoch": 2754, "lr": 2.3043372039994905e-06}
{"train_loss": 0.032748498022556305, "global_step": 245164, "epoch": 2754, "lr": 2.3041632384026733e-06}
{"train_loss": 0.06824144721031189, "global_step": 245165, "epoch": 2754, "lr": 2.3039892792179717e-06}
{"train_loss": 0.05352851748466492, "global_step": 245166, "epoch": 2754, "lr": 2.3038153264454243e-06}
{"train_loss": 0.057171642780303955, "global_step": 245167, "epoch": 2754, "lr": 2.303641380085042e-06}
{"train_loss": 0.04726795479655266, "global_step": 245168, "epoch": 2754, "lr": 2.303467440136858e-06}
{"train_loss": 0.023248357698321342, "global_step": 245169, "epoch": 2754, "lr": 2.3032935066008843e-06}
{"train_loss": 0.03658033162355423, "global_step": 245170, "epoch": 2754, "lr": 2.303119579477164e-06}
{"train_loss": 0.051028747111558914, "global_step": 245171, "epoch": 2754, "lr": 2.3029456587656986e-06}
{"train_loss": 0.040136780589818954, "global_step": 245172, "epoch": 2754, "lr": 2.302771744466525e-06}
{"train_loss": 0.045291345566511154, "global_step": 245173, "epoch": 2754, "lr": 2.3025978365796564e-06}
{"train_loss": 0.061523012816905975, "global_step": 245174, "epoch": 2754, "lr": 2.30242393510513e-06}
{"train_loss": 0.043856970965862274, "global_step": 245175, "epoch": 2754, "lr": 2.3022500400429526e-06}
{"train_loss": 0.008483318611979485, "global_step": 245176, "epoch": 2754, "lr": 2.302076151393162e-06}
{"train_loss": 0.0553712360560894, "global_step": 245177, "epoch": 2754, "lr": 2.3019022691557757e-06}
{"train_loss": 0.034714408218860626, "global_step": 245178, "epoch": 2754, "lr": 2.3017283933308154e-06}
{"train_loss": 0.01891360431909561, "global_step": 245179, "epoch": 2754, "lr": 2.3015545239183036e-06}
{"train_loss": 0.03653165325522423, "global_step": 245180, "epoch": 2754, "lr": 2.301380660918279e-06}
{"train_loss": 0.057909540832042694, "global_step": 245181, "epoch": 2754, "lr": 2.3012068043307357e-06}
{"train_loss": 0.04309779405593872, "global_step": 245182, "epoch": 2754, "lr": 2.301032954155724e-06}
{"train_loss": 0.04941343143582344, "global_step": 245183, "epoch": 2754, "lr": 2.3008591103932553e-06}
{"train_loss": 0.014475670643150806, "global_step": 245184, "epoch": 2754, "lr": 2.300685273043357e-06}
{"train_loss": 0.04518963769078255, "global_step": 245185, "epoch": 2754, "lr": 2.3005114421060515e-06}
{"train_loss": 0.03502388298511505, "global_step": 245186, "epoch": 2754, "lr": 2.3003376175813494e-06}
{"train_loss": 0.03918786346912384, "global_step": 245187, "epoch": 2754, "lr": 2.3001637994693017e-06}
{"train_loss": 0.023601001128554344, "global_step": 245188, "epoch": 2754, "lr": 2.2999899877699015e-06}
{"train_loss": 0.03432898968458176, "global_step": 245189, "epoch": 2754, "lr": 2.299816182483194e-06}
{"train_loss": 0.05409502610564232, "global_step": 245190, "epoch": 2754, "lr": 2.299642383609191e-06}
{"train_loss": 0.061326757073402405, "global_step": 245191, "epoch": 2754, "lr": 2.299468591147924e-06}
{"train_loss": 0.0511159785091877, "global_step": 245192, "epoch": 2754, "lr": 2.2992948050994055e-06}
{"train_loss": 0.02084553986787796, "global_step": 245193, "epoch": 2754, "lr": 2.299121025463674e-06}
{"train_loss": 0.0439241442075941, "global_step": 245194, "epoch": 2754, "lr": 2.2989472522407352e-06, "val_loss": 9.46588134765625}
{"train_loss": 0.056863803416490555, "global_step": 245195, "epoch": 2755, "lr": 2.2987734854306275e-06}
{"train_loss": 0.027780456468462944, "global_step": 245196, "epoch": 2755, "lr": 2.2985997250333626e-06}
{"train_loss": 0.08329121768474579, "global_step": 245197, "epoch": 2755, "lr": 2.298425971048973e-06}
{"train_loss": 0.07279935479164124, "global_step": 245198, "epoch": 2755, "lr": 2.298252223477482e-06}
{"train_loss": 0.06338804960250854, "global_step": 245199, "epoch": 2755, "lr": 2.2980784823188996e-06}
{"train_loss": 0.0568900927901268, "global_step": 245200, "epoch": 2755, "lr": 2.2979047475732706e-06}
{"train_loss": 0.056151892989873886, "global_step": 245201, "epoch": 2755, "lr": 2.297731019240601e-06}
{"train_loss": 0.03253087028861046, "global_step": 245202, "epoch": 2755, "lr": 2.2975572973209237e-06}
{"train_loss": 0.027017781510949135, "global_step": 245203, "epoch": 2755, "lr": 2.297383581814255e-06}
{"train_loss": 0.027192655950784683, "global_step": 245204, "epoch": 2755, "lr": 2.297209872720624e-06}
{"train_loss": 0.012627989985048771, "global_step": 245205, "epoch": 2755, "lr": 2.2970361700400455e-06}
{"train_loss": 0.02830079197883606, "global_step": 245206, "epoch": 2755, "lr": 2.29686247377256e-06}
{"train_loss": 0.02666308544576168, "global_step": 245207, "epoch": 2755, "lr": 2.2966887839181715e-06}
{"train_loss": 0.07727965712547302, "global_step": 245208, "epoch": 2755, "lr": 2.296515100476915e-06}
{"train_loss": 0.052534203976392746, "global_step": 245209, "epoch": 2755, "lr": 2.2963414234488114e-06}
{"train_loss": 0.05488673970103264, "global_step": 245210, "epoch": 2755, "lr": 2.2961677528338832e-06}
{"train_loss": 0.0629214495420456, "global_step": 245211, "epoch": 2755, "lr": 2.2959940886321473e-06}
{"train_loss": 0.07608190178871155, "global_step": 245212, "epoch": 2755, "lr": 2.2958204308436426e-06}
{"train_loss": 0.05339249595999718, "global_step": 245213, "epoch": 2755, "lr": 2.29564677946838e-06}
{"train_loss": 0.03237142041325569, "global_step": 245214, "epoch": 2755, "lr": 2.2954731345063817e-06}
{"train_loss": 0.022794248536229134, "global_step": 245215, "epoch": 2755, "lr": 2.295299495957681e-06}
{"train_loss": 0.03265925124287605, "global_step": 245216, "epoch": 2755, "lr": 2.2951258638222896e-06}
{"train_loss": 0.028847577050328255, "global_step": 245217, "epoch": 2755, "lr": 2.2949522381002453e-06}
{"train_loss": 0.052041079849004745, "global_step": 245218, "epoch": 2755, "lr": 2.2947786187915656e-06}
{"train_loss": 0.075821653008461, "global_step": 245219, "epoch": 2755, "lr": 2.294605005896261e-06}
{"train_loss": 0.04246192425489426, "global_step": 245220, "epoch": 2755, "lr": 2.2944313994143762e-06}
{"train_loss": 0.06539459526538849, "global_step": 245221, "epoch": 2755, "lr": 2.2942577993459114e-06}
{"train_loss": 0.0609298013150692, "global_step": 245222, "epoch": 2755, "lr": 2.2940842056909163e-06}
{"train_loss": 0.0253303200006485, "global_step": 245223, "epoch": 2755, "lr": 2.2939106184493962e-06}
{"train_loss": 0.036400653421878815, "global_step": 245224, "epoch": 2755, "lr": 2.2937370376213683e-06}
{"train_loss": 0.03483540937304497, "global_step": 245225, "epoch": 2755, "lr": 2.293563463206877e-06}
{"train_loss": 0.06748587638139725, "global_step": 245226, "epoch": 2755, "lr": 2.2933898952059275e-06}
{"train_loss": 0.04641687870025635, "global_step": 245227, "epoch": 2755, "lr": 2.293216333618553e-06}
{"train_loss": 0.048875246196985245, "global_step": 245228, "epoch": 2755, "lr": 2.2930427784447705e-06}
{"train_loss": 0.07652798295021057, "global_step": 245229, "epoch": 2755, "lr": 2.2928692296846133e-06}
{"train_loss": 0.0446888767182827, "global_step": 245230, "epoch": 2755, "lr": 2.292695687338092e-06}
{"train_loss": 0.03578284755349159, "global_step": 245231, "epoch": 2755, "lr": 2.292522151405241e-06}
{"train_loss": 0.047738123685121536, "global_step": 245232, "epoch": 2755, "lr": 2.292348621886076e-06}
{"train_loss": 0.04105205088853836, "global_step": 245233, "epoch": 2755, "lr": 2.292175098780619e-06}
{"train_loss": 0.08648683875799179, "global_step": 245234, "epoch": 2755, "lr": 2.2920015820889095e-06}
{"train_loss": 0.05728844553232193, "global_step": 245235, "epoch": 2755, "lr": 2.2918280718109473e-06}
{"train_loss": 0.05830661952495575, "global_step": 245236, "epoch": 2755, "lr": 2.291654567946777e-06}
{"train_loss": 0.026891864836215973, "global_step": 245237, "epoch": 2755, "lr": 2.2914810704964042e-06}
{"train_loss": 0.020613765344023705, "global_step": 245238, "epoch": 2755, "lr": 2.291307579459867e-06}
{"train_loss": 0.09485917538404465, "global_step": 245239, "epoch": 2755, "lr": 2.2911340948371775e-06}
{"train_loss": 0.025546103715896606, "global_step": 245240, "epoch": 2755, "lr": 2.290960616628368e-06}
{"train_loss": 0.05052277073264122, "global_step": 245241, "epoch": 2755, "lr": 2.2907871448334505e-06}
{"train_loss": 0.023591605946421623, "global_step": 245242, "epoch": 2755, "lr": 2.2906136794524637e-06}
{"train_loss": 0.0542764775454998, "global_step": 245243, "epoch": 2755, "lr": 2.2904402204854125e-06}
{"train_loss": 0.022580498829483986, "global_step": 245244, "epoch": 2755, "lr": 2.290266767932342e-06}
{"train_loss": 0.05458603799343109, "global_step": 245245, "epoch": 2755, "lr": 2.290093321793252e-06}
{"train_loss": 0.029648717492818832, "global_step": 245246, "epoch": 2755, "lr": 2.289919882068187e-06}
{"train_loss": 0.04256677255034447, "global_step": 245247, "epoch": 2755, "lr": 2.289746448757152e-06}
{"train_loss": 0.021453548222780228, "global_step": 245248, "epoch": 2755, "lr": 2.2895730218601864e-06}
{"train_loss": 0.029280712828040123, "global_step": 245249, "epoch": 2755, "lr": 2.2893996013773013e-06}
{"train_loss": 0.050787292420864105, "global_step": 245250, "epoch": 2755, "lr": 2.2892261873085295e-06}
{"train_loss": 0.03412162885069847, "global_step": 245251, "epoch": 2755, "lr": 2.289052779653883e-06}
{"train_loss": 0.053388699889183044, "global_step": 245252, "epoch": 2755, "lr": 2.2888793784133887e-06}
{"train_loss": 0.03349525108933449, "global_step": 245253, "epoch": 2755, "lr": 2.288705983587086e-06}
{"train_loss": 0.04645223915576935, "global_step": 245254, "epoch": 2755, "lr": 2.288532595174975e-06}
{"train_loss": 0.09609007835388184, "global_step": 245255, "epoch": 2755, "lr": 2.288359213177099e-06}
{"train_loss": 0.042029254138469696, "global_step": 245256, "epoch": 2755, "lr": 2.2881858375934704e-06}
{"train_loss": 0.027805132791399956, "global_step": 245257, "epoch": 2755, "lr": 2.2880124684241054e-06}
{"train_loss": 0.05300632491707802, "global_step": 245258, "epoch": 2755, "lr": 2.2878391056690428e-06}
{"train_loss": 0.033243030309677124, "global_step": 245259, "epoch": 2755, "lr": 2.2876657493282938e-06}
{"train_loss": 0.017458437010645866, "global_step": 245260, "epoch": 2755, "lr": 2.2874923994018915e-06}
{"train_loss": 0.06437912583351135, "global_step": 245261, "epoch": 2755, "lr": 2.287319055889853e-06}
{"train_loss": 0.030255407094955444, "global_step": 245262, "epoch": 2755, "lr": 2.2871457187921995e-06}
{"train_loss": 0.030388010665774345, "global_step": 245263, "epoch": 2755, "lr": 2.28697238810896e-06}
{"train_loss": 0.02441929094493389, "global_step": 245264, "epoch": 2755, "lr": 2.2867990638401503e-06}
{"train_loss": 0.025077663362026215, "global_step": 245265, "epoch": 2755, "lr": 2.28662574598581e-06}
{"train_loss": 0.03531292825937271, "global_step": 245266, "epoch": 2755, "lr": 2.2864524345459382e-06}
{"train_loss": 0.024720540270209312, "global_step": 245267, "epoch": 2755, "lr": 2.2862791295205856e-06}
{"train_loss": 0.03634769842028618, "global_step": 245268, "epoch": 2755, "lr": 2.286105830909746e-06}
{"train_loss": 0.0470118448138237, "global_step": 245269, "epoch": 2755, "lr": 2.2859325387134644e-06}
{"train_loss": 0.03219769522547722, "global_step": 245270, "epoch": 2755, "lr": 2.2857592529317627e-06}
{"train_loss": 0.03998709097504616, "global_step": 245271, "epoch": 2755, "lr": 2.2855859735646525e-06}
{"train_loss": 0.07424023747444153, "global_step": 245272, "epoch": 2755, "lr": 2.285412700612166e-06}
{"train_loss": 0.07365518808364868, "global_step": 245273, "epoch": 2755, "lr": 2.2852394340743266e-06}
{"train_loss": 0.06464405357837677, "global_step": 245274, "epoch": 2755, "lr": 2.2850661739511557e-06}
{"train_loss": 0.08104424923658371, "global_step": 245275, "epoch": 2755, "lr": 2.2848929202426704e-06}
{"train_loss": 0.06522926688194275, "global_step": 245276, "epoch": 2755, "lr": 2.284719672948904e-06}
{"train_loss": 0.020052464678883553, "global_step": 245277, "epoch": 2755, "lr": 2.2845464320698728e-06}
{"train_loss": 0.04635915905237198, "global_step": 245278, "epoch": 2755, "lr": 2.284373197605605e-06}
{"train_loss": 0.011668629013001919, "global_step": 245279, "epoch": 2755, "lr": 2.284199969556117e-06}
{"train_loss": 0.01837904006242752, "global_step": 245280, "epoch": 2755, "lr": 2.284026747921447e-06}
{"train_loss": 0.04569272696971893, "global_step": 245281, "epoch": 2755, "lr": 2.2838535327015964e-06}
{"train_loss": 0.027853287756443024, "global_step": 245282, "epoch": 2755, "lr": 2.2836803238966087e-06}
{"train_loss": 0.04483662053858966, "global_step": 245283, "epoch": 2755, "lr": 2.2835071215064953e-06, "val_loss": 9.44279670715332, "train_action_mse_error": 11.86993408203125}
{"train_loss": 0.028503956273198128, "global_step": 245284, "epoch": 2756, "lr": 2.2833339255312836e-06}
{"train_loss": 0.02897215075790882, "global_step": 245285, "epoch": 2756, "lr": 2.2831607359709907e-06}
{"train_loss": 0.04292023926973343, "global_step": 245286, "epoch": 2756, "lr": 2.282987552825655e-06}
{"train_loss": 0.04413043335080147, "global_step": 245287, "epoch": 2756, "lr": 2.2828143760952823e-06}
{"train_loss": 0.020660456269979477, "global_step": 245288, "epoch": 2756, "lr": 2.2826412057799008e-06}
{"train_loss": 0.0821039229631424, "global_step": 245289, "epoch": 2756, "lr": 2.282468041879543e-06}
{"train_loss": 0.02900102362036705, "global_step": 245290, "epoch": 2756, "lr": 2.282294884394226e-06}
{"train_loss": 0.009386523626744747, "global_step": 245291, "epoch": 2756, "lr": 2.282121733323972e-06}
{"train_loss": 0.07288315147161484, "global_step": 245292, "epoch": 2756, "lr": 2.281948588668803e-06}
{"train_loss": 0.06049487739801407, "global_step": 245293, "epoch": 2756, "lr": 2.281775450428747e-06}
{"train_loss": 0.04764349013566971, "global_step": 245294, "epoch": 2756, "lr": 2.2816023186038316e-06}
{"train_loss": 0.02450665645301342, "global_step": 245295, "epoch": 2756, "lr": 2.2814291931940567e-06}
{"train_loss": 0.06521739065647125, "global_step": 245296, "epoch": 2756, "lr": 2.281256074199478e-06}
{"train_loss": 0.025833502411842346, "global_step": 245297, "epoch": 2756, "lr": 2.2810829616200904e-06}
{"train_loss": 0.029100125655531883, "global_step": 245298, "epoch": 2756, "lr": 2.280909855455937e-06}
{"train_loss": 0.011163599789142609, "global_step": 245299, "epoch": 2756, "lr": 2.2807367557070303e-06}
{"train_loss": 0.037208035588264465, "global_step": 245300, "epoch": 2756, "lr": 2.280563662373403e-06}
{"train_loss": 0.08310659229755402, "global_step": 245301, "epoch": 2756, "lr": 2.280390575455066e-06}
{"train_loss": 0.0747775211930275, "global_step": 245302, "epoch": 2756, "lr": 2.280217494952047e-06}
{"train_loss": 0.042692456394433975, "global_step": 245303, "epoch": 2756, "lr": 2.28004442086438e-06}
{"train_loss": 0.05667725205421448, "global_step": 245304, "epoch": 2756, "lr": 2.27987135319207e-06}
{"train_loss": 0.06857747584581375, "global_step": 245305, "epoch": 2756, "lr": 2.2796982919351504e-06}
{"train_loss": 0.02774990163743496, "global_step": 245306, "epoch": 2756, "lr": 2.279525237093649e-06}
{"train_loss": 0.06366454809904099, "global_step": 245307, "epoch": 2756, "lr": 2.279352188667577e-06}
{"train_loss": 0.03053227812051773, "global_step": 245308, "epoch": 2756, "lr": 2.2791791466569732e-06}
{"train_loss": 0.028035486117005348, "global_step": 245309, "epoch": 2756, "lr": 2.279006111061843e-06}
{"train_loss": 0.07714696228504181, "global_step": 245310, "epoch": 2756, "lr": 2.278833081882231e-06}
{"train_loss": 0.05756638944149017, "global_step": 245311, "epoch": 2756, "lr": 2.2786600591181373e-06}
{"train_loss": 0.05470265448093414, "global_step": 245312, "epoch": 2756, "lr": 2.2784870427696002e-06}
{"train_loss": 0.05940091982483864, "global_step": 245313, "epoch": 2756, "lr": 2.278314032836637e-06}
{"train_loss": 0.06883881241083145, "global_step": 245314, "epoch": 2756, "lr": 2.278141029319275e-06}
{"train_loss": 0.06262068450450897, "global_step": 245315, "epoch": 2756, "lr": 2.277968032217537e-06}
{"train_loss": 0.046441152691841125, "global_step": 245316, "epoch": 2756, "lr": 2.2777950415314443e-06}
{"train_loss": 0.013518569059669971, "global_step": 245317, "epoch": 2756, "lr": 2.277622057261014e-06}
{"train_loss": 0.06586971133947372, "global_step": 245318, "epoch": 2756, "lr": 2.2774490794062796e-06}
{"train_loss": 0.06982391327619553, "global_step": 245319, "epoch": 2756, "lr": 2.277276107967258e-06}
{"train_loss": 0.06494631618261337, "global_step": 245320, "epoch": 2756, "lr": 2.2771031429439817e-06}
{"train_loss": 0.04556965082883835, "global_step": 245321, "epoch": 2756, "lr": 2.2769301843364622e-06}
{"train_loss": 0.05676186457276344, "global_step": 245322, "epoch": 2756, "lr": 2.2767572321447328e-06}
{"train_loss": 0.05253757908940315, "global_step": 245323, "epoch": 2756, "lr": 2.276584286368805e-06}
{"train_loss": 0.04314018785953522, "global_step": 245324, "epoch": 2756, "lr": 2.2764113470087055e-06}
{"train_loss": 0.06876929104328156, "global_step": 245325, "epoch": 2756, "lr": 2.276238414064474e-06}
{"train_loss": 0.04813435673713684, "global_step": 245326, "epoch": 2756, "lr": 2.2760654875361053e-06}
{"train_loss": 0.04137219861149788, "global_step": 245327, "epoch": 2756, "lr": 2.275892567423654e-06}
{"train_loss": 0.04832315817475319, "global_step": 245328, "epoch": 2756, "lr": 2.2757196537271152e-06}
{"train_loss": 0.06343579292297363, "global_step": 245329, "epoch": 2756, "lr": 2.275546746446533e-06}
{"train_loss": 0.04007192701101303, "global_step": 245330, "epoch": 2756, "lr": 2.275373845581913e-06}
{"train_loss": 0.04921609163284302, "global_step": 245331, "epoch": 2756, "lr": 2.2752009511332938e-06}
{"train_loss": 0.03770092874765396, "global_step": 245332, "epoch": 2756, "lr": 2.275028063100687e-06}
{"train_loss": 0.039865802973508835, "global_step": 245333, "epoch": 2756, "lr": 2.274855181484126e-06}
{"train_loss": 0.09256164729595184, "global_step": 245334, "epoch": 2756, "lr": 2.274682306283632e-06}
{"train_loss": 0.05479836091399193, "global_step": 245335, "epoch": 2756, "lr": 2.274509437499217e-06}
{"train_loss": 0.0498957633972168, "global_step": 245336, "epoch": 2756, "lr": 2.2743365751309197e-06}
{"train_loss": 0.019319480285048485, "global_step": 245337, "epoch": 2756, "lr": 2.2741637191787455e-06}
{"train_loss": 0.05745851248502731, "global_step": 245338, "epoch": 2756, "lr": 2.273990869642739e-06}
{"train_loss": 0.06071482598781586, "global_step": 245339, "epoch": 2756, "lr": 2.2738180265229112e-06}
{"train_loss": 0.04844890907406807, "global_step": 245340, "epoch": 2756, "lr": 2.2736451898192846e-06}
{"train_loss": 0.029610438272356987, "global_step": 245341, "epoch": 2756, "lr": 2.273472359531881e-06}
{"train_loss": 0.04358425363898277, "global_step": 245342, "epoch": 2756, "lr": 2.2732995356607336e-06}
{"train_loss": 0.0499090850353241, "global_step": 245343, "epoch": 2756, "lr": 2.2731267182058536e-06}
{"train_loss": 0.03238622099161148, "global_step": 245344, "epoch": 2756, "lr": 2.2729539071672745e-06}
{"train_loss": 0.057476501911878586, "global_step": 245345, "epoch": 2756, "lr": 2.272781102545013e-06}
{"train_loss": 0.028601109981536865, "global_step": 245346, "epoch": 2756, "lr": 2.2726083043390966e-06}
{"train_loss": 0.022837018594145775, "global_step": 245347, "epoch": 2756, "lr": 2.2724355125495424e-06}
{"train_loss": 0.03332005441188812, "global_step": 245348, "epoch": 2756, "lr": 2.272262727176383e-06}
{"train_loss": 0.07736283540725708, "global_step": 245349, "epoch": 2756, "lr": 2.27208994821963e-06}
{"train_loss": 0.06011968478560448, "global_step": 245350, "epoch": 2756, "lr": 2.2719171756793166e-06}
{"train_loss": 0.029187357053160667, "global_step": 245351, "epoch": 2756, "lr": 2.2717444095554595e-06}
{"train_loss": 0.040460579097270966, "global_step": 245352, "epoch": 2756, "lr": 2.2715716498480866e-06}
{"train_loss": 0.037332452833652496, "global_step": 245353, "epoch": 2756, "lr": 2.2713988965572196e-06}
{"train_loss": 0.03803599625825882, "global_step": 245354, "epoch": 2756, "lr": 2.271226149682881e-06}
{"train_loss": 0.04490865767002106, "global_step": 245355, "epoch": 2756, "lr": 2.2710534092250935e-06}
{"train_loss": 0.029619889333844185, "global_step": 245356, "epoch": 2756, "lr": 2.270880675183884e-06}
{"train_loss": 0.04613865911960602, "global_step": 245357, "epoch": 2756, "lr": 2.2707079475592694e-06}
{"train_loss": 0.04262613132596016, "global_step": 245358, "epoch": 2756, "lr": 2.270535226351278e-06}
{"train_loss": 0.03667166084051132, "global_step": 245359, "epoch": 2756, "lr": 2.270362511559926e-06}
{"train_loss": 0.05280855670571327, "global_step": 245360, "epoch": 2756, "lr": 2.2701898031852464e-06}
{"train_loss": 0.02716808207333088, "global_step": 245361, "epoch": 2756, "lr": 2.270017101227262e-06}
{"train_loss": 0.025118086487054825, "global_step": 245362, "epoch": 2756, "lr": 2.2698444056859835e-06}
{"train_loss": 0.09286917001008987, "global_step": 245363, "epoch": 2756, "lr": 2.2696717165614502e-06}
{"train_loss": 0.04778430238366127, "global_step": 245364, "epoch": 2756, "lr": 2.2694990338536727e-06}
{"train_loss": 0.05042339488863945, "global_step": 245365, "epoch": 2756, "lr": 2.2693263575626842e-06}
{"train_loss": 0.014284554868936539, "global_step": 245366, "epoch": 2756, "lr": 2.2691536876885023e-06}
{"train_loss": 0.06241641193628311, "global_step": 245367, "epoch": 2756, "lr": 2.268981024231148e-06}
{"train_loss": 0.022514907643198967, "global_step": 245368, "epoch": 2756, "lr": 2.2688083671906503e-06}
{"train_loss": 0.0503355897963047, "global_step": 245369, "epoch": 2756, "lr": 2.2686357165670303e-06}
{"train_loss": 0.026639556512236595, "global_step": 245370, "epoch": 2756, "lr": 2.268463072360305e-06}
{"train_loss": 0.056013286113739014, "global_step": 245371, "epoch": 2756, "lr": 2.2682904345705082e-06}
{"train_loss": 0.046443279121029246, "global_step": 245372, "epoch": 2756, "lr": 2.2681178031976613e-06, "val_loss": 9.437087059020996}
{"train_loss": 0.01847456581890583, "global_step": 245373, "epoch": 2757, "lr": 2.267945178241776e-06}
{"train_loss": 0.04418424889445305, "global_step": 245374, "epoch": 2757, "lr": 2.2677725597028855e-06}
{"train_loss": 0.06027236580848694, "global_step": 245375, "epoch": 2757, "lr": 2.2675999475810117e-06}
{"train_loss": 0.04723569378256798, "global_step": 245376, "epoch": 2757, "lr": 2.267427341876183e-06}
{"train_loss": 0.05004626885056496, "global_step": 245377, "epoch": 2757, "lr": 2.267254742588415e-06}
{"train_loss": 0.06039898842573166, "global_step": 245378, "epoch": 2757, "lr": 2.267082149717725e-06}
{"train_loss": 0.028649309650063515, "global_step": 245379, "epoch": 2757, "lr": 2.2669095632641467e-06}
{"train_loss": 0.03634224086999893, "global_step": 245380, "epoch": 2757, "lr": 2.2667369832277018e-06}
{"train_loss": 0.0690150037407875, "global_step": 245381, "epoch": 2757, "lr": 2.2665644096084127e-06}
{"train_loss": 0.06512369960546494, "global_step": 245382, "epoch": 2757, "lr": 2.2663918424063067e-06}
{"train_loss": 0.12419688701629639, "global_step": 245383, "epoch": 2757, "lr": 2.2662192816213956e-06}
{"train_loss": 0.04724154248833656, "global_step": 245384, "epoch": 2757, "lr": 2.2660467272537124e-06}
{"train_loss": 0.04572678357362747, "global_step": 245385, "epoch": 2757, "lr": 2.265874179303279e-06}
{"train_loss": 0.009455627761781216, "global_step": 245386, "epoch": 2757, "lr": 2.2657016377701178e-06}
{"train_loss": 0.0843634307384491, "global_step": 245387, "epoch": 2757, "lr": 2.2655291026542458e-06}
{"train_loss": 0.03461050987243652, "global_step": 245388, "epoch": 2757, "lr": 2.2653565739556957e-06}
{"train_loss": 0.025433529168367386, "global_step": 245389, "epoch": 2757, "lr": 2.265184051674485e-06}
{"train_loss": 0.04942762851715088, "global_step": 245390, "epoch": 2757, "lr": 2.2650115358106405e-06}
{"train_loss": 0.03561415150761604, "global_step": 245391, "epoch": 2757, "lr": 2.2648390263641795e-06}
{"train_loss": 0.047068364918231964, "global_step": 245392, "epoch": 2757, "lr": 2.264666523335135e-06}
{"train_loss": 0.027121616527438164, "global_step": 245393, "epoch": 2757, "lr": 2.264494026723518e-06}
{"train_loss": 0.012291526421904564, "global_step": 245394, "epoch": 2757, "lr": 2.2643215365293624e-06}
{"train_loss": 0.03545621410012245, "global_step": 245395, "epoch": 2757, "lr": 2.264149052752684e-06}
{"train_loss": 0.03442462906241417, "global_step": 245396, "epoch": 2757, "lr": 2.263976575393506e-06}
{"train_loss": 0.057540412992239, "global_step": 245397, "epoch": 2757, "lr": 2.263804104451861e-06}
{"train_loss": 0.007854841649532318, "global_step": 245398, "epoch": 2757, "lr": 2.26363163992776e-06}
{"train_loss": 0.0598551444709301, "global_step": 245399, "epoch": 2757, "lr": 2.2634591818212426e-06}
{"train_loss": 0.06485043466091156, "global_step": 245400, "epoch": 2757, "lr": 2.263286730132308e-06}
{"train_loss": 0.03334886208176613, "global_step": 245401, "epoch": 2757, "lr": 2.263114284861001e-06}
{"train_loss": 0.05181020125746727, "global_step": 245402, "epoch": 2757, "lr": 2.2629418460073327e-06}
{"train_loss": 0.07958319783210754, "global_step": 245403, "epoch": 2757, "lr": 2.262769413571336e-06}
{"train_loss": 0.019248198717832565, "global_step": 245404, "epoch": 2757, "lr": 2.262596987553017e-06}
{"train_loss": 0.04472968727350235, "global_step": 245405, "epoch": 2757, "lr": 2.2624245679524204e-06}
{"train_loss": 0.03543075919151306, "global_step": 245406, "epoch": 2757, "lr": 2.2622521547695563e-06}
{"train_loss": 0.05753413215279579, "global_step": 245407, "epoch": 2757, "lr": 2.262079748004453e-06}
{"train_loss": 0.029630502685904503, "global_step": 245408, "epoch": 2757, "lr": 2.2619073476571218e-06}
{"train_loss": 0.027678724378347397, "global_step": 245409, "epoch": 2757, "lr": 2.2617349537276067e-06}
{"train_loss": 0.013822010718286037, "global_step": 245410, "epoch": 2757, "lr": 2.2615625662159134e-06}
{"train_loss": 0.04047700762748718, "global_step": 245411, "epoch": 2757, "lr": 2.2613901851220697e-06}
{"train_loss": 0.06191916763782501, "global_step": 245412, "epoch": 2757, "lr": 2.261217810446109e-06}
{"train_loss": 0.07241837680339813, "global_step": 245413, "epoch": 2757, "lr": 2.261045442188031e-06}
{"train_loss": 0.07063263654708862, "global_step": 245414, "epoch": 2757, "lr": 2.260873080347886e-06}
{"train_loss": 0.03358086943626404, "global_step": 245415, "epoch": 2757, "lr": 2.2607007249256794e-06}
{"train_loss": 0.033454008400440216, "global_step": 245416, "epoch": 2757, "lr": 2.2605283759214447e-06}
{"train_loss": 0.02245890349149704, "global_step": 245417, "epoch": 2757, "lr": 2.2603560333351926e-06}
{"train_loss": 0.04562430456280708, "global_step": 245418, "epoch": 2757, "lr": 2.260183697166962e-06}
{"train_loss": 0.03228531405329704, "global_step": 245419, "epoch": 2757, "lr": 2.260011367416759e-06}
{"train_loss": 0.045397188514471054, "global_step": 245420, "epoch": 2757, "lr": 2.2598390440846217e-06}
{"train_loss": 0.019691791385412216, "global_step": 245421, "epoch": 2757, "lr": 2.2596667271705618e-06}
{"train_loss": 0.025925537571310997, "global_step": 245422, "epoch": 2757, "lr": 2.259494416674618e-06}
{"train_loss": 0.02353656478226185, "global_step": 245423, "epoch": 2757, "lr": 2.25932211259679e-06}
{"train_loss": 0.039292044937610626, "global_step": 245424, "epoch": 2757, "lr": 2.2591498149371227e-06}
{"train_loss": 0.0799185037612915, "global_step": 245425, "epoch": 2757, "lr": 2.258977523695627e-06}
{"train_loss": 0.03156428411602974, "global_step": 245426, "epoch": 2757, "lr": 2.2588052388723355e-06}
{"train_loss": 0.08724995702505112, "global_step": 245427, "epoch": 2757, "lr": 2.2586329604672606e-06}
{"train_loss": 0.027642395347356796, "global_step": 245428, "epoch": 2757, "lr": 2.2584606884804347e-06}
{"train_loss": 0.04982994124293327, "global_step": 245429, "epoch": 2757, "lr": 2.2582884229118697e-06}
{"train_loss": 0.010600744746625423, "global_step": 245430, "epoch": 2757, "lr": 2.2581161637616032e-06}
{"train_loss": 0.03825271129608154, "global_step": 245431, "epoch": 2757, "lr": 2.2579439110296476e-06}
{"train_loss": 0.03467961400747299, "global_step": 245432, "epoch": 2757, "lr": 2.257771664716024e-06}
{"train_loss": 0.038688480854034424, "global_step": 245433, "epoch": 2757, "lr": 2.257599424820772e-06}
{"train_loss": 0.04836645722389221, "global_step": 245434, "epoch": 2757, "lr": 2.2574271913438973e-06}
{"train_loss": 0.039887186139822006, "global_step": 245435, "epoch": 2757, "lr": 2.2572549642854324e-06}
{"train_loss": 0.0255744606256485, "global_step": 245436, "epoch": 2757, "lr": 2.2570827436453944e-06}
{"train_loss": 0.09405026584863663, "global_step": 245437, "epoch": 2757, "lr": 2.2569105294238167e-06}
{"train_loss": 0.041329268366098404, "global_step": 245438, "epoch": 2757, "lr": 2.2567383216207103e-06}
{"train_loss": 0.03804849088191986, "global_step": 245439, "epoch": 2757, "lr": 2.2565661202361033e-06}
{"train_loss": 0.05600496381521225, "global_step": 245440, "epoch": 2757, "lr": 2.256393925270017e-06}
{"train_loss": 0.02805767022073269, "global_step": 245441, "epoch": 2757, "lr": 2.256221736722486e-06}
{"train_loss": 0.037196915596723557, "global_step": 245442, "epoch": 2757, "lr": 2.2560495545935144e-06}
{"train_loss": 0.037320684641599655, "global_step": 245443, "epoch": 2757, "lr": 2.255877378883142e-06}
{"train_loss": 0.06572827696800232, "global_step": 245444, "epoch": 2757, "lr": 2.2557052095913743e-06}
{"train_loss": 0.0377480573952198, "global_step": 245445, "epoch": 2757, "lr": 2.2555330467182556e-06}
{"train_loss": 0.03191643953323364, "global_step": 245446, "epoch": 2757, "lr": 2.255360890263797e-06}
{"train_loss": 0.05393538624048233, "global_step": 245447, "epoch": 2757, "lr": 2.2551887402280203e-06}
{"train_loss": 0.03096434846520424, "global_step": 245448, "epoch": 2757, "lr": 2.2550165966109484e-06}
{"train_loss": 0.047218941152095795, "global_step": 245449, "epoch": 2757, "lr": 2.2548444594126194e-06}
{"train_loss": 0.01899685524404049, "global_step": 245450, "epoch": 2757, "lr": 2.2546723286330396e-06}
{"train_loss": 0.057060595601797104, "global_step": 245451, "epoch": 2757, "lr": 2.2545002042722305e-06}
{"train_loss": 0.04034288227558136, "global_step": 245452, "epoch": 2757, "lr": 2.254328086330226e-06}
{"train_loss": 0.030963245779275894, "global_step": 245453, "epoch": 2757, "lr": 2.254155974807043e-06}
{"train_loss": 0.062628835439682, "global_step": 245454, "epoch": 2757, "lr": 2.253983869702714e-06}
{"train_loss": 0.016488034278154373, "global_step": 245455, "epoch": 2757, "lr": 2.2538117710172447e-06}
{"train_loss": 0.05827827751636505, "global_step": 245456, "epoch": 2757, "lr": 2.2536396787506796e-06}
{"train_loss": 0.03831862658262253, "global_step": 245457, "epoch": 2757, "lr": 2.253467592903019e-06}
{"train_loss": 0.061996810138225555, "global_step": 245458, "epoch": 2757, "lr": 2.2532955134743073e-06}
{"train_loss": 0.03859606757760048, "global_step": 245459, "epoch": 2757, "lr": 2.2531234404645497e-06}
{"train_loss": 0.038497649133205414, "global_step": 245460, "epoch": 2757, "lr": 2.252951373873785e-06}
{"train_loss": 0.04319699191268575, "global_step": 245461, "epoch": 2757, "lr": 2.252779313702025e-06, "val_loss": 9.542134284973145}
{"train_loss": 0.0541270412504673, "global_step": 245462, "epoch": 2758, "lr": 2.252607259949302e-06}
{"train_loss": 0.04134830832481384, "global_step": 245463, "epoch": 2758, "lr": 2.252435212615628e-06}
{"train_loss": 0.09836090356111526, "global_step": 245464, "epoch": 2758, "lr": 2.2522631717010357e-06}
{"train_loss": 0.09441486746072769, "global_step": 245465, "epoch": 2758, "lr": 2.252091137205542e-06}
{"train_loss": 0.053520504385232925, "global_step": 245466, "epoch": 2758, "lr": 2.251919109129175e-06}
{"train_loss": 0.0671309381723404, "global_step": 245467, "epoch": 2758, "lr": 2.2517470874719505e-06}
{"train_loss": 0.028807999566197395, "global_step": 245468, "epoch": 2758, "lr": 2.2515750722338967e-06}
{"train_loss": 0.028645044192671776, "global_step": 245469, "epoch": 2758, "lr": 2.2514030634150474e-06}
{"train_loss": 0.027493424713611603, "global_step": 245470, "epoch": 2758, "lr": 2.251231061015402e-06}
{"train_loss": 0.04376884922385216, "global_step": 245471, "epoch": 2758, "lr": 2.251059065035005e-06}
{"train_loss": 0.05575287342071533, "global_step": 245472, "epoch": 2758, "lr": 2.2508870754738676e-06}
{"train_loss": 0.033427126705646515, "global_step": 245473, "epoch": 2758, "lr": 2.2507150923320174e-06}
{"train_loss": 0.022787312045693398, "global_step": 245474, "epoch": 2758, "lr": 2.2505431156094767e-06}
{"train_loss": 0.04608829319477081, "global_step": 245475, "epoch": 2758, "lr": 2.2503711453062735e-06}
{"train_loss": 0.025557521730661392, "global_step": 245476, "epoch": 2758, "lr": 2.250199181422413e-06}
{"train_loss": 0.08274422585964203, "global_step": 245477, "epoch": 2758, "lr": 2.250027223957946e-06}
{"train_loss": 0.08418364822864532, "global_step": 245478, "epoch": 2758, "lr": 2.249855272912871e-06}
{"train_loss": 0.01569582335650921, "global_step": 245479, "epoch": 2758, "lr": 2.2496833282872275e-06}
{"train_loss": 0.08316631615161896, "global_step": 245480, "epoch": 2758, "lr": 2.2495113900810215e-06}
{"train_loss": 0.11650370806455612, "global_step": 245481, "epoch": 2758, "lr": 2.2493394582942974e-06}
{"train_loss": 0.05406251177191734, "global_step": 245482, "epoch": 2758, "lr": 2.24916753292706e-06}
{"train_loss": 0.05919657647609711, "global_step": 245483, "epoch": 2758, "lr": 2.248995613979349e-06}
{"train_loss": 0.08700759708881378, "global_step": 245484, "epoch": 2758, "lr": 2.2488237014511694e-06}
{"train_loss": 0.03446052968502045, "global_step": 245485, "epoch": 2758, "lr": 2.2486517953425654e-06}
{"train_loss": 0.06412974745035172, "global_step": 245486, "epoch": 2758, "lr": 2.2484798956535323e-06}
{"train_loss": 0.10422901064157486, "global_step": 245487, "epoch": 2758, "lr": 2.2483080023841254e-06}
{"train_loss": 0.04399070143699646, "global_step": 245488, "epoch": 2758, "lr": 2.248136115534344e-06}
{"train_loss": 0.04951098561286926, "global_step": 245489, "epoch": 2758, "lr": 2.2479642351042163e-06}
{"train_loss": 0.04016783460974693, "global_step": 245490, "epoch": 2758, "lr": 2.24779236109377e-06}
{"train_loss": 0.031458478420972824, "global_step": 245491, "epoch": 2758, "lr": 2.247620493503022e-06}
{"train_loss": 0.028883738443255424, "global_step": 245492, "epoch": 2758, "lr": 2.2474486323320052e-06}
{"train_loss": 0.017666270956397057, "global_step": 245493, "epoch": 2758, "lr": 2.2472767775807313e-06}
{"train_loss": 0.08045962452888489, "global_step": 245494, "epoch": 2758, "lr": 2.2471049292492386e-06}
{"train_loss": 0.058221813291311264, "global_step": 245495, "epoch": 2758, "lr": 2.2469330873375273e-06}
{"train_loss": 0.05215361341834068, "global_step": 245496, "epoch": 2758, "lr": 2.246761251845647e-06}
{"train_loss": 0.03145860880613327, "global_step": 245497, "epoch": 2758, "lr": 2.2465894227735983e-06}
{"train_loss": 0.06922854483127594, "global_step": 245498, "epoch": 2758, "lr": 2.2464176001214197e-06}
{"train_loss": 0.025926630944013596, "global_step": 245499, "epoch": 2758, "lr": 2.246245783889123e-06}
{"train_loss": 0.05224644020199776, "global_step": 245500, "epoch": 2758, "lr": 2.24607397407674e-06}
{"train_loss": 0.03602852672338486, "global_step": 245501, "epoch": 2758, "lr": 2.2459021706842887e-06}
{"train_loss": 0.053709667176008224, "global_step": 245502, "epoch": 2758, "lr": 2.2457303737117964e-06}
{"train_loss": 0.014893537387251854, "global_step": 245503, "epoch": 2758, "lr": 2.2455585831592794e-06}
{"train_loss": 0.03681078553199768, "global_step": 245504, "epoch": 2758, "lr": 2.2453867990267665e-06}
{"train_loss": 0.028713000938296318, "global_step": 245505, "epoch": 2758, "lr": 2.2452150213142787e-06}
{"train_loss": 0.06058615446090698, "global_step": 245506, "epoch": 2758, "lr": 2.2450432500218387e-06}
{"train_loss": 0.04996277764439583, "global_step": 245507, "epoch": 2758, "lr": 2.2448714851494747e-06}
{"train_loss": 0.021106578409671783, "global_step": 245508, "epoch": 2758, "lr": 2.244699726697197e-06}
{"train_loss": 0.05920284986495972, "global_step": 245509, "epoch": 2758, "lr": 2.2445279746650504e-06}
{"train_loss": 0.040914613753557205, "global_step": 245510, "epoch": 2758, "lr": 2.2443562290530353e-06}
{"train_loss": 0.03759627044200897, "global_step": 245511, "epoch": 2758, "lr": 2.24418448986119e-06}
{"train_loss": 0.07689688354730606, "global_step": 245512, "epoch": 2758, "lr": 2.2440127570895255e-06}
{"train_loss": 0.04364718124270439, "global_step": 245513, "epoch": 2758, "lr": 2.2438410307380808e-06}
{"train_loss": 0.03725164756178856, "global_step": 245514, "epoch": 2758, "lr": 2.2436693108068616e-06}
{"train_loss": 0.04832516983151436, "global_step": 245515, "epoch": 2758, "lr": 2.2434975972959015e-06}
{"train_loss": 0.05465439707040787, "global_step": 245516, "epoch": 2758, "lr": 2.2433258902052224e-06}
{"train_loss": 0.056209366768598557, "global_step": 245517, "epoch": 2758, "lr": 2.2431541895348464e-06}
{"train_loss": 0.013448229990899563, "global_step": 245518, "epoch": 2758, "lr": 2.24298249528479e-06}
{"train_loss": 0.051876846700906754, "global_step": 245519, "epoch": 2758, "lr": 2.2428108074550925e-06}
{"train_loss": 0.04178183898329735, "global_step": 245520, "epoch": 2758, "lr": 2.242639126045759e-06}
{"train_loss": 0.032561417669057846, "global_step": 245521, "epoch": 2758, "lr": 2.242467451056829e-06}
{"train_loss": 0.01984892226755619, "global_step": 245522, "epoch": 2758, "lr": 2.242295782488307e-06}
{"train_loss": 0.03540504351258278, "global_step": 245523, "epoch": 2758, "lr": 2.242124120340233e-06}
{"train_loss": 0.061057645827531815, "global_step": 245524, "epoch": 2758, "lr": 2.2419524646126177e-06}
{"train_loss": 0.05697447061538696, "global_step": 245525, "epoch": 2758, "lr": 2.2417808153054996e-06}
{"train_loss": 0.0309884212911129, "global_step": 245526, "epoch": 2758, "lr": 2.241609172418879e-06}
{"train_loss": 0.04281655326485634, "global_step": 245527, "epoch": 2758, "lr": 2.2414375359528063e-06}
{"train_loss": 0.049118418246507645, "global_step": 245528, "epoch": 2758, "lr": 2.241265905907286e-06}
{"train_loss": 0.041016239672899246, "global_step": 245529, "epoch": 2758, "lr": 2.2410942822823354e-06}
{"train_loss": 0.02906370535492897, "global_step": 245530, "epoch": 2758, "lr": 2.240922665077999e-06}
{"train_loss": 0.037336770445108414, "global_step": 245531, "epoch": 2758, "lr": 2.240751054294282e-06}
{"train_loss": 0.03039894811809063, "global_step": 245532, "epoch": 2758, "lr": 2.2405794499312184e-06}
{"train_loss": 0.03982780501246452, "global_step": 245533, "epoch": 2758, "lr": 2.240407851988818e-06}
{"train_loss": 0.021831385791301727, "global_step": 245534, "epoch": 2758, "lr": 2.2402362604671213e-06}
{"train_loss": 0.04269886761903763, "global_step": 245535, "epoch": 2758, "lr": 2.2400646753661383e-06}
{"train_loss": 0.04179437831044197, "global_step": 245536, "epoch": 2758, "lr": 2.239893096685902e-06}
{"train_loss": 0.04666971415281296, "global_step": 245537, "epoch": 2758, "lr": 2.239721524426425e-06}
{"train_loss": 0.018143625929951668, "global_step": 245538, "epoch": 2758, "lr": 2.239549958587739e-06}
{"train_loss": 0.05026546120643616, "global_step": 245539, "epoch": 2758, "lr": 2.2393783991698557e-06}
{"train_loss": 0.02767237275838852, "global_step": 245540, "epoch": 2758, "lr": 2.2392068461728144e-06}
{"train_loss": 0.01902562938630581, "global_step": 245541, "epoch": 2758, "lr": 2.2390352995966258e-06}
{"train_loss": 0.039403315633535385, "global_step": 245542, "epoch": 2758, "lr": 2.238863759441312e-06}
{"train_loss": 0.07930731028318405, "global_step": 245543, "epoch": 2758, "lr": 2.2386922257069065e-06}
{"train_loss": 0.0361323282122612, "global_step": 245544, "epoch": 2758, "lr": 2.2385206983934258e-06}
{"train_loss": 0.05849263072013855, "global_step": 245545, "epoch": 2758, "lr": 2.238349177500898e-06}
{"train_loss": 0.044715821743011475, "global_step": 245546, "epoch": 2758, "lr": 2.238177663029334e-06}
{"train_loss": 0.054023902863264084, "global_step": 245547, "epoch": 2758, "lr": 2.2380061549787725e-06}
{"train_loss": 0.052578628063201904, "global_step": 245548, "epoch": 2758, "lr": 2.2378346533492243e-06}
{"train_loss": 0.04165297746658325, "global_step": 245549, "epoch": 2758, "lr": 2.237663158140718e-06}
{"train_loss": 0.04747889219165853, "global_step": 245550, "epoch": 2758, "lr": 2.2374916693532753e-06, "val_loss": 9.490684509277344}
{"train_loss": 0.03150647133588791, "global_step": 245551, "epoch": 2759, "lr": 2.237320186986924e-06}
{"train_loss": 0.027079986408352852, "global_step": 245552, "epoch": 2759, "lr": 2.2371487110416756e-06}
{"train_loss": 0.024942344054579735, "global_step": 245553, "epoch": 2759, "lr": 2.2369772415175683e-06}
{"train_loss": 0.05153333395719528, "global_step": 245554, "epoch": 2759, "lr": 2.2368057784146077e-06}
{"train_loss": 0.03600055351853371, "global_step": 245555, "epoch": 2759, "lr": 2.236634321732839e-06}
{"train_loss": 0.02367793209850788, "global_step": 245556, "epoch": 2759, "lr": 2.2364628714722613e-06}
{"train_loss": 0.04795266315340996, "global_step": 245557, "epoch": 2759, "lr": 2.2362914276329195e-06}
{"train_loss": 0.02687399834394455, "global_step": 245558, "epoch": 2759, "lr": 2.2361199902148134e-06}
{"train_loss": 0.03092205338180065, "global_step": 245559, "epoch": 2759, "lr": 2.235948559217982e-06}
{"train_loss": 0.011175896972417831, "global_step": 245560, "epoch": 2759, "lr": 2.2357771346424526e-06}
{"train_loss": 0.0564822219312191, "global_step": 245561, "epoch": 2759, "lr": 2.2356057164882427e-06}
{"train_loss": 0.019784998148679733, "global_step": 245562, "epoch": 2759, "lr": 2.235434304755363e-06}
{"train_loss": 0.03319542482495308, "global_step": 245563, "epoch": 2759, "lr": 2.235262899443852e-06}
{"train_loss": 0.04597816243767738, "global_step": 245564, "epoch": 2759, "lr": 2.2350915005537267e-06}
{"train_loss": 0.02228189818561077, "global_step": 245565, "epoch": 2759, "lr": 2.2349201080850147e-06}
{"train_loss": 0.0358409509062767, "global_step": 245566, "epoch": 2759, "lr": 2.2347487220377327e-06}
{"train_loss": 0.05865458399057388, "global_step": 245567, "epoch": 2759, "lr": 2.234577342411903e-06}
{"train_loss": 0.050782397389411926, "global_step": 245568, "epoch": 2759, "lr": 2.234405969207559e-06}
{"train_loss": 0.05543341860175133, "global_step": 245569, "epoch": 2759, "lr": 2.2342346024247117e-06}
{"train_loss": 0.032322145998477936, "global_step": 245570, "epoch": 2759, "lr": 2.2340632420633943e-06}
{"train_loss": 0.03520572930574417, "global_step": 245571, "epoch": 2759, "lr": 2.2338918881236182e-06}
{"train_loss": 0.01131995115429163, "global_step": 245572, "epoch": 2759, "lr": 2.2337205406054217e-06}
{"train_loss": 0.04415883496403694, "global_step": 245573, "epoch": 2759, "lr": 2.2335491995088107e-06}
{"train_loss": 0.023687167093157768, "global_step": 245574, "epoch": 2759, "lr": 2.2333778648338246e-06}
{"train_loss": 0.03516938164830208, "global_step": 245575, "epoch": 2759, "lr": 2.2332065365804734e-06}
{"train_loss": 0.07894130796194077, "global_step": 245576, "epoch": 2759, "lr": 2.2330352147487853e-06}
{"train_loss": 0.030295219272375107, "global_step": 245577, "epoch": 2759, "lr": 2.232863899338783e-06}
{"train_loss": 0.07020267844200134, "global_step": 245578, "epoch": 2759, "lr": 2.232692590350488e-06}
{"train_loss": 0.07464788109064102, "global_step": 245579, "epoch": 2759, "lr": 2.232521287783934e-06}
{"train_loss": 0.0072820959612727165, "global_step": 245580, "epoch": 2759, "lr": 2.2323499916391266e-06}
{"train_loss": 0.023716796189546585, "global_step": 245581, "epoch": 2759, "lr": 2.2321787019161044e-06}
{"train_loss": 0.02743915282189846, "global_step": 245582, "epoch": 2759, "lr": 2.2320074186148786e-06}
{"train_loss": 0.07689125835895538, "global_step": 245583, "epoch": 2759, "lr": 2.231836141735477e-06}
{"train_loss": 0.03936559334397316, "global_step": 245584, "epoch": 2759, "lr": 2.231664871277922e-06}
{"train_loss": 0.06671374291181564, "global_step": 245585, "epoch": 2759, "lr": 2.2314936072422467e-06}
{"train_loss": 0.051378313452005386, "global_step": 245586, "epoch": 2759, "lr": 2.2313223496284563e-06}
{"train_loss": 0.04802035540342331, "global_step": 245587, "epoch": 2759, "lr": 2.2311510984365847e-06}
{"train_loss": 0.0831691101193428, "global_step": 245588, "epoch": 2759, "lr": 2.2309798536666484e-06}
{"train_loss": 0.05553682893514633, "global_step": 245589, "epoch": 2759, "lr": 2.2308086153186804e-06}
{"train_loss": 0.043563831597566605, "global_step": 245590, "epoch": 2759, "lr": 2.2306373833926977e-06}
{"train_loss": 0.03538253530859947, "global_step": 245591, "epoch": 2759, "lr": 2.2304661578887222e-06}
{"train_loss": 0.05283185839653015, "global_step": 245592, "epoch": 2759, "lr": 2.230294938806776e-06}
{"train_loss": 0.016243774443864822, "global_step": 245593, "epoch": 2759, "lr": 2.2301237261468877e-06}
{"train_loss": 0.04351557791233063, "global_step": 245594, "epoch": 2759, "lr": 2.229952519909073e-06}
{"train_loss": 0.0600699782371521, "global_step": 245595, "epoch": 2759, "lr": 2.22978132009336e-06}
{"train_loss": 0.06860402226448059, "global_step": 245596, "epoch": 2759, "lr": 2.229610126699777e-06}
{"train_loss": 0.06413007527589798, "global_step": 245597, "epoch": 2759, "lr": 2.229438939728329e-06}
{"train_loss": 0.04067673161625862, "global_step": 245598, "epoch": 2759, "lr": 2.2292677591790658e-06}
{"train_loss": 0.06513899564743042, "global_step": 245599, "epoch": 2759, "lr": 2.2290965850519873e-06}
{"train_loss": 0.03622309863567352, "global_step": 245600, "epoch": 2759, "lr": 2.228925417347122e-06}
{"train_loss": 0.020692026242613792, "global_step": 245601, "epoch": 2759, "lr": 2.2287542560645026e-06}
{"train_loss": 0.043709490448236465, "global_step": 245602, "epoch": 2759, "lr": 2.2285831012041347e-06}
{"train_loss": 0.12089914083480835, "global_step": 245603, "epoch": 2759, "lr": 2.228411952766063e-06}
{"train_loss": 0.043833039700984955, "global_step": 245604, "epoch": 2759, "lr": 2.2282408107502926e-06}
{"train_loss": 0.019972853362560272, "global_step": 245605, "epoch": 2759, "lr": 2.228069675156852e-06}
{"train_loss": 0.0431535430252552, "global_step": 245606, "epoch": 2759, "lr": 2.227898545985768e-06}
{"train_loss": 0.006288446951657534, "global_step": 245607, "epoch": 2759, "lr": 2.227727423237058e-06}
{"train_loss": 0.07320153713226318, "global_step": 245608, "epoch": 2759, "lr": 2.2275563069107554e-06}
{"train_loss": 0.03848445415496826, "global_step": 245609, "epoch": 2759, "lr": 2.227385197006865e-06}
{"train_loss": 0.03542322292923927, "global_step": 245610, "epoch": 2759, "lr": 2.2272140935254258e-06}
{"train_loss": 0.0361754335463047, "global_step": 245611, "epoch": 2759, "lr": 2.2270429964664553e-06}
{"train_loss": 0.03835517540574074, "global_step": 245612, "epoch": 2759, "lr": 2.226871905829975e-06}
{"train_loss": 0.042479343712329865, "global_step": 245613, "epoch": 2759, "lr": 2.226700821616007e-06}
{"train_loss": 0.026832707226276398, "global_step": 245614, "epoch": 2759, "lr": 2.226529743824579e-06}
{"train_loss": 0.045589108020067215, "global_step": 245615, "epoch": 2759, "lr": 2.226358672455714e-06}
{"train_loss": 0.04921279102563858, "global_step": 245616, "epoch": 2759, "lr": 2.226187607509428e-06}
{"train_loss": 0.034294117242097855, "global_step": 245617, "epoch": 2759, "lr": 2.22601654898576e-06}
{"train_loss": 0.0408191978931427, "global_step": 245618, "epoch": 2759, "lr": 2.2258454968847096e-06}
{"train_loss": 0.032828040421009064, "global_step": 245619, "epoch": 2759, "lr": 2.2256744512063164e-06}
{"train_loss": 0.03171828016638756, "global_step": 245620, "epoch": 2759, "lr": 2.225503411950597e-06}
{"train_loss": 0.03910325467586517, "global_step": 245621, "epoch": 2759, "lr": 2.2253323791175838e-06}
{"train_loss": 0.02584102936089039, "global_step": 245622, "epoch": 2759, "lr": 2.2251613527072834e-06}
{"train_loss": 0.02362298220396042, "global_step": 245623, "epoch": 2759, "lr": 2.224990332719734e-06}
{"train_loss": 0.04924803599715233, "global_step": 245624, "epoch": 2759, "lr": 2.224819319154947e-06}
{"train_loss": 0.028881171718239784, "global_step": 245625, "epoch": 2759, "lr": 2.2246483120129556e-06}
{"train_loss": 0.03734442591667175, "global_step": 245626, "epoch": 2759, "lr": 2.224477311293777e-06}
{"train_loss": 0.023884057998657227, "global_step": 245627, "epoch": 2759, "lr": 2.2243063169974377e-06}
{"train_loss": 0.01960168220102787, "global_step": 245628, "epoch": 2759, "lr": 2.22413532912395e-06}
{"train_loss": 0.04392215609550476, "global_step": 245629, "epoch": 2759, "lr": 2.223964347673352e-06}
{"train_loss": 0.01463125180453062, "global_step": 245630, "epoch": 2759, "lr": 2.223793372645655e-06}
{"train_loss": 0.07429084181785583, "global_step": 245631, "epoch": 2759, "lr": 2.2236224040408928e-06}
{"train_loss": 0.054609254002571106, "global_step": 245632, "epoch": 2759, "lr": 2.2234514418590755e-06}
{"train_loss": 0.026783963665366173, "global_step": 245633, "epoch": 2759, "lr": 2.223280486100232e-06}
{"train_loss": 0.053580041974782944, "global_step": 245634, "epoch": 2759, "lr": 2.2231095367643894e-06}
{"train_loss": 0.08651013672351837, "global_step": 245635, "epoch": 2759, "lr": 2.2229385938515646e-06}
{"train_loss": 0.019889352843165398, "global_step": 245636, "epoch": 2759, "lr": 2.2227676573617907e-06}
{"train_loss": 0.06406198441982269, "global_step": 245637, "epoch": 2759, "lr": 2.2225967272950845e-06}
{"train_loss": 0.03137708455324173, "global_step": 245638, "epoch": 2759, "lr": 2.2224258036514566e-06}
{"train_loss": 0.042018483209769045, "global_step": 245639, "epoch": 2759, "lr": 2.222254886430952e-06, "val_loss": 9.48079776763916}
{"train_loss": 0.041435498744249344, "global_step": 245640, "epoch": 2760, "lr": 2.2220839756335767e-06}
{"train_loss": 0.05837896093726158, "global_step": 245641, "epoch": 2760, "lr": 2.2219130712593627e-06}
{"train_loss": 0.02504819445312023, "global_step": 245642, "epoch": 2760, "lr": 2.221742173308328e-06}
{"train_loss": 0.03717556223273277, "global_step": 245643, "epoch": 2760, "lr": 2.221571281780499e-06}
{"train_loss": 0.05011996254324913, "global_step": 245644, "epoch": 2760, "lr": 2.221400396675899e-06}
{"train_loss": 0.08649488538503647, "global_step": 245645, "epoch": 2760, "lr": 2.221229517994544e-06}
{"train_loss": 0.0557418018579483, "global_step": 245646, "epoch": 2760, "lr": 2.221058645736468e-06}
{"train_loss": 0.04659312963485718, "global_step": 245647, "epoch": 2760, "lr": 2.220887779901687e-06}
{"train_loss": 0.06707736849784851, "global_step": 245648, "epoch": 2760, "lr": 2.2207169204902234e-06}
{"train_loss": 0.051159437745809555, "global_step": 245649, "epoch": 2760, "lr": 2.2205460675020996e-06}
{"train_loss": 0.03728162869811058, "global_step": 245650, "epoch": 2760, "lr": 2.220375220937343e-06}
{"train_loss": 0.01987990364432335, "global_step": 245651, "epoch": 2760, "lr": 2.2202043807959816e-06}
{"train_loss": 0.0572618804872036, "global_step": 245652, "epoch": 2760, "lr": 2.220033547078021e-06}
{"train_loss": 0.03565854951739311, "global_step": 245653, "epoch": 2760, "lr": 2.2198627197834996e-06}
{"train_loss": 0.05959144979715347, "global_step": 245654, "epoch": 2760, "lr": 2.219691898912435e-06}
{"train_loss": 0.08945127576589584, "global_step": 245655, "epoch": 2760, "lr": 2.219521084464854e-06}
{"train_loss": 0.015806743875145912, "global_step": 245656, "epoch": 2760, "lr": 2.219350276440768e-06}
{"train_loss": 0.032193537801504135, "global_step": 245657, "epoch": 2760, "lr": 2.219179474840216e-06}
{"train_loss": 0.03500392660498619, "global_step": 245658, "epoch": 2760, "lr": 2.2190086796632036e-06}
{"train_loss": 0.026374779641628265, "global_step": 245659, "epoch": 2760, "lr": 2.218837890909775e-06}
{"train_loss": 0.04986025020480156, "global_step": 245660, "epoch": 2760, "lr": 2.2186671085799306e-06}
{"train_loss": 0.01762116141617298, "global_step": 245661, "epoch": 2760, "lr": 2.218496332673714e-06}
{"train_loss": 0.050849806517362595, "global_step": 245662, "epoch": 2760, "lr": 2.218325563191126e-06}
{"train_loss": 0.06739961355924606, "global_step": 245663, "epoch": 2760, "lr": 2.2181548001322106e-06}
{"train_loss": 0.05207415670156479, "global_step": 245664, "epoch": 2760, "lr": 2.217984043496979e-06}
{"train_loss": 0.03281354904174805, "global_step": 245665, "epoch": 2760, "lr": 2.217813293285459e-06}
{"train_loss": 0.005129266530275345, "global_step": 245666, "epoch": 2760, "lr": 2.2176425494976673e-06}
{"train_loss": 0.11953619122505188, "global_step": 245667, "epoch": 2760, "lr": 2.2174718121336425e-06}
{"train_loss": 0.06164532154798508, "global_step": 245668, "epoch": 2760, "lr": 2.2173010811933848e-06}
{"train_loss": 0.03332345932722092, "global_step": 245669, "epoch": 2760, "lr": 2.2171303566769275e-06}
{"train_loss": 0.022306136786937714, "global_step": 245670, "epoch": 2760, "lr": 2.216959638584298e-06}
{"train_loss": 0.014232741668820381, "global_step": 245671, "epoch": 2760, "lr": 2.216788926915514e-06}
{"train_loss": 0.013427997007966042, "global_step": 245672, "epoch": 2760, "lr": 2.216618221670608e-06}
{"train_loss": 0.033516447991132736, "global_step": 245673, "epoch": 2760, "lr": 2.216447522849591e-06}
{"train_loss": 0.1024792343378067, "global_step": 245674, "epoch": 2760, "lr": 2.2162768304524904e-06}
{"train_loss": 0.05615359544754028, "global_step": 245675, "epoch": 2760, "lr": 2.216106144479324e-06}
{"train_loss": 0.07005001604557037, "global_step": 245676, "epoch": 2760, "lr": 2.21593546493013e-06}
{"train_loss": 0.03732370212674141, "global_step": 245677, "epoch": 2760, "lr": 2.215764791804914e-06}
{"train_loss": 0.03825416415929794, "global_step": 245678, "epoch": 2760, "lr": 2.2155941251037092e-06}
{"train_loss": 0.01718333177268505, "global_step": 245679, "epoch": 2760, "lr": 2.2154234648265326e-06}
{"train_loss": 0.02475040778517723, "global_step": 245680, "epoch": 2760, "lr": 2.215252810973406e-06}
{"train_loss": 0.04955223947763443, "global_step": 245681, "epoch": 2760, "lr": 2.2150821635443686e-06}
{"train_loss": 0.06303637474775314, "global_step": 245682, "epoch": 2760, "lr": 2.2149115225394256e-06}
{"train_loss": 0.040245555341243744, "global_step": 245683, "epoch": 2760, "lr": 2.2147408879585994e-06}
{"train_loss": 0.056621529161930084, "global_step": 245684, "epoch": 2760, "lr": 2.2145702598019235e-06}
{"train_loss": 0.022312715649604797, "global_step": 245685, "epoch": 2760, "lr": 2.2143996380694144e-06}
{"train_loss": 0.03960702195763588, "global_step": 245686, "epoch": 2760, "lr": 2.214229022761094e-06}
{"train_loss": 0.044083304703235626, "global_step": 245687, "epoch": 2760, "lr": 2.214058413876996e-06}
{"train_loss": 0.07544755935668945, "global_step": 245688, "epoch": 2760, "lr": 2.2138878114171256e-06}
{"train_loss": 0.03520117327570915, "global_step": 245689, "epoch": 2760, "lr": 2.2137172153815222e-06}
{"train_loss": 0.07506240159273148, "global_step": 245690, "epoch": 2760, "lr": 2.213546625770202e-06}
{"train_loss": 0.02477092854678631, "global_step": 245691, "epoch": 2760, "lr": 2.2133760425831872e-06}
{"train_loss": 0.07561200857162476, "global_step": 245692, "epoch": 2760, "lr": 2.2132054658205006e-06}
{"train_loss": 0.034790828824043274, "global_step": 245693, "epoch": 2760, "lr": 2.2130348954821687e-06}
{"train_loss": 0.022725384682416916, "global_step": 245694, "epoch": 2760, "lr": 2.212864331568204e-06}
{"train_loss": 0.05456746742129326, "global_step": 245695, "epoch": 2760, "lr": 2.21269377407865e-06}
{"train_loss": 0.01697761006653309, "global_step": 245696, "epoch": 2760, "lr": 2.2125232230135073e-06}
{"train_loss": 0.05726981908082962, "global_step": 245697, "epoch": 2760, "lr": 2.212352678372814e-06}
{"train_loss": 0.07982363551855087, "global_step": 245698, "epoch": 2760, "lr": 2.2121821401565823e-06}
{"train_loss": 0.08359994739294052, "global_step": 245699, "epoch": 2760, "lr": 2.2120116083648445e-06}
{"train_loss": 0.023191960528492928, "global_step": 245700, "epoch": 2760, "lr": 2.211841082997612e-06}
{"train_loss": 0.030488329008221626, "global_step": 245701, "epoch": 2760, "lr": 2.211670564054924e-06}
{"train_loss": 0.058555278927087784, "global_step": 245702, "epoch": 2760, "lr": 2.2115000515367912e-06}
{"train_loss": 0.029254736378788948, "global_step": 245703, "epoch": 2760, "lr": 2.2113295454432416e-06}
{"train_loss": 0.04619309678673744, "global_step": 245704, "epoch": 2760, "lr": 2.2111590457742915e-06}
{"train_loss": 0.038896121084690094, "global_step": 245705, "epoch": 2760, "lr": 2.2109885525299636e-06}
{"train_loss": 0.09068384021520615, "global_step": 245706, "epoch": 2760, "lr": 2.2108180657102963e-06}
{"train_loss": 0.02364829182624817, "global_step": 245707, "epoch": 2760, "lr": 2.210647585315295e-06}
{"train_loss": 0.06261422485113144, "global_step": 245708, "epoch": 2760, "lr": 2.2104771113449994e-06}
{"train_loss": 0.04570498317480087, "global_step": 245709, "epoch": 2760, "lr": 2.2103066437994148e-06}
{"train_loss": 0.06368935853242874, "global_step": 245710, "epoch": 2760, "lr": 2.2101361826785793e-06}
{"train_loss": 0.09052630513906479, "global_step": 245711, "epoch": 2760, "lr": 2.2099657279824992e-06}
{"train_loss": 0.0515810027718544, "global_step": 245712, "epoch": 2760, "lr": 2.2097952797112133e-06}
{"train_loss": 0.06558824330568314, "global_step": 245713, "epoch": 2760, "lr": 2.2096248378647323e-06}
{"train_loss": 0.04130702093243599, "global_step": 245714, "epoch": 2760, "lr": 2.2094544024430895e-06}
{"train_loss": 0.0553245022892952, "global_step": 245715, "epoch": 2760, "lr": 2.209283973446308e-06}
{"train_loss": 0.048469774425029755, "global_step": 245716, "epoch": 2760, "lr": 2.209113550874392e-06}
{"train_loss": 0.0630674883723259, "global_step": 245717, "epoch": 2760, "lr": 2.2089431347273926e-06}
{"train_loss": 0.059412725269794464, "global_step": 245718, "epoch": 2760, "lr": 2.2087727250053037e-06}
{"train_loss": 0.03101191855967045, "global_step": 245719, "epoch": 2760, "lr": 2.208602321708175e-06}
{"train_loss": 0.022930847480893135, "global_step": 245720, "epoch": 2760, "lr": 2.2084319248360186e-06}
{"train_loss": 0.028627343475818634, "global_step": 245721, "epoch": 2760, "lr": 2.2082615343888446e-06}
{"train_loss": 0.05034516006708145, "global_step": 245722, "epoch": 2760, "lr": 2.2080911503666924e-06}
{"train_loss": 0.014859352260828018, "global_step": 245723, "epoch": 2760, "lr": 2.2079207727695837e-06}
{"train_loss": 0.0743759423494339, "global_step": 245724, "epoch": 2760, "lr": 2.2077504015975303e-06}
{"train_loss": 0.04645264893770218, "global_step": 245725, "epoch": 2760, "lr": 2.207580036850565e-06}
{"train_loss": 0.042597752064466476, "global_step": 245726, "epoch": 2760, "lr": 2.2074096785287103e-06}
{"train_loss": 0.04958818852901459, "global_step": 245727, "epoch": 2760, "lr": 2.2072393266319877e-06}
{"train_loss": 0.0469161362669776, "global_step": 245728, "epoch": 2760, "lr": 2.207068981160415e-06, "val_loss": 9.654693603515625, "train_action_mse_error": 11.340278625488281}
{"train_loss": 0.05250124633312225, "global_step": 245729, "epoch": 2761, "lr": 2.2068986421140246e-06}
{"train_loss": 0.028693292289972305, "global_step": 245730, "epoch": 2761, "lr": 2.206728309492828e-06}
{"train_loss": 0.060649972409009933, "global_step": 245731, "epoch": 2761, "lr": 2.2065579832968587e-06}
{"train_loss": 0.03457827866077423, "global_step": 245732, "epoch": 2761, "lr": 2.206387663526133e-06}
{"train_loss": 0.03605928644537926, "global_step": 245733, "epoch": 2761, "lr": 2.2062173501806783e-06}
{"train_loss": 0.11333311349153519, "global_step": 245734, "epoch": 2761, "lr": 2.206047043260512e-06}
{"train_loss": 0.06356186419725418, "global_step": 245735, "epoch": 2761, "lr": 2.205876742765667e-06}
{"train_loss": 0.08778193593025208, "global_step": 245736, "epoch": 2761, "lr": 2.205706448696149e-06}
{"train_loss": 0.05637245252728462, "global_step": 245737, "epoch": 2761, "lr": 2.2055361610519965e-06}
{"train_loss": 0.06116652116179466, "global_step": 245738, "epoch": 2761, "lr": 2.205365879833227e-06}
{"train_loss": 0.037042658776044846, "global_step": 245739, "epoch": 2761, "lr": 2.205195605039867e-06}
{"train_loss": 0.026144716888666153, "global_step": 245740, "epoch": 2761, "lr": 2.205025336671929e-06}
{"train_loss": 0.039860960096120834, "global_step": 245741, "epoch": 2761, "lr": 2.204855074729445e-06}
{"train_loss": 0.04539512097835541, "global_step": 245742, "epoch": 2761, "lr": 2.204684819212438e-06}
{"train_loss": 0.021428432315587997, "global_step": 245743, "epoch": 2761, "lr": 2.2045145701209245e-06}
{"train_loss": 0.05174919590353966, "global_step": 245744, "epoch": 2761, "lr": 2.204344327454938e-06}
{"train_loss": 0.020416464656591415, "global_step": 245745, "epoch": 2761, "lr": 2.2041740912144892e-06}
{"train_loss": 0.023140138015151024, "global_step": 245746, "epoch": 2761, "lr": 2.204003861399612e-06}
{"train_loss": 0.02754370868206024, "global_step": 245747, "epoch": 2761, "lr": 2.2038336380103162e-06}
{"train_loss": 0.03970234468579292, "global_step": 245748, "epoch": 2761, "lr": 2.2036634210466422e-06}
{"train_loss": 0.03628464415669441, "global_step": 245749, "epoch": 2761, "lr": 2.203493210508595e-06}
{"train_loss": 0.0175382811576128, "global_step": 245750, "epoch": 2761, "lr": 2.2033230063962072e-06}
{"train_loss": 0.020510252565145493, "global_step": 245751, "epoch": 2761, "lr": 2.203152808709502e-06}
{"train_loss": 0.03281905874609947, "global_step": 245752, "epoch": 2761, "lr": 2.2029826174485013e-06}
{"train_loss": 0.06654075533151627, "global_step": 245753, "epoch": 2761, "lr": 2.202812432613227e-06}
{"train_loss": 0.01769251376390457, "global_step": 245754, "epoch": 2761, "lr": 2.2026422542037017e-06}
{"train_loss": 0.05040169879794121, "global_step": 245755, "epoch": 2761, "lr": 2.202472082219947e-06}
{"train_loss": 0.03512444347143173, "global_step": 245756, "epoch": 2761, "lr": 2.202301916661986e-06}
{"train_loss": 0.03423205390572548, "global_step": 245757, "epoch": 2761, "lr": 2.202131757529846e-06}
{"train_loss": 0.044726043939590454, "global_step": 245758, "epoch": 2761, "lr": 2.2019616048235438e-06}
{"train_loss": 0.08323055505752563, "global_step": 245759, "epoch": 2761, "lr": 2.201791458543112e-06}
{"train_loss": 0.06128980219364166, "global_step": 245760, "epoch": 2761, "lr": 2.201621318688557e-06}
{"train_loss": 0.03352128341794014, "global_step": 245761, "epoch": 2761, "lr": 2.2014511852599174e-06}
{"train_loss": 0.053112953901290894, "global_step": 245762, "epoch": 2761, "lr": 2.2012810582572097e-06}
{"train_loss": 0.016979681327939034, "global_step": 245763, "epoch": 2761, "lr": 2.201110937680456e-06}
{"train_loss": 0.03894459828734398, "global_step": 245764, "epoch": 2761, "lr": 2.2009408235296793e-06}
{"train_loss": 0.0747249573469162, "global_step": 245765, "epoch": 2761, "lr": 2.200770715804906e-06}
{"train_loss": 0.06637106090784073, "global_step": 245766, "epoch": 2761, "lr": 2.2006006145061543e-06}
{"train_loss": 0.05981678143143654, "global_step": 245767, "epoch": 2761, "lr": 2.2004305196334506e-06}
{"train_loss": 0.02411782182753086, "global_step": 245768, "epoch": 2761, "lr": 2.2002604311868123e-06}
{"train_loss": 0.02553333155810833, "global_step": 245769, "epoch": 2761, "lr": 2.2000903491662784e-06}
{"train_loss": 0.0758085548877716, "global_step": 245770, "epoch": 2761, "lr": 2.199920273571848e-06}
{"train_loss": 0.04092314839363098, "global_step": 245771, "epoch": 2761, "lr": 2.199750204403561e-06}
{"train_loss": 0.048285696655511856, "global_step": 245772, "epoch": 2761, "lr": 2.199580141661428e-06}
{"train_loss": 0.033027954399585724, "global_step": 245773, "epoch": 2761, "lr": 2.199410085345488e-06}
{"train_loss": 0.05173119530081749, "global_step": 245774, "epoch": 2761, "lr": 2.199240035455746e-06}
{"train_loss": 0.06297438591718674, "global_step": 245775, "epoch": 2761, "lr": 2.199069991992242e-06}
{"train_loss": 0.029892833903431892, "global_step": 245776, "epoch": 2761, "lr": 2.198899954954986e-06}
{"train_loss": 0.05403081700205803, "global_step": 245777, "epoch": 2761, "lr": 2.1987299243440008e-06}
{"train_loss": 0.07187124341726303, "global_step": 245778, "epoch": 2761, "lr": 2.198559900159325e-06}
{"train_loss": 0.028875963762402534, "global_step": 245779, "epoch": 2761, "lr": 2.1983898824009586e-06}
{"train_loss": 0.03424220159649849, "global_step": 245780, "epoch": 2761, "lr": 2.198219871068946e-06}
{"train_loss": 0.04196830093860626, "global_step": 245781, "epoch": 2761, "lr": 2.198049866163293e-06}
{"train_loss": 0.04334418475627899, "global_step": 245782, "epoch": 2761, "lr": 2.197879867684033e-06}
{"train_loss": 0.07506538182497025, "global_step": 245783, "epoch": 2761, "lr": 2.197709875631182e-06}
{"train_loss": 0.03943640738725662, "global_step": 245784, "epoch": 2761, "lr": 2.197539890004774e-06}
{"train_loss": 0.0649404376745224, "global_step": 245785, "epoch": 2761, "lr": 2.197369910804814e-06}
{"train_loss": 0.054610393941402435, "global_step": 245786, "epoch": 2761, "lr": 2.197199938031347e-06}
{"train_loss": 0.08984199911355972, "global_step": 245787, "epoch": 2761, "lr": 2.1970299716843724e-06}
{"train_loss": 0.05922061577439308, "global_step": 245788, "epoch": 2761, "lr": 2.1968600117639348e-06}
{"train_loss": 0.023044981062412262, "global_step": 245789, "epoch": 2761, "lr": 2.1966900582700343e-06}
{"train_loss": 0.046157095581293106, "global_step": 245790, "epoch": 2761, "lr": 2.1965201112027156e-06}
{"train_loss": 0.03715091571211815, "global_step": 245791, "epoch": 2761, "lr": 2.196350170561989e-06}
{"train_loss": 0.07304172962903976, "global_step": 245792, "epoch": 2761, "lr": 2.1961802363478833e-06}
{"train_loss": 0.0410563163459301, "global_step": 245793, "epoch": 2761, "lr": 2.19601030856042e-06}
{"train_loss": 0.04689661040902138, "global_step": 245794, "epoch": 2761, "lr": 2.1958403871996104e-06}
{"train_loss": 0.03617434576153755, "global_step": 245795, "epoch": 2761, "lr": 2.1956704722654987e-06}
{"train_loss": 0.056532856076955795, "global_step": 245796, "epoch": 2761, "lr": 2.1955005637580907e-06}
{"train_loss": 0.0590163879096508, "global_step": 245797, "epoch": 2761, "lr": 2.1953306616774194e-06}
{"train_loss": 0.04146935045719147, "global_step": 245798, "epoch": 2761, "lr": 2.1951607660234962e-06}
{"train_loss": 0.07795262336730957, "global_step": 245799, "epoch": 2761, "lr": 2.1949908767963546e-06}
{"train_loss": 0.04267701134085655, "global_step": 245800, "epoch": 2761, "lr": 2.1948209939960108e-06}
{"train_loss": 0.08385287970304489, "global_step": 245801, "epoch": 2761, "lr": 2.194651117622498e-06}
{"train_loss": 0.07078007608652115, "global_step": 245802, "epoch": 2761, "lr": 2.1944812476758224e-06}
{"train_loss": 0.06020231544971466, "global_step": 245803, "epoch": 2761, "lr": 2.1943113841560227e-06}
{"train_loss": 0.04921148344874382, "global_step": 245804, "epoch": 2761, "lr": 2.1941415270631093e-06}
{"train_loss": 0.025934433564543724, "global_step": 245805, "epoch": 2761, "lr": 2.193971676397122e-06}
{"train_loss": 0.05769052356481552, "global_step": 245806, "epoch": 2761, "lr": 2.19380183215806e-06}
{"train_loss": 0.05411981791257858, "global_step": 245807, "epoch": 2761, "lr": 2.1936319943459683e-06}
{"train_loss": 0.019091738387942314, "global_step": 245808, "epoch": 2761, "lr": 2.193462162960852e-06}
{"train_loss": 0.02529917284846306, "global_step": 245809, "epoch": 2761, "lr": 2.1932923380027505e-06}
{"train_loss": 0.047888923436403275, "global_step": 245810, "epoch": 2761, "lr": 2.193122519471669e-06}
{"train_loss": 0.056497201323509216, "global_step": 245811, "epoch": 2761, "lr": 2.192952707367646e-06}
{"train_loss": 0.034501586109399796, "global_step": 245812, "epoch": 2761, "lr": 2.1927829016906932e-06}
{"train_loss": 0.016047460958361626, "global_step": 245813, "epoch": 2761, "lr": 2.192613102440838e-06}
{"train_loss": 0.038450758904218674, "global_step": 245814, "epoch": 2761, "lr": 2.1924433096181086e-06}
{"train_loss": 0.06600326299667358, "global_step": 245815, "epoch": 2761, "lr": 2.1922735232225155e-06}
{"train_loss": 0.05780506506562233, "global_step": 245816, "epoch": 2761, "lr": 2.192103743254098e-06}
{"train_loss": 0.04713871438851517, "global_step": 245817, "epoch": 2761, "lr": 2.1919339697128617e-06, "val_loss": 9.548909187316895}
{"train_loss": 0.008317924104630947, "global_step": 245818, "epoch": 2762, "lr": 2.191764202598845e-06}
{"train_loss": 0.041878409683704376, "global_step": 245819, "epoch": 2762, "lr": 2.1915944419120537e-06}
{"train_loss": 0.033488258719444275, "global_step": 245820, "epoch": 2762, "lr": 2.191424687652527e-06}
{"train_loss": 0.05347089841961861, "global_step": 245821, "epoch": 2762, "lr": 2.1912549398202754e-06}
{"train_loss": 0.05439078062772751, "global_step": 245822, "epoch": 2762, "lr": 2.1910851984153324e-06}
{"train_loss": 0.05490182712674141, "global_step": 245823, "epoch": 2762, "lr": 2.1909154634377038e-06}
{"train_loss": 0.035408925265073776, "global_step": 245824, "epoch": 2762, "lr": 2.1907457348874395e-06}
{"train_loss": 0.042840126901865005, "global_step": 245825, "epoch": 2762, "lr": 2.1905760127645337e-06}
{"train_loss": 0.030720727518200874, "global_step": 245826, "epoch": 2762, "lr": 2.1904062970690307e-06}
{"train_loss": 0.07939617335796356, "global_step": 245827, "epoch": 2762, "lr": 2.1902365878009366e-06}
{"train_loss": 0.04374827444553375, "global_step": 245828, "epoch": 2762, "lr": 2.190066884960296e-06}
{"train_loss": 0.04158347472548485, "global_step": 245829, "epoch": 2762, "lr": 2.189897188547102e-06}
{"train_loss": 0.04481014609336853, "global_step": 245830, "epoch": 2762, "lr": 2.189727498561406e-06}
{"train_loss": 0.06390324980020523, "global_step": 245831, "epoch": 2762, "lr": 2.189557815003218e-06}
{"train_loss": 0.03938876464962959, "global_step": 245832, "epoch": 2762, "lr": 2.1893881378725555e-06}
{"train_loss": 0.01047213189303875, "global_step": 245833, "epoch": 2762, "lr": 2.1892184671694514e-06}
{"train_loss": 0.09874656051397324, "global_step": 245834, "epoch": 2762, "lr": 2.189048802893917e-06}
{"train_loss": 0.018967675045132637, "global_step": 245835, "epoch": 2762, "lr": 2.1888791450459913e-06}
{"train_loss": 0.05696546286344528, "global_step": 245836, "epoch": 2762, "lr": 2.188709493625679e-06}
{"train_loss": 0.05758657306432724, "global_step": 245837, "epoch": 2762, "lr": 2.18853984863302e-06}
{"train_loss": 0.08416581153869629, "global_step": 245838, "epoch": 2762, "lr": 2.188370210068019e-06}
{"train_loss": 0.03690650686621666, "global_step": 245839, "epoch": 2762, "lr": 2.1882005779307214e-06}
{"train_loss": 0.009893816895782948, "global_step": 245840, "epoch": 2762, "lr": 2.188030952221126e-06}
{"train_loss": 0.04484270140528679, "global_step": 245841, "epoch": 2762, "lr": 2.1878613329392784e-06}
{"train_loss": 0.05584632605314255, "global_step": 245842, "epoch": 2762, "lr": 2.1876917200851775e-06}
{"train_loss": 0.04204239323735237, "global_step": 245843, "epoch": 2762, "lr": 2.1875221136588687e-06}
{"train_loss": 0.03465142101049423, "global_step": 245844, "epoch": 2762, "lr": 2.187352513660362e-06}
{"train_loss": 0.07520975172519684, "global_step": 245845, "epoch": 2762, "lr": 2.1871829200896867e-06}
{"train_loss": 0.08845476061105728, "global_step": 245846, "epoch": 2762, "lr": 2.1870133329468525e-06}
{"train_loss": 0.02211776003241539, "global_step": 245847, "epoch": 2762, "lr": 2.1868437522318984e-06}
{"train_loss": 0.05024809017777443, "global_step": 245848, "epoch": 2762, "lr": 2.1866741779448364e-06}
{"train_loss": 0.01661214418709278, "global_step": 245849, "epoch": 2762, "lr": 2.186504610085693e-06}
{"train_loss": 0.026104746386408806, "global_step": 245850, "epoch": 2762, "lr": 2.1863350486544975e-06}
{"train_loss": 0.02982814982533455, "global_step": 245851, "epoch": 2762, "lr": 2.18616549365126e-06}
{"train_loss": 0.03741953894495964, "global_step": 245852, "epoch": 2762, "lr": 2.1859959450760135e-06}
{"train_loss": 0.02625168487429619, "global_step": 245853, "epoch": 2762, "lr": 2.1858264029287756e-06}
{"train_loss": 0.015334658324718475, "global_step": 245854, "epoch": 2762, "lr": 2.1856568672095733e-06}
{"train_loss": 0.06691401451826096, "global_step": 245855, "epoch": 2762, "lr": 2.1854873379184236e-06}
{"train_loss": 0.04676857218146324, "global_step": 245856, "epoch": 2762, "lr": 2.1853178150553544e-06}
{"train_loss": 0.005321312230080366, "global_step": 245857, "epoch": 2762, "lr": 2.185148298620382e-06}
{"train_loss": 0.029851054772734642, "global_step": 245858, "epoch": 2762, "lr": 2.18497878861354e-06}
{"train_loss": 0.033232640475034714, "global_step": 245859, "epoch": 2762, "lr": 2.184809285034839e-06}
{"train_loss": 0.04764493927359581, "global_step": 245860, "epoch": 2762, "lr": 2.184639787884313e-06}
{"train_loss": 0.03809323534369469, "global_step": 245861, "epoch": 2762, "lr": 2.184470297161978e-06}
{"train_loss": 0.01943022944033146, "global_step": 245862, "epoch": 2762, "lr": 2.184300812867857e-06}
{"train_loss": 0.09577426314353943, "global_step": 245863, "epoch": 2762, "lr": 2.184131335001971e-06}
{"train_loss": 0.026381200179457664, "global_step": 245864, "epoch": 2762, "lr": 2.1839618635643543e-06}
{"train_loss": 0.036171309649944305, "global_step": 245865, "epoch": 2762, "lr": 2.1837923985550123e-06}
{"train_loss": 0.026456698775291443, "global_step": 245866, "epoch": 2762, "lr": 2.1836229399739837e-06}
{"train_loss": 0.07493969053030014, "global_step": 245867, "epoch": 2762, "lr": 2.183453487821274e-06}
{"train_loss": 0.036853544414043427, "global_step": 245868, "epoch": 2762, "lr": 2.183284042096928e-06}
{"train_loss": 0.04772467166185379, "global_step": 245869, "epoch": 2762, "lr": 2.183114602800945e-06}
{"train_loss": 0.10642529278993607, "global_step": 245870, "epoch": 2762, "lr": 2.1829451699333703e-06}
{"train_loss": 0.07803072780370712, "global_step": 245871, "epoch": 2762, "lr": 2.1827757434942144e-06}
{"train_loss": 0.05878257751464844, "global_step": 245872, "epoch": 2762, "lr": 2.182606323483494e-06}
{"train_loss": 0.03400357812643051, "global_step": 245873, "epoch": 2762, "lr": 2.1824369099012477e-06}
{"train_loss": 0.0589635856449604, "global_step": 245874, "epoch": 2762, "lr": 2.1822675027474815e-06}
{"train_loss": 0.04341107979416847, "global_step": 245875, "epoch": 2762, "lr": 2.182098102022234e-06}
{"train_loss": 0.06086457148194313, "global_step": 245876, "epoch": 2762, "lr": 2.1819287077255166e-06}
{"train_loss": 0.04918103292584419, "global_step": 245877, "epoch": 2762, "lr": 2.181759319857363e-06}
{"train_loss": 0.057244013994932175, "global_step": 245878, "epoch": 2762, "lr": 2.1815899384177775e-06}
{"train_loss": 0.042669057846069336, "global_step": 245879, "epoch": 2762, "lr": 2.181420563406805e-06}
{"train_loss": 0.03813496604561806, "global_step": 245880, "epoch": 2762, "lr": 2.1812511948244464e-06}
{"train_loss": 0.035942427814006805, "global_step": 245881, "epoch": 2762, "lr": 2.181081832670745e-06}
{"train_loss": 0.1145978718996048, "global_step": 245882, "epoch": 2762, "lr": 2.180912476945707e-06}
{"train_loss": 0.02527875453233719, "global_step": 245883, "epoch": 2762, "lr": 2.1807431276493706e-06}
{"train_loss": 0.01651514135301113, "global_step": 245884, "epoch": 2762, "lr": 2.1805737847817476e-06}
{"train_loss": 0.017671160399913788, "global_step": 245885, "epoch": 2762, "lr": 2.1804044483428653e-06}
{"train_loss": 0.045620907098054886, "global_step": 245886, "epoch": 2762, "lr": 2.1802351183327353e-06}
{"train_loss": 0.09173521399497986, "global_step": 245887, "epoch": 2762, "lr": 2.1800657947513956e-06}
{"train_loss": 0.036717962473630905, "global_step": 245888, "epoch": 2762, "lr": 2.1798964775988696e-06}
{"train_loss": 0.07672488689422607, "global_step": 245889, "epoch": 2762, "lr": 2.1797271668751618e-06}
{"train_loss": 0.0780566930770874, "global_step": 245890, "epoch": 2762, "lr": 2.179557862580317e-06}
{"train_loss": 0.030008554458618164, "global_step": 245891, "epoch": 2762, "lr": 2.179388564714341e-06}
{"train_loss": 0.023591594770550728, "global_step": 245892, "epoch": 2762, "lr": 2.179219273277272e-06}
{"train_loss": 0.024274060502648354, "global_step": 245893, "epoch": 2762, "lr": 2.1790499882691162e-06}
{"train_loss": 0.02741067297756672, "global_step": 245894, "epoch": 2762, "lr": 2.1788807096899064e-06}
{"train_loss": 0.016060495749115944, "global_step": 245895, "epoch": 2762, "lr": 2.17871143753966e-06}
{"train_loss": 0.031908441334962845, "global_step": 245896, "epoch": 2762, "lr": 2.1785421718184153e-06}
{"train_loss": 0.036264337599277496, "global_step": 245897, "epoch": 2762, "lr": 2.1783729125261665e-06}
{"train_loss": 0.042859263718128204, "global_step": 245898, "epoch": 2762, "lr": 2.1782036596629642e-06}
{"train_loss": 0.03571115806698799, "global_step": 245899, "epoch": 2762, "lr": 2.178034413228813e-06}
{"train_loss": 0.045570820569992065, "global_step": 245900, "epoch": 2762, "lr": 2.1778651732237477e-06}
{"train_loss": 0.05259380489587784, "global_step": 245901, "epoch": 2762, "lr": 2.177695939647778e-06}
{"train_loss": 0.04855623468756676, "global_step": 245902, "epoch": 2762, "lr": 2.1775267125009436e-06}
{"train_loss": 0.06847121566534042, "global_step": 245903, "epoch": 2762, "lr": 2.1773574917832494e-06}
{"train_loss": 0.04076797887682915, "global_step": 245904, "epoch": 2762, "lr": 2.177188277494735e-06}
{"train_loss": 0.05316510051488876, "global_step": 245905, "epoch": 2762, "lr": 2.1770190696354053e-06}
{"train_loss": 0.046219593662182604, "global_step": 245906, "epoch": 2762, "lr": 2.1768498682052997e-06, "val_loss": 9.556790351867676}
{"train_loss": 0.04826117679476738, "global_step": 245907, "epoch": 2763, "lr": 2.176680673204429e-06}
{"train_loss": 0.041589848697185516, "global_step": 245908, "epoch": 2763, "lr": 2.176511484632826e-06}
{"train_loss": 0.08645971864461899, "global_step": 245909, "epoch": 2763, "lr": 2.1763423024905085e-06}
{"train_loss": 0.0506853312253952, "global_step": 245910, "epoch": 2763, "lr": 2.176173126777492e-06}
{"train_loss": 0.02821624092757702, "global_step": 245911, "epoch": 2763, "lr": 2.1760039574938107e-06}
{"train_loss": 0.032149605453014374, "global_step": 245912, "epoch": 2763, "lr": 2.175834794639481e-06}
{"train_loss": 0.04845463111996651, "global_step": 245913, "epoch": 2763, "lr": 2.17566563821453e-06}
{"train_loss": 0.02139636129140854, "global_step": 245914, "epoch": 2763, "lr": 2.1754964882189755e-06}
{"train_loss": 0.09482520818710327, "global_step": 245915, "epoch": 2763, "lr": 2.1753273446528445e-06}
{"train_loss": 0.018522251397371292, "global_step": 245916, "epoch": 2763, "lr": 2.1751582075161536e-06}
{"train_loss": 0.04124099388718605, "global_step": 245917, "epoch": 2763, "lr": 2.1749890768089365e-06}
{"train_loss": 0.0882013812661171, "global_step": 245918, "epoch": 2763, "lr": 2.174819952531204e-06}
{"train_loss": 0.01804535835981369, "global_step": 245919, "epoch": 2763, "lr": 2.174650834682984e-06}
{"train_loss": 0.029733076691627502, "global_step": 245920, "epoch": 2763, "lr": 2.174481723264299e-06}
{"train_loss": 0.05480128899216652, "global_step": 245921, "epoch": 2763, "lr": 2.174312618275176e-06}
{"train_loss": 0.04328379034996033, "global_step": 245922, "epoch": 2763, "lr": 2.1741435197156324e-06}
{"train_loss": 0.04484882578253746, "global_step": 245923, "epoch": 2763, "lr": 2.1739744275856843e-06}
{"train_loss": 0.029263895004987717, "global_step": 245924, "epoch": 2763, "lr": 2.1738053418853765e-06}
{"train_loss": 0.03766821324825287, "global_step": 245925, "epoch": 2763, "lr": 2.1736362626147035e-06}
{"train_loss": 0.021462751552462578, "global_step": 245926, "epoch": 2763, "lr": 2.1734671897737145e-06}
{"train_loss": 0.0480765625834465, "global_step": 245927, "epoch": 2763, "lr": 2.1732981233624108e-06}
{"train_loss": 0.028893593698740005, "global_step": 245928, "epoch": 2763, "lr": 2.17312906338083e-06}
{"train_loss": 0.024167997762560844, "global_step": 245929, "epoch": 2763, "lr": 2.172960009828984e-06}
{"train_loss": 0.07501407712697983, "global_step": 245930, "epoch": 2763, "lr": 2.172790962706911e-06}
{"train_loss": 0.055083293467760086, "global_step": 245931, "epoch": 2763, "lr": 2.1726219220146116e-06}
{"train_loss": 0.05150945112109184, "global_step": 245932, "epoch": 2763, "lr": 2.1724528877521244e-06}
{"train_loss": 0.0437089167535305, "global_step": 245933, "epoch": 2763, "lr": 2.172283859919466e-06}
{"train_loss": 0.056189242750406265, "global_step": 245934, "epoch": 2763, "lr": 2.17211483851667e-06}
{"train_loss": 0.026078835129737854, "global_step": 245935, "epoch": 2763, "lr": 2.171945823543736e-06}
{"train_loss": 0.034855809062719345, "global_step": 245936, "epoch": 2763, "lr": 2.1717768150007144e-06}
{"train_loss": 0.07305119186639786, "global_step": 245937, "epoch": 2763, "lr": 2.171607812887605e-06}
{"train_loss": 0.07730234414339066, "global_step": 245938, "epoch": 2763, "lr": 2.1714388172044464e-06}
{"train_loss": 0.0675857812166214, "global_step": 245939, "epoch": 2763, "lr": 2.17126982795125e-06}
{"train_loss": 0.06505856662988663, "global_step": 245940, "epoch": 2763, "lr": 2.1711008451280433e-06}
{"train_loss": 0.038905560970306396, "global_step": 245941, "epoch": 2763, "lr": 2.1709318687348547e-06}
{"train_loss": 0.03682471066713333, "global_step": 245942, "epoch": 2763, "lr": 2.1707628987717e-06}
{"train_loss": 0.0391046367585659, "global_step": 245943, "epoch": 2763, "lr": 2.170593935238596e-06}
{"train_loss": 0.047218214720487595, "global_step": 245944, "epoch": 2763, "lr": 2.1704249781355822e-06}
{"train_loss": 0.04944713041186333, "global_step": 245945, "epoch": 2763, "lr": 2.170256027462664e-06}
{"train_loss": 0.03530413284897804, "global_step": 245946, "epoch": 2763, "lr": 2.1700870832198796e-06}
{"train_loss": 0.04197252541780472, "global_step": 245947, "epoch": 2763, "lr": 2.169918145407246e-06}
{"train_loss": 0.06032111868262291, "global_step": 245948, "epoch": 2763, "lr": 2.169749214024774e-06}
{"train_loss": 0.022017473354935646, "global_step": 245949, "epoch": 2763, "lr": 2.169580289072504e-06}
{"train_loss": 0.029306674376130104, "global_step": 245950, "epoch": 2763, "lr": 2.169411370550439e-06}
{"train_loss": 0.05790415778756142, "global_step": 245951, "epoch": 2763, "lr": 2.1692424584586255e-06}
{"train_loss": 0.029315216466784477, "global_step": 245952, "epoch": 2763, "lr": 2.1690735527970685e-06}
{"train_loss": 0.014357443898916245, "global_step": 245953, "epoch": 2763, "lr": 2.168904653565801e-06}
{"train_loss": 0.05122816190123558, "global_step": 245954, "epoch": 2763, "lr": 2.1687357607648396e-06}
{"train_loss": 0.058535125106573105, "global_step": 245955, "epoch": 2763, "lr": 2.1685668743942123e-06}
{"train_loss": 0.022969504818320274, "global_step": 245956, "epoch": 2763, "lr": 2.16839799445393e-06}
{"train_loss": 0.033682193607091904, "global_step": 245957, "epoch": 2763, "lr": 2.1682291209440263e-06}
{"train_loss": 0.05504535511136055, "global_step": 245958, "epoch": 2763, "lr": 2.1680602538645235e-06}
{"train_loss": 0.033152904361486435, "global_step": 245959, "epoch": 2763, "lr": 2.167891393215438e-06}
{"train_loss": 0.041313204914331436, "global_step": 245960, "epoch": 2763, "lr": 2.1677225389968026e-06}
{"train_loss": 0.05917323753237724, "global_step": 245961, "epoch": 2763, "lr": 2.1675536912086234e-06}
{"train_loss": 0.03029526397585869, "global_step": 245962, "epoch": 2763, "lr": 2.167384849850945e-06}
{"train_loss": 0.052506282925605774, "global_step": 245963, "epoch": 2763, "lr": 2.167216014923773e-06}
{"train_loss": 0.06460326910018921, "global_step": 245964, "epoch": 2763, "lr": 2.16704718642714e-06}
{"train_loss": 0.0258510522544384, "global_step": 245965, "epoch": 2763, "lr": 2.1668783643610578e-06}
{"train_loss": 0.047164756804704666, "global_step": 245966, "epoch": 2763, "lr": 2.166709548725565e-06}
{"train_loss": 0.07828473299741745, "global_step": 245967, "epoch": 2763, "lr": 2.1665407395206673e-06}
{"train_loss": 0.03864269703626633, "global_step": 245968, "epoch": 2763, "lr": 2.166371936746403e-06}
{"train_loss": 0.044191595166921616, "global_step": 245969, "epoch": 2763, "lr": 2.1662031404027782e-06}
{"train_loss": 0.03543640300631523, "global_step": 245970, "epoch": 2763, "lr": 2.166034350489832e-06}
{"train_loss": 0.01581529714167118, "global_step": 245971, "epoch": 2763, "lr": 2.165865567007569e-06}
{"train_loss": 0.02972007915377617, "global_step": 245972, "epoch": 2763, "lr": 2.165696789956034e-06}
{"train_loss": 0.03372681885957718, "global_step": 245973, "epoch": 2763, "lr": 2.165528019335228e-06}
{"train_loss": 0.05136571452021599, "global_step": 245974, "epoch": 2763, "lr": 2.165359255145194e-06}
{"train_loss": 0.04201754927635193, "global_step": 245975, "epoch": 2763, "lr": 2.1651904973859385e-06}
{"train_loss": 0.025571631267666817, "global_step": 245976, "epoch": 2763, "lr": 2.1650217460574885e-06}
{"train_loss": 0.058317072689533234, "global_step": 245977, "epoch": 2763, "lr": 2.1648530011598723e-06}
{"train_loss": 0.016825823113322258, "global_step": 245978, "epoch": 2763, "lr": 2.164684262693101e-06}
{"train_loss": 0.06442674249410629, "global_step": 245979, "epoch": 2763, "lr": 2.1645155306572184e-06}
{"train_loss": 0.0681372657418251, "global_step": 245980, "epoch": 2763, "lr": 2.164346805052225e-06}
{"train_loss": 0.04850900545716286, "global_step": 245981, "epoch": 2763, "lr": 2.164178085878149e-06}
{"train_loss": 0.05293506011366844, "global_step": 245982, "epoch": 2763, "lr": 2.1640093731350284e-06}
{"train_loss": 0.04406801611185074, "global_step": 245983, "epoch": 2763, "lr": 2.1638406668228585e-06}
{"train_loss": 0.05282706767320633, "global_step": 245984, "epoch": 2763, "lr": 2.1636719669416883e-06}
{"train_loss": 0.04299704357981682, "global_step": 245985, "epoch": 2763, "lr": 2.163503273491524e-06}
{"train_loss": 0.03942951560020447, "global_step": 245986, "epoch": 2763, "lr": 2.1633345864723987e-06}
{"train_loss": 0.060940876603126526, "global_step": 245987, "epoch": 2763, "lr": 2.163165905884329e-06}
{"train_loss": 0.06116173043847084, "global_step": 245988, "epoch": 2763, "lr": 2.1629972317273316e-06}
{"train_loss": 0.050943851470947266, "global_step": 245989, "epoch": 2763, "lr": 2.1628285640014456e-06}
{"train_loss": 0.035037457942962646, "global_step": 245990, "epoch": 2763, "lr": 2.1626599027066763e-06}
{"train_loss": 0.08956672996282578, "global_step": 245991, "epoch": 2763, "lr": 2.162491247843057e-06}
{"train_loss": 0.018815893679857254, "global_step": 245992, "epoch": 2763, "lr": 2.16232259941061e-06}
{"train_loss": 0.06508037447929382, "global_step": 245993, "epoch": 2763, "lr": 2.162153957409352e-06}
{"train_loss": 0.046498414129018784, "global_step": 245994, "epoch": 2763, "lr": 2.1619853218393104e-06}
{"train_loss": 0.04501394832276561, "global_step": 245995, "epoch": 2763, "lr": 2.1618166927005024e-06, "val_loss": 9.571043014526367}
{"train_loss": 0.038913797587156296, "global_step": 245996, "epoch": 2764, "lr": 2.1616480699929665e-06}
{"train_loss": 0.053132399916648865, "global_step": 245997, "epoch": 2764, "lr": 2.161479453716703e-06}
{"train_loss": 0.04762880504131317, "global_step": 245998, "epoch": 2764, "lr": 2.16131084387175e-06}
{"train_loss": 0.040373072028160095, "global_step": 245999, "epoch": 2764, "lr": 2.161142240458125e-06}
{"train_loss": 0.06103004142642021, "global_step": 246000, "epoch": 2764, "lr": 2.1609736434758554e-06}
{"train_loss": 0.06769158691167831, "global_step": 246001, "epoch": 2764, "lr": 2.160805052924952e-06}
{"train_loss": 0.06261423975229263, "global_step": 246002, "epoch": 2764, "lr": 2.160636468805455e-06}
{"train_loss": 0.03365166112780571, "global_step": 246003, "epoch": 2764, "lr": 2.1604678911173683e-06}
{"train_loss": 0.049953147768974304, "global_step": 246004, "epoch": 2764, "lr": 2.1602993198607312e-06}
{"train_loss": 0.05132610350847244, "global_step": 246005, "epoch": 2764, "lr": 2.1601307550355497e-06}
{"train_loss": 0.04568915069103241, "global_step": 246006, "epoch": 2764, "lr": 2.159962196641868e-06}
{"train_loss": 0.02046493999660015, "global_step": 246007, "epoch": 2764, "lr": 2.1597936446796807e-06}
{"train_loss": 0.08493902534246445, "global_step": 246008, "epoch": 2764, "lr": 2.1596250991490428e-06}
{"train_loss": 0.015065164305269718, "global_step": 246009, "epoch": 2764, "lr": 2.159456560049944e-06}
{"train_loss": 0.09223875403404236, "global_step": 246010, "epoch": 2764, "lr": 2.159288027382439e-06}
{"train_loss": 0.04847615584731102, "global_step": 246011, "epoch": 2764, "lr": 2.1591195011465224e-06}
{"train_loss": 0.044506147503852844, "global_step": 246012, "epoch": 2764, "lr": 2.158950981342234e-06}
{"train_loss": 0.08221155405044556, "global_step": 246013, "epoch": 2764, "lr": 2.158782467969589e-06}
{"train_loss": 0.048654939979314804, "global_step": 246014, "epoch": 2764, "lr": 2.1586139610286103e-06}
{"train_loss": 0.011729386635124683, "global_step": 246015, "epoch": 2764, "lr": 2.1584454605193317e-06}
{"train_loss": 0.04252297431230545, "global_step": 246016, "epoch": 2764, "lr": 2.158276966441758e-06}
{"train_loss": 0.020675774663686752, "global_step": 246017, "epoch": 2764, "lr": 2.1581084787959283e-06}
{"train_loss": 0.03198573365807533, "global_step": 246018, "epoch": 2764, "lr": 2.157939997581854e-06}
{"train_loss": 0.02059510536491871, "global_step": 246019, "epoch": 2764, "lr": 2.157771522799562e-06}
{"train_loss": 0.03890971839427948, "global_step": 246020, "epoch": 2764, "lr": 2.157603054449081e-06}
{"train_loss": 0.02704809606075287, "global_step": 246021, "epoch": 2764, "lr": 2.1574345925304163e-06}
{"train_loss": 0.023014893755316734, "global_step": 246022, "epoch": 2764, "lr": 2.157266137043612e-06}
{"train_loss": 0.04874500632286072, "global_step": 246023, "epoch": 2764, "lr": 2.1570976879886685e-06}
{"train_loss": 0.03705864027142525, "global_step": 246024, "epoch": 2764, "lr": 2.15692924536563e-06}
{"train_loss": 0.03154721111059189, "global_step": 246025, "epoch": 2764, "lr": 2.1567608091745073e-06}
{"train_loss": 0.04132482036948204, "global_step": 246026, "epoch": 2764, "lr": 2.1565923794153174e-06}
{"train_loss": 0.02401289902627468, "global_step": 246027, "epoch": 2764, "lr": 2.1564239560880994e-06}
{"train_loss": 0.027013469487428665, "global_step": 246028, "epoch": 2764, "lr": 2.156255539192864e-06}
{"train_loss": 0.06346580386161804, "global_step": 246029, "epoch": 2764, "lr": 2.156087128729639e-06}
{"train_loss": 0.03457029536366463, "global_step": 246030, "epoch": 2764, "lr": 2.155918724698436e-06}
{"train_loss": 0.034111786633729935, "global_step": 246031, "epoch": 2764, "lr": 2.155750327099293e-06}
{"train_loss": 0.040584810078144073, "global_step": 246032, "epoch": 2764, "lr": 2.155581935932227e-06}
{"train_loss": 0.049352262169122696, "global_step": 246033, "epoch": 2764, "lr": 2.155413551197255e-06}
{"train_loss": 0.025319628417491913, "global_step": 246034, "epoch": 2764, "lr": 2.1552451728944156e-06}
{"train_loss": 0.049370668828487396, "global_step": 246035, "epoch": 2764, "lr": 2.155076801023709e-06}
{"train_loss": 0.04821796715259552, "global_step": 246036, "epoch": 2764, "lr": 2.1549084355851732e-06}
{"train_loss": 0.031701602041721344, "global_step": 246037, "epoch": 2764, "lr": 2.1547400765788263e-06}
{"train_loss": 0.0597984679043293, "global_step": 246038, "epoch": 2764, "lr": 2.1545717240046947e-06}
{"train_loss": 0.08445281535387039, "global_step": 246039, "epoch": 2764, "lr": 2.1544033778627902e-06}
{"train_loss": 0.02724471315741539, "global_step": 246040, "epoch": 2764, "lr": 2.154235038153152e-06}
{"train_loss": 0.03276577964425087, "global_step": 246041, "epoch": 2764, "lr": 2.15406670487579e-06}
{"train_loss": 0.06234182417392731, "global_step": 246042, "epoch": 2764, "lr": 2.153898378030733e-06}
{"train_loss": 0.02902209758758545, "global_step": 246043, "epoch": 2764, "lr": 2.153730057617992e-06}
{"train_loss": 0.02941937744617462, "global_step": 246044, "epoch": 2764, "lr": 2.153561743637611e-06}
{"train_loss": 0.09416422992944717, "global_step": 246045, "epoch": 2764, "lr": 2.15339343608959e-06}
{"train_loss": 0.02334613725543022, "global_step": 246046, "epoch": 2764, "lr": 2.153225134973974e-06}
{"train_loss": 0.04693005234003067, "global_step": 246047, "epoch": 2764, "lr": 2.1530568402907626e-06}
{"train_loss": 0.051585741341114044, "global_step": 246048, "epoch": 2764, "lr": 2.1528885520399945e-06}
{"train_loss": 0.0724867656826973, "global_step": 246049, "epoch": 2764, "lr": 2.1527202702216865e-06}
{"train_loss": 0.04714768007397652, "global_step": 246050, "epoch": 2764, "lr": 2.152551994835855e-06}
{"train_loss": 0.03834123909473419, "global_step": 246051, "epoch": 2764, "lr": 2.15238372588254e-06}
{"train_loss": 0.0339074432849884, "global_step": 246052, "epoch": 2764, "lr": 2.152215463361751e-06}
{"train_loss": 0.03417178988456726, "global_step": 246053, "epoch": 2764, "lr": 2.1520472072735164e-06}
{"train_loss": 0.019512830302119255, "global_step": 246054, "epoch": 2764, "lr": 2.1518789576178477e-06}
{"train_loss": 0.07335758954286575, "global_step": 246055, "epoch": 2764, "lr": 2.1517107143947836e-06}
{"train_loss": 0.04648679122328758, "global_step": 246056, "epoch": 2764, "lr": 2.1515424776043346e-06}
{"train_loss": 0.09365089237689972, "global_step": 246057, "epoch": 2764, "lr": 2.151374247246529e-06}
{"train_loss": 0.024292094632983208, "global_step": 246058, "epoch": 2764, "lr": 2.1512060233213895e-06}
{"train_loss": 0.029722126200795174, "global_step": 246059, "epoch": 2764, "lr": 2.1510378058289315e-06}
{"train_loss": 0.09289895743131638, "global_step": 246060, "epoch": 2764, "lr": 2.150869594769195e-06}
{"train_loss": 0.021420998498797417, "global_step": 246061, "epoch": 2764, "lr": 2.150701390142179e-06}
{"train_loss": 0.052715856581926346, "global_step": 246062, "epoch": 2764, "lr": 2.1505331919479233e-06}
{"train_loss": 0.03768632933497429, "global_step": 246063, "epoch": 2764, "lr": 2.1503650001864496e-06}
{"train_loss": 0.1421789526939392, "global_step": 246064, "epoch": 2764, "lr": 2.1501968148577634e-06}
{"train_loss": 0.05980324372649193, "global_step": 246065, "epoch": 2764, "lr": 2.150028635961909e-06}
{"train_loss": 0.015653634443879128, "global_step": 246066, "epoch": 2764, "lr": 2.1498604634988984e-06}
{"train_loss": 0.031150566413998604, "global_step": 246067, "epoch": 2764, "lr": 2.149692297468753e-06}
{"train_loss": 0.016914576292037964, "global_step": 246068, "epoch": 2764, "lr": 2.1495241378715005e-06}
{"train_loss": 0.024934688583016396, "global_step": 246069, "epoch": 2764, "lr": 2.1493559847071574e-06}
{"train_loss": 0.06694582849740982, "global_step": 246070, "epoch": 2764, "lr": 2.149187837975758e-06}
{"train_loss": 0.018505243584513664, "global_step": 246071, "epoch": 2764, "lr": 2.1490196976773124e-06}
{"train_loss": 0.0316971018910408, "global_step": 246072, "epoch": 2764, "lr": 2.148851563811849e-06}
{"train_loss": 0.04190848767757416, "global_step": 246073, "epoch": 2764, "lr": 2.1486834363793894e-06}
{"train_loss": 0.027946339920163155, "global_step": 246074, "epoch": 2764, "lr": 2.1485153153799566e-06}
{"train_loss": 0.039385709911584854, "global_step": 246075, "epoch": 2764, "lr": 2.1483472008135717e-06}
{"train_loss": 0.025928247720003128, "global_step": 246076, "epoch": 2764, "lr": 2.148179092680258e-06}
{"train_loss": 0.027946989983320236, "global_step": 246077, "epoch": 2764, "lr": 2.1480109909800373e-06}
{"train_loss": 0.011633191257715225, "global_step": 246078, "epoch": 2764, "lr": 2.147842895712937e-06}
{"train_loss": 0.04166412353515625, "global_step": 246079, "epoch": 2764, "lr": 2.1476748068789687e-06}
{"train_loss": 0.02839987352490425, "global_step": 246080, "epoch": 2764, "lr": 2.147506724478171e-06}
{"train_loss": 0.04918767139315605, "global_step": 246081, "epoch": 2764, "lr": 2.1473386485105552e-06}
{"train_loss": 0.054743681102991104, "global_step": 246082, "epoch": 2764, "lr": 2.147170578976149e-06}
{"train_loss": 0.06957344710826874, "global_step": 246083, "epoch": 2764, "lr": 2.147002515874963e-06}
{"train_loss": 0.04401659115814091, "global_step": 246084, "epoch": 2764, "lr": 2.1468344592070365e-06, "val_loss": 9.478592872619629}
{"train_loss": 0.05085330083966255, "global_step": 246085, "epoch": 2765, "lr": 2.146666408972381e-06}
{"train_loss": 0.07456514984369278, "global_step": 246086, "epoch": 2765, "lr": 2.1464983651710236e-06}
{"train_loss": 0.023563610389828682, "global_step": 246087, "epoch": 2765, "lr": 2.1463303278029924e-06}
{"train_loss": 0.018082531169056892, "global_step": 246088, "epoch": 2765, "lr": 2.1461622968682925e-06}
{"train_loss": 0.0323500856757164, "global_step": 246089, "epoch": 2765, "lr": 2.1459942723669692e-06}
{"train_loss": 0.061781592667102814, "global_step": 246090, "epoch": 2765, "lr": 2.145826254299027e-06}
{"train_loss": 0.04841215908527374, "global_step": 246091, "epoch": 2765, "lr": 2.1456582426645e-06}
{"train_loss": 0.023014919832348824, "global_step": 246092, "epoch": 2765, "lr": 2.145490237463399e-06}
{"train_loss": 0.0522046834230423, "global_step": 246093, "epoch": 2765, "lr": 2.1453222386957626e-06}
{"train_loss": 0.08267462998628616, "global_step": 246094, "epoch": 2765, "lr": 2.145154246361597e-06}
{"train_loss": 0.08656737208366394, "global_step": 246095, "epoch": 2765, "lr": 2.1449862604609404e-06}
{"train_loss": 0.040461670607328415, "global_step": 246096, "epoch": 2765, "lr": 2.1448182809938043e-06}
{"train_loss": 0.053366076201200485, "global_step": 246097, "epoch": 2765, "lr": 2.1446503079602053e-06}
{"train_loss": 0.022332491353154182, "global_step": 246098, "epoch": 2765, "lr": 2.144482341360188e-06}
{"train_loss": 0.024642102420330048, "global_step": 246099, "epoch": 2765, "lr": 2.1443143811937515e-06}
{"train_loss": 0.039853841066360474, "global_step": 246100, "epoch": 2765, "lr": 2.1441464274609357e-06}
{"train_loss": 0.07904075086116791, "global_step": 246101, "epoch": 2765, "lr": 2.1439784801617456e-06}
{"train_loss": 0.04889526590704918, "global_step": 246102, "epoch": 2765, "lr": 2.143810539296226e-06}
{"train_loss": 0.03819046542048454, "global_step": 246103, "epoch": 2765, "lr": 2.1436426048643875e-06}
{"train_loss": 0.05593640357255936, "global_step": 246104, "epoch": 2765, "lr": 2.143474676866242e-06}
{"train_loss": 0.028630252927541733, "global_step": 246105, "epoch": 2765, "lr": 2.143306755301827e-06}
{"train_loss": 0.049507927149534225, "global_step": 246106, "epoch": 2765, "lr": 2.143138840171166e-06}
{"train_loss": 0.0288598220795393, "global_step": 246107, "epoch": 2765, "lr": 2.142970931474275e-06}
{"train_loss": 0.07357658445835114, "global_step": 246108, "epoch": 2765, "lr": 2.1428030292111767e-06}
{"train_loss": 0.08789325505495071, "global_step": 246109, "epoch": 2765, "lr": 2.1426351333818927e-06}
{"train_loss": 0.043975360691547394, "global_step": 246110, "epoch": 2765, "lr": 2.142467243986457e-06}
{"train_loss": 0.09687931090593338, "global_step": 246111, "epoch": 2765, "lr": 2.1422993610248743e-06}
{"train_loss": 0.04199249669909477, "global_step": 246112, "epoch": 2765, "lr": 2.142131484497184e-06}
{"train_loss": 0.09990416467189789, "global_step": 246113, "epoch": 2765, "lr": 2.1419636144033917e-06}
{"train_loss": 0.022731082513928413, "global_step": 246114, "epoch": 2765, "lr": 2.141795750743536e-06}
{"train_loss": 0.04969125613570213, "global_step": 246115, "epoch": 2765, "lr": 2.141627893517628e-06}
{"train_loss": 0.034255314618349075, "global_step": 246116, "epoch": 2765, "lr": 2.1414600427257013e-06}
{"train_loss": 0.04425441101193428, "global_step": 246117, "epoch": 2765, "lr": 2.141292198367767e-06}
{"train_loss": 0.049998898059129715, "global_step": 246118, "epoch": 2765, "lr": 2.1411243604438524e-06}
{"train_loss": 0.05043953284621239, "global_step": 246119, "epoch": 2765, "lr": 2.14095652895398e-06}
{"train_loss": 0.0613386295735836, "global_step": 246120, "epoch": 2765, "lr": 2.1407887038981778e-06}
{"train_loss": 0.030413873493671417, "global_step": 246121, "epoch": 2765, "lr": 2.1406208852764565e-06}
{"train_loss": 0.0811307355761528, "global_step": 246122, "epoch": 2765, "lr": 2.140453073088844e-06}
{"train_loss": 0.04410100355744362, "global_step": 246123, "epoch": 2765, "lr": 2.1402852673353735e-06}
{"train_loss": 0.043378569185733795, "global_step": 246124, "epoch": 2765, "lr": 2.1401174680160507e-06}
{"train_loss": 0.04248429089784622, "global_step": 246125, "epoch": 2765, "lr": 2.139949675130909e-06}
{"train_loss": 0.0365343801677227, "global_step": 246126, "epoch": 2765, "lr": 2.1397818886799648e-06}
{"train_loss": 0.020390916615724564, "global_step": 246127, "epoch": 2765, "lr": 2.1396141086632515e-06}
{"train_loss": 0.045428771525621414, "global_step": 246128, "epoch": 2765, "lr": 2.1394463350807746e-06}
{"train_loss": 0.07747430354356766, "global_step": 246129, "epoch": 2765, "lr": 2.1392785679325734e-06}
{"train_loss": 0.03948870301246643, "global_step": 246130, "epoch": 2765, "lr": 2.1391108072186526e-06}
{"train_loss": 0.08910483121871948, "global_step": 246131, "epoch": 2765, "lr": 2.1389430529390574e-06}
{"train_loss": 0.015456881374120712, "global_step": 246132, "epoch": 2765, "lr": 2.1387753050937876e-06}
{"train_loss": 0.04883969947695732, "global_step": 246133, "epoch": 2765, "lr": 2.1386075636828815e-06}
{"train_loss": 0.05080568045377731, "global_step": 246134, "epoch": 2765, "lr": 2.138439828706351e-06}
{"train_loss": 0.032534241676330566, "global_step": 246135, "epoch": 2765, "lr": 2.1382721001642345e-06}
{"train_loss": 0.013746300712227821, "global_step": 246136, "epoch": 2765, "lr": 2.138104378056538e-06}
{"train_loss": 0.03279072418808937, "global_step": 246137, "epoch": 2765, "lr": 2.1379366623832885e-06}
{"train_loss": 0.047884322702884674, "global_step": 246138, "epoch": 2765, "lr": 2.1377689531445144e-06}
{"train_loss": 0.03671478107571602, "global_step": 246139, "epoch": 2765, "lr": 2.1376012503402266e-06}
{"train_loss": 0.03975115716457367, "global_step": 246140, "epoch": 2765, "lr": 2.1374335539704636e-06}
{"train_loss": 0.0583958625793457, "global_step": 246141, "epoch": 2765, "lr": 2.1372658640352316e-06}
{"train_loss": 0.07495203614234924, "global_step": 246142, "epoch": 2765, "lr": 2.1370981805345638e-06}
{"train_loss": 0.0498071052134037, "global_step": 246143, "epoch": 2765, "lr": 2.1369305034684817e-06}
{"train_loss": 0.04575642943382263, "global_step": 246144, "epoch": 2765, "lr": 2.1367628328370027e-06}
{"train_loss": 0.03214147314429283, "global_step": 246145, "epoch": 2765, "lr": 2.1365951686401543e-06}
{"train_loss": 0.055079858750104904, "global_step": 246146, "epoch": 2765, "lr": 2.1364275108779586e-06}
{"train_loss": 0.0739712342619896, "global_step": 246147, "epoch": 2765, "lr": 2.1362598595504325e-06}
{"train_loss": 0.052485279738903046, "global_step": 246148, "epoch": 2765, "lr": 2.136092214657609e-06}
{"train_loss": 0.020881598815321922, "global_step": 246149, "epoch": 2765, "lr": 2.1359245761994995e-06}
{"train_loss": 0.06248418614268303, "global_step": 246150, "epoch": 2765, "lr": 2.135756944176137e-06}
{"train_loss": 0.044238362461328506, "global_step": 246151, "epoch": 2765, "lr": 2.1355893185875274e-06}
{"train_loss": 0.045010633766651154, "global_step": 246152, "epoch": 2765, "lr": 2.135421699433715e-06}
{"train_loss": 0.026959918439388275, "global_step": 246153, "epoch": 2765, "lr": 2.1352540867147054e-06}
{"train_loss": 0.023314934223890305, "global_step": 246154, "epoch": 2765, "lr": 2.135086480430537e-06}
{"train_loss": 0.0514276921749115, "global_step": 246155, "epoch": 2765, "lr": 2.1349188805812103e-06}
{"train_loss": 0.034215398132801056, "global_step": 246156, "epoch": 2765, "lr": 2.1347512871667697e-06}
{"train_loss": 0.036126282066106796, "global_step": 246157, "epoch": 2765, "lr": 2.134583700187226e-06}
{"train_loss": 0.08260410279035568, "global_step": 246158, "epoch": 2765, "lr": 2.134416119642596e-06}
{"train_loss": 0.032749563455581665, "global_step": 246159, "epoch": 2765, "lr": 2.1342485455329244e-06}
{"train_loss": 0.04243077337741852, "global_step": 246160, "epoch": 2765, "lr": 2.1340809778582103e-06}
{"train_loss": 0.04102989658713341, "global_step": 246161, "epoch": 2765, "lr": 2.133913416618488e-06}
{"train_loss": 0.06186605244874954, "global_step": 246162, "epoch": 2765, "lr": 2.1337458618137795e-06}
{"train_loss": 0.02274923212826252, "global_step": 246163, "epoch": 2765, "lr": 2.1335783134441068e-06}
{"train_loss": 0.015486490912735462, "global_step": 246164, "epoch": 2765, "lr": 2.133410771509481e-06}
{"train_loss": 0.04737440496683121, "global_step": 246165, "epoch": 2765, "lr": 2.133243236009946e-06}
{"train_loss": 0.04011019319295883, "global_step": 246166, "epoch": 2765, "lr": 2.1330757069455087e-06}
{"train_loss": 0.048137813806533813, "global_step": 246167, "epoch": 2765, "lr": 2.1329081843161957e-06}
{"train_loss": 0.06663236767053604, "global_step": 246168, "epoch": 2765, "lr": 2.1327406681220296e-06}
{"train_loss": 0.05757084861397743, "global_step": 246169, "epoch": 2765, "lr": 2.1325731583630383e-06}
{"train_loss": 0.022404799237847328, "global_step": 246170, "epoch": 2765, "lr": 2.132405655039227e-06}
{"train_loss": 0.05026780813932419, "global_step": 246171, "epoch": 2765, "lr": 2.1322381581506455e-06}
{"train_loss": 0.05759107694029808, "global_step": 246172, "epoch": 2765, "lr": 2.132070667697289e-06}
{"train_loss": 0.04709302739690194, "global_step": 246173, "epoch": 2765, "lr": 2.1319031836792015e-06, "val_loss": 9.564146041870117, "train_action_mse_error": 21.765295028686523}
{"train_loss": 0.011097602546215057, "global_step": 246174, "epoch": 2766, "lr": 2.131735706096394e-06}
{"train_loss": 0.03378789499402046, "global_step": 246175, "epoch": 2766, "lr": 2.131568234948883e-06}
{"train_loss": 0.06044662371277809, "global_step": 246176, "epoch": 2766, "lr": 2.1314007702367077e-06}
{"train_loss": 0.07901467382907867, "global_step": 246177, "epoch": 2766, "lr": 2.131233311959879e-06}
{"train_loss": 0.06464394181966782, "global_step": 246178, "epoch": 2766, "lr": 2.1310658601184254e-06}
{"train_loss": 0.06828556954860687, "global_step": 246179, "epoch": 2766, "lr": 2.130898414712362e-06}
{"train_loss": 0.029091117903590202, "global_step": 246180, "epoch": 2766, "lr": 2.1307309757417184e-06}
{"train_loss": 0.03938128799200058, "global_step": 246181, "epoch": 2766, "lr": 2.1305635432065096e-06}
{"train_loss": 0.07175232470035553, "global_step": 246182, "epoch": 2766, "lr": 2.1303961171067755e-06}
{"train_loss": 0.07045801728963852, "global_step": 246183, "epoch": 2766, "lr": 2.13022869744251e-06}
{"train_loss": 0.061511680483818054, "global_step": 246184, "epoch": 2766, "lr": 2.1300612842137635e-06}
{"train_loss": 0.020983561873435974, "global_step": 246185, "epoch": 2766, "lr": 2.129893877420541e-06}
{"train_loss": 0.014965717680752277, "global_step": 246186, "epoch": 2766, "lr": 2.1297264770628765e-06}
{"train_loss": 0.019265854731202126, "global_step": 246187, "epoch": 2766, "lr": 2.129559083140781e-06}
{"train_loss": 0.07917425781488419, "global_step": 246188, "epoch": 2766, "lr": 2.129391695654287e-06}
{"train_loss": 0.03262152150273323, "global_step": 246189, "epoch": 2766, "lr": 2.129224314603412e-06}
{"train_loss": 0.043522678315639496, "global_step": 246190, "epoch": 2766, "lr": 2.1290569399881777e-06}
{"train_loss": 0.028551872819662094, "global_step": 246191, "epoch": 2766, "lr": 2.1288895718086065e-06}
{"train_loss": 0.014103841967880726, "global_step": 246192, "epoch": 2766, "lr": 2.128722210064732e-06}
{"train_loss": 0.03252805396914482, "global_step": 246193, "epoch": 2766, "lr": 2.1285548547565535e-06}
{"train_loss": 0.017588282003998756, "global_step": 246194, "epoch": 2766, "lr": 2.1283875058841107e-06}
{"train_loss": 0.042579978704452515, "global_step": 246195, "epoch": 2766, "lr": 2.128220163447431e-06}
{"train_loss": 0.03317331522703171, "global_step": 246196, "epoch": 2766, "lr": 2.1280528274465194e-06}
{"train_loss": 0.05442836880683899, "global_step": 246197, "epoch": 2766, "lr": 2.1278854978814155e-06}
{"train_loss": 0.06434988975524902, "global_step": 246198, "epoch": 2766, "lr": 2.12771817475213e-06}
{"train_loss": 0.026615742594003677, "global_step": 246199, "epoch": 2766, "lr": 2.1275508580586913e-06}
{"train_loss": 0.06700838357210159, "global_step": 246200, "epoch": 2766, "lr": 2.127383547801115e-06}
{"train_loss": 0.08136304467916489, "global_step": 246201, "epoch": 2766, "lr": 2.127216243979435e-06}
{"train_loss": 0.05923354998230934, "global_step": 246202, "epoch": 2766, "lr": 2.1270489465936625e-06}
{"train_loss": 0.03753708302974701, "global_step": 246203, "epoch": 2766, "lr": 2.126881655643831e-06}
{"train_loss": 0.04917950555682182, "global_step": 246204, "epoch": 2766, "lr": 2.126714371129951e-06}
{"train_loss": 0.020600486546754837, "global_step": 246205, "epoch": 2766, "lr": 2.1265470930520505e-06}
{"train_loss": 0.03001369908452034, "global_step": 246206, "epoch": 2766, "lr": 2.1263798214101516e-06}
{"train_loss": 0.04471202567219734, "global_step": 246207, "epoch": 2766, "lr": 2.126212556204282e-06}
{"train_loss": 0.032035090029239655, "global_step": 246208, "epoch": 2766, "lr": 2.1260452974344537e-06}
{"train_loss": 0.02473096363246441, "global_step": 246209, "epoch": 2766, "lr": 2.1258780451007043e-06}
{"train_loss": 0.03950870409607887, "global_step": 246210, "epoch": 2766, "lr": 2.12571079920304e-06}
{"train_loss": 0.024076765403151512, "global_step": 246211, "epoch": 2766, "lr": 2.1255435597414943e-06}
{"train_loss": 0.05035786330699921, "global_step": 246212, "epoch": 2766, "lr": 2.1253763267160832e-06}
{"train_loss": 0.03145831078290939, "global_step": 246213, "epoch": 2766, "lr": 2.125209100126829e-06}
{"train_loss": 0.024221146479249, "global_step": 246214, "epoch": 2766, "lr": 2.125041879973766e-06}
{"train_loss": 0.0505308099091053, "global_step": 246215, "epoch": 2766, "lr": 2.1248746662568987e-06}
{"train_loss": 0.06717698276042938, "global_step": 246216, "epoch": 2766, "lr": 2.1247074589762604e-06}
{"train_loss": 0.05678796023130417, "global_step": 246217, "epoch": 2766, "lr": 2.124540258131874e-06}
{"train_loss": 0.038158200681209564, "global_step": 246218, "epoch": 2766, "lr": 2.1243730637237614e-06}
{"train_loss": 0.02675248496234417, "global_step": 246219, "epoch": 2766, "lr": 2.1242058757519333e-06}
{"train_loss": 0.06993944197893143, "global_step": 246220, "epoch": 2766, "lr": 2.1240386942164346e-06}
{"train_loss": 0.05103001743555069, "global_step": 246221, "epoch": 2766, "lr": 2.123871519117265e-06}
{"train_loss": 0.03392766788601875, "global_step": 246222, "epoch": 2766, "lr": 2.123704350454464e-06}
{"train_loss": 0.05505197122693062, "global_step": 246223, "epoch": 2766, "lr": 2.123537188228042e-06}
{"train_loss": 0.07557206600904465, "global_step": 246224, "epoch": 2766, "lr": 2.1233700324380325e-06}
{"train_loss": 0.03051515482366085, "global_step": 246225, "epoch": 2766, "lr": 2.123202883084452e-06}
{"train_loss": 0.0608881339430809, "global_step": 246226, "epoch": 2766, "lr": 2.1230357401673228e-06}
{"train_loss": 0.02999621257185936, "global_step": 246227, "epoch": 2766, "lr": 2.122868603686662e-06}
{"train_loss": 0.056141577661037445, "global_step": 246228, "epoch": 2766, "lr": 2.1227014736425077e-06}
{"train_loss": 0.04418810456991196, "global_step": 246229, "epoch": 2766, "lr": 2.122534350034866e-06}
{"train_loss": 0.03556825593113899, "global_step": 246230, "epoch": 2766, "lr": 2.1223672328637702e-06}
{"train_loss": 0.02618272975087166, "global_step": 246231, "epoch": 2766, "lr": 2.122200122129231e-06}
{"train_loss": 0.02543729729950428, "global_step": 246232, "epoch": 2766, "lr": 2.122033017831282e-06}
{"train_loss": 0.040957000106573105, "global_step": 246233, "epoch": 2766, "lr": 2.121865919969945e-06}
{"train_loss": 0.02329256571829319, "global_step": 246234, "epoch": 2766, "lr": 2.121698828545232e-06}
{"train_loss": 0.01630360633134842, "global_step": 246235, "epoch": 2766, "lr": 2.1215317435571867e-06}
{"train_loss": 0.04485301300883293, "global_step": 246236, "epoch": 2766, "lr": 2.121364665005804e-06}
{"train_loss": 0.030769288539886475, "global_step": 246237, "epoch": 2766, "lr": 2.121197592891133e-06}
{"train_loss": 0.06577169895172119, "global_step": 246238, "epoch": 2766, "lr": 2.121030527213175e-06}
{"train_loss": 0.0774792805314064, "global_step": 246239, "epoch": 2766, "lr": 2.120863467971962e-06}
{"train_loss": 0.032369837164878845, "global_step": 246240, "epoch": 2766, "lr": 2.120696415167517e-06}
{"train_loss": 0.059665337204933167, "global_step": 246241, "epoch": 2766, "lr": 2.1205293687998618e-06}
{"train_loss": 0.05543489009141922, "global_step": 246242, "epoch": 2766, "lr": 2.1203623288690132e-06}
{"train_loss": 0.03074592538177967, "global_step": 246243, "epoch": 2766, "lr": 2.120195295375005e-06}
{"train_loss": 0.06999696046113968, "global_step": 246244, "epoch": 2766, "lr": 2.120028268317842e-06}
{"train_loss": 0.02173461578786373, "global_step": 246245, "epoch": 2766, "lr": 2.119861247697569e-06}
{"train_loss": 0.06274969130754471, "global_step": 246246, "epoch": 2766, "lr": 2.1196942335141913e-06}
{"train_loss": 0.036815185099840164, "global_step": 246247, "epoch": 2766, "lr": 2.1195272257677424e-06}
{"train_loss": 0.02746851183474064, "global_step": 246248, "epoch": 2766, "lr": 2.1193602244582335e-06}
{"train_loss": 0.03884587064385414, "global_step": 246249, "epoch": 2766, "lr": 2.1191932295856977e-06}
{"train_loss": 0.0639587938785553, "global_step": 246250, "epoch": 2766, "lr": 2.119026241150146e-06}
{"train_loss": 0.027220390737056732, "global_step": 246251, "epoch": 2766, "lr": 2.1188592591516175e-06}
{"train_loss": 0.06134861707687378, "global_step": 246252, "epoch": 2766, "lr": 2.118692283590118e-06}
{"train_loss": 0.042275067418813705, "global_step": 246253, "epoch": 2766, "lr": 2.118525314465675e-06}
{"train_loss": 0.05299472063779831, "global_step": 246254, "epoch": 2766, "lr": 2.1183583517783213e-06}
{"train_loss": 0.03784642368555069, "global_step": 246255, "epoch": 2766, "lr": 2.118191395528063e-06}
{"train_loss": 0.015563467517495155, "global_step": 246256, "epoch": 2766, "lr": 2.1180244457149334e-06}
{"train_loss": 0.028211984783411026, "global_step": 246257, "epoch": 2766, "lr": 2.117857502338949e-06}
{"train_loss": 0.03179797902703285, "global_step": 246258, "epoch": 2766, "lr": 2.1176905654001376e-06}
{"train_loss": 0.058904893696308136, "global_step": 246259, "epoch": 2766, "lr": 2.117523634898516e-06}
{"train_loss": 0.06253635138273239, "global_step": 246260, "epoch": 2766, "lr": 2.1173567108341175e-06}
{"train_loss": 0.03215605393052101, "global_step": 246261, "epoch": 2766, "lr": 2.1171897932069475e-06}
{"train_loss": 0.04383774903299433, "global_step": 246262, "epoch": 2766, "lr": 2.1170228820170444e-06, "val_loss": 9.482213020324707}
{"train_loss": 0.0493144728243351, "global_step": 246263, "epoch": 2767, "lr": 2.1168559772644203e-06}
{"train_loss": 0.056485142558813095, "global_step": 246264, "epoch": 2767, "lr": 2.116689078949102e-06}
{"train_loss": 0.03918066248297691, "global_step": 246265, "epoch": 2767, "lr": 2.1165221870711072e-06}
{"train_loss": 0.028384601697325706, "global_step": 246266, "epoch": 2767, "lr": 2.116355301630474e-06}
{"train_loss": 0.03922557458281517, "global_step": 246267, "epoch": 2767, "lr": 2.1161884226272024e-06}
{"train_loss": 0.02214733138680458, "global_step": 246268, "epoch": 2767, "lr": 2.116021550061326e-06}
{"train_loss": 0.05650826171040535, "global_step": 246269, "epoch": 2767, "lr": 2.115854683932872e-06}
{"train_loss": 0.01384774036705494, "global_step": 246270, "epoch": 2767, "lr": 2.1156878242418523e-06}
{"train_loss": 0.024936627596616745, "global_step": 246271, "epoch": 2767, "lr": 2.1155209709882994e-06}
{"train_loss": 0.04024915769696236, "global_step": 246272, "epoch": 2767, "lr": 2.1153541241722307e-06}
{"train_loss": 0.02040979638695717, "global_step": 246273, "epoch": 2767, "lr": 2.115187283793668e-06}
{"train_loss": 0.05662255361676216, "global_step": 246274, "epoch": 2767, "lr": 2.115020449852634e-06}
{"train_loss": 0.029851585626602173, "global_step": 246275, "epoch": 2767, "lr": 2.1148536223491554e-06}
{"train_loss": 0.03165889531373978, "global_step": 246276, "epoch": 2767, "lr": 2.1146868012832446e-06}
{"train_loss": 0.07250655442476273, "global_step": 246277, "epoch": 2767, "lr": 2.114519986654939e-06}
{"train_loss": 0.03143361583352089, "global_step": 246278, "epoch": 2767, "lr": 2.11435317846424e-06}
{"train_loss": 0.03463473916053772, "global_step": 246279, "epoch": 2767, "lr": 2.114186376711197e-06}
{"train_loss": 0.02869722805917263, "global_step": 246280, "epoch": 2767, "lr": 2.11401958139581e-06}
{"train_loss": 0.0350516140460968, "global_step": 246281, "epoch": 2767, "lr": 2.113852792518112e-06}
{"train_loss": 0.05182021111249924, "global_step": 246282, "epoch": 2767, "lr": 2.1136860100781198e-06}
{"train_loss": 0.059827301651239395, "global_step": 246283, "epoch": 2767, "lr": 2.1135192340758614e-06}
{"train_loss": 0.0400557816028595, "global_step": 246284, "epoch": 2767, "lr": 2.1133524645113533e-06}
{"train_loss": 0.016268517822027206, "global_step": 246285, "epoch": 2767, "lr": 2.113185701384629e-06}
{"train_loss": 0.029535368084907532, "global_step": 246286, "epoch": 2767, "lr": 2.113018944695694e-06}
{"train_loss": 0.031607501208782196, "global_step": 246287, "epoch": 2767, "lr": 2.1128521944445866e-06}
{"train_loss": 0.0547746941447258, "global_step": 246288, "epoch": 2767, "lr": 2.1126854506313187e-06}
{"train_loss": 0.03165794163942337, "global_step": 246289, "epoch": 2767, "lr": 2.1125187132559176e-06}
{"train_loss": 0.012833709828555584, "global_step": 246290, "epoch": 2767, "lr": 2.112351982318411e-06}
{"train_loss": 0.031114596873521805, "global_step": 246291, "epoch": 2767, "lr": 2.1121852578188052e-06}
{"train_loss": 0.06584714353084564, "global_step": 246292, "epoch": 2767, "lr": 2.112018539757138e-06}
{"train_loss": 0.04509682208299637, "global_step": 246293, "epoch": 2767, "lr": 2.111851828133421e-06}
{"train_loss": 0.046192143112421036, "global_step": 246294, "epoch": 2767, "lr": 2.111685122947693e-06}
{"train_loss": 0.04365475848317146, "global_step": 246295, "epoch": 2767, "lr": 2.1115184241999544e-06}
{"train_loss": 0.03111700341105461, "global_step": 246296, "epoch": 2767, "lr": 2.1113517318902432e-06}
{"train_loss": 0.024904850870370865, "global_step": 246297, "epoch": 2767, "lr": 2.111185046018571e-06}
{"train_loss": 0.04792750999331474, "global_step": 246298, "epoch": 2767, "lr": 2.111018366584977e-06}
{"train_loss": 0.06554777175188065, "global_step": 246299, "epoch": 2767, "lr": 2.110851693589461e-06}
{"train_loss": 0.04867394641041756, "global_step": 246300, "epoch": 2767, "lr": 2.1106850270320665e-06}
{"train_loss": 0.006245281081646681, "global_step": 246301, "epoch": 2767, "lr": 2.1105183669128003e-06}
{"train_loss": 0.051841333508491516, "global_step": 246302, "epoch": 2767, "lr": 2.110351713231695e-06}
{"train_loss": 0.047417741268873215, "global_step": 246303, "epoch": 2767, "lr": 2.110185065988768e-06}
{"train_loss": 0.034172751009464264, "global_step": 246304, "epoch": 2767, "lr": 2.1100184251840403e-06}
{"train_loss": 0.059345707297325134, "global_step": 246305, "epoch": 2767, "lr": 2.1098517908175407e-06}
{"train_loss": 0.09852343797683716, "global_step": 246306, "epoch": 2767, "lr": 2.1096851628892856e-06}
{"train_loss": 0.05995458737015724, "global_step": 246307, "epoch": 2767, "lr": 2.109518541399308e-06}
{"train_loss": 0.07179185748100281, "global_step": 246308, "epoch": 2767, "lr": 2.109351926347608e-06}
{"train_loss": 0.035118646919727325, "global_step": 246309, "epoch": 2767, "lr": 2.1091853177342357e-06}
{"train_loss": 0.08240844309329987, "global_step": 246310, "epoch": 2767, "lr": 2.1090187155591856e-06}
{"train_loss": 0.08710977435112, "global_step": 246311, "epoch": 2767, "lr": 2.1088521198225074e-06}
{"train_loss": 0.040733203291893005, "global_step": 246312, "epoch": 2767, "lr": 2.1086855305242016e-06}
{"train_loss": 0.04820401594042778, "global_step": 246313, "epoch": 2767, "lr": 2.1085189476643064e-06}
{"train_loss": 0.03179824352264404, "global_step": 246314, "epoch": 2767, "lr": 2.1083523712428275e-06}
{"train_loss": 0.040559832006692886, "global_step": 246315, "epoch": 2767, "lr": 2.1081858012598045e-06}
{"train_loss": 0.04270340874791145, "global_step": 246316, "epoch": 2767, "lr": 2.1080192377152475e-06}
{"train_loss": 0.06599078327417374, "global_step": 246317, "epoch": 2767, "lr": 2.1078526806091904e-06}
{"train_loss": 0.05827966704964638, "global_step": 246318, "epoch": 2767, "lr": 2.107686129941644e-06}
{"train_loss": 0.032455384731292725, "global_step": 246319, "epoch": 2767, "lr": 2.1075195857126363e-06}
{"train_loss": 0.033236946910619736, "global_step": 246320, "epoch": 2767, "lr": 2.107353047922184e-06}
{"train_loss": 0.03918921574950218, "global_step": 246321, "epoch": 2767, "lr": 2.1071865165703197e-06}
{"train_loss": 0.0635901466012001, "global_step": 246322, "epoch": 2767, "lr": 2.107019991657061e-06}
{"train_loss": 0.04020213708281517, "global_step": 246323, "epoch": 2767, "lr": 2.1068534731824295e-06}
{"train_loss": 0.03249967470765114, "global_step": 246324, "epoch": 2767, "lr": 2.106686961146448e-06}
{"train_loss": 0.02172200381755829, "global_step": 246325, "epoch": 2767, "lr": 2.106520455549138e-06}
{"train_loss": 0.048074983060359955, "global_step": 246326, "epoch": 2767, "lr": 2.106353956390522e-06}
{"train_loss": 0.04102586582303047, "global_step": 246327, "epoch": 2767, "lr": 2.1061874636706225e-06}
{"train_loss": 0.02861582301557064, "global_step": 246328, "epoch": 2767, "lr": 2.106020977389461e-06}
{"train_loss": 0.041192587465047836, "global_step": 246329, "epoch": 2767, "lr": 2.105854497547066e-06}
{"train_loss": 0.013171682134270668, "global_step": 246330, "epoch": 2767, "lr": 2.105688024143454e-06}
{"train_loss": 0.030689511448144913, "global_step": 246331, "epoch": 2767, "lr": 2.105521557178647e-06}
{"train_loss": 0.027318423613905907, "global_step": 246332, "epoch": 2767, "lr": 2.1053550966526723e-06}
{"train_loss": 0.043010953813791275, "global_step": 246333, "epoch": 2767, "lr": 2.105188642565542e-06}
{"train_loss": 0.03737761825323105, "global_step": 246334, "epoch": 2767, "lr": 2.105022194917289e-06}
{"train_loss": 0.0596124492585659, "global_step": 246335, "epoch": 2767, "lr": 2.1048557537079292e-06}
{"train_loss": 0.027170268818736076, "global_step": 246336, "epoch": 2767, "lr": 2.1046893189374918e-06}
{"train_loss": 0.03296941518783569, "global_step": 246337, "epoch": 2767, "lr": 2.104522890605992e-06}
{"train_loss": 0.039168521761894226, "global_step": 246338, "epoch": 2767, "lr": 2.1043564687134586e-06}
{"train_loss": 0.03862413391470909, "global_step": 246339, "epoch": 2767, "lr": 2.1041900532599077e-06}
{"train_loss": 0.0424879714846611, "global_step": 246340, "epoch": 2767, "lr": 2.104023644245362e-06}
{"train_loss": 0.07550614327192307, "global_step": 246341, "epoch": 2767, "lr": 2.1038572416698545e-06}
{"train_loss": 0.03723347932100296, "global_step": 246342, "epoch": 2767, "lr": 2.10369084553339e-06}
{"train_loss": 0.02047581411898136, "global_step": 246343, "epoch": 2767, "lr": 2.103524455836009e-06}
{"train_loss": 0.028295136988162994, "global_step": 246344, "epoch": 2767, "lr": 2.1033580725777214e-06}
{"train_loss": 0.0589812733232975, "global_step": 246345, "epoch": 2767, "lr": 2.103191695758555e-06}
{"train_loss": 0.067283995449543, "global_step": 246346, "epoch": 2767, "lr": 2.103025325378527e-06}
{"train_loss": 0.0270409993827343, "global_step": 246347, "epoch": 2767, "lr": 2.1028589614376703e-06}
{"train_loss": 0.04711894690990448, "global_step": 246348, "epoch": 2767, "lr": 2.102692603935996e-06}
{"train_loss": 0.07958859205245972, "global_step": 246349, "epoch": 2767, "lr": 2.1025262528735325e-06}
{"train_loss": 0.0681075006723404, "global_step": 246350, "epoch": 2767, "lr": 2.1023599082502955e-06}
{"train_loss": 0.04308582953794786, "global_step": 246351, "epoch": 2767, "lr": 2.102193570066324e-06, "val_loss": 9.551796913146973}
{"train_loss": 0.060136035084724426, "global_step": 246352, "epoch": 2768, "lr": 2.1020272383216187e-06}
{"train_loss": 0.05907755717635155, "global_step": 246353, "epoch": 2768, "lr": 2.1018609130162183e-06}
{"train_loss": 0.05299984663724899, "global_step": 246354, "epoch": 2768, "lr": 2.1016945941501275e-06}
{"train_loss": 0.010413945652544498, "global_step": 246355, "epoch": 2768, "lr": 2.1015282817233917e-06}
{"train_loss": 0.05521627515554428, "global_step": 246356, "epoch": 2768, "lr": 2.101361975736016e-06}
{"train_loss": 0.016045117750763893, "global_step": 246357, "epoch": 2768, "lr": 2.1011956761880336e-06}
{"train_loss": 0.060882776975631714, "global_step": 246358, "epoch": 2768, "lr": 2.1010293830794558e-06}
{"train_loss": 0.06175483390688896, "global_step": 246359, "epoch": 2768, "lr": 2.1008630964103104e-06}
{"train_loss": 0.039257802069187164, "global_step": 246360, "epoch": 2768, "lr": 2.100696816180625e-06}
{"train_loss": 0.07872039824724197, "global_step": 246361, "epoch": 2768, "lr": 2.100530542390411e-06}
{"train_loss": 0.031966257840394974, "global_step": 246362, "epoch": 2768, "lr": 2.1003642750397067e-06}
{"train_loss": 0.020155860111117363, "global_step": 246363, "epoch": 2768, "lr": 2.1001980141285237e-06}
{"train_loss": 0.04416274279356003, "global_step": 246364, "epoch": 2768, "lr": 2.100031759656873e-06}
{"train_loss": 0.009284410625696182, "global_step": 246365, "epoch": 2768, "lr": 2.0998655116248045e-06}
{"train_loss": 0.058476369827985764, "global_step": 246366, "epoch": 2768, "lr": 2.0996992700323127e-06}
{"train_loss": 0.028637630864977837, "global_step": 246367, "epoch": 2768, "lr": 2.099533034879442e-06}
{"train_loss": 0.022813430055975914, "global_step": 246368, "epoch": 2768, "lr": 2.099366806166203e-06}
{"train_loss": 0.03865253925323486, "global_step": 246369, "epoch": 2768, "lr": 2.0992005838926133e-06}
{"train_loss": 0.047580745071172714, "global_step": 246370, "epoch": 2768, "lr": 2.099034368058711e-06}
{"train_loss": 0.027417989447712898, "global_step": 246371, "epoch": 2768, "lr": 2.098868158664502e-06}
{"train_loss": 0.01392114907503128, "global_step": 246372, "epoch": 2768, "lr": 2.0987019557100198e-06}
{"train_loss": 0.0350441075861454, "global_step": 246373, "epoch": 2768, "lr": 2.09853575919528e-06}
{"train_loss": 0.0642881989479065, "global_step": 246374, "epoch": 2768, "lr": 2.0983695691203174e-06}
{"train_loss": 0.035950351506471634, "global_step": 246375, "epoch": 2768, "lr": 2.098203385485137e-06}
{"train_loss": 0.05159692466259003, "global_step": 246376, "epoch": 2768, "lr": 2.098037208289766e-06}
{"train_loss": 0.05790017545223236, "global_step": 246377, "epoch": 2768, "lr": 2.0978710375342382e-06}
{"train_loss": 0.0810590386390686, "global_step": 246378, "epoch": 2768, "lr": 2.0977048732185646e-06}
{"train_loss": 0.044015754014253616, "global_step": 246379, "epoch": 2768, "lr": 2.097538715342773e-06}
{"train_loss": 0.05040543153882027, "global_step": 246380, "epoch": 2768, "lr": 2.0973725639068798e-06}
{"train_loss": 0.003358446527272463, "global_step": 246381, "epoch": 2768, "lr": 2.097206418910913e-06}
{"train_loss": 0.02261953055858612, "global_step": 246382, "epoch": 2768, "lr": 2.097040280354895e-06}
{"train_loss": 0.05148253217339516, "global_step": 246383, "epoch": 2768, "lr": 2.096874148238842e-06}
{"train_loss": 0.04773103445768356, "global_step": 246384, "epoch": 2768, "lr": 2.096708022562782e-06}
{"train_loss": 0.0664757788181305, "global_step": 246385, "epoch": 2768, "lr": 2.0965419033267377e-06}
{"train_loss": 0.039152391254901886, "global_step": 246386, "epoch": 2768, "lr": 2.0963757905307246e-06}
{"train_loss": 0.0692446306347847, "global_step": 246387, "epoch": 2768, "lr": 2.096209684174777e-06}
{"train_loss": 0.02973668836057186, "global_step": 246388, "epoch": 2768, "lr": 2.0960435842589054e-06}
{"train_loss": 0.04841307923197746, "global_step": 246389, "epoch": 2768, "lr": 2.095877490783138e-06}
{"train_loss": 0.03648087754845619, "global_step": 246390, "epoch": 2768, "lr": 2.095711403747497e-06}
{"train_loss": 0.031920384615659714, "global_step": 246391, "epoch": 2768, "lr": 2.0955453231520037e-06}
{"train_loss": 0.04186802729964256, "global_step": 246392, "epoch": 2768, "lr": 2.095379248996676e-06}
{"train_loss": 0.021957915276288986, "global_step": 246393, "epoch": 2768, "lr": 2.0952131812815467e-06}
{"train_loss": 0.01795784942805767, "global_step": 246394, "epoch": 2768, "lr": 2.095047120006627e-06}
{"train_loss": 0.03308584541082382, "global_step": 246395, "epoch": 2768, "lr": 2.094881065171944e-06}
{"train_loss": 0.05187111720442772, "global_step": 246396, "epoch": 2768, "lr": 2.0947150167775266e-06}
{"train_loss": 0.08538664132356644, "global_step": 246397, "epoch": 2768, "lr": 2.094548974823385e-06}
{"train_loss": 0.021277375519275665, "global_step": 246398, "epoch": 2768, "lr": 2.0943829393095527e-06}
{"train_loss": 0.05192803964018822, "global_step": 246399, "epoch": 2768, "lr": 2.094216910236041e-06}
{"train_loss": 0.07656532526016235, "global_step": 246400, "epoch": 2768, "lr": 2.094050887602883e-06}
{"train_loss": 0.049299802631139755, "global_step": 246401, "epoch": 2768, "lr": 2.0938848714100955e-06}
{"train_loss": 0.030088480561971664, "global_step": 246402, "epoch": 2768, "lr": 2.093718861657701e-06}
{"train_loss": 0.018746638670563698, "global_step": 246403, "epoch": 2768, "lr": 2.0935528583457208e-06}
{"train_loss": 0.03211667388677597, "global_step": 246404, "epoch": 2768, "lr": 2.093386861474178e-06}
{"train_loss": 0.04130616784095764, "global_step": 246405, "epoch": 2768, "lr": 2.0932208710431e-06}
{"train_loss": 0.048352498561143875, "global_step": 246406, "epoch": 2768, "lr": 2.0930548870525036e-06}
{"train_loss": 0.010571381077170372, "global_step": 246407, "epoch": 2768, "lr": 2.092888909502405e-06}
{"train_loss": 0.03726648911833763, "global_step": 246408, "epoch": 2768, "lr": 2.092722938392838e-06}
{"train_loss": 0.07019061595201492, "global_step": 246409, "epoch": 2768, "lr": 2.0925569737238195e-06}
{"train_loss": 0.021984528750181198, "global_step": 246410, "epoch": 2768, "lr": 2.0923910154953764e-06}
{"train_loss": 0.0625549778342247, "global_step": 246411, "epoch": 2768, "lr": 2.0922250637075203e-06}
{"train_loss": 0.047820061445236206, "global_step": 246412, "epoch": 2768, "lr": 2.092059118360279e-06}
{"train_loss": 0.05305267497897148, "global_step": 246413, "epoch": 2768, "lr": 2.0918931794536854e-06}
{"train_loss": 0.05663188546895981, "global_step": 246414, "epoch": 2768, "lr": 2.0917272469877456e-06}
{"train_loss": 0.025027044117450714, "global_step": 246415, "epoch": 2768, "lr": 2.0915613209624987e-06}
{"train_loss": 0.03884445130825043, "global_step": 246416, "epoch": 2768, "lr": 2.0913954013779492e-06}
{"train_loss": 0.04315955936908722, "global_step": 246417, "epoch": 2768, "lr": 2.091229488234131e-06}
{"train_loss": 0.022444332018494606, "global_step": 246418, "epoch": 2768, "lr": 2.091063581531061e-06}
{"train_loss": 0.10325374454259872, "global_step": 246419, "epoch": 2768, "lr": 2.0908976812687662e-06}
{"train_loss": 0.009872171096503735, "global_step": 246420, "epoch": 2768, "lr": 2.0907317874472643e-06}
{"train_loss": 0.06128985807299614, "global_step": 246421, "epoch": 2768, "lr": 2.090565900066582e-06}
{"train_loss": 0.08805839717388153, "global_step": 246422, "epoch": 2768, "lr": 2.090400019126737e-06}
{"train_loss": 0.06119673699140549, "global_step": 246423, "epoch": 2768, "lr": 2.090234144627756e-06}
{"train_loss": 0.03789254650473595, "global_step": 246424, "epoch": 2768, "lr": 2.0900682765696566e-06}
{"train_loss": 0.05024181678891182, "global_step": 246425, "epoch": 2768, "lr": 2.089902414952466e-06}
{"train_loss": 0.054441675543785095, "global_step": 246426, "epoch": 2768, "lr": 2.089736559776201e-06}
{"train_loss": 0.02842474915087223, "global_step": 246427, "epoch": 2768, "lr": 2.0895707110408893e-06}
{"train_loss": 0.025472424924373627, "global_step": 246428, "epoch": 2768, "lr": 2.0894048687465475e-06}
{"train_loss": 0.06776217371225357, "global_step": 246429, "epoch": 2768, "lr": 2.089239032893209e-06}
{"train_loss": 0.015147162601351738, "global_step": 246430, "epoch": 2768, "lr": 2.089073203480879e-06}
{"train_loss": 0.047276563942432404, "global_step": 246431, "epoch": 2768, "lr": 2.088907380509586e-06}
{"train_loss": 0.07843214273452759, "global_step": 246432, "epoch": 2768, "lr": 2.0887415639793684e-06}
{"train_loss": 0.02601728029549122, "global_step": 246433, "epoch": 2768, "lr": 2.0885757538902264e-06}
{"train_loss": 0.02729438617825508, "global_step": 246434, "epoch": 2768, "lr": 2.088409950242198e-06}
{"train_loss": 0.08705989271402359, "global_step": 246435, "epoch": 2768, "lr": 2.0882441530352905e-06}
{"train_loss": 0.12443851679563522, "global_step": 246436, "epoch": 2768, "lr": 2.0880783622695467e-06}
{"train_loss": 0.05119972676038742, "global_step": 246437, "epoch": 2768, "lr": 2.087912577944967e-06}
{"train_loss": 0.05440139397978783, "global_step": 246438, "epoch": 2768, "lr": 2.087746800061585e-06}
{"train_loss": 0.053274426609277725, "global_step": 246439, "epoch": 2768, "lr": 2.0875810286194288e-06}
{"train_loss": 0.04505279867941241, "global_step": 246440, "epoch": 2768, "lr": 2.0874152636185027e-06, "val_loss": 9.566083908081055}
{"train_loss": 0.07235638797283173, "global_step": 246441, "epoch": 2769, "lr": 2.0872495050588468e-06}
{"train_loss": 0.026208698749542236, "global_step": 246442, "epoch": 2769, "lr": 2.087083752940472e-06}
{"train_loss": 0.04597015306353569, "global_step": 246443, "epoch": 2769, "lr": 2.086918007263411e-06}
{"train_loss": 0.03402726724743843, "global_step": 246444, "epoch": 2769, "lr": 2.0867522680276753e-06}
{"train_loss": 0.04775957763195038, "global_step": 246445, "epoch": 2769, "lr": 2.0865865352332924e-06}
{"train_loss": 0.04451530799269676, "global_step": 246446, "epoch": 2769, "lr": 2.086420808880285e-06}
{"train_loss": 0.02218402735888958, "global_step": 246447, "epoch": 2769, "lr": 2.0862550889686694e-06}
{"train_loss": 0.0460795983672142, "global_step": 246448, "epoch": 2769, "lr": 2.0860893754984733e-06}
{"train_loss": 0.0356026217341423, "global_step": 246449, "epoch": 2769, "lr": 2.0859236684697248e-06}
{"train_loss": 0.03511558100581169, "global_step": 246450, "epoch": 2769, "lr": 2.085757967882435e-06}
{"train_loss": 0.10010085999965668, "global_step": 246451, "epoch": 2769, "lr": 2.0855922737366364e-06}
{"train_loss": 0.05483163893222809, "global_step": 246452, "epoch": 2769, "lr": 2.0854265860323406e-06}
{"train_loss": 0.08453668653964996, "global_step": 246453, "epoch": 2769, "lr": 2.085260904769576e-06}
{"train_loss": 0.043783921748399734, "global_step": 246454, "epoch": 2769, "lr": 2.085095229948364e-06}
{"train_loss": 0.01882309652864933, "global_step": 246455, "epoch": 2769, "lr": 2.0849295615687323e-06}
{"train_loss": 0.02412581443786621, "global_step": 246456, "epoch": 2769, "lr": 2.084763899630687e-06}
{"train_loss": 0.05870820954442024, "global_step": 246457, "epoch": 2769, "lr": 2.084598244134273e-06}
{"train_loss": 0.030504897236824036, "global_step": 246458, "epoch": 2769, "lr": 2.084432595079494e-06}
{"train_loss": 0.04634859040379524, "global_step": 246459, "epoch": 2769, "lr": 2.0842669524663794e-06}
{"train_loss": 0.03399454429745674, "global_step": 246460, "epoch": 2769, "lr": 2.0841013162949506e-06}
{"train_loss": 0.02442781627178192, "global_step": 246461, "epoch": 2769, "lr": 2.083935686565236e-06}
{"train_loss": 0.02356371097266674, "global_step": 246462, "epoch": 2769, "lr": 2.083770063277246e-06}
{"train_loss": 0.06694816052913666, "global_step": 246463, "epoch": 2769, "lr": 2.0836044464310145e-06}
{"train_loss": 0.034798599779605865, "global_step": 246464, "epoch": 2769, "lr": 2.083438836026552e-06}
{"train_loss": 0.04960199445486069, "global_step": 246465, "epoch": 2769, "lr": 2.0832732320638925e-06}
{"train_loss": 0.019828226417303085, "global_step": 246466, "epoch": 2769, "lr": 2.0831076345430468e-06}
{"train_loss": 0.049287568777799606, "global_step": 246467, "epoch": 2769, "lr": 2.0829420434640425e-06}
{"train_loss": 0.04218817874789238, "global_step": 246468, "epoch": 2769, "lr": 2.0827764588269127e-06}
{"train_loss": 0.04267343506217003, "global_step": 246469, "epoch": 2769, "lr": 2.082610880631658e-06}
{"train_loss": 0.05509655177593231, "global_step": 246470, "epoch": 2769, "lr": 2.0824453088783225e-06}
{"train_loss": 0.06792668253183365, "global_step": 246471, "epoch": 2769, "lr": 2.082279743566912e-06}
{"train_loss": 0.05054755508899689, "global_step": 246472, "epoch": 2769, "lr": 2.082114184697459e-06}
{"train_loss": 0.03733573853969574, "global_step": 246473, "epoch": 2769, "lr": 2.0819486322699754e-06}
{"train_loss": 0.040935080498456955, "global_step": 246474, "epoch": 2769, "lr": 2.0817830862845e-06}
{"train_loss": 0.02433936670422554, "global_step": 246475, "epoch": 2769, "lr": 2.0816175467410327e-06}
{"train_loss": 0.04171167314052582, "global_step": 246476, "epoch": 2769, "lr": 2.0814520136396177e-06}
{"train_loss": 0.06505756825208664, "global_step": 246477, "epoch": 2769, "lr": 2.081286486980261e-06}
{"train_loss": 0.021821318194270134, "global_step": 246478, "epoch": 2769, "lr": 2.081120966762995e-06}
{"train_loss": 0.07156061381101608, "global_step": 246479, "epoch": 2769, "lr": 2.0809554529878427e-06}
{"train_loss": 0.009695122949779034, "global_step": 246480, "epoch": 2769, "lr": 2.0807899456548097e-06}
{"train_loss": 0.036777231842279434, "global_step": 246481, "epoch": 2769, "lr": 2.0806244447639456e-06}
{"train_loss": 0.03358624503016472, "global_step": 246482, "epoch": 2769, "lr": 2.080458950315245e-06}
{"train_loss": 0.05592948943376541, "global_step": 246483, "epoch": 2769, "lr": 2.0802934623087465e-06}
{"train_loss": 0.04926515743136406, "global_step": 246484, "epoch": 2769, "lr": 2.080127980744473e-06}
{"train_loss": 0.04391343146562576, "global_step": 246485, "epoch": 2769, "lr": 2.0799625056224348e-06}
{"train_loss": 0.08216051757335663, "global_step": 246486, "epoch": 2769, "lr": 2.0797970369426602e-06}
{"train_loss": 0.055906400084495544, "global_step": 246487, "epoch": 2769, "lr": 2.079631574705182e-06}
{"train_loss": 0.042876169085502625, "global_step": 246488, "epoch": 2769, "lr": 2.0794661189100063e-06}
{"train_loss": 0.05543685704469681, "global_step": 246489, "epoch": 2769, "lr": 2.079300669557166e-06}
{"train_loss": 0.018746521323919296, "global_step": 246490, "epoch": 2769, "lr": 2.079135226646678e-06}
{"train_loss": 0.02227611467242241, "global_step": 246491, "epoch": 2769, "lr": 2.078969790178564e-06}
{"train_loss": 0.04000040888786316, "global_step": 246492, "epoch": 2769, "lr": 2.0788043601528526e-06}
{"train_loss": 0.011268037371337414, "global_step": 246493, "epoch": 2769, "lr": 2.0786389365695603e-06}
{"train_loss": 0.04631896689534187, "global_step": 246494, "epoch": 2769, "lr": 2.0784735194287086e-06}
{"train_loss": 0.0706566795706749, "global_step": 246495, "epoch": 2769, "lr": 2.0783081087303256e-06}
{"train_loss": 0.061183828860521317, "global_step": 246496, "epoch": 2769, "lr": 2.078142704474428e-06}
{"train_loss": 0.037964362651109695, "global_step": 246497, "epoch": 2769, "lr": 2.0779773066610443e-06}
{"train_loss": 0.0527757965028286, "global_step": 246498, "epoch": 2769, "lr": 2.077811915290184e-06}
{"train_loss": 0.024770785123109818, "global_step": 246499, "epoch": 2769, "lr": 2.0776465303618875e-06}
{"train_loss": 0.0339091494679451, "global_step": 246500, "epoch": 2769, "lr": 2.0774811518761594e-06}
{"train_loss": 0.06041623651981354, "global_step": 246501, "epoch": 2769, "lr": 2.0773157798330334e-06}
{"train_loss": 0.05985505133867264, "global_step": 246502, "epoch": 2769, "lr": 2.077150414232526e-06}
{"train_loss": 0.08487515896558762, "global_step": 246503, "epoch": 2769, "lr": 2.0769850550746595e-06}
{"train_loss": 0.03008902259171009, "global_step": 246504, "epoch": 2769, "lr": 2.076819702359467e-06}
{"train_loss": 0.04609448462724686, "global_step": 246505, "epoch": 2769, "lr": 2.0766543560869488e-06}
{"train_loss": 0.08488670736551285, "global_step": 246506, "epoch": 2769, "lr": 2.076489016257155e-06}
{"train_loss": 0.053382907062768936, "global_step": 246507, "epoch": 2769, "lr": 2.076323682870085e-06}
{"train_loss": 0.0446413978934288, "global_step": 246508, "epoch": 2769, "lr": 2.0761583559257726e-06}
{"train_loss": 0.04893060401082039, "global_step": 246509, "epoch": 2769, "lr": 2.0759930354242283e-06}
{"train_loss": 0.05334082618355751, "global_step": 246510, "epoch": 2769, "lr": 2.0758277213654974e-06}
{"train_loss": 0.016120364889502525, "global_step": 246511, "epoch": 2769, "lr": 2.075662413749574e-06}
{"train_loss": 0.06310207396745682, "global_step": 246512, "epoch": 2769, "lr": 2.075497112576502e-06}
{"train_loss": 0.05609084293246269, "global_step": 246513, "epoch": 2769, "lr": 2.0753318178462875e-06}
{"train_loss": 0.004256510641425848, "global_step": 246514, "epoch": 2769, "lr": 2.0751665295589694e-06}
{"train_loss": 0.05769333615899086, "global_step": 246515, "epoch": 2769, "lr": 2.075001247714553e-06}
{"train_loss": 0.05865500122308731, "global_step": 246516, "epoch": 2769, "lr": 2.0748359723130774e-06}
{"train_loss": 0.09023049473762512, "global_step": 246517, "epoch": 2769, "lr": 2.074670703354553e-06}
{"train_loss": 0.03952937200665474, "global_step": 246518, "epoch": 2769, "lr": 2.0745054408389976e-06}
{"train_loss": 0.06065855175256729, "global_step": 246519, "epoch": 2769, "lr": 2.074340184766449e-06}
{"train_loss": 0.042251475155353546, "global_step": 246520, "epoch": 2769, "lr": 2.074174935136919e-06}
{"train_loss": 0.037057824432849884, "global_step": 246521, "epoch": 2769, "lr": 2.074009691950435e-06}
{"train_loss": 0.05369819700717926, "global_step": 246522, "epoch": 2769, "lr": 2.073844455207008e-06}
{"train_loss": 0.018855595961213112, "global_step": 246523, "epoch": 2769, "lr": 2.0736792249066774e-06}
{"train_loss": 0.044328927993774414, "global_step": 246524, "epoch": 2769, "lr": 2.073514001049448e-06}
{"train_loss": 0.023262759670615196, "global_step": 246525, "epoch": 2769, "lr": 2.073348783635359e-06}
{"train_loss": 0.06166975200176239, "global_step": 246526, "epoch": 2769, "lr": 2.0731835726644165e-06}
{"train_loss": 0.08451040089130402, "global_step": 246527, "epoch": 2769, "lr": 2.073018368136653e-06}
{"train_loss": 0.034356094896793365, "global_step": 246528, "epoch": 2769, "lr": 2.072853170052086e-06}
{"train_loss": 0.04543959397006403, "global_step": 246529, "epoch": 2769, "lr": 2.0726879784107478e-06, "val_loss": 9.538334846496582}
{"train_loss": 0.020748766139149666, "global_step": 246530, "epoch": 2770, "lr": 2.0725227932126447e-06}
{"train_loss": 0.02350730262696743, "global_step": 246531, "epoch": 2770, "lr": 2.0723576144578095e-06}
{"train_loss": 0.08213763684034348, "global_step": 246532, "epoch": 2770, "lr": 2.072192442146259e-06}
{"train_loss": 0.030584610998630524, "global_step": 246533, "epoch": 2770, "lr": 2.0720272762780213e-06}
{"train_loss": 0.0449402742087841, "global_step": 246534, "epoch": 2770, "lr": 2.0718621168531128e-06}
{"train_loss": 0.01599741168320179, "global_step": 246535, "epoch": 2770, "lr": 2.0716969638715665e-06}
{"train_loss": 0.04876132309436798, "global_step": 246536, "epoch": 2770, "lr": 2.071531817333383e-06}
{"train_loss": 0.024601725861430168, "global_step": 246537, "epoch": 2770, "lr": 2.071366677238612e-06}
{"train_loss": 0.03288900479674339, "global_step": 246538, "epoch": 2770, "lr": 2.071201543587248e-06}
{"train_loss": 0.034310463815927505, "global_step": 246539, "epoch": 2770, "lr": 2.0710364163793294e-06}
{"train_loss": 0.02306997776031494, "global_step": 246540, "epoch": 2770, "lr": 2.0708712956148846e-06}
{"train_loss": 0.05976273864507675, "global_step": 246541, "epoch": 2770, "lr": 2.070706181293919e-06}
{"train_loss": 0.03680575639009476, "global_step": 246542, "epoch": 2770, "lr": 2.0705410734164655e-06}
{"train_loss": 0.05179948732256889, "global_step": 246543, "epoch": 2770, "lr": 2.070375971982541e-06}
{"train_loss": 0.04512280598282814, "global_step": 246544, "epoch": 2770, "lr": 2.070210876992179e-06}
{"train_loss": 0.027359193190932274, "global_step": 246545, "epoch": 2770, "lr": 2.070045788445385e-06}
{"train_loss": 0.07523924112319946, "global_step": 246546, "epoch": 2770, "lr": 2.069880706342192e-06}
{"train_loss": 0.05260153114795685, "global_step": 246547, "epoch": 2770, "lr": 2.069715630682617e-06}
{"train_loss": 0.07575520873069763, "global_step": 246548, "epoch": 2770, "lr": 2.069550561466688e-06}
{"train_loss": 0.02385244332253933, "global_step": 246549, "epoch": 2770, "lr": 2.0693854986944206e-06}
{"train_loss": 0.06792893260717392, "global_step": 246550, "epoch": 2770, "lr": 2.0692204423658435e-06}
{"train_loss": 0.0636151134967804, "global_step": 246551, "epoch": 2770, "lr": 2.0690553924809675e-06}
{"train_loss": 0.06429903209209442, "global_step": 246552, "epoch": 2770, "lr": 2.0688903490398316e-06}
{"train_loss": 0.03212910145521164, "global_step": 246553, "epoch": 2770, "lr": 2.0687253120424466e-06}
{"train_loss": 0.11217837780714035, "global_step": 246554, "epoch": 2770, "lr": 2.0685602814888348e-06}
{"train_loss": 0.08123552799224854, "global_step": 246555, "epoch": 2770, "lr": 2.06839525737903e-06}
{"train_loss": 0.06170612573623657, "global_step": 246556, "epoch": 2770, "lr": 2.0682302397130315e-06}
{"train_loss": 0.08598535507917404, "global_step": 246557, "epoch": 2770, "lr": 2.0680652284908897e-06}
{"train_loss": 0.03670146316289902, "global_step": 246558, "epoch": 2770, "lr": 2.067900223712599e-06}
{"train_loss": 0.05320773646235466, "global_step": 246559, "epoch": 2770, "lr": 2.0677352253782035e-06}
{"train_loss": 0.04155784845352173, "global_step": 246560, "epoch": 2770, "lr": 2.067570233487709e-06}
{"train_loss": 0.06416469812393188, "global_step": 246561, "epoch": 2770, "lr": 2.0674052480411542e-06}
{"train_loss": 0.042632248252630234, "global_step": 246562, "epoch": 2770, "lr": 2.0672402690385452e-06}
{"train_loss": 0.06332270801067352, "global_step": 246563, "epoch": 2770, "lr": 2.0670752964799202e-06}
{"train_loss": 0.059503037482500076, "global_step": 246564, "epoch": 2770, "lr": 2.0669103303652847e-06}
{"train_loss": 0.05885716527700424, "global_step": 246565, "epoch": 2770, "lr": 2.0667453706946726e-06}
{"train_loss": 0.025205742567777634, "global_step": 246566, "epoch": 2770, "lr": 2.0665804174681003e-06}
{"train_loss": 0.042313531041145325, "global_step": 246567, "epoch": 2770, "lr": 2.0664154706855955e-06}
{"train_loss": 0.08389720320701599, "global_step": 246568, "epoch": 2770, "lr": 2.066250530347169e-06}
{"train_loss": 0.03077658824622631, "global_step": 246569, "epoch": 2770, "lr": 2.06608559645286e-06}
{"train_loss": 0.07102843374013901, "global_step": 246570, "epoch": 2770, "lr": 2.0659206690026744e-06}
{"train_loss": 0.03390130028128624, "global_step": 246571, "epoch": 2770, "lr": 2.0657557479966506e-06}
{"train_loss": 0.04599933326244354, "global_step": 246572, "epoch": 2770, "lr": 2.065590833434794e-06}
{"train_loss": 0.05137721076607704, "global_step": 246573, "epoch": 2770, "lr": 2.065425925317138e-06}
{"train_loss": 0.01636684499680996, "global_step": 246574, "epoch": 2770, "lr": 2.0652610236436943e-06}
{"train_loss": 0.07500060647726059, "global_step": 246575, "epoch": 2770, "lr": 2.0650961284144954e-06}
{"train_loss": 0.03622042015194893, "global_step": 246576, "epoch": 2770, "lr": 2.0649312396295694e-06}
{"train_loss": 0.04345046356320381, "global_step": 246577, "epoch": 2770, "lr": 2.0647663572889165e-06}
{"train_loss": 0.024271609261631966, "global_step": 246578, "epoch": 2770, "lr": 2.0646014813925806e-06}
{"train_loss": 0.04096714407205582, "global_step": 246579, "epoch": 2770, "lr": 2.064436611940568e-06}
{"train_loss": 0.046107109636068344, "global_step": 246580, "epoch": 2770, "lr": 2.064271748932911e-06}
{"train_loss": 0.05188668519258499, "global_step": 246581, "epoch": 2770, "lr": 2.0641068923696273e-06}
{"train_loss": 0.03123510256409645, "global_step": 246582, "epoch": 2770, "lr": 2.063942042250744e-06}
{"train_loss": 0.03748909756541252, "global_step": 246583, "epoch": 2770, "lr": 2.0637771985762722e-06}
{"train_loss": 0.032766737043857574, "global_step": 246584, "epoch": 2770, "lr": 2.063612361346251e-06}
{"train_loss": 0.021257853135466576, "global_step": 246585, "epoch": 2770, "lr": 2.063447530560686e-06}
{"train_loss": 0.026375994086265564, "global_step": 246586, "epoch": 2770, "lr": 2.06328270621961e-06}
{"train_loss": 0.03290695324540138, "global_step": 246587, "epoch": 2770, "lr": 2.063117888323035e-06}
{"train_loss": 0.03889111056923866, "global_step": 246588, "epoch": 2770, "lr": 2.0629530768709993e-06}
{"train_loss": 0.025121545419096947, "global_step": 246589, "epoch": 2770, "lr": 2.062788271863503e-06}
{"train_loss": 0.056303489953279495, "global_step": 246590, "epoch": 2770, "lr": 2.06262347330059e-06}
{"train_loss": 0.08097606897354126, "global_step": 246591, "epoch": 2770, "lr": 2.0624586811822668e-06}
{"train_loss": 0.018824651837348938, "global_step": 246592, "epoch": 2770, "lr": 2.0622938955085714e-06}
{"train_loss": 0.09700091183185577, "global_step": 246593, "epoch": 2770, "lr": 2.0621291162795042e-06}
{"train_loss": 0.05542176589369774, "global_step": 246594, "epoch": 2770, "lr": 2.0619643434951097e-06}
{"train_loss": 0.08157198876142502, "global_step": 246595, "epoch": 2770, "lr": 2.0617995771553987e-06}
{"train_loss": 0.04393566772341728, "global_step": 246596, "epoch": 2770, "lr": 2.0616348172603884e-06}
{"train_loss": 0.01882408931851387, "global_step": 246597, "epoch": 2770, "lr": 2.0614700638101116e-06}
{"train_loss": 0.07039318233728409, "global_step": 246598, "epoch": 2770, "lr": 2.061305316804579e-06}
{"train_loss": 0.049263738095760345, "global_step": 246599, "epoch": 2770, "lr": 2.0611405762438306e-06}
{"train_loss": 0.03244321048259735, "global_step": 246600, "epoch": 2770, "lr": 2.0609758421278656e-06}
{"train_loss": 0.031467799097299576, "global_step": 246601, "epoch": 2770, "lr": 2.0608111144567287e-06}
{"train_loss": 0.023369470611214638, "global_step": 246602, "epoch": 2770, "lr": 2.0606463932304254e-06}
{"train_loss": 0.03678429499268532, "global_step": 246603, "epoch": 2770, "lr": 2.0604816784489887e-06}
{"train_loss": 0.07093384116888046, "global_step": 246604, "epoch": 2770, "lr": 2.0603169701124304e-06}
{"train_loss": 0.034697677940130234, "global_step": 246605, "epoch": 2770, "lr": 2.060152268220783e-06}
{"train_loss": 0.044975947588682175, "global_step": 246606, "epoch": 2770, "lr": 2.059987572774058e-06}
{"train_loss": 0.02191341295838356, "global_step": 246607, "epoch": 2770, "lr": 2.059822883772289e-06}
{"train_loss": 0.037841856479644775, "global_step": 246608, "epoch": 2770, "lr": 2.0596582012154863e-06}
{"train_loss": 0.022890081629157066, "global_step": 246609, "epoch": 2770, "lr": 2.059493525103684e-06}
{"train_loss": 0.019086532294750214, "global_step": 246610, "epoch": 2770, "lr": 2.059328855436893e-06}
{"train_loss": 0.04581829160451889, "global_step": 246611, "epoch": 2770, "lr": 2.0591641922151517e-06}
{"train_loss": 0.02609633095562458, "global_step": 246612, "epoch": 2770, "lr": 2.0589995354384605e-06}
{"train_loss": 0.03511251509189606, "global_step": 246613, "epoch": 2770, "lr": 2.0588348851068527e-06}
{"train_loss": 0.0812036544084549, "global_step": 246614, "epoch": 2770, "lr": 2.0586702412203563e-06}
{"train_loss": 0.03733519837260246, "global_step": 246615, "epoch": 2770, "lr": 2.058505603778982e-06}
{"train_loss": 0.06003730371594429, "global_step": 246616, "epoch": 2770, "lr": 2.058340972782763e-06}
{"train_loss": 0.02517942525446415, "global_step": 246617, "epoch": 2770, "lr": 2.0581763482317107e-06}
{"train_loss": 0.04587552246501606, "global_step": 246618, "epoch": 2770, "lr": 2.0580117301258526e-06, "val_loss": 9.523916244506836, "train_action_mse_error": 14.454998016357422}
{"train_loss": 0.098772794008255, "global_step": 246619, "epoch": 2771, "lr": 2.0578471184652113e-06}
{"train_loss": 0.028557024896144867, "global_step": 246620, "epoch": 2771, "lr": 2.0576825132498144e-06}
{"train_loss": 0.042771562933921814, "global_step": 246621, "epoch": 2771, "lr": 2.0575179144796673e-06}
{"train_loss": 0.022404659539461136, "global_step": 246622, "epoch": 2771, "lr": 2.057353322154809e-06}
{"train_loss": 0.030899809673428535, "global_step": 246623, "epoch": 2771, "lr": 2.0571887362752507e-06}
{"train_loss": 0.03093056008219719, "global_step": 246624, "epoch": 2771, "lr": 2.0570241568410253e-06}
{"train_loss": 0.05918193235993385, "global_step": 246625, "epoch": 2771, "lr": 2.056859583852144e-06}
{"train_loss": 0.044190581887960434, "global_step": 246626, "epoch": 2771, "lr": 2.0566950173086408e-06}
{"train_loss": 0.03407810255885124, "global_step": 246627, "epoch": 2771, "lr": 2.0565304572105203e-06}
{"train_loss": 0.04201251640915871, "global_step": 246628, "epoch": 2771, "lr": 2.056365903557822e-06}
{"train_loss": 0.07083472609519958, "global_step": 246629, "epoch": 2771, "lr": 2.0562013563505566e-06}
{"train_loss": 0.046035535633563995, "global_step": 246630, "epoch": 2771, "lr": 2.056036815588752e-06}
{"train_loss": 0.11597012728452682, "global_step": 246631, "epoch": 2771, "lr": 2.0558722812724306e-06}
{"train_loss": 0.03798237815499306, "global_step": 246632, "epoch": 2771, "lr": 2.0557077534016145e-06}
{"train_loss": 0.05818043276667595, "global_step": 246633, "epoch": 2771, "lr": 2.05554323197632e-06}
{"train_loss": 0.09112830460071564, "global_step": 246634, "epoch": 2771, "lr": 2.0553787169965755e-06}
{"train_loss": 0.04105589911341667, "global_step": 246635, "epoch": 2771, "lr": 2.0552142084624026e-06}
{"train_loss": 0.043066948652267456, "global_step": 246636, "epoch": 2771, "lr": 2.055049706373813e-06}
{"train_loss": 0.016139212995767593, "global_step": 246637, "epoch": 2771, "lr": 2.0548852107308505e-06}
{"train_loss": 0.04649851843714714, "global_step": 246638, "epoch": 2771, "lr": 2.0547207215335152e-06}
{"train_loss": 0.04230620712041855, "global_step": 246639, "epoch": 2771, "lr": 2.0545562387818407e-06}
{"train_loss": 0.028061075136065483, "global_step": 246640, "epoch": 2771, "lr": 2.0543917624758434e-06}
{"train_loss": 0.06018837168812752, "global_step": 246641, "epoch": 2771, "lr": 2.054227292615557e-06}
{"train_loss": 0.05211949720978737, "global_step": 246642, "epoch": 2771, "lr": 2.0540628292009866e-06}
{"train_loss": 0.02041013538837433, "global_step": 246643, "epoch": 2771, "lr": 2.0538983722321713e-06}
{"train_loss": 0.06706055998802185, "global_step": 246644, "epoch": 2771, "lr": 2.0537339217091165e-06}
{"train_loss": 0.01843799278140068, "global_step": 246645, "epoch": 2771, "lr": 2.0535694776318614e-06}
{"train_loss": 0.04543357342481613, "global_step": 246646, "epoch": 2771, "lr": 2.0534050400004114e-06}
{"train_loss": 0.0644133985042572, "global_step": 246647, "epoch": 2771, "lr": 2.053240608814805e-06}
{"train_loss": 0.036747537553310394, "global_step": 246648, "epoch": 2771, "lr": 2.053076184075048e-06}
{"train_loss": 0.041687995195388794, "global_step": 246649, "epoch": 2771, "lr": 2.052911765781168e-06}
{"train_loss": 0.026072954759001732, "global_step": 246650, "epoch": 2771, "lr": 2.0527473539331986e-06}
{"train_loss": 0.04524565860629082, "global_step": 246651, "epoch": 2771, "lr": 2.052582948531151e-06}
{"train_loss": 0.022588450461626053, "global_step": 246652, "epoch": 2771, "lr": 2.052418549575047e-06}
{"train_loss": 0.05139010772109032, "global_step": 246653, "epoch": 2771, "lr": 2.052254157064909e-06}
{"train_loss": 0.030973920598626137, "global_step": 246654, "epoch": 2771, "lr": 2.0520897710007704e-06}
{"train_loss": 0.06492702662944794, "global_step": 246655, "epoch": 2771, "lr": 2.0519253913826307e-06}
{"train_loss": 0.09819212555885315, "global_step": 246656, "epoch": 2771, "lr": 2.051761018210541e-06}
{"train_loss": 0.047886595129966736, "global_step": 246657, "epoch": 2771, "lr": 2.051596651484494e-06}
{"train_loss": 0.04285082966089249, "global_step": 246658, "epoch": 2771, "lr": 2.0514322912045304e-06}
{"train_loss": 0.05012161284685135, "global_step": 246659, "epoch": 2771, "lr": 2.0512679373706656e-06}
{"train_loss": 0.055067867040634155, "global_step": 246660, "epoch": 2771, "lr": 2.051103589982928e-06}
{"train_loss": 0.009197606705129147, "global_step": 246661, "epoch": 2771, "lr": 2.050939249041328e-06}
{"train_loss": 0.06057443469762802, "global_step": 246662, "epoch": 2771, "lr": 2.0507749145459053e-06}
{"train_loss": 0.012570631690323353, "global_step": 246663, "epoch": 2771, "lr": 2.0506105864966594e-06}
{"train_loss": 0.046977583318948746, "global_step": 246664, "epoch": 2771, "lr": 2.050446264893635e-06}
{"train_loss": 0.02802826650440693, "global_step": 246665, "epoch": 2771, "lr": 2.050281949736832e-06}
{"train_loss": 0.03666031360626221, "global_step": 246666, "epoch": 2771, "lr": 2.0501176410262945e-06}
{"train_loss": 0.03364359587430954, "global_step": 246667, "epoch": 2771, "lr": 2.0499533387620283e-06}
{"train_loss": 0.06487348675727844, "global_step": 246668, "epoch": 2771, "lr": 2.049789042944067e-06}
{"train_loss": 0.07712282985448837, "global_step": 246669, "epoch": 2771, "lr": 2.049624753572421e-06}
{"train_loss": 0.022904731333255768, "global_step": 246670, "epoch": 2771, "lr": 2.0494604706471243e-06}
{"train_loss": 0.06593643873929977, "global_step": 246671, "epoch": 2771, "lr": 2.0492961941681877e-06}
{"train_loss": 0.04795555770397186, "global_step": 246672, "epoch": 2771, "lr": 2.0491319241356445e-06}
{"train_loss": 0.058624669909477234, "global_step": 246673, "epoch": 2771, "lr": 2.048967660549511e-06}
{"train_loss": 0.022148549556732178, "global_step": 246674, "epoch": 2771, "lr": 2.0488034034098046e-06}
{"train_loss": 0.06153620779514313, "global_step": 246675, "epoch": 2771, "lr": 2.0486391527165584e-06}
{"train_loss": 0.030761167407035828, "global_step": 246676, "epoch": 2771, "lr": 2.0484749084697776e-06}
{"train_loss": 0.02418622374534607, "global_step": 246677, "epoch": 2771, "lr": 2.0483106706695066e-06}
{"train_loss": 0.0353546105325222, "global_step": 246678, "epoch": 2771, "lr": 2.048146439315746e-06}
{"train_loss": 0.022464165464043617, "global_step": 246679, "epoch": 2771, "lr": 2.0479822144085335e-06}
{"train_loss": 0.04297078400850296, "global_step": 246680, "epoch": 2771, "lr": 2.0478179959478814e-06}
{"train_loss": 0.02955528348684311, "global_step": 246681, "epoch": 2771, "lr": 2.047653783933823e-06}
{"train_loss": 0.037316471338272095, "global_step": 246682, "epoch": 2771, "lr": 2.047489578366363e-06}
{"train_loss": 0.06551183760166168, "global_step": 246683, "epoch": 2771, "lr": 2.0473253792455406e-06}
{"train_loss": 0.00911640003323555, "global_step": 246684, "epoch": 2771, "lr": 2.047161186571367e-06}
{"train_loss": 0.021033022552728653, "global_step": 246685, "epoch": 2771, "lr": 2.0469970003438697e-06}
{"train_loss": 0.03935309126973152, "global_step": 246686, "epoch": 2771, "lr": 2.0468328205630716e-06}
{"train_loss": 0.04094476252794266, "global_step": 246687, "epoch": 2771, "lr": 2.0466686472289887e-06}
{"train_loss": 0.08185817301273346, "global_step": 246688, "epoch": 2771, "lr": 2.046504480341649e-06}
{"train_loss": 0.04535567760467529, "global_step": 246689, "epoch": 2771, "lr": 2.046340319901069e-06}
{"train_loss": 0.05173944681882858, "global_step": 246690, "epoch": 2771, "lr": 2.046176165907282e-06}
{"train_loss": 0.05488615483045578, "global_step": 246691, "epoch": 2771, "lr": 2.0460120183602937e-06}
{"train_loss": 0.038700297474861145, "global_step": 246692, "epoch": 2771, "lr": 2.0458478772601377e-06}
{"train_loss": 0.027275769039988518, "global_step": 246693, "epoch": 2771, "lr": 2.04568374260683e-06}
{"train_loss": 0.019805599004030228, "global_step": 246694, "epoch": 2771, "lr": 2.0455196144004043e-06}
{"train_loss": 0.06593477725982666, "global_step": 246695, "epoch": 2771, "lr": 2.045355492640866e-06}
{"train_loss": 0.05868847295641899, "global_step": 246696, "epoch": 2771, "lr": 2.0451913773282484e-06}
{"train_loss": 0.02935180254280567, "global_step": 246697, "epoch": 2771, "lr": 2.0450272684625682e-06}
{"train_loss": 0.05616125091910362, "global_step": 246698, "epoch": 2771, "lr": 2.044863166043853e-06}
{"train_loss": 0.009662723168730736, "global_step": 246699, "epoch": 2771, "lr": 2.0446990700721146e-06}
{"train_loss": 0.06395245343446732, "global_step": 246700, "epoch": 2771, "lr": 2.044534980547391e-06}
{"train_loss": 0.04205112159252167, "global_step": 246701, "epoch": 2771, "lr": 2.0443708974696883e-06}
{"train_loss": 0.02791973203420639, "global_step": 246702, "epoch": 2771, "lr": 2.044206820839034e-06}
{"train_loss": 0.03844301402568817, "global_step": 246703, "epoch": 2771, "lr": 2.0440427506554562e-06}
{"train_loss": 0.07245636731386185, "global_step": 246704, "epoch": 2771, "lr": 2.0438786869189706e-06}
{"train_loss": 0.020233070477843285, "global_step": 246705, "epoch": 2771, "lr": 2.043714629629606e-06}
{"train_loss": 0.039625849574804306, "global_step": 246706, "epoch": 2771, "lr": 2.043550578787379e-06}
{"train_loss": 0.045018670473540766, "global_step": 246707, "epoch": 2771, "lr": 2.0433865343923053e-06, "val_loss": 9.586369514465332}
{"train_loss": 0.031373582780361176, "global_step": 246708, "epoch": 2772, "lr": 2.043222496444419e-06}
{"train_loss": 0.02402922697365284, "global_step": 246709, "epoch": 2772, "lr": 2.043058464943731e-06}
{"train_loss": 0.03745044395327568, "global_step": 246710, "epoch": 2772, "lr": 2.0428944398902806e-06}
{"train_loss": 0.017170706763863564, "global_step": 246711, "epoch": 2772, "lr": 2.0427304212840725e-06}
{"train_loss": 0.015535643324255943, "global_step": 246712, "epoch": 2772, "lr": 2.0425664091251294e-06}
{"train_loss": 0.04119161143898964, "global_step": 246713, "epoch": 2772, "lr": 2.0424024034134847e-06}
{"train_loss": 0.028067246079444885, "global_step": 246714, "epoch": 2772, "lr": 2.0422384041491495e-06}
{"train_loss": 0.04779941961169243, "global_step": 246715, "epoch": 2772, "lr": 2.0420744113321567e-06}
{"train_loss": 0.0839032530784607, "global_step": 246716, "epoch": 2772, "lr": 2.0419104249625174e-06}
{"train_loss": 0.0491979643702507, "global_step": 246717, "epoch": 2772, "lr": 2.0417464450402655e-06}
{"train_loss": 0.02066895179450512, "global_step": 246718, "epoch": 2772, "lr": 2.041582471565412e-06}
{"train_loss": 0.08662185817956924, "global_step": 246719, "epoch": 2772, "lr": 2.041418504537984e-06}
{"train_loss": 0.03552227094769478, "global_step": 246720, "epoch": 2772, "lr": 2.0412545439579987e-06}
{"train_loss": 0.054376933723688126, "global_step": 246721, "epoch": 2772, "lr": 2.041090589825484e-06}
{"train_loss": 0.012078475207090378, "global_step": 246722, "epoch": 2772, "lr": 2.0409266421404615e-06}
{"train_loss": 0.010608133859932423, "global_step": 246723, "epoch": 2772, "lr": 2.0407627009029484e-06}
{"train_loss": 0.030190162360668182, "global_step": 246724, "epoch": 2772, "lr": 2.0405987661129776e-06}
{"train_loss": 0.05351189151406288, "global_step": 246725, "epoch": 2772, "lr": 2.0404348377705552e-06}
{"train_loss": 0.08958908170461655, "global_step": 246726, "epoch": 2772, "lr": 2.0402709158757192e-06}
{"train_loss": 0.10766799747943878, "global_step": 246727, "epoch": 2772, "lr": 2.0401070004284815e-06}
{"train_loss": 0.029125019907951355, "global_step": 246728, "epoch": 2772, "lr": 2.0399430914288696e-06}
{"train_loss": 0.03952924534678459, "global_step": 246729, "epoch": 2772, "lr": 2.0397791888768945e-06}
{"train_loss": 0.04004782810807228, "global_step": 246730, "epoch": 2772, "lr": 2.039615292772595e-06}
{"train_loss": 0.03517182916402817, "global_step": 246731, "epoch": 2772, "lr": 2.039451403115977e-06}
{"train_loss": 0.038861531764268875, "global_step": 246732, "epoch": 2772, "lr": 2.0392875199070794e-06}
{"train_loss": 0.05359657108783722, "global_step": 246733, "epoch": 2772, "lr": 2.039123643145907e-06}
{"train_loss": 0.04001259803771973, "global_step": 246734, "epoch": 2772, "lr": 2.038959772832499e-06}
{"train_loss": 0.02906913310289383, "global_step": 246735, "epoch": 2772, "lr": 2.038795908966856e-06}
{"train_loss": 0.04551633447408676, "global_step": 246736, "epoch": 2772, "lr": 2.0386320515490274e-06}
{"train_loss": 0.037136636674404144, "global_step": 246737, "epoch": 2772, "lr": 2.038468200579008e-06}
{"train_loss": 0.025262057781219482, "global_step": 246738, "epoch": 2772, "lr": 2.038304356056836e-06}
{"train_loss": 0.06393394619226456, "global_step": 246739, "epoch": 2772, "lr": 2.0381405179825285e-06}
{"train_loss": 0.030745767056941986, "global_step": 246740, "epoch": 2772, "lr": 2.037976686356102e-06}
{"train_loss": 0.057058241218328476, "global_step": 246741, "epoch": 2772, "lr": 2.037812861177596e-06}
{"train_loss": 0.032459113746881485, "global_step": 246742, "epoch": 2772, "lr": 2.0376490424470153e-06}
{"train_loss": 0.040844887495040894, "global_step": 246743, "epoch": 2772, "lr": 2.0374852301643934e-06}
{"train_loss": 0.06849224865436554, "global_step": 246744, "epoch": 2772, "lr": 2.0373214243297467e-06}
{"train_loss": 0.03450578451156616, "global_step": 246745, "epoch": 2772, "lr": 2.037157624943092e-06}
{"train_loss": 0.00875547993928194, "global_step": 246746, "epoch": 2772, "lr": 2.0369938320044637e-06}
{"train_loss": 0.012913288548588753, "global_step": 246747, "epoch": 2772, "lr": 2.0368300455138713e-06}
{"train_loss": 0.018452562391757965, "global_step": 246748, "epoch": 2772, "lr": 2.036666265471343e-06}
{"train_loss": 0.03857406601309776, "global_step": 246749, "epoch": 2772, "lr": 2.036502491876907e-06}
{"train_loss": 0.04451411962509155, "global_step": 246750, "epoch": 2772, "lr": 2.0363387247305686e-06}
{"train_loss": 0.07227139919996262, "global_step": 246751, "epoch": 2772, "lr": 2.0361749640323667e-06}
{"train_loss": 0.04062435030937195, "global_step": 246752, "epoch": 2772, "lr": 2.0360112097823124e-06}
{"train_loss": 0.02801041677594185, "global_step": 246753, "epoch": 2772, "lr": 2.035847461980439e-06}
{"train_loss": 0.04926857352256775, "global_step": 246754, "epoch": 2772, "lr": 2.035683720626752e-06}
{"train_loss": 0.03125111386179924, "global_step": 246755, "epoch": 2772, "lr": 2.0355199857212903e-06}
{"train_loss": 0.05306996405124664, "global_step": 246756, "epoch": 2772, "lr": 2.0353562572640596e-06}
{"train_loss": 0.07161270827054977, "global_step": 246757, "epoch": 2772, "lr": 2.0351925352550926e-06}
{"train_loss": 0.03503558412194252, "global_step": 246758, "epoch": 2772, "lr": 2.035028819694418e-06}
{"train_loss": 0.05548633262515068, "global_step": 246759, "epoch": 2772, "lr": 2.0348651105820404e-06}
{"train_loss": 0.06366567313671112, "global_step": 246760, "epoch": 2772, "lr": 2.034701407917994e-06}
{"train_loss": 0.13035674393177032, "global_step": 246761, "epoch": 2772, "lr": 2.0345377117022948e-06}
{"train_loss": 0.04963785782456398, "global_step": 246762, "epoch": 2772, "lr": 2.034374021934976e-06}
{"train_loss": 0.027338465675711632, "global_step": 246763, "epoch": 2772, "lr": 2.034210338616038e-06}
{"train_loss": 0.03238585218787193, "global_step": 246764, "epoch": 2772, "lr": 2.0340466617455255e-06}
{"train_loss": 0.043188463896512985, "global_step": 246765, "epoch": 2772, "lr": 2.0338829913234437e-06}
{"train_loss": 0.050503022968769073, "global_step": 246766, "epoch": 2772, "lr": 2.033719327349831e-06}
{"train_loss": 0.04510148614645004, "global_step": 246767, "epoch": 2772, "lr": 2.0335556698246936e-06}
{"train_loss": 0.05964670330286026, "global_step": 246768, "epoch": 2772, "lr": 2.0333920187480647e-06}
{"train_loss": 0.0417979471385479, "global_step": 246769, "epoch": 2772, "lr": 2.033228374119955e-06}
{"train_loss": 0.050067417323589325, "global_step": 246770, "epoch": 2772, "lr": 2.0330647359403986e-06}
{"train_loss": 0.063088059425354, "global_step": 246771, "epoch": 2772, "lr": 2.032901104209406e-06}
{"train_loss": 0.02615199238061905, "global_step": 246772, "epoch": 2772, "lr": 2.0327374789270105e-06}
{"train_loss": 0.03093373030424118, "global_step": 246773, "epoch": 2772, "lr": 2.0325738600932285e-06}
{"train_loss": 0.027389049530029297, "global_step": 246774, "epoch": 2772, "lr": 2.032410247708083e-06}
{"train_loss": 0.004719734191894531, "global_step": 246775, "epoch": 2772, "lr": 2.03224664177159e-06}
{"train_loss": 0.06543432176113129, "global_step": 246776, "epoch": 2772, "lr": 2.0320830422837777e-06}
{"train_loss": 0.05947403237223625, "global_step": 246777, "epoch": 2772, "lr": 2.0319194492446736e-06}
{"train_loss": 0.023976311087608337, "global_step": 246778, "epoch": 2772, "lr": 2.0317558626542833e-06}
{"train_loss": 0.029778312891721725, "global_step": 246779, "epoch": 2772, "lr": 2.031592282512651e-06}
{"train_loss": 0.033886417746543884, "global_step": 246780, "epoch": 2772, "lr": 2.031428708819777e-06}
{"train_loss": 0.020228102803230286, "global_step": 246781, "epoch": 2772, "lr": 2.0312651415756944e-06}
{"train_loss": 0.02219323068857193, "global_step": 246782, "epoch": 2772, "lr": 2.0311015807804313e-06}
{"train_loss": 0.09755376726388931, "global_step": 246783, "epoch": 2772, "lr": 2.030938026433993e-06}
{"train_loss": 0.051861561834812164, "global_step": 246784, "epoch": 2772, "lr": 2.030774478536418e-06}
{"train_loss": 0.026349134743213654, "global_step": 246785, "epoch": 2772, "lr": 2.0306109370877127e-06}
{"train_loss": 0.08652189373970032, "global_step": 246786, "epoch": 2772, "lr": 2.0304474020879096e-06}
{"train_loss": 0.05827432498335838, "global_step": 246787, "epoch": 2772, "lr": 2.0302838735370254e-06}
{"train_loss": 0.030605195090174675, "global_step": 246788, "epoch": 2772, "lr": 2.0301203514350942e-06}
{"train_loss": 0.02588784508407116, "global_step": 246789, "epoch": 2772, "lr": 2.0299568357821264e-06}
{"train_loss": 0.08294860273599625, "global_step": 246790, "epoch": 2772, "lr": 2.0297933265781387e-06}
{"train_loss": 0.10120594501495361, "global_step": 246791, "epoch": 2772, "lr": 2.0296298238231702e-06}
{"train_loss": 0.02668491192162037, "global_step": 246792, "epoch": 2772, "lr": 2.029466327517221e-06}
{"train_loss": 0.028371667489409447, "global_step": 246793, "epoch": 2772, "lr": 2.0293028376603297e-06}
{"train_loss": 0.04672287404537201, "global_step": 246794, "epoch": 2772, "lr": 2.0291393542525183e-06}
{"train_loss": 0.04336370527744293, "global_step": 246795, "epoch": 2772, "lr": 2.0289758772937985e-06}
{"train_loss": 0.044110313371828434, "global_step": 246796, "epoch": 2772, "lr": 2.0288124067842084e-06, "val_loss": 9.582728385925293}
{"train_loss": 0.02889432944357395, "global_step": 246797, "epoch": 2773, "lr": 2.028648942723749e-06}
{"train_loss": 0.05399613454937935, "global_step": 246798, "epoch": 2773, "lr": 2.0284854851124634e-06}
{"train_loss": 0.02043529972434044, "global_step": 246799, "epoch": 2773, "lr": 2.0283220339503527e-06}
{"train_loss": 0.0517028346657753, "global_step": 246800, "epoch": 2773, "lr": 2.0281585892374555e-06}
{"train_loss": 0.05361532047390938, "global_step": 246801, "epoch": 2773, "lr": 2.027995150973783e-06}
{"train_loss": 0.07786563783884048, "global_step": 246802, "epoch": 2773, "lr": 2.027831719159368e-06}
{"train_loss": 0.0797998458147049, "global_step": 246803, "epoch": 2773, "lr": 2.0276682937942214e-06}
{"train_loss": 0.015354890376329422, "global_step": 246804, "epoch": 2773, "lr": 2.0275048748783776e-06}
{"train_loss": 0.019095083698630333, "global_step": 246805, "epoch": 2773, "lr": 2.0273414624118413e-06}
{"train_loss": 0.05885007977485657, "global_step": 246806, "epoch": 2773, "lr": 2.027178056394652e-06}
{"train_loss": 0.02147722989320755, "global_step": 246807, "epoch": 2773, "lr": 2.02701465682682e-06}
{"train_loss": 0.08006692677736282, "global_step": 246808, "epoch": 2773, "lr": 2.0268512637083736e-06}
{"train_loss": 0.03027399070560932, "global_step": 246809, "epoch": 2773, "lr": 2.026687877039324e-06}
{"train_loss": 0.04869402199983597, "global_step": 246810, "epoch": 2773, "lr": 2.026524496819715e-06}
{"train_loss": 0.016758255660533905, "global_step": 246811, "epoch": 2773, "lr": 2.0263611230495416e-06}
{"train_loss": 0.05214889347553253, "global_step": 246812, "epoch": 2773, "lr": 2.0261977557288424e-06}
{"train_loss": 0.03789789602160454, "global_step": 246813, "epoch": 2773, "lr": 2.0260343948576456e-06}
{"train_loss": 0.09042977541685104, "global_step": 246814, "epoch": 2773, "lr": 2.0258710404359504e-06}
{"train_loss": 0.07401478290557861, "global_step": 246815, "epoch": 2773, "lr": 2.0257076924638076e-06}
{"train_loss": 0.05923328921198845, "global_step": 246816, "epoch": 2773, "lr": 2.025544350941211e-06}
{"train_loss": 0.0368751659989357, "global_step": 246817, "epoch": 2773, "lr": 2.0253810158681994e-06}
{"train_loss": 0.040611088275909424, "global_step": 246818, "epoch": 2773, "lr": 2.0252176872447904e-06}
{"train_loss": 0.05762356147170067, "global_step": 246819, "epoch": 2773, "lr": 2.025054365071011e-06}
{"train_loss": 0.03146573156118393, "global_step": 246820, "epoch": 2773, "lr": 2.0248910493468663e-06}
{"train_loss": 0.062305524945259094, "global_step": 246821, "epoch": 2773, "lr": 2.024727740072402e-06}
{"train_loss": 0.028649205341935158, "global_step": 246822, "epoch": 2773, "lr": 2.024564437247628e-06}
{"train_loss": 0.02898728847503662, "global_step": 246823, "epoch": 2773, "lr": 2.0244011408725616e-06}
{"train_loss": 0.038990166038274765, "global_step": 246824, "epoch": 2773, "lr": 2.0242378509472305e-06}
{"train_loss": 0.0488271489739418, "global_step": 246825, "epoch": 2773, "lr": 2.0240745674716567e-06}
{"train_loss": 0.05785351246595383, "global_step": 246826, "epoch": 2773, "lr": 2.0239112904458624e-06}
{"train_loss": 0.04943061247467995, "global_step": 246827, "epoch": 2773, "lr": 2.02374801986987e-06}
{"train_loss": 0.05268198624253273, "global_step": 246828, "epoch": 2773, "lr": 2.023584755743696e-06}
{"train_loss": 0.050820108503103256, "global_step": 246829, "epoch": 2773, "lr": 2.0234214980673627e-06}
{"train_loss": 0.024138379842042923, "global_step": 246830, "epoch": 2773, "lr": 2.0232582468409035e-06}
{"train_loss": 0.035113416612148285, "global_step": 246831, "epoch": 2773, "lr": 2.0230950020643237e-06}
{"train_loss": 0.018233228474855423, "global_step": 246832, "epoch": 2773, "lr": 2.0229317637376622e-06}
{"train_loss": 0.06551438570022583, "global_step": 246833, "epoch": 2773, "lr": 2.02276853186093e-06}
{"train_loss": 0.07671660929918289, "global_step": 246834, "epoch": 2773, "lr": 2.0226053064341555e-06}
{"train_loss": 0.07234479486942291, "global_step": 246835, "epoch": 2773, "lr": 2.0224420874573493e-06}
{"train_loss": 0.06518884748220444, "global_step": 246836, "epoch": 2773, "lr": 2.0222788749305444e-06}
{"train_loss": 0.03115270659327507, "global_step": 246837, "epoch": 2773, "lr": 2.0221156688537577e-06}
{"train_loss": 0.06447233259677887, "global_step": 246838, "epoch": 2773, "lr": 2.0219524692270177e-06}
{"train_loss": 0.08899813890457153, "global_step": 246839, "epoch": 2773, "lr": 2.0217892760503344e-06}
{"train_loss": 0.062405966222286224, "global_step": 246840, "epoch": 2773, "lr": 2.021626089323747e-06}
{"train_loss": 0.03936052694916725, "global_step": 246841, "epoch": 2773, "lr": 2.021462909047256e-06}
{"train_loss": 0.009636553935706615, "global_step": 246842, "epoch": 2773, "lr": 2.021299735220905e-06}
{"train_loss": 0.03162836283445358, "global_step": 246843, "epoch": 2773, "lr": 2.021136567844695e-06}
{"train_loss": 0.043592266738414764, "global_step": 246844, "epoch": 2773, "lr": 2.020973406918669e-06}
{"train_loss": 0.03332360088825226, "global_step": 246845, "epoch": 2773, "lr": 2.0208102524428285e-06}
{"train_loss": 0.04005789756774902, "global_step": 246846, "epoch": 2773, "lr": 2.0206471044172115e-06}
{"train_loss": 0.012189183384180069, "global_step": 246847, "epoch": 2773, "lr": 2.0204839628418292e-06}
{"train_loss": 0.08389460295438766, "global_step": 246848, "epoch": 2773, "lr": 2.0203208277167095e-06}
{"train_loss": 0.0419834665954113, "global_step": 246849, "epoch": 2773, "lr": 2.020157699041875e-06}
{"train_loss": 0.03688683733344078, "global_step": 246850, "epoch": 2773, "lr": 2.019994576817341e-06}
{"train_loss": 0.02667086198925972, "global_step": 246851, "epoch": 2773, "lr": 2.019831461043142e-06}
{"train_loss": 0.062380507588386536, "global_step": 246852, "epoch": 2773, "lr": 2.019668351719284e-06}
{"train_loss": 0.007711789570748806, "global_step": 246853, "epoch": 2773, "lr": 2.019505248845804e-06}
{"train_loss": 0.03434537723660469, "global_step": 246854, "epoch": 2773, "lr": 2.0193421524227096e-06}
{"train_loss": 0.058072060346603394, "global_step": 246855, "epoch": 2773, "lr": 2.0191790624500385e-06}
{"train_loss": 0.011421955190598965, "global_step": 246856, "epoch": 2773, "lr": 2.019015978927796e-06}
{"train_loss": 0.09939354658126831, "global_step": 246857, "epoch": 2773, "lr": 2.0188529018560166e-06}
{"train_loss": 0.038839247077703476, "global_step": 246858, "epoch": 2773, "lr": 2.01868983123471e-06}
{"train_loss": 0.046454448252916336, "global_step": 246859, "epoch": 2773, "lr": 2.0185267670639163e-06}
{"train_loss": 0.043855562806129456, "global_step": 246860, "epoch": 2773, "lr": 2.018363709343646e-06}
{"train_loss": 0.04923015087842941, "global_step": 246861, "epoch": 2773, "lr": 2.0182006580739155e-06}
{"train_loss": 0.037927351891994476, "global_step": 246862, "epoch": 2773, "lr": 2.0180376132547587e-06}
{"train_loss": 0.05018356814980507, "global_step": 246863, "epoch": 2773, "lr": 2.017874574886186e-06}
{"train_loss": 0.05605732277035713, "global_step": 246864, "epoch": 2773, "lr": 2.0177115429682314e-06}
{"train_loss": 0.045377857983112335, "global_step": 246865, "epoch": 2773, "lr": 2.0175485175009055e-06}
{"train_loss": 0.04049926623702049, "global_step": 246866, "epoch": 2773, "lr": 2.0173854984842365e-06}
{"train_loss": 0.028947846964001656, "global_step": 246867, "epoch": 2773, "lr": 2.017222485918241e-06}
{"train_loss": 0.06544993072748184, "global_step": 246868, "epoch": 2773, "lr": 2.0170594798029574e-06}
{"train_loss": 0.016432369127869606, "global_step": 246869, "epoch": 2773, "lr": 2.0168964801383807e-06}
{"train_loss": 0.024596136063337326, "global_step": 246870, "epoch": 2773, "lr": 2.0167334869245604e-06}
{"train_loss": 0.035595767199993134, "global_step": 246871, "epoch": 2773, "lr": 2.0165705001614966e-06}
{"train_loss": 0.06591136008501053, "global_step": 246872, "epoch": 2773, "lr": 2.0164075198492227e-06}
{"train_loss": 0.038328517228364944, "global_step": 246873, "epoch": 2773, "lr": 2.0162445459877556e-06}
{"train_loss": 0.09220615029335022, "global_step": 246874, "epoch": 2773, "lr": 2.016081578577128e-06}
{"train_loss": 0.05756223201751709, "global_step": 246875, "epoch": 2773, "lr": 2.015918617617346e-06}
{"train_loss": 0.027482042089104652, "global_step": 246876, "epoch": 2773, "lr": 2.0157556631084427e-06}
{"train_loss": 0.046059027314186096, "global_step": 246877, "epoch": 2773, "lr": 2.0155927150504295e-06}
{"train_loss": 0.051541756838560104, "global_step": 246878, "epoch": 2773, "lr": 2.0154297734433448e-06}
{"train_loss": 0.03095797821879387, "global_step": 246879, "epoch": 2773, "lr": 2.015266838287189e-06}
{"train_loss": 0.048702824860811234, "global_step": 246880, "epoch": 2773, "lr": 2.0151039095820057e-06}
{"train_loss": 0.041832420974969864, "global_step": 246881, "epoch": 2773, "lr": 2.0149409873278014e-06}
{"train_loss": 0.0605083592236042, "global_step": 246882, "epoch": 2773, "lr": 2.014778071524609e-06}
{"train_loss": 0.05725632235407829, "global_step": 246883, "epoch": 2773, "lr": 2.0146151621724397e-06}
{"train_loss": 0.10041339695453644, "global_step": 246884, "epoch": 2773, "lr": 2.0144522592713153e-06}
{"train_loss": 0.04705278683202655, "global_step": 246885, "epoch": 2773, "lr": 2.0142893628212756e-06, "val_loss": 9.514998435974121}
{"train_loss": 0.07333581149578094, "global_step": 246886, "epoch": 2774, "lr": 2.0141264728223197e-06}
{"train_loss": 0.03702990710735321, "global_step": 246887, "epoch": 2774, "lr": 2.0139635892744866e-06}
{"train_loss": 0.08871980756521225, "global_step": 246888, "epoch": 2774, "lr": 2.013800712177788e-06}
{"train_loss": 0.043240804225206375, "global_step": 246889, "epoch": 2774, "lr": 2.013637841532251e-06}
{"train_loss": 0.0315609835088253, "global_step": 246890, "epoch": 2774, "lr": 2.0134749773378925e-06}
{"train_loss": 0.06168793514370918, "global_step": 246891, "epoch": 2774, "lr": 2.0133121195947402e-06}
{"train_loss": 0.02004118449985981, "global_step": 246892, "epoch": 2774, "lr": 2.0131492683028107e-06}
{"train_loss": 0.04179452732205391, "global_step": 246893, "epoch": 2774, "lr": 2.012986423462132e-06}
{"train_loss": 0.014949135482311249, "global_step": 246894, "epoch": 2774, "lr": 2.012823585072715e-06}
{"train_loss": 0.056300122290849686, "global_step": 246895, "epoch": 2774, "lr": 2.0126607531345987e-06}
{"train_loss": 0.07635559886693954, "global_step": 246896, "epoch": 2774, "lr": 2.0124979276477884e-06}
{"train_loss": 0.06887266039848328, "global_step": 246897, "epoch": 2774, "lr": 2.0123351086123177e-06}
{"train_loss": 0.07929594814777374, "global_step": 246898, "epoch": 2774, "lr": 2.012172296028203e-06}
{"train_loss": 0.03769310563802719, "global_step": 246899, "epoch": 2774, "lr": 2.012009489895461e-06}
{"train_loss": 0.03391949459910393, "global_step": 246900, "epoch": 2774, "lr": 2.0118466902141254e-06}
{"train_loss": 0.11722490191459656, "global_step": 246901, "epoch": 2774, "lr": 2.0116838969842067e-06}
{"train_loss": 0.048499733209609985, "global_step": 246902, "epoch": 2774, "lr": 2.0115211102057386e-06}
{"train_loss": 0.05490656942129135, "global_step": 246903, "epoch": 2774, "lr": 2.011358329878732e-06}
{"train_loss": 0.057351309806108475, "global_step": 246904, "epoch": 2774, "lr": 2.0111955560032203e-06}
{"train_loss": 0.05972716584801674, "global_step": 246905, "epoch": 2774, "lr": 2.011032788579209e-06}
{"train_loss": 0.04952605813741684, "global_step": 246906, "epoch": 2774, "lr": 2.0108700276067372e-06}
{"train_loss": 0.044809166342020035, "global_step": 246907, "epoch": 2774, "lr": 2.01070727308581e-06}
{"train_loss": 0.048351190984249115, "global_step": 246908, "epoch": 2774, "lr": 2.010544525016467e-06}
{"train_loss": 0.06360170245170593, "global_step": 246909, "epoch": 2774, "lr": 2.0103817833987182e-06}
{"train_loss": 0.03235442191362381, "global_step": 246910, "epoch": 2774, "lr": 2.0102190482325867e-06}
{"train_loss": 0.040289830416440964, "global_step": 246911, "epoch": 2774, "lr": 2.0100563195181e-06}
{"train_loss": 0.02783198095858097, "global_step": 246912, "epoch": 2774, "lr": 2.0098935972552745e-06}
{"train_loss": 0.049930717796087265, "global_step": 246913, "epoch": 2774, "lr": 2.009730881444133e-06}
{"train_loss": 0.04458874836564064, "global_step": 246914, "epoch": 2774, "lr": 2.0095681720847024e-06}
{"train_loss": 0.07289274036884308, "global_step": 246915, "epoch": 2774, "lr": 2.0094054691769948e-06}
{"train_loss": 0.06346104294061661, "global_step": 246916, "epoch": 2774, "lr": 2.0092427727210426e-06}
{"train_loss": 0.09047120064496994, "global_step": 246917, "epoch": 2774, "lr": 2.0090800827168575e-06}
{"train_loss": 0.055018629878759384, "global_step": 246918, "epoch": 2774, "lr": 2.008917399164467e-06}
{"train_loss": 0.06532436609268188, "global_step": 246919, "epoch": 2774, "lr": 2.0087547220638934e-06}
{"train_loss": 0.04709416627883911, "global_step": 246920, "epoch": 2774, "lr": 2.008592051415159e-06}
{"train_loss": 0.043003033846616745, "global_step": 246921, "epoch": 2774, "lr": 2.0084293872182856e-06}
{"train_loss": 0.0532393716275692, "global_step": 246922, "epoch": 2774, "lr": 2.0082667294732905e-06}
{"train_loss": 0.06260863691568375, "global_step": 246923, "epoch": 2774, "lr": 2.0081040781802063e-06}
{"train_loss": 0.016499629244208336, "global_step": 246924, "epoch": 2774, "lr": 2.007941433339039e-06}
{"train_loss": 0.0195888951420784, "global_step": 246925, "epoch": 2774, "lr": 2.007778794949827e-06}
{"train_loss": 0.02961960807442665, "global_step": 246926, "epoch": 2774, "lr": 2.0076161630125767e-06}
{"train_loss": 0.06199786812067032, "global_step": 246927, "epoch": 2774, "lr": 2.0074535375273205e-06}
{"train_loss": 0.05624646693468094, "global_step": 246928, "epoch": 2774, "lr": 2.0072909184940757e-06}
{"train_loss": 0.04458320140838623, "global_step": 246929, "epoch": 2774, "lr": 2.0071283059128697e-06}
{"train_loss": 0.06015467271208763, "global_step": 246930, "epoch": 2774, "lr": 2.0069656997837196e-06}
{"train_loss": 0.060532346367836, "global_step": 246931, "epoch": 2774, "lr": 2.0068031001066468e-06}
{"train_loss": 0.04375065863132477, "global_step": 246932, "epoch": 2774, "lr": 2.0066405068816742e-06}
{"train_loss": 0.06459451466798782, "global_step": 246933, "epoch": 2774, "lr": 2.0064779201088235e-06}
{"train_loss": 0.04225941747426987, "global_step": 246934, "epoch": 2774, "lr": 2.006315339788117e-06}
{"train_loss": 0.03610125556588173, "global_step": 246935, "epoch": 2774, "lr": 2.006152765919578e-06}
{"train_loss": 0.021673476323485374, "global_step": 246936, "epoch": 2774, "lr": 2.0059901985032216e-06}
{"train_loss": 0.03947034105658531, "global_step": 246937, "epoch": 2774, "lr": 2.0058276375390815e-06}
{"train_loss": 0.030120091512799263, "global_step": 246938, "epoch": 2774, "lr": 2.005665083027175e-06}
{"train_loss": 0.0534047894179821, "global_step": 246939, "epoch": 2774, "lr": 2.0055025349675126e-06}
{"train_loss": 0.04444736987352371, "global_step": 246940, "epoch": 2774, "lr": 2.0053399933601336e-06}
{"train_loss": 0.06830839067697525, "global_step": 246941, "epoch": 2774, "lr": 2.0051774582050485e-06}
{"train_loss": 0.057053517550230026, "global_step": 246942, "epoch": 2774, "lr": 2.00501492950228e-06}
{"train_loss": 0.054843805730342865, "global_step": 246943, "epoch": 2774, "lr": 2.004852407251856e-06}
{"train_loss": 0.0861922949552536, "global_step": 246944, "epoch": 2774, "lr": 2.0046898914537927e-06}
{"train_loss": 0.039471182972192764, "global_step": 246945, "epoch": 2774, "lr": 2.0045273821081124e-06}
{"train_loss": 0.01652011089026928, "global_step": 246946, "epoch": 2774, "lr": 2.0043648792148427e-06}
{"train_loss": 0.037301041185855865, "global_step": 246947, "epoch": 2774, "lr": 2.004202382773995e-06}
{"train_loss": 0.043233275413513184, "global_step": 246948, "epoch": 2774, "lr": 2.004039892785603e-06}
{"train_loss": 0.039103198796510696, "global_step": 246949, "epoch": 2774, "lr": 2.003877409249677e-06}
{"train_loss": 0.038305558264255524, "global_step": 246950, "epoch": 2774, "lr": 2.003714932166251e-06}
{"train_loss": 0.043074652552604675, "global_step": 246951, "epoch": 2774, "lr": 2.003552461535335e-06}
{"train_loss": 0.04048383980989456, "global_step": 246952, "epoch": 2774, "lr": 2.0033899973569636e-06}
{"train_loss": 0.05491073802113533, "global_step": 246953, "epoch": 2774, "lr": 2.003227539631142e-06}
{"train_loss": 0.10344845056533813, "global_step": 246954, "epoch": 2774, "lr": 2.003065088357908e-06}
{"train_loss": 0.042508840560913086, "global_step": 246955, "epoch": 2774, "lr": 2.0029026435372746e-06}
{"train_loss": 0.044412534683942795, "global_step": 246956, "epoch": 2774, "lr": 2.002740205169262e-06}
{"train_loss": 0.0730179101228714, "global_step": 246957, "epoch": 2774, "lr": 2.002577773253905e-06}
{"train_loss": 0.026808638125658035, "global_step": 246958, "epoch": 2774, "lr": 2.0024153477912088e-06}
{"train_loss": 0.026420777663588524, "global_step": 246959, "epoch": 2774, "lr": 2.0022529287812064e-06}
{"train_loss": 0.07260376214981079, "global_step": 246960, "epoch": 2774, "lr": 2.0020905162239146e-06}
{"train_loss": 0.036420151591300964, "global_step": 246961, "epoch": 2774, "lr": 2.001928110119361e-06}
{"train_loss": 0.04625188559293747, "global_step": 246962, "epoch": 2774, "lr": 2.001765710467557e-06}
{"train_loss": 0.03463646024465561, "global_step": 246963, "epoch": 2774, "lr": 2.0016033172685354e-06}
{"train_loss": 0.033346984535455704, "global_step": 246964, "epoch": 2774, "lr": 2.001440930522308e-06}
{"train_loss": 0.0579044483602047, "global_step": 246965, "epoch": 2774, "lr": 2.001278550228908e-06}
{"train_loss": 0.03610865771770477, "global_step": 246966, "epoch": 2774, "lr": 2.0011161763883456e-06}
{"train_loss": 0.03885314241051674, "global_step": 246967, "epoch": 2774, "lr": 2.0009538090006552e-06}
{"train_loss": 0.035131461918354034, "global_step": 246968, "epoch": 2774, "lr": 2.000791448065842e-06}
{"train_loss": 0.035212442278862, "global_step": 246969, "epoch": 2774, "lr": 2.0006290935839444e-06}
{"train_loss": 0.03133608400821686, "global_step": 246970, "epoch": 2774, "lr": 2.000466745554974e-06}
{"train_loss": 0.018383262678980827, "global_step": 246971, "epoch": 2774, "lr": 2.000304403978959e-06}
{"train_loss": 0.025640254840254784, "global_step": 246972, "epoch": 2774, "lr": 2.0001420688559148e-06}
{"train_loss": 0.05745285749435425, "global_step": 246973, "epoch": 2774, "lr": 1.99997974018587e-06}
{"train_loss": 0.048200629808510954, "global_step": 246974, "epoch": 2774, "lr": 1.9998174179688357e-06, "val_loss": 9.53986644744873}
{"train_loss": 0.0253779087215662, "global_step": 246975, "epoch": 2775, "lr": 1.9996551022048503e-06}
{"train_loss": 0.0194074809551239, "global_step": 246976, "epoch": 2775, "lr": 1.9994927928939257e-06}
{"train_loss": 0.03198140859603882, "global_step": 246977, "epoch": 2775, "lr": 1.9993304900360776e-06}
{"train_loss": 0.10635498911142349, "global_step": 246978, "epoch": 2775, "lr": 1.99916819363134e-06}
{"train_loss": 0.01978175900876522, "global_step": 246979, "epoch": 2775, "lr": 1.999005903679724e-06}
{"train_loss": 0.02210148796439171, "global_step": 246980, "epoch": 2775, "lr": 1.998843620181262e-06}
{"train_loss": 0.059545014053583145, "global_step": 246981, "epoch": 2775, "lr": 1.998681343135961e-06}
{"train_loss": 0.02135387249290943, "global_step": 246982, "epoch": 2775, "lr": 1.998519072543864e-06}
{"train_loss": 0.07017408311367035, "global_step": 246983, "epoch": 2775, "lr": 1.9983568084049718e-06}
{"train_loss": 0.031211553141474724, "global_step": 246984, "epoch": 2775, "lr": 1.9981945507193234e-06}
{"train_loss": 0.05998242273926735, "global_step": 246985, "epoch": 2775, "lr": 1.998032299486924e-06}
{"train_loss": 0.059845779091119766, "global_step": 246986, "epoch": 2775, "lr": 1.997870054707812e-06}
{"train_loss": 0.06906376779079437, "global_step": 246987, "epoch": 2775, "lr": 1.997707816381994e-06}
{"train_loss": 0.07584724575281143, "global_step": 246988, "epoch": 2775, "lr": 1.9975455845095082e-06}
{"train_loss": 0.026991602033376694, "global_step": 246989, "epoch": 2775, "lr": 1.9973833590903602e-06}
{"train_loss": 0.054513152688741684, "global_step": 246990, "epoch": 2775, "lr": 1.9972211401245835e-06}
{"train_loss": 0.024500105530023575, "global_step": 246991, "epoch": 2775, "lr": 1.997058927612189e-06}
{"train_loss": 0.032597292214632034, "global_step": 246992, "epoch": 2775, "lr": 1.9968967215532097e-06}
{"train_loss": 0.013851391151547432, "global_step": 246993, "epoch": 2775, "lr": 1.996734521947663e-06}
{"train_loss": 0.04260465130209923, "global_step": 246994, "epoch": 2775, "lr": 1.996572328795565e-06}
{"train_loss": 0.039955928921699524, "global_step": 246995, "epoch": 2775, "lr": 1.9964101420969495e-06}
{"train_loss": 0.0243258997797966, "global_step": 246996, "epoch": 2775, "lr": 1.996247961851827e-06}
{"train_loss": 0.039237163960933685, "global_step": 246997, "epoch": 2775, "lr": 1.9960857880602257e-06}
{"train_loss": 0.02715414948761463, "global_step": 246998, "epoch": 2775, "lr": 1.995923620722162e-06}
{"train_loss": 0.03159312158823013, "global_step": 246999, "epoch": 2775, "lr": 1.9957614598376695e-06}
{"train_loss": 0.05719047039747238, "global_step": 247000, "epoch": 2775, "lr": 1.9955993054067536e-06}
{"train_loss": 0.043825965374708176, "global_step": 247001, "epoch": 2775, "lr": 1.995437157429453e-06}
{"train_loss": 0.014704037457704544, "global_step": 247002, "epoch": 2775, "lr": 1.9952750159057678e-06}
{"train_loss": 0.04956560954451561, "global_step": 247003, "epoch": 2775, "lr": 1.995112880835748e-06}
{"train_loss": 0.038195859640836716, "global_step": 247004, "epoch": 2775, "lr": 1.994950752219388e-06}
{"train_loss": 0.0336441695690155, "global_step": 247005, "epoch": 2775, "lr": 1.9947886300567266e-06}
{"train_loss": 0.05556125566363335, "global_step": 247006, "epoch": 2775, "lr": 1.9946265143477805e-06}
{"train_loss": 0.042600907385349274, "global_step": 247007, "epoch": 2775, "lr": 1.994464405092572e-06}
{"train_loss": 0.021472332999110222, "global_step": 247008, "epoch": 2775, "lr": 1.9943023022911233e-06}
{"train_loss": 0.0338522344827652, "global_step": 247009, "epoch": 2775, "lr": 1.9941402059434568e-06}
{"train_loss": 0.037383440881967545, "global_step": 247010, "epoch": 2775, "lr": 1.9939781160495884e-06}
{"train_loss": 0.04116661101579666, "global_step": 247011, "epoch": 2775, "lr": 1.993816032609547e-06}
{"train_loss": 0.0436076782643795, "global_step": 247012, "epoch": 2775, "lr": 1.993653955623348e-06}
{"train_loss": 0.08233898878097534, "global_step": 247013, "epoch": 2775, "lr": 1.9934918850910257e-06}
{"train_loss": 0.04812046140432358, "global_step": 247014, "epoch": 2775, "lr": 1.993329821012585e-06}
{"train_loss": 0.04539650306105614, "global_step": 247015, "epoch": 2775, "lr": 1.993167763388065e-06}
{"train_loss": 0.04958629980683327, "global_step": 247016, "epoch": 2775, "lr": 1.993005712217477e-06}
{"train_loss": 0.021247297525405884, "global_step": 247017, "epoch": 2775, "lr": 1.9928436675008378e-06}
{"train_loss": 0.018445676192641258, "global_step": 247018, "epoch": 2775, "lr": 1.99268162923818e-06}
{"train_loss": 0.04288151115179062, "global_step": 247019, "epoch": 2775, "lr": 1.9925195974295154e-06}
{"train_loss": 0.05759722366929054, "global_step": 247020, "epoch": 2775, "lr": 1.9923575720748765e-06}
{"train_loss": 0.09121344238519669, "global_step": 247021, "epoch": 2775, "lr": 1.992195553174281e-06}
{"train_loss": 0.052103180438280106, "global_step": 247022, "epoch": 2775, "lr": 1.9920335407277502e-06}
{"train_loss": 0.023764194920659065, "global_step": 247023, "epoch": 2775, "lr": 1.9918715347352956e-06}
{"train_loss": 0.03888607770204544, "global_step": 247024, "epoch": 2775, "lr": 1.9917095351969617e-06}
{"train_loss": 0.04914239048957825, "global_step": 247025, "epoch": 2775, "lr": 1.9915475421127483e-06}
{"train_loss": 0.045289039611816406, "global_step": 247026, "epoch": 2775, "lr": 1.991385555482689e-06}
{"train_loss": 0.030059488490223885, "global_step": 247027, "epoch": 2775, "lr": 1.9912235753067997e-06}
{"train_loss": 0.042595990002155304, "global_step": 247028, "epoch": 2775, "lr": 1.9910616015851145e-06}
{"train_loss": 0.06446624547243118, "global_step": 247029, "epoch": 2775, "lr": 1.990899634317639e-06}
{"train_loss": 0.027220554649829865, "global_step": 247030, "epoch": 2775, "lr": 1.9907376735044005e-06}
{"train_loss": 0.0076682656072080135, "global_step": 247031, "epoch": 2775, "lr": 1.990575719145427e-06}
{"train_loss": 0.04221801459789276, "global_step": 247032, "epoch": 2775, "lr": 1.990413771240729e-06}
{"train_loss": 0.0452028289437294, "global_step": 247033, "epoch": 2775, "lr": 1.9902518297903405e-06}
{"train_loss": 0.03361694514751434, "global_step": 247034, "epoch": 2775, "lr": 1.9900898947942727e-06}
{"train_loss": 0.0766032338142395, "global_step": 247035, "epoch": 2775, "lr": 1.9899279662525583e-06}
{"train_loss": 0.03763272240757942, "global_step": 247036, "epoch": 2775, "lr": 1.9897660441652088e-06}
{"train_loss": 0.03013785183429718, "global_step": 247037, "epoch": 2775, "lr": 1.989604128532252e-06}
{"train_loss": 0.057028740644454956, "global_step": 247038, "epoch": 2775, "lr": 1.9894422193537043e-06}
{"train_loss": 0.03835871070623398, "global_step": 247039, "epoch": 2775, "lr": 1.9892803166295936e-06}
{"train_loss": 0.049708880484104156, "global_step": 247040, "epoch": 2775, "lr": 1.9891184203599365e-06}
{"train_loss": 0.03460155799984932, "global_step": 247041, "epoch": 2775, "lr": 1.9889565305447667e-06}
{"train_loss": 0.05328375846147537, "global_step": 247042, "epoch": 2775, "lr": 1.9887946471840836e-06}
{"train_loss": 0.04965159669518471, "global_step": 247043, "epoch": 2775, "lr": 1.988632770277932e-06}
{"train_loss": 0.07644715160131454, "global_step": 247044, "epoch": 2775, "lr": 1.9884708998263167e-06}
{"train_loss": 0.01940285600721836, "global_step": 247045, "epoch": 2775, "lr": 1.9883090358292723e-06}
{"train_loss": 0.047122880816459656, "global_step": 247046, "epoch": 2775, "lr": 1.9881471782868086e-06}
{"train_loss": 0.070169597864151, "global_step": 247047, "epoch": 2775, "lr": 1.987985327198949e-06}
{"train_loss": 0.033325739204883575, "global_step": 247048, "epoch": 2775, "lr": 1.9878234825657317e-06}
{"train_loss": 0.026262760162353516, "global_step": 247049, "epoch": 2775, "lr": 1.987661644387162e-06}
{"train_loss": 0.025844711810350418, "global_step": 247050, "epoch": 2775, "lr": 1.9874998126632626e-06}
{"train_loss": 0.024595942348241806, "global_step": 247051, "epoch": 2775, "lr": 1.9873379873940667e-06}
{"train_loss": 0.04847267270088196, "global_step": 247052, "epoch": 2775, "lr": 1.98717616857958e-06}
{"train_loss": 0.07007536292076111, "global_step": 247053, "epoch": 2775, "lr": 1.9870143562198353e-06}
{"train_loss": 0.0578962117433548, "global_step": 247054, "epoch": 2775, "lr": 1.986852550314855e-06}
{"train_loss": 0.04171733930706978, "global_step": 247055, "epoch": 2775, "lr": 1.9866907508646505e-06}
{"train_loss": 0.02652927301824093, "global_step": 247056, "epoch": 2775, "lr": 1.9865289578692547e-06}
{"train_loss": 0.03200478479266167, "global_step": 247057, "epoch": 2775, "lr": 1.986367171328679e-06}
{"train_loss": 0.028567209839820862, "global_step": 247058, "epoch": 2775, "lr": 1.986205391242957e-06}
{"train_loss": 0.027211511507630348, "global_step": 247059, "epoch": 2775, "lr": 1.9860436176121044e-06}
{"train_loss": 0.054582543671131134, "global_step": 247060, "epoch": 2775, "lr": 1.985881850436139e-06}
{"train_loss": 0.04321128875017166, "global_step": 247061, "epoch": 2775, "lr": 1.9857200897150875e-06}
{"train_loss": 0.021726883947849274, "global_step": 247062, "epoch": 2775, "lr": 1.985558335448978e-06}
{"train_loss": 0.04178087444180686, "global_step": 247063, "epoch": 2775, "lr": 1.9853965876378112e-06, "val_loss": 9.414606094360352, "train_action_mse_error": 10.596994400024414}
{"train_loss": 0.028394747525453568, "global_step": 247064, "epoch": 2776, "lr": 1.985234846281636e-06}
{"train_loss": 0.03077639825642109, "global_step": 247065, "epoch": 2776, "lr": 1.985073111380453e-06}
{"train_loss": 0.04923458397388458, "global_step": 247066, "epoch": 2776, "lr": 1.9849113829342903e-06}
{"train_loss": 0.023212986066937447, "global_step": 247067, "epoch": 2776, "lr": 1.984749660943175e-06}
{"train_loss": 0.027732908725738525, "global_step": 247068, "epoch": 2776, "lr": 1.9845879454071237e-06}
{"train_loss": 0.03798910602927208, "global_step": 247069, "epoch": 2776, "lr": 1.984426236326159e-06}
{"train_loss": 0.08634991943836212, "global_step": 247070, "epoch": 2776, "lr": 1.984264533700303e-06}
{"train_loss": 0.06712281703948975, "global_step": 247071, "epoch": 2776, "lr": 1.9841028375295834e-06}
{"train_loss": 0.04776211455464363, "global_step": 247072, "epoch": 2776, "lr": 1.983941147814006e-06}
{"train_loss": 0.0625479444861412, "global_step": 247073, "epoch": 2776, "lr": 1.983779464553609e-06}
{"train_loss": 0.06948018819093704, "global_step": 247074, "epoch": 2776, "lr": 1.983617787748404e-06}
{"train_loss": 0.053667303174734116, "global_step": 247075, "epoch": 2776, "lr": 1.983456117398419e-06}
{"train_loss": 0.06333010643720627, "global_step": 247076, "epoch": 2776, "lr": 1.98329445350367e-06}
{"train_loss": 0.08456216007471085, "global_step": 247077, "epoch": 2776, "lr": 1.983132796064191e-06}
{"train_loss": 0.036072514951229095, "global_step": 247078, "epoch": 2776, "lr": 1.9829711450799817e-06}
{"train_loss": 0.029403524473309517, "global_step": 247079, "epoch": 2776, "lr": 1.982809500551086e-06}
{"train_loss": 0.053576838225126266, "global_step": 247080, "epoch": 2776, "lr": 1.9826478624775102e-06}
{"train_loss": 0.055300381034612656, "global_step": 247081, "epoch": 2776, "lr": 1.9824862308592873e-06}
{"train_loss": 0.04337060824036598, "global_step": 247082, "epoch": 2776, "lr": 1.9823246056964282e-06}
{"train_loss": 0.08516993373632431, "global_step": 247083, "epoch": 2776, "lr": 1.9821629869889614e-06}
{"train_loss": 0.11961679905653, "global_step": 247084, "epoch": 2776, "lr": 1.9820013747369135e-06}
{"train_loss": 0.0371427983045578, "global_step": 247085, "epoch": 2776, "lr": 1.9818397689402914e-06}
{"train_loss": 0.049061473459005356, "global_step": 247086, "epoch": 2776, "lr": 1.981678169599138e-06}
{"train_loss": 0.059702299535274506, "global_step": 247087, "epoch": 2776, "lr": 1.9815165767134546e-06}
{"train_loss": 0.01358511671423912, "global_step": 247088, "epoch": 2776, "lr": 1.9813549902832683e-06}
{"train_loss": 0.036562480032444, "global_step": 247089, "epoch": 2776, "lr": 1.981193410308613e-06}
{"train_loss": 0.04308035224676132, "global_step": 247090, "epoch": 2776, "lr": 1.981031836789493e-06}
{"train_loss": 0.04929797723889351, "global_step": 247091, "epoch": 2776, "lr": 1.980870269725943e-06}
{"train_loss": 0.02869855985045433, "global_step": 247092, "epoch": 2776, "lr": 1.9807087091179786e-06}
{"train_loss": 0.03214952349662781, "global_step": 247093, "epoch": 2776, "lr": 1.9805471549656172e-06}
{"train_loss": 0.035763341933488846, "global_step": 247094, "epoch": 2776, "lr": 1.980385607268892e-06}
{"train_loss": 0.016240354627370834, "global_step": 247095, "epoch": 2776, "lr": 1.9802240660278136e-06}
{"train_loss": 0.04548360034823418, "global_step": 247096, "epoch": 2776, "lr": 1.9800625312424158e-06}
{"train_loss": 0.03563413769006729, "global_step": 247097, "epoch": 2776, "lr": 1.9799010029127097e-06}
{"train_loss": 0.06781721115112305, "global_step": 247098, "epoch": 2776, "lr": 1.9797394810387228e-06}
{"train_loss": 0.05127999931573868, "global_step": 247099, "epoch": 2776, "lr": 1.979577965620466e-06}
{"train_loss": 0.037320561707019806, "global_step": 247100, "epoch": 2776, "lr": 1.9794164566579786e-06}
{"train_loss": 0.06069400906562805, "global_step": 247101, "epoch": 2776, "lr": 1.9792549541512662e-06}
{"train_loss": 0.07202116400003433, "global_step": 247102, "epoch": 2776, "lr": 1.9790934581003616e-06}
{"train_loss": 0.04905484616756439, "global_step": 247103, "epoch": 2776, "lr": 1.978931968505282e-06}
{"train_loss": 0.0751112550497055, "global_step": 247104, "epoch": 2776, "lr": 1.9787704853660495e-06}
{"train_loss": 0.02264370396733284, "global_step": 247105, "epoch": 2776, "lr": 1.9786090086826913e-06}
{"train_loss": 0.05546776205301285, "global_step": 247106, "epoch": 2776, "lr": 1.9784475384552136e-06}
{"train_loss": 0.0245803389698267, "global_step": 247107, "epoch": 2776, "lr": 1.9782860746836605e-06}
{"train_loss": 0.02651873603463173, "global_step": 247108, "epoch": 2776, "lr": 1.9781246173680324e-06}
{"train_loss": 0.0439164862036705, "global_step": 247109, "epoch": 2776, "lr": 1.977963166508362e-06}
{"train_loss": 0.029332732781767845, "global_step": 247110, "epoch": 2776, "lr": 1.9778017221046717e-06}
{"train_loss": 0.020575810223817825, "global_step": 247111, "epoch": 2776, "lr": 1.9776402841569784e-06}
{"train_loss": 0.047183386981487274, "global_step": 247112, "epoch": 2776, "lr": 1.9774788526653042e-06}
{"train_loss": 0.04514103755354881, "global_step": 247113, "epoch": 2776, "lr": 1.977317427629677e-06}
{"train_loss": 0.027840709313750267, "global_step": 247114, "epoch": 2776, "lr": 1.977156009050113e-06}
{"train_loss": 0.06963150203227997, "global_step": 247115, "epoch": 2776, "lr": 1.9769945969266347e-06}
{"train_loss": 0.04352710768580437, "global_step": 247116, "epoch": 2776, "lr": 1.976833191259264e-06}
{"train_loss": 0.058963432908058167, "global_step": 247117, "epoch": 2776, "lr": 1.9766717920480236e-06}
{"train_loss": 0.03444252535700798, "global_step": 247118, "epoch": 2776, "lr": 1.97651039929293e-06}
{"train_loss": 0.04675884544849396, "global_step": 247119, "epoch": 2776, "lr": 1.9763490129940164e-06}
{"train_loss": 0.03414378687739372, "global_step": 247120, "epoch": 2776, "lr": 1.9761876331512887e-06}
{"train_loss": 0.032461486756801605, "global_step": 247121, "epoch": 2776, "lr": 1.9760262597647796e-06}
{"train_loss": 0.04470209777355194, "global_step": 247122, "epoch": 2776, "lr": 1.975864892834517e-06}
{"train_loss": 0.047024693340063095, "global_step": 247123, "epoch": 2776, "lr": 1.975703532360501e-06}
{"train_loss": 0.019806593656539917, "global_step": 247124, "epoch": 2776, "lr": 1.9755421783427763e-06}
{"train_loss": 0.03559700772166252, "global_step": 247125, "epoch": 2776, "lr": 1.9753808307813594e-06}
{"train_loss": 0.020215075463056564, "global_step": 247126, "epoch": 2776, "lr": 1.9752194896762555e-06}
{"train_loss": 0.03395881503820419, "global_step": 247127, "epoch": 2776, "lr": 1.975058155027504e-06}
{"train_loss": 0.05848163738846779, "global_step": 247128, "epoch": 2776, "lr": 1.9748968268351153e-06}
{"train_loss": 0.08766201138496399, "global_step": 247129, "epoch": 2776, "lr": 1.9747355050991235e-06}
{"train_loss": 0.048809465020895004, "global_step": 247130, "epoch": 2776, "lr": 1.9745741898195336e-06}
{"train_loss": 0.05520763248205185, "global_step": 247131, "epoch": 2776, "lr": 1.97441288099639e-06}
{"train_loss": 0.08638601750135422, "global_step": 247132, "epoch": 2776, "lr": 1.9742515786296934e-06}
{"train_loss": 0.05777016654610634, "global_step": 247133, "epoch": 2776, "lr": 1.9740902827194706e-06}
{"train_loss": 0.06586083024740219, "global_step": 247134, "epoch": 2776, "lr": 1.9739289932657556e-06}
{"train_loss": 0.04692631959915161, "global_step": 247135, "epoch": 2776, "lr": 1.973767710268548e-06}
{"train_loss": 0.05839709937572479, "global_step": 247136, "epoch": 2776, "lr": 1.9736064337278927e-06}
{"train_loss": 0.05271126329898834, "global_step": 247137, "epoch": 2776, "lr": 1.9734451636437947e-06}
{"train_loss": 0.04503810405731201, "global_step": 247138, "epoch": 2776, "lr": 1.973283900016276e-06}
{"train_loss": 0.016419805586338043, "global_step": 247139, "epoch": 2776, "lr": 1.9731226428453766e-06}
{"train_loss": 0.02855287492275238, "global_step": 247140, "epoch": 2776, "lr": 1.9729613921310954e-06}
{"train_loss": 0.0587804913520813, "global_step": 247141, "epoch": 2776, "lr": 1.9728001478734716e-06}
{"train_loss": 0.05508997291326523, "global_step": 247142, "epoch": 2776, "lr": 1.9726389100725106e-06}
{"train_loss": 0.06780898571014404, "global_step": 247143, "epoch": 2776, "lr": 1.9724776787282517e-06}
{"train_loss": 0.05939612165093422, "global_step": 247144, "epoch": 2776, "lr": 1.9723164538407056e-06}
{"train_loss": 0.07213608920574188, "global_step": 247145, "epoch": 2776, "lr": 1.9721552354098948e-06}
{"train_loss": 0.05723465979099274, "global_step": 247146, "epoch": 2776, "lr": 1.9719940234358416e-06}
{"train_loss": 0.08288587629795074, "global_step": 247147, "epoch": 2776, "lr": 1.971832817918573e-06}
{"train_loss": 0.05251411348581314, "global_step": 247148, "epoch": 2776, "lr": 1.971671618858101e-06}
{"train_loss": 0.04809235408902168, "global_step": 247149, "epoch": 2776, "lr": 1.971510426254458e-06}
{"train_loss": 0.06502949446439743, "global_step": 247150, "epoch": 2776, "lr": 1.9713492401076507e-06}
{"train_loss": 0.06723232567310333, "global_step": 247151, "epoch": 2776, "lr": 1.971188060417717e-06}
{"train_loss": 0.049004437925105684, "global_step": 247152, "epoch": 2776, "lr": 1.971026887184668e-06, "val_loss": 9.486555099487305}
{"train_loss": 0.07221564650535583, "global_step": 247153, "epoch": 2777, "lr": 1.9708657204085378e-06}
{"train_loss": 0.038852933794260025, "global_step": 247154, "epoch": 2777, "lr": 1.970704560089326e-06}
{"train_loss": 0.0448402501642704, "global_step": 247155, "epoch": 2777, "lr": 1.9705434062270765e-06}
{"train_loss": 0.05939633026719093, "global_step": 247156, "epoch": 2777, "lr": 1.970382258821801e-06}
{"train_loss": 0.05478375032544136, "global_step": 247157, "epoch": 2777, "lr": 1.9702211178735163e-06}
{"train_loss": 0.08569834381341934, "global_step": 247158, "epoch": 2777, "lr": 1.970059983382255e-06}
{"train_loss": 0.055557966232299805, "global_step": 247159, "epoch": 2777, "lr": 1.9698988553480346e-06}
{"train_loss": 0.05788816511631012, "global_step": 247160, "epoch": 2777, "lr": 1.9697377337708766e-06}
{"train_loss": 0.09070604294538498, "global_step": 247161, "epoch": 2777, "lr": 1.969576618650798e-06}
{"train_loss": 0.01583205722272396, "global_step": 247162, "epoch": 2777, "lr": 1.9694155099878263e-06}
{"train_loss": 0.024889305233955383, "global_step": 247163, "epoch": 2777, "lr": 1.9692544077819785e-06}
{"train_loss": 0.07888144999742508, "global_step": 247164, "epoch": 2777, "lr": 1.9690933120332878e-06}
{"train_loss": 0.016317546367645264, "global_step": 247165, "epoch": 2777, "lr": 1.968932222741765e-06}
{"train_loss": 0.05636560171842575, "global_step": 247166, "epoch": 2777, "lr": 1.968771139907427e-06}
{"train_loss": 0.04047838971018791, "global_step": 247167, "epoch": 2777, "lr": 1.9686100635303073e-06}
{"train_loss": 0.0396684929728508, "global_step": 247168, "epoch": 2777, "lr": 1.9684489936104167e-06}
{"train_loss": 0.08508278429508209, "global_step": 247169, "epoch": 2777, "lr": 1.968287930147794e-06}
{"train_loss": 0.044716279953718185, "global_step": 247170, "epoch": 2777, "lr": 1.9681268731424453e-06}
{"train_loss": 0.0550251305103302, "global_step": 247171, "epoch": 2777, "lr": 1.9679658225943863e-06}
{"train_loss": 0.05480538308620453, "global_step": 247172, "epoch": 2777, "lr": 1.9678047785036625e-06}
{"train_loss": 0.04336542636156082, "global_step": 247173, "epoch": 2777, "lr": 1.967643740870273e-06}
{"train_loss": 0.04641690477728844, "global_step": 247174, "epoch": 2777, "lr": 1.967482709694246e-06}
{"train_loss": 0.043486446142196655, "global_step": 247175, "epoch": 2777, "lr": 1.9673216849756148e-06}
{"train_loss": 0.022063786163926125, "global_step": 247176, "epoch": 2777, "lr": 1.967160666714385e-06}
{"train_loss": 0.05952756851911545, "global_step": 247177, "epoch": 2777, "lr": 1.9669996549105894e-06}
{"train_loss": 0.012674874626100063, "global_step": 247178, "epoch": 2777, "lr": 1.96683864956424e-06}
{"train_loss": 0.029827967286109924, "global_step": 247179, "epoch": 2777, "lr": 1.966677650675369e-06}
{"train_loss": 0.04184413328766823, "global_step": 247180, "epoch": 2777, "lr": 1.9665166582439885e-06}
{"train_loss": 0.014088010415434837, "global_step": 247181, "epoch": 2777, "lr": 1.9663556722701316e-06}
{"train_loss": 0.03645024821162224, "global_step": 247182, "epoch": 2777, "lr": 1.966194692753803e-06}
{"train_loss": 0.0656857118010521, "global_step": 247183, "epoch": 2777, "lr": 1.966033719695043e-06}
{"train_loss": 0.05323590338230133, "global_step": 247184, "epoch": 2777, "lr": 1.9658727530938557e-06}
{"train_loss": 0.04574683681130409, "global_step": 247185, "epoch": 2777, "lr": 1.9657117929502813e-06}
{"train_loss": 0.05785885453224182, "global_step": 247186, "epoch": 2777, "lr": 1.9655508392643187e-06}
{"train_loss": 0.05024874210357666, "global_step": 247187, "epoch": 2777, "lr": 1.9653898920360125e-06}
{"train_loss": 0.04357365146279335, "global_step": 247188, "epoch": 2777, "lr": 1.9652289512653686e-06}
{"train_loss": 0.021512169390916824, "global_step": 247189, "epoch": 2777, "lr": 1.9650680169524204e-06}
{"train_loss": 0.0771883875131607, "global_step": 247190, "epoch": 2777, "lr": 1.9649070890971786e-06}
{"train_loss": 0.08822184801101685, "global_step": 247191, "epoch": 2777, "lr": 1.9647461676996715e-06}
{"train_loss": 0.03501826152205467, "global_step": 247192, "epoch": 2777, "lr": 1.9645852527599153e-06}
{"train_loss": 0.04317634552717209, "global_step": 247193, "epoch": 2777, "lr": 1.9644243442779373e-06}
{"train_loss": 0.057408303022384644, "global_step": 247194, "epoch": 2777, "lr": 1.964263442253761e-06}
{"train_loss": 0.022217657417058945, "global_step": 247195, "epoch": 2777, "lr": 1.9641025466873966e-06}
{"train_loss": 0.030130604282021523, "global_step": 247196, "epoch": 2777, "lr": 1.963941657578883e-06}
{"train_loss": 0.053351011127233505, "global_step": 247197, "epoch": 2777, "lr": 1.96378077492822e-06}
{"train_loss": 0.03952447324991226, "global_step": 247198, "epoch": 2777, "lr": 1.9636198987354526e-06}
{"train_loss": 0.048663634806871414, "global_step": 247199, "epoch": 2777, "lr": 1.9634590290005805e-06}
{"train_loss": 0.04633062332868576, "global_step": 247200, "epoch": 2777, "lr": 1.9632981657236426e-06}
{"train_loss": 0.03579351678490639, "global_step": 247201, "epoch": 2777, "lr": 1.96313730890465e-06}
{"train_loss": 0.062967948615551, "global_step": 247202, "epoch": 2777, "lr": 1.962976458543636e-06}
{"train_loss": 0.03886917233467102, "global_step": 247203, "epoch": 2777, "lr": 1.9628156146406116e-06}
{"train_loss": 0.05303305387496948, "global_step": 247204, "epoch": 2777, "lr": 1.962654777195594e-06}
{"train_loss": 0.010990021750330925, "global_step": 247205, "epoch": 2777, "lr": 1.962493946208621e-06}
{"train_loss": 0.07531298696994781, "global_step": 247206, "epoch": 2777, "lr": 1.9623331216796936e-06}
{"train_loss": 0.0492626391351223, "global_step": 247207, "epoch": 2777, "lr": 1.962172303608856e-06}
{"train_loss": 0.06032109260559082, "global_step": 247208, "epoch": 2777, "lr": 1.9620114919961187e-06}
{"train_loss": 0.03127000853419304, "global_step": 247209, "epoch": 2777, "lr": 1.9618506868414987e-06}
{"train_loss": 0.05544246733188629, "global_step": 247210, "epoch": 2777, "lr": 1.9616898881450183e-06}
{"train_loss": 0.10481642186641693, "global_step": 247211, "epoch": 2777, "lr": 1.961529095906711e-06}
{"train_loss": 0.04605637863278389, "global_step": 247212, "epoch": 2777, "lr": 1.9613683101265823e-06}
{"train_loss": 0.0190668236464262, "global_step": 247213, "epoch": 2777, "lr": 1.9612075308046707e-06}
{"train_loss": 0.023659855127334595, "global_step": 247214, "epoch": 2777, "lr": 1.9610467579409876e-06}
{"train_loss": 0.05114709958434105, "global_step": 247215, "epoch": 2777, "lr": 1.960885991535555e-06}
{"train_loss": 0.05814806744456291, "global_step": 247216, "epoch": 2777, "lr": 1.960725231588395e-06}
{"train_loss": 0.039177604019641876, "global_step": 247217, "epoch": 2777, "lr": 1.96056447809953e-06}
{"train_loss": 0.043390460312366486, "global_step": 247218, "epoch": 2777, "lr": 1.9604037310689828e-06}
{"train_loss": 0.09833434224128723, "global_step": 247219, "epoch": 2777, "lr": 1.9602429904967744e-06}
{"train_loss": 0.022601502016186714, "global_step": 247220, "epoch": 2777, "lr": 1.9600822563829225e-06}
{"train_loss": 0.06041737645864487, "global_step": 247221, "epoch": 2777, "lr": 1.959921528727454e-06}
{"train_loss": 0.01335384976118803, "global_step": 247222, "epoch": 2777, "lr": 1.9597608075303863e-06}
{"train_loss": 0.06914830952882767, "global_step": 247223, "epoch": 2777, "lr": 1.959600092791747e-06}
{"train_loss": 0.042905475944280624, "global_step": 247224, "epoch": 2777, "lr": 1.9594393845115524e-06}
{"train_loss": 0.03319961205124855, "global_step": 247225, "epoch": 2777, "lr": 1.9592786826898245e-06}
{"train_loss": 0.04644468426704407, "global_step": 247226, "epoch": 2777, "lr": 1.9591179873265864e-06}
{"train_loss": 0.08564768731594086, "global_step": 247227, "epoch": 2777, "lr": 1.9589572984218653e-06}
{"train_loss": 0.04205593839287758, "global_step": 247228, "epoch": 2777, "lr": 1.9587966159756666e-06}
{"train_loss": 0.01445840485394001, "global_step": 247229, "epoch": 2777, "lr": 1.958635939988024e-06}
{"train_loss": 0.05134027823805809, "global_step": 247230, "epoch": 2777, "lr": 1.9584752704589646e-06}
{"train_loss": 0.08020692318677902, "global_step": 247231, "epoch": 2777, "lr": 1.958314607388495e-06}
{"train_loss": 0.04507974907755852, "global_step": 247232, "epoch": 2777, "lr": 1.9581539507766477e-06}
{"train_loss": 0.029312409460544586, "global_step": 247233, "epoch": 2777, "lr": 1.9579933006234395e-06}
{"train_loss": 0.02384323999285698, "global_step": 247234, "epoch": 2777, "lr": 1.9578326569288987e-06}
{"train_loss": 0.042768120765686035, "global_step": 247235, "epoch": 2777, "lr": 1.9576720196930353e-06}
{"train_loss": 0.05375991761684418, "global_step": 247236, "epoch": 2777, "lr": 1.9575113889158837e-06}
{"train_loss": 0.05087941512465477, "global_step": 247237, "epoch": 2777, "lr": 1.957350764597454e-06}
{"train_loss": 0.03829406946897507, "global_step": 247238, "epoch": 2777, "lr": 1.957190146737775e-06}
{"train_loss": 0.03812607377767563, "global_step": 247239, "epoch": 2777, "lr": 1.957029535336863e-06}
{"train_loss": 0.09039292484521866, "global_step": 247240, "epoch": 2777, "lr": 1.956868930394751e-06}
{"train_loss": 0.04803508580819275, "global_step": 247241, "epoch": 2777, "lr": 1.95670833191145e-06, "val_loss": 9.533282279968262}
{"train_loss": 0.08092381060123444, "global_step": 247242, "epoch": 2778, "lr": 1.9565477398869768e-06}
{"train_loss": 0.02588394097983837, "global_step": 247243, "epoch": 2778, "lr": 1.956387154321371e-06}
{"train_loss": 0.021411992609500885, "global_step": 247244, "epoch": 2778, "lr": 1.9562265752146313e-06}
{"train_loss": 0.04880892112851143, "global_step": 247245, "epoch": 2778, "lr": 1.956066002566803e-06}
{"train_loss": 0.03641348332166672, "global_step": 247246, "epoch": 2778, "lr": 1.955905436377886e-06}
{"train_loss": 0.06156323105096817, "global_step": 247247, "epoch": 2778, "lr": 1.955744876647919e-06}
{"train_loss": 0.03902898728847504, "global_step": 247248, "epoch": 2778, "lr": 1.9555843233769076e-06}
{"train_loss": 0.04861733317375183, "global_step": 247249, "epoch": 2778, "lr": 1.9554237765648907e-06}
{"train_loss": 0.03139811009168625, "global_step": 247250, "epoch": 2778, "lr": 1.9552632362118796e-06}
{"train_loss": 0.03472781181335449, "global_step": 247251, "epoch": 2778, "lr": 1.955102702317901e-06}
{"train_loss": 0.011442817747592926, "global_step": 247252, "epoch": 2778, "lr": 1.954942174882962e-06}
{"train_loss": 0.04471202939748764, "global_step": 247253, "epoch": 2778, "lr": 1.9547816539071064e-06}
{"train_loss": 0.025152970105409622, "global_step": 247254, "epoch": 2778, "lr": 1.954621139390339e-06}
{"train_loss": 0.07003528624773026, "global_step": 247255, "epoch": 2778, "lr": 1.954460631332694e-06}
{"train_loss": 0.023783694952726364, "global_step": 247256, "epoch": 2778, "lr": 1.954300129734177e-06}
{"train_loss": 0.020111916586756706, "global_step": 247257, "epoch": 2778, "lr": 1.954139634594826e-06}
{"train_loss": 0.019467435777187347, "global_step": 247258, "epoch": 2778, "lr": 1.9539791459146528e-06}
{"train_loss": 0.07096147537231445, "global_step": 247259, "epoch": 2778, "lr": 1.953818663693685e-06}
{"train_loss": 0.029220305383205414, "global_step": 247260, "epoch": 2778, "lr": 1.9536581879319337e-06}
{"train_loss": 0.0331934317946434, "global_step": 247261, "epoch": 2778, "lr": 1.9534977186294323e-06}
{"train_loss": 0.04163359850645065, "global_step": 247262, "epoch": 2778, "lr": 1.9533372557861916e-06}
{"train_loss": 0.030107660219073296, "global_step": 247263, "epoch": 2778, "lr": 1.953176799402251e-06}
{"train_loss": 0.04259692132472992, "global_step": 247264, "epoch": 2778, "lr": 1.95301634947761e-06}
{"train_loss": 0.03261411935091019, "global_step": 247265, "epoch": 2778, "lr": 1.9528559060122963e-06}
{"train_loss": 0.006828123237937689, "global_step": 247266, "epoch": 2778, "lr": 1.9526954690063493e-06}
{"train_loss": 0.06472154706716537, "global_step": 247267, "epoch": 2778, "lr": 1.952535038459763e-06}
{"train_loss": 0.04592546820640564, "global_step": 247268, "epoch": 2778, "lr": 1.952374614372582e-06}
{"train_loss": 0.04489864781498909, "global_step": 247269, "epoch": 2778, "lr": 1.952214196744817e-06}
{"train_loss": 0.034876689314842224, "global_step": 247270, "epoch": 2778, "lr": 1.952053785576491e-06}
{"train_loss": 0.04102132469415665, "global_step": 247271, "epoch": 2778, "lr": 1.9518933808676197e-06}
{"train_loss": 0.013499150052666664, "global_step": 247272, "epoch": 2778, "lr": 1.951732982618243e-06}
{"train_loss": 0.03788156434893608, "global_step": 247273, "epoch": 2778, "lr": 1.95157259082836e-06}
{"train_loss": 0.018480416387319565, "global_step": 247274, "epoch": 2778, "lr": 1.95141220549801e-06}
{"train_loss": 0.06295232474803925, "global_step": 247275, "epoch": 2778, "lr": 1.9512518266271985e-06}
{"train_loss": 0.08820196241140366, "global_step": 247276, "epoch": 2778, "lr": 1.951091454215964e-06}
{"train_loss": 0.04596273973584175, "global_step": 247277, "epoch": 2778, "lr": 1.9509310882643128e-06}
{"train_loss": 0.044490840286016464, "global_step": 247278, "epoch": 2778, "lr": 1.9507707287722778e-06}
{"train_loss": 0.08943616598844528, "global_step": 247279, "epoch": 2778, "lr": 1.9506103757398698e-06}
{"train_loss": 0.02610546164214611, "global_step": 247280, "epoch": 2778, "lr": 1.9504500291671222e-06}
{"train_loss": 0.05392574891448021, "global_step": 247281, "epoch": 2778, "lr": 1.950289689054052e-06}
{"train_loss": 0.04732229560613632, "global_step": 247282, "epoch": 2778, "lr": 1.9501293554006763e-06}
{"train_loss": 0.03791080042719841, "global_step": 247283, "epoch": 2778, "lr": 1.949969028207027e-06}
{"train_loss": 0.06921127438545227, "global_step": 247284, "epoch": 2778, "lr": 1.949808707473111e-06}
{"train_loss": 0.007506830617785454, "global_step": 247285, "epoch": 2778, "lr": 1.9496483931989605e-06}
{"train_loss": 0.017025664448738098, "global_step": 247286, "epoch": 2778, "lr": 1.9494880853845877e-06}
{"train_loss": 0.07986316084861755, "global_step": 247287, "epoch": 2778, "lr": 1.9493277840300304e-06}
{"train_loss": 0.03382004797458649, "global_step": 247288, "epoch": 2778, "lr": 1.9491674891352953e-06}
{"train_loss": 0.06901905685663223, "global_step": 247289, "epoch": 2778, "lr": 1.949007200700409e-06}
{"train_loss": 0.028691258281469345, "global_step": 247290, "epoch": 2778, "lr": 1.9488469187253944e-06}
{"train_loss": 0.06220905855298042, "global_step": 247291, "epoch": 2778, "lr": 1.9486866432102734e-06}
{"train_loss": 0.035713836550712585, "global_step": 247292, "epoch": 2778, "lr": 1.9485263741550577e-06}
{"train_loss": 0.07619697600603104, "global_step": 247293, "epoch": 2778, "lr": 1.9483661115597853e-06}
{"train_loss": 0.038602400571107864, "global_step": 247294, "epoch": 2778, "lr": 1.948205855424462e-06}
{"train_loss": 0.021185437217354774, "global_step": 247295, "epoch": 2778, "lr": 1.948045605749127e-06}
{"train_loss": 0.06474942713975906, "global_step": 247296, "epoch": 2778, "lr": 1.94788536253378e-06}
{"train_loss": 0.03390160948038101, "global_step": 247297, "epoch": 2778, "lr": 1.9477251257784656e-06}
{"train_loss": 0.020671945065259933, "global_step": 247298, "epoch": 2778, "lr": 1.9475648954831836e-06}
{"train_loss": 0.05479716882109642, "global_step": 247299, "epoch": 2778, "lr": 1.9474046716479733e-06}
{"train_loss": 0.05557726323604584, "global_step": 247300, "epoch": 2778, "lr": 1.9472444542728453e-06}
{"train_loss": 0.010430884547531605, "global_step": 247301, "epoch": 2778, "lr": 1.947084243357822e-06}
{"train_loss": 0.06101301312446594, "global_step": 247302, "epoch": 2778, "lr": 1.9469240389029364e-06}
{"train_loss": 0.06032143533229828, "global_step": 247303, "epoch": 2778, "lr": 1.946763840908189e-06}
{"train_loss": 0.054987601935863495, "global_step": 247304, "epoch": 2778, "lr": 1.946603649373624e-06}
{"train_loss": 0.06158509850502014, "global_step": 247305, "epoch": 2778, "lr": 1.946443464299247e-06}
{"train_loss": 0.01954745501279831, "global_step": 247306, "epoch": 2778, "lr": 1.946283285685091e-06}
{"train_loss": 0.03852932155132294, "global_step": 247307, "epoch": 2778, "lr": 1.946123113531162e-06}
{"train_loss": 0.07030363380908966, "global_step": 247308, "epoch": 2778, "lr": 1.9459629478374984e-06}
{"train_loss": 0.05657193437218666, "global_step": 247309, "epoch": 2778, "lr": 1.945802788604112e-06}
{"train_loss": 0.03973814845085144, "global_step": 247310, "epoch": 2778, "lr": 1.9456426358310298e-06}
{"train_loss": 0.06663265079259872, "global_step": 247311, "epoch": 2778, "lr": 1.945482489518263e-06}
{"train_loss": 0.021746279671788216, "global_step": 247312, "epoch": 2778, "lr": 1.9453223496658457e-06}
{"train_loss": 0.06581959128379822, "global_step": 247313, "epoch": 2778, "lr": 1.945162216273794e-06}
{"train_loss": 0.03856021910905838, "global_step": 247314, "epoch": 2778, "lr": 1.9450020893421296e-06}
{"train_loss": 0.036039210855960846, "global_step": 247315, "epoch": 2778, "lr": 1.9448419688708695e-06}
{"train_loss": 0.050300225615501404, "global_step": 247316, "epoch": 2778, "lr": 1.9446818548600474e-06}
{"train_loss": 0.05890960991382599, "global_step": 247317, "epoch": 2778, "lr": 1.944521747309669e-06}
{"train_loss": 0.03174564614892006, "global_step": 247318, "epoch": 2778, "lr": 1.9443616462197724e-06}
{"train_loss": 0.04539193585515022, "global_step": 247319, "epoch": 2778, "lr": 1.944201551590369e-06}
{"train_loss": 0.03211512416601181, "global_step": 247320, "epoch": 2778, "lr": 1.9440414634214752e-06}
{"train_loss": 0.024321189150214195, "global_step": 247321, "epoch": 2778, "lr": 1.9438813817131253e-06}
{"train_loss": 0.03869161754846573, "global_step": 247322, "epoch": 2778, "lr": 1.943721306465329e-06}
{"train_loss": 0.032181911170482635, "global_step": 247323, "epoch": 2778, "lr": 1.943561237678121e-06}
{"train_loss": 0.0373661145567894, "global_step": 247324, "epoch": 2778, "lr": 1.9434011753515115e-06}
{"train_loss": 0.03898118808865547, "global_step": 247325, "epoch": 2778, "lr": 1.9432411194855282e-06}
{"train_loss": 0.05808175727725029, "global_step": 247326, "epoch": 2778, "lr": 1.9430810700801884e-06}
{"train_loss": 0.026146266609430313, "global_step": 247327, "epoch": 2778, "lr": 1.9429210271355136e-06}
{"train_loss": 0.03297541290521622, "global_step": 247328, "epoch": 2778, "lr": 1.9427609906515263e-06}
{"train_loss": 0.045814063400030136, "global_step": 247329, "epoch": 2778, "lr": 1.9426009606282604e-06}
{"train_loss": 0.04252483039110731, "global_step": 247330, "epoch": 2778, "lr": 1.942440937065715e-06, "val_loss": 9.464599609375}
{"train_loss": 0.021225593984127045, "global_step": 247331, "epoch": 2779, "lr": 1.942280919963929e-06}
{"train_loss": 0.03755475580692291, "global_step": 247332, "epoch": 2779, "lr": 1.942120909322909e-06}
{"train_loss": 0.01768413744866848, "global_step": 247333, "epoch": 2779, "lr": 1.941960905142692e-06}
{"train_loss": 0.03695313259959221, "global_step": 247334, "epoch": 2779, "lr": 1.9418009074232914e-06}
{"train_loss": 0.05510549247264862, "global_step": 247335, "epoch": 2779, "lr": 1.9416409161647275e-06}
{"train_loss": 0.03453405946493149, "global_step": 247336, "epoch": 2779, "lr": 1.9414809313670235e-06}
{"train_loss": 0.0675056129693985, "global_step": 247337, "epoch": 2779, "lr": 1.9413209530302014e-06}
{"train_loss": 0.03432197496294975, "global_step": 247338, "epoch": 2779, "lr": 1.9411609811542887e-06}
{"train_loss": 0.04935230687260628, "global_step": 247339, "epoch": 2779, "lr": 1.9410010157392964e-06}
{"train_loss": 0.03523000702261925, "global_step": 247340, "epoch": 2779, "lr": 1.940841056785259e-06}
{"train_loss": 0.036993321031332016, "global_step": 247341, "epoch": 2779, "lr": 1.94068110429218e-06}
{"train_loss": 0.055552925914525986, "global_step": 247342, "epoch": 2779, "lr": 1.9405211582600948e-06}
{"train_loss": 0.03080371394753456, "global_step": 247343, "epoch": 2779, "lr": 1.9403612186890188e-06}
{"train_loss": 0.02964208275079727, "global_step": 247344, "epoch": 2779, "lr": 1.9402012855789796e-06}
{"train_loss": 0.0626458153128624, "global_step": 247345, "epoch": 2779, "lr": 1.940041358929989e-06}
{"train_loss": 0.04230163246393204, "global_step": 247346, "epoch": 2779, "lr": 1.9398814387420805e-06}
{"train_loss": 0.03678637742996216, "global_step": 247347, "epoch": 2779, "lr": 1.9397215250152588e-06}
{"train_loss": 0.07314901053905487, "global_step": 247348, "epoch": 2779, "lr": 1.939561617749569e-06}
{"train_loss": 0.028750810772180557, "global_step": 247349, "epoch": 2779, "lr": 1.9394017169450053e-06}
{"train_loss": 0.04684320464730263, "global_step": 247350, "epoch": 2779, "lr": 1.9392418226016173e-06}
{"train_loss": 0.03912867605686188, "global_step": 247351, "epoch": 2779, "lr": 1.9390819347194e-06}
{"train_loss": 0.08408630639314651, "global_step": 247352, "epoch": 2779, "lr": 1.9389220532983975e-06}
{"train_loss": 0.05345935374498367, "global_step": 247353, "epoch": 2779, "lr": 1.9387621783386156e-06}
{"train_loss": 0.04753706231713295, "global_step": 247354, "epoch": 2779, "lr": 1.938602309840082e-06}
{"train_loss": 0.0440964549779892, "global_step": 247355, "epoch": 2779, "lr": 1.9384424478028185e-06}
{"train_loss": 0.03493499010801315, "global_step": 247356, "epoch": 2779, "lr": 1.938282592226848e-06}
{"train_loss": 0.032834459096193314, "global_step": 247357, "epoch": 2779, "lr": 1.9381227431121863e-06}
{"train_loss": 0.06299864500761032, "global_step": 247358, "epoch": 2779, "lr": 1.9379629004588564e-06}
{"train_loss": 0.07513180375099182, "global_step": 247359, "epoch": 2779, "lr": 1.9378030642668908e-06}
{"train_loss": 0.00925976037979126, "global_step": 247360, "epoch": 2779, "lr": 1.9376432345362907e-06}
{"train_loss": 0.05176827684044838, "global_step": 247361, "epoch": 2779, "lr": 1.9374834112670938e-06}
{"train_loss": 0.03737518936395645, "global_step": 247362, "epoch": 2779, "lr": 1.9373235944593116e-06}
{"train_loss": 0.060105081647634506, "global_step": 247363, "epoch": 2779, "lr": 1.9371637841129774e-06}
{"train_loss": 0.0399760827422142, "global_step": 247364, "epoch": 2779, "lr": 1.9370039802281025e-06}
{"train_loss": 0.08288975059986115, "global_step": 247365, "epoch": 2779, "lr": 1.9368441828047145e-06}
{"train_loss": 0.034367434680461884, "global_step": 247366, "epoch": 2779, "lr": 1.9366843918428247e-06}
{"train_loss": 0.022232506424188614, "global_step": 247367, "epoch": 2779, "lr": 1.9365246073424715e-06}
{"train_loss": 0.05766817927360535, "global_step": 247368, "epoch": 2779, "lr": 1.936364829303661e-06}
{"train_loss": 0.03250463306903839, "global_step": 247369, "epoch": 2779, "lr": 1.9362050577264203e-06}
{"train_loss": 0.047983601689338684, "global_step": 247370, "epoch": 2779, "lr": 1.936045292610772e-06}
{"train_loss": 0.05703249201178551, "global_step": 247371, "epoch": 2779, "lr": 1.9358855339567384e-06}
{"train_loss": 0.036534883081912994, "global_step": 247372, "epoch": 2779, "lr": 1.935725781764336e-06}
{"train_loss": 0.036826036870479584, "global_step": 247373, "epoch": 2779, "lr": 1.9355660360335927e-06}
{"train_loss": 0.03289441764354706, "global_step": 247374, "epoch": 2779, "lr": 1.9354062967645193e-06}
{"train_loss": 0.057267479598522186, "global_step": 247375, "epoch": 2779, "lr": 1.9352465639571495e-06}
{"train_loss": 0.0530153252184391, "global_step": 247376, "epoch": 2779, "lr": 1.935086837611505e-06}
{"train_loss": 0.03556310385465622, "global_step": 247377, "epoch": 2779, "lr": 1.934927117727592e-06}
{"train_loss": 0.03081965446472168, "global_step": 247378, "epoch": 2779, "lr": 1.934767404305454e-06}
{"train_loss": 0.060641758143901825, "global_step": 247379, "epoch": 2779, "lr": 1.934607697345092e-06}
{"train_loss": 0.034958358854055405, "global_step": 247380, "epoch": 2779, "lr": 1.934447996846539e-06}
{"train_loss": 0.05650261044502258, "global_step": 247381, "epoch": 2779, "lr": 1.9342883028098113e-06}
{"train_loss": 0.06802557408809662, "global_step": 247382, "epoch": 2779, "lr": 1.934128615234937e-06}
{"train_loss": 0.05026843026280403, "global_step": 247383, "epoch": 2779, "lr": 1.933968934121927e-06}
{"train_loss": 0.05354255437850952, "global_step": 247384, "epoch": 2779, "lr": 1.93380925947082e-06}
{"train_loss": 0.04111764580011368, "global_step": 247385, "epoch": 2779, "lr": 1.9336495912816164e-06}
{"train_loss": 0.08145271241664886, "global_step": 247386, "epoch": 2779, "lr": 1.933489929554355e-06}
{"train_loss": 0.04496417194604874, "global_step": 247387, "epoch": 2779, "lr": 1.933330274289041e-06}
{"train_loss": 0.06710099428892136, "global_step": 247388, "epoch": 2779, "lr": 1.933170625485714e-06}
{"train_loss": 0.08367352187633514, "global_step": 247389, "epoch": 2779, "lr": 1.933010983144379e-06}
{"train_loss": 0.1201355829834938, "global_step": 247390, "epoch": 2779, "lr": 1.9328513472650688e-06}
{"train_loss": 0.05377238243818283, "global_step": 247391, "epoch": 2779, "lr": 1.932691717847801e-06}
{"train_loss": 0.06232253834605217, "global_step": 247392, "epoch": 2779, "lr": 1.9325320948925973e-06}
{"train_loss": 0.02275276370346546, "global_step": 247393, "epoch": 2779, "lr": 1.9323724783994745e-06}
{"train_loss": 0.05956125259399414, "global_step": 247394, "epoch": 2779, "lr": 1.932212868368466e-06}
{"train_loss": 0.06144629046320915, "global_step": 247395, "epoch": 2779, "lr": 1.932053264799577e-06}
{"train_loss": 0.05512531101703644, "global_step": 247396, "epoch": 2779, "lr": 1.9318936676928468e-06}
{"train_loss": 0.032148901373147964, "global_step": 247397, "epoch": 2779, "lr": 1.9317340770482863e-06}
{"train_loss": 0.05080730468034744, "global_step": 247398, "epoch": 2779, "lr": 1.9315744928659117e-06}
{"train_loss": 0.03836728259921074, "global_step": 247399, "epoch": 2779, "lr": 1.9314149151457573e-06}
{"train_loss": 0.049786072224378586, "global_step": 247400, "epoch": 2779, "lr": 1.9312553438878277e-06}
{"train_loss": 0.03341135382652283, "global_step": 247401, "epoch": 2779, "lr": 1.931095779092168e-06}
{"train_loss": 0.057666558772325516, "global_step": 247402, "epoch": 2779, "lr": 1.9309362207587777e-06}
{"train_loss": 0.02017270028591156, "global_step": 247403, "epoch": 2779, "lr": 1.9307766688876904e-06}
{"train_loss": 0.06985550373792648, "global_step": 247404, "epoch": 2779, "lr": 1.9306171234789228e-06}
{"train_loss": 0.04794365540146828, "global_step": 247405, "epoch": 2779, "lr": 1.9304575845325024e-06}
{"train_loss": 0.03234928101301193, "global_step": 247406, "epoch": 2779, "lr": 1.9302980520484402e-06}
{"train_loss": 0.0766865462064743, "global_step": 247407, "epoch": 2779, "lr": 1.93013852602677e-06}
{"train_loss": 0.04053165018558502, "global_step": 247408, "epoch": 2779, "lr": 1.929979006467497e-06}
{"train_loss": 0.06002705544233322, "global_step": 247409, "epoch": 2779, "lr": 1.929819493370666e-06}
{"train_loss": 0.01967477984726429, "global_step": 247410, "epoch": 2779, "lr": 1.929659986736271e-06}
{"train_loss": 0.03128022700548172, "global_step": 247411, "epoch": 2779, "lr": 1.9295004865643507e-06}
{"train_loss": 0.031999342143535614, "global_step": 247412, "epoch": 2779, "lr": 1.9293409928549277e-06}
{"train_loss": 0.06459623575210571, "global_step": 247413, "epoch": 2779, "lr": 1.9291815056080187e-06}
{"train_loss": 0.025448091328144073, "global_step": 247414, "epoch": 2779, "lr": 1.929022024823646e-06}
{"train_loss": 0.04079177603125572, "global_step": 247415, "epoch": 2779, "lr": 1.9288625505018253e-06}
{"train_loss": 0.054239336401224136, "global_step": 247416, "epoch": 2779, "lr": 1.9287030826425857e-06}
{"train_loss": 0.07470040768384933, "global_step": 247417, "epoch": 2779, "lr": 1.9285436212459483e-06}
{"train_loss": 0.04495526850223541, "global_step": 247418, "epoch": 2779, "lr": 1.9283841663119305e-06}
{"train_loss": 0.048413489704554, "global_step": 247419, "epoch": 2779, "lr": 1.928224717840554e-06, "val_loss": 9.491745948791504}
{"train_loss": 0.027780191972851753, "global_step": 247420, "epoch": 2780, "lr": 1.9280652758318475e-06}
{"train_loss": 0.0988716185092926, "global_step": 247421, "epoch": 2780, "lr": 1.9279058402858154e-06}
{"train_loss": 0.03229441121220589, "global_step": 247422, "epoch": 2780, "lr": 1.9277464112025024e-06}
{"train_loss": 0.04732128232717514, "global_step": 247423, "epoch": 2780, "lr": 1.927586988581909e-06}
{"train_loss": 0.05750299617648125, "global_step": 247424, "epoch": 2780, "lr": 1.927427572424073e-06}
{"train_loss": 0.0726548582315445, "global_step": 247425, "epoch": 2780, "lr": 1.927268162729001e-06}
{"train_loss": 0.05047500506043434, "global_step": 247426, "epoch": 2780, "lr": 1.9271087594967317e-06}
{"train_loss": 0.024243701249361038, "global_step": 247427, "epoch": 2780, "lr": 1.9269493627272704e-06}
{"train_loss": 0.01692393608391285, "global_step": 247428, "epoch": 2780, "lr": 1.9267899724206395e-06}
{"train_loss": 0.04789190739393234, "global_step": 247429, "epoch": 2780, "lr": 1.9266305885768776e-06}
{"train_loss": 0.0640946626663208, "global_step": 247430, "epoch": 2780, "lr": 1.9264712111959903e-06}
{"train_loss": 0.06056118756532669, "global_step": 247431, "epoch": 2780, "lr": 1.9263118402779944e-06}
{"train_loss": 0.04665928706526756, "global_step": 247432, "epoch": 2780, "lr": 1.9261524758229344e-06}
{"train_loss": 0.08199282735586166, "global_step": 247433, "epoch": 2780, "lr": 1.9259931178308044e-06}
{"train_loss": 0.04715275391936302, "global_step": 247434, "epoch": 2780, "lr": 1.925833766301649e-06}
{"train_loss": 0.06626272946596146, "global_step": 247435, "epoch": 2780, "lr": 1.925674421235474e-06}
{"train_loss": 0.05074677616357803, "global_step": 247436, "epoch": 2780, "lr": 1.9255150826323066e-06}
{"train_loss": 0.03257133066654205, "global_step": 247437, "epoch": 2780, "lr": 1.9253557504921694e-06}
{"train_loss": 0.0704970583319664, "global_step": 247438, "epoch": 2780, "lr": 1.9251964248150735e-06}
{"train_loss": 0.041034113615751266, "global_step": 247439, "epoch": 2780, "lr": 1.925037105601063e-06}
{"train_loss": 0.02215603180229664, "global_step": 247440, "epoch": 2780, "lr": 1.9248777928501326e-06}
{"train_loss": 0.015821117907762527, "global_step": 247441, "epoch": 2780, "lr": 1.924718486562327e-06}
{"train_loss": 0.0324450358748436, "global_step": 247442, "epoch": 2780, "lr": 1.9245591867376456e-06}
{"train_loss": 0.05027265101671219, "global_step": 247443, "epoch": 2780, "lr": 1.924399893376133e-06}
{"train_loss": 0.08634999394416809, "global_step": 247444, "epoch": 2780, "lr": 1.9242406064777895e-06}
{"train_loss": 0.03378931060433388, "global_step": 247445, "epoch": 2780, "lr": 1.9240813260426538e-06}
{"train_loss": 0.05999308079481125, "global_step": 247446, "epoch": 2780, "lr": 1.9239220520707313e-06}
{"train_loss": 0.040732491761446, "global_step": 247447, "epoch": 2780, "lr": 1.9237627845620553e-06}
{"train_loss": 0.06220296397805214, "global_step": 247448, "epoch": 2780, "lr": 1.923603523516643e-06}
{"train_loss": 0.024574412032961845, "global_step": 247449, "epoch": 2780, "lr": 1.9234442689345155e-06}
{"train_loss": 0.0477224662899971, "global_step": 247450, "epoch": 2780, "lr": 1.9232850208157018e-06}
{"train_loss": 0.04716154932975769, "global_step": 247451, "epoch": 2780, "lr": 1.9231257791602064e-06}
{"train_loss": 0.08393220603466034, "global_step": 247452, "epoch": 2780, "lr": 1.922966543968069e-06}
{"train_loss": 0.019716428592801094, "global_step": 247453, "epoch": 2780, "lr": 1.9228073152392944e-06}
{"train_loss": 0.03191149979829788, "global_step": 247454, "epoch": 2780, "lr": 1.9226480929739223e-06}
{"train_loss": 0.03936631605029106, "global_step": 247455, "epoch": 2780, "lr": 1.922488877171952e-06}
{"train_loss": 0.06546759605407715, "global_step": 247456, "epoch": 2780, "lr": 1.922329667833428e-06}
{"train_loss": 0.022115763276815414, "global_step": 247457, "epoch": 2780, "lr": 1.922170464958356e-06}
{"train_loss": 0.043860238045454025, "global_step": 247458, "epoch": 2780, "lr": 1.9220112685467638e-06}
{"train_loss": 0.04037738963961601, "global_step": 247459, "epoch": 2780, "lr": 1.9218520785986683e-06}
{"train_loss": 0.02592703513801098, "global_step": 247460, "epoch": 2780, "lr": 1.921692895114102e-06}
{"train_loss": 0.042297277599573135, "global_step": 247461, "epoch": 2780, "lr": 1.921533718093066e-06}
{"train_loss": 0.024037763476371765, "global_step": 247462, "epoch": 2780, "lr": 1.921374547535604e-06}
{"train_loss": 0.03264419734477997, "global_step": 247463, "epoch": 2780, "lr": 1.9212153834417213e-06}
{"train_loss": 0.042340029031038284, "global_step": 247464, "epoch": 2780, "lr": 1.9210562258114407e-06}
{"train_loss": 0.051264166831970215, "global_step": 247465, "epoch": 2780, "lr": 1.9208970746448007e-06}
{"train_loss": 0.02888055518269539, "global_step": 247466, "epoch": 2780, "lr": 1.9207379299418015e-06}
{"train_loss": 0.06001925468444824, "global_step": 247467, "epoch": 2780, "lr": 1.9205787917024763e-06}
{"train_loss": 0.028111808001995087, "global_step": 247468, "epoch": 2780, "lr": 1.9204196599268477e-06}
{"train_loss": 0.04251205548644066, "global_step": 247469, "epoch": 2780, "lr": 1.9202605346149203e-06}
{"train_loss": 0.04093904048204422, "global_step": 247470, "epoch": 2780, "lr": 1.9201014157667395e-06}
{"train_loss": 0.039106398820877075, "global_step": 247471, "epoch": 2780, "lr": 1.9199423033823107e-06}
{"train_loss": 0.027917761355638504, "global_step": 247472, "epoch": 2780, "lr": 1.919783197461661e-06}
{"train_loss": 0.07516884803771973, "global_step": 247473, "epoch": 2780, "lr": 1.919624098004802e-06}
{"train_loss": 0.035432346165180206, "global_step": 247474, "epoch": 2780, "lr": 1.919465005011778e-06}
{"train_loss": 0.061824869364500046, "global_step": 247475, "epoch": 2780, "lr": 1.919305918482589e-06}
{"train_loss": 0.03548872098326683, "global_step": 247476, "epoch": 2780, "lr": 1.919146838417257e-06}
{"train_loss": 0.04856488108634949, "global_step": 247477, "epoch": 2780, "lr": 1.9189877648158217e-06}
{"train_loss": 0.011402199044823647, "global_step": 247478, "epoch": 2780, "lr": 1.918828697678282e-06}
{"train_loss": 0.05671199783682823, "global_step": 247479, "epoch": 2780, "lr": 1.9186696370046775e-06}
{"train_loss": 0.031393978744745255, "global_step": 247480, "epoch": 2780, "lr": 1.9185105827950133e-06}
{"train_loss": 0.03283272683620453, "global_step": 247481, "epoch": 2780, "lr": 1.9183515350493284e-06}
{"train_loss": 0.06100117787718773, "global_step": 247482, "epoch": 2780, "lr": 1.9181924937676286e-06}
{"train_loss": 0.034013938158750534, "global_step": 247483, "epoch": 2780, "lr": 1.9180334589499417e-06}
{"train_loss": 0.0290233064442873, "global_step": 247484, "epoch": 2780, "lr": 1.9178744305962948e-06}
{"train_loss": 0.03462255373597145, "global_step": 247485, "epoch": 2780, "lr": 1.917715408706694e-06}
{"train_loss": 0.06399322301149368, "global_step": 247486, "epoch": 2780, "lr": 1.9175563932811834e-06}
{"train_loss": 0.05059552937746048, "global_step": 247487, "epoch": 2780, "lr": 1.9173973843197578e-06}
{"train_loss": 0.08975692093372345, "global_step": 247488, "epoch": 2780, "lr": 1.9172383818224616e-06}
{"train_loss": 0.10460412502288818, "global_step": 247489, "epoch": 2780, "lr": 1.9170793857893e-06}
{"train_loss": 0.10991723090410233, "global_step": 247490, "epoch": 2780, "lr": 1.916920396220312e-06}
{"train_loss": 0.03799143806099892, "global_step": 247491, "epoch": 2780, "lr": 1.9167614131154977e-06}
{"train_loss": 0.051157575100660324, "global_step": 247492, "epoch": 2780, "lr": 1.9166024364748957e-06}
{"train_loss": 0.04745366424322128, "global_step": 247493, "epoch": 2780, "lr": 1.916443466298512e-06}
{"train_loss": 0.05907980352640152, "global_step": 247494, "epoch": 2780, "lr": 1.9162845025863853e-06}
{"train_loss": 0.05654691904783249, "global_step": 247495, "epoch": 2780, "lr": 1.916125545338521e-06}
{"train_loss": 0.09316875040531158, "global_step": 247496, "epoch": 2780, "lr": 1.915966594554952e-06}
{"train_loss": 0.04536201432347298, "global_step": 247497, "epoch": 2780, "lr": 1.9158076502356903e-06}
{"train_loss": 0.04083085060119629, "global_step": 247498, "epoch": 2780, "lr": 1.915648712380769e-06}
{"train_loss": 0.04945540800690651, "global_step": 247499, "epoch": 2780, "lr": 1.9154897809901984e-06}
{"train_loss": 0.029145216569304466, "global_step": 247500, "epoch": 2780, "lr": 1.915330856064007e-06}
{"train_loss": 0.032055340707302094, "global_step": 247501, "epoch": 2780, "lr": 1.9151719376022058e-06}
{"train_loss": 0.014429496601223946, "global_step": 247502, "epoch": 2780, "lr": 1.915013025604828e-06}
{"train_loss": 0.027969105169177055, "global_step": 247503, "epoch": 2780, "lr": 1.9148541200718962e-06}
{"train_loss": 0.03296847641468048, "global_step": 247504, "epoch": 2780, "lr": 1.9146952210034152e-06}
{"train_loss": 0.024729471653699875, "global_step": 247505, "epoch": 2780, "lr": 1.91453632839943e-06}
{"train_loss": 0.052946969866752625, "global_step": 247506, "epoch": 2780, "lr": 1.9143774422599404e-06}
{"train_loss": 0.018964743241667747, "global_step": 247507, "epoch": 2780, "lr": 1.9142185625849796e-06}
{"train_loss": 0.04645228268725149, "global_step": 247508, "epoch": 2780, "lr": 1.91405968937457e-06, "val_loss": 9.545110702514648, "train_action_mse_error": 6.687889099121094}
{"train_loss": 0.03261217847466469, "global_step": 247509, "epoch": 2781, "lr": 1.9139008226287223e-06}
{"train_loss": 0.07045843452215195, "global_step": 247510, "epoch": 2781, "lr": 1.9137419623474707e-06}
{"train_loss": 0.010072275996208191, "global_step": 247511, "epoch": 2781, "lr": 1.9135831085308253e-06}
{"train_loss": 0.06908728182315826, "global_step": 247512, "epoch": 2781, "lr": 1.9134242611788146e-06}
{"train_loss": 0.05888741835951805, "global_step": 247513, "epoch": 2781, "lr": 1.91326542029146e-06}
{"train_loss": 0.027051957324147224, "global_step": 247514, "epoch": 2781, "lr": 1.9131065858687793e-06}
{"train_loss": 0.02095257118344307, "global_step": 247515, "epoch": 2781, "lr": 1.9129477579107936e-06}
{"train_loss": 0.04596973583102226, "global_step": 247516, "epoch": 2781, "lr": 1.9127889364175256e-06}
{"train_loss": 0.050406262278556824, "global_step": 247517, "epoch": 2781, "lr": 1.9126301213889974e-06}
{"train_loss": 0.07488691061735153, "global_step": 247518, "epoch": 2781, "lr": 1.9124713128252316e-06}
{"train_loss": 0.07260103523731232, "global_step": 247519, "epoch": 2781, "lr": 1.912312510726244e-06}
{"train_loss": 0.03046206571161747, "global_step": 247520, "epoch": 2781, "lr": 1.912153715092063e-06}
{"train_loss": 0.029159799218177795, "global_step": 247521, "epoch": 2781, "lr": 1.9119949259227054e-06}
{"train_loss": 0.05445997416973114, "global_step": 247522, "epoch": 2781, "lr": 1.9118361432181986e-06}
{"train_loss": 0.04175585135817528, "global_step": 247523, "epoch": 2781, "lr": 1.9116773669785536e-06}
{"train_loss": 0.04468390718102455, "global_step": 247524, "epoch": 2781, "lr": 1.9115185972037985e-06}
{"train_loss": 0.031164975836873055, "global_step": 247525, "epoch": 2781, "lr": 1.9113598338939555e-06}
{"train_loss": 0.024153949692845345, "global_step": 247526, "epoch": 2781, "lr": 1.911201077049046e-06}
{"train_loss": 0.04158513993024826, "global_step": 247527, "epoch": 2781, "lr": 1.9110423266690823e-06}
{"train_loss": 0.04665543511509895, "global_step": 247528, "epoch": 2781, "lr": 1.9108835827541026e-06}
{"train_loss": 0.0049291434697806835, "global_step": 247529, "epoch": 2781, "lr": 1.9107248453041073e-06}
{"train_loss": 0.025332383811473846, "global_step": 247530, "epoch": 2781, "lr": 1.91056611431914e-06}
{"train_loss": 0.02674654871225357, "global_step": 247531, "epoch": 2781, "lr": 1.9104073897992014e-06}
{"train_loss": 0.04266473650932312, "global_step": 247532, "epoch": 2781, "lr": 1.91024867174433e-06}
{"train_loss": 0.01165869738906622, "global_step": 247533, "epoch": 2781, "lr": 1.910089960154532e-06}
{"train_loss": 0.04709908366203308, "global_step": 247534, "epoch": 2781, "lr": 1.9099312550298453e-06}
{"train_loss": 0.01240527629852295, "global_step": 247535, "epoch": 2781, "lr": 1.909772556370276e-06}
{"train_loss": 0.050718292593955994, "global_step": 247536, "epoch": 2781, "lr": 1.909613864175858e-06}
{"train_loss": 0.04481081664562225, "global_step": 247537, "epoch": 2781, "lr": 1.909455178446595e-06}
{"train_loss": 0.037665314972400665, "global_step": 247538, "epoch": 2781, "lr": 1.9092964991825222e-06}
{"train_loss": 0.043285615742206573, "global_step": 247539, "epoch": 2781, "lr": 1.909137826383667e-06}
{"train_loss": 0.04997601732611656, "global_step": 247540, "epoch": 2781, "lr": 1.908979160050034e-06}
{"train_loss": 0.03732256591320038, "global_step": 247541, "epoch": 2781, "lr": 1.9088205001816573e-06}
{"train_loss": 0.019692881032824516, "global_step": 247542, "epoch": 2781, "lr": 1.908661846778553e-06}
{"train_loss": 0.05210549384355545, "global_step": 247543, "epoch": 2781, "lr": 1.9085031998407444e-06}
{"train_loss": 0.055099062621593475, "global_step": 247544, "epoch": 2781, "lr": 1.908344559368247e-06}
{"train_loss": 0.024624761193990707, "global_step": 247545, "epoch": 2781, "lr": 1.908185925361089e-06}
{"train_loss": 0.04112805798649788, "global_step": 247546, "epoch": 2781, "lr": 1.9080272978192926e-06}
{"train_loss": 0.061573728919029236, "global_step": 247547, "epoch": 2781, "lr": 1.907868676742869e-06}
{"train_loss": 0.04072348400950432, "global_step": 247548, "epoch": 2781, "lr": 1.907710062131851e-06}
{"train_loss": 0.07816334068775177, "global_step": 247549, "epoch": 2781, "lr": 1.9075514539862506e-06}
{"train_loss": 0.0457315593957901, "global_step": 247550, "epoch": 2781, "lr": 1.9073928523061004e-06}
{"train_loss": 0.019307665526866913, "global_step": 247551, "epoch": 2781, "lr": 1.907234257091417e-06}
{"train_loss": 0.04862118139863014, "global_step": 247552, "epoch": 2781, "lr": 1.907075668342212e-06}
{"train_loss": 0.06852467358112335, "global_step": 247553, "epoch": 2781, "lr": 1.9069170860585184e-06}
{"train_loss": 0.03144954890012741, "global_step": 247554, "epoch": 2781, "lr": 1.9067585102403475e-06}
{"train_loss": 0.03364359214901924, "global_step": 247555, "epoch": 2781, "lr": 1.9065999408877266e-06}
{"train_loss": 0.03673486039042473, "global_step": 247556, "epoch": 2781, "lr": 1.906441378000684e-06}
{"train_loss": 0.0527178980410099, "global_step": 247557, "epoch": 2781, "lr": 1.9062828215792305e-06}
{"train_loss": 0.038088276982307434, "global_step": 247558, "epoch": 2781, "lr": 1.906124271623394e-06}
{"train_loss": 0.06786657869815826, "global_step": 247559, "epoch": 2781, "lr": 1.905965728133191e-06}
{"train_loss": 0.037381190806627274, "global_step": 247560, "epoch": 2781, "lr": 1.9058071911086495e-06}
{"train_loss": 0.07098434120416641, "global_step": 247561, "epoch": 2781, "lr": 1.9056486605497747e-06}
{"train_loss": 0.044698767364025116, "global_step": 247562, "epoch": 2781, "lr": 1.9054901364566113e-06}
{"train_loss": 0.03796880319714546, "global_step": 247563, "epoch": 2781, "lr": 1.9053316188291592e-06}
{"train_loss": 0.061428818851709366, "global_step": 247564, "epoch": 2781, "lr": 1.9051731076674573e-06}
{"train_loss": 0.060123227536678314, "global_step": 247565, "epoch": 2781, "lr": 1.905014602971511e-06}
{"train_loss": 0.024571560323238373, "global_step": 247566, "epoch": 2781, "lr": 1.9048561047413537e-06}
{"train_loss": 0.0320337750017643, "global_step": 247567, "epoch": 2781, "lr": 1.9046976129770023e-06}
{"train_loss": 0.05744762718677521, "global_step": 247568, "epoch": 2781, "lr": 1.9045391276784786e-06}
{"train_loss": 0.03753507137298584, "global_step": 247569, "epoch": 2781, "lr": 1.9043806488457993e-06}
{"train_loss": 0.022192571312189102, "global_step": 247570, "epoch": 2781, "lr": 1.904222176478998e-06}
{"train_loss": 0.08365463465452194, "global_step": 247571, "epoch": 2781, "lr": 1.90406371057808e-06}
{"train_loss": 0.04632897675037384, "global_step": 247572, "epoch": 2781, "lr": 1.9039052511430789e-06}
{"train_loss": 0.02393103949725628, "global_step": 247573, "epoch": 2781, "lr": 1.9037467981740054e-06}
{"train_loss": 0.08904553204774857, "global_step": 247574, "epoch": 2781, "lr": 1.9035883516708874e-06}
{"train_loss": 0.030340945348143578, "global_step": 247575, "epoch": 2781, "lr": 1.9034299116337528e-06}
{"train_loss": 0.0332145132124424, "global_step": 247576, "epoch": 2781, "lr": 1.9032714780626126e-06}
{"train_loss": 0.049909282475709915, "global_step": 247577, "epoch": 2781, "lr": 1.903113050957489e-06}
{"train_loss": 0.0489351823925972, "global_step": 247578, "epoch": 2781, "lr": 1.9029546303184098e-06}
{"train_loss": 0.10105106979608536, "global_step": 247579, "epoch": 2781, "lr": 1.9027962161453916e-06}
{"train_loss": 0.03132029250264168, "global_step": 247580, "epoch": 2781, "lr": 1.902637808438451e-06}
{"train_loss": 0.054168928414583206, "global_step": 247581, "epoch": 2781, "lr": 1.9024794071976216e-06}
{"train_loss": 0.10949201136827469, "global_step": 247582, "epoch": 2781, "lr": 1.9023210124229085e-06}
{"train_loss": 0.049480140209198, "global_step": 247583, "epoch": 2781, "lr": 1.9021626241143564e-06}
{"train_loss": 0.07123653590679169, "global_step": 247584, "epoch": 2781, "lr": 1.9020042422719652e-06}
{"train_loss": 0.058583226054906845, "global_step": 247585, "epoch": 2781, "lr": 1.9018458668957573e-06}
{"train_loss": 0.040360141545534134, "global_step": 247586, "epoch": 2781, "lr": 1.9016874979857657e-06}
{"train_loss": 0.0574292354285717, "global_step": 247587, "epoch": 2781, "lr": 1.9015291355420018e-06}
{"train_loss": 0.02537255361676216, "global_step": 247588, "epoch": 2781, "lr": 1.9013707795644987e-06}
{"train_loss": 0.06158416345715523, "global_step": 247589, "epoch": 2781, "lr": 1.901212430053262e-06}
{"train_loss": 0.03205838426947594, "global_step": 247590, "epoch": 2781, "lr": 1.9010540870083305e-06}
{"train_loss": 0.09846844524145126, "global_step": 247591, "epoch": 2781, "lr": 1.9008957504297042e-06}
{"train_loss": 0.02295602113008499, "global_step": 247592, "epoch": 2781, "lr": 1.9007374203174278e-06}
{"train_loss": 0.027430111542344093, "global_step": 247593, "epoch": 2781, "lr": 1.9005790966715009e-06}
{"train_loss": 0.04386096075177193, "global_step": 247594, "epoch": 2781, "lr": 1.9004207794919626e-06}
{"train_loss": 0.04174182936549187, "global_step": 247595, "epoch": 2781, "lr": 1.9002624687788185e-06}
{"train_loss": 0.07263240963220596, "global_step": 247596, "epoch": 2781, "lr": 1.9001041645321073e-06}
{"train_loss": 0.045625263453576335, "global_step": 247597, "epoch": 2781, "lr": 1.8999458667518344e-06, "val_loss": 9.467463493347168}
{"train_loss": 0.0437689945101738, "global_step": 247598, "epoch": 2782, "lr": 1.8997875754380335e-06}
{"train_loss": 0.03302265331149101, "global_step": 247599, "epoch": 2782, "lr": 1.89962929059071e-06}
{"train_loss": 0.03409824147820473, "global_step": 247600, "epoch": 2782, "lr": 1.8994710122099025e-06}
{"train_loss": 0.03739308938384056, "global_step": 247601, "epoch": 2782, "lr": 1.8993127402956224e-06}
{"train_loss": 0.028018010780215263, "global_step": 247602, "epoch": 2782, "lr": 1.8991544748478973e-06}
{"train_loss": 0.060042474418878555, "global_step": 247603, "epoch": 2782, "lr": 1.8989962158667384e-06}
{"train_loss": 0.033860646188259125, "global_step": 247604, "epoch": 2782, "lr": 1.898837963352179e-06}
{"train_loss": 0.033907536417245865, "global_step": 247605, "epoch": 2782, "lr": 1.8986797173042302e-06}
{"train_loss": 0.05103453993797302, "global_step": 247606, "epoch": 2782, "lr": 1.89852147772292e-06}
{"train_loss": 0.036444276571273804, "global_step": 247607, "epoch": 2782, "lr": 1.8983632446082645e-06}
{"train_loss": 0.015660148113965988, "global_step": 247608, "epoch": 2782, "lr": 1.8982050179602916e-06}
{"train_loss": 0.07651041448116302, "global_step": 247609, "epoch": 2782, "lr": 1.898046797779013e-06}
{"train_loss": 0.029040472581982613, "global_step": 247610, "epoch": 2782, "lr": 1.8978885840644556e-06}
{"train_loss": 0.04768464341759682, "global_step": 247611, "epoch": 2782, "lr": 1.8977303768166532e-06}
{"train_loss": 0.07790318131446838, "global_step": 247612, "epoch": 2782, "lr": 1.8975721760356003e-06}
{"train_loss": 0.05849771574139595, "global_step": 247613, "epoch": 2782, "lr": 1.897413981721341e-06}
{"train_loss": 0.0595746748149395, "global_step": 247614, "epoch": 2782, "lr": 1.897255793873881e-06}
{"train_loss": 0.039379965513944626, "global_step": 247615, "epoch": 2782, "lr": 1.8970976124932593e-06}
{"train_loss": 0.05724973604083061, "global_step": 247616, "epoch": 2782, "lr": 1.8969394375794758e-06}
{"train_loss": 0.03660483658313751, "global_step": 247617, "epoch": 2782, "lr": 1.8967812691325748e-06}
{"train_loss": 0.04858749732375145, "global_step": 247618, "epoch": 2782, "lr": 1.8966231071525509e-06}
{"train_loss": 0.04595106840133667, "global_step": 247619, "epoch": 2782, "lr": 1.8964649516394485e-06}
{"train_loss": 0.05417032167315483, "global_step": 247620, "epoch": 2782, "lr": 1.8963068025932785e-06}
{"train_loss": 0.06680262833833694, "global_step": 247621, "epoch": 2782, "lr": 1.8961486600140631e-06}
{"train_loss": 0.06934785097837448, "global_step": 247622, "epoch": 2782, "lr": 1.895990523901825e-06}
{"train_loss": 0.04781722649931908, "global_step": 247623, "epoch": 2782, "lr": 1.8958323942565859e-06}
{"train_loss": 0.04968579486012459, "global_step": 247624, "epoch": 2782, "lr": 1.8956742710783682e-06}
{"train_loss": 0.054836489260196686, "global_step": 247625, "epoch": 2782, "lr": 1.895516154367183e-06}
{"train_loss": 0.05549873411655426, "global_step": 247626, "epoch": 2782, "lr": 1.8953580441230635e-06}
{"train_loss": 0.049136191606521606, "global_step": 247627, "epoch": 2782, "lr": 1.8951999403460264e-06}
{"train_loss": 0.033059969544410706, "global_step": 247628, "epoch": 2782, "lr": 1.895041843036094e-06}
{"train_loss": 0.012919940054416656, "global_step": 247629, "epoch": 2782, "lr": 1.8948837521932827e-06}
{"train_loss": 0.03531128540635109, "global_step": 247630, "epoch": 2782, "lr": 1.8947256678176262e-06}
{"train_loss": 0.06093868613243103, "global_step": 247631, "epoch": 2782, "lr": 1.8945675899091297e-06}
{"train_loss": 0.02659003809094429, "global_step": 247632, "epoch": 2782, "lr": 1.8944095184678267e-06}
{"train_loss": 0.04201938211917877, "global_step": 247633, "epoch": 2782, "lr": 1.8942514534937338e-06}
{"train_loss": 0.017833616584539413, "global_step": 247634, "epoch": 2782, "lr": 1.8940933949868734e-06}
{"train_loss": 0.019697174429893494, "global_step": 247635, "epoch": 2782, "lr": 1.8939353429472618e-06}
{"train_loss": 0.021200986579060555, "global_step": 247636, "epoch": 2782, "lr": 1.8937772973749267e-06}
{"train_loss": 0.013772957026958466, "global_step": 247637, "epoch": 2782, "lr": 1.8936192582698854e-06}
{"train_loss": 0.026818014681339264, "global_step": 247638, "epoch": 2782, "lr": 1.8934612256321648e-06}
{"train_loss": 0.044932108372449875, "global_step": 247639, "epoch": 2782, "lr": 1.8933031994617767e-06}
{"train_loss": 0.08413440734148026, "global_step": 247640, "epoch": 2782, "lr": 1.8931451797587484e-06}
{"train_loss": 0.038303155452013016, "global_step": 247641, "epoch": 2782, "lr": 1.8929871665231025e-06}
{"train_loss": 0.07484573125839233, "global_step": 247642, "epoch": 2782, "lr": 1.8928291597548608e-06}
{"train_loss": 0.04622936621308327, "global_step": 247643, "epoch": 2782, "lr": 1.8926711594540346e-06}
{"train_loss": 0.025873644277453423, "global_step": 247644, "epoch": 2782, "lr": 1.8925131656206573e-06}
{"train_loss": 0.0642252266407013, "global_step": 247645, "epoch": 2782, "lr": 1.89235517825474e-06}
{"train_loss": 0.032743439078330994, "global_step": 247646, "epoch": 2782, "lr": 1.89219719735631e-06}
{"train_loss": 0.04960104078054428, "global_step": 247647, "epoch": 2782, "lr": 1.8920392229253958e-06}
{"train_loss": 0.059866346418857574, "global_step": 247648, "epoch": 2782, "lr": 1.8918812549620024e-06}
{"train_loss": 0.03023115172982216, "global_step": 247649, "epoch": 2782, "lr": 1.8917232934661687e-06}
{"train_loss": 0.04964146018028259, "global_step": 247650, "epoch": 2782, "lr": 1.8915653384378951e-06}
{"train_loss": 0.08930786699056625, "global_step": 247651, "epoch": 2782, "lr": 1.8914073898772255e-06}
{"train_loss": 0.05051356181502342, "global_step": 247652, "epoch": 2782, "lr": 1.8912494477841603e-06}
{"train_loss": 0.027046239003539085, "global_step": 247653, "epoch": 2782, "lr": 1.8910915121587381e-06}
{"train_loss": 0.046836502850055695, "global_step": 247654, "epoch": 2782, "lr": 1.8909335830009645e-06}
{"train_loss": 0.038296740502119064, "global_step": 247655, "epoch": 2782, "lr": 1.890775660310873e-06}
{"train_loss": 0.046429235488176346, "global_step": 247656, "epoch": 2782, "lr": 1.8906177440884797e-06}
{"train_loss": 0.03728622943162918, "global_step": 247657, "epoch": 2782, "lr": 1.8904598343338076e-06}
{"train_loss": 0.05997657775878906, "global_step": 247658, "epoch": 2782, "lr": 1.8903019310468727e-06}
{"train_loss": 0.04523497819900513, "global_step": 247659, "epoch": 2782, "lr": 1.8901440342277032e-06}
{"train_loss": 0.06343547254800797, "global_step": 247660, "epoch": 2782, "lr": 1.8899861438763157e-06}
{"train_loss": 0.029205208644270897, "global_step": 247661, "epoch": 2782, "lr": 1.8898282599927375e-06}
{"train_loss": 0.04865312948822975, "global_step": 247662, "epoch": 2782, "lr": 1.8896703825769857e-06}
{"train_loss": 0.04283427819609642, "global_step": 247663, "epoch": 2782, "lr": 1.8895125116290768e-06}
{"train_loss": 0.03321198374032974, "global_step": 247664, "epoch": 2782, "lr": 1.8893546471490386e-06}
{"train_loss": 0.07880233973264694, "global_step": 247665, "epoch": 2782, "lr": 1.8891967891368879e-06}
{"train_loss": 0.0427103266119957, "global_step": 247666, "epoch": 2782, "lr": 1.889038937592652e-06}
{"train_loss": 0.048559095710515976, "global_step": 247667, "epoch": 2782, "lr": 1.888881092516348e-06}
{"train_loss": 0.051301006227731705, "global_step": 247668, "epoch": 2782, "lr": 1.888723253907998e-06}
{"train_loss": 0.03369787335395813, "global_step": 247669, "epoch": 2782, "lr": 1.8885654217676186e-06}
{"train_loss": 0.06883983314037323, "global_step": 247670, "epoch": 2782, "lr": 1.8884075960952374e-06}
{"train_loss": 0.029719313606619835, "global_step": 247671, "epoch": 2782, "lr": 1.8882497768908713e-06}
{"train_loss": 0.08940710127353668, "global_step": 247672, "epoch": 2782, "lr": 1.888091964154548e-06}
{"train_loss": 0.06685052812099457, "global_step": 247673, "epoch": 2782, "lr": 1.8879341578862786e-06}
{"train_loss": 0.03985828161239624, "global_step": 247674, "epoch": 2782, "lr": 1.887776358086102e-06}
{"train_loss": 0.013403243385255337, "global_step": 247675, "epoch": 2782, "lr": 1.8876185647540124e-06}
{"train_loss": 0.027703242376446724, "global_step": 247676, "epoch": 2782, "lr": 1.8874607778900544e-06}
{"train_loss": 0.052225567400455475, "global_step": 247677, "epoch": 2782, "lr": 1.8873029974942392e-06}
{"train_loss": 0.024690410122275352, "global_step": 247678, "epoch": 2782, "lr": 1.8871452235665943e-06}
{"train_loss": 0.02051091380417347, "global_step": 247679, "epoch": 2782, "lr": 1.8869874561071254e-06}
{"train_loss": 0.07127349823713303, "global_step": 247680, "epoch": 2782, "lr": 1.886829695115877e-06}
{"train_loss": 0.03221442922949791, "global_step": 247681, "epoch": 2782, "lr": 1.886671940592849e-06}
{"train_loss": 0.04289801046252251, "global_step": 247682, "epoch": 2782, "lr": 1.886514192538069e-06}
{"train_loss": 0.048257164657115936, "global_step": 247683, "epoch": 2782, "lr": 1.8863564509515708e-06}
{"train_loss": 0.05697082355618477, "global_step": 247684, "epoch": 2782, "lr": 1.8861987158333594e-06}
{"train_loss": 0.03740186616778374, "global_step": 247685, "epoch": 2782, "lr": 1.8860409871834628e-06}
{"train_loss": 0.04508942444128602, "global_step": 247686, "epoch": 2782, "lr": 1.885883265001903e-06, "val_loss": 9.419827461242676}
{"train_loss": 0.056682225316762924, "global_step": 247687, "epoch": 2783, "lr": 1.8857255492887027e-06}
{"train_loss": 0.04192688688635826, "global_step": 247688, "epoch": 2783, "lr": 1.8855678400438725e-06}
{"train_loss": 0.07612717151641846, "global_step": 247689, "epoch": 2783, "lr": 1.885410137267446e-06}
{"train_loss": 0.04879213497042656, "global_step": 247690, "epoch": 2783, "lr": 1.8852524409594395e-06}
{"train_loss": 0.02001803182065487, "global_step": 247691, "epoch": 2783, "lr": 1.8850947511198758e-06}
{"train_loss": 0.0225827619433403, "global_step": 247692, "epoch": 2783, "lr": 1.884937067748771e-06}
{"train_loss": 0.03466981649398804, "global_step": 247693, "epoch": 2783, "lr": 1.884779390846153e-06}
{"train_loss": 0.026428114622831345, "global_step": 247694, "epoch": 2783, "lr": 1.8846217204120387e-06}
{"train_loss": 0.025018854066729546, "global_step": 247695, "epoch": 2783, "lr": 1.8844640564464499e-06}
{"train_loss": 0.06692440062761307, "global_step": 247696, "epoch": 2783, "lr": 1.8843063989494092e-06}
{"train_loss": 0.06258466094732285, "global_step": 247697, "epoch": 2783, "lr": 1.8841487479209385e-06}
{"train_loss": 0.01967005990445614, "global_step": 247698, "epoch": 2783, "lr": 1.8839911033610547e-06}
{"train_loss": 0.030649319291114807, "global_step": 247699, "epoch": 2783, "lr": 1.8838334652697853e-06}
{"train_loss": 0.03834180161356926, "global_step": 247700, "epoch": 2783, "lr": 1.8836758336471471e-06}
{"train_loss": 0.01726139895617962, "global_step": 247701, "epoch": 2783, "lr": 1.8835182084931623e-06}
{"train_loss": 0.06634826958179474, "global_step": 247702, "epoch": 2783, "lr": 1.8833605898078531e-06}
{"train_loss": 0.11263399571180344, "global_step": 247703, "epoch": 2783, "lr": 1.8832029775912363e-06}
{"train_loss": 0.02640024945139885, "global_step": 247704, "epoch": 2783, "lr": 1.8830453718433394e-06}
{"train_loss": 0.05489305779337883, "global_step": 247705, "epoch": 2783, "lr": 1.8828877725641736e-06}
{"train_loss": 0.020088866353034973, "global_step": 247706, "epoch": 2783, "lr": 1.8827301797537777e-06}
{"train_loss": 0.058398839086294174, "global_step": 247707, "epoch": 2783, "lr": 1.8825725934121518e-06}
{"train_loss": 0.03689088672399521, "global_step": 247708, "epoch": 2783, "lr": 1.8824150135393348e-06}
{"train_loss": 0.027970321476459503, "global_step": 247709, "epoch": 2783, "lr": 1.8822574401353376e-06}
{"train_loss": 0.040321268141269684, "global_step": 247710, "epoch": 2783, "lr": 1.8820998732001882e-06}
{"train_loss": 0.02561165764927864, "global_step": 247711, "epoch": 2783, "lr": 1.8819423127338976e-06}
{"train_loss": 0.0773349478840828, "global_step": 247712, "epoch": 2783, "lr": 1.8817847587364989e-06}
{"train_loss": 0.017032749950885773, "global_step": 247713, "epoch": 2783, "lr": 1.8816272112080036e-06}
{"train_loss": 0.12189950048923492, "global_step": 247714, "epoch": 2783, "lr": 1.8814696701484391e-06}
{"train_loss": 0.019543426111340523, "global_step": 247715, "epoch": 2783, "lr": 1.8813121355578222e-06}
{"train_loss": 0.040685441344976425, "global_step": 247716, "epoch": 2783, "lr": 1.8811546074361807e-06}
{"train_loss": 0.12104242295026779, "global_step": 247717, "epoch": 2783, "lr": 1.8809970857835313e-06}
{"train_loss": 0.024690307676792145, "global_step": 247718, "epoch": 2783, "lr": 1.8808395705998904e-06}
{"train_loss": 0.050275109708309174, "global_step": 247719, "epoch": 2783, "lr": 1.8806820618852861e-06}
{"train_loss": 0.027231577783823013, "global_step": 247720, "epoch": 2783, "lr": 1.8805245596397346e-06}
{"train_loss": 0.034308403730392456, "global_step": 247721, "epoch": 2783, "lr": 1.8803670638632698e-06}
{"train_loss": 0.04582422599196434, "global_step": 247722, "epoch": 2783, "lr": 1.8802095745558967e-06}
{"train_loss": 0.043038126081228256, "global_step": 247723, "epoch": 2783, "lr": 1.8800520917176433e-06}
{"train_loss": 0.018990937620401382, "global_step": 247724, "epoch": 2783, "lr": 1.8798946153485319e-06}
{"train_loss": 0.0419633574783802, "global_step": 247725, "epoch": 2783, "lr": 1.879737145448579e-06}
{"train_loss": 0.029851708561182022, "global_step": 247726, "epoch": 2783, "lr": 1.8795796820178124e-06}
{"train_loss": 0.03165629506111145, "global_step": 247727, "epoch": 2783, "lr": 1.8794222250562487e-06}
{"train_loss": 0.036346323788166046, "global_step": 247728, "epoch": 2783, "lr": 1.8792647745639103e-06}
{"train_loss": 0.08147524297237396, "global_step": 247729, "epoch": 2783, "lr": 1.8791073305408192e-06}
{"train_loss": 0.010167242027819157, "global_step": 247730, "epoch": 2783, "lr": 1.878949892986992e-06}
{"train_loss": 0.0358116514980793, "global_step": 247731, "epoch": 2783, "lr": 1.8787924619024566e-06}
{"train_loss": 0.02393624186515808, "global_step": 247732, "epoch": 2783, "lr": 1.8786350372872298e-06}
{"train_loss": 0.03886454552412033, "global_step": 247733, "epoch": 2783, "lr": 1.8784776191413333e-06}
{"train_loss": 0.04474084824323654, "global_step": 247734, "epoch": 2783, "lr": 1.8783202074647899e-06}
{"train_loss": 0.043873850256204605, "global_step": 247735, "epoch": 2783, "lr": 1.8781628022576215e-06}
{"train_loss": 0.03108419105410576, "global_step": 247736, "epoch": 2783, "lr": 1.8780054035198447e-06}
{"train_loss": 0.0476534478366375, "global_step": 247737, "epoch": 2783, "lr": 1.8778480112514873e-06}
{"train_loss": 0.026916326954960823, "global_step": 247738, "epoch": 2783, "lr": 1.8776906254525606e-06}
{"train_loss": 0.07421775162220001, "global_step": 247739, "epoch": 2783, "lr": 1.8775332461230977e-06}
{"train_loss": 0.048119932413101196, "global_step": 247740, "epoch": 2783, "lr": 1.8773758732631152e-06}
{"train_loss": 0.04770919308066368, "global_step": 247741, "epoch": 2783, "lr": 1.87721850687263e-06}
{"train_loss": 0.051444556564092636, "global_step": 247742, "epoch": 2783, "lr": 1.8770611469516698e-06}
{"train_loss": 0.035164106637239456, "global_step": 247743, "epoch": 2783, "lr": 1.8769037935002455e-06}
{"train_loss": 0.03903599828481674, "global_step": 247744, "epoch": 2783, "lr": 1.8767464465183904e-06}
{"train_loss": 0.07169454544782639, "global_step": 247745, "epoch": 2783, "lr": 1.8765891060061103e-06}
{"train_loss": 0.04861054569482803, "global_step": 247746, "epoch": 2783, "lr": 1.8764317719634495e-06}
{"train_loss": 0.04032110422849655, "global_step": 247747, "epoch": 2783, "lr": 1.8762744443904078e-06}
{"train_loss": 0.030082466080784798, "global_step": 247748, "epoch": 2783, "lr": 1.876117123287019e-06}
{"train_loss": 0.04548683762550354, "global_step": 247749, "epoch": 2783, "lr": 1.8759598086532937e-06}
{"train_loss": 0.03064187988638878, "global_step": 247750, "epoch": 2783, "lr": 1.8758025004892654e-06}
{"train_loss": 0.031796619296073914, "global_step": 247751, "epoch": 2783, "lr": 1.8756451987949397e-06}
{"train_loss": 0.021295366808772087, "global_step": 247752, "epoch": 2783, "lr": 1.875487903570361e-06}
{"train_loss": 0.07723432034254074, "global_step": 247753, "epoch": 2783, "lr": 1.8753306148155237e-06}
{"train_loss": 0.005117658991366625, "global_step": 247754, "epoch": 2783, "lr": 1.875173332530472e-06}
{"train_loss": 0.10661132633686066, "global_step": 247755, "epoch": 2783, "lr": 1.8750160567152065e-06}
{"train_loss": 0.05736580491065979, "global_step": 247756, "epoch": 2783, "lr": 1.8748587873697599e-06}
{"train_loss": 0.015194983221590519, "global_step": 247757, "epoch": 2783, "lr": 1.8747015244941601e-06}
{"train_loss": 0.02264164760708809, "global_step": 247758, "epoch": 2783, "lr": 1.874544268088413e-06}
{"train_loss": 0.030899392440915108, "global_step": 247759, "epoch": 2783, "lr": 1.8743870181525514e-06}
{"train_loss": 0.037577733397483826, "global_step": 247760, "epoch": 2783, "lr": 1.8742297746865867e-06}
{"train_loss": 0.007107384968549013, "global_step": 247761, "epoch": 2783, "lr": 1.874072537690552e-06}
{"train_loss": 0.04632721468806267, "global_step": 247762, "epoch": 2783, "lr": 1.8739153071644533e-06}
{"train_loss": 0.03353676572442055, "global_step": 247763, "epoch": 2783, "lr": 1.873758083108329e-06}
{"train_loss": 0.07172542810440063, "global_step": 247764, "epoch": 2783, "lr": 1.8736008655221848e-06}
{"train_loss": 0.053611647337675095, "global_step": 247765, "epoch": 2783, "lr": 1.8734436544060541e-06}
{"train_loss": 0.050236694514751434, "global_step": 247766, "epoch": 2783, "lr": 1.8732864497599423e-06}
{"train_loss": 0.029374239966273308, "global_step": 247767, "epoch": 2783, "lr": 1.873129251583894e-06}
{"train_loss": 0.055869508534669876, "global_step": 247768, "epoch": 2783, "lr": 1.8729720598779034e-06}
{"train_loss": 0.05144554376602173, "global_step": 247769, "epoch": 2783, "lr": 1.8728148746420148e-06}
{"train_loss": 0.12722629308700562, "global_step": 247770, "epoch": 2783, "lr": 1.8726576958762342e-06}
{"train_loss": 0.04284173995256424, "global_step": 247771, "epoch": 2783, "lr": 1.8725005235805949e-06}
{"train_loss": 0.028944989666342735, "global_step": 247772, "epoch": 2783, "lr": 1.872343357755102e-06}
{"train_loss": 0.06939031183719635, "global_step": 247773, "epoch": 2783, "lr": 1.8721861983997946e-06}
{"train_loss": 0.04977206513285637, "global_step": 247774, "epoch": 2783, "lr": 1.8720290455146782e-06}
{"train_loss": 0.04447651689014073, "global_step": 247775, "epoch": 2783, "lr": 1.871871899099792e-06, "val_loss": 9.580154418945312}
{"train_loss": 0.03604363650083542, "global_step": 247776, "epoch": 2784, "lr": 1.8717147591551354e-06}
{"train_loss": 0.046561866998672485, "global_step": 247777, "epoch": 2784, "lr": 1.871557625680742e-06}
{"train_loss": 0.036277152597904205, "global_step": 247778, "epoch": 2784, "lr": 1.8714004986766342e-06}
{"train_loss": 0.0541229248046875, "global_step": 247779, "epoch": 2784, "lr": 1.8712433781428284e-06}
{"train_loss": 0.05737077444791794, "global_step": 247780, "epoch": 2784, "lr": 1.871086264079347e-06}
{"train_loss": 0.03337306156754494, "global_step": 247781, "epoch": 2784, "lr": 1.870929156486212e-06}
{"train_loss": 0.11627630889415741, "global_step": 247782, "epoch": 2784, "lr": 1.8707720553634456e-06}
{"train_loss": 0.02736843377351761, "global_step": 247783, "epoch": 2784, "lr": 1.870614960711059e-06}
{"train_loss": 0.02688783034682274, "global_step": 247784, "epoch": 2784, "lr": 1.870457872529091e-06}
{"train_loss": 0.0733490139245987, "global_step": 247785, "epoch": 2784, "lr": 1.8703007908175473e-06}
{"train_loss": 0.021968694403767586, "global_step": 247786, "epoch": 2784, "lr": 1.8701437155764612e-06}
{"train_loss": 0.03840886056423187, "global_step": 247787, "epoch": 2784, "lr": 1.8699866468058436e-06}
{"train_loss": 0.04161079600453377, "global_step": 247788, "epoch": 2784, "lr": 1.8698295845057223e-06}
{"train_loss": 0.06126577779650688, "global_step": 247789, "epoch": 2784, "lr": 1.8696725286761086e-06}
{"train_loss": 0.018372314050793648, "global_step": 247790, "epoch": 2784, "lr": 1.869515479317041e-06}
{"train_loss": 0.046905335038900375, "global_step": 247791, "epoch": 2784, "lr": 1.86935843642852e-06}
{"train_loss": 0.03486113250255585, "global_step": 247792, "epoch": 2784, "lr": 1.8692014000105784e-06}
{"train_loss": 0.03447875380516052, "global_step": 247793, "epoch": 2784, "lr": 1.8690443700632444e-06}
{"train_loss": 0.0334877111017704, "global_step": 247794, "epoch": 2784, "lr": 1.8688873465865232e-06}
{"train_loss": 0.013216814957559109, "global_step": 247795, "epoch": 2784, "lr": 1.8687303295804482e-06}
{"train_loss": 0.026880715042352676, "global_step": 247796, "epoch": 2784, "lr": 1.868573319045036e-06}
{"train_loss": 0.04595908150076866, "global_step": 247797, "epoch": 2784, "lr": 1.8684163149803092e-06}
{"train_loss": 0.06892191618680954, "global_step": 247798, "epoch": 2784, "lr": 1.8682593173862785e-06}
{"train_loss": 0.037668079137802124, "global_step": 247799, "epoch": 2784, "lr": 1.8681023262629827e-06}
{"train_loss": 0.02703000418841839, "global_step": 247800, "epoch": 2784, "lr": 1.8679453416104276e-06}
{"train_loss": 0.0631946474313736, "global_step": 247801, "epoch": 2784, "lr": 1.8677883634286463e-06}
{"train_loss": 0.012097119353711605, "global_step": 247802, "epoch": 2784, "lr": 1.8676313917176446e-06}
{"train_loss": 0.036419741809368134, "global_step": 247803, "epoch": 2784, "lr": 1.8674744264774613e-06}
{"train_loss": 0.025127125903964043, "global_step": 247804, "epoch": 2784, "lr": 1.8673174677081073e-06}
{"train_loss": 0.019739391282200813, "global_step": 247805, "epoch": 2784, "lr": 1.8671605154096106e-06}
{"train_loss": 0.02723376452922821, "global_step": 247806, "epoch": 2784, "lr": 1.8670035695819764e-06}
{"train_loss": 0.04729657992720604, "global_step": 247807, "epoch": 2784, "lr": 1.8668466302252496e-06}
{"train_loss": 0.0387125201523304, "global_step": 247808, "epoch": 2784, "lr": 1.8666896973394299e-06}
{"train_loss": 0.030941110104322433, "global_step": 247809, "epoch": 2784, "lr": 1.866532770924545e-06}
{"train_loss": 0.03260514885187149, "global_step": 247810, "epoch": 2784, "lr": 1.8663758509806284e-06}
{"train_loss": 0.028741812333464622, "global_step": 247811, "epoch": 2784, "lr": 1.8662189375076855e-06}
{"train_loss": 0.01714581623673439, "global_step": 247812, "epoch": 2784, "lr": 1.8660620305057385e-06}
{"train_loss": 0.0252403374761343, "global_step": 247813, "epoch": 2784, "lr": 1.8659051299748209e-06}
{"train_loss": 0.04913896694779396, "global_step": 247814, "epoch": 2784, "lr": 1.865748235914938e-06}
{"train_loss": 0.05173647403717041, "global_step": 247815, "epoch": 2784, "lr": 1.865591348326129e-06}
{"train_loss": 0.023673057556152344, "global_step": 247816, "epoch": 2784, "lr": 1.8654344672083933e-06}
{"train_loss": 0.05844685435295105, "global_step": 247817, "epoch": 2784, "lr": 1.8652775925617704e-06}
{"train_loss": 0.022254878655076027, "global_step": 247818, "epoch": 2784, "lr": 1.8651207243862768e-06}
{"train_loss": 0.06458202749490738, "global_step": 247819, "epoch": 2784, "lr": 1.8649638626819233e-06}
{"train_loss": 0.06646287441253662, "global_step": 247820, "epoch": 2784, "lr": 1.8648070074487435e-06}
{"train_loss": 0.03699720650911331, "global_step": 247821, "epoch": 2784, "lr": 1.8646501586867482e-06}
{"train_loss": 0.03219318017363548, "global_step": 247822, "epoch": 2784, "lr": 1.864493316395971e-06}
{"train_loss": 0.03305305168032646, "global_step": 247823, "epoch": 2784, "lr": 1.8643364805764174e-06}
{"train_loss": 0.016700560227036476, "global_step": 247824, "epoch": 2784, "lr": 1.8641796512281262e-06}
{"train_loss": 0.07764232158660889, "global_step": 247825, "epoch": 2784, "lr": 1.864022828351103e-06}
{"train_loss": 0.010739817284047604, "global_step": 247826, "epoch": 2784, "lr": 1.863866011945381e-06}
{"train_loss": 0.01672510989010334, "global_step": 247827, "epoch": 2784, "lr": 1.8637092020109659e-06}
{"train_loss": 0.061090100556612015, "global_step": 247828, "epoch": 2784, "lr": 1.8635523985478964e-06}
{"train_loss": 0.021295929327607155, "global_step": 247829, "epoch": 2784, "lr": 1.8633956015561837e-06}
{"train_loss": 0.03444244712591171, "global_step": 247830, "epoch": 2784, "lr": 1.8632388110358501e-06}
{"train_loss": 0.03969944268465042, "global_step": 247831, "epoch": 2784, "lr": 1.8630820269869175e-06}
{"train_loss": 0.052631959319114685, "global_step": 247832, "epoch": 2784, "lr": 1.8629252494094085e-06}
{"train_loss": 0.10295695066452026, "global_step": 247833, "epoch": 2784, "lr": 1.8627684783033394e-06}
{"train_loss": 0.03661833703517914, "global_step": 247834, "epoch": 2784, "lr": 1.8626117136687382e-06}
{"train_loss": 0.06899496167898178, "global_step": 247835, "epoch": 2784, "lr": 1.8624549555056214e-06}
{"train_loss": 0.03265063092112541, "global_step": 247836, "epoch": 2784, "lr": 1.8622982038140057e-06}
{"train_loss": 0.025840962305665016, "global_step": 247837, "epoch": 2784, "lr": 1.8621414585939244e-06}
{"train_loss": 0.03192431479692459, "global_step": 247838, "epoch": 2784, "lr": 1.8619847198453832e-06}
{"train_loss": 0.07163260132074356, "global_step": 247839, "epoch": 2784, "lr": 1.8618279875684208e-06}
{"train_loss": 0.04005082696676254, "global_step": 247840, "epoch": 2784, "lr": 1.8616712617630428e-06}
{"train_loss": 0.04255997762084007, "global_step": 247841, "epoch": 2784, "lr": 1.8615145424292823e-06}
{"train_loss": 0.07475924491882324, "global_step": 247842, "epoch": 2784, "lr": 1.8613578295671452e-06}
{"train_loss": 0.05133301019668579, "global_step": 247843, "epoch": 2784, "lr": 1.8612011231766701e-06}
{"train_loss": 0.0188656747341156, "global_step": 247844, "epoch": 2784, "lr": 1.8610444232578628e-06}
{"train_loss": 0.02658260241150856, "global_step": 247845, "epoch": 2784, "lr": 1.8608877298107508e-06}
{"train_loss": 0.021703040227293968, "global_step": 247846, "epoch": 2784, "lr": 1.8607310428353674e-06}
{"train_loss": 0.05772508680820465, "global_step": 247847, "epoch": 2784, "lr": 1.8605743623317074e-06}
{"train_loss": 0.0395013727247715, "global_step": 247848, "epoch": 2784, "lr": 1.8604176882998202e-06}
{"train_loss": 0.01196865551173687, "global_step": 247849, "epoch": 2784, "lr": 1.8602610207397008e-06}
{"train_loss": 0.04076708108186722, "global_step": 247850, "epoch": 2784, "lr": 1.8601043596513933e-06}
{"train_loss": 0.037313297390937805, "global_step": 247851, "epoch": 2784, "lr": 1.8599477050349034e-06}
{"train_loss": 0.07385826110839844, "global_step": 247852, "epoch": 2784, "lr": 1.8597910568902531e-06}
{"train_loss": 0.07881690561771393, "global_step": 247853, "epoch": 2784, "lr": 1.859634415217476e-06}
{"train_loss": 0.03976373001933098, "global_step": 247854, "epoch": 2784, "lr": 1.8594777800165775e-06}
{"train_loss": 0.023505274206399918, "global_step": 247855, "epoch": 2784, "lr": 1.8593211512875909e-06}
{"train_loss": 0.08024178445339203, "global_step": 247856, "epoch": 2784, "lr": 1.8591645290305272e-06}
{"train_loss": 0.02338891662657261, "global_step": 247857, "epoch": 2784, "lr": 1.8590079132454087e-06}
{"train_loss": 0.015155073255300522, "global_step": 247858, "epoch": 2784, "lr": 1.8588513039322686e-06}
{"train_loss": 0.0358797125518322, "global_step": 247859, "epoch": 2784, "lr": 1.8586947010911126e-06}
{"train_loss": 0.021406076848506927, "global_step": 247860, "epoch": 2784, "lr": 1.858538104721974e-06}
{"train_loss": 0.005196006502956152, "global_step": 247861, "epoch": 2784, "lr": 1.8583815148248584e-06}
{"train_loss": 0.040188394486904144, "global_step": 247862, "epoch": 2784, "lr": 1.8582249313998046e-06}
{"train_loss": 0.017011886462569237, "global_step": 247863, "epoch": 2784, "lr": 1.8580683544468237e-06}
{"train_loss": 0.04029809736406033, "global_step": 247864, "epoch": 2784, "lr": 1.8579117839659321e-06, "val_loss": 9.503655433654785}
{"train_loss": 0.020302588120102882, "global_step": 247865, "epoch": 2785, "lr": 1.8577552199571691e-06}
{"train_loss": 0.024168066680431366, "global_step": 247866, "epoch": 2785, "lr": 1.8575986624205343e-06}
{"train_loss": 0.05317245423793793, "global_step": 247867, "epoch": 2785, "lr": 1.857442111356067e-06}
{"train_loss": 0.030682334676384926, "global_step": 247868, "epoch": 2785, "lr": 1.8572855667637723e-06}
{"train_loss": 0.05237005650997162, "global_step": 247869, "epoch": 2785, "lr": 1.8571290286436894e-06}
{"train_loss": 0.06724617630243301, "global_step": 247870, "epoch": 2785, "lr": 1.856972496995818e-06}
{"train_loss": 0.061927471309900284, "global_step": 247871, "epoch": 2785, "lr": 1.8568159718201972e-06}
{"train_loss": 0.035850800573825836, "global_step": 247872, "epoch": 2785, "lr": 1.8566594531168325e-06}
{"train_loss": 0.03152104839682579, "global_step": 247873, "epoch": 2785, "lr": 1.8565029408857627e-06}
{"train_loss": 0.10657286643981934, "global_step": 247874, "epoch": 2785, "lr": 1.8563464351269877e-06}
{"train_loss": 0.07034175097942352, "global_step": 247875, "epoch": 2785, "lr": 1.8561899358405522e-06}
{"train_loss": 0.05147717148065567, "global_step": 247876, "epoch": 2785, "lr": 1.8560334430264558e-06}
{"train_loss": 0.02423734962940216, "global_step": 247877, "epoch": 2785, "lr": 1.8558769566847378e-06}
{"train_loss": 0.04244603216648102, "global_step": 247878, "epoch": 2785, "lr": 1.8557204768154035e-06}
{"train_loss": 0.028325533494353294, "global_step": 247879, "epoch": 2785, "lr": 1.8555640034184862e-06}
{"train_loss": 0.047428473830223083, "global_step": 247880, "epoch": 2785, "lr": 1.8554075364939916e-06}
{"train_loss": 0.046790555119514465, "global_step": 247881, "epoch": 2785, "lr": 1.855251076041964e-06}
{"train_loss": 0.02877149172127247, "global_step": 247882, "epoch": 2785, "lr": 1.8550946220623977e-06}
{"train_loss": 0.028895234689116478, "global_step": 247883, "epoch": 2785, "lr": 1.8549381745553317e-06}
{"train_loss": 0.028536507859826088, "global_step": 247884, "epoch": 2785, "lr": 1.8547817335207884e-06}
{"train_loss": 0.04031095281243324, "global_step": 247885, "epoch": 2785, "lr": 1.8546252989587732e-06}
{"train_loss": 0.07356271147727966, "global_step": 247886, "epoch": 2785, "lr": 1.8544688708693304e-06}
{"train_loss": 0.05509146302938461, "global_step": 247887, "epoch": 2785, "lr": 1.8543124492524545e-06}
{"train_loss": 0.020290985703468323, "global_step": 247888, "epoch": 2785, "lr": 1.8541560341081898e-06}
{"train_loss": 0.038076117634773254, "global_step": 247889, "epoch": 2785, "lr": 1.8539996254365421e-06}
{"train_loss": 0.05125084146857262, "global_step": 247890, "epoch": 2785, "lr": 1.8538432232375336e-06}
{"train_loss": 0.02583131566643715, "global_step": 247891, "epoch": 2785, "lr": 1.8536868275111974e-06}
{"train_loss": 0.052896108478307724, "global_step": 247892, "epoch": 2785, "lr": 1.853530438257539e-06}
{"train_loss": 0.030033079907298088, "global_step": 247893, "epoch": 2785, "lr": 1.8533740554765921e-06}
{"train_loss": 0.06832136958837509, "global_step": 247894, "epoch": 2785, "lr": 1.8532176791683732e-06}
{"train_loss": 0.02592785842716694, "global_step": 247895, "epoch": 2785, "lr": 1.853061309332893e-06}
{"train_loss": 0.028502367436885834, "global_step": 247896, "epoch": 2785, "lr": 1.8529049459701908e-06}
{"train_loss": 0.023009592667222023, "global_step": 247897, "epoch": 2785, "lr": 1.852748589080272e-06}
{"train_loss": 0.06991574168205261, "global_step": 247898, "epoch": 2785, "lr": 1.85259223866317e-06}
{"train_loss": 0.030074747279286385, "global_step": 247899, "epoch": 2785, "lr": 1.8524358947188957e-06}
{"train_loss": 0.03475956618785858, "global_step": 247900, "epoch": 2785, "lr": 1.8522795572474716e-06}
{"train_loss": 0.030013617128133774, "global_step": 247901, "epoch": 2785, "lr": 1.8521232262489307e-06}
{"train_loss": 0.057229362428188324, "global_step": 247902, "epoch": 2785, "lr": 1.8519669017232788e-06}
{"train_loss": 0.00786724966019392, "global_step": 247903, "epoch": 2785, "lr": 1.8518105836705434e-06}
{"train_loss": 0.03893885761499405, "global_step": 247904, "epoch": 2785, "lr": 1.851654272090747e-06}
{"train_loss": 0.061441145837306976, "global_step": 247905, "epoch": 2785, "lr": 1.8514979669839116e-06}
{"train_loss": 0.08097974956035614, "global_step": 247906, "epoch": 2785, "lr": 1.8513416683500484e-06}
{"train_loss": 0.02715701051056385, "global_step": 247907, "epoch": 2785, "lr": 1.8511853761891906e-06}
{"train_loss": 0.03051052615046501, "global_step": 247908, "epoch": 2785, "lr": 1.851029090501355e-06}
{"train_loss": 0.05867570638656616, "global_step": 247909, "epoch": 2785, "lr": 1.8508728112865581e-06}
{"train_loss": 0.038181424140930176, "global_step": 247910, "epoch": 2785, "lr": 1.8507165385448277e-06}
{"train_loss": 0.04267159849405289, "global_step": 247911, "epoch": 2785, "lr": 1.8505602722761806e-06}
{"train_loss": 0.02558341808617115, "global_step": 247912, "epoch": 2785, "lr": 1.8504040124806332e-06}
{"train_loss": 0.045219168066978455, "global_step": 247913, "epoch": 2785, "lr": 1.8502477591582245e-06}
{"train_loss": 0.019557345658540726, "global_step": 247914, "epoch": 2785, "lr": 1.8500915123089491e-06}
{"train_loss": 0.05615713447332382, "global_step": 247915, "epoch": 2785, "lr": 1.8499352719328512e-06}
{"train_loss": 0.015594321303069592, "global_step": 247916, "epoch": 2785, "lr": 1.8497790380299418e-06}
{"train_loss": 0.06730751693248749, "global_step": 247917, "epoch": 2785, "lr": 1.8496228106002433e-06}
{"train_loss": 0.0765809565782547, "global_step": 247918, "epoch": 2785, "lr": 1.849466589643767e-06}
{"train_loss": 0.02477117069065571, "global_step": 247919, "epoch": 2785, "lr": 1.8493103751605511e-06}
{"train_loss": 0.09756514430046082, "global_step": 247920, "epoch": 2785, "lr": 1.8491541671506074e-06}
{"train_loss": 0.04746982827782631, "global_step": 247921, "epoch": 2785, "lr": 1.8489979656139578e-06}
{"train_loss": 0.04730002209544182, "global_step": 247922, "epoch": 2785, "lr": 1.8488417705506245e-06}
{"train_loss": 0.07342154532670975, "global_step": 247923, "epoch": 2785, "lr": 1.8486855819606242e-06}
{"train_loss": 0.07670484483242035, "global_step": 247924, "epoch": 2785, "lr": 1.8485293998439902e-06}
{"train_loss": 0.04069826751947403, "global_step": 247925, "epoch": 2785, "lr": 1.8483732242007223e-06}
{"train_loss": 0.02961590327322483, "global_step": 247926, "epoch": 2785, "lr": 1.8482170550308653e-06}
{"train_loss": 0.024459468200802803, "global_step": 247927, "epoch": 2785, "lr": 1.8480608923344244e-06}
{"train_loss": 0.028226835653185844, "global_step": 247928, "epoch": 2785, "lr": 1.847904736111422e-06}
{"train_loss": 0.05794171243906021, "global_step": 247929, "epoch": 2785, "lr": 1.847748586361886e-06}
{"train_loss": 0.08262334018945694, "global_step": 247930, "epoch": 2785, "lr": 1.8475924430858272e-06}
{"train_loss": 0.06505802273750305, "global_step": 247931, "epoch": 2785, "lr": 1.8474363062832788e-06}
{"train_loss": 0.030858753249049187, "global_step": 247932, "epoch": 2785, "lr": 1.8472801759542468e-06}
{"train_loss": 0.03495445474982262, "global_step": 247933, "epoch": 2785, "lr": 1.8471240520987753e-06}
{"train_loss": 0.02411622181534767, "global_step": 247934, "epoch": 2785, "lr": 1.8469679347168645e-06}
{"train_loss": 0.03544165566563606, "global_step": 247935, "epoch": 2785, "lr": 1.8468118238085364e-06}
{"train_loss": 0.032192423939704895, "global_step": 247936, "epoch": 2785, "lr": 1.8466557193738187e-06}
{"train_loss": 0.0658920556306839, "global_step": 247937, "epoch": 2785, "lr": 1.8464996214127395e-06}
{"train_loss": 0.034722454845905304, "global_step": 247938, "epoch": 2785, "lr": 1.8463435299252984e-06}
{"train_loss": 0.07388130575418472, "global_step": 247939, "epoch": 2785, "lr": 1.8461874449115457e-06}
{"train_loss": 0.04304088279604912, "global_step": 247940, "epoch": 2785, "lr": 1.8460313663714702e-06}
{"train_loss": 0.06545450538396835, "global_step": 247941, "epoch": 2785, "lr": 1.8458752943051215e-06}
{"train_loss": 0.032432593405246735, "global_step": 247942, "epoch": 2785, "lr": 1.8457192287125002e-06}
{"train_loss": 0.030308017507195473, "global_step": 247943, "epoch": 2785, "lr": 1.8455631695936393e-06}
{"train_loss": 0.03143121302127838, "global_step": 247944, "epoch": 2785, "lr": 1.8454071169485498e-06}
{"train_loss": 0.04344834014773369, "global_step": 247945, "epoch": 2785, "lr": 1.8452510707772708e-06}
{"train_loss": 0.040054600685834885, "global_step": 247946, "epoch": 2785, "lr": 1.8450950310797964e-06}
{"train_loss": 0.05362587422132492, "global_step": 247947, "epoch": 2785, "lr": 1.8449389978561716e-06}
{"train_loss": 0.021161047741770744, "global_step": 247948, "epoch": 2785, "lr": 1.8447829711064014e-06}
{"train_loss": 0.05134672671556473, "global_step": 247949, "epoch": 2785, "lr": 1.8446269508305192e-06}
{"train_loss": 0.08425020426511765, "global_step": 247950, "epoch": 2785, "lr": 1.8444709370285363e-06}
{"train_loss": 0.04873520880937576, "global_step": 247951, "epoch": 2785, "lr": 1.8443149297004803e-06}
{"train_loss": 0.053490299731492996, "global_step": 247952, "epoch": 2785, "lr": 1.8441589288463622e-06}
{"train_loss": 0.04496521264147223, "global_step": 247953, "epoch": 2785, "lr": 1.8440029344662212e-06, "val_loss": 9.4473295211792, "train_action_mse_error": 19.084848403930664}
{"train_loss": 0.03523322567343712, "global_step": 247954, "epoch": 2786, "lr": 1.8438469465600572e-06}
{"train_loss": 0.03556939214468002, "global_step": 247955, "epoch": 2786, "lr": 1.8436909651278977e-06}
{"train_loss": 0.027079321444034576, "global_step": 247956, "epoch": 2786, "lr": 1.843534990169782e-06}
{"train_loss": 0.09906875342130661, "global_step": 247957, "epoch": 2786, "lr": 1.843379021685704e-06}
{"train_loss": 0.05687609314918518, "global_step": 247958, "epoch": 2786, "lr": 1.8432230596757028e-06}
{"train_loss": 0.03998105973005295, "global_step": 247959, "epoch": 2786, "lr": 1.8430671041397895e-06}
{"train_loss": 0.024292660877108574, "global_step": 247960, "epoch": 2786, "lr": 1.8429111550779921e-06}
{"train_loss": 0.028888029977679253, "global_step": 247961, "epoch": 2786, "lr": 1.8427552124903214e-06}
{"train_loss": 0.038152437657117844, "global_step": 247962, "epoch": 2786, "lr": 1.8425992763768162e-06}
{"train_loss": 0.011194351129233837, "global_step": 247963, "epoch": 2786, "lr": 1.842443346737477e-06}
{"train_loss": 0.01662808656692505, "global_step": 247964, "epoch": 2786, "lr": 1.842287423572342e-06}
{"train_loss": 0.02899295650422573, "global_step": 247965, "epoch": 2786, "lr": 1.842131506881417e-06}
{"train_loss": 0.04639100283384323, "global_step": 247966, "epoch": 2786, "lr": 1.8419755966647412e-06}
{"train_loss": 0.0589836947619915, "global_step": 247967, "epoch": 2786, "lr": 1.8418196929223197e-06}
{"train_loss": 0.08057032525539398, "global_step": 247968, "epoch": 2786, "lr": 1.841663795654175e-06}
{"train_loss": 0.07109753042459488, "global_step": 247969, "epoch": 2786, "lr": 1.8415079048603345e-06}
{"train_loss": 0.056372348219156265, "global_step": 247970, "epoch": 2786, "lr": 1.8413520205408097e-06}
{"train_loss": 0.029643161222338676, "global_step": 247971, "epoch": 2786, "lr": 1.8411961426956392e-06}
{"train_loss": 0.05600685626268387, "global_step": 247972, "epoch": 2786, "lr": 1.841040271324823e-06}
{"train_loss": 0.049540601670742035, "global_step": 247973, "epoch": 2786, "lr": 1.8408844064284003e-06}
{"train_loss": 0.042221490293741226, "global_step": 247974, "epoch": 2786, "lr": 1.8407285480063763e-06}
{"train_loss": 0.05713315308094025, "global_step": 247975, "epoch": 2786, "lr": 1.8405726960587844e-06}
{"train_loss": 0.03206279128789902, "global_step": 247976, "epoch": 2786, "lr": 1.8404168505856412e-06}
{"train_loss": 0.06462646275758743, "global_step": 247977, "epoch": 2786, "lr": 1.8402610115869634e-06}
{"train_loss": 0.02513740211725235, "global_step": 247978, "epoch": 2786, "lr": 1.8401051790627732e-06}
{"train_loss": 0.06738349795341492, "global_step": 247979, "epoch": 2786, "lr": 1.839949353013104e-06}
{"train_loss": 0.06687019020318985, "global_step": 247980, "epoch": 2786, "lr": 1.8397935334379557e-06}
{"train_loss": 0.054586153477430344, "global_step": 247981, "epoch": 2786, "lr": 1.839637720337367e-06}
{"train_loss": 0.05563099682331085, "global_step": 247982, "epoch": 2786, "lr": 1.8394819137113496e-06}
{"train_loss": 0.042419467121362686, "global_step": 247983, "epoch": 2786, "lr": 1.8393261135599304e-06}
{"train_loss": 0.07470528036355972, "global_step": 247984, "epoch": 2786, "lr": 1.8391703198831211e-06}
{"train_loss": 0.054802000522613525, "global_step": 247985, "epoch": 2786, "lr": 1.8390145326809494e-06}
{"train_loss": 0.05120144411921501, "global_step": 247986, "epoch": 2786, "lr": 1.8388587519534373e-06}
{"train_loss": 0.0447835810482502, "global_step": 247987, "epoch": 2786, "lr": 1.8387029777006015e-06}
{"train_loss": 0.06539815664291382, "global_step": 247988, "epoch": 2786, "lr": 1.8385472099224643e-06}
{"train_loss": 0.039235107600688934, "global_step": 247989, "epoch": 2786, "lr": 1.8383914486190533e-06}
{"train_loss": 0.04212335869669914, "global_step": 247990, "epoch": 2786, "lr": 1.8382356937903745e-06}
{"train_loss": 0.05547728016972542, "global_step": 247991, "epoch": 2786, "lr": 1.8380799454364606e-06}
{"train_loss": 0.08989246934652328, "global_step": 247992, "epoch": 2786, "lr": 1.8379242035573397e-06}
{"train_loss": 0.024494005367159843, "global_step": 247993, "epoch": 2786, "lr": 1.8377684681530116e-06}
{"train_loss": 0.05311606079339981, "global_step": 247994, "epoch": 2786, "lr": 1.8376127392235154e-06}
{"train_loss": 0.04467524588108063, "global_step": 247995, "epoch": 2786, "lr": 1.8374570167688566e-06}
{"train_loss": 0.06887537986040115, "global_step": 247996, "epoch": 2786, "lr": 1.837301300789074e-06}
{"train_loss": 0.03445518761873245, "global_step": 247997, "epoch": 2786, "lr": 1.8371455912841728e-06}
{"train_loss": 0.03229088336229324, "global_step": 247998, "epoch": 2786, "lr": 1.8369898882541869e-06}
{"train_loss": 0.06985004991292953, "global_step": 247999, "epoch": 2786, "lr": 1.8368341916991217e-06}
{"train_loss": 0.03467100113630295, "global_step": 248000, "epoch": 2786, "lr": 1.8366785016190159e-06}
{"train_loss": 0.07448731362819672, "global_step": 248001, "epoch": 2786, "lr": 1.836522818013875e-06}
{"train_loss": 0.03724915161728859, "global_step": 248002, "epoch": 2786, "lr": 1.8363671408837325e-06}
{"train_loss": 0.03741123899817467, "global_step": 248003, "epoch": 2786, "lr": 1.8362114702285994e-06}
{"train_loss": 0.05974852666258812, "global_step": 248004, "epoch": 2786, "lr": 1.8360558060485033e-06}
{"train_loss": 0.03333278372883797, "global_step": 248005, "epoch": 2786, "lr": 1.8359001483434668e-06}
{"train_loss": 0.0332377664744854, "global_step": 248006, "epoch": 2786, "lr": 1.8357444971134952e-06}
{"train_loss": 0.056352436542510986, "global_step": 248007, "epoch": 2786, "lr": 1.8355888523586272e-06}
{"train_loss": 0.04480117931962013, "global_step": 248008, "epoch": 2786, "lr": 1.8354332140788745e-06}
{"train_loss": 0.041605088859796524, "global_step": 248009, "epoch": 2786, "lr": 1.835277582274264e-06}
{"train_loss": 0.03642717003822327, "global_step": 248010, "epoch": 2786, "lr": 1.835121956944813e-06}
{"train_loss": 0.02421915903687477, "global_step": 248011, "epoch": 2786, "lr": 1.8349663380905435e-06}
{"train_loss": 0.06127255782485008, "global_step": 248012, "epoch": 2786, "lr": 1.8348107257114721e-06}
{"train_loss": 0.018044300377368927, "global_step": 248013, "epoch": 2786, "lr": 1.8346551198076268e-06}
{"train_loss": 0.057698898017406464, "global_step": 248014, "epoch": 2786, "lr": 1.8344995203790184e-06}
{"train_loss": 0.012814505025744438, "global_step": 248015, "epoch": 2786, "lr": 1.8343439274256857e-06}
{"train_loss": 0.030884837731719017, "global_step": 248016, "epoch": 2786, "lr": 1.8341883409476291e-06}
{"train_loss": 0.055887412279844284, "global_step": 248017, "epoch": 2786, "lr": 1.8340327609448815e-06}
{"train_loss": 0.05945301800966263, "global_step": 248018, "epoch": 2786, "lr": 1.8338771874174598e-06}
{"train_loss": 0.037719301879405975, "global_step": 248019, "epoch": 2786, "lr": 1.8337216203653918e-06}
{"train_loss": 0.02088642679154873, "global_step": 248020, "epoch": 2786, "lr": 1.8335660597886829e-06}
{"train_loss": 0.08841873705387115, "global_step": 248021, "epoch": 2786, "lr": 1.8334105056873774e-06}
{"train_loss": 0.07134069502353668, "global_step": 248022, "epoch": 2786, "lr": 1.83325495806147e-06}
{"train_loss": 0.09936623275279999, "global_step": 248023, "epoch": 2786, "lr": 1.833099416911005e-06}
{"train_loss": 0.07105724513530731, "global_step": 248024, "epoch": 2786, "lr": 1.8329438822359823e-06}
{"train_loss": 0.0504264160990715, "global_step": 248025, "epoch": 2786, "lr": 1.832788354036441e-06}
{"train_loss": 0.062192343175411224, "global_step": 248026, "epoch": 2786, "lr": 1.8326328323123865e-06}
{"train_loss": 0.029318813234567642, "global_step": 248027, "epoch": 2786, "lr": 1.8324773170638521e-06}
{"train_loss": 0.03389055281877518, "global_step": 248028, "epoch": 2786, "lr": 1.83232180829086e-06}
{"train_loss": 0.048372454941272736, "global_step": 248029, "epoch": 2786, "lr": 1.8321663059934158e-06}
{"train_loss": 0.040772732347249985, "global_step": 248030, "epoch": 2786, "lr": 1.8320108101715528e-06}
{"train_loss": 0.021247798576951027, "global_step": 248031, "epoch": 2786, "lr": 1.8318553208252876e-06}
{"train_loss": 0.05483122915029526, "global_step": 248032, "epoch": 2786, "lr": 1.831699837954648e-06}
{"train_loss": 0.08897936344146729, "global_step": 248033, "epoch": 2786, "lr": 1.831544361559645e-06}
{"train_loss": 0.07340490818023682, "global_step": 248034, "epoch": 2786, "lr": 1.8313888916403065e-06}
{"train_loss": 0.05917954817414284, "global_step": 248035, "epoch": 2786, "lr": 1.8312334281966436e-06}
{"train_loss": 0.017639871686697006, "global_step": 248036, "epoch": 2786, "lr": 1.8310779712286953e-06}
{"train_loss": 0.05077133700251579, "global_step": 248037, "epoch": 2786, "lr": 1.8309225207364611e-06}
{"train_loss": 0.05708741396665573, "global_step": 248038, "epoch": 2786, "lr": 1.8307670767199802e-06}
{"train_loss": 0.09087696671485901, "global_step": 248039, "epoch": 2786, "lr": 1.8306116391792583e-06}
{"train_loss": 0.02585752308368683, "global_step": 248040, "epoch": 2786, "lr": 1.8304562081143338e-06}
{"train_loss": 0.0628555566072464, "global_step": 248041, "epoch": 2786, "lr": 1.8303007835252073e-06}
{"train_loss": 0.04880154066917936, "global_step": 248042, "epoch": 2786, "lr": 1.8301453654119172e-06, "val_loss": 9.563867568969727}
{"train_loss": 0.032529909163713455, "global_step": 248043, "epoch": 2787, "lr": 1.8299899537744748e-06}
{"train_loss": 0.032669126987457275, "global_step": 248044, "epoch": 2787, "lr": 1.8298345486128965e-06}
{"train_loss": 0.10161871463060379, "global_step": 248045, "epoch": 2787, "lr": 1.829679149927216e-06}
{"train_loss": 0.031120503321290016, "global_step": 248046, "epoch": 2787, "lr": 1.829523757717444e-06}
{"train_loss": 0.06403469294309616, "global_step": 248047, "epoch": 2787, "lr": 1.8293683719836142e-06}
{"train_loss": 0.03600114956498146, "global_step": 248048, "epoch": 2787, "lr": 1.8292129927257318e-06}
{"train_loss": 0.04246307536959648, "global_step": 248049, "epoch": 2787, "lr": 1.8290576199438248e-06}
{"train_loss": 0.030314065515995026, "global_step": 248050, "epoch": 2787, "lr": 1.8289022536379153e-06}
{"train_loss": 0.04198363423347473, "global_step": 248051, "epoch": 2787, "lr": 1.8287468938080198e-06}
{"train_loss": 0.05369812995195389, "global_step": 248052, "epoch": 2787, "lr": 1.8285915404541665e-06}
{"train_loss": 0.019614990800619125, "global_step": 248053, "epoch": 2787, "lr": 1.8284361935763717e-06}
{"train_loss": 0.04492964968085289, "global_step": 248054, "epoch": 2787, "lr": 1.828280853174652e-06}
{"train_loss": 0.05913320183753967, "global_step": 248055, "epoch": 2787, "lr": 1.8281255192490354e-06}
{"train_loss": 0.04000624269247055, "global_step": 248056, "epoch": 2787, "lr": 1.8279701917995384e-06}
{"train_loss": 0.04389830678701401, "global_step": 248057, "epoch": 2787, "lr": 1.8278148708261888e-06}
{"train_loss": 0.021512752398848534, "global_step": 248058, "epoch": 2787, "lr": 1.8276595563289977e-06}
{"train_loss": 0.03221074491739273, "global_step": 248059, "epoch": 2787, "lr": 1.8275042483079929e-06}
{"train_loss": 0.06020766869187355, "global_step": 248060, "epoch": 2787, "lr": 1.827348946763191e-06}
{"train_loss": 0.04420600086450577, "global_step": 248061, "epoch": 2787, "lr": 1.8271936516946197e-06}
{"train_loss": 0.030735908076167107, "global_step": 248062, "epoch": 2787, "lr": 1.8270383631022847e-06}
{"train_loss": 0.014912160113453865, "global_step": 248063, "epoch": 2787, "lr": 1.8268830809862247e-06}
{"train_loss": 0.05775921046733856, "global_step": 248064, "epoch": 2787, "lr": 1.8267278053464566e-06}
{"train_loss": 0.07331159710884094, "global_step": 248065, "epoch": 2787, "lr": 1.8265725361829855e-06}
{"train_loss": 0.028810152783989906, "global_step": 248066, "epoch": 2787, "lr": 1.8264172734958561e-06}
{"train_loss": 0.044581253081560135, "global_step": 248067, "epoch": 2787, "lr": 1.8262620172850742e-06}
{"train_loss": 0.0551522932946682, "global_step": 248068, "epoch": 2787, "lr": 1.8261067675506672e-06}
{"train_loss": 0.07633203268051147, "global_step": 248069, "epoch": 2787, "lr": 1.8259515242926462e-06}
{"train_loss": 0.03028251603245735, "global_step": 248070, "epoch": 2787, "lr": 1.8257962875110446e-06}
{"train_loss": 0.059507809579372406, "global_step": 248071, "epoch": 2787, "lr": 1.8256410572058736e-06}
{"train_loss": 0.014369978569447994, "global_step": 248072, "epoch": 2787, "lr": 1.8254858333771663e-06}
{"train_loss": 0.05610346794128418, "global_step": 248073, "epoch": 2787, "lr": 1.825330616024923e-06}
{"train_loss": 0.0544380322098732, "global_step": 248074, "epoch": 2787, "lr": 1.8251754051491876e-06}
{"train_loss": 0.04217006638646126, "global_step": 248075, "epoch": 2787, "lr": 1.8250202007499605e-06}
{"train_loss": 0.04126835614442825, "global_step": 248076, "epoch": 2787, "lr": 1.8248650028272808e-06}
{"train_loss": 0.08417332172393799, "global_step": 248077, "epoch": 2787, "lr": 1.8247098113811534e-06}
{"train_loss": 0.05093206465244293, "global_step": 248078, "epoch": 2787, "lr": 1.824554626411612e-06}
{"train_loss": 0.05416503921151161, "global_step": 248079, "epoch": 2787, "lr": 1.8243994479186676e-06}
{"train_loss": 0.06835252791643143, "global_step": 248080, "epoch": 2787, "lr": 1.8242442759023538e-06}
{"train_loss": 0.06322921812534332, "global_step": 248081, "epoch": 2787, "lr": 1.8240891103626755e-06}
{"train_loss": 0.03601755574345589, "global_step": 248082, "epoch": 2787, "lr": 1.8239339512996667e-06}
{"train_loss": 0.05613759905099869, "global_step": 248083, "epoch": 2787, "lr": 1.8237787987133436e-06}
{"train_loss": 0.08605615794658661, "global_step": 248084, "epoch": 2787, "lr": 1.8236236526037177e-06}
{"train_loss": 0.038147274404764175, "global_step": 248085, "epoch": 2787, "lr": 1.8234685129708274e-06}
{"train_loss": 0.09006549417972565, "global_step": 248086, "epoch": 2787, "lr": 1.8233133798146785e-06}
{"train_loss": 0.02412370592355728, "global_step": 248087, "epoch": 2787, "lr": 1.8231582531352987e-06}
{"train_loss": 0.04891900718212128, "global_step": 248088, "epoch": 2787, "lr": 1.8230031329327047e-06}
{"train_loss": 0.03455394133925438, "global_step": 248089, "epoch": 2787, "lr": 1.8228480192069297e-06}
{"train_loss": 0.08411802351474762, "global_step": 248090, "epoch": 2787, "lr": 1.8226929119579794e-06}
{"train_loss": 0.04521951079368591, "global_step": 248091, "epoch": 2787, "lr": 1.822537811185887e-06}
{"train_loss": 0.04921451210975647, "global_step": 248092, "epoch": 2787, "lr": 1.8223827168906582e-06}
{"train_loss": 0.029076872393488884, "global_step": 248093, "epoch": 2787, "lr": 1.8222276290723316e-06}
{"train_loss": 0.046186283230781555, "global_step": 248094, "epoch": 2787, "lr": 1.8220725477309074e-06}
{"train_loss": 0.04240184649825096, "global_step": 248095, "epoch": 2787, "lr": 1.8219174728664301e-06}
{"train_loss": 0.026885729283094406, "global_step": 248096, "epoch": 2787, "lr": 1.821762404478905e-06}
{"train_loss": 0.1344602406024933, "global_step": 248097, "epoch": 2787, "lr": 1.8216073425683545e-06}
{"train_loss": 0.07296346127986908, "global_step": 248098, "epoch": 2787, "lr": 1.8214522871348005e-06}
{"train_loss": 0.01905057579278946, "global_step": 248099, "epoch": 2787, "lr": 1.8212972381782657e-06}
{"train_loss": 0.0812947228550911, "global_step": 248100, "epoch": 2787, "lr": 1.8211421956987773e-06}
{"train_loss": 0.07744929194450378, "global_step": 248101, "epoch": 2787, "lr": 1.8209871596963413e-06}
{"train_loss": 0.04852299019694328, "global_step": 248102, "epoch": 2787, "lr": 1.8208321301709908e-06}
{"train_loss": 0.026161938905715942, "global_step": 248103, "epoch": 2787, "lr": 1.8206771071227368e-06}
{"train_loss": 0.03803260251879692, "global_step": 248104, "epoch": 2787, "lr": 1.820522090551613e-06}
{"train_loss": 0.04740078002214432, "global_step": 248105, "epoch": 2787, "lr": 1.82036708045763e-06}
{"train_loss": 0.04285810887813568, "global_step": 248106, "epoch": 2787, "lr": 1.8202120768408104e-06}
{"train_loss": 0.04743199795484543, "global_step": 248107, "epoch": 2787, "lr": 1.8200570797011762e-06}
{"train_loss": 0.03728073835372925, "global_step": 248108, "epoch": 2787, "lr": 1.8199020890387496e-06}
{"train_loss": 0.07593967765569687, "global_step": 248109, "epoch": 2787, "lr": 1.8197471048535476e-06}
{"train_loss": 0.06773708760738373, "global_step": 248110, "epoch": 2787, "lr": 1.8195921271455974e-06}
{"train_loss": 0.06328688561916351, "global_step": 248111, "epoch": 2787, "lr": 1.8194371559149105e-06}
{"train_loss": 0.017529787495732307, "global_step": 248112, "epoch": 2787, "lr": 1.8192821911615143e-06}
{"train_loss": 0.01523054763674736, "global_step": 248113, "epoch": 2787, "lr": 1.819127232885426e-06}
{"train_loss": 0.027225444093346596, "global_step": 248114, "epoch": 2787, "lr": 1.8189722810866784e-06}
{"train_loss": 0.02890850231051445, "global_step": 248115, "epoch": 2787, "lr": 1.8188173357652716e-06}
{"train_loss": 0.09466314315795898, "global_step": 248116, "epoch": 2787, "lr": 1.8186623969212447e-06}
{"train_loss": 0.0925869420170784, "global_step": 248117, "epoch": 2787, "lr": 1.8185074645546086e-06}
{"train_loss": 0.040716346353292465, "global_step": 248118, "epoch": 2787, "lr": 1.818352538665391e-06}
{"train_loss": 0.13971583545207977, "global_step": 248119, "epoch": 2787, "lr": 1.8181976192535976e-06}
{"train_loss": 0.048640988767147064, "global_step": 248120, "epoch": 2787, "lr": 1.8180427063192729e-06}
{"train_loss": 0.08405598998069763, "global_step": 248121, "epoch": 2787, "lr": 1.8178877998624222e-06}
{"train_loss": 0.03819388151168823, "global_step": 248122, "epoch": 2787, "lr": 1.8177328998830678e-06}
{"train_loss": 0.07291562110185623, "global_step": 248123, "epoch": 2787, "lr": 1.817578006381232e-06}
{"train_loss": 0.014614141546189785, "global_step": 248124, "epoch": 2787, "lr": 1.8174231193569313e-06}
{"train_loss": 0.0582403689622879, "global_step": 248125, "epoch": 2787, "lr": 1.8172682388101992e-06}
{"train_loss": 0.02305910922586918, "global_step": 248126, "epoch": 2787, "lr": 1.817113364741041e-06}
{"train_loss": 0.06199291720986366, "global_step": 248127, "epoch": 2787, "lr": 1.8169584971494846e-06}
{"train_loss": 0.06431709975004196, "global_step": 248128, "epoch": 2787, "lr": 1.8168036360355523e-06}
{"train_loss": 0.06526267528533936, "global_step": 248129, "epoch": 2787, "lr": 1.816648781399266e-06}
{"train_loss": 0.03676168620586395, "global_step": 248130, "epoch": 2787, "lr": 1.8164939332406427e-06}
{"train_loss": 0.05053839300957958, "global_step": 248131, "epoch": 2787, "lr": 1.8163390915597045e-06, "val_loss": 9.438483238220215}
{"train_loss": 0.07357645779848099, "global_step": 248132, "epoch": 2788, "lr": 1.816184256356468e-06}
{"train_loss": 0.03979131951928139, "global_step": 248133, "epoch": 2788, "lr": 1.8160294276309665e-06}
{"train_loss": 0.07123047113418579, "global_step": 248134, "epoch": 2788, "lr": 1.8158746053832055e-06}
{"train_loss": 0.07994502782821655, "global_step": 248135, "epoch": 2788, "lr": 1.8157197896132183e-06}
{"train_loss": 0.024802573025226593, "global_step": 248136, "epoch": 2788, "lr": 1.8155649803210162e-06}
{"train_loss": 0.03895149007439613, "global_step": 248137, "epoch": 2788, "lr": 1.8154101775066212e-06}
{"train_loss": 0.022853566333651543, "global_step": 248138, "epoch": 2788, "lr": 1.8152553811700612e-06}
{"train_loss": 0.04508397728204727, "global_step": 248139, "epoch": 2788, "lr": 1.8151005913113527e-06}
{"train_loss": 0.05151054635643959, "global_step": 248140, "epoch": 2788, "lr": 1.8149458079305182e-06}
{"train_loss": 0.046766817569732666, "global_step": 248141, "epoch": 2788, "lr": 1.814791031027574e-06}
{"train_loss": 0.007576265372335911, "global_step": 248142, "epoch": 2788, "lr": 1.8146362606025479e-06}
{"train_loss": 0.0442839115858078, "global_step": 248143, "epoch": 2788, "lr": 1.8144814966554514e-06}
{"train_loss": 0.04922837018966675, "global_step": 248144, "epoch": 2788, "lr": 1.8143267391863172e-06}
{"train_loss": 0.0673983097076416, "global_step": 248145, "epoch": 2788, "lr": 1.8141719881951513e-06}
{"train_loss": 0.05354296416044235, "global_step": 248146, "epoch": 2788, "lr": 1.8140172436819925e-06}
{"train_loss": 0.03718499839305878, "global_step": 248147, "epoch": 2788, "lr": 1.8138625056468405e-06}
{"train_loss": 0.02960823103785515, "global_step": 248148, "epoch": 2788, "lr": 1.8137077740897346e-06}
{"train_loss": 0.05611705034971237, "global_step": 248149, "epoch": 2788, "lr": 1.8135530490106857e-06}
{"train_loss": 0.02813386358320713, "global_step": 248150, "epoch": 2788, "lr": 1.8133983304097212e-06}
{"train_loss": 0.027147850021719933, "global_step": 248151, "epoch": 2788, "lr": 1.8132436182868529e-06}
{"train_loss": 0.04737531393766403, "global_step": 248152, "epoch": 2788, "lr": 1.8130889126421135e-06}
{"train_loss": 0.07121734321117401, "global_step": 248153, "epoch": 2788, "lr": 1.812934213475509e-06}
{"train_loss": 0.01918015256524086, "global_step": 248154, "epoch": 2788, "lr": 1.8127795207870724e-06}
{"train_loss": 0.04472930356860161, "global_step": 248155, "epoch": 2788, "lr": 1.8126248345768204e-06}
{"train_loss": 0.060228098183870316, "global_step": 248156, "epoch": 2788, "lr": 1.8124701548447753e-06}
{"train_loss": 0.013467186130583286, "global_step": 248157, "epoch": 2788, "lr": 1.8123154815909481e-06}
{"train_loss": 0.07296523451805115, "global_step": 248158, "epoch": 2788, "lr": 1.812160814815378e-06}
{"train_loss": 0.047261692583560944, "global_step": 248159, "epoch": 2788, "lr": 1.8120061545180755e-06}
{"train_loss": 0.03376729413866997, "global_step": 248160, "epoch": 2788, "lr": 1.8118515006990523e-06}
{"train_loss": 0.040767304599285126, "global_step": 248161, "epoch": 2788, "lr": 1.8116968533583467e-06}
{"train_loss": 0.025227868929505348, "global_step": 248162, "epoch": 2788, "lr": 1.8115422124959646e-06}
{"train_loss": 0.035292815417051315, "global_step": 248163, "epoch": 2788, "lr": 1.8113875781119394e-06}
{"train_loss": 0.04438605532050133, "global_step": 248164, "epoch": 2788, "lr": 1.8112329502062765e-06}
{"train_loss": 0.021300464868545532, "global_step": 248165, "epoch": 2788, "lr": 1.8110783287790146e-06}
{"train_loss": 0.06704410165548325, "global_step": 248166, "epoch": 2788, "lr": 1.8109237138301595e-06}
{"train_loss": 0.034754689782857895, "global_step": 248167, "epoch": 2788, "lr": 1.8107691053597441e-06}
{"train_loss": 0.018347065895795822, "global_step": 248168, "epoch": 2788, "lr": 1.81061450336778e-06}
{"train_loss": 0.03380986675620079, "global_step": 248169, "epoch": 2788, "lr": 1.810459907854295e-06}
{"train_loss": 0.023631837218999863, "global_step": 248170, "epoch": 2788, "lr": 1.8103053188193054e-06}
{"train_loss": 0.04392962530255318, "global_step": 248171, "epoch": 2788, "lr": 1.8101507362628333e-06}
{"train_loss": 0.04220966249704361, "global_step": 248172, "epoch": 2788, "lr": 1.8099961601848957e-06}
{"train_loss": 0.053868960589170456, "global_step": 248173, "epoch": 2788, "lr": 1.8098415905855148e-06}
{"train_loss": 0.02140526846051216, "global_step": 248174, "epoch": 2788, "lr": 1.8096870274647181e-06}
{"train_loss": 0.08226537704467773, "global_step": 248175, "epoch": 2788, "lr": 1.8095324708225169e-06}
{"train_loss": 0.02331106923520565, "global_step": 248176, "epoch": 2788, "lr": 1.8093779206589446e-06}
{"train_loss": 0.02927485853433609, "global_step": 248177, "epoch": 2788, "lr": 1.8092233769740064e-06}
{"train_loss": 0.03451908007264137, "global_step": 248178, "epoch": 2788, "lr": 1.8090688397677357e-06}
{"train_loss": 0.03516769036650658, "global_step": 248179, "epoch": 2788, "lr": 1.8089143090401494e-06}
{"train_loss": 0.041279591619968414, "global_step": 248180, "epoch": 2788, "lr": 1.808759784791264e-06}
{"train_loss": 0.03902740031480789, "global_step": 248181, "epoch": 2788, "lr": 1.8086052670211018e-06}
{"train_loss": 0.04470311850309372, "global_step": 248182, "epoch": 2788, "lr": 1.8084507557296904e-06}
{"train_loss": 0.03336990997195244, "global_step": 248183, "epoch": 2788, "lr": 1.808296250917041e-06}
{"train_loss": 0.028845157474279404, "global_step": 248184, "epoch": 2788, "lr": 1.8081417525831813e-06}
{"train_loss": 0.01248509157449007, "global_step": 248185, "epoch": 2788, "lr": 1.8079872607281278e-06}
{"train_loss": 0.04769091680645943, "global_step": 248186, "epoch": 2788, "lr": 1.8078327753519086e-06}
{"train_loss": 0.03943682461977005, "global_step": 248187, "epoch": 2788, "lr": 1.8076782964545292e-06}
{"train_loss": 0.027889076620340347, "global_step": 248188, "epoch": 2788, "lr": 1.8075238240360282e-06}
{"train_loss": 0.045804671943187714, "global_step": 248189, "epoch": 2788, "lr": 1.8073693580964113e-06}
{"train_loss": 0.049429114907979965, "global_step": 248190, "epoch": 2788, "lr": 1.8072148986357064e-06}
{"train_loss": 0.022868135944008827, "global_step": 248191, "epoch": 2788, "lr": 1.8070604456539408e-06}
{"train_loss": 0.058315739035606384, "global_step": 248192, "epoch": 2788, "lr": 1.8069059991511262e-06}
{"train_loss": 0.028541428968310356, "global_step": 248193, "epoch": 2788, "lr": 1.8067515591272844e-06}
{"train_loss": 0.02415972389280796, "global_step": 248194, "epoch": 2788, "lr": 1.8065971255824432e-06}
{"train_loss": 0.019400762394070625, "global_step": 248195, "epoch": 2788, "lr": 1.8064426985166082e-06}
{"train_loss": 0.03386121615767479, "global_step": 248196, "epoch": 2788, "lr": 1.8062882779298185e-06}
{"train_loss": 0.01571127399802208, "global_step": 248197, "epoch": 2788, "lr": 1.8061338638220792e-06}
{"train_loss": 0.060352157801389694, "global_step": 248198, "epoch": 2788, "lr": 1.8059794561934185e-06}
{"train_loss": 0.037405356764793396, "global_step": 248199, "epoch": 2788, "lr": 1.8058250550438639e-06}
{"train_loss": 0.059770628809928894, "global_step": 248200, "epoch": 2788, "lr": 1.8056706603734208e-06}
{"train_loss": 0.06907356530427933, "global_step": 248201, "epoch": 2788, "lr": 1.8055162721821227e-06}
{"train_loss": 0.012447549030184746, "global_step": 248202, "epoch": 2788, "lr": 1.8053618904699753e-06}
{"train_loss": 0.0634663999080658, "global_step": 248203, "epoch": 2788, "lr": 1.8052075152370228e-06}
{"train_loss": 0.05288856849074364, "global_step": 248204, "epoch": 2788, "lr": 1.8050531464832655e-06}
{"train_loss": 0.06946078687906265, "global_step": 248205, "epoch": 2788, "lr": 1.8048987842087306e-06}
{"train_loss": 0.055117473006248474, "global_step": 248206, "epoch": 2788, "lr": 1.8047444284134407e-06}
{"train_loss": 0.017808301374316216, "global_step": 248207, "epoch": 2788, "lr": 1.8045900790974179e-06}
{"train_loss": 0.06577254086732864, "global_step": 248208, "epoch": 2788, "lr": 1.804435736260679e-06}
{"train_loss": 0.07978174090385437, "global_step": 248209, "epoch": 2788, "lr": 1.8042813999032404e-06}
{"train_loss": 0.03806574270129204, "global_step": 248210, "epoch": 2788, "lr": 1.8041270700251412e-06}
{"train_loss": 0.08291108906269073, "global_step": 248211, "epoch": 2788, "lr": 1.8039727466263811e-06}
{"train_loss": 0.04598666727542877, "global_step": 248212, "epoch": 2788, "lr": 1.8038184297069939e-06}
{"train_loss": 0.02199879102408886, "global_step": 248213, "epoch": 2788, "lr": 1.80366411926699e-06}
{"train_loss": 0.019892094656825066, "global_step": 248214, "epoch": 2788, "lr": 1.8035098153064035e-06}
{"train_loss": 0.04276783764362335, "global_step": 248215, "epoch": 2788, "lr": 1.8033555178252393e-06}
{"train_loss": 0.02094988152384758, "global_step": 248216, "epoch": 2788, "lr": 1.803201226823531e-06}
{"train_loss": 0.07670728117227554, "global_step": 248217, "epoch": 2788, "lr": 1.8030469423012953e-06}
{"train_loss": 0.04029852896928787, "global_step": 248218, "epoch": 2788, "lr": 1.8028926642585543e-06}
{"train_loss": 0.08717558532953262, "global_step": 248219, "epoch": 2788, "lr": 1.8027383926953245e-06}
{"train_loss": 0.04286268446594477, "global_step": 248220, "epoch": 2788, "lr": 1.8025841276116283e-06, "val_loss": 9.531885147094727}
{"train_loss": 0.040303632616996765, "global_step": 248221, "epoch": 2789, "lr": 1.8024298690074881e-06}
{"train_loss": 0.0319916270673275, "global_step": 248222, "epoch": 2789, "lr": 1.8022756168829257e-06}
{"train_loss": 0.03561592847108841, "global_step": 248223, "epoch": 2789, "lr": 1.8021213712379525e-06}
{"train_loss": 0.09003306180238724, "global_step": 248224, "epoch": 2789, "lr": 1.8019671320726073e-06}
{"train_loss": 0.033118829131126404, "global_step": 248225, "epoch": 2789, "lr": 1.8018128993868899e-06}
{"train_loss": 0.011301673017442226, "global_step": 248226, "epoch": 2789, "lr": 1.8016586731808339e-06}
{"train_loss": 0.05246788635849953, "global_step": 248227, "epoch": 2789, "lr": 1.8015044534544666e-06}
{"train_loss": 0.031045490875840187, "global_step": 248228, "epoch": 2789, "lr": 1.8013502402077886e-06}
{"train_loss": 0.07186587154865265, "global_step": 248229, "epoch": 2789, "lr": 1.8011960334408384e-06}
{"train_loss": 0.061926741153001785, "global_step": 248230, "epoch": 2789, "lr": 1.8010418331536272e-06}
{"train_loss": 0.035876717418432236, "global_step": 248231, "epoch": 2789, "lr": 1.800887639346177e-06}
{"train_loss": 0.03864867612719536, "global_step": 248232, "epoch": 2789, "lr": 1.800733452018516e-06}
{"train_loss": 0.07808543741703033, "global_step": 248233, "epoch": 2789, "lr": 1.8005792711706547e-06}
{"train_loss": 0.08652760833501816, "global_step": 248234, "epoch": 2789, "lr": 1.800425096802616e-06}
{"train_loss": 0.034775421023368835, "global_step": 248235, "epoch": 2789, "lr": 1.800270928914427e-06}
{"train_loss": 0.06147925555706024, "global_step": 248236, "epoch": 2789, "lr": 1.8001167675061048e-06}
{"train_loss": 0.03360431641340256, "global_step": 248237, "epoch": 2789, "lr": 1.799962612577666e-06}
{"train_loss": 0.04318001866340637, "global_step": 248238, "epoch": 2789, "lr": 1.7998084641291324e-06}
{"train_loss": 0.026403039693832397, "global_step": 248239, "epoch": 2789, "lr": 1.7996543221605322e-06}
{"train_loss": 0.04826776310801506, "global_step": 248240, "epoch": 2789, "lr": 1.7995001866718764e-06}
{"train_loss": 0.06279677152633667, "global_step": 248241, "epoch": 2789, "lr": 1.7993460576631982e-06}
{"train_loss": 0.01115466095507145, "global_step": 248242, "epoch": 2789, "lr": 1.7991919351344978e-06}
{"train_loss": 0.06659185141324997, "global_step": 248243, "epoch": 2789, "lr": 1.7990378190858192e-06}
{"train_loss": 0.050200991332530975, "global_step": 248244, "epoch": 2789, "lr": 1.7988837095171684e-06}
{"train_loss": 0.04011490195989609, "global_step": 248245, "epoch": 2789, "lr": 1.7987296064285674e-06}
{"train_loss": 0.045504286885261536, "global_step": 248246, "epoch": 2789, "lr": 1.7985755098200442e-06}
{"train_loss": 0.06932337582111359, "global_step": 248247, "epoch": 2789, "lr": 1.7984214196916093e-06}
{"train_loss": 0.055922992527484894, "global_step": 248248, "epoch": 2789, "lr": 1.7982673360432966e-06}
{"train_loss": 0.06673771142959595, "global_step": 248249, "epoch": 2789, "lr": 1.798113258875117e-06}
{"train_loss": 0.03362491354346275, "global_step": 248250, "epoch": 2789, "lr": 1.7979591881870928e-06}
{"train_loss": 0.03884515538811684, "global_step": 248251, "epoch": 2789, "lr": 1.797805123979246e-06}
{"train_loss": 0.053915396332740784, "global_step": 248252, "epoch": 2789, "lr": 1.7976510662515988e-06}
{"train_loss": 0.02209223061800003, "global_step": 248253, "epoch": 2789, "lr": 1.7974970150041626e-06}
{"train_loss": 0.02410551905632019, "global_step": 248254, "epoch": 2789, "lr": 1.797342970236976e-06}
{"train_loss": 0.06135427579283714, "global_step": 248255, "epoch": 2789, "lr": 1.797188931950039e-06}
{"train_loss": 0.05364156886935234, "global_step": 248256, "epoch": 2789, "lr": 1.7970349001433907e-06}
{"train_loss": 0.07780126482248306, "global_step": 248257, "epoch": 2789, "lr": 1.7968808748170363e-06}
{"train_loss": 0.06094755604863167, "global_step": 248258, "epoch": 2789, "lr": 1.7967268559710093e-06}
{"train_loss": 0.020231353119015694, "global_step": 248259, "epoch": 2789, "lr": 1.7965728436053208e-06}
{"train_loss": 0.02058417908847332, "global_step": 248260, "epoch": 2789, "lr": 1.7964188377199987e-06}
{"train_loss": 0.05985115095973015, "global_step": 248261, "epoch": 2789, "lr": 1.7962648383150594e-06}
{"train_loss": 0.05683670565485954, "global_step": 248262, "epoch": 2789, "lr": 1.796110845390525e-06}
{"train_loss": 0.042312365025281906, "global_step": 248263, "epoch": 2789, "lr": 1.7959568589464126e-06}
{"train_loss": 0.034518901258707047, "global_step": 248264, "epoch": 2789, "lr": 1.7958028789827497e-06}
{"train_loss": 0.04094596207141876, "global_step": 248265, "epoch": 2789, "lr": 1.795648905499553e-06}
{"train_loss": 0.06339143216609955, "global_step": 248266, "epoch": 2789, "lr": 1.7954949384968444e-06}
{"train_loss": 0.030030064284801483, "global_step": 248267, "epoch": 2789, "lr": 1.7953409779746466e-06}
{"train_loss": 0.050552427768707275, "global_step": 248268, "epoch": 2789, "lr": 1.7951870239329704e-06}
{"train_loss": 0.03266606107354164, "global_step": 248269, "epoch": 2789, "lr": 1.7950330763718493e-06}
{"train_loss": 0.024512123316526413, "global_step": 248270, "epoch": 2789, "lr": 1.7948791352912997e-06}
{"train_loss": 0.08080298453569412, "global_step": 248271, "epoch": 2789, "lr": 1.7947252006913384e-06}
{"train_loss": 0.014788331463932991, "global_step": 248272, "epoch": 2789, "lr": 1.7945712725719932e-06}
{"train_loss": 0.04791880026459694, "global_step": 248273, "epoch": 2789, "lr": 1.7944173509332695e-06}
{"train_loss": 0.052369676530361176, "global_step": 248274, "epoch": 2789, "lr": 1.794263435775212e-06}
{"train_loss": 0.04079349711537361, "global_step": 248275, "epoch": 2789, "lr": 1.7941095270978148e-06}
{"train_loss": 0.05114259198307991, "global_step": 248276, "epoch": 2789, "lr": 1.7939556249011224e-06}
{"train_loss": 0.04729902744293213, "global_step": 248277, "epoch": 2789, "lr": 1.793801729185146e-06}
{"train_loss": 0.07988633215427399, "global_step": 248278, "epoch": 2789, "lr": 1.793647839949897e-06}
{"train_loss": 0.031794026494026184, "global_step": 248279, "epoch": 2789, "lr": 1.7934939571954135e-06}
{"train_loss": 0.03918419033288956, "global_step": 248280, "epoch": 2789, "lr": 1.7933400809216961e-06}
{"train_loss": 0.0488492026925087, "global_step": 248281, "epoch": 2789, "lr": 1.7931862111287834e-06}
{"train_loss": 0.05296457186341286, "global_step": 248282, "epoch": 2789, "lr": 1.7930323478166922e-06}
{"train_loss": 0.026205847039818764, "global_step": 248283, "epoch": 2789, "lr": 1.7928784909854334e-06}
{"train_loss": 0.035662975162267685, "global_step": 248284, "epoch": 2789, "lr": 1.7927246406350407e-06}
{"train_loss": 0.031083544716238976, "global_step": 248285, "epoch": 2789, "lr": 1.792570796765519e-06}
{"train_loss": 0.04322919622063637, "global_step": 248286, "epoch": 2789, "lr": 1.792416959376908e-06}
{"train_loss": 0.026774350553750992, "global_step": 248287, "epoch": 2789, "lr": 1.7922631284692182e-06}
{"train_loss": 0.05254897102713585, "global_step": 248288, "epoch": 2789, "lr": 1.792109304042472e-06}
{"train_loss": 0.09466090798377991, "global_step": 248289, "epoch": 2789, "lr": 1.7919554860966804e-06}
{"train_loss": 0.05498925969004631, "global_step": 248290, "epoch": 2789, "lr": 1.7918016746318822e-06}
{"train_loss": 0.03404000401496887, "global_step": 248291, "epoch": 2789, "lr": 1.7916478696480832e-06}
{"train_loss": 0.08841659128665924, "global_step": 248292, "epoch": 2789, "lr": 1.791494071145311e-06}
{"train_loss": 0.050673048943281174, "global_step": 248293, "epoch": 2789, "lr": 1.7913402791235823e-06}
{"train_loss": 0.05676950141787529, "global_step": 248294, "epoch": 2789, "lr": 1.7911864935829303e-06}
{"train_loss": 0.03524099290370941, "global_step": 248295, "epoch": 2789, "lr": 1.7910327145233552e-06}
{"train_loss": 0.062207307666540146, "global_step": 248296, "epoch": 2789, "lr": 1.7908789419448958e-06}
{"train_loss": 0.06432150304317474, "global_step": 248297, "epoch": 2789, "lr": 1.7907251758475575e-06}
{"train_loss": 0.05546127259731293, "global_step": 248298, "epoch": 2789, "lr": 1.7905714162313735e-06}
{"train_loss": 0.028546057641506195, "global_step": 248299, "epoch": 2789, "lr": 1.7904176630963554e-06}
{"train_loss": 0.07710443437099457, "global_step": 248300, "epoch": 2789, "lr": 1.790263916442525e-06}
{"train_loss": 0.04862909018993378, "global_step": 248301, "epoch": 2789, "lr": 1.7901101762699158e-06}
{"train_loss": 0.06325513869524002, "global_step": 248302, "epoch": 2789, "lr": 1.7899564425785332e-06}
{"train_loss": 0.05453411117196083, "global_step": 248303, "epoch": 2789, "lr": 1.7898027153684104e-06}
{"train_loss": 0.05292036756873131, "global_step": 248304, "epoch": 2789, "lr": 1.7896489946395478e-06}
{"train_loss": 0.03319111466407776, "global_step": 248305, "epoch": 2789, "lr": 1.789495280391995e-06}
{"train_loss": 0.037110336124897, "global_step": 248306, "epoch": 2789, "lr": 1.7893415726257411e-06}
{"train_loss": 0.04618620499968529, "global_step": 248307, "epoch": 2789, "lr": 1.789187871340836e-06}
{"train_loss": 0.044388625770807266, "global_step": 248308, "epoch": 2789, "lr": 1.7890341765372798e-06}
{"train_loss": 0.04764708714413174, "global_step": 248309, "epoch": 2789, "lr": 1.7888804882151056e-06, "val_loss": 9.48213005065918}
{"train_loss": 0.03739417344331741, "global_step": 248310, "epoch": 2790, "lr": 1.7887268063743246e-06}
{"train_loss": 0.036368608474731445, "global_step": 248311, "epoch": 2790, "lr": 1.788573131014959e-06}
{"train_loss": 0.047007448971271515, "global_step": 248312, "epoch": 2790, "lr": 1.788419462137042e-06}
{"train_loss": 0.057314615696668625, "global_step": 248313, "epoch": 2790, "lr": 1.7882657997405738e-06}
{"train_loss": 0.04289373755455017, "global_step": 248314, "epoch": 2790, "lr": 1.7881121438255933e-06}
{"train_loss": 0.07774785906076431, "global_step": 248315, "epoch": 2790, "lr": 1.7879584943921113e-06}
{"train_loss": 0.05828014016151428, "global_step": 248316, "epoch": 2790, "lr": 1.7878048514401446e-06}
{"train_loss": 0.016476763412356377, "global_step": 248317, "epoch": 2790, "lr": 1.7876512149697211e-06}
{"train_loss": 0.06659036129713058, "global_step": 248318, "epoch": 2790, "lr": 1.7874975849808683e-06}
{"train_loss": 0.0046978057362139225, "global_step": 248319, "epoch": 2790, "lr": 1.7873439614735921e-06}
{"train_loss": 0.03614161163568497, "global_step": 248320, "epoch": 2790, "lr": 1.7871903444479254e-06}
{"train_loss": 0.033364735543727875, "global_step": 248321, "epoch": 2790, "lr": 1.7870367339038796e-06}
{"train_loss": 0.04666641727089882, "global_step": 248322, "epoch": 2790, "lr": 1.7868831298414822e-06}
{"train_loss": 0.049759671092033386, "global_step": 248323, "epoch": 2790, "lr": 1.7867295322607446e-06}
{"train_loss": 0.07158999145030975, "global_step": 248324, "epoch": 2790, "lr": 1.7865759411617001e-06}
{"train_loss": 0.05209561809897423, "global_step": 248325, "epoch": 2790, "lr": 1.786422356544354e-06}
{"train_loss": 0.053883932530879974, "global_step": 248326, "epoch": 2790, "lr": 1.786268778408745e-06}
{"train_loss": 0.055145494639873505, "global_step": 248327, "epoch": 2790, "lr": 1.7861152067548792e-06}
{"train_loss": 0.0301673524081707, "global_step": 248328, "epoch": 2790, "lr": 1.785961641582784e-06}
{"train_loss": 0.046570565551519394, "global_step": 248329, "epoch": 2790, "lr": 1.785808082892476e-06}
{"train_loss": 0.07949286699295044, "global_step": 248330, "epoch": 2790, "lr": 1.785654530683989e-06}
{"train_loss": 0.08409944176673889, "global_step": 248331, "epoch": 2790, "lr": 1.7855009849573224e-06}
{"train_loss": 0.04345228150486946, "global_step": 248332, "epoch": 2790, "lr": 1.7853474457125096e-06}
{"train_loss": 0.08130262047052383, "global_step": 248333, "epoch": 2790, "lr": 1.7851939129495676e-06}
{"train_loss": 0.022258004173636436, "global_step": 248334, "epoch": 2790, "lr": 1.785040386668524e-06}
{"train_loss": 0.032292090356349945, "global_step": 248335, "epoch": 2790, "lr": 1.7848868668693896e-06}
{"train_loss": 0.04616789519786835, "global_step": 248336, "epoch": 2790, "lr": 1.7847333535521871e-06}
{"train_loss": 0.061880793422460556, "global_step": 248337, "epoch": 2790, "lr": 1.7845798467169494e-06}
{"train_loss": 0.03976796939969063, "global_step": 248338, "epoch": 2790, "lr": 1.7844263463636769e-06}
{"train_loss": 0.02730754390358925, "global_step": 248339, "epoch": 2790, "lr": 1.784272852492408e-06}
{"train_loss": 0.05775490403175354, "global_step": 248340, "epoch": 2790, "lr": 1.7841193651031485e-06}
{"train_loss": 0.02685411460697651, "global_step": 248341, "epoch": 2790, "lr": 1.7839658841959373e-06}
{"train_loss": 0.07063916325569153, "global_step": 248342, "epoch": 2790, "lr": 1.7838124097707741e-06}
{"train_loss": 0.015593651682138443, "global_step": 248343, "epoch": 2790, "lr": 1.7836589418276984e-06}
{"train_loss": 0.03758382797241211, "global_step": 248344, "epoch": 2790, "lr": 1.7835054803667095e-06}
{"train_loss": 0.06159676983952522, "global_step": 248345, "epoch": 2790, "lr": 1.7833520253878522e-06}
{"train_loss": 0.03193618357181549, "global_step": 248346, "epoch": 2790, "lr": 1.7831985768911319e-06}
{"train_loss": 0.0789773091673851, "global_step": 248347, "epoch": 2790, "lr": 1.7830451348765763e-06}
{"train_loss": 0.029528599232435226, "global_step": 248348, "epoch": 2790, "lr": 1.7828916993442024e-06}
{"train_loss": 0.05250938609242439, "global_step": 248349, "epoch": 2790, "lr": 1.782738270294021e-06}
{"train_loss": 0.04380922019481659, "global_step": 248350, "epoch": 2790, "lr": 1.782584847726071e-06}
{"train_loss": 0.06276822090148926, "global_step": 248351, "epoch": 2790, "lr": 1.7824314316403633e-06}
{"train_loss": 0.024260137230157852, "global_step": 248352, "epoch": 2790, "lr": 1.7822780220369207e-06}
{"train_loss": 0.0649167001247406, "global_step": 248353, "epoch": 2790, "lr": 1.7821246189157593e-06}
{"train_loss": 0.022610096260905266, "global_step": 248354, "epoch": 2790, "lr": 1.7819712222769125e-06}
{"train_loss": 0.06088457256555557, "global_step": 248355, "epoch": 2790, "lr": 1.7818178321203805e-06}
{"train_loss": 0.020676324144005775, "global_step": 248356, "epoch": 2790, "lr": 1.7816644484462074e-06}
{"train_loss": 0.044044945389032364, "global_step": 248357, "epoch": 2790, "lr": 1.7815110712543882e-06}
{"train_loss": 0.07081513106822968, "global_step": 248358, "epoch": 2790, "lr": 1.7813577005449667e-06}
{"train_loss": 0.020406272262334824, "global_step": 248359, "epoch": 2790, "lr": 1.7812043363179487e-06}
{"train_loss": 0.028194524347782135, "global_step": 248360, "epoch": 2790, "lr": 1.7810509785733676e-06}
{"train_loss": 0.033075615763664246, "global_step": 248361, "epoch": 2790, "lr": 1.7808976273112288e-06}
{"train_loss": 0.05263878032565117, "global_step": 248362, "epoch": 2790, "lr": 1.7807442825315658e-06}
{"train_loss": 0.029242226853966713, "global_step": 248363, "epoch": 2790, "lr": 1.7805909442343894e-06}
{"train_loss": 0.0347900353372097, "global_step": 248364, "epoch": 2790, "lr": 1.780437612419733e-06}
{"train_loss": 0.014960097149014473, "global_step": 248365, "epoch": 2790, "lr": 1.780284287087597e-06}
{"train_loss": 0.03972576931118965, "global_step": 248366, "epoch": 2790, "lr": 1.7801309682380252e-06}
{"train_loss": 0.07398626208305359, "global_step": 248367, "epoch": 2790, "lr": 1.779977655871018e-06}
{"train_loss": 0.07585499435663223, "global_step": 248368, "epoch": 2790, "lr": 1.779824349986614e-06}
{"train_loss": 0.03449419513344765, "global_step": 248369, "epoch": 2790, "lr": 1.7796710505848192e-06}
{"train_loss": 0.03328971937298775, "global_step": 248370, "epoch": 2790, "lr": 1.7795177576656608e-06}
{"train_loss": 0.027351124212145805, "global_step": 248371, "epoch": 2790, "lr": 1.7793644712291557e-06}
{"train_loss": 0.03500168398022652, "global_step": 248372, "epoch": 2790, "lr": 1.7792111912753262e-06}
{"train_loss": 0.06470179557800293, "global_step": 248373, "epoch": 2790, "lr": 1.7790579178042e-06}
{"train_loss": 0.047822676599025726, "global_step": 248374, "epoch": 2790, "lr": 1.778904650815788e-06}
{"train_loss": 0.03132332116365433, "global_step": 248375, "epoch": 2790, "lr": 1.7787513903101183e-06}
{"train_loss": 0.05165755748748779, "global_step": 248376, "epoch": 2790, "lr": 1.7785981362872073e-06}
{"train_loss": 0.0650300532579422, "global_step": 248377, "epoch": 2790, "lr": 1.7784448887470773e-06}
{"train_loss": 0.04033924266695976, "global_step": 248378, "epoch": 2790, "lr": 1.7782916476897394e-06}
{"train_loss": 0.04590420424938202, "global_step": 248379, "epoch": 2790, "lr": 1.7781384131152323e-06}
{"train_loss": 0.05580170452594757, "global_step": 248380, "epoch": 2790, "lr": 1.7779851850235563e-06}
{"train_loss": 0.06093692034482956, "global_step": 248381, "epoch": 2790, "lr": 1.7778319634147556e-06}
{"train_loss": 0.02329830639064312, "global_step": 248382, "epoch": 2790, "lr": 1.7776787482888246e-06}
{"train_loss": 0.011474675498902798, "global_step": 248383, "epoch": 2790, "lr": 1.777525539645808e-06}
{"train_loss": 0.03994576632976532, "global_step": 248384, "epoch": 2790, "lr": 1.7773723374857055e-06}
{"train_loss": 0.0578191876411438, "global_step": 248385, "epoch": 2790, "lr": 1.777219141808556e-06}
{"train_loss": 0.06934262067079544, "global_step": 248386, "epoch": 2790, "lr": 1.7770659526143706e-06}
{"train_loss": 0.025127898901700974, "global_step": 248387, "epoch": 2790, "lr": 1.7769127699031662e-06}
{"train_loss": 0.03806253895163536, "global_step": 248388, "epoch": 2790, "lr": 1.7767595936749704e-06}
{"train_loss": 0.046737994998693466, "global_step": 248389, "epoch": 2790, "lr": 1.7766064239297997e-06}
{"train_loss": 0.06632093340158463, "global_step": 248390, "epoch": 2790, "lr": 1.7764532606676821e-06}
{"train_loss": 0.03912673145532608, "global_step": 248391, "epoch": 2790, "lr": 1.7763001038886228e-06}
{"train_loss": 0.03983050584793091, "global_step": 248392, "epoch": 2790, "lr": 1.7761469535926611e-06}
{"train_loss": 0.05713556334376335, "global_step": 248393, "epoch": 2790, "lr": 1.7759938097798079e-06}
{"train_loss": 0.029553435742855072, "global_step": 248394, "epoch": 2790, "lr": 1.7758406724500798e-06}
{"train_loss": 0.055710043758153915, "global_step": 248395, "epoch": 2790, "lr": 1.7756875416035046e-06}
{"train_loss": 0.014634545892477036, "global_step": 248396, "epoch": 2790, "lr": 1.7755344172401044e-06}
{"train_loss": 0.04743894562125206, "global_step": 248397, "epoch": 2790, "lr": 1.775381299359885e-06}
{"train_loss": 0.045054115307925455, "global_step": 248398, "epoch": 2790, "lr": 1.7752281879628908e-06, "val_loss": 9.54046630859375, "train_action_mse_error": 29.470176696777344}
{"train_loss": 0.022326132282614708, "global_step": 248399, "epoch": 2791, "lr": 1.7750750830491214e-06}
{"train_loss": 0.05657607689499855, "global_step": 248400, "epoch": 2791, "lr": 1.7749219846186105e-06}
{"train_loss": 0.022535104304552078, "global_step": 248401, "epoch": 2791, "lr": 1.7747688926713636e-06}
{"train_loss": 0.014939254149794579, "global_step": 248402, "epoch": 2791, "lr": 1.7746158072074248e-06}
{"train_loss": 0.06625691056251526, "global_step": 248403, "epoch": 2791, "lr": 1.7744627282267889e-06}
{"train_loss": 0.03859305381774902, "global_step": 248404, "epoch": 2791, "lr": 1.7743096557294947e-06}
{"train_loss": 0.07487423717975616, "global_step": 248405, "epoch": 2791, "lr": 1.7741565897155532e-06}
{"train_loss": 0.07651925832033157, "global_step": 248406, "epoch": 2791, "lr": 1.7740035301849977e-06}
{"train_loss": 0.044189903885126114, "global_step": 248407, "epoch": 2791, "lr": 1.7738504771378285e-06}
{"train_loss": 0.03993665426969528, "global_step": 248408, "epoch": 2791, "lr": 1.7736974305740783e-06}
{"train_loss": 0.021703796461224556, "global_step": 248409, "epoch": 2791, "lr": 1.7735443904937754e-06}
{"train_loss": 0.04310908541083336, "global_step": 248410, "epoch": 2791, "lr": 1.7733913568969251e-06}
{"train_loss": 0.040986254811286926, "global_step": 248411, "epoch": 2791, "lr": 1.7732383297835553e-06}
{"train_loss": 0.00973835401237011, "global_step": 248412, "epoch": 2791, "lr": 1.7730853091536882e-06}
{"train_loss": 0.0646870955824852, "global_step": 248413, "epoch": 2791, "lr": 1.7729322950073401e-06}
{"train_loss": 0.03220275789499283, "global_step": 248414, "epoch": 2791, "lr": 1.7727792873445338e-06}
{"train_loss": 0.034013472497463226, "global_step": 248415, "epoch": 2791, "lr": 1.772626286165291e-06}
{"train_loss": 0.0452125184237957, "global_step": 248416, "epoch": 2791, "lr": 1.7724732914696285e-06}
{"train_loss": 0.04541722685098648, "global_step": 248417, "epoch": 2791, "lr": 1.772320303257574e-06}
{"train_loss": 0.06248673051595688, "global_step": 248418, "epoch": 2791, "lr": 1.7721673215291335e-06}
{"train_loss": 0.030987709760665894, "global_step": 248419, "epoch": 2791, "lr": 1.7720143462843452e-06}
{"train_loss": 0.07348540425300598, "global_step": 248420, "epoch": 2791, "lr": 1.7718613775232151e-06}
{"train_loss": 0.037654146552085876, "global_step": 248421, "epoch": 2791, "lr": 1.771708415245782e-06}
{"train_loss": 0.03481513261795044, "global_step": 248422, "epoch": 2791, "lr": 1.7715554594520456e-06}
{"train_loss": 0.0432286411523819, "global_step": 248423, "epoch": 2791, "lr": 1.7714025101420395e-06}
{"train_loss": 0.037757448852062225, "global_step": 248424, "epoch": 2791, "lr": 1.7712495673157747e-06}
{"train_loss": 0.043132662773132324, "global_step": 248425, "epoch": 2791, "lr": 1.7710966309732846e-06}
{"train_loss": 0.07531518489122391, "global_step": 248426, "epoch": 2791, "lr": 1.7709437011145856e-06}
{"train_loss": 0.014986133202910423, "global_step": 248427, "epoch": 2791, "lr": 1.7707907777396838e-06}
{"train_loss": 0.053854383528232574, "global_step": 248428, "epoch": 2791, "lr": 1.770637860848623e-06}
{"train_loss": 0.04705287516117096, "global_step": 248429, "epoch": 2791, "lr": 1.7704849504414033e-06}
{"train_loss": 0.07399220764636993, "global_step": 248430, "epoch": 2791, "lr": 1.7703320465180585e-06}
{"train_loss": 0.11565150320529938, "global_step": 248431, "epoch": 2791, "lr": 1.770179149078599e-06}
{"train_loss": 0.06739836186170578, "global_step": 248432, "epoch": 2791, "lr": 1.7700262581230586e-06}
{"train_loss": 0.06656810641288757, "global_step": 248433, "epoch": 2791, "lr": 1.7698733736514484e-06}
{"train_loss": 0.048682842403650284, "global_step": 248434, "epoch": 2791, "lr": 1.7697204956637904e-06}
{"train_loss": 0.04578794911503792, "global_step": 248435, "epoch": 2791, "lr": 1.7695676241601011e-06}
{"train_loss": 0.06571371853351593, "global_step": 248436, "epoch": 2791, "lr": 1.7694147591404142e-06}
{"train_loss": 0.034828245639801025, "global_step": 248437, "epoch": 2791, "lr": 1.7692619006047352e-06}
{"train_loss": 0.045315761119127274, "global_step": 248438, "epoch": 2791, "lr": 1.7691090485530915e-06}
{"train_loss": 0.05453133583068848, "global_step": 248439, "epoch": 2791, "lr": 1.7689562029855055e-06}
{"train_loss": 0.05147448927164078, "global_step": 248440, "epoch": 2791, "lr": 1.768803363901994e-06}
{"train_loss": 0.10215193778276443, "global_step": 248441, "epoch": 2791, "lr": 1.7686505313025791e-06}
{"train_loss": 0.06391587108373642, "global_step": 248442, "epoch": 2791, "lr": 1.7684977051872831e-06}
{"train_loss": 0.02989366464316845, "global_step": 248443, "epoch": 2791, "lr": 1.7683448855561224e-06}
{"train_loss": 0.06565812230110168, "global_step": 248444, "epoch": 2791, "lr": 1.7681920724091138e-06}
{"train_loss": 0.0752350240945816, "global_step": 248445, "epoch": 2791, "lr": 1.7680392657462962e-06}
{"train_loss": 0.06607721745967865, "global_step": 248446, "epoch": 2791, "lr": 1.7678864655676696e-06}
{"train_loss": 0.02865036390721798, "global_step": 248447, "epoch": 2791, "lr": 1.7677336718732673e-06}
{"train_loss": 0.0909653976559639, "global_step": 248448, "epoch": 2791, "lr": 1.7675808846631003e-06}
{"train_loss": 0.04217737913131714, "global_step": 248449, "epoch": 2791, "lr": 1.7674281039372021e-06}
{"train_loss": 0.0687141865491867, "global_step": 248450, "epoch": 2791, "lr": 1.767275329695578e-06}
{"train_loss": 0.01909489929676056, "global_step": 248451, "epoch": 2791, "lr": 1.7671225619382613e-06}
{"train_loss": 0.035334132611751556, "global_step": 248452, "epoch": 2791, "lr": 1.7669698006652636e-06}
{"train_loss": 0.07876171171665192, "global_step": 248453, "epoch": 2791, "lr": 1.766817045876612e-06}
{"train_loss": 0.01121530495584011, "global_step": 248454, "epoch": 2791, "lr": 1.766664297572318e-06}
{"train_loss": 0.02928856573998928, "global_step": 248455, "epoch": 2791, "lr": 1.7665115557524147e-06}
{"train_loss": 0.04897288605570793, "global_step": 248456, "epoch": 2791, "lr": 1.766358820416908e-06}
{"train_loss": 0.05755665898323059, "global_step": 248457, "epoch": 2791, "lr": 1.7662060915658363e-06}
{"train_loss": 0.0580592043697834, "global_step": 248458, "epoch": 2791, "lr": 1.7660533691992055e-06}
{"train_loss": 0.06420436501502991, "global_step": 248459, "epoch": 2791, "lr": 1.765900653317043e-06}
{"train_loss": 0.06447172909975052, "global_step": 248460, "epoch": 2791, "lr": 1.7657479439193603e-06}
{"train_loss": 0.014096557162702084, "global_step": 248461, "epoch": 2791, "lr": 1.765595241006196e-06}
{"train_loss": 0.07858482748270035, "global_step": 248462, "epoch": 2791, "lr": 1.7654425445775503e-06}
{"train_loss": 0.059957005083560944, "global_step": 248463, "epoch": 2791, "lr": 1.7652898546334562e-06}
{"train_loss": 0.10840369760990143, "global_step": 248464, "epoch": 2791, "lr": 1.7651371711739362e-06}
{"train_loss": 0.020960165187716484, "global_step": 248465, "epoch": 2791, "lr": 1.7649844941989957e-06}
{"train_loss": 0.042879246175289154, "global_step": 248466, "epoch": 2791, "lr": 1.7648318237086737e-06}
{"train_loss": 0.031978897750377655, "global_step": 248467, "epoch": 2791, "lr": 1.7646791597029754e-06}
{"train_loss": 0.04411373659968376, "global_step": 248468, "epoch": 2791, "lr": 1.7645265021819346e-06}
{"train_loss": 0.030289340764284134, "global_step": 248469, "epoch": 2791, "lr": 1.7643738511455566e-06}
{"train_loss": 0.03418091684579849, "global_step": 248470, "epoch": 2791, "lr": 1.76422120659388e-06}
{"train_loss": 0.0179434884339571, "global_step": 248471, "epoch": 2791, "lr": 1.764068568526911e-06}
{"train_loss": 0.042007263749837875, "global_step": 248472, "epoch": 2791, "lr": 1.7639159369446767e-06}
{"train_loss": 0.026289617642760277, "global_step": 248473, "epoch": 2791, "lr": 1.763763311847194e-06}
{"train_loss": 0.042211197316646576, "global_step": 248474, "epoch": 2791, "lr": 1.7636106932344853e-06}
{"train_loss": 0.03574950620532036, "global_step": 248475, "epoch": 2791, "lr": 1.7634580811065726e-06}
{"train_loss": 0.04572081193327904, "global_step": 248476, "epoch": 2791, "lr": 1.7633054754634782e-06}
{"train_loss": 0.041449759155511856, "global_step": 248477, "epoch": 2791, "lr": 1.763152876305213e-06}
{"train_loss": 0.04372866824269295, "global_step": 248478, "epoch": 2791, "lr": 1.7630002836318105e-06}
{"train_loss": 0.024256443604826927, "global_step": 248479, "epoch": 2791, "lr": 1.762847697443276e-06}
{"train_loss": 0.012834909372031689, "global_step": 248480, "epoch": 2791, "lr": 1.7626951177396434e-06}
{"train_loss": 0.05549882724881172, "global_step": 248481, "epoch": 2791, "lr": 1.7625425445209287e-06}
{"train_loss": 0.08213360607624054, "global_step": 248482, "epoch": 2791, "lr": 1.7623899777871488e-06}
{"train_loss": 0.05935605615377426, "global_step": 248483, "epoch": 2791, "lr": 1.7622374175383372e-06}
{"train_loss": 0.030034950003027916, "global_step": 248484, "epoch": 2791, "lr": 1.7620848637744935e-06}
{"train_loss": 0.041755128651857376, "global_step": 248485, "epoch": 2791, "lr": 1.761932316495657e-06}
{"train_loss": 0.05699659883975983, "global_step": 248486, "epoch": 2791, "lr": 1.7617797757018384e-06}
{"train_loss": 0.04844023841904121, "global_step": 248487, "epoch": 2791, "lr": 1.7616272413930657e-06, "val_loss": 9.492873191833496}
{"train_loss": 0.08592331409454346, "global_step": 248488, "epoch": 2792, "lr": 1.7614747135693443e-06}
{"train_loss": 0.04837672784924507, "global_step": 248489, "epoch": 2792, "lr": 1.7613221922307134e-06}
{"train_loss": 0.05533191189169884, "global_step": 248490, "epoch": 2792, "lr": 1.761169677377178e-06}
{"train_loss": 0.046440936625003815, "global_step": 248491, "epoch": 2792, "lr": 1.7610171690087717e-06}
{"train_loss": 0.01844514161348343, "global_step": 248492, "epoch": 2792, "lr": 1.7608646671255002e-06}
{"train_loss": 0.04599085822701454, "global_step": 248493, "epoch": 2792, "lr": 1.7607121717274022e-06}
{"train_loss": 0.048180967569351196, "global_step": 248494, "epoch": 2792, "lr": 1.7605596828144832e-06}
{"train_loss": 0.06196298450231552, "global_step": 248495, "epoch": 2792, "lr": 1.760407200386771e-06}
{"train_loss": 0.09586037695407867, "global_step": 248496, "epoch": 2792, "lr": 1.7602547244442823e-06}
{"train_loss": 0.012505832128226757, "global_step": 248497, "epoch": 2792, "lr": 1.760102254987045e-06}
{"train_loss": 0.027861129492521286, "global_step": 248498, "epoch": 2792, "lr": 1.7599497920150643e-06}
{"train_loss": 0.013821878470480442, "global_step": 248499, "epoch": 2792, "lr": 1.7597973355283792e-06}
{"train_loss": 0.0657777488231659, "global_step": 248500, "epoch": 2792, "lr": 1.7596448855269955e-06}
{"train_loss": 0.02323487587273121, "global_step": 248501, "epoch": 2792, "lr": 1.7594924420109404e-06}
{"train_loss": 0.03635908663272858, "global_step": 248502, "epoch": 2792, "lr": 1.7593400049802366e-06}
{"train_loss": 0.0685601532459259, "global_step": 248503, "epoch": 2792, "lr": 1.7591875744349007e-06}
{"train_loss": 0.049407120794057846, "global_step": 248504, "epoch": 2792, "lr": 1.7590351503749546e-06}
{"train_loss": 0.06023566424846649, "global_step": 248505, "epoch": 2792, "lr": 1.7588827328004155e-06}
{"train_loss": 0.043464288115501404, "global_step": 248506, "epoch": 2792, "lr": 1.7587303217113105e-06}
{"train_loss": 0.05742218717932701, "global_step": 248507, "epoch": 2792, "lr": 1.7585779171076511e-06}
{"train_loss": 0.029686173424124718, "global_step": 248508, "epoch": 2792, "lr": 1.758425518989465e-06}
{"train_loss": 0.03192714601755142, "global_step": 248509, "epoch": 2792, "lr": 1.7582731273567688e-06}
{"train_loss": 0.05854874476790428, "global_step": 248510, "epoch": 2792, "lr": 1.7581207422095903e-06}
{"train_loss": 0.05941210687160492, "global_step": 248511, "epoch": 2792, "lr": 1.7579683635479404e-06}
{"train_loss": 0.06534834206104279, "global_step": 248512, "epoch": 2792, "lr": 1.7578159913718473e-06}
{"train_loss": 0.017400262877345085, "global_step": 248513, "epoch": 2792, "lr": 1.7576636256813217e-06}
{"train_loss": 0.059676412492990494, "global_step": 248514, "epoch": 2792, "lr": 1.757511266476397e-06}
{"train_loss": 0.04922513663768768, "global_step": 248515, "epoch": 2792, "lr": 1.757358913757079e-06}
{"train_loss": 0.05100228637456894, "global_step": 248516, "epoch": 2792, "lr": 1.757206567523406e-06}
{"train_loss": 0.01337907649576664, "global_step": 248517, "epoch": 2792, "lr": 1.7570542277753788e-06}
{"train_loss": 0.05896034091711044, "global_step": 248518, "epoch": 2792, "lr": 1.75690189451303e-06}
{"train_loss": 0.01969991624355316, "global_step": 248519, "epoch": 2792, "lr": 1.756749567736382e-06}
{"train_loss": 0.04916095733642578, "global_step": 248520, "epoch": 2792, "lr": 1.7565972474454463e-06}
{"train_loss": 0.015294390730559826, "global_step": 248521, "epoch": 2792, "lr": 1.75644493364025e-06}
{"train_loss": 0.04855602979660034, "global_step": 248522, "epoch": 2792, "lr": 1.75629262632081e-06}
{"train_loss": 0.0182516947388649, "global_step": 248523, "epoch": 2792, "lr": 1.7561403254871546e-06}
{"train_loss": 0.06914588809013367, "global_step": 248524, "epoch": 2792, "lr": 1.7559880311392885e-06}
{"train_loss": 0.0698355883359909, "global_step": 248525, "epoch": 2792, "lr": 1.755835743277251e-06}
{"train_loss": 0.04194527491927147, "global_step": 248526, "epoch": 2792, "lr": 1.7556834619010476e-06}
{"train_loss": 0.08990275859832764, "global_step": 248527, "epoch": 2792, "lr": 1.7555311870107117e-06}
{"train_loss": 0.047990188002586365, "global_step": 248528, "epoch": 2792, "lr": 1.7553789186062485e-06}
{"train_loss": 0.06132519245147705, "global_step": 248529, "epoch": 2792, "lr": 1.7552266566876917e-06}
{"train_loss": 0.04172690212726593, "global_step": 248530, "epoch": 2792, "lr": 1.7550744012550523e-06}
{"train_loss": 0.05986710637807846, "global_step": 248531, "epoch": 2792, "lr": 1.7549221523083581e-06}
{"train_loss": 0.03207533806562424, "global_step": 248532, "epoch": 2792, "lr": 1.7547699098476256e-06}
{"train_loss": 0.030671346932649612, "global_step": 248533, "epoch": 2792, "lr": 1.7546176738728826e-06}
{"train_loss": 0.07397568970918655, "global_step": 248534, "epoch": 2792, "lr": 1.7544654443841291e-06}
{"train_loss": 0.03527793288230896, "global_step": 248535, "epoch": 2792, "lr": 1.7543132213814095e-06}
{"train_loss": 0.04239242151379585, "global_step": 248536, "epoch": 2792, "lr": 1.754161004864735e-06}
{"train_loss": 0.018378624692559242, "global_step": 248537, "epoch": 2792, "lr": 1.7540087948341278e-06}
{"train_loss": 0.05396930128335953, "global_step": 248538, "epoch": 2792, "lr": 1.7538565912895987e-06}
{"train_loss": 0.059569597244262695, "global_step": 248539, "epoch": 2792, "lr": 1.7537043942311814e-06}
{"train_loss": 0.05275014787912369, "global_step": 248540, "epoch": 2792, "lr": 1.7535522036588815e-06}
{"train_loss": 0.038128308951854706, "global_step": 248541, "epoch": 2792, "lr": 1.7534000195727374e-06}
{"train_loss": 0.0386134535074234, "global_step": 248542, "epoch": 2792, "lr": 1.7532478419727604e-06}
{"train_loss": 0.04549723491072655, "global_step": 248543, "epoch": 2792, "lr": 1.7530956708589674e-06}
{"train_loss": 0.03621792048215866, "global_step": 248544, "epoch": 2792, "lr": 1.752943506231386e-06}
{"train_loss": 0.0671122744679451, "global_step": 248545, "epoch": 2792, "lr": 1.752791348090027e-06}
{"train_loss": 0.027590714395046234, "global_step": 248546, "epoch": 2792, "lr": 1.7526391964349242e-06}
{"train_loss": 0.04482992738485336, "global_step": 248547, "epoch": 2792, "lr": 1.7524870512660885e-06}
{"train_loss": 0.034344326704740524, "global_step": 248548, "epoch": 2792, "lr": 1.752334912583542e-06}
{"train_loss": 0.063505619764328, "global_step": 248549, "epoch": 2792, "lr": 1.752182780387307e-06}
{"train_loss": 0.017544494941830635, "global_step": 248550, "epoch": 2792, "lr": 1.7520306546774056e-06}
{"train_loss": 0.046693671494722366, "global_step": 248551, "epoch": 2792, "lr": 1.7518785354538492e-06}
{"train_loss": 0.02943236567080021, "global_step": 248552, "epoch": 2792, "lr": 1.7517264227166708e-06}
{"train_loss": 0.008165253326296806, "global_step": 248553, "epoch": 2792, "lr": 1.751574316465876e-06}
{"train_loss": 0.06750109046697617, "global_step": 248554, "epoch": 2792, "lr": 1.7514222167014983e-06}
{"train_loss": 0.0327729694545269, "global_step": 248555, "epoch": 2792, "lr": 1.7512701234235595e-06}
{"train_loss": 0.03648810461163521, "global_step": 248556, "epoch": 2792, "lr": 1.7511180366320656e-06}
{"train_loss": 0.04288467764854431, "global_step": 248557, "epoch": 2792, "lr": 1.7509659563270553e-06}
{"train_loss": 0.09383879601955414, "global_step": 248558, "epoch": 2792, "lr": 1.7508138825085285e-06}
{"train_loss": 0.035487495362758636, "global_step": 248559, "epoch": 2792, "lr": 1.7506618151765298e-06}
{"train_loss": 0.06022869423031807, "global_step": 248560, "epoch": 2792, "lr": 1.7505097543310533e-06}
{"train_loss": 0.03340063989162445, "global_step": 248561, "epoch": 2792, "lr": 1.750357699972144e-06}
{"train_loss": 0.03200746327638626, "global_step": 248562, "epoch": 2792, "lr": 1.7502056520998012e-06}
{"train_loss": 0.030643517151474953, "global_step": 248563, "epoch": 2792, "lr": 1.750053610714064e-06}
{"train_loss": 0.07550495117902756, "global_step": 248564, "epoch": 2792, "lr": 1.7499015758149385e-06}
{"train_loss": 0.031309932470321655, "global_step": 248565, "epoch": 2792, "lr": 1.7497495474024516e-06}
{"train_loss": 0.0332677848637104, "global_step": 248566, "epoch": 2792, "lr": 1.7495975254766205e-06}
{"train_loss": 0.07988040149211884, "global_step": 248567, "epoch": 2792, "lr": 1.749445510037473e-06}
{"train_loss": 0.06759078800678253, "global_step": 248568, "epoch": 2792, "lr": 1.7492935010850198e-06}
{"train_loss": 0.04755482077598572, "global_step": 248569, "epoch": 2792, "lr": 1.7491414986192945e-06}
{"train_loss": 0.04898606985807419, "global_step": 248570, "epoch": 2792, "lr": 1.748989502640297e-06}
{"train_loss": 0.05370040237903595, "global_step": 248571, "epoch": 2792, "lr": 1.7488375131480661e-06}
{"train_loss": 0.041359465569257736, "global_step": 248572, "epoch": 2792, "lr": 1.7486855301426187e-06}
{"train_loss": 0.046282537281513214, "global_step": 248573, "epoch": 2792, "lr": 1.7485335536239655e-06}
{"train_loss": 0.040851738303899765, "global_step": 248574, "epoch": 2792, "lr": 1.7483815835921403e-06}
{"train_loss": 0.021104594692587852, "global_step": 248575, "epoch": 2792, "lr": 1.7482296200471593e-06}
{"train_loss": 0.04652386789701963, "global_step": 248576, "epoch": 2792, "lr": 1.7480776629890284e-06, "val_loss": 9.503866195678711}
{"train_loss": 0.04907100647687912, "global_step": 248577, "epoch": 2793, "lr": 1.7479257124177917e-06}
{"train_loss": 0.05408090353012085, "global_step": 248578, "epoch": 2793, "lr": 1.747773768333455e-06}
{"train_loss": 0.04709656909108162, "global_step": 248579, "epoch": 2793, "lr": 1.7476218307360403e-06}
{"train_loss": 0.008552216924726963, "global_step": 248580, "epoch": 2793, "lr": 1.7474698996255755e-06}
{"train_loss": 0.09357369691133499, "global_step": 248581, "epoch": 2793, "lr": 1.7473179750020662e-06}
{"train_loss": 0.03165373206138611, "global_step": 248582, "epoch": 2793, "lr": 1.7471660568655513e-06}
{"train_loss": 0.04498279467225075, "global_step": 248583, "epoch": 2793, "lr": 1.7470141452160304e-06}
{"train_loss": 0.07733067125082016, "global_step": 248584, "epoch": 2793, "lr": 1.7468622400535484e-06}
{"train_loss": 0.035928841680288315, "global_step": 248585, "epoch": 2793, "lr": 1.7467103413780995e-06}
{"train_loss": 0.04588913172483444, "global_step": 248586, "epoch": 2793, "lr": 1.746558449189728e-06}
{"train_loss": 0.0307577196508646, "global_step": 248587, "epoch": 2793, "lr": 1.746406563488434e-06}
{"train_loss": 0.09506165236234665, "global_step": 248588, "epoch": 2793, "lr": 1.7462546842742566e-06}
{"train_loss": 0.09451542049646378, "global_step": 248589, "epoch": 2793, "lr": 1.746102811547201e-06}
{"train_loss": 0.05988519266247749, "global_step": 248590, "epoch": 2793, "lr": 1.7459509453072952e-06}
{"train_loss": 0.04195815324783325, "global_step": 248591, "epoch": 2793, "lr": 1.7457990855545614e-06}
{"train_loss": 0.06928586214780807, "global_step": 248592, "epoch": 2793, "lr": 1.7456472322890105e-06}
{"train_loss": 0.009115299209952354, "global_step": 248593, "epoch": 2793, "lr": 1.7454953855106759e-06}
{"train_loss": 0.040906477719545364, "global_step": 248594, "epoch": 2793, "lr": 1.7453435452195631e-06}
{"train_loss": 0.01151559129357338, "global_step": 248595, "epoch": 2793, "lr": 1.7451917114157112e-06}
{"train_loss": 0.025546202436089516, "global_step": 248596, "epoch": 2793, "lr": 1.74503988409912e-06}
{"train_loss": 0.014164429157972336, "global_step": 248597, "epoch": 2793, "lr": 1.7448880632698283e-06}
{"train_loss": 0.04836071655154228, "global_step": 248598, "epoch": 2793, "lr": 1.7447362489278418e-06}
{"train_loss": 0.0297780092805624, "global_step": 248599, "epoch": 2793, "lr": 1.7445844410731938e-06}
{"train_loss": 0.025798633694648743, "global_step": 248600, "epoch": 2793, "lr": 1.7444326397058951e-06}
{"train_loss": 0.02831694856286049, "global_step": 248601, "epoch": 2793, "lr": 1.744280844825974e-06}
{"train_loss": 0.05694136768579483, "global_step": 248602, "epoch": 2793, "lr": 1.7441290564334356e-06}
{"train_loss": 0.03429930657148361, "global_step": 248603, "epoch": 2793, "lr": 1.743977274528319e-06}
{"train_loss": 0.015422542579472065, "global_step": 248604, "epoch": 2793, "lr": 1.7438254991106295e-06}
{"train_loss": 0.02655051089823246, "global_step": 248605, "epoch": 2793, "lr": 1.7436737301804063e-06}
{"train_loss": 0.026255473494529724, "global_step": 248606, "epoch": 2793, "lr": 1.7435219677376436e-06}
{"train_loss": 0.033048637211322784, "global_step": 248607, "epoch": 2793, "lr": 1.7433702117823859e-06}
{"train_loss": 0.07367759943008423, "global_step": 248608, "epoch": 2793, "lr": 1.743218462314644e-06}
{"train_loss": 0.08019965887069702, "global_step": 248609, "epoch": 2793, "lr": 1.7430667193344297e-06}
{"train_loss": 0.03294496610760689, "global_step": 248610, "epoch": 2793, "lr": 1.7429149828417811e-06}
{"train_loss": 0.052668001502752304, "global_step": 248611, "epoch": 2793, "lr": 1.7427632528367044e-06}
{"train_loss": 0.018069930374622345, "global_step": 248612, "epoch": 2793, "lr": 1.7426115293192269e-06}
{"train_loss": 0.06008252501487732, "global_step": 248613, "epoch": 2793, "lr": 1.742459812289371e-06}
{"train_loss": 0.05867742374539375, "global_step": 248614, "epoch": 2793, "lr": 1.7423081017471477e-06}
{"train_loss": 0.05343608185648918, "global_step": 248615, "epoch": 2793, "lr": 1.7421563976925847e-06}
{"train_loss": 0.05154396593570709, "global_step": 248616, "epoch": 2793, "lr": 1.742004700125699e-06}
{"train_loss": 0.05366821587085724, "global_step": 248617, "epoch": 2793, "lr": 1.7418530090465179e-06}
{"train_loss": 0.02769426256418228, "global_step": 248618, "epoch": 2793, "lr": 1.7417013244550474e-06}
{"train_loss": 0.04546723887324333, "global_step": 248619, "epoch": 2793, "lr": 1.7415496463513258e-06}
{"train_loss": 0.03624409809708595, "global_step": 248620, "epoch": 2793, "lr": 1.7413979747353648e-06}
{"train_loss": 0.025534244254231453, "global_step": 248621, "epoch": 2793, "lr": 1.7412463096071806e-06}
{"train_loss": 0.03624600172042847, "global_step": 248622, "epoch": 2793, "lr": 1.7410946509667958e-06}
{"train_loss": 0.04059365391731262, "global_step": 248623, "epoch": 2793, "lr": 1.7409429988142323e-06}
{"train_loss": 0.053930796682834625, "global_step": 248624, "epoch": 2793, "lr": 1.7407913531495178e-06}
{"train_loss": 0.02937699854373932, "global_step": 248625, "epoch": 2793, "lr": 1.7406397139726582e-06}
{"train_loss": 0.044657401740550995, "global_step": 248626, "epoch": 2793, "lr": 1.740488081283681e-06}
{"train_loss": 0.03148047253489494, "global_step": 248627, "epoch": 2793, "lr": 1.740336455082614e-06}
{"train_loss": 0.049454156309366226, "global_step": 248628, "epoch": 2793, "lr": 1.7401848353694627e-06}
{"train_loss": 0.04379229247570038, "global_step": 248629, "epoch": 2793, "lr": 1.7400332221442606e-06}
{"train_loss": 0.031176552176475525, "global_step": 248630, "epoch": 2793, "lr": 1.739881615407024e-06}
{"train_loss": 0.05553712695837021, "global_step": 248631, "epoch": 2793, "lr": 1.73973001515777e-06}
{"train_loss": 0.019163452088832855, "global_step": 248632, "epoch": 2793, "lr": 1.7395784213965205e-06}
{"train_loss": 0.04729366675019264, "global_step": 248633, "epoch": 2793, "lr": 1.7394268341232977e-06}
{"train_loss": 0.05414882302284241, "global_step": 248634, "epoch": 2793, "lr": 1.7392752533381185e-06}
{"train_loss": 0.024491511285305023, "global_step": 248635, "epoch": 2793, "lr": 1.7391236790410104e-06}
{"train_loss": 0.029141917824745178, "global_step": 248636, "epoch": 2793, "lr": 1.738972111231979e-06}
{"train_loss": 0.05121925473213196, "global_step": 248637, "epoch": 2793, "lr": 1.7388205499110632e-06}
{"train_loss": 0.006329294294118881, "global_step": 248638, "epoch": 2793, "lr": 1.7386689950782686e-06}
{"train_loss": 0.09196342527866364, "global_step": 248639, "epoch": 2793, "lr": 1.738517446733634e-06}
{"train_loss": 0.034849364310503006, "global_step": 248640, "epoch": 2793, "lr": 1.7383659048771538e-06}
{"train_loss": 0.016963019967079163, "global_step": 248641, "epoch": 2793, "lr": 1.7382143695088727e-06}
{"train_loss": 0.027312539517879486, "global_step": 248642, "epoch": 2793, "lr": 1.7380628406287903e-06}
{"train_loss": 0.03419569507241249, "global_step": 248643, "epoch": 2793, "lr": 1.7379113182369455e-06}
{"train_loss": 0.05983791872859001, "global_step": 248644, "epoch": 2793, "lr": 1.737759802333344e-06}
{"train_loss": 0.05842329561710358, "global_step": 248645, "epoch": 2793, "lr": 1.7376082929180138e-06}
{"train_loss": 0.027263715863227844, "global_step": 248646, "epoch": 2793, "lr": 1.7374567899909766e-06}
{"train_loss": 0.07476289570331573, "global_step": 248647, "epoch": 2793, "lr": 1.7373052935522438e-06}
{"train_loss": 0.025684872642159462, "global_step": 248648, "epoch": 2793, "lr": 1.7371538036018543e-06}
{"train_loss": 0.019593264907598495, "global_step": 248649, "epoch": 2793, "lr": 1.7370023201398022e-06}
{"train_loss": 0.04165160655975342, "global_step": 248650, "epoch": 2793, "lr": 1.7368508431661323e-06}
{"train_loss": 0.05784999951720238, "global_step": 248651, "epoch": 2793, "lr": 1.73669937268085e-06}
{"train_loss": 0.029768850654363632, "global_step": 248652, "epoch": 2793, "lr": 1.736547908683983e-06}
{"train_loss": 0.052144743502140045, "global_step": 248653, "epoch": 2793, "lr": 1.7363964511755482e-06}
{"train_loss": 0.08538539707660675, "global_step": 248654, "epoch": 2793, "lr": 1.7362450001555618e-06}
{"train_loss": 0.06828825175762177, "global_step": 248655, "epoch": 2793, "lr": 1.7360935556240577e-06}
{"train_loss": 0.052469443529844284, "global_step": 248656, "epoch": 2793, "lr": 1.7359421175810409e-06}
{"train_loss": 0.0407363623380661, "global_step": 248657, "epoch": 2793, "lr": 1.735790686026545e-06}
{"train_loss": 0.03821944817900658, "global_step": 248658, "epoch": 2793, "lr": 1.735639260960581e-06}
{"train_loss": 0.02834075689315796, "global_step": 248659, "epoch": 2793, "lr": 1.7354878423831656e-06}
{"train_loss": 0.030191874131560326, "global_step": 248660, "epoch": 2793, "lr": 1.735336430294332e-06}
{"train_loss": 0.07233867794275284, "global_step": 248661, "epoch": 2793, "lr": 1.7351850246940914e-06}
{"train_loss": 0.029277443885803223, "global_step": 248662, "epoch": 2793, "lr": 1.7350336255824662e-06}
{"train_loss": 0.06461162120103836, "global_step": 248663, "epoch": 2793, "lr": 1.7348822329594783e-06}
{"train_loss": 0.03195299208164215, "global_step": 248664, "epoch": 2793, "lr": 1.7347308468251443e-06}
{"train_loss": 0.04306999688151847, "global_step": 248665, "epoch": 2793, "lr": 1.734579467179498e-06, "val_loss": 9.519144058227539}
{"train_loss": 0.021947108209133148, "global_step": 248666, "epoch": 2794, "lr": 1.7344280940225388e-06}
{"train_loss": 0.011421061120927334, "global_step": 248667, "epoch": 2794, "lr": 1.7342767273543004e-06}
{"train_loss": 0.04069104790687561, "global_step": 248668, "epoch": 2794, "lr": 1.7341253671747993e-06}
{"train_loss": 0.07792645692825317, "global_step": 248669, "epoch": 2794, "lr": 1.7339740134840631e-06}
{"train_loss": 0.02236856147646904, "global_step": 248670, "epoch": 2794, "lr": 1.7338226662820978e-06}
{"train_loss": 0.043916475027799606, "global_step": 248671, "epoch": 2794, "lr": 1.7336713255689362e-06}
{"train_loss": 0.029980527237057686, "global_step": 248672, "epoch": 2794, "lr": 1.73351999134459e-06}
{"train_loss": 0.056077346205711365, "global_step": 248673, "epoch": 2794, "lr": 1.7333686636090918e-06}
{"train_loss": 0.03692301735281944, "global_step": 248674, "epoch": 2794, "lr": 1.733217342362442e-06}
{"train_loss": 0.060949262231588364, "global_step": 248675, "epoch": 2794, "lr": 1.733066027604685e-06}
{"train_loss": 0.06137614697217941, "global_step": 248676, "epoch": 2794, "lr": 1.732914719335821e-06}
{"train_loss": 0.03546522557735443, "global_step": 248677, "epoch": 2794, "lr": 1.7327634175558826e-06}
{"train_loss": 0.09962748736143112, "global_step": 248678, "epoch": 2794, "lr": 1.7326121222648817e-06}
{"train_loss": 0.08377460390329361, "global_step": 248679, "epoch": 2794, "lr": 1.7324608334628455e-06}
{"train_loss": 0.032743386924266815, "global_step": 248680, "epoch": 2794, "lr": 1.7323095511497912e-06}
{"train_loss": 0.04426554590463638, "global_step": 248681, "epoch": 2794, "lr": 1.7321582753257348e-06}
{"train_loss": 0.07086969166994095, "global_step": 248682, "epoch": 2794, "lr": 1.73200700599071e-06}
{"train_loss": 0.07027853280305862, "global_step": 248683, "epoch": 2794, "lr": 1.7318557431447225e-06}
{"train_loss": 0.02175496704876423, "global_step": 248684, "epoch": 2794, "lr": 1.731704486787805e-06}
{"train_loss": 0.06858143210411072, "global_step": 248685, "epoch": 2794, "lr": 1.7315532369199638e-06}
{"train_loss": 0.0710393488407135, "global_step": 248686, "epoch": 2794, "lr": 1.7314019935412318e-06}
{"train_loss": 0.017347438260912895, "global_step": 248687, "epoch": 2794, "lr": 1.73125075665162e-06}
{"train_loss": 0.06571517884731293, "global_step": 248688, "epoch": 2794, "lr": 1.731099526251162e-06}
{"train_loss": 0.062664695084095, "global_step": 248689, "epoch": 2794, "lr": 1.7309483023398632e-06}
{"train_loss": 0.015459111891686916, "global_step": 248690, "epoch": 2794, "lr": 1.7307970849177513e-06}
{"train_loss": 0.046425361186265945, "global_step": 248691, "epoch": 2794, "lr": 1.7306458739848487e-06}
{"train_loss": 0.05375986546278, "global_step": 248692, "epoch": 2794, "lr": 1.7304946695411662e-06}
{"train_loss": 0.03327926993370056, "global_step": 248693, "epoch": 2794, "lr": 1.7303434715867372e-06}
{"train_loss": 0.06453187763690948, "global_step": 248694, "epoch": 2794, "lr": 1.7301922801215676e-06}
{"train_loss": 0.03321956843137741, "global_step": 248695, "epoch": 2794, "lr": 1.7300410951456901e-06}
{"train_loss": 0.03986384719610214, "global_step": 248696, "epoch": 2794, "lr": 1.729889916659122e-06}
{"train_loss": 0.012081664055585861, "global_step": 248697, "epoch": 2794, "lr": 1.729738744661874e-06}
{"train_loss": 0.09482656419277191, "global_step": 248698, "epoch": 2794, "lr": 1.7295875791539794e-06}
{"train_loss": 0.06789519637823105, "global_step": 248699, "epoch": 2794, "lr": 1.729436420135455e-06}
{"train_loss": 0.05418500304222107, "global_step": 248700, "epoch": 2794, "lr": 1.729285267606312e-06}
{"train_loss": 0.025200219824910164, "global_step": 248701, "epoch": 2794, "lr": 1.7291341215665891e-06}
{"train_loss": 0.027273988351225853, "global_step": 248702, "epoch": 2794, "lr": 1.728982982016286e-06}
{"train_loss": 0.050761424005031586, "global_step": 248703, "epoch": 2794, "lr": 1.7288318489554423e-06}
{"train_loss": 0.052027154713869095, "global_step": 248704, "epoch": 2794, "lr": 1.728680722384063e-06}
{"train_loss": 0.0981360524892807, "global_step": 248705, "epoch": 2794, "lr": 1.7285296023021757e-06}
{"train_loss": 0.04914278909564018, "global_step": 248706, "epoch": 2794, "lr": 1.7283784887097977e-06}
{"train_loss": 0.0640401616692543, "global_step": 248707, "epoch": 2794, "lr": 1.7282273816069505e-06}
{"train_loss": 0.050891123712062836, "global_step": 248708, "epoch": 2794, "lr": 1.7280762809936568e-06}
{"train_loss": 0.0804663747549057, "global_step": 248709, "epoch": 2794, "lr": 1.727925186869933e-06}
{"train_loss": 0.03139377012848854, "global_step": 248710, "epoch": 2794, "lr": 1.7277740992358016e-06}
{"train_loss": 0.05073774978518486, "global_step": 248711, "epoch": 2794, "lr": 1.7276230180912844e-06}
{"train_loss": 0.028544997796416283, "global_step": 248712, "epoch": 2794, "lr": 1.7274719434363983e-06}
{"train_loss": 0.043537382036447525, "global_step": 248713, "epoch": 2794, "lr": 1.7273208752711655e-06}
{"train_loss": 0.04609440639615059, "global_step": 248714, "epoch": 2794, "lr": 1.7271698135956082e-06}
{"train_loss": 0.02147889882326126, "global_step": 248715, "epoch": 2794, "lr": 1.727018758409743e-06}
{"train_loss": 0.05736929923295975, "global_step": 248716, "epoch": 2794, "lr": 1.7268677097135922e-06}
{"train_loss": 0.024025999009609222, "global_step": 248717, "epoch": 2794, "lr": 1.7267166675071721e-06}
{"train_loss": 0.06094992533326149, "global_step": 248718, "epoch": 2794, "lr": 1.726565631790511e-06}
{"train_loss": 0.030870186164975166, "global_step": 248719, "epoch": 2794, "lr": 1.7264146025636197e-06}
{"train_loss": 0.029738551005721092, "global_step": 248720, "epoch": 2794, "lr": 1.7262635798265314e-06}
{"train_loss": 0.02745576947927475, "global_step": 248721, "epoch": 2794, "lr": 1.726112563579252e-06}
{"train_loss": 0.05690857023000717, "global_step": 248722, "epoch": 2794, "lr": 1.7259615538218088e-06}
{"train_loss": 0.0736185610294342, "global_step": 248723, "epoch": 2794, "lr": 1.7258105505542245e-06}
{"train_loss": 0.01132846251130104, "global_step": 248724, "epoch": 2794, "lr": 1.7256595537765153e-06}
{"train_loss": 0.03098873980343342, "global_step": 248725, "epoch": 2794, "lr": 1.7255085634887036e-06}
{"train_loss": 0.053619302809238434, "global_step": 248726, "epoch": 2794, "lr": 1.7253575796908117e-06}
{"train_loss": 0.0496055893599987, "global_step": 248727, "epoch": 2794, "lr": 1.7252066023828507e-06}
{"train_loss": 0.08580710738897324, "global_step": 248728, "epoch": 2794, "lr": 1.7250556315648536e-06}
{"train_loss": 0.0255768820643425, "global_step": 248729, "epoch": 2794, "lr": 1.7249046672368318e-06}
{"train_loss": 0.0768836960196495, "global_step": 248730, "epoch": 2794, "lr": 1.7247537093988076e-06}
{"train_loss": 0.06827659904956818, "global_step": 248731, "epoch": 2794, "lr": 1.724602758050803e-06}
{"train_loss": 0.0704106017947197, "global_step": 248732, "epoch": 2794, "lr": 1.7244518131928344e-06}
{"train_loss": 0.031654760241508484, "global_step": 248733, "epoch": 2794, "lr": 1.72430087482493e-06}
{"train_loss": 0.048143334686756134, "global_step": 248734, "epoch": 2794, "lr": 1.7241499429470953e-06}
{"train_loss": 0.042861394584178925, "global_step": 248735, "epoch": 2794, "lr": 1.723999017559369e-06}
{"train_loss": 0.018479928374290466, "global_step": 248736, "epoch": 2794, "lr": 1.7238480986617567e-06}
{"train_loss": 0.06988992542028427, "global_step": 248737, "epoch": 2794, "lr": 1.7236971862542916e-06}
{"train_loss": 0.06940483301877975, "global_step": 248738, "epoch": 2794, "lr": 1.7235462803369794e-06}
{"train_loss": 0.029651079326868057, "global_step": 248739, "epoch": 2794, "lr": 1.7233953809098591e-06}
{"train_loss": 0.07198626548051834, "global_step": 248740, "epoch": 2794, "lr": 1.7232444879729304e-06}
{"train_loss": 0.07966631650924683, "global_step": 248741, "epoch": 2794, "lr": 1.7230936015262267e-06}
{"train_loss": 0.029042255133390427, "global_step": 248742, "epoch": 2794, "lr": 1.7229427215697592e-06}
{"train_loss": 0.03743186965584755, "global_step": 248743, "epoch": 2794, "lr": 1.722791848103561e-06}
{"train_loss": 0.018075894564390182, "global_step": 248744, "epoch": 2794, "lr": 1.7226409811276378e-06}
{"train_loss": 0.08040980994701385, "global_step": 248745, "epoch": 2794, "lr": 1.722490120642023e-06}
{"train_loss": 0.08361661434173584, "global_step": 248746, "epoch": 2794, "lr": 1.7223392666467276e-06}
{"train_loss": 0.024212799966335297, "global_step": 248747, "epoch": 2794, "lr": 1.7221884191417792e-06}
{"train_loss": 0.016728660091757774, "global_step": 248748, "epoch": 2794, "lr": 1.7220375781271891e-06}
{"train_loss": 0.03294030949473381, "global_step": 248749, "epoch": 2794, "lr": 1.7218867436029905e-06}
{"train_loss": 0.03577273711562157, "global_step": 248750, "epoch": 2794, "lr": 1.7217359155691893e-06}
{"train_loss": 0.02702445164322853, "global_step": 248751, "epoch": 2794, "lr": 1.7215850940258128e-06}
{"train_loss": 0.07425326108932495, "global_step": 248752, "epoch": 2794, "lr": 1.7214342789728777e-06}
{"train_loss": 0.05346764624118805, "global_step": 248753, "epoch": 2794, "lr": 1.721283470410412e-06}
{"train_loss": 0.04838394300405229, "global_step": 248754, "epoch": 2794, "lr": 1.721132668338432e-06, "val_loss": 9.480470657348633}
{"train_loss": 0.0466240718960762, "global_step": 248755, "epoch": 2795, "lr": 1.7209818727569548e-06}
{"train_loss": 0.087760791182518, "global_step": 248756, "epoch": 2795, "lr": 1.7208310836660024e-06}
{"train_loss": 0.06883841753005981, "global_step": 248757, "epoch": 2795, "lr": 1.7206803010655969e-06}
{"train_loss": 0.05855691432952881, "global_step": 248758, "epoch": 2795, "lr": 1.7205295249557607e-06}
{"train_loss": 0.028832202777266502, "global_step": 248759, "epoch": 2795, "lr": 1.7203787553365048e-06}
{"train_loss": 0.07042050361633301, "global_step": 248760, "epoch": 2795, "lr": 1.7202279922078623e-06}
{"train_loss": 0.023786066100001335, "global_step": 248761, "epoch": 2795, "lr": 1.720077235569839e-06}
{"train_loss": 0.053300537168979645, "global_step": 248762, "epoch": 2795, "lr": 1.7199264854224739e-06}
{"train_loss": 0.04455750063061714, "global_step": 248763, "epoch": 2795, "lr": 1.7197757417657667e-06}
{"train_loss": 0.03582651913166046, "global_step": 248764, "epoch": 2795, "lr": 1.7196250045997509e-06}
{"train_loss": 0.014058241620659828, "global_step": 248765, "epoch": 2795, "lr": 1.7194742739244373e-06}
{"train_loss": 0.06299764662981033, "global_step": 248766, "epoch": 2795, "lr": 1.7193235497398598e-06}
{"train_loss": 0.02613639086484909, "global_step": 248767, "epoch": 2795, "lr": 1.7191728320460287e-06}
{"train_loss": 0.041749507188797, "global_step": 248768, "epoch": 2795, "lr": 1.719022120842967e-06}
{"train_loss": 0.07767564058303833, "global_step": 248769, "epoch": 2795, "lr": 1.7188714161306962e-06}
{"train_loss": 0.0318111777305603, "global_step": 248770, "epoch": 2795, "lr": 1.7187207179092279e-06}
{"train_loss": 0.06366467475891113, "global_step": 248771, "epoch": 2795, "lr": 1.7185700261785953e-06}
{"train_loss": 0.07134726643562317, "global_step": 248772, "epoch": 2795, "lr": 1.7184193409388039e-06}
{"train_loss": 0.01811516284942627, "global_step": 248773, "epoch": 2795, "lr": 1.7182686621898925e-06}
{"train_loss": 0.03142577409744263, "global_step": 248774, "epoch": 2795, "lr": 1.7181179899318667e-06}
{"train_loss": 0.07662484794855118, "global_step": 248775, "epoch": 2795, "lr": 1.7179673241647542e-06}
{"train_loss": 0.05265594646334648, "global_step": 248776, "epoch": 2795, "lr": 1.7178166648885663e-06}
{"train_loss": 0.05786692723631859, "global_step": 248777, "epoch": 2795, "lr": 1.7176660121033362e-06}
{"train_loss": 0.0488026887178421, "global_step": 248778, "epoch": 2795, "lr": 1.7175153658090749e-06}
{"train_loss": 0.03216465562582016, "global_step": 248779, "epoch": 2795, "lr": 1.7173647260058045e-06}
{"train_loss": 0.030088385567069054, "global_step": 248780, "epoch": 2795, "lr": 1.717214092693542e-06}
{"train_loss": 0.0454959012567997, "global_step": 248781, "epoch": 2795, "lr": 1.7170634658723207e-06}
{"train_loss": 0.046061478555202484, "global_step": 248782, "epoch": 2795, "lr": 1.7169128455421457e-06}
{"train_loss": 0.03196784853935242, "global_step": 248783, "epoch": 2795, "lr": 1.716762231703045e-06}
{"train_loss": 0.053476255387067795, "global_step": 248784, "epoch": 2795, "lr": 1.7166116243550356e-06}
{"train_loss": 0.058466751128435135, "global_step": 248785, "epoch": 2795, "lr": 1.716461023498145e-06}
{"train_loss": 0.05379517748951912, "global_step": 248786, "epoch": 2795, "lr": 1.7163104291323783e-06}
{"train_loss": 0.06352092325687408, "global_step": 248787, "epoch": 2795, "lr": 1.716159841257775e-06}
{"train_loss": 0.02804233878850937, "global_step": 248788, "epoch": 2795, "lr": 1.716009259874335e-06}
{"train_loss": 0.022093426436185837, "global_step": 248789, "epoch": 2795, "lr": 1.7158586849820913e-06}
{"train_loss": 0.04629620909690857, "global_step": 248790, "epoch": 2795, "lr": 1.7157081165810662e-06}
{"train_loss": 0.024433346465229988, "global_step": 248791, "epoch": 2795, "lr": 1.7155575546712765e-06}
{"train_loss": 0.056050386279821396, "global_step": 248792, "epoch": 2795, "lr": 1.7154069992527387e-06}
{"train_loss": 0.06195874139666557, "global_step": 248793, "epoch": 2795, "lr": 1.7152564503254753e-06}
{"train_loss": 0.016516992822289467, "global_step": 248794, "epoch": 2795, "lr": 1.7151059078895138e-06}
{"train_loss": 0.02802484668791294, "global_step": 248795, "epoch": 2795, "lr": 1.7149553719448596e-06}
{"train_loss": 0.07698046416044235, "global_step": 248796, "epoch": 2795, "lr": 1.7148048424915464e-06}
{"train_loss": 0.05155694857239723, "global_step": 248797, "epoch": 2795, "lr": 1.714654319529585e-06}
{"train_loss": 0.030785726383328438, "global_step": 248798, "epoch": 2795, "lr": 1.7145038030590032e-06}
{"train_loss": 0.02533041685819626, "global_step": 248799, "epoch": 2795, "lr": 1.7143532930798122e-06}
{"train_loss": 0.03615086153149605, "global_step": 248800, "epoch": 2795, "lr": 1.7142027895920454e-06}
{"train_loss": 0.056342530995607376, "global_step": 248801, "epoch": 2795, "lr": 1.714052292595708e-06}
{"train_loss": 0.051765862852334976, "global_step": 248802, "epoch": 2795, "lr": 1.7139018020908392e-06}
{"train_loss": 0.0673457533121109, "global_step": 248803, "epoch": 2795, "lr": 1.7137513180774334e-06}
{"train_loss": 0.029751598834991455, "global_step": 248804, "epoch": 2795, "lr": 1.7136008405555404e-06}
{"train_loss": 0.11109041422605515, "global_step": 248805, "epoch": 2795, "lr": 1.7134503695251548e-06}
{"train_loss": 0.037810392677783966, "global_step": 248806, "epoch": 2795, "lr": 1.7132999049863096e-06}
{"train_loss": 0.022735562175512314, "global_step": 248807, "epoch": 2795, "lr": 1.7131494469390275e-06}
{"train_loss": 0.037978630512952805, "global_step": 248808, "epoch": 2795, "lr": 1.7129989953833138e-06}
{"train_loss": 0.04671536758542061, "global_step": 248809, "epoch": 2795, "lr": 1.7128485503192072e-06}
{"train_loss": 0.045800406485795975, "global_step": 248810, "epoch": 2795, "lr": 1.7126981117467133e-06}
{"train_loss": 0.03029460273683071, "global_step": 248811, "epoch": 2795, "lr": 1.7125476796658657e-06}
{"train_loss": 0.03597400709986687, "global_step": 248812, "epoch": 2795, "lr": 1.7123972540766697e-06}
{"train_loss": 0.02527538314461708, "global_step": 248813, "epoch": 2795, "lr": 1.7122468349791643e-06}
{"train_loss": 0.07341935485601425, "global_step": 248814, "epoch": 2795, "lr": 1.7120964223733493e-06}
{"train_loss": 0.07827528566122055, "global_step": 248815, "epoch": 2795, "lr": 1.7119460162592583e-06}
{"train_loss": 0.043377794325351715, "global_step": 248816, "epoch": 2795, "lr": 1.7117956166369021e-06}
{"train_loss": 0.043543778359889984, "global_step": 248817, "epoch": 2795, "lr": 1.711645223506314e-06}
{"train_loss": 0.038053493946790695, "global_step": 248818, "epoch": 2795, "lr": 1.7114948368675e-06}
{"train_loss": 0.039795882999897, "global_step": 248819, "epoch": 2795, "lr": 1.7113444567204927e-06}
{"train_loss": 0.05930742248892784, "global_step": 248820, "epoch": 2795, "lr": 1.711194083065304e-06}
{"train_loss": 0.06023726984858513, "global_step": 248821, "epoch": 2795, "lr": 1.711043715901961e-06}
{"train_loss": 0.06125421077013016, "global_step": 248822, "epoch": 2795, "lr": 1.7108933552304695e-06}
{"train_loss": 0.0529833547770977, "global_step": 248823, "epoch": 2795, "lr": 1.7107430010508686e-06}
{"train_loss": 0.032763078808784485, "global_step": 248824, "epoch": 2795, "lr": 1.7105926533631634e-06}
{"train_loss": 0.06486032158136368, "global_step": 248825, "epoch": 2795, "lr": 1.7104423121673819e-06}
{"train_loss": 0.017788277938961983, "global_step": 248826, "epoch": 2795, "lr": 1.7102919774635462e-06}
{"train_loss": 0.05857324227690697, "global_step": 248827, "epoch": 2795, "lr": 1.7101416492516731e-06}
{"train_loss": 0.04073217138648033, "global_step": 248828, "epoch": 2795, "lr": 1.7099913275317848e-06}
{"train_loss": 0.06114692986011505, "global_step": 248829, "epoch": 2795, "lr": 1.7098410123038922e-06}
{"train_loss": 0.030184799805283546, "global_step": 248830, "epoch": 2795, "lr": 1.7096907035680287e-06}
{"train_loss": 0.0519142784178257, "global_step": 248831, "epoch": 2795, "lr": 1.7095404013242055e-06}
{"train_loss": 0.027086561545729637, "global_step": 248832, "epoch": 2795, "lr": 1.7093901055724504e-06}
{"train_loss": 0.06467238813638687, "global_step": 248833, "epoch": 2795, "lr": 1.7092398163127742e-06}
{"train_loss": 0.03423650562763214, "global_step": 248834, "epoch": 2795, "lr": 1.7090895335452052e-06}
{"train_loss": 0.034395597875118256, "global_step": 248835, "epoch": 2795, "lr": 1.7089392572697593e-06}
{"train_loss": 0.035861290991306305, "global_step": 248836, "epoch": 2795, "lr": 1.7087889874864593e-06}
{"train_loss": 0.03633412718772888, "global_step": 248837, "epoch": 2795, "lr": 1.7086387241953217e-06}
{"train_loss": 0.027118127793073654, "global_step": 248838, "epoch": 2795, "lr": 1.7084884673963742e-06}
{"train_loss": 0.055445510894060135, "global_step": 248839, "epoch": 2795, "lr": 1.7083382170896223e-06}
{"train_loss": 0.023051658645272255, "global_step": 248840, "epoch": 2795, "lr": 1.708187973275105e-06}
{"train_loss": 0.0727616474032402, "global_step": 248841, "epoch": 2795, "lr": 1.7080377359528278e-06}
{"train_loss": 0.015018188394606113, "global_step": 248842, "epoch": 2795, "lr": 1.7078875051228239e-06}
{"train_loss": 0.04592033512262481, "global_step": 248843, "epoch": 2795, "lr": 1.7077372807850933e-06, "val_loss": 9.538484573364258, "train_action_mse_error": 24.898399353027344}
{"train_loss": 0.07755068689584732, "global_step": 248844, "epoch": 2796, "lr": 1.7075870629396806e-06}
{"train_loss": 0.013723092153668404, "global_step": 248845, "epoch": 2796, "lr": 1.7074368515865912e-06}
{"train_loss": 0.031427767127752304, "global_step": 248846, "epoch": 2796, "lr": 1.7072866467258475e-06}
{"train_loss": 0.050724975764751434, "global_step": 248847, "epoch": 2796, "lr": 1.7071364483574714e-06}
{"train_loss": 0.044009894132614136, "global_step": 248848, "epoch": 2796, "lr": 1.7069862564814743e-06}
{"train_loss": 0.06642497330904007, "global_step": 248849, "epoch": 2796, "lr": 1.7068360710978948e-06}
{"train_loss": 0.08202652633190155, "global_step": 248850, "epoch": 2796, "lr": 1.706685892206733e-06}
{"train_loss": 0.03555355593562126, "global_step": 248851, "epoch": 2796, "lr": 1.7065357198080279e-06}
{"train_loss": 0.07869090139865875, "global_step": 248852, "epoch": 2796, "lr": 1.7063855539017848e-06}
{"train_loss": 0.0525931678712368, "global_step": 248853, "epoch": 2796, "lr": 1.7062353944880315e-06}
{"train_loss": 0.03661009669303894, "global_step": 248854, "epoch": 2796, "lr": 1.7060852415667849e-06}
{"train_loss": 0.02746177650988102, "global_step": 248855, "epoch": 2796, "lr": 1.7059350951380725e-06}
{"train_loss": 0.02640189230442047, "global_step": 248856, "epoch": 2796, "lr": 1.7057849552019e-06}
{"train_loss": 0.06341343373060226, "global_step": 248857, "epoch": 2796, "lr": 1.7056348217583008e-06}
{"train_loss": 0.06920024752616882, "global_step": 248858, "epoch": 2796, "lr": 1.7054846948072855e-06}
{"train_loss": 0.06183357909321785, "global_step": 248859, "epoch": 2796, "lr": 1.705334574348888e-06}
{"train_loss": 0.04789939522743225, "global_step": 248860, "epoch": 2796, "lr": 1.7051844603831136e-06}
{"train_loss": 0.06748057901859283, "global_step": 248861, "epoch": 2796, "lr": 1.7050343529099843e-06}
{"train_loss": 0.043525416404008865, "global_step": 248862, "epoch": 2796, "lr": 1.7048842519295339e-06}
{"train_loss": 0.04777540639042854, "global_step": 248863, "epoch": 2796, "lr": 1.7047341574417619e-06}
{"train_loss": 0.072331503033638, "global_step": 248864, "epoch": 2796, "lr": 1.704584069446713e-06}
{"train_loss": 0.06371448934078217, "global_step": 248865, "epoch": 2796, "lr": 1.7044339879443817e-06}
{"train_loss": 0.04952135682106018, "global_step": 248866, "epoch": 2796, "lr": 1.704283912934812e-06}
{"train_loss": 0.023623542860150337, "global_step": 248867, "epoch": 2796, "lr": 1.7041338444180045e-06}
{"train_loss": 0.01737254485487938, "global_step": 248868, "epoch": 2796, "lr": 1.703983782393992e-06}
{"train_loss": 0.04828760772943497, "global_step": 248869, "epoch": 2796, "lr": 1.7038337268627857e-06}
{"train_loss": 0.07860825210809708, "global_step": 248870, "epoch": 2796, "lr": 1.7036836778244191e-06}
{"train_loss": 0.045841705054044724, "global_step": 248871, "epoch": 2796, "lr": 1.7035336352788922e-06}
{"train_loss": 0.06589191406965256, "global_step": 248872, "epoch": 2796, "lr": 1.7033835992262437e-06}
{"train_loss": 0.04385100677609444, "global_step": 248873, "epoch": 2796, "lr": 1.7032335696664847e-06}
{"train_loss": 0.05380461364984512, "global_step": 248874, "epoch": 2796, "lr": 1.703083546599643e-06}
{"train_loss": 0.023154471069574356, "global_step": 248875, "epoch": 2796, "lr": 1.7029335300257243e-06}
{"train_loss": 0.034808121621608734, "global_step": 248876, "epoch": 2796, "lr": 1.7027835199447618e-06}
{"train_loss": 0.057440489530563354, "global_step": 248877, "epoch": 2796, "lr": 1.7026335163567663e-06}
{"train_loss": 0.02154264785349369, "global_step": 248878, "epoch": 2796, "lr": 1.7024835192617716e-06}
{"train_loss": 0.02210286445915699, "global_step": 248879, "epoch": 2796, "lr": 1.702333528659783e-06}
{"train_loss": 0.03589450567960739, "global_step": 248880, "epoch": 2796, "lr": 1.7021835445508338e-06}
{"train_loss": 0.06916847825050354, "global_step": 248881, "epoch": 2796, "lr": 1.7020335669349297e-06}
{"train_loss": 0.02389718033373356, "global_step": 248882, "epoch": 2796, "lr": 1.7018835958121037e-06}
{"train_loss": 0.03226618096232414, "global_step": 248883, "epoch": 2796, "lr": 1.7017336311823618e-06}
{"train_loss": 0.05152449011802673, "global_step": 248884, "epoch": 2796, "lr": 1.7015836730457424e-06}
{"train_loss": 0.053930751979351044, "global_step": 248885, "epoch": 2796, "lr": 1.701433721402257e-06}
{"train_loss": 0.023776741698384285, "global_step": 248886, "epoch": 2796, "lr": 1.7012837762519217e-06}
{"train_loss": 0.03890014812350273, "global_step": 248887, "epoch": 2796, "lr": 1.7011338375947649e-06}
{"train_loss": 0.026058003306388855, "global_step": 248888, "epoch": 2796, "lr": 1.700983905430792e-06}
{"train_loss": 0.07309279590845108, "global_step": 248889, "epoch": 2796, "lr": 1.7008339797600414e-06}
{"train_loss": 0.02535465732216835, "global_step": 248890, "epoch": 2796, "lr": 1.7006840605825192e-06}
{"train_loss": 0.08633790165185928, "global_step": 248891, "epoch": 2796, "lr": 1.7005341478982583e-06}
{"train_loss": 0.07235264033079147, "global_step": 248892, "epoch": 2796, "lr": 1.7003842417072646e-06}
{"train_loss": 0.02661276049911976, "global_step": 248893, "epoch": 2796, "lr": 1.7002343420095712e-06}
{"train_loss": 0.012519222684204578, "global_step": 248894, "epoch": 2796, "lr": 1.7000844488051893e-06}
{"train_loss": 0.09726446866989136, "global_step": 248895, "epoch": 2796, "lr": 1.6999345620941465e-06}
{"train_loss": 0.0263518113642931, "global_step": 248896, "epoch": 2796, "lr": 1.699784681876454e-06}
{"train_loss": 0.05286290496587753, "global_step": 248897, "epoch": 2796, "lr": 1.6996348081521397e-06}
{"train_loss": 0.04262460768222809, "global_step": 248898, "epoch": 2796, "lr": 1.6994849409212144e-06}
{"train_loss": 0.019256092607975006, "global_step": 248899, "epoch": 2796, "lr": 1.699335080183706e-06}
{"train_loss": 0.09743968397378922, "global_step": 248900, "epoch": 2796, "lr": 1.6991852259396423e-06}
{"train_loss": 0.043157909065485, "global_step": 248901, "epoch": 2796, "lr": 1.6990353781890234e-06}
{"train_loss": 0.07838504761457443, "global_step": 248902, "epoch": 2796, "lr": 1.6988855369318934e-06}
{"train_loss": 0.034166838973760605, "global_step": 248903, "epoch": 2796, "lr": 1.698735702168247e-06}
{"train_loss": 0.05918211117386818, "global_step": 248904, "epoch": 2796, "lr": 1.6985858738981285e-06}
{"train_loss": 0.03862062841653824, "global_step": 248905, "epoch": 2796, "lr": 1.6984360521215325e-06}
{"train_loss": 0.07067608088254929, "global_step": 248906, "epoch": 2796, "lr": 1.6982862368385033e-06}
{"train_loss": 0.013691172003746033, "global_step": 248907, "epoch": 2796, "lr": 1.6981364280490463e-06}
{"train_loss": 0.04597918689250946, "global_step": 248908, "epoch": 2796, "lr": 1.6979866257531952e-06}
{"train_loss": 0.038503795862197876, "global_step": 248909, "epoch": 2796, "lr": 1.6978368299509494e-06}
{"train_loss": 0.046572837978601456, "global_step": 248910, "epoch": 2796, "lr": 1.6976870406423483e-06}
{"train_loss": 0.0350986048579216, "global_step": 248911, "epoch": 2796, "lr": 1.6975372578273974e-06}
{"train_loss": 0.0160947535187006, "global_step": 248912, "epoch": 2796, "lr": 1.6973874815061298e-06}
{"train_loss": 0.0394010990858078, "global_step": 248913, "epoch": 2796, "lr": 1.6972377116785565e-06}
{"train_loss": 0.05962651968002319, "global_step": 248914, "epoch": 2796, "lr": 1.6970879483447055e-06}
{"train_loss": 0.06548977643251419, "global_step": 248915, "epoch": 2796, "lr": 1.6969381915045879e-06}
{"train_loss": 0.033897705376148224, "global_step": 248916, "epoch": 2796, "lr": 1.6967884411582259e-06}
{"train_loss": 0.02724076248705387, "global_step": 248917, "epoch": 2796, "lr": 1.696638697305647e-06}
{"train_loss": 0.046513207256793976, "global_step": 248918, "epoch": 2796, "lr": 1.6964889599468681e-06}
{"train_loss": 0.04099252447485924, "global_step": 248919, "epoch": 2796, "lr": 1.6963392290819003e-06}
{"train_loss": 0.022483816370368004, "global_step": 248920, "epoch": 2796, "lr": 1.6961895047107767e-06}
{"train_loss": 0.06190470606088638, "global_step": 248921, "epoch": 2796, "lr": 1.6960397868335086e-06}
{"train_loss": 0.04211015626788139, "global_step": 248922, "epoch": 2796, "lr": 1.695890075450124e-06}
{"train_loss": 0.03728746250271797, "global_step": 248923, "epoch": 2796, "lr": 1.6957403705606334e-06}
{"train_loss": 0.04695922136306763, "global_step": 248924, "epoch": 2796, "lr": 1.6955906721650649e-06}
{"train_loss": 0.04566100984811783, "global_step": 248925, "epoch": 2796, "lr": 1.6954409802634352e-06}
{"train_loss": 0.06980221718549728, "global_step": 248926, "epoch": 2796, "lr": 1.6952912948557608e-06}
{"train_loss": 0.07941458374261856, "global_step": 248927, "epoch": 2796, "lr": 1.6951416159420696e-06}
{"train_loss": 0.05146017670631409, "global_step": 248928, "epoch": 2796, "lr": 1.6949919435223726e-06}
{"train_loss": 0.016739264130592346, "global_step": 248929, "epoch": 2796, "lr": 1.6948422775967032e-06}
{"train_loss": 0.03875629976391792, "global_step": 248930, "epoch": 2796, "lr": 1.6946926181650668e-06}
{"train_loss": 0.037682920694351196, "global_step": 248931, "epoch": 2796, "lr": 1.6945429652274914e-06}
{"train_loss": 0.04691955586372132, "global_step": 248932, "epoch": 2796, "lr": 1.6943933187839988e-06, "val_loss": 9.478355407714844}
{"train_loss": 0.01791549101471901, "global_step": 248933, "epoch": 2797, "lr": 1.6942436788346062e-06}
{"train_loss": 0.03681878000497818, "global_step": 248934, "epoch": 2797, "lr": 1.6940940453793296e-06}
{"train_loss": 0.031939249485731125, "global_step": 248935, "epoch": 2797, "lr": 1.6939444184181918e-06}
{"train_loss": 0.05745190754532814, "global_step": 248936, "epoch": 2797, "lr": 1.6937947979512204e-06}
{"train_loss": 0.035701099783182144, "global_step": 248937, "epoch": 2797, "lr": 1.6936451839784263e-06}
{"train_loss": 0.05371994525194168, "global_step": 248938, "epoch": 2797, "lr": 1.6934955764998373e-06}
{"train_loss": 0.07476607710123062, "global_step": 248939, "epoch": 2797, "lr": 1.6933459755154591e-06}
{"train_loss": 0.06700711697340012, "global_step": 248940, "epoch": 2797, "lr": 1.6931963810253359e-06}
{"train_loss": 0.0441727340221405, "global_step": 248941, "epoch": 2797, "lr": 1.6930467930294625e-06}
{"train_loss": 0.07218877971172333, "global_step": 248942, "epoch": 2797, "lr": 1.6928972115278774e-06}
{"train_loss": 0.05251147225499153, "global_step": 248943, "epoch": 2797, "lr": 1.6927476365205864e-06}
{"train_loss": 0.0910874456167221, "global_step": 248944, "epoch": 2797, "lr": 1.6925980680076225e-06}
{"train_loss": 0.046459272503852844, "global_step": 248945, "epoch": 2797, "lr": 1.6924485059889971e-06}
{"train_loss": 0.056276943534612656, "global_step": 248946, "epoch": 2797, "lr": 1.6922989504647323e-06}
{"train_loss": 0.034318506717681885, "global_step": 248947, "epoch": 2797, "lr": 1.6921494014348504e-06}
{"train_loss": 0.020097831264138222, "global_step": 248948, "epoch": 2797, "lr": 1.6919998588993734e-06}
{"train_loss": 0.016945987939834595, "global_step": 248949, "epoch": 2797, "lr": 1.6918503228583127e-06}
{"train_loss": 0.03833852708339691, "global_step": 248950, "epoch": 2797, "lr": 1.6917007933117012e-06}
{"train_loss": 0.06133245676755905, "global_step": 248951, "epoch": 2797, "lr": 1.6915512702595392e-06}
{"train_loss": 0.02647930383682251, "global_step": 248952, "epoch": 2797, "lr": 1.6914017537018656e-06}
{"train_loss": 0.06109190732240677, "global_step": 248953, "epoch": 2797, "lr": 1.6912522436387023e-06}
{"train_loss": 0.026580482721328735, "global_step": 248954, "epoch": 2797, "lr": 1.6911027400700496e-06}
{"train_loss": 0.04357939958572388, "global_step": 248955, "epoch": 2797, "lr": 1.6909532429959462e-06}
{"train_loss": 0.08552674204111099, "global_step": 248956, "epoch": 2797, "lr": 1.6908037524164088e-06}
{"train_loss": 0.020616983994841576, "global_step": 248957, "epoch": 2797, "lr": 1.690654268331443e-06}
{"train_loss": 0.003877088660374284, "global_step": 248958, "epoch": 2797, "lr": 1.6905047907410932e-06}
{"train_loss": 0.028140151873230934, "global_step": 248959, "epoch": 2797, "lr": 1.6903553196453537e-06}
{"train_loss": 0.04737629368901253, "global_step": 248960, "epoch": 2797, "lr": 1.6902058550442635e-06}
{"train_loss": 0.03215306997299194, "global_step": 248961, "epoch": 2797, "lr": 1.6900563969378335e-06}
{"train_loss": 0.0133822001516819, "global_step": 248962, "epoch": 2797, "lr": 1.6899069453260918e-06}
{"train_loss": 0.02216927707195282, "global_step": 248963, "epoch": 2797, "lr": 1.6897575002090548e-06}
{"train_loss": 0.0619945265352726, "global_step": 248964, "epoch": 2797, "lr": 1.6896080615867338e-06}
{"train_loss": 0.06566018611192703, "global_step": 248965, "epoch": 2797, "lr": 1.689458629459162e-06}
{"train_loss": 0.03233473002910614, "global_step": 248966, "epoch": 2797, "lr": 1.6893092038263447e-06}
{"train_loss": 0.02317439578473568, "global_step": 248967, "epoch": 2797, "lr": 1.689159784688321e-06}
{"train_loss": 0.08301205933094025, "global_step": 248968, "epoch": 2797, "lr": 1.6890103720450911e-06}
{"train_loss": 0.05764910206198692, "global_step": 248969, "epoch": 2797, "lr": 1.6888609658966936e-06}
{"train_loss": 0.05584206432104111, "global_step": 248970, "epoch": 2797, "lr": 1.688711566243134e-06}
{"train_loss": 0.03895067423582077, "global_step": 248971, "epoch": 2797, "lr": 1.6885621730844347e-06}
{"train_loss": 0.046187836676836014, "global_step": 248972, "epoch": 2797, "lr": 1.688412786420629e-06}
{"train_loss": 0.044095057994127274, "global_step": 248973, "epoch": 2797, "lr": 1.6882634062517221e-06}
{"train_loss": 0.039224352687597275, "global_step": 248974, "epoch": 2797, "lr": 1.6881140325777422e-06}
{"train_loss": 0.04682613164186478, "global_step": 248975, "epoch": 2797, "lr": 1.6879646653987058e-06}
{"train_loss": 0.07747533917427063, "global_step": 248976, "epoch": 2797, "lr": 1.687815304714635e-06}
{"train_loss": 0.018989525735378265, "global_step": 248977, "epoch": 2797, "lr": 1.6876659505255466e-06}
{"train_loss": 0.04603899270296097, "global_step": 248978, "epoch": 2797, "lr": 1.6875166028314626e-06}
{"train_loss": 0.04436377063393593, "global_step": 248979, "epoch": 2797, "lr": 1.6873672616324e-06}
{"train_loss": 0.03520214185118675, "global_step": 248980, "epoch": 2797, "lr": 1.6872179269283917e-06}
{"train_loss": 0.04711850360035896, "global_step": 248981, "epoch": 2797, "lr": 1.687068598719438e-06}
{"train_loss": 0.01478128507733345, "global_step": 248982, "epoch": 2797, "lr": 1.6869192770055775e-06}
{"train_loss": 0.07860688120126724, "global_step": 248983, "epoch": 2797, "lr": 1.6867699617868105e-06}
{"train_loss": 0.04227285832166672, "global_step": 248984, "epoch": 2797, "lr": 1.6866206530631812e-06}
{"train_loss": 0.04526004195213318, "global_step": 248985, "epoch": 2797, "lr": 1.6864713508346842e-06}
{"train_loss": 0.05026792734861374, "global_step": 248986, "epoch": 2797, "lr": 1.6863220551013637e-06}
{"train_loss": 0.06922749429941177, "global_step": 248987, "epoch": 2797, "lr": 1.6861727658632198e-06}
{"train_loss": 0.0238809697329998, "global_step": 248988, "epoch": 2797, "lr": 1.6860234831202859e-06}
{"train_loss": 0.07666096091270447, "global_step": 248989, "epoch": 2797, "lr": 1.6858742068725787e-06}
{"train_loss": 0.03391004353761673, "global_step": 248990, "epoch": 2797, "lr": 1.685724937120109e-06}
{"train_loss": 0.06205250322818756, "global_step": 248991, "epoch": 2797, "lr": 1.6855756738629158e-06}
{"train_loss": 0.06881201267242432, "global_step": 248992, "epoch": 2797, "lr": 1.6854264171009991e-06}
{"train_loss": 0.042765527963638306, "global_step": 248993, "epoch": 2797, "lr": 1.6852771668343924e-06}
{"train_loss": 0.06621533632278442, "global_step": 248994, "epoch": 2797, "lr": 1.6851279230631066e-06}
{"train_loss": 0.04950716719031334, "global_step": 248995, "epoch": 2797, "lr": 1.6849786857871753e-06}
{"train_loss": 0.08445736765861511, "global_step": 248996, "epoch": 2797, "lr": 1.6848294550066091e-06}
{"train_loss": 0.02065899968147278, "global_step": 248997, "epoch": 2797, "lr": 1.6846802307214192e-06}
{"train_loss": 0.09751815348863602, "global_step": 248998, "epoch": 2797, "lr": 1.684531012931645e-06}
{"train_loss": 0.011850147508084774, "global_step": 248999, "epoch": 2797, "lr": 1.6843818016372859e-06}
{"train_loss": 0.03246565908193588, "global_step": 249000, "epoch": 2797, "lr": 1.684232596838381e-06}
{"train_loss": 0.051709022372961044, "global_step": 249001, "epoch": 2797, "lr": 1.6840833985349413e-06}
{"train_loss": 0.044465675950050354, "global_step": 249002, "epoch": 2797, "lr": 1.6839342067269837e-06}
{"train_loss": 0.01838058792054653, "global_step": 249003, "epoch": 2797, "lr": 1.6837850214145412e-06}
{"train_loss": 0.06056496128439903, "global_step": 249004, "epoch": 2797, "lr": 1.683635842597614e-06}
{"train_loss": 0.0711374431848526, "global_step": 249005, "epoch": 2797, "lr": 1.683486670276241e-06}
{"train_loss": 0.040797725319862366, "global_step": 249006, "epoch": 2797, "lr": 1.6833375044504273e-06}
{"train_loss": 0.08224353939294815, "global_step": 249007, "epoch": 2797, "lr": 1.6831883451202014e-06}
{"train_loss": 0.05772559344768524, "global_step": 249008, "epoch": 2797, "lr": 1.6830391922855849e-06}
{"train_loss": 0.020770784467458725, "global_step": 249009, "epoch": 2797, "lr": 1.6828900459465947e-06}
{"train_loss": 0.07003436982631683, "global_step": 249010, "epoch": 2797, "lr": 1.6827409061032528e-06}
{"train_loss": 0.037151891738176346, "global_step": 249011, "epoch": 2797, "lr": 1.6825917727555708e-06}
{"train_loss": 0.04845372959971428, "global_step": 249012, "epoch": 2797, "lr": 1.6824426459035813e-06}
{"train_loss": 0.045510802417993546, "global_step": 249013, "epoch": 2797, "lr": 1.682293525547296e-06}
{"train_loss": 0.06332242488861084, "global_step": 249014, "epoch": 2797, "lr": 1.6821444116867424e-06}
{"train_loss": 0.026476126164197922, "global_step": 249015, "epoch": 2797, "lr": 1.6819953043219262e-06}
{"train_loss": 0.04005807638168335, "global_step": 249016, "epoch": 2797, "lr": 1.681846203452886e-06}
{"train_loss": 0.028058944270014763, "global_step": 249017, "epoch": 2797, "lr": 1.6816971090796274e-06}
{"train_loss": 0.03346589207649231, "global_step": 249018, "epoch": 2797, "lr": 1.6815480212021784e-06}
{"train_loss": 0.039194505661726, "global_step": 249019, "epoch": 2797, "lr": 1.6813989398205553e-06}
{"train_loss": 0.03587209805846214, "global_step": 249020, "epoch": 2797, "lr": 1.6812498649347808e-06}
{"train_loss": 0.04595172084250561, "global_step": 249021, "epoch": 2797, "lr": 1.6811007965448656e-06, "val_loss": 9.519943237304688}
{"train_loss": 0.0628938153386116, "global_step": 249022, "epoch": 2798, "lr": 1.6809517346508486e-06}
{"train_loss": 0.05066254734992981, "global_step": 249023, "epoch": 2798, "lr": 1.6808026792527298e-06}
{"train_loss": 0.049082495272159576, "global_step": 249024, "epoch": 2798, "lr": 1.6806536303505427e-06}
{"train_loss": 0.025251343846321106, "global_step": 249025, "epoch": 2798, "lr": 1.6805045879442982e-06}
{"train_loss": 0.06873546540737152, "global_step": 249026, "epoch": 2798, "lr": 1.6803555520340242e-06}
{"train_loss": 0.015474211424589157, "global_step": 249027, "epoch": 2798, "lr": 1.6802065226197428e-06}
{"train_loss": 0.01483165007084608, "global_step": 249028, "epoch": 2798, "lr": 1.6800574997014594e-06}
{"train_loss": 0.046924374997615814, "global_step": 249029, "epoch": 2798, "lr": 1.6799084832792079e-06}
{"train_loss": 0.03355339914560318, "global_step": 249030, "epoch": 2798, "lr": 1.6797594733530042e-06}
{"train_loss": 0.029096191748976707, "global_step": 249031, "epoch": 2798, "lr": 1.679610469922871e-06}
{"train_loss": 0.019290102645754814, "global_step": 249032, "epoch": 2798, "lr": 1.6794614729888192e-06}
{"train_loss": 0.03532775118947029, "global_step": 249033, "epoch": 2798, "lr": 1.6793124825508767e-06}
{"train_loss": 0.04287038370966911, "global_step": 249034, "epoch": 2798, "lr": 1.6791634986090655e-06}
{"train_loss": 0.05530359596014023, "global_step": 249035, "epoch": 2798, "lr": 1.6790145211633967e-06}
{"train_loss": 0.019865458831191063, "global_step": 249036, "epoch": 2798, "lr": 1.6788655502138984e-06}
{"train_loss": 0.060809701681137085, "global_step": 249037, "epoch": 2798, "lr": 1.678716585760587e-06}
{"train_loss": 0.028273940086364746, "global_step": 249038, "epoch": 2798, "lr": 1.6785676278034845e-06}
{"train_loss": 0.05512571707367897, "global_step": 249039, "epoch": 2798, "lr": 1.678418676342608e-06}
{"train_loss": 0.037792425602674484, "global_step": 249040, "epoch": 2798, "lr": 1.6782697313779794e-06}
{"train_loss": 0.0551721416413784, "global_step": 249041, "epoch": 2798, "lr": 1.678120792909621e-06}
{"train_loss": 0.037248797714710236, "global_step": 249042, "epoch": 2798, "lr": 1.677971860937544e-06}
{"train_loss": 0.047570254653692245, "global_step": 249043, "epoch": 2798, "lr": 1.6778229354617758e-06}
{"train_loss": 0.05630015581846237, "global_step": 249044, "epoch": 2798, "lr": 1.677674016482339e-06}
{"train_loss": 0.05793697014451027, "global_step": 249045, "epoch": 2798, "lr": 1.6775251039992445e-06}
{"train_loss": 0.1053583100438118, "global_step": 249046, "epoch": 2798, "lr": 1.6773761980125258e-06}
{"train_loss": 0.08685796707868576, "global_step": 249047, "epoch": 2798, "lr": 1.6772272985221881e-06}
{"train_loss": 0.02219865284860134, "global_step": 249048, "epoch": 2798, "lr": 1.6770784055282651e-06}
{"train_loss": 0.040616802871227264, "global_step": 249049, "epoch": 2798, "lr": 1.6769295190307621e-06}
{"train_loss": 0.04341352730989456, "global_step": 249050, "epoch": 2798, "lr": 1.6767806390297126e-06}
{"train_loss": 0.02811776101589203, "global_step": 249051, "epoch": 2798, "lr": 1.6766317655251273e-06}
{"train_loss": 0.08414056897163391, "global_step": 249052, "epoch": 2798, "lr": 1.6764828985170344e-06}
{"train_loss": 0.05265836790204048, "global_step": 249053, "epoch": 2798, "lr": 1.676334038005445e-06}
{"train_loss": 0.08145838975906372, "global_step": 249054, "epoch": 2798, "lr": 1.6761851839903864e-06}
{"train_loss": 0.04444838687777519, "global_step": 249055, "epoch": 2798, "lr": 1.67603633647187e-06}
{"train_loss": 0.09333550930023193, "global_step": 249056, "epoch": 2798, "lr": 1.6758874954499293e-06}
{"train_loss": 0.03374236822128296, "global_step": 249057, "epoch": 2798, "lr": 1.6757386609245694e-06}
{"train_loss": 0.03995687514543533, "global_step": 249058, "epoch": 2798, "lr": 1.6755898328958242e-06}
{"train_loss": 0.015552803874015808, "global_step": 249059, "epoch": 2798, "lr": 1.6754410113637042e-06}
{"train_loss": 0.02558945119380951, "global_step": 249060, "epoch": 2798, "lr": 1.675292196328232e-06}
{"train_loss": 0.030669573694467545, "global_step": 249061, "epoch": 2798, "lr": 1.6751433877894296e-06}
{"train_loss": 0.04718047007918358, "global_step": 249062, "epoch": 2798, "lr": 1.6749945857473082e-06}
{"train_loss": 0.050979018211364746, "global_step": 249063, "epoch": 2798, "lr": 1.674845790201901e-06}
{"train_loss": 0.07120320945978165, "global_step": 249064, "epoch": 2798, "lr": 1.6746970011532192e-06}
{"train_loss": 0.014511636458337307, "global_step": 249065, "epoch": 2798, "lr": 1.6745482186012907e-06}
{"train_loss": 0.08968472480773926, "global_step": 249066, "epoch": 2798, "lr": 1.6743994425461208e-06}
{"train_loss": 0.06185488775372505, "global_step": 249067, "epoch": 2798, "lr": 1.6742506729877482e-06}
{"train_loss": 0.04229995980858803, "global_step": 249068, "epoch": 2798, "lr": 1.6741019099261789e-06}
{"train_loss": 0.04155227914452553, "global_step": 249069, "epoch": 2798, "lr": 1.6739531533614406e-06}
{"train_loss": 0.06518059223890305, "global_step": 249070, "epoch": 2798, "lr": 1.673804403293544e-06}
{"train_loss": 0.04103522002696991, "global_step": 249071, "epoch": 2798, "lr": 1.6736556597225229e-06}
{"train_loss": 0.07125894725322723, "global_step": 249072, "epoch": 2798, "lr": 1.6735069226483934e-06}
{"train_loss": 0.09991934150457382, "global_step": 249073, "epoch": 2798, "lr": 1.6733581920711617e-06}
{"train_loss": 0.05472516268491745, "global_step": 249074, "epoch": 2798, "lr": 1.6732094679908605e-06}
{"train_loss": 0.03063904121518135, "global_step": 249075, "epoch": 2798, "lr": 1.673060750407507e-06}
{"train_loss": 0.019849734380841255, "global_step": 249076, "epoch": 2798, "lr": 1.6729120393211227e-06}
{"train_loss": 0.028998147696256638, "global_step": 249077, "epoch": 2798, "lr": 1.6727633347317252e-06}
{"train_loss": 0.041834693402051926, "global_step": 249078, "epoch": 2798, "lr": 1.6726146366393413e-06}
{"train_loss": 0.0696486309170723, "global_step": 249079, "epoch": 2798, "lr": 1.6724659450439772e-06}
{"train_loss": 0.035692423582077026, "global_step": 249080, "epoch": 2798, "lr": 1.672317259945666e-06}
{"train_loss": 0.09591568261384964, "global_step": 249081, "epoch": 2798, "lr": 1.6721685813444187e-06}
{"train_loss": 0.08788034319877625, "global_step": 249082, "epoch": 2798, "lr": 1.6720199092402634e-06}
{"train_loss": 0.04945313185453415, "global_step": 249083, "epoch": 2798, "lr": 1.6718712436332107e-06}
{"train_loss": 0.042966578155756, "global_step": 249084, "epoch": 2798, "lr": 1.6717225845232942e-06}
{"train_loss": 0.052012450993061066, "global_step": 249085, "epoch": 2798, "lr": 1.6715739319105195e-06}
{"train_loss": 0.06513457745313644, "global_step": 249086, "epoch": 2798, "lr": 1.6714252857949143e-06}
{"train_loss": 0.0527062714099884, "global_step": 249087, "epoch": 2798, "lr": 1.671276646176495e-06}
{"train_loss": 0.03659608215093613, "global_step": 249088, "epoch": 2798, "lr": 1.6711280130552897e-06}
{"train_loss": 0.06563734263181686, "global_step": 249089, "epoch": 2798, "lr": 1.6709793864313038e-06}
{"train_loss": 0.028897840529680252, "global_step": 249090, "epoch": 2798, "lr": 1.6708307663045708e-06}
{"train_loss": 0.03186546266078949, "global_step": 249091, "epoch": 2798, "lr": 1.6706821526751016e-06}
{"train_loss": 0.023987673223018646, "global_step": 249092, "epoch": 2798, "lr": 1.6705335455429294e-06}
{"train_loss": 0.0701819360256195, "global_step": 249093, "epoch": 2798, "lr": 1.6703849449080544e-06}
{"train_loss": 0.07590597867965698, "global_step": 249094, "epoch": 2798, "lr": 1.6702363507705154e-06}
{"train_loss": 0.05930833891034126, "global_step": 249095, "epoch": 2798, "lr": 1.6700877631303125e-06}
{"train_loss": 0.04079083725810051, "global_step": 249096, "epoch": 2798, "lr": 1.6699391819874899e-06}
{"train_loss": 0.0914580225944519, "global_step": 249097, "epoch": 2798, "lr": 1.6697906073420478e-06}
{"train_loss": 0.0842515081167221, "global_step": 249098, "epoch": 2798, "lr": 1.6696420391940082e-06}
{"train_loss": 0.0359998494386673, "global_step": 249099, "epoch": 2798, "lr": 1.6694934775434102e-06}
{"train_loss": 0.05520731210708618, "global_step": 249100, "epoch": 2798, "lr": 1.669344922390248e-06}
{"train_loss": 0.04194009676575661, "global_step": 249101, "epoch": 2798, "lr": 1.6691963737345606e-06}
{"train_loss": 0.052430253475904465, "global_step": 249102, "epoch": 2798, "lr": 1.6690478315763536e-06}
{"train_loss": 0.044905420392751694, "global_step": 249103, "epoch": 2798, "lr": 1.66889929591566e-06}
{"train_loss": 0.040399208664894104, "global_step": 249104, "epoch": 2798, "lr": 1.6687507667524915e-06}
{"train_loss": 0.06881842762231827, "global_step": 249105, "epoch": 2798, "lr": 1.6686022440868754e-06}
{"train_loss": 0.05944555625319481, "global_step": 249106, "epoch": 2798, "lr": 1.6684537279188173e-06}
{"train_loss": 0.039300914853811264, "global_step": 249107, "epoch": 2798, "lr": 1.668305218248356e-06}
{"train_loss": 0.03227115049958229, "global_step": 249108, "epoch": 2798, "lr": 1.6681567150754918e-06}
{"train_loss": 0.04021633043885231, "global_step": 249109, "epoch": 2798, "lr": 1.6680082184002632e-06}
{"train_loss": 0.04985748800668824, "global_step": 249110, "epoch": 2798, "lr": 1.667859728222676e-06, "val_loss": 9.611991882324219}
{"train_loss": 0.07381568104028702, "global_step": 249111, "epoch": 2799, "lr": 1.6677112445427633e-06}
{"train_loss": 0.07594233751296997, "global_step": 249112, "epoch": 2799, "lr": 1.6675627673605365e-06}
{"train_loss": 0.03622496500611305, "global_step": 249113, "epoch": 2799, "lr": 1.667414296676012e-06}
{"train_loss": 0.05086002126336098, "global_step": 249114, "epoch": 2799, "lr": 1.6672658324892177e-06}
{"train_loss": 0.05246391519904137, "global_step": 249115, "epoch": 2799, "lr": 1.6671173748001645e-06}
{"train_loss": 0.08017026633024216, "global_step": 249116, "epoch": 2799, "lr": 1.666968923608886e-06}
{"train_loss": 0.028432445600628853, "global_step": 249117, "epoch": 2799, "lr": 1.6668204789153875e-06}
{"train_loss": 0.02614603377878666, "global_step": 249118, "epoch": 2799, "lr": 1.6666720407197079e-06}
{"train_loss": 0.0354565754532814, "global_step": 249119, "epoch": 2799, "lr": 1.6665236090218416e-06}
{"train_loss": 0.09506616741418839, "global_step": 249120, "epoch": 2799, "lr": 1.6663751838218277e-06}
{"train_loss": 0.042398083955049515, "global_step": 249121, "epoch": 2799, "lr": 1.6662267651196773e-06}
{"train_loss": 0.047318849712610245, "global_step": 249122, "epoch": 2799, "lr": 1.6660783529154233e-06}
{"train_loss": 0.04938523843884468, "global_step": 249123, "epoch": 2799, "lr": 1.665929947209066e-06}
{"train_loss": 0.04787388816475868, "global_step": 249124, "epoch": 2799, "lr": 1.6657815480006389e-06}
{"train_loss": 0.05323956906795502, "global_step": 249125, "epoch": 2799, "lr": 1.6656331552901582e-06}
{"train_loss": 0.07455562055110931, "global_step": 249126, "epoch": 2799, "lr": 1.6654847690776466e-06}
{"train_loss": 0.0875825360417366, "global_step": 249127, "epoch": 2799, "lr": 1.6653363893631147e-06}
{"train_loss": 0.05049272999167442, "global_step": 249128, "epoch": 2799, "lr": 1.665188016146596e-06}
{"train_loss": 0.022633221000432968, "global_step": 249129, "epoch": 2799, "lr": 1.6650396494280963e-06}
{"train_loss": 0.03252232074737549, "global_step": 249130, "epoch": 2799, "lr": 1.664891289207654e-06}
{"train_loss": 0.04215191677212715, "global_step": 249131, "epoch": 2799, "lr": 1.664742935485264e-06}
{"train_loss": 0.049486104398965836, "global_step": 249132, "epoch": 2799, "lr": 1.6645945882609704e-06}
{"train_loss": 0.03835992142558098, "global_step": 249133, "epoch": 2799, "lr": 1.6644462475347787e-06}
{"train_loss": 0.10171777009963989, "global_step": 249134, "epoch": 2799, "lr": 1.6642979133067115e-06}
{"train_loss": 0.0636649876832962, "global_step": 249135, "epoch": 2799, "lr": 1.6641495855767963e-06}
{"train_loss": 0.03583703935146332, "global_step": 249136, "epoch": 2799, "lr": 1.6640012643450387e-06}
{"train_loss": 0.031936369836330414, "global_step": 249137, "epoch": 2799, "lr": 1.6638529496114775e-06}
{"train_loss": 0.0887235552072525, "global_step": 249138, "epoch": 2799, "lr": 1.6637046413761125e-06}
{"train_loss": 0.027810897678136826, "global_step": 249139, "epoch": 2799, "lr": 1.663556339638983e-06}
{"train_loss": 0.04706468805670738, "global_step": 249140, "epoch": 2799, "lr": 1.6634080444000888e-06}
{"train_loss": 0.0384601429104805, "global_step": 249141, "epoch": 2799, "lr": 1.6632597556594688e-06}
{"train_loss": 0.0261551346629858, "global_step": 249142, "epoch": 2799, "lr": 1.6631114734171282e-06}
{"train_loss": 0.04955248907208443, "global_step": 249143, "epoch": 2799, "lr": 1.6629631976731009e-06}
{"train_loss": 0.03041698783636093, "global_step": 249144, "epoch": 2799, "lr": 1.662814928427392e-06}
{"train_loss": 0.057050690054893494, "global_step": 249145, "epoch": 2799, "lr": 1.6626666656800295e-06}
{"train_loss": 0.06778725236654282, "global_step": 249146, "epoch": 2799, "lr": 1.6625184094310298e-06}
{"train_loss": 0.03995932266116142, "global_step": 249147, "epoch": 2799, "lr": 1.6623701596804265e-06}
{"train_loss": 0.03591761365532875, "global_step": 249148, "epoch": 2799, "lr": 1.6622219164282137e-06}
{"train_loss": 0.026496725156903267, "global_step": 249149, "epoch": 2799, "lr": 1.6620736796744363e-06}
{"train_loss": 0.023394620046019554, "global_step": 249150, "epoch": 2799, "lr": 1.661925449419105e-06}
{"train_loss": 0.008315665647387505, "global_step": 249151, "epoch": 2799, "lr": 1.661777225662231e-06}
{"train_loss": 0.024275990203022957, "global_step": 249152, "epoch": 2799, "lr": 1.6616290084038478e-06}
{"train_loss": 0.016967928037047386, "global_step": 249153, "epoch": 2799, "lr": 1.6614807976439661e-06}
{"train_loss": 0.06130508705973625, "global_step": 249154, "epoch": 2799, "lr": 1.6613325933826141e-06}
{"train_loss": 0.033788491040468216, "global_step": 249155, "epoch": 2799, "lr": 1.661184395619797e-06}
{"train_loss": 0.07434327155351639, "global_step": 249156, "epoch": 2799, "lr": 1.6610362043555538e-06}
{"train_loss": 0.061229266226291656, "global_step": 249157, "epoch": 2799, "lr": 1.6608880195898902e-06}
{"train_loss": 0.025881104171276093, "global_step": 249158, "epoch": 2799, "lr": 1.6607398413228337e-06}
{"train_loss": 0.02447330206632614, "global_step": 249159, "epoch": 2799, "lr": 1.6605916695543955e-06}
{"train_loss": 0.04015228524804115, "global_step": 249160, "epoch": 2799, "lr": 1.6604435042846089e-06}
{"train_loss": 0.01191803626716137, "global_step": 249161, "epoch": 2799, "lr": 1.6602953455134796e-06}
{"train_loss": 0.03888997435569763, "global_step": 249162, "epoch": 2799, "lr": 1.660147193241046e-06}
{"train_loss": 0.037500735372304916, "global_step": 249163, "epoch": 2799, "lr": 1.6599990474673032e-06}
{"train_loss": 0.05754631757736206, "global_step": 249164, "epoch": 2799, "lr": 1.6598509081922953e-06}
{"train_loss": 0.04895463585853577, "global_step": 249165, "epoch": 2799, "lr": 1.6597027754160221e-06}
{"train_loss": 0.05200327932834625, "global_step": 249166, "epoch": 2799, "lr": 1.6595546491385173e-06}
{"train_loss": 0.026755549013614655, "global_step": 249167, "epoch": 2799, "lr": 1.6594065293597972e-06}
{"train_loss": 0.050499092787504196, "global_step": 249168, "epoch": 2799, "lr": 1.6592584160798785e-06}
{"train_loss": 0.03262526914477348, "global_step": 249169, "epoch": 2799, "lr": 1.6591103092987836e-06}
{"train_loss": 0.020104818046092987, "global_step": 249170, "epoch": 2799, "lr": 1.6589622090165292e-06}
{"train_loss": 0.04702059179544449, "global_step": 249171, "epoch": 2799, "lr": 1.6588141152331426e-06}
{"train_loss": 0.024210793897509575, "global_step": 249172, "epoch": 2799, "lr": 1.6586660279486354e-06}
{"train_loss": 0.06063838675618172, "global_step": 249173, "epoch": 2799, "lr": 1.6585179471630353e-06}
{"train_loss": 0.009001588448882103, "global_step": 249174, "epoch": 2799, "lr": 1.658369872876353e-06}
{"train_loss": 0.06602348387241364, "global_step": 249175, "epoch": 2799, "lr": 1.6582218050886222e-06}
{"train_loss": 0.027924006804823875, "global_step": 249176, "epoch": 2799, "lr": 1.6580737437998483e-06}
{"train_loss": 0.07679188251495361, "global_step": 249177, "epoch": 2799, "lr": 1.657925689010059e-06}
{"train_loss": 0.08502468466758728, "global_step": 249178, "epoch": 2799, "lr": 1.6577776407192714e-06}
{"train_loss": 0.028640562668442726, "global_step": 249179, "epoch": 2799, "lr": 1.657629598927507e-06}
{"train_loss": 0.02540816366672516, "global_step": 249180, "epoch": 2799, "lr": 1.657481563634783e-06}
{"train_loss": 0.04550277069211006, "global_step": 249181, "epoch": 2799, "lr": 1.6573335348411267e-06}
{"train_loss": 0.07145972549915314, "global_step": 249182, "epoch": 2799, "lr": 1.6571855125465497e-06}
{"train_loss": 0.029445068910717964, "global_step": 249183, "epoch": 2799, "lr": 1.6570374967510737e-06}
{"train_loss": 0.04703563079237938, "global_step": 249184, "epoch": 2799, "lr": 1.6568894874547214e-06}
{"train_loss": 0.05173090100288391, "global_step": 249185, "epoch": 2799, "lr": 1.6567414846575092e-06}
{"train_loss": 0.03833659365773201, "global_step": 249186, "epoch": 2799, "lr": 1.6565934883594592e-06}
{"train_loss": 0.014736868441104889, "global_step": 249187, "epoch": 2799, "lr": 1.6564454985605937e-06}
{"train_loss": 0.06667126715183258, "global_step": 249188, "epoch": 2799, "lr": 1.6562975152609295e-06}
{"train_loss": 0.04005620628595352, "global_step": 249189, "epoch": 2799, "lr": 1.6561495384604831e-06}
{"train_loss": 0.019994227215647697, "global_step": 249190, "epoch": 2799, "lr": 1.6560015681592823e-06}
{"train_loss": 0.022195953875780106, "global_step": 249191, "epoch": 2799, "lr": 1.6558536043573382e-06}
{"train_loss": 0.039228249341249466, "global_step": 249192, "epoch": 2799, "lr": 1.655705647054684e-06}
{"train_loss": 0.056623559445142746, "global_step": 249193, "epoch": 2799, "lr": 1.6555576962513197e-06}
{"train_loss": 0.10514800250530243, "global_step": 249194, "epoch": 2799, "lr": 1.6554097519472843e-06}
{"train_loss": 0.07928232848644257, "global_step": 249195, "epoch": 2799, "lr": 1.6552618141425891e-06}
{"train_loss": 0.05673253908753395, "global_step": 249196, "epoch": 2799, "lr": 1.6551138828372559e-06}
{"train_loss": 0.0496719554066658, "global_step": 249197, "epoch": 2799, "lr": 1.654965958031296e-06}
{"train_loss": 0.02270686998963356, "global_step": 249198, "epoch": 2799, "lr": 1.6548180397247425e-06}
{"train_loss": 0.04554008630763614, "global_step": 249199, "epoch": 2799, "lr": 1.6546701279176068e-06, "val_loss": 9.509968757629395}
{"train_loss": 0.059846315532922745, "global_step": 249200, "epoch": 2800, "lr": 1.6545222226099166e-06}
{"train_loss": 0.023893458768725395, "global_step": 249201, "epoch": 2800, "lr": 1.6543743238016774e-06}
{"train_loss": 0.03168825805187225, "global_step": 249202, "epoch": 2800, "lr": 1.6542264314929279e-06}
{"train_loss": 0.047423914074897766, "global_step": 249203, "epoch": 2800, "lr": 1.6540785456836682e-06}
{"train_loss": 0.04992629587650299, "global_step": 249204, "epoch": 2800, "lr": 1.6539306663739373e-06}
{"train_loss": 0.05300235003232956, "global_step": 249205, "epoch": 2800, "lr": 1.6537827935637408e-06}
{"train_loss": 0.04344170168042183, "global_step": 249206, "epoch": 2800, "lr": 1.653634927253106e-06}
{"train_loss": 0.035406798124313354, "global_step": 249207, "epoch": 2800, "lr": 1.6534870674420556e-06}
{"train_loss": 0.017117587849497795, "global_step": 249208, "epoch": 2800, "lr": 1.6533392141305948e-06}
{"train_loss": 0.051096633076667786, "global_step": 249209, "epoch": 2800, "lr": 1.6531913673187627e-06}
{"train_loss": 0.06877493858337402, "global_step": 249210, "epoch": 2800, "lr": 1.6530435270065646e-06}
{"train_loss": 0.0647478774189949, "global_step": 249211, "epoch": 2800, "lr": 1.6528956931940288e-06}
{"train_loss": 0.03657666593790054, "global_step": 249212, "epoch": 2800, "lr": 1.6527478658811658e-06}
{"train_loss": 0.04161527380347252, "global_step": 249213, "epoch": 2800, "lr": 1.652600045068009e-06}
{"train_loss": 0.024838637560606003, "global_step": 249214, "epoch": 2800, "lr": 1.6524522307545643e-06}
{"train_loss": 0.06592407822608948, "global_step": 249215, "epoch": 2800, "lr": 1.6523044229408647e-06}
{"train_loss": 0.0217377170920372, "global_step": 249216, "epoch": 2800, "lr": 1.6521566216269157e-06}
{"train_loss": 0.029454994946718216, "global_step": 249217, "epoch": 2800, "lr": 1.6520088268127508e-06}
{"train_loss": 0.0321902297437191, "global_step": 249218, "epoch": 2800, "lr": 1.6518610384983812e-06}
{"train_loss": 0.059630442410707474, "global_step": 249219, "epoch": 2800, "lr": 1.6517132566838345e-06}
{"train_loss": 0.0556923933327198, "global_step": 249220, "epoch": 2800, "lr": 1.6515654813691162e-06}
{"train_loss": 0.08169152587652206, "global_step": 249221, "epoch": 2800, "lr": 1.6514177125542652e-06}
{"train_loss": 0.044603582471609116, "global_step": 249222, "epoch": 2800, "lr": 1.651269950239287e-06}
{"train_loss": 0.05037052556872368, "global_step": 249223, "epoch": 2800, "lr": 1.6511221944242096e-06}
{"train_loss": 0.06306600570678711, "global_step": 249224, "epoch": 2800, "lr": 1.6509744451090437e-06}
{"train_loss": 0.033439554274082184, "global_step": 249225, "epoch": 2800, "lr": 1.6508267022938228e-06}
{"train_loss": 0.013318886049091816, "global_step": 249226, "epoch": 2800, "lr": 1.6506789659785526e-06}
{"train_loss": 0.0406741164624691, "global_step": 249227, "epoch": 2800, "lr": 1.6505312361632609e-06}
{"train_loss": 0.06088646501302719, "global_step": 249228, "epoch": 2800, "lr": 1.6503835128479694e-06}
{"train_loss": 0.03353935480117798, "global_step": 249229, "epoch": 2800, "lr": 1.6502357960326897e-06}
{"train_loss": 0.03892182931303978, "global_step": 249230, "epoch": 2800, "lr": 1.6500880857174494e-06}
{"train_loss": 0.02888491563498974, "global_step": 249231, "epoch": 2800, "lr": 1.6499403819022651e-06}
{"train_loss": 0.04033754765987396, "global_step": 249232, "epoch": 2800, "lr": 1.6497926845871592e-06}
{"train_loss": 0.05778409540653229, "global_step": 249233, "epoch": 2800, "lr": 1.6496449937721426e-06}
{"train_loss": 0.02291894145309925, "global_step": 249234, "epoch": 2800, "lr": 1.6494973094572486e-06}
{"train_loss": 0.06084928289055824, "global_step": 249235, "epoch": 2800, "lr": 1.6493496316424827e-06}
{"train_loss": 0.030071904882788658, "global_step": 249236, "epoch": 2800, "lr": 1.649201960327884e-06}
{"train_loss": 0.03258046507835388, "global_step": 249237, "epoch": 2800, "lr": 1.6490542955134524e-06}
{"train_loss": 0.07256289571523666, "global_step": 249238, "epoch": 2800, "lr": 1.6489066371992156e-06}
{"train_loss": 0.05101238191127777, "global_step": 249239, "epoch": 2800, "lr": 1.6487589853851958e-06}
{"train_loss": 0.05226321518421173, "global_step": 249240, "epoch": 2800, "lr": 1.6486113400714153e-06}
{"train_loss": 0.05700167641043663, "global_step": 249241, "epoch": 2800, "lr": 1.6484637012578797e-06}
{"train_loss": 0.06396062672138214, "global_step": 249242, "epoch": 2800, "lr": 1.6483160689446275e-06}
{"train_loss": 0.035943157970905304, "global_step": 249243, "epoch": 2800, "lr": 1.6481684431316701e-06}
{"train_loss": 0.06524521857500076, "global_step": 249244, "epoch": 2800, "lr": 1.6480208238190186e-06}
{"train_loss": 0.03184597194194794, "global_step": 249245, "epoch": 2800, "lr": 1.647873211006712e-06}
{"train_loss": 0.06867510080337524, "global_step": 249246, "epoch": 2800, "lr": 1.6477256046947498e-06}
{"train_loss": 0.029024925082921982, "global_step": 249247, "epoch": 2800, "lr": 1.6475780048831713e-06}
{"train_loss": 0.032477691769599915, "global_step": 249248, "epoch": 2800, "lr": 1.6474304115719763e-06}
{"train_loss": 0.04882354289293289, "global_step": 249249, "epoch": 2800, "lr": 1.6472828247612038e-06}
{"train_loss": 0.030198555439710617, "global_step": 249250, "epoch": 2800, "lr": 1.6471352444508591e-06}
{"train_loss": 0.08279058337211609, "global_step": 249251, "epoch": 2800, "lr": 1.646987670640976e-06}
{"train_loss": 0.03007315844297409, "global_step": 249252, "epoch": 2800, "lr": 1.6468401033315538e-06}
{"train_loss": 0.018596122041344643, "global_step": 249253, "epoch": 2800, "lr": 1.6466925425226321e-06}
{"train_loss": 0.08285266160964966, "global_step": 249254, "epoch": 2800, "lr": 1.6465449882142214e-06}
{"train_loss": 0.060190074145793915, "global_step": 249255, "epoch": 2800, "lr": 1.6463974404063497e-06}
{"train_loss": 0.021484166383743286, "global_step": 249256, "epoch": 2800, "lr": 1.6462498990990172e-06}
{"train_loss": 0.07420185208320618, "global_step": 249257, "epoch": 2800, "lr": 1.646102364292268e-06}
{"train_loss": 0.07083619385957718, "global_step": 249258, "epoch": 2800, "lr": 1.6459548359861077e-06}
{"train_loss": 0.10044267028570175, "global_step": 249259, "epoch": 2800, "lr": 1.6458073141805642e-06}
{"train_loss": 0.015965865924954414, "global_step": 249260, "epoch": 2800, "lr": 1.6456597988756428e-06}
{"train_loss": 0.023821163922548294, "global_step": 249261, "epoch": 2800, "lr": 1.6455122900713827e-06}
{"train_loss": 0.031088612973690033, "global_step": 249262, "epoch": 2800, "lr": 1.6453647877677835e-06}
{"train_loss": 0.013706483878195286, "global_step": 249263, "epoch": 2800, "lr": 1.6452172919648845e-06}
{"train_loss": 0.057604677975177765, "global_step": 249264, "epoch": 2800, "lr": 1.645069802662691e-06}
{"train_loss": 0.04145006090402603, "global_step": 249265, "epoch": 2800, "lr": 1.6449223198612363e-06}
{"train_loss": 0.04825080558657646, "global_step": 249266, "epoch": 2800, "lr": 1.6447748435605315e-06}
{"train_loss": 0.06821522116661072, "global_step": 249267, "epoch": 2800, "lr": 1.6446273737605877e-06}
{"train_loss": 0.04575579985976219, "global_step": 249268, "epoch": 2800, "lr": 1.644479910461444e-06}
{"train_loss": 0.030650615692138672, "global_step": 249269, "epoch": 2800, "lr": 1.6443324536631054e-06}
{"train_loss": 0.042236000299453735, "global_step": 249270, "epoch": 2800, "lr": 1.6441850033656004e-06}
{"train_loss": 0.030722225084900856, "global_step": 249271, "epoch": 2800, "lr": 1.6440375595689394e-06}
{"train_loss": 0.030622582882642746, "global_step": 249272, "epoch": 2800, "lr": 1.6438901222731506e-06}
{"train_loss": 0.022989176213741302, "global_step": 249273, "epoch": 2800, "lr": 1.6437426914782505e-06}
{"train_loss": 0.08859415352344513, "global_step": 249274, "epoch": 2800, "lr": 1.6435952671842668e-06}
{"train_loss": 0.049600549042224884, "global_step": 249275, "epoch": 2800, "lr": 1.6434478493912054e-06}
{"train_loss": 0.07074306160211563, "global_step": 249276, "epoch": 2800, "lr": 1.6433004380990992e-06}
{"train_loss": 0.014617405831813812, "global_step": 249277, "epoch": 2800, "lr": 1.6431530333079536e-06}
{"train_loss": 0.04944095388054848, "global_step": 249278, "epoch": 2800, "lr": 1.6430056350178025e-06}
{"train_loss": 0.030336027964949608, "global_step": 249279, "epoch": 2800, "lr": 1.6428582432286566e-06}
{"train_loss": 0.0719296783208847, "global_step": 249280, "epoch": 2800, "lr": 1.642710857940538e-06}
{"train_loss": 0.06279005110263824, "global_step": 249281, "epoch": 2800, "lr": 1.6425634791534749e-06}
{"train_loss": 0.0681580901145935, "global_step": 249282, "epoch": 2800, "lr": 1.6424161068674671e-06}
{"train_loss": 0.049716800451278687, "global_step": 249283, "epoch": 2800, "lr": 1.6422687410825588e-06}
{"train_loss": 0.058229707181453705, "global_step": 249284, "epoch": 2800, "lr": 1.6421213817987557e-06}
{"train_loss": 0.04450438171625137, "global_step": 249285, "epoch": 2800, "lr": 1.6419740290160801e-06}
{"train_loss": 0.047426361590623856, "global_step": 249286, "epoch": 2800, "lr": 1.6418266827345486e-06}
{"train_loss": 0.052946675568819046, "global_step": 249287, "epoch": 2800, "lr": 1.6416793429541889e-06}
{"train_loss": 0.04640707243861777, "global_step": 249288, "epoch": 2800, "lr": 1.6415320096750063e-06, "train/sim_max_reward_0": 0.49349579721930514, "train/sim_max_reward_1": 0.9794887129635318, "train/sim_max_reward_2": 0.6480029415184853, "train/sim_max_reward_3": 0.6883949890195573, "train/sim_max_reward_4": 0.9237773828754468, "train/sim_max_reward_5": 0.9591448526453403, "test/sim_max_reward_4500000": 0.41433120904562803, "test/sim_max_reward_4500001": 0.9703693169105897, "test/sim_max_reward_4500002": 0.0, "test/sim_max_reward_4500003": 0.16195438594400785, "test/sim_max_reward_4500004": 0.20213854121770689, "test/sim_max_reward_4500005": 0.3962107011898521, "test/sim_max_reward_4500006": 0.7750596600180256, "test/sim_max_reward_4500007": 0.9646235050047078, "test/sim_max_reward_4500008": 0.30629629403552605, "test/sim_max_reward_4500009": 0.05080250540718008, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5341155929133375, "test/sim_max_reward_4500014": 0.9766561642289244, "test/sim_max_reward_4500015": 0.9830133349355006, "test/sim_max_reward_4500016": 0.9285039369758115, "test/sim_max_reward_4500017": 0.6696409389858822, "test/sim_max_reward_4500018": 0.9818470546466048, "test/sim_max_reward_4500019": 0.9931939348640572, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9602337958650125, "test/sim_max_reward_4500022": 0.995137884815144, "test/sim_max_reward_4500023": 0.6436549140667202, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.39818367578748526, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6673967616380225, "test/sim_max_reward_4500029": 0.5853225024805226, "test/sim_max_reward_4500030": 0.8808446342672929, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6210428798395, "test/sim_max_reward_4500033": 0.9729834436973681, "test/sim_max_reward_4500034": 0.3981930998327364, "test/sim_max_reward_4500035": 0.5002572707272496, "test/sim_max_reward_4500036": 0.9971556359639331, "test/sim_max_reward_4500037": 1.0, "test/sim_max_reward_4500038": 0.1628869717221862, "test/sim_max_reward_4500039": 0.971257478774688, "test/sim_max_reward_4500040": 0.9752066086213722, "test/sim_max_reward_4500041": 0.3924771967310573, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9752330925706484, "test/sim_max_reward_4500044": 0.6181538467093708, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.4108774641263671, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.970365600153371, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7820507793736112, "test/mean_score": 0.6081124366942678, "val_loss": 9.562032699584961, "train_action_mse_error": 20.0693359375}
{"train_loss": 0.03187299519777298, "global_step": 249289, "epoch": 2801, "lr": 1.6413846828970403e-06}
{"train_loss": 0.028416156768798828, "global_step": 249290, "epoch": 2801, "lr": 1.6412373626202958e-06}
{"train_loss": 0.022728025913238525, "global_step": 249291, "epoch": 2801, "lr": 1.6410900488448012e-06}
{"train_loss": 0.09452932327985764, "global_step": 249292, "epoch": 2801, "lr": 1.6409427415705669e-06}
{"train_loss": 0.02990178018808365, "global_step": 249293, "epoch": 2801, "lr": 1.6407954407976268e-06}
{"train_loss": 0.034229595214128494, "global_step": 249294, "epoch": 2801, "lr": 1.6406481465259803e-06}
{"train_loss": 0.02910926192998886, "global_step": 249295, "epoch": 2801, "lr": 1.6405008587556726e-06}
{"train_loss": 0.05240271985530853, "global_step": 249296, "epoch": 2801, "lr": 1.6403535774866973e-06}
{"train_loss": 0.061773937195539474, "global_step": 249297, "epoch": 2801, "lr": 1.6402063027190939e-06}
{"train_loss": 0.022214673459529877, "global_step": 249298, "epoch": 2801, "lr": 1.6400590344528787e-06}
{"train_loss": 0.06974605470895767, "global_step": 249299, "epoch": 2801, "lr": 1.6399117726880687e-06}
{"train_loss": 0.04544910788536072, "global_step": 249300, "epoch": 2801, "lr": 1.6397645174246745e-06}
{"train_loss": 0.05502472445368767, "global_step": 249301, "epoch": 2801, "lr": 1.6396172686627297e-06}
{"train_loss": 0.07771304249763489, "global_step": 249302, "epoch": 2801, "lr": 1.6394700264022456e-06}
{"train_loss": 0.03054690547287464, "global_step": 249303, "epoch": 2801, "lr": 1.6393227906432495e-06}
{"train_loss": 0.04325580596923828, "global_step": 249304, "epoch": 2801, "lr": 1.6391755613857584e-06}
{"train_loss": 0.05360287427902222, "global_step": 249305, "epoch": 2801, "lr": 1.6390283386297889e-06}
{"train_loss": 0.0314079113304615, "global_step": 249306, "epoch": 2801, "lr": 1.638881122375363e-06}
{"train_loss": 0.043691862374544144, "global_step": 249307, "epoch": 2801, "lr": 1.6387339126224977e-06}
{"train_loss": 0.020283915102481842, "global_step": 249308, "epoch": 2801, "lr": 1.6385867093712148e-06}
{"train_loss": 0.04572401940822601, "global_step": 249309, "epoch": 2801, "lr": 1.638439512621537e-06}
{"train_loss": 0.07090510427951813, "global_step": 249310, "epoch": 2801, "lr": 1.6382923223734803e-06}
{"train_loss": 0.025082677602767944, "global_step": 249311, "epoch": 2801, "lr": 1.638145138627062e-06}
{"train_loss": 0.04807349294424057, "global_step": 249312, "epoch": 2801, "lr": 1.6379979613823148e-06}
{"train_loss": 0.020243797451257706, "global_step": 249313, "epoch": 2801, "lr": 1.6378507906392393e-06}
{"train_loss": 0.021938826888799667, "global_step": 249314, "epoch": 2801, "lr": 1.637703626397874e-06}
{"train_loss": 0.10754039883613586, "global_step": 249315, "epoch": 2801, "lr": 1.6375564686582245e-06}
{"train_loss": 0.054189980030059814, "global_step": 249316, "epoch": 2801, "lr": 1.6374093174203132e-06}
{"train_loss": 0.09135723859071732, "global_step": 249317, "epoch": 2801, "lr": 1.6372621726841675e-06}
{"train_loss": 0.026089057326316833, "global_step": 249318, "epoch": 2801, "lr": 1.6371150344497987e-06}
{"train_loss": 0.05232319235801697, "global_step": 249319, "epoch": 2801, "lr": 1.6369679027172402e-06}
{"train_loss": 0.04425077512860298, "global_step": 249320, "epoch": 2801, "lr": 1.636820777486492e-06}
{"train_loss": 0.03755856305360794, "global_step": 249321, "epoch": 2801, "lr": 1.636673658757587e-06}
{"train_loss": 0.055256739258766174, "global_step": 249322, "epoch": 2801, "lr": 1.6365265465305423e-06}
{"train_loss": 0.02644723281264305, "global_step": 249323, "epoch": 2801, "lr": 1.6363794408053802e-06}
{"train_loss": 0.03987371176481247, "global_step": 249324, "epoch": 2801, "lr": 1.6362323415821113e-06}
{"train_loss": 0.09445483982563019, "global_step": 249325, "epoch": 2801, "lr": 1.6360852488607637e-06}
{"train_loss": 0.03017246723175049, "global_step": 249326, "epoch": 2801, "lr": 1.6359381626413539e-06}
{"train_loss": 0.011583118699491024, "global_step": 249327, "epoch": 2801, "lr": 1.6357910829239043e-06}
{"train_loss": 0.10314996540546417, "global_step": 249328, "epoch": 2801, "lr": 1.6356440097084314e-06}
{"train_loss": 0.054749272763729095, "global_step": 249329, "epoch": 2801, "lr": 1.635496942994963e-06}
{"train_loss": 0.055349864065647125, "global_step": 249330, "epoch": 2801, "lr": 1.6353498827835046e-06}
{"train_loss": 0.05789677053689957, "global_step": 249331, "epoch": 2801, "lr": 1.6352028290740896e-06}
{"train_loss": 0.0462435707449913, "global_step": 249332, "epoch": 2801, "lr": 1.6350557818667288e-06}
{"train_loss": 0.07329826802015305, "global_step": 249333, "epoch": 2801, "lr": 1.6349087411614395e-06}
{"train_loss": 0.04007536917924881, "global_step": 249334, "epoch": 2801, "lr": 1.63476170695826e-06}
{"train_loss": 0.02581503428518772, "global_step": 249335, "epoch": 2801, "lr": 1.6346146792571848e-06}
{"train_loss": 0.05131971463561058, "global_step": 249336, "epoch": 2801, "lr": 1.6344676580582585e-06}
{"train_loss": 0.02819770947098732, "global_step": 249337, "epoch": 2801, "lr": 1.634320643361481e-06}
{"train_loss": 0.040470559149980545, "global_step": 249338, "epoch": 2801, "lr": 1.6341736351668802e-06}
{"train_loss": 0.04519544541835785, "global_step": 249339, "epoch": 2801, "lr": 1.6340266334744836e-06}
{"train_loss": 0.01440755371004343, "global_step": 249340, "epoch": 2801, "lr": 1.6338796382842913e-06}
{"train_loss": 0.052256789058446884, "global_step": 249341, "epoch": 2801, "lr": 1.6337326495963424e-06}
{"train_loss": 0.056300073862075806, "global_step": 249342, "epoch": 2801, "lr": 1.6335856674106364e-06}
{"train_loss": 0.03372099995613098, "global_step": 249343, "epoch": 2801, "lr": 1.6334386917272182e-06}
{"train_loss": 0.044878195971250534, "global_step": 249344, "epoch": 2801, "lr": 1.633291722546093e-06}
{"train_loss": 0.06201957166194916, "global_step": 249345, "epoch": 2801, "lr": 1.6331447598672778e-06}
{"train_loss": 0.029674069955945015, "global_step": 249346, "epoch": 2801, "lr": 1.6329978036908e-06}
{"train_loss": 0.025893688201904297, "global_step": 249347, "epoch": 2801, "lr": 1.632850854016671e-06}
{"train_loss": 0.02132387086749077, "global_step": 249348, "epoch": 2801, "lr": 1.6327039108449239e-06}
{"train_loss": 0.05267690494656563, "global_step": 249349, "epoch": 2801, "lr": 1.6325569741755586e-06}
{"train_loss": 0.048767175525426865, "global_step": 249350, "epoch": 2801, "lr": 1.6324100440086199e-06}
{"train_loss": 0.057739295065402985, "global_step": 249351, "epoch": 2801, "lr": 1.632263120344102e-06}
{"train_loss": 0.12964151799678802, "global_step": 249352, "epoch": 2801, "lr": 1.632116203182038e-06}
{"train_loss": 0.033238109201192856, "global_step": 249353, "epoch": 2801, "lr": 1.631969292522456e-06}
{"train_loss": 0.05656426399946213, "global_step": 249354, "epoch": 2801, "lr": 1.6318223883653561e-06}
{"train_loss": 0.05011940374970436, "global_step": 249355, "epoch": 2801, "lr": 1.6316754907107767e-06}
{"train_loss": 0.05578897148370743, "global_step": 249356, "epoch": 2801, "lr": 1.6315285995587237e-06}
{"train_loss": 0.040492571890354156, "global_step": 249357, "epoch": 2801, "lr": 1.6313817149092248e-06}
{"train_loss": 0.07417766004800797, "global_step": 249358, "epoch": 2801, "lr": 1.6312348367622965e-06}
{"train_loss": 0.04154398292303085, "global_step": 249359, "epoch": 2801, "lr": 1.6310879651179612e-06}
{"train_loss": 0.041413482278585434, "global_step": 249360, "epoch": 2801, "lr": 1.6309410999762299e-06}
{"train_loss": 0.012448590248823166, "global_step": 249361, "epoch": 2801, "lr": 1.6307942413371357e-06}
{"train_loss": 0.03853427618741989, "global_step": 249362, "epoch": 2801, "lr": 1.6306473892006845e-06}
{"train_loss": 0.05595012381672859, "global_step": 249363, "epoch": 2801, "lr": 1.6305005435669152e-06}
{"train_loss": 0.04764733836054802, "global_step": 249364, "epoch": 2801, "lr": 1.6303537044358218e-06}
{"train_loss": 0.033094484359025955, "global_step": 249365, "epoch": 2801, "lr": 1.6302068718074493e-06}
{"train_loss": 0.013305884785950184, "global_step": 249366, "epoch": 2801, "lr": 1.6300600456817972e-06}
{"train_loss": 0.046809446066617966, "global_step": 249367, "epoch": 2801, "lr": 1.629913226058899e-06}
{"train_loss": 0.037863168865442276, "global_step": 249368, "epoch": 2801, "lr": 1.6297664129387657e-06}
{"train_loss": 0.016905736178159714, "global_step": 249369, "epoch": 2801, "lr": 1.6296196063214198e-06}
{"train_loss": 0.02768130972981453, "global_step": 249370, "epoch": 2801, "lr": 1.6294728062068887e-06}
{"train_loss": 0.030778413638472557, "global_step": 249371, "epoch": 2801, "lr": 1.6293260125951782e-06}
{"train_loss": 0.042351897805929184, "global_step": 249372, "epoch": 2801, "lr": 1.6291792254863213e-06}
{"train_loss": 0.0442785918712616, "global_step": 249373, "epoch": 2801, "lr": 1.6290324448803295e-06}
{"train_loss": 0.0846283882856369, "global_step": 249374, "epoch": 2801, "lr": 1.6288856707772248e-06}
{"train_loss": 0.03307661786675453, "global_step": 249375, "epoch": 2801, "lr": 1.628738903177024e-06}
{"train_loss": 0.048096537590026855, "global_step": 249376, "epoch": 2801, "lr": 1.62859214207976e-06}
{"train_loss": 0.0457251217906897, "global_step": 249377, "epoch": 2801, "lr": 1.6284453874854333e-06, "val_loss": 9.557042121887207}
{"train_loss": 0.025066664442420006, "global_step": 249378, "epoch": 2802, "lr": 1.6282986393940713e-06}
{"train_loss": 0.027209410443902016, "global_step": 249379, "epoch": 2802, "lr": 1.628151897805702e-06}
{"train_loss": 0.05414101108908653, "global_step": 249380, "epoch": 2802, "lr": 1.6280051627203308e-06}
{"train_loss": 0.04977363348007202, "global_step": 249381, "epoch": 2802, "lr": 1.627858434137991e-06}
{"train_loss": 0.049387406557798386, "global_step": 249382, "epoch": 2802, "lr": 1.6277117120586937e-06}
{"train_loss": 0.06723078340291977, "global_step": 249383, "epoch": 2802, "lr": 1.6275649964824612e-06}
{"train_loss": 0.05598132684826851, "global_step": 249384, "epoch": 2802, "lr": 1.6274182874093103e-06}
{"train_loss": 0.04717056080698967, "global_step": 249385, "epoch": 2802, "lr": 1.6272715848392628e-06}
{"train_loss": 0.047245822846889496, "global_step": 249386, "epoch": 2802, "lr": 1.6271248887723468e-06}
{"train_loss": 0.06386706978082657, "global_step": 249387, "epoch": 2802, "lr": 1.626978199208562e-06}
{"train_loss": 0.02128494530916214, "global_step": 249388, "epoch": 2802, "lr": 1.6268315161479474e-06}
{"train_loss": 0.023407436907291412, "global_step": 249389, "epoch": 2802, "lr": 1.6266848395905198e-06}
{"train_loss": 0.04718879982829094, "global_step": 249390, "epoch": 2802, "lr": 1.6265381695362846e-06}
{"train_loss": 0.023803912103176117, "global_step": 249391, "epoch": 2802, "lr": 1.6263915059852808e-06}
{"train_loss": 0.03857409581542015, "global_step": 249392, "epoch": 2802, "lr": 1.6262448489375137e-06}
{"train_loss": 0.05624617636203766, "global_step": 249393, "epoch": 2802, "lr": 1.6260981983930112e-06}
{"train_loss": 0.07748968154191971, "global_step": 249394, "epoch": 2802, "lr": 1.62595155435179e-06}
{"train_loss": 0.037351224571466446, "global_step": 249395, "epoch": 2802, "lr": 1.6258049168138722e-06}
{"train_loss": 0.058068931102752686, "global_step": 249396, "epoch": 2802, "lr": 1.625658285779269e-06}
{"train_loss": 0.03890860080718994, "global_step": 249397, "epoch": 2802, "lr": 1.6255116612480138e-06}
{"train_loss": 0.0628552958369255, "global_step": 249398, "epoch": 2802, "lr": 1.6253650432201117e-06}
{"train_loss": 0.0961754322052002, "global_step": 249399, "epoch": 2802, "lr": 1.6252184316955965e-06}
{"train_loss": 0.058480095118284225, "global_step": 249400, "epoch": 2802, "lr": 1.6250718266744735e-06}
{"train_loss": 0.017875803634524345, "global_step": 249401, "epoch": 2802, "lr": 1.6249252281567762e-06}
{"train_loss": 0.03657381236553192, "global_step": 249402, "epoch": 2802, "lr": 1.6247786361425155e-06}
{"train_loss": 0.04992227256298065, "global_step": 249403, "epoch": 2802, "lr": 1.6246320506317192e-06}
{"train_loss": 0.040903232991695404, "global_step": 249404, "epoch": 2802, "lr": 1.6244854716243929e-06}
{"train_loss": 0.024492641910910606, "global_step": 249405, "epoch": 2802, "lr": 1.6243388991205699e-06}
{"train_loss": 0.0373237207531929, "global_step": 249406, "epoch": 2802, "lr": 1.6241923331202614e-06}
{"train_loss": 0.047673992812633514, "global_step": 249407, "epoch": 2802, "lr": 1.6240457736234893e-06}
{"train_loss": 0.0732891857624054, "global_step": 249408, "epoch": 2802, "lr": 1.6238992206302816e-06}
{"train_loss": 0.03481004014611244, "global_step": 249409, "epoch": 2802, "lr": 1.623752674140644e-06}
{"train_loss": 0.035090018063783646, "global_step": 249410, "epoch": 2802, "lr": 1.6236061341546094e-06}
{"train_loss": 0.05152786895632744, "global_step": 249411, "epoch": 2802, "lr": 1.6234596006721892e-06}
{"train_loss": 0.025450855493545532, "global_step": 249412, "epoch": 2802, "lr": 1.6233130736934054e-06}
{"train_loss": 0.03777583688497543, "global_step": 249413, "epoch": 2802, "lr": 1.623166553218275e-06}
{"train_loss": 0.01943226344883442, "global_step": 249414, "epoch": 2802, "lr": 1.6230200392468253e-06}
{"train_loss": 0.040173713117837906, "global_step": 249415, "epoch": 2802, "lr": 1.6228735317790733e-06}
{"train_loss": 0.04003977030515671, "global_step": 249416, "epoch": 2802, "lr": 1.6227270308150244e-06}
{"train_loss": 0.044192858040332794, "global_step": 249417, "epoch": 2802, "lr": 1.622580536354723e-06}
{"train_loss": 0.05186902731657028, "global_step": 249418, "epoch": 2802, "lr": 1.6224340483981637e-06}
{"train_loss": 0.07146428525447845, "global_step": 249419, "epoch": 2802, "lr": 1.6222875669453907e-06}
{"train_loss": 0.09289819747209549, "global_step": 249420, "epoch": 2802, "lr": 1.6221410919963987e-06}
{"train_loss": 0.02504199557006359, "global_step": 249421, "epoch": 2802, "lr": 1.6219946235512319e-06}
{"train_loss": 0.03529968112707138, "global_step": 249422, "epoch": 2802, "lr": 1.6218481616098957e-06}
{"train_loss": 0.03995635360479355, "global_step": 249423, "epoch": 2802, "lr": 1.6217017061724071e-06}
{"train_loss": 0.04563937336206436, "global_step": 249424, "epoch": 2802, "lr": 1.6215552572387882e-06}
{"train_loss": 0.0850495845079422, "global_step": 249425, "epoch": 2802, "lr": 1.6214088148090722e-06}
{"train_loss": 0.05847742035984993, "global_step": 249426, "epoch": 2802, "lr": 1.6212623788832592e-06}
{"train_loss": 0.05538554489612579, "global_step": 249427, "epoch": 2802, "lr": 1.6211159494613881e-06}
{"train_loss": 0.04677087441086769, "global_step": 249428, "epoch": 2802, "lr": 1.620969526543459e-06}
{"train_loss": 0.04483349621295929, "global_step": 249429, "epoch": 2802, "lr": 1.6208231101295045e-06}
{"train_loss": 0.059082865715026855, "global_step": 249430, "epoch": 2802, "lr": 1.6206767002195367e-06}
{"train_loss": 0.027079982683062553, "global_step": 249431, "epoch": 2802, "lr": 1.6205302968135883e-06}
{"train_loss": 0.026626380160450935, "global_step": 249432, "epoch": 2802, "lr": 1.6203838999116594e-06}
{"train_loss": 0.054930996149778366, "global_step": 249433, "epoch": 2802, "lr": 1.620237509513789e-06}
{"train_loss": 0.04469609633088112, "global_step": 249434, "epoch": 2802, "lr": 1.6200911256199824e-06}
{"train_loss": 0.11611399054527283, "global_step": 249435, "epoch": 2802, "lr": 1.6199447482302677e-06}
{"train_loss": 0.05474086478352547, "global_step": 249436, "epoch": 2802, "lr": 1.6197983773446557e-06}
{"train_loss": 0.0681934505701065, "global_step": 249437, "epoch": 2802, "lr": 1.6196520129631797e-06}
{"train_loss": 0.06398655474185944, "global_step": 249438, "epoch": 2802, "lr": 1.6195056550858457e-06}
{"train_loss": 0.043050672858953476, "global_step": 249439, "epoch": 2802, "lr": 1.6193593037126865e-06}
{"train_loss": 0.04471488669514656, "global_step": 249440, "epoch": 2802, "lr": 1.6192129588437078e-06}
{"train_loss": 0.07126741111278534, "global_step": 249441, "epoch": 2802, "lr": 1.6190666204789428e-06}
{"train_loss": 0.060539886355400085, "global_step": 249442, "epoch": 2802, "lr": 1.6189202886183973e-06}
{"train_loss": 0.029848549515008926, "global_step": 249443, "epoch": 2802, "lr": 1.618773963262099e-06}
{"train_loss": 0.06852126121520996, "global_step": 249444, "epoch": 2802, "lr": 1.6186276444100757e-06}
{"train_loss": 0.038004957139492035, "global_step": 249445, "epoch": 2802, "lr": 1.6184813320623326e-06}
{"train_loss": 0.05204305052757263, "global_step": 249446, "epoch": 2802, "lr": 1.618335026218898e-06}
{"train_loss": 0.03012535721063614, "global_step": 249447, "epoch": 2802, "lr": 1.6181887268797824e-06}
{"train_loss": 0.03899388760328293, "global_step": 249448, "epoch": 2802, "lr": 1.618042434045014e-06}
{"train_loss": 0.03657372295856476, "global_step": 249449, "epoch": 2802, "lr": 1.6178961477146092e-06}
{"train_loss": 0.024713123217225075, "global_step": 249450, "epoch": 2802, "lr": 1.6177498678885962e-06}
{"train_loss": 0.043583907186985016, "global_step": 249451, "epoch": 2802, "lr": 1.6176035945669743e-06}
{"train_loss": 0.03995799273252487, "global_step": 249452, "epoch": 2802, "lr": 1.6174573277497883e-06}
{"train_loss": 0.04379623755812645, "global_step": 249453, "epoch": 2802, "lr": 1.6173110674370385e-06}
{"train_loss": 0.0498943105340004, "global_step": 249454, "epoch": 2802, "lr": 1.617164813628752e-06}
{"train_loss": 0.02990637719631195, "global_step": 249455, "epoch": 2802, "lr": 1.6170185663249514e-06}
{"train_loss": 0.10013572126626968, "global_step": 249456, "epoch": 2802, "lr": 1.616872325525648e-06}
{"train_loss": 0.011924689635634422, "global_step": 249457, "epoch": 2802, "lr": 1.6167260912308745e-06}
{"train_loss": 0.06101391091942787, "global_step": 249458, "epoch": 2802, "lr": 1.6165798634406314e-06}
{"train_loss": 0.04439309984445572, "global_step": 249459, "epoch": 2802, "lr": 1.6164336421549574e-06}
{"train_loss": 0.007109996397048235, "global_step": 249460, "epoch": 2802, "lr": 1.6162874273738637e-06}
{"train_loss": 0.057403355836868286, "global_step": 249461, "epoch": 2802, "lr": 1.6161412190973669e-06}
{"train_loss": 0.06637152284383774, "global_step": 249462, "epoch": 2802, "lr": 1.615995017325489e-06}
{"train_loss": 0.0449308380484581, "global_step": 249463, "epoch": 2802, "lr": 1.615848822058258e-06}
{"train_loss": 0.022056635469198227, "global_step": 249464, "epoch": 2802, "lr": 1.6157026332956793e-06}
{"train_loss": 0.03629084303975105, "global_step": 249465, "epoch": 2802, "lr": 1.6155564510377863e-06}
{"train_loss": 0.04701021983905622, "global_step": 249466, "epoch": 2802, "lr": 1.6154102752845846e-06, "val_loss": 9.563202857971191}
{"train_loss": 0.0455779992043972, "global_step": 249467, "epoch": 2803, "lr": 1.6152641060361074e-06}
{"train_loss": 0.02235851250588894, "global_step": 249468, "epoch": 2803, "lr": 1.6151179432923657e-06}
{"train_loss": 0.04403730109333992, "global_step": 249469, "epoch": 2803, "lr": 1.614971787053382e-06}
{"train_loss": 0.027982911095023155, "global_step": 249470, "epoch": 2803, "lr": 1.6148256373191727e-06}
{"train_loss": 0.022853948175907135, "global_step": 249471, "epoch": 2803, "lr": 1.614679494089766e-06}
{"train_loss": 0.04265444725751877, "global_step": 249472, "epoch": 2803, "lr": 1.6145333573651721e-06}
{"train_loss": 0.06446271389722824, "global_step": 249473, "epoch": 2803, "lr": 1.614387227145414e-06}
{"train_loss": 0.011787368915975094, "global_step": 249474, "epoch": 2803, "lr": 1.6142411034305139e-06}
{"train_loss": 0.031040413305163383, "global_step": 249475, "epoch": 2803, "lr": 1.614094986220488e-06}
{"train_loss": 0.019155152142047882, "global_step": 249476, "epoch": 2803, "lr": 1.6139488755153531e-06}
{"train_loss": 0.0401671826839447, "global_step": 249477, "epoch": 2803, "lr": 1.6138027713151427e-06}
{"train_loss": 0.08216238766908646, "global_step": 249478, "epoch": 2803, "lr": 1.6136566736198566e-06}
{"train_loss": 0.019917085766792297, "global_step": 249479, "epoch": 2803, "lr": 1.6135105824295282e-06}
{"train_loss": 0.05747087299823761, "global_step": 249480, "epoch": 2803, "lr": 1.6133644977441798e-06}
{"train_loss": 0.027256114408373833, "global_step": 249481, "epoch": 2803, "lr": 1.6132184195638168e-06}
{"train_loss": 0.02446068450808525, "global_step": 249482, "epoch": 2803, "lr": 1.6130723478884724e-06}
{"train_loss": 0.057580575346946716, "global_step": 249483, "epoch": 2803, "lr": 1.6129262827181578e-06}
{"train_loss": 0.05257695913314819, "global_step": 249484, "epoch": 2803, "lr": 1.6127802240528956e-06}
{"train_loss": 0.0408160500228405, "global_step": 249485, "epoch": 2803, "lr": 1.6126341718927018e-06}
{"train_loss": 0.04210026189684868, "global_step": 249486, "epoch": 2803, "lr": 1.61248812623761e-06}
{"train_loss": 0.010934235528111458, "global_step": 249487, "epoch": 2803, "lr": 1.6123420870876204e-06}
{"train_loss": 0.03377359360456467, "global_step": 249488, "epoch": 2803, "lr": 1.612196054442766e-06}
{"train_loss": 0.04312705993652344, "global_step": 249489, "epoch": 2803, "lr": 1.6120500283030582e-06}
{"train_loss": 0.024982046335935593, "global_step": 249490, "epoch": 2803, "lr": 1.6119040086685245e-06}
{"train_loss": 0.047343168407678604, "global_step": 249491, "epoch": 2803, "lr": 1.611757995539176e-06}
{"train_loss": 0.04893146827816963, "global_step": 249492, "epoch": 2803, "lr": 1.6116119889150404e-06}
{"train_loss": 0.030475890263915062, "global_step": 249493, "epoch": 2803, "lr": 1.6114659887961348e-06}
{"train_loss": 0.0352325476706028, "global_step": 249494, "epoch": 2803, "lr": 1.6113199951824754e-06}
{"train_loss": 0.03841090574860573, "global_step": 249495, "epoch": 2803, "lr": 1.6111740080740845e-06}
{"train_loss": 0.039325691759586334, "global_step": 249496, "epoch": 2803, "lr": 1.6110280274709787e-06}
{"train_loss": 0.054476261138916016, "global_step": 249497, "epoch": 2803, "lr": 1.6108820533731861e-06}
{"train_loss": 0.03224625810980797, "global_step": 249498, "epoch": 2803, "lr": 1.6107360857807175e-06}
{"train_loss": 0.036881957203149796, "global_step": 249499, "epoch": 2803, "lr": 1.6105901246936006e-06}
{"train_loss": 0.04268542304635048, "global_step": 249500, "epoch": 2803, "lr": 1.6104441701118412e-06}
{"train_loss": 0.0509846992790699, "global_step": 249501, "epoch": 2803, "lr": 1.610298222035478e-06}
{"train_loss": 0.030618760734796524, "global_step": 249502, "epoch": 2803, "lr": 1.610152280464511e-06}
{"train_loss": 0.05212373286485672, "global_step": 249503, "epoch": 2803, "lr": 1.6100063453989734e-06}
{"train_loss": 0.037040624767541885, "global_step": 249504, "epoch": 2803, "lr": 1.6098604168388765e-06}
{"train_loss": 0.0653885006904602, "global_step": 249505, "epoch": 2803, "lr": 1.6097144947842535e-06}
{"train_loss": 0.023122770711779594, "global_step": 249506, "epoch": 2803, "lr": 1.6095685792351044e-06}
{"train_loss": 0.043867189437150955, "global_step": 249507, "epoch": 2803, "lr": 1.6094226701914683e-06}
{"train_loss": 0.0575009360909462, "global_step": 249508, "epoch": 2803, "lr": 1.609276767653345e-06}
{"train_loss": 0.03908964619040489, "global_step": 249509, "epoch": 2803, "lr": 1.6091308716207731e-06}
{"train_loss": 0.06838827580213547, "global_step": 249510, "epoch": 2803, "lr": 1.6089849820937586e-06}
{"train_loss": 0.0777536928653717, "global_step": 249511, "epoch": 2803, "lr": 1.608839099072329e-06}
{"train_loss": 0.04439307376742363, "global_step": 249512, "epoch": 2803, "lr": 1.608693222556501e-06}
{"train_loss": 0.06498492509126663, "global_step": 249513, "epoch": 2803, "lr": 1.6085473525462968e-06}
{"train_loss": 0.04976703226566315, "global_step": 249514, "epoch": 2803, "lr": 1.6084014890417275e-06}
{"train_loss": 0.015179810114204884, "global_step": 249515, "epoch": 2803, "lr": 1.608255632042821e-06}
{"train_loss": 0.018287038430571556, "global_step": 249516, "epoch": 2803, "lr": 1.6081097815495939e-06}
{"train_loss": 0.056604400277137756, "global_step": 249517, "epoch": 2803, "lr": 1.607963937562068e-06}
{"train_loss": 0.0241914801299572, "global_step": 249518, "epoch": 2803, "lr": 1.6078181000802662e-06}
{"train_loss": 0.06285165250301361, "global_step": 249519, "epoch": 2803, "lr": 1.6076722691041935e-06}
{"train_loss": 0.01415170542895794, "global_step": 249520, "epoch": 2803, "lr": 1.6075264446338888e-06}
{"train_loss": 0.038267865777015686, "global_step": 249521, "epoch": 2803, "lr": 1.6073806266693581e-06}
{"train_loss": 0.03812111169099808, "global_step": 249522, "epoch": 2803, "lr": 1.6072348152106287e-06}
{"train_loss": 0.03197740018367767, "global_step": 249523, "epoch": 2803, "lr": 1.6070890102577119e-06}
{"train_loss": 0.025546882301568985, "global_step": 249524, "epoch": 2803, "lr": 1.6069432118106353e-06}
{"train_loss": 0.025491654872894287, "global_step": 249525, "epoch": 2803, "lr": 1.6067974198694102e-06}
{"train_loss": 0.0435771644115448, "global_step": 249526, "epoch": 2803, "lr": 1.6066516344340698e-06}
{"train_loss": 0.04771006107330322, "global_step": 249527, "epoch": 2803, "lr": 1.6065058555046198e-06}
{"train_loss": 0.048247821629047394, "global_step": 249528, "epoch": 2803, "lr": 1.6063600830810877e-06}
{"train_loss": 0.040173668414354324, "global_step": 249529, "epoch": 2803, "lr": 1.6062143171634902e-06}
{"train_loss": 0.047438737004995346, "global_step": 249530, "epoch": 2803, "lr": 1.6060685577518497e-06}
{"train_loss": 0.03216525912284851, "global_step": 249531, "epoch": 2803, "lr": 1.6059228048461828e-06}
{"train_loss": 0.037712231278419495, "global_step": 249532, "epoch": 2803, "lr": 1.6057770584465004e-06}
{"train_loss": 0.06838249415159225, "global_step": 249533, "epoch": 2803, "lr": 1.6056313185528415e-06}
{"train_loss": 0.05703184753656387, "global_step": 249534, "epoch": 2803, "lr": 1.6054855851652117e-06}
{"train_loss": 0.07120661437511444, "global_step": 249535, "epoch": 2803, "lr": 1.6053398582836388e-06}
{"train_loss": 0.03432905673980713, "global_step": 249536, "epoch": 2803, "lr": 1.605194137908128e-06}
{"train_loss": 0.058174967765808105, "global_step": 249537, "epoch": 2803, "lr": 1.6050484240387187e-06}
{"train_loss": 0.06207610294222832, "global_step": 249538, "epoch": 2803, "lr": 1.604902716675416e-06}
{"train_loss": 0.04603167623281479, "global_step": 249539, "epoch": 2803, "lr": 1.604757015818248e-06}
{"train_loss": 0.030566735193133354, "global_step": 249540, "epoch": 2803, "lr": 1.6046113214672254e-06}
{"train_loss": 0.04726937413215637, "global_step": 249541, "epoch": 2803, "lr": 1.6044656336223818e-06}
{"train_loss": 0.036379147320985794, "global_step": 249542, "epoch": 2803, "lr": 1.604319952283717e-06}
{"train_loss": 0.021692520007491112, "global_step": 249543, "epoch": 2803, "lr": 1.6041742774512701e-06}
{"train_loss": 0.032565079629421234, "global_step": 249544, "epoch": 2803, "lr": 1.6040286091250468e-06}
{"train_loss": 0.05610935017466545, "global_step": 249545, "epoch": 2803, "lr": 1.6038829473050743e-06}
{"train_loss": 0.01652868464589119, "global_step": 249546, "epoch": 2803, "lr": 1.6037372919913695e-06}
{"train_loss": 0.08132549375295639, "global_step": 249547, "epoch": 2803, "lr": 1.6035916431839549e-06}
{"train_loss": 0.07321605831384659, "global_step": 249548, "epoch": 2803, "lr": 1.6034460008828412e-06}
{"train_loss": 0.04073818027973175, "global_step": 249549, "epoch": 2803, "lr": 1.603300365088062e-06}
{"train_loss": 0.02993905544281006, "global_step": 249550, "epoch": 2803, "lr": 1.6031547357996224e-06}
{"train_loss": 0.0068705445155501366, "global_step": 249551, "epoch": 2803, "lr": 1.6030091130175506e-06}
{"train_loss": 0.03646649792790413, "global_step": 249552, "epoch": 2803, "lr": 1.6028634967418633e-06}
{"train_loss": 0.04522598162293434, "global_step": 249553, "epoch": 2803, "lr": 1.6027178869725823e-06}
{"train_loss": 0.04851953685283661, "global_step": 249554, "epoch": 2803, "lr": 1.60257228370973e-06}
{"train_loss": 0.041656106637183866, "global_step": 249555, "epoch": 2803, "lr": 1.6024266869533177e-06, "val_loss": 9.675910949707031}
{"train_loss": 0.05981874465942383, "global_step": 249556, "epoch": 2804, "lr": 1.602281096703373e-06}
{"train_loss": 0.018110312521457672, "global_step": 249557, "epoch": 2804, "lr": 1.602135512959907e-06}
{"train_loss": 0.041977282613515854, "global_step": 249558, "epoch": 2804, "lr": 1.6019899357229474e-06}
{"train_loss": 0.06327970325946808, "global_step": 249559, "epoch": 2804, "lr": 1.6018443649925052e-06}
{"train_loss": 0.04300101846456528, "global_step": 249560, "epoch": 2804, "lr": 1.601698800768614e-06}
{"train_loss": 0.039148859679698944, "global_step": 249561, "epoch": 2804, "lr": 1.6015532430512736e-06}
{"train_loss": 0.06017260253429413, "global_step": 249562, "epoch": 2804, "lr": 1.601407691840523e-06}
{"train_loss": 0.03106461465358734, "global_step": 249563, "epoch": 2804, "lr": 1.6012621471363675e-06}
{"train_loss": 0.05952809005975723, "global_step": 249564, "epoch": 2804, "lr": 1.6011166089388408e-06}
{"train_loss": 0.015781963244080544, "global_step": 249565, "epoch": 2804, "lr": 1.600971077247948e-06}
{"train_loss": 0.08108536154031754, "global_step": 249566, "epoch": 2804, "lr": 1.6008255520637171e-06}
{"train_loss": 0.06694699078798294, "global_step": 249567, "epoch": 2804, "lr": 1.6006800333861594e-06}
{"train_loss": 0.0414484366774559, "global_step": 249568, "epoch": 2804, "lr": 1.6005345212153079e-06}
{"train_loss": 0.04922385513782501, "global_step": 249569, "epoch": 2804, "lr": 1.600389015551168e-06}
{"train_loss": 0.04288841038942337, "global_step": 249570, "epoch": 2804, "lr": 1.6002435163937734e-06}
{"train_loss": 0.047270968556404114, "global_step": 249571, "epoch": 2804, "lr": 1.6000980237431351e-06}
{"train_loss": 0.04472026601433754, "global_step": 249572, "epoch": 2804, "lr": 1.5999525375992697e-06}
{"train_loss": 0.04016055166721344, "global_step": 249573, "epoch": 2804, "lr": 1.5998070579621994e-06}
{"train_loss": 0.027834990993142128, "global_step": 249574, "epoch": 2804, "lr": 1.5996615848319463e-06}
{"train_loss": 0.019704757258296013, "global_step": 249575, "epoch": 2804, "lr": 1.5995161182085329e-06}
{"train_loss": 0.05254840478301048, "global_step": 249576, "epoch": 2804, "lr": 1.5993706580919697e-06}
{"train_loss": 0.030899319797754288, "global_step": 249577, "epoch": 2804, "lr": 1.5992252044822853e-06}
{"train_loss": 0.07577935606241226, "global_step": 249578, "epoch": 2804, "lr": 1.5990797573794903e-06}
{"train_loss": 0.03001459501683712, "global_step": 249579, "epoch": 2804, "lr": 1.598934316783618e-06}
{"train_loss": 0.048662323504686356, "global_step": 249580, "epoch": 2804, "lr": 1.5987888826946684e-06}
{"train_loss": 0.0484718419611454, "global_step": 249581, "epoch": 2804, "lr": 1.5986434551126806e-06}
{"train_loss": 0.052931834012269974, "global_step": 249582, "epoch": 2804, "lr": 1.5984980340376543e-06}
{"train_loss": 0.07339899241924286, "global_step": 249583, "epoch": 2804, "lr": 1.5983526194696285e-06}
{"train_loss": 0.03640807792544365, "global_step": 249584, "epoch": 2804, "lr": 1.598207211408609e-06}
{"train_loss": 0.0581306628882885, "global_step": 249585, "epoch": 2804, "lr": 1.5980618098546286e-06}
{"train_loss": 0.0041804239153862, "global_step": 249586, "epoch": 2804, "lr": 1.5979164148076932e-06}
{"train_loss": 0.04763048514723778, "global_step": 249587, "epoch": 2804, "lr": 1.5977710262678247e-06}
{"train_loss": 0.02394920215010643, "global_step": 249588, "epoch": 2804, "lr": 1.5976256442350512e-06}
{"train_loss": 0.041185494512319565, "global_step": 249589, "epoch": 2804, "lr": 1.5974802687093837e-06}
{"train_loss": 0.02779129520058632, "global_step": 249590, "epoch": 2804, "lr": 1.59733489969085e-06}
{"train_loss": 0.0333719439804554, "global_step": 249591, "epoch": 2804, "lr": 1.5971895371794609e-06}
{"train_loss": 0.05924785137176514, "global_step": 249592, "epoch": 2804, "lr": 1.5970441811752445e-06}
{"train_loss": 0.04556713253259659, "global_step": 249593, "epoch": 2804, "lr": 1.5968988316782064e-06}
{"train_loss": 0.03999042510986328, "global_step": 249594, "epoch": 2804, "lr": 1.5967534886883851e-06}
{"train_loss": 0.04255380854010582, "global_step": 249595, "epoch": 2804, "lr": 1.5966081522057863e-06}
{"train_loss": 0.03779306262731552, "global_step": 249596, "epoch": 2804, "lr": 1.5964628222304323e-06}
{"train_loss": 0.0769881159067154, "global_step": 249597, "epoch": 2804, "lr": 1.5963174987623453e-06}
{"train_loss": 0.04567064344882965, "global_step": 249598, "epoch": 2804, "lr": 1.5961721818015418e-06}
{"train_loss": 0.023762337863445282, "global_step": 249599, "epoch": 2804, "lr": 1.5960268713480442e-06}
{"train_loss": 0.034447476267814636, "global_step": 249600, "epoch": 2804, "lr": 1.5958815674018746e-06}
{"train_loss": 0.059752173721790314, "global_step": 249601, "epoch": 2804, "lr": 1.5957362699630385e-06}
{"train_loss": 0.056497007608413696, "global_step": 249602, "epoch": 2804, "lr": 1.5955909790315749e-06}
{"train_loss": 0.05679209530353546, "global_step": 249603, "epoch": 2804, "lr": 1.5954456946074892e-06}
{"train_loss": 0.027321718633174896, "global_step": 249604, "epoch": 2804, "lr": 1.5953004166908147e-06}
{"train_loss": 0.03466751426458359, "global_step": 249605, "epoch": 2804, "lr": 1.5951551452815516e-06}
{"train_loss": 0.020722215995192528, "global_step": 249606, "epoch": 2804, "lr": 1.5950098803797387e-06}
{"train_loss": 0.028143635019659996, "global_step": 249607, "epoch": 2804, "lr": 1.5948646219853757e-06}
{"train_loss": 0.022883206605911255, "global_step": 249608, "epoch": 2804, "lr": 1.594719370098502e-06}
{"train_loss": 0.04399576783180237, "global_step": 249609, "epoch": 2804, "lr": 1.5945741247191281e-06}
{"train_loss": 0.054099857807159424, "global_step": 249610, "epoch": 2804, "lr": 1.5944288858472712e-06}
{"train_loss": 0.02660166472196579, "global_step": 249611, "epoch": 2804, "lr": 1.594283653482953e-06}
{"train_loss": 0.03537236899137497, "global_step": 249612, "epoch": 2804, "lr": 1.5941384276261906e-06}
{"train_loss": 0.0394497886300087, "global_step": 249613, "epoch": 2804, "lr": 1.593993208277017e-06}
{"train_loss": 0.015783710405230522, "global_step": 249614, "epoch": 2804, "lr": 1.5938479954354268e-06}
{"train_loss": 0.06925547122955322, "global_step": 249615, "epoch": 2804, "lr": 1.5937027891014643e-06}
{"train_loss": 0.057815708220005035, "global_step": 249616, "epoch": 2804, "lr": 1.5935575892751352e-06}
{"train_loss": 0.040781453251838684, "global_step": 249617, "epoch": 2804, "lr": 1.5934123959564618e-06}
{"train_loss": 0.05290114879608154, "global_step": 249618, "epoch": 2804, "lr": 1.5932672091454603e-06}
{"train_loss": 0.0831189826130867, "global_step": 249619, "epoch": 2804, "lr": 1.5931220288421645e-06}
{"train_loss": 0.05339366942644119, "global_step": 249620, "epoch": 2804, "lr": 1.592976855046574e-06}
{"train_loss": 0.057939186692237854, "global_step": 249621, "epoch": 2804, "lr": 1.5928316877587224e-06}
{"train_loss": 0.050234317779541016, "global_step": 249622, "epoch": 2804, "lr": 1.5926865269786206e-06}
{"train_loss": 0.02683454379439354, "global_step": 249623, "epoch": 2804, "lr": 1.592541372706291e-06}
{"train_loss": 0.03968152403831482, "global_step": 249624, "epoch": 2804, "lr": 1.5923962249417611e-06}
{"train_loss": 0.05784207955002785, "global_step": 249625, "epoch": 2804, "lr": 1.592251083685037e-06}
{"train_loss": 0.03908602520823479, "global_step": 249626, "epoch": 2804, "lr": 1.5921059489361456e-06}
{"train_loss": 0.04036423936486244, "global_step": 249627, "epoch": 2804, "lr": 1.5919608206951042e-06}
{"train_loss": 0.05626361072063446, "global_step": 249628, "epoch": 2804, "lr": 1.5918156989619403e-06}
{"train_loss": 0.05459843575954437, "global_step": 249629, "epoch": 2804, "lr": 1.5916705837366653e-06}
{"train_loss": 0.008200783282518387, "global_step": 249630, "epoch": 2804, "lr": 1.591525475019301e-06}
{"train_loss": 0.041706930845975876, "global_step": 249631, "epoch": 2804, "lr": 1.5913803728098587e-06}
{"train_loss": 0.014117663726210594, "global_step": 249632, "epoch": 2804, "lr": 1.5912352771083716e-06}
{"train_loss": 0.047662004828453064, "global_step": 249633, "epoch": 2804, "lr": 1.591090187914851e-06}
{"train_loss": 0.07322568446397781, "global_step": 249634, "epoch": 2804, "lr": 1.590945105229319e-06}
{"train_loss": 0.06405634433031082, "global_step": 249635, "epoch": 2804, "lr": 1.5908000290517921e-06}
{"train_loss": 0.06576565653085709, "global_step": 249636, "epoch": 2804, "lr": 1.5906549593822984e-06}
{"train_loss": 0.049436427652835846, "global_step": 249637, "epoch": 2804, "lr": 1.590509896220843e-06}
{"train_loss": 0.04474172368645668, "global_step": 249638, "epoch": 2804, "lr": 1.5903648395674597e-06}
{"train_loss": 0.02787754125893116, "global_step": 249639, "epoch": 2804, "lr": 1.5902197894221537e-06}
{"train_loss": 0.04156564176082611, "global_step": 249640, "epoch": 2804, "lr": 1.590074745784964e-06}
{"train_loss": 0.027048639953136444, "global_step": 249641, "epoch": 2804, "lr": 1.5899297086558906e-06}
{"train_loss": 0.0470605194568634, "global_step": 249642, "epoch": 2804, "lr": 1.5897846780349667e-06}
{"train_loss": 0.07471303641796112, "global_step": 249643, "epoch": 2804, "lr": 1.589639653922198e-06}
{"train_loss": 0.04414425055716145, "global_step": 249644, "epoch": 2804, "lr": 1.5894946363176234e-06, "val_loss": 9.491358757019043}
{"train_loss": 0.08398443460464478, "global_step": 249645, "epoch": 2805, "lr": 1.589349625221237e-06}
{"train_loss": 0.04216260462999344, "global_step": 249646, "epoch": 2805, "lr": 1.5892046206330834e-06}
{"train_loss": 0.05006266012787819, "global_step": 249647, "epoch": 2805, "lr": 1.589059622553174e-06}
{"train_loss": 0.009952468797564507, "global_step": 249648, "epoch": 2805, "lr": 1.588914630981514e-06}
{"train_loss": 0.06018325313925743, "global_step": 249649, "epoch": 2805, "lr": 1.5887696459181478e-06}
{"train_loss": 0.05691808462142944, "global_step": 249650, "epoch": 2805, "lr": 1.58862466736307e-06}
{"train_loss": 0.048100795596838, "global_step": 249651, "epoch": 2805, "lr": 1.5884796953163194e-06}
{"train_loss": 0.041253551840782166, "global_step": 249652, "epoch": 2805, "lr": 1.5883347297778962e-06}
{"train_loss": 0.03692251816391945, "global_step": 249653, "epoch": 2805, "lr": 1.5881897707478444e-06}
{"train_loss": 0.022119401022791862, "global_step": 249654, "epoch": 2805, "lr": 1.5880448182261642e-06}
{"train_loss": 0.05077161639928818, "global_step": 249655, "epoch": 2805, "lr": 1.5878998722128836e-06}
{"train_loss": 0.040134795010089874, "global_step": 249656, "epoch": 2805, "lr": 1.5877549327080133e-06}
{"train_loss": 0.08406466990709305, "global_step": 249657, "epoch": 2805, "lr": 1.5876099997115867e-06}
{"train_loss": 0.0113521758466959, "global_step": 249658, "epoch": 2805, "lr": 1.5874650732236096e-06}
{"train_loss": 0.09785792976617813, "global_step": 249659, "epoch": 2805, "lr": 1.587320153244115e-06}
{"train_loss": 0.027736254036426544, "global_step": 249660, "epoch": 2805, "lr": 1.5871752397731089e-06}
{"train_loss": 0.045243971049785614, "global_step": 249661, "epoch": 2805, "lr": 1.5870303328106185e-06}
{"train_loss": 0.03184583783149719, "global_step": 249662, "epoch": 2805, "lr": 1.5868854323566662e-06}
{"train_loss": 0.03759064897894859, "global_step": 249663, "epoch": 2805, "lr": 1.5867405384112577e-06}
{"train_loss": 0.040145449340343475, "global_step": 249664, "epoch": 2805, "lr": 1.5865956509744317e-06}
{"train_loss": 0.04059305414557457, "global_step": 249665, "epoch": 2805, "lr": 1.586450770046194e-06}
{"train_loss": 0.053619276732206345, "global_step": 249666, "epoch": 2805, "lr": 1.586305895626572e-06}
{"train_loss": 0.06349599361419678, "global_step": 249667, "epoch": 2805, "lr": 1.5861610277155715e-06}
{"train_loss": 0.06935162842273712, "global_step": 249668, "epoch": 2805, "lr": 1.5860161663132312e-06}
{"train_loss": 0.04920259863138199, "global_step": 249669, "epoch": 2805, "lr": 1.5858713114195567e-06}
{"train_loss": 0.05768829211592674, "global_step": 249670, "epoch": 2805, "lr": 1.5857264630345703e-06}
{"train_loss": 0.025820257142186165, "global_step": 249671, "epoch": 2805, "lr": 1.5855816211582942e-06}
{"train_loss": 0.02837555855512619, "global_step": 249672, "epoch": 2805, "lr": 1.5854367857907504e-06}
{"train_loss": 0.02994670532643795, "global_step": 249673, "epoch": 2805, "lr": 1.5852919569319447e-06}
{"train_loss": 0.0412130281329155, "global_step": 249674, "epoch": 2805, "lr": 1.5851471345819158e-06}
{"train_loss": 0.033537667244672775, "global_step": 249675, "epoch": 2805, "lr": 1.5850023187406693e-06}
{"train_loss": 0.028369002044200897, "global_step": 249676, "epoch": 2805, "lr": 1.584857509408233e-06}
{"train_loss": 0.06014915183186531, "global_step": 249677, "epoch": 2805, "lr": 1.5847127065846179e-06}
{"train_loss": 0.04311602935194969, "global_step": 249678, "epoch": 2805, "lr": 1.5845679102698464e-06}
{"train_loss": 0.0590897761285305, "global_step": 249679, "epoch": 2805, "lr": 1.5844231204639458e-06}
{"train_loss": 0.034194011241197586, "global_step": 249680, "epoch": 2805, "lr": 1.5842783371669279e-06}
{"train_loss": 0.0482272133231163, "global_step": 249681, "epoch": 2805, "lr": 1.5841335603788143e-06}
{"train_loss": 0.044074926525354385, "global_step": 249682, "epoch": 2805, "lr": 1.5839887900996276e-06}
{"train_loss": 0.041785743087530136, "global_step": 249683, "epoch": 2805, "lr": 1.583844026329373e-06}
{"train_loss": 0.021027693524956703, "global_step": 249684, "epoch": 2805, "lr": 1.5836992690680896e-06}
{"train_loss": 0.06399647891521454, "global_step": 249685, "epoch": 2805, "lr": 1.583554518315783e-06}
{"train_loss": 0.04330817982554436, "global_step": 249686, "epoch": 2805, "lr": 1.5834097740724807e-06}
{"train_loss": 0.031090963631868362, "global_step": 249687, "epoch": 2805, "lr": 1.5832650363381995e-06}
{"train_loss": 0.029359951615333557, "global_step": 249688, "epoch": 2805, "lr": 1.5831203051129506e-06}
{"train_loss": 0.016701146960258484, "global_step": 249689, "epoch": 2805, "lr": 1.5829755803967672e-06}
{"train_loss": 0.0399448536336422, "global_step": 249690, "epoch": 2805, "lr": 1.5828308621896605e-06}
{"train_loss": 0.056303735822439194, "global_step": 249691, "epoch": 2805, "lr": 1.5826861504916523e-06}
{"train_loss": 0.05263006314635277, "global_step": 249692, "epoch": 2805, "lr": 1.5825414453027598e-06}
{"train_loss": 0.04391882196068764, "global_step": 249693, "epoch": 2805, "lr": 1.5823967466230105e-06}
{"train_loss": 0.059249021112918854, "global_step": 249694, "epoch": 2805, "lr": 1.58225205445241e-06}
{"train_loss": 0.027829142287373543, "global_step": 249695, "epoch": 2805, "lr": 1.5821073687909972e-06}
{"train_loss": 0.04240903630852699, "global_step": 249696, "epoch": 2805, "lr": 1.5819626896387663e-06}
{"train_loss": 0.042384881526231766, "global_step": 249697, "epoch": 2805, "lr": 1.5818180169957564e-06}
{"train_loss": 0.05836793780326843, "global_step": 249698, "epoch": 2805, "lr": 1.581673350861984e-06}
{"train_loss": 0.06905870139598846, "global_step": 249699, "epoch": 2805, "lr": 1.5815286912374604e-06}
{"train_loss": 0.06181619316339493, "global_step": 249700, "epoch": 2805, "lr": 1.5813840381222134e-06}
{"train_loss": 0.06539209932088852, "global_step": 249701, "epoch": 2805, "lr": 1.5812393915162537e-06}
{"train_loss": 0.053629517555236816, "global_step": 249702, "epoch": 2805, "lr": 1.5810947514196152e-06}
{"train_loss": 0.044623274356126785, "global_step": 249703, "epoch": 2805, "lr": 1.580950117832297e-06}
{"train_loss": 0.05838434398174286, "global_step": 249704, "epoch": 2805, "lr": 1.5808054907543447e-06}
{"train_loss": 0.023013919591903687, "global_step": 249705, "epoch": 2805, "lr": 1.5806608701857517e-06}
{"train_loss": 0.04293258488178253, "global_step": 249706, "epoch": 2805, "lr": 1.5805162561265518e-06}
{"train_loss": 0.05648798495531082, "global_step": 249707, "epoch": 2805, "lr": 1.5803716485767617e-06}
{"train_loss": 0.0566917285323143, "global_step": 249708, "epoch": 2805, "lr": 1.5802270475364034e-06}
{"train_loss": 0.04366973787546158, "global_step": 249709, "epoch": 2805, "lr": 1.5800824530054825e-06}
{"train_loss": 0.01989852450788021, "global_step": 249710, "epoch": 2805, "lr": 1.5799378649840435e-06}
{"train_loss": 0.06697142124176025, "global_step": 249711, "epoch": 2805, "lr": 1.5797932834720809e-06}
{"train_loss": 0.022896233946084976, "global_step": 249712, "epoch": 2805, "lr": 1.5796487084696331e-06}
{"train_loss": 0.08649931848049164, "global_step": 249713, "epoch": 2805, "lr": 1.5795041399767062e-06}
{"train_loss": 0.0762358158826828, "global_step": 249714, "epoch": 2805, "lr": 1.5793595779933224e-06}
{"train_loss": 0.007922055199742317, "global_step": 249715, "epoch": 2805, "lr": 1.579215022519509e-06}
{"train_loss": 0.029586149379611015, "global_step": 249716, "epoch": 2805, "lr": 1.5790704735552775e-06}
{"train_loss": 0.07817140966653824, "global_step": 249717, "epoch": 2805, "lr": 1.57892593110065e-06}
{"train_loss": 0.06408926844596863, "global_step": 249718, "epoch": 2805, "lr": 1.578781395155643e-06}
{"train_loss": 0.06417886912822723, "global_step": 249719, "epoch": 2805, "lr": 1.57863686572029e-06}
{"train_loss": 0.04483991861343384, "global_step": 249720, "epoch": 2805, "lr": 1.5784923427945908e-06}
{"train_loss": 0.03854510933160782, "global_step": 249721, "epoch": 2805, "lr": 1.5783478263785734e-06}
{"train_loss": 0.03300675377249718, "global_step": 249722, "epoch": 2805, "lr": 1.5782033164722598e-06}
{"train_loss": 0.042101286351680756, "global_step": 249723, "epoch": 2805, "lr": 1.5780588130756612e-06}
{"train_loss": 0.04716929420828819, "global_step": 249724, "epoch": 2805, "lr": 1.5779143161888056e-06}
{"train_loss": 0.040063656866550446, "global_step": 249725, "epoch": 2805, "lr": 1.577769825811709e-06}
{"train_loss": 0.04599635303020477, "global_step": 249726, "epoch": 2805, "lr": 1.5776253419443887e-06}
{"train_loss": 0.05027835816144943, "global_step": 249727, "epoch": 2805, "lr": 1.5774808645868723e-06}
{"train_loss": 0.20781639218330383, "global_step": 249728, "epoch": 2805, "lr": 1.577336393739165e-06}
{"train_loss": 0.03201109170913696, "global_step": 249729, "epoch": 2805, "lr": 1.577191929401306e-06}
{"train_loss": 0.05978298932313919, "global_step": 249730, "epoch": 2805, "lr": 1.5770474715732897e-06}
{"train_loss": 0.02636125683784485, "global_step": 249731, "epoch": 2805, "lr": 1.5769030202551605e-06}
{"train_loss": 0.018530942499637604, "global_step": 249732, "epoch": 2805, "lr": 1.5767585754469184e-06}
{"train_loss": 0.04715477629156595, "global_step": 249733, "epoch": 2805, "lr": 1.576614137148591e-06, "val_loss": 9.474576950073242, "train_action_mse_error": 29.400863647460938}
{"train_loss": 0.022067315876483917, "global_step": 249734, "epoch": 2806, "lr": 1.5764697053602062e-06}
{"train_loss": 0.027748415246605873, "global_step": 249735, "epoch": 2806, "lr": 1.5763252800817697e-06}
{"train_loss": 0.06721270084381104, "global_step": 249736, "epoch": 2806, "lr": 1.576180861313309e-06}
{"train_loss": 0.05171385407447815, "global_step": 249737, "epoch": 2806, "lr": 1.5760364490548352e-06}
{"train_loss": 0.0521029457449913, "global_step": 249738, "epoch": 2806, "lr": 1.5758920433063761e-06}
{"train_loss": 0.024216197431087494, "global_step": 249739, "epoch": 2806, "lr": 1.575747644067943e-06}
{"train_loss": 0.05136905238032341, "global_step": 249740, "epoch": 2806, "lr": 1.5756032513395692e-06}
{"train_loss": 0.08553022891283035, "global_step": 249741, "epoch": 2806, "lr": 1.57545886512126e-06}
{"train_loss": 0.014740705490112305, "global_step": 249742, "epoch": 2806, "lr": 1.575314485413043e-06}
{"train_loss": 0.030999289825558662, "global_step": 249743, "epoch": 2806, "lr": 1.57517011221493e-06}
{"train_loss": 0.037204477936029434, "global_step": 249744, "epoch": 2806, "lr": 1.5750257455269534e-06}
{"train_loss": 0.05228598043322563, "global_step": 249745, "epoch": 2806, "lr": 1.574881385349114e-06}
{"train_loss": 0.043780408799648285, "global_step": 249746, "epoch": 2806, "lr": 1.5747370316814502e-06}
{"train_loss": 0.05791214108467102, "global_step": 249747, "epoch": 2806, "lr": 1.5745926845239679e-06}
{"train_loss": 0.05721145495772362, "global_step": 249748, "epoch": 2806, "lr": 1.5744483438766944e-06}
{"train_loss": 0.12228075414896011, "global_step": 249749, "epoch": 2806, "lr": 1.5743040097396412e-06}
{"train_loss": 0.06068355590105057, "global_step": 249750, "epoch": 2806, "lr": 1.5741596821128358e-06}
{"train_loss": 0.033953651785850525, "global_step": 249751, "epoch": 2806, "lr": 1.5740153609962949e-06}
{"train_loss": 0.08623509109020233, "global_step": 249752, "epoch": 2806, "lr": 1.5738710463900353e-06}
{"train_loss": 0.03570445999503136, "global_step": 249753, "epoch": 2806, "lr": 1.5737267382940846e-06}
{"train_loss": 0.05532141029834747, "global_step": 249754, "epoch": 2806, "lr": 1.5735824367084485e-06}
{"train_loss": 0.02740578167140484, "global_step": 249755, "epoch": 2806, "lr": 1.5734381416331601e-06}
{"train_loss": 0.05435406416654587, "global_step": 249756, "epoch": 2806, "lr": 1.573293853068225e-06}
{"train_loss": 0.041609663516283035, "global_step": 249757, "epoch": 2806, "lr": 1.5731495710136823e-06}
{"train_loss": 0.011338476091623306, "global_step": 249758, "epoch": 2806, "lr": 1.5730052954695318e-06}
{"train_loss": 0.06869648396968842, "global_step": 249759, "epoch": 2806, "lr": 1.5728610264358013e-06}
{"train_loss": 0.040952276438474655, "global_step": 249760, "epoch": 2806, "lr": 1.5727167639125072e-06}
{"train_loss": 0.04321923479437828, "global_step": 249761, "epoch": 2806, "lr": 1.5725725078996723e-06}
{"train_loss": 0.04852786287665367, "global_step": 249762, "epoch": 2806, "lr": 1.5724282583973182e-06}
{"train_loss": 0.04943732172250748, "global_step": 249763, "epoch": 2806, "lr": 1.5722840154054564e-06}
{"train_loss": 0.052228525280952454, "global_step": 249764, "epoch": 2806, "lr": 1.5721397789241143e-06}
{"train_loss": 0.014666097238659859, "global_step": 249765, "epoch": 2806, "lr": 1.571995548953309e-06}
{"train_loss": 0.047812219709157944, "global_step": 249766, "epoch": 2806, "lr": 1.5718513254930512e-06}
{"train_loss": 0.02777988463640213, "global_step": 249767, "epoch": 2806, "lr": 1.57170710854338e-06}
{"train_loss": 0.05932687222957611, "global_step": 249768, "epoch": 2806, "lr": 1.5715628981042897e-06}
{"train_loss": 0.07800579071044922, "global_step": 249769, "epoch": 2806, "lr": 1.5714186941758135e-06}
{"train_loss": 0.0522015206515789, "global_step": 249770, "epoch": 2806, "lr": 1.5712744967579796e-06}
{"train_loss": 0.06548092514276505, "global_step": 249771, "epoch": 2806, "lr": 1.5711303058507876e-06}
{"train_loss": 0.04446423053741455, "global_step": 249772, "epoch": 2806, "lr": 1.570986121454271e-06}
{"train_loss": 0.040816061198711395, "global_step": 249773, "epoch": 2806, "lr": 1.5708419435684462e-06}
{"train_loss": 0.03136036545038223, "global_step": 249774, "epoch": 2806, "lr": 1.5706977721933302e-06}
{"train_loss": 0.05765543878078461, "global_step": 249775, "epoch": 2806, "lr": 1.5705536073289395e-06}
{"train_loss": 0.08009522408246994, "global_step": 249776, "epoch": 2806, "lr": 1.5704094489753074e-06}
{"train_loss": 0.02836558409035206, "global_step": 249777, "epoch": 2806, "lr": 1.5702652971324338e-06}
{"train_loss": 0.06555690616369247, "global_step": 249778, "epoch": 2806, "lr": 1.5701211518003578e-06}
{"train_loss": 0.02467285469174385, "global_step": 249779, "epoch": 2806, "lr": 1.569977012979079e-06}
{"train_loss": 0.04253020137548447, "global_step": 249780, "epoch": 2806, "lr": 1.5698328806686313e-06}
{"train_loss": 0.02051209658384323, "global_step": 249781, "epoch": 2806, "lr": 1.5696887548690254e-06}
{"train_loss": 0.006293793674558401, "global_step": 249782, "epoch": 2806, "lr": 1.569544635580289e-06}
{"train_loss": 0.030804188922047615, "global_step": 249783, "epoch": 2806, "lr": 1.5694005228024334e-06}
{"train_loss": 0.019569827243685722, "global_step": 249784, "epoch": 2806, "lr": 1.5692564165354862e-06}
{"train_loss": 0.038056109100580215, "global_step": 249785, "epoch": 2806, "lr": 1.5691123167794585e-06}
{"train_loss": 0.027689671143889427, "global_step": 249786, "epoch": 2806, "lr": 1.5689682235343784e-06}
{"train_loss": 0.06726281344890594, "global_step": 249787, "epoch": 2806, "lr": 1.568824136800251e-06}
{"train_loss": 0.029586000367999077, "global_step": 249788, "epoch": 2806, "lr": 1.5686800565771098e-06}
{"train_loss": 0.02420632913708687, "global_step": 249789, "epoch": 2806, "lr": 1.5685359828649716e-06}
{"train_loss": 0.055708013474941254, "global_step": 249790, "epoch": 2806, "lr": 1.5683919156638471e-06}
{"train_loss": 0.0261251088231802, "global_step": 249791, "epoch": 2806, "lr": 1.56824785497377e-06}
{"train_loss": 0.07387026399374008, "global_step": 249792, "epoch": 2806, "lr": 1.5681038007947457e-06}
{"train_loss": 0.0480278879404068, "global_step": 249793, "epoch": 2806, "lr": 1.5679597531268075e-06}
{"train_loss": 0.023192308843135834, "global_step": 249794, "epoch": 2806, "lr": 1.5678157119699554e-06}
{"train_loss": 0.018068360164761543, "global_step": 249795, "epoch": 2806, "lr": 1.5676716773242284e-06}
{"train_loss": 0.06233248487114906, "global_step": 249796, "epoch": 2806, "lr": 1.5675276491896374e-06}
{"train_loss": 0.05897480621933937, "global_step": 249797, "epoch": 2806, "lr": 1.567383627566199e-06}
{"train_loss": 0.011921081691980362, "global_step": 249798, "epoch": 2806, "lr": 1.5672396124539412e-06}
{"train_loss": 0.0207721758633852, "global_step": 249799, "epoch": 2806, "lr": 1.5670956038528695e-06}
{"train_loss": 0.045499734580516815, "global_step": 249800, "epoch": 2806, "lr": 1.566951601763017e-06}
{"train_loss": 0.05499235913157463, "global_step": 249801, "epoch": 2806, "lr": 1.566807606184395e-06}
{"train_loss": 0.04434714466333389, "global_step": 249802, "epoch": 2806, "lr": 1.5666636171170257e-06}
{"train_loss": 0.014807576313614845, "global_step": 249803, "epoch": 2806, "lr": 1.5665196345609313e-06}
{"train_loss": 0.03928089514374733, "global_step": 249804, "epoch": 2806, "lr": 1.5663756585161227e-06}
{"train_loss": 0.018145540729165077, "global_step": 249805, "epoch": 2806, "lr": 1.5662316889826278e-06}
{"train_loss": 0.02126406878232956, "global_step": 249806, "epoch": 2806, "lr": 1.5660877259604633e-06}
{"train_loss": 0.02558993361890316, "global_step": 249807, "epoch": 2806, "lr": 1.565943769449646e-06}
{"train_loss": 0.04355371743440628, "global_step": 249808, "epoch": 2806, "lr": 1.5657998194501978e-06}
{"train_loss": 0.024703316390514374, "global_step": 249809, "epoch": 2806, "lr": 1.5656558759621354e-06}
{"train_loss": 0.040573205798864365, "global_step": 249810, "epoch": 2806, "lr": 1.5655119389854866e-06}
{"train_loss": 0.036604009568691254, "global_step": 249811, "epoch": 2806, "lr": 1.565368008520257e-06}
{"train_loss": 0.03821982815861702, "global_step": 249812, "epoch": 2806, "lr": 1.5652240845664802e-06}
{"train_loss": 0.033316224813461304, "global_step": 249813, "epoch": 2806, "lr": 1.565080167124161e-06}
{"train_loss": 0.06166698411107063, "global_step": 249814, "epoch": 2806, "lr": 1.5649362561933334e-06}
{"train_loss": 0.051335886120796204, "global_step": 249815, "epoch": 2806, "lr": 1.5647923517740025e-06}
{"train_loss": 0.04567436873912811, "global_step": 249816, "epoch": 2806, "lr": 1.564648453866202e-06}
{"train_loss": 0.021780824288725853, "global_step": 249817, "epoch": 2806, "lr": 1.5645045624699373e-06}
{"train_loss": 0.0371977761387825, "global_step": 249818, "epoch": 2806, "lr": 1.5643606775852415e-06}
{"train_loss": 0.05114276334643364, "global_step": 249819, "epoch": 2806, "lr": 1.5642167992121204e-06}
{"train_loss": 0.0428289920091629, "global_step": 249820, "epoch": 2806, "lr": 1.5640729273506072e-06}
{"train_loss": 0.039602652192115784, "global_step": 249821, "epoch": 2806, "lr": 1.5639290620007074e-06}
{"train_loss": 0.04289827447154381, "global_step": 249822, "epoch": 2806, "lr": 1.5637852031624545e-06, "val_loss": 9.617165565490723}
{"train_loss": 0.028779171407222748, "global_step": 249823, "epoch": 2807, "lr": 1.5636413508358538e-06}
{"train_loss": 0.036577217280864716, "global_step": 249824, "epoch": 2807, "lr": 1.5634975050209332e-06}
{"train_loss": 0.04100850597023964, "global_step": 249825, "epoch": 2807, "lr": 1.5633536657177095e-06}
{"train_loss": 0.029208935797214508, "global_step": 249826, "epoch": 2807, "lr": 1.5632098329262047e-06}
{"train_loss": 0.04188353195786476, "global_step": 249827, "epoch": 2807, "lr": 1.5630660066464353e-06}
{"train_loss": 0.02995704859495163, "global_step": 249828, "epoch": 2807, "lr": 1.5629221868784184e-06}
{"train_loss": 0.02912314608693123, "global_step": 249829, "epoch": 2807, "lr": 1.5627783736221812e-06}
{"train_loss": 0.03965126350522041, "global_step": 249830, "epoch": 2807, "lr": 1.5626345668777297e-06}
{"train_loss": 0.027982451021671295, "global_step": 249831, "epoch": 2807, "lr": 1.5624907666451027e-06}
{"train_loss": 0.04523775354027748, "global_step": 249832, "epoch": 2807, "lr": 1.5623469729243001e-06}
{"train_loss": 0.03367544710636139, "global_step": 249833, "epoch": 2807, "lr": 1.5622031857153552e-06}
{"train_loss": 0.08539745211601257, "global_step": 249834, "epoch": 2807, "lr": 1.5620594050182735e-06}
{"train_loss": 0.023965362459421158, "global_step": 249835, "epoch": 2807, "lr": 1.561915630833094e-06}
{"train_loss": 0.030661359429359436, "global_step": 249836, "epoch": 2807, "lr": 1.5617718631598221e-06}
{"train_loss": 0.046717412769794464, "global_step": 249837, "epoch": 2807, "lr": 1.561628101998469e-06}
{"train_loss": 0.06678294390439987, "global_step": 249838, "epoch": 2807, "lr": 1.5614843473490736e-06}
{"train_loss": 0.05142015218734741, "global_step": 249839, "epoch": 2807, "lr": 1.561340599211647e-06}
{"train_loss": 0.03147294744849205, "global_step": 249840, "epoch": 2807, "lr": 1.5611968575862058e-06}
{"train_loss": 0.030842188745737076, "global_step": 249841, "epoch": 2807, "lr": 1.5610531224727665e-06}
{"train_loss": 0.029207341372966766, "global_step": 249842, "epoch": 2807, "lr": 1.5609093938713625e-06}
{"train_loss": 0.07117992639541626, "global_step": 249843, "epoch": 2807, "lr": 1.560765671781994e-06}
{"train_loss": 0.028589077293872833, "global_step": 249844, "epoch": 2807, "lr": 1.5606219562046997e-06}
{"train_loss": 0.030912021175026894, "global_step": 249845, "epoch": 2807, "lr": 1.5604782471394798e-06}
{"train_loss": 0.029709000140428543, "global_step": 249846, "epoch": 2807, "lr": 1.5603345445863726e-06}
{"train_loss": 0.012125427834689617, "global_step": 249847, "epoch": 2807, "lr": 1.5601908485453786e-06}
{"train_loss": 0.020903194323182106, "global_step": 249848, "epoch": 2807, "lr": 1.5600471590165366e-06}
{"train_loss": 0.033085305243730545, "global_step": 249849, "epoch": 2807, "lr": 1.5599034759998465e-06}
{"train_loss": 0.03529311344027519, "global_step": 249850, "epoch": 2807, "lr": 1.5597597994953417e-06}
{"train_loss": 0.09358061105012894, "global_step": 249851, "epoch": 2807, "lr": 1.5596161295030332e-06}
{"train_loss": 0.05995437502861023, "global_step": 249852, "epoch": 2807, "lr": 1.5594724660229488e-06}
{"train_loss": 0.05266138166189194, "global_step": 249853, "epoch": 2807, "lr": 1.5593288090550995e-06}
{"train_loss": 0.05943094193935394, "global_step": 249854, "epoch": 2807, "lr": 1.5591851585995132e-06}
{"train_loss": 0.057596929371356964, "global_step": 249855, "epoch": 2807, "lr": 1.5590415146561954e-06}
{"train_loss": 0.01669984869658947, "global_step": 249856, "epoch": 2807, "lr": 1.558897877225185e-06}
{"train_loss": 0.028253793716430664, "global_step": 249857, "epoch": 2807, "lr": 1.558754246306482e-06}
{"train_loss": 0.029923709109425545, "global_step": 249858, "epoch": 2807, "lr": 1.5586106219001196e-06}
{"train_loss": 0.0480036698281765, "global_step": 249859, "epoch": 2807, "lr": 1.5584670040061034e-06}
{"train_loss": 0.04801429435610771, "global_step": 249860, "epoch": 2807, "lr": 1.5583233926244612e-06}
{"train_loss": 0.023847442120313644, "global_step": 249861, "epoch": 2807, "lr": 1.5581797877552207e-06}
{"train_loss": 0.048066746443510056, "global_step": 249862, "epoch": 2807, "lr": 1.5580361893983874e-06}
{"train_loss": 0.010193614289164543, "global_step": 249863, "epoch": 2807, "lr": 1.5578925975539893e-06}
{"train_loss": 0.04715825244784355, "global_step": 249864, "epoch": 2807, "lr": 1.5577490122220372e-06}
{"train_loss": 0.07119359076023102, "global_step": 249865, "epoch": 2807, "lr": 1.5576054334025592e-06}
{"train_loss": 0.02692793309688568, "global_step": 249866, "epoch": 2807, "lr": 1.557461861095566e-06}
{"train_loss": 0.06657466292381287, "global_step": 249867, "epoch": 2807, "lr": 1.5573182953010856e-06}
{"train_loss": 0.03306102380156517, "global_step": 249868, "epoch": 2807, "lr": 1.5571747360191346e-06}
{"train_loss": 0.06651093810796738, "global_step": 249869, "epoch": 2807, "lr": 1.5570311832497297e-06}
{"train_loss": 0.08740533143281937, "global_step": 249870, "epoch": 2807, "lr": 1.5568876369928875e-06}
{"train_loss": 0.02427484467625618, "global_step": 249871, "epoch": 2807, "lr": 1.5567440972486358e-06}
{"train_loss": 0.06559497117996216, "global_step": 249872, "epoch": 2807, "lr": 1.5566005640169855e-06}
{"train_loss": 0.03473089635372162, "global_step": 249873, "epoch": 2807, "lr": 1.5564570372979703e-06}
{"train_loss": 0.031664006412029266, "global_step": 249874, "epoch": 2807, "lr": 1.55631351709159e-06}
{"train_loss": 0.05272163078188896, "global_step": 249875, "epoch": 2807, "lr": 1.556170003397872e-06}
{"train_loss": 0.009614119306206703, "global_step": 249876, "epoch": 2807, "lr": 1.5560264962168447e-06}
{"train_loss": 0.02095991186797619, "global_step": 249877, "epoch": 2807, "lr": 1.5558829955485077e-06}
{"train_loss": 0.05924815312027931, "global_step": 249878, "epoch": 2807, "lr": 1.5557395013928999e-06}
{"train_loss": 0.03921842202544212, "global_step": 249879, "epoch": 2807, "lr": 1.555596013750027e-06}
{"train_loss": 0.03894137963652611, "global_step": 249880, "epoch": 2807, "lr": 1.5554525326199222e-06}
{"train_loss": 0.03616301715373993, "global_step": 249881, "epoch": 2807, "lr": 1.5553090580025853e-06}
{"train_loss": 0.028746191412210464, "global_step": 249882, "epoch": 2807, "lr": 1.5551655898980554e-06}
{"train_loss": 0.04862375557422638, "global_step": 249883, "epoch": 2807, "lr": 1.555022128306338e-06}
{"train_loss": 0.04259171336889267, "global_step": 249884, "epoch": 2807, "lr": 1.5548786732274611e-06}
{"train_loss": 0.04720583185553551, "global_step": 249885, "epoch": 2807, "lr": 1.5547352246614355e-06}
{"train_loss": 0.01405036635696888, "global_step": 249886, "epoch": 2807, "lr": 1.554591782608289e-06}
{"train_loss": 0.0349724255502224, "global_step": 249887, "epoch": 2807, "lr": 1.5544483470680327e-06}
{"train_loss": 0.07118889689445496, "global_step": 249888, "epoch": 2807, "lr": 1.5543049180406999e-06}
{"train_loss": 0.05721618980169296, "global_step": 249889, "epoch": 2807, "lr": 1.5541614955262906e-06}
{"train_loss": 0.06695085763931274, "global_step": 249890, "epoch": 2807, "lr": 1.5540180795248382e-06}
{"train_loss": 0.0531732402741909, "global_step": 249891, "epoch": 2807, "lr": 1.5538746700363539e-06}
{"train_loss": 0.04325046390295029, "global_step": 249892, "epoch": 2807, "lr": 1.5537312670608706e-06}
{"train_loss": 0.0477202832698822, "global_step": 249893, "epoch": 2807, "lr": 1.5535878705983886e-06}
{"train_loss": 0.07725200802087784, "global_step": 249894, "epoch": 2807, "lr": 1.5534444806489412e-06}
{"train_loss": 0.03416458144783974, "global_step": 249895, "epoch": 2807, "lr": 1.553301097212534e-06}
{"train_loss": 0.04107605293393135, "global_step": 249896, "epoch": 2807, "lr": 1.5531577202892e-06}
{"train_loss": 0.0325019396841526, "global_step": 249897, "epoch": 2807, "lr": 1.5530143498789562e-06}
{"train_loss": 0.025326423346996307, "global_step": 249898, "epoch": 2807, "lr": 1.5528709859818136e-06}
{"train_loss": 0.018799681216478348, "global_step": 249899, "epoch": 2807, "lr": 1.5527276285978054e-06}
{"train_loss": 0.0335155613720417, "global_step": 249900, "epoch": 2807, "lr": 1.5525842777269317e-06}
{"train_loss": 0.06283245235681534, "global_step": 249901, "epoch": 2807, "lr": 1.5524409333692313e-06}
{"train_loss": 0.02706674486398697, "global_step": 249902, "epoch": 2807, "lr": 1.5522975955247098e-06}
{"train_loss": 0.027840832248330116, "global_step": 249903, "epoch": 2807, "lr": 1.5521542641934005e-06}
{"train_loss": 0.06080470234155655, "global_step": 249904, "epoch": 2807, "lr": 1.5520109393753035e-06}
{"train_loss": 0.010823721997439861, "global_step": 249905, "epoch": 2807, "lr": 1.5518676210704518e-06}
{"train_loss": 0.05139676854014397, "global_step": 249906, "epoch": 2807, "lr": 1.5517243092788569e-06}
{"train_loss": 0.059203945100307465, "global_step": 249907, "epoch": 2807, "lr": 1.5515810040005519e-06}
{"train_loss": 0.05748818442225456, "global_step": 249908, "epoch": 2807, "lr": 1.5514377052355367e-06}
{"train_loss": 0.030930908396840096, "global_step": 249909, "epoch": 2807, "lr": 1.5512944129838448e-06}
{"train_loss": 0.04030400142073631, "global_step": 249910, "epoch": 2807, "lr": 1.551151127245487e-06}
{"train_loss": 0.04167398921308223, "global_step": 249911, "epoch": 2807, "lr": 1.5510078480204916e-06, "val_loss": 9.490036964416504}
{"train_loss": 0.07580527663230896, "global_step": 249912, "epoch": 2808, "lr": 1.550864575308869e-06}
{"train_loss": 0.03660449758172035, "global_step": 249913, "epoch": 2808, "lr": 1.5507213091106476e-06}
{"train_loss": 0.019790692254900932, "global_step": 249914, "epoch": 2808, "lr": 1.5505780494258381e-06}
{"train_loss": 0.04020834341645241, "global_step": 249915, "epoch": 2808, "lr": 1.5504347962544574e-06}
{"train_loss": 0.009959963150322437, "global_step": 249916, "epoch": 2808, "lr": 1.5502915495965386e-06}
{"train_loss": 0.06205474212765694, "global_step": 249917, "epoch": 2808, "lr": 1.5501483094520875e-06}
{"train_loss": 0.03810325637459755, "global_step": 249918, "epoch": 2808, "lr": 1.5500050758211316e-06}
{"train_loss": 0.06327565014362335, "global_step": 249919, "epoch": 2808, "lr": 1.549861848703682e-06}
{"train_loss": 0.03753220662474632, "global_step": 249920, "epoch": 2808, "lr": 1.5497186280997667e-06}
{"train_loss": 0.05374044179916382, "global_step": 249921, "epoch": 2808, "lr": 1.5495754140094022e-06}
{"train_loss": 0.036275964230298996, "global_step": 249922, "epoch": 2808, "lr": 1.549432206432605e-06}
{"train_loss": 0.06912348419427872, "global_step": 249923, "epoch": 2808, "lr": 1.549289005369392e-06}
{"train_loss": 0.058144815266132355, "global_step": 249924, "epoch": 2808, "lr": 1.5491458108197965e-06}
{"train_loss": 0.03884562477469444, "global_step": 249925, "epoch": 2808, "lr": 1.5490026227838183e-06}
{"train_loss": 0.03867724910378456, "global_step": 249926, "epoch": 2808, "lr": 1.548859441261491e-06}
{"train_loss": 0.050833504647016525, "global_step": 249927, "epoch": 2808, "lr": 1.5487162662528254e-06}
{"train_loss": 0.056458860635757446, "global_step": 249928, "epoch": 2808, "lr": 1.5485730977578495e-06}
{"train_loss": 0.08843924105167389, "global_step": 249929, "epoch": 2808, "lr": 1.5484299357765741e-06}
{"train_loss": 0.04057029262185097, "global_step": 249930, "epoch": 2808, "lr": 1.5482867803090217e-06}
{"train_loss": 0.016793332993984222, "global_step": 249931, "epoch": 2808, "lr": 1.5481436313552088e-06}
{"train_loss": 0.05102415755391121, "global_step": 249932, "epoch": 2808, "lr": 1.5480004889151633e-06}
{"train_loss": 0.08183085918426514, "global_step": 249933, "epoch": 2808, "lr": 1.5478573529888963e-06}
{"train_loss": 0.028608474880456924, "global_step": 249934, "epoch": 2808, "lr": 1.5477142235764242e-06}
{"train_loss": 0.0623779371380806, "global_step": 249935, "epoch": 2808, "lr": 1.5475711006777805e-06}
{"train_loss": 0.030596772208809853, "global_step": 249936, "epoch": 2808, "lr": 1.5474279842929706e-06}
{"train_loss": 0.04564319923520088, "global_step": 249937, "epoch": 2808, "lr": 1.5472848744220225e-06}
{"train_loss": 0.051989007741212845, "global_step": 249938, "epoch": 2808, "lr": 1.5471417710649416e-06}
{"train_loss": 0.06097126752138138, "global_step": 249939, "epoch": 2808, "lr": 1.5469986742217668e-06}
{"train_loss": 0.03611040487885475, "global_step": 249940, "epoch": 2808, "lr": 1.5468555838925037e-06}
{"train_loss": 0.02856271341443062, "global_step": 249941, "epoch": 2808, "lr": 1.5467125000771742e-06}
{"train_loss": 0.06186142563819885, "global_step": 249942, "epoch": 2808, "lr": 1.546569422775801e-06}
{"train_loss": 0.04961664602160454, "global_step": 249943, "epoch": 2808, "lr": 1.5464263519884003e-06}
{"train_loss": 0.029467323794960976, "global_step": 249944, "epoch": 2808, "lr": 1.5462832877149892e-06}
{"train_loss": 0.0916559025645256, "global_step": 249945, "epoch": 2808, "lr": 1.5461402299555949e-06}
{"train_loss": 0.07511322945356369, "global_step": 249946, "epoch": 2808, "lr": 1.545997178710229e-06}
{"train_loss": 0.024291737005114555, "global_step": 249947, "epoch": 2808, "lr": 1.5458541339789135e-06}
{"train_loss": 0.026614006608724594, "global_step": 249948, "epoch": 2808, "lr": 1.545711095761665e-06}
{"train_loss": 0.0254583228379488, "global_step": 249949, "epoch": 2808, "lr": 1.5455680640585112e-06}
{"train_loss": 0.05203958973288536, "global_step": 249950, "epoch": 2808, "lr": 1.545425038869458e-06}
{"train_loss": 0.03665915131568909, "global_step": 249951, "epoch": 2808, "lr": 1.5452820201945328e-06}
{"train_loss": 0.033165838569402695, "global_step": 249952, "epoch": 2808, "lr": 1.5451390080337635e-06}
{"train_loss": 0.046274155378341675, "global_step": 249953, "epoch": 2808, "lr": 1.5449960023871446e-06}
{"train_loss": 0.024926455691456795, "global_step": 249954, "epoch": 2808, "lr": 1.5448530032547203e-06}
{"train_loss": 0.04763541743159294, "global_step": 249955, "epoch": 2808, "lr": 1.5447100106364965e-06}
{"train_loss": 0.08655145019292831, "global_step": 249956, "epoch": 2808, "lr": 1.5445670245325006e-06}
{"train_loss": 0.03692268952727318, "global_step": 249957, "epoch": 2808, "lr": 1.5444240449427383e-06}
{"train_loss": 0.09537853300571442, "global_step": 249958, "epoch": 2808, "lr": 1.5442810718672485e-06}
{"train_loss": 0.04697602242231369, "global_step": 249959, "epoch": 2808, "lr": 1.5441381053060311e-06}
{"train_loss": 0.03831622749567032, "global_step": 249960, "epoch": 2808, "lr": 1.5439951452591195e-06}
{"train_loss": 0.025460321456193924, "global_step": 249961, "epoch": 2808, "lr": 1.5438521917265193e-06}
{"train_loss": 0.024416502565145493, "global_step": 249962, "epoch": 2808, "lr": 1.5437092447082635e-06}
{"train_loss": 0.02293098159134388, "global_step": 249963, "epoch": 2808, "lr": 1.5435663042043635e-06}
{"train_loss": 0.04772617295384407, "global_step": 249964, "epoch": 2808, "lr": 1.5434233702148414e-06}
{"train_loss": 0.02743406407535076, "global_step": 249965, "epoch": 2808, "lr": 1.543280442739714e-06}
{"train_loss": 0.04689374566078186, "global_step": 249966, "epoch": 2808, "lr": 1.5431375217790034e-06}
{"train_loss": 0.01818283088505268, "global_step": 249967, "epoch": 2808, "lr": 1.542994607332726e-06}
{"train_loss": 0.055425986647605896, "global_step": 249968, "epoch": 2808, "lr": 1.542851699400899e-06}
{"train_loss": 0.03473059460520744, "global_step": 249969, "epoch": 2808, "lr": 1.5427087979835553e-06}
{"train_loss": 0.0477469265460968, "global_step": 249970, "epoch": 2808, "lr": 1.5425659030806949e-06}
{"train_loss": 0.06518056243658066, "global_step": 249971, "epoch": 2808, "lr": 1.5424230146923513e-06}
{"train_loss": 0.025123197585344315, "global_step": 249972, "epoch": 2808, "lr": 1.5422801328185298e-06}
{"train_loss": 0.0829429104924202, "global_step": 249973, "epoch": 2808, "lr": 1.5421372574592697e-06}
{"train_loss": 0.05494367703795433, "global_step": 249974, "epoch": 2808, "lr": 1.5419943886145705e-06}
{"train_loss": 0.09524841606616974, "global_step": 249975, "epoch": 2808, "lr": 1.54185152628446e-06}
{"train_loss": 0.015676990151405334, "global_step": 249976, "epoch": 2808, "lr": 1.541708670468961e-06}
{"train_loss": 0.01867331936955452, "global_step": 249977, "epoch": 2808, "lr": 1.5415658211680894e-06}
{"train_loss": 0.04432455822825432, "global_step": 249978, "epoch": 2808, "lr": 1.5414229783818567e-06}
{"train_loss": 0.05586663633584976, "global_step": 249979, "epoch": 2808, "lr": 1.5412801421102963e-06}
{"train_loss": 0.015056315809488297, "global_step": 249980, "epoch": 2808, "lr": 1.5411373123534135e-06}
{"train_loss": 0.03942319378256798, "global_step": 249981, "epoch": 2808, "lr": 1.5409944891112416e-06}
{"train_loss": 0.02796265482902527, "global_step": 249982, "epoch": 2808, "lr": 1.5408516723837864e-06}
{"train_loss": 0.04117750748991966, "global_step": 249983, "epoch": 2808, "lr": 1.5407088621710753e-06}
{"train_loss": 0.05637463927268982, "global_step": 249984, "epoch": 2808, "lr": 1.5405660584731196e-06}
{"train_loss": 0.09138736873865128, "global_step": 249985, "epoch": 2808, "lr": 1.5404232612899528e-06}
{"train_loss": 0.04574553295969963, "global_step": 249986, "epoch": 2808, "lr": 1.54028047062158e-06}
{"train_loss": 0.020179079845547676, "global_step": 249987, "epoch": 2808, "lr": 1.540137686468035e-06}
{"train_loss": 0.05597482621669769, "global_step": 249988, "epoch": 2808, "lr": 1.5399949088293175e-06}
{"train_loss": 0.08778830617666245, "global_step": 249989, "epoch": 2808, "lr": 1.539852137705461e-06}
{"train_loss": 0.06277314573526382, "global_step": 249990, "epoch": 2808, "lr": 1.5397093730964818e-06}
{"train_loss": 0.03545651212334633, "global_step": 249991, "epoch": 2808, "lr": 1.5395666150023969e-06}
{"train_loss": 0.022677579894661903, "global_step": 249992, "epoch": 2808, "lr": 1.5394238634232284e-06}
{"train_loss": 0.025673162192106247, "global_step": 249993, "epoch": 2808, "lr": 1.5392811183589873e-06}
{"train_loss": 0.08254843950271606, "global_step": 249994, "epoch": 2808, "lr": 1.539138379809707e-06}
{"train_loss": 0.05780545249581337, "global_step": 249995, "epoch": 2808, "lr": 1.538995647775393e-06}
{"train_loss": 0.032240401953458786, "global_step": 249996, "epoch": 2808, "lr": 1.5388529222560732e-06}
{"train_loss": 0.043323859572410583, "global_step": 249997, "epoch": 2808, "lr": 1.5387102032517643e-06}
{"train_loss": 0.05694335326552391, "global_step": 249998, "epoch": 2808, "lr": 1.5385674907624825e-06}
{"train_loss": 0.024750737473368645, "global_step": 249999, "epoch": 2808, "lr": 1.5384247847882504e-06}
{"train_loss": 0.04687384558904372, "global_step": 250000, "epoch": 2808, "lr": 1.5382820853290902e-06, "val_loss": 9.636140823364258}
{"train_loss": 0.021633928641676903, "global_step": 250001, "epoch": 2809, "lr": 1.538139392385013e-06}
{"train_loss": 0.06403893232345581, "global_step": 250002, "epoch": 2809, "lr": 1.5379967059560462e-06}
{"train_loss": 0.029531056061387062, "global_step": 250003, "epoch": 2809, "lr": 1.5378540260422013e-06}
{"train_loss": 0.09503038972616196, "global_step": 250004, "epoch": 2809, "lr": 1.5377113526435005e-06}
{"train_loss": 0.04441128671169281, "global_step": 250005, "epoch": 2809, "lr": 1.537568685759966e-06}
{"train_loss": 0.01722199097275734, "global_step": 250006, "epoch": 2809, "lr": 1.537426025391614e-06}
{"train_loss": 0.09782581776380539, "global_step": 250007, "epoch": 2809, "lr": 1.5372833715384671e-06}
{"train_loss": 0.05390739068388939, "global_step": 250008, "epoch": 2809, "lr": 1.5371407242005365e-06}
{"train_loss": 0.01804644800722599, "global_step": 250009, "epoch": 2809, "lr": 1.5369980833778497e-06}
{"train_loss": 0.012663248926401138, "global_step": 250010, "epoch": 2809, "lr": 1.5368554490704235e-06}
{"train_loss": 0.03428410738706589, "global_step": 250011, "epoch": 2809, "lr": 1.5367128212782745e-06}
{"train_loss": 0.026740146800875664, "global_step": 250012, "epoch": 2809, "lr": 1.5365702000014249e-06}
{"train_loss": 0.04827946797013283, "global_step": 250013, "epoch": 2809, "lr": 1.536427585239897e-06}
{"train_loss": 0.017534082755446434, "global_step": 250014, "epoch": 2809, "lr": 1.536284976993696e-06}
{"train_loss": 0.058392442762851715, "global_step": 250015, "epoch": 2809, "lr": 1.5361423752628613e-06}
{"train_loss": 0.05925232172012329, "global_step": 250016, "epoch": 2809, "lr": 1.5359997800473924e-06}
{"train_loss": 0.02802678756415844, "global_step": 250017, "epoch": 2809, "lr": 1.5358571913473286e-06}
{"train_loss": 0.01938769593834877, "global_step": 250018, "epoch": 2809, "lr": 1.535714609162664e-06}
{"train_loss": 0.017922203987836838, "global_step": 250019, "epoch": 2809, "lr": 1.5355720334934431e-06}
{"train_loss": 0.03141963481903076, "global_step": 250020, "epoch": 2809, "lr": 1.535429464339666e-06}
{"train_loss": 0.023720502853393555, "global_step": 250021, "epoch": 2809, "lr": 1.5352869017013716e-06}
{"train_loss": 0.05482754856348038, "global_step": 250022, "epoch": 2809, "lr": 1.5351443455785542e-06}
{"train_loss": 0.061180196702480316, "global_step": 250023, "epoch": 2809, "lr": 1.5350017959712471e-06}
{"train_loss": 0.030790625140070915, "global_step": 250024, "epoch": 2809, "lr": 1.534859252879478e-06}
{"train_loss": 0.05606311559677124, "global_step": 250025, "epoch": 2809, "lr": 1.5347167163032528e-06}
{"train_loss": 0.02026544138789177, "global_step": 250026, "epoch": 2809, "lr": 1.5345741862425877e-06}
{"train_loss": 0.06212226301431656, "global_step": 250027, "epoch": 2809, "lr": 1.5344316626975163e-06}
{"train_loss": 0.07671289145946503, "global_step": 250028, "epoch": 2809, "lr": 1.534289145668044e-06}
{"train_loss": 0.05599468946456909, "global_step": 250029, "epoch": 2809, "lr": 1.5341466351541989e-06}
{"train_loss": 0.07533412426710129, "global_step": 250030, "epoch": 2809, "lr": 1.5340041311560028e-06}
{"train_loss": 0.015289517119526863, "global_step": 250031, "epoch": 2809, "lr": 1.5338616336734557e-06}
{"train_loss": 0.04687781259417534, "global_step": 250032, "epoch": 2809, "lr": 1.5337191427066023e-06}
{"train_loss": 0.019305802881717682, "global_step": 250033, "epoch": 2809, "lr": 1.5335766582554423e-06}
{"train_loss": 0.046514593064785004, "global_step": 250034, "epoch": 2809, "lr": 1.5334341803200037e-06}
{"train_loss": 0.01776753179728985, "global_step": 250035, "epoch": 2809, "lr": 1.5332917089003028e-06}
{"train_loss": 0.04547324776649475, "global_step": 250036, "epoch": 2809, "lr": 1.533149243996368e-06}
{"train_loss": 0.03411952033638954, "global_step": 250037, "epoch": 2809, "lr": 1.5330067856081986e-06}
{"train_loss": 0.02219250053167343, "global_step": 250038, "epoch": 2809, "lr": 1.5328643337358394e-06}
{"train_loss": 0.056125055998563766, "global_step": 250039, "epoch": 2809, "lr": 1.5327218883792848e-06}
{"train_loss": 0.029110467061400414, "global_step": 250040, "epoch": 2809, "lr": 1.5325794495385682e-06}
{"train_loss": 0.03781691566109657, "global_step": 250041, "epoch": 2809, "lr": 1.5324370172137058e-06}
{"train_loss": 0.05455252155661583, "global_step": 250042, "epoch": 2809, "lr": 1.5322945914047148e-06}
{"train_loss": 0.039403147995471954, "global_step": 250043, "epoch": 2809, "lr": 1.5321521721116172e-06}
{"train_loss": 0.022742217406630516, "global_step": 250044, "epoch": 2809, "lr": 1.5320097593344296e-06}
{"train_loss": 0.0731419250369072, "global_step": 250045, "epoch": 2809, "lr": 1.5318673530731798e-06}
{"train_loss": 0.013318045996129513, "global_step": 250046, "epoch": 2809, "lr": 1.5317249533278677e-06}
{"train_loss": 0.05022888630628586, "global_step": 250047, "epoch": 2809, "lr": 1.5315825600985378e-06}
{"train_loss": 0.0241255946457386, "global_step": 250048, "epoch": 2809, "lr": 1.5314401733851846e-06}
{"train_loss": 0.030740266665816307, "global_step": 250049, "epoch": 2809, "lr": 1.531297793187847e-06}
{"train_loss": 0.06179478019475937, "global_step": 250050, "epoch": 2809, "lr": 1.5311554195065247e-06}
{"train_loss": 0.05025836452841759, "global_step": 250051, "epoch": 2809, "lr": 1.5310130523412625e-06}
{"train_loss": 0.08485570549964905, "global_step": 250052, "epoch": 2809, "lr": 1.5308706916920546e-06}
{"train_loss": 0.0554586723446846, "global_step": 250053, "epoch": 2809, "lr": 1.5307283375589344e-06}
{"train_loss": 0.06247810274362564, "global_step": 250054, "epoch": 2809, "lr": 1.5305859899419128e-06}
{"train_loss": 0.04298700764775276, "global_step": 250055, "epoch": 2809, "lr": 1.5304436488410179e-06}
{"train_loss": 0.030612966045737267, "global_step": 250056, "epoch": 2809, "lr": 1.5303013142562606e-06}
{"train_loss": 0.05109257623553276, "global_step": 250057, "epoch": 2809, "lr": 1.5301589861876687e-06}
{"train_loss": 0.05531487986445427, "global_step": 250058, "epoch": 2809, "lr": 1.5300166646352477e-06}
{"train_loss": 0.039693232625722885, "global_step": 250059, "epoch": 2809, "lr": 1.5298743495990253e-06}
{"train_loss": 0.03046518564224243, "global_step": 250060, "epoch": 2809, "lr": 1.5297320410790295e-06}
{"train_loss": 0.020155299454927444, "global_step": 250061, "epoch": 2809, "lr": 1.5295897390752655e-06}
{"train_loss": 0.07384684681892395, "global_step": 250062, "epoch": 2809, "lr": 1.5294474435877614e-06}
{"train_loss": 0.017306338995695114, "global_step": 250063, "epoch": 2809, "lr": 1.5293051546165281e-06}
{"train_loss": 0.0379757285118103, "global_step": 250064, "epoch": 2809, "lr": 1.529162872161588e-06}
{"train_loss": 0.061255358159542084, "global_step": 250065, "epoch": 2809, "lr": 1.5290205962229686e-06}
{"train_loss": 0.07449363172054291, "global_step": 250066, "epoch": 2809, "lr": 1.5288783268006702e-06}
{"train_loss": 0.03958472982048988, "global_step": 250067, "epoch": 2809, "lr": 1.5287360638947312e-06}
{"train_loss": 0.06523711234331131, "global_step": 250068, "epoch": 2809, "lr": 1.5285938075051687e-06}
{"train_loss": 0.02971571683883667, "global_step": 250069, "epoch": 2809, "lr": 1.5284515576319825e-06}
{"train_loss": 0.055056605488061905, "global_step": 250070, "epoch": 2809, "lr": 1.528309314275217e-06}
{"train_loss": 0.05352766066789627, "global_step": 250071, "epoch": 2809, "lr": 1.5281670774348667e-06}
{"train_loss": 0.03757159411907196, "global_step": 250072, "epoch": 2809, "lr": 1.528024847110976e-06}
{"train_loss": 0.023914875462651253, "global_step": 250073, "epoch": 2809, "lr": 1.5278826233035393e-06}
{"train_loss": 0.09484773874282837, "global_step": 250074, "epoch": 2809, "lr": 1.5277404060126011e-06}
{"train_loss": 0.03735573589801788, "global_step": 250075, "epoch": 2809, "lr": 1.5275981952381557e-06}
{"train_loss": 0.081121526658535, "global_step": 250076, "epoch": 2809, "lr": 1.5274559909802422e-06}
{"train_loss": 0.035811133682727814, "global_step": 250077, "epoch": 2809, "lr": 1.527313793238866e-06}
{"train_loss": 0.042889971286058426, "global_step": 250078, "epoch": 2809, "lr": 1.5271716020140492e-06}
{"train_loss": 0.054683152586221695, "global_step": 250079, "epoch": 2809, "lr": 1.5270294173058198e-06}
{"train_loss": 0.05734844505786896, "global_step": 250080, "epoch": 2809, "lr": 1.5268872391141886e-06}
{"train_loss": 0.06783576309680939, "global_step": 250081, "epoch": 2809, "lr": 1.5267450674391781e-06}
{"train_loss": 0.08125786483287811, "global_step": 250082, "epoch": 2809, "lr": 1.5266029022808048e-06}
{"train_loss": 0.04729309678077698, "global_step": 250083, "epoch": 2809, "lr": 1.526460743639091e-06}
{"train_loss": 0.03886612504720688, "global_step": 250084, "epoch": 2809, "lr": 1.5263185915140477e-06}
{"train_loss": 0.05427877977490425, "global_step": 250085, "epoch": 2809, "lr": 1.5261764459057083e-06}
{"train_loss": 0.0372629389166832, "global_step": 250086, "epoch": 2809, "lr": 1.5260343068140726e-06}
{"train_loss": 0.03735283762216568, "global_step": 250087, "epoch": 2809, "lr": 1.5258921742391852e-06}
{"train_loss": 0.053525276482105255, "global_step": 250088, "epoch": 2809, "lr": 1.525750048181035e-06}
{"train_loss": 0.04487539094276308, "global_step": 250089, "epoch": 2809, "lr": 1.5256079286396719e-06, "val_loss": 9.551773071289062}
{"train_loss": 0.06406885385513306, "global_step": 250090, "epoch": 2810, "lr": 1.5254658156150904e-06}
{"train_loss": 0.08181975781917572, "global_step": 250091, "epoch": 2810, "lr": 1.5253237091073235e-06}
{"train_loss": 0.06614737957715988, "global_step": 250092, "epoch": 2810, "lr": 1.5251816091163829e-06}
{"train_loss": 0.03359370306134224, "global_step": 250093, "epoch": 2810, "lr": 1.5250395156422904e-06}
{"train_loss": 0.036444686353206635, "global_step": 250094, "epoch": 2810, "lr": 1.5248974286850682e-06}
{"train_loss": 0.04868460074067116, "global_step": 250095, "epoch": 2810, "lr": 1.5247553482447275e-06}
{"train_loss": 0.017219645902514458, "global_step": 250096, "epoch": 2810, "lr": 1.5246132743213015e-06}
{"train_loss": 0.03578723967075348, "global_step": 250097, "epoch": 2810, "lr": 1.5244712069147904e-06}
{"train_loss": 0.02589278295636177, "global_step": 250098, "epoch": 2810, "lr": 1.524329146025233e-06}
{"train_loss": 0.04144289717078209, "global_step": 250099, "epoch": 2810, "lr": 1.5241870916526292e-06}
{"train_loss": 0.09533959627151489, "global_step": 250100, "epoch": 2810, "lr": 1.5240450437970177e-06}
{"train_loss": 0.049552541226148605, "global_step": 250101, "epoch": 2810, "lr": 1.5239030024584045e-06}
{"train_loss": 0.04444439336657524, "global_step": 250102, "epoch": 2810, "lr": 1.5237609676368058e-06}
{"train_loss": 0.03827008232474327, "global_step": 250103, "epoch": 2810, "lr": 1.5236189393322553e-06}
{"train_loss": 0.02195131964981556, "global_step": 250104, "epoch": 2810, "lr": 1.5234769175447584e-06}
{"train_loss": 0.04587217792868614, "global_step": 250105, "epoch": 2810, "lr": 1.5233349022743426e-06}
{"train_loss": 0.029912933707237244, "global_step": 250106, "epoch": 2810, "lr": 1.5231928935210193e-06}
{"train_loss": 0.01672295294702053, "global_step": 250107, "epoch": 2810, "lr": 1.5230508912848107e-06}
{"train_loss": 0.03383065387606621, "global_step": 250108, "epoch": 2810, "lr": 1.5229088955657444e-06}
{"train_loss": 0.03874291479587555, "global_step": 250109, "epoch": 2810, "lr": 1.5227669063638206e-06}
{"train_loss": 0.032606638967990875, "global_step": 250110, "epoch": 2810, "lr": 1.5226249236790834e-06}
{"train_loss": 0.01701771281659603, "global_step": 250111, "epoch": 2810, "lr": 1.5224829475115275e-06}
{"train_loss": 0.027827637270092964, "global_step": 250112, "epoch": 2810, "lr": 1.5223409778611919e-06}
{"train_loss": 0.02364235185086727, "global_step": 250113, "epoch": 2810, "lr": 1.5221990147280763e-06}
{"train_loss": 0.03264511004090309, "global_step": 250114, "epoch": 2810, "lr": 1.522057058112214e-06}
{"train_loss": 0.023113789036870003, "global_step": 250115, "epoch": 2810, "lr": 1.5219151080136217e-06}
{"train_loss": 0.03743450716137886, "global_step": 250116, "epoch": 2810, "lr": 1.5217731644323163e-06}
{"train_loss": 0.03351837396621704, "global_step": 250117, "epoch": 2810, "lr": 1.5216312273683198e-06}
{"train_loss": 0.031059082597494125, "global_step": 250118, "epoch": 2810, "lr": 1.5214892968216487e-06}
{"train_loss": 0.038594331592321396, "global_step": 250119, "epoch": 2810, "lr": 1.52134737279232e-06}
{"train_loss": 0.025624314323067665, "global_step": 250120, "epoch": 2810, "lr": 1.5212054552803555e-06}
{"train_loss": 0.01670973375439644, "global_step": 250121, "epoch": 2810, "lr": 1.521063544285778e-06}
{"train_loss": 0.028399944305419922, "global_step": 250122, "epoch": 2810, "lr": 1.520921639808598e-06}
{"train_loss": 0.06367266923189163, "global_step": 250123, "epoch": 2810, "lr": 1.5207797418488435e-06}
{"train_loss": 0.03945453092455864, "global_step": 250124, "epoch": 2810, "lr": 1.5206378504065256e-06}
{"train_loss": 0.015475024469196796, "global_step": 250125, "epoch": 2810, "lr": 1.520495965481672e-06}
{"train_loss": 0.02937776781618595, "global_step": 250126, "epoch": 2810, "lr": 1.5203540870742883e-06}
{"train_loss": 0.04547097533941269, "global_step": 250127, "epoch": 2810, "lr": 1.5202122151844133e-06}
{"train_loss": 0.013303504325449467, "global_step": 250128, "epoch": 2810, "lr": 1.5200703498120473e-06}
{"train_loss": 0.012871123850345612, "global_step": 250129, "epoch": 2810, "lr": 1.5199284909572176e-06}
{"train_loss": 0.03420249745249748, "global_step": 250130, "epoch": 2810, "lr": 1.5197866386199412e-06}
{"train_loss": 0.030950147658586502, "global_step": 250131, "epoch": 2810, "lr": 1.51964479280024e-06}
{"train_loss": 0.06605613231658936, "global_step": 250132, "epoch": 2810, "lr": 1.5195029534981365e-06}
{"train_loss": 0.038754016160964966, "global_step": 250133, "epoch": 2810, "lr": 1.519361120713636e-06}
{"train_loss": 0.025630423799157143, "global_step": 250134, "epoch": 2810, "lr": 1.519219294446772e-06}
{"train_loss": 0.03574308007955551, "global_step": 250135, "epoch": 2810, "lr": 1.5190774746975556e-06}
{"train_loss": 0.028448021039366722, "global_step": 250136, "epoch": 2810, "lr": 1.5189356614660145e-06}
{"train_loss": 0.04000912234187126, "global_step": 250137, "epoch": 2810, "lr": 1.5187938547521542e-06}
{"train_loss": 0.05490122735500336, "global_step": 250138, "epoch": 2810, "lr": 1.518652054556008e-06}
{"train_loss": 0.04264673590660095, "global_step": 250139, "epoch": 2810, "lr": 1.5185102608775815e-06}
{"train_loss": 0.029965590685606003, "global_step": 250140, "epoch": 2810, "lr": 1.518368473716908e-06}
{"train_loss": 0.03246363252401352, "global_step": 250141, "epoch": 2810, "lr": 1.518226693073993e-06}
{"train_loss": 0.036840781569480896, "global_step": 250142, "epoch": 2810, "lr": 1.5180849189488589e-06}
{"train_loss": 0.04120328277349472, "global_step": 250143, "epoch": 2810, "lr": 1.5179431513415388e-06}
{"train_loss": 0.07328087836503983, "global_step": 250144, "epoch": 2810, "lr": 1.517801390252027e-06}
{"train_loss": 0.0264102965593338, "global_step": 250145, "epoch": 2810, "lr": 1.517659635680363e-06}
{"train_loss": 0.03975626826286316, "global_step": 250146, "epoch": 2810, "lr": 1.5175178876265628e-06}
{"train_loss": 0.03496340662240982, "global_step": 250147, "epoch": 2810, "lr": 1.5173761460906323e-06}
{"train_loss": 0.026588231325149536, "global_step": 250148, "epoch": 2810, "lr": 1.5172344110726045e-06}
{"train_loss": 0.08801144361495972, "global_step": 250149, "epoch": 2810, "lr": 1.5170926825724907e-06}
{"train_loss": 0.04969814047217369, "global_step": 250150, "epoch": 2810, "lr": 1.5169509605903132e-06}
{"train_loss": 0.05507488176226616, "global_step": 250151, "epoch": 2810, "lr": 1.516809245126094e-06}
{"train_loss": 0.03937477245926857, "global_step": 250152, "epoch": 2810, "lr": 1.5166675361798444e-06}
{"train_loss": 0.053116824477910995, "global_step": 250153, "epoch": 2810, "lr": 1.516525833751592e-06}
{"train_loss": 0.0472160167992115, "global_step": 250154, "epoch": 2810, "lr": 1.5163841378413478e-06}
{"train_loss": 0.036615993827581406, "global_step": 250155, "epoch": 2810, "lr": 1.51624244844914e-06}
{"train_loss": 0.040368448942899704, "global_step": 250156, "epoch": 2810, "lr": 1.5161007655749792e-06}
{"train_loss": 0.056130506098270416, "global_step": 250157, "epoch": 2810, "lr": 1.5159590892188935e-06}
{"train_loss": 0.035372305661439896, "global_step": 250158, "epoch": 2810, "lr": 1.5158174193808882e-06}
{"train_loss": 0.05750499293208122, "global_step": 250159, "epoch": 2810, "lr": 1.5156757560609913e-06}
{"train_loss": 0.05394045263528824, "global_step": 250160, "epoch": 2810, "lr": 1.5155340992592248e-06}
{"train_loss": 0.028794948011636734, "global_step": 250161, "epoch": 2810, "lr": 1.5153924489756e-06}
{"train_loss": 0.04309888929128647, "global_step": 250162, "epoch": 2810, "lr": 1.5152508052101388e-06}
{"train_loss": 0.021186960861086845, "global_step": 250163, "epoch": 2810, "lr": 1.5151091679628694e-06}
{"train_loss": 0.06276555359363556, "global_step": 250164, "epoch": 2810, "lr": 1.5149675372337912e-06}
{"train_loss": 0.04043973609805107, "global_step": 250165, "epoch": 2810, "lr": 1.5148259130229435e-06}
{"train_loss": 0.04227996990084648, "global_step": 250166, "epoch": 2810, "lr": 1.514684295330332e-06}
{"train_loss": 0.05067294090986252, "global_step": 250167, "epoch": 2810, "lr": 1.514542684155984e-06}
{"train_loss": 0.02721482887864113, "global_step": 250168, "epoch": 2810, "lr": 1.5144010794999109e-06}
{"train_loss": 0.037375498563051224, "global_step": 250169, "epoch": 2810, "lr": 1.5142594813621347e-06}
{"train_loss": 0.04040732607245445, "global_step": 250170, "epoch": 2810, "lr": 1.5141178897426777e-06}
{"train_loss": 0.037593673914670944, "global_step": 250171, "epoch": 2810, "lr": 1.5139763046415568e-06}
{"train_loss": 0.03815526142716408, "global_step": 250172, "epoch": 2810, "lr": 1.5138347260587938e-06}
{"train_loss": 0.03856592997908592, "global_step": 250173, "epoch": 2810, "lr": 1.5136931539944e-06}
{"train_loss": 0.07335053384304047, "global_step": 250174, "epoch": 2810, "lr": 1.513551588448403e-06}
{"train_loss": 0.012973717413842678, "global_step": 250175, "epoch": 2810, "lr": 1.5134100294208143e-06}
{"train_loss": 0.05804283916950226, "global_step": 250176, "epoch": 2810, "lr": 1.5132684769116611e-06}
{"train_loss": 0.04026337340474129, "global_step": 250177, "epoch": 2810, "lr": 1.5131269309209495e-06}
{"train_loss": 0.03968264820744817, "global_step": 250178, "epoch": 2810, "lr": 1.512985391448718e-06, "val_loss": 9.506903648376465, "train_action_mse_error": 23.53562355041504}
{"train_loss": 0.04398561269044876, "global_step": 250179, "epoch": 2811, "lr": 1.5128438584949722e-06}
{"train_loss": 0.032403573393821716, "global_step": 250180, "epoch": 2811, "lr": 1.5127023320597289e-06}
{"train_loss": 0.04658834636211395, "global_step": 250181, "epoch": 2811, "lr": 1.5125608121430157e-06}
{"train_loss": 0.024140862748026848, "global_step": 250182, "epoch": 2811, "lr": 1.5124192987448438e-06}
{"train_loss": 0.0294171329587698, "global_step": 250183, "epoch": 2811, "lr": 1.5122777918652409e-06}
{"train_loss": 0.051876962184906006, "global_step": 250184, "epoch": 2811, "lr": 1.512136291504218e-06}
{"train_loss": 0.06840801984071732, "global_step": 250185, "epoch": 2811, "lr": 1.5119947976617976e-06}
{"train_loss": 0.07093890011310577, "global_step": 250186, "epoch": 2811, "lr": 1.5118533103379961e-06}
{"train_loss": 0.02739890292286873, "global_step": 250187, "epoch": 2811, "lr": 1.511711829532836e-06}
{"train_loss": 0.13382560014724731, "global_step": 250188, "epoch": 2811, "lr": 1.511570355246339e-06}
{"train_loss": 0.028466328978538513, "global_step": 250189, "epoch": 2811, "lr": 1.5114288874785166e-06}
{"train_loss": 0.033446699380874634, "global_step": 250190, "epoch": 2811, "lr": 1.5112874262293909e-06}
{"train_loss": 0.022437041625380516, "global_step": 250191, "epoch": 2811, "lr": 1.5111459714989895e-06}
{"train_loss": 0.0749332383275032, "global_step": 250192, "epoch": 2811, "lr": 1.5110045232873127e-06}
{"train_loss": 0.033377595245838165, "global_step": 250193, "epoch": 2811, "lr": 1.5108630815943993e-06}
{"train_loss": 0.04303202033042908, "global_step": 250194, "epoch": 2811, "lr": 1.5107216464202545e-06}
{"train_loss": 0.013826635666191578, "global_step": 250195, "epoch": 2811, "lr": 1.5105802177649064e-06}
{"train_loss": 0.02258855476975441, "global_step": 250196, "epoch": 2811, "lr": 1.5104387956283606e-06}
{"train_loss": 0.07914808392524719, "global_step": 250197, "epoch": 2811, "lr": 1.5102973800106557e-06}
{"train_loss": 0.06632794439792633, "global_step": 250198, "epoch": 2811, "lr": 1.510155970911792e-06}
{"train_loss": 0.06926115602254868, "global_step": 250199, "epoch": 2811, "lr": 1.5100145683318023e-06}
{"train_loss": 0.03515365347266197, "global_step": 250200, "epoch": 2811, "lr": 1.5098731722706982e-06}
{"train_loss": 0.03263389319181442, "global_step": 250201, "epoch": 2811, "lr": 1.5097317827285017e-06}
{"train_loss": 0.0499076247215271, "global_step": 250202, "epoch": 2811, "lr": 1.5095903997052296e-06}
{"train_loss": 0.0325615368783474, "global_step": 250203, "epoch": 2811, "lr": 1.5094490232009039e-06}
{"train_loss": 0.04674898460507393, "global_step": 250204, "epoch": 2811, "lr": 1.5093076532155358e-06}
{"train_loss": 0.013600205071270466, "global_step": 250205, "epoch": 2811, "lr": 1.509166289749153e-06}
{"train_loss": 0.018587419763207436, "global_step": 250206, "epoch": 2811, "lr": 1.509024932801778e-06}
{"train_loss": 0.10004055500030518, "global_step": 250207, "epoch": 2811, "lr": 1.508883582373416e-06}
{"train_loss": 0.049617473036050797, "global_step": 250208, "epoch": 2811, "lr": 1.5087422384641003e-06}
{"train_loss": 0.05283379927277565, "global_step": 250209, "epoch": 2811, "lr": 1.5086009010738367e-06}
{"train_loss": 0.026954714208841324, "global_step": 250210, "epoch": 2811, "lr": 1.5084595702026583e-06}
{"train_loss": 0.03489777073264122, "global_step": 250211, "epoch": 2811, "lr": 1.5083182458505651e-06}
{"train_loss": 0.046377263963222504, "global_step": 250212, "epoch": 2811, "lr": 1.5081769280175962e-06}
{"train_loss": 0.05170685052871704, "global_step": 250213, "epoch": 2811, "lr": 1.5080356167037624e-06}
{"train_loss": 0.07510485500097275, "global_step": 250214, "epoch": 2811, "lr": 1.5078943119090804e-06}
{"train_loss": 0.05812538042664528, "global_step": 250215, "epoch": 2811, "lr": 1.5077530136335672e-06}
{"train_loss": 0.06634687632322311, "global_step": 250216, "epoch": 2811, "lr": 1.50761172187725e-06}
{"train_loss": 0.0552067756652832, "global_step": 250217, "epoch": 2811, "lr": 1.507470436640146e-06}
{"train_loss": 0.07224644720554352, "global_step": 250218, "epoch": 2811, "lr": 1.507329157922266e-06}
{"train_loss": 0.07585696130990982, "global_step": 250219, "epoch": 2811, "lr": 1.5071878857236378e-06}
{"train_loss": 0.07822465896606445, "global_step": 250220, "epoch": 2811, "lr": 1.5070466200442724e-06}
{"train_loss": 0.04883597791194916, "global_step": 250221, "epoch": 2811, "lr": 1.5069053608841977e-06}
{"train_loss": 0.0367790088057518, "global_step": 250222, "epoch": 2811, "lr": 1.5067641082434247e-06}
{"train_loss": 0.10210470855236053, "global_step": 250223, "epoch": 2811, "lr": 1.5066228621219812e-06}
{"train_loss": 0.037197429686784744, "global_step": 250224, "epoch": 2811, "lr": 1.5064816225198785e-06}
{"train_loss": 0.043630827218294144, "global_step": 250225, "epoch": 2811, "lr": 1.5063403894371386e-06}
{"train_loss": 0.05778881162405014, "global_step": 250226, "epoch": 2811, "lr": 1.5061991628737782e-06}
{"train_loss": 0.08366677165031433, "global_step": 250227, "epoch": 2811, "lr": 1.506057942829825e-06}
{"train_loss": 0.03423956409096718, "global_step": 250228, "epoch": 2811, "lr": 1.5059167293052845e-06}
{"train_loss": 0.018110714852809906, "global_step": 250229, "epoch": 2811, "lr": 1.5057755223001846e-06}
{"train_loss": 0.06478387117385864, "global_step": 250230, "epoch": 2811, "lr": 1.505634321814542e-06}
{"train_loss": 0.031731899827718735, "global_step": 250231, "epoch": 2811, "lr": 1.5054931278483786e-06}
{"train_loss": 0.05072688311338425, "global_step": 250232, "epoch": 2811, "lr": 1.5053519404017058e-06}
{"train_loss": 0.024557117372751236, "global_step": 250233, "epoch": 2811, "lr": 1.5052107594745512e-06}
{"train_loss": 0.060580283403396606, "global_step": 250234, "epoch": 2811, "lr": 1.505069585066926e-06}
{"train_loss": 0.027275316417217255, "global_step": 250235, "epoch": 2811, "lr": 1.504928417178858e-06}
{"train_loss": 0.07484997808933258, "global_step": 250236, "epoch": 2811, "lr": 1.5047872558103526e-06}
{"train_loss": 0.04088722541928291, "global_step": 250237, "epoch": 2811, "lr": 1.5046461009614488e-06}
{"train_loss": 0.04597623273730278, "global_step": 250238, "epoch": 2811, "lr": 1.5045049526321463e-06}
{"train_loss": 0.03662969172000885, "global_step": 250239, "epoch": 2811, "lr": 1.504363810822479e-06}
{"train_loss": 0.04650655761361122, "global_step": 250240, "epoch": 2811, "lr": 1.5042226755324518e-06}
{"train_loss": 0.053429313004016876, "global_step": 250241, "epoch": 2811, "lr": 1.5040815467620873e-06}
{"train_loss": 0.0658627599477768, "global_step": 250242, "epoch": 2811, "lr": 1.5039404245114186e-06}
{"train_loss": 0.034755609929561615, "global_step": 250243, "epoch": 2811, "lr": 1.503799308780446e-06}
{"train_loss": 0.059428054839372635, "global_step": 250244, "epoch": 2811, "lr": 1.5036581995692023e-06}
{"train_loss": 0.03473865985870361, "global_step": 250245, "epoch": 2811, "lr": 1.503517096877699e-06}
{"train_loss": 0.03901698812842369, "global_step": 250246, "epoch": 2811, "lr": 1.5033760007059584e-06}
{"train_loss": 0.05187133699655533, "global_step": 250247, "epoch": 2811, "lr": 1.5032349110539911e-06}
{"train_loss": 0.03861893340945244, "global_step": 250248, "epoch": 2811, "lr": 1.503093827921831e-06}
{"train_loss": 0.07898368686437607, "global_step": 250249, "epoch": 2811, "lr": 1.5029527513094833e-06}
{"train_loss": 0.02617308869957924, "global_step": 250250, "epoch": 2811, "lr": 1.5028116812169757e-06}
{"train_loss": 0.052280351519584656, "global_step": 250251, "epoch": 2811, "lr": 1.5026706176443194e-06}
{"train_loss": 0.06717298924922943, "global_step": 250252, "epoch": 2811, "lr": 1.5025295605915423e-06}
{"train_loss": 0.07417722791433334, "global_step": 250253, "epoch": 2811, "lr": 1.5023885100586554e-06}
{"train_loss": 0.08969416469335556, "global_step": 250254, "epoch": 2811, "lr": 1.5022474660456864e-06}
{"train_loss": 0.04618537425994873, "global_step": 250255, "epoch": 2811, "lr": 1.5021064285526464e-06}
{"train_loss": 0.0333026759326458, "global_step": 250256, "epoch": 2811, "lr": 1.5019653975795579e-06}
{"train_loss": 0.038729194551706314, "global_step": 250257, "epoch": 2811, "lr": 1.5018243731264425e-06}
{"train_loss": 0.03916013985872269, "global_step": 250258, "epoch": 2811, "lr": 1.5016833551933063e-06}
{"train_loss": 0.12404223531484604, "global_step": 250259, "epoch": 2811, "lr": 1.501542343780188e-06}
{"train_loss": 0.022129638120532036, "global_step": 250260, "epoch": 2811, "lr": 1.5014013388870874e-06}
{"train_loss": 0.027619976550340652, "global_step": 250261, "epoch": 2811, "lr": 1.5012603405140379e-06}
{"train_loss": 0.08572444319725037, "global_step": 250262, "epoch": 2811, "lr": 1.5011193486610508e-06}
{"train_loss": 0.043911162763834, "global_step": 250263, "epoch": 2811, "lr": 1.5009783633281483e-06}
{"train_loss": 0.03539896011352539, "global_step": 250264, "epoch": 2811, "lr": 1.5008373845153412e-06}
{"train_loss": 0.03643365576863289, "global_step": 250265, "epoch": 2811, "lr": 1.5006964122226686e-06}
{"train_loss": 0.037917014211416245, "global_step": 250266, "epoch": 2811, "lr": 1.5005554464501248e-06}
{"train_loss": 0.049737657305229915, "global_step": 250267, "epoch": 2811, "lr": 1.5004144871977487e-06, "val_loss": 9.54604434967041}
{"train_loss": 0.018386922776699066, "global_step": 250268, "epoch": 2812, "lr": 1.5002735344655405e-06}
{"train_loss": 0.04277577996253967, "global_step": 250269, "epoch": 2812, "lr": 1.5001325882535389e-06}
{"train_loss": 0.04407641664147377, "global_step": 250270, "epoch": 2812, "lr": 1.4999916485617438e-06}
{"train_loss": 0.021375780925154686, "global_step": 250271, "epoch": 2812, "lr": 1.4998507153901943e-06}
{"train_loss": 0.0017622173763811588, "global_step": 250272, "epoch": 2812, "lr": 1.49970978873889e-06}
{"train_loss": 0.043915823101997375, "global_step": 250273, "epoch": 2812, "lr": 1.4995688686078647e-06}
{"train_loss": 0.024309735745191574, "global_step": 250274, "epoch": 2812, "lr": 1.499427954997129e-06}
{"train_loss": 0.029888151213526726, "global_step": 250275, "epoch": 2812, "lr": 1.4992870479067057e-06}
{"train_loss": 0.051998429000377655, "global_step": 250276, "epoch": 2812, "lr": 1.4991461473366109e-06}
{"train_loss": 0.04657291620969772, "global_step": 250277, "epoch": 2812, "lr": 1.4990052532868615e-06}
{"train_loss": 0.09069984406232834, "global_step": 250278, "epoch": 2812, "lr": 1.4988643657574852e-06}
{"train_loss": 0.054751645773649216, "global_step": 250279, "epoch": 2812, "lr": 1.4987234847484877e-06}
{"train_loss": 0.047705862671136856, "global_step": 250280, "epoch": 2812, "lr": 1.498582610259902e-06}
{"train_loss": 0.051778875291347504, "global_step": 250281, "epoch": 2812, "lr": 1.4984417422917397e-06}
{"train_loss": 0.05608063563704491, "global_step": 250282, "epoch": 2812, "lr": 1.4983008808440225e-06}
{"train_loss": 0.02589230239391327, "global_step": 250283, "epoch": 2812, "lr": 1.4981600259167616e-06}
{"train_loss": 0.06338668614625931, "global_step": 250284, "epoch": 2812, "lr": 1.4980191775099906e-06}
{"train_loss": 0.013345503248274326, "global_step": 250285, "epoch": 2812, "lr": 1.497878335623709e-06}
{"train_loss": 0.047918375581502914, "global_step": 250286, "epoch": 2812, "lr": 1.4977375002579564e-06}
{"train_loss": 0.06945133209228516, "global_step": 250287, "epoch": 2812, "lr": 1.4975966714127377e-06}
{"train_loss": 0.04097674414515495, "global_step": 250288, "epoch": 2812, "lr": 1.4974558490880807e-06}
{"train_loss": 0.06378771364688873, "global_step": 250289, "epoch": 2812, "lr": 1.4973150332839914e-06}
{"train_loss": 0.08107766509056091, "global_step": 250290, "epoch": 2812, "lr": 1.4971742240005026e-06}
{"train_loss": 0.016771236434578896, "global_step": 250291, "epoch": 2812, "lr": 1.4970334212376203e-06}
{"train_loss": 0.05219656229019165, "global_step": 250292, "epoch": 2812, "lr": 1.4968926249953829e-06}
{"train_loss": 0.04216834902763367, "global_step": 250293, "epoch": 2812, "lr": 1.4967518352737853e-06}
{"train_loss": 0.025306029245257378, "global_step": 250294, "epoch": 2812, "lr": 1.496611052072866e-06}
{"train_loss": 0.02145303040742874, "global_step": 250295, "epoch": 2812, "lr": 1.4964702753926362e-06}
{"train_loss": 0.04023140296339989, "global_step": 250296, "epoch": 2812, "lr": 1.4963295052331072e-06}
{"train_loss": 0.026579653844237328, "global_step": 250297, "epoch": 2812, "lr": 1.4961887415943176e-06}
{"train_loss": 0.0651492327451706, "global_step": 250298, "epoch": 2812, "lr": 1.496047984476262e-06}
{"train_loss": 0.0346544086933136, "global_step": 250299, "epoch": 2812, "lr": 1.4959072338789792e-06}
{"train_loss": 0.058634527027606964, "global_step": 250300, "epoch": 2812, "lr": 1.4957664898024748e-06}
{"train_loss": 0.07345455139875412, "global_step": 250301, "epoch": 2812, "lr": 1.495625752246782e-06}
{"train_loss": 0.027678456157445908, "global_step": 250302, "epoch": 2812, "lr": 1.4954850212119009e-06}
{"train_loss": 0.020687591284513474, "global_step": 250303, "epoch": 2812, "lr": 1.4953442966978703e-06}
{"train_loss": 0.029395798221230507, "global_step": 250304, "epoch": 2812, "lr": 1.4952035787046903e-06}
{"train_loss": 0.03390512242913246, "global_step": 250305, "epoch": 2812, "lr": 1.495062867232394e-06}
{"train_loss": 0.047793686389923096, "global_step": 250306, "epoch": 2812, "lr": 1.4949221622809928e-06}
{"train_loss": 0.058843567967414856, "global_step": 250307, "epoch": 2812, "lr": 1.494781463850514e-06}
{"train_loss": 0.051563337445259094, "global_step": 250308, "epoch": 2812, "lr": 1.4946407719409639e-06}
{"train_loss": 0.018098296597599983, "global_step": 250309, "epoch": 2812, "lr": 1.4945000865523696e-06}
{"train_loss": 0.04050811752676964, "global_step": 250310, "epoch": 2812, "lr": 1.4943594076847479e-06}
{"train_loss": 0.051353566348552704, "global_step": 250311, "epoch": 2812, "lr": 1.4942187353381266e-06}
{"train_loss": 0.05725352466106415, "global_step": 250312, "epoch": 2812, "lr": 1.4940780695125055e-06}
{"train_loss": 0.05250740796327591, "global_step": 250313, "epoch": 2812, "lr": 1.4939374102079184e-06}
{"train_loss": 0.024274829775094986, "global_step": 250314, "epoch": 2812, "lr": 1.4937967574243817e-06}
{"train_loss": 0.04403260350227356, "global_step": 250315, "epoch": 2812, "lr": 1.4936561111619062e-06}
{"train_loss": 0.0689295306801796, "global_step": 250316, "epoch": 2812, "lr": 1.4935154714205257e-06}
{"train_loss": 0.030292173847556114, "global_step": 250317, "epoch": 2812, "lr": 1.4933748382002455e-06}
{"train_loss": 0.04337628185749054, "global_step": 250318, "epoch": 2812, "lr": 1.4932342115010988e-06}
{"train_loss": 0.023694347590208054, "global_step": 250319, "epoch": 2812, "lr": 1.493093591323086e-06}
{"train_loss": 0.08594291657209396, "global_step": 250320, "epoch": 2812, "lr": 1.4929529776662398e-06}
{"train_loss": 0.03912457451224327, "global_step": 250321, "epoch": 2812, "lr": 1.4928123705305718e-06}
{"train_loss": 0.03320392593741417, "global_step": 250322, "epoch": 2812, "lr": 1.4926717699161096e-06}
{"train_loss": 0.053855050355196, "global_step": 250323, "epoch": 2812, "lr": 1.4925311758228589e-06}
{"train_loss": 0.044094186276197433, "global_step": 250324, "epoch": 2812, "lr": 1.4923905882508527e-06}
{"train_loss": 0.05378108099102974, "global_step": 250325, "epoch": 2812, "lr": 1.4922500072000966e-06}
{"train_loss": 0.060373421758413315, "global_step": 250326, "epoch": 2812, "lr": 1.4921094326706241e-06}
{"train_loss": 0.024024223908782005, "global_step": 250327, "epoch": 2812, "lr": 1.4919688646624407e-06}
{"train_loss": 0.056605275720357895, "global_step": 250328, "epoch": 2812, "lr": 1.4918283031755742e-06}
{"train_loss": 0.04885117709636688, "global_step": 250329, "epoch": 2812, "lr": 1.4916877482100354e-06}
{"train_loss": 0.07622463256120682, "global_step": 250330, "epoch": 2812, "lr": 1.491547199765858e-06}
{"train_loss": 0.07410642504692078, "global_step": 250331, "epoch": 2812, "lr": 1.4914066578430419e-06}
{"train_loss": 0.014458193443715572, "global_step": 250332, "epoch": 2812, "lr": 1.4912661224416203e-06}
{"train_loss": 0.0587826743721962, "global_step": 250333, "epoch": 2812, "lr": 1.4911255935616042e-06}
{"train_loss": 0.059291671961545944, "global_step": 250334, "epoch": 2812, "lr": 1.4909850712030105e-06}
{"train_loss": 0.06372435390949249, "global_step": 250335, "epoch": 2812, "lr": 1.4908445553658722e-06}
{"train_loss": 0.01955457217991352, "global_step": 250336, "epoch": 2812, "lr": 1.4907040460501897e-06}
{"train_loss": 0.027191732078790665, "global_step": 250337, "epoch": 2812, "lr": 1.4905635432559962e-06}
{"train_loss": 0.04468288645148277, "global_step": 250338, "epoch": 2812, "lr": 1.490423046983297e-06}
{"train_loss": 0.08251305669546127, "global_step": 250339, "epoch": 2812, "lr": 1.4902825572321312e-06}
{"train_loss": 0.016463248059153557, "global_step": 250340, "epoch": 2812, "lr": 1.490142074002493e-06}
{"train_loss": 0.030437445268034935, "global_step": 250341, "epoch": 2812, "lr": 1.490001597294427e-06}
{"train_loss": 0.0424494594335556, "global_step": 250342, "epoch": 2812, "lr": 1.4898611271079277e-06}
{"train_loss": 0.05660678446292877, "global_step": 250343, "epoch": 2812, "lr": 1.4897206634430338e-06}
{"train_loss": 0.044744499027729034, "global_step": 250344, "epoch": 2812, "lr": 1.489580206299751e-06}
{"train_loss": 0.015325836837291718, "global_step": 250345, "epoch": 2812, "lr": 1.489439755678107e-06}
{"train_loss": 0.058183249086141586, "global_step": 250346, "epoch": 2812, "lr": 1.4892993115781074e-06}
{"train_loss": 0.08707264810800552, "global_step": 250347, "epoch": 2812, "lr": 1.4891588739997907e-06}
{"train_loss": 0.05714678391814232, "global_step": 250348, "epoch": 2812, "lr": 1.4890184429431575e-06}
{"train_loss": 0.05021340399980545, "global_step": 250349, "epoch": 2812, "lr": 1.488878018408235e-06}
{"train_loss": 0.03766936436295509, "global_step": 250350, "epoch": 2812, "lr": 1.4887376003950516e-06}
{"train_loss": 0.08620529621839523, "global_step": 250351, "epoch": 2812, "lr": 1.488597188903601e-06}
{"train_loss": 0.03987520560622215, "global_step": 250352, "epoch": 2812, "lr": 1.488456783933928e-06}
{"train_loss": 0.08290732651948929, "global_step": 250353, "epoch": 2812, "lr": 1.4883163854860383e-06}
{"train_loss": 0.058883391320705414, "global_step": 250354, "epoch": 2812, "lr": 1.4881759935599538e-06}
{"train_loss": 0.04548298940062523, "global_step": 250355, "epoch": 2812, "lr": 1.4880356081556912e-06}
{"train_loss": 0.04619956077756674, "global_step": 250356, "epoch": 2812, "lr": 1.4878952292732729e-06, "val_loss": 9.546160697937012}
{"train_loss": 0.035932764410972595, "global_step": 250357, "epoch": 2813, "lr": 1.4877548569127154e-06}
{"train_loss": 0.04349330812692642, "global_step": 250358, "epoch": 2813, "lr": 1.4876144910740352e-06}
{"train_loss": 0.020657340064644814, "global_step": 250359, "epoch": 2813, "lr": 1.4874741317572549e-06}
{"train_loss": 0.056385114789009094, "global_step": 250360, "epoch": 2813, "lr": 1.4873337789623964e-06}
{"train_loss": 0.039723098278045654, "global_step": 250361, "epoch": 2813, "lr": 1.487193432689471e-06}
{"train_loss": 0.044828541576862335, "global_step": 250362, "epoch": 2813, "lr": 1.4870530929385007e-06}
{"train_loss": 0.03275234252214432, "global_step": 250363, "epoch": 2813, "lr": 1.4869127597095077e-06}
{"train_loss": 0.04468649625778198, "global_step": 250364, "epoch": 2813, "lr": 1.486772433002509e-06}
{"train_loss": 0.03583286330103874, "global_step": 250365, "epoch": 2813, "lr": 1.4866321128175154e-06}
{"train_loss": 0.040755268186330795, "global_step": 250366, "epoch": 2813, "lr": 1.48649179915456e-06}
{"train_loss": 0.06461313366889954, "global_step": 250367, "epoch": 2813, "lr": 1.486351492013649e-06}
{"train_loss": 0.04262235388159752, "global_step": 250368, "epoch": 2813, "lr": 1.4862111913948096e-06}
{"train_loss": 0.03211203217506409, "global_step": 250369, "epoch": 2813, "lr": 1.4860708972980587e-06}
{"train_loss": 0.025186587125062943, "global_step": 250370, "epoch": 2813, "lr": 1.4859306097234127e-06}
{"train_loss": 0.026243358850479126, "global_step": 250371, "epoch": 2813, "lr": 1.4857903286708885e-06}
{"train_loss": 0.030971409752964973, "global_step": 250372, "epoch": 2813, "lr": 1.4856500541405195e-06}
{"train_loss": 0.0222639087587595, "global_step": 250373, "epoch": 2813, "lr": 1.4855097861323052e-06}
{"train_loss": 0.03170974925160408, "global_step": 250374, "epoch": 2813, "lr": 1.485369524646274e-06}
{"train_loss": 0.02499212883412838, "global_step": 250375, "epoch": 2813, "lr": 1.4852292696824476e-06}
{"train_loss": 0.05916294828057289, "global_step": 250376, "epoch": 2813, "lr": 1.4850890212408319e-06}
{"train_loss": 0.03930431231856346, "global_step": 250377, "epoch": 2813, "lr": 1.48494877932146e-06}
{"train_loss": 0.03767828270792961, "global_step": 250378, "epoch": 2813, "lr": 1.484808543924343e-06}
{"train_loss": 0.021402206271886826, "global_step": 250379, "epoch": 2813, "lr": 1.484668315049509e-06}
{"train_loss": 0.03401757776737213, "global_step": 250380, "epoch": 2813, "lr": 1.4845280926969628e-06}
{"train_loss": 0.04258212819695473, "global_step": 250381, "epoch": 2813, "lr": 1.484387876866733e-06}
{"train_loss": 0.030076494440436363, "global_step": 250382, "epoch": 2813, "lr": 1.4842476675588356e-06}
{"train_loss": 0.04987383261322975, "global_step": 250383, "epoch": 2813, "lr": 1.4841074647732933e-06}
{"train_loss": 0.025208009406924248, "global_step": 250384, "epoch": 2813, "lr": 1.4839672685101114e-06}
{"train_loss": 0.03723005950450897, "global_step": 250385, "epoch": 2813, "lr": 1.4838270787693232e-06}
{"train_loss": 0.050635166466236115, "global_step": 250386, "epoch": 2813, "lr": 1.4836868955509508e-06}
{"train_loss": 0.03185658156871796, "global_step": 250387, "epoch": 2813, "lr": 1.4835467188549946e-06}
{"train_loss": 0.05510682612657547, "global_step": 250388, "epoch": 2813, "lr": 1.483406548681493e-06}
{"train_loss": 0.058895766735076904, "global_step": 250389, "epoch": 2813, "lr": 1.4832663850304519e-06}
{"train_loss": 0.02679094299674034, "global_step": 250390, "epoch": 2813, "lr": 1.4831262279018987e-06}
{"train_loss": 0.018797842785716057, "global_step": 250391, "epoch": 2813, "lr": 1.4829860772958392e-06}
{"train_loss": 0.026126574724912643, "global_step": 250392, "epoch": 2813, "lr": 1.4828459332123068e-06}
{"train_loss": 0.049651920795440674, "global_step": 250393, "epoch": 2813, "lr": 1.4827057956513124e-06}
{"train_loss": 0.04129658639431, "global_step": 250394, "epoch": 2813, "lr": 1.4825656646128838e-06}
{"train_loss": 0.04712438955903053, "global_step": 250395, "epoch": 2813, "lr": 1.4824255400970212e-06}
{"train_loss": 0.03548642620444298, "global_step": 250396, "epoch": 2813, "lr": 1.482285422103763e-06}
{"train_loss": 0.061857614666223526, "global_step": 250397, "epoch": 2813, "lr": 1.4821453106331152e-06}
{"train_loss": 0.015647299587726593, "global_step": 250398, "epoch": 2813, "lr": 1.4820052056851108e-06}
{"train_loss": 0.027969934046268463, "global_step": 250399, "epoch": 2813, "lr": 1.48186510725975e-06}
{"train_loss": 0.041343796998262405, "global_step": 250400, "epoch": 2813, "lr": 1.481725015357066e-06}
{"train_loss": 0.06380146741867065, "global_step": 250401, "epoch": 2813, "lr": 1.48158492997707e-06}
{"train_loss": 0.05013107880949974, "global_step": 250402, "epoch": 2813, "lr": 1.4814448511197843e-06}
{"train_loss": 0.07335980236530304, "global_step": 250403, "epoch": 2813, "lr": 1.4813047787852252e-06}
{"train_loss": 0.05308448523283005, "global_step": 250404, "epoch": 2813, "lr": 1.4811647129734151e-06}
{"train_loss": 0.014671778306365013, "global_step": 250405, "epoch": 2813, "lr": 1.481024653684371e-06}
{"train_loss": 0.06971094012260437, "global_step": 250406, "epoch": 2813, "lr": 1.4808846009181144e-06}
{"train_loss": 0.03607126325368881, "global_step": 250407, "epoch": 2813, "lr": 1.4807445546746568e-06}
{"train_loss": 0.0772596076130867, "global_step": 250408, "epoch": 2813, "lr": 1.480604514954026e-06}
{"train_loss": 0.05338142067193985, "global_step": 250409, "epoch": 2813, "lr": 1.480464481756233e-06}
{"train_loss": 0.022892460227012634, "global_step": 250410, "epoch": 2813, "lr": 1.4803244550813056e-06}
{"train_loss": 0.09019488841295242, "global_step": 250411, "epoch": 2813, "lr": 1.4801844349292548e-06}
{"train_loss": 0.018576323986053467, "global_step": 250412, "epoch": 2813, "lr": 1.4800444213000975e-06}
{"train_loss": 0.055969204753637314, "global_step": 250413, "epoch": 2813, "lr": 1.4799044141938612e-06}
{"train_loss": 0.033189866691827774, "global_step": 250414, "epoch": 2813, "lr": 1.479764413610557e-06}
{"train_loss": 0.025118665769696236, "global_step": 250415, "epoch": 2813, "lr": 1.4796244195502129e-06}
{"train_loss": 0.045002300292253494, "global_step": 250416, "epoch": 2813, "lr": 1.4794844320128343e-06}
{"train_loss": 0.09112069010734558, "global_step": 250417, "epoch": 2813, "lr": 1.4793444509984543e-06}
{"train_loss": 0.023209668695926666, "global_step": 250418, "epoch": 2813, "lr": 1.479204476507079e-06}
{"train_loss": 0.05677309259772301, "global_step": 250419, "epoch": 2813, "lr": 1.479064508538741e-06}
{"train_loss": 0.0357418917119503, "global_step": 250420, "epoch": 2813, "lr": 1.4789245470934465e-06}
{"train_loss": 0.037974342703819275, "global_step": 250421, "epoch": 2813, "lr": 1.478784592171223e-06}
{"train_loss": 0.025578094646334648, "global_step": 250422, "epoch": 2813, "lr": 1.4786446437720758e-06}
{"train_loss": 0.02302529849112034, "global_step": 250423, "epoch": 2813, "lr": 1.4785047018960386e-06}
{"train_loss": 0.06409172713756561, "global_step": 250424, "epoch": 2813, "lr": 1.4783647665431277e-06}
{"train_loss": 0.06804762780666351, "global_step": 250425, "epoch": 2813, "lr": 1.4782248377133545e-06}
{"train_loss": 0.09142150729894638, "global_step": 250426, "epoch": 2813, "lr": 1.478084915406752e-06}
{"train_loss": 0.02328038029372692, "global_step": 250427, "epoch": 2813, "lr": 1.4779449996233207e-06}
{"train_loss": 0.057492949068546295, "global_step": 250428, "epoch": 2813, "lr": 1.4778050903630935e-06}
{"train_loss": 0.030117040500044823, "global_step": 250429, "epoch": 2813, "lr": 1.477665187626076e-06}
{"train_loss": 0.052685532718896866, "global_step": 250430, "epoch": 2813, "lr": 1.4775252914123017e-06}
{"train_loss": 0.026342634111642838, "global_step": 250431, "epoch": 2813, "lr": 1.4773854017217815e-06}
{"train_loss": 0.027582520619034767, "global_step": 250432, "epoch": 2813, "lr": 1.4772455185545376e-06}
{"train_loss": 0.04770853370428085, "global_step": 250433, "epoch": 2813, "lr": 1.477105641910581e-06}
{"train_loss": 0.03903907164931297, "global_step": 250434, "epoch": 2813, "lr": 1.4769657717899455e-06}
{"train_loss": 0.07893004268407822, "global_step": 250435, "epoch": 2813, "lr": 1.4768259081926305e-06}
{"train_loss": 0.013208597898483276, "global_step": 250436, "epoch": 2813, "lr": 1.4766860511186698e-06}
{"train_loss": 0.04974500462412834, "global_step": 250437, "epoch": 2813, "lr": 1.4765462005680741e-06}
{"train_loss": 0.05562470853328705, "global_step": 250438, "epoch": 2813, "lr": 1.4764063565408716e-06}
{"train_loss": 0.0693364292383194, "global_step": 250439, "epoch": 2813, "lr": 1.4762665190370672e-06}
{"train_loss": 0.037183500826358795, "global_step": 250440, "epoch": 2813, "lr": 1.4761266880566893e-06}
{"train_loss": 0.012393003329634666, "global_step": 250441, "epoch": 2813, "lr": 1.4759868635997597e-06}
{"train_loss": 0.05244803428649902, "global_step": 250442, "epoch": 2813, "lr": 1.4758470456662844e-06}
{"train_loss": 0.016454050317406654, "global_step": 250443, "epoch": 2813, "lr": 1.4757072342562962e-06}
{"train_loss": 0.05466306954622269, "global_step": 250444, "epoch": 2813, "lr": 1.4755674293698118e-06}
{"train_loss": 0.04232076023904125, "global_step": 250445, "epoch": 2813, "lr": 1.475427631006837e-06, "val_loss": 9.544767379760742}
{"train_loss": 0.02358061820268631, "global_step": 250446, "epoch": 2814, "lr": 1.4752878391674052e-06}
{"train_loss": 0.09001431614160538, "global_step": 250447, "epoch": 2814, "lr": 1.4751480538515273e-06}
{"train_loss": 0.044588200747966766, "global_step": 250448, "epoch": 2814, "lr": 1.4750082750592253e-06}
{"train_loss": 0.03194679692387581, "global_step": 250449, "epoch": 2814, "lr": 1.4748685027905163e-06}
{"train_loss": 0.024897228926420212, "global_step": 250450, "epoch": 2814, "lr": 1.4747287370454166e-06}
{"train_loss": 0.027708809822797775, "global_step": 250451, "epoch": 2814, "lr": 1.474588977823954e-06}
{"train_loss": 0.034198492765426636, "global_step": 250452, "epoch": 2814, "lr": 1.4744492251261344e-06}
{"train_loss": 0.04089517891407013, "global_step": 250453, "epoch": 2814, "lr": 1.4743094789519906e-06}
{"train_loss": 0.040525082498788834, "global_step": 250454, "epoch": 2814, "lr": 1.4741697393015285e-06}
{"train_loss": 0.03051299974322319, "global_step": 250455, "epoch": 2814, "lr": 1.4740300061747814e-06}
{"train_loss": 0.05183360353112221, "global_step": 250456, "epoch": 2814, "lr": 1.4738902795717491e-06}
{"train_loss": 0.038922280073165894, "global_step": 250457, "epoch": 2814, "lr": 1.4737505594924706e-06}
{"train_loss": 0.04956398904323578, "global_step": 250458, "epoch": 2814, "lr": 1.4736108459369457e-06}
{"train_loss": 0.07723423093557358, "global_step": 250459, "epoch": 2814, "lr": 1.473471138905208e-06}
{"train_loss": 0.04751907289028168, "global_step": 250460, "epoch": 2814, "lr": 1.4733314383972685e-06}
{"train_loss": 0.031491469591856, "global_step": 250461, "epoch": 2814, "lr": 1.4731917444131493e-06}
{"train_loss": 0.07133609801530838, "global_step": 250462, "epoch": 2814, "lr": 1.4730520569528672e-06}
{"train_loss": 0.04631996899843216, "global_step": 250463, "epoch": 2814, "lr": 1.4729123760164387e-06}
{"train_loss": 0.02623720094561577, "global_step": 250464, "epoch": 2814, "lr": 1.4727727016038916e-06}
{"train_loss": 0.050471216440200806, "global_step": 250465, "epoch": 2814, "lr": 1.4726330337152371e-06}
{"train_loss": 0.03180085867643356, "global_step": 250466, "epoch": 2814, "lr": 1.4724933723504976e-06}
{"train_loss": 0.04084162041544914, "global_step": 250467, "epoch": 2814, "lr": 1.4723537175096835e-06}
{"train_loss": 0.019807811826467514, "global_step": 250468, "epoch": 2814, "lr": 1.4722140691928287e-06}
{"train_loss": 0.04979481175541878, "global_step": 250469, "epoch": 2814, "lr": 1.4720744273999331e-06}
{"train_loss": 0.04383682459592819, "global_step": 250470, "epoch": 2814, "lr": 1.4719347921310356e-06}
{"train_loss": 0.019262196496129036, "global_step": 250471, "epoch": 2814, "lr": 1.471795163386136e-06}
{"train_loss": 0.033879611641168594, "global_step": 250472, "epoch": 2814, "lr": 1.4716555411652678e-06}
{"train_loss": 0.04002607241272926, "global_step": 250473, "epoch": 2814, "lr": 1.471515925468442e-06}
{"train_loss": 0.02882138267159462, "global_step": 250474, "epoch": 2814, "lr": 1.4713763162956806e-06}
{"train_loss": 0.06604022532701492, "global_step": 250475, "epoch": 2814, "lr": 1.4712367136469952e-06}
{"train_loss": 0.026290996000170708, "global_step": 250476, "epoch": 2814, "lr": 1.471097117522413e-06}
{"train_loss": 0.013542534783482552, "global_step": 250477, "epoch": 2814, "lr": 1.4709575279219568e-06}
{"train_loss": 0.030332472175359726, "global_step": 250478, "epoch": 2814, "lr": 1.4708179448456316e-06}
{"train_loss": 0.024540474638342857, "global_step": 250479, "epoch": 2814, "lr": 1.4706783682934711e-06}
{"train_loss": 0.08363986760377884, "global_step": 250480, "epoch": 2814, "lr": 1.4705387982654805e-06}
{"train_loss": 0.031326860189437866, "global_step": 250481, "epoch": 2814, "lr": 1.470399234761688e-06}
{"train_loss": 0.05235988646745682, "global_step": 250482, "epoch": 2814, "lr": 1.4702596777821044e-06}
{"train_loss": 0.04794001206755638, "global_step": 250483, "epoch": 2814, "lr": 1.4701201273267573e-06}
{"train_loss": 0.01578516885638237, "global_step": 250484, "epoch": 2814, "lr": 1.4699805833956638e-06}
{"train_loss": 0.060975201427936554, "global_step": 250485, "epoch": 2814, "lr": 1.4698410459888345e-06}
{"train_loss": 0.06769271194934845, "global_step": 250486, "epoch": 2814, "lr": 1.4697015151062975e-06}
{"train_loss": 0.03932110592722893, "global_step": 250487, "epoch": 2814, "lr": 1.4695619907480585e-06}
{"train_loss": 0.06803194433450699, "global_step": 250488, "epoch": 2814, "lr": 1.469422472914156e-06}
{"train_loss": 0.029017318040132523, "global_step": 250489, "epoch": 2814, "lr": 1.4692829616045956e-06}
{"train_loss": 0.027570579200983047, "global_step": 250490, "epoch": 2814, "lr": 1.469143456819394e-06}
{"train_loss": 0.01803349144756794, "global_step": 250491, "epoch": 2814, "lr": 1.469003958558579e-06}
{"train_loss": 0.06851661950349808, "global_step": 250492, "epoch": 2814, "lr": 1.4688644668221618e-06}
{"train_loss": 0.014461827464401722, "global_step": 250493, "epoch": 2814, "lr": 1.46872498161017e-06}
{"train_loss": 0.08167950809001923, "global_step": 250494, "epoch": 2814, "lr": 1.4685855029226092e-06}
{"train_loss": 0.06186848506331444, "global_step": 250495, "epoch": 2814, "lr": 1.4684460307595015e-06}
{"train_loss": 0.021146556362509727, "global_step": 250496, "epoch": 2814, "lr": 1.4683065651208805e-06}
{"train_loss": 0.027234353125095367, "global_step": 250497, "epoch": 2814, "lr": 1.4681671060067514e-06}
{"train_loss": 0.07045096158981323, "global_step": 250498, "epoch": 2814, "lr": 1.4680276534171366e-06}
{"train_loss": 0.03890230506658554, "global_step": 250499, "epoch": 2814, "lr": 1.4678882073520473e-06}
{"train_loss": 0.04074093699455261, "global_step": 250500, "epoch": 2814, "lr": 1.4677487678115166e-06}
{"train_loss": 0.09262584894895554, "global_step": 250501, "epoch": 2814, "lr": 1.46760933479555e-06}
{"train_loss": 0.0389070026576519, "global_step": 250502, "epoch": 2814, "lr": 1.4674699083041755e-06}
{"train_loss": 0.011862861923873425, "global_step": 250503, "epoch": 2814, "lr": 1.467330488337404e-06}
{"train_loss": 0.03095056861639023, "global_step": 250504, "epoch": 2814, "lr": 1.4671910748952689e-06}
{"train_loss": 0.03886847943067551, "global_step": 250505, "epoch": 2814, "lr": 1.4670516679777647e-06}
{"train_loss": 0.03641945496201515, "global_step": 250506, "epoch": 2814, "lr": 1.4669122675849356e-06}
{"train_loss": 0.05159449949860573, "global_step": 250507, "epoch": 2814, "lr": 1.4667728737167763e-06}
{"train_loss": 0.0376153327524662, "global_step": 250508, "epoch": 2814, "lr": 1.466633486373331e-06}
{"train_loss": 0.03719697147607803, "global_step": 250509, "epoch": 2814, "lr": 1.4664941055545944e-06}
{"train_loss": 0.027799557894468307, "global_step": 250510, "epoch": 2814, "lr": 1.4663547312606052e-06}
{"train_loss": 0.06263862550258636, "global_step": 250511, "epoch": 2814, "lr": 1.4662153634913634e-06}
{"train_loss": 0.036458570510149, "global_step": 250512, "epoch": 2814, "lr": 1.4660760022469023e-06}
{"train_loss": 0.04303749278187752, "global_step": 250513, "epoch": 2814, "lr": 1.4659366475272385e-06}
{"train_loss": 0.040339283645153046, "global_step": 250514, "epoch": 2814, "lr": 1.4657972993323832e-06}
{"train_loss": 0.033815544098615646, "global_step": 250515, "epoch": 2814, "lr": 1.4656579576623642e-06}
{"train_loss": 0.05853641778230667, "global_step": 250516, "epoch": 2814, "lr": 1.4655186225171924e-06}
{"train_loss": 0.05671914666891098, "global_step": 250517, "epoch": 2814, "lr": 1.4653792938968903e-06}
{"train_loss": 0.02720773033797741, "global_step": 250518, "epoch": 2814, "lr": 1.4652399718014797e-06}
{"train_loss": 0.04754854738712311, "global_step": 250519, "epoch": 2814, "lr": 1.4651006562309721e-06}
{"train_loss": 0.01842389814555645, "global_step": 250520, "epoch": 2814, "lr": 1.4649613471853895e-06}
{"train_loss": 0.06139632686972618, "global_step": 250521, "epoch": 2814, "lr": 1.4648220446647598e-06}
{"train_loss": 0.032031599432229996, "global_step": 250522, "epoch": 2814, "lr": 1.4646827486690883e-06}
{"train_loss": 0.036650169640779495, "global_step": 250523, "epoch": 2814, "lr": 1.4645434591983975e-06}
{"train_loss": 0.07071872055530548, "global_step": 250524, "epoch": 2814, "lr": 1.4644041762527093e-06}
{"train_loss": 0.023547004908323288, "global_step": 250525, "epoch": 2814, "lr": 1.4642648998320351e-06}
{"train_loss": 0.04132562875747681, "global_step": 250526, "epoch": 2814, "lr": 1.4641256299364025e-06}
{"train_loss": 0.020862864330410957, "global_step": 250527, "epoch": 2814, "lr": 1.4639863665658339e-06}
{"train_loss": 0.12053924798965454, "global_step": 250528, "epoch": 2814, "lr": 1.4638471097203289e-06}
{"train_loss": 0.03846796974539757, "global_step": 250529, "epoch": 2814, "lr": 1.4637078593999265e-06}
{"train_loss": 0.034672804176807404, "global_step": 250530, "epoch": 2814, "lr": 1.463568615604627e-06}
{"train_loss": 0.0873517096042633, "global_step": 250531, "epoch": 2814, "lr": 1.4634293783344633e-06}
{"train_loss": 0.035986971110105515, "global_step": 250532, "epoch": 2814, "lr": 1.4632901475894578e-06}
{"train_loss": 0.037529535591602325, "global_step": 250533, "epoch": 2814, "lr": 1.4631509233696106e-06}
{"train_loss": 0.042946920447637524, "global_step": 250534, "epoch": 2814, "lr": 1.4630117056749603e-06, "val_loss": 9.619056701660156}
{"train_loss": 0.06252777576446533, "global_step": 250535, "epoch": 2815, "lr": 1.4628724945055073e-06}
{"train_loss": 0.03984726965427399, "global_step": 250536, "epoch": 2815, "lr": 1.4627332898612844e-06}
{"train_loss": 0.06208967790007591, "global_step": 250537, "epoch": 2815, "lr": 1.462594091742303e-06}
{"train_loss": 0.04936505854129791, "global_step": 250538, "epoch": 2815, "lr": 1.462454900148591e-06}
{"train_loss": 0.045683711767196655, "global_step": 250539, "epoch": 2815, "lr": 1.4623157150801536e-06}
{"train_loss": 0.02728814259171486, "global_step": 250540, "epoch": 2815, "lr": 1.4621765365370189e-06}
{"train_loss": 0.05690653994679451, "global_step": 250541, "epoch": 2815, "lr": 1.4620373645192032e-06}
{"train_loss": 0.07931771129369736, "global_step": 250542, "epoch": 2815, "lr": 1.4618981990267288e-06}
{"train_loss": 0.09123805910348892, "global_step": 250543, "epoch": 2815, "lr": 1.4617590400596015e-06}
{"train_loss": 0.06487908214330673, "global_step": 250544, "epoch": 2815, "lr": 1.4616198876178544e-06}
{"train_loss": 0.02328687347471714, "global_step": 250545, "epoch": 2815, "lr": 1.4614807417014985e-06}
{"train_loss": 0.011239258572459221, "global_step": 250546, "epoch": 2815, "lr": 1.461341602310562e-06}
{"train_loss": 0.07442682981491089, "global_step": 250547, "epoch": 2815, "lr": 1.4612024694450443e-06}
{"train_loss": 0.02260418049991131, "global_step": 250548, "epoch": 2815, "lr": 1.4610633431049904e-06}
{"train_loss": 0.037294503301382065, "global_step": 250549, "epoch": 2815, "lr": 1.4609242232903942e-06}
{"train_loss": 0.03367949277162552, "global_step": 250550, "epoch": 2815, "lr": 1.4607851100012837e-06}
{"train_loss": 0.0432053878903389, "global_step": 250551, "epoch": 2815, "lr": 1.4606460032376867e-06}
{"train_loss": 0.020186422392725945, "global_step": 250552, "epoch": 2815, "lr": 1.4605069029996089e-06}
{"train_loss": 0.033690646290779114, "global_step": 250553, "epoch": 2815, "lr": 1.4603678092870832e-06}
{"train_loss": 0.07383222877979279, "global_step": 250554, "epoch": 2815, "lr": 1.4602287221001098e-06}
{"train_loss": 0.017343105748295784, "global_step": 250555, "epoch": 2815, "lr": 1.4600896414387222e-06}
{"train_loss": 0.052910637110471725, "global_step": 250556, "epoch": 2815, "lr": 1.4599505673029313e-06}
{"train_loss": 0.06803595274686813, "global_step": 250557, "epoch": 2815, "lr": 1.4598114996927591e-06}
{"train_loss": 0.056500084698200226, "global_step": 250558, "epoch": 2815, "lr": 1.4596724386082228e-06}
{"train_loss": 0.036847665905952454, "global_step": 250559, "epoch": 2815, "lr": 1.459533384049344e-06}
{"train_loss": 0.06217031180858612, "global_step": 250560, "epoch": 2815, "lr": 1.45939433601614e-06}
{"train_loss": 0.0589510016143322, "global_step": 250561, "epoch": 2815, "lr": 1.459255294508627e-06}
{"train_loss": 0.08062896132469177, "global_step": 250562, "epoch": 2815, "lr": 1.4591162595268271e-06}
{"train_loss": 0.020190296694636345, "global_step": 250563, "epoch": 2815, "lr": 1.4589772310707516e-06}
{"train_loss": 0.02682684361934662, "global_step": 250564, "epoch": 2815, "lr": 1.458838209140434e-06}
{"train_loss": 0.06063222885131836, "global_step": 250565, "epoch": 2815, "lr": 1.4586991937358796e-06}
{"train_loss": 0.03206417337059975, "global_step": 250566, "epoch": 2815, "lr": 1.4585601848571106e-06}
{"train_loss": 0.04132337495684624, "global_step": 250567, "epoch": 2815, "lr": 1.458421182504144e-06}
{"train_loss": 0.025526244193315506, "global_step": 250568, "epoch": 2815, "lr": 1.4582821866770068e-06}
{"train_loss": 0.051432814449071884, "global_step": 250569, "epoch": 2815, "lr": 1.4581431973757054e-06}
{"train_loss": 0.07469010353088379, "global_step": 250570, "epoch": 2815, "lr": 1.4580042146002727e-06}
{"train_loss": 0.023411711677908897, "global_step": 250571, "epoch": 2815, "lr": 1.4578652383507142e-06}
{"train_loss": 0.02543853409588337, "global_step": 250572, "epoch": 2815, "lr": 1.4577262686270577e-06}
{"train_loss": 0.05826573818922043, "global_step": 250573, "epoch": 2815, "lr": 1.4575873054293142e-06}
{"train_loss": 0.009534729644656181, "global_step": 250574, "epoch": 2815, "lr": 1.457448348757512e-06}
{"train_loss": 0.045276056975126266, "global_step": 250575, "epoch": 2815, "lr": 1.4573093986116558e-06}
{"train_loss": 0.021054431796073914, "global_step": 250576, "epoch": 2815, "lr": 1.4571704549917798e-06}
{"train_loss": 0.04557814449071884, "global_step": 250577, "epoch": 2815, "lr": 1.4570315178978944e-06}
{"train_loss": 0.07317454367876053, "global_step": 250578, "epoch": 2815, "lr": 1.4568925873300166e-06}
{"train_loss": 0.04262646287679672, "global_step": 250579, "epoch": 2815, "lr": 1.4567536632881684e-06}
{"train_loss": 0.07696592062711716, "global_step": 250580, "epoch": 2815, "lr": 1.4566147457723723e-06}
{"train_loss": 0.03454413637518883, "global_step": 250581, "epoch": 2815, "lr": 1.4564758347826391e-06}
{"train_loss": 0.04985252395272255, "global_step": 250582, "epoch": 2815, "lr": 1.4563369303189912e-06}
{"train_loss": 0.01227918267250061, "global_step": 250583, "epoch": 2815, "lr": 1.456198032381445e-06}
{"train_loss": 0.03187466412782669, "global_step": 250584, "epoch": 2815, "lr": 1.4560591409700286e-06}
{"train_loss": 0.07199856638908386, "global_step": 250585, "epoch": 2815, "lr": 1.4559202560847475e-06}
{"train_loss": 0.01526904571801424, "global_step": 250586, "epoch": 2815, "lr": 1.4557813777256235e-06}
{"train_loss": 0.036461200565099716, "global_step": 250587, "epoch": 2815, "lr": 1.4556425058926849e-06}
{"train_loss": 0.0818350687623024, "global_step": 250588, "epoch": 2815, "lr": 1.4555036405859369e-06}
{"train_loss": 0.04341655597090721, "global_step": 250589, "epoch": 2815, "lr": 1.4553647818054128e-06}
{"train_loss": 0.02510988339781761, "global_step": 250590, "epoch": 2815, "lr": 1.4552259295511184e-06}
{"train_loss": 0.03575956076383591, "global_step": 250591, "epoch": 2815, "lr": 1.4550870838230758e-06}
{"train_loss": 0.06578202545642853, "global_step": 250592, "epoch": 2815, "lr": 1.454948244621307e-06}
{"train_loss": 0.027659021317958832, "global_step": 250593, "epoch": 2815, "lr": 1.454809411945829e-06}
{"train_loss": 0.04468002915382385, "global_step": 250594, "epoch": 2815, "lr": 1.4546705857966581e-06}
{"train_loss": 0.11857935041189194, "global_step": 250595, "epoch": 2815, "lr": 1.4545317661738222e-06}
{"train_loss": 0.032825928181409836, "global_step": 250596, "epoch": 2815, "lr": 1.4543929530773216e-06}
{"train_loss": 0.035602279007434845, "global_step": 250597, "epoch": 2815, "lr": 1.4542541465071945e-06}
{"train_loss": 0.05440231040120125, "global_step": 250598, "epoch": 2815, "lr": 1.4541153464634473e-06}
{"train_loss": 0.08723282814025879, "global_step": 250599, "epoch": 2815, "lr": 1.4539765529461068e-06}
{"train_loss": 0.03773832321166992, "global_step": 250600, "epoch": 2815, "lr": 1.453837765955185e-06}
{"train_loss": 0.05771509185433388, "global_step": 250601, "epoch": 2815, "lr": 1.4536989854907035e-06}
{"train_loss": 0.014516917988657951, "global_step": 250602, "epoch": 2815, "lr": 1.4535602115526792e-06}
{"train_loss": 0.04589563235640526, "global_step": 250603, "epoch": 2815, "lr": 1.4534214441411343e-06}
{"train_loss": 0.0755883976817131, "global_step": 250604, "epoch": 2815, "lr": 1.4532826832560853e-06}
{"train_loss": 0.06585324555635452, "global_step": 250605, "epoch": 2815, "lr": 1.4531439288975434e-06}
{"train_loss": 0.04220841825008392, "global_step": 250606, "epoch": 2815, "lr": 1.453005181065542e-06}
{"train_loss": 0.04826667904853821, "global_step": 250607, "epoch": 2815, "lr": 1.4528664397600922e-06}
{"train_loss": 0.019883088767528534, "global_step": 250608, "epoch": 2815, "lr": 1.4527277049812104e-06}
{"train_loss": 0.09081714600324631, "global_step": 250609, "epoch": 2815, "lr": 1.4525889767289192e-06}
{"train_loss": 0.05330505967140198, "global_step": 250610, "epoch": 2815, "lr": 1.4524502550032348e-06}
{"train_loss": 0.0554702989757061, "global_step": 250611, "epoch": 2815, "lr": 1.4523115398041742e-06}
{"train_loss": 0.03804844990372658, "global_step": 250612, "epoch": 2815, "lr": 1.4521728311317651e-06}
{"train_loss": 0.03174690157175064, "global_step": 250613, "epoch": 2815, "lr": 1.4520341289860128e-06}
{"train_loss": 0.06117874011397362, "global_step": 250614, "epoch": 2815, "lr": 1.4518954333669453e-06}
{"train_loss": 0.03869491443037987, "global_step": 250615, "epoch": 2815, "lr": 1.4517567442745795e-06}
{"train_loss": 0.030177876353263855, "global_step": 250616, "epoch": 2815, "lr": 1.451618061708937e-06}
{"train_loss": 0.042918723076581955, "global_step": 250617, "epoch": 2815, "lr": 1.4514793856700237e-06}
{"train_loss": 0.04025097191333771, "global_step": 250618, "epoch": 2815, "lr": 1.4513407161578728e-06}
{"train_loss": 0.03239733353257179, "global_step": 250619, "epoch": 2815, "lr": 1.4512020531724956e-06}
{"train_loss": 0.030361419543623924, "global_step": 250620, "epoch": 2815, "lr": 1.4510633967139143e-06}
{"train_loss": 0.05290013179183006, "global_step": 250621, "epoch": 2815, "lr": 1.4509247467821453e-06}
{"train_loss": 0.025292333215475082, "global_step": 250622, "epoch": 2815, "lr": 1.4507861033772052e-06}
{"train_loss": 0.046282988575318554, "global_step": 250623, "epoch": 2815, "lr": 1.4506474664991222e-06, "val_loss": 9.551576614379883, "train_action_mse_error": 13.297906875610352}
{"train_loss": 0.08684328198432922, "global_step": 250624, "epoch": 2816, "lr": 1.4505088361478958e-06}
{"train_loss": 0.037597429007291794, "global_step": 250625, "epoch": 2816, "lr": 1.4503702123235707e-06}
{"train_loss": 0.06158924102783203, "global_step": 250626, "epoch": 2816, "lr": 1.4502315950261414e-06}
{"train_loss": 0.07207507640123367, "global_step": 250627, "epoch": 2816, "lr": 1.450092984255641e-06}
{"train_loss": 0.0838838592171669, "global_step": 250628, "epoch": 2816, "lr": 1.4499543800120808e-06}
{"train_loss": 0.02112853340804577, "global_step": 250629, "epoch": 2816, "lr": 1.4498157822954882e-06}
{"train_loss": 0.018006550148129463, "global_step": 250630, "epoch": 2816, "lr": 1.449677191105875e-06}
{"train_loss": 0.04372360184788704, "global_step": 250631, "epoch": 2816, "lr": 1.449538606443257e-06}
{"train_loss": 0.04008841514587402, "global_step": 250632, "epoch": 2816, "lr": 1.449400028307657e-06}
{"train_loss": 0.08541544526815414, "global_step": 250633, "epoch": 2816, "lr": 1.449261456699097e-06}
{"train_loss": 0.051249340176582336, "global_step": 250634, "epoch": 2816, "lr": 1.449122891617588e-06}
{"train_loss": 0.018920527771115303, "global_step": 250635, "epoch": 2816, "lr": 1.448984333063158e-06}
{"train_loss": 0.04934162646532059, "global_step": 250636, "epoch": 2816, "lr": 1.448845781035818e-06}
{"train_loss": 0.04729089513421059, "global_step": 250637, "epoch": 2816, "lr": 1.4487072355355902e-06}
{"train_loss": 0.05288931354880333, "global_step": 250638, "epoch": 2816, "lr": 1.4485686965624912e-06}
{"train_loss": 0.04934722185134888, "global_step": 250639, "epoch": 2816, "lr": 1.4484301641165377e-06}
{"train_loss": 0.025204846635460854, "global_step": 250640, "epoch": 2816, "lr": 1.4482916381977518e-06}
{"train_loss": 0.045289985835552216, "global_step": 250641, "epoch": 2816, "lr": 1.4481531188061504e-06}
{"train_loss": 0.04728014022111893, "global_step": 250642, "epoch": 2816, "lr": 1.4480146059417555e-06}
{"train_loss": 0.07475914806127548, "global_step": 250643, "epoch": 2816, "lr": 1.4478760996045837e-06}
{"train_loss": 0.04208601638674736, "global_step": 250644, "epoch": 2816, "lr": 1.4477375997946518e-06}
{"train_loss": 0.02725970186293125, "global_step": 250645, "epoch": 2816, "lr": 1.4475991065119764e-06}
{"train_loss": 0.05240897089242935, "global_step": 250646, "epoch": 2816, "lr": 1.4474606197565854e-06}
{"train_loss": 0.05136054754257202, "global_step": 250647, "epoch": 2816, "lr": 1.4473221395284897e-06}
{"train_loss": 0.017882486805319786, "global_step": 250648, "epoch": 2816, "lr": 1.4471836658277115e-06}
{"train_loss": 0.04231374338269234, "global_step": 250649, "epoch": 2816, "lr": 1.4470451986542621e-06}
{"train_loss": 0.01890568621456623, "global_step": 250650, "epoch": 2816, "lr": 1.4469067380081692e-06}
{"train_loss": 0.02901287004351616, "global_step": 250651, "epoch": 2816, "lr": 1.4467682838894493e-06}
{"train_loss": 0.06797719746828079, "global_step": 250652, "epoch": 2816, "lr": 1.446629836298119e-06}
{"train_loss": 0.04879986494779587, "global_step": 250653, "epoch": 2816, "lr": 1.4464913952341951e-06}
{"train_loss": 0.04624837264418602, "global_step": 250654, "epoch": 2816, "lr": 1.4463529606977e-06}
{"train_loss": 0.05182432383298874, "global_step": 250655, "epoch": 2816, "lr": 1.44621453268865e-06}
{"train_loss": 0.10939151048660278, "global_step": 250656, "epoch": 2816, "lr": 1.4460761112070674e-06}
{"train_loss": 0.020950475707650185, "global_step": 250657, "epoch": 2816, "lr": 1.4459376962529636e-06}
{"train_loss": 0.043706998229026794, "global_step": 250658, "epoch": 2816, "lr": 1.4457992878263605e-06}
{"train_loss": 0.061439547687768936, "global_step": 250659, "epoch": 2816, "lr": 1.4456608859272858e-06}
{"train_loss": 0.026092033833265305, "global_step": 250660, "epoch": 2816, "lr": 1.4455224905557451e-06}
{"train_loss": 0.05430071800947189, "global_step": 250661, "epoch": 2816, "lr": 1.4453841017117664e-06}
{"train_loss": 0.047494564205408096, "global_step": 250662, "epoch": 2816, "lr": 1.445245719395355e-06}
{"train_loss": 0.04449485242366791, "global_step": 250663, "epoch": 2816, "lr": 1.44510734360655e-06}
{"train_loss": 0.03457242250442505, "global_step": 250664, "epoch": 2816, "lr": 1.444968974345351e-06}
{"train_loss": 0.03258465602993965, "global_step": 250665, "epoch": 2816, "lr": 1.444830611611786e-06}
{"train_loss": 0.02842603623867035, "global_step": 250666, "epoch": 2816, "lr": 1.4446922554058718e-06}
{"train_loss": 0.08423339575529099, "global_step": 250667, "epoch": 2816, "lr": 1.4445539057276247e-06}
{"train_loss": 0.04818327724933624, "global_step": 250668, "epoch": 2816, "lr": 1.4444155625770672e-06}
{"train_loss": 0.04346396028995514, "global_step": 250669, "epoch": 2816, "lr": 1.4442772259542215e-06}
{"train_loss": 0.041118230670690536, "global_step": 250670, "epoch": 2816, "lr": 1.4441388958590928e-06}
{"train_loss": 0.014408934861421585, "global_step": 250671, "epoch": 2816, "lr": 1.444000572291715e-06}
{"train_loss": 0.02797062136232853, "global_step": 250672, "epoch": 2816, "lr": 1.4438622552520931e-06}
{"train_loss": 0.03848107531666756, "global_step": 250673, "epoch": 2816, "lr": 1.443723944740255e-06}
{"train_loss": 0.07210967689752579, "global_step": 250674, "epoch": 2816, "lr": 1.443585640756212e-06}
{"train_loss": 0.06942840665578842, "global_step": 250675, "epoch": 2816, "lr": 1.4434473432999974e-06}
{"train_loss": 0.06307908892631531, "global_step": 250676, "epoch": 2816, "lr": 1.4433090523716107e-06}
{"train_loss": 0.032260261476039886, "global_step": 250677, "epoch": 2816, "lr": 1.4431707679710804e-06}
{"train_loss": 0.040915597230196, "global_step": 250678, "epoch": 2816, "lr": 1.4430324900984283e-06}
{"train_loss": 0.05417542904615402, "global_step": 250679, "epoch": 2816, "lr": 1.4428942187536653e-06}
{"train_loss": 0.03705696016550064, "global_step": 250680, "epoch": 2816, "lr": 1.442755953936814e-06}
{"train_loss": 0.02465445175766945, "global_step": 250681, "epoch": 2816, "lr": 1.442617695647891e-06}
{"train_loss": 0.02313416823744774, "global_step": 250682, "epoch": 2816, "lr": 1.4424794438869183e-06}
{"train_loss": 0.02862381935119629, "global_step": 250683, "epoch": 2816, "lr": 1.4423411986539125e-06}
{"train_loss": 0.024571256712079048, "global_step": 250684, "epoch": 2816, "lr": 1.4422029599488907e-06}
{"train_loss": 0.028870368376374245, "global_step": 250685, "epoch": 2816, "lr": 1.4420647277718691e-06}
{"train_loss": 0.06690143793821335, "global_step": 250686, "epoch": 2816, "lr": 1.441926502122881e-06}
{"train_loss": 0.028999201953411102, "global_step": 250687, "epoch": 2816, "lr": 1.4417882830019213e-06}
{"train_loss": 0.03879670053720474, "global_step": 250688, "epoch": 2816, "lr": 1.4416500704090286e-06}
{"train_loss": 0.060520950704813004, "global_step": 250689, "epoch": 2816, "lr": 1.4415118643442138e-06}
{"train_loss": 0.05039455369114876, "global_step": 250690, "epoch": 2816, "lr": 1.4413736648074937e-06}
{"train_loss": 0.07122939825057983, "global_step": 250691, "epoch": 2816, "lr": 1.4412354717988907e-06}
{"train_loss": 0.0287270899862051, "global_step": 250692, "epoch": 2816, "lr": 1.4410972853184213e-06}
{"train_loss": 0.03529791161417961, "global_step": 250693, "epoch": 2816, "lr": 1.440959105366102e-06}
{"train_loss": 0.0726689025759697, "global_step": 250694, "epoch": 2816, "lr": 1.4408209319419553e-06}
{"train_loss": 0.07246336340904236, "global_step": 250695, "epoch": 2816, "lr": 1.440682765046003e-06}
{"train_loss": 0.009214717894792557, "global_step": 250696, "epoch": 2816, "lr": 1.440544604678251e-06}
{"train_loss": 0.037712614983320236, "global_step": 250697, "epoch": 2816, "lr": 1.4404064508387328e-06}
{"train_loss": 0.028281789273023605, "global_step": 250698, "epoch": 2816, "lr": 1.440268303527459e-06}
{"train_loss": 0.0994037613272667, "global_step": 250699, "epoch": 2816, "lr": 1.4401301627444464e-06}
{"train_loss": 0.01842494308948517, "global_step": 250700, "epoch": 2816, "lr": 1.4399920284897173e-06}
{"train_loss": 0.06588920950889587, "global_step": 250701, "epoch": 2816, "lr": 1.4398539007632938e-06}
{"train_loss": 0.036672718822956085, "global_step": 250702, "epoch": 2816, "lr": 1.4397157795651816e-06}
{"train_loss": 0.04409489780664444, "global_step": 250703, "epoch": 2816, "lr": 1.439577664895414e-06}
{"train_loss": 0.050231315195560455, "global_step": 250704, "epoch": 2816, "lr": 1.439439556754002e-06}
{"train_loss": 0.05767379328608513, "global_step": 250705, "epoch": 2816, "lr": 1.439301455140968e-06}
{"train_loss": 0.06597553938627243, "global_step": 250706, "epoch": 2816, "lr": 1.4391633600563226e-06}
{"train_loss": 0.015105816535651684, "global_step": 250707, "epoch": 2816, "lr": 1.439025271500094e-06}
{"train_loss": 0.045172762125730515, "global_step": 250708, "epoch": 2816, "lr": 1.4388871894722932e-06}
{"train_loss": 0.007943393662571907, "global_step": 250709, "epoch": 2816, "lr": 1.438749113972948e-06}
{"train_loss": 0.04798050597310066, "global_step": 250710, "epoch": 2816, "lr": 1.438611045002064e-06}
{"train_loss": 0.041978560388088226, "global_step": 250711, "epoch": 2816, "lr": 1.4384729825596744e-06}
{"train_loss": 0.04516051104815488, "global_step": 250712, "epoch": 2816, "lr": 1.438334926645779e-06, "val_loss": 9.479809761047363}
{"train_loss": 0.06742275506258011, "global_step": 250713, "epoch": 2817, "lr": 1.4381968772604172e-06}
{"train_loss": 0.04644537344574928, "global_step": 250714, "epoch": 2817, "lr": 1.4380588344035938e-06}
{"train_loss": 0.07002800703048706, "global_step": 250715, "epoch": 2817, "lr": 1.4379207980753373e-06}
{"train_loss": 0.03496336564421654, "global_step": 250716, "epoch": 2817, "lr": 1.4377827682756584e-06}
{"train_loss": 0.051558900624513626, "global_step": 250717, "epoch": 2817, "lr": 1.4376447450045738e-06}
{"train_loss": 0.05774703994393349, "global_step": 250718, "epoch": 2817, "lr": 1.4375067282621112e-06}
{"train_loss": 0.07653430104255676, "global_step": 250719, "epoch": 2817, "lr": 1.4373687180482765e-06}
{"train_loss": 0.033842552453279495, "global_step": 250720, "epoch": 2817, "lr": 1.4372307143631025e-06}
{"train_loss": 0.04729500785470009, "global_step": 250721, "epoch": 2817, "lr": 1.437092717206595e-06}
{"train_loss": 0.06096512824296951, "global_step": 250722, "epoch": 2817, "lr": 1.436954726578782e-06}
{"train_loss": 0.07087203115224838, "global_step": 250723, "epoch": 2817, "lr": 1.436816742479674e-06}
{"train_loss": 0.06988552212715149, "global_step": 250724, "epoch": 2817, "lr": 1.4366787649093049e-06}
{"train_loss": 0.052980728447437286, "global_step": 250725, "epoch": 2817, "lr": 1.436540793867669e-06}
{"train_loss": 0.02590283192694187, "global_step": 250726, "epoch": 2817, "lr": 1.4364028293548104e-06}
{"train_loss": 0.04145321622490883, "global_step": 250727, "epoch": 2817, "lr": 1.4362648713707294e-06}
{"train_loss": 0.038789741694927216, "global_step": 250728, "epoch": 2817, "lr": 1.4361269199154538e-06}
{"train_loss": 0.021496178582310677, "global_step": 250729, "epoch": 2817, "lr": 1.4359889749889888e-06}
{"train_loss": 0.015074195340275764, "global_step": 250730, "epoch": 2817, "lr": 1.435851036591368e-06}
{"train_loss": 0.034320440143346786, "global_step": 250731, "epoch": 2817, "lr": 1.4357131047226135e-06}
{"train_loss": 0.03378582373261452, "global_step": 250732, "epoch": 2817, "lr": 1.4355751793827254e-06}
{"train_loss": 0.04610436409711838, "global_step": 250733, "epoch": 2817, "lr": 1.4354372605717425e-06}
{"train_loss": 0.06332404911518097, "global_step": 250734, "epoch": 2817, "lr": 1.4352993482896647e-06}
{"train_loss": 0.05559340864419937, "global_step": 250735, "epoch": 2817, "lr": 1.4351614425365256e-06}
{"train_loss": 0.03738533705472946, "global_step": 250736, "epoch": 2817, "lr": 1.4350235433123305e-06}
{"train_loss": 0.10909546166658401, "global_step": 250737, "epoch": 2817, "lr": 1.4348856506171072e-06}
{"train_loss": 0.03785015270113945, "global_step": 250738, "epoch": 2817, "lr": 1.4347477644508723e-06}
{"train_loss": 0.03978782147169113, "global_step": 250739, "epoch": 2817, "lr": 1.4346098848136425e-06}
{"train_loss": 0.020365845412015915, "global_step": 250740, "epoch": 2817, "lr": 1.4344720117054344e-06}
{"train_loss": 0.1257174164056778, "global_step": 250741, "epoch": 2817, "lr": 1.434334145126276e-06}
{"train_loss": 0.047344524413347244, "global_step": 250742, "epoch": 2817, "lr": 1.434196285076178e-06}
{"train_loss": 0.030635394155979156, "global_step": 250743, "epoch": 2817, "lr": 1.4340584315551575e-06}
{"train_loss": 0.061393607407808304, "global_step": 250744, "epoch": 2817, "lr": 1.4339205845632363e-06}
{"train_loss": 0.028579743579030037, "global_step": 250745, "epoch": 2817, "lr": 1.433782744100437e-06}
{"train_loss": 0.05169190093874931, "global_step": 250746, "epoch": 2817, "lr": 1.4336449101667704e-06}
{"train_loss": 0.031319618225097656, "global_step": 250747, "epoch": 2817, "lr": 1.4335070827622587e-06}
{"train_loss": 0.0401710607111454, "global_step": 250748, "epoch": 2817, "lr": 1.4333692618869188e-06}
{"train_loss": 0.04735754057765007, "global_step": 250749, "epoch": 2817, "lr": 1.4332314475407727e-06}
{"train_loss": 0.06174430623650551, "global_step": 250750, "epoch": 2817, "lr": 1.4330936397238316e-06}
{"train_loss": 0.027542101219296455, "global_step": 250751, "epoch": 2817, "lr": 1.432955838436123e-06}
{"train_loss": 0.04946207255125046, "global_step": 250752, "epoch": 2817, "lr": 1.4328180436776638e-06}
{"train_loss": 0.07121900469064713, "global_step": 250753, "epoch": 2817, "lr": 1.4326802554484653e-06}
{"train_loss": 0.07681061327457428, "global_step": 250754, "epoch": 2817, "lr": 1.4325424737485604e-06}
{"train_loss": 0.03656434640288353, "global_step": 250755, "epoch": 2817, "lr": 1.4324046985779494e-06}
{"train_loss": 0.05648511275649071, "global_step": 250756, "epoch": 2817, "lr": 1.4322669299366599e-06}
{"train_loss": 0.01664208620786667, "global_step": 250757, "epoch": 2817, "lr": 1.4321291678247084e-06}
{"train_loss": 0.0413905493915081, "global_step": 250758, "epoch": 2817, "lr": 1.4319914122421229e-06}
{"train_loss": 0.05758435279130936, "global_step": 250759, "epoch": 2817, "lr": 1.431853663188909e-06}
{"train_loss": 0.06227593496441841, "global_step": 250760, "epoch": 2817, "lr": 1.4317159206650942e-06}
{"train_loss": 0.03522267937660217, "global_step": 250761, "epoch": 2817, "lr": 1.4315781846706843e-06}
{"train_loss": 0.08404555171728134, "global_step": 250762, "epoch": 2817, "lr": 1.4314404552057181e-06}
{"train_loss": 0.026903681457042694, "global_step": 250763, "epoch": 2817, "lr": 1.4313027322701956e-06}
{"train_loss": 0.042690444737672806, "global_step": 250764, "epoch": 2817, "lr": 1.4311650158641443e-06}
{"train_loss": 0.029154695570468903, "global_step": 250765, "epoch": 2817, "lr": 1.4310273059875811e-06}
{"train_loss": 0.04917306452989578, "global_step": 250766, "epoch": 2817, "lr": 1.4308896026405171e-06}
{"train_loss": 0.05814684182405472, "global_step": 250767, "epoch": 2817, "lr": 1.4307519058229912e-06}
{"train_loss": 0.030888091772794724, "global_step": 250768, "epoch": 2817, "lr": 1.4306142155349978e-06}
{"train_loss": 0.043884117156267166, "global_step": 250769, "epoch": 2817, "lr": 1.4304765317765757e-06}
{"train_loss": 0.019189467653632164, "global_step": 250770, "epoch": 2817, "lr": 1.4303388545477248e-06}
{"train_loss": 0.04773079976439476, "global_step": 250771, "epoch": 2817, "lr": 1.4302011838484842e-06}
{"train_loss": 0.028294522315263748, "global_step": 250772, "epoch": 2817, "lr": 1.4300635196788482e-06}
{"train_loss": 0.0801728144288063, "global_step": 250773, "epoch": 2817, "lr": 1.4299258620388556e-06}
{"train_loss": 0.05634302645921707, "global_step": 250774, "epoch": 2817, "lr": 1.4297882109285176e-06}
{"train_loss": 0.10033154487609863, "global_step": 250775, "epoch": 2817, "lr": 1.429650566347851e-06}
{"train_loss": 0.030785197392106056, "global_step": 250776, "epoch": 2817, "lr": 1.4295129282968723e-06}
{"train_loss": 0.062180034816265106, "global_step": 250777, "epoch": 2817, "lr": 1.4293752967756147e-06}
{"train_loss": 0.03237993270158768, "global_step": 250778, "epoch": 2817, "lr": 1.4292376717840728e-06}
{"train_loss": 0.03644482046365738, "global_step": 250779, "epoch": 2817, "lr": 1.4291000533222853e-06}
{"train_loss": 0.035824231803417206, "global_step": 250780, "epoch": 2817, "lr": 1.428962441390258e-06}
{"train_loss": 0.04197198152542114, "global_step": 250781, "epoch": 2817, "lr": 1.428824835988024e-06}
{"train_loss": 0.04140808433294296, "global_step": 250782, "epoch": 2817, "lr": 1.4286872371155835e-06}
{"train_loss": 0.015937039628624916, "global_step": 250783, "epoch": 2817, "lr": 1.4285496447729695e-06}
{"train_loss": 0.07561904191970825, "global_step": 250784, "epoch": 2817, "lr": 1.428412058960188e-06}
{"train_loss": 0.02664046175777912, "global_step": 250785, "epoch": 2817, "lr": 1.4282744796772662e-06}
{"train_loss": 0.10494759678840637, "global_step": 250786, "epoch": 2817, "lr": 1.4281369069242268e-06}
{"train_loss": 0.037332791835069656, "global_step": 250787, "epoch": 2817, "lr": 1.4279993407010806e-06}
{"train_loss": 0.0826980471611023, "global_step": 250788, "epoch": 2817, "lr": 1.4278617810078442e-06}
{"train_loss": 0.005864434875547886, "global_step": 250789, "epoch": 2817, "lr": 1.4277242278445456e-06}
{"train_loss": 0.026244957000017166, "global_step": 250790, "epoch": 2817, "lr": 1.4275866812111905e-06}
{"train_loss": 0.08488170057535172, "global_step": 250791, "epoch": 2817, "lr": 1.4274491411078118e-06}
{"train_loss": 0.05655385181307793, "global_step": 250792, "epoch": 2817, "lr": 1.4273116075344206e-06}
{"train_loss": 0.020830640569329262, "global_step": 250793, "epoch": 2817, "lr": 1.4271740804910282e-06}
{"train_loss": 0.08170882612466812, "global_step": 250794, "epoch": 2817, "lr": 1.427036559977668e-06}
{"train_loss": 0.04443321004509926, "global_step": 250795, "epoch": 2817, "lr": 1.4268990459943399e-06}
{"train_loss": 0.05745402351021767, "global_step": 250796, "epoch": 2817, "lr": 1.4267615385410826e-06}
{"train_loss": 0.042105622589588165, "global_step": 250797, "epoch": 2817, "lr": 1.426624037617902e-06}
{"train_loss": 0.04445784166455269, "global_step": 250798, "epoch": 2817, "lr": 1.4264865432248197e-06}
{"train_loss": 0.038053613156080246, "global_step": 250799, "epoch": 2817, "lr": 1.426349055361853e-06}
{"train_loss": 0.03489690646529198, "global_step": 250800, "epoch": 2817, "lr": 1.4262115740290239e-06}
{"train_loss": 0.04850908805187164, "global_step": 250801, "epoch": 2817, "lr": 1.4260740992263488e-06, "val_loss": 9.516281127929688}
{"train_loss": 0.09128274768590927, "global_step": 250802, "epoch": 2818, "lr": 1.4259366309538447e-06}
{"train_loss": 0.06040865182876587, "global_step": 250803, "epoch": 2818, "lr": 1.425799169211528e-06}
{"train_loss": 0.02594766952097416, "global_step": 250804, "epoch": 2818, "lr": 1.425661713999421e-06}
{"train_loss": 0.053670961409807205, "global_step": 250805, "epoch": 2818, "lr": 1.4255242653175516e-06}
{"train_loss": 0.0467197559773922, "global_step": 250806, "epoch": 2818, "lr": 1.4253868231659195e-06}
{"train_loss": 0.03932185843586922, "global_step": 250807, "epoch": 2818, "lr": 1.4252493875445526e-06}
{"train_loss": 0.03224216401576996, "global_step": 250808, "epoch": 2818, "lr": 1.4251119584534734e-06}
{"train_loss": 0.03302645683288574, "global_step": 250809, "epoch": 2818, "lr": 1.4249745358926924e-06}
{"train_loss": 0.07857894152402878, "global_step": 250810, "epoch": 2818, "lr": 1.4248371198622323e-06}
{"train_loss": 0.0388476587831974, "global_step": 250811, "epoch": 2818, "lr": 1.4246997103621097e-06}
{"train_loss": 0.03335033357143402, "global_step": 250812, "epoch": 2818, "lr": 1.4245623073923463e-06}
{"train_loss": 0.08160844445228577, "global_step": 250813, "epoch": 2818, "lr": 1.4244249109529595e-06}
{"train_loss": 0.02642570622265339, "global_step": 250814, "epoch": 2818, "lr": 1.42428752104396e-06}
{"train_loss": 0.022361090406775475, "global_step": 250815, "epoch": 2818, "lr": 1.424150137665381e-06}
{"train_loss": 0.03772670030593872, "global_step": 250816, "epoch": 2818, "lr": 1.4240127608172281e-06}
{"train_loss": 0.05448579788208008, "global_step": 250817, "epoch": 2818, "lr": 1.4238753904995238e-06}
{"train_loss": 0.0569341778755188, "global_step": 250818, "epoch": 2818, "lr": 1.4237380267122902e-06}
{"train_loss": 0.016206957399845123, "global_step": 250819, "epoch": 2818, "lr": 1.4236006694555436e-06}
{"train_loss": 0.006510693579912186, "global_step": 250820, "epoch": 2818, "lr": 1.4234633187293012e-06}
{"train_loss": 0.02169070392847061, "global_step": 250821, "epoch": 2818, "lr": 1.4233259745335793e-06}
{"train_loss": 0.010497091338038445, "global_step": 250822, "epoch": 2818, "lr": 1.4231886368684e-06}
{"train_loss": 0.030631328001618385, "global_step": 250823, "epoch": 2818, "lr": 1.4230513057337802e-06}
{"train_loss": 0.07466545701026917, "global_step": 250824, "epoch": 2818, "lr": 1.4229139811297421e-06}
{"train_loss": 0.02746700309216976, "global_step": 250825, "epoch": 2818, "lr": 1.4227766630563022e-06}
{"train_loss": 0.04949096590280533, "global_step": 250826, "epoch": 2818, "lr": 1.4226393515134774e-06}
{"train_loss": 0.050371523946523666, "global_step": 250827, "epoch": 2818, "lr": 1.422502046501284e-06}
{"train_loss": 0.049285225570201874, "global_step": 250828, "epoch": 2818, "lr": 1.4223647480197445e-06}
{"train_loss": 0.05216491222381592, "global_step": 250829, "epoch": 2818, "lr": 1.4222274560688754e-06}
{"train_loss": 0.05348479747772217, "global_step": 250830, "epoch": 2818, "lr": 1.422090170648699e-06}
{"train_loss": 0.09849482029676437, "global_step": 250831, "epoch": 2818, "lr": 1.4219528917592261e-06}
{"train_loss": 0.05002228170633316, "global_step": 250832, "epoch": 2818, "lr": 1.4218156194004851e-06}
{"train_loss": 0.062073711305856705, "global_step": 250833, "epoch": 2818, "lr": 1.421678353572481e-06}
{"train_loss": 0.040228214114904404, "global_step": 250834, "epoch": 2818, "lr": 1.4215410942752472e-06}
{"train_loss": 0.04694913700222969, "global_step": 250835, "epoch": 2818, "lr": 1.4214038415087893e-06}
{"train_loss": 0.10084383189678192, "global_step": 250836, "epoch": 2818, "lr": 1.4212665952731352e-06}
{"train_loss": 0.04753263667225838, "global_step": 250837, "epoch": 2818, "lr": 1.421129355568296e-06}
{"train_loss": 0.031018387526273727, "global_step": 250838, "epoch": 2818, "lr": 1.420992122394299e-06}
{"train_loss": 0.04827548563480377, "global_step": 250839, "epoch": 2818, "lr": 1.4208548957511558e-06}
{"train_loss": 0.021679209545254707, "global_step": 250840, "epoch": 2818, "lr": 1.420717675638883e-06}
{"train_loss": 0.05044487491250038, "global_step": 250841, "epoch": 2818, "lr": 1.420580462057508e-06}
{"train_loss": 0.004605171270668507, "global_step": 250842, "epoch": 2818, "lr": 1.4204432550070368e-06}
{"train_loss": 0.040944941341876984, "global_step": 250843, "epoch": 2818, "lr": 1.4203060544875023e-06}
{"train_loss": 0.07300315797328949, "global_step": 250844, "epoch": 2818, "lr": 1.4201688604989105e-06}
{"train_loss": 0.04900415241718292, "global_step": 250845, "epoch": 2818, "lr": 1.4200316730412888e-06}
{"train_loss": 0.023058343678712845, "global_step": 250846, "epoch": 2818, "lr": 1.4198944921146483e-06}
{"train_loss": 0.06606242060661316, "global_step": 250847, "epoch": 2818, "lr": 1.419757317719017e-06}
{"train_loss": 0.05198286473751068, "global_step": 250848, "epoch": 2818, "lr": 1.4196201498544003e-06}
{"train_loss": 0.032503869384527206, "global_step": 250849, "epoch": 2818, "lr": 1.419482988520826e-06}
{"train_loss": 0.019722437486052513, "global_step": 250850, "epoch": 2818, "lr": 1.4193458337183107e-06}
{"train_loss": 0.035894736647605896, "global_step": 250851, "epoch": 2818, "lr": 1.419208685446871e-06}
{"train_loss": 0.04529707878828049, "global_step": 250852, "epoch": 2818, "lr": 1.4190715437065238e-06}
{"train_loss": 0.03198767825961113, "global_step": 250853, "epoch": 2818, "lr": 1.4189344084972966e-06}
{"train_loss": 0.03687402978539467, "global_step": 250854, "epoch": 2818, "lr": 1.418797279819195e-06}
{"train_loss": 0.06308361142873764, "global_step": 250855, "epoch": 2818, "lr": 1.4186601576722524e-06}
{"train_loss": 0.027347926050424576, "global_step": 250856, "epoch": 2818, "lr": 1.4185230420564744e-06}
{"train_loss": 0.04717572033405304, "global_step": 250857, "epoch": 2818, "lr": 1.4183859329718774e-06}
{"train_loss": 0.047711797058582306, "global_step": 250858, "epoch": 2818, "lr": 1.418248830418495e-06}
{"train_loss": 0.0482567697763443, "global_step": 250859, "epoch": 2818, "lr": 1.4181117343963325e-06}
{"train_loss": 0.0786404013633728, "global_step": 250860, "epoch": 2818, "lr": 1.4179746449054176e-06}
{"train_loss": 0.05429305508732796, "global_step": 250861, "epoch": 2818, "lr": 1.4178375619457563e-06}
{"train_loss": 0.06446523219347, "global_step": 250862, "epoch": 2818, "lr": 1.4177004855173814e-06}
{"train_loss": 0.028825033456087112, "global_step": 250863, "epoch": 2818, "lr": 1.4175634156203044e-06}
{"train_loss": 0.03608062118291855, "global_step": 250864, "epoch": 2818, "lr": 1.4174263522545417e-06}
{"train_loss": 0.04874559864401817, "global_step": 250865, "epoch": 2818, "lr": 1.4172892954201156e-06}
{"train_loss": 0.007482847198843956, "global_step": 250866, "epoch": 2818, "lr": 1.4171522451170426e-06}
{"train_loss": 0.037594668567180634, "global_step": 250867, "epoch": 2818, "lr": 1.4170152013453397e-06}
{"train_loss": 0.03185449168086052, "global_step": 250868, "epoch": 2818, "lr": 1.4168781641050233e-06}
{"train_loss": 0.026776036247611046, "global_step": 250869, "epoch": 2818, "lr": 1.416741133396121e-06}
{"train_loss": 0.0730353593826294, "global_step": 250870, "epoch": 2818, "lr": 1.41660410921865e-06}
{"train_loss": 0.06734328716993332, "global_step": 250871, "epoch": 2818, "lr": 1.4164670915726153e-06}
{"train_loss": 0.06781881302595139, "global_step": 250872, "epoch": 2818, "lr": 1.4163300804580504e-06}
{"train_loss": 0.0462186336517334, "global_step": 250873, "epoch": 2818, "lr": 1.416193075874961e-06}
{"train_loss": 0.02089715749025345, "global_step": 250874, "epoch": 2818, "lr": 1.41605607782338e-06}
{"train_loss": 0.028607560321688652, "global_step": 250875, "epoch": 2818, "lr": 1.4159190863033135e-06}
{"train_loss": 0.056739699095487595, "global_step": 250876, "epoch": 2818, "lr": 1.4157821013147831e-06}
{"train_loss": 0.022719085216522217, "global_step": 250877, "epoch": 2818, "lr": 1.4156451228578117e-06}
{"train_loss": 0.037228796631097794, "global_step": 250878, "epoch": 2818, "lr": 1.4155081509324098e-06}
{"train_loss": 0.02034451998770237, "global_step": 250879, "epoch": 2818, "lr": 1.415371185538611e-06}
{"train_loss": 0.0459614060819149, "global_step": 250880, "epoch": 2818, "lr": 1.4152342266764097e-06}
{"train_loss": 0.05157848820090294, "global_step": 250881, "epoch": 2818, "lr": 1.4150972743458502e-06}
{"train_loss": 0.029016617685556412, "global_step": 250882, "epoch": 2818, "lr": 1.4149603285469327e-06}
{"train_loss": 0.08355932682752609, "global_step": 250883, "epoch": 2818, "lr": 1.4148233892796848e-06}
{"train_loss": 0.02845616266131401, "global_step": 250884, "epoch": 2818, "lr": 1.4146864565441176e-06}
{"train_loss": 0.03357846289873123, "global_step": 250885, "epoch": 2818, "lr": 1.414549530340259e-06}
{"train_loss": 0.06819590926170349, "global_step": 250886, "epoch": 2818, "lr": 1.4144126106681144e-06}
{"train_loss": 0.023437513038516045, "global_step": 250887, "epoch": 2818, "lr": 1.4142756975277172e-06}
{"train_loss": 0.04188557341694832, "global_step": 250888, "epoch": 2818, "lr": 1.414138790919073e-06}
{"train_loss": 0.025827210396528244, "global_step": 250889, "epoch": 2818, "lr": 1.4140018908422093e-06}
{"train_loss": 0.04402300111656444, "global_step": 250890, "epoch": 2818, "lr": 1.4138649972971374e-06, "val_loss": 9.50722599029541}
{"train_loss": 0.06142948567867279, "global_step": 250891, "epoch": 2819, "lr": 1.413728110283885e-06}
{"train_loss": 0.045870907604694366, "global_step": 250892, "epoch": 2819, "lr": 1.4135912298024578e-06}
{"train_loss": 0.026229919865727425, "global_step": 250893, "epoch": 2819, "lr": 1.4134543558528834e-06}
{"train_loss": 0.0174082163721323, "global_step": 250894, "epoch": 2819, "lr": 1.4133174884351841e-06}
{"train_loss": 0.08552229404449463, "global_step": 250895, "epoch": 2819, "lr": 1.4131806275493597e-06}
{"train_loss": 0.038043312728405, "global_step": 250896, "epoch": 2819, "lr": 1.413043773195455e-06}
{"train_loss": 0.039465393871068954, "global_step": 250897, "epoch": 2819, "lr": 1.412906925373464e-06}
{"train_loss": 0.026224814355373383, "global_step": 250898, "epoch": 2819, "lr": 1.4127700840834202e-06}
{"train_loss": 0.0466882698237896, "global_step": 250899, "epoch": 2819, "lr": 1.4126332493253346e-06}
{"train_loss": 0.06555071473121643, "global_step": 250900, "epoch": 2819, "lr": 1.4124964210992298e-06}
{"train_loss": 0.05920438468456268, "global_step": 250901, "epoch": 2819, "lr": 1.412359599405122e-06}
{"train_loss": 0.042562276124954224, "global_step": 250902, "epoch": 2819, "lr": 1.4122227842430335e-06}
{"train_loss": 0.057183653116226196, "global_step": 250903, "epoch": 2819, "lr": 1.412085975612981e-06}
{"train_loss": 0.05156272277235985, "global_step": 250904, "epoch": 2819, "lr": 1.4119491735149759e-06}
{"train_loss": 0.03332394361495972, "global_step": 250905, "epoch": 2819, "lr": 1.4118123779490454e-06}
{"train_loss": 0.09561970084905624, "global_step": 250906, "epoch": 2819, "lr": 1.4116755889151955e-06}
{"train_loss": 0.06709133833646774, "global_step": 250907, "epoch": 2819, "lr": 1.4115388064134648e-06}
{"train_loss": 0.06414236128330231, "global_step": 250908, "epoch": 2819, "lr": 1.4114020304438591e-06}
{"train_loss": 0.054707273840904236, "global_step": 250909, "epoch": 2819, "lr": 1.4112652610063947e-06}
{"train_loss": 0.06786130368709564, "global_step": 250910, "epoch": 2819, "lr": 1.4111284981010942e-06}
{"train_loss": 0.02635551244020462, "global_step": 250911, "epoch": 2819, "lr": 1.410991741727974e-06}
{"train_loss": 0.09082458168268204, "global_step": 250912, "epoch": 2819, "lr": 1.4108549918870506e-06}
{"train_loss": 0.03362804651260376, "global_step": 250913, "epoch": 2819, "lr": 1.4107182485783522e-06}
{"train_loss": 0.01450512558221817, "global_step": 250914, "epoch": 2819, "lr": 1.4105815118018895e-06}
{"train_loss": 0.02876870520412922, "global_step": 250915, "epoch": 2819, "lr": 1.4104447815576793e-06}
{"train_loss": 0.05450655519962311, "global_step": 250916, "epoch": 2819, "lr": 1.410308057845744e-06}
{"train_loss": 0.06238715723156929, "global_step": 250917, "epoch": 2819, "lr": 1.4101713406661e-06}
{"train_loss": 0.04734419658780098, "global_step": 250918, "epoch": 2819, "lr": 1.410034630018764e-06}
{"train_loss": 0.06985116004943848, "global_step": 250919, "epoch": 2819, "lr": 1.4098979259037636e-06}
{"train_loss": 0.034471143037080765, "global_step": 250920, "epoch": 2819, "lr": 1.409761228321105e-06}
{"train_loss": 0.035468488931655884, "global_step": 250921, "epoch": 2819, "lr": 1.409624537270815e-06}
{"train_loss": 0.0581679604947567, "global_step": 250922, "epoch": 2819, "lr": 1.4094878527529053e-06}
{"train_loss": 0.09646622091531754, "global_step": 250923, "epoch": 2819, "lr": 1.4093511747673982e-06}
{"train_loss": 0.04492408409714699, "global_step": 250924, "epoch": 2819, "lr": 1.40921450331431e-06}
{"train_loss": 0.05195233225822449, "global_step": 250925, "epoch": 2819, "lr": 1.409077838393663e-06}
{"train_loss": 0.07541825622320175, "global_step": 250926, "epoch": 2819, "lr": 1.4089411800054742e-06}
{"train_loss": 0.05522140860557556, "global_step": 250927, "epoch": 2819, "lr": 1.4088045281497597e-06}
{"train_loss": 0.039366114884614944, "global_step": 250928, "epoch": 2819, "lr": 1.4086678828265366e-06}
{"train_loss": 0.06230611354112625, "global_step": 250929, "epoch": 2819, "lr": 1.4085312440358322e-06}
{"train_loss": 0.04011044651269913, "global_step": 250930, "epoch": 2819, "lr": 1.4083946117776525e-06}
{"train_loss": 0.035639688372612, "global_step": 250931, "epoch": 2819, "lr": 1.4082579860520252e-06}
{"train_loss": 0.0397908054292202, "global_step": 250932, "epoch": 2819, "lr": 1.4081213668589665e-06}
{"train_loss": 0.06294910609722137, "global_step": 250933, "epoch": 2819, "lr": 1.407984754198488e-06}
{"train_loss": 0.028693970292806625, "global_step": 250934, "epoch": 2819, "lr": 1.4078481480706174e-06}
{"train_loss": 0.04307307302951813, "global_step": 250935, "epoch": 2819, "lr": 1.407711548475371e-06}
{"train_loss": 0.03850775212049484, "global_step": 250936, "epoch": 2819, "lr": 1.407574955412766e-06}
{"train_loss": 0.061076343059539795, "global_step": 250937, "epoch": 2819, "lr": 1.407438368882813e-06}
{"train_loss": 0.0360337533056736, "global_step": 250938, "epoch": 2819, "lr": 1.4073017888855456e-06}
{"train_loss": 0.04664217308163643, "global_step": 250939, "epoch": 2819, "lr": 1.4071652154209692e-06}
{"train_loss": 0.04535989463329315, "global_step": 250940, "epoch": 2819, "lr": 1.4070286484891115e-06}
{"train_loss": 0.0428374782204628, "global_step": 250941, "epoch": 2819, "lr": 1.4068920880899838e-06}
{"train_loss": 0.055750396102666855, "global_step": 250942, "epoch": 2819, "lr": 1.4067555342236083e-06}
{"train_loss": 0.04655349999666214, "global_step": 250943, "epoch": 2819, "lr": 1.406618986890007e-06}
{"train_loss": 0.09452369809150696, "global_step": 250944, "epoch": 2819, "lr": 1.4064824460891856e-06}
{"train_loss": 0.03899356350302696, "global_step": 250945, "epoch": 2819, "lr": 1.406345911821172e-06}
{"train_loss": 0.0839945375919342, "global_step": 250946, "epoch": 2819, "lr": 1.4062093840859825e-06}
{"train_loss": 0.049234405159950256, "global_step": 250947, "epoch": 2819, "lr": 1.4060728628836396e-06}
{"train_loss": 0.05358744412660599, "global_step": 250948, "epoch": 2819, "lr": 1.4059363482141542e-06}
{"train_loss": 0.1220085620880127, "global_step": 250949, "epoch": 2819, "lr": 1.4057998400775541e-06}
{"train_loss": 0.027367383241653442, "global_step": 250950, "epoch": 2819, "lr": 1.405663338473845e-06}
{"train_loss": 0.042181577533483505, "global_step": 250951, "epoch": 2819, "lr": 1.4055268434030545e-06}
{"train_loss": 0.07057476788759232, "global_step": 250952, "epoch": 2819, "lr": 1.4053903548651991e-06}
{"train_loss": 0.03393646329641342, "global_step": 250953, "epoch": 2819, "lr": 1.4052538728602959e-06}
{"train_loss": 0.01630774326622486, "global_step": 250954, "epoch": 2819, "lr": 1.4051173973883614e-06}
{"train_loss": 0.06134152039885521, "global_step": 250955, "epoch": 2819, "lr": 1.4049809284494231e-06}
{"train_loss": 0.11282171308994293, "global_step": 250956, "epoch": 2819, "lr": 1.4048444660434867e-06}
{"train_loss": 0.04780052229762077, "global_step": 250957, "epoch": 2819, "lr": 1.4047080101705856e-06}
{"train_loss": 0.047885384410619736, "global_step": 250958, "epoch": 2819, "lr": 1.4045715608307198e-06}
{"train_loss": 0.026471737772226334, "global_step": 250959, "epoch": 2819, "lr": 1.4044351180239223e-06}
{"train_loss": 0.045051198452711105, "global_step": 250960, "epoch": 2819, "lr": 1.4042986817501991e-06}
{"train_loss": 0.0712844580411911, "global_step": 250961, "epoch": 2819, "lr": 1.4041622520095832e-06}
{"train_loss": 0.054396502673625946, "global_step": 250962, "epoch": 2819, "lr": 1.4040258288020857e-06}
{"train_loss": 0.050348591059446335, "global_step": 250963, "epoch": 2819, "lr": 1.4038894121277236e-06}
{"train_loss": 0.01763475313782692, "global_step": 250964, "epoch": 2819, "lr": 1.4037530019865131e-06}
{"train_loss": 0.024626385420560837, "global_step": 250965, "epoch": 2819, "lr": 1.4036165983784766e-06}
{"train_loss": 0.06131305545568466, "global_step": 250966, "epoch": 2819, "lr": 1.4034802013036309e-06}
{"train_loss": 0.03474610298871994, "global_step": 250967, "epoch": 2819, "lr": 1.4033438107619922e-06}
{"train_loss": 0.02150806598365307, "global_step": 250968, "epoch": 2819, "lr": 1.403207426753589e-06}
{"train_loss": 0.04572555795311928, "global_step": 250969, "epoch": 2819, "lr": 1.403071049278426e-06}
{"train_loss": 0.027712510898709297, "global_step": 250970, "epoch": 2819, "lr": 1.4029346783365316e-06}
{"train_loss": 0.03595865145325661, "global_step": 250971, "epoch": 2819, "lr": 1.4027983139279166e-06}
{"train_loss": 0.050894491374492645, "global_step": 250972, "epoch": 2819, "lr": 1.4026619560526088e-06}
{"train_loss": 0.054070234298706055, "global_step": 250973, "epoch": 2819, "lr": 1.4025256047106194e-06}
{"train_loss": 0.01964046247303486, "global_step": 250974, "epoch": 2819, "lr": 1.402389259901965e-06}
{"train_loss": 0.017962513491511345, "global_step": 250975, "epoch": 2819, "lr": 1.4022529216266677e-06}
{"train_loss": 0.0670170933008194, "global_step": 250976, "epoch": 2819, "lr": 1.4021165898847499e-06}
{"train_loss": 0.030878949910402298, "global_step": 250977, "epoch": 2819, "lr": 1.401980264676217e-06}
{"train_loss": 0.04269732907414436, "global_step": 250978, "epoch": 2819, "lr": 1.4018439460011024e-06}
{"train_loss": 0.050162671740805166, "global_step": 250979, "epoch": 2819, "lr": 1.4017076338594116e-06, "val_loss": 9.49580192565918}
{"train_loss": 0.07904493063688278, "global_step": 250980, "epoch": 2820, "lr": 1.4015713282511777e-06}
{"train_loss": 0.03086930513381958, "global_step": 250981, "epoch": 2820, "lr": 1.4014350291764066e-06}
{"train_loss": 0.020906813442707062, "global_step": 250982, "epoch": 2820, "lr": 1.401298736635115e-06}
{"train_loss": 0.006990513298660517, "global_step": 250983, "epoch": 2820, "lr": 1.4011624506273357e-06}
{"train_loss": 0.03440358117222786, "global_step": 250984, "epoch": 2820, "lr": 1.4010261711530692e-06}
{"train_loss": 0.011289678514003754, "global_step": 250985, "epoch": 2820, "lr": 1.4008898982123431e-06}
{"train_loss": 0.04316068813204765, "global_step": 250986, "epoch": 2820, "lr": 1.4007536318051794e-06}
{"train_loss": 0.02169402502477169, "global_step": 250987, "epoch": 2820, "lr": 1.4006173719315895e-06}
{"train_loss": 0.0488930344581604, "global_step": 250988, "epoch": 2820, "lr": 1.4004811185915901e-06}
{"train_loss": 0.017709849402308464, "global_step": 250989, "epoch": 2820, "lr": 1.4003448717852141e-06}
{"train_loss": 0.0217269379645586, "global_step": 250990, "epoch": 2820, "lr": 1.400208631512462e-06}
{"train_loss": 0.06496360152959824, "global_step": 250991, "epoch": 2820, "lr": 1.400072397773361e-06}
{"train_loss": 0.037386953830718994, "global_step": 250992, "epoch": 2820, "lr": 1.3999361705679225e-06}
{"train_loss": 0.03255453333258629, "global_step": 250993, "epoch": 2820, "lr": 1.39979994989618e-06}
{"train_loss": 0.086715467274189, "global_step": 250994, "epoch": 2820, "lr": 1.3996637357581332e-06}
{"train_loss": 0.03218688443303108, "global_step": 250995, "epoch": 2820, "lr": 1.3995275281538212e-06}
{"train_loss": 0.011399004608392715, "global_step": 250996, "epoch": 2820, "lr": 1.3993913270832382e-06}
{"train_loss": 0.07916230708360672, "global_step": 250997, "epoch": 2820, "lr": 1.3992551325464233e-06}
{"train_loss": 0.05012335628271103, "global_step": 250998, "epoch": 2820, "lr": 1.399118944543376e-06}
{"train_loss": 0.05252234637737274, "global_step": 250999, "epoch": 2820, "lr": 1.3989827630741358e-06}
{"train_loss": 0.036537375301122665, "global_step": 251000, "epoch": 2820, "lr": 1.3988465881387026e-06}
{"train_loss": 0.04315509274601936, "global_step": 251001, "epoch": 2820, "lr": 1.3987104197371038e-06}
{"train_loss": 0.07065024971961975, "global_step": 251002, "epoch": 2820, "lr": 1.3985742578693562e-06}
{"train_loss": 0.03239546716213226, "global_step": 251003, "epoch": 2820, "lr": 1.3984381025354764e-06}
{"train_loss": 0.029685519635677338, "global_step": 251004, "epoch": 2820, "lr": 1.398301953735487e-06}
{"train_loss": 0.0366964153945446, "global_step": 251005, "epoch": 2820, "lr": 1.398165811469404e-06}
{"train_loss": 0.06358209252357483, "global_step": 251006, "epoch": 2820, "lr": 1.3980296757372447e-06}
{"train_loss": 0.0228769201785326, "global_step": 251007, "epoch": 2820, "lr": 1.3978935465390252e-06}
{"train_loss": 0.08014516532421112, "global_step": 251008, "epoch": 2820, "lr": 1.3977574238747737e-06}
{"train_loss": 0.06061607599258423, "global_step": 251009, "epoch": 2820, "lr": 1.39762130774449e-06}
{"train_loss": 0.027879580855369568, "global_step": 251010, "epoch": 2820, "lr": 1.3974851981482128e-06}
{"train_loss": 0.03898776322603226, "global_step": 251011, "epoch": 2820, "lr": 1.397349095085948e-06}
{"train_loss": 0.06235774606466293, "global_step": 251012, "epoch": 2820, "lr": 1.3972129985577175e-06}
{"train_loss": 0.013768634758889675, "global_step": 251013, "epoch": 2820, "lr": 1.3970769085635382e-06}
{"train_loss": 0.03123677708208561, "global_step": 251014, "epoch": 2820, "lr": 1.3969408251034321e-06}
{"train_loss": 0.03834497928619385, "global_step": 251015, "epoch": 2820, "lr": 1.3968047481774161e-06}
{"train_loss": 0.09331810474395752, "global_step": 251016, "epoch": 2820, "lr": 1.3966686777855065e-06}
{"train_loss": 0.07277607917785645, "global_step": 251017, "epoch": 2820, "lr": 1.3965326139277147e-06}
{"train_loss": 0.02058257721364498, "global_step": 251018, "epoch": 2820, "lr": 1.3963965566040792e-06}
{"train_loss": 0.06239263713359833, "global_step": 251019, "epoch": 2820, "lr": 1.3962605058146005e-06}
{"train_loss": 0.03510825335979462, "global_step": 251020, "epoch": 2820, "lr": 1.3961244615593005e-06}
{"train_loss": 0.05169551074504852, "global_step": 251021, "epoch": 2820, "lr": 1.3959884238382015e-06}
{"train_loss": 0.03551654890179634, "global_step": 251022, "epoch": 2820, "lr": 1.3958523926513145e-06}
{"train_loss": 0.017253952100872993, "global_step": 251023, "epoch": 2820, "lr": 1.3957163679986673e-06}
{"train_loss": 0.01814981922507286, "global_step": 251024, "epoch": 2820, "lr": 1.3955803498802712e-06}
{"train_loss": 0.07977207750082016, "global_step": 251025, "epoch": 2820, "lr": 1.3954443382961535e-06}
{"train_loss": 0.01737814024090767, "global_step": 251026, "epoch": 2820, "lr": 1.3953083332463147e-06}
{"train_loss": 0.03367214649915695, "global_step": 251027, "epoch": 2820, "lr": 1.3951723347307931e-06}
{"train_loss": 0.043362412601709366, "global_step": 251028, "epoch": 2820, "lr": 1.395036342749595e-06}
{"train_loss": 0.034424688667058945, "global_step": 251029, "epoch": 2820, "lr": 1.394900357302742e-06}
{"train_loss": 0.04222942888736725, "global_step": 251030, "epoch": 2820, "lr": 1.3947643783902509e-06}
{"train_loss": 0.03184465318918228, "global_step": 251031, "epoch": 2820, "lr": 1.394628406012144e-06}
{"train_loss": 0.05585057660937309, "global_step": 251032, "epoch": 2820, "lr": 1.3944924401684323e-06}
{"train_loss": 0.05938757210969925, "global_step": 251033, "epoch": 2820, "lr": 1.3943564808591437e-06}
{"train_loss": 0.08517691493034363, "global_step": 251034, "epoch": 2820, "lr": 1.394220528084289e-06}
{"train_loss": 0.04796634986996651, "global_step": 251035, "epoch": 2820, "lr": 1.3940845818438909e-06}
{"train_loss": 0.014338037930428982, "global_step": 251036, "epoch": 2820, "lr": 1.3939486421379655e-06}
{"train_loss": 0.037139929831027985, "global_step": 251037, "epoch": 2820, "lr": 1.3938127089665299e-06}
{"train_loss": 0.05686932057142258, "global_step": 251038, "epoch": 2820, "lr": 1.3936767823296004e-06}
{"train_loss": 0.06759348511695862, "global_step": 251039, "epoch": 2820, "lr": 1.3935408622271995e-06}
{"train_loss": 0.04178619757294655, "global_step": 251040, "epoch": 2820, "lr": 1.3934049486593548e-06}
{"train_loss": 0.05280134081840515, "global_step": 251041, "epoch": 2820, "lr": 1.3932690416260608e-06}
{"train_loss": 0.04924161732196808, "global_step": 251042, "epoch": 2820, "lr": 1.3931331411273562e-06}
{"train_loss": 0.05578302592039108, "global_step": 251043, "epoch": 2820, "lr": 1.3929972471632524e-06}
{"train_loss": 0.01557372510433197, "global_step": 251044, "epoch": 2820, "lr": 1.3928613597337715e-06}
{"train_loss": 0.06858961284160614, "global_step": 251045, "epoch": 2820, "lr": 1.3927254788389187e-06}
{"train_loss": 0.032563064247369766, "global_step": 251046, "epoch": 2820, "lr": 1.392589604478728e-06}
{"train_loss": 0.04234170913696289, "global_step": 251047, "epoch": 2820, "lr": 1.3924537366532097e-06}
{"train_loss": 0.06840919703245163, "global_step": 251048, "epoch": 2820, "lr": 1.3923178753623812e-06}
{"train_loss": 0.032730672508478165, "global_step": 251049, "epoch": 2820, "lr": 1.392182020606264e-06}
{"train_loss": 0.03956887125968933, "global_step": 251050, "epoch": 2820, "lr": 1.392046172384881e-06}
{"train_loss": 0.078012615442276, "global_step": 251051, "epoch": 2820, "lr": 1.3919103306982374e-06}
{"train_loss": 0.025468859821558, "global_step": 251052, "epoch": 2820, "lr": 1.3917744955463664e-06}
{"train_loss": 0.044587746262550354, "global_step": 251053, "epoch": 2820, "lr": 1.3916386669292735e-06}
{"train_loss": 0.035508040338754654, "global_step": 251054, "epoch": 2820, "lr": 1.3915028448469814e-06}
{"train_loss": 0.030346518382430077, "global_step": 251055, "epoch": 2820, "lr": 1.3913670292995118e-06}
{"train_loss": 0.061529505997896194, "global_step": 251056, "epoch": 2820, "lr": 1.3912312202868817e-06}
{"train_loss": 0.024369297549128532, "global_step": 251057, "epoch": 2820, "lr": 1.3910954178091019e-06}
{"train_loss": 0.02344720996916294, "global_step": 251058, "epoch": 2820, "lr": 1.3909596218662058e-06}
{"train_loss": 0.04530605673789978, "global_step": 251059, "epoch": 2820, "lr": 1.390823832458199e-06}
{"train_loss": 0.03084610588848591, "global_step": 251060, "epoch": 2820, "lr": 1.3906880495851037e-06}
{"train_loss": 0.024826323613524437, "global_step": 251061, "epoch": 2820, "lr": 1.3905522732469366e-06}
{"train_loss": 0.018824147060513496, "global_step": 251062, "epoch": 2820, "lr": 1.39041650344372e-06}
{"train_loss": 0.03545266017317772, "global_step": 251063, "epoch": 2820, "lr": 1.3902807401754646e-06}
{"train_loss": 0.03366139158606529, "global_step": 251064, "epoch": 2820, "lr": 1.3901449834421987e-06}
{"train_loss": 0.03590040281414986, "global_step": 251065, "epoch": 2820, "lr": 1.390009233243933e-06}
{"train_loss": 0.03224596381187439, "global_step": 251066, "epoch": 2820, "lr": 1.3898734895806842e-06}
{"train_loss": 0.0571218878030777, "global_step": 251067, "epoch": 2820, "lr": 1.38973775245248e-06}
{"train_loss": 0.04198144109961524, "global_step": 251068, "epoch": 2820, "lr": 1.3896020218593319e-06, "val_loss": 9.611368179321289, "train_action_mse_error": 27.066774368286133}
{"train_loss": 0.05067785084247589, "global_step": 251069, "epoch": 2821, "lr": 1.3894662978012619e-06}
{"train_loss": 0.06301075965166092, "global_step": 251070, "epoch": 2821, "lr": 1.3893305802782753e-06}
{"train_loss": 0.0419897623360157, "global_step": 251071, "epoch": 2821, "lr": 1.389194869290411e-06}
{"train_loss": 0.07640926539897919, "global_step": 251072, "epoch": 2821, "lr": 1.3890591648376695e-06}
{"train_loss": 0.019034042954444885, "global_step": 251073, "epoch": 2821, "lr": 1.3889234669200836e-06}
{"train_loss": 0.03643832728266716, "global_step": 251074, "epoch": 2821, "lr": 1.3887877755376588e-06}
{"train_loss": 0.04059490188956261, "global_step": 251075, "epoch": 2821, "lr": 1.3886520906904233e-06}
{"train_loss": 0.0199489276856184, "global_step": 251076, "epoch": 2821, "lr": 1.3885164123783933e-06}
{"train_loss": 0.02770317904651165, "global_step": 251077, "epoch": 2821, "lr": 1.3883807406015748e-06}
{"train_loss": 0.048231929540634155, "global_step": 251078, "epoch": 2821, "lr": 1.3882450753600062e-06}
{"train_loss": 0.03290276974439621, "global_step": 251079, "epoch": 2821, "lr": 1.3881094166536879e-06}
{"train_loss": 0.024968698620796204, "global_step": 251080, "epoch": 2821, "lr": 1.387973764482653e-06}
{"train_loss": 0.013834104873239994, "global_step": 251081, "epoch": 2821, "lr": 1.387838118846907e-06}
{"train_loss": 0.058101586997509, "global_step": 251082, "epoch": 2821, "lr": 1.3877024797464777e-06}
{"train_loss": 0.025511527433991432, "global_step": 251083, "epoch": 2821, "lr": 1.3875668471813764e-06}
{"train_loss": 0.037727050483226776, "global_step": 251084, "epoch": 2821, "lr": 1.387431221151625e-06}
{"train_loss": 0.0955691933631897, "global_step": 251085, "epoch": 2821, "lr": 1.3872956016572403e-06}
{"train_loss": 0.01673944480717182, "global_step": 251086, "epoch": 2821, "lr": 1.3871599886982446e-06}
{"train_loss": 0.01347869448363781, "global_step": 251087, "epoch": 2821, "lr": 1.3870243822746487e-06}
{"train_loss": 0.02931182272732258, "global_step": 251088, "epoch": 2821, "lr": 1.386888782386475e-06}
{"train_loss": 0.021958407014608383, "global_step": 251089, "epoch": 2821, "lr": 1.3867531890337404e-06}
{"train_loss": 0.032844994217157364, "global_step": 251090, "epoch": 2821, "lr": 1.3866176022164723e-06}
{"train_loss": 0.03764023259282112, "global_step": 251091, "epoch": 2821, "lr": 1.3864820219346709e-06}
{"train_loss": 0.04837111011147499, "global_step": 251092, "epoch": 2821, "lr": 1.3863464481883692e-06}
{"train_loss": 0.03318376839160919, "global_step": 251093, "epoch": 2821, "lr": 1.3862108809775788e-06}
{"train_loss": 0.020581115037202835, "global_step": 251094, "epoch": 2821, "lr": 1.3860753203023268e-06}
{"train_loss": 0.037999942898750305, "global_step": 251095, "epoch": 2821, "lr": 1.385939766162614e-06}
{"train_loss": 0.04104328900575638, "global_step": 251096, "epoch": 2821, "lr": 1.3858042185584785e-06}
{"train_loss": 0.0951673611998558, "global_step": 251097, "epoch": 2821, "lr": 1.3856686774899263e-06}
{"train_loss": 0.0523163303732872, "global_step": 251098, "epoch": 2821, "lr": 1.385533142956974e-06}
{"train_loss": 0.025102466344833374, "global_step": 251099, "epoch": 2821, "lr": 1.3853976149596493e-06}
{"train_loss": 0.04973869398236275, "global_step": 251100, "epoch": 2821, "lr": 1.3852620934979632e-06}
{"train_loss": 0.04736369103193283, "global_step": 251101, "epoch": 2821, "lr": 1.385126578571938e-06}
{"train_loss": 0.043405719101428986, "global_step": 251102, "epoch": 2821, "lr": 1.3849910701815849e-06}
{"train_loss": 0.053567834198474884, "global_step": 251103, "epoch": 2821, "lr": 1.3848555683269315e-06}
{"train_loss": 0.04615117236971855, "global_step": 251104, "epoch": 2821, "lr": 1.384720073007989e-06}
{"train_loss": 0.06254538148641586, "global_step": 251105, "epoch": 2821, "lr": 1.384584584224785e-06}
{"train_loss": 0.07021240144968033, "global_step": 251106, "epoch": 2821, "lr": 1.3844491019773254e-06}
{"train_loss": 0.035369738936424255, "global_step": 251107, "epoch": 2821, "lr": 1.3843136262656375e-06}
{"train_loss": 0.05215267464518547, "global_step": 251108, "epoch": 2821, "lr": 1.3841781570897271e-06}
{"train_loss": 0.03766225278377533, "global_step": 251109, "epoch": 2821, "lr": 1.3840426944496331e-06}
{"train_loss": 0.05047331005334854, "global_step": 251110, "epoch": 2821, "lr": 1.38390723834535e-06}
{"train_loss": 0.041895221918821335, "global_step": 251111, "epoch": 2821, "lr": 1.3837717887769163e-06}
{"train_loss": 0.025132734328508377, "global_step": 251112, "epoch": 2821, "lr": 1.3836363457443435e-06}
{"train_loss": 0.053075119853019714, "global_step": 251113, "epoch": 2821, "lr": 1.383500909247637e-06}
{"train_loss": 0.044961825013160706, "global_step": 251114, "epoch": 2821, "lr": 1.3833654792868411e-06}
{"train_loss": 0.03161833807826042, "global_step": 251115, "epoch": 2821, "lr": 1.3832300558619505e-06}
{"train_loss": 0.08586349338293076, "global_step": 251116, "epoch": 2821, "lr": 1.3830946389729926e-06}
{"train_loss": 0.06298360973596573, "global_step": 251117, "epoch": 2821, "lr": 1.3829592286199843e-06}
{"train_loss": 0.031474556773900986, "global_step": 251118, "epoch": 2821, "lr": 1.382823824802948e-06}
{"train_loss": 0.022202158346772194, "global_step": 251119, "epoch": 2821, "lr": 1.3826884275218942e-06}
{"train_loss": 0.06063004955649376, "global_step": 251120, "epoch": 2821, "lr": 1.3825530367768514e-06}
{"train_loss": 0.020499423146247864, "global_step": 251121, "epoch": 2821, "lr": 1.38241765256783e-06}
{"train_loss": 0.07622744143009186, "global_step": 251122, "epoch": 2821, "lr": 1.3822822748948473e-06}
{"train_loss": 0.030701983720064163, "global_step": 251123, "epoch": 2821, "lr": 1.382146903757925e-06}
{"train_loss": 0.04153824970126152, "global_step": 251124, "epoch": 2821, "lr": 1.3820115391570853e-06}
{"train_loss": 0.0315583162009716, "global_step": 251125, "epoch": 2821, "lr": 1.3818761810923342e-06}
{"train_loss": 0.031167883425951004, "global_step": 251126, "epoch": 2821, "lr": 1.3817408295636992e-06}
{"train_loss": 0.011027037166059017, "global_step": 251127, "epoch": 2821, "lr": 1.3816054845711967e-06}
{"train_loss": 0.047797977924346924, "global_step": 251128, "epoch": 2821, "lr": 1.3814701461148494e-06}
{"train_loss": 0.06979517638683319, "global_step": 251129, "epoch": 2821, "lr": 1.3813348141946625e-06}
{"train_loss": 0.03225821256637573, "global_step": 251130, "epoch": 2821, "lr": 1.3811994888106695e-06}
{"train_loss": 0.05562048405408859, "global_step": 251131, "epoch": 2821, "lr": 1.3810641699628812e-06}
{"train_loss": 0.04240166395902634, "global_step": 251132, "epoch": 2821, "lr": 1.3809288576513147e-06}
{"train_loss": 0.05086686462163925, "global_step": 251133, "epoch": 2821, "lr": 1.3807935518759862e-06}
{"train_loss": 0.06034361571073532, "global_step": 251134, "epoch": 2821, "lr": 1.3806582526369238e-06}
{"train_loss": 0.036651041358709335, "global_step": 251135, "epoch": 2821, "lr": 1.3805229599341385e-06}
{"train_loss": 0.04428233951330185, "global_step": 251136, "epoch": 2821, "lr": 1.380387673767647e-06}
{"train_loss": 0.03483118861913681, "global_step": 251137, "epoch": 2821, "lr": 1.3802523941374712e-06}
{"train_loss": 0.016358627006411552, "global_step": 251138, "epoch": 2821, "lr": 1.3801171210436225e-06}
{"train_loss": 0.04888356104493141, "global_step": 251139, "epoch": 2821, "lr": 1.3799818544861342e-06}
{"train_loss": 0.05694633722305298, "global_step": 251140, "epoch": 2821, "lr": 1.3798465944650062e-06}
{"train_loss": 0.0324259027838707, "global_step": 251141, "epoch": 2821, "lr": 1.3797113409802665e-06}
{"train_loss": 0.0633746087551117, "global_step": 251142, "epoch": 2821, "lr": 1.3795760940319314e-06}
{"train_loss": 0.04390670359134674, "global_step": 251143, "epoch": 2821, "lr": 1.3794408536200287e-06}
{"train_loss": 0.09441802650690079, "global_step": 251144, "epoch": 2821, "lr": 1.3793056197445587e-06}
{"train_loss": 0.0347566157579422, "global_step": 251145, "epoch": 2821, "lr": 1.3791703924055488e-06}
{"train_loss": 0.07777085900306702, "global_step": 251146, "epoch": 2821, "lr": 1.379035171603016e-06}
{"train_loss": 0.07066522538661957, "global_step": 251147, "epoch": 2821, "lr": 1.378899957336982e-06}
{"train_loss": 0.06045102700591087, "global_step": 251148, "epoch": 2821, "lr": 1.378764749607464e-06}
{"train_loss": 0.049454256892204285, "global_step": 251149, "epoch": 2821, "lr": 1.378629548414473e-06}
{"train_loss": 0.036369964480400085, "global_step": 251150, "epoch": 2821, "lr": 1.378494353758042e-06}
{"train_loss": 0.0499589666724205, "global_step": 251151, "epoch": 2821, "lr": 1.3783591656381711e-06}
{"train_loss": 0.07705813646316528, "global_step": 251152, "epoch": 2821, "lr": 1.3782239840548883e-06}
{"train_loss": 0.08103402704000473, "global_step": 251153, "epoch": 2821, "lr": 1.37808880900821e-06}
{"train_loss": 0.06174958869814873, "global_step": 251154, "epoch": 2821, "lr": 1.3779536404981642e-06}
{"train_loss": 0.0338832326233387, "global_step": 251155, "epoch": 2821, "lr": 1.3778184785247505e-06}
{"train_loss": 0.09046576172113419, "global_step": 251156, "epoch": 2821, "lr": 1.3776833230880026e-06}
{"train_loss": 0.045645079920800885, "global_step": 251157, "epoch": 2821, "lr": 1.377548174187926e-06, "val_loss": 9.556608200073242}
{"train_loss": 0.02637345902621746, "global_step": 251158, "epoch": 2822, "lr": 1.3774130318245481e-06}
{"train_loss": 0.06328489631414413, "global_step": 251159, "epoch": 2822, "lr": 1.377277895997886e-06}
{"train_loss": 0.043978821486234665, "global_step": 251160, "epoch": 2822, "lr": 1.377142766707956e-06}
{"train_loss": 0.01415653619915247, "global_step": 251161, "epoch": 2822, "lr": 1.3770076439547753e-06}
{"train_loss": 0.0453387089073658, "global_step": 251162, "epoch": 2822, "lr": 1.3768725277383655e-06}
{"train_loss": 0.05862941965460777, "global_step": 251163, "epoch": 2822, "lr": 1.3767374180587378e-06}
{"train_loss": 0.029043670743703842, "global_step": 251164, "epoch": 2822, "lr": 1.3766023149159202e-06}
{"train_loss": 0.06970782577991486, "global_step": 251165, "epoch": 2822, "lr": 1.376467218309918e-06}
{"train_loss": 0.02944912761449814, "global_step": 251166, "epoch": 2822, "lr": 1.3763321282407648e-06}
{"train_loss": 0.050489384680986404, "global_step": 251167, "epoch": 2822, "lr": 1.3761970447084716e-06}
{"train_loss": 0.013348420150578022, "global_step": 251168, "epoch": 2822, "lr": 1.3760619677130493e-06}
{"train_loss": 0.03526625037193298, "global_step": 251169, "epoch": 2822, "lr": 1.3759268972545314e-06}
{"train_loss": 0.066885806620121, "global_step": 251170, "epoch": 2822, "lr": 1.3757918333329235e-06}
{"train_loss": 0.039756111800670624, "global_step": 251171, "epoch": 2822, "lr": 1.3756567759482475e-06}
{"train_loss": 0.042198196053504944, "global_step": 251172, "epoch": 2822, "lr": 1.3755217251005203e-06}
{"train_loss": 0.02975904941558838, "global_step": 251173, "epoch": 2822, "lr": 1.3753866807897642e-06}
{"train_loss": 0.04326185956597328, "global_step": 251174, "epoch": 2822, "lr": 1.3752516430159956e-06}
{"train_loss": 0.005544270854443312, "global_step": 251175, "epoch": 2822, "lr": 1.3751166117792313e-06}
{"train_loss": 0.028260255232453346, "global_step": 251176, "epoch": 2822, "lr": 1.3749815870794824e-06}
{"train_loss": 0.04394208639860153, "global_step": 251177, "epoch": 2822, "lr": 1.3748465689167877e-06}
{"train_loss": 0.052192822098731995, "global_step": 251178, "epoch": 2822, "lr": 1.3747115572911418e-06}
{"train_loss": 0.045664504170417786, "global_step": 251179, "epoch": 2822, "lr": 1.3745765522025777e-06}
{"train_loss": 0.03485000878572464, "global_step": 251180, "epoch": 2822, "lr": 1.3744415536511068e-06}
{"train_loss": 0.08974616974592209, "global_step": 251181, "epoch": 2822, "lr": 1.3743065616367512e-06}
{"train_loss": 0.07163459807634354, "global_step": 251182, "epoch": 2822, "lr": 1.3741715761595276e-06}
{"train_loss": 0.054154764860868454, "global_step": 251183, "epoch": 2822, "lr": 1.3740365972194525e-06}
{"train_loss": 0.03156118094921112, "global_step": 251184, "epoch": 2822, "lr": 1.3739016248165426e-06}
{"train_loss": 0.04397409036755562, "global_step": 251185, "epoch": 2822, "lr": 1.3737666589508202e-06}
{"train_loss": 0.03400016948580742, "global_step": 251186, "epoch": 2822, "lr": 1.3736316996223075e-06}
{"train_loss": 0.03738895803689957, "global_step": 251187, "epoch": 2822, "lr": 1.3734967468310157e-06}
{"train_loss": 0.09518744051456451, "global_step": 251188, "epoch": 2822, "lr": 1.3733618005769611e-06}
{"train_loss": 0.0336758978664875, "global_step": 251189, "epoch": 2822, "lr": 1.3732268608601663e-06}
{"train_loss": 0.020068323239684105, "global_step": 251190, "epoch": 2822, "lr": 1.373091927680653e-06}
{"train_loss": 0.03464137762784958, "global_step": 251191, "epoch": 2822, "lr": 1.3729570010384274e-06}
{"train_loss": 0.06150451675057411, "global_step": 251192, "epoch": 2822, "lr": 1.3728220809335225e-06}
{"train_loss": 0.02417309395968914, "global_step": 251193, "epoch": 2822, "lr": 1.3726871673659436e-06}
{"train_loss": 0.060811176896095276, "global_step": 251194, "epoch": 2822, "lr": 1.3725522603357188e-06}
{"train_loss": 0.046631671488285065, "global_step": 251195, "epoch": 2822, "lr": 1.3724173598428591e-06}
{"train_loss": 0.029004449024796486, "global_step": 251196, "epoch": 2822, "lr": 1.3722824658873867e-06}
{"train_loss": 0.03442086651921272, "global_step": 251197, "epoch": 2822, "lr": 1.3721475784693128e-06}
{"train_loss": 0.03791987895965576, "global_step": 251198, "epoch": 2822, "lr": 1.3720126975886705e-06}
{"train_loss": 0.037440571933984756, "global_step": 251199, "epoch": 2822, "lr": 1.3718778232454598e-06}
{"train_loss": 0.02638227678835392, "global_step": 251200, "epoch": 2822, "lr": 1.3717429554397088e-06}
{"train_loss": 0.027792761102318764, "global_step": 251201, "epoch": 2822, "lr": 1.3716080941714338e-06}
{"train_loss": 0.03190963715314865, "global_step": 251202, "epoch": 2822, "lr": 1.3714732394406572e-06}
{"train_loss": 0.05957658961415291, "global_step": 251203, "epoch": 2822, "lr": 1.3713383912473899e-06}
{"train_loss": 0.05849895626306534, "global_step": 251204, "epoch": 2822, "lr": 1.3712035495916542e-06}
{"train_loss": 0.0569814033806324, "global_step": 251205, "epoch": 2822, "lr": 1.3710687144734725e-06}
{"train_loss": 0.04618271067738533, "global_step": 251206, "epoch": 2822, "lr": 1.37093388589285e-06}
{"train_loss": 0.014764571562409401, "global_step": 251207, "epoch": 2822, "lr": 1.3707990638498203e-06}
{"train_loss": 0.06005348265171051, "global_step": 251208, "epoch": 2822, "lr": 1.3706642483443942e-06}
{"train_loss": 0.02535475417971611, "global_step": 251209, "epoch": 2822, "lr": 1.3705294393765832e-06}
{"train_loss": 0.04725639149546623, "global_step": 251210, "epoch": 2822, "lr": 1.370394636946415e-06}
{"train_loss": 0.0752032920718193, "global_step": 251211, "epoch": 2822, "lr": 1.3702598410539002e-06}
{"train_loss": 0.025062233209609985, "global_step": 251212, "epoch": 2822, "lr": 1.370125051699067e-06}
{"train_loss": 0.08976467698812485, "global_step": 251213, "epoch": 2822, "lr": 1.3699902688819323e-06}
{"train_loss": 0.045388802886009216, "global_step": 251214, "epoch": 2822, "lr": 1.3698554926025008e-06}
{"train_loss": 0.06467161327600479, "global_step": 251215, "epoch": 2822, "lr": 1.369720722860801e-06}
{"train_loss": 0.03650887310504913, "global_step": 251216, "epoch": 2822, "lr": 1.3695859596568495e-06}
{"train_loss": 0.03562579303979874, "global_step": 251217, "epoch": 2822, "lr": 1.3694512029906681e-06}
{"train_loss": 0.02160714752972126, "global_step": 251218, "epoch": 2822, "lr": 1.3693164528622625e-06}
{"train_loss": 0.026496905833482742, "global_step": 251219, "epoch": 2822, "lr": 1.3691817092716718e-06}
{"train_loss": 0.06696057319641113, "global_step": 251220, "epoch": 2822, "lr": 1.3690469722188903e-06}
{"train_loss": 0.022622989490628242, "global_step": 251221, "epoch": 2822, "lr": 1.3689122417039513e-06}
{"train_loss": 0.030392559245228767, "global_step": 251222, "epoch": 2822, "lr": 1.3687775177268714e-06}
{"train_loss": 0.048596933484077454, "global_step": 251223, "epoch": 2822, "lr": 1.3686428002876672e-06}
{"train_loss": 0.055891651660203934, "global_step": 251224, "epoch": 2822, "lr": 1.3685080893863556e-06}
{"train_loss": 0.03852982446551323, "global_step": 251225, "epoch": 2822, "lr": 1.368373385022953e-06}
{"train_loss": 0.05734840780496597, "global_step": 251226, "epoch": 2822, "lr": 1.3682386871974873e-06}
{"train_loss": 0.03454960137605667, "global_step": 251227, "epoch": 2822, "lr": 1.3681039959099585e-06}
{"train_loss": 0.057342126965522766, "global_step": 251228, "epoch": 2822, "lr": 1.3679693111604052e-06}
{"train_loss": 0.06437618285417557, "global_step": 251229, "epoch": 2822, "lr": 1.3678346329488278e-06}
{"train_loss": 0.028652328997850418, "global_step": 251230, "epoch": 2822, "lr": 1.3676999612752595e-06}
{"train_loss": 0.012059200555086136, "global_step": 251231, "epoch": 2822, "lr": 1.3675652961397e-06}
{"train_loss": 0.02244170755147934, "global_step": 251232, "epoch": 2822, "lr": 1.3674306375421886e-06}
{"train_loss": 0.03147377446293831, "global_step": 251233, "epoch": 2822, "lr": 1.3672959854827306e-06}
{"train_loss": 0.05233870819211006, "global_step": 251234, "epoch": 2822, "lr": 1.3671613399613482e-06}
{"train_loss": 0.05611913651227951, "global_step": 251235, "epoch": 2822, "lr": 1.367026700978058e-06}
{"train_loss": 0.02491876855492592, "global_step": 251236, "epoch": 2822, "lr": 1.3668920685328767e-06}
{"train_loss": 0.05656062811613083, "global_step": 251237, "epoch": 2822, "lr": 1.3667574426258212e-06}
{"train_loss": 0.03704657033085823, "global_step": 251238, "epoch": 2822, "lr": 1.3666228232569134e-06}
{"train_loss": 0.032071802765131, "global_step": 251239, "epoch": 2822, "lr": 1.3664882104261755e-06}
{"train_loss": 0.10702096670866013, "global_step": 251240, "epoch": 2822, "lr": 1.3663536041336188e-06}
{"train_loss": 0.06625673174858093, "global_step": 251241, "epoch": 2822, "lr": 1.3662190043792656e-06}
{"train_loss": 0.04317754879593849, "global_step": 251242, "epoch": 2822, "lr": 1.366084411163121e-06}
{"train_loss": 0.029190238565206528, "global_step": 251243, "epoch": 2822, "lr": 1.3659498244852243e-06}
{"train_loss": 0.07187238335609436, "global_step": 251244, "epoch": 2822, "lr": 1.3658152443455752e-06}
{"train_loss": 0.07888083159923553, "global_step": 251245, "epoch": 2822, "lr": 1.365680670744207e-06}
{"train_loss": 0.044044392157262295, "global_step": 251246, "epoch": 2822, "lr": 1.3655461036811313e-06, "val_loss": 9.53370189666748}
{"train_loss": 0.012345188297331333, "global_step": 251247, "epoch": 2823, "lr": 1.3654115431563586e-06}
{"train_loss": 0.03847023844718933, "global_step": 251248, "epoch": 2823, "lr": 1.3652769891699169e-06}
{"train_loss": 0.031218450516462326, "global_step": 251249, "epoch": 2823, "lr": 1.3651424417218173e-06}
{"train_loss": 0.044934481382369995, "global_step": 251250, "epoch": 2823, "lr": 1.3650079008120875e-06}
{"train_loss": 0.07171902060508728, "global_step": 251251, "epoch": 2823, "lr": 1.3648733664407388e-06}
{"train_loss": 0.08835469186306, "global_step": 251252, "epoch": 2823, "lr": 1.364738838607782e-06}
{"train_loss": 0.057033464312553406, "global_step": 251253, "epoch": 2823, "lr": 1.3646043173132505e-06}
{"train_loss": 0.03579358011484146, "global_step": 251254, "epoch": 2823, "lr": 1.36446980255715e-06}
{"train_loss": 0.10120254009962082, "global_step": 251255, "epoch": 2823, "lr": 1.3643352943395138e-06}
{"train_loss": 0.029267091304063797, "global_step": 251256, "epoch": 2823, "lr": 1.3642007926603362e-06}
{"train_loss": 0.0528423972427845, "global_step": 251257, "epoch": 2823, "lr": 1.364066297519656e-06}
{"train_loss": 0.06338511407375336, "global_step": 251258, "epoch": 2823, "lr": 1.3639318089174847e-06}
{"train_loss": 0.03592844307422638, "global_step": 251259, "epoch": 2823, "lr": 1.3637973268538384e-06}
{"train_loss": 0.0272996686398983, "global_step": 251260, "epoch": 2823, "lr": 1.3636628513287396e-06}
{"train_loss": 0.03579717129468918, "global_step": 251261, "epoch": 2823, "lr": 1.3635283823421996e-06}
{"train_loss": 0.04834727942943573, "global_step": 251262, "epoch": 2823, "lr": 1.3633939198942457e-06}
{"train_loss": 0.03477751463651657, "global_step": 251263, "epoch": 2823, "lr": 1.363259463984884e-06}
{"train_loss": 0.05255290865898132, "global_step": 251264, "epoch": 2823, "lr": 1.3631250146141472e-06}
{"train_loss": 0.04547781124711037, "global_step": 251265, "epoch": 2823, "lr": 1.3629905717820358e-06}
{"train_loss": 0.0334697961807251, "global_step": 251266, "epoch": 2823, "lr": 1.3628561354885882e-06}
{"train_loss": 0.07777828723192215, "global_step": 251267, "epoch": 2823, "lr": 1.362721705733805e-06}
{"train_loss": 0.02632872946560383, "global_step": 251268, "epoch": 2823, "lr": 1.3625872825177132e-06}
{"train_loss": 0.04742088168859482, "global_step": 251269, "epoch": 2823, "lr": 1.3624528658403245e-06}
{"train_loss": 0.03904001787304878, "global_step": 251270, "epoch": 2823, "lr": 1.3623184557016666e-06}
{"train_loss": 0.028335534036159515, "global_step": 251271, "epoch": 2823, "lr": 1.3621840521017503e-06}
{"train_loss": 0.0630381628870964, "global_step": 251272, "epoch": 2823, "lr": 1.3620496550405981e-06}
{"train_loss": 0.07212855666875839, "global_step": 251273, "epoch": 2823, "lr": 1.3619152645182153e-06}
{"train_loss": 0.08029805123806, "global_step": 251274, "epoch": 2823, "lr": 1.3617808805346355e-06}
{"train_loss": 0.031438909471035004, "global_step": 251275, "epoch": 2823, "lr": 1.3616465030898806e-06}
{"train_loss": 0.03897611424326897, "global_step": 251276, "epoch": 2823, "lr": 1.3615121321839508e-06}
{"train_loss": 0.01650979556143284, "global_step": 251277, "epoch": 2823, "lr": 1.3613777678168738e-06}
{"train_loss": 0.03629448264837265, "global_step": 251278, "epoch": 2823, "lr": 1.3612434099886662e-06}
{"train_loss": 0.06100020557641983, "global_step": 251279, "epoch": 2823, "lr": 1.3611090586993502e-06}
{"train_loss": 0.03199581801891327, "global_step": 251280, "epoch": 2823, "lr": 1.360974713948937e-06}
{"train_loss": 0.07489873468875885, "global_step": 251281, "epoch": 2823, "lr": 1.3608403757374489e-06}
{"train_loss": 0.05221005156636238, "global_step": 251282, "epoch": 2823, "lr": 1.3607060440649023e-06}
{"train_loss": 0.020426161587238312, "global_step": 251283, "epoch": 2823, "lr": 1.3605717189313193e-06}
{"train_loss": 0.05743488296866417, "global_step": 251284, "epoch": 2823, "lr": 1.360437400336706e-06}
{"train_loss": 0.02129640430212021, "global_step": 251285, "epoch": 2823, "lr": 1.3603030882811007e-06}
{"train_loss": 0.07502221316099167, "global_step": 251286, "epoch": 2823, "lr": 1.3601687827645037e-06}
{"train_loss": 0.048109348863363266, "global_step": 251287, "epoch": 2823, "lr": 1.3600344837869372e-06}
{"train_loss": 0.04506068676710129, "global_step": 251288, "epoch": 2823, "lr": 1.3599001913484287e-06}
{"train_loss": 0.052166879177093506, "global_step": 251289, "epoch": 2823, "lr": 1.3597659054489842e-06}
{"train_loss": 0.033403366804122925, "global_step": 251290, "epoch": 2823, "lr": 1.3596316260886255e-06}
{"train_loss": 0.02345709688961506, "global_step": 251291, "epoch": 2823, "lr": 1.359497353267375e-06}
{"train_loss": 0.07885058224201202, "global_step": 251292, "epoch": 2823, "lr": 1.3593630869852435e-06}
{"train_loss": 0.04751456901431084, "global_step": 251293, "epoch": 2823, "lr": 1.3592288272422481e-06}
{"train_loss": 0.0061168246902525425, "global_step": 251294, "epoch": 2823, "lr": 1.359094574038422e-06}
{"train_loss": 0.028260760009288788, "global_step": 251295, "epoch": 2823, "lr": 1.358960327373765e-06}
{"train_loss": 0.034286294132471085, "global_step": 251296, "epoch": 2823, "lr": 1.358826087248305e-06}
{"train_loss": 0.07210459560155869, "global_step": 251297, "epoch": 2823, "lr": 1.3586918536620585e-06}
{"train_loss": 0.010928490199148655, "global_step": 251298, "epoch": 2823, "lr": 1.3585576266150423e-06}
{"train_loss": 0.037939995527267456, "global_step": 251299, "epoch": 2823, "lr": 1.358423406107273e-06}
{"train_loss": 0.047440335154533386, "global_step": 251300, "epoch": 2823, "lr": 1.3582891921387787e-06}
{"train_loss": 0.08486353605985641, "global_step": 251301, "epoch": 2823, "lr": 1.3581549847095643e-06}
{"train_loss": 0.016218796372413635, "global_step": 251302, "epoch": 2823, "lr": 1.3580207838196525e-06}
{"train_loss": 0.08157753199338913, "global_step": 251303, "epoch": 2823, "lr": 1.3578865894690596e-06}
{"train_loss": 0.057154763489961624, "global_step": 251304, "epoch": 2823, "lr": 1.3577524016578136e-06}
{"train_loss": 0.02942177839577198, "global_step": 251305, "epoch": 2823, "lr": 1.35761822038592e-06}
{"train_loss": 0.03205503895878792, "global_step": 251306, "epoch": 2823, "lr": 1.357484045653401e-06}
{"train_loss": 0.07573797553777695, "global_step": 251307, "epoch": 2823, "lr": 1.3573498774602734e-06}
{"train_loss": 0.03118916042149067, "global_step": 251308, "epoch": 2823, "lr": 1.3572157158065645e-06}
{"train_loss": 0.023434769362211227, "global_step": 251309, "epoch": 2823, "lr": 1.3570815606922804e-06}
{"train_loss": 0.05053019896149635, "global_step": 251310, "epoch": 2823, "lr": 1.3569474121174487e-06}
{"train_loss": 0.07268183678388596, "global_step": 251311, "epoch": 2823, "lr": 1.3568132700820747e-06}
{"train_loss": 0.024280596524477005, "global_step": 251312, "epoch": 2823, "lr": 1.3566791345861862e-06}
{"train_loss": 0.030422652140259743, "global_step": 251313, "epoch": 2823, "lr": 1.3565450056298e-06}
{"train_loss": 0.03551309183239937, "global_step": 251314, "epoch": 2823, "lr": 1.3564108832129329e-06}
{"train_loss": 0.052248723804950714, "global_step": 251315, "epoch": 2823, "lr": 1.3562767673356069e-06}
{"train_loss": 0.038502439856529236, "global_step": 251316, "epoch": 2823, "lr": 1.356142657997833e-06}
{"train_loss": 0.018200527876615524, "global_step": 251317, "epoch": 2823, "lr": 1.3560085551996393e-06}
{"train_loss": 0.06474693864583969, "global_step": 251318, "epoch": 2823, "lr": 1.355874458941031e-06}
{"train_loss": 0.03021315298974514, "global_step": 251319, "epoch": 2823, "lr": 1.3557403692220361e-06}
{"train_loss": 0.03860870376229286, "global_step": 251320, "epoch": 2823, "lr": 1.3556062860426654e-06}
{"train_loss": 0.04951148107647896, "global_step": 251321, "epoch": 2823, "lr": 1.3554722094029415e-06}
{"train_loss": 0.06751617044210434, "global_step": 251322, "epoch": 2823, "lr": 1.3553381393028807e-06}
{"train_loss": 0.024356095120310783, "global_step": 251323, "epoch": 2823, "lr": 1.3552040757425055e-06}
{"train_loss": 0.033663202077150345, "global_step": 251324, "epoch": 2823, "lr": 1.3550700187218324e-06}
{"train_loss": 0.01739225909113884, "global_step": 251325, "epoch": 2823, "lr": 1.354935968240867e-06}
{"train_loss": 0.028576413169503212, "global_step": 251326, "epoch": 2823, "lr": 1.354801924299648e-06}
{"train_loss": 0.03921247646212578, "global_step": 251327, "epoch": 2823, "lr": 1.3546678868981754e-06}
{"train_loss": 0.02415527030825615, "global_step": 251328, "epoch": 2823, "lr": 1.3545338560364829e-06}
{"train_loss": 0.023514628410339355, "global_step": 251329, "epoch": 2823, "lr": 1.3543998317145702e-06}
{"train_loss": 0.02106633223593235, "global_step": 251330, "epoch": 2823, "lr": 1.354265813932476e-06}
{"train_loss": 0.01732861064374447, "global_step": 251331, "epoch": 2823, "lr": 1.354131802690195e-06}
{"train_loss": 0.04652822017669678, "global_step": 251332, "epoch": 2823, "lr": 1.3539977979877715e-06}
{"train_loss": 0.023881815373897552, "global_step": 251333, "epoch": 2823, "lr": 1.3538637998252002e-06}
{"train_loss": 0.041226789355278015, "global_step": 251334, "epoch": 2823, "lr": 1.3537298082025141e-06}
{"train_loss": 0.04312947595994292, "global_step": 251335, "epoch": 2823, "lr": 1.3535958231197243e-06, "val_loss": 9.619643211364746}
{"train_loss": 0.010926314629614353, "global_step": 251336, "epoch": 2824, "lr": 1.3534618445768533e-06}
{"train_loss": 0.04487593099474907, "global_step": 251337, "epoch": 2824, "lr": 1.3533278725739117e-06}
{"train_loss": 0.04001687467098236, "global_step": 251338, "epoch": 2824, "lr": 1.353193907110928e-06}
{"train_loss": 0.008902646601200104, "global_step": 251339, "epoch": 2824, "lr": 1.3530599481879068e-06}
{"train_loss": 0.04552949219942093, "global_step": 251340, "epoch": 2824, "lr": 1.3529259958048824e-06}
{"train_loss": 0.11310505867004395, "global_step": 251341, "epoch": 2824, "lr": 1.3527920499618595e-06}
{"train_loss": 0.10227982699871063, "global_step": 251342, "epoch": 2824, "lr": 1.352658110658861e-06}
{"train_loss": 0.027451323345303535, "global_step": 251343, "epoch": 2824, "lr": 1.352524177895903e-06}
{"train_loss": 0.06733270734548569, "global_step": 251344, "epoch": 2824, "lr": 1.3523902516730135e-06}
{"train_loss": 0.039378371089696884, "global_step": 251345, "epoch": 2824, "lr": 1.3522563319901926e-06}
{"train_loss": 0.061285361647605896, "global_step": 251346, "epoch": 2824, "lr": 1.3521224188474736e-06}
{"train_loss": 0.0052935476414859295, "global_step": 251347, "epoch": 2824, "lr": 1.3519885122448617e-06}
{"train_loss": 0.050663553178310394, "global_step": 251348, "epoch": 2824, "lr": 1.351854612182385e-06}
{"train_loss": 0.027488678693771362, "global_step": 251349, "epoch": 2824, "lr": 1.3517207186600656e-06}
{"train_loss": 0.07845748960971832, "global_step": 251350, "epoch": 2824, "lr": 1.3515868316779034e-06}
{"train_loss": 0.019520491361618042, "global_step": 251351, "epoch": 2824, "lr": 1.3514529512359374e-06}
{"train_loss": 0.044795867055654526, "global_step": 251352, "epoch": 2824, "lr": 1.3513190773341678e-06}
{"train_loss": 0.022539474070072174, "global_step": 251353, "epoch": 2824, "lr": 1.3511852099726275e-06}
{"train_loss": 0.03198868781328201, "global_step": 251354, "epoch": 2824, "lr": 1.3510513491513221e-06}
{"train_loss": 0.017316080629825592, "global_step": 251355, "epoch": 2824, "lr": 1.3509174948702795e-06}
{"train_loss": 0.0863332599401474, "global_step": 251356, "epoch": 2824, "lr": 1.3507836471295055e-06}
{"train_loss": 0.05952359735965729, "global_step": 251357, "epoch": 2824, "lr": 1.350649805929033e-06}
{"train_loss": 0.028060907498002052, "global_step": 251358, "epoch": 2824, "lr": 1.3505159712688676e-06}
{"train_loss": 0.02151649259030819, "global_step": 251359, "epoch": 2824, "lr": 1.350382143149037e-06}
{"train_loss": 0.04689051955938339, "global_step": 251360, "epoch": 2824, "lr": 1.3502483215695526e-06}
{"train_loss": 0.024354901164770126, "global_step": 251361, "epoch": 2824, "lr": 1.3501145065304365e-06}
{"train_loss": 0.019029946997761726, "global_step": 251362, "epoch": 2824, "lr": 1.3499806980317054e-06}
{"train_loss": 0.04714900627732277, "global_step": 251363, "epoch": 2824, "lr": 1.34984689607337e-06}
{"train_loss": 0.04464573785662651, "global_step": 251364, "epoch": 2824, "lr": 1.3497131006554587e-06}
{"train_loss": 0.04029778391122818, "global_step": 251365, "epoch": 2824, "lr": 1.349579311777982e-06}
{"train_loss": 0.05516623333096504, "global_step": 251366, "epoch": 2824, "lr": 1.349445529440968e-06}
{"train_loss": 0.06055489927530289, "global_step": 251367, "epoch": 2824, "lr": 1.3493117536444221e-06}
{"train_loss": 0.008445135317742825, "global_step": 251368, "epoch": 2824, "lr": 1.3491779843883723e-06}
{"train_loss": 0.05553016811609268, "global_step": 251369, "epoch": 2824, "lr": 1.3490442216728293e-06}
{"train_loss": 0.018156008794903755, "global_step": 251370, "epoch": 2824, "lr": 1.3489104654978212e-06}
{"train_loss": 0.04379379749298096, "global_step": 251371, "epoch": 2824, "lr": 1.3487767158633479e-06}
{"train_loss": 0.04506336525082588, "global_step": 251372, "epoch": 2824, "lr": 1.3486429727694482e-06}
{"train_loss": 0.02773819863796234, "global_step": 251373, "epoch": 2824, "lr": 1.3485092362161223e-06}
{"train_loss": 0.044214844703674316, "global_step": 251374, "epoch": 2824, "lr": 1.348375506203403e-06}
{"train_loss": 0.025132019072771072, "global_step": 251375, "epoch": 2824, "lr": 1.3482417827312965e-06}
{"train_loss": 0.047318100929260254, "global_step": 251376, "epoch": 2824, "lr": 1.3481080657998356e-06}
{"train_loss": 0.018013423308730125, "global_step": 251377, "epoch": 2824, "lr": 1.3479743554090152e-06}
{"train_loss": 0.0188059713691473, "global_step": 251378, "epoch": 2824, "lr": 1.3478406515588793e-06}
{"train_loss": 0.01223320048302412, "global_step": 251379, "epoch": 2824, "lr": 1.3477069542494225e-06}
{"train_loss": 0.10495062172412872, "global_step": 251380, "epoch": 2824, "lr": 1.3475732634806837e-06}
{"train_loss": 0.03337325528264046, "global_step": 251381, "epoch": 2824, "lr": 1.3474395792526628e-06}
{"train_loss": 0.024047933518886566, "global_step": 251382, "epoch": 2824, "lr": 1.3473059015653876e-06}
{"train_loss": 0.02094384841620922, "global_step": 251383, "epoch": 2824, "lr": 1.347172230418875e-06}
{"train_loss": 0.09692548215389252, "global_step": 251384, "epoch": 2824, "lr": 1.3470385658131358e-06}
{"train_loss": 0.07379455864429474, "global_step": 251385, "epoch": 2824, "lr": 1.3469049077482031e-06}
{"train_loss": 0.06413695216178894, "global_step": 251386, "epoch": 2824, "lr": 1.346771256224083e-06}
{"train_loss": 0.08684882521629333, "global_step": 251387, "epoch": 2824, "lr": 1.3466376112407974e-06}
{"train_loss": 0.03731217607855797, "global_step": 251388, "epoch": 2824, "lr": 1.3465039727983575e-06}
{"train_loss": 0.02477281540632248, "global_step": 251389, "epoch": 2824, "lr": 1.3463703408967965e-06}
{"train_loss": 0.057018790394067764, "global_step": 251390, "epoch": 2824, "lr": 1.3462367155361145e-06}
{"train_loss": 0.034372881054878235, "global_step": 251391, "epoch": 2824, "lr": 1.3461030967163446e-06}
{"train_loss": 0.04424367472529411, "global_step": 251392, "epoch": 2824, "lr": 1.3459694844374927e-06}
{"train_loss": 0.044715482741594315, "global_step": 251393, "epoch": 2824, "lr": 1.3458358786995916e-06}
{"train_loss": 0.016433363780379295, "global_step": 251394, "epoch": 2824, "lr": 1.3457022795026418e-06}
{"train_loss": 0.03855420649051666, "global_step": 251395, "epoch": 2824, "lr": 1.345568686846671e-06}
{"train_loss": 0.01285515259951353, "global_step": 251396, "epoch": 2824, "lr": 1.34543510073169e-06}
{"train_loss": 0.024734871461987495, "global_step": 251397, "epoch": 2824, "lr": 1.3453015211577325e-06}
{"train_loss": 0.02904052846133709, "global_step": 251398, "epoch": 2824, "lr": 1.345167948124798e-06}
{"train_loss": 0.03980419412255287, "global_step": 251399, "epoch": 2824, "lr": 1.3450343816329203e-06}
{"train_loss": 0.0434279590845108, "global_step": 251400, "epoch": 2824, "lr": 1.3449008216821046e-06}
{"train_loss": 0.04641222953796387, "global_step": 251401, "epoch": 2824, "lr": 1.3447672682723733e-06}
{"train_loss": 0.02732757478952408, "global_step": 251402, "epoch": 2824, "lr": 1.3446337214037485e-06}
{"train_loss": 0.05223296582698822, "global_step": 251403, "epoch": 2824, "lr": 1.3445001810762414e-06}
{"train_loss": 0.0336872860789299, "global_step": 251404, "epoch": 2824, "lr": 1.3443666472898797e-06}
{"train_loss": 0.08170153200626373, "global_step": 251405, "epoch": 2824, "lr": 1.3442331200446635e-06}
{"train_loss": 0.03614233061671257, "global_step": 251406, "epoch": 2824, "lr": 1.3440995993406314e-06}
{"train_loss": 0.044292762875556946, "global_step": 251407, "epoch": 2824, "lr": 1.3439660851777891e-06}
{"train_loss": 0.048143669962882996, "global_step": 251408, "epoch": 2824, "lr": 1.343832577556159e-06}
{"train_loss": 0.02855903096497059, "global_step": 251409, "epoch": 2824, "lr": 1.3436990764757517e-06}
{"train_loss": 0.037391625344753265, "global_step": 251410, "epoch": 2824, "lr": 1.343565581936601e-06}
{"train_loss": 0.04177508130669594, "global_step": 251411, "epoch": 2824, "lr": 1.3434320939387069e-06}
{"train_loss": 0.027708273380994797, "global_step": 251412, "epoch": 2824, "lr": 1.3432986124820968e-06}
{"train_loss": 0.04463673010468483, "global_step": 251413, "epoch": 2824, "lr": 1.3431651375667876e-06}
{"train_loss": 0.03313755989074707, "global_step": 251414, "epoch": 2824, "lr": 1.3430316691928013e-06}
{"train_loss": 0.04452834650874138, "global_step": 251415, "epoch": 2824, "lr": 1.3428982073601436e-06}
{"train_loss": 0.04428046941757202, "global_step": 251416, "epoch": 2824, "lr": 1.3427647520688425e-06}
{"train_loss": 0.04728000611066818, "global_step": 251417, "epoch": 2824, "lr": 1.3426313033189143e-06}
{"train_loss": 0.04252735525369644, "global_step": 251418, "epoch": 2824, "lr": 1.3424978611103811e-06}
{"train_loss": 0.05060829222202301, "global_step": 251419, "epoch": 2824, "lr": 1.3423644254432488e-06}
{"train_loss": 0.01618010364472866, "global_step": 251420, "epoch": 2824, "lr": 1.3422309963175396e-06}
{"train_loss": 0.027912579476833344, "global_step": 251421, "epoch": 2824, "lr": 1.3420975737332864e-06}
{"train_loss": 0.0142339076846838, "global_step": 251422, "epoch": 2824, "lr": 1.341964157690484e-06}
{"train_loss": 0.05759909376502037, "global_step": 251423, "epoch": 2824, "lr": 1.3418307481891712e-06}
{"train_loss": 0.041351695508392676, "global_step": 251424, "epoch": 2824, "lr": 1.341697345229348e-06, "val_loss": 9.587255477905273}
{"train_loss": 0.02013043314218521, "global_step": 251425, "epoch": 2825, "lr": 1.3415639488110477e-06}
{"train_loss": 0.08340323716402054, "global_step": 251426, "epoch": 2825, "lr": 1.3414305589342757e-06}
{"train_loss": 0.04852898046374321, "global_step": 251427, "epoch": 2825, "lr": 1.34129717559906e-06}
{"train_loss": 0.07529948651790619, "global_step": 251428, "epoch": 2825, "lr": 1.341163798805406e-06}
{"train_loss": 0.08529266715049744, "global_step": 251429, "epoch": 2825, "lr": 1.3410304285533471e-06}
{"train_loss": 0.08715083450078964, "global_step": 251430, "epoch": 2825, "lr": 1.3408970648428887e-06}
{"train_loss": 0.023054635152220726, "global_step": 251431, "epoch": 2825, "lr": 1.3407637076740588e-06}
{"train_loss": 0.015705185011029243, "global_step": 251432, "epoch": 2825, "lr": 1.3406303570468626e-06}
{"train_loss": 0.05290065333247185, "global_step": 251433, "epoch": 2825, "lr": 1.3404970129613338e-06}
{"train_loss": 0.03277158364653587, "global_step": 251434, "epoch": 2825, "lr": 1.340363675417472e-06}
{"train_loss": 0.043457936495542526, "global_step": 251435, "epoch": 2825, "lr": 1.3402303444153164e-06}
{"train_loss": 0.028327738866209984, "global_step": 251436, "epoch": 2825, "lr": 1.3400970199548668e-06}
{"train_loss": 0.014098226092755795, "global_step": 251437, "epoch": 2825, "lr": 1.339963702036151e-06}
{"train_loss": 0.028224555775523186, "global_step": 251438, "epoch": 2825, "lr": 1.3398303906591803e-06}
{"train_loss": 0.01682726852595806, "global_step": 251439, "epoch": 2825, "lr": 1.3396970858239822e-06}
{"train_loss": 0.03638298064470291, "global_step": 251440, "epoch": 2825, "lr": 1.3395637875305678e-06}
{"train_loss": 0.04685499519109726, "global_step": 251441, "epoch": 2825, "lr": 1.3394304957789538e-06}
{"train_loss": 0.05097782611846924, "global_step": 251442, "epoch": 2825, "lr": 1.3392972105691626e-06}
{"train_loss": 0.02380984276533127, "global_step": 251443, "epoch": 2825, "lr": 1.339163931901205e-06}
{"train_loss": 0.012499970383942127, "global_step": 251444, "epoch": 2825, "lr": 1.339030659775109e-06}
{"train_loss": 0.07409953325986862, "global_step": 251445, "epoch": 2825, "lr": 1.33889739419088e-06}
{"train_loss": 0.04686180129647255, "global_step": 251446, "epoch": 2825, "lr": 1.3387641351485514e-06}
{"train_loss": 0.05114518478512764, "global_step": 251447, "epoch": 2825, "lr": 1.3386308826481286e-06}
{"train_loss": 0.061212558299303055, "global_step": 251448, "epoch": 2825, "lr": 1.338497636689634e-06}
{"train_loss": 0.0404747910797596, "global_step": 251449, "epoch": 2825, "lr": 1.3383643972730842e-06}
{"train_loss": 0.024852197617292404, "global_step": 251450, "epoch": 2825, "lr": 1.3382311643985012e-06}
{"train_loss": 0.061415888369083405, "global_step": 251451, "epoch": 2825, "lr": 1.3380979380658965e-06}
{"train_loss": 0.06166646629571915, "global_step": 251452, "epoch": 2825, "lr": 1.3379647182752974e-06}
{"train_loss": 0.046906016767024994, "global_step": 251453, "epoch": 2825, "lr": 1.3378315050267099e-06}
{"train_loss": 0.07316743582487106, "global_step": 251454, "epoch": 2825, "lr": 1.3376982983201614e-06}
{"train_loss": 0.034579385071992874, "global_step": 251455, "epoch": 2825, "lr": 1.3375650981556575e-06}
{"train_loss": 0.05356915295124054, "global_step": 251456, "epoch": 2825, "lr": 1.3374319045332318e-06}
{"train_loss": 0.05365600809454918, "global_step": 251457, "epoch": 2825, "lr": 1.3372987174529007e-06}
{"train_loss": 0.0332278273999691, "global_step": 251458, "epoch": 2825, "lr": 1.337165536914664e-06}
{"train_loss": 0.017506545409560204, "global_step": 251459, "epoch": 2825, "lr": 1.3370323629185665e-06}
{"train_loss": 0.048375796526670456, "global_step": 251460, "epoch": 2825, "lr": 1.3368991954645972e-06}
{"train_loss": 0.03773906081914902, "global_step": 251461, "epoch": 2825, "lr": 1.3367660345527998e-06}
{"train_loss": 0.040443047881126404, "global_step": 251462, "epoch": 2825, "lr": 1.3366328801831752e-06}
{"train_loss": 0.050697822123765945, "global_step": 251463, "epoch": 2825, "lr": 1.336499732355756e-06}
{"train_loss": 0.00947950966656208, "global_step": 251464, "epoch": 2825, "lr": 1.3363665910705425e-06}
{"train_loss": 0.026525916531682014, "global_step": 251465, "epoch": 2825, "lr": 1.3362334563275625e-06}
{"train_loss": 0.030528049916028976, "global_step": 251466, "epoch": 2825, "lr": 1.3361003281268325e-06}
{"train_loss": 0.008202812634408474, "global_step": 251467, "epoch": 2825, "lr": 1.3359672064683748e-06}
{"train_loss": 0.04822758585214615, "global_step": 251468, "epoch": 2825, "lr": 1.3358340913522004e-06}
{"train_loss": 0.043363723903894424, "global_step": 251469, "epoch": 2825, "lr": 1.3357009827783317e-06}
{"train_loss": 0.05435478687286377, "global_step": 251470, "epoch": 2825, "lr": 1.3355678807467797e-06}
{"train_loss": 0.024996081367135048, "global_step": 251471, "epoch": 2825, "lr": 1.3354347852575722e-06}
{"train_loss": 0.03608929365873337, "global_step": 251472, "epoch": 2825, "lr": 1.33530169631072e-06}
{"train_loss": 0.08000006526708603, "global_step": 251473, "epoch": 2825, "lr": 1.3351686139062457e-06}
{"train_loss": 0.043874967843294144, "global_step": 251474, "epoch": 2825, "lr": 1.3350355380441659e-06}
{"train_loss": 0.07584278285503387, "global_step": 251475, "epoch": 2825, "lr": 1.3349024687244971e-06}
{"train_loss": 0.04292478784918785, "global_step": 251476, "epoch": 2825, "lr": 1.3347694059472505e-06}
{"train_loss": 0.03213879093527794, "global_step": 251477, "epoch": 2825, "lr": 1.3346363497124592e-06}
{"train_loss": 0.046586278825998306, "global_step": 251478, "epoch": 2825, "lr": 1.3345033000201346e-06}
{"train_loss": 0.03256681561470032, "global_step": 251479, "epoch": 2825, "lr": 1.3343702568702876e-06}
{"train_loss": 0.02835727669298649, "global_step": 251480, "epoch": 2825, "lr": 1.3342372202629406e-06}
{"train_loss": 0.04956929013133049, "global_step": 251481, "epoch": 2825, "lr": 1.33410419019811e-06}
{"train_loss": 0.053572915494441986, "global_step": 251482, "epoch": 2825, "lr": 1.3339711666758237e-06}
{"train_loss": 0.07148841768503189, "global_step": 251483, "epoch": 2825, "lr": 1.3338381496960873e-06}
{"train_loss": 0.04613764211535454, "global_step": 251484, "epoch": 2825, "lr": 1.3337051392589229e-06}
{"train_loss": 0.042236436158418655, "global_step": 251485, "epoch": 2825, "lr": 1.333572135364347e-06}
{"train_loss": 0.05004549399018288, "global_step": 251486, "epoch": 2825, "lr": 1.3334391380123878e-06}
{"train_loss": 0.035359375178813934, "global_step": 251487, "epoch": 2825, "lr": 1.333306147203045e-06}
{"train_loss": 0.033131036907434464, "global_step": 251488, "epoch": 2825, "lr": 1.333173162936352e-06}
{"train_loss": 0.030069099739193916, "global_step": 251489, "epoch": 2825, "lr": 1.3330401852123143e-06}
{"train_loss": 0.04721102863550186, "global_step": 251490, "epoch": 2825, "lr": 1.332907214030965e-06}
{"train_loss": 0.050510842353105545, "global_step": 251491, "epoch": 2825, "lr": 1.3327742493923046e-06}
{"train_loss": 0.029549865052103996, "global_step": 251492, "epoch": 2825, "lr": 1.3326412912963604e-06}
{"train_loss": 0.045878708362579346, "global_step": 251493, "epoch": 2825, "lr": 1.3325083397431548e-06}
{"train_loss": 0.03542306274175644, "global_step": 251494, "epoch": 2825, "lr": 1.3323753947326933e-06}
{"train_loss": 0.00883023627102375, "global_step": 251495, "epoch": 2825, "lr": 1.3322424562650093e-06}
{"train_loss": 0.06647881865501404, "global_step": 251496, "epoch": 2825, "lr": 1.3321095243401027e-06}
{"train_loss": 0.024330703541636467, "global_step": 251497, "epoch": 2825, "lr": 1.331976598958007e-06}
{"train_loss": 0.034395549446344376, "global_step": 251498, "epoch": 2825, "lr": 1.3318436801187328e-06}
{"train_loss": 0.08296403288841248, "global_step": 251499, "epoch": 2825, "lr": 1.3317107678223028e-06}
{"train_loss": 0.015409065410494804, "global_step": 251500, "epoch": 2825, "lr": 1.3315778620687226e-06}
{"train_loss": 0.04590805247426033, "global_step": 251501, "epoch": 2825, "lr": 1.331444962858025e-06}
{"train_loss": 0.04811624810099602, "global_step": 251502, "epoch": 2825, "lr": 1.3313120701902159e-06}
{"train_loss": 0.007928429171442986, "global_step": 251503, "epoch": 2825, "lr": 1.3311791840653286e-06}
{"train_loss": 0.030717894434928894, "global_step": 251504, "epoch": 2825, "lr": 1.3310463044833633e-06}
{"train_loss": 0.03119068220257759, "global_step": 251505, "epoch": 2825, "lr": 1.3309134314443473e-06}
{"train_loss": 0.0260991919785738, "global_step": 251506, "epoch": 2825, "lr": 1.3307805649482975e-06}
{"train_loss": 0.04620261490345001, "global_step": 251507, "epoch": 2825, "lr": 1.3306477049952303e-06}
{"train_loss": 0.028658552095294, "global_step": 251508, "epoch": 2825, "lr": 1.330514851585163e-06}
{"train_loss": 0.016612017527222633, "global_step": 251509, "epoch": 2825, "lr": 1.330382004718117e-06}
{"train_loss": 0.06026669219136238, "global_step": 251510, "epoch": 2825, "lr": 1.330249164394104e-06}
{"train_loss": 0.019975902512669563, "global_step": 251511, "epoch": 2825, "lr": 1.330116330613146e-06}
{"train_loss": 0.04893668740987778, "global_step": 251512, "epoch": 2825, "lr": 1.3299835033752705e-06}
{"train_loss": 0.04131905104588257, "global_step": 251513, "epoch": 2825, "lr": 1.3298506826804779e-06, "val_loss": 9.481674194335938, "train_action_mse_error": 7.059462070465088}
{"train_loss": 0.03830522298812866, "global_step": 251514, "epoch": 2826, "lr": 1.3297178685287904e-06}
{"train_loss": 0.046555012464523315, "global_step": 251515, "epoch": 2826, "lr": 1.3295850609202354e-06}
{"train_loss": 0.029635446146130562, "global_step": 251516, "epoch": 2826, "lr": 1.3294522598548242e-06}
{"train_loss": 0.05146323889493942, "global_step": 251517, "epoch": 2826, "lr": 1.3293194653325736e-06}
{"train_loss": 0.029369693249464035, "global_step": 251518, "epoch": 2826, "lr": 1.3291866773535056e-06}
{"train_loss": 0.032454170286655426, "global_step": 251519, "epoch": 2826, "lr": 1.3290538959176258e-06}
{"train_loss": 0.07080383598804474, "global_step": 251520, "epoch": 2826, "lr": 1.328921121024973e-06}
{"train_loss": 0.06008952111005783, "global_step": 251521, "epoch": 2826, "lr": 1.3287883526755418e-06}
{"train_loss": 0.04070455580949783, "global_step": 251522, "epoch": 2826, "lr": 1.328655590869371e-06}
{"train_loss": 0.03637678548693657, "global_step": 251523, "epoch": 2826, "lr": 1.328522835606466e-06}
{"train_loss": 0.05243799090385437, "global_step": 251524, "epoch": 2826, "lr": 1.3283900868868493e-06}
{"train_loss": 0.03982017934322357, "global_step": 251525, "epoch": 2826, "lr": 1.3282573447105318e-06}
{"train_loss": 0.03833598271012306, "global_step": 251526, "epoch": 2826, "lr": 1.3281246090775468e-06}
{"train_loss": 0.08243784308433533, "global_step": 251527, "epoch": 2826, "lr": 1.3279918799878942e-06}
{"train_loss": 0.06749412417411804, "global_step": 251528, "epoch": 2826, "lr": 1.3278591574415966e-06}
{"train_loss": 0.06229011341929436, "global_step": 251529, "epoch": 2826, "lr": 1.327726441438687e-06}
{"train_loss": 0.027031077072024345, "global_step": 251530, "epoch": 2826, "lr": 1.3275937319791597e-06}
{"train_loss": 0.030814286321401596, "global_step": 251531, "epoch": 2826, "lr": 1.327461029063054e-06}
{"train_loss": 0.0745200589299202, "global_step": 251532, "epoch": 2826, "lr": 1.3273283326903696e-06}
{"train_loss": 0.018056387081742287, "global_step": 251533, "epoch": 2826, "lr": 1.32719564286114e-06}
{"train_loss": 0.04343589395284653, "global_step": 251534, "epoch": 2826, "lr": 1.3270629595753704e-06}
{"train_loss": 0.025111418217420578, "global_step": 251535, "epoch": 2826, "lr": 1.3269302828330888e-06}
{"train_loss": 0.021552519872784615, "global_step": 251536, "epoch": 2826, "lr": 1.3267976126343006e-06}
{"train_loss": 0.0658411979675293, "global_step": 251537, "epoch": 2826, "lr": 1.3266649489790395e-06}
{"train_loss": 0.020379161462187767, "global_step": 251538, "epoch": 2826, "lr": 1.3265322918673106e-06}
{"train_loss": 0.07431282848119736, "global_step": 251539, "epoch": 2826, "lr": 1.3263996412991364e-06}
{"train_loss": 0.057055629789829254, "global_step": 251540, "epoch": 2826, "lr": 1.3262669972745335e-06}
{"train_loss": 0.07717984169721603, "global_step": 251541, "epoch": 2826, "lr": 1.3261343597935239e-06}
{"train_loss": 0.057924091815948486, "global_step": 251542, "epoch": 2826, "lr": 1.3260017288561189e-06}
{"train_loss": 0.05584188550710678, "global_step": 251543, "epoch": 2826, "lr": 1.3258691044623462e-06}
{"train_loss": 0.01735304668545723, "global_step": 251544, "epoch": 2826, "lr": 1.3257364866122056e-06}
{"train_loss": 0.050766997039318085, "global_step": 251545, "epoch": 2826, "lr": 1.3256038753057366e-06}
{"train_loss": 0.025098370388150215, "global_step": 251546, "epoch": 2826, "lr": 1.325471270542944e-06}
{"train_loss": 0.04388020187616348, "global_step": 251547, "epoch": 2826, "lr": 1.3253386723238448e-06}
{"train_loss": 0.06020069867372513, "global_step": 251548, "epoch": 2826, "lr": 1.3252060806484667e-06}
{"train_loss": 0.060538746416568756, "global_step": 251549, "epoch": 2826, "lr": 1.3250734955168154e-06}
{"train_loss": 0.1078350618481636, "global_step": 251550, "epoch": 2826, "lr": 1.3249409169289184e-06}
{"train_loss": 0.05835406854748726, "global_step": 251551, "epoch": 2826, "lr": 1.3248083448847926e-06}
{"train_loss": 0.047287262976169586, "global_step": 251552, "epoch": 2826, "lr": 1.3246757793844489e-06}
{"train_loss": 0.04302551597356796, "global_step": 251553, "epoch": 2826, "lr": 1.3245432204279152e-06}
{"train_loss": 0.06741970777511597, "global_step": 251554, "epoch": 2826, "lr": 1.3244106680151912e-06}
{"train_loss": 0.027828499674797058, "global_step": 251555, "epoch": 2826, "lr": 1.3242781221463163e-06}
{"train_loss": 0.0878763496875763, "global_step": 251556, "epoch": 2826, "lr": 1.3241455828213013e-06}
{"train_loss": 0.05263354629278183, "global_step": 251557, "epoch": 2826, "lr": 1.324013050040157e-06}
{"train_loss": 0.012192778289318085, "global_step": 251558, "epoch": 2826, "lr": 1.3238805238029062e-06}
{"train_loss": 0.0364447720348835, "global_step": 251559, "epoch": 2826, "lr": 1.3237480041095652e-06}
{"train_loss": 0.059332285076379776, "global_step": 251560, "epoch": 2826, "lr": 1.323615490960156e-06}
{"train_loss": 0.04680836945772171, "global_step": 251561, "epoch": 2826, "lr": 1.3234829843546847e-06}
{"train_loss": 0.040380097925662994, "global_step": 251562, "epoch": 2826, "lr": 1.32335048429319e-06}
{"train_loss": 0.05130312964320183, "global_step": 251563, "epoch": 2826, "lr": 1.3232179907756714e-06}
{"train_loss": 0.04731537401676178, "global_step": 251564, "epoch": 2826, "lr": 1.3230855038021518e-06}
{"train_loss": 0.06482290476560593, "global_step": 251565, "epoch": 2826, "lr": 1.3229530233726528e-06}
{"train_loss": 0.03648514673113823, "global_step": 251566, "epoch": 2826, "lr": 1.322820549487186e-06}
{"train_loss": 0.042637959122657776, "global_step": 251567, "epoch": 2826, "lr": 1.3226880821457731e-06}
{"train_loss": 0.05535053089261055, "global_step": 251568, "epoch": 2826, "lr": 1.3225556213484313e-06}
{"train_loss": 0.08875565230846405, "global_step": 251569, "epoch": 2826, "lr": 1.3224231670951882e-06}
{"train_loss": 0.051606930792331696, "global_step": 251570, "epoch": 2826, "lr": 1.322290719386038e-06}
{"train_loss": 0.034694306552410126, "global_step": 251571, "epoch": 2826, "lr": 1.322158278221025e-06}
{"train_loss": 0.10956008732318878, "global_step": 251572, "epoch": 2826, "lr": 1.3220258436001443e-06}
{"train_loss": 0.07303662598133087, "global_step": 251573, "epoch": 2826, "lr": 1.3218934155234286e-06}
{"train_loss": 0.052399642765522, "global_step": 251574, "epoch": 2826, "lr": 1.3217609939908893e-06}
{"train_loss": 0.052101705223321915, "global_step": 251575, "epoch": 2826, "lr": 1.3216285790025485e-06}
{"train_loss": 0.04401473328471184, "global_step": 251576, "epoch": 2826, "lr": 1.3214961705584173e-06}
{"train_loss": 0.05826890468597412, "global_step": 251577, "epoch": 2826, "lr": 1.3213637686585235e-06}
{"train_loss": 0.05663388594985008, "global_step": 251578, "epoch": 2826, "lr": 1.321231373302878e-06}
{"train_loss": 0.07781919836997986, "global_step": 251579, "epoch": 2826, "lr": 1.321098984491498e-06}
{"train_loss": 0.06294777989387512, "global_step": 251580, "epoch": 2826, "lr": 1.3209666022243995e-06}
{"train_loss": 0.06471051275730133, "global_step": 251581, "epoch": 2826, "lr": 1.3208342265016106e-06}
{"train_loss": 0.02773856371641159, "global_step": 251582, "epoch": 2826, "lr": 1.3207018573231367e-06}
{"train_loss": 0.021911902353167534, "global_step": 251583, "epoch": 2826, "lr": 1.320569494689e-06}
{"train_loss": 0.02828887104988098, "global_step": 251584, "epoch": 2826, "lr": 1.3204371385992232e-06}
{"train_loss": 0.041725292801856995, "global_step": 251585, "epoch": 2826, "lr": 1.3203047890538167e-06}
{"train_loss": 0.03147142007946968, "global_step": 251586, "epoch": 2826, "lr": 1.3201724460528086e-06}
{"train_loss": 0.03411521017551422, "global_step": 251587, "epoch": 2826, "lr": 1.3200401095962044e-06}
{"train_loss": 0.016071565449237823, "global_step": 251588, "epoch": 2826, "lr": 1.319907779684032e-06}
{"train_loss": 0.02121836133301258, "global_step": 251589, "epoch": 2826, "lr": 1.3197754563163024e-06}
{"train_loss": 0.009167521260678768, "global_step": 251590, "epoch": 2826, "lr": 1.319643139493032e-06}
{"train_loss": 0.03086484968662262, "global_step": 251591, "epoch": 2826, "lr": 1.3195108292142488e-06}
{"train_loss": 0.05376775562763214, "global_step": 251592, "epoch": 2826, "lr": 1.3193785254799585e-06}
{"train_loss": 0.06406983733177185, "global_step": 251593, "epoch": 2826, "lr": 1.3192462282901886e-06}
{"train_loss": 0.05356162041425705, "global_step": 251594, "epoch": 2826, "lr": 1.3191139376449502e-06}
{"train_loss": 0.028492208570241928, "global_step": 251595, "epoch": 2826, "lr": 1.3189816535442657e-06}
{"train_loss": 0.010457347147166729, "global_step": 251596, "epoch": 2826, "lr": 1.3188493759881514e-06}
{"train_loss": 0.014428986236453056, "global_step": 251597, "epoch": 2826, "lr": 1.318717104976619e-06}
{"train_loss": 0.026204071938991547, "global_step": 251598, "epoch": 2826, "lr": 1.3185848405096957e-06}
{"train_loss": 0.0509532205760479, "global_step": 251599, "epoch": 2826, "lr": 1.3184525825873929e-06}
{"train_loss": 0.03458164259791374, "global_step": 251600, "epoch": 2826, "lr": 1.3183203312097324e-06}
{"train_loss": 0.04058272764086723, "global_step": 251601, "epoch": 2826, "lr": 1.3181880863767316e-06}
{"train_loss": 0.04690674633792277, "global_step": 251602, "epoch": 2826, "lr": 1.3180558480884009e-06, "val_loss": 9.610795974731445}
{"train_loss": 0.058521248400211334, "global_step": 251603, "epoch": 2827, "lr": 1.317923616344774e-06}
{"train_loss": 0.06953779608011246, "global_step": 251604, "epoch": 2827, "lr": 1.3177913911458505e-06}
{"train_loss": 0.026010319590568542, "global_step": 251605, "epoch": 2827, "lr": 1.3176591724916643e-06}
{"train_loss": 0.09426218271255493, "global_step": 251606, "epoch": 2827, "lr": 1.317526960382215e-06}
{"train_loss": 0.03999864682555199, "global_step": 251607, "epoch": 2827, "lr": 1.3173947548175413e-06}
{"train_loss": 0.05454482138156891, "global_step": 251608, "epoch": 2827, "lr": 1.3172625557976436e-06}
{"train_loss": 0.02790830470621586, "global_step": 251609, "epoch": 2827, "lr": 1.3171303633225495e-06}
{"train_loss": 0.020938286557793617, "global_step": 251610, "epoch": 2827, "lr": 1.3169981773922757e-06}
{"train_loss": 0.027905603870749474, "global_step": 251611, "epoch": 2827, "lr": 1.3168659980068387e-06}
{"train_loss": 0.04556524008512497, "global_step": 251612, "epoch": 2827, "lr": 1.3167338251662497e-06}
{"train_loss": 0.028671633452177048, "global_step": 251613, "epoch": 2827, "lr": 1.3166016588705365e-06}
{"train_loss": 0.013370082713663578, "global_step": 251614, "epoch": 2827, "lr": 1.3164694991197101e-06}
{"train_loss": 0.08843598514795303, "global_step": 251615, "epoch": 2827, "lr": 1.3163373459137985e-06}
{"train_loss": 0.019348228350281715, "global_step": 251616, "epoch": 2827, "lr": 1.3162051992528013e-06}
{"train_loss": 0.03834240883588791, "global_step": 251617, "epoch": 2827, "lr": 1.3160730591367577e-06}
{"train_loss": 0.06636705994606018, "global_step": 251618, "epoch": 2827, "lr": 1.3159409255656673e-06}
{"train_loss": 0.040063630789518356, "global_step": 251619, "epoch": 2827, "lr": 1.315808798539553e-06}
{"train_loss": 0.02232125960290432, "global_step": 251620, "epoch": 2827, "lr": 1.3156766780584417e-06}
{"train_loss": 0.03836970776319504, "global_step": 251621, "epoch": 2827, "lr": 1.3155445641223395e-06}
{"train_loss": 0.04491959139704704, "global_step": 251622, "epoch": 2827, "lr": 1.3154124567312743e-06}
{"train_loss": 0.018162153661251068, "global_step": 251623, "epoch": 2827, "lr": 1.3152803558852512e-06}
{"train_loss": 0.032990917563438416, "global_step": 251624, "epoch": 2827, "lr": 1.3151482615843036e-06}
{"train_loss": 0.02501128613948822, "global_step": 251625, "epoch": 2827, "lr": 1.3150161738284317e-06}
{"train_loss": 0.02953490987420082, "global_step": 251626, "epoch": 2827, "lr": 1.3148840926176687e-06}
{"train_loss": 0.041554998606443405, "global_step": 251627, "epoch": 2827, "lr": 1.3147520179520256e-06}
{"train_loss": 0.08086427301168442, "global_step": 251628, "epoch": 2827, "lr": 1.3146199498315192e-06}
{"train_loss": 0.03399212658405304, "global_step": 251629, "epoch": 2827, "lr": 1.3144878882561773e-06}
{"train_loss": 0.05897792428731918, "global_step": 251630, "epoch": 2827, "lr": 1.3143558332259942e-06}
{"train_loss": 0.06503735482692719, "global_step": 251631, "epoch": 2827, "lr": 1.3142237847410144e-06}
{"train_loss": 0.07064425945281982, "global_step": 251632, "epoch": 2827, "lr": 1.3140917428012378e-06}
{"train_loss": 0.03295242786407471, "global_step": 251633, "epoch": 2827, "lr": 1.3139597074066924e-06}
{"train_loss": 0.04768148437142372, "global_step": 251634, "epoch": 2827, "lr": 1.3138276785573888e-06}
{"train_loss": 0.0645592212677002, "global_step": 251635, "epoch": 2827, "lr": 1.3136956562533443e-06}
{"train_loss": 0.04077277332544327, "global_step": 251636, "epoch": 2827, "lr": 1.3135636404945862e-06}
{"train_loss": 0.037074100226163864, "global_step": 251637, "epoch": 2827, "lr": 1.3134316312811256e-06}
{"train_loss": 0.06196953356266022, "global_step": 251638, "epoch": 2827, "lr": 1.313299628612974e-06}
{"train_loss": 0.03702836111187935, "global_step": 251639, "epoch": 2827, "lr": 1.3131676324901643e-06}
{"train_loss": 0.027253298088908195, "global_step": 251640, "epoch": 2827, "lr": 1.3130356429126967e-06}
{"train_loss": 0.042242858558893204, "global_step": 251641, "epoch": 2827, "lr": 1.3129036598806045e-06}
{"train_loss": 0.049023665487766266, "global_step": 251642, "epoch": 2827, "lr": 1.3127716833938985e-06}
{"train_loss": 0.029913870617747307, "global_step": 251643, "epoch": 2827, "lr": 1.3126397134525959e-06}
{"train_loss": 0.08348026871681213, "global_step": 251644, "epoch": 2827, "lr": 1.312507750056713e-06}
{"train_loss": 0.0295303612947464, "global_step": 251645, "epoch": 2827, "lr": 1.3123757932062774e-06}
{"train_loss": 0.04074050486087799, "global_step": 251646, "epoch": 2827, "lr": 1.3122438429012896e-06}
{"train_loss": 0.04866182431578636, "global_step": 251647, "epoch": 2827, "lr": 1.312111899141788e-06}
{"train_loss": 0.07316014170646667, "global_step": 251648, "epoch": 2827, "lr": 1.311979961927773e-06}
{"train_loss": 0.017214128747582436, "global_step": 251649, "epoch": 2827, "lr": 1.3118480312592719e-06}
{"train_loss": 0.052780166268348694, "global_step": 251650, "epoch": 2827, "lr": 1.3117161071362905e-06}
{"train_loss": 0.03337182477116585, "global_step": 251651, "epoch": 2827, "lr": 1.3115841895588677e-06}
{"train_loss": 0.05410820618271828, "global_step": 251652, "epoch": 2827, "lr": 1.3114522785269978e-06}
{"train_loss": 0.0592426098883152, "global_step": 251653, "epoch": 2827, "lr": 1.3113203740407199e-06}
{"train_loss": 0.02241826429963112, "global_step": 251654, "epoch": 2827, "lr": 1.3111884761000337e-06}
{"train_loss": 0.023267235606908798, "global_step": 251655, "epoch": 2827, "lr": 1.3110565847049672e-06}
{"train_loss": 0.027642447501420975, "global_step": 251656, "epoch": 2827, "lr": 1.3109246998555369e-06}
{"train_loss": 0.02999080717563629, "global_step": 251657, "epoch": 2827, "lr": 1.3107928215517595e-06}
{"train_loss": 0.04942942038178444, "global_step": 251658, "epoch": 2827, "lr": 1.3106609497936518e-06}
{"train_loss": 0.01898273639380932, "global_step": 251659, "epoch": 2827, "lr": 1.3105290845812302e-06}
{"train_loss": 0.05672973394393921, "global_step": 251660, "epoch": 2827, "lr": 1.3103972259145226e-06}
{"train_loss": 0.04795113578438759, "global_step": 251661, "epoch": 2827, "lr": 1.3102653737935289e-06}
{"train_loss": 0.03364000469446182, "global_step": 251662, "epoch": 2827, "lr": 1.3101335282182825e-06}
{"train_loss": 0.044200312346220016, "global_step": 251663, "epoch": 2827, "lr": 1.3100016891887889e-06}
{"train_loss": 0.07653334736824036, "global_step": 251664, "epoch": 2827, "lr": 1.3098698567050815e-06}
{"train_loss": 0.058037176728248596, "global_step": 251665, "epoch": 2827, "lr": 1.3097380307671603e-06}
{"train_loss": 0.05689976364374161, "global_step": 251666, "epoch": 2827, "lr": 1.3096062113750585e-06}
{"train_loss": 0.07973886281251907, "global_step": 251667, "epoch": 2827, "lr": 1.3094743985287872e-06}
{"train_loss": 0.06971383094787598, "global_step": 251668, "epoch": 2827, "lr": 1.3093425922283576e-06}
{"train_loss": 0.04526718333363533, "global_step": 251669, "epoch": 2827, "lr": 1.3092107924737973e-06}
{"train_loss": 0.037448134273290634, "global_step": 251670, "epoch": 2827, "lr": 1.309078999265112e-06}
{"train_loss": 0.03359634801745415, "global_step": 251671, "epoch": 2827, "lr": 1.3089472126023406e-06}
{"train_loss": 0.043449584394693375, "global_step": 251672, "epoch": 2827, "lr": 1.3088154324854828e-06}
{"train_loss": 0.03831184655427933, "global_step": 251673, "epoch": 2827, "lr": 1.3086836589145557e-06}
{"train_loss": 0.07339572161436081, "global_step": 251674, "epoch": 2827, "lr": 1.3085518918895812e-06}
{"train_loss": 0.07777734845876694, "global_step": 251675, "epoch": 2827, "lr": 1.308420131410587e-06}
{"train_loss": 0.05723758414387703, "global_step": 251676, "epoch": 2827, "lr": 1.3082883774775789e-06}
{"train_loss": 0.026033218950033188, "global_step": 251677, "epoch": 2827, "lr": 1.308156630090579e-06}
{"train_loss": 0.040101565420627594, "global_step": 251678, "epoch": 2827, "lr": 1.3080248892495984e-06}
{"train_loss": 0.057387273758649826, "global_step": 251679, "epoch": 2827, "lr": 1.3078931549546702e-06}
{"train_loss": 0.06900060921907425, "global_step": 251680, "epoch": 2827, "lr": 1.3077614272057949e-06}
{"train_loss": 0.040786538273096085, "global_step": 251681, "epoch": 2827, "lr": 1.3076297060029995e-06}
{"train_loss": 0.01888456381857395, "global_step": 251682, "epoch": 2827, "lr": 1.3074979913462959e-06}
{"train_loss": 0.02285042591392994, "global_step": 251683, "epoch": 2827, "lr": 1.3073662832357114e-06}
{"train_loss": 0.040534235537052155, "global_step": 251684, "epoch": 2827, "lr": 1.3072345816712572e-06}
{"train_loss": 0.0508158802986145, "global_step": 251685, "epoch": 2827, "lr": 1.30710288665295e-06}
{"train_loss": 0.04448298364877701, "global_step": 251686, "epoch": 2827, "lr": 1.306971198180812e-06}
{"train_loss": 0.03417100012302399, "global_step": 251687, "epoch": 2827, "lr": 1.3068395162548596e-06}
{"train_loss": 0.05506691336631775, "global_step": 251688, "epoch": 2827, "lr": 1.3067078408750987e-06}
{"train_loss": 0.04713909700512886, "global_step": 251689, "epoch": 2827, "lr": 1.3065761720415682e-06}
{"train_loss": 0.049827590584754944, "global_step": 251690, "epoch": 2827, "lr": 1.3064445097542732e-06}
{"train_loss": 0.04552136293581028, "global_step": 251691, "epoch": 2827, "lr": 1.3063128540132308e-06, "val_loss": 9.568436622619629}
{"train_loss": 0.06355728954076767, "global_step": 251692, "epoch": 2828, "lr": 1.306181204818463e-06}
{"train_loss": 0.061974409967660904, "global_step": 251693, "epoch": 2828, "lr": 1.306049562169981e-06}
{"train_loss": 0.0508008636534214, "global_step": 251694, "epoch": 2828, "lr": 1.3059179260678123e-06}
{"train_loss": 0.061324458569288254, "global_step": 251695, "epoch": 2828, "lr": 1.3057862965119627e-06}
{"train_loss": 0.03063160553574562, "global_step": 251696, "epoch": 2828, "lr": 1.3056546735024655e-06}
{"train_loss": 0.044855497777462006, "global_step": 251697, "epoch": 2828, "lr": 1.3055230570393262e-06}
{"train_loss": 0.026191875338554382, "global_step": 251698, "epoch": 2828, "lr": 1.305391447122567e-06}
{"train_loss": 0.015862401574850082, "global_step": 251699, "epoch": 2828, "lr": 1.305259843752199e-06}
{"train_loss": 0.028454884886741638, "global_step": 251700, "epoch": 2828, "lr": 1.3051282469282555e-06}
{"train_loss": 0.053710803389549255, "global_step": 251701, "epoch": 2828, "lr": 1.304996656650731e-06}
{"train_loss": 0.018617408350110054, "global_step": 251702, "epoch": 2828, "lr": 1.30486507291967e-06}
{"train_loss": 0.05151024088263512, "global_step": 251703, "epoch": 2828, "lr": 1.3047334957350665e-06}
{"train_loss": 0.026248721405863762, "global_step": 251704, "epoch": 2828, "lr": 1.3046019250969487e-06}
{"train_loss": 0.03296476975083351, "global_step": 251705, "epoch": 2828, "lr": 1.3044703610053388e-06}
{"train_loss": 0.05894828960299492, "global_step": 251706, "epoch": 2828, "lr": 1.3043388034602476e-06}
{"train_loss": 0.028050832450389862, "global_step": 251707, "epoch": 2828, "lr": 1.3042072524616922e-06}
{"train_loss": 0.022508172318339348, "global_step": 251708, "epoch": 2828, "lr": 1.3040757080096944e-06}
{"train_loss": 0.045282378792762756, "global_step": 251709, "epoch": 2828, "lr": 1.3039441701042709e-06}
{"train_loss": 0.04109581187367439, "global_step": 251710, "epoch": 2828, "lr": 1.303812638745433e-06}
{"train_loss": 0.06439695507287979, "global_step": 251711, "epoch": 2828, "lr": 1.303681113933214e-06}
{"train_loss": 0.04355273023247719, "global_step": 251712, "epoch": 2828, "lr": 1.3035495956676135e-06}
{"train_loss": 0.07234032452106476, "global_step": 251713, "epoch": 2828, "lr": 1.30341808394866e-06}
{"train_loss": 0.022327763959765434, "global_step": 251714, "epoch": 2828, "lr": 1.3032865787763638e-06}
{"train_loss": 0.026404697448015213, "global_step": 251715, "epoch": 2828, "lr": 1.303155080150753e-06}
{"train_loss": 0.022949259728193283, "global_step": 251716, "epoch": 2828, "lr": 1.3030235880718388e-06}
{"train_loss": 0.11877568811178207, "global_step": 251717, "epoch": 2828, "lr": 1.302892102539638e-06}
{"train_loss": 0.034770723432302475, "global_step": 251718, "epoch": 2828, "lr": 1.3027606235541667e-06}
{"train_loss": 0.011011014692485332, "global_step": 251719, "epoch": 2828, "lr": 1.3026291511154531e-06}
{"train_loss": 0.05960232764482498, "global_step": 251720, "epoch": 2828, "lr": 1.3024976852235028e-06}
{"train_loss": 0.041802745312452316, "global_step": 251721, "epoch": 2828, "lr": 1.302366225878343e-06}
{"train_loss": 0.029672320932149887, "global_step": 251722, "epoch": 2828, "lr": 1.30223477307998e-06}
{"train_loss": 0.030745919793844223, "global_step": 251723, "epoch": 2828, "lr": 1.302103326828441e-06}
{"train_loss": 0.02104790508747101, "global_step": 251724, "epoch": 2828, "lr": 1.3019718871237375e-06}
{"train_loss": 0.05215775966644287, "global_step": 251725, "epoch": 2828, "lr": 1.3018404539658967e-06}
{"train_loss": 0.05492297559976578, "global_step": 251726, "epoch": 2828, "lr": 1.3017090273549193e-06}
{"train_loss": 0.09050072729587555, "global_step": 251727, "epoch": 2828, "lr": 1.3015776072908437e-06}
{"train_loss": 0.052953533828258514, "global_step": 251728, "epoch": 2828, "lr": 1.3014461937736756e-06}
{"train_loss": 0.016500068828463554, "global_step": 251729, "epoch": 2828, "lr": 1.301314786803426e-06}
{"train_loss": 0.0913066416978836, "global_step": 251730, "epoch": 2828, "lr": 1.3011833863801338e-06}
{"train_loss": 0.03208121284842491, "global_step": 251731, "epoch": 2828, "lr": 1.3010519925037934e-06}
{"train_loss": 0.05856224149465561, "global_step": 251732, "epoch": 2828, "lr": 1.300920605174444e-06}
{"train_loss": 0.042166002094745636, "global_step": 251733, "epoch": 2828, "lr": 1.3007892243920795e-06}
{"train_loss": 0.08397514373064041, "global_step": 251734, "epoch": 2828, "lr": 1.300657850156739e-06}
{"train_loss": 0.05823057144880295, "global_step": 251735, "epoch": 2828, "lr": 1.3005264824684283e-06}
{"train_loss": 0.08212507516145706, "global_step": 251736, "epoch": 2828, "lr": 1.3003951213271748e-06}
{"train_loss": 0.018583040684461594, "global_step": 251737, "epoch": 2828, "lr": 1.3002637667329786e-06}
{"train_loss": 0.009312784299254417, "global_step": 251738, "epoch": 2828, "lr": 1.3001324186858787e-06}
{"train_loss": 0.041199516505002975, "global_step": 251739, "epoch": 2828, "lr": 1.300001077185875e-06}
{"train_loss": 0.02006477862596512, "global_step": 251740, "epoch": 2828, "lr": 1.2998697422329952e-06}
{"train_loss": 0.02929249405860901, "global_step": 251741, "epoch": 2828, "lr": 1.2997384138272561e-06}
{"train_loss": 0.07603368908166885, "global_step": 251742, "epoch": 2828, "lr": 1.2996070919686743e-06}
{"train_loss": 0.10129798948764801, "global_step": 251743, "epoch": 2828, "lr": 1.2994757766572607e-06}
{"train_loss": 0.0751345306634903, "global_step": 251744, "epoch": 2828, "lr": 1.2993444678930434e-06}
{"train_loss": 0.036840081214904785, "global_step": 251745, "epoch": 2828, "lr": 1.2992131656760386e-06}
{"train_loss": 0.016248224303126335, "global_step": 251746, "epoch": 2828, "lr": 1.2990818700062524e-06}
{"train_loss": 0.06985314935445786, "global_step": 251747, "epoch": 2828, "lr": 1.2989505808837233e-06}
{"train_loss": 0.025918222963809967, "global_step": 251748, "epoch": 2828, "lr": 1.2988192983084456e-06}
{"train_loss": 0.04532841965556145, "global_step": 251749, "epoch": 2828, "lr": 1.2986880222804532e-06}
{"train_loss": 0.031509049236774445, "global_step": 251750, "epoch": 2828, "lr": 1.2985567527997566e-06}
{"train_loss": 0.047631457448005676, "global_step": 251751, "epoch": 2828, "lr": 1.2984254898663783e-06}
{"train_loss": 0.009230432100594044, "global_step": 251752, "epoch": 2828, "lr": 1.2982942334803294e-06}
{"train_loss": 0.031155524775385857, "global_step": 251753, "epoch": 2828, "lr": 1.2981629836416321e-06}
{"train_loss": 0.038405098021030426, "global_step": 251754, "epoch": 2828, "lr": 1.2980317403503028e-06}
{"train_loss": 0.023196974769234657, "global_step": 251755, "epoch": 2828, "lr": 1.2979005036063641e-06}
{"train_loss": 0.040239300578832626, "global_step": 251756, "epoch": 2828, "lr": 1.2977692734098213e-06}
{"train_loss": 0.018937217071652412, "global_step": 251757, "epoch": 2828, "lr": 1.2976380497607077e-06}
{"train_loss": 0.008537529036402702, "global_step": 251758, "epoch": 2828, "lr": 1.297506832659029e-06}
{"train_loss": 0.04234001040458679, "global_step": 251759, "epoch": 2828, "lr": 1.2973756221048071e-06}
{"train_loss": 0.034491002559661865, "global_step": 251760, "epoch": 2828, "lr": 1.2972444180980591e-06}
{"train_loss": 0.0450742207467556, "global_step": 251761, "epoch": 2828, "lr": 1.297113220638807e-06}
{"train_loss": 0.06271746009588242, "global_step": 251762, "epoch": 2828, "lr": 1.2969820297270562e-06}
{"train_loss": 0.03903286159038544, "global_step": 251763, "epoch": 2828, "lr": 1.2968508453628402e-06}
{"train_loss": 0.09712155163288116, "global_step": 251764, "epoch": 2828, "lr": 1.2967196675461646e-06}
{"train_loss": 0.07026290148496628, "global_step": 251765, "epoch": 2828, "lr": 1.2965884962770513e-06}
{"train_loss": 0.03384324535727501, "global_step": 251766, "epoch": 2828, "lr": 1.2964573315555229e-06}
{"train_loss": 0.03626550734043121, "global_step": 251767, "epoch": 2828, "lr": 1.2963261733815845e-06}
{"train_loss": 0.05982799455523491, "global_step": 251768, "epoch": 2828, "lr": 1.2961950217552699e-06}
{"train_loss": 0.01646840199828148, "global_step": 251769, "epoch": 2828, "lr": 1.2960638766765788e-06}
{"train_loss": 0.0644908919930458, "global_step": 251770, "epoch": 2828, "lr": 1.29593273814555e-06}
{"train_loss": 0.023227330297231674, "global_step": 251771, "epoch": 2828, "lr": 1.295801606162178e-06}
{"train_loss": 0.021729597821831703, "global_step": 251772, "epoch": 2828, "lr": 1.2956704807265019e-06}
{"train_loss": 0.044770248234272, "global_step": 251773, "epoch": 2828, "lr": 1.2955393618385214e-06}
{"train_loss": 0.04417709633708, "global_step": 251774, "epoch": 2828, "lr": 1.29540824949827e-06}
{"train_loss": 0.079413503408432, "global_step": 251775, "epoch": 2828, "lr": 1.2952771437057475e-06}
{"train_loss": 0.03299722075462341, "global_step": 251776, "epoch": 2828, "lr": 1.2951460444609876e-06}
{"train_loss": 0.051694996654987335, "global_step": 251777, "epoch": 2828, "lr": 1.2950149517640009e-06}
{"train_loss": 0.034888625144958496, "global_step": 251778, "epoch": 2828, "lr": 1.2948838656148044e-06}
{"train_loss": 0.05614105612039566, "global_step": 251779, "epoch": 2828, "lr": 1.29475278601342e-06}
{"train_loss": 0.04417865074584993, "global_step": 251780, "epoch": 2828, "lr": 1.2946217129598593e-06, "val_loss": 9.613462448120117}
{"train_loss": 0.04491046816110611, "global_step": 251781, "epoch": 2829, "lr": 1.294490646454144e-06}
{"train_loss": 0.019416002556681633, "global_step": 251782, "epoch": 2829, "lr": 1.294359586496291e-06}
{"train_loss": 0.042116113007068634, "global_step": 251783, "epoch": 2829, "lr": 1.2942285330863225e-06}
{"train_loss": 0.03350456804037094, "global_step": 251784, "epoch": 2829, "lr": 1.294097486224244e-06}
{"train_loss": 0.03339514508843422, "global_step": 251785, "epoch": 2829, "lr": 1.293966445910083e-06}
{"train_loss": 0.030904486775398254, "global_step": 251786, "epoch": 2829, "lr": 1.2938354121438568e-06}
{"train_loss": 0.04895491153001785, "global_step": 251787, "epoch": 2829, "lr": 1.2937043849255815e-06}
{"train_loss": 0.028777770698070526, "global_step": 251788, "epoch": 2829, "lr": 1.2935733642552683e-06}
{"train_loss": 0.05591130256652832, "global_step": 251789, "epoch": 2829, "lr": 1.293442350132945e-06}
{"train_loss": 0.05627303570508957, "global_step": 251790, "epoch": 2829, "lr": 1.2933113425586174e-06}
{"train_loss": 0.017675969749689102, "global_step": 251791, "epoch": 2829, "lr": 1.2931803415323184e-06}
{"train_loss": 0.0352330356836319, "global_step": 251792, "epoch": 2829, "lr": 1.293049347054054e-06}
{"train_loss": 0.06270589679479599, "global_step": 251793, "epoch": 2829, "lr": 1.2929183591238514e-06}
{"train_loss": 0.0603514164686203, "global_step": 251794, "epoch": 2829, "lr": 1.2927873777417165e-06}
{"train_loss": 0.06315819174051285, "global_step": 251795, "epoch": 2829, "lr": 1.292656402907677e-06}
{"train_loss": 0.06542743742465973, "global_step": 251796, "epoch": 2829, "lr": 1.2925254346217385e-06}
{"train_loss": 0.06299804896116257, "global_step": 251797, "epoch": 2829, "lr": 1.2923944728839343e-06}
{"train_loss": 0.03748483210802078, "global_step": 251798, "epoch": 2829, "lr": 1.29226351769427e-06}
{"train_loss": 0.0664253905415535, "global_step": 251799, "epoch": 2829, "lr": 1.2921325690527676e-06}
{"train_loss": 0.07200776040554047, "global_step": 251800, "epoch": 2829, "lr": 1.2920016269594438e-06}
{"train_loss": 0.047353293746709824, "global_step": 251801, "epoch": 2829, "lr": 1.2918706914143153e-06}
{"train_loss": 0.026866726577281952, "global_step": 251802, "epoch": 2829, "lr": 1.2917397624174044e-06}
{"train_loss": 0.03236158564686775, "global_step": 251803, "epoch": 2829, "lr": 1.291608839968722e-06}
{"train_loss": 0.057119160890579224, "global_step": 251804, "epoch": 2829, "lr": 1.2914779240682906e-06}
{"train_loss": 0.023048190400004387, "global_step": 251805, "epoch": 2829, "lr": 1.2913470147161267e-06}
{"train_loss": 0.09301966428756714, "global_step": 251806, "epoch": 2829, "lr": 1.2912161119122523e-06}
{"train_loss": 0.03928111121058464, "global_step": 251807, "epoch": 2829, "lr": 1.2910852156566734e-06}
{"train_loss": 0.03913053125143051, "global_step": 251808, "epoch": 2829, "lr": 1.2909543259494172e-06}
{"train_loss": 0.03436549752950668, "global_step": 251809, "epoch": 2829, "lr": 1.2908234427904953e-06}
{"train_loss": 0.03570074588060379, "global_step": 251810, "epoch": 2829, "lr": 1.2906925661799351e-06}
{"train_loss": 0.04502451792359352, "global_step": 251811, "epoch": 2829, "lr": 1.2905616961177424e-06}
{"train_loss": 0.02359764277935028, "global_step": 251812, "epoch": 2829, "lr": 1.2904308326039449e-06}
{"train_loss": 0.06787577271461487, "global_step": 251813, "epoch": 2829, "lr": 1.290299975638548e-06}
{"train_loss": 0.07091172784566879, "global_step": 251814, "epoch": 2829, "lr": 1.2901691252215797e-06}
{"train_loss": 0.05167233198881149, "global_step": 251815, "epoch": 2829, "lr": 1.2900382813530564e-06}
{"train_loss": 0.09366337954998016, "global_step": 251816, "epoch": 2829, "lr": 1.289907444032995e-06}
{"train_loss": 0.042391568422317505, "global_step": 251817, "epoch": 2829, "lr": 1.2897766132614063e-06}
{"train_loss": 0.01673891395330429, "global_step": 251818, "epoch": 2829, "lr": 1.2896457890383184e-06}
{"train_loss": 0.06421955674886703, "global_step": 251819, "epoch": 2829, "lr": 1.2895149713637423e-06}
{"train_loss": 0.05801931023597717, "global_step": 251820, "epoch": 2829, "lr": 1.2893841602376944e-06}
{"train_loss": 0.06392087787389755, "global_step": 251821, "epoch": 2829, "lr": 1.2892533556602027e-06}
{"train_loss": 0.026331568136811256, "global_step": 251822, "epoch": 2829, "lr": 1.289122557631267e-06}
{"train_loss": 0.06832519173622131, "global_step": 251823, "epoch": 2829, "lr": 1.2889917661509265e-06}
{"train_loss": 0.03292731195688248, "global_step": 251824, "epoch": 2829, "lr": 1.2888609812191754e-06}
{"train_loss": 0.049833666533231735, "global_step": 251825, "epoch": 2829, "lr": 1.2887302028360526e-06}
{"train_loss": 0.016321886330842972, "global_step": 251826, "epoch": 2829, "lr": 1.2885994310015636e-06}
{"train_loss": 0.03743647411465645, "global_step": 251827, "epoch": 2829, "lr": 1.2884686657157307e-06}
{"train_loss": 0.07492879778146744, "global_step": 251828, "epoch": 2829, "lr": 1.2883379069785595e-06}
{"train_loss": 0.09749704599380493, "global_step": 251829, "epoch": 2829, "lr": 1.2882071547900887e-06}
{"train_loss": 0.06648866832256317, "global_step": 251830, "epoch": 2829, "lr": 1.2880764091503184e-06}
{"train_loss": 0.053297258913517, "global_step": 251831, "epoch": 2829, "lr": 1.2879456700592763e-06}
{"train_loss": 0.07831152528524399, "global_step": 251832, "epoch": 2829, "lr": 1.2878149375169735e-06}
{"train_loss": 0.0719653069972992, "global_step": 251833, "epoch": 2829, "lr": 1.2876842115234322e-06}
{"train_loss": 0.0421585813164711, "global_step": 251834, "epoch": 2829, "lr": 1.2875534920786636e-06}
{"train_loss": 0.05121570825576782, "global_step": 251835, "epoch": 2829, "lr": 1.2874227791827009e-06}
{"train_loss": 0.05395737290382385, "global_step": 251836, "epoch": 2829, "lr": 1.2872920728355387e-06}
{"train_loss": 0.03986497223377228, "global_step": 251837, "epoch": 2829, "lr": 1.28716137303721e-06}
{"train_loss": 0.03254189342260361, "global_step": 251838, "epoch": 2829, "lr": 1.2870306797877319e-06}
{"train_loss": 0.058499544858932495, "global_step": 251839, "epoch": 2829, "lr": 1.2868999930871095e-06}
{"train_loss": 0.05557156726717949, "global_step": 251840, "epoch": 2829, "lr": 1.2867693129353818e-06}
{"train_loss": 0.020624330267310143, "global_step": 251841, "epoch": 2829, "lr": 1.2866386393325436e-06}
{"train_loss": 0.03265554830431938, "global_step": 251842, "epoch": 2829, "lr": 1.2865079722786334e-06}
{"train_loss": 0.02666429989039898, "global_step": 251843, "epoch": 2829, "lr": 1.2863773117736511e-06}
{"train_loss": 0.048361532390117645, "global_step": 251844, "epoch": 2829, "lr": 1.2862466578176246e-06}
{"train_loss": 0.05043458938598633, "global_step": 251845, "epoch": 2829, "lr": 1.2861160104105652e-06}
{"train_loss": 0.07996794581413269, "global_step": 251846, "epoch": 2829, "lr": 1.2859853695525003e-06}
{"train_loss": 0.03138865903019905, "global_step": 251847, "epoch": 2829, "lr": 1.2858547352434359e-06}
{"train_loss": 0.03890791907906532, "global_step": 251848, "epoch": 2829, "lr": 1.2857241074833936e-06}
{"train_loss": 0.02635270170867443, "global_step": 251849, "epoch": 2829, "lr": 1.2855934862723962e-06}
{"train_loss": 0.05049164220690727, "global_step": 251850, "epoch": 2829, "lr": 1.2854628716104544e-06}
{"train_loss": 0.11914458125829697, "global_step": 251851, "epoch": 2829, "lr": 1.285332263497585e-06}
{"train_loss": 0.034414142370224, "global_step": 251852, "epoch": 2829, "lr": 1.285201661933816e-06}
{"train_loss": 0.057564906775951385, "global_step": 251853, "epoch": 2829, "lr": 1.2850710669191524e-06}
{"train_loss": 0.0242146048694849, "global_step": 251854, "epoch": 2829, "lr": 1.2849404784536224e-06}
{"train_loss": 0.033628690987825394, "global_step": 251855, "epoch": 2829, "lr": 1.2848098965372314e-06}
{"train_loss": 0.02109912596642971, "global_step": 251856, "epoch": 2829, "lr": 1.2846793211700127e-06}
{"train_loss": 0.04878836125135422, "global_step": 251857, "epoch": 2829, "lr": 1.284548752351966e-06}
{"train_loss": 0.05655088648200035, "global_step": 251858, "epoch": 2829, "lr": 1.2844181900831253e-06}
{"train_loss": 0.028187496587634087, "global_step": 251859, "epoch": 2829, "lr": 1.2842876343634957e-06}
{"train_loss": 0.02695663459599018, "global_step": 251860, "epoch": 2829, "lr": 1.284157085193105e-06}
{"train_loss": 0.03124335967004299, "global_step": 251861, "epoch": 2829, "lr": 1.2840265425719643e-06}
{"train_loss": 0.025987450033426285, "global_step": 251862, "epoch": 2829, "lr": 1.2838960065000904e-06}
{"train_loss": 0.03033585660159588, "global_step": 251863, "epoch": 2829, "lr": 1.2837654769775055e-06}
{"train_loss": 0.049875348806381226, "global_step": 251864, "epoch": 2829, "lr": 1.2836349540042204e-06}
{"train_loss": 0.059120986610651016, "global_step": 251865, "epoch": 2829, "lr": 1.2835044375802629e-06}
{"train_loss": 0.03986417129635811, "global_step": 251866, "epoch": 2829, "lr": 1.2833739277056388e-06}
{"train_loss": 0.06259158253669739, "global_step": 251867, "epoch": 2829, "lr": 1.2832434243803703e-06}
{"train_loss": 0.008390824310481548, "global_step": 251868, "epoch": 2829, "lr": 1.2831129276044796e-06}
{"train_loss": 0.04655929476943579, "global_step": 251869, "epoch": 2829, "lr": 1.2829824373779831e-06, "val_loss": 9.61711597442627}
{"train_loss": 0.035581640899181366, "global_step": 251870, "epoch": 2830, "lr": 1.2828519537008864e-06}
{"train_loss": 0.08707499504089355, "global_step": 251871, "epoch": 2830, "lr": 1.2827214765732287e-06}
{"train_loss": 0.03431376814842224, "global_step": 251872, "epoch": 2830, "lr": 1.2825910059950042e-06}
{"train_loss": 0.030957330018281937, "global_step": 251873, "epoch": 2830, "lr": 1.2824605419662462e-06}
{"train_loss": 0.03408646583557129, "global_step": 251874, "epoch": 2830, "lr": 1.2823300844869657e-06}
{"train_loss": 0.029898101463913918, "global_step": 251875, "epoch": 2830, "lr": 1.282199633557185e-06}
{"train_loss": 0.055291492491960526, "global_step": 251876, "epoch": 2830, "lr": 1.282069189176921e-06}
{"train_loss": 0.058934327214956284, "global_step": 251877, "epoch": 2830, "lr": 1.2819387513461845e-06}
{"train_loss": 0.029162606224417686, "global_step": 251878, "epoch": 2830, "lr": 1.2818083200649978e-06}
{"train_loss": 0.028956374153494835, "global_step": 251879, "epoch": 2830, "lr": 1.2816778953333775e-06}
{"train_loss": 0.035578418523073196, "global_step": 251880, "epoch": 2830, "lr": 1.2815474771513458e-06}
{"train_loss": 0.03882887214422226, "global_step": 251881, "epoch": 2830, "lr": 1.2814170655189139e-06}
{"train_loss": 0.03491848334670067, "global_step": 251882, "epoch": 2830, "lr": 1.281286660436104e-06}
{"train_loss": 0.01980133168399334, "global_step": 251883, "epoch": 2830, "lr": 1.2811562619029271e-06}
{"train_loss": 0.05935114622116089, "global_step": 251884, "epoch": 2830, "lr": 1.281025869919411e-06}
{"train_loss": 0.03683644160628319, "global_step": 251885, "epoch": 2830, "lr": 1.280895484485556e-06}
{"train_loss": 0.03505023196339607, "global_step": 251886, "epoch": 2830, "lr": 1.2807651056014002e-06}
{"train_loss": 0.05926285311579704, "global_step": 251887, "epoch": 2830, "lr": 1.28063473326695e-06}
{"train_loss": 0.04245145246386528, "global_step": 251888, "epoch": 2830, "lr": 1.280504367482227e-06}
{"train_loss": 0.04739725962281227, "global_step": 251889, "epoch": 2830, "lr": 1.2803740082472426e-06}
{"train_loss": 0.030019886791706085, "global_step": 251890, "epoch": 2830, "lr": 1.280243655562019e-06}
{"train_loss": 0.04312711954116821, "global_step": 251891, "epoch": 2830, "lr": 1.2801133094265727e-06}
{"train_loss": 0.03992238640785217, "global_step": 251892, "epoch": 2830, "lr": 1.279982969840915e-06}
{"train_loss": 0.036491744220256805, "global_step": 251893, "epoch": 2830, "lr": 1.279852636805079e-06}
{"train_loss": 0.028331924229860306, "global_step": 251894, "epoch": 2830, "lr": 1.2797223103190759e-06}
{"train_loss": 0.035385992377996445, "global_step": 251895, "epoch": 2830, "lr": 1.2795919903829111e-06}
{"train_loss": 0.07087916135787964, "global_step": 251896, "epoch": 2830, "lr": 1.2794616769966184e-06}
{"train_loss": 0.0547475591301918, "global_step": 251897, "epoch": 2830, "lr": 1.2793313701602028e-06}
{"train_loss": 0.031184270977973938, "global_step": 251898, "epoch": 2830, "lr": 1.2792010698736922e-06}
{"train_loss": 0.06844776123762131, "global_step": 251899, "epoch": 2830, "lr": 1.2790707761370978e-06}
{"train_loss": 0.030111730098724365, "global_step": 251900, "epoch": 2830, "lr": 1.2789404889504365e-06}
{"train_loss": 0.04330752417445183, "global_step": 251901, "epoch": 2830, "lr": 1.27881020831373e-06}
{"train_loss": 0.013080191798508167, "global_step": 251902, "epoch": 2830, "lr": 1.2786799342269896e-06}
{"train_loss": 0.020783448591828346, "global_step": 251903, "epoch": 2830, "lr": 1.278549666690243e-06}
{"train_loss": 0.017373211681842804, "global_step": 251904, "epoch": 2830, "lr": 1.2784194057034904e-06}
{"train_loss": 0.06378477066755295, "global_step": 251905, "epoch": 2830, "lr": 1.2782891512667705e-06}
{"train_loss": 0.037965357303619385, "global_step": 251906, "epoch": 2830, "lr": 1.2781589033800889e-06}
{"train_loss": 0.043593525886535645, "global_step": 251907, "epoch": 2830, "lr": 1.278028662043468e-06}
{"train_loss": 0.05412252992391586, "global_step": 251908, "epoch": 2830, "lr": 1.2778984272569127e-06}
{"train_loss": 0.01477221678942442, "global_step": 251909, "epoch": 2830, "lr": 1.2777681990204571e-06}
{"train_loss": 0.05523674190044403, "global_step": 251910, "epoch": 2830, "lr": 1.2776379773341118e-06}
{"train_loss": 0.047573015093803406, "global_step": 251911, "epoch": 2830, "lr": 1.2775077621978881e-06}
{"train_loss": 0.017169928178191185, "global_step": 251912, "epoch": 2830, "lr": 1.2773775536118193e-06}
{"train_loss": 0.05025231093168259, "global_step": 251913, "epoch": 2830, "lr": 1.2772473515759053e-06}
{"train_loss": 0.04947400465607643, "global_step": 251914, "epoch": 2830, "lr": 1.2771171560901796e-06}
{"train_loss": 0.05545514449477196, "global_step": 251915, "epoch": 2830, "lr": 1.276986967154642e-06}
{"train_loss": 0.04361545667052269, "global_step": 251916, "epoch": 2830, "lr": 1.2768567847693259e-06}
{"train_loss": 0.06500668078660965, "global_step": 251917, "epoch": 2830, "lr": 1.2767266089342421e-06}
{"train_loss": 0.028082575649023056, "global_step": 251918, "epoch": 2830, "lr": 1.276596439649408e-06}
{"train_loss": 0.02107415348291397, "global_step": 251919, "epoch": 2830, "lr": 1.2764662769148395e-06}
{"train_loss": 0.04220542311668396, "global_step": 251920, "epoch": 2830, "lr": 1.276336120730559e-06}
{"train_loss": 0.04475686326622963, "global_step": 251921, "epoch": 2830, "lr": 1.276205971096578e-06}
{"train_loss": 0.02676229551434517, "global_step": 251922, "epoch": 2830, "lr": 1.2760758280129237e-06}
{"train_loss": 0.05095687136054039, "global_step": 251923, "epoch": 2830, "lr": 1.2759456914796019e-06}
{"train_loss": 0.02546069398522377, "global_step": 251924, "epoch": 2830, "lr": 1.275815561496635e-06}
{"train_loss": 0.03383094072341919, "global_step": 251925, "epoch": 2830, "lr": 1.2756854380640393e-06}
{"train_loss": 0.05283840373158455, "global_step": 251926, "epoch": 2830, "lr": 1.2755553211818427e-06}
{"train_loss": 0.04934419319033623, "global_step": 251927, "epoch": 2830, "lr": 1.2754252108500453e-06}
{"train_loss": 0.0716894194483757, "global_step": 251928, "epoch": 2830, "lr": 1.2752951070686747e-06}
{"train_loss": 0.08585841953754425, "global_step": 251929, "epoch": 2830, "lr": 1.2751650098377477e-06}
{"train_loss": 0.06532919406890869, "global_step": 251930, "epoch": 2830, "lr": 1.275034919157281e-06}
{"train_loss": 0.05008009076118469, "global_step": 251931, "epoch": 2830, "lr": 1.2749048350272907e-06}
{"train_loss": 0.11012870818376541, "global_step": 251932, "epoch": 2830, "lr": 1.2747747574477997e-06}
{"train_loss": 0.06665265560150146, "global_step": 251933, "epoch": 2830, "lr": 1.2746446864188188e-06}
{"train_loss": 0.030623145401477814, "global_step": 251934, "epoch": 2830, "lr": 1.2745146219403703e-06}
{"train_loss": 0.038056470453739166, "global_step": 251935, "epoch": 2830, "lr": 1.2743845640124653e-06}
{"train_loss": 0.0486820749938488, "global_step": 251936, "epoch": 2830, "lr": 1.2742545126351258e-06}
{"train_loss": 0.0570109561085701, "global_step": 251937, "epoch": 2830, "lr": 1.274124467808374e-06}
{"train_loss": 0.04975103214383125, "global_step": 251938, "epoch": 2830, "lr": 1.273994429532216e-06}
{"train_loss": 0.03104252740740776, "global_step": 251939, "epoch": 2830, "lr": 1.273864397806679e-06}
{"train_loss": 0.020884724333882332, "global_step": 251940, "epoch": 2830, "lr": 1.273734372631774e-06}
{"train_loss": 0.05994899198412895, "global_step": 251941, "epoch": 2830, "lr": 1.2736043540075294e-06}
{"train_loss": 0.0397331565618515, "global_step": 251942, "epoch": 2830, "lr": 1.2734743419339446e-06}
{"train_loss": 0.105037622153759, "global_step": 251943, "epoch": 2830, "lr": 1.273344336411053e-06}
{"train_loss": 0.030232351273298264, "global_step": 251944, "epoch": 2830, "lr": 1.2732143374388604e-06}
{"train_loss": 0.057812534272670746, "global_step": 251945, "epoch": 2830, "lr": 1.2730843450174002e-06}
{"train_loss": 0.028141243383288383, "global_step": 251946, "epoch": 2830, "lr": 1.2729543591466719e-06}
{"train_loss": 0.055813971906900406, "global_step": 251947, "epoch": 2830, "lr": 1.2728243798266982e-06}
{"train_loss": 0.07976420223712921, "global_step": 251948, "epoch": 2830, "lr": 1.2726944070575119e-06}
{"train_loss": 0.06030287966132164, "global_step": 251949, "epoch": 2830, "lr": 1.272564440839108e-06}
{"train_loss": 0.039199475198984146, "global_step": 251950, "epoch": 2830, "lr": 1.2724344811715194e-06}
{"train_loss": 0.023218460381031036, "global_step": 251951, "epoch": 2830, "lr": 1.2723045280547518e-06}
{"train_loss": 0.010324343107640743, "global_step": 251952, "epoch": 2830, "lr": 1.272174581488833e-06}
{"train_loss": 0.03911255672574043, "global_step": 251953, "epoch": 2830, "lr": 1.272044641473774e-06}
{"train_loss": 0.06369195878505707, "global_step": 251954, "epoch": 2830, "lr": 1.2719147080095971e-06}
{"train_loss": 0.02217000350356102, "global_step": 251955, "epoch": 2830, "lr": 1.271784781096319e-06}
{"train_loss": 0.06237269565463066, "global_step": 251956, "epoch": 2830, "lr": 1.2716548607339561e-06}
{"train_loss": 0.041035182774066925, "global_step": 251957, "epoch": 2830, "lr": 1.27152494692252e-06}
{"train_loss": 0.04426595695274935, "global_step": 251958, "epoch": 2830, "lr": 1.2713950396620377e-06, "val_loss": 9.60142707824707, "train_action_mse_error": 20.956146240234375}
{"train_loss": 0.03238082304596901, "global_step": 251959, "epoch": 2831, "lr": 1.2712651389525154e-06}
{"train_loss": 0.008670918643474579, "global_step": 251960, "epoch": 2831, "lr": 1.2711352447939862e-06}
{"train_loss": 0.007232550997287035, "global_step": 251961, "epoch": 2831, "lr": 1.2710053571864555e-06}
{"train_loss": 0.06774566322565079, "global_step": 251962, "epoch": 2831, "lr": 1.2708754761299458e-06}
{"train_loss": 0.04623805731534958, "global_step": 251963, "epoch": 2831, "lr": 1.270745601624468e-06}
{"train_loss": 0.0389845035970211, "global_step": 251964, "epoch": 2831, "lr": 1.270615733670044e-06}
{"train_loss": 0.04924795776605606, "global_step": 251965, "epoch": 2831, "lr": 1.2704858722667022e-06}
{"train_loss": 0.08649803698062897, "global_step": 251966, "epoch": 2831, "lr": 1.2703560174144369e-06}
{"train_loss": 0.01864904910326004, "global_step": 251967, "epoch": 2831, "lr": 1.2702261691132922e-06}
{"train_loss": 0.04791247472167015, "global_step": 251968, "epoch": 2831, "lr": 1.2700963273632571e-06}
{"train_loss": 0.0247955322265625, "global_step": 251969, "epoch": 2831, "lr": 1.2699664921643762e-06}
{"train_loss": 0.019447894766926765, "global_step": 251970, "epoch": 2831, "lr": 1.2698366635166437e-06}
{"train_loss": 0.03532318025827408, "global_step": 251971, "epoch": 2831, "lr": 1.2697068414200985e-06}
{"train_loss": 0.011320480145514011, "global_step": 251972, "epoch": 2831, "lr": 1.2695770258747407e-06}
{"train_loss": 0.03631023317575455, "global_step": 251973, "epoch": 2831, "lr": 1.2694472168805926e-06}
{"train_loss": 0.03962751105427742, "global_step": 251974, "epoch": 2831, "lr": 1.2693174144376818e-06}
{"train_loss": 0.05655418708920479, "global_step": 251975, "epoch": 2831, "lr": 1.2691876185460083e-06}
{"train_loss": 0.032109010964632034, "global_step": 251976, "epoch": 2831, "lr": 1.2690578292056055e-06}
{"train_loss": 0.04084044694900513, "global_step": 251977, "epoch": 2831, "lr": 1.2689280464164788e-06}
{"train_loss": 0.03961566090583801, "global_step": 251978, "epoch": 2831, "lr": 1.2687982701786506e-06}
{"train_loss": 0.05419974774122238, "global_step": 251979, "epoch": 2831, "lr": 1.2686685004921428e-06}
{"train_loss": 0.02773323282599449, "global_step": 251980, "epoch": 2831, "lr": 1.2685387373569668e-06}
{"train_loss": 0.03905763477087021, "global_step": 251981, "epoch": 2831, "lr": 1.2684089807731392e-06}
{"train_loss": 0.031623098999261856, "global_step": 251982, "epoch": 2831, "lr": 1.268279230740682e-06}
{"train_loss": 0.030902819707989693, "global_step": 251983, "epoch": 2831, "lr": 1.2681494872596067e-06}
{"train_loss": 0.08803877979516983, "global_step": 251984, "epoch": 2831, "lr": 1.2680197503299408e-06}
{"train_loss": 0.055372755974531174, "global_step": 251985, "epoch": 2831, "lr": 1.2678900199516952e-06}
{"train_loss": 0.02566877380013466, "global_step": 251986, "epoch": 2831, "lr": 1.2677602961248869e-06}
{"train_loss": 0.03618475794792175, "global_step": 251987, "epoch": 2831, "lr": 1.2676305788495324e-06}
{"train_loss": 0.03728030249476433, "global_step": 251988, "epoch": 2831, "lr": 1.2675008681256539e-06}
{"train_loss": 0.06940701603889465, "global_step": 251989, "epoch": 2831, "lr": 1.2673711639532626e-06}
{"train_loss": 0.055475134402513504, "global_step": 251990, "epoch": 2831, "lr": 1.2672414663323806e-06}
{"train_loss": 0.034790653735399246, "global_step": 251991, "epoch": 2831, "lr": 1.2671117752630191e-06}
{"train_loss": 0.05656816065311432, "global_step": 251992, "epoch": 2831, "lr": 1.2669820907452113e-06}
{"train_loss": 0.03260162100195885, "global_step": 251993, "epoch": 2831, "lr": 1.2668524127789572e-06}
{"train_loss": 0.02642211504280567, "global_step": 251994, "epoch": 2831, "lr": 1.2667227413642846e-06}
{"train_loss": 0.02536855638027191, "global_step": 251995, "epoch": 2831, "lr": 1.266593076501199e-06}
{"train_loss": 0.04885281249880791, "global_step": 251996, "epoch": 2831, "lr": 1.2664634181897338e-06}
{"train_loss": 0.021098405122756958, "global_step": 251997, "epoch": 2831, "lr": 1.2663337664298946e-06}
{"train_loss": 0.042686257511377335, "global_step": 251998, "epoch": 2831, "lr": 1.2662041212217035e-06}
{"train_loss": 0.04063975438475609, "global_step": 251999, "epoch": 2831, "lr": 1.266074482565177e-06}
{"train_loss": 0.04186133295297623, "global_step": 252000, "epoch": 2831, "lr": 1.2659448504603322e-06}
{"train_loss": 0.033270224928855896, "global_step": 252001, "epoch": 2831, "lr": 1.265815224907191e-06}
{"train_loss": 0.03863373398780823, "global_step": 252002, "epoch": 2831, "lr": 1.2656856059057643e-06}
{"train_loss": 0.02052478678524494, "global_step": 252003, "epoch": 2831, "lr": 1.2655559934560747e-06}
{"train_loss": 0.050885170698165894, "global_step": 252004, "epoch": 2831, "lr": 1.265426387558133e-06}
{"train_loss": 0.01949651539325714, "global_step": 252005, "epoch": 2831, "lr": 1.2652967882119615e-06}
{"train_loss": 0.03600010648369789, "global_step": 252006, "epoch": 2831, "lr": 1.265167195417577e-06}
{"train_loss": 0.060894351452589035, "global_step": 252007, "epoch": 2831, "lr": 1.2650376091750016e-06}
{"train_loss": 0.03364672139286995, "global_step": 252008, "epoch": 2831, "lr": 1.2649080294842409e-06}
{"train_loss": 0.033243365585803986, "global_step": 252009, "epoch": 2831, "lr": 1.2647784563453223e-06}
{"train_loss": 0.034784842282533646, "global_step": 252010, "epoch": 2831, "lr": 1.2646488897582632e-06}
{"train_loss": 0.036712974309921265, "global_step": 252011, "epoch": 2831, "lr": 1.264519329723074e-06}
{"train_loss": 0.051899123936891556, "global_step": 252012, "epoch": 2831, "lr": 1.2643897762397826e-06}
{"train_loss": 0.025866782292723656, "global_step": 252013, "epoch": 2831, "lr": 1.2642602293083894e-06}
{"train_loss": 0.02630445919930935, "global_step": 252014, "epoch": 2831, "lr": 1.2641306889289328e-06}
{"train_loss": 0.009861791506409645, "global_step": 252015, "epoch": 2831, "lr": 1.2640011551014186e-06}
{"train_loss": 0.07254068553447723, "global_step": 252016, "epoch": 2831, "lr": 1.2638716278258577e-06}
{"train_loss": 0.032217640429735184, "global_step": 252017, "epoch": 2831, "lr": 1.2637421071022837e-06}
{"train_loss": 0.04817264899611473, "global_step": 252018, "epoch": 2831, "lr": 1.2636125929306963e-06}
{"train_loss": 0.045519400388002396, "global_step": 252019, "epoch": 2831, "lr": 1.2634830853111235e-06}
{"train_loss": 0.026610732078552246, "global_step": 252020, "epoch": 2831, "lr": 1.2633535842435872e-06}
{"train_loss": 0.052542027086019516, "global_step": 252021, "epoch": 2831, "lr": 1.263224089728099e-06}
{"train_loss": 0.030329497531056404, "global_step": 252022, "epoch": 2831, "lr": 1.263094601764675e-06}
{"train_loss": 0.0209223460406065, "global_step": 252023, "epoch": 2831, "lr": 1.2629651203533266e-06}
{"train_loss": 0.0879134088754654, "global_step": 252024, "epoch": 2831, "lr": 1.2628356454940926e-06}
{"train_loss": 0.03821863234043121, "global_step": 252025, "epoch": 2831, "lr": 1.2627061771869618e-06}
{"train_loss": 0.04351815581321716, "global_step": 252026, "epoch": 2831, "lr": 1.2625767154319789e-06}
{"train_loss": 0.026800040155649185, "global_step": 252027, "epoch": 2831, "lr": 1.2624472602291382e-06}
{"train_loss": 0.023853052407503128, "global_step": 252028, "epoch": 2831, "lr": 1.2623178115784728e-06}
{"train_loss": 0.03714027628302574, "global_step": 252029, "epoch": 2831, "lr": 1.262188369479994e-06}
{"train_loss": 0.03670094162225723, "global_step": 252030, "epoch": 2831, "lr": 1.2620589339337185e-06}
{"train_loss": 0.025902224704623222, "global_step": 252031, "epoch": 2831, "lr": 1.2619295049396685e-06}
{"train_loss": 0.06601594388484955, "global_step": 252032, "epoch": 2831, "lr": 1.2618000824978549e-06}
{"train_loss": 0.06954912841320038, "global_step": 252033, "epoch": 2831, "lr": 1.2616706666083001e-06}
{"train_loss": 0.04440030828118324, "global_step": 252034, "epoch": 2831, "lr": 1.2615412572710206e-06}
{"train_loss": 0.022208085283637047, "global_step": 252035, "epoch": 2831, "lr": 1.2614118544860276e-06}
{"train_loss": 0.04856423661112785, "global_step": 252036, "epoch": 2831, "lr": 1.2612824582533433e-06}
{"train_loss": 0.0246894471347332, "global_step": 252037, "epoch": 2831, "lr": 1.2611530685729954e-06}
{"train_loss": 0.058801766484975815, "global_step": 252038, "epoch": 2831, "lr": 1.2610236854449787e-06}
{"train_loss": 0.01823507249355316, "global_step": 252039, "epoch": 2831, "lr": 1.260894308869337e-06}
{"train_loss": 0.0409102737903595, "global_step": 252040, "epoch": 2831, "lr": 1.2607649388460596e-06}
{"train_loss": 0.03902881592512131, "global_step": 252041, "epoch": 2831, "lr": 1.2606355753751907e-06}
{"train_loss": 0.0730195865035057, "global_step": 252042, "epoch": 2831, "lr": 1.2605062184567307e-06}
{"train_loss": 0.05541514232754707, "global_step": 252043, "epoch": 2831, "lr": 1.2603768680907013e-06}
{"train_loss": 0.04596051946282387, "global_step": 252044, "epoch": 2831, "lr": 1.2602475242771194e-06}
{"train_loss": 0.05249698460102081, "global_step": 252045, "epoch": 2831, "lr": 1.2601181870160017e-06}
{"train_loss": 0.03272603824734688, "global_step": 252046, "epoch": 2831, "lr": 1.2599888563073703e-06}
{"train_loss": 0.039864572073762004, "global_step": 252047, "epoch": 2831, "lr": 1.259859532151242e-06, "val_loss": 9.575250625610352}
{"train_loss": 0.03384920954704285, "global_step": 252048, "epoch": 2832, "lr": 1.2597302145476274e-06}
{"train_loss": 0.04190797358751297, "global_step": 252049, "epoch": 2832, "lr": 1.2596009034965495e-06}
{"train_loss": 0.1035296842455864, "global_step": 252050, "epoch": 2832, "lr": 1.2594715989980243e-06}
{"train_loss": 0.09884485602378845, "global_step": 252051, "epoch": 2832, "lr": 1.2593423010520633e-06}
{"train_loss": 0.009555737487971783, "global_step": 252052, "epoch": 2832, "lr": 1.2592130096586996e-06}
{"train_loss": 0.030837086960673332, "global_step": 252053, "epoch": 2832, "lr": 1.2590837248179333e-06}
{"train_loss": 0.04385711997747421, "global_step": 252054, "epoch": 2832, "lr": 1.258954446529792e-06}
{"train_loss": 0.027096852660179138, "global_step": 252055, "epoch": 2832, "lr": 1.258825174794287e-06}
{"train_loss": 0.03556274622678757, "global_step": 252056, "epoch": 2832, "lr": 1.2586959096114403e-06}
{"train_loss": 0.05568677932024002, "global_step": 252057, "epoch": 2832, "lr": 1.2585666509812689e-06}
{"train_loss": 0.1180342510342598, "global_step": 252058, "epoch": 2832, "lr": 1.258437398903789e-06}
{"train_loss": 0.03888233006000519, "global_step": 252059, "epoch": 2832, "lr": 1.2583081533790175e-06}
{"train_loss": 0.051619794219732285, "global_step": 252060, "epoch": 2832, "lr": 1.2581789144069767e-06}
{"train_loss": 0.04484647512435913, "global_step": 252061, "epoch": 2832, "lr": 1.258049681987672e-06}
{"train_loss": 0.03318190947175026, "global_step": 252062, "epoch": 2832, "lr": 1.2579204561211367e-06}
{"train_loss": 0.014816578477621078, "global_step": 252063, "epoch": 2832, "lr": 1.2577912368073709e-06}
{"train_loss": 0.02428164705634117, "global_step": 252064, "epoch": 2832, "lr": 1.2576620240464076e-06}
{"train_loss": 0.023820748552680016, "global_step": 252065, "epoch": 2832, "lr": 1.2575328178382528e-06}
{"train_loss": 0.03864230588078499, "global_step": 252066, "epoch": 2832, "lr": 1.2574036181829341e-06}
{"train_loss": 0.06738662719726562, "global_step": 252067, "epoch": 2832, "lr": 1.2572744250804625e-06}
{"train_loss": 0.04460941255092621, "global_step": 252068, "epoch": 2832, "lr": 1.2571452385308546e-06}
{"train_loss": 0.02522924728691578, "global_step": 252069, "epoch": 2832, "lr": 1.2570160585341273e-06}
{"train_loss": 0.016394108533859253, "global_step": 252070, "epoch": 2832, "lr": 1.2568868850903026e-06}
{"train_loss": 0.027227090671658516, "global_step": 252071, "epoch": 2832, "lr": 1.2567577181993917e-06}
{"train_loss": 0.11278238147497177, "global_step": 252072, "epoch": 2832, "lr": 1.2566285578614222e-06}
{"train_loss": 0.05231329798698425, "global_step": 252073, "epoch": 2832, "lr": 1.2564994040764e-06}
{"train_loss": 0.07161931693553925, "global_step": 252074, "epoch": 2832, "lr": 1.256370256844347e-06}
{"train_loss": 0.034169819205999374, "global_step": 252075, "epoch": 2832, "lr": 1.2562411161652854e-06}
{"train_loss": 0.04395071789622307, "global_step": 252076, "epoch": 2832, "lr": 1.2561119820392208e-06}
{"train_loss": 0.04450698569417, "global_step": 252077, "epoch": 2832, "lr": 1.2559828544661866e-06}
{"train_loss": 0.09872130304574966, "global_step": 252078, "epoch": 2832, "lr": 1.2558537334461885e-06}
{"train_loss": 0.06254584342241287, "global_step": 252079, "epoch": 2832, "lr": 1.2557246189792483e-06}
{"train_loss": 0.02401858940720558, "global_step": 252080, "epoch": 2832, "lr": 1.2555955110653717e-06}
{"train_loss": 0.04210986942052841, "global_step": 252081, "epoch": 2832, "lr": 1.2554664097045977e-06}
{"train_loss": 0.07099783420562744, "global_step": 252082, "epoch": 2832, "lr": 1.255337314896926e-06}
{"train_loss": 0.04146644473075867, "global_step": 252083, "epoch": 2832, "lr": 1.2552082266423848e-06}
{"train_loss": 0.06065349280834198, "global_step": 252084, "epoch": 2832, "lr": 1.2550791449409793e-06}
{"train_loss": 0.01822751760482788, "global_step": 252085, "epoch": 2832, "lr": 1.2549500697927429e-06}
{"train_loss": 0.026028359308838844, "global_step": 252086, "epoch": 2832, "lr": 1.2548210011976813e-06}
{"train_loss": 0.039802517741918564, "global_step": 252087, "epoch": 2832, "lr": 1.2546919391558165e-06}
{"train_loss": 0.04478374123573303, "global_step": 252088, "epoch": 2832, "lr": 1.2545628836671597e-06}
{"train_loss": 0.04840424284338951, "global_step": 252089, "epoch": 2832, "lr": 1.254433834731733e-06}
{"train_loss": 0.05203989893198013, "global_step": 252090, "epoch": 2832, "lr": 1.2543047923495589e-06}
{"train_loss": 0.036272622644901276, "global_step": 252091, "epoch": 2832, "lr": 1.2541757565206425e-06}
{"train_loss": 0.022893916815519333, "global_step": 252092, "epoch": 2832, "lr": 1.2540467272450174e-06}
{"train_loss": 0.044109199196100235, "global_step": 252093, "epoch": 2832, "lr": 1.2539177045226835e-06}
{"train_loss": 0.03357924148440361, "global_step": 252094, "epoch": 2832, "lr": 1.2537886883536687e-06}
{"train_loss": 0.026396270841360092, "global_step": 252095, "epoch": 2832, "lr": 1.253659678737984e-06}
{"train_loss": 0.0639127865433693, "global_step": 252096, "epoch": 2832, "lr": 1.2535306756756572e-06}
{"train_loss": 0.02767910622060299, "global_step": 252097, "epoch": 2832, "lr": 1.2534016791666935e-06}
{"train_loss": 0.018762700259685516, "global_step": 252098, "epoch": 2832, "lr": 1.2532726892111213e-06}
{"train_loss": 0.04756834730505943, "global_step": 252099, "epoch": 2832, "lr": 1.2531437058089456e-06}
{"train_loss": 0.04411300644278526, "global_step": 252100, "epoch": 2832, "lr": 1.2530147289602001e-06}
{"train_loss": 0.059716466814279556, "global_step": 252101, "epoch": 2832, "lr": 1.2528857586648846e-06}
{"train_loss": 0.04332438111305237, "global_step": 252102, "epoch": 2832, "lr": 1.2527567949230267e-06}
{"train_loss": 0.03137369453907013, "global_step": 252103, "epoch": 2832, "lr": 1.2526278377346435e-06}
{"train_loss": 0.026024943217635155, "global_step": 252104, "epoch": 2832, "lr": 1.2524988870997512e-06}
{"train_loss": 0.024907156825065613, "global_step": 252105, "epoch": 2832, "lr": 1.2523699430183611e-06}
{"train_loss": 0.07438329607248306, "global_step": 252106, "epoch": 2832, "lr": 1.252241005490501e-06}
{"train_loss": 0.028134383261203766, "global_step": 252107, "epoch": 2832, "lr": 1.2521120745161764e-06}
{"train_loss": 0.060388483107089996, "global_step": 252108, "epoch": 2832, "lr": 1.251983150095415e-06}
{"train_loss": 0.06042416766285896, "global_step": 252109, "epoch": 2832, "lr": 1.2518542322282278e-06}
{"train_loss": 0.039472129195928574, "global_step": 252110, "epoch": 2832, "lr": 1.2517253209146374e-06}
{"train_loss": 0.038347695022821426, "global_step": 252111, "epoch": 2832, "lr": 1.25159641615466e-06}
{"train_loss": 0.03278164565563202, "global_step": 252112, "epoch": 2832, "lr": 1.251467517948307e-06}
{"train_loss": 0.05146820843219757, "global_step": 252113, "epoch": 2832, "lr": 1.2513386262956062e-06}
{"train_loss": 0.04238331690430641, "global_step": 252114, "epoch": 2832, "lr": 1.251209741196563e-06}
{"train_loss": 0.03193452209234238, "global_step": 252115, "epoch": 2832, "lr": 1.2510808626512049e-06}
{"train_loss": 0.04439380392432213, "global_step": 252116, "epoch": 2832, "lr": 1.2509519906595378e-06}
{"train_loss": 0.024330701678991318, "global_step": 252117, "epoch": 2832, "lr": 1.2508231252215951e-06}
{"train_loss": 0.04904170706868172, "global_step": 252118, "epoch": 2832, "lr": 1.2506942663373767e-06}
{"train_loss": 0.030744951218366623, "global_step": 252119, "epoch": 2832, "lr": 1.2505654140069156e-06}
{"train_loss": 0.07403979450464249, "global_step": 252120, "epoch": 2832, "lr": 1.2504365682302121e-06}
{"train_loss": 0.06534808874130249, "global_step": 252121, "epoch": 2832, "lr": 1.2503077290073051e-06}
{"train_loss": 0.050093624740839005, "global_step": 252122, "epoch": 2832, "lr": 1.250178896338189e-06}
{"train_loss": 0.06458967179059982, "global_step": 252123, "epoch": 2832, "lr": 1.2500500702229023e-06}
{"train_loss": 0.03428736329078674, "global_step": 252124, "epoch": 2832, "lr": 1.2499212506614456e-06}
{"train_loss": 0.0664224773645401, "global_step": 252125, "epoch": 2832, "lr": 1.2497924376538462e-06}
{"train_loss": 0.06139937788248062, "global_step": 252126, "epoch": 2832, "lr": 1.2496636312001208e-06}
{"train_loss": 0.019166821613907814, "global_step": 252127, "epoch": 2832, "lr": 1.2495348313002752e-06}
{"train_loss": 0.046809133142232895, "global_step": 252128, "epoch": 2832, "lr": 1.2494060379543425e-06}
{"train_loss": 0.01557967346161604, "global_step": 252129, "epoch": 2832, "lr": 1.2492772511623286e-06}
{"train_loss": 0.04786744713783264, "global_step": 252130, "epoch": 2832, "lr": 1.2491484709242606e-06}
{"train_loss": 0.03904174268245697, "global_step": 252131, "epoch": 2832, "lr": 1.2490196972401446e-06}
{"train_loss": 0.04394887015223503, "global_step": 252132, "epoch": 2832, "lr": 1.2488909301100083e-06}
{"train_loss": 0.026922745630145073, "global_step": 252133, "epoch": 2832, "lr": 1.248762169533857e-06}
{"train_loss": 0.06837937235832214, "global_step": 252134, "epoch": 2832, "lr": 1.2486334155117241e-06}
{"train_loss": 0.032907627522945404, "global_step": 252135, "epoch": 2832, "lr": 1.2485046680436097e-06}
{"train_loss": 0.044896791239133045, "global_step": 252136, "epoch": 2832, "lr": 1.248375927129547e-06, "val_loss": 9.582300186157227}
{"train_loss": 0.021942490711808205, "global_step": 252137, "epoch": 2833, "lr": 1.2482471927695416e-06}
{"train_loss": 0.041111160069704056, "global_step": 252138, "epoch": 2833, "lr": 1.2481184649636213e-06}
{"train_loss": 0.04019208997488022, "global_step": 252139, "epoch": 2833, "lr": 1.2479897437117916e-06}
{"train_loss": 0.07332272082567215, "global_step": 252140, "epoch": 2833, "lr": 1.2478610290140802e-06}
{"train_loss": 0.06527084112167358, "global_step": 252141, "epoch": 2833, "lr": 1.2477323208704927e-06}
{"train_loss": 0.04307795688509941, "global_step": 252142, "epoch": 2833, "lr": 1.247603619281057e-06}
{"train_loss": 0.052302710711956024, "global_step": 252143, "epoch": 2833, "lr": 1.247474924245784e-06}
{"train_loss": 0.03228290006518364, "global_step": 252144, "epoch": 2833, "lr": 1.2473462357647014e-06}
{"train_loss": 0.02668275125324726, "global_step": 252145, "epoch": 2833, "lr": 1.2472175538378095e-06}
{"train_loss": 0.05135272443294525, "global_step": 252146, "epoch": 2833, "lr": 1.2470888784651414e-06}
{"train_loss": 0.02788555808365345, "global_step": 252147, "epoch": 2833, "lr": 1.246960209646708e-06}
{"train_loss": 0.0501285195350647, "global_step": 252148, "epoch": 2833, "lr": 1.246831547382521e-06}
{"train_loss": 0.024106871336698532, "global_step": 252149, "epoch": 2833, "lr": 1.2467028916726075e-06}
{"train_loss": 0.017336200922727585, "global_step": 252150, "epoch": 2833, "lr": 1.2465742425169792e-06}
{"train_loss": 0.04708978161215782, "global_step": 252151, "epoch": 2833, "lr": 1.2464455999156576e-06}
{"train_loss": 0.02045210264623165, "global_step": 252152, "epoch": 2833, "lr": 1.2463169638686546e-06}
{"train_loss": 0.07490415126085281, "global_step": 252153, "epoch": 2833, "lr": 1.2461883343759916e-06}
{"train_loss": 0.04706043750047684, "global_step": 252154, "epoch": 2833, "lr": 1.2460597114376804e-06}
{"train_loss": 0.053747378289699554, "global_step": 252155, "epoch": 2833, "lr": 1.2459310950537483e-06}
{"train_loss": 0.033583104610443115, "global_step": 252156, "epoch": 2833, "lr": 1.2458024852242012e-06}
{"train_loss": 0.03280939161777496, "global_step": 252157, "epoch": 2833, "lr": 1.2456738819490665e-06}
{"train_loss": 0.03638125956058502, "global_step": 252158, "epoch": 2833, "lr": 1.2455452852283557e-06}
{"train_loss": 0.07504269480705261, "global_step": 252159, "epoch": 2833, "lr": 1.2454166950620848e-06}
{"train_loss": 0.025741200894117355, "global_step": 252160, "epoch": 2833, "lr": 1.2452881114502712e-06}
{"train_loss": 0.1143038347363472, "global_step": 252161, "epoch": 2833, "lr": 1.2451595343929423e-06}
{"train_loss": 0.026510747149586678, "global_step": 252162, "epoch": 2833, "lr": 1.2450309638900981e-06}
{"train_loss": 0.029483512043952942, "global_step": 252163, "epoch": 2833, "lr": 1.244902399941772e-06}
{"train_loss": 0.021948376670479774, "global_step": 252164, "epoch": 2833, "lr": 1.244773842547975e-06}
{"train_loss": 0.06317514926195145, "global_step": 252165, "epoch": 2833, "lr": 1.2446452917087182e-06}
{"train_loss": 0.0619308277964592, "global_step": 252166, "epoch": 2833, "lr": 1.2445167474240348e-06}
{"train_loss": 0.04313063248991966, "global_step": 252167, "epoch": 2833, "lr": 1.2443882096939197e-06}
{"train_loss": 0.03500134497880936, "global_step": 252168, "epoch": 2833, "lr": 1.2442596785184112e-06}
{"train_loss": 0.027587084099650383, "global_step": 252169, "epoch": 2833, "lr": 1.2441311538975099e-06}
{"train_loss": 0.03182660788297653, "global_step": 252170, "epoch": 2833, "lr": 1.2440026358312485e-06}
{"train_loss": 0.05623849481344223, "global_step": 252171, "epoch": 2833, "lr": 1.2438741243196328e-06}
{"train_loss": 0.03543233126401901, "global_step": 252172, "epoch": 2833, "lr": 1.2437456193626852e-06}
{"train_loss": 0.10213106125593185, "global_step": 252173, "epoch": 2833, "lr": 1.2436171209604165e-06}
{"train_loss": 0.03147130459547043, "global_step": 252174, "epoch": 2833, "lr": 1.2434886291128544e-06}
{"train_loss": 0.03473011404275894, "global_step": 252175, "epoch": 2833, "lr": 1.2433601438200105e-06}
{"train_loss": 0.05677437409758568, "global_step": 252176, "epoch": 2833, "lr": 1.2432316650819064e-06}
{"train_loss": 0.03631690889596939, "global_step": 252177, "epoch": 2833, "lr": 1.2431031928985482e-06}
{"train_loss": 0.04380609840154648, "global_step": 252178, "epoch": 2833, "lr": 1.2429747272699633e-06}
{"train_loss": 0.01971922069787979, "global_step": 252179, "epoch": 2833, "lr": 1.2428462681961683e-06}
{"train_loss": 0.025896184146404266, "global_step": 252180, "epoch": 2833, "lr": 1.2427178156771746e-06}
{"train_loss": 0.019267069175839424, "global_step": 252181, "epoch": 2833, "lr": 1.2425893697130042e-06}
{"train_loss": 0.01866224966943264, "global_step": 252182, "epoch": 2833, "lr": 1.242460930303674e-06}
{"train_loss": 0.047942496836185455, "global_step": 252183, "epoch": 2833, "lr": 1.2423324974492002e-06}
{"train_loss": 0.04933027923107147, "global_step": 252184, "epoch": 2833, "lr": 1.2422040711495998e-06}
{"train_loss": 0.05836542323231697, "global_step": 252185, "epoch": 2833, "lr": 1.2420756514048949e-06}
{"train_loss": 0.03810602426528931, "global_step": 252186, "epoch": 2833, "lr": 1.2419472382150966e-06}
{"train_loss": 0.0439864918589592, "global_step": 252187, "epoch": 2833, "lr": 1.241818831580227e-06}
{"train_loss": 0.05749486759305, "global_step": 252188, "epoch": 2833, "lr": 1.241690431500292e-06}
{"train_loss": 0.07312510162591934, "global_step": 252189, "epoch": 2833, "lr": 1.2415620379753245e-06}
{"train_loss": 0.021679146215319633, "global_step": 252190, "epoch": 2833, "lr": 1.2414336510053305e-06}
{"train_loss": 0.012573037296533585, "global_step": 252191, "epoch": 2833, "lr": 1.2413052705903372e-06}
{"train_loss": 0.07898711413145065, "global_step": 252192, "epoch": 2833, "lr": 1.2411768967303505e-06}
{"train_loss": 0.056043483316898346, "global_step": 252193, "epoch": 2833, "lr": 1.2410485294253982e-06}
{"train_loss": 0.07289327681064606, "global_step": 252194, "epoch": 2833, "lr": 1.2409201686754856e-06}
{"train_loss": 0.043472982943058014, "global_step": 252195, "epoch": 2833, "lr": 1.2407918144806463e-06}
{"train_loss": 0.08517767488956451, "global_step": 252196, "epoch": 2833, "lr": 1.2406634668408802e-06}
{"train_loss": 0.02262408286333084, "global_step": 252197, "epoch": 2833, "lr": 1.2405351257562203e-06}
{"train_loss": 0.05414107069373131, "global_step": 252198, "epoch": 2833, "lr": 1.240406791226667e-06}
{"train_loss": 0.04430895671248436, "global_step": 252199, "epoch": 2833, "lr": 1.2402784632522535e-06}
{"train_loss": 0.0487079992890358, "global_step": 252200, "epoch": 2833, "lr": 1.2401501418329852e-06}
{"train_loss": 0.016983618959784508, "global_step": 252201, "epoch": 2833, "lr": 1.24002182696889e-06}
{"train_loss": 0.01916036568582058, "global_step": 252202, "epoch": 2833, "lr": 1.2398935186599735e-06}
{"train_loss": 0.04491026699542999, "global_step": 252203, "epoch": 2833, "lr": 1.239765216906269e-06}
{"train_loss": 0.06802623718976974, "global_step": 252204, "epoch": 2833, "lr": 1.2396369217077764e-06}
{"train_loss": 0.06090746074914932, "global_step": 252205, "epoch": 2833, "lr": 1.2395086330645234e-06}
{"train_loss": 0.014392049051821232, "global_step": 252206, "epoch": 2833, "lr": 1.2393803509765212e-06}
{"train_loss": 0.030346332117915154, "global_step": 252207, "epoch": 2833, "lr": 1.2392520754437864e-06}
{"train_loss": 0.04304726794362068, "global_step": 252208, "epoch": 2833, "lr": 1.2391238064663468e-06}
{"train_loss": 0.03849249705672264, "global_step": 252209, "epoch": 2833, "lr": 1.238995544044208e-06}
{"train_loss": 0.05383116379380226, "global_step": 252210, "epoch": 2833, "lr": 1.2388672881773977e-06}
{"train_loss": 0.036402083933353424, "global_step": 252211, "epoch": 2833, "lr": 1.2387390388659214e-06}
{"train_loss": 0.03774624317884445, "global_step": 252212, "epoch": 2833, "lr": 1.238610796109807e-06}
{"train_loss": 0.021118128672242165, "global_step": 252213, "epoch": 2833, "lr": 1.23848255990906e-06}
{"train_loss": 0.032479576766490936, "global_step": 252214, "epoch": 2833, "lr": 1.2383543302637135e-06}
{"train_loss": 0.054691798985004425, "global_step": 252215, "epoch": 2833, "lr": 1.2382261071737678e-06}
{"train_loss": 0.02207709103822708, "global_step": 252216, "epoch": 2833, "lr": 1.238097890639256e-06}
{"train_loss": 0.03195452690124512, "global_step": 252217, "epoch": 2833, "lr": 1.2379696806601838e-06}
{"train_loss": 0.04796493425965309, "global_step": 252218, "epoch": 2833, "lr": 1.2378414772365677e-06}
{"train_loss": 0.0983751192688942, "global_step": 252219, "epoch": 2833, "lr": 1.2377132803684355e-06}
{"train_loss": 0.054910920560359955, "global_step": 252220, "epoch": 2833, "lr": 1.2375850900557984e-06}
{"train_loss": 0.027749886736273766, "global_step": 252221, "epoch": 2833, "lr": 1.2374569062986729e-06}
{"train_loss": 0.037796907126903534, "global_step": 252222, "epoch": 2833, "lr": 1.2373287290970758e-06}
{"train_loss": 0.014665177091956139, "global_step": 252223, "epoch": 2833, "lr": 1.2372005584510293e-06}
{"train_loss": 0.03298739343881607, "global_step": 252224, "epoch": 2833, "lr": 1.237072394360539e-06}
{"train_loss": 0.04298210461111216, "global_step": 252225, "epoch": 2833, "lr": 1.2369442368256379e-06, "val_loss": 9.5360107421875}
{"train_loss": 0.03524576500058174, "global_step": 252226, "epoch": 2834, "lr": 1.2368160858463263e-06}
{"train_loss": 0.02020339109003544, "global_step": 252227, "epoch": 2834, "lr": 1.236687941422643e-06}
{"train_loss": 0.04802393540740013, "global_step": 252228, "epoch": 2834, "lr": 1.2365598035545823e-06}
{"train_loss": 0.027975494042038918, "global_step": 252229, "epoch": 2834, "lr": 1.2364316722421776e-06}
{"train_loss": 0.04130399227142334, "global_step": 252230, "epoch": 2834, "lr": 1.2363035474854346e-06}
{"train_loss": 0.030968250706791878, "global_step": 252231, "epoch": 2834, "lr": 1.2361754292843808e-06}
{"train_loss": 0.05491600185632706, "global_step": 252232, "epoch": 2834, "lr": 1.2360473176390274e-06}
{"train_loss": 0.028848936781287193, "global_step": 252233, "epoch": 2834, "lr": 1.2359192125493967e-06}
{"train_loss": 0.06312856823205948, "global_step": 252234, "epoch": 2834, "lr": 1.2357911140154944e-06}
{"train_loss": 0.0435696616768837, "global_step": 252235, "epoch": 2834, "lr": 1.2356630220373533e-06}
{"train_loss": 0.02552531473338604, "global_step": 252236, "epoch": 2834, "lr": 1.2355349366149793e-06}
{"train_loss": 0.041797272861003876, "global_step": 252237, "epoch": 2834, "lr": 1.2354068577483945e-06}
{"train_loss": 0.023410888388752937, "global_step": 252238, "epoch": 2834, "lr": 1.2352787854376101e-06}
{"train_loss": 0.07349015027284622, "global_step": 252239, "epoch": 2834, "lr": 1.235150719682654e-06}
{"train_loss": 0.07176898419857025, "global_step": 252240, "epoch": 2834, "lr": 1.2350226604835369e-06}
{"train_loss": 0.04852801933884621, "global_step": 252241, "epoch": 2834, "lr": 1.2348946078402756e-06}
{"train_loss": 0.043445125222206116, "global_step": 252242, "epoch": 2834, "lr": 1.2347665617528926e-06}
{"train_loss": 0.06383905559778214, "global_step": 252243, "epoch": 2834, "lr": 1.234638522221393e-06}
{"train_loss": 0.07171878963708878, "global_step": 252244, "epoch": 2834, "lr": 1.2345104892458103e-06}
{"train_loss": 0.07052333652973175, "global_step": 252245, "epoch": 2834, "lr": 1.2343824628261446e-06}
{"train_loss": 0.049214281141757965, "global_step": 252246, "epoch": 2834, "lr": 1.2342544429624292e-06}
{"train_loss": 0.02862357348203659, "global_step": 252247, "epoch": 2834, "lr": 1.234126429654664e-06}
{"train_loss": 0.07584123313426971, "global_step": 252248, "epoch": 2834, "lr": 1.2339984229028877e-06}
{"train_loss": 0.07048351317644119, "global_step": 252249, "epoch": 2834, "lr": 1.2338704227071008e-06}
{"train_loss": 0.07249873876571655, "global_step": 252250, "epoch": 2834, "lr": 1.233742429067325e-06}
{"train_loss": 0.04688606411218643, "global_step": 252251, "epoch": 2834, "lr": 1.233614441983577e-06}
{"train_loss": 0.008975888602435589, "global_step": 252252, "epoch": 2834, "lr": 1.2334864614558795e-06}
{"train_loss": 0.028906110674142838, "global_step": 252253, "epoch": 2834, "lr": 1.2333584874842375e-06}
{"train_loss": 0.04211842268705368, "global_step": 252254, "epoch": 2834, "lr": 1.233230520068679e-06}
{"train_loss": 0.054181184619665146, "global_step": 252255, "epoch": 2834, "lr": 1.233102559209226e-06}
{"train_loss": 0.03582905977964401, "global_step": 252256, "epoch": 2834, "lr": 1.2329746049058787e-06}
{"train_loss": 0.05046321824193001, "global_step": 252257, "epoch": 2834, "lr": 1.2328466571586705e-06}
{"train_loss": 0.03753869608044624, "global_step": 252258, "epoch": 2834, "lr": 1.2327187159676067e-06}
{"train_loss": 0.06809432804584503, "global_step": 252259, "epoch": 2834, "lr": 1.2325907813327154e-06}
{"train_loss": 0.06407823413610458, "global_step": 252260, "epoch": 2834, "lr": 1.2324628532540018e-06}
{"train_loss": 0.053847961127758026, "global_step": 252261, "epoch": 2834, "lr": 1.2323349317314936e-06}
{"train_loss": 0.042873281985521317, "global_step": 252262, "epoch": 2834, "lr": 1.2322070167651967e-06}
{"train_loss": 0.021540703251957893, "global_step": 252263, "epoch": 2834, "lr": 1.2320791083551442e-06}
{"train_loss": 0.02878100983798504, "global_step": 252264, "epoch": 2834, "lr": 1.2319512065013362e-06}
{"train_loss": 0.03409385681152344, "global_step": 252265, "epoch": 2834, "lr": 1.2318233112038058e-06}
{"train_loss": 0.030595438554883003, "global_step": 252266, "epoch": 2834, "lr": 1.2316954224625532e-06}
{"train_loss": 0.04255609214305878, "global_step": 252267, "epoch": 2834, "lr": 1.2315675402776116e-06}
{"train_loss": 0.03289451077580452, "global_step": 252268, "epoch": 2834, "lr": 1.2314396646489868e-06}
{"train_loss": 0.0584665983915329, "global_step": 252269, "epoch": 2834, "lr": 1.2313117955767063e-06}
{"train_loss": 0.019681766629219055, "global_step": 252270, "epoch": 2834, "lr": 1.2311839330607755e-06}
{"train_loss": 0.05855965241789818, "global_step": 252271, "epoch": 2834, "lr": 1.2310560771012225e-06}
{"train_loss": 0.01584816165268421, "global_step": 252272, "epoch": 2834, "lr": 1.2309282276980582e-06}
{"train_loss": 0.07326529920101166, "global_step": 252273, "epoch": 2834, "lr": 1.230800384851294e-06}
{"train_loss": 0.05844094976782799, "global_step": 252274, "epoch": 2834, "lr": 1.2306725485609682e-06}
{"train_loss": 0.04271678999066353, "global_step": 252275, "epoch": 2834, "lr": 1.2305447188270757e-06}
{"train_loss": 0.028998762369155884, "global_step": 252276, "epoch": 2834, "lr": 1.2304168956496443e-06}
{"train_loss": 0.03732096403837204, "global_step": 252277, "epoch": 2834, "lr": 1.2302890790286848e-06}
{"train_loss": 0.03309763967990875, "global_step": 252278, "epoch": 2834, "lr": 1.2301612689642195e-06}
{"train_loss": 0.08687830716371536, "global_step": 252279, "epoch": 2834, "lr": 1.2300334654562706e-06}
{"train_loss": 0.06508230417966843, "global_step": 252280, "epoch": 2834, "lr": 1.2299056685048493e-06}
{"train_loss": 0.018368493765592575, "global_step": 252281, "epoch": 2834, "lr": 1.2297778781099668e-06}
{"train_loss": 0.04142114520072937, "global_step": 252282, "epoch": 2834, "lr": 1.2296500942716505e-06}
{"train_loss": 0.07105731219053268, "global_step": 252283, "epoch": 2834, "lr": 1.2295223169899061e-06}
{"train_loss": 0.05830007791519165, "global_step": 252284, "epoch": 2834, "lr": 1.2293945462647672e-06}
{"train_loss": 0.032169096171855927, "global_step": 252285, "epoch": 2834, "lr": 1.2292667820962333e-06}
{"train_loss": 0.03576676547527313, "global_step": 252286, "epoch": 2834, "lr": 1.2291390244843382e-06}
{"train_loss": 0.03611014783382416, "global_step": 252287, "epoch": 2834, "lr": 1.2290112734290816e-06}
{"train_loss": 0.12732543051242828, "global_step": 252288, "epoch": 2834, "lr": 1.2288835289304968e-06}
{"train_loss": 0.016539299860596657, "global_step": 252289, "epoch": 2834, "lr": 1.228755790988595e-06}
{"train_loss": 0.017539290711283684, "global_step": 252290, "epoch": 2834, "lr": 1.2286280596033873e-06}
{"train_loss": 0.03461701422929764, "global_step": 252291, "epoch": 2834, "lr": 1.2285003347749014e-06}
{"train_loss": 0.03813272714614868, "global_step": 252292, "epoch": 2834, "lr": 1.2283726165031428e-06}
{"train_loss": 0.07371082901954651, "global_step": 252293, "epoch": 2834, "lr": 1.2282449047881394e-06}
{"train_loss": 0.03484994173049927, "global_step": 252294, "epoch": 2834, "lr": 1.2281171996299024e-06}
{"train_loss": 0.04294182360172272, "global_step": 252295, "epoch": 2834, "lr": 1.2279895010284536e-06}
{"train_loss": 0.0715557411313057, "global_step": 252296, "epoch": 2834, "lr": 1.2278618089838045e-06}
{"train_loss": 0.08407150208950043, "global_step": 252297, "epoch": 2834, "lr": 1.227734123495977e-06}
{"train_loss": 0.047391463071107864, "global_step": 252298, "epoch": 2834, "lr": 1.2276064445649826e-06}
{"train_loss": 0.0636180117726326, "global_step": 252299, "epoch": 2834, "lr": 1.2274787721908488e-06}
{"train_loss": 0.027578646317124367, "global_step": 252300, "epoch": 2834, "lr": 1.2273511063735754e-06}
{"train_loss": 0.049738749861717224, "global_step": 252301, "epoch": 2834, "lr": 1.2272234471132016e-06}
{"train_loss": 0.03474726900458336, "global_step": 252302, "epoch": 2834, "lr": 1.2270957944097273e-06}
{"train_loss": 0.038956671953201294, "global_step": 252303, "epoch": 2834, "lr": 1.2269681482631746e-06}
{"train_loss": 0.053512390702962875, "global_step": 252304, "epoch": 2834, "lr": 1.2268405086735657e-06}
{"train_loss": 0.04598621279001236, "global_step": 252305, "epoch": 2834, "lr": 1.2267128756409118e-06}
{"train_loss": 0.08525488525629044, "global_step": 252306, "epoch": 2834, "lr": 1.2265852491652297e-06}
{"train_loss": 0.042768921703100204, "global_step": 252307, "epoch": 2834, "lr": 1.2264576292465412e-06}
{"train_loss": 0.03203475475311279, "global_step": 252308, "epoch": 2834, "lr": 1.2263300158848578e-06}
{"train_loss": 0.07596044987440109, "global_step": 252309, "epoch": 2834, "lr": 1.2262024090802015e-06}
{"train_loss": 0.029714886099100113, "global_step": 252310, "epoch": 2834, "lr": 1.226074808832589e-06}
{"train_loss": 0.025124607607722282, "global_step": 252311, "epoch": 2834, "lr": 1.2259472151420314e-06}
{"train_loss": 0.05388256534934044, "global_step": 252312, "epoch": 2834, "lr": 1.2258196280085565e-06}
{"train_loss": 0.0511951707303524, "global_step": 252313, "epoch": 2834, "lr": 1.2256920474321753e-06}
{"train_loss": 0.046719774760808165, "global_step": 252314, "epoch": 2834, "lr": 1.2255644734129046e-06, "val_loss": 9.67080020904541}
{"train_loss": 0.043168436735868454, "global_step": 252315, "epoch": 2835, "lr": 1.2254369059507664e-06}
{"train_loss": 0.08066654950380325, "global_step": 252316, "epoch": 2835, "lr": 1.2253093450457665e-06}
{"train_loss": 0.07018522173166275, "global_step": 252317, "epoch": 2835, "lr": 1.2251817906979324e-06}
{"train_loss": 0.04409230872988701, "global_step": 252318, "epoch": 2835, "lr": 1.2250542429072754e-06}
{"train_loss": 0.02390149049460888, "global_step": 252319, "epoch": 2835, "lr": 1.224926701673823e-06}
{"train_loss": 0.1082143560051918, "global_step": 252320, "epoch": 2835, "lr": 1.224799166997581e-06}
{"train_loss": 0.04758531227707863, "global_step": 252321, "epoch": 2835, "lr": 1.2246716388785662e-06}
{"train_loss": 0.02936776727437973, "global_step": 252322, "epoch": 2835, "lr": 1.224544117316806e-06}
{"train_loss": 0.024101993069052696, "global_step": 252323, "epoch": 2835, "lr": 1.224416602312306e-06}
{"train_loss": 0.05914195999503136, "global_step": 252324, "epoch": 2835, "lr": 1.2242890938650885e-06}
{"train_loss": 0.01728336326777935, "global_step": 252325, "epoch": 2835, "lr": 1.2241615919751704e-06}
{"train_loss": 0.0403360091149807, "global_step": 252326, "epoch": 2835, "lr": 1.224034096642579e-06}
{"train_loss": 0.034735191613435745, "global_step": 252327, "epoch": 2835, "lr": 1.223906607867309e-06}
{"train_loss": 0.05807722732424736, "global_step": 252328, "epoch": 2835, "lr": 1.2237791256493935e-06}
{"train_loss": 0.050469037145376205, "global_step": 252329, "epoch": 2835, "lr": 1.2236516499888494e-06}
{"train_loss": 0.048744551837444305, "global_step": 252330, "epoch": 2835, "lr": 1.2235241808856878e-06}
{"train_loss": 0.045361705124378204, "global_step": 252331, "epoch": 2835, "lr": 1.2233967183399363e-06}
{"train_loss": 0.017268899828195572, "global_step": 252332, "epoch": 2835, "lr": 1.2232692623515952e-06}
{"train_loss": 0.046034444123506546, "global_step": 252333, "epoch": 2835, "lr": 1.2231418129206973e-06}
{"train_loss": 0.04601301625370979, "global_step": 252334, "epoch": 2835, "lr": 1.2230143700472486e-06}
{"train_loss": 0.005303189158439636, "global_step": 252335, "epoch": 2835, "lr": 1.2228869337312765e-06}
{"train_loss": 0.036451760679483414, "global_step": 252336, "epoch": 2835, "lr": 1.2227595039727868e-06}
{"train_loss": 0.029687020927667618, "global_step": 252337, "epoch": 2835, "lr": 1.2226320807718129e-06}
{"train_loss": 0.05693735554814339, "global_step": 252338, "epoch": 2835, "lr": 1.222504664128349e-06}
{"train_loss": 0.031087614595890045, "global_step": 252339, "epoch": 2835, "lr": 1.2223772540424338e-06}
{"train_loss": 0.026367852464318275, "global_step": 252340, "epoch": 2835, "lr": 1.2222498505140678e-06}
{"train_loss": 0.037573933601379395, "global_step": 252341, "epoch": 2835, "lr": 1.2221224535432841e-06}
{"train_loss": 0.04318671301007271, "global_step": 252342, "epoch": 2835, "lr": 1.2219950631300826e-06}
{"train_loss": 0.02665789984166622, "global_step": 252343, "epoch": 2835, "lr": 1.2218676792744966e-06}
{"train_loss": 0.03757798299193382, "global_step": 252344, "epoch": 2835, "lr": 1.2217403019765316e-06}
{"train_loss": 0.019930727779865265, "global_step": 252345, "epoch": 2835, "lr": 1.22161293123621e-06}
{"train_loss": 0.016124602407217026, "global_step": 252346, "epoch": 2835, "lr": 1.221485567053554e-06}
{"train_loss": 0.04460826888680458, "global_step": 252347, "epoch": 2835, "lr": 1.2213582094285635e-06}
{"train_loss": 0.03982260078191757, "global_step": 252348, "epoch": 2835, "lr": 1.2212308583612775e-06}
{"train_loss": 0.031957898288965225, "global_step": 252349, "epoch": 2835, "lr": 1.2211035138516958e-06}
{"train_loss": 0.04861484840512276, "global_step": 252350, "epoch": 2835, "lr": 1.2209761758998462e-06}
{"train_loss": 0.012230702675879002, "global_step": 252351, "epoch": 2835, "lr": 1.2208488445057398e-06}
{"train_loss": 0.043519627302885056, "global_step": 252352, "epoch": 2835, "lr": 1.2207215196693987e-06}
{"train_loss": 0.05217155069112778, "global_step": 252353, "epoch": 2835, "lr": 1.2205942013908344e-06}
{"train_loss": 0.03144479915499687, "global_step": 252354, "epoch": 2835, "lr": 1.2204668896700633e-06}
{"train_loss": 0.08109438419342041, "global_step": 252355, "epoch": 2835, "lr": 1.2203395845071131e-06}
{"train_loss": 0.03987647220492363, "global_step": 252356, "epoch": 2835, "lr": 1.2202122859019893e-06}
{"train_loss": 0.0402996651828289, "global_step": 252357, "epoch": 2835, "lr": 1.2200849938547143e-06}
{"train_loss": 0.0369744710624218, "global_step": 252358, "epoch": 2835, "lr": 1.2199577083653101e-06}
{"train_loss": 0.01939850114285946, "global_step": 252359, "epoch": 2835, "lr": 1.219830429433777e-06}
{"train_loss": 0.05416130647063255, "global_step": 252360, "epoch": 2835, "lr": 1.2197031570601536e-06}
{"train_loss": 0.07657473534345627, "global_step": 252361, "epoch": 2835, "lr": 1.2195758912444345e-06}
{"train_loss": 0.04259719327092171, "global_step": 252362, "epoch": 2835, "lr": 1.2194486319866583e-06}
{"train_loss": 0.0814225897192955, "global_step": 252363, "epoch": 2835, "lr": 1.2193213792868308e-06}
{"train_loss": 0.015094184316694736, "global_step": 252364, "epoch": 2835, "lr": 1.2191941331449685e-06}
{"train_loss": 0.08360585570335388, "global_step": 252365, "epoch": 2835, "lr": 1.2190668935610939e-06}
{"train_loss": 0.044886767864227295, "global_step": 252366, "epoch": 2835, "lr": 1.2189396605352176e-06}
{"train_loss": 0.04848537594079971, "global_step": 252367, "epoch": 2835, "lr": 1.2188124340673623e-06}
{"train_loss": 0.06219089403748512, "global_step": 252368, "epoch": 2835, "lr": 1.2186852141575445e-06}
{"train_loss": 0.027831465005874634, "global_step": 252369, "epoch": 2835, "lr": 1.2185580008057807e-06}
{"train_loss": 0.05071509629487991, "global_step": 252370, "epoch": 2835, "lr": 1.218430794012082e-06}
{"train_loss": 0.08611585199832916, "global_step": 252371, "epoch": 2835, "lr": 1.2183035937764764e-06}
{"train_loss": 0.029856981709599495, "global_step": 252372, "epoch": 2835, "lr": 1.2181764000989692e-06}
{"train_loss": 0.036630358546972275, "global_step": 252373, "epoch": 2835, "lr": 1.2180492129795883e-06}
{"train_loss": 0.06647977977991104, "global_step": 252374, "epoch": 2835, "lr": 1.2179220324183448e-06}
{"train_loss": 0.060247816145420074, "global_step": 252375, "epoch": 2835, "lr": 1.217794858415261e-06}
{"train_loss": 0.030636753886938095, "global_step": 252376, "epoch": 2835, "lr": 1.2176676909703422e-06}
{"train_loss": 0.032421648502349854, "global_step": 252377, "epoch": 2835, "lr": 1.2175405300836217e-06}
{"train_loss": 0.06309813261032104, "global_step": 252378, "epoch": 2835, "lr": 1.2174133757551e-06}
{"train_loss": 0.032154373824596405, "global_step": 252379, "epoch": 2835, "lr": 1.2172862279848097e-06}
{"train_loss": 0.046852048486471176, "global_step": 252380, "epoch": 2835, "lr": 1.2171590867727566e-06}
{"train_loss": 0.01582331396639347, "global_step": 252381, "epoch": 2835, "lr": 1.2170319521189633e-06}
{"train_loss": 0.039808083325624466, "global_step": 252382, "epoch": 2835, "lr": 1.216904824023446e-06}
{"train_loss": 0.05224362015724182, "global_step": 252383, "epoch": 2835, "lr": 1.216777702486216e-06}
{"train_loss": 0.03958113491535187, "global_step": 252384, "epoch": 2835, "lr": 1.2166505875073064e-06}
{"train_loss": 0.07139233499765396, "global_step": 252385, "epoch": 2835, "lr": 1.2165234790867119e-06}
{"train_loss": 0.012181204743683338, "global_step": 252386, "epoch": 2835, "lr": 1.2163963772244714e-06}
{"train_loss": 0.04507405683398247, "global_step": 252387, "epoch": 2835, "lr": 1.2162692819205846e-06}
{"train_loss": 0.03743259236216545, "global_step": 252388, "epoch": 2835, "lr": 1.2161421931750793e-06}
{"train_loss": 0.010067330673336983, "global_step": 252389, "epoch": 2835, "lr": 1.216015110987967e-06}
{"train_loss": 0.07004670798778534, "global_step": 252390, "epoch": 2835, "lr": 1.2158880353592694e-06}
{"train_loss": 0.04396849125623703, "global_step": 252391, "epoch": 2835, "lr": 1.2157609662890036e-06}
{"train_loss": 0.022261904552578926, "global_step": 252392, "epoch": 2835, "lr": 1.2156339037771747e-06}
{"train_loss": 0.09755894541740417, "global_step": 252393, "epoch": 2835, "lr": 1.2155068478238162e-06}
{"train_loss": 0.053084999322891235, "global_step": 252394, "epoch": 2835, "lr": 1.2153797984289394e-06}
{"train_loss": 0.034965261816978455, "global_step": 252395, "epoch": 2835, "lr": 1.2152527555925553e-06}
{"train_loss": 0.025727272033691406, "global_step": 252396, "epoch": 2835, "lr": 1.2151257193146914e-06}
{"train_loss": 0.03181946277618408, "global_step": 252397, "epoch": 2835, "lr": 1.2149986895953536e-06}
{"train_loss": 0.024583525955677032, "global_step": 252398, "epoch": 2835, "lr": 1.2148716664345693e-06}
{"train_loss": 0.07326002418994904, "global_step": 252399, "epoch": 2835, "lr": 1.2147446498323501e-06}
{"train_loss": 0.035674698650836945, "global_step": 252400, "epoch": 2835, "lr": 1.2146176397887066e-06}
{"train_loss": 0.0542585626244545, "global_step": 252401, "epoch": 2835, "lr": 1.2144906363036724e-06}
{"train_loss": 0.04206693917512894, "global_step": 252402, "epoch": 2835, "lr": 1.2143636393772472e-06}
{"train_loss": 0.04341469674758362, "global_step": 252403, "epoch": 2835, "lr": 1.2142366490094648e-06, "val_loss": 9.624924659729004, "train_action_mse_error": 15.414243698120117}
{"train_loss": 0.022622426971793175, "global_step": 252404, "epoch": 2836, "lr": 1.2141096652003302e-06}
{"train_loss": 0.03653034195303917, "global_step": 252405, "epoch": 2836, "lr": 1.213982687949866e-06}
{"train_loss": 0.04165426641702652, "global_step": 252406, "epoch": 2836, "lr": 1.2138557172580777e-06}
{"train_loss": 0.06637092679738998, "global_step": 252407, "epoch": 2836, "lr": 1.2137287531250042e-06}
{"train_loss": 0.049618493765592575, "global_step": 252408, "epoch": 2836, "lr": 1.2136017955506396e-06}
{"train_loss": 0.028745975345373154, "global_step": 252409, "epoch": 2836, "lr": 1.213474844535023e-06}
{"train_loss": 0.035930801182985306, "global_step": 252410, "epoch": 2836, "lr": 1.213347900078149e-06}
{"train_loss": 0.03131728991866112, "global_step": 252411, "epoch": 2836, "lr": 1.213220962180056e-06}
{"train_loss": 0.08132748305797577, "global_step": 252412, "epoch": 2836, "lr": 1.213094030840739e-06}
{"train_loss": 0.021019984036684036, "global_step": 252413, "epoch": 2836, "lr": 1.2129671060602365e-06}
{"train_loss": 0.028683656826615334, "global_step": 252414, "epoch": 2836, "lr": 1.2128401878385542e-06}
{"train_loss": 0.07309992611408234, "global_step": 252415, "epoch": 2836, "lr": 1.2127132761757087e-06}
{"train_loss": 0.029525715857744217, "global_step": 252416, "epoch": 2836, "lr": 1.2125863710717166e-06}
{"train_loss": 0.047711607068777084, "global_step": 252417, "epoch": 2836, "lr": 1.2124594725266004e-06}
{"train_loss": 0.07601027935743332, "global_step": 252418, "epoch": 2836, "lr": 1.2123325805403761e-06}
{"train_loss": 0.038401514291763306, "global_step": 252419, "epoch": 2836, "lr": 1.2122056951130556e-06}
{"train_loss": 0.025141337886452675, "global_step": 252420, "epoch": 2836, "lr": 1.2120788162446661e-06}
{"train_loss": 0.05540992319583893, "global_step": 252421, "epoch": 2836, "lr": 1.2119519439352134e-06}
{"train_loss": 0.06367891281843185, "global_step": 252422, "epoch": 2836, "lr": 1.2118250781847196e-06}
{"train_loss": 0.03059527836740017, "global_step": 252423, "epoch": 2836, "lr": 1.2116982189932013e-06}
{"train_loss": 0.083549864590168, "global_step": 252424, "epoch": 2836, "lr": 1.2115713663606754e-06}
{"train_loss": 0.051963191479444504, "global_step": 252425, "epoch": 2836, "lr": 1.2114445202871582e-06}
{"train_loss": 0.06817112118005753, "global_step": 252426, "epoch": 2836, "lr": 1.2113176807726668e-06}
{"train_loss": 0.024034637957811356, "global_step": 252427, "epoch": 2836, "lr": 1.211190847817223e-06}
{"train_loss": 0.05723128840327263, "global_step": 252428, "epoch": 2836, "lr": 1.211064021420838e-06}
{"train_loss": 0.01145228836685419, "global_step": 252429, "epoch": 2836, "lr": 1.2109372015835285e-06}
{"train_loss": 0.05231792479753494, "global_step": 252430, "epoch": 2836, "lr": 1.2108103883053167e-06}
{"train_loss": 0.02744384855031967, "global_step": 252431, "epoch": 2836, "lr": 1.2106835815862194e-06}
{"train_loss": 0.0444885790348053, "global_step": 252432, "epoch": 2836, "lr": 1.2105567814262474e-06}
{"train_loss": 0.046763818711042404, "global_step": 252433, "epoch": 2836, "lr": 1.2104299878254234e-06}
{"train_loss": 0.026437800377607346, "global_step": 252434, "epoch": 2836, "lr": 1.2103032007837577e-06}
{"train_loss": 0.07944408804178238, "global_step": 252435, "epoch": 2836, "lr": 1.2101764203012734e-06}
{"train_loss": 0.07937628775835037, "global_step": 252436, "epoch": 2836, "lr": 1.2100496463779865e-06}
{"train_loss": 0.07092652469873428, "global_step": 252437, "epoch": 2836, "lr": 1.2099228790139194e-06}
{"train_loss": 0.04225393012166023, "global_step": 252438, "epoch": 2836, "lr": 1.2097961182090777e-06}
{"train_loss": 0.05442405492067337, "global_step": 252439, "epoch": 2836, "lr": 1.2096693639634894e-06}
{"train_loss": 0.04539842903614044, "global_step": 252440, "epoch": 2836, "lr": 1.2095426162771594e-06}
{"train_loss": 0.04312293604016304, "global_step": 252441, "epoch": 2836, "lr": 1.2094158751501162e-06}
{"train_loss": 0.07990457117557526, "global_step": 252442, "epoch": 2836, "lr": 1.2092891405823703e-06}
{"train_loss": 0.06573987752199173, "global_step": 252443, "epoch": 2836, "lr": 1.2091624125739443e-06}
{"train_loss": 0.047396160662174225, "global_step": 252444, "epoch": 2836, "lr": 1.2090356911248435e-06}
{"train_loss": 0.04662715643644333, "global_step": 252445, "epoch": 2836, "lr": 1.2089089762351013e-06}
{"train_loss": 0.023347023874521255, "global_step": 252446, "epoch": 2836, "lr": 1.2087822679047234e-06}
{"train_loss": 0.03997331112623215, "global_step": 252447, "epoch": 2836, "lr": 1.2086555661337318e-06}
{"train_loss": 0.02082393877208233, "global_step": 252448, "epoch": 2836, "lr": 1.2085288709221375e-06}
{"train_loss": 0.025196777656674385, "global_step": 252449, "epoch": 2836, "lr": 1.2084021822699687e-06}
{"train_loss": 0.010471832938492298, "global_step": 252450, "epoch": 2836, "lr": 1.2082755001772306e-06}
{"train_loss": 0.035824209451675415, "global_step": 252451, "epoch": 2836, "lr": 1.2081488246439454e-06}
{"train_loss": 0.03328796103596687, "global_step": 252452, "epoch": 2836, "lr": 1.2080221556701298e-06}
{"train_loss": 0.0332278236746788, "global_step": 252453, "epoch": 2836, "lr": 1.2078954932558062e-06}
{"train_loss": 0.05277741700410843, "global_step": 252454, "epoch": 2836, "lr": 1.20776883740098e-06}
{"train_loss": 0.05048473924398422, "global_step": 252455, "epoch": 2836, "lr": 1.2076421881056731e-06}
{"train_loss": 0.05121328681707382, "global_step": 252456, "epoch": 2836, "lr": 1.2075155453699082e-06}
{"train_loss": 0.06697992235422134, "global_step": 252457, "epoch": 2836, "lr": 1.2073889091936963e-06}
{"train_loss": 0.04368508234620094, "global_step": 252458, "epoch": 2836, "lr": 1.2072622795770593e-06}
{"train_loss": 0.01224514376372099, "global_step": 252459, "epoch": 2836, "lr": 1.2071356565200088e-06}
{"train_loss": 0.04052294045686722, "global_step": 252460, "epoch": 2836, "lr": 1.2070090400225665e-06}
{"train_loss": 0.024219999089837074, "global_step": 252461, "epoch": 2836, "lr": 1.2068824300847436e-06}
{"train_loss": 0.030603857710957527, "global_step": 252462, "epoch": 2836, "lr": 1.2067558267065626e-06}
{"train_loss": 0.03994908556342125, "global_step": 252463, "epoch": 2836, "lr": 1.20662922988804e-06}
{"train_loss": 0.056326329708099365, "global_step": 252464, "epoch": 2836, "lr": 1.2065026396291923e-06}
{"train_loss": 0.048520077019929886, "global_step": 252465, "epoch": 2836, "lr": 1.2063760559300308e-06}
{"train_loss": 0.04364652559161186, "global_step": 252466, "epoch": 2836, "lr": 1.2062494787905777e-06}
{"train_loss": 0.0415361225605011, "global_step": 252467, "epoch": 2836, "lr": 1.2061229082108494e-06}
{"train_loss": 0.013467424549162388, "global_step": 252468, "epoch": 2836, "lr": 1.2059963441908683e-06}
{"train_loss": 0.04610734060406685, "global_step": 252469, "epoch": 2836, "lr": 1.2058697867306456e-06}
{"train_loss": 0.06959298253059387, "global_step": 252470, "epoch": 2836, "lr": 1.205743235830198e-06}
{"train_loss": 0.05548502504825592, "global_step": 252471, "epoch": 2836, "lr": 1.2056166914895417e-06}
{"train_loss": 0.03465615585446358, "global_step": 252472, "epoch": 2836, "lr": 1.2054901537086937e-06}
{"train_loss": 0.058895040303468704, "global_step": 252473, "epoch": 2836, "lr": 1.2053636224876764e-06}
{"train_loss": 0.04387141391634941, "global_step": 252474, "epoch": 2836, "lr": 1.2052370978265004e-06}
{"train_loss": 0.037018805742263794, "global_step": 252475, "epoch": 2836, "lr": 1.2051105797251939e-06}
{"train_loss": 0.026278797537088394, "global_step": 252476, "epoch": 2836, "lr": 1.2049840681837565e-06}
{"train_loss": 0.0407305508852005, "global_step": 252477, "epoch": 2836, "lr": 1.2048575632022164e-06}
{"train_loss": 0.04691169783473015, "global_step": 252478, "epoch": 2836, "lr": 1.2047310647805844e-06}
{"train_loss": 0.03848135098814964, "global_step": 252479, "epoch": 2836, "lr": 1.2046045729188882e-06}
{"train_loss": 0.05887313559651375, "global_step": 252480, "epoch": 2836, "lr": 1.2044780876171335e-06}
{"train_loss": 0.04781123623251915, "global_step": 252481, "epoch": 2836, "lr": 1.204351608875348e-06}
{"train_loss": 0.06361741572618484, "global_step": 252482, "epoch": 2836, "lr": 1.2042251366935376e-06}
{"train_loss": 0.06555669754743576, "global_step": 252483, "epoch": 2836, "lr": 1.2040986710717295e-06}
{"train_loss": 0.04374656081199646, "global_step": 252484, "epoch": 2836, "lr": 1.2039722120099296e-06}
{"train_loss": 0.021995874121785164, "global_step": 252485, "epoch": 2836, "lr": 1.2038457595081654e-06}
{"train_loss": 0.07069622725248337, "global_step": 252486, "epoch": 2836, "lr": 1.2037193135664426e-06}
{"train_loss": 0.04140692204236984, "global_step": 252487, "epoch": 2836, "lr": 1.2035928741847946e-06}
{"train_loss": 0.046622905880212784, "global_step": 252488, "epoch": 2836, "lr": 1.2034664413632212e-06}
{"train_loss": 0.09611731767654419, "global_step": 252489, "epoch": 2836, "lr": 1.2033400151017504e-06}
{"train_loss": 0.06661927700042725, "global_step": 252490, "epoch": 2836, "lr": 1.2032135954003932e-06}
{"train_loss": 0.06308681517839432, "global_step": 252491, "epoch": 2836, "lr": 1.203087182259166e-06}
{"train_loss": 0.045943399828471496, "global_step": 252492, "epoch": 2836, "lr": 1.202960775678097e-06, "val_loss": 9.621164321899414}
{"train_loss": 0.032281529158353806, "global_step": 252493, "epoch": 2837, "lr": 1.2028343756571913e-06}
{"train_loss": 0.06960852444171906, "global_step": 252494, "epoch": 2837, "lr": 1.2027079821964715e-06}
{"train_loss": 0.03737274557352066, "global_step": 252495, "epoch": 2837, "lr": 1.2025815952959541e-06}
{"train_loss": 0.07100135087966919, "global_step": 252496, "epoch": 2837, "lr": 1.20245521495565e-06}
{"train_loss": 0.04721798747777939, "global_step": 252497, "epoch": 2837, "lr": 1.2023288411755817e-06}
{"train_loss": 0.0846424326300621, "global_step": 252498, "epoch": 2837, "lr": 1.2022024739557713e-06}
{"train_loss": 0.028041645884513855, "global_step": 252499, "epoch": 2837, "lr": 1.2020761132962243e-06}
{"train_loss": 0.014700614847242832, "global_step": 252500, "epoch": 2837, "lr": 1.2019497591969687e-06}
{"train_loss": 0.0375087596476078, "global_step": 252501, "epoch": 2837, "lr": 1.2018234116580097e-06}
{"train_loss": 0.033845845609903336, "global_step": 252502, "epoch": 2837, "lr": 1.201697070679375e-06}
{"train_loss": 0.021914226934313774, "global_step": 252503, "epoch": 2837, "lr": 1.201570736261076e-06}
{"train_loss": 0.07578088343143463, "global_step": 252504, "epoch": 2837, "lr": 1.2014444084031352e-06}
{"train_loss": 0.034554269164800644, "global_step": 252505, "epoch": 2837, "lr": 1.2013180871055574e-06}
{"train_loss": 0.061418190598487854, "global_step": 252506, "epoch": 2837, "lr": 1.2011917723683764e-06}
{"train_loss": 0.051147688180208206, "global_step": 252507, "epoch": 2837, "lr": 1.2010654641915975e-06}
{"train_loss": 0.040600087493658066, "global_step": 252508, "epoch": 2837, "lr": 1.2009391625752376e-06}
{"train_loss": 0.047029681503772736, "global_step": 252509, "epoch": 2837, "lr": 1.2008128675193186e-06}
{"train_loss": 0.11006657034158707, "global_step": 252510, "epoch": 2837, "lr": 1.200686579023852e-06}
{"train_loss": 0.03274866193532944, "global_step": 252511, "epoch": 2837, "lr": 1.2005602970888653e-06}
{"train_loss": 0.05469601973891258, "global_step": 252512, "epoch": 2837, "lr": 1.2004340217143584e-06}
{"train_loss": 0.01378617063164711, "global_step": 252513, "epoch": 2837, "lr": 1.2003077529003703e-06}
{"train_loss": 0.04498514533042908, "global_step": 252514, "epoch": 2837, "lr": 1.2001814906468955e-06}
{"train_loss": 0.03710152953863144, "global_step": 252515, "epoch": 2837, "lr": 1.2000552349539672e-06}
{"train_loss": 0.08308283984661102, "global_step": 252516, "epoch": 2837, "lr": 1.1999289858215967e-06}
{"train_loss": 0.05411522090435028, "global_step": 252517, "epoch": 2837, "lr": 1.1998027432498004e-06}
{"train_loss": 0.04660838469862938, "global_step": 252518, "epoch": 2837, "lr": 1.1996765072385951e-06}
{"train_loss": 0.01606752537190914, "global_step": 252519, "epoch": 2837, "lr": 1.1995502777879974e-06}
{"train_loss": 0.07761019468307495, "global_step": 252520, "epoch": 2837, "lr": 1.199424054898024e-06}
{"train_loss": 0.025642460212111473, "global_step": 252521, "epoch": 2837, "lr": 1.199297838568697e-06}
{"train_loss": 0.06590604037046432, "global_step": 252522, "epoch": 2837, "lr": 1.1991716288000276e-06}
{"train_loss": 0.08673780411481857, "global_step": 252523, "epoch": 2837, "lr": 1.1990454255920381e-06}
{"train_loss": 0.07149764895439148, "global_step": 252524, "epoch": 2837, "lr": 1.1989192289447392e-06}
{"train_loss": 0.022566234692931175, "global_step": 252525, "epoch": 2837, "lr": 1.1987930388581537e-06}
{"train_loss": 0.10770631581544876, "global_step": 252526, "epoch": 2837, "lr": 1.198666855332292e-06}
{"train_loss": 0.06275404244661331, "global_step": 252527, "epoch": 2837, "lr": 1.1985406783671716e-06}
{"train_loss": 0.03848513960838318, "global_step": 252528, "epoch": 2837, "lr": 1.1984145079628194e-06}
{"train_loss": 0.04494326561689377, "global_step": 252529, "epoch": 2837, "lr": 1.1982883441192416e-06}
{"train_loss": 0.0707506537437439, "global_step": 252530, "epoch": 2837, "lr": 1.1981621868364601e-06}
{"train_loss": 0.04564082995057106, "global_step": 252531, "epoch": 2837, "lr": 1.1980360361144916e-06}
{"train_loss": 0.04025750979781151, "global_step": 252532, "epoch": 2837, "lr": 1.1979098919533527e-06}
{"train_loss": 0.02978079579770565, "global_step": 252533, "epoch": 2837, "lr": 1.1977837543530602e-06}
{"train_loss": 0.0648859441280365, "global_step": 252534, "epoch": 2837, "lr": 1.1976576233136306e-06}
{"train_loss": 0.028772857040166855, "global_step": 252535, "epoch": 2837, "lr": 1.1975314988350806e-06}
{"train_loss": 0.04932396113872528, "global_step": 252536, "epoch": 2837, "lr": 1.1974053809174269e-06}
{"train_loss": 0.021690305322408676, "global_step": 252537, "epoch": 2837, "lr": 1.197279269560686e-06}
{"train_loss": 0.048184748739004135, "global_step": 252538, "epoch": 2837, "lr": 1.1971531647648804e-06}
{"train_loss": 0.05845476686954498, "global_step": 252539, "epoch": 2837, "lr": 1.197027066530021e-06}
{"train_loss": 0.07769732922315598, "global_step": 252540, "epoch": 2837, "lr": 1.1969009748561244e-06}
{"train_loss": 0.05041981860995293, "global_step": 252541, "epoch": 2837, "lr": 1.196774889743213e-06}
{"train_loss": 0.07769688963890076, "global_step": 252542, "epoch": 2837, "lr": 1.1966488111912978e-06}
{"train_loss": 0.04042832553386688, "global_step": 252543, "epoch": 2837, "lr": 1.1965227392003952e-06}
{"train_loss": 0.0669604167342186, "global_step": 252544, "epoch": 2837, "lr": 1.1963966737705334e-06}
{"train_loss": 0.031182289123535156, "global_step": 252545, "epoch": 2837, "lr": 1.1962706149017178e-06}
{"train_loss": 0.01969214715063572, "global_step": 252546, "epoch": 2837, "lr": 1.1961445625939704e-06}
{"train_loss": 0.06236397102475166, "global_step": 252547, "epoch": 2837, "lr": 1.1960185168473026e-06}
{"train_loss": 0.03216369450092316, "global_step": 252548, "epoch": 2837, "lr": 1.1958924776617364e-06}
{"train_loss": 0.06683864444494247, "global_step": 252549, "epoch": 2837, "lr": 1.1957664450372885e-06}
{"train_loss": 0.026616984978318214, "global_step": 252550, "epoch": 2837, "lr": 1.1956404189739756e-06}
{"train_loss": 0.04689344763755798, "global_step": 252551, "epoch": 2837, "lr": 1.1955143994718144e-06}
{"train_loss": 0.03292285278439522, "global_step": 252552, "epoch": 2837, "lr": 1.1953883865308157e-06}
{"train_loss": 0.016007274389266968, "global_step": 252553, "epoch": 2837, "lr": 1.1952623801510077e-06}
{"train_loss": 0.04177573695778847, "global_step": 252554, "epoch": 2837, "lr": 1.1951363803324012e-06}
{"train_loss": 0.03437776863574982, "global_step": 252555, "epoch": 2837, "lr": 1.195010387075013e-06}
{"train_loss": 0.01834917813539505, "global_step": 252556, "epoch": 2837, "lr": 1.1948844003788596e-06}
{"train_loss": 0.04968257620930672, "global_step": 252557, "epoch": 2837, "lr": 1.1947584202439632e-06}
{"train_loss": 0.05447565019130707, "global_step": 252558, "epoch": 2837, "lr": 1.1946324466703295e-06}
{"train_loss": 0.05552174896001816, "global_step": 252559, "epoch": 2837, "lr": 1.194506479657992e-06}
{"train_loss": 0.025368260219693184, "global_step": 252560, "epoch": 2837, "lr": 1.1943805192069502e-06}
{"train_loss": 0.02459714189171791, "global_step": 252561, "epoch": 2837, "lr": 1.1942545653172377e-06}
{"train_loss": 0.04377645626664162, "global_step": 252562, "epoch": 2837, "lr": 1.1941286179888545e-06}
{"train_loss": 0.09274093061685562, "global_step": 252563, "epoch": 2837, "lr": 1.1940026772218227e-06}
{"train_loss": 0.060856856405735016, "global_step": 252564, "epoch": 2837, "lr": 1.1938767430161702e-06}
{"train_loss": 0.025301916524767876, "global_step": 252565, "epoch": 2837, "lr": 1.1937508153719024e-06}
{"train_loss": 0.06809075176715851, "global_step": 252566, "epoch": 2837, "lr": 1.193624894289047e-06}
{"train_loss": 0.04302703216671944, "global_step": 252567, "epoch": 2837, "lr": 1.1934989797676044e-06}
{"train_loss": 0.0468437485396862, "global_step": 252568, "epoch": 2837, "lr": 1.1933730718076075e-06}
{"train_loss": 0.0451255738735199, "global_step": 252569, "epoch": 2837, "lr": 1.193247170409062e-06}
{"train_loss": 0.04627430438995361, "global_step": 252570, "epoch": 2837, "lr": 1.1931212755719957e-06}
{"train_loss": 0.06092977151274681, "global_step": 252571, "epoch": 2837, "lr": 1.1929953872964138e-06}
{"train_loss": 0.04701266810297966, "global_step": 252572, "epoch": 2837, "lr": 1.1928695055823448e-06}
{"train_loss": 0.030226880684494972, "global_step": 252573, "epoch": 2837, "lr": 1.1927436304297935e-06}
{"train_loss": 0.06965075433254242, "global_step": 252574, "epoch": 2837, "lr": 1.1926177618387824e-06}
{"train_loss": 0.025921890512108803, "global_step": 252575, "epoch": 2837, "lr": 1.1924918998093338e-06}
{"train_loss": 0.08193464577198029, "global_step": 252576, "epoch": 2837, "lr": 1.1923660443414585e-06}
{"train_loss": 0.09089599549770355, "global_step": 252577, "epoch": 2837, "lr": 1.192240195435168e-06}
{"train_loss": 0.014927775599062443, "global_step": 252578, "epoch": 2837, "lr": 1.1921143530904954e-06}
{"train_loss": 0.04115291312336922, "global_step": 252579, "epoch": 2837, "lr": 1.1919885173074407e-06}
{"train_loss": 0.04853234440088272, "global_step": 252580, "epoch": 2837, "lr": 1.1918626880860373e-06}
{"train_loss": 0.048273394972588236, "global_step": 252581, "epoch": 2837, "lr": 1.191736865426285e-06, "val_loss": 9.601625442504883}
{"train_loss": 0.026822438463568687, "global_step": 252582, "epoch": 2838, "lr": 1.1916110493282118e-06}
{"train_loss": 0.047291550785303116, "global_step": 252583, "epoch": 2838, "lr": 1.1914852397918285e-06}
{"train_loss": 0.05202814191579819, "global_step": 252584, "epoch": 2838, "lr": 1.1913594368171631e-06}
{"train_loss": 0.03265303745865822, "global_step": 252585, "epoch": 2838, "lr": 1.1912336404042213e-06}
{"train_loss": 0.040194835513830185, "global_step": 252586, "epoch": 2838, "lr": 1.1911078505530194e-06}
{"train_loss": 0.037958212196826935, "global_step": 252587, "epoch": 2838, "lr": 1.1909820672635797e-06}
{"train_loss": 0.05812542885541916, "global_step": 252588, "epoch": 2838, "lr": 1.1908562905359188e-06}
{"train_loss": 0.0456773042678833, "global_step": 252589, "epoch": 2838, "lr": 1.1907305203700537e-06}
{"train_loss": 0.04844798147678375, "global_step": 252590, "epoch": 2838, "lr": 1.190604756765995e-06}
{"train_loss": 0.005792577285319567, "global_step": 252591, "epoch": 2838, "lr": 1.1904789997237708e-06}
{"train_loss": 0.038857121020555496, "global_step": 252592, "epoch": 2838, "lr": 1.1903532492433867e-06}
{"train_loss": 0.015600615181028843, "global_step": 252593, "epoch": 2838, "lr": 1.1902275053248647e-06}
{"train_loss": 0.039877139031887054, "global_step": 252594, "epoch": 2838, "lr": 1.1901017679682213e-06}
{"train_loss": 0.028509194031357765, "global_step": 252595, "epoch": 2838, "lr": 1.1899760371734792e-06}
{"train_loss": 0.06175380200147629, "global_step": 252596, "epoch": 2838, "lr": 1.1898503129406434e-06}
{"train_loss": 0.045547906309366226, "global_step": 252597, "epoch": 2838, "lr": 1.1897245952697423e-06}
{"train_loss": 0.05075469985604286, "global_step": 252598, "epoch": 2838, "lr": 1.1895988841607863e-06}
{"train_loss": 0.043264418840408325, "global_step": 252599, "epoch": 2838, "lr": 1.189473179613787e-06}
{"train_loss": 0.010617868043482304, "global_step": 252600, "epoch": 2838, "lr": 1.1893474816287775e-06}
{"train_loss": 0.04783489555120468, "global_step": 252601, "epoch": 2838, "lr": 1.1892217902057634e-06}
{"train_loss": 0.05614006146788597, "global_step": 252602, "epoch": 2838, "lr": 1.1890961053447614e-06}
{"train_loss": 0.035749513655900955, "global_step": 252603, "epoch": 2838, "lr": 1.1889704270457935e-06}
{"train_loss": 0.039214082062244415, "global_step": 252604, "epoch": 2838, "lr": 1.188844755308871e-06}
{"train_loss": 0.060172874480485916, "global_step": 252605, "epoch": 2838, "lr": 1.1887190901340105e-06}
{"train_loss": 0.025955805554986, "global_step": 252606, "epoch": 2838, "lr": 1.1885934315212399e-06}
{"train_loss": 0.07127046585083008, "global_step": 252607, "epoch": 2838, "lr": 1.188467779470559e-06}
{"train_loss": 0.048582810908555984, "global_step": 252608, "epoch": 2838, "lr": 1.188342133982001e-06}
{"train_loss": 0.042408525943756104, "global_step": 252609, "epoch": 2838, "lr": 1.1882164950555718e-06}
{"train_loss": 0.00690450007095933, "global_step": 252610, "epoch": 2838, "lr": 1.1880908626912934e-06}
{"train_loss": 0.015617823228240013, "global_step": 252611, "epoch": 2838, "lr": 1.1879652368891826e-06}
{"train_loss": 0.0402316153049469, "global_step": 252612, "epoch": 2838, "lr": 1.1878396176492557e-06}
{"train_loss": 0.02119951881468296, "global_step": 252613, "epoch": 2838, "lr": 1.187714004971524e-06}
{"train_loss": 0.09735886752605438, "global_step": 252614, "epoch": 2838, "lr": 1.18758839885601e-06}
{"train_loss": 0.061582062393426895, "global_step": 252615, "epoch": 2838, "lr": 1.18746279930273e-06}
{"train_loss": 0.008730670437216759, "global_step": 252616, "epoch": 2838, "lr": 1.1873372063117061e-06}
{"train_loss": 0.07970555871725082, "global_step": 252617, "epoch": 2838, "lr": 1.1872116198829441e-06}
{"train_loss": 0.024426404386758804, "global_step": 252618, "epoch": 2838, "lr": 1.1870860400164663e-06}
{"train_loss": 0.017686927691102028, "global_step": 252619, "epoch": 2838, "lr": 1.186960466712289e-06}
{"train_loss": 0.04117451608181, "global_step": 252620, "epoch": 2838, "lr": 1.186834899970435e-06}
{"train_loss": 0.03643227741122246, "global_step": 252621, "epoch": 2838, "lr": 1.1867093397909146e-06}
{"train_loss": 0.02866603247821331, "global_step": 252622, "epoch": 2838, "lr": 1.186583786173745e-06}
{"train_loss": 0.09042824804782867, "global_step": 252623, "epoch": 2838, "lr": 1.1864582391189482e-06}
{"train_loss": 0.09194686263799667, "global_step": 252624, "epoch": 2838, "lr": 1.18633269862653e-06}
{"train_loss": 0.07263893634080887, "global_step": 252625, "epoch": 2838, "lr": 1.1862071646965179e-06}
{"train_loss": 0.009188510477542877, "global_step": 252626, "epoch": 2838, "lr": 1.186081637328923e-06}
{"train_loss": 0.04484415054321289, "global_step": 252627, "epoch": 2838, "lr": 1.1859561165237732e-06}
{"train_loss": 0.01640177145600319, "global_step": 252628, "epoch": 2838, "lr": 1.1858306022810629e-06}
{"train_loss": 0.02403164841234684, "global_step": 252629, "epoch": 2838, "lr": 1.1857050946008362e-06}
{"train_loss": 0.03308279812335968, "global_step": 252630, "epoch": 2838, "lr": 1.1855795934830882e-06}
{"train_loss": 0.01597367227077484, "global_step": 252631, "epoch": 2838, "lr": 1.1854540989278463e-06}
{"train_loss": 0.0795908197760582, "global_step": 252632, "epoch": 2838, "lr": 1.1853286109351213e-06}
{"train_loss": 0.048579126596450806, "global_step": 252633, "epoch": 2838, "lr": 1.1852031295049415e-06}
{"train_loss": 0.08652261644601822, "global_step": 252634, "epoch": 2838, "lr": 1.1850776546373067e-06}
{"train_loss": 0.04806061461567879, "global_step": 252635, "epoch": 2838, "lr": 1.18495218633225e-06}
{"train_loss": 0.05582131817936897, "global_step": 252636, "epoch": 2838, "lr": 1.1848267245897825e-06}
{"train_loss": 0.07811146974563599, "global_step": 252637, "epoch": 2838, "lr": 1.1847012694099158e-06}
{"train_loss": 0.04471437260508537, "global_step": 252638, "epoch": 2838, "lr": 1.1845758207926717e-06}
{"train_loss": 0.020039668306708336, "global_step": 252639, "epoch": 2838, "lr": 1.1844503787380666e-06}
{"train_loss": 0.07021249830722809, "global_step": 252640, "epoch": 2838, "lr": 1.184324943246118e-06}
{"train_loss": 0.02261660061776638, "global_step": 252641, "epoch": 2838, "lr": 1.1841995143168415e-06}
{"train_loss": 0.04066108539700508, "global_step": 252642, "epoch": 2838, "lr": 1.1840740919502602e-06}
{"train_loss": 0.04493816941976547, "global_step": 252643, "epoch": 2838, "lr": 1.1839486761463736e-06}
{"train_loss": 0.027458403259515762, "global_step": 252644, "epoch": 2838, "lr": 1.1838232669052206e-06}
{"train_loss": 0.04941301420331001, "global_step": 252645, "epoch": 2838, "lr": 1.1836978642268014e-06}
{"train_loss": 0.07849390059709549, "global_step": 252646, "epoch": 2838, "lr": 1.1835724681111437e-06}
{"train_loss": 0.07106336206197739, "global_step": 252647, "epoch": 2838, "lr": 1.183447078558253e-06}
{"train_loss": 0.03313650190830231, "global_step": 252648, "epoch": 2838, "lr": 1.1833216955681626e-06}
{"train_loss": 0.04223344475030899, "global_step": 252649, "epoch": 2838, "lr": 1.1831963191408724e-06}
{"train_loss": 0.041575007140636444, "global_step": 252650, "epoch": 2838, "lr": 1.1830709492764103e-06}
{"train_loss": 0.029806621372699738, "global_step": 252651, "epoch": 2838, "lr": 1.1829455859747818e-06}
{"train_loss": 0.06859841197729111, "global_step": 252652, "epoch": 2838, "lr": 1.1828202292360202e-06}
{"train_loss": 0.05475229397416115, "global_step": 252653, "epoch": 2838, "lr": 1.1826948790601254e-06}
{"train_loss": 0.028229504823684692, "global_step": 252654, "epoch": 2838, "lr": 1.1825695354471257e-06}
{"train_loss": 0.028106650337576866, "global_step": 252655, "epoch": 2838, "lr": 1.1824441983970369e-06}
{"train_loss": 0.04572467878460884, "global_step": 252656, "epoch": 2838, "lr": 1.1823188679098708e-06}
{"train_loss": 0.03580498695373535, "global_step": 252657, "epoch": 2838, "lr": 1.1821935439856491e-06}
{"train_loss": 0.046946004033088684, "global_step": 252658, "epoch": 2838, "lr": 1.1820682266243887e-06}
{"train_loss": 0.041128311306238174, "global_step": 252659, "epoch": 2838, "lr": 1.1819429158261009e-06}
{"train_loss": 0.025570986792445183, "global_step": 252660, "epoch": 2838, "lr": 1.1818176115908075e-06}
{"train_loss": 0.06136738508939743, "global_step": 252661, "epoch": 2838, "lr": 1.1816923139185198e-06}
{"train_loss": 0.06839925795793533, "global_step": 252662, "epoch": 2838, "lr": 1.1815670228092602e-06}
{"train_loss": 0.039284057915210724, "global_step": 252663, "epoch": 2838, "lr": 1.1814417382630506e-06}
{"train_loss": 0.029668239876627922, "global_step": 252664, "epoch": 2838, "lr": 1.181316460279891e-06}
{"train_loss": 0.06853503733873367, "global_step": 252665, "epoch": 2838, "lr": 1.181191188859815e-06}
{"train_loss": 0.008725998923182487, "global_step": 252666, "epoch": 2838, "lr": 1.1810659240028277e-06}
{"train_loss": 0.08604824542999268, "global_step": 252667, "epoch": 2838, "lr": 1.1809406657089573e-06}
{"train_loss": 0.05037233605980873, "global_step": 252668, "epoch": 2838, "lr": 1.180815413978209e-06}
{"train_loss": 0.05878705903887749, "global_step": 252669, "epoch": 2838, "lr": 1.1806901688106053e-06}
{"train_loss": 0.04447732132263063, "global_step": 252670, "epoch": 2838, "lr": 1.1805649302061628e-06, "val_loss": 9.6458740234375}
{"train_loss": 0.02693699672818184, "global_step": 252671, "epoch": 2839, "lr": 1.180439698164898e-06}
{"train_loss": 0.07266956567764282, "global_step": 252672, "epoch": 2839, "lr": 1.1803144726868331e-06}
{"train_loss": 0.03658483177423477, "global_step": 252673, "epoch": 2839, "lr": 1.1801892537719738e-06}
{"train_loss": 0.06343255192041397, "global_step": 252674, "epoch": 2839, "lr": 1.180064041420348e-06}
{"train_loss": 0.062352295964956284, "global_step": 252675, "epoch": 2839, "lr": 1.1799388356319663e-06}
{"train_loss": 0.037905093282461166, "global_step": 252676, "epoch": 2839, "lr": 1.1798136364068456e-06}
{"train_loss": 0.04921199753880501, "global_step": 252677, "epoch": 2839, "lr": 1.179688443745003e-06}
{"train_loss": 0.05118468403816223, "global_step": 252678, "epoch": 2839, "lr": 1.17956325764646e-06}
{"train_loss": 0.052086323499679565, "global_step": 252679, "epoch": 2839, "lr": 1.1794380781112224e-06}
{"train_loss": 0.012126862071454525, "global_step": 252680, "epoch": 2839, "lr": 1.1793129051393238e-06}
{"train_loss": 0.03885797783732414, "global_step": 252681, "epoch": 2839, "lr": 1.1791877387307638e-06}
{"train_loss": 0.06365863978862762, "global_step": 252682, "epoch": 2839, "lr": 1.1790625788855758e-06}
{"train_loss": 0.08285986632108688, "global_step": 252683, "epoch": 2839, "lr": 1.17893742560376e-06}
{"train_loss": 0.01980537921190262, "global_step": 252684, "epoch": 2839, "lr": 1.178812278885344e-06}
{"train_loss": 0.03655105456709862, "global_step": 252685, "epoch": 2839, "lr": 1.1786871387303389e-06}
{"train_loss": 0.08611898869276047, "global_step": 252686, "epoch": 2839, "lr": 1.1785620051387668e-06}
{"train_loss": 0.03009673021733761, "global_step": 252687, "epoch": 2839, "lr": 1.1784368781106392e-06}
{"train_loss": 0.06433980166912079, "global_step": 252688, "epoch": 2839, "lr": 1.1783117576459778e-06}
{"train_loss": 0.06393394619226456, "global_step": 252689, "epoch": 2839, "lr": 1.1781866437447996e-06}
{"train_loss": 0.05381406843662262, "global_step": 252690, "epoch": 2839, "lr": 1.17806153640711e-06}
{"train_loss": 0.06937611848115921, "global_step": 252691, "epoch": 2839, "lr": 1.177936435632948e-06}
{"train_loss": 0.054480500519275665, "global_step": 252692, "epoch": 2839, "lr": 1.1778113414223079e-06}
{"train_loss": 0.05602249875664711, "global_step": 252693, "epoch": 2839, "lr": 1.1776862537752232e-06}
{"train_loss": 0.05729197338223457, "global_step": 252694, "epoch": 2839, "lr": 1.1775611726916936e-06}
{"train_loss": 0.054829660803079605, "global_step": 252695, "epoch": 2839, "lr": 1.177436098171758e-06}
{"train_loss": 0.04732455313205719, "global_step": 252696, "epoch": 2839, "lr": 1.1773110302154167e-06}
{"train_loss": 0.06514987349510193, "global_step": 252697, "epoch": 2839, "lr": 1.177185968822686e-06}
{"train_loss": 0.06916848570108414, "global_step": 252698, "epoch": 2839, "lr": 1.177060913993594e-06}
{"train_loss": 0.08313041925430298, "global_step": 252699, "epoch": 2839, "lr": 1.176935865728146e-06}
{"train_loss": 0.04533344879746437, "global_step": 252700, "epoch": 2839, "lr": 1.1768108240263643e-06}
{"train_loss": 0.03665779158473015, "global_step": 252701, "epoch": 2839, "lr": 1.176685788888271e-06}
{"train_loss": 0.01677742972970009, "global_step": 252702, "epoch": 2839, "lr": 1.1765607603138717e-06}
{"train_loss": 0.054879724979400635, "global_step": 252703, "epoch": 2839, "lr": 1.1764357383031888e-06}
{"train_loss": 0.055726293474435806, "global_step": 252704, "epoch": 2839, "lr": 1.1763107228562387e-06}
{"train_loss": 0.04974910989403725, "global_step": 252705, "epoch": 2839, "lr": 1.1761857139730437e-06}
{"train_loss": 0.038255542516708374, "global_step": 252706, "epoch": 2839, "lr": 1.1760607116536093e-06}
{"train_loss": 0.04463047534227371, "global_step": 252707, "epoch": 2839, "lr": 1.1759357158979635e-06}
{"train_loss": 0.05224089324474335, "global_step": 252708, "epoch": 2839, "lr": 1.1758107267061114e-06}
{"train_loss": 0.05572402477264404, "global_step": 252709, "epoch": 2839, "lr": 1.1756857440780812e-06}
{"train_loss": 0.05708572268486023, "global_step": 252710, "epoch": 2839, "lr": 1.1755607680138835e-06}
{"train_loss": 0.049049679189920425, "global_step": 252711, "epoch": 2839, "lr": 1.1754357985135356e-06}
{"train_loss": 0.046376947313547134, "global_step": 252712, "epoch": 2839, "lr": 1.1753108355770592e-06}
{"train_loss": 0.01114202756434679, "global_step": 252713, "epoch": 2839, "lr": 1.1751858792044656e-06}
{"train_loss": 0.0550401508808136, "global_step": 252714, "epoch": 2839, "lr": 1.175060929395777e-06}
{"train_loss": 0.03840186819434166, "global_step": 252715, "epoch": 2839, "lr": 1.174935986150999e-06}
{"train_loss": 0.017357513308525085, "global_step": 252716, "epoch": 2839, "lr": 1.174811049470159e-06}
{"train_loss": 0.06836061179637909, "global_step": 252717, "epoch": 2839, "lr": 1.1746861193532687e-06}
{"train_loss": 0.0372629351913929, "global_step": 252718, "epoch": 2839, "lr": 1.1745611958003499e-06}
{"train_loss": 0.061726003885269165, "global_step": 252719, "epoch": 2839, "lr": 1.1744362788114138e-06}
{"train_loss": 0.09538497030735016, "global_step": 252720, "epoch": 2839, "lr": 1.1743113683864826e-06}
{"train_loss": 0.04057256504893303, "global_step": 252721, "epoch": 2839, "lr": 1.1741864645255674e-06}
{"train_loss": 0.0533149354159832, "global_step": 252722, "epoch": 2839, "lr": 1.1740615672286904e-06}
{"train_loss": 0.037497732788324356, "global_step": 252723, "epoch": 2839, "lr": 1.173936676495857e-06}
{"train_loss": 0.10391213744878769, "global_step": 252724, "epoch": 2839, "lr": 1.1738117923271064e-06}
{"train_loss": 0.08066745102405548, "global_step": 252725, "epoch": 2839, "lr": 1.1736869147224328e-06}
{"train_loss": 0.045289888978004456, "global_step": 252726, "epoch": 2839, "lr": 1.1735620436818585e-06}
{"train_loss": 0.05727503076195717, "global_step": 252727, "epoch": 2839, "lr": 1.1734371792054111e-06}
{"train_loss": 0.03951362147927284, "global_step": 252728, "epoch": 2839, "lr": 1.1733123212930962e-06}
{"train_loss": 0.05692174285650253, "global_step": 252729, "epoch": 2839, "lr": 1.1731874699449364e-06}
{"train_loss": 0.034121472388505936, "global_step": 252730, "epoch": 2839, "lr": 1.1730626251609478e-06}
{"train_loss": 0.03134119510650635, "global_step": 252731, "epoch": 2839, "lr": 1.1729377869411418e-06}
{"train_loss": 0.03206154704093933, "global_step": 252732, "epoch": 2839, "lr": 1.1728129552855404e-06}
{"train_loss": 0.023352693766355515, "global_step": 252733, "epoch": 2839, "lr": 1.1726881301941662e-06}
{"train_loss": 0.03544413670897484, "global_step": 252734, "epoch": 2839, "lr": 1.1725633116670243e-06}
{"train_loss": 0.05897635594010353, "global_step": 252735, "epoch": 2839, "lr": 1.1724384997041315e-06}
{"train_loss": 0.08036439120769501, "global_step": 252736, "epoch": 2839, "lr": 1.1723136943055158e-06}
{"train_loss": 0.02629692107439041, "global_step": 252737, "epoch": 2839, "lr": 1.1721888954711824e-06}
{"train_loss": 0.03830643370747566, "global_step": 252738, "epoch": 2839, "lr": 1.1720641032011537e-06}
{"train_loss": 0.030851297080516815, "global_step": 252739, "epoch": 2839, "lr": 1.1719393174954518e-06}
{"train_loss": 0.04754827916622162, "global_step": 252740, "epoch": 2839, "lr": 1.1718145383540769e-06}
{"train_loss": 0.05592789128422737, "global_step": 252741, "epoch": 2839, "lr": 1.1716897657770676e-06}
{"train_loss": 0.01815682463347912, "global_step": 252742, "epoch": 2839, "lr": 1.171564999764424e-06}
{"train_loss": 0.051570601761341095, "global_step": 252743, "epoch": 2839, "lr": 1.1714402403161684e-06}
{"train_loss": 0.05911672115325928, "global_step": 252744, "epoch": 2839, "lr": 1.1713154874323174e-06}
{"train_loss": 0.06048065796494484, "global_step": 252745, "epoch": 2839, "lr": 1.171190741112882e-06}
{"train_loss": 0.03437141329050064, "global_step": 252746, "epoch": 2839, "lr": 1.1710660013578955e-06}
{"train_loss": 0.03624803200364113, "global_step": 252747, "epoch": 2839, "lr": 1.1709412681673581e-06}
{"train_loss": 0.04908067733049393, "global_step": 252748, "epoch": 2839, "lr": 1.170816541541292e-06}
{"train_loss": 0.021910397335886955, "global_step": 252749, "epoch": 2839, "lr": 1.1706918214797137e-06}
{"train_loss": 0.05114360153675079, "global_step": 252750, "epoch": 2839, "lr": 1.1705671079826508e-06}
{"train_loss": 0.04554400593042374, "global_step": 252751, "epoch": 2839, "lr": 1.170442401050098e-06}
{"train_loss": 0.039225392043590546, "global_step": 252752, "epoch": 2839, "lr": 1.1703177006820886e-06}
{"train_loss": 0.07047832012176514, "global_step": 252753, "epoch": 2839, "lr": 1.1701930068786337e-06}
{"train_loss": 0.04176710173487663, "global_step": 252754, "epoch": 2839, "lr": 1.1700683196397555e-06}
{"train_loss": 0.023691486567258835, "global_step": 252755, "epoch": 2839, "lr": 1.1699436389654595e-06}
{"train_loss": 0.02639789506793022, "global_step": 252756, "epoch": 2839, "lr": 1.169818964855779e-06}
{"train_loss": 0.04621215537190437, "global_step": 252757, "epoch": 2839, "lr": 1.1696942973107139e-06}
{"train_loss": 0.06597276777029037, "global_step": 252758, "epoch": 2839, "lr": 1.1695696363302922e-06}
{"train_loss": 0.048689019876752004, "global_step": 252759, "epoch": 2839, "lr": 1.1694449819145192e-06, "val_loss": 9.546875953674316}
{"train_loss": 0.037077393382787704, "global_step": 252760, "epoch": 2840, "lr": 1.1693203340634284e-06}
{"train_loss": 0.03442554920911789, "global_step": 252761, "epoch": 2840, "lr": 1.1691956927770197e-06}
{"train_loss": 0.06561817973852158, "global_step": 252762, "epoch": 2840, "lr": 1.1690710580553155e-06}
{"train_loss": 0.05806232616305351, "global_step": 252763, "epoch": 2840, "lr": 1.1689464298983431e-06}
{"train_loss": 0.0537966750562191, "global_step": 252764, "epoch": 2840, "lr": 1.1688218083061087e-06}
{"train_loss": 0.011683459393680096, "global_step": 252765, "epoch": 2840, "lr": 1.1686971932786283e-06}
{"train_loss": 0.02300017699599266, "global_step": 252766, "epoch": 2840, "lr": 1.1685725848159246e-06}
{"train_loss": 0.07196763157844543, "global_step": 252767, "epoch": 2840, "lr": 1.1684479829180085e-06}
{"train_loss": 0.02228379435837269, "global_step": 252768, "epoch": 2840, "lr": 1.168323387584902e-06}
{"train_loss": 0.0362703911960125, "global_step": 252769, "epoch": 2840, "lr": 1.1681987988166166e-06}
{"train_loss": 0.06575465947389603, "global_step": 252770, "epoch": 2840, "lr": 1.168074216613174e-06}
{"train_loss": 0.024325279518961906, "global_step": 252771, "epoch": 2840, "lr": 1.167949640974586e-06}
{"train_loss": 0.016962196677923203, "global_step": 252772, "epoch": 2840, "lr": 1.1678250719008742e-06}
{"train_loss": 0.06162947788834572, "global_step": 252773, "epoch": 2840, "lr": 1.1677005093920557e-06}
{"train_loss": 0.05479639396071434, "global_step": 252774, "epoch": 2840, "lr": 1.1675759534481412e-06}
{"train_loss": 0.017163969576358795, "global_step": 252775, "epoch": 2840, "lr": 1.1674514040691475e-06}
{"train_loss": 0.03286303952336311, "global_step": 252776, "epoch": 2840, "lr": 1.1673268612551026e-06}
{"train_loss": 0.016002990305423737, "global_step": 252777, "epoch": 2840, "lr": 1.167202325006006e-06}
{"train_loss": 0.02875097654759884, "global_step": 252778, "epoch": 2840, "lr": 1.1670777953218914e-06}
{"train_loss": 0.05707506090402603, "global_step": 252779, "epoch": 2840, "lr": 1.1669532722027699e-06}
{"train_loss": 0.04383955895900726, "global_step": 252780, "epoch": 2840, "lr": 1.1668287556486523e-06}
{"train_loss": 0.05901738256216049, "global_step": 252781, "epoch": 2840, "lr": 1.1667042456595556e-06}
{"train_loss": 0.057135194540023804, "global_step": 252782, "epoch": 2840, "lr": 1.1665797422355073e-06}
{"train_loss": 0.048148926347494125, "global_step": 252783, "epoch": 2840, "lr": 1.1664552453765076e-06}
{"train_loss": 0.050552140921354294, "global_step": 252784, "epoch": 2840, "lr": 1.1663307550825953e-06}
{"train_loss": 0.045243918895721436, "global_step": 252785, "epoch": 2840, "lr": 1.1662062713537647e-06}
{"train_loss": 0.057453762739896774, "global_step": 252786, "epoch": 2840, "lr": 1.1660817941900492e-06}
{"train_loss": 0.06816792488098145, "global_step": 252787, "epoch": 2840, "lr": 1.165957323591449e-06}
{"train_loss": 0.05427839979529381, "global_step": 252788, "epoch": 2840, "lr": 1.1658328595580026e-06}
{"train_loss": 0.09475614875555038, "global_step": 252789, "epoch": 2840, "lr": 1.1657084020897047e-06}
{"train_loss": 0.0292697474360466, "global_step": 252790, "epoch": 2840, "lr": 1.1655839511865885e-06}
{"train_loss": 0.04663071781396866, "global_step": 252791, "epoch": 2840, "lr": 1.1654595068486651e-06}
{"train_loss": 0.031070690602064133, "global_step": 252792, "epoch": 2840, "lr": 1.1653350690759458e-06}
{"train_loss": 0.025138918310403824, "global_step": 252793, "epoch": 2840, "lr": 1.1652106378684523e-06}
{"train_loss": 0.056231893599033356, "global_step": 252794, "epoch": 2840, "lr": 1.1650862132262075e-06}
{"train_loss": 0.04449194297194481, "global_step": 252795, "epoch": 2840, "lr": 1.1649617951492164e-06}
{"train_loss": 0.026969479396939278, "global_step": 252796, "epoch": 2840, "lr": 1.1648373836375016e-06}
{"train_loss": 0.06364265084266663, "global_step": 252797, "epoch": 2840, "lr": 1.1647129786910793e-06}
{"train_loss": 0.028543170541524887, "global_step": 252798, "epoch": 2840, "lr": 1.164588580309961e-06}
{"train_loss": 0.05002885311841965, "global_step": 252799, "epoch": 2840, "lr": 1.1644641884941742e-06}
{"train_loss": 0.05674159899353981, "global_step": 252800, "epoch": 2840, "lr": 1.1643398032437302e-06}
{"train_loss": 0.027541767805814743, "global_step": 252801, "epoch": 2840, "lr": 1.1642154245586456e-06}
{"train_loss": 0.039614077657461166, "global_step": 252802, "epoch": 2840, "lr": 1.1640910524389316e-06}
{"train_loss": 0.01759878173470497, "global_step": 252803, "epoch": 2840, "lr": 1.1639666868846212e-06}
{"train_loss": 0.014540344476699829, "global_step": 252804, "epoch": 2840, "lr": 1.1638423278957089e-06}
{"train_loss": 0.047405023127794266, "global_step": 252805, "epoch": 2840, "lr": 1.1637179754722283e-06}
{"train_loss": 0.05753418803215027, "global_step": 252806, "epoch": 2840, "lr": 1.1635936296141902e-06}
{"train_loss": 0.04816271364688873, "global_step": 252807, "epoch": 2840, "lr": 1.1634692903216116e-06}
{"train_loss": 0.04869605228304863, "global_step": 252808, "epoch": 2840, "lr": 1.1633449575945088e-06}
{"train_loss": 0.028479762375354767, "global_step": 252809, "epoch": 2840, "lr": 1.1632206314329043e-06}
{"train_loss": 0.0458696223795414, "global_step": 252810, "epoch": 2840, "lr": 1.1630963118367976e-06}
{"train_loss": 0.044703591614961624, "global_step": 252811, "epoch": 2840, "lr": 1.1629719988062282e-06}
{"train_loss": 0.048387523740530014, "global_step": 252812, "epoch": 2840, "lr": 1.1628476923412013e-06}
{"train_loss": 0.02668612077832222, "global_step": 252813, "epoch": 2840, "lr": 1.162723392441728e-06}
{"train_loss": 0.03204961121082306, "global_step": 252814, "epoch": 2840, "lr": 1.162599099107836e-06}
{"train_loss": 0.07064201682806015, "global_step": 252815, "epoch": 2840, "lr": 1.1624748123395367e-06}
{"train_loss": 0.028205687180161476, "global_step": 252816, "epoch": 2840, "lr": 1.1623505321368466e-06}
{"train_loss": 0.06579151004552841, "global_step": 252817, "epoch": 2840, "lr": 1.1622262584997823e-06}
{"train_loss": 0.03436126187443733, "global_step": 252818, "epoch": 2840, "lr": 1.162101991428366e-06}
{"train_loss": 0.021566733717918396, "global_step": 252819, "epoch": 2840, "lr": 1.1619777309226033e-06}
{"train_loss": 0.03710341826081276, "global_step": 252820, "epoch": 2840, "lr": 1.1618534769825219e-06}
{"train_loss": 0.07307198643684387, "global_step": 252821, "epoch": 2840, "lr": 1.1617292296081272e-06}
{"train_loss": 0.049499865621328354, "global_step": 252822, "epoch": 2840, "lr": 1.161604988799453e-06}
{"train_loss": 0.03546682745218277, "global_step": 252823, "epoch": 2840, "lr": 1.1614807545564988e-06}
{"train_loss": 0.07403174787759781, "global_step": 252824, "epoch": 2840, "lr": 1.1613565268792925e-06}
{"train_loss": 0.06260625272989273, "global_step": 252825, "epoch": 2840, "lr": 1.1612323057678398e-06}
{"train_loss": 0.06904643774032593, "global_step": 252826, "epoch": 2840, "lr": 1.161108091222174e-06}
{"train_loss": 0.022710110992193222, "global_step": 252827, "epoch": 2840, "lr": 1.1609838832422947e-06}
{"train_loss": 0.05489715561270714, "global_step": 252828, "epoch": 2840, "lr": 1.1608596818282302e-06}
{"train_loss": 0.04138622432947159, "global_step": 252829, "epoch": 2840, "lr": 1.1607354869799857e-06}
{"train_loss": 0.07858294248580933, "global_step": 252830, "epoch": 2840, "lr": 1.1606112986975893e-06}
{"train_loss": 0.041032660752534866, "global_step": 252831, "epoch": 2840, "lr": 1.1604871169810516e-06}
{"train_loss": 0.04656736180186272, "global_step": 252832, "epoch": 2840, "lr": 1.1603629418304007e-06}
{"train_loss": 0.03384663164615631, "global_step": 252833, "epoch": 2840, "lr": 1.160238773245631e-06}
{"train_loss": 0.06458865851163864, "global_step": 252834, "epoch": 2840, "lr": 1.1601146112267813e-06}
{"train_loss": 0.025468068197369576, "global_step": 252835, "epoch": 2840, "lr": 1.1599904557738517e-06}
{"train_loss": 0.06000784412026405, "global_step": 252836, "epoch": 2840, "lr": 1.1598663068868643e-06}
{"train_loss": 0.06132877990603447, "global_step": 252837, "epoch": 2840, "lr": 1.1597421645658468e-06}
{"train_loss": 0.06431691348552704, "global_step": 252838, "epoch": 2840, "lr": 1.1596180288107995e-06}
{"train_loss": 0.12584851682186127, "global_step": 252839, "epoch": 2840, "lr": 1.1594938996217553e-06}
{"train_loss": 0.03408385440707207, "global_step": 252840, "epoch": 2840, "lr": 1.159369776998709e-06}
{"train_loss": 0.018790237605571747, "global_step": 252841, "epoch": 2840, "lr": 1.1592456609416991e-06}
{"train_loss": 0.05976023152470589, "global_step": 252842, "epoch": 2840, "lr": 1.1591215514507314e-06}
{"train_loss": 0.06323035061359406, "global_step": 252843, "epoch": 2840, "lr": 1.1589974485258282e-06}
{"train_loss": 0.10081804543733597, "global_step": 252844, "epoch": 2840, "lr": 1.1588733521669948e-06}
{"train_loss": 0.02833501063287258, "global_step": 252845, "epoch": 2840, "lr": 1.1587492623742647e-06}
{"train_loss": 0.02542298473417759, "global_step": 252846, "epoch": 2840, "lr": 1.1586251791476376e-06}
{"train_loss": 0.042123377323150635, "global_step": 252847, "epoch": 2840, "lr": 1.1585011024871418e-06}
{"train_loss": 0.04592141824031479, "global_step": 252848, "epoch": 2840, "lr": 1.158377032392788e-06, "val_loss": 9.539209365844727, "train_action_mse_error": 22.514251708984375}
{"train_loss": 0.04385348781943321, "global_step": 252849, "epoch": 2841, "lr": 1.1582529688645981e-06}
{"train_loss": 0.03726714476943016, "global_step": 252850, "epoch": 2841, "lr": 1.158128911902584e-06}
{"train_loss": 0.032405778765678406, "global_step": 252851, "epoch": 2841, "lr": 1.1580048615067617e-06}
{"train_loss": 0.030827779322862625, "global_step": 252852, "epoch": 2841, "lr": 1.1578808176771539e-06}
{"train_loss": 0.032643891870975494, "global_step": 252853, "epoch": 2841, "lr": 1.1577567804137712e-06}
{"train_loss": 0.03870834410190582, "global_step": 252854, "epoch": 2841, "lr": 1.157632749716636e-06}
{"train_loss": 0.04031931981444359, "global_step": 252855, "epoch": 2841, "lr": 1.1575087255857598e-06}
{"train_loss": 0.036278750747442245, "global_step": 252856, "epoch": 2841, "lr": 1.1573847080211641e-06}
{"train_loss": 0.023935791105031967, "global_step": 252857, "epoch": 2841, "lr": 1.157260697022855e-06}
{"train_loss": 0.06537449359893799, "global_step": 252858, "epoch": 2841, "lr": 1.1571366925908656e-06}
{"train_loss": 0.055985525250434875, "global_step": 252859, "epoch": 2841, "lr": 1.1570126947251957e-06}
{"train_loss": 0.086307592689991, "global_step": 252860, "epoch": 2841, "lr": 1.1568887034258792e-06}
{"train_loss": 0.028286118060350418, "global_step": 252861, "epoch": 2841, "lr": 1.1567647186929153e-06}
{"train_loss": 0.039283812046051025, "global_step": 252862, "epoch": 2841, "lr": 1.156640740526338e-06}
{"train_loss": 0.04834194853901863, "global_step": 252863, "epoch": 2841, "lr": 1.156516768926147e-06}
{"train_loss": 0.01974749192595482, "global_step": 252864, "epoch": 2841, "lr": 1.1563928038923699e-06}
{"train_loss": 0.0615796260535717, "global_step": 252865, "epoch": 2841, "lr": 1.156268845425018e-06}
{"train_loss": 0.05077669396996498, "global_step": 252866, "epoch": 2841, "lr": 1.1561448935241136e-06}
{"train_loss": 0.05638856440782547, "global_step": 252867, "epoch": 2841, "lr": 1.1560209481896678e-06}
{"train_loss": 0.024579539895057678, "global_step": 252868, "epoch": 2841, "lr": 1.1558970094217081e-06}
{"train_loss": 0.018756402656435966, "global_step": 252869, "epoch": 2841, "lr": 1.155773077220229e-06}
{"train_loss": 0.09375639259815216, "global_step": 252870, "epoch": 2841, "lr": 1.1556491515852753e-06}
{"train_loss": 0.03211435675621033, "global_step": 252871, "epoch": 2841, "lr": 1.1555252325168354e-06}
{"train_loss": 0.08761962503194809, "global_step": 252872, "epoch": 2841, "lr": 1.1554013200149426e-06}
{"train_loss": 0.04182487726211548, "global_step": 252873, "epoch": 2841, "lr": 1.1552774140796196e-06}
{"train_loss": 0.04891687631607056, "global_step": 252874, "epoch": 2841, "lr": 1.1551535147108662e-06}
{"train_loss": 0.022324664518237114, "global_step": 252875, "epoch": 2841, "lr": 1.1550296219087098e-06}
{"train_loss": 0.02104608528316021, "global_step": 252876, "epoch": 2841, "lr": 1.1549057356731617e-06}
{"train_loss": 0.05777434632182121, "global_step": 252877, "epoch": 2841, "lr": 1.15478185600425e-06}
{"train_loss": 0.028602883219718933, "global_step": 252878, "epoch": 2841, "lr": 1.1546579829019744e-06}
{"train_loss": 0.07109721004962921, "global_step": 252879, "epoch": 2841, "lr": 1.1545341163663625e-06}
{"train_loss": 0.09509368240833282, "global_step": 252880, "epoch": 2841, "lr": 1.1544102563974258e-06}
{"train_loss": 0.014941219240427017, "global_step": 252881, "epoch": 2841, "lr": 1.1542864029951861e-06}
{"train_loss": 0.034452881664037704, "global_step": 252882, "epoch": 2841, "lr": 1.1541625561596547e-06}
{"train_loss": 0.05607520416378975, "global_step": 252883, "epoch": 2841, "lr": 1.1540387158908538e-06}
{"train_loss": 0.061559032648801804, "global_step": 252884, "epoch": 2841, "lr": 1.1539148821887946e-06}
{"train_loss": 0.024970768019557, "global_step": 252885, "epoch": 2841, "lr": 1.1537910550534992e-06}
{"train_loss": 0.046218447387218475, "global_step": 252886, "epoch": 2841, "lr": 1.153667234484973e-06}
{"train_loss": 0.023611949756741524, "global_step": 252887, "epoch": 2841, "lr": 1.1535434204832497e-06}
{"train_loss": 0.03510364517569542, "global_step": 252888, "epoch": 2841, "lr": 1.1534196130483344e-06}
{"train_loss": 0.02952686697244644, "global_step": 252889, "epoch": 2841, "lr": 1.1532958121802495e-06}
{"train_loss": 0.04154367744922638, "global_step": 252890, "epoch": 2841, "lr": 1.1531720178790062e-06}
{"train_loss": 0.05031967535614967, "global_step": 252891, "epoch": 2841, "lr": 1.153048230144621e-06}
{"train_loss": 0.02808464504778385, "global_step": 252892, "epoch": 2841, "lr": 1.1529244489771162e-06}
{"train_loss": 0.04896813631057739, "global_step": 252893, "epoch": 2841, "lr": 1.1528006743765029e-06}
{"train_loss": 0.027752069756388664, "global_step": 252894, "epoch": 2841, "lr": 1.1526769063428034e-06}
{"train_loss": 0.03307963162660599, "global_step": 252895, "epoch": 2841, "lr": 1.1525531448760284e-06}
{"train_loss": 0.049933675676584244, "global_step": 252896, "epoch": 2841, "lr": 1.1524293899762006e-06}
{"train_loss": 0.03967021778225899, "global_step": 252897, "epoch": 2841, "lr": 1.1523056416433253e-06}
{"train_loss": 0.021915197372436523, "global_step": 252898, "epoch": 2841, "lr": 1.1521818998774358e-06}
{"train_loss": 0.03833116963505745, "global_step": 252899, "epoch": 2841, "lr": 1.1520581646785323e-06}
{"train_loss": 0.06167447566986084, "global_step": 252900, "epoch": 2841, "lr": 1.1519344360466477e-06}
{"train_loss": 0.042952924966812134, "global_step": 252901, "epoch": 2841, "lr": 1.1518107139817825e-06}
{"train_loss": 0.01540346723049879, "global_step": 252902, "epoch": 2841, "lr": 1.1516869984839696e-06}
{"train_loss": 0.018576471135020256, "global_step": 252903, "epoch": 2841, "lr": 1.1515632895532091e-06}
{"train_loss": 0.015009209513664246, "global_step": 252904, "epoch": 2841, "lr": 1.151439587189529e-06}
{"train_loss": 0.05098668485879898, "global_step": 252905, "epoch": 2841, "lr": 1.15131589139294e-06}
{"train_loss": 0.045347705483436584, "global_step": 252906, "epoch": 2841, "lr": 1.1511922021634646e-06}
{"train_loss": 0.054038576781749725, "global_step": 252907, "epoch": 2841, "lr": 1.151068519501114e-06}
{"train_loss": 0.06802593916654587, "global_step": 252908, "epoch": 2841, "lr": 1.1509448434059045e-06}
{"train_loss": 0.03480443358421326, "global_step": 252909, "epoch": 2841, "lr": 1.1508211738778585e-06}
{"train_loss": 0.061225004494190216, "global_step": 252910, "epoch": 2841, "lr": 1.150697510916987e-06}
{"train_loss": 0.04084629565477371, "global_step": 252911, "epoch": 2841, "lr": 1.1505738545233125e-06}
{"train_loss": 0.06035234034061432, "global_step": 252912, "epoch": 2841, "lr": 1.1504502046968458e-06}
{"train_loss": 0.10247453302145004, "global_step": 252913, "epoch": 2841, "lr": 1.1503265614376091e-06}
{"train_loss": 0.009124258533120155, "global_step": 252914, "epoch": 2841, "lr": 1.150202924745608e-06}
{"train_loss": 0.0519169345498085, "global_step": 252915, "epoch": 2841, "lr": 1.150079294620876e-06}
{"train_loss": 0.05478439852595329, "global_step": 252916, "epoch": 2841, "lr": 1.149955671063413e-06}
{"train_loss": 0.04163646325469017, "global_step": 252917, "epoch": 2841, "lr": 1.1498320540732465e-06}
{"train_loss": 0.031047143042087555, "global_step": 252918, "epoch": 2841, "lr": 1.149708443650388e-06}
{"train_loss": 0.040857601910829544, "global_step": 252919, "epoch": 2841, "lr": 1.149584839794854e-06}
{"train_loss": 0.023062441498041153, "global_step": 252920, "epoch": 2841, "lr": 1.1494612425066665e-06}
{"train_loss": 0.03821970894932747, "global_step": 252921, "epoch": 2841, "lr": 1.1493376517858368e-06}
{"train_loss": 0.027417723089456558, "global_step": 252922, "epoch": 2841, "lr": 1.1492140676323815e-06}
{"train_loss": 0.0746995210647583, "global_step": 252923, "epoch": 2841, "lr": 1.149090490046323e-06}
{"train_loss": 0.04808322340250015, "global_step": 252924, "epoch": 2841, "lr": 1.148966919027672e-06}
{"train_loss": 0.07650180906057358, "global_step": 252925, "epoch": 2841, "lr": 1.148843354576451e-06}
{"train_loss": 0.04939880967140198, "global_step": 252926, "epoch": 2841, "lr": 1.1487197966926656e-06}
{"train_loss": 0.11470651626586914, "global_step": 252927, "epoch": 2841, "lr": 1.1485962453763432e-06}
{"train_loss": 0.07223466783761978, "global_step": 252928, "epoch": 2841, "lr": 1.1484727006275008e-06}
{"train_loss": 0.031842637807130814, "global_step": 252929, "epoch": 2841, "lr": 1.148349162446144e-06}
{"train_loss": 0.05717787146568298, "global_step": 252930, "epoch": 2841, "lr": 1.1482256308323004e-06}
{"train_loss": 0.057868193835020065, "global_step": 252931, "epoch": 2841, "lr": 1.1481021057859754e-06}
{"train_loss": 0.02376563474535942, "global_step": 252932, "epoch": 2841, "lr": 1.1479785873072023e-06}
{"train_loss": 0.03706740587949753, "global_step": 252933, "epoch": 2841, "lr": 1.1478550753959816e-06}
{"train_loss": 0.04254244267940521, "global_step": 252934, "epoch": 2841, "lr": 1.1477315700523405e-06}
{"train_loss": 0.04412026330828667, "global_step": 252935, "epoch": 2841, "lr": 1.147608071276285e-06}
{"train_loss": 0.057845063507556915, "global_step": 252936, "epoch": 2841, "lr": 1.1474845790678479e-06}
{"train_loss": 0.04487979274996546, "global_step": 252937, "epoch": 2841, "lr": 1.147361093427024e-06, "val_loss": 9.563416481018066}
{"train_loss": 0.03894467651844025, "global_step": 252938, "epoch": 2842, "lr": 1.1472376143538521e-06}
{"train_loss": 0.02735026553273201, "global_step": 252939, "epoch": 2842, "lr": 1.1471141418483322e-06}
{"train_loss": 0.08024393767118454, "global_step": 252940, "epoch": 2842, "lr": 1.146990675910492e-06}
{"train_loss": 0.03860625624656677, "global_step": 252941, "epoch": 2842, "lr": 1.1468672165403427e-06}
{"train_loss": 0.07068395614624023, "global_step": 252942, "epoch": 2842, "lr": 1.1467437637379008e-06}
{"train_loss": 0.05185364931821823, "global_step": 252943, "epoch": 2842, "lr": 1.1466203175031832e-06}
{"train_loss": 0.0337500162422657, "global_step": 252944, "epoch": 2842, "lr": 1.1464968778362006e-06}
{"train_loss": 0.02307366393506527, "global_step": 252945, "epoch": 2842, "lr": 1.1463734447369868e-06}
{"train_loss": 0.08823796361684799, "global_step": 252946, "epoch": 2842, "lr": 1.1462500182055413e-06}
{"train_loss": 0.05389560014009476, "global_step": 252947, "epoch": 2842, "lr": 1.1461265982418923e-06}
{"train_loss": 0.049655988812446594, "global_step": 252948, "epoch": 2842, "lr": 1.1460031848460451e-06}
{"train_loss": 0.06070453301072121, "global_step": 252949, "epoch": 2842, "lr": 1.1458797780180274e-06}
{"train_loss": 0.05532851442694664, "global_step": 252950, "epoch": 2842, "lr": 1.1457563777578507e-06}
{"train_loss": 0.024636179208755493, "global_step": 252951, "epoch": 2842, "lr": 1.145632984065531e-06}
{"train_loss": 0.026787981390953064, "global_step": 252952, "epoch": 2842, "lr": 1.14550959694108e-06}
{"train_loss": 0.028912514448165894, "global_step": 252953, "epoch": 2842, "lr": 1.1453862163845253e-06}
{"train_loss": 0.04653569310903549, "global_step": 252954, "epoch": 2842, "lr": 1.1452628423958777e-06}
{"train_loss": 0.052454814314842224, "global_step": 252955, "epoch": 2842, "lr": 1.1451394749751542e-06}
{"train_loss": 0.04014522582292557, "global_step": 252956, "epoch": 2842, "lr": 1.1450161141223658e-06}
{"train_loss": 0.06893908232450485, "global_step": 252957, "epoch": 2842, "lr": 1.1448927598375404e-06}
{"train_loss": 0.039605457335710526, "global_step": 252958, "epoch": 2842, "lr": 1.1447694121206887e-06}
{"train_loss": 0.024834970012307167, "global_step": 252959, "epoch": 2842, "lr": 1.1446460709718276e-06}
{"train_loss": 0.059940315783023834, "global_step": 252960, "epoch": 2842, "lr": 1.1445227363909683e-06}
{"train_loss": 0.023104380816221237, "global_step": 252961, "epoch": 2842, "lr": 1.1443994083781384e-06}
{"train_loss": 0.03265795856714249, "global_step": 252962, "epoch": 2842, "lr": 1.1442760869333436e-06}
{"train_loss": 0.047848187386989594, "global_step": 252963, "epoch": 2842, "lr": 1.144152772056606e-06}
{"train_loss": 0.04731489717960358, "global_step": 252964, "epoch": 2842, "lr": 1.144029463747942e-06}
{"train_loss": 0.07037763297557831, "global_step": 252965, "epoch": 2842, "lr": 1.1439061620073744e-06}
{"train_loss": 0.01792050525546074, "global_step": 252966, "epoch": 2842, "lr": 1.1437828668349082e-06}
{"train_loss": 0.07363826036453247, "global_step": 252967, "epoch": 2842, "lr": 1.1436595782305603e-06}
{"train_loss": 0.029789399355649948, "global_step": 252968, "epoch": 2842, "lr": 1.1435362961943586e-06}
{"train_loss": 0.04636384919285774, "global_step": 252969, "epoch": 2842, "lr": 1.1434130207263082e-06}
{"train_loss": 0.03830491378903389, "global_step": 252970, "epoch": 2842, "lr": 1.1432897518264375e-06}
{"train_loss": 0.03922980651259422, "global_step": 252971, "epoch": 2842, "lr": 1.1431664894947458e-06}
{"train_loss": 0.00480936374515295, "global_step": 252972, "epoch": 2842, "lr": 1.143043233731267e-06}
{"train_loss": 0.06881542503833771, "global_step": 252973, "epoch": 2842, "lr": 1.1429199845360062e-06}
{"train_loss": 0.0499144084751606, "global_step": 252974, "epoch": 2842, "lr": 1.1427967419089913e-06}
{"train_loss": 0.05513196066021919, "global_step": 252975, "epoch": 2842, "lr": 1.1426735058502225e-06}
{"train_loss": 0.008873933926224709, "global_step": 252976, "epoch": 2842, "lr": 1.142550276359733e-06}
{"train_loss": 0.023438483476638794, "global_step": 252977, "epoch": 2842, "lr": 1.1424270534375281e-06}
{"train_loss": 0.03650829568505287, "global_step": 252978, "epoch": 2842, "lr": 1.142303837083636e-06}
{"train_loss": 0.034331224858760834, "global_step": 252979, "epoch": 2842, "lr": 1.1421806272980563e-06}
{"train_loss": 0.04870332032442093, "global_step": 252980, "epoch": 2842, "lr": 1.1420574240808169e-06}
{"train_loss": 0.041572727262973785, "global_step": 252981, "epoch": 2842, "lr": 1.1419342274319345e-06}
{"train_loss": 0.03225026652216911, "global_step": 252982, "epoch": 2842, "lr": 1.1418110373514201e-06}
{"train_loss": 0.09240839630365372, "global_step": 252983, "epoch": 2842, "lr": 1.1416878538393018e-06}
{"train_loss": 0.06851529330015182, "global_step": 252984, "epoch": 2842, "lr": 1.141564676895579e-06}
{"train_loss": 0.05433179438114166, "global_step": 252985, "epoch": 2842, "lr": 1.1414415065202855e-06}
{"train_loss": 0.02334635704755783, "global_step": 252986, "epoch": 2842, "lr": 1.1413183427134211e-06}
{"train_loss": 0.008931257762014866, "global_step": 252987, "epoch": 2842, "lr": 1.1411951854750192e-06}
{"train_loss": 0.046729229390621185, "global_step": 252988, "epoch": 2842, "lr": 1.1410720348050796e-06}
{"train_loss": 0.008629083633422852, "global_step": 252989, "epoch": 2842, "lr": 1.1409488907036358e-06}
{"train_loss": 0.043931830674409866, "global_step": 252990, "epoch": 2842, "lr": 1.1408257531706878e-06}
{"train_loss": 0.05130312219262123, "global_step": 252991, "epoch": 2842, "lr": 1.140702622206269e-06}
{"train_loss": 0.0703263208270073, "global_step": 252992, "epoch": 2842, "lr": 1.140579497810379e-06}
{"train_loss": 0.04446295276284218, "global_step": 252993, "epoch": 2842, "lr": 1.1404563799830459e-06}
{"train_loss": 0.0527389720082283, "global_step": 252994, "epoch": 2842, "lr": 1.1403332687242806e-06}
{"train_loss": 0.0401955209672451, "global_step": 252995, "epoch": 2842, "lr": 1.1402101640341112e-06}
{"train_loss": 0.03403196483850479, "global_step": 252996, "epoch": 2842, "lr": 1.1400870659125318e-06}
{"train_loss": 0.058547090739011765, "global_step": 252997, "epoch": 2842, "lr": 1.1399639743595814e-06}
{"train_loss": 0.04834655299782753, "global_step": 252998, "epoch": 2842, "lr": 1.1398408893752655e-06}
{"train_loss": 0.043083105236291885, "global_step": 252999, "epoch": 2842, "lr": 1.1397178109596008e-06}
{"train_loss": 0.06372981518507004, "global_step": 253000, "epoch": 2842, "lr": 1.1395947391126039e-06}
{"train_loss": 0.021543366834521294, "global_step": 253001, "epoch": 2842, "lr": 1.1394716738343026e-06}
{"train_loss": 0.05073362588882446, "global_step": 253002, "epoch": 2842, "lr": 1.1393486151246912e-06}
{"train_loss": 0.05026181787252426, "global_step": 253003, "epoch": 2842, "lr": 1.1392255629838089e-06}
{"train_loss": 0.053291093558073044, "global_step": 253004, "epoch": 2842, "lr": 1.1391025174116554e-06}
{"train_loss": 0.02900078520178795, "global_step": 253005, "epoch": 2842, "lr": 1.1389794784082586e-06}
{"train_loss": 0.043967749923467636, "global_step": 253006, "epoch": 2842, "lr": 1.1388564459736295e-06}
{"train_loss": 0.03762577474117279, "global_step": 253007, "epoch": 2842, "lr": 1.1387334201077793e-06}
{"train_loss": 0.053523678332567215, "global_step": 253008, "epoch": 2842, "lr": 1.1386104008107412e-06}
{"train_loss": 0.042054433375597, "global_step": 253009, "epoch": 2842, "lr": 1.1384873880825097e-06}
{"train_loss": 0.02325686812400818, "global_step": 253010, "epoch": 2842, "lr": 1.1383643819231239e-06}
{"train_loss": 0.029223229736089706, "global_step": 253011, "epoch": 2842, "lr": 1.1382413823325832e-06}
{"train_loss": 0.040412165224552155, "global_step": 253012, "epoch": 2842, "lr": 1.138118389310916e-06}
{"train_loss": 0.06098451465368271, "global_step": 253013, "epoch": 2842, "lr": 1.1379954028581274e-06}
{"train_loss": 0.038252685219049454, "global_step": 253014, "epoch": 2842, "lr": 1.13787242297424e-06}
{"train_loss": 0.06980237364768982, "global_step": 253015, "epoch": 2842, "lr": 1.13774944965927e-06}
{"train_loss": 0.03654620423913002, "global_step": 253016, "epoch": 2842, "lr": 1.1376264829132343e-06}
{"train_loss": 0.07938453555107117, "global_step": 253017, "epoch": 2842, "lr": 1.1375035227361553e-06}
{"train_loss": 0.05860792472958565, "global_step": 253018, "epoch": 2842, "lr": 1.1373805691280327e-06}
{"train_loss": 0.03290392830967903, "global_step": 253019, "epoch": 2842, "lr": 1.1372576220889053e-06}
{"train_loss": 0.04655683785676956, "global_step": 253020, "epoch": 2842, "lr": 1.1371346816187677e-06}
{"train_loss": 0.015818145126104355, "global_step": 253021, "epoch": 2842, "lr": 1.1370117477176533e-06}
{"train_loss": 0.021011317148804665, "global_step": 253022, "epoch": 2842, "lr": 1.1368888203855732e-06}
{"train_loss": 0.02150607295334339, "global_step": 253023, "epoch": 2842, "lr": 1.1367658996225382e-06}
{"train_loss": 0.04001972824335098, "global_step": 253024, "epoch": 2842, "lr": 1.136642985428571e-06}
{"train_loss": 0.0500587522983551, "global_step": 253025, "epoch": 2842, "lr": 1.1365200778036877e-06}
{"train_loss": 0.04436669230795978, "global_step": 253026, "epoch": 2842, "lr": 1.1363971767479053e-06, "val_loss": 9.527115821838379}
{"train_loss": 0.04634680226445198, "global_step": 253027, "epoch": 2843, "lr": 1.1362742822612348e-06}
{"train_loss": 0.046411167830228806, "global_step": 253028, "epoch": 2843, "lr": 1.1361513943436985e-06}
{"train_loss": 0.03172683343291283, "global_step": 253029, "epoch": 2843, "lr": 1.136028512995313e-06}
{"train_loss": 0.018224947154521942, "global_step": 253030, "epoch": 2843, "lr": 1.1359056382160892e-06}
{"train_loss": 0.0284564308822155, "global_step": 253031, "epoch": 2843, "lr": 1.1357827700060497e-06}
{"train_loss": 0.04345728084445, "global_step": 253032, "epoch": 2843, "lr": 1.1356599083652053e-06}
{"train_loss": 0.0639115571975708, "global_step": 253033, "epoch": 2843, "lr": 1.1355370532935783e-06}
{"train_loss": 0.036302756518125534, "global_step": 253034, "epoch": 2843, "lr": 1.1354142047911853e-06}
{"train_loss": 0.06506121903657913, "global_step": 253035, "epoch": 2843, "lr": 1.1352913628580318e-06}
{"train_loss": 0.015603662468492985, "global_step": 253036, "epoch": 2843, "lr": 1.1351685274941514e-06}
{"train_loss": 0.06831260770559311, "global_step": 253037, "epoch": 2843, "lr": 1.1350456986995494e-06}
{"train_loss": 0.051508501172065735, "global_step": 253038, "epoch": 2843, "lr": 1.1349228764742481e-06}
{"train_loss": 0.030827302485704422, "global_step": 253039, "epoch": 2843, "lr": 1.1348000608182585e-06}
{"train_loss": 0.018141692504286766, "global_step": 253040, "epoch": 2843, "lr": 1.1346772517315973e-06}
{"train_loss": 0.018115729093551636, "global_step": 253041, "epoch": 2843, "lr": 1.1345544492142868e-06}
{"train_loss": 0.04335992783308029, "global_step": 253042, "epoch": 2843, "lr": 1.1344316532663324e-06}
{"train_loss": 0.028195245191454887, "global_step": 253043, "epoch": 2843, "lr": 1.1343088638877676e-06}
{"train_loss": 0.04659438133239746, "global_step": 253044, "epoch": 2843, "lr": 1.1341860810785976e-06}
{"train_loss": 0.025991568341851234, "global_step": 253045, "epoch": 2843, "lr": 1.1340633048388338e-06}
{"train_loss": 0.03684535250067711, "global_step": 253046, "epoch": 2843, "lr": 1.1339405351685094e-06}
{"train_loss": 0.016820188611745834, "global_step": 253047, "epoch": 2843, "lr": 1.1338177720676247e-06}
{"train_loss": 0.030271049588918686, "global_step": 253048, "epoch": 2843, "lr": 1.1336950155362014e-06}
{"train_loss": 0.03108731471002102, "global_step": 253049, "epoch": 2843, "lr": 1.133572265574262e-06}
{"train_loss": 0.06320597231388092, "global_step": 253050, "epoch": 2843, "lr": 1.1334495221818176e-06}
{"train_loss": 0.022504117339849472, "global_step": 253051, "epoch": 2843, "lr": 1.133326785358879e-06}
{"train_loss": 0.042681120336055756, "global_step": 253052, "epoch": 2843, "lr": 1.133204055105469e-06}
{"train_loss": 0.018758611753582954, "global_step": 253053, "epoch": 2843, "lr": 1.1330813314216148e-06}
{"train_loss": 0.03262307122349739, "global_step": 253054, "epoch": 2843, "lr": 1.1329586143073112e-06}
{"train_loss": 0.08709452301263809, "global_step": 253055, "epoch": 2843, "lr": 1.1328359037625913e-06}
{"train_loss": 0.07534915953874588, "global_step": 253056, "epoch": 2843, "lr": 1.1327131997874663e-06}
{"train_loss": 0.03214104846119881, "global_step": 253057, "epoch": 2843, "lr": 1.1325905023819528e-06}
{"train_loss": 0.029717300087213516, "global_step": 253058, "epoch": 2843, "lr": 1.1324678115460618e-06}
{"train_loss": 0.05419383570551872, "global_step": 253059, "epoch": 2843, "lr": 1.1323451272798214e-06}
{"train_loss": 0.04457763209939003, "global_step": 253060, "epoch": 2843, "lr": 1.1322224495832368e-06}
{"train_loss": 0.034716226160526276, "global_step": 253061, "epoch": 2843, "lr": 1.1320997784563303e-06}
{"train_loss": 0.11848460882902145, "global_step": 253062, "epoch": 2843, "lr": 1.1319771138991187e-06}
{"train_loss": 0.05296565964818001, "global_step": 253063, "epoch": 2843, "lr": 1.1318544559116183e-06}
{"train_loss": 0.030050717294216156, "global_step": 253064, "epoch": 2843, "lr": 1.1317318044938408e-06}
{"train_loss": 0.03927665948867798, "global_step": 253065, "epoch": 2843, "lr": 1.1316091596458078e-06}
{"train_loss": 0.0389421321451664, "global_step": 253066, "epoch": 2843, "lr": 1.1314865213675363e-06}
{"train_loss": 0.016848353669047356, "global_step": 253067, "epoch": 2843, "lr": 1.1313638896590372e-06}
{"train_loss": 0.04829166457056999, "global_step": 253068, "epoch": 2843, "lr": 1.131241264520333e-06}
{"train_loss": 0.02661930024623871, "global_step": 253069, "epoch": 2843, "lr": 1.13111864595144e-06}
{"train_loss": 0.052336275577545166, "global_step": 253070, "epoch": 2843, "lr": 1.130996033952364e-06}
{"train_loss": 0.04830428957939148, "global_step": 253071, "epoch": 2843, "lr": 1.1308734285231381e-06}
{"train_loss": 0.05590686574578285, "global_step": 253072, "epoch": 2843, "lr": 1.1307508296637681e-06}
{"train_loss": 0.07205545902252197, "global_step": 253073, "epoch": 2843, "lr": 1.1306282373742705e-06}
{"train_loss": 0.05059332773089409, "global_step": 253074, "epoch": 2843, "lr": 1.1305056516546731e-06}
{"train_loss": 0.047572895884513855, "global_step": 253075, "epoch": 2843, "lr": 1.1303830725049701e-06}
{"train_loss": 0.056829214096069336, "global_step": 253076, "epoch": 2843, "lr": 1.1302604999252064e-06}
{"train_loss": 0.00895980279892683, "global_step": 253077, "epoch": 2843, "lr": 1.130137933915376e-06}
{"train_loss": 0.06534505635499954, "global_step": 253078, "epoch": 2843, "lr": 1.1300153744755015e-06}
{"train_loss": 0.04580489173531532, "global_step": 253079, "epoch": 2843, "lr": 1.129892821605605e-06}
{"train_loss": 0.049644943326711655, "global_step": 253080, "epoch": 2843, "lr": 1.1297702753056915e-06}
{"train_loss": 0.06787261366844177, "global_step": 253081, "epoch": 2843, "lr": 1.129647735575795e-06}
{"train_loss": 0.04180596396327019, "global_step": 253082, "epoch": 2843, "lr": 1.1295252024159153e-06}
{"train_loss": 0.05129250884056091, "global_step": 253083, "epoch": 2843, "lr": 1.1294026758260746e-06}
{"train_loss": 0.032621294260025024, "global_step": 253084, "epoch": 2843, "lr": 1.129280155806295e-06}
{"train_loss": 0.0544310100376606, "global_step": 253085, "epoch": 2843, "lr": 1.129157642356582e-06}
{"train_loss": 0.04435352981090546, "global_step": 253086, "epoch": 2843, "lr": 1.1290351354769634e-06}
{"train_loss": 0.009182166308164597, "global_step": 253087, "epoch": 2843, "lr": 1.128912635167445e-06}
{"train_loss": 0.06218798831105232, "global_step": 253088, "epoch": 2843, "lr": 1.1287901414280545e-06}
{"train_loss": 0.059292297810316086, "global_step": 253089, "epoch": 2843, "lr": 1.128667654258797e-06}
{"train_loss": 0.13925990462303162, "global_step": 253090, "epoch": 2843, "lr": 1.1285451736596897e-06}
{"train_loss": 0.023256246000528336, "global_step": 253091, "epoch": 2843, "lr": 1.1284226996307657e-06}
{"train_loss": 0.024603508412837982, "global_step": 253092, "epoch": 2843, "lr": 1.1283002321720248e-06}
{"train_loss": 0.042488422244787216, "global_step": 253093, "epoch": 2843, "lr": 1.1281777712834896e-06}
{"train_loss": 0.037176426500082016, "global_step": 253094, "epoch": 2843, "lr": 1.128055316965171e-06}
{"train_loss": 0.017184918746352196, "global_step": 253095, "epoch": 2843, "lr": 1.1279328692170966e-06}
{"train_loss": 0.02318507432937622, "global_step": 253096, "epoch": 2843, "lr": 1.1278104280392665e-06}
{"train_loss": 0.06431930512189865, "global_step": 253097, "epoch": 2843, "lr": 1.1276879934317143e-06}
{"train_loss": 0.03276683762669563, "global_step": 253098, "epoch": 2843, "lr": 1.127565565394445e-06}
{"train_loss": 0.0322817899286747, "global_step": 253099, "epoch": 2843, "lr": 1.1274431439274813e-06}
{"train_loss": 0.059896886348724365, "global_step": 253100, "epoch": 2843, "lr": 1.127320729030834e-06}
{"train_loss": 0.11426284909248352, "global_step": 253101, "epoch": 2843, "lr": 1.1271983207045256e-06}
{"train_loss": 0.02647061087191105, "global_step": 253102, "epoch": 2843, "lr": 1.1270759189485669e-06}
{"train_loss": 0.04418175294995308, "global_step": 253103, "epoch": 2843, "lr": 1.1269535237629857e-06}
{"train_loss": 0.03249197080731392, "global_step": 253104, "epoch": 2843, "lr": 1.1268311351477822e-06}
{"train_loss": 0.03224356472492218, "global_step": 253105, "epoch": 2843, "lr": 1.126708753102984e-06}
{"train_loss": 0.023976245895028114, "global_step": 253106, "epoch": 2843, "lr": 1.1265863776285967e-06}
{"train_loss": 0.06132357940077782, "global_step": 253107, "epoch": 2843, "lr": 1.1264640087246481e-06}
{"train_loss": 0.03439931944012642, "global_step": 253108, "epoch": 2843, "lr": 1.1263416463911547e-06}
{"train_loss": 0.025310782715678215, "global_step": 253109, "epoch": 2843, "lr": 1.1262192906281222e-06}
{"train_loss": 0.08668436855077744, "global_step": 253110, "epoch": 2843, "lr": 1.1260969414355836e-06}
{"train_loss": 0.05979321151971817, "global_step": 253111, "epoch": 2843, "lr": 1.125974598813534e-06}
{"train_loss": 0.040917541831731796, "global_step": 253112, "epoch": 2843, "lr": 1.1258522627620115e-06}
{"train_loss": 0.042534321546554565, "global_step": 253113, "epoch": 2843, "lr": 1.1257299332810167e-06}
{"train_loss": 0.03174961730837822, "global_step": 253114, "epoch": 2843, "lr": 1.125607610370577e-06}
{"train_loss": 0.04381663806401612, "global_step": 253115, "epoch": 2843, "lr": 1.125485294030698e-06, "val_loss": 9.636092185974121}
{"train_loss": 0.040406156331300735, "global_step": 253116, "epoch": 2844, "lr": 1.1253629842614077e-06}
{"train_loss": 0.020785098895430565, "global_step": 253117, "epoch": 2844, "lr": 1.1252406810627115e-06}
{"train_loss": 0.02499277889728546, "global_step": 253118, "epoch": 2844, "lr": 1.1251183844346313e-06}
{"train_loss": 0.03248346596956253, "global_step": 253119, "epoch": 2844, "lr": 1.1249960943771897e-06}
{"train_loss": 0.09685833007097244, "global_step": 253120, "epoch": 2844, "lr": 1.1248738108903867e-06}
{"train_loss": 0.020526470616459846, "global_step": 253121, "epoch": 2844, "lr": 1.1247515339742555e-06}
{"train_loss": 0.03905604034662247, "global_step": 253122, "epoch": 2844, "lr": 1.124629263628807e-06}
{"train_loss": 0.027005739510059357, "global_step": 253123, "epoch": 2844, "lr": 1.1245069998540526e-06}
{"train_loss": 0.046755049377679825, "global_step": 253124, "epoch": 2844, "lr": 1.1243847426500143e-06}
{"train_loss": 0.04149410501122475, "global_step": 253125, "epoch": 2844, "lr": 1.1242624920167034e-06}
{"train_loss": 0.03814084455370903, "global_step": 253126, "epoch": 2844, "lr": 1.1241402479541364e-06}
{"train_loss": 0.0324094220995903, "global_step": 253127, "epoch": 2844, "lr": 1.124018010462341e-06}
{"train_loss": 0.061406541615724564, "global_step": 253128, "epoch": 2844, "lr": 1.123895779541323e-06}
{"train_loss": 0.016774997115135193, "global_step": 253129, "epoch": 2844, "lr": 1.1237735551911044e-06}
{"train_loss": 0.07104546576738358, "global_step": 253130, "epoch": 2844, "lr": 1.1236513374116907e-06}
{"train_loss": 0.09814261645078659, "global_step": 253131, "epoch": 2844, "lr": 1.1235291262031157e-06}
{"train_loss": 0.04425930231809616, "global_step": 253132, "epoch": 2844, "lr": 1.1234069215653785e-06}
{"train_loss": 0.09377741068601608, "global_step": 253133, "epoch": 2844, "lr": 1.1232847234985133e-06}
{"train_loss": 0.08179794996976852, "global_step": 253134, "epoch": 2844, "lr": 1.1231625320025141e-06}
{"train_loss": 0.01745949126780033, "global_step": 253135, "epoch": 2844, "lr": 1.1230403470774199e-06}
{"train_loss": 0.040274012833833694, "global_step": 253136, "epoch": 2844, "lr": 1.1229181687232303e-06}
{"train_loss": 0.035227466374635696, "global_step": 253137, "epoch": 2844, "lr": 1.1227959969399793e-06}
{"train_loss": 0.07077967375516891, "global_step": 253138, "epoch": 2844, "lr": 1.122673831727661e-06}
{"train_loss": 0.04319743067026138, "global_step": 253139, "epoch": 2844, "lr": 1.1225516730863084e-06}
{"train_loss": 0.04637185111641884, "global_step": 253140, "epoch": 2844, "lr": 1.122429521015933e-06}
{"train_loss": 0.050088487565517426, "global_step": 253141, "epoch": 2844, "lr": 1.1223073755165514e-06}
{"train_loss": 0.011230849660933018, "global_step": 253142, "epoch": 2844, "lr": 1.1221852365881802e-06}
{"train_loss": 0.04841909185051918, "global_step": 253143, "epoch": 2844, "lr": 1.1220631042308305e-06}
{"train_loss": 0.04756449908018112, "global_step": 253144, "epoch": 2844, "lr": 1.12194097844453e-06}
{"train_loss": 0.05344078689813614, "global_step": 253145, "epoch": 2844, "lr": 1.1218188592292845e-06}
{"train_loss": 0.08765891194343567, "global_step": 253146, "epoch": 2844, "lr": 1.1216967465851158e-06}
{"train_loss": 0.033010803163051605, "global_step": 253147, "epoch": 2844, "lr": 1.1215746405120408e-06}
{"train_loss": 0.030981816351413727, "global_step": 253148, "epoch": 2844, "lr": 1.1214525410100762e-06}
{"train_loss": 0.046299148350954056, "global_step": 253149, "epoch": 2844, "lr": 1.121330448079233e-06}
{"train_loss": 0.02713564969599247, "global_step": 253150, "epoch": 2844, "lr": 1.1212083617195334e-06}
{"train_loss": 0.0706818476319313, "global_step": 253151, "epoch": 2844, "lr": 1.1210862819309886e-06}
{"train_loss": 0.03353138267993927, "global_step": 253152, "epoch": 2844, "lr": 1.1209642087136207e-06}
{"train_loss": 0.03503500297665596, "global_step": 253153, "epoch": 2844, "lr": 1.1208421420674409e-06}
{"train_loss": 0.046596359461545944, "global_step": 253154, "epoch": 2844, "lr": 1.1207200819924713e-06}
{"train_loss": 0.061953380703926086, "global_step": 253155, "epoch": 2844, "lr": 1.1205980284887284e-06}
{"train_loss": 0.03944792225956917, "global_step": 253156, "epoch": 2844, "lr": 1.1204759815562183e-06}
{"train_loss": 0.07741223275661469, "global_step": 253157, "epoch": 2844, "lr": 1.1203539411949681e-06}
{"train_loss": 0.033933401107788086, "global_step": 253158, "epoch": 2844, "lr": 1.120231907404984e-06}
{"train_loss": 0.03121344931423664, "global_step": 253159, "epoch": 2844, "lr": 1.1201098801862987e-06}
{"train_loss": 0.011849837377667427, "global_step": 253160, "epoch": 2844, "lr": 1.119987859538918e-06}
{"train_loss": 0.027076764032244682, "global_step": 253161, "epoch": 2844, "lr": 1.119865845462853e-06}
{"train_loss": 0.01618019863963127, "global_step": 253162, "epoch": 2844, "lr": 1.119743837958126e-06}
{"train_loss": 0.02947482466697693, "global_step": 253163, "epoch": 2844, "lr": 1.1196218370247592e-06}
{"train_loss": 0.03475319966673851, "global_step": 253164, "epoch": 2844, "lr": 1.1194998426627579e-06}
{"train_loss": 0.06238122284412384, "global_step": 253165, "epoch": 2844, "lr": 1.1193778548721502e-06}
{"train_loss": 0.06337826699018478, "global_step": 253166, "epoch": 2844, "lr": 1.1192558736529412e-06}
{"train_loss": 0.071449413895607, "global_step": 253167, "epoch": 2844, "lr": 1.1191338990051536e-06}
{"train_loss": 0.08270769566297531, "global_step": 253168, "epoch": 2844, "lr": 1.1190119309288039e-06}
{"train_loss": 0.07041586190462112, "global_step": 253169, "epoch": 2844, "lr": 1.1188899694239086e-06}
{"train_loss": 0.02912754938006401, "global_step": 253170, "epoch": 2844, "lr": 1.1187680144904789e-06}
{"train_loss": 0.05460117384791374, "global_step": 253171, "epoch": 2844, "lr": 1.118646066128537e-06}
{"train_loss": 0.04399268701672554, "global_step": 253172, "epoch": 2844, "lr": 1.1185241243380939e-06}
{"train_loss": 0.09373243898153305, "global_step": 253173, "epoch": 2844, "lr": 1.1184021891191777e-06}
{"train_loss": 0.04317906126379967, "global_step": 253174, "epoch": 2844, "lr": 1.118280260471788e-06}
{"train_loss": 0.03176314756274223, "global_step": 253175, "epoch": 2844, "lr": 1.1181583383959583e-06}
{"train_loss": 0.019819997251033783, "global_step": 253176, "epoch": 2844, "lr": 1.1180364228916884e-06}
{"train_loss": 0.031053833663463593, "global_step": 253177, "epoch": 2844, "lr": 1.1179145139590064e-06}
{"train_loss": 0.039549924433231354, "global_step": 253178, "epoch": 2844, "lr": 1.1177926115979232e-06}
{"train_loss": 0.04375039413571358, "global_step": 253179, "epoch": 2844, "lr": 1.1176707158084553e-06}
{"train_loss": 0.016274498775601387, "global_step": 253180, "epoch": 2844, "lr": 1.1175488265906252e-06}
{"train_loss": 0.023227469995617867, "global_step": 253181, "epoch": 2844, "lr": 1.117426943944444e-06}
{"train_loss": 0.08598396182060242, "global_step": 253182, "epoch": 2844, "lr": 1.1173050678699282e-06}
{"train_loss": 0.04880882799625397, "global_step": 253183, "epoch": 2844, "lr": 1.1171831983670945e-06}
{"train_loss": 0.07020189613103867, "global_step": 253184, "epoch": 2844, "lr": 1.117061335435965e-06}
{"train_loss": 0.027860775589942932, "global_step": 253185, "epoch": 2844, "lr": 1.1169394790765398e-06}
{"train_loss": 0.06956510990858078, "global_step": 253186, "epoch": 2844, "lr": 1.1168176292888577e-06}
{"train_loss": 0.04229535162448883, "global_step": 253187, "epoch": 2844, "lr": 1.1166957860729187e-06}
{"train_loss": 0.10581459105014801, "global_step": 253188, "epoch": 2844, "lr": 1.1165739494287453e-06}
{"train_loss": 0.018806278705596924, "global_step": 253189, "epoch": 2844, "lr": 1.1164521193563481e-06}
{"train_loss": 0.011171695776283741, "global_step": 253190, "epoch": 2844, "lr": 1.1163302958557554e-06}
{"train_loss": 0.030826102942228317, "global_step": 253191, "epoch": 2844, "lr": 1.1162084789269667e-06}
{"train_loss": 0.02728847600519657, "global_step": 253192, "epoch": 2844, "lr": 1.1160866685700156e-06}
{"train_loss": 0.04974209889769554, "global_step": 253193, "epoch": 2844, "lr": 1.115964864784913e-06}
{"train_loss": 0.09298934042453766, "global_step": 253194, "epoch": 2844, "lr": 1.1158430675716702e-06}
{"train_loss": 0.025715481489896774, "global_step": 253195, "epoch": 2844, "lr": 1.1157212769303038e-06}
{"train_loss": 0.02323591336607933, "global_step": 253196, "epoch": 2844, "lr": 1.115599492860836e-06}
{"train_loss": 0.024544835090637207, "global_step": 253197, "epoch": 2844, "lr": 1.115477715363278e-06}
{"train_loss": 0.06028328835964203, "global_step": 253198, "epoch": 2844, "lr": 1.115355944437646e-06}
{"train_loss": 0.08010751008987427, "global_step": 253199, "epoch": 2844, "lr": 1.1152341800839628e-06}
{"train_loss": 0.027675488963723183, "global_step": 253200, "epoch": 2844, "lr": 1.115112422302239e-06}
{"train_loss": 0.020889056846499443, "global_step": 253201, "epoch": 2844, "lr": 1.1149906710924918e-06}
{"train_loss": 0.04245052859187126, "global_step": 253202, "epoch": 2844, "lr": 1.1148689264547373e-06}
{"train_loss": 0.03279522806406021, "global_step": 253203, "epoch": 2844, "lr": 1.1147471883889982e-06}
{"train_loss": 0.04530773311853409, "global_step": 253204, "epoch": 2844, "lr": 1.1146254568952796e-06, "val_loss": 9.652766227722168}
{"train_loss": 0.032265882939100266, "global_step": 253205, "epoch": 2845, "lr": 1.1145037319736095e-06}
{"train_loss": 0.09509315341711044, "global_step": 253206, "epoch": 2845, "lr": 1.1143820136239934e-06}
{"train_loss": 0.05304253101348877, "global_step": 253207, "epoch": 2845, "lr": 1.1142603018464537e-06}
{"train_loss": 0.06768950074911118, "global_step": 253208, "epoch": 2845, "lr": 1.1141385966410067e-06}
{"train_loss": 0.046209950000047684, "global_step": 253209, "epoch": 2845, "lr": 1.1140168980076692e-06}
{"train_loss": 0.022626878693699837, "global_step": 253210, "epoch": 2845, "lr": 1.1138952059464525e-06}
{"train_loss": 0.028135253116488457, "global_step": 253211, "epoch": 2845, "lr": 1.113773520457384e-06}
{"train_loss": 0.0350787490606308, "global_step": 253212, "epoch": 2845, "lr": 1.113651841540464e-06}
{"train_loss": 0.035109661519527435, "global_step": 253213, "epoch": 2845, "lr": 1.1135301691957257e-06}
{"train_loss": 0.0635506883263588, "global_step": 253214, "epoch": 2845, "lr": 1.113408503423169e-06}
{"train_loss": 0.05121378228068352, "global_step": 253215, "epoch": 2845, "lr": 1.1132868442228272e-06}
{"train_loss": 0.024709591642022133, "global_step": 253216, "epoch": 2845, "lr": 1.1131651915947006e-06}
{"train_loss": 0.03984814137220383, "global_step": 253217, "epoch": 2845, "lr": 1.113043545538811e-06}
{"train_loss": 0.05566101521253586, "global_step": 253218, "epoch": 2845, "lr": 1.1129219060551865e-06}
{"train_loss": 0.04231861233711243, "global_step": 253219, "epoch": 2845, "lr": 1.112800273143827e-06}
{"train_loss": 0.04794810712337494, "global_step": 253220, "epoch": 2845, "lr": 1.11267864680476e-06}
{"train_loss": 0.026759009808301926, "global_step": 253221, "epoch": 2845, "lr": 1.112557027037997e-06}
{"train_loss": 0.02885027788579464, "global_step": 253222, "epoch": 2845, "lr": 1.1124354138435544e-06}
{"train_loss": 0.04769650846719742, "global_step": 253223, "epoch": 2845, "lr": 1.1123138072214433e-06}
{"train_loss": 0.043199241161346436, "global_step": 253224, "epoch": 2845, "lr": 1.1121922071716973e-06}
{"train_loss": 0.061456792056560516, "global_step": 253225, "epoch": 2845, "lr": 1.1120706136943105e-06}
{"train_loss": 0.12377076596021652, "global_step": 253226, "epoch": 2845, "lr": 1.1119490267893162e-06}
{"train_loss": 0.0345255509018898, "global_step": 253227, "epoch": 2845, "lr": 1.1118274464567203e-06}
{"train_loss": 0.06754297018051147, "global_step": 253228, "epoch": 2845, "lr": 1.1117058726965501e-06}
{"train_loss": 0.060133546590805054, "global_step": 253229, "epoch": 2845, "lr": 1.111584305508806e-06}
{"train_loss": 0.03039199858903885, "global_step": 253230, "epoch": 2845, "lr": 1.1114627448935211e-06}
{"train_loss": 0.03346875309944153, "global_step": 253231, "epoch": 2845, "lr": 1.1113411908506955e-06}
{"train_loss": 0.04559430852532387, "global_step": 253232, "epoch": 2845, "lr": 1.1112196433803623e-06}
{"train_loss": 0.02439100667834282, "global_step": 253233, "epoch": 2845, "lr": 1.1110981024825328e-06}
{"train_loss": 0.029147617518901825, "global_step": 253234, "epoch": 2845, "lr": 1.1109765681572127e-06}
{"train_loss": 0.04032590612769127, "global_step": 253235, "epoch": 2845, "lr": 1.1108550404044294e-06}
{"train_loss": 0.03943069651722908, "global_step": 253236, "epoch": 2845, "lr": 1.1107335192241886e-06}
{"train_loss": 0.05017125606536865, "global_step": 253237, "epoch": 2845, "lr": 1.1106120046165237e-06}
{"train_loss": 0.049428101629018784, "global_step": 253238, "epoch": 2845, "lr": 1.1104904965814344e-06}
{"train_loss": 0.0519842654466629, "global_step": 253239, "epoch": 2845, "lr": 1.110368995118949e-06}
{"train_loss": 0.04651789739727974, "global_step": 253240, "epoch": 2845, "lr": 1.1102475002290723e-06}
{"train_loss": 0.025290079414844513, "global_step": 253241, "epoch": 2845, "lr": 1.1101260119118329e-06}
{"train_loss": 0.048015888780355453, "global_step": 253242, "epoch": 2845, "lr": 1.1100045301672412e-06}
{"train_loss": 0.03408799692988396, "global_step": 253243, "epoch": 2845, "lr": 1.1098830549953087e-06}
{"train_loss": 0.06543154269456863, "global_step": 253244, "epoch": 2845, "lr": 1.1097615863960575e-06}
{"train_loss": 0.050461411476135254, "global_step": 253245, "epoch": 2845, "lr": 1.1096401243695099e-06}
{"train_loss": 0.057956255972385406, "global_step": 253246, "epoch": 2845, "lr": 1.1095186689156656e-06}
{"train_loss": 0.032648466527462006, "global_step": 253247, "epoch": 2845, "lr": 1.109397220034558e-06}
{"train_loss": 0.022223029285669327, "global_step": 253248, "epoch": 2845, "lr": 1.1092757777261874e-06}
{"train_loss": 0.06653957813978195, "global_step": 253249, "epoch": 2845, "lr": 1.109154341990587e-06}
{"train_loss": 0.11201514303684235, "global_step": 253250, "epoch": 2845, "lr": 1.1090329128277622e-06}
{"train_loss": 0.022048620507121086, "global_step": 253251, "epoch": 2845, "lr": 1.108911490237735e-06}
{"train_loss": 0.082878977060318, "global_step": 253252, "epoch": 2845, "lr": 1.1087900742205115e-06}
{"train_loss": 0.07103310525417328, "global_step": 253253, "epoch": 2845, "lr": 1.108668664776119e-06}
{"train_loss": 0.05179746076464653, "global_step": 253254, "epoch": 2845, "lr": 1.1085472619045745e-06}
{"train_loss": 0.036658186465501785, "global_step": 253255, "epoch": 2845, "lr": 1.1084258656058833e-06}
{"train_loss": 0.046092499047517776, "global_step": 253256, "epoch": 2845, "lr": 1.1083044758800732e-06}
{"train_loss": 0.08477522432804108, "global_step": 253257, "epoch": 2845, "lr": 1.1081830927271497e-06}
{"train_loss": 0.0736229419708252, "global_step": 253258, "epoch": 2845, "lr": 1.1080617161471407e-06}
{"train_loss": 0.06863225251436234, "global_step": 253259, "epoch": 2845, "lr": 1.1079403461400516e-06}
{"train_loss": 0.05434007942676544, "global_step": 253260, "epoch": 2845, "lr": 1.1078189827059105e-06}
{"train_loss": 0.07243335247039795, "global_step": 253261, "epoch": 2845, "lr": 1.1076976258447224e-06}
{"train_loss": 0.044696319848299026, "global_step": 253262, "epoch": 2845, "lr": 1.1075762755565155e-06}
{"train_loss": 0.02009042538702488, "global_step": 253263, "epoch": 2845, "lr": 1.1074549318412897e-06}
{"train_loss": 0.042606621980667114, "global_step": 253264, "epoch": 2845, "lr": 1.1073335946990782e-06}
{"train_loss": 0.03254248946905136, "global_step": 253265, "epoch": 2845, "lr": 1.107212264129881e-06}
{"train_loss": 0.09814244508743286, "global_step": 253266, "epoch": 2845, "lr": 1.1070909401337316e-06}
{"train_loss": 0.03460514917969704, "global_step": 253267, "epoch": 2845, "lr": 1.106969622710635e-06}
{"train_loss": 0.03425535559654236, "global_step": 253268, "epoch": 2845, "lr": 1.1068483118606087e-06}
{"train_loss": 0.038503266870975494, "global_step": 253269, "epoch": 2845, "lr": 1.1067270075836745e-06}
{"train_loss": 0.03817243501543999, "global_step": 253270, "epoch": 2845, "lr": 1.1066057098798432e-06}
{"train_loss": 0.027008721604943275, "global_step": 253271, "epoch": 2845, "lr": 1.1064844187491375e-06}
{"train_loss": 0.06813710182905197, "global_step": 253272, "epoch": 2845, "lr": 1.106363134191557e-06}
{"train_loss": 0.05365249142050743, "global_step": 253273, "epoch": 2845, "lr": 1.1062418562071407e-06}
{"train_loss": 0.06778240203857422, "global_step": 253274, "epoch": 2845, "lr": 1.106120584795889e-06}
{"train_loss": 0.03349926322698593, "global_step": 253275, "epoch": 2845, "lr": 1.105999319957829e-06}
{"train_loss": 0.03956218808889389, "global_step": 253276, "epoch": 2845, "lr": 1.1058780616929608e-06}
{"train_loss": 0.033785685896873474, "global_step": 253277, "epoch": 2845, "lr": 1.1057568100013238e-06}
{"train_loss": 0.03906741738319397, "global_step": 253278, "epoch": 2845, "lr": 1.1056355648829119e-06}
{"train_loss": 0.08019732683897018, "global_step": 253279, "epoch": 2845, "lr": 1.1055143263377588e-06}
{"train_loss": 0.04055451601743698, "global_step": 253280, "epoch": 2845, "lr": 1.1053930943658697e-06}
{"train_loss": 0.08415830880403519, "global_step": 253281, "epoch": 2845, "lr": 1.105271868967267e-06}
{"train_loss": 0.040527671575546265, "global_step": 253282, "epoch": 2845, "lr": 1.105150650141956e-06}
{"train_loss": 0.03684372454881668, "global_step": 253283, "epoch": 2845, "lr": 1.1050294378899707e-06}
{"train_loss": 0.013665027916431427, "global_step": 253284, "epoch": 2845, "lr": 1.1049082322113159e-06}
{"train_loss": 0.034010306000709534, "global_step": 253285, "epoch": 2845, "lr": 1.1047870331060139e-06}
{"train_loss": 0.061452142894268036, "global_step": 253286, "epoch": 2845, "lr": 1.1046658405740706e-06}
{"train_loss": 0.04619145765900612, "global_step": 253287, "epoch": 2845, "lr": 1.1045446546155136e-06}
{"train_loss": 0.04072953388094902, "global_step": 253288, "epoch": 2845, "lr": 1.1044234752303484e-06}
{"train_loss": 0.033822935074567795, "global_step": 253289, "epoch": 2845, "lr": 1.1043023024186027e-06}
{"train_loss": 0.07904038578271866, "global_step": 253290, "epoch": 2845, "lr": 1.104181136180288e-06}
{"train_loss": 0.07877754420042038, "global_step": 253291, "epoch": 2845, "lr": 1.1040599765154147e-06}
{"train_loss": 0.022398294880986214, "global_step": 253292, "epoch": 2845, "lr": 1.1039388234240112e-06}
{"train_loss": 0.04872373466411333, "global_step": 253293, "epoch": 2845, "lr": 1.1038176769060827e-06, "val_loss": 9.530903816223145, "train_action_mse_error": 22.245582580566406}
{"train_loss": 0.02631964161992073, "global_step": 253294, "epoch": 2846, "lr": 1.1036965369616569e-06}
{"train_loss": 0.04914228245615959, "global_step": 253295, "epoch": 2846, "lr": 1.1035754035907343e-06}
{"train_loss": 0.0720948576927185, "global_step": 253296, "epoch": 2846, "lr": 1.1034542767933476e-06}
{"train_loss": 0.037141840904951096, "global_step": 253297, "epoch": 2846, "lr": 1.1033331565694972e-06}
{"train_loss": 0.07367347180843353, "global_step": 253298, "epoch": 2846, "lr": 1.1032120429192162e-06}
{"train_loss": 0.03636723384261131, "global_step": 253299, "epoch": 2846, "lr": 1.1030909358425046e-06}
{"train_loss": 0.04155609756708145, "global_step": 253300, "epoch": 2846, "lr": 1.1029698353393957e-06}
{"train_loss": 0.08390433341264725, "global_step": 253301, "epoch": 2846, "lr": 1.1028487414098898e-06}
{"train_loss": 0.02489742450416088, "global_step": 253302, "epoch": 2846, "lr": 1.1027276540540144e-06}
{"train_loss": 0.05683736130595207, "global_step": 253303, "epoch": 2846, "lr": 1.102606573271775e-06}
{"train_loss": 0.026976831257343292, "global_step": 253304, "epoch": 2846, "lr": 1.1024854990631995e-06}
{"train_loss": 0.05061263591051102, "global_step": 253305, "epoch": 2846, "lr": 1.1023644314282932e-06}
{"train_loss": 0.10988249629735947, "global_step": 253306, "epoch": 2846, "lr": 1.1022433703670842e-06}
{"train_loss": 0.07235021144151688, "global_step": 253307, "epoch": 2846, "lr": 1.102122315879578e-06}
{"train_loss": 0.06570253521203995, "global_step": 253308, "epoch": 2846, "lr": 1.102001267965802e-06}
{"train_loss": 0.0594952329993248, "global_step": 253309, "epoch": 2846, "lr": 1.1018802266257621e-06}
{"train_loss": 0.03843904659152031, "global_step": 253310, "epoch": 2846, "lr": 1.1017591918594749e-06}
{"train_loss": 0.026373660191893578, "global_step": 253311, "epoch": 2846, "lr": 1.1016381636669627e-06}
{"train_loss": 0.00797464419156313, "global_step": 253312, "epoch": 2846, "lr": 1.1015171420482418e-06}
{"train_loss": 0.0649191215634346, "global_step": 253313, "epoch": 2846, "lr": 1.1013961270033235e-06}
{"train_loss": 0.02772035263478756, "global_step": 253314, "epoch": 2846, "lr": 1.1012751185322246e-06}
{"train_loss": 0.060824647545814514, "global_step": 253315, "epoch": 2846, "lr": 1.101154116634967e-06}
{"train_loss": 0.06653201580047607, "global_step": 253316, "epoch": 2846, "lr": 1.1010331213115621e-06}
{"train_loss": 0.0504806712269783, "global_step": 253317, "epoch": 2846, "lr": 1.1009121325620264e-06}
{"train_loss": 0.024854248389601707, "global_step": 253318, "epoch": 2846, "lr": 1.1007911503863766e-06}
{"train_loss": 0.03878422826528549, "global_step": 253319, "epoch": 2846, "lr": 1.100670174784635e-06}
{"train_loss": 0.00939356442540884, "global_step": 253320, "epoch": 2846, "lr": 1.1005492057568012e-06}
{"train_loss": 0.06466268002986908, "global_step": 253321, "epoch": 2846, "lr": 1.1004282433029146e-06}
{"train_loss": 0.04285208135843277, "global_step": 253322, "epoch": 2846, "lr": 1.1003072874229692e-06}
{"train_loss": 0.03471796214580536, "global_step": 253323, "epoch": 2846, "lr": 1.1001863381169986e-06}
{"train_loss": 0.10631637275218964, "global_step": 253324, "epoch": 2846, "lr": 1.1000653953850082e-06}
{"train_loss": 0.05512980744242668, "global_step": 253325, "epoch": 2846, "lr": 1.0999444592270146e-06}
{"train_loss": 0.021489467471837997, "global_step": 253326, "epoch": 2846, "lr": 1.099823529643046e-06}
{"train_loss": 0.05540735274553299, "global_step": 253327, "epoch": 2846, "lr": 1.0997026066331073e-06}
{"train_loss": 0.037816133350133896, "global_step": 253328, "epoch": 2846, "lr": 1.0995816901972156e-06}
{"train_loss": 0.04456339403986931, "global_step": 253329, "epoch": 2846, "lr": 1.0994607803353874e-06}
{"train_loss": 0.0559694766998291, "global_step": 253330, "epoch": 2846, "lr": 1.0993398770476449e-06}
{"train_loss": 0.05460601672530174, "global_step": 253331, "epoch": 2846, "lr": 1.099218980333999e-06}
{"train_loss": 0.05942295864224434, "global_step": 253332, "epoch": 2846, "lr": 1.099098090194467e-06}
{"train_loss": 0.01517088245600462, "global_step": 253333, "epoch": 2846, "lr": 1.098977206629065e-06}
{"train_loss": 0.06991461664438248, "global_step": 253334, "epoch": 2846, "lr": 1.0988563296378095e-06}
{"train_loss": 0.04022682085633278, "global_step": 253335, "epoch": 2846, "lr": 1.0987354592207177e-06}
{"train_loss": 0.018281493335962296, "global_step": 253336, "epoch": 2846, "lr": 1.098614595377806e-06}
{"train_loss": 0.049170009791851044, "global_step": 253337, "epoch": 2846, "lr": 1.0984937381090854e-06}
{"train_loss": 0.027528317645192146, "global_step": 253338, "epoch": 2846, "lr": 1.0983728874145783e-06}
{"train_loss": 0.02450413815677166, "global_step": 253339, "epoch": 2846, "lr": 1.0982520432942956e-06}
{"train_loss": 0.037147946655750275, "global_step": 253340, "epoch": 2846, "lr": 1.098131205748265e-06}
{"train_loss": 0.03591003268957138, "global_step": 253341, "epoch": 2846, "lr": 1.098010374776487e-06}
{"train_loss": 0.06071173772215843, "global_step": 253342, "epoch": 2846, "lr": 1.0978895503789887e-06}
{"train_loss": 0.03852250799536705, "global_step": 253343, "epoch": 2846, "lr": 1.0977687325557817e-06}
{"train_loss": 0.05300750583410263, "global_step": 253344, "epoch": 2846, "lr": 1.097647921306888e-06}
{"train_loss": 0.06205550953745842, "global_step": 253345, "epoch": 2846, "lr": 1.097527116632313e-06}
{"train_loss": 0.03444413095712662, "global_step": 253346, "epoch": 2846, "lr": 1.0974063185320848e-06}
{"train_loss": 0.063136026263237, "global_step": 253347, "epoch": 2846, "lr": 1.0972855270062087e-06}
{"train_loss": 0.10158582031726837, "global_step": 253348, "epoch": 2846, "lr": 1.0971647420547127e-06}
{"train_loss": 0.0436076894402504, "global_step": 253349, "epoch": 2846, "lr": 1.0970439636776075e-06}
{"train_loss": 0.017955804243683815, "global_step": 253350, "epoch": 2846, "lr": 1.0969231918749046e-06}
{"train_loss": 0.05113180726766586, "global_step": 253351, "epoch": 2846, "lr": 1.096802426646626e-06}
{"train_loss": 0.0500238761305809, "global_step": 253352, "epoch": 2846, "lr": 1.0966816679927827e-06}
{"train_loss": 0.029994871467351913, "global_step": 253353, "epoch": 2846, "lr": 1.0965609159133971e-06}
{"train_loss": 0.055965397506952286, "global_step": 253354, "epoch": 2846, "lr": 1.0964401704084804e-06}
{"train_loss": 0.03156193718314171, "global_step": 253355, "epoch": 2846, "lr": 1.09631943147806e-06}
{"train_loss": 0.0403558723628521, "global_step": 253356, "epoch": 2846, "lr": 1.0961986991221307e-06}
{"train_loss": 0.04452672600746155, "global_step": 253357, "epoch": 2846, "lr": 1.096077973340731e-06}
{"train_loss": 0.04934890568256378, "global_step": 253358, "epoch": 2846, "lr": 1.0959572541338614e-06}
{"train_loss": 0.0515717938542366, "global_step": 253359, "epoch": 2846, "lr": 1.0958365415015492e-06}
{"train_loss": 0.04178826883435249, "global_step": 253360, "epoch": 2846, "lr": 1.0957158354438058e-06}
{"train_loss": 0.07322412729263306, "global_step": 253361, "epoch": 2846, "lr": 1.095595135960642e-06}
{"train_loss": 0.05036165192723274, "global_step": 253362, "epoch": 2846, "lr": 1.0954744430520858e-06}
{"train_loss": 0.048981379717588425, "global_step": 253363, "epoch": 2846, "lr": 1.0953537567181426e-06}
{"train_loss": 0.011215581558644772, "global_step": 253364, "epoch": 2846, "lr": 1.0952330769588348e-06}
{"train_loss": 0.04936183616518974, "global_step": 253365, "epoch": 2846, "lr": 1.0951124037741734e-06}
{"train_loss": 0.039276521652936935, "global_step": 253366, "epoch": 2846, "lr": 1.094991737164186e-06}
{"train_loss": 0.022282548248767853, "global_step": 253367, "epoch": 2846, "lr": 1.0948710771288728e-06}
{"train_loss": 0.04450615867972374, "global_step": 253368, "epoch": 2846, "lr": 1.0947504236682615e-06}
{"train_loss": 0.01217673346400261, "global_step": 253369, "epoch": 2846, "lr": 1.0946297767823631e-06}
{"train_loss": 0.05653272196650505, "global_step": 253370, "epoch": 2846, "lr": 1.0945091364712002e-06}
{"train_loss": 0.0378788560628891, "global_step": 253371, "epoch": 2846, "lr": 1.0943885027347779e-06}
{"train_loss": 0.02335541509091854, "global_step": 253372, "epoch": 2846, "lr": 1.094267875573124e-06}
{"train_loss": 0.05814632400870323, "global_step": 253373, "epoch": 2846, "lr": 1.0941472549862497e-06}
{"train_loss": 0.03298904746770859, "global_step": 253374, "epoch": 2846, "lr": 1.0940266409741716e-06}
{"train_loss": 0.025468498468399048, "global_step": 253375, "epoch": 2846, "lr": 1.0939060335369011e-06}
{"train_loss": 0.05001305416226387, "global_step": 253376, "epoch": 2846, "lr": 1.09378543267446e-06}
{"train_loss": 0.07229241728782654, "global_step": 253377, "epoch": 2846, "lr": 1.0936648383868653e-06}
{"train_loss": 0.03617837280035019, "global_step": 253378, "epoch": 2846, "lr": 1.0935442506741334e-06}
{"train_loss": 0.0968429446220398, "global_step": 253379, "epoch": 2846, "lr": 1.0934236695362754e-06}
{"train_loss": 0.01928797736763954, "global_step": 253380, "epoch": 2846, "lr": 1.0933030949733082e-06}
{"train_loss": 0.00965949147939682, "global_step": 253381, "epoch": 2846, "lr": 1.0931825269852536e-06}
{"train_loss": 0.04567483739320482, "global_step": 253382, "epoch": 2846, "lr": 1.0930619655721286e-06, "val_loss": 9.631546974182129}
{"train_loss": 0.025658462196588516, "global_step": 253383, "epoch": 2847, "lr": 1.0929414107339386e-06}
{"train_loss": 0.03512066975235939, "global_step": 253384, "epoch": 2847, "lr": 1.092820862470706e-06}
{"train_loss": 0.06579761207103729, "global_step": 253385, "epoch": 2847, "lr": 1.0927003207824526e-06}
{"train_loss": 0.039489440619945526, "global_step": 253386, "epoch": 2847, "lr": 1.0925797856691844e-06}
{"train_loss": 0.008309364318847656, "global_step": 253387, "epoch": 2847, "lr": 1.0924592571309289e-06}
{"train_loss": 0.02385614812374115, "global_step": 253388, "epoch": 2847, "lr": 1.0923387351676917e-06}
{"train_loss": 0.0819944441318512, "global_step": 253389, "epoch": 2847, "lr": 1.092218219779495e-06}
{"train_loss": 0.018355675041675568, "global_step": 253390, "epoch": 2847, "lr": 1.0920977109663556e-06}
{"train_loss": 0.02609018236398697, "global_step": 253391, "epoch": 2847, "lr": 1.0919772087282842e-06}
{"train_loss": 0.05591810494661331, "global_step": 253392, "epoch": 2847, "lr": 1.0918567130652978e-06}
{"train_loss": 0.02651183307170868, "global_step": 253393, "epoch": 2847, "lr": 1.0917362239774242e-06}
{"train_loss": 0.04662177711725235, "global_step": 253394, "epoch": 2847, "lr": 1.0916157414646578e-06}
{"train_loss": 0.029937390238046646, "global_step": 253395, "epoch": 2847, "lr": 1.0914952655270372e-06}
{"train_loss": 0.04548701271414757, "global_step": 253396, "epoch": 2847, "lr": 1.0913747961645682e-06}
{"train_loss": 0.057821113616228104, "global_step": 253397, "epoch": 2847, "lr": 1.0912543333772618e-06}
{"train_loss": 0.055449143052101135, "global_step": 253398, "epoch": 2847, "lr": 1.0911338771651458e-06}
{"train_loss": 0.04812406376004219, "global_step": 253399, "epoch": 2847, "lr": 1.0910134275282258e-06}
{"train_loss": 0.0371357724070549, "global_step": 253400, "epoch": 2847, "lr": 1.0908929844665294e-06}
{"train_loss": 0.04719483479857445, "global_step": 253401, "epoch": 2847, "lr": 1.0907725479800623e-06}
{"train_loss": 0.03920022025704384, "global_step": 253402, "epoch": 2847, "lr": 1.0906521180688468e-06}
{"train_loss": 0.03965386748313904, "global_step": 253403, "epoch": 2847, "lr": 1.0905316947328935e-06}
{"train_loss": 0.05626315996050835, "global_step": 253404, "epoch": 2847, "lr": 1.0904112779722254e-06}
{"train_loss": 0.05582514777779579, "global_step": 253405, "epoch": 2847, "lr": 1.090290867786853e-06}
{"train_loss": 0.06581347435712814, "global_step": 253406, "epoch": 2847, "lr": 1.0901704641767984e-06}
{"train_loss": 0.0313340499997139, "global_step": 253407, "epoch": 2847, "lr": 1.0900500671420732e-06}
{"train_loss": 0.07734846323728561, "global_step": 253408, "epoch": 2847, "lr": 1.089929676682694e-06}
{"train_loss": 0.033692870289087296, "global_step": 253409, "epoch": 2847, "lr": 1.0898092927986714e-06}
{"train_loss": 0.010390162467956543, "global_step": 253410, "epoch": 2847, "lr": 1.089688915490039e-06}
{"train_loss": 0.05596238002181053, "global_step": 253411, "epoch": 2847, "lr": 1.0895685447567917e-06}
{"train_loss": 0.01708356849849224, "global_step": 253412, "epoch": 2847, "lr": 1.089448180598962e-06}
{"train_loss": 0.05185995250940323, "global_step": 253413, "epoch": 2847, "lr": 1.089327823016556e-06}
{"train_loss": 0.02082722820341587, "global_step": 253414, "epoch": 2847, "lr": 1.0892074720096012e-06}
{"train_loss": 0.024055998772382736, "global_step": 253415, "epoch": 2847, "lr": 1.0890871275780977e-06}
{"train_loss": 0.038570940494537354, "global_step": 253416, "epoch": 2847, "lr": 1.0889667897220733e-06}
{"train_loss": 0.039074745029211044, "global_step": 253417, "epoch": 2847, "lr": 1.0888464584415448e-06}
{"train_loss": 0.07498326152563095, "global_step": 253418, "epoch": 2847, "lr": 1.0887261337365173e-06}
{"train_loss": 0.05020051822066307, "global_step": 253419, "epoch": 2847, "lr": 1.0886058156070188e-06}
{"train_loss": 0.023869166150689125, "global_step": 253420, "epoch": 2847, "lr": 1.0884855040530662e-06}
{"train_loss": 0.03257951885461807, "global_step": 253421, "epoch": 2847, "lr": 1.0883651990746646e-06}
{"train_loss": 0.04100308567285538, "global_step": 253422, "epoch": 2847, "lr": 1.0882449006718365e-06}
{"train_loss": 0.01863279938697815, "global_step": 253423, "epoch": 2847, "lr": 1.0881246088445985e-06}
{"train_loss": 0.027467461302876472, "global_step": 253424, "epoch": 2847, "lr": 1.088004323592967e-06}
{"train_loss": 0.06051728129386902, "global_step": 253425, "epoch": 2847, "lr": 1.0878840449169593e-06}
{"train_loss": 0.02155754342675209, "global_step": 253426, "epoch": 2847, "lr": 1.08776377281658e-06}
{"train_loss": 0.04835401847958565, "global_step": 253427, "epoch": 2847, "lr": 1.0876435072918633e-06}
{"train_loss": 0.09507345408201218, "global_step": 253428, "epoch": 2847, "lr": 1.0875232483428144e-06}
{"train_loss": 0.022067906334996223, "global_step": 253429, "epoch": 2847, "lr": 1.0874029959694499e-06}
{"train_loss": 0.07305237650871277, "global_step": 253430, "epoch": 2847, "lr": 1.0872827501717864e-06}
{"train_loss": 0.05666046589612961, "global_step": 253431, "epoch": 2847, "lr": 1.0871625109498518e-06}
{"train_loss": 0.02485264092683792, "global_step": 253432, "epoch": 2847, "lr": 1.0870422783036405e-06}
{"train_loss": 0.049330249428749084, "global_step": 253433, "epoch": 2847, "lr": 1.0869220522331802e-06}
{"train_loss": 0.031737830489873886, "global_step": 253434, "epoch": 2847, "lr": 1.0868018327384933e-06}
{"train_loss": 0.028517043218016624, "global_step": 253435, "epoch": 2847, "lr": 1.0866816198195906e-06}
{"train_loss": 0.012283743359148502, "global_step": 253436, "epoch": 2847, "lr": 1.0865614134764834e-06}
{"train_loss": 0.048710525035858154, "global_step": 253437, "epoch": 2847, "lr": 1.086441213709194e-06}
{"train_loss": 0.03480679914355278, "global_step": 253438, "epoch": 2847, "lr": 1.0863210205177388e-06}
{"train_loss": 0.06358090043067932, "global_step": 253439, "epoch": 2847, "lr": 1.0862008339021234e-06}
{"train_loss": 0.04523983970284462, "global_step": 253440, "epoch": 2847, "lr": 1.0860806538623813e-06}
{"train_loss": 0.03403414413332939, "global_step": 253441, "epoch": 2847, "lr": 1.0859604803985123e-06}
{"train_loss": 0.02204822190105915, "global_step": 253442, "epoch": 2847, "lr": 1.0858403135105443e-06}
{"train_loss": 0.02836107462644577, "global_step": 253443, "epoch": 2847, "lr": 1.0857201531984829e-06}
{"train_loss": 0.06461948156356812, "global_step": 253444, "epoch": 2847, "lr": 1.0855999994623556e-06}
{"train_loss": 0.02979186922311783, "global_step": 253445, "epoch": 2847, "lr": 1.0854798523021737e-06}
{"train_loss": 0.05068181827664375, "global_step": 253446, "epoch": 2847, "lr": 1.0853597117179537e-06}
{"train_loss": 0.04169468581676483, "global_step": 253447, "epoch": 2847, "lr": 1.0852395777097068e-06}
{"train_loss": 0.0446658618748188, "global_step": 253448, "epoch": 2847, "lr": 1.0851194502774554e-06}
{"train_loss": 0.02744126319885254, "global_step": 253449, "epoch": 2847, "lr": 1.0849993294212102e-06}
{"train_loss": 0.03756159543991089, "global_step": 253450, "epoch": 2847, "lr": 1.0848792151409993e-06}
{"train_loss": 0.0240958072245121, "global_step": 253451, "epoch": 2847, "lr": 1.0847591074368168e-06}
{"train_loss": 0.03978079929947853, "global_step": 253452, "epoch": 2847, "lr": 1.084639006308702e-06}
{"train_loss": 0.022753970697522163, "global_step": 253453, "epoch": 2847, "lr": 1.08451891175666e-06}
{"train_loss": 0.05347602814435959, "global_step": 253454, "epoch": 2847, "lr": 1.0843988237807024e-06}
{"train_loss": 0.06149531528353691, "global_step": 253455, "epoch": 2847, "lr": 1.0842787423808621e-06}
{"train_loss": 0.051034361124038696, "global_step": 253456, "epoch": 2847, "lr": 1.0841586675571336e-06}
{"train_loss": 0.02861175127327442, "global_step": 253457, "epoch": 2847, "lr": 1.0840385993095504e-06}
{"train_loss": 0.052626293152570724, "global_step": 253458, "epoch": 2847, "lr": 1.0839185376381177e-06}
{"train_loss": 0.027277590706944466, "global_step": 253459, "epoch": 2847, "lr": 1.0837984825428637e-06}
{"train_loss": 0.054525889456272125, "global_step": 253460, "epoch": 2847, "lr": 1.0836784340237938e-06}
{"train_loss": 0.11292272806167603, "global_step": 253461, "epoch": 2847, "lr": 1.0835583920809244e-06}
{"train_loss": 0.026355138048529625, "global_step": 253462, "epoch": 2847, "lr": 1.0834383567142725e-06}
{"train_loss": 0.044232260435819626, "global_step": 253463, "epoch": 2847, "lr": 1.08331832792386e-06}
{"train_loss": 0.018617058172822, "global_step": 253464, "epoch": 2847, "lr": 1.0831983057096984e-06}
{"train_loss": 0.04578922316431999, "global_step": 253465, "epoch": 2847, "lr": 1.083078290071804e-06}
{"train_loss": 0.06032155454158783, "global_step": 253466, "epoch": 2847, "lr": 1.0829582810101934e-06}
{"train_loss": 0.07076156139373779, "global_step": 253467, "epoch": 2847, "lr": 1.0828382785248836e-06}
{"train_loss": 0.03580595180392265, "global_step": 253468, "epoch": 2847, "lr": 1.0827182826158856e-06}
{"train_loss": 0.061160363256931305, "global_step": 253469, "epoch": 2847, "lr": 1.0825982932832267e-06}
{"train_loss": 0.044755350798368454, "global_step": 253470, "epoch": 2847, "lr": 1.0824783105269133e-06}
{"train_loss": 0.04241122959411881, "global_step": 253471, "epoch": 2847, "lr": 1.0823583343469613e-06, "val_loss": 9.634178161621094}
{"train_loss": 0.04649878293275833, "global_step": 253472, "epoch": 2848, "lr": 1.0822383647433932e-06}
{"train_loss": 0.08009479939937592, "global_step": 253473, "epoch": 2848, "lr": 1.08211840171622e-06}
{"train_loss": 0.03568919003009796, "global_step": 253474, "epoch": 2848, "lr": 1.081998445265464e-06}
{"train_loss": 0.04126375541090965, "global_step": 253475, "epoch": 2848, "lr": 1.081878495391131e-06}
{"train_loss": 0.07221081852912903, "global_step": 253476, "epoch": 2848, "lr": 1.0817585520932483e-06}
{"train_loss": 0.09266991168260574, "global_step": 253477, "epoch": 2848, "lr": 1.0816386153718216e-06}
{"train_loss": 0.056185249239206314, "global_step": 253478, "epoch": 2848, "lr": 1.0815186852268789e-06}
{"train_loss": 0.050461191684007645, "global_step": 253479, "epoch": 2848, "lr": 1.0813987616584254e-06}
{"train_loss": 0.04178405553102493, "global_step": 253480, "epoch": 2848, "lr": 1.081278844666489e-06}
{"train_loss": 0.02876237966120243, "global_step": 253481, "epoch": 2848, "lr": 1.0811589342510697e-06}
{"train_loss": 0.041080765426158905, "global_step": 253482, "epoch": 2848, "lr": 1.0810390304121953e-06}
{"train_loss": 0.02669502981007099, "global_step": 253483, "epoch": 2848, "lr": 1.080919133149877e-06}
{"train_loss": 0.04706858843564987, "global_step": 253484, "epoch": 2848, "lr": 1.0807992424641367e-06}
{"train_loss": 0.011000534519553185, "global_step": 253485, "epoch": 2848, "lr": 1.08067935835498e-06}
{"train_loss": 0.02967676892876625, "global_step": 253486, "epoch": 2848, "lr": 1.0805594808224407e-06}
{"train_loss": 0.01671656034886837, "global_step": 253487, "epoch": 2848, "lr": 1.0804396098665127e-06}
{"train_loss": 0.012118439190089703, "global_step": 253488, "epoch": 2848, "lr": 1.0803197454872237e-06}
{"train_loss": 0.023728560656309128, "global_step": 253489, "epoch": 2848, "lr": 1.0801998876845966e-06}
{"train_loss": 0.05387916788458824, "global_step": 253490, "epoch": 2848, "lr": 1.080080036458636e-06}
{"train_loss": 0.057740144431591034, "global_step": 253491, "epoch": 2848, "lr": 1.0799601918093704e-06}
{"train_loss": 0.035532932728528976, "global_step": 253492, "epoch": 2848, "lr": 1.0798403537367996e-06}
{"train_loss": 0.06375887989997864, "global_step": 253493, "epoch": 2848, "lr": 1.0797205222409511e-06}
{"train_loss": 0.03082975558936596, "global_step": 253494, "epoch": 2848, "lr": 1.0796006973218364e-06}
{"train_loss": 0.034754205495119095, "global_step": 253495, "epoch": 2848, "lr": 1.079480878979472e-06}
{"train_loss": 0.054811447858810425, "global_step": 253496, "epoch": 2848, "lr": 1.0793610672138742e-06}
{"train_loss": 0.049379173666238785, "global_step": 253497, "epoch": 2848, "lr": 1.0792412620250658e-06}
{"train_loss": 0.02961583621799946, "global_step": 253498, "epoch": 2848, "lr": 1.0791214634130575e-06}
{"train_loss": 0.0827227458357811, "global_step": 253499, "epoch": 2848, "lr": 1.0790016713778605e-06}
{"train_loss": 0.014535823836922646, "global_step": 253500, "epoch": 2848, "lr": 1.078881885919497e-06}
{"train_loss": 0.034722328186035156, "global_step": 253501, "epoch": 2848, "lr": 1.0787621070379782e-06}
{"train_loss": 0.04373474419116974, "global_step": 253502, "epoch": 2848, "lr": 1.0786423347333319e-06}
{"train_loss": 0.0982891321182251, "global_step": 253503, "epoch": 2848, "lr": 1.0785225690055578e-06}
{"train_loss": 0.018425283953547478, "global_step": 253504, "epoch": 2848, "lr": 1.0784028098546838e-06}
{"train_loss": 0.021475261077284813, "global_step": 253505, "epoch": 2848, "lr": 1.0782830572807213e-06}
{"train_loss": 0.05007103458046913, "global_step": 253506, "epoch": 2848, "lr": 1.0781633112836865e-06}
{"train_loss": 0.05733130872249603, "global_step": 253507, "epoch": 2848, "lr": 1.0780435718635961e-06}
{"train_loss": 0.029167136177420616, "global_step": 253508, "epoch": 2848, "lr": 1.0779238390204671e-06}
{"train_loss": 0.052509743720293045, "global_step": 253509, "epoch": 2848, "lr": 1.0778041127543158e-06}
{"train_loss": 0.04003779590129852, "global_step": 253510, "epoch": 2848, "lr": 1.0776843930651592e-06}
{"train_loss": 0.06376905739307404, "global_step": 253511, "epoch": 2848, "lr": 1.0775646799530082e-06}
{"train_loss": 0.09925790876150131, "global_step": 253512, "epoch": 2848, "lr": 1.0774449734178848e-06}
{"train_loss": 0.0389198400080204, "global_step": 253513, "epoch": 2848, "lr": 1.077325273459795e-06}
{"train_loss": 0.050036314874887466, "global_step": 253514, "epoch": 2848, "lr": 1.0772055800787717e-06}
{"train_loss": 0.06726782768964767, "global_step": 253515, "epoch": 2848, "lr": 1.0770858932748207e-06}
{"train_loss": 0.040030669420957565, "global_step": 253516, "epoch": 2848, "lr": 1.0769662130479585e-06}
{"train_loss": 0.05435269698500633, "global_step": 253517, "epoch": 2848, "lr": 1.0768465393981963e-06}
{"train_loss": 0.04902871698141098, "global_step": 253518, "epoch": 2848, "lr": 1.0767268723255676e-06}
{"train_loss": 0.06083989143371582, "global_step": 253519, "epoch": 2848, "lr": 1.0766072118300662e-06}
{"train_loss": 0.042164579033851624, "global_step": 253520, "epoch": 2848, "lr": 1.0764875579117207e-06}
{"train_loss": 0.04550480470061302, "global_step": 253521, "epoch": 2848, "lr": 1.076367910570547e-06}
{"train_loss": 0.03764050826430321, "global_step": 253522, "epoch": 2848, "lr": 1.0762482698065624e-06}
{"train_loss": 0.0495733805000782, "global_step": 253523, "epoch": 2848, "lr": 1.0761286356197719e-06}
{"train_loss": 0.03864375129342079, "global_step": 253524, "epoch": 2848, "lr": 1.0760090080102036e-06}
{"train_loss": 0.022748226299881935, "global_step": 253525, "epoch": 2848, "lr": 1.0758893869778686e-06}
{"train_loss": 0.03607909008860588, "global_step": 253526, "epoch": 2848, "lr": 1.0757697725227834e-06}
{"train_loss": 0.05077686533331871, "global_step": 253527, "epoch": 2848, "lr": 1.07565016464497e-06}
{"train_loss": 0.01918318308889866, "global_step": 253528, "epoch": 2848, "lr": 1.0755305633444346e-06}
{"train_loss": 0.034853920340538025, "global_step": 253529, "epoch": 2848, "lr": 1.0754109686211988e-06}
{"train_loss": 0.03782620653510094, "global_step": 253530, "epoch": 2848, "lr": 1.0752913804752795e-06}
{"train_loss": 0.08442242443561554, "global_step": 253531, "epoch": 2848, "lr": 1.0751717989066879e-06}
{"train_loss": 0.025106025859713554, "global_step": 253532, "epoch": 2848, "lr": 1.075052223915446e-06}
{"train_loss": 0.04832383990287781, "global_step": 253533, "epoch": 2848, "lr": 1.074932655501565e-06}
{"train_loss": 0.04768666997551918, "global_step": 253534, "epoch": 2848, "lr": 1.0748130936650613e-06}
{"train_loss": 0.07005970925092697, "global_step": 253535, "epoch": 2848, "lr": 1.0746935384059576e-06}
{"train_loss": 0.04375781863927841, "global_step": 253536, "epoch": 2848, "lr": 1.0745739897242647e-06}
{"train_loss": 0.047689519822597504, "global_step": 253537, "epoch": 2848, "lr": 1.0744544476199937e-06}
{"train_loss": 0.03619500622153282, "global_step": 253538, "epoch": 2848, "lr": 1.0743349120931722e-06}
{"train_loss": 0.047398753464221954, "global_step": 253539, "epoch": 2848, "lr": 1.074215383143806e-06}
{"train_loss": 0.0327882394194603, "global_step": 253540, "epoch": 2848, "lr": 1.0740958607719177e-06}
{"train_loss": 0.07943297922611237, "global_step": 253541, "epoch": 2848, "lr": 1.073976344977523e-06}
{"train_loss": 0.07780732214450836, "global_step": 253542, "epoch": 2848, "lr": 1.0738568357606282e-06}
{"train_loss": 0.013019525445997715, "global_step": 253543, "epoch": 2848, "lr": 1.0737373331212608e-06}
{"train_loss": 0.04777804762125015, "global_step": 253544, "epoch": 2848, "lr": 1.0736178370594318e-06}
{"train_loss": 0.037843406200408936, "global_step": 253545, "epoch": 2848, "lr": 1.073498347575158e-06}
{"train_loss": 0.044332049787044525, "global_step": 253546, "epoch": 2848, "lr": 1.0733788646684616e-06}
{"train_loss": 0.06203874573111534, "global_step": 253547, "epoch": 2848, "lr": 1.0732593883393483e-06}
{"train_loss": 0.05235810577869415, "global_step": 253548, "epoch": 2848, "lr": 1.07313991858784e-06}
{"train_loss": 0.03814772516489029, "global_step": 253549, "epoch": 2848, "lr": 1.0730204554139534e-06}
{"train_loss": 0.03078332170844078, "global_step": 253550, "epoch": 2848, "lr": 1.0729009988176997e-06}
{"train_loss": 0.09763696789741516, "global_step": 253551, "epoch": 2848, "lr": 1.0727815487991012e-06}
{"train_loss": 0.07083846628665924, "global_step": 253552, "epoch": 2848, "lr": 1.0726621053581687e-06}
{"train_loss": 0.016638049855828285, "global_step": 253553, "epoch": 2848, "lr": 1.0725426684949191e-06}
{"train_loss": 0.02466442435979843, "global_step": 253554, "epoch": 2848, "lr": 1.0724232382093746e-06}
{"train_loss": 0.02384871616959572, "global_step": 253555, "epoch": 2848, "lr": 1.0723038145015407e-06}
{"train_loss": 0.03589344769716263, "global_step": 253556, "epoch": 2848, "lr": 1.0721843973714453e-06}
{"train_loss": 0.075424924492836, "global_step": 253557, "epoch": 2848, "lr": 1.0720649868190935e-06}
{"train_loss": 0.05871148779988289, "global_step": 253558, "epoch": 2848, "lr": 1.071945582844508e-06}
{"train_loss": 0.006235095672309399, "global_step": 253559, "epoch": 2848, "lr": 1.0718261854476995e-06}
{"train_loss": 0.046148731038392926, "global_step": 253560, "epoch": 2848, "lr": 1.0717067946286906e-06, "val_loss": 9.690191268920898}
{"train_loss": 0.04632236808538437, "global_step": 253561, "epoch": 2849, "lr": 1.0715874103874978e-06}
{"train_loss": 0.05428318306803703, "global_step": 253562, "epoch": 2849, "lr": 1.0714680327241267e-06}
{"train_loss": 0.030040496960282326, "global_step": 253563, "epoch": 2849, "lr": 1.0713486616386104e-06}
{"train_loss": 0.01850401982665062, "global_step": 253564, "epoch": 2849, "lr": 1.0712292971309435e-06}
{"train_loss": 0.03724311664700508, "global_step": 253565, "epoch": 2849, "lr": 1.0711099392011648e-06}
{"train_loss": 0.05637921765446663, "global_step": 253566, "epoch": 2849, "lr": 1.0709905878492687e-06}
{"train_loss": 0.05570555105805397, "global_step": 253567, "epoch": 2849, "lr": 1.0708712430752887e-06}
{"train_loss": 0.04602903127670288, "global_step": 253568, "epoch": 2849, "lr": 1.0707519048792304e-06}
{"train_loss": 0.02548874169588089, "global_step": 253569, "epoch": 2849, "lr": 1.0706325732611155e-06}
{"train_loss": 0.05599832534790039, "global_step": 253570, "epoch": 2849, "lr": 1.0705132482209557e-06}
{"train_loss": 0.030718382447957993, "global_step": 253571, "epoch": 2849, "lr": 1.0703939297587673e-06}
{"train_loss": 0.019176576286554337, "global_step": 253572, "epoch": 2849, "lr": 1.0702746178745671e-06}
{"train_loss": 0.029532957822084427, "global_step": 253573, "epoch": 2849, "lr": 1.0701553125683771e-06}
{"train_loss": 0.036808256059885025, "global_step": 253574, "epoch": 2849, "lr": 1.0700360138402088e-06}
{"train_loss": 0.0389137826859951, "global_step": 253575, "epoch": 2849, "lr": 1.069916721690073e-06}
{"train_loss": 0.09237467497587204, "global_step": 253576, "epoch": 2849, "lr": 1.0697974361179973e-06}
{"train_loss": 0.03561633080244064, "global_step": 253577, "epoch": 2849, "lr": 1.0696781571239823e-06}
{"train_loss": 0.03500836715102196, "global_step": 253578, "epoch": 2849, "lr": 1.0695588847080607e-06}
{"train_loss": 0.05185338854789734, "global_step": 253579, "epoch": 2849, "lr": 1.0694396188702328e-06}
{"train_loss": 0.056154508143663406, "global_step": 253580, "epoch": 2849, "lr": 1.0693203596105317e-06}
{"train_loss": 0.04748046398162842, "global_step": 253581, "epoch": 2849, "lr": 1.0692011069289576e-06}
{"train_loss": 0.05170532688498497, "global_step": 253582, "epoch": 2849, "lr": 1.0690818608255326e-06}
{"train_loss": 0.02921939827501774, "global_step": 253583, "epoch": 2849, "lr": 1.0689626213002735e-06}
{"train_loss": 0.05244280397891998, "global_step": 253584, "epoch": 2849, "lr": 1.0688433883532022e-06}
{"train_loss": 0.02985186129808426, "global_step": 253585, "epoch": 2849, "lr": 1.068724161984319e-06}
{"train_loss": 0.0377606637775898, "global_step": 253586, "epoch": 2849, "lr": 1.0686049421936573e-06}
{"train_loss": 0.042863328009843826, "global_step": 253587, "epoch": 2849, "lr": 1.0684857289812166e-06}
{"train_loss": 0.02720365673303604, "global_step": 253588, "epoch": 2849, "lr": 1.0683665223470308e-06}
{"train_loss": 0.04420379549264908, "global_step": 253589, "epoch": 2849, "lr": 1.0682473222910993e-06}
{"train_loss": 0.05973266437649727, "global_step": 253590, "epoch": 2849, "lr": 1.0681281288134504e-06}
{"train_loss": 0.02441137470304966, "global_step": 253591, "epoch": 2849, "lr": 1.0680089419140948e-06}
{"train_loss": 0.052142344415187836, "global_step": 253592, "epoch": 2849, "lr": 1.0678897615930495e-06}
{"train_loss": 0.07418860495090485, "global_step": 253593, "epoch": 2849, "lr": 1.0677705878503253e-06}
{"train_loss": 0.023931438103318214, "global_step": 253594, "epoch": 2849, "lr": 1.06765142068595e-06}
{"train_loss": 0.03116360493004322, "global_step": 253595, "epoch": 2849, "lr": 1.0675322600999294e-06}
{"train_loss": 0.06340409070253372, "global_step": 253596, "epoch": 2849, "lr": 1.0674131060922798e-06}
{"train_loss": 0.08009269833564758, "global_step": 253597, "epoch": 2849, "lr": 1.0672939586630238e-06}
{"train_loss": 0.02907206304371357, "global_step": 253598, "epoch": 2849, "lr": 1.0671748178121665e-06}
{"train_loss": 0.046813592314720154, "global_step": 253599, "epoch": 2849, "lr": 1.0670556835397417e-06}
{"train_loss": 0.08026662468910217, "global_step": 253600, "epoch": 2849, "lr": 1.0669365558457489e-06}
{"train_loss": 0.046917468309402466, "global_step": 253601, "epoch": 2849, "lr": 1.0668174347302106e-06}
{"train_loss": 0.026497963815927505, "global_step": 253602, "epoch": 2849, "lr": 1.0666983201931436e-06}
{"train_loss": 0.041937023401260376, "global_step": 253603, "epoch": 2849, "lr": 1.0665792122345642e-06}
{"train_loss": 0.0830182433128357, "global_step": 253604, "epoch": 2849, "lr": 1.0664601108544837e-06}
{"train_loss": 0.06098815053701401, "global_step": 253605, "epoch": 2849, "lr": 1.066341016052924e-06}
{"train_loss": 0.08092040568590164, "global_step": 253606, "epoch": 2849, "lr": 1.0662219278298912e-06}
{"train_loss": 0.029014827683568, "global_step": 253607, "epoch": 2849, "lr": 1.0661028461854183e-06}
{"train_loss": 0.0704183280467987, "global_step": 253608, "epoch": 2849, "lr": 1.0659837711195053e-06}
{"train_loss": 0.027235060930252075, "global_step": 253609, "epoch": 2849, "lr": 1.0658647026321745e-06}
{"train_loss": 0.037206195294857025, "global_step": 253610, "epoch": 2849, "lr": 1.0657456407234424e-06}
{"train_loss": 0.032813750207424164, "global_step": 253611, "epoch": 2849, "lr": 1.0656265853933257e-06}
{"train_loss": 0.04039251431822777, "global_step": 253612, "epoch": 2849, "lr": 1.065507536641841e-06}
{"train_loss": 0.04208387807011604, "global_step": 253613, "epoch": 2849, "lr": 1.0653884944689996e-06}
{"train_loss": 0.057062648236751556, "global_step": 253614, "epoch": 2849, "lr": 1.0652694588748236e-06}
{"train_loss": 0.020274464040994644, "global_step": 253615, "epoch": 2849, "lr": 1.0651504298593184e-06}
{"train_loss": 0.009046182036399841, "global_step": 253616, "epoch": 2849, "lr": 1.0650314074225177e-06}
{"train_loss": 0.03406219184398651, "global_step": 253617, "epoch": 2849, "lr": 1.0649123915644154e-06}
{"train_loss": 0.02165878936648369, "global_step": 253618, "epoch": 2849, "lr": 1.0647933822850508e-06}
{"train_loss": 0.049027346074581146, "global_step": 253619, "epoch": 2849, "lr": 1.0646743795844183e-06}
{"train_loss": 0.022594338282942772, "global_step": 253620, "epoch": 2849, "lr": 1.0645553834625511e-06}
{"train_loss": 0.06971345841884613, "global_step": 253621, "epoch": 2849, "lr": 1.064436393919449e-06}
{"train_loss": 0.04154614731669426, "global_step": 253622, "epoch": 2849, "lr": 1.0643174109551457e-06}
{"train_loss": 0.04301751032471657, "global_step": 253623, "epoch": 2849, "lr": 1.0641984345696465e-06}
{"train_loss": 0.02710498869419098, "global_step": 253624, "epoch": 2849, "lr": 1.0640794647629683e-06}
{"train_loss": 0.06801710277795792, "global_step": 253625, "epoch": 2849, "lr": 1.0639605015351273e-06}
{"train_loss": 0.059628814458847046, "global_step": 253626, "epoch": 2849, "lr": 1.0638415448861405e-06}
{"train_loss": 0.03468829393386841, "global_step": 253627, "epoch": 2849, "lr": 1.0637225948160246e-06}
{"train_loss": 0.02252417802810669, "global_step": 253628, "epoch": 2849, "lr": 1.063603651324796e-06}
{"train_loss": 0.037173446267843246, "global_step": 253629, "epoch": 2849, "lr": 1.063484714412466e-06}
{"train_loss": 0.017373284325003624, "global_step": 253630, "epoch": 2849, "lr": 1.0633657840790623e-06}
{"train_loss": 0.04330304265022278, "global_step": 253631, "epoch": 2849, "lr": 1.0632468603245794e-06}
{"train_loss": 0.06130427122116089, "global_step": 253632, "epoch": 2849, "lr": 1.0631279431490558e-06}
{"train_loss": 0.09075512737035751, "global_step": 253633, "epoch": 2849, "lr": 1.0630090325524978e-06}
{"train_loss": 0.03519270941615105, "global_step": 253634, "epoch": 2849, "lr": 1.0628901285349157e-06}
{"train_loss": 0.044750481843948364, "global_step": 253635, "epoch": 2849, "lr": 1.0627712310963379e-06}
{"train_loss": 0.03038075380027294, "global_step": 253636, "epoch": 2849, "lr": 1.0626523402367693e-06}
{"train_loss": 0.04893305525183678, "global_step": 253637, "epoch": 2849, "lr": 1.0625334559562383e-06}
{"train_loss": 0.064065121114254, "global_step": 253638, "epoch": 2849, "lr": 1.0624145782547444e-06}
{"train_loss": 0.026073647662997246, "global_step": 253639, "epoch": 2849, "lr": 1.0622957071323213e-06}
{"train_loss": 0.06679847836494446, "global_step": 253640, "epoch": 2849, "lr": 1.062176842588969e-06}
{"train_loss": 0.031627122312784195, "global_step": 253641, "epoch": 2849, "lr": 1.0620579846247148e-06}
{"train_loss": 0.02625991590321064, "global_step": 253642, "epoch": 2849, "lr": 1.0619391332395646e-06}
{"train_loss": 0.028715120628476143, "global_step": 253643, "epoch": 2849, "lr": 1.061820288433546e-06}
{"train_loss": 0.06578008085489273, "global_step": 253644, "epoch": 2849, "lr": 1.0617014502066647e-06}
{"train_loss": 0.10002289712429047, "global_step": 253645, "epoch": 2849, "lr": 1.0615826185589428e-06}
{"train_loss": 0.05752846226096153, "global_step": 253646, "epoch": 2849, "lr": 1.0614637934903971e-06}
{"train_loss": 0.02126753143966198, "global_step": 253647, "epoch": 2849, "lr": 1.0613449750010386e-06}
{"train_loss": 0.010566729120910168, "global_step": 253648, "epoch": 2849, "lr": 1.0612261630908837e-06}
{"train_loss": 0.043985716046325944, "global_step": 253649, "epoch": 2849, "lr": 1.061107357759955e-06, "val_loss": 9.591567993164062}
{"train_loss": 0.036826420575380325, "global_step": 253650, "epoch": 2850, "lr": 1.0609885590082636e-06}
{"train_loss": 0.035419534891843796, "global_step": 253651, "epoch": 2850, "lr": 1.0608697668358257e-06}
{"train_loss": 0.027721017599105835, "global_step": 253652, "epoch": 2850, "lr": 1.0607509812426587e-06}
{"train_loss": 0.025998089462518692, "global_step": 253653, "epoch": 2850, "lr": 1.060632202228773e-06}
{"train_loss": 0.061377543956041336, "global_step": 253654, "epoch": 2850, "lr": 1.060513429794191e-06}
{"train_loss": 0.015059839934110641, "global_step": 253655, "epoch": 2850, "lr": 1.060394663938924e-06}
{"train_loss": 0.024907344952225685, "global_step": 253656, "epoch": 2850, "lr": 1.060275904662994e-06}
{"train_loss": 0.04961470514535904, "global_step": 253657, "epoch": 2850, "lr": 1.0601571519664122e-06}
{"train_loss": 0.046954818069934845, "global_step": 253658, "epoch": 2850, "lr": 1.0600384058491953e-06}
{"train_loss": 0.047232579439878464, "global_step": 253659, "epoch": 2850, "lr": 1.0599196663113598e-06}
{"train_loss": 0.028716281056404114, "global_step": 253660, "epoch": 2850, "lr": 1.0598009333529223e-06}
{"train_loss": 0.08563202619552612, "global_step": 253661, "epoch": 2850, "lr": 1.0596822069738944e-06}
{"train_loss": 0.022201379761099815, "global_step": 253662, "epoch": 2850, "lr": 1.0595634871743033e-06}
{"train_loss": 0.051598288118839264, "global_step": 253663, "epoch": 2850, "lr": 1.059444773954149e-06}
{"train_loss": 0.05813920497894287, "global_step": 253664, "epoch": 2850, "lr": 1.0593260673134598e-06}
{"train_loss": 0.03961174190044403, "global_step": 253665, "epoch": 2850, "lr": 1.0592073672522463e-06}
{"train_loss": 0.023067135363817215, "global_step": 253666, "epoch": 2850, "lr": 1.0590886737705309e-06}
{"train_loss": 0.018544021993875504, "global_step": 253667, "epoch": 2850, "lr": 1.058969986868319e-06}
{"train_loss": 0.05518776923418045, "global_step": 253668, "epoch": 2850, "lr": 1.058851306545633e-06}
{"train_loss": 0.05618222802877426, "global_step": 253669, "epoch": 2850, "lr": 1.0587326328024893e-06}
{"train_loss": 0.06302130222320557, "global_step": 253670, "epoch": 2850, "lr": 1.0586139656388994e-06}
{"train_loss": 0.027523359283804893, "global_step": 253671, "epoch": 2850, "lr": 1.0584953050548851e-06}
{"train_loss": 0.024169981479644775, "global_step": 253672, "epoch": 2850, "lr": 1.0583766510504578e-06}
{"train_loss": 0.06681613624095917, "global_step": 253673, "epoch": 2850, "lr": 1.0582580036256396e-06}
{"train_loss": 0.05630381777882576, "global_step": 253674, "epoch": 2850, "lr": 1.058139362780436e-06}
{"train_loss": 0.026134980842471123, "global_step": 253675, "epoch": 2850, "lr": 1.0580207285148746e-06}
{"train_loss": 0.061781249940395355, "global_step": 253676, "epoch": 2850, "lr": 1.0579021008289614e-06}
{"train_loss": 0.02862612158060074, "global_step": 253677, "epoch": 2850, "lr": 1.0577834797227238e-06}
{"train_loss": 0.026867272332310677, "global_step": 253678, "epoch": 2850, "lr": 1.0576648651961618e-06}
{"train_loss": 0.0348639041185379, "global_step": 253679, "epoch": 2850, "lr": 1.057546257249309e-06}
{"train_loss": 0.04448753967881203, "global_step": 253680, "epoch": 2850, "lr": 1.057427655882165e-06}
{"train_loss": 0.03665870055556297, "global_step": 253681, "epoch": 2850, "lr": 1.0573090610947578e-06}
{"train_loss": 0.034067705273628235, "global_step": 253682, "epoch": 2850, "lr": 1.057190472887093e-06}
{"train_loss": 0.013923410326242447, "global_step": 253683, "epoch": 2850, "lr": 1.0570718912591982e-06}
{"train_loss": 0.030490966513752937, "global_step": 253684, "epoch": 2850, "lr": 1.0569533162110845e-06}
{"train_loss": 0.06172484531998634, "global_step": 253685, "epoch": 2850, "lr": 1.0568347477427631e-06}
{"train_loss": 0.0698077529668808, "global_step": 253686, "epoch": 2850, "lr": 1.0567161858542562e-06}
{"train_loss": 0.025078263133764267, "global_step": 253687, "epoch": 2850, "lr": 1.0565976305455748e-06}
{"train_loss": 0.04591762274503708, "global_step": 253688, "epoch": 2850, "lr": 1.0564790818167358e-06}
{"train_loss": 0.027708526700735092, "global_step": 253689, "epoch": 2850, "lr": 1.0563605396677612e-06}
{"train_loss": 0.01963561214506626, "global_step": 253690, "epoch": 2850, "lr": 1.056242004098662e-06}
{"train_loss": 0.05906635895371437, "global_step": 253691, "epoch": 2850, "lr": 1.0561234751094551e-06}
{"train_loss": 0.03556007891893387, "global_step": 253692, "epoch": 2850, "lr": 1.056004952700157e-06}
{"train_loss": 0.07634513825178146, "global_step": 253693, "epoch": 2850, "lr": 1.0558864368707732e-06}
{"train_loss": 0.02233835682272911, "global_step": 253694, "epoch": 2850, "lr": 1.0557679276213373e-06}
{"train_loss": 0.05823884904384613, "global_step": 253695, "epoch": 2850, "lr": 1.0556494249518546e-06}
{"train_loss": 0.06481599062681198, "global_step": 253696, "epoch": 2850, "lr": 1.0555309288623472e-06}
{"train_loss": 0.05219746753573418, "global_step": 253697, "epoch": 2850, "lr": 1.0554124393528208e-06}
{"train_loss": 0.03322334960103035, "global_step": 253698, "epoch": 2850, "lr": 1.0552939564233032e-06}
{"train_loss": 0.028088344261050224, "global_step": 253699, "epoch": 2850, "lr": 1.0551754800738001e-06}
{"train_loss": 0.06042492762207985, "global_step": 253700, "epoch": 2850, "lr": 1.0550570103043334e-06}
{"train_loss": 0.03451994061470032, "global_step": 253701, "epoch": 2850, "lr": 1.0549385471149142e-06}
{"train_loss": 0.048927176743745804, "global_step": 253702, "epoch": 2850, "lr": 1.0548200905055705e-06}
{"train_loss": 0.0321829654276371, "global_step": 253703, "epoch": 2850, "lr": 1.0547016404763021e-06}
{"train_loss": 0.0204622820019722, "global_step": 253704, "epoch": 2850, "lr": 1.054583197027137e-06}
{"train_loss": 0.035913143306970596, "global_step": 253705, "epoch": 2850, "lr": 1.0544647601580859e-06}
{"train_loss": 0.024477457627654076, "global_step": 253706, "epoch": 2850, "lr": 1.0543463298691602e-06}
{"train_loss": 0.046752166002988815, "global_step": 253707, "epoch": 2850, "lr": 1.0542279061603876e-06}
{"train_loss": 0.04816491901874542, "global_step": 253708, "epoch": 2850, "lr": 1.0541094890317737e-06}
{"train_loss": 0.0508531779050827, "global_step": 253709, "epoch": 2850, "lr": 1.0539910784833407e-06}
{"train_loss": 0.023749206215143204, "global_step": 253710, "epoch": 2850, "lr": 1.0538726745150995e-06}
{"train_loss": 0.04265046492218971, "global_step": 253711, "epoch": 2850, "lr": 1.0537542771270725e-06}
{"train_loss": 0.04216912388801575, "global_step": 253712, "epoch": 2850, "lr": 1.0536358863192653e-06}
{"train_loss": 0.0443219356238842, "global_step": 253713, "epoch": 2850, "lr": 1.0535175020917054e-06}
{"train_loss": 0.04350375011563301, "global_step": 253714, "epoch": 2850, "lr": 1.053399124444404e-06}
{"train_loss": 0.08309893310070038, "global_step": 253715, "epoch": 2850, "lr": 1.053280753377378e-06}
{"train_loss": 0.057029932737350464, "global_step": 253716, "epoch": 2850, "lr": 1.0531623888906328e-06}
{"train_loss": 0.07027630507946014, "global_step": 253717, "epoch": 2850, "lr": 1.0530440309842016e-06}
{"train_loss": 0.06334516406059265, "global_step": 253718, "epoch": 2850, "lr": 1.05292567965809e-06}
{"train_loss": 0.05247974395751953, "global_step": 253719, "epoch": 2850, "lr": 1.0528073349123147e-06}
{"train_loss": 0.026861263439059258, "global_step": 253720, "epoch": 2850, "lr": 1.0526889967468923e-06}
{"train_loss": 0.016156328842043877, "global_step": 253721, "epoch": 2850, "lr": 1.052570665161845e-06}
{"train_loss": 0.04414217174053192, "global_step": 253722, "epoch": 2850, "lr": 1.0524523401571784e-06}
{"train_loss": 0.04155619442462921, "global_step": 253723, "epoch": 2850, "lr": 1.0523340217329147e-06}
{"train_loss": 0.0900077149271965, "global_step": 253724, "epoch": 2850, "lr": 1.052215709889065e-06}
{"train_loss": 0.05053292214870453, "global_step": 253725, "epoch": 2850, "lr": 1.0520974046256515e-06}
{"train_loss": 0.04324865713715553, "global_step": 253726, "epoch": 2850, "lr": 1.0519791059426799e-06}
{"train_loss": 0.022903727367520332, "global_step": 253727, "epoch": 2850, "lr": 1.051860813840183e-06}
{"train_loss": 0.02970491535961628, "global_step": 253728, "epoch": 2850, "lr": 1.0517425283181614e-06}
{"train_loss": 0.05430613085627556, "global_step": 253729, "epoch": 2850, "lr": 1.051624249376637e-06}
{"train_loss": 0.01906641013920307, "global_step": 253730, "epoch": 2850, "lr": 1.0515059770156322e-06}
{"train_loss": 0.05519895628094673, "global_step": 253731, "epoch": 2850, "lr": 1.0513877112351467e-06}
{"train_loss": 0.029746513813734055, "global_step": 253732, "epoch": 2850, "lr": 1.0512694520352083e-06}
{"train_loss": 0.0228304211050272, "global_step": 253733, "epoch": 2850, "lr": 1.0511511994158285e-06}
{"train_loss": 0.06227849796414375, "global_step": 253734, "epoch": 2850, "lr": 1.0510329533770292e-06}
{"train_loss": 0.06541960686445236, "global_step": 253735, "epoch": 2850, "lr": 1.050914713918816e-06}
{"train_loss": 0.04015403613448143, "global_step": 253736, "epoch": 2850, "lr": 1.0507964810412162e-06}
{"train_loss": 0.025583554059267044, "global_step": 253737, "epoch": 2850, "lr": 1.0506782547442362e-06}
{"train_loss": 0.04218370404638601, "global_step": 253738, "epoch": 2850, "lr": 1.0505600350278976e-06, "train/sim_max_reward_0": 0.6047543012549216, "train/sim_max_reward_1": 0.959141588067592, "train/sim_max_reward_2": 0.4787764640791213, "train/sim_max_reward_3": 1.0, "train/sim_max_reward_4": 0.9651268216647143, "train/sim_max_reward_5": 0.5621616322035807, "test/sim_max_reward_4500000": 0.4733774102906665, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9699188754714678, "test/sim_max_reward_4500003": 0.1498258514530413, "test/sim_max_reward_4500004": 0.9659958193409286, "test/sim_max_reward_4500005": 0.40374251767777897, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9018101746816511, "test/sim_max_reward_4500008": 0.47572920174190453, "test/sim_max_reward_4500009": 0.05013346105517987, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 0.9537275399671202, "test/sim_max_reward_4500012": 1.0, "test/sim_max_reward_4500013": 0.4554880717688603, "test/sim_max_reward_4500014": 0.4175996549375326, "test/sim_max_reward_4500015": 0.9774749263018379, "test/sim_max_reward_4500016": 0.6979499638245563, "test/sim_max_reward_4500017": 0.08132895622439862, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9895452737081195, "test/sim_max_reward_4500022": 0.960890546046734, "test/sim_max_reward_4500023": 0.5806358423919554, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.8595085264446927, "test/sim_max_reward_4500026": 0.9643632267303657, "test/sim_max_reward_4500027": 0.9934127573291814, "test/sim_max_reward_4500028": 0.9857091657113822, "test/sim_max_reward_4500029": 0.6694511318668459, "test/sim_max_reward_4500030": 0.9784662526399763, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.5566910116739859, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.4122335745301168, "test/sim_max_reward_4500035": 0.8732975291882602, "test/sim_max_reward_4500036": 0.05900372777320473, "test/sim_max_reward_4500037": 0.9697980369829908, "test/sim_max_reward_4500038": 0.16217193982329736, "test/sim_max_reward_4500039": 0.9617564166643786, "test/sim_max_reward_4500040": 0.9457131078803297, "test/sim_max_reward_4500041": 0.9902371778813209, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 1.0, "test/sim_max_reward_4500044": 0.007286013661705386, "test/sim_max_reward_4500045": 1.0, "test/sim_max_reward_4500046": 0.6418739594613376, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7616601345449884, "test/mean_score": 0.6507229528625422, "val_loss": 9.564681053161621, "train_action_mse_error": 20.59957504272461}
{"train_loss": 0.09346669912338257, "global_step": 253739, "epoch": 2851, "lr": 1.050441821892212e-06}
{"train_loss": 0.03560417518019676, "global_step": 253740, "epoch": 2851, "lr": 1.050323615337201e-06}
{"train_loss": 0.030416296795010567, "global_step": 253741, "epoch": 2851, "lr": 1.0502054153628759e-06}
{"train_loss": 0.03273783251643181, "global_step": 253742, "epoch": 2851, "lr": 1.0500872219692537e-06}
{"train_loss": 0.03446738421916962, "global_step": 253743, "epoch": 2851, "lr": 1.0499690351563563e-06}
{"train_loss": 0.037315040826797485, "global_step": 253744, "epoch": 2851, "lr": 1.0498508549241836e-06}
{"train_loss": 0.04560089111328125, "global_step": 253745, "epoch": 2851, "lr": 1.0497326812727693e-06}
{"train_loss": 0.053271885961294174, "global_step": 253746, "epoch": 2851, "lr": 1.0496145142021186e-06}
{"train_loss": 0.053032681345939636, "global_step": 253747, "epoch": 2851, "lr": 1.0494963537122538e-06}
{"train_loss": 0.01836841180920601, "global_step": 253748, "epoch": 2851, "lr": 1.049378199803186e-06}
{"train_loss": 0.022855786606669426, "global_step": 253749, "epoch": 2851, "lr": 1.0492600524749375e-06}
{"train_loss": 0.03494228422641754, "global_step": 253750, "epoch": 2851, "lr": 1.049141911727508e-06}
{"train_loss": 0.05229447782039642, "global_step": 253751, "epoch": 2851, "lr": 1.0490237775609368e-06}
{"train_loss": 0.019832579419016838, "global_step": 253752, "epoch": 2851, "lr": 1.0489056499752181e-06}
{"train_loss": 0.0717429593205452, "global_step": 253753, "epoch": 2851, "lr": 1.0487875289703852e-06}
{"train_loss": 0.02832026407122612, "global_step": 253754, "epoch": 2851, "lr": 1.0486694145464381e-06}
{"train_loss": 0.05287611857056618, "global_step": 253755, "epoch": 2851, "lr": 1.0485513067034102e-06}
{"train_loss": 0.05760929360985756, "global_step": 253756, "epoch": 2851, "lr": 1.048433205441296e-06}
{"train_loss": 0.04718989133834839, "global_step": 253757, "epoch": 2851, "lr": 1.048315110760134e-06}
{"train_loss": 0.028638806194067, "global_step": 253758, "epoch": 2851, "lr": 1.0481970226599246e-06}
{"train_loss": 0.041011542081832886, "global_step": 253759, "epoch": 2851, "lr": 1.04807894114069e-06}
{"train_loss": 0.04701092094182968, "global_step": 253760, "epoch": 2851, "lr": 1.047960866202441e-06}
{"train_loss": 0.0383625365793705, "global_step": 253761, "epoch": 2851, "lr": 1.0478427978451998e-06}
{"train_loss": 0.08668673783540726, "global_step": 253762, "epoch": 2851, "lr": 1.0477247360689835e-06}
{"train_loss": 0.057113874703645706, "global_step": 253763, "epoch": 2851, "lr": 1.0476066808738028e-06}
{"train_loss": 0.029345151036977768, "global_step": 253764, "epoch": 2851, "lr": 1.047488632259669e-06}
{"train_loss": 0.013432856649160385, "global_step": 253765, "epoch": 2851, "lr": 1.0473705902266096e-06}
{"train_loss": 0.08862295746803284, "global_step": 253766, "epoch": 2851, "lr": 1.0472525547746304e-06}
{"train_loss": 0.06159590184688568, "global_step": 253767, "epoch": 2851, "lr": 1.0471345259037535e-06}
{"train_loss": 0.052673567086458206, "global_step": 253768, "epoch": 2851, "lr": 1.0470165036139957e-06}
{"train_loss": 0.02614288032054901, "global_step": 253769, "epoch": 2851, "lr": 1.0468984879053622e-06}
{"train_loss": 0.04975348338484764, "global_step": 253770, "epoch": 2851, "lr": 1.0467804787778812e-06}
{"train_loss": 0.022386232390999794, "global_step": 253771, "epoch": 2851, "lr": 1.0466624762315635e-06}
{"train_loss": 0.04918403550982475, "global_step": 253772, "epoch": 2851, "lr": 1.046544480266426e-06}
{"train_loss": 0.03401042893528938, "global_step": 253773, "epoch": 2851, "lr": 1.0464264908824794e-06}
{"train_loss": 0.09349910914897919, "global_step": 253774, "epoch": 2851, "lr": 1.046308508079752e-06}
{"train_loss": 0.03915984183549881, "global_step": 253775, "epoch": 2851, "lr": 1.0461905318582433e-06}
{"train_loss": 0.08523827791213989, "global_step": 253776, "epoch": 2851, "lr": 1.0460725622179867e-06}
{"train_loss": 0.05279475077986717, "global_step": 253777, "epoch": 2851, "lr": 1.0459545991589826e-06}
{"train_loss": 0.0132167162373662, "global_step": 253778, "epoch": 2851, "lr": 1.0458366426812527e-06}
{"train_loss": 0.04846300929784775, "global_step": 253779, "epoch": 2851, "lr": 1.0457186927848195e-06}
{"train_loss": 0.017228977754712105, "global_step": 253780, "epoch": 2851, "lr": 1.0456007494696828e-06}
{"train_loss": 0.028026970103383064, "global_step": 253781, "epoch": 2851, "lr": 1.0454828127358763e-06}
{"train_loss": 0.014273423701524734, "global_step": 253782, "epoch": 2851, "lr": 1.045364882583405e-06}
{"train_loss": 0.07043391466140747, "global_step": 253783, "epoch": 2851, "lr": 1.0452469590122915e-06}
{"train_loss": 0.03239668533205986, "global_step": 253784, "epoch": 2851, "lr": 1.0451290420225413e-06}
{"train_loss": 0.04714398831129074, "global_step": 253785, "epoch": 2851, "lr": 1.0450111316141819e-06}
{"train_loss": 0.030575891956686974, "global_step": 253786, "epoch": 2851, "lr": 1.0448932277872248e-06}
{"train_loss": 0.01948186755180359, "global_step": 253787, "epoch": 2851, "lr": 1.0447753305416808e-06}
{"train_loss": 0.031096089631319046, "global_step": 253788, "epoch": 2851, "lr": 1.044657439877572e-06}
{"train_loss": 0.03150145709514618, "global_step": 253789, "epoch": 2851, "lr": 1.0445395557949156e-06}
{"train_loss": 0.047998782247304916, "global_step": 253790, "epoch": 2851, "lr": 1.0444216782937167e-06}
{"train_loss": 0.08277135342359543, "global_step": 253791, "epoch": 2851, "lr": 1.0443038073740085e-06}
{"train_loss": 0.03869784623384476, "global_step": 253792, "epoch": 2851, "lr": 1.0441859430357914e-06}
{"train_loss": 0.047790661454200745, "global_step": 253793, "epoch": 2851, "lr": 1.0440680852790874e-06}
{"train_loss": 0.006105762906372547, "global_step": 253794, "epoch": 2851, "lr": 1.0439502341039076e-06}
{"train_loss": 0.04810104891657829, "global_step": 253795, "epoch": 2851, "lr": 1.0438323895102797e-06}
{"train_loss": 0.06313537806272507, "global_step": 253796, "epoch": 2851, "lr": 1.0437145514982038e-06}
{"train_loss": 0.018748972564935684, "global_step": 253797, "epoch": 2851, "lr": 1.0435967200677078e-06}
{"train_loss": 0.04782361909747124, "global_step": 253798, "epoch": 2851, "lr": 1.043478895218808e-06}
{"train_loss": 0.047231245785951614, "global_step": 253799, "epoch": 2851, "lr": 1.0433610769515046e-06}
{"train_loss": 0.0696408823132515, "global_step": 253800, "epoch": 2851, "lr": 1.0432432652658364e-06}
{"train_loss": 0.020415421575307846, "global_step": 253801, "epoch": 2851, "lr": 1.043125460161798e-06}
{"train_loss": 0.14034651219844818, "global_step": 253802, "epoch": 2851, "lr": 1.0430076616394224e-06}
{"train_loss": 0.06555400043725967, "global_step": 253803, "epoch": 2851, "lr": 1.0428898696987156e-06}
{"train_loss": 0.02926892414689064, "global_step": 253804, "epoch": 2851, "lr": 1.0427720843396938e-06}
{"train_loss": 0.035065658390522, "global_step": 253805, "epoch": 2851, "lr": 1.0426543055623794e-06}
{"train_loss": 0.0505654402077198, "global_step": 253806, "epoch": 2851, "lr": 1.042536533366778e-06}
{"train_loss": 0.05754005163908005, "global_step": 253807, "epoch": 2851, "lr": 1.0424187677529119e-06}
{"train_loss": 0.11524592339992523, "global_step": 253808, "epoch": 2851, "lr": 1.0423010087207973e-06}
{"train_loss": 0.051411859691143036, "global_step": 253809, "epoch": 2851, "lr": 1.042183256270446e-06}
{"train_loss": 0.038144778460264206, "global_step": 253810, "epoch": 2851, "lr": 1.0420655104018794e-06}
{"train_loss": 0.05346869304776192, "global_step": 253811, "epoch": 2851, "lr": 1.0419477711151093e-06}
{"train_loss": 0.016554981470108032, "global_step": 253812, "epoch": 2851, "lr": 1.041830038410152e-06}
{"train_loss": 0.012726008892059326, "global_step": 253813, "epoch": 2851, "lr": 1.0417123122870242e-06}
{"train_loss": 0.03566668555140495, "global_step": 253814, "epoch": 2851, "lr": 1.041594592745737e-06}
{"train_loss": 0.055321935564279556, "global_step": 253815, "epoch": 2851, "lr": 1.0414768797863185e-06}
{"train_loss": 0.06072332337498665, "global_step": 253816, "epoch": 2851, "lr": 1.0413591734087736e-06}
{"train_loss": 0.06492158025503159, "global_step": 253817, "epoch": 2851, "lr": 1.0412414736131194e-06}
{"train_loss": 0.024254441261291504, "global_step": 253818, "epoch": 2851, "lr": 1.0411237803993724e-06}
{"train_loss": 0.06232994422316551, "global_step": 253819, "epoch": 2851, "lr": 1.0410060937675548e-06}
{"train_loss": 0.028705690056085587, "global_step": 253820, "epoch": 2851, "lr": 1.0408884137176722e-06}
{"train_loss": 0.033213865011930466, "global_step": 253821, "epoch": 2851, "lr": 1.0407707402497523e-06}
{"train_loss": 0.033533841371536255, "global_step": 253822, "epoch": 2851, "lr": 1.0406530733637953e-06}
{"train_loss": 0.048768069595098495, "global_step": 253823, "epoch": 2851, "lr": 1.0405354130598345e-06}
{"train_loss": 0.04096394404768944, "global_step": 253824, "epoch": 2851, "lr": 1.0404177593378695e-06}
{"train_loss": 0.049770720303058624, "global_step": 253825, "epoch": 2851, "lr": 1.0403001121979283e-06}
{"train_loss": 0.027270259335637093, "global_step": 253826, "epoch": 2851, "lr": 1.0401824716400167e-06}
{"train_loss": 0.0456577551791842, "global_step": 253827, "epoch": 2851, "lr": 1.040064837664162e-06, "val_loss": 9.682394027709961}
{"train_loss": 0.060135211795568466, "global_step": 253828, "epoch": 2852, "lr": 1.03994721027037e-06}
{"train_loss": 0.046534463763237, "global_step": 253829, "epoch": 2852, "lr": 1.0398295894586629e-06}
{"train_loss": 0.04883761331439018, "global_step": 253830, "epoch": 2852, "lr": 1.0397119752290518e-06}
{"train_loss": 0.031011290848255157, "global_step": 253831, "epoch": 2852, "lr": 1.039594367581559e-06}
{"train_loss": 0.06874051690101624, "global_step": 253832, "epoch": 2852, "lr": 1.0394767665161897e-06}
{"train_loss": 0.0492163710296154, "global_step": 253833, "epoch": 2852, "lr": 1.0393591720329666e-06}
{"train_loss": 0.05033918470144272, "global_step": 253834, "epoch": 2852, "lr": 1.0392415841319114e-06}
{"train_loss": 0.07419305294752121, "global_step": 253835, "epoch": 2852, "lr": 1.0391240028130245e-06}
{"train_loss": 0.04086527228355408, "global_step": 253836, "epoch": 2852, "lr": 1.039006428076339e-06}
{"train_loss": 0.06719184666872025, "global_step": 253837, "epoch": 2852, "lr": 1.038888859921855e-06}
{"train_loss": 0.03251848742365837, "global_step": 253838, "epoch": 2852, "lr": 1.0387712983496056e-06}
{"train_loss": 0.04438190534710884, "global_step": 253839, "epoch": 2852, "lr": 1.0386537433595855e-06}
{"train_loss": 0.03613142669200897, "global_step": 253840, "epoch": 2852, "lr": 1.038536194951828e-06}
{"train_loss": 0.05978701636195183, "global_step": 253841, "epoch": 2852, "lr": 1.038418653126344e-06}
{"train_loss": 0.0239071324467659, "global_step": 253842, "epoch": 2852, "lr": 1.0383011178831447e-06}
{"train_loss": 0.03242797777056694, "global_step": 253843, "epoch": 2852, "lr": 1.0381835892222525e-06}
{"train_loss": 0.04451214522123337, "global_step": 253844, "epoch": 2852, "lr": 1.0380660671436726e-06}
{"train_loss": 0.07677002251148224, "global_step": 253845, "epoch": 2852, "lr": 1.037948551647433e-06}
{"train_loss": 0.026344163343310356, "global_step": 253846, "epoch": 2852, "lr": 1.0378310427335447e-06}
{"train_loss": 0.026843009516596794, "global_step": 253847, "epoch": 2852, "lr": 1.0377135404020243e-06}
{"train_loss": 0.059019602835178375, "global_step": 253848, "epoch": 2852, "lr": 1.0375960446528831e-06}
{"train_loss": 0.04309283196926117, "global_step": 253849, "epoch": 2852, "lr": 1.0374785554861432e-06}
{"train_loss": 0.023423049598932266, "global_step": 253850, "epoch": 2852, "lr": 1.0373610729018157e-06}
{"train_loss": 0.044972971081733704, "global_step": 253851, "epoch": 2852, "lr": 1.0372435968999172e-06}
{"train_loss": 0.06186586990952492, "global_step": 253852, "epoch": 2852, "lr": 1.0371261274804644e-06}
{"train_loss": 0.04449963569641113, "global_step": 253853, "epoch": 2852, "lr": 1.037008664643474e-06}
{"train_loss": 0.03881032019853592, "global_step": 253854, "epoch": 2852, "lr": 1.036891208388957e-06}
{"train_loss": 0.06610992550849915, "global_step": 253855, "epoch": 2852, "lr": 1.0367737587169412e-06}
{"train_loss": 0.0741618275642395, "global_step": 253856, "epoch": 2852, "lr": 1.0366563156274322e-06}
{"train_loss": 0.022981783375144005, "global_step": 253857, "epoch": 2852, "lr": 1.0365388791204467e-06}
{"train_loss": 0.07084786891937256, "global_step": 253858, "epoch": 2852, "lr": 1.0364214491960012e-06}
{"train_loss": 0.021789059042930603, "global_step": 253859, "epoch": 2852, "lr": 1.0363040258541123e-06}
{"train_loss": 0.017101788893342018, "global_step": 253860, "epoch": 2852, "lr": 1.036186609094797e-06}
{"train_loss": 0.02638089656829834, "global_step": 253861, "epoch": 2852, "lr": 1.036069198918066e-06}
{"train_loss": 0.056852929294109344, "global_step": 253862, "epoch": 2852, "lr": 1.0359517953239417e-06}
{"train_loss": 0.055555302649736404, "global_step": 253863, "epoch": 2852, "lr": 1.0358343983124408e-06}
{"train_loss": 0.021932756528258324, "global_step": 253864, "epoch": 2852, "lr": 1.0357170078835632e-06}
{"train_loss": 0.05151892825961113, "global_step": 253865, "epoch": 2852, "lr": 1.0355996240373478e-06}
{"train_loss": 0.02448061667382717, "global_step": 253866, "epoch": 2852, "lr": 1.0354822467737945e-06}
{"train_loss": 0.03502289205789566, "global_step": 253867, "epoch": 2852, "lr": 1.0353648760929258e-06}
{"train_loss": 0.029963482171297073, "global_step": 253868, "epoch": 2852, "lr": 1.0352475119947525e-06}
{"train_loss": 0.03165990859270096, "global_step": 253869, "epoch": 2852, "lr": 1.035130154479297e-06}
{"train_loss": 0.06873831897974014, "global_step": 253870, "epoch": 2852, "lr": 1.03501280354657e-06}
{"train_loss": 0.033720891922712326, "global_step": 253871, "epoch": 2852, "lr": 1.0348954591965887e-06}
{"train_loss": 0.02626507170498371, "global_step": 253872, "epoch": 2852, "lr": 1.0347781214293694e-06}
{"train_loss": 0.07238351553678513, "global_step": 253873, "epoch": 2852, "lr": 1.0346607902449235e-06}
{"train_loss": 0.06322459131479263, "global_step": 253874, "epoch": 2852, "lr": 1.0345434656432783e-06}
{"train_loss": 0.045549724251031876, "global_step": 253875, "epoch": 2852, "lr": 1.0344261476244343e-06}
{"train_loss": 0.05459696426987648, "global_step": 253876, "epoch": 2852, "lr": 1.034308836188419e-06}
{"train_loss": 0.059478748589754105, "global_step": 253877, "epoch": 2852, "lr": 1.0341915313352434e-06}
{"train_loss": 0.06868679821491241, "global_step": 253878, "epoch": 2852, "lr": 1.0340742330649244e-06}
{"train_loss": 0.049846794456243515, "global_step": 253879, "epoch": 2852, "lr": 1.0339569413774785e-06}
{"train_loss": 0.03788229078054428, "global_step": 253880, "epoch": 2852, "lr": 1.033839656272917e-06}
{"train_loss": 0.04968726262450218, "global_step": 253881, "epoch": 2852, "lr": 1.0337223777512616e-06}
{"train_loss": 0.029907912015914917, "global_step": 253882, "epoch": 2852, "lr": 1.033605105812524e-06}
{"train_loss": 0.02470879815518856, "global_step": 253883, "epoch": 2852, "lr": 1.0334878404567262e-06}
{"train_loss": 0.09637812525033951, "global_step": 253884, "epoch": 2852, "lr": 1.0333705816838735e-06}
{"train_loss": 0.06264681369066238, "global_step": 253885, "epoch": 2852, "lr": 1.0332533294939883e-06}
{"train_loss": 0.055837225168943405, "global_step": 253886, "epoch": 2852, "lr": 1.0331360838870875e-06}
{"train_loss": 0.013755875639617443, "global_step": 253887, "epoch": 2852, "lr": 1.0330188448631817e-06}
{"train_loss": 0.021961867809295654, "global_step": 253888, "epoch": 2852, "lr": 1.0329016124222879e-06}
{"train_loss": 0.027580203488469124, "global_step": 253889, "epoch": 2852, "lr": 1.0327843865644282e-06}
{"train_loss": 0.02210613340139389, "global_step": 253890, "epoch": 2852, "lr": 1.0326671672896083e-06}
{"train_loss": 0.05298275500535965, "global_step": 253891, "epoch": 2852, "lr": 1.0325499545978556e-06}
{"train_loss": 0.03338983282446861, "global_step": 253892, "epoch": 2852, "lr": 1.032432748489176e-06}
{"train_loss": 0.03854444622993469, "global_step": 253893, "epoch": 2852, "lr": 1.0323155489635916e-06}
{"train_loss": 0.03136599436402321, "global_step": 253894, "epoch": 2852, "lr": 1.0321983560211136e-06}
{"train_loss": 0.05284739285707474, "global_step": 253895, "epoch": 2852, "lr": 1.0320811696617638e-06}
{"train_loss": 0.053157273679971695, "global_step": 253896, "epoch": 2852, "lr": 1.0319639898855481e-06}
{"train_loss": 0.10299011319875717, "global_step": 253897, "epoch": 2852, "lr": 1.0318468166924888e-06}
{"train_loss": 0.014220385812222958, "global_step": 253898, "epoch": 2852, "lr": 1.0317296500826024e-06}
{"train_loss": 0.008412668481469154, "global_step": 253899, "epoch": 2852, "lr": 1.0316124900559054e-06}
{"train_loss": 0.04927589371800423, "global_step": 253900, "epoch": 2852, "lr": 1.031495336612409e-06}
{"train_loss": 0.03556274622678757, "global_step": 253901, "epoch": 2852, "lr": 1.0313781897521303e-06}
{"train_loss": 0.044087957590818405, "global_step": 253902, "epoch": 2852, "lr": 1.0312610494750851e-06}
{"train_loss": 0.04944838583469391, "global_step": 253903, "epoch": 2852, "lr": 1.0311439157812963e-06}
{"train_loss": 0.04804704710841179, "global_step": 253904, "epoch": 2852, "lr": 1.0310267886707636e-06}
{"train_loss": 0.05666627734899521, "global_step": 253905, "epoch": 2852, "lr": 1.0309096681435148e-06}
{"train_loss": 0.025422563776373863, "global_step": 253906, "epoch": 2852, "lr": 1.0307925541995722e-06}
{"train_loss": 0.06412612646818161, "global_step": 253907, "epoch": 2852, "lr": 1.0306754468389357e-06}
{"train_loss": 0.029263347387313843, "global_step": 253908, "epoch": 2852, "lr": 1.0305583460616274e-06}
{"train_loss": 0.03745941072702408, "global_step": 253909, "epoch": 2852, "lr": 1.0304412518676643e-06}
{"train_loss": 0.022854432463645935, "global_step": 253910, "epoch": 2852, "lr": 1.0303241642570627e-06}
{"train_loss": 0.04018831625580788, "global_step": 253911, "epoch": 2852, "lr": 1.0302070832298393e-06}
{"train_loss": 0.0515044741332531, "global_step": 253912, "epoch": 2852, "lr": 1.0300900087860056e-06}
{"train_loss": 0.02922852151095867, "global_step": 253913, "epoch": 2852, "lr": 1.0299729409255777e-06}
{"train_loss": 0.04203270003199577, "global_step": 253914, "epoch": 2852, "lr": 1.029855879648578e-06}
{"train_loss": 0.025228260084986687, "global_step": 253915, "epoch": 2852, "lr": 1.0297388249550122e-06}
{"train_loss": 0.044511475254980365, "global_step": 253916, "epoch": 2852, "lr": 1.0296217768449023e-06, "val_loss": 9.62517261505127}
{"train_loss": 0.04616484418511391, "global_step": 253917, "epoch": 2853, "lr": 1.0295047353182597e-06}
{"train_loss": 0.01415469404309988, "global_step": 253918, "epoch": 2853, "lr": 1.0293877003751118e-06}
{"train_loss": 0.02371528185904026, "global_step": 253919, "epoch": 2853, "lr": 1.0292706720154644e-06}
{"train_loss": 0.06458112597465515, "global_step": 253920, "epoch": 2853, "lr": 1.0291536502393283e-06}
{"train_loss": 0.04581684246659279, "global_step": 253921, "epoch": 2853, "lr": 1.0290366350467317e-06}
{"train_loss": 0.023322541266679764, "global_step": 253922, "epoch": 2853, "lr": 1.02891962643768e-06}
{"train_loss": 0.0437089167535305, "global_step": 253923, "epoch": 2853, "lr": 1.028802624412195e-06}
{"train_loss": 0.03124413453042507, "global_step": 253924, "epoch": 2853, "lr": 1.0286856289702885e-06}
{"train_loss": 0.033649031072854996, "global_step": 253925, "epoch": 2853, "lr": 1.0285686401119821e-06}
{"train_loss": 0.009608260355889797, "global_step": 253926, "epoch": 2853, "lr": 1.0284516578372815e-06}
{"train_loss": 0.018271280452609062, "global_step": 253927, "epoch": 2853, "lr": 1.0283346821462148e-06}
{"train_loss": 0.04656019061803818, "global_step": 253928, "epoch": 2853, "lr": 1.028217713038787e-06}
{"train_loss": 0.07906660437583923, "global_step": 253929, "epoch": 2853, "lr": 1.0281007505150264e-06}
{"train_loss": 0.09002802520990372, "global_step": 253930, "epoch": 2853, "lr": 1.0279837945749327e-06}
{"train_loss": 0.05286594107747078, "global_step": 253931, "epoch": 2853, "lr": 1.0278668452185337e-06}
{"train_loss": 0.03465771675109863, "global_step": 253932, "epoch": 2853, "lr": 1.0277499024458403e-06}
{"train_loss": 0.03557952865958214, "global_step": 253933, "epoch": 2853, "lr": 1.0276329662568696e-06}
{"train_loss": 0.08154445141553879, "global_step": 253934, "epoch": 2853, "lr": 1.0275160366516322e-06}
{"train_loss": 0.04210938885807991, "global_step": 253935, "epoch": 2853, "lr": 1.0273991136301565e-06}
{"train_loss": 0.04842720180749893, "global_step": 253936, "epoch": 2853, "lr": 1.0272821971924419e-06}
{"train_loss": 0.015503617003560066, "global_step": 253937, "epoch": 2853, "lr": 1.0271652873385161e-06}
{"train_loss": 0.04778091982007027, "global_step": 253938, "epoch": 2853, "lr": 1.027048384068391e-06}
{"train_loss": 0.024923644959926605, "global_step": 253939, "epoch": 2853, "lr": 1.0269314873820824e-06}
{"train_loss": 0.044970981776714325, "global_step": 253940, "epoch": 2853, "lr": 1.0268145972796072e-06}
{"train_loss": 0.04830755293369293, "global_step": 253941, "epoch": 2853, "lr": 1.0266977137609769e-06}
{"train_loss": 0.09749983251094818, "global_step": 253942, "epoch": 2853, "lr": 1.026580836826213e-06}
{"train_loss": 0.04033612087368965, "global_step": 253943, "epoch": 2853, "lr": 1.0264639664753216e-06}
{"train_loss": 0.03860945254564285, "global_step": 253944, "epoch": 2853, "lr": 1.026347102708336e-06}
{"train_loss": 0.050089288502931595, "global_step": 253945, "epoch": 2853, "lr": 1.0262302455252559e-06}
{"train_loss": 0.03216109797358513, "global_step": 253946, "epoch": 2853, "lr": 1.0261133949261038e-06}
{"train_loss": 0.06594754755496979, "global_step": 253947, "epoch": 2853, "lr": 1.0259965509108903e-06}
{"train_loss": 0.04061245918273926, "global_step": 253948, "epoch": 2853, "lr": 1.0258797134796383e-06}
{"train_loss": 0.045526545494794846, "global_step": 253949, "epoch": 2853, "lr": 1.0257628826323529e-06}
{"train_loss": 0.05421794205904007, "global_step": 253950, "epoch": 2853, "lr": 1.0256460583690675e-06}
{"train_loss": 0.030731843784451485, "global_step": 253951, "epoch": 2853, "lr": 1.0255292406897766e-06}
{"train_loss": 0.02832474373281002, "global_step": 253952, "epoch": 2853, "lr": 1.0254124295945134e-06}
{"train_loss": 0.060724228620529175, "global_step": 253953, "epoch": 2853, "lr": 1.0252956250832835e-06}
{"train_loss": 0.0161841232329607, "global_step": 253954, "epoch": 2853, "lr": 1.0251788271561091e-06}
{"train_loss": 0.04270428046584129, "global_step": 253955, "epoch": 2853, "lr": 1.0250620358130015e-06}
{"train_loss": 0.026639772579073906, "global_step": 253956, "epoch": 2853, "lr": 1.024945251053977e-06}
{"train_loss": 0.016522828489542007, "global_step": 253957, "epoch": 2853, "lr": 1.0248284728790526e-06}
{"train_loss": 0.047231245785951614, "global_step": 253958, "epoch": 2853, "lr": 1.024711701288239e-06}
{"train_loss": 0.04561146721243858, "global_step": 253959, "epoch": 2853, "lr": 1.0245949362815588e-06}
{"train_loss": 0.037305958569049835, "global_step": 253960, "epoch": 2853, "lr": 1.024478177859023e-06}
{"train_loss": 0.022732384502887726, "global_step": 253961, "epoch": 2853, "lr": 1.0243614260206536e-06}
{"train_loss": 0.02957378700375557, "global_step": 253962, "epoch": 2853, "lr": 1.0242446807664563e-06}
{"train_loss": 0.066419817507267, "global_step": 253963, "epoch": 2853, "lr": 1.0241279420964588e-06}
{"train_loss": 0.029383402317762375, "global_step": 253964, "epoch": 2853, "lr": 1.0240112100106615e-06}
{"train_loss": 0.0346396341919899, "global_step": 253965, "epoch": 2853, "lr": 1.023894484509097e-06}
{"train_loss": 0.04403701797127724, "global_step": 253966, "epoch": 2853, "lr": 1.0237777655917658e-06}
{"train_loss": 0.04913509264588356, "global_step": 253967, "epoch": 2853, "lr": 1.023661053258701e-06}
{"train_loss": 0.03973916172981262, "global_step": 253968, "epoch": 2853, "lr": 1.0235443475098972e-06}
{"train_loss": 0.052850391715765, "global_step": 253969, "epoch": 2853, "lr": 1.0234276483453876e-06}
{"train_loss": 0.07933451235294342, "global_step": 253970, "epoch": 2853, "lr": 1.0233109557651776e-06}
{"train_loss": 0.06124892085790634, "global_step": 253971, "epoch": 2853, "lr": 1.0231942697692898e-06}
{"train_loss": 0.03170383721590042, "global_step": 253972, "epoch": 2853, "lr": 1.023077590357735e-06}
{"train_loss": 0.023010967299342155, "global_step": 253973, "epoch": 2853, "lr": 1.02296091753053e-06}
{"train_loss": 0.09006047993898392, "global_step": 253974, "epoch": 2853, "lr": 1.0228442512876912e-06}
{"train_loss": 0.086616151034832, "global_step": 253975, "epoch": 2853, "lr": 1.0227275916292356e-06}
{"train_loss": 0.019007014110684395, "global_step": 253976, "epoch": 2853, "lr": 1.022610938555174e-06}
{"train_loss": 0.048217758536338806, "global_step": 253977, "epoch": 2853, "lr": 1.022494292065529e-06}
{"train_loss": 0.044441722333431244, "global_step": 253978, "epoch": 2853, "lr": 1.0223776521603056e-06}
{"train_loss": 0.04694782570004463, "global_step": 253979, "epoch": 2853, "lr": 1.022261018839532e-06}
{"train_loss": 0.05702993646264076, "global_step": 253980, "epoch": 2853, "lr": 1.0221443921032192e-06}
{"train_loss": 0.019446613267064095, "global_step": 253981, "epoch": 2853, "lr": 1.0220277719513783e-06}
{"train_loss": 0.02494749054312706, "global_step": 253982, "epoch": 2853, "lr": 1.0219111583840368e-06}
{"train_loss": 0.07523436844348907, "global_step": 253983, "epoch": 2853, "lr": 1.021794551401195e-06}
{"train_loss": 0.043963123112916946, "global_step": 253984, "epoch": 2853, "lr": 1.0216779510028806e-06}
{"train_loss": 0.05172008275985718, "global_step": 253985, "epoch": 2853, "lr": 1.0215613571890993e-06}
{"train_loss": 0.03925444930791855, "global_step": 253986, "epoch": 2853, "lr": 1.0214447699598784e-06}
{"train_loss": 0.047485075891017914, "global_step": 253987, "epoch": 2853, "lr": 1.0213281893152237e-06}
{"train_loss": 0.04788386821746826, "global_step": 253988, "epoch": 2853, "lr": 1.0212116152551576e-06}
{"train_loss": 0.06717385351657867, "global_step": 253989, "epoch": 2853, "lr": 1.0210950477796856e-06}
{"train_loss": 0.0249742791056633, "global_step": 253990, "epoch": 2853, "lr": 1.020978486888835e-06}
{"train_loss": 0.04524759575724602, "global_step": 253991, "epoch": 2853, "lr": 1.0208619325826174e-06}
{"train_loss": 0.04531693458557129, "global_step": 253992, "epoch": 2853, "lr": 1.0207453848610493e-06}
{"train_loss": 0.05186579003930092, "global_step": 253993, "epoch": 2853, "lr": 1.0206288437241417e-06}
{"train_loss": 0.031023908406496048, "global_step": 253994, "epoch": 2853, "lr": 1.020512309171917e-06}
{"train_loss": 0.006916247308254242, "global_step": 253995, "epoch": 2853, "lr": 1.0203957812043864e-06}
{"train_loss": 0.03872789442539215, "global_step": 253996, "epoch": 2853, "lr": 1.0202792598215604e-06}
{"train_loss": 0.04799038544297218, "global_step": 253997, "epoch": 2853, "lr": 1.0201627450234675e-06}
{"train_loss": 0.033377744257450104, "global_step": 253998, "epoch": 2853, "lr": 1.0200462368101127e-06}
{"train_loss": 0.03603329882025719, "global_step": 253999, "epoch": 2853, "lr": 1.019929735181524e-06}
{"train_loss": 0.04484447091817856, "global_step": 254000, "epoch": 2853, "lr": 1.0198132401377014e-06}
{"train_loss": 0.013612382113933563, "global_step": 254001, "epoch": 2853, "lr": 1.019696751678667e-06}
{"train_loss": 0.05602237954735756, "global_step": 254002, "epoch": 2853, "lr": 1.0195802698044376e-06}
{"train_loss": 0.032116565853357315, "global_step": 254003, "epoch": 2853, "lr": 1.0194637945150353e-06}
{"train_loss": 0.04671565815806389, "global_step": 254004, "epoch": 2853, "lr": 1.0193473258104601e-06}
{"train_loss": 0.04312291373027844, "global_step": 254005, "epoch": 2853, "lr": 1.0192308636907456e-06, "val_loss": 9.594799995422363}
{"train_loss": 0.03670487552881241, "global_step": 254006, "epoch": 2854, "lr": 1.0191144081558912e-06}
{"train_loss": 0.04184502735733986, "global_step": 254007, "epoch": 2854, "lr": 1.0189979592059252e-06}
{"train_loss": 0.028813477605581284, "global_step": 254008, "epoch": 2854, "lr": 1.0188815168408528e-06}
{"train_loss": 0.029509687796235085, "global_step": 254009, "epoch": 2854, "lr": 1.018765081060702e-06}
{"train_loss": 0.04860333353281021, "global_step": 254010, "epoch": 2854, "lr": 1.0186486518654725e-06}
{"train_loss": 0.03913865238428116, "global_step": 254011, "epoch": 2854, "lr": 1.0185322292551979e-06}
{"train_loss": 0.02906230464577675, "global_step": 254012, "epoch": 2854, "lr": 1.0184158132298726e-06}
{"train_loss": 0.08188075572252274, "global_step": 254013, "epoch": 2854, "lr": 1.0182994037895354e-06}
{"train_loss": 0.061922766268253326, "global_step": 254014, "epoch": 2854, "lr": 1.0181830009341863e-06}
{"train_loss": 0.0543353445827961, "global_step": 254015, "epoch": 2854, "lr": 1.018066604663842e-06}
{"train_loss": 0.021798409521579742, "global_step": 254016, "epoch": 2854, "lr": 1.0179502149785302e-06}
{"train_loss": 0.13960807025432587, "global_step": 254017, "epoch": 2854, "lr": 1.017833831878251e-06}
{"train_loss": 0.04902569577097893, "global_step": 254018, "epoch": 2854, "lr": 1.0177174553630319e-06}
{"train_loss": 0.035003937780857086, "global_step": 254019, "epoch": 2854, "lr": 1.0176010854328788e-06}
{"train_loss": 0.03435997664928436, "global_step": 254020, "epoch": 2854, "lr": 1.0174847220878137e-06}
{"train_loss": 0.0467110276222229, "global_step": 254021, "epoch": 2854, "lr": 1.0173683653278531e-06}
{"train_loss": 0.027699273079633713, "global_step": 254022, "epoch": 2854, "lr": 1.0172520151530085e-06}
{"train_loss": 0.08732810616493225, "global_step": 254023, "epoch": 2854, "lr": 1.0171356715632963e-06}
{"train_loss": 0.04547566547989845, "global_step": 254024, "epoch": 2854, "lr": 1.0170193345587387e-06}
{"train_loss": 0.02430480159819126, "global_step": 254025, "epoch": 2854, "lr": 1.0169030041393357e-06}
{"train_loss": 0.04090166464447975, "global_step": 254026, "epoch": 2854, "lr": 1.0167866803051207e-06}
{"train_loss": 0.046133898198604584, "global_step": 254027, "epoch": 2854, "lr": 1.016670363056099e-06}
{"train_loss": 0.07899880409240723, "global_step": 254028, "epoch": 2854, "lr": 1.0165540523922935e-06}
{"train_loss": 0.06027393788099289, "global_step": 254029, "epoch": 2854, "lr": 1.0164377483137089e-06}
{"train_loss": 0.04896435886621475, "global_step": 254030, "epoch": 2854, "lr": 1.0163214508203734e-06}
{"train_loss": 0.06287645548582077, "global_step": 254031, "epoch": 2854, "lr": 1.0162051599122925e-06}
{"train_loss": 0.019411209970712662, "global_step": 254032, "epoch": 2854, "lr": 1.0160888755894882e-06}
{"train_loss": 0.012734485790133476, "global_step": 254033, "epoch": 2854, "lr": 1.0159725978519664e-06}
{"train_loss": 0.033272020518779755, "global_step": 254034, "epoch": 2854, "lr": 1.01585632669976e-06}
{"train_loss": 0.04408174008131027, "global_step": 254035, "epoch": 2854, "lr": 1.015740062132875e-06}
{"train_loss": 0.03610020503401756, "global_step": 254036, "epoch": 2854, "lr": 1.0156238041513167e-06}
{"train_loss": 0.0770144909620285, "global_step": 254037, "epoch": 2854, "lr": 1.0155075527551184e-06}
{"train_loss": 0.0660237967967987, "global_step": 254038, "epoch": 2854, "lr": 1.0153913079442856e-06}
{"train_loss": 0.03472781553864479, "global_step": 254039, "epoch": 2854, "lr": 1.015275069718835e-06}
{"train_loss": 0.06683829426765442, "global_step": 254040, "epoch": 2854, "lr": 1.0151588380787836e-06}
{"train_loss": 0.026310274377465248, "global_step": 254041, "epoch": 2854, "lr": 1.015042613024153e-06}
{"train_loss": 0.03766203671693802, "global_step": 254042, "epoch": 2854, "lr": 1.0149263945549491e-06}
{"train_loss": 0.05590580403804779, "global_step": 254043, "epoch": 2854, "lr": 1.0148101826711886e-06}
{"train_loss": 0.024104202166199684, "global_step": 254044, "epoch": 2854, "lr": 1.0146939773728936e-06}
{"train_loss": 0.02973068691790104, "global_step": 254045, "epoch": 2854, "lr": 1.0145777786600753e-06}
{"train_loss": 0.0832943394780159, "global_step": 254046, "epoch": 2854, "lr": 1.0144615865327445e-06}
{"train_loss": 0.05538700148463249, "global_step": 254047, "epoch": 2854, "lr": 1.0143454009909291e-06}
{"train_loss": 0.03948543965816498, "global_step": 254048, "epoch": 2854, "lr": 1.014229222034635e-06}
{"train_loss": 0.04158703237771988, "global_step": 254049, "epoch": 2854, "lr": 1.014113049663884e-06}
{"train_loss": 0.05592066049575806, "global_step": 254050, "epoch": 2854, "lr": 1.0139968838786817e-06}
{"train_loss": 0.02981741540133953, "global_step": 254051, "epoch": 2854, "lr": 1.013880724679056e-06}
{"train_loss": 0.02303396165370941, "global_step": 254052, "epoch": 2854, "lr": 1.013764572065018e-06}
{"train_loss": 0.06922659277915955, "global_step": 254053, "epoch": 2854, "lr": 1.013648426036573e-06}
{"train_loss": 0.03283633291721344, "global_step": 254054, "epoch": 2854, "lr": 1.0135322865937546e-06}
{"train_loss": 0.0676015093922615, "global_step": 254055, "epoch": 2854, "lr": 1.0134161537365684e-06}
{"train_loss": 0.05880917236208916, "global_step": 254056, "epoch": 2854, "lr": 1.0133000274650305e-06}
{"train_loss": 0.030828850343823433, "global_step": 254057, "epoch": 2854, "lr": 1.0131839077791584e-06}
{"train_loss": 0.030965229496359825, "global_step": 254058, "epoch": 2854, "lr": 1.013067794678968e-06}
{"train_loss": 0.08206857740879059, "global_step": 254059, "epoch": 2854, "lr": 1.0129516881644707e-06}
{"train_loss": 0.048243194818496704, "global_step": 254060, "epoch": 2854, "lr": 1.0128355882356833e-06}
{"train_loss": 0.08646528422832489, "global_step": 254061, "epoch": 2854, "lr": 1.0127194948926277e-06}
{"train_loss": 0.019728826358914375, "global_step": 254062, "epoch": 2854, "lr": 1.0126034081353153e-06}
{"train_loss": 0.04811156168580055, "global_step": 254063, "epoch": 2854, "lr": 1.012487327963757e-06}
{"train_loss": 0.034083519130945206, "global_step": 254064, "epoch": 2854, "lr": 1.012371254377975e-06}
{"train_loss": 0.02283766306936741, "global_step": 254065, "epoch": 2854, "lr": 1.0122551873779806e-06}
{"train_loss": 0.04342781752347946, "global_step": 254066, "epoch": 2854, "lr": 1.0121391269637959e-06}
{"train_loss": 0.018837949261069298, "global_step": 254067, "epoch": 2854, "lr": 1.0120230731354263e-06}
{"train_loss": 0.029043134301900864, "global_step": 254068, "epoch": 2854, "lr": 1.0119070258928942e-06}
{"train_loss": 0.07335386425256729, "global_step": 254069, "epoch": 2854, "lr": 1.0117909852362161e-06}
{"train_loss": 0.04563911259174347, "global_step": 254070, "epoch": 2854, "lr": 1.0116749511654089e-06}
{"train_loss": 0.08997747302055359, "global_step": 254071, "epoch": 2854, "lr": 1.0115589236804778e-06}
{"train_loss": 0.09655401855707169, "global_step": 254072, "epoch": 2854, "lr": 1.011442902781451e-06}
{"train_loss": 0.054406024515628815, "global_step": 254073, "epoch": 2854, "lr": 1.0113268884683392e-06}
{"train_loss": 0.06820185482501984, "global_step": 254074, "epoch": 2854, "lr": 1.0112108807411536e-06}
{"train_loss": 0.02540450356900692, "global_step": 254075, "epoch": 2854, "lr": 1.0110948795999165e-06}
{"train_loss": 0.032375600188970566, "global_step": 254076, "epoch": 2854, "lr": 1.0109788850446334e-06}
{"train_loss": 0.07957081496715546, "global_step": 254077, "epoch": 2854, "lr": 1.0108628970753376e-06}
{"train_loss": 0.053839344531297684, "global_step": 254078, "epoch": 2854, "lr": 1.0107469156920235e-06}
{"train_loss": 0.04379871487617493, "global_step": 254079, "epoch": 2854, "lr": 1.0106309408947246e-06}
{"train_loss": 0.049048878252506256, "global_step": 254080, "epoch": 2854, "lr": 1.0105149726834461e-06}
{"train_loss": 0.03693285211920738, "global_step": 254081, "epoch": 2854, "lr": 1.0103990110582107e-06}
{"train_loss": 0.04551941901445389, "global_step": 254082, "epoch": 2854, "lr": 1.010283056019029e-06}
{"train_loss": 0.0755627229809761, "global_step": 254083, "epoch": 2854, "lr": 1.010167107565918e-06}
{"train_loss": 0.038937244564294815, "global_step": 254084, "epoch": 2854, "lr": 1.0100511656988886e-06}
{"train_loss": 0.03904949501156807, "global_step": 254085, "epoch": 2854, "lr": 1.0099352304179633e-06}
{"train_loss": 0.05027903988957405, "global_step": 254086, "epoch": 2854, "lr": 1.0098193017231528e-06}
{"train_loss": 0.05657914653420448, "global_step": 254087, "epoch": 2854, "lr": 1.0097033796144795e-06}
{"train_loss": 0.03588629513978958, "global_step": 254088, "epoch": 2854, "lr": 1.0095874640919544e-06}
{"train_loss": 0.060157474130392075, "global_step": 254089, "epoch": 2854, "lr": 1.0094715551555888e-06}
{"train_loss": 0.046956080943346024, "global_step": 254090, "epoch": 2854, "lr": 1.009355652805405e-06}
{"train_loss": 0.0330684669315815, "global_step": 254091, "epoch": 2854, "lr": 1.0092397570414137e-06}
{"train_loss": 0.03133339062333107, "global_step": 254092, "epoch": 2854, "lr": 1.0091238678636372e-06}
{"train_loss": 0.029007811099290848, "global_step": 254093, "epoch": 2854, "lr": 1.009007985272087e-06}
{"train_loss": 0.047872498191976816, "global_step": 254094, "epoch": 2854, "lr": 1.0088921092667792e-06, "val_loss": 9.727502822875977}
{"train_loss": 0.026863068342208862, "global_step": 254095, "epoch": 2855, "lr": 1.0087762398477195e-06}
{"train_loss": 0.054549138993024826, "global_step": 254096, "epoch": 2855, "lr": 1.0086603770149416e-06}
{"train_loss": 0.04952089115977287, "global_step": 254097, "epoch": 2855, "lr": 1.0085445207684508e-06}
{"train_loss": 0.03320992365479469, "global_step": 254098, "epoch": 2855, "lr": 1.0084286711082636e-06}
{"train_loss": 0.046846501529216766, "global_step": 254099, "epoch": 2855, "lr": 1.0083128280343968e-06}
{"train_loss": 0.09978818893432617, "global_step": 254100, "epoch": 2855, "lr": 1.008196991546867e-06}
{"train_loss": 0.021351078525185585, "global_step": 254101, "epoch": 2855, "lr": 1.0080811616456798e-06}
{"train_loss": 0.114947110414505, "global_step": 254102, "epoch": 2855, "lr": 1.0079653383308684e-06}
{"train_loss": 0.04419192299246788, "global_step": 254103, "epoch": 2855, "lr": 1.007849521602433e-06}
{"train_loss": 0.04527159035205841, "global_step": 254104, "epoch": 2855, "lr": 1.0077337114603958e-06}
{"train_loss": 0.05389579012989998, "global_step": 254105, "epoch": 2855, "lr": 1.0076179079047732e-06}
{"train_loss": 0.07919856905937195, "global_step": 254106, "epoch": 2855, "lr": 1.007502110935582e-06}
{"train_loss": 0.05301091447472572, "global_step": 254107, "epoch": 2855, "lr": 1.0073863205528277e-06}
{"train_loss": 0.03931156173348427, "global_step": 254108, "epoch": 2855, "lr": 1.0072705367565382e-06}
{"train_loss": 0.022680632770061493, "global_step": 254109, "epoch": 2855, "lr": 1.0071547595467246e-06}
{"train_loss": 0.07245160639286041, "global_step": 254110, "epoch": 2855, "lr": 1.0070389889234034e-06}
{"train_loss": 0.02806234359741211, "global_step": 254111, "epoch": 2855, "lr": 1.0069232248865856e-06}
{"train_loss": 0.030431699007749557, "global_step": 254112, "epoch": 2855, "lr": 1.0068074674362882e-06}
{"train_loss": 0.04913584887981415, "global_step": 254113, "epoch": 2855, "lr": 1.0066917165725331e-06}
{"train_loss": 0.06835124641656876, "global_step": 254114, "epoch": 2855, "lr": 1.006575972295326e-06}
{"train_loss": 0.052255962044000626, "global_step": 254115, "epoch": 2855, "lr": 1.0064602346046892e-06}
{"train_loss": 0.06507077813148499, "global_step": 254116, "epoch": 2855, "lr": 1.0063445035006392e-06}
{"train_loss": 0.059122562408447266, "global_step": 254117, "epoch": 2855, "lr": 1.006228778983187e-06}
{"train_loss": 0.027901804074645042, "global_step": 254118, "epoch": 2855, "lr": 1.0061130610523494e-06}
{"train_loss": 0.052827928215265274, "global_step": 254119, "epoch": 2855, "lr": 1.0059973497081433e-06}
{"train_loss": 0.055812086910009384, "global_step": 254120, "epoch": 2855, "lr": 1.0058816449505847e-06}
{"train_loss": 0.05951306223869324, "global_step": 254121, "epoch": 2855, "lr": 1.005765946779691e-06}
{"train_loss": 0.04322223365306854, "global_step": 254122, "epoch": 2855, "lr": 1.0056502551954673e-06}
{"train_loss": 0.059168439358472824, "global_step": 254123, "epoch": 2855, "lr": 1.0055345701979412e-06}
{"train_loss": 0.03753768652677536, "global_step": 254124, "epoch": 2855, "lr": 1.0054188917871244e-06}
{"train_loss": 0.04142722114920616, "global_step": 254125, "epoch": 2855, "lr": 1.0053032199630274e-06}
{"train_loss": 0.07416350394487381, "global_step": 254126, "epoch": 2855, "lr": 1.0051875547256784e-06}
{"train_loss": 0.05096318945288658, "global_step": 254127, "epoch": 2855, "lr": 1.0050718960750772e-06}
{"train_loss": 0.041845593601465225, "global_step": 254128, "epoch": 2855, "lr": 1.0049562440112514e-06}
{"train_loss": 0.05199962481856346, "global_step": 254129, "epoch": 2855, "lr": 1.004840598534207e-06}
{"train_loss": 0.054725103080272675, "global_step": 254130, "epoch": 2855, "lr": 1.004724959643971e-06}
{"train_loss": 0.07526540011167526, "global_step": 254131, "epoch": 2855, "lr": 1.00460932734055e-06}
{"train_loss": 0.028936052694916725, "global_step": 254132, "epoch": 2855, "lr": 1.0044937016239598e-06}
{"train_loss": 0.05058935284614563, "global_step": 254133, "epoch": 2855, "lr": 1.0043780824942227e-06}
{"train_loss": 0.045631617307662964, "global_step": 254134, "epoch": 2855, "lr": 1.0042624699513503e-06}
{"train_loss": 0.062105223536491394, "global_step": 254135, "epoch": 2855, "lr": 1.0041468639953477e-06}
{"train_loss": 0.01402311772108078, "global_step": 254136, "epoch": 2855, "lr": 1.004031264626254e-06}
{"train_loss": 0.046815574169158936, "global_step": 254137, "epoch": 2855, "lr": 1.003915671844058e-06}
{"train_loss": 0.016517087817192078, "global_step": 254138, "epoch": 2855, "lr": 1.0038000856487982e-06}
{"train_loss": 0.029379965737462044, "global_step": 254139, "epoch": 2855, "lr": 1.0036845060404754e-06}
{"train_loss": 0.031804926693439484, "global_step": 254140, "epoch": 2855, "lr": 1.0035689330191168e-06}
{"train_loss": 0.08485551923513412, "global_step": 254141, "epoch": 2855, "lr": 1.0034533665847223e-06}
{"train_loss": 0.05135689675807953, "global_step": 254142, "epoch": 2855, "lr": 1.00333780673732e-06}
{"train_loss": 0.0559048093855381, "global_step": 254143, "epoch": 2855, "lr": 1.003222253476921e-06}
{"train_loss": 0.07370160520076752, "global_step": 254144, "epoch": 2855, "lr": 1.0031067068035416e-06}
{"train_loss": 0.08739352226257324, "global_step": 254145, "epoch": 2855, "lr": 1.0029911667171988e-06}
{"train_loss": 0.012669703923165798, "global_step": 254146, "epoch": 2855, "lr": 1.002875633217909e-06}
{"train_loss": 0.03420067951083183, "global_step": 254147, "epoch": 2855, "lr": 1.002760106305678e-06}
{"train_loss": 0.03795542195439339, "global_step": 254148, "epoch": 2855, "lr": 1.0026445859805389e-06}
{"train_loss": 0.07437408715486526, "global_step": 254149, "epoch": 2855, "lr": 1.0025290722424862e-06}
{"train_loss": 0.028260627761483192, "global_step": 254150, "epoch": 2855, "lr": 1.0024135650915534e-06}
{"train_loss": 0.04250849410891533, "global_step": 254151, "epoch": 2855, "lr": 1.0022980645277514e-06}
{"train_loss": 0.03045620582997799, "global_step": 254152, "epoch": 2855, "lr": 1.0021825705510856e-06}
{"train_loss": 0.06323657929897308, "global_step": 254153, "epoch": 2855, "lr": 1.002067083161584e-06}
{"train_loss": 0.03038480319082737, "global_step": 254154, "epoch": 2855, "lr": 1.0019516023592524e-06}
{"train_loss": 0.022798534482717514, "global_step": 254155, "epoch": 2855, "lr": 1.001836128144118e-06}
{"train_loss": 0.05735253170132637, "global_step": 254156, "epoch": 2855, "lr": 1.001720660516181e-06}
{"train_loss": 0.027356985956430435, "global_step": 254157, "epoch": 2855, "lr": 1.001605199475475e-06}
{"train_loss": 0.0331890843808651, "global_step": 254158, "epoch": 2855, "lr": 1.0014897450219996e-06}
{"train_loss": 0.047940801829099655, "global_step": 254159, "epoch": 2855, "lr": 1.001374297155777e-06}
{"train_loss": 0.08357585966587067, "global_step": 254160, "epoch": 2855, "lr": 1.0012588558768243e-06}
{"train_loss": 0.041076578199863434, "global_step": 254161, "epoch": 2855, "lr": 1.0011434211851522e-06}
{"train_loss": 0.09625943005084991, "global_step": 254162, "epoch": 2855, "lr": 1.001027993080783e-06}
{"train_loss": 0.046812061220407486, "global_step": 254163, "epoch": 2855, "lr": 1.0009125715637224e-06}
{"train_loss": 0.019112365320324898, "global_step": 254164, "epoch": 2855, "lr": 1.0007971566340036e-06}
{"train_loss": 0.02929062955081463, "global_step": 254165, "epoch": 2855, "lr": 1.0006817482916208e-06}
{"train_loss": 0.04117496684193611, "global_step": 254166, "epoch": 2855, "lr": 1.0005663465366023e-06}
{"train_loss": 0.060467641800642014, "global_step": 254167, "epoch": 2855, "lr": 1.0004509513689587e-06}
{"train_loss": 0.021066972985863686, "global_step": 254168, "epoch": 2855, "lr": 1.0003355627887067e-06}
{"train_loss": 0.05272119119763374, "global_step": 254169, "epoch": 2855, "lr": 1.0002201807958634e-06}
{"train_loss": 0.011572531424462795, "global_step": 254170, "epoch": 2855, "lr": 1.0001048053904448e-06}
{"train_loss": 0.04434870928525925, "global_step": 254171, "epoch": 2855, "lr": 9.999894365724627e-07}
{"train_loss": 0.034406937658786774, "global_step": 254172, "epoch": 2855, "lr": 9.998740743419389e-07}
{"train_loss": 0.06663443893194199, "global_step": 254173, "epoch": 2855, "lr": 9.997587186988787e-07}
{"train_loss": 0.022622866556048393, "global_step": 254174, "epoch": 2855, "lr": 9.996433696433105e-07}
{"train_loss": 0.017527585849165916, "global_step": 254175, "epoch": 2855, "lr": 9.995280271752338e-07}
{"train_loss": 0.0873701423406601, "global_step": 254176, "epoch": 2855, "lr": 9.994126912946822e-07}
{"train_loss": 0.04865673929452896, "global_step": 254177, "epoch": 2855, "lr": 9.992973620016554e-07}
{"train_loss": 0.06674789637327194, "global_step": 254178, "epoch": 2855, "lr": 9.991820392961814e-07}
{"train_loss": 0.0184852946549654, "global_step": 254179, "epoch": 2855, "lr": 9.990667231782657e-07}
{"train_loss": 0.0635506734251976, "global_step": 254180, "epoch": 2855, "lr": 9.989514136479306e-07}
{"train_loss": 0.046300556510686874, "global_step": 254181, "epoch": 2855, "lr": 9.988361107051924e-07}
{"train_loss": 0.02976749651134014, "global_step": 254182, "epoch": 2855, "lr": 9.98720814350057e-07}
{"train_loss": 0.04824167977641808, "global_step": 254183, "epoch": 2855, "lr": 9.98605524582552e-07, "val_loss": 9.658792495727539, "train_action_mse_error": 28.59870147705078}
{"train_loss": 0.04890356585383415, "global_step": 254184, "epoch": 2856, "lr": 9.984902414026887e-07}
{"train_loss": 0.045574087649583817, "global_step": 254185, "epoch": 2856, "lr": 9.983749648104723e-07}
{"train_loss": 0.012016081251204014, "global_step": 254186, "epoch": 2856, "lr": 9.982596948059363e-07}
{"train_loss": 0.07491295039653778, "global_step": 254187, "epoch": 2856, "lr": 9.98144431389081e-07}
{"train_loss": 0.13299249112606049, "global_step": 254188, "epoch": 2856, "lr": 9.980291745599334e-07}
{"train_loss": 0.012689121067523956, "global_step": 254189, "epoch": 2856, "lr": 9.979139243184998e-07}
{"train_loss": 0.05251695215702057, "global_step": 254190, "epoch": 2856, "lr": 9.977986806647964e-07}
{"train_loss": 0.03345515951514244, "global_step": 254191, "epoch": 2856, "lr": 9.976834435988514e-07}
{"train_loss": 0.034277044236660004, "global_step": 254192, "epoch": 2856, "lr": 9.975682131206588e-07}
{"train_loss": 0.05741231143474579, "global_step": 254193, "epoch": 2856, "lr": 9.97452989230252e-07}
{"train_loss": 0.0427824966609478, "global_step": 254194, "epoch": 2856, "lr": 9.973377719276422e-07}
{"train_loss": 0.06781382113695145, "global_step": 254195, "epoch": 2856, "lr": 9.97222561212835e-07}
{"train_loss": 0.028342172503471375, "global_step": 254196, "epoch": 2856, "lr": 9.971073570858636e-07}
{"train_loss": 0.017596883699297905, "global_step": 254197, "epoch": 2856, "lr": 9.96992159546728e-07}
{"train_loss": 0.04470326751470566, "global_step": 254198, "epoch": 2856, "lr": 9.968769685954504e-07}
{"train_loss": 0.03192072734236717, "global_step": 254199, "epoch": 2856, "lr": 9.967617842320476e-07}
{"train_loss": 0.02586429938673973, "global_step": 254200, "epoch": 2856, "lr": 9.966466064565304e-07}
{"train_loss": 0.05946763977408409, "global_step": 254201, "epoch": 2856, "lr": 9.96531435268916e-07}
{"train_loss": 0.025311142206192017, "global_step": 254202, "epoch": 2856, "lr": 9.96416270669226e-07}
{"train_loss": 0.059349846094846725, "global_step": 254203, "epoch": 2856, "lr": 9.963011126574662e-07}
{"train_loss": 0.04514327272772789, "global_step": 254204, "epoch": 2856, "lr": 9.961859612336588e-07}
{"train_loss": 0.010403532534837723, "global_step": 254205, "epoch": 2856, "lr": 9.96070816397815e-07}
{"train_loss": 0.04455680772662163, "global_step": 254206, "epoch": 2856, "lr": 9.959556781499513e-07}
{"train_loss": 0.0243275985121727, "global_step": 254207, "epoch": 2856, "lr": 9.958405464900843e-07}
{"train_loss": 0.033560026437044144, "global_step": 254208, "epoch": 2856, "lr": 9.957254214182365e-07}
{"train_loss": 0.03348107635974884, "global_step": 254209, "epoch": 2856, "lr": 9.956103029344077e-07}
{"train_loss": 0.09343123435974121, "global_step": 254210, "epoch": 2856, "lr": 9.954951910386256e-07}
{"train_loss": 0.06385574489831924, "global_step": 254211, "epoch": 2856, "lr": 9.95380085730896e-07}
{"train_loss": 0.05941491574048996, "global_step": 254212, "epoch": 2856, "lr": 9.952649870112462e-07}
{"train_loss": 0.034819744527339935, "global_step": 254213, "epoch": 2856, "lr": 9.951498948796822e-07}
{"train_loss": 0.05552027374505997, "global_step": 254214, "epoch": 2856, "lr": 9.95034809336226e-07}
{"train_loss": 0.018165770918130875, "global_step": 254215, "epoch": 2856, "lr": 9.94919730380889e-07}
{"train_loss": 0.020245924592018127, "global_step": 254216, "epoch": 2856, "lr": 9.948046580136871e-07}
{"train_loss": 0.04447184503078461, "global_step": 254217, "epoch": 2856, "lr": 9.946895922346433e-07}
{"train_loss": 0.0350034274160862, "global_step": 254218, "epoch": 2856, "lr": 9.945745330437573e-07}
{"train_loss": 0.024405479431152344, "global_step": 254219, "epoch": 2856, "lr": 9.944594804410568e-07}
{"train_loss": 0.06115659326314926, "global_step": 254220, "epoch": 2856, "lr": 9.943444344265529e-07}
{"train_loss": 0.05746548995375633, "global_step": 254221, "epoch": 2856, "lr": 9.942293950002623e-07}
{"train_loss": 0.058634884655475616, "global_step": 254222, "epoch": 2856, "lr": 9.941143621622017e-07}
{"train_loss": 0.03010810911655426, "global_step": 254223, "epoch": 2856, "lr": 9.939993359123823e-07}
{"train_loss": 0.05267331004142761, "global_step": 254224, "epoch": 2856, "lr": 9.93884316250826e-07}
{"train_loss": 0.03356345370411873, "global_step": 254225, "epoch": 2856, "lr": 9.93769303177544e-07}
{"train_loss": 0.03689512610435486, "global_step": 254226, "epoch": 2856, "lr": 9.936542966925533e-07}
{"train_loss": 0.06481391936540604, "global_step": 254227, "epoch": 2856, "lr": 9.935392967958702e-07}
{"train_loss": 0.07288550585508347, "global_step": 254228, "epoch": 2856, "lr": 9.934243034875001e-07}
{"train_loss": 0.03393842652440071, "global_step": 254229, "epoch": 2856, "lr": 9.933093167674766e-07}
{"train_loss": 0.0092273885384202, "global_step": 254230, "epoch": 2856, "lr": 9.931943366357942e-07}
{"train_loss": 0.03494271636009216, "global_step": 254231, "epoch": 2856, "lr": 9.930793630924916e-07}
{"train_loss": 0.04529613256454468, "global_step": 254232, "epoch": 2856, "lr": 9.929643961375635e-07}
{"train_loss": 0.06883473694324493, "global_step": 254233, "epoch": 2856, "lr": 9.928494357710316e-07}
{"train_loss": 0.0529259517788887, "global_step": 254234, "epoch": 2856, "lr": 9.92734481992924e-07}
{"train_loss": 0.07385890930891037, "global_step": 254235, "epoch": 2856, "lr": 9.926195348032352e-07}
{"train_loss": 0.07628326863050461, "global_step": 254236, "epoch": 2856, "lr": 9.925045942019985e-07}
{"train_loss": 0.058895956724882126, "global_step": 254237, "epoch": 2856, "lr": 9.92389660189219e-07}
{"train_loss": 0.08148427307605743, "global_step": 254238, "epoch": 2856, "lr": 9.922747327649195e-07}
{"train_loss": 0.03980991616845131, "global_step": 254239, "epoch": 2856, "lr": 9.921598119291054e-07}
{"train_loss": 0.030661389231681824, "global_step": 254240, "epoch": 2856, "lr": 9.920448976818043e-07}
{"train_loss": 0.02099936455488205, "global_step": 254241, "epoch": 2856, "lr": 9.919299900230216e-07}
{"train_loss": 0.017306935042142868, "global_step": 254242, "epoch": 2856, "lr": 9.9181508895278e-07}
{"train_loss": 0.03042764775454998, "global_step": 254243, "epoch": 2856, "lr": 9.917001944710901e-07}
{"train_loss": 0.0600716732442379, "global_step": 254244, "epoch": 2856, "lr": 9.91585306577969e-07}
{"train_loss": 0.03146371990442276, "global_step": 254245, "epoch": 2856, "lr": 9.914704252734275e-07}
{"train_loss": 0.06473708897829056, "global_step": 254246, "epoch": 2856, "lr": 9.913555505574934e-07}
{"train_loss": 0.10009825229644775, "global_step": 254247, "epoch": 2856, "lr": 9.912406824301669e-07}
{"train_loss": 0.019424322992563248, "global_step": 254248, "epoch": 2856, "lr": 9.911258208914753e-07}
{"train_loss": 0.05632844567298889, "global_step": 254249, "epoch": 2856, "lr": 9.910109659414246e-07}
{"train_loss": 0.05406675115227699, "global_step": 254250, "epoch": 2856, "lr": 9.908961175800368e-07}
{"train_loss": 0.031938694417476654, "global_step": 254251, "epoch": 2856, "lr": 9.907812758073288e-07}
{"train_loss": 0.03946556895971298, "global_step": 254252, "epoch": 2856, "lr": 9.906664406233056e-07}
{"train_loss": 0.08375686407089233, "global_step": 254253, "epoch": 2856, "lr": 9.905516120280011e-07}
{"train_loss": 0.017949528992176056, "global_step": 254254, "epoch": 2856, "lr": 9.904367900214095e-07}
{"train_loss": 0.026144549250602722, "global_step": 254255, "epoch": 2856, "lr": 9.90321974603564e-07}
{"train_loss": 0.054534055292606354, "global_step": 254256, "epoch": 2856, "lr": 9.902071657744705e-07}
{"train_loss": 0.043917205184698105, "global_step": 254257, "epoch": 2856, "lr": 9.900923635341453e-07}
{"train_loss": 0.04985366761684418, "global_step": 254258, "epoch": 2856, "lr": 9.899775678825995e-07}
{"train_loss": 0.06467115879058838, "global_step": 254259, "epoch": 2856, "lr": 9.89862778819861e-07}
{"train_loss": 0.03923642635345459, "global_step": 254260, "epoch": 2856, "lr": 9.897479963459355e-07}
{"train_loss": 0.05198090896010399, "global_step": 254261, "epoch": 2856, "lr": 9.89633220460845e-07}
{"train_loss": 0.05300282686948776, "global_step": 254262, "epoch": 2856, "lr": 9.895184511646004e-07}
{"train_loss": 0.057426560670137405, "global_step": 254263, "epoch": 2856, "lr": 9.894036884572134e-07}
{"train_loss": 0.02006874792277813, "global_step": 254264, "epoch": 2856, "lr": 9.892889323387055e-07}
{"train_loss": 0.028985880315303802, "global_step": 254265, "epoch": 2856, "lr": 9.891741828090883e-07}
{"train_loss": 0.10272463411092758, "global_step": 254266, "epoch": 2856, "lr": 9.890594398683838e-07}
{"train_loss": 0.0480976477265358, "global_step": 254267, "epoch": 2856, "lr": 9.889447035166033e-07}
{"train_loss": 0.07377509027719498, "global_step": 254268, "epoch": 2856, "lr": 9.888299737537521e-07}
{"train_loss": 0.02589072473347187, "global_step": 254269, "epoch": 2856, "lr": 9.887152505798635e-07}
{"train_loss": 0.058875154703855515, "global_step": 254270, "epoch": 2856, "lr": 9.886005339949434e-07}
{"train_loss": 0.03890587389469147, "global_step": 254271, "epoch": 2856, "lr": 9.88485823999008e-07}
{"train_loss": 0.04673263633602791, "global_step": 254272, "epoch": 2856, "lr": 9.883711205920743e-07, "val_loss": 9.582892417907715}
{"train_loss": 0.05315880477428436, "global_step": 254273, "epoch": 2857, "lr": 9.88256423774153e-07}
{"train_loss": 0.044303055852651596, "global_step": 254274, "epoch": 2857, "lr": 9.881417335452669e-07}
{"train_loss": 0.03746366128325462, "global_step": 254275, "epoch": 2857, "lr": 9.880270499054266e-07}
{"train_loss": 0.016977444291114807, "global_step": 254276, "epoch": 2857, "lr": 9.879123728546546e-07}
{"train_loss": 0.03027677908539772, "global_step": 254277, "epoch": 2857, "lr": 9.877977023929507e-07}
{"train_loss": 0.08596907556056976, "global_step": 254278, "epoch": 2857, "lr": 9.876830385203484e-07}
{"train_loss": 0.0905148908495903, "global_step": 254279, "epoch": 2857, "lr": 9.875683812368474e-07}
{"train_loss": 0.05542308837175369, "global_step": 254280, "epoch": 2857, "lr": 9.874537305424758e-07}
{"train_loss": 0.047213420271873474, "global_step": 254281, "epoch": 2857, "lr": 9.87339086437239e-07}
{"train_loss": 0.07209552079439163, "global_step": 254282, "epoch": 2857, "lr": 9.872244489211646e-07}
{"train_loss": 0.03552202507853508, "global_step": 254283, "epoch": 2857, "lr": 9.871098179942528e-07}
{"train_loss": 0.04627801477909088, "global_step": 254284, "epoch": 2857, "lr": 9.869951936565313e-07}
{"train_loss": 0.023709718137979507, "global_step": 254285, "epoch": 2857, "lr": 9.868805759080057e-07}
{"train_loss": 0.03695671260356903, "global_step": 254286, "epoch": 2857, "lr": 9.867659647486982e-07}
{"train_loss": 0.09030169993638992, "global_step": 254287, "epoch": 2857, "lr": 9.866513601786255e-07}
{"train_loss": 0.03207913413643837, "global_step": 254288, "epoch": 2857, "lr": 9.865367621977982e-07}
{"train_loss": 0.01055015716701746, "global_step": 254289, "epoch": 2857, "lr": 9.864221708062394e-07}
{"train_loss": 0.047341443598270416, "global_step": 254290, "epoch": 2857, "lr": 9.863075860039484e-07}
{"train_loss": 0.07022500783205032, "global_step": 254291, "epoch": 2857, "lr": 9.861930077909586e-07}
{"train_loss": 0.05016922950744629, "global_step": 254292, "epoch": 2857, "lr": 9.86078436167276e-07}
{"train_loss": 0.05435042455792427, "global_step": 254293, "epoch": 2857, "lr": 9.859638711329167e-07}
{"train_loss": 0.02086842805147171, "global_step": 254294, "epoch": 2857, "lr": 9.858493126878976e-07}
{"train_loss": 0.041730474680662155, "global_step": 254295, "epoch": 2857, "lr": 9.857347608322353e-07}
{"train_loss": 0.038556601852178574, "global_step": 254296, "epoch": 2857, "lr": 9.856202155659411e-07}
{"train_loss": 0.08849713951349258, "global_step": 254297, "epoch": 2857, "lr": 9.855056768890369e-07}
{"train_loss": 0.04125475883483887, "global_step": 254298, "epoch": 2857, "lr": 9.853911448015284e-07}
{"train_loss": 0.04764685779809952, "global_step": 254299, "epoch": 2857, "lr": 9.85276619303438e-07}
{"train_loss": 0.02539479173719883, "global_step": 254300, "epoch": 2857, "lr": 9.851621003947875e-07}
{"train_loss": 0.030870093032717705, "global_step": 254301, "epoch": 2857, "lr": 9.850475880755716e-07}
{"train_loss": 0.02450299635529518, "global_step": 254302, "epoch": 2857, "lr": 9.849330823458292e-07}
{"train_loss": 0.04110751301050186, "global_step": 254303, "epoch": 2857, "lr": 9.848185832055601e-07}
{"train_loss": 0.05562019720673561, "global_step": 254304, "epoch": 2857, "lr": 9.847040906547866e-07}
{"train_loss": 0.04899151995778084, "global_step": 254305, "epoch": 2857, "lr": 9.8458960469352e-07}
{"train_loss": 0.03293699398636818, "global_step": 254306, "epoch": 2857, "lr": 9.84475125321782e-07}
{"train_loss": 0.020471569150686264, "global_step": 254307, "epoch": 2857, "lr": 9.843606525395787e-07}
{"train_loss": 0.06893651187419891, "global_step": 254308, "epoch": 2857, "lr": 9.842461863469376e-07}
{"train_loss": 0.05404330790042877, "global_step": 254309, "epoch": 2857, "lr": 9.841317267438587e-07}
{"train_loss": 0.009196820668876171, "global_step": 254310, "epoch": 2857, "lr": 9.8401727373037e-07}
{"train_loss": 0.04612629860639572, "global_step": 254311, "epoch": 2857, "lr": 9.839028273064821e-07}
{"train_loss": 0.03170171007514, "global_step": 254312, "epoch": 2857, "lr": 9.83788387472212e-07}
{"train_loss": 0.05200676620006561, "global_step": 254313, "epoch": 2857, "lr": 9.836739542275763e-07}
{"train_loss": 0.045261647552251816, "global_step": 254314, "epoch": 2857, "lr": 9.835595275725863e-07}
{"train_loss": 0.04746352881193161, "global_step": 254315, "epoch": 2857, "lr": 9.834451075072582e-07}
{"train_loss": 0.0425189733505249, "global_step": 254316, "epoch": 2857, "lr": 9.833306940316089e-07}
{"train_loss": 0.04823276028037071, "global_step": 254317, "epoch": 2857, "lr": 9.832162871456497e-07}
{"train_loss": 0.08383285254240036, "global_step": 254318, "epoch": 2857, "lr": 9.83101886849408e-07}
{"train_loss": 0.026363776996731758, "global_step": 254319, "epoch": 2857, "lr": 9.829874931428839e-07}
{"train_loss": 0.10137321799993515, "global_step": 254320, "epoch": 2857, "lr": 9.828731060261053e-07}
{"train_loss": 0.030708936974406242, "global_step": 254321, "epoch": 2857, "lr": 9.827587254990723e-07}
{"train_loss": 0.02627047710120678, "global_step": 254322, "epoch": 2857, "lr": 9.826443515618179e-07}
{"train_loss": 0.04374893009662628, "global_step": 254323, "epoch": 2857, "lr": 9.825299842143477e-07}
{"train_loss": 0.045935988426208496, "global_step": 254324, "epoch": 2857, "lr": 9.824156234566784e-07}
{"train_loss": 0.06078370288014412, "global_step": 254325, "epoch": 2857, "lr": 9.823012692888267e-07}
{"train_loss": 0.09967764467000961, "global_step": 254326, "epoch": 2857, "lr": 9.821869217108038e-07}
{"train_loss": 0.02458065189421177, "global_step": 254327, "epoch": 2857, "lr": 9.820725807226372e-07}
{"train_loss": 0.027611542493104935, "global_step": 254328, "epoch": 2857, "lr": 9.81958246324327e-07}
{"train_loss": 0.015438825823366642, "global_step": 254329, "epoch": 2857, "lr": 9.818439185158957e-07}
{"train_loss": 0.0641496479511261, "global_step": 254330, "epoch": 2857, "lr": 9.817295972973538e-07}
{"train_loss": 0.0236385315656662, "global_step": 254331, "epoch": 2857, "lr": 9.816152826687298e-07}
{"train_loss": 0.041667383164167404, "global_step": 254332, "epoch": 2857, "lr": 9.81500974630023e-07}
{"train_loss": 0.06506717205047607, "global_step": 254333, "epoch": 2857, "lr": 9.813866731812615e-07}
{"train_loss": 0.03966265544295311, "global_step": 254334, "epoch": 2857, "lr": 9.812723783224508e-07}
{"train_loss": 0.03952313959598541, "global_step": 254335, "epoch": 2857, "lr": 9.811580900536132e-07}
{"train_loss": 0.017171205952763557, "global_step": 254336, "epoch": 2857, "lr": 9.810438083747597e-07}
{"train_loss": 0.04687075316905975, "global_step": 254337, "epoch": 2857, "lr": 9.809295332859125e-07}
{"train_loss": 0.04807006195187569, "global_step": 254338, "epoch": 2857, "lr": 9.808152647870773e-07}
{"train_loss": 0.038244783878326416, "global_step": 254339, "epoch": 2857, "lr": 9.807010028782705e-07}
{"train_loss": 0.05143846198916435, "global_step": 254340, "epoch": 2857, "lr": 9.805867475595198e-07}
{"train_loss": 0.0644042119383812, "global_step": 254341, "epoch": 2857, "lr": 9.804724988308256e-07}
{"train_loss": 0.08306533843278885, "global_step": 254342, "epoch": 2857, "lr": 9.803582566922099e-07}
{"train_loss": 0.029243210330605507, "global_step": 254343, "epoch": 2857, "lr": 9.802440211436893e-07}
{"train_loss": 0.02110777422785759, "global_step": 254344, "epoch": 2857, "lr": 9.801297921852747e-07}
{"train_loss": 0.028480688109993935, "global_step": 254345, "epoch": 2857, "lr": 9.800155698169834e-07}
{"train_loss": 0.025447465479373932, "global_step": 254346, "epoch": 2857, "lr": 9.799013540388369e-07}
{"train_loss": 0.03556055948138237, "global_step": 254347, "epoch": 2857, "lr": 9.79787144850841e-07}
{"train_loss": 0.03073311783373356, "global_step": 254348, "epoch": 2857, "lr": 9.796729422530183e-07}
{"train_loss": 0.002836857922375202, "global_step": 254349, "epoch": 2857, "lr": 9.795587462453793e-07}
{"train_loss": 0.09026730060577393, "global_step": 254350, "epoch": 2857, "lr": 9.79444556827941e-07}
{"train_loss": 0.04351072758436203, "global_step": 254351, "epoch": 2857, "lr": 9.793303740007142e-07}
{"train_loss": 0.030752569437026978, "global_step": 254352, "epoch": 2857, "lr": 9.792161977637271e-07}
{"train_loss": 0.06587859988212585, "global_step": 254353, "epoch": 2857, "lr": 9.79102028116985e-07}
{"train_loss": 0.06739838421344757, "global_step": 254354, "epoch": 2857, "lr": 9.789878650605044e-07}
{"train_loss": 0.09255525469779968, "global_step": 254355, "epoch": 2857, "lr": 9.788737085942967e-07}
{"train_loss": 0.08601649850606918, "global_step": 254356, "epoch": 2857, "lr": 9.787595587183895e-07}
{"train_loss": 0.02922006882727146, "global_step": 254357, "epoch": 2857, "lr": 9.78645415432783e-07}
{"train_loss": 0.05711749196052551, "global_step": 254358, "epoch": 2857, "lr": 9.7853127873751e-07}
{"train_loss": 0.04128246009349823, "global_step": 254359, "epoch": 2857, "lr": 9.784171486325654e-07}
{"train_loss": 0.016588499769568443, "global_step": 254360, "epoch": 2857, "lr": 9.78303025117977e-07}
{"train_loss": 0.04566355970468414, "global_step": 254361, "epoch": 2857, "lr": 9.781889081937668e-07, "val_loss": 9.576231956481934}
{"train_loss": 0.03989064693450928, "global_step": 254362, "epoch": 2858, "lr": 9.780747978599292e-07}
{"train_loss": 0.03594400733709335, "global_step": 254363, "epoch": 2858, "lr": 9.779606941165032e-07}
{"train_loss": 0.038343753665685654, "global_step": 254364, "epoch": 2858, "lr": 9.778465969634888e-07}
{"train_loss": 0.04813206195831299, "global_step": 254365, "epoch": 2858, "lr": 9.777325064009023e-07}
{"train_loss": 0.02631012164056301, "global_step": 254366, "epoch": 2858, "lr": 9.776184224287665e-07}
{"train_loss": 0.06508588045835495, "global_step": 254367, "epoch": 2858, "lr": 9.775043450470922e-07}
{"train_loss": 0.03796663135290146, "global_step": 254368, "epoch": 2858, "lr": 9.773902742558904e-07}
{"train_loss": 0.07898161560297012, "global_step": 254369, "epoch": 2858, "lr": 9.772762100551891e-07}
{"train_loss": 0.00849783606827259, "global_step": 254370, "epoch": 2858, "lr": 9.771621524449882e-07}
{"train_loss": 0.050744444131851196, "global_step": 254371, "epoch": 2858, "lr": 9.770481014253152e-07}
{"train_loss": 0.07717803865671158, "global_step": 254372, "epoch": 2858, "lr": 9.769340569961705e-07}
{"train_loss": 0.03135814145207405, "global_step": 254373, "epoch": 2858, "lr": 9.768200191575926e-07}
{"train_loss": 0.030947459861636162, "global_step": 254374, "epoch": 2858, "lr": 9.767059879095763e-07}
{"train_loss": 0.03759433329105377, "global_step": 254375, "epoch": 2858, "lr": 9.765919632521493e-07}
{"train_loss": 0.03356992080807686, "global_step": 254376, "epoch": 2858, "lr": 9.764779451853169e-07}
{"train_loss": 0.05241882801055908, "global_step": 254377, "epoch": 2858, "lr": 9.763639337091013e-07}
{"train_loss": 0.03496461361646652, "global_step": 254378, "epoch": 2858, "lr": 9.762499288235194e-07}
{"train_loss": 0.014456107281148434, "global_step": 254379, "epoch": 2858, "lr": 9.761359305285767e-07}
{"train_loss": 0.03637637570500374, "global_step": 254380, "epoch": 2858, "lr": 9.760219388243007e-07}
{"train_loss": 0.037128329277038574, "global_step": 254381, "epoch": 2858, "lr": 9.759079537106975e-07}
{"train_loss": 0.04442894458770752, "global_step": 254382, "epoch": 2858, "lr": 9.757939751877888e-07}
{"train_loss": 0.11093413084745407, "global_step": 254383, "epoch": 2858, "lr": 9.756800032555857e-07}
{"train_loss": 0.02905815839767456, "global_step": 254384, "epoch": 2858, "lr": 9.755660379141052e-07}
{"train_loss": 0.012993531301617622, "global_step": 254385, "epoch": 2858, "lr": 9.75452079163358e-07}
{"train_loss": 0.022852113470435143, "global_step": 254386, "epoch": 2858, "lr": 9.753381270033724e-07}
{"train_loss": 0.026851771399378777, "global_step": 254387, "epoch": 2858, "lr": 9.75224181434148e-07}
{"train_loss": 0.04249310865998268, "global_step": 254388, "epoch": 2858, "lr": 9.751102424557123e-07}
{"train_loss": 0.05797307938337326, "global_step": 254389, "epoch": 2858, "lr": 9.749963100680658e-07}
{"train_loss": 0.030603298917412758, "global_step": 254390, "epoch": 2858, "lr": 9.748823842712473e-07}
{"train_loss": 0.061885766685009, "global_step": 254391, "epoch": 2858, "lr": 9.747684650652455e-07}
{"train_loss": 0.05989920347929001, "global_step": 254392, "epoch": 2858, "lr": 9.746545524500994e-07}
{"train_loss": 0.06902257353067398, "global_step": 254393, "epoch": 2858, "lr": 9.745406464258033e-07}
{"train_loss": 0.0866631492972374, "global_step": 254394, "epoch": 2858, "lr": 9.744267469923908e-07}
{"train_loss": 0.05018142610788345, "global_step": 254395, "epoch": 2858, "lr": 9.74312854149867e-07}
{"train_loss": 0.0826406255364418, "global_step": 254396, "epoch": 2858, "lr": 9.741989678982434e-07}
{"train_loss": 0.028363589197397232, "global_step": 254397, "epoch": 2858, "lr": 9.740850882375475e-07}
{"train_loss": 0.04661819711327553, "global_step": 254398, "epoch": 2858, "lr": 9.73971215167785e-07}
{"train_loss": 0.03862542659044266, "global_step": 254399, "epoch": 2858, "lr": 9.738573486889778e-07}
{"train_loss": 0.018560651689767838, "global_step": 254400, "epoch": 2858, "lr": 9.737434888011376e-07}
{"train_loss": 0.022163983434438705, "global_step": 254401, "epoch": 2858, "lr": 9.736296355042806e-07}
{"train_loss": 0.030328288674354553, "global_step": 254402, "epoch": 2858, "lr": 9.73515788798418e-07}
{"train_loss": 0.05157176032662392, "global_step": 254403, "epoch": 2858, "lr": 9.73401948683572e-07}
{"train_loss": 0.04218528792262077, "global_step": 254404, "epoch": 2858, "lr": 9.732881151597538e-07}
{"train_loss": 0.04869568720459938, "global_step": 254405, "epoch": 2858, "lr": 9.7317428822698e-07}
{"train_loss": 0.07633082568645477, "global_step": 254406, "epoch": 2858, "lr": 9.730604678852618e-07}
{"train_loss": 0.04762503504753113, "global_step": 254407, "epoch": 2858, "lr": 9.729466541346211e-07}
{"train_loss": 0.03417309373617172, "global_step": 254408, "epoch": 2858, "lr": 9.728328469750691e-07}
{"train_loss": 0.07220743596553802, "global_step": 254409, "epoch": 2858, "lr": 9.727190464066227e-07}
{"train_loss": 0.009072900749742985, "global_step": 254410, "epoch": 2858, "lr": 9.726052524292983e-07}
{"train_loss": 0.03488261252641678, "global_step": 254411, "epoch": 2858, "lr": 9.724914650431073e-07}
{"train_loss": 0.03494548052549362, "global_step": 254412, "epoch": 2858, "lr": 9.723776842480658e-07}
{"train_loss": 0.05268615484237671, "global_step": 254413, "epoch": 2858, "lr": 9.722639100441966e-07}
{"train_loss": 0.04325119033455849, "global_step": 254414, "epoch": 2858, "lr": 9.721501424314993e-07}
{"train_loss": 0.048562128096818924, "global_step": 254415, "epoch": 2858, "lr": 9.720363814100075e-07}
{"train_loss": 0.03459396958351135, "global_step": 254416, "epoch": 2858, "lr": 9.719226269797267e-07}
{"train_loss": 0.018394172191619873, "global_step": 254417, "epoch": 2858, "lr": 9.718088791406732e-07}
{"train_loss": 0.07299790531396866, "global_step": 254418, "epoch": 2858, "lr": 9.71695137892864e-07}
{"train_loss": 0.06457585096359253, "global_step": 254419, "epoch": 2858, "lr": 9.715814032363045e-07}
{"train_loss": 0.045331794768571854, "global_step": 254420, "epoch": 2858, "lr": 9.714676751710283e-07}
{"train_loss": 0.07074162364006042, "global_step": 254421, "epoch": 2858, "lr": 9.713539536970296e-07}
{"train_loss": 0.070346400141716, "global_step": 254422, "epoch": 2858, "lr": 9.712402388143472e-07}
{"train_loss": 0.046827394515275955, "global_step": 254423, "epoch": 2858, "lr": 9.711265305229755e-07}
{"train_loss": 0.04180316999554634, "global_step": 254424, "epoch": 2858, "lr": 9.710128288229425e-07}
{"train_loss": 0.035409197211265564, "global_step": 254425, "epoch": 2858, "lr": 9.708991337142536e-07}
{"train_loss": 0.0667620301246643, "global_step": 254426, "epoch": 2858, "lr": 9.707854451969368e-07}
{"train_loss": 0.03114280104637146, "global_step": 254427, "epoch": 2858, "lr": 9.706717632709917e-07}
{"train_loss": 0.040765244513750076, "global_step": 254428, "epoch": 2858, "lr": 9.705580879364519e-07}
{"train_loss": 0.06655244529247284, "global_step": 254429, "epoch": 2858, "lr": 9.70444419193317e-07}
{"train_loss": 0.08597516268491745, "global_step": 254430, "epoch": 2858, "lr": 9.703307570416098e-07}
{"train_loss": 0.09040829539299011, "global_step": 254431, "epoch": 2858, "lr": 9.70217101481341e-07}
{"train_loss": 0.009586534462869167, "global_step": 254432, "epoch": 2858, "lr": 9.70103452512533e-07}
{"train_loss": 0.029762623831629753, "global_step": 254433, "epoch": 2858, "lr": 9.699898101351967e-07}
{"train_loss": 0.053535785526037216, "global_step": 254434, "epoch": 2858, "lr": 9.698761743493435e-07}
{"train_loss": 0.0462917685508728, "global_step": 254435, "epoch": 2858, "lr": 9.697625451550007e-07}
{"train_loss": 0.004029144532978535, "global_step": 254436, "epoch": 2858, "lr": 9.696489225521687e-07}
{"train_loss": 0.0476832315325737, "global_step": 254437, "epoch": 2858, "lr": 9.695353065408752e-07}
{"train_loss": 0.044262710958719254, "global_step": 254438, "epoch": 2858, "lr": 9.694216971211255e-07}
{"train_loss": 0.041754018515348434, "global_step": 254439, "epoch": 2858, "lr": 9.693080942929422e-07}
{"train_loss": 0.01904171332716942, "global_step": 254440, "epoch": 2858, "lr": 9.69194498056336e-07}
{"train_loss": 0.031113050878047943, "global_step": 254441, "epoch": 2858, "lr": 9.69080908411324e-07}
{"train_loss": 0.03321002796292305, "global_step": 254442, "epoch": 2858, "lr": 9.689673253579224e-07}
{"train_loss": 0.03687582537531853, "global_step": 254443, "epoch": 2858, "lr": 9.68853748896148e-07}
{"train_loss": 0.03164670988917351, "global_step": 254444, "epoch": 2858, "lr": 9.687401790260066e-07}
{"train_loss": 0.02610556222498417, "global_step": 254445, "epoch": 2858, "lr": 9.686266157475254e-07}
{"train_loss": 0.03567136824131012, "global_step": 254446, "epoch": 2858, "lr": 9.685130590607161e-07}
{"train_loss": 0.03639695420861244, "global_step": 254447, "epoch": 2858, "lr": 9.683995089655894e-07}
{"train_loss": 0.1074836328625679, "global_step": 254448, "epoch": 2858, "lr": 9.682859654621623e-07}
{"train_loss": 0.06752517819404602, "global_step": 254449, "epoch": 2858, "lr": 9.681724285504567e-07}
{"train_loss": 0.04529777786621217, "global_step": 254450, "epoch": 2858, "lr": 9.680588982304784e-07, "val_loss": 9.653538703918457}
{"train_loss": 0.028760790824890137, "global_step": 254451, "epoch": 2859, "lr": 9.679453745022549e-07}
{"train_loss": 0.01962895318865776, "global_step": 254452, "epoch": 2859, "lr": 9.678318573657807e-07}
{"train_loss": 0.06847236305475235, "global_step": 254453, "epoch": 2859, "lr": 9.677183468210948e-07}
{"train_loss": 0.0392746701836586, "global_step": 254454, "epoch": 2859, "lr": 9.676048428682027e-07}
{"train_loss": 0.01630743220448494, "global_step": 254455, "epoch": 2859, "lr": 9.6749134550711e-07}
{"train_loss": 0.08481528609991074, "global_step": 254456, "epoch": 2859, "lr": 9.673778547378442e-07}
{"train_loss": 0.03683719038963318, "global_step": 254457, "epoch": 2859, "lr": 9.672643705604167e-07}
{"train_loss": 0.0597073957324028, "global_step": 254458, "epoch": 2859, "lr": 9.67150892974844e-07}
{"train_loss": 0.06001129001379013, "global_step": 254459, "epoch": 2859, "lr": 9.670374219811372e-07}
{"train_loss": 0.013966832309961319, "global_step": 254460, "epoch": 2859, "lr": 9.669239575793189e-07}
{"train_loss": 0.049202173948287964, "global_step": 254461, "epoch": 2859, "lr": 9.668104997693938e-07}
{"train_loss": 0.019762584939599037, "global_step": 254462, "epoch": 2859, "lr": 9.666970485513904e-07}
{"train_loss": 0.08205112814903259, "global_step": 254463, "epoch": 2859, "lr": 9.665836039253084e-07}
{"train_loss": 0.010390999726951122, "global_step": 254464, "epoch": 2859, "lr": 9.664701658911812e-07}
{"train_loss": 0.034784168004989624, "global_step": 254465, "epoch": 2859, "lr": 9.663567344490087e-07}
{"train_loss": 0.023936470970511436, "global_step": 254466, "epoch": 2859, "lr": 9.662433095988132e-07}
{"train_loss": 0.020603975281119347, "global_step": 254467, "epoch": 2859, "lr": 9.661298913406058e-07}
{"train_loss": 0.023974105715751648, "global_step": 254468, "epoch": 2859, "lr": 9.66016479674403e-07}
{"train_loss": 0.02425975725054741, "global_step": 254469, "epoch": 2859, "lr": 9.659030746002328e-07}
{"train_loss": 0.03959592431783676, "global_step": 254470, "epoch": 2859, "lr": 9.657896761180895e-07}
{"train_loss": 0.050574611872434616, "global_step": 254471, "epoch": 2859, "lr": 9.65676284228001e-07}
{"train_loss": 0.03910199925303459, "global_step": 254472, "epoch": 2859, "lr": 9.65562898929978e-07}
{"train_loss": 0.033743880689144135, "global_step": 254473, "epoch": 2859, "lr": 9.654495202240431e-07}
{"train_loss": 0.04370899870991707, "global_step": 254474, "epoch": 2859, "lr": 9.653361481102019e-07}
{"train_loss": 0.04074656963348389, "global_step": 254475, "epoch": 2859, "lr": 9.652227825884708e-07}
{"train_loss": 0.06036161631345749, "global_step": 254476, "epoch": 2859, "lr": 9.65109423658872e-07}
{"train_loss": 0.03343444690108299, "global_step": 254477, "epoch": 2859, "lr": 9.649960713214167e-07}
{"train_loss": 0.06485065817832947, "global_step": 254478, "epoch": 2859, "lr": 9.648827255761162e-07}
{"train_loss": 0.04018926993012428, "global_step": 254479, "epoch": 2859, "lr": 9.647693864229923e-07}
{"train_loss": 0.07674551755189896, "global_step": 254480, "epoch": 2859, "lr": 9.646560538620563e-07}
{"train_loss": 0.03356091305613518, "global_step": 254481, "epoch": 2859, "lr": 9.645427278933306e-07}
{"train_loss": 0.04692969098687172, "global_step": 254482, "epoch": 2859, "lr": 9.644294085168148e-07}
{"train_loss": 0.043096330016851425, "global_step": 254483, "epoch": 2859, "lr": 9.643160957325425e-07}
{"train_loss": 0.04323849454522133, "global_step": 254484, "epoch": 2859, "lr": 9.642027895405136e-07}
{"train_loss": 0.06074720621109009, "global_step": 254485, "epoch": 2859, "lr": 9.64089489940756e-07}
{"train_loss": 0.03953614830970764, "global_step": 254486, "epoch": 2859, "lr": 9.639761969332694e-07}
{"train_loss": 0.04560095816850662, "global_step": 254487, "epoch": 2859, "lr": 9.638629105180874e-07}
{"train_loss": 0.05658312886953354, "global_step": 254488, "epoch": 2859, "lr": 9.637496306952154e-07}
{"train_loss": 0.027648605406284332, "global_step": 254489, "epoch": 2859, "lr": 9.6363635746467e-07}
{"train_loss": 0.024634037166833878, "global_step": 254490, "epoch": 2859, "lr": 9.63523090826457e-07}
{"train_loss": 0.0328395776450634, "global_step": 254491, "epoch": 2859, "lr": 9.634098307806095e-07}
{"train_loss": 0.04212883859872818, "global_step": 254492, "epoch": 2859, "lr": 9.63296577327133e-07}
{"train_loss": 0.03905452415347099, "global_step": 254493, "epoch": 2859, "lr": 9.631833304660443e-07}
{"train_loss": 0.045410867780447006, "global_step": 254494, "epoch": 2859, "lr": 9.6307009019736e-07}
{"train_loss": 0.014394315890967846, "global_step": 254495, "epoch": 2859, "lr": 9.629568565210855e-07}
{"train_loss": 0.03350455313920975, "global_step": 254496, "epoch": 2859, "lr": 9.628436294372489e-07}
{"train_loss": 0.02461511828005314, "global_step": 254497, "epoch": 2859, "lr": 9.627304089458555e-07}
{"train_loss": 0.020046481862664223, "global_step": 254498, "epoch": 2859, "lr": 9.626171950469332e-07}
{"train_loss": 0.008610301651060581, "global_step": 254499, "epoch": 2859, "lr": 9.625039877404818e-07}
{"train_loss": 0.0655180960893631, "global_step": 254500, "epoch": 2859, "lr": 9.62390787026529e-07}
{"train_loss": 0.11061092466115952, "global_step": 254501, "epoch": 2859, "lr": 9.622775929050809e-07}
{"train_loss": 0.016166927292943, "global_step": 254502, "epoch": 2859, "lr": 9.621644053761646e-07}
{"train_loss": 0.03180498629808426, "global_step": 254503, "epoch": 2859, "lr": 9.62051224439775e-07}
{"train_loss": 0.025047602131962776, "global_step": 254504, "epoch": 2859, "lr": 9.619380500959451e-07}
{"train_loss": 0.045643825083971024, "global_step": 254505, "epoch": 2859, "lr": 9.61824882344692e-07}
{"train_loss": 0.04979721084237099, "global_step": 254506, "epoch": 2859, "lr": 9.617117211860149e-07}
{"train_loss": 0.007558940444141626, "global_step": 254507, "epoch": 2859, "lr": 9.615985666199478e-07}
{"train_loss": 0.03988495469093323, "global_step": 254508, "epoch": 2859, "lr": 9.61485418646485e-07}
{"train_loss": 0.025126783177256584, "global_step": 254509, "epoch": 2859, "lr": 9.613722772656597e-07}
{"train_loss": 0.035501182079315186, "global_step": 254510, "epoch": 2859, "lr": 9.612591424774776e-07}
{"train_loss": 0.024420827627182007, "global_step": 254511, "epoch": 2859, "lr": 9.611460142819606e-07}
{"train_loss": 0.07495247572660446, "global_step": 254512, "epoch": 2859, "lr": 9.610328926791145e-07}
{"train_loss": 0.044849470257759094, "global_step": 254513, "epoch": 2859, "lr": 9.609197776689617e-07}
{"train_loss": 0.06291576474905014, "global_step": 254514, "epoch": 2859, "lr": 9.608066692515128e-07}
{"train_loss": 0.03179582580924034, "global_step": 254515, "epoch": 2859, "lr": 9.606935674267958e-07}
{"train_loss": 0.03021080605685711, "global_step": 254516, "epoch": 2859, "lr": 9.605804721948054e-07}
{"train_loss": 0.024178633466362953, "global_step": 254517, "epoch": 2859, "lr": 9.604673835555743e-07}
{"train_loss": 0.005018456373363733, "global_step": 254518, "epoch": 2859, "lr": 9.603543015091033e-07}
{"train_loss": 0.07754795253276825, "global_step": 254519, "epoch": 2859, "lr": 9.602412260554194e-07}
{"train_loss": 0.03338305652141571, "global_step": 254520, "epoch": 2859, "lr": 9.601281571945286e-07}
{"train_loss": 0.053010500967502594, "global_step": 254521, "epoch": 2859, "lr": 9.600150949264586e-07}
{"train_loss": 0.04346301779150963, "global_step": 254522, "epoch": 2859, "lr": 9.599020392512092e-07}
{"train_loss": 0.07090508937835693, "global_step": 254523, "epoch": 2859, "lr": 9.597889901688084e-07}
{"train_loss": 0.030647678300738335, "global_step": 254524, "epoch": 2859, "lr": 9.596759476792673e-07}
{"train_loss": 0.037385452538728714, "global_step": 254525, "epoch": 2859, "lr": 9.595629117825965e-07}
{"train_loss": 0.028052950277924538, "global_step": 254526, "epoch": 2859, "lr": 9.59449882478819e-07}
{"train_loss": 0.03334325551986694, "global_step": 254527, "epoch": 2859, "lr": 9.593368597679398e-07}
{"train_loss": 0.042335160076618195, "global_step": 254528, "epoch": 2859, "lr": 9.592238436499812e-07}
{"train_loss": 0.03850006312131882, "global_step": 254529, "epoch": 2859, "lr": 9.591108341249655e-07}
{"train_loss": 0.04332607984542847, "global_step": 254530, "epoch": 2859, "lr": 9.58997831192887e-07}
{"train_loss": 0.01948816329240799, "global_step": 254531, "epoch": 2859, "lr": 9.58884834853785e-07}
{"train_loss": 0.05110783129930496, "global_step": 254532, "epoch": 2859, "lr": 9.587718451076588e-07}
{"train_loss": 0.06420844048261642, "global_step": 254533, "epoch": 2859, "lr": 9.586588619545256e-07}
{"train_loss": 0.04263888671994209, "global_step": 254534, "epoch": 2859, "lr": 9.585458853944075e-07}
{"train_loss": 0.06084596738219261, "global_step": 254535, "epoch": 2859, "lr": 9.584329154273098e-07}
{"train_loss": 0.056408658623695374, "global_step": 254536, "epoch": 2859, "lr": 9.583199520532548e-07}
{"train_loss": 0.06090358644723892, "global_step": 254537, "epoch": 2859, "lr": 9.582069952722538e-07}
{"train_loss": 0.04955962300300598, "global_step": 254538, "epoch": 2859, "lr": 9.58094045084329e-07}
{"train_loss": 0.04079729917176654, "global_step": 254539, "epoch": 2859, "lr": 9.579811014894857e-07, "val_loss": 9.566019058227539}
{"train_loss": 0.05852915346622467, "global_step": 254540, "epoch": 2860, "lr": 9.578681644877462e-07}
{"train_loss": 0.04338044673204422, "global_step": 254541, "epoch": 2860, "lr": 9.577552340791273e-07}
{"train_loss": 0.0557112954556942, "global_step": 254542, "epoch": 2860, "lr": 9.576423102636345e-07}
{"train_loss": 0.021121105179190636, "global_step": 254543, "epoch": 2860, "lr": 9.575293930412953e-07}
{"train_loss": 0.01125184353441, "global_step": 254544, "epoch": 2860, "lr": 9.574164824121156e-07}
{"train_loss": 0.03821459412574768, "global_step": 254545, "epoch": 2860, "lr": 9.57303578376112e-07}
{"train_loss": 0.09548737853765488, "global_step": 254546, "epoch": 2860, "lr": 9.57190680933301e-07}
{"train_loss": 0.042013321071863174, "global_step": 254547, "epoch": 2860, "lr": 9.570777900836992e-07}
{"train_loss": 0.04299909994006157, "global_step": 254548, "epoch": 2860, "lr": 9.569649058273233e-07}
{"train_loss": 0.04262898862361908, "global_step": 254549, "epoch": 2860, "lr": 9.568520281641846e-07}
{"train_loss": 0.057595547288656235, "global_step": 254550, "epoch": 2860, "lr": 9.56739157094294e-07}
{"train_loss": 0.06225980818271637, "global_step": 254551, "epoch": 2860, "lr": 9.566262926176795e-07}
{"train_loss": 0.03475469723343849, "global_step": 254552, "epoch": 2860, "lr": 9.565134347343408e-07}
{"train_loss": 0.05708261579275131, "global_step": 254553, "epoch": 2860, "lr": 9.564005834443113e-07}
{"train_loss": 0.023136232048273087, "global_step": 254554, "epoch": 2860, "lr": 9.562877387475854e-07}
{"train_loss": 0.05563519150018692, "global_step": 254555, "epoch": 2860, "lr": 9.561749006441968e-07}
{"train_loss": 0.07186207175254822, "global_step": 254556, "epoch": 2860, "lr": 9.560620691341448e-07}
{"train_loss": 0.05541939288377762, "global_step": 254557, "epoch": 2860, "lr": 9.559492442174634e-07}
{"train_loss": 0.04225020855665207, "global_step": 254558, "epoch": 2860, "lr": 9.558364258941465e-07}
{"train_loss": 0.029384536668658257, "global_step": 254559, "epoch": 2860, "lr": 9.557236141642222e-07}
{"train_loss": 0.0409589558839798, "global_step": 254560, "epoch": 2860, "lr": 9.556108090277072e-07}
{"train_loss": 0.059797514230012894, "global_step": 254561, "epoch": 2860, "lr": 9.554980104846068e-07}
{"train_loss": 0.04094884172081947, "global_step": 254562, "epoch": 2860, "lr": 9.553852185349487e-07}
{"train_loss": 0.04693219065666199, "global_step": 254563, "epoch": 2860, "lr": 9.552724331787333e-07}
{"train_loss": 0.031247057020664215, "global_step": 254564, "epoch": 2860, "lr": 9.551596544159935e-07}
{"train_loss": 0.03288154676556587, "global_step": 254565, "epoch": 2860, "lr": 9.550468822467295e-07}
{"train_loss": 0.018078314140439034, "global_step": 254566, "epoch": 2860, "lr": 9.54934116670958e-07}
{"train_loss": 0.042879868298769, "global_step": 254567, "epoch": 2860, "lr": 9.548213576887066e-07}
{"train_loss": 0.047145694494247437, "global_step": 254568, "epoch": 2860, "lr": 9.547086052999754e-07}
{"train_loss": 0.06935864686965942, "global_step": 254569, "epoch": 2860, "lr": 9.545958595047866e-07}
{"train_loss": 0.05950988829135895, "global_step": 254570, "epoch": 2860, "lr": 9.544831203031568e-07}
{"train_loss": 0.09150747209787369, "global_step": 254571, "epoch": 2860, "lr": 9.543703876950972e-07}
{"train_loss": 0.02104177139699459, "global_step": 254572, "epoch": 2860, "lr": 9.542576616806242e-07}
{"train_loss": 0.022103825584053993, "global_step": 254573, "epoch": 2860, "lr": 9.541449422597548e-07}
{"train_loss": 0.029582032933831215, "global_step": 254574, "epoch": 2860, "lr": 9.540322294325054e-07}
{"train_loss": 0.048819348216056824, "global_step": 254575, "epoch": 2860, "lr": 9.539195231988817e-07}
{"train_loss": 0.03568542003631592, "global_step": 254576, "epoch": 2860, "lr": 9.53806823558906e-07}
{"train_loss": 0.04771342873573303, "global_step": 254577, "epoch": 2860, "lr": 9.536941305126002e-07}
{"train_loss": 0.06331366300582886, "global_step": 254578, "epoch": 2860, "lr": 9.535814440599644e-07}
{"train_loss": 0.02450489066541195, "global_step": 254579, "epoch": 2860, "lr": 9.534687642010321e-07}
{"train_loss": 0.05008057504892349, "global_step": 254580, "epoch": 2860, "lr": 9.533560909357975e-07}
{"train_loss": 0.061585377901792526, "global_step": 254581, "epoch": 2860, "lr": 9.53243424264294e-07}
{"train_loss": 0.029034461826086044, "global_step": 254582, "epoch": 2860, "lr": 9.531307641865217e-07}
{"train_loss": 0.0676381066441536, "global_step": 254583, "epoch": 2860, "lr": 9.530181107025137e-07}
{"train_loss": 0.00937475636601448, "global_step": 254584, "epoch": 2860, "lr": 9.529054638122647e-07}
{"train_loss": 0.059048958122730255, "global_step": 254585, "epoch": 2860, "lr": 9.527928235158023e-07}
{"train_loss": 0.02591932937502861, "global_step": 254586, "epoch": 2860, "lr": 9.526801898131432e-07}
{"train_loss": 0.04331535845994949, "global_step": 254587, "epoch": 2860, "lr": 9.525675627042929e-07}
{"train_loss": 0.05911651626229286, "global_step": 254588, "epoch": 2860, "lr": 9.524549421892737e-07}
{"train_loss": 0.034877073019742966, "global_step": 254589, "epoch": 2860, "lr": 9.523423282681021e-07}
{"train_loss": 0.022715212777256966, "global_step": 254590, "epoch": 2860, "lr": 9.522297209407893e-07}
{"train_loss": 0.058812834322452545, "global_step": 254591, "epoch": 2860, "lr": 9.52117120207352e-07}
{"train_loss": 0.06065068021416664, "global_step": 254592, "epoch": 2860, "lr": 9.520045260678012e-07}
{"train_loss": 0.06126176565885544, "global_step": 254593, "epoch": 2860, "lr": 9.518919385221592e-07}
{"train_loss": 0.07055787742137909, "global_step": 254594, "epoch": 2860, "lr": 9.517793575704315e-07}
{"train_loss": 0.02082027681171894, "global_step": 254595, "epoch": 2860, "lr": 9.516667832126457e-07}
{"train_loss": 0.08289183676242828, "global_step": 254596, "epoch": 2860, "lr": 9.515542154488078e-07}
{"train_loss": 0.032524459064006805, "global_step": 254597, "epoch": 2860, "lr": 9.514416542789339e-07}
{"train_loss": 0.08141405135393143, "global_step": 254598, "epoch": 2860, "lr": 9.513290997030467e-07}
{"train_loss": 0.034943293780088425, "global_step": 254599, "epoch": 2860, "lr": 9.512165517211514e-07}
{"train_loss": 0.04428574815392494, "global_step": 254600, "epoch": 2860, "lr": 9.511040103332702e-07}
{"train_loss": 0.04691696539521217, "global_step": 254601, "epoch": 2860, "lr": 9.50991475539409e-07}
{"train_loss": 0.04333390295505524, "global_step": 254602, "epoch": 2860, "lr": 9.508789473395951e-07}
{"train_loss": 0.018893662840127945, "global_step": 254603, "epoch": 2860, "lr": 9.507664257338344e-07}
{"train_loss": 0.05170896276831627, "global_step": 254604, "epoch": 2860, "lr": 9.50653910722149e-07}
{"train_loss": 0.07270480692386627, "global_step": 254605, "epoch": 2860, "lr": 9.5054140230455e-07}
{"train_loss": 0.06507252156734467, "global_step": 254606, "epoch": 2860, "lr": 9.504289004810485e-07}
{"train_loss": 0.09790904819965363, "global_step": 254607, "epoch": 2860, "lr": 9.503164052516667e-07}
{"train_loss": 0.04193861037492752, "global_step": 254608, "epoch": 2860, "lr": 9.502039166164156e-07}
{"train_loss": 0.032255493104457855, "global_step": 254609, "epoch": 2860, "lr": 9.50091434575312e-07}
{"train_loss": 0.044340189546346664, "global_step": 254610, "epoch": 2860, "lr": 9.499789591283781e-07}
{"train_loss": 0.02380617894232273, "global_step": 254611, "epoch": 2860, "lr": 9.498664902756138e-07}
{"train_loss": 0.0840965136885643, "global_step": 254612, "epoch": 2860, "lr": 9.497540280170413e-07}
{"train_loss": 0.0426819734275341, "global_step": 254613, "epoch": 2860, "lr": 9.496415723526775e-07}
{"train_loss": 0.05556963384151459, "global_step": 254614, "epoch": 2860, "lr": 9.495291232825387e-07}
{"train_loss": 0.06904221326112747, "global_step": 254615, "epoch": 2860, "lr": 9.494166808066363e-07}
{"train_loss": 0.029643740504980087, "global_step": 254616, "epoch": 2860, "lr": 9.493042449249867e-07}
{"train_loss": 0.06460423767566681, "global_step": 254617, "epoch": 2860, "lr": 9.491918156376067e-07}
{"train_loss": 0.043038636445999146, "global_step": 254618, "epoch": 2860, "lr": 9.490793929445074e-07}
{"train_loss": 0.09818976372480392, "global_step": 254619, "epoch": 2860, "lr": 9.48966976845711e-07}
{"train_loss": 0.029794715344905853, "global_step": 254620, "epoch": 2860, "lr": 9.488545673412231e-07}
{"train_loss": 0.028040308505296707, "global_step": 254621, "epoch": 2860, "lr": 9.487421644310712e-07}
{"train_loss": 0.04919221252202988, "global_step": 254622, "epoch": 2860, "lr": 9.486297681152557e-07}
{"train_loss": 0.014985292218625546, "global_step": 254623, "epoch": 2860, "lr": 9.48517378393804e-07}
{"train_loss": 0.016717039048671722, "global_step": 254624, "epoch": 2860, "lr": 9.484049952667218e-07}
{"train_loss": 0.038334209471940994, "global_step": 254625, "epoch": 2860, "lr": 9.482926187340313e-07}
{"train_loss": 0.025728989392518997, "global_step": 254626, "epoch": 2860, "lr": 9.481802487957436e-07}
{"train_loss": 0.022117050364613533, "global_step": 254627, "epoch": 2860, "lr": 9.48067885451881e-07}
{"train_loss": 0.045910175408372715, "global_step": 254628, "epoch": 2860, "lr": 9.479555287024433e-07, "val_loss": 9.687135696411133, "train_action_mse_error": 34.05487823486328}
{"train_loss": 0.07346805930137634, "global_step": 254629, "epoch": 2861, "lr": 9.47843178547464e-07}
{"train_loss": 0.07737758755683899, "global_step": 254630, "epoch": 2861, "lr": 9.47730834986943e-07}
{"train_loss": 0.06858840584754944, "global_step": 254631, "epoch": 2861, "lr": 9.476184980209025e-07}
{"train_loss": 0.04196646064519882, "global_step": 254632, "epoch": 2861, "lr": 9.47506167649359e-07}
{"train_loss": 0.04287875071167946, "global_step": 254633, "epoch": 2861, "lr": 9.47393843872324e-07}
{"train_loss": 0.059955328702926636, "global_step": 254634, "epoch": 2861, "lr": 9.472815266898138e-07}
{"train_loss": 0.0630389153957367, "global_step": 254635, "epoch": 2861, "lr": 9.471692161018453e-07}
{"train_loss": 0.03591807186603546, "global_step": 254636, "epoch": 2861, "lr": 9.470569121084294e-07}
{"train_loss": 0.030312063172459602, "global_step": 254637, "epoch": 2861, "lr": 9.469446147095884e-07}
{"train_loss": 0.08203131705522537, "global_step": 254638, "epoch": 2861, "lr": 9.468323239053278e-07}
{"train_loss": 0.05776205286383629, "global_step": 254639, "epoch": 2861, "lr": 9.467200396956699e-07}
{"train_loss": 0.02957683801651001, "global_step": 254640, "epoch": 2861, "lr": 9.466077620806312e-07}
{"train_loss": 0.04047805815935135, "global_step": 254641, "epoch": 2861, "lr": 9.464954910602175e-07}
{"train_loss": 0.027890726923942566, "global_step": 254642, "epoch": 2861, "lr": 9.463832266344507e-07}
{"train_loss": 0.08755306899547577, "global_step": 254643, "epoch": 2861, "lr": 9.462709688033533e-07}
{"train_loss": 0.05136813595890999, "global_step": 254644, "epoch": 2861, "lr": 9.461587175669196e-07}
{"train_loss": 0.03344365209341049, "global_step": 254645, "epoch": 2861, "lr": 9.460464729251828e-07}
{"train_loss": 0.022902822121977806, "global_step": 254646, "epoch": 2861, "lr": 9.459342348781486e-07}
{"train_loss": 0.05211334303021431, "global_step": 254647, "epoch": 2861, "lr": 9.458220034258448e-07}
{"train_loss": 0.04353279247879982, "global_step": 254648, "epoch": 2861, "lr": 9.457097785682711e-07}
{"train_loss": 0.07332169264554977, "global_step": 254649, "epoch": 2861, "lr": 9.455975603054446e-07}
{"train_loss": 0.025320516899228096, "global_step": 254650, "epoch": 2861, "lr": 9.454853486373871e-07}
{"train_loss": 0.06763055920600891, "global_step": 254651, "epoch": 2861, "lr": 9.453731435641155e-07}
{"train_loss": 0.06075994670391083, "global_step": 254652, "epoch": 2861, "lr": 9.452609450856353e-07}
{"train_loss": 0.031376998871564865, "global_step": 254653, "epoch": 2861, "lr": 9.451487532019743e-07}
{"train_loss": 0.013356231153011322, "global_step": 254654, "epoch": 2861, "lr": 9.450365679131324e-07}
{"train_loss": 0.040358319878578186, "global_step": 254655, "epoch": 2861, "lr": 9.449243892191373e-07}
{"train_loss": 0.03734204173088074, "global_step": 254656, "epoch": 2861, "lr": 9.448122171199947e-07}
{"train_loss": 0.04597865417599678, "global_step": 254657, "epoch": 2861, "lr": 9.447000516157323e-07}
{"train_loss": 0.013477128930389881, "global_step": 254658, "epoch": 2861, "lr": 9.4458789270635e-07}
{"train_loss": 0.027477048337459564, "global_step": 254659, "epoch": 2861, "lr": 9.444757403918703e-07}
{"train_loss": 0.03159289434552193, "global_step": 254660, "epoch": 2861, "lr": 9.443635946723095e-07}
{"train_loss": 0.08741357922554016, "global_step": 254661, "epoch": 2861, "lr": 9.442514555476789e-07}
{"train_loss": 0.039189498871564865, "global_step": 254662, "epoch": 2861, "lr": 9.441393230180006e-07}
{"train_loss": 0.045673903077840805, "global_step": 254663, "epoch": 2861, "lr": 9.440271970832803e-07}
{"train_loss": 0.05232434347271919, "global_step": 254664, "epoch": 2861, "lr": 9.439150777435402e-07}
{"train_loss": 0.03671785816550255, "global_step": 254665, "epoch": 2861, "lr": 9.438029649987912e-07}
{"train_loss": 0.040045805275440216, "global_step": 254666, "epoch": 2861, "lr": 9.436908588490501e-07}
{"train_loss": 0.05762956663966179, "global_step": 254667, "epoch": 2861, "lr": 9.435787592943279e-07}
{"train_loss": 0.023628439754247665, "global_step": 254668, "epoch": 2861, "lr": 9.434666663346525e-07}
{"train_loss": 0.086088627576828, "global_step": 254669, "epoch": 2861, "lr": 9.433545799700239e-07}
{"train_loss": 0.03362540155649185, "global_step": 254670, "epoch": 2861, "lr": 9.432425002004697e-07}
{"train_loss": 0.018992595374584198, "global_step": 254671, "epoch": 2861, "lr": 9.4313042702599e-07}
{"train_loss": 0.041235703974962234, "global_step": 254672, "epoch": 2861, "lr": 9.430183604466125e-07}
{"train_loss": 0.056099146604537964, "global_step": 254673, "epoch": 2861, "lr": 9.429063004623484e-07}
{"train_loss": 0.031934257596731186, "global_step": 254674, "epoch": 2861, "lr": 9.427942470732143e-07}
{"train_loss": 0.05156492441892624, "global_step": 254675, "epoch": 2861, "lr": 9.426822002792212e-07}
{"train_loss": 0.06289064139127731, "global_step": 254676, "epoch": 2861, "lr": 9.42570160080386e-07}
{"train_loss": 0.024858664721250534, "global_step": 254677, "epoch": 2861, "lr": 9.424581264767252e-07}
{"train_loss": 0.05138126760721207, "global_step": 254678, "epoch": 2861, "lr": 9.423460994682554e-07}
{"train_loss": 0.10373150557279587, "global_step": 254679, "epoch": 2861, "lr": 9.422340790549877e-07}
{"train_loss": 0.01464416366070509, "global_step": 254680, "epoch": 2861, "lr": 9.421220652369389e-07}
{"train_loss": 0.14240865409374237, "global_step": 254681, "epoch": 2861, "lr": 9.420100580141255e-07}
{"train_loss": 0.06399146467447281, "global_step": 254682, "epoch": 2861, "lr": 9.418980573865532e-07}
{"train_loss": 0.043889712542295456, "global_step": 254683, "epoch": 2861, "lr": 9.417860633542552e-07}
{"train_loss": 0.02734457142651081, "global_step": 254684, "epoch": 2861, "lr": 9.41674075917226e-07}
{"train_loss": 0.06747331470251083, "global_step": 254685, "epoch": 2861, "lr": 9.415620950754989e-07}
{"train_loss": 0.044318389147520065, "global_step": 254686, "epoch": 2861, "lr": 9.414501208290794e-07}
{"train_loss": 0.02181786298751831, "global_step": 254687, "epoch": 2861, "lr": 9.413381531779841e-07}
{"train_loss": 0.045946210622787476, "global_step": 254688, "epoch": 2861, "lr": 9.412261921222243e-07}
{"train_loss": 0.040690742433071136, "global_step": 254689, "epoch": 2861, "lr": 9.41114237661822e-07}
{"train_loss": 0.04020826146006584, "global_step": 254690, "epoch": 2861, "lr": 9.410022897967829e-07}
{"train_loss": 0.06961330771446228, "global_step": 254691, "epoch": 2861, "lr": 9.408903485271348e-07}
{"train_loss": 0.028390055522322655, "global_step": 254692, "epoch": 2861, "lr": 9.40778413852883e-07}
{"train_loss": 0.04923316463828087, "global_step": 254693, "epoch": 2861, "lr": 9.4066648577405e-07}
{"train_loss": 0.07162737846374512, "global_step": 254694, "epoch": 2861, "lr": 9.40554564290641e-07}
{"train_loss": 0.02513270080089569, "global_step": 254695, "epoch": 2861, "lr": 9.404426494026786e-07}
{"train_loss": 0.050475385040044785, "global_step": 254696, "epoch": 2861, "lr": 9.403307411101736e-07}
{"train_loss": 0.04318786785006523, "global_step": 254697, "epoch": 2861, "lr": 9.402188394131483e-07}
{"train_loss": 0.06809744238853455, "global_step": 254698, "epoch": 2861, "lr": 9.401069443116084e-07}
{"train_loss": 0.031786493957042694, "global_step": 254699, "epoch": 2861, "lr": 9.399950558055759e-07}
{"train_loss": 0.05435969680547714, "global_step": 254700, "epoch": 2861, "lr": 9.398831738950564e-07}
{"train_loss": 0.0626755878329277, "global_step": 254701, "epoch": 2861, "lr": 9.397712985800833e-07}
{"train_loss": 0.027521681040525436, "global_step": 254702, "epoch": 2861, "lr": 9.396594298606509e-07}
{"train_loss": 0.0361173078417778, "global_step": 254703, "epoch": 2861, "lr": 9.395475677367816e-07}
{"train_loss": 0.029210248962044716, "global_step": 254704, "epoch": 2861, "lr": 9.394357122084974e-07}
{"train_loss": 0.02866901457309723, "global_step": 254705, "epoch": 2861, "lr": 9.393238632758095e-07}
{"train_loss": 0.018581459298729897, "global_step": 254706, "epoch": 2861, "lr": 9.39212020938729e-07}
{"train_loss": 0.012949429452419281, "global_step": 254707, "epoch": 2861, "lr": 9.391001851972725e-07}
{"train_loss": 0.02045770175755024, "global_step": 254708, "epoch": 2861, "lr": 9.389883560514568e-07}
{"train_loss": 0.051053691655397415, "global_step": 254709, "epoch": 2861, "lr": 9.388765335012984e-07}
{"train_loss": 0.05483957380056381, "global_step": 254710, "epoch": 2861, "lr": 9.387647175468084e-07}
{"train_loss": 0.0394880473613739, "global_step": 254711, "epoch": 2861, "lr": 9.386529081880036e-07}
{"train_loss": 0.04418305307626724, "global_step": 254712, "epoch": 2861, "lr": 9.385411054249004e-07}
{"train_loss": 0.03421202674508095, "global_step": 254713, "epoch": 2861, "lr": 9.384293092575047e-07}
{"train_loss": 0.03618321195244789, "global_step": 254714, "epoch": 2861, "lr": 9.383175196858496e-07}
{"train_loss": 0.05452017858624458, "global_step": 254715, "epoch": 2861, "lr": 9.382057367099351e-07}
{"train_loss": 0.040602125227451324, "global_step": 254716, "epoch": 2861, "lr": 9.380939603297778e-07}
{"train_loss": 0.0463338615268134, "global_step": 254717, "epoch": 2861, "lr": 9.379821905454e-07, "val_loss": 9.630291938781738}
{"train_loss": 0.07001513242721558, "global_step": 254718, "epoch": 2862, "lr": 9.378704273568129e-07}
{"train_loss": 0.059024129062891006, "global_step": 254719, "epoch": 2862, "lr": 9.377586707640274e-07}
{"train_loss": 0.10106988251209259, "global_step": 254720, "epoch": 2862, "lr": 9.376469207670657e-07}
{"train_loss": 0.033041417598724365, "global_step": 254721, "epoch": 2862, "lr": 9.375351773659446e-07}
{"train_loss": 0.040022656321525574, "global_step": 254722, "epoch": 2862, "lr": 9.374234405606642e-07}
{"train_loss": 0.030750157311558723, "global_step": 254723, "epoch": 2862, "lr": 9.373117103512519e-07}
{"train_loss": 0.04730803146958351, "global_step": 254724, "epoch": 2862, "lr": 9.371999867377191e-07}
{"train_loss": 0.06924305856227875, "global_step": 254725, "epoch": 2862, "lr": 9.370882697200878e-07}
{"train_loss": 0.061081208288669586, "global_step": 254726, "epoch": 2862, "lr": 9.369765592983581e-07}
{"train_loss": 0.01787092536687851, "global_step": 254727, "epoch": 2862, "lr": 9.368648554725635e-07}
{"train_loss": 0.08389826864004135, "global_step": 254728, "epoch": 2862, "lr": 9.367531582427036e-07}
{"train_loss": 0.029423244297504425, "global_step": 254729, "epoch": 2862, "lr": 9.366414676088009e-07}
{"train_loss": 0.057590171694755554, "global_step": 254730, "epoch": 2862, "lr": 9.365297835708664e-07}
{"train_loss": 0.04695767164230347, "global_step": 254731, "epoch": 2862, "lr": 9.364181061289168e-07}
{"train_loss": 0.07290799915790558, "global_step": 254732, "epoch": 2862, "lr": 9.363064352829687e-07}
{"train_loss": 0.05238153412938118, "global_step": 254733, "epoch": 2862, "lr": 9.361947710330387e-07}
{"train_loss": 0.03255083039402962, "global_step": 254734, "epoch": 2862, "lr": 9.360831133791382e-07}
{"train_loss": 0.04444735124707222, "global_step": 254735, "epoch": 2862, "lr": 9.359714623212835e-07}
{"train_loss": 0.0624249204993248, "global_step": 254736, "epoch": 2862, "lr": 9.358598178594913e-07}
{"train_loss": 0.06981271505355835, "global_step": 254737, "epoch": 2862, "lr": 9.357481799937729e-07}
{"train_loss": 0.030587023124098778, "global_step": 254738, "epoch": 2862, "lr": 9.356365487241448e-07}
{"train_loss": 0.06206890568137169, "global_step": 254739, "epoch": 2862, "lr": 9.355249240506236e-07}
{"train_loss": 0.017635870724916458, "global_step": 254740, "epoch": 2862, "lr": 9.354133059732206e-07}
{"train_loss": 0.03224676474928856, "global_step": 254741, "epoch": 2862, "lr": 9.353016944919524e-07}
{"train_loss": 0.038799069821834564, "global_step": 254742, "epoch": 2862, "lr": 9.351900896068355e-07}
{"train_loss": 0.038487959653139114, "global_step": 254743, "epoch": 2862, "lr": 9.350784913178868e-07}
{"train_loss": 0.09158802777528763, "global_step": 254744, "epoch": 2862, "lr": 9.34966899625117e-07}
{"train_loss": 0.06837406754493713, "global_step": 254745, "epoch": 2862, "lr": 9.348553145285432e-07}
{"train_loss": 0.05249762907624245, "global_step": 254746, "epoch": 2862, "lr": 9.347437360281819e-07}
{"train_loss": 0.03813548758625984, "global_step": 254747, "epoch": 2862, "lr": 9.346321641240441e-07}
{"train_loss": 0.023527413606643677, "global_step": 254748, "epoch": 2862, "lr": 9.345205988161465e-07}
{"train_loss": 0.03910326212644577, "global_step": 254749, "epoch": 2862, "lr": 9.344090401045058e-07}
{"train_loss": 0.09228969365358353, "global_step": 254750, "epoch": 2862, "lr": 9.342974879891386e-07}
{"train_loss": 0.011903108097612858, "global_step": 254751, "epoch": 2862, "lr": 9.341859424700506e-07}
{"train_loss": 0.023928314447402954, "global_step": 254752, "epoch": 2862, "lr": 9.340744035472692e-07}
{"train_loss": 0.05790410190820694, "global_step": 254753, "epoch": 2862, "lr": 9.339628712208004e-07}
{"train_loss": 0.04365835711359978, "global_step": 254754, "epoch": 2862, "lr": 9.338513454906661e-07}
{"train_loss": 0.029681656509637833, "global_step": 254755, "epoch": 2862, "lr": 9.337398263568719e-07}
{"train_loss": 0.034314002841711044, "global_step": 254756, "epoch": 2862, "lr": 9.336283138194402e-07}
{"train_loss": 0.030891790986061096, "global_step": 254757, "epoch": 2862, "lr": 9.335168078783818e-07}
{"train_loss": 0.04148125275969505, "global_step": 254758, "epoch": 2862, "lr": 9.334053085337191e-07}
{"train_loss": 0.033501237630844116, "global_step": 254759, "epoch": 2862, "lr": 9.332938157854631e-07}
{"train_loss": 0.04532140493392944, "global_step": 254760, "epoch": 2862, "lr": 9.331823296336251e-07}
{"train_loss": 0.03417419642210007, "global_step": 254761, "epoch": 2862, "lr": 9.330708500782215e-07}
{"train_loss": 0.05469411611557007, "global_step": 254762, "epoch": 2862, "lr": 9.32959377119269e-07}
{"train_loss": 0.035306915640830994, "global_step": 254763, "epoch": 2862, "lr": 9.328479107567845e-07}
{"train_loss": 0.028574025258421898, "global_step": 254764, "epoch": 2862, "lr": 9.327364509907732e-07}
{"train_loss": 0.009230019524693489, "global_step": 254765, "epoch": 2862, "lr": 9.326249978212687e-07}
{"train_loss": 0.05459539592266083, "global_step": 254766, "epoch": 2862, "lr": 9.325135512482652e-07}
{"train_loss": 0.04480316862463951, "global_step": 254767, "epoch": 2862, "lr": 9.324021112717906e-07}
{"train_loss": 0.06961644440889359, "global_step": 254768, "epoch": 2862, "lr": 9.32290677891856e-07}
{"train_loss": 0.04628355801105499, "global_step": 254769, "epoch": 2862, "lr": 9.321792511084782e-07}
{"train_loss": 0.034813717007637024, "global_step": 254770, "epoch": 2862, "lr": 9.320678309216679e-07}
{"train_loss": 0.018305357545614243, "global_step": 254771, "epoch": 2862, "lr": 9.319564173314477e-07}
{"train_loss": 0.021134458482265472, "global_step": 254772, "epoch": 2862, "lr": 9.318450103378229e-07}
{"train_loss": 0.028168844059109688, "global_step": 254773, "epoch": 2862, "lr": 9.317336099408214e-07}
{"train_loss": 0.03326372802257538, "global_step": 254774, "epoch": 2862, "lr": 9.316222161404376e-07}
{"train_loss": 0.010878469794988632, "global_step": 254775, "epoch": 2862, "lr": 9.315108289367102e-07}
{"train_loss": 0.038409747183322906, "global_step": 254776, "epoch": 2862, "lr": 9.313994483296395e-07}
{"train_loss": 0.10281872004270554, "global_step": 254777, "epoch": 2862, "lr": 9.312880743192421e-07}
{"train_loss": 0.01033863052725792, "global_step": 254778, "epoch": 2862, "lr": 9.311767069055344e-07}
{"train_loss": 0.037084419280290604, "global_step": 254779, "epoch": 2862, "lr": 9.310653460885332e-07}
{"train_loss": 0.056601520627737045, "global_step": 254780, "epoch": 2862, "lr": 9.309539918682553e-07}
{"train_loss": 0.0796477273106575, "global_step": 254781, "epoch": 2862, "lr": 9.308426442447116e-07}
{"train_loss": 0.018262354657053947, "global_step": 254782, "epoch": 2862, "lr": 9.307313032179188e-07}
{"train_loss": 0.07386618107557297, "global_step": 254783, "epoch": 2862, "lr": 9.306199687878825e-07}
{"train_loss": 0.05331771820783615, "global_step": 254784, "epoch": 2862, "lr": 9.30508640954636e-07}
{"train_loss": 0.08169381320476532, "global_step": 254785, "epoch": 2862, "lr": 9.303973197181793e-07}
{"train_loss": 0.09688465297222137, "global_step": 254786, "epoch": 2862, "lr": 9.3028600507854e-07}
{"train_loss": 0.044951312243938446, "global_step": 254787, "epoch": 2862, "lr": 9.301746970357183e-07}
{"train_loss": 0.09209109097719193, "global_step": 254788, "epoch": 2862, "lr": 9.300633955897364e-07}
{"train_loss": 0.017165634781122208, "global_step": 254789, "epoch": 2862, "lr": 9.299521007406109e-07}
{"train_loss": 0.021783696487545967, "global_step": 254790, "epoch": 2862, "lr": 9.298408124883584e-07}
{"train_loss": 0.04938845708966255, "global_step": 254791, "epoch": 2862, "lr": 9.297295308329845e-07}
{"train_loss": 0.050250452011823654, "global_step": 254792, "epoch": 2862, "lr": 9.29618255774517e-07}
{"train_loss": 0.061258651316165924, "global_step": 254793, "epoch": 2862, "lr": 9.295069873129614e-07}
{"train_loss": 0.04613736271858215, "global_step": 254794, "epoch": 2862, "lr": 9.293957254483398e-07}
{"train_loss": 0.029810210689902306, "global_step": 254795, "epoch": 2862, "lr": 9.292844701806525e-07}
{"train_loss": 0.04793227091431618, "global_step": 254796, "epoch": 2862, "lr": 9.291732215099325e-07}
{"train_loss": 0.049093205481767654, "global_step": 254797, "epoch": 2862, "lr": 9.29061979436191e-07}
{"train_loss": 0.02135547436773777, "global_step": 254798, "epoch": 2862, "lr": 9.28950743959428e-07}
{"train_loss": 0.0778556689620018, "global_step": 254799, "epoch": 2862, "lr": 9.288395150796769e-07}
{"train_loss": 0.01937296614050865, "global_step": 254800, "epoch": 2862, "lr": 9.287282927969432e-07}
{"train_loss": 0.029472416266798973, "global_step": 254801, "epoch": 2862, "lr": 9.28617077111249e-07}
{"train_loss": 0.04840317741036415, "global_step": 254802, "epoch": 2862, "lr": 9.285058680225944e-07}
{"train_loss": 0.0525488406419754, "global_step": 254803, "epoch": 2862, "lr": 9.283946655310127e-07}
{"train_loss": 0.06394342333078384, "global_step": 254804, "epoch": 2862, "lr": 9.282834696365039e-07}
{"train_loss": 0.03477132320404053, "global_step": 254805, "epoch": 2862, "lr": 9.281722803390903e-07}
{"train_loss": 0.04632218091047547, "global_step": 254806, "epoch": 2862, "lr": 9.280610976387883e-07, "val_loss": 9.682790756225586}
{"train_loss": 0.015996115282177925, "global_step": 254807, "epoch": 2863, "lr": 9.279499215356091e-07}
{"train_loss": 0.09083139151334763, "global_step": 254808, "epoch": 2863, "lr": 9.278387520295695e-07}
{"train_loss": 0.038037028163671494, "global_step": 254809, "epoch": 2863, "lr": 9.27727589120686e-07}
{"train_loss": 0.07006214559078217, "global_step": 254810, "epoch": 2863, "lr": 9.276164328089642e-07}
{"train_loss": 0.03511996567249298, "global_step": 254811, "epoch": 2863, "lr": 9.275052830944375e-07}
{"train_loss": 0.055964116007089615, "global_step": 254812, "epoch": 2863, "lr": 9.273941399771002e-07}
{"train_loss": 0.056818995624780655, "global_step": 254813, "epoch": 2863, "lr": 9.272830034569746e-07}
{"train_loss": 0.07361184060573578, "global_step": 254814, "epoch": 2863, "lr": 9.271718735340884e-07}
{"train_loss": 0.03288792073726654, "global_step": 254815, "epoch": 2863, "lr": 9.270607502084361e-07}
{"train_loss": 0.030387450009584427, "global_step": 254816, "epoch": 2863, "lr": 9.269496334800454e-07}
{"train_loss": 0.07892909646034241, "global_step": 254817, "epoch": 2863, "lr": 9.268385233489274e-07}
{"train_loss": 0.10364721715450287, "global_step": 254818, "epoch": 2863, "lr": 9.267274198151044e-07}
{"train_loss": 0.05066267028450966, "global_step": 254819, "epoch": 2863, "lr": 9.266163228785762e-07}
{"train_loss": 0.05776141956448555, "global_step": 254820, "epoch": 2863, "lr": 9.265052325393764e-07}
{"train_loss": 0.049051620066165924, "global_step": 254821, "epoch": 2863, "lr": 9.263941487974992e-07}
{"train_loss": 0.04662372171878815, "global_step": 254822, "epoch": 2863, "lr": 9.26283071652978e-07}
{"train_loss": 0.05733892694115639, "global_step": 254823, "epoch": 2863, "lr": 9.261720011058128e-07}
{"train_loss": 0.03193143755197525, "global_step": 254824, "epoch": 2863, "lr": 9.260609371560314e-07}
{"train_loss": 0.0467984676361084, "global_step": 254825, "epoch": 2863, "lr": 9.259498798036392e-07}
{"train_loss": 0.04028923436999321, "global_step": 254826, "epoch": 2863, "lr": 9.258388290486586e-07}
{"train_loss": 0.037769757211208344, "global_step": 254827, "epoch": 2863, "lr": 9.25727784891095e-07}
{"train_loss": 0.04301677271723747, "global_step": 254828, "epoch": 2863, "lr": 9.256167473309763e-07}
{"train_loss": 0.049604933708906174, "global_step": 254829, "epoch": 2863, "lr": 9.255057163683079e-07}
{"train_loss": 0.05672137811779976, "global_step": 254830, "epoch": 2863, "lr": 9.25394692003101e-07}
{"train_loss": 0.038771551102399826, "global_step": 254831, "epoch": 2863, "lr": 9.252836742353832e-07}
{"train_loss": 0.018666256219148636, "global_step": 254832, "epoch": 2863, "lr": 9.251726630651658e-07}
{"train_loss": 0.025760095566511154, "global_step": 254833, "epoch": 2863, "lr": 9.250616584924543e-07}
{"train_loss": 0.025236276909708977, "global_step": 254834, "epoch": 2863, "lr": 9.249506605172764e-07}
{"train_loss": 0.05322263389825821, "global_step": 254835, "epoch": 2863, "lr": 9.248396691396321e-07}
{"train_loss": 0.044949933886528015, "global_step": 254836, "epoch": 2863, "lr": 9.247286843595493e-07}
{"train_loss": 0.03453728184103966, "global_step": 254837, "epoch": 2863, "lr": 9.246177061770445e-07}
{"train_loss": 0.032232582569122314, "global_step": 254838, "epoch": 2863, "lr": 9.245067345921177e-07}
{"train_loss": 0.026859434321522713, "global_step": 254839, "epoch": 2863, "lr": 9.243957696047967e-07}
{"train_loss": 0.06719604879617691, "global_step": 254840, "epoch": 2863, "lr": 9.242848112150926e-07}
{"train_loss": 0.026078831404447556, "global_step": 254841, "epoch": 2863, "lr": 9.24173859423022e-07}
{"train_loss": 0.06154991686344147, "global_step": 254842, "epoch": 2863, "lr": 9.240629142285906e-07}
{"train_loss": 0.004540556576102972, "global_step": 254843, "epoch": 2863, "lr": 9.239519756318316e-07}
{"train_loss": 0.090297631919384, "global_step": 254844, "epoch": 2863, "lr": 9.238410436327394e-07}
{"train_loss": 0.05348076671361923, "global_step": 254845, "epoch": 2863, "lr": 9.237301182313474e-07}
{"train_loss": 0.05210823565721512, "global_step": 254846, "epoch": 2863, "lr": 9.236191994276555e-07}
{"train_loss": 0.04367454722523689, "global_step": 254847, "epoch": 2863, "lr": 9.23508287221686e-07}
{"train_loss": 0.03016085736453533, "global_step": 254848, "epoch": 2863, "lr": 9.2339738161345e-07}
{"train_loss": 0.054954759776592255, "global_step": 254849, "epoch": 2863, "lr": 9.232864826029641e-07}
{"train_loss": 0.02104167267680168, "global_step": 254850, "epoch": 2863, "lr": 9.231755901902506e-07}
{"train_loss": 0.041916295886039734, "global_step": 254851, "epoch": 2863, "lr": 9.230647043753149e-07}
{"train_loss": 0.04789907857775688, "global_step": 254852, "epoch": 2863, "lr": 9.229538251581793e-07}
{"train_loss": 0.049118708819150925, "global_step": 254853, "epoch": 2863, "lr": 9.228429525388494e-07}
{"train_loss": 0.07573746144771576, "global_step": 254854, "epoch": 2863, "lr": 9.227320865173472e-07}
{"train_loss": 0.058155059814453125, "global_step": 254855, "epoch": 2863, "lr": 9.226212270936841e-07}
{"train_loss": 0.04689665511250496, "global_step": 254856, "epoch": 2863, "lr": 9.225103742678765e-07}
{"train_loss": 0.005644559860229492, "global_step": 254857, "epoch": 2863, "lr": 9.223995280399411e-07}
{"train_loss": 0.05776150897145271, "global_step": 254858, "epoch": 2863, "lr": 9.222886884098892e-07}
{"train_loss": 0.03235587850213051, "global_step": 254859, "epoch": 2863, "lr": 9.221778553777372e-07}
{"train_loss": 0.0360652357339859, "global_step": 254860, "epoch": 2863, "lr": 9.220670289435018e-07}
{"train_loss": 0.04488832876086235, "global_step": 254861, "epoch": 2863, "lr": 9.219562091071942e-07}
{"train_loss": 0.0454888641834259, "global_step": 254862, "epoch": 2863, "lr": 9.218453958688312e-07}
{"train_loss": 0.08393311500549316, "global_step": 254863, "epoch": 2863, "lr": 9.21734589228429e-07}
{"train_loss": 0.03160056471824646, "global_step": 254864, "epoch": 2863, "lr": 9.216237891860046e-07}
{"train_loss": 0.03423157334327698, "global_step": 254865, "epoch": 2863, "lr": 9.215129957415636e-07}
{"train_loss": 0.03966711089015007, "global_step": 254866, "epoch": 2863, "lr": 9.214022088951336e-07}
{"train_loss": 0.030258888378739357, "global_step": 254867, "epoch": 2863, "lr": 9.212914286467144e-07}
{"train_loss": 0.0383509024977684, "global_step": 254868, "epoch": 2863, "lr": 9.211806549963342e-07}
{"train_loss": 0.031094808131456375, "global_step": 254869, "epoch": 2863, "lr": 9.210698879440038e-07}
{"train_loss": 0.03289417922496796, "global_step": 254870, "epoch": 2863, "lr": 9.2095912748974e-07}
{"train_loss": 0.020220143720507622, "global_step": 254871, "epoch": 2863, "lr": 9.208483736335482e-07}
{"train_loss": 0.036883097141981125, "global_step": 254872, "epoch": 2863, "lr": 9.207376263754564e-07}
{"train_loss": 0.02536778710782528, "global_step": 254873, "epoch": 2863, "lr": 9.206268857154698e-07}
{"train_loss": 0.02805691584944725, "global_step": 254874, "epoch": 2863, "lr": 9.205161516536109e-07}
{"train_loss": 0.03733083978295326, "global_step": 254875, "epoch": 2863, "lr": 9.204054241898852e-07}
{"train_loss": 0.03137587383389473, "global_step": 254876, "epoch": 2863, "lr": 9.202947033243148e-07}
{"train_loss": 0.028272677212953568, "global_step": 254877, "epoch": 2863, "lr": 9.201839890569164e-07}
{"train_loss": 0.0471065528690815, "global_step": 254878, "epoch": 2863, "lr": 9.200732813876955e-07}
{"train_loss": 0.05406497046351433, "global_step": 254879, "epoch": 2863, "lr": 9.199625803166745e-07}
{"train_loss": 0.033538687974214554, "global_step": 254880, "epoch": 2863, "lr": 9.198518858438643e-07}
{"train_loss": 0.09774108231067657, "global_step": 254881, "epoch": 2863, "lr": 9.197411979692871e-07}
{"train_loss": 0.06707826256752014, "global_step": 254882, "epoch": 2863, "lr": 9.19630516692943e-07}
{"train_loss": 0.04741576313972473, "global_step": 254883, "epoch": 2863, "lr": 9.195198420148655e-07}
{"train_loss": 0.02752758376300335, "global_step": 254884, "epoch": 2863, "lr": 9.194091739350541e-07}
{"train_loss": 0.021821392700076103, "global_step": 254885, "epoch": 2863, "lr": 9.192985124535313e-07}
{"train_loss": 0.057652149349451065, "global_step": 254886, "epoch": 2863, "lr": 9.191878575703139e-07}
{"train_loss": 0.053879912942647934, "global_step": 254887, "epoch": 2863, "lr": 9.190772092854128e-07}
{"train_loss": 0.04366498813033104, "global_step": 254888, "epoch": 2863, "lr": 9.189665675988446e-07}
{"train_loss": 0.025499427691102028, "global_step": 254889, "epoch": 2863, "lr": 9.18855932510615e-07}
{"train_loss": 0.026857750490307808, "global_step": 254890, "epoch": 2863, "lr": 9.187453040207572e-07}
{"train_loss": 0.049538273364305496, "global_step": 254891, "epoch": 2863, "lr": 9.186346821292713e-07}
{"train_loss": 0.04786531627178192, "global_step": 254892, "epoch": 2863, "lr": 9.185240668361794e-07}
{"train_loss": 0.013013677671551704, "global_step": 254893, "epoch": 2863, "lr": 9.184134581414872e-07}
{"train_loss": 0.04511721432209015, "global_step": 254894, "epoch": 2863, "lr": 9.183028560452223e-07}
{"train_loss": 0.04422445760077138, "global_step": 254895, "epoch": 2863, "lr": 9.181922605473903e-07, "val_loss": 9.60084056854248}
{"train_loss": 0.06086374446749687, "global_step": 254896, "epoch": 2864, "lr": 9.180816716480134e-07}
{"train_loss": 0.029475275427103043, "global_step": 254897, "epoch": 2864, "lr": 9.179710893470971e-07}
{"train_loss": 0.014514317736029625, "global_step": 254898, "epoch": 2864, "lr": 9.178605136446639e-07}
{"train_loss": 0.08489920943975449, "global_step": 254899, "epoch": 2864, "lr": 9.177499445407245e-07}
{"train_loss": 0.07615408301353455, "global_step": 254900, "epoch": 2864, "lr": 9.176393820352958e-07}
{"train_loss": 0.03586474433541298, "global_step": 254901, "epoch": 2864, "lr": 9.175288261283943e-07}
{"train_loss": 0.031163614243268967, "global_step": 254902, "epoch": 2864, "lr": 9.174182768200312e-07}
{"train_loss": 0.02463817223906517, "global_step": 254903, "epoch": 2864, "lr": 9.173077341102232e-07}
{"train_loss": 0.04649674892425537, "global_step": 254904, "epoch": 2864, "lr": 9.171971979989813e-07}
{"train_loss": 0.022019144147634506, "global_step": 254905, "epoch": 2864, "lr": 9.170866684863333e-07}
{"train_loss": 0.03614518418908119, "global_step": 254906, "epoch": 2864, "lr": 9.169761455722736e-07}
{"train_loss": 0.09205474704504013, "global_step": 254907, "epoch": 2864, "lr": 9.168656292568356e-07}
{"train_loss": 0.025532739236950874, "global_step": 254908, "epoch": 2864, "lr": 9.167551195400304e-07}
{"train_loss": 0.01864944025874138, "global_step": 254909, "epoch": 2864, "lr": 9.166446164218579e-07}
{"train_loss": 0.0337861031293869, "global_step": 254910, "epoch": 2864, "lr": 9.165341199023514e-07}
{"train_loss": 0.02972816489636898, "global_step": 254911, "epoch": 2864, "lr": 9.164236299815166e-07}
{"train_loss": 0.03483220562338829, "global_step": 254912, "epoch": 2864, "lr": 9.163131466593755e-07}
{"train_loss": 0.04111814498901367, "global_step": 254913, "epoch": 2864, "lr": 9.162026699359339e-07}
{"train_loss": 0.05609184131026268, "global_step": 254914, "epoch": 2864, "lr": 9.160921998112082e-07}
{"train_loss": 0.03002571500837803, "global_step": 254915, "epoch": 2864, "lr": 9.159817362852208e-07}
{"train_loss": 0.09403766691684723, "global_step": 254916, "epoch": 2864, "lr": 9.158712793579771e-07}
{"train_loss": 0.048674967139959335, "global_step": 254917, "epoch": 2864, "lr": 9.157608290294995e-07}
{"train_loss": 0.021736357361078262, "global_step": 254918, "epoch": 2864, "lr": 9.156503852997933e-07}
{"train_loss": 0.07528681308031082, "global_step": 254919, "epoch": 2864, "lr": 9.155399481688864e-07}
{"train_loss": 0.03826981037855148, "global_step": 254920, "epoch": 2864, "lr": 9.154295176367844e-07}
{"train_loss": 0.023303238674998283, "global_step": 254921, "epoch": 2864, "lr": 9.153190937035039e-07}
{"train_loss": 0.0650717094540596, "global_step": 254922, "epoch": 2864, "lr": 9.152086763690615e-07}
{"train_loss": 0.04105164855718613, "global_step": 254923, "epoch": 2864, "lr": 9.150982656334683e-07}
{"train_loss": 0.026792189106345177, "global_step": 254924, "epoch": 2864, "lr": 9.149878614967466e-07}
{"train_loss": 0.05444653704762459, "global_step": 254925, "epoch": 2864, "lr": 9.14877463958902e-07}
{"train_loss": 0.06490454822778702, "global_step": 254926, "epoch": 2864, "lr": 9.147670730199564e-07}
{"train_loss": 0.06655232608318329, "global_step": 254927, "epoch": 2864, "lr": 9.146566886799213e-07}
{"train_loss": 0.0598607137799263, "global_step": 254928, "epoch": 2864, "lr": 9.14546310938813e-07}
{"train_loss": 0.013845516368746758, "global_step": 254929, "epoch": 2864, "lr": 9.144359397966485e-07}
{"train_loss": 0.04136613756418228, "global_step": 254930, "epoch": 2864, "lr": 9.143255752534385e-07}
{"train_loss": 0.0408952422440052, "global_step": 254931, "epoch": 2864, "lr": 9.142152173091945e-07}
{"train_loss": 0.05510544776916504, "global_step": 254932, "epoch": 2864, "lr": 9.141048659639385e-07}
{"train_loss": 0.022532159462571144, "global_step": 254933, "epoch": 2864, "lr": 9.139945212176815e-07}
{"train_loss": 0.04873964563012123, "global_step": 254934, "epoch": 2864, "lr": 9.138841830704459e-07}
{"train_loss": 0.03880392387509346, "global_step": 254935, "epoch": 2864, "lr": 9.137738515222317e-07}
{"train_loss": 0.02592441812157631, "global_step": 254936, "epoch": 2864, "lr": 9.13663526573072e-07}
{"train_loss": 0.028478160500526428, "global_step": 254937, "epoch": 2864, "lr": 9.135532082229614e-07}
{"train_loss": 0.03904890641570091, "global_step": 254938, "epoch": 2864, "lr": 9.134428964719333e-07}
{"train_loss": 0.07615499943494797, "global_step": 254939, "epoch": 2864, "lr": 9.133325913199875e-07}
{"train_loss": 0.04496445134282112, "global_step": 254940, "epoch": 2864, "lr": 9.132222927671519e-07}
{"train_loss": 0.04134948551654816, "global_step": 254941, "epoch": 2864, "lr": 9.13112000813432e-07}
{"train_loss": 0.04456668347120285, "global_step": 254942, "epoch": 2864, "lr": 9.130017154588444e-07}
{"train_loss": 0.05479259788990021, "global_step": 254943, "epoch": 2864, "lr": 9.128914367034113e-07}
{"train_loss": 0.06816824525594711, "global_step": 254944, "epoch": 2864, "lr": 9.127811645471384e-07}
{"train_loss": 0.035419102758169174, "global_step": 254945, "epoch": 2864, "lr": 9.126708989900479e-07}
{"train_loss": 0.06267542392015457, "global_step": 254946, "epoch": 2864, "lr": 9.125606400321451e-07}
{"train_loss": 0.058524686843156815, "global_step": 254947, "epoch": 2864, "lr": 9.124503876734525e-07}
{"train_loss": 0.02894386649131775, "global_step": 254948, "epoch": 2864, "lr": 9.12340141913981e-07}
{"train_loss": 0.046098366379737854, "global_step": 254949, "epoch": 2864, "lr": 9.122299027537473e-07}
{"train_loss": 0.0697825476527214, "global_step": 254950, "epoch": 2864, "lr": 9.121196701927681e-07}
{"train_loss": 0.04603429511189461, "global_step": 254951, "epoch": 2864, "lr": 9.120094442310545e-07}
{"train_loss": 0.03528741002082825, "global_step": 254952, "epoch": 2864, "lr": 9.118992248686231e-07}
{"train_loss": 0.051477883011102676, "global_step": 254953, "epoch": 2864, "lr": 9.117890121054906e-07}
{"train_loss": 0.025021608918905258, "global_step": 254954, "epoch": 2864, "lr": 9.116788059416681e-07}
{"train_loss": 0.05975768342614174, "global_step": 254955, "epoch": 2864, "lr": 9.115686063771723e-07}
{"train_loss": 0.015540674328804016, "global_step": 254956, "epoch": 2864, "lr": 9.114584134120141e-07}
{"train_loss": 0.01921357959508896, "global_step": 254957, "epoch": 2864, "lr": 9.113482270462159e-07}
{"train_loss": 0.02924339845776558, "global_step": 254958, "epoch": 2864, "lr": 9.112380472797888e-07}
{"train_loss": 0.09103376418352127, "global_step": 254959, "epoch": 2864, "lr": 9.111278741127438e-07}
{"train_loss": 0.08663875609636307, "global_step": 254960, "epoch": 2864, "lr": 9.110177075451032e-07}
{"train_loss": 0.05588170140981674, "global_step": 254961, "epoch": 2864, "lr": 9.10907547576878e-07}
{"train_loss": 0.05233556032180786, "global_step": 254962, "epoch": 2864, "lr": 9.107973942080794e-07}
{"train_loss": 0.03209511935710907, "global_step": 254963, "epoch": 2864, "lr": 9.106872474387295e-07}
{"train_loss": 0.04012606292963028, "global_step": 254964, "epoch": 2864, "lr": 9.105771072688396e-07}
{"train_loss": 0.04706965386867523, "global_step": 254965, "epoch": 2864, "lr": 9.104669736984207e-07}
{"train_loss": 0.027572209015488625, "global_step": 254966, "epoch": 2864, "lr": 9.103568467274947e-07}
{"train_loss": 0.030286531895399094, "global_step": 254967, "epoch": 2864, "lr": 9.102467263560676e-07}
{"train_loss": 0.10769105702638626, "global_step": 254968, "epoch": 2864, "lr": 9.101366125841671e-07}
{"train_loss": 0.02664851024746895, "global_step": 254969, "epoch": 2864, "lr": 9.100265054117929e-07}
{"train_loss": 0.04592859745025635, "global_step": 254970, "epoch": 2864, "lr": 9.09916404838973e-07}
{"train_loss": 0.02701123058795929, "global_step": 254971, "epoch": 2864, "lr": 9.098063108657129e-07}
{"train_loss": 0.05317780748009682, "global_step": 254972, "epoch": 2864, "lr": 9.096962234920348e-07}
{"train_loss": 0.06104998663067818, "global_step": 254973, "epoch": 2864, "lr": 9.095861427179497e-07}
{"train_loss": 0.04810066521167755, "global_step": 254974, "epoch": 2864, "lr": 9.094760685434689e-07}
{"train_loss": 0.030402429401874542, "global_step": 254975, "epoch": 2864, "lr": 9.093660009686089e-07}
{"train_loss": 0.04447287693619728, "global_step": 254976, "epoch": 2864, "lr": 9.092559399933919e-07}
{"train_loss": 0.04422236606478691, "global_step": 254977, "epoch": 2864, "lr": 9.091458856178236e-07}
{"train_loss": 0.041388701647520065, "global_step": 254978, "epoch": 2864, "lr": 9.090358378419206e-07}
{"train_loss": 0.0353548638522625, "global_step": 254979, "epoch": 2864, "lr": 9.08925796665705e-07}
{"train_loss": 0.0061292750760912895, "global_step": 254980, "epoch": 2864, "lr": 9.088157620891824e-07}
{"train_loss": 0.03756042569875717, "global_step": 254981, "epoch": 2864, "lr": 9.08705734112375e-07}
{"train_loss": 0.042673539370298386, "global_step": 254982, "epoch": 2864, "lr": 9.085957127352884e-07}
{"train_loss": 0.013694653287529945, "global_step": 254983, "epoch": 2864, "lr": 9.084856979579448e-07}
{"train_loss": 0.044201850503934235, "global_step": 254984, "epoch": 2864, "lr": 9.083756897803552e-07, "val_loss": 9.630621910095215}
{"train_loss": 0.04912516474723816, "global_step": 254985, "epoch": 2865, "lr": 9.08265688202542e-07}
{"train_loss": 0.03614363074302673, "global_step": 254986, "epoch": 2865, "lr": 9.081556932245106e-07}
{"train_loss": 0.041996266692876816, "global_step": 254987, "epoch": 2865, "lr": 9.080457048462776e-07}
{"train_loss": 0.03528992831707001, "global_step": 254988, "epoch": 2865, "lr": 9.079357230678654e-07}
{"train_loss": 0.03632818162441254, "global_step": 254989, "epoch": 2865, "lr": 9.078257478892738e-07}
{"train_loss": 0.028560370206832886, "global_step": 254990, "epoch": 2865, "lr": 9.077157793105362e-07}
{"train_loss": 0.055208586156368256, "global_step": 254991, "epoch": 2865, "lr": 9.076058173316526e-07}
{"train_loss": 0.07260933518409729, "global_step": 254992, "epoch": 2865, "lr": 9.074958619526453e-07}
{"train_loss": 0.04949014261364937, "global_step": 254993, "epoch": 2865, "lr": 9.073859131735252e-07}
{"train_loss": 0.05031426623463631, "global_step": 254994, "epoch": 2865, "lr": 9.072759709943091e-07}
{"train_loss": 0.022063953801989555, "global_step": 254995, "epoch": 2865, "lr": 9.07166035415008e-07}
{"train_loss": 0.03326753154397011, "global_step": 254996, "epoch": 2865, "lr": 9.070561064356442e-07}
{"train_loss": 0.03898286819458008, "global_step": 254997, "epoch": 2865, "lr": 9.069461840562233e-07}
{"train_loss": 0.03181236609816551, "global_step": 254998, "epoch": 2865, "lr": 9.068362682767728e-07}
{"train_loss": 0.0200640931725502, "global_step": 254999, "epoch": 2865, "lr": 9.06726359097293e-07}
{"train_loss": 0.06514057517051697, "global_step": 255000, "epoch": 2865, "lr": 9.066164565178115e-07}
{"train_loss": 0.03269755095243454, "global_step": 255001, "epoch": 2865, "lr": 9.065065605383283e-07}
{"train_loss": 0.05004756897687912, "global_step": 255002, "epoch": 2865, "lr": 9.063966711588767e-07}
{"train_loss": 0.020037595182657242, "global_step": 255003, "epoch": 2865, "lr": 9.062867883794568e-07}
{"train_loss": 0.04733141139149666, "global_step": 255004, "epoch": 2865, "lr": 9.061769122000907e-07}
{"train_loss": 0.04728374257683754, "global_step": 255005, "epoch": 2865, "lr": 9.06067042620784e-07}
{"train_loss": 0.03876306489109993, "global_step": 255006, "epoch": 2865, "lr": 9.0595717964157e-07}
{"train_loss": 0.027549946680665016, "global_step": 255007, "epoch": 2865, "lr": 9.058473232624431e-07}
{"train_loss": 0.04073333367705345, "global_step": 255008, "epoch": 2865, "lr": 9.057374734834257e-07}
{"train_loss": 0.057028841227293015, "global_step": 255009, "epoch": 2865, "lr": 9.056276303045341e-07}
{"train_loss": 0.024388935416936874, "global_step": 255010, "epoch": 2865, "lr": 9.055177937257908e-07}
{"train_loss": 0.056578703224658966, "global_step": 255011, "epoch": 2865, "lr": 9.054079637471901e-07}
{"train_loss": 0.06365546584129333, "global_step": 255012, "epoch": 2865, "lr": 9.052981403687655e-07}
{"train_loss": 0.0345371812582016, "global_step": 255013, "epoch": 2865, "lr": 9.051883235905279e-07}
{"train_loss": 0.007954568602144718, "global_step": 255014, "epoch": 2865, "lr": 9.050785134124828e-07}
{"train_loss": 0.045357465744018555, "global_step": 255015, "epoch": 2865, "lr": 9.049687098346582e-07}
{"train_loss": 0.03151634335517883, "global_step": 255016, "epoch": 2865, "lr": 9.048589128570539e-07}
{"train_loss": 0.03309314697980881, "global_step": 255017, "epoch": 2865, "lr": 9.047491224797033e-07}
{"train_loss": 0.03623144328594208, "global_step": 255018, "epoch": 2865, "lr": 9.046393387026009e-07}
{"train_loss": 0.06398189812898636, "global_step": 255019, "epoch": 2865, "lr": 9.045295615257799e-07}
{"train_loss": 0.04671253636479378, "global_step": 255020, "epoch": 2865, "lr": 9.044197909492403e-07}
{"train_loss": 0.04906131327152252, "global_step": 255021, "epoch": 2865, "lr": 9.043100269730042e-07}
{"train_loss": 0.027985868975520134, "global_step": 255022, "epoch": 2865, "lr": 9.04200269597083e-07}
{"train_loss": 0.040395595133304596, "global_step": 255023, "epoch": 2865, "lr": 9.040905188215043e-07}
{"train_loss": 0.0670032948255539, "global_step": 255024, "epoch": 2865, "lr": 9.039807746462625e-07}
{"train_loss": 0.06599455326795578, "global_step": 255025, "epoch": 2865, "lr": 9.038710370713854e-07}
{"train_loss": 0.04834384098649025, "global_step": 255026, "epoch": 2865, "lr": 9.037613060968841e-07}
{"train_loss": 0.061276212334632874, "global_step": 255027, "epoch": 2865, "lr": 9.036515817227698e-07}
{"train_loss": 0.058725208044052124, "global_step": 255028, "epoch": 2865, "lr": 9.035418639490701e-07}
{"train_loss": 0.06967990100383759, "global_step": 255029, "epoch": 2865, "lr": 9.034321527757849e-07}
{"train_loss": 0.04739232361316681, "global_step": 255030, "epoch": 2865, "lr": 9.033224482029312e-07}
{"train_loss": 0.04091939702630043, "global_step": 255031, "epoch": 2865, "lr": 9.03212750230531e-07}
{"train_loss": 0.0743539035320282, "global_step": 255032, "epoch": 2865, "lr": 9.031030588585953e-07}
{"train_loss": 0.04791906476020813, "global_step": 255033, "epoch": 2865, "lr": 9.02993374087141e-07}
{"train_loss": 0.04774666950106621, "global_step": 255034, "epoch": 2865, "lr": 9.02883695916179e-07}
{"train_loss": 0.06184009835124016, "global_step": 255035, "epoch": 2865, "lr": 9.027740243457261e-07}
{"train_loss": 0.02490338310599327, "global_step": 255036, "epoch": 2865, "lr": 9.026643593757988e-07}
{"train_loss": 0.06818193942308426, "global_step": 255037, "epoch": 2865, "lr": 9.025547010064028e-07}
{"train_loss": 0.028173942118883133, "global_step": 255038, "epoch": 2865, "lr": 9.024450492375713e-07}
{"train_loss": 0.041784267872571945, "global_step": 255039, "epoch": 2865, "lr": 9.023354040692989e-07}
{"train_loss": 0.048192527145147324, "global_step": 255040, "epoch": 2865, "lr": 9.022257655016131e-07}
{"train_loss": 0.030566692352294922, "global_step": 255041, "epoch": 2865, "lr": 9.021161335345196e-07}
{"train_loss": 0.05390964075922966, "global_step": 255042, "epoch": 2865, "lr": 9.020065081680406e-07}
{"train_loss": 0.022490447387099266, "global_step": 255043, "epoch": 2865, "lr": 9.018968894021873e-07}
{"train_loss": 0.030374804511666298, "global_step": 255044, "epoch": 2865, "lr": 9.017872772369817e-07}
{"train_loss": 0.03860985115170479, "global_step": 255045, "epoch": 2865, "lr": 9.01677671672424e-07}
{"train_loss": 0.0216944832354784, "global_step": 255046, "epoch": 2865, "lr": 9.015680727085474e-07}
{"train_loss": 0.05654844269156456, "global_step": 255047, "epoch": 2865, "lr": 9.014584803453463e-07}
{"train_loss": 0.0630025640130043, "global_step": 255048, "epoch": 2865, "lr": 9.013488945828485e-07}
{"train_loss": 0.04492979496717453, "global_step": 255049, "epoch": 2865, "lr": 9.012393154210707e-07}
{"train_loss": 0.027861695736646652, "global_step": 255050, "epoch": 2865, "lr": 9.011297428600185e-07}
{"train_loss": 0.0450436994433403, "global_step": 255051, "epoch": 2865, "lr": 9.01020176899714e-07}
{"train_loss": 0.044301874935626984, "global_step": 255052, "epoch": 2865, "lr": 9.009106175401628e-07}
{"train_loss": 0.05210636183619499, "global_step": 255053, "epoch": 2865, "lr": 9.008010647813925e-07}
{"train_loss": 0.047165144234895706, "global_step": 255054, "epoch": 2865, "lr": 9.006915186234088e-07}
{"train_loss": 0.05800395831465721, "global_step": 255055, "epoch": 2865, "lr": 9.005819790662284e-07}
{"train_loss": 0.02059890516102314, "global_step": 255056, "epoch": 2865, "lr": 9.004724461098624e-07}
{"train_loss": 0.026977602392435074, "global_step": 255057, "epoch": 2865, "lr": 9.003629197543384e-07}
{"train_loss": 0.017574340105056763, "global_step": 255058, "epoch": 2865, "lr": 9.002533999996566e-07}
{"train_loss": 0.05738130956888199, "global_step": 255059, "epoch": 2865, "lr": 9.00143886845839e-07}
{"train_loss": 0.07509669661521912, "global_step": 255060, "epoch": 2865, "lr": 9.000343802928967e-07}
{"train_loss": 0.0374201275408268, "global_step": 255061, "epoch": 2865, "lr": 8.999248803408467e-07}
{"train_loss": 0.07874982804059982, "global_step": 255062, "epoch": 2865, "lr": 8.998153869896997e-07}
{"train_loss": 0.05358648672699928, "global_step": 255063, "epoch": 2865, "lr": 8.997059002394837e-07}
{"train_loss": 0.025065220892429352, "global_step": 255064, "epoch": 2865, "lr": 8.995964200901985e-07}
{"train_loss": 0.07273784279823303, "global_step": 255065, "epoch": 2865, "lr": 8.99486946541861e-07}
{"train_loss": 0.026945093646645546, "global_step": 255066, "epoch": 2865, "lr": 8.993774795944931e-07}
{"train_loss": 0.044243935495615005, "global_step": 255067, "epoch": 2865, "lr": 8.992680192481063e-07}
{"train_loss": 0.03296992927789688, "global_step": 255068, "epoch": 2865, "lr": 8.991585655027112e-07}
{"train_loss": 0.024423634633421898, "global_step": 255069, "epoch": 2865, "lr": 8.99049118358325e-07}
{"train_loss": 0.03337814658880234, "global_step": 255070, "epoch": 2865, "lr": 8.989396778149695e-07}
{"train_loss": 0.032929930835962296, "global_step": 255071, "epoch": 2865, "lr": 8.988302438726448e-07}
{"train_loss": 0.04486484453082085, "global_step": 255072, "epoch": 2865, "lr": 8.987208165313787e-07}
{"train_loss": 0.04384126676393024, "global_step": 255073, "epoch": 2865, "lr": 8.986113957911823e-07, "val_loss": 9.717549324035645, "train_action_mse_error": 11.953544616699219}
{"train_loss": 0.05309372767806053, "global_step": 255074, "epoch": 2866, "lr": 8.985019816520668e-07}
{"train_loss": 0.06667885929346085, "global_step": 255075, "epoch": 2866, "lr": 8.983925741140487e-07}
{"train_loss": 0.049125511199235916, "global_step": 255076, "epoch": 2866, "lr": 8.982831731771446e-07}
{"train_loss": 0.04197290167212486, "global_step": 255077, "epoch": 2866, "lr": 8.981737788413658e-07}
{"train_loss": 0.021955519914627075, "global_step": 255078, "epoch": 2866, "lr": 8.980643911067343e-07}
{"train_loss": 0.08093629032373428, "global_step": 255079, "epoch": 2866, "lr": 8.979550099732504e-07}
{"train_loss": 0.030359726399183273, "global_step": 255080, "epoch": 2866, "lr": 8.97845635440947e-07}
{"train_loss": 0.0957680493593216, "global_step": 255081, "epoch": 2866, "lr": 8.977362675098244e-07}
{"train_loss": 0.05555442348122597, "global_step": 255082, "epoch": 2866, "lr": 8.976269061799048e-07}
{"train_loss": 0.02316257171332836, "global_step": 255083, "epoch": 2866, "lr": 8.975175514511991e-07}
{"train_loss": 0.06682420521974564, "global_step": 255084, "epoch": 2866, "lr": 8.974082033237241e-07}
{"train_loss": 0.037072159349918365, "global_step": 255085, "epoch": 2866, "lr": 8.972988617974964e-07}
{"train_loss": 0.04013460874557495, "global_step": 255086, "epoch": 2866, "lr": 8.971895268725272e-07}
{"train_loss": 0.05923863500356674, "global_step": 255087, "epoch": 2866, "lr": 8.970801985488331e-07}
{"train_loss": 0.07079091668128967, "global_step": 255088, "epoch": 2866, "lr": 8.969708768264251e-07}
{"train_loss": 0.027149928733706474, "global_step": 255089, "epoch": 2866, "lr": 8.968615617053255e-07}
{"train_loss": 0.05245846509933472, "global_step": 255090, "epoch": 2866, "lr": 8.967522531855399e-07}
{"train_loss": 0.040387239307165146, "global_step": 255091, "epoch": 2866, "lr": 8.966429512670904e-07}
{"train_loss": 0.062314439564943314, "global_step": 255092, "epoch": 2866, "lr": 8.965336559499881e-07}
{"train_loss": 0.05182500183582306, "global_step": 255093, "epoch": 2866, "lr": 8.964243672342498e-07}
{"train_loss": 0.0180846955627203, "global_step": 255094, "epoch": 2866, "lr": 8.963150851198865e-07}
{"train_loss": 0.03777637332677841, "global_step": 255095, "epoch": 2866, "lr": 8.962058096069148e-07}
{"train_loss": 0.03548843041062355, "global_step": 255096, "epoch": 2866, "lr": 8.960965406953514e-07}
{"train_loss": 0.02705339528620243, "global_step": 255097, "epoch": 2866, "lr": 8.959872783852075e-07}
{"train_loss": 0.023029888048768044, "global_step": 255098, "epoch": 2866, "lr": 8.958780226764996e-07}
{"train_loss": 0.02736297994852066, "global_step": 255099, "epoch": 2866, "lr": 8.957687735692443e-07}
{"train_loss": 0.033241547644138336, "global_step": 255100, "epoch": 2866, "lr": 8.95659531063453e-07}
{"train_loss": 0.031562674790620804, "global_step": 255101, "epoch": 2866, "lr": 8.955502951591477e-07}
{"train_loss": 0.04505152627825737, "global_step": 255102, "epoch": 2866, "lr": 8.954410658563339e-07}
{"train_loss": 0.07552427053451538, "global_step": 255103, "epoch": 2866, "lr": 8.953318431550228e-07}
{"train_loss": 0.046649497002363205, "global_step": 255104, "epoch": 2866, "lr": 8.952226270552422e-07}
{"train_loss": 0.010337894782423973, "global_step": 255105, "epoch": 2866, "lr": 8.951134175569975e-07}
{"train_loss": 0.04861923307180405, "global_step": 255106, "epoch": 2866, "lr": 8.95004214660311e-07}
{"train_loss": 0.02696850709617138, "global_step": 255107, "epoch": 2866, "lr": 8.948950183651827e-07}
{"train_loss": 0.04648740962147713, "global_step": 255108, "epoch": 2866, "lr": 8.947858286716459e-07}
{"train_loss": 0.05261598899960518, "global_step": 255109, "epoch": 2866, "lr": 8.946766455797062e-07}
{"train_loss": 0.02343752235174179, "global_step": 255110, "epoch": 2866, "lr": 8.945674690893802e-07}
{"train_loss": 0.02158094011247158, "global_step": 255111, "epoch": 2866, "lr": 8.944582992006733e-07}
{"train_loss": 0.04956936091184616, "global_step": 255112, "epoch": 2866, "lr": 8.943491359136135e-07}
{"train_loss": 0.03614885360002518, "global_step": 255113, "epoch": 2866, "lr": 8.942399792282064e-07}
{"train_loss": 0.046773266047239304, "global_step": 255114, "epoch": 2866, "lr": 8.941308291444794e-07}
{"train_loss": 0.08429846167564392, "global_step": 255115, "epoch": 2866, "lr": 8.940216856624272e-07}
{"train_loss": 0.03547653928399086, "global_step": 255116, "epoch": 2866, "lr": 8.939125487820832e-07}
{"train_loss": 0.08172913640737534, "global_step": 255117, "epoch": 2866, "lr": 8.938034185034472e-07}
{"train_loss": 0.044542498886585236, "global_step": 255118, "epoch": 2866, "lr": 8.93694294826547e-07}
{"train_loss": 0.0243163350969553, "global_step": 255119, "epoch": 2866, "lr": 8.935851777513826e-07}
{"train_loss": 0.054929666221141815, "global_step": 255120, "epoch": 2866, "lr": 8.934760672779874e-07}
{"train_loss": 0.05635267123579979, "global_step": 255121, "epoch": 2866, "lr": 8.933669634063613e-07}
{"train_loss": 0.08366353064775467, "global_step": 255122, "epoch": 2866, "lr": 8.93257866136521e-07}
{"train_loss": 0.052662212401628494, "global_step": 255123, "epoch": 2866, "lr": 8.931487754684831e-07}
{"train_loss": 0.04453422874212265, "global_step": 255124, "epoch": 2866, "lr": 8.930396914022643e-07}
{"train_loss": 0.03896750509738922, "global_step": 255125, "epoch": 2866, "lr": 8.929306139378813e-07}
{"train_loss": 0.0246402770280838, "global_step": 255126, "epoch": 2866, "lr": 8.92821543075345e-07}
{"train_loss": 0.04668961092829704, "global_step": 255127, "epoch": 2866, "lr": 8.927124788146668e-07}
{"train_loss": 0.04158145561814308, "global_step": 255128, "epoch": 2866, "lr": 8.926034211558687e-07}
{"train_loss": 0.06587662547826767, "global_step": 255129, "epoch": 2866, "lr": 8.924943700989564e-07}
{"train_loss": 0.0299840085208416, "global_step": 255130, "epoch": 2866, "lr": 8.923853256439519e-07}
{"train_loss": 0.07055855542421341, "global_step": 255131, "epoch": 2866, "lr": 8.922762877908719e-07}
{"train_loss": 0.038167644292116165, "global_step": 255132, "epoch": 2866, "lr": 8.921672565397221e-07}
{"train_loss": 0.056938476860523224, "global_step": 255133, "epoch": 2866, "lr": 8.920582318905246e-07}
{"train_loss": 0.0585063099861145, "global_step": 255134, "epoch": 2866, "lr": 8.91949213843285e-07}
{"train_loss": 0.035865891724824905, "global_step": 255135, "epoch": 2866, "lr": 8.91840202398031e-07}
{"train_loss": 0.04996979609131813, "global_step": 255136, "epoch": 2866, "lr": 8.917311975547682e-07}
{"train_loss": 0.027556611225008965, "global_step": 255137, "epoch": 2866, "lr": 8.916221993135188e-07}
{"train_loss": 0.04839623346924782, "global_step": 255138, "epoch": 2866, "lr": 8.915132076742827e-07}
{"train_loss": 0.07351889461278915, "global_step": 255139, "epoch": 2866, "lr": 8.914042226370933e-07}
{"train_loss": 0.0722656175494194, "global_step": 255140, "epoch": 2866, "lr": 8.912952442019506e-07}
{"train_loss": 0.044374436140060425, "global_step": 255141, "epoch": 2866, "lr": 8.911862723688768e-07}
{"train_loss": 0.02824166975915432, "global_step": 255142, "epoch": 2866, "lr": 8.910773071378831e-07}
{"train_loss": 0.03495863825082779, "global_step": 255143, "epoch": 2866, "lr": 8.909683485089859e-07}
{"train_loss": 0.037034060806035995, "global_step": 255144, "epoch": 2866, "lr": 8.908593964822021e-07}
{"train_loss": 0.025991518050432205, "global_step": 255145, "epoch": 2866, "lr": 8.907504510575371e-07}
{"train_loss": 0.021889809519052505, "global_step": 255146, "epoch": 2866, "lr": 8.906415122350187e-07}
{"train_loss": 0.03655020520091057, "global_step": 255147, "epoch": 2866, "lr": 8.905325800146469e-07}
{"train_loss": 0.03586115688085556, "global_step": 255148, "epoch": 2866, "lr": 8.904236543964551e-07}
{"train_loss": 0.030260875821113586, "global_step": 255149, "epoch": 2866, "lr": 8.903147353804375e-07}
{"train_loss": 0.014756557531654835, "global_step": 255150, "epoch": 2866, "lr": 8.902058229666277e-07}
{"train_loss": 0.020493343472480774, "global_step": 255151, "epoch": 2866, "lr": 8.900969171550256e-07}
{"train_loss": 0.0327412374317646, "global_step": 255152, "epoch": 2866, "lr": 8.899880179456532e-07}
{"train_loss": 0.06696652621030807, "global_step": 255153, "epoch": 2866, "lr": 8.898791253385163e-07}
{"train_loss": 0.03351923078298569, "global_step": 255154, "epoch": 2866, "lr": 8.897702393336483e-07}
{"train_loss": 0.07899321615695953, "global_step": 255155, "epoch": 2866, "lr": 8.896613599310433e-07}
{"train_loss": 0.03535627946257591, "global_step": 255156, "epoch": 2866, "lr": 8.895524871307292e-07}
{"train_loss": 0.05489974841475487, "global_step": 255157, "epoch": 2866, "lr": 8.894436209327118e-07}
{"train_loss": 0.04388564079999924, "global_step": 255158, "epoch": 2866, "lr": 8.893347613370129e-07}
{"train_loss": 0.04483405128121376, "global_step": 255159, "epoch": 2866, "lr": 8.892259083436494e-07}
{"train_loss": 0.05884362384676933, "global_step": 255160, "epoch": 2866, "lr": 8.891170619526213e-07}
{"train_loss": 0.06376530230045319, "global_step": 255161, "epoch": 2866, "lr": 8.890082221639617e-07}
{"train_loss": 0.0454363226785921, "global_step": 255162, "epoch": 2866, "lr": 8.888993889776709e-07, "val_loss": 9.57863712310791}
{"train_loss": 0.048645347356796265, "global_step": 255163, "epoch": 2867, "lr": 8.887905623937709e-07}
{"train_loss": 0.04069714993238449, "global_step": 255164, "epoch": 2867, "lr": 8.886817424122729e-07}
{"train_loss": 0.05605476722121239, "global_step": 255165, "epoch": 2867, "lr": 8.885729290331934e-07}
{"train_loss": 0.06666186451911926, "global_step": 255166, "epoch": 2867, "lr": 8.884641222565493e-07}
{"train_loss": 0.03335855156183243, "global_step": 255167, "epoch": 2867, "lr": 8.883553220823515e-07}
{"train_loss": 0.04182491824030876, "global_step": 255168, "epoch": 2867, "lr": 8.882465285106112e-07}
{"train_loss": 0.02300519309937954, "global_step": 255169, "epoch": 2867, "lr": 8.881377415413561e-07}
{"train_loss": 0.01978115551173687, "global_step": 255170, "epoch": 2867, "lr": 8.880289611745862e-07}
{"train_loss": 0.045029643923044205, "global_step": 255171, "epoch": 2867, "lr": 8.879201874103238e-07}
{"train_loss": 0.04965202137827873, "global_step": 255172, "epoch": 2867, "lr": 8.878114202485798e-07}
{"train_loss": 0.05321011319756508, "global_step": 255173, "epoch": 2867, "lr": 8.877026596893767e-07}
{"train_loss": 0.021948616951704025, "global_step": 255174, "epoch": 2867, "lr": 8.875939057327143e-07}
{"train_loss": 0.03782316669821739, "global_step": 255175, "epoch": 2867, "lr": 8.874851583786259e-07}
{"train_loss": 0.018424017354846, "global_step": 255176, "epoch": 2867, "lr": 8.873764176271115e-07}
{"train_loss": 0.06312208622694016, "global_step": 255177, "epoch": 2867, "lr": 8.872676834781934e-07}
{"train_loss": 0.030514877289533615, "global_step": 255178, "epoch": 2867, "lr": 8.871589559318771e-07}
{"train_loss": 0.08251266926527023, "global_step": 255179, "epoch": 2867, "lr": 8.870502349881904e-07}
{"train_loss": 0.02552480809390545, "global_step": 255180, "epoch": 2867, "lr": 8.869415206471443e-07}
{"train_loss": 0.04204488918185234, "global_step": 255181, "epoch": 2867, "lr": 8.868328129087389e-07}
{"train_loss": 0.08189889788627625, "global_step": 255182, "epoch": 2867, "lr": 8.86724111773013e-07}
{"train_loss": 0.05817097797989845, "global_step": 255183, "epoch": 2867, "lr": 8.866154172399555e-07}
{"train_loss": 0.08037291467189789, "global_step": 255184, "epoch": 2867, "lr": 8.865067293096052e-07}
{"train_loss": 0.027455061674118042, "global_step": 255185, "epoch": 2867, "lr": 8.863980479819567e-07}
{"train_loss": 0.025153595954179764, "global_step": 255186, "epoch": 2867, "lr": 8.862893732570376e-07}
{"train_loss": 0.02265515923500061, "global_step": 255187, "epoch": 2867, "lr": 8.861807051348592e-07}
{"train_loss": 0.039219122380018234, "global_step": 255188, "epoch": 2867, "lr": 8.860720436154324e-07}
{"train_loss": 0.07300276309251785, "global_step": 255189, "epoch": 2867, "lr": 8.859633886987739e-07}
{"train_loss": 0.030546491965651512, "global_step": 255190, "epoch": 2867, "lr": 8.858547403849061e-07}
{"train_loss": 0.024309569969773293, "global_step": 255191, "epoch": 2867, "lr": 8.857460986738287e-07}
{"train_loss": 0.06423188745975494, "global_step": 255192, "epoch": 2867, "lr": 8.856374635655695e-07}
{"train_loss": 0.053520046174526215, "global_step": 255193, "epoch": 2867, "lr": 8.855288350601343e-07}
{"train_loss": 0.06029682978987694, "global_step": 255194, "epoch": 2867, "lr": 8.854202131575395e-07}
{"train_loss": 0.01760963909327984, "global_step": 255195, "epoch": 2867, "lr": 8.853115978578075e-07}
{"train_loss": 0.018565597012639046, "global_step": 255196, "epoch": 2867, "lr": 8.852029891609381e-07}
{"train_loss": 0.049197595566511154, "global_step": 255197, "epoch": 2867, "lr": 8.850943870669648e-07}
{"train_loss": 0.045081477612257004, "global_step": 255198, "epoch": 2867, "lr": 8.849857915758819e-07}
{"train_loss": 0.04731535539031029, "global_step": 255199, "epoch": 2867, "lr": 8.848772026877228e-07}
{"train_loss": 0.07541577517986298, "global_step": 255200, "epoch": 2867, "lr": 8.847686204024875e-07}
{"train_loss": 0.0642194002866745, "global_step": 255201, "epoch": 2867, "lr": 8.84660044720198e-07}
{"train_loss": 0.056854814291000366, "global_step": 255202, "epoch": 2867, "lr": 8.845514756408657e-07}
{"train_loss": 0.10499703884124756, "global_step": 255203, "epoch": 2867, "lr": 8.844429131645126e-07}
{"train_loss": 0.06088157370686531, "global_step": 255204, "epoch": 2867, "lr": 8.843343572911389e-07}
{"train_loss": 0.06978190690279007, "global_step": 255205, "epoch": 2867, "lr": 8.842258080207777e-07}
{"train_loss": 0.01700090616941452, "global_step": 255206, "epoch": 2867, "lr": 8.841172653534235e-07}
{"train_loss": 0.06375689059495926, "global_step": 255207, "epoch": 2867, "lr": 8.840087292891097e-07}
{"train_loss": 0.0241069495677948, "global_step": 255208, "epoch": 2867, "lr": 8.839001998278362e-07}
{"train_loss": 0.03642149642109871, "global_step": 255209, "epoch": 2867, "lr": 8.837916769696308e-07}
{"train_loss": 0.019933559000492096, "global_step": 255210, "epoch": 2867, "lr": 8.836831607144935e-07}
{"train_loss": 0.05572854354977608, "global_step": 255211, "epoch": 2867, "lr": 8.835746510624465e-07}
{"train_loss": 0.03021291270852089, "global_step": 255212, "epoch": 2867, "lr": 8.834661480135065e-07}
{"train_loss": 0.05834199860692024, "global_step": 255213, "epoch": 2867, "lr": 8.8335765156769e-07}
{"train_loss": 0.04697517678141594, "global_step": 255214, "epoch": 2867, "lr": 8.832491617250027e-07}
{"train_loss": 0.05880022048950195, "global_step": 255215, "epoch": 2867, "lr": 8.831406784854668e-07}
{"train_loss": 0.06482701748609543, "global_step": 255216, "epoch": 2867, "lr": 8.830322018490877e-07}
{"train_loss": 0.0339750312268734, "global_step": 255217, "epoch": 2867, "lr": 8.829237318158934e-07}
{"train_loss": 0.07237498462200165, "global_step": 255218, "epoch": 2867, "lr": 8.828152683858892e-07}
{"train_loss": 0.03232120722532272, "global_step": 255219, "epoch": 2867, "lr": 8.82706811559092e-07}
{"train_loss": 0.03840910270810127, "global_step": 255220, "epoch": 2867, "lr": 8.825983613355182e-07}
{"train_loss": 0.04069971665740013, "global_step": 255221, "epoch": 2867, "lr": 8.824899177151735e-07}
{"train_loss": 0.044850319623947144, "global_step": 255222, "epoch": 2867, "lr": 8.823814806980857e-07}
{"train_loss": 0.007682880852371454, "global_step": 255223, "epoch": 2867, "lr": 8.822730502842602e-07}
{"train_loss": 0.0757216289639473, "global_step": 255224, "epoch": 2867, "lr": 8.821646264737138e-07}
{"train_loss": 0.07728694379329681, "global_step": 255225, "epoch": 2867, "lr": 8.820562092664631e-07}
{"train_loss": 0.045097511261701584, "global_step": 255226, "epoch": 2867, "lr": 8.819477986625246e-07}
{"train_loss": 0.026817642152309418, "global_step": 255227, "epoch": 2867, "lr": 8.818393946619041e-07}
{"train_loss": 0.01246652752161026, "global_step": 255228, "epoch": 2867, "lr": 8.817309972646237e-07}
{"train_loss": 0.036884721368551254, "global_step": 255229, "epoch": 2867, "lr": 8.816226064706945e-07}
{"train_loss": 0.0810873731970787, "global_step": 255230, "epoch": 2867, "lr": 8.815142222801331e-07}
{"train_loss": 0.022677632048726082, "global_step": 255231, "epoch": 2867, "lr": 8.814058446929562e-07}
{"train_loss": 0.02421608380973339, "global_step": 255232, "epoch": 2867, "lr": 8.812974737091695e-07}
{"train_loss": 0.02518404833972454, "global_step": 255233, "epoch": 2867, "lr": 8.811891093288005e-07}
{"train_loss": 0.046010613441467285, "global_step": 255234, "epoch": 2867, "lr": 8.810807515518549e-07}
{"train_loss": 0.02323925867676735, "global_step": 255235, "epoch": 2867, "lr": 8.809724003783492e-07}
{"train_loss": 0.06760279089212418, "global_step": 255236, "epoch": 2867, "lr": 8.808640558082948e-07}
{"train_loss": 0.04750566557049751, "global_step": 255237, "epoch": 2867, "lr": 8.807557178417137e-07}
{"train_loss": 0.033754561096429825, "global_step": 255238, "epoch": 2867, "lr": 8.806473864786113e-07}
{"train_loss": 0.04655817523598671, "global_step": 255239, "epoch": 2867, "lr": 8.805390617190157e-07}
{"train_loss": 0.05869758129119873, "global_step": 255240, "epoch": 2867, "lr": 8.804307435629267e-07}
{"train_loss": 0.03164414316415787, "global_step": 255241, "epoch": 2867, "lr": 8.803224320103664e-07}
{"train_loss": 0.036342937499284744, "global_step": 255242, "epoch": 2867, "lr": 8.802141270613462e-07}
{"train_loss": 0.013499006628990173, "global_step": 255243, "epoch": 2867, "lr": 8.801058287158881e-07}
{"train_loss": 0.04302959144115448, "global_step": 255244, "epoch": 2867, "lr": 8.799975369739921e-07}
{"train_loss": 0.020634660497307777, "global_step": 255245, "epoch": 2867, "lr": 8.798892518356916e-07}
{"train_loss": 0.03897666186094284, "global_step": 255246, "epoch": 2867, "lr": 8.797809733009809e-07}
{"train_loss": 0.047284819185733795, "global_step": 255247, "epoch": 2867, "lr": 8.796727013698936e-07}
{"train_loss": 0.05622510612010956, "global_step": 255248, "epoch": 2867, "lr": 8.795644360424348e-07}
{"train_loss": 0.05720360949635506, "global_step": 255249, "epoch": 2867, "lr": 8.794561773186161e-07}
{"train_loss": 0.020590122789144516, "global_step": 255250, "epoch": 2867, "lr": 8.793479251984593e-07}
{"train_loss": 0.044674543083090795, "global_step": 255251, "epoch": 2867, "lr": 8.792396796819757e-07, "val_loss": 9.656064987182617}
{"train_loss": 0.05603119730949402, "global_step": 255252, "epoch": 2868, "lr": 8.791314407691708e-07}
{"train_loss": 0.031129207462072372, "global_step": 255253, "epoch": 2868, "lr": 8.79023208460078e-07}
{"train_loss": 0.052507184445858, "global_step": 255254, "epoch": 2868, "lr": 8.789149827546972e-07}
{"train_loss": 0.04554618149995804, "global_step": 255255, "epoch": 2868, "lr": 8.788067636530506e-07}
{"train_loss": 0.026737047359347343, "global_step": 255256, "epoch": 2868, "lr": 8.786985511551493e-07}
{"train_loss": 0.03483375906944275, "global_step": 255257, "epoch": 2868, "lr": 8.7859034526101e-07}
{"train_loss": 0.046633392572402954, "global_step": 255258, "epoch": 2868, "lr": 8.784821459706438e-07}
{"train_loss": 0.020230863243341446, "global_step": 255259, "epoch": 2868, "lr": 8.783739532840618e-07}
{"train_loss": 0.03561275079846382, "global_step": 255260, "epoch": 2868, "lr": 8.782657672012917e-07}
{"train_loss": 0.06353655457496643, "global_step": 255261, "epoch": 2868, "lr": 8.781575877223336e-07}
{"train_loss": 0.044830549508333206, "global_step": 255262, "epoch": 2868, "lr": 8.780494148472095e-07}
{"train_loss": 0.03413008525967598, "global_step": 255263, "epoch": 2868, "lr": 8.779412485759364e-07}
{"train_loss": 0.0424923300743103, "global_step": 255264, "epoch": 2868, "lr": 8.778330889085252e-07}
{"train_loss": 0.039600592106580734, "global_step": 255265, "epoch": 2868, "lr": 8.777249358449868e-07}
{"train_loss": 0.06794475764036179, "global_step": 255266, "epoch": 2868, "lr": 8.776167893853381e-07}
{"train_loss": 0.0386444516479969, "global_step": 255267, "epoch": 2868, "lr": 8.775086495296014e-07}
{"train_loss": 0.0793672576546669, "global_step": 255268, "epoch": 2868, "lr": 8.77400516277782e-07}
{"train_loss": 0.018085671588778496, "global_step": 255269, "epoch": 2868, "lr": 8.772923896298968e-07}
{"train_loss": 0.029373835772275925, "global_step": 255270, "epoch": 2868, "lr": 8.771842695859622e-07}
{"train_loss": 0.03635590523481369, "global_step": 255271, "epoch": 2868, "lr": 8.770761561459895e-07}
{"train_loss": 0.04966099560260773, "global_step": 255272, "epoch": 2868, "lr": 8.769680493099952e-07}
{"train_loss": 0.039928536862134933, "global_step": 255273, "epoch": 2868, "lr": 8.768599490779961e-07}
{"train_loss": 0.06853917986154556, "global_step": 255274, "epoch": 2868, "lr": 8.767518554500032e-07}
{"train_loss": 0.03404178470373154, "global_step": 255275, "epoch": 2868, "lr": 8.766437684260332e-07}
{"train_loss": 0.03995683044195175, "global_step": 255276, "epoch": 2868, "lr": 8.765356880060971e-07}
{"train_loss": 0.06477247923612595, "global_step": 255277, "epoch": 2868, "lr": 8.764276141902172e-07}
{"train_loss": 0.011833551339805126, "global_step": 255278, "epoch": 2868, "lr": 8.763195469783991e-07}
{"train_loss": 0.04596627503633499, "global_step": 255279, "epoch": 2868, "lr": 8.762114863706649e-07}
{"train_loss": 0.01579584926366806, "global_step": 255280, "epoch": 2868, "lr": 8.761034323670204e-07}
{"train_loss": 0.009886263869702816, "global_step": 255281, "epoch": 2868, "lr": 8.759953849674873e-07}
{"train_loss": 0.06607100367546082, "global_step": 255282, "epoch": 2868, "lr": 8.758873441720772e-07}
{"train_loss": 0.051315099000930786, "global_step": 255283, "epoch": 2868, "lr": 8.757793099808065e-07}
{"train_loss": 0.023517843335866928, "global_step": 255284, "epoch": 2868, "lr": 8.756712823936864e-07}
{"train_loss": 0.039742812514305115, "global_step": 255285, "epoch": 2868, "lr": 8.755632614107334e-07}
{"train_loss": 0.07253032922744751, "global_step": 255286, "epoch": 2868, "lr": 8.754552470319644e-07}
{"train_loss": 0.054943401366472244, "global_step": 255287, "epoch": 2868, "lr": 8.753472392573903e-07}
{"train_loss": 0.04743894562125206, "global_step": 255288, "epoch": 2868, "lr": 8.752392380870278e-07}
{"train_loss": 0.06187695264816284, "global_step": 255289, "epoch": 2868, "lr": 8.75131243520888e-07}
{"train_loss": 0.02761879749596119, "global_step": 255290, "epoch": 2868, "lr": 8.750232555589932e-07}
{"train_loss": 0.042528122663497925, "global_step": 255291, "epoch": 2868, "lr": 8.749152742013488e-07}
{"train_loss": 0.029758207499980927, "global_step": 255292, "epoch": 2868, "lr": 8.748072994479773e-07}
{"train_loss": 0.025216959416866302, "global_step": 255293, "epoch": 2868, "lr": 8.746993312988838e-07}
{"train_loss": 0.0703481137752533, "global_step": 255294, "epoch": 2868, "lr": 8.745913697540909e-07}
{"train_loss": 0.028943216428160667, "global_step": 255295, "epoch": 2868, "lr": 8.74483414813615e-07}
{"train_loss": 0.03108297847211361, "global_step": 255296, "epoch": 2868, "lr": 8.743754664774617e-07}
{"train_loss": 0.021278642117977142, "global_step": 255297, "epoch": 2868, "lr": 8.742675247456477e-07}
{"train_loss": 0.036516815423965454, "global_step": 255298, "epoch": 2868, "lr": 8.741595896181953e-07}
{"train_loss": 0.06561644375324249, "global_step": 255299, "epoch": 2868, "lr": 8.740516610951043e-07}
{"train_loss": 0.04594644159078598, "global_step": 255300, "epoch": 2868, "lr": 8.739437391764082e-07}
{"train_loss": 0.051301032304763794, "global_step": 255301, "epoch": 2868, "lr": 8.738358238621014e-07}
{"train_loss": 0.05267468839883804, "global_step": 255302, "epoch": 2868, "lr": 8.73727915152217e-07}
{"train_loss": 0.05710087716579437, "global_step": 255303, "epoch": 2868, "lr": 8.736200130467553e-07}
{"train_loss": 0.057422615587711334, "global_step": 255304, "epoch": 2868, "lr": 8.735121175457383e-07}
{"train_loss": 0.04934200644493103, "global_step": 255305, "epoch": 2868, "lr": 8.734042286491828e-07}
{"train_loss": 0.03193507716059685, "global_step": 255306, "epoch": 2868, "lr": 8.732963463570942e-07}
{"train_loss": 0.020515523850917816, "global_step": 255307, "epoch": 2868, "lr": 8.731884706695004e-07}
{"train_loss": 0.04347836226224899, "global_step": 255308, "epoch": 2868, "lr": 8.730806015863957e-07}
{"train_loss": 0.057142093777656555, "global_step": 255309, "epoch": 2868, "lr": 8.729727391078191e-07}
{"train_loss": 0.0555589459836483, "global_step": 255310, "epoch": 2868, "lr": 8.728648832337649e-07}
{"train_loss": 0.01807943917810917, "global_step": 255311, "epoch": 2868, "lr": 8.727570339642554e-07}
{"train_loss": 0.0207215603441, "global_step": 255312, "epoch": 2868, "lr": 8.726491912993073e-07}
{"train_loss": 0.023993195965886116, "global_step": 255313, "epoch": 2868, "lr": 8.725413552389317e-07}
{"train_loss": 0.07994799315929413, "global_step": 255314, "epoch": 2868, "lr": 8.724335257831395e-07}
{"train_loss": 0.038665734231472015, "global_step": 255315, "epoch": 2868, "lr": 8.723257029319588e-07}
{"train_loss": 0.03640086576342583, "global_step": 255316, "epoch": 2868, "lr": 8.722178866853892e-07}
{"train_loss": 0.04117294400930405, "global_step": 255317, "epoch": 2868, "lr": 8.721100770434531e-07}
{"train_loss": 0.04183715581893921, "global_step": 255318, "epoch": 2868, "lr": 8.720022740061617e-07}
{"train_loss": 0.0781150832772255, "global_step": 255319, "epoch": 2868, "lr": 8.718944775735316e-07}
{"train_loss": 0.008337443694472313, "global_step": 255320, "epoch": 2868, "lr": 8.717866877455738e-07}
{"train_loss": 0.038637492805719376, "global_step": 255321, "epoch": 2868, "lr": 8.716789045223051e-07}
{"train_loss": 0.04431959614157677, "global_step": 255322, "epoch": 2868, "lr": 8.715711279037475e-07}
{"train_loss": 0.040966521948575974, "global_step": 255323, "epoch": 2868, "lr": 8.714633578899012e-07}
{"train_loss": 0.07468917965888977, "global_step": 255324, "epoch": 2868, "lr": 8.713555944807939e-07}
{"train_loss": 0.029483769088983536, "global_step": 255325, "epoch": 2868, "lr": 8.71247837676431e-07}
{"train_loss": 0.07125841826200485, "global_step": 255326, "epoch": 2868, "lr": 8.711400874768294e-07}
{"train_loss": 0.0838870108127594, "global_step": 255327, "epoch": 2868, "lr": 8.710323438820056e-07}
{"train_loss": 0.026180049404501915, "global_step": 255328, "epoch": 2868, "lr": 8.709246068919708e-07}
{"train_loss": 0.04574781656265259, "global_step": 255329, "epoch": 2868, "lr": 8.70816876506747e-07}
{"train_loss": 0.04129642993211746, "global_step": 255330, "epoch": 2868, "lr": 8.707091527263345e-07}
{"train_loss": 0.04333817586302757, "global_step": 255331, "epoch": 2868, "lr": 8.706014355507663e-07}
{"train_loss": 0.04158325120806694, "global_step": 255332, "epoch": 2868, "lr": 8.704937249800371e-07}
{"train_loss": 0.011810480616986752, "global_step": 255333, "epoch": 2868, "lr": 8.703860210141801e-07}
{"train_loss": 0.021087920293211937, "global_step": 255334, "epoch": 2868, "lr": 8.702783236531953e-07}
{"train_loss": 0.01921461522579193, "global_step": 255335, "epoch": 2868, "lr": 8.701706328970993e-07}
{"train_loss": 0.04077411815524101, "global_step": 255336, "epoch": 2868, "lr": 8.7006294874592e-07}
{"train_loss": 0.03123561665415764, "global_step": 255337, "epoch": 2868, "lr": 8.699552711996517e-07}
{"train_loss": 0.059906862676143646, "global_step": 255338, "epoch": 2868, "lr": 8.698476002583223e-07}
{"train_loss": 0.05408134683966637, "global_step": 255339, "epoch": 2868, "lr": 8.697399359219483e-07}
{"train_loss": 0.04226728799763355, "global_step": 255340, "epoch": 2868, "lr": 8.696322781905297e-07, "val_loss": 9.60517406463623}
{"train_loss": 0.05253944173455238, "global_step": 255341, "epoch": 2869, "lr": 8.695246270641e-07}
{"train_loss": 0.014923491515219212, "global_step": 255342, "epoch": 2869, "lr": 8.69416982542659e-07}
{"train_loss": 0.03458932787179947, "global_step": 255343, "epoch": 2869, "lr": 8.69309344626229e-07}
{"train_loss": 0.056969475001096725, "global_step": 255344, "epoch": 2869, "lr": 8.692017133148156e-07}
{"train_loss": 0.06194241717457771, "global_step": 255345, "epoch": 2869, "lr": 8.690940886084464e-07}
{"train_loss": 0.03699255362153053, "global_step": 255346, "epoch": 2869, "lr": 8.689864705071216e-07}
{"train_loss": 0.04268556088209152, "global_step": 255347, "epoch": 2869, "lr": 8.688788590108687e-07}
{"train_loss": 0.03445371612906456, "global_step": 255348, "epoch": 2869, "lr": 8.687712541196879e-07}
{"train_loss": 0.07883476465940475, "global_step": 255349, "epoch": 2869, "lr": 8.686636558336126e-07}
{"train_loss": 0.02269001305103302, "global_step": 255350, "epoch": 2869, "lr": 8.685560641526424e-07}
{"train_loss": 0.05523771420121193, "global_step": 255351, "epoch": 2869, "lr": 8.684484790767943e-07}
{"train_loss": 0.06418564170598984, "global_step": 255352, "epoch": 2869, "lr": 8.683409006060849e-07}
{"train_loss": 0.02559059113264084, "global_step": 255353, "epoch": 2869, "lr": 8.682333287405308e-07}
{"train_loss": 0.051183078438043594, "global_step": 255354, "epoch": 2869, "lr": 8.681257634801376e-07}
{"train_loss": 0.024573879316449165, "global_step": 255355, "epoch": 2869, "lr": 8.68018204824933e-07}
{"train_loss": 0.04787096381187439, "global_step": 255356, "epoch": 2869, "lr": 8.679106527749226e-07}
{"train_loss": 0.060980308800935745, "global_step": 255357, "epoch": 2869, "lr": 8.678031073301173e-07}
{"train_loss": 0.041095029562711716, "global_step": 255358, "epoch": 2869, "lr": 8.676955684905452e-07}
{"train_loss": 0.030315298587083817, "global_step": 255359, "epoch": 2869, "lr": 8.675880362562062e-07}
{"train_loss": 0.01820892095565796, "global_step": 255360, "epoch": 2869, "lr": 8.674805106271277e-07}
{"train_loss": 0.06806737184524536, "global_step": 255361, "epoch": 2869, "lr": 8.673729916033102e-07}
{"train_loss": 0.039453789591789246, "global_step": 255362, "epoch": 2869, "lr": 8.672654791847812e-07}
{"train_loss": 0.041106414049863815, "global_step": 255363, "epoch": 2869, "lr": 8.671579733715463e-07}
{"train_loss": 0.05086379125714302, "global_step": 255364, "epoch": 2869, "lr": 8.670504741636277e-07}
{"train_loss": 0.059722330421209335, "global_step": 255365, "epoch": 2869, "lr": 8.669429815610309e-07}
{"train_loss": 0.030990052968263626, "global_step": 255366, "epoch": 2869, "lr": 8.66835495563778e-07}
{"train_loss": 0.050897594541311264, "global_step": 255367, "epoch": 2869, "lr": 8.667280161718804e-07}
{"train_loss": 0.04261605069041252, "global_step": 255368, "epoch": 2869, "lr": 8.666205433853491e-07}
{"train_loss": 0.024168608710169792, "global_step": 255369, "epoch": 2869, "lr": 8.665130772042063e-07}
{"train_loss": 0.04387032613158226, "global_step": 255370, "epoch": 2869, "lr": 8.664056176284574e-07}
{"train_loss": 0.025097046047449112, "global_step": 255371, "epoch": 2869, "lr": 8.662981646581248e-07}
{"train_loss": 0.04821775108575821, "global_step": 255372, "epoch": 2869, "lr": 8.661907182932194e-07}
{"train_loss": 0.06124979630112648, "global_step": 255373, "epoch": 2869, "lr": 8.66083278533747e-07}
{"train_loss": 0.017261594533920288, "global_step": 255374, "epoch": 2869, "lr": 8.659758453797406e-07}
{"train_loss": 0.035052340477705, "global_step": 255375, "epoch": 2869, "lr": 8.658684188312005e-07}
{"train_loss": 0.04557541385293007, "global_step": 255376, "epoch": 2869, "lr": 8.657609988881432e-07}
{"train_loss": 0.05369594693183899, "global_step": 255377, "epoch": 2869, "lr": 8.656535855505965e-07}
{"train_loss": 0.01736553944647312, "global_step": 255378, "epoch": 2869, "lr": 8.655461788185548e-07}
{"train_loss": 0.05422500520944595, "global_step": 255379, "epoch": 2869, "lr": 8.654387786920404e-07}
{"train_loss": 0.04803153872489929, "global_step": 255380, "epoch": 2869, "lr": 8.653313851710698e-07}
{"train_loss": 0.02811851166188717, "global_step": 255381, "epoch": 2869, "lr": 8.652239982556654e-07}
{"train_loss": 0.03604349493980408, "global_step": 255382, "epoch": 2869, "lr": 8.651166179458215e-07}
{"train_loss": 0.03924413025379181, "global_step": 255383, "epoch": 2869, "lr": 8.650092442415714e-07}
{"train_loss": 0.02239909954369068, "global_step": 255384, "epoch": 2869, "lr": 8.649018771429151e-07}
{"train_loss": 0.06051575765013695, "global_step": 255385, "epoch": 2869, "lr": 8.647945166498805e-07}
{"train_loss": 0.036786798387765884, "global_step": 255386, "epoch": 2869, "lr": 8.646871627624731e-07}
{"train_loss": 0.027854613959789276, "global_step": 255387, "epoch": 2869, "lr": 8.645798154807094e-07}
{"train_loss": 0.05248871445655823, "global_step": 255388, "epoch": 2869, "lr": 8.644724748046007e-07}
{"train_loss": 0.036038413643836975, "global_step": 255389, "epoch": 2869, "lr": 8.643651407341691e-07}
{"train_loss": 0.03477523848414421, "global_step": 255390, "epoch": 2869, "lr": 8.642578132694201e-07}
{"train_loss": 0.06721759587526321, "global_step": 255391, "epoch": 2869, "lr": 8.641504924103816e-07}
{"train_loss": 0.0461001917719841, "global_step": 255392, "epoch": 2869, "lr": 8.640431781570479e-07}
{"train_loss": 0.07431710511445999, "global_step": 255393, "epoch": 2869, "lr": 8.639358705094525e-07}
{"train_loss": 0.061097852885723114, "global_step": 255394, "epoch": 2869, "lr": 8.638285694676007e-07}
{"train_loss": 0.04612389951944351, "global_step": 255395, "epoch": 2869, "lr": 8.637212750315038e-07}
{"train_loss": 0.05991433560848236, "global_step": 255396, "epoch": 2869, "lr": 8.636139872011895e-07}
{"train_loss": 0.03044290840625763, "global_step": 255397, "epoch": 2869, "lr": 8.635067059766522e-07}
{"train_loss": 0.04693945124745369, "global_step": 255398, "epoch": 2869, "lr": 8.633994313579252e-07}
{"train_loss": 0.02526714839041233, "global_step": 255399, "epoch": 2869, "lr": 8.632921633450142e-07}
{"train_loss": 0.04446035623550415, "global_step": 255400, "epoch": 2869, "lr": 8.631849019379357e-07}
{"train_loss": 0.07170472294092178, "global_step": 255401, "epoch": 2869, "lr": 8.630776471367008e-07}
{"train_loss": 0.052190497517585754, "global_step": 255402, "epoch": 2869, "lr": 8.629703989413263e-07}
{"train_loss": 0.025900956243276596, "global_step": 255403, "epoch": 2869, "lr": 8.628631573518287e-07}
{"train_loss": 0.06584864854812622, "global_step": 255404, "epoch": 2869, "lr": 8.627559223682191e-07}
{"train_loss": 0.050598181784152985, "global_step": 255405, "epoch": 2869, "lr": 8.626486939905087e-07}
{"train_loss": 0.041669707745313644, "global_step": 255406, "epoch": 2869, "lr": 8.625414722187253e-07}
{"train_loss": 0.03864239901304245, "global_step": 255407, "epoch": 2869, "lr": 8.624342570528687e-07}
{"train_loss": 0.04310625419020653, "global_step": 255408, "epoch": 2869, "lr": 8.623270484929613e-07}
{"train_loss": 0.051516324281692505, "global_step": 255409, "epoch": 2869, "lr": 8.622198465390142e-07}
{"train_loss": 0.0163718331605196, "global_step": 255410, "epoch": 2869, "lr": 8.621126511910382e-07}
{"train_loss": 0.06384336203336716, "global_step": 255411, "epoch": 2869, "lr": 8.620054624490615e-07}
{"train_loss": 0.04158162325620651, "global_step": 255412, "epoch": 2869, "lr": 8.618982803130782e-07}
{"train_loss": 0.04768922179937363, "global_step": 255413, "epoch": 2869, "lr": 8.617911047831217e-07}
{"train_loss": 0.03745252266526222, "global_step": 255414, "epoch": 2869, "lr": 8.616839358591977e-07}
{"train_loss": 0.02626737393438816, "global_step": 255415, "epoch": 2869, "lr": 8.615767735413171e-07}
{"train_loss": 0.03667346388101578, "global_step": 255416, "epoch": 2869, "lr": 8.614696178295023e-07}
{"train_loss": 0.07466214150190353, "global_step": 255417, "epoch": 2869, "lr": 8.613624687237643e-07}
{"train_loss": 0.04493660479784012, "global_step": 255418, "epoch": 2869, "lr": 8.612553262241141e-07}
{"train_loss": 0.039059530943632126, "global_step": 255419, "epoch": 2869, "lr": 8.61148190330574e-07}
{"train_loss": 0.04501968249678612, "global_step": 255420, "epoch": 2869, "lr": 8.610410610431496e-07}
{"train_loss": 0.04942217469215393, "global_step": 255421, "epoch": 2869, "lr": 8.609339383618575e-07}
{"train_loss": 0.07022351771593094, "global_step": 255422, "epoch": 2869, "lr": 8.608268222867144e-07}
{"train_loss": 0.03189373388886452, "global_step": 255423, "epoch": 2869, "lr": 8.607197128177425e-07}
{"train_loss": 0.06150206923484802, "global_step": 255424, "epoch": 2869, "lr": 8.606126099549361e-07}
{"train_loss": 0.06551467627286911, "global_step": 255425, "epoch": 2869, "lr": 8.605055136983342e-07}
{"train_loss": 0.03951140493154526, "global_step": 255426, "epoch": 2869, "lr": 8.603984240479257e-07}
{"train_loss": 0.06492190808057785, "global_step": 255427, "epoch": 2869, "lr": 8.602913410037439e-07}
{"train_loss": 0.09146463871002197, "global_step": 255428, "epoch": 2869, "lr": 8.601842645657998e-07}
{"train_loss": 0.04518137902565551, "global_step": 255429, "epoch": 2869, "lr": 8.600771947340991e-07, "val_loss": 9.652055740356445}
{"train_loss": 0.054704178124666214, "global_step": 255430, "epoch": 2870, "lr": 8.599701315086639e-07}
{"train_loss": 0.012514090165495872, "global_step": 255431, "epoch": 2870, "lr": 8.598630748895053e-07}
{"train_loss": 0.06526947021484375, "global_step": 255432, "epoch": 2870, "lr": 8.597560248766457e-07}
{"train_loss": 0.025737643241882324, "global_step": 255433, "epoch": 2870, "lr": 8.596489814700848e-07}
{"train_loss": 0.04970366135239601, "global_step": 255434, "epoch": 2870, "lr": 8.595419446698505e-07}
{"train_loss": 0.05578901246190071, "global_step": 255435, "epoch": 2870, "lr": 8.594349144759483e-07}
{"train_loss": 0.02465209737420082, "global_step": 255436, "epoch": 2870, "lr": 8.593278908884006e-07}
{"train_loss": 0.03967825323343277, "global_step": 255437, "epoch": 2870, "lr": 8.592208739072127e-07}
{"train_loss": 0.0523468442261219, "global_step": 255438, "epoch": 2870, "lr": 8.59113863532407e-07}
{"train_loss": 0.08195658773183823, "global_step": 255439, "epoch": 2870, "lr": 8.590068597639944e-07}
{"train_loss": 0.055900197476148605, "global_step": 255440, "epoch": 2870, "lr": 8.588998626019917e-07}
{"train_loss": 0.046039875596761703, "global_step": 255441, "epoch": 2870, "lr": 8.587928720464045e-07}
{"train_loss": 0.05215878039598465, "global_step": 255442, "epoch": 2870, "lr": 8.586858880972603e-07}
{"train_loss": 0.08433011174201965, "global_step": 255443, "epoch": 2870, "lr": 8.585789107545595e-07}
{"train_loss": 0.03546835482120514, "global_step": 255444, "epoch": 2870, "lr": 8.584719400183295e-07}
{"train_loss": 0.03133207932114601, "global_step": 255445, "epoch": 2870, "lr": 8.583649758885814e-07}
{"train_loss": 0.04378699138760567, "global_step": 255446, "epoch": 2870, "lr": 8.58258018365321e-07}
{"train_loss": 0.029161738231778145, "global_step": 255447, "epoch": 2870, "lr": 8.581510674485704e-07}
{"train_loss": 0.019739845767617226, "global_step": 255448, "epoch": 2870, "lr": 8.580441231383408e-07}
{"train_loss": 0.06175004318356514, "global_step": 255449, "epoch": 2870, "lr": 8.579371854346541e-07}
{"train_loss": 0.06272123754024506, "global_step": 255450, "epoch": 2870, "lr": 8.578302543375106e-07}
{"train_loss": 0.0282694473862648, "global_step": 255451, "epoch": 2870, "lr": 8.577233298469379e-07}
{"train_loss": 0.017418643459677696, "global_step": 255452, "epoch": 2870, "lr": 8.576164119629415e-07}
{"train_loss": 0.0240024384111166, "global_step": 255453, "epoch": 2870, "lr": 8.575095006855494e-07}
{"train_loss": 0.035453785210847855, "global_step": 255454, "epoch": 2870, "lr": 8.574025960147558e-07}
{"train_loss": 0.03721432387828827, "global_step": 255455, "epoch": 2870, "lr": 8.572956979505886e-07}
{"train_loss": 0.05132736638188362, "global_step": 255456, "epoch": 2870, "lr": 8.571888064930589e-07}
{"train_loss": 0.03874501958489418, "global_step": 255457, "epoch": 2870, "lr": 8.570819216421833e-07}
{"train_loss": 0.04607035592198372, "global_step": 255458, "epoch": 2870, "lr": 8.569750433979728e-07}
{"train_loss": 0.038948506116867065, "global_step": 255459, "epoch": 2870, "lr": 8.568681717604443e-07}
{"train_loss": 0.03987237066030502, "global_step": 255460, "epoch": 2870, "lr": 8.567613067296088e-07}
{"train_loss": 0.020214295014739037, "global_step": 255461, "epoch": 2870, "lr": 8.566544483054828e-07}
{"train_loss": 0.04878651723265648, "global_step": 255462, "epoch": 2870, "lr": 8.565475964880831e-07}
{"train_loss": 0.04183894395828247, "global_step": 255463, "epoch": 2870, "lr": 8.564407512774208e-07}
{"train_loss": 0.030892444774508476, "global_step": 255464, "epoch": 2870, "lr": 8.563339126735071e-07}
{"train_loss": 0.05508097633719444, "global_step": 255465, "epoch": 2870, "lr": 8.562270806763584e-07}
{"train_loss": 0.02500511147081852, "global_step": 255466, "epoch": 2870, "lr": 8.561202552860026e-07}
{"train_loss": 0.03300756588578224, "global_step": 255467, "epoch": 2870, "lr": 8.560134365024342e-07}
{"train_loss": 0.05158562585711479, "global_step": 255468, "epoch": 2870, "lr": 8.559066243256753e-07}
{"train_loss": 0.02410198375582695, "global_step": 255469, "epoch": 2870, "lr": 8.557998187557426e-07}
{"train_loss": 0.021646669134497643, "global_step": 255470, "epoch": 2870, "lr": 8.556930197926527e-07}
{"train_loss": 0.04185466468334198, "global_step": 255471, "epoch": 2870, "lr": 8.555862274364057e-07}
{"train_loss": 0.05973677709698677, "global_step": 255472, "epoch": 2870, "lr": 8.554794416870348e-07}
{"train_loss": 0.061458587646484375, "global_step": 255473, "epoch": 2870, "lr": 8.553726625445458e-07}
{"train_loss": 0.04455848038196564, "global_step": 255474, "epoch": 2870, "lr": 8.552658900089495e-07}
{"train_loss": 0.043165359646081924, "global_step": 255475, "epoch": 2870, "lr": 8.551591240802626e-07}
{"train_loss": 0.021744051948189735, "global_step": 255476, "epoch": 2870, "lr": 8.550523647585074e-07}
{"train_loss": 0.10564574599266052, "global_step": 255477, "epoch": 2870, "lr": 8.549456120436839e-07}
{"train_loss": 0.04661114141345024, "global_step": 255478, "epoch": 2870, "lr": 8.548388659358197e-07}
{"train_loss": 0.02665194310247898, "global_step": 255479, "epoch": 2870, "lr": 8.547321264349151e-07}
{"train_loss": 0.09697999060153961, "global_step": 255480, "epoch": 2870, "lr": 8.546253935410032e-07}
{"train_loss": 0.06589984893798828, "global_step": 255481, "epoch": 2870, "lr": 8.545186672540783e-07}
{"train_loss": 0.05268506705760956, "global_step": 255482, "epoch": 2870, "lr": 8.54411947574174e-07}
{"train_loss": 0.05699737370014191, "global_step": 255483, "epoch": 2870, "lr": 8.543052345012902e-07}
{"train_loss": 0.030405623838305473, "global_step": 255484, "epoch": 2870, "lr": 8.541985280354436e-07}
{"train_loss": 0.03298712894320488, "global_step": 255485, "epoch": 2870, "lr": 8.540918281766564e-07}
{"train_loss": 0.027396684512495995, "global_step": 255486, "epoch": 2870, "lr": 8.539851349249339e-07}
{"train_loss": 0.018483372405171394, "global_step": 255487, "epoch": 2870, "lr": 8.538784482802931e-07}
{"train_loss": 0.032792896032333374, "global_step": 255488, "epoch": 2870, "lr": 8.537717682427504e-07}
{"train_loss": 0.0271349735558033, "global_step": 255489, "epoch": 2870, "lr": 8.536650948123226e-07}
{"train_loss": 0.023787178099155426, "global_step": 255490, "epoch": 2870, "lr": 8.535584279890152e-07}
{"train_loss": 0.030174344778060913, "global_step": 255491, "epoch": 2870, "lr": 8.534517677728505e-07}
{"train_loss": 0.02417822927236557, "global_step": 255492, "epoch": 2870, "lr": 8.533451141638394e-07}
{"train_loss": 0.0524803102016449, "global_step": 255493, "epoch": 2870, "lr": 8.532384671619986e-07}
{"train_loss": 0.05416323244571686, "global_step": 255494, "epoch": 2870, "lr": 8.531318267673394e-07}
{"train_loss": 0.08450723439455032, "global_step": 255495, "epoch": 2870, "lr": 8.530251929798782e-07}
{"train_loss": 0.07769083231687546, "global_step": 255496, "epoch": 2870, "lr": 8.529185657996263e-07}
{"train_loss": 0.04078558832406998, "global_step": 255497, "epoch": 2870, "lr": 8.528119452266059e-07}
{"train_loss": 0.0672784075140953, "global_step": 255498, "epoch": 2870, "lr": 8.527053312608169e-07}
{"train_loss": 0.03648076206445694, "global_step": 255499, "epoch": 2870, "lr": 8.525987239022926e-07}
{"train_loss": 0.026870673522353172, "global_step": 255500, "epoch": 2870, "lr": 8.52492123151033e-07}
{"train_loss": 0.05273556709289551, "global_step": 255501, "epoch": 2870, "lr": 8.523855290070548e-07}
{"train_loss": 0.05309367552399635, "global_step": 255502, "epoch": 2870, "lr": 8.522789414703747e-07}
{"train_loss": 0.044465210288763046, "global_step": 255503, "epoch": 2870, "lr": 8.521723605410092e-07}
{"train_loss": 0.05087176337838173, "global_step": 255504, "epoch": 2870, "lr": 8.520657862189696e-07}
{"train_loss": 0.057952750474214554, "global_step": 255505, "epoch": 2870, "lr": 8.519592185042668e-07}
{"train_loss": 0.04749506711959839, "global_step": 255506, "epoch": 2870, "lr": 8.518526573969288e-07}
{"train_loss": 0.042100585997104645, "global_step": 255507, "epoch": 2870, "lr": 8.517461028969497e-07}
{"train_loss": 0.04013075307011604, "global_step": 255508, "epoch": 2870, "lr": 8.516395550043632e-07}
{"train_loss": 0.023152697831392288, "global_step": 255509, "epoch": 2870, "lr": 8.51533013719169e-07}
{"train_loss": 0.04104242101311684, "global_step": 255510, "epoch": 2870, "lr": 8.514264790413895e-07}
{"train_loss": 0.03870200365781784, "global_step": 255511, "epoch": 2870, "lr": 8.513199509710302e-07}
{"train_loss": 0.03962528333067894, "global_step": 255512, "epoch": 2870, "lr": 8.512134295081187e-07}
{"train_loss": 0.02541397139430046, "global_step": 255513, "epoch": 2870, "lr": 8.511069146526607e-07}
{"train_loss": 0.04186118021607399, "global_step": 255514, "epoch": 2870, "lr": 8.510004064046783e-07}
{"train_loss": 0.0753825306892395, "global_step": 255515, "epoch": 2870, "lr": 8.508939047641717e-07}
{"train_loss": 0.04452293738722801, "global_step": 255516, "epoch": 2870, "lr": 8.507874097311685e-07}
{"train_loss": 0.03333016112446785, "global_step": 255517, "epoch": 2870, "lr": 8.506809213056688e-07}
{"train_loss": 0.04375962345871363, "global_step": 255518, "epoch": 2870, "lr": 8.505744394877058e-07, "val_loss": 9.523589134216309, "train_action_mse_error": 20.916656494140625}
{"train_loss": 0.047525469213724136, "global_step": 255519, "epoch": 2871, "lr": 8.504679642772795e-07}
{"train_loss": 0.04496299475431442, "global_step": 255520, "epoch": 2871, "lr": 8.503614956744121e-07}
{"train_loss": 0.05411907657980919, "global_step": 255521, "epoch": 2871, "lr": 8.502550336791093e-07}
{"train_loss": 0.041868310421705246, "global_step": 255522, "epoch": 2871, "lr": 8.501485782913987e-07}
{"train_loss": 0.01826840452849865, "global_step": 255523, "epoch": 2871, "lr": 8.500421295112804e-07}
{"train_loss": 0.045627716928720474, "global_step": 255524, "epoch": 2871, "lr": 8.499356873387765e-07}
{"train_loss": 0.03789358586072922, "global_step": 255525, "epoch": 2871, "lr": 8.498292517738981e-07}
{"train_loss": 0.07078748196363449, "global_step": 255526, "epoch": 2871, "lr": 8.49722822816662e-07}
{"train_loss": 0.019804703071713448, "global_step": 255527, "epoch": 2871, "lr": 8.496164004670847e-07}
{"train_loss": 0.07260604947805405, "global_step": 255528, "epoch": 2871, "lr": 8.49509984725172e-07}
{"train_loss": 0.029614828526973724, "global_step": 255529, "epoch": 2871, "lr": 8.494035755909513e-07}
{"train_loss": 0.0318511426448822, "global_step": 255530, "epoch": 2871, "lr": 8.492971730644173e-07}
{"train_loss": 0.05288340896368027, "global_step": 255531, "epoch": 2871, "lr": 8.491907771456087e-07}
{"train_loss": 0.014033692888915539, "global_step": 255532, "epoch": 2871, "lr": 8.4908438783452e-07}
{"train_loss": 0.04989118501543999, "global_step": 255533, "epoch": 2871, "lr": 8.489780051311735e-07}
{"train_loss": 0.06327172368764877, "global_step": 255534, "epoch": 2871, "lr": 8.488716290355858e-07}
{"train_loss": 0.062193162739276886, "global_step": 255535, "epoch": 2871, "lr": 8.48765259547768e-07}
{"train_loss": 0.041534312069416046, "global_step": 255536, "epoch": 2871, "lr": 8.486588966677311e-07}
{"train_loss": 0.04435334354639053, "global_step": 255537, "epoch": 2871, "lr": 8.485525403954975e-07}
{"train_loss": 0.031159911304712296, "global_step": 255538, "epoch": 2871, "lr": 8.484461907310726e-07}
{"train_loss": 0.04059828072786331, "global_step": 255539, "epoch": 2871, "lr": 8.483398476744731e-07}
{"train_loss": 0.030860159546136856, "global_step": 255540, "epoch": 2871, "lr": 8.482335112257211e-07}
{"train_loss": 0.04195748642086983, "global_step": 255541, "epoch": 2871, "lr": 8.481271813848224e-07}
{"train_loss": 0.06025129184126854, "global_step": 255542, "epoch": 2871, "lr": 8.480208581517934e-07}
{"train_loss": 0.09172888845205307, "global_step": 255543, "epoch": 2871, "lr": 8.47914541526651e-07}
{"train_loss": 0.09283611923456192, "global_step": 255544, "epoch": 2871, "lr": 8.47808231509406e-07}
{"train_loss": 0.06054406613111496, "global_step": 255545, "epoch": 2871, "lr": 8.477019281000698e-07}
{"train_loss": 0.0717267096042633, "global_step": 255546, "epoch": 2871, "lr": 8.475956312986699e-07}
{"train_loss": 0.06379581987857819, "global_step": 255547, "epoch": 2871, "lr": 8.474893411052065e-07}
{"train_loss": 0.03433011472225189, "global_step": 255548, "epoch": 2871, "lr": 8.473830575197017e-07}
{"train_loss": 0.02052912674844265, "global_step": 255549, "epoch": 2871, "lr": 8.472767805421611e-07}
{"train_loss": 0.022992286831140518, "global_step": 255550, "epoch": 2871, "lr": 8.471705101726124e-07}
{"train_loss": 0.05864480510354042, "global_step": 255551, "epoch": 2871, "lr": 8.470642464110557e-07}
{"train_loss": 0.026303164660930634, "global_step": 255552, "epoch": 2871, "lr": 8.469579892575185e-07}
{"train_loss": 0.03877653554081917, "global_step": 255553, "epoch": 2871, "lr": 8.468517387120067e-07}
{"train_loss": 0.057037994265556335, "global_step": 255554, "epoch": 2871, "lr": 8.467454947745368e-07}
{"train_loss": 0.03335395082831383, "global_step": 255555, "epoch": 2871, "lr": 8.466392574451198e-07}
{"train_loss": 0.019903043285012245, "global_step": 255556, "epoch": 2871, "lr": 8.465330267237781e-07}
{"train_loss": 0.027426017448306084, "global_step": 255557, "epoch": 2871, "lr": 8.464268026105171e-07}
{"train_loss": 0.07563256472349167, "global_step": 255558, "epoch": 2871, "lr": 8.46320585105359e-07}
{"train_loss": 0.05380712449550629, "global_step": 255559, "epoch": 2871, "lr": 8.462143742083039e-07}
{"train_loss": 0.01498859841376543, "global_step": 255560, "epoch": 2871, "lr": 8.461081699193907e-07}
{"train_loss": 0.03217659518122673, "global_step": 255561, "epoch": 2871, "lr": 8.460019722386136e-07}
{"train_loss": 0.04069796949625015, "global_step": 255562, "epoch": 2871, "lr": 8.45895781165984e-07}
{"train_loss": 0.05558403208851814, "global_step": 255563, "epoch": 2871, "lr": 8.45789596701535e-07}
{"train_loss": 0.046229518949985504, "global_step": 255564, "epoch": 2871, "lr": 8.456834188452667e-07}
{"train_loss": 0.016648048534989357, "global_step": 255565, "epoch": 2871, "lr": 8.455772475971957e-07}
{"train_loss": 0.026893023401498795, "global_step": 255566, "epoch": 2871, "lr": 8.454710829573387e-07}
{"train_loss": 0.015115870162844658, "global_step": 255567, "epoch": 2871, "lr": 8.453649249257123e-07}
{"train_loss": 0.1270090788602829, "global_step": 255568, "epoch": 2871, "lr": 8.452587735023221e-07}
{"train_loss": 0.03289322555065155, "global_step": 255569, "epoch": 2871, "lr": 8.451526286871903e-07}
{"train_loss": 0.05547217279672623, "global_step": 255570, "epoch": 2871, "lr": 8.45046490480328e-07}
{"train_loss": 0.02733653411269188, "global_step": 255571, "epoch": 2871, "lr": 8.449403588817517e-07}
{"train_loss": 0.028804531320929527, "global_step": 255572, "epoch": 2871, "lr": 8.448342338914728e-07}
{"train_loss": 0.0754595398902893, "global_step": 255573, "epoch": 2871, "lr": 8.447281155095077e-07}
{"train_loss": 0.027785886079072952, "global_step": 255574, "epoch": 2871, "lr": 8.446220037358676e-07}
{"train_loss": 0.03917165845632553, "global_step": 255575, "epoch": 2871, "lr": 8.445158985705692e-07}
{"train_loss": 0.026170536875724792, "global_step": 255576, "epoch": 2871, "lr": 8.444098000136292e-07}
{"train_loss": 0.023630933836102486, "global_step": 255577, "epoch": 2871, "lr": 8.443037080650529e-07}
{"train_loss": 0.036170877516269684, "global_step": 255578, "epoch": 2871, "lr": 8.441976227248682e-07}
{"train_loss": 0.06431657820940018, "global_step": 255579, "epoch": 2871, "lr": 8.440915439930807e-07}
{"train_loss": 0.026585867628455162, "global_step": 255580, "epoch": 2871, "lr": 8.439854718697071e-07}
{"train_loss": 0.02843240275979042, "global_step": 255581, "epoch": 2871, "lr": 8.438794063547528e-07}
{"train_loss": 0.017637446522712708, "global_step": 255582, "epoch": 2871, "lr": 8.437733474482512e-07}
{"train_loss": 0.036062657833099365, "global_step": 255583, "epoch": 2871, "lr": 8.436672951501967e-07}
{"train_loss": 0.07712975889444351, "global_step": 255584, "epoch": 2871, "lr": 8.435612494606171e-07}
{"train_loss": 0.033414244651794434, "global_step": 255585, "epoch": 2871, "lr": 8.434552103795124e-07}
{"train_loss": 0.07784996181726456, "global_step": 255586, "epoch": 2871, "lr": 8.433491779069158e-07}
{"train_loss": 0.042976509779691696, "global_step": 255587, "epoch": 2871, "lr": 8.432431520428274e-07}
{"train_loss": 0.03182840719819069, "global_step": 255588, "epoch": 2871, "lr": 8.431371327872694e-07}
{"train_loss": 0.06585638970136642, "global_step": 255589, "epoch": 2871, "lr": 8.430311201402474e-07}
{"train_loss": 0.023598574101924896, "global_step": 255590, "epoch": 2871, "lr": 8.429251141017836e-07}
{"train_loss": 0.09293206036090851, "global_step": 255591, "epoch": 2871, "lr": 8.428191146718889e-07}
{"train_loss": 0.07513207942247391, "global_step": 255592, "epoch": 2871, "lr": 8.427131218505747e-07}
{"train_loss": 0.04939620569348335, "global_step": 255593, "epoch": 2871, "lr": 8.426071356378629e-07}
{"train_loss": 0.02462468482553959, "global_step": 255594, "epoch": 2871, "lr": 8.425011560337648e-07}
{"train_loss": 0.0690140575170517, "global_step": 255595, "epoch": 2871, "lr": 8.423951830382915e-07}
{"train_loss": 0.028865806758403778, "global_step": 255596, "epoch": 2871, "lr": 8.422892166514595e-07}
{"train_loss": 0.05279750004410744, "global_step": 255597, "epoch": 2871, "lr": 8.421832568732802e-07}
{"train_loss": 0.06702325493097305, "global_step": 255598, "epoch": 2871, "lr": 8.420773037037754e-07}
{"train_loss": 0.04136498272418976, "global_step": 255599, "epoch": 2871, "lr": 8.41971357142951e-07}
{"train_loss": 0.05384425073862076, "global_step": 255600, "epoch": 2871, "lr": 8.418654171908236e-07}
{"train_loss": 0.05699796974658966, "global_step": 255601, "epoch": 2871, "lr": 8.417594838474097e-07}
{"train_loss": 0.047344036400318146, "global_step": 255602, "epoch": 2871, "lr": 8.416535571127204e-07}
{"train_loss": 0.022588441148400307, "global_step": 255603, "epoch": 2871, "lr": 8.415476369867725e-07}
{"train_loss": 0.02010219357907772, "global_step": 255604, "epoch": 2871, "lr": 8.41441723469577e-07}
{"train_loss": 0.039569076150655746, "global_step": 255605, "epoch": 2871, "lr": 8.413358165611507e-07}
{"train_loss": 0.025327567011117935, "global_step": 255606, "epoch": 2871, "lr": 8.412299162615101e-07}
{"train_loss": 0.04469290051316278, "global_step": 255607, "epoch": 2871, "lr": 8.411240225706662e-07, "val_loss": 9.609721183776855}
{"train_loss": 0.04448127746582031, "global_step": 255608, "epoch": 2872, "lr": 8.41018135488636e-07}
{"train_loss": 0.0738048106431961, "global_step": 255609, "epoch": 2872, "lr": 8.409122550154302e-07}
{"train_loss": 0.03906310722231865, "global_step": 255610, "epoch": 2872, "lr": 8.408063811510603e-07}
{"train_loss": 0.03929755464196205, "global_step": 255611, "epoch": 2872, "lr": 8.407005138955481e-07}
{"train_loss": 0.023389235138893127, "global_step": 255612, "epoch": 2872, "lr": 8.405946532489051e-07}
{"train_loss": 0.03411456570029259, "global_step": 255613, "epoch": 2872, "lr": 8.404887992111477e-07}
{"train_loss": 0.027404582127928734, "global_step": 255614, "epoch": 2872, "lr": 8.403829517822814e-07}
{"train_loss": 0.08336810022592545, "global_step": 255615, "epoch": 2872, "lr": 8.402771109623286e-07}
{"train_loss": 0.04886694252490997, "global_step": 255616, "epoch": 2872, "lr": 8.40171276751306e-07}
{"train_loss": 0.036573778837919235, "global_step": 255617, "epoch": 2872, "lr": 8.400654491492188e-07}
{"train_loss": 0.049478523433208466, "global_step": 255618, "epoch": 2872, "lr": 8.399596281560895e-07}
{"train_loss": 0.05235051363706589, "global_step": 255619, "epoch": 2872, "lr": 8.398538137719236e-07}
{"train_loss": 0.02017691172659397, "global_step": 255620, "epoch": 2872, "lr": 8.397480059967433e-07}
{"train_loss": 0.021315407007932663, "global_step": 255621, "epoch": 2872, "lr": 8.396422048305541e-07}
{"train_loss": 0.05774412676692009, "global_step": 255622, "epoch": 2872, "lr": 8.395364102733838e-07}
{"train_loss": 0.020150672644376755, "global_step": 255623, "epoch": 2872, "lr": 8.394306223252324e-07}
{"train_loss": 0.011229963041841984, "global_step": 255624, "epoch": 2872, "lr": 8.393248409861277e-07}
{"train_loss": 0.018402887508273125, "global_step": 255625, "epoch": 2872, "lr": 8.392190662560695e-07}
{"train_loss": 0.0091707156971097, "global_step": 255626, "epoch": 2872, "lr": 8.391132981350803e-07}
{"train_loss": 0.10465806722640991, "global_step": 255627, "epoch": 2872, "lr": 8.390075366231765e-07}
{"train_loss": 0.028591439127922058, "global_step": 255628, "epoch": 2872, "lr": 8.389017817203638e-07}
{"train_loss": 0.038348130881786346, "global_step": 255629, "epoch": 2872, "lr": 8.387960334266643e-07}
{"train_loss": 0.06106889992952347, "global_step": 255630, "epoch": 2872, "lr": 8.386902917420891e-07}
{"train_loss": 0.07477187365293503, "global_step": 255631, "epoch": 2872, "lr": 8.38584556666655e-07}
{"train_loss": 0.04521432891488075, "global_step": 255632, "epoch": 2872, "lr": 8.38478828200373e-07}
{"train_loss": 0.038469940423965454, "global_step": 255633, "epoch": 2872, "lr": 8.383731063432598e-07}
{"train_loss": 0.01484476588666439, "global_step": 255634, "epoch": 2872, "lr": 8.382673910953265e-07}
{"train_loss": 0.03885766863822937, "global_step": 255635, "epoch": 2872, "lr": 8.381616824565841e-07}
{"train_loss": 0.08117661625146866, "global_step": 255636, "epoch": 2872, "lr": 8.380559804270604e-07}
{"train_loss": 0.05418117716908455, "global_step": 255637, "epoch": 2872, "lr": 8.379502850067555e-07}
{"train_loss": 0.03356577455997467, "global_step": 255638, "epoch": 2872, "lr": 8.378445961956915e-07}
{"train_loss": 0.04810179024934769, "global_step": 255639, "epoch": 2872, "lr": 8.377389139938796e-07}
{"train_loss": 0.0321531780064106, "global_step": 255640, "epoch": 2872, "lr": 8.376332384013308e-07}
{"train_loss": 0.08232787251472473, "global_step": 255641, "epoch": 2872, "lr": 8.375275694180673e-07}
{"train_loss": 0.04181237891316414, "global_step": 255642, "epoch": 2872, "lr": 8.374219070440947e-07}
{"train_loss": 0.059832170605659485, "global_step": 255643, "epoch": 2872, "lr": 8.373162512794408e-07}
{"train_loss": 0.033220402896404266, "global_step": 255644, "epoch": 2872, "lr": 8.372106021241e-07}
{"train_loss": 0.052044883370399475, "global_step": 255645, "epoch": 2872, "lr": 8.371049595781055e-07}
{"train_loss": 0.024985946714878082, "global_step": 255646, "epoch": 2872, "lr": 8.36999323641452e-07}
{"train_loss": 0.07424142956733704, "global_step": 255647, "epoch": 2872, "lr": 8.368936943141725e-07}
{"train_loss": 0.09894311428070068, "global_step": 255648, "epoch": 2872, "lr": 8.367880715962728e-07}
{"train_loss": 0.07292823493480682, "global_step": 255649, "epoch": 2872, "lr": 8.366824554877639e-07}
{"train_loss": 0.04535560682415962, "global_step": 255650, "epoch": 2872, "lr": 8.36576845988668e-07}
{"train_loss": 0.03748926892876625, "global_step": 255651, "epoch": 2872, "lr": 8.364712430989963e-07}
{"train_loss": 0.05430665239691734, "global_step": 255652, "epoch": 2872, "lr": 8.363656468187598e-07}
{"train_loss": 0.02630252204835415, "global_step": 255653, "epoch": 2872, "lr": 8.362600571479751e-07}
{"train_loss": 0.07698450982570648, "global_step": 255654, "epoch": 2872, "lr": 8.361544740866533e-07}
{"train_loss": 0.04378785192966461, "global_step": 255655, "epoch": 2872, "lr": 8.360488976348113e-07}
{"train_loss": 0.07246784120798111, "global_step": 255656, "epoch": 2872, "lr": 8.359433277924711e-07}
{"train_loss": 0.035894013941287994, "global_step": 255657, "epoch": 2872, "lr": 8.358377645596326e-07}
{"train_loss": 0.031230855733156204, "global_step": 255658, "epoch": 2872, "lr": 8.357322079363183e-07}
{"train_loss": 0.04414409026503563, "global_step": 255659, "epoch": 2872, "lr": 8.356266579225392e-07}
{"train_loss": 0.09687742590904236, "global_step": 255660, "epoch": 2872, "lr": 8.355211145183118e-07}
{"train_loss": 0.03891720622777939, "global_step": 255661, "epoch": 2872, "lr": 8.354155777236472e-07}
{"train_loss": 0.03349883854389191, "global_step": 255662, "epoch": 2872, "lr": 8.353100475385623e-07}
{"train_loss": 0.026583444327116013, "global_step": 255663, "epoch": 2872, "lr": 8.352045239630735e-07}
{"train_loss": 0.042316000908613205, "global_step": 255664, "epoch": 2872, "lr": 8.35099006997192e-07}
{"train_loss": 0.07516401261091232, "global_step": 255665, "epoch": 2872, "lr": 8.349934966409289e-07}
{"train_loss": 0.0442437082529068, "global_step": 255666, "epoch": 2872, "lr": 8.34887992894301e-07}
{"train_loss": 0.03227691724896431, "global_step": 255667, "epoch": 2872, "lr": 8.347824957573302e-07}
{"train_loss": 0.0630268007516861, "global_step": 255668, "epoch": 2872, "lr": 8.346770052300168e-07}
{"train_loss": 0.052257318049669266, "global_step": 255669, "epoch": 2872, "lr": 8.345715213123828e-07}
{"train_loss": 0.049620311707258224, "global_step": 255670, "epoch": 2872, "lr": 8.34466044004445e-07}
{"train_loss": 0.05667455866932869, "global_step": 255671, "epoch": 2872, "lr": 8.343605733062143e-07}
{"train_loss": 0.06648093461990356, "global_step": 255672, "epoch": 2872, "lr": 8.34255109217702e-07}
{"train_loss": 0.04617103561758995, "global_step": 255673, "epoch": 2872, "lr": 8.341496517389246e-07}
{"train_loss": 0.017934156581759453, "global_step": 255674, "epoch": 2872, "lr": 8.34044200869899e-07}
{"train_loss": 0.05958021059632301, "global_step": 255675, "epoch": 2872, "lr": 8.339387566106304e-07}
{"train_loss": 0.03212829679250717, "global_step": 255676, "epoch": 2872, "lr": 8.338333189611469e-07}
{"train_loss": 0.02754301205277443, "global_step": 255677, "epoch": 2872, "lr": 8.337278879214538e-07}
{"train_loss": 0.024421244859695435, "global_step": 255678, "epoch": 2872, "lr": 8.336224634915623e-07}
{"train_loss": 0.022888202220201492, "global_step": 255679, "epoch": 2872, "lr": 8.335170456714947e-07}
{"train_loss": 0.06549161672592163, "global_step": 255680, "epoch": 2872, "lr": 8.33411634461262e-07}
{"train_loss": 0.027347778901457787, "global_step": 255681, "epoch": 2872, "lr": 8.333062298608751e-07}
{"train_loss": 0.023442715406417847, "global_step": 255682, "epoch": 2872, "lr": 8.33200831870351e-07}
{"train_loss": 0.030954036861658096, "global_step": 255683, "epoch": 2872, "lr": 8.330954404897063e-07}
{"train_loss": 0.05140681192278862, "global_step": 255684, "epoch": 2872, "lr": 8.329900557189519e-07}
{"train_loss": 0.04176139459013939, "global_step": 255685, "epoch": 2872, "lr": 8.32884677558099e-07}
{"train_loss": 0.05409768968820572, "global_step": 255686, "epoch": 2872, "lr": 8.327793060071753e-07}
{"train_loss": 0.056080013513565063, "global_step": 255687, "epoch": 2872, "lr": 8.326739410661755e-07}
{"train_loss": 0.04247869551181793, "global_step": 255688, "epoch": 2872, "lr": 8.32568582735127e-07}
{"train_loss": 0.0572618693113327, "global_step": 255689, "epoch": 2872, "lr": 8.324632310140413e-07}
{"train_loss": 0.02619672380387783, "global_step": 255690, "epoch": 2872, "lr": 8.32357885902929e-07}
{"train_loss": 0.047301411628723145, "global_step": 255691, "epoch": 2872, "lr": 8.322525474018072e-07}
{"train_loss": 0.04807862266898155, "global_step": 255692, "epoch": 2872, "lr": 8.321472155106924e-07}
{"train_loss": 0.03144851699471474, "global_step": 255693, "epoch": 2872, "lr": 8.320418902295956e-07}
{"train_loss": 0.03038950264453888, "global_step": 255694, "epoch": 2872, "lr": 8.319365715585282e-07}
{"train_loss": 0.022182369604706764, "global_step": 255695, "epoch": 2872, "lr": 8.318312594975119e-07}
{"train_loss": 0.04552858716316437, "global_step": 255696, "epoch": 2872, "lr": 8.317259540465528e-07, "val_loss": 9.617732048034668}
{"train_loss": 0.06718333810567856, "global_step": 255697, "epoch": 2873, "lr": 8.316206552056727e-07}
{"train_loss": 0.033189743757247925, "global_step": 255698, "epoch": 2873, "lr": 8.315153629748774e-07}
{"train_loss": 0.050529517233371735, "global_step": 255699, "epoch": 2873, "lr": 8.314100773541889e-07}
{"train_loss": 0.06240685656666756, "global_step": 255700, "epoch": 2873, "lr": 8.313047983436184e-07}
{"train_loss": 0.0433042012155056, "global_step": 255701, "epoch": 2873, "lr": 8.311995259431771e-07}
{"train_loss": 0.035812243819236755, "global_step": 255702, "epoch": 2873, "lr": 8.310942601528815e-07}
{"train_loss": 0.03235559165477753, "global_step": 255703, "epoch": 2873, "lr": 8.309890009727484e-07}
{"train_loss": 0.01401822455227375, "global_step": 255704, "epoch": 2873, "lr": 8.308837484027887e-07}
{"train_loss": 0.038088396191596985, "global_step": 255705, "epoch": 2873, "lr": 8.307785024430193e-07}
{"train_loss": 0.014578839763998985, "global_step": 255706, "epoch": 2873, "lr": 8.306732630934455e-07}
{"train_loss": 0.03233540430665016, "global_step": 255707, "epoch": 2873, "lr": 8.305680303541008e-07}
{"train_loss": 0.027005132287740707, "global_step": 255708, "epoch": 2873, "lr": 8.304628042249741e-07}
{"train_loss": 0.05122595280408859, "global_step": 255709, "epoch": 2873, "lr": 8.303575847060985e-07}
{"train_loss": 0.07046432048082352, "global_step": 255710, "epoch": 2873, "lr": 8.302523717974853e-07}
{"train_loss": 0.03568880259990692, "global_step": 255711, "epoch": 2873, "lr": 8.3014716549914e-07}
{"train_loss": 0.05694984644651413, "global_step": 255712, "epoch": 2873, "lr": 8.300419658110847e-07}
{"train_loss": 0.09206484258174896, "global_step": 255713, "epoch": 2873, "lr": 8.299367727333307e-07}
{"train_loss": 0.05396011844277382, "global_step": 255714, "epoch": 2873, "lr": 8.298315862658945e-07}
{"train_loss": 0.03756559267640114, "global_step": 255715, "epoch": 2873, "lr": 8.297264064087873e-07}
{"train_loss": 0.054802313446998596, "global_step": 255716, "epoch": 2873, "lr": 8.296212331620202e-07}
{"train_loss": 0.04259246215224266, "global_step": 255717, "epoch": 2873, "lr": 8.295160665256152e-07}
{"train_loss": 0.04027266055345535, "global_step": 255718, "epoch": 2873, "lr": 8.29410906499578e-07}
{"train_loss": 0.0263062734156847, "global_step": 255719, "epoch": 2873, "lr": 8.293057530839255e-07}
{"train_loss": 0.04848068952560425, "global_step": 255720, "epoch": 2873, "lr": 8.292006062786795e-07}
{"train_loss": 0.044932980090379715, "global_step": 255721, "epoch": 2873, "lr": 8.290954660838457e-07}
{"train_loss": 0.04662212356925011, "global_step": 255722, "epoch": 2873, "lr": 8.289903324994408e-07}
{"train_loss": 0.036399468779563904, "global_step": 255723, "epoch": 2873, "lr": 8.288852055254759e-07}
{"train_loss": 0.014295587316155434, "global_step": 255724, "epoch": 2873, "lr": 8.287800851619787e-07}
{"train_loss": 0.023223627358675003, "global_step": 255725, "epoch": 2873, "lr": 8.286749714089436e-07}
{"train_loss": 0.07735421508550644, "global_step": 255726, "epoch": 2873, "lr": 8.285698642663931e-07}
{"train_loss": 0.031536903232336044, "global_step": 255727, "epoch": 2873, "lr": 8.284647637343434e-07}
{"train_loss": 0.014014722779393196, "global_step": 255728, "epoch": 2873, "lr": 8.283596698128116e-07}
{"train_loss": 0.020740048959851265, "global_step": 255729, "epoch": 2873, "lr": 8.282545825018028e-07}
{"train_loss": 0.036289703100919724, "global_step": 255730, "epoch": 2873, "lr": 8.281495018013397e-07}
{"train_loss": 0.019355062395334244, "global_step": 255731, "epoch": 2873, "lr": 8.280444277114274e-07}
{"train_loss": 0.06289656460285187, "global_step": 255732, "epoch": 2873, "lr": 8.279393602320939e-07}
{"train_loss": 0.023691421374678612, "global_step": 255733, "epoch": 2873, "lr": 8.278342993633337e-07}
{"train_loss": 0.03640099614858627, "global_step": 255734, "epoch": 2873, "lr": 8.277292451051799e-07}
{"train_loss": 0.049218785017728806, "global_step": 255735, "epoch": 2873, "lr": 8.276241974576327e-07}
{"train_loss": 0.04326144978404045, "global_step": 255736, "epoch": 2873, "lr": 8.275191564207196e-07}
{"train_loss": 0.1150466576218605, "global_step": 255737, "epoch": 2873, "lr": 8.274141219944409e-07}
{"train_loss": 0.04367688670754433, "global_step": 255738, "epoch": 2873, "lr": 8.273090941788131e-07}
{"train_loss": 0.049499206244945526, "global_step": 255739, "epoch": 2873, "lr": 8.272040729738639e-07}
{"train_loss": 0.05058141052722931, "global_step": 255740, "epoch": 2873, "lr": 8.270990583795934e-07}
{"train_loss": 0.030030053108930588, "global_step": 255741, "epoch": 2873, "lr": 8.269940503960238e-07}
{"train_loss": 0.031665410846471786, "global_step": 255742, "epoch": 2873, "lr": 8.268890490231607e-07}
{"train_loss": 0.018565190955996513, "global_step": 255743, "epoch": 2873, "lr": 8.267840542610262e-07}
{"train_loss": 0.03262614086270332, "global_step": 255744, "epoch": 2873, "lr": 8.266790661096258e-07}
{"train_loss": 0.01260386686772108, "global_step": 255745, "epoch": 2873, "lr": 8.265740845689874e-07}
{"train_loss": 0.044147226959466934, "global_step": 255746, "epoch": 2873, "lr": 8.264691096391108e-07}
{"train_loss": 0.01357342954725027, "global_step": 255747, "epoch": 2873, "lr": 8.263641413200185e-07}
{"train_loss": 0.027259010821580887, "global_step": 255748, "epoch": 2873, "lr": 8.262591796117213e-07}
{"train_loss": 0.04593998193740845, "global_step": 255749, "epoch": 2873, "lr": 8.261542245142362e-07}
{"train_loss": 0.026313139125704765, "global_step": 255750, "epoch": 2873, "lr": 8.26049276027574e-07}
{"train_loss": 0.055501699447631836, "global_step": 255751, "epoch": 2873, "lr": 8.259443341517458e-07}
{"train_loss": 0.04041389375925064, "global_step": 255752, "epoch": 2873, "lr": 8.258393988867741e-07}
{"train_loss": 0.029059506952762604, "global_step": 255753, "epoch": 2873, "lr": 8.257344702326697e-07}
{"train_loss": 0.022419117391109467, "global_step": 255754, "epoch": 2873, "lr": 8.256295481894438e-07}
{"train_loss": 0.02991153486073017, "global_step": 255755, "epoch": 2873, "lr": 8.255246327571187e-07}
{"train_loss": 0.02348710410296917, "global_step": 255756, "epoch": 2873, "lr": 8.254197239356943e-07}
{"train_loss": 0.023292716592550278, "global_step": 255757, "epoch": 2873, "lr": 8.253148217251927e-07}
{"train_loss": 0.04268952086567879, "global_step": 255758, "epoch": 2873, "lr": 8.252099261256363e-07}
{"train_loss": 0.0645667165517807, "global_step": 255759, "epoch": 2873, "lr": 8.251050371370251e-07}
{"train_loss": 0.021121833473443985, "global_step": 255760, "epoch": 2873, "lr": 8.250001547593811e-07}
{"train_loss": 0.046314917504787445, "global_step": 255761, "epoch": 2873, "lr": 8.248952789927156e-07}
{"train_loss": 0.012754585593938828, "global_step": 255762, "epoch": 2873, "lr": 8.247904098370451e-07}
{"train_loss": 0.022876447066664696, "global_step": 255763, "epoch": 2873, "lr": 8.246855472923753e-07}
{"train_loss": 0.052887752652168274, "global_step": 255764, "epoch": 2873, "lr": 8.245806913587395e-07}
{"train_loss": 0.035008642822504044, "global_step": 255765, "epoch": 2873, "lr": 8.244758420361265e-07}
{"train_loss": 0.046517398208379745, "global_step": 255766, "epoch": 2873, "lr": 8.243709993245751e-07}
{"train_loss": 0.03063359670341015, "global_step": 255767, "epoch": 2873, "lr": 8.2426616322408e-07}
{"train_loss": 0.009252005256712437, "global_step": 255768, "epoch": 2873, "lr": 8.241613337346632e-07}
{"train_loss": 0.03178267925977707, "global_step": 255769, "epoch": 2873, "lr": 8.240565108563414e-07}
{"train_loss": 0.02779478020966053, "global_step": 255770, "epoch": 2873, "lr": 8.239516945891257e-07}
{"train_loss": 0.04401100426912308, "global_step": 255771, "epoch": 2873, "lr": 8.238468849330272e-07}
{"train_loss": 0.09618783742189407, "global_step": 255772, "epoch": 2873, "lr": 8.237420818880681e-07}
{"train_loss": 0.04020014777779579, "global_step": 255773, "epoch": 2873, "lr": 8.23637285454254e-07}
{"train_loss": 0.029812466353178024, "global_step": 255774, "epoch": 2873, "lr": 8.235324956316015e-07}
{"train_loss": 0.046643901616334915, "global_step": 255775, "epoch": 2873, "lr": 8.234277124201273e-07}
{"train_loss": 0.026096977293491364, "global_step": 255776, "epoch": 2873, "lr": 8.233229358198424e-07}
{"train_loss": 0.03536985442042351, "global_step": 255777, "epoch": 2873, "lr": 8.232181658307636e-07}
{"train_loss": 0.055881042033433914, "global_step": 255778, "epoch": 2873, "lr": 8.231134024529019e-07}
{"train_loss": 0.032930780202150345, "global_step": 255779, "epoch": 2873, "lr": 8.230086456862796e-07}
{"train_loss": 0.08048929274082184, "global_step": 255780, "epoch": 2873, "lr": 8.229038955308966e-07}
{"train_loss": 0.014650669880211353, "global_step": 255781, "epoch": 2873, "lr": 8.227991519867805e-07}
{"train_loss": 0.05288637802004814, "global_step": 255782, "epoch": 2873, "lr": 8.226944150539373e-07}
{"train_loss": 0.04326726123690605, "global_step": 255783, "epoch": 2873, "lr": 8.225896847323833e-07}
{"train_loss": 0.05464380234479904, "global_step": 255784, "epoch": 2873, "lr": 8.224849610221352e-07}
{"train_loss": 0.03974872844272785, "global_step": 255785, "epoch": 2873, "lr": 8.223802439232043e-07, "val_loss": 9.61106014251709}
{"train_loss": 0.054032839834690094, "global_step": 255786, "epoch": 2874, "lr": 8.222755334356013e-07}
{"train_loss": 0.015253923833370209, "global_step": 255787, "epoch": 2874, "lr": 8.221708295593433e-07}
{"train_loss": 0.03930908441543579, "global_step": 255788, "epoch": 2874, "lr": 8.220661322944523e-07}
{"train_loss": 0.054107166826725006, "global_step": 255789, "epoch": 2874, "lr": 8.219614416409283e-07}
{"train_loss": 0.02398730255663395, "global_step": 255790, "epoch": 2874, "lr": 8.218567575987934e-07}
{"train_loss": 0.07943913340568542, "global_step": 255791, "epoch": 2874, "lr": 8.217520801680645e-07}
{"train_loss": 0.035616569221019745, "global_step": 255792, "epoch": 2874, "lr": 8.21647409348747e-07}
{"train_loss": 0.03448477014899254, "global_step": 255793, "epoch": 2874, "lr": 8.215427451408575e-07}
{"train_loss": 0.06029767543077469, "global_step": 255794, "epoch": 2874, "lr": 8.214380875444183e-07}
{"train_loss": 0.032470326870679855, "global_step": 255795, "epoch": 2874, "lr": 8.213334365594349e-07}
{"train_loss": 0.026222653687000275, "global_step": 255796, "epoch": 2874, "lr": 8.212287921859241e-07}
{"train_loss": 0.03126182034611702, "global_step": 255797, "epoch": 2874, "lr": 8.211241544238968e-07}
{"train_loss": 0.08161582797765732, "global_step": 255798, "epoch": 2874, "lr": 8.210195232733753e-07}
{"train_loss": 0.05901079997420311, "global_step": 255799, "epoch": 2874, "lr": 8.209148987343651e-07}
{"train_loss": 0.08369502425193787, "global_step": 255800, "epoch": 2874, "lr": 8.208102808068829e-07}
{"train_loss": 0.03975146636366844, "global_step": 255801, "epoch": 2874, "lr": 8.207056694909454e-07}
{"train_loss": 0.025056293234229088, "global_step": 255802, "epoch": 2874, "lr": 8.206010647865637e-07}
{"train_loss": 0.03429676964879036, "global_step": 255803, "epoch": 2874, "lr": 8.204964666937542e-07}
{"train_loss": 0.021234730258584023, "global_step": 255804, "epoch": 2874, "lr": 8.203918752125284e-07}
{"train_loss": 0.055432550609111786, "global_step": 255805, "epoch": 2874, "lr": 8.202872903428971e-07}
{"train_loss": 0.020029837265610695, "global_step": 255806, "epoch": 2874, "lr": 8.201827120848881e-07}
{"train_loss": 0.0671779215335846, "global_step": 255807, "epoch": 2874, "lr": 8.20078140438496e-07}
{"train_loss": 0.022800292819738388, "global_step": 255808, "epoch": 2874, "lr": 8.199735754037541e-07}
{"train_loss": 0.045439716428518295, "global_step": 255809, "epoch": 2874, "lr": 8.198690169806567e-07}
{"train_loss": 0.0416577085852623, "global_step": 255810, "epoch": 2874, "lr": 8.197644651692371e-07}
{"train_loss": 0.03638951852917671, "global_step": 255811, "epoch": 2874, "lr": 8.196599199694954e-07}
{"train_loss": 0.021179188042879105, "global_step": 255812, "epoch": 2874, "lr": 8.195553813814539e-07}
{"train_loss": 0.04914960265159607, "global_step": 255813, "epoch": 2874, "lr": 8.194508494051289e-07}
{"train_loss": 0.05182806774973869, "global_step": 255814, "epoch": 2874, "lr": 8.193463240405208e-07}
{"train_loss": 0.05120785906910896, "global_step": 255815, "epoch": 2874, "lr": 8.192418052876571e-07}
{"train_loss": 0.05295097082853317, "global_step": 255816, "epoch": 2874, "lr": 8.191372931465436e-07}
{"train_loss": 0.04860934615135193, "global_step": 255817, "epoch": 2874, "lr": 8.190327876172021e-07}
{"train_loss": 0.028343521058559418, "global_step": 255818, "epoch": 2874, "lr": 8.189282886996385e-07}
{"train_loss": 0.029230812564492226, "global_step": 255819, "epoch": 2874, "lr": 8.188237963938694e-07}
{"train_loss": 0.058276623487472534, "global_step": 255820, "epoch": 2874, "lr": 8.187193106999113e-07}
{"train_loss": 0.023715144023299217, "global_step": 255821, "epoch": 2874, "lr": 8.18614831617781e-07}
{"train_loss": 0.026240751147270203, "global_step": 255822, "epoch": 2874, "lr": 8.185103591474841e-07}
{"train_loss": 0.06739290058612823, "global_step": 255823, "epoch": 2874, "lr": 8.184058932890426e-07}
{"train_loss": 0.032857928425073624, "global_step": 255824, "epoch": 2874, "lr": 8.183014340424622e-07}
{"train_loss": 0.07437439262866974, "global_step": 255825, "epoch": 2874, "lr": 8.181969814077651e-07}
{"train_loss": 0.026802053675055504, "global_step": 255826, "epoch": 2874, "lr": 8.180925353849623e-07}
{"train_loss": 0.0689237117767334, "global_step": 255827, "epoch": 2874, "lr": 8.179880959740649e-07}
{"train_loss": 0.05009083449840546, "global_step": 255828, "epoch": 2874, "lr": 8.178836631750952e-07}
{"train_loss": 0.029487621039152145, "global_step": 255829, "epoch": 2874, "lr": 8.177792369880532e-07}
{"train_loss": 0.013860045932233334, "global_step": 255830, "epoch": 2874, "lr": 8.176748174129667e-07}
{"train_loss": 0.052112460136413574, "global_step": 255831, "epoch": 2874, "lr": 8.175704044498411e-07}
{"train_loss": 0.049580544233322144, "global_step": 255832, "epoch": 2874, "lr": 8.174659980986987e-07}
{"train_loss": 0.04706060513854027, "global_step": 255833, "epoch": 2874, "lr": 8.173615983595451e-07}
{"train_loss": 0.032862961292266846, "global_step": 255834, "epoch": 2874, "lr": 8.172572052323968e-07}
{"train_loss": 0.060938842594623566, "global_step": 255835, "epoch": 2874, "lr": 8.171528187172706e-07}
{"train_loss": 0.017095208168029785, "global_step": 255836, "epoch": 2874, "lr": 8.170484388141775e-07}
{"train_loss": 0.05901053547859192, "global_step": 255837, "epoch": 2874, "lr": 8.169440655231341e-07}
{"train_loss": 0.040967125445604324, "global_step": 255838, "epoch": 2874, "lr": 8.168396988441573e-07}
{"train_loss": 0.017588602378964424, "global_step": 255839, "epoch": 2874, "lr": 8.16735338777247e-07}
{"train_loss": 0.03802694380283356, "global_step": 255840, "epoch": 2874, "lr": 8.166309853224363e-07}
{"train_loss": 0.0393868088722229, "global_step": 255841, "epoch": 2874, "lr": 8.165266384797255e-07}
{"train_loss": 0.06487593799829483, "global_step": 255842, "epoch": 2874, "lr": 8.164222982491366e-07}
{"train_loss": 0.04783462733030319, "global_step": 255843, "epoch": 2874, "lr": 8.163179646306751e-07}
{"train_loss": 0.022163597866892815, "global_step": 255844, "epoch": 2874, "lr": 8.162136376243634e-07}
{"train_loss": 0.05677730217576027, "global_step": 255845, "epoch": 2874, "lr": 8.16109317230207e-07}
{"train_loss": 0.015066220425069332, "global_step": 255846, "epoch": 2874, "lr": 8.160050034482281e-07}
{"train_loss": 0.07761488109827042, "global_step": 255847, "epoch": 2874, "lr": 8.159006962784432e-07}
{"train_loss": 0.056460123509168625, "global_step": 255848, "epoch": 2874, "lr": 8.157963957208581e-07}
{"train_loss": 0.04613091051578522, "global_step": 255849, "epoch": 2874, "lr": 8.156921017754893e-07}
{"train_loss": 0.03287810832262039, "global_step": 255850, "epoch": 2874, "lr": 8.155878144423478e-07}
{"train_loss": 0.026021085679531097, "global_step": 255851, "epoch": 2874, "lr": 8.154835337214561e-07}
{"train_loss": 0.04155464470386505, "global_step": 255852, "epoch": 2874, "lr": 8.153792596128195e-07}
{"train_loss": 0.03752126544713974, "global_step": 255853, "epoch": 2874, "lr": 8.152749921164604e-07}
{"train_loss": 0.06811534613370895, "global_step": 255854, "epoch": 2874, "lr": 8.151707312323842e-07}
{"train_loss": 0.03326850011944771, "global_step": 255855, "epoch": 2874, "lr": 8.150664769606132e-07}
{"train_loss": 0.03454338386654854, "global_step": 255856, "epoch": 2874, "lr": 8.149622293011527e-07}
{"train_loss": 0.05110863968729973, "global_step": 255857, "epoch": 2874, "lr": 8.148579882540197e-07}
{"train_loss": 0.026399845257401466, "global_step": 255858, "epoch": 2874, "lr": 8.147537538192306e-07}
{"train_loss": 0.05620057135820389, "global_step": 255859, "epoch": 2874, "lr": 8.146495259968023e-07}
{"train_loss": 0.02884593978524208, "global_step": 255860, "epoch": 2874, "lr": 8.145453047867402e-07}
{"train_loss": 0.05723453685641289, "global_step": 255861, "epoch": 2874, "lr": 8.144410901890664e-07}
{"train_loss": 0.06953708827495575, "global_step": 255862, "epoch": 2874, "lr": 8.143368822037866e-07}
{"train_loss": 0.06418298929929733, "global_step": 255863, "epoch": 2874, "lr": 8.142326808309287e-07}
{"train_loss": 0.04033844918012619, "global_step": 255864, "epoch": 2874, "lr": 8.141284860704923e-07}
{"train_loss": 0.046300824731588364, "global_step": 255865, "epoch": 2874, "lr": 8.140242979224944e-07}
{"train_loss": 0.035232704132795334, "global_step": 255866, "epoch": 2874, "lr": 8.13920116386957e-07}
{"train_loss": 0.011552237905561924, "global_step": 255867, "epoch": 2874, "lr": 8.138159414638857e-07}
{"train_loss": 0.04389461502432823, "global_step": 255868, "epoch": 2874, "lr": 8.137117731532973e-07}
{"train_loss": 0.01992407813668251, "global_step": 255869, "epoch": 2874, "lr": 8.136076114552027e-07}
{"train_loss": 0.03872958570718765, "global_step": 255870, "epoch": 2874, "lr": 8.135034563696242e-07}
{"train_loss": 0.028135228902101517, "global_step": 255871, "epoch": 2874, "lr": 8.133993078965674e-07}
{"train_loss": 0.03544776886701584, "global_step": 255872, "epoch": 2874, "lr": 8.132951660360544e-07}
{"train_loss": 0.05523321032524109, "global_step": 255873, "epoch": 2874, "lr": 8.131910307880852e-07}
{"train_loss": 0.042381427648445866, "global_step": 255874, "epoch": 2874, "lr": 8.130869021526933e-07, "val_loss": 9.621505737304688}
{"train_loss": 0.03533681109547615, "global_step": 255875, "epoch": 2875, "lr": 8.129827801298728e-07}
{"train_loss": 0.03399307653307915, "global_step": 255876, "epoch": 2875, "lr": 8.128786647196574e-07}
{"train_loss": 0.043557122349739075, "global_step": 255877, "epoch": 2875, "lr": 8.127745559220412e-07}
{"train_loss": 0.06810574978590012, "global_step": 255878, "epoch": 2875, "lr": 8.126704537370522e-07}
{"train_loss": 0.06944461166858673, "global_step": 255879, "epoch": 2875, "lr": 8.125663581647014e-07}
{"train_loss": 0.03443092107772827, "global_step": 255880, "epoch": 2875, "lr": 8.12462269205e-07}
{"train_loss": 0.03928795084357262, "global_step": 255881, "epoch": 2875, "lr": 8.123581868579644e-07}
{"train_loss": 0.04570263996720314, "global_step": 255882, "epoch": 2875, "lr": 8.12254111123606e-07}
{"train_loss": 0.0882987529039383, "global_step": 255883, "epoch": 2875, "lr": 8.121500420019413e-07}
{"train_loss": 0.02467513084411621, "global_step": 255884, "epoch": 2875, "lr": 8.120459794929814e-07}
{"train_loss": 0.05299331992864609, "global_step": 255885, "epoch": 2875, "lr": 8.119419235967485e-07}
{"train_loss": 0.06641717255115509, "global_step": 255886, "epoch": 2875, "lr": 8.118378743132427e-07}
{"train_loss": 0.04132232815027237, "global_step": 255887, "epoch": 2875, "lr": 8.117338316424916e-07}
{"train_loss": 0.06912872940301895, "global_step": 255888, "epoch": 2875, "lr": 8.116297955845009e-07}
{"train_loss": 0.02434680238366127, "global_step": 255889, "epoch": 2875, "lr": 8.115257661392872e-07}
{"train_loss": 0.11250326782464981, "global_step": 255890, "epoch": 2875, "lr": 8.114217433068672e-07}
{"train_loss": 0.027276823297142982, "global_step": 255891, "epoch": 2875, "lr": 8.113177270872518e-07}
{"train_loss": 0.059990961104631424, "global_step": 255892, "epoch": 2875, "lr": 8.112137174804468e-07}
{"train_loss": 0.0381564199924469, "global_step": 255893, "epoch": 2875, "lr": 8.111097144864855e-07}
{"train_loss": 0.04911879450082779, "global_step": 255894, "epoch": 2875, "lr": 8.11005718105362e-07}
{"train_loss": 0.06551913917064667, "global_step": 255895, "epoch": 2875, "lr": 8.109017283371101e-07}
{"train_loss": 0.051179271191358566, "global_step": 255896, "epoch": 2875, "lr": 8.107977451817239e-07}
{"train_loss": 0.021678708493709564, "global_step": 255897, "epoch": 2875, "lr": 8.106937686392313e-07}
{"train_loss": 0.05399857088923454, "global_step": 255898, "epoch": 2875, "lr": 8.105897987096378e-07}
{"train_loss": 0.02222374640405178, "global_step": 255899, "epoch": 2875, "lr": 8.104858353929601e-07}
{"train_loss": 0.02618190459907055, "global_step": 255900, "epoch": 2875, "lr": 8.103818786892147e-07}
{"train_loss": 0.032507214695215225, "global_step": 255901, "epoch": 2875, "lr": 8.102779285984186e-07}
{"train_loss": 0.05717889592051506, "global_step": 255902, "epoch": 2875, "lr": 8.101739851205714e-07}
{"train_loss": 0.025902897119522095, "global_step": 255903, "epoch": 2875, "lr": 8.100700482557066e-07}
{"train_loss": 0.02191915176808834, "global_step": 255904, "epoch": 2875, "lr": 8.099661180038242e-07}
{"train_loss": 0.034111183136701584, "global_step": 255905, "epoch": 2875, "lr": 8.098621943649409e-07}
{"train_loss": 0.07265712320804596, "global_step": 255906, "epoch": 2875, "lr": 8.097582773390788e-07}
{"train_loss": 0.028240198269486427, "global_step": 255907, "epoch": 2875, "lr": 8.09654366926238e-07}
{"train_loss": 0.07915765792131424, "global_step": 255908, "epoch": 2875, "lr": 8.095504631264406e-07}
{"train_loss": 0.01177176833152771, "global_step": 255909, "epoch": 2875, "lr": 8.094465659396977e-07}
{"train_loss": 0.06198810413479805, "global_step": 255910, "epoch": 2875, "lr": 8.093426753660316e-07}
{"train_loss": 0.058383211493492126, "global_step": 255911, "epoch": 2875, "lr": 8.092387914054422e-07}
{"train_loss": 0.04823847860097885, "global_step": 255912, "epoch": 2875, "lr": 8.091349140579574e-07}
{"train_loss": 0.034348729997873306, "global_step": 255913, "epoch": 2875, "lr": 8.090310433235826e-07}
{"train_loss": 0.008079378865659237, "global_step": 255914, "epoch": 2875, "lr": 8.089271792023345e-07}
{"train_loss": 0.06005806848406792, "global_step": 255915, "epoch": 2875, "lr": 8.088233216942243e-07}
{"train_loss": 0.06591606140136719, "global_step": 255916, "epoch": 2875, "lr": 8.087194707992684e-07}
{"train_loss": 0.02946598455309868, "global_step": 255917, "epoch": 2875, "lr": 8.086156265174838e-07}
{"train_loss": 0.06587563455104828, "global_step": 255918, "epoch": 2875, "lr": 8.085117888488813e-07}
{"train_loss": 0.0540776401758194, "global_step": 255919, "epoch": 2875, "lr": 8.084079577934722e-07}
{"train_loss": 0.033729273825883865, "global_step": 255920, "epoch": 2875, "lr": 8.083041333512675e-07}
{"train_loss": 0.029171818867325783, "global_step": 255921, "epoch": 2875, "lr": 8.08200315522295e-07}
{"train_loss": 0.03171819821000099, "global_step": 255922, "epoch": 2875, "lr": 8.080965043065602e-07}
{"train_loss": 0.09496499598026276, "global_step": 255923, "epoch": 2875, "lr": 8.079926997040743e-07}
{"train_loss": 0.015310296788811684, "global_step": 255924, "epoch": 2875, "lr": 8.078889017148539e-07}
{"train_loss": 0.04062237590551376, "global_step": 255925, "epoch": 2875, "lr": 8.077851103389211e-07}
{"train_loss": 0.03763475641608238, "global_step": 255926, "epoch": 2875, "lr": 8.076813255762705e-07}
{"train_loss": 0.055123038589954376, "global_step": 255927, "epoch": 2875, "lr": 8.075775474269354e-07}
{"train_loss": 0.04400855675339699, "global_step": 255928, "epoch": 2875, "lr": 8.074737758909157e-07}
{"train_loss": 0.045518986880779266, "global_step": 255929, "epoch": 2875, "lr": 8.073700109682392e-07}
{"train_loss": 0.03340514004230499, "global_step": 255930, "epoch": 2875, "lr": 8.072662526589058e-07}
{"train_loss": 0.015804696828126907, "global_step": 255931, "epoch": 2875, "lr": 8.071625009629436e-07}
{"train_loss": 0.06754383444786072, "global_step": 255932, "epoch": 2875, "lr": 8.070587558803466e-07}
{"train_loss": 0.05135979875922203, "global_step": 255933, "epoch": 2875, "lr": 8.06955017411154e-07}
{"train_loss": 0.051159653812646866, "global_step": 255934, "epoch": 2875, "lr": 8.068512855553601e-07}
{"train_loss": 0.041566792875528336, "global_step": 255935, "epoch": 2875, "lr": 8.06747560312987e-07}
{"train_loss": 0.03149470314383507, "global_step": 255936, "epoch": 2875, "lr": 8.066438416840406e-07}
{"train_loss": 0.045012086629867554, "global_step": 255937, "epoch": 2875, "lr": 8.065401296685537e-07}
{"train_loss": 0.03341764211654663, "global_step": 255938, "epoch": 2875, "lr": 8.064364242665156e-07}
{"train_loss": 0.03337777405977249, "global_step": 255939, "epoch": 2875, "lr": 8.063327254779595e-07}
{"train_loss": 0.05434027314186096, "global_step": 255940, "epoch": 2875, "lr": 8.06229033302891e-07}
{"train_loss": 0.03837846219539642, "global_step": 255941, "epoch": 2875, "lr": 8.061253477413267e-07}
{"train_loss": 0.03956824913620949, "global_step": 255942, "epoch": 2875, "lr": 8.060216687932775e-07}
{"train_loss": 0.06953900307416916, "global_step": 255943, "epoch": 2875, "lr": 8.059179964587604e-07}
{"train_loss": 0.036375537514686584, "global_step": 255944, "epoch": 2875, "lr": 8.058143307377919e-07}
{"train_loss": 0.04470401257276535, "global_step": 255945, "epoch": 2875, "lr": 8.057106716303719e-07}
{"train_loss": 0.008781468495726585, "global_step": 255946, "epoch": 2875, "lr": 8.056070191365339e-07}
{"train_loss": 0.03997339680790901, "global_step": 255947, "epoch": 2875, "lr": 8.055033732562778e-07}
{"train_loss": 0.0189752746373415, "global_step": 255948, "epoch": 2875, "lr": 8.053997339896202e-07}
{"train_loss": 0.02543948031961918, "global_step": 255949, "epoch": 2875, "lr": 8.052961013365778e-07}
{"train_loss": 0.06038700044155121, "global_step": 255950, "epoch": 2875, "lr": 8.051924752971674e-07}
{"train_loss": 0.0232063177973032, "global_step": 255951, "epoch": 2875, "lr": 8.050888558713998e-07}
{"train_loss": 0.041087765246629715, "global_step": 255952, "epoch": 2875, "lr": 8.049852430592863e-07}
{"train_loss": 0.03382045403122902, "global_step": 255953, "epoch": 2875, "lr": 8.048816368608381e-07}
{"train_loss": 0.02078372798860073, "global_step": 255954, "epoch": 2875, "lr": 8.047780372760827e-07}
{"train_loss": 0.023656299337744713, "global_step": 255955, "epoch": 2875, "lr": 8.046744443050202e-07}
{"train_loss": 0.0696554183959961, "global_step": 255956, "epoch": 2875, "lr": 8.045708579476674e-07}
{"train_loss": 0.0485801100730896, "global_step": 255957, "epoch": 2875, "lr": 8.044672782040463e-07}
{"train_loss": 0.03426680341362953, "global_step": 255958, "epoch": 2875, "lr": 8.043637050741626e-07}
{"train_loss": 0.0776546373963356, "global_step": 255959, "epoch": 2875, "lr": 8.042601385580328e-07}
{"train_loss": 0.029605010524392128, "global_step": 255960, "epoch": 2875, "lr": 8.041565786556682e-07}
{"train_loss": 0.07475969195365906, "global_step": 255961, "epoch": 2875, "lr": 8.040530253670909e-07}
{"train_loss": 0.05606180429458618, "global_step": 255962, "epoch": 2875, "lr": 8.039494786923007e-07}
{"train_loss": 0.045011340221913346, "global_step": 255963, "epoch": 2875, "lr": 8.038459386313258e-07, "val_loss": 9.700350761413574, "train_action_mse_error": 12.643205642700195}
{"train_loss": 0.012505150400102139, "global_step": 255964, "epoch": 2876, "lr": 8.037424051841713e-07}
{"train_loss": 0.01231105625629425, "global_step": 255965, "epoch": 2876, "lr": 8.036388783508597e-07}
{"train_loss": 0.019546348601579666, "global_step": 255966, "epoch": 2876, "lr": 8.03535358131391e-07}
{"train_loss": 0.023773573338985443, "global_step": 255967, "epoch": 2876, "lr": 8.034318445257926e-07}
{"train_loss": 0.06046934053301811, "global_step": 255968, "epoch": 2876, "lr": 8.033283375340706e-07}
{"train_loss": 0.06242179498076439, "global_step": 255969, "epoch": 2876, "lr": 8.032248371562468e-07}
{"train_loss": 0.04232853651046753, "global_step": 255970, "epoch": 2876, "lr": 8.031213433923212e-07}
{"train_loss": 0.02614368125796318, "global_step": 255971, "epoch": 2876, "lr": 8.030178562423219e-07}
{"train_loss": 0.0404483862221241, "global_step": 255972, "epoch": 2876, "lr": 8.029143757062541e-07}
{"train_loss": 0.038381174206733704, "global_step": 255973, "epoch": 2876, "lr": 8.028109017841345e-07}
{"train_loss": 0.03714293614029884, "global_step": 255974, "epoch": 2876, "lr": 8.027074344759855e-07}
{"train_loss": 0.007541087456047535, "global_step": 255975, "epoch": 2876, "lr": 8.026039737818014e-07}
{"train_loss": 0.020982718095183372, "global_step": 255976, "epoch": 2876, "lr": 8.025005197016155e-07}
{"train_loss": 0.03365031257271767, "global_step": 255977, "epoch": 2876, "lr": 8.023970722354334e-07}
{"train_loss": 0.04053858667612076, "global_step": 255978, "epoch": 2876, "lr": 8.022936313832663e-07}
{"train_loss": 0.03994867950677872, "global_step": 255979, "epoch": 2876, "lr": 8.021901971451362e-07}
{"train_loss": 0.04272504150867462, "global_step": 255980, "epoch": 2876, "lr": 8.020867695210432e-07}
{"train_loss": 0.02656264416873455, "global_step": 255981, "epoch": 2876, "lr": 8.019833485110206e-07}
{"train_loss": 0.028219401836395264, "global_step": 255982, "epoch": 2876, "lr": 8.018799341150684e-07}
{"train_loss": 0.02315678261220455, "global_step": 255983, "epoch": 2876, "lr": 8.017765263331977e-07}
{"train_loss": 0.08080845326185226, "global_step": 255984, "epoch": 2876, "lr": 8.016731251654363e-07}
{"train_loss": 0.02618039958178997, "global_step": 255985, "epoch": 2876, "lr": 8.015697306117842e-07}
{"train_loss": 0.02410970628261566, "global_step": 255986, "epoch": 2876, "lr": 8.014663426722635e-07}
{"train_loss": 0.10045695304870605, "global_step": 255987, "epoch": 2876, "lr": 8.013629613468854e-07}
{"train_loss": 0.014908397570252419, "global_step": 255988, "epoch": 2876, "lr": 8.012595866356665e-07}
{"train_loss": 0.040924232453107834, "global_step": 255989, "epoch": 2876, "lr": 8.011562185386179e-07}
{"train_loss": 0.026184864342212677, "global_step": 255990, "epoch": 2876, "lr": 8.010528570557507e-07}
{"train_loss": 0.09769425541162491, "global_step": 255991, "epoch": 2876, "lr": 8.009495021870872e-07}
{"train_loss": 0.03813211992383003, "global_step": 255992, "epoch": 2876, "lr": 8.008461539326329e-07}
{"train_loss": 0.03891105577349663, "global_step": 255993, "epoch": 2876, "lr": 8.007428122924043e-07}
{"train_loss": 0.017177104949951172, "global_step": 255994, "epoch": 2876, "lr": 8.006394772664183e-07}
{"train_loss": 0.053830329328775406, "global_step": 255995, "epoch": 2876, "lr": 8.005361488546914e-07}
{"train_loss": 0.05726947262883186, "global_step": 255996, "epoch": 2876, "lr": 8.004328270572237e-07}
{"train_loss": 0.024918971583247185, "global_step": 255997, "epoch": 2876, "lr": 8.003295118740428e-07}
{"train_loss": 0.021190684288740158, "global_step": 255998, "epoch": 2876, "lr": 8.0022620330516e-07}
{"train_loss": 0.015454869717359543, "global_step": 255999, "epoch": 2876, "lr": 8.001229013505862e-07}
{"train_loss": 0.044551435858011246, "global_step": 256000, "epoch": 2876, "lr": 8.000196060103327e-07}
{"train_loss": 0.03247477859258652, "global_step": 256001, "epoch": 2876, "lr": 7.999163172844216e-07}
{"train_loss": 0.052703674882650375, "global_step": 256002, "epoch": 2876, "lr": 7.998130351728528e-07}
{"train_loss": 0.04047983139753342, "global_step": 256003, "epoch": 2876, "lr": 7.997097596756597e-07}
{"train_loss": 0.04762192443013191, "global_step": 256004, "epoch": 2876, "lr": 7.996064907928424e-07}
{"train_loss": 0.06943944841623306, "global_step": 256005, "epoch": 2876, "lr": 7.995032285244175e-07}
{"train_loss": 0.0729323998093605, "global_step": 256006, "epoch": 2876, "lr": 7.993999728704016e-07}
{"train_loss": 0.03794766962528229, "global_step": 256007, "epoch": 2876, "lr": 7.992967238308058e-07}
{"train_loss": 0.006691241171211004, "global_step": 256008, "epoch": 2876, "lr": 7.991934814056412e-07}
{"train_loss": 0.0928998589515686, "global_step": 256009, "epoch": 2876, "lr": 7.990902455949356e-07}
{"train_loss": 0.05608344450592995, "global_step": 256010, "epoch": 2876, "lr": 7.989870163986835e-07}
{"train_loss": 0.043472521007061005, "global_step": 256011, "epoch": 2876, "lr": 7.988837938169069e-07}
{"train_loss": 0.03308100625872612, "global_step": 256012, "epoch": 2876, "lr": 7.987805778496282e-07}
{"train_loss": 0.054815638810396194, "global_step": 256013, "epoch": 2876, "lr": 7.986773684968473e-07}
{"train_loss": 0.01900787465274334, "global_step": 256014, "epoch": 2876, "lr": 7.985741657585866e-07}
{"train_loss": 0.054290954023599625, "global_step": 256015, "epoch": 2876, "lr": 7.984709696348624e-07}
{"train_loss": 0.03686033561825752, "global_step": 256016, "epoch": 2876, "lr": 7.983677801256751e-07}
{"train_loss": 0.02247556857764721, "global_step": 256017, "epoch": 2876, "lr": 7.982645972310576e-07}
{"train_loss": 0.07901976257562637, "global_step": 256018, "epoch": 2876, "lr": 7.981614209510046e-07}
{"train_loss": 0.02894030325114727, "global_step": 256019, "epoch": 2876, "lr": 7.980582512855439e-07}
{"train_loss": 0.06608094274997711, "global_step": 256020, "epoch": 2876, "lr": 7.979550882346864e-07}
{"train_loss": 0.0151231549680233, "global_step": 256021, "epoch": 2876, "lr": 7.978519317984435e-07}
{"train_loss": 0.0615713857114315, "global_step": 256022, "epoch": 2876, "lr": 7.977487819768258e-07}
{"train_loss": 0.030510643497109413, "global_step": 256023, "epoch": 2876, "lr": 7.976456387698505e-07}
{"train_loss": 0.03393151983618736, "global_step": 256024, "epoch": 2876, "lr": 7.975425021775396e-07}
{"train_loss": 0.039325371384620667, "global_step": 256025, "epoch": 2876, "lr": 7.97439372199893e-07}
{"train_loss": 0.02945329062640667, "global_step": 256026, "epoch": 2876, "lr": 7.973362488369329e-07}
{"train_loss": 0.061955686658620834, "global_step": 256027, "epoch": 2876, "lr": 7.972331320886706e-07}
{"train_loss": 0.05336875841021538, "global_step": 256028, "epoch": 2876, "lr": 7.97130021955117e-07}
{"train_loss": 0.04767515882849693, "global_step": 256029, "epoch": 2876, "lr": 7.970269184362944e-07}
{"train_loss": 0.02341890148818493, "global_step": 256030, "epoch": 2876, "lr": 7.969238215322083e-07}
{"train_loss": 0.03670176863670349, "global_step": 256031, "epoch": 2876, "lr": 7.96820731242881e-07}
{"train_loss": 0.04049258306622505, "global_step": 256032, "epoch": 2876, "lr": 7.96717647568318e-07}
{"train_loss": 0.07767553627490997, "global_step": 256033, "epoch": 2876, "lr": 7.966145705085415e-07}
{"train_loss": 0.024608446285128593, "global_step": 256034, "epoch": 2876, "lr": 7.965115000635514e-07}
{"train_loss": 0.0404108427464962, "global_step": 256035, "epoch": 2876, "lr": 7.964084362333812e-07}
{"train_loss": 0.044612184166908264, "global_step": 256036, "epoch": 2876, "lr": 7.963053790180253e-07}
{"train_loss": 0.04966486245393753, "global_step": 256037, "epoch": 2876, "lr": 7.962023284175113e-07}
{"train_loss": 0.03608158230781555, "global_step": 256038, "epoch": 2876, "lr": 7.960992844318449e-07}
{"train_loss": 0.06862773001194, "global_step": 256039, "epoch": 2876, "lr": 7.959962470610482e-07}
{"train_loss": 0.021971795707941055, "global_step": 256040, "epoch": 2876, "lr": 7.958932163051269e-07}
{"train_loss": 0.015051421709358692, "global_step": 256041, "epoch": 2876, "lr": 7.957901921640976e-07}
{"train_loss": 0.04628319293260574, "global_step": 256042, "epoch": 2876, "lr": 7.956871746379713e-07}
{"train_loss": 0.06282074004411697, "global_step": 256043, "epoch": 2876, "lr": 7.955841637267758e-07}
{"train_loss": 0.04733671247959137, "global_step": 256044, "epoch": 2876, "lr": 7.954811594305057e-07}
{"train_loss": 0.06611788272857666, "global_step": 256045, "epoch": 2876, "lr": 7.953781617491829e-07}
{"train_loss": 0.01366709265857935, "global_step": 256046, "epoch": 2876, "lr": 7.952751706828243e-07}
{"train_loss": 0.06318070739507675, "global_step": 256047, "epoch": 2876, "lr": 7.95172186231441e-07}
{"train_loss": 0.021150289103388786, "global_step": 256048, "epoch": 2876, "lr": 7.950692083950495e-07}
{"train_loss": 0.01854567602276802, "global_step": 256049, "epoch": 2876, "lr": 7.949662371736555e-07}
{"train_loss": 0.019095104187726974, "global_step": 256050, "epoch": 2876, "lr": 7.948632725672811e-07}
{"train_loss": 0.06079169735312462, "global_step": 256051, "epoch": 2876, "lr": 7.947603145759374e-07}
{"train_loss": 0.040796155544269956, "global_step": 256052, "epoch": 2876, "lr": 7.946573631996412e-07, "val_loss": 9.651124000549316}
{"train_loss": 0.035363662987947464, "global_step": 256053, "epoch": 2877, "lr": 7.945544184384035e-07}
{"train_loss": 0.10102888941764832, "global_step": 256054, "epoch": 2877, "lr": 7.944514802922354e-07}
{"train_loss": 0.05309112370014191, "global_step": 256055, "epoch": 2877, "lr": 7.94348548761159e-07}
{"train_loss": 0.014803288504481316, "global_step": 256056, "epoch": 2877, "lr": 7.942456238451745e-07}
{"train_loss": 0.07411199808120728, "global_step": 256057, "epoch": 2877, "lr": 7.941427055443096e-07}
{"train_loss": 0.0353555828332901, "global_step": 256058, "epoch": 2877, "lr": 7.940397938585753e-07}
{"train_loss": 0.02381124347448349, "global_step": 256059, "epoch": 2877, "lr": 7.939368887879772e-07}
{"train_loss": 0.01329141017049551, "global_step": 256060, "epoch": 2877, "lr": 7.938339903325376e-07}
{"train_loss": 0.05299777537584305, "global_step": 256061, "epoch": 2877, "lr": 7.937310984922619e-07}
{"train_loss": 0.056873057037591934, "global_step": 256062, "epoch": 2877, "lr": 7.936282132671724e-07}
{"train_loss": 0.03922655060887337, "global_step": 256063, "epoch": 2877, "lr": 7.935253346572802e-07}
{"train_loss": 0.045061562210321426, "global_step": 256064, "epoch": 2877, "lr": 7.934224626625963e-07}
{"train_loss": 0.050737638026475906, "global_step": 256065, "epoch": 2877, "lr": 7.933195972831431e-07}
{"train_loss": 0.023619284853339195, "global_step": 256066, "epoch": 2877, "lr": 7.932167385189205e-07}
{"train_loss": 0.038216590881347656, "global_step": 256067, "epoch": 2877, "lr": 7.931138863699561e-07}
{"train_loss": 0.04997657984495163, "global_step": 256068, "epoch": 2877, "lr": 7.930110408362556e-07}
{"train_loss": 0.020306682214140892, "global_step": 256069, "epoch": 2877, "lr": 7.929082019178358e-07}
{"train_loss": 0.02053636498749256, "global_step": 256070, "epoch": 2877, "lr": 7.928053696147075e-07}
{"train_loss": 0.0374227799475193, "global_step": 256071, "epoch": 2877, "lr": 7.927025439268876e-07}
{"train_loss": 0.03268561512231827, "global_step": 256072, "epoch": 2877, "lr": 7.925997248543926e-07}
{"train_loss": 0.02387806959450245, "global_step": 256073, "epoch": 2877, "lr": 7.924969123972281e-07}
{"train_loss": 0.041602540761232376, "global_step": 256074, "epoch": 2877, "lr": 7.923941065554164e-07}
{"train_loss": 0.02792363613843918, "global_step": 256075, "epoch": 2877, "lr": 7.922913073289684e-07}
{"train_loss": 0.04313494637608528, "global_step": 256076, "epoch": 2877, "lr": 7.921885147178898e-07}
{"train_loss": 0.017896709963679314, "global_step": 256077, "epoch": 2877, "lr": 7.920857287222084e-07}
{"train_loss": 0.0206526480615139, "global_step": 256078, "epoch": 2877, "lr": 7.919829493419295e-07}
{"train_loss": 0.034225691109895706, "global_step": 256079, "epoch": 2877, "lr": 7.918801765770701e-07}
{"train_loss": 0.033196572214365005, "global_step": 256080, "epoch": 2877, "lr": 7.91777410427641e-07}
{"train_loss": 0.0429350920021534, "global_step": 256081, "epoch": 2877, "lr": 7.91674650893659e-07}
{"train_loss": 0.10169392824172974, "global_step": 256082, "epoch": 2877, "lr": 7.915718979751296e-07}
{"train_loss": 0.04683936387300491, "global_step": 256083, "epoch": 2877, "lr": 7.914691516720807e-07}
{"train_loss": 0.01097220741212368, "global_step": 256084, "epoch": 2877, "lr": 7.913664119845177e-07}
{"train_loss": 0.02538163959980011, "global_step": 256085, "epoch": 2877, "lr": 7.912636789124572e-07}
{"train_loss": 0.05571021884679794, "global_step": 256086, "epoch": 2877, "lr": 7.911609524559105e-07}
{"train_loss": 0.022068386897444725, "global_step": 256087, "epoch": 2877, "lr": 7.910582326148886e-07}
{"train_loss": 0.03092910349369049, "global_step": 256088, "epoch": 2877, "lr": 7.909555193894136e-07}
{"train_loss": 0.028441043570637703, "global_step": 256089, "epoch": 2877, "lr": 7.908528127794912e-07}
{"train_loss": 0.03483792021870613, "global_step": 256090, "epoch": 2877, "lr": 7.907501127851435e-07}
{"train_loss": 0.05294560641050339, "global_step": 256091, "epoch": 2877, "lr": 7.906474194063762e-07}
{"train_loss": 0.028333358466625214, "global_step": 256092, "epoch": 2877, "lr": 7.905447326432114e-07}
{"train_loss": 0.07680584490299225, "global_step": 256093, "epoch": 2877, "lr": 7.904420524956546e-07}
{"train_loss": 0.06798673421144485, "global_step": 256094, "epoch": 2877, "lr": 7.903393789637225e-07}
{"train_loss": 0.03521792218089104, "global_step": 256095, "epoch": 2877, "lr": 7.902367120474319e-07}
{"train_loss": 0.05597276985645294, "global_step": 256096, "epoch": 2877, "lr": 7.901340517467937e-07}
{"train_loss": 0.03340129181742668, "global_step": 256097, "epoch": 2877, "lr": 7.90031398061819e-07}
{"train_loss": 0.03750045597553253, "global_step": 256098, "epoch": 2877, "lr": 7.899287509925302e-07}
{"train_loss": 0.03378232195973396, "global_step": 256099, "epoch": 2877, "lr": 7.898261105389327e-07}
{"train_loss": 0.04936760291457176, "global_step": 256100, "epoch": 2877, "lr": 7.897234767010376e-07}
{"train_loss": 0.10140378028154373, "global_step": 256101, "epoch": 2877, "lr": 7.896208494788726e-07}
{"train_loss": 0.030874064192175865, "global_step": 256102, "epoch": 2877, "lr": 7.895182288724379e-07}
{"train_loss": 0.034892577677965164, "global_step": 256103, "epoch": 2877, "lr": 7.894156148817555e-07}
{"train_loss": 0.05195297300815582, "global_step": 256104, "epoch": 2877, "lr": 7.893130075068366e-07}
{"train_loss": 0.06965315341949463, "global_step": 256105, "epoch": 2877, "lr": 7.89210406747698e-07}
{"train_loss": 0.02830308862030506, "global_step": 256106, "epoch": 2877, "lr": 7.891078126043449e-07}
{"train_loss": 0.050890807062387466, "global_step": 256107, "epoch": 2877, "lr": 7.890052250767998e-07}
{"train_loss": 0.0333641842007637, "global_step": 256108, "epoch": 2877, "lr": 7.889026441650682e-07}
{"train_loss": 0.04394126683473587, "global_step": 256109, "epoch": 2877, "lr": 7.888000698691722e-07}
{"train_loss": 0.027919355779886246, "global_step": 256110, "epoch": 2877, "lr": 7.88697502189123e-07}
{"train_loss": 0.047402236610651016, "global_step": 256111, "epoch": 2877, "lr": 7.885949411249316e-07}
{"train_loss": 0.05835049971938133, "global_step": 256112, "epoch": 2877, "lr": 7.884923866766148e-07}
{"train_loss": 0.036446575075387955, "global_step": 256113, "epoch": 2877, "lr": 7.883898388441835e-07}
{"train_loss": 0.055039819329977036, "global_step": 256114, "epoch": 2877, "lr": 7.882872976276545e-07}
{"train_loss": 0.03231217339634895, "global_step": 256115, "epoch": 2877, "lr": 7.881847630270445e-07}
{"train_loss": 0.057962168008089066, "global_step": 256116, "epoch": 2877, "lr": 7.880822350423589e-07}
{"train_loss": 0.020229771733283997, "global_step": 256117, "epoch": 2877, "lr": 7.8797971367362e-07}
{"train_loss": 0.03561484441161156, "global_step": 256118, "epoch": 2877, "lr": 7.878771989208278e-07}
{"train_loss": 0.03138234093785286, "global_step": 256119, "epoch": 2877, "lr": 7.8777469078401e-07}
{"train_loss": 0.0626453310251236, "global_step": 256120, "epoch": 2877, "lr": 7.876721892631833e-07}
{"train_loss": 0.029361817985773087, "global_step": 256121, "epoch": 2877, "lr": 7.875696943583477e-07}
{"train_loss": 0.03524080291390419, "global_step": 256122, "epoch": 2877, "lr": 7.874672060695254e-07}
{"train_loss": 0.027582718059420586, "global_step": 256123, "epoch": 2877, "lr": 7.873647243967275e-07}
{"train_loss": 0.025121012702584267, "global_step": 256124, "epoch": 2877, "lr": 7.872622493399706e-07}
{"train_loss": 0.08018136769533157, "global_step": 256125, "epoch": 2877, "lr": 7.871597808992659e-07}
{"train_loss": 0.037540510296821594, "global_step": 256126, "epoch": 2877, "lr": 7.870573190746245e-07}
{"train_loss": 0.050066497176885605, "global_step": 256127, "epoch": 2877, "lr": 7.869548638660629e-07}
{"train_loss": 0.06702706217765808, "global_step": 256128, "epoch": 2877, "lr": 7.868524152736034e-07}
{"train_loss": 0.03438752517104149, "global_step": 256129, "epoch": 2877, "lr": 7.867499732972461e-07}
{"train_loss": 0.08374977856874466, "global_step": 256130, "epoch": 2877, "lr": 7.86647537937013e-07}
{"train_loss": 0.04714395850896835, "global_step": 256131, "epoch": 2877, "lr": 7.865451091929154e-07}
{"train_loss": 0.05652380734682083, "global_step": 256132, "epoch": 2877, "lr": 7.864426870649588e-07}
{"train_loss": 0.020430928096175194, "global_step": 256133, "epoch": 2877, "lr": 7.863402715531764e-07}
{"train_loss": 0.05855106562376022, "global_step": 256134, "epoch": 2877, "lr": 7.862378626575628e-07}
{"train_loss": 0.03415950387716293, "global_step": 256135, "epoch": 2877, "lr": 7.861354603781456e-07}
{"train_loss": 0.05379917100071907, "global_step": 256136, "epoch": 2877, "lr": 7.860330647149306e-07}
{"train_loss": 0.10357120633125305, "global_step": 256137, "epoch": 2877, "lr": 7.859306756679285e-07}
{"train_loss": 0.05608018860220909, "global_step": 256138, "epoch": 2877, "lr": 7.858282932371619e-07}
{"train_loss": 0.035416942089796066, "global_step": 256139, "epoch": 2877, "lr": 7.857259174226416e-07}
{"train_loss": 0.057212479412555695, "global_step": 256140, "epoch": 2877, "lr": 7.85623548224379e-07}
{"train_loss": 0.04355745485282681, "global_step": 256141, "epoch": 2877, "lr": 7.855211856423905e-07, "val_loss": 9.54609203338623}
{"train_loss": 0.039558086544275284, "global_step": 256142, "epoch": 2878, "lr": 7.854188296766874e-07}
{"train_loss": 0.035027291625738144, "global_step": 256143, "epoch": 2878, "lr": 7.853164803272916e-07}
{"train_loss": 0.04094085842370987, "global_step": 256144, "epoch": 2878, "lr": 7.852141375942035e-07}
{"train_loss": 0.018837139010429382, "global_step": 256145, "epoch": 2878, "lr": 7.85111801477445e-07}
{"train_loss": 0.014240670949220657, "global_step": 256146, "epoch": 2878, "lr": 7.850094719770274e-07}
{"train_loss": 0.052495189011096954, "global_step": 256147, "epoch": 2878, "lr": 7.849071490929671e-07}
{"train_loss": 0.050620049238204956, "global_step": 256148, "epoch": 2878, "lr": 7.848048328252755e-07}
{"train_loss": 0.06329537183046341, "global_step": 256149, "epoch": 2878, "lr": 7.84702523173969e-07}
{"train_loss": 0.06145671010017395, "global_step": 256150, "epoch": 2878, "lr": 7.846002201390534e-07}
{"train_loss": 0.09628480672836304, "global_step": 256151, "epoch": 2878, "lr": 7.844979237205563e-07}
{"train_loss": 0.045497629791498184, "global_step": 256152, "epoch": 2878, "lr": 7.843956339184777e-07}
{"train_loss": 0.09085628390312195, "global_step": 256153, "epoch": 2878, "lr": 7.842933507328398e-07}
{"train_loss": 0.03552534431219101, "global_step": 256154, "epoch": 2878, "lr": 7.841910741636538e-07}
{"train_loss": 0.062196314334869385, "global_step": 256155, "epoch": 2878, "lr": 7.840888042109307e-07}
{"train_loss": 0.05274740979075432, "global_step": 256156, "epoch": 2878, "lr": 7.839865408746927e-07}
{"train_loss": 0.0233706496655941, "global_step": 256157, "epoch": 2878, "lr": 7.838842841549399e-07}
{"train_loss": 0.03804861381649971, "global_step": 256158, "epoch": 2878, "lr": 7.837820340516999e-07}
{"train_loss": 0.018106039613485336, "global_step": 256159, "epoch": 2878, "lr": 7.836797905649785e-07}
{"train_loss": 0.020127393305301666, "global_step": 256160, "epoch": 2878, "lr": 7.83577553694792e-07}
{"train_loss": 0.018327683210372925, "global_step": 256161, "epoch": 2878, "lr": 7.834753234411519e-07}
{"train_loss": 0.01689128763973713, "global_step": 256162, "epoch": 2878, "lr": 7.8337309980408e-07}
{"train_loss": 0.03330749273300171, "global_step": 256163, "epoch": 2878, "lr": 7.832708827835766e-07}
{"train_loss": 0.057909075170755386, "global_step": 256164, "epoch": 2878, "lr": 7.831686723796694e-07}
{"train_loss": 0.021596405655145645, "global_step": 256165, "epoch": 2878, "lr": 7.830664685923583e-07}
{"train_loss": 0.05656047537922859, "global_step": 256166, "epoch": 2878, "lr": 7.829642714216657e-07}
{"train_loss": 0.08332491666078568, "global_step": 256167, "epoch": 2878, "lr": 7.828620808676079e-07}
{"train_loss": 0.04067293182015419, "global_step": 256168, "epoch": 2878, "lr": 7.827598969301908e-07}
{"train_loss": 0.051192719489336014, "global_step": 256169, "epoch": 2878, "lr": 7.826577196094365e-07}
{"train_loss": 0.011326550506055355, "global_step": 256170, "epoch": 2878, "lr": 7.825555489053449e-07}
{"train_loss": 0.03885969519615173, "global_step": 256171, "epoch": 2878, "lr": 7.824533848179493e-07}
{"train_loss": 0.07302486896514893, "global_step": 256172, "epoch": 2878, "lr": 7.823512273472445e-07}
{"train_loss": 0.04699220135807991, "global_step": 256173, "epoch": 2878, "lr": 7.822490764932577e-07}
{"train_loss": 0.0501740500330925, "global_step": 256174, "epoch": 2878, "lr": 7.821469322559893e-07}
{"train_loss": 0.03797563165426254, "global_step": 256175, "epoch": 2878, "lr": 7.820447946354726e-07}
{"train_loss": 0.04954887181520462, "global_step": 256176, "epoch": 2878, "lr": 7.819426636317018e-07}
{"train_loss": 0.03411468490958214, "global_step": 256177, "epoch": 2878, "lr": 7.818405392447048e-07}
{"train_loss": 0.04102274030447006, "global_step": 256178, "epoch": 2878, "lr": 7.817384214744872e-07}
{"train_loss": 0.06105043366551399, "global_step": 256179, "epoch": 2878, "lr": 7.816363103210656e-07}
{"train_loss": 0.06754482537508011, "global_step": 256180, "epoch": 2878, "lr": 7.81534205784451e-07}
{"train_loss": 0.052563052624464035, "global_step": 256181, "epoch": 2878, "lr": 7.814321078646603e-07}
{"train_loss": 0.05428597331047058, "global_step": 256182, "epoch": 2878, "lr": 7.813300165617044e-07}
{"train_loss": 0.04923710972070694, "global_step": 256183, "epoch": 2878, "lr": 7.812279318756055e-07}
{"train_loss": 0.056287914514541626, "global_step": 256184, "epoch": 2878, "lr": 7.811258538063637e-07}
{"train_loss": 0.018051426857709885, "global_step": 256185, "epoch": 2878, "lr": 7.810237823540012e-07}
{"train_loss": 0.017994439229369164, "global_step": 256186, "epoch": 2878, "lr": 7.80921717518529e-07}
{"train_loss": 0.04137822985649109, "global_step": 256187, "epoch": 2878, "lr": 7.80819659299964e-07}
{"train_loss": 0.009538335725665092, "global_step": 256188, "epoch": 2878, "lr": 7.80717607698317e-07}
{"train_loss": 0.08056609332561493, "global_step": 256189, "epoch": 2878, "lr": 7.806155627136047e-07}
{"train_loss": 0.041748739778995514, "global_step": 256190, "epoch": 2878, "lr": 7.805135243458384e-07}
{"train_loss": 0.057249389588832855, "global_step": 256191, "epoch": 2878, "lr": 7.804114925950289e-07}
{"train_loss": 0.056186024099588394, "global_step": 256192, "epoch": 2878, "lr": 7.803094674611989e-07}
{"train_loss": 0.04074868559837341, "global_step": 256193, "epoch": 2878, "lr": 7.802074489443478e-07}
{"train_loss": 0.05562962591648102, "global_step": 256194, "epoch": 2878, "lr": 7.801054370445038e-07}
{"train_loss": 0.037380363792181015, "global_step": 256195, "epoch": 2878, "lr": 7.800034317616723e-07}
{"train_loss": 0.09109105169773102, "global_step": 256196, "epoch": 2878, "lr": 7.799014330958754e-07}
{"train_loss": 0.02767348848283291, "global_step": 256197, "epoch": 2878, "lr": 7.797994410471133e-07}
{"train_loss": 0.029722368344664574, "global_step": 256198, "epoch": 2878, "lr": 7.796974556154135e-07}
{"train_loss": 0.03544774651527405, "global_step": 256199, "epoch": 2878, "lr": 7.795954768007763e-07}
{"train_loss": 0.047219447791576385, "global_step": 256200, "epoch": 2878, "lr": 7.794935046032292e-07}
{"train_loss": 0.04331289231777191, "global_step": 256201, "epoch": 2878, "lr": 7.793915390227724e-07}
{"train_loss": 0.08197785168886185, "global_step": 256202, "epoch": 2878, "lr": 7.792895800594335e-07}
{"train_loss": 0.07899811863899231, "global_step": 256203, "epoch": 2878, "lr": 7.791876277132182e-07}
{"train_loss": 0.007605799473822117, "global_step": 256204, "epoch": 2878, "lr": 7.790856819841374e-07}
{"train_loss": 0.03450990095734596, "global_step": 256205, "epoch": 2878, "lr": 7.78983742872208e-07}
{"train_loss": 0.09039968997240067, "global_step": 256206, "epoch": 2878, "lr": 7.78881810377452e-07}
{"train_loss": 0.08990266919136047, "global_step": 256207, "epoch": 2878, "lr": 7.787798844998694e-07}
{"train_loss": 0.022034838795661926, "global_step": 256208, "epoch": 2878, "lr": 7.786779652394826e-07}
{"train_loss": 0.0250612311065197, "global_step": 256209, "epoch": 2878, "lr": 7.785760525963026e-07}
{"train_loss": 0.03962773457169533, "global_step": 256210, "epoch": 2878, "lr": 7.784741465703405e-07}
{"train_loss": 0.06471921503543854, "global_step": 256211, "epoch": 2878, "lr": 7.783722471616129e-07}
{"train_loss": 0.04865144565701485, "global_step": 256212, "epoch": 2878, "lr": 7.782703543701308e-07}
{"train_loss": 0.0701986774802208, "global_step": 256213, "epoch": 2878, "lr": 7.781684681959167e-07}
{"train_loss": 0.03876499831676483, "global_step": 256214, "epoch": 2878, "lr": 7.78066588638976e-07}
{"train_loss": 0.030871864408254623, "global_step": 256215, "epoch": 2878, "lr": 7.779647156993197e-07}
{"train_loss": 0.030322253704071045, "global_step": 256216, "epoch": 2878, "lr": 7.778628493769702e-07}
{"train_loss": 0.047579098492860794, "global_step": 256217, "epoch": 2878, "lr": 7.777609896719385e-07}
{"train_loss": 0.037834614515304565, "global_step": 256218, "epoch": 2878, "lr": 7.7765913658423e-07}
{"train_loss": 0.07184837013483047, "global_step": 256219, "epoch": 2878, "lr": 7.775572901138728e-07}
{"train_loss": 0.04742714390158653, "global_step": 256220, "epoch": 2878, "lr": 7.774554502608722e-07}
{"train_loss": 0.08946932852268219, "global_step": 256221, "epoch": 2878, "lr": 7.773536170252393e-07}
{"train_loss": 0.025756273418664932, "global_step": 256222, "epoch": 2878, "lr": 7.77251790406991e-07}
{"train_loss": 0.07926174253225327, "global_step": 256223, "epoch": 2878, "lr": 7.771499704061436e-07}
{"train_loss": 0.043270789086818695, "global_step": 256224, "epoch": 2878, "lr": 7.770481570227083e-07}
{"train_loss": 0.030644966289401054, "global_step": 256225, "epoch": 2878, "lr": 7.76946350256702e-07}
{"train_loss": 0.035401858389377594, "global_step": 256226, "epoch": 2878, "lr": 7.768445501081301e-07}
{"train_loss": 0.03277546167373657, "global_step": 256227, "epoch": 2878, "lr": 7.767427565770091e-07}
{"train_loss": 0.05120295658707619, "global_step": 256228, "epoch": 2878, "lr": 7.766409696633614e-07}
{"train_loss": 0.026850126683712006, "global_step": 256229, "epoch": 2878, "lr": 7.765391893671924e-07}
{"train_loss": 0.04607638021868266, "global_step": 256230, "epoch": 2878, "lr": 7.764374156885191e-07, "val_loss": 9.59116268157959}
{"train_loss": 0.022854333743453026, "global_step": 256231, "epoch": 2879, "lr": 7.763356486273521e-07}
{"train_loss": 0.05829056352376938, "global_step": 256232, "epoch": 2879, "lr": 7.762338881837083e-07}
{"train_loss": 0.02702593244612217, "global_step": 256233, "epoch": 2879, "lr": 7.761321343575933e-07}
{"train_loss": 0.056879401206970215, "global_step": 256234, "epoch": 2879, "lr": 7.760303871490349e-07}
{"train_loss": 0.03720973804593086, "global_step": 256235, "epoch": 2879, "lr": 7.759286465580384e-07}
{"train_loss": 0.04812433943152428, "global_step": 256236, "epoch": 2879, "lr": 7.758269125846151e-07}
{"train_loss": 0.06889820098876953, "global_step": 256237, "epoch": 2879, "lr": 7.757251852287817e-07}
{"train_loss": 0.03613338619470596, "global_step": 256238, "epoch": 2879, "lr": 7.756234644905602e-07}
{"train_loss": 0.03589394688606262, "global_step": 256239, "epoch": 2879, "lr": 7.755217503699452e-07}
{"train_loss": 0.04586401209235191, "global_step": 256240, "epoch": 2879, "lr": 7.7542004286697e-07}
{"train_loss": 0.027173055335879326, "global_step": 256241, "epoch": 2879, "lr": 7.753183419816346e-07}
{"train_loss": 0.018618816509842873, "global_step": 256242, "epoch": 2879, "lr": 7.75216647713961e-07}
{"train_loss": 0.03544865921139717, "global_step": 256243, "epoch": 2879, "lr": 7.751149600639552e-07}
{"train_loss": 0.032287951558828354, "global_step": 256244, "epoch": 2879, "lr": 7.750132790316389e-07}
{"train_loss": 0.03164267539978027, "global_step": 256245, "epoch": 2879, "lr": 7.749116046170179e-07}
{"train_loss": 0.062100693583488464, "global_step": 256246, "epoch": 2879, "lr": 7.748099368201145e-07}
{"train_loss": 0.02586064673960209, "global_step": 256247, "epoch": 2879, "lr": 7.747082756409396e-07}
{"train_loss": 0.047646716237068176, "global_step": 256248, "epoch": 2879, "lr": 7.746066210794989e-07}
{"train_loss": 0.030257543548941612, "global_step": 256249, "epoch": 2879, "lr": 7.7450497313582e-07}
{"train_loss": 0.03241981565952301, "global_step": 256250, "epoch": 2879, "lr": 7.744033318099031e-07}
{"train_loss": 0.07108372449874878, "global_step": 256251, "epoch": 2879, "lr": 7.743016971017702e-07}
{"train_loss": 0.04172283783555031, "global_step": 256252, "epoch": 2879, "lr": 7.742000690114271e-07}
{"train_loss": 0.11165132373571396, "global_step": 256253, "epoch": 2879, "lr": 7.740984475389012e-07}
{"train_loss": 0.043004363775253296, "global_step": 256254, "epoch": 2879, "lr": 7.739968326841928e-07}
{"train_loss": 0.04629722982645035, "global_step": 256255, "epoch": 2879, "lr": 7.738952244473186e-07}
{"train_loss": 0.0635467916727066, "global_step": 256256, "epoch": 2879, "lr": 7.737936228282949e-07}
{"train_loss": 0.04517238587141037, "global_step": 256257, "epoch": 2879, "lr": 7.736920278271387e-07}
{"train_loss": 0.027788711711764336, "global_step": 256258, "epoch": 2879, "lr": 7.735904394438554e-07}
{"train_loss": 0.03139324113726616, "global_step": 256259, "epoch": 2879, "lr": 7.734888576784672e-07}
{"train_loss": 0.04564077779650688, "global_step": 256260, "epoch": 2879, "lr": 7.733872825309795e-07}
{"train_loss": 0.04096788167953491, "global_step": 256261, "epoch": 2879, "lr": 7.732857140014149e-07}
{"train_loss": 0.047550324350595474, "global_step": 256262, "epoch": 2879, "lr": 7.731841520897787e-07}
{"train_loss": 0.04041482135653496, "global_step": 256263, "epoch": 2879, "lr": 7.730825967960875e-07}
{"train_loss": 0.04170260950922966, "global_step": 256264, "epoch": 2879, "lr": 7.729810481203581e-07}
{"train_loss": 0.025486113503575325, "global_step": 256265, "epoch": 2879, "lr": 7.72879506062596e-07}
{"train_loss": 0.01604240946471691, "global_step": 256266, "epoch": 2879, "lr": 7.727779706228233e-07}
{"train_loss": 0.017755698412656784, "global_step": 256267, "epoch": 2879, "lr": 7.726764418010512e-07}
{"train_loss": 0.045529305934906006, "global_step": 256268, "epoch": 2879, "lr": 7.725749195972909e-07}
{"train_loss": 0.05029640719294548, "global_step": 256269, "epoch": 2879, "lr": 7.724734040115589e-07}
{"train_loss": 0.039314914494752884, "global_step": 256270, "epoch": 2879, "lr": 7.723718950438719e-07}
{"train_loss": 0.029304329305887222, "global_step": 256271, "epoch": 2879, "lr": 7.722703926942354e-07}
{"train_loss": 0.052142661064863205, "global_step": 256272, "epoch": 2879, "lr": 7.721688969626661e-07}
{"train_loss": 0.04155508801341057, "global_step": 256273, "epoch": 2879, "lr": 7.720674078491807e-07}
{"train_loss": 0.06450553983449936, "global_step": 256274, "epoch": 2879, "lr": 7.719659253537903e-07}
{"train_loss": 0.04794486612081528, "global_step": 256275, "epoch": 2879, "lr": 7.71864449476506e-07}
{"train_loss": 0.033453259617090225, "global_step": 256276, "epoch": 2879, "lr": 7.717629802173499e-07}
{"train_loss": 0.014798130840063095, "global_step": 256277, "epoch": 2879, "lr": 7.716615175763276e-07}
{"train_loss": 0.04931710660457611, "global_step": 256278, "epoch": 2879, "lr": 7.715600615534557e-07}
{"train_loss": 0.041390568017959595, "global_step": 256279, "epoch": 2879, "lr": 7.714586121487455e-07}
{"train_loss": 0.07068321853876114, "global_step": 256280, "epoch": 2879, "lr": 7.71357169362219e-07}
{"train_loss": 0.027305621653795242, "global_step": 256281, "epoch": 2879, "lr": 7.712557331938764e-07}
{"train_loss": 0.01669364422559738, "global_step": 256282, "epoch": 2879, "lr": 7.711543036437452e-07}
{"train_loss": 0.03451769798994064, "global_step": 256283, "epoch": 2879, "lr": 7.710528807118256e-07}
{"train_loss": 0.06847486644983292, "global_step": 256284, "epoch": 2879, "lr": 7.709514643981397e-07}
{"train_loss": 0.03054993785917759, "global_step": 256285, "epoch": 2879, "lr": 7.708500547027042e-07}
{"train_loss": 0.03592098131775856, "global_step": 256286, "epoch": 2879, "lr": 7.707486516255247e-07}
{"train_loss": 0.028070297092199326, "global_step": 256287, "epoch": 2879, "lr": 7.706472551666177e-07}
{"train_loss": 0.027135875076055527, "global_step": 256288, "epoch": 2879, "lr": 7.705458653259945e-07}
{"train_loss": 0.07253766059875488, "global_step": 256289, "epoch": 2879, "lr": 7.704444821036771e-07}
{"train_loss": 0.03215654194355011, "global_step": 256290, "epoch": 2879, "lr": 7.703431054996657e-07}
{"train_loss": 0.045595694333314896, "global_step": 256291, "epoch": 2879, "lr": 7.702417355139879e-07}
{"train_loss": 0.05242922157049179, "global_step": 256292, "epoch": 2879, "lr": 7.701403721466493e-07}
{"train_loss": 0.026435839012265205, "global_step": 256293, "epoch": 2879, "lr": 7.700390153976666e-07}
{"train_loss": 0.05429593101143837, "global_step": 256294, "epoch": 2879, "lr": 7.699376652670509e-07}
{"train_loss": 0.02515033259987831, "global_step": 256295, "epoch": 2879, "lr": 7.698363217548188e-07}
{"train_loss": 0.028676988556981087, "global_step": 256296, "epoch": 2879, "lr": 7.697349848609758e-07}
{"train_loss": 0.03485514968633652, "global_step": 256297, "epoch": 2879, "lr": 7.696336545855499e-07}
{"train_loss": 0.027579592540860176, "global_step": 256298, "epoch": 2879, "lr": 7.695323309285407e-07}
{"train_loss": 0.02623918652534485, "global_step": 256299, "epoch": 2879, "lr": 7.694310138899708e-07}
{"train_loss": 0.035607509315013885, "global_step": 256300, "epoch": 2879, "lr": 7.69329703469851e-07}
{"train_loss": 0.046850044280290604, "global_step": 256301, "epoch": 2879, "lr": 7.692283996681926e-07}
{"train_loss": 0.03280416876077652, "global_step": 256302, "epoch": 2879, "lr": 7.691271024850122e-07}
{"train_loss": 0.040138550102710724, "global_step": 256303, "epoch": 2879, "lr": 7.690258119203209e-07}
{"train_loss": 0.11804553866386414, "global_step": 256304, "epoch": 2879, "lr": 7.689245279741408e-07}
{"train_loss": 0.06474032253026962, "global_step": 256305, "epoch": 2879, "lr": 7.688232506464721e-07}
{"train_loss": 0.04651854559779167, "global_step": 256306, "epoch": 2879, "lr": 7.687219799373424e-07}
{"train_loss": 0.04508598893880844, "global_step": 256307, "epoch": 2879, "lr": 7.686207158467462e-07}
{"train_loss": 0.061407070606946945, "global_step": 256308, "epoch": 2879, "lr": 7.685194583747224e-07}
{"train_loss": 0.02773543819785118, "global_step": 256309, "epoch": 2879, "lr": 7.684182075212598e-07}
{"train_loss": 0.04561404883861542, "global_step": 256310, "epoch": 2879, "lr": 7.683169632863918e-07}
{"train_loss": 0.04202669858932495, "global_step": 256311, "epoch": 2879, "lr": 7.682157256701183e-07}
{"train_loss": 0.07434214651584625, "global_step": 256312, "epoch": 2879, "lr": 7.681144946724616e-07}
{"train_loss": 0.01317371055483818, "global_step": 256313, "epoch": 2879, "lr": 7.680132702934273e-07}
{"train_loss": 0.043867338448762894, "global_step": 256314, "epoch": 2879, "lr": 7.679120525330374e-07}
{"train_loss": 0.028045490384101868, "global_step": 256315, "epoch": 2879, "lr": 7.678108413912977e-07}
{"train_loss": 0.03477664664387703, "global_step": 256316, "epoch": 2879, "lr": 7.677096368682301e-07}
{"train_loss": 0.033981803804636, "global_step": 256317, "epoch": 2879, "lr": 7.676084389638405e-07}
{"train_loss": 0.06248454749584198, "global_step": 256318, "epoch": 2879, "lr": 7.675072476781453e-07}
{"train_loss": 0.04230375677933184, "global_step": 256319, "epoch": 2879, "lr": 7.674060630111613e-07, "val_loss": 9.581679344177246}
{"train_loss": 0.03008200228214264, "global_step": 256320, "epoch": 2880, "lr": 7.67304884962905e-07}
{"train_loss": 0.03612604737281799, "global_step": 256321, "epoch": 2880, "lr": 7.672037135333765e-07}
{"train_loss": 0.06171450391411781, "global_step": 256322, "epoch": 2880, "lr": 7.67102548722598e-07}
{"train_loss": 0.03256130591034889, "global_step": 256323, "epoch": 2880, "lr": 7.670013905305862e-07}
{"train_loss": 0.050460800528526306, "global_step": 256324, "epoch": 2880, "lr": 7.669002389573465e-07}
{"train_loss": 0.01313259731978178, "global_step": 256325, "epoch": 2880, "lr": 7.667990940029013e-07}
{"train_loss": 0.02410309389233589, "global_step": 256326, "epoch": 2880, "lr": 7.66697955667256e-07}
{"train_loss": 0.01801973208785057, "global_step": 256327, "epoch": 2880, "lr": 7.665968239504328e-07}
{"train_loss": 0.02296086587011814, "global_step": 256328, "epoch": 2880, "lr": 7.664956988524374e-07}
{"train_loss": 0.024494357407093048, "global_step": 256329, "epoch": 2880, "lr": 7.663945803732864e-07}
{"train_loss": 0.06842977553606033, "global_step": 256330, "epoch": 2880, "lr": 7.662934685129963e-07}
{"train_loss": 0.03128530830144882, "global_step": 256331, "epoch": 2880, "lr": 7.661923632715784e-07}
{"train_loss": 0.03688136488199234, "global_step": 256332, "epoch": 2880, "lr": 7.660912646490382e-07}
{"train_loss": 0.06757216155529022, "global_step": 256333, "epoch": 2880, "lr": 7.659901726454033e-07}
{"train_loss": 0.0499001145362854, "global_step": 256334, "epoch": 2880, "lr": 7.658890872606794e-07}
{"train_loss": 0.018778204917907715, "global_step": 256335, "epoch": 2880, "lr": 7.657880084948832e-07}
{"train_loss": 0.059438079595565796, "global_step": 256336, "epoch": 2880, "lr": 7.656869363480257e-07}
{"train_loss": 0.05560462176799774, "global_step": 256337, "epoch": 2880, "lr": 7.65585870820118e-07}
{"train_loss": 0.06961692869663239, "global_step": 256338, "epoch": 2880, "lr": 7.654848119111823e-07}
{"train_loss": 0.027122005820274353, "global_step": 256339, "epoch": 2880, "lr": 7.653837596212243e-07}
{"train_loss": 0.054884955286979675, "global_step": 256340, "epoch": 2880, "lr": 7.652827139502605e-07}
{"train_loss": 0.01224568858742714, "global_step": 256341, "epoch": 2880, "lr": 7.651816748983076e-07}
{"train_loss": 0.02120986580848694, "global_step": 256342, "epoch": 2880, "lr": 7.650806424653712e-07}
{"train_loss": 0.03342312574386597, "global_step": 256343, "epoch": 2880, "lr": 7.649796166514733e-07}
{"train_loss": 0.01580769009888172, "global_step": 256344, "epoch": 2880, "lr": 7.648785974566252e-07}
{"train_loss": 0.053427860140800476, "global_step": 256345, "epoch": 2880, "lr": 7.647775848808325e-07}
{"train_loss": 0.07278413325548172, "global_step": 256346, "epoch": 2880, "lr": 7.646765789241228e-07}
{"train_loss": 0.035726532340049744, "global_step": 256347, "epoch": 2880, "lr": 7.645755795864962e-07}
{"train_loss": 0.04336724430322647, "global_step": 256348, "epoch": 2880, "lr": 7.644745868679748e-07}
{"train_loss": 0.04337427392601967, "global_step": 256349, "epoch": 2880, "lr": 7.643736007685697e-07}
{"train_loss": 0.0381365530192852, "global_step": 256350, "epoch": 2880, "lr": 7.642726212882922e-07}
{"train_loss": 0.02558913081884384, "global_step": 256351, "epoch": 2880, "lr": 7.641716484271588e-07}
{"train_loss": 0.05838058143854141, "global_step": 256352, "epoch": 2880, "lr": 7.640706821851861e-07}
{"train_loss": 0.07312449812889099, "global_step": 256353, "epoch": 2880, "lr": 7.639697225623799e-07}
{"train_loss": 0.04577363282442093, "global_step": 256354, "epoch": 2880, "lr": 7.638687695587621e-07}
{"train_loss": 0.050723597407341, "global_step": 256355, "epoch": 2880, "lr": 7.637678231743384e-07}
{"train_loss": 0.02773122489452362, "global_step": 256356, "epoch": 2880, "lr": 7.636668834091254e-07}
{"train_loss": 0.059150632470846176, "global_step": 256357, "epoch": 2880, "lr": 7.635659502631398e-07}
{"train_loss": 0.04606829956173897, "global_step": 256358, "epoch": 2880, "lr": 7.634650237363983e-07}
{"train_loss": 0.04691672697663307, "global_step": 256359, "epoch": 2880, "lr": 7.633641038289008e-07}
{"train_loss": 0.05183148384094238, "global_step": 256360, "epoch": 2880, "lr": 7.632631905406695e-07}
{"train_loss": 0.038540761917829514, "global_step": 256361, "epoch": 2880, "lr": 7.631622838717156e-07}
{"train_loss": 0.056931812316179276, "global_step": 256362, "epoch": 2880, "lr": 7.630613838220613e-07}
{"train_loss": 0.0463523231446743, "global_step": 256363, "epoch": 2880, "lr": 7.629604903917121e-07}
{"train_loss": 0.04330449551343918, "global_step": 256364, "epoch": 2880, "lr": 7.62859603580679e-07}
{"train_loss": 0.035454265773296356, "global_step": 256365, "epoch": 2880, "lr": 7.627587233889788e-07}
{"train_loss": 0.08396949619054794, "global_step": 256366, "epoch": 2880, "lr": 7.626578498166282e-07}
{"train_loss": 0.0538102388381958, "global_step": 256367, "epoch": 2880, "lr": 7.625569828636381e-07}
{"train_loss": 0.027483830228447914, "global_step": 256368, "epoch": 2880, "lr": 7.624561225300197e-07}
{"train_loss": 0.03942866995930672, "global_step": 256369, "epoch": 2880, "lr": 7.623552688157898e-07}
{"train_loss": 0.09683135896921158, "global_step": 256370, "epoch": 2880, "lr": 7.622544217209593e-07}
{"train_loss": 0.008426565676927567, "global_step": 256371, "epoch": 2880, "lr": 7.621535812455505e-07}
{"train_loss": 0.04473601654171944, "global_step": 256372, "epoch": 2880, "lr": 7.620527473895634e-07}
{"train_loss": 0.0226188525557518, "global_step": 256373, "epoch": 2880, "lr": 7.6195192015302e-07}
{"train_loss": 0.0686907023191452, "global_step": 256374, "epoch": 2880, "lr": 7.618510995359319e-07}
{"train_loss": 0.03931752219796181, "global_step": 256375, "epoch": 2880, "lr": 7.617502855383096e-07}
{"train_loss": 0.025512393563985825, "global_step": 256376, "epoch": 2880, "lr": 7.616494781601757e-07}
{"train_loss": 0.030512621626257896, "global_step": 256377, "epoch": 2880, "lr": 7.615486774015357e-07}
{"train_loss": 0.055261094123125076, "global_step": 256378, "epoch": 2880, "lr": 7.614478832624061e-07}
{"train_loss": 0.0695413202047348, "global_step": 256379, "epoch": 2880, "lr": 7.613470957427982e-07}
{"train_loss": 0.04410260543227196, "global_step": 256380, "epoch": 2880, "lr": 7.612463148427284e-07}
{"train_loss": 0.07187318056821823, "global_step": 256381, "epoch": 2880, "lr": 7.611455405622081e-07}
{"train_loss": 0.05836108699440956, "global_step": 256382, "epoch": 2880, "lr": 7.610447729012537e-07}
{"train_loss": 0.05711307004094124, "global_step": 256383, "epoch": 2880, "lr": 7.60944011859871e-07}
{"train_loss": 0.054144054651260376, "global_step": 256384, "epoch": 2880, "lr": 7.608432574380819e-07}
{"train_loss": 0.035959579050540924, "global_step": 256385, "epoch": 2880, "lr": 7.607425096358977e-07}
{"train_loss": 0.019925402477383614, "global_step": 256386, "epoch": 2880, "lr": 7.606417684533351e-07}
{"train_loss": 0.08515633642673492, "global_step": 256387, "epoch": 2880, "lr": 7.605410338903995e-07}
{"train_loss": 0.033127620816230774, "global_step": 256388, "epoch": 2880, "lr": 7.604403059471132e-07}
{"train_loss": 0.05131011828780174, "global_step": 256389, "epoch": 2880, "lr": 7.603395846234818e-07}
{"train_loss": 0.009107270278036594, "global_step": 256390, "epoch": 2880, "lr": 7.602388699195217e-07}
{"train_loss": 0.05475945398211479, "global_step": 256391, "epoch": 2880, "lr": 7.601381618352498e-07}
{"train_loss": 0.051674578338861465, "global_step": 256392, "epoch": 2880, "lr": 7.600374603706773e-07}
{"train_loss": 0.07317089289426804, "global_step": 256393, "epoch": 2880, "lr": 7.59936765525815e-07}
{"train_loss": 0.09055023640394211, "global_step": 256394, "epoch": 2880, "lr": 7.598360773006797e-07}
{"train_loss": 0.044734708964824677, "global_step": 256395, "epoch": 2880, "lr": 7.597353956952879e-07}
{"train_loss": 0.05392693355679512, "global_step": 256396, "epoch": 2880, "lr": 7.59634720709651e-07}
{"train_loss": 0.059631381183862686, "global_step": 256397, "epoch": 2880, "lr": 7.595340523437744e-07}
{"train_loss": 0.06601565331220627, "global_step": 256398, "epoch": 2880, "lr": 7.594333905976858e-07}
{"train_loss": 0.0963268131017685, "global_step": 256399, "epoch": 2880, "lr": 7.593327354713797e-07}
{"train_loss": 0.03404078632593155, "global_step": 256400, "epoch": 2880, "lr": 7.59232086964895e-07}
{"train_loss": 0.022980159148573875, "global_step": 256401, "epoch": 2880, "lr": 7.591314450782261e-07}
{"train_loss": 0.025911739096045494, "global_step": 256402, "epoch": 2880, "lr": 7.590308098113841e-07}
{"train_loss": 0.014289028942584991, "global_step": 256403, "epoch": 2880, "lr": 7.589301811644023e-07}
{"train_loss": 0.07341533899307251, "global_step": 256404, "epoch": 2880, "lr": 7.588295591372752e-07}
{"train_loss": 0.010948832146823406, "global_step": 256405, "epoch": 2880, "lr": 7.587289437300249e-07}
{"train_loss": 0.05780612304806709, "global_step": 256406, "epoch": 2880, "lr": 7.58628334942657e-07}
{"train_loss": 0.06805625557899475, "global_step": 256407, "epoch": 2880, "lr": 7.585277327751994e-07}
{"train_loss": 0.045523385538227774, "global_step": 256408, "epoch": 2880, "lr": 7.584271372276575e-07, "val_loss": 9.585257530212402, "train_action_mse_error": 25.141647338867188}
{"train_loss": 0.04430079460144043, "global_step": 256409, "epoch": 2881, "lr": 7.583265483000368e-07}
{"train_loss": 0.026192322373390198, "global_step": 256410, "epoch": 2881, "lr": 7.582259659923707e-07}
{"train_loss": 0.0361107736825943, "global_step": 256411, "epoch": 2881, "lr": 7.581253903046537e-07}
{"train_loss": 0.0572236105799675, "global_step": 256412, "epoch": 2881, "lr": 7.580248212369078e-07}
{"train_loss": 0.04484713822603226, "global_step": 256413, "epoch": 2881, "lr": 7.579242587891444e-07}
{"train_loss": 0.028022024780511856, "global_step": 256414, "epoch": 2881, "lr": 7.578237029613799e-07}
{"train_loss": 0.02137714996933937, "global_step": 256415, "epoch": 2881, "lr": 7.5772315375362e-07}
{"train_loss": 0.06311575323343277, "global_step": 256416, "epoch": 2881, "lr": 7.576226111658924e-07}
{"train_loss": 0.042954955250024796, "global_step": 256417, "epoch": 2881, "lr": 7.575220751981971e-07}
{"train_loss": 0.056846145540475845, "global_step": 256418, "epoch": 2881, "lr": 7.574215458505562e-07}
{"train_loss": 0.07568205893039703, "global_step": 256419, "epoch": 2881, "lr": 7.573210231229755e-07}
{"train_loss": 0.025242306292057037, "global_step": 256420, "epoch": 2881, "lr": 7.57220507015477e-07}
{"train_loss": 0.06425822526216507, "global_step": 256421, "epoch": 2881, "lr": 7.571199975280663e-07}
{"train_loss": 0.034862227737903595, "global_step": 256422, "epoch": 2881, "lr": 7.570194946607656e-07}
{"train_loss": 0.047542132437229156, "global_step": 256423, "epoch": 2881, "lr": 7.569189984135805e-07}
{"train_loss": 0.03595982491970062, "global_step": 256424, "epoch": 2881, "lr": 7.568185087865276e-07}
{"train_loss": 0.0491332933306694, "global_step": 256425, "epoch": 2881, "lr": 7.567180257796181e-07}
{"train_loss": 0.06557024270296097, "global_step": 256426, "epoch": 2881, "lr": 7.56617549392874e-07}
{"train_loss": 0.04208977147936821, "global_step": 256427, "epoch": 2881, "lr": 7.565170796262955e-07}
{"train_loss": 0.035598255693912506, "global_step": 256428, "epoch": 2881, "lr": 7.564166164799047e-07}
{"train_loss": 0.034695930778980255, "global_step": 256429, "epoch": 2881, "lr": 7.563161599537128e-07}
{"train_loss": 0.047524552792310715, "global_step": 256430, "epoch": 2881, "lr": 7.562157100477363e-07}
{"train_loss": 0.08549255132675171, "global_step": 256431, "epoch": 2881, "lr": 7.561152667619864e-07}
{"train_loss": 0.03933326154947281, "global_step": 256432, "epoch": 2881, "lr": 7.560148300964742e-07}
{"train_loss": 0.045917682349681854, "global_step": 256433, "epoch": 2881, "lr": 7.55914400051222e-07}
{"train_loss": 0.04207553714513779, "global_step": 256434, "epoch": 2881, "lr": 7.558139766262296e-07}
{"train_loss": 0.032194506376981735, "global_step": 256435, "epoch": 2881, "lr": 7.557135598215248e-07}
{"train_loss": 0.02655479684472084, "global_step": 256436, "epoch": 2881, "lr": 7.556131496371078e-07}
{"train_loss": 0.014797637239098549, "global_step": 256437, "epoch": 2881, "lr": 7.555127460730005e-07}
{"train_loss": 0.04252968728542328, "global_step": 256438, "epoch": 2881, "lr": 7.554123491292143e-07}
{"train_loss": 0.06014321371912956, "global_step": 256439, "epoch": 2881, "lr": 7.5531195880576e-07}
{"train_loss": 0.03159565106034279, "global_step": 256440, "epoch": 2881, "lr": 7.5521157510266e-07}
{"train_loss": 0.04504893720149994, "global_step": 256441, "epoch": 2881, "lr": 7.551111980199199e-07}
{"train_loss": 0.03998513147234917, "global_step": 256442, "epoch": 2881, "lr": 7.550108275575507e-07}
{"train_loss": 0.03078901208937168, "global_step": 256443, "epoch": 2881, "lr": 7.549104637155746e-07}
{"train_loss": 0.043986380100250244, "global_step": 256444, "epoch": 2881, "lr": 7.548101064939972e-07}
{"train_loss": 0.07586050778627396, "global_step": 256445, "epoch": 2881, "lr": 7.547097558928351e-07}
{"train_loss": 0.02233177050948143, "global_step": 256446, "epoch": 2881, "lr": 7.54609411912105e-07}
{"train_loss": 0.06754585355520248, "global_step": 256447, "epoch": 2881, "lr": 7.545090745518125e-07}
{"train_loss": 0.055566925555467606, "global_step": 256448, "epoch": 2881, "lr": 7.544087438119851e-07}
{"train_loss": 0.05183025449514389, "global_step": 256449, "epoch": 2881, "lr": 7.543084196926175e-07}
{"train_loss": 0.04386952519416809, "global_step": 256450, "epoch": 2881, "lr": 7.542081021937375e-07}
{"train_loss": 0.09614332020282745, "global_step": 256451, "epoch": 2881, "lr": 7.54107791315356e-07}
{"train_loss": 0.03538388013839722, "global_step": 256452, "epoch": 2881, "lr": 7.540074870574843e-07}
{"train_loss": 0.046343524008989334, "global_step": 256453, "epoch": 2881, "lr": 7.539071894201277e-07}
{"train_loss": 0.01122212689369917, "global_step": 256454, "epoch": 2881, "lr": 7.538068984033197e-07}
{"train_loss": 0.0491262711584568, "global_step": 256455, "epoch": 2881, "lr": 7.537066140070548e-07}
{"train_loss": 0.04529031366109848, "global_step": 256456, "epoch": 2881, "lr": 7.536063362313551e-07}
{"train_loss": 0.031121566891670227, "global_step": 256457, "epoch": 2881, "lr": 7.535060650762315e-07}
{"train_loss": 0.028980519622564316, "global_step": 256458, "epoch": 2881, "lr": 7.534058005417011e-07}
{"train_loss": 0.026276452466845512, "global_step": 256459, "epoch": 2881, "lr": 7.533055426277746e-07}
{"train_loss": 0.035013049840927124, "global_step": 256460, "epoch": 2881, "lr": 7.532052913344689e-07}
{"train_loss": 0.03992696478962898, "global_step": 256461, "epoch": 2881, "lr": 7.531050466617895e-07}
{"train_loss": 0.09028507024049759, "global_step": 256462, "epoch": 2881, "lr": 7.530048086097585e-07}
{"train_loss": 0.0563650019466877, "global_step": 256463, "epoch": 2881, "lr": 7.529045771783815e-07}
{"train_loss": 0.030545536428689957, "global_step": 256464, "epoch": 2881, "lr": 7.528043523676753e-07}
{"train_loss": 0.03520897403359413, "global_step": 256465, "epoch": 2881, "lr": 7.52704134177662e-07}
{"train_loss": 0.049192141741514206, "global_step": 256466, "epoch": 2881, "lr": 7.526039226083414e-07}
{"train_loss": 0.06596486270427704, "global_step": 256467, "epoch": 2881, "lr": 7.52503717659736e-07}
{"train_loss": 0.021523943170905113, "global_step": 256468, "epoch": 2881, "lr": 7.524035193318513e-07}
{"train_loss": 0.02751968614757061, "global_step": 256469, "epoch": 2881, "lr": 7.523033276247149e-07}
{"train_loss": 0.058176882565021515, "global_step": 256470, "epoch": 2881, "lr": 7.522031425383213e-07}
{"train_loss": 0.06846056133508682, "global_step": 256471, "epoch": 2881, "lr": 7.521029640727039e-07}
{"train_loss": 0.0219428651034832, "global_step": 256472, "epoch": 2881, "lr": 7.520027922278572e-07}
{"train_loss": 0.024494152516126633, "global_step": 256473, "epoch": 2881, "lr": 7.519026270038087e-07}
{"train_loss": 0.05486977472901344, "global_step": 256474, "epoch": 2881, "lr": 7.518024684005642e-07}
{"train_loss": 0.04945679008960724, "global_step": 256475, "epoch": 2881, "lr": 7.517023164181402e-07}
{"train_loss": 0.0667109340429306, "global_step": 256476, "epoch": 2881, "lr": 7.51602171056548e-07}
{"train_loss": 0.039599984884262085, "global_step": 256477, "epoch": 2881, "lr": 7.515020323158039e-07}
{"train_loss": 0.024653980508446693, "global_step": 256478, "epoch": 2881, "lr": 7.514019001959194e-07}
{"train_loss": 0.024246426299214363, "global_step": 256479, "epoch": 2881, "lr": 7.513017746969109e-07}
{"train_loss": 0.019793890416622162, "global_step": 256480, "epoch": 2881, "lr": 7.512016558187896e-07}
{"train_loss": 0.12460149824619293, "global_step": 256481, "epoch": 2881, "lr": 7.511015435615665e-07}
{"train_loss": 0.042073555290699005, "global_step": 256482, "epoch": 2881, "lr": 7.510014379252584e-07}
{"train_loss": 0.017093677073717117, "global_step": 256483, "epoch": 2881, "lr": 7.509013389098762e-07}
{"train_loss": 0.07727760076522827, "global_step": 256484, "epoch": 2881, "lr": 7.508012465154369e-07}
{"train_loss": 0.06603275239467621, "global_step": 256485, "epoch": 2881, "lr": 7.507011607419511e-07}
{"train_loss": 0.017757641151547432, "global_step": 256486, "epoch": 2881, "lr": 7.50601081589436e-07}
{"train_loss": 0.014162909239530563, "global_step": 256487, "epoch": 2881, "lr": 7.505010090579023e-07}
{"train_loss": 0.04558936879038811, "global_step": 256488, "epoch": 2881, "lr": 7.504009431473613e-07}
{"train_loss": 0.023845059797167778, "global_step": 256489, "epoch": 2881, "lr": 7.50300883857824e-07}
{"train_loss": 0.03577200695872307, "global_step": 256490, "epoch": 2881, "lr": 7.502008311893183e-07}
{"train_loss": 0.04406851902604103, "global_step": 256491, "epoch": 2881, "lr": 7.501007851418384e-07}
{"train_loss": 0.015807470306754112, "global_step": 256492, "epoch": 2881, "lr": 7.500007457154179e-07}
{"train_loss": 0.05049344152212143, "global_step": 256493, "epoch": 2881, "lr": 7.499007129100511e-07}
{"train_loss": 0.05222400650382042, "global_step": 256494, "epoch": 2881, "lr": 7.498006867257656e-07}
{"train_loss": 0.03864274546504021, "global_step": 256495, "epoch": 2881, "lr": 7.497006671625617e-07}
{"train_loss": 0.053286708891391754, "global_step": 256496, "epoch": 2881, "lr": 7.49600654220467e-07}
{"train_loss": 0.04387411507620905, "global_step": 256497, "epoch": 2881, "lr": 7.495006478994814e-07, "val_loss": 9.597769737243652}
{"train_loss": 0.034123413264751434, "global_step": 256498, "epoch": 2882, "lr": 7.49400648199633e-07}
{"train_loss": 0.06878707557916641, "global_step": 256499, "epoch": 2882, "lr": 7.493006551209214e-07}
{"train_loss": 0.04334966093301773, "global_step": 256500, "epoch": 2882, "lr": 7.492006686633634e-07}
{"train_loss": 0.018155135214328766, "global_step": 256501, "epoch": 2882, "lr": 7.491006888269869e-07}
{"train_loss": 0.06430057436227798, "global_step": 256502, "epoch": 2882, "lr": 7.490007156117807e-07}
{"train_loss": 0.013556338846683502, "global_step": 256503, "epoch": 2882, "lr": 7.48900749017778e-07}
{"train_loss": 0.009463761933147907, "global_step": 256504, "epoch": 2882, "lr": 7.488007890449844e-07}
{"train_loss": 0.010409895330667496, "global_step": 256505, "epoch": 2882, "lr": 7.487008356934166e-07}
{"train_loss": 0.01995326206088066, "global_step": 256506, "epoch": 2882, "lr": 7.486008889630802e-07}
{"train_loss": 0.06713799387216568, "global_step": 256507, "epoch": 2882, "lr": 7.485009488539973e-07}
{"train_loss": 0.031161939725279808, "global_step": 256508, "epoch": 2882, "lr": 7.484010153661791e-07}
{"train_loss": 0.024326691403985023, "global_step": 256509, "epoch": 2882, "lr": 7.483010884996366e-07}
{"train_loss": 0.0570838637650013, "global_step": 256510, "epoch": 2882, "lr": 7.482011682543865e-07}
{"train_loss": 0.030235817655920982, "global_step": 256511, "epoch": 2882, "lr": 7.4810125463044e-07}
{"train_loss": 0.008435982279479504, "global_step": 256512, "epoch": 2882, "lr": 7.480013476278136e-07}
{"train_loss": 0.02634412981569767, "global_step": 256513, "epoch": 2882, "lr": 7.479014472465129e-07}
{"train_loss": 0.04854199290275574, "global_step": 256514, "epoch": 2882, "lr": 7.478015534865602e-07}
{"train_loss": 0.04420990124344826, "global_step": 256515, "epoch": 2882, "lr": 7.477016663479608e-07}
{"train_loss": 0.019067462533712387, "global_step": 256516, "epoch": 2882, "lr": 7.476017858307372e-07}
{"train_loss": 0.028612913563847542, "global_step": 256517, "epoch": 2882, "lr": 7.475019119348947e-07}
{"train_loss": 0.01662655733525753, "global_step": 256518, "epoch": 2882, "lr": 7.474020446604502e-07}
{"train_loss": 0.04975344240665436, "global_step": 256519, "epoch": 2882, "lr": 7.473021840074201e-07}
{"train_loss": 0.025450320914387703, "global_step": 256520, "epoch": 2882, "lr": 7.472023299758157e-07}
{"train_loss": 0.0587509348988533, "global_step": 256521, "epoch": 2882, "lr": 7.471024825656425e-07}
{"train_loss": 0.03121102787554264, "global_step": 256522, "epoch": 2882, "lr": 7.470026417769282e-07}
{"train_loss": 0.03739611804485321, "global_step": 256523, "epoch": 2882, "lr": 7.469028076096729e-07}
{"train_loss": 0.04426395520567894, "global_step": 256524, "epoch": 2882, "lr": 7.468029800639043e-07}
{"train_loss": 0.0360732339322567, "global_step": 256525, "epoch": 2882, "lr": 7.467031591396223e-07}
{"train_loss": 0.0451173298060894, "global_step": 256526, "epoch": 2882, "lr": 7.466033448368492e-07}
{"train_loss": 0.03286676108837128, "global_step": 256527, "epoch": 2882, "lr": 7.465035371555906e-07}
{"train_loss": 0.044469233602285385, "global_step": 256528, "epoch": 2882, "lr": 7.464037360958686e-07}
{"train_loss": 0.04268629476428032, "global_step": 256529, "epoch": 2882, "lr": 7.46303941657689e-07}
{"train_loss": 0.0482296347618103, "global_step": 256530, "epoch": 2882, "lr": 7.46204153841068e-07}
{"train_loss": 0.057441774755716324, "global_step": 256531, "epoch": 2882, "lr": 7.461043726460226e-07}
{"train_loss": 0.061587732285261154, "global_step": 256532, "epoch": 2882, "lr": 7.460045980725638e-07}
{"train_loss": 0.029316913336515427, "global_step": 256533, "epoch": 2882, "lr": 7.459048301206972e-07}
{"train_loss": 0.021794624626636505, "global_step": 256534, "epoch": 2882, "lr": 7.458050687904505e-07}
{"train_loss": 0.031616050750017166, "global_step": 256535, "epoch": 2882, "lr": 7.457053140818293e-07}
{"train_loss": 0.04389354586601257, "global_step": 256536, "epoch": 2882, "lr": 7.456055659948446e-07}
{"train_loss": 0.05403423309326172, "global_step": 256537, "epoch": 2882, "lr": 7.455058245295188e-07}
{"train_loss": 0.05571405217051506, "global_step": 256538, "epoch": 2882, "lr": 7.454060896858516e-07}
{"train_loss": 0.03898952156305313, "global_step": 256539, "epoch": 2882, "lr": 7.453063614638711e-07}
{"train_loss": 0.04298041760921478, "global_step": 256540, "epoch": 2882, "lr": 7.452066398635771e-07}
{"train_loss": 0.06572894006967545, "global_step": 256541, "epoch": 2882, "lr": 7.451069248849974e-07}
{"train_loss": 0.05700330063700676, "global_step": 256542, "epoch": 2882, "lr": 7.450072165281319e-07}
{"train_loss": 0.029541587457060814, "global_step": 256543, "epoch": 2882, "lr": 7.449075147930029e-07}
{"train_loss": 0.06414614617824554, "global_step": 256544, "epoch": 2882, "lr": 7.44807819679616e-07}
{"train_loss": 0.05447635427117348, "global_step": 256545, "epoch": 2882, "lr": 7.447081311879934e-07}
{"train_loss": 0.020772531628608704, "global_step": 256546, "epoch": 2882, "lr": 7.446084493181405e-07}
{"train_loss": 0.01070485170930624, "global_step": 256547, "epoch": 2882, "lr": 7.445087740700851e-07}
{"train_loss": 0.03155019134283066, "global_step": 256548, "epoch": 2882, "lr": 7.444091054438163e-07}
{"train_loss": 0.02092117629945278, "global_step": 256549, "epoch": 2882, "lr": 7.443094434393727e-07}
{"train_loss": 0.018988756462931633, "global_step": 256550, "epoch": 2882, "lr": 7.442097880567489e-07}
{"train_loss": 0.055984657257795334, "global_step": 256551, "epoch": 2882, "lr": 7.441101392959726e-07}
{"train_loss": 0.06583412736654282, "global_step": 256552, "epoch": 2882, "lr": 7.440104971570494e-07}
{"train_loss": 0.045612212270498276, "global_step": 256553, "epoch": 2882, "lr": 7.439108616399848e-07}
{"train_loss": 0.05346490070223808, "global_step": 256554, "epoch": 2882, "lr": 7.438112327448121e-07}
{"train_loss": 0.04460867866873741, "global_step": 256555, "epoch": 2882, "lr": 7.437116104715258e-07}
{"train_loss": 0.03727097436785698, "global_step": 256556, "epoch": 2882, "lr": 7.436119948201481e-07}
{"train_loss": 0.04197490215301514, "global_step": 256557, "epoch": 2882, "lr": 7.435123857906956e-07}
{"train_loss": 0.031189939007163048, "global_step": 256558, "epoch": 2882, "lr": 7.434127833831739e-07}
{"train_loss": 0.06908759474754333, "global_step": 256559, "epoch": 2882, "lr": 7.433131875975996e-07}
{"train_loss": 0.04165174812078476, "global_step": 256560, "epoch": 2882, "lr": 7.432135984339894e-07}
{"train_loss": 0.018549881875514984, "global_step": 256561, "epoch": 2882, "lr": 7.431140158923488e-07}
{"train_loss": 0.038619935512542725, "global_step": 256562, "epoch": 2882, "lr": 7.430144399727057e-07}
{"train_loss": 0.06433773785829544, "global_step": 256563, "epoch": 2882, "lr": 7.429148706750544e-07}
{"train_loss": 0.02165621891617775, "global_step": 256564, "epoch": 2882, "lr": 7.428153079994227e-07}
{"train_loss": 0.03738338500261307, "global_step": 256565, "epoch": 2882, "lr": 7.427157519458161e-07}
{"train_loss": 0.0265516247600317, "global_step": 256566, "epoch": 2882, "lr": 7.426162025142569e-07}
{"train_loss": 0.03435777872800827, "global_step": 256567, "epoch": 2882, "lr": 7.42516659704745e-07}
{"train_loss": 0.04948021471500397, "global_step": 256568, "epoch": 2882, "lr": 7.424171235173083e-07}
{"train_loss": 0.030622242018580437, "global_step": 256569, "epoch": 2882, "lr": 7.423175939519467e-07}
{"train_loss": 0.0587732195854187, "global_step": 256570, "epoch": 2882, "lr": 7.422180710086824e-07}
{"train_loss": 0.018301013857126236, "global_step": 256571, "epoch": 2882, "lr": 7.421185546875264e-07}
{"train_loss": 0.03041318617761135, "global_step": 256572, "epoch": 2882, "lr": 7.420190449884901e-07}
{"train_loss": 0.04662492871284485, "global_step": 256573, "epoch": 2882, "lr": 7.419195419115954e-07}
{"train_loss": 0.03564468398690224, "global_step": 256574, "epoch": 2882, "lr": 7.418200454568425e-07}
{"train_loss": 0.04045321047306061, "global_step": 256575, "epoch": 2882, "lr": 7.41720555624259e-07}
{"train_loss": 0.022548800334334373, "global_step": 256576, "epoch": 2882, "lr": 7.416210724138451e-07}
{"train_loss": 0.02625482901930809, "global_step": 256577, "epoch": 2882, "lr": 7.415215958256227e-07}
{"train_loss": 0.0380864292383194, "global_step": 256578, "epoch": 2882, "lr": 7.414221258595977e-07}
{"train_loss": 0.020125582814216614, "global_step": 256579, "epoch": 2882, "lr": 7.413226625157921e-07}
{"train_loss": 0.06295520812273026, "global_step": 256580, "epoch": 2882, "lr": 7.41223205794217e-07}
{"train_loss": 0.043444693088531494, "global_step": 256581, "epoch": 2882, "lr": 7.411237556948836e-07}
{"train_loss": 0.05711551010608673, "global_step": 256582, "epoch": 2882, "lr": 7.410243122177973e-07}
{"train_loss": 0.04227576404809952, "global_step": 256583, "epoch": 2882, "lr": 7.409248753629916e-07}
{"train_loss": 0.029550934210419655, "global_step": 256584, "epoch": 2882, "lr": 7.408254451304608e-07}
{"train_loss": 0.05308767408132553, "global_step": 256585, "epoch": 2882, "lr": 7.407260215202272e-07}
{"train_loss": 0.03864006192705939, "global_step": 256586, "epoch": 2882, "lr": 7.406266045323019e-07, "val_loss": 9.711993217468262}
{"train_loss": 0.03375359997153282, "global_step": 256587, "epoch": 2883, "lr": 7.405271941667014e-07}
{"train_loss": 0.043333638459444046, "global_step": 256588, "epoch": 2883, "lr": 7.404277904234369e-07}
{"train_loss": 0.025653162971138954, "global_step": 256589, "epoch": 2883, "lr": 7.403283933025196e-07}
{"train_loss": 0.03865981847047806, "global_step": 256590, "epoch": 2883, "lr": 7.40229002803966e-07}
{"train_loss": 0.06195657700300217, "global_step": 256591, "epoch": 2883, "lr": 7.401296189277874e-07}
{"train_loss": 0.039197735488414764, "global_step": 256592, "epoch": 2883, "lr": 7.400302416740001e-07}
{"train_loss": 0.09251238405704498, "global_step": 256593, "epoch": 2883, "lr": 7.399308710426101e-07}
{"train_loss": 0.024374986067414284, "global_step": 256594, "epoch": 2883, "lr": 7.398315070336448e-07}
{"train_loss": 0.038306672126054764, "global_step": 256595, "epoch": 2883, "lr": 7.397321496470988e-07}
{"train_loss": 0.06101704761385918, "global_step": 256596, "epoch": 2883, "lr": 7.396327988829998e-07}
{"train_loss": 0.05703476816415787, "global_step": 256597, "epoch": 2883, "lr": 7.39533454741359e-07}
{"train_loss": 0.023208534345030785, "global_step": 256598, "epoch": 2883, "lr": 7.394341172221875e-07}
{"train_loss": 0.03970741108059883, "global_step": 256599, "epoch": 2883, "lr": 7.393347863254907e-07}
{"train_loss": 0.02072860486805439, "global_step": 256600, "epoch": 2883, "lr": 7.392354620512965e-07}
{"train_loss": 0.06409533321857452, "global_step": 256601, "epoch": 2883, "lr": 7.391361443996103e-07}
{"train_loss": 0.022856654599308968, "global_step": 256602, "epoch": 2883, "lr": 7.390368333704489e-07}
{"train_loss": 0.04780368506908417, "global_step": 256603, "epoch": 2883, "lr": 7.389375289638234e-07}
{"train_loss": 0.044443488121032715, "global_step": 256604, "epoch": 2883, "lr": 7.388382311797448e-07}
{"train_loss": 0.05022747814655304, "global_step": 256605, "epoch": 2883, "lr": 7.387389400182299e-07}
{"train_loss": 0.032200686633586884, "global_step": 256606, "epoch": 2883, "lr": 7.386396554792951e-07}
{"train_loss": 0.07998823374509811, "global_step": 256607, "epoch": 2883, "lr": 7.385403775629407e-07}
{"train_loss": 0.04096570238471031, "global_step": 256608, "epoch": 2883, "lr": 7.384411062691943e-07}
{"train_loss": 0.02375531941652298, "global_step": 256609, "epoch": 2883, "lr": 7.383418415980669e-07}
{"train_loss": 0.042528510093688965, "global_step": 256610, "epoch": 2883, "lr": 7.382425835495644e-07}
{"train_loss": 0.041331060230731964, "global_step": 256611, "epoch": 2883, "lr": 7.381433321237086e-07}
{"train_loss": 0.029971498996019363, "global_step": 256612, "epoch": 2883, "lr": 7.380440873205052e-07}
{"train_loss": 0.04164796322584152, "global_step": 256613, "epoch": 2883, "lr": 7.379448491399766e-07}
{"train_loss": 0.04357879236340523, "global_step": 256614, "epoch": 2883, "lr": 7.378456175821225e-07}
{"train_loss": 0.06890203803777695, "global_step": 256615, "epoch": 2883, "lr": 7.377463926469707e-07}
{"train_loss": 0.05554905906319618, "global_step": 256616, "epoch": 2883, "lr": 7.37647174334527e-07}
{"train_loss": 0.016354108229279518, "global_step": 256617, "epoch": 2883, "lr": 7.375479626448078e-07}
{"train_loss": 0.023235496133565903, "global_step": 256618, "epoch": 2883, "lr": 7.374487575778244e-07}
{"train_loss": 0.09251532703638077, "global_step": 256619, "epoch": 2883, "lr": 7.373495591335933e-07}
{"train_loss": 0.039095208048820496, "global_step": 256620, "epoch": 2883, "lr": 7.372503673121201e-07}
{"train_loss": 0.07225149869918823, "global_step": 256621, "epoch": 2883, "lr": 7.37151182113427e-07}
{"train_loss": 0.025445308536291122, "global_step": 256622, "epoch": 2883, "lr": 7.370520035375194e-07}
{"train_loss": 0.02492525614798069, "global_step": 256623, "epoch": 2883, "lr": 7.369528315844143e-07}
{"train_loss": 0.06361446529626846, "global_step": 256624, "epoch": 2883, "lr": 7.368536662541281e-07}
{"train_loss": 0.028352908790111542, "global_step": 256625, "epoch": 2883, "lr": 7.36754507546672e-07}
{"train_loss": 0.030247468501329422, "global_step": 256626, "epoch": 2883, "lr": 7.36655355462057e-07}
{"train_loss": 0.07494864612817764, "global_step": 256627, "epoch": 2883, "lr": 7.365562100002999e-07}
{"train_loss": 0.07139658182859421, "global_step": 256628, "epoch": 2883, "lr": 7.364570711614171e-07}
{"train_loss": 0.011038482189178467, "global_step": 256629, "epoch": 2883, "lr": 7.36357938945409e-07}
{"train_loss": 0.06851503998041153, "global_step": 256630, "epoch": 2883, "lr": 7.36258813352303e-07}
{"train_loss": 0.04302913323044777, "global_step": 256631, "epoch": 2883, "lr": 7.361596943820992e-07}
{"train_loss": 0.09072829782962799, "global_step": 256632, "epoch": 2883, "lr": 7.360605820348254e-07}
{"train_loss": 0.037059321999549866, "global_step": 256633, "epoch": 2883, "lr": 7.359614763104817e-07}
{"train_loss": 0.03105023503303528, "global_step": 256634, "epoch": 2883, "lr": 7.358623772090956e-07}
{"train_loss": 0.02697872556746006, "global_step": 256635, "epoch": 2883, "lr": 7.357632847306673e-07}
{"train_loss": 0.0649651437997818, "global_step": 256636, "epoch": 2883, "lr": 7.356641988752134e-07}
{"train_loss": 0.006687816698104143, "global_step": 256637, "epoch": 2883, "lr": 7.355651196427504e-07}
{"train_loss": 0.043114710599184036, "global_step": 256638, "epoch": 2883, "lr": 7.354660470332952e-07}
{"train_loss": 0.02071506157517433, "global_step": 256639, "epoch": 2883, "lr": 7.353669810468478e-07}
{"train_loss": 0.021061841398477554, "global_step": 256640, "epoch": 2883, "lr": 7.352679216834357e-07}
{"train_loss": 0.055448051542043686, "global_step": 256641, "epoch": 2883, "lr": 7.351688689430647e-07}
{"train_loss": 0.04742961376905441, "global_step": 256642, "epoch": 2883, "lr": 7.350698228257513e-07}
{"train_loss": 0.045265741646289825, "global_step": 256643, "epoch": 2883, "lr": 7.349707833315067e-07}
{"train_loss": 0.06920192390680313, "global_step": 256644, "epoch": 2883, "lr": 7.348717504603419e-07}
{"train_loss": 0.07459855824708939, "global_step": 256645, "epoch": 2883, "lr": 7.347727242122738e-07}
{"train_loss": 0.028105417266488075, "global_step": 256646, "epoch": 2883, "lr": 7.346737045873131e-07}
{"train_loss": 0.035317160189151764, "global_step": 256647, "epoch": 2883, "lr": 7.345746915854823e-07}
{"train_loss": 0.019266214221715927, "global_step": 256648, "epoch": 2883, "lr": 7.344756852067813e-07}
{"train_loss": 0.02835942804813385, "global_step": 256649, "epoch": 2883, "lr": 7.343766854512324e-07}
{"train_loss": 0.013174843974411488, "global_step": 256650, "epoch": 2883, "lr": 7.342776923188466e-07}
{"train_loss": 0.043149612843990326, "global_step": 256651, "epoch": 2883, "lr": 7.34178705809635e-07}
{"train_loss": 0.0484788678586483, "global_step": 256652, "epoch": 2883, "lr": 7.340797259236143e-07}
{"train_loss": 0.020076289772987366, "global_step": 256653, "epoch": 2883, "lr": 7.339807526607955e-07}
{"train_loss": 0.06139791011810303, "global_step": 256654, "epoch": 2883, "lr": 7.3388178602119e-07}
{"train_loss": 0.06988893449306488, "global_step": 256655, "epoch": 2883, "lr": 7.337828260048196e-07}
{"train_loss": 0.06579061597585678, "global_step": 256656, "epoch": 2883, "lr": 7.336838726116846e-07}
{"train_loss": 0.05010085180401802, "global_step": 256657, "epoch": 2883, "lr": 7.335849258418071e-07}
{"train_loss": 0.04418572783470154, "global_step": 256658, "epoch": 2883, "lr": 7.334859856951981e-07}
{"train_loss": 0.060405418276786804, "global_step": 256659, "epoch": 2883, "lr": 7.3338705217188e-07}
{"train_loss": 0.021030772477388382, "global_step": 256660, "epoch": 2883, "lr": 7.332881252718471e-07}
{"train_loss": 0.04752165079116821, "global_step": 256661, "epoch": 2883, "lr": 7.331892049951272e-07}
{"train_loss": 0.0696057453751564, "global_step": 256662, "epoch": 2883, "lr": 7.330902913417259e-07}
{"train_loss": 0.03626348078250885, "global_step": 256663, "epoch": 2883, "lr": 7.329913843116653e-07}
{"train_loss": 0.048606015741825104, "global_step": 256664, "epoch": 2883, "lr": 7.328924839049511e-07}
{"train_loss": 0.055248234421014786, "global_step": 256665, "epoch": 2883, "lr": 7.327935901216054e-07}
{"train_loss": 0.025709856301546097, "global_step": 256666, "epoch": 2883, "lr": 7.326947029616227e-07}
{"train_loss": 0.04777345806360245, "global_step": 256667, "epoch": 2883, "lr": 7.325958224250417e-07}
{"train_loss": 0.05175220966339111, "global_step": 256668, "epoch": 2883, "lr": 7.324969485118572e-07}
{"train_loss": 0.03222820162773132, "global_step": 256669, "epoch": 2883, "lr": 7.323980812220855e-07}
{"train_loss": 0.01878347247838974, "global_step": 256670, "epoch": 2883, "lr": 7.322992205557433e-07}
{"train_loss": 0.02938426472246647, "global_step": 256671, "epoch": 2883, "lr": 7.322003665128418e-07}
{"train_loss": 0.08994153141975403, "global_step": 256672, "epoch": 2883, "lr": 7.321015190934033e-07}
{"train_loss": 0.012743405997753143, "global_step": 256673, "epoch": 2883, "lr": 7.320026782974221e-07}
{"train_loss": 0.025436164811253548, "global_step": 256674, "epoch": 2883, "lr": 7.319038441249315e-07}
{"train_loss": 0.043902147849079934, "global_step": 256675, "epoch": 2883, "lr": 7.318050165759316e-07, "val_loss": 9.655081748962402}
{"train_loss": 0.06429115682840347, "global_step": 256676, "epoch": 2884, "lr": 7.31706195650439e-07}
{"train_loss": 0.064719557762146, "global_step": 256677, "epoch": 2884, "lr": 7.316073813484703e-07}
{"train_loss": 0.0341520719230175, "global_step": 256678, "epoch": 2884, "lr": 7.315085736700367e-07}
{"train_loss": 0.053555987775325775, "global_step": 256679, "epoch": 2884, "lr": 7.314097726151492e-07}
{"train_loss": 0.07798547297716141, "global_step": 256680, "epoch": 2884, "lr": 7.313109781838246e-07}
{"train_loss": 0.03237970173358917, "global_step": 256681, "epoch": 2884, "lr": 7.312121903760738e-07}
{"train_loss": 0.03639623522758484, "global_step": 256682, "epoch": 2884, "lr": 7.31113409191908e-07}
{"train_loss": 0.03728672116994858, "global_step": 256683, "epoch": 2884, "lr": 7.310146346313496e-07}
{"train_loss": 0.054667554795742035, "global_step": 256684, "epoch": 2884, "lr": 7.309158666944038e-07}
{"train_loss": 0.030270248651504517, "global_step": 256685, "epoch": 2884, "lr": 7.308171053810875e-07}
{"train_loss": 0.058225858956575394, "global_step": 256686, "epoch": 2884, "lr": 7.307183506914061e-07}
{"train_loss": 0.017976418137550354, "global_step": 256687, "epoch": 2884, "lr": 7.306196026253875e-07}
{"train_loss": 0.03500322252511978, "global_step": 256688, "epoch": 2884, "lr": 7.305208611830316e-07}
{"train_loss": 0.05040711909532547, "global_step": 256689, "epoch": 2884, "lr": 7.304221263643607e-07}
{"train_loss": 0.05995475873351097, "global_step": 256690, "epoch": 2884, "lr": 7.303233981693747e-07}
{"train_loss": 0.029285883530974388, "global_step": 256691, "epoch": 2884, "lr": 7.302246765981069e-07}
{"train_loss": 0.10724044591188431, "global_step": 256692, "epoch": 2884, "lr": 7.301259616505518e-07}
{"train_loss": 0.019631575793027878, "global_step": 256693, "epoch": 2884, "lr": 7.300272533267372e-07}
{"train_loss": 0.06665869802236557, "global_step": 256694, "epoch": 2884, "lr": 7.299285516266629e-07}
{"train_loss": 0.05043300241231918, "global_step": 256695, "epoch": 2884, "lr": 7.29829856550357e-07}
{"train_loss": 0.02277829684317112, "global_step": 256696, "epoch": 2884, "lr": 7.297311680978191e-07}
{"train_loss": 0.011804861016571522, "global_step": 256697, "epoch": 2884, "lr": 7.296324862690717e-07}
{"train_loss": 0.0761069655418396, "global_step": 256698, "epoch": 2884, "lr": 7.295338110641203e-07}
{"train_loss": 0.025245793163776398, "global_step": 256699, "epoch": 2884, "lr": 7.29435142482987e-07}
{"train_loss": 0.03687846660614014, "global_step": 256700, "epoch": 2884, "lr": 7.293364805256831e-07}
{"train_loss": 0.04710851609706879, "global_step": 256701, "epoch": 2884, "lr": 7.292378251922138e-07}
{"train_loss": 0.012728357687592506, "global_step": 256702, "epoch": 2884, "lr": 7.291391764826016e-07}
{"train_loss": 0.06266739964485168, "global_step": 256703, "epoch": 2884, "lr": 7.290405343968521e-07}
{"train_loss": 0.05633380636572838, "global_step": 256704, "epoch": 2884, "lr": 7.289418989349872e-07}
{"train_loss": 0.037184182554483414, "global_step": 256705, "epoch": 2884, "lr": 7.288432700970127e-07}
{"train_loss": 0.041918907314538956, "global_step": 256706, "epoch": 2884, "lr": 7.287446478829452e-07}
{"train_loss": 0.060375042259693146, "global_step": 256707, "epoch": 2884, "lr": 7.286460322927957e-07}
{"train_loss": 0.0377248153090477, "global_step": 256708, "epoch": 2884, "lr": 7.285474233265865e-07}
{"train_loss": 0.03863603249192238, "global_step": 256709, "epoch": 2884, "lr": 7.284488209843122e-07}
{"train_loss": 0.025318223983049393, "global_step": 256710, "epoch": 2884, "lr": 7.283502252660057e-07}
{"train_loss": 0.031301792711019516, "global_step": 256711, "epoch": 2884, "lr": 7.282516361716674e-07}
{"train_loss": 0.022815920412540436, "global_step": 256712, "epoch": 2884, "lr": 7.281530537013193e-07}
{"train_loss": 0.015802675858139992, "global_step": 256713, "epoch": 2884, "lr": 7.280544778549669e-07}
{"train_loss": 0.044831231236457825, "global_step": 256714, "epoch": 2884, "lr": 7.279559086326326e-07}
{"train_loss": 0.04404405504465103, "global_step": 256715, "epoch": 2884, "lr": 7.278573460343163e-07}
{"train_loss": 0.04417949169874191, "global_step": 256716, "epoch": 2884, "lr": 7.277587900600458e-07}
{"train_loss": 0.020945677533745766, "global_step": 256717, "epoch": 2884, "lr": 7.276602407098209e-07}
{"train_loss": 0.13890933990478516, "global_step": 256718, "epoch": 2884, "lr": 7.275616979836641e-07}
{"train_loss": 0.0391588918864727, "global_step": 256719, "epoch": 2884, "lr": 7.274631618815864e-07}
{"train_loss": 0.03372422233223915, "global_step": 256720, "epoch": 2884, "lr": 7.273646324035987e-07}
{"train_loss": 0.049991607666015625, "global_step": 256721, "epoch": 2884, "lr": 7.272661095497235e-07}
{"train_loss": 0.05031035095453262, "global_step": 256722, "epoch": 2884, "lr": 7.271675933199606e-07}
{"train_loss": 0.03564271703362465, "global_step": 256723, "epoch": 2884, "lr": 7.270690837143324e-07}
{"train_loss": 0.03718056529760361, "global_step": 256724, "epoch": 2884, "lr": 7.269705807328442e-07}
{"train_loss": 0.08385897427797318, "global_step": 256725, "epoch": 2884, "lr": 7.268720843755239e-07}
{"train_loss": 0.07218549400568008, "global_step": 256726, "epoch": 2884, "lr": 7.26773594642366e-07}
{"train_loss": 0.01187247596681118, "global_step": 256727, "epoch": 2884, "lr": 7.26675111533398e-07}
{"train_loss": 0.08023574948310852, "global_step": 256728, "epoch": 2884, "lr": 7.265766350486258e-07}
{"train_loss": 0.036821115761995316, "global_step": 256729, "epoch": 2884, "lr": 7.264781651880659e-07}
{"train_loss": 0.013247666880488396, "global_step": 256730, "epoch": 2884, "lr": 7.263797019517293e-07}
{"train_loss": 0.08006947487592697, "global_step": 256731, "epoch": 2884, "lr": 7.262812453396328e-07}
{"train_loss": 0.02316109836101532, "global_step": 256732, "epoch": 2884, "lr": 7.261827953517819e-07}
{"train_loss": 0.06452187150716782, "global_step": 256733, "epoch": 2884, "lr": 7.260843519881988e-07}
{"train_loss": 0.047921378165483475, "global_step": 256734, "epoch": 2884, "lr": 7.25985915248889e-07}
{"train_loss": 0.04199530556797981, "global_step": 256735, "epoch": 2884, "lr": 7.258874851338749e-07}
{"train_loss": 0.07127963751554489, "global_step": 256736, "epoch": 2884, "lr": 7.257890616431618e-07}
{"train_loss": 0.03831847384572029, "global_step": 256737, "epoch": 2884, "lr": 7.256906447767664e-07}
{"train_loss": 0.0805569514632225, "global_step": 256738, "epoch": 2884, "lr": 7.255922345347055e-07}
{"train_loss": 0.0369160994887352, "global_step": 256739, "epoch": 2884, "lr": 7.254938309169846e-07}
{"train_loss": 0.024718159809708595, "global_step": 256740, "epoch": 2884, "lr": 7.253954339236202e-07}
{"train_loss": 0.02075166068971157, "global_step": 256741, "epoch": 2884, "lr": 7.252970435546292e-07}
{"train_loss": 0.06374301016330719, "global_step": 256742, "epoch": 2884, "lr": 7.251986598100169e-07}
{"train_loss": 0.052626561373472214, "global_step": 256743, "epoch": 2884, "lr": 7.251002826898057e-07}
{"train_loss": 0.053300607949495316, "global_step": 256744, "epoch": 2884, "lr": 7.25001912194001e-07}
{"train_loss": 0.04375879839062691, "global_step": 256745, "epoch": 2884, "lr": 7.249035483226196e-07}
{"train_loss": 0.03093634732067585, "global_step": 256746, "epoch": 2884, "lr": 7.248051910756726e-07}
{"train_loss": 0.07141777873039246, "global_step": 256747, "epoch": 2884, "lr": 7.247068404531765e-07}
{"train_loss": 0.055818330496549606, "global_step": 256748, "epoch": 2884, "lr": 7.246084964551425e-07}
{"train_loss": 0.035205405205488205, "global_step": 256749, "epoch": 2884, "lr": 7.245101590815873e-07}
{"train_loss": 0.019305657595396042, "global_step": 256750, "epoch": 2884, "lr": 7.244118283325163e-07}
{"train_loss": 0.06130940839648247, "global_step": 256751, "epoch": 2884, "lr": 7.243135042079518e-07}
{"train_loss": 0.027277395129203796, "global_step": 256752, "epoch": 2884, "lr": 7.242151867078994e-07}
{"train_loss": 0.023145660758018494, "global_step": 256753, "epoch": 2884, "lr": 7.241168758323758e-07}
{"train_loss": 0.013098959811031818, "global_step": 256754, "epoch": 2884, "lr": 7.240185715813919e-07}
{"train_loss": 0.048133645206689835, "global_step": 256755, "epoch": 2884, "lr": 7.2392027395497e-07}
{"train_loss": 0.05082044005393982, "global_step": 256756, "epoch": 2884, "lr": 7.238219829531101e-07}
{"train_loss": 0.036255352199077606, "global_step": 256757, "epoch": 2884, "lr": 7.2372369857584e-07}
{"train_loss": 0.03475751727819443, "global_step": 256758, "epoch": 2884, "lr": 7.236254208231541e-07}
{"train_loss": 0.027500979602336884, "global_step": 256759, "epoch": 2884, "lr": 7.235271496950858e-07}
{"train_loss": 0.04796909540891647, "global_step": 256760, "epoch": 2884, "lr": 7.234288851916293e-07}
{"train_loss": 0.06659779697656631, "global_step": 256761, "epoch": 2884, "lr": 7.233306273128182e-07}
{"train_loss": 0.04898582026362419, "global_step": 256762, "epoch": 2884, "lr": 7.232323760586468e-07}
{"train_loss": 0.029379958286881447, "global_step": 256763, "epoch": 2884, "lr": 7.231341314291373e-07}
{"train_loss": 0.04470544927910473, "global_step": 256764, "epoch": 2884, "lr": 7.230358934243009e-07, "val_loss": 9.646458625793457}
{"train_loss": 0.05362989008426666, "global_step": 256765, "epoch": 2885, "lr": 7.229376620441542e-07}
{"train_loss": 0.08538366109132767, "global_step": 256766, "epoch": 2885, "lr": 7.228394372887082e-07}
{"train_loss": 0.03920058533549309, "global_step": 256767, "epoch": 2885, "lr": 7.227412191579741e-07}
{"train_loss": 0.05291304737329483, "global_step": 256768, "epoch": 2885, "lr": 7.226430076519686e-07}
{"train_loss": 0.042863886803388596, "global_step": 256769, "epoch": 2885, "lr": 7.225448027707026e-07}
{"train_loss": 0.07102679461240768, "global_step": 256770, "epoch": 2885, "lr": 7.224466045141876e-07}
{"train_loss": 0.07234937697649002, "global_step": 256771, "epoch": 2885, "lr": 7.223484128824454e-07}
{"train_loss": 0.023898620158433914, "global_step": 256772, "epoch": 2885, "lr": 7.222502278754761e-07}
{"train_loss": 0.04015128314495087, "global_step": 256773, "epoch": 2885, "lr": 7.221520494932965e-07}
{"train_loss": 0.03952214494347572, "global_step": 256774, "epoch": 2885, "lr": 7.220538777359287e-07}
{"train_loss": 0.07024610042572021, "global_step": 256775, "epoch": 2885, "lr": 7.219557126033782e-07}
{"train_loss": 0.07389536499977112, "global_step": 256776, "epoch": 2885, "lr": 7.218575540956674e-07}
{"train_loss": 0.11835730820894241, "global_step": 256777, "epoch": 2885, "lr": 7.217594022127905e-07}
{"train_loss": 0.015751538798213005, "global_step": 256778, "epoch": 2885, "lr": 7.216612569547809e-07}
{"train_loss": 0.11840280890464783, "global_step": 256779, "epoch": 2885, "lr": 7.215631183216443e-07}
{"train_loss": 0.026814395561814308, "global_step": 256780, "epoch": 2885, "lr": 7.21464986313386e-07}
{"train_loss": 0.021188678219914436, "global_step": 256781, "epoch": 2885, "lr": 7.21366860930034e-07}
{"train_loss": 0.03883041441440582, "global_step": 256782, "epoch": 2885, "lr": 7.21268742171588e-07}
{"train_loss": 0.07005686312913895, "global_step": 256783, "epoch": 2885, "lr": 7.211706300380705e-07}
{"train_loss": 0.074139803647995, "global_step": 256784, "epoch": 2885, "lr": 7.210725245294869e-07}
{"train_loss": 0.043155599385499954, "global_step": 256785, "epoch": 2885, "lr": 7.20974425645854e-07}
{"train_loss": 0.02840016596019268, "global_step": 256786, "epoch": 2885, "lr": 7.208763333871881e-07}
{"train_loss": 0.07045678049325943, "global_step": 256787, "epoch": 2885, "lr": 7.207782477535007e-07}
{"train_loss": 0.04465986415743828, "global_step": 256788, "epoch": 2885, "lr": 7.206801687448028e-07}
{"train_loss": 0.05975917726755142, "global_step": 256789, "epoch": 2885, "lr": 7.205820963611054e-07}
{"train_loss": 0.026822630316019058, "global_step": 256790, "epoch": 2885, "lr": 7.204840306024252e-07}
{"train_loss": 0.01843588426709175, "global_step": 256791, "epoch": 2885, "lr": 7.203859714687844e-07}
{"train_loss": 0.07229458540678024, "global_step": 256792, "epoch": 2885, "lr": 7.202879189601774e-07}
{"train_loss": 0.021646305918693542, "global_step": 256793, "epoch": 2885, "lr": 7.201898730766321e-07}
{"train_loss": 0.030645648017525673, "global_step": 256794, "epoch": 2885, "lr": 7.200918338181539e-07}
{"train_loss": 0.03636925667524338, "global_step": 256795, "epoch": 2885, "lr": 7.199938011847595e-07}
{"train_loss": 0.0723046138882637, "global_step": 256796, "epoch": 2885, "lr": 7.198957751764602e-07}
{"train_loss": 0.041204772889614105, "global_step": 256797, "epoch": 2885, "lr": 7.197977557932723e-07}
{"train_loss": 0.05708305165171623, "global_step": 256798, "epoch": 2885, "lr": 7.19699743035207e-07}
{"train_loss": 0.04978271201252937, "global_step": 256799, "epoch": 2885, "lr": 7.196017369022812e-07}
{"train_loss": 0.05272909626364708, "global_step": 256800, "epoch": 2885, "lr": 7.195037373944946e-07}
{"train_loss": 0.038344159722328186, "global_step": 256801, "epoch": 2885, "lr": 7.194057445118807e-07}
{"train_loss": 0.02046629786491394, "global_step": 256802, "epoch": 2885, "lr": 7.193077582544339e-07}
{"train_loss": 0.03735847771167755, "global_step": 256803, "epoch": 2885, "lr": 7.192097786221763e-07}
{"train_loss": 0.05584806203842163, "global_step": 256804, "epoch": 2885, "lr": 7.191118056151247e-07}
{"train_loss": 0.020890485495328903, "global_step": 256805, "epoch": 2885, "lr": 7.190138392332846e-07}
{"train_loss": 0.05117231234908104, "global_step": 256806, "epoch": 2885, "lr": 7.189158794766726e-07}
{"train_loss": 0.0391143299639225, "global_step": 256807, "epoch": 2885, "lr": 7.188179263453054e-07}
{"train_loss": 0.04017173871397972, "global_step": 256808, "epoch": 2885, "lr": 7.187199798391886e-07}
{"train_loss": 0.04870867729187012, "global_step": 256809, "epoch": 2885, "lr": 7.186220399583388e-07}
{"train_loss": 0.03878204897046089, "global_step": 256810, "epoch": 2885, "lr": 7.185241067027726e-07}
{"train_loss": 0.04002956300973892, "global_step": 256811, "epoch": 2885, "lr": 7.184261800724956e-07}
{"train_loss": 0.028644025325775146, "global_step": 256812, "epoch": 2885, "lr": 7.183282600675301e-07}
{"train_loss": 0.014734111726284027, "global_step": 256813, "epoch": 2885, "lr": 7.182303466878815e-07}
{"train_loss": 0.060325320810079575, "global_step": 256814, "epoch": 2885, "lr": 7.181324399335665e-07}
{"train_loss": 0.07663052529096603, "global_step": 256815, "epoch": 2885, "lr": 7.180345398046018e-07}
{"train_loss": 0.042151592671871185, "global_step": 256816, "epoch": 2885, "lr": 7.179366463009929e-07}
{"train_loss": 0.0671425387263298, "global_step": 256817, "epoch": 2885, "lr": 7.17838759422762e-07}
{"train_loss": 0.01898823492228985, "global_step": 256818, "epoch": 2885, "lr": 7.177408791699092e-07}
{"train_loss": 0.03867088630795479, "global_step": 256819, "epoch": 2885, "lr": 7.176430055424621e-07}
{"train_loss": 0.02288675121963024, "global_step": 256820, "epoch": 2885, "lr": 7.175451385404208e-07}
{"train_loss": 0.013364381156861782, "global_step": 256821, "epoch": 2885, "lr": 7.174472781638131e-07}
{"train_loss": 0.04649488255381584, "global_step": 256822, "epoch": 2885, "lr": 7.173494244126389e-07}
{"train_loss": 0.021322906017303467, "global_step": 256823, "epoch": 2885, "lr": 7.172515772869148e-07}
{"train_loss": 0.025847548618912697, "global_step": 256824, "epoch": 2885, "lr": 7.171537367866632e-07}
{"train_loss": 0.049683358520269394, "global_step": 256825, "epoch": 2885, "lr": 7.170559029118784e-07}
{"train_loss": 0.11006460338830948, "global_step": 256826, "epoch": 2885, "lr": 7.169580756625882e-07}
{"train_loss": 0.04660680890083313, "global_step": 256827, "epoch": 2885, "lr": 7.168602550388092e-07}
{"train_loss": 0.04589855298399925, "global_step": 256828, "epoch": 2885, "lr": 7.167624410405416e-07}
{"train_loss": 0.03042479045689106, "global_step": 256829, "epoch": 2885, "lr": 7.166646336678073e-07}
{"train_loss": 0.07827508449554443, "global_step": 256830, "epoch": 2885, "lr": 7.165668329206121e-07}
{"train_loss": 0.050824619829654694, "global_step": 256831, "epoch": 2885, "lr": 7.164690387989781e-07}
{"train_loss": 0.0364692248404026, "global_step": 256832, "epoch": 2885, "lr": 7.163712513029109e-07}
{"train_loss": 0.037916794419288635, "global_step": 256833, "epoch": 2885, "lr": 7.162734704324326e-07}
{"train_loss": 0.02538587711751461, "global_step": 256834, "epoch": 2885, "lr": 7.161756961875432e-07}
{"train_loss": 0.025572480633854866, "global_step": 256835, "epoch": 2885, "lr": 7.160779285682706e-07}
{"train_loss": 0.039420709013938904, "global_step": 256836, "epoch": 2885, "lr": 7.159801675746147e-07}
{"train_loss": 0.06936317682266235, "global_step": 256837, "epoch": 2885, "lr": 7.158824132066033e-07}
{"train_loss": 0.10001973807811737, "global_step": 256838, "epoch": 2885, "lr": 7.157846654642309e-07}
{"train_loss": 0.05783155933022499, "global_step": 256839, "epoch": 2885, "lr": 7.15686924347525e-07}
{"train_loss": 0.023618577048182487, "global_step": 256840, "epoch": 2885, "lr": 7.155891898564915e-07}
{"train_loss": 0.026104139164090157, "global_step": 256841, "epoch": 2885, "lr": 7.154914619911523e-07}
{"train_loss": 0.05056620016694069, "global_step": 256842, "epoch": 2885, "lr": 7.153937407515077e-07}
{"train_loss": 0.057963185012340546, "global_step": 256843, "epoch": 2885, "lr": 7.152960261375852e-07}
{"train_loss": 0.048302583396434784, "global_step": 256844, "epoch": 2885, "lr": 7.151983181493848e-07}
{"train_loss": 0.06934480369091034, "global_step": 256845, "epoch": 2885, "lr": 7.151006167869234e-07}
{"train_loss": 0.055338721722364426, "global_step": 256846, "epoch": 2885, "lr": 7.150029220502175e-07}
{"train_loss": 0.018679993227124214, "global_step": 256847, "epoch": 2885, "lr": 7.149052339392837e-07}
{"train_loss": 0.03352515771985054, "global_step": 256848, "epoch": 2885, "lr": 7.148075524541276e-07}
{"train_loss": 0.08405166864395142, "global_step": 256849, "epoch": 2885, "lr": 7.147098775947602e-07}
{"train_loss": 0.030843259766697884, "global_step": 256850, "epoch": 2885, "lr": 7.14612209361204e-07}
{"train_loss": 0.025899022817611694, "global_step": 256851, "epoch": 2885, "lr": 7.145145477534698e-07}
{"train_loss": 0.04223524406552315, "global_step": 256852, "epoch": 2885, "lr": 7.144168927715633e-07}
{"train_loss": 0.04706492979246914, "global_step": 256853, "epoch": 2885, "lr": 7.143192444155067e-07, "val_loss": 9.604865074157715, "train_action_mse_error": 7.997434616088867}
{"train_loss": 0.03616657108068466, "global_step": 256854, "epoch": 2886, "lr": 7.142216026853054e-07}
{"train_loss": 0.0321451835334301, "global_step": 256855, "epoch": 2886, "lr": 7.141239675809819e-07}
{"train_loss": 0.05581824854016304, "global_step": 256856, "epoch": 2886, "lr": 7.140263391025359e-07}
{"train_loss": 0.03788677603006363, "global_step": 256857, "epoch": 2886, "lr": 7.139287172499954e-07}
{"train_loss": 0.027943100780248642, "global_step": 256858, "epoch": 2886, "lr": 7.138311020233656e-07}
{"train_loss": 0.08381878584623337, "global_step": 256859, "epoch": 2886, "lr": 7.137334934226581e-07}
{"train_loss": 0.11459577828645706, "global_step": 256860, "epoch": 2886, "lr": 7.136358914478891e-07}
{"train_loss": 0.024286217987537384, "global_step": 256861, "epoch": 2886, "lr": 7.1353829609907e-07}
{"train_loss": 0.015344536863267422, "global_step": 256862, "epoch": 2886, "lr": 7.134407073762173e-07}
{"train_loss": 0.05723432078957558, "global_step": 256863, "epoch": 2886, "lr": 7.133431252793421e-07}
{"train_loss": 0.05728258937597275, "global_step": 256864, "epoch": 2886, "lr": 7.1324554980845e-07}
{"train_loss": 0.023463120684027672, "global_step": 256865, "epoch": 2886, "lr": 7.131479809635688e-07}
{"train_loss": 0.03063737414777279, "global_step": 256866, "epoch": 2886, "lr": 7.130504187446985e-07}
{"train_loss": 0.05126861855387688, "global_step": 256867, "epoch": 2886, "lr": 7.129528631518668e-07}
{"train_loss": 0.04450926184654236, "global_step": 256868, "epoch": 2886, "lr": 7.12855314185068e-07}
{"train_loss": 0.03117324784398079, "global_step": 256869, "epoch": 2886, "lr": 7.127577718443357e-07}
{"train_loss": 0.04756962135434151, "global_step": 256870, "epoch": 2886, "lr": 7.126602361296642e-07}
{"train_loss": 0.04260209575295448, "global_step": 256871, "epoch": 2886, "lr": 7.125627070410756e-07}
{"train_loss": 0.01053767092525959, "global_step": 256872, "epoch": 2886, "lr": 7.124651845785813e-07}
{"train_loss": 0.048744846135377884, "global_step": 256873, "epoch": 2886, "lr": 7.123676687421977e-07}
{"train_loss": 0.05378532037138939, "global_step": 256874, "epoch": 2886, "lr": 7.122701595319358e-07}
{"train_loss": 0.04149676859378815, "global_step": 256875, "epoch": 2886, "lr": 7.121726569478071e-07}
{"train_loss": 0.02508801594376564, "global_step": 256876, "epoch": 2886, "lr": 7.12075160989828e-07}
{"train_loss": 0.06273864954710007, "global_step": 256877, "epoch": 2886, "lr": 7.119776716580095e-07}
{"train_loss": 0.07571078091859818, "global_step": 256878, "epoch": 2886, "lr": 7.11880188952363e-07}
{"train_loss": 0.038362182676792145, "global_step": 256879, "epoch": 2886, "lr": 7.117827128729049e-07}
{"train_loss": 0.019561927765607834, "global_step": 256880, "epoch": 2886, "lr": 7.116852434196464e-07}
{"train_loss": 0.04589664563536644, "global_step": 256881, "epoch": 2886, "lr": 7.115877805925986e-07}
{"train_loss": 0.017172804102301598, "global_step": 256882, "epoch": 2886, "lr": 7.114903243917781e-07}
{"train_loss": 0.036941785365343094, "global_step": 256883, "epoch": 2886, "lr": 7.113928748172016e-07}
{"train_loss": 0.04130642116069794, "global_step": 256884, "epoch": 2886, "lr": 7.112954318688747e-07}
{"train_loss": 0.10185937583446503, "global_step": 256885, "epoch": 2886, "lr": 7.111979955468085e-07}
{"train_loss": 0.03542670980095863, "global_step": 256886, "epoch": 2886, "lr": 7.111005658510306e-07}
{"train_loss": 0.05883224681019783, "global_step": 256887, "epoch": 2886, "lr": 7.110031427815356e-07}
{"train_loss": 0.02390395849943161, "global_step": 256888, "epoch": 2886, "lr": 7.109057263383512e-07}
{"train_loss": 0.07001571357250214, "global_step": 256889, "epoch": 2886, "lr": 7.10808316521483e-07}
{"train_loss": 0.0582394115626812, "global_step": 256890, "epoch": 2886, "lr": 7.107109133309475e-07}
{"train_loss": 0.0655260756611824, "global_step": 256891, "epoch": 2886, "lr": 7.106135167667505e-07}
{"train_loss": 0.03441344574093819, "global_step": 256892, "epoch": 2886, "lr": 7.105161268289195e-07}
{"train_loss": 0.05198527127504349, "global_step": 256893, "epoch": 2886, "lr": 7.104187435174492e-07}
{"train_loss": 0.04893583804368973, "global_step": 256894, "epoch": 2886, "lr": 7.103213668323672e-07}
{"train_loss": 0.015509174205362797, "global_step": 256895, "epoch": 2886, "lr": 7.102239967736846e-07}
{"train_loss": 0.03153705969452858, "global_step": 256896, "epoch": 2886, "lr": 7.101266333414069e-07}
{"train_loss": 0.025376595556735992, "global_step": 256897, "epoch": 2886, "lr": 7.100292765355565e-07}
{"train_loss": 0.03973223268985748, "global_step": 256898, "epoch": 2886, "lr": 7.099319263561388e-07}
{"train_loss": 0.025044923648238182, "global_step": 256899, "epoch": 2886, "lr": 7.098345828031705e-07}
{"train_loss": 0.047089505940675735, "global_step": 256900, "epoch": 2886, "lr": 7.097372458766626e-07}
{"train_loss": 0.005838317796587944, "global_step": 256901, "epoch": 2886, "lr": 7.09639915576632e-07}
{"train_loss": 0.01743541844189167, "global_step": 256902, "epoch": 2886, "lr": 7.095425919030895e-07}
{"train_loss": 0.030144309625029564, "global_step": 256903, "epoch": 2886, "lr": 7.094452748560465e-07}
{"train_loss": 0.05239136517047882, "global_step": 256904, "epoch": 2886, "lr": 7.093479644355194e-07}
{"train_loss": 0.07916460931301117, "global_step": 256905, "epoch": 2886, "lr": 7.092506606415194e-07}
{"train_loss": 0.03450891003012657, "global_step": 256906, "epoch": 2886, "lr": 7.091533634740577e-07}
{"train_loss": 0.01725439541041851, "global_step": 256907, "epoch": 2886, "lr": 7.090560729331508e-07}
{"train_loss": 0.041100163012742996, "global_step": 256908, "epoch": 2886, "lr": 7.089587890188099e-07}
{"train_loss": 0.0713171437382698, "global_step": 256909, "epoch": 2886, "lr": 7.088615117310515e-07}
{"train_loss": 0.0694502741098404, "global_step": 256910, "epoch": 2886, "lr": 7.08764241069887e-07}
{"train_loss": 0.042856570333242416, "global_step": 256911, "epoch": 2886, "lr": 7.086669770353271e-07}
{"train_loss": 0.06169234216213226, "global_step": 256912, "epoch": 2886, "lr": 7.085697196273833e-07}
{"train_loss": 0.11867605894804001, "global_step": 256913, "epoch": 2886, "lr": 7.08472468846072e-07}
{"train_loss": 0.04446837306022644, "global_step": 256914, "epoch": 2886, "lr": 7.083752246914044e-07}
{"train_loss": 0.058894407004117966, "global_step": 256915, "epoch": 2886, "lr": 7.082779871634027e-07}
{"train_loss": 0.030509505420923233, "global_step": 256916, "epoch": 2886, "lr": 7.081807562620612e-07}
{"train_loss": 0.030897503718733788, "global_step": 256917, "epoch": 2886, "lr": 7.080835319874079e-07}
{"train_loss": 0.032246995717287064, "global_step": 256918, "epoch": 2886, "lr": 7.079863143394594e-07}
{"train_loss": 0.009325037710368633, "global_step": 256919, "epoch": 2886, "lr": 7.0788910331821e-07}
{"train_loss": 0.020953156054019928, "global_step": 256920, "epoch": 2886, "lr": 7.077918989236931e-07}
{"train_loss": 0.04438996687531471, "global_step": 256921, "epoch": 2886, "lr": 7.076947011559088e-07}
{"train_loss": 0.05075609311461449, "global_step": 256922, "epoch": 2886, "lr": 7.075975100148736e-07}
{"train_loss": 0.06573837995529175, "global_step": 256923, "epoch": 2886, "lr": 7.075003255006041e-07}
{"train_loss": 0.0471549853682518, "global_step": 256924, "epoch": 2886, "lr": 7.074031476131115e-07}
{"train_loss": 0.06063733249902725, "global_step": 256925, "epoch": 2886, "lr": 7.073059763524015e-07}
{"train_loss": 0.047341860830783844, "global_step": 256926, "epoch": 2886, "lr": 7.072088117184961e-07}
{"train_loss": 0.052002061158418655, "global_step": 256927, "epoch": 2886, "lr": 7.071116537114065e-07}
{"train_loss": 0.061091914772987366, "global_step": 256928, "epoch": 2886, "lr": 7.070145023311492e-07}
{"train_loss": 0.04572940990328789, "global_step": 256929, "epoch": 2886, "lr": 7.069173575777244e-07}
{"train_loss": 0.025196146219968796, "global_step": 256930, "epoch": 2886, "lr": 7.068202194511597e-07}
{"train_loss": 0.08321303129196167, "global_step": 256931, "epoch": 2886, "lr": 7.067230879514608e-07}
{"train_loss": 0.01936906762421131, "global_step": 256932, "epoch": 2886, "lr": 7.066259630786442e-07}
{"train_loss": 0.09086617827415466, "global_step": 256933, "epoch": 2886, "lr": 7.065288448327212e-07}
{"train_loss": 0.027168216183781624, "global_step": 256934, "epoch": 2886, "lr": 7.064317332136971e-07}
{"train_loss": 0.025113912299275398, "global_step": 256935, "epoch": 2886, "lr": 7.063346282215999e-07}
{"train_loss": 0.011670246720314026, "global_step": 256936, "epoch": 2886, "lr": 7.062375298564294e-07}
{"train_loss": 0.039593011140823364, "global_step": 256937, "epoch": 2886, "lr": 7.061404381182136e-07}
{"train_loss": 0.0284804105758667, "global_step": 256938, "epoch": 2886, "lr": 7.060433530069465e-07}
{"train_loss": 0.03165426850318909, "global_step": 256939, "epoch": 2886, "lr": 7.059462745226564e-07}
{"train_loss": 0.04710870981216431, "global_step": 256940, "epoch": 2886, "lr": 7.058492026653485e-07}
{"train_loss": 0.02717491239309311, "global_step": 256941, "epoch": 2886, "lr": 7.057521374350395e-07}
{"train_loss": 0.04411806627647596, "global_step": 256942, "epoch": 2886, "lr": 7.056550788317407e-07, "val_loss": 9.598294258117676}
{"train_loss": 0.034538011997938156, "global_step": 256943, "epoch": 2887, "lr": 7.055580268554685e-07}
{"train_loss": 0.05198584124445915, "global_step": 256944, "epoch": 2887, "lr": 7.054609815062285e-07}
{"train_loss": 0.056884150952100754, "global_step": 256945, "epoch": 2887, "lr": 7.05363942784043e-07}
{"train_loss": 0.02893155999481678, "global_step": 256946, "epoch": 2887, "lr": 7.052669106889176e-07}
{"train_loss": 0.10524538904428482, "global_step": 256947, "epoch": 2887, "lr": 7.051698852208744e-07}
{"train_loss": 0.04940888658165932, "global_step": 256948, "epoch": 2887, "lr": 7.050728663799133e-07}
{"train_loss": 0.05319653078913689, "global_step": 256949, "epoch": 2887, "lr": 7.049758541660567e-07}
{"train_loss": 0.07712428271770477, "global_step": 256950, "epoch": 2887, "lr": 7.048788485793156e-07}
{"train_loss": 0.024448703974485397, "global_step": 256951, "epoch": 2887, "lr": 7.047818496197012e-07}
{"train_loss": 0.025404544547200203, "global_step": 256952, "epoch": 2887, "lr": 7.046848572872244e-07}
{"train_loss": 0.05484471842646599, "global_step": 256953, "epoch": 2887, "lr": 7.045878715819077e-07}
{"train_loss": 0.07826081663370132, "global_step": 256954, "epoch": 2887, "lr": 7.044908925037563e-07}
{"train_loss": 0.018027545884251595, "global_step": 256955, "epoch": 2887, "lr": 7.043939200527871e-07}
{"train_loss": 0.033398717641830444, "global_step": 256956, "epoch": 2887, "lr": 7.042969542290112e-07}
{"train_loss": 0.05190347880125046, "global_step": 256957, "epoch": 2887, "lr": 7.041999950324396e-07}
{"train_loss": 0.014683215878903866, "global_step": 256958, "epoch": 2887, "lr": 7.04103042463089e-07}
{"train_loss": 0.0662134438753128, "global_step": 256959, "epoch": 2887, "lr": 7.040060965209705e-07}
{"train_loss": 0.02938033454120159, "global_step": 256960, "epoch": 2887, "lr": 7.039091572060952e-07}
{"train_loss": 0.021894201636314392, "global_step": 256961, "epoch": 2887, "lr": 7.038122245184797e-07}
{"train_loss": 0.05426881089806557, "global_step": 256962, "epoch": 2887, "lr": 7.037152984581408e-07}
{"train_loss": 0.013696973212063313, "global_step": 256963, "epoch": 2887, "lr": 7.036183790250783e-07}
{"train_loss": 0.05919511243700981, "global_step": 256964, "epoch": 2887, "lr": 7.035214662193201e-07}
{"train_loss": 0.056642964482307434, "global_step": 256965, "epoch": 2887, "lr": 7.034245600408717e-07}
{"train_loss": 0.045252300798892975, "global_step": 256966, "epoch": 2887, "lr": 7.033276604897443e-07}
{"train_loss": 0.02775285765528679, "global_step": 256967, "epoch": 2887, "lr": 7.032307675659544e-07}
{"train_loss": 0.01585197262465954, "global_step": 256968, "epoch": 2887, "lr": 7.031338812695188e-07}
{"train_loss": 0.04664405435323715, "global_step": 256969, "epoch": 2887, "lr": 7.030370016004373e-07}
{"train_loss": 0.04068422317504883, "global_step": 256970, "epoch": 2887, "lr": 7.029401285587378e-07}
{"train_loss": 0.0855548307299614, "global_step": 256971, "epoch": 2887, "lr": 7.028432621444314e-07}
{"train_loss": 0.042071565985679626, "global_step": 256972, "epoch": 2887, "lr": 7.027464023575181e-07}
{"train_loss": 0.034399840980768204, "global_step": 256973, "epoch": 2887, "lr": 7.026495491980256e-07}
{"train_loss": 0.05292639508843422, "global_step": 256974, "epoch": 2887, "lr": 7.025527026659539e-07}
{"train_loss": 0.017242643982172012, "global_step": 256975, "epoch": 2887, "lr": 7.024558627613309e-07}
{"train_loss": 0.043198101222515106, "global_step": 256976, "epoch": 2887, "lr": 7.023590294841565e-07}
{"train_loss": 0.04387667030096054, "global_step": 256977, "epoch": 2887, "lr": 7.022622028344527e-07}
{"train_loss": 0.031038809567689896, "global_step": 256978, "epoch": 2887, "lr": 7.021653828122254e-07}
{"train_loss": 0.03185300901532173, "global_step": 256979, "epoch": 2887, "lr": 7.020685694174966e-07}
{"train_loss": 0.044684503227472305, "global_step": 256980, "epoch": 2887, "lr": 7.019717626502664e-07}
{"train_loss": 0.029008636251091957, "global_step": 256981, "epoch": 2887, "lr": 7.018749625105625e-07}
{"train_loss": 0.02084214985370636, "global_step": 256982, "epoch": 2887, "lr": 7.017781689983849e-07}
{"train_loss": 0.09103511273860931, "global_step": 256983, "epoch": 2887, "lr": 7.016813821137558e-07}
{"train_loss": 0.062185633927583694, "global_step": 256984, "epoch": 2887, "lr": 7.015846018566808e-07}
{"train_loss": 0.10007794946432114, "global_step": 256985, "epoch": 2887, "lr": 7.014878282271819e-07}
{"train_loss": 0.019489280879497528, "global_step": 256986, "epoch": 2887, "lr": 7.01391061225265e-07}
{"train_loss": 0.03204191476106644, "global_step": 256987, "epoch": 2887, "lr": 7.012943008509465e-07}
{"train_loss": 0.03707084804773331, "global_step": 256988, "epoch": 2887, "lr": 7.01197547104232e-07}
{"train_loss": 0.04047746956348419, "global_step": 256989, "epoch": 2887, "lr": 7.011007999851438e-07}
{"train_loss": 0.09667536616325378, "global_step": 256990, "epoch": 2887, "lr": 7.010040594936984e-07}
{"train_loss": 0.04815935343503952, "global_step": 256991, "epoch": 2887, "lr": 7.009073256298904e-07}
{"train_loss": 0.03588559851050377, "global_step": 256992, "epoch": 2887, "lr": 7.008105983937529e-07}
{"train_loss": 0.03017338365316391, "global_step": 256993, "epoch": 2887, "lr": 7.007138777852917e-07}
{"train_loss": 0.006653874646872282, "global_step": 256994, "epoch": 2887, "lr": 7.006171638045122e-07}
{"train_loss": 0.03724999353289604, "global_step": 256995, "epoch": 2887, "lr": 7.005204564514367e-07}
{"train_loss": 0.06254711002111435, "global_step": 256996, "epoch": 2887, "lr": 7.004237557260818e-07}
{"train_loss": 0.043601952493190765, "global_step": 256997, "epoch": 2887, "lr": 7.003270616284419e-07}
{"train_loss": 0.02202575094997883, "global_step": 256998, "epoch": 2887, "lr": 7.002303741585503e-07}
{"train_loss": 0.029027380049228668, "global_step": 256999, "epoch": 2887, "lr": 7.001336933164127e-07}
{"train_loss": 0.026275213807821274, "global_step": 257000, "epoch": 2887, "lr": 7.0003701910204e-07}
{"train_loss": 0.02679044008255005, "global_step": 257001, "epoch": 2887, "lr": 6.999403515154435e-07}
{"train_loss": 0.030104853212833405, "global_step": 257002, "epoch": 2887, "lr": 6.998436905566397e-07}
{"train_loss": 0.09719064086675644, "global_step": 257003, "epoch": 2887, "lr": 6.997470362256397e-07}
{"train_loss": 0.049717411398887634, "global_step": 257004, "epoch": 2887, "lr": 6.996503885224604e-07}
{"train_loss": 0.05375458300113678, "global_step": 257005, "epoch": 2887, "lr": 6.995537474471126e-07}
{"train_loss": 0.049641214311122894, "global_step": 257006, "epoch": 2887, "lr": 6.994571129996075e-07}
{"train_loss": 0.038512784987688065, "global_step": 257007, "epoch": 2887, "lr": 6.993604851799617e-07}
{"train_loss": 0.07414856553077698, "global_step": 257008, "epoch": 2887, "lr": 6.992638639881866e-07}
{"train_loss": 0.07644852250814438, "global_step": 257009, "epoch": 2887, "lr": 6.991672494242874e-07}
{"train_loss": 0.058330751955509186, "global_step": 257010, "epoch": 2887, "lr": 6.99070641488292e-07}
{"train_loss": 0.05157708749175072, "global_step": 257011, "epoch": 2887, "lr": 6.989740401802059e-07}
{"train_loss": 0.09709630906581879, "global_step": 257012, "epoch": 2887, "lr": 6.988774455000402e-07}
{"train_loss": 0.028253283351659775, "global_step": 257013, "epoch": 2887, "lr": 6.987808574478061e-07}
{"train_loss": 0.025045571848750114, "global_step": 257014, "epoch": 2887, "lr": 6.986842760235202e-07}
{"train_loss": 0.10058335959911346, "global_step": 257015, "epoch": 2887, "lr": 6.98587701227199e-07}
{"train_loss": 0.014370814897119999, "global_step": 257016, "epoch": 2887, "lr": 6.984911330588483e-07}
{"train_loss": 0.05569199472665787, "global_step": 257017, "epoch": 2887, "lr": 6.983945715184903e-07}
{"train_loss": 0.041638243943452835, "global_step": 257018, "epoch": 2887, "lr": 6.982980166061249e-07}
{"train_loss": 0.03456519916653633, "global_step": 257019, "epoch": 2887, "lr": 6.982014683217796e-07}
{"train_loss": 0.03797123208642006, "global_step": 257020, "epoch": 2887, "lr": 6.981049266654549e-07}
{"train_loss": 0.04518010839819908, "global_step": 257021, "epoch": 2887, "lr": 6.980083916371727e-07}
{"train_loss": 0.034202735871076584, "global_step": 257022, "epoch": 2887, "lr": 6.979118632369386e-07}
{"train_loss": 0.07246420532464981, "global_step": 257023, "epoch": 2887, "lr": 6.978153414647693e-07}
{"train_loss": 0.010442418977618217, "global_step": 257024, "epoch": 2887, "lr": 6.977188263206813e-07}
{"train_loss": 0.02537418343126774, "global_step": 257025, "epoch": 2887, "lr": 6.976223178046859e-07}
{"train_loss": 0.03312209993600845, "global_step": 257026, "epoch": 2887, "lr": 6.975258159167886e-07}
{"train_loss": 0.05197567865252495, "global_step": 257027, "epoch": 2887, "lr": 6.974293206570059e-07}
{"train_loss": 0.037169888615608215, "global_step": 257028, "epoch": 2887, "lr": 6.973328320253602e-07}
{"train_loss": 0.05802997201681137, "global_step": 257029, "epoch": 2887, "lr": 6.972363500218516e-07}
{"train_loss": 0.06042257323861122, "global_step": 257030, "epoch": 2887, "lr": 6.971398746465019e-07}
{"train_loss": 0.04555390323669221, "global_step": 257031, "epoch": 2887, "lr": 6.97043405899317e-07, "val_loss": 9.596062660217285}
{"train_loss": 0.05332997068762779, "global_step": 257032, "epoch": 2888, "lr": 6.969469437803189e-07}
{"train_loss": 0.08172062039375305, "global_step": 257033, "epoch": 2888, "lr": 6.968504882895132e-07}
{"train_loss": 0.051264576613903046, "global_step": 257034, "epoch": 2888, "lr": 6.967540394269168e-07}
{"train_loss": 0.04584040492773056, "global_step": 257035, "epoch": 2888, "lr": 6.966575971925404e-07}
{"train_loss": 0.044236332178115845, "global_step": 257036, "epoch": 2888, "lr": 6.965611615863954e-07}
{"train_loss": 0.05758757144212723, "global_step": 257037, "epoch": 2888, "lr": 6.964647326084983e-07}
{"train_loss": 0.05126768723130226, "global_step": 257038, "epoch": 2888, "lr": 6.963683102588603e-07}
{"train_loss": 0.04780864343047142, "global_step": 257039, "epoch": 2888, "lr": 6.962718945374924e-07}
{"train_loss": 0.07539580017328262, "global_step": 257040, "epoch": 2888, "lr": 6.961754854444113e-07}
{"train_loss": 0.09092685580253601, "global_step": 257041, "epoch": 2888, "lr": 6.960790829796283e-07}
{"train_loss": 0.020812660455703735, "global_step": 257042, "epoch": 2888, "lr": 6.959826871431597e-07}
{"train_loss": 0.03337373211979866, "global_step": 257043, "epoch": 2888, "lr": 6.958862979350111e-07}
{"train_loss": 0.036780692636966705, "global_step": 257044, "epoch": 2888, "lr": 6.95789915355205e-07}
{"train_loss": 0.03446017578244209, "global_step": 257045, "epoch": 2888, "lr": 6.956935394037411e-07}
{"train_loss": 0.015435040928423405, "global_step": 257046, "epoch": 2888, "lr": 6.955971700806474e-07}
{"train_loss": 0.03334908187389374, "global_step": 257047, "epoch": 2888, "lr": 6.955008073859237e-07}
{"train_loss": 0.05924994498491287, "global_step": 257048, "epoch": 2888, "lr": 6.954044513195923e-07}
{"train_loss": 0.06488007307052612, "global_step": 257049, "epoch": 2888, "lr": 6.953081018816643e-07}
{"train_loss": 0.02913571335375309, "global_step": 257050, "epoch": 2888, "lr": 6.952117590721507e-07}
{"train_loss": 0.009810700081288815, "global_step": 257051, "epoch": 2888, "lr": 6.951154228910628e-07}
{"train_loss": 0.04790278151631355, "global_step": 257052, "epoch": 2888, "lr": 6.950190933384171e-07}
{"train_loss": 0.04329593852162361, "global_step": 257053, "epoch": 2888, "lr": 6.949227704142248e-07}
{"train_loss": 0.06595032662153244, "global_step": 257054, "epoch": 2888, "lr": 6.948264541184968e-07}
{"train_loss": 0.051029812544584274, "global_step": 257055, "epoch": 2888, "lr": 6.947301444512555e-07}
{"train_loss": 0.06113890931010246, "global_step": 257056, "epoch": 2888, "lr": 6.946338414125008e-07}
{"train_loss": 0.050112009048461914, "global_step": 257057, "epoch": 2888, "lr": 6.945375450022551e-07}
{"train_loss": 0.0562082938849926, "global_step": 257058, "epoch": 2888, "lr": 6.944412552205238e-07}
{"train_loss": 0.06279624253511429, "global_step": 257059, "epoch": 2888, "lr": 6.943449720673234e-07}
{"train_loss": 0.045591793954372406, "global_step": 257060, "epoch": 2888, "lr": 6.942486955426708e-07}
{"train_loss": 0.058497000485658646, "global_step": 257061, "epoch": 2888, "lr": 6.941524256465771e-07}
{"train_loss": 0.047584958374500275, "global_step": 257062, "epoch": 2888, "lr": 6.940561623790476e-07}
{"train_loss": 0.05529404059052467, "global_step": 257063, "epoch": 2888, "lr": 6.939599057401048e-07}
{"train_loss": 0.0343669131398201, "global_step": 257064, "epoch": 2888, "lr": 6.938636557297596e-07}
{"train_loss": 0.0661504790186882, "global_step": 257065, "epoch": 2888, "lr": 6.937674123480175e-07}
{"train_loss": 0.032222457230091095, "global_step": 257066, "epoch": 2888, "lr": 6.936711755949066e-07}
{"train_loss": 0.048029232770204544, "global_step": 257067, "epoch": 2888, "lr": 6.93574945470421e-07}
{"train_loss": 0.06941763311624527, "global_step": 257068, "epoch": 2888, "lr": 6.934787219745942e-07}
{"train_loss": 0.04530962556600571, "global_step": 257069, "epoch": 2888, "lr": 6.93382505107415e-07}
{"train_loss": 0.07803547382354736, "global_step": 257070, "epoch": 2888, "lr": 6.932862948689222e-07}
{"train_loss": 0.030610395595431328, "global_step": 257071, "epoch": 2888, "lr": 6.931900912591105e-07}
{"train_loss": 0.019791144877672195, "global_step": 257072, "epoch": 2888, "lr": 6.930938942779963e-07}
{"train_loss": 0.05437551438808441, "global_step": 257073, "epoch": 2888, "lr": 6.929977039255964e-07}
{"train_loss": 0.08242180943489075, "global_step": 257074, "epoch": 2888, "lr": 6.929015202019218e-07}
{"train_loss": 0.041713543236255646, "global_step": 257075, "epoch": 2888, "lr": 6.928053431069892e-07}
{"train_loss": 0.03856641426682472, "global_step": 257076, "epoch": 2888, "lr": 6.927091726408041e-07}
{"train_loss": 0.046759311109781265, "global_step": 257077, "epoch": 2888, "lr": 6.926130088033833e-07}
{"train_loss": 0.0347352996468544, "global_step": 257078, "epoch": 2888, "lr": 6.925168515947433e-07}
{"train_loss": 0.015117930248379707, "global_step": 257079, "epoch": 2888, "lr": 6.924207010148897e-07}
{"train_loss": 0.08119924366474152, "global_step": 257080, "epoch": 2888, "lr": 6.923245570638392e-07}
{"train_loss": 0.06539943069219589, "global_step": 257081, "epoch": 2888, "lr": 6.92228419741603e-07}
{"train_loss": 0.05381236970424652, "global_step": 257082, "epoch": 2888, "lr": 6.921322890482029e-07}
{"train_loss": 0.021128160879015923, "global_step": 257083, "epoch": 2888, "lr": 6.920361649836394e-07}
{"train_loss": 0.014939749613404274, "global_step": 257084, "epoch": 2888, "lr": 6.919400475479343e-07}
{"train_loss": 0.06879453361034393, "global_step": 257085, "epoch": 2888, "lr": 6.918439367410878e-07}
{"train_loss": 0.037754643708467484, "global_step": 257086, "epoch": 2888, "lr": 6.917478325631333e-07}
{"train_loss": 0.043066948652267456, "global_step": 257087, "epoch": 2888, "lr": 6.91651735014065e-07}
{"train_loss": 0.03595469146966934, "global_step": 257088, "epoch": 2888, "lr": 6.915556440939053e-07}
{"train_loss": 0.03803187608718872, "global_step": 257089, "epoch": 2888, "lr": 6.914595598026652e-07}
{"train_loss": 0.008985795080661774, "global_step": 257090, "epoch": 2888, "lr": 6.913634821403558e-07}
{"train_loss": 0.03634529933333397, "global_step": 257091, "epoch": 2888, "lr": 6.912674111069939e-07}
{"train_loss": 0.05990038067102432, "global_step": 257092, "epoch": 2888, "lr": 6.911713467025904e-07}
{"train_loss": 0.04093198850750923, "global_step": 257093, "epoch": 2888, "lr": 6.910752889271565e-07}
{"train_loss": 0.03710724785923958, "global_step": 257094, "epoch": 2888, "lr": 6.90979237780709e-07}
{"train_loss": 0.07723314315080643, "global_step": 257095, "epoch": 2888, "lr": 6.908831932632531e-07}
{"train_loss": 0.03522998467087746, "global_step": 257096, "epoch": 2888, "lr": 6.907871553748113e-07}
{"train_loss": 0.03490732982754707, "global_step": 257097, "epoch": 2888, "lr": 6.906911241153891e-07}
{"train_loss": 0.01756259985268116, "global_step": 257098, "epoch": 2888, "lr": 6.90595099485003e-07}
{"train_loss": 0.02252863720059395, "global_step": 257099, "epoch": 2888, "lr": 6.904990814836642e-07}
{"train_loss": 0.03765633702278137, "global_step": 257100, "epoch": 2888, "lr": 6.90403070111395e-07}
{"train_loss": 0.04333627596497536, "global_step": 257101, "epoch": 2888, "lr": 6.903070653681898e-07}
{"train_loss": 0.02093786746263504, "global_step": 257102, "epoch": 2888, "lr": 6.902110672540762e-07}
{"train_loss": 0.09183496236801147, "global_step": 257103, "epoch": 2888, "lr": 6.901150757690655e-07}
{"train_loss": 0.035268090665340424, "global_step": 257104, "epoch": 2888, "lr": 6.900190909131632e-07}
{"train_loss": 0.05211291462182999, "global_step": 257105, "epoch": 2888, "lr": 6.899231126863859e-07}
{"train_loss": 0.032045695930719376, "global_step": 257106, "epoch": 2888, "lr": 6.898271410887557e-07}
{"train_loss": 0.0624585896730423, "global_step": 257107, "epoch": 2888, "lr": 6.897311761202674e-07}
{"train_loss": 0.07469955086708069, "global_step": 257108, "epoch": 2888, "lr": 6.89635217780954e-07}
{"train_loss": 0.07499917596578598, "global_step": 257109, "epoch": 2888, "lr": 6.8953926607081e-07}
{"train_loss": 0.05634012073278427, "global_step": 257110, "epoch": 2888, "lr": 6.894433209898577e-07}
{"train_loss": 0.03474961221218109, "global_step": 257111, "epoch": 2888, "lr": 6.89347382538108e-07}
{"train_loss": 0.0318339504301548, "global_step": 257112, "epoch": 2888, "lr": 6.892514507155779e-07}
{"train_loss": 0.017051929607987404, "global_step": 257113, "epoch": 2888, "lr": 6.891555255222726e-07}
{"train_loss": 0.05492036044597626, "global_step": 257114, "epoch": 2888, "lr": 6.890596069582145e-07}
{"train_loss": 0.04377034306526184, "global_step": 257115, "epoch": 2888, "lr": 6.88963695023409e-07}
{"train_loss": 0.05029047280550003, "global_step": 257116, "epoch": 2888, "lr": 6.888677897178674e-07}
{"train_loss": 0.04204642027616501, "global_step": 257117, "epoch": 2888, "lr": 6.887718910416118e-07}
{"train_loss": 0.034932445734739304, "global_step": 257118, "epoch": 2888, "lr": 6.886759989946479e-07}
{"train_loss": 0.03377852588891983, "global_step": 257119, "epoch": 2888, "lr": 6.88580113576992e-07}
{"train_loss": 0.04674382744294204, "global_step": 257120, "epoch": 2888, "lr": 6.884842347886556e-07, "val_loss": 9.65373420715332}
{"train_loss": 0.017422311007976532, "global_step": 257121, "epoch": 2889, "lr": 6.883883626296494e-07}
{"train_loss": 0.04345694184303284, "global_step": 257122, "epoch": 2889, "lr": 6.882924970999905e-07}
{"train_loss": 0.07892505824565887, "global_step": 257123, "epoch": 2889, "lr": 6.881966381996896e-07}
{"train_loss": 0.009156061336398125, "global_step": 257124, "epoch": 2889, "lr": 6.881007859287581e-07}
{"train_loss": 0.05612112209200859, "global_step": 257125, "epoch": 2889, "lr": 6.880049402872069e-07}
{"train_loss": 0.03980256989598274, "global_step": 257126, "epoch": 2889, "lr": 6.879091012750583e-07}
{"train_loss": 0.014738153666257858, "global_step": 257127, "epoch": 2889, "lr": 6.878132688923178e-07}
{"train_loss": 0.041239768266677856, "global_step": 257128, "epoch": 2889, "lr": 6.877174431390021e-07}
{"train_loss": 0.0285075344145298, "global_step": 257129, "epoch": 2889, "lr": 6.876216240151168e-07}
{"train_loss": 0.04929358512163162, "global_step": 257130, "epoch": 2889, "lr": 6.87525811520684e-07}
{"train_loss": 0.024082958698272705, "global_step": 257131, "epoch": 2889, "lr": 6.874300056557092e-07}
{"train_loss": 0.019570181146264076, "global_step": 257132, "epoch": 2889, "lr": 6.873342064202093e-07}
{"train_loss": 0.039493437856435776, "global_step": 257133, "epoch": 2889, "lr": 6.872384138141951e-07}
{"train_loss": 0.036041807383298874, "global_step": 257134, "epoch": 2889, "lr": 6.871426278376836e-07}
{"train_loss": 0.07267636805772781, "global_step": 257135, "epoch": 2889, "lr": 6.8704684849068e-07}
{"train_loss": 0.019885744899511337, "global_step": 257136, "epoch": 2889, "lr": 6.869510757732067e-07}
{"train_loss": 0.04925275593996048, "global_step": 257137, "epoch": 2889, "lr": 6.868553096852693e-07}
{"train_loss": 0.04194588214159012, "global_step": 257138, "epoch": 2889, "lr": 6.867595502268842e-07}
{"train_loss": 0.049429237842559814, "global_step": 257139, "epoch": 2889, "lr": 6.866637973980572e-07}
{"train_loss": 0.04732247442007065, "global_step": 257140, "epoch": 2889, "lr": 6.86568051198816e-07}
{"train_loss": 0.02901676669716835, "global_step": 257141, "epoch": 2889, "lr": 6.864723116291605e-07}
{"train_loss": 0.04033145308494568, "global_step": 257142, "epoch": 2889, "lr": 6.863765786891129e-07}
{"train_loss": 0.0508798286318779, "global_step": 257143, "epoch": 2889, "lr": 6.862808523786735e-07}
{"train_loss": 0.04443146660923958, "global_step": 257144, "epoch": 2889, "lr": 6.861851326978697e-07}
{"train_loss": 0.028194952756166458, "global_step": 257145, "epoch": 2889, "lr": 6.860894196467016e-07}
{"train_loss": 0.05038854479789734, "global_step": 257146, "epoch": 2889, "lr": 6.859937132251915e-07}
{"train_loss": 0.03465079888701439, "global_step": 257147, "epoch": 2889, "lr": 6.858980134333448e-07}
{"train_loss": 0.03742730990052223, "global_step": 257148, "epoch": 2889, "lr": 6.858023202711784e-07}
{"train_loss": 0.055094219744205475, "global_step": 257149, "epoch": 2889, "lr": 6.85706633738703e-07}
{"train_loss": 0.05385186895728111, "global_step": 257150, "epoch": 2889, "lr": 6.856109538359412e-07}
{"train_loss": 0.04773740470409393, "global_step": 257151, "epoch": 2889, "lr": 6.855152805628929e-07}
{"train_loss": 0.022372927516698837, "global_step": 257152, "epoch": 2889, "lr": 6.854196139195746e-07}
{"train_loss": 0.06699805706739426, "global_step": 257153, "epoch": 2889, "lr": 6.853239539060031e-07}
{"train_loss": 0.045522481203079224, "global_step": 257154, "epoch": 2889, "lr": 6.852283005221838e-07}
{"train_loss": 0.06944095343351364, "global_step": 257155, "epoch": 2889, "lr": 6.851326537681391e-07}
{"train_loss": 0.027684278786182404, "global_step": 257156, "epoch": 2889, "lr": 6.850370136438799e-07}
{"train_loss": 0.016659744083881378, "global_step": 257157, "epoch": 2889, "lr": 6.84941380149412e-07}
{"train_loss": 0.037237267941236496, "global_step": 257158, "epoch": 2889, "lr": 6.848457532847519e-07}
{"train_loss": 0.078521229326725, "global_step": 257159, "epoch": 2889, "lr": 6.847501330499162e-07}
{"train_loss": 0.056920554488897324, "global_step": 257160, "epoch": 2889, "lr": 6.846545194449161e-07}
{"train_loss": 0.047176554799079895, "global_step": 257161, "epoch": 2889, "lr": 6.845589124697627e-07}
{"train_loss": 0.05600498616695404, "global_step": 257162, "epoch": 2889, "lr": 6.84463312124467e-07}
{"train_loss": 0.053201381117105484, "global_step": 257163, "epoch": 2889, "lr": 6.843677184090402e-07}
{"train_loss": 0.02062823995947838, "global_step": 257164, "epoch": 2889, "lr": 6.842721313235101e-07}
{"train_loss": 0.10828791558742523, "global_step": 257165, "epoch": 2889, "lr": 6.84176550867871e-07}
{"train_loss": 0.03188565373420715, "global_step": 257166, "epoch": 2889, "lr": 6.840809770421453e-07}
{"train_loss": 0.076326385140419, "global_step": 257167, "epoch": 2889, "lr": 6.839854098463438e-07}
{"train_loss": 0.02045762725174427, "global_step": 257168, "epoch": 2889, "lr": 6.83889849280478e-07}
{"train_loss": 0.07270265370607376, "global_step": 257169, "epoch": 2889, "lr": 6.837942953445642e-07}
{"train_loss": 0.0902077853679657, "global_step": 257170, "epoch": 2889, "lr": 6.836987480386081e-07}
{"train_loss": 0.02184055745601654, "global_step": 257171, "epoch": 2889, "lr": 6.836032073626264e-07}
{"train_loss": 0.056263312697410583, "global_step": 257172, "epoch": 2889, "lr": 6.835076733166412e-07}
{"train_loss": 0.03075038269162178, "global_step": 257173, "epoch": 2889, "lr": 6.834121459006526e-07}
{"train_loss": 0.046868469566106796, "global_step": 257174, "epoch": 2889, "lr": 6.833166251146828e-07}
{"train_loss": 0.06933824717998505, "global_step": 257175, "epoch": 2889, "lr": 6.832211109587316e-07}
{"train_loss": 0.044344231486320496, "global_step": 257176, "epoch": 2889, "lr": 6.83125603432827e-07}
{"train_loss": 0.05839278921484947, "global_step": 257177, "epoch": 2889, "lr": 6.830301025369745e-07}
{"train_loss": 0.015513163059949875, "global_step": 257178, "epoch": 2889, "lr": 6.829346082711851e-07}
{"train_loss": 0.041372328996658325, "global_step": 257179, "epoch": 2889, "lr": 6.828391206354756e-07}
{"train_loss": 0.05319105088710785, "global_step": 257180, "epoch": 2889, "lr": 6.82743639629857e-07}
{"train_loss": 0.0626465454697609, "global_step": 257181, "epoch": 2889, "lr": 6.826481652543404e-07}
{"train_loss": 0.03899630531668663, "global_step": 257182, "epoch": 2889, "lr": 6.82552697508948e-07}
{"train_loss": 0.026276912540197372, "global_step": 257183, "epoch": 2889, "lr": 6.824572363936744e-07}
{"train_loss": 0.017314516007900238, "global_step": 257184, "epoch": 2889, "lr": 6.823617819085526e-07}
{"train_loss": 0.040487729012966156, "global_step": 257185, "epoch": 2889, "lr": 6.822663340535773e-07}
{"train_loss": 0.05097290500998497, "global_step": 257186, "epoch": 2889, "lr": 6.821708928287818e-07}
{"train_loss": 0.04174241051077843, "global_step": 257187, "epoch": 2889, "lr": 6.820754582341549e-07}
{"train_loss": 0.06819634139537811, "global_step": 257188, "epoch": 2889, "lr": 6.819800302697299e-07}
{"train_loss": 0.03762652352452278, "global_step": 257189, "epoch": 2889, "lr": 6.818846089355069e-07}
{"train_loss": 0.05171841382980347, "global_step": 257190, "epoch": 2889, "lr": 6.81789194231508e-07}
{"train_loss": 0.023452209308743477, "global_step": 257191, "epoch": 2889, "lr": 6.816937861577388e-07}
{"train_loss": 0.03990239277482033, "global_step": 257192, "epoch": 2889, "lr": 6.81598384714216e-07}
{"train_loss": 0.02809738554060459, "global_step": 257193, "epoch": 2889, "lr": 6.815029899009506e-07}
{"train_loss": 0.039187002927064896, "global_step": 257194, "epoch": 2889, "lr": 6.814076017179538e-07}
{"train_loss": 0.013679091818630695, "global_step": 257195, "epoch": 2889, "lr": 6.813122201652478e-07}
{"train_loss": 0.053918201476335526, "global_step": 257196, "epoch": 2889, "lr": 6.812168452428269e-07}
{"train_loss": 0.07806837558746338, "global_step": 257197, "epoch": 2889, "lr": 6.811214769507246e-07}
{"train_loss": 0.03976839780807495, "global_step": 257198, "epoch": 2889, "lr": 6.810261152889463e-07}
{"train_loss": 0.06726587563753128, "global_step": 257199, "epoch": 2889, "lr": 6.809307602574921e-07}
{"train_loss": 0.019430480897426605, "global_step": 257200, "epoch": 2889, "lr": 6.808354118563953e-07}
{"train_loss": 0.047526612877845764, "global_step": 257201, "epoch": 2889, "lr": 6.807400700856559e-07}
{"train_loss": 0.034279465675354004, "global_step": 257202, "epoch": 2889, "lr": 6.806447349452904e-07}
{"train_loss": 0.022730305790901184, "global_step": 257203, "epoch": 2889, "lr": 6.805494064353102e-07}
{"train_loss": 0.08242509514093399, "global_step": 257204, "epoch": 2889, "lr": 6.804540845557261e-07}
{"train_loss": 0.0190273467451334, "global_step": 257205, "epoch": 2889, "lr": 6.803587693065605e-07}
{"train_loss": 0.0800846740603447, "global_step": 257206, "epoch": 2889, "lr": 6.802634606878133e-07}
{"train_loss": 0.026060543954372406, "global_step": 257207, "epoch": 2889, "lr": 6.801681586995068e-07}
{"train_loss": 0.027340836822986603, "global_step": 257208, "epoch": 2889, "lr": 6.80072863341652e-07}
{"train_loss": 0.04384937376997779, "global_step": 257209, "epoch": 2889, "lr": 6.799775746142544e-07, "val_loss": 9.608304977416992}
{"train_loss": 0.04330521821975708, "global_step": 257210, "epoch": 2890, "lr": 6.798822925173365e-07}
{"train_loss": 0.02877490222454071, "global_step": 257211, "epoch": 2890, "lr": 6.797870170509036e-07}
{"train_loss": 0.03258286416530609, "global_step": 257212, "epoch": 2890, "lr": 6.796917482149779e-07}
{"train_loss": 0.010943908244371414, "global_step": 257213, "epoch": 2890, "lr": 6.795964860095649e-07}
{"train_loss": 0.056973837316036224, "global_step": 257214, "epoch": 2890, "lr": 6.795012304346815e-07}
{"train_loss": 0.047393590211868286, "global_step": 257215, "epoch": 2890, "lr": 6.794059814903331e-07}
{"train_loss": 0.04777415096759796, "global_step": 257216, "epoch": 2890, "lr": 6.793107391765419e-07}
{"train_loss": 0.04134542867541313, "global_step": 257217, "epoch": 2890, "lr": 6.792155034933079e-07}
{"train_loss": 0.06753033399581909, "global_step": 257218, "epoch": 2890, "lr": 6.791202744406589e-07}
{"train_loss": 0.06847476214170456, "global_step": 257219, "epoch": 2890, "lr": 6.790250520186003e-07}
{"train_loss": 0.07588325440883636, "global_step": 257220, "epoch": 2890, "lr": 6.78929836227149e-07}
{"train_loss": 0.045975662767887115, "global_step": 257221, "epoch": 2890, "lr": 6.788346270663049e-07}
{"train_loss": 0.045675016939640045, "global_step": 257222, "epoch": 2890, "lr": 6.787394245361012e-07}
{"train_loss": 0.052185747772455215, "global_step": 257223, "epoch": 2890, "lr": 6.786442286365324e-07}
{"train_loss": 0.04407479614019394, "global_step": 257224, "epoch": 2890, "lr": 6.785490393676208e-07}
{"train_loss": 0.039158716797828674, "global_step": 257225, "epoch": 2890, "lr": 6.784538567293775e-07}
{"train_loss": 0.02046092040836811, "global_step": 257226, "epoch": 2890, "lr": 6.783586807218134e-07}
{"train_loss": 0.059667155146598816, "global_step": 257227, "epoch": 2890, "lr": 6.782635113449454e-07}
{"train_loss": 0.009957998059689999, "global_step": 257228, "epoch": 2890, "lr": 6.781683485987844e-07}
{"train_loss": 0.05469868704676628, "global_step": 257229, "epoch": 2890, "lr": 6.780731924833416e-07}
{"train_loss": 0.03730660304427147, "global_step": 257230, "epoch": 2890, "lr": 6.779780429986282e-07}
{"train_loss": 0.05034684017300606, "global_step": 257231, "epoch": 2890, "lr": 6.778829001446608e-07}
{"train_loss": 0.0467650480568409, "global_step": 257232, "epoch": 2890, "lr": 6.777877639214503e-07}
{"train_loss": 0.05560210719704628, "global_step": 257233, "epoch": 2890, "lr": 6.776926343290135e-07}
{"train_loss": 0.027709297835826874, "global_step": 257234, "epoch": 2890, "lr": 6.77597511367356e-07}
{"train_loss": 0.05773996561765671, "global_step": 257235, "epoch": 2890, "lr": 6.775023950365e-07}
{"train_loss": 0.03508993238210678, "global_step": 257236, "epoch": 2890, "lr": 6.774072853364455e-07}
{"train_loss": 0.04307396337389946, "global_step": 257237, "epoch": 2890, "lr": 6.773121822672146e-07}
{"train_loss": 0.04593029245734215, "global_step": 257238, "epoch": 2890, "lr": 6.772170858288241e-07}
{"train_loss": 0.03827448561787605, "global_step": 257239, "epoch": 2890, "lr": 6.771219960212738e-07}
{"train_loss": 0.07478895038366318, "global_step": 257240, "epoch": 2890, "lr": 6.770269128445861e-07}
{"train_loss": 0.035179413855075836, "global_step": 257241, "epoch": 2890, "lr": 6.769318362987664e-07}
{"train_loss": 0.06731344759464264, "global_step": 257242, "epoch": 2890, "lr": 6.768367663838371e-07}
{"train_loss": 0.04533417895436287, "global_step": 257243, "epoch": 2890, "lr": 6.767417030998036e-07}
{"train_loss": 0.04073672369122505, "global_step": 257244, "epoch": 2890, "lr": 6.766466464466825e-07}
{"train_loss": 0.03399647772312164, "global_step": 257245, "epoch": 2890, "lr": 6.765515964244795e-07}
{"train_loss": 0.02085856720805168, "global_step": 257246, "epoch": 2890, "lr": 6.764565530332223e-07}
{"train_loss": 0.07225368916988373, "global_step": 257247, "epoch": 2890, "lr": 6.763615162729054e-07}
{"train_loss": 0.11919384449720383, "global_step": 257248, "epoch": 2890, "lr": 6.762664861435564e-07}
{"train_loss": 0.0440131276845932, "global_step": 257249, "epoch": 2890, "lr": 6.76171462645181e-07}
{"train_loss": 0.04230962321162224, "global_step": 257250, "epoch": 2890, "lr": 6.760764457777957e-07}
{"train_loss": 0.04121699556708336, "global_step": 257251, "epoch": 2890, "lr": 6.759814355414063e-07}
{"train_loss": 0.05982181429862976, "global_step": 257252, "epoch": 2890, "lr": 6.758864319360292e-07}
{"train_loss": 0.04465920105576515, "global_step": 257253, "epoch": 2890, "lr": 6.757914349616812e-07}
{"train_loss": 0.06959008425474167, "global_step": 257254, "epoch": 2890, "lr": 6.756964446183733e-07}
{"train_loss": 0.050684429705142975, "global_step": 257255, "epoch": 2890, "lr": 6.756014609061112e-07}
{"train_loss": 0.0572480633854866, "global_step": 257256, "epoch": 2890, "lr": 6.75506483824917e-07}
{"train_loss": 0.051683202385902405, "global_step": 257257, "epoch": 2890, "lr": 6.754115133748018e-07}
{"train_loss": 0.10607567429542542, "global_step": 257258, "epoch": 2890, "lr": 6.753165495557712e-07}
{"train_loss": 0.021263698115944862, "global_step": 257259, "epoch": 2890, "lr": 6.752215923678473e-07}
{"train_loss": 0.06894154101610184, "global_step": 257260, "epoch": 2890, "lr": 6.751266418110358e-07}
{"train_loss": 0.04366615414619446, "global_step": 257261, "epoch": 2890, "lr": 6.750316978853532e-07}
{"train_loss": 0.036180708557367325, "global_step": 257262, "epoch": 2890, "lr": 6.749367605908108e-07}
{"train_loss": 0.019361604005098343, "global_step": 257263, "epoch": 2890, "lr": 6.74841829927425e-07}
{"train_loss": 0.07026276737451553, "global_step": 257264, "epoch": 2890, "lr": 6.747469058952017e-07}
{"train_loss": 0.029994528740644455, "global_step": 257265, "epoch": 2890, "lr": 6.746519884941627e-07}
{"train_loss": 0.05222153291106224, "global_step": 257266, "epoch": 2890, "lr": 6.745570777243082e-07}
{"train_loss": 0.060505662113428116, "global_step": 257267, "epoch": 2890, "lr": 6.74462173585666e-07}
{"train_loss": 0.023966152220964432, "global_step": 257268, "epoch": 2890, "lr": 6.74367276078236e-07}
{"train_loss": 0.07825736701488495, "global_step": 257269, "epoch": 2890, "lr": 6.74272385202035e-07}
{"train_loss": 0.04376041516661644, "global_step": 257270, "epoch": 2890, "lr": 6.741775009570794e-07}
{"train_loss": 0.02878115326166153, "global_step": 257271, "epoch": 2890, "lr": 6.740826233433806e-07}
{"train_loss": 0.03838689997792244, "global_step": 257272, "epoch": 2890, "lr": 6.739877523609495e-07}
{"train_loss": 0.014535218477249146, "global_step": 257273, "epoch": 2890, "lr": 6.738928880098028e-07}
{"train_loss": 0.02227677032351494, "global_step": 257274, "epoch": 2890, "lr": 6.737980302899405e-07}
{"train_loss": 0.03482784330844879, "global_step": 257275, "epoch": 2890, "lr": 6.73703179201396e-07}
{"train_loss": 0.02172955498099327, "global_step": 257276, "epoch": 2890, "lr": 6.736083347441635e-07}
{"train_loss": 0.018221398815512657, "global_step": 257277, "epoch": 2890, "lr": 6.735134969182655e-07}
{"train_loss": 0.031820230185985565, "global_step": 257278, "epoch": 2890, "lr": 6.734186657237129e-07}
{"train_loss": 0.04968629032373428, "global_step": 257279, "epoch": 2890, "lr": 6.733238411605169e-07}
{"train_loss": 0.0477604940533638, "global_step": 257280, "epoch": 2890, "lr": 6.732290232286942e-07}
{"train_loss": 0.042079176753759384, "global_step": 257281, "epoch": 2890, "lr": 6.731342119282502e-07}
{"train_loss": 0.041582126170396805, "global_step": 257282, "epoch": 2890, "lr": 6.730394072592072e-07}
{"train_loss": 0.06007407605648041, "global_step": 257283, "epoch": 2890, "lr": 6.729446092215652e-07}
{"train_loss": 0.03680974245071411, "global_step": 257284, "epoch": 2890, "lr": 6.728498178153519e-07}
{"train_loss": 0.02709629014134407, "global_step": 257285, "epoch": 2890, "lr": 6.727550330405675e-07}
{"train_loss": 0.027542278170585632, "global_step": 257286, "epoch": 2890, "lr": 6.726602548972338e-07}
{"train_loss": 0.014292114414274693, "global_step": 257287, "epoch": 2890, "lr": 6.725654833853568e-07}
{"train_loss": 0.05333837866783142, "global_step": 257288, "epoch": 2890, "lr": 6.724707185049528e-07}
{"train_loss": 0.05017601326107979, "global_step": 257289, "epoch": 2890, "lr": 6.723759602560331e-07}
{"train_loss": 0.048556193709373474, "global_step": 257290, "epoch": 2890, "lr": 6.722812086386144e-07}
{"train_loss": 0.048038091510534286, "global_step": 257291, "epoch": 2890, "lr": 6.721864636527075e-07}
{"train_loss": 0.06328821927309036, "global_step": 257292, "epoch": 2890, "lr": 6.720917252983183e-07}
{"train_loss": 0.025049777701497078, "global_step": 257293, "epoch": 2890, "lr": 6.719969935754689e-07}
{"train_loss": 0.03154560923576355, "global_step": 257294, "epoch": 2890, "lr": 6.719022684841702e-07}
{"train_loss": 0.021087223663926125, "global_step": 257295, "epoch": 2890, "lr": 6.71807550024428e-07}
{"train_loss": 0.035944487899541855, "global_step": 257296, "epoch": 2890, "lr": 6.717128381962645e-07}
{"train_loss": 0.05194951966404915, "global_step": 257297, "epoch": 2890, "lr": 6.716181329996796e-07}
{"train_loss": 0.044870641647597374, "global_step": 257298, "epoch": 2890, "lr": 6.71523434434701e-07, "val_loss": 9.512868881225586, "train_action_mse_error": 19.046951293945312}
{"train_loss": 0.03817025572061539, "global_step": 257299, "epoch": 2891, "lr": 6.714287425013344e-07}
{"train_loss": 0.019491925835609436, "global_step": 257300, "epoch": 2891, "lr": 6.713340571995907e-07}
{"train_loss": 0.06807460635900497, "global_step": 257301, "epoch": 2891, "lr": 6.712393785294924e-07}
{"train_loss": 0.048051245510578156, "global_step": 257302, "epoch": 2891, "lr": 6.711447064910337e-07}
{"train_loss": 0.02961093746125698, "global_step": 257303, "epoch": 2891, "lr": 6.71050041084248e-07}
{"train_loss": 0.03822093829512596, "global_step": 257304, "epoch": 2891, "lr": 6.709553823091352e-07}
{"train_loss": 0.04901733249425888, "global_step": 257305, "epoch": 2891, "lr": 6.708607301657121e-07}
{"train_loss": 0.07788032293319702, "global_step": 257306, "epoch": 2891, "lr": 6.707660846539843e-07}
{"train_loss": 0.05925925448536873, "global_step": 257307, "epoch": 2891, "lr": 6.706714457739794e-07}
{"train_loss": 0.02578592672944069, "global_step": 257308, "epoch": 2891, "lr": 6.705768135256918e-07}
{"train_loss": 0.043253250420093536, "global_step": 257309, "epoch": 2891, "lr": 6.704821879091549e-07}
{"train_loss": 0.0455782487988472, "global_step": 257310, "epoch": 2891, "lr": 6.703875689243633e-07}
{"train_loss": 0.029344523325562477, "global_step": 257311, "epoch": 2891, "lr": 6.70292956571339e-07}
{"train_loss": 0.07374190539121628, "global_step": 257312, "epoch": 2891, "lr": 6.701983508500931e-07}
{"train_loss": 0.0471862368285656, "global_step": 257313, "epoch": 2891, "lr": 6.701037517606368e-07}
{"train_loss": 0.018694940954446793, "global_step": 257314, "epoch": 2891, "lr": 6.700091593029867e-07}
{"train_loss": 0.1050189733505249, "global_step": 257315, "epoch": 2891, "lr": 6.699145734771483e-07}
{"train_loss": 0.010551313869655132, "global_step": 257316, "epoch": 2891, "lr": 6.698199942831385e-07}
{"train_loss": 0.025471048429608345, "global_step": 257317, "epoch": 2891, "lr": 6.697254217209736e-07}
{"train_loss": 0.022425806149840355, "global_step": 257318, "epoch": 2891, "lr": 6.696308557906594e-07}
{"train_loss": 0.03487319499254227, "global_step": 257319, "epoch": 2891, "lr": 6.695362964922125e-07}
{"train_loss": 0.06761187314987183, "global_step": 257320, "epoch": 2891, "lr": 6.694417438256495e-07}
{"train_loss": 0.00945200677961111, "global_step": 257321, "epoch": 2891, "lr": 6.69347197790976e-07}
{"train_loss": 0.04991639405488968, "global_step": 257322, "epoch": 2891, "lr": 6.692526583882086e-07}
{"train_loss": 0.06793145090341568, "global_step": 257323, "epoch": 2891, "lr": 6.691581256173529e-07}
{"train_loss": 0.06982407718896866, "global_step": 257324, "epoch": 2891, "lr": 6.690635994784367e-07}
{"train_loss": 0.05555669218301773, "global_step": 257325, "epoch": 2891, "lr": 6.689690799714543e-07}
{"train_loss": 0.08138199895620346, "global_step": 257326, "epoch": 2891, "lr": 6.688745670964336e-07}
{"train_loss": 0.05075165629386902, "global_step": 257327, "epoch": 2891, "lr": 6.687800608533801e-07}
{"train_loss": 0.034278661012649536, "global_step": 257328, "epoch": 2891, "lr": 6.686855612423104e-07}
{"train_loss": 0.036863118410110474, "global_step": 257329, "epoch": 2891, "lr": 6.685910682632301e-07}
{"train_loss": 0.0121831726282835, "global_step": 257330, "epoch": 2891, "lr": 6.684965819161614e-07}
{"train_loss": 0.04061145707964897, "global_step": 257331, "epoch": 2891, "lr": 6.6840210220111e-07}
{"train_loss": 0.03200917690992355, "global_step": 257332, "epoch": 2891, "lr": 6.683076291180923e-07}
{"train_loss": 0.05669200047850609, "global_step": 257333, "epoch": 2891, "lr": 6.682131626671139e-07}
{"train_loss": 0.02867969125509262, "global_step": 257334, "epoch": 2891, "lr": 6.681187028481972e-07}
{"train_loss": 0.06413827836513519, "global_step": 257335, "epoch": 2891, "lr": 6.680242496613475e-07}
{"train_loss": 0.0422159880399704, "global_step": 257336, "epoch": 2891, "lr": 6.679298031065817e-07}
{"train_loss": 0.061694271862506866, "global_step": 257337, "epoch": 2891, "lr": 6.678353631839163e-07}
{"train_loss": 0.03265543654561043, "global_step": 257338, "epoch": 2891, "lr": 6.677409298933513e-07}
{"train_loss": 0.0298494603484869, "global_step": 257339, "epoch": 2891, "lr": 6.676465032349144e-07}
{"train_loss": 0.05293755233287811, "global_step": 257340, "epoch": 2891, "lr": 6.675520832086057e-07}
{"train_loss": 0.021416235715150833, "global_step": 257341, "epoch": 2891, "lr": 6.674576698144474e-07}
{"train_loss": 0.053290244191884995, "global_step": 257342, "epoch": 2891, "lr": 6.673632630524451e-07}
{"train_loss": 0.017012013122439384, "global_step": 257343, "epoch": 2891, "lr": 6.67268862922621e-07}
{"train_loss": 0.051389627158641815, "global_step": 257344, "epoch": 2891, "lr": 6.671744694249748e-07}
{"train_loss": 0.03174266591668129, "global_step": 257345, "epoch": 2891, "lr": 6.670800825595291e-07}
{"train_loss": 0.03632774576544762, "global_step": 257346, "epoch": 2891, "lr": 6.669857023262893e-07}
{"train_loss": 0.07852333039045334, "global_step": 257347, "epoch": 2891, "lr": 6.668913287252776e-07}
{"train_loss": 0.039127741008996964, "global_step": 257348, "epoch": 2891, "lr": 6.66796961756494e-07}
{"train_loss": 0.05917913094162941, "global_step": 257349, "epoch": 2891, "lr": 6.667026014199663e-07}
{"train_loss": 0.0693393275141716, "global_step": 257350, "epoch": 2891, "lr": 6.666082477156888e-07}
{"train_loss": 0.05893935635685921, "global_step": 257351, "epoch": 2891, "lr": 6.665139006436949e-07}
{"train_loss": 0.04470844566822052, "global_step": 257352, "epoch": 2891, "lr": 6.664195602039791e-07}
{"train_loss": 0.05310443416237831, "global_step": 257353, "epoch": 2891, "lr": 6.663252263965692e-07}
{"train_loss": 0.031384266912937164, "global_step": 257354, "epoch": 2891, "lr": 6.66230899221465e-07}
{"train_loss": 0.06957277655601501, "global_step": 257355, "epoch": 2891, "lr": 6.661365786786888e-07}
{"train_loss": 0.05949900671839714, "global_step": 257356, "epoch": 2891, "lr": 6.660422647682462e-07}
{"train_loss": 0.050229571759700775, "global_step": 257357, "epoch": 2891, "lr": 6.659479574901539e-07}
{"train_loss": 0.018748747184872627, "global_step": 257358, "epoch": 2891, "lr": 6.658536568444229e-07}
{"train_loss": 0.04937666282057762, "global_step": 257359, "epoch": 2891, "lr": 6.657593628310643e-07}
{"train_loss": 0.03374277055263519, "global_step": 257360, "epoch": 2891, "lr": 6.656650754501004e-07}
{"train_loss": 0.03405360132455826, "global_step": 257361, "epoch": 2891, "lr": 6.655707947015311e-07}
{"train_loss": 0.02811191789805889, "global_step": 257362, "epoch": 2891, "lr": 6.65476520585373e-07}
{"train_loss": 0.03459826484322548, "global_step": 257363, "epoch": 2891, "lr": 6.653822531016429e-07}
{"train_loss": 0.04176294058561325, "global_step": 257364, "epoch": 2891, "lr": 6.652879922503519e-07}
{"train_loss": 0.059423044323921204, "global_step": 257365, "epoch": 2891, "lr": 6.651937380315109e-07}
{"train_loss": 0.08205784857273102, "global_step": 257366, "epoch": 2891, "lr": 6.650994904451313e-07}
{"train_loss": 0.042108308523893356, "global_step": 257367, "epoch": 2891, "lr": 6.650052494912295e-07}
{"train_loss": 0.057400502264499664, "global_step": 257368, "epoch": 2891, "lr": 6.649110151698168e-07}
{"train_loss": 0.05052809789776802, "global_step": 257369, "epoch": 2891, "lr": 6.648167874809041e-07}
{"train_loss": 0.019228894263505936, "global_step": 257370, "epoch": 2891, "lr": 6.647225664245027e-07}
{"train_loss": 0.01171096507459879, "global_step": 257371, "epoch": 2891, "lr": 6.646283520006346e-07}
{"train_loss": 0.044440463185310364, "global_step": 257372, "epoch": 2891, "lr": 6.645341442093001e-07}
{"train_loss": 0.027904469519853592, "global_step": 257373, "epoch": 2891, "lr": 6.644399430505211e-07}
{"train_loss": 0.04635358601808548, "global_step": 257374, "epoch": 2891, "lr": 6.643457485243032e-07}
{"train_loss": 0.03011149913072586, "global_step": 257375, "epoch": 2891, "lr": 6.642515606306687e-07}
{"train_loss": 0.052399370819330215, "global_step": 257376, "epoch": 2891, "lr": 6.641573793696177e-07}
{"train_loss": 0.053933240473270416, "global_step": 257377, "epoch": 2891, "lr": 6.640632047411777e-07}
{"train_loss": 0.034215692430734634, "global_step": 257378, "epoch": 2891, "lr": 6.639690367453433e-07}
{"train_loss": 0.03985637426376343, "global_step": 257379, "epoch": 2891, "lr": 6.638748753821477e-07}
{"train_loss": 0.035561028867959976, "global_step": 257380, "epoch": 2891, "lr": 6.637807206515855e-07}
{"train_loss": 0.029391489923000336, "global_step": 257381, "epoch": 2891, "lr": 6.63686572553679e-07}
{"train_loss": 0.10269378125667572, "global_step": 257382, "epoch": 2891, "lr": 6.635924310884334e-07}
{"train_loss": 0.059817712754011154, "global_step": 257383, "epoch": 2891, "lr": 6.634982962558711e-07}
{"train_loss": 0.06151794269680977, "global_step": 257384, "epoch": 2891, "lr": 6.634041680559977e-07}
{"train_loss": 0.05696101859211922, "global_step": 257385, "epoch": 2891, "lr": 6.633100464888353e-07}
{"train_loss": 0.07919832319021225, "global_step": 257386, "epoch": 2891, "lr": 6.63215931554384e-07}
{"train_loss": 0.045492576320101036, "global_step": 257387, "epoch": 2891, "lr": 6.631218232526604e-07, "val_loss": 9.574807167053223}
{"train_loss": 0.06663503497838974, "global_step": 257388, "epoch": 2892, "lr": 6.630277215836811e-07}
{"train_loss": 0.01698078028857708, "global_step": 257389, "epoch": 2892, "lr": 6.629336265474572e-07}
{"train_loss": 0.06256867200136185, "global_step": 257390, "epoch": 2892, "lr": 6.62839538144e-07}
{"train_loss": 0.05890502408146858, "global_step": 257391, "epoch": 2892, "lr": 6.627454563733204e-07}
{"train_loss": 0.04751136526465416, "global_step": 257392, "epoch": 2892, "lr": 6.626513812354407e-07}
{"train_loss": 0.01974901743233204, "global_step": 257393, "epoch": 2892, "lr": 6.625573127303553e-07}
{"train_loss": 0.036572251468896866, "global_step": 257394, "epoch": 2892, "lr": 6.624632508580974e-07}
{"train_loss": 0.04985219985246658, "global_step": 257395, "epoch": 2892, "lr": 6.623691956186617e-07}
{"train_loss": 0.04297366738319397, "global_step": 257396, "epoch": 2892, "lr": 6.622751470120758e-07}
{"train_loss": 0.019988849759101868, "global_step": 257397, "epoch": 2892, "lr": 6.621811050383397e-07}
{"train_loss": 0.05901939421892166, "global_step": 257398, "epoch": 2892, "lr": 6.620870696974812e-07}
{"train_loss": 0.03456109017133713, "global_step": 257399, "epoch": 2892, "lr": 6.619930409894948e-07}
{"train_loss": 0.04430652782320976, "global_step": 257400, "epoch": 2892, "lr": 6.618990189144081e-07}
{"train_loss": 0.02834285981953144, "global_step": 257401, "epoch": 2892, "lr": 6.618050034722212e-07}
{"train_loss": 0.03485366702079773, "global_step": 257402, "epoch": 2892, "lr": 6.617109946629563e-07}
{"train_loss": 0.03965376317501068, "global_step": 257403, "epoch": 2892, "lr": 6.616169924866244e-07}
{"train_loss": 0.058883704245090485, "global_step": 257404, "epoch": 2892, "lr": 6.615229969432369e-07}
{"train_loss": 0.08117743581533432, "global_step": 257405, "epoch": 2892, "lr": 6.614290080328045e-07}
{"train_loss": 0.039105646312236786, "global_step": 257406, "epoch": 2892, "lr": 6.613350257553441e-07}
{"train_loss": 0.03925316780805588, "global_step": 257407, "epoch": 2892, "lr": 6.612410501108612e-07}
{"train_loss": 0.07892535626888275, "global_step": 257408, "epoch": 2892, "lr": 6.611470810993725e-07}
{"train_loss": 0.0488261878490448, "global_step": 257409, "epoch": 2892, "lr": 6.610531187209001e-07}
{"train_loss": 0.020834701135754585, "global_step": 257410, "epoch": 2892, "lr": 6.609591629754386e-07}
{"train_loss": 0.04586678370833397, "global_step": 257411, "epoch": 2892, "lr": 6.608652138630156e-07}
{"train_loss": 0.07047846913337708, "global_step": 257412, "epoch": 2892, "lr": 6.607712713836311e-07}
{"train_loss": 0.037509895861148834, "global_step": 257413, "epoch": 2892, "lr": 6.60677335537313e-07}
{"train_loss": 0.02789647877216339, "global_step": 257414, "epoch": 2892, "lr": 6.605834063240613e-07}
{"train_loss": 0.032465897500514984, "global_step": 257415, "epoch": 2892, "lr": 6.604894837438924e-07}
{"train_loss": 0.03284899890422821, "global_step": 257416, "epoch": 2892, "lr": 6.603955677968176e-07}
{"train_loss": 0.03858833760023117, "global_step": 257417, "epoch": 2892, "lr": 6.603016584828592e-07}
{"train_loss": 0.03487764298915863, "global_step": 257418, "epoch": 2892, "lr": 6.602077558020115e-07}
{"train_loss": 0.06830618530511856, "global_step": 257419, "epoch": 2892, "lr": 6.601138597543077e-07}
{"train_loss": 0.028202148154377937, "global_step": 257420, "epoch": 2892, "lr": 6.600199703397425e-07}
{"train_loss": 0.036002203822135925, "global_step": 257421, "epoch": 2892, "lr": 6.59926087558338e-07}
{"train_loss": 0.024303600192070007, "global_step": 257422, "epoch": 2892, "lr": 6.598322114101052e-07}
{"train_loss": 0.03763186186552048, "global_step": 257423, "epoch": 2892, "lr": 6.597383418950608e-07}
{"train_loss": 0.05874558538198471, "global_step": 257424, "epoch": 2892, "lr": 6.59644479013205e-07}
{"train_loss": 0.007978927344083786, "global_step": 257425, "epoch": 2892, "lr": 6.595506227645653e-07}
{"train_loss": 0.04449160769581795, "global_step": 257426, "epoch": 2892, "lr": 6.594567731491419e-07}
{"train_loss": 0.03111453354358673, "global_step": 257427, "epoch": 2892, "lr": 6.593629301669624e-07}
{"train_loss": 0.021115422248840332, "global_step": 257428, "epoch": 2892, "lr": 6.592690938180212e-07}
{"train_loss": 0.02100658416748047, "global_step": 257429, "epoch": 2892, "lr": 6.591752641023463e-07}
{"train_loss": 0.0920749306678772, "global_step": 257430, "epoch": 2892, "lr": 6.590814410199431e-07}
{"train_loss": 0.03590855002403259, "global_step": 257431, "epoch": 2892, "lr": 6.58987624570817e-07}
{"train_loss": 0.06885135173797607, "global_step": 257432, "epoch": 2892, "lr": 6.588938147549961e-07}
{"train_loss": 0.09178102761507034, "global_step": 257433, "epoch": 2892, "lr": 6.588000115724857e-07}
{"train_loss": 0.02480509877204895, "global_step": 257434, "epoch": 2892, "lr": 6.58706215023297e-07}
{"train_loss": 0.031058091670274734, "global_step": 257435, "epoch": 2892, "lr": 6.58612425107441e-07}
{"train_loss": 0.04402170702815056, "global_step": 257436, "epoch": 2892, "lr": 6.585186418249401e-07}
{"train_loss": 0.09122344106435776, "global_step": 257437, "epoch": 2892, "lr": 6.584248651757941e-07}
{"train_loss": 0.03328719735145569, "global_step": 257438, "epoch": 2892, "lr": 6.583310951600197e-07}
{"train_loss": 0.06947619467973709, "global_step": 257439, "epoch": 2892, "lr": 6.582373317776335e-07}
{"train_loss": 0.030431821942329407, "global_step": 257440, "epoch": 2892, "lr": 6.581435750286525e-07}
{"train_loss": 0.046951305121183395, "global_step": 257441, "epoch": 2892, "lr": 6.580498249130707e-07}
{"train_loss": 0.03622018173336983, "global_step": 257442, "epoch": 2892, "lr": 6.579560814309216e-07}
{"train_loss": 0.016903700307011604, "global_step": 257443, "epoch": 2892, "lr": 6.578623445822052e-07}
{"train_loss": 0.042124662548303604, "global_step": 257444, "epoch": 2892, "lr": 6.577686143669381e-07}
{"train_loss": 0.04645425081253052, "global_step": 257445, "epoch": 2892, "lr": 6.576748907851316e-07}
{"train_loss": 0.047759760171175, "global_step": 257446, "epoch": 2892, "lr": 6.57581173836802e-07}
{"train_loss": 0.05127065256237984, "global_step": 257447, "epoch": 2892, "lr": 6.574874635219608e-07}
{"train_loss": 0.034162990748882294, "global_step": 257448, "epoch": 2892, "lr": 6.573937598406132e-07}
{"train_loss": 0.041042305529117584, "global_step": 257449, "epoch": 2892, "lr": 6.573000627927817e-07}
{"train_loss": 0.04129466041922569, "global_step": 257450, "epoch": 2892, "lr": 6.572063723784716e-07}
{"train_loss": 0.062313809990882874, "global_step": 257451, "epoch": 2892, "lr": 6.571126885977053e-07}
{"train_loss": 0.06168794259428978, "global_step": 257452, "epoch": 2892, "lr": 6.570190114504826e-07}
{"train_loss": 0.03189649432897568, "global_step": 257453, "epoch": 2892, "lr": 6.569253409368259e-07}
{"train_loss": 0.02759571745991707, "global_step": 257454, "epoch": 2892, "lr": 6.568316770567407e-07}
{"train_loss": 0.046933725476264954, "global_step": 257455, "epoch": 2892, "lr": 6.567380198102491e-07}
{"train_loss": 0.03294220566749573, "global_step": 257456, "epoch": 2892, "lr": 6.566443691973512e-07}
{"train_loss": 0.012016753666102886, "global_step": 257457, "epoch": 2892, "lr": 6.565507252180691e-07}
{"train_loss": 0.026342878118157387, "global_step": 257458, "epoch": 2892, "lr": 6.564570878724085e-07}
{"train_loss": 0.007898558862507343, "global_step": 257459, "epoch": 2892, "lr": 6.563634571603916e-07}
{"train_loss": 0.04542071372270584, "global_step": 257460, "epoch": 2892, "lr": 6.562698330820183e-07}
{"train_loss": 0.07054809480905533, "global_step": 257461, "epoch": 2892, "lr": 6.561762156373108e-07}
{"train_loss": 0.026569752022624016, "global_step": 257462, "epoch": 2892, "lr": 6.560826048262858e-07}
{"train_loss": 0.031725019216537476, "global_step": 257463, "epoch": 2892, "lr": 6.559890006489433e-07}
{"train_loss": 0.05156718194484711, "global_step": 257464, "epoch": 2892, "lr": 6.558954031053e-07}
{"train_loss": 0.028462860733270645, "global_step": 257465, "epoch": 2892, "lr": 6.558018121953779e-07}
{"train_loss": 0.0355958491563797, "global_step": 257466, "epoch": 2892, "lr": 6.557082279191718e-07}
{"train_loss": 0.03971163555979729, "global_step": 257467, "epoch": 2892, "lr": 6.556146502767091e-07}
{"train_loss": 0.013303166255354881, "global_step": 257468, "epoch": 2892, "lr": 6.555210792679955e-07}
{"train_loss": 0.0363420695066452, "global_step": 257469, "epoch": 2892, "lr": 6.554275148930533e-07}
{"train_loss": 0.06171680614352226, "global_step": 257470, "epoch": 2892, "lr": 6.553339571518824e-07}
{"train_loss": 0.0585818774998188, "global_step": 257471, "epoch": 2892, "lr": 6.552404060444939e-07}
{"train_loss": 0.003402155824005604, "global_step": 257472, "epoch": 2892, "lr": 6.551468615709155e-07}
{"train_loss": 0.07838515937328339, "global_step": 257473, "epoch": 2892, "lr": 6.550533237311474e-07}
{"train_loss": 0.03332684561610222, "global_step": 257474, "epoch": 2892, "lr": 6.54959792525206e-07}
{"train_loss": 0.04269924759864807, "global_step": 257475, "epoch": 2892, "lr": 6.548662679531082e-07}
{"train_loss": 0.0421905933533994, "global_step": 257476, "epoch": 2892, "lr": 6.547727500148593e-07, "val_loss": 9.60627269744873}
{"train_loss": 0.02442980371415615, "global_step": 257477, "epoch": 2893, "lr": 6.546792387104705e-07}
{"train_loss": 0.057856135070323944, "global_step": 257478, "epoch": 2893, "lr": 6.545857340399641e-07}
{"train_loss": 0.03786633536219597, "global_step": 257479, "epoch": 2893, "lr": 6.544922360033456e-07}
{"train_loss": 0.05898721516132355, "global_step": 257480, "epoch": 2893, "lr": 6.543987446006261e-07}
{"train_loss": 0.03826067969202995, "global_step": 257481, "epoch": 2893, "lr": 6.543052598318278e-07}
{"train_loss": 0.03774358332157135, "global_step": 257482, "epoch": 2893, "lr": 6.542117816969507e-07}
{"train_loss": 0.015430765226483345, "global_step": 257483, "epoch": 2893, "lr": 6.541183101960169e-07}
{"train_loss": 0.037546705454587936, "global_step": 257484, "epoch": 2893, "lr": 6.540248453290321e-07}
{"train_loss": 0.026283495128154755, "global_step": 257485, "epoch": 2893, "lr": 6.539313870960129e-07}
{"train_loss": 0.03379030525684357, "global_step": 257486, "epoch": 2893, "lr": 6.538379354969704e-07}
{"train_loss": 0.05197438225150108, "global_step": 257487, "epoch": 2893, "lr": 6.537444905319267e-07}
{"train_loss": 0.06185447797179222, "global_step": 257488, "epoch": 2893, "lr": 6.536510522008765e-07}
{"train_loss": 0.039274707436561584, "global_step": 257489, "epoch": 2893, "lr": 6.535576205038419e-07}
{"train_loss": 0.04405656456947327, "global_step": 257490, "epoch": 2893, "lr": 6.534641954408394e-07}
{"train_loss": 0.026968082413077354, "global_step": 257491, "epoch": 2893, "lr": 6.533707770118747e-07}
{"train_loss": 0.044192202389240265, "global_step": 257492, "epoch": 2893, "lr": 6.532773652169643e-07}
{"train_loss": 0.054686740040779114, "global_step": 257493, "epoch": 2893, "lr": 6.531839600561196e-07}
{"train_loss": 0.03942827135324478, "global_step": 257494, "epoch": 2893, "lr": 6.530905615293459e-07}
{"train_loss": 0.02725343219935894, "global_step": 257495, "epoch": 2893, "lr": 6.52997169636671e-07}
{"train_loss": 0.07309232652187347, "global_step": 257496, "epoch": 2893, "lr": 6.52903784378095e-07}
{"train_loss": 0.05231121554970741, "global_step": 257497, "epoch": 2893, "lr": 6.528104057536344e-07}
{"train_loss": 0.05328459292650223, "global_step": 257498, "epoch": 2893, "lr": 6.527170337633059e-07}
{"train_loss": 0.033228855580091476, "global_step": 257499, "epoch": 2893, "lr": 6.526236684071152e-07}
{"train_loss": 0.04698510468006134, "global_step": 257500, "epoch": 2893, "lr": 6.525303096850788e-07}
{"train_loss": 0.02623862214386463, "global_step": 257501, "epoch": 2893, "lr": 6.524369575972078e-07}
{"train_loss": 0.034105535596609116, "global_step": 257502, "epoch": 2893, "lr": 6.523436121435133e-07}
{"train_loss": 0.06988520920276642, "global_step": 257503, "epoch": 2893, "lr": 6.522502733240121e-07}
{"train_loss": 0.06533315777778625, "global_step": 257504, "epoch": 2893, "lr": 6.521569411387152e-07}
{"train_loss": 0.0681973546743393, "global_step": 257505, "epoch": 2893, "lr": 6.520636155876337e-07}
{"train_loss": 0.053085971623659134, "global_step": 257506, "epoch": 2893, "lr": 6.519702966707786e-07}
{"train_loss": 0.05021972209215164, "global_step": 257507, "epoch": 2893, "lr": 6.518769843881667e-07}
{"train_loss": 0.03699836507439613, "global_step": 257508, "epoch": 2893, "lr": 6.517836787398091e-07}
{"train_loss": 0.03512332960963249, "global_step": 257509, "epoch": 2893, "lr": 6.516903797257167e-07}
{"train_loss": 0.03132305666804314, "global_step": 257510, "epoch": 2893, "lr": 6.515970873459065e-07}
{"train_loss": 0.033016543835401535, "global_step": 257511, "epoch": 2893, "lr": 6.515038016003838e-07}
{"train_loss": 0.023520220071077347, "global_step": 257512, "epoch": 2893, "lr": 6.514105224891653e-07}
{"train_loss": 0.0809016078710556, "global_step": 257513, "epoch": 2893, "lr": 6.513172500122622e-07}
{"train_loss": 0.038636282086372375, "global_step": 257514, "epoch": 2893, "lr": 6.51223984169691e-07}
{"train_loss": 0.04888970032334328, "global_step": 257515, "epoch": 2893, "lr": 6.511307249614573e-07}
{"train_loss": 0.03279213607311249, "global_step": 257516, "epoch": 2893, "lr": 6.510374723875778e-07}
{"train_loss": 0.019353223964571953, "global_step": 257517, "epoch": 2893, "lr": 6.509442264480636e-07}
{"train_loss": 0.03319019079208374, "global_step": 257518, "epoch": 2893, "lr": 6.508509871429314e-07}
{"train_loss": 0.026421405375003815, "global_step": 257519, "epoch": 2893, "lr": 6.507577544721921e-07}
{"train_loss": 0.04219529405236244, "global_step": 257520, "epoch": 2893, "lr": 6.50664528435857e-07}
{"train_loss": 0.07585745304822922, "global_step": 257521, "epoch": 2893, "lr": 6.505713090339372e-07}
{"train_loss": 0.022055579349398613, "global_step": 257522, "epoch": 2893, "lr": 6.504780962664437e-07}
{"train_loss": 0.05667174234986305, "global_step": 257523, "epoch": 2893, "lr": 6.503848901333987e-07}
{"train_loss": 0.03277629241347313, "global_step": 257524, "epoch": 2893, "lr": 6.502916906348022e-07}
{"train_loss": 0.04692336916923523, "global_step": 257525, "epoch": 2893, "lr": 6.501984977706765e-07}
{"train_loss": 0.03787138685584068, "global_step": 257526, "epoch": 2893, "lr": 6.501053115410216e-07}
{"train_loss": 0.03537663072347641, "global_step": 257527, "epoch": 2893, "lr": 6.500121319458708e-07}
{"train_loss": 0.0379239097237587, "global_step": 257528, "epoch": 2893, "lr": 6.499189589852128e-07}
{"train_loss": 0.024427318945527077, "global_step": 257529, "epoch": 2893, "lr": 6.498257926590812e-07}
{"train_loss": 0.02968701720237732, "global_step": 257530, "epoch": 2893, "lr": 6.497326329674757e-07}
{"train_loss": 0.04747389256954193, "global_step": 257531, "epoch": 2893, "lr": 6.496394799104133e-07}
{"train_loss": 0.057510122656822205, "global_step": 257532, "epoch": 2893, "lr": 6.495463334878993e-07}
{"train_loss": 0.043768078088760376, "global_step": 257533, "epoch": 2893, "lr": 6.494531936999615e-07}
{"train_loss": 0.03708270564675331, "global_step": 257534, "epoch": 2893, "lr": 6.493600605465943e-07}
{"train_loss": 0.03465208783745766, "global_step": 257535, "epoch": 2893, "lr": 6.492669340278201e-07}
{"train_loss": 0.037528738379478455, "global_step": 257536, "epoch": 2893, "lr": 6.491738141436555e-07}
{"train_loss": 0.06639518588781357, "global_step": 257537, "epoch": 2893, "lr": 6.490807008941058e-07}
{"train_loss": 0.07258740067481995, "global_step": 257538, "epoch": 2893, "lr": 6.489875942791878e-07}
{"train_loss": 0.027171585708856583, "global_step": 257539, "epoch": 2893, "lr": 6.488944942989072e-07}
{"train_loss": 0.025726966559886932, "global_step": 257540, "epoch": 2893, "lr": 6.488014009532862e-07}
{"train_loss": 0.04944627359509468, "global_step": 257541, "epoch": 2893, "lr": 6.487083142423356e-07}
{"train_loss": 0.03966633230447769, "global_step": 257542, "epoch": 2893, "lr": 6.486152341660556e-07}
{"train_loss": 0.04515623301267624, "global_step": 257543, "epoch": 2893, "lr": 6.48522160724474e-07}
{"train_loss": 0.0635976642370224, "global_step": 257544, "epoch": 2893, "lr": 6.484290939175908e-07}
{"train_loss": 0.04809857904911041, "global_step": 257545, "epoch": 2893, "lr": 6.483360337454336e-07}
{"train_loss": 0.033416517078876495, "global_step": 257546, "epoch": 2893, "lr": 6.482429802080025e-07}
{"train_loss": 0.05271496623754501, "global_step": 257547, "epoch": 2893, "lr": 6.481499333053087e-07}
{"train_loss": 0.061336543411016464, "global_step": 257548, "epoch": 2893, "lr": 6.480568930373743e-07}
{"train_loss": 0.07385041564702988, "global_step": 257549, "epoch": 2893, "lr": 6.479638594042048e-07}
{"train_loss": 0.040107179433107376, "global_step": 257550, "epoch": 2893, "lr": 6.478708324058169e-07}
{"train_loss": 0.07683637738227844, "global_step": 257551, "epoch": 2893, "lr": 6.477778120422218e-07}
{"train_loss": 0.048512957990169525, "global_step": 257552, "epoch": 2893, "lr": 6.476847983134249e-07}
{"train_loss": 0.10480227321386337, "global_step": 257553, "epoch": 2893, "lr": 6.47591791219454e-07}
{"train_loss": 0.035372648388147354, "global_step": 257554, "epoch": 2893, "lr": 6.474987907603092e-07}
{"train_loss": 0.038512226194143295, "global_step": 257555, "epoch": 2893, "lr": 6.47405796936007e-07}
{"train_loss": 0.02937423624098301, "global_step": 257556, "epoch": 2893, "lr": 6.473128097465587e-07}
{"train_loss": 0.057156968861818314, "global_step": 257557, "epoch": 2893, "lr": 6.472198291919806e-07}
{"train_loss": 0.05257922783493996, "global_step": 257558, "epoch": 2893, "lr": 6.471268552722787e-07}
{"train_loss": 0.0625537857413292, "global_step": 257559, "epoch": 2893, "lr": 6.470338879874693e-07}
{"train_loss": 0.03847375139594078, "global_step": 257560, "epoch": 2893, "lr": 6.469409273375637e-07}
{"train_loss": 0.042621925473213196, "global_step": 257561, "epoch": 2893, "lr": 6.468479733225785e-07}
{"train_loss": 0.05752645432949066, "global_step": 257562, "epoch": 2893, "lr": 6.467550259425248e-07}
{"train_loss": 0.06015709042549133, "global_step": 257563, "epoch": 2893, "lr": 6.466620851974081e-07}
{"train_loss": 0.03300725296139717, "global_step": 257564, "epoch": 2893, "lr": 6.465691510872507e-07}
{"train_loss": 0.045517924147542944, "global_step": 257565, "epoch": 2893, "lr": 6.464762236120581e-07, "val_loss": 9.626203536987305}
{"train_loss": 0.03105792962014675, "global_step": 257566, "epoch": 2894, "lr": 6.463833027718469e-07}
{"train_loss": 0.03665568679571152, "global_step": 257567, "epoch": 2894, "lr": 6.462903885666282e-07}
{"train_loss": 0.07023521512746811, "global_step": 257568, "epoch": 2894, "lr": 6.461974809964078e-07}
{"train_loss": 0.011279099620878696, "global_step": 257569, "epoch": 2894, "lr": 6.461045800612131e-07}
{"train_loss": 0.04360291734337807, "global_step": 257570, "epoch": 2894, "lr": 6.460116857610444e-07}
{"train_loss": 0.06909936666488647, "global_step": 257571, "epoch": 2894, "lr": 6.459187980959125e-07}
{"train_loss": 0.05503213405609131, "global_step": 257572, "epoch": 2894, "lr": 6.458259170658454e-07}
{"train_loss": 0.03094053454697132, "global_step": 257573, "epoch": 2894, "lr": 6.457330426708374e-07}
{"train_loss": 0.03009996935725212, "global_step": 257574, "epoch": 2894, "lr": 6.456401749109164e-07}
{"train_loss": 0.04657876119017601, "global_step": 257575, "epoch": 2894, "lr": 6.455473137860824e-07}
{"train_loss": 0.039648495614528656, "global_step": 257576, "epoch": 2894, "lr": 6.454544592963518e-07}
{"train_loss": 0.029967134818434715, "global_step": 257577, "epoch": 2894, "lr": 6.453616114417415e-07}
{"train_loss": 0.01962943933904171, "global_step": 257578, "epoch": 2894, "lr": 6.452687702222626e-07}
{"train_loss": 0.036008428782224655, "global_step": 257579, "epoch": 2894, "lr": 6.451759356379206e-07}
{"train_loss": 0.07163053750991821, "global_step": 257580, "epoch": 2894, "lr": 6.450831076887376e-07}
{"train_loss": 0.06269218772649765, "global_step": 257581, "epoch": 2894, "lr": 6.449902863747192e-07}
{"train_loss": 0.07959544658660889, "global_step": 257582, "epoch": 2894, "lr": 6.448974716958822e-07}
{"train_loss": 0.042314209043979645, "global_step": 257583, "epoch": 2894, "lr": 6.448046636522375e-07}
{"train_loss": 0.037979211658239365, "global_step": 257584, "epoch": 2894, "lr": 6.447118622437909e-07}
{"train_loss": 0.10662954300642014, "global_step": 257585, "epoch": 2894, "lr": 6.446190674705699e-07}
{"train_loss": 0.028542064130306244, "global_step": 257586, "epoch": 2894, "lr": 6.445262793325746e-07}
{"train_loss": 0.14610368013381958, "global_step": 257587, "epoch": 2894, "lr": 6.444334978298216e-07}
{"train_loss": 0.04596193879842758, "global_step": 257588, "epoch": 2894, "lr": 6.443407229623222e-07}
{"train_loss": 0.059650860726833344, "global_step": 257589, "epoch": 2894, "lr": 6.442479547300872e-07}
{"train_loss": 0.045317113399505615, "global_step": 257590, "epoch": 2894, "lr": 6.441551931331336e-07}
{"train_loss": 0.026437362655997276, "global_step": 257591, "epoch": 2894, "lr": 6.440624381714722e-07}
{"train_loss": 0.060134366154670715, "global_step": 257592, "epoch": 2894, "lr": 6.439696898451142e-07}
{"train_loss": 0.06674518436193466, "global_step": 257593, "epoch": 2894, "lr": 6.438769481540763e-07}
{"train_loss": 0.03857456147670746, "global_step": 257594, "epoch": 2894, "lr": 6.43784213098364e-07}
{"train_loss": 0.06787388771772385, "global_step": 257595, "epoch": 2894, "lr": 6.43691484677994e-07}
{"train_loss": 0.032458242028951645, "global_step": 257596, "epoch": 2894, "lr": 6.435987628929774e-07}
{"train_loss": 0.0838278979063034, "global_step": 257597, "epoch": 2894, "lr": 6.435060477433308e-07}
{"train_loss": 0.054615143686532974, "global_step": 257598, "epoch": 2894, "lr": 6.434133392290598e-07}
{"train_loss": 0.033746086061000824, "global_step": 257599, "epoch": 2894, "lr": 6.43320637350181e-07}
{"train_loss": 0.03894895687699318, "global_step": 257600, "epoch": 2894, "lr": 6.432279421067056e-07}
{"train_loss": 0.03534439206123352, "global_step": 257601, "epoch": 2894, "lr": 6.431352534986501e-07}
{"train_loss": 0.02539221942424774, "global_step": 257602, "epoch": 2894, "lr": 6.430425715260202e-07}
{"train_loss": 0.058718569576740265, "global_step": 257603, "epoch": 2894, "lr": 6.429498961888325e-07}
{"train_loss": 0.022662881761789322, "global_step": 257604, "epoch": 2894, "lr": 6.428572274870981e-07}
{"train_loss": 0.05041452497243881, "global_step": 257605, "epoch": 2894, "lr": 6.427645654208336e-07}
{"train_loss": 0.06226072832942009, "global_step": 257606, "epoch": 2894, "lr": 6.426719099900392e-07}
{"train_loss": 0.012697323225438595, "global_step": 257607, "epoch": 2894, "lr": 6.425792611947423e-07}
{"train_loss": 0.03536505624651909, "global_step": 257608, "epoch": 2894, "lr": 6.424866190349487e-07}
{"train_loss": 0.09036128968000412, "global_step": 257609, "epoch": 2894, "lr": 6.423939835106696e-07}
{"train_loss": 0.026325080543756485, "global_step": 257610, "epoch": 2894, "lr": 6.423013546219215e-07}
{"train_loss": 0.0429295189678669, "global_step": 257611, "epoch": 2894, "lr": 6.422087323687098e-07}
{"train_loss": 0.04371533915400505, "global_step": 257612, "epoch": 2894, "lr": 6.42116116751057e-07}
{"train_loss": 0.015267997048795223, "global_step": 257613, "epoch": 2894, "lr": 6.420235077689685e-07}
{"train_loss": 0.02130831964313984, "global_step": 257614, "epoch": 2894, "lr": 6.419309054224609e-07}
{"train_loss": 0.02626320719718933, "global_step": 257615, "epoch": 2894, "lr": 6.418383097115343e-07}
{"train_loss": 0.03966953232884407, "global_step": 257616, "epoch": 2894, "lr": 6.417457206362221e-07}
{"train_loss": 0.06134333461523056, "global_step": 257617, "epoch": 2894, "lr": 6.416531381965185e-07}
{"train_loss": 0.04396675154566765, "global_step": 257618, "epoch": 2894, "lr": 6.415605623924459e-07}
{"train_loss": 0.034827765077352524, "global_step": 257619, "epoch": 2894, "lr": 6.414679932240153e-07}
{"train_loss": 0.037162452936172485, "global_step": 257620, "epoch": 2894, "lr": 6.413754306912379e-07}
{"train_loss": 0.052693918347358704, "global_step": 257621, "epoch": 2894, "lr": 6.412828747941247e-07}
{"train_loss": 0.021849164739251137, "global_step": 257622, "epoch": 2894, "lr": 6.411903255326868e-07}
{"train_loss": 0.0653938353061676, "global_step": 257623, "epoch": 2894, "lr": 6.410977829069409e-07}
{"train_loss": 0.02561163902282715, "global_step": 257624, "epoch": 2894, "lr": 6.410052469168981e-07}
{"train_loss": 0.0547730028629303, "global_step": 257625, "epoch": 2894, "lr": 6.409127175625696e-07}
{"train_loss": 0.02524627558887005, "global_step": 257626, "epoch": 2894, "lr": 6.408201948439718e-07}
{"train_loss": 0.034642595797777176, "global_step": 257627, "epoch": 2894, "lr": 6.407276787611105e-07}
{"train_loss": 0.10480436682701111, "global_step": 257628, "epoch": 2894, "lr": 6.406351693140023e-07}
{"train_loss": 0.11396469175815582, "global_step": 257629, "epoch": 2894, "lr": 6.405426665026637e-07}
{"train_loss": 0.035411711782217026, "global_step": 257630, "epoch": 2894, "lr": 6.404501703270948e-07}
{"train_loss": 0.07602208107709885, "global_step": 257631, "epoch": 2894, "lr": 6.403576807873235e-07}
{"train_loss": 0.08526771515607834, "global_step": 257632, "epoch": 2894, "lr": 6.402651978833496e-07}
{"train_loss": 0.042274318635463715, "global_step": 257633, "epoch": 2894, "lr": 6.401727216151953e-07}
{"train_loss": 0.06442778557538986, "global_step": 257634, "epoch": 2894, "lr": 6.400802519828608e-07}
{"train_loss": 0.035377755761146545, "global_step": 257635, "epoch": 2894, "lr": 6.399877889863737e-07}
{"train_loss": 0.05135532468557358, "global_step": 257636, "epoch": 2894, "lr": 6.398953326257285e-07}
{"train_loss": 0.06387560069561005, "global_step": 257637, "epoch": 2894, "lr": 6.398028829009583e-07}
{"train_loss": 0.03245635703206062, "global_step": 257638, "epoch": 2894, "lr": 6.39710439812058e-07}
{"train_loss": 0.0223979689180851, "global_step": 257639, "epoch": 2894, "lr": 6.396180033590493e-07}
{"train_loss": 0.04270295798778534, "global_step": 257640, "epoch": 2894, "lr": 6.395255735419436e-07}
{"train_loss": 0.06207771599292755, "global_step": 257641, "epoch": 2894, "lr": 6.39433150360752e-07}
{"train_loss": 0.04112021252512932, "global_step": 257642, "epoch": 2894, "lr": 6.393407338154855e-07}
{"train_loss": 0.026840416714549065, "global_step": 257643, "epoch": 2894, "lr": 6.392483239061553e-07}
{"train_loss": 0.09721002727746964, "global_step": 257644, "epoch": 2894, "lr": 6.391559206327835e-07}
{"train_loss": 0.0920943021774292, "global_step": 257645, "epoch": 2894, "lr": 6.390635239953702e-07}
{"train_loss": 0.029625380411744118, "global_step": 257646, "epoch": 2894, "lr": 6.389711339939375e-07}
{"train_loss": 0.011378251016139984, "global_step": 257647, "epoch": 2894, "lr": 6.388787506284855e-07}
{"train_loss": 0.03727578744292259, "global_step": 257648, "epoch": 2894, "lr": 6.387863738990419e-07}
{"train_loss": 0.017834920436143875, "global_step": 257649, "epoch": 2894, "lr": 6.386940038056122e-07}
{"train_loss": 0.04015825316309929, "global_step": 257650, "epoch": 2894, "lr": 6.386016403482075e-07}
{"train_loss": 0.06057550013065338, "global_step": 257651, "epoch": 2894, "lr": 6.385092835268392e-07}
{"train_loss": 0.0906352773308754, "global_step": 257652, "epoch": 2894, "lr": 6.384169333415235e-07}
{"train_loss": 0.020643364638090134, "global_step": 257653, "epoch": 2894, "lr": 6.383245897922663e-07}
{"train_loss": 0.04828246778107426, "global_step": 257654, "epoch": 2894, "lr": 6.382322528790896e-07, "val_loss": 9.639211654663086}
{"train_loss": 0.031012361869215965, "global_step": 257655, "epoch": 2895, "lr": 6.381399226019991e-07}
{"train_loss": 0.03992694243788719, "global_step": 257656, "epoch": 2895, "lr": 6.380475989610113e-07}
{"train_loss": 0.045027099549770355, "global_step": 257657, "epoch": 2895, "lr": 6.379552819561374e-07}
{"train_loss": 0.022274937480688095, "global_step": 257658, "epoch": 2895, "lr": 6.378629715873829e-07}
{"train_loss": 0.026703499257564545, "global_step": 257659, "epoch": 2895, "lr": 6.377706678547701e-07}
{"train_loss": 0.04021768644452095, "global_step": 257660, "epoch": 2895, "lr": 6.376783707583045e-07}
{"train_loss": 0.01735309511423111, "global_step": 257661, "epoch": 2895, "lr": 6.375860802980083e-07}
{"train_loss": 0.04037592187523842, "global_step": 257662, "epoch": 2895, "lr": 6.374937964738815e-07}
{"train_loss": 0.017274465411901474, "global_step": 257663, "epoch": 2895, "lr": 6.374015192859406e-07}
{"train_loss": 0.038183894008398056, "global_step": 257664, "epoch": 2895, "lr": 6.373092487342026e-07}
{"train_loss": 0.09036625921726227, "global_step": 257665, "epoch": 2895, "lr": 6.372169848186782e-07}
{"train_loss": 0.05844023823738098, "global_step": 257666, "epoch": 2895, "lr": 6.371247275393732e-07}
{"train_loss": 0.050817131996154785, "global_step": 257667, "epoch": 2895, "lr": 6.370324768963099e-07}
{"train_loss": 0.0430261455476284, "global_step": 257668, "epoch": 2895, "lr": 6.36940232889488e-07}
{"train_loss": 0.055533505976200104, "global_step": 257669, "epoch": 2895, "lr": 6.368479955189355e-07}
{"train_loss": 0.01181179191917181, "global_step": 257670, "epoch": 2895, "lr": 6.367557647846523e-07}
{"train_loss": 0.056119102984666824, "global_step": 257671, "epoch": 2895, "lr": 6.366635406866606e-07}
{"train_loss": 0.033904921263456345, "global_step": 257672, "epoch": 2895, "lr": 6.365713232249659e-07}
{"train_loss": 0.03940020129084587, "global_step": 257673, "epoch": 2895, "lr": 6.36479112399585e-07}
{"train_loss": 0.03713785856962204, "global_step": 257674, "epoch": 2895, "lr": 6.363869082105234e-07}
{"train_loss": 0.05426592752337456, "global_step": 257675, "epoch": 2895, "lr": 6.362947106578032e-07}
{"train_loss": 0.02680717594921589, "global_step": 257676, "epoch": 2895, "lr": 6.362025197414245e-07}
{"train_loss": 0.077633336186409, "global_step": 257677, "epoch": 2895, "lr": 6.361103354614095e-07}
{"train_loss": 0.043634895235300064, "global_step": 257678, "epoch": 2895, "lr": 6.360181578177693e-07}
{"train_loss": 0.05442437529563904, "global_step": 257679, "epoch": 2895, "lr": 6.359259868105094e-07}
{"train_loss": 0.04931996762752533, "global_step": 257680, "epoch": 2895, "lr": 6.358338224396576e-07}
{"train_loss": 0.042367782443761826, "global_step": 257681, "epoch": 2895, "lr": 6.357416647052083e-07}
{"train_loss": 0.04510427638888359, "global_step": 257682, "epoch": 2895, "lr": 6.356495136071839e-07}
{"train_loss": 0.044260893017053604, "global_step": 257683, "epoch": 2895, "lr": 6.355573691455952e-07}
{"train_loss": 0.040528539568185806, "global_step": 257684, "epoch": 2895, "lr": 6.35465231320459e-07}
{"train_loss": 0.061171144247055054, "global_step": 257685, "epoch": 2895, "lr": 6.353731001317753e-07}
{"train_loss": 0.046878620982170105, "global_step": 257686, "epoch": 2895, "lr": 6.352809755795663e-07}
{"train_loss": 0.08920825272798538, "global_step": 257687, "epoch": 2895, "lr": 6.351888576638432e-07}
{"train_loss": 0.03057665005326271, "global_step": 257688, "epoch": 2895, "lr": 6.350967463846169e-07}
{"train_loss": 0.05054617300629616, "global_step": 257689, "epoch": 2895, "lr": 6.350046417418986e-07}
{"train_loss": 0.02431493252515793, "global_step": 257690, "epoch": 2895, "lr": 6.349125437357051e-07}
{"train_loss": 0.0227425005286932, "global_step": 257691, "epoch": 2895, "lr": 6.348204523660417e-07}
{"train_loss": 0.02189001999795437, "global_step": 257692, "epoch": 2895, "lr": 6.347283676329308e-07}
{"train_loss": 0.04501589387655258, "global_step": 257693, "epoch": 2895, "lr": 6.346362895363722e-07}
{"train_loss": 0.06782471388578415, "global_step": 257694, "epoch": 2895, "lr": 6.345442180763938e-07}
{"train_loss": 0.07233574986457825, "global_step": 257695, "epoch": 2895, "lr": 6.344521532529901e-07}
{"train_loss": 0.04747084528207779, "global_step": 257696, "epoch": 2895, "lr": 6.343600950661887e-07}
{"train_loss": 0.04508522152900696, "global_step": 257697, "epoch": 2895, "lr": 6.342680435159953e-07}
{"train_loss": 0.059284862130880356, "global_step": 257698, "epoch": 2895, "lr": 6.341759986024209e-07}
{"train_loss": 0.060343679040670395, "global_step": 257699, "epoch": 2895, "lr": 6.340839603254822e-07}
{"train_loss": 0.04648739844560623, "global_step": 257700, "epoch": 2895, "lr": 6.339919286851848e-07}
{"train_loss": 0.04506291076540947, "global_step": 257701, "epoch": 2895, "lr": 6.338999036815507e-07}
{"train_loss": 0.03377081826329231, "global_step": 257702, "epoch": 2895, "lr": 6.338078853145857e-07}
{"train_loss": 0.04476437345147133, "global_step": 257703, "epoch": 2895, "lr": 6.337158735843007e-07}
{"train_loss": 0.02139701321721077, "global_step": 257704, "epoch": 2895, "lr": 6.336238684907126e-07}
{"train_loss": 0.015787670388817787, "global_step": 257705, "epoch": 2895, "lr": 6.335318700338322e-07}
{"train_loss": 0.05018678680062294, "global_step": 257706, "epoch": 2895, "lr": 6.334398782136708e-07}
{"train_loss": 0.06334763765335083, "global_step": 257707, "epoch": 2895, "lr": 6.33347893030245e-07}
{"train_loss": 0.015365296974778175, "global_step": 257708, "epoch": 2895, "lr": 6.332559144835604e-07}
{"train_loss": 0.026167484000325203, "global_step": 257709, "epoch": 2895, "lr": 6.331639425736336e-07}
{"train_loss": 0.02730310522019863, "global_step": 257710, "epoch": 2895, "lr": 6.330719773004756e-07}
{"train_loss": 0.049733951687812805, "global_step": 257711, "epoch": 2895, "lr": 6.329800186641033e-07}
{"train_loss": 0.04754897579550743, "global_step": 257712, "epoch": 2895, "lr": 6.32888066664522e-07}
{"train_loss": 0.02019045129418373, "global_step": 257713, "epoch": 2895, "lr": 6.327961213017486e-07}
{"train_loss": 0.019970901310443878, "global_step": 257714, "epoch": 2895, "lr": 6.327041825757885e-07}
{"train_loss": 0.02042870596051216, "global_step": 257715, "epoch": 2895, "lr": 6.326122504866638e-07}
{"train_loss": 0.04940808564424515, "global_step": 257716, "epoch": 2895, "lr": 6.325203250343858e-07}
{"train_loss": 0.0436190627515316, "global_step": 257717, "epoch": 2895, "lr": 6.3242840621896e-07}
{"train_loss": 0.05627192556858063, "global_step": 257718, "epoch": 2895, "lr": 6.32336494040403e-07}
{"train_loss": 0.053812213242053986, "global_step": 257719, "epoch": 2895, "lr": 6.32244588498726e-07}
{"train_loss": 0.05982358381152153, "global_step": 257720, "epoch": 2895, "lr": 6.321526895939456e-07}
{"train_loss": 0.057543832808732986, "global_step": 257721, "epoch": 2895, "lr": 6.320607973260617e-07}
{"train_loss": 0.03655734285712242, "global_step": 257722, "epoch": 2895, "lr": 6.319689116951078e-07}
{"train_loss": 0.05246299132704735, "global_step": 257723, "epoch": 2895, "lr": 6.318770327010726e-07}
{"train_loss": 0.012443291023373604, "global_step": 257724, "epoch": 2895, "lr": 6.317851603439895e-07}
{"train_loss": 0.033626411110162735, "global_step": 257725, "epoch": 2895, "lr": 6.316932946238529e-07}
{"train_loss": 0.053936317563056946, "global_step": 257726, "epoch": 2895, "lr": 6.316014355406907e-07}
{"train_loss": 0.0602143295109272, "global_step": 257727, "epoch": 2895, "lr": 6.315095830945028e-07}
{"train_loss": 0.011875588446855545, "global_step": 257728, "epoch": 2895, "lr": 6.314177372853114e-07}
{"train_loss": 0.025361046195030212, "global_step": 257729, "epoch": 2895, "lr": 6.313258981131165e-07}
{"train_loss": 0.023660287261009216, "global_step": 257730, "epoch": 2895, "lr": 6.312340655779458e-07}
{"train_loss": 0.03261909633874893, "global_step": 257731, "epoch": 2895, "lr": 6.311422396797994e-07}
{"train_loss": 0.032598257064819336, "global_step": 257732, "epoch": 2895, "lr": 6.310504204186995e-07}
{"train_loss": 0.027217892929911613, "global_step": 257733, "epoch": 2895, "lr": 6.309586077946461e-07}
{"train_loss": 0.02691255696117878, "global_step": 257734, "epoch": 2895, "lr": 6.308668018076613e-07}
{"train_loss": 0.03908945992588997, "global_step": 257735, "epoch": 2895, "lr": 6.307750024577619e-07}
{"train_loss": 0.03796839714050293, "global_step": 257736, "epoch": 2895, "lr": 6.306832097449422e-07}
{"train_loss": 0.08985419571399689, "global_step": 257737, "epoch": 2895, "lr": 6.305914236692301e-07}
{"train_loss": 0.04150870814919472, "global_step": 257738, "epoch": 2895, "lr": 6.30499644230631e-07}
{"train_loss": 0.04882080480456352, "global_step": 257739, "epoch": 2895, "lr": 6.304078714291672e-07}
{"train_loss": 0.06810344755649567, "global_step": 257740, "epoch": 2895, "lr": 6.303161052648332e-07}
{"train_loss": 0.0469026155769825, "global_step": 257741, "epoch": 2895, "lr": 6.302243457376567e-07}
{"train_loss": 0.029659627005457878, "global_step": 257742, "epoch": 2895, "lr": 6.301325928476431e-07}
{"train_loss": 0.04221844751638978, "global_step": 257743, "epoch": 2895, "lr": 6.300408465948093e-07, "val_loss": 9.619108200073242, "train_action_mse_error": 38.97510528564453}
{"train_loss": 0.06311749666929245, "global_step": 257744, "epoch": 2896, "lr": 6.299491069791608e-07}
{"train_loss": 0.03436325863003731, "global_step": 257745, "epoch": 2896, "lr": 6.298573740007196e-07}
{"train_loss": 0.0767727941274643, "global_step": 257746, "epoch": 2896, "lr": 6.29765647659486e-07}
{"train_loss": 0.023738296702504158, "global_step": 257747, "epoch": 2896, "lr": 6.29673927955482e-07}
{"train_loss": 0.05020787939429283, "global_step": 257748, "epoch": 2896, "lr": 6.295822148887132e-07}
{"train_loss": 0.06649962067604065, "global_step": 257749, "epoch": 2896, "lr": 6.294905084592018e-07}
{"train_loss": 0.011855601333081722, "global_step": 257750, "epoch": 2896, "lr": 6.293988086669478e-07}
{"train_loss": 0.04096711799502373, "global_step": 257751, "epoch": 2896, "lr": 6.293071155119678e-07}
{"train_loss": 0.02648109197616577, "global_step": 257752, "epoch": 2896, "lr": 6.292154289942842e-07}
{"train_loss": 0.02904999442398548, "global_step": 257753, "epoch": 2896, "lr": 6.291237491138913e-07}
{"train_loss": 0.031069330871105194, "global_step": 257754, "epoch": 2896, "lr": 6.290320758708168e-07}
{"train_loss": 0.044684603810310364, "global_step": 257755, "epoch": 2896, "lr": 6.289404092650663e-07}
{"train_loss": 0.048258453607559204, "global_step": 257756, "epoch": 2896, "lr": 6.288487492966566e-07}
{"train_loss": 0.03654095530509949, "global_step": 257757, "epoch": 2896, "lr": 6.287570959655875e-07}
{"train_loss": 0.01852157525718212, "global_step": 257758, "epoch": 2896, "lr": 6.286654492718868e-07}
{"train_loss": 0.05962246283888817, "global_step": 257759, "epoch": 2896, "lr": 6.285738092155602e-07}
{"train_loss": 0.03920932114124298, "global_step": 257760, "epoch": 2896, "lr": 6.284821757966186e-07}
{"train_loss": 0.024652158841490746, "global_step": 257761, "epoch": 2896, "lr": 6.283905490150732e-07}
{"train_loss": 0.048621345311403275, "global_step": 257762, "epoch": 2896, "lr": 6.282989288709462e-07}
{"train_loss": 0.024134699255228043, "global_step": 257763, "epoch": 2896, "lr": 6.28207315364232e-07}
{"train_loss": 0.03545417636632919, "global_step": 257764, "epoch": 2896, "lr": 6.28115708494964e-07}
{"train_loss": 0.037736572325229645, "global_step": 257765, "epoch": 2896, "lr": 6.280241082631367e-07}
{"train_loss": 0.029178719967603683, "global_step": 257766, "epoch": 2896, "lr": 6.279325146687776e-07}
{"train_loss": 0.012345113791525364, "global_step": 257767, "epoch": 2896, "lr": 6.278409277118813e-07}
{"train_loss": 0.08516314625740051, "global_step": 257768, "epoch": 2896, "lr": 6.277493473924812e-07}
{"train_loss": 0.04046427085995674, "global_step": 257769, "epoch": 2896, "lr": 6.27657773710566e-07}
{"train_loss": 0.08521862328052521, "global_step": 257770, "epoch": 2896, "lr": 6.275662066661691e-07}
{"train_loss": 0.03900187462568283, "global_step": 257771, "epoch": 2896, "lr": 6.274746462592906e-07}
{"train_loss": 0.033042166382074356, "global_step": 257772, "epoch": 2896, "lr": 6.273830924899526e-07}
{"train_loss": 0.021391689777374268, "global_step": 257773, "epoch": 2896, "lr": 6.272915453581551e-07}
{"train_loss": 0.027885677292943, "global_step": 257774, "epoch": 2896, "lr": 6.272000048639149e-07}
{"train_loss": 0.03784024715423584, "global_step": 257775, "epoch": 2896, "lr": 6.271084710072539e-07}
{"train_loss": 0.056736141443252563, "global_step": 257776, "epoch": 2896, "lr": 6.270169437881667e-07}
{"train_loss": 0.015035908669233322, "global_step": 257777, "epoch": 2896, "lr": 6.269254232066812e-07}
{"train_loss": 0.03199392929673195, "global_step": 257778, "epoch": 2896, "lr": 6.268339092628028e-07}
{"train_loss": 0.024725498631596565, "global_step": 257779, "epoch": 2896, "lr": 6.267424019565482e-07}
{"train_loss": 0.06997852772474289, "global_step": 257780, "epoch": 2896, "lr": 6.266509012879174e-07}
{"train_loss": 0.025565097108483315, "global_step": 257781, "epoch": 2896, "lr": 6.265594072569381e-07}
{"train_loss": 0.0273829884827137, "global_step": 257782, "epoch": 2896, "lr": 6.264679198636159e-07}
{"train_loss": 0.06022728607058525, "global_step": 257783, "epoch": 2896, "lr": 6.263764391079619e-07}
{"train_loss": 0.036762312054634094, "global_step": 257784, "epoch": 2896, "lr": 6.262849649899871e-07}
{"train_loss": 0.07132585346698761, "global_step": 257785, "epoch": 2896, "lr": 6.261934975097139e-07}
{"train_loss": 0.018471045419573784, "global_step": 257786, "epoch": 2896, "lr": 6.261020366671422e-07}
{"train_loss": 0.029420744627714157, "global_step": 257787, "epoch": 2896, "lr": 6.260105824622886e-07}
{"train_loss": 0.08437763899564743, "global_step": 257788, "epoch": 2896, "lr": 6.259191348951643e-07}
{"train_loss": 0.03413618728518486, "global_step": 257789, "epoch": 2896, "lr": 6.258276939657859e-07}
{"train_loss": 0.03533883020281792, "global_step": 257790, "epoch": 2896, "lr": 6.257362596741645e-07}
{"train_loss": 0.05999873951077461, "global_step": 257791, "epoch": 2896, "lr": 6.256448320203112e-07}
{"train_loss": 0.08983375877141953, "global_step": 257792, "epoch": 2896, "lr": 6.255534110042371e-07}
{"train_loss": 0.043832458555698395, "global_step": 257793, "epoch": 2896, "lr": 6.254619966259534e-07}
{"train_loss": 0.06071570888161659, "global_step": 257794, "epoch": 2896, "lr": 6.253705888854822e-07}
{"train_loss": 0.03399384766817093, "global_step": 257795, "epoch": 2896, "lr": 6.252791877828179e-07}
{"train_loss": 0.0320323184132576, "global_step": 257796, "epoch": 2896, "lr": 6.251877933179884e-07}
{"train_loss": 0.03185034170746803, "global_step": 257797, "epoch": 2896, "lr": 6.250964054909991e-07}
{"train_loss": 0.036713965237140656, "global_step": 257798, "epoch": 2896, "lr": 6.250050243018668e-07}
{"train_loss": 0.03182775527238846, "global_step": 257799, "epoch": 2896, "lr": 6.24913649750597e-07}
{"train_loss": 0.0393127016723156, "global_step": 257800, "epoch": 2896, "lr": 6.248222818372118e-07}
{"train_loss": 0.03169776499271393, "global_step": 257801, "epoch": 2896, "lr": 6.247309205617113e-07}
{"train_loss": 0.04595467448234558, "global_step": 257802, "epoch": 2896, "lr": 6.246395659241178e-07}
{"train_loss": 0.058084771037101746, "global_step": 257803, "epoch": 2896, "lr": 6.245482179244366e-07}
{"train_loss": 0.06615213304758072, "global_step": 257804, "epoch": 2896, "lr": 6.244568765626846e-07}
{"train_loss": 0.028216205537319183, "global_step": 257805, "epoch": 2896, "lr": 6.243655418388672e-07}
{"train_loss": 0.07095282524824142, "global_step": 257806, "epoch": 2896, "lr": 6.242742137530122e-07}
{"train_loss": 0.02208871766924858, "global_step": 257807, "epoch": 2896, "lr": 6.241828923051141e-07}
{"train_loss": 0.028875285759568214, "global_step": 257808, "epoch": 2896, "lr": 6.24091577495195e-07}
{"train_loss": 0.05083410441875458, "global_step": 257809, "epoch": 2896, "lr": 6.240002693232661e-07}
{"train_loss": 0.09124501794576645, "global_step": 257810, "epoch": 2896, "lr": 6.239089677893383e-07}
{"train_loss": 0.03129882365465164, "global_step": 257811, "epoch": 2896, "lr": 6.23817672893423e-07}
{"train_loss": 0.02893897145986557, "global_step": 257812, "epoch": 2896, "lr": 6.237263846355368e-07}
{"train_loss": 0.03635929897427559, "global_step": 257813, "epoch": 2896, "lr": 6.236351030156851e-07}
{"train_loss": 0.03797764331102371, "global_step": 257814, "epoch": 2896, "lr": 6.235438280338846e-07}
{"train_loss": 0.05036850646138191, "global_step": 257815, "epoch": 2896, "lr": 6.234525596901464e-07}
{"train_loss": 0.04686230048537254, "global_step": 257816, "epoch": 2896, "lr": 6.233612979844817e-07}
{"train_loss": 0.04185982793569565, "global_step": 257817, "epoch": 2896, "lr": 6.232700429169069e-07}
{"train_loss": 0.030646171420812607, "global_step": 257818, "epoch": 2896, "lr": 6.231787944874334e-07}
{"train_loss": 0.05223660171031952, "global_step": 257819, "epoch": 2896, "lr": 6.230875526960723e-07}
{"train_loss": 0.022293435409665108, "global_step": 257820, "epoch": 2896, "lr": 6.229963175428288e-07}
{"train_loss": 0.019573098048567772, "global_step": 257821, "epoch": 2896, "lr": 6.229050890277254e-07}
{"train_loss": 0.05890868976712227, "global_step": 257822, "epoch": 2896, "lr": 6.228138671507677e-07}
{"train_loss": 0.029839489609003067, "global_step": 257823, "epoch": 2896, "lr": 6.227226519119778e-07}
{"train_loss": 0.04506285861134529, "global_step": 257824, "epoch": 2896, "lr": 6.226314433113556e-07}
{"train_loss": 0.08299524337053299, "global_step": 257825, "epoch": 2896, "lr": 6.225402413489179e-07}
{"train_loss": 0.0397028811275959, "global_step": 257826, "epoch": 2896, "lr": 6.224490460246813e-07}
{"train_loss": 0.08748199045658112, "global_step": 257827, "epoch": 2896, "lr": 6.223578573386513e-07}
{"train_loss": 0.05585651472210884, "global_step": 257828, "epoch": 2896, "lr": 6.222666752908446e-07}
{"train_loss": 0.03550827130675316, "global_step": 257829, "epoch": 2896, "lr": 6.221754998812724e-07}
{"train_loss": 0.035011034458875656, "global_step": 257830, "epoch": 2896, "lr": 6.220843311099511e-07}
{"train_loss": 0.054829832166433334, "global_step": 257831, "epoch": 2896, "lr": 6.21993168976881e-07}
{"train_loss": 0.04279791630720824, "global_step": 257832, "epoch": 2896, "lr": 6.219020134820897e-07, "val_loss": 9.59106731414795}
{"train_loss": 0.054870788007974625, "global_step": 257833, "epoch": 2897, "lr": 6.218108646255772e-07}
{"train_loss": 0.02669515460729599, "global_step": 257834, "epoch": 2897, "lr": 6.2171972240736e-07}
{"train_loss": 0.051276128739118576, "global_step": 257835, "epoch": 2897, "lr": 6.216285868274552e-07}
{"train_loss": 0.042805131524801254, "global_step": 257836, "epoch": 2897, "lr": 6.21537457885868e-07}
{"train_loss": 0.040609296411275864, "global_step": 257837, "epoch": 2897, "lr": 6.214463355826094e-07}
{"train_loss": 0.02901589125394821, "global_step": 257838, "epoch": 2897, "lr": 6.213552199177019e-07}
{"train_loss": 0.024773485958576202, "global_step": 257839, "epoch": 2897, "lr": 6.212641108911455e-07}
{"train_loss": 0.06889921426773071, "global_step": 257840, "epoch": 2897, "lr": 6.211730085029621e-07}
{"train_loss": 0.04554174467921257, "global_step": 257841, "epoch": 2897, "lr": 6.210819127531576e-07}
{"train_loss": 0.03782076761126518, "global_step": 257842, "epoch": 2897, "lr": 6.209908236417483e-07}
{"train_loss": 0.060508936643600464, "global_step": 257843, "epoch": 2897, "lr": 6.208997411687457e-07}
{"train_loss": 0.05405726656317711, "global_step": 257844, "epoch": 2897, "lr": 6.208086653341605e-07}
{"train_loss": 0.02327524870634079, "global_step": 257845, "epoch": 2897, "lr": 6.207175961380096e-07}
{"train_loss": 0.04954318329691887, "global_step": 257846, "epoch": 2897, "lr": 6.206265335802985e-07}
{"train_loss": 0.018668506294488907, "global_step": 257847, "epoch": 2897, "lr": 6.205354776610384e-07}
{"train_loss": 0.010402947664260864, "global_step": 257848, "epoch": 2897, "lr": 6.204444283802513e-07}
{"train_loss": 0.05192957818508148, "global_step": 257849, "epoch": 2897, "lr": 6.203533857379373e-07}
{"train_loss": 0.06414099037647247, "global_step": 257850, "epoch": 2897, "lr": 6.202623497341187e-07}
{"train_loss": 0.011776539497077465, "global_step": 257851, "epoch": 2897, "lr": 6.201713203688065e-07}
{"train_loss": 0.03950478136539459, "global_step": 257852, "epoch": 2897, "lr": 6.200802976420062e-07}
{"train_loss": 0.032380249351263046, "global_step": 257853, "epoch": 2897, "lr": 6.1998928155374e-07}
{"train_loss": 0.05500181391835213, "global_step": 257854, "epoch": 2897, "lr": 6.198982721040081e-07}
{"train_loss": 0.05816998705267906, "global_step": 257855, "epoch": 2897, "lr": 6.198072692928325e-07}
{"train_loss": 0.030080270022153854, "global_step": 257856, "epoch": 2897, "lr": 6.197162731202189e-07}
{"train_loss": 0.07649940997362137, "global_step": 257857, "epoch": 2897, "lr": 6.196252835861837e-07}
{"train_loss": 0.012011677026748657, "global_step": 257858, "epoch": 2897, "lr": 6.195343006907384e-07}
{"train_loss": 0.019998319447040558, "global_step": 257859, "epoch": 2897, "lr": 6.194433244338993e-07}
{"train_loss": 0.04887108877301216, "global_step": 257860, "epoch": 2897, "lr": 6.193523548156666e-07}
{"train_loss": 0.03601616621017456, "global_step": 257861, "epoch": 2897, "lr": 6.192613918360624e-07}
{"train_loss": 0.012863324023783207, "global_step": 257862, "epoch": 2897, "lr": 6.191704354950978e-07}
{"train_loss": 0.04927733540534973, "global_step": 257863, "epoch": 2897, "lr": 6.190794857927839e-07}
{"train_loss": 0.06192111596465111, "global_step": 257864, "epoch": 2897, "lr": 6.18988542729132e-07}
{"train_loss": 0.03328876569867134, "global_step": 257865, "epoch": 2897, "lr": 6.188976063041529e-07}
{"train_loss": 0.07708857208490372, "global_step": 257866, "epoch": 2897, "lr": 6.18806676517869e-07}
{"train_loss": 0.04233383387327194, "global_step": 257867, "epoch": 2897, "lr": 6.187157533702748e-07}
{"train_loss": 0.042693354189395905, "global_step": 257868, "epoch": 2897, "lr": 6.186248368613979e-07}
{"train_loss": 0.037902940064668655, "global_step": 257869, "epoch": 2897, "lr": 6.185339269912438e-07}
{"train_loss": 0.029276598244905472, "global_step": 257870, "epoch": 2897, "lr": 6.184430237598293e-07}
{"train_loss": 0.02545849233865738, "global_step": 257871, "epoch": 2897, "lr": 6.1835212716716e-07}
{"train_loss": 0.08513601124286652, "global_step": 257872, "epoch": 2897, "lr": 6.182612372132524e-07}
{"train_loss": 0.043777238577604294, "global_step": 257873, "epoch": 2897, "lr": 6.181703538981176e-07}
{"train_loss": 0.03669971600174904, "global_step": 257874, "epoch": 2897, "lr": 6.180794772217668e-07}
{"train_loss": 0.06793496012687683, "global_step": 257875, "epoch": 2897, "lr": 6.179886071842112e-07}
{"train_loss": 0.03362587094306946, "global_step": 257876, "epoch": 2897, "lr": 6.178977437854727e-07}
{"train_loss": 0.05746858939528465, "global_step": 257877, "epoch": 2897, "lr": 6.178068870255515e-07}
{"train_loss": 0.05825459584593773, "global_step": 257878, "epoch": 2897, "lr": 6.177160369044588e-07}
{"train_loss": 0.10455843061208725, "global_step": 257879, "epoch": 2897, "lr": 6.176251934222221e-07}
{"train_loss": 0.044633544981479645, "global_step": 257880, "epoch": 2897, "lr": 6.175343565788361e-07}
{"train_loss": 0.05395707115530968, "global_step": 257881, "epoch": 2897, "lr": 6.174435263743284e-07}
{"train_loss": 0.02596479281783104, "global_step": 257882, "epoch": 2897, "lr": 6.173527028086934e-07}
{"train_loss": 0.01898740977048874, "global_step": 257883, "epoch": 2897, "lr": 6.172618858819645e-07}
{"train_loss": 0.05073751136660576, "global_step": 257884, "epoch": 2897, "lr": 6.171710755941362e-07}
{"train_loss": 0.10217700898647308, "global_step": 257885, "epoch": 2897, "lr": 6.170802719452251e-07}
{"train_loss": 0.048058949410915375, "global_step": 257886, "epoch": 2897, "lr": 6.169894749352534e-07}
{"train_loss": 0.05422048643231392, "global_step": 257887, "epoch": 2897, "lr": 6.168986845642211e-07}
{"train_loss": 0.03987601026892662, "global_step": 257888, "epoch": 2897, "lr": 6.168079008321448e-07}
{"train_loss": 0.026125382632017136, "global_step": 257889, "epoch": 2897, "lr": 6.167171237390412e-07}
{"train_loss": 0.02620883658528328, "global_step": 257890, "epoch": 2897, "lr": 6.166263532849104e-07}
{"train_loss": 0.039619140326976776, "global_step": 257891, "epoch": 2897, "lr": 6.1653558946978e-07}
{"train_loss": 0.07165517657995224, "global_step": 257892, "epoch": 2897, "lr": 6.164448322936444e-07}
{"train_loss": 0.031393833458423615, "global_step": 257893, "epoch": 2897, "lr": 6.163540817565372e-07}
{"train_loss": 0.0543687604367733, "global_step": 257894, "epoch": 2897, "lr": 6.162633378584526e-07}
{"train_loss": 0.0445198155939579, "global_step": 257895, "epoch": 2897, "lr": 6.161726005994073e-07}
{"train_loss": 0.06874412298202515, "global_step": 257896, "epoch": 2897, "lr": 6.16081869979418e-07}
{"train_loss": 0.06337964534759521, "global_step": 257897, "epoch": 2897, "lr": 6.159911459984957e-07}
{"train_loss": 0.030756302177906036, "global_step": 257898, "epoch": 2897, "lr": 6.159004286566517e-07}
{"train_loss": 0.035576242953538895, "global_step": 257899, "epoch": 2897, "lr": 6.158097179538968e-07}
{"train_loss": 0.030204640701413155, "global_step": 257900, "epoch": 2897, "lr": 6.15719013890248e-07}
{"train_loss": 0.037355273962020874, "global_step": 257901, "epoch": 2897, "lr": 6.156283164657107e-07}
{"train_loss": 0.03221988305449486, "global_step": 257902, "epoch": 2897, "lr": 6.155376256803014e-07}
{"train_loss": 0.017782654613256454, "global_step": 257903, "epoch": 2897, "lr": 6.154469415340314e-07}
{"train_loss": 0.05190949887037277, "global_step": 257904, "epoch": 2897, "lr": 6.153562640269117e-07}
{"train_loss": 0.03519779071211815, "global_step": 257905, "epoch": 2897, "lr": 6.152655931589534e-07}
{"train_loss": 0.07341564446687698, "global_step": 257906, "epoch": 2897, "lr": 6.151749289301734e-07}
{"train_loss": 0.030182182788848877, "global_step": 257907, "epoch": 2897, "lr": 6.150842713405825e-07}
{"train_loss": 0.08623464405536652, "global_step": 257908, "epoch": 2897, "lr": 6.149936203901918e-07}
{"train_loss": 0.047045860439538956, "global_step": 257909, "epoch": 2897, "lr": 6.14902976079007e-07}
{"train_loss": 0.0208250992000103, "global_step": 257910, "epoch": 2897, "lr": 6.148123384070559e-07}
{"train_loss": 0.05750397592782974, "global_step": 257911, "epoch": 2897, "lr": 6.147217073743328e-07}
{"train_loss": 0.03482414036989212, "global_step": 257912, "epoch": 2897, "lr": 6.146310829808654e-07}
{"train_loss": 0.04677276313304901, "global_step": 257913, "epoch": 2897, "lr": 6.14540465226654e-07}
{"train_loss": 0.06406257301568985, "global_step": 257914, "epoch": 2897, "lr": 6.14449854111715e-07}
{"train_loss": 0.05080120265483856, "global_step": 257915, "epoch": 2897, "lr": 6.143592496360651e-07}
{"train_loss": 0.03930806741118431, "global_step": 257916, "epoch": 2897, "lr": 6.142686517997098e-07}
{"train_loss": 0.07446984201669693, "global_step": 257917, "epoch": 2897, "lr": 6.14178060602666e-07}
{"train_loss": 0.07441262900829315, "global_step": 257918, "epoch": 2897, "lr": 6.140874760449389e-07}
{"train_loss": 0.057172954082489014, "global_step": 257919, "epoch": 2897, "lr": 6.139968981265565e-07}
{"train_loss": 0.029455037787556648, "global_step": 257920, "epoch": 2897, "lr": 6.139063268475076e-07}
{"train_loss": 0.04520407427897614, "global_step": 257921, "epoch": 2897, "lr": 6.138157622078255e-07, "val_loss": 9.625580787658691}
{"train_loss": 0.022972188889980316, "global_step": 257922, "epoch": 2898, "lr": 6.137252042075104e-07}
{"train_loss": 0.06003541126847267, "global_step": 257923, "epoch": 2898, "lr": 6.136346528465842e-07}
{"train_loss": 0.053913414478302, "global_step": 257924, "epoch": 2898, "lr": 6.13544108125047e-07}
{"train_loss": 0.03213536739349365, "global_step": 257925, "epoch": 2898, "lr": 6.134535700429211e-07}
{"train_loss": 0.035894885659217834, "global_step": 257926, "epoch": 2898, "lr": 6.133630386002121e-07}
{"train_loss": 0.06113853678107262, "global_step": 257927, "epoch": 2898, "lr": 6.132725137969309e-07}
{"train_loss": 0.04317709058523178, "global_step": 257928, "epoch": 2898, "lr": 6.131819956330998e-07}
{"train_loss": 0.021177373826503754, "global_step": 257929, "epoch": 2898, "lr": 6.130914841087243e-07}
{"train_loss": 0.06209598109126091, "global_step": 257930, "epoch": 2898, "lr": 6.130009792238101e-07}
{"train_loss": 0.049376875162124634, "global_step": 257931, "epoch": 2898, "lr": 6.129104809783793e-07}
{"train_loss": 0.04232253134250641, "global_step": 257932, "epoch": 2898, "lr": 6.128199893724374e-07}
{"train_loss": 0.020285649225115776, "global_step": 257933, "epoch": 2898, "lr": 6.127295044060011e-07}
{"train_loss": 0.06532712280750275, "global_step": 257934, "epoch": 2898, "lr": 6.126390260790871e-07}
{"train_loss": 0.03530178964138031, "global_step": 257935, "epoch": 2898, "lr": 6.125485543916954e-07}
{"train_loss": 0.02515365183353424, "global_step": 257936, "epoch": 2898, "lr": 6.124580893438536e-07}
{"train_loss": 0.032744646072387695, "global_step": 257937, "epoch": 2898, "lr": 6.123676309355564e-07}
{"train_loss": 0.03245778754353523, "global_step": 257938, "epoch": 2898, "lr": 6.122771791668259e-07}
{"train_loss": 0.041905343532562256, "global_step": 257939, "epoch": 2898, "lr": 6.12186734037673e-07}
{"train_loss": 0.030921725556254387, "global_step": 257940, "epoch": 2898, "lr": 6.120962955481091e-07}
{"train_loss": 0.041017886251211166, "global_step": 257941, "epoch": 2898, "lr": 6.120058636981507e-07}
{"train_loss": 0.04521394520998001, "global_step": 257942, "epoch": 2898, "lr": 6.119154384878034e-07}
{"train_loss": 0.051423877477645874, "global_step": 257943, "epoch": 2898, "lr": 6.118250199170783e-07}
{"train_loss": 0.05969145894050598, "global_step": 257944, "epoch": 2898, "lr": 6.117346079859976e-07}
{"train_loss": 0.03993794322013855, "global_step": 257945, "epoch": 2898, "lr": 6.116442026945668e-07}
{"train_loss": 0.048362646251916885, "global_step": 257946, "epoch": 2898, "lr": 6.11553804042797e-07}
{"train_loss": 0.04878242686390877, "global_step": 257947, "epoch": 2898, "lr": 6.114634120306995e-07}
{"train_loss": 0.05686916038393974, "global_step": 257948, "epoch": 2898, "lr": 6.113730266582907e-07}
{"train_loss": 0.06987352669239044, "global_step": 257949, "epoch": 2898, "lr": 6.112826479255817e-07}
{"train_loss": 0.030355535447597504, "global_step": 257950, "epoch": 2898, "lr": 6.111922758325838e-07}
{"train_loss": 0.026155538856983185, "global_step": 257951, "epoch": 2898, "lr": 6.111019103793081e-07}
{"train_loss": 0.027633758261799812, "global_step": 257952, "epoch": 2898, "lr": 6.110115515657655e-07}
{"train_loss": 0.029903193935751915, "global_step": 257953, "epoch": 2898, "lr": 6.109211993919728e-07}
{"train_loss": 0.018047599121928215, "global_step": 257954, "epoch": 2898, "lr": 6.108308538579355e-07}
{"train_loss": 0.05137898772954941, "global_step": 257955, "epoch": 2898, "lr": 6.107405149636758e-07}
{"train_loss": 0.029973722994327545, "global_step": 257956, "epoch": 2898, "lr": 6.106501827091993e-07}
{"train_loss": 0.012427751906216145, "global_step": 257957, "epoch": 2898, "lr": 6.105598570945225e-07}
{"train_loss": 0.07971958816051483, "global_step": 257958, "epoch": 2898, "lr": 6.104695381196457e-07}
{"train_loss": 0.07578161358833313, "global_step": 257959, "epoch": 2898, "lr": 6.103792257845964e-07}
{"train_loss": 0.018226491287350655, "global_step": 257960, "epoch": 2898, "lr": 6.102889200893746e-07}
{"train_loss": 0.04909655079245567, "global_step": 257961, "epoch": 2898, "lr": 6.101986210339971e-07}
{"train_loss": 0.03072846308350563, "global_step": 257962, "epoch": 2898, "lr": 6.101083286184805e-07}
{"train_loss": 0.04990522563457489, "global_step": 257963, "epoch": 2898, "lr": 6.100180428428304e-07}
{"train_loss": 0.04968111962080002, "global_step": 257964, "epoch": 2898, "lr": 6.099277637070633e-07}
{"train_loss": 0.05581972002983093, "global_step": 257965, "epoch": 2898, "lr": 6.098374912111904e-07}
{"train_loss": 0.06996187567710876, "global_step": 257966, "epoch": 2898, "lr": 6.097472253552228e-07}
{"train_loss": 0.041628431528806686, "global_step": 257967, "epoch": 2898, "lr": 6.096569661391716e-07}
{"train_loss": 0.05045262724161148, "global_step": 257968, "epoch": 2898, "lr": 6.095667135630478e-07}
{"train_loss": 0.02707262523472309, "global_step": 257969, "epoch": 2898, "lr": 6.094764676268627e-07}
{"train_loss": 0.03157451003789902, "global_step": 257970, "epoch": 2898, "lr": 6.093862283306384e-07}
{"train_loss": 0.08020768314599991, "global_step": 257971, "epoch": 2898, "lr": 6.092959956743749e-07}
{"train_loss": 0.027924612164497375, "global_step": 257972, "epoch": 2898, "lr": 6.092057696580944e-07}
{"train_loss": 0.05975744500756264, "global_step": 257973, "epoch": 2898, "lr": 6.091155502818025e-07}
{"train_loss": 0.02626977302134037, "global_step": 257974, "epoch": 2898, "lr": 6.090253375455102e-07}
{"train_loss": 0.049355484545230865, "global_step": 257975, "epoch": 2898, "lr": 6.089351314492342e-07}
{"train_loss": 0.03585875406861305, "global_step": 257976, "epoch": 2898, "lr": 6.088449319929857e-07}
{"train_loss": 0.04638265073299408, "global_step": 257977, "epoch": 2898, "lr": 6.087547391767756e-07}
{"train_loss": 0.029102366417646408, "global_step": 257978, "epoch": 2898, "lr": 6.086645530006207e-07}
{"train_loss": 0.0609569326043129, "global_step": 257979, "epoch": 2898, "lr": 6.085743734645211e-07}
{"train_loss": 0.07466543465852737, "global_step": 257980, "epoch": 2898, "lr": 6.084842005685043e-07}
{"train_loss": 0.060182325541973114, "global_step": 257981, "epoch": 2898, "lr": 6.083940343125649e-07}
{"train_loss": 0.04458612576127052, "global_step": 257982, "epoch": 2898, "lr": 6.083038746967362e-07}
{"train_loss": 0.038911834359169006, "global_step": 257983, "epoch": 2898, "lr": 6.082137217210071e-07}
{"train_loss": 0.11873486638069153, "global_step": 257984, "epoch": 2898, "lr": 6.081235753854109e-07}
{"train_loss": 0.07225432991981506, "global_step": 257985, "epoch": 2898, "lr": 6.080334356899475e-07}
{"train_loss": 0.04320692643523216, "global_step": 257986, "epoch": 2898, "lr": 6.079433026346337e-07}
{"train_loss": 0.03625717759132385, "global_step": 257987, "epoch": 2898, "lr": 6.07853176219475e-07}
{"train_loss": 0.015580752864480019, "global_step": 257988, "epoch": 2898, "lr": 6.077630564444936e-07}
{"train_loss": 0.03215307369828224, "global_step": 257989, "epoch": 2898, "lr": 6.076729433096951e-07}
{"train_loss": 0.07363466918468475, "global_step": 257990, "epoch": 2898, "lr": 6.075828368150904e-07}
{"train_loss": 0.02636670507490635, "global_step": 257991, "epoch": 2898, "lr": 6.074927369606964e-07}
{"train_loss": 0.043127287179231644, "global_step": 257992, "epoch": 2898, "lr": 6.074026437465241e-07}
{"train_loss": 0.06300580501556396, "global_step": 257993, "epoch": 2898, "lr": 6.073125571725846e-07}
{"train_loss": 0.06034490838646889, "global_step": 257994, "epoch": 2898, "lr": 6.07222477238889e-07}
{"train_loss": 0.027416670694947243, "global_step": 257995, "epoch": 2898, "lr": 6.071324039454484e-07}
{"train_loss": 0.0327632874250412, "global_step": 257996, "epoch": 2898, "lr": 6.070423372922795e-07}
{"train_loss": 0.07721525430679321, "global_step": 257997, "epoch": 2898, "lr": 6.069522772793934e-07}
{"train_loss": 0.031810030341148376, "global_step": 257998, "epoch": 2898, "lr": 6.068622239067956e-07}
{"train_loss": 0.05233241617679596, "global_step": 257999, "epoch": 2898, "lr": 6.067721771745083e-07}
{"train_loss": 0.043826039880514145, "global_step": 258000, "epoch": 2898, "lr": 6.066821370825371e-07}
{"train_loss": 0.10083360970020294, "global_step": 258001, "epoch": 2898, "lr": 6.06592103630893e-07}
{"train_loss": 0.03150021284818649, "global_step": 258002, "epoch": 2898, "lr": 6.065020768195928e-07}
{"train_loss": 0.0258741844445467, "global_step": 258003, "epoch": 2898, "lr": 6.06412056648642e-07}
{"train_loss": 0.04074689745903015, "global_step": 258004, "epoch": 2898, "lr": 6.063220431180628e-07}
{"train_loss": 0.06752808392047882, "global_step": 258005, "epoch": 2898, "lr": 6.062320362278551e-07}
{"train_loss": 0.01800660975277424, "global_step": 258006, "epoch": 2898, "lr": 6.061420359780412e-07}
{"train_loss": 0.057637397199869156, "global_step": 258007, "epoch": 2898, "lr": 6.060520423686322e-07}
{"train_loss": 0.04842841625213623, "global_step": 258008, "epoch": 2898, "lr": 6.059620553996336e-07}
{"train_loss": 0.09284739196300507, "global_step": 258009, "epoch": 2898, "lr": 6.058720750710623e-07}
{"train_loss": 0.04610272412177887, "global_step": 258010, "epoch": 2898, "lr": 6.05782101382929e-07, "val_loss": 9.647294044494629}
{"train_loss": 0.04153713583946228, "global_step": 258011, "epoch": 2899, "lr": 6.056921343352451e-07}
{"train_loss": 0.045984238386154175, "global_step": 258012, "epoch": 2899, "lr": 6.056021739280326e-07}
{"train_loss": 0.033081699162721634, "global_step": 258013, "epoch": 2899, "lr": 6.055122201612861e-07}
{"train_loss": 0.01229863055050373, "global_step": 258014, "epoch": 2899, "lr": 6.054222730350278e-07}
{"train_loss": 0.038605716079473495, "global_step": 258015, "epoch": 2899, "lr": 6.053323325492688e-07}
{"train_loss": 0.031193388625979424, "global_step": 258016, "epoch": 2899, "lr": 6.052423987040201e-07}
{"train_loss": 0.05430231988430023, "global_step": 258017, "epoch": 2899, "lr": 6.051524714992929e-07}
{"train_loss": 0.07038265466690063, "global_step": 258018, "epoch": 2899, "lr": 6.050625509351094e-07}
{"train_loss": 0.029371516779065132, "global_step": 258019, "epoch": 2899, "lr": 6.049726370114639e-07}
{"train_loss": 0.022139692679047585, "global_step": 258020, "epoch": 2899, "lr": 6.048827297283843e-07}
{"train_loss": 0.025301186367869377, "global_step": 258021, "epoch": 2899, "lr": 6.047928290858707e-07}
{"train_loss": 0.04291114956140518, "global_step": 258022, "epoch": 2899, "lr": 6.047029350839451e-07}
{"train_loss": 0.04597526416182518, "global_step": 258023, "epoch": 2899, "lr": 6.04613047722613e-07}
{"train_loss": 0.026181062683463097, "global_step": 258024, "epoch": 2899, "lr": 6.045231670018858e-07}
{"train_loss": 0.05180152505636215, "global_step": 258025, "epoch": 2899, "lr": 6.0443329292178e-07}
{"train_loss": 0.016625389456748962, "global_step": 258026, "epoch": 2899, "lr": 6.043434254823065e-07}
{"train_loss": 0.04415290430188179, "global_step": 258027, "epoch": 2899, "lr": 6.042535646834824e-07}
{"train_loss": 0.07009150832891464, "global_step": 258028, "epoch": 2899, "lr": 6.041637105253073e-07}
{"train_loss": 0.04668949544429779, "global_step": 258029, "epoch": 2899, "lr": 6.040738630078035e-07}
{"train_loss": 0.03108196146786213, "global_step": 258030, "epoch": 2899, "lr": 6.039840221309767e-07}
{"train_loss": 0.05234955996274948, "global_step": 258031, "epoch": 2899, "lr": 6.03894187894849e-07}
{"train_loss": 0.047429561614990234, "global_step": 258032, "epoch": 2899, "lr": 6.038043602994203e-07}
{"train_loss": 0.019322969019412994, "global_step": 258033, "epoch": 2899, "lr": 6.037145393447074e-07}
{"train_loss": 0.06651798635721207, "global_step": 258034, "epoch": 2899, "lr": 6.03624725030727e-07}
{"train_loss": 0.07324088364839554, "global_step": 258035, "epoch": 2899, "lr": 6.035349173574845e-07}
{"train_loss": 0.03398330509662628, "global_step": 258036, "epoch": 2899, "lr": 6.034451163249966e-07}
{"train_loss": 0.032970622181892395, "global_step": 258037, "epoch": 2899, "lr": 6.033553219332744e-07}
{"train_loss": 0.04767115041613579, "global_step": 258038, "epoch": 2899, "lr": 6.032655341823234e-07}
{"train_loss": 0.047142501920461655, "global_step": 258039, "epoch": 2899, "lr": 6.03175753072166e-07}
{"train_loss": 0.09276439249515533, "global_step": 258040, "epoch": 2899, "lr": 6.030859786028076e-07}
{"train_loss": 0.046389300376176834, "global_step": 258041, "epoch": 2899, "lr": 6.029962107742648e-07}
{"train_loss": 0.032226089388132095, "global_step": 258042, "epoch": 2899, "lr": 6.029064495865433e-07}
{"train_loss": 0.023653347045183182, "global_step": 258043, "epoch": 2899, "lr": 6.028166950396596e-07}
{"train_loss": 0.047673359513282776, "global_step": 258044, "epoch": 2899, "lr": 6.027269471336306e-07}
{"train_loss": 0.08100111037492752, "global_step": 258045, "epoch": 2899, "lr": 6.026372058684559e-07}
{"train_loss": 0.04854954034090042, "global_step": 258046, "epoch": 2899, "lr": 6.025474712441581e-07}
{"train_loss": 0.007922107353806496, "global_step": 258047, "epoch": 2899, "lr": 6.024577432607426e-07}
{"train_loss": 0.033219560980796814, "global_step": 258048, "epoch": 2899, "lr": 6.023680219182314e-07}
{"train_loss": 0.069914810359478, "global_step": 258049, "epoch": 2899, "lr": 6.022783072166194e-07}
{"train_loss": 0.041157156229019165, "global_step": 258050, "epoch": 2899, "lr": 6.021885991559395e-07}
{"train_loss": 0.04561655968427658, "global_step": 258051, "epoch": 2899, "lr": 6.020988977361863e-07}
{"train_loss": 0.023616548627614975, "global_step": 258052, "epoch": 2899, "lr": 6.02009202957382e-07}
{"train_loss": 0.03133757784962654, "global_step": 258053, "epoch": 2899, "lr": 6.019195148195322e-07}
{"train_loss": 0.04292059317231178, "global_step": 258054, "epoch": 2899, "lr": 6.018298333226591e-07}
{"train_loss": 0.056312091648578644, "global_step": 258055, "epoch": 2899, "lr": 6.017401584667626e-07}
{"train_loss": 0.08867563307285309, "global_step": 258056, "epoch": 2899, "lr": 6.016504902518594e-07}
{"train_loss": 0.04606293886899948, "global_step": 258057, "epoch": 2899, "lr": 6.015608286779606e-07}
{"train_loss": 0.03472200781106949, "global_step": 258058, "epoch": 2899, "lr": 6.014711737450884e-07}
{"train_loss": 0.044892702251672745, "global_step": 258059, "epoch": 2899, "lr": 6.013815254532374e-07}
{"train_loss": 0.12270016968250275, "global_step": 258060, "epoch": 2899, "lr": 6.012918838024295e-07}
{"train_loss": 0.06744100153446198, "global_step": 258061, "epoch": 2899, "lr": 6.012022487926817e-07}
{"train_loss": 0.050644006580114365, "global_step": 258062, "epoch": 2899, "lr": 6.011126204239937e-07}
{"train_loss": 0.04306502640247345, "global_step": 258063, "epoch": 2899, "lr": 6.010229986963878e-07}
{"train_loss": 0.05299828201532364, "global_step": 258064, "epoch": 2899, "lr": 6.009333836098697e-07}
{"train_loss": 0.027715379372239113, "global_step": 258065, "epoch": 2899, "lr": 6.008437751644613e-07}
{"train_loss": 0.042436979711055756, "global_step": 258066, "epoch": 2899, "lr": 6.007541733601574e-07}
{"train_loss": 0.0470123253762722, "global_step": 258067, "epoch": 2899, "lr": 6.006645781969855e-07}
{"train_loss": 0.036097388714551926, "global_step": 258068, "epoch": 2899, "lr": 6.005749896749512e-07}
{"train_loss": 0.056542590260505676, "global_step": 258069, "epoch": 2899, "lr": 6.004854077940714e-07}
{"train_loss": 0.08583739399909973, "global_step": 258070, "epoch": 2899, "lr": 6.003958325543457e-07}
{"train_loss": 0.03844992071390152, "global_step": 258071, "epoch": 2899, "lr": 6.003062639558021e-07}
{"train_loss": 0.039854373782873154, "global_step": 258072, "epoch": 2899, "lr": 6.002167019984406e-07}
{"train_loss": 0.05332368612289429, "global_step": 258073, "epoch": 2899, "lr": 6.001271466822833e-07}
{"train_loss": 0.007253037299960852, "global_step": 258074, "epoch": 2899, "lr": 6.000375980073303e-07}
{"train_loss": 0.04774761199951172, "global_step": 258075, "epoch": 2899, "lr": 5.999480559736037e-07}
{"train_loss": 0.06576650589704514, "global_step": 258076, "epoch": 2899, "lr": 5.998585205811091e-07}
{"train_loss": 0.06348846107721329, "global_step": 258077, "epoch": 2899, "lr": 5.997689918298688e-07}
{"train_loss": 0.01813613250851631, "global_step": 258078, "epoch": 2899, "lr": 5.996794697198826e-07}
{"train_loss": 0.04428500309586525, "global_step": 258079, "epoch": 2899, "lr": 5.995899542511674e-07}
{"train_loss": 0.04347706586122513, "global_step": 258080, "epoch": 2899, "lr": 5.995004454237341e-07}
{"train_loss": 0.04178042709827423, "global_step": 258081, "epoch": 2899, "lr": 5.994109432375938e-07}
{"train_loss": 0.027610212564468384, "global_step": 258082, "epoch": 2899, "lr": 5.993214476927689e-07}
{"train_loss": 0.05025231093168259, "global_step": 258083, "epoch": 2899, "lr": 5.992319587892537e-07}
{"train_loss": 0.04599972441792488, "global_step": 258084, "epoch": 2899, "lr": 5.991424765270759e-07}
{"train_loss": 0.033492039889097214, "global_step": 258085, "epoch": 2899, "lr": 5.990530009062357e-07}
{"train_loss": 0.03097834810614586, "global_step": 258086, "epoch": 2899, "lr": 5.989635319267606e-07}
{"train_loss": 0.017138388007879257, "global_step": 258087, "epoch": 2899, "lr": 5.988740695886397e-07}
{"train_loss": 0.02822580747306347, "global_step": 258088, "epoch": 2899, "lr": 5.987846138919062e-07}
{"train_loss": 0.03046768717467785, "global_step": 258089, "epoch": 2899, "lr": 5.986951648365658e-07}
{"train_loss": 0.04301870986819267, "global_step": 258090, "epoch": 2899, "lr": 5.986057224226238e-07}
{"train_loss": 0.05182750150561333, "global_step": 258091, "epoch": 2899, "lr": 5.985162866500971e-07}
{"train_loss": 0.0370931476354599, "global_step": 258092, "epoch": 2899, "lr": 5.984268575190022e-07}
{"train_loss": 0.04697676748037338, "global_step": 258093, "epoch": 2899, "lr": 5.983374350293391e-07}
{"train_loss": 0.015919599682092667, "global_step": 258094, "epoch": 2899, "lr": 5.982480191811357e-07}
{"train_loss": 0.04498651996254921, "global_step": 258095, "epoch": 2899, "lr": 5.981586099743919e-07}
{"train_loss": 0.04161284491419792, "global_step": 258096, "epoch": 2899, "lr": 5.980692074091188e-07}
{"train_loss": 0.04966401308774948, "global_step": 258097, "epoch": 2899, "lr": 5.979798114853441e-07}
{"train_loss": 0.0731184110045433, "global_step": 258098, "epoch": 2899, "lr": 5.978904222030624e-07}
{"train_loss": 0.04417970698598898, "global_step": 258099, "epoch": 2899, "lr": 5.978010395622902e-07, "val_loss": 9.63268756866455}
{"train_loss": 0.04000720754265785, "global_step": 258100, "epoch": 2900, "lr": 5.977116635630442e-07}
{"train_loss": 0.0772363543510437, "global_step": 258101, "epoch": 2900, "lr": 5.976222942053356e-07}
{"train_loss": 0.06031804159283638, "global_step": 258102, "epoch": 2900, "lr": 5.975329314891698e-07}
{"train_loss": 0.06992706656455994, "global_step": 258103, "epoch": 2900, "lr": 5.97443575414569e-07}
{"train_loss": 0.06904149055480957, "global_step": 258104, "epoch": 2900, "lr": 5.973542259815335e-07}
{"train_loss": 0.046770721673965454, "global_step": 258105, "epoch": 2900, "lr": 5.972648831900907e-07}
{"train_loss": 0.036120276898145676, "global_step": 258106, "epoch": 2900, "lr": 5.971755470402351e-07}
{"train_loss": 0.03652021661400795, "global_step": 258107, "epoch": 2900, "lr": 5.970862175319892e-07}
{"train_loss": 0.0616023875772953, "global_step": 258108, "epoch": 2900, "lr": 5.969968946653637e-07}
{"train_loss": 0.029636669903993607, "global_step": 258109, "epoch": 2900, "lr": 5.969075784403699e-07}
{"train_loss": 0.038159701973199844, "global_step": 258110, "epoch": 2900, "lr": 5.96818268857019e-07}
{"train_loss": 0.030011316761374474, "global_step": 258111, "epoch": 2900, "lr": 5.967289659153275e-07}
{"train_loss": 0.06523068249225616, "global_step": 258112, "epoch": 2900, "lr": 5.96639669615301e-07}
{"train_loss": 0.028660133481025696, "global_step": 258113, "epoch": 2900, "lr": 5.965503799569505e-07}
{"train_loss": 0.05501530319452286, "global_step": 258114, "epoch": 2900, "lr": 5.964610969402928e-07}
{"train_loss": 0.0186958946287632, "global_step": 258115, "epoch": 2900, "lr": 5.963718205653446e-07}
{"train_loss": 0.020869165658950806, "global_step": 258116, "epoch": 2900, "lr": 5.962825508321112e-07}
{"train_loss": 0.02698819898068905, "global_step": 258117, "epoch": 2900, "lr": 5.961932877405985e-07}
{"train_loss": 0.06584814935922623, "global_step": 258118, "epoch": 2900, "lr": 5.96104031290834e-07}
{"train_loss": 0.03864961490035057, "global_step": 258119, "epoch": 2900, "lr": 5.960147814828121e-07}
{"train_loss": 0.05602150410413742, "global_step": 258120, "epoch": 2900, "lr": 5.959255383165608e-07}
{"train_loss": 0.050546981394290924, "global_step": 258121, "epoch": 2900, "lr": 5.958363017920853e-07}
{"train_loss": 0.037891559302806854, "global_step": 258122, "epoch": 2900, "lr": 5.957470719093917e-07}
{"train_loss": 0.03179123252630234, "global_step": 258123, "epoch": 2900, "lr": 5.956578486685016e-07}
{"train_loss": 0.03960474953055382, "global_step": 258124, "epoch": 2900, "lr": 5.955686320694265e-07}
{"train_loss": 0.054988835006952286, "global_step": 258125, "epoch": 2900, "lr": 5.954794221121663e-07}
{"train_loss": 0.030609924346208572, "global_step": 258126, "epoch": 2900, "lr": 5.953902187967487e-07}
{"train_loss": 0.0328284315764904, "global_step": 258127, "epoch": 2900, "lr": 5.953010221231737e-07}
{"train_loss": 0.018364476040005684, "global_step": 258128, "epoch": 2900, "lr": 5.952118320914634e-07}
{"train_loss": 0.05374636873602867, "global_step": 258129, "epoch": 2900, "lr": 5.951226487016237e-07}
{"train_loss": 0.05591956526041031, "global_step": 258130, "epoch": 2900, "lr": 5.950334719536654e-07}
{"train_loss": 0.023597894236445427, "global_step": 258131, "epoch": 2900, "lr": 5.949443018476053e-07}
{"train_loss": 0.02449744939804077, "global_step": 258132, "epoch": 2900, "lr": 5.948551383834489e-07}
{"train_loss": 0.05475369840860367, "global_step": 258133, "epoch": 2900, "lr": 5.947659815612127e-07}
{"train_loss": 0.044065650552511215, "global_step": 258134, "epoch": 2900, "lr": 5.946768313809081e-07}
{"train_loss": 0.04133489355444908, "global_step": 258135, "epoch": 2900, "lr": 5.945876878425516e-07}
{"train_loss": 0.034926433116197586, "global_step": 258136, "epoch": 2900, "lr": 5.944985509461487e-07}
{"train_loss": 0.05814065411686897, "global_step": 258137, "epoch": 2900, "lr": 5.944094206917106e-07}
{"train_loss": 0.0671411007642746, "global_step": 258138, "epoch": 2900, "lr": 5.943202970792538e-07}
{"train_loss": 0.04070635512471199, "global_step": 258139, "epoch": 2900, "lr": 5.942311801087896e-07}
{"train_loss": 0.023046890273690224, "global_step": 258140, "epoch": 2900, "lr": 5.941420697803291e-07}
{"train_loss": 0.046450089663267136, "global_step": 258141, "epoch": 2900, "lr": 5.940529660938831e-07}
{"train_loss": 0.08134952187538147, "global_step": 258142, "epoch": 2900, "lr": 5.939638690494631e-07}
{"train_loss": 0.012026363983750343, "global_step": 258143, "epoch": 2900, "lr": 5.938747786470799e-07}
{"train_loss": 0.04305877536535263, "global_step": 258144, "epoch": 2900, "lr": 5.937856948867505e-07}
{"train_loss": 0.06857217103242874, "global_step": 258145, "epoch": 2900, "lr": 5.93696617768491e-07}
{"train_loss": 0.08125779032707214, "global_step": 258146, "epoch": 2900, "lr": 5.936075472922964e-07}
{"train_loss": 0.022324280813336372, "global_step": 258147, "epoch": 2900, "lr": 5.935184834581941e-07}
{"train_loss": 0.08563955128192902, "global_step": 258148, "epoch": 2900, "lr": 5.934294262661899e-07}
{"train_loss": 0.019452260807156563, "global_step": 258149, "epoch": 2900, "lr": 5.933403757163003e-07}
{"train_loss": 0.07123927772045135, "global_step": 258150, "epoch": 2900, "lr": 5.932513318085309e-07}
{"train_loss": 0.043399594724178314, "global_step": 258151, "epoch": 2900, "lr": 5.931622945428983e-07}
{"train_loss": 0.08388764411211014, "global_step": 258152, "epoch": 2900, "lr": 5.930732639194081e-07}
{"train_loss": 0.03570419177412987, "global_step": 258153, "epoch": 2900, "lr": 5.929842399380825e-07}
{"train_loss": 0.03485042601823807, "global_step": 258154, "epoch": 2900, "lr": 5.92895222598927e-07}
{"train_loss": 0.041626423597335815, "global_step": 258155, "epoch": 2900, "lr": 5.928062119019528e-07}
{"train_loss": 0.019303003326058388, "global_step": 258156, "epoch": 2900, "lr": 5.927172078471765e-07}
{"train_loss": 0.09927144646644592, "global_step": 258157, "epoch": 2900, "lr": 5.926282104346037e-07}
{"train_loss": 0.07107070088386536, "global_step": 258158, "epoch": 2900, "lr": 5.925392196642565e-07}
{"train_loss": 0.0637429729104042, "global_step": 258159, "epoch": 2900, "lr": 5.92450235536135e-07}
{"train_loss": 0.050420843064785004, "global_step": 258160, "epoch": 2900, "lr": 5.923612580502557e-07}
{"train_loss": 0.03299157693982124, "global_step": 258161, "epoch": 2900, "lr": 5.9227228720663e-07}
{"train_loss": 0.012126648798584938, "global_step": 258162, "epoch": 2900, "lr": 5.921833230052742e-07}
{"train_loss": 0.043590694665908813, "global_step": 258163, "epoch": 2900, "lr": 5.920943654461941e-07}
{"train_loss": 0.03533870726823807, "global_step": 258164, "epoch": 2900, "lr": 5.920054145294119e-07}
{"train_loss": 0.07734033465385437, "global_step": 258165, "epoch": 2900, "lr": 5.919164702549218e-07}
{"train_loss": 0.0224562156945467, "global_step": 258166, "epoch": 2900, "lr": 5.918275326227518e-07}
{"train_loss": 0.08617477864027023, "global_step": 258167, "epoch": 2900, "lr": 5.917386016329074e-07}
{"train_loss": 0.022461343556642532, "global_step": 258168, "epoch": 2900, "lr": 5.916496772854052e-07}
{"train_loss": 0.035125236958265305, "global_step": 258169, "epoch": 2900, "lr": 5.915607595802509e-07}
{"train_loss": 0.05073922872543335, "global_step": 258170, "epoch": 2900, "lr": 5.914718485174553e-07}
{"train_loss": 0.03605300933122635, "global_step": 258171, "epoch": 2900, "lr": 5.913829440970409e-07}
{"train_loss": 0.03077925182878971, "global_step": 258172, "epoch": 2900, "lr": 5.912940463190075e-07}
{"train_loss": 0.06711957603693008, "global_step": 258173, "epoch": 2900, "lr": 5.912051551833775e-07}
{"train_loss": 0.014696650207042694, "global_step": 258174, "epoch": 2900, "lr": 5.911162706901508e-07}
{"train_loss": 0.037659719586372375, "global_step": 258175, "epoch": 2900, "lr": 5.91027392839355e-07}
{"train_loss": 0.05077207460999489, "global_step": 258176, "epoch": 2900, "lr": 5.909385216309848e-07}
{"train_loss": 0.013856982812285423, "global_step": 258177, "epoch": 2900, "lr": 5.908496570650679e-07}
{"train_loss": 0.03519006818532944, "global_step": 258178, "epoch": 2900, "lr": 5.907607991416042e-07}
{"train_loss": 0.05452260747551918, "global_step": 258179, "epoch": 2900, "lr": 5.906719478606105e-07}
{"train_loss": 0.03460870683193207, "global_step": 258180, "epoch": 2900, "lr": 5.905831032220977e-07}
{"train_loss": 0.01778445765376091, "global_step": 258181, "epoch": 2900, "lr": 5.904942652260826e-07}
{"train_loss": 0.07102315127849579, "global_step": 258182, "epoch": 2900, "lr": 5.904054338725707e-07}
{"train_loss": 0.0364643819630146, "global_step": 258183, "epoch": 2900, "lr": 5.903166091615841e-07}
{"train_loss": 0.0470324382185936, "global_step": 258184, "epoch": 2900, "lr": 5.902277910931175e-07}
{"train_loss": 0.07147932797670364, "global_step": 258185, "epoch": 2900, "lr": 5.901389796671986e-07}
{"train_loss": 0.03800993412733078, "global_step": 258186, "epoch": 2900, "lr": 5.900501748838272e-07}
{"train_loss": 0.0682329460978508, "global_step": 258187, "epoch": 2900, "lr": 5.899613767430256e-07}
{"train_loss": 0.04528278275654557, "global_step": 258188, "epoch": 2900, "lr": 5.898725852447995e-07, "train/sim_max_reward_0": 0.6315256370946402, "train/sim_max_reward_1": 0.910134470354703, "train/sim_max_reward_2": 0.40417081344754296, "train/sim_max_reward_3": 0.6440206832038002, "train/sim_max_reward_4": 0.9033803456937738, "train/sim_max_reward_5": 0.983442319866121, "test/sim_max_reward_4500000": 0.4379702551246098, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 1.0, "test/sim_max_reward_4500003": 0.18091962665128797, "test/sim_max_reward_4500004": 0.1996671806449519, "test/sim_max_reward_4500005": 0.4030420011993842, "test/sim_max_reward_4500006": 1.0, "test/sim_max_reward_4500007": 0.9362228967371227, "test/sim_max_reward_4500008": 1.0, "test/sim_max_reward_4500009": 0.05039520329503581, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.5991895059224047, "test/sim_max_reward_4500014": 0.3748437231337764, "test/sim_max_reward_4500015": 0.8695428481072802, "test/sim_max_reward_4500016": 0.9876037628471904, "test/sim_max_reward_4500017": 0.9843363132460174, "test/sim_max_reward_4500018": 0.9598372502058183, "test/sim_max_reward_4500019": 1.0, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9690689951660535, "test/sim_max_reward_4500023": 0.61264328541393, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.9973702492174286, "test/sim_max_reward_4500026": 0.9917067324624669, "test/sim_max_reward_4500027": 1.0, "test/sim_max_reward_4500028": 0.6616460229800247, "test/sim_max_reward_4500029": 0.5640790650522463, "test/sim_max_reward_4500030": 0.9936331199628019, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.6555660566643561, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 1.0, "test/sim_max_reward_4500035": 0.8175045622215961, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.9961404712602898, "test/sim_max_reward_4500038": 0.16228084894068848, "test/sim_max_reward_4500039": 1.0, "test/sim_max_reward_4500040": 0.9951244998330494, "test/sim_max_reward_4500041": 0.519599405545571, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9866336815253407, "test/sim_max_reward_4500044": 0.5342977344650998, "test/sim_max_reward_4500045": 0.44913486083366644, "test/sim_max_reward_4500046": 0.4123754048199863, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 1.0, "train/mean_score": 0.7461123782767635, "test/mean_score": 0.6660475112695895, "val_loss": 9.578186988830566, "train_action_mse_error": 13.131032943725586}
{"train_loss": 0.05406574532389641, "global_step": 258189, "epoch": 2901, "lr": 5.897838003891653e-07}
{"train_loss": 0.03756901994347572, "global_step": 258190, "epoch": 2901, "lr": 5.896950221761288e-07}
{"train_loss": 0.02502765879034996, "global_step": 258191, "epoch": 2901, "lr": 5.89606250605712e-07}
{"train_loss": 0.03208020329475403, "global_step": 258192, "epoch": 2901, "lr": 5.895174856779151e-07}
{"train_loss": 0.06858276575803757, "global_step": 258193, "epoch": 2901, "lr": 5.894287273927601e-07}
{"train_loss": 0.06595884263515472, "global_step": 258194, "epoch": 2901, "lr": 5.893399757502527e-07}
{"train_loss": 0.005447462201118469, "global_step": 258195, "epoch": 2901, "lr": 5.892512307503984e-07}
{"train_loss": 0.0482206866145134, "global_step": 258196, "epoch": 2901, "lr": 5.891624923932249e-07}
{"train_loss": 0.03841571509838104, "global_step": 258197, "epoch": 2901, "lr": 5.890737606787322e-07}
{"train_loss": 0.03716418147087097, "global_step": 258198, "epoch": 2901, "lr": 5.889850356069426e-07}
{"train_loss": 0.042964886873960495, "global_step": 258199, "epoch": 2901, "lr": 5.888963171778561e-07}
{"train_loss": 0.04656184837222099, "global_step": 258200, "epoch": 2901, "lr": 5.888076053914949e-07}
{"train_loss": 0.06481482088565826, "global_step": 258201, "epoch": 2901, "lr": 5.887189002478588e-07}
{"train_loss": 0.006980365142226219, "global_step": 258202, "epoch": 2901, "lr": 5.886302017469758e-07}
{"train_loss": 0.01615782640874386, "global_step": 258203, "epoch": 2901, "lr": 5.885415098888402e-07}
{"train_loss": 0.0772509053349495, "global_step": 258204, "epoch": 2901, "lr": 5.8845282467348e-07}
{"train_loss": 0.0461098775267601, "global_step": 258205, "epoch": 2901, "lr": 5.883641461008949e-07}
{"train_loss": 0.08209047466516495, "global_step": 258206, "epoch": 2901, "lr": 5.882754741711071e-07}
{"train_loss": 0.035578757524490356, "global_step": 258207, "epoch": 2901, "lr": 5.881868088841224e-07}
{"train_loss": 0.04330170154571533, "global_step": 258208, "epoch": 2901, "lr": 5.880981502399463e-07}
{"train_loss": 0.04813854396343231, "global_step": 258209, "epoch": 2901, "lr": 5.880094982386064e-07}
{"train_loss": 0.03679850697517395, "global_step": 258210, "epoch": 2901, "lr": 5.879208528801028e-07}
{"train_loss": 0.029191069304943085, "global_step": 258211, "epoch": 2901, "lr": 5.878322141644577e-07}
{"train_loss": 0.06073787808418274, "global_step": 258212, "epoch": 2901, "lr": 5.877435820916655e-07}
{"train_loss": 0.0348353236913681, "global_step": 258213, "epoch": 2901, "lr": 5.876549566617595e-07}
{"train_loss": 0.01633552275598049, "global_step": 258214, "epoch": 2901, "lr": 5.875663378747343e-07}
{"train_loss": 0.061683960258960724, "global_step": 258215, "epoch": 2901, "lr": 5.874777257306118e-07}
{"train_loss": 0.041900038719177246, "global_step": 258216, "epoch": 2901, "lr": 5.873891202293979e-07}
{"train_loss": 0.031028445810079575, "global_step": 258217, "epoch": 2901, "lr": 5.873005213711091e-07}
{"train_loss": 0.028337543830275536, "global_step": 258218, "epoch": 2901, "lr": 5.87211929155751e-07}
{"train_loss": 0.07504041492938995, "global_step": 258219, "epoch": 2901, "lr": 5.871233435833456e-07}
{"train_loss": 0.05042370781302452, "global_step": 258220, "epoch": 2901, "lr": 5.870347646538987e-07}
{"train_loss": 0.02351091429591179, "global_step": 258221, "epoch": 2901, "lr": 5.869461923674213e-07}
{"train_loss": 0.05868856608867645, "global_step": 258222, "epoch": 2901, "lr": 5.868576267239301e-07}
{"train_loss": 0.06531021744012833, "global_step": 258223, "epoch": 2901, "lr": 5.86769067723425e-07}
{"train_loss": 0.05051223561167717, "global_step": 258224, "epoch": 2901, "lr": 5.866805153659339e-07}
{"train_loss": 0.03558773174881935, "global_step": 258225, "epoch": 2901, "lr": 5.865919696514565e-07}
{"train_loss": 0.0347863994538784, "global_step": 258226, "epoch": 2901, "lr": 5.865034305800154e-07}
{"train_loss": 0.028732556849718094, "global_step": 258227, "epoch": 2901, "lr": 5.86414898151616e-07}
{"train_loss": 0.027220407500863075, "global_step": 258228, "epoch": 2901, "lr": 5.863263723662693e-07}
{"train_loss": 0.0765417218208313, "global_step": 258229, "epoch": 2901, "lr": 5.862378532239865e-07}
{"train_loss": 0.04599246382713318, "global_step": 258230, "epoch": 2901, "lr": 5.861493407247842e-07}
{"train_loss": 0.054146476089954376, "global_step": 258231, "epoch": 2901, "lr": 5.860608348686736e-07}
{"train_loss": 0.049215711653232574, "global_step": 258232, "epoch": 2901, "lr": 5.859723356556601e-07}
{"train_loss": 0.02911239117383957, "global_step": 258233, "epoch": 2901, "lr": 5.85883843085766e-07}
{"train_loss": 0.023505818098783493, "global_step": 258234, "epoch": 2901, "lr": 5.857953571589913e-07}
{"train_loss": 0.0741136223077774, "global_step": 258235, "epoch": 2901, "lr": 5.857068778753583e-07}
{"train_loss": 0.04899223893880844, "global_step": 258236, "epoch": 2901, "lr": 5.856184052348723e-07}
{"train_loss": 0.04006913676857948, "global_step": 258237, "epoch": 2901, "lr": 5.855299392375502e-07}
{"train_loss": 0.05166802927851677, "global_step": 258238, "epoch": 2901, "lr": 5.854414798833974e-07}
{"train_loss": 0.040179673582315445, "global_step": 258239, "epoch": 2901, "lr": 5.853530271724306e-07}
{"train_loss": 0.12073107808828354, "global_step": 258240, "epoch": 2901, "lr": 5.852645811046665e-07}
{"train_loss": 0.03877851366996765, "global_step": 258241, "epoch": 2901, "lr": 5.85176141680105e-07}
{"train_loss": 0.08067089319229126, "global_step": 258242, "epoch": 2901, "lr": 5.850877088987627e-07}
{"train_loss": 0.028974276036024094, "global_step": 258243, "epoch": 2901, "lr": 5.849992827606621e-07}
{"train_loss": 0.044506803154945374, "global_step": 258244, "epoch": 2901, "lr": 5.849108632657974e-07}
{"train_loss": 0.052542511373758316, "global_step": 258245, "epoch": 2901, "lr": 5.848224504141964e-07}
{"train_loss": 0.045944489538669586, "global_step": 258246, "epoch": 2901, "lr": 5.847340442058535e-07}
{"train_loss": 0.07388409972190857, "global_step": 258247, "epoch": 2901, "lr": 5.846456446408022e-07}
{"train_loss": 0.06906825304031372, "global_step": 258248, "epoch": 2901, "lr": 5.845572517190368e-07}
{"train_loss": 0.03005260042846203, "global_step": 258249, "epoch": 2901, "lr": 5.844688654405794e-07}
{"train_loss": 0.030849719420075417, "global_step": 258250, "epoch": 2901, "lr": 5.843804858054303e-07}
{"train_loss": 0.021425852552056313, "global_step": 258251, "epoch": 2901, "lr": 5.84292112813617e-07}
{"train_loss": 0.05806644260883331, "global_step": 258252, "epoch": 2901, "lr": 5.842037464651396e-07}
{"train_loss": 0.03638054430484772, "global_step": 258253, "epoch": 2901, "lr": 5.841153867600147e-07}
{"train_loss": 0.051073964685201645, "global_step": 258254, "epoch": 2901, "lr": 5.840270336982479e-07}
{"train_loss": 0.05134992673993111, "global_step": 258255, "epoch": 2901, "lr": 5.839386872798669e-07}
{"train_loss": 0.03600990027189255, "global_step": 258256, "epoch": 2901, "lr": 5.838503475048607e-07}
{"train_loss": 0.08273850381374359, "global_step": 258257, "epoch": 2901, "lr": 5.837620143732625e-07}
{"train_loss": 0.024435220286250114, "global_step": 258258, "epoch": 2901, "lr": 5.836736878850724e-07}
{"train_loss": 0.01943177916109562, "global_step": 258259, "epoch": 2901, "lr": 5.835853680403014e-07}
{"train_loss": 0.060884345322847366, "global_step": 258260, "epoch": 2901, "lr": 5.834970548389717e-07}
{"train_loss": 0.0652891993522644, "global_step": 258261, "epoch": 2901, "lr": 5.834087482810835e-07}
{"train_loss": 0.03900071606040001, "global_step": 258262, "epoch": 2901, "lr": 5.833204483666588e-07}
{"train_loss": 0.10656998306512833, "global_step": 258263, "epoch": 2901, "lr": 5.832321550957031e-07}
{"train_loss": 0.01074595469981432, "global_step": 258264, "epoch": 2901, "lr": 5.831438684682278e-07}
{"train_loss": 0.03383826091885567, "global_step": 258265, "epoch": 2901, "lr": 5.830555884842437e-07}
{"train_loss": 0.03539091348648071, "global_step": 258266, "epoch": 2901, "lr": 5.829673151437731e-07}
{"train_loss": 0.07338246703147888, "global_step": 258267, "epoch": 2901, "lr": 5.828790484468161e-07}
{"train_loss": 0.06232864782214165, "global_step": 258268, "epoch": 2901, "lr": 5.827907883933892e-07}
{"train_loss": 0.028298884630203247, "global_step": 258269, "epoch": 2901, "lr": 5.827025349835035e-07}
{"train_loss": 0.04640255123376846, "global_step": 258270, "epoch": 2901, "lr": 5.826142882171648e-07}
{"train_loss": 0.027692968025803566, "global_step": 258271, "epoch": 2901, "lr": 5.825260480944006e-07}
{"train_loss": 0.026782069355249405, "global_step": 258272, "epoch": 2901, "lr": 5.824378146152109e-07}
{"train_loss": 0.05126018077135086, "global_step": 258273, "epoch": 2901, "lr": 5.82349587779607e-07}
{"train_loss": 0.05064128339290619, "global_step": 258274, "epoch": 2901, "lr": 5.82261367587611e-07}
{"train_loss": 0.04967832565307617, "global_step": 258275, "epoch": 2901, "lr": 5.821731540392172e-07}
{"train_loss": 0.08720123022794724, "global_step": 258276, "epoch": 2901, "lr": 5.820849471344592e-07}
{"train_loss": 0.04633628942197963, "global_step": 258277, "epoch": 2901, "lr": 5.819967468733256e-07, "val_loss": 9.717732429504395}
{"train_loss": 0.07049848139286041, "global_step": 258278, "epoch": 2902, "lr": 5.819085532558444e-07}
{"train_loss": 0.0336991585791111, "global_step": 258279, "epoch": 2902, "lr": 5.818203662820266e-07}
{"train_loss": 0.046294186264276505, "global_step": 258280, "epoch": 2902, "lr": 5.817321859518776e-07}
{"train_loss": 0.053026966750621796, "global_step": 258281, "epoch": 2902, "lr": 5.816440122654198e-07}
{"train_loss": 0.05350370705127716, "global_step": 258282, "epoch": 2902, "lr": 5.815558452226477e-07}
{"train_loss": 0.021465690806508064, "global_step": 258283, "epoch": 2902, "lr": 5.814676848235889e-07}
{"train_loss": 0.04173867776989937, "global_step": 258284, "epoch": 2902, "lr": 5.813795310682491e-07}
{"train_loss": 0.08275773376226425, "global_step": 258285, "epoch": 2902, "lr": 5.812913839566391e-07}
{"train_loss": 0.04349413141608238, "global_step": 258286, "epoch": 2902, "lr": 5.812032434887705e-07}
{"train_loss": 0.0805116519331932, "global_step": 258287, "epoch": 2902, "lr": 5.811151096646594e-07}
{"train_loss": 0.03623601421713829, "global_step": 258288, "epoch": 2902, "lr": 5.810269824843118e-07}
{"train_loss": 0.028767798095941544, "global_step": 258289, "epoch": 2902, "lr": 5.809388619477496e-07}
{"train_loss": 0.04151984304189682, "global_step": 258290, "epoch": 2902, "lr": 5.808507480549674e-07}
{"train_loss": 0.043706826865673065, "global_step": 258291, "epoch": 2902, "lr": 5.807626408059986e-07}
{"train_loss": 0.04266037046909332, "global_step": 258292, "epoch": 2902, "lr": 5.806745402008373e-07}
{"train_loss": 0.04246997460722923, "global_step": 258293, "epoch": 2902, "lr": 5.80586446239506e-07}
{"train_loss": 0.042552269995212555, "global_step": 258294, "epoch": 2902, "lr": 5.804983589220047e-07}
{"train_loss": 0.06222442537546158, "global_step": 258295, "epoch": 2902, "lr": 5.804102782483612e-07}
{"train_loss": 0.04414331912994385, "global_step": 258296, "epoch": 2902, "lr": 5.803222042185752e-07}
{"train_loss": 0.03198546916246414, "global_step": 258297, "epoch": 2902, "lr": 5.802341368326636e-07}
{"train_loss": 0.032062239944934845, "global_step": 258298, "epoch": 2902, "lr": 5.801460760906374e-07}
{"train_loss": 0.024485943838953972, "global_step": 258299, "epoch": 2902, "lr": 5.800580219925078e-07}
{"train_loss": 0.09262033551931381, "global_step": 258300, "epoch": 2902, "lr": 5.799699745382858e-07}
{"train_loss": 0.09019946306943893, "global_step": 258301, "epoch": 2902, "lr": 5.798819337279826e-07}
{"train_loss": 0.06227332353591919, "global_step": 258302, "epoch": 2902, "lr": 5.797938995616204e-07}
{"train_loss": 0.0346335731446743, "global_step": 258303, "epoch": 2902, "lr": 5.797058720391934e-07}
{"train_loss": 0.025061480700969696, "global_step": 258304, "epoch": 2902, "lr": 5.796178511607298e-07}
{"train_loss": 0.05003784969449043, "global_step": 258305, "epoch": 2902, "lr": 5.795298369262347e-07}
{"train_loss": 0.05461621657013893, "global_step": 258306, "epoch": 2902, "lr": 5.79441829335714e-07}
{"train_loss": 0.018751023337244987, "global_step": 258307, "epoch": 2902, "lr": 5.793538283891897e-07}
{"train_loss": 0.10355033725500107, "global_step": 258308, "epoch": 2902, "lr": 5.792658340866619e-07}
{"train_loss": 0.06513644009828568, "global_step": 258309, "epoch": 2902, "lr": 5.791778464281528e-07}
{"train_loss": 0.07368861883878708, "global_step": 258310, "epoch": 2902, "lr": 5.790898654136734e-07}
{"train_loss": 0.04343584552407265, "global_step": 258311, "epoch": 2902, "lr": 5.790018910432349e-07}
{"train_loss": 0.07728217542171478, "global_step": 258312, "epoch": 2902, "lr": 5.789139233168428e-07}
{"train_loss": 0.062312472611665726, "global_step": 258313, "epoch": 2902, "lr": 5.788259622345138e-07}
{"train_loss": 0.031146200373768806, "global_step": 258314, "epoch": 2902, "lr": 5.78738007796259e-07}
{"train_loss": 0.04297402873635292, "global_step": 258315, "epoch": 2902, "lr": 5.786500600020894e-07}
{"train_loss": 0.05962614715099335, "global_step": 258316, "epoch": 2902, "lr": 5.785621188520218e-07}
{"train_loss": 0.05296654999256134, "global_step": 258317, "epoch": 2902, "lr": 5.784741843460617e-07}
{"train_loss": 0.05240889638662338, "global_step": 258318, "epoch": 2902, "lr": 5.783862564842202e-07}
{"train_loss": 0.04220903664827347, "global_step": 258319, "epoch": 2902, "lr": 5.782983352665195e-07}
{"train_loss": 0.054341644048690796, "global_step": 258320, "epoch": 2902, "lr": 5.782104206929595e-07}
{"train_loss": 0.061316102743148804, "global_step": 258321, "epoch": 2902, "lr": 5.78122512763557e-07}
{"train_loss": 0.07657253742218018, "global_step": 258322, "epoch": 2902, "lr": 5.780346114783231e-07}
{"train_loss": 0.047671880573034286, "global_step": 258323, "epoch": 2902, "lr": 5.779467168372743e-07}
{"train_loss": 0.04039493948221207, "global_step": 258324, "epoch": 2902, "lr": 5.778588288404163e-07}
{"train_loss": 0.03540894016623497, "global_step": 258325, "epoch": 2902, "lr": 5.777709474877602e-07}
{"train_loss": 0.05940042436122894, "global_step": 258326, "epoch": 2902, "lr": 5.776830727793225e-07}
{"train_loss": 0.03034873865544796, "global_step": 258327, "epoch": 2902, "lr": 5.775952047151145e-07}
{"train_loss": 0.05469675362110138, "global_step": 258328, "epoch": 2902, "lr": 5.775073432951417e-07}
{"train_loss": 0.04638909175992012, "global_step": 258329, "epoch": 2902, "lr": 5.774194885194262e-07}
{"train_loss": 0.04185912758111954, "global_step": 258330, "epoch": 2902, "lr": 5.77331640387968e-07}
{"train_loss": 0.05050849914550781, "global_step": 258331, "epoch": 2902, "lr": 5.772437989007895e-07}
{"train_loss": 0.04102645814418793, "global_step": 258332, "epoch": 2902, "lr": 5.771559640578961e-07}
{"train_loss": 0.029777824878692627, "global_step": 258333, "epoch": 2902, "lr": 5.770681358593044e-07}
{"train_loss": 0.022402551025152206, "global_step": 258334, "epoch": 2902, "lr": 5.769803143050201e-07}
{"train_loss": 0.038205940276384354, "global_step": 258335, "epoch": 2902, "lr": 5.768924993950597e-07}
{"train_loss": 0.04197338968515396, "global_step": 258336, "epoch": 2902, "lr": 5.768046911294345e-07}
{"train_loss": 0.029407795518636703, "global_step": 258337, "epoch": 2902, "lr": 5.767168895081554e-07}
{"train_loss": 0.023829372599720955, "global_step": 258338, "epoch": 2902, "lr": 5.766290945312336e-07}
{"train_loss": 0.02011321485042572, "global_step": 258339, "epoch": 2902, "lr": 5.765413061986857e-07}
{"train_loss": 0.061993785202503204, "global_step": 258340, "epoch": 2902, "lr": 5.764535245105173e-07}
{"train_loss": 0.038403403013944626, "global_step": 258341, "epoch": 2902, "lr": 5.763657494667396e-07}
{"train_loss": 0.028230534866452217, "global_step": 258342, "epoch": 2902, "lr": 5.76277981067369e-07}
{"train_loss": 0.0817732959985733, "global_step": 258343, "epoch": 2902, "lr": 5.761902193124169e-07}
{"train_loss": 0.08083325624465942, "global_step": 258344, "epoch": 2902, "lr": 5.761024642018886e-07}
{"train_loss": 0.03382455185055733, "global_step": 258345, "epoch": 2902, "lr": 5.760147157358065e-07}
{"train_loss": 0.06705524772405624, "global_step": 258346, "epoch": 2902, "lr": 5.75926973914176e-07}
{"train_loss": 0.03242456167936325, "global_step": 258347, "epoch": 2902, "lr": 5.758392387370082e-07}
{"train_loss": 0.04769587516784668, "global_step": 258348, "epoch": 2902, "lr": 5.757515102043198e-07}
{"train_loss": 0.05768835172057152, "global_step": 258349, "epoch": 2902, "lr": 5.756637883161109e-07}
{"train_loss": 0.03271689638495445, "global_step": 258350, "epoch": 2902, "lr": 5.755760730724036e-07}
{"train_loss": 0.024004308506846428, "global_step": 258351, "epoch": 2902, "lr": 5.754883644732145e-07}
{"train_loss": 0.026575490832328796, "global_step": 258352, "epoch": 2902, "lr": 5.754006625185437e-07}
{"train_loss": 0.036842647939920425, "global_step": 258353, "epoch": 2902, "lr": 5.753129672084078e-07}
{"train_loss": 0.018791399896144867, "global_step": 258354, "epoch": 2902, "lr": 5.75225278542818e-07}
{"train_loss": 0.024364935234189034, "global_step": 258355, "epoch": 2902, "lr": 5.751375965217908e-07}
{"train_loss": 0.06271163374185562, "global_step": 258356, "epoch": 2902, "lr": 5.750499211453319e-07}
{"train_loss": 0.039042167365550995, "global_step": 258357, "epoch": 2902, "lr": 5.749622524134579e-07}
{"train_loss": 0.0949043482542038, "global_step": 258358, "epoch": 2902, "lr": 5.748745903261743e-07}
{"train_loss": 0.03057154268026352, "global_step": 258359, "epoch": 2902, "lr": 5.747869348834977e-07}
{"train_loss": 0.06448124349117279, "global_step": 258360, "epoch": 2902, "lr": 5.746992860854339e-07}
{"train_loss": 0.08702703565359116, "global_step": 258361, "epoch": 2902, "lr": 5.746116439320048e-07}
{"train_loss": 0.05248182266950607, "global_step": 258362, "epoch": 2902, "lr": 5.745240084232162e-07}
{"train_loss": 0.06832452863454819, "global_step": 258363, "epoch": 2902, "lr": 5.74436379559079e-07}
{"train_loss": 0.02063758485019207, "global_step": 258364, "epoch": 2902, "lr": 5.743487573396045e-07}
{"train_loss": 0.06425010412931442, "global_step": 258365, "epoch": 2902, "lr": 5.742611417648091e-07}
{"train_loss": 0.04850321217031961, "global_step": 258366, "epoch": 2902, "lr": 5.741735328347042e-07, "val_loss": 9.646918296813965}
{"train_loss": 0.05700720101594925, "global_step": 258367, "epoch": 2903, "lr": 5.740859305492952e-07}
{"train_loss": 0.04773377627134323, "global_step": 258368, "epoch": 2903, "lr": 5.739983349085987e-07}
{"train_loss": 0.022311948239803314, "global_step": 258369, "epoch": 2903, "lr": 5.739107459126258e-07}
{"train_loss": 0.0538274310529232, "global_step": 258370, "epoch": 2903, "lr": 5.738231635613878e-07}
{"train_loss": 0.03487350046634674, "global_step": 258371, "epoch": 2903, "lr": 5.737355878548956e-07}
{"train_loss": 0.05302540957927704, "global_step": 258372, "epoch": 2903, "lr": 5.736480187931659e-07}
{"train_loss": 0.05469880253076553, "global_step": 258373, "epoch": 2903, "lr": 5.735604563762043e-07}
{"train_loss": 0.08886191993951797, "global_step": 258374, "epoch": 2903, "lr": 5.734729006040274e-07}
{"train_loss": 0.08773044496774673, "global_step": 258375, "epoch": 2903, "lr": 5.733853514766463e-07}
{"train_loss": 0.09547430276870728, "global_step": 258376, "epoch": 2903, "lr": 5.732978089940666e-07}
{"train_loss": 0.028498830273747444, "global_step": 258377, "epoch": 2903, "lr": 5.73210273156305e-07}
{"train_loss": 0.043419867753982544, "global_step": 258378, "epoch": 2903, "lr": 5.73122743963378e-07}
{"train_loss": 0.02182888239622116, "global_step": 258379, "epoch": 2903, "lr": 5.730352214152857e-07}
{"train_loss": 0.03288497403264046, "global_step": 258380, "epoch": 2903, "lr": 5.729477055120503e-07}
{"train_loss": 0.029863009229302406, "global_step": 258381, "epoch": 2903, "lr": 5.728601962536772e-07}
{"train_loss": 0.030246693640947342, "global_step": 258382, "epoch": 2903, "lr": 5.727726936401834e-07}
{"train_loss": 0.04640958830714226, "global_step": 258383, "epoch": 2903, "lr": 5.726851976715741e-07}
{"train_loss": 0.030533449724316597, "global_step": 258384, "epoch": 2903, "lr": 5.725977083478662e-07}
{"train_loss": 0.04150140658020973, "global_step": 258385, "epoch": 2903, "lr": 5.725102256690761e-07}
{"train_loss": 0.053019750863313675, "global_step": 258386, "epoch": 2903, "lr": 5.724227496351986e-07}
{"train_loss": 0.048750996589660645, "global_step": 258387, "epoch": 2903, "lr": 5.723352802462667e-07}
{"train_loss": 0.07948437333106995, "global_step": 258388, "epoch": 2903, "lr": 5.72247817502275e-07}
{"train_loss": 0.027081269770860672, "global_step": 258389, "epoch": 2903, "lr": 5.721603614032456e-07}
{"train_loss": 0.05285117030143738, "global_step": 258390, "epoch": 2903, "lr": 5.720729119491842e-07}
{"train_loss": 0.037095461040735245, "global_step": 258391, "epoch": 2903, "lr": 5.719854691401072e-07}
{"train_loss": 0.036476653069257736, "global_step": 258392, "epoch": 2903, "lr": 5.718980329760204e-07}
{"train_loss": 0.052711550146341324, "global_step": 258393, "epoch": 2903, "lr": 5.718106034569459e-07}
{"train_loss": 0.043676022440195084, "global_step": 258394, "epoch": 2903, "lr": 5.717231805828838e-07}
{"train_loss": 0.0673040822148323, "global_step": 258395, "epoch": 2903, "lr": 5.716357643538562e-07}
{"train_loss": 0.0356876514852047, "global_step": 258396, "epoch": 2903, "lr": 5.71548354769863e-07}
{"train_loss": 0.04921352118253708, "global_step": 258397, "epoch": 2903, "lr": 5.714609518309322e-07}
{"train_loss": 0.023112783208489418, "global_step": 258398, "epoch": 2903, "lr": 5.713735555370581e-07}
{"train_loss": 0.02186167612671852, "global_step": 258399, "epoch": 2903, "lr": 5.712861658882628e-07}
{"train_loss": 0.062259741127491, "global_step": 258400, "epoch": 2903, "lr": 5.711987828845522e-07}
{"train_loss": 0.05259184539318085, "global_step": 258401, "epoch": 2903, "lr": 5.711114065259482e-07}
{"train_loss": 0.04293634369969368, "global_step": 258402, "epoch": 2903, "lr": 5.710240368124509e-07}
{"train_loss": 0.010033589787781239, "global_step": 258403, "epoch": 2903, "lr": 5.709366737440769e-07}
{"train_loss": 0.03351898491382599, "global_step": 258404, "epoch": 2903, "lr": 5.708493173208373e-07}
{"train_loss": 0.035679835826158524, "global_step": 258405, "epoch": 2903, "lr": 5.70761967542749e-07}
{"train_loss": 0.04494618996977806, "global_step": 258406, "epoch": 2903, "lr": 5.706746244098172e-07}
{"train_loss": 0.02301417477428913, "global_step": 258407, "epoch": 2903, "lr": 5.705872879220531e-07}
{"train_loss": 0.040138568729162216, "global_step": 258408, "epoch": 2903, "lr": 5.704999580794734e-07}
{"train_loss": 0.029171785339713097, "global_step": 258409, "epoch": 2903, "lr": 5.704126348820893e-07}
{"train_loss": 0.05995221063494682, "global_step": 258410, "epoch": 2903, "lr": 5.703253183299062e-07}
{"train_loss": 0.057967253029346466, "global_step": 258411, "epoch": 2903, "lr": 5.702380084229464e-07}
{"train_loss": 0.03652271255850792, "global_step": 258412, "epoch": 2903, "lr": 5.701507051612098e-07}
{"train_loss": 0.03480834141373634, "global_step": 258413, "epoch": 2903, "lr": 5.700634085447187e-07}
{"train_loss": 0.07423408329486847, "global_step": 258414, "epoch": 2903, "lr": 5.699761185734787e-07}
{"train_loss": 0.04783244803547859, "global_step": 258415, "epoch": 2903, "lr": 5.698888352475007e-07}
{"train_loss": 0.035937827080488205, "global_step": 258416, "epoch": 2903, "lr": 5.698015585668015e-07}
{"train_loss": 0.04123155400156975, "global_step": 258417, "epoch": 2903, "lr": 5.697142885313866e-07}
{"train_loss": 0.05140529200434685, "global_step": 258418, "epoch": 2903, "lr": 5.696270251412727e-07}
{"train_loss": 0.06231037527322769, "global_step": 258419, "epoch": 2903, "lr": 5.69539768396471e-07}
{"train_loss": 0.051771264523267746, "global_step": 258420, "epoch": 2903, "lr": 5.694525182969923e-07}
{"train_loss": 0.05937749519944191, "global_step": 258421, "epoch": 2903, "lr": 5.69365274842848e-07}
{"train_loss": 0.026117201894521713, "global_step": 258422, "epoch": 2903, "lr": 5.69278038034049e-07}
{"train_loss": 0.027956295758485794, "global_step": 258423, "epoch": 2903, "lr": 5.691908078706121e-07}
{"train_loss": 0.024810653179883957, "global_step": 258424, "epoch": 2903, "lr": 5.691035843525372e-07}
{"train_loss": 0.05632130056619644, "global_step": 258425, "epoch": 2903, "lr": 5.69016367479852e-07}
{"train_loss": 0.045214492827653885, "global_step": 258426, "epoch": 2903, "lr": 5.689291572525568e-07}
{"train_loss": 0.07734943181276321, "global_step": 258427, "epoch": 2903, "lr": 5.68841953670668e-07}
{"train_loss": 0.07030825316905975, "global_step": 258428, "epoch": 2903, "lr": 5.687547567341911e-07}
{"train_loss": 0.06175121292471886, "global_step": 258429, "epoch": 2903, "lr": 5.686675664431484e-07}
{"train_loss": 0.025240564718842506, "global_step": 258430, "epoch": 2903, "lr": 5.6858038279754e-07}
{"train_loss": 0.03248300030827522, "global_step": 258431, "epoch": 2903, "lr": 5.68493205797388e-07}
{"train_loss": 0.046997323632240295, "global_step": 258432, "epoch": 2903, "lr": 5.684060354426979e-07}
{"train_loss": 0.04123377799987793, "global_step": 258433, "epoch": 2903, "lr": 5.683188717334865e-07}
{"train_loss": 0.02947118692100048, "global_step": 258434, "epoch": 2903, "lr": 5.682317146697536e-07}
{"train_loss": 0.041814882308244705, "global_step": 258435, "epoch": 2903, "lr": 5.681445642515271e-07}
{"train_loss": 0.016697866842150688, "global_step": 258436, "epoch": 2903, "lr": 5.680574204788069e-07}
{"train_loss": 0.04323841631412506, "global_step": 258437, "epoch": 2903, "lr": 5.679702833516154e-07}
{"train_loss": 0.011095323599874973, "global_step": 258438, "epoch": 2903, "lr": 5.678831528699525e-07}
{"train_loss": 0.05474419891834259, "global_step": 258439, "epoch": 2903, "lr": 5.677960290338347e-07}
{"train_loss": 0.049670156091451645, "global_step": 258440, "epoch": 2903, "lr": 5.677089118432732e-07}
{"train_loss": 0.045625753700733185, "global_step": 258441, "epoch": 2903, "lr": 5.676218012982848e-07}
{"train_loss": 0.02608015574514866, "global_step": 258442, "epoch": 2903, "lr": 5.675346973988748e-07}
{"train_loss": 0.06787864863872528, "global_step": 258443, "epoch": 2903, "lr": 5.674476001450546e-07}
{"train_loss": 0.014742547646164894, "global_step": 258444, "epoch": 2903, "lr": 5.673605095368462e-07}
{"train_loss": 0.041912950575351715, "global_step": 258445, "epoch": 2903, "lr": 5.672734255742496e-07}
{"train_loss": 0.034067302942276, "global_step": 258446, "epoch": 2903, "lr": 5.671863482572815e-07}
{"train_loss": 0.029102563858032227, "global_step": 258447, "epoch": 2903, "lr": 5.670992775859474e-07}
{"train_loss": 0.07859782874584198, "global_step": 258448, "epoch": 2903, "lr": 5.670122135602696e-07}
{"train_loss": 0.028648575767874718, "global_step": 258449, "epoch": 2903, "lr": 5.669251561802535e-07}
{"train_loss": 0.09350705146789551, "global_step": 258450, "epoch": 2903, "lr": 5.668381054459104e-07}
{"train_loss": 0.047859076410532, "global_step": 258451, "epoch": 2903, "lr": 5.667510613572568e-07}
{"train_loss": 0.03417011722922325, "global_step": 258452, "epoch": 2903, "lr": 5.666640239142984e-07}
{"train_loss": 0.026716342195868492, "global_step": 258453, "epoch": 2903, "lr": 5.665769931170517e-07}
{"train_loss": 0.015327059663832188, "global_step": 258454, "epoch": 2903, "lr": 5.664899689655223e-07}
{"train_loss": 0.044680594696841214, "global_step": 258455, "epoch": 2903, "lr": 5.664029514597269e-07, "val_loss": 9.633971214294434}
{"train_loss": 0.05392862483859062, "global_step": 258456, "epoch": 2904, "lr": 5.663159405996821e-07}
{"train_loss": 0.04519367218017578, "global_step": 258457, "epoch": 2904, "lr": 5.662289363853823e-07}
{"train_loss": 0.023219970986247063, "global_step": 258458, "epoch": 2904, "lr": 5.661419388168609e-07}
{"train_loss": 0.04191925749182701, "global_step": 258459, "epoch": 2904, "lr": 5.660549478941179e-07}
{"train_loss": 0.03675980493426323, "global_step": 258460, "epoch": 2904, "lr": 5.659679636171589e-07}
{"train_loss": 0.031099984422326088, "global_step": 258461, "epoch": 2904, "lr": 5.658809859860115e-07}
{"train_loss": 0.0495627298951149, "global_step": 258462, "epoch": 2904, "lr": 5.657940150006758e-07}
{"train_loss": 0.07698886841535568, "global_step": 258463, "epoch": 2904, "lr": 5.657070506611684e-07}
{"train_loss": 0.06578657031059265, "global_step": 258464, "epoch": 2904, "lr": 5.656200929674949e-07}
{"train_loss": 0.058034684509038925, "global_step": 258465, "epoch": 2904, "lr": 5.655331419196719e-07}
{"train_loss": 0.038823358714580536, "global_step": 258466, "epoch": 2904, "lr": 5.654461975177106e-07}
{"train_loss": 0.02845044620335102, "global_step": 258467, "epoch": 2904, "lr": 5.653592597616275e-07}
{"train_loss": 0.05316690355539322, "global_step": 258468, "epoch": 2904, "lr": 5.652723286514228e-07}
{"train_loss": 0.04825722426176071, "global_step": 258469, "epoch": 2904, "lr": 5.651854041871185e-07}
{"train_loss": 0.02864067256450653, "global_step": 258470, "epoch": 2904, "lr": 5.650984863687203e-07}
{"train_loss": 0.033474769443273544, "global_step": 258471, "epoch": 2904, "lr": 5.650115751962448e-07}
{"train_loss": 0.05020200088620186, "global_step": 258472, "epoch": 2904, "lr": 5.649246706696976e-07}
{"train_loss": 0.020480848848819733, "global_step": 258473, "epoch": 2904, "lr": 5.648377727890952e-07}
{"train_loss": 0.0464850589632988, "global_step": 258474, "epoch": 2904, "lr": 5.647508815544489e-07}
{"train_loss": 0.050814442336559296, "global_step": 258475, "epoch": 2904, "lr": 5.646639969657696e-07}
{"train_loss": 0.02303384616971016, "global_step": 258476, "epoch": 2904, "lr": 5.645771190230631e-07}
{"train_loss": 0.06024647876620293, "global_step": 258477, "epoch": 2904, "lr": 5.644902477263514e-07}
{"train_loss": 0.06687822192907333, "global_step": 258478, "epoch": 2904, "lr": 5.644033830756401e-07}
{"train_loss": 0.11429563909769058, "global_step": 258479, "epoch": 2904, "lr": 5.643165250709403e-07}
{"train_loss": 0.029684696346521378, "global_step": 258480, "epoch": 2904, "lr": 5.642296737122743e-07}
{"train_loss": 0.08020907640457153, "global_step": 258481, "epoch": 2904, "lr": 5.641428289996365e-07}
{"train_loss": 0.07230803370475769, "global_step": 258482, "epoch": 2904, "lr": 5.64055990933049e-07}
{"train_loss": 0.04660242050886154, "global_step": 258483, "epoch": 2904, "lr": 5.63969159512523e-07}
{"train_loss": 0.024532344192266464, "global_step": 258484, "epoch": 2904, "lr": 5.638823347380696e-07}
{"train_loss": 0.0479089692234993, "global_step": 258485, "epoch": 2904, "lr": 5.637955166096998e-07}
{"train_loss": 0.04844655469059944, "global_step": 258486, "epoch": 2904, "lr": 5.637087051274248e-07}
{"train_loss": 0.018300529569387436, "global_step": 258487, "epoch": 2904, "lr": 5.636219002912502e-07}
{"train_loss": 0.008423334918916225, "global_step": 258488, "epoch": 2904, "lr": 5.635351021012037e-07}
{"train_loss": 0.07946212589740753, "global_step": 258489, "epoch": 2904, "lr": 5.634483105572797e-07}
{"train_loss": 0.07845081388950348, "global_step": 258490, "epoch": 2904, "lr": 5.63361525659506e-07}
{"train_loss": 0.046769093722105026, "global_step": 258491, "epoch": 2904, "lr": 5.63274747407877e-07}
{"train_loss": 0.029758142307400703, "global_step": 258492, "epoch": 2904, "lr": 5.631879758024205e-07}
{"train_loss": 0.04918335750699043, "global_step": 258493, "epoch": 2904, "lr": 5.631012108431366e-07}
{"train_loss": 0.03884539008140564, "global_step": 258494, "epoch": 2904, "lr": 5.630144525300418e-07}
{"train_loss": 0.026634298264980316, "global_step": 258495, "epoch": 2904, "lr": 5.629277008631473e-07}
{"train_loss": 0.06647007167339325, "global_step": 258496, "epoch": 2904, "lr": 5.628409558424641e-07}
{"train_loss": 0.029777107760310173, "global_step": 258497, "epoch": 2904, "lr": 5.627542174680034e-07}
{"train_loss": 0.051665350794792175, "global_step": 258498, "epoch": 2904, "lr": 5.626674857397818e-07}
{"train_loss": 0.03432990983128548, "global_step": 258499, "epoch": 2904, "lr": 5.625807606578048e-07}
{"train_loss": 0.04399758577346802, "global_step": 258500, "epoch": 2904, "lr": 5.624940422220837e-07}
{"train_loss": 0.05373959243297577, "global_step": 258501, "epoch": 2904, "lr": 5.624073304326405e-07}
{"train_loss": 0.02359694242477417, "global_step": 258502, "epoch": 2904, "lr": 5.623206252894698e-07}
{"train_loss": 0.04491327330470085, "global_step": 258503, "epoch": 2904, "lr": 5.622339267925991e-07}
{"train_loss": 0.018073169514536858, "global_step": 258504, "epoch": 2904, "lr": 5.621472349420287e-07}
{"train_loss": 0.037734996527433395, "global_step": 258505, "epoch": 2904, "lr": 5.620605497377751e-07}
{"train_loss": 0.02581832744181156, "global_step": 258506, "epoch": 2904, "lr": 5.61973871179855e-07}
{"train_loss": 0.014166186563670635, "global_step": 258507, "epoch": 2904, "lr": 5.618871992682684e-07}
{"train_loss": 0.0855894386768341, "global_step": 258508, "epoch": 2904, "lr": 5.618005340030375e-07}
{"train_loss": 0.03340167924761772, "global_step": 258509, "epoch": 2904, "lr": 5.617138753841733e-07}
{"train_loss": 0.016804199665784836, "global_step": 258510, "epoch": 2904, "lr": 5.616272234116759e-07}
{"train_loss": 0.054685547947883606, "global_step": 258511, "epoch": 2904, "lr": 5.615405780855732e-07}
{"train_loss": 0.02946772612631321, "global_step": 258512, "epoch": 2904, "lr": 5.614539394058594e-07}
{"train_loss": 0.049615390598773956, "global_step": 258513, "epoch": 2904, "lr": 5.613673073725623e-07}
{"train_loss": 0.06682336330413818, "global_step": 258514, "epoch": 2904, "lr": 5.612806819856875e-07}
{"train_loss": 0.04263162985444069, "global_step": 258515, "epoch": 2904, "lr": 5.611940632452406e-07}
{"train_loss": 0.035937655717134476, "global_step": 258516, "epoch": 2904, "lr": 5.611074511512437e-07}
{"train_loss": 0.034487269818782806, "global_step": 258517, "epoch": 2904, "lr": 5.610208457037025e-07}
{"train_loss": 0.06771951913833618, "global_step": 258518, "epoch": 2904, "lr": 5.60934246902628e-07}
{"train_loss": 0.024499190971255302, "global_step": 258519, "epoch": 2904, "lr": 5.608476547480313e-07}
{"train_loss": 0.06723298132419586, "global_step": 258520, "epoch": 2904, "lr": 5.607610692399346e-07}
{"train_loss": 0.03382791206240654, "global_step": 258521, "epoch": 2904, "lr": 5.606744903783323e-07}
{"train_loss": 0.011293736286461353, "global_step": 258522, "epoch": 2904, "lr": 5.605879181632467e-07}
{"train_loss": 0.05146000161767006, "global_step": 258523, "epoch": 2904, "lr": 5.605013525946889e-07}
{"train_loss": 0.041056275367736816, "global_step": 258524, "epoch": 2904, "lr": 5.6041479367267e-07}
{"train_loss": 0.02353532239794731, "global_step": 258525, "epoch": 2904, "lr": 5.603282413971956e-07}
{"train_loss": 0.0235764030367136, "global_step": 258526, "epoch": 2904, "lr": 5.602416957682877e-07}
{"train_loss": 0.04842134565114975, "global_step": 258527, "epoch": 2904, "lr": 5.60155156785952e-07}
{"train_loss": 0.08079170435667038, "global_step": 258528, "epoch": 2904, "lr": 5.600686244501995e-07}
{"train_loss": 0.08738519251346588, "global_step": 258529, "epoch": 2904, "lr": 5.599820987610415e-07}
{"train_loss": 0.026254776865243912, "global_step": 258530, "epoch": 2904, "lr": 5.598955797184946e-07}
{"train_loss": 0.0329616479575634, "global_step": 258531, "epoch": 2904, "lr": 5.598090673225642e-07}
{"train_loss": 0.038237933069467545, "global_step": 258532, "epoch": 2904, "lr": 5.597225615732727e-07}
{"train_loss": 0.061955638229846954, "global_step": 258533, "epoch": 2904, "lr": 5.596360624706143e-07}
{"train_loss": 0.03275115415453911, "global_step": 258534, "epoch": 2904, "lr": 5.59549570014617e-07}
{"train_loss": 0.037310753017663956, "global_step": 258535, "epoch": 2904, "lr": 5.594630842052806e-07}
{"train_loss": 0.04388761520385742, "global_step": 258536, "epoch": 2904, "lr": 5.593766050426219e-07}
{"train_loss": 0.03027850016951561, "global_step": 258537, "epoch": 2904, "lr": 5.592901325266575e-07}
{"train_loss": 0.032221339643001556, "global_step": 258538, "epoch": 2904, "lr": 5.592036666573874e-07}
{"train_loss": 0.07245568186044693, "global_step": 258539, "epoch": 2904, "lr": 5.591172074348394e-07}
{"train_loss": 0.029674869030714035, "global_step": 258540, "epoch": 2904, "lr": 5.590307548590079e-07}
{"train_loss": 0.039936892688274384, "global_step": 258541, "epoch": 2904, "lr": 5.589443089299151e-07}
{"train_loss": 0.017141757532954216, "global_step": 258542, "epoch": 2904, "lr": 5.58857869647561e-07}
{"train_loss": 0.07142475992441177, "global_step": 258543, "epoch": 2904, "lr": 5.587714370119789e-07}
{"train_loss": 0.044542439669119506, "global_step": 258544, "epoch": 2904, "lr": 5.586850110231579e-07, "val_loss": 9.634267807006836}
{"train_loss": 0.05445433035492897, "global_step": 258545, "epoch": 2905, "lr": 5.585985916811254e-07}
{"train_loss": 0.04739357531070709, "global_step": 258546, "epoch": 2905, "lr": 5.585121789858816e-07}
{"train_loss": 0.02674172632396221, "global_step": 258547, "epoch": 2905, "lr": 5.584257729374487e-07}
{"train_loss": 0.05857940390706062, "global_step": 258548, "epoch": 2905, "lr": 5.583393735358267e-07}
{"train_loss": 0.04823143035173416, "global_step": 258549, "epoch": 2905, "lr": 5.582529807810377e-07}
{"train_loss": 0.023518986999988556, "global_step": 258550, "epoch": 2905, "lr": 5.581665946730874e-07}
{"train_loss": 0.03232021629810333, "global_step": 258551, "epoch": 2905, "lr": 5.580802152119868e-07}
{"train_loss": 0.06697660684585571, "global_step": 258552, "epoch": 2905, "lr": 5.579938423977527e-07}
{"train_loss": 0.06964622437953949, "global_step": 258553, "epoch": 2905, "lr": 5.579074762303904e-07}
{"train_loss": 0.05191374570131302, "global_step": 258554, "epoch": 2905, "lr": 5.578211167099168e-07}
{"train_loss": 0.062041301280260086, "global_step": 258555, "epoch": 2905, "lr": 5.577347638363428e-07}
{"train_loss": 0.0453956313431263, "global_step": 258556, "epoch": 2905, "lr": 5.576484176096796e-07}
{"train_loss": 0.07550477981567383, "global_step": 258557, "epoch": 2905, "lr": 5.575620780299329e-07}
{"train_loss": 0.012638821266591549, "global_step": 258558, "epoch": 2905, "lr": 5.574757450971247e-07}
{"train_loss": 0.027804438024759293, "global_step": 258559, "epoch": 2905, "lr": 5.573894188112605e-07}
{"train_loss": 0.0525253564119339, "global_step": 258560, "epoch": 2905, "lr": 5.573030991723516e-07}
{"train_loss": 0.05935629829764366, "global_step": 258561, "epoch": 2905, "lr": 5.57216786180409e-07}
{"train_loss": 0.03436652943491936, "global_step": 258562, "epoch": 2905, "lr": 5.571304798354493e-07}
{"train_loss": 0.05651545152068138, "global_step": 258563, "epoch": 2905, "lr": 5.570441801374782e-07}
{"train_loss": 0.07460915297269821, "global_step": 258564, "epoch": 2905, "lr": 5.569578870865122e-07}
{"train_loss": 0.07445866614580154, "global_step": 258565, "epoch": 2905, "lr": 5.56871600682557e-07}
{"train_loss": 0.07283511757850647, "global_step": 258566, "epoch": 2905, "lr": 5.567853209256347e-07}
{"train_loss": 0.08311165124177933, "global_step": 258567, "epoch": 2905, "lr": 5.566990478157397e-07}
{"train_loss": 0.026327531784772873, "global_step": 258568, "epoch": 2905, "lr": 5.566127813528998e-07}
{"train_loss": 0.08480007201433182, "global_step": 258569, "epoch": 2905, "lr": 5.565265215371207e-07}
{"train_loss": 0.06784907728433609, "global_step": 258570, "epoch": 2905, "lr": 5.564402683684189e-07}
{"train_loss": 0.07178757339715958, "global_step": 258571, "epoch": 2905, "lr": 5.563540218467945e-07}
{"train_loss": 0.07885725796222687, "global_step": 258572, "epoch": 2905, "lr": 5.562677819722639e-07}
{"train_loss": 0.06275284290313721, "global_step": 258573, "epoch": 2905, "lr": 5.561815487448441e-07}
{"train_loss": 0.02822710946202278, "global_step": 258574, "epoch": 2905, "lr": 5.560953221645459e-07}
{"train_loss": 0.027711303904652596, "global_step": 258575, "epoch": 2905, "lr": 5.560091022313751e-07}
{"train_loss": 0.06677398085594177, "global_step": 258576, "epoch": 2905, "lr": 5.559228889453427e-07}
{"train_loss": 0.030735723674297333, "global_step": 258577, "epoch": 2905, "lr": 5.558366823064709e-07}
{"train_loss": 0.06087883189320564, "global_step": 258578, "epoch": 2905, "lr": 5.557504823147597e-07}
{"train_loss": 0.02820594795048237, "global_step": 258579, "epoch": 2905, "lr": 5.556642889702257e-07}
{"train_loss": 0.036102600395679474, "global_step": 258580, "epoch": 2905, "lr": 5.555781022728801e-07}
{"train_loss": 0.02985089085996151, "global_step": 258581, "epoch": 2905, "lr": 5.554919222227339e-07}
{"train_loss": 0.03219759836792946, "global_step": 258582, "epoch": 2905, "lr": 5.554057488197983e-07}
{"train_loss": 0.029916781932115555, "global_step": 258583, "epoch": 2905, "lr": 5.5531958206409e-07}
{"train_loss": 0.0269713643938303, "global_step": 258584, "epoch": 2905, "lr": 5.552334219556088e-07}
{"train_loss": 0.03163112699985504, "global_step": 258585, "epoch": 2905, "lr": 5.551472684943826e-07}
{"train_loss": 0.09424395114183426, "global_step": 258586, "epoch": 2905, "lr": 5.550611216804058e-07}
{"train_loss": 0.031230637803673744, "global_step": 258587, "epoch": 2905, "lr": 5.549749815137007e-07}
{"train_loss": 0.011196726933121681, "global_step": 258588, "epoch": 2905, "lr": 5.548888479942838e-07}
{"train_loss": 0.03704630210995674, "global_step": 258589, "epoch": 2905, "lr": 5.548027211221496e-07}
{"train_loss": 0.040971480309963226, "global_step": 258590, "epoch": 2905, "lr": 5.547166008973259e-07}
{"train_loss": 0.0459919236600399, "global_step": 258591, "epoch": 2905, "lr": 5.546304873198128e-07}
{"train_loss": 0.024436412379145622, "global_step": 258592, "epoch": 2905, "lr": 5.545443803896322e-07}
{"train_loss": 0.030651679262518883, "global_step": 258593, "epoch": 2905, "lr": 5.544582801067899e-07}
{"train_loss": 0.022721033543348312, "global_step": 258594, "epoch": 2905, "lr": 5.543721864712969e-07}
{"train_loss": 0.0423320047557354, "global_step": 258595, "epoch": 2905, "lr": 5.542860994831644e-07}
{"train_loss": 0.04968862980604172, "global_step": 258596, "epoch": 2905, "lr": 5.54200019142409e-07}
{"train_loss": 0.047766316682100296, "global_step": 258597, "epoch": 2905, "lr": 5.541139454490307e-07}
{"train_loss": 0.03406576067209244, "global_step": 258598, "epoch": 2905, "lr": 5.540278784030573e-07}
{"train_loss": 0.019646992906928062, "global_step": 258599, "epoch": 2905, "lr": 5.539418180044886e-07}
{"train_loss": 0.025164786726236343, "global_step": 258600, "epoch": 2905, "lr": 5.538557642533415e-07}
{"train_loss": 0.03079015016555786, "global_step": 258601, "epoch": 2905, "lr": 5.53769717149627e-07}
{"train_loss": 0.07694679498672485, "global_step": 258602, "epoch": 2905, "lr": 5.536836766933561e-07}
{"train_loss": 0.056582920253276825, "global_step": 258603, "epoch": 2905, "lr": 5.535976428845346e-07}
{"train_loss": 0.017411990091204643, "global_step": 258604, "epoch": 2905, "lr": 5.53511615723179e-07}
{"train_loss": 0.07797292619943619, "global_step": 258605, "epoch": 2905, "lr": 5.534255952093059e-07}
{"train_loss": 0.025382276624441147, "global_step": 258606, "epoch": 2905, "lr": 5.533395813429209e-07}
{"train_loss": 0.05153222009539604, "global_step": 258607, "epoch": 2905, "lr": 5.532535741240353e-07}
{"train_loss": 0.025530554354190826, "global_step": 258608, "epoch": 2905, "lr": 5.531675735526598e-07}
{"train_loss": 0.017828557640314102, "global_step": 258609, "epoch": 2905, "lr": 5.530815796288169e-07}
{"train_loss": 0.01709905080497265, "global_step": 258610, "epoch": 2905, "lr": 5.52995592352501e-07}
{"train_loss": 0.03200087323784828, "global_step": 258611, "epoch": 2905, "lr": 5.529096117237342e-07}
{"train_loss": 0.02997528947889805, "global_step": 258612, "epoch": 2905, "lr": 5.528236377425277e-07}
{"train_loss": 0.04740934818983078, "global_step": 258613, "epoch": 2905, "lr": 5.527376704088926e-07}
{"train_loss": 0.05619030445814133, "global_step": 258614, "epoch": 2905, "lr": 5.526517097228401e-07}
{"train_loss": 0.046277377754449844, "global_step": 258615, "epoch": 2905, "lr": 5.525657556843755e-07}
{"train_loss": 0.043741751462221146, "global_step": 258616, "epoch": 2905, "lr": 5.524798082935212e-07}
{"train_loss": 0.04949084296822548, "global_step": 258617, "epoch": 2905, "lr": 5.523938675502772e-07}
{"train_loss": 0.02343200333416462, "global_step": 258618, "epoch": 2905, "lr": 5.523079334546654e-07}
{"train_loss": 0.029341891407966614, "global_step": 258619, "epoch": 2905, "lr": 5.522220060066919e-07}
{"train_loss": 0.05325177684426308, "global_step": 258620, "epoch": 2905, "lr": 5.521360852063673e-07}
{"train_loss": 0.025561990216374397, "global_step": 258621, "epoch": 2905, "lr": 5.52050171053714e-07}
{"train_loss": 0.05488955229520798, "global_step": 258622, "epoch": 2905, "lr": 5.519642635487265e-07}
{"train_loss": 0.03577972576022148, "global_step": 258623, "epoch": 2905, "lr": 5.518783626914215e-07}
{"train_loss": 0.015453563071787357, "global_step": 258624, "epoch": 2905, "lr": 5.51792468481821e-07}
{"train_loss": 0.04805291071534157, "global_step": 258625, "epoch": 2905, "lr": 5.517065809199251e-07}
{"train_loss": 0.015492637641727924, "global_step": 258626, "epoch": 2905, "lr": 5.516207000057561e-07}
{"train_loss": 0.05861799791455269, "global_step": 258627, "epoch": 2905, "lr": 5.515348257393138e-07}
{"train_loss": 0.028405262157320976, "global_step": 258628, "epoch": 2905, "lr": 5.51448958120615e-07}
{"train_loss": 0.03936558961868286, "global_step": 258629, "epoch": 2905, "lr": 5.513630971496708e-07}
{"train_loss": 0.029445603489875793, "global_step": 258630, "epoch": 2905, "lr": 5.512772428264978e-07}
{"train_loss": 0.035780128091573715, "global_step": 258631, "epoch": 2905, "lr": 5.511913951510961e-07}
{"train_loss": 0.012447299435734749, "global_step": 258632, "epoch": 2905, "lr": 5.511055541234877e-07}
{"train_loss": 0.04294386509219917, "global_step": 258633, "epoch": 2905, "lr": 5.510197197436784e-07, "val_loss": 9.554610252380371, "train_action_mse_error": 24.68592071533203}
{"train_loss": 0.042692188173532486, "global_step": 258634, "epoch": 2906, "lr": 5.509338920116846e-07}
{"train_loss": 0.07449128478765488, "global_step": 258635, "epoch": 2906, "lr": 5.508480709275122e-07}
{"train_loss": 0.041197046637535095, "global_step": 258636, "epoch": 2906, "lr": 5.507622564911829e-07}
{"train_loss": 0.08267910033464432, "global_step": 258637, "epoch": 2906, "lr": 5.506764487026916e-07}
{"train_loss": 0.020091421902179718, "global_step": 258638, "epoch": 2906, "lr": 5.505906475620659e-07}
{"train_loss": 0.05493373051285744, "global_step": 258639, "epoch": 2906, "lr": 5.505048530693058e-07}
{"train_loss": 0.027139799669384956, "global_step": 258640, "epoch": 2906, "lr": 5.504190652244279e-07}
{"train_loss": 0.04181588441133499, "global_step": 258641, "epoch": 2906, "lr": 5.503332840274433e-07}
{"train_loss": 0.035903722047805786, "global_step": 258642, "epoch": 2906, "lr": 5.502475094783687e-07}
{"train_loss": 0.08839765936136246, "global_step": 258643, "epoch": 2906, "lr": 5.501617415772042e-07}
{"train_loss": 0.0547313317656517, "global_step": 258644, "epoch": 2906, "lr": 5.500759803239719e-07}
{"train_loss": 0.06379516422748566, "global_step": 258645, "epoch": 2906, "lr": 5.499902257186773e-07}
{"train_loss": 0.04272596910595894, "global_step": 258646, "epoch": 2906, "lr": 5.499044777613371e-07}
{"train_loss": 0.06484370678663254, "global_step": 258647, "epoch": 2906, "lr": 5.498187364519569e-07}
{"train_loss": 0.038990799337625504, "global_step": 258648, "epoch": 2906, "lr": 5.497330017905533e-07}
{"train_loss": 0.01936567947268486, "global_step": 258649, "epoch": 2906, "lr": 5.496472737771318e-07}
{"train_loss": 0.031079959124326706, "global_step": 258650, "epoch": 2906, "lr": 5.495615524117092e-07}
{"train_loss": 0.03215651214122772, "global_step": 258651, "epoch": 2906, "lr": 5.49475837694291e-07}
{"train_loss": 0.025317925959825516, "global_step": 258652, "epoch": 2906, "lr": 5.493901296248994e-07}
{"train_loss": 0.07068607956171036, "global_step": 258653, "epoch": 2906, "lr": 5.493044282035397e-07}
{"train_loss": 0.039980217814445496, "global_step": 258654, "epoch": 2906, "lr": 5.492187334302179e-07}
{"train_loss": 0.031724248081445694, "global_step": 258655, "epoch": 2906, "lr": 5.491330453049559e-07}
{"train_loss": 0.0669282078742981, "global_step": 258656, "epoch": 2906, "lr": 5.490473638277594e-07}
{"train_loss": 0.0381280891597271, "global_step": 258657, "epoch": 2906, "lr": 5.489616889986393e-07}
{"train_loss": 0.07764435559511185, "global_step": 258658, "epoch": 2906, "lr": 5.488760208176069e-07}
{"train_loss": 0.061534564942121506, "global_step": 258659, "epoch": 2906, "lr": 5.487903592846733e-07}
{"train_loss": 0.0361669585108757, "global_step": 258660, "epoch": 2906, "lr": 5.487047043998606e-07}
{"train_loss": 0.050586115568876266, "global_step": 258661, "epoch": 2906, "lr": 5.486190561631632e-07}
{"train_loss": 0.04058801010251045, "global_step": 258662, "epoch": 2906, "lr": 5.48533414574609e-07}
{"train_loss": 0.01952829211950302, "global_step": 258663, "epoch": 2906, "lr": 5.48447779634198e-07}
{"train_loss": 0.04959043860435486, "global_step": 258664, "epoch": 2906, "lr": 5.483621513419468e-07}
{"train_loss": 0.02859880030155182, "global_step": 258665, "epoch": 2906, "lr": 5.482765296978609e-07}
{"train_loss": 0.03434986621141434, "global_step": 258666, "epoch": 2906, "lr": 5.481909147019626e-07}
{"train_loss": 0.055516138672828674, "global_step": 258667, "epoch": 2906, "lr": 5.481053063542518e-07}
{"train_loss": 0.05539616942405701, "global_step": 258668, "epoch": 2906, "lr": 5.480197046547508e-07}
{"train_loss": 0.04497089982032776, "global_step": 258669, "epoch": 2906, "lr": 5.479341096034596e-07}
{"train_loss": 0.051309458911418915, "global_step": 258670, "epoch": 2906, "lr": 5.478485212004003e-07}
{"train_loss": 0.05545078590512276, "global_step": 258671, "epoch": 2906, "lr": 5.477629394455785e-07}
{"train_loss": 0.05252443253993988, "global_step": 258672, "epoch": 2906, "lr": 5.47677364339011e-07}
{"train_loss": 0.03799585998058319, "global_step": 258673, "epoch": 2906, "lr": 5.475917958807031e-07}
{"train_loss": 0.05448102205991745, "global_step": 258674, "epoch": 2906, "lr": 5.475062340706661e-07}
{"train_loss": 0.025739900767803192, "global_step": 258675, "epoch": 2906, "lr": 5.474206789089164e-07}
{"train_loss": 0.06634791940450668, "global_step": 258676, "epoch": 2906, "lr": 5.473351303954655e-07}
{"train_loss": 0.02305496111512184, "global_step": 258677, "epoch": 2906, "lr": 5.472495885303186e-07}
{"train_loss": 0.024132899940013885, "global_step": 258678, "epoch": 2906, "lr": 5.471640533134925e-07}
{"train_loss": 0.03950431942939758, "global_step": 258679, "epoch": 2906, "lr": 5.470785247449983e-07}
{"train_loss": 0.054546598345041275, "global_step": 258680, "epoch": 2906, "lr": 5.469930028248416e-07}
{"train_loss": 0.04122805967926979, "global_step": 258681, "epoch": 2906, "lr": 5.4690748755305e-07}
{"train_loss": 0.05077238008379936, "global_step": 258682, "epoch": 2906, "lr": 5.468219789296125e-07}
{"train_loss": 0.049557413905858994, "global_step": 258683, "epoch": 2906, "lr": 5.467364769545625e-07}
{"train_loss": 0.02657741867005825, "global_step": 258684, "epoch": 2906, "lr": 5.466509816278886e-07}
{"train_loss": 0.010236025787889957, "global_step": 258685, "epoch": 2906, "lr": 5.465654929496245e-07}
{"train_loss": 0.062093961983919144, "global_step": 258686, "epoch": 2906, "lr": 5.464800109197699e-07}
{"train_loss": 0.061524081975221634, "global_step": 258687, "epoch": 2906, "lr": 5.46394535538336e-07}
{"train_loss": 0.07585303485393524, "global_step": 258688, "epoch": 2906, "lr": 5.463090668053395e-07}
{"train_loss": 0.06589803844690323, "global_step": 258689, "epoch": 2906, "lr": 5.462236047207858e-07}
{"train_loss": 0.04077742248773575, "global_step": 258690, "epoch": 2906, "lr": 5.461381492846974e-07}
{"train_loss": 0.055110059678554535, "global_step": 258691, "epoch": 2906, "lr": 5.460527004970739e-07}
{"train_loss": 0.09834473580121994, "global_step": 258692, "epoch": 2906, "lr": 5.459672583579268e-07}
{"train_loss": 0.030037691816687584, "global_step": 258693, "epoch": 2906, "lr": 5.458818228672724e-07}
{"train_loss": 0.04446268081665039, "global_step": 258694, "epoch": 2906, "lr": 5.45796394025122e-07}
{"train_loss": 0.03447199985384941, "global_step": 258695, "epoch": 2906, "lr": 5.457109718314868e-07}
{"train_loss": 0.021712535992264748, "global_step": 258696, "epoch": 2906, "lr": 5.456255562863777e-07}
{"train_loss": 0.07003330439329147, "global_step": 258697, "epoch": 2906, "lr": 5.455401473898059e-07}
{"train_loss": 0.04042663425207138, "global_step": 258698, "epoch": 2906, "lr": 5.45454745141788e-07}
{"train_loss": 0.04384647309780121, "global_step": 258699, "epoch": 2906, "lr": 5.45369349542324e-07}
{"train_loss": 0.05519355833530426, "global_step": 258700, "epoch": 2906, "lr": 5.452839605914417e-07}
{"train_loss": 0.06144682317972183, "global_step": 258701, "epoch": 2906, "lr": 5.451985782891356e-07}
{"train_loss": 0.032803554087877274, "global_step": 258702, "epoch": 2906, "lr": 5.451132026354277e-07}
{"train_loss": 0.03661046177148819, "global_step": 258703, "epoch": 2906, "lr": 5.450278336303238e-07}
{"train_loss": 0.0745648443698883, "global_step": 258704, "epoch": 2906, "lr": 5.449424712738404e-07}
{"train_loss": 0.020380640402436256, "global_step": 258705, "epoch": 2906, "lr": 5.448571155659832e-07}
{"train_loss": 0.0430811271071434, "global_step": 258706, "epoch": 2906, "lr": 5.447717665067742e-07}
{"train_loss": 0.02737531252205372, "global_step": 258707, "epoch": 2906, "lr": 5.446864240962136e-07}
{"train_loss": 0.07641132920980453, "global_step": 258708, "epoch": 2906, "lr": 5.446010883343177e-07}
{"train_loss": 0.04444241523742676, "global_step": 258709, "epoch": 2906, "lr": 5.445157592210981e-07}
{"train_loss": 0.04228038713335991, "global_step": 258710, "epoch": 2906, "lr": 5.444304367565656e-07}
{"train_loss": 0.026050209999084473, "global_step": 258711, "epoch": 2906, "lr": 5.443451209407258e-07}
{"train_loss": 0.031681161373853683, "global_step": 258712, "epoch": 2906, "lr": 5.442598117736064e-07}
{"train_loss": 0.027681704610586166, "global_step": 258713, "epoch": 2906, "lr": 5.44174509255202e-07}
{"train_loss": 0.01893317885696888, "global_step": 258714, "epoch": 2906, "lr": 5.440892133855291e-07}
{"train_loss": 0.02285010926425457, "global_step": 258715, "epoch": 2906, "lr": 5.440039241646045e-07}
{"train_loss": 0.04635712504386902, "global_step": 258716, "epoch": 2906, "lr": 5.439186415924335e-07}
{"train_loss": 0.035785313695669174, "global_step": 258717, "epoch": 2906, "lr": 5.438333656690331e-07}
{"train_loss": 0.028823906555771828, "global_step": 258718, "epoch": 2906, "lr": 5.437480963944031e-07}
{"train_loss": 0.06555389612913132, "global_step": 258719, "epoch": 2906, "lr": 5.436628337685712e-07}
{"train_loss": 0.07172061502933502, "global_step": 258720, "epoch": 2906, "lr": 5.435775777915375e-07}
{"train_loss": 0.06072156876325607, "global_step": 258721, "epoch": 2906, "lr": 5.434923284633242e-07}
{"train_loss": 0.04560577575368493, "global_step": 258722, "epoch": 2906, "lr": 5.434070857839257e-07, "val_loss": 9.617327690124512}
{"train_loss": 0.05421796068549156, "global_step": 258723, "epoch": 2907, "lr": 5.433218497533699e-07}
{"train_loss": 0.08185888826847076, "global_step": 258724, "epoch": 2907, "lr": 5.432366203716566e-07}
{"train_loss": 0.01647038385272026, "global_step": 258725, "epoch": 2907, "lr": 5.43151397638808e-07}
{"train_loss": 0.05164889991283417, "global_step": 258726, "epoch": 2907, "lr": 5.4306618155483e-07}
{"train_loss": 0.0849914476275444, "global_step": 258727, "epoch": 2907, "lr": 5.429809721197276e-07}
{"train_loss": 0.04021399840712547, "global_step": 258728, "epoch": 2907, "lr": 5.428957693335235e-07}
{"train_loss": 0.059569913893938065, "global_step": 258729, "epoch": 2907, "lr": 5.42810573196223e-07}
{"train_loss": 0.02605707198381424, "global_step": 258730, "epoch": 2907, "lr": 5.427253837078427e-07}
{"train_loss": 0.024489374831318855, "global_step": 258731, "epoch": 2907, "lr": 5.426402008683829e-07}
{"train_loss": 0.08576948195695877, "global_step": 258732, "epoch": 2907, "lr": 5.425550246778654e-07}
{"train_loss": 0.04355546087026596, "global_step": 258733, "epoch": 2907, "lr": 5.424698551362961e-07}
{"train_loss": 0.03606342151761055, "global_step": 258734, "epoch": 2907, "lr": 5.42384692243697e-07}
{"train_loss": 0.0467802993953228, "global_step": 258735, "epoch": 2907, "lr": 5.422995360000627e-07}
{"train_loss": 0.021208658814430237, "global_step": 258736, "epoch": 2907, "lr": 5.422143864054207e-07}
{"train_loss": 0.046362169086933136, "global_step": 258737, "epoch": 2907, "lr": 5.421292434597658e-07}
{"train_loss": 0.07122138887643814, "global_step": 258738, "epoch": 2907, "lr": 5.42044107163131e-07}
{"train_loss": 0.05969662219285965, "global_step": 258739, "epoch": 2907, "lr": 5.419589775155054e-07}
{"train_loss": 0.019323578104376793, "global_step": 258740, "epoch": 2907, "lr": 5.418738545169167e-07}
{"train_loss": 0.023712806403636932, "global_step": 258741, "epoch": 2907, "lr": 5.417887381673648e-07}
{"train_loss": 0.02751135267317295, "global_step": 258742, "epoch": 2907, "lr": 5.417036284668719e-07}
{"train_loss": 0.030107423663139343, "global_step": 258743, "epoch": 2907, "lr": 5.416185254154382e-07}
{"train_loss": 0.023275168612599373, "global_step": 258744, "epoch": 2907, "lr": 5.415334290130858e-07}
{"train_loss": 0.0247591994702816, "global_step": 258745, "epoch": 2907, "lr": 5.414483392598202e-07}
{"train_loss": 0.044060319662094116, "global_step": 258746, "epoch": 2907, "lr": 5.413632561556525e-07}
{"train_loss": 0.018895018845796585, "global_step": 258747, "epoch": 2907, "lr": 5.412781797005994e-07}
{"train_loss": 0.043224528431892395, "global_step": 258748, "epoch": 2907, "lr": 5.411931098946666e-07}
{"train_loss": 0.05987986922264099, "global_step": 258749, "epoch": 2907, "lr": 5.411080467378649e-07}
{"train_loss": 0.030937261879444122, "global_step": 258750, "epoch": 2907, "lr": 5.410229902302111e-07}
{"train_loss": 0.05323992297053337, "global_step": 258751, "epoch": 2907, "lr": 5.409379403717163e-07}
{"train_loss": 0.07938966155052185, "global_step": 258752, "epoch": 2907, "lr": 5.408528971623861e-07}
{"train_loss": 0.09765498340129852, "global_step": 258753, "epoch": 2907, "lr": 5.407678606022371e-07}
{"train_loss": 0.03934573754668236, "global_step": 258754, "epoch": 2907, "lr": 5.406828306912747e-07}
{"train_loss": 0.03780471906065941, "global_step": 258755, "epoch": 2907, "lr": 5.405978074295215e-07}
{"train_loss": 0.03887849301099777, "global_step": 258756, "epoch": 2907, "lr": 5.405127908169771e-07}
{"train_loss": 0.01625879481434822, "global_step": 258757, "epoch": 2907, "lr": 5.404277808536585e-07}
{"train_loss": 0.0356261245906353, "global_step": 258758, "epoch": 2907, "lr": 5.403427775395764e-07}
{"train_loss": 0.015555127523839474, "global_step": 258759, "epoch": 2907, "lr": 5.402577808747477e-07}
{"train_loss": 0.027746235951781273, "global_step": 258760, "epoch": 2907, "lr": 5.401727908591725e-07}
{"train_loss": 0.06284788995981216, "global_step": 258761, "epoch": 2907, "lr": 5.400878074928727e-07}
{"train_loss": 0.013961431570351124, "global_step": 258762, "epoch": 2907, "lr": 5.400028307758486e-07}
{"train_loss": 0.04439057409763336, "global_step": 258763, "epoch": 2907, "lr": 5.399178607081279e-07}
{"train_loss": 0.056250423192977905, "global_step": 258764, "epoch": 2907, "lr": 5.398328972897049e-07}
{"train_loss": 0.020522641018033028, "global_step": 258765, "epoch": 2907, "lr": 5.397479405206018e-07}
{"train_loss": 0.06168537959456444, "global_step": 258766, "epoch": 2907, "lr": 5.396629904008244e-07}
{"train_loss": 0.04707537591457367, "global_step": 258767, "epoch": 2907, "lr": 5.39578046930389e-07}
{"train_loss": 0.059928253293037415, "global_step": 258768, "epoch": 2907, "lr": 5.394931101093015e-07}
{"train_loss": 0.05646413192152977, "global_step": 258769, "epoch": 2907, "lr": 5.394081799375783e-07}
{"train_loss": 0.04227808490395546, "global_step": 258770, "epoch": 2907, "lr": 5.393232564152306e-07}
{"train_loss": 0.057007305324077606, "global_step": 258771, "epoch": 2907, "lr": 5.392383395422639e-07}
{"train_loss": 0.06762322783470154, "global_step": 258772, "epoch": 2907, "lr": 5.391534293187006e-07}
{"train_loss": 0.052515700459480286, "global_step": 258773, "epoch": 2907, "lr": 5.390685257445349e-07}
{"train_loss": 0.03668961673974991, "global_step": 258774, "epoch": 2907, "lr": 5.389836288197946e-07}
{"train_loss": 0.06956585496664047, "global_step": 258775, "epoch": 2907, "lr": 5.388987385444855e-07}
{"train_loss": 0.02598860301077366, "global_step": 258776, "epoch": 2907, "lr": 5.388138549186183e-07}
{"train_loss": 0.08651882410049438, "global_step": 258777, "epoch": 2907, "lr": 5.387289779422044e-07}
{"train_loss": 0.06233122572302818, "global_step": 258778, "epoch": 2907, "lr": 5.386441076152549e-07}
{"train_loss": 0.04223715886473656, "global_step": 258779, "epoch": 2907, "lr": 5.385592439377806e-07}
{"train_loss": 0.026503268629312515, "global_step": 258780, "epoch": 2907, "lr": 5.384743869097986e-07}
{"train_loss": 0.08342113345861435, "global_step": 258781, "epoch": 2907, "lr": 5.383895365313141e-07}
{"train_loss": 0.06275688111782074, "global_step": 258782, "epoch": 2907, "lr": 5.383046928023383e-07}
{"train_loss": 0.034133851528167725, "global_step": 258783, "epoch": 2907, "lr": 5.382198557228824e-07}
{"train_loss": 0.027968302369117737, "global_step": 258784, "epoch": 2907, "lr": 5.381350252929685e-07}
{"train_loss": 0.03628189489245415, "global_step": 258785, "epoch": 2907, "lr": 5.38050201512591e-07}
{"train_loss": 0.028053754940629005, "global_step": 258786, "epoch": 2907, "lr": 5.379653843817722e-07}
{"train_loss": 0.09460816532373428, "global_step": 258787, "epoch": 2907, "lr": 5.378805739005233e-07}
{"train_loss": 0.02300408110022545, "global_step": 258788, "epoch": 2907, "lr": 5.377957700688497e-07}
{"train_loss": 0.043249525129795074, "global_step": 258789, "epoch": 2907, "lr": 5.377109728867736e-07}
{"train_loss": 0.04812805354595184, "global_step": 258790, "epoch": 2907, "lr": 5.376261823542894e-07}
{"train_loss": 0.046202197670936584, "global_step": 258791, "epoch": 2907, "lr": 5.37541398471425e-07}
{"train_loss": 0.08775597810745239, "global_step": 258792, "epoch": 2907, "lr": 5.37456621238186e-07}
{"train_loss": 0.03643861040472984, "global_step": 258793, "epoch": 2907, "lr": 5.373718506545832e-07}
{"train_loss": 0.04902312904596329, "global_step": 258794, "epoch": 2907, "lr": 5.372870867206226e-07}
{"train_loss": 0.019924430176615715, "global_step": 258795, "epoch": 2907, "lr": 5.372023294363259e-07}
{"train_loss": 0.017917079851031303, "global_step": 258796, "epoch": 2907, "lr": 5.371175788016991e-07}
{"train_loss": 0.04323441907763481, "global_step": 258797, "epoch": 2907, "lr": 5.37032834816753e-07}
{"train_loss": 0.037691742181777954, "global_step": 258798, "epoch": 2907, "lr": 5.369480974814989e-07}
{"train_loss": 0.03187773376703262, "global_step": 258799, "epoch": 2907, "lr": 5.368633667959477e-07}
{"train_loss": 0.00847562961280346, "global_step": 258800, "epoch": 2907, "lr": 5.367786427601163e-07}
{"train_loss": 0.06548187136650085, "global_step": 258801, "epoch": 2907, "lr": 5.3669392537401e-07}
{"train_loss": 0.03318588435649872, "global_step": 258802, "epoch": 2907, "lr": 5.366092146376456e-07}
{"train_loss": 0.057537540793418884, "global_step": 258803, "epoch": 2907, "lr": 5.365245105510285e-07}
{"train_loss": 0.05395008996129036, "global_step": 258804, "epoch": 2907, "lr": 5.364398131141702e-07}
{"train_loss": 0.023433459922671318, "global_step": 258805, "epoch": 2907, "lr": 5.363551223270869e-07}
{"train_loss": 0.0488097257912159, "global_step": 258806, "epoch": 2907, "lr": 5.362704381897899e-07}
{"train_loss": 0.015198487788438797, "global_step": 258807, "epoch": 2907, "lr": 5.361857607022846e-07}
{"train_loss": 0.07191427797079086, "global_step": 258808, "epoch": 2907, "lr": 5.361010898645935e-07}
{"train_loss": 0.035606369376182556, "global_step": 258809, "epoch": 2907, "lr": 5.360164256767108e-07}
{"train_loss": 0.06583169102668762, "global_step": 258810, "epoch": 2907, "lr": 5.359317681386644e-07}
{"train_loss": 0.044793178965703825, "global_step": 258811, "epoch": 2907, "lr": 5.358471172504542e-07, "val_loss": 9.563294410705566}
{"train_loss": 0.027424776926636696, "global_step": 258812, "epoch": 2908, "lr": 5.357624730121025e-07}
{"train_loss": 0.07888685166835785, "global_step": 258813, "epoch": 2908, "lr": 5.356778354236092e-07}
{"train_loss": 0.013775601983070374, "global_step": 258814, "epoch": 2908, "lr": 5.355932044849965e-07}
{"train_loss": 0.0448286309838295, "global_step": 258815, "epoch": 2908, "lr": 5.355085801962645e-07}
{"train_loss": 0.060311079025268555, "global_step": 258816, "epoch": 2908, "lr": 5.354239625574353e-07}
{"train_loss": 0.0551641546189785, "global_step": 258817, "epoch": 2908, "lr": 5.353393515685145e-07}
{"train_loss": 0.05179143697023392, "global_step": 258818, "epoch": 2908, "lr": 5.352547472295133e-07}
{"train_loss": 0.05706248804926872, "global_step": 258819, "epoch": 2908, "lr": 5.351701495404426e-07}
{"train_loss": 0.02715073898434639, "global_step": 258820, "epoch": 2908, "lr": 5.350855585013193e-07}
{"train_loss": 0.05002012476325035, "global_step": 258821, "epoch": 2908, "lr": 5.350009741121486e-07}
{"train_loss": 0.02873210236430168, "global_step": 258822, "epoch": 2908, "lr": 5.34916396372942e-07}
{"train_loss": 0.020350869745016098, "global_step": 258823, "epoch": 2908, "lr": 5.348318252837159e-07}
{"train_loss": 0.025976117700338364, "global_step": 258824, "epoch": 2908, "lr": 5.347472608444814e-07}
{"train_loss": 0.05823148787021637, "global_step": 258825, "epoch": 2908, "lr": 5.346627030552443e-07}
{"train_loss": 0.05979003757238388, "global_step": 258826, "epoch": 2908, "lr": 5.345781519160154e-07}
{"train_loss": 0.07659441232681274, "global_step": 258827, "epoch": 2908, "lr": 5.34493607426817e-07}
{"train_loss": 0.018383629620075226, "global_step": 258828, "epoch": 2908, "lr": 5.344090695876436e-07}
{"train_loss": 0.06524816155433655, "global_step": 258829, "epoch": 2908, "lr": 5.343245383985229e-07}
{"train_loss": 0.020732099190354347, "global_step": 258830, "epoch": 2908, "lr": 5.34240013859455e-07}
{"train_loss": 0.07981261610984802, "global_step": 258831, "epoch": 2908, "lr": 5.34155495970462e-07}
{"train_loss": 0.04615408927202225, "global_step": 258832, "epoch": 2908, "lr": 5.340709847315439e-07}
{"train_loss": 0.06084198132157326, "global_step": 258833, "epoch": 2908, "lr": 5.33986480142723e-07}
{"train_loss": 0.03299213573336601, "global_step": 258834, "epoch": 2908, "lr": 5.339019822039937e-07}
{"train_loss": 0.04942019283771515, "global_step": 258835, "epoch": 2908, "lr": 5.338174909153892e-07}
{"train_loss": 0.05890687555074692, "global_step": 258836, "epoch": 2908, "lr": 5.337330062769041e-07}
{"train_loss": 0.04406854510307312, "global_step": 258837, "epoch": 2908, "lr": 5.336485282885606e-07}
{"train_loss": 0.045614197850227356, "global_step": 258838, "epoch": 2908, "lr": 5.335640569503586e-07}
{"train_loss": 0.04360269382596016, "global_step": 258839, "epoch": 2908, "lr": 5.334795922623203e-07}
{"train_loss": 0.0412244014441967, "global_step": 258840, "epoch": 2908, "lr": 5.333951342244514e-07}
{"train_loss": 0.057688239961862564, "global_step": 258841, "epoch": 2908, "lr": 5.333106828367629e-07}
{"train_loss": 0.01876705512404442, "global_step": 258842, "epoch": 2908, "lr": 5.332262380992714e-07}
{"train_loss": 0.020599979907274246, "global_step": 258843, "epoch": 2908, "lr": 5.331418000119826e-07}
{"train_loss": 0.03491847589612007, "global_step": 258844, "epoch": 2908, "lr": 5.33057368574913e-07}
{"train_loss": 0.009133891202509403, "global_step": 258845, "epoch": 2908, "lr": 5.329729437880681e-07}
{"train_loss": 0.046141281723976135, "global_step": 258846, "epoch": 2908, "lr": 5.328885256514649e-07}
{"train_loss": 0.052035003900527954, "global_step": 258847, "epoch": 2908, "lr": 5.32804114165103e-07}
{"train_loss": 0.035053204745054245, "global_step": 258848, "epoch": 2908, "lr": 5.327197093290104e-07}
{"train_loss": 0.05917700380086899, "global_step": 258849, "epoch": 2908, "lr": 5.326353111431926e-07}
{"train_loss": 0.0427824892103672, "global_step": 258850, "epoch": 2908, "lr": 5.32550919607655e-07}
{"train_loss": 0.02847648225724697, "global_step": 258851, "epoch": 2908, "lr": 5.324665347224145e-07}
{"train_loss": 0.03519926592707634, "global_step": 258852, "epoch": 2908, "lr": 5.32382156487482e-07}
{"train_loss": 0.032171111553907394, "global_step": 258853, "epoch": 2908, "lr": 5.322977849028632e-07}
{"train_loss": 0.05429277941584587, "global_step": 258854, "epoch": 2908, "lr": 5.322134199685802e-07}
{"train_loss": 0.07799947261810303, "global_step": 258855, "epoch": 2908, "lr": 5.32129061684633e-07}
{"train_loss": 0.03918898478150368, "global_step": 258856, "epoch": 2908, "lr": 5.320447100510384e-07}
{"train_loss": 0.04524216055870056, "global_step": 258857, "epoch": 2908, "lr": 5.319603650678074e-07}
{"train_loss": 0.0705738291144371, "global_step": 258858, "epoch": 2908, "lr": 5.31876026734951e-07}
{"train_loss": 0.08617495000362396, "global_step": 258859, "epoch": 2908, "lr": 5.31791695052486e-07}
{"train_loss": 0.05585230141878128, "global_step": 258860, "epoch": 2908, "lr": 5.317073700204123e-07}
{"train_loss": 0.02452506497502327, "global_step": 258861, "epoch": 2908, "lr": 5.316230516387521e-07}
{"train_loss": 0.023344891145825386, "global_step": 258862, "epoch": 2908, "lr": 5.315387399075111e-07}
{"train_loss": 0.06939925253391266, "global_step": 258863, "epoch": 2908, "lr": 5.314544348267059e-07}
{"train_loss": 0.03566208854317665, "global_step": 258864, "epoch": 2908, "lr": 5.313701363963363e-07}
{"train_loss": 0.025883251801133156, "global_step": 258865, "epoch": 2908, "lr": 5.312858446164248e-07}
{"train_loss": 0.033412110060453415, "global_step": 258866, "epoch": 2908, "lr": 5.312015594869769e-07}
{"train_loss": 0.038233064115047455, "global_step": 258867, "epoch": 2908, "lr": 5.311172810080089e-07}
{"train_loss": 0.018016358837485313, "global_step": 258868, "epoch": 2908, "lr": 5.310330091795268e-07}
{"train_loss": 0.06871233135461807, "global_step": 258869, "epoch": 2908, "lr": 5.309487440015471e-07}
{"train_loss": 0.05809347331523895, "global_step": 258870, "epoch": 2908, "lr": 5.308644854740752e-07}
{"train_loss": 0.05621793866157532, "global_step": 258871, "epoch": 2908, "lr": 5.307802335971279e-07}
{"train_loss": 0.040908411145210266, "global_step": 258872, "epoch": 2908, "lr": 5.306959883707164e-07}
{"train_loss": 0.02303636074066162, "global_step": 258873, "epoch": 2908, "lr": 5.30611749794846e-07}
{"train_loss": 0.06326302886009216, "global_step": 258874, "epoch": 2908, "lr": 5.305275178695334e-07}
{"train_loss": 0.038913629949092865, "global_step": 258875, "epoch": 2908, "lr": 5.3044329259479e-07}
{"train_loss": 0.0657234787940979, "global_step": 258876, "epoch": 2908, "lr": 5.30359073970621e-07}
{"train_loss": 0.029321514070034027, "global_step": 258877, "epoch": 2908, "lr": 5.302748619970488e-07}
{"train_loss": 0.04239431768655777, "global_step": 258878, "epoch": 2908, "lr": 5.301906566740678e-07}
{"train_loss": 0.012856951914727688, "global_step": 258879, "epoch": 2908, "lr": 5.301064580017113e-07}
{"train_loss": 0.05540621280670166, "global_step": 258880, "epoch": 2908, "lr": 5.300222659799736e-07}
{"train_loss": 0.040936969220638275, "global_step": 258881, "epoch": 2908, "lr": 5.299380806088716e-07}
{"train_loss": 0.020577292889356613, "global_step": 258882, "epoch": 2908, "lr": 5.298539018884163e-07}
{"train_loss": 0.04437913000583649, "global_step": 258883, "epoch": 2908, "lr": 5.297697298186188e-07}
{"train_loss": 0.03472035005688667, "global_step": 258884, "epoch": 2908, "lr": 5.296855643994958e-07}
{"train_loss": 0.035373393446207047, "global_step": 258885, "epoch": 2908, "lr": 5.296014056310472e-07}
{"train_loss": 0.06503797322511673, "global_step": 258886, "epoch": 2908, "lr": 5.295172535132953e-07}
{"train_loss": 0.027907252311706543, "global_step": 258887, "epoch": 2908, "lr": 5.2943310804624e-07}
{"train_loss": 0.022002633661031723, "global_step": 258888, "epoch": 2908, "lr": 5.293489692299036e-07}
{"train_loss": 0.051483262330293655, "global_step": 258889, "epoch": 2908, "lr": 5.292648370642917e-07}
{"train_loss": 0.049422696232795715, "global_step": 258890, "epoch": 2908, "lr": 5.291807115494207e-07}
{"train_loss": 0.035986851900815964, "global_step": 258891, "epoch": 2908, "lr": 5.290965926852964e-07}
{"train_loss": 0.017845401540398598, "global_step": 258892, "epoch": 2908, "lr": 5.290124804719354e-07}
{"train_loss": 0.062397636473178864, "global_step": 258893, "epoch": 2908, "lr": 5.289283749093432e-07}
{"train_loss": 0.021674197167158127, "global_step": 258894, "epoch": 2908, "lr": 5.288442759975309e-07}
{"train_loss": 0.029480725526809692, "global_step": 258895, "epoch": 2908, "lr": 5.287601837365153e-07}
{"train_loss": 0.03499208018183708, "global_step": 258896, "epoch": 2908, "lr": 5.286760981263017e-07}
{"train_loss": 0.050173792988061905, "global_step": 258897, "epoch": 2908, "lr": 5.28592019166907e-07}
{"train_loss": 0.040149394422769547, "global_step": 258898, "epoch": 2908, "lr": 5.28507946858342e-07}
{"train_loss": 0.04530024155974388, "global_step": 258899, "epoch": 2908, "lr": 5.284238812006126e-07}
{"train_loss": 0.04308539551546734, "global_step": 258900, "epoch": 2908, "lr": 5.283398221937352e-07, "val_loss": 9.610025405883789}
{"train_loss": 0.02996646799147129, "global_step": 258901, "epoch": 2909, "lr": 5.28255769837721e-07}
{"train_loss": 0.05418016389012337, "global_step": 258902, "epoch": 2909, "lr": 5.281717241325812e-07}
{"train_loss": 0.061390288174152374, "global_step": 258903, "epoch": 2909, "lr": 5.280876850783212e-07}
{"train_loss": 0.04164104536175728, "global_step": 258904, "epoch": 2909, "lr": 5.280036526749577e-07}
{"train_loss": 0.02264644391834736, "global_step": 258905, "epoch": 2909, "lr": 5.279196269225073e-07}
{"train_loss": 0.029030699282884598, "global_step": 258906, "epoch": 2909, "lr": 5.278356078209645e-07}
{"train_loss": 0.02798144333064556, "global_step": 258907, "epoch": 2909, "lr": 5.277515953703626e-07}
{"train_loss": 0.019146861508488655, "global_step": 258908, "epoch": 2909, "lr": 5.276675895706907e-07}
{"train_loss": 0.035238709300756454, "global_step": 258909, "epoch": 2909, "lr": 5.275835904219817e-07}
{"train_loss": 0.04634440690279007, "global_step": 258910, "epoch": 2909, "lr": 5.274995979242248e-07}
{"train_loss": 0.02228546142578125, "global_step": 258911, "epoch": 2909, "lr": 5.274156120774531e-07}
{"train_loss": 0.017631279304623604, "global_step": 258912, "epoch": 2909, "lr": 5.273316328816613e-07}
{"train_loss": 0.023462757468223572, "global_step": 258913, "epoch": 2909, "lr": 5.27247660336866e-07}
{"train_loss": 0.035010963678359985, "global_step": 258914, "epoch": 2909, "lr": 5.271636944430836e-07}
{"train_loss": 0.027670027688145638, "global_step": 258915, "epoch": 2909, "lr": 5.270797352003199e-07}
{"train_loss": 0.047472286969423294, "global_step": 258916, "epoch": 2909, "lr": 5.269957826085803e-07}
{"train_loss": 0.023821713402867317, "global_step": 258917, "epoch": 2909, "lr": 5.269118366678927e-07}
{"train_loss": 0.09337988495826721, "global_step": 258918, "epoch": 2909, "lr": 5.26827897378257e-07}
{"train_loss": 0.05199854448437691, "global_step": 258919, "epoch": 2909, "lr": 5.267439647396843e-07}
{"train_loss": 0.05869240686297417, "global_step": 258920, "epoch": 2909, "lr": 5.266600387521859e-07}
{"train_loss": 0.027869973331689835, "global_step": 258921, "epoch": 2909, "lr": 5.265761194157781e-07}
{"train_loss": 0.059931546449661255, "global_step": 258922, "epoch": 2909, "lr": 5.264922067304668e-07}
{"train_loss": 0.02602408081293106, "global_step": 258923, "epoch": 2909, "lr": 5.264083006962684e-07}
{"train_loss": 0.07448238879442215, "global_step": 258924, "epoch": 2909, "lr": 5.263244013131885e-07}
{"train_loss": 0.04549359157681465, "global_step": 258925, "epoch": 2909, "lr": 5.262405085812439e-07}
{"train_loss": 0.09629837423563004, "global_step": 258926, "epoch": 2909, "lr": 5.261566225004455e-07}
{"train_loss": 0.06496735662221909, "global_step": 258927, "epoch": 2909, "lr": 5.260727430707935e-07}
{"train_loss": 0.03448475897312164, "global_step": 258928, "epoch": 2909, "lr": 5.259888702923155e-07}
{"train_loss": 0.05188922584056854, "global_step": 258929, "epoch": 2909, "lr": 5.259050041650115e-07}
{"train_loss": 0.09691005200147629, "global_step": 258930, "epoch": 2909, "lr": 5.258211446889039e-07}
{"train_loss": 0.02908516861498356, "global_step": 258931, "epoch": 2909, "lr": 5.257372918639869e-07}
{"train_loss": 0.06052515283226967, "global_step": 258932, "epoch": 2909, "lr": 5.256534456902828e-07}
{"train_loss": 0.027828069403767586, "global_step": 258933, "epoch": 2909, "lr": 5.255696061678084e-07}
{"train_loss": 0.034407585859298706, "global_step": 258934, "epoch": 2909, "lr": 5.254857732965634e-07}
{"train_loss": 0.049936145544052124, "global_step": 258935, "epoch": 2909, "lr": 5.254019470765647e-07}
{"train_loss": 0.04188377782702446, "global_step": 258936, "epoch": 2909, "lr": 5.253181275078234e-07}
{"train_loss": 0.0475597083568573, "global_step": 258937, "epoch": 2909, "lr": 5.252343145903504e-07}
{"train_loss": 0.05411522835493088, "global_step": 258938, "epoch": 2909, "lr": 5.251505083241515e-07}
{"train_loss": 0.05243131145834923, "global_step": 258939, "epoch": 2909, "lr": 5.250667087092486e-07}
{"train_loss": 0.08051953464746475, "global_step": 258940, "epoch": 2909, "lr": 5.249829157456476e-07}
{"train_loss": 0.05159943923354149, "global_step": 258941, "epoch": 2909, "lr": 5.248991294333594e-07}
{"train_loss": 0.05588716268539429, "global_step": 258942, "epoch": 2909, "lr": 5.248153497723896e-07}
{"train_loss": 0.027570052072405815, "global_step": 258943, "epoch": 2909, "lr": 5.247315767627603e-07}
{"train_loss": 0.050659164786338806, "global_step": 258944, "epoch": 2909, "lr": 5.246478104044772e-07}
{"train_loss": 0.05521769821643829, "global_step": 258945, "epoch": 2909, "lr": 5.245640506975569e-07}
{"train_loss": 0.0438002273440361, "global_step": 258946, "epoch": 2909, "lr": 5.244802976419993e-07}
{"train_loss": 0.04201667383313179, "global_step": 258947, "epoch": 2909, "lr": 5.243965512378268e-07}
{"train_loss": 0.06685768812894821, "global_step": 258948, "epoch": 2909, "lr": 5.243128114850393e-07}
{"train_loss": 0.06022069603204727, "global_step": 258949, "epoch": 2909, "lr": 5.242290783836589e-07}
{"train_loss": 0.03795357048511505, "global_step": 258950, "epoch": 2909, "lr": 5.24145351933697e-07}
{"train_loss": 0.04121243581175804, "global_step": 258951, "epoch": 2909, "lr": 5.240616321351588e-07}
{"train_loss": 0.015993699431419373, "global_step": 258952, "epoch": 2909, "lr": 5.239779189880501e-07}
{"train_loss": 0.03116922453045845, "global_step": 258953, "epoch": 2909, "lr": 5.238942124923984e-07}
{"train_loss": 0.052755437791347504, "global_step": 258954, "epoch": 2909, "lr": 5.23810512648204e-07}
{"train_loss": 0.03492506965994835, "global_step": 258955, "epoch": 2909, "lr": 5.237268194554778e-07}
{"train_loss": 0.041077326983213425, "global_step": 258956, "epoch": 2909, "lr": 5.236431329142366e-07}
{"train_loss": 0.04671119898557663, "global_step": 258957, "epoch": 2909, "lr": 5.235594530244859e-07}
{"train_loss": 0.0484338104724884, "global_step": 258958, "epoch": 2909, "lr": 5.234757797862422e-07}
{"train_loss": 0.03770856931805611, "global_step": 258959, "epoch": 2909, "lr": 5.233921131995112e-07}
{"train_loss": 0.06324493885040283, "global_step": 258960, "epoch": 2909, "lr": 5.233084532643095e-07}
{"train_loss": 0.0193694569170475, "global_step": 258961, "epoch": 2909, "lr": 5.232247999806427e-07}
{"train_loss": 0.0288563072681427, "global_step": 258962, "epoch": 2909, "lr": 5.231411533485275e-07}
{"train_loss": 0.03561259061098099, "global_step": 258963, "epoch": 2909, "lr": 5.230575133679749e-07}
{"train_loss": 0.04652462899684906, "global_step": 258964, "epoch": 2909, "lr": 5.229738800389905e-07}
{"train_loss": 0.016554899513721466, "global_step": 258965, "epoch": 2909, "lr": 5.22890253361591e-07}
{"train_loss": 0.06071990728378296, "global_step": 258966, "epoch": 2909, "lr": 5.228066333357873e-07}
{"train_loss": 0.031967952847480774, "global_step": 258967, "epoch": 2909, "lr": 5.227230199615851e-07}
{"train_loss": 0.05204194411635399, "global_step": 258968, "epoch": 2909, "lr": 5.226394132390012e-07}
{"train_loss": 0.07318910956382751, "global_step": 258969, "epoch": 2909, "lr": 5.225558131680519e-07}
{"train_loss": 0.06814109534025192, "global_step": 258970, "epoch": 2909, "lr": 5.224722197487319e-07}
{"train_loss": 0.05294707790017128, "global_step": 258971, "epoch": 2909, "lr": 5.223886329810689e-07}
{"train_loss": 0.020275669172406197, "global_step": 258972, "epoch": 2909, "lr": 5.223050528650631e-07}
{"train_loss": 0.007766291033476591, "global_step": 258973, "epoch": 2909, "lr": 5.222214794007363e-07}
{"train_loss": 0.0538736954331398, "global_step": 258974, "epoch": 2909, "lr": 5.221379125880888e-07}
{"train_loss": 0.028346145525574684, "global_step": 258975, "epoch": 2909, "lr": 5.220543524271426e-07}
{"train_loss": 0.05581970885396004, "global_step": 258976, "epoch": 2909, "lr": 5.219707989178923e-07}
{"train_loss": 0.010478763841092587, "global_step": 258977, "epoch": 2909, "lr": 5.218872520603713e-07}
{"train_loss": 0.07253719121217728, "global_step": 258978, "epoch": 2909, "lr": 5.218037118545738e-07}
{"train_loss": 0.03811442106962204, "global_step": 258979, "epoch": 2909, "lr": 5.217201783005165e-07}
{"train_loss": 0.03782806918025017, "global_step": 258980, "epoch": 2909, "lr": 5.216366513982107e-07}
{"train_loss": 0.061098676174879074, "global_step": 258981, "epoch": 2909, "lr": 5.215531311476729e-07}
{"train_loss": 0.03962346911430359, "global_step": 258982, "epoch": 2909, "lr": 5.214696175489031e-07}
{"train_loss": 0.05865058675408363, "global_step": 258983, "epoch": 2909, "lr": 5.213861106019235e-07}
{"train_loss": 0.03997565805912018, "global_step": 258984, "epoch": 2909, "lr": 5.213026103067398e-07}
{"train_loss": 0.030174095183610916, "global_step": 258985, "epoch": 2909, "lr": 5.212191166633573e-07}
{"train_loss": 0.030598783865571022, "global_step": 258986, "epoch": 2909, "lr": 5.21135629671804e-07}
{"train_loss": 0.03698478639125824, "global_step": 258987, "epoch": 2909, "lr": 5.210521493320741e-07}
{"train_loss": 0.10248222947120667, "global_step": 258988, "epoch": 2909, "lr": 5.209686756441845e-07}
{"train_loss": 0.04441474528271663, "global_step": 258989, "epoch": 2909, "lr": 5.208852086081517e-07, "val_loss": 9.610978126525879}
{"train_loss": 0.05779522284865379, "global_step": 258990, "epoch": 2910, "lr": 5.208017482239813e-07}
{"train_loss": 0.061255283653736115, "global_step": 258991, "epoch": 2910, "lr": 5.207182944916899e-07}
{"train_loss": 0.05060635134577751, "global_step": 258992, "epoch": 2910, "lr": 5.206348474112776e-07}
{"train_loss": 0.02748130075633526, "global_step": 258993, "epoch": 2910, "lr": 5.205514069827722e-07}
{"train_loss": 0.03286914899945259, "global_step": 258994, "epoch": 2910, "lr": 5.204679732061679e-07}
{"train_loss": 0.02150942198932171, "global_step": 258995, "epoch": 2910, "lr": 5.203845460814871e-07}
{"train_loss": 0.048923999071121216, "global_step": 258996, "epoch": 2910, "lr": 5.203011256087353e-07}
{"train_loss": 0.07761583477258682, "global_step": 258997, "epoch": 2910, "lr": 5.202177117879237e-07}
{"train_loss": 0.06406307965517044, "global_step": 258998, "epoch": 2910, "lr": 5.201343046190743e-07}
{"train_loss": 0.07261202484369278, "global_step": 258999, "epoch": 2910, "lr": 5.200509041021817e-07}
{"train_loss": 0.09837092459201813, "global_step": 259000, "epoch": 2910, "lr": 5.199675102372681e-07}
{"train_loss": 0.03784610703587532, "global_step": 259001, "epoch": 2910, "lr": 5.19884123024339e-07}
{"train_loss": 0.0522567480802536, "global_step": 259002, "epoch": 2910, "lr": 5.198007424634166e-07}
{"train_loss": 0.05670253932476044, "global_step": 259003, "epoch": 2910, "lr": 5.197173685544954e-07}
{"train_loss": 0.03985414281487465, "global_step": 259004, "epoch": 2910, "lr": 5.196340012975976e-07}
{"train_loss": 0.06301421672105789, "global_step": 259005, "epoch": 2910, "lr": 5.195506406927342e-07}
{"train_loss": 0.06472202390432358, "global_step": 259006, "epoch": 2910, "lr": 5.194672867399053e-07}
{"train_loss": 0.04501127451658249, "global_step": 259007, "epoch": 2910, "lr": 5.193839394391442e-07}
{"train_loss": 0.05532193183898926, "global_step": 259008, "epoch": 2910, "lr": 5.193005987904398e-07}
{"train_loss": 0.04759320244193077, "global_step": 259009, "epoch": 2910, "lr": 5.192172647938143e-07}
{"train_loss": 0.056312888860702515, "global_step": 259010, "epoch": 2910, "lr": 5.191339374492788e-07}
{"train_loss": 0.041955992579460144, "global_step": 259011, "epoch": 2910, "lr": 5.190506167568387e-07}
{"train_loss": 0.056990548968315125, "global_step": 259012, "epoch": 2910, "lr": 5.18967302716511e-07}
{"train_loss": 0.02350175939500332, "global_step": 259013, "epoch": 2910, "lr": 5.18883995328312e-07}
{"train_loss": 0.05599680915474892, "global_step": 259014, "epoch": 2910, "lr": 5.188006945922363e-07}
{"train_loss": 0.04316166788339615, "global_step": 259015, "epoch": 2910, "lr": 5.187174005083118e-07}
{"train_loss": 0.04577665776014328, "global_step": 259016, "epoch": 2910, "lr": 5.186341130765382e-07}
{"train_loss": 0.07051581889390945, "global_step": 259017, "epoch": 2910, "lr": 5.185508322969323e-07}
{"train_loss": 0.03734773024916649, "global_step": 259018, "epoch": 2910, "lr": 5.184675581694998e-07}
{"train_loss": 0.00654798885807395, "global_step": 259019, "epoch": 2910, "lr": 5.183842906942627e-07}
{"train_loss": 0.029784977436065674, "global_step": 259020, "epoch": 2910, "lr": 5.183010298712209e-07}
{"train_loss": 0.06193069368600845, "global_step": 259021, "epoch": 2910, "lr": 5.182177757003914e-07}
{"train_loss": 0.0385880172252655, "global_step": 259022, "epoch": 2910, "lr": 5.181345281817907e-07}
{"train_loss": 0.040155086666345596, "global_step": 259023, "epoch": 2910, "lr": 5.180512873154186e-07}
{"train_loss": 0.0290371086448431, "global_step": 259024, "epoch": 2910, "lr": 5.17968053101292e-07}
{"train_loss": 0.05988677963614464, "global_step": 259025, "epoch": 2910, "lr": 5.17884825539422e-07}
{"train_loss": 0.08344525098800659, "global_step": 259026, "epoch": 2910, "lr": 5.178016046298195e-07}
{"train_loss": 0.03725855052471161, "global_step": 259027, "epoch": 2910, "lr": 5.177183903724903e-07}
{"train_loss": 0.037946917116642, "global_step": 259028, "epoch": 2910, "lr": 5.176351827674564e-07}
{"train_loss": 0.02751968614757061, "global_step": 259029, "epoch": 2910, "lr": 5.175519818147234e-07}
{"train_loss": 0.036907244473695755, "global_step": 259030, "epoch": 2910, "lr": 5.174687875143025e-07}
{"train_loss": 0.07657206803560257, "global_step": 259031, "epoch": 2910, "lr": 5.173855998662048e-07}
{"train_loss": 0.019895553588867188, "global_step": 259032, "epoch": 2910, "lr": 5.173024188704356e-07}
{"train_loss": 0.07902632653713226, "global_step": 259033, "epoch": 2910, "lr": 5.172192445270175e-07}
{"train_loss": 0.0434691421687603, "global_step": 259034, "epoch": 2910, "lr": 5.171360768359613e-07}
{"train_loss": 0.04201503098011017, "global_step": 259035, "epoch": 2910, "lr": 5.170529157972615e-07}
{"train_loss": 0.03980018198490143, "global_step": 259036, "epoch": 2910, "lr": 5.169697614109515e-07}
{"train_loss": 0.061088766902685165, "global_step": 259037, "epoch": 2910, "lr": 5.168866136770256e-07}
{"train_loss": 0.017319070175290108, "global_step": 259038, "epoch": 2910, "lr": 5.168034725955007e-07}
{"train_loss": 0.05866190791130066, "global_step": 259039, "epoch": 2910, "lr": 5.167203381663877e-07}
{"train_loss": 0.050261080265045166, "global_step": 259040, "epoch": 2910, "lr": 5.166372103896977e-07}
{"train_loss": 0.0708351731300354, "global_step": 259041, "epoch": 2910, "lr": 5.165540892654475e-07}
{"train_loss": 0.046991825103759766, "global_step": 259042, "epoch": 2910, "lr": 5.164709747936425e-07}
{"train_loss": 0.0255966167896986, "global_step": 259043, "epoch": 2910, "lr": 5.163878669742939e-07}
{"train_loss": 0.037101924419403076, "global_step": 259044, "epoch": 2910, "lr": 5.163047658074127e-07}
{"train_loss": 0.04722512140870094, "global_step": 259045, "epoch": 2910, "lr": 5.162216712930157e-07}
{"train_loss": 0.04884197562932968, "global_step": 259046, "epoch": 2910, "lr": 5.161385834311027e-07}
{"train_loss": 0.043200958520174026, "global_step": 259047, "epoch": 2910, "lr": 5.160555022216962e-07}
{"train_loss": 0.042764805257320404, "global_step": 259048, "epoch": 2910, "lr": 5.159724276648015e-07}
{"train_loss": 0.01800786703824997, "global_step": 259049, "epoch": 2910, "lr": 5.158893597604353e-07}
{"train_loss": 0.03519352898001671, "global_step": 259050, "epoch": 2910, "lr": 5.158062985085976e-07}
{"train_loss": 0.09055399149656296, "global_step": 259051, "epoch": 2910, "lr": 5.157232439093108e-07}
{"train_loss": 0.02622019685804844, "global_step": 259052, "epoch": 2910, "lr": 5.156401959625801e-07}
{"train_loss": 0.0572240948677063, "global_step": 259053, "epoch": 2910, "lr": 5.155571546684224e-07}
{"train_loss": 0.01641680672764778, "global_step": 259054, "epoch": 2910, "lr": 5.154741200268431e-07}
{"train_loss": 0.03884249925613403, "global_step": 259055, "epoch": 2910, "lr": 5.153910920378535e-07}
{"train_loss": 0.03846808522939682, "global_step": 259056, "epoch": 2910, "lr": 5.153080707014646e-07}
{"train_loss": 0.03910890966653824, "global_step": 259057, "epoch": 2910, "lr": 5.15225056017693e-07}
{"train_loss": 0.02921966463327408, "global_step": 259058, "epoch": 2910, "lr": 5.151420479865443e-07}
{"train_loss": 0.050013791769742966, "global_step": 259059, "epoch": 2910, "lr": 5.150590466080296e-07}
{"train_loss": 0.027570724487304688, "global_step": 259060, "epoch": 2910, "lr": 5.149760518821711e-07}
{"train_loss": 0.015365317463874817, "global_step": 259061, "epoch": 2910, "lr": 5.148930638089633e-07}
{"train_loss": 0.04730639234185219, "global_step": 259062, "epoch": 2910, "lr": 5.148100823884283e-07}
{"train_loss": 0.027404960244894028, "global_step": 259063, "epoch": 2910, "lr": 5.147271076205662e-07}
{"train_loss": 0.05108501762151718, "global_step": 259064, "epoch": 2910, "lr": 5.146441395054047e-07}
{"train_loss": 0.048175252974033356, "global_step": 259065, "epoch": 2910, "lr": 5.145611780429438e-07}
{"train_loss": 0.049415647983551025, "global_step": 259066, "epoch": 2910, "lr": 5.144782232331946e-07}
{"train_loss": 0.05116073414683342, "global_step": 259067, "epoch": 2910, "lr": 5.143952750761738e-07}
{"train_loss": 0.018801288679242134, "global_step": 259068, "epoch": 2910, "lr": 5.143123335718869e-07}
{"train_loss": 0.04657774791121483, "global_step": 259069, "epoch": 2910, "lr": 5.142293987203506e-07}
{"train_loss": 0.08280885964632034, "global_step": 259070, "epoch": 2910, "lr": 5.141464705215704e-07}
{"train_loss": 0.04515688493847847, "global_step": 259071, "epoch": 2910, "lr": 5.14063548975563e-07}
{"train_loss": 0.047369856387376785, "global_step": 259072, "epoch": 2910, "lr": 5.139806340823339e-07}
{"train_loss": 0.027354268357157707, "global_step": 259073, "epoch": 2910, "lr": 5.138977258418998e-07}
{"train_loss": 0.0296647846698761, "global_step": 259074, "epoch": 2910, "lr": 5.138148242542662e-07}
{"train_loss": 0.023716988041996956, "global_step": 259075, "epoch": 2910, "lr": 5.137319293194442e-07}
{"train_loss": 0.043237537145614624, "global_step": 259076, "epoch": 2910, "lr": 5.136490410374506e-07}
{"train_loss": 0.03031661920249462, "global_step": 259077, "epoch": 2910, "lr": 5.135661594082964e-07}
{"train_loss": 0.04532543789030293, "global_step": 259078, "epoch": 2910, "lr": 5.13483284431987e-07, "val_loss": 9.698708534240723, "train_action_mse_error": 29.1790828704834}
{"train_loss": 0.02610154077410698, "global_step": 259079, "epoch": 2911, "lr": 5.134004161085393e-07}
{"train_loss": 0.02447865530848503, "global_step": 259080, "epoch": 2911, "lr": 5.133175544379587e-07}
{"train_loss": 0.03596123680472374, "global_step": 259081, "epoch": 2911, "lr": 5.13234699420262e-07}
{"train_loss": 0.04080191254615784, "global_step": 259082, "epoch": 2911, "lr": 5.131518510554545e-07}
{"train_loss": 0.034018680453300476, "global_step": 259083, "epoch": 2911, "lr": 5.130690093435531e-07}
{"train_loss": 0.01080171950161457, "global_step": 259084, "epoch": 2911, "lr": 5.129861742845632e-07}
{"train_loss": 0.06141410768032074, "global_step": 259085, "epoch": 2911, "lr": 5.129033458785071e-07}
{"train_loss": 0.04752223193645477, "global_step": 259086, "epoch": 2911, "lr": 5.128205241253792e-07}
{"train_loss": 0.03945871815085411, "global_step": 259087, "epoch": 2911, "lr": 5.127377090252072e-07}
{"train_loss": 0.100446917116642, "global_step": 259088, "epoch": 2911, "lr": 5.126549005779857e-07}
{"train_loss": 0.027275698259472847, "global_step": 259089, "epoch": 2911, "lr": 5.125720987837423e-07}
{"train_loss": 0.05630195885896683, "global_step": 259090, "epoch": 2911, "lr": 5.124893036424772e-07}
{"train_loss": 0.04260971397161484, "global_step": 259091, "epoch": 2911, "lr": 5.124065151542013e-07}
{"train_loss": 0.02464059554040432, "global_step": 259092, "epoch": 2911, "lr": 5.123237333189313e-07}
{"train_loss": 0.03942039608955383, "global_step": 259093, "epoch": 2911, "lr": 5.122409581366839e-07}
{"train_loss": 0.04449491947889328, "global_step": 259094, "epoch": 2911, "lr": 5.121581896074534e-07}
{"train_loss": 0.07308182120323181, "global_step": 259095, "epoch": 2911, "lr": 5.120754277312622e-07}
{"train_loss": 0.046137310564517975, "global_step": 259096, "epoch": 2911, "lr": 5.119926725081214e-07}
{"train_loss": 0.046373795717954636, "global_step": 259097, "epoch": 2911, "lr": 5.119099239380365e-07}
{"train_loss": 0.024984143674373627, "global_step": 259098, "epoch": 2911, "lr": 5.11827182021024e-07}
{"train_loss": 0.010309807024896145, "global_step": 259099, "epoch": 2911, "lr": 5.117444467570898e-07}
{"train_loss": 0.0342150554060936, "global_step": 259100, "epoch": 2911, "lr": 5.116617181462558e-07}
{"train_loss": 0.03326791524887085, "global_step": 259101, "epoch": 2911, "lr": 5.11578996188522e-07}
{"train_loss": 0.08224860578775406, "global_step": 259102, "epoch": 2911, "lr": 5.114962808838997e-07}
{"train_loss": 0.08333861082792282, "global_step": 259103, "epoch": 2911, "lr": 5.114135722324054e-07}
{"train_loss": 0.07212067395448685, "global_step": 259104, "epoch": 2911, "lr": 5.113308702340558e-07}
{"train_loss": 0.05762319266796112, "global_step": 259105, "epoch": 2911, "lr": 5.112481748888454e-07}
{"train_loss": 0.042253993451595306, "global_step": 259106, "epoch": 2911, "lr": 5.111654861967963e-07}
{"train_loss": 0.06054572016000748, "global_step": 259107, "epoch": 2911, "lr": 5.110828041579252e-07}
{"train_loss": 0.08109056204557419, "global_step": 259108, "epoch": 2911, "lr": 5.110001287722266e-07}
{"train_loss": 0.03339869901537895, "global_step": 259109, "epoch": 2911, "lr": 5.109174600397226e-07}
{"train_loss": 0.03313538059592247, "global_step": 259110, "epoch": 2911, "lr": 5.108347979604245e-07}
{"train_loss": 0.046120233833789825, "global_step": 259111, "epoch": 2911, "lr": 5.107521425343431e-07}
{"train_loss": 0.0707785040140152, "global_step": 259112, "epoch": 2911, "lr": 5.106694937614842e-07}
{"train_loss": 0.02802659384906292, "global_step": 259113, "epoch": 2911, "lr": 5.105868516418644e-07}
{"train_loss": 0.03478725999593735, "global_step": 259114, "epoch": 2911, "lr": 5.105042161754947e-07}
{"train_loss": 0.05504782125353813, "global_step": 259115, "epoch": 2911, "lr": 5.104215873623808e-07}
{"train_loss": 0.05622304603457451, "global_step": 259116, "epoch": 2911, "lr": 5.103389652025392e-07}
{"train_loss": 0.022497547790408134, "global_step": 259117, "epoch": 2911, "lr": 5.102563496959811e-07}
{"train_loss": 0.0974867045879364, "global_step": 259118, "epoch": 2911, "lr": 5.101737408427121e-07}
{"train_loss": 0.059287719428539276, "global_step": 259119, "epoch": 2911, "lr": 5.100911386427488e-07}
{"train_loss": 0.039847955107688904, "global_step": 259120, "epoch": 2911, "lr": 5.100085430961021e-07}
{"train_loss": 0.03456153720617294, "global_step": 259121, "epoch": 2911, "lr": 5.099259542027834e-07}
{"train_loss": 0.03405027091503143, "global_step": 259122, "epoch": 2911, "lr": 5.098433719627927e-07}
{"train_loss": 0.033533088862895966, "global_step": 259123, "epoch": 2911, "lr": 5.097607963761574e-07}
{"train_loss": 0.07661429047584534, "global_step": 259124, "epoch": 2911, "lr": 5.096782274428835e-07}
{"train_loss": 0.08095353841781616, "global_step": 259125, "epoch": 2911, "lr": 5.095956651629763e-07}
{"train_loss": 0.03305111452937126, "global_step": 259126, "epoch": 2911, "lr": 5.095131095364525e-07}
{"train_loss": 0.06154254451394081, "global_step": 259127, "epoch": 2911, "lr": 5.094305605633231e-07}
{"train_loss": 0.021169209852814674, "global_step": 259128, "epoch": 2911, "lr": 5.093480182435939e-07}
{"train_loss": 0.0226760096848011, "global_step": 259129, "epoch": 2911, "lr": 5.092654825772813e-07}
{"train_loss": 0.026899304240942, "global_step": 259130, "epoch": 2911, "lr": 5.091829535643911e-07}
{"train_loss": 0.008320914581418037, "global_step": 259131, "epoch": 2911, "lr": 5.091004312049396e-07}
{"train_loss": 0.028262989595532417, "global_step": 259132, "epoch": 2911, "lr": 5.090179154989383e-07}
{"train_loss": 0.07381626963615417, "global_step": 259133, "epoch": 2911, "lr": 5.08935406446398e-07}
{"train_loss": 0.039159443229436874, "global_step": 259134, "epoch": 2911, "lr": 5.088529040473244e-07}
{"train_loss": 0.06819096207618713, "global_step": 259135, "epoch": 2911, "lr": 5.087704083017341e-07}
{"train_loss": 0.06341248750686646, "global_step": 259136, "epoch": 2911, "lr": 5.086879192096383e-07}
{"train_loss": 0.05270941182971001, "global_step": 259137, "epoch": 2911, "lr": 5.086054367710425e-07}
{"train_loss": 0.03378146514296532, "global_step": 259138, "epoch": 2911, "lr": 5.085229609859632e-07}
{"train_loss": 0.041981540620326996, "global_step": 259139, "epoch": 2911, "lr": 5.084404918544117e-07}
{"train_loss": 0.013374307192862034, "global_step": 259140, "epoch": 2911, "lr": 5.083580293763935e-07}
{"train_loss": 0.02400447055697441, "global_step": 259141, "epoch": 2911, "lr": 5.082755735519252e-07}
{"train_loss": 0.055363915860652924, "global_step": 259142, "epoch": 2911, "lr": 5.081931243810179e-07}
{"train_loss": 0.06006244570016861, "global_step": 259143, "epoch": 2911, "lr": 5.081106818636772e-07}
{"train_loss": 0.08042186498641968, "global_step": 259144, "epoch": 2911, "lr": 5.080282459999198e-07}
{"train_loss": 0.025990040972828865, "global_step": 259145, "epoch": 2911, "lr": 5.079458167897566e-07}
{"train_loss": 0.06497010588645935, "global_step": 259146, "epoch": 2911, "lr": 5.078633942331934e-07}
{"train_loss": 0.01862173341214657, "global_step": 259147, "epoch": 2911, "lr": 5.077809783302468e-07}
{"train_loss": 0.027653872966766357, "global_step": 259148, "epoch": 2911, "lr": 5.076985690809277e-07}
{"train_loss": 0.048078734427690506, "global_step": 259149, "epoch": 2911, "lr": 5.076161664852419e-07}
{"train_loss": 0.06317579746246338, "global_step": 259150, "epoch": 2911, "lr": 5.075337705432059e-07}
{"train_loss": 0.03594820946455002, "global_step": 259151, "epoch": 2911, "lr": 5.074513812548254e-07}
{"train_loss": 0.049040332436561584, "global_step": 259152, "epoch": 2911, "lr": 5.073689986201169e-07}
{"train_loss": 0.024238228797912598, "global_step": 259153, "epoch": 2911, "lr": 5.072866226390915e-07}
{"train_loss": 0.028454162180423737, "global_step": 259154, "epoch": 2911, "lr": 5.072042533117549e-07}
{"train_loss": 0.034408584237098694, "global_step": 259155, "epoch": 2911, "lr": 5.071218906381237e-07}
{"train_loss": 0.029066748917102814, "global_step": 259156, "epoch": 2911, "lr": 5.070395346182033e-07}
{"train_loss": 0.04096002131700516, "global_step": 259157, "epoch": 2911, "lr": 5.069571852520161e-07}
{"train_loss": 0.059908755123615265, "global_step": 259158, "epoch": 2911, "lr": 5.068748425395565e-07}
{"train_loss": 0.04288020730018616, "global_step": 259159, "epoch": 2911, "lr": 5.067925064808465e-07}
{"train_loss": 0.038134168833494186, "global_step": 259160, "epoch": 2911, "lr": 5.067101770758976e-07}
{"train_loss": 0.08715906739234924, "global_step": 259161, "epoch": 2911, "lr": 5.066278543247205e-07}
{"train_loss": 0.03481574356555939, "global_step": 259162, "epoch": 2911, "lr": 5.065455382273154e-07}
{"train_loss": 0.027908682823181152, "global_step": 259163, "epoch": 2911, "lr": 5.064632287837101e-07}
{"train_loss": 0.06311207264661789, "global_step": 259164, "epoch": 2911, "lr": 5.063809259939045e-07}
{"train_loss": 0.02018926665186882, "global_step": 259165, "epoch": 2911, "lr": 5.062986298579097e-07}
{"train_loss": 0.04662301018834114, "global_step": 259166, "epoch": 2911, "lr": 5.062163403757425e-07}
{"train_loss": 0.04506555195437389, "global_step": 259167, "epoch": 2911, "lr": 5.061340575474139e-07, "val_loss": 9.587697982788086}
{"train_loss": 0.028578616678714752, "global_step": 259168, "epoch": 2912, "lr": 5.060517813729293e-07}
{"train_loss": 0.026892155408859253, "global_step": 259169, "epoch": 2912, "lr": 5.059695118523001e-07}
{"train_loss": 0.061815712600946426, "global_step": 259170, "epoch": 2912, "lr": 5.058872489855481e-07}
{"train_loss": 0.037136271595954895, "global_step": 259171, "epoch": 2912, "lr": 5.058049927726682e-07}
{"train_loss": 0.03656560927629471, "global_step": 259172, "epoch": 2912, "lr": 5.05722743213688e-07}
{"train_loss": 0.039028946310281754, "global_step": 259173, "epoch": 2912, "lr": 5.056405003086018e-07}
{"train_loss": 0.028375208377838135, "global_step": 259174, "epoch": 2912, "lr": 5.055582640574319e-07}
{"train_loss": 0.01039156224578619, "global_step": 259175, "epoch": 2912, "lr": 5.054760344601895e-07}
{"train_loss": 0.026172691956162453, "global_step": 259176, "epoch": 2912, "lr": 5.053938115168799e-07}
{"train_loss": 0.03467733412981033, "global_step": 259177, "epoch": 2912, "lr": 5.053115952275145e-07}
{"train_loss": 0.02767111547291279, "global_step": 259178, "epoch": 2912, "lr": 5.052293855921097e-07}
{"train_loss": 0.044054266065359116, "global_step": 259179, "epoch": 2912, "lr": 5.051471826106713e-07}
{"train_loss": 0.0351124107837677, "global_step": 259180, "epoch": 2912, "lr": 5.050649862832157e-07}
{"train_loss": 0.030853694304823875, "global_step": 259181, "epoch": 2912, "lr": 5.049827966097487e-07}
{"train_loss": 0.005767446011304855, "global_step": 259182, "epoch": 2912, "lr": 5.049006135902867e-07}
{"train_loss": 0.06580100208520889, "global_step": 259183, "epoch": 2912, "lr": 5.048184372248355e-07}
{"train_loss": 0.02448447048664093, "global_step": 259184, "epoch": 2912, "lr": 5.04736267513406e-07}
{"train_loss": 0.04185701906681061, "global_step": 259185, "epoch": 2912, "lr": 5.046541044560149e-07}
{"train_loss": 0.05558544024825096, "global_step": 259186, "epoch": 2912, "lr": 5.045719480526679e-07}
{"train_loss": 0.0596696101129055, "global_step": 259187, "epoch": 2912, "lr": 5.044897983033814e-07}
{"train_loss": 0.011187087744474411, "global_step": 259188, "epoch": 2912, "lr": 5.044076552081556e-07}
{"train_loss": 0.040504295378923416, "global_step": 259189, "epoch": 2912, "lr": 5.043255187670182e-07}
{"train_loss": 0.04692120850086212, "global_step": 259190, "epoch": 2912, "lr": 5.042433889799636e-07}
{"train_loss": 0.04164908453822136, "global_step": 259191, "epoch": 2912, "lr": 5.041612658470141e-07}
{"train_loss": 0.05952848866581917, "global_step": 259192, "epoch": 2912, "lr": 5.040791493681751e-07}
{"train_loss": 0.021168405190110207, "global_step": 259193, "epoch": 2912, "lr": 5.039970395434579e-07}
{"train_loss": 0.06236576288938522, "global_step": 259194, "epoch": 2912, "lr": 5.039149363728735e-07}
{"train_loss": 0.123630091547966, "global_step": 259195, "epoch": 2912, "lr": 5.038328398564441e-07}
{"train_loss": 0.05379209667444229, "global_step": 259196, "epoch": 2912, "lr": 5.037507499941585e-07}
{"train_loss": 0.07354956865310669, "global_step": 259197, "epoch": 2912, "lr": 5.036686667860502e-07}
{"train_loss": 0.06287075579166412, "global_step": 259198, "epoch": 2912, "lr": 5.035865902321135e-07}
{"train_loss": 0.05509910732507706, "global_step": 259199, "epoch": 2912, "lr": 5.035045203323708e-07}
{"train_loss": 0.06538856029510498, "global_step": 259200, "epoch": 2912, "lr": 5.034224570868273e-07}
{"train_loss": 0.03876126930117607, "global_step": 259201, "epoch": 2912, "lr": 5.033404004954945e-07}
{"train_loss": 0.0548010878264904, "global_step": 259202, "epoch": 2912, "lr": 5.032583505583888e-07}
{"train_loss": 0.02508672885596752, "global_step": 259203, "epoch": 2912, "lr": 5.031763072755102e-07}
{"train_loss": 0.03421671316027641, "global_step": 259204, "epoch": 2912, "lr": 5.03094270646881e-07}
{"train_loss": 0.024452833458781242, "global_step": 259205, "epoch": 2912, "lr": 5.030122406725013e-07}
{"train_loss": 0.03368030861020088, "global_step": 259206, "epoch": 2912, "lr": 5.029302173523931e-07}
{"train_loss": 0.03850860893726349, "global_step": 259207, "epoch": 2912, "lr": 5.02848200686562e-07}
{"train_loss": 0.030399935320019722, "global_step": 259208, "epoch": 2912, "lr": 5.027661906750247e-07}
{"train_loss": 0.04348968341946602, "global_step": 259209, "epoch": 2912, "lr": 5.026841873177812e-07}
{"train_loss": 0.06556380540132523, "global_step": 259210, "epoch": 2912, "lr": 5.026021906148482e-07}
{"train_loss": 0.06101510673761368, "global_step": 259211, "epoch": 2912, "lr": 5.025202005662366e-07}
{"train_loss": 0.025148678570985794, "global_step": 259212, "epoch": 2912, "lr": 5.024382171719632e-07}
{"train_loss": 0.033328138291835785, "global_step": 259213, "epoch": 2912, "lr": 5.023562404320281e-07}
{"train_loss": 0.09704457223415375, "global_step": 259214, "epoch": 2912, "lr": 5.022742703464534e-07}
{"train_loss": 0.06410106271505356, "global_step": 259215, "epoch": 2912, "lr": 5.02192306915239e-07}
{"train_loss": 0.018249299377202988, "global_step": 259216, "epoch": 2912, "lr": 5.021103501384073e-07}
{"train_loss": 0.035206347703933716, "global_step": 259217, "epoch": 2912, "lr": 5.020284000159581e-07}
{"train_loss": 0.031641945242881775, "global_step": 259218, "epoch": 2912, "lr": 5.019464565479082e-07}
{"train_loss": 0.05655137076973915, "global_step": 259219, "epoch": 2912, "lr": 5.018645197342687e-07}
{"train_loss": 0.030592158436775208, "global_step": 259220, "epoch": 2912, "lr": 5.017825895750561e-07}
{"train_loss": 0.014536651782691479, "global_step": 259221, "epoch": 2912, "lr": 5.01700666070265e-07}
{"train_loss": 0.026633767411112785, "global_step": 259222, "epoch": 2912, "lr": 5.016187492199286e-07}
{"train_loss": 0.049832310527563095, "global_step": 259223, "epoch": 2912, "lr": 5.015368390240415e-07}
{"train_loss": 0.03562253713607788, "global_step": 259224, "epoch": 2912, "lr": 5.014549354826148e-07}
{"train_loss": 0.0592644177377224, "global_step": 259225, "epoch": 2912, "lr": 5.013730385956705e-07}
{"train_loss": 0.03965497761964798, "global_step": 259226, "epoch": 2912, "lr": 5.012911483632088e-07}
{"train_loss": 0.04587964341044426, "global_step": 259227, "epoch": 2912, "lr": 5.012092647852517e-07}
{"train_loss": 0.054333701729774475, "global_step": 259228, "epoch": 2912, "lr": 5.01127387861794e-07}
{"train_loss": 0.08381693810224533, "global_step": 259229, "epoch": 2912, "lr": 5.01045517592863e-07}
{"train_loss": 0.057872697710990906, "global_step": 259230, "epoch": 2912, "lr": 5.009636539784645e-07}
{"train_loss": 0.07451291382312775, "global_step": 259231, "epoch": 2912, "lr": 5.008817970186042e-07}
{"train_loss": 0.05466718226671219, "global_step": 259232, "epoch": 2912, "lr": 5.007999467132985e-07}
{"train_loss": 0.05197571963071823, "global_step": 259233, "epoch": 2912, "lr": 5.007181030625585e-07}
{"train_loss": 0.03523896634578705, "global_step": 259234, "epoch": 2912, "lr": 5.006362660663899e-07}
{"train_loss": 0.06130572408437729, "global_step": 259235, "epoch": 2912, "lr": 5.005544357248093e-07}
{"train_loss": 0.03873533383011818, "global_step": 259236, "epoch": 2912, "lr": 5.004726120378222e-07}
{"train_loss": 0.02117580734193325, "global_step": 259237, "epoch": 2912, "lr": 5.003907950054509e-07}
{"train_loss": 0.008127287030220032, "global_step": 259238, "epoch": 2912, "lr": 5.003089846276953e-07}
{"train_loss": 0.036566585302352905, "global_step": 259239, "epoch": 2912, "lr": 5.002271809045666e-07}
{"train_loss": 0.06013686582446098, "global_step": 259240, "epoch": 2912, "lr": 5.001453838360814e-07}
{"train_loss": 0.05493205413222313, "global_step": 259241, "epoch": 2912, "lr": 5.000635934222452e-07}
{"train_loss": 0.033553462475538254, "global_step": 259242, "epoch": 2912, "lr": 4.999818096630804e-07}
{"train_loss": 0.043579667806625366, "global_step": 259243, "epoch": 2912, "lr": 4.999000325585812e-07}
{"train_loss": 0.05758701637387276, "global_step": 259244, "epoch": 2912, "lr": 4.998182621087755e-07}
{"train_loss": 0.022864369675517082, "global_step": 259245, "epoch": 2912, "lr": 4.997364983136577e-07}
{"train_loss": 0.03920547291636467, "global_step": 259246, "epoch": 2912, "lr": 4.9965474117325e-07}
{"train_loss": 0.09911108762025833, "global_step": 259247, "epoch": 2912, "lr": 4.995729906875634e-07}
{"train_loss": 0.08585318922996521, "global_step": 259248, "epoch": 2912, "lr": 4.994912468566037e-07}
{"train_loss": 0.10757957398891449, "global_step": 259249, "epoch": 2912, "lr": 4.994095096803819e-07}
{"train_loss": 0.051987674087285995, "global_step": 259250, "epoch": 2912, "lr": 4.993277791589146e-07}
{"train_loss": 0.05634526163339615, "global_step": 259251, "epoch": 2912, "lr": 4.992460552922073e-07}
{"train_loss": 0.03739987313747406, "global_step": 259252, "epoch": 2912, "lr": 4.991643380802769e-07}
{"train_loss": 0.05451836436986923, "global_step": 259253, "epoch": 2912, "lr": 4.990826275231231e-07}
{"train_loss": 0.05141497775912285, "global_step": 259254, "epoch": 2912, "lr": 4.990009236207682e-07}
{"train_loss": 0.03465311601758003, "global_step": 259255, "epoch": 2912, "lr": 4.989192263732178e-07}
{"train_loss": 0.04621985669802414, "global_step": 259256, "epoch": 2912, "lr": 4.988375357804887e-07, "val_loss": 9.617505073547363}
{"train_loss": 0.041087958961725235, "global_step": 259257, "epoch": 2913, "lr": 4.987558518425861e-07}
{"train_loss": 0.06403696537017822, "global_step": 259258, "epoch": 2913, "lr": 4.986741745595214e-07}
{"train_loss": 0.0391840785741806, "global_step": 259259, "epoch": 2913, "lr": 4.985925039313055e-07}
{"train_loss": 0.0371164008975029, "global_step": 259260, "epoch": 2913, "lr": 4.985108399579497e-07}
{"train_loss": 0.035117391496896744, "global_step": 259261, "epoch": 2913, "lr": 4.984291826394704e-07}
{"train_loss": 0.03852759674191475, "global_step": 259262, "epoch": 2913, "lr": 4.983475319758679e-07}
{"train_loss": 0.07158738374710083, "global_step": 259263, "epoch": 2913, "lr": 4.982658879671642e-07}
{"train_loss": 0.03167387843132019, "global_step": 259264, "epoch": 2913, "lr": 4.981842506133649e-07}
{"train_loss": 0.044518809765577316, "global_step": 259265, "epoch": 2913, "lr": 4.981026199144811e-07}
{"train_loss": 0.015218231827020645, "global_step": 259266, "epoch": 2913, "lr": 4.980209958705184e-07}
{"train_loss": 0.04561636969447136, "global_step": 259267, "epoch": 2913, "lr": 4.979393784815045e-07}
{"train_loss": 0.027091117575764656, "global_step": 259268, "epoch": 2913, "lr": 4.978577677474283e-07}
{"train_loss": 0.022174589335918427, "global_step": 259269, "epoch": 2913, "lr": 4.977761636683176e-07}
{"train_loss": 0.13342389464378357, "global_step": 259270, "epoch": 2913, "lr": 4.976945662441779e-07}
{"train_loss": 0.05243869498372078, "global_step": 259271, "epoch": 2913, "lr": 4.976129754750203e-07}
{"train_loss": 0.044355690479278564, "global_step": 259272, "epoch": 2913, "lr": 4.975313913608504e-07}
{"train_loss": 0.03514597564935684, "global_step": 259273, "epoch": 2913, "lr": 4.974498139016847e-07}
{"train_loss": 0.036520566791296005, "global_step": 259274, "epoch": 2913, "lr": 4.973682430975346e-07}
{"train_loss": 0.015559431165456772, "global_step": 259275, "epoch": 2913, "lr": 4.97286678948411e-07}
{"train_loss": 0.039933942258358, "global_step": 259276, "epoch": 2913, "lr": 4.97205121454325e-07}
{"train_loss": 0.02867499180138111, "global_step": 259277, "epoch": 2913, "lr": 4.971235706152821e-07}
{"train_loss": 0.04519634693861008, "global_step": 259278, "epoch": 2913, "lr": 4.970420264313047e-07}
{"train_loss": 0.059339478611946106, "global_step": 259279, "epoch": 2913, "lr": 4.969604889023927e-07}
{"train_loss": 0.037258658558130264, "global_step": 259280, "epoch": 2913, "lr": 4.968789580285627e-07}
{"train_loss": 0.024251917377114296, "global_step": 259281, "epoch": 2913, "lr": 4.967974338098202e-07}
{"train_loss": 0.04388369247317314, "global_step": 259282, "epoch": 2913, "lr": 4.96715916246182e-07}
{"train_loss": 0.011171579360961914, "global_step": 259283, "epoch": 2913, "lr": 4.966344053376537e-07}
{"train_loss": 0.01989986002445221, "global_step": 259284, "epoch": 2913, "lr": 4.965529010842573e-07}
{"train_loss": 0.017816508188843727, "global_step": 259285, "epoch": 2913, "lr": 4.964714034859874e-07}
{"train_loss": 0.032330870628356934, "global_step": 259286, "epoch": 2913, "lr": 4.963899125428717e-07}
{"train_loss": 0.029378382489085197, "global_step": 259287, "epoch": 2913, "lr": 4.963084282549046e-07}
{"train_loss": 0.0392957367002964, "global_step": 259288, "epoch": 2913, "lr": 4.96226950622114e-07}
{"train_loss": 0.06323372572660446, "global_step": 259289, "epoch": 2913, "lr": 4.961454796444942e-07}
{"train_loss": 0.0751013234257698, "global_step": 259290, "epoch": 2913, "lr": 4.960640153220731e-07}
{"train_loss": 0.050465572625398636, "global_step": 259291, "epoch": 2913, "lr": 4.959825576548449e-07}
{"train_loss": 0.02384122833609581, "global_step": 259292, "epoch": 2913, "lr": 4.95901106642832e-07}
{"train_loss": 0.038181696087121964, "global_step": 259293, "epoch": 2913, "lr": 4.9581966228604e-07}
{"train_loss": 0.05318649485707283, "global_step": 259294, "epoch": 2913, "lr": 4.957382245844855e-07}
{"train_loss": 0.02220577374100685, "global_step": 259295, "epoch": 2913, "lr": 4.95656793538174e-07}
{"train_loss": 0.030467074364423752, "global_step": 259296, "epoch": 2913, "lr": 4.955753691471166e-07}
{"train_loss": 0.045769426971673965, "global_step": 259297, "epoch": 2913, "lr": 4.954939514113244e-07}
{"train_loss": 0.07023640722036362, "global_step": 259298, "epoch": 2913, "lr": 4.95412540330814e-07}
{"train_loss": 0.012865914031863213, "global_step": 259299, "epoch": 2913, "lr": 4.953311359055857e-07}
{"train_loss": 0.05686540529131889, "global_step": 259300, "epoch": 2913, "lr": 4.952497381356669e-07}
{"train_loss": 0.04657710716128349, "global_step": 259301, "epoch": 2913, "lr": 4.951683470210521e-07}
{"train_loss": 0.05281988903880119, "global_step": 259302, "epoch": 2913, "lr": 4.950869625617582e-07}
{"train_loss": 0.04757985100150108, "global_step": 259303, "epoch": 2913, "lr": 4.950055847577962e-07}
{"train_loss": 0.037152595818042755, "global_step": 259304, "epoch": 2913, "lr": 4.94924213609177e-07}
{"train_loss": 0.023479949682950974, "global_step": 259305, "epoch": 2913, "lr": 4.948428491159118e-07}
{"train_loss": 0.011600579135119915, "global_step": 259306, "epoch": 2913, "lr": 4.947614912780119e-07}
{"train_loss": 0.039108388125896454, "global_step": 259307, "epoch": 2913, "lr": 4.946801400954937e-07}
{"train_loss": 0.04687357321381569, "global_step": 259308, "epoch": 2913, "lr": 4.945987955683518e-07}
{"train_loss": 0.0322117917239666, "global_step": 259309, "epoch": 2913, "lr": 4.945174576966194e-07}
{"train_loss": 0.0568920373916626, "global_step": 259310, "epoch": 2913, "lr": 4.944361264802855e-07}
{"train_loss": 0.060056380927562714, "global_step": 259311, "epoch": 2913, "lr": 4.943548019193777e-07}
{"train_loss": 0.05062349885702133, "global_step": 259312, "epoch": 2913, "lr": 4.942734840138962e-07}
{"train_loss": 0.03508290275931358, "global_step": 259313, "epoch": 2913, "lr": 4.941921727638576e-07}
{"train_loss": 0.04616333544254303, "global_step": 259314, "epoch": 2913, "lr": 4.94110868169273e-07}
{"train_loss": 0.054167285561561584, "global_step": 259315, "epoch": 2913, "lr": 4.940295702301478e-07}
{"train_loss": 0.07956834137439728, "global_step": 259316, "epoch": 2913, "lr": 4.939482789465044e-07}
{"train_loss": 0.03179201856255531, "global_step": 259317, "epoch": 2913, "lr": 4.938669943183372e-07}
{"train_loss": 0.04551297426223755, "global_step": 259318, "epoch": 2913, "lr": 4.937857163456738e-07}
{"train_loss": 0.024081364274024963, "global_step": 259319, "epoch": 2913, "lr": 4.937044450285145e-07}
{"train_loss": 0.03662079945206642, "global_step": 259320, "epoch": 2913, "lr": 4.936231803668756e-07}
{"train_loss": 0.06916496157646179, "global_step": 259321, "epoch": 2913, "lr": 4.935419223607629e-07}
{"train_loss": 0.05301539972424507, "global_step": 259322, "epoch": 2913, "lr": 4.934606710101874e-07}
{"train_loss": 0.016984030604362488, "global_step": 259323, "epoch": 2913, "lr": 4.933794263151658e-07}
{"train_loss": 0.051169659942388535, "global_step": 259324, "epoch": 2913, "lr": 4.932981882757093e-07}
{"train_loss": 0.05647336319088936, "global_step": 259325, "epoch": 2913, "lr": 4.932169568918177e-07}
{"train_loss": 0.03967368230223656, "global_step": 259326, "epoch": 2913, "lr": 4.931357321635189e-07}
{"train_loss": 0.027635864913463593, "global_step": 259327, "epoch": 2913, "lr": 4.930545140908072e-07}
{"train_loss": 0.020768163725733757, "global_step": 259328, "epoch": 2913, "lr": 4.929733026737049e-07}
{"train_loss": 0.059843480587005615, "global_step": 259329, "epoch": 2913, "lr": 4.928920979122175e-07}
{"train_loss": 0.02678786963224411, "global_step": 259330, "epoch": 2913, "lr": 4.928108998063563e-07}
{"train_loss": 0.024439053609967232, "global_step": 259331, "epoch": 2913, "lr": 4.927297083561322e-07}
{"train_loss": 0.04444878548383713, "global_step": 259332, "epoch": 2913, "lr": 4.926485235615619e-07}
{"train_loss": 0.04775954782962799, "global_step": 259333, "epoch": 2913, "lr": 4.925673454226509e-07}
{"train_loss": 0.047662246972322464, "global_step": 259334, "epoch": 2913, "lr": 4.924861739394104e-07}
{"train_loss": 0.041080113500356674, "global_step": 259335, "epoch": 2913, "lr": 4.924050091118459e-07}
{"train_loss": 0.03655552119016647, "global_step": 259336, "epoch": 2913, "lr": 4.923238509399797e-07}
{"train_loss": 0.04519088938832283, "global_step": 259337, "epoch": 2913, "lr": 4.922426994238172e-07}
{"train_loss": 0.1050826758146286, "global_step": 259338, "epoch": 2913, "lr": 4.921615545633696e-07}
{"train_loss": 0.04844706505537033, "global_step": 259339, "epoch": 2913, "lr": 4.920804163586479e-07}
{"train_loss": 0.038904886692762375, "global_step": 259340, "epoch": 2913, "lr": 4.919992848096578e-07}
{"train_loss": 0.020884346216917038, "global_step": 259341, "epoch": 2913, "lr": 4.919181599164213e-07}
{"train_loss": 0.03961434215307236, "global_step": 259342, "epoch": 2913, "lr": 4.918370416789331e-07}
{"train_loss": 0.03164224326610565, "global_step": 259343, "epoch": 2913, "lr": 4.917559300972263e-07}
{"train_loss": 0.02884467877447605, "global_step": 259344, "epoch": 2913, "lr": 4.916748251712899e-07}
{"train_loss": 0.04172088047803453, "global_step": 259345, "epoch": 2913, "lr": 4.915937269011517e-07, "val_loss": 9.667055130004883}
{"train_loss": 0.04314336180686951, "global_step": 259346, "epoch": 2914, "lr": 4.915126352868061e-07}
{"train_loss": 0.07681075483560562, "global_step": 259347, "epoch": 2914, "lr": 4.914315503282807e-07}
{"train_loss": 0.01881476491689682, "global_step": 259348, "epoch": 2914, "lr": 4.913504720255757e-07}
{"train_loss": 0.03212737664580345, "global_step": 259349, "epoch": 2914, "lr": 4.912694003787077e-07}
{"train_loss": 0.023204464465379715, "global_step": 259350, "epoch": 2914, "lr": 4.911883353876823e-07}
{"train_loss": 0.05697763338685036, "global_step": 259351, "epoch": 2914, "lr": 4.91107277052516e-07}
{"train_loss": 0.03290729969739914, "global_step": 259352, "epoch": 2914, "lr": 4.910262253732146e-07}
{"train_loss": 0.0489242859184742, "global_step": 259353, "epoch": 2914, "lr": 4.909451803497945e-07}
{"train_loss": 0.03754422441124916, "global_step": 259354, "epoch": 2914, "lr": 4.908641419822613e-07}
{"train_loss": 0.052541524171829224, "global_step": 259355, "epoch": 2914, "lr": 4.907831102706261e-07}
{"train_loss": 0.02853042259812355, "global_step": 259356, "epoch": 2914, "lr": 4.907020852149058e-07}
{"train_loss": 0.022725725546479225, "global_step": 259357, "epoch": 2914, "lr": 4.906210668151001e-07}
{"train_loss": 0.014075867831707, "global_step": 259358, "epoch": 2914, "lr": 4.905400550712369e-07}
{"train_loss": 0.016687646508216858, "global_step": 259359, "epoch": 2914, "lr": 4.904590499833106e-07}
{"train_loss": 0.05063928663730621, "global_step": 259360, "epoch": 2914, "lr": 4.903780515513379e-07}
{"train_loss": 0.05052569881081581, "global_step": 259361, "epoch": 2914, "lr": 4.902970597753353e-07}
{"train_loss": 0.05092266947031021, "global_step": 259362, "epoch": 2914, "lr": 4.902160746553086e-07}
{"train_loss": 0.04080726206302643, "global_step": 259363, "epoch": 2914, "lr": 4.901350961912632e-07}
{"train_loss": 0.04265975207090378, "global_step": 259364, "epoch": 2914, "lr": 4.900541243832213e-07}
{"train_loss": 0.05631731078028679, "global_step": 259365, "epoch": 2914, "lr": 4.899731592311829e-07}
{"train_loss": 0.02184966206550598, "global_step": 259366, "epoch": 2914, "lr": 4.898922007351647e-07}
{"train_loss": 0.04722558334469795, "global_step": 259367, "epoch": 2914, "lr": 4.898112488951834e-07}
{"train_loss": 0.055955830961465836, "global_step": 259368, "epoch": 2914, "lr": 4.897303037112333e-07}
{"train_loss": 0.029163232073187828, "global_step": 259369, "epoch": 2914, "lr": 4.896493651833478e-07}
{"train_loss": 0.035704728215932846, "global_step": 259370, "epoch": 2914, "lr": 4.895684333115158e-07}
{"train_loss": 0.04456270858645439, "global_step": 259371, "epoch": 2914, "lr": 4.89487508095765e-07}
{"train_loss": 0.03607246279716492, "global_step": 259372, "epoch": 2914, "lr": 4.894065895360955e-07}
{"train_loss": 0.041639670729637146, "global_step": 259373, "epoch": 2914, "lr": 4.893256776325183e-07}
{"train_loss": 0.05969108268618584, "global_step": 259374, "epoch": 2914, "lr": 4.892447723850557e-07}
{"train_loss": 0.06858547776937485, "global_step": 259375, "epoch": 2914, "lr": 4.891638737937021e-07}
{"train_loss": 0.09489333629608154, "global_step": 259376, "epoch": 2914, "lr": 4.890829818584853e-07}
{"train_loss": 0.019204622134566307, "global_step": 259377, "epoch": 2914, "lr": 4.890020965794051e-07}
{"train_loss": 0.08959677070379257, "global_step": 259378, "epoch": 2914, "lr": 4.889212179564729e-07}
{"train_loss": 0.05115972459316254, "global_step": 259379, "epoch": 2914, "lr": 4.888403459897051e-07}
{"train_loss": 0.09168308973312378, "global_step": 259380, "epoch": 2914, "lr": 4.887594806791074e-07}
{"train_loss": 0.011975801549851894, "global_step": 259381, "epoch": 2914, "lr": 4.886786220246908e-07}
{"train_loss": 0.04900617152452469, "global_step": 259382, "epoch": 2914, "lr": 4.885977700264665e-07}
{"train_loss": 0.08299697935581207, "global_step": 259383, "epoch": 2914, "lr": 4.885169246844512e-07}
{"train_loss": 0.043428126722574234, "global_step": 259384, "epoch": 2914, "lr": 4.884360859986503e-07}
{"train_loss": 0.11545582115650177, "global_step": 259385, "epoch": 2914, "lr": 4.88355253969075e-07}
{"train_loss": 0.013177263550460339, "global_step": 259386, "epoch": 2914, "lr": 4.882744285957363e-07}
{"train_loss": 0.04511038959026337, "global_step": 259387, "epoch": 2914, "lr": 4.881936098786455e-07}
{"train_loss": 0.041449304670095444, "global_step": 259388, "epoch": 2914, "lr": 4.881127978178135e-07}
{"train_loss": 0.03147372975945473, "global_step": 259389, "epoch": 2914, "lr": 4.880319924132515e-07}
{"train_loss": 0.03679225221276283, "global_step": 259390, "epoch": 2914, "lr": 4.879511936649705e-07}
{"train_loss": 0.04390192776918411, "global_step": 259391, "epoch": 2914, "lr": 4.878704015729818e-07}
{"train_loss": 0.009279143996536732, "global_step": 259392, "epoch": 2914, "lr": 4.877896161372964e-07}
{"train_loss": 0.04936258867383003, "global_step": 259393, "epoch": 2914, "lr": 4.877088373579197e-07}
{"train_loss": 0.04954872652888298, "global_step": 259394, "epoch": 2914, "lr": 4.876280652348741e-07}
{"train_loss": 0.042491115629673004, "global_step": 259395, "epoch": 2914, "lr": 4.875472997681596e-07}
{"train_loss": 0.03187412768602371, "global_step": 259396, "epoch": 2914, "lr": 4.874665409577872e-07}
{"train_loss": 0.025384003296494484, "global_step": 259397, "epoch": 2914, "lr": 4.873857888037736e-07}
{"train_loss": 0.046772174537181854, "global_step": 259398, "epoch": 2914, "lr": 4.873050433061354e-07}
{"train_loss": 0.04382263123989105, "global_step": 259399, "epoch": 2914, "lr": 4.872243044648672e-07}
{"train_loss": 0.040745433419942856, "global_step": 259400, "epoch": 2914, "lr": 4.871435722799911e-07}
{"train_loss": 0.05383182317018509, "global_step": 259401, "epoch": 2914, "lr": 4.870628467515126e-07}
{"train_loss": 0.10499384254217148, "global_step": 259402, "epoch": 2914, "lr": 4.869821278794429e-07}
{"train_loss": 0.059705957770347595, "global_step": 259403, "epoch": 2914, "lr": 4.869014156637985e-07}
{"train_loss": 0.055392563343048096, "global_step": 259404, "epoch": 2914, "lr": 4.868207101045852e-07}
{"train_loss": 0.038911957293748856, "global_step": 259405, "epoch": 2914, "lr": 4.867400112018195e-07}
{"train_loss": 0.040099792182445526, "global_step": 259406, "epoch": 2914, "lr": 4.866593189555069e-07}
{"train_loss": 0.04432729631662369, "global_step": 259407, "epoch": 2914, "lr": 4.865786333656585e-07}
{"train_loss": 0.04898218810558319, "global_step": 259408, "epoch": 2914, "lr": 4.864979544322856e-07}
{"train_loss": 0.06501726061105728, "global_step": 259409, "epoch": 2914, "lr": 4.864172821553991e-07}
{"train_loss": 0.049355387687683105, "global_step": 259410, "epoch": 2914, "lr": 4.863366165350103e-07}
{"train_loss": 0.02166132628917694, "global_step": 259411, "epoch": 2914, "lr": 4.862559575711301e-07}
{"train_loss": 0.04618360102176666, "global_step": 259412, "epoch": 2914, "lr": 4.861753052637697e-07}
{"train_loss": 0.04739001765847206, "global_step": 259413, "epoch": 2914, "lr": 4.860946596129401e-07}
{"train_loss": 0.022310325875878334, "global_step": 259414, "epoch": 2914, "lr": 4.860140206186526e-07}
{"train_loss": 0.056466296315193176, "global_step": 259415, "epoch": 2914, "lr": 4.859333882809125e-07}
{"train_loss": 0.07215362787246704, "global_step": 259416, "epoch": 2914, "lr": 4.858527625997422e-07}
{"train_loss": 0.07769924402236938, "global_step": 259417, "epoch": 2914, "lr": 4.857721435751417e-07}
{"train_loss": 0.028777463361620903, "global_step": 259418, "epoch": 2914, "lr": 4.856915312071276e-07}
{"train_loss": 0.032352618873119354, "global_step": 259419, "epoch": 2914, "lr": 4.856109254957053e-07}
{"train_loss": 0.05997130647301674, "global_step": 259420, "epoch": 2914, "lr": 4.855303264408917e-07}
{"train_loss": 0.029063016176223755, "global_step": 259421, "epoch": 2914, "lr": 4.854497340426922e-07}
{"train_loss": 0.05629856511950493, "global_step": 259422, "epoch": 2914, "lr": 4.853691483011235e-07}
{"train_loss": 0.03324894234538078, "global_step": 259423, "epoch": 2914, "lr": 4.852885692161913e-07}
{"train_loss": 0.040854815393686295, "global_step": 259424, "epoch": 2914, "lr": 4.852079967879064e-07}
{"train_loss": 0.013818064704537392, "global_step": 259425, "epoch": 2914, "lr": 4.851274310162857e-07}
{"train_loss": 0.06534209102392197, "global_step": 259426, "epoch": 2914, "lr": 4.850468719013345e-07}
{"train_loss": 0.02613738924264908, "global_step": 259427, "epoch": 2914, "lr": 4.849663194430642e-07}
{"train_loss": 0.03994807228446007, "global_step": 259428, "epoch": 2914, "lr": 4.848857736414914e-07}
{"train_loss": 0.03717052564024925, "global_step": 259429, "epoch": 2914, "lr": 4.848052344966159e-07}
{"train_loss": 0.043684106320142746, "global_step": 259430, "epoch": 2914, "lr": 4.8472470200846e-07}
{"train_loss": 0.06036169454455376, "global_step": 259431, "epoch": 2914, "lr": 4.846441761770293e-07}
{"train_loss": 0.054158274084329605, "global_step": 259432, "epoch": 2914, "lr": 4.845636570023348e-07}
{"train_loss": 0.04096014052629471, "global_step": 259433, "epoch": 2914, "lr": 4.844831444843823e-07}
{"train_loss": 0.04554665139928628, "global_step": 259434, "epoch": 2914, "lr": 4.844026386231937e-07, "val_loss": 9.585118293762207}
{"train_loss": 0.03512874245643616, "global_step": 259435, "epoch": 2915, "lr": 4.843221394187692e-07}
{"train_loss": 0.022011905908584595, "global_step": 259436, "epoch": 2915, "lr": 4.842416468711253e-07}
{"train_loss": 0.06559999287128448, "global_step": 259437, "epoch": 2915, "lr": 4.841611609802676e-07}
{"train_loss": 0.022515127435326576, "global_step": 259438, "epoch": 2915, "lr": 4.840806817462184e-07}
{"train_loss": 0.08244149386882782, "global_step": 259439, "epoch": 2915, "lr": 4.840002091689777e-07}
{"train_loss": 0.04796193912625313, "global_step": 259440, "epoch": 2915, "lr": 4.839197432485565e-07}
{"train_loss": 0.014433836564421654, "global_step": 259441, "epoch": 2915, "lr": 4.838392839849714e-07}
{"train_loss": 0.05418840050697327, "global_step": 259442, "epoch": 2915, "lr": 4.837588313782281e-07}
{"train_loss": 0.029522866010665894, "global_step": 259443, "epoch": 2915, "lr": 4.836783854283489e-07}
{"train_loss": 0.04152876138687134, "global_step": 259444, "epoch": 2915, "lr": 4.835979461353279e-07}
{"train_loss": 0.008135448209941387, "global_step": 259445, "epoch": 2915, "lr": 4.83517513499182e-07}
{"train_loss": 0.05216863751411438, "global_step": 259446, "epoch": 2915, "lr": 4.834370875199278e-07}
{"train_loss": 0.02214088663458824, "global_step": 259447, "epoch": 2915, "lr": 4.83356668197571e-07}
{"train_loss": 0.048721663653850555, "global_step": 259448, "epoch": 2915, "lr": 4.832762555321224e-07}
{"train_loss": 0.07325813174247742, "global_step": 259449, "epoch": 2915, "lr": 4.831958495235989e-07}
{"train_loss": 0.046080105006694794, "global_step": 259450, "epoch": 2915, "lr": 4.831154501720003e-07}
{"train_loss": 0.04883747175335884, "global_step": 259451, "epoch": 2915, "lr": 4.830350574773435e-07}
{"train_loss": 0.04625475034117699, "global_step": 259452, "epoch": 2915, "lr": 4.829546714396449e-07}
{"train_loss": 0.10135022550821304, "global_step": 259453, "epoch": 2915, "lr": 4.828742920589047e-07}
{"train_loss": 0.035158321261405945, "global_step": 259454, "epoch": 2915, "lr": 4.827939193351394e-07}
{"train_loss": 0.07615412771701813, "global_step": 259455, "epoch": 2915, "lr": 4.827135532683602e-07}
{"train_loss": 0.04562617093324661, "global_step": 259456, "epoch": 2915, "lr": 4.826331938585727e-07}
{"train_loss": 0.032315272837877274, "global_step": 259457, "epoch": 2915, "lr": 4.825528411057933e-07}
{"train_loss": 0.021074572578072548, "global_step": 259458, "epoch": 2915, "lr": 4.824724950100334e-07}
{"train_loss": 0.07110903412103653, "global_step": 259459, "epoch": 2915, "lr": 4.823921555712985e-07}
{"train_loss": 0.08083251118659973, "global_step": 259460, "epoch": 2915, "lr": 4.823118227896051e-07}
{"train_loss": 0.024082493036985397, "global_step": 259461, "epoch": 2915, "lr": 4.822314966649588e-07}
{"train_loss": 0.02976004220545292, "global_step": 259462, "epoch": 2915, "lr": 4.821511771973763e-07}
{"train_loss": 0.055485427379608154, "global_step": 259463, "epoch": 2915, "lr": 4.820708643868632e-07}
{"train_loss": 0.028144346550107002, "global_step": 259464, "epoch": 2915, "lr": 4.819905582334306e-07}
{"train_loss": 0.06149845942854881, "global_step": 259465, "epoch": 2915, "lr": 4.819102587370894e-07}
{"train_loss": 0.022630810737609863, "global_step": 259466, "epoch": 2915, "lr": 4.818299658978565e-07}
{"train_loss": 0.03154098242521286, "global_step": 259467, "epoch": 2915, "lr": 4.817496797157372e-07}
{"train_loss": 0.016758032143115997, "global_step": 259468, "epoch": 2915, "lr": 4.816694001907429e-07}
{"train_loss": 0.0373520664870739, "global_step": 259469, "epoch": 2915, "lr": 4.815891273228845e-07}
{"train_loss": 0.03216348588466644, "global_step": 259470, "epoch": 2915, "lr": 4.815088611121732e-07}
{"train_loss": 0.07470419257879257, "global_step": 259471, "epoch": 2915, "lr": 4.814286015586145e-07}
{"train_loss": 0.031019942834973335, "global_step": 259472, "epoch": 2915, "lr": 4.813483486622306e-07}
{"train_loss": 0.06517760455608368, "global_step": 259473, "epoch": 2915, "lr": 4.812681024230215e-07}
{"train_loss": 0.036277513951063156, "global_step": 259474, "epoch": 2915, "lr": 4.811878628410038e-07}
{"train_loss": 0.007194928824901581, "global_step": 259475, "epoch": 2915, "lr": 4.811076299161887e-07}
{"train_loss": 0.0757569670677185, "global_step": 259476, "epoch": 2915, "lr": 4.810274036485818e-07}
{"train_loss": 0.04918086156249046, "global_step": 259477, "epoch": 2915, "lr": 4.809471840381996e-07}
{"train_loss": 0.030287861824035645, "global_step": 259478, "epoch": 2915, "lr": 4.808669710850478e-07}
{"train_loss": 0.05198263004422188, "global_step": 259479, "epoch": 2915, "lr": 4.807867647891429e-07}
{"train_loss": 0.0639588013291359, "global_step": 259480, "epoch": 2915, "lr": 4.807065651504905e-07}
{"train_loss": 0.03007686324417591, "global_step": 259481, "epoch": 2915, "lr": 4.806263721691073e-07}
{"train_loss": 0.04215819761157036, "global_step": 259482, "epoch": 2915, "lr": 4.805461858449989e-07}
{"train_loss": 0.01908065378665924, "global_step": 259483, "epoch": 2915, "lr": 4.804660061781763e-07}
{"train_loss": 0.03981906920671463, "global_step": 259484, "epoch": 2915, "lr": 4.803858331686506e-07}
{"train_loss": 0.05536117032170296, "global_step": 259485, "epoch": 2915, "lr": 4.803056668164385e-07}
{"train_loss": 0.06874478608369827, "global_step": 259486, "epoch": 2915, "lr": 4.8022550712154e-07}
{"train_loss": 0.03952968120574951, "global_step": 259487, "epoch": 2915, "lr": 4.801453540839717e-07}
{"train_loss": 0.05829727277159691, "global_step": 259488, "epoch": 2915, "lr": 4.800652077037504e-07}
{"train_loss": 0.057584475725889206, "global_step": 259489, "epoch": 2915, "lr": 4.799850679808759e-07}
{"train_loss": 0.048128243535757065, "global_step": 259490, "epoch": 2915, "lr": 4.799049349153651e-07}
{"train_loss": 0.05401519685983658, "global_step": 259491, "epoch": 2915, "lr": 4.798248085072232e-07}
{"train_loss": 0.06504206359386444, "global_step": 259492, "epoch": 2915, "lr": 4.797446887564727e-07}
{"train_loss": 0.04933594539761543, "global_step": 259493, "epoch": 2915, "lr": 4.79664575663108e-07}
{"train_loss": 0.058071259409189224, "global_step": 259494, "epoch": 2915, "lr": 4.795844692271567e-07}
{"train_loss": 0.06593603640794754, "global_step": 259495, "epoch": 2915, "lr": 4.79504369448619e-07}
{"train_loss": 0.054992347955703735, "global_step": 259496, "epoch": 2915, "lr": 4.794242763275114e-07}
{"train_loss": 0.056007836014032364, "global_step": 259497, "epoch": 2915, "lr": 4.79344189863834e-07}
{"train_loss": 0.036229804158210754, "global_step": 259498, "epoch": 2915, "lr": 4.792641100576145e-07}
{"train_loss": 0.049547262489795685, "global_step": 259499, "epoch": 2915, "lr": 4.791840369088474e-07}
{"train_loss": 0.04127166420221329, "global_step": 259500, "epoch": 2915, "lr": 4.791039704175548e-07}
{"train_loss": 0.06076209992170334, "global_step": 259501, "epoch": 2915, "lr": 4.790239105837369e-07}
{"train_loss": 0.04002055153250694, "global_step": 259502, "epoch": 2915, "lr": 4.789438574074157e-07}
{"train_loss": 0.05761272832751274, "global_step": 259503, "epoch": 2915, "lr": 4.788638108885967e-07}
{"train_loss": 0.05577501282095909, "global_step": 259504, "epoch": 2915, "lr": 4.787837710272914e-07}
{"train_loss": 0.03260203078389168, "global_step": 259505, "epoch": 2915, "lr": 4.787037378235049e-07}
{"train_loss": 0.04506789520382881, "global_step": 259506, "epoch": 2915, "lr": 4.786237112772596e-07}
{"train_loss": 0.03474580869078636, "global_step": 259507, "epoch": 2915, "lr": 4.785436913885554e-07}
{"train_loss": 0.08829447627067566, "global_step": 259508, "epoch": 2915, "lr": 4.784636781574092e-07}
{"train_loss": 0.04565022140741348, "global_step": 259509, "epoch": 2915, "lr": 4.783836715838319e-07}
{"train_loss": 0.0367177277803421, "global_step": 259510, "epoch": 2915, "lr": 4.783036716678291e-07}
{"train_loss": 0.02994673326611519, "global_step": 259511, "epoch": 2915, "lr": 4.782236784094174e-07}
{"train_loss": 0.0614706389605999, "global_step": 259512, "epoch": 2915, "lr": 4.781436918086024e-07}
{"train_loss": 0.07797179371118546, "global_step": 259513, "epoch": 2915, "lr": 4.780637118654007e-07}
{"train_loss": 0.04513414204120636, "global_step": 259514, "epoch": 2915, "lr": 4.779837385798125e-07}
{"train_loss": 0.025417158380150795, "global_step": 259515, "epoch": 2915, "lr": 4.779037719518653e-07}
{"train_loss": 0.01648501493036747, "global_step": 259516, "epoch": 2915, "lr": 4.778238119815537e-07}
{"train_loss": 0.025409184396266937, "global_step": 259517, "epoch": 2915, "lr": 4.777438586688999e-07}
{"train_loss": 0.06644729524850845, "global_step": 259518, "epoch": 2915, "lr": 4.776639120139037e-07}
{"train_loss": 0.08046659082174301, "global_step": 259519, "epoch": 2915, "lr": 4.775839720165875e-07}
{"train_loss": 0.016760125756263733, "global_step": 259520, "epoch": 2915, "lr": 4.775040386769514e-07}
{"train_loss": 0.06295250356197357, "global_step": 259521, "epoch": 2915, "lr": 4.774241119950173e-07}
{"train_loss": 0.04313645139336586, "global_step": 259522, "epoch": 2915, "lr": 4.773441919707855e-07}
{"train_loss": 0.04673629514663742, "global_step": 259523, "epoch": 2915, "lr": 4.77264278604278e-07, "val_loss": 9.640207290649414, "train_action_mse_error": 25.402223587036133}
{"train_loss": 0.024619441479444504, "global_step": 259524, "epoch": 2916, "lr": 4.771843718954894e-07}
{"train_loss": 0.03671242296695709, "global_step": 259525, "epoch": 2916, "lr": 4.771044718444472e-07}
{"train_loss": 0.07628046721220016, "global_step": 259526, "epoch": 2916, "lr": 4.770245784511462e-07}
{"train_loss": 0.03754673898220062, "global_step": 259527, "epoch": 2916, "lr": 4.769446917156139e-07}
{"train_loss": 0.019989976659417152, "global_step": 259528, "epoch": 2916, "lr": 4.768648116378504e-07}
{"train_loss": 0.05768002197146416, "global_step": 259529, "epoch": 2916, "lr": 4.7678493821786686e-07}
{"train_loss": 0.04160919412970543, "global_step": 259530, "epoch": 2916, "lr": 4.767050714556798e-07}
{"train_loss": 0.05566491559147835, "global_step": 259531, "epoch": 2916, "lr": 4.766252113512948e-07}
{"train_loss": 0.03889404237270355, "global_step": 259532, "epoch": 2916, "lr": 4.76545357904723e-07}
{"train_loss": 0.06650514155626297, "global_step": 259533, "epoch": 2916, "lr": 4.764655111159755e-07}
{"train_loss": 0.07240721583366394, "global_step": 259534, "epoch": 2916, "lr": 4.763856709850634e-07}
{"train_loss": 0.031259357929229736, "global_step": 259535, "epoch": 2916, "lr": 4.7630583751199774e-07}
{"train_loss": 0.09245751798152924, "global_step": 259536, "epoch": 2916, "lr": 4.7622601069678974e-07}
{"train_loss": 0.04041792079806328, "global_step": 259537, "epoch": 2916, "lr": 4.7614619053945044e-07}
{"train_loss": 0.04527108743786812, "global_step": 259538, "epoch": 2916, "lr": 4.760663770399909e-07}
{"train_loss": 0.04971357062458992, "global_step": 259539, "epoch": 2916, "lr": 4.7598657019841674e-07}
{"train_loss": 0.07936587929725647, "global_step": 259540, "epoch": 2916, "lr": 4.759067700147446e-07}
{"train_loss": 0.09595125913619995, "global_step": 259541, "epoch": 2916, "lr": 4.7582697648898e-07}
{"train_loss": 0.025939295068383217, "global_step": 259542, "epoch": 2916, "lr": 4.757471896211396e-07}
{"train_loss": 0.05769452825188637, "global_step": 259543, "epoch": 2916, "lr": 4.75667409411229e-07}
{"train_loss": 0.04506886005401611, "global_step": 259544, "epoch": 2916, "lr": 4.7558763585927033e-07}
{"train_loss": 0.04684774950146675, "global_step": 259545, "epoch": 2916, "lr": 4.755078689652526e-07}
{"train_loss": 0.043325863778591156, "global_step": 259546, "epoch": 2916, "lr": 4.7542810872920895e-07}
{"train_loss": 0.06144673749804497, "global_step": 259547, "epoch": 2916, "lr": 4.7534835515113396e-07}
{"train_loss": 0.055252447724342346, "global_step": 259548, "epoch": 2916, "lr": 4.7526860823104425e-07}
{"train_loss": 0.034824151545763016, "global_step": 259549, "epoch": 2916, "lr": 4.7518886796895643e-07}
{"train_loss": 0.03166468068957329, "global_step": 259550, "epoch": 2916, "lr": 4.751091343648761e-07}
{"train_loss": 0.04192110523581505, "global_step": 259551, "epoch": 2916, "lr": 4.750294074188089e-07}
{"train_loss": 0.03114720620214939, "global_step": 259552, "epoch": 2916, "lr": 4.749496871307713e-07}
{"train_loss": 0.055522430688142776, "global_step": 259553, "epoch": 2916, "lr": 4.7486997350078e-07}
{"train_loss": 0.03113185055553913, "global_step": 259554, "epoch": 2916, "lr": 4.747902665288295e-07}
{"train_loss": 0.032060377299785614, "global_step": 259555, "epoch": 2916, "lr": 4.7471056621494756e-07}
{"train_loss": 0.04222672060132027, "global_step": 259556, "epoch": 2916, "lr": 4.746308725591286e-07}
{"train_loss": 0.04679534211754799, "global_step": 259557, "epoch": 2916, "lr": 4.7455118556140045e-07}
{"train_loss": 0.01495453529059887, "global_step": 259558, "epoch": 2916, "lr": 4.7447150522175745e-07}
{"train_loss": 0.02719726599752903, "global_step": 259559, "epoch": 2916, "lr": 4.7439183154022737e-07}
{"train_loss": 0.032148025929927826, "global_step": 259560, "epoch": 2916, "lr": 4.743121645168047e-07}
{"train_loss": 0.030381904914975166, "global_step": 259561, "epoch": 2916, "lr": 4.742325041515061e-07}
{"train_loss": 0.07578743994235992, "global_step": 259562, "epoch": 2916, "lr": 4.741528504443482e-07}
{"train_loss": 0.058637116104364395, "global_step": 259563, "epoch": 2916, "lr": 4.7407320339533657e-07}
{"train_loss": 0.053739774972200394, "global_step": 259564, "epoch": 2916, "lr": 4.739935630044767e-07}
{"train_loss": 0.06726668030023575, "global_step": 259565, "epoch": 2916, "lr": 4.739139292717909e-07}
{"train_loss": 0.01712914928793907, "global_step": 259566, "epoch": 2916, "lr": 4.7383430219728466e-07}
{"train_loss": 0.024657366797327995, "global_step": 259567, "epoch": 2916, "lr": 4.737546817809579e-07}
{"train_loss": 0.03828173503279686, "global_step": 259568, "epoch": 2916, "lr": 4.736750680228441e-07}
{"train_loss": 0.02181106247007847, "global_step": 259569, "epoch": 2916, "lr": 4.735954609229321e-07}
{"train_loss": 0.01721140742301941, "global_step": 259570, "epoch": 2916, "lr": 4.73515860481244e-07}
{"train_loss": 0.04812372475862503, "global_step": 259571, "epoch": 2916, "lr": 4.734362666977854e-07}
{"train_loss": 0.04011678323149681, "global_step": 259572, "epoch": 2916, "lr": 4.73356679572573e-07}
{"train_loss": 0.06004372984170914, "global_step": 259573, "epoch": 2916, "lr": 4.7327709910561233e-07}
{"train_loss": 0.04135781154036522, "global_step": 259574, "epoch": 2916, "lr": 4.731975252969201e-07}
{"train_loss": 0.08746824413537979, "global_step": 259575, "epoch": 2916, "lr": 4.731179581464962e-07}
{"train_loss": 0.029968874529004097, "global_step": 259576, "epoch": 2916, "lr": 4.730383976543629e-07}
{"train_loss": 0.07667672634124756, "global_step": 259577, "epoch": 2916, "lr": 4.729588438205257e-07}
{"train_loss": 0.08202267438173294, "global_step": 259578, "epoch": 2916, "lr": 4.7287929664499576e-07}
{"train_loss": 0.05154038220643997, "global_step": 259579, "epoch": 2916, "lr": 4.727997561277786e-07}
{"train_loss": 0.051721785217523575, "global_step": 259580, "epoch": 2916, "lr": 4.7272022226889646e-07}
{"train_loss": 0.011859260499477386, "global_step": 259581, "epoch": 2916, "lr": 4.7264069506834927e-07}
{"train_loss": 0.04595370963215828, "global_step": 259582, "epoch": 2916, "lr": 4.725611745261538e-07}
{"train_loss": 0.029174692928791046, "global_step": 259583, "epoch": 2916, "lr": 4.7248166064231545e-07}
{"train_loss": 0.031107617542147636, "global_step": 259584, "epoch": 2916, "lr": 4.72402153416851e-07}
{"train_loss": 0.032505493611097336, "global_step": 259585, "epoch": 2916, "lr": 4.72322652849766e-07}
{"train_loss": 0.05529162287712097, "global_step": 259586, "epoch": 2916, "lr": 4.7224315894107697e-07}
{"train_loss": 0.041569579392671585, "global_step": 259587, "epoch": 2916, "lr": 4.7216367169078956e-07}
{"train_loss": 0.02882162667810917, "global_step": 259588, "epoch": 2916, "lr": 4.720841910989149e-07}
{"train_loss": 0.03723709285259247, "global_step": 259589, "epoch": 2916, "lr": 4.720047171654696e-07}
{"train_loss": 0.03165578469634056, "global_step": 259590, "epoch": 2916, "lr": 4.719252498904536e-07}
{"train_loss": 0.06101056560873985, "global_step": 259591, "epoch": 2916, "lr": 4.718457892738892e-07}
{"train_loss": 0.05062498897314072, "global_step": 259592, "epoch": 2916, "lr": 4.717663353157764e-07}
{"train_loss": 0.05800497159361839, "global_step": 259593, "epoch": 2916, "lr": 4.7168688801613735e-07}
{"train_loss": 0.059718698263168335, "global_step": 259594, "epoch": 2916, "lr": 4.7160744737496653e-07}
{"train_loss": 0.0541500486433506, "global_step": 259595, "epoch": 2916, "lr": 4.715280133922917e-07}
{"train_loss": 0.03577061742544174, "global_step": 259596, "epoch": 2916, "lr": 4.7144858606811835e-07}
{"train_loss": 0.03182942047715187, "global_step": 259597, "epoch": 2916, "lr": 4.7136916540245215e-07}
{"train_loss": 0.07102478295564651, "global_step": 259598, "epoch": 2916, "lr": 4.7128975139530406e-07}
{"train_loss": 0.05548804625868797, "global_step": 259599, "epoch": 2916, "lr": 4.7121034404669086e-07}
{"train_loss": 0.045912280678749084, "global_step": 259600, "epoch": 2916, "lr": 4.71130943356618e-07}
{"train_loss": 0.04848146438598633, "global_step": 259601, "epoch": 2916, "lr": 4.710515493250967e-07}
{"train_loss": 0.060933344066143036, "global_step": 259602, "epoch": 2916, "lr": 4.7097216195213787e-07}
{"train_loss": 0.07248159497976303, "global_step": 259603, "epoch": 2916, "lr": 4.7089278123775837e-07}
{"train_loss": 0.005468939431011677, "global_step": 259604, "epoch": 2916, "lr": 4.7081340718196363e-07}
{"train_loss": 0.02547958679497242, "global_step": 259605, "epoch": 2916, "lr": 4.7073403978475926e-07}
{"train_loss": 0.030229192227125168, "global_step": 259606, "epoch": 2916, "lr": 4.706546790461619e-07}
{"train_loss": 0.051711175590753555, "global_step": 259607, "epoch": 2916, "lr": 4.7057532496618264e-07}
{"train_loss": 0.06766172498464584, "global_step": 259608, "epoch": 2916, "lr": 4.7049597754483256e-07}
{"train_loss": 0.0283680222928524, "global_step": 259609, "epoch": 2916, "lr": 4.704166367821172e-07}
{"train_loss": 0.07325444370508194, "global_step": 259610, "epoch": 2916, "lr": 4.7033730267805334e-07}
{"train_loss": 0.052173539996147156, "global_step": 259611, "epoch": 2916, "lr": 4.702579752326464e-07}
{"train_loss": 0.04659933093516679, "global_step": 259612, "epoch": 2916, "lr": 4.70178654445913e-07, "val_loss": 9.604318618774414}
{"train_loss": 0.04093944653868675, "global_step": 259613, "epoch": 2917, "lr": 4.7009934031785885e-07}
{"train_loss": 0.04228592664003372, "global_step": 259614, "epoch": 2917, "lr": 4.700200328485005e-07}
{"train_loss": 0.031540486961603165, "global_step": 259615, "epoch": 2917, "lr": 4.6994073203783796e-07}
{"train_loss": 0.03578609228134155, "global_step": 259616, "epoch": 2917, "lr": 4.6986143788588786e-07}
{"train_loss": 0.04065241664648056, "global_step": 259617, "epoch": 2917, "lr": 4.6978215039266137e-07}
{"train_loss": 0.09236441552639008, "global_step": 259618, "epoch": 2917, "lr": 4.6970286955817513e-07}
{"train_loss": 0.015621576458215714, "global_step": 259619, "epoch": 2917, "lr": 4.696235953824291e-07}
{"train_loss": 0.02060912735760212, "global_step": 259620, "epoch": 2917, "lr": 4.6954432786543435e-07}
{"train_loss": 0.04526231437921524, "global_step": 259621, "epoch": 2917, "lr": 4.6946506700721315e-07}
{"train_loss": 0.07405905425548553, "global_step": 259622, "epoch": 2917, "lr": 4.693858128077655e-07}
{"train_loss": 0.05257023498415947, "global_step": 259623, "epoch": 2917, "lr": 4.6930656526710246e-07}
{"train_loss": 0.06324763596057892, "global_step": 259624, "epoch": 2917, "lr": 4.6922732438524076e-07}
{"train_loss": 0.026842990890145302, "global_step": 259625, "epoch": 2917, "lr": 4.691480901621914e-07}
{"train_loss": 0.05641546845436096, "global_step": 259626, "epoch": 2917, "lr": 4.6906886259795444e-07}
{"train_loss": 0.030470142140984535, "global_step": 259627, "epoch": 2917, "lr": 4.689896416925521e-07}
{"train_loss": 0.04321323707699776, "global_step": 259628, "epoch": 2917, "lr": 4.6891042744598434e-07}
{"train_loss": 0.04986110329627991, "global_step": 259629, "epoch": 2917, "lr": 4.688312198582789e-07}
{"train_loss": 0.04911023750901222, "global_step": 259630, "epoch": 2917, "lr": 4.6875201892942477e-07}
{"train_loss": 0.028126554563641548, "global_step": 259631, "epoch": 2917, "lr": 4.686728246594496e-07}
{"train_loss": 0.05385884270071983, "global_step": 259632, "epoch": 2917, "lr": 4.6859363704835346e-07}
{"train_loss": 0.07151973992586136, "global_step": 259633, "epoch": 2917, "lr": 4.68514456096153e-07}
{"train_loss": 0.02955993078649044, "global_step": 259634, "epoch": 2917, "lr": 4.6843528180285924e-07}
{"train_loss": 0.020053382962942123, "global_step": 259635, "epoch": 2917, "lr": 4.683561141684778e-07}
{"train_loss": 0.015781337395310402, "global_step": 259636, "epoch": 2917, "lr": 4.6827695319301977e-07}
{"train_loss": 0.040594231337308884, "global_step": 259637, "epoch": 2917, "lr": 4.6819779887650184e-07}
{"train_loss": 0.06671097129583359, "global_step": 259638, "epoch": 2917, "lr": 4.681186512189295e-07}
{"train_loss": 0.027195941656827927, "global_step": 259639, "epoch": 2917, "lr": 4.6803951022031943e-07}
{"train_loss": 0.02966664358973503, "global_step": 259640, "epoch": 2917, "lr": 4.6796037588067164e-07}
{"train_loss": 0.07322222739458084, "global_step": 259641, "epoch": 2917, "lr": 4.6788124820000836e-07}
{"train_loss": 0.023292409256100655, "global_step": 259642, "epoch": 2917, "lr": 4.678021271783295e-07}
{"train_loss": 0.03344978019595146, "global_step": 259643, "epoch": 2917, "lr": 4.6772301281565734e-07}
{"train_loss": 0.048313628882169724, "global_step": 259644, "epoch": 2917, "lr": 4.6764390511199184e-07}
{"train_loss": 0.03249456360936165, "global_step": 259645, "epoch": 2917, "lr": 4.6756480406734414e-07}
{"train_loss": 0.07857958227396011, "global_step": 259646, "epoch": 2917, "lr": 4.6748570968173644e-07}
{"train_loss": 0.040939681231975555, "global_step": 259647, "epoch": 2917, "lr": 4.6740662195516316e-07}
{"train_loss": 0.013563733547925949, "global_step": 259648, "epoch": 2917, "lr": 4.6732754088765205e-07}
{"train_loss": 0.0317765474319458, "global_step": 259649, "epoch": 2917, "lr": 4.672484664791976e-07}
{"train_loss": 0.03314420208334923, "global_step": 259650, "epoch": 2917, "lr": 4.67169398729822e-07}
{"train_loss": 0.0514083169400692, "global_step": 259651, "epoch": 2917, "lr": 4.6709033763953083e-07}
{"train_loss": 0.09731762111186981, "global_step": 259652, "epoch": 2917, "lr": 4.6701128320834066e-07}
{"train_loss": 0.08248206228017807, "global_step": 259653, "epoch": 2917, "lr": 4.6693223543624597e-07}
{"train_loss": 0.08076544106006622, "global_step": 259654, "epoch": 2917, "lr": 4.668531943232801e-07}
{"train_loss": 0.08588019758462906, "global_step": 259655, "epoch": 2917, "lr": 4.66774159869432e-07}
{"train_loss": 0.05158860981464386, "global_step": 259656, "epoch": 2917, "lr": 4.6669513207472926e-07}
{"train_loss": 0.02634017914533615, "global_step": 259657, "epoch": 2917, "lr": 4.6661611093917203e-07}
{"train_loss": 0.03285028040409088, "global_step": 259658, "epoch": 2917, "lr": 4.6653709646277687e-07}
{"train_loss": 0.02941303700208664, "global_step": 259659, "epoch": 2917, "lr": 4.664580886455494e-07}
{"train_loss": 0.05826261267066002, "global_step": 259660, "epoch": 2917, "lr": 4.6637908748750626e-07}
{"train_loss": 0.04955320805311203, "global_step": 259661, "epoch": 2917, "lr": 4.6630009298865294e-07}
{"train_loss": 0.05033819377422333, "global_step": 259662, "epoch": 2917, "lr": 4.6622110514900065e-07}
{"train_loss": 0.05728501081466675, "global_step": 259663, "epoch": 2917, "lr": 4.661421239685659e-07}
{"train_loss": 0.03260935842990875, "global_step": 259664, "epoch": 2917, "lr": 4.660631494473489e-07}
{"train_loss": 0.055038630962371826, "global_step": 259665, "epoch": 2917, "lr": 4.6598418158536607e-07}
{"train_loss": 0.02398744598031044, "global_step": 259666, "epoch": 2917, "lr": 4.6590522038262865e-07}
{"train_loss": 0.023667937144637108, "global_step": 259667, "epoch": 2917, "lr": 4.6582626583915325e-07}
{"train_loss": 0.04564284905791283, "global_step": 259668, "epoch": 2917, "lr": 4.657473179549343e-07}
{"train_loss": 0.06842782348394394, "global_step": 259669, "epoch": 2917, "lr": 4.656683767299996e-07}
{"train_loss": 0.09144506603479385, "global_step": 259670, "epoch": 2917, "lr": 4.655894421643492e-07}
{"train_loss": 0.026930443942546844, "global_step": 259671, "epoch": 2917, "lr": 4.6551051425799406e-07}
{"train_loss": 0.03978690505027771, "global_step": 259672, "epoch": 2917, "lr": 4.65431593010951e-07}
{"train_loss": 0.01483907736837864, "global_step": 259673, "epoch": 2917, "lr": 4.653526784232254e-07}
{"train_loss": 0.07498162984848022, "global_step": 259674, "epoch": 2917, "lr": 4.6527377049482844e-07}
{"train_loss": 0.025329846888780594, "global_step": 259675, "epoch": 2917, "lr": 4.6519486922577126e-07}
{"train_loss": 0.04635302349925041, "global_step": 259676, "epoch": 2917, "lr": 4.6511597461606494e-07}
{"train_loss": 0.039356961846351624, "global_step": 259677, "epoch": 2917, "lr": 4.650370866657261e-07}
{"train_loss": 0.0486154742538929, "global_step": 259678, "epoch": 2917, "lr": 4.649582053747492e-07}
{"train_loss": 0.02839496172964573, "global_step": 259679, "epoch": 2917, "lr": 4.64879330743162e-07}
{"train_loss": 0.053712278604507446, "global_step": 259680, "epoch": 2917, "lr": 4.648004627709646e-07}
{"train_loss": 0.04236042499542236, "global_step": 259681, "epoch": 2917, "lr": 4.6472160145817344e-07}
{"train_loss": 0.0331764742732048, "global_step": 259682, "epoch": 2917, "lr": 4.646427468047998e-07}
{"train_loss": 0.013328799977898598, "global_step": 259683, "epoch": 2917, "lr": 4.645638988108436e-07}
{"train_loss": 0.04713348671793938, "global_step": 259684, "epoch": 2917, "lr": 4.644850574763271e-07}
{"train_loss": 0.05045666545629501, "global_step": 259685, "epoch": 2917, "lr": 4.6440622280125576e-07}
{"train_loss": 0.028036033734679222, "global_step": 259686, "epoch": 2917, "lr": 4.6432739478564635e-07}
{"train_loss": 0.04292295128107071, "global_step": 259687, "epoch": 2917, "lr": 4.6424857342949325e-07}
{"train_loss": 0.04536622017621994, "global_step": 259688, "epoch": 2917, "lr": 4.6416975873282975e-07}
{"train_loss": 0.07456252723932266, "global_step": 259689, "epoch": 2917, "lr": 4.640909506956448e-07}
{"train_loss": 0.03146474063396454, "global_step": 259690, "epoch": 2917, "lr": 4.640121493179661e-07}
{"train_loss": 0.027410654351115227, "global_step": 259691, "epoch": 2917, "lr": 4.639333545997937e-07}
{"train_loss": 0.04578757658600807, "global_step": 259692, "epoch": 2917, "lr": 4.638545665411442e-07}
{"train_loss": 0.06179521977901459, "global_step": 259693, "epoch": 2917, "lr": 4.6377578514202323e-07}
{"train_loss": 0.017082739621400833, "global_step": 259694, "epoch": 2917, "lr": 4.636970104024474e-07}
{"train_loss": 0.03953438997268677, "global_step": 259695, "epoch": 2917, "lr": 4.6361824232242224e-07}
{"train_loss": 0.030549393966794014, "global_step": 259696, "epoch": 2917, "lr": 4.6353948090195885e-07}
{"train_loss": 0.06643690913915634, "global_step": 259697, "epoch": 2917, "lr": 4.634607261410684e-07}
{"train_loss": 0.035317666828632355, "global_step": 259698, "epoch": 2917, "lr": 4.6338197803976194e-07}
{"train_loss": 0.040848102420568466, "global_step": 259699, "epoch": 2917, "lr": 4.6330323659805054e-07}
{"train_loss": 0.04245775565505028, "global_step": 259700, "epoch": 2917, "lr": 4.6322450181594537e-07}
{"train_loss": 0.04441870334610511, "global_step": 259701, "epoch": 2917, "lr": 4.6314577369345744e-07, "val_loss": 9.573305130004883}
{"train_loss": 0.04054884612560272, "global_step": 259702, "epoch": 2918, "lr": 4.630670522305924e-07}
{"train_loss": 0.06205318123102188, "global_step": 259703, "epoch": 2918, "lr": 4.629883374273669e-07}
{"train_loss": 0.026389138773083687, "global_step": 259704, "epoch": 2918, "lr": 4.6290962928378646e-07}
{"train_loss": 0.06415410339832306, "global_step": 259705, "epoch": 2918, "lr": 4.6283092779986216e-07}
{"train_loss": 0.07475554198026657, "global_step": 259706, "epoch": 2918, "lr": 4.627522329756107e-07}
{"train_loss": 0.04160791635513306, "global_step": 259707, "epoch": 2918, "lr": 4.6267354481103753e-07}
{"train_loss": 0.04606157913804054, "global_step": 259708, "epoch": 2918, "lr": 4.6259486330614833e-07}
{"train_loss": 0.011413190513849258, "global_step": 259709, "epoch": 2918, "lr": 4.6251618846097077e-07}
{"train_loss": 0.03647565841674805, "global_step": 259710, "epoch": 2918, "lr": 4.624375202754938e-07}
{"train_loss": 0.06454687565565109, "global_step": 259711, "epoch": 2918, "lr": 4.623588587497396e-07}
{"train_loss": 0.031600914895534515, "global_step": 259712, "epoch": 2918, "lr": 4.6228020388372486e-07}
{"train_loss": 0.05286245048046112, "global_step": 259713, "epoch": 2918, "lr": 4.6220155567744395e-07}
{"train_loss": 0.014915748499333858, "global_step": 259714, "epoch": 2918, "lr": 4.6212291413092466e-07}
{"train_loss": 0.0682334378361702, "global_step": 259715, "epoch": 2918, "lr": 4.62044279244167e-07}
{"train_loss": 0.02575777843594551, "global_step": 259716, "epoch": 2918, "lr": 4.619656510171766e-07}
{"train_loss": 0.02239776961505413, "global_step": 259717, "epoch": 2918, "lr": 4.618870294499811e-07}
{"train_loss": 0.05239684134721756, "global_step": 259718, "epoch": 2918, "lr": 4.6180841454257494e-07}
{"train_loss": 0.02663493901491165, "global_step": 259719, "epoch": 2918, "lr": 4.6172980629497486e-07}
{"train_loss": 0.05611760914325714, "global_step": 259720, "epoch": 2918, "lr": 4.616512047071975e-07}
{"train_loss": 0.03459864482283592, "global_step": 259721, "epoch": 2918, "lr": 4.6157260977923724e-07}
{"train_loss": 0.03129523992538452, "global_step": 259722, "epoch": 2918, "lr": 4.614940215111219e-07}
{"train_loss": 0.03154507279396057, "global_step": 259723, "epoch": 2918, "lr": 4.614154399028514e-07}
{"train_loss": 0.061259087175130844, "global_step": 259724, "epoch": 2918, "lr": 4.6133686495444253e-07}
{"train_loss": 0.05487097427248955, "global_step": 259725, "epoch": 2918, "lr": 4.612582966659007e-07}
{"train_loss": 0.026941390708088875, "global_step": 259726, "epoch": 2918, "lr": 4.6117973503724266e-07}
{"train_loss": 0.021472778171300888, "global_step": 259727, "epoch": 2918, "lr": 4.611011800684739e-07}
{"train_loss": 0.07804606854915619, "global_step": 259728, "epoch": 2918, "lr": 4.610226317596056e-07}
{"train_loss": 0.07326453179121017, "global_step": 259729, "epoch": 2918, "lr": 4.6094409011064877e-07}
{"train_loss": 0.04547157511115074, "global_step": 259730, "epoch": 2918, "lr": 4.60865555121609e-07}
{"train_loss": 0.05175911262631416, "global_step": 259731, "epoch": 2918, "lr": 4.6078702679250853e-07}
{"train_loss": 0.060532744973897934, "global_step": 259732, "epoch": 2918, "lr": 4.607085051233473e-07}
{"train_loss": 0.06180093064904213, "global_step": 259733, "epoch": 2918, "lr": 4.606299901141475e-07}
{"train_loss": 0.04290551692247391, "global_step": 259734, "epoch": 2918, "lr": 4.605514817649037e-07}
{"train_loss": 0.04687077924609184, "global_step": 259735, "epoch": 2918, "lr": 4.6047298007564354e-07}
{"train_loss": 0.06164267659187317, "global_step": 259736, "epoch": 2918, "lr": 4.6039448504636153e-07}
{"train_loss": 0.05475463345646858, "global_step": 259737, "epoch": 2918, "lr": 4.603159966770798e-07}
{"train_loss": 0.026010435074567795, "global_step": 259738, "epoch": 2918, "lr": 4.6023751496779844e-07}
{"train_loss": 0.04460551589727402, "global_step": 259739, "epoch": 2918, "lr": 4.6015903991854515e-07}
{"train_loss": 0.03412420302629471, "global_step": 259740, "epoch": 2918, "lr": 4.6008057152930886e-07}
{"train_loss": 0.05005481839179993, "global_step": 259741, "epoch": 2918, "lr": 4.6000210980011724e-07}
{"train_loss": 0.05955115333199501, "global_step": 259742, "epoch": 2918, "lr": 4.5992365473097045e-07}
{"train_loss": 0.10850455611944199, "global_step": 259743, "epoch": 2918, "lr": 4.598452063218905e-07}
{"train_loss": 0.060019489377737045, "global_step": 259744, "epoch": 2918, "lr": 4.5976676457287205e-07}
{"train_loss": 0.011821644380688667, "global_step": 259745, "epoch": 2918, "lr": 4.596883294839427e-07}
{"train_loss": 0.038439664989709854, "global_step": 259746, "epoch": 2918, "lr": 4.5960990105509693e-07}
{"train_loss": 0.03588293492794037, "global_step": 259747, "epoch": 2918, "lr": 4.595314792863514e-07}
{"train_loss": 0.06853262335062027, "global_step": 259748, "epoch": 2918, "lr": 4.594530641777228e-07}
{"train_loss": 0.02084486186504364, "global_step": 259749, "epoch": 2918, "lr": 4.5937465572921666e-07}
{"train_loss": 0.07224737107753754, "global_step": 259750, "epoch": 2918, "lr": 4.5929625394084407e-07}
{"train_loss": 0.039573051035404205, "global_step": 259751, "epoch": 2918, "lr": 4.5921785881261616e-07}
{"train_loss": 0.037067960947752, "global_step": 259752, "epoch": 2918, "lr": 4.5913947034453844e-07}
{"train_loss": 0.017274264246225357, "global_step": 259753, "epoch": 2918, "lr": 4.590610885366331e-07}
{"train_loss": 0.01809706725180149, "global_step": 259754, "epoch": 2918, "lr": 4.5898271338889463e-07}
{"train_loss": 0.046429358422756195, "global_step": 259755, "epoch": 2918, "lr": 4.5890434490135084e-07}
{"train_loss": 0.04609116166830063, "global_step": 259756, "epoch": 2918, "lr": 4.5882598307399603e-07}
{"train_loss": 0.015953589230775833, "global_step": 259757, "epoch": 2918, "lr": 4.5874762790685255e-07}
{"train_loss": 0.04622286930680275, "global_step": 259758, "epoch": 2918, "lr": 4.586692793999203e-07}
{"train_loss": 0.045336488634347916, "global_step": 259759, "epoch": 2918, "lr": 4.5859093755322156e-07}
{"train_loss": 0.05517953634262085, "global_step": 259760, "epoch": 2918, "lr": 4.5851260236676183e-07}
{"train_loss": 0.04092171788215637, "global_step": 259761, "epoch": 2918, "lr": 4.5843427384055224e-07}
{"train_loss": 0.04450900852680206, "global_step": 259762, "epoch": 2918, "lr": 4.5835595197459835e-07}
{"train_loss": 0.029463615268468857, "global_step": 259763, "epoch": 2918, "lr": 4.5827763676891676e-07}
{"train_loss": 0.07676150649785995, "global_step": 259764, "epoch": 2918, "lr": 4.581993282235131e-07}
{"train_loss": 0.00710173137485981, "global_step": 259765, "epoch": 2918, "lr": 4.5812102633840395e-07}
{"train_loss": 0.06249970197677612, "global_step": 259766, "epoch": 2918, "lr": 4.5804273111359485e-07}
{"train_loss": 0.023034706711769104, "global_step": 259767, "epoch": 2918, "lr": 4.5796444254910253e-07}
{"train_loss": 0.06749819964170456, "global_step": 259768, "epoch": 2918, "lr": 4.5788616064492694e-07}
{"train_loss": 0.02463678829371929, "global_step": 259769, "epoch": 2918, "lr": 4.5780788540108476e-07}
{"train_loss": 0.025631822645664215, "global_step": 259770, "epoch": 2918, "lr": 4.5772961681758706e-07}
{"train_loss": 0.006018848158419132, "global_step": 259771, "epoch": 2918, "lr": 4.5765135489445055e-07}
{"train_loss": 0.05809803679585457, "global_step": 259772, "epoch": 2918, "lr": 4.575730996316696e-07}
{"train_loss": 0.013596727512776852, "global_step": 259773, "epoch": 2918, "lr": 4.5749485102927205e-07}
{"train_loss": 0.11281249672174454, "global_step": 259774, "epoch": 2918, "lr": 4.574166090872578e-07}
{"train_loss": 0.08774835616350174, "global_step": 259775, "epoch": 2918, "lr": 4.5733837380563804e-07}
{"train_loss": 0.01980588585138321, "global_step": 259776, "epoch": 2918, "lr": 4.5726014518442385e-07}
{"train_loss": 0.06447611004114151, "global_step": 259777, "epoch": 2918, "lr": 4.5718192322363185e-07}
{"train_loss": 0.05529520660638809, "global_step": 259778, "epoch": 2918, "lr": 4.5710370792326764e-07}
{"train_loss": 0.024167383089661598, "global_step": 259779, "epoch": 2918, "lr": 4.5702549928334225e-07}
{"train_loss": 0.09030896425247192, "global_step": 259780, "epoch": 2918, "lr": 4.5694729730386133e-07}
{"train_loss": 0.04567811265587807, "global_step": 259781, "epoch": 2918, "lr": 4.5686910198484146e-07}
{"train_loss": 0.06268002837896347, "global_step": 259782, "epoch": 2918, "lr": 4.5679091332629375e-07}
{"train_loss": 0.02447783574461937, "global_step": 259783, "epoch": 2918, "lr": 4.567127313282238e-07}
{"train_loss": 0.04969606176018715, "global_step": 259784, "epoch": 2918, "lr": 4.566345559906482e-07}
{"train_loss": 0.043581828474998474, "global_step": 259785, "epoch": 2918, "lr": 4.5655638731357253e-07}
{"train_loss": 0.07618947327136993, "global_step": 259786, "epoch": 2918, "lr": 4.564782252970079e-07}
{"train_loss": 0.016218481585383415, "global_step": 259787, "epoch": 2918, "lr": 4.564000699409654e-07}
{"train_loss": 0.016950875520706177, "global_step": 259788, "epoch": 2918, "lr": 4.563219212454617e-07}
{"train_loss": 0.024802418425679207, "global_step": 259789, "epoch": 2918, "lr": 4.5624377921049675e-07}
{"train_loss": 0.04567148162868251, "global_step": 259790, "epoch": 2918, "lr": 4.561656438360873e-07, "val_loss": 9.715899467468262}
{"train_loss": 0.057321928441524506, "global_step": 259791, "epoch": 2919, "lr": 4.5608751512224437e-07}
{"train_loss": 0.03708820417523384, "global_step": 259792, "epoch": 2919, "lr": 4.560093930689735e-07}
{"train_loss": 0.00736258877441287, "global_step": 259793, "epoch": 2919, "lr": 4.559312776762914e-07}
{"train_loss": 0.03580212965607643, "global_step": 259794, "epoch": 2919, "lr": 4.558531689442036e-07}
{"train_loss": 0.06508010625839233, "global_step": 259795, "epoch": 2919, "lr": 4.557750668727212e-07}
{"train_loss": 0.029010893777012825, "global_step": 259796, "epoch": 2919, "lr": 4.5569697146186087e-07}
{"train_loss": 0.04038791358470917, "global_step": 259797, "epoch": 2919, "lr": 4.556188827116226e-07}
{"train_loss": 0.01726704090833664, "global_step": 259798, "epoch": 2919, "lr": 4.5554080062202855e-07}
{"train_loss": 0.03667726740241051, "global_step": 259799, "epoch": 2919, "lr": 4.5546272519307874e-07}
{"train_loss": 0.021224359050393105, "global_step": 259800, "epoch": 2919, "lr": 4.553846564247899e-07}
{"train_loss": 0.0440155491232872, "global_step": 259801, "epoch": 2919, "lr": 4.553065943171675e-07}
{"train_loss": 0.03654346242547035, "global_step": 259802, "epoch": 2919, "lr": 4.5522853887022267e-07}
{"train_loss": 0.027908867225050926, "global_step": 259803, "epoch": 2919, "lr": 4.5515049008397757e-07}
{"train_loss": 0.03253216668963432, "global_step": 259804, "epoch": 2919, "lr": 4.550724479584267e-07}
{"train_loss": 0.03755840286612511, "global_step": 259805, "epoch": 2919, "lr": 4.549944124935923e-07}
{"train_loss": 0.050374340265989304, "global_step": 259806, "epoch": 2919, "lr": 4.549163836894743e-07}
{"train_loss": 0.05468650162220001, "global_step": 259807, "epoch": 2919, "lr": 4.5483836154609495e-07}
{"train_loss": 0.06143192574381828, "global_step": 259808, "epoch": 2919, "lr": 4.5476034606345416e-07}
{"train_loss": 0.015919078141450882, "global_step": 259809, "epoch": 2919, "lr": 4.546823372415687e-07}
{"train_loss": 0.042382217943668365, "global_step": 259810, "epoch": 2919, "lr": 4.546043350804441e-07}
{"train_loss": 0.030797502025961876, "global_step": 259811, "epoch": 2919, "lr": 4.5452633958010247e-07}
{"train_loss": 0.03569089621305466, "global_step": 259812, "epoch": 2919, "lr": 4.544483507405384e-07}
{"train_loss": 0.09451833367347717, "global_step": 259813, "epoch": 2919, "lr": 4.543703685617739e-07}
{"train_loss": 0.024963179603219032, "global_step": 259814, "epoch": 2919, "lr": 4.542923930438092e-07}
{"train_loss": 0.05727342888712883, "global_step": 259815, "epoch": 2919, "lr": 4.542144241866719e-07}
{"train_loss": 0.07060961425304413, "global_step": 259816, "epoch": 2919, "lr": 4.5413646199035096e-07}
{"train_loss": 0.034457847476005554, "global_step": 259817, "epoch": 2919, "lr": 4.540585064548741e-07}
{"train_loss": 0.037704821676015854, "global_step": 259818, "epoch": 2919, "lr": 4.539805575802414e-07}
{"train_loss": 0.032560694962739944, "global_step": 259819, "epoch": 2919, "lr": 4.539026153664694e-07}
{"train_loss": 0.04097888246178627, "global_step": 259820, "epoch": 2919, "lr": 4.538246798135637e-07}
{"train_loss": 0.04699475318193436, "global_step": 259821, "epoch": 2919, "lr": 4.5374675092154094e-07}
{"train_loss": 0.07540374994277954, "global_step": 259822, "epoch": 2919, "lr": 4.536688286904067e-07}
{"train_loss": 0.0643446072936058, "global_step": 259823, "epoch": 2919, "lr": 4.535909131201721e-07}
{"train_loss": 0.01946032978594303, "global_step": 259824, "epoch": 2919, "lr": 4.535130042108482e-07}
{"train_loss": 0.041047103703022, "global_step": 259825, "epoch": 2919, "lr": 4.534351019624461e-07}
{"train_loss": 0.05797535553574562, "global_step": 259826, "epoch": 2919, "lr": 4.5335720637497694e-07}
{"train_loss": 0.05317038297653198, "global_step": 259827, "epoch": 2919, "lr": 4.532793174484462e-07}
{"train_loss": 0.0772973895072937, "global_step": 259828, "epoch": 2919, "lr": 4.5320143518287063e-07}
{"train_loss": 0.062283214181661606, "global_step": 259829, "epoch": 2919, "lr": 4.531235595782557e-07}
{"train_loss": 0.11163143068552017, "global_step": 259830, "epoch": 2919, "lr": 4.5304569063462364e-07}
{"train_loss": 0.05954935774207115, "global_step": 259831, "epoch": 2919, "lr": 4.5296782835196895e-07}
{"train_loss": 0.0750008150935173, "global_step": 259832, "epoch": 2919, "lr": 4.528899727303082e-07}
{"train_loss": 0.07511749863624573, "global_step": 259833, "epoch": 2919, "lr": 4.5281212376965253e-07}
{"train_loss": 0.05911130830645561, "global_step": 259834, "epoch": 2919, "lr": 4.5273428147001305e-07}
{"train_loss": 0.02757936343550682, "global_step": 259835, "epoch": 2919, "lr": 4.526564458314009e-07}
{"train_loss": 0.06098482012748718, "global_step": 259836, "epoch": 2919, "lr": 4.5257861685382705e-07}
{"train_loss": 0.018319064751267433, "global_step": 259837, "epoch": 2919, "lr": 4.5250079453729166e-07}
{"train_loss": 0.07190696150064468, "global_step": 259838, "epoch": 2919, "lr": 4.5242297888181686e-07}
{"train_loss": 0.07574670761823654, "global_step": 259839, "epoch": 2919, "lr": 4.523451698874137e-07}
{"train_loss": 0.02996750921010971, "global_step": 259840, "epoch": 2919, "lr": 4.522673675540878e-07}
{"train_loss": 0.05224999785423279, "global_step": 259841, "epoch": 2919, "lr": 4.521895718818503e-07}
{"train_loss": 0.03043895773589611, "global_step": 259842, "epoch": 2919, "lr": 4.521117828707122e-07}
{"train_loss": 0.05192476138472557, "global_step": 259843, "epoch": 2919, "lr": 4.5203400052068466e-07}
{"train_loss": 0.06610409915447235, "global_step": 259844, "epoch": 2919, "lr": 4.5195622483177327e-07}
{"train_loss": 0.03494821488857269, "global_step": 259845, "epoch": 2919, "lr": 4.5187845580400013e-07}
{"train_loss": 0.03378567844629288, "global_step": 259846, "epoch": 2919, "lr": 4.518006934373598e-07}
{"train_loss": 0.0137149877846241, "global_step": 259847, "epoch": 2919, "lr": 4.5172293773187436e-07}
{"train_loss": 0.013874121010303497, "global_step": 259848, "epoch": 2919, "lr": 4.5164518868754947e-07}
{"train_loss": 0.021930847316980362, "global_step": 259849, "epoch": 2919, "lr": 4.5156744630440174e-07}
{"train_loss": 0.03804900497198105, "global_step": 259850, "epoch": 2919, "lr": 4.5148971058243117e-07}
{"train_loss": 0.035349614918231964, "global_step": 259851, "epoch": 2919, "lr": 4.5141198152165443e-07}
{"train_loss": 0.06617753952741623, "global_step": 259852, "epoch": 2919, "lr": 4.513342591220826e-07}
{"train_loss": 0.07463759183883667, "global_step": 259853, "epoch": 2919, "lr": 4.512565433837268e-07}
{"train_loss": 0.03767042234539986, "global_step": 259854, "epoch": 2919, "lr": 4.511788343065926e-07}
{"train_loss": 0.023245370015501976, "global_step": 259855, "epoch": 2919, "lr": 4.511011318906966e-07}
{"train_loss": 0.02677082270383835, "global_step": 259856, "epoch": 2919, "lr": 4.5102343613604434e-07}
{"train_loss": 0.02326708287000656, "global_step": 259857, "epoch": 2919, "lr": 4.50945747042647e-07}
{"train_loss": 0.07830909639596939, "global_step": 259858, "epoch": 2919, "lr": 4.5086806461051565e-07}
{"train_loss": 0.05442216992378235, "global_step": 259859, "epoch": 2919, "lr": 4.507903888396614e-07}
{"train_loss": 0.01744856685400009, "global_step": 259860, "epoch": 2919, "lr": 4.5071271973010085e-07}
{"train_loss": 0.06417376548051834, "global_step": 259861, "epoch": 2919, "lr": 4.5063505728182854e-07}
{"train_loss": 0.03772033751010895, "global_step": 259862, "epoch": 2919, "lr": 4.5055740149487214e-07}
{"train_loss": 0.018747709691524506, "global_step": 259863, "epoch": 2919, "lr": 4.5047975236923166e-07}
{"train_loss": 0.016341673210263252, "global_step": 259864, "epoch": 2919, "lr": 4.5040210990491826e-07}
{"train_loss": 0.07854441553354263, "global_step": 259865, "epoch": 2919, "lr": 4.5032447410194857e-07}
{"train_loss": 0.03756023943424225, "global_step": 259866, "epoch": 2919, "lr": 4.502468449603281e-07}
{"train_loss": 0.02799292467534542, "global_step": 259867, "epoch": 2919, "lr": 4.5016922248006243e-07}
{"train_loss": 0.12689758837223053, "global_step": 259868, "epoch": 2919, "lr": 4.500916066611738e-07}
{"train_loss": 0.08103962987661362, "global_step": 259869, "epoch": 2919, "lr": 4.5001399750366213e-07}
{"train_loss": 0.013278543949127197, "global_step": 259870, "epoch": 2919, "lr": 4.4993639500754417e-07}
{"train_loss": 0.022406695410609245, "global_step": 259871, "epoch": 2919, "lr": 4.498587991728309e-07}
{"train_loss": 0.06074381619691849, "global_step": 259872, "epoch": 2919, "lr": 4.49781209999528e-07}
{"train_loss": 0.042100660502910614, "global_step": 259873, "epoch": 2919, "lr": 4.497036274876465e-07}
{"train_loss": 0.07827793806791306, "global_step": 259874, "epoch": 2919, "lr": 4.4962605163719753e-07}
{"train_loss": 0.054929621517658234, "global_step": 259875, "epoch": 2919, "lr": 4.495484824481977e-07}
{"train_loss": 0.03328407555818558, "global_step": 259876, "epoch": 2919, "lr": 4.4947091992064706e-07}
{"train_loss": 0.03753067925572395, "global_step": 259877, "epoch": 2919, "lr": 4.493933640545622e-07}
{"train_loss": 0.020360389724373817, "global_step": 259878, "epoch": 2919, "lr": 4.493158148499543e-07}
{"train_loss": 0.045395443587448826, "global_step": 259879, "epoch": 2919, "lr": 4.492382723068289e-07, "val_loss": 9.676045417785645}
{"train_loss": 0.027277737855911255, "global_step": 259880, "epoch": 2920, "lr": 4.491607364252026e-07}
{"train_loss": 0.11921900510787964, "global_step": 259881, "epoch": 2920, "lr": 4.49083207205081e-07}
{"train_loss": 0.05709052085876465, "global_step": 259882, "epoch": 2920, "lr": 4.4900568464647517e-07}
{"train_loss": 0.06260579824447632, "global_step": 259883, "epoch": 2920, "lr": 4.489281687494018e-07}
{"train_loss": 0.033969078212976456, "global_step": 259884, "epoch": 2920, "lr": 4.4885065951386086e-07}
{"train_loss": 0.03273716941475868, "global_step": 259885, "epoch": 2920, "lr": 4.48773156939869e-07}
{"train_loss": 0.013345248997211456, "global_step": 259886, "epoch": 2920, "lr": 4.4869566102743177e-07}
{"train_loss": 0.010318727232515812, "global_step": 259887, "epoch": 2920, "lr": 4.486181717765714e-07}
{"train_loss": 0.03504584729671478, "global_step": 259888, "epoch": 2920, "lr": 4.4854068918728785e-07}
{"train_loss": 0.024380577728152275, "global_step": 259889, "epoch": 2920, "lr": 4.484632132595923e-07}
{"train_loss": 0.0485580712556839, "global_step": 259890, "epoch": 2920, "lr": 4.4838574399349577e-07}
{"train_loss": 0.051670901477336884, "global_step": 259891, "epoch": 2920, "lr": 4.4830828138901495e-07}
{"train_loss": 0.0640786811709404, "global_step": 259892, "epoch": 2920, "lr": 4.482308254461498e-07}
{"train_loss": 0.05606900900602341, "global_step": 259893, "epoch": 2920, "lr": 4.48153376164917e-07}
{"train_loss": 0.08875157684087753, "global_step": 259894, "epoch": 2920, "lr": 4.4807593354532774e-07}
{"train_loss": 0.09039102494716644, "global_step": 259895, "epoch": 2920, "lr": 4.4799849758738743e-07}
{"train_loss": 0.044419243931770325, "global_step": 259896, "epoch": 2920, "lr": 4.479210682911128e-07}
{"train_loss": 0.07253950089216232, "global_step": 259897, "epoch": 2920, "lr": 4.4784364565650936e-07}
{"train_loss": 0.04920557141304016, "global_step": 259898, "epoch": 2920, "lr": 4.4776622968359383e-07}
{"train_loss": 0.04536067321896553, "global_step": 259899, "epoch": 2920, "lr": 4.4768882037236613e-07}
{"train_loss": 0.022062554955482483, "global_step": 259900, "epoch": 2920, "lr": 4.476114177228485e-07}
{"train_loss": 0.01487954892218113, "global_step": 259901, "epoch": 2920, "lr": 4.47534021735041e-07}
{"train_loss": 0.04446367546916008, "global_step": 259902, "epoch": 2920, "lr": 4.4745663240896574e-07}
{"train_loss": 0.02412348985671997, "global_step": 259903, "epoch": 2920, "lr": 4.473792497446172e-07}
{"train_loss": 0.09394866228103638, "global_step": 259904, "epoch": 2920, "lr": 4.473018737420176e-07}
{"train_loss": 0.05109967663884163, "global_step": 259905, "epoch": 2920, "lr": 4.4722450440117247e-07}
{"train_loss": 0.01957165077328682, "global_step": 259906, "epoch": 2920, "lr": 4.471471417220985e-07}
{"train_loss": 0.044704269617795944, "global_step": 259907, "epoch": 2920, "lr": 4.470697857048012e-07}
{"train_loss": 0.013564654625952244, "global_step": 259908, "epoch": 2920, "lr": 4.4699243634929167e-07}
{"train_loss": 0.014229013584554195, "global_step": 259909, "epoch": 2920, "lr": 4.4691509365557546e-07}
{"train_loss": 0.01667340099811554, "global_step": 259910, "epoch": 2920, "lr": 4.468377576236693e-07}
{"train_loss": 0.030271707102656364, "global_step": 259911, "epoch": 2920, "lr": 4.467604282535842e-07}
{"train_loss": 0.06890174001455307, "global_step": 259912, "epoch": 2920, "lr": 4.4668310554532575e-07}
{"train_loss": 0.04078269749879837, "global_step": 259913, "epoch": 2920, "lr": 4.4660578949891064e-07}
{"train_loss": 0.06440246850252151, "global_step": 259914, "epoch": 2920, "lr": 4.4652848011433876e-07}
{"train_loss": 0.05486166104674339, "global_step": 259915, "epoch": 2920, "lr": 4.4645117739163247e-07}
{"train_loss": 0.03993823006749153, "global_step": 259916, "epoch": 2920, "lr": 4.4637388133079717e-07}
{"train_loss": 0.030076945200562477, "global_step": 259917, "epoch": 2920, "lr": 4.4629659193183846e-07}
{"train_loss": 0.03357253223657608, "global_step": 259918, "epoch": 2920, "lr": 4.4621930919477307e-07}
{"train_loss": 0.02592226304113865, "global_step": 259919, "epoch": 2920, "lr": 4.46142033119612e-07}
{"train_loss": 0.029372168704867363, "global_step": 259920, "epoch": 2920, "lr": 4.4606476370636087e-07}
{"train_loss": 0.06759358197450638, "global_step": 259921, "epoch": 2920, "lr": 4.459875009550363e-07}
{"train_loss": 0.049666862934827805, "global_step": 259922, "epoch": 2920, "lr": 4.459102448656383e-07}
{"train_loss": 0.051348358392715454, "global_step": 259923, "epoch": 2920, "lr": 4.45832995438189e-07}
{"train_loss": 0.02800460159778595, "global_step": 259924, "epoch": 2920, "lr": 4.4575575267268856e-07}
{"train_loss": 0.044287845492362976, "global_step": 259925, "epoch": 2920, "lr": 4.456785165691535e-07}
{"train_loss": 0.05157787725329399, "global_step": 259926, "epoch": 2920, "lr": 4.4560128712759496e-07}
{"train_loss": 0.03535999730229378, "global_step": 259927, "epoch": 2920, "lr": 4.455240643480241e-07}
{"train_loss": 0.0946650579571724, "global_step": 259928, "epoch": 2920, "lr": 4.4544684823044083e-07}
{"train_loss": 0.054097022861242294, "global_step": 259929, "epoch": 2920, "lr": 4.4536963877486736e-07}
{"train_loss": 0.03958888351917267, "global_step": 259930, "epoch": 2920, "lr": 4.452924359813093e-07}
{"train_loss": 0.03227153420448303, "global_step": 259931, "epoch": 2920, "lr": 4.4521523984977776e-07}
{"train_loss": 0.05674978345632553, "global_step": 259932, "epoch": 2920, "lr": 4.451380503802893e-07}
{"train_loss": 0.03319120407104492, "global_step": 259933, "epoch": 2920, "lr": 4.450608675728385e-07}
{"train_loss": 0.0292991790920496, "global_step": 259934, "epoch": 2920, "lr": 4.44983691427453e-07}
{"train_loss": 0.04677422344684601, "global_step": 259935, "epoch": 2920, "lr": 4.449065219441273e-07}
{"train_loss": 0.03682689741253853, "global_step": 259936, "epoch": 2920, "lr": 4.448293591228891e-07}
{"train_loss": 0.07800951600074768, "global_step": 259937, "epoch": 2920, "lr": 4.4475220296373297e-07}
{"train_loss": 0.05296707898378372, "global_step": 259938, "epoch": 2920, "lr": 4.446750534666755e-07}
{"train_loss": 0.034846194088459015, "global_step": 259939, "epoch": 2920, "lr": 4.4459791063172775e-07}
{"train_loss": 0.019166631624102592, "global_step": 259940, "epoch": 2920, "lr": 4.4452077445890637e-07}
{"train_loss": 0.05865278095006943, "global_step": 259941, "epoch": 2920, "lr": 4.4444364494821147e-07}
{"train_loss": 0.04704413190484047, "global_step": 259942, "epoch": 2920, "lr": 4.44366522099654e-07}
{"train_loss": 0.03221779316663742, "global_step": 259943, "epoch": 2920, "lr": 4.4428940591325073e-07}
{"train_loss": 0.05745738372206688, "global_step": 259944, "epoch": 2920, "lr": 4.442122963890072e-07}
{"train_loss": 0.07062504440546036, "global_step": 259945, "epoch": 2920, "lr": 4.4413519352693447e-07}
{"train_loss": 0.0874929204583168, "global_step": 259946, "epoch": 2920, "lr": 4.4405809732704916e-07}
{"train_loss": 0.027013292536139488, "global_step": 259947, "epoch": 2920, "lr": 4.4398100778935136e-07}
{"train_loss": 0.036953479051589966, "global_step": 259948, "epoch": 2920, "lr": 4.439039249138577e-07}
{"train_loss": 0.02778029441833496, "global_step": 259949, "epoch": 2920, "lr": 4.4382684870057923e-07}
{"train_loss": 0.04293156415224075, "global_step": 259950, "epoch": 2920, "lr": 4.43749779149516e-07}
{"train_loss": 0.047518640756607056, "global_step": 259951, "epoch": 2920, "lr": 4.4367271626069573e-07}
{"train_loss": 0.04162360727787018, "global_step": 259952, "epoch": 2920, "lr": 4.435956600341129e-07}
{"train_loss": 0.07234720885753632, "global_step": 259953, "epoch": 2920, "lr": 4.435186104697897e-07}
{"train_loss": 0.009023287333548069, "global_step": 259954, "epoch": 2920, "lr": 4.434415675677317e-07}
{"train_loss": 0.03686405345797539, "global_step": 259955, "epoch": 2920, "lr": 4.433645313279444e-07}
{"train_loss": 0.05236125737428665, "global_step": 259956, "epoch": 2920, "lr": 4.4328750175044454e-07}
{"train_loss": 0.03367013484239578, "global_step": 259957, "epoch": 2920, "lr": 4.4321047883524315e-07}
{"train_loss": 0.03070387803018093, "global_step": 259958, "epoch": 2920, "lr": 4.431334625823458e-07}
{"train_loss": 0.03964608907699585, "global_step": 259959, "epoch": 2920, "lr": 4.430564529917691e-07}
{"train_loss": 0.05551667883992195, "global_step": 259960, "epoch": 2920, "lr": 4.4297945006351315e-07}
{"train_loss": 0.042242906987667084, "global_step": 259961, "epoch": 2920, "lr": 4.429024537976001e-07}
{"train_loss": 0.05025095120072365, "global_step": 259962, "epoch": 2920, "lr": 4.428254641940299e-07}
{"train_loss": 0.0262533538043499, "global_step": 259963, "epoch": 2920, "lr": 4.4274848125281934e-07}
{"train_loss": 0.03521715849637985, "global_step": 259964, "epoch": 2920, "lr": 4.4267150497397935e-07}
{"train_loss": 0.02143261581659317, "global_step": 259965, "epoch": 2920, "lr": 4.425945353575156e-07}
{"train_loss": 0.02362724579870701, "global_step": 259966, "epoch": 2920, "lr": 4.425175724034447e-07}
{"train_loss": 0.07015691697597504, "global_step": 259967, "epoch": 2920, "lr": 4.424406161117667e-07}
{"train_loss": 0.044444456110509595, "global_step": 259968, "epoch": 2920, "lr": 4.423636664825037e-07, "val_loss": 9.586263656616211, "train_action_mse_error": 15.472343444824219}
{"train_loss": 0.03159347549080849, "global_step": 259969, "epoch": 2921, "lr": 4.4228672351566137e-07}
{"train_loss": 0.04731200262904167, "global_step": 259970, "epoch": 2921, "lr": 4.4220978721124516e-07}
{"train_loss": 0.02325616218149662, "global_step": 259971, "epoch": 2921, "lr": 4.4213285756927183e-07}
{"train_loss": 0.06464792788028717, "global_step": 259972, "epoch": 2921, "lr": 4.4205593458975235e-07}
{"train_loss": 0.05559312552213669, "global_step": 259973, "epoch": 2921, "lr": 4.419790182726924e-07}
{"train_loss": 0.038033708930015564, "global_step": 259974, "epoch": 2921, "lr": 4.4190210861810856e-07}
{"train_loss": 0.06905586272478104, "global_step": 259975, "epoch": 2921, "lr": 4.418252056260008e-07}
{"train_loss": 0.066210076212883, "global_step": 259976, "epoch": 2921, "lr": 4.417483092963859e-07}
{"train_loss": 0.04900261014699936, "global_step": 259977, "epoch": 2921, "lr": 4.4167141962927483e-07}
{"train_loss": 0.07664284110069275, "global_step": 259978, "epoch": 2921, "lr": 4.4159453662467877e-07}
{"train_loss": 0.05754478648304939, "global_step": 259979, "epoch": 2921, "lr": 4.415176602826032e-07}
{"train_loss": 0.06639323383569717, "global_step": 259980, "epoch": 2921, "lr": 4.4144079060307043e-07}
{"train_loss": 0.02951420284807682, "global_step": 259981, "epoch": 2921, "lr": 4.413639275860693e-07}
{"train_loss": 0.04551278427243233, "global_step": 259982, "epoch": 2921, "lr": 4.4128707123163304e-07}
{"train_loss": 0.038996923714876175, "global_step": 259983, "epoch": 2921, "lr": 4.412102215397562e-07}
{"train_loss": 0.038769930601119995, "global_step": 259984, "epoch": 2921, "lr": 4.4113337851045546e-07}
{"train_loss": 0.024279246106743813, "global_step": 259985, "epoch": 2921, "lr": 4.410565421437418e-07}
{"train_loss": 0.017390284687280655, "global_step": 259986, "epoch": 2921, "lr": 4.409797124396209e-07}
{"train_loss": 0.07381066679954529, "global_step": 259987, "epoch": 2921, "lr": 4.4090288939810933e-07}
{"train_loss": 0.02178281918168068, "global_step": 259988, "epoch": 2921, "lr": 4.408260730192126e-07}
{"train_loss": 0.11082138866186142, "global_step": 259989, "epoch": 2921, "lr": 4.4074926330294753e-07}
{"train_loss": 0.05531027913093567, "global_step": 259990, "epoch": 2921, "lr": 4.4067246024931395e-07}
{"train_loss": 0.017546478658914566, "global_step": 259991, "epoch": 2921, "lr": 4.405956638583286e-07}
{"train_loss": 0.025867288932204247, "global_step": 259992, "epoch": 2921, "lr": 4.405188741300026e-07}
{"train_loss": 0.056990403681993484, "global_step": 259993, "epoch": 2921, "lr": 4.40442091064347e-07}
{"train_loss": 0.03265492618083954, "global_step": 259994, "epoch": 2921, "lr": 4.4036531466136175e-07}
{"train_loss": 0.04488499462604523, "global_step": 259995, "epoch": 2921, "lr": 4.4028854492107474e-07}
{"train_loss": 0.05014951154589653, "global_step": 259996, "epoch": 2921, "lr": 4.4021178184348033e-07}
{"train_loss": 0.017703182995319366, "global_step": 259997, "epoch": 2921, "lr": 4.4013502542860076e-07}
{"train_loss": 0.03518906980752945, "global_step": 259998, "epoch": 2921, "lr": 4.40058275676436e-07}
{"train_loss": 0.10374392569065094, "global_step": 259999, "epoch": 2921, "lr": 4.399815325870027e-07}
{"train_loss": 0.0601411797106266, "global_step": 260000, "epoch": 2921, "lr": 4.39904796160312e-07}
{"train_loss": 0.028547853231430054, "global_step": 260001, "epoch": 2921, "lr": 4.398280663963694e-07}
{"train_loss": 0.036970071494579315, "global_step": 260002, "epoch": 2921, "lr": 4.397513432951916e-07}
{"train_loss": 0.04827378690242767, "global_step": 260003, "epoch": 2921, "lr": 4.396746268567786e-07}
{"train_loss": 0.03197088837623596, "global_step": 260004, "epoch": 2921, "lr": 4.395979170811526e-07}
{"train_loss": 0.04748384654521942, "global_step": 260005, "epoch": 2921, "lr": 4.3952121396831913e-07}
{"train_loss": 0.01851695403456688, "global_step": 260006, "epoch": 2921, "lr": 4.3944451751828373e-07}
{"train_loss": 0.02859918773174286, "global_step": 260007, "epoch": 2921, "lr": 4.3936782773106313e-07}
{"train_loss": 0.007729421369731426, "global_step": 260008, "epoch": 2921, "lr": 4.392911446066683e-07}
{"train_loss": 0.08631210029125214, "global_step": 260009, "epoch": 2921, "lr": 4.3921446814510493e-07}
{"train_loss": 0.0425712876021862, "global_step": 260010, "epoch": 2921, "lr": 4.39137798346384e-07}
{"train_loss": 0.04999998211860657, "global_step": 260011, "epoch": 2921, "lr": 4.390611352105167e-07}
{"train_loss": 0.05163021385669708, "global_step": 260012, "epoch": 2921, "lr": 4.3898447873751415e-07}
{"train_loss": 0.04551573842763901, "global_step": 260013, "epoch": 2921, "lr": 4.389078289273818e-07}
{"train_loss": 0.04286971688270569, "global_step": 260014, "epoch": 2921, "lr": 4.3883118578014194e-07}
{"train_loss": 0.06530134379863739, "global_step": 260015, "epoch": 2921, "lr": 4.387545492957945e-07}
{"train_loss": 0.028930071741342545, "global_step": 260016, "epoch": 2921, "lr": 4.3867791947435066e-07}
{"train_loss": 0.03698999807238579, "global_step": 260017, "epoch": 2921, "lr": 4.386012963158215e-07}
{"train_loss": 0.018989168107509613, "global_step": 260018, "epoch": 2921, "lr": 4.3852467982022363e-07}
{"train_loss": 0.04631906375288963, "global_step": 260019, "epoch": 2921, "lr": 4.3844806998755703e-07}
{"train_loss": 0.01014386024326086, "global_step": 260020, "epoch": 2921, "lr": 4.3837146681783847e-07}
{"train_loss": 0.06130112335085869, "global_step": 260021, "epoch": 2921, "lr": 4.3829487031107896e-07}
{"train_loss": 0.06275748461484909, "global_step": 260022, "epoch": 2921, "lr": 4.3821828046728407e-07}
{"train_loss": 0.041159749031066895, "global_step": 260023, "epoch": 2921, "lr": 4.3814169728646495e-07}
{"train_loss": 0.028404373675584793, "global_step": 260024, "epoch": 2921, "lr": 4.380651207686381e-07}
{"train_loss": 0.04176424443721771, "global_step": 260025, "epoch": 2921, "lr": 4.379885509138093e-07}
{"train_loss": 0.02515946701169014, "global_step": 260026, "epoch": 2921, "lr": 4.3791198772198393e-07}
{"train_loss": 0.06594941020011902, "global_step": 260027, "epoch": 2921, "lr": 4.3783543119318424e-07}
{"train_loss": 0.07364565879106522, "global_step": 260028, "epoch": 2921, "lr": 4.3775888132741025e-07}
{"train_loss": 0.04435362294316292, "global_step": 260029, "epoch": 2921, "lr": 4.376823381246731e-07}
{"train_loss": 0.02317347191274166, "global_step": 260030, "epoch": 2921, "lr": 4.3760580158498374e-07}
{"train_loss": 0.009761631488800049, "global_step": 260031, "epoch": 2921, "lr": 4.37529271708359e-07}
{"train_loss": 0.05375044420361519, "global_step": 260032, "epoch": 2921, "lr": 4.374527484948043e-07}
{"train_loss": 0.03780943527817726, "global_step": 260033, "epoch": 2921, "lr": 4.373762319443309e-07}
{"train_loss": 0.050109393894672394, "global_step": 260034, "epoch": 2921, "lr": 4.3729972205694415e-07}
{"train_loss": 0.02873481996357441, "global_step": 260035, "epoch": 2921, "lr": 4.3722321883266083e-07}
{"train_loss": 0.020643427968025208, "global_step": 260036, "epoch": 2921, "lr": 4.3714672227148644e-07}
{"train_loss": 0.013587811961770058, "global_step": 260037, "epoch": 2921, "lr": 4.3707023237343213e-07}
{"train_loss": 0.03173615783452988, "global_step": 260038, "epoch": 2921, "lr": 4.3699374913851455e-07}
{"train_loss": 0.10596626996994019, "global_step": 260039, "epoch": 2921, "lr": 4.369172725667336e-07}
{"train_loss": 0.017716242000460625, "global_step": 260040, "epoch": 2921, "lr": 4.3684080265811166e-07}
{"train_loss": 0.08256971836090088, "global_step": 260041, "epoch": 2921, "lr": 4.367643394126486e-07}
{"train_loss": 0.05412070080637932, "global_step": 260042, "epoch": 2921, "lr": 4.366878828303611e-07}
{"train_loss": 0.017567167058587074, "global_step": 260043, "epoch": 2921, "lr": 4.366114329112547e-07}
{"train_loss": 0.042564332485198975, "global_step": 260044, "epoch": 2921, "lr": 4.3653498965534055e-07}
{"train_loss": 0.039611902087926865, "global_step": 260045, "epoch": 2921, "lr": 4.3645855306262975e-07}
{"train_loss": 0.035316430032253265, "global_step": 260046, "epoch": 2921, "lr": 4.363821231331389e-07}
{"train_loss": 0.03822414204478264, "global_step": 260047, "epoch": 2921, "lr": 4.3630569986686795e-07}
{"train_loss": 0.02735831029713154, "global_step": 260048, "epoch": 2921, "lr": 4.3622928326383375e-07}
{"train_loss": 0.01941845193505287, "global_step": 260049, "epoch": 2921, "lr": 4.3615287332404166e-07}
{"train_loss": 0.02794233337044716, "global_step": 260050, "epoch": 2921, "lr": 4.3607647004750286e-07}
{"train_loss": 0.0723523497581482, "global_step": 260051, "epoch": 2921, "lr": 4.3600007343423397e-07}
{"train_loss": 0.05238759145140648, "global_step": 260052, "epoch": 2921, "lr": 4.359236834842406e-07}
{"train_loss": 0.029795708134770393, "global_step": 260053, "epoch": 2921, "lr": 4.3584730019752827e-07}
{"train_loss": 0.01744217425584793, "global_step": 260054, "epoch": 2921, "lr": 4.357709235741192e-07}
{"train_loss": 0.04134390875697136, "global_step": 260055, "epoch": 2921, "lr": 4.3569455361401334e-07}
{"train_loss": 0.0367857851088047, "global_step": 260056, "epoch": 2921, "lr": 4.3561819031722186e-07}
{"train_loss": 0.04275672704902258, "global_step": 260057, "epoch": 2921, "lr": 4.3554183368376134e-07, "val_loss": 9.595038414001465}
{"train_loss": 0.029668092727661133, "global_step": 260058, "epoch": 2922, "lr": 4.354654837136374e-07}
{"train_loss": 0.04809349775314331, "global_step": 260059, "epoch": 2922, "lr": 4.3538914040685553e-07}
{"train_loss": 0.016133781522512436, "global_step": 260060, "epoch": 2922, "lr": 4.35312803763438e-07}
{"train_loss": 0.04189930111169815, "global_step": 260061, "epoch": 2922, "lr": 4.352364737833847e-07}
{"train_loss": 0.0371728353202343, "global_step": 260062, "epoch": 2922, "lr": 4.3516015046671246e-07}
{"train_loss": 0.06348849087953568, "global_step": 260063, "epoch": 2922, "lr": 4.350838338134322e-07}
{"train_loss": 0.05674351006746292, "global_step": 260064, "epoch": 2922, "lr": 4.35007523823544e-07}
{"train_loss": 0.028755204752087593, "global_step": 260065, "epoch": 2922, "lr": 4.3493122049707014e-07}
{"train_loss": 0.02486686035990715, "global_step": 260066, "epoch": 2922, "lr": 4.348549238340105e-07}
{"train_loss": 0.04501180350780487, "global_step": 260067, "epoch": 2922, "lr": 4.3477863383438177e-07}
{"train_loss": 0.013993311673402786, "global_step": 260068, "epoch": 2922, "lr": 4.347023504981951e-07}
{"train_loss": 0.05536521226167679, "global_step": 260069, "epoch": 2922, "lr": 4.3462607382545596e-07}
{"train_loss": 0.04106423631310463, "global_step": 260070, "epoch": 2922, "lr": 4.3454980381618105e-07}
{"train_loss": 0.06948564946651459, "global_step": 260071, "epoch": 2922, "lr": 4.344735404703759e-07}
{"train_loss": 0.03974086791276932, "global_step": 260072, "epoch": 2922, "lr": 4.343972837880461e-07}
{"train_loss": 0.033343780785799026, "global_step": 260073, "epoch": 2922, "lr": 4.3432103376921384e-07}
{"train_loss": 0.023427722975611687, "global_step": 260074, "epoch": 2922, "lr": 4.342447904138791e-07}
{"train_loss": 0.011521265842020512, "global_step": 260075, "epoch": 2922, "lr": 4.3416855372205855e-07}
{"train_loss": 0.033011868596076965, "global_step": 260076, "epoch": 2922, "lr": 4.3409232369375773e-07}
{"train_loss": 0.04930594563484192, "global_step": 260077, "epoch": 2922, "lr": 4.340161003289933e-07}
{"train_loss": 0.024087341502308846, "global_step": 260078, "epoch": 2922, "lr": 4.3393988362776527e-07}
{"train_loss": 0.07869924604892731, "global_step": 260079, "epoch": 2922, "lr": 4.3386367359009586e-07}
{"train_loss": 0.04879896715283394, "global_step": 260080, "epoch": 2922, "lr": 4.33787470215985e-07}
{"train_loss": 0.0666329637169838, "global_step": 260081, "epoch": 2922, "lr": 4.337112735054494e-07}
{"train_loss": 0.02137167565524578, "global_step": 260082, "epoch": 2922, "lr": 4.3363508345849456e-07}
{"train_loss": 0.05336145684123039, "global_step": 260083, "epoch": 2922, "lr": 4.3355890007513723e-07}
{"train_loss": 0.06807791441679001, "global_step": 260084, "epoch": 2922, "lr": 4.3348272335538286e-07}
{"train_loss": 0.032038364559412, "global_step": 260085, "epoch": 2922, "lr": 4.334065532992426e-07}
{"train_loss": 0.040004875510931015, "global_step": 260086, "epoch": 2922, "lr": 4.3333038990672205e-07}
{"train_loss": 0.010636791586875916, "global_step": 260087, "epoch": 2922, "lr": 4.332542331778433e-07}
{"train_loss": 0.030067196115851402, "global_step": 260088, "epoch": 2922, "lr": 4.331780831126064e-07}
{"train_loss": 0.060712676495313644, "global_step": 260089, "epoch": 2922, "lr": 4.331019397110225e-07}
{"train_loss": 0.08060828596353531, "global_step": 260090, "epoch": 2922, "lr": 4.330258029731082e-07}
{"train_loss": 0.07389576733112335, "global_step": 260091, "epoch": 2922, "lr": 4.329496728988636e-07}
{"train_loss": 0.03674019128084183, "global_step": 260092, "epoch": 2922, "lr": 4.3287354948830517e-07}
{"train_loss": 0.04833744838833809, "global_step": 260093, "epoch": 2922, "lr": 4.3279743274144967e-07}
{"train_loss": 0.020821554586291313, "global_step": 260094, "epoch": 2922, "lr": 4.3272132265829156e-07}
{"train_loss": 0.042953990399837494, "global_step": 260095, "epoch": 2922, "lr": 4.326452192388586e-07}
{"train_loss": 0.0920569971203804, "global_step": 260096, "epoch": 2922, "lr": 4.325691224831507e-07}
{"train_loss": 0.07001164555549622, "global_step": 260097, "epoch": 2922, "lr": 4.3249303239117354e-07}
{"train_loss": 0.058171968907117844, "global_step": 260098, "epoch": 2922, "lr": 4.3241694896294924e-07}
{"train_loss": 0.09213892370462418, "global_step": 260099, "epoch": 2922, "lr": 4.323408721984779e-07}
{"train_loss": 0.06977825611829758, "global_step": 260100, "epoch": 2922, "lr": 4.322648020977815e-07}
{"train_loss": 0.031248413026332855, "global_step": 260101, "epoch": 2922, "lr": 4.321887386608547e-07}
{"train_loss": 0.0447046123445034, "global_step": 260102, "epoch": 2922, "lr": 4.321126818877197e-07}
{"train_loss": 0.04214990884065628, "global_step": 260103, "epoch": 2922, "lr": 4.320366317783875e-07}
{"train_loss": 0.03275410085916519, "global_step": 260104, "epoch": 2922, "lr": 4.3196058833285814e-07}
{"train_loss": 0.03222794830799103, "global_step": 260105, "epoch": 2922, "lr": 4.3188455155114826e-07}
{"train_loss": 0.02554045245051384, "global_step": 260106, "epoch": 2922, "lr": 4.31808521433269e-07}
{"train_loss": 0.047233376652002335, "global_step": 260107, "epoch": 2922, "lr": 4.317324979792259e-07}
{"train_loss": 0.02065766043961048, "global_step": 260108, "epoch": 2922, "lr": 4.3165648118903555e-07}
{"train_loss": 0.0362028107047081, "global_step": 260109, "epoch": 2922, "lr": 4.315804710627036e-07}
{"train_loss": 0.037084631621837616, "global_step": 260110, "epoch": 2922, "lr": 4.315044676002411e-07}
{"train_loss": 0.025994407013058662, "global_step": 260111, "epoch": 2922, "lr": 4.314284708016536e-07}
{"train_loss": 0.060330770909786224, "global_step": 260112, "epoch": 2922, "lr": 4.3135248066696333e-07}
{"train_loss": 0.04486636444926262, "global_step": 260113, "epoch": 2922, "lr": 4.3127649719617027e-07}
{"train_loss": 0.03808451443910599, "global_step": 260114, "epoch": 2922, "lr": 4.31200520389291e-07}
{"train_loss": 0.02801489271223545, "global_step": 260115, "epoch": 2922, "lr": 4.3112455024633126e-07}
{"train_loss": 0.04483295604586601, "global_step": 260116, "epoch": 2922, "lr": 4.31048586767302e-07}
{"train_loss": 0.028191708028316498, "global_step": 260117, "epoch": 2922, "lr": 4.3097262995221433e-07}
{"train_loss": 0.05460832267999649, "global_step": 260118, "epoch": 2922, "lr": 4.308966798010794e-07}
{"train_loss": 0.007741896435618401, "global_step": 260119, "epoch": 2922, "lr": 4.3082073631390275e-07}
{"train_loss": 0.032022710889577866, "global_step": 260120, "epoch": 2922, "lr": 4.30744799490701e-07}
{"train_loss": 0.03665335103869438, "global_step": 260121, "epoch": 2922, "lr": 4.306688693314798e-07}
{"train_loss": 0.019142385572195053, "global_step": 260122, "epoch": 2922, "lr": 4.3059294583625565e-07}
{"train_loss": 0.08707118779420853, "global_step": 260123, "epoch": 2922, "lr": 4.3051702900502866e-07}
{"train_loss": 0.03292818367481232, "global_step": 260124, "epoch": 2922, "lr": 4.30441118837821e-07}
{"train_loss": 0.07697034627199173, "global_step": 260125, "epoch": 2922, "lr": 4.3036521533462716e-07}
{"train_loss": 0.029719633981585503, "global_step": 260126, "epoch": 2922, "lr": 4.302893184954748e-07}
{"train_loss": 0.024118315428495407, "global_step": 260127, "epoch": 2922, "lr": 4.302134283203585e-07}
{"train_loss": 0.06668060272932053, "global_step": 260128, "epoch": 2922, "lr": 4.301375448093003e-07}
{"train_loss": 0.035003215074539185, "global_step": 260129, "epoch": 2922, "lr": 4.3006166796230597e-07}
{"train_loss": 0.017759006470441818, "global_step": 260130, "epoch": 2922, "lr": 4.299857977793864e-07}
{"train_loss": 0.014821435324847698, "global_step": 260131, "epoch": 2922, "lr": 4.2990993426054727e-07}
{"train_loss": 0.06995584070682526, "global_step": 260132, "epoch": 2922, "lr": 4.2983407740580517e-07}
{"train_loss": 0.047316279262304306, "global_step": 260133, "epoch": 2922, "lr": 4.2975822721517123e-07}
{"train_loss": 0.03980403020977974, "global_step": 260134, "epoch": 2922, "lr": 4.2968238368864544e-07}
{"train_loss": 0.017040450125932693, "global_step": 260135, "epoch": 2922, "lr": 4.2960654682625e-07}
{"train_loss": 0.042211130261421204, "global_step": 260136, "epoch": 2922, "lr": 4.2953071662798495e-07}
{"train_loss": 0.0647716298699379, "global_step": 260137, "epoch": 2922, "lr": 4.294548930938669e-07}
{"train_loss": 0.030823297798633575, "global_step": 260138, "epoch": 2922, "lr": 4.2937907622390695e-07}
{"train_loss": 0.03945570066571236, "global_step": 260139, "epoch": 2922, "lr": 4.293032660181107e-07}
{"train_loss": 0.06181436404585838, "global_step": 260140, "epoch": 2922, "lr": 4.292274624764947e-07}
{"train_loss": 0.030602840706706047, "global_step": 260141, "epoch": 2922, "lr": 4.291516655990591e-07}
{"train_loss": 0.06073835864663124, "global_step": 260142, "epoch": 2922, "lr": 4.290758753858204e-07}
{"train_loss": 0.049896128475666046, "global_step": 260143, "epoch": 2922, "lr": 4.2900009183679534e-07}
{"train_loss": 0.05811765417456627, "global_step": 260144, "epoch": 2922, "lr": 4.2892431495197835e-07}
{"train_loss": 0.039561450481414795, "global_step": 260145, "epoch": 2922, "lr": 4.288485447313917e-07}
{"train_loss": 0.04297665556829967, "global_step": 260146, "epoch": 2922, "lr": 4.287727811750408e-07, "val_loss": 9.669914245605469}
{"train_loss": 0.041858404874801636, "global_step": 260147, "epoch": 2923, "lr": 4.286970242829369e-07}
{"train_loss": 0.02424255758523941, "global_step": 260148, "epoch": 2923, "lr": 4.286212740550966e-07}
{"train_loss": 0.024961020797491074, "global_step": 260149, "epoch": 2923, "lr": 4.2854553049151426e-07}
{"train_loss": 0.04861730337142944, "global_step": 260150, "epoch": 2923, "lr": 4.284697935922177e-07}
{"train_loss": 0.03649597987532616, "global_step": 260151, "epoch": 2923, "lr": 4.28394063357207e-07}
{"train_loss": 0.038197554647922516, "global_step": 260152, "epoch": 2923, "lr": 4.2831833978649314e-07}
{"train_loss": 0.04170046001672745, "global_step": 260153, "epoch": 2923, "lr": 4.282426228800873e-07}
{"train_loss": 0.03370463848114014, "global_step": 260154, "epoch": 2923, "lr": 4.281669126380006e-07}
{"train_loss": 0.028071455657482147, "global_step": 260155, "epoch": 2923, "lr": 4.28091209060244e-07}
{"train_loss": 0.06977576017379761, "global_step": 260156, "epoch": 2923, "lr": 4.280155121468232e-07}
{"train_loss": 0.11340552568435669, "global_step": 260157, "epoch": 2923, "lr": 4.279398218977548e-07}
{"train_loss": 0.038987062871456146, "global_step": 260158, "epoch": 2923, "lr": 4.2786413831304437e-07}
{"train_loss": 0.037297822535037994, "global_step": 260159, "epoch": 2923, "lr": 4.2778846139270303e-07}
{"train_loss": 0.08740901201963425, "global_step": 260160, "epoch": 2923, "lr": 4.2771279113674735e-07}
{"train_loss": 0.058280427008867264, "global_step": 260161, "epoch": 2923, "lr": 4.2763712754517184e-07}
{"train_loss": 0.029005544260144234, "global_step": 260162, "epoch": 2923, "lr": 4.275614706180042e-07}
{"train_loss": 0.036333002150058746, "global_step": 260163, "epoch": 2923, "lr": 4.27485820355239e-07}
{"train_loss": 0.033374425023794174, "global_step": 260164, "epoch": 2923, "lr": 4.274101767568983e-07}
{"train_loss": 0.06933487951755524, "global_step": 260165, "epoch": 2923, "lr": 4.2733453982298775e-07}
{"train_loss": 0.033536504954099655, "global_step": 260166, "epoch": 2923, "lr": 4.2725890955351844e-07}
{"train_loss": 0.05141769349575043, "global_step": 260167, "epoch": 2923, "lr": 4.2718328594850146e-07}
{"train_loss": 0.07140833884477615, "global_step": 260168, "epoch": 2923, "lr": 4.271076690079423e-07}
{"train_loss": 0.030999761074781418, "global_step": 260169, "epoch": 2923, "lr": 4.270320587318577e-07}
{"train_loss": 0.036369096487760544, "global_step": 260170, "epoch": 2923, "lr": 4.2695645512025317e-07}
{"train_loss": 0.0644778311252594, "global_step": 260171, "epoch": 2923, "lr": 4.2688085817313984e-07}
{"train_loss": 0.06977065652608871, "global_step": 260172, "epoch": 2923, "lr": 4.268052678905232e-07}
{"train_loss": 0.027310315519571304, "global_step": 260173, "epoch": 2923, "lr": 4.2672968427242557e-07}
{"train_loss": 0.038025353103876114, "global_step": 260174, "epoch": 2923, "lr": 4.266541073188468e-07}
{"train_loss": 0.026463832706212997, "global_step": 260175, "epoch": 2923, "lr": 4.2657853702980363e-07}
{"train_loss": 0.030914541333913803, "global_step": 260176, "epoch": 2923, "lr": 4.2650297340530166e-07}
{"train_loss": 0.035764019936323166, "global_step": 260177, "epoch": 2923, "lr": 4.2642741644534635e-07}
{"train_loss": 0.04036981239914894, "global_step": 260178, "epoch": 2923, "lr": 4.263518661499599e-07}
{"train_loss": 0.0478852279484272, "global_step": 260179, "epoch": 2923, "lr": 4.262763225191424e-07}
{"train_loss": 0.07540373504161835, "global_step": 260180, "epoch": 2923, "lr": 4.262007855529104e-07}
{"train_loss": 0.03178337216377258, "global_step": 260181, "epoch": 2923, "lr": 4.261252552512696e-07}
{"train_loss": 0.04702058061957359, "global_step": 260182, "epoch": 2923, "lr": 4.2604973161423643e-07}
{"train_loss": 0.05098104849457741, "global_step": 260183, "epoch": 2923, "lr": 4.259742146418111e-07}
{"train_loss": 0.008215624839067459, "global_step": 260184, "epoch": 2923, "lr": 4.258987043340101e-07}
{"train_loss": 0.06397950649261475, "global_step": 260185, "epoch": 2923, "lr": 4.258232006908447e-07}
{"train_loss": 0.0557912141084671, "global_step": 260186, "epoch": 2923, "lr": 4.257477037123203e-07}
{"train_loss": 0.019658910110592842, "global_step": 260187, "epoch": 2923, "lr": 4.256722133984536e-07}
{"train_loss": 0.028126800432801247, "global_step": 260188, "epoch": 2923, "lr": 4.2559672974925024e-07}
{"train_loss": 0.0762491524219513, "global_step": 260189, "epoch": 2923, "lr": 4.255212527647212e-07}
{"train_loss": 0.04062223806977272, "global_step": 260190, "epoch": 2923, "lr": 4.254457824448776e-07}
{"train_loss": 0.05710102617740631, "global_step": 260191, "epoch": 2923, "lr": 4.253703187897251e-07}
{"train_loss": 0.03990400582551956, "global_step": 260192, "epoch": 2923, "lr": 4.252948617992802e-07}
{"train_loss": 0.050324007868766785, "global_step": 260193, "epoch": 2923, "lr": 4.2521941147354306e-07}
{"train_loss": 0.03841157257556915, "global_step": 260194, "epoch": 2923, "lr": 4.251439678125413e-07}
{"train_loss": 0.04330975562334061, "global_step": 260195, "epoch": 2923, "lr": 4.250685308162694e-07}
{"train_loss": 0.05453352630138397, "global_step": 260196, "epoch": 2923, "lr": 4.2499310048474406e-07}
{"train_loss": 0.05052519962191582, "global_step": 260197, "epoch": 2923, "lr": 4.249176768179708e-07}
{"train_loss": 0.030386220663785934, "global_step": 260198, "epoch": 2923, "lr": 4.248422598159718e-07}
{"train_loss": 0.04289272055029869, "global_step": 260199, "epoch": 2923, "lr": 4.2476684947874156e-07}
{"train_loss": 0.08396253734827042, "global_step": 260200, "epoch": 2923, "lr": 4.246914458062967e-07}
{"train_loss": 0.029163016006350517, "global_step": 260201, "epoch": 2923, "lr": 4.2461604879864837e-07}
{"train_loss": 0.02368961088359356, "global_step": 260202, "epoch": 2923, "lr": 4.245406584558076e-07}
{"train_loss": 0.07530765235424042, "global_step": 260203, "epoch": 2923, "lr": 4.244652747777855e-07}
{"train_loss": 0.06977023929357529, "global_step": 260204, "epoch": 2923, "lr": 4.243898977645822e-07}
{"train_loss": 0.0616716705262661, "global_step": 260205, "epoch": 2923, "lr": 4.2431452741622524e-07}
{"train_loss": 0.02152646891772747, "global_step": 260206, "epoch": 2923, "lr": 4.242391637327092e-07}
{"train_loss": 0.04697592929005623, "global_step": 260207, "epoch": 2923, "lr": 4.241638067140508e-07}
{"train_loss": 0.11966816335916519, "global_step": 260208, "epoch": 2923, "lr": 4.240884563602554e-07}
{"train_loss": 0.07013433426618576, "global_step": 260209, "epoch": 2923, "lr": 4.2401311267134535e-07}
{"train_loss": 0.051047418266534805, "global_step": 260210, "epoch": 2923, "lr": 4.2393777564731507e-07}
{"train_loss": 0.05122969672083855, "global_step": 260211, "epoch": 2923, "lr": 4.238624452881868e-07}
{"train_loss": 0.06170181930065155, "global_step": 260212, "epoch": 2923, "lr": 4.237871215939659e-07}
{"train_loss": 0.03245126083493233, "global_step": 260213, "epoch": 2923, "lr": 4.237118045646582e-07}
{"train_loss": 0.03440024331212044, "global_step": 260214, "epoch": 2923, "lr": 4.2363649420028575e-07}
{"train_loss": 0.06114717572927475, "global_step": 260215, "epoch": 2923, "lr": 4.2356119050084295e-07}
{"train_loss": 0.03696003183722496, "global_step": 260216, "epoch": 2923, "lr": 4.2348589346635217e-07}
{"train_loss": 0.050040535628795624, "global_step": 260217, "epoch": 2923, "lr": 4.234106030968188e-07}
{"train_loss": 0.023945922031998634, "global_step": 260218, "epoch": 2923, "lr": 4.23335319392254e-07}
{"train_loss": 0.03781841695308685, "global_step": 260219, "epoch": 2923, "lr": 4.2326004235266893e-07}
{"train_loss": 0.06289637088775635, "global_step": 260220, "epoch": 2923, "lr": 4.2318477197806905e-07}
{"train_loss": 0.05689231678843498, "global_step": 260221, "epoch": 2923, "lr": 4.2310950826847106e-07}
{"train_loss": 0.09823238849639893, "global_step": 260222, "epoch": 2923, "lr": 4.230342512238805e-07}
{"train_loss": 0.03714046999812126, "global_step": 260223, "epoch": 2923, "lr": 4.229590008443085e-07}
{"train_loss": 0.06471452116966248, "global_step": 260224, "epoch": 2923, "lr": 4.2288375712976614e-07}
{"train_loss": 0.025226688012480736, "global_step": 260225, "epoch": 2923, "lr": 4.2280852008025896e-07}
{"train_loss": 0.022334931418299675, "global_step": 260226, "epoch": 2923, "lr": 4.227332896958036e-07}
{"train_loss": 0.034392982721328735, "global_step": 260227, "epoch": 2923, "lr": 4.2265806597640566e-07}
{"train_loss": 0.042016416788101196, "global_step": 260228, "epoch": 2923, "lr": 4.225828489220818e-07}
{"train_loss": 0.02232286147773266, "global_step": 260229, "epoch": 2923, "lr": 4.2250763853283195e-07}
{"train_loss": 0.030119605362415314, "global_step": 260230, "epoch": 2923, "lr": 4.2243243480867834e-07}
{"train_loss": 0.03299727290868759, "global_step": 260231, "epoch": 2923, "lr": 4.22357237749621e-07}
{"train_loss": 0.0423683226108551, "global_step": 260232, "epoch": 2923, "lr": 4.22282047355671e-07}
{"train_loss": 0.030963126569986343, "global_step": 260233, "epoch": 2923, "lr": 4.22206863626845e-07}
{"train_loss": 0.06200374662876129, "global_step": 260234, "epoch": 2923, "lr": 4.221316865631486e-07}
{"train_loss": 0.0469978204771374, "global_step": 260235, "epoch": 2923, "lr": 4.220565161645873e-07, "val_loss": 9.548990249633789}
{"train_loss": 0.0409565232694149, "global_step": 260236, "epoch": 2924, "lr": 4.219813524311833e-07}
{"train_loss": 0.022764956578612328, "global_step": 260237, "epoch": 2924, "lr": 4.219061953629311e-07}
{"train_loss": 0.044312987476587296, "global_step": 260238, "epoch": 2924, "lr": 4.218310449598528e-07}
{"train_loss": 0.032901644706726074, "global_step": 260239, "epoch": 2924, "lr": 4.2175590122195963e-07}
{"train_loss": 0.05582220107316971, "global_step": 260240, "epoch": 2924, "lr": 4.216807641492515e-07}
{"train_loss": 0.04582623019814491, "global_step": 260241, "epoch": 2924, "lr": 4.216056337417451e-07}
{"train_loss": 0.021520568057894707, "global_step": 260242, "epoch": 2924, "lr": 4.2153050999945153e-07}
{"train_loss": 0.04761399328708649, "global_step": 260243, "epoch": 2924, "lr": 4.214553929223819e-07}
{"train_loss": 0.14131200313568115, "global_step": 260244, "epoch": 2924, "lr": 4.213802825105362e-07}
{"train_loss": 0.038954347372055054, "global_step": 260245, "epoch": 2924, "lr": 4.213051787639366e-07}
{"train_loss": 0.04201655462384224, "global_step": 260246, "epoch": 2924, "lr": 4.2123008168258314e-07}
{"train_loss": 0.03214116021990776, "global_step": 260247, "epoch": 2924, "lr": 4.2115499126649804e-07}
{"train_loss": 0.043949685990810394, "global_step": 260248, "epoch": 2924, "lr": 4.2107990751568127e-07}
{"train_loss": 0.04095971956849098, "global_step": 260249, "epoch": 2924, "lr": 4.2100483043014395e-07}
{"train_loss": 0.015734083950519562, "global_step": 260250, "epoch": 2924, "lr": 4.209297600099027e-07}
{"train_loss": 0.02166719362139702, "global_step": 260251, "epoch": 2924, "lr": 4.2085469625495755e-07}
{"train_loss": 0.029807191342115402, "global_step": 260252, "epoch": 2924, "lr": 4.207796391653307e-07}
{"train_loss": 0.07509028166532516, "global_step": 260253, "epoch": 2924, "lr": 4.2070458874102214e-07}
{"train_loss": 0.02585618756711483, "global_step": 260254, "epoch": 2924, "lr": 4.2062954498204297e-07}
{"train_loss": 0.02780461311340332, "global_step": 260255, "epoch": 2924, "lr": 4.205545078884099e-07}
{"train_loss": 0.05214310809969902, "global_step": 260256, "epoch": 2924, "lr": 4.204794774601284e-07}
{"train_loss": 0.054161954671144485, "global_step": 260257, "epoch": 2924, "lr": 4.204044536972096e-07}
{"train_loss": 0.06957341730594635, "global_step": 260258, "epoch": 2924, "lr": 4.2032943659965906e-07}
{"train_loss": 0.04240252822637558, "global_step": 260259, "epoch": 2924, "lr": 4.2025442616749344e-07}
{"train_loss": 0.020073678344488144, "global_step": 260260, "epoch": 2924, "lr": 4.2017942240072384e-07}
{"train_loss": 0.030306871980428696, "global_step": 260261, "epoch": 2924, "lr": 4.2010442529935025e-07}
{"train_loss": 0.023434963077306747, "global_step": 260262, "epoch": 2924, "lr": 4.2002943486339487e-07}
{"train_loss": 0.020879432559013367, "global_step": 260263, "epoch": 2924, "lr": 4.199544510928577e-07}
{"train_loss": 0.031813837587833405, "global_step": 260264, "epoch": 2924, "lr": 4.198794739877609e-07}
{"train_loss": 0.09565486013889313, "global_step": 260265, "epoch": 2924, "lr": 4.198045035480991e-07}
{"train_loss": 0.0251579899340868, "global_step": 260266, "epoch": 2924, "lr": 4.197295397738943e-07}
{"train_loss": 0.05111372098326683, "global_step": 260267, "epoch": 2924, "lr": 4.1965458266514657e-07}
{"train_loss": 0.06628929078578949, "global_step": 260268, "epoch": 2924, "lr": 4.195796322218781e-07}
{"train_loss": 0.03858174383640289, "global_step": 260269, "epoch": 2924, "lr": 4.1950468844409453e-07}
{"train_loss": 0.018571406602859497, "global_step": 260270, "epoch": 2924, "lr": 4.1942975133180127e-07}
{"train_loss": 0.0463191457092762, "global_step": 260271, "epoch": 2924, "lr": 4.1935482088500953e-07}
{"train_loss": 0.051366180181503296, "global_step": 260272, "epoch": 2924, "lr": 4.1927989710373594e-07}
{"train_loss": 0.05135982856154442, "global_step": 260273, "epoch": 2924, "lr": 4.1920497998798045e-07}
{"train_loss": 0.03176680952310562, "global_step": 260274, "epoch": 2924, "lr": 4.191300695377598e-07}
{"train_loss": 0.06058325245976448, "global_step": 260275, "epoch": 2924, "lr": 4.19055165753085e-07}
{"train_loss": 0.058120496571063995, "global_step": 260276, "epoch": 2924, "lr": 4.189802686339617e-07}
{"train_loss": 0.05493875592947006, "global_step": 260277, "epoch": 2924, "lr": 4.1890537818040086e-07}
{"train_loss": 0.02506500855088234, "global_step": 260278, "epoch": 2924, "lr": 4.188304943924193e-07}
{"train_loss": 0.06723302602767944, "global_step": 260279, "epoch": 2924, "lr": 4.187556172700169e-07}
{"train_loss": 0.008415179327130318, "global_step": 260280, "epoch": 2924, "lr": 4.1868074681321033e-07}
{"train_loss": 0.09441547095775604, "global_step": 260281, "epoch": 2924, "lr": 4.1860588302201074e-07}
{"train_loss": 0.04267730936408043, "global_step": 260282, "epoch": 2924, "lr": 4.1853102589641813e-07}
{"train_loss": 0.0706566795706749, "global_step": 260283, "epoch": 2924, "lr": 4.1845617543645464e-07}
{"train_loss": 0.030551180243492126, "global_step": 260284, "epoch": 2924, "lr": 4.183813316421259e-07}
{"train_loss": 0.03104962408542633, "global_step": 260285, "epoch": 2924, "lr": 4.183064945134429e-07}
{"train_loss": 0.031429726630449295, "global_step": 260286, "epoch": 2924, "lr": 4.182316640504114e-07}
{"train_loss": 0.05540366470813751, "global_step": 260287, "epoch": 2924, "lr": 4.1815684025304226e-07}
{"train_loss": 0.024532072246074677, "global_step": 260288, "epoch": 2924, "lr": 4.1808202312135227e-07}
{"train_loss": 0.04874826595187187, "global_step": 260289, "epoch": 2924, "lr": 4.1800721265534136e-07}
{"train_loss": 0.0334521122276783, "global_step": 260290, "epoch": 2924, "lr": 4.1793240885503183e-07}
{"train_loss": 0.0252867229282856, "global_step": 260291, "epoch": 2924, "lr": 4.1785761172042357e-07}
{"train_loss": 0.056565191596746445, "global_step": 260292, "epoch": 2924, "lr": 4.177828212515278e-07}
{"train_loss": 0.07768351584672928, "global_step": 260293, "epoch": 2924, "lr": 4.177080374483611e-07}
{"train_loss": 0.0257001593708992, "global_step": 260294, "epoch": 2924, "lr": 4.17633260310929e-07}
{"train_loss": 0.05830992013216019, "global_step": 260295, "epoch": 2924, "lr": 4.175584898392371e-07}
{"train_loss": 0.04431559517979622, "global_step": 260296, "epoch": 2924, "lr": 4.1748372603330197e-07}
{"train_loss": 0.030675925314426422, "global_step": 260297, "epoch": 2924, "lr": 4.174089688931349e-07}
{"train_loss": 0.029159052297472954, "global_step": 260298, "epoch": 2924, "lr": 4.173342184187412e-07}
{"train_loss": 0.032847948372364044, "global_step": 260299, "epoch": 2924, "lr": 4.172594746101266e-07}
{"train_loss": 0.016627535223960876, "global_step": 260300, "epoch": 2924, "lr": 4.171847374673188e-07}
{"train_loss": 0.0887465700507164, "global_step": 260301, "epoch": 2924, "lr": 4.171100069903067e-07}
{"train_loss": 0.027814097702503204, "global_step": 260302, "epoch": 2924, "lr": 4.1703528317911244e-07}
{"train_loss": 0.04710565134882927, "global_step": 260303, "epoch": 2924, "lr": 4.1696056603374165e-07}
{"train_loss": 0.0418194904923439, "global_step": 260304, "epoch": 2924, "lr": 4.16885855554211e-07}
{"train_loss": 0.041327740997076035, "global_step": 260305, "epoch": 2924, "lr": 4.16811151740526e-07}
{"train_loss": 0.03176552802324295, "global_step": 260306, "epoch": 2924, "lr": 4.1673645459269215e-07}
{"train_loss": 0.05989047884941101, "global_step": 260307, "epoch": 2924, "lr": 4.1666176411072623e-07}
{"train_loss": 0.08712630718946457, "global_step": 260308, "epoch": 2924, "lr": 4.165870802946337e-07}
{"train_loss": 0.019126618281006813, "global_step": 260309, "epoch": 2924, "lr": 4.165124031444256e-07}
{"train_loss": 0.05029291659593582, "global_step": 260310, "epoch": 2924, "lr": 4.1643773266011323e-07}
{"train_loss": 0.052578337490558624, "global_step": 260311, "epoch": 2924, "lr": 4.1636306884171307e-07}
{"train_loss": 0.03573552519083023, "global_step": 260312, "epoch": 2924, "lr": 4.1628841168921963e-07}
{"train_loss": 0.03053760528564453, "global_step": 260313, "epoch": 2924, "lr": 4.1621376120266066e-07}
{"train_loss": 0.029147803783416748, "global_step": 260314, "epoch": 2924, "lr": 4.161391173820306e-07}
{"train_loss": 0.06524915248155594, "global_step": 260315, "epoch": 2924, "lr": 4.1606448022735167e-07}
{"train_loss": 0.05814005061984062, "global_step": 260316, "epoch": 2924, "lr": 4.159898497386239e-07}
{"train_loss": 0.024719785898923874, "global_step": 260317, "epoch": 2924, "lr": 4.159152259158639e-07}
{"train_loss": 0.059020042419433594, "global_step": 260318, "epoch": 2924, "lr": 4.1584060875907713e-07}
{"train_loss": 0.07566969841718674, "global_step": 260319, "epoch": 2924, "lr": 4.1576599826828044e-07}
{"train_loss": 0.06645727157592773, "global_step": 260320, "epoch": 2924, "lr": 4.1569139444347925e-07}
{"train_loss": 0.03781403228640556, "global_step": 260321, "epoch": 2924, "lr": 4.1561679728468474e-07}
{"train_loss": 0.03923853859305382, "global_step": 260322, "epoch": 2924, "lr": 4.1554220679190236e-07}
{"train_loss": 0.06624962389469147, "global_step": 260323, "epoch": 2924, "lr": 4.1546762296514884e-07}
{"train_loss": 0.04458090729927749, "global_step": 260324, "epoch": 2924, "lr": 4.1539304580442973e-07, "val_loss": 9.635602951049805}
{"train_loss": 0.03380166366696358, "global_step": 260325, "epoch": 2925, "lr": 4.153184753097561e-07}
{"train_loss": 0.024153059348464012, "global_step": 260326, "epoch": 2925, "lr": 4.1524391148113907e-07}
{"train_loss": 0.058746784925460815, "global_step": 260327, "epoch": 2925, "lr": 4.151693543185897e-07}
{"train_loss": 0.01884935237467289, "global_step": 260328, "epoch": 2925, "lr": 4.150948038221136e-07}
{"train_loss": 0.0560065433382988, "global_step": 260329, "epoch": 2925, "lr": 4.1502025999172746e-07}
{"train_loss": 0.03575597703456879, "global_step": 260330, "epoch": 2925, "lr": 4.149457228274367e-07}
{"train_loss": 0.056439369916915894, "global_step": 260331, "epoch": 2925, "lr": 4.14871192329247e-07}
{"train_loss": 0.08730174601078033, "global_step": 260332, "epoch": 2925, "lr": 4.1479666849718046e-07}
{"train_loss": 0.02461676113307476, "global_step": 260333, "epoch": 2925, "lr": 4.147221513312372e-07}
{"train_loss": 0.02850729413330555, "global_step": 260334, "epoch": 2925, "lr": 4.1464764083142815e-07}
{"train_loss": 0.048903074115514755, "global_step": 260335, "epoch": 2925, "lr": 4.145731369977701e-07}
{"train_loss": 0.04092073068022728, "global_step": 260336, "epoch": 2925, "lr": 4.1449863983026307e-07}
{"train_loss": 0.04439689591526985, "global_step": 260337, "epoch": 2925, "lr": 4.144241493289236e-07}
{"train_loss": 0.04422938451170921, "global_step": 260338, "epoch": 2925, "lr": 4.143496654937629e-07}
{"train_loss": 0.033006370067596436, "global_step": 260339, "epoch": 2925, "lr": 4.142751883247864e-07}
{"train_loss": 0.027427680790424347, "global_step": 260340, "epoch": 2925, "lr": 4.1420071782201085e-07}
{"train_loss": 0.03430032357573509, "global_step": 260341, "epoch": 2925, "lr": 4.1412625398543625e-07}
{"train_loss": 0.05317964404821396, "global_step": 260342, "epoch": 2925, "lr": 4.140517968150792e-07}
{"train_loss": 0.09943372756242752, "global_step": 260343, "epoch": 2925, "lr": 4.139773463109509e-07}
{"train_loss": 0.05741199105978012, "global_step": 260344, "epoch": 2925, "lr": 4.1390290247305674e-07}
{"train_loss": 0.057840652763843536, "global_step": 260345, "epoch": 2925, "lr": 4.13828465301408e-07}
{"train_loss": 0.03378932178020477, "global_step": 260346, "epoch": 2925, "lr": 4.137540347960156e-07}
{"train_loss": 0.058391720056533813, "global_step": 260347, "epoch": 2925, "lr": 4.136796109568908e-07}
{"train_loss": 0.04903796687722206, "global_step": 260348, "epoch": 2925, "lr": 4.136051937840446e-07}
{"train_loss": 0.035332173109054565, "global_step": 260349, "epoch": 2925, "lr": 4.135307832774826e-07}
{"train_loss": 0.028165878728032112, "global_step": 260350, "epoch": 2925, "lr": 4.1345637943721595e-07}
{"train_loss": 0.035994965583086014, "global_step": 260351, "epoch": 2925, "lr": 4.133819822632556e-07}
{"train_loss": 0.06992749124765396, "global_step": 260352, "epoch": 2925, "lr": 4.133075917556128e-07}
{"train_loss": 0.023110251873731613, "global_step": 260353, "epoch": 2925, "lr": 4.1323320791429863e-07}
{"train_loss": 0.023580171167850494, "global_step": 260354, "epoch": 2925, "lr": 4.131588307393186e-07}
{"train_loss": 0.028192149475216866, "global_step": 260355, "epoch": 2925, "lr": 4.1308446023068383e-07}
{"train_loss": 0.028861138969659805, "global_step": 260356, "epoch": 2925, "lr": 4.1301009638840536e-07}
{"train_loss": 0.016990646719932556, "global_step": 260357, "epoch": 2925, "lr": 4.1293573921249997e-07}
{"train_loss": 0.04422621801495552, "global_step": 260358, "epoch": 2925, "lr": 4.1286138870296197e-07}
{"train_loss": 0.043037496507167816, "global_step": 260359, "epoch": 2925, "lr": 4.1278704485981923e-07}
{"train_loss": 0.026924680918455124, "global_step": 260360, "epoch": 2925, "lr": 4.1271270768306613e-07}
{"train_loss": 0.03506717458367348, "global_step": 260361, "epoch": 2925, "lr": 4.1263837717271935e-07}
{"train_loss": 0.025417925789952278, "global_step": 260362, "epoch": 2925, "lr": 4.1256405332879e-07}
{"train_loss": 0.04463425651192665, "global_step": 260363, "epoch": 2925, "lr": 4.1248973615128914e-07}
{"train_loss": 0.033160049468278885, "global_step": 260364, "epoch": 2925, "lr": 4.124154256402224e-07}
{"train_loss": 0.08729798346757889, "global_step": 260365, "epoch": 2925, "lr": 4.1234112179560634e-07}
{"train_loss": 0.06595104932785034, "global_step": 260366, "epoch": 2925, "lr": 4.1226682461744106e-07}
{"train_loss": 0.05050651729106903, "global_step": 260367, "epoch": 2925, "lr": 4.121925341057431e-07}
{"train_loss": 0.049754973500967026, "global_step": 260368, "epoch": 2925, "lr": 4.1211825026052364e-07}
{"train_loss": 0.031088197603821754, "global_step": 260369, "epoch": 2925, "lr": 4.120439730817882e-07}
{"train_loss": 0.046266213059425354, "global_step": 260370, "epoch": 2925, "lr": 4.1196970256955346e-07}
{"train_loss": 0.06870041787624359, "global_step": 260371, "epoch": 2925, "lr": 4.118954387238194e-07}
{"train_loss": 0.0586780309677124, "global_step": 260372, "epoch": 2925, "lr": 4.1182118154460824e-07}
{"train_loss": 0.09699632972478867, "global_step": 260373, "epoch": 2925, "lr": 4.1174693103191996e-07}
{"train_loss": 0.039502352476119995, "global_step": 260374, "epoch": 2925, "lr": 4.116726871857657e-07}
{"train_loss": 0.043429434299468994, "global_step": 260375, "epoch": 2925, "lr": 4.115984500061565e-07}
{"train_loss": 0.02887953445315361, "global_step": 260376, "epoch": 2925, "lr": 4.1152421949310904e-07}
{"train_loss": 0.07581701129674911, "global_step": 260377, "epoch": 2925, "lr": 4.1144999564662334e-07}
{"train_loss": 0.04930652305483818, "global_step": 260378, "epoch": 2925, "lr": 4.1137577846671606e-07}
{"train_loss": 0.08196163177490234, "global_step": 260379, "epoch": 2925, "lr": 4.113015679533927e-07}
{"train_loss": 0.04796837642788887, "global_step": 260380, "epoch": 2925, "lr": 4.1122736410666996e-07}
{"train_loss": 0.049664244055747986, "global_step": 260381, "epoch": 2925, "lr": 4.111531669265478e-07}
{"train_loss": 0.06098947301506996, "global_step": 260382, "epoch": 2925, "lr": 4.1107897641304295e-07}
{"train_loss": 0.05674659460783005, "global_step": 260383, "epoch": 2925, "lr": 4.1100479256616644e-07}
{"train_loss": 0.08811547607183456, "global_step": 260384, "epoch": 2925, "lr": 4.109306153859238e-07}
{"train_loss": 0.04368698596954346, "global_step": 260385, "epoch": 2925, "lr": 4.1085644487232623e-07}
{"train_loss": 0.08705928921699524, "global_step": 260386, "epoch": 2925, "lr": 4.107822810253903e-07}
{"train_loss": 0.08939322829246521, "global_step": 260387, "epoch": 2925, "lr": 4.1070812384511603e-07}
{"train_loss": 0.0653606653213501, "global_step": 260388, "epoch": 2925, "lr": 4.1063397333151454e-07}
{"train_loss": 0.035993751138448715, "global_step": 260389, "epoch": 2925, "lr": 4.1055982948460804e-07}
{"train_loss": 0.03173625469207764, "global_step": 260390, "epoch": 2925, "lr": 4.104856923043854e-07}
{"train_loss": 0.04796670377254486, "global_step": 260391, "epoch": 2925, "lr": 4.1041156179087994e-07}
{"train_loss": 0.06443767249584198, "global_step": 260392, "epoch": 2925, "lr": 4.103374379440805e-07}
{"train_loss": 0.04602069407701492, "global_step": 260393, "epoch": 2925, "lr": 4.10263320764015e-07}
{"train_loss": 0.08029219508171082, "global_step": 260394, "epoch": 2925, "lr": 4.101892102506777e-07}
{"train_loss": 0.02737712673842907, "global_step": 260395, "epoch": 2925, "lr": 4.1011510640409645e-07}
{"train_loss": 0.03572959080338478, "global_step": 260396, "epoch": 2925, "lr": 4.100410092242601e-07}
{"train_loss": 0.08668812364339828, "global_step": 260397, "epoch": 2925, "lr": 4.099669187111965e-07}
{"train_loss": 0.014904080890119076, "global_step": 260398, "epoch": 2925, "lr": 4.0989283486490003e-07}
{"train_loss": 0.043400079011917114, "global_step": 260399, "epoch": 2925, "lr": 4.0981875768539846e-07}
{"train_loss": 0.02685205265879631, "global_step": 260400, "epoch": 2925, "lr": 4.097446871726862e-07}
{"train_loss": 0.06185811385512352, "global_step": 260401, "epoch": 2925, "lr": 4.096706233267855e-07}
{"train_loss": 0.07743405550718307, "global_step": 260402, "epoch": 2925, "lr": 4.0959656614769083e-07}
{"train_loss": 0.05364610254764557, "global_step": 260403, "epoch": 2925, "lr": 4.0952251563542987e-07}
{"train_loss": 0.04423556476831436, "global_step": 260404, "epoch": 2925, "lr": 4.0944847179000266e-07}
{"train_loss": 0.05518394336104393, "global_step": 260405, "epoch": 2925, "lr": 4.093744346114203e-07}
{"train_loss": 0.06420297175645828, "global_step": 260406, "epoch": 2925, "lr": 4.0930040409969395e-07}
{"train_loss": 0.06773033738136292, "global_step": 260407, "epoch": 2925, "lr": 4.092263802548291e-07}
{"train_loss": 0.017697451636195183, "global_step": 260408, "epoch": 2925, "lr": 4.0915236307684233e-07}
{"train_loss": 0.035538580268621445, "global_step": 260409, "epoch": 2925, "lr": 4.090783525657338e-07}
{"train_loss": 0.03949074447154999, "global_step": 260410, "epoch": 2925, "lr": 4.0900434872153117e-07}
{"train_loss": 0.03711753338575363, "global_step": 260411, "epoch": 2925, "lr": 4.089303515442233e-07}
{"train_loss": 0.028733577579259872, "global_step": 260412, "epoch": 2925, "lr": 4.0885636103383807e-07}
{"train_loss": 0.04830902118989256, "global_step": 260413, "epoch": 2925, "lr": 4.087823771903754e-07, "val_loss": 9.698348045349121, "train_action_mse_error": 13.889860153198242}
{"train_loss": 0.07482115179300308, "global_step": 260414, "epoch": 2926, "lr": 4.0870840001384637e-07}
{"train_loss": 0.014340604655444622, "global_step": 260415, "epoch": 2926, "lr": 4.086344295042621e-07}
{"train_loss": 0.03723372891545296, "global_step": 260416, "epoch": 2926, "lr": 4.0856046566163373e-07}
{"train_loss": 0.03325599059462547, "global_step": 260417, "epoch": 2926, "lr": 4.084865084859668e-07}
{"train_loss": 0.07346097379922867, "global_step": 260418, "epoch": 2926, "lr": 4.08412557977278e-07}
{"train_loss": 0.042872875928878784, "global_step": 260419, "epoch": 2926, "lr": 4.083386141355727e-07}
{"train_loss": 0.045253414660692215, "global_step": 260420, "epoch": 2926, "lr": 4.082646769608567e-07}
{"train_loss": 0.06131841987371445, "global_step": 260421, "epoch": 2926, "lr": 4.0819074645315205e-07}
{"train_loss": 0.04210339114069939, "global_step": 260422, "epoch": 2926, "lr": 4.0811682261245876e-07}
{"train_loss": 0.03209998086094856, "global_step": 260423, "epoch": 2926, "lr": 4.0804290543879356e-07}
{"train_loss": 0.03132244944572449, "global_step": 260424, "epoch": 2926, "lr": 4.0796899493215634e-07}
{"train_loss": 0.041736900806427, "global_step": 260425, "epoch": 2926, "lr": 4.078950910925694e-07}
{"train_loss": 0.03980991989374161, "global_step": 260426, "epoch": 2926, "lr": 4.0782119392003273e-07}
{"train_loss": 0.07016544044017792, "global_step": 260427, "epoch": 2926, "lr": 4.077473034145629e-07}
{"train_loss": 0.05171671509742737, "global_step": 260428, "epoch": 2926, "lr": 4.0767341957616003e-07}
{"train_loss": 0.03966556489467621, "global_step": 260429, "epoch": 2926, "lr": 4.075995424048518e-07}
{"train_loss": 0.06539475172758102, "global_step": 260430, "epoch": 2926, "lr": 4.0752567190062707e-07}
{"train_loss": 0.0519181452691555, "global_step": 260431, "epoch": 2926, "lr": 4.0745180806351366e-07}
{"train_loss": 0.027255913242697716, "global_step": 260432, "epoch": 2926, "lr": 4.0737795089350606e-07}
{"train_loss": 0.06710679829120636, "global_step": 260433, "epoch": 2926, "lr": 4.0730410039063195e-07}
{"train_loss": 0.05854064226150513, "global_step": 260434, "epoch": 2926, "lr": 4.072302565548858e-07}
{"train_loss": 0.03777620196342468, "global_step": 260435, "epoch": 2926, "lr": 4.071564193862842e-07}
{"train_loss": 0.08797742426395416, "global_step": 260436, "epoch": 2926, "lr": 4.070825888848329e-07}
{"train_loss": 0.031137336045503616, "global_step": 260437, "epoch": 2926, "lr": 4.0700876505054273e-07}
{"train_loss": 0.028064049780368805, "global_step": 260438, "epoch": 2926, "lr": 4.069349478834361e-07}
{"train_loss": 0.05492924153804779, "global_step": 260439, "epoch": 2926, "lr": 4.068611373835074e-07}
{"train_loss": 0.03363431617617607, "global_step": 260440, "epoch": 2926, "lr": 4.067873335507677e-07}
{"train_loss": 0.03777211904525757, "global_step": 260441, "epoch": 2926, "lr": 4.0671353638523925e-07}
{"train_loss": 0.051810987293720245, "global_step": 260442, "epoch": 2926, "lr": 4.0663974588691647e-07}
{"train_loss": 0.05281412601470947, "global_step": 260443, "epoch": 2926, "lr": 4.065659620558215e-07}
{"train_loss": 0.0467582568526268, "global_step": 260444, "epoch": 2926, "lr": 4.064921848919545e-07}
{"train_loss": 0.03210053965449333, "global_step": 260445, "epoch": 2926, "lr": 4.06418414395332e-07}
{"train_loss": 0.059060219675302505, "global_step": 260446, "epoch": 2926, "lr": 4.0634465056596513e-07}
{"train_loss": 0.08605112880468369, "global_step": 260447, "epoch": 2926, "lr": 4.0627089340385393e-07}
{"train_loss": 0.05893012508749962, "global_step": 260448, "epoch": 2926, "lr": 4.0619714290902055e-07}
{"train_loss": 0.022604690864682198, "global_step": 260449, "epoch": 2926, "lr": 4.061233990814706e-07}
{"train_loss": 0.034121934324502945, "global_step": 260450, "epoch": 2926, "lr": 4.0604966192120953e-07}
{"train_loss": 0.02922980487346649, "global_step": 260451, "epoch": 2926, "lr": 4.059759314282541e-07}
{"train_loss": 0.059422966092824936, "global_step": 260452, "epoch": 2926, "lr": 4.0590220760260424e-07}
{"train_loss": 0.03141724690794945, "global_step": 260453, "epoch": 2926, "lr": 4.058284904442822e-07}
{"train_loss": 0.045429527759552, "global_step": 260454, "epoch": 2926, "lr": 4.0575477995329346e-07}
{"train_loss": 0.0485815554857254, "global_step": 260455, "epoch": 2926, "lr": 4.0568107612963813e-07}
{"train_loss": 0.06705936044454575, "global_step": 260456, "epoch": 2926, "lr": 4.056073789733383e-07}
{"train_loss": 0.061919793486595154, "global_step": 260457, "epoch": 2926, "lr": 4.055336884844052e-07}
{"train_loss": 0.08905448764562607, "global_step": 260458, "epoch": 2926, "lr": 4.0546000466283874e-07}
{"train_loss": 0.017822980880737305, "global_step": 260459, "epoch": 2926, "lr": 4.053863275086556e-07}
{"train_loss": 0.07153873890638351, "global_step": 260460, "epoch": 2926, "lr": 4.0531265702186126e-07}
{"train_loss": 0.05080987513065338, "global_step": 260461, "epoch": 2926, "lr": 4.052389932024725e-07}
{"train_loss": 0.043776288628578186, "global_step": 260462, "epoch": 2926, "lr": 4.0516533605048924e-07}
{"train_loss": 0.01733974739909172, "global_step": 260463, "epoch": 2926, "lr": 4.050916855659337e-07}
{"train_loss": 0.01911410316824913, "global_step": 260464, "epoch": 2926, "lr": 4.050180417488003e-07}
{"train_loss": 0.039325036108493805, "global_step": 260465, "epoch": 2926, "lr": 4.0494440459911685e-07}
{"train_loss": 0.028676195070147514, "global_step": 260466, "epoch": 2926, "lr": 4.048707741168778e-07}
{"train_loss": 0.044041574001312256, "global_step": 260467, "epoch": 2926, "lr": 4.047971503021053e-07}
{"train_loss": 0.027636561542749405, "global_step": 260468, "epoch": 2926, "lr": 4.047235331547994e-07}
{"train_loss": 0.02609052136540413, "global_step": 260469, "epoch": 2926, "lr": 4.046499226749767e-07}
{"train_loss": 0.03745463117957115, "global_step": 260470, "epoch": 2926, "lr": 4.045763188626372e-07}
{"train_loss": 0.023307159543037415, "global_step": 260471, "epoch": 2926, "lr": 4.045027217178032e-07}
{"train_loss": 0.017434511333703995, "global_step": 260472, "epoch": 2926, "lr": 4.044291312404802e-07}
{"train_loss": 0.0314684696495533, "global_step": 260473, "epoch": 2926, "lr": 4.0435554743067374e-07}
{"train_loss": 0.03964885324239731, "global_step": 260474, "epoch": 2926, "lr": 4.0428197028840044e-07}
{"train_loss": 0.048849549144506454, "global_step": 260475, "epoch": 2926, "lr": 4.042083998136659e-07}
{"train_loss": 0.052106358110904694, "global_step": 260476, "epoch": 2926, "lr": 4.0413483600648115e-07}
{"train_loss": 0.06441091746091843, "global_step": 260477, "epoch": 2926, "lr": 4.040612788668574e-07}
{"train_loss": 0.05598997324705124, "global_step": 260478, "epoch": 2926, "lr": 4.039877283948001e-07}
{"train_loss": 0.04381532222032547, "global_step": 260479, "epoch": 2926, "lr": 4.0391418459032607e-07}
{"train_loss": 0.05679690092802048, "global_step": 260480, "epoch": 2926, "lr": 4.0384064745343507e-07}
{"train_loss": 0.014700724743306637, "global_step": 260481, "epoch": 2926, "lr": 4.0376711698414947e-07}
{"train_loss": 0.04466654360294342, "global_step": 260482, "epoch": 2926, "lr": 4.0369359318246926e-07}
{"train_loss": 0.04179166257381439, "global_step": 260483, "epoch": 2926, "lr": 4.03620076048411e-07}
{"train_loss": 0.014538759365677834, "global_step": 260484, "epoch": 2926, "lr": 4.035465655819803e-07}
{"train_loss": 0.01670086942613125, "global_step": 260485, "epoch": 2926, "lr": 4.0347306178318835e-07}
{"train_loss": 0.06880361586809158, "global_step": 260486, "epoch": 2926, "lr": 4.033995646520461e-07}
{"train_loss": 0.07180388271808624, "global_step": 260487, "epoch": 2926, "lr": 4.0332607418855917e-07}
{"train_loss": 0.028125392273068428, "global_step": 260488, "epoch": 2926, "lr": 4.0325259039274423e-07}
{"train_loss": 0.034564051777124405, "global_step": 260489, "epoch": 2926, "lr": 4.0317911326460123e-07}
{"train_loss": 0.07753551006317139, "global_step": 260490, "epoch": 2926, "lr": 4.031056428041524e-07}
{"train_loss": 0.039631329476833344, "global_step": 260491, "epoch": 2926, "lr": 4.030321790113978e-07}
{"train_loss": 0.03824939206242561, "global_step": 260492, "epoch": 2926, "lr": 4.0295872188634843e-07}
{"train_loss": 0.043790798634290695, "global_step": 260493, "epoch": 2926, "lr": 4.0288527142902656e-07}
{"train_loss": 0.0656646192073822, "global_step": 260494, "epoch": 2926, "lr": 4.0281182763942104e-07}
{"train_loss": 0.037168290466070175, "global_step": 260495, "epoch": 2926, "lr": 4.027383905175597e-07}
{"train_loss": 0.05926011875271797, "global_step": 260496, "epoch": 2926, "lr": 4.02664960063448e-07}
{"train_loss": 0.04138565808534622, "global_step": 260497, "epoch": 2926, "lr": 4.0259153627709155e-07}
{"train_loss": 0.030232537537813187, "global_step": 260498, "epoch": 2926, "lr": 4.025181191584959e-07}
{"train_loss": 0.05488060787320137, "global_step": 260499, "epoch": 2926, "lr": 4.0244470870768324e-07}
{"train_loss": 0.06103264540433884, "global_step": 260500, "epoch": 2926, "lr": 4.023713049246536e-07}
{"train_loss": 0.05787317827343941, "global_step": 260501, "epoch": 2926, "lr": 4.0229790780942355e-07}
{"train_loss": 0.04499644507769118, "global_step": 260502, "epoch": 2926, "lr": 4.0222451736199873e-07, "val_loss": 9.616147994995117}
{"train_loss": 0.02823304384946823, "global_step": 260503, "epoch": 2927, "lr": 4.0215113358239573e-07}
{"train_loss": 0.043756626546382904, "global_step": 260504, "epoch": 2927, "lr": 4.020777564706091e-07}
{"train_loss": 0.04576892778277397, "global_step": 260505, "epoch": 2927, "lr": 4.0200438602666645e-07}
{"train_loss": 0.016838613897562027, "global_step": 260506, "epoch": 2927, "lr": 4.019310222505679e-07}
{"train_loss": 0.05994219705462456, "global_step": 260507, "epoch": 2927, "lr": 4.018576651423245e-07}
{"train_loss": 0.06787538528442383, "global_step": 260508, "epoch": 2927, "lr": 4.0178431470194734e-07}
{"train_loss": 0.010452207177877426, "global_step": 260509, "epoch": 2927, "lr": 4.01710970929442e-07}
{"train_loss": 0.04294665530323982, "global_step": 260510, "epoch": 2927, "lr": 4.016376338248251e-07}
{"train_loss": 0.05072929337620735, "global_step": 260511, "epoch": 2927, "lr": 4.0156430338810223e-07}
{"train_loss": 0.07077081501483917, "global_step": 260512, "epoch": 2927, "lr": 4.0149097961929007e-07}
{"train_loss": 0.03979941084980965, "global_step": 260513, "epoch": 2927, "lr": 4.0141766251838855e-07}
{"train_loss": 0.05186121165752411, "global_step": 260514, "epoch": 2927, "lr": 4.0134435208540874e-07}
{"train_loss": 0.04476410523056984, "global_step": 260515, "epoch": 2927, "lr": 4.0127104832036743e-07}
{"train_loss": 0.02004890888929367, "global_step": 260516, "epoch": 2927, "lr": 4.0119775122327006e-07}
{"train_loss": 0.050651464611291885, "global_step": 260517, "epoch": 2927, "lr": 4.0112446079413337e-07}
{"train_loss": 0.057358723133802414, "global_step": 260518, "epoch": 2927, "lr": 4.010511770329517e-07}
{"train_loss": 0.01823052391409874, "global_step": 260519, "epoch": 2927, "lr": 4.009778999397473e-07}
{"train_loss": 0.05207119137048721, "global_step": 260520, "epoch": 2927, "lr": 4.009046295145258e-07}
{"train_loss": 0.061070866882801056, "global_step": 260521, "epoch": 2927, "lr": 4.0083136575730373e-07}
{"train_loss": 0.03849082067608833, "global_step": 260522, "epoch": 2927, "lr": 4.0075810866808116e-07}
{"train_loss": 0.05009298399090767, "global_step": 260523, "epoch": 2927, "lr": 4.006848582468692e-07}
{"train_loss": 0.03913610056042671, "global_step": 260524, "epoch": 2927, "lr": 4.0061161449368444e-07}
{"train_loss": 0.05113987624645233, "global_step": 260525, "epoch": 2927, "lr": 4.005383774085325e-07}
{"train_loss": 0.04620775207877159, "global_step": 260526, "epoch": 2927, "lr": 4.0046514699142444e-07}
{"train_loss": 0.08209509402513504, "global_step": 260527, "epoch": 2927, "lr": 4.0039192324236587e-07}
{"train_loss": 0.014778975397348404, "global_step": 260528, "epoch": 2927, "lr": 4.003187061613733e-07}
{"train_loss": 0.05710462108254433, "global_step": 260529, "epoch": 2927, "lr": 4.002454957484525e-07}
{"train_loss": 0.08229465782642365, "global_step": 260530, "epoch": 2927, "lr": 4.001722920036144e-07}
{"train_loss": 0.033805619925260544, "global_step": 260531, "epoch": 2927, "lr": 4.0009909492686457e-07}
{"train_loss": 0.04033553600311279, "global_step": 260532, "epoch": 2927, "lr": 4.0002590451821977e-07}
{"train_loss": 0.04349144175648689, "global_step": 260533, "epoch": 2927, "lr": 3.9995272077769096e-07}
{"train_loss": 0.028409386053681374, "global_step": 260534, "epoch": 2927, "lr": 3.9987954370527827e-07}
{"train_loss": 0.032797571271657944, "global_step": 260535, "epoch": 2927, "lr": 3.998063733010038e-07}
{"train_loss": 0.03786686435341835, "global_step": 260536, "epoch": 2927, "lr": 3.9973320956486207e-07}
{"train_loss": 0.018455354496836662, "global_step": 260537, "epoch": 2927, "lr": 3.9966005249688075e-07}
{"train_loss": 0.06254660338163376, "global_step": 260538, "epoch": 2927, "lr": 3.995869020970544e-07}
{"train_loss": 0.0520840659737587, "global_step": 260539, "epoch": 2927, "lr": 3.995137583653996e-07}
{"train_loss": 0.027981063351035118, "global_step": 260540, "epoch": 2927, "lr": 3.994406213019275e-07}
{"train_loss": 0.0162284504622221, "global_step": 260541, "epoch": 2927, "lr": 3.993674909066436e-07}
{"train_loss": 0.032885290682315826, "global_step": 260542, "epoch": 2927, "lr": 3.9929436717955906e-07}
{"train_loss": 0.054234422743320465, "global_step": 260543, "epoch": 2927, "lr": 3.992212501206905e-07}
{"train_loss": 0.015719378367066383, "global_step": 260544, "epoch": 2927, "lr": 3.9914813973003784e-07}
{"train_loss": 0.04578954726457596, "global_step": 260545, "epoch": 2927, "lr": 3.990750360076179e-07}
{"train_loss": 0.02567487582564354, "global_step": 260546, "epoch": 2927, "lr": 3.9900193895343606e-07}
{"train_loss": 0.0466296523809433, "global_step": 260547, "epoch": 2927, "lr": 3.9892884856750354e-07}
{"train_loss": 0.06656891107559204, "global_step": 260548, "epoch": 2927, "lr": 3.9885576484983144e-07}
{"train_loss": 0.013693316839635372, "global_step": 260549, "epoch": 2927, "lr": 3.9878268780042526e-07}
{"train_loss": 0.032904110848903656, "global_step": 260550, "epoch": 2927, "lr": 3.9870961741930724e-07}
{"train_loss": 0.04651525244116783, "global_step": 260551, "epoch": 2927, "lr": 3.9863655370646626e-07}
{"train_loss": 0.03316708281636238, "global_step": 260552, "epoch": 2927, "lr": 3.985634966619356e-07}
{"train_loss": 0.055202629417181015, "global_step": 260553, "epoch": 2927, "lr": 3.984904462857042e-07}
{"train_loss": 0.031754616647958755, "global_step": 260554, "epoch": 2927, "lr": 3.9841740257779425e-07}
{"train_loss": 0.028733626008033752, "global_step": 260555, "epoch": 2927, "lr": 3.9834436553821687e-07}
{"train_loss": 0.040407694876194, "global_step": 260556, "epoch": 2927, "lr": 3.98271335166972e-07}
{"train_loss": 0.03013215959072113, "global_step": 260557, "epoch": 2927, "lr": 3.981983114640764e-07}
{"train_loss": 0.06358040869235992, "global_step": 260558, "epoch": 2927, "lr": 3.9812529442954106e-07}
{"train_loss": 0.029894458130002022, "global_step": 260559, "epoch": 2927, "lr": 3.980522840633716e-07}
{"train_loss": 0.04448912665247917, "global_step": 260560, "epoch": 2927, "lr": 3.9797928036557905e-07}
{"train_loss": 0.028086166828870773, "global_step": 260561, "epoch": 2927, "lr": 3.9790628333617463e-07}
{"train_loss": 0.037325285375118256, "global_step": 260562, "epoch": 2927, "lr": 3.978332929751638e-07}
{"train_loss": 0.01456619892269373, "global_step": 260563, "epoch": 2927, "lr": 3.9776030928256325e-07}
{"train_loss": 0.10415618866682053, "global_step": 260564, "epoch": 2927, "lr": 3.9768733225837297e-07}
{"train_loss": 0.030929945409297943, "global_step": 260565, "epoch": 2927, "lr": 3.9761436190261516e-07}
{"train_loss": 0.0669671893119812, "global_step": 260566, "epoch": 2927, "lr": 3.9754139821529533e-07}
{"train_loss": 0.05286814644932747, "global_step": 260567, "epoch": 2927, "lr": 3.9746844119641914e-07}
{"train_loss": 0.05042830854654312, "global_step": 260568, "epoch": 2927, "lr": 3.973954908459976e-07}
{"train_loss": 0.04302164167165756, "global_step": 260569, "epoch": 2927, "lr": 3.9732254716404184e-07}
{"train_loss": 0.1091858372092247, "global_step": 260570, "epoch": 2927, "lr": 3.9724961015056297e-07}
{"train_loss": 0.04777060076594353, "global_step": 260571, "epoch": 2927, "lr": 3.971766798055665e-07}
{"train_loss": 0.07864400744438171, "global_step": 260572, "epoch": 2927, "lr": 3.9710375612906357e-07}
{"train_loss": 0.06752892583608627, "global_step": 260573, "epoch": 2927, "lr": 3.9703083912107084e-07}
{"train_loss": 0.031163794919848442, "global_step": 260574, "epoch": 2927, "lr": 3.9695792878158835e-07}
{"train_loss": 0.04988401383161545, "global_step": 260575, "epoch": 2927, "lr": 3.968850251106382e-07}
{"train_loss": 0.0347001850605011, "global_step": 260576, "epoch": 2927, "lr": 3.9681212810821487e-07}
{"train_loss": 0.05086590722203255, "global_step": 260577, "epoch": 2927, "lr": 3.967392377743351e-07}
{"train_loss": 0.07095809280872345, "global_step": 260578, "epoch": 2927, "lr": 3.966663541090099e-07}
{"train_loss": 0.034756533801555634, "global_step": 260579, "epoch": 2927, "lr": 3.965934771122504e-07}
{"train_loss": 0.04448182135820389, "global_step": 260580, "epoch": 2927, "lr": 3.965206067840621e-07}
{"train_loss": 0.04152623936533928, "global_step": 260581, "epoch": 2927, "lr": 3.9644774312446175e-07}
{"train_loss": 0.043147288262844086, "global_step": 260582, "epoch": 2927, "lr": 3.963748861334493e-07}
{"train_loss": 0.022649923339486122, "global_step": 260583, "epoch": 2927, "lr": 3.963020358110414e-07}
{"train_loss": 0.03222927823662758, "global_step": 260584, "epoch": 2927, "lr": 3.962291921572492e-07}
{"train_loss": 0.021242769435048103, "global_step": 260585, "epoch": 2927, "lr": 3.961563551720726e-07}
{"train_loss": 0.03397759422659874, "global_step": 260586, "epoch": 2927, "lr": 3.960835248555339e-07}
{"train_loss": 0.048413392156362534, "global_step": 260587, "epoch": 2927, "lr": 3.9601070120763304e-07}
{"train_loss": 0.032538387924432755, "global_step": 260588, "epoch": 2927, "lr": 3.959378842283923e-07}
{"train_loss": 0.047134753316640854, "global_step": 260589, "epoch": 2927, "lr": 3.9586507391780046e-07}
{"train_loss": 0.03658859804272652, "global_step": 260590, "epoch": 2927, "lr": 3.9579227027589097e-07}
{"train_loss": 0.04322167156326972, "global_step": 260591, "epoch": 2927, "lr": 3.9571947330265815e-07, "val_loss": 9.617914199829102}
{"train_loss": 0.06929798424243927, "global_step": 260592, "epoch": 2928, "lr": 3.956466829981187e-07}
{"train_loss": 0.037435147911310196, "global_step": 260593, "epoch": 2928, "lr": 3.955738993622726e-07}
{"train_loss": 0.03488790616393089, "global_step": 260594, "epoch": 2928, "lr": 3.9550112239514215e-07}
{"train_loss": 0.05613620951771736, "global_step": 260595, "epoch": 2928, "lr": 3.9542835209673835e-07}
{"train_loss": 0.03730297088623047, "global_step": 260596, "epoch": 2928, "lr": 3.9535558846705566e-07}
{"train_loss": 0.08154571056365967, "global_step": 260597, "epoch": 2928, "lr": 3.9528283150611635e-07}
{"train_loss": 0.015262008644640446, "global_step": 260598, "epoch": 2928, "lr": 3.952100812139259e-07}
{"train_loss": 0.019283805042505264, "global_step": 260599, "epoch": 2928, "lr": 3.951373375904954e-07}
{"train_loss": 0.06982353329658508, "global_step": 260600, "epoch": 2928, "lr": 3.950646006358305e-07}
{"train_loss": 0.052336398512125015, "global_step": 260601, "epoch": 2928, "lr": 3.9499187034994777e-07}
{"train_loss": 0.054485343396663666, "global_step": 260602, "epoch": 2928, "lr": 3.949191467328528e-07}
{"train_loss": 0.03839946910738945, "global_step": 260603, "epoch": 2928, "lr": 3.948464297845567e-07}
{"train_loss": 0.01588256284594536, "global_step": 260604, "epoch": 2928, "lr": 3.9477371950507047e-07}
{"train_loss": 0.0306536965072155, "global_step": 260605, "epoch": 2928, "lr": 3.947010158943998e-07}
{"train_loss": 0.06395584344863892, "global_step": 260606, "epoch": 2928, "lr": 3.946283189525557e-07}
{"train_loss": 0.04286247119307518, "global_step": 260607, "epoch": 2928, "lr": 3.9455562867955487e-07}
{"train_loss": 0.03540102019906044, "global_step": 260608, "epoch": 2928, "lr": 3.9448294507539727e-07}
{"train_loss": 0.07521352916955948, "global_step": 260609, "epoch": 2928, "lr": 3.9441026814009964e-07}
{"train_loss": 0.07771002501249313, "global_step": 260610, "epoch": 2928, "lr": 3.943375978736674e-07}
{"train_loss": 0.03170190006494522, "global_step": 260611, "epoch": 2928, "lr": 3.9426493427611177e-07}
{"train_loss": 0.03309077396988869, "global_step": 260612, "epoch": 2928, "lr": 3.9419227734743826e-07}
{"train_loss": 0.028123561292886734, "global_step": 260613, "epoch": 2928, "lr": 3.9411962708766903e-07}
{"train_loss": 0.029110711067914963, "global_step": 260614, "epoch": 2928, "lr": 3.9404698349679856e-07}
{"train_loss": 0.04493067041039467, "global_step": 260615, "epoch": 2928, "lr": 3.939743465748491e-07}
{"train_loss": 0.05991446599364281, "global_step": 260616, "epoch": 2928, "lr": 3.939017163218206e-07}
{"train_loss": 0.0693235695362091, "global_step": 260617, "epoch": 2928, "lr": 3.9382909273773527e-07}
{"train_loss": 0.04825792834162712, "global_step": 260618, "epoch": 2928, "lr": 3.9375647582258755e-07}
{"train_loss": 0.029189903289079666, "global_step": 260619, "epoch": 2928, "lr": 3.936838655763941e-07}
{"train_loss": 0.026732251048088074, "global_step": 260620, "epoch": 2928, "lr": 3.9361126199917167e-07}
{"train_loss": 0.0387052558362484, "global_step": 260621, "epoch": 2928, "lr": 3.935386650909145e-07}
{"train_loss": 0.06136516481637955, "global_step": 260622, "epoch": 2928, "lr": 3.9346607485165054e-07}
{"train_loss": 0.03463967144489288, "global_step": 260623, "epoch": 2928, "lr": 3.9339349128137414e-07}
{"train_loss": 0.06532548367977142, "global_step": 260624, "epoch": 2928, "lr": 3.9332091438010753e-07}
{"train_loss": 0.03984149172902107, "global_step": 260625, "epoch": 2928, "lr": 3.932483441478452e-07}
{"train_loss": 0.0549580454826355, "global_step": 260626, "epoch": 2928, "lr": 3.931757805846148e-07}
{"train_loss": 0.06755130738019943, "global_step": 260627, "epoch": 2928, "lr": 3.931032236904109e-07}
{"train_loss": 0.06711839139461517, "global_step": 260628, "epoch": 2928, "lr": 3.930306734652556e-07}
{"train_loss": 0.052635014057159424, "global_step": 260629, "epoch": 2928, "lr": 3.929581299091489e-07}
{"train_loss": 0.08407922089099884, "global_step": 260630, "epoch": 2928, "lr": 3.928855930221076e-07}
{"train_loss": 0.03495822474360466, "global_step": 260631, "epoch": 2928, "lr": 3.928130628041371e-07}
{"train_loss": 0.022402161732316017, "global_step": 260632, "epoch": 2928, "lr": 3.9274053925524857e-07}
{"train_loss": 0.05660543963313103, "global_step": 260633, "epoch": 2928, "lr": 3.926680223754475e-07}
{"train_loss": 0.03313222900032997, "global_step": 260634, "epoch": 2928, "lr": 3.9259551216475064e-07}
{"train_loss": 0.0491047240793705, "global_step": 260635, "epoch": 2928, "lr": 3.925230086231635e-07}
{"train_loss": 0.04363095387816429, "global_step": 260636, "epoch": 2928, "lr": 3.9245051175069714e-07}
{"train_loss": 0.04523397982120514, "global_step": 260637, "epoch": 2928, "lr": 3.923780215473627e-07}
{"train_loss": 0.06094451993703842, "global_step": 260638, "epoch": 2928, "lr": 3.923055380131657e-07}
{"train_loss": 0.050887253135442734, "global_step": 260639, "epoch": 2928, "lr": 3.922330611481229e-07}
{"train_loss": 0.029739590361714363, "global_step": 260640, "epoch": 2928, "lr": 3.9216059095223413e-07}
{"train_loss": 0.049378808587789536, "global_step": 260641, "epoch": 2928, "lr": 3.920881274255217e-07}
{"train_loss": 0.029269546270370483, "global_step": 260642, "epoch": 2928, "lr": 3.9201567056798006e-07}
{"train_loss": 0.02643345296382904, "global_step": 260643, "epoch": 2928, "lr": 3.919432203796369e-07}
{"train_loss": 0.027433127164840698, "global_step": 260644, "epoch": 2928, "lr": 3.918707768604812e-07}
{"train_loss": 0.025826197117567062, "global_step": 260645, "epoch": 2928, "lr": 3.9179834001054626e-07}
{"train_loss": 0.027317441999912262, "global_step": 260646, "epoch": 2928, "lr": 3.9172590982982093e-07}
{"train_loss": 0.01454550214111805, "global_step": 260647, "epoch": 2928, "lr": 3.9165348631832746e-07}
{"train_loss": 0.033724259585142136, "global_step": 260648, "epoch": 2928, "lr": 3.9158106947606574e-07}
{"train_loss": 0.03628699481487274, "global_step": 260649, "epoch": 2928, "lr": 3.915086593030581e-07}
{"train_loss": 0.012409472838044167, "global_step": 260650, "epoch": 2928, "lr": 3.9143625579929897e-07}
{"train_loss": 0.06558061391115189, "global_step": 260651, "epoch": 2928, "lr": 3.913638589648161e-07}
{"train_loss": 0.0323711559176445, "global_step": 260652, "epoch": 2928, "lr": 3.912914687996039e-07}
{"train_loss": 0.040691886097192764, "global_step": 260653, "epoch": 2928, "lr": 3.91219085303679e-07}
{"train_loss": 0.061885952949523926, "global_step": 260654, "epoch": 2928, "lr": 3.911467084770526e-07}
{"train_loss": 0.06731560826301575, "global_step": 260655, "epoch": 2928, "lr": 3.9107433831972464e-07}
{"train_loss": 0.0341963991522789, "global_step": 260656, "epoch": 2928, "lr": 3.9100197483172283e-07}
{"train_loss": 0.035060279071331024, "global_step": 260657, "epoch": 2928, "lr": 3.9092961801303617e-07}
{"train_loss": 0.028022294864058495, "global_step": 260658, "epoch": 2928, "lr": 3.9085726786369237e-07}
{"train_loss": 0.05782141163945198, "global_step": 260659, "epoch": 2928, "lr": 3.9078492438368584e-07}
{"train_loss": 0.06666196137666702, "global_step": 260660, "epoch": 2928, "lr": 3.9071258757303886e-07}
{"train_loss": 0.03965262323617935, "global_step": 260661, "epoch": 2928, "lr": 3.9064025743175693e-07}
{"train_loss": 0.05337188392877579, "global_step": 260662, "epoch": 2928, "lr": 3.905679339598456e-07}
{"train_loss": 0.008230605162680149, "global_step": 260663, "epoch": 2928, "lr": 3.90495617157316e-07}
{"train_loss": 0.02141420543193817, "global_step": 260664, "epoch": 2928, "lr": 3.904233070241847e-07}
{"train_loss": 0.004838231019675732, "global_step": 260665, "epoch": 2928, "lr": 3.9035100356045183e-07}
{"train_loss": 0.027017811313271523, "global_step": 260666, "epoch": 2928, "lr": 3.90278706766134e-07}
{"train_loss": 0.09067481011152267, "global_step": 260667, "epoch": 2928, "lr": 3.902064166412367e-07}
{"train_loss": 0.04343612864613533, "global_step": 260668, "epoch": 2928, "lr": 3.9013413318577664e-07}
{"train_loss": 0.038182180374860764, "global_step": 260669, "epoch": 2928, "lr": 3.900618563997538e-07}
{"train_loss": 0.031468264758586884, "global_step": 260670, "epoch": 2928, "lr": 3.899895862831848e-07}
{"train_loss": 0.03566829487681389, "global_step": 260671, "epoch": 2928, "lr": 3.8991732283606976e-07}
{"train_loss": 0.026247018948197365, "global_step": 260672, "epoch": 2928, "lr": 3.898450660584363e-07}
{"train_loss": 0.0895962044596672, "global_step": 260673, "epoch": 2928, "lr": 3.8977281595027893e-07}
{"train_loss": 0.02404121309518814, "global_step": 260674, "epoch": 2928, "lr": 3.8970057251160876e-07}
{"train_loss": 0.0755733847618103, "global_step": 260675, "epoch": 2928, "lr": 3.896283357424424e-07}
{"train_loss": 0.030643705278635025, "global_step": 260676, "epoch": 2928, "lr": 3.8955610564278547e-07}
{"train_loss": 0.04340187460184097, "global_step": 260677, "epoch": 2928, "lr": 3.8948388221264896e-07}
{"train_loss": 0.042346615344285965, "global_step": 260678, "epoch": 2928, "lr": 3.8941166545203856e-07}
{"train_loss": 0.042245179414749146, "global_step": 260679, "epoch": 2928, "lr": 3.893394553609708e-07}
{"train_loss": 0.04343920928331908, "global_step": 260680, "epoch": 2928, "lr": 3.892672519394458e-07, "val_loss": 9.662332534790039}
{"train_loss": 0.041280947625637054, "global_step": 260681, "epoch": 2929, "lr": 3.891950551874857e-07}
{"train_loss": 0.03370778262615204, "global_step": 260682, "epoch": 2929, "lr": 3.891228651050904e-07}
{"train_loss": 0.03699024021625519, "global_step": 260683, "epoch": 2929, "lr": 3.890506816922768e-07}
{"train_loss": 0.048627857118844986, "global_step": 260684, "epoch": 2929, "lr": 3.889785049490446e-07}
{"train_loss": 0.0371784083545208, "global_step": 260685, "epoch": 2929, "lr": 3.889063348754163e-07}
{"train_loss": 0.0337192639708519, "global_step": 260686, "epoch": 2929, "lr": 3.8883417147138613e-07}
{"train_loss": 0.04996044933795929, "global_step": 260687, "epoch": 2929, "lr": 3.887620147369819e-07}
{"train_loss": 0.03551230952143669, "global_step": 260688, "epoch": 2929, "lr": 3.8868986467219815e-07}
{"train_loss": 0.042477983981370926, "global_step": 260689, "epoch": 2929, "lr": 3.8861772127705144e-07}
{"train_loss": 0.07845699787139893, "global_step": 260690, "epoch": 2929, "lr": 3.8854558455154735e-07}
{"train_loss": 0.08243037015199661, "global_step": 260691, "epoch": 2929, "lr": 3.884734544957025e-07}
{"train_loss": 0.025103513151407242, "global_step": 260692, "epoch": 2929, "lr": 3.884013311095225e-07}
{"train_loss": 0.04598535969853401, "global_step": 260693, "epoch": 2929, "lr": 3.8832921439301285e-07}
{"train_loss": 0.04587971046566963, "global_step": 260694, "epoch": 2929, "lr": 3.882571043461958e-07}
{"train_loss": 0.07376346737146378, "global_step": 260695, "epoch": 2929, "lr": 3.8818500096906573e-07}
{"train_loss": 0.03486745432019234, "global_step": 260696, "epoch": 2929, "lr": 3.8811290426164493e-07}
{"train_loss": 0.038064438849687576, "global_step": 260697, "epoch": 2929, "lr": 3.8804081422393334e-07}
{"train_loss": 0.06320200115442276, "global_step": 260698, "epoch": 2929, "lr": 3.8796873085594765e-07}
{"train_loss": 0.04973370209336281, "global_step": 260699, "epoch": 2929, "lr": 3.878966541576934e-07}
{"train_loss": 0.05376677215099335, "global_step": 260700, "epoch": 2929, "lr": 3.878245841291817e-07}
{"train_loss": 0.04460154473781586, "global_step": 260701, "epoch": 2929, "lr": 3.877525207704236e-07}
{"train_loss": 0.05319242551922798, "global_step": 260702, "epoch": 2929, "lr": 3.8768046408143023e-07}
{"train_loss": 0.02050050161778927, "global_step": 260703, "epoch": 2929, "lr": 3.8760841406220163e-07}
{"train_loss": 0.02907099574804306, "global_step": 260704, "epoch": 2929, "lr": 3.8753637071275996e-07}
{"train_loss": 0.020542804151773453, "global_step": 260705, "epoch": 2929, "lr": 3.874643340331052e-07}
{"train_loss": 0.06616850942373276, "global_step": 260706, "epoch": 2929, "lr": 3.873923040232541e-07}
{"train_loss": 0.01789799891412258, "global_step": 260707, "epoch": 2929, "lr": 3.8732028068321213e-07}
{"train_loss": 0.028854820877313614, "global_step": 260708, "epoch": 2929, "lr": 3.872482640129904e-07}
{"train_loss": 0.02811707742512226, "global_step": 260709, "epoch": 2929, "lr": 3.8717625401260006e-07}
{"train_loss": 0.030356913805007935, "global_step": 260710, "epoch": 2929, "lr": 3.871042506820466e-07}
{"train_loss": 0.04210793599486351, "global_step": 260711, "epoch": 2929, "lr": 3.870322540213467e-07}
{"train_loss": 0.030916618183255196, "global_step": 260712, "epoch": 2929, "lr": 3.8696026403050036e-07}
{"train_loss": 0.02713111974298954, "global_step": 260713, "epoch": 2929, "lr": 3.8688828070952975e-07}
{"train_loss": 0.03598053753376007, "global_step": 260714, "epoch": 2929, "lr": 3.868163040584294e-07}
{"train_loss": 0.07334548979997635, "global_step": 260715, "epoch": 2929, "lr": 3.8674433407722143e-07}
{"train_loss": 0.02640051767230034, "global_step": 260716, "epoch": 2929, "lr": 3.866723707659059e-07}
{"train_loss": 0.023731164634227753, "global_step": 260717, "epoch": 2929, "lr": 3.86600414124505e-07}
{"train_loss": 0.06185593456029892, "global_step": 260718, "epoch": 2929, "lr": 3.8652846415301316e-07}
{"train_loss": 0.041592370718717575, "global_step": 260719, "epoch": 2929, "lr": 3.8645652085145256e-07}
{"train_loss": 0.06459062546491623, "global_step": 260720, "epoch": 2929, "lr": 3.863845842198288e-07}
{"train_loss": 0.015260225161910057, "global_step": 260721, "epoch": 2929, "lr": 3.8631265425814744e-07}
{"train_loss": 0.024042494595050812, "global_step": 260722, "epoch": 2929, "lr": 3.862407309664251e-07}
{"train_loss": 0.05087655410170555, "global_step": 260723, "epoch": 2929, "lr": 3.8616881434466735e-07}
{"train_loss": 0.08165931701660156, "global_step": 260724, "epoch": 2929, "lr": 3.860969043928797e-07}
{"train_loss": 0.038382939994335175, "global_step": 260725, "epoch": 2929, "lr": 3.860250011110844e-07}
{"train_loss": 0.044284265488386154, "global_step": 260726, "epoch": 2929, "lr": 3.859531044992759e-07}
{"train_loss": 0.09493172913789749, "global_step": 260727, "epoch": 2929, "lr": 3.858812145574764e-07}
{"train_loss": 0.05869855359196663, "global_step": 260728, "epoch": 2929, "lr": 3.8580933128568584e-07}
{"train_loss": 0.0685797855257988, "global_step": 260729, "epoch": 2929, "lr": 3.8573745468392096e-07}
{"train_loss": 0.044728200882673264, "global_step": 260730, "epoch": 2929, "lr": 3.8566558475219285e-07}
{"train_loss": 0.08337651938199997, "global_step": 260731, "epoch": 2929, "lr": 3.855937214905014e-07}
{"train_loss": 0.05149054154753685, "global_step": 260732, "epoch": 2929, "lr": 3.85521864898869e-07}
{"train_loss": 0.041331946849823, "global_step": 260733, "epoch": 2929, "lr": 3.8545001497728994e-07}
{"train_loss": 0.07402638345956802, "global_step": 260734, "epoch": 2929, "lr": 3.853781717257865e-07}
{"train_loss": 0.05469633638858795, "global_step": 260735, "epoch": 2929, "lr": 3.853063351443642e-07}
{"train_loss": 0.06883177906274796, "global_step": 260736, "epoch": 2929, "lr": 3.8523450523303417e-07}
{"train_loss": 0.07539638876914978, "global_step": 260737, "epoch": 2929, "lr": 3.8516268199180197e-07}
{"train_loss": 0.04687787592411041, "global_step": 260738, "epoch": 2929, "lr": 3.850908654206842e-07}
{"train_loss": 0.00346094137057662, "global_step": 260739, "epoch": 2929, "lr": 3.8501905551968087e-07}
{"train_loss": 0.016932664439082146, "global_step": 260740, "epoch": 2929, "lr": 3.8494725228881425e-07}
{"train_loss": 0.04947579652070999, "global_step": 260741, "epoch": 2929, "lr": 3.8487545572807873e-07}
{"train_loss": 0.009475299157202244, "global_step": 260742, "epoch": 2929, "lr": 3.848036658374965e-07}
{"train_loss": 0.040653880685567856, "global_step": 260743, "epoch": 2929, "lr": 3.847318826170676e-07}
{"train_loss": 0.05440211296081543, "global_step": 260744, "epoch": 2929, "lr": 3.8466010606681424e-07}
{"train_loss": 0.015196938998997211, "global_step": 260745, "epoch": 2929, "lr": 3.845883361867364e-07}
{"train_loss": 0.05084014683961868, "global_step": 260746, "epoch": 2929, "lr": 3.845165729768452e-07}
{"train_loss": 0.02400277927517891, "global_step": 260747, "epoch": 2929, "lr": 3.8444481643714615e-07}
{"train_loss": 0.06765176355838776, "global_step": 260748, "epoch": 2929, "lr": 3.8437306656766147e-07}
{"train_loss": 0.058930933475494385, "global_step": 260749, "epoch": 2929, "lr": 3.843013233683912e-07}
{"train_loss": 0.02842320315539837, "global_step": 260750, "epoch": 2929, "lr": 3.8422958683934086e-07}
{"train_loss": 0.04296554997563362, "global_step": 260751, "epoch": 2929, "lr": 3.841578569805326e-07}
{"train_loss": 0.007321691606193781, "global_step": 260752, "epoch": 2929, "lr": 3.8408613379196654e-07}
{"train_loss": 0.05897527560591698, "global_step": 260753, "epoch": 2929, "lr": 3.840144172736593e-07}
{"train_loss": 0.044028859585523605, "global_step": 260754, "epoch": 2929, "lr": 3.8394270742561077e-07}
{"train_loss": 0.0340532548725605, "global_step": 260755, "epoch": 2929, "lr": 3.8387100424784327e-07}
{"train_loss": 0.04381777346134186, "global_step": 260756, "epoch": 2929, "lr": 3.837993077403512e-07}
{"train_loss": 0.0723157450556755, "global_step": 260757, "epoch": 2929, "lr": 3.8372761790316237e-07}
{"train_loss": 0.0350116565823555, "global_step": 260758, "epoch": 2929, "lr": 3.836559347362656e-07}
{"train_loss": 0.05621116980910301, "global_step": 260759, "epoch": 2929, "lr": 3.835842582396887e-07}
{"train_loss": 0.06528446823358536, "global_step": 260760, "epoch": 2929, "lr": 3.8351258841343164e-07}
{"train_loss": 0.033735331147909164, "global_step": 260761, "epoch": 2929, "lr": 3.8344092525751106e-07}
{"train_loss": 0.06234167516231537, "global_step": 260762, "epoch": 2929, "lr": 3.8336926877192703e-07}
{"train_loss": 0.03901740908622742, "global_step": 260763, "epoch": 2929, "lr": 3.832976189566961e-07}
{"train_loss": 0.054491303861141205, "global_step": 260764, "epoch": 2929, "lr": 3.8322597581182394e-07}
{"train_loss": 0.08087573945522308, "global_step": 260765, "epoch": 2929, "lr": 3.831543393373216e-07}
{"train_loss": 0.10090398788452148, "global_step": 260766, "epoch": 2929, "lr": 3.830827095332057e-07}
{"train_loss": 0.02415870502591133, "global_step": 260767, "epoch": 2929, "lr": 3.8301108639947626e-07}
{"train_loss": 0.07925301790237427, "global_step": 260768, "epoch": 2929, "lr": 3.8293946993614435e-07}
{"train_loss": 0.0463428290330627, "global_step": 260769, "epoch": 2929, "lr": 3.8286786014322115e-07, "val_loss": 9.576592445373535}
{"train_loss": 0.05599689483642578, "global_step": 260770, "epoch": 2930, "lr": 3.827962570207233e-07}
{"train_loss": 0.06395470350980759, "global_step": 260771, "epoch": 2930, "lr": 3.827246605686452e-07}
{"train_loss": 0.050306860357522964, "global_step": 260772, "epoch": 2930, "lr": 3.826530707870091e-07}
{"train_loss": 0.029017331078648567, "global_step": 260773, "epoch": 2930, "lr": 3.8258148767582045e-07}
{"train_loss": 0.03527594357728958, "global_step": 260774, "epoch": 2930, "lr": 3.825099112350905e-07}
{"train_loss": 0.04451849311590195, "global_step": 260775, "epoch": 2930, "lr": 3.8243834146482473e-07}
{"train_loss": 0.06167031452059746, "global_step": 260776, "epoch": 2930, "lr": 3.823667783650342e-07}
{"train_loss": 0.06373191624879837, "global_step": 260777, "epoch": 2930, "lr": 3.822952219357356e-07}
{"train_loss": 0.04265172407031059, "global_step": 260778, "epoch": 2930, "lr": 3.82223672176929e-07}
{"train_loss": 0.03835202008485794, "global_step": 260779, "epoch": 2930, "lr": 3.821521290886254e-07}
{"train_loss": 0.050707656890153885, "global_step": 260780, "epoch": 2930, "lr": 3.820805926708415e-07}
{"train_loss": 0.05317842960357666, "global_step": 260781, "epoch": 2930, "lr": 3.820090629235773e-07}
{"train_loss": 0.06810451298952103, "global_step": 260782, "epoch": 2930, "lr": 3.819375398468494e-07}
{"train_loss": 0.04124465957283974, "global_step": 260783, "epoch": 2930, "lr": 3.818660234406635e-07}
{"train_loss": 0.040977735072374344, "global_step": 260784, "epoch": 2930, "lr": 3.8179451370503607e-07}
{"train_loss": 0.06219195947051048, "global_step": 260785, "epoch": 2930, "lr": 3.8172301063996163e-07}
{"train_loss": 0.035259928554296494, "global_step": 260786, "epoch": 2930, "lr": 3.8165151424546794e-07}
{"train_loss": 0.053114913403987885, "global_step": 260787, "epoch": 2930, "lr": 3.8158002452155507e-07}
{"train_loss": 0.07475152611732483, "global_step": 260788, "epoch": 2930, "lr": 3.81508541468234e-07}
{"train_loss": 0.052219703793525696, "global_step": 260789, "epoch": 2930, "lr": 3.814370650855159e-07}
{"train_loss": 0.05618203431367874, "global_step": 260790, "epoch": 2930, "lr": 3.813655953734063e-07}
{"train_loss": 0.01893577165901661, "global_step": 260791, "epoch": 2930, "lr": 3.812941323319219e-07}
{"train_loss": 0.033810608088970184, "global_step": 260792, "epoch": 2930, "lr": 3.812226759610626e-07}
{"train_loss": 0.06195507571101189, "global_step": 260793, "epoch": 2930, "lr": 3.811512262608452e-07}
{"train_loss": 0.026093892753124237, "global_step": 260794, "epoch": 2930, "lr": 3.810797832312751e-07}
{"train_loss": 0.013862109743058681, "global_step": 260795, "epoch": 2930, "lr": 3.8100834687236355e-07}
{"train_loss": 0.04788339138031006, "global_step": 260796, "epoch": 2930, "lr": 3.809369171841215e-07}
{"train_loss": 0.03567872568964958, "global_step": 260797, "epoch": 2930, "lr": 3.8086549416656014e-07}
{"train_loss": 0.028735749423503876, "global_step": 260798, "epoch": 2930, "lr": 3.8079407781968504e-07}
{"train_loss": 0.018033569678664207, "global_step": 260799, "epoch": 2930, "lr": 3.807226681435072e-07}
{"train_loss": 0.015837913379073143, "global_step": 260800, "epoch": 2930, "lr": 3.8065126513803783e-07}
{"train_loss": 0.05744599923491478, "global_step": 260801, "epoch": 2930, "lr": 3.8057986880328246e-07}
{"train_loss": 0.04696105793118477, "global_step": 260802, "epoch": 2930, "lr": 3.8050847913925216e-07}
{"train_loss": 0.024125127121806145, "global_step": 260803, "epoch": 2930, "lr": 3.8043709614596356e-07}
{"train_loss": 0.03127634525299072, "global_step": 260804, "epoch": 2930, "lr": 3.803657198234167e-07}
{"train_loss": 0.06457231193780899, "global_step": 260805, "epoch": 2930, "lr": 3.802943501716227e-07}
{"train_loss": 0.025333749130368233, "global_step": 260806, "epoch": 2930, "lr": 3.802229871905982e-07}
{"train_loss": 0.038141075521707535, "global_step": 260807, "epoch": 2930, "lr": 3.801516308803432e-07}
{"train_loss": 0.06994952261447906, "global_step": 260808, "epoch": 2930, "lr": 3.8008028124087993e-07}
{"train_loss": 0.011766827665269375, "global_step": 260809, "epoch": 2930, "lr": 3.8000893827220276e-07}
{"train_loss": 0.03527645021677017, "global_step": 260810, "epoch": 2930, "lr": 3.7993760197432837e-07}
{"train_loss": 0.03667081892490387, "global_step": 260811, "epoch": 2930, "lr": 3.798662723472679e-07}
{"train_loss": 0.05377605929970741, "global_step": 260812, "epoch": 2930, "lr": 3.7979494939103244e-07}
{"train_loss": 0.022783607244491577, "global_step": 260813, "epoch": 2930, "lr": 3.797236331056275e-07}
{"train_loss": 0.021262045949697495, "global_step": 260814, "epoch": 2930, "lr": 3.796523234910643e-07}
{"train_loss": 0.03119608759880066, "global_step": 260815, "epoch": 2930, "lr": 3.7958102054734825e-07}
{"train_loss": 0.035526204854249954, "global_step": 260816, "epoch": 2930, "lr": 3.795097242745016e-07}
{"train_loss": 0.029764477163553238, "global_step": 260817, "epoch": 2930, "lr": 3.794384346725133e-07}
{"train_loss": 0.029400644823908806, "global_step": 260818, "epoch": 2930, "lr": 3.7936715174141104e-07}
{"train_loss": 0.017190972343087196, "global_step": 260819, "epoch": 2930, "lr": 3.792958754812004e-07}
{"train_loss": 0.054410114884376526, "global_step": 260820, "epoch": 2930, "lr": 3.7922460589188134e-07}
{"train_loss": 0.02090270258486271, "global_step": 260821, "epoch": 2930, "lr": 3.791533429734762e-07}
{"train_loss": 0.03839384391903877, "global_step": 260822, "epoch": 2930, "lr": 3.790820867259903e-07}
{"train_loss": 0.057906463742256165, "global_step": 260823, "epoch": 2930, "lr": 3.7901083714942945e-07}
{"train_loss": 0.06352658569812775, "global_step": 260824, "epoch": 2930, "lr": 3.789395942438101e-07}
{"train_loss": 0.05141448602080345, "global_step": 260825, "epoch": 2930, "lr": 3.788683580091323e-07}
{"train_loss": 0.03145122155547142, "global_step": 260826, "epoch": 2930, "lr": 3.787971284454128e-07}
{"train_loss": 0.043569955974817276, "global_step": 260827, "epoch": 2930, "lr": 3.787259055526571e-07}
{"train_loss": 0.020331356674432755, "global_step": 260828, "epoch": 2930, "lr": 3.786546893308762e-07}
{"train_loss": 0.045592814683914185, "global_step": 260829, "epoch": 2930, "lr": 3.785834797800869e-07}
{"train_loss": 0.029418449848890305, "global_step": 260830, "epoch": 2930, "lr": 3.7851227690028354e-07}
{"train_loss": 0.08735039830207825, "global_step": 260831, "epoch": 2930, "lr": 3.78441080691494e-07}
{"train_loss": 0.020348625257611275, "global_step": 260832, "epoch": 2930, "lr": 3.783698911537126e-07}
{"train_loss": 0.029501773416996002, "global_step": 260833, "epoch": 2930, "lr": 3.7829870828695604e-07}
{"train_loss": 0.025324221700429916, "global_step": 260834, "epoch": 2930, "lr": 3.7822753209122986e-07}
{"train_loss": 0.0471356175839901, "global_step": 260835, "epoch": 2930, "lr": 3.7815636256654516e-07}
{"train_loss": 0.038619257509708405, "global_step": 260836, "epoch": 2930, "lr": 3.7808519971291314e-07}
{"train_loss": 0.037577833980321884, "global_step": 260837, "epoch": 2930, "lr": 3.7801404353034473e-07}
{"train_loss": 0.06711938977241516, "global_step": 260838, "epoch": 2930, "lr": 3.7794289401885117e-07}
{"train_loss": 0.043776482343673706, "global_step": 260839, "epoch": 2930, "lr": 3.7787175117843245e-07}
{"train_loss": 0.0361025333404541, "global_step": 260840, "epoch": 2930, "lr": 3.7780061500910514e-07}
{"train_loss": 0.049475204199552536, "global_step": 260841, "epoch": 2930, "lr": 3.7772948551088037e-07}
{"train_loss": 0.05142053961753845, "global_step": 260842, "epoch": 2930, "lr": 3.7765836268376374e-07}
{"train_loss": 0.04315020143985748, "global_step": 260843, "epoch": 2930, "lr": 3.775872465277663e-07}
{"train_loss": 0.06849829107522964, "global_step": 260844, "epoch": 2930, "lr": 3.775161370428992e-07}
{"train_loss": 0.039023257791996, "global_step": 260845, "epoch": 2930, "lr": 3.7744503422916245e-07}
{"train_loss": 0.025910211727023125, "global_step": 260846, "epoch": 2930, "lr": 3.773739380865837e-07}
{"train_loss": 0.04458282142877579, "global_step": 260847, "epoch": 2930, "lr": 3.773028486151575e-07}
{"train_loss": 0.03404978662729263, "global_step": 260848, "epoch": 2930, "lr": 3.772317658149005e-07}
{"train_loss": 0.02452627755701542, "global_step": 260849, "epoch": 2930, "lr": 3.7716068968581265e-07}
{"train_loss": 0.027223240584135056, "global_step": 260850, "epoch": 2930, "lr": 3.770896202279162e-07}
{"train_loss": 0.03529145568609238, "global_step": 260851, "epoch": 2930, "lr": 3.7701855744121663e-07}
{"train_loss": 0.03775405138731003, "global_step": 260852, "epoch": 2930, "lr": 3.7694750132571957e-07}
{"train_loss": 0.004377281758934259, "global_step": 260853, "epoch": 2930, "lr": 3.7687645188143605e-07}
{"train_loss": 0.036320555955171585, "global_step": 260854, "epoch": 2930, "lr": 3.7680540910837725e-07}
{"train_loss": 0.02624858170747757, "global_step": 260855, "epoch": 2930, "lr": 3.7673437300655423e-07}
{"train_loss": 0.04128330200910568, "global_step": 260856, "epoch": 2930, "lr": 3.7666334357597255e-07}
{"train_loss": 0.07586327195167542, "global_step": 260857, "epoch": 2930, "lr": 3.765923208166489e-07}
{"train_loss": 0.040849862290609086, "global_step": 260858, "epoch": 2930, "lr": 3.765213047285776e-07, "val_loss": 9.616073608398438, "train_action_mse_error": 8.211527824401855}
{"train_loss": 0.0498550683259964, "global_step": 260859, "epoch": 2931, "lr": 3.7645029531178655e-07}
{"train_loss": 0.06904343515634537, "global_step": 260860, "epoch": 2931, "lr": 3.763792925662757e-07}
{"train_loss": 0.022021789103746414, "global_step": 260861, "epoch": 2931, "lr": 3.7630829649205057e-07}
{"train_loss": 0.022638551890850067, "global_step": 260862, "epoch": 2931, "lr": 3.762373070891334e-07}
{"train_loss": 0.054125163704156876, "global_step": 260863, "epoch": 2931, "lr": 3.761663243575242e-07}
{"train_loss": 0.054109156131744385, "global_step": 260864, "epoch": 2931, "lr": 3.76095348297234e-07}
{"train_loss": 0.017056986689567566, "global_step": 260865, "epoch": 2931, "lr": 3.7602437890826845e-07}
{"train_loss": 0.012639377266168594, "global_step": 260866, "epoch": 2931, "lr": 3.7595341619064417e-07}
{"train_loss": 0.08195512741804123, "global_step": 260867, "epoch": 2931, "lr": 3.7588246014437224e-07}
{"train_loss": 0.044607847929000854, "global_step": 260868, "epoch": 2931, "lr": 3.7581151076945267e-07}
{"train_loss": 0.049070898443460464, "global_step": 260869, "epoch": 2931, "lr": 3.7574056806590763e-07}
{"train_loss": 0.025600435212254524, "global_step": 260870, "epoch": 2931, "lr": 3.756696320337316e-07}
{"train_loss": 0.050082460045814514, "global_step": 260871, "epoch": 2931, "lr": 3.755987026729468e-07}
{"train_loss": 0.0664590448141098, "global_step": 260872, "epoch": 2931, "lr": 3.755277799835533e-07}
{"train_loss": 0.02371688187122345, "global_step": 260873, "epoch": 2931, "lr": 3.7545686396556754e-07}
{"train_loss": 0.032025109976530075, "global_step": 260874, "epoch": 2931, "lr": 3.753859546189953e-07}
{"train_loss": 0.06744342297315598, "global_step": 260875, "epoch": 2931, "lr": 3.7531505194384754e-07}
{"train_loss": 0.028892818838357925, "global_step": 260876, "epoch": 2931, "lr": 3.7524415594014093e-07}
{"train_loss": 0.0212932750582695, "global_step": 260877, "epoch": 2931, "lr": 3.7517326660787e-07}
{"train_loss": 0.06219416856765747, "global_step": 260878, "epoch": 2931, "lr": 3.751023839470569e-07}
{"train_loss": 0.02847057394683361, "global_step": 260879, "epoch": 2931, "lr": 3.7503150795770156e-07}
{"train_loss": 0.0403607115149498, "global_step": 260880, "epoch": 2931, "lr": 3.749606386398208e-07}
{"train_loss": 0.06506521254777908, "global_step": 260881, "epoch": 2931, "lr": 3.7488977599342557e-07}
{"train_loss": 0.05176122486591339, "global_step": 260882, "epoch": 2931, "lr": 3.7481892001851595e-07}
{"train_loss": 0.031053142622113228, "global_step": 260883, "epoch": 2931, "lr": 3.7474807071510855e-07}
{"train_loss": 0.040876854211091995, "global_step": 260884, "epoch": 2931, "lr": 3.746772280832145e-07}
{"train_loss": 0.04894300550222397, "global_step": 260885, "epoch": 2931, "lr": 3.746063921228338e-07}
{"train_loss": 0.032850462943315506, "global_step": 260886, "epoch": 2931, "lr": 3.745355628339886e-07}
{"train_loss": 0.04545307159423828, "global_step": 260887, "epoch": 2931, "lr": 3.74464740216679e-07}
{"train_loss": 0.04678254947066307, "global_step": 260888, "epoch": 2931, "lr": 3.743939242709216e-07}
{"train_loss": 0.035580866038799286, "global_step": 260889, "epoch": 2931, "lr": 3.743231149967164e-07}
{"train_loss": 0.02483769878745079, "global_step": 260890, "epoch": 2931, "lr": 3.7425231239408e-07}
{"train_loss": 0.01472366601228714, "global_step": 260891, "epoch": 2931, "lr": 3.741815164630236e-07}
{"train_loss": 0.016220103949308395, "global_step": 260892, "epoch": 2931, "lr": 3.7411072720355265e-07}
{"train_loss": 0.027329104021191597, "global_step": 260893, "epoch": 2931, "lr": 3.740399446156784e-07}
{"train_loss": 0.08456315100193024, "global_step": 260894, "epoch": 2931, "lr": 3.7396916869940623e-07}
{"train_loss": 0.02746245265007019, "global_step": 260895, "epoch": 2931, "lr": 3.7389839945475293e-07}
{"train_loss": 0.03566202148795128, "global_step": 260896, "epoch": 2931, "lr": 3.7382763688172393e-07}
{"train_loss": 0.04514233022928238, "global_step": 260897, "epoch": 2931, "lr": 3.737568809803249e-07}
{"train_loss": 0.036524590104818344, "global_step": 260898, "epoch": 2931, "lr": 3.736861317505724e-07}
{"train_loss": 0.036096397787332535, "global_step": 260899, "epoch": 2931, "lr": 3.7361538919247207e-07}
{"train_loss": 0.026794949546456337, "global_step": 260900, "epoch": 2931, "lr": 3.735446533060405e-07}
{"train_loss": 0.0702885240316391, "global_step": 260901, "epoch": 2931, "lr": 3.7347392409127216e-07}
{"train_loss": 0.01383389625698328, "global_step": 260902, "epoch": 2931, "lr": 3.734032015481892e-07}
{"train_loss": 0.07382774353027344, "global_step": 260903, "epoch": 2931, "lr": 3.7333248567679725e-07}
{"train_loss": 0.04407564178109169, "global_step": 260904, "epoch": 2931, "lr": 3.732617764771018e-07}
{"train_loss": 0.03470413386821747, "global_step": 260905, "epoch": 2931, "lr": 3.7319107394912514e-07}
{"train_loss": 0.04284996911883354, "global_step": 260906, "epoch": 2931, "lr": 3.731203780928616e-07}
{"train_loss": 0.04140955209732056, "global_step": 260907, "epoch": 2931, "lr": 3.730496889083279e-07}
{"train_loss": 0.04748082906007767, "global_step": 260908, "epoch": 2931, "lr": 3.7297900639552965e-07}
{"train_loss": 0.04638392850756645, "global_step": 260909, "epoch": 2931, "lr": 3.729083305544834e-07}
{"train_loss": 0.06307276338338852, "global_step": 260910, "epoch": 2931, "lr": 3.728376613851947e-07}
{"train_loss": 0.06151454895734787, "global_step": 260911, "epoch": 2931, "lr": 3.7276699888767474e-07}
{"train_loss": 0.03874154016375542, "global_step": 260912, "epoch": 2931, "lr": 3.7269634306192903e-07}
{"train_loss": 0.036024414002895355, "global_step": 260913, "epoch": 2931, "lr": 3.726256939079686e-07}
{"train_loss": 0.058802831918001175, "global_step": 260914, "epoch": 2931, "lr": 3.725550514258047e-07}
{"train_loss": 0.06496582925319672, "global_step": 260915, "epoch": 2931, "lr": 3.7248441561544835e-07}
{"train_loss": 0.057610321789979935, "global_step": 260916, "epoch": 2931, "lr": 3.7241378647690507e-07}
{"train_loss": 0.018419787287712097, "global_step": 260917, "epoch": 2931, "lr": 3.7234316401018044e-07}
{"train_loss": 0.026859385892748833, "global_step": 260918, "epoch": 2931, "lr": 3.722725482152967e-07}
{"train_loss": 0.019154028967022896, "global_step": 260919, "epoch": 2931, "lr": 3.7220193909225375e-07}
{"train_loss": 0.04023546352982521, "global_step": 260920, "epoch": 2931, "lr": 3.721313366410628e-07}
{"train_loss": 0.04357666149735451, "global_step": 260921, "epoch": 2931, "lr": 3.7206074086173495e-07}
{"train_loss": 0.04430568963289261, "global_step": 260922, "epoch": 2931, "lr": 3.719901517542812e-07}
{"train_loss": 0.051322951912879944, "global_step": 260923, "epoch": 2931, "lr": 3.7191956931870163e-07}
{"train_loss": 0.046648990362882614, "global_step": 260924, "epoch": 2931, "lr": 3.7184899355501843e-07}
{"train_loss": 0.01336140651255846, "global_step": 260925, "epoch": 2931, "lr": 3.7177842446323164e-07}
{"train_loss": 0.024940522387623787, "global_step": 260926, "epoch": 2931, "lr": 3.7170786204335227e-07}
{"train_loss": 0.038894668221473694, "global_step": 260927, "epoch": 2931, "lr": 3.7163730629539706e-07}
{"train_loss": 0.039410293102264404, "global_step": 260928, "epoch": 2931, "lr": 3.7156675721936594e-07}
{"train_loss": 0.08770068734884262, "global_step": 260929, "epoch": 2931, "lr": 3.7149621481528116e-07}
{"train_loss": 0.025453977286815643, "global_step": 260930, "epoch": 2931, "lr": 3.7142567908313716e-07}
{"train_loss": 0.03748471662402153, "global_step": 260931, "epoch": 2931, "lr": 3.713551500229506e-07}
{"train_loss": 0.07782482355833054, "global_step": 260932, "epoch": 2931, "lr": 3.7128462763472703e-07}
{"train_loss": 0.06334022432565689, "global_step": 260933, "epoch": 2931, "lr": 3.7121411191848863e-07}
{"train_loss": 0.019415317103266716, "global_step": 260934, "epoch": 2931, "lr": 3.7114360287422987e-07}
{"train_loss": 0.06805675476789474, "global_step": 260935, "epoch": 2931, "lr": 3.710731005019674e-07}
{"train_loss": 0.04102141037583351, "global_step": 260936, "epoch": 2931, "lr": 3.710026048017068e-07}
{"train_loss": 0.04913807660341263, "global_step": 260937, "epoch": 2931, "lr": 3.709321157734591e-07}
{"train_loss": 0.0346960574388504, "global_step": 260938, "epoch": 2931, "lr": 3.70861633417241e-07}
{"train_loss": 0.072249636054039, "global_step": 260939, "epoch": 2931, "lr": 3.7079115773304694e-07}
{"train_loss": 0.06333580613136292, "global_step": 260940, "epoch": 2931, "lr": 3.707206887209047e-07}
{"train_loss": 0.022429585456848145, "global_step": 260941, "epoch": 2931, "lr": 3.7065022638080873e-07}
{"train_loss": 0.0365196131169796, "global_step": 260942, "epoch": 2931, "lr": 3.705797707127756e-07}
{"train_loss": 0.04755597189068794, "global_step": 260943, "epoch": 2931, "lr": 3.70509321716811e-07}
{"train_loss": 0.04206174612045288, "global_step": 260944, "epoch": 2931, "lr": 3.7043887939292586e-07}
{"train_loss": 0.0387125127017498, "global_step": 260945, "epoch": 2931, "lr": 3.7036844374113145e-07}
{"train_loss": 0.026211896911263466, "global_step": 260946, "epoch": 2931, "lr": 3.7029801476143876e-07}
{"train_loss": 0.0429964131476839, "global_step": 260947, "epoch": 2931, "lr": 3.7022759245385344e-07, "val_loss": 9.63269329071045}
{"train_loss": 0.03346385434269905, "global_step": 260948, "epoch": 2932, "lr": 3.7015717681838645e-07}
{"train_loss": 0.06291482597589493, "global_step": 260949, "epoch": 2932, "lr": 3.7008676785504347e-07}
{"train_loss": 0.0690394788980484, "global_step": 260950, "epoch": 2932, "lr": 3.700163655638411e-07}
{"train_loss": 0.016460685059428215, "global_step": 260951, "epoch": 2932, "lr": 3.699459699447794e-07}
{"train_loss": 0.0445474311709404, "global_step": 260952, "epoch": 2932, "lr": 3.6987558099788044e-07}
{"train_loss": 0.04983673244714737, "global_step": 260953, "epoch": 2932, "lr": 3.698051987231443e-07}
{"train_loss": 0.03611975163221359, "global_step": 260954, "epoch": 2932, "lr": 3.6973482312058217e-07}
{"train_loss": 0.06726811081171036, "global_step": 260955, "epoch": 2932, "lr": 3.69664454190205e-07}
{"train_loss": 0.0479513444006443, "global_step": 260956, "epoch": 2932, "lr": 3.69594091932024e-07}
{"train_loss": 0.020881474018096924, "global_step": 260957, "epoch": 2932, "lr": 3.695237363460391e-07}
{"train_loss": 0.037562668323516846, "global_step": 260958, "epoch": 2932, "lr": 3.6945338743227254e-07}
{"train_loss": 0.04128085821866989, "global_step": 260959, "epoch": 2932, "lr": 3.693830451907299e-07}
{"train_loss": 0.060150545090436935, "global_step": 260960, "epoch": 2932, "lr": 3.693127096214166e-07}
{"train_loss": 0.03590063378214836, "global_step": 260961, "epoch": 2932, "lr": 3.6924238072433837e-07}
{"train_loss": 0.036372750997543335, "global_step": 260962, "epoch": 2932, "lr": 3.691720584995173e-07}
{"train_loss": 0.03957849740982056, "global_step": 260963, "epoch": 2932, "lr": 3.691017429469534e-07}
{"train_loss": 0.04189978167414665, "global_step": 260964, "epoch": 2932, "lr": 3.6903143406665787e-07}
{"train_loss": 0.08037062734365463, "global_step": 260965, "epoch": 2932, "lr": 3.689611318586472e-07}
{"train_loss": 0.018234584480524063, "global_step": 260966, "epoch": 2932, "lr": 3.6889083632292155e-07}
{"train_loss": 0.04025210440158844, "global_step": 260967, "epoch": 2932, "lr": 3.688205474594919e-07}
{"train_loss": 0.053831856697797775, "global_step": 260968, "epoch": 2932, "lr": 3.6875026526836944e-07}
{"train_loss": 0.01723880134522915, "global_step": 260969, "epoch": 2932, "lr": 3.686799897495652e-07}
{"train_loss": 0.03801435977220535, "global_step": 260970, "epoch": 2932, "lr": 3.686097209030848e-07}
{"train_loss": 0.01684517040848732, "global_step": 260971, "epoch": 2932, "lr": 3.685394587289448e-07}
{"train_loss": 0.027478639036417007, "global_step": 260972, "epoch": 2932, "lr": 3.6846920322714526e-07}
{"train_loss": 0.055949319154024124, "global_step": 260973, "epoch": 2932, "lr": 3.683989543977029e-07}
{"train_loss": 0.056698355823755264, "global_step": 260974, "epoch": 2932, "lr": 3.683287122406176e-07}
{"train_loss": 0.030650300905108452, "global_step": 260975, "epoch": 2932, "lr": 3.6825847675591165e-07}
{"train_loss": 0.07818924635648727, "global_step": 260976, "epoch": 2932, "lr": 3.6818824794359053e-07}
{"train_loss": 0.024176055565476418, "global_step": 260977, "epoch": 2932, "lr": 3.6811802580365986e-07}
{"train_loss": 0.02837555482983589, "global_step": 260978, "epoch": 2932, "lr": 3.680478103361307e-07}
{"train_loss": 0.017841294407844543, "global_step": 260979, "epoch": 2932, "lr": 3.679776015410086e-07}
{"train_loss": 0.034400418400764465, "global_step": 260980, "epoch": 2932, "lr": 3.679073994183102e-07}
{"train_loss": 0.03874124214053154, "global_step": 260981, "epoch": 2932, "lr": 3.678372039680411e-07}
{"train_loss": 0.04224449023604393, "global_step": 260982, "epoch": 2932, "lr": 3.677670151902124e-07}
{"train_loss": 0.0349162258207798, "global_step": 260983, "epoch": 2932, "lr": 3.676968330848296e-07}
{"train_loss": 0.06248555704951286, "global_step": 260984, "epoch": 2932, "lr": 3.676266576519094e-07}
{"train_loss": 0.07392334192991257, "global_step": 260985, "epoch": 2932, "lr": 3.675564888914518e-07}
{"train_loss": 0.02976372465491295, "global_step": 260986, "epoch": 2932, "lr": 3.674863268034734e-07}
{"train_loss": 0.07942531257867813, "global_step": 260987, "epoch": 2932, "lr": 3.674161713879853e-07}
{"train_loss": 0.036526672542095184, "global_step": 260988, "epoch": 2932, "lr": 3.673460226449876e-07}
{"train_loss": 0.058343555778265, "global_step": 260989, "epoch": 2932, "lr": 3.672758805744969e-07}
{"train_loss": 0.017675289884209633, "global_step": 260990, "epoch": 2932, "lr": 3.672057451765243e-07}
{"train_loss": 0.08564919978380203, "global_step": 260991, "epoch": 2932, "lr": 3.6713561645106974e-07}
{"train_loss": 0.07009574770927429, "global_step": 260992, "epoch": 2932, "lr": 3.6706549439815554e-07}
{"train_loss": 0.07909433543682098, "global_step": 260993, "epoch": 2932, "lr": 3.6699537901778156e-07}
{"train_loss": 0.10204263776540756, "global_step": 260994, "epoch": 2932, "lr": 3.6692527030995907e-07}
{"train_loss": 0.042430199682712555, "global_step": 260995, "epoch": 2932, "lr": 3.66855168274699e-07}
{"train_loss": 0.07364102452993393, "global_step": 260996, "epoch": 2932, "lr": 3.667850729120126e-07}
{"train_loss": 0.047170139849185944, "global_step": 260997, "epoch": 2932, "lr": 3.6671498422190533e-07}
{"train_loss": 0.05466481298208237, "global_step": 260998, "epoch": 2932, "lr": 3.666449022043883e-07}
{"train_loss": 0.047102320939302444, "global_step": 260999, "epoch": 2932, "lr": 3.665748268594671e-07}
{"train_loss": 0.04584567993879318, "global_step": 261000, "epoch": 2932, "lr": 3.6650475818715834e-07}
{"train_loss": 0.030897952616214752, "global_step": 261001, "epoch": 2932, "lr": 3.664346961874732e-07}
{"train_loss": 0.04726755619049072, "global_step": 261002, "epoch": 2932, "lr": 3.66364640860406e-07}
{"train_loss": 0.050394605845212936, "global_step": 261003, "epoch": 2932, "lr": 3.662945922059846e-07}
{"train_loss": 0.029672065749764442, "global_step": 261004, "epoch": 2932, "lr": 3.662245502242034e-07}
{"train_loss": 0.040697336196899414, "global_step": 261005, "epoch": 2932, "lr": 3.661545149150847e-07}
{"train_loss": 0.03181081265211105, "global_step": 261006, "epoch": 2932, "lr": 3.6608448627862834e-07}
{"train_loss": 0.047791991382837296, "global_step": 261007, "epoch": 2932, "lr": 3.6601446431484556e-07}
{"train_loss": 0.05016708746552467, "global_step": 261008, "epoch": 2932, "lr": 3.6594444902374736e-07}
{"train_loss": 0.030611097812652588, "global_step": 261009, "epoch": 2932, "lr": 3.658744404053449e-07}
{"train_loss": 0.09130031615495682, "global_step": 261010, "epoch": 2932, "lr": 3.6580443845964376e-07}
{"train_loss": 0.04498725384473801, "global_step": 261011, "epoch": 2932, "lr": 3.657344431866605e-07}
{"train_loss": 0.03323690965771675, "global_step": 261012, "epoch": 2932, "lr": 3.656644545863952e-07}
{"train_loss": 0.07661642879247665, "global_step": 261013, "epoch": 2932, "lr": 3.655944726588589e-07}
{"train_loss": 0.0595572255551815, "global_step": 261014, "epoch": 2932, "lr": 3.6552449740406835e-07}
{"train_loss": 0.0735366940498352, "global_step": 261015, "epoch": 2932, "lr": 3.6545452882202345e-07}
{"train_loss": 0.04893366992473602, "global_step": 261016, "epoch": 2932, "lr": 3.6538456691274095e-07}
{"train_loss": 0.06809915602207184, "global_step": 261017, "epoch": 2932, "lr": 3.653146116762263e-07}
{"train_loss": 0.01624579168856144, "global_step": 261018, "epoch": 2932, "lr": 3.652446631124962e-07}
{"train_loss": 0.041440948843955994, "global_step": 261019, "epoch": 2932, "lr": 3.651747212215451e-07}
{"train_loss": 0.04754351079463959, "global_step": 261020, "epoch": 2932, "lr": 3.651047860033951e-07}
{"train_loss": 0.04575764760375023, "global_step": 261021, "epoch": 2932, "lr": 3.65034857458052e-07}
{"train_loss": 0.00861138105392456, "global_step": 261022, "epoch": 2932, "lr": 3.649649355855267e-07}
{"train_loss": 0.057051852345466614, "global_step": 261023, "epoch": 2932, "lr": 3.648950203858248e-07}
{"train_loss": 0.036848943680524826, "global_step": 261024, "epoch": 2932, "lr": 3.648251118589574e-07}
{"train_loss": 0.026293091475963593, "global_step": 261025, "epoch": 2932, "lr": 3.6475521000493564e-07}
{"train_loss": 0.035201482474803925, "global_step": 261026, "epoch": 2932, "lr": 3.64685314823765e-07}
{"train_loss": 0.03616096079349518, "global_step": 261027, "epoch": 2932, "lr": 3.6461542631545664e-07}
{"train_loss": 0.06386867165565491, "global_step": 261028, "epoch": 2932, "lr": 3.645455444800272e-07}
{"train_loss": 0.03986825793981552, "global_step": 261029, "epoch": 2932, "lr": 3.6447566931747113e-07}
{"train_loss": 0.014522109180688858, "global_step": 261030, "epoch": 2932, "lr": 3.644058008278162e-07}
{"train_loss": 0.03589310497045517, "global_step": 261031, "epoch": 2932, "lr": 3.643359390110512e-07}
{"train_loss": 0.044584982097148895, "global_step": 261032, "epoch": 2932, "lr": 3.642660838672041e-07}
{"train_loss": 0.02173696644604206, "global_step": 261033, "epoch": 2932, "lr": 3.6419623539627467e-07}
{"train_loss": 0.09187688678503036, "global_step": 261034, "epoch": 2932, "lr": 3.6412639359827415e-07}
{"train_loss": 0.07514970749616623, "global_step": 261035, "epoch": 2932, "lr": 3.640565584732081e-07}
{"train_loss": 0.04625025563193171, "global_step": 261036, "epoch": 2932, "lr": 3.639867300210931e-07, "val_loss": 9.588800430297852}
{"train_loss": 0.05000581592321396, "global_step": 261037, "epoch": 2933, "lr": 3.639169082419347e-07}
{"train_loss": 0.03224610164761543, "global_step": 261038, "epoch": 2933, "lr": 3.638470931357441e-07}
{"train_loss": 0.08057110011577606, "global_step": 261039, "epoch": 2933, "lr": 3.637772847025267e-07}
{"train_loss": 0.07825959473848343, "global_step": 261040, "epoch": 2933, "lr": 3.6370748294229376e-07}
{"train_loss": 0.029209166765213013, "global_step": 261041, "epoch": 2933, "lr": 3.636376878550618e-07}
{"train_loss": 0.03389750421047211, "global_step": 261042, "epoch": 2933, "lr": 3.6356789944082535e-07}
{"train_loss": 0.03281339257955551, "global_step": 261043, "epoch": 2933, "lr": 3.6349811769960664e-07}
{"train_loss": 0.040351249277591705, "global_step": 261044, "epoch": 2933, "lr": 3.634283426314111e-07}
{"train_loss": 0.05219672992825508, "global_step": 261045, "epoch": 2933, "lr": 3.6335857423625e-07}
{"train_loss": 0.03971355780959129, "global_step": 261046, "epoch": 2933, "lr": 3.6328881251412316e-07}
{"train_loss": 0.11470186710357666, "global_step": 261047, "epoch": 2933, "lr": 3.632190574650529e-07}
{"train_loss": 0.06177908182144165, "global_step": 261048, "epoch": 2933, "lr": 3.631493090890392e-07}
{"train_loss": 0.032785046845674515, "global_step": 261049, "epoch": 2933, "lr": 3.630795673860987e-07}
{"train_loss": 0.01510191522538662, "global_step": 261050, "epoch": 2933, "lr": 3.6300983235623145e-07}
{"train_loss": 0.07207843661308289, "global_step": 261051, "epoch": 2933, "lr": 3.6294010399945954e-07}
{"train_loss": 0.012246559374034405, "global_step": 261052, "epoch": 2933, "lr": 3.6287038231578307e-07}
{"train_loss": 0.012703181244432926, "global_step": 261053, "epoch": 2933, "lr": 3.628006673052131e-07}
{"train_loss": 0.03325403109192848, "global_step": 261054, "epoch": 2933, "lr": 3.6273095896776077e-07}
{"train_loss": 0.036173202097415924, "global_step": 261055, "epoch": 2933, "lr": 3.626612573034316e-07}
{"train_loss": 0.04507340490818024, "global_step": 261056, "epoch": 2933, "lr": 3.625915623122367e-07}
{"train_loss": 0.021334437653422356, "global_step": 261057, "epoch": 2933, "lr": 3.6252187399418715e-07}
{"train_loss": 0.08228202909231186, "global_step": 261058, "epoch": 2933, "lr": 3.624521923492941e-07}
{"train_loss": 0.026589572429656982, "global_step": 261059, "epoch": 2933, "lr": 3.62382517377563e-07}
{"train_loss": 0.052887145429849625, "global_step": 261060, "epoch": 2933, "lr": 3.6231284907900507e-07}
{"train_loss": 0.05542619898915291, "global_step": 261061, "epoch": 2933, "lr": 3.622431874536314e-07}
{"train_loss": 0.06331527978181839, "global_step": 261062, "epoch": 2933, "lr": 3.621735325014475e-07}
{"train_loss": 0.02714032121002674, "global_step": 261063, "epoch": 2933, "lr": 3.621038842224589e-07}
{"train_loss": 0.03630413860082626, "global_step": 261064, "epoch": 2933, "lr": 3.6203424261668785e-07}
{"train_loss": 0.03303099423646927, "global_step": 261065, "epoch": 2933, "lr": 3.619646076841343e-07}
{"train_loss": 0.007239116821438074, "global_step": 261066, "epoch": 2933, "lr": 3.6189497942480945e-07}
{"train_loss": 0.04121590778231621, "global_step": 261067, "epoch": 2933, "lr": 3.6182535783871873e-07}
{"train_loss": 0.04001858830451965, "global_step": 261068, "epoch": 2933, "lr": 3.617557429258789e-07}
{"train_loss": 0.022966021671891212, "global_step": 261069, "epoch": 2933, "lr": 3.6168613468629543e-07}
{"train_loss": 0.041609231382608414, "global_step": 261070, "epoch": 2933, "lr": 3.6161653311997947e-07}
{"train_loss": 0.02641056850552559, "global_step": 261071, "epoch": 2933, "lr": 3.6154693822693655e-07}
{"train_loss": 0.028812112286686897, "global_step": 261072, "epoch": 2933, "lr": 3.614773500071833e-07}
{"train_loss": 0.10614016652107239, "global_step": 261073, "epoch": 2933, "lr": 3.6140776846071976e-07}
{"train_loss": 0.03978954628109932, "global_step": 261074, "epoch": 2933, "lr": 3.613381935875626e-07}
{"train_loss": 0.05309485271573067, "global_step": 261075, "epoch": 2933, "lr": 3.612686253877173e-07}
{"train_loss": 0.02716889977455139, "global_step": 261076, "epoch": 2933, "lr": 3.6119906386119504e-07}
{"train_loss": 0.04646821320056915, "global_step": 261077, "epoch": 2933, "lr": 3.611295090080069e-07}
{"train_loss": 0.009568720124661922, "global_step": 261078, "epoch": 2933, "lr": 3.6105996082815284e-07}
{"train_loss": 0.06069773808121681, "global_step": 261079, "epoch": 2933, "lr": 3.6099041932165514e-07}
{"train_loss": 0.015212610363960266, "global_step": 261080, "epoch": 2933, "lr": 3.609208844885192e-07}
{"train_loss": 0.04542739689350128, "global_step": 261081, "epoch": 2933, "lr": 3.608513563287508e-07}
{"train_loss": 0.07205608487129211, "global_step": 261082, "epoch": 2933, "lr": 3.607818348423553e-07}
{"train_loss": 0.024188777431845665, "global_step": 261083, "epoch": 2933, "lr": 3.6071232002935494e-07}
{"train_loss": 0.06405595690011978, "global_step": 261084, "epoch": 2933, "lr": 3.606428118897442e-07}
{"train_loss": 0.05233074724674225, "global_step": 261085, "epoch": 2933, "lr": 3.605733104235454e-07}
{"train_loss": 0.04079803079366684, "global_step": 261086, "epoch": 2933, "lr": 3.6050381563076384e-07}
{"train_loss": 0.034434013068675995, "global_step": 261087, "epoch": 2933, "lr": 3.604343275114053e-07}
{"train_loss": 0.03467543050646782, "global_step": 261088, "epoch": 2933, "lr": 3.603648460654807e-07}
{"train_loss": 0.034613799303770065, "global_step": 261089, "epoch": 2933, "lr": 3.602953712930013e-07}
{"train_loss": 0.056760262697935104, "global_step": 261090, "epoch": 2933, "lr": 3.6022590319397256e-07}
{"train_loss": 0.09714809060096741, "global_step": 261091, "epoch": 2933, "lr": 3.601564417684111e-07}
{"train_loss": 0.010991279035806656, "global_step": 261092, "epoch": 2933, "lr": 3.6008698701631705e-07}
{"train_loss": 0.031974103301763535, "global_step": 261093, "epoch": 2933, "lr": 3.6001753893770696e-07}
{"train_loss": 0.03249436616897583, "global_step": 261094, "epoch": 2933, "lr": 3.599480975325864e-07}
{"train_loss": 0.027195677161216736, "global_step": 261095, "epoch": 2933, "lr": 3.598786628009665e-07}
{"train_loss": 0.016461534425616264, "global_step": 261096, "epoch": 2933, "lr": 3.598092347428528e-07}
{"train_loss": 0.047080617398023605, "global_step": 261097, "epoch": 2933, "lr": 3.597398133582619e-07}
{"train_loss": 0.04757336527109146, "global_step": 261098, "epoch": 2933, "lr": 3.596703986471939e-07}
{"train_loss": 0.10253481566905975, "global_step": 261099, "epoch": 2933, "lr": 3.5960099060966537e-07}
{"train_loss": 0.08967708796262741, "global_step": 261100, "epoch": 2933, "lr": 3.595315892456874e-07}
{"train_loss": 0.04304289072751999, "global_step": 261101, "epoch": 2933, "lr": 3.594621945552601e-07}
{"train_loss": 0.03390040621161461, "global_step": 261102, "epoch": 2933, "lr": 3.593928065384056e-07}
{"train_loss": 0.03765355795621872, "global_step": 261103, "epoch": 2933, "lr": 3.593234251951183e-07}
{"train_loss": 0.042025111615657806, "global_step": 261104, "epoch": 2933, "lr": 3.592540505254205e-07}
{"train_loss": 0.12510542571544647, "global_step": 261105, "epoch": 2933, "lr": 3.591846825293066e-07}
{"train_loss": 0.040091462433338165, "global_step": 261106, "epoch": 2933, "lr": 3.5911532120680433e-07}
{"train_loss": 0.03936455771327019, "global_step": 261107, "epoch": 2933, "lr": 3.590459665579082e-07}
{"train_loss": 0.042933229357004166, "global_step": 261108, "epoch": 2933, "lr": 3.5897661858263486e-07}
{"train_loss": 0.04223001003265381, "global_step": 261109, "epoch": 2933, "lr": 3.589072772809954e-07}
{"train_loss": 0.05731812119483948, "global_step": 261110, "epoch": 2933, "lr": 3.5883794265299533e-07}
{"train_loss": 0.028639670461416245, "global_step": 261111, "epoch": 2933, "lr": 3.587686146986402e-07}
{"train_loss": 0.033038247376680374, "global_step": 261112, "epoch": 2933, "lr": 3.5869929341794673e-07}
{"train_loss": 0.04459768161177635, "global_step": 261113, "epoch": 2933, "lr": 3.5862997881092044e-07}
{"train_loss": 0.03433482348918915, "global_step": 261114, "epoch": 2933, "lr": 3.585606708775724e-07}
{"train_loss": 0.03555731847882271, "global_step": 261115, "epoch": 2933, "lr": 3.5849136961791373e-07}
{"train_loss": 0.039952557533979416, "global_step": 261116, "epoch": 2933, "lr": 3.584220750319445e-07}
{"train_loss": 0.026702405884861946, "global_step": 261117, "epoch": 2933, "lr": 3.5835278711968123e-07}
{"train_loss": 0.059270940721035004, "global_step": 261118, "epoch": 2933, "lr": 3.5828350588113514e-07}
{"train_loss": 0.03144390136003494, "global_step": 261119, "epoch": 2933, "lr": 3.582142313163117e-07}
{"train_loss": 0.027951467782258987, "global_step": 261120, "epoch": 2933, "lr": 3.581449634252221e-07}
{"train_loss": 0.06927261501550674, "global_step": 261121, "epoch": 2933, "lr": 3.580757022078718e-07}
{"train_loss": 0.04155391454696655, "global_step": 261122, "epoch": 2933, "lr": 3.580064476642775e-07}
{"train_loss": 0.05213062837719917, "global_step": 261123, "epoch": 2933, "lr": 3.5793719979444474e-07}
{"train_loss": 0.04461118206381798, "global_step": 261124, "epoch": 2933, "lr": 3.5786795859837353e-07}
{"train_loss": 0.044131129493497394, "global_step": 261125, "epoch": 2933, "lr": 3.577987240760916e-07, "val_loss": 9.52175235748291}
{"train_loss": 0.03740499168634415, "global_step": 261126, "epoch": 2934, "lr": 3.5772949622759345e-07}
{"train_loss": 0.03192334622144699, "global_step": 261127, "epoch": 2934, "lr": 3.5766027505289565e-07}
{"train_loss": 0.06663540005683899, "global_step": 261128, "epoch": 2934, "lr": 3.5759106055200385e-07}
{"train_loss": 0.030098890885710716, "global_step": 261129, "epoch": 2934, "lr": 3.5752185272493467e-07}
{"train_loss": 0.05154938995838165, "global_step": 261130, "epoch": 2934, "lr": 3.574526515716825e-07}
{"train_loss": 0.06015929579734802, "global_step": 261131, "epoch": 2934, "lr": 3.573834570922696e-07}
{"train_loss": 0.056465838104486465, "global_step": 261132, "epoch": 2934, "lr": 3.573142692867071e-07}
{"train_loss": 0.052807584404945374, "global_step": 261133, "epoch": 2934, "lr": 3.5724508815498937e-07}
{"train_loss": 0.014774071052670479, "global_step": 261134, "epoch": 2934, "lr": 3.571759136971442e-07}
{"train_loss": 0.04160512983798981, "global_step": 261135, "epoch": 2934, "lr": 3.571067459131605e-07}
{"train_loss": 0.06183597445487976, "global_step": 261136, "epoch": 2934, "lr": 3.5703758480306606e-07}
{"train_loss": 0.0205693319439888, "global_step": 261137, "epoch": 2934, "lr": 3.569684303668608e-07}
{"train_loss": 0.03957695513963699, "global_step": 261138, "epoch": 2934, "lr": 3.568992826045614e-07}
{"train_loss": 0.025982443243265152, "global_step": 261139, "epoch": 2934, "lr": 3.568301415161623e-07}
{"train_loss": 0.047710735350847244, "global_step": 261140, "epoch": 2934, "lr": 3.5676100710169136e-07}
{"train_loss": 0.041648149490356445, "global_step": 261141, "epoch": 2934, "lr": 3.566918793611429e-07}
{"train_loss": 0.03449542820453644, "global_step": 261142, "epoch": 2934, "lr": 3.566227582945336e-07}
{"train_loss": 0.048063647001981735, "global_step": 261143, "epoch": 2934, "lr": 3.565536439018746e-07}
{"train_loss": 0.05888547748327255, "global_step": 261144, "epoch": 2934, "lr": 3.564845361831659e-07}
{"train_loss": 0.06129993498325348, "global_step": 261145, "epoch": 2934, "lr": 3.564154351384241e-07}
{"train_loss": 0.018916180357336998, "global_step": 261146, "epoch": 2934, "lr": 3.563463407676604e-07}
{"train_loss": 0.06830530613660812, "global_step": 261147, "epoch": 2934, "lr": 3.5627725307088025e-07}
{"train_loss": 0.05567287281155586, "global_step": 261148, "epoch": 2934, "lr": 3.562081720480892e-07}
{"train_loss": 0.05159885436296463, "global_step": 261149, "epoch": 2934, "lr": 3.561390976993095e-07}
{"train_loss": 0.03974802792072296, "global_step": 261150, "epoch": 2934, "lr": 3.560700300245357e-07}
{"train_loss": 0.01697850041091442, "global_step": 261151, "epoch": 2934, "lr": 3.560009690237842e-07}
{"train_loss": 0.02006327547132969, "global_step": 261152, "epoch": 2934, "lr": 3.559319146970608e-07}
{"train_loss": 0.03814942017197609, "global_step": 261153, "epoch": 2934, "lr": 3.5586286704438197e-07}
{"train_loss": 0.06759181618690491, "global_step": 261154, "epoch": 2934, "lr": 3.5579382606574786e-07}
{"train_loss": 0.03254532068967819, "global_step": 261155, "epoch": 2934, "lr": 3.55724791761175e-07}
{"train_loss": 0.0628795251250267, "global_step": 261156, "epoch": 2934, "lr": 3.5565576413066906e-07}
{"train_loss": 0.07048580050468445, "global_step": 261157, "epoch": 2934, "lr": 3.55586743174241e-07}
{"train_loss": 0.039013393223285675, "global_step": 261158, "epoch": 2934, "lr": 3.555177288918965e-07}
{"train_loss": 0.05457475036382675, "global_step": 261159, "epoch": 2934, "lr": 3.5544872128365216e-07}
{"train_loss": 0.05755645036697388, "global_step": 261160, "epoch": 2934, "lr": 3.5537972034950796e-07}
{"train_loss": 0.023583708330988884, "global_step": 261161, "epoch": 2934, "lr": 3.5531072608948057e-07}
{"train_loss": 0.009055234491825104, "global_step": 261162, "epoch": 2934, "lr": 3.5524173850357554e-07}
{"train_loss": 0.05576430633664131, "global_step": 261163, "epoch": 2934, "lr": 3.55172757591804e-07}
{"train_loss": 0.03985399007797241, "global_step": 261164, "epoch": 2934, "lr": 3.5510378335417704e-07}
{"train_loss": 0.007194016128778458, "global_step": 261165, "epoch": 2934, "lr": 3.550348157907002e-07}
{"train_loss": 0.018728449940681458, "global_step": 261166, "epoch": 2934, "lr": 3.54965854901379e-07}
{"train_loss": 0.07725898921489716, "global_step": 261167, "epoch": 2934, "lr": 3.5489690068623016e-07}
{"train_loss": 0.030514895915985107, "global_step": 261168, "epoch": 2934, "lr": 3.548279531452592e-07}
{"train_loss": 0.053466636687517166, "global_step": 261169, "epoch": 2934, "lr": 3.547590122784827e-07}
{"train_loss": 0.04036913812160492, "global_step": 261170, "epoch": 2934, "lr": 3.546900780859008e-07}
{"train_loss": 0.045544758439064026, "global_step": 261171, "epoch": 2934, "lr": 3.54621150567519e-07}
{"train_loss": 0.021352963522076607, "global_step": 261172, "epoch": 2934, "lr": 3.5455222972335946e-07}
{"train_loss": 0.024681204929947853, "global_step": 261173, "epoch": 2934, "lr": 3.544833155534277e-07}
{"train_loss": 0.051988158375024796, "global_step": 261174, "epoch": 2934, "lr": 3.544144080577239e-07}
{"train_loss": 0.053035397082567215, "global_step": 261175, "epoch": 2934, "lr": 3.5434550723627005e-07}
{"train_loss": 0.03724130243062973, "global_step": 261176, "epoch": 2934, "lr": 3.542766130890662e-07}
{"train_loss": 0.039452824741601944, "global_step": 261177, "epoch": 2934, "lr": 3.542077256161236e-07}
{"train_loss": 0.03258373960852623, "global_step": 261178, "epoch": 2934, "lr": 3.541388448174532e-07}
{"train_loss": 0.0603184849023819, "global_step": 261179, "epoch": 2934, "lr": 3.540699706930661e-07}
{"train_loss": 0.05385030061006546, "global_step": 261180, "epoch": 2934, "lr": 3.5400110324296245e-07}
{"train_loss": 0.0670359879732132, "global_step": 261181, "epoch": 2934, "lr": 3.539322424671698e-07}
{"train_loss": 0.05298454686999321, "global_step": 261182, "epoch": 2934, "lr": 3.5386338836567723e-07}
{"train_loss": 0.04581114649772644, "global_step": 261183, "epoch": 2934, "lr": 3.5379454093850684e-07}
{"train_loss": 0.026023155078291893, "global_step": 261184, "epoch": 2934, "lr": 3.5372570018565866e-07}
{"train_loss": 0.04026006907224655, "global_step": 261185, "epoch": 2934, "lr": 3.536568661071493e-07}
{"train_loss": 0.019076919183135033, "global_step": 261186, "epoch": 2934, "lr": 3.535880387029844e-07}
{"train_loss": 0.0799245685338974, "global_step": 261187, "epoch": 2934, "lr": 3.5351921797318053e-07}
{"train_loss": 0.04718887805938721, "global_step": 261188, "epoch": 2934, "lr": 3.5345040391773776e-07}
{"train_loss": 0.06227394938468933, "global_step": 261189, "epoch": 2934, "lr": 3.5338159653666713e-07}
{"train_loss": 0.016210464760661125, "global_step": 261190, "epoch": 2934, "lr": 3.533127958299798e-07}
{"train_loss": 0.09580209851264954, "global_step": 261191, "epoch": 2934, "lr": 3.532440017976812e-07}
{"train_loss": 0.03569624572992325, "global_step": 261192, "epoch": 2934, "lr": 3.531752144397882e-07}
{"train_loss": 0.03499947488307953, "global_step": 261193, "epoch": 2934, "lr": 3.5310643375630613e-07}
{"train_loss": 0.04224175214767456, "global_step": 261194, "epoch": 2934, "lr": 3.530376597472407e-07}
{"train_loss": 0.037436045706272125, "global_step": 261195, "epoch": 2934, "lr": 3.5296889241260846e-07}
{"train_loss": 0.01834721304476261, "global_step": 261196, "epoch": 2934, "lr": 3.5290013175240946e-07}
{"train_loss": 0.029764406383037567, "global_step": 261197, "epoch": 2934, "lr": 3.528313777666603e-07}
{"train_loss": 0.05273928865790367, "global_step": 261198, "epoch": 2934, "lr": 3.527626304553666e-07}
{"train_loss": 0.05081460252404213, "global_step": 261199, "epoch": 2934, "lr": 3.5269388981853944e-07}
{"train_loss": 0.03271912410855293, "global_step": 261200, "epoch": 2934, "lr": 3.5262515585618996e-07}
{"train_loss": 0.05892959609627724, "global_step": 261201, "epoch": 2934, "lr": 3.5255642856831807e-07}
{"train_loss": 0.054591644555330276, "global_step": 261202, "epoch": 2934, "lr": 3.52487707954946e-07}
{"train_loss": 0.02392750419676304, "global_step": 261203, "epoch": 2934, "lr": 3.5241899401607936e-07}
{"train_loss": 0.05805390328168869, "global_step": 261204, "epoch": 2934, "lr": 3.523502867517181e-07}
{"train_loss": 0.04802318662405014, "global_step": 261205, "epoch": 2934, "lr": 3.5228158616188447e-07}
{"train_loss": 0.04868290573358536, "global_step": 261206, "epoch": 2934, "lr": 3.522128922465784e-07}
{"train_loss": 0.023203318938612938, "global_step": 261207, "epoch": 2934, "lr": 3.52144205005811e-07}
{"train_loss": 0.019107144325971603, "global_step": 261208, "epoch": 2934, "lr": 3.5207552443959345e-07}
{"train_loss": 0.05009588971734047, "global_step": 261209, "epoch": 2934, "lr": 3.520068505479368e-07}
{"train_loss": 0.051840707659721375, "global_step": 261210, "epoch": 2934, "lr": 3.519381833308466e-07}
{"train_loss": 0.025789065286517143, "global_step": 261211, "epoch": 2934, "lr": 3.5186952278832843e-07}
{"train_loss": 0.04416155070066452, "global_step": 261212, "epoch": 2934, "lr": 3.5180086892039885e-07}
{"train_loss": 0.06533458828926086, "global_step": 261213, "epoch": 2934, "lr": 3.517322217270691e-07}
{"train_loss": 0.04324701003562868, "global_step": 261214, "epoch": 2934, "lr": 3.5166358120833906e-07, "val_loss": 9.674138069152832}
{"train_loss": 0.03890186920762062, "global_step": 261215, "epoch": 2935, "lr": 3.515949473642199e-07}
{"train_loss": 0.031641289591789246, "global_step": 261216, "epoch": 2935, "lr": 3.5152632019473384e-07}
{"train_loss": 0.038282059133052826, "global_step": 261217, "epoch": 2935, "lr": 3.514576996998697e-07}
{"train_loss": 0.05715782195329666, "global_step": 261218, "epoch": 2935, "lr": 3.513890858796498e-07}
{"train_loss": 0.0332350917160511, "global_step": 261219, "epoch": 2935, "lr": 3.513204787340851e-07}
{"train_loss": 0.04295849800109863, "global_step": 261220, "epoch": 2935, "lr": 3.5125187826317574e-07}
{"train_loss": 0.05217377841472626, "global_step": 261221, "epoch": 2935, "lr": 3.5118328446693274e-07}
{"train_loss": 0.09182759374380112, "global_step": 261222, "epoch": 2935, "lr": 3.511146973453727e-07}
{"train_loss": 0.028011435642838478, "global_step": 261223, "epoch": 2935, "lr": 3.510461168985013e-07}
{"train_loss": 0.057672008872032166, "global_step": 261224, "epoch": 2935, "lr": 3.50977543126324e-07}
{"train_loss": 0.031008265912532806, "global_step": 261225, "epoch": 2935, "lr": 3.5090897602885196e-07}
{"train_loss": 0.028038837015628815, "global_step": 261226, "epoch": 2935, "lr": 3.508404156060963e-07}
{"train_loss": 0.08728957921266556, "global_step": 261227, "epoch": 2935, "lr": 3.5077186185806797e-07}
{"train_loss": 0.023425396531820297, "global_step": 261228, "epoch": 2935, "lr": 3.507033147847671e-07}
{"train_loss": 0.033678703010082245, "global_step": 261229, "epoch": 2935, "lr": 3.5063477438621597e-07}
{"train_loss": 0.10651455074548721, "global_step": 261230, "epoch": 2935, "lr": 3.5056624066240884e-07}
{"train_loss": 0.03904653340578079, "global_step": 261231, "epoch": 2935, "lr": 3.5049771361336804e-07}
{"train_loss": 0.052966199815273285, "global_step": 261232, "epoch": 2935, "lr": 3.504291932390935e-07}
{"train_loss": 0.03963007032871246, "global_step": 261233, "epoch": 2935, "lr": 3.5036067953960747e-07}
{"train_loss": 0.04513281211256981, "global_step": 261234, "epoch": 2935, "lr": 3.502921725148989e-07}
{"train_loss": 0.05350974574685097, "global_step": 261235, "epoch": 2935, "lr": 3.502236721649954e-07}
{"train_loss": 0.011404857970774174, "global_step": 261236, "epoch": 2935, "lr": 3.501551784898971e-07}
{"train_loss": 0.08024226129055023, "global_step": 261237, "epoch": 2935, "lr": 3.5008669148961505e-07}
{"train_loss": 0.01161801628768444, "global_step": 261238, "epoch": 2935, "lr": 3.5001821116416036e-07}
{"train_loss": 0.021792074665427208, "global_step": 261239, "epoch": 2935, "lr": 3.499497375135441e-07}
{"train_loss": 0.02975103072822094, "global_step": 261240, "epoch": 2935, "lr": 3.498812705377663e-07}
{"train_loss": 0.03741057589650154, "global_step": 261241, "epoch": 2935, "lr": 3.4981281023684366e-07}
{"train_loss": 0.03266220539808273, "global_step": 261242, "epoch": 2935, "lr": 3.497443566107816e-07}
{"train_loss": 0.0346672348678112, "global_step": 261243, "epoch": 2935, "lr": 3.4967590965959697e-07}
{"train_loss": 0.05708321928977966, "global_step": 261244, "epoch": 2935, "lr": 3.4960746938328405e-07}
{"train_loss": 0.03082176484167576, "global_step": 261245, "epoch": 2935, "lr": 3.4953903578187064e-07}
{"train_loss": 0.08474931865930557, "global_step": 261246, "epoch": 2935, "lr": 3.4947060885535123e-07}
{"train_loss": 0.0736570954322815, "global_step": 261247, "epoch": 2935, "lr": 3.494021886037424e-07}
{"train_loss": 0.02488758973777294, "global_step": 261248, "epoch": 2935, "lr": 3.493337750270498e-07}
{"train_loss": 0.06183439865708351, "global_step": 261249, "epoch": 2935, "lr": 3.492653681252844e-07}
{"train_loss": 0.06947535276412964, "global_step": 261250, "epoch": 2935, "lr": 3.4919696789845747e-07}
{"train_loss": 0.07834147661924362, "global_step": 261251, "epoch": 2935, "lr": 3.491285743465744e-07}
{"train_loss": 0.04776714742183685, "global_step": 261252, "epoch": 2935, "lr": 3.4906018746964644e-07}
{"train_loss": 0.04746623709797859, "global_step": 261253, "epoch": 2935, "lr": 3.48991807267679e-07}
{"train_loss": 0.07520122081041336, "global_step": 261254, "epoch": 2935, "lr": 3.4892343374068324e-07}
{"train_loss": 0.010241646319627762, "global_step": 261255, "epoch": 2935, "lr": 3.4885506688867585e-07}
{"train_loss": 0.03162233158946037, "global_step": 261256, "epoch": 2935, "lr": 3.4878670671165125e-07}
{"train_loss": 0.13085640966892242, "global_step": 261257, "epoch": 2935, "lr": 3.487183532096372e-07}
{"train_loss": 0.04335590824484825, "global_step": 261258, "epoch": 2935, "lr": 3.4865000638262814e-07}
{"train_loss": 0.015296628698706627, "global_step": 261259, "epoch": 2935, "lr": 3.4858166623063513e-07}
{"train_loss": 0.04999605938792229, "global_step": 261260, "epoch": 2935, "lr": 3.4851333275367494e-07}
{"train_loss": 0.03738686442375183, "global_step": 261261, "epoch": 2935, "lr": 3.4844500595174746e-07}
{"train_loss": 0.06748977303504944, "global_step": 261262, "epoch": 2935, "lr": 3.4837668582486937e-07}
{"train_loss": 0.04417584836483002, "global_step": 261263, "epoch": 2935, "lr": 3.483083723730463e-07}
{"train_loss": 0.05622130259871483, "global_step": 261264, "epoch": 2935, "lr": 3.482400655962892e-07}
{"train_loss": 0.025906987488269806, "global_step": 261265, "epoch": 2935, "lr": 3.4817176549460374e-07}
{"train_loss": 0.03497575223445892, "global_step": 261266, "epoch": 2935, "lr": 3.4810347206800096e-07}
{"train_loss": 0.06321839243173599, "global_step": 261267, "epoch": 2935, "lr": 3.4803518531649205e-07}
{"train_loss": 0.03819483146071434, "global_step": 261268, "epoch": 2935, "lr": 3.47966905240088e-07}
{"train_loss": 0.023699956014752388, "global_step": 261269, "epoch": 2935, "lr": 3.478986318387889e-07}
{"train_loss": 0.05015263706445694, "global_step": 261270, "epoch": 2935, "lr": 3.4783036511261135e-07}
{"train_loss": 0.023580532521009445, "global_step": 261271, "epoch": 2935, "lr": 3.477621050615609e-07}
{"train_loss": 0.07488120347261429, "global_step": 261272, "epoch": 2935, "lr": 3.476938516856543e-07}
{"train_loss": 0.06279550492763519, "global_step": 261273, "epoch": 2935, "lr": 3.476256049848914e-07}
{"train_loss": 0.03348051756620407, "global_step": 261274, "epoch": 2935, "lr": 3.4755736495928894e-07}
{"train_loss": 0.03792356327176094, "global_step": 261275, "epoch": 2935, "lr": 3.474891316088469e-07}
{"train_loss": 0.02383236773312092, "global_step": 261276, "epoch": 2935, "lr": 3.4742090493358195e-07}
{"train_loss": 0.02864118665456772, "global_step": 261277, "epoch": 2935, "lr": 3.473526849334996e-07}
{"train_loss": 0.03596814349293709, "global_step": 261278, "epoch": 2935, "lr": 3.47284471608611e-07}
{"train_loss": 0.03672713041305542, "global_step": 261279, "epoch": 2935, "lr": 3.472162649589272e-07}
{"train_loss": 0.06390213221311569, "global_step": 261280, "epoch": 2935, "lr": 3.471480649844538e-07}
{"train_loss": 0.049872081726789474, "global_step": 261281, "epoch": 2935, "lr": 3.4707987168520197e-07}
{"train_loss": 0.05222633481025696, "global_step": 261282, "epoch": 2935, "lr": 3.470116850611771e-07}
{"train_loss": 0.05697154998779297, "global_step": 261283, "epoch": 2935, "lr": 3.4694350511239594e-07}
{"train_loss": 0.013958542607724667, "global_step": 261284, "epoch": 2935, "lr": 3.4687533183885844e-07}
{"train_loss": 0.048256438225507736, "global_step": 261285, "epoch": 2935, "lr": 3.4680716524058134e-07}
{"train_loss": 0.057187117636203766, "global_step": 261286, "epoch": 2935, "lr": 3.467390053175701e-07}
{"train_loss": 0.07591172307729721, "global_step": 261287, "epoch": 2935, "lr": 3.4667085206983584e-07}
{"train_loss": 0.025304557755589485, "global_step": 261288, "epoch": 2935, "lr": 3.4660270549738416e-07}
{"train_loss": 0.059249840676784515, "global_step": 261289, "epoch": 2935, "lr": 3.4653456560022614e-07}
{"train_loss": 0.059193287044763565, "global_step": 261290, "epoch": 2935, "lr": 3.464664323783728e-07}
{"train_loss": 0.017833532765507698, "global_step": 261291, "epoch": 2935, "lr": 3.4639830583182984e-07}
{"train_loss": 0.023233788087964058, "global_step": 261292, "epoch": 2935, "lr": 3.463301859606083e-07}
{"train_loss": 0.020307334139943123, "global_step": 261293, "epoch": 2935, "lr": 3.462620727647192e-07}
{"train_loss": 0.04789147153496742, "global_step": 261294, "epoch": 2935, "lr": 3.461939662441682e-07}
{"train_loss": 0.10191232711076736, "global_step": 261295, "epoch": 2935, "lr": 3.461258663989719e-07}
{"train_loss": 0.058286480605602264, "global_step": 261296, "epoch": 2935, "lr": 3.4605777322913034e-07}
{"train_loss": 0.03531044349074364, "global_step": 261297, "epoch": 2935, "lr": 3.459896867346546e-07}
{"train_loss": 0.01994878612458706, "global_step": 261298, "epoch": 2935, "lr": 3.4592160691555576e-07}
{"train_loss": 0.040595658123493195, "global_step": 261299, "epoch": 2935, "lr": 3.4585353377184495e-07}
{"train_loss": 0.06513801217079163, "global_step": 261300, "epoch": 2935, "lr": 3.457854673035277e-07}
{"train_loss": 0.03916795179247856, "global_step": 261301, "epoch": 2935, "lr": 3.457174075106151e-07}
{"train_loss": 0.04109513387084007, "global_step": 261302, "epoch": 2935, "lr": 3.4564935439311277e-07}
{"train_loss": 0.04624217225427038, "global_step": 261303, "epoch": 2935, "lr": 3.4558130795103174e-07, "val_loss": 9.546191215515137, "train_action_mse_error": 16.050304412841797}
{"train_loss": 0.03227614983916283, "global_step": 261304, "epoch": 2936, "lr": 3.455132681843831e-07}
{"train_loss": 0.03526084125041962, "global_step": 261305, "epoch": 2936, "lr": 3.4544523509317806e-07}
{"train_loss": 0.020470030605793, "global_step": 261306, "epoch": 2936, "lr": 3.453772086774221e-07}
{"train_loss": 0.03228825703263283, "global_step": 261307, "epoch": 2936, "lr": 3.453091889371207e-07}
{"train_loss": 0.027127137407660484, "global_step": 261308, "epoch": 2936, "lr": 3.452411758722962e-07}
{"train_loss": 0.05515608936548233, "global_step": 261309, "epoch": 2936, "lr": 3.45173169482943e-07}
{"train_loss": 0.02162395976483822, "global_step": 261310, "epoch": 2936, "lr": 3.451051697690777e-07}
{"train_loss": 0.04577723518013954, "global_step": 261311, "epoch": 2936, "lr": 3.4503717673070033e-07}
{"train_loss": 0.05807022005319595, "global_step": 261312, "epoch": 2936, "lr": 3.449691903678387e-07}
{"train_loss": 0.04525495320558548, "global_step": 261313, "epoch": 2936, "lr": 3.449012106804816e-07}
{"train_loss": 0.04300679266452789, "global_step": 261314, "epoch": 2936, "lr": 3.448332376686514e-07}
{"train_loss": 0.02211911603808403, "global_step": 261315, "epoch": 2936, "lr": 3.447652713323535e-07}
{"train_loss": 0.07250611484050751, "global_step": 261316, "epoch": 2936, "lr": 3.44697311671599e-07}
{"train_loss": 0.023152384907007217, "global_step": 261317, "epoch": 2936, "lr": 3.44629358686388e-07}
{"train_loss": 0.02856382168829441, "global_step": 261318, "epoch": 2936, "lr": 3.445614123767427e-07}
{"train_loss": 0.1072198897600174, "global_step": 261319, "epoch": 2936, "lr": 3.444934727426685e-07}
{"train_loss": 0.050235915929079056, "global_step": 261320, "epoch": 2936, "lr": 3.4442553978416556e-07}
{"train_loss": 0.05613385885953903, "global_step": 261321, "epoch": 2936, "lr": 3.443576135012505e-07}
{"train_loss": 0.07962054014205933, "global_step": 261322, "epoch": 2936, "lr": 3.4428969389392884e-07}
{"train_loss": 0.031073013320565224, "global_step": 261323, "epoch": 2936, "lr": 3.4422178096221724e-07}
{"train_loss": 0.027774931862950325, "global_step": 261324, "epoch": 2936, "lr": 3.441538747061213e-07}
{"train_loss": 0.027245087549090385, "global_step": 261325, "epoch": 2936, "lr": 3.440859751256409e-07}
{"train_loss": 0.06317600607872009, "global_step": 261326, "epoch": 2936, "lr": 3.4401808222079836e-07}
{"train_loss": 0.02666269615292549, "global_step": 261327, "epoch": 2936, "lr": 3.439501959915936e-07}
{"train_loss": 0.025365278124809265, "global_step": 261328, "epoch": 2936, "lr": 3.4388231643804337e-07}
{"train_loss": 0.030511103570461273, "global_step": 261329, "epoch": 2936, "lr": 3.4381444356015313e-07}
{"train_loss": 0.07058143615722656, "global_step": 261330, "epoch": 2936, "lr": 3.437465773579285e-07}
{"train_loss": 0.09272627532482147, "global_step": 261331, "epoch": 2936, "lr": 3.4367871783138605e-07}
{"train_loss": 0.026675749570131302, "global_step": 261332, "epoch": 2936, "lr": 3.4361086498052587e-07}
{"train_loss": 0.03443678095936775, "global_step": 261333, "epoch": 2936, "lr": 3.4354301880536454e-07}
{"train_loss": 0.020218830555677414, "global_step": 261334, "epoch": 2936, "lr": 3.434751793059077e-07}
{"train_loss": 0.036945413798093796, "global_step": 261335, "epoch": 2936, "lr": 3.4340734648216634e-07}
{"train_loss": 0.05189981311559677, "global_step": 261336, "epoch": 2936, "lr": 3.433395203341516e-07}
{"train_loss": 0.025454219430685043, "global_step": 261337, "epoch": 2936, "lr": 3.4327170086186354e-07}
{"train_loss": 0.02393185906112194, "global_step": 261338, "epoch": 2936, "lr": 3.432038880653243e-07}
{"train_loss": 0.05228161811828613, "global_step": 261339, "epoch": 2936, "lr": 3.431360819445284e-07}
{"train_loss": 0.058449722826480865, "global_step": 261340, "epoch": 2936, "lr": 3.4306828249949795e-07}
{"train_loss": 0.03822718560695648, "global_step": 261341, "epoch": 2936, "lr": 3.4300048973023855e-07}
{"train_loss": 0.034185681492090225, "global_step": 261342, "epoch": 2936, "lr": 3.429327036367502e-07}
{"train_loss": 0.03463355451822281, "global_step": 261343, "epoch": 2936, "lr": 3.428649242190551e-07}
{"train_loss": 0.05172356218099594, "global_step": 261344, "epoch": 2936, "lr": 3.427971514771533e-07}
{"train_loss": 0.04556075111031532, "global_step": 261345, "epoch": 2936, "lr": 3.427293854110558e-07}
{"train_loss": 0.030195040628314018, "global_step": 261346, "epoch": 2936, "lr": 3.4266162602077935e-07}
{"train_loss": 0.029246294870972633, "global_step": 261347, "epoch": 2936, "lr": 3.4259387330632387e-07}
{"train_loss": 0.01928653195500374, "global_step": 261348, "epoch": 2936, "lr": 3.425261272677005e-07}
{"train_loss": 0.05211000144481659, "global_step": 261349, "epoch": 2936, "lr": 3.424583879049203e-07}
{"train_loss": 0.07571757584810257, "global_step": 261350, "epoch": 2936, "lr": 3.42390655217989e-07}
{"train_loss": 0.021396566182374954, "global_step": 261351, "epoch": 2936, "lr": 3.4232292920691744e-07}
{"train_loss": 0.022851070389151573, "global_step": 261352, "epoch": 2936, "lr": 3.4225520987172246e-07}
{"train_loss": 0.08497311174869537, "global_step": 261353, "epoch": 2936, "lr": 3.421874972123984e-07}
{"train_loss": 0.09613916277885437, "global_step": 261354, "epoch": 2936, "lr": 3.4211979122896754e-07}
{"train_loss": 0.02471226081252098, "global_step": 261355, "epoch": 2936, "lr": 3.4205209192142985e-07}
{"train_loss": 0.04953407496213913, "global_step": 261356, "epoch": 2936, "lr": 3.419843992897964e-07}
{"train_loss": 0.006171934772282839, "global_step": 261357, "epoch": 2936, "lr": 3.419167133340784e-07}
{"train_loss": 0.02640795148909092, "global_step": 261358, "epoch": 2936, "lr": 3.418490340542868e-07}
{"train_loss": 0.08373953402042389, "global_step": 261359, "epoch": 2936, "lr": 3.4178136145043285e-07}
{"train_loss": 0.04169689863920212, "global_step": 261360, "epoch": 2936, "lr": 3.4171369552251643e-07}
{"train_loss": 0.05495026707649231, "global_step": 261361, "epoch": 2936, "lr": 3.416460362705487e-07}
{"train_loss": 0.04491330310702324, "global_step": 261362, "epoch": 2936, "lr": 3.415783836945463e-07}
{"train_loss": 0.02208838425576687, "global_step": 261363, "epoch": 2936, "lr": 3.4151073779450925e-07}
{"train_loss": 0.10917528718709946, "global_step": 261364, "epoch": 2936, "lr": 3.414430985704542e-07}
{"train_loss": 0.02136903442442417, "global_step": 261365, "epoch": 2936, "lr": 3.413754660223867e-07}
{"train_loss": 0.02883312478661537, "global_step": 261366, "epoch": 2936, "lr": 3.4130784015031227e-07}
{"train_loss": 0.03876235708594322, "global_step": 261367, "epoch": 2936, "lr": 3.412402209542476e-07}
{"train_loss": 0.03937462344765663, "global_step": 261368, "epoch": 2936, "lr": 3.4117260843419265e-07}
{"train_loss": 0.07501641660928726, "global_step": 261369, "epoch": 2936, "lr": 3.4110500259016967e-07}
{"train_loss": 0.033380430191755295, "global_step": 261370, "epoch": 2936, "lr": 3.4103740342217307e-07}
{"train_loss": 0.03221144527196884, "global_step": 261371, "epoch": 2936, "lr": 3.4096981093021954e-07}
{"train_loss": 0.029583975672721863, "global_step": 261372, "epoch": 2936, "lr": 3.4090222511432013e-07}
{"train_loss": 0.03729373216629028, "global_step": 261373, "epoch": 2936, "lr": 3.4083464597448047e-07}
{"train_loss": 0.07511237263679504, "global_step": 261374, "epoch": 2936, "lr": 3.4076707351071155e-07}
{"train_loss": 0.08044160902500153, "global_step": 261375, "epoch": 2936, "lr": 3.4069950772301905e-07}
{"train_loss": 0.03154724836349487, "global_step": 261376, "epoch": 2936, "lr": 3.4063194861141403e-07}
{"train_loss": 0.048428598791360855, "global_step": 261377, "epoch": 2936, "lr": 3.405643961759075e-07}
{"train_loss": 0.03699560835957527, "global_step": 261378, "epoch": 2936, "lr": 3.404968504165051e-07}
{"train_loss": 0.030224263668060303, "global_step": 261379, "epoch": 2936, "lr": 3.4042931133321797e-07}
{"train_loss": 0.03627639263868332, "global_step": 261380, "epoch": 2936, "lr": 3.4036177892605715e-07}
{"train_loss": 0.03397727385163307, "global_step": 261381, "epoch": 2936, "lr": 3.402942531950226e-07}
{"train_loss": 0.05090707540512085, "global_step": 261382, "epoch": 2936, "lr": 3.402267341401366e-07}
{"train_loss": 0.07129599153995514, "global_step": 261383, "epoch": 2936, "lr": 3.4015922176139916e-07}
{"train_loss": 0.05326715111732483, "global_step": 261384, "epoch": 2936, "lr": 3.4009171605882686e-07}
{"train_loss": 0.046958789229393005, "global_step": 261385, "epoch": 2936, "lr": 3.400242170324197e-07}
{"train_loss": 0.009655569680035114, "global_step": 261386, "epoch": 2936, "lr": 3.399567246821944e-07}
{"train_loss": 0.0555528849363327, "global_step": 261387, "epoch": 2936, "lr": 3.3988923900815095e-07}
{"train_loss": 0.024059414863586426, "global_step": 261388, "epoch": 2936, "lr": 3.3982176001031153e-07}
{"train_loss": 0.036100953817367554, "global_step": 261389, "epoch": 2936, "lr": 3.3975428768867055e-07}
{"train_loss": 0.05712888017296791, "global_step": 261390, "epoch": 2936, "lr": 3.396868220432503e-07}
{"train_loss": 0.07856286317110062, "global_step": 261391, "epoch": 2936, "lr": 3.3961936307404517e-07}
{"train_loss": 0.043830620349918525, "global_step": 261392, "epoch": 2936, "lr": 3.3955191078108295e-07, "val_loss": 9.628105163574219}
{"train_loss": 0.02794503979384899, "global_step": 261393, "epoch": 2937, "lr": 3.394844651643581e-07}
{"train_loss": 0.07394202053546906, "global_step": 261394, "epoch": 2937, "lr": 3.394170262238816e-07}
{"train_loss": 0.07181639224290848, "global_step": 261395, "epoch": 2937, "lr": 3.3934959395967026e-07}
{"train_loss": 0.041031546890735626, "global_step": 261396, "epoch": 2937, "lr": 3.3928216837172955e-07}
{"train_loss": 0.042757224291563034, "global_step": 261397, "epoch": 2937, "lr": 3.392147494600595e-07}
{"train_loss": 0.014625468291342258, "global_step": 261398, "epoch": 2937, "lr": 3.3914733722468226e-07}
{"train_loss": 0.09040869027376175, "global_step": 261399, "epoch": 2937, "lr": 3.3907993166560347e-07}
{"train_loss": 0.04768798500299454, "global_step": 261400, "epoch": 2937, "lr": 3.390125327828231e-07}
{"train_loss": 0.02060285210609436, "global_step": 261401, "epoch": 2937, "lr": 3.389451405763633e-07}
{"train_loss": 0.06037133187055588, "global_step": 261402, "epoch": 2937, "lr": 3.388777550462241e-07}
{"train_loss": 0.020497558638453484, "global_step": 261403, "epoch": 2937, "lr": 3.388103761924166e-07}
{"train_loss": 0.08473970741033554, "global_step": 261404, "epoch": 2937, "lr": 3.3874300401495195e-07}
{"train_loss": 0.053670916706323624, "global_step": 261405, "epoch": 2937, "lr": 3.3867563851383567e-07}
{"train_loss": 0.0671909973025322, "global_step": 261406, "epoch": 2937, "lr": 3.386082796890844e-07}
{"train_loss": 0.0646965280175209, "global_step": 261407, "epoch": 2937, "lr": 3.385409275406981e-07}
{"train_loss": 0.0275604035705328, "global_step": 261408, "epoch": 2937, "lr": 3.38473582068688e-07}
{"train_loss": 0.06103487312793732, "global_step": 261409, "epoch": 2937, "lr": 3.3840624327307056e-07}
{"train_loss": 0.07580772042274475, "global_step": 261410, "epoch": 2937, "lr": 3.38338911153846e-07}
{"train_loss": 0.06029820069670677, "global_step": 261411, "epoch": 2937, "lr": 3.3827158571102524e-07}
{"train_loss": 0.058182574808597565, "global_step": 261412, "epoch": 2937, "lr": 3.3820426694461947e-07}
{"train_loss": 0.07107450067996979, "global_step": 261413, "epoch": 2937, "lr": 3.3813695485463983e-07}
{"train_loss": 0.044437870383262634, "global_step": 261414, "epoch": 2937, "lr": 3.380696494410862e-07}
{"train_loss": 0.05679863691329956, "global_step": 261415, "epoch": 2937, "lr": 3.3800235070398087e-07}
{"train_loss": 0.013178020715713501, "global_step": 261416, "epoch": 2937, "lr": 3.3793505864331833e-07}
{"train_loss": 0.04400281608104706, "global_step": 261417, "epoch": 2937, "lr": 3.378677732591207e-07}
{"train_loss": 0.12476309388875961, "global_step": 261418, "epoch": 2937, "lr": 3.3780049455139353e-07}
{"train_loss": 0.02251722291111946, "global_step": 261419, "epoch": 2937, "lr": 3.377332225201368e-07}
{"train_loss": 0.03448896482586861, "global_step": 261420, "epoch": 2937, "lr": 3.3766595716537284e-07}
{"train_loss": 0.060347896069288254, "global_step": 261421, "epoch": 2937, "lr": 3.3759869848709603e-07}
{"train_loss": 0.053581953048706055, "global_step": 261422, "epoch": 2937, "lr": 3.375314464853341e-07}
{"train_loss": 0.014329247176647186, "global_step": 261423, "epoch": 2937, "lr": 3.3746420116007594e-07}
{"train_loss": 0.0257156640291214, "global_step": 261424, "epoch": 2937, "lr": 3.373969625113493e-07}
{"train_loss": 0.05997510626912117, "global_step": 261425, "epoch": 2937, "lr": 3.3732973053914874e-07}
{"train_loss": 0.03268919512629509, "global_step": 261426, "epoch": 2937, "lr": 3.372625052434908e-07}
{"train_loss": 0.04748016223311424, "global_step": 261427, "epoch": 2937, "lr": 3.3719528662438105e-07}
{"train_loss": 0.04311930760741234, "global_step": 261428, "epoch": 2937, "lr": 3.3712807468183617e-07}
{"train_loss": 0.0169940497726202, "global_step": 261429, "epoch": 2937, "lr": 3.3706086941585057e-07}
{"train_loss": 0.0316314697265625, "global_step": 261430, "epoch": 2937, "lr": 3.3699367082644653e-07}
{"train_loss": 0.037774331867694855, "global_step": 261431, "epoch": 2937, "lr": 3.369264789136295e-07}
{"train_loss": 0.01345765870064497, "global_step": 261432, "epoch": 2937, "lr": 3.368592936774051e-07}
{"train_loss": 0.05346677824854851, "global_step": 261433, "epoch": 2937, "lr": 3.367921151177844e-07}
{"train_loss": 0.024483295157551765, "global_step": 261434, "epoch": 2937, "lr": 3.367249432347785e-07}
{"train_loss": 0.03847716376185417, "global_step": 261435, "epoch": 2937, "lr": 3.3665777802839306e-07}
{"train_loss": 0.03733731061220169, "global_step": 261436, "epoch": 2937, "lr": 3.36590619498639e-07}
{"train_loss": 0.028176067396998405, "global_step": 261437, "epoch": 2937, "lr": 3.3652346764552754e-07}
{"train_loss": 0.03726045787334442, "global_step": 261438, "epoch": 2937, "lr": 3.3645632246905866e-07}
{"train_loss": 0.03365850821137428, "global_step": 261439, "epoch": 2937, "lr": 3.3638918396925457e-07}
{"train_loss": 0.08222807943820953, "global_step": 261440, "epoch": 2937, "lr": 3.363220521461097e-07}
{"train_loss": 0.03193289786577225, "global_step": 261441, "epoch": 2937, "lr": 3.3625492699965175e-07}
{"train_loss": 0.029430512338876724, "global_step": 261442, "epoch": 2937, "lr": 3.3618780852986976e-07}
{"train_loss": 0.028195315971970558, "global_step": 261443, "epoch": 2937, "lr": 3.361206967367858e-07}
{"train_loss": 0.018331637606024742, "global_step": 261444, "epoch": 2937, "lr": 3.360535916204055e-07}
{"train_loss": 0.04333445802330971, "global_step": 261445, "epoch": 2937, "lr": 3.3598649318073996e-07}
{"train_loss": 0.061947986483573914, "global_step": 261446, "epoch": 2937, "lr": 3.359194014177891e-07}
{"train_loss": 0.021366212517023087, "global_step": 261447, "epoch": 2937, "lr": 3.358523163315752e-07}
{"train_loss": 0.0586528405547142, "global_step": 261448, "epoch": 2937, "lr": 3.357852379220983e-07}
{"train_loss": 0.06058346480131149, "global_step": 261449, "epoch": 2937, "lr": 3.3571816618936936e-07}
{"train_loss": 0.06071417033672333, "global_step": 261450, "epoch": 2937, "lr": 3.35651101133394e-07}
{"train_loss": 0.051641106605529785, "global_step": 261451, "epoch": 2937, "lr": 3.3558404275418897e-07}
{"train_loss": 0.06117599084973335, "global_step": 261452, "epoch": 2937, "lr": 3.355169910517597e-07}
{"train_loss": 0.03270406648516655, "global_step": 261453, "epoch": 2937, "lr": 3.3544994602611177e-07}
{"train_loss": 0.04053211212158203, "global_step": 261454, "epoch": 2937, "lr": 3.3538290767726186e-07}
{"train_loss": 0.01177911926060915, "global_step": 261455, "epoch": 2937, "lr": 3.3531587600521e-07}
{"train_loss": 0.04816509783267975, "global_step": 261456, "epoch": 2937, "lr": 3.352488510099727e-07}
{"train_loss": 0.05670809745788574, "global_step": 261457, "epoch": 2937, "lr": 3.351818326915557e-07}
{"train_loss": 0.038395434617996216, "global_step": 261458, "epoch": 2937, "lr": 3.351148210499699e-07}
{"train_loss": 0.039996955543756485, "global_step": 261459, "epoch": 2937, "lr": 3.3504781608521553e-07}
{"train_loss": 0.007091222330927849, "global_step": 261460, "epoch": 2937, "lr": 3.349808177973146e-07}
{"train_loss": 0.05208044871687889, "global_step": 261461, "epoch": 2937, "lr": 3.349138261862672e-07}
{"train_loss": 0.0578557513654232, "global_step": 261462, "epoch": 2937, "lr": 3.3484684125208996e-07}
{"train_loss": 0.05659835785627365, "global_step": 261463, "epoch": 2937, "lr": 3.3477986299478293e-07}
{"train_loss": 0.043134428560733795, "global_step": 261464, "epoch": 2937, "lr": 3.3471289141436266e-07}
{"train_loss": 0.04585058242082596, "global_step": 261465, "epoch": 2937, "lr": 3.3464592651082925e-07}
{"train_loss": 0.05981490761041641, "global_step": 261466, "epoch": 2937, "lr": 3.345789682842049e-07}
{"train_loss": 0.048576027154922485, "global_step": 261467, "epoch": 2937, "lr": 3.34512016734484e-07}
{"train_loss": 0.027248553931713104, "global_step": 261468, "epoch": 2937, "lr": 3.3444507186168873e-07}
{"train_loss": 0.04501922056078911, "global_step": 261469, "epoch": 2937, "lr": 3.343781336658136e-07}
{"train_loss": 0.05025529861450195, "global_step": 261470, "epoch": 2937, "lr": 3.3431120214688637e-07}
{"train_loss": 0.06665696948766708, "global_step": 261471, "epoch": 2937, "lr": 3.342442773048959e-07}
{"train_loss": 0.042728789150714874, "global_step": 261472, "epoch": 2937, "lr": 3.3417735913987e-07}
{"train_loss": 0.04223603010177612, "global_step": 261473, "epoch": 2937, "lr": 3.3411044765180313e-07}
{"train_loss": 0.034896403551101685, "global_step": 261474, "epoch": 2937, "lr": 3.3404354284071183e-07}
{"train_loss": 0.03226510062813759, "global_step": 261475, "epoch": 2937, "lr": 3.3397664470660175e-07}
{"train_loss": 0.07741107791662216, "global_step": 261476, "epoch": 2937, "lr": 3.3390975324948393e-07}
{"train_loss": 0.04522869363427162, "global_step": 261477, "epoch": 2937, "lr": 3.338428684693695e-07}
{"train_loss": 0.06395208090543747, "global_step": 261478, "epoch": 2937, "lr": 3.337759903662585e-07}
{"train_loss": 0.029602618888020515, "global_step": 261479, "epoch": 2937, "lr": 3.337091189401675e-07}
{"train_loss": 0.07190687209367752, "global_step": 261480, "epoch": 2937, "lr": 3.336422541911077e-07}
{"train_loss": 0.045758933802082964, "global_step": 261481, "epoch": 2937, "lr": 3.3357539611907904e-07, "val_loss": 9.65067195892334}
{"train_loss": 0.027201704680919647, "global_step": 261482, "epoch": 2938, "lr": 3.335085447240982e-07}
{"train_loss": 0.05281298980116844, "global_step": 261483, "epoch": 2938, "lr": 3.3344170000617627e-07}
{"train_loss": 0.049631230533123016, "global_step": 261484, "epoch": 2938, "lr": 3.333748619653132e-07}
{"train_loss": 0.028029009699821472, "global_step": 261485, "epoch": 2938, "lr": 3.3330803060152014e-07}
{"train_loss": 0.0732090026140213, "global_step": 261486, "epoch": 2938, "lr": 3.3324120591481376e-07}
{"train_loss": 0.06264180690050125, "global_step": 261487, "epoch": 2938, "lr": 3.331743879051941e-07}
{"train_loss": 0.03352266550064087, "global_step": 261488, "epoch": 2938, "lr": 3.331075765726721e-07}
{"train_loss": 0.0385119654238224, "global_step": 261489, "epoch": 2938, "lr": 3.330407719172646e-07}
{"train_loss": 0.08687597513198853, "global_step": 261490, "epoch": 2938, "lr": 3.3297397393897145e-07}
{"train_loss": 0.04131809249520302, "global_step": 261491, "epoch": 2938, "lr": 3.329071826378039e-07}
{"train_loss": 0.07095441967248917, "global_step": 261492, "epoch": 2938, "lr": 3.328403980137729e-07}
{"train_loss": 0.03714875504374504, "global_step": 261493, "epoch": 2938, "lr": 3.327736200668841e-07}
{"train_loss": 0.027664706110954285, "global_step": 261494, "epoch": 2938, "lr": 3.3270684879714855e-07}
{"train_loss": 0.03805823624134064, "global_step": 261495, "epoch": 2938, "lr": 3.326400842045774e-07}
{"train_loss": 0.034569330513477325, "global_step": 261496, "epoch": 2938, "lr": 3.3257332628917616e-07}
{"train_loss": 0.04267274588346481, "global_step": 261497, "epoch": 2938, "lr": 3.3250657505095593e-07}
{"train_loss": 0.08978334814310074, "global_step": 261498, "epoch": 2938, "lr": 3.324398304899223e-07}
{"train_loss": 0.025245383381843567, "global_step": 261499, "epoch": 2938, "lr": 3.3237309260609193e-07}
{"train_loss": 0.05618665739893913, "global_step": 261500, "epoch": 2938, "lr": 3.323063613994648e-07}
{"train_loss": 0.037206511944532394, "global_step": 261501, "epoch": 2938, "lr": 3.32239636870052e-07}
{"train_loss": 0.05189639702439308, "global_step": 261502, "epoch": 2938, "lr": 3.3217291901787016e-07}
{"train_loss": 0.028373509645462036, "global_step": 261503, "epoch": 2938, "lr": 3.321062078429138e-07}
{"train_loss": 0.052236754447221756, "global_step": 261504, "epoch": 2938, "lr": 3.320395033452106e-07}
{"train_loss": 0.022379107773303986, "global_step": 261505, "epoch": 2938, "lr": 3.319728055247495e-07}
{"train_loss": 0.04174867644906044, "global_step": 261506, "epoch": 2938, "lr": 3.3190611438155827e-07}
{"train_loss": 0.03819280117750168, "global_step": 261507, "epoch": 2938, "lr": 3.318394299156313e-07}
{"train_loss": 0.04225413501262665, "global_step": 261508, "epoch": 2938, "lr": 3.317727521269853e-07}
{"train_loss": 0.057469144463539124, "global_step": 261509, "epoch": 2938, "lr": 3.3170608101562026e-07}
{"train_loss": 0.03275479003787041, "global_step": 261510, "epoch": 2938, "lr": 3.3163941658156395e-07}
{"train_loss": 0.03590861335396767, "global_step": 261511, "epoch": 2938, "lr": 3.315727588248052e-07}
{"train_loss": 0.06244613230228424, "global_step": 261512, "epoch": 2938, "lr": 3.315061077453607e-07}
{"train_loss": 0.05452447757124901, "global_step": 261513, "epoch": 2938, "lr": 3.3143946334324715e-07}
{"train_loss": 0.02132391929626465, "global_step": 261514, "epoch": 2938, "lr": 3.3137282561845895e-07}
{"train_loss": 0.015928450971841812, "global_step": 261515, "epoch": 2938, "lr": 3.3130619457101273e-07}
{"train_loss": 0.03704476356506348, "global_step": 261516, "epoch": 2938, "lr": 3.312395702009197e-07}
{"train_loss": 0.0626753568649292, "global_step": 261517, "epoch": 2938, "lr": 3.3117295250818523e-07}
{"train_loss": 0.02251710742712021, "global_step": 261518, "epoch": 2938, "lr": 3.311063414928206e-07}
{"train_loss": 0.07789096981287003, "global_step": 261519, "epoch": 2938, "lr": 3.310397371548313e-07}
{"train_loss": 0.029321154579520226, "global_step": 261520, "epoch": 2938, "lr": 3.309731394942284e-07}
{"train_loss": 0.047393057495355606, "global_step": 261521, "epoch": 2938, "lr": 3.30906548511023e-07}
{"train_loss": 0.022452453151345253, "global_step": 261522, "epoch": 2938, "lr": 3.3083996420522067e-07}
{"train_loss": 0.03200286254286766, "global_step": 261523, "epoch": 2938, "lr": 3.3077338657683255e-07}
{"train_loss": 0.04245448857545853, "global_step": 261524, "epoch": 2938, "lr": 3.3070681562586414e-07}
{"train_loss": 0.019068913534283638, "global_step": 261525, "epoch": 2938, "lr": 3.3064025135232654e-07}
{"train_loss": 0.03773610293865204, "global_step": 261526, "epoch": 2938, "lr": 3.3057369375623094e-07}
{"train_loss": 0.013138169422745705, "global_step": 261527, "epoch": 2938, "lr": 3.305071428375828e-07}
{"train_loss": 0.04431341588497162, "global_step": 261528, "epoch": 2938, "lr": 3.304405985963932e-07}
{"train_loss": 0.05780476704239845, "global_step": 261529, "epoch": 2938, "lr": 3.3037406103266776e-07}
{"train_loss": 0.014661149121820927, "global_step": 261530, "epoch": 2938, "lr": 3.3030753014642313e-07}
{"train_loss": 0.05193888396024704, "global_step": 261531, "epoch": 2938, "lr": 3.302410059376593e-07}
{"train_loss": 0.0879472866654396, "global_step": 261532, "epoch": 2938, "lr": 3.301744884063929e-07}
{"train_loss": 0.028503360226750374, "global_step": 261533, "epoch": 2938, "lr": 3.3010797755262944e-07}
{"train_loss": 0.039871055632829666, "global_step": 261534, "epoch": 2938, "lr": 3.300414733763746e-07}
{"train_loss": 0.04464145004749298, "global_step": 261535, "epoch": 2938, "lr": 3.2997497587763935e-07}
{"train_loss": 0.029376093298196793, "global_step": 261536, "epoch": 2938, "lr": 3.299084850564349e-07}
{"train_loss": 0.02552040107548237, "global_step": 261537, "epoch": 2938, "lr": 3.298420009127723e-07}
{"train_loss": 0.020333852618932724, "global_step": 261538, "epoch": 2938, "lr": 3.2977552344665153e-07}
{"train_loss": 0.04483495652675629, "global_step": 261539, "epoch": 2938, "lr": 3.297090526580893e-07}
{"train_loss": 0.06533151865005493, "global_step": 261540, "epoch": 2938, "lr": 3.296425885470966e-07}
{"train_loss": 0.05118207633495331, "global_step": 261541, "epoch": 2938, "lr": 3.29576131113668e-07}
{"train_loss": 0.08314922451972961, "global_step": 261542, "epoch": 2938, "lr": 3.2950968035783123e-07}
{"train_loss": 0.04291384667158127, "global_step": 261543, "epoch": 2938, "lr": 3.2944323627958075e-07}
{"train_loss": 0.06136645749211311, "global_step": 261544, "epoch": 2938, "lr": 3.2937679887893314e-07}
{"train_loss": 0.0287982989102602, "global_step": 261545, "epoch": 2938, "lr": 3.293103681558996e-07}
{"train_loss": 0.0928066223859787, "global_step": 261546, "epoch": 2938, "lr": 3.292439441104855e-07}
{"train_loss": 0.039848990738391876, "global_step": 261547, "epoch": 2938, "lr": 3.291775267426911e-07}
{"train_loss": 0.08202685415744781, "global_step": 261548, "epoch": 2938, "lr": 3.291111160525384e-07}
{"train_loss": 0.06965336948633194, "global_step": 261549, "epoch": 2938, "lr": 3.290447120400331e-07}
{"train_loss": 0.034134265035390854, "global_step": 261550, "epoch": 2938, "lr": 3.289783147051806e-07}
{"train_loss": 0.06874299049377441, "global_step": 261551, "epoch": 2938, "lr": 3.289119240479921e-07}
{"train_loss": 0.08862383663654327, "global_step": 261552, "epoch": 2938, "lr": 3.2884554006847314e-07}
{"train_loss": 0.04080558568239212, "global_step": 261553, "epoch": 2938, "lr": 3.2877916276664035e-07}
{"train_loss": 0.05264488235116005, "global_step": 261554, "epoch": 2938, "lr": 3.287127921424882e-07}
{"train_loss": 0.062006622552871704, "global_step": 261555, "epoch": 2938, "lr": 3.2864642819604443e-07}
{"train_loss": 0.031588830053806305, "global_step": 261556, "epoch": 2938, "lr": 3.285800709273035e-07}
{"train_loss": 0.03676459193229675, "global_step": 261557, "epoch": 2938, "lr": 3.2851372033628755e-07}
{"train_loss": 0.0742129236459732, "global_step": 261558, "epoch": 2938, "lr": 3.284473764229856e-07}
{"train_loss": 0.028806736692786217, "global_step": 261559, "epoch": 2938, "lr": 3.283810391874309e-07}
{"train_loss": 0.05140393227338791, "global_step": 261560, "epoch": 2938, "lr": 3.2831470862961233e-07}
{"train_loss": 0.026269955560564995, "global_step": 261561, "epoch": 2938, "lr": 3.2824838474954655e-07}
{"train_loss": 0.01601850427687168, "global_step": 261562, "epoch": 2938, "lr": 3.281820675472447e-07}
{"train_loss": 0.08183389157056808, "global_step": 261563, "epoch": 2938, "lr": 3.2811575702271225e-07}
{"train_loss": 0.04321923479437828, "global_step": 261564, "epoch": 2938, "lr": 3.2804945317596035e-07}
{"train_loss": 0.07308278232812881, "global_step": 261565, "epoch": 2938, "lr": 3.279831560069946e-07}
{"train_loss": 0.05549382045865059, "global_step": 261566, "epoch": 2938, "lr": 3.27916865515826e-07}
{"train_loss": 0.03365734592080116, "global_step": 261567, "epoch": 2938, "lr": 3.2785058170246576e-07}
{"train_loss": 0.052115414291620255, "global_step": 261568, "epoch": 2938, "lr": 3.2778430456691935e-07}
{"train_loss": 0.022141113877296448, "global_step": 261569, "epoch": 2938, "lr": 3.277180341091979e-07}
{"train_loss": 0.0461367020550906, "global_step": 261570, "epoch": 2938, "lr": 3.2765177032930694e-07, "val_loss": 9.598291397094727}
{"train_loss": 0.04915057495236397, "global_step": 261571, "epoch": 2939, "lr": 3.2758551322725763e-07}
{"train_loss": 0.08743583410978317, "global_step": 261572, "epoch": 2939, "lr": 3.27519262803061e-07}
{"train_loss": 0.029796577990055084, "global_step": 261573, "epoch": 2939, "lr": 3.2745301905672265e-07}
{"train_loss": 0.08208777010440826, "global_step": 261574, "epoch": 2939, "lr": 3.273867819882592e-07}
{"train_loss": 0.05007617175579071, "global_step": 261575, "epoch": 2939, "lr": 3.2732055159766516e-07}
{"train_loss": 0.05311420187354088, "global_step": 261576, "epoch": 2939, "lr": 3.272543278849627e-07}
{"train_loss": 0.062103837728500366, "global_step": 261577, "epoch": 2939, "lr": 3.2718811085015176e-07}
{"train_loss": 0.01649968884885311, "global_step": 261578, "epoch": 2939, "lr": 3.271219004932491e-07}
{"train_loss": 0.06945962458848953, "global_step": 261579, "epoch": 2939, "lr": 3.2705569681425465e-07}
{"train_loss": 0.03695768862962723, "global_step": 261580, "epoch": 2939, "lr": 3.269894998131795e-07}
{"train_loss": 0.03469296544790268, "global_step": 261581, "epoch": 2939, "lr": 3.269233094900459e-07}
{"train_loss": 0.06875601410865784, "global_step": 261582, "epoch": 2939, "lr": 3.2685712584484274e-07}
{"train_loss": 0.03937666118144989, "global_step": 261583, "epoch": 2939, "lr": 3.2679094887759223e-07}
{"train_loss": 0.09767080843448639, "global_step": 261584, "epoch": 2939, "lr": 3.267247785882999e-07}
{"train_loss": 0.03954653441905975, "global_step": 261585, "epoch": 2939, "lr": 3.2665861497697123e-07}
{"train_loss": 0.045181095600128174, "global_step": 261586, "epoch": 2939, "lr": 3.2659245804361747e-07}
{"train_loss": 0.01953207701444626, "global_step": 261587, "epoch": 2939, "lr": 3.2652630778824965e-07}
{"train_loss": 0.05042112246155739, "global_step": 261588, "epoch": 2939, "lr": 3.264601642108789e-07}
{"train_loss": 0.04554055631160736, "global_step": 261589, "epoch": 2939, "lr": 3.2639402731150515e-07}
{"train_loss": 0.02698388695716858, "global_step": 261590, "epoch": 2939, "lr": 3.263278970901451e-07}
{"train_loss": 0.10641241073608398, "global_step": 261591, "epoch": 2939, "lr": 3.2626177354680433e-07}
{"train_loss": 0.05138808488845825, "global_step": 261592, "epoch": 2939, "lr": 3.261956566814883e-07}
{"train_loss": 0.07446161657571793, "global_step": 261593, "epoch": 2939, "lr": 3.261295464942138e-07}
{"train_loss": 0.037320513278245926, "global_step": 261594, "epoch": 2939, "lr": 3.260634429849807e-07}
{"train_loss": 0.044596150517463684, "global_step": 261595, "epoch": 2939, "lr": 3.2599734615381127e-07}
{"train_loss": 0.015176302753388882, "global_step": 261596, "epoch": 2939, "lr": 3.259312560007e-07}
{"train_loss": 0.06546091288328171, "global_step": 261597, "epoch": 2939, "lr": 3.258651725256634e-07}
{"train_loss": 0.04195161908864975, "global_step": 261598, "epoch": 2939, "lr": 3.257990957287072e-07}
{"train_loss": 0.06603718549013138, "global_step": 261599, "epoch": 2939, "lr": 3.257330256098423e-07}
{"train_loss": 0.02645215392112732, "global_step": 261600, "epoch": 2939, "lr": 3.2566696216908e-07}
{"train_loss": 0.012384464964270592, "global_step": 261601, "epoch": 2939, "lr": 3.256009054064202e-07}
{"train_loss": 0.044605616480112076, "global_step": 261602, "epoch": 2939, "lr": 3.255348553218851e-07}
{"train_loss": 0.0439118817448616, "global_step": 261603, "epoch": 2939, "lr": 3.2546881191547473e-07}
{"train_loss": 0.023228202015161514, "global_step": 261604, "epoch": 2939, "lr": 3.254027751872002e-07}
{"train_loss": 0.03174104169011116, "global_step": 261605, "epoch": 2939, "lr": 3.253367451370615e-07}
{"train_loss": 0.04839015379548073, "global_step": 261606, "epoch": 2939, "lr": 3.2527072176508634e-07}
{"train_loss": 0.0349726527929306, "global_step": 261607, "epoch": 2939, "lr": 3.252047050712692e-07}
{"train_loss": 0.01774238422513008, "global_step": 261608, "epoch": 2939, "lr": 3.251386950556212e-07}
{"train_loss": 0.05481445789337158, "global_step": 261609, "epoch": 2939, "lr": 3.2507269171815345e-07}
{"train_loss": 0.02738971635699272, "global_step": 261610, "epoch": 2939, "lr": 3.2500669505887705e-07}
{"train_loss": 0.0378047339618206, "global_step": 261611, "epoch": 2939, "lr": 3.249407050777975e-07}
{"train_loss": 0.057318661361932755, "global_step": 261612, "epoch": 2939, "lr": 3.248747217749204e-07}
{"train_loss": 0.021068738773465157, "global_step": 261613, "epoch": 2939, "lr": 3.248087451502624e-07}
{"train_loss": 0.028719503432512283, "global_step": 261614, "epoch": 2939, "lr": 3.24742775203829e-07}
{"train_loss": 0.08407445251941681, "global_step": 261615, "epoch": 2939, "lr": 3.2467681193562025e-07}
{"train_loss": 0.0355306938290596, "global_step": 261616, "epoch": 2939, "lr": 3.246108553456584e-07}
{"train_loss": 0.031102530658245087, "global_step": 261617, "epoch": 2939, "lr": 3.2454490543395443e-07}
{"train_loss": 0.04738357290625572, "global_step": 261618, "epoch": 2939, "lr": 3.2447896220049736e-07}
{"train_loss": 0.026546748355031013, "global_step": 261619, "epoch": 2939, "lr": 3.244130256453204e-07}
{"train_loss": 0.019143393263220787, "global_step": 261620, "epoch": 2939, "lr": 3.243470957684125e-07}
{"train_loss": 0.026846855878829956, "global_step": 261621, "epoch": 2939, "lr": 3.2428117256979584e-07}
{"train_loss": 0.057662513107061386, "global_step": 261622, "epoch": 2939, "lr": 3.2421525604947045e-07}
{"train_loss": 0.03351901099085808, "global_step": 261623, "epoch": 2939, "lr": 3.2414934620744743e-07}
{"train_loss": 0.05912943184375763, "global_step": 261624, "epoch": 2939, "lr": 3.240834430437434e-07}
{"train_loss": 0.05385798215866089, "global_step": 261625, "epoch": 2939, "lr": 3.240175465583528e-07}
{"train_loss": 0.023984311148524284, "global_step": 261626, "epoch": 2939, "lr": 3.2395165675130346e-07}
{"train_loss": 0.059941571205854416, "global_step": 261627, "epoch": 2939, "lr": 3.238857736225842e-07}
{"train_loss": 0.02049839310348034, "global_step": 261628, "epoch": 2939, "lr": 3.238198971722173e-07}
{"train_loss": 0.054429445415735245, "global_step": 261629, "epoch": 2939, "lr": 3.2375402740020265e-07}
{"train_loss": 0.03528042137622833, "global_step": 261630, "epoch": 2939, "lr": 3.23688164306557e-07}
{"train_loss": 0.02175203152000904, "global_step": 261631, "epoch": 2939, "lr": 3.236223078912859e-07}
{"train_loss": 0.03526388108730316, "global_step": 261632, "epoch": 2939, "lr": 3.235564581544004e-07}
{"train_loss": 0.041871339082717896, "global_step": 261633, "epoch": 2939, "lr": 3.2349061509590605e-07}
{"train_loss": 0.045350681990385056, "global_step": 261634, "epoch": 2939, "lr": 3.234247787158085e-07}
{"train_loss": 0.04497483745217323, "global_step": 261635, "epoch": 2939, "lr": 3.233589490141242e-07}
{"train_loss": 0.06278006732463837, "global_step": 261636, "epoch": 2939, "lr": 3.232931259908589e-07}
{"train_loss": 0.03319353982806206, "global_step": 261637, "epoch": 2939, "lr": 3.2322730964602366e-07}
{"train_loss": 0.04242498427629471, "global_step": 261638, "epoch": 2939, "lr": 3.2316149997962397e-07}
{"train_loss": 0.04372531548142433, "global_step": 261639, "epoch": 2939, "lr": 3.2309569699166544e-07}
{"train_loss": 0.046533968299627304, "global_step": 261640, "epoch": 2939, "lr": 3.2302990068216466e-07}
{"train_loss": 0.09445127099752426, "global_step": 261641, "epoch": 2939, "lr": 3.2296411105112724e-07}
{"train_loss": 0.02860993891954422, "global_step": 261642, "epoch": 2939, "lr": 3.2289832809856423e-07}
{"train_loss": 0.032531991600990295, "global_step": 261643, "epoch": 2939, "lr": 3.228325518244757e-07}
{"train_loss": 0.056004829704761505, "global_step": 261644, "epoch": 2939, "lr": 3.227667822288838e-07}
{"train_loss": 0.04078212380409241, "global_step": 261645, "epoch": 2939, "lr": 3.2270101931178297e-07}
{"train_loss": 0.05814947187900543, "global_step": 261646, "epoch": 2939, "lr": 3.2263526307319546e-07}
{"train_loss": 0.03593263030052185, "global_step": 261647, "epoch": 2939, "lr": 3.2256951351312127e-07}
{"train_loss": 0.02406347543001175, "global_step": 261648, "epoch": 2939, "lr": 3.22503770631577e-07}
{"train_loss": 0.05540598928928375, "global_step": 261649, "epoch": 2939, "lr": 3.224380344285627e-07}
{"train_loss": 0.02923424541950226, "global_step": 261650, "epoch": 2939, "lr": 3.2237230490408944e-07}
{"train_loss": 0.036922719329595566, "global_step": 261651, "epoch": 2939, "lr": 3.223065820581683e-07}
{"train_loss": 0.04628756642341614, "global_step": 261652, "epoch": 2939, "lr": 3.222408658908105e-07}
{"train_loss": 0.04498199373483658, "global_step": 261653, "epoch": 2939, "lr": 3.2217515640202147e-07}
{"train_loss": 0.04040863364934921, "global_step": 261654, "epoch": 2939, "lr": 3.2210945359180677e-07}
{"train_loss": 0.029521731659770012, "global_step": 261655, "epoch": 2939, "lr": 3.220437574601831e-07}
{"train_loss": 0.04137152060866356, "global_step": 261656, "epoch": 2939, "lr": 3.2197806800715045e-07}
{"train_loss": 0.046627894043922424, "global_step": 261657, "epoch": 2939, "lr": 3.219123852327255e-07}
{"train_loss": 0.0556056834757328, "global_step": 261658, "epoch": 2939, "lr": 3.2184670913691373e-07}
{"train_loss": 0.04426269966774107, "global_step": 261659, "epoch": 2939, "lr": 3.217810397197263e-07, "val_loss": 9.685725212097168}
{"train_loss": 0.028284717351198196, "global_step": 261660, "epoch": 2940, "lr": 3.2171537698116316e-07}
{"train_loss": 0.09654515981674194, "global_step": 261661, "epoch": 2940, "lr": 3.2164972092124655e-07}
{"train_loss": 0.08490343391895294, "global_step": 261662, "epoch": 2940, "lr": 3.2158407153998207e-07}
{"train_loss": 0.017160819843411446, "global_step": 261663, "epoch": 2940, "lr": 3.2151842883736406e-07}
{"train_loss": 0.017078937962651253, "global_step": 261664, "epoch": 2940, "lr": 3.214527928134203e-07}
{"train_loss": 0.027526620775461197, "global_step": 261665, "epoch": 2940, "lr": 3.213871634681509e-07}
{"train_loss": 0.057383883744478226, "global_step": 261666, "epoch": 2940, "lr": 3.213215408015613e-07}
{"train_loss": 0.024155864492058754, "global_step": 261667, "epoch": 2940, "lr": 3.212559248136737e-07}
{"train_loss": 0.03013349138200283, "global_step": 261668, "epoch": 2940, "lr": 3.211903155044771e-07}
{"train_loss": 0.0187259279191494, "global_step": 261669, "epoch": 2940, "lr": 3.2112471287399914e-07}
{"train_loss": 0.07344526797533035, "global_step": 261670, "epoch": 2940, "lr": 3.2105911692223435e-07}
{"train_loss": 0.05449439585208893, "global_step": 261671, "epoch": 2940, "lr": 3.209935276491993e-07}
{"train_loss": 0.05377570167183876, "global_step": 261672, "epoch": 2940, "lr": 3.2092794505489966e-07}
{"train_loss": 0.04357636347413063, "global_step": 261673, "epoch": 2940, "lr": 3.2086236913934644e-07}
{"train_loss": 0.04950537532567978, "global_step": 261674, "epoch": 2940, "lr": 3.207967999025507e-07}
{"train_loss": 0.04524276405572891, "global_step": 261675, "epoch": 2940, "lr": 3.2073123734451813e-07}
{"train_loss": 0.06702589243650436, "global_step": 261676, "epoch": 2940, "lr": 3.206656814652542e-07}
{"train_loss": 0.04748658463358879, "global_step": 261677, "epoch": 2940, "lr": 3.2060013226477557e-07}
{"train_loss": 0.08392541110515594, "global_step": 261678, "epoch": 2940, "lr": 3.2053458974308225e-07}
{"train_loss": 0.04622674733400345, "global_step": 261679, "epoch": 2940, "lr": 3.204690539001909e-07}
{"train_loss": 0.07098297774791718, "global_step": 261680, "epoch": 2940, "lr": 3.20403524736107e-07}
{"train_loss": 0.04680865630507469, "global_step": 261681, "epoch": 2940, "lr": 3.2033800225083623e-07}
{"train_loss": 0.035900235176086426, "global_step": 261682, "epoch": 2940, "lr": 3.202724864443951e-07}
{"train_loss": 0.05096067860722542, "global_step": 261683, "epoch": 2940, "lr": 3.2020697731678373e-07}
{"train_loss": 0.029596198350191116, "global_step": 261684, "epoch": 2940, "lr": 3.201414748680187e-07}
{"train_loss": 0.016734231263399124, "global_step": 261685, "epoch": 2940, "lr": 3.2007597909810004e-07}
{"train_loss": 0.07240381836891174, "global_step": 261686, "epoch": 2940, "lr": 3.200104900070444e-07}
{"train_loss": 0.032795362174510956, "global_step": 261687, "epoch": 2940, "lr": 3.1994500759485736e-07}
{"train_loss": 0.020692378282546997, "global_step": 261688, "epoch": 2940, "lr": 3.1987953186154995e-07}
{"train_loss": 0.030218979343771935, "global_step": 261689, "epoch": 2940, "lr": 3.1981406280712776e-07}
{"train_loss": 0.06678963452577591, "global_step": 261690, "epoch": 2940, "lr": 3.197486004316019e-07}
{"train_loss": 0.013594371266663074, "global_step": 261691, "epoch": 2940, "lr": 3.196831447349835e-07}
{"train_loss": 0.033071476966142654, "global_step": 261692, "epoch": 2940, "lr": 3.1961769571727253e-07}
{"train_loss": 0.051843296736478806, "global_step": 261693, "epoch": 2940, "lr": 3.195522533784856e-07}
{"train_loss": 0.0334911122918129, "global_step": 261694, "epoch": 2940, "lr": 3.194868177186283e-07}
{"train_loss": 0.04867076873779297, "global_step": 261695, "epoch": 2940, "lr": 3.194213887377118e-07}
{"train_loss": 0.02720167301595211, "global_step": 261696, "epoch": 2940, "lr": 3.193559664357415e-07}
{"train_loss": 0.03464678302407265, "global_step": 261697, "epoch": 2940, "lr": 3.192905508127342e-07}
{"train_loss": 0.032164763659238815, "global_step": 261698, "epoch": 2940, "lr": 3.1922514186868426e-07}
{"train_loss": 0.0450252965092659, "global_step": 261699, "epoch": 2940, "lr": 3.191597396036139e-07}
{"train_loss": 0.03616570681333542, "global_step": 261700, "epoch": 2940, "lr": 3.190943440175287e-07}
{"train_loss": 0.037405285984277725, "global_step": 261701, "epoch": 2940, "lr": 3.190289551104342e-07}
{"train_loss": 0.05297993868589401, "global_step": 261702, "epoch": 2940, "lr": 3.1896357288234144e-07}
{"train_loss": 0.016793537884950638, "global_step": 261703, "epoch": 2940, "lr": 3.188981973332561e-07}
{"train_loss": 0.08820444345474243, "global_step": 261704, "epoch": 2940, "lr": 3.1883282846318916e-07}
{"train_loss": 0.0538766123354435, "global_step": 261705, "epoch": 2940, "lr": 3.187674662721518e-07}
{"train_loss": 0.0820363387465477, "global_step": 261706, "epoch": 2940, "lr": 3.1870211076014954e-07}
{"train_loss": 0.06388795375823975, "global_step": 261707, "epoch": 2940, "lr": 3.1863676192718794e-07}
{"train_loss": 0.02603684738278389, "global_step": 261708, "epoch": 2940, "lr": 3.185714197732892e-07}
{"train_loss": 0.055409640073776245, "global_step": 261709, "epoch": 2940, "lr": 3.1850608429844776e-07}
{"train_loss": 0.10819407552480698, "global_step": 261710, "epoch": 2940, "lr": 3.184407555026747e-07}
{"train_loss": 0.0329839326441288, "global_step": 261711, "epoch": 2940, "lr": 3.1837543338598675e-07}
{"train_loss": 0.047848232090473175, "global_step": 261712, "epoch": 2940, "lr": 3.183101179483894e-07}
{"train_loss": 0.019591324031352997, "global_step": 261713, "epoch": 2940, "lr": 3.1824480918988266e-07}
{"train_loss": 0.03251613304018974, "global_step": 261714, "epoch": 2940, "lr": 3.181795071104887e-07}
{"train_loss": 0.0550377182662487, "global_step": 261715, "epoch": 2940, "lr": 3.1811421171020204e-07}
{"train_loss": 0.029197843745350838, "global_step": 261716, "epoch": 2940, "lr": 3.1804892298905043e-07}
{"train_loss": 0.012818064540624619, "global_step": 261717, "epoch": 2940, "lr": 3.179836409470227e-07}
{"train_loss": 0.03224074840545654, "global_step": 261718, "epoch": 2940, "lr": 3.179183655841411e-07}
{"train_loss": 0.06539282947778702, "global_step": 261719, "epoch": 2940, "lr": 3.1785309690040567e-07}
{"train_loss": 0.047880250960588455, "global_step": 261720, "epoch": 2940, "lr": 3.1778783489583297e-07}
{"train_loss": 0.030799703672528267, "global_step": 261721, "epoch": 2940, "lr": 3.177225795704286e-07}
{"train_loss": 0.03803901746869087, "global_step": 261722, "epoch": 2940, "lr": 3.1765733092419813e-07}
{"train_loss": 0.019463833421468735, "global_step": 261723, "epoch": 2940, "lr": 3.1759208895715266e-07}
{"train_loss": 0.020766407251358032, "global_step": 261724, "epoch": 2940, "lr": 3.175268536693088e-07}
{"train_loss": 0.040757764130830765, "global_step": 261725, "epoch": 2940, "lr": 3.17461625060661e-07}
{"train_loss": 0.031071782112121582, "global_step": 261726, "epoch": 2940, "lr": 3.17396403131226e-07}
{"train_loss": 0.034664735198020935, "global_step": 261727, "epoch": 2940, "lr": 3.1733118788100925e-07}
{"train_loss": 0.036645784974098206, "global_step": 261728, "epoch": 2940, "lr": 3.1726597931002746e-07}
{"train_loss": 0.05224956199526787, "global_step": 261729, "epoch": 2940, "lr": 3.1720077741828056e-07}
{"train_loss": 0.03687366470694542, "global_step": 261730, "epoch": 2940, "lr": 3.171355822057798e-07}
{"train_loss": 0.01985042728483677, "global_step": 261731, "epoch": 2940, "lr": 3.1707039367253613e-07}
{"train_loss": 0.032737378031015396, "global_step": 261732, "epoch": 2940, "lr": 3.170052118185551e-07}
{"train_loss": 0.07066153734922409, "global_step": 261733, "epoch": 2940, "lr": 3.1694003664384796e-07}
{"train_loss": 0.03998131304979324, "global_step": 261734, "epoch": 2940, "lr": 3.1687486814842016e-07}
{"train_loss": 0.01139448769390583, "global_step": 261735, "epoch": 2940, "lr": 3.1680970633228836e-07}
{"train_loss": 0.05018476024270058, "global_step": 261736, "epoch": 2940, "lr": 3.1674455119545254e-07}
{"train_loss": 0.020722942426800728, "global_step": 261737, "epoch": 2940, "lr": 3.1667940273792383e-07}
{"train_loss": 0.03431890159845352, "global_step": 261738, "epoch": 2940, "lr": 3.1661426095971334e-07}
{"train_loss": 0.03196471184492111, "global_step": 261739, "epoch": 2940, "lr": 3.165491258608322e-07}
{"train_loss": 0.05327220633625984, "global_step": 261740, "epoch": 2940, "lr": 3.164839974412803e-07}
{"train_loss": 0.028782756999135017, "global_step": 261741, "epoch": 2940, "lr": 3.164188757010744e-07}
{"train_loss": 0.04559142887592316, "global_step": 261742, "epoch": 2940, "lr": 3.1635376064022004e-07}
{"train_loss": 0.04257318377494812, "global_step": 261743, "epoch": 2940, "lr": 3.1628865225872274e-07}
{"train_loss": 0.034494027495384216, "global_step": 261744, "epoch": 2940, "lr": 3.1622355055659914e-07}
{"train_loss": 0.05199894681572914, "global_step": 261745, "epoch": 2940, "lr": 3.161584555338548e-07}
{"train_loss": 0.03329292684793472, "global_step": 261746, "epoch": 2940, "lr": 3.1609336719049533e-07}
{"train_loss": 0.06159387528896332, "global_step": 261747, "epoch": 2940, "lr": 3.1602828552653175e-07}
{"train_loss": 0.04260689312110791, "global_step": 261748, "epoch": 2940, "lr": 3.1596321054196967e-07, "val_loss": 9.603616714477539, "train_action_mse_error": 20.277929306030273}
{"train_loss": 0.028334330767393112, "global_step": 261749, "epoch": 2941, "lr": 3.158981422368257e-07}
{"train_loss": 0.017304839566349983, "global_step": 261750, "epoch": 2941, "lr": 3.1583308061109985e-07}
{"train_loss": 0.025497548282146454, "global_step": 261751, "epoch": 2941, "lr": 3.1576802566480324e-07}
{"train_loss": 0.07447513192892075, "global_step": 261752, "epoch": 2941, "lr": 3.157029773979525e-07}
{"train_loss": 0.055559419095516205, "global_step": 261753, "epoch": 2941, "lr": 3.1563793581054215e-07}
{"train_loss": 0.046728599816560745, "global_step": 261754, "epoch": 2941, "lr": 3.155729009025943e-07}
{"train_loss": 0.0594603531062603, "global_step": 261755, "epoch": 2941, "lr": 3.155078726741145e-07}
{"train_loss": 0.027988256886601448, "global_step": 261756, "epoch": 2941, "lr": 3.1544285112510284e-07}
{"train_loss": 0.0712004005908966, "global_step": 261757, "epoch": 2941, "lr": 3.1537783625557593e-07}
{"train_loss": 0.04254841431975365, "global_step": 261758, "epoch": 2941, "lr": 3.1531282806554486e-07}
{"train_loss": 0.04140157997608185, "global_step": 261759, "epoch": 2941, "lr": 3.152478265550096e-07}
{"train_loss": 0.08046875149011612, "global_step": 261760, "epoch": 2941, "lr": 3.151828317239869e-07}
{"train_loss": 0.019218984991312027, "global_step": 261761, "epoch": 2941, "lr": 3.1511784357247663e-07}
{"train_loss": 0.054492589086294174, "global_step": 261762, "epoch": 2941, "lr": 3.1505286210049555e-07}
{"train_loss": 0.06885665655136108, "global_step": 261763, "epoch": 2941, "lr": 3.149878873080547e-07}
{"train_loss": 0.028159931302070618, "global_step": 261764, "epoch": 2941, "lr": 3.1492291919515413e-07}
{"train_loss": 0.03427254408597946, "global_step": 261765, "epoch": 2941, "lr": 3.148579577618105e-07}
{"train_loss": 0.02911548502743244, "global_step": 261766, "epoch": 2941, "lr": 3.1479300300802375e-07}
{"train_loss": 0.048163678497076035, "global_step": 261767, "epoch": 2941, "lr": 3.147280549338105e-07}
{"train_loss": 0.06748361140489578, "global_step": 261768, "epoch": 2941, "lr": 3.1466311353917644e-07}
{"train_loss": 0.04922245442867279, "global_step": 261769, "epoch": 2941, "lr": 3.14598178824127e-07}
{"train_loss": 0.03788091242313385, "global_step": 261770, "epoch": 2941, "lr": 3.145332507886789e-07}
{"train_loss": 0.03666123002767563, "global_step": 261771, "epoch": 2941, "lr": 3.144683294328321e-07}
{"train_loss": 0.06483978778123856, "global_step": 261772, "epoch": 2941, "lr": 3.144034147566033e-07}
{"train_loss": 0.07445317506790161, "global_step": 261773, "epoch": 2941, "lr": 3.1433850675999245e-07}
{"train_loss": 0.08423394709825516, "global_step": 261774, "epoch": 2941, "lr": 3.142736054430162e-07}
{"train_loss": 0.01537152100354433, "global_step": 261775, "epoch": 2941, "lr": 3.142087108056857e-07}
{"train_loss": 0.03883142024278641, "global_step": 261776, "epoch": 2941, "lr": 3.1414382284799536e-07}
{"train_loss": 0.04724707081913948, "global_step": 261777, "epoch": 2941, "lr": 3.140789415699674e-07}
{"train_loss": 0.017572639510035515, "global_step": 261778, "epoch": 2941, "lr": 3.140140669716074e-07}
{"train_loss": 0.05765375867486, "global_step": 261779, "epoch": 2941, "lr": 3.139491990529153e-07}
{"train_loss": 0.042080044746398926, "global_step": 261780, "epoch": 2941, "lr": 3.1388433781391334e-07}
{"train_loss": 0.05943837761878967, "global_step": 261781, "epoch": 2941, "lr": 3.1381948325460155e-07}
{"train_loss": 0.03397436812520027, "global_step": 261782, "epoch": 2941, "lr": 3.137546353749965e-07}
{"train_loss": 0.02096255123615265, "global_step": 261783, "epoch": 2941, "lr": 3.136897941750927e-07}
{"train_loss": 0.03792715445160866, "global_step": 261784, "epoch": 2941, "lr": 3.136249596549179e-07}
{"train_loss": 0.027859488502144814, "global_step": 261785, "epoch": 2941, "lr": 3.13560131814461e-07}
{"train_loss": 0.029745876789093018, "global_step": 261786, "epoch": 2941, "lr": 3.1349531065374413e-07}
{"train_loss": 0.06709211319684982, "global_step": 261787, "epoch": 2941, "lr": 3.13430496172773e-07}
{"train_loss": 0.030054423958063126, "global_step": 261788, "epoch": 2941, "lr": 3.1336568837155855e-07}
{"train_loss": 0.08800682425498962, "global_step": 261789, "epoch": 2941, "lr": 3.133008872501009e-07}
{"train_loss": 0.03285212814807892, "global_step": 261790, "epoch": 2941, "lr": 3.132360928084166e-07}
{"train_loss": 0.0608295202255249, "global_step": 261791, "epoch": 2941, "lr": 3.131713050465113e-07}
{"train_loss": 0.05328180268406868, "global_step": 261792, "epoch": 2941, "lr": 3.13106523964396e-07}
{"train_loss": 0.03546471148729324, "global_step": 261793, "epoch": 2941, "lr": 3.1304174956207634e-07}
{"train_loss": 0.04323193430900574, "global_step": 261794, "epoch": 2941, "lr": 3.129769818395634e-07}
{"train_loss": 0.049073003232479095, "global_step": 261795, "epoch": 2941, "lr": 3.129122207968627e-07}
{"train_loss": 0.0544901043176651, "global_step": 261796, "epoch": 2941, "lr": 3.128474664339909e-07}
{"train_loss": 0.040190689265728, "global_step": 261797, "epoch": 2941, "lr": 3.12782718750948e-07}
{"train_loss": 0.03526223823428154, "global_step": 261798, "epoch": 2941, "lr": 3.127179777477396e-07}
{"train_loss": 0.05480870231986046, "global_step": 261799, "epoch": 2941, "lr": 3.126532434243934e-07}
{"train_loss": 0.017217563465237617, "global_step": 261800, "epoch": 2941, "lr": 3.125885157808928e-07}
{"train_loss": 0.02547420933842659, "global_step": 261801, "epoch": 2941, "lr": 3.12523794817271e-07}
{"train_loss": 0.03149037063121796, "global_step": 261802, "epoch": 2941, "lr": 3.12459080533517e-07}
{"train_loss": 0.030572226271033287, "global_step": 261803, "epoch": 2941, "lr": 3.1239437292964744e-07}
{"train_loss": 0.03543980047106743, "global_step": 261804, "epoch": 2941, "lr": 3.1232967200567345e-07}
{"train_loss": 0.01052356231957674, "global_step": 261805, "epoch": 2941, "lr": 3.122649777616005e-07}
{"train_loss": 0.03063955530524254, "global_step": 261806, "epoch": 2941, "lr": 3.122002901974341e-07}
{"train_loss": 0.06773662567138672, "global_step": 261807, "epoch": 2941, "lr": 3.1213560931319106e-07}
{"train_loss": 0.026413768529891968, "global_step": 261808, "epoch": 2941, "lr": 3.1207093510887684e-07}
{"train_loss": 0.06719721853733063, "global_step": 261809, "epoch": 2941, "lr": 3.12006267584497e-07}
{"train_loss": 0.04343242943286896, "global_step": 261810, "epoch": 2941, "lr": 3.1194160674006267e-07}
{"train_loss": 0.0504627600312233, "global_step": 261811, "epoch": 2941, "lr": 3.118769525755849e-07}
{"train_loss": 0.049100328236818314, "global_step": 261812, "epoch": 2941, "lr": 3.1181230509106375e-07}
{"train_loss": 0.07135052978992462, "global_step": 261813, "epoch": 2941, "lr": 3.117476642865158e-07}
{"train_loss": 0.04920743405818939, "global_step": 261814, "epoch": 2941, "lr": 3.116830301619522e-07}
{"train_loss": 0.031700994819402695, "global_step": 261815, "epoch": 2941, "lr": 3.1161840271737296e-07}
{"train_loss": 0.05569949001073837, "global_step": 261816, "epoch": 2941, "lr": 3.1155378195278915e-07}
{"train_loss": 0.037570465356111526, "global_step": 261817, "epoch": 2941, "lr": 3.114891678682175e-07}
{"train_loss": 0.034186702221632004, "global_step": 261818, "epoch": 2941, "lr": 3.114245604636579e-07}
{"train_loss": 0.021135199815034866, "global_step": 261819, "epoch": 2941, "lr": 3.1135995973912144e-07}
{"train_loss": 0.02223009057343006, "global_step": 261820, "epoch": 2941, "lr": 3.112953656946138e-07}
{"train_loss": 0.05580432713031769, "global_step": 261821, "epoch": 2941, "lr": 3.1123077833015156e-07}
{"train_loss": 0.041075680404901505, "global_step": 261822, "epoch": 2941, "lr": 3.1116619764573474e-07}
{"train_loss": 0.03389904275536537, "global_step": 261823, "epoch": 2941, "lr": 3.1110162364137995e-07}
{"train_loss": 0.030394544824957848, "global_step": 261824, "epoch": 2941, "lr": 3.1103705631708723e-07}
{"train_loss": 0.05922931432723999, "global_step": 261825, "epoch": 2941, "lr": 3.109724956728732e-07}
{"train_loss": 0.03221195563673973, "global_step": 261826, "epoch": 2941, "lr": 3.109079417087435e-07}
{"train_loss": 0.0396587997674942, "global_step": 261827, "epoch": 2941, "lr": 3.1084339442470356e-07}
{"train_loss": 0.021720467135310173, "global_step": 261828, "epoch": 2941, "lr": 3.107788538207701e-07}
{"train_loss": 0.04406800866127014, "global_step": 261829, "epoch": 2941, "lr": 3.107143198969431e-07}
{"train_loss": 0.04315663501620293, "global_step": 261830, "epoch": 2941, "lr": 3.1064979265323925e-07}
{"train_loss": 0.07316900789737701, "global_step": 261831, "epoch": 2941, "lr": 3.1058527208965847e-07}
{"train_loss": 0.08266483247280121, "global_step": 261832, "epoch": 2941, "lr": 3.1052075820621197e-07}
{"train_loss": 0.050592247396707535, "global_step": 261833, "epoch": 2941, "lr": 3.1045625100291633e-07}
{"train_loss": 0.015449777245521545, "global_step": 261834, "epoch": 2941, "lr": 3.10391750479766e-07}
{"train_loss": 0.0468999482691288, "global_step": 261835, "epoch": 2941, "lr": 3.1032725663678317e-07}
{"train_loss": 0.0475037582218647, "global_step": 261836, "epoch": 2941, "lr": 3.102627694739735e-07}
{"train_loss": 0.04391088732256648, "global_step": 261837, "epoch": 2941, "lr": 3.101982889913424e-07, "val_loss": 9.595210075378418}
{"train_loss": 0.03429113328456879, "global_step": 261838, "epoch": 2942, "lr": 3.101338151888955e-07}
{"train_loss": 0.041216205805540085, "global_step": 261839, "epoch": 2942, "lr": 3.1006934806664946e-07}
{"train_loss": 0.04364148899912834, "global_step": 261840, "epoch": 2942, "lr": 3.100048876246042e-07}
{"train_loss": 0.04264620691537857, "global_step": 261841, "epoch": 2942, "lr": 3.0994043386278206e-07}
{"train_loss": 0.03774991258978844, "global_step": 261842, "epoch": 2942, "lr": 3.098759867811718e-07}
{"train_loss": 0.03520742058753967, "global_step": 261843, "epoch": 2942, "lr": 3.0981154637980125e-07}
{"train_loss": 0.07678134739398956, "global_step": 261844, "epoch": 2942, "lr": 3.097471126586649e-07}
{"train_loss": 0.023897850885987282, "global_step": 261845, "epoch": 2942, "lr": 3.096826856177848e-07}
{"train_loss": 0.013626118190586567, "global_step": 261846, "epoch": 2942, "lr": 3.0961826525715555e-07}
{"train_loss": 0.03943303972482681, "global_step": 261847, "epoch": 2942, "lr": 3.0955385157679927e-07}
{"train_loss": 0.04087473452091217, "global_step": 261848, "epoch": 2942, "lr": 3.094894445767105e-07}
{"train_loss": 0.016199002042412758, "global_step": 261849, "epoch": 2942, "lr": 3.0942504425691135e-07}
{"train_loss": 0.023674001917243004, "global_step": 261850, "epoch": 2942, "lr": 3.0936065061739627e-07}
{"train_loss": 0.08725380897521973, "global_step": 261851, "epoch": 2942, "lr": 3.0929626365818753e-07}
{"train_loss": 0.06807135790586472, "global_step": 261852, "epoch": 2942, "lr": 3.0923188337929063e-07}
{"train_loss": 0.036440446972846985, "global_step": 261853, "epoch": 2942, "lr": 3.0916750978070563e-07}
{"train_loss": 0.02806439995765686, "global_step": 261854, "epoch": 2942, "lr": 3.0910314286245466e-07}
{"train_loss": 0.021690666675567627, "global_step": 261855, "epoch": 2942, "lr": 3.090387826245322e-07}
{"train_loss": 0.040410690009593964, "global_step": 261856, "epoch": 2942, "lr": 3.089744290669605e-07}
{"train_loss": 0.04839376360177994, "global_step": 261857, "epoch": 2942, "lr": 3.0891008218973396e-07}
{"train_loss": 0.03788325563073158, "global_step": 261858, "epoch": 2942, "lr": 3.0884574199287473e-07}
{"train_loss": 0.03246387466788292, "global_step": 261859, "epoch": 2942, "lr": 3.0878140847638294e-07}
{"train_loss": 0.017528917640447617, "global_step": 261860, "epoch": 2942, "lr": 3.0871708164027514e-07}
{"train_loss": 0.04380141198635101, "global_step": 261861, "epoch": 2942, "lr": 3.086527614845458e-07}
{"train_loss": 0.062281280755996704, "global_step": 261862, "epoch": 2942, "lr": 3.0858844800921717e-07}
{"train_loss": 0.062420010566711426, "global_step": 261863, "epoch": 2942, "lr": 3.0852414121429473e-07}
{"train_loss": 0.029552258551120758, "global_step": 261864, "epoch": 2942, "lr": 3.084598410997841e-07}
{"train_loss": 0.07747720927000046, "global_step": 261865, "epoch": 2942, "lr": 3.083955476656908e-07}
{"train_loss": 0.06575389951467514, "global_step": 261866, "epoch": 2942, "lr": 3.083312609120315e-07}
{"train_loss": 0.032999660819768906, "global_step": 261867, "epoch": 2942, "lr": 3.082669808388117e-07}
{"train_loss": 0.05637148395180702, "global_step": 261868, "epoch": 2942, "lr": 3.082027074460425e-07}
{"train_loss": 0.022665519267320633, "global_step": 261869, "epoch": 2942, "lr": 3.08138440733724e-07}
{"train_loss": 0.03657892718911171, "global_step": 261870, "epoch": 2942, "lr": 3.080741807018728e-07}
{"train_loss": 0.029270509257912636, "global_step": 261871, "epoch": 2942, "lr": 3.080099273504999e-07}
{"train_loss": 0.045226696878671646, "global_step": 261872, "epoch": 2942, "lr": 3.079456806796055e-07}
{"train_loss": 0.06337007135152817, "global_step": 261873, "epoch": 2942, "lr": 3.0788144068920054e-07}
{"train_loss": 0.04783640429377556, "global_step": 261874, "epoch": 2942, "lr": 3.078172073792962e-07}
{"train_loss": 0.09590884298086166, "global_step": 261875, "epoch": 2942, "lr": 3.0775298074990357e-07}
{"train_loss": 0.05071747675538063, "global_step": 261876, "epoch": 2942, "lr": 3.076887608010282e-07}
{"train_loss": 0.03208693116903305, "global_step": 261877, "epoch": 2942, "lr": 3.0762454753267554e-07}
{"train_loss": 0.07254906743764877, "global_step": 261878, "epoch": 2942, "lr": 3.075603409448569e-07}
{"train_loss": 0.06995430588722229, "global_step": 261879, "epoch": 2942, "lr": 3.0749614103758317e-07}
{"train_loss": 0.06776529550552368, "global_step": 261880, "epoch": 2942, "lr": 3.074319478108545e-07}
{"train_loss": 0.02632712945342064, "global_step": 261881, "epoch": 2942, "lr": 3.0736776126469304e-07}
{"train_loss": 0.01716718263924122, "global_step": 261882, "epoch": 2942, "lr": 3.0730358139909876e-07}
{"train_loss": 0.06480774283409119, "global_step": 261883, "epoch": 2942, "lr": 3.072394082140828e-07}
{"train_loss": 0.07924658805131912, "global_step": 261884, "epoch": 2942, "lr": 3.0717524170965074e-07}
{"train_loss": 0.07242149114608765, "global_step": 261885, "epoch": 2942, "lr": 3.071110818858136e-07}
{"train_loss": 0.025406476110219955, "global_step": 261886, "epoch": 2942, "lr": 3.07046928742577e-07}
{"train_loss": 0.03263973444700241, "global_step": 261887, "epoch": 2942, "lr": 3.0698278227995756e-07}
{"train_loss": 0.05844639241695404, "global_step": 261888, "epoch": 2942, "lr": 3.069186424979553e-07}
{"train_loss": 0.06503675132989883, "global_step": 261889, "epoch": 2942, "lr": 3.068545093965813e-07}
{"train_loss": 0.047691710293293, "global_step": 261890, "epoch": 2942, "lr": 3.067903829758467e-07}
{"train_loss": 0.011050904169678688, "global_step": 261891, "epoch": 2942, "lr": 3.067262632357626e-07}
{"train_loss": 0.03846006840467453, "global_step": 261892, "epoch": 2942, "lr": 3.066621501763234e-07}
{"train_loss": 0.06525208801031113, "global_step": 261893, "epoch": 2942, "lr": 3.0659804379755685e-07}
{"train_loss": 0.043545130640268326, "global_step": 261894, "epoch": 2942, "lr": 3.06533944099463e-07}
{"train_loss": 0.043196696788072586, "global_step": 261895, "epoch": 2942, "lr": 3.0646985108204187e-07}
{"train_loss": 0.039196308702230453, "global_step": 261896, "epoch": 2942, "lr": 3.0640576474531555e-07}
{"train_loss": 0.048250917345285416, "global_step": 261897, "epoch": 2942, "lr": 3.063416850892897e-07}
{"train_loss": 0.019276509061455727, "global_step": 261898, "epoch": 2942, "lr": 3.0627761211396435e-07}
{"train_loss": 0.04725724458694458, "global_step": 261899, "epoch": 2942, "lr": 3.06213545819356e-07}
{"train_loss": 0.05927757918834686, "global_step": 261900, "epoch": 2942, "lr": 3.061494862054759e-07}
{"train_loss": 0.06704385578632355, "global_step": 261901, "epoch": 2942, "lr": 3.0608543327232396e-07}
{"train_loss": 0.032778382301330566, "global_step": 261902, "epoch": 2942, "lr": 3.060213870199169e-07}
{"train_loss": 0.029538221657276154, "global_step": 261903, "epoch": 2942, "lr": 3.059573474482547e-07}
{"train_loss": 0.01835288479924202, "global_step": 261904, "epoch": 2942, "lr": 3.058933145573539e-07}
{"train_loss": 0.058922410011291504, "global_step": 261905, "epoch": 2942, "lr": 3.0582928834722024e-07}
{"train_loss": 0.06703660637140274, "global_step": 261906, "epoch": 2942, "lr": 3.057652688178592e-07}
{"train_loss": 0.026170283555984497, "global_step": 261907, "epoch": 2942, "lr": 3.057012559692873e-07}
{"train_loss": 0.04725147783756256, "global_step": 261908, "epoch": 2942, "lr": 3.0563724980150477e-07}
{"train_loss": 0.05592946708202362, "global_step": 261909, "epoch": 2942, "lr": 3.0557325031452254e-07}
{"train_loss": 0.020450975745916367, "global_step": 261910, "epoch": 2942, "lr": 3.055092575083518e-07}
{"train_loss": 0.029145507141947746, "global_step": 261911, "epoch": 2942, "lr": 3.0544527138299804e-07}
{"train_loss": 0.032715097069740295, "global_step": 261912, "epoch": 2942, "lr": 3.053812919384724e-07}
{"train_loss": 0.03000960871577263, "global_step": 261913, "epoch": 2942, "lr": 3.05317319174786e-07}
{"train_loss": 0.06018628552556038, "global_step": 261914, "epoch": 2942, "lr": 3.052533530919388e-07}
{"train_loss": 0.04255243018269539, "global_step": 261915, "epoch": 2942, "lr": 3.051893936899475e-07}
{"train_loss": 0.0311236921697855, "global_step": 261916, "epoch": 2942, "lr": 3.0512544096881755e-07}
{"train_loss": 0.03157772868871689, "global_step": 261917, "epoch": 2942, "lr": 3.0506149492856016e-07}
{"train_loss": 0.05221190303564072, "global_step": 261918, "epoch": 2942, "lr": 3.049975555691753e-07}
{"train_loss": 0.019187845289707184, "global_step": 261919, "epoch": 2942, "lr": 3.049336228906852e-07}
{"train_loss": 0.03585979342460632, "global_step": 261920, "epoch": 2942, "lr": 3.048696968930842e-07}
{"train_loss": 0.0760243833065033, "global_step": 261921, "epoch": 2942, "lr": 3.048057775763946e-07}
{"train_loss": 0.03695378080010414, "global_step": 261922, "epoch": 2942, "lr": 3.0474186494061087e-07}
{"train_loss": 0.028586342930793762, "global_step": 261923, "epoch": 2942, "lr": 3.0467795898575514e-07}
{"train_loss": 0.024887345731258392, "global_step": 261924, "epoch": 2942, "lr": 3.0461405971182746e-07}
{"train_loss": 0.10968393087387085, "global_step": 261925, "epoch": 2942, "lr": 3.0455016711883886e-07}
{"train_loss": 0.04454416847588976, "global_step": 261926, "epoch": 2942, "lr": 3.04486281206795e-07, "val_loss": 9.672738075256348}
{"train_loss": 0.023947695270180702, "global_step": 261927, "epoch": 2943, "lr": 3.044224019757125e-07}
{"train_loss": 0.06393250823020935, "global_step": 261928, "epoch": 2943, "lr": 3.043585294255913e-07}
{"train_loss": 0.034741200506687164, "global_step": 261929, "epoch": 2943, "lr": 3.0429466355644253e-07}
{"train_loss": 0.03870903700590134, "global_step": 261930, "epoch": 2943, "lr": 3.0423080436827735e-07}
{"train_loss": 0.018974879756569862, "global_step": 261931, "epoch": 2943, "lr": 3.041669518611068e-07}
{"train_loss": 0.025601886212825775, "global_step": 261932, "epoch": 2943, "lr": 3.041031060349253e-07}
{"train_loss": 0.03662698343396187, "global_step": 261933, "epoch": 2943, "lr": 3.040392668897607e-07}
{"train_loss": 0.057089898735284805, "global_step": 261934, "epoch": 2943, "lr": 3.039754344256074e-07}
{"train_loss": 0.045546405017375946, "global_step": 261935, "epoch": 2943, "lr": 3.0391160864248204e-07}
{"train_loss": 0.02503509446978569, "global_step": 261936, "epoch": 2943, "lr": 3.0384778954039017e-07}
{"train_loss": 0.047719165682792664, "global_step": 261937, "epoch": 2943, "lr": 3.0378397711933737e-07}
{"train_loss": 0.016745958477258682, "global_step": 261938, "epoch": 2943, "lr": 3.037201713793347e-07}
{"train_loss": 0.023790599778294563, "global_step": 261939, "epoch": 2943, "lr": 3.0365637232039336e-07}
{"train_loss": 0.05086826533079147, "global_step": 261940, "epoch": 2943, "lr": 3.035925799425188e-07}
{"train_loss": 0.02975984290242195, "global_step": 261941, "epoch": 2943, "lr": 3.0352879424571657e-07}
{"train_loss": 0.033502254635095596, "global_step": 261942, "epoch": 2943, "lr": 3.0346501523000335e-07}
{"train_loss": 0.021218756213784218, "global_step": 261943, "epoch": 2943, "lr": 3.0340124289538475e-07}
{"train_loss": 0.029897280037403107, "global_step": 261944, "epoch": 2943, "lr": 3.0333747724186626e-07}
{"train_loss": 0.064960777759552, "global_step": 261945, "epoch": 2943, "lr": 3.0327371826945894e-07}
{"train_loss": 0.10257823765277863, "global_step": 261946, "epoch": 2943, "lr": 3.032099659781684e-07}
{"train_loss": 0.03146565705537796, "global_step": 261947, "epoch": 2943, "lr": 3.031462203680113e-07}
{"train_loss": 0.05471937358379364, "global_step": 261948, "epoch": 2943, "lr": 3.030824814389821e-07}
{"train_loss": 0.06002485752105713, "global_step": 261949, "epoch": 2943, "lr": 3.0301874919110295e-07}
{"train_loss": 0.039537712931632996, "global_step": 261950, "epoch": 2943, "lr": 3.0295502362437944e-07}
{"train_loss": 0.043293751776218414, "global_step": 261951, "epoch": 2943, "lr": 3.0289130473881713e-07}
{"train_loss": 0.060002319514751434, "global_step": 261952, "epoch": 2943, "lr": 3.028275925344215e-07}
{"train_loss": 0.014955862425267696, "global_step": 261953, "epoch": 2943, "lr": 3.027638870112093e-07}
{"train_loss": 0.014668883755803108, "global_step": 261954, "epoch": 2943, "lr": 3.0270018816918046e-07}
{"train_loss": 0.09271050989627838, "global_step": 261955, "epoch": 2943, "lr": 3.0263649600835165e-07}
{"train_loss": 0.025400277227163315, "global_step": 261956, "epoch": 2943, "lr": 3.025728105287229e-07}
{"train_loss": 0.012928081676363945, "global_step": 261957, "epoch": 2943, "lr": 3.025091317303108e-07}
{"train_loss": 0.07103193551301956, "global_step": 261958, "epoch": 2943, "lr": 3.02445459613121e-07}
{"train_loss": 0.05302109569311142, "global_step": 261959, "epoch": 2943, "lr": 3.0238179417715893e-07}
{"train_loss": 0.036556683480739594, "global_step": 261960, "epoch": 2943, "lr": 3.023181354224358e-07}
{"train_loss": 0.039292894303798676, "global_step": 261961, "epoch": 2943, "lr": 3.0225448334896265e-07}
{"train_loss": 0.027359716594219208, "global_step": 261962, "epoch": 2943, "lr": 3.021908379567451e-07}
{"train_loss": 0.06785038113594055, "global_step": 261963, "epoch": 2943, "lr": 3.021271992457941e-07}
{"train_loss": 0.041281022131443024, "global_step": 261964, "epoch": 2943, "lr": 3.020635672161154e-07}
{"train_loss": 0.06939703971147537, "global_step": 261965, "epoch": 2943, "lr": 3.019999418677144e-07}
{"train_loss": 0.033220961689949036, "global_step": 261966, "epoch": 2943, "lr": 3.019363232006078e-07}
{"train_loss": 0.04919550567865372, "global_step": 261967, "epoch": 2943, "lr": 3.0187271121480123e-07}
{"train_loss": 0.047110818326473236, "global_step": 261968, "epoch": 2943, "lr": 3.0180910591030007e-07}
{"train_loss": 0.07881532609462738, "global_step": 261969, "epoch": 2943, "lr": 3.017455072871156e-07}
{"train_loss": 0.061175692826509476, "global_step": 261970, "epoch": 2943, "lr": 3.016819153452588e-07}
{"train_loss": 0.018361857160925865, "global_step": 261971, "epoch": 2943, "lr": 3.016183300847297e-07}
{"train_loss": 0.017893807962536812, "global_step": 261972, "epoch": 2943, "lr": 3.01554751505545e-07}
{"train_loss": 0.0682191327214241, "global_step": 261973, "epoch": 2943, "lr": 3.014911796077102e-07}
{"train_loss": 0.04892570525407791, "global_step": 261974, "epoch": 2943, "lr": 3.0142761439123647e-07}
{"train_loss": 0.07070871442556381, "global_step": 261975, "epoch": 2943, "lr": 3.013640558561237e-07}
{"train_loss": 0.04089067876338959, "global_step": 261976, "epoch": 2943, "lr": 3.013005040023942e-07}
{"train_loss": 0.04008428379893303, "global_step": 261977, "epoch": 2943, "lr": 3.012369588300423e-07}
{"train_loss": 0.032231561839580536, "global_step": 261978, "epoch": 2943, "lr": 3.011734203390903e-07}
{"train_loss": 0.06614428758621216, "global_step": 261979, "epoch": 2943, "lr": 3.011098885295327e-07}
{"train_loss": 0.06595389544963837, "global_step": 261980, "epoch": 2943, "lr": 3.0104636340138604e-07}
{"train_loss": 0.03618255630135536, "global_step": 261981, "epoch": 2943, "lr": 3.0098284495466143e-07}
{"train_loss": 0.026176393032073975, "global_step": 261982, "epoch": 2943, "lr": 3.0091933318936447e-07}
{"train_loss": 0.047804176807403564, "global_step": 261983, "epoch": 2943, "lr": 3.008558281055007e-07}
{"train_loss": 0.05719519034028053, "global_step": 261984, "epoch": 2943, "lr": 3.0079232970308125e-07}
{"train_loss": 0.05746406316757202, "global_step": 261985, "epoch": 2943, "lr": 3.0072883798211716e-07}
{"train_loss": 0.04969630762934685, "global_step": 261986, "epoch": 2943, "lr": 3.0066535294260847e-07}
{"train_loss": 0.07312704622745514, "global_step": 261987, "epoch": 2943, "lr": 3.0060187458457735e-07}
{"train_loss": 0.041626811027526855, "global_step": 261988, "epoch": 2943, "lr": 3.0053840290801825e-07}
{"train_loss": 0.029028160497546196, "global_step": 261989, "epoch": 2943, "lr": 3.004749379129479e-07}
{"train_loss": 0.07699577510356903, "global_step": 261990, "epoch": 2943, "lr": 3.004114795993718e-07}
{"train_loss": 0.06623221933841705, "global_step": 261991, "epoch": 2943, "lr": 3.0034802796730655e-07}
{"train_loss": 0.01907687447965145, "global_step": 261992, "epoch": 2943, "lr": 3.002845830167467e-07}
{"train_loss": 0.012955011799931526, "global_step": 261993, "epoch": 2943, "lr": 3.002211447477088e-07}
{"train_loss": 0.042117614299058914, "global_step": 261994, "epoch": 2943, "lr": 3.0015771316019847e-07}
{"train_loss": 0.03371872752904892, "global_step": 261995, "epoch": 2943, "lr": 3.0009428825423236e-07}
{"train_loss": 0.03624625876545906, "global_step": 261996, "epoch": 2943, "lr": 3.000308700298105e-07}
{"train_loss": 0.0476282499730587, "global_step": 261997, "epoch": 2943, "lr": 2.999674584869383e-07}
{"train_loss": 0.02141920104622841, "global_step": 261998, "epoch": 2943, "lr": 2.9990405362563814e-07}
{"train_loss": 0.08771945536136627, "global_step": 261999, "epoch": 2943, "lr": 2.9984065544590436e-07}
{"train_loss": 0.03036203421652317, "global_step": 262000, "epoch": 2943, "lr": 2.997772639477536e-07}
{"train_loss": 0.06430164724588394, "global_step": 262001, "epoch": 2943, "lr": 2.9971387913119155e-07}
{"train_loss": 0.0437152162194252, "global_step": 262002, "epoch": 2943, "lr": 2.996505009962236e-07}
{"train_loss": 0.02392263524234295, "global_step": 262003, "epoch": 2943, "lr": 2.995871295428665e-07}
{"train_loss": 0.04027102515101433, "global_step": 262004, "epoch": 2943, "lr": 2.9952376477112574e-07}
{"train_loss": 0.0663866400718689, "global_step": 262005, "epoch": 2943, "lr": 2.994604066810069e-07}
{"train_loss": 0.03436921536922455, "global_step": 262006, "epoch": 2943, "lr": 2.993970552725156e-07}
{"train_loss": 0.043043844401836395, "global_step": 262007, "epoch": 2943, "lr": 2.993337105456684e-07}
{"train_loss": 0.025694821029901505, "global_step": 262008, "epoch": 2943, "lr": 2.992703725004653e-07}
{"train_loss": 0.03970764949917793, "global_step": 262009, "epoch": 2943, "lr": 2.9920704113692854e-07}
{"train_loss": 0.06396681070327759, "global_step": 262010, "epoch": 2943, "lr": 2.9914371645505257e-07}
{"train_loss": 0.02812812849879265, "global_step": 262011, "epoch": 2943, "lr": 2.9908039845484845e-07}
{"train_loss": 0.034268688410520554, "global_step": 262012, "epoch": 2943, "lr": 2.990170871363329e-07}
{"train_loss": 0.04791072756052017, "global_step": 262013, "epoch": 2943, "lr": 2.9895378249950033e-07}
{"train_loss": 0.03790860250592232, "global_step": 262014, "epoch": 2943, "lr": 2.9889048454437294e-07}
{"train_loss": 0.04389536314842741, "global_step": 262015, "epoch": 2943, "lr": 2.9882719327095633e-07, "val_loss": 9.767440795898438}
{"train_loss": 0.04191930592060089, "global_step": 262016, "epoch": 2944, "lr": 2.9876390867925044e-07}
{"train_loss": 0.023298585787415504, "global_step": 262017, "epoch": 2944, "lr": 2.9870063076927744e-07}
{"train_loss": 0.04200991988182068, "global_step": 262018, "epoch": 2944, "lr": 2.986373595410319e-07}
{"train_loss": 0.038538865745067596, "global_step": 262019, "epoch": 2944, "lr": 2.985740949945304e-07}
{"train_loss": 0.026840368285775185, "global_step": 262020, "epoch": 2944, "lr": 2.9851083712977843e-07}
{"train_loss": 0.06550253182649612, "global_step": 262021, "epoch": 2944, "lr": 2.984475859467928e-07}
{"train_loss": 0.012980048544704914, "global_step": 262022, "epoch": 2944, "lr": 2.9838434144556784e-07}
{"train_loss": 0.06499534100294113, "global_step": 262023, "epoch": 2944, "lr": 2.9832110362612577e-07}
{"train_loss": 0.04819265380501747, "global_step": 262024, "epoch": 2944, "lr": 2.9825787248846107e-07}
{"train_loss": 0.050515130162239075, "global_step": 262025, "epoch": 2944, "lr": 2.981946480325959e-07}
{"train_loss": 0.07725675404071808, "global_step": 262026, "epoch": 2944, "lr": 2.981314302585303e-07}
{"train_loss": 0.07525092363357544, "global_step": 262027, "epoch": 2944, "lr": 2.9806821916627534e-07}
{"train_loss": 0.051669053733348846, "global_step": 262028, "epoch": 2944, "lr": 2.9800501475583664e-07}
{"train_loss": 0.03722355514764786, "global_step": 262029, "epoch": 2944, "lr": 2.979418170272308e-07}
{"train_loss": 0.03126923367381096, "global_step": 262030, "epoch": 2944, "lr": 2.978786259804578e-07}
{"train_loss": 0.09395302087068558, "global_step": 262031, "epoch": 2944, "lr": 2.9781544161552877e-07}
{"train_loss": 0.06885845959186554, "global_step": 262032, "epoch": 2944, "lr": 2.9775226393244926e-07}
{"train_loss": 0.03588191419839859, "global_step": 262033, "epoch": 2944, "lr": 2.976890929312359e-07}
{"train_loss": 0.07128393650054932, "global_step": 262034, "epoch": 2944, "lr": 2.976259286118943e-07}
{"train_loss": 0.040168847888708115, "global_step": 262035, "epoch": 2944, "lr": 2.9756277097442444e-07}
{"train_loss": 0.02675686776638031, "global_step": 262036, "epoch": 2944, "lr": 2.9749962001884845e-07}
{"train_loss": 0.054401714354753494, "global_step": 262037, "epoch": 2944, "lr": 2.974364757451664e-07}
{"train_loss": 0.03371583670377731, "global_step": 262038, "epoch": 2944, "lr": 2.973733381533894e-07}
{"train_loss": 0.020334456115961075, "global_step": 262039, "epoch": 2944, "lr": 2.9731020724351744e-07}
{"train_loss": 0.05912008136510849, "global_step": 262040, "epoch": 2944, "lr": 2.972470830155727e-07}
{"train_loss": 0.02559092827141285, "global_step": 262041, "epoch": 2944, "lr": 2.971839654695552e-07}
{"train_loss": 0.05268554016947746, "global_step": 262042, "epoch": 2944, "lr": 2.97120854605476e-07}
{"train_loss": 0.028339099138975143, "global_step": 262043, "epoch": 2944, "lr": 2.9705775042334625e-07}
{"train_loss": 0.016083646565675735, "global_step": 262044, "epoch": 2944, "lr": 2.9699465292316596e-07}
{"train_loss": 0.04680347070097923, "global_step": 262045, "epoch": 2944, "lr": 2.969315621049518e-07}
{"train_loss": 0.020159419625997543, "global_step": 262046, "epoch": 2944, "lr": 2.968684779687092e-07}
{"train_loss": 0.044992201030254364, "global_step": 262047, "epoch": 2944, "lr": 2.968054005144494e-07}
{"train_loss": 0.06801940500736237, "global_step": 262048, "epoch": 2944, "lr": 2.967423297421723e-07}
{"train_loss": 0.06443408876657486, "global_step": 262049, "epoch": 2944, "lr": 2.966792656518946e-07}
{"train_loss": 0.04511146619915962, "global_step": 262050, "epoch": 2944, "lr": 2.9661620824362746e-07}
{"train_loss": 0.05465362221002579, "global_step": 262051, "epoch": 2944, "lr": 2.965531575173708e-07}
{"train_loss": 0.039729561656713486, "global_step": 262052, "epoch": 2944, "lr": 2.964901134731302e-07}
{"train_loss": 0.05847412347793579, "global_step": 262053, "epoch": 2944, "lr": 2.9642707611093336e-07}
{"train_loss": 0.06579546630382538, "global_step": 262054, "epoch": 2944, "lr": 2.963640454307637e-07}
{"train_loss": 0.0423491895198822, "global_step": 262055, "epoch": 2944, "lr": 2.9630102143264894e-07}
{"train_loss": 0.05774832144379616, "global_step": 262056, "epoch": 2944, "lr": 2.962380041165891e-07}
{"train_loss": 0.03450407460331917, "global_step": 262057, "epoch": 2944, "lr": 2.961749934826008e-07}
{"train_loss": 0.058186110109090805, "global_step": 262058, "epoch": 2944, "lr": 2.961119895306785e-07}
{"train_loss": 0.023371083661913872, "global_step": 262059, "epoch": 2944, "lr": 2.9604899226083893e-07}
{"train_loss": 0.04066706821322441, "global_step": 262060, "epoch": 2944, "lr": 2.959860016730875e-07}
{"train_loss": 0.03346884250640869, "global_step": 262061, "epoch": 2944, "lr": 2.95923017767441e-07}
{"train_loss": 0.05768260359764099, "global_step": 262062, "epoch": 2944, "lr": 2.958600405438994e-07}
{"train_loss": 0.0888797864317894, "global_step": 262063, "epoch": 2944, "lr": 2.957970700024737e-07}
{"train_loss": 0.05936045944690704, "global_step": 262064, "epoch": 2944, "lr": 2.9573410614316957e-07}
{"train_loss": 0.06456668674945831, "global_step": 262065, "epoch": 2944, "lr": 2.956711489660036e-07}
{"train_loss": 0.02930716797709465, "global_step": 262066, "epoch": 2944, "lr": 2.9560819847097024e-07}
{"train_loss": 0.04429273307323456, "global_step": 262067, "epoch": 2944, "lr": 2.9554525465809723e-07}
{"train_loss": 0.03573932126164436, "global_step": 262068, "epoch": 2944, "lr": 2.9548231752737355e-07}
{"train_loss": 0.11550865322351456, "global_step": 262069, "epoch": 2944, "lr": 2.954193870788213e-07}
{"train_loss": 0.0383414626121521, "global_step": 262070, "epoch": 2944, "lr": 2.953564633124406e-07}
{"train_loss": 0.012198339216411114, "global_step": 262071, "epoch": 2944, "lr": 2.9529354622824244e-07}
{"train_loss": 0.08727433532476425, "global_step": 262072, "epoch": 2944, "lr": 2.9523063582624353e-07}
{"train_loss": 0.03723309561610222, "global_step": 262073, "epoch": 2944, "lr": 2.951677321064383e-07}
{"train_loss": 0.05329880118370056, "global_step": 262074, "epoch": 2944, "lr": 2.95104835068849e-07}
{"train_loss": 0.06897024810314178, "global_step": 262075, "epoch": 2944, "lr": 2.9504194471347e-07}
{"train_loss": 0.042668458074331284, "global_step": 262076, "epoch": 2944, "lr": 2.94979061040318e-07}
{"train_loss": 0.042596299201250076, "global_step": 262077, "epoch": 2944, "lr": 2.949161840494041e-07}
{"train_loss": 0.041379135102033615, "global_step": 262078, "epoch": 2944, "lr": 2.948533137407283e-07}
{"train_loss": 0.023666922003030777, "global_step": 262079, "epoch": 2944, "lr": 2.947904501143073e-07}
{"train_loss": 0.05585780367255211, "global_step": 262080, "epoch": 2944, "lr": 2.9472759317014096e-07}
{"train_loss": 0.022643232718110085, "global_step": 262081, "epoch": 2944, "lr": 2.946647429082461e-07}
{"train_loss": 0.04882580786943436, "global_step": 262082, "epoch": 2944, "lr": 2.9460189932862815e-07}
{"train_loss": 0.028836442157626152, "global_step": 262083, "epoch": 2944, "lr": 2.9453906243129825e-07}
{"train_loss": 0.041210830211639404, "global_step": 262084, "epoch": 2944, "lr": 2.944762322162564e-07}
{"train_loss": 0.060602881014347076, "global_step": 262085, "epoch": 2944, "lr": 2.944134086835193e-07}
{"train_loss": 0.035264771431684494, "global_step": 262086, "epoch": 2944, "lr": 2.9435059183309244e-07}
{"train_loss": 0.04366098344326019, "global_step": 262087, "epoch": 2944, "lr": 2.942877816649814e-07}
{"train_loss": 0.045145899057388306, "global_step": 262088, "epoch": 2944, "lr": 2.9422497817919723e-07}
{"train_loss": 0.03202509135007858, "global_step": 262089, "epoch": 2944, "lr": 2.941621813757511e-07}
{"train_loss": 0.030115902423858643, "global_step": 262090, "epoch": 2944, "lr": 2.940993912546486e-07}
{"train_loss": 0.037247974425554276, "global_step": 262091, "epoch": 2944, "lr": 2.940366078159007e-07}
{"train_loss": 0.0546315461397171, "global_step": 262092, "epoch": 2944, "lr": 2.93973831059513e-07}
{"train_loss": 0.0340685211122036, "global_step": 262093, "epoch": 2944, "lr": 2.9391106098549117e-07}
{"train_loss": 0.09442972391843796, "global_step": 262094, "epoch": 2944, "lr": 2.938482975938517e-07}
{"train_loss": 0.009755637496709824, "global_step": 262095, "epoch": 2944, "lr": 2.9378554088459463e-07}
{"train_loss": 0.040706001222133636, "global_step": 262096, "epoch": 2944, "lr": 2.9372279085773115e-07}
{"train_loss": 0.041605498641729355, "global_step": 262097, "epoch": 2944, "lr": 2.936600475132778e-07}
{"train_loss": 0.033043272793293, "global_step": 262098, "epoch": 2944, "lr": 2.935973108512291e-07}
{"train_loss": 0.01574019156396389, "global_step": 262099, "epoch": 2944, "lr": 2.935345808716072e-07}
{"train_loss": 0.02999185398221016, "global_step": 262100, "epoch": 2944, "lr": 2.9347185757440666e-07}
{"train_loss": 0.022828299552202225, "global_step": 262101, "epoch": 2944, "lr": 2.9340914095964956e-07}
{"train_loss": 0.015735309571027756, "global_step": 262102, "epoch": 2944, "lr": 2.933464310273304e-07}
{"train_loss": 0.0981491357088089, "global_step": 262103, "epoch": 2944, "lr": 2.932837277774714e-07}
{"train_loss": 0.04566456500984979, "global_step": 262104, "epoch": 2944, "lr": 2.9322103121006693e-07, "val_loss": 9.56602954864502}
{"train_loss": 0.07090774923563004, "global_step": 262105, "epoch": 2945, "lr": 2.931583413251393e-07}
{"train_loss": 0.050901420414447784, "global_step": 262106, "epoch": 2945, "lr": 2.9309565812268846e-07}
{"train_loss": 0.034021586179733276, "global_step": 262107, "epoch": 2945, "lr": 2.930329816027255e-07}
{"train_loss": 0.051835134625434875, "global_step": 262108, "epoch": 2945, "lr": 2.929703117652616e-07}
{"train_loss": 0.020532608032226562, "global_step": 262109, "epoch": 2945, "lr": 2.9290764861029665e-07}
{"train_loss": 0.027919430285692215, "global_step": 262110, "epoch": 2945, "lr": 2.9284499213784733e-07}
{"train_loss": 0.05692538246512413, "global_step": 262111, "epoch": 2945, "lr": 2.9278234234791924e-07}
{"train_loss": 0.045966170728206635, "global_step": 262112, "epoch": 2945, "lr": 2.9271969924052343e-07}
{"train_loss": 0.04699639230966568, "global_step": 262113, "epoch": 2945, "lr": 2.9265706281565997e-07}
{"train_loss": 0.03276614099740982, "global_step": 262114, "epoch": 2945, "lr": 2.925944330733454e-07}
{"train_loss": 0.026562301442027092, "global_step": 262115, "epoch": 2945, "lr": 2.9253181001358544e-07}
{"train_loss": 0.0417957529425621, "global_step": 262116, "epoch": 2945, "lr": 2.92469193636391e-07}
{"train_loss": 0.03678442910313606, "global_step": 262117, "epoch": 2945, "lr": 2.9240658394176226e-07}
{"train_loss": 0.049982670694589615, "global_step": 262118, "epoch": 2945, "lr": 2.923439809297213e-07}
{"train_loss": 0.029932769015431404, "global_step": 262119, "epoch": 2945, "lr": 2.9228138460026256e-07}
{"train_loss": 0.10510008037090302, "global_step": 262120, "epoch": 2945, "lr": 2.922187949534083e-07}
{"train_loss": 0.03670026734471321, "global_step": 262121, "epoch": 2945, "lr": 2.92156211989153e-07}
{"train_loss": 0.01902829483151436, "global_step": 262122, "epoch": 2945, "lr": 2.9209363570751325e-07}
{"train_loss": 0.04503084346652031, "global_step": 262123, "epoch": 2945, "lr": 2.920310661084946e-07}
{"train_loss": 0.05472259223461151, "global_step": 262124, "epoch": 2945, "lr": 2.9196850319210825e-07}
{"train_loss": 0.034796830266714096, "global_step": 262125, "epoch": 2945, "lr": 2.919059469583596e-07}
{"train_loss": 0.05113597214221954, "global_step": 262126, "epoch": 2945, "lr": 2.9184339740725983e-07}
{"train_loss": 0.021604781970381737, "global_step": 262127, "epoch": 2945, "lr": 2.9178085453881456e-07}
{"train_loss": 0.027019530534744263, "global_step": 262128, "epoch": 2945, "lr": 2.917183183530292e-07}
{"train_loss": 0.028793660923838615, "global_step": 262129, "epoch": 2945, "lr": 2.9165578884992607e-07}
{"train_loss": 0.06946481019258499, "global_step": 262130, "epoch": 2945, "lr": 2.9159326602949397e-07}
{"train_loss": 0.053866900503635406, "global_step": 262131, "epoch": 2945, "lr": 2.9153074989176077e-07}
{"train_loss": 0.04675281420350075, "global_step": 262132, "epoch": 2945, "lr": 2.914682404367153e-07}
{"train_loss": 0.0544574148952961, "global_step": 262133, "epoch": 2945, "lr": 2.9140573766438526e-07}
{"train_loss": 0.0349099300801754, "global_step": 262134, "epoch": 2945, "lr": 2.9134324157476524e-07}
{"train_loss": 0.023300206288695335, "global_step": 262135, "epoch": 2945, "lr": 2.9128075216786623e-07}
{"train_loss": 0.048330433666706085, "global_step": 262136, "epoch": 2945, "lr": 2.9121826944370487e-07}
{"train_loss": 0.02493136376142502, "global_step": 262137, "epoch": 2945, "lr": 2.9115579340228126e-07}
{"train_loss": 0.06789349764585495, "global_step": 262138, "epoch": 2945, "lr": 2.910933240436009e-07}
{"train_loss": 0.025286363437771797, "global_step": 262139, "epoch": 2945, "lr": 2.910308613676804e-07}
{"train_loss": 0.004541181959211826, "global_step": 262140, "epoch": 2945, "lr": 2.909684053745254e-07}
{"train_loss": 0.06031044200062752, "global_step": 262141, "epoch": 2945, "lr": 2.9090595606414697e-07}
{"train_loss": 0.042761702090501785, "global_step": 262142, "epoch": 2945, "lr": 2.90843513436545e-07}
{"train_loss": 0.03628639131784439, "global_step": 262143, "epoch": 2945, "lr": 2.9078107749173635e-07}
{"train_loss": 0.050042521208524704, "global_step": 262144, "epoch": 2945, "lr": 2.9071864822972636e-07}
{"train_loss": 0.043248727917671204, "global_step": 262145, "epoch": 2945, "lr": 2.906562256505208e-07}
{"train_loss": 0.046987760812044144, "global_step": 262146, "epoch": 2945, "lr": 2.9059380975413054e-07}
{"train_loss": 0.03585338965058327, "global_step": 262147, "epoch": 2945, "lr": 2.9053140054056684e-07}
{"train_loss": 0.04248125106096268, "global_step": 262148, "epoch": 2945, "lr": 2.9046899800983516e-07}
{"train_loss": 0.03473711386322975, "global_step": 262149, "epoch": 2945, "lr": 2.9040660216194113e-07}
{"train_loss": 0.05738144367933273, "global_step": 262150, "epoch": 2945, "lr": 2.903442129969014e-07}
{"train_loss": 0.05758019536733627, "global_step": 262151, "epoch": 2945, "lr": 2.9028183051471037e-07}
{"train_loss": 0.04116971790790558, "global_step": 262152, "epoch": 2945, "lr": 2.9021945471539026e-07}
{"train_loss": 0.038482148200273514, "global_step": 262153, "epoch": 2945, "lr": 2.901570855989466e-07}
{"train_loss": 0.07236811518669128, "global_step": 262154, "epoch": 2945, "lr": 2.90094723165385e-07}
{"train_loss": 0.050441183149814606, "global_step": 262155, "epoch": 2945, "lr": 2.9003236741471095e-07}
{"train_loss": 0.036347195506095886, "global_step": 262156, "epoch": 2945, "lr": 2.899700183469356e-07}
{"train_loss": 0.03334662690758705, "global_step": 262157, "epoch": 2945, "lr": 2.8990767596207004e-07}
{"train_loss": 0.018357053399086, "global_step": 262158, "epoch": 2945, "lr": 2.8984534026012535e-07}
{"train_loss": 0.04712839797139168, "global_step": 262159, "epoch": 2945, "lr": 2.8978301124109597e-07}
{"train_loss": 0.01774176023900509, "global_step": 262160, "epoch": 2945, "lr": 2.8972068890500413e-07}
{"train_loss": 0.04738280922174454, "global_step": 262161, "epoch": 2945, "lr": 2.896583732518554e-07}
{"train_loss": 0.03580475598573685, "global_step": 262162, "epoch": 2945, "lr": 2.895960642816553e-07}
{"train_loss": 0.03682849928736687, "global_step": 262163, "epoch": 2945, "lr": 2.895337619944094e-07}
{"train_loss": 0.05129501596093178, "global_step": 262164, "epoch": 2945, "lr": 2.894714663901288e-07}
{"train_loss": 0.008225579746067524, "global_step": 262165, "epoch": 2945, "lr": 2.894091774688301e-07}
{"train_loss": 0.040803249925374985, "global_step": 262166, "epoch": 2945, "lr": 2.8934689523050783e-07}
{"train_loss": 0.04030146077275276, "global_step": 262167, "epoch": 2945, "lr": 2.892846196751842e-07}
{"train_loss": 0.031391691416502, "global_step": 262168, "epoch": 2945, "lr": 2.8922235080285355e-07}
{"train_loss": 0.06974925845861435, "global_step": 262169, "epoch": 2945, "lr": 2.891600886135326e-07}
{"train_loss": 0.04031994193792343, "global_step": 262170, "epoch": 2945, "lr": 2.8909783310722695e-07}
{"train_loss": 0.05310102179646492, "global_step": 262171, "epoch": 2945, "lr": 2.8903558428395315e-07}
{"train_loss": 0.04383932426571846, "global_step": 262172, "epoch": 2945, "lr": 2.8897334214370573e-07}
{"train_loss": 0.03613540530204773, "global_step": 262173, "epoch": 2945, "lr": 2.8891110668650136e-07}
{"train_loss": 0.016803063452243805, "global_step": 262174, "epoch": 2945, "lr": 2.888488779123455e-07}
{"train_loss": 0.027089564129710197, "global_step": 262175, "epoch": 2945, "lr": 2.887866558212493e-07}
{"train_loss": 0.07472610473632812, "global_step": 262176, "epoch": 2945, "lr": 2.887244404132239e-07}
{"train_loss": 0.04284774139523506, "global_step": 262177, "epoch": 2945, "lr": 2.886622316882692e-07}
{"train_loss": 0.0973062515258789, "global_step": 262178, "epoch": 2945, "lr": 2.8860002964639644e-07}
{"train_loss": 0.07510265707969666, "global_step": 262179, "epoch": 2945, "lr": 2.885378342876166e-07}
{"train_loss": 0.04055512696504593, "global_step": 262180, "epoch": 2945, "lr": 2.8847564561194084e-07}
{"train_loss": 0.05783447250723839, "global_step": 262181, "epoch": 2945, "lr": 2.8841346361936917e-07}
{"train_loss": 0.013414726592600346, "global_step": 262182, "epoch": 2945, "lr": 2.883512883099182e-07}
{"train_loss": 0.024309903383255005, "global_step": 262183, "epoch": 2945, "lr": 2.8828911968358796e-07}
{"train_loss": 0.027183551341295242, "global_step": 262184, "epoch": 2945, "lr": 2.882269577403951e-07}
{"train_loss": 0.07067637890577316, "global_step": 262185, "epoch": 2945, "lr": 2.8816480248034514e-07}
{"train_loss": 0.03207312151789665, "global_step": 262186, "epoch": 2945, "lr": 2.8810265390344373e-07}
{"train_loss": 0.034962236881256104, "global_step": 262187, "epoch": 2945, "lr": 2.8804051200970184e-07}
{"train_loss": 0.026118649169802666, "global_step": 262188, "epoch": 2945, "lr": 2.879783767991251e-07}
{"train_loss": 0.07362006604671478, "global_step": 262189, "epoch": 2945, "lr": 2.8791624827172456e-07}
{"train_loss": 0.027847252786159515, "global_step": 262190, "epoch": 2945, "lr": 2.8785412642750585e-07}
{"train_loss": 0.03703875467181206, "global_step": 262191, "epoch": 2945, "lr": 2.8779201126648e-07}
{"train_loss": 0.05778445303440094, "global_step": 262192, "epoch": 2945, "lr": 2.877299027886582e-07}
{"train_loss": 0.042716667903691864, "global_step": 262193, "epoch": 2945, "lr": 2.8766780099404033e-07, "val_loss": 9.707290649414062, "train_action_mse_error": 13.05067253112793}
{"train_loss": 0.030015740543603897, "global_step": 262194, "epoch": 2946, "lr": 2.876057058826487e-07}
{"train_loss": 0.08188170939683914, "global_step": 262195, "epoch": 2946, "lr": 2.875436174544721e-07}
{"train_loss": 0.04723481461405754, "global_step": 262196, "epoch": 2946, "lr": 2.8748153570953286e-07}
{"train_loss": 0.0473979152739048, "global_step": 262197, "epoch": 2946, "lr": 2.8741946064783644e-07}
{"train_loss": 0.03892898187041283, "global_step": 262198, "epoch": 2946, "lr": 2.87357392269394e-07}
{"train_loss": 0.05162688344717026, "global_step": 262199, "epoch": 2946, "lr": 2.872953305742054e-07}
{"train_loss": 0.05178691819310188, "global_step": 262200, "epoch": 2946, "lr": 2.872332755622875e-07}
{"train_loss": 0.03242257237434387, "global_step": 262201, "epoch": 2946, "lr": 2.871712272336402e-07}
{"train_loss": 0.04158424213528633, "global_step": 262202, "epoch": 2946, "lr": 2.8710918558828015e-07}
{"train_loss": 0.014972931705415249, "global_step": 262203, "epoch": 2946, "lr": 2.870471506262129e-07}
{"train_loss": 0.058399952948093414, "global_step": 262204, "epoch": 2946, "lr": 2.86985122347444e-07}
{"train_loss": 0.025852512568235397, "global_step": 262205, "epoch": 2946, "lr": 2.869231007519846e-07}
{"train_loss": 0.06591946631669998, "global_step": 262206, "epoch": 2946, "lr": 2.8686108583984575e-07}
{"train_loss": 0.03718191385269165, "global_step": 262207, "epoch": 2946, "lr": 2.8679907761102743e-07}
{"train_loss": 0.06785643100738525, "global_step": 262208, "epoch": 2946, "lr": 2.8673707606554634e-07}
{"train_loss": 0.04730360954999924, "global_step": 262209, "epoch": 2946, "lr": 2.8667508120340805e-07}
{"train_loss": 0.09742709994316101, "global_step": 262210, "epoch": 2946, "lr": 2.8661309302461805e-07}
{"train_loss": 0.05239184573292732, "global_step": 262211, "epoch": 2946, "lr": 2.865511115291875e-07}
{"train_loss": 0.042098384350538254, "global_step": 262212, "epoch": 2946, "lr": 2.8648913671712743e-07}
{"train_loss": 0.07536426186561584, "global_step": 262213, "epoch": 2946, "lr": 2.8642716858843787e-07}
{"train_loss": 0.02097185142338276, "global_step": 262214, "epoch": 2946, "lr": 2.863652071431355e-07}
{"train_loss": 0.03814918175339699, "global_step": 262215, "epoch": 2946, "lr": 2.863032523812259e-07}
{"train_loss": 0.05589386075735092, "global_step": 262216, "epoch": 2946, "lr": 2.8624130430271456e-07}
{"train_loss": 0.027965931221842766, "global_step": 262217, "epoch": 2946, "lr": 2.861793629076126e-07}
{"train_loss": 0.06905059516429901, "global_step": 262218, "epoch": 2946, "lr": 2.861174281959311e-07}
{"train_loss": 0.024348696693778038, "global_step": 262219, "epoch": 2946, "lr": 2.860555001676701e-07}
{"train_loss": 0.03200433403253555, "global_step": 262220, "epoch": 2946, "lr": 2.859935788228463e-07}
{"train_loss": 0.05833936482667923, "global_step": 262221, "epoch": 2946, "lr": 2.859316641614651e-07}
{"train_loss": 0.044595129787921906, "global_step": 262222, "epoch": 2946, "lr": 2.8586975618353215e-07}
{"train_loss": 0.05806154012680054, "global_step": 262223, "epoch": 2946, "lr": 2.858078548890586e-07}
{"train_loss": 0.017313314601778984, "global_step": 262224, "epoch": 2946, "lr": 2.8574596027805544e-07}
{"train_loss": 0.08463034778833389, "global_step": 262225, "epoch": 2946, "lr": 2.856840723505227e-07}
{"train_loss": 0.037243399769067764, "global_step": 262226, "epoch": 2946, "lr": 2.8562219110647717e-07}
{"train_loss": 0.07727483659982681, "global_step": 262227, "epoch": 2946, "lr": 2.855603165459242e-07}
{"train_loss": 0.02219916507601738, "global_step": 262228, "epoch": 2946, "lr": 2.8549844866886946e-07}
{"train_loss": 0.029149239882826805, "global_step": 262229, "epoch": 2946, "lr": 2.85436587475324e-07}
{"train_loss": 0.03397658094763756, "global_step": 262230, "epoch": 2946, "lr": 2.8537473296529895e-07}
{"train_loss": 0.0549144372344017, "global_step": 262231, "epoch": 2946, "lr": 2.853128851387943e-07}
{"train_loss": 0.030010933056473732, "global_step": 262232, "epoch": 2946, "lr": 2.852510439958267e-07}
{"train_loss": 0.05756571516394615, "global_step": 262233, "epoch": 2946, "lr": 2.851892095363962e-07}
{"train_loss": 0.03395257517695427, "global_step": 262234, "epoch": 2946, "lr": 2.851273817605249e-07}
{"train_loss": 0.07115264236927032, "global_step": 262235, "epoch": 2946, "lr": 2.850655606682018e-07}
{"train_loss": 0.02318316139280796, "global_step": 262236, "epoch": 2946, "lr": 2.8500374625945456e-07}
{"train_loss": 0.029723986983299255, "global_step": 262237, "epoch": 2946, "lr": 2.849419385342833e-07}
{"train_loss": 0.04133132845163345, "global_step": 262238, "epoch": 2946, "lr": 2.8488013749268795e-07}
{"train_loss": 0.02040243335068226, "global_step": 262239, "epoch": 2946, "lr": 2.848183431346907e-07}
{"train_loss": 0.044280193746089935, "global_step": 262240, "epoch": 2946, "lr": 2.847565554602916e-07}
{"train_loss": 0.046794723719358444, "global_step": 262241, "epoch": 2946, "lr": 2.846947744695017e-07}
{"train_loss": 0.03963937237858772, "global_step": 262242, "epoch": 2946, "lr": 2.846330001623265e-07}
{"train_loss": 0.037347983568906784, "global_step": 262243, "epoch": 2946, "lr": 2.8457123253877726e-07}
{"train_loss": 0.09911434352397919, "global_step": 262244, "epoch": 2946, "lr": 2.845094715988594e-07}
{"train_loss": 0.019815582782030106, "global_step": 262245, "epoch": 2946, "lr": 2.844477173425897e-07}
{"train_loss": 0.04245473071932793, "global_step": 262246, "epoch": 2946, "lr": 2.8438596976996243e-07}
{"train_loss": 0.03437318280339241, "global_step": 262247, "epoch": 2946, "lr": 2.8432422888099995e-07}
{"train_loss": 0.04149086773395538, "global_step": 262248, "epoch": 2946, "lr": 2.8426249467569663e-07}
{"train_loss": 0.04314235970377922, "global_step": 262249, "epoch": 2946, "lr": 2.8420076715407476e-07}
{"train_loss": 0.034704409539699554, "global_step": 262250, "epoch": 2946, "lr": 2.8413904631613417e-07}
{"train_loss": 0.05142972618341446, "global_step": 262251, "epoch": 2946, "lr": 2.8407733216188615e-07}
{"train_loss": 0.08396999537944794, "global_step": 262252, "epoch": 2946, "lr": 2.8401562469133615e-07}
{"train_loss": 0.03938468173146248, "global_step": 262253, "epoch": 2946, "lr": 2.839539239044953e-07}
{"train_loss": 0.028571883216500282, "global_step": 262254, "epoch": 2946, "lr": 2.838922298013691e-07}
{"train_loss": 0.0359833799302578, "global_step": 262255, "epoch": 2946, "lr": 2.838305423819687e-07}
{"train_loss": 0.04487183690071106, "global_step": 262256, "epoch": 2946, "lr": 2.8376886164630525e-07}
{"train_loss": 0.024043086916208267, "global_step": 262257, "epoch": 2946, "lr": 2.8370718759437864e-07}
{"train_loss": 0.05060094594955444, "global_step": 262258, "epoch": 2946, "lr": 2.8364552022620563e-07}
{"train_loss": 0.0471934899687767, "global_step": 262259, "epoch": 2946, "lr": 2.835838595417861e-07}
{"train_loss": 0.054640062153339386, "global_step": 262260, "epoch": 2946, "lr": 2.8352220554113684e-07}
{"train_loss": 0.030211010947823524, "global_step": 262261, "epoch": 2946, "lr": 2.834605582242578e-07}
{"train_loss": 0.015545598231256008, "global_step": 262262, "epoch": 2946, "lr": 2.833989175911655e-07}
{"train_loss": 0.07384732365608215, "global_step": 262263, "epoch": 2946, "lr": 2.8333728364186017e-07}
{"train_loss": 0.04261796176433563, "global_step": 262264, "epoch": 2946, "lr": 2.8327565637635834e-07}
{"train_loss": 0.033759478479623795, "global_step": 262265, "epoch": 2946, "lr": 2.8321403579466e-07}
{"train_loss": 0.05766088515520096, "global_step": 262266, "epoch": 2946, "lr": 2.8315242189677627e-07}
{"train_loss": 0.02315712906420231, "global_step": 262267, "epoch": 2946, "lr": 2.830908146827238e-07}
{"train_loss": 0.0738983303308487, "global_step": 262268, "epoch": 2946, "lr": 2.830292141524971e-07}
{"train_loss": 0.0362686961889267, "global_step": 262269, "epoch": 2946, "lr": 2.8296762030611267e-07}
{"train_loss": 0.05975645035505295, "global_step": 262270, "epoch": 2946, "lr": 2.8290603314358177e-07}
{"train_loss": 0.02061944454908371, "global_step": 262271, "epoch": 2946, "lr": 2.8284445266490436e-07}
{"train_loss": 0.03427219018340111, "global_step": 262272, "epoch": 2946, "lr": 2.82782878870097e-07}
{"train_loss": 0.06946738064289093, "global_step": 262273, "epoch": 2946, "lr": 2.827213117591543e-07}
{"train_loss": 0.056114356964826584, "global_step": 262274, "epoch": 2946, "lr": 2.826597513321039e-07}
{"train_loss": 0.047775670886039734, "global_step": 262275, "epoch": 2946, "lr": 2.825981975889347e-07}
{"train_loss": 0.04633051156997681, "global_step": 262276, "epoch": 2946, "lr": 2.8253665052966893e-07}
{"train_loss": 0.024094553664326668, "global_step": 262277, "epoch": 2946, "lr": 2.8247511015431217e-07}
{"train_loss": 0.04944629594683647, "global_step": 262278, "epoch": 2946, "lr": 2.8241357646286994e-07}
{"train_loss": 0.046387553215026855, "global_step": 262279, "epoch": 2946, "lr": 2.8235204945534777e-07}
{"train_loss": 0.07402213662862778, "global_step": 262280, "epoch": 2946, "lr": 2.8229052913176236e-07}
{"train_loss": 0.04995843395590782, "global_step": 262281, "epoch": 2946, "lr": 2.822290154921137e-07}
{"train_loss": 0.0455373814224862, "global_step": 262282, "epoch": 2946, "lr": 2.8216750853641284e-07, "val_loss": 9.602531433105469}
{"train_loss": 0.061874862760305405, "global_step": 262283, "epoch": 2947, "lr": 2.821060082646709e-07}
{"train_loss": 0.04814637824892998, "global_step": 262284, "epoch": 2947, "lr": 2.82044514676888e-07}
{"train_loss": 0.03297463059425354, "global_step": 262285, "epoch": 2947, "lr": 2.819830277730862e-07}
{"train_loss": 0.01259341649711132, "global_step": 262286, "epoch": 2947, "lr": 2.819215475532599e-07}
{"train_loss": 0.030191650614142418, "global_step": 262287, "epoch": 2947, "lr": 2.8186007401742597e-07}
{"train_loss": 0.028968892991542816, "global_step": 262288, "epoch": 2947, "lr": 2.8179860716558983e-07}
{"train_loss": 0.09223025292158127, "global_step": 262289, "epoch": 2947, "lr": 2.81737146997757e-07}
{"train_loss": 0.05959491804242134, "global_step": 262290, "epoch": 2947, "lr": 2.816756935139442e-07}
{"train_loss": 0.04499628022313118, "global_step": 262291, "epoch": 2947, "lr": 2.8161424671415136e-07}
{"train_loss": 0.06575795263051987, "global_step": 262292, "epoch": 2947, "lr": 2.815528065983897e-07}
{"train_loss": 0.036392632871866226, "global_step": 262293, "epoch": 2947, "lr": 2.814913731666646e-07}
{"train_loss": 0.01469944417476654, "global_step": 262294, "epoch": 2947, "lr": 2.814299464189929e-07}
{"train_loss": 0.03792744129896164, "global_step": 262295, "epoch": 2947, "lr": 2.813685263553689e-07}
{"train_loss": 0.049088235944509506, "global_step": 262296, "epoch": 2947, "lr": 2.813071129758149e-07}
{"train_loss": 0.02457260712981224, "global_step": 262297, "epoch": 2947, "lr": 2.8124570628033084e-07}
{"train_loss": 0.01928931102156639, "global_step": 262298, "epoch": 2947, "lr": 2.811843062689334e-07}
{"train_loss": 0.047008831053972244, "global_step": 262299, "epoch": 2947, "lr": 2.811229129416171e-07}
{"train_loss": 0.025984087958931923, "global_step": 262300, "epoch": 2947, "lr": 2.810615262983984e-07}
{"train_loss": 0.03394316881895065, "global_step": 262301, "epoch": 2947, "lr": 2.810001463392886e-07}
{"train_loss": 0.01671098731458187, "global_step": 262302, "epoch": 2947, "lr": 2.8093877306429316e-07}
{"train_loss": 0.028319114819169044, "global_step": 262303, "epoch": 2947, "lr": 2.808774064734121e-07}
{"train_loss": 0.056217994540929794, "global_step": 262304, "epoch": 2947, "lr": 2.808160465666676e-07}
{"train_loss": 0.02661873586475849, "global_step": 262305, "epoch": 2947, "lr": 2.807546933440597e-07}
{"train_loss": 0.03540806099772453, "global_step": 262306, "epoch": 2947, "lr": 2.8069334680559945e-07}
{"train_loss": 0.03958970680832863, "global_step": 262307, "epoch": 2947, "lr": 2.8063200695129245e-07}
{"train_loss": 0.031240440905094147, "global_step": 262308, "epoch": 2947, "lr": 2.805706737811442e-07}
{"train_loss": 0.009019256569445133, "global_step": 262309, "epoch": 2947, "lr": 2.80509347295177e-07}
{"train_loss": 0.022785460576415062, "global_step": 262310, "epoch": 2947, "lr": 2.804480274933852e-07}
{"train_loss": 0.051483798772096634, "global_step": 262311, "epoch": 2947, "lr": 2.803867143757799e-07}
{"train_loss": 0.052404534071683884, "global_step": 262312, "epoch": 2947, "lr": 2.8032540794237226e-07}
{"train_loss": 0.024847563356161118, "global_step": 262313, "epoch": 2947, "lr": 2.802641081931623e-07}
{"train_loss": 0.020274568349123, "global_step": 262314, "epoch": 2947, "lr": 2.8020281512817215e-07}
{"train_loss": 0.04474573954939842, "global_step": 262315, "epoch": 2947, "lr": 2.8014152874740186e-07}
{"train_loss": 0.07810003310441971, "global_step": 262316, "epoch": 2947, "lr": 2.800802490508569e-07}
{"train_loss": 0.05617442727088928, "global_step": 262317, "epoch": 2947, "lr": 2.800189760385541e-07}
{"train_loss": 0.042565204203128815, "global_step": 262318, "epoch": 2947, "lr": 2.799577097104933e-07}
{"train_loss": 0.0836726501584053, "global_step": 262319, "epoch": 2947, "lr": 2.798964500666912e-07}
{"train_loss": 0.04309757053852081, "global_step": 262320, "epoch": 2947, "lr": 2.7983519710714225e-07}
{"train_loss": 0.048410575836896896, "global_step": 262321, "epoch": 2947, "lr": 2.7977395083186863e-07}
{"train_loss": 0.03829849511384964, "global_step": 262322, "epoch": 2947, "lr": 2.7971271124087594e-07}
{"train_loss": 0.0316450335085392, "global_step": 262323, "epoch": 2947, "lr": 2.7965147833416414e-07}
{"train_loss": 0.02488282136619091, "global_step": 262324, "epoch": 2947, "lr": 2.7959025211174995e-07}
{"train_loss": 0.03553140535950661, "global_step": 262325, "epoch": 2947, "lr": 2.795290325736388e-07}
{"train_loss": 0.050525832921266556, "global_step": 262326, "epoch": 2947, "lr": 2.7946781971984193e-07}
{"train_loss": 0.06188249960541725, "global_step": 262327, "epoch": 2947, "lr": 2.7940661355035927e-07}
{"train_loss": 0.045451872050762177, "global_step": 262328, "epoch": 2947, "lr": 2.793454140652074e-07}
{"train_loss": 0.022106721997261047, "global_step": 262329, "epoch": 2947, "lr": 2.7928422126438647e-07}
{"train_loss": 0.0389723964035511, "global_step": 262330, "epoch": 2947, "lr": 2.792230351479186e-07}
{"train_loss": 0.0333702526986599, "global_step": 262331, "epoch": 2947, "lr": 2.7916185571579824e-07}
{"train_loss": 0.03758557140827179, "global_step": 262332, "epoch": 2947, "lr": 2.7910068296803646e-07}
{"train_loss": 0.031057870015501976, "global_step": 262333, "epoch": 2947, "lr": 2.7903951690465e-07}
{"train_loss": 0.03704996034502983, "global_step": 262334, "epoch": 2947, "lr": 2.789783575256333e-07}
{"train_loss": 0.0449935607612133, "global_step": 262335, "epoch": 2947, "lr": 2.789172048310085e-07}
{"train_loss": 0.046055279672145844, "global_step": 262336, "epoch": 2947, "lr": 2.788560588207756e-07}
{"train_loss": 0.021973736584186554, "global_step": 262337, "epoch": 2947, "lr": 2.7879491949494017e-07}
{"train_loss": 0.09433114528656006, "global_step": 262338, "epoch": 2947, "lr": 2.787337868535189e-07}
{"train_loss": 0.017176872119307518, "global_step": 262339, "epoch": 2947, "lr": 2.786726608965118e-07}
{"train_loss": 0.06540092080831528, "global_step": 262340, "epoch": 2947, "lr": 2.7861154162393544e-07}
{"train_loss": 0.012829786166548729, "global_step": 262341, "epoch": 2947, "lr": 2.785504290357899e-07}
{"train_loss": 0.045423753559589386, "global_step": 262342, "epoch": 2947, "lr": 2.7848932313209175e-07}
{"train_loss": 0.0377623587846756, "global_step": 262343, "epoch": 2947, "lr": 2.7842822391284105e-07}
{"train_loss": 0.04776840656995773, "global_step": 262344, "epoch": 2947, "lr": 2.7836713137805446e-07}
{"train_loss": 0.02196578122675419, "global_step": 262345, "epoch": 2947, "lr": 2.78306045527732e-07}
{"train_loss": 0.050450216978788376, "global_step": 262346, "epoch": 2947, "lr": 2.7824496636188466e-07}
{"train_loss": 0.04432213678956032, "global_step": 262347, "epoch": 2947, "lr": 2.781838938805237e-07}
{"train_loss": 0.03198540583252907, "global_step": 262348, "epoch": 2947, "lr": 2.7812282808365456e-07}
{"train_loss": 0.0329478457570076, "global_step": 262349, "epoch": 2947, "lr": 2.780617689712828e-07}
{"train_loss": 0.061067622154951096, "global_step": 262350, "epoch": 2947, "lr": 2.780007165434195e-07}
{"train_loss": 0.04994091019034386, "global_step": 262351, "epoch": 2947, "lr": 2.7793967080007586e-07}
{"train_loss": 0.02920866385102272, "global_step": 262352, "epoch": 2947, "lr": 2.778786317412574e-07}
{"train_loss": 0.05322294309735298, "global_step": 262353, "epoch": 2947, "lr": 2.778175993669696e-07}
{"train_loss": 0.048611532896757126, "global_step": 262354, "epoch": 2947, "lr": 2.777565736772236e-07}
{"train_loss": 0.03878563642501831, "global_step": 262355, "epoch": 2947, "lr": 2.7769555467203056e-07}
{"train_loss": 0.03878571465611458, "global_step": 262356, "epoch": 2947, "lr": 2.776345423513904e-07}
{"train_loss": 0.03325739502906799, "global_step": 262357, "epoch": 2947, "lr": 2.7757353671531983e-07}
{"train_loss": 0.04538923129439354, "global_step": 262358, "epoch": 2947, "lr": 2.775125377638188e-07}
{"train_loss": 0.04579037427902222, "global_step": 262359, "epoch": 2947, "lr": 2.7745154549690953e-07}
{"train_loss": 0.026565631851553917, "global_step": 262360, "epoch": 2947, "lr": 2.7739055991458095e-07}
{"train_loss": 0.02189340628683567, "global_step": 262361, "epoch": 2947, "lr": 2.7732958101685524e-07}
{"train_loss": 0.03224559128284454, "global_step": 262362, "epoch": 2947, "lr": 2.7726860880373794e-07}
{"train_loss": 0.08277163654565811, "global_step": 262363, "epoch": 2947, "lr": 2.772076432752346e-07}
{"train_loss": 0.032958514988422394, "global_step": 262364, "epoch": 2947, "lr": 2.7714668443135085e-07}
{"train_loss": 0.03631475567817688, "global_step": 262365, "epoch": 2947, "lr": 2.770857322721032e-07}
{"train_loss": 0.049027323722839355, "global_step": 262366, "epoch": 2947, "lr": 2.770247867974973e-07}
{"train_loss": 0.014096655882894993, "global_step": 262367, "epoch": 2947, "lr": 2.769638480075332e-07}
{"train_loss": 0.021381476894021034, "global_step": 262368, "epoch": 2947, "lr": 2.769029159022274e-07}
{"train_loss": 0.057851642370224, "global_step": 262369, "epoch": 2947, "lr": 2.768419904815911e-07}
{"train_loss": 0.03046901524066925, "global_step": 262370, "epoch": 2947, "lr": 2.767810717456243e-07}
{"train_loss": 0.04004912149537815, "global_step": 262371, "epoch": 2947, "lr": 2.767201596943325e-07, "val_loss": 9.527570724487305}
{"train_loss": 0.03629319742321968, "global_step": 262372, "epoch": 2948, "lr": 2.76659254327738e-07}
{"train_loss": 0.019454337656497955, "global_step": 262373, "epoch": 2948, "lr": 2.7659835564583514e-07}
{"train_loss": 0.05527573823928833, "global_step": 262374, "epoch": 2948, "lr": 2.7653746364864065e-07}
{"train_loss": 0.020765012130141258, "global_step": 262375, "epoch": 2948, "lr": 2.764765783361545e-07}
{"train_loss": 0.04099638760089874, "global_step": 262376, "epoch": 2948, "lr": 2.7641569970839885e-07}
{"train_loss": 0.05476924031972885, "global_step": 262377, "epoch": 2948, "lr": 2.763548277653683e-07}
{"train_loss": 0.013536449521780014, "global_step": 262378, "epoch": 2948, "lr": 2.7629396250707375e-07}
{"train_loss": 0.037658873945474625, "global_step": 262379, "epoch": 2948, "lr": 2.762331039335264e-07}
{"train_loss": 0.04467711225152016, "global_step": 262380, "epoch": 2948, "lr": 2.7617225204473184e-07}
{"train_loss": 0.0486530140042305, "global_step": 262381, "epoch": 2948, "lr": 2.761114068407067e-07}
{"train_loss": 0.029385047033429146, "global_step": 262382, "epoch": 2948, "lr": 2.760505683214454e-07}
{"train_loss": 0.061204493045806885, "global_step": 262383, "epoch": 2948, "lr": 2.759897364869701e-07}
{"train_loss": 0.06070404499769211, "global_step": 262384, "epoch": 2948, "lr": 2.7592891133727537e-07}
{"train_loss": 0.01485347654670477, "global_step": 262385, "epoch": 2948, "lr": 2.758680928723778e-07}
{"train_loss": 0.03335262089967728, "global_step": 262386, "epoch": 2948, "lr": 2.7580728109228293e-07}
{"train_loss": 0.036364223808050156, "global_step": 262387, "epoch": 2948, "lr": 2.7574647599700187e-07}
{"train_loss": 0.03866587206721306, "global_step": 262388, "epoch": 2948, "lr": 2.756856775865402e-07}
{"train_loss": 0.0412178561091423, "global_step": 262389, "epoch": 2948, "lr": 2.756248858609034e-07}
{"train_loss": 0.013937688432633877, "global_step": 262390, "epoch": 2948, "lr": 2.755641008201082e-07}
{"train_loss": 0.049551233649253845, "global_step": 262391, "epoch": 2948, "lr": 2.7550332246415455e-07}
{"train_loss": 0.017944853752851486, "global_step": 262392, "epoch": 2948, "lr": 2.7544255079305357e-07}
{"train_loss": 0.03275363892316818, "global_step": 262393, "epoch": 2948, "lr": 2.753817858068164e-07}
{"train_loss": 0.11641392856836319, "global_step": 262394, "epoch": 2948, "lr": 2.7532102750544296e-07}
{"train_loss": 0.028132552281022072, "global_step": 262395, "epoch": 2948, "lr": 2.7526027588895e-07}
{"train_loss": 0.07476364821195602, "global_step": 262396, "epoch": 2948, "lr": 2.751995309573374e-07}
{"train_loss": 0.04442012310028076, "global_step": 262397, "epoch": 2948, "lr": 2.751387927106219e-07}
{"train_loss": 0.015131314285099506, "global_step": 262398, "epoch": 2948, "lr": 2.750780611488091e-07}
{"train_loss": 0.004019925370812416, "global_step": 262399, "epoch": 2948, "lr": 2.750173362719044e-07}
{"train_loss": 0.037141572684049606, "global_step": 262400, "epoch": 2948, "lr": 2.7495661807991346e-07}
{"train_loss": 0.023226862773299217, "global_step": 262401, "epoch": 2948, "lr": 2.748959065728529e-07}
{"train_loss": 0.028189020231366158, "global_step": 262402, "epoch": 2948, "lr": 2.748352017507283e-07}
{"train_loss": 0.050412654876708984, "global_step": 262403, "epoch": 2948, "lr": 2.747745036135452e-07}
{"train_loss": 0.01738475263118744, "global_step": 262404, "epoch": 2948, "lr": 2.747138121613091e-07}
{"train_loss": 0.046490687876939774, "global_step": 262405, "epoch": 2948, "lr": 2.7465312739403115e-07}
{"train_loss": 0.04153959080576897, "global_step": 262406, "epoch": 2948, "lr": 2.74592449311728e-07}
{"train_loss": 0.04435247555375099, "global_step": 262407, "epoch": 2948, "lr": 2.745317779143941e-07}
{"train_loss": 0.021327396854758263, "global_step": 262408, "epoch": 2948, "lr": 2.7447111320204055e-07}
{"train_loss": 0.016330016776919365, "global_step": 262409, "epoch": 2948, "lr": 2.74410455174684e-07}
{"train_loss": 0.044191084802150726, "global_step": 262410, "epoch": 2948, "lr": 2.743498038323244e-07}
{"train_loss": 0.10205921530723572, "global_step": 262411, "epoch": 2948, "lr": 2.74289159174973e-07}
{"train_loss": 0.060174401849508286, "global_step": 262412, "epoch": 2948, "lr": 2.742285212026352e-07}
{"train_loss": 0.04156757891178131, "global_step": 262413, "epoch": 2948, "lr": 2.7416788991532216e-07}
{"train_loss": 0.029828276485204697, "global_step": 262414, "epoch": 2948, "lr": 2.7410726531303944e-07}
{"train_loss": 0.06477177888154984, "global_step": 262415, "epoch": 2948, "lr": 2.740466473958037e-07}
{"train_loss": 0.03109065629541874, "global_step": 262416, "epoch": 2948, "lr": 2.739860361636093e-07}
{"train_loss": 0.05629316717386246, "global_step": 262417, "epoch": 2948, "lr": 2.73925431616473e-07}
{"train_loss": 0.04338527098298073, "global_step": 262418, "epoch": 2948, "lr": 2.7386483375440587e-07}
{"train_loss": 0.037552360445261, "global_step": 262419, "epoch": 2948, "lr": 2.7380424257740787e-07}
{"train_loss": 0.06319314986467361, "global_step": 262420, "epoch": 2948, "lr": 2.737436580854902e-07}
{"train_loss": 0.030952628701925278, "global_step": 262421, "epoch": 2948, "lr": 2.7368308027866384e-07}
{"train_loss": 0.023120058700442314, "global_step": 262422, "epoch": 2948, "lr": 2.736225091569289e-07}
{"train_loss": 0.03300686180591583, "global_step": 262423, "epoch": 2948, "lr": 2.735619447203075e-07}
{"train_loss": 0.06569918990135193, "global_step": 262424, "epoch": 2948, "lr": 2.7350138696879415e-07}
{"train_loss": 0.045515548437833786, "global_step": 262425, "epoch": 2948, "lr": 2.7344083590240545e-07}
{"train_loss": 0.06628110259771347, "global_step": 262426, "epoch": 2948, "lr": 2.7338029152114695e-07}
{"train_loss": 0.059207573533058167, "global_step": 262427, "epoch": 2948, "lr": 2.733197538250243e-07}
{"train_loss": 0.022308802232146263, "global_step": 262428, "epoch": 2948, "lr": 2.7325922281404846e-07}
{"train_loss": 0.009904267266392708, "global_step": 262429, "epoch": 2948, "lr": 2.731986984882251e-07}
{"train_loss": 0.02081139013171196, "global_step": 262430, "epoch": 2948, "lr": 2.7313818084756524e-07}
{"train_loss": 0.030534788966178894, "global_step": 262431, "epoch": 2948, "lr": 2.7307766989208003e-07}
{"train_loss": 0.05705006793141365, "global_step": 262432, "epoch": 2948, "lr": 2.730171656217695e-07}
{"train_loss": 0.03847839683294296, "global_step": 262433, "epoch": 2948, "lr": 2.7295666803664467e-07}
{"train_loss": 0.04682573303580284, "global_step": 262434, "epoch": 2948, "lr": 2.728961771367167e-07}
{"train_loss": 0.05110716074705124, "global_step": 262435, "epoch": 2948, "lr": 2.728356929219855e-07}
{"train_loss": 0.027465369552373886, "global_step": 262436, "epoch": 2948, "lr": 2.727752153924734e-07}
{"train_loss": 0.06738854199647903, "global_step": 262437, "epoch": 2948, "lr": 2.7271474454818036e-07}
{"train_loss": 0.03847697004675865, "global_step": 262438, "epoch": 2948, "lr": 2.726542803891119e-07}
{"train_loss": 0.05280717834830284, "global_step": 262439, "epoch": 2948, "lr": 2.725938229152791e-07}
{"train_loss": 0.01615665853023529, "global_step": 262440, "epoch": 2948, "lr": 2.725333721266876e-07}
{"train_loss": 0.024951621890068054, "global_step": 262441, "epoch": 2948, "lr": 2.72472928023354e-07}
{"train_loss": 0.032231900840997696, "global_step": 262442, "epoch": 2948, "lr": 2.724124906052783e-07}
{"train_loss": 0.05212897062301636, "global_step": 262443, "epoch": 2948, "lr": 2.7235205987246604e-07}
{"train_loss": 0.03066365048289299, "global_step": 262444, "epoch": 2948, "lr": 2.7229163582493387e-07}
{"train_loss": 0.0973723754286766, "global_step": 262445, "epoch": 2948, "lr": 2.722312184626874e-07}
{"train_loss": 0.03438352420926094, "global_step": 262446, "epoch": 2948, "lr": 2.721708077857321e-07}
{"train_loss": 0.03575504198670387, "global_step": 262447, "epoch": 2948, "lr": 2.721104037940736e-07}
{"train_loss": 0.040137771517038345, "global_step": 262448, "epoch": 2948, "lr": 2.720500064877285e-07}
{"train_loss": 0.041896991431713104, "global_step": 262449, "epoch": 2948, "lr": 2.719896158666968e-07}
{"train_loss": 0.05289687588810921, "global_step": 262450, "epoch": 2948, "lr": 2.719292319309896e-07}
{"train_loss": 0.07652714848518372, "global_step": 262451, "epoch": 2948, "lr": 2.7186885468061806e-07}
{"train_loss": 0.03875155746936798, "global_step": 262452, "epoch": 2948, "lr": 2.7180848411558765e-07}
{"train_loss": 0.05459548160433769, "global_step": 262453, "epoch": 2948, "lr": 2.71748120235904e-07}
{"train_loss": 0.024869220331311226, "global_step": 262454, "epoch": 2948, "lr": 2.7168776304157815e-07}
{"train_loss": 0.06747513264417648, "global_step": 262455, "epoch": 2948, "lr": 2.7162741253262124e-07}
{"train_loss": 0.00924815982580185, "global_step": 262456, "epoch": 2948, "lr": 2.715670687090333e-07}
{"train_loss": 0.07754795253276825, "global_step": 262457, "epoch": 2948, "lr": 2.7150673157083083e-07}
{"train_loss": 0.029800672084093094, "global_step": 262458, "epoch": 2948, "lr": 2.7144640111801957e-07}
{"train_loss": 0.04803640395402908, "global_step": 262459, "epoch": 2948, "lr": 2.7138607735060494e-07}
{"train_loss": 0.04158386517985818, "global_step": 262460, "epoch": 2948, "lr": 2.7132576026859257e-07, "val_loss": 9.627302169799805}
{"train_loss": 0.049278371036052704, "global_step": 262461, "epoch": 2949, "lr": 2.712654498719991e-07}
{"train_loss": 0.08178852498531342, "global_step": 262462, "epoch": 2949, "lr": 2.712051461608245e-07}
{"train_loss": 0.06966923177242279, "global_step": 262463, "epoch": 2949, "lr": 2.711448491350799e-07}
{"train_loss": 0.03600875288248062, "global_step": 262464, "epoch": 2949, "lr": 2.7108455879477634e-07}
{"train_loss": 0.02658355049788952, "global_step": 262465, "epoch": 2949, "lr": 2.7102427513991946e-07}
{"train_loss": 0.04947645962238312, "global_step": 262466, "epoch": 2949, "lr": 2.7096399817051476e-07}
{"train_loss": 0.04864773154258728, "global_step": 262467, "epoch": 2949, "lr": 2.709037278865734e-07}
{"train_loss": 0.04222029820084572, "global_step": 262468, "epoch": 2949, "lr": 2.7084346428810637e-07}
{"train_loss": 0.06068357080221176, "global_step": 262469, "epoch": 2949, "lr": 2.7078320737511376e-07}
{"train_loss": 0.05345124751329422, "global_step": 262470, "epoch": 2949, "lr": 2.707229571476122e-07}
{"train_loss": 0.03351929411292076, "global_step": 262471, "epoch": 2949, "lr": 2.706627136056017e-07}
{"train_loss": 0.04650575667619705, "global_step": 262472, "epoch": 2949, "lr": 2.706024767490989e-07}
{"train_loss": 0.08300496637821198, "global_step": 262473, "epoch": 2949, "lr": 2.705422465781038e-07}
{"train_loss": 0.027425838634371758, "global_step": 262474, "epoch": 2949, "lr": 2.704820230926275e-07}
{"train_loss": 0.08904372900724411, "global_step": 262475, "epoch": 2949, "lr": 2.704218062926811e-07}
{"train_loss": 0.052117303013801575, "global_step": 262476, "epoch": 2949, "lr": 2.7036159617827017e-07}
{"train_loss": 0.04102737829089165, "global_step": 262477, "epoch": 2949, "lr": 2.703013927494003e-07}
{"train_loss": 0.03752906247973442, "global_step": 262478, "epoch": 2949, "lr": 2.70241196006088e-07}
{"train_loss": 0.06734519451856613, "global_step": 262479, "epoch": 2949, "lr": 2.701810059483334e-07}
{"train_loss": 0.040882308036088943, "global_step": 262480, "epoch": 2949, "lr": 2.701208225761476e-07}
{"train_loss": 0.026163557544350624, "global_step": 262481, "epoch": 2949, "lr": 2.7006064588953053e-07}
{"train_loss": 0.02030663751065731, "global_step": 262482, "epoch": 2949, "lr": 2.700004758885044e-07}
{"train_loss": 0.023360351100564003, "global_step": 262483, "epoch": 2949, "lr": 2.699403125730693e-07}
{"train_loss": 0.05040383338928223, "global_step": 262484, "epoch": 2949, "lr": 2.6988015594323626e-07}
{"train_loss": 0.006341897416859865, "global_step": 262485, "epoch": 2949, "lr": 2.698200059990108e-07}
{"train_loss": 0.03426029905676842, "global_step": 262486, "epoch": 2949, "lr": 2.697598627404041e-07}
{"train_loss": 0.07067656517028809, "global_step": 262487, "epoch": 2949, "lr": 2.696997261674161e-07}
{"train_loss": 0.029729316011071205, "global_step": 262488, "epoch": 2949, "lr": 2.6963959628006354e-07}
{"train_loss": 0.010977612808346748, "global_step": 262489, "epoch": 2949, "lr": 2.695794730783519e-07}
{"train_loss": 0.04785379394888878, "global_step": 262490, "epoch": 2949, "lr": 2.6951935656228666e-07}
{"train_loss": 0.043713755905628204, "global_step": 262491, "epoch": 2949, "lr": 2.694592467318846e-07}
{"train_loss": 0.01501479372382164, "global_step": 262492, "epoch": 2949, "lr": 2.693991435871457e-07}
{"train_loss": 0.05641893669962883, "global_step": 262493, "epoch": 2949, "lr": 2.6933904712807544e-07}
{"train_loss": 0.00813677441328764, "global_step": 262494, "epoch": 2949, "lr": 2.6927895735469056e-07}
{"train_loss": 0.048098862171173096, "global_step": 262495, "epoch": 2949, "lr": 2.6921887426699655e-07}
{"train_loss": 0.0481114536523819, "global_step": 262496, "epoch": 2949, "lr": 2.691587978649934e-07}
{"train_loss": 0.035376667976379395, "global_step": 262497, "epoch": 2949, "lr": 2.690987281487034e-07}
{"train_loss": 0.04313535615801811, "global_step": 262498, "epoch": 2949, "lr": 2.690386651181209e-07}
{"train_loss": 0.028972569853067398, "global_step": 262499, "epoch": 2949, "lr": 2.689786087732626e-07}
{"train_loss": 0.08061540126800537, "global_step": 262500, "epoch": 2949, "lr": 2.689185591141341e-07}
{"train_loss": 0.06012988090515137, "global_step": 262501, "epoch": 2949, "lr": 2.6885851614074644e-07}
{"train_loss": 0.030901800841093063, "global_step": 262502, "epoch": 2949, "lr": 2.6879847985309957e-07}
{"train_loss": 0.022308049723505974, "global_step": 262503, "epoch": 2949, "lr": 2.687384502512102e-07}
{"train_loss": 0.05631856247782707, "global_step": 262504, "epoch": 2949, "lr": 2.686784273350784e-07}
{"train_loss": 0.028899189084768295, "global_step": 262505, "epoch": 2949, "lr": 2.6861841110472073e-07}
{"train_loss": 0.0383216068148613, "global_step": 262506, "epoch": 2949, "lr": 2.685584015601372e-07}
{"train_loss": 0.01770337112247944, "global_step": 262507, "epoch": 2949, "lr": 2.684983987013445e-07}
{"train_loss": 0.07454150915145874, "global_step": 262508, "epoch": 2949, "lr": 2.6843840252834264e-07}
{"train_loss": 0.07148279249668121, "global_step": 262509, "epoch": 2949, "lr": 2.6837841304114265e-07}
{"train_loss": 0.11274566501379013, "global_step": 262510, "epoch": 2949, "lr": 2.6831843023975567e-07}
{"train_loss": 0.03688132390379906, "global_step": 262511, "epoch": 2949, "lr": 2.6825845412418726e-07}
{"train_loss": 0.07141201198101044, "global_step": 262512, "epoch": 2949, "lr": 2.6819848469444297e-07}
{"train_loss": 0.022437969222664833, "global_step": 262513, "epoch": 2949, "lr": 2.681385219505339e-07}
{"train_loss": 0.018150456249713898, "global_step": 262514, "epoch": 2949, "lr": 2.6807856589247117e-07}
{"train_loss": 0.024617750197649002, "global_step": 262515, "epoch": 2949, "lr": 2.680186165202547e-07}
{"train_loss": 0.0323033407330513, "global_step": 262516, "epoch": 2949, "lr": 2.679586738338957e-07}
{"train_loss": 0.05456895753741264, "global_step": 262517, "epoch": 2949, "lr": 2.6789873783340523e-07}
{"train_loss": 0.053453512489795685, "global_step": 262518, "epoch": 2949, "lr": 2.678388085187944e-07}
{"train_loss": 0.018641412258148193, "global_step": 262519, "epoch": 2949, "lr": 2.677788858900632e-07}
{"train_loss": 0.04965422675013542, "global_step": 262520, "epoch": 2949, "lr": 2.677189699472227e-07}
{"train_loss": 0.06558682024478912, "global_step": 262521, "epoch": 2949, "lr": 2.6765906069027847e-07}
{"train_loss": 0.039465490728616714, "global_step": 262522, "epoch": 2949, "lr": 2.675991581192472e-07}
{"train_loss": 0.03989136219024658, "global_step": 262523, "epoch": 2949, "lr": 2.675392622341233e-07}
{"train_loss": 0.01337207667529583, "global_step": 262524, "epoch": 2949, "lr": 2.6747937303492345e-07}
{"train_loss": 0.01679486408829689, "global_step": 262525, "epoch": 2949, "lr": 2.6741949052166424e-07}
{"train_loss": 0.055833227932453156, "global_step": 262526, "epoch": 2949, "lr": 2.673596146943347e-07}
{"train_loss": 0.05136698856949806, "global_step": 262527, "epoch": 2949, "lr": 2.672997455529569e-07}
{"train_loss": 0.02323297969996929, "global_step": 262528, "epoch": 2949, "lr": 2.672398830975309e-07}
{"train_loss": 0.0281705092638731, "global_step": 262529, "epoch": 2949, "lr": 2.6718002732807336e-07}
{"train_loss": 0.061972398310899734, "global_step": 262530, "epoch": 2949, "lr": 2.6712017824458424e-07}
{"train_loss": 0.059142109006643295, "global_step": 262531, "epoch": 2949, "lr": 2.6706033584708023e-07}
{"train_loss": 0.026776526123285294, "global_step": 262532, "epoch": 2949, "lr": 2.670005001355558e-07}
{"train_loss": 0.02610514685511589, "global_step": 262533, "epoch": 2949, "lr": 2.6694067111003305e-07}
{"train_loss": 0.02916809171438217, "global_step": 262534, "epoch": 2949, "lr": 2.6688084877050656e-07}
{"train_loss": 0.04663871228694916, "global_step": 262535, "epoch": 2949, "lr": 2.668210331169985e-07}
{"train_loss": 0.03193727508187294, "global_step": 262536, "epoch": 2949, "lr": 2.6676122414950876e-07}
{"train_loss": 0.02621414326131344, "global_step": 262537, "epoch": 2949, "lr": 2.6670142186804857e-07}
{"train_loss": 0.02290855534374714, "global_step": 262538, "epoch": 2949, "lr": 2.666416262726179e-07}
{"train_loss": 0.07422091066837311, "global_step": 262539, "epoch": 2949, "lr": 2.6658183736323895e-07}
{"train_loss": 0.030217545107007027, "global_step": 262540, "epoch": 2949, "lr": 2.6652205513990615e-07}
{"train_loss": 0.09260367602109909, "global_step": 262541, "epoch": 2949, "lr": 2.664622796026361e-07}
{"train_loss": 0.022222142666578293, "global_step": 262542, "epoch": 2949, "lr": 2.6640251075143454e-07}
{"train_loss": 0.03813905268907547, "global_step": 262543, "epoch": 2949, "lr": 2.663427485863068e-07}
{"train_loss": 0.08351756632328033, "global_step": 262544, "epoch": 2949, "lr": 2.662829931072641e-07}
{"train_loss": 0.050829190760850906, "global_step": 262545, "epoch": 2949, "lr": 2.66223244314312e-07}
{"train_loss": 0.02611795999109745, "global_step": 262546, "epoch": 2949, "lr": 2.6616350220746155e-07}
{"train_loss": 0.040152598172426224, "global_step": 262547, "epoch": 2949, "lr": 2.661037667867183e-07}
{"train_loss": 0.05327538028359413, "global_step": 262548, "epoch": 2949, "lr": 2.6604403805209343e-07}
{"train_loss": 0.043429025412233695, "global_step": 262549, "epoch": 2949, "lr": 2.6598431600358686e-07, "val_loss": 9.68445110321045}
{"train_loss": 0.02159077674150467, "global_step": 262550, "epoch": 2950, "lr": 2.6592460064122083e-07}
{"train_loss": 0.08012530207633972, "global_step": 262551, "epoch": 2950, "lr": 2.658648919649898e-07}
{"train_loss": 0.0354667603969574, "global_step": 262552, "epoch": 2950, "lr": 2.6580518997491035e-07}
{"train_loss": 0.06545596569776535, "global_step": 262553, "epoch": 2950, "lr": 2.657454946709825e-07}
{"train_loss": 0.03400026634335518, "global_step": 262554, "epoch": 2950, "lr": 2.65685806053223e-07}
{"train_loss": 0.05755338445305824, "global_step": 262555, "epoch": 2950, "lr": 2.656261241216318e-07}
{"train_loss": 0.10209614038467407, "global_step": 262556, "epoch": 2950, "lr": 2.655664488762255e-07}
{"train_loss": 0.01645669899880886, "global_step": 262557, "epoch": 2950, "lr": 2.6550678031700416e-07}
{"train_loss": 0.04443102329969406, "global_step": 262558, "epoch": 2950, "lr": 2.654471184439844e-07}
{"train_loss": 0.040053825825452805, "global_step": 262559, "epoch": 2950, "lr": 2.6538746325716066e-07}
{"train_loss": 0.07375895231962204, "global_step": 262560, "epoch": 2950, "lr": 2.6532781475655524e-07}
{"train_loss": 0.020787378773093224, "global_step": 262561, "epoch": 2950, "lr": 2.652681729421735e-07}
{"train_loss": 0.040220946073532104, "global_step": 262562, "epoch": 2950, "lr": 2.652085378140101e-07}
{"train_loss": 0.03068680688738823, "global_step": 262563, "epoch": 2950, "lr": 2.651489093720927e-07}
{"train_loss": 0.01988433301448822, "global_step": 262564, "epoch": 2950, "lr": 2.6508928761641574e-07}
{"train_loss": 0.06674100458621979, "global_step": 262565, "epoch": 2950, "lr": 2.6502967254699584e-07}
{"train_loss": 0.06349427998065948, "global_step": 262566, "epoch": 2950, "lr": 2.6497006416382754e-07}
{"train_loss": 0.07589222490787506, "global_step": 262567, "epoch": 2950, "lr": 2.649104624669385e-07}
{"train_loss": 0.027155639603734016, "global_step": 262568, "epoch": 2950, "lr": 2.6485086745631773e-07}
{"train_loss": 0.050065889954566956, "global_step": 262569, "epoch": 2950, "lr": 2.6479127913198733e-07}
{"train_loss": 0.02622677944600582, "global_step": 262570, "epoch": 2950, "lr": 2.647316974939418e-07}
{"train_loss": 0.053866833448410034, "global_step": 262571, "epoch": 2950, "lr": 2.646721225422033e-07}
{"train_loss": 0.036630529910326004, "global_step": 262572, "epoch": 2950, "lr": 2.646125542767719e-07}
{"train_loss": 0.03025202266871929, "global_step": 262573, "epoch": 2950, "lr": 2.6455299269765866e-07}
{"train_loss": 0.04009583219885826, "global_step": 262574, "epoch": 2950, "lr": 2.644934378048691e-07}
{"train_loss": 0.023517277091741562, "global_step": 262575, "epoch": 2950, "lr": 2.644338895984089e-07}
{"train_loss": 0.0536792054772377, "global_step": 262576, "epoch": 2950, "lr": 2.643743480782945e-07}
{"train_loss": 0.06596346199512482, "global_step": 262577, "epoch": 2950, "lr": 2.6431481324452613e-07}
{"train_loss": 0.04386124759912491, "global_step": 262578, "epoch": 2950, "lr": 2.642552850971147e-07}
{"train_loss": 0.06575733423233032, "global_step": 262579, "epoch": 2950, "lr": 2.641957636360659e-07}
{"train_loss": 0.02914416417479515, "global_step": 262580, "epoch": 2950, "lr": 2.641362488613963e-07}
{"train_loss": 0.06261075288057327, "global_step": 262581, "epoch": 2950, "lr": 2.640767407731004e-07}
{"train_loss": 0.0339236706495285, "global_step": 262582, "epoch": 2950, "lr": 2.640172393711948e-07}
{"train_loss": 0.04428288713097572, "global_step": 262583, "epoch": 2950, "lr": 2.6395774465569065e-07}
{"train_loss": 0.032233063131570816, "global_step": 262584, "epoch": 2950, "lr": 2.6389825662658796e-07}
{"train_loss": 0.0653899610042572, "global_step": 262585, "epoch": 2950, "lr": 2.638387752838922e-07}
{"train_loss": 0.009033537469804287, "global_step": 262586, "epoch": 2950, "lr": 2.637793006276257e-07}
{"train_loss": 0.051862966269254684, "global_step": 262587, "epoch": 2950, "lr": 2.6371983265778277e-07}
{"train_loss": 0.023743515834212303, "global_step": 262588, "epoch": 2950, "lr": 2.6366037137438013e-07}
{"train_loss": 0.048637982457876205, "global_step": 262589, "epoch": 2950, "lr": 2.6360091677741784e-07}
{"train_loss": 0.05619211122393608, "global_step": 262590, "epoch": 2950, "lr": 2.6354146886691246e-07}
{"train_loss": 0.0690172091126442, "global_step": 262591, "epoch": 2950, "lr": 2.63482027642864e-07}
{"train_loss": 0.0679917111992836, "global_step": 262592, "epoch": 2950, "lr": 2.6342259310528916e-07}
{"train_loss": 0.06835296750068665, "global_step": 262593, "epoch": 2950, "lr": 2.633631652541879e-07}
{"train_loss": 0.03075333870947361, "global_step": 262594, "epoch": 2950, "lr": 2.6330374408957136e-07}
{"train_loss": 0.039359234273433685, "global_step": 262595, "epoch": 2950, "lr": 2.632443296114451e-07}
{"train_loss": 0.1138724684715271, "global_step": 262596, "epoch": 2950, "lr": 2.631849218198201e-07}
{"train_loss": 0.04890386760234833, "global_step": 262597, "epoch": 2950, "lr": 2.6312552071470766e-07}
{"train_loss": 0.06990434974431992, "global_step": 262598, "epoch": 2950, "lr": 2.630661262961076e-07}
{"train_loss": 0.008106364868581295, "global_step": 262599, "epoch": 2950, "lr": 2.630067385640367e-07}
{"train_loss": 0.08239728212356567, "global_step": 262600, "epoch": 2950, "lr": 2.6294735751849486e-07}
{"train_loss": 0.0756620243191719, "global_step": 262601, "epoch": 2950, "lr": 2.6288798315949327e-07}
{"train_loss": 0.0333714485168457, "global_step": 262602, "epoch": 2950, "lr": 2.628286154870374e-07}
{"train_loss": 0.013473771512508392, "global_step": 262603, "epoch": 2950, "lr": 2.62769254501144e-07}
{"train_loss": 0.023656249046325684, "global_step": 262604, "epoch": 2950, "lr": 2.6270990020181297e-07}
{"train_loss": 0.05871538445353508, "global_step": 262605, "epoch": 2950, "lr": 2.626505525890555e-07}
{"train_loss": 0.05742477998137474, "global_step": 262606, "epoch": 2950, "lr": 2.6259121166287705e-07}
{"train_loss": 0.09360390901565552, "global_step": 262607, "epoch": 2950, "lr": 2.6253187742328877e-07}
{"train_loss": 0.027749735862016678, "global_step": 262608, "epoch": 2950, "lr": 2.624725498702907e-07}
{"train_loss": 0.05227663367986679, "global_step": 262609, "epoch": 2950, "lr": 2.6241322900390497e-07}
{"train_loss": 0.056639738380908966, "global_step": 262610, "epoch": 2950, "lr": 2.623539148241261e-07}
{"train_loss": 0.061533574014902115, "global_step": 262611, "epoch": 2950, "lr": 2.6229460733097064e-07}
{"train_loss": 0.029021594673395157, "global_step": 262612, "epoch": 2950, "lr": 2.622353065244387e-07}
{"train_loss": 0.045233193784952164, "global_step": 262613, "epoch": 2950, "lr": 2.6217601240454693e-07}
{"train_loss": 0.07095088809728622, "global_step": 262614, "epoch": 2950, "lr": 2.621167249713008e-07}
{"train_loss": 0.03280286863446236, "global_step": 262615, "epoch": 2950, "lr": 2.6205744422470593e-07}
{"train_loss": 0.01714932732284069, "global_step": 262616, "epoch": 2950, "lr": 2.619981701647678e-07}
{"train_loss": 0.05811968073248863, "global_step": 262617, "epoch": 2950, "lr": 2.619389027915031e-07}
{"train_loss": 0.05090866610407829, "global_step": 262618, "epoch": 2950, "lr": 2.618796421049119e-07}
{"train_loss": 0.01841023750603199, "global_step": 262619, "epoch": 2950, "lr": 2.618203881050052e-07}
{"train_loss": 0.04762786626815796, "global_step": 262620, "epoch": 2950, "lr": 2.617611407917886e-07}
{"train_loss": 0.06807432323694229, "global_step": 262621, "epoch": 2950, "lr": 2.617019001652732e-07}
{"train_loss": 0.03441748768091202, "global_step": 262622, "epoch": 2950, "lr": 2.6164266622546454e-07}
{"train_loss": 0.02074897661805153, "global_step": 262623, "epoch": 2950, "lr": 2.615834389723737e-07}
{"train_loss": 0.01910291239619255, "global_step": 262624, "epoch": 2950, "lr": 2.6152421840600627e-07}
{"train_loss": 0.03259020671248436, "global_step": 262625, "epoch": 2950, "lr": 2.6146500452637333e-07}
{"train_loss": 0.05713120847940445, "global_step": 262626, "epoch": 2950, "lr": 2.614057973334749e-07}
{"train_loss": 0.049642618745565414, "global_step": 262627, "epoch": 2950, "lr": 2.6134659682732766e-07}
{"train_loss": 0.0473618321120739, "global_step": 262628, "epoch": 2950, "lr": 2.612874030079371e-07}
{"train_loss": 0.03663308173418045, "global_step": 262629, "epoch": 2950, "lr": 2.612282158753032e-07}
{"train_loss": 0.04705413803458214, "global_step": 262630, "epoch": 2950, "lr": 2.6116903542944824e-07}
{"train_loss": 0.024946179240942, "global_step": 262631, "epoch": 2950, "lr": 2.611098616703722e-07}
{"train_loss": 0.04607263579964638, "global_step": 262632, "epoch": 2950, "lr": 2.6105069459808065e-07}
{"train_loss": 0.05998525395989418, "global_step": 262633, "epoch": 2950, "lr": 2.609915342125846e-07}
{"train_loss": 0.028835127130150795, "global_step": 262634, "epoch": 2950, "lr": 2.6093238051389526e-07}
{"train_loss": 0.06714137643575668, "global_step": 262635, "epoch": 2950, "lr": 2.6087323350201807e-07}
{"train_loss": 0.04113996401429176, "global_step": 262636, "epoch": 2950, "lr": 2.6081409317695317e-07}
{"train_loss": 0.03105001337826252, "global_step": 262637, "epoch": 2950, "lr": 2.6075495953872263e-07}
{"train_loss": 0.04658523193571005, "global_step": 262638, "epoch": 2950, "lr": 2.606958325873265e-07, "train/sim_max_reward_0": 0.969569941290762, "train/sim_max_reward_1": 0.8690252665042076, "train/sim_max_reward_2": 0.48809434785577327, "train/sim_max_reward_3": 0.7178503635472601, "train/sim_max_reward_4": 0.9995336075189198, "train/sim_max_reward_5": 0.9952549888847996, "test/sim_max_reward_4500000": 0.4501213967229618, "test/sim_max_reward_4500001": 1.0, "test/sim_max_reward_4500002": 0.9303770194340388, "test/sim_max_reward_4500003": 0.519044480639616, "test/sim_max_reward_4500004": 0.9663455366268004, "test/sim_max_reward_4500005": 0.3917133834441786, "test/sim_max_reward_4500006": 0.9988199142760412, "test/sim_max_reward_4500007": 0.9187875742309932, "test/sim_max_reward_4500008": 0.2295204077088277, "test/sim_max_reward_4500009": 0.04940730422871775, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 0.0, "test/sim_max_reward_4500013": 0.49919735583076136, "test/sim_max_reward_4500014": 0.3926543889630716, "test/sim_max_reward_4500015": 0.9879912709907023, "test/sim_max_reward_4500016": 0.973193794747659, "test/sim_max_reward_4500017": 0.5199918094212627, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.9879480776866816, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 1.0, "test/sim_max_reward_4500022": 0.9134043988247091, "test/sim_max_reward_4500023": 0.6132335664250131, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 1.0, "test/sim_max_reward_4500026": 0.0, "test/sim_max_reward_4500027": 0.9822682786766269, "test/sim_max_reward_4500028": 0.5887321070488334, "test/sim_max_reward_4500029": 0.7211772986278148, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.0, "test/sim_max_reward_4500032": 0.36188047579357735, "test/sim_max_reward_4500033": 1.0, "test/sim_max_reward_4500034": 0.444006187514827, "test/sim_max_reward_4500035": 0.3905283399840129, "test/sim_max_reward_4500036": 0.0709947044204028, "test/sim_max_reward_4500037": 0.0, "test/sim_max_reward_4500038": 0.16246027888131, "test/sim_max_reward_4500039": 0.9383128912372154, "test/sim_max_reward_4500040": 0.816130048352738, "test/sim_max_reward_4500041": 0.48589780198217764, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.9892887932436295, "test/sim_max_reward_4500044": 0.4349875067780698, "test/sim_max_reward_4500045": 0.9975290708634882, "test/sim_max_reward_4500046": 0.6229810717812079, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 1.0, "test/sim_max_reward_4500049": 0.9710815595263067, "train/mean_score": 0.8398880859336204, "test/mean_score": 0.5864001618982855, "val_loss": 9.596595764160156, "train_action_mse_error": 14.827189445495605}
{"train_loss": 0.022544845938682556, "global_step": 262639, "epoch": 2951, "lr": 2.6063671232277045e-07}
{"train_loss": 0.04035436362028122, "global_step": 262640, "epoch": 2951, "lr": 2.6057759874506536e-07}
{"train_loss": 0.017907770350575447, "global_step": 262641, "epoch": 2951, "lr": 2.605184918542225e-07}
{"train_loss": 0.05513115972280502, "global_step": 262642, "epoch": 2951, "lr": 2.6045939165024735e-07}
{"train_loss": 0.06345119327306747, "global_step": 262643, "epoch": 2951, "lr": 2.604002981331455e-07}
{"train_loss": 0.06232237070798874, "global_step": 262644, "epoch": 2951, "lr": 2.6034121130292245e-07}
{"train_loss": 0.04383133724331856, "global_step": 262645, "epoch": 2951, "lr": 2.602821311595949e-07}
{"train_loss": 0.031237546354532242, "global_step": 262646, "epoch": 2951, "lr": 2.6022305770316833e-07}
{"train_loss": 0.03172754868865013, "global_step": 262647, "epoch": 2951, "lr": 2.6016399093364285e-07}
{"train_loss": 0.048767831176519394, "global_step": 262648, "epoch": 2951, "lr": 2.6010493085103507e-07}
{"train_loss": 0.030391253530979156, "global_step": 262649, "epoch": 2951, "lr": 2.600458774553505e-07}
{"train_loss": 0.0836385190486908, "global_step": 262650, "epoch": 2951, "lr": 2.5998683074659467e-07}
{"train_loss": 0.03830089420080185, "global_step": 262651, "epoch": 2951, "lr": 2.599277907247788e-07}
{"train_loss": 0.06475192308425903, "global_step": 262652, "epoch": 2951, "lr": 2.598687573899083e-07}
{"train_loss": 0.05955825746059418, "global_step": 262653, "epoch": 2951, "lr": 2.598097307419944e-07}
{"train_loss": 0.028562691062688828, "global_step": 262654, "epoch": 2951, "lr": 2.5975071078104264e-07}
{"train_loss": 0.07536069303750992, "global_step": 262655, "epoch": 2951, "lr": 2.5969169750705847e-07}
{"train_loss": 0.07658090442419052, "global_step": 262656, "epoch": 2951, "lr": 2.5963269092005303e-07}
{"train_loss": 0.031296275556087494, "global_step": 262657, "epoch": 2951, "lr": 2.595736910200375e-07}
{"train_loss": 0.042539410293102264, "global_step": 262658, "epoch": 2951, "lr": 2.5951469780701177e-07}
{"train_loss": 0.07207443565130234, "global_step": 262659, "epoch": 2951, "lr": 2.59455711280987e-07}
{"train_loss": 0.05605974793434143, "global_step": 262660, "epoch": 2951, "lr": 2.593967314419743e-07}
{"train_loss": 0.06090664118528366, "global_step": 262661, "epoch": 2951, "lr": 2.593377582899792e-07}
{"train_loss": 0.06523051857948303, "global_step": 262662, "epoch": 2951, "lr": 2.592787918250128e-07}
{"train_loss": 0.02600741945207119, "global_step": 262663, "epoch": 2951, "lr": 2.592198320470751e-07}
{"train_loss": 0.03957322984933853, "global_step": 262664, "epoch": 2951, "lr": 2.591608789561828e-07}
{"train_loss": 0.028228692710399628, "global_step": 262665, "epoch": 2951, "lr": 2.591019325523414e-07}
{"train_loss": 0.029934238642454147, "global_step": 262666, "epoch": 2951, "lr": 2.5904299283555647e-07}
{"train_loss": 0.04050673916935921, "global_step": 262667, "epoch": 2951, "lr": 2.589840598058335e-07}
{"train_loss": 0.019852373749017715, "global_step": 262668, "epoch": 2951, "lr": 2.589251334631837e-07}
{"train_loss": 0.021319905295968056, "global_step": 262669, "epoch": 2951, "lr": 2.588662138076181e-07}
{"train_loss": 0.02308533899486065, "global_step": 262670, "epoch": 2951, "lr": 2.588073008391423e-07}
{"train_loss": 0.034685660153627396, "global_step": 262671, "epoch": 2951, "lr": 2.587483945577618e-07}
{"train_loss": 0.014956463128328323, "global_step": 262672, "epoch": 2951, "lr": 2.5868949496348215e-07}
{"train_loss": 0.03749716654419899, "global_step": 262673, "epoch": 2951, "lr": 2.586306020563256e-07}
{"train_loss": 0.0808454230427742, "global_step": 262674, "epoch": 2951, "lr": 2.58571715836281e-07}
{"train_loss": 0.016023531556129456, "global_step": 262675, "epoch": 2951, "lr": 2.5851283630337064e-07}
{"train_loss": 0.056929536163806915, "global_step": 262676, "epoch": 2951, "lr": 2.584539634575944e-07}
{"train_loss": 0.037478502839803696, "global_step": 262677, "epoch": 2951, "lr": 2.583950972989635e-07}
{"train_loss": 0.06486787647008896, "global_step": 262678, "epoch": 2951, "lr": 2.583362378274834e-07}
{"train_loss": 0.0415656715631485, "global_step": 262679, "epoch": 2951, "lr": 2.582773850431652e-07}
{"train_loss": 0.02866343781352043, "global_step": 262680, "epoch": 2951, "lr": 2.5821853894601455e-07}
{"train_loss": 0.05751581862568855, "global_step": 262681, "epoch": 2951, "lr": 2.5815969953604247e-07}
{"train_loss": 0.027985842898488045, "global_step": 262682, "epoch": 2951, "lr": 2.5810086681324897e-07}
{"train_loss": 0.026498544961214066, "global_step": 262683, "epoch": 2951, "lr": 2.580420407776563e-07}
{"train_loss": 0.020901773124933243, "global_step": 262684, "epoch": 2951, "lr": 2.5798322142925324e-07}
{"train_loss": 0.05563028156757355, "global_step": 262685, "epoch": 2951, "lr": 2.579244087680677e-07}
{"train_loss": 0.06053636223077774, "global_step": 262686, "epoch": 2951, "lr": 2.5786560279409397e-07}
{"train_loss": 0.03637731075286865, "global_step": 262687, "epoch": 2951, "lr": 2.5780680350733774e-07}
{"train_loss": 0.03841051086783409, "global_step": 262688, "epoch": 2951, "lr": 2.577480109078212e-07}
{"train_loss": 0.044695254415273666, "global_step": 262689, "epoch": 2951, "lr": 2.5768922499554426e-07}
{"train_loss": 0.020061368122696877, "global_step": 262690, "epoch": 2951, "lr": 2.5763044577051256e-07}
{"train_loss": 0.01687280274927616, "global_step": 262691, "epoch": 2951, "lr": 2.575716732327371e-07}
{"train_loss": 0.03211568295955658, "global_step": 262692, "epoch": 2951, "lr": 2.5751290738222356e-07}
{"train_loss": 0.03799642622470856, "global_step": 262693, "epoch": 2951, "lr": 2.5745414821898293e-07}
{"train_loss": 0.03007740154862404, "global_step": 262694, "epoch": 2951, "lr": 2.573953957430153e-07}
{"train_loss": 0.03921320289373398, "global_step": 262695, "epoch": 2951, "lr": 2.5733664995434283e-07}
{"train_loss": 0.034668125212192535, "global_step": 262696, "epoch": 2951, "lr": 2.5727791085295993e-07}
{"train_loss": 0.06029585748910904, "global_step": 262697, "epoch": 2951, "lr": 2.572191784388833e-07}
{"train_loss": 0.03433440998196602, "global_step": 262698, "epoch": 2951, "lr": 2.57160452712113e-07}
{"train_loss": 0.042566534131765366, "global_step": 262699, "epoch": 2951, "lr": 2.5710173367266553e-07}
{"train_loss": 0.04309340566396713, "global_step": 262700, "epoch": 2951, "lr": 2.57043021320541e-07}
{"train_loss": 0.03339210897684097, "global_step": 262701, "epoch": 2951, "lr": 2.5698431565575057e-07}
{"train_loss": 0.042542532086372375, "global_step": 262702, "epoch": 2951, "lr": 2.569256166783052e-07}
{"train_loss": 0.04502267763018608, "global_step": 262703, "epoch": 2951, "lr": 2.5686692438821045e-07}
{"train_loss": 0.03249868005514145, "global_step": 262704, "epoch": 2951, "lr": 2.56808238785472e-07}
{"train_loss": 0.04973556101322174, "global_step": 262705, "epoch": 2951, "lr": 2.5674955987010084e-07}
{"train_loss": 0.02478170022368431, "global_step": 262706, "epoch": 2951, "lr": 2.56690887642097e-07}
{"train_loss": 0.06382828950881958, "global_step": 262707, "epoch": 2951, "lr": 2.5663222210148276e-07}
{"train_loss": 0.041741084307432175, "global_step": 262708, "epoch": 2951, "lr": 2.565735632482524e-07}
{"train_loss": 0.065361388027668, "global_step": 262709, "epoch": 2951, "lr": 2.5651491108242275e-07}
{"train_loss": 0.023030132055282593, "global_step": 262710, "epoch": 2951, "lr": 2.5645626560399925e-07}
{"train_loss": 0.05915762484073639, "global_step": 262711, "epoch": 2951, "lr": 2.5639762681298753e-07}
{"train_loss": 0.03159144148230553, "global_step": 262712, "epoch": 2951, "lr": 2.5633899470939304e-07}
{"train_loss": 0.04832170158624649, "global_step": 262713, "epoch": 2951, "lr": 2.562803692932325e-07}
{"train_loss": 0.029726583510637283, "global_step": 262714, "epoch": 2951, "lr": 2.5622175056450595e-07}
{"train_loss": 0.06508864462375641, "global_step": 262715, "epoch": 2951, "lr": 2.561631385232299e-07}
{"train_loss": 0.04407884180545807, "global_step": 262716, "epoch": 2951, "lr": 2.5610453316939897e-07}
{"train_loss": 0.02195700630545616, "global_step": 262717, "epoch": 2951, "lr": 2.560459345030297e-07}
{"train_loss": 0.02964629977941513, "global_step": 262718, "epoch": 2951, "lr": 2.5598734252413324e-07}
{"train_loss": 0.044674839824438095, "global_step": 262719, "epoch": 2951, "lr": 2.559287572327096e-07}
{"train_loss": 0.04235232621431351, "global_step": 262720, "epoch": 2951, "lr": 2.5587017862876985e-07}
{"train_loss": 0.05376356467604637, "global_step": 262721, "epoch": 2951, "lr": 2.558116067123251e-07}
{"train_loss": 0.10778465867042542, "global_step": 262722, "epoch": 2951, "lr": 2.557530414833753e-07}
{"train_loss": 0.07044228166341782, "global_step": 262723, "epoch": 2951, "lr": 2.556944829419372e-07}
{"train_loss": 0.0094227846711874, "global_step": 262724, "epoch": 2951, "lr": 2.5563593108801073e-07}
{"train_loss": 0.011952647939324379, "global_step": 262725, "epoch": 2951, "lr": 2.555773859216126e-07}
{"train_loss": 0.06789696961641312, "global_step": 262726, "epoch": 2951, "lr": 2.555188474427428e-07}
{"train_loss": 0.042637887797998576, "global_step": 262727, "epoch": 2951, "lr": 2.554603156514124e-07, "val_loss": 9.570244789123535}
{"train_loss": 0.039741870015859604, "global_step": 262728, "epoch": 2952, "lr": 2.554017905476325e-07}
{"train_loss": 0.04848247021436691, "global_step": 262729, "epoch": 2952, "lr": 2.553432721314031e-07}
{"train_loss": 0.022684959694743156, "global_step": 262730, "epoch": 2952, "lr": 2.552847604027353e-07}
{"train_loss": 0.02414010837674141, "global_step": 262731, "epoch": 2952, "lr": 2.552262553616458e-07}
{"train_loss": 0.04361380636692047, "global_step": 262732, "epoch": 2952, "lr": 2.55167757008129e-07}
{"train_loss": 0.05647813528776169, "global_step": 262733, "epoch": 2952, "lr": 2.55109265342196e-07}
{"train_loss": 0.03532438725233078, "global_step": 262734, "epoch": 2952, "lr": 2.5505078036386356e-07}
{"train_loss": 0.05024313926696777, "global_step": 262735, "epoch": 2952, "lr": 2.549923020731315e-07}
{"train_loss": 0.031268324702978134, "global_step": 262736, "epoch": 2952, "lr": 2.549338304700055e-07}
{"train_loss": 0.02113325148820877, "global_step": 262737, "epoch": 2952, "lr": 2.548753655545022e-07}
{"train_loss": 0.04285639524459839, "global_step": 262738, "epoch": 2952, "lr": 2.548169073266216e-07}
{"train_loss": 0.05756237357854843, "global_step": 262739, "epoch": 2952, "lr": 2.547584557863747e-07}
{"train_loss": 0.03145495802164078, "global_step": 262740, "epoch": 2952, "lr": 2.5470001093377273e-07}
{"train_loss": 0.042788200080394745, "global_step": 262741, "epoch": 2952, "lr": 2.5464157276881564e-07}
{"train_loss": 0.024209408089518547, "global_step": 262742, "epoch": 2952, "lr": 2.545831412915145e-07}
{"train_loss": 0.04033989459276199, "global_step": 262743, "epoch": 2952, "lr": 2.54524716501886e-07}
{"train_loss": 0.0775735154747963, "global_step": 262744, "epoch": 2952, "lr": 2.5446629839992464e-07}
{"train_loss": 0.02761450596153736, "global_step": 262745, "epoch": 2952, "lr": 2.5440788698564144e-07}
{"train_loss": 0.022777531296014786, "global_step": 262746, "epoch": 2952, "lr": 2.543494822590531e-07}
{"train_loss": 0.05623132735490799, "global_step": 262747, "epoch": 2952, "lr": 2.542910842201596e-07}
{"train_loss": 0.057893071323633194, "global_step": 262748, "epoch": 2952, "lr": 2.5423269286896643e-07}
{"train_loss": 0.016948223114013672, "global_step": 262749, "epoch": 2952, "lr": 2.5417430820549037e-07}
{"train_loss": 0.03433511033654213, "global_step": 262750, "epoch": 2952, "lr": 2.5411593022973133e-07}
{"train_loss": 0.07506868243217468, "global_step": 262751, "epoch": 2952, "lr": 2.540575589417005e-07}
{"train_loss": 0.05859437212347984, "global_step": 262752, "epoch": 2952, "lr": 2.539991943414033e-07}
{"train_loss": 0.03104967251420021, "global_step": 262753, "epoch": 2952, "lr": 2.539408364288509e-07}
{"train_loss": 0.0671028345823288, "global_step": 262754, "epoch": 2952, "lr": 2.5388248520404887e-07}
{"train_loss": 0.0547247976064682, "global_step": 262755, "epoch": 2952, "lr": 2.538241406670139e-07}
{"train_loss": 0.02790435589849949, "global_step": 262756, "epoch": 2952, "lr": 2.5376580281773476e-07}
{"train_loss": 0.1303085833787918, "global_step": 262757, "epoch": 2952, "lr": 2.537074716562393e-07}
{"train_loss": 0.07299407571554184, "global_step": 262758, "epoch": 2952, "lr": 2.5364914718252197e-07}
{"train_loss": 0.008103231899440289, "global_step": 262759, "epoch": 2952, "lr": 2.535908293965938e-07}
{"train_loss": 0.039020512253046036, "global_step": 262760, "epoch": 2952, "lr": 2.5353251829846603e-07}
{"train_loss": 0.02083456888794899, "global_step": 262761, "epoch": 2952, "lr": 2.5347421388814405e-07}
{"train_loss": 0.013399737887084484, "global_step": 262762, "epoch": 2952, "lr": 2.534159161656391e-07}
{"train_loss": 0.04705298691987991, "global_step": 262763, "epoch": 2952, "lr": 2.533576251309511e-07}
{"train_loss": 0.057671695947647095, "global_step": 262764, "epoch": 2952, "lr": 2.5329934078409667e-07}
{"train_loss": 0.04628089442849159, "global_step": 262765, "epoch": 2952, "lr": 2.532410631250759e-07}
{"train_loss": 0.03501909598708153, "global_step": 262766, "epoch": 2952, "lr": 2.5318279215390537e-07}
{"train_loss": 0.028735866770148277, "global_step": 262767, "epoch": 2952, "lr": 2.5312452787058516e-07}
{"train_loss": 0.05697619542479515, "global_step": 262768, "epoch": 2952, "lr": 2.5306627027513186e-07}
{"train_loss": 0.014989878050982952, "global_step": 262769, "epoch": 2952, "lr": 2.530080193675399e-07}
{"train_loss": 0.048746492713689804, "global_step": 262770, "epoch": 2952, "lr": 2.52949775147826e-07}
{"train_loss": 0.03473186120390892, "global_step": 262771, "epoch": 2952, "lr": 2.5289153761600127e-07}
{"train_loss": 0.0233557540923357, "global_step": 262772, "epoch": 2952, "lr": 2.528333067720656e-07}
{"train_loss": 0.04070466756820679, "global_step": 262773, "epoch": 2952, "lr": 2.5277508261603025e-07}
{"train_loss": 0.037933140993118286, "global_step": 262774, "epoch": 2952, "lr": 2.527168651479006e-07}
{"train_loss": 0.04035096988081932, "global_step": 262775, "epoch": 2952, "lr": 2.5265865436768787e-07}
{"train_loss": 0.10751844197511673, "global_step": 262776, "epoch": 2952, "lr": 2.5260045027540314e-07}
{"train_loss": 0.05895249545574188, "global_step": 262777, "epoch": 2952, "lr": 2.5254225287104636e-07}
{"train_loss": 0.03646446019411087, "global_step": 262778, "epoch": 2952, "lr": 2.5248406215462874e-07}
{"train_loss": 0.04302700608968735, "global_step": 262779, "epoch": 2952, "lr": 2.5242587812616124e-07}
{"train_loss": 0.05283365771174431, "global_step": 262780, "epoch": 2952, "lr": 2.5236770078564397e-07}
{"train_loss": 0.057850245386362076, "global_step": 262781, "epoch": 2952, "lr": 2.5230953013309355e-07}
{"train_loss": 0.06638607382774353, "global_step": 262782, "epoch": 2952, "lr": 2.5225136616851e-07}
{"train_loss": 0.021258018910884857, "global_step": 262783, "epoch": 2952, "lr": 2.521932088919099e-07}
{"train_loss": 0.030351633206009865, "global_step": 262784, "epoch": 2952, "lr": 2.5213505830329333e-07}
{"train_loss": 0.0353391133248806, "global_step": 262785, "epoch": 2952, "lr": 2.5207691440267136e-07}
{"train_loss": 0.03172930330038071, "global_step": 262786, "epoch": 2952, "lr": 2.5201877719004954e-07}
{"train_loss": 0.028611959889531136, "global_step": 262787, "epoch": 2952, "lr": 2.5196064666543893e-07}
{"train_loss": 0.05267907679080963, "global_step": 262788, "epoch": 2952, "lr": 2.519025228288452e-07}
{"train_loss": 0.09117923676967621, "global_step": 262789, "epoch": 2952, "lr": 2.518444056802793e-07}
{"train_loss": 0.0492950901389122, "global_step": 262790, "epoch": 2952, "lr": 2.517862952197414e-07}
{"train_loss": 0.024474620819091797, "global_step": 262791, "epoch": 2952, "lr": 2.51728191447248e-07}
{"train_loss": 0.025310508906841278, "global_step": 262792, "epoch": 2952, "lr": 2.5167009436280475e-07}
{"train_loss": 0.02405541017651558, "global_step": 262793, "epoch": 2952, "lr": 2.5161200396641713e-07}
{"train_loss": 0.06086723506450653, "global_step": 262794, "epoch": 2952, "lr": 2.5155392025809076e-07}
{"train_loss": 0.02356782555580139, "global_step": 262795, "epoch": 2952, "lr": 2.514958432378367e-07}
{"train_loss": 0.038786210119724274, "global_step": 262796, "epoch": 2952, "lr": 2.5143777290566605e-07}
{"train_loss": 0.05373635143041611, "global_step": 262797, "epoch": 2952, "lr": 2.513797092615844e-07}
{"train_loss": 0.06129428744316101, "global_step": 262798, "epoch": 2952, "lr": 2.513216523055972e-07}
{"train_loss": 0.0469413585960865, "global_step": 262799, "epoch": 2952, "lr": 2.5126360203771015e-07}
{"train_loss": 0.09265825152397156, "global_step": 262800, "epoch": 2952, "lr": 2.512055584579398e-07}
{"train_loss": 0.02857021801173687, "global_step": 262801, "epoch": 2952, "lr": 2.511475215662806e-07}
{"train_loss": 0.0356767401099205, "global_step": 262802, "epoch": 2952, "lr": 2.5108949136275484e-07}
{"train_loss": 0.023824721574783325, "global_step": 262803, "epoch": 2952, "lr": 2.510314678473624e-07}
{"train_loss": 0.048277586698532104, "global_step": 262804, "epoch": 2952, "lr": 2.509734510201145e-07}
{"train_loss": 0.05098079517483711, "global_step": 262805, "epoch": 2952, "lr": 2.5091544088101104e-07}
{"train_loss": 0.034603018313646317, "global_step": 262806, "epoch": 2952, "lr": 2.5085743743006873e-07}
{"train_loss": 0.03590593487024307, "global_step": 262807, "epoch": 2952, "lr": 2.5079944066729867e-07}
{"train_loss": 0.04436340183019638, "global_step": 262808, "epoch": 2952, "lr": 2.5074145059269526e-07}
{"train_loss": 0.07086562365293503, "global_step": 262809, "epoch": 2952, "lr": 2.5068346720627523e-07}
{"train_loss": 0.04272669553756714, "global_step": 262810, "epoch": 2952, "lr": 2.506254905080385e-07}
{"train_loss": 0.03224716708064079, "global_step": 262811, "epoch": 2952, "lr": 2.5056752049800736e-07}
{"train_loss": 0.020334366708993912, "global_step": 262812, "epoch": 2952, "lr": 2.505095571761762e-07}
{"train_loss": 0.025091655552387238, "global_step": 262813, "epoch": 2952, "lr": 2.5045160054256166e-07}
{"train_loss": 0.018394481390714645, "global_step": 262814, "epoch": 2952, "lr": 2.5039365059715824e-07}
{"train_loss": 0.04316435754299164, "global_step": 262815, "epoch": 2952, "lr": 2.5033570733999367e-07}
{"train_loss": 0.04266294983498166, "global_step": 262816, "epoch": 2952, "lr": 2.502777707710624e-07, "val_loss": 9.670371055603027}
{"train_loss": 0.0464787483215332, "global_step": 262817, "epoch": 2953, "lr": 2.5021984089037e-07}
{"train_loss": 0.0560884103178978, "global_step": 262818, "epoch": 2953, "lr": 2.5016191769793305e-07}
{"train_loss": 0.048437539488077164, "global_step": 262819, "epoch": 2953, "lr": 2.501040011937572e-07}
{"train_loss": 0.03547137603163719, "global_step": 262820, "epoch": 2953, "lr": 2.500460913778424e-07}
{"train_loss": 0.04086895287036896, "global_step": 262821, "epoch": 2953, "lr": 2.4998818825021085e-07}
{"train_loss": 0.03580459579825401, "global_step": 262822, "epoch": 2953, "lr": 2.4993029181085703e-07}
{"train_loss": 0.03182243928313255, "global_step": 262823, "epoch": 2953, "lr": 2.4987240205979203e-07}
{"train_loss": 0.034727104008197784, "global_step": 262824, "epoch": 2953, "lr": 2.498145189970269e-07}
{"train_loss": 0.038364410400390625, "global_step": 262825, "epoch": 2953, "lr": 2.497566426225728e-07}
{"train_loss": 0.0584474541246891, "global_step": 262826, "epoch": 2953, "lr": 2.496987729364242e-07}
{"train_loss": 0.06244153901934624, "global_step": 262827, "epoch": 2953, "lr": 2.496409099386032e-07}
{"train_loss": 0.032785944640636444, "global_step": 262828, "epoch": 2953, "lr": 2.4958305362910995e-07}
{"train_loss": 0.056575123220682144, "global_step": 262829, "epoch": 2953, "lr": 2.4952520400795544e-07}
{"train_loss": 0.036866866052150726, "global_step": 262830, "epoch": 2953, "lr": 2.4946736107514525e-07}
{"train_loss": 0.04827957972884178, "global_step": 262831, "epoch": 2953, "lr": 2.494095248306849e-07}
{"train_loss": 0.022339971736073494, "global_step": 262832, "epoch": 2953, "lr": 2.493516952745856e-07}
{"train_loss": 0.07720708847045898, "global_step": 262833, "epoch": 2953, "lr": 2.4929387240685274e-07}
{"train_loss": 0.07478661090135574, "global_step": 262834, "epoch": 2953, "lr": 2.492360562275031e-07}
{"train_loss": 0.04246849566698074, "global_step": 262835, "epoch": 2953, "lr": 2.491782467365311e-07}
{"train_loss": 0.035360343754291534, "global_step": 262836, "epoch": 2953, "lr": 2.491204439339534e-07}
{"train_loss": 0.03547525778412819, "global_step": 262837, "epoch": 2953, "lr": 2.490626478197755e-07}
{"train_loss": 0.028813250362873077, "global_step": 262838, "epoch": 2953, "lr": 2.49004858394003e-07}
{"train_loss": 0.03982788696885109, "global_step": 262839, "epoch": 2953, "lr": 2.48947075656647e-07}
{"train_loss": 0.056684959679841995, "global_step": 262840, "epoch": 2953, "lr": 2.48889299607713e-07}
{"train_loss": 0.11542392522096634, "global_step": 262841, "epoch": 2953, "lr": 2.4883153024720665e-07}
{"train_loss": 0.06848136335611343, "global_step": 262842, "epoch": 2953, "lr": 2.4877376757514445e-07}
{"train_loss": 0.04215928912162781, "global_step": 262843, "epoch": 2953, "lr": 2.4871601159152657e-07}
{"train_loss": 0.05160646140575409, "global_step": 262844, "epoch": 2953, "lr": 2.4865826229635846e-07}
{"train_loss": 0.02999148517847061, "global_step": 262845, "epoch": 2953, "lr": 2.486005196896568e-07}
{"train_loss": 0.07598068565130234, "global_step": 262846, "epoch": 2953, "lr": 2.485427837714216e-07}
{"train_loss": 0.060808323323726654, "global_step": 262847, "epoch": 2953, "lr": 2.4848505454166394e-07}
{"train_loss": 0.0423005111515522, "global_step": 262848, "epoch": 2953, "lr": 2.484273320003894e-07}
{"train_loss": 0.03642166778445244, "global_step": 262849, "epoch": 2953, "lr": 2.48369616147609e-07}
{"train_loss": 0.022341972216963768, "global_step": 262850, "epoch": 2953, "lr": 2.483119069833284e-07}
{"train_loss": 0.0449482761323452, "global_step": 262851, "epoch": 2953, "lr": 2.482542045075586e-07}
{"train_loss": 0.050672776997089386, "global_step": 262852, "epoch": 2953, "lr": 2.4819650872029977e-07}
{"train_loss": 0.018221531063318253, "global_step": 262853, "epoch": 2953, "lr": 2.481388196215684e-07}
{"train_loss": 0.02420835942029953, "global_step": 262854, "epoch": 2953, "lr": 2.480811372113645e-07}
{"train_loss": 0.053813520818948746, "global_step": 262855, "epoch": 2953, "lr": 2.4802346148970477e-07}
{"train_loss": 0.08217231929302216, "global_step": 262856, "epoch": 2953, "lr": 2.4796579245658924e-07}
{"train_loss": 0.04623715579509735, "global_step": 262857, "epoch": 2953, "lr": 2.4790813011202894e-07}
{"train_loss": 0.042150724679231644, "global_step": 262858, "epoch": 2953, "lr": 2.478504744560295e-07}
{"train_loss": 0.061752792447805405, "global_step": 262859, "epoch": 2953, "lr": 2.4779282548860196e-07}
{"train_loss": 0.013880331069231033, "global_step": 262860, "epoch": 2953, "lr": 2.4773518320975184e-07}
{"train_loss": 0.046835482120513916, "global_step": 262861, "epoch": 2953, "lr": 2.4767754761949036e-07}
{"train_loss": 0.039751213043928146, "global_step": 262862, "epoch": 2953, "lr": 2.4761991871781745e-07}
{"train_loss": 0.013354089111089706, "global_step": 262863, "epoch": 2953, "lr": 2.4756229650474417e-07}
{"train_loss": 0.0364036038517952, "global_step": 262864, "epoch": 2953, "lr": 2.4750468098028165e-07}
{"train_loss": 0.024523500353097916, "global_step": 262865, "epoch": 2953, "lr": 2.4744707214444106e-07}
{"train_loss": 0.031620610505342484, "global_step": 262866, "epoch": 2953, "lr": 2.473894699972168e-07}
{"train_loss": 0.024629153311252594, "global_step": 262867, "epoch": 2953, "lr": 2.4733187453863106e-07}
{"train_loss": 0.015939386561512947, "global_step": 262868, "epoch": 2953, "lr": 2.4727428576867827e-07}
{"train_loss": 0.044467147439718246, "global_step": 262869, "epoch": 2953, "lr": 2.472167036873807e-07}
{"train_loss": 0.047519490122795105, "global_step": 262870, "epoch": 2953, "lr": 2.4715912829473275e-07}
{"train_loss": 0.06150738522410393, "global_step": 262871, "epoch": 2953, "lr": 2.471015595907511e-07}
{"train_loss": 0.012504027225077152, "global_step": 262872, "epoch": 2953, "lr": 2.4704399757544127e-07}
{"train_loss": 0.07016942650079727, "global_step": 262873, "epoch": 2953, "lr": 2.469864422488033e-07}
{"train_loss": 0.02980879321694374, "global_step": 262874, "epoch": 2953, "lr": 2.4692889361085937e-07}
{"train_loss": 0.0287712924182415, "global_step": 262875, "epoch": 2953, "lr": 2.468713516616039e-07}
{"train_loss": 0.01747049391269684, "global_step": 262876, "epoch": 2953, "lr": 2.4681381640105363e-07}
{"train_loss": 0.036995213478803635, "global_step": 262877, "epoch": 2953, "lr": 2.46756287829214e-07}
{"train_loss": 0.02796700783073902, "global_step": 262878, "epoch": 2953, "lr": 2.4669876594609067e-07}
{"train_loss": 0.012474864721298218, "global_step": 262879, "epoch": 2953, "lr": 2.466412507516891e-07}
{"train_loss": 0.030933711677789688, "global_step": 262880, "epoch": 2953, "lr": 2.4658374224602045e-07}
{"train_loss": 0.08226609975099564, "global_step": 262881, "epoch": 2953, "lr": 2.4652624042909576e-07}
{"train_loss": 0.07562635093927383, "global_step": 262882, "epoch": 2953, "lr": 2.4646874530091513e-07}
{"train_loss": 0.0356585867702961, "global_step": 262883, "epoch": 2953, "lr": 2.4641125686149513e-07}
{"train_loss": 0.07801254093647003, "global_step": 262884, "epoch": 2953, "lr": 2.463537751108358e-07}
{"train_loss": 0.07099219411611557, "global_step": 262885, "epoch": 2953, "lr": 2.4629630004894823e-07}
{"train_loss": 0.03335373476147652, "global_step": 262886, "epoch": 2953, "lr": 2.4623883167583796e-07}
{"train_loss": 0.0460597462952137, "global_step": 262887, "epoch": 2953, "lr": 2.461813699915161e-07}
{"train_loss": 0.05692240968346596, "global_step": 262888, "epoch": 2953, "lr": 2.461239149959882e-07}
{"train_loss": 0.0427580326795578, "global_step": 262889, "epoch": 2953, "lr": 2.4606646668926536e-07}
{"train_loss": 0.05036354437470436, "global_step": 262890, "epoch": 2953, "lr": 2.460090250713476e-07}
{"train_loss": 0.05335071682929993, "global_step": 262891, "epoch": 2953, "lr": 2.4595159014225156e-07}
{"train_loss": 0.041391804814338684, "global_step": 262892, "epoch": 2953, "lr": 2.458941619019772e-07}
{"train_loss": 0.032843589782714844, "global_step": 262893, "epoch": 2953, "lr": 2.4583674035054127e-07}
{"train_loss": 0.09092831611633301, "global_step": 262894, "epoch": 2953, "lr": 2.457793254879437e-07}
{"train_loss": 0.046694427728652954, "global_step": 262895, "epoch": 2953, "lr": 2.4572191731419557e-07}
{"train_loss": 0.02363225817680359, "global_step": 262896, "epoch": 2953, "lr": 2.4566451582929694e-07}
{"train_loss": 0.060931719839572906, "global_step": 262897, "epoch": 2953, "lr": 2.4560712103327e-07}
{"train_loss": 0.06802497804164886, "global_step": 262898, "epoch": 2953, "lr": 2.4554973292610914e-07}
{"train_loss": 0.05141763016581535, "global_step": 262899, "epoch": 2953, "lr": 2.4549235150783666e-07}
{"train_loss": 0.02065235748887062, "global_step": 262900, "epoch": 2953, "lr": 2.4543497677844143e-07}
{"train_loss": 0.06040671095252037, "global_step": 262901, "epoch": 2953, "lr": 2.453776087379456e-07}
{"train_loss": 0.018254216760396957, "global_step": 262902, "epoch": 2953, "lr": 2.4532024738634917e-07}
{"train_loss": 0.04451632499694824, "global_step": 262903, "epoch": 2953, "lr": 2.4526289272366887e-07}
{"train_loss": 0.050775814801454544, "global_step": 262904, "epoch": 2953, "lr": 2.452055447499046e-07}
{"train_loss": 0.0444554227320582, "global_step": 262905, "epoch": 2953, "lr": 2.4514820346506205e-07, "val_loss": 9.608022689819336}
{"train_loss": 0.022052085027098656, "global_step": 262906, "epoch": 2954, "lr": 2.4509086886915777e-07}
{"train_loss": 0.014968379400670528, "global_step": 262907, "epoch": 2954, "lr": 2.450335409621918e-07}
{"train_loss": 0.0541403666138649, "global_step": 262908, "epoch": 2954, "lr": 2.4497621974417515e-07}
{"train_loss": 0.055957578122615814, "global_step": 262909, "epoch": 2954, "lr": 2.449189052151135e-07}
{"train_loss": 0.03384346142411232, "global_step": 262910, "epoch": 2954, "lr": 2.448615973750179e-07}
{"train_loss": 0.046167876571416855, "global_step": 262911, "epoch": 2954, "lr": 2.448042962238939e-07}
{"train_loss": 0.08604049682617188, "global_step": 262912, "epoch": 2954, "lr": 2.447470017617526e-07}
{"train_loss": 0.08569168299436569, "global_step": 262913, "epoch": 2954, "lr": 2.44689713988594e-07}
{"train_loss": 0.031141554936766624, "global_step": 262914, "epoch": 2954, "lr": 2.446324329044347e-07}
{"train_loss": 0.050143275409936905, "global_step": 262915, "epoch": 2954, "lr": 2.445751585092693e-07}
{"train_loss": 0.08057067543268204, "global_step": 262916, "epoch": 2954, "lr": 2.445178908031254e-07}
{"train_loss": 0.05272935703396797, "global_step": 262917, "epoch": 2954, "lr": 2.4446062978599195e-07}
{"train_loss": 0.04607534781098366, "global_step": 262918, "epoch": 2954, "lr": 2.4440337545789114e-07}
{"train_loss": 0.051822979003190994, "global_step": 262919, "epoch": 2954, "lr": 2.443461278188175e-07}
{"train_loss": 0.05519770085811615, "global_step": 262920, "epoch": 2954, "lr": 2.4428888686878757e-07}
{"train_loss": 0.06088579446077347, "global_step": 262921, "epoch": 2954, "lr": 2.44231652607807e-07}
{"train_loss": 0.07813195139169693, "global_step": 262922, "epoch": 2954, "lr": 2.441744250358813e-07}
{"train_loss": 0.017906559631228447, "global_step": 262923, "epoch": 2954, "lr": 2.4411720415302154e-07}
{"train_loss": 0.018869537860155106, "global_step": 262924, "epoch": 2954, "lr": 2.440599899592333e-07}
{"train_loss": 0.02882617525756359, "global_step": 262925, "epoch": 2954, "lr": 2.4400278245452767e-07}
{"train_loss": 0.04207265004515648, "global_step": 262926, "epoch": 2954, "lr": 2.439455816389047e-07}
{"train_loss": 0.01960352621972561, "global_step": 262927, "epoch": 2954, "lr": 2.4388838751238094e-07}
{"train_loss": 0.05030093342065811, "global_step": 262928, "epoch": 2954, "lr": 2.4383120007495653e-07}
{"train_loss": 0.08671697229146957, "global_step": 262929, "epoch": 2954, "lr": 2.4377401932664247e-07}
{"train_loss": 0.03781994432210922, "global_step": 262930, "epoch": 2954, "lr": 2.4371684526744986e-07}
{"train_loss": 0.06791365891695023, "global_step": 262931, "epoch": 2954, "lr": 2.436596778973788e-07}
{"train_loss": 0.013601998798549175, "global_step": 262932, "epoch": 2954, "lr": 2.436025172164458e-07}
{"train_loss": 0.027092956006526947, "global_step": 262933, "epoch": 2954, "lr": 2.435453632246509e-07}
{"train_loss": 0.024489562958478928, "global_step": 262934, "epoch": 2954, "lr": 2.434882159220053e-07}
{"train_loss": 0.06786837428808212, "global_step": 262935, "epoch": 2954, "lr": 2.4343107530852004e-07}
{"train_loss": 0.06790966540575027, "global_step": 262936, "epoch": 2954, "lr": 2.4337394138419513e-07}
{"train_loss": 0.041628867387771606, "global_step": 262937, "epoch": 2954, "lr": 2.433168141490416e-07}
{"train_loss": 0.057476580142974854, "global_step": 262938, "epoch": 2954, "lr": 2.432596936030651e-07}
{"train_loss": 0.051965679973363876, "global_step": 262939, "epoch": 2954, "lr": 2.4320257974628223e-07}
{"train_loss": 0.035359445959329605, "global_step": 262940, "epoch": 2954, "lr": 2.43145472578693e-07}
{"train_loss": 0.04552655667066574, "global_step": 262941, "epoch": 2954, "lr": 2.43088372100303e-07}
{"train_loss": 0.0637604221701622, "global_step": 262942, "epoch": 2954, "lr": 2.4303127831112325e-07}
{"train_loss": 0.03220629319548607, "global_step": 262943, "epoch": 2954, "lr": 2.429741912111649e-07}
{"train_loss": 0.04192991182208061, "global_step": 262944, "epoch": 2954, "lr": 2.4291711080043354e-07}
{"train_loss": 0.02379448711872101, "global_step": 262945, "epoch": 2954, "lr": 2.4286003707892914e-07}
{"train_loss": 0.06023663654923439, "global_step": 262946, "epoch": 2954, "lr": 2.4280297004667385e-07}
{"train_loss": 0.046074848622083664, "global_step": 262947, "epoch": 2954, "lr": 2.427459097036622e-07}
{"train_loss": 0.02860884554684162, "global_step": 262948, "epoch": 2954, "lr": 2.426888560499052e-07}
{"train_loss": 0.03252654895186424, "global_step": 262949, "epoch": 2954, "lr": 2.4263180908541406e-07}
{"train_loss": 0.024749819189310074, "global_step": 262950, "epoch": 2954, "lr": 2.4257476881019423e-07}
{"train_loss": 0.05770069360733032, "global_step": 262951, "epoch": 2954, "lr": 2.425177352242569e-07}
{"train_loss": 0.05262228101491928, "global_step": 262952, "epoch": 2954, "lr": 2.4246070832760205e-07}
{"train_loss": 0.033823978155851364, "global_step": 262953, "epoch": 2954, "lr": 2.4240368812024626e-07}
{"train_loss": 0.07313094288110733, "global_step": 262954, "epoch": 2954, "lr": 2.423466746021896e-07}
{"train_loss": 0.021660616621375084, "global_step": 262955, "epoch": 2954, "lr": 2.422896677734432e-07}
{"train_loss": 0.07204217463731766, "global_step": 262956, "epoch": 2954, "lr": 2.422326676340181e-07}
{"train_loss": 0.04162384569644928, "global_step": 262957, "epoch": 2954, "lr": 2.421756741839143e-07}
{"train_loss": 0.015902522951364517, "global_step": 262958, "epoch": 2954, "lr": 2.421186874231485e-07}
{"train_loss": 0.04624677449464798, "global_step": 262959, "epoch": 2954, "lr": 2.4206170735171506e-07}
{"train_loss": 0.04630913585424423, "global_step": 262960, "epoch": 2954, "lr": 2.420047339696363e-07}
{"train_loss": 0.04157709330320358, "global_step": 262961, "epoch": 2954, "lr": 2.4194776727691214e-07}
{"train_loss": 0.08248969167470932, "global_step": 262962, "epoch": 2954, "lr": 2.4189080727355376e-07}
{"train_loss": 0.024555474519729614, "global_step": 262963, "epoch": 2954, "lr": 2.418338539595666e-07}
{"train_loss": 0.03885817527770996, "global_step": 262964, "epoch": 2954, "lr": 2.4177690733495626e-07}
{"train_loss": 0.07007795572280884, "global_step": 262965, "epoch": 2954, "lr": 2.4171996739972836e-07}
{"train_loss": 0.04359223321080208, "global_step": 262966, "epoch": 2954, "lr": 2.416630341538995e-07}
{"train_loss": 0.03214595466852188, "global_step": 262967, "epoch": 2954, "lr": 2.4160610759747514e-07}
{"train_loss": 0.01154295727610588, "global_step": 262968, "epoch": 2954, "lr": 2.4154918773045544e-07}
{"train_loss": 0.013234784826636314, "global_step": 262969, "epoch": 2954, "lr": 2.4149227455285694e-07}
{"train_loss": 0.04720370098948479, "global_step": 262970, "epoch": 2954, "lr": 2.414353680646797e-07}
{"train_loss": 0.04905925691127777, "global_step": 262971, "epoch": 2954, "lr": 2.413784682659348e-07}
{"train_loss": 0.048252172768116, "global_step": 262972, "epoch": 2954, "lr": 2.413215751566333e-07}
{"train_loss": 0.10827214270830154, "global_step": 262973, "epoch": 2954, "lr": 2.4126468873678085e-07}
{"train_loss": 0.04721854254603386, "global_step": 262974, "epoch": 2954, "lr": 2.412078090063774e-07}
{"train_loss": 0.05614348500967026, "global_step": 262975, "epoch": 2954, "lr": 2.411509359654396e-07}
{"train_loss": 0.051924534142017365, "global_step": 262976, "epoch": 2954, "lr": 2.41094069613973e-07}
{"train_loss": 0.032994844019412994, "global_step": 262977, "epoch": 2954, "lr": 2.410372099519886e-07}
{"train_loss": 0.06232120841741562, "global_step": 262978, "epoch": 2954, "lr": 2.409803569794866e-07}
{"train_loss": 0.08021803200244904, "global_step": 262979, "epoch": 2954, "lr": 2.40923510696478e-07}
{"train_loss": 0.042956557124853134, "global_step": 262980, "epoch": 2954, "lr": 2.408666711029739e-07}
{"train_loss": 0.054150450974702835, "global_step": 262981, "epoch": 2954, "lr": 2.408098381989743e-07}
{"train_loss": 0.0878034383058548, "global_step": 262982, "epoch": 2954, "lr": 2.4075301198449586e-07}
{"train_loss": 0.03808173909783363, "global_step": 262983, "epoch": 2954, "lr": 2.4069619245953854e-07}
{"train_loss": 0.08684592694044113, "global_step": 262984, "epoch": 2954, "lr": 2.4063937962411356e-07}
{"train_loss": 0.041417647153139114, "global_step": 262985, "epoch": 2954, "lr": 2.4058257347822635e-07}
{"train_loss": 0.02418375574052334, "global_step": 262986, "epoch": 2954, "lr": 2.405257740218936e-07}
{"train_loss": 0.03681385889649391, "global_step": 262987, "epoch": 2954, "lr": 2.4046898125510975e-07}
{"train_loss": 0.06767718493938446, "global_step": 262988, "epoch": 2954, "lr": 2.4041219517788596e-07}
{"train_loss": 0.039622776210308075, "global_step": 262989, "epoch": 2954, "lr": 2.403554157902388e-07}
{"train_loss": 0.028222499415278435, "global_step": 262990, "epoch": 2954, "lr": 2.402986430921683e-07}
{"train_loss": 0.03538615256547928, "global_step": 262991, "epoch": 2954, "lr": 2.4024187708368004e-07}
{"train_loss": 0.03522215038537979, "global_step": 262992, "epoch": 2954, "lr": 2.4018511776478515e-07}
{"train_loss": 0.025369010865688324, "global_step": 262993, "epoch": 2954, "lr": 2.401283651354891e-07}
{"train_loss": 0.04646423061409693, "global_step": 262994, "epoch": 2954, "lr": 2.40071619195803e-07, "val_loss": 9.66894817352295}
{"train_loss": 0.04928424954414368, "global_step": 262995, "epoch": 2955, "lr": 2.4001487994573246e-07}
{"train_loss": 0.022921793162822723, "global_step": 262996, "epoch": 2955, "lr": 2.3995814738528854e-07}
{"train_loss": 0.06982727348804474, "global_step": 262997, "epoch": 2955, "lr": 2.399014215144768e-07}
{"train_loss": 0.022417334839701653, "global_step": 262998, "epoch": 2955, "lr": 2.3984470233329727e-07}
{"train_loss": 0.03430154547095299, "global_step": 262999, "epoch": 2955, "lr": 2.3978798984176654e-07}
{"train_loss": 0.045645397156476974, "global_step": 263000, "epoch": 2955, "lr": 2.3973128403989577e-07}
{"train_loss": 0.02833506278693676, "global_step": 263001, "epoch": 2955, "lr": 2.3967458492767936e-07}
{"train_loss": 0.03764650225639343, "global_step": 263002, "epoch": 2955, "lr": 2.39617892505134e-07}
{"train_loss": 0.013452237471938133, "global_step": 263003, "epoch": 2955, "lr": 2.395612067722708e-07}
{"train_loss": 0.011918638832867146, "global_step": 263004, "epoch": 2955, "lr": 2.3950452772908414e-07}
{"train_loss": 0.018123652786016464, "global_step": 263005, "epoch": 2955, "lr": 2.394478553755963e-07}
{"train_loss": 0.06811950355768204, "global_step": 263006, "epoch": 2955, "lr": 2.3939118971180176e-07}
{"train_loss": 0.03685170039534569, "global_step": 263007, "epoch": 2955, "lr": 2.393345307377171e-07}
{"train_loss": 0.0394597053527832, "global_step": 263008, "epoch": 2955, "lr": 2.392778784533478e-07}
{"train_loss": 0.061003465205430984, "global_step": 263009, "epoch": 2955, "lr": 2.3922123285870513e-07}
{"train_loss": 0.01830727979540825, "global_step": 263010, "epoch": 2955, "lr": 2.391645939537834e-07}
{"train_loss": 0.02193678356707096, "global_step": 263011, "epoch": 2955, "lr": 2.391079617386105e-07}
{"train_loss": 0.014829274266958237, "global_step": 263012, "epoch": 2955, "lr": 2.390513362131752e-07}
{"train_loss": 0.07078643143177032, "global_step": 263013, "epoch": 2955, "lr": 2.389947173774942e-07}
{"train_loss": 0.03477078303694725, "global_step": 263014, "epoch": 2955, "lr": 2.389381052315731e-07}
{"train_loss": 0.026881976053118706, "global_step": 263015, "epoch": 2955, "lr": 2.3888149977542295e-07}
{"train_loss": 0.036462001502513885, "global_step": 263016, "epoch": 2955, "lr": 2.388249010090493e-07}
{"train_loss": 0.02999839559197426, "global_step": 263017, "epoch": 2955, "lr": 2.387683089324577e-07}
{"train_loss": 0.045307278633117676, "global_step": 263018, "epoch": 2955, "lr": 2.3871172354565376e-07}
{"train_loss": 0.04925648123025894, "global_step": 263019, "epoch": 2955, "lr": 2.38655144848654e-07}
{"train_loss": 0.041988108307123184, "global_step": 263020, "epoch": 2955, "lr": 2.385985728414586e-07}
{"train_loss": 0.010299043729901314, "global_step": 263021, "epoch": 2955, "lr": 2.38542007524073e-07}
{"train_loss": 0.057136159390211105, "global_step": 263022, "epoch": 2955, "lr": 2.384854488965138e-07}
{"train_loss": 0.03293924778699875, "global_step": 263023, "epoch": 2955, "lr": 2.384288969587811e-07}
{"train_loss": 0.04290119931101799, "global_step": 263024, "epoch": 2955, "lr": 2.3837235171089155e-07}
{"train_loss": 0.047878433018922806, "global_step": 263025, "epoch": 2955, "lr": 2.3831581315283956e-07}
{"train_loss": 0.050158414989709854, "global_step": 263026, "epoch": 2955, "lr": 2.382592812846418e-07}
{"train_loss": 0.04911356046795845, "global_step": 263027, "epoch": 2955, "lr": 2.3820275610630382e-07}
{"train_loss": 0.03011779859662056, "global_step": 263028, "epoch": 2955, "lr": 2.3814623761783116e-07}
{"train_loss": 0.08113431930541992, "global_step": 263029, "epoch": 2955, "lr": 2.3808972581923493e-07}
{"train_loss": 0.03268471360206604, "global_step": 263030, "epoch": 2955, "lr": 2.380332207105207e-07}
{"train_loss": 0.009185285307466984, "global_step": 263031, "epoch": 2955, "lr": 2.37976722291694e-07}
{"train_loss": 0.10190500319004059, "global_step": 263032, "epoch": 2955, "lr": 2.379202305627659e-07}
{"train_loss": 0.032458510249853134, "global_step": 263033, "epoch": 2955, "lr": 2.3786374552374757e-07}
{"train_loss": 0.03565371409058571, "global_step": 263034, "epoch": 2955, "lr": 2.3780726717463897e-07}
{"train_loss": 0.0633939579129219, "global_step": 263035, "epoch": 2955, "lr": 2.377507955154512e-07}
{"train_loss": 0.05301246792078018, "global_step": 263036, "epoch": 2955, "lr": 2.3769433054618984e-07}
{"train_loss": 0.05182039737701416, "global_step": 263037, "epoch": 2955, "lr": 2.3763787226686042e-07}
{"train_loss": 0.03232228755950928, "global_step": 263038, "epoch": 2955, "lr": 2.3758142067747956e-07}
{"train_loss": 0.036161571741104126, "global_step": 263039, "epoch": 2955, "lr": 2.3752497577805288e-07}
{"train_loss": 0.029943546280264854, "global_step": 263040, "epoch": 2955, "lr": 2.374685375685748e-07}
{"train_loss": 0.09070921689271927, "global_step": 263041, "epoch": 2955, "lr": 2.374121060490675e-07}
{"train_loss": 0.03200242295861244, "global_step": 263042, "epoch": 2955, "lr": 2.3735568121953655e-07}
{"train_loss": 0.034336525946855545, "global_step": 263043, "epoch": 2955, "lr": 2.3729926307998195e-07}
{"train_loss": 0.026676993817090988, "global_step": 263044, "epoch": 2955, "lr": 2.3724285163041482e-07}
{"train_loss": 0.0781756043434143, "global_step": 263045, "epoch": 2955, "lr": 2.3718644687084624e-07}
{"train_loss": 0.02164224162697792, "global_step": 263046, "epoch": 2955, "lr": 2.3713004880128175e-07}
{"train_loss": 0.04622061923146248, "global_step": 263047, "epoch": 2955, "lr": 2.3707365742173249e-07}
{"train_loss": 0.018458815291523933, "global_step": 263048, "epoch": 2955, "lr": 2.3701727273219287e-07}
{"train_loss": 0.03266296908259392, "global_step": 263049, "epoch": 2955, "lr": 2.3696089473268513e-07}
{"train_loss": 0.08775431662797928, "global_step": 263050, "epoch": 2955, "lr": 2.3690452342321478e-07}
{"train_loss": 0.05037626624107361, "global_step": 263051, "epoch": 2955, "lr": 2.3684815880378185e-07}
{"train_loss": 0.045951779931783676, "global_step": 263052, "epoch": 2955, "lr": 2.3679180087439744e-07}
{"train_loss": 0.038231026381254196, "global_step": 263053, "epoch": 2955, "lr": 2.3673544963507266e-07}
{"train_loss": 0.03768295422196388, "global_step": 263054, "epoch": 2955, "lr": 2.3667910508581305e-07}
{"train_loss": 0.044758860021829605, "global_step": 263055, "epoch": 2955, "lr": 2.3662276722661857e-07}
{"train_loss": 0.01585046574473381, "global_step": 263056, "epoch": 2955, "lr": 2.365664360575115e-07}
{"train_loss": 0.006408921908587217, "global_step": 263057, "epoch": 2955, "lr": 2.3651011157848623e-07}
{"train_loss": 0.021353237330913544, "global_step": 263058, "epoch": 2955, "lr": 2.3645379378955946e-07}
{"train_loss": 0.0540245920419693, "global_step": 263059, "epoch": 2955, "lr": 2.3639748269073115e-07}
{"train_loss": 0.08363805711269379, "global_step": 263060, "epoch": 2955, "lr": 2.3634117828201795e-07}
{"train_loss": 0.061650145798921585, "global_step": 263061, "epoch": 2955, "lr": 2.3628488056341436e-07}
{"train_loss": 0.07931124418973923, "global_step": 263062, "epoch": 2955, "lr": 2.3622858953494253e-07}
{"train_loss": 0.06846926361322403, "global_step": 263063, "epoch": 2955, "lr": 2.361723051966025e-07}
{"train_loss": 0.01632757857441902, "global_step": 263064, "epoch": 2955, "lr": 2.361160275483998e-07}
{"train_loss": 0.042310379445552826, "global_step": 263065, "epoch": 2955, "lr": 2.3605975659034551e-07}
{"train_loss": 0.032811637967824936, "global_step": 263066, "epoch": 2955, "lr": 2.3600349232244523e-07}
{"train_loss": 0.040274374186992645, "global_step": 263067, "epoch": 2955, "lr": 2.3594723474471003e-07}
{"train_loss": 0.036286622285842896, "global_step": 263068, "epoch": 2955, "lr": 2.3589098385714548e-07}
{"train_loss": 0.02974342554807663, "global_step": 263069, "epoch": 2955, "lr": 2.358347396597571e-07}
{"train_loss": 0.035360146313905716, "global_step": 263070, "epoch": 2955, "lr": 2.3577850215255604e-07}
{"train_loss": 0.0538078136742115, "global_step": 263071, "epoch": 2955, "lr": 2.3572227133554782e-07}
{"train_loss": 0.06135695055127144, "global_step": 263072, "epoch": 2955, "lr": 2.3566604720873798e-07}
{"train_loss": 0.04475219547748566, "global_step": 263073, "epoch": 2955, "lr": 2.3560982977213762e-07}
{"train_loss": 0.025709230452775955, "global_step": 263074, "epoch": 2955, "lr": 2.355536190257579e-07}
{"train_loss": 0.06533645838499069, "global_step": 263075, "epoch": 2955, "lr": 2.3549741496959322e-07}
{"train_loss": 0.043501559644937515, "global_step": 263076, "epoch": 2955, "lr": 2.354412176036658e-07}
{"train_loss": 0.03651599586009979, "global_step": 263077, "epoch": 2955, "lr": 2.3538502692797004e-07}
{"train_loss": 0.017063748091459274, "global_step": 263078, "epoch": 2955, "lr": 2.3532884294252822e-07}
{"train_loss": 0.05341823026537895, "global_step": 263079, "epoch": 2955, "lr": 2.3527266564733473e-07}
{"train_loss": 0.06548991054296494, "global_step": 263080, "epoch": 2955, "lr": 2.3521649504240072e-07}
{"train_loss": 0.022930080071091652, "global_step": 263081, "epoch": 2955, "lr": 2.3516033112774282e-07}
{"train_loss": 0.06295551359653473, "global_step": 263082, "epoch": 2955, "lr": 2.3510417390334992e-07}
{"train_loss": 0.041726991068571806, "global_step": 263083, "epoch": 2955, "lr": 2.350480233692498e-07, "val_loss": 9.593533515930176, "train_action_mse_error": 24.647315979003906}
{"train_loss": 0.027282189577817917, "global_step": 263084, "epoch": 2956, "lr": 2.3499187952543688e-07}
{"train_loss": 0.017840847373008728, "global_step": 263085, "epoch": 2956, "lr": 2.3493574237192782e-07}
{"train_loss": 0.03435206785798073, "global_step": 263086, "epoch": 2956, "lr": 2.348796119087171e-07}
{"train_loss": 0.0711253210902214, "global_step": 263087, "epoch": 2956, "lr": 2.3482348813582134e-07}
{"train_loss": 0.027945619076490402, "global_step": 263088, "epoch": 2956, "lr": 2.3476737105325163e-07}
{"train_loss": 0.03671160712838173, "global_step": 263089, "epoch": 2956, "lr": 2.34711260661008e-07}
{"train_loss": 0.05039576068520546, "global_step": 263090, "epoch": 2956, "lr": 2.3465515695910712e-07}
{"train_loss": 0.04561620205640793, "global_step": 263091, "epoch": 2956, "lr": 2.3459905994754338e-07}
{"train_loss": 0.040330275893211365, "global_step": 263092, "epoch": 2956, "lr": 2.345429696263335e-07}
{"train_loss": 0.07136853039264679, "global_step": 263093, "epoch": 2956, "lr": 2.3448688599548297e-07}
{"train_loss": 0.024883100762963295, "global_step": 263094, "epoch": 2956, "lr": 2.3443080905499736e-07}
{"train_loss": 0.05308673158288002, "global_step": 263095, "epoch": 2956, "lr": 2.343747388048878e-07}
{"train_loss": 0.06705467402935028, "global_step": 263096, "epoch": 2956, "lr": 2.3431867524515982e-07}
{"train_loss": 0.09507425874471664, "global_step": 263097, "epoch": 2956, "lr": 2.3426261837581897e-07}
{"train_loss": 0.08141571283340454, "global_step": 263098, "epoch": 2956, "lr": 2.3420656819687637e-07}
{"train_loss": 0.03399987146258354, "global_step": 263099, "epoch": 2956, "lr": 2.3415052470833753e-07}
{"train_loss": 0.03411887586116791, "global_step": 263100, "epoch": 2956, "lr": 2.340944879102136e-07}
{"train_loss": 0.039985496550798416, "global_step": 263101, "epoch": 2956, "lr": 2.3403845780250454e-07}
{"train_loss": 0.021089892834424973, "global_step": 263102, "epoch": 2956, "lr": 2.3398243438522703e-07}
{"train_loss": 0.050404142588377, "global_step": 263103, "epoch": 2956, "lr": 2.3392641765838108e-07}
{"train_loss": 0.052292369306087494, "global_step": 263104, "epoch": 2956, "lr": 2.3387040762197776e-07}
{"train_loss": 0.02403923124074936, "global_step": 263105, "epoch": 2956, "lr": 2.3381440427602264e-07}
{"train_loss": 0.012995708733797073, "global_step": 263106, "epoch": 2956, "lr": 2.337584076205268e-07}
{"train_loss": 0.031073516234755516, "global_step": 263107, "epoch": 2956, "lr": 2.3370241765549584e-07}
{"train_loss": 0.04052488133311272, "global_step": 263108, "epoch": 2956, "lr": 2.3364643438093524e-07}
{"train_loss": 0.03319444879889488, "global_step": 263109, "epoch": 2956, "lr": 2.3359045779685618e-07}
{"train_loss": 0.03198718652129173, "global_step": 263110, "epoch": 2956, "lr": 2.3353448790326415e-07}
{"train_loss": 0.03341885656118393, "global_step": 263111, "epoch": 2956, "lr": 2.3347852470016473e-07}
{"train_loss": 0.0446128211915493, "global_step": 263112, "epoch": 2956, "lr": 2.3342256818756902e-07}
{"train_loss": 0.033522576093673706, "global_step": 263113, "epoch": 2956, "lr": 2.3336661836548256e-07}
{"train_loss": 0.0241562407463789, "global_step": 263114, "epoch": 2956, "lr": 2.3331067523391646e-07}
{"train_loss": 0.016077499836683273, "global_step": 263115, "epoch": 2956, "lr": 2.3325473879287073e-07}
{"train_loss": 0.07874026894569397, "global_step": 263116, "epoch": 2956, "lr": 2.3319880904236201e-07}
{"train_loss": 0.05523098260164261, "global_step": 263117, "epoch": 2956, "lr": 2.331428859823903e-07}
{"train_loss": 0.011586640030145645, "global_step": 263118, "epoch": 2956, "lr": 2.3308696961296672e-07}
{"train_loss": 0.047285255044698715, "global_step": 263119, "epoch": 2956, "lr": 2.3303105993409678e-07}
{"train_loss": 0.03512609750032425, "global_step": 263120, "epoch": 2956, "lr": 2.329751569457861e-07}
{"train_loss": 0.07542357593774796, "global_step": 263121, "epoch": 2956, "lr": 2.3291926064805126e-07}
{"train_loss": 0.06204541400074959, "global_step": 263122, "epoch": 2956, "lr": 2.328633710408923e-07}
{"train_loss": 0.07207246124744415, "global_step": 263123, "epoch": 2956, "lr": 2.3280748812431474e-07}
{"train_loss": 0.02807983197271824, "global_step": 263124, "epoch": 2956, "lr": 2.3275161189833528e-07}
{"train_loss": 0.02469719760119915, "global_step": 263125, "epoch": 2956, "lr": 2.3269574236294832e-07}
{"train_loss": 0.08563917130231857, "global_step": 263126, "epoch": 2956, "lr": 2.3263987951817612e-07}
{"train_loss": 0.029653415083885193, "global_step": 263127, "epoch": 2956, "lr": 2.325840233640131e-07}
{"train_loss": 0.038028791546821594, "global_step": 263128, "epoch": 2956, "lr": 2.3252817390047588e-07}
{"train_loss": 0.028562676161527634, "global_step": 263129, "epoch": 2956, "lr": 2.3247233112757005e-07}
{"train_loss": 0.03168480843305588, "global_step": 263130, "epoch": 2956, "lr": 2.3241649504530115e-07}
{"train_loss": 0.06995765119791031, "global_step": 263131, "epoch": 2956, "lr": 2.3236066565367476e-07}
{"train_loss": 0.040421947836875916, "global_step": 263132, "epoch": 2956, "lr": 2.3230484295270195e-07}
{"train_loss": 0.06074826419353485, "global_step": 263133, "epoch": 2956, "lr": 2.3224902694238826e-07}
{"train_loss": 0.03651992976665497, "global_step": 263134, "epoch": 2956, "lr": 2.3219321762274483e-07}
{"train_loss": 0.012980345636606216, "global_step": 263135, "epoch": 2956, "lr": 2.3213741499377717e-07}
{"train_loss": 0.0757177397608757, "global_step": 263136, "epoch": 2956, "lr": 2.3208161905549087e-07}
{"train_loss": 0.033391550183296204, "global_step": 263137, "epoch": 2956, "lr": 2.3202582980789145e-07}
{"train_loss": 0.05170372873544693, "global_step": 263138, "epoch": 2956, "lr": 2.3197004725099004e-07}
{"train_loss": 0.041353002190589905, "global_step": 263139, "epoch": 2956, "lr": 2.3191427138479772e-07}
{"train_loss": 0.037068333476781845, "global_step": 263140, "epoch": 2956, "lr": 2.318585022093145e-07}
{"train_loss": 0.05419672653079033, "global_step": 263141, "epoch": 2956, "lr": 2.3180273972455146e-07}
{"train_loss": 0.02989872172474861, "global_step": 263142, "epoch": 2956, "lr": 2.317469839305142e-07}
{"train_loss": 0.0343768335878849, "global_step": 263143, "epoch": 2956, "lr": 2.3169123482721932e-07}
{"train_loss": 0.035318728536367416, "global_step": 263144, "epoch": 2956, "lr": 2.316354924146613e-07}
{"train_loss": 0.035165078938007355, "global_step": 263145, "epoch": 2956, "lr": 2.3157975669285125e-07}
{"train_loss": 0.036815229803323746, "global_step": 263146, "epoch": 2956, "lr": 2.3152402766180582e-07}
{"train_loss": 0.020457357168197632, "global_step": 263147, "epoch": 2956, "lr": 2.3146830532151943e-07}
{"train_loss": 0.07954113185405731, "global_step": 263148, "epoch": 2956, "lr": 2.3141258967200873e-07}
{"train_loss": 0.050510894507169724, "global_step": 263149, "epoch": 2956, "lr": 2.3135688071327378e-07}
{"train_loss": 0.05075763538479805, "global_step": 263150, "epoch": 2956, "lr": 2.3130117844533117e-07}
{"train_loss": 0.0778096541762352, "global_step": 263151, "epoch": 2956, "lr": 2.3124548286818093e-07}
{"train_loss": 0.025330055505037308, "global_step": 263152, "epoch": 2956, "lr": 2.3118979398183415e-07}
{"train_loss": 0.047587111592292786, "global_step": 263153, "epoch": 2956, "lr": 2.311341117862964e-07}
{"train_loss": 0.038566336035728455, "global_step": 263154, "epoch": 2956, "lr": 2.3107843628157876e-07}
{"train_loss": 0.0482080914080143, "global_step": 263155, "epoch": 2956, "lr": 2.310227674676868e-07}
{"train_loss": 0.05238426476716995, "global_step": 263156, "epoch": 2956, "lr": 2.309671053446205e-07}
{"train_loss": 0.0577307790517807, "global_step": 263157, "epoch": 2956, "lr": 2.309114499124021e-07}
{"train_loss": 0.0247017964720726, "global_step": 263158, "epoch": 2956, "lr": 2.3085580117102602e-07}
{"train_loss": 0.053145632147789, "global_step": 263159, "epoch": 2956, "lr": 2.3080015912050335e-07}
{"train_loss": 0.0494244284927845, "global_step": 263160, "epoch": 2956, "lr": 2.307445237608452e-07}
{"train_loss": 0.03485759720206261, "global_step": 263161, "epoch": 2956, "lr": 2.3068889509205715e-07}
{"train_loss": 0.04083016887307167, "global_step": 263162, "epoch": 2956, "lr": 2.3063327311414474e-07}
{"train_loss": 0.06927566975355148, "global_step": 263163, "epoch": 2956, "lr": 2.3057765782711903e-07}
{"train_loss": 0.043833911418914795, "global_step": 263164, "epoch": 2956, "lr": 2.3052204923098564e-07}
{"train_loss": 0.0414905846118927, "global_step": 263165, "epoch": 2956, "lr": 2.3046644732575008e-07}
{"train_loss": 0.015394194982945919, "global_step": 263166, "epoch": 2956, "lr": 2.3041085211142343e-07}
{"train_loss": 0.02953200601041317, "global_step": 263167, "epoch": 2956, "lr": 2.303552635880113e-07}
{"train_loss": 0.07372554391622543, "global_step": 263168, "epoch": 2956, "lr": 2.302996817555192e-07}
{"train_loss": 0.05350693315267563, "global_step": 263169, "epoch": 2956, "lr": 2.3024410661395822e-07}
{"train_loss": 0.0370083749294281, "global_step": 263170, "epoch": 2956, "lr": 2.3018853816333396e-07}
{"train_loss": 0.04040946066379547, "global_step": 263171, "epoch": 2956, "lr": 2.301329764036575e-07}
{"train_loss": 0.04372576363594009, "global_step": 263172, "epoch": 2956, "lr": 2.300774213349288e-07, "val_loss": 9.61870288848877}
{"train_loss": 0.030886348336935043, "global_step": 263173, "epoch": 2957, "lr": 2.3002187295715904e-07}
{"train_loss": 0.06315198540687561, "global_step": 263174, "epoch": 2957, "lr": 2.2996633127035928e-07}
{"train_loss": 0.027834421023726463, "global_step": 263175, "epoch": 2957, "lr": 2.2991079627453505e-07}
{"train_loss": 0.04072089120745659, "global_step": 263176, "epoch": 2957, "lr": 2.2985526796968637e-07}
{"train_loss": 0.04568679630756378, "global_step": 263177, "epoch": 2957, "lr": 2.2979974635583546e-07}
{"train_loss": 0.037379901856184006, "global_step": 263178, "epoch": 2957, "lr": 2.2974423143297673e-07}
{"train_loss": 0.034357063472270966, "global_step": 263179, "epoch": 2957, "lr": 2.2968872320112134e-07}
{"train_loss": 0.06067074462771416, "global_step": 263180, "epoch": 2957, "lr": 2.2963322166028035e-07}
{"train_loss": 0.048408493399620056, "global_step": 263181, "epoch": 2957, "lr": 2.2957772681045931e-07}
{"train_loss": 0.035188570618629456, "global_step": 263182, "epoch": 2957, "lr": 2.2952223865165823e-07}
{"train_loss": 0.041564635932445526, "global_step": 263183, "epoch": 2957, "lr": 2.2946675718389931e-07}
{"train_loss": 0.06348546594381332, "global_step": 263184, "epoch": 2957, "lr": 2.2941128240717703e-07}
{"train_loss": 0.04125982150435448, "global_step": 263185, "epoch": 2957, "lr": 2.29355814321508e-07}
{"train_loss": 0.021116288378834724, "global_step": 263186, "epoch": 2957, "lr": 2.2930035292689222e-07}
{"train_loss": 0.03608856350183487, "global_step": 263187, "epoch": 2957, "lr": 2.2924489822334084e-07}
{"train_loss": 0.041105303913354874, "global_step": 263188, "epoch": 2957, "lr": 2.2918945021085937e-07}
{"train_loss": 0.0629940778017044, "global_step": 263189, "epoch": 2957, "lr": 2.2913400888945892e-07}
{"train_loss": 0.028713103383779526, "global_step": 263190, "epoch": 2957, "lr": 2.2907857425914502e-07}
{"train_loss": 0.05367652326822281, "global_step": 263191, "epoch": 2957, "lr": 2.2902314631992328e-07}
{"train_loss": 0.07602623105049133, "global_step": 263192, "epoch": 2957, "lr": 2.2896772507180475e-07}
{"train_loss": 0.02536897175014019, "global_step": 263193, "epoch": 2957, "lr": 2.28912310514795e-07}
{"train_loss": 0.03406663239002228, "global_step": 263194, "epoch": 2957, "lr": 2.288569026488996e-07}
{"train_loss": 0.07429590076208115, "global_step": 263195, "epoch": 2957, "lr": 2.2880150147412405e-07}
{"train_loss": 0.035787392407655716, "global_step": 263196, "epoch": 2957, "lr": 2.2874610699048503e-07}
{"train_loss": 0.03315923735499382, "global_step": 263197, "epoch": 2957, "lr": 2.2869071919798258e-07}
{"train_loss": 0.013910279609262943, "global_step": 263198, "epoch": 2957, "lr": 2.2863533809662774e-07}
{"train_loss": 0.07035556435585022, "global_step": 263199, "epoch": 2957, "lr": 2.285799636864261e-07}
{"train_loss": 0.06948263943195343, "global_step": 263200, "epoch": 2957, "lr": 2.2852459596738318e-07}
{"train_loss": 0.05485553294420242, "global_step": 263201, "epoch": 2957, "lr": 2.284692349395101e-07}
{"train_loss": 0.05104384198784828, "global_step": 263202, "epoch": 2957, "lr": 2.2841388060281245e-07}
{"train_loss": 0.03581727296113968, "global_step": 263203, "epoch": 2957, "lr": 2.283585329572957e-07}
{"train_loss": 0.07340382039546967, "global_step": 263204, "epoch": 2957, "lr": 2.2830319200297656e-07}
{"train_loss": 0.029650205746293068, "global_step": 263205, "epoch": 2957, "lr": 2.2824785773984948e-07}
{"train_loss": 0.09706024080514908, "global_step": 263206, "epoch": 2957, "lr": 2.2819253016792553e-07}
{"train_loss": 0.06215192377567291, "global_step": 263207, "epoch": 2957, "lr": 2.2813720928722137e-07}
{"train_loss": 0.04703158140182495, "global_step": 263208, "epoch": 2957, "lr": 2.280818950977315e-07}
{"train_loss": 0.029322145506739616, "global_step": 263209, "epoch": 2957, "lr": 2.2802658759947247e-07}
{"train_loss": 0.06641726195812225, "global_step": 263210, "epoch": 2957, "lr": 2.2797128679244993e-07}
{"train_loss": 0.05511889234185219, "global_step": 263211, "epoch": 2957, "lr": 2.2791599267666385e-07}
{"train_loss": 0.01439522672444582, "global_step": 263212, "epoch": 2957, "lr": 2.2786070525213642e-07}
{"train_loss": 0.03367219492793083, "global_step": 263213, "epoch": 2957, "lr": 2.2780542451885657e-07}
{"train_loss": 0.06766737997531891, "global_step": 263214, "epoch": 2957, "lr": 2.2775015047684644e-07}
{"train_loss": 0.024587005376815796, "global_step": 263215, "epoch": 2957, "lr": 2.2769488312611164e-07}
{"train_loss": 0.022470755502581596, "global_step": 263216, "epoch": 2957, "lr": 2.2763962246665215e-07}
{"train_loss": 0.03740197420120239, "global_step": 263217, "epoch": 2957, "lr": 2.2758436849848463e-07}
{"train_loss": 0.019942810758948326, "global_step": 263218, "epoch": 2957, "lr": 2.2752912122160909e-07}
{"train_loss": 0.019980037584900856, "global_step": 263219, "epoch": 2957, "lr": 2.274738806360366e-07}
{"train_loss": 0.04860696569085121, "global_step": 263220, "epoch": 2957, "lr": 2.274186467417727e-07}
{"train_loss": 0.02249530516564846, "global_step": 263221, "epoch": 2957, "lr": 2.27363419538823e-07}
{"train_loss": 0.05814255401492119, "global_step": 263222, "epoch": 2957, "lr": 2.2730819902719857e-07}
{"train_loss": 0.0366012379527092, "global_step": 263223, "epoch": 2957, "lr": 2.272529852069105e-07}
{"train_loss": 0.04279855266213417, "global_step": 263224, "epoch": 2957, "lr": 2.2719777807795883e-07}
{"train_loss": 0.03398776426911354, "global_step": 263225, "epoch": 2957, "lr": 2.2714257764035462e-07}
{"train_loss": 0.04818875342607498, "global_step": 263226, "epoch": 2957, "lr": 2.2708738389410346e-07}
{"train_loss": 0.04309052973985672, "global_step": 263227, "epoch": 2957, "lr": 2.2703219683921085e-07}
{"train_loss": 0.04365255683660507, "global_step": 263228, "epoch": 2957, "lr": 2.2697701647569348e-07}
{"train_loss": 0.041368741542100906, "global_step": 263229, "epoch": 2957, "lr": 2.2692184280354578e-07}
{"train_loss": 0.03284216299653053, "global_step": 263230, "epoch": 2957, "lr": 2.2686667582278997e-07}
{"train_loss": 0.07354807108640671, "global_step": 263231, "epoch": 2957, "lr": 2.2681151553341494e-07}
{"train_loss": 0.06315838545560837, "global_step": 263232, "epoch": 2957, "lr": 2.2675636193544848e-07}
{"train_loss": 0.05798422545194626, "global_step": 263233, "epoch": 2957, "lr": 2.2670121502887943e-07}
{"train_loss": 0.04131041467189789, "global_step": 263234, "epoch": 2957, "lr": 2.2664607481373e-07}
{"train_loss": 0.0781986266374588, "global_step": 263235, "epoch": 2957, "lr": 2.2659094129000025e-07}
{"train_loss": 0.026006445288658142, "global_step": 263236, "epoch": 2957, "lr": 2.2653581445769568e-07}
{"train_loss": 0.036104921251535416, "global_step": 263237, "epoch": 2957, "lr": 2.2648069431682738e-07}
{"train_loss": 0.054927755147218704, "global_step": 263238, "epoch": 2957, "lr": 2.264255808674065e-07}
{"train_loss": 0.01427626796066761, "global_step": 263239, "epoch": 2957, "lr": 2.26370474109433e-07}
{"train_loss": 0.06624449044466019, "global_step": 263240, "epoch": 2957, "lr": 2.2631537404291803e-07}
{"train_loss": 0.08932948112487793, "global_step": 263241, "epoch": 2957, "lr": 2.2626028066786152e-07}
{"train_loss": 0.031517393887043, "global_step": 263242, "epoch": 2957, "lr": 2.2620519398428573e-07}
{"train_loss": 0.045572783797979355, "global_step": 263243, "epoch": 2957, "lr": 2.2615011399219066e-07}
{"train_loss": 0.0321979746222496, "global_step": 263244, "epoch": 2957, "lr": 2.2609504069157628e-07}
{"train_loss": 0.012768295593559742, "global_step": 263245, "epoch": 2957, "lr": 2.260399740824648e-07}
{"train_loss": 0.033354293555021286, "global_step": 263246, "epoch": 2957, "lr": 2.259849141648507e-07}
{"train_loss": 0.020897939801216125, "global_step": 263247, "epoch": 2957, "lr": 2.2592986093874502e-07}
{"train_loss": 0.03526362031698227, "global_step": 263248, "epoch": 2957, "lr": 2.2587481440415893e-07}
{"train_loss": 0.07827005535364151, "global_step": 263249, "epoch": 2957, "lr": 2.2581977456109794e-07}
{"train_loss": 0.039659008383750916, "global_step": 263250, "epoch": 2957, "lr": 2.2576474140956205e-07}
{"train_loss": 0.05138964578509331, "global_step": 263251, "epoch": 2957, "lr": 2.2570971494957348e-07}
{"train_loss": 0.038968075066804886, "global_step": 263252, "epoch": 2957, "lr": 2.256546951811267e-07}
{"train_loss": 0.06191069260239601, "global_step": 263253, "epoch": 2957, "lr": 2.2559968210423833e-07}
{"train_loss": 0.042713433504104614, "global_step": 263254, "epoch": 2957, "lr": 2.2554467571890837e-07}
{"train_loss": 0.035361148416996, "global_step": 263255, "epoch": 2957, "lr": 2.2548967602515347e-07}
{"train_loss": 0.03999193757772446, "global_step": 263256, "epoch": 2957, "lr": 2.2543468302296811e-07}
{"train_loss": 0.04803762212395668, "global_step": 263257, "epoch": 2957, "lr": 2.253796967123689e-07}
{"train_loss": 0.02108651213347912, "global_step": 263258, "epoch": 2957, "lr": 2.253247170933559e-07}
{"train_loss": 0.02334265038371086, "global_step": 263259, "epoch": 2957, "lr": 2.252697441659457e-07}
{"train_loss": 0.026215972378849983, "global_step": 263260, "epoch": 2957, "lr": 2.252147779301439e-07}
{"train_loss": 0.043966313499664324, "global_step": 263261, "epoch": 2957, "lr": 2.2515981838595047e-07, "val_loss": 9.64611530303955}
{"train_loss": 0.02613312005996704, "global_step": 263262, "epoch": 2958, "lr": 2.251048655333765e-07}
{"train_loss": 0.043437957763671875, "global_step": 263263, "epoch": 2958, "lr": 2.2504991937243868e-07}
{"train_loss": 0.05470738187432289, "global_step": 263264, "epoch": 2958, "lr": 2.249949799031259e-07}
{"train_loss": 0.06124446913599968, "global_step": 263265, "epoch": 2958, "lr": 2.249400471254659e-07}
{"train_loss": 0.019706565886735916, "global_step": 263266, "epoch": 2958, "lr": 2.2488512103945313e-07}
{"train_loss": 0.027534617111086845, "global_step": 263267, "epoch": 2958, "lr": 2.2483020164509317e-07}
{"train_loss": 0.049697257578372955, "global_step": 263268, "epoch": 2958, "lr": 2.2477528894240262e-07}
{"train_loss": 0.05951223522424698, "global_step": 263269, "epoch": 2958, "lr": 2.2472038293138155e-07}
{"train_loss": 0.029736004769802094, "global_step": 263270, "epoch": 2958, "lr": 2.2466548361204098e-07}
{"train_loss": 0.05664650350809097, "global_step": 263271, "epoch": 2958, "lr": 2.2461059098438653e-07}
{"train_loss": 0.03171835467219353, "global_step": 263272, "epoch": 2958, "lr": 2.2455570504842927e-07}
{"train_loss": 0.03862718120217323, "global_step": 263273, "epoch": 2958, "lr": 2.245008258041692e-07}
{"train_loss": 0.06525171548128128, "global_step": 263274, "epoch": 2958, "lr": 2.2444595325162298e-07}
{"train_loss": 0.05155402421951294, "global_step": 263275, "epoch": 2958, "lr": 2.2439108739079063e-07}
{"train_loss": 0.049880146980285645, "global_step": 263276, "epoch": 2958, "lr": 2.243362282216832e-07}
{"train_loss": 0.02575133740901947, "global_step": 263277, "epoch": 2958, "lr": 2.242813757443063e-07}
{"train_loss": 0.019304875284433365, "global_step": 263278, "epoch": 2958, "lr": 2.24226529958671e-07}
{"train_loss": 0.05270117521286011, "global_step": 263279, "epoch": 2958, "lr": 2.241716908647773e-07}
{"train_loss": 0.05064576864242554, "global_step": 263280, "epoch": 2958, "lr": 2.2411685846264186e-07}
{"train_loss": 0.057298898696899414, "global_step": 263281, "epoch": 2958, "lr": 2.2406203275226467e-07}
{"train_loss": 0.030925214290618896, "global_step": 263282, "epoch": 2958, "lr": 2.2400721373365685e-07}
{"train_loss": 0.02775799110531807, "global_step": 263283, "epoch": 2958, "lr": 2.239524014068184e-07}
{"train_loss": 0.03468553349375725, "global_step": 263284, "epoch": 2958, "lr": 2.238975957717715e-07}
{"train_loss": 0.03382669761776924, "global_step": 263285, "epoch": 2958, "lr": 2.2384279682851062e-07}
{"train_loss": 0.035021547228097916, "global_step": 263286, "epoch": 2958, "lr": 2.2378800457704684e-07}
{"train_loss": 0.05236782878637314, "global_step": 263287, "epoch": 2958, "lr": 2.2373321901739132e-07}
{"train_loss": 0.0535924956202507, "global_step": 263288, "epoch": 2958, "lr": 2.2367844014954397e-07}
{"train_loss": 0.06235054135322571, "global_step": 263289, "epoch": 2958, "lr": 2.2362366797352152e-07}
{"train_loss": 0.02530193142592907, "global_step": 263290, "epoch": 2958, "lr": 2.2356890248931838e-07}
{"train_loss": 0.029694141820073128, "global_step": 263291, "epoch": 2958, "lr": 2.2351414369695678e-07}
{"train_loss": 0.035740841180086136, "global_step": 263292, "epoch": 2958, "lr": 2.234593915964367e-07}
{"train_loss": 0.029074540361762047, "global_step": 263293, "epoch": 2958, "lr": 2.234046461877637e-07}
{"train_loss": 0.03132384270429611, "global_step": 263294, "epoch": 2958, "lr": 2.233499074709433e-07}
{"train_loss": 0.09748730808496475, "global_step": 263295, "epoch": 2958, "lr": 2.2329517544599222e-07}
{"train_loss": 0.0630173459649086, "global_step": 263296, "epoch": 2958, "lr": 2.232404501129104e-07}
{"train_loss": 0.05555734038352966, "global_step": 263297, "epoch": 2958, "lr": 2.2318573147170897e-07}
{"train_loss": 0.06318734586238861, "global_step": 263298, "epoch": 2958, "lr": 2.2313101952239347e-07}
{"train_loss": 0.0269879549741745, "global_step": 263299, "epoch": 2958, "lr": 2.2307631426496943e-07}
{"train_loss": 0.022065358236432076, "global_step": 263300, "epoch": 2958, "lr": 2.2302161569944802e-07}
{"train_loss": 0.04708927124738693, "global_step": 263301, "epoch": 2958, "lr": 2.2296692382583473e-07}
{"train_loss": 0.013553984463214874, "global_step": 263302, "epoch": 2958, "lr": 2.2291223864413512e-07}
{"train_loss": 0.029025554656982422, "global_step": 263303, "epoch": 2958, "lr": 2.2285756015436033e-07}
{"train_loss": 0.019917162135243416, "global_step": 263304, "epoch": 2958, "lr": 2.2280288835651032e-07}
{"train_loss": 0.07326118648052216, "global_step": 263305, "epoch": 2958, "lr": 2.2274822325060173e-07}
{"train_loss": 0.0680081844329834, "global_step": 263306, "epoch": 2958, "lr": 2.2269356483664017e-07}
{"train_loss": 0.04686194285750389, "global_step": 263307, "epoch": 2958, "lr": 2.226389131146256e-07}
{"train_loss": 0.06578202545642853, "global_step": 263308, "epoch": 2958, "lr": 2.2258426808457467e-07}
{"train_loss": 0.021095728501677513, "global_step": 263309, "epoch": 2958, "lr": 2.2252962974649294e-07}
{"train_loss": 0.04554591700434685, "global_step": 263310, "epoch": 2958, "lr": 2.2247499810038042e-07}
{"train_loss": 0.0585830993950367, "global_step": 263311, "epoch": 2958, "lr": 2.2242037314624818e-07}
{"train_loss": 0.05354178696870804, "global_step": 263312, "epoch": 2958, "lr": 2.2236575488410737e-07}
{"train_loss": 0.05607302486896515, "global_step": 263313, "epoch": 2958, "lr": 2.223111433139635e-07}
{"train_loss": 0.01723574474453926, "global_step": 263314, "epoch": 2958, "lr": 2.2225653843582216e-07}
{"train_loss": 0.048919208347797394, "global_step": 263315, "epoch": 2958, "lr": 2.2220194024969442e-07}
{"train_loss": 0.0526781789958477, "global_step": 263316, "epoch": 2958, "lr": 2.221473487555803e-07}
{"train_loss": 0.045787833631038666, "global_step": 263317, "epoch": 2958, "lr": 2.2209276395349643e-07}
{"train_loss": 0.05121386796236038, "global_step": 263318, "epoch": 2958, "lr": 2.2203818584344283e-07}
{"train_loss": 0.04694744572043419, "global_step": 263319, "epoch": 2958, "lr": 2.219836144254306e-07}
{"train_loss": 0.046321868896484375, "global_step": 263320, "epoch": 2958, "lr": 2.2192904969946527e-07}
{"train_loss": 0.06272319704294205, "global_step": 263321, "epoch": 2958, "lr": 2.218744916655524e-07}
{"train_loss": 0.04665791243314743, "global_step": 263322, "epoch": 2958, "lr": 2.2181994032370312e-07}
{"train_loss": 0.05539445951581001, "global_step": 263323, "epoch": 2958, "lr": 2.2176539567392295e-07}
{"train_loss": 0.06605639308691025, "global_step": 263324, "epoch": 2958, "lr": 2.2171085771622303e-07}
{"train_loss": 0.03728540241718292, "global_step": 263325, "epoch": 2958, "lr": 2.216563264506033e-07}
{"train_loss": 0.04046765714883804, "global_step": 263326, "epoch": 2958, "lr": 2.2160180187708046e-07}
{"train_loss": 0.046895481646060944, "global_step": 263327, "epoch": 2958, "lr": 2.215472839956545e-07}
{"train_loss": 0.04701405018568039, "global_step": 263328, "epoch": 2958, "lr": 2.2149277280633095e-07}
{"train_loss": 0.05594146251678467, "global_step": 263329, "epoch": 2958, "lr": 2.214382683091265e-07}
{"train_loss": 0.03190109133720398, "global_step": 263330, "epoch": 2958, "lr": 2.213837705040356e-07}
{"train_loss": 0.01709764637053013, "global_step": 263331, "epoch": 2958, "lr": 2.213292793910804e-07}
{"train_loss": 0.04174603894352913, "global_step": 263332, "epoch": 2958, "lr": 2.2127479497025538e-07}
{"train_loss": 0.05786662921309471, "global_step": 263333, "epoch": 2958, "lr": 2.2122031724157721e-07}
{"train_loss": 0.04437597468495369, "global_step": 263334, "epoch": 2958, "lr": 2.2116584620504588e-07}
{"train_loss": 0.03698647394776344, "global_step": 263335, "epoch": 2958, "lr": 2.2111138186067805e-07}
{"train_loss": 0.04854894429445267, "global_step": 263336, "epoch": 2958, "lr": 2.2105692420846813e-07}
{"train_loss": 0.032099731266498566, "global_step": 263337, "epoch": 2958, "lr": 2.210024732484328e-07}
{"train_loss": 0.033334337174892426, "global_step": 263338, "epoch": 2958, "lr": 2.2094802898057764e-07}
{"train_loss": 0.08047804236412048, "global_step": 263339, "epoch": 2958, "lr": 2.2089359140491372e-07}
{"train_loss": 0.05414487421512604, "global_step": 263340, "epoch": 2958, "lr": 2.208391605214355e-07}
{"train_loss": 0.048466458916664124, "global_step": 263341, "epoch": 2958, "lr": 2.2078473633016515e-07}
{"train_loss": 0.09284744411706924, "global_step": 263342, "epoch": 2958, "lr": 2.207303188310972e-07}
{"train_loss": 0.06814217567443848, "global_step": 263343, "epoch": 2958, "lr": 2.2067590802425375e-07}
{"train_loss": 0.011259066872298717, "global_step": 263344, "epoch": 2958, "lr": 2.2062150390962932e-07}
{"train_loss": 0.028257174417376518, "global_step": 263345, "epoch": 2958, "lr": 2.20567106487235e-07}
{"train_loss": 0.047751326113939285, "global_step": 263346, "epoch": 2958, "lr": 2.2051271575707634e-07}
{"train_loss": 0.03255847096443176, "global_step": 263347, "epoch": 2958, "lr": 2.2045833171916441e-07}
{"train_loss": 0.0557207316160202, "global_step": 263348, "epoch": 2958, "lr": 2.2040395437351036e-07}
{"train_loss": 0.024259790778160095, "global_step": 263349, "epoch": 2958, "lr": 2.2034958372010861e-07}
{"train_loss": 0.04438442232484898, "global_step": 263350, "epoch": 2958, "lr": 2.2029521975898139e-07, "val_loss": 9.673853874206543}
{"train_loss": 0.011441090144217014, "global_step": 263351, "epoch": 2959, "lr": 2.202408624901231e-07}
{"train_loss": 0.0866466835141182, "global_step": 263352, "epoch": 2959, "lr": 2.2018651191355043e-07}
{"train_loss": 0.03729572147130966, "global_step": 263353, "epoch": 2959, "lr": 2.201321680292634e-07}
{"train_loss": 0.036990053951740265, "global_step": 263354, "epoch": 2959, "lr": 2.200778308372786e-07}
{"train_loss": 0.04947541281580925, "global_step": 263355, "epoch": 2959, "lr": 2.2002350033759056e-07}
{"train_loss": 0.045755535364151, "global_step": 263356, "epoch": 2959, "lr": 2.1996917653022143e-07}
{"train_loss": 0.06602153927087784, "global_step": 263357, "epoch": 2959, "lr": 2.199148594151601e-07}
{"train_loss": 0.06875315308570862, "global_step": 263358, "epoch": 2959, "lr": 2.1986054899243435e-07}
{"train_loss": 0.06769579648971558, "global_step": 263359, "epoch": 2959, "lr": 2.1980624526203863e-07}
{"train_loss": 0.030652491375803947, "global_step": 263360, "epoch": 2959, "lr": 2.1975194822397848e-07}
{"train_loss": 0.014049334451556206, "global_step": 263361, "epoch": 2959, "lr": 2.1969765787827056e-07}
{"train_loss": 0.06880509108304977, "global_step": 263362, "epoch": 2959, "lr": 2.1964337422491487e-07}
{"train_loss": 0.023075636476278305, "global_step": 263363, "epoch": 2959, "lr": 2.1958909726392806e-07}
{"train_loss": 0.036247219890356064, "global_step": 263364, "epoch": 2959, "lr": 2.195348269953046e-07}
{"train_loss": 0.028494393453001976, "global_step": 263365, "epoch": 2959, "lr": 2.194805634190611e-07}
{"train_loss": 0.040102824568748474, "global_step": 263366, "epoch": 2959, "lr": 2.1942630653519758e-07}
{"train_loss": 0.021300045773386955, "global_step": 263367, "epoch": 2959, "lr": 2.1937205634373071e-07}
{"train_loss": 0.07307226955890656, "global_step": 263368, "epoch": 2959, "lr": 2.1931781284465491e-07}
{"train_loss": 0.02858457714319229, "global_step": 263369, "epoch": 2959, "lr": 2.192635760379924e-07}
{"train_loss": 0.0611763671040535, "global_step": 263370, "epoch": 2959, "lr": 2.1920934592373765e-07}
{"train_loss": 0.042759839445352554, "global_step": 263371, "epoch": 2959, "lr": 2.1915512250191284e-07}
{"train_loss": 0.014383266679942608, "global_step": 263372, "epoch": 2959, "lr": 2.1910090577250686e-07}
{"train_loss": 0.023364730179309845, "global_step": 263373, "epoch": 2959, "lr": 2.1904669573553638e-07}
{"train_loss": 0.022297604009509087, "global_step": 263374, "epoch": 2959, "lr": 2.189924923910125e-07}
{"train_loss": 0.03250626474618912, "global_step": 263375, "epoch": 2959, "lr": 2.189382957389352e-07}
{"train_loss": 0.02391086146235466, "global_step": 263376, "epoch": 2959, "lr": 2.1888410577931563e-07}
{"train_loss": 0.022232137620449066, "global_step": 263377, "epoch": 2959, "lr": 2.1882992251216484e-07}
{"train_loss": 0.05219943821430206, "global_step": 263378, "epoch": 2959, "lr": 2.187757459374773e-07}
{"train_loss": 0.02545889839529991, "global_step": 263379, "epoch": 2959, "lr": 2.1872157605527521e-07}
{"train_loss": 0.022295912727713585, "global_step": 263380, "epoch": 2959, "lr": 2.1866741286555858e-07}
{"train_loss": 0.042582910507917404, "global_step": 263381, "epoch": 2959, "lr": 2.1861325636833297e-07}
{"train_loss": 0.10031574219465256, "global_step": 263382, "epoch": 2959, "lr": 2.1855910656360946e-07}
{"train_loss": 0.04445876181125641, "global_step": 263383, "epoch": 2959, "lr": 2.185049634513936e-07}
{"train_loss": 0.03831038251519203, "global_step": 263384, "epoch": 2959, "lr": 2.1845082703169095e-07}
{"train_loss": 0.04044486954808235, "global_step": 263385, "epoch": 2959, "lr": 2.183966973045126e-07}
{"train_loss": 0.018402669578790665, "global_step": 263386, "epoch": 2959, "lr": 2.1834257426986414e-07}
{"train_loss": 0.025791993364691734, "global_step": 263387, "epoch": 2959, "lr": 2.182884579277511e-07}
{"train_loss": 0.07222584635019302, "global_step": 263388, "epoch": 2959, "lr": 2.1823434827819012e-07}
{"train_loss": 0.026281891390681267, "global_step": 263389, "epoch": 2959, "lr": 2.181802453211701e-07}
{"train_loss": 0.03155341371893883, "global_step": 263390, "epoch": 2959, "lr": 2.1812614905671324e-07}
{"train_loss": 0.059730615466833115, "global_step": 263391, "epoch": 2959, "lr": 2.180720594848251e-07}
{"train_loss": 0.03613830730319023, "global_step": 263392, "epoch": 2959, "lr": 2.1801797660550572e-07}
{"train_loss": 0.015734074637293816, "global_step": 263393, "epoch": 2959, "lr": 2.1796390041877167e-07}
{"train_loss": 0.05273953825235367, "global_step": 263394, "epoch": 2959, "lr": 2.1790983092462303e-07}
{"train_loss": 0.03332889825105667, "global_step": 263395, "epoch": 2959, "lr": 2.1785576812307084e-07}
{"train_loss": 0.06059609353542328, "global_step": 263396, "epoch": 2959, "lr": 2.178017120141207e-07}
{"train_loss": 0.04304639995098114, "global_step": 263397, "epoch": 2959, "lr": 2.1774766259777811e-07}
{"train_loss": 0.029603270813822746, "global_step": 263398, "epoch": 2959, "lr": 2.1769361987405424e-07}
{"train_loss": 0.03562936931848526, "global_step": 263399, "epoch": 2959, "lr": 2.1763958384295457e-07}
{"train_loss": 0.03292575851082802, "global_step": 263400, "epoch": 2959, "lr": 2.1758555450449025e-07}
{"train_loss": 0.030687646940350533, "global_step": 263401, "epoch": 2959, "lr": 2.1753153185866127e-07}
{"train_loss": 0.0677804946899414, "global_step": 263402, "epoch": 2959, "lr": 2.1747751590547872e-07}
{"train_loss": 0.060888927429914474, "global_step": 263403, "epoch": 2959, "lr": 2.1742350664494815e-07}
{"train_loss": 0.02056203782558441, "global_step": 263404, "epoch": 2959, "lr": 2.1736950407708068e-07}
{"train_loss": 0.07367206364870071, "global_step": 263405, "epoch": 2959, "lr": 2.1731550820188183e-07}
{"train_loss": 0.05976775661110878, "global_step": 263406, "epoch": 2959, "lr": 2.1726151901935166e-07}
{"train_loss": 0.09186288714408875, "global_step": 263407, "epoch": 2959, "lr": 2.172075365295123e-07}
{"train_loss": 0.04962032288312912, "global_step": 263408, "epoch": 2959, "lr": 2.1715356073235827e-07}
{"train_loss": 0.05263563618063927, "global_step": 263409, "epoch": 2959, "lr": 2.170995916279006e-07}
{"train_loss": 0.026532795280218124, "global_step": 263410, "epoch": 2959, "lr": 2.1704562921615045e-07}
{"train_loss": 0.019346699118614197, "global_step": 263411, "epoch": 2959, "lr": 2.169916734971078e-07}
{"train_loss": 0.043380338698625565, "global_step": 263412, "epoch": 2959, "lr": 2.1693772447078376e-07}
{"train_loss": 0.02838600054383278, "global_step": 263413, "epoch": 2959, "lr": 2.168837821371894e-07}
{"train_loss": 0.022002045065164566, "global_step": 263414, "epoch": 2959, "lr": 2.1682984649632477e-07}
{"train_loss": 0.034544311463832855, "global_step": 263415, "epoch": 2959, "lr": 2.1677591754820647e-07}
{"train_loss": 0.0423562154173851, "global_step": 263416, "epoch": 2959, "lr": 2.1672199529282899e-07}
{"train_loss": 0.03385016322135925, "global_step": 263417, "epoch": 2959, "lr": 2.1666807973020899e-07}
{"train_loss": 0.020080441609025, "global_step": 263418, "epoch": 2959, "lr": 2.1661417086035195e-07}
{"train_loss": 0.038227379322052, "global_step": 263419, "epoch": 2959, "lr": 2.1656026868326352e-07}
{"train_loss": 0.06661348044872284, "global_step": 263420, "epoch": 2959, "lr": 2.1650637319894916e-07}
{"train_loss": 0.03721361979842186, "global_step": 263421, "epoch": 2959, "lr": 2.164524844074256e-07}
{"train_loss": 0.039836399257183075, "global_step": 263422, "epoch": 2959, "lr": 2.163986023086928e-07}
{"train_loss": 0.0769822746515274, "global_step": 263423, "epoch": 2959, "lr": 2.1634472690275076e-07}
{"train_loss": 0.028950084000825882, "global_step": 263424, "epoch": 2959, "lr": 2.162908581896217e-07}
{"train_loss": 0.04522581398487091, "global_step": 263425, "epoch": 2959, "lr": 2.1623699616930004e-07}
{"train_loss": 0.03666599467396736, "global_step": 263426, "epoch": 2959, "lr": 2.1618314084180247e-07}
{"train_loss": 0.020732510834932327, "global_step": 263427, "epoch": 2959, "lr": 2.1612929220712895e-07}
{"train_loss": 0.052864424884319305, "global_step": 263428, "epoch": 2959, "lr": 2.1607545026529064e-07}
{"train_loss": 0.0521526075899601, "global_step": 263429, "epoch": 2959, "lr": 2.160216150162986e-07}
{"train_loss": 0.0025433721020817757, "global_step": 263430, "epoch": 2959, "lr": 2.1596778646014727e-07}
{"train_loss": 0.06934936344623566, "global_step": 263431, "epoch": 2959, "lr": 2.1591396459685887e-07}
{"train_loss": 0.04216672480106354, "global_step": 263432, "epoch": 2959, "lr": 2.158601494264334e-07}
{"train_loss": 0.06990444660186768, "global_step": 263433, "epoch": 2959, "lr": 2.158063409488764e-07}
{"train_loss": 0.04888768121600151, "global_step": 263434, "epoch": 2959, "lr": 2.15752539164199e-07}
{"train_loss": 0.04919177666306496, "global_step": 263435, "epoch": 2959, "lr": 2.1569874407240676e-07}
{"train_loss": 0.07551296055316925, "global_step": 263436, "epoch": 2959, "lr": 2.1564495567351072e-07}
{"train_loss": 0.03893289342522621, "global_step": 263437, "epoch": 2959, "lr": 2.1559117396751095e-07}
{"train_loss": 0.04312746971845627, "global_step": 263438, "epoch": 2959, "lr": 2.1553739895441848e-07}
{"train_loss": 0.04235943512509713, "global_step": 263439, "epoch": 2959, "lr": 2.1548363063423893e-07, "val_loss": 9.660772323608398}
{"train_loss": 0.040482647716999054, "global_step": 263440, "epoch": 2960, "lr": 2.1542986900697782e-07}
{"train_loss": 0.03145071864128113, "global_step": 263441, "epoch": 2960, "lr": 2.153761140726518e-07}
{"train_loss": 0.045924097299575806, "global_step": 263442, "epoch": 2960, "lr": 2.1532236583126085e-07}
{"train_loss": 0.04372479021549225, "global_step": 263443, "epoch": 2960, "lr": 2.152686242828106e-07}
{"train_loss": 0.0667543113231659, "global_step": 263444, "epoch": 2960, "lr": 2.1521488942731206e-07}
{"train_loss": 0.04930128529667854, "global_step": 263445, "epoch": 2960, "lr": 2.151611612647708e-07}
{"train_loss": 0.03426915407180786, "global_step": 263446, "epoch": 2960, "lr": 2.1510743979519244e-07}
{"train_loss": 0.019332734867930412, "global_step": 263447, "epoch": 2960, "lr": 2.15053725018588e-07}
{"train_loss": 0.08222855627536774, "global_step": 263448, "epoch": 2960, "lr": 2.1500001693496308e-07}
{"train_loss": 0.04865799471735954, "global_step": 263449, "epoch": 2960, "lr": 2.149463155443232e-07}
{"train_loss": 0.09762494266033173, "global_step": 263450, "epoch": 2960, "lr": 2.1489262084667948e-07}
{"train_loss": 0.04427911341190338, "global_step": 263451, "epoch": 2960, "lr": 2.1483893284203193e-07}
{"train_loss": 0.07126813381910324, "global_step": 263452, "epoch": 2960, "lr": 2.147852515303972e-07}
{"train_loss": 0.0509272925555706, "global_step": 263453, "epoch": 2960, "lr": 2.1473157691177527e-07}
{"train_loss": 0.03960351273417473, "global_step": 263454, "epoch": 2960, "lr": 2.1467790898617724e-07}
{"train_loss": 0.027924511581659317, "global_step": 263455, "epoch": 2960, "lr": 2.1462424775360868e-07}
{"train_loss": 0.06409160792827606, "global_step": 263456, "epoch": 2960, "lr": 2.1457059321407514e-07}
{"train_loss": 0.025513682514429092, "global_step": 263457, "epoch": 2960, "lr": 2.145169453675877e-07}
{"train_loss": 0.0743715912103653, "global_step": 263458, "epoch": 2960, "lr": 2.1446330421415195e-07}
{"train_loss": 0.03741908818483353, "global_step": 263459, "epoch": 2960, "lr": 2.1440966975377342e-07}
{"train_loss": 0.06407562643289566, "global_step": 263460, "epoch": 2960, "lr": 2.143560419864632e-07}
{"train_loss": 0.05590854585170746, "global_step": 263461, "epoch": 2960, "lr": 2.1430242091222685e-07}
{"train_loss": 0.01638794131577015, "global_step": 263462, "epoch": 2960, "lr": 2.1424880653106992e-07}
{"train_loss": 0.03938135877251625, "global_step": 263463, "epoch": 2960, "lr": 2.14195198842998e-07}
{"train_loss": 0.021177005022764206, "global_step": 263464, "epoch": 2960, "lr": 2.1414159784802213e-07}
{"train_loss": 0.08123522996902466, "global_step": 263465, "epoch": 2960, "lr": 2.1408800354614232e-07}
{"train_loss": 0.043334122747182846, "global_step": 263466, "epoch": 2960, "lr": 2.1403441593738084e-07}
{"train_loss": 0.03878839313983917, "global_step": 263467, "epoch": 2960, "lr": 2.1398083502173206e-07}
{"train_loss": 0.01800018548965454, "global_step": 263468, "epoch": 2960, "lr": 2.1392726079920155e-07}
{"train_loss": 0.07712572813034058, "global_step": 263469, "epoch": 2960, "lr": 2.1387369326981154e-07}
{"train_loss": 0.03892342001199722, "global_step": 263470, "epoch": 2960, "lr": 2.1382013243355092e-07}
{"train_loss": 0.028268005698919296, "global_step": 263471, "epoch": 2960, "lr": 2.137665782904419e-07}
{"train_loss": 0.04412918910384178, "global_step": 263472, "epoch": 2960, "lr": 2.1371303084047888e-07}
{"train_loss": 0.05444658175110817, "global_step": 263473, "epoch": 2960, "lr": 2.136594900836786e-07}
{"train_loss": 0.020268358290195465, "global_step": 263474, "epoch": 2960, "lr": 2.136059560200465e-07}
{"train_loss": 0.03160689026117325, "global_step": 263475, "epoch": 2960, "lr": 2.1355242864958824e-07}
{"train_loss": 0.026109809055924416, "global_step": 263476, "epoch": 2960, "lr": 2.1349890797230375e-07}
{"train_loss": 0.05476514622569084, "global_step": 263477, "epoch": 2960, "lr": 2.1344539398821527e-07}
{"train_loss": 0.07667026668787003, "global_step": 263478, "epoch": 2960, "lr": 2.133918866973228e-07}
{"train_loss": 0.030519185587763786, "global_step": 263479, "epoch": 2960, "lr": 2.1333838609962632e-07}
{"train_loss": 0.04966328293085098, "global_step": 263480, "epoch": 2960, "lr": 2.1328489219514803e-07}
{"train_loss": 0.02199353463947773, "global_step": 263481, "epoch": 2960, "lr": 2.132314049838824e-07}
{"train_loss": 0.04214656725525856, "global_step": 263482, "epoch": 2960, "lr": 2.131779244658405e-07}
{"train_loss": 0.055675968527793884, "global_step": 263483, "epoch": 2960, "lr": 2.1312445064102793e-07}
{"train_loss": 0.02230153977870941, "global_step": 263484, "epoch": 2960, "lr": 2.1307098350945576e-07}
{"train_loss": 0.052349917590618134, "global_step": 263485, "epoch": 2960, "lr": 2.1301752307112954e-07}
{"train_loss": 0.06710750609636307, "global_step": 263486, "epoch": 2960, "lr": 2.129640693260604e-07}
{"train_loss": 0.041923247277736664, "global_step": 263487, "epoch": 2960, "lr": 2.1291062227424274e-07}
{"train_loss": 0.035912029445171356, "global_step": 263488, "epoch": 2960, "lr": 2.1285718191569882e-07}
{"train_loss": 0.021107779815793037, "global_step": 263489, "epoch": 2960, "lr": 2.128037482504286e-07}
{"train_loss": 0.015612920746207237, "global_step": 263490, "epoch": 2960, "lr": 2.1275032127844318e-07}
{"train_loss": 0.06528833508491516, "global_step": 263491, "epoch": 2960, "lr": 2.126969009997426e-07}
{"train_loss": 0.07389510422945023, "global_step": 263492, "epoch": 2960, "lr": 2.1264348741433792e-07}
{"train_loss": 0.028060434386134148, "global_step": 263493, "epoch": 2960, "lr": 2.1259008052224028e-07}
{"train_loss": 0.04292331635951996, "global_step": 263494, "epoch": 2960, "lr": 2.1253668032344964e-07}
{"train_loss": 0.05749504268169403, "global_step": 263495, "epoch": 2960, "lr": 2.1248328681797714e-07}
{"train_loss": 0.042867112904787064, "global_step": 263496, "epoch": 2960, "lr": 2.1242990000582829e-07}
{"train_loss": 0.07008399814367294, "global_step": 263497, "epoch": 2960, "lr": 2.1237651988701422e-07}
{"train_loss": 0.04435740411281586, "global_step": 263498, "epoch": 2960, "lr": 2.1232314646153495e-07}
{"train_loss": 0.035264842212200165, "global_step": 263499, "epoch": 2960, "lr": 2.122697797294071e-07}
{"train_loss": 0.07682017982006073, "global_step": 263500, "epoch": 2960, "lr": 2.1221641969063065e-07}
{"train_loss": 0.02972029522061348, "global_step": 263501, "epoch": 2960, "lr": 2.121630663452112e-07}
{"train_loss": 0.048076193779706955, "global_step": 263502, "epoch": 2960, "lr": 2.121097196931654e-07}
{"train_loss": 0.0860995352268219, "global_step": 263503, "epoch": 2960, "lr": 2.120563797344932e-07}
{"train_loss": 0.025419441983103752, "global_step": 263504, "epoch": 2960, "lr": 2.1200304646920022e-07}
{"train_loss": 0.06290556490421295, "global_step": 263505, "epoch": 2960, "lr": 2.119497198972975e-07}
{"train_loss": 0.05598403513431549, "global_step": 263506, "epoch": 2960, "lr": 2.1189640001879063e-07}
{"train_loss": 0.06232358142733574, "global_step": 263507, "epoch": 2960, "lr": 2.118430868336907e-07}
{"train_loss": 0.01813032664358616, "global_step": 263508, "epoch": 2960, "lr": 2.117897803419977e-07}
{"train_loss": 0.027521049603819847, "global_step": 263509, "epoch": 2960, "lr": 2.117364805437283e-07}
{"train_loss": 0.021272892132401466, "global_step": 263510, "epoch": 2960, "lr": 2.116831874388825e-07}
{"train_loss": 0.08347907662391663, "global_step": 263511, "epoch": 2960, "lr": 2.1162990102746582e-07}
{"train_loss": 0.0072946697473526, "global_step": 263512, "epoch": 2960, "lr": 2.1157662130948942e-07}
{"train_loss": 0.030067872256040573, "global_step": 263513, "epoch": 2960, "lr": 2.1152334828496433e-07}
{"train_loss": 0.025431044399738312, "global_step": 263514, "epoch": 2960, "lr": 2.1147008195388507e-07}
{"train_loss": 0.03300857916474342, "global_step": 263515, "epoch": 2960, "lr": 2.114168223162738e-07}
{"train_loss": 0.08798358589410782, "global_step": 263516, "epoch": 2960, "lr": 2.1136356937212498e-07}
{"train_loss": 0.039300743490457535, "global_step": 263517, "epoch": 2960, "lr": 2.1131032312145527e-07}
{"train_loss": 0.0652238056063652, "global_step": 263518, "epoch": 2960, "lr": 2.112570835642702e-07}
{"train_loss": 0.06183253601193428, "global_step": 263519, "epoch": 2960, "lr": 2.112038507005698e-07}
{"train_loss": 0.035907041281461716, "global_step": 263520, "epoch": 2960, "lr": 2.111506245303707e-07}
{"train_loss": 0.04000120982527733, "global_step": 263521, "epoch": 2960, "lr": 2.1109740505366739e-07}
{"train_loss": 0.05527010187506676, "global_step": 263522, "epoch": 2960, "lr": 2.11044192270482e-07}
{"train_loss": 0.023301810026168823, "global_step": 263523, "epoch": 2960, "lr": 2.1099098618081458e-07}
{"train_loss": 0.030261371284723282, "global_step": 263524, "epoch": 2960, "lr": 2.109377867846707e-07}
{"train_loss": 0.07100452482700348, "global_step": 263525, "epoch": 2960, "lr": 2.108845940820614e-07}
{"train_loss": 0.06881293654441833, "global_step": 263526, "epoch": 2960, "lr": 2.1083140807299228e-07}
{"train_loss": 0.05105310305953026, "global_step": 263527, "epoch": 2960, "lr": 2.107782287574689e-07}
{"train_loss": 0.045948878746856464, "global_step": 263528, "epoch": 2960, "lr": 2.1072505613549676e-07, "val_loss": 9.633599281311035, "train_action_mse_error": 36.381038665771484}
{"train_loss": 0.021702364087104797, "global_step": 263529, "epoch": 2961, "lr": 2.1067189020709256e-07}
{"train_loss": 0.04303780570626259, "global_step": 263530, "epoch": 2961, "lr": 2.1061873097225072e-07}
{"train_loss": 0.041135046631097794, "global_step": 263531, "epoch": 2961, "lr": 2.105655784309879e-07}
{"train_loss": 0.04112369939684868, "global_step": 263532, "epoch": 2961, "lr": 2.105124325833041e-07}
{"train_loss": 0.0652293860912323, "global_step": 263533, "epoch": 2961, "lr": 2.1045929342921044e-07}
{"train_loss": 0.024952838197350502, "global_step": 263534, "epoch": 2961, "lr": 2.10406160968718e-07}
{"train_loss": 0.079477459192276, "global_step": 263535, "epoch": 2961, "lr": 2.103530352018268e-07}
{"train_loss": 0.06743639707565308, "global_step": 263536, "epoch": 2961, "lr": 2.1029991612854793e-07}
{"train_loss": 0.07109657675027847, "global_step": 263537, "epoch": 2961, "lr": 2.1024680374888694e-07}
{"train_loss": 0.03897278010845184, "global_step": 263538, "epoch": 2961, "lr": 2.101936980628494e-07}
{"train_loss": 0.01737920194864273, "global_step": 263539, "epoch": 2961, "lr": 2.1014059907044636e-07}
{"train_loss": 0.016993973404169083, "global_step": 263540, "epoch": 2961, "lr": 2.1008750677167788e-07}
{"train_loss": 0.04653418809175491, "global_step": 263541, "epoch": 2961, "lr": 2.100344211665606e-07}
{"train_loss": 0.04915987700223923, "global_step": 263542, "epoch": 2961, "lr": 2.0998134225510002e-07}
{"train_loss": 0.041171953082084656, "global_step": 263543, "epoch": 2961, "lr": 2.099282700372962e-07}
{"train_loss": 0.01579093560576439, "global_step": 263544, "epoch": 2961, "lr": 2.0987520451316577e-07}
{"train_loss": 0.028902601450681686, "global_step": 263545, "epoch": 2961, "lr": 2.0982214568270875e-07}
{"train_loss": 0.04555423930287361, "global_step": 263546, "epoch": 2961, "lr": 2.0976909354593065e-07}
{"train_loss": 0.044790118932724, "global_step": 263547, "epoch": 2961, "lr": 2.0971604810284262e-07}
{"train_loss": 0.033026326447725296, "global_step": 263548, "epoch": 2961, "lr": 2.0966300935345573e-07}
{"train_loss": 0.04578316584229469, "global_step": 263549, "epoch": 2961, "lr": 2.0960997729777e-07}
{"train_loss": 0.017924616113305092, "global_step": 263550, "epoch": 2961, "lr": 2.095569519357965e-07}
{"train_loss": 0.02366911806166172, "global_step": 263551, "epoch": 2961, "lr": 2.095039332675408e-07}
{"train_loss": 0.04646017774939537, "global_step": 263552, "epoch": 2961, "lr": 2.0945092129300846e-07}
{"train_loss": 0.03746814653277397, "global_step": 263553, "epoch": 2961, "lr": 2.0939791601221058e-07}
{"train_loss": 0.04124053940176964, "global_step": 263554, "epoch": 2961, "lr": 2.0934491742515272e-07}
{"train_loss": 0.036543820053339005, "global_step": 263555, "epoch": 2961, "lr": 2.092919255318404e-07}
{"train_loss": 0.039075978100299835, "global_step": 263556, "epoch": 2961, "lr": 2.0923894033228476e-07}
{"train_loss": 0.03855586796998978, "global_step": 263557, "epoch": 2961, "lr": 2.0918596182648575e-07}
{"train_loss": 0.040973961353302, "global_step": 263558, "epoch": 2961, "lr": 2.0913299001445453e-07}
{"train_loss": 0.04032158479094505, "global_step": 263559, "epoch": 2961, "lr": 2.0908002489620215e-07}
{"train_loss": 0.030620375648140907, "global_step": 263560, "epoch": 2961, "lr": 2.0902706647172865e-07}
{"train_loss": 0.017079118639230728, "global_step": 263561, "epoch": 2961, "lr": 2.0897411474105067e-07}
{"train_loss": 0.03263753652572632, "global_step": 263562, "epoch": 2961, "lr": 2.0892116970416266e-07}
{"train_loss": 0.05417025461792946, "global_step": 263563, "epoch": 2961, "lr": 2.0886823136108126e-07}
{"train_loss": 0.10939273983240128, "global_step": 263564, "epoch": 2961, "lr": 2.08815299711812e-07}
{"train_loss": 0.04656491428613663, "global_step": 263565, "epoch": 2961, "lr": 2.0876237475635495e-07}
{"train_loss": 0.040657248347997665, "global_step": 263566, "epoch": 2961, "lr": 2.0870945649473228e-07}
{"train_loss": 0.03443954512476921, "global_step": 263567, "epoch": 2961, "lr": 2.0865654492693288e-07}
{"train_loss": 0.03701092675328255, "global_step": 263568, "epoch": 2961, "lr": 2.0860364005297894e-07}
{"train_loss": 0.07481653988361359, "global_step": 263569, "epoch": 2961, "lr": 2.0855074187286495e-07}
{"train_loss": 0.03626597300171852, "global_step": 263570, "epoch": 2961, "lr": 2.0849785038661308e-07}
{"train_loss": 0.0710015594959259, "global_step": 263571, "epoch": 2961, "lr": 2.0844496559421777e-07}
{"train_loss": 0.04560915753245354, "global_step": 263572, "epoch": 2961, "lr": 2.083920874956846e-07}
{"train_loss": 0.05124644190073013, "global_step": 263573, "epoch": 2961, "lr": 2.083392160910358e-07}
{"train_loss": 0.06169036403298378, "global_step": 263574, "epoch": 2961, "lr": 2.0828635138026021e-07}
{"train_loss": 0.022304194048047066, "global_step": 263575, "epoch": 2961, "lr": 2.0823349336338004e-07}
{"train_loss": 0.015070665627717972, "global_step": 263576, "epoch": 2961, "lr": 2.081806420403898e-07}
{"train_loss": 0.031171265989542007, "global_step": 263577, "epoch": 2961, "lr": 2.0812779741130607e-07}
{"train_loss": 0.04283393546938896, "global_step": 263578, "epoch": 2961, "lr": 2.0807495947613442e-07}
{"train_loss": 0.055397626012563705, "global_step": 263579, "epoch": 2961, "lr": 2.0802212823488043e-07}
{"train_loss": 0.05406872183084488, "global_step": 263580, "epoch": 2961, "lr": 2.0796930368754963e-07}
{"train_loss": 0.059993162751197815, "global_step": 263581, "epoch": 2961, "lr": 2.0791648583414757e-07}
{"train_loss": 0.040017370134592056, "global_step": 263582, "epoch": 2961, "lr": 2.0786367467469094e-07}
{"train_loss": 0.05164644494652748, "global_step": 263583, "epoch": 2961, "lr": 2.078108702091741e-07}
{"train_loss": 0.0571831651031971, "global_step": 263584, "epoch": 2961, "lr": 2.077580724376138e-07}
{"train_loss": 0.029913216829299927, "global_step": 263585, "epoch": 2961, "lr": 2.0770528136001e-07}
{"train_loss": 0.08529593795537949, "global_step": 263586, "epoch": 2961, "lr": 2.0765249697637378e-07}
{"train_loss": 0.011455457657575607, "global_step": 263587, "epoch": 2961, "lr": 2.0759971928671628e-07}
{"train_loss": 0.05572662875056267, "global_step": 263588, "epoch": 2961, "lr": 2.075469482910375e-07}
{"train_loss": 0.029026011005043983, "global_step": 263589, "epoch": 2961, "lr": 2.074941839893485e-07}
{"train_loss": 0.042753301560878754, "global_step": 263590, "epoch": 2961, "lr": 2.0744142638164932e-07}
{"train_loss": 0.05649961531162262, "global_step": 263591, "epoch": 2961, "lr": 2.073886754679566e-07}
{"train_loss": 0.03762216493487358, "global_step": 263592, "epoch": 2961, "lr": 2.0733593124827588e-07}
{"train_loss": 0.02704009972512722, "global_step": 263593, "epoch": 2961, "lr": 2.0728319372260717e-07}
{"train_loss": 0.046562571078538895, "global_step": 263594, "epoch": 2961, "lr": 2.0723046289096714e-07}
{"train_loss": 0.0368671752512455, "global_step": 263595, "epoch": 2961, "lr": 2.0717773875335577e-07}
{"train_loss": 0.03290342539548874, "global_step": 263596, "epoch": 2961, "lr": 2.0712502130978416e-07}
{"train_loss": 0.07587466388940811, "global_step": 263597, "epoch": 2961, "lr": 2.0707231056025234e-07}
{"train_loss": 0.059941042214632034, "global_step": 263598, "epoch": 2961, "lr": 2.070196065047769e-07}
{"train_loss": 0.07696162909269333, "global_step": 263599, "epoch": 2961, "lr": 2.0696690914336347e-07}
{"train_loss": 0.022413088008761406, "global_step": 263600, "epoch": 2961, "lr": 2.06914218476012e-07}
{"train_loss": 0.06339196115732193, "global_step": 263601, "epoch": 2961, "lr": 2.068615345027336e-07}
{"train_loss": 0.05975736677646637, "global_step": 263602, "epoch": 2961, "lr": 2.0680885722353937e-07}
{"train_loss": 0.03262382745742798, "global_step": 263603, "epoch": 2961, "lr": 2.0675618663842934e-07}
{"train_loss": 0.035479236394166946, "global_step": 263604, "epoch": 2961, "lr": 2.067035227474201e-07}
{"train_loss": 0.06432640552520752, "global_step": 263605, "epoch": 2961, "lr": 2.0665086555050617e-07}
{"train_loss": 0.04579148814082146, "global_step": 263606, "epoch": 2961, "lr": 2.0659821504770415e-07}
{"train_loss": 0.03148652985692024, "global_step": 263607, "epoch": 2961, "lr": 2.0654557123901407e-07}
{"train_loss": 0.023487744852900505, "global_step": 263608, "epoch": 2961, "lr": 2.0649293412445259e-07}
{"train_loss": 0.053305432200431824, "global_step": 263609, "epoch": 2961, "lr": 2.0644030370401967e-07}
{"train_loss": 0.049201156944036484, "global_step": 263610, "epoch": 2961, "lr": 2.063876799777209e-07}
{"train_loss": 0.03303854912519455, "global_step": 263611, "epoch": 2961, "lr": 2.063350629455729e-07}
{"train_loss": 0.02340327575802803, "global_step": 263612, "epoch": 2961, "lr": 2.062824526075702e-07}
{"train_loss": 0.018444553017616272, "global_step": 263613, "epoch": 2961, "lr": 2.0622984896372933e-07}
{"train_loss": 0.06939820945262909, "global_step": 263614, "epoch": 2961, "lr": 2.0617725201405037e-07}
{"train_loss": 0.028399286791682243, "global_step": 263615, "epoch": 2961, "lr": 2.0612466175854994e-07}
{"train_loss": 0.04932175204157829, "global_step": 263616, "epoch": 2961, "lr": 2.0607207819722253e-07}
{"train_loss": 0.04325595014634427, "global_step": 263617, "epoch": 2961, "lr": 2.0601950133008475e-07, "val_loss": 9.622940063476562}
{"train_loss": 0.04407791793346405, "global_step": 263618, "epoch": 2962, "lr": 2.0596693115714216e-07}
{"train_loss": 0.014952306635677814, "global_step": 263619, "epoch": 2962, "lr": 2.0591436767839478e-07}
{"train_loss": 0.04382331296801567, "global_step": 263620, "epoch": 2962, "lr": 2.0586181089385925e-07}
{"train_loss": 0.03813188895583153, "global_step": 263621, "epoch": 2962, "lr": 2.0580926080354112e-07}
{"train_loss": 0.03215096518397331, "global_step": 263622, "epoch": 2962, "lr": 2.057567174074404e-07}
{"train_loss": 0.05183091014623642, "global_step": 263623, "epoch": 2962, "lr": 2.0570418070557372e-07}
{"train_loss": 0.07240834087133408, "global_step": 263624, "epoch": 2962, "lr": 2.0565165069794111e-07}
{"train_loss": 0.03378736600279808, "global_step": 263625, "epoch": 2962, "lr": 2.055991273845481e-07}
{"train_loss": 0.017987707629799843, "global_step": 263626, "epoch": 2962, "lr": 2.0554661076541137e-07}
{"train_loss": 0.03191345930099487, "global_step": 263627, "epoch": 2962, "lr": 2.054941008405309e-07}
{"train_loss": 0.028527000918984413, "global_step": 263628, "epoch": 2962, "lr": 2.0544159760991223e-07}
{"train_loss": 0.05372782424092293, "global_step": 263629, "epoch": 2962, "lr": 2.0538910107356646e-07}
{"train_loss": 0.043982598930597305, "global_step": 263630, "epoch": 2962, "lr": 2.053366112314936e-07}
{"train_loss": 0.03035408817231655, "global_step": 263631, "epoch": 2962, "lr": 2.0528412808371035e-07}
{"train_loss": 0.05352454632520676, "global_step": 263632, "epoch": 2962, "lr": 2.0523165163022218e-07}
{"train_loss": 0.017529770731925964, "global_step": 263633, "epoch": 2962, "lr": 2.051791818710347e-07}
{"train_loss": 0.05385471507906914, "global_step": 263634, "epoch": 2962, "lr": 2.0512671880614788e-07}
{"train_loss": 0.04429841786623001, "global_step": 263635, "epoch": 2962, "lr": 2.0507426243557838e-07}
{"train_loss": 0.04944828897714615, "global_step": 263636, "epoch": 2962, "lr": 2.0502181275933174e-07}
{"train_loss": 0.011462573893368244, "global_step": 263637, "epoch": 2962, "lr": 2.04969369777408e-07}
{"train_loss": 0.056634340435266495, "global_step": 263638, "epoch": 2962, "lr": 2.0491693348982378e-07}
{"train_loss": 0.05139491707086563, "global_step": 263639, "epoch": 2962, "lr": 2.0486450389657908e-07}
{"train_loss": 0.05536676570773125, "global_step": 263640, "epoch": 2962, "lr": 2.04812080997685e-07}
{"train_loss": 0.049863021820783615, "global_step": 263641, "epoch": 2962, "lr": 2.0475966479314713e-07}
{"train_loss": 0.05076334998011589, "global_step": 263642, "epoch": 2962, "lr": 2.0470725528297098e-07}
{"train_loss": 0.038693640381097794, "global_step": 263643, "epoch": 2962, "lr": 2.0465485246716765e-07}
{"train_loss": 0.05090394988656044, "global_step": 263644, "epoch": 2962, "lr": 2.0460245634573715e-07}
{"train_loss": 0.019070567563176155, "global_step": 263645, "epoch": 2962, "lr": 2.0455006691869616e-07}
{"train_loss": 0.036099664866924286, "global_step": 263646, "epoch": 2962, "lr": 2.0449768418604465e-07}
{"train_loss": 0.06809766590595245, "global_step": 263647, "epoch": 2962, "lr": 2.0444530814778817e-07}
{"train_loss": 0.08822523057460785, "global_step": 263648, "epoch": 2962, "lr": 2.0439293880393783e-07}
{"train_loss": 0.08565763384103775, "global_step": 263649, "epoch": 2962, "lr": 2.0434057615450475e-07}
{"train_loss": 0.03606447950005531, "global_step": 263650, "epoch": 2962, "lr": 2.042882201994889e-07}
{"train_loss": 0.05845670774579048, "global_step": 263651, "epoch": 2962, "lr": 2.0423587093889585e-07}
{"train_loss": 0.013269053772091866, "global_step": 263652, "epoch": 2962, "lr": 2.0418352837274223e-07}
{"train_loss": 0.0541544146835804, "global_step": 263653, "epoch": 2962, "lr": 2.0413119250102809e-07}
{"train_loss": 0.015776976943016052, "global_step": 263654, "epoch": 2962, "lr": 2.0407886332375337e-07}
{"train_loss": 0.022062886506319046, "global_step": 263655, "epoch": 2962, "lr": 2.0402654084094585e-07}
{"train_loss": 0.03574112802743912, "global_step": 263656, "epoch": 2962, "lr": 2.039742250525889e-07}
{"train_loss": 0.06850680708885193, "global_step": 263657, "epoch": 2962, "lr": 2.0392191595871024e-07}
{"train_loss": 0.02594798244535923, "global_step": 263658, "epoch": 2962, "lr": 2.0386961355929878e-07}
{"train_loss": 0.08381619304418564, "global_step": 263659, "epoch": 2962, "lr": 2.0381731785437674e-07}
{"train_loss": 0.04108768701553345, "global_step": 263660, "epoch": 2962, "lr": 2.0376502884393854e-07}
{"train_loss": 0.05565303936600685, "global_step": 263661, "epoch": 2962, "lr": 2.0371274652800087e-07}
{"train_loss": 0.028429647907614708, "global_step": 263662, "epoch": 2962, "lr": 2.0366047090656925e-07}
{"train_loss": 0.03178710490465164, "global_step": 263663, "epoch": 2962, "lr": 2.036082019796437e-07}
{"train_loss": 0.02616935595870018, "global_step": 263664, "epoch": 2962, "lr": 2.0355593974724086e-07}
{"train_loss": 0.02350565418601036, "global_step": 263665, "epoch": 2962, "lr": 2.0350368420936072e-07}
{"train_loss": 0.022310130298137665, "global_step": 263666, "epoch": 2962, "lr": 2.0345143536600887e-07}
{"train_loss": 0.06409204006195068, "global_step": 263667, "epoch": 2962, "lr": 2.033991932172019e-07}
{"train_loss": 0.037238262593746185, "global_step": 263668, "epoch": 2962, "lr": 2.033469577629399e-07}
{"train_loss": 0.015646370127797127, "global_step": 263669, "epoch": 2962, "lr": 2.0329472900322833e-07}
{"train_loss": 0.026107963174581528, "global_step": 263670, "epoch": 2962, "lr": 2.0324250693807835e-07}
{"train_loss": 0.018755123019218445, "global_step": 263671, "epoch": 2962, "lr": 2.0319029156749548e-07}
{"train_loss": 0.029893871396780014, "global_step": 263672, "epoch": 2962, "lr": 2.0313808289148527e-07}
{"train_loss": 0.04901786893606186, "global_step": 263673, "epoch": 2962, "lr": 2.0308588091005888e-07}
{"train_loss": 0.06934984773397446, "global_step": 263674, "epoch": 2962, "lr": 2.0303368562321623e-07}
{"train_loss": 0.012585647404193878, "global_step": 263675, "epoch": 2962, "lr": 2.0298149703097402e-07}
{"train_loss": 0.04653141647577286, "global_step": 263676, "epoch": 2962, "lr": 2.0292931513333223e-07}
{"train_loss": 0.02731524594128132, "global_step": 263677, "epoch": 2962, "lr": 2.0287713993029644e-07}
{"train_loss": 0.053620219230651855, "global_step": 263678, "epoch": 2962, "lr": 2.0282497142188328e-07}
{"train_loss": 0.038618333637714386, "global_step": 263679, "epoch": 2962, "lr": 2.027728096080872e-07}
{"train_loss": 0.062198519706726074, "global_step": 263680, "epoch": 2962, "lr": 2.0272065448892486e-07}
{"train_loss": 0.029788626357913017, "global_step": 263681, "epoch": 2962, "lr": 2.0266850606439624e-07}
{"train_loss": 0.020443903282284737, "global_step": 263682, "epoch": 2962, "lr": 2.0261636433451802e-07}
{"train_loss": 0.04971439763903618, "global_step": 263683, "epoch": 2962, "lr": 2.0256422929928464e-07}
{"train_loss": 0.052455656230449677, "global_step": 263684, "epoch": 2962, "lr": 2.0251210095871276e-07}
{"train_loss": 0.03792933002114296, "global_step": 263685, "epoch": 2962, "lr": 2.0245997931280792e-07}
{"train_loss": 0.02695949375629425, "global_step": 263686, "epoch": 2962, "lr": 2.0240786436157566e-07}
{"train_loss": 0.034648071974515915, "global_step": 263687, "epoch": 2962, "lr": 2.0235575610502155e-07}
{"train_loss": 0.09947597980499268, "global_step": 263688, "epoch": 2962, "lr": 2.0230365454315114e-07}
{"train_loss": 0.0821397453546524, "global_step": 263689, "epoch": 2962, "lr": 2.0225155967597554e-07}
{"train_loss": 0.04323304072022438, "global_step": 263690, "epoch": 2962, "lr": 2.0219947150350027e-07}
{"train_loss": 0.029054870828986168, "global_step": 263691, "epoch": 2962, "lr": 2.021473900257309e-07}
{"train_loss": 0.019048646092414856, "global_step": 263692, "epoch": 2962, "lr": 2.0209531524267855e-07}
{"train_loss": 0.08581264317035675, "global_step": 263693, "epoch": 2962, "lr": 2.020432471543432e-07}
{"train_loss": 0.05067842826247215, "global_step": 263694, "epoch": 2962, "lr": 2.019911857607415e-07}
{"train_loss": 0.039495885372161865, "global_step": 263695, "epoch": 2962, "lr": 2.0193913106187345e-07}
{"train_loss": 0.04434473067522049, "global_step": 263696, "epoch": 2962, "lr": 2.0188708305774461e-07}
{"train_loss": 0.08083629608154297, "global_step": 263697, "epoch": 2962, "lr": 2.0183504174837163e-07}
{"train_loss": 0.057649318128824234, "global_step": 263698, "epoch": 2962, "lr": 2.01783007133749e-07}
{"train_loss": 0.042351048439741135, "global_step": 263699, "epoch": 2962, "lr": 2.017309792138933e-07}
{"train_loss": 0.0031383626628667116, "global_step": 263700, "epoch": 2962, "lr": 2.0167895798880455e-07}
{"train_loss": 0.10116077214479446, "global_step": 263701, "epoch": 2962, "lr": 2.0162694345849387e-07}
{"train_loss": 0.027309060096740723, "global_step": 263702, "epoch": 2962, "lr": 2.015749356229668e-07}
{"train_loss": 0.026514098048210144, "global_step": 263703, "epoch": 2962, "lr": 2.0152293448223446e-07}
{"train_loss": 0.035015568137168884, "global_step": 263704, "epoch": 2962, "lr": 2.0147094003629686e-07}
{"train_loss": 0.03654560446739197, "global_step": 263705, "epoch": 2962, "lr": 2.014189522851706e-07}
{"train_loss": 0.042752401141993954, "global_step": 263706, "epoch": 2962, "lr": 2.0136697122885018e-07, "val_loss": 9.744600296020508}
{"train_loss": 0.027472905814647675, "global_step": 263707, "epoch": 2963, "lr": 2.0131499686735222e-07}
{"train_loss": 0.030370792374014854, "global_step": 263708, "epoch": 2963, "lr": 2.0126302920067674e-07}
{"train_loss": 0.036631494760513306, "global_step": 263709, "epoch": 2963, "lr": 2.012110682288404e-07}
{"train_loss": 0.048646535724401474, "global_step": 263710, "epoch": 2963, "lr": 2.0115911395183763e-07}
{"train_loss": 0.050112996250391006, "global_step": 263711, "epoch": 2963, "lr": 2.0110716636968506e-07}
{"train_loss": 0.04930754005908966, "global_step": 263712, "epoch": 2963, "lr": 2.010552254823883e-07}
{"train_loss": 0.05550070479512215, "global_step": 263713, "epoch": 2963, "lr": 2.0100329128995286e-07}
{"train_loss": 0.05026348680257797, "global_step": 263714, "epoch": 2963, "lr": 2.0095136379238432e-07}
{"train_loss": 0.04137253388762474, "global_step": 263715, "epoch": 2963, "lr": 2.0089944298968822e-07}
{"train_loss": 0.03493810072541237, "global_step": 263716, "epoch": 2963, "lr": 2.008475288818812e-07}
{"train_loss": 0.05433804914355278, "global_step": 263717, "epoch": 2963, "lr": 2.007956214689577e-07}
{"train_loss": 0.03306107223033905, "global_step": 263718, "epoch": 2963, "lr": 2.0074372075092883e-07}
{"train_loss": 0.031122110784053802, "global_step": 263719, "epoch": 2963, "lr": 2.0069182672780574e-07}
{"train_loss": 0.04883335530757904, "global_step": 263720, "epoch": 2963, "lr": 2.006399393995939e-07}
{"train_loss": 0.04795149713754654, "global_step": 263721, "epoch": 2963, "lr": 2.0058805876629895e-07}
{"train_loss": 0.07113645225763321, "global_step": 263722, "epoch": 2963, "lr": 2.0053618482792634e-07}
{"train_loss": 0.05504663661122322, "global_step": 263723, "epoch": 2963, "lr": 2.0048431758448171e-07}
{"train_loss": 0.054468195885419846, "global_step": 263724, "epoch": 2963, "lr": 2.004324570359817e-07}
{"train_loss": 0.017100336030125618, "global_step": 263725, "epoch": 2963, "lr": 2.0038060318242624e-07}
{"train_loss": 0.07782246917486191, "global_step": 263726, "epoch": 2963, "lr": 2.003287560238154e-07}
{"train_loss": 0.021900242194533348, "global_step": 263727, "epoch": 2963, "lr": 2.0027691556017137e-07}
{"train_loss": 0.03250947967171669, "global_step": 263728, "epoch": 2963, "lr": 2.002250817914886e-07}
{"train_loss": 0.04030638560652733, "global_step": 263729, "epoch": 2963, "lr": 2.0017325471777815e-07}
{"train_loss": 0.06268847733736038, "global_step": 263730, "epoch": 2963, "lr": 2.0012143433904562e-07}
{"train_loss": 0.038292113691568375, "global_step": 263731, "epoch": 2963, "lr": 2.0006962065530766e-07}
{"train_loss": 0.042713768780231476, "global_step": 263732, "epoch": 2963, "lr": 2.0001781366655314e-07}
{"train_loss": 0.016695158556103706, "global_step": 263733, "epoch": 2963, "lr": 1.9996601337280985e-07}
{"train_loss": 0.03293270245194435, "global_step": 263734, "epoch": 2963, "lr": 1.9991421977406667e-07}
{"train_loss": 0.055705178529024124, "global_step": 263735, "epoch": 2963, "lr": 1.9986243287034022e-07}
{"train_loss": 0.03850654885172844, "global_step": 263736, "epoch": 2963, "lr": 1.9981065266163612e-07}
{"train_loss": 0.03215549886226654, "global_step": 263737, "epoch": 2963, "lr": 1.9975887914795987e-07}
{"train_loss": 0.03513558208942413, "global_step": 263738, "epoch": 2963, "lr": 1.9970711232931705e-07}
{"train_loss": 0.04844176024198532, "global_step": 263739, "epoch": 2963, "lr": 1.9965535220571874e-07}
{"train_loss": 0.049860838800668716, "global_step": 263740, "epoch": 2963, "lr": 1.996035987771705e-07}
{"train_loss": 0.032562196254730225, "global_step": 263741, "epoch": 2963, "lr": 1.9955185204367234e-07}
{"train_loss": 0.041490588337183, "global_step": 263742, "epoch": 2963, "lr": 1.9950011200524643e-07}
{"train_loss": 0.044023241847753525, "global_step": 263743, "epoch": 2963, "lr": 1.9944837866188726e-07}
{"train_loss": 0.0681251809000969, "global_step": 263744, "epoch": 2963, "lr": 1.9939665201360592e-07}
{"train_loss": 0.0352209098637104, "global_step": 263745, "epoch": 2963, "lr": 1.9934493206040794e-07}
{"train_loss": 0.02802734449505806, "global_step": 263746, "epoch": 2963, "lr": 1.992932188022989e-07}
{"train_loss": 0.03251213580369949, "global_step": 263747, "epoch": 2963, "lr": 1.9924151223928989e-07}
{"train_loss": 0.060578908771276474, "global_step": 263748, "epoch": 2963, "lr": 1.9918981237138644e-07}
{"train_loss": 0.08247272670269012, "global_step": 263749, "epoch": 2963, "lr": 1.9913811919859415e-07}
{"train_loss": 0.028652912005782127, "global_step": 263750, "epoch": 2963, "lr": 1.9908643272092408e-07}
{"train_loss": 0.04449235275387764, "global_step": 263751, "epoch": 2963, "lr": 1.9903475293837072e-07}
{"train_loss": 0.0225413478910923, "global_step": 263752, "epoch": 2963, "lr": 1.9898307985096178e-07}
{"train_loss": 0.04711813107132912, "global_step": 263753, "epoch": 2963, "lr": 1.9893141345868617e-07}
{"train_loss": 0.028719503432512283, "global_step": 263754, "epoch": 2963, "lr": 1.9887975376155498e-07}
{"train_loss": 0.05149789899587631, "global_step": 263755, "epoch": 2963, "lr": 1.9882810075957937e-07}
{"train_loss": 0.047992169857025146, "global_step": 263756, "epoch": 2963, "lr": 1.9877645445277038e-07}
{"train_loss": 0.09233088791370392, "global_step": 263757, "epoch": 2963, "lr": 1.987248148411225e-07}
{"train_loss": 0.090439073741436, "global_step": 263758, "epoch": 2963, "lr": 1.9867318192465234e-07}
{"train_loss": 0.05405876412987709, "global_step": 263759, "epoch": 2963, "lr": 1.9862155570335995e-07}
{"train_loss": 0.06811846047639847, "global_step": 263760, "epoch": 2963, "lr": 1.9856993617725638e-07}
{"train_loss": 0.05306476727128029, "global_step": 263761, "epoch": 2963, "lr": 1.9851832334634723e-07}
{"train_loss": 0.0326998233795166, "global_step": 263762, "epoch": 2963, "lr": 1.9846671721064357e-07}
{"train_loss": 0.058713242411613464, "global_step": 263763, "epoch": 2963, "lr": 1.98415117770151e-07}
{"train_loss": 0.028566721826791763, "global_step": 263764, "epoch": 2963, "lr": 1.9836352502486943e-07}
{"train_loss": 0.04431124031543732, "global_step": 263765, "epoch": 2963, "lr": 1.983119389748156e-07}
{"train_loss": 0.055856384336948395, "global_step": 263766, "epoch": 2963, "lr": 1.9826035961998945e-07}
{"train_loss": 0.032675422728061676, "global_step": 263767, "epoch": 2963, "lr": 1.9820878696039657e-07}
{"train_loss": 0.024271151050925255, "global_step": 263768, "epoch": 2963, "lr": 1.981572209960536e-07}
{"train_loss": 0.019722439348697662, "global_step": 263769, "epoch": 2963, "lr": 1.9810566172696055e-07}
{"train_loss": 0.06618544459342957, "global_step": 263770, "epoch": 2963, "lr": 1.9805410915311738e-07}
{"train_loss": 0.047702401876449585, "global_step": 263771, "epoch": 2963, "lr": 1.9800256327454636e-07}
{"train_loss": 0.06871792674064636, "global_step": 263772, "epoch": 2963, "lr": 1.9795102409124745e-07}
{"train_loss": 0.029372327029705048, "global_step": 263773, "epoch": 2963, "lr": 1.9789949160322619e-07}
{"train_loss": 0.05113697052001953, "global_step": 263774, "epoch": 2963, "lr": 1.9784796581048815e-07}
{"train_loss": 0.04258880019187927, "global_step": 263775, "epoch": 2963, "lr": 1.9779644671303888e-07}
{"train_loss": 0.06518115848302841, "global_step": 263776, "epoch": 2963, "lr": 1.9774493431089503e-07}
{"train_loss": 0.04240965098142624, "global_step": 263777, "epoch": 2963, "lr": 1.976934286040566e-07}
{"train_loss": 0.02314586006104946, "global_step": 263778, "epoch": 2963, "lr": 1.9764192959252914e-07}
{"train_loss": 0.05791229382157326, "global_step": 263779, "epoch": 2963, "lr": 1.9759043727632375e-07}
{"train_loss": 0.060353416949510574, "global_step": 263780, "epoch": 2963, "lr": 1.97538951655446e-07}
{"train_loss": 0.06854631006717682, "global_step": 263781, "epoch": 2963, "lr": 1.9748747272989588e-07}
{"train_loss": 0.08404150605201721, "global_step": 263782, "epoch": 2963, "lr": 1.9743600049969557e-07}
{"train_loss": 0.06014872342348099, "global_step": 263783, "epoch": 2963, "lr": 1.97384534964834e-07}
{"train_loss": 0.06251411885023117, "global_step": 263784, "epoch": 2963, "lr": 1.9733307612533337e-07}
{"train_loss": 0.0497899167239666, "global_step": 263785, "epoch": 2963, "lr": 1.9728162398119366e-07}
{"train_loss": 0.01962951198220253, "global_step": 263786, "epoch": 2963, "lr": 1.9723017853242044e-07}
{"train_loss": 0.027470994740724564, "global_step": 263787, "epoch": 2963, "lr": 1.9717873977902478e-07}
{"train_loss": 0.031160429120063782, "global_step": 263788, "epoch": 2963, "lr": 1.9712730772101228e-07}
{"train_loss": 0.03496652469038963, "global_step": 263789, "epoch": 2963, "lr": 1.9707588235838293e-07}
{"train_loss": 0.030861714854836464, "global_step": 263790, "epoch": 2963, "lr": 1.970244636911589e-07}
{"train_loss": 0.04699992388486862, "global_step": 263791, "epoch": 2963, "lr": 1.9697305171932912e-07}
{"train_loss": 0.02686930075287819, "global_step": 263792, "epoch": 2963, "lr": 1.9692164644291579e-07}
{"train_loss": 0.05887511000037193, "global_step": 263793, "epoch": 2963, "lr": 1.9687024786191332e-07}
{"train_loss": 0.028911469504237175, "global_step": 263794, "epoch": 2963, "lr": 1.9681885597633842e-07}
{"train_loss": 0.045304370248753036, "global_step": 263795, "epoch": 2963, "lr": 1.9676747078619106e-07, "val_loss": 9.640524864196777}
{"train_loss": 0.0430024228990078, "global_step": 263796, "epoch": 2964, "lr": 1.967160922914879e-07}
{"train_loss": 0.05225074663758278, "global_step": 263797, "epoch": 2964, "lr": 1.966647204922234e-07}
{"train_loss": 0.018524549901485443, "global_step": 263798, "epoch": 2964, "lr": 1.966133553884142e-07}
{"train_loss": 0.024296622723340988, "global_step": 263799, "epoch": 2964, "lr": 1.965619969800603e-07}
{"train_loss": 0.03795326128602028, "global_step": 263800, "epoch": 2964, "lr": 1.965106452671728e-07}
{"train_loss": 0.02194211632013321, "global_step": 263801, "epoch": 2964, "lr": 1.9645930024975722e-07}
{"train_loss": 0.07619000971317291, "global_step": 263802, "epoch": 2964, "lr": 1.9640796192781918e-07}
{"train_loss": 0.07144425064325333, "global_step": 263803, "epoch": 2964, "lr": 1.9635663030136974e-07}
{"train_loss": 0.025811344385147095, "global_step": 263804, "epoch": 2964, "lr": 1.9630530537040892e-07}
{"train_loss": 0.0583210252225399, "global_step": 263805, "epoch": 2964, "lr": 1.9625398713495335e-07}
{"train_loss": 0.05268196016550064, "global_step": 263806, "epoch": 2964, "lr": 1.9620267559500304e-07}
{"train_loss": 0.030087517574429512, "global_step": 263807, "epoch": 2964, "lr": 1.9615137075056356e-07}
{"train_loss": 0.09308362752199173, "global_step": 263808, "epoch": 2964, "lr": 1.9610007260164598e-07}
{"train_loss": 0.07331368327140808, "global_step": 263809, "epoch": 2964, "lr": 1.9604878114825587e-07}
{"train_loss": 0.03315006196498871, "global_step": 263810, "epoch": 2964, "lr": 1.9599749639039876e-07}
{"train_loss": 0.05921068787574768, "global_step": 263811, "epoch": 2964, "lr": 1.959462183280858e-07}
{"train_loss": 0.03667598217725754, "global_step": 263812, "epoch": 2964, "lr": 1.9589494696131695e-07}
{"train_loss": 0.043872661888599396, "global_step": 263813, "epoch": 2964, "lr": 1.958436822901033e-07}
{"train_loss": 0.03812761977314949, "global_step": 263814, "epoch": 2964, "lr": 1.95792424314456e-07}
{"train_loss": 0.035312894731760025, "global_step": 263815, "epoch": 2964, "lr": 1.95741173034375e-07}
{"train_loss": 0.029296543449163437, "global_step": 263816, "epoch": 2964, "lr": 1.956899284498659e-07}
{"train_loss": 0.03654489666223526, "global_step": 263817, "epoch": 2964, "lr": 1.956386905609453e-07}
{"train_loss": 0.0466698482632637, "global_step": 263818, "epoch": 2964, "lr": 1.9558745936761324e-07}
{"train_loss": 0.042229875922203064, "global_step": 263819, "epoch": 2964, "lr": 1.955362348698697e-07}
{"train_loss": 0.026001926511526108, "global_step": 263820, "epoch": 2964, "lr": 1.9548501706773692e-07}
{"train_loss": 0.04728362709283829, "global_step": 263821, "epoch": 2964, "lr": 1.9543380596121485e-07}
{"train_loss": 0.047129590064287186, "global_step": 263822, "epoch": 2964, "lr": 1.9538260155030352e-07}
{"train_loss": 0.0387195460498333, "global_step": 263823, "epoch": 2964, "lr": 1.9533140383501958e-07}
{"train_loss": 0.04775174707174301, "global_step": 263824, "epoch": 2964, "lr": 1.9528021281536857e-07}
{"train_loss": 0.047735150903463364, "global_step": 263825, "epoch": 2964, "lr": 1.952290284913505e-07}
{"train_loss": 0.011106230318546295, "global_step": 263826, "epoch": 2964, "lr": 1.95177850862982e-07}
{"train_loss": 0.04753638431429863, "global_step": 263827, "epoch": 2964, "lr": 1.9512667993025758e-07}
{"train_loss": 0.02684760093688965, "global_step": 263828, "epoch": 2964, "lr": 1.9507551569319938e-07}
{"train_loss": 0.05581680312752724, "global_step": 263829, "epoch": 2964, "lr": 1.9502435815180187e-07}
{"train_loss": 0.03957433998584747, "global_step": 263830, "epoch": 2964, "lr": 1.9497320730607617e-07}
{"train_loss": 0.07414694875478745, "global_step": 263831, "epoch": 2964, "lr": 1.9492206315602778e-07}
{"train_loss": 0.02129312977194786, "global_step": 263832, "epoch": 2964, "lr": 1.9487092570166787e-07}
{"train_loss": 0.033736951649188995, "global_step": 263833, "epoch": 2964, "lr": 1.9481979494300195e-07}
{"train_loss": 0.11798512190580368, "global_step": 263834, "epoch": 2964, "lr": 1.9476867088003002e-07}
{"train_loss": 0.04952683299779892, "global_step": 263835, "epoch": 2964, "lr": 1.9471755351276876e-07}
{"train_loss": 0.013129157945513725, "global_step": 263836, "epoch": 2964, "lr": 1.946664428412237e-07}
{"train_loss": 0.030172444880008698, "global_step": 263837, "epoch": 2964, "lr": 1.946153388653893e-07}
{"train_loss": 0.03356390446424484, "global_step": 263838, "epoch": 2964, "lr": 1.9456424158528774e-07}
{"train_loss": 0.014557375572621822, "global_step": 263839, "epoch": 2964, "lr": 1.9451315100091904e-07}
{"train_loss": 0.02185356430709362, "global_step": 263840, "epoch": 2964, "lr": 1.9446206711229432e-07}
{"train_loss": 0.03427630662918091, "global_step": 263841, "epoch": 2964, "lr": 1.9441098991941352e-07}
{"train_loss": 0.02677999809384346, "global_step": 263842, "epoch": 2964, "lr": 1.9435991942228782e-07}
{"train_loss": 0.07706525921821594, "global_step": 263843, "epoch": 2964, "lr": 1.9430885562092272e-07}
{"train_loss": 0.05442837253212929, "global_step": 263844, "epoch": 2964, "lr": 1.942577985153293e-07}
{"train_loss": 0.03148920461535454, "global_step": 263845, "epoch": 2964, "lr": 1.9420674810550764e-07}
{"train_loss": 0.05257768929004669, "global_step": 263846, "epoch": 2964, "lr": 1.941557043914688e-07}
{"train_loss": 0.030909664928913116, "global_step": 263847, "epoch": 2964, "lr": 1.9410466737321832e-07}
{"train_loss": 0.02967418171465397, "global_step": 263848, "epoch": 2964, "lr": 1.9405363705076174e-07}
{"train_loss": 0.01980498991906643, "global_step": 263849, "epoch": 2964, "lr": 1.9400261342411018e-07}
{"train_loss": 0.03712989762425423, "global_step": 263850, "epoch": 2964, "lr": 1.9395159649326366e-07}
{"train_loss": 0.03984124958515167, "global_step": 263851, "epoch": 2964, "lr": 1.9390058625823882e-07}
{"train_loss": 0.03958265483379364, "global_step": 263852, "epoch": 2964, "lr": 1.9384958271903563e-07}
{"train_loss": 0.023523960262537003, "global_step": 263853, "epoch": 2964, "lr": 1.937985858756597e-07}
{"train_loss": 0.04282489791512489, "global_step": 263854, "epoch": 2964, "lr": 1.9374759572812208e-07}
{"train_loss": 0.021013377234339714, "global_step": 263855, "epoch": 2964, "lr": 1.9369661227642831e-07}
{"train_loss": 0.04812829568982124, "global_step": 263856, "epoch": 2964, "lr": 1.93645635520584e-07}
{"train_loss": 0.09135489165782928, "global_step": 263857, "epoch": 2964, "lr": 1.935946654605947e-07}
{"train_loss": 0.05525614321231842, "global_step": 263858, "epoch": 2964, "lr": 1.9354370209647698e-07}
{"train_loss": 0.02635897323489189, "global_step": 263859, "epoch": 2964, "lr": 1.9349274542822537e-07}
{"train_loss": 0.05283072963356972, "global_step": 263860, "epoch": 2964, "lr": 1.9344179545585094e-07}
{"train_loss": 0.028343254700303078, "global_step": 263861, "epoch": 2964, "lr": 1.9339085217935926e-07}
{"train_loss": 0.03732220456004143, "global_step": 263862, "epoch": 2964, "lr": 1.9333991559876697e-07}
{"train_loss": 0.0263401847332716, "global_step": 263863, "epoch": 2964, "lr": 1.932889857140685e-07}
{"train_loss": 0.044636212289333344, "global_step": 263864, "epoch": 2964, "lr": 1.9323806252527497e-07}
{"train_loss": 0.06660661101341248, "global_step": 263865, "epoch": 2964, "lr": 1.9318714603239196e-07}
{"train_loss": 0.016822699457406998, "global_step": 263866, "epoch": 2964, "lr": 1.9313623623542497e-07}
{"train_loss": 0.028316205367445946, "global_step": 263867, "epoch": 2964, "lr": 1.930853331343907e-07}
{"train_loss": 0.007504059933125973, "global_step": 263868, "epoch": 2964, "lr": 1.9303443672928912e-07}
{"train_loss": 0.05396154895424843, "global_step": 263869, "epoch": 2964, "lr": 1.9298354702012577e-07}
{"train_loss": 0.07957694679498672, "global_step": 263870, "epoch": 2964, "lr": 1.9293266400690623e-07}
{"train_loss": 0.062969870865345, "global_step": 263871, "epoch": 2964, "lr": 1.9288178768964161e-07}
{"train_loss": 0.07905234396457672, "global_step": 263872, "epoch": 2964, "lr": 1.9283091806833188e-07}
{"train_loss": 0.035045966506004333, "global_step": 263873, "epoch": 2964, "lr": 1.927800551429937e-07}
{"train_loss": 0.02170909196138382, "global_step": 263874, "epoch": 2964, "lr": 1.9272919891363262e-07}
{"train_loss": 0.039844341576099396, "global_step": 263875, "epoch": 2964, "lr": 1.9267834938024865e-07}
{"train_loss": 0.037133459001779556, "global_step": 263876, "epoch": 2964, "lr": 1.9262750654285289e-07}
{"train_loss": 0.041361093521118164, "global_step": 263877, "epoch": 2964, "lr": 1.9257667040144535e-07}
{"train_loss": 0.027985287830233574, "global_step": 263878, "epoch": 2964, "lr": 1.925258409560482e-07}
{"train_loss": 0.049658581614494324, "global_step": 263879, "epoch": 2964, "lr": 1.9247501820665036e-07}
{"train_loss": 0.03798382356762886, "global_step": 263880, "epoch": 2964, "lr": 1.9242420215327405e-07}
{"train_loss": 0.05518178641796112, "global_step": 263881, "epoch": 2964, "lr": 1.9237339279591926e-07}
{"train_loss": 0.038206230849027634, "global_step": 263882, "epoch": 2964, "lr": 1.9232259013458598e-07}
{"train_loss": 0.04974501207470894, "global_step": 263883, "epoch": 2964, "lr": 1.9227179416929087e-07}
{"train_loss": 0.04190693975143721, "global_step": 263884, "epoch": 2964, "lr": 1.9222100490003947e-07, "val_loss": 9.662985801696777}
{"train_loss": 0.042934779077768326, "global_step": 263885, "epoch": 2965, "lr": 1.9217022232683734e-07}
{"train_loss": 0.020449498668313026, "global_step": 263886, "epoch": 2965, "lr": 1.9211944644969003e-07}
{"train_loss": 0.048779819160699844, "global_step": 263887, "epoch": 2965, "lr": 1.920686772686031e-07}
{"train_loss": 0.05828464776277542, "global_step": 263888, "epoch": 2965, "lr": 1.9201791478358766e-07}
{"train_loss": 0.05465618520975113, "global_step": 263889, "epoch": 2965, "lr": 1.919671589946437e-07}
{"train_loss": 0.05490140989422798, "global_step": 263890, "epoch": 2965, "lr": 1.9191640990178784e-07}
{"train_loss": 0.04410345479846001, "global_step": 263891, "epoch": 2965, "lr": 1.9186566750502012e-07}
{"train_loss": 0.03731713443994522, "global_step": 263892, "epoch": 2965, "lr": 1.918149318043516e-07}
{"train_loss": 0.028107695281505585, "global_step": 263893, "epoch": 2965, "lr": 1.9176420279978237e-07}
{"train_loss": 0.012485404498875141, "global_step": 263894, "epoch": 2965, "lr": 1.91713480491329e-07}
{"train_loss": 0.023889902979135513, "global_step": 263895, "epoch": 2965, "lr": 1.9166276487898595e-07}
{"train_loss": 0.016592063009738922, "global_step": 263896, "epoch": 2965, "lr": 1.9161205596276988e-07}
{"train_loss": 0.026306023821234703, "global_step": 263897, "epoch": 2965, "lr": 1.9156135374268636e-07}
{"train_loss": 0.041931990534067154, "global_step": 263898, "epoch": 2965, "lr": 1.9151065821874092e-07}
{"train_loss": 0.04581872746348381, "global_step": 263899, "epoch": 2965, "lr": 1.914599693909336e-07}
{"train_loss": 0.04012232646346092, "global_step": 263900, "epoch": 2965, "lr": 1.9140928725928097e-07}
{"train_loss": 0.042329661548137665, "global_step": 263901, "epoch": 2965, "lr": 1.9135861182378867e-07}
{"train_loss": 0.06066499650478363, "global_step": 263902, "epoch": 2965, "lr": 1.9130794308445666e-07}
{"train_loss": 0.0320512019097805, "global_step": 263903, "epoch": 2965, "lr": 1.9125728104130158e-07}
{"train_loss": 0.03476600721478462, "global_step": 263904, "epoch": 2965, "lr": 1.912066256943179e-07}
{"train_loss": 0.011267141439020634, "global_step": 263905, "epoch": 2965, "lr": 1.911559770435223e-07}
{"train_loss": 0.03759975731372833, "global_step": 263906, "epoch": 2965, "lr": 1.9110533508892026e-07}
{"train_loss": 0.051409389823675156, "global_step": 263907, "epoch": 2965, "lr": 1.9105469983051737e-07}
{"train_loss": 0.04329109191894531, "global_step": 263908, "epoch": 2965, "lr": 1.910040712683192e-07}
{"train_loss": 0.03182069957256317, "global_step": 263909, "epoch": 2965, "lr": 1.909534494023313e-07}
{"train_loss": 0.021745959296822548, "global_step": 263910, "epoch": 2965, "lr": 1.9090283423256473e-07}
{"train_loss": 0.02798810414969921, "global_step": 263911, "epoch": 2965, "lr": 1.9085222575902505e-07}
{"train_loss": 0.028937753289937973, "global_step": 263912, "epoch": 2965, "lr": 1.9080162398171786e-07}
{"train_loss": 0.05491482466459274, "global_step": 263913, "epoch": 2965, "lr": 1.9075102890064866e-07}
{"train_loss": 0.05256703495979309, "global_step": 263914, "epoch": 2965, "lr": 1.9070044051582857e-07}
{"train_loss": 0.024980520829558372, "global_step": 263915, "epoch": 2965, "lr": 1.906498588272576e-07}
{"train_loss": 0.04148631542921066, "global_step": 263916, "epoch": 2965, "lr": 1.905992838349524e-07}
{"train_loss": 0.050817061215639114, "global_step": 263917, "epoch": 2965, "lr": 1.9054871553891295e-07}
{"train_loss": 0.024297010153532028, "global_step": 263918, "epoch": 2965, "lr": 1.9049815393914483e-07}
{"train_loss": 0.04848066717386246, "global_step": 263919, "epoch": 2965, "lr": 1.9044759903565912e-07}
{"train_loss": 0.03082178346812725, "global_step": 263920, "epoch": 2965, "lr": 1.9039705082845582e-07}
{"train_loss": 0.08392231166362762, "global_step": 263921, "epoch": 2965, "lr": 1.9034650931754606e-07}
{"train_loss": 0.05115560069680214, "global_step": 263922, "epoch": 2965, "lr": 1.902959745029409e-07}
{"train_loss": 0.05296003073453903, "global_step": 263923, "epoch": 2965, "lr": 1.9024544638464592e-07}
{"train_loss": 0.0659172385931015, "global_step": 263924, "epoch": 2965, "lr": 1.901949249626611e-07}
{"train_loss": 0.043489936739206314, "global_step": 263925, "epoch": 2965, "lr": 1.9014441023699759e-07}
{"train_loss": 0.05737326666712761, "global_step": 263926, "epoch": 2965, "lr": 1.9009390220766643e-07}
{"train_loss": 0.04300110787153244, "global_step": 263927, "epoch": 2965, "lr": 1.9004340087466766e-07}
{"train_loss": 0.04475775733590126, "global_step": 263928, "epoch": 2965, "lr": 1.899929062380068e-07}
{"train_loss": 0.06234709545969963, "global_step": 263929, "epoch": 2965, "lr": 1.89942418297695e-07}
{"train_loss": 0.054235342890024185, "global_step": 263930, "epoch": 2965, "lr": 1.8989193705374331e-07}
{"train_loss": 0.04048469290137291, "global_step": 263931, "epoch": 2965, "lr": 1.8984146250615176e-07}
{"train_loss": 0.055916737765073776, "global_step": 263932, "epoch": 2965, "lr": 1.897909946549259e-07}
{"train_loss": 0.05179222300648689, "global_step": 263933, "epoch": 2965, "lr": 1.8974053350007682e-07}
{"train_loss": 0.0076242294162511826, "global_step": 263934, "epoch": 2965, "lr": 1.8969007904161562e-07}
{"train_loss": 0.03563252463936806, "global_step": 263935, "epoch": 2965, "lr": 1.8963963127953676e-07}
{"train_loss": 0.03971449285745621, "global_step": 263936, "epoch": 2965, "lr": 1.895891902138569e-07}
{"train_loss": 0.03619268536567688, "global_step": 263937, "epoch": 2965, "lr": 1.8953875584457602e-07}
{"train_loss": 0.012604452669620514, "global_step": 263938, "epoch": 2965, "lr": 1.8948832817170526e-07}
{"train_loss": 0.003714923048391938, "global_step": 263939, "epoch": 2965, "lr": 1.8943790719525566e-07}
{"train_loss": 0.048953212797641754, "global_step": 263940, "epoch": 2965, "lr": 1.8938749291522173e-07}
{"train_loss": 0.01911478117108345, "global_step": 263941, "epoch": 2965, "lr": 1.8933708533162564e-07}
{"train_loss": 0.007894427515566349, "global_step": 263942, "epoch": 2965, "lr": 1.8928668444446184e-07}
{"train_loss": 0.03329659625887871, "global_step": 263943, "epoch": 2965, "lr": 1.8923629025374146e-07}
{"train_loss": 0.061048246920108795, "global_step": 263944, "epoch": 2965, "lr": 1.8918590275947002e-07}
{"train_loss": 0.07923877239227295, "global_step": 263945, "epoch": 2965, "lr": 1.8913552196165862e-07}
{"train_loss": 0.024419793859124184, "global_step": 263946, "epoch": 2965, "lr": 1.8908514786030728e-07}
{"train_loss": 0.022826047614216805, "global_step": 263947, "epoch": 2965, "lr": 1.890347804554271e-07}
{"train_loss": 0.04418419301509857, "global_step": 263948, "epoch": 2965, "lr": 1.8898441974702364e-07}
{"train_loss": 0.029550321400165558, "global_step": 263949, "epoch": 2965, "lr": 1.8893406573510796e-07}
{"train_loss": 0.02551059052348137, "global_step": 263950, "epoch": 2965, "lr": 1.8888371841968012e-07}
{"train_loss": 0.050652772188186646, "global_step": 263951, "epoch": 2965, "lr": 1.8883337780075117e-07}
{"train_loss": 0.06686613708734512, "global_step": 263952, "epoch": 2965, "lr": 1.8878304387832669e-07}
{"train_loss": 0.02197783999145031, "global_step": 263953, "epoch": 2965, "lr": 1.8873271665240667e-07}
{"train_loss": 0.03269370272755623, "global_step": 263954, "epoch": 2965, "lr": 1.886823961230133e-07}
{"train_loss": 0.05026503652334213, "global_step": 263955, "epoch": 2965, "lr": 1.8863208229014108e-07}
{"train_loss": 0.03246721252799034, "global_step": 263956, "epoch": 2965, "lr": 1.885817751537955e-07}
{"train_loss": 0.015595164149999619, "global_step": 263957, "epoch": 2965, "lr": 1.8853147471399325e-07}
{"train_loss": 0.04219008609652519, "global_step": 263958, "epoch": 2965, "lr": 1.884811809707343e-07}
{"train_loss": 0.018698040395975113, "global_step": 263959, "epoch": 2965, "lr": 1.884308939240298e-07}
{"train_loss": 0.03221438080072403, "global_step": 263960, "epoch": 2965, "lr": 1.883806135738797e-07}
{"train_loss": 0.026496518403291702, "global_step": 263961, "epoch": 2965, "lr": 1.8833033992029513e-07}
{"train_loss": 0.03355482220649719, "global_step": 263962, "epoch": 2965, "lr": 1.882800729632872e-07}
{"train_loss": 0.022622302174568176, "global_step": 263963, "epoch": 2965, "lr": 1.8822981270285033e-07}
{"train_loss": 0.06910624355077744, "global_step": 263964, "epoch": 2965, "lr": 1.8817955913900676e-07}
{"train_loss": 0.03825875371694565, "global_step": 263965, "epoch": 2965, "lr": 1.881293122717509e-07}
{"train_loss": 0.02697879448533058, "global_step": 263966, "epoch": 2965, "lr": 1.8807907210109388e-07}
{"train_loss": 0.032004762440919876, "global_step": 263967, "epoch": 2965, "lr": 1.8802883862704125e-07}
{"train_loss": 0.04075134918093681, "global_step": 263968, "epoch": 2965, "lr": 1.8797861184960407e-07}
{"train_loss": 0.012156374752521515, "global_step": 263969, "epoch": 2965, "lr": 1.879283917687824e-07}
{"train_loss": 0.030892280861735344, "global_step": 263970, "epoch": 2965, "lr": 1.8787817838459287e-07}
{"train_loss": 0.021944856271147728, "global_step": 263971, "epoch": 2965, "lr": 1.8782797169702993e-07}
{"train_loss": 0.039495982229709625, "global_step": 263972, "epoch": 2965, "lr": 1.8777777170611022e-07}
{"train_loss": 0.03826384722736528, "global_step": 263973, "epoch": 2965, "lr": 1.8772757841183375e-07, "val_loss": 9.667315483093262, "train_action_mse_error": 7.523408889770508}
{"train_loss": 0.06322585791349411, "global_step": 263974, "epoch": 2966, "lr": 1.8767739181421162e-07}
{"train_loss": 0.032407838851213455, "global_step": 263975, "epoch": 2966, "lr": 1.876272119132494e-07}
{"train_loss": 0.04121826961636543, "global_step": 263976, "epoch": 2966, "lr": 1.875770387089526e-07}
{"train_loss": 0.04071396589279175, "global_step": 263977, "epoch": 2966, "lr": 1.8752687220133237e-07}
{"train_loss": 0.0393315926194191, "global_step": 263978, "epoch": 2966, "lr": 1.8747671239038866e-07}
{"train_loss": 0.033183492720127106, "global_step": 263979, "epoch": 2966, "lr": 1.8742655927613262e-07}
{"train_loss": 0.05332879349589348, "global_step": 263980, "epoch": 2966, "lr": 1.8737641285856422e-07}
{"train_loss": 0.045275334268808365, "global_step": 263981, "epoch": 2966, "lr": 1.8732627313770567e-07}
{"train_loss": 0.03263228386640549, "global_step": 263982, "epoch": 2966, "lr": 1.8727614011354587e-07}
{"train_loss": 0.031180337071418762, "global_step": 263983, "epoch": 2966, "lr": 1.8722601378610704e-07}
{"train_loss": 0.023891925811767578, "global_step": 263984, "epoch": 2966, "lr": 1.871758941553836e-07}
{"train_loss": 0.025576353073120117, "global_step": 263985, "epoch": 2966, "lr": 1.8712578122138668e-07}
{"train_loss": 0.055313728749752045, "global_step": 263986, "epoch": 2966, "lr": 1.8707567498412736e-07}
{"train_loss": 0.03179877623915672, "global_step": 263987, "epoch": 2966, "lr": 1.8702557544360565e-07}
{"train_loss": 0.053805023431777954, "global_step": 263988, "epoch": 2966, "lr": 1.8697548259983267e-07}
{"train_loss": 0.030680257827043533, "global_step": 263989, "epoch": 2966, "lr": 1.8692539645281394e-07}
{"train_loss": 0.05106488615274429, "global_step": 263990, "epoch": 2966, "lr": 1.8687531700255502e-07}
{"train_loss": 0.0627312958240509, "global_step": 263991, "epoch": 2966, "lr": 1.8682524424906145e-07}
{"train_loss": 0.03944428265094757, "global_step": 263992, "epoch": 2966, "lr": 1.8677517819234436e-07}
{"train_loss": 0.025854703038930893, "global_step": 263993, "epoch": 2966, "lr": 1.867251188324093e-07}
{"train_loss": 0.03023187443614006, "global_step": 263994, "epoch": 2966, "lr": 1.8667506616926177e-07}
{"train_loss": 0.04021912440657616, "global_step": 263995, "epoch": 2966, "lr": 1.866250202029074e-07}
{"train_loss": 0.048578690737485886, "global_step": 263996, "epoch": 2966, "lr": 1.865749809333517e-07}
{"train_loss": 0.029824040830135345, "global_step": 263997, "epoch": 2966, "lr": 1.8652494836060576e-07}
{"train_loss": 0.04104838892817497, "global_step": 263998, "epoch": 2966, "lr": 1.8647492248468069e-07}
{"train_loss": 0.040396034717559814, "global_step": 263999, "epoch": 2966, "lr": 1.8642490330557094e-07}
{"train_loss": 0.033739447593688965, "global_step": 264000, "epoch": 2966, "lr": 1.8637489082328762e-07}
{"train_loss": 0.0065759518183767796, "global_step": 264001, "epoch": 2966, "lr": 1.8632488503784183e-07}
{"train_loss": 0.025100797414779663, "global_step": 264002, "epoch": 2966, "lr": 1.8627488594923914e-07}
{"train_loss": 0.0363067165017128, "global_step": 264003, "epoch": 2966, "lr": 1.8622489355747952e-07}
{"train_loss": 0.036983292549848557, "global_step": 264004, "epoch": 2966, "lr": 1.8617490786257962e-07}
{"train_loss": 0.0881587341427803, "global_step": 264005, "epoch": 2966, "lr": 1.8612492886453393e-07}
{"train_loss": 0.08147980272769928, "global_step": 264006, "epoch": 2966, "lr": 1.860749565633646e-07}
{"train_loss": 0.04206842929124832, "global_step": 264007, "epoch": 2966, "lr": 1.8602499095906612e-07}
{"train_loss": 0.038312699645757675, "global_step": 264008, "epoch": 2966, "lr": 1.8597503205165512e-07}
{"train_loss": 0.044103752821683884, "global_step": 264009, "epoch": 2966, "lr": 1.8592507984112607e-07}
{"train_loss": 0.04267464205622673, "global_step": 264010, "epoch": 2966, "lr": 1.858751343274956e-07}
{"train_loss": 0.06557055562734604, "global_step": 264011, "epoch": 2966, "lr": 1.858251955107637e-07}
{"train_loss": 0.02455901727080345, "global_step": 264012, "epoch": 2966, "lr": 1.8577526339094153e-07}
{"train_loss": 0.02962440997362137, "global_step": 264013, "epoch": 2966, "lr": 1.8572533796804014e-07}
{"train_loss": 0.06559770554304123, "global_step": 264014, "epoch": 2966, "lr": 1.85675419242054e-07}
{"train_loss": 0.052936308085918427, "global_step": 264015, "epoch": 2966, "lr": 1.8562550721299976e-07}
{"train_loss": 0.028023652732372284, "global_step": 264016, "epoch": 2966, "lr": 1.8557560188088295e-07}
{"train_loss": 0.03284308314323425, "global_step": 264017, "epoch": 2966, "lr": 1.855257032457036e-07}
{"train_loss": 0.024859309196472168, "global_step": 264018, "epoch": 2966, "lr": 1.8547581130747837e-07}
{"train_loss": 0.032289646565914154, "global_step": 264019, "epoch": 2966, "lr": 1.8542592606620722e-07}
{"train_loss": 0.021356144919991493, "global_step": 264020, "epoch": 2966, "lr": 1.8537604752189574e-07}
{"train_loss": 0.051615674048662186, "global_step": 264021, "epoch": 2966, "lr": 1.8532617567455502e-07}
{"train_loss": 0.03572177514433861, "global_step": 264022, "epoch": 2966, "lr": 1.8527631052419058e-07}
{"train_loss": 0.030211860314011574, "global_step": 264023, "epoch": 2966, "lr": 1.8522645207080803e-07}
{"train_loss": 0.02988482266664505, "global_step": 264024, "epoch": 2966, "lr": 1.8517660031441286e-07}
{"train_loss": 0.05039115622639656, "global_step": 264025, "epoch": 2966, "lr": 1.851267552550162e-07}
{"train_loss": 0.038433101028203964, "global_step": 264026, "epoch": 2966, "lr": 1.8507691689261809e-07}
{"train_loss": 0.025093114003539085, "global_step": 264027, "epoch": 2966, "lr": 1.8502708522723511e-07}
{"train_loss": 0.04774694889783859, "global_step": 264028, "epoch": 2966, "lr": 1.8497726025886176e-07}
{"train_loss": 0.051499031484127045, "global_step": 264029, "epoch": 2966, "lr": 1.8492744198751467e-07}
{"train_loss": 0.0479433573782444, "global_step": 264030, "epoch": 2966, "lr": 1.8487763041319938e-07}
{"train_loss": 0.058536626398563385, "global_step": 264031, "epoch": 2966, "lr": 1.8482782553591594e-07}
{"train_loss": 0.07425472140312195, "global_step": 264032, "epoch": 2966, "lr": 1.8477802735568095e-07}
{"train_loss": 0.03808257356286049, "global_step": 264033, "epoch": 2966, "lr": 1.847282358724889e-07}
{"train_loss": 0.05602191388607025, "global_step": 264034, "epoch": 2966, "lr": 1.846784510863564e-07}
{"train_loss": 0.05728275701403618, "global_step": 264035, "epoch": 2966, "lr": 1.8462867299728347e-07}
{"train_loss": 0.035969384014606476, "global_step": 264036, "epoch": 2966, "lr": 1.8457890160528123e-07}
{"train_loss": 0.05180547758936882, "global_step": 264037, "epoch": 2966, "lr": 1.8452913691035524e-07}
{"train_loss": 0.04831374064087868, "global_step": 264038, "epoch": 2966, "lr": 1.8447937891251655e-07}
{"train_loss": 0.07306525111198425, "global_step": 264039, "epoch": 2966, "lr": 1.844296276117652e-07}
{"train_loss": 0.053860489279031754, "global_step": 264040, "epoch": 2966, "lr": 1.843798830081067e-07}
{"train_loss": 0.019439905881881714, "global_step": 264041, "epoch": 2966, "lr": 1.8433014510155222e-07}
{"train_loss": 0.020560452714562416, "global_step": 264042, "epoch": 2966, "lr": 1.842804138921128e-07}
{"train_loss": 0.05577540770173073, "global_step": 264043, "epoch": 2966, "lr": 1.8423068937978293e-07}
{"train_loss": 0.054013945162296295, "global_step": 264044, "epoch": 2966, "lr": 1.8418097156457925e-07}
{"train_loss": 0.06696158647537231, "global_step": 264045, "epoch": 2966, "lr": 1.8413126044650176e-07}
{"train_loss": 0.017472684383392334, "global_step": 264046, "epoch": 2966, "lr": 1.840815560255671e-07}
{"train_loss": 0.07965181022882462, "global_step": 264047, "epoch": 2966, "lr": 1.8403185830176973e-07}
{"train_loss": 0.032095301896333694, "global_step": 264048, "epoch": 2966, "lr": 1.8398216727512073e-07}
{"train_loss": 0.05933001637458801, "global_step": 264049, "epoch": 2966, "lr": 1.839324829456368e-07}
{"train_loss": 0.02563057281076908, "global_step": 264050, "epoch": 2966, "lr": 1.838828053133068e-07}
{"train_loss": 0.08387254178524017, "global_step": 264051, "epoch": 2966, "lr": 1.8383313437815297e-07}
{"train_loss": 0.09824620932340622, "global_step": 264052, "epoch": 2966, "lr": 1.8378347014016972e-07}
{"train_loss": 0.07691305130720139, "global_step": 264053, "epoch": 2966, "lr": 1.8373381259937372e-07}
{"train_loss": 0.020726682618260384, "global_step": 264054, "epoch": 2966, "lr": 1.8368416175577051e-07}
{"train_loss": 0.014869839884340763, "global_step": 264055, "epoch": 2966, "lr": 1.8363451760936012e-07}
{"train_loss": 0.04078643396496773, "global_step": 264056, "epoch": 2966, "lr": 1.8358488016014807e-07}
{"train_loss": 0.06001659482717514, "global_step": 264057, "epoch": 2966, "lr": 1.8353524940815658e-07}
{"train_loss": 0.026293771341443062, "global_step": 264058, "epoch": 2966, "lr": 1.8348562535337454e-07}
{"train_loss": 0.0500810332596302, "global_step": 264059, "epoch": 2966, "lr": 1.8343600799581307e-07}
{"train_loss": 0.020989803597331047, "global_step": 264060, "epoch": 2966, "lr": 1.833863973354888e-07}
{"train_loss": 0.033557672053575516, "global_step": 264061, "epoch": 2966, "lr": 1.8333679337239617e-07}
{"train_loss": 0.04297406748041845, "global_step": 264062, "epoch": 2966, "lr": 1.8328719610654633e-07, "val_loss": 9.715352058410645}
{"train_loss": 0.02821941487491131, "global_step": 264063, "epoch": 2967, "lr": 1.8323760553795032e-07}
{"train_loss": 0.017694754526019096, "global_step": 264064, "epoch": 2967, "lr": 1.831880216666082e-07}
{"train_loss": 0.04704229161143303, "global_step": 264065, "epoch": 2967, "lr": 1.8313844449253103e-07}
{"train_loss": 0.07014733552932739, "global_step": 264066, "epoch": 2967, "lr": 1.8308887401571884e-07}
{"train_loss": 0.04908842220902443, "global_step": 264067, "epoch": 2967, "lr": 1.8303931023618824e-07}
{"train_loss": 0.060938891023397446, "global_step": 264068, "epoch": 2967, "lr": 1.829897531539393e-07}
{"train_loss": 0.017529046162962914, "global_step": 264069, "epoch": 2967, "lr": 1.8294020276898306e-07}
{"train_loss": 0.020130623131990433, "global_step": 264070, "epoch": 2967, "lr": 1.8289065908131953e-07}
{"train_loss": 0.005342427175492048, "global_step": 264071, "epoch": 2967, "lr": 1.8284112209095983e-07}
{"train_loss": 0.030742017552256584, "global_step": 264072, "epoch": 2967, "lr": 1.827915917979095e-07}
{"train_loss": 0.040885116904973984, "global_step": 264073, "epoch": 2967, "lr": 1.827420682021741e-07}
{"train_loss": 0.0517885722219944, "global_step": 264074, "epoch": 2967, "lr": 1.8269255130377028e-07}
{"train_loss": 0.03455454483628273, "global_step": 264075, "epoch": 2967, "lr": 1.8264304110268694e-07}
{"train_loss": 0.08692371845245361, "global_step": 264076, "epoch": 2967, "lr": 1.8259353759894625e-07}
{"train_loss": 0.025808723643422127, "global_step": 264077, "epoch": 2967, "lr": 1.8254404079254272e-07}
{"train_loss": 0.05215080454945564, "global_step": 264078, "epoch": 2967, "lr": 1.8249455068349296e-07}
{"train_loss": 0.042262572795152664, "global_step": 264079, "epoch": 2967, "lr": 1.8244506727179699e-07}
{"train_loss": 0.03378531336784363, "global_step": 264080, "epoch": 2967, "lr": 1.823955905574659e-07}
{"train_loss": 0.02119462378323078, "global_step": 264081, "epoch": 2967, "lr": 1.8234612054050527e-07}
{"train_loss": 0.039113257080316544, "global_step": 264082, "epoch": 2967, "lr": 1.822966572209206e-07}
{"train_loss": 0.039832957088947296, "global_step": 264083, "epoch": 2967, "lr": 1.8224720059871747e-07}
{"train_loss": 0.047070957720279694, "global_step": 264084, "epoch": 2967, "lr": 1.8219775067390145e-07}
{"train_loss": 0.013162197545170784, "global_step": 264085, "epoch": 2967, "lr": 1.8214830744648914e-07}
{"train_loss": 0.03660145029425621, "global_step": 264086, "epoch": 2967, "lr": 1.8209887091647504e-07}
{"train_loss": 0.03833595663309097, "global_step": 264087, "epoch": 2967, "lr": 1.8204944108387024e-07}
{"train_loss": 0.020401662215590477, "global_step": 264088, "epoch": 2967, "lr": 1.8200001794868026e-07}
{"train_loss": 0.0471743643283844, "global_step": 264089, "epoch": 2967, "lr": 1.8195060151091624e-07}
{"train_loss": 0.02557850442826748, "global_step": 264090, "epoch": 2967, "lr": 1.8190119177058374e-07}
{"train_loss": 0.017218979075551033, "global_step": 264091, "epoch": 2967, "lr": 1.8185178872768273e-07}
{"train_loss": 0.03438210114836693, "global_step": 264092, "epoch": 2967, "lr": 1.8180239238222986e-07}
{"train_loss": 0.027422325685620308, "global_step": 264093, "epoch": 2967, "lr": 1.8175300273421957e-07}
{"train_loss": 0.02719220705330372, "global_step": 264094, "epoch": 2967, "lr": 1.8170361978366857e-07}
{"train_loss": 0.04101581498980522, "global_step": 264095, "epoch": 2967, "lr": 1.8165424353058237e-07}
{"train_loss": 0.044815775007009506, "global_step": 264096, "epoch": 2967, "lr": 1.816048739749665e-07}
{"train_loss": 0.022266298532485962, "global_step": 264097, "epoch": 2967, "lr": 1.81555511116821e-07}
{"train_loss": 0.012707811780273914, "global_step": 264098, "epoch": 2967, "lr": 1.8150615495616252e-07}
{"train_loss": 0.03721385449171066, "global_step": 264099, "epoch": 2967, "lr": 1.8145680549299103e-07}
{"train_loss": 0.0900002270936966, "global_step": 264100, "epoch": 2967, "lr": 1.814074627273177e-07}
{"train_loss": 0.018961109220981598, "global_step": 264101, "epoch": 2967, "lr": 1.8135812665914797e-07}
{"train_loss": 0.05181955546140671, "global_step": 264102, "epoch": 2967, "lr": 1.8130879728848193e-07}
{"train_loss": 0.032495275139808655, "global_step": 264103, "epoch": 2967, "lr": 1.812594746153362e-07}
{"train_loss": 0.05339576676487923, "global_step": 264104, "epoch": 2967, "lr": 1.812101586397108e-07}
{"train_loss": 0.03340757638216019, "global_step": 264105, "epoch": 2967, "lr": 1.811608493616168e-07}
{"train_loss": 0.021070696413517, "global_step": 264106, "epoch": 2967, "lr": 1.8111154678105978e-07}
{"train_loss": 0.035759564489126205, "global_step": 264107, "epoch": 2967, "lr": 1.8106225089803975e-07}
{"train_loss": 0.058911483734846115, "global_step": 264108, "epoch": 2967, "lr": 1.8101296171257332e-07}
{"train_loss": 0.06351965665817261, "global_step": 264109, "epoch": 2967, "lr": 1.809636792246605e-07}
{"train_loss": 0.0465470515191555, "global_step": 264110, "epoch": 2967, "lr": 1.8091440343431242e-07}
{"train_loss": 0.044759731739759445, "global_step": 264111, "epoch": 2967, "lr": 1.8086513434152907e-07}
{"train_loss": 0.04933686926960945, "global_step": 264112, "epoch": 2967, "lr": 1.8081587194632155e-07}
{"train_loss": 0.03090842254459858, "global_step": 264113, "epoch": 2967, "lr": 1.8076661624870095e-07}
{"train_loss": 0.047124091535806656, "global_step": 264114, "epoch": 2967, "lr": 1.8071736724866727e-07}
{"train_loss": 0.07124799489974976, "global_step": 264115, "epoch": 2967, "lr": 1.806681249462261e-07}
{"train_loss": 0.05031463876366615, "global_step": 264116, "epoch": 2967, "lr": 1.8061888934138848e-07}
{"train_loss": 0.09133704751729965, "global_step": 264117, "epoch": 2967, "lr": 1.8056966043415446e-07}
{"train_loss": 0.018015125766396523, "global_step": 264118, "epoch": 2967, "lr": 1.8052043822454622e-07}
{"train_loss": 0.028242338448762894, "global_step": 264119, "epoch": 2967, "lr": 1.804712227125527e-07}
{"train_loss": 0.013957693241536617, "global_step": 264120, "epoch": 2967, "lr": 1.8042201389818492e-07}
{"train_loss": 0.03433792293071747, "global_step": 264121, "epoch": 2967, "lr": 1.803728117814596e-07}
{"train_loss": 0.07668809592723846, "global_step": 264122, "epoch": 2967, "lr": 1.8032361636237115e-07}
{"train_loss": 0.03080098144710064, "global_step": 264123, "epoch": 2967, "lr": 1.8027442764093072e-07}
{"train_loss": 0.07101727277040482, "global_step": 264124, "epoch": 2967, "lr": 1.8022524561714381e-07}
{"train_loss": 0.039042286574840546, "global_step": 264125, "epoch": 2967, "lr": 1.8017607029102713e-07}
{"train_loss": 0.06793159991502762, "global_step": 264126, "epoch": 2967, "lr": 1.8012690166256952e-07}
{"train_loss": 0.03695443272590637, "global_step": 264127, "epoch": 2967, "lr": 1.800777397317932e-07}
{"train_loss": 0.11016576737165451, "global_step": 264128, "epoch": 2967, "lr": 1.8002858449869264e-07}
{"train_loss": 0.044185787439346313, "global_step": 264129, "epoch": 2967, "lr": 1.7997943596328448e-07}
{"train_loss": 0.030818769708275795, "global_step": 264130, "epoch": 2967, "lr": 1.7993029412556316e-07}
{"train_loss": 0.02831483818590641, "global_step": 264131, "epoch": 2967, "lr": 1.798811589855509e-07}
{"train_loss": 0.041465193033218384, "global_step": 264132, "epoch": 2967, "lr": 1.7983203054324217e-07}
{"train_loss": 0.03865800425410271, "global_step": 264133, "epoch": 2967, "lr": 1.7978290879865355e-07}
{"train_loss": 0.04286278411746025, "global_step": 264134, "epoch": 2967, "lr": 1.7973379375177957e-07}
{"train_loss": 0.057053931057453156, "global_step": 264135, "epoch": 2967, "lr": 1.7968468540263683e-07}
{"train_loss": 0.03442353755235672, "global_step": 264136, "epoch": 2967, "lr": 1.7963558375122535e-07}
{"train_loss": 0.009574119001626968, "global_step": 264137, "epoch": 2967, "lr": 1.7958648879755624e-07}
{"train_loss": 0.08172334730625153, "global_step": 264138, "epoch": 2967, "lr": 1.7953740054163503e-07}
{"train_loss": 0.026885170489549637, "global_step": 264139, "epoch": 2967, "lr": 1.7948831898346728e-07}
{"train_loss": 0.030090775340795517, "global_step": 264140, "epoch": 2967, "lr": 1.7943924412305857e-07}
{"train_loss": 0.07917235791683197, "global_step": 264141, "epoch": 2967, "lr": 1.7939017596041995e-07}
{"train_loss": 0.07385002076625824, "global_step": 264142, "epoch": 2967, "lr": 1.7934111449555147e-07}
{"train_loss": 0.06328511238098145, "global_step": 264143, "epoch": 2967, "lr": 1.7929205972846974e-07}
{"train_loss": 0.07561001181602478, "global_step": 264144, "epoch": 2967, "lr": 1.7924301165916924e-07}
{"train_loss": 0.046160727739334106, "global_step": 264145, "epoch": 2967, "lr": 1.7919397028766105e-07}
{"train_loss": 0.03641362860798836, "global_step": 264146, "epoch": 2967, "lr": 1.7914493561395628e-07}
{"train_loss": 0.06136531010270119, "global_step": 264147, "epoch": 2967, "lr": 1.790959076380605e-07}
{"train_loss": 0.02252468653023243, "global_step": 264148, "epoch": 2967, "lr": 1.790468863599737e-07}
{"train_loss": 0.05577852204442024, "global_step": 264149, "epoch": 2967, "lr": 1.7899787177970694e-07}
{"train_loss": 0.04992831498384476, "global_step": 264150, "epoch": 2967, "lr": 1.789488638972714e-07}
{"train_loss": 0.042740762186644785, "global_step": 264151, "epoch": 2967, "lr": 1.7889986271266146e-07, "val_loss": 9.605325698852539}
{"train_loss": 0.0788009762763977, "global_step": 264152, "epoch": 2968, "lr": 1.7885086822589936e-07}
{"train_loss": 0.02876792848110199, "global_step": 264153, "epoch": 2968, "lr": 1.7880188043697954e-07}
{"train_loss": 0.06748420745134354, "global_step": 264154, "epoch": 2968, "lr": 1.787528993459131e-07}
{"train_loss": 0.011680475436151028, "global_step": 264155, "epoch": 2968, "lr": 1.787039249527056e-07}
{"train_loss": 0.03333038091659546, "global_step": 264156, "epoch": 2968, "lr": 1.7865495725736258e-07}
{"train_loss": 0.024920858442783356, "global_step": 264157, "epoch": 2968, "lr": 1.7860599625989516e-07}
{"train_loss": 0.021982070058584213, "global_step": 264158, "epoch": 2968, "lr": 1.7855704196030332e-07}
{"train_loss": 0.08231139928102493, "global_step": 264159, "epoch": 2968, "lr": 1.7850809435859817e-07}
{"train_loss": 0.05644237622618675, "global_step": 264160, "epoch": 2968, "lr": 1.7845915345478525e-07}
{"train_loss": 0.04682439938187599, "global_step": 264161, "epoch": 2968, "lr": 1.784102192488757e-07}
{"train_loss": 0.008673246949911118, "global_step": 264162, "epoch": 2968, "lr": 1.7836129174086392e-07}
{"train_loss": 0.0639994740486145, "global_step": 264163, "epoch": 2968, "lr": 1.7831237093077214e-07}
{"train_loss": 0.07058156281709671, "global_step": 264164, "epoch": 2968, "lr": 1.782634568185948e-07}
{"train_loss": 0.016552940011024475, "global_step": 264165, "epoch": 2968, "lr": 1.78214549404343e-07}
{"train_loss": 0.038714345544576645, "global_step": 264166, "epoch": 2968, "lr": 1.781656486880223e-07}
{"train_loss": 0.024912457913160324, "global_step": 264167, "epoch": 2968, "lr": 1.7811675466964383e-07}
{"train_loss": 0.0590546652674675, "global_step": 264168, "epoch": 2968, "lr": 1.7806786734920754e-07}
{"train_loss": 0.0387018620967865, "global_step": 264169, "epoch": 2968, "lr": 1.7801898672672457e-07}
{"train_loss": 0.031860217452049255, "global_step": 264170, "epoch": 2968, "lr": 1.779701128021949e-07}
{"train_loss": 0.010538184083998203, "global_step": 264171, "epoch": 2968, "lr": 1.7792124557563516e-07}
{"train_loss": 0.03773031011223793, "global_step": 264172, "epoch": 2968, "lr": 1.778723850470454e-07}
{"train_loss": 0.052383217960596085, "global_step": 264173, "epoch": 2968, "lr": 1.7782353121643114e-07}
{"train_loss": 0.02238401025533676, "global_step": 264174, "epoch": 2968, "lr": 1.777746840838035e-07}
{"train_loss": 0.04672323167324066, "global_step": 264175, "epoch": 2968, "lr": 1.7772584364916245e-07}
{"train_loss": 0.06027264520525932, "global_step": 264176, "epoch": 2968, "lr": 1.7767700991252468e-07}
{"train_loss": 0.05539585277438164, "global_step": 264177, "epoch": 2968, "lr": 1.7762818287389017e-07}
{"train_loss": 0.03491964563727379, "global_step": 264178, "epoch": 2968, "lr": 1.7757936253326446e-07}
{"train_loss": 0.05816233158111572, "global_step": 264179, "epoch": 2968, "lr": 1.7753054889065867e-07}
{"train_loss": 0.04520018398761749, "global_step": 264180, "epoch": 2968, "lr": 1.7748174194607282e-07}
{"train_loss": 0.00545481638982892, "global_step": 264181, "epoch": 2968, "lr": 1.7743294169951797e-07}
{"train_loss": 0.021151240915060043, "global_step": 264182, "epoch": 2968, "lr": 1.773841481509997e-07}
{"train_loss": 0.020253416150808334, "global_step": 264183, "epoch": 2968, "lr": 1.7733536130052907e-07}
{"train_loss": 0.016506558284163475, "global_step": 264184, "epoch": 2968, "lr": 1.7728658114810615e-07}
{"train_loss": 0.06389369070529938, "global_step": 264185, "epoch": 2968, "lr": 1.7723780769373642e-07}
{"train_loss": 0.08342202752828598, "global_step": 264186, "epoch": 2968, "lr": 1.7718904093743661e-07}
{"train_loss": 0.038925375789403915, "global_step": 264187, "epoch": 2968, "lr": 1.7714028087920108e-07}
{"train_loss": 0.045503243803977966, "global_step": 264188, "epoch": 2968, "lr": 1.7709152751904656e-07}
{"train_loss": 0.02596590481698513, "global_step": 264189, "epoch": 2968, "lr": 1.7704278085696745e-07}
{"train_loss": 0.04001106321811676, "global_step": 264190, "epoch": 2968, "lr": 1.7699404089298598e-07}
{"train_loss": 0.025720996782183647, "global_step": 264191, "epoch": 2968, "lr": 1.7694530762709105e-07}
{"train_loss": 0.05610799416899681, "global_step": 264192, "epoch": 2968, "lr": 1.7689658105930484e-07}
{"train_loss": 0.060287997126579285, "global_step": 264193, "epoch": 2968, "lr": 1.7684786118962738e-07}
{"train_loss": 0.06927908957004547, "global_step": 264194, "epoch": 2968, "lr": 1.767991480180642e-07}
{"train_loss": 0.014005842618644238, "global_step": 264195, "epoch": 2968, "lr": 1.7675044154462083e-07}
{"train_loss": 0.05822678282856941, "global_step": 264196, "epoch": 2968, "lr": 1.7670174176930844e-07}
{"train_loss": 0.027481138706207275, "global_step": 264197, "epoch": 2968, "lr": 1.7665304869213251e-07}
{"train_loss": 0.06153486296534538, "global_step": 264198, "epoch": 2968, "lr": 1.7660436231309864e-07}
{"train_loss": 0.036990076303482056, "global_step": 264199, "epoch": 2968, "lr": 1.7655568263221234e-07}
{"train_loss": 0.060027170926332474, "global_step": 264200, "epoch": 2968, "lr": 1.765070096494792e-07}
{"train_loss": 0.03204412758350372, "global_step": 264201, "epoch": 2968, "lr": 1.764583433649103e-07}
{"train_loss": 0.08586008101701736, "global_step": 264202, "epoch": 2968, "lr": 1.7640968377850564e-07}
{"train_loss": 0.02767479419708252, "global_step": 264203, "epoch": 2968, "lr": 1.7636103089027634e-07}
{"train_loss": 0.03362669795751572, "global_step": 264204, "epoch": 2968, "lr": 1.7631238470022794e-07}
{"train_loss": 0.023790258914232254, "global_step": 264205, "epoch": 2968, "lr": 1.7626374520837153e-07}
{"train_loss": 0.03349726274609566, "global_step": 264206, "epoch": 2968, "lr": 1.7621511241470156e-07}
{"train_loss": 0.045858222991228104, "global_step": 264207, "epoch": 2968, "lr": 1.7616648631924027e-07}
{"train_loss": 0.0902019739151001, "global_step": 264208, "epoch": 2968, "lr": 1.7611786692198205e-07}
{"train_loss": 0.023121988400816917, "global_step": 264209, "epoch": 2968, "lr": 1.7606925422293807e-07}
{"train_loss": 0.024019064381718636, "global_step": 264210, "epoch": 2968, "lr": 1.7602064822211383e-07}
{"train_loss": 0.012088405899703503, "global_step": 264211, "epoch": 2968, "lr": 1.759720489195149e-07}
{"train_loss": 0.050929635763168335, "global_step": 264212, "epoch": 2968, "lr": 1.759234563151524e-07}
{"train_loss": 0.026486018672585487, "global_step": 264213, "epoch": 2968, "lr": 1.758748704090263e-07}
{"train_loss": 0.0677151307463646, "global_step": 264214, "epoch": 2968, "lr": 1.758262912011477e-07}
{"train_loss": 0.02072717808187008, "global_step": 264215, "epoch": 2968, "lr": 1.7577771869152216e-07}
{"train_loss": 0.046562857925891876, "global_step": 264216, "epoch": 2968, "lr": 1.7572915288015524e-07}
{"train_loss": 0.03092716820538044, "global_step": 264217, "epoch": 2968, "lr": 1.7568059376705803e-07}
{"train_loss": 0.04051883518695831, "global_step": 264218, "epoch": 2968, "lr": 1.7563204135223054e-07}
{"train_loss": 0.0704663097858429, "global_step": 264219, "epoch": 2968, "lr": 1.755834956356839e-07}
{"train_loss": 0.03989216685295105, "global_step": 264220, "epoch": 2968, "lr": 1.7553495661741803e-07}
{"train_loss": 0.034188173711299896, "global_step": 264221, "epoch": 2968, "lr": 1.7548642429744966e-07}
{"train_loss": 0.08759135752916336, "global_step": 264222, "epoch": 2968, "lr": 1.7543789867577875e-07}
{"train_loss": 0.0453139990568161, "global_step": 264223, "epoch": 2968, "lr": 1.7538937975241088e-07}
{"train_loss": 0.029404882341623306, "global_step": 264224, "epoch": 2968, "lr": 1.7534086752735713e-07}
{"train_loss": 0.042712751775979996, "global_step": 264225, "epoch": 2968, "lr": 1.7529236200061748e-07}
{"train_loss": 0.054624225944280624, "global_step": 264226, "epoch": 2968, "lr": 1.7524386317220864e-07}
{"train_loss": 0.03788243979215622, "global_step": 264227, "epoch": 2968, "lr": 1.75195371042125e-07}
{"train_loss": 0.029039893299341202, "global_step": 264228, "epoch": 2968, "lr": 1.7514688561038328e-07}
{"train_loss": 0.03417963907122612, "global_step": 264229, "epoch": 2968, "lr": 1.750984068769834e-07}
{"train_loss": 0.0453670434653759, "global_step": 264230, "epoch": 2968, "lr": 1.75049934841931e-07}
{"train_loss": 0.05770730599761009, "global_step": 264231, "epoch": 2968, "lr": 1.7500146950524266e-07}
{"train_loss": 0.041463013738393784, "global_step": 264232, "epoch": 2968, "lr": 1.7495301086691284e-07}
{"train_loss": 0.03644780069589615, "global_step": 264233, "epoch": 2968, "lr": 1.7490455892695824e-07}
{"train_loss": 0.0450386106967926, "global_step": 264234, "epoch": 2968, "lr": 1.7485611368537324e-07}
{"train_loss": 0.0370698943734169, "global_step": 264235, "epoch": 2968, "lr": 1.7480767514218012e-07}
{"train_loss": 0.042901016771793365, "global_step": 264236, "epoch": 2968, "lr": 1.7475924329737325e-07}
{"train_loss": 0.03629043698310852, "global_step": 264237, "epoch": 2968, "lr": 1.7471081815095825e-07}
{"train_loss": 0.08718817681074142, "global_step": 264238, "epoch": 2968, "lr": 1.7466239970295174e-07}
{"train_loss": 0.050419457256793976, "global_step": 264239, "epoch": 2968, "lr": 1.746139879533537e-07}
{"train_loss": 0.042345153424219135, "global_step": 264240, "epoch": 2968, "lr": 1.7456558290216973e-07, "val_loss": 9.682998657226562}
{"train_loss": 0.055323950946331024, "global_step": 264241, "epoch": 2969, "lr": 1.745171845494109e-07}
{"train_loss": 0.05976446345448494, "global_step": 264242, "epoch": 2969, "lr": 1.7446879289507722e-07}
{"train_loss": 0.03263590857386589, "global_step": 264243, "epoch": 2969, "lr": 1.744204079391798e-07}
{"train_loss": 0.07430815696716309, "global_step": 264244, "epoch": 2969, "lr": 1.7437202968172416e-07}
{"train_loss": 0.03541814163327217, "global_step": 264245, "epoch": 2969, "lr": 1.7432365812272144e-07}
{"train_loss": 0.01898159645497799, "global_step": 264246, "epoch": 2969, "lr": 1.7427529326217162e-07}
{"train_loss": 0.047151412814855576, "global_step": 264247, "epoch": 2969, "lr": 1.7422693510008027e-07}
{"train_loss": 0.010529424995183945, "global_step": 264248, "epoch": 2969, "lr": 1.7417858363645845e-07}
{"train_loss": 0.03474636375904083, "global_step": 264249, "epoch": 2969, "lr": 1.7413023887130619e-07}
{"train_loss": 0.05291557312011719, "global_step": 264250, "epoch": 2969, "lr": 1.7408190080464572e-07}
{"train_loss": 0.07288306206464767, "global_step": 264251, "epoch": 2969, "lr": 1.740335694364603e-07}
{"train_loss": 0.06472595036029816, "global_step": 264252, "epoch": 2969, "lr": 1.739852447667778e-07}
{"train_loss": 0.015140515752136707, "global_step": 264253, "epoch": 2969, "lr": 1.7393692679559258e-07}
{"train_loss": 0.03603339567780495, "global_step": 264254, "epoch": 2969, "lr": 1.738886155229158e-07}
{"train_loss": 0.05868376046419144, "global_step": 264255, "epoch": 2969, "lr": 1.7384031094874743e-07}
{"train_loss": 0.02653205581009388, "global_step": 264256, "epoch": 2969, "lr": 1.737920130731041e-07}
{"train_loss": 0.04441104829311371, "global_step": 264257, "epoch": 2969, "lr": 1.7374372189598586e-07}
{"train_loss": 0.04536508023738861, "global_step": 264258, "epoch": 2969, "lr": 1.736954374174038e-07}
{"train_loss": 0.06088946387171745, "global_step": 264259, "epoch": 2969, "lr": 1.736471596373579e-07}
{"train_loss": 0.01904701441526413, "global_step": 264260, "epoch": 2969, "lr": 1.735988885558537e-07}
{"train_loss": 0.06766951084136963, "global_step": 264261, "epoch": 2969, "lr": 1.735506241729079e-07}
{"train_loss": 0.05447278916835785, "global_step": 264262, "epoch": 2969, "lr": 1.7350236648852047e-07}
{"train_loss": 0.03257198631763458, "global_step": 264263, "epoch": 2969, "lr": 1.7345411550269696e-07}
{"train_loss": 0.04764406010508537, "global_step": 264264, "epoch": 2969, "lr": 1.7340587121544293e-07}
{"train_loss": 0.06284104287624359, "global_step": 264265, "epoch": 2969, "lr": 1.733576336267695e-07}
{"train_loss": 0.023847166448831558, "global_step": 264266, "epoch": 2969, "lr": 1.733094027366766e-07}
{"train_loss": 0.023619910702109337, "global_step": 264267, "epoch": 2969, "lr": 1.7326117854518097e-07}
{"train_loss": 0.034602854400873184, "global_step": 264268, "epoch": 2969, "lr": 1.7321296105228257e-07}
{"train_loss": 0.032312557101249695, "global_step": 264269, "epoch": 2969, "lr": 1.7316475025798696e-07}
{"train_loss": 0.03531673923134804, "global_step": 264270, "epoch": 2969, "lr": 1.7311654616229966e-07}
{"train_loss": 0.060827042907476425, "global_step": 264271, "epoch": 2969, "lr": 1.7306834876523182e-07}
{"train_loss": 0.03914670646190643, "global_step": 264272, "epoch": 2969, "lr": 1.7302015806678896e-07}
{"train_loss": 0.04655500501394272, "global_step": 264273, "epoch": 2969, "lr": 1.7297197406697662e-07}
{"train_loss": 0.04871620610356331, "global_step": 264274, "epoch": 2969, "lr": 1.7292379676579483e-07}
{"train_loss": 0.08708535134792328, "global_step": 264275, "epoch": 2969, "lr": 1.7287562616326024e-07}
{"train_loss": 0.07890398055315018, "global_step": 264276, "epoch": 2969, "lr": 1.728274622593784e-07}
{"train_loss": 0.01812516525387764, "global_step": 264277, "epoch": 2969, "lr": 1.727793050541493e-07}
{"train_loss": 0.04236578568816185, "global_step": 264278, "epoch": 2969, "lr": 1.727311545475785e-07}
{"train_loss": 0.04977341741323471, "global_step": 264279, "epoch": 2969, "lr": 1.7268301073968263e-07}
{"train_loss": 0.03455565869808197, "global_step": 264280, "epoch": 2969, "lr": 1.726348736304617e-07}
{"train_loss": 0.06352537870407104, "global_step": 264281, "epoch": 2969, "lr": 1.725867432199213e-07}
{"train_loss": 0.06399057805538177, "global_step": 264282, "epoch": 2969, "lr": 1.725386195080725e-07}
{"train_loss": 0.019420601427555084, "global_step": 264283, "epoch": 2969, "lr": 1.7249050249490973e-07}
{"train_loss": 0.046946268528699875, "global_step": 264284, "epoch": 2969, "lr": 1.724423921804552e-07}
{"train_loss": 0.02577621303498745, "global_step": 264285, "epoch": 2969, "lr": 1.7239428856470897e-07}
{"train_loss": 0.037875667214393616, "global_step": 264286, "epoch": 2969, "lr": 1.7234619164767652e-07}
{"train_loss": 0.028814855962991714, "global_step": 264287, "epoch": 2969, "lr": 1.7229810142936343e-07}
{"train_loss": 0.0662611797451973, "global_step": 264288, "epoch": 2969, "lr": 1.7225001790978078e-07}
{"train_loss": 0.043885163962841034, "global_step": 264289, "epoch": 2969, "lr": 1.7220194108892863e-07}
{"train_loss": 0.06708330661058426, "global_step": 264290, "epoch": 2969, "lr": 1.7215387096681802e-07}
{"train_loss": 0.021405696868896484, "global_step": 264291, "epoch": 2969, "lr": 1.7210580754344896e-07}
{"train_loss": 0.05515960231423378, "global_step": 264292, "epoch": 2969, "lr": 1.7205775081883812e-07}
{"train_loss": 0.04373091086745262, "global_step": 264293, "epoch": 2969, "lr": 1.720097007929855e-07}
{"train_loss": 0.050373442471027374, "global_step": 264294, "epoch": 2969, "lr": 1.7196165746589664e-07}
{"train_loss": 0.05960161238908768, "global_step": 264295, "epoch": 2969, "lr": 1.7191362083758822e-07}
{"train_loss": 0.03200164809823036, "global_step": 264296, "epoch": 2969, "lr": 1.718655909080491e-07}
{"train_loss": 0.04622173681855202, "global_step": 264297, "epoch": 2969, "lr": 1.718175676773015e-07}
{"train_loss": 0.03203938528895378, "global_step": 264298, "epoch": 2969, "lr": 1.717695511453399e-07}
{"train_loss": 0.04871925711631775, "global_step": 264299, "epoch": 2969, "lr": 1.7172154131218642e-07}
{"train_loss": 0.02987700141966343, "global_step": 264300, "epoch": 2969, "lr": 1.7167353817783004e-07}
{"train_loss": 0.021004512906074524, "global_step": 264301, "epoch": 2969, "lr": 1.716255417422874e-07}
{"train_loss": 0.03325311094522476, "global_step": 264302, "epoch": 2969, "lr": 1.7157755200555846e-07}
{"train_loss": 0.06114627793431282, "global_step": 264303, "epoch": 2969, "lr": 1.7152956896765438e-07}
{"train_loss": 0.021252473816275597, "global_step": 264304, "epoch": 2969, "lr": 1.714815926285862e-07}
{"train_loss": 0.026379384100437164, "global_step": 264305, "epoch": 2969, "lr": 1.714336229883484e-07}
{"train_loss": 0.0397789441049099, "global_step": 264306, "epoch": 2969, "lr": 1.7138566004695766e-07}
{"train_loss": 0.05494046211242676, "global_step": 264307, "epoch": 2969, "lr": 1.713377038044195e-07}
{"train_loss": 0.019377030432224274, "global_step": 264308, "epoch": 2969, "lr": 1.712897542607339e-07}
{"train_loss": 0.054970648139715195, "global_step": 264309, "epoch": 2969, "lr": 1.71241811415912e-07}
{"train_loss": 0.02982783131301403, "global_step": 264310, "epoch": 2969, "lr": 1.7119387526995934e-07}
{"train_loss": 0.047561004757881165, "global_step": 264311, "epoch": 2969, "lr": 1.7114594582288147e-07}
{"train_loss": 0.02573123760521412, "global_step": 264312, "epoch": 2969, "lr": 1.7109802307468392e-07}
{"train_loss": 0.05535218492150307, "global_step": 264313, "epoch": 2969, "lr": 1.7105010702537783e-07}
{"train_loss": 0.01270900946110487, "global_step": 264314, "epoch": 2969, "lr": 1.7100219767496317e-07}
{"train_loss": 0.02481149695813656, "global_step": 264315, "epoch": 2969, "lr": 1.7095429502345661e-07}
{"train_loss": 0.08053115755319595, "global_step": 264316, "epoch": 2969, "lr": 1.7090639907085259e-07}
{"train_loss": 0.08258623629808426, "global_step": 264317, "epoch": 2969, "lr": 1.7085850981716778e-07}
{"train_loss": 0.032716237008571625, "global_step": 264318, "epoch": 2969, "lr": 1.708106272623966e-07}
{"train_loss": 0.027060942724347115, "global_step": 264319, "epoch": 2969, "lr": 1.7076275140655573e-07}
{"train_loss": 0.060977108776569366, "global_step": 264320, "epoch": 2969, "lr": 1.707148822496507e-07}
{"train_loss": 0.03558376803994179, "global_step": 264321, "epoch": 2969, "lr": 1.7066701979168153e-07}
{"train_loss": 0.05740763619542122, "global_step": 264322, "epoch": 2969, "lr": 1.7061916403266486e-07}
{"train_loss": 0.05454111844301224, "global_step": 264323, "epoch": 2969, "lr": 1.7057131497259515e-07}
{"train_loss": 0.06437154114246368, "global_step": 264324, "epoch": 2969, "lr": 1.7052347261148904e-07}
{"train_loss": 0.03738347440958023, "global_step": 264325, "epoch": 2969, "lr": 1.7047563694934653e-07}
{"train_loss": 0.033404234796762466, "global_step": 264326, "epoch": 2969, "lr": 1.7042780798617874e-07}
{"train_loss": 0.02437359094619751, "global_step": 264327, "epoch": 2969, "lr": 1.7037998572198565e-07}
{"train_loss": 0.039792921394109726, "global_step": 264328, "epoch": 2969, "lr": 1.703321701567784e-07}
{"train_loss": 0.043487171611089386, "global_step": 264329, "epoch": 2969, "lr": 1.7028436129056245e-07, "val_loss": 9.665196418762207}
{"train_loss": 0.04851600527763367, "global_step": 264330, "epoch": 2970, "lr": 1.7023655912334902e-07}
{"train_loss": 0.033324483782052994, "global_step": 264331, "epoch": 2970, "lr": 1.7018876365513247e-07}
{"train_loss": 0.04674912244081497, "global_step": 264332, "epoch": 2970, "lr": 1.7014097488593505e-07}
{"train_loss": 0.06003793329000473, "global_step": 264333, "epoch": 2970, "lr": 1.700931928157512e-07}
{"train_loss": 0.055551473051309586, "global_step": 264334, "epoch": 2970, "lr": 1.7004541744459202e-07}
{"train_loss": 0.057411208748817444, "global_step": 264335, "epoch": 2970, "lr": 1.699976487724575e-07}
{"train_loss": 0.07668071985244751, "global_step": 264336, "epoch": 2970, "lr": 1.699498867993643e-07}
{"train_loss": 0.03472660481929779, "global_step": 264337, "epoch": 2970, "lr": 1.6990213152531242e-07}
{"train_loss": 0.02312861941754818, "global_step": 264338, "epoch": 2970, "lr": 1.6985438295031296e-07}
{"train_loss": 0.01376744732260704, "global_step": 264339, "epoch": 2970, "lr": 1.6980664107436595e-07}
{"train_loss": 0.03540238365530968, "global_step": 264340, "epoch": 2970, "lr": 1.6975890589748245e-07}
{"train_loss": 0.05651461333036423, "global_step": 264341, "epoch": 2970, "lr": 1.6971117741966802e-07}
{"train_loss": 0.028067948296666145, "global_step": 264342, "epoch": 2970, "lr": 1.6966345564092268e-07}
{"train_loss": 0.019017307087779045, "global_step": 264343, "epoch": 2970, "lr": 1.696157405612686e-07}
{"train_loss": 0.023700745776295662, "global_step": 264344, "epoch": 2970, "lr": 1.6956803218069472e-07}
{"train_loss": 0.034875039011240005, "global_step": 264345, "epoch": 2970, "lr": 1.6952033049921767e-07}
{"train_loss": 0.020747315138578415, "global_step": 264346, "epoch": 2970, "lr": 1.69472635516843e-07}
{"train_loss": 0.01570480316877365, "global_step": 264347, "epoch": 2970, "lr": 1.6942494723357073e-07}
{"train_loss": 0.029078863561153412, "global_step": 264348, "epoch": 2970, "lr": 1.6937726564941746e-07}
{"train_loss": 0.04434286057949066, "global_step": 264349, "epoch": 2970, "lr": 1.6932959076438327e-07}
{"train_loss": 0.0274826567620039, "global_step": 264350, "epoch": 2970, "lr": 1.6928192257846808e-07}
{"train_loss": 0.07151354104280472, "global_step": 264351, "epoch": 2970, "lr": 1.6923426109169416e-07}
{"train_loss": 0.04736023023724556, "global_step": 264352, "epoch": 2970, "lr": 1.6918660630405591e-07}
{"train_loss": 0.037213072180747986, "global_step": 264353, "epoch": 2970, "lr": 1.6913895821556448e-07}
{"train_loss": 0.07394938915967941, "global_step": 264354, "epoch": 2970, "lr": 1.690913168262198e-07}
{"train_loss": 0.020680001005530357, "global_step": 264355, "epoch": 2970, "lr": 1.690436821360386e-07}
{"train_loss": 0.03092413954436779, "global_step": 264356, "epoch": 2970, "lr": 1.6899605414502085e-07}
{"train_loss": 0.04163636639714241, "global_step": 264357, "epoch": 2970, "lr": 1.6894843285317763e-07}
{"train_loss": 0.0760761946439743, "global_step": 264358, "epoch": 2970, "lr": 1.6890081826050897e-07}
{"train_loss": 0.017156049609184265, "global_step": 264359, "epoch": 2970, "lr": 1.6885321036702039e-07}
{"train_loss": 0.08891177177429199, "global_step": 264360, "epoch": 2970, "lr": 1.688056091727286e-07}
{"train_loss": 0.03871486708521843, "global_step": 264361, "epoch": 2970, "lr": 1.6875801467762798e-07}
{"train_loss": 0.016504250466823578, "global_step": 264362, "epoch": 2970, "lr": 1.6871042688173522e-07}
{"train_loss": 0.06469328701496124, "global_step": 264363, "epoch": 2970, "lr": 1.686628457850503e-07}
{"train_loss": 0.05207759141921997, "global_step": 264364, "epoch": 2970, "lr": 1.6861527138757883e-07}
{"train_loss": 0.03511988744139671, "global_step": 264365, "epoch": 2970, "lr": 1.6856770368933184e-07}
{"train_loss": 0.037224866449832916, "global_step": 264366, "epoch": 2970, "lr": 1.685201426903149e-07}
{"train_loss": 0.021833386272192, "global_step": 264367, "epoch": 2970, "lr": 1.6847258839053358e-07}
{"train_loss": 0.04509399086236954, "global_step": 264368, "epoch": 2970, "lr": 1.6842504078998789e-07}
{"train_loss": 0.060927778482437134, "global_step": 264369, "epoch": 2970, "lr": 1.6837749988869446e-07}
{"train_loss": 0.038458772003650665, "global_step": 264370, "epoch": 2970, "lr": 1.6832996568665883e-07}
{"train_loss": 0.0830121710896492, "global_step": 264371, "epoch": 2970, "lr": 1.6828243818387545e-07}
{"train_loss": 0.02641439437866211, "global_step": 264372, "epoch": 2970, "lr": 1.6823491738036656e-07}
{"train_loss": 0.03606999292969704, "global_step": 264373, "epoch": 2970, "lr": 1.6818740327612658e-07}
{"train_loss": 0.02842702530324459, "global_step": 264374, "epoch": 2970, "lr": 1.6813989587117218e-07}
{"train_loss": 0.043845728039741516, "global_step": 264375, "epoch": 2970, "lr": 1.680923951654978e-07}
{"train_loss": 0.07810056954622269, "global_step": 264376, "epoch": 2970, "lr": 1.680449011591201e-07}
{"train_loss": 0.07243619859218597, "global_step": 264377, "epoch": 2970, "lr": 1.6799741385203905e-07}
{"train_loss": 0.03116270713508129, "global_step": 264378, "epoch": 2970, "lr": 1.6794993324426023e-07}
{"train_loss": 0.05243174359202385, "global_step": 264379, "epoch": 2970, "lr": 1.679024593358003e-07}
{"train_loss": 0.02172335423529148, "global_step": 264380, "epoch": 2970, "lr": 1.6785499212665367e-07}
{"train_loss": 0.07497008889913559, "global_step": 264381, "epoch": 2970, "lr": 1.678075316168315e-07}
{"train_loss": 0.03566262125968933, "global_step": 264382, "epoch": 2970, "lr": 1.6776007780634485e-07}
{"train_loss": 0.04895477369427681, "global_step": 264383, "epoch": 2970, "lr": 1.677126306951937e-07}
{"train_loss": 0.03714531287550926, "global_step": 264384, "epoch": 2970, "lr": 1.6766519028338367e-07}
{"train_loss": 0.027474306523799896, "global_step": 264385, "epoch": 2970, "lr": 1.6761775657092583e-07}
{"train_loss": 0.0494556687772274, "global_step": 264386, "epoch": 2970, "lr": 1.6757032955782015e-07}
{"train_loss": 0.019951555877923965, "global_step": 264387, "epoch": 2970, "lr": 1.6752290924408332e-07}
{"train_loss": 0.07946963608264923, "global_step": 264388, "epoch": 2970, "lr": 1.6747549562970976e-07}
{"train_loss": 0.030792146921157837, "global_step": 264389, "epoch": 2970, "lr": 1.6742808871471617e-07}
{"train_loss": 0.04910595342516899, "global_step": 264390, "epoch": 2970, "lr": 1.6738068849910248e-07}
{"train_loss": 0.03591328486800194, "global_step": 264391, "epoch": 2970, "lr": 1.6733329498287987e-07}
{"train_loss": 0.07320371270179749, "global_step": 264392, "epoch": 2970, "lr": 1.6728590816604828e-07}
{"train_loss": 0.09903357923030853, "global_step": 264393, "epoch": 2970, "lr": 1.6723852804861883e-07}
{"train_loss": 0.04725397005677223, "global_step": 264394, "epoch": 2970, "lr": 1.6719115463059708e-07}
{"train_loss": 0.016240796074271202, "global_step": 264395, "epoch": 2970, "lr": 1.671437879119886e-07}
{"train_loss": 0.032870057970285416, "global_step": 264396, "epoch": 2970, "lr": 1.670964278927989e-07}
{"train_loss": 0.054829131811857224, "global_step": 264397, "epoch": 2970, "lr": 1.6704907457303908e-07}
{"train_loss": 0.027245841920375824, "global_step": 264398, "epoch": 2970, "lr": 1.670017279527092e-07}
{"train_loss": 0.05868211016058922, "global_step": 264399, "epoch": 2970, "lr": 1.6695438803182028e-07}
{"train_loss": 0.038262657821178436, "global_step": 264400, "epoch": 2970, "lr": 1.6690705481037795e-07}
{"train_loss": 0.0565076544880867, "global_step": 264401, "epoch": 2970, "lr": 1.6685972828838769e-07}
{"train_loss": 0.06860494613647461, "global_step": 264402, "epoch": 2970, "lr": 1.6681240846585512e-07}
{"train_loss": 0.08996061980724335, "global_step": 264403, "epoch": 2970, "lr": 1.6676509534278574e-07}
{"train_loss": 0.01979525201022625, "global_step": 264404, "epoch": 2970, "lr": 1.6671778891918512e-07}
{"train_loss": 0.039099592715501785, "global_step": 264405, "epoch": 2970, "lr": 1.6667048919506433e-07}
{"train_loss": 0.03084504045546055, "global_step": 264406, "epoch": 2970, "lr": 1.6662319617042898e-07}
{"train_loss": 0.02349410578608513, "global_step": 264407, "epoch": 2970, "lr": 1.665759098452846e-07}
{"train_loss": 0.01840021461248398, "global_step": 264408, "epoch": 2970, "lr": 1.665286302196367e-07}
{"train_loss": 0.06447891145944595, "global_step": 264409, "epoch": 2970, "lr": 1.6648135729348534e-07}
{"train_loss": 0.046967070549726486, "global_step": 264410, "epoch": 2970, "lr": 1.6643409106685272e-07}
{"train_loss": 0.02106098085641861, "global_step": 264411, "epoch": 2970, "lr": 1.6638683153972766e-07}
{"train_loss": 0.02253248170018196, "global_step": 264412, "epoch": 2970, "lr": 1.663395787121269e-07}
{"train_loss": 0.06539089977741241, "global_step": 264413, "epoch": 2970, "lr": 1.6629233258405596e-07}
{"train_loss": 0.062090467661619186, "global_step": 264414, "epoch": 2970, "lr": 1.662450931555204e-07}
{"train_loss": 0.03915794566273689, "global_step": 264415, "epoch": 2970, "lr": 1.6619786042652573e-07}
{"train_loss": 0.04295254126191139, "global_step": 264416, "epoch": 2970, "lr": 1.66150634397072e-07}
{"train_loss": 0.03168437257409096, "global_step": 264417, "epoch": 2970, "lr": 1.6610341506718142e-07}
{"train_loss": 0.044463712701134465, "global_step": 264418, "epoch": 2970, "lr": 1.6605620243684282e-07, "val_loss": 9.672564506530762, "train_action_mse_error": 11.557514190673828}
{"train_loss": 0.10103970766067505, "global_step": 264419, "epoch": 2971, "lr": 1.6600899650607294e-07}
{"train_loss": 0.03882772475481033, "global_step": 264420, "epoch": 2971, "lr": 1.6596179727487727e-07}
{"train_loss": 0.0797068327665329, "global_step": 264421, "epoch": 2971, "lr": 1.6591460474326136e-07}
{"train_loss": 0.02722352184355259, "global_step": 264422, "epoch": 2971, "lr": 1.658674189112308e-07}
{"train_loss": 0.09563064575195312, "global_step": 264423, "epoch": 2971, "lr": 1.6582023977879113e-07}
{"train_loss": 0.031096097081899643, "global_step": 264424, "epoch": 2971, "lr": 1.6577306734594789e-07}
{"train_loss": 0.03889206796884537, "global_step": 264425, "epoch": 2971, "lr": 1.6572590161271218e-07}
{"train_loss": 0.06598608195781708, "global_step": 264426, "epoch": 2971, "lr": 1.65678742579084e-07}
{"train_loss": 0.02350085973739624, "global_step": 264427, "epoch": 2971, "lr": 1.6563159024507446e-07}
{"train_loss": 0.0758344978094101, "global_step": 264428, "epoch": 2971, "lr": 1.6558444461068358e-07}
{"train_loss": 0.04552935063838959, "global_step": 264429, "epoch": 2971, "lr": 1.6553730567592795e-07}
{"train_loss": 0.03565641865134239, "global_step": 264430, "epoch": 2971, "lr": 1.6549017344080765e-07}
{"train_loss": 0.010935544036328793, "global_step": 264431, "epoch": 2971, "lr": 1.6544304790532816e-07}
{"train_loss": 0.06842074543237686, "global_step": 264432, "epoch": 2971, "lr": 1.6539592906950063e-07}
{"train_loss": 0.05385614186525345, "global_step": 264433, "epoch": 2971, "lr": 1.6534881693332504e-07}
{"train_loss": 0.05122764781117439, "global_step": 264434, "epoch": 2971, "lr": 1.653017114968125e-07}
{"train_loss": 0.0637732446193695, "global_step": 264435, "epoch": 2971, "lr": 1.6525461275996858e-07}
{"train_loss": 0.03522540256381035, "global_step": 264436, "epoch": 2971, "lr": 1.6520752072279877e-07}
{"train_loss": 0.03873066604137421, "global_step": 264437, "epoch": 2971, "lr": 1.6516043538530866e-07}
{"train_loss": 0.045526280999183655, "global_step": 264438, "epoch": 2971, "lr": 1.6511335674750382e-07}
{"train_loss": 0.07549457252025604, "global_step": 264439, "epoch": 2971, "lr": 1.6506628480939533e-07}
{"train_loss": 0.06897114217281342, "global_step": 264440, "epoch": 2971, "lr": 1.6501921957098322e-07}
{"train_loss": 0.031048830598592758, "global_step": 264441, "epoch": 2971, "lr": 1.6497216103227852e-07}
{"train_loss": 0.04490882530808449, "global_step": 264442, "epoch": 2971, "lr": 1.6492510919328686e-07}
{"train_loss": 0.042908113449811935, "global_step": 264443, "epoch": 2971, "lr": 1.648780640540082e-07}
{"train_loss": 0.020507555454969406, "global_step": 264444, "epoch": 2971, "lr": 1.6483102561445918e-07}
{"train_loss": 0.07879801839590073, "global_step": 264445, "epoch": 2971, "lr": 1.6478399387463982e-07}
{"train_loss": 0.07751720398664474, "global_step": 264446, "epoch": 2971, "lr": 1.6473696883456125e-07}
{"train_loss": 0.028544440865516663, "global_step": 264447, "epoch": 2971, "lr": 1.6468995049422343e-07}
{"train_loss": 0.04734768718481064, "global_step": 264448, "epoch": 2971, "lr": 1.646429388536319e-07}
{"train_loss": 0.01795663870871067, "global_step": 264449, "epoch": 2971, "lr": 1.6459593391280336e-07}
{"train_loss": 0.02436046116054058, "global_step": 264450, "epoch": 2971, "lr": 1.6454893567173223e-07}
{"train_loss": 0.028645072132349014, "global_step": 264451, "epoch": 2971, "lr": 1.645019441304352e-07}
{"train_loss": 0.018195580691099167, "global_step": 264452, "epoch": 2971, "lr": 1.6445495928890664e-07}
{"train_loss": 0.01717144250869751, "global_step": 264453, "epoch": 2971, "lr": 1.6440798114716883e-07}
{"train_loss": 0.06809723377227783, "global_step": 264454, "epoch": 2971, "lr": 1.6436100970521063e-07}
{"train_loss": 0.05065629631280899, "global_step": 264455, "epoch": 2971, "lr": 1.643140449630487e-07}
{"train_loss": 0.04558224231004715, "global_step": 264456, "epoch": 2971, "lr": 1.642670869206886e-07}
{"train_loss": 0.026257170364260674, "global_step": 264457, "epoch": 2971, "lr": 1.6422013557813586e-07}
{"train_loss": 0.03770516812801361, "global_step": 264458, "epoch": 2971, "lr": 1.6417319093539606e-07}
{"train_loss": 0.016013827174901962, "global_step": 264459, "epoch": 2971, "lr": 1.6412625299247476e-07}
{"train_loss": 0.0871809646487236, "global_step": 264460, "epoch": 2971, "lr": 1.6407932174938301e-07}
{"train_loss": 0.03947240486741066, "global_step": 264461, "epoch": 2971, "lr": 1.6403239720612084e-07}
{"train_loss": 0.03271611034870148, "global_step": 264462, "epoch": 2971, "lr": 1.6398547936269383e-07}
{"train_loss": 0.06744453310966492, "global_step": 264463, "epoch": 2971, "lr": 1.6393856821911857e-07}
{"train_loss": 0.05834600701928139, "global_step": 264464, "epoch": 2971, "lr": 1.6389166377538957e-07}
{"train_loss": 0.07317785173654556, "global_step": 264465, "epoch": 2971, "lr": 1.6384476603151232e-07}
{"train_loss": 0.018328983336687088, "global_step": 264466, "epoch": 2971, "lr": 1.637978749875091e-07}
{"train_loss": 0.037851180881261826, "global_step": 264467, "epoch": 2971, "lr": 1.6375099064336873e-07}
{"train_loss": 0.028018828481435776, "global_step": 264468, "epoch": 2971, "lr": 1.6370411299910792e-07}
{"train_loss": 0.024877939373254776, "global_step": 264469, "epoch": 2971, "lr": 1.636572420547322e-07}
{"train_loss": 0.0619656965136528, "global_step": 264470, "epoch": 2971, "lr": 1.6361037781024158e-07}
{"train_loss": 0.09438139200210571, "global_step": 264471, "epoch": 2971, "lr": 1.6356352026564714e-07}
{"train_loss": 0.014529438689351082, "global_step": 264472, "epoch": 2971, "lr": 1.6351666942095445e-07}
{"train_loss": 0.05004967376589775, "global_step": 264473, "epoch": 2971, "lr": 1.6346982527616905e-07}
{"train_loss": 0.027889493852853775, "global_step": 264474, "epoch": 2971, "lr": 1.634229878312965e-07}
{"train_loss": 0.046866584569215775, "global_step": 264475, "epoch": 2971, "lr": 1.6337615708634236e-07}
{"train_loss": 0.08269888162612915, "global_step": 264476, "epoch": 2971, "lr": 1.6332933304132325e-07}
{"train_loss": 0.024955708533525467, "global_step": 264477, "epoch": 2971, "lr": 1.632825156962281e-07}
{"train_loss": 0.01822688803076744, "global_step": 264478, "epoch": 2971, "lr": 1.632357050510791e-07}
{"train_loss": 0.09438340365886688, "global_step": 264479, "epoch": 2971, "lr": 1.631889011058707e-07}
{"train_loss": 0.04463006556034088, "global_step": 264480, "epoch": 2971, "lr": 1.63142103860614e-07}
{"train_loss": 0.02800735831260681, "global_step": 264481, "epoch": 2971, "lr": 1.6309531331532013e-07}
{"train_loss": 0.02544655278325081, "global_step": 264482, "epoch": 2971, "lr": 1.6304852946998905e-07}
{"train_loss": 0.02552860602736473, "global_step": 264483, "epoch": 2971, "lr": 1.6300175232462632e-07}
{"train_loss": 0.03983078524470329, "global_step": 264484, "epoch": 2971, "lr": 1.6295498187924307e-07}
{"train_loss": 0.08101245015859604, "global_step": 264485, "epoch": 2971, "lr": 1.6290821813383927e-07}
{"train_loss": 0.04660521820187569, "global_step": 264486, "epoch": 2971, "lr": 1.6286146108842603e-07}
{"train_loss": 0.013806654140353203, "global_step": 264487, "epoch": 2971, "lr": 1.628147107430089e-07}
{"train_loss": 0.050612661987543106, "global_step": 264488, "epoch": 2971, "lr": 1.62767967097599e-07}
{"train_loss": 0.042189545929431915, "global_step": 264489, "epoch": 2971, "lr": 1.6272123015219631e-07}
{"train_loss": 0.048142608255147934, "global_step": 264490, "epoch": 2971, "lr": 1.6267449990680084e-07}
{"train_loss": 0.017133934423327446, "global_step": 264491, "epoch": 2971, "lr": 1.6262777636143478e-07}
{"train_loss": 0.03277772292494774, "global_step": 264492, "epoch": 2971, "lr": 1.625810595160926e-07}
{"train_loss": 0.05457251891493797, "global_step": 264493, "epoch": 2971, "lr": 1.6253434937077982e-07}
{"train_loss": 0.09813959151506424, "global_step": 264494, "epoch": 2971, "lr": 1.6248764592551315e-07}
{"train_loss": 0.035094793885946274, "global_step": 264495, "epoch": 2971, "lr": 1.6244094918029252e-07}
{"train_loss": 0.013907453045248985, "global_step": 264496, "epoch": 2971, "lr": 1.6239425913511796e-07}
{"train_loss": 0.05259397253394127, "global_step": 264497, "epoch": 2971, "lr": 1.6234757579000615e-07}
{"train_loss": 0.06806951016187668, "global_step": 264498, "epoch": 2971, "lr": 1.623008991449626e-07}
{"train_loss": 0.06355036795139313, "global_step": 264499, "epoch": 2971, "lr": 1.6225422919998734e-07}
{"train_loss": 0.037428583949804306, "global_step": 264500, "epoch": 2971, "lr": 1.622075659550859e-07}
{"train_loss": 0.06252581626176834, "global_step": 264501, "epoch": 2971, "lr": 1.6216090941026941e-07}
{"train_loss": 0.027286076918244362, "global_step": 264502, "epoch": 2971, "lr": 1.6211425956554892e-07}
{"train_loss": 0.02232237160205841, "global_step": 264503, "epoch": 2971, "lr": 1.6206761642091894e-07}
{"train_loss": 0.057904861867427826, "global_step": 264504, "epoch": 2971, "lr": 1.6202097997639054e-07}
{"train_loss": 0.0367477722465992, "global_step": 264505, "epoch": 2971, "lr": 1.619743502319748e-07}
{"train_loss": 0.029165618121623993, "global_step": 264506, "epoch": 2971, "lr": 1.6192772718767734e-07}
{"train_loss": 0.04548510213204649, "global_step": 264507, "epoch": 2971, "lr": 1.6188111084349256e-07, "val_loss": 9.62034797668457}
{"train_loss": 0.040326859802007675, "global_step": 264508, "epoch": 2972, "lr": 1.6183450119944265e-07}
{"train_loss": 0.08293672651052475, "global_step": 264509, "epoch": 2972, "lr": 1.617878982555221e-07}
{"train_loss": 0.0222631748765707, "global_step": 264510, "epoch": 2972, "lr": 1.6174130201174197e-07}
{"train_loss": 0.05831390246748924, "global_step": 264511, "epoch": 2972, "lr": 1.616947124681134e-07}
{"train_loss": 0.09332410246133804, "global_step": 264512, "epoch": 2972, "lr": 1.6164812962463083e-07}
{"train_loss": 0.021821703761816025, "global_step": 264513, "epoch": 2972, "lr": 1.616015534813109e-07}
{"train_loss": 0.037628356367349625, "global_step": 264514, "epoch": 2972, "lr": 1.6155498403815917e-07}
{"train_loss": 0.0561494305729866, "global_step": 264515, "epoch": 2972, "lr": 1.6150842129517008e-07}
{"train_loss": 0.029806194826960564, "global_step": 264516, "epoch": 2972, "lr": 1.6146186525236585e-07}
{"train_loss": 0.03550376743078232, "global_step": 264517, "epoch": 2972, "lr": 1.6141531590974644e-07}
{"train_loss": 0.05051214247941971, "global_step": 264518, "epoch": 2972, "lr": 1.6136877326731192e-07}
{"train_loss": 0.03104584664106369, "global_step": 264519, "epoch": 2972, "lr": 1.613222373250789e-07}
{"train_loss": 0.04611070454120636, "global_step": 264520, "epoch": 2972, "lr": 1.6127570808304183e-07}
{"train_loss": 0.04113630950450897, "global_step": 264521, "epoch": 2972, "lr": 1.612291855412229e-07}
{"train_loss": 0.031752705574035645, "global_step": 264522, "epoch": 2972, "lr": 1.611826696996166e-07}
{"train_loss": 0.03125804662704468, "global_step": 264523, "epoch": 2972, "lr": 1.6113616055822845e-07}
{"train_loss": 0.054218556731939316, "global_step": 264524, "epoch": 2972, "lr": 1.6108965811706956e-07}
{"train_loss": 0.024505365639925003, "global_step": 264525, "epoch": 2972, "lr": 1.6104316237614547e-07}
{"train_loss": 0.04160577431321144, "global_step": 264526, "epoch": 2972, "lr": 1.6099667333546176e-07}
{"train_loss": 0.07268457859754562, "global_step": 264527, "epoch": 2972, "lr": 1.6095019099502395e-07}
{"train_loss": 0.039186008274555206, "global_step": 264528, "epoch": 2972, "lr": 1.609037153548376e-07}
{"train_loss": 0.02696308307349682, "global_step": 264529, "epoch": 2972, "lr": 1.6085724641491384e-07}
{"train_loss": 0.02190650813281536, "global_step": 264530, "epoch": 2972, "lr": 1.6081078417524707e-07}
{"train_loss": 0.03876296803355217, "global_step": 264531, "epoch": 2972, "lr": 1.6076432863585956e-07}
{"train_loss": 0.024705154821276665, "global_step": 264532, "epoch": 2972, "lr": 1.6071787979675123e-07}
{"train_loss": 0.09526518732309341, "global_step": 264533, "epoch": 2972, "lr": 1.6067143765792214e-07}
{"train_loss": 0.01993628963828087, "global_step": 264534, "epoch": 2972, "lr": 1.6062500221938336e-07}
{"train_loss": 0.027250686660408974, "global_step": 264535, "epoch": 2972, "lr": 1.60578573481146e-07}
{"train_loss": 0.009522872045636177, "global_step": 264536, "epoch": 2972, "lr": 1.6053215144320455e-07}
{"train_loss": 0.061433471739292145, "global_step": 264537, "epoch": 2972, "lr": 1.604857361055756e-07}
{"train_loss": 0.07498504221439362, "global_step": 264538, "epoch": 2972, "lr": 1.6043932746826474e-07}
{"train_loss": 0.031014880165457726, "global_step": 264539, "epoch": 2972, "lr": 1.6039292553127195e-07}
{"train_loss": 0.034314483404159546, "global_step": 264540, "epoch": 2972, "lr": 1.6034653029460834e-07}
{"train_loss": 0.0372745618224144, "global_step": 264541, "epoch": 2972, "lr": 1.603001417582739e-07}
{"train_loss": 0.06891889125108719, "global_step": 264542, "epoch": 2972, "lr": 1.602537599222853e-07}
{"train_loss": 0.030925877392292023, "global_step": 264543, "epoch": 2972, "lr": 1.6020738478664253e-07}
{"train_loss": 0.04503122717142105, "global_step": 264544, "epoch": 2972, "lr": 1.6016101635135118e-07}
{"train_loss": 0.07379166781902313, "global_step": 264545, "epoch": 2972, "lr": 1.6011465461641673e-07}
{"train_loss": 0.03701689466834068, "global_step": 264546, "epoch": 2972, "lr": 1.6006829958185033e-07}
{"train_loss": 0.06527484953403473, "global_step": 264547, "epoch": 2972, "lr": 1.6002195124765196e-07}
{"train_loss": 0.018673522397875786, "global_step": 264548, "epoch": 2972, "lr": 1.5997560961383828e-07}
{"train_loss": 0.01333867572247982, "global_step": 264549, "epoch": 2972, "lr": 1.599292746803982e-07}
{"train_loss": 0.015245672315359116, "global_step": 264550, "epoch": 2972, "lr": 1.5988294644735946e-07}
{"train_loss": 0.038269445300102234, "global_step": 264551, "epoch": 2972, "lr": 1.5983662491470542e-07}
{"train_loss": 0.025424201041460037, "global_step": 264552, "epoch": 2972, "lr": 1.597903100824638e-07}
{"train_loss": 0.0500156506896019, "global_step": 264553, "epoch": 2972, "lr": 1.5974400195062355e-07}
{"train_loss": 0.07431622594594955, "global_step": 264554, "epoch": 2972, "lr": 1.5969770051920685e-07}
{"train_loss": 0.04704009369015694, "global_step": 264555, "epoch": 2972, "lr": 1.5965140578820258e-07}
{"train_loss": 0.05826180800795555, "global_step": 264556, "epoch": 2972, "lr": 1.5960511775762743e-07}
{"train_loss": 0.01539489347487688, "global_step": 264557, "epoch": 2972, "lr": 1.5955883642748693e-07}
{"train_loss": 0.02239670790731907, "global_step": 264558, "epoch": 2972, "lr": 1.595125617977866e-07}
{"train_loss": 0.009513200260698795, "global_step": 264559, "epoch": 2972, "lr": 1.5946629386853762e-07}
{"train_loss": 0.02533154934644699, "global_step": 264560, "epoch": 2972, "lr": 1.5942003263973438e-07}
{"train_loss": 0.017830146476626396, "global_step": 264561, "epoch": 2972, "lr": 1.5937377811138798e-07}
{"train_loss": 0.0966116189956665, "global_step": 264562, "epoch": 2972, "lr": 1.593275302835151e-07}
{"train_loss": 0.04676651209592819, "global_step": 264563, "epoch": 2972, "lr": 1.5928128915610462e-07}
{"train_loss": 0.07204676419496536, "global_step": 264564, "epoch": 2972, "lr": 1.5923505472917323e-07}
{"train_loss": 0.05166502296924591, "global_step": 264565, "epoch": 2972, "lr": 1.591888270027264e-07}
{"train_loss": 0.04789349436759949, "global_step": 264566, "epoch": 2972, "lr": 1.591426059767698e-07}
{"train_loss": 0.0474373921751976, "global_step": 264567, "epoch": 2972, "lr": 1.5909639165130886e-07}
{"train_loss": 0.027134176343679428, "global_step": 264568, "epoch": 2972, "lr": 1.590501840263492e-07}
{"train_loss": 0.041342493146657944, "global_step": 264569, "epoch": 2972, "lr": 1.590039831019019e-07}
{"train_loss": 0.036548517644405365, "global_step": 264570, "epoch": 2972, "lr": 1.5895778887796141e-07}
{"train_loss": 0.00968803558498621, "global_step": 264571, "epoch": 2972, "lr": 1.589116013545444e-07}
{"train_loss": 0.030361630022525787, "global_step": 264572, "epoch": 2972, "lr": 1.5886542053165643e-07}
{"train_loss": 0.05632131174206734, "global_step": 264573, "epoch": 2972, "lr": 1.5881924640929747e-07}
{"train_loss": 0.024186687543988228, "global_step": 264574, "epoch": 2972, "lr": 1.5877307898748417e-07}
{"train_loss": 0.04685520753264427, "global_step": 264575, "epoch": 2972, "lr": 1.58726918266211e-07}
{"train_loss": 0.035620905458927155, "global_step": 264576, "epoch": 2972, "lr": 1.5868076424548905e-07}
{"train_loss": 0.05552184209227562, "global_step": 264577, "epoch": 2972, "lr": 1.5863461692532944e-07}
{"train_loss": 0.04353322461247444, "global_step": 264578, "epoch": 2972, "lr": 1.5858847630573214e-07}
{"train_loss": 0.07714837044477463, "global_step": 264579, "epoch": 2972, "lr": 1.5854234238670274e-07}
{"train_loss": 0.05639002099633217, "global_step": 264580, "epoch": 2972, "lr": 1.584962151682523e-07}
{"train_loss": 0.04436693713068962, "global_step": 264581, "epoch": 2972, "lr": 1.5845009465038084e-07}
{"train_loss": 0.0637374073266983, "global_step": 264582, "epoch": 2972, "lr": 1.5840398083310503e-07}
{"train_loss": 0.06571156531572342, "global_step": 264583, "epoch": 2972, "lr": 1.583578737164193e-07}
{"train_loss": 0.015267626382410526, "global_step": 264584, "epoch": 2972, "lr": 1.5831177330033474e-07}
{"train_loss": 0.0347316674888134, "global_step": 264585, "epoch": 2972, "lr": 1.5826567958486248e-07}
{"train_loss": 0.04710770398378372, "global_step": 264586, "epoch": 2972, "lr": 1.582195925700025e-07}
{"train_loss": 0.08585479855537415, "global_step": 264587, "epoch": 2972, "lr": 1.5817351225575482e-07}
{"train_loss": 0.09482639282941818, "global_step": 264588, "epoch": 2972, "lr": 1.5812743864214163e-07}
{"train_loss": 0.020273244008421898, "global_step": 264589, "epoch": 2972, "lr": 1.5808137172915737e-07}
{"train_loss": 0.06692851334810257, "global_step": 264590, "epoch": 2972, "lr": 1.5803531151681317e-07}
{"train_loss": 0.0870310440659523, "global_step": 264591, "epoch": 2972, "lr": 1.5798925800511455e-07}
{"train_loss": 0.02869141660630703, "global_step": 264592, "epoch": 2972, "lr": 1.5794321119406152e-07}
{"train_loss": 0.03421768546104431, "global_step": 264593, "epoch": 2972, "lr": 1.5789717108367075e-07}
{"train_loss": 0.0939084142446518, "global_step": 264594, "epoch": 2972, "lr": 1.5785113767393668e-07}
{"train_loss": 0.045970041304826736, "global_step": 264595, "epoch": 2972, "lr": 1.578051109648815e-07}
{"train_loss": 0.044594192626268674, "global_step": 264596, "epoch": 2972, "lr": 1.577590909564941e-07, "val_loss": 9.579144477844238}
{"train_loss": 0.05283280834555626, "global_step": 264597, "epoch": 2973, "lr": 1.5771307764879118e-07}
{"train_loss": 0.033815622329711914, "global_step": 264598, "epoch": 2973, "lr": 1.5766707104177826e-07}
{"train_loss": 0.020173165947198868, "global_step": 264599, "epoch": 2973, "lr": 1.5762107113546087e-07}
{"train_loss": 0.060760993510484695, "global_step": 264600, "epoch": 2973, "lr": 1.5757507792983904e-07}
{"train_loss": 0.02891293726861477, "global_step": 264601, "epoch": 2973, "lr": 1.575290914249239e-07}
{"train_loss": 0.043658576905727386, "global_step": 264602, "epoch": 2973, "lr": 1.5748311162072648e-07}
{"train_loss": 0.03453006222844124, "global_step": 264603, "epoch": 2973, "lr": 1.5743713851724128e-07}
{"train_loss": 0.03104206547141075, "global_step": 264604, "epoch": 2973, "lr": 1.5739117211448494e-07}
{"train_loss": 0.06262324750423431, "global_step": 264605, "epoch": 2973, "lr": 1.5734521241246302e-07}
{"train_loss": 0.07865893840789795, "global_step": 264606, "epoch": 2973, "lr": 1.5729925941117552e-07}
{"train_loss": 0.05095125362277031, "global_step": 264607, "epoch": 2973, "lr": 1.5725331311062796e-07}
{"train_loss": 0.03655119240283966, "global_step": 264608, "epoch": 2973, "lr": 1.5720737351083148e-07}
{"train_loss": 0.023270875215530396, "global_step": 264609, "epoch": 2973, "lr": 1.5716144061179715e-07}
{"train_loss": 0.015779653564095497, "global_step": 264610, "epoch": 2973, "lr": 1.5711551441351947e-07}
{"train_loss": 0.06308303028345108, "global_step": 264611, "epoch": 2973, "lr": 1.5706959491600949e-07}
{"train_loss": 0.051055118441581726, "global_step": 264612, "epoch": 2973, "lr": 1.570236821192783e-07}
{"train_loss": 0.03341999650001526, "global_step": 264613, "epoch": 2973, "lr": 1.5697777602332043e-07}
{"train_loss": 0.06645545363426208, "global_step": 264614, "epoch": 2973, "lr": 1.56931876628158e-07}
{"train_loss": 0.0363180935382843, "global_step": 264615, "epoch": 2973, "lr": 1.5688598393378552e-07}
{"train_loss": 0.10430646687746048, "global_step": 264616, "epoch": 2973, "lr": 1.5684009794020848e-07}
{"train_loss": 0.05746332183480263, "global_step": 264617, "epoch": 2973, "lr": 1.5679421864743804e-07}
{"train_loss": 0.04918871447443962, "global_step": 264618, "epoch": 2973, "lr": 1.5674834605548527e-07}
{"train_loss": 0.03670717775821686, "global_step": 264619, "epoch": 2973, "lr": 1.5670248016434463e-07}
{"train_loss": 0.05572899058461189, "global_step": 264620, "epoch": 2973, "lr": 1.5665662097402722e-07}
{"train_loss": 0.06562621891498566, "global_step": 264621, "epoch": 2973, "lr": 1.5661076848454414e-07}
{"train_loss": 0.033649202436208725, "global_step": 264622, "epoch": 2973, "lr": 1.565649226958954e-07}
{"train_loss": 0.03250296786427498, "global_step": 264623, "epoch": 2973, "lr": 1.5651908360808652e-07}
{"train_loss": 0.03967903181910515, "global_step": 264624, "epoch": 2973, "lr": 1.5647325122112866e-07}
{"train_loss": 0.024561023339629173, "global_step": 264625, "epoch": 2973, "lr": 1.5642742553502178e-07}
{"train_loss": 0.06713145226240158, "global_step": 264626, "epoch": 2973, "lr": 1.5638160654978251e-07}
{"train_loss": 0.058981768786907196, "global_step": 264627, "epoch": 2973, "lr": 1.5633579426540533e-07}
{"train_loss": 0.029698578640818596, "global_step": 264628, "epoch": 2973, "lr": 1.5628998868190137e-07}
{"train_loss": 0.02763572707772255, "global_step": 264629, "epoch": 2973, "lr": 1.5624418979927612e-07}
{"train_loss": 0.04571027308702469, "global_step": 264630, "epoch": 2973, "lr": 1.5619839761753518e-07}
{"train_loss": 0.03097015619277954, "global_step": 264631, "epoch": 2973, "lr": 1.5615261213668963e-07}
{"train_loss": 0.034188877791166306, "global_step": 264632, "epoch": 2973, "lr": 1.5610683335673948e-07}
{"train_loss": 0.08287199586629868, "global_step": 264633, "epoch": 2973, "lr": 1.5606106127769583e-07}
{"train_loss": 0.09300630539655685, "global_step": 264634, "epoch": 2973, "lr": 1.560152958995642e-07}
{"train_loss": 0.04341943934559822, "global_step": 264635, "epoch": 2973, "lr": 1.5596953722234464e-07}
{"train_loss": 0.022977063432335854, "global_step": 264636, "epoch": 2973, "lr": 1.5592378524604822e-07}
{"train_loss": 0.03290069103240967, "global_step": 264637, "epoch": 2973, "lr": 1.558780399706805e-07}
{"train_loss": 0.04643513634800911, "global_step": 264638, "epoch": 2973, "lr": 1.5583230139624704e-07}
{"train_loss": 0.06058830767869949, "global_step": 264639, "epoch": 2973, "lr": 1.557865695227534e-07}
{"train_loss": 0.01832214929163456, "global_step": 264640, "epoch": 2973, "lr": 1.5574084435021062e-07}
{"train_loss": 0.02430632896721363, "global_step": 264641, "epoch": 2973, "lr": 1.5569512587861878e-07}
{"train_loss": 0.014190860092639923, "global_step": 264642, "epoch": 2973, "lr": 1.556494141079834e-07}
{"train_loss": 0.027251221239566803, "global_step": 264643, "epoch": 2973, "lr": 1.5560370903832112e-07}
{"train_loss": 0.04064261540770531, "global_step": 264644, "epoch": 2973, "lr": 1.5555801066962084e-07}
{"train_loss": 0.035637274384498596, "global_step": 264645, "epoch": 2973, "lr": 1.5551231900190478e-07}
{"train_loss": 0.024812830612063408, "global_step": 264646, "epoch": 2973, "lr": 1.5546663403516737e-07}
{"train_loss": 0.020005110651254654, "global_step": 264647, "epoch": 2973, "lr": 1.5542095576942527e-07}
{"train_loss": 0.05554397776722908, "global_step": 264648, "epoch": 2973, "lr": 1.553752842046785e-07}
{"train_loss": 0.040838904678821564, "global_step": 264649, "epoch": 2973, "lr": 1.553296193409326e-07}
{"train_loss": 0.04495357349514961, "global_step": 264650, "epoch": 2973, "lr": 1.552839611781931e-07}
{"train_loss": 0.04109659045934677, "global_step": 264651, "epoch": 2973, "lr": 1.5523830971647114e-07}
{"train_loss": 0.04260944575071335, "global_step": 264652, "epoch": 2973, "lr": 1.5519266495576668e-07}
{"train_loss": 0.015955917537212372, "global_step": 264653, "epoch": 2973, "lr": 1.5514702689609084e-07}
{"train_loss": 0.05416596308350563, "global_step": 264654, "epoch": 2973, "lr": 1.551013955374492e-07}
{"train_loss": 0.053974807262420654, "global_step": 264655, "epoch": 2973, "lr": 1.5505577087984724e-07}
{"train_loss": 0.06225094571709633, "global_step": 264656, "epoch": 2973, "lr": 1.550101529232906e-07}
{"train_loss": 0.05124621465802193, "global_step": 264657, "epoch": 2973, "lr": 1.549645416677792e-07}
{"train_loss": 0.036399371922016144, "global_step": 264658, "epoch": 2973, "lr": 1.5491893711333528e-07}
{"train_loss": 0.07037489861249924, "global_step": 264659, "epoch": 2973, "lr": 1.5487333925994774e-07}
{"train_loss": 0.0527605339884758, "global_step": 264660, "epoch": 2973, "lr": 1.5482774810763322e-07}
{"train_loss": 0.0412323884665966, "global_step": 264661, "epoch": 2973, "lr": 1.5478216365639174e-07}
{"train_loss": 0.05063745006918907, "global_step": 264662, "epoch": 2973, "lr": 1.547365859062344e-07}
{"train_loss": 0.02085857279598713, "global_step": 264663, "epoch": 2973, "lr": 1.546910148571612e-07}
{"train_loss": 0.006705311127007008, "global_step": 264664, "epoch": 2973, "lr": 1.5464545050918878e-07}
{"train_loss": 0.013250492513179779, "global_step": 264665, "epoch": 2973, "lr": 1.5459989286231157e-07}
{"train_loss": 0.04906207323074341, "global_step": 264666, "epoch": 2973, "lr": 1.5455434191654072e-07}
{"train_loss": 0.013609710149466991, "global_step": 264667, "epoch": 2973, "lr": 1.5450879767188731e-07}
{"train_loss": 0.0466962493956089, "global_step": 264668, "epoch": 2973, "lr": 1.5446326012835134e-07}
{"train_loss": 0.02760312706232071, "global_step": 264669, "epoch": 2973, "lr": 1.5441772928593833e-07}
{"train_loss": 0.05918598175048828, "global_step": 264670, "epoch": 2973, "lr": 1.5437220514465388e-07}
{"train_loss": 0.07208607345819473, "global_step": 264671, "epoch": 2973, "lr": 1.543266877045091e-07}
{"train_loss": 0.0727781280875206, "global_step": 264672, "epoch": 2973, "lr": 1.542811769655095e-07}
{"train_loss": 0.09126589447259903, "global_step": 264673, "epoch": 2973, "lr": 1.5423567292766062e-07}
{"train_loss": 0.03966682404279709, "global_step": 264674, "epoch": 2973, "lr": 1.541901755909625e-07}
{"train_loss": 0.07344137877225876, "global_step": 264675, "epoch": 2973, "lr": 1.5414468495542622e-07}
{"train_loss": 0.049638863652944565, "global_step": 264676, "epoch": 2973, "lr": 1.540992010210629e-07}
{"train_loss": 0.05254295840859413, "global_step": 264677, "epoch": 2973, "lr": 1.5405372378786697e-07}
{"train_loss": 0.05847462639212608, "global_step": 264678, "epoch": 2973, "lr": 1.5400825325585512e-07}
{"train_loss": 0.02826927788555622, "global_step": 264679, "epoch": 2973, "lr": 1.5396278942502728e-07}
{"train_loss": 0.045671939849853516, "global_step": 264680, "epoch": 2973, "lr": 1.5391733229539462e-07}
{"train_loss": 0.03229331970214844, "global_step": 264681, "epoch": 2973, "lr": 1.538718818669571e-07}
{"train_loss": 0.04897608235478401, "global_step": 264682, "epoch": 2973, "lr": 1.5382643813972586e-07}
{"train_loss": 0.0690038651227951, "global_step": 264683, "epoch": 2973, "lr": 1.5378100111370085e-07}
{"train_loss": 0.056546494364738464, "global_step": 264684, "epoch": 2973, "lr": 1.5373557078889877e-07}
{"train_loss": 0.04482998380835137, "global_step": 264685, "epoch": 2973, "lr": 1.5369014716531404e-07, "val_loss": 9.644609451293945}
{"train_loss": 0.044336311519145966, "global_step": 264686, "epoch": 2974, "lr": 1.5364473024296332e-07}
{"train_loss": 0.028383834287524223, "global_step": 264687, "epoch": 2974, "lr": 1.5359932002184108e-07}
{"train_loss": 0.038208093494176865, "global_step": 264688, "epoch": 2974, "lr": 1.5355391650196394e-07}
{"train_loss": 0.03360579162836075, "global_step": 264689, "epoch": 2974, "lr": 1.535085196833319e-07}
{"train_loss": 0.029266400262713432, "global_step": 264690, "epoch": 2974, "lr": 1.534631295659561e-07}
{"train_loss": 0.03588072210550308, "global_step": 264691, "epoch": 2974, "lr": 1.534177461498365e-07}
{"train_loss": 0.10583385825157166, "global_step": 264692, "epoch": 2974, "lr": 1.5337236943498423e-07}
{"train_loss": 0.03733786195516586, "global_step": 264693, "epoch": 2974, "lr": 1.5332699942140482e-07}
{"train_loss": 0.07916537672281265, "global_step": 264694, "epoch": 2974, "lr": 1.5328163610910384e-07}
{"train_loss": 0.08779793232679367, "global_step": 264695, "epoch": 2974, "lr": 1.5323627949808128e-07}
{"train_loss": 0.07867058366537094, "global_step": 264696, "epoch": 2974, "lr": 1.531909295883538e-07}
{"train_loss": 0.08321139961481094, "global_step": 264697, "epoch": 2974, "lr": 1.5314558637991582e-07}
{"train_loss": 0.02312353067100048, "global_step": 264698, "epoch": 2974, "lr": 1.531002498727896e-07}
{"train_loss": 0.041619885712862015, "global_step": 264699, "epoch": 2974, "lr": 1.5305492006696398e-07}
{"train_loss": 0.05588289722800255, "global_step": 264700, "epoch": 2974, "lr": 1.530095969624501e-07}
{"train_loss": 0.11263665556907654, "global_step": 264701, "epoch": 2974, "lr": 1.529642805592646e-07}
{"train_loss": 0.025358840823173523, "global_step": 264702, "epoch": 2974, "lr": 1.5291897085740192e-07}
{"train_loss": 0.04702598601579666, "global_step": 264703, "epoch": 2974, "lr": 1.5287366785686764e-07}
{"train_loss": 0.02009231224656105, "global_step": 264704, "epoch": 2974, "lr": 1.5282837155767839e-07}
{"train_loss": 0.060233309864997864, "global_step": 264705, "epoch": 2974, "lr": 1.5278308195983415e-07}
{"train_loss": 0.051471736282110214, "global_step": 264706, "epoch": 2974, "lr": 1.5273779906333497e-07}
{"train_loss": 0.010532407090067863, "global_step": 264707, "epoch": 2974, "lr": 1.5269252286819746e-07}
{"train_loss": 0.06451985239982605, "global_step": 264708, "epoch": 2974, "lr": 1.5264725337441609e-07}
{"train_loss": 0.047266241163015366, "global_step": 264709, "epoch": 2974, "lr": 1.526019905820131e-07}
{"train_loss": 0.05293063819408417, "global_step": 264710, "epoch": 2974, "lr": 1.525567344909773e-07}
{"train_loss": 0.05288940295577049, "global_step": 264711, "epoch": 2974, "lr": 1.5251148510133094e-07}
{"train_loss": 0.0736757442355156, "global_step": 264712, "epoch": 2974, "lr": 1.5246624241306296e-07}
{"train_loss": 0.0412956103682518, "global_step": 264713, "epoch": 2974, "lr": 1.5242100642619551e-07}
{"train_loss": 0.01123069692403078, "global_step": 264714, "epoch": 2974, "lr": 1.5237577714072304e-07}
{"train_loss": 0.016459831967949867, "global_step": 264715, "epoch": 2974, "lr": 1.5233055455665667e-07}
{"train_loss": 0.06396622955799103, "global_step": 264716, "epoch": 2974, "lr": 1.5228533867400196e-07}
{"train_loss": 0.020182687789201736, "global_step": 264717, "epoch": 2974, "lr": 1.5224012949276444e-07}
{"train_loss": 0.054156310856342316, "global_step": 264718, "epoch": 2974, "lr": 1.5219492701295523e-07}
{"train_loss": 0.04925478994846344, "global_step": 264719, "epoch": 2974, "lr": 1.521497312345743e-07}
{"train_loss": 0.029472874477505684, "global_step": 264720, "epoch": 2974, "lr": 1.5210454215762725e-07}
{"train_loss": 0.0411667563021183, "global_step": 264721, "epoch": 2974, "lr": 1.520593597821196e-07}
{"train_loss": 0.031474921852350235, "global_step": 264722, "epoch": 2974, "lr": 1.5201418410806246e-07}
{"train_loss": 0.02819298766553402, "global_step": 264723, "epoch": 2974, "lr": 1.5196901513546135e-07}
{"train_loss": 0.030570512637495995, "global_step": 264724, "epoch": 2974, "lr": 1.519238528643163e-07}
{"train_loss": 0.052075374871492386, "global_step": 264725, "epoch": 2974, "lr": 1.51878697294644e-07}
{"train_loss": 0.03048706240952015, "global_step": 264726, "epoch": 2974, "lr": 1.5183354842643883e-07}
{"train_loss": 0.03649163618683815, "global_step": 264727, "epoch": 2974, "lr": 1.5178840625971192e-07}
{"train_loss": 0.03170095384120941, "global_step": 264728, "epoch": 2974, "lr": 1.5174327079447438e-07}
{"train_loss": 0.07461623847484589, "global_step": 264729, "epoch": 2974, "lr": 1.5169814203072064e-07}
{"train_loss": 0.03536205366253853, "global_step": 264730, "epoch": 2974, "lr": 1.5165301996846738e-07}
{"train_loss": 0.024003883823752403, "global_step": 264731, "epoch": 2974, "lr": 1.5160790460771457e-07}
{"train_loss": 0.03698810562491417, "global_step": 264732, "epoch": 2974, "lr": 1.5156279594847334e-07}
{"train_loss": 0.08681399375200272, "global_step": 264733, "epoch": 2974, "lr": 1.5151769399074923e-07}
{"train_loss": 0.046498462557792664, "global_step": 264734, "epoch": 2974, "lr": 1.5147259873454223e-07}
{"train_loss": 0.03195745125412941, "global_step": 264735, "epoch": 2974, "lr": 1.514275101798579e-07}
{"train_loss": 0.023960944265127182, "global_step": 264736, "epoch": 2974, "lr": 1.513824283267129e-07}
{"train_loss": 0.03869481012225151, "global_step": 264737, "epoch": 2974, "lr": 1.513373531751072e-07}
{"train_loss": 0.05159702152013779, "global_step": 264738, "epoch": 2974, "lr": 1.5129228472504088e-07}
{"train_loss": 0.014777169562876225, "global_step": 264739, "epoch": 2974, "lr": 1.5124722297653048e-07}
{"train_loss": 0.01695040985941887, "global_step": 264740, "epoch": 2974, "lr": 1.512021679295761e-07}
{"train_loss": 0.057856183499097824, "global_step": 264741, "epoch": 2974, "lr": 1.5115711958418877e-07}
{"train_loss": 0.04492472857236862, "global_step": 264742, "epoch": 2974, "lr": 1.5111207794036298e-07}
{"train_loss": 0.10424728691577911, "global_step": 264743, "epoch": 2974, "lr": 1.5106704299812092e-07}
{"train_loss": 0.08969514816999435, "global_step": 264744, "epoch": 2974, "lr": 1.5102201475745703e-07}
{"train_loss": 0.029584651812911034, "global_step": 264745, "epoch": 2974, "lr": 1.509769932183769e-07}
{"train_loss": 0.05604659020900726, "global_step": 264746, "epoch": 2974, "lr": 1.5093197838089157e-07}
{"train_loss": 0.03317706659436226, "global_step": 264747, "epoch": 2974, "lr": 1.5088697024501218e-07}
{"train_loss": 0.03728139027953148, "global_step": 264748, "epoch": 2974, "lr": 1.508419688107332e-07}
{"train_loss": 0.022667525336146355, "global_step": 264749, "epoch": 2974, "lr": 1.507969740780657e-07}
{"train_loss": 0.03945866599678993, "global_step": 264750, "epoch": 2974, "lr": 1.5075198604701523e-07}
{"train_loss": 0.041313398629426956, "global_step": 264751, "epoch": 2974, "lr": 1.507070047175929e-07}
{"train_loss": 0.05468963086605072, "global_step": 264752, "epoch": 2974, "lr": 1.5066203008979873e-07}
{"train_loss": 0.016534648835659027, "global_step": 264753, "epoch": 2974, "lr": 1.5061706216363823e-07}
{"train_loss": 0.05205124244093895, "global_step": 264754, "epoch": 2974, "lr": 1.5057210093912255e-07}
{"train_loss": 0.03942335024476051, "global_step": 264755, "epoch": 2974, "lr": 1.5052714641625164e-07}
{"train_loss": 0.02544540911912918, "global_step": 264756, "epoch": 2974, "lr": 1.5048219859503666e-07}
{"train_loss": 0.018247514963150024, "global_step": 264757, "epoch": 2974, "lr": 1.504372574754831e-07}
{"train_loss": 0.10004868358373642, "global_step": 264758, "epoch": 2974, "lr": 1.5039232305759654e-07}
{"train_loss": 0.0281673651188612, "global_step": 264759, "epoch": 2974, "lr": 1.5034739534137698e-07}
{"train_loss": 0.030073879286646843, "global_step": 264760, "epoch": 2974, "lr": 1.5030247432684107e-07}
{"train_loss": 0.03437694162130356, "global_step": 264761, "epoch": 2974, "lr": 1.5025756001398328e-07}
{"train_loss": 0.030225928872823715, "global_step": 264762, "epoch": 2974, "lr": 1.5021265240282023e-07}
{"train_loss": 0.05736717954277992, "global_step": 264763, "epoch": 2974, "lr": 1.5016775149335193e-07}
{"train_loss": 0.054002612829208374, "global_step": 264764, "epoch": 2974, "lr": 1.501228572855895e-07}
{"train_loss": 0.039533548057079315, "global_step": 264765, "epoch": 2974, "lr": 1.5007796977953291e-07}
{"train_loss": 0.03204734995961189, "global_step": 264766, "epoch": 2974, "lr": 1.5003308897518774e-07}
{"train_loss": 0.051490865647792816, "global_step": 264767, "epoch": 2974, "lr": 1.499882148725651e-07}
{"train_loss": 0.07934442907571793, "global_step": 264768, "epoch": 2974, "lr": 1.499433474716705e-07}
{"train_loss": 0.04739323630928993, "global_step": 264769, "epoch": 2974, "lr": 1.498984867725095e-07}
{"train_loss": 0.047399844974279404, "global_step": 264770, "epoch": 2974, "lr": 1.4985363277508213e-07}
{"train_loss": 0.03479017689824104, "global_step": 264771, "epoch": 2974, "lr": 1.4980878547940503e-07}
{"train_loss": 0.026255708187818527, "global_step": 264772, "epoch": 2974, "lr": 1.4976394488547818e-07}
{"train_loss": 0.05073991417884827, "global_step": 264773, "epoch": 2974, "lr": 1.497191109933016e-07}
{"train_loss": 0.04482101428249244, "global_step": 264774, "epoch": 2974, "lr": 1.4967428380289196e-07, "val_loss": 9.63765811920166}
{"train_loss": 0.11566554754972458, "global_step": 264775, "epoch": 2975, "lr": 1.496294633142492e-07}
{"train_loss": 0.02850448712706566, "global_step": 264776, "epoch": 2975, "lr": 1.495846495273845e-07}
{"train_loss": 0.028050193563103676, "global_step": 264777, "epoch": 2975, "lr": 1.495398424422978e-07}
{"train_loss": 0.04613871872425079, "global_step": 264778, "epoch": 2975, "lr": 1.4949504205900024e-07}
{"train_loss": 0.046808600425720215, "global_step": 264779, "epoch": 2975, "lr": 1.4945024837749178e-07}
{"train_loss": 0.039656590670347214, "global_step": 264780, "epoch": 2975, "lr": 1.4940546139778354e-07}
{"train_loss": 0.03400857001543045, "global_step": 264781, "epoch": 2975, "lr": 1.4936068111988665e-07}
{"train_loss": 0.025294281542301178, "global_step": 264782, "epoch": 2975, "lr": 1.4931590754378998e-07}
{"train_loss": 0.06442436575889587, "global_step": 264783, "epoch": 2975, "lr": 1.4927114066951574e-07}
{"train_loss": 0.017015986144542694, "global_step": 264784, "epoch": 2975, "lr": 1.4922638049706394e-07}
{"train_loss": 0.051973387598991394, "global_step": 264785, "epoch": 2975, "lr": 1.4918162702644568e-07}
{"train_loss": 0.14146026968955994, "global_step": 264786, "epoch": 2975, "lr": 1.491368802576554e-07}
{"train_loss": 0.036872267723083496, "global_step": 264787, "epoch": 2975, "lr": 1.490921401907097e-07}
{"train_loss": 0.0423579216003418, "global_step": 264788, "epoch": 2975, "lr": 1.490474068256087e-07}
{"train_loss": 0.018250616267323494, "global_step": 264789, "epoch": 2975, "lr": 1.490026801623634e-07}
{"train_loss": 0.05707351490855217, "global_step": 264790, "epoch": 2975, "lr": 1.4895796020097386e-07}
{"train_loss": 0.07669799029827118, "global_step": 264791, "epoch": 2975, "lr": 1.4891324694145115e-07}
{"train_loss": 0.03087548352777958, "global_step": 264792, "epoch": 2975, "lr": 1.4886854038380083e-07}
{"train_loss": 0.03289375081658363, "global_step": 264793, "epoch": 2975, "lr": 1.488238405280229e-07}
{"train_loss": 0.07189679890871048, "global_step": 264794, "epoch": 2975, "lr": 1.4877914737413402e-07}
{"train_loss": 0.03299708664417267, "global_step": 264795, "epoch": 2975, "lr": 1.4873446092212863e-07}
{"train_loss": 0.0584084689617157, "global_step": 264796, "epoch": 2975, "lr": 1.4868978117202336e-07}
{"train_loss": 0.02923266962170601, "global_step": 264797, "epoch": 2975, "lr": 1.4864510812381272e-07}
{"train_loss": 0.07361467182636261, "global_step": 264798, "epoch": 2975, "lr": 1.486004417775133e-07}
{"train_loss": 0.04117341339588165, "global_step": 264799, "epoch": 2975, "lr": 1.4855578213312515e-07}
{"train_loss": 0.029602883383631706, "global_step": 264800, "epoch": 2975, "lr": 1.4851112919065935e-07}
{"train_loss": 0.051717035472393036, "global_step": 264801, "epoch": 2975, "lr": 1.4846648295011588e-07}
{"train_loss": 0.0562918558716774, "global_step": 264802, "epoch": 2975, "lr": 1.4842184341150588e-07}
{"train_loss": 0.01764126867055893, "global_step": 264803, "epoch": 2975, "lr": 1.4837721057482933e-07}
{"train_loss": 0.056884586811065674, "global_step": 264804, "epoch": 2975, "lr": 1.483325844401029e-07}
{"train_loss": 0.04178152605891228, "global_step": 264805, "epoch": 2975, "lr": 1.4828796500731544e-07}
{"train_loss": 0.07295908778905869, "global_step": 264806, "epoch": 2975, "lr": 1.4824335227648923e-07}
{"train_loss": 0.03696940466761589, "global_step": 264807, "epoch": 2975, "lr": 1.481987462476242e-07}
{"train_loss": 0.06915292888879776, "global_step": 264808, "epoch": 2975, "lr": 1.4815414692072594e-07}
{"train_loss": 0.021491754800081253, "global_step": 264809, "epoch": 2975, "lr": 1.481095542958e-07}
{"train_loss": 0.043638043105602264, "global_step": 264810, "epoch": 2975, "lr": 1.480649683728519e-07}
{"train_loss": 0.031858496367931366, "global_step": 264811, "epoch": 2975, "lr": 1.4802038915189276e-07}
{"train_loss": 0.04382631927728653, "global_step": 264812, "epoch": 2975, "lr": 1.4797581663291704e-07}
{"train_loss": 0.029651785269379616, "global_step": 264813, "epoch": 2975, "lr": 1.4793125081594695e-07}
{"train_loss": 0.0809883251786232, "global_step": 264814, "epoch": 2975, "lr": 1.478866917009769e-07}
{"train_loss": 0.06903767585754395, "global_step": 264815, "epoch": 2975, "lr": 1.478421392880125e-07}
{"train_loss": 0.0595097579061985, "global_step": 264816, "epoch": 2975, "lr": 1.477975935770648e-07}
{"train_loss": 0.03614640608429909, "global_step": 264817, "epoch": 2975, "lr": 1.4775305456813938e-07}
{"train_loss": 0.0361795648932457, "global_step": 264818, "epoch": 2975, "lr": 1.4770852226124177e-07}
{"train_loss": 0.010644334368407726, "global_step": 264819, "epoch": 2975, "lr": 1.4766399665637753e-07}
{"train_loss": 0.05868542194366455, "global_step": 264820, "epoch": 2975, "lr": 1.4761947775355222e-07}
{"train_loss": 0.053025875240564346, "global_step": 264821, "epoch": 2975, "lr": 1.475749655527714e-07}
{"train_loss": 0.04340096563100815, "global_step": 264822, "epoch": 2975, "lr": 1.4753046005403504e-07}
{"train_loss": 0.015759391710162163, "global_step": 264823, "epoch": 2975, "lr": 1.4748596125736537e-07}
{"train_loss": 0.04668622463941574, "global_step": 264824, "epoch": 2975, "lr": 1.4744146916275125e-07}
{"train_loss": 0.03937850520014763, "global_step": 264825, "epoch": 2975, "lr": 1.4739698377020939e-07}
{"train_loss": 0.03407446667551994, "global_step": 264826, "epoch": 2975, "lr": 1.4735250507973975e-07}
{"train_loss": 0.03023291751742363, "global_step": 264827, "epoch": 2975, "lr": 1.47308033091359e-07}
{"train_loss": 0.05473357066512108, "global_step": 264828, "epoch": 2975, "lr": 1.47263567805056e-07}
{"train_loss": 0.035209134221076965, "global_step": 264829, "epoch": 2975, "lr": 1.4721910922084747e-07}
{"train_loss": 0.01851438730955124, "global_step": 264830, "epoch": 2975, "lr": 1.4717465733873892e-07}
{"train_loss": 0.03505663946270943, "global_step": 264831, "epoch": 2975, "lr": 1.471302121587359e-07}
{"train_loss": 0.021853860467672348, "global_step": 264832, "epoch": 2975, "lr": 1.4708577368084952e-07}
{"train_loss": 0.04257611185312271, "global_step": 264833, "epoch": 2975, "lr": 1.4704134190506867e-07}
{"train_loss": 0.11563582718372345, "global_step": 264834, "epoch": 2975, "lr": 1.4699691683141558e-07}
{"train_loss": 0.06916795670986176, "global_step": 264835, "epoch": 2975, "lr": 1.4695249845989022e-07}
{"train_loss": 0.05023012310266495, "global_step": 264836, "epoch": 2975, "lr": 1.4690808679050372e-07}
{"train_loss": 0.032134488224983215, "global_step": 264837, "epoch": 2975, "lr": 1.468636818232505e-07}
{"train_loss": 0.028388554230332375, "global_step": 264838, "epoch": 2975, "lr": 1.4681928355815276e-07}
{"train_loss": 0.04783795028924942, "global_step": 264839, "epoch": 2975, "lr": 1.4677489199519944e-07}
{"train_loss": 0.05515771359205246, "global_step": 264840, "epoch": 2975, "lr": 1.4673050713440717e-07}
{"train_loss": 0.017613735049962997, "global_step": 264841, "epoch": 2975, "lr": 1.466861289757815e-07}
{"train_loss": 0.030261101201176643, "global_step": 264842, "epoch": 2975, "lr": 1.466417575193224e-07}
{"train_loss": 0.03013038821518421, "global_step": 264843, "epoch": 2975, "lr": 1.4659739276504102e-07}
{"train_loss": 0.027368780225515366, "global_step": 264844, "epoch": 2975, "lr": 1.465530347129429e-07}
{"train_loss": 0.05801966041326523, "global_step": 264845, "epoch": 2975, "lr": 1.4650868336303358e-07}
{"train_loss": 0.020330658182501793, "global_step": 264846, "epoch": 2975, "lr": 1.4646433871531862e-07}
{"train_loss": 0.04962688311934471, "global_step": 264847, "epoch": 2975, "lr": 1.4642000076980356e-07}
{"train_loss": 0.04502066224813461, "global_step": 264848, "epoch": 2975, "lr": 1.4637566952649395e-07}
{"train_loss": 0.02971627749502659, "global_step": 264849, "epoch": 2975, "lr": 1.4633134498539535e-07}
{"train_loss": 0.032821059226989746, "global_step": 264850, "epoch": 2975, "lr": 1.4628702714651887e-07}
{"train_loss": 0.042723219841718674, "global_step": 264851, "epoch": 2975, "lr": 1.462427160098645e-07}
{"train_loss": 0.050038598477840424, "global_step": 264852, "epoch": 2975, "lr": 1.461984115754378e-07}
{"train_loss": 0.09820955991744995, "global_step": 264853, "epoch": 2975, "lr": 1.4615411384324985e-07}
{"train_loss": 0.024534154683351517, "global_step": 264854, "epoch": 2975, "lr": 1.4610982281330065e-07}
{"train_loss": 0.06844492256641388, "global_step": 264855, "epoch": 2975, "lr": 1.4606553848560134e-07}
{"train_loss": 0.0693647563457489, "global_step": 264856, "epoch": 2975, "lr": 1.4602126086015743e-07}
{"train_loss": 0.05939098820090294, "global_step": 264857, "epoch": 2975, "lr": 1.459769899369745e-07}
{"train_loss": 0.026310939341783524, "global_step": 264858, "epoch": 2975, "lr": 1.4593272571605253e-07}
{"train_loss": 0.042336657643318176, "global_step": 264859, "epoch": 2975, "lr": 1.458884681974082e-07}
{"train_loss": 0.07136530429124832, "global_step": 264860, "epoch": 2975, "lr": 1.4584421738103593e-07}
{"train_loss": 0.040322862565517426, "global_step": 264861, "epoch": 2975, "lr": 1.4579997326694683e-07}
{"train_loss": 0.03694402053952217, "global_step": 264862, "epoch": 2975, "lr": 1.45755735855152e-07}
{"train_loss": 0.04589236645927925, "global_step": 264863, "epoch": 2975, "lr": 1.4571150514565145e-07, "val_loss": 9.542264938354492, "train_action_mse_error": 24.0590763092041}
{"train_loss": 0.03176813945174217, "global_step": 264864, "epoch": 2976, "lr": 1.4566728113845073e-07}
{"train_loss": 0.018137669190764427, "global_step": 264865, "epoch": 2976, "lr": 1.456230638335554e-07}
{"train_loss": 0.039170391857624054, "global_step": 264866, "epoch": 2976, "lr": 1.4557885323097652e-07}
{"train_loss": 0.05894767493009567, "global_step": 264867, "epoch": 2976, "lr": 1.455346493307197e-07}
{"train_loss": 0.02093834988772869, "global_step": 264868, "epoch": 2976, "lr": 1.454904521327849e-07}
{"train_loss": 0.045546114444732666, "global_step": 264869, "epoch": 2976, "lr": 1.4544626163717766e-07}
{"train_loss": 0.024731943383812904, "global_step": 264870, "epoch": 2976, "lr": 1.4540207784391468e-07}
{"train_loss": 0.03281643986701965, "global_step": 264871, "epoch": 2976, "lr": 1.4535790075299038e-07}
{"train_loss": 0.04913734644651413, "global_step": 264872, "epoch": 2976, "lr": 1.4531373036441586e-07}
{"train_loss": 0.06856489926576614, "global_step": 264873, "epoch": 2976, "lr": 1.452695666781967e-07}
{"train_loss": 0.03809918090701103, "global_step": 264874, "epoch": 2976, "lr": 1.452254096943384e-07}
{"train_loss": 0.06466169655323029, "global_step": 264875, "epoch": 2976, "lr": 1.4518125941284655e-07}
{"train_loss": 0.024228427559137344, "global_step": 264876, "epoch": 2976, "lr": 1.451371158337267e-07}
{"train_loss": 0.06596031785011292, "global_step": 264877, "epoch": 2976, "lr": 1.4509297895698438e-07}
{"train_loss": 0.03992465138435364, "global_step": 264878, "epoch": 2976, "lr": 1.450488487826307e-07}
{"train_loss": 0.03409038484096527, "global_step": 264879, "epoch": 2976, "lr": 1.4500472531066567e-07}
{"train_loss": 0.037985481321811676, "global_step": 264880, "epoch": 2976, "lr": 1.4496060854109484e-07}
{"train_loss": 0.03537765517830849, "global_step": 264881, "epoch": 2976, "lr": 1.4491649847392929e-07}
{"train_loss": 0.09545902162790298, "global_step": 264882, "epoch": 2976, "lr": 1.4487239510916905e-07}
{"train_loss": 0.06899555772542953, "global_step": 264883, "epoch": 2976, "lr": 1.448282984468252e-07}
{"train_loss": 0.01902725175023079, "global_step": 264884, "epoch": 2976, "lr": 1.4478420848690333e-07}
{"train_loss": 0.0305969025939703, "global_step": 264885, "epoch": 2976, "lr": 1.4474012522940338e-07}
{"train_loss": 0.04847555607557297, "global_step": 264886, "epoch": 2976, "lr": 1.446960486743365e-07}
{"train_loss": 0.02814633399248123, "global_step": 264887, "epoch": 2976, "lr": 1.4465197882170821e-07}
{"train_loss": 0.03806903585791588, "global_step": 264888, "epoch": 2976, "lr": 1.446079156715241e-07}
{"train_loss": 0.023229191079735756, "global_step": 264889, "epoch": 2976, "lr": 1.4456385922379522e-07}
{"train_loss": 0.020942943170666695, "global_step": 264890, "epoch": 2976, "lr": 1.4451980947851606e-07}
{"train_loss": 0.024457260966300964, "global_step": 264891, "epoch": 2976, "lr": 1.444757664356977e-07}
{"train_loss": 0.06082744151353836, "global_step": 264892, "epoch": 2976, "lr": 1.4443173009535127e-07}
{"train_loss": 0.0107637420296669, "global_step": 264893, "epoch": 2976, "lr": 1.4438770045747673e-07}
{"train_loss": 0.04374987259507179, "global_step": 264894, "epoch": 2976, "lr": 1.4434367752207967e-07}
{"train_loss": 0.02095530368387699, "global_step": 264895, "epoch": 2976, "lr": 1.4429966128916562e-07}
{"train_loss": 0.06618773937225342, "global_step": 264896, "epoch": 2976, "lr": 1.4425565175874567e-07}
{"train_loss": 0.04036089405417442, "global_step": 264897, "epoch": 2976, "lr": 1.442116489308254e-07}
{"train_loss": 0.060320090502500534, "global_step": 264898, "epoch": 2976, "lr": 1.4416765280540478e-07}
{"train_loss": 0.050070807337760925, "global_step": 264899, "epoch": 2976, "lr": 1.4412366338249495e-07}
{"train_loss": 0.04757120460271835, "global_step": 264900, "epoch": 2976, "lr": 1.4407968066210142e-07}
{"train_loss": 0.029860449954867363, "global_step": 264901, "epoch": 2976, "lr": 1.4403570464422421e-07}
{"train_loss": 0.09371127933263779, "global_step": 264902, "epoch": 2976, "lr": 1.4399173532888e-07}
{"train_loss": 0.02640191651880741, "global_step": 264903, "epoch": 2976, "lr": 1.4394777271606318e-07}
{"train_loss": 0.020274080336093903, "global_step": 264904, "epoch": 2976, "lr": 1.4390381680578491e-07}
{"train_loss": 0.03275681659579277, "global_step": 264905, "epoch": 2976, "lr": 1.4385986759805625e-07}
{"train_loss": 0.038939401507377625, "global_step": 264906, "epoch": 2976, "lr": 1.4381592509287167e-07}
{"train_loss": 0.03457123413681984, "global_step": 264907, "epoch": 2976, "lr": 1.4377198929024782e-07}
{"train_loss": 0.023806819692254066, "global_step": 264908, "epoch": 2976, "lr": 1.437280601901847e-07}
{"train_loss": 0.04737299680709839, "global_step": 264909, "epoch": 2976, "lr": 1.4368413779268787e-07}
{"train_loss": 0.03519531339406967, "global_step": 264910, "epoch": 2976, "lr": 1.436402220977684e-07}
{"train_loss": 0.037643998861312866, "global_step": 264911, "epoch": 2976, "lr": 1.435963131054263e-07}
{"train_loss": 0.06053917109966278, "global_step": 264912, "epoch": 2976, "lr": 1.435524108156727e-07}
{"train_loss": 0.03227028250694275, "global_step": 264913, "epoch": 2976, "lr": 1.435085152285076e-07}
{"train_loss": 0.05977905914187431, "global_step": 264914, "epoch": 2976, "lr": 1.4346462634394208e-07}
{"train_loss": 0.04208601266145706, "global_step": 264915, "epoch": 2976, "lr": 1.434207441619817e-07}
{"train_loss": 0.059689756482839584, "global_step": 264916, "epoch": 2976, "lr": 1.4337686868262644e-07}
{"train_loss": 0.027605265378952026, "global_step": 264917, "epoch": 2976, "lr": 1.4333299990588744e-07}
{"train_loss": 0.0634349063038826, "global_step": 264918, "epoch": 2976, "lr": 1.4328913783177023e-07}
{"train_loss": 0.08563153445720673, "global_step": 264919, "epoch": 2976, "lr": 1.4324528246028035e-07}
{"train_loss": 0.03050706721842289, "global_step": 264920, "epoch": 2976, "lr": 1.4320143379142338e-07}
{"train_loss": 0.03145468235015869, "global_step": 264921, "epoch": 2976, "lr": 1.4315759182521038e-07}
{"train_loss": 0.0716228187084198, "global_step": 264922, "epoch": 2976, "lr": 1.4311375656163585e-07}
{"train_loss": 0.03178079426288605, "global_step": 264923, "epoch": 2976, "lr": 1.4306992800071638e-07}
{"train_loss": 0.03466630354523659, "global_step": 264924, "epoch": 2976, "lr": 1.430261061424465e-07}
{"train_loss": 0.023413270711898804, "global_step": 264925, "epoch": 2976, "lr": 1.429822909868428e-07}
{"train_loss": 0.033600952476263046, "global_step": 264926, "epoch": 2976, "lr": 1.4293848253391084e-07}
{"train_loss": 0.04842439293861389, "global_step": 264927, "epoch": 2976, "lr": 1.4289468078365065e-07}
{"train_loss": 0.02513754926621914, "global_step": 264928, "epoch": 2976, "lr": 1.4285088573606776e-07}
{"train_loss": 0.03610233590006828, "global_step": 264929, "epoch": 2976, "lr": 1.4280709739117326e-07}
{"train_loss": 0.05484325811266899, "global_step": 264930, "epoch": 2976, "lr": 1.4276331574897273e-07}
{"train_loss": 0.03988569602370262, "global_step": 264931, "epoch": 2976, "lr": 1.4271954080946616e-07}
{"train_loss": 0.0575060211122036, "global_step": 264932, "epoch": 2976, "lr": 1.4267577257266462e-07}
{"train_loss": 0.023425454273819923, "global_step": 264933, "epoch": 2976, "lr": 1.4263201103857927e-07}
{"train_loss": 0.049805898219347, "global_step": 264934, "epoch": 2976, "lr": 1.4258825620719896e-07}
{"train_loss": 0.03513417765498161, "global_step": 264935, "epoch": 2976, "lr": 1.4254450807854592e-07}
{"train_loss": 0.0509318932890892, "global_step": 264936, "epoch": 2976, "lr": 1.4250076665262013e-07}
{"train_loss": 0.01881999336183071, "global_step": 264937, "epoch": 2976, "lr": 1.4245703192942161e-07}
{"train_loss": 0.040332432836294174, "global_step": 264938, "epoch": 2976, "lr": 1.4241330390896702e-07}
{"train_loss": 0.03911601006984711, "global_step": 264939, "epoch": 2976, "lr": 1.4236958259125633e-07}
{"train_loss": 0.034445155411958694, "global_step": 264940, "epoch": 2976, "lr": 1.4232586797630065e-07}
{"train_loss": 0.07416322827339172, "global_step": 264941, "epoch": 2976, "lr": 1.4228216006409445e-07}
{"train_loss": 0.05808624252676964, "global_step": 264942, "epoch": 2976, "lr": 1.4223845885465436e-07}
{"train_loss": 0.01723468489944935, "global_step": 264943, "epoch": 2976, "lr": 1.4219476434798596e-07}
{"train_loss": 0.03622593730688095, "global_step": 264944, "epoch": 2976, "lr": 1.4215107654408367e-07}
{"train_loss": 0.042755335569381714, "global_step": 264945, "epoch": 2976, "lr": 1.421073954429697e-07}
{"train_loss": 0.035129938274621964, "global_step": 264946, "epoch": 2976, "lr": 1.4206372104463296e-07}
{"train_loss": 0.04373548924922943, "global_step": 264947, "epoch": 2976, "lr": 1.4202005334909563e-07}
{"train_loss": 0.04276612773537636, "global_step": 264948, "epoch": 2976, "lr": 1.4197639235635218e-07}
{"train_loss": 0.07120472937822342, "global_step": 264949, "epoch": 2976, "lr": 1.4193273806641372e-07}
{"train_loss": 0.06418299674987793, "global_step": 264950, "epoch": 2976, "lr": 1.4188909047928577e-07}
{"train_loss": 0.05308965966105461, "global_step": 264951, "epoch": 2976, "lr": 1.4184544959497391e-07}
{"train_loss": 0.042506104586332034, "global_step": 264952, "epoch": 2976, "lr": 1.4180181541347813e-07, "val_loss": 9.53985595703125}
{"train_loss": 0.034248266369104385, "global_step": 264953, "epoch": 2977, "lr": 1.417581879348151e-07}
{"train_loss": 0.04871879518032074, "global_step": 264954, "epoch": 2977, "lr": 1.4171456715897923e-07}
{"train_loss": 0.022053886204957962, "global_step": 264955, "epoch": 2977, "lr": 1.4167095308598722e-07}
{"train_loss": 0.028178604319691658, "global_step": 264956, "epoch": 2977, "lr": 1.4162734571583348e-07}
{"train_loss": 0.03296743705868721, "global_step": 264957, "epoch": 2977, "lr": 1.415837450485402e-07}
{"train_loss": 0.058210525661706924, "global_step": 264958, "epoch": 2977, "lr": 1.4154015108409634e-07}
{"train_loss": 0.06955190747976303, "global_step": 264959, "epoch": 2977, "lr": 1.4149656382251297e-07}
{"train_loss": 0.083457812666893, "global_step": 264960, "epoch": 2977, "lr": 1.414529832638012e-07}
{"train_loss": 0.06668721139431, "global_step": 264961, "epoch": 2977, "lr": 1.4140940940796653e-07}
{"train_loss": 0.02688463032245636, "global_step": 264962, "epoch": 2977, "lr": 1.4136584225500348e-07}
{"train_loss": 0.050524529069662094, "global_step": 264963, "epoch": 2977, "lr": 1.4132228180493423e-07}
{"train_loss": 0.07061433792114258, "global_step": 264964, "epoch": 2977, "lr": 1.4127872805775322e-07}
{"train_loss": 0.05163475498557091, "global_step": 264965, "epoch": 2977, "lr": 1.4123518101347156e-07}
{"train_loss": 0.04963502287864685, "global_step": 264966, "epoch": 2977, "lr": 1.4119164067208923e-07}
{"train_loss": 0.05752350017428398, "global_step": 264967, "epoch": 2977, "lr": 1.4114810703361737e-07}
{"train_loss": 0.039156440645456314, "global_step": 264968, "epoch": 2977, "lr": 1.4110458009806148e-07}
{"train_loss": 0.09914050996303558, "global_step": 264969, "epoch": 2977, "lr": 1.4106105986542716e-07}
{"train_loss": 0.03696384280920029, "global_step": 264970, "epoch": 2977, "lr": 1.410175463357144e-07}
{"train_loss": 0.04232095927000046, "global_step": 264971, "epoch": 2977, "lr": 1.4097403950893984e-07}
{"train_loss": 0.0635618045926094, "global_step": 264972, "epoch": 2977, "lr": 1.4093053938509792e-07}
{"train_loss": 0.02344316989183426, "global_step": 264973, "epoch": 2977, "lr": 1.4088704596420533e-07}
{"train_loss": 0.04334712028503418, "global_step": 264974, "epoch": 2977, "lr": 1.4084355924626202e-07}
{"train_loss": 0.031190114095807076, "global_step": 264975, "epoch": 2977, "lr": 1.408000792312736e-07}
{"train_loss": 0.033890802413225174, "global_step": 264976, "epoch": 2977, "lr": 1.4075660591924557e-07}
{"train_loss": 0.05931525304913521, "global_step": 264977, "epoch": 2977, "lr": 1.4071313931018902e-07}
{"train_loss": 0.03303306922316551, "global_step": 264978, "epoch": 2977, "lr": 1.4066967940410402e-07}
{"train_loss": 0.01771562732756138, "global_step": 264979, "epoch": 2977, "lr": 1.4062622620099608e-07}
{"train_loss": 0.02255914732813835, "global_step": 264980, "epoch": 2977, "lr": 1.405827797008763e-07}
{"train_loss": 0.03221195563673973, "global_step": 264981, "epoch": 2977, "lr": 1.4053933990374469e-07}
{"train_loss": 0.06974063813686371, "global_step": 264982, "epoch": 2977, "lr": 1.4049590680961232e-07}
{"train_loss": 0.023341331630945206, "global_step": 264983, "epoch": 2977, "lr": 1.404524804184848e-07}
{"train_loss": 0.06336050480604172, "global_step": 264984, "epoch": 2977, "lr": 1.4040906073036208e-07}
{"train_loss": 0.046549711376428604, "global_step": 264985, "epoch": 2977, "lr": 1.4036564774525529e-07}
{"train_loss": 0.05647194758057594, "global_step": 264986, "epoch": 2977, "lr": 1.4032224146316442e-07}
{"train_loss": 0.03105352073907852, "global_step": 264987, "epoch": 2977, "lr": 1.402788418841061e-07}
{"train_loss": 0.025350766256451607, "global_step": 264988, "epoch": 2977, "lr": 1.4023544900807484e-07}
{"train_loss": 0.04066632315516472, "global_step": 264989, "epoch": 2977, "lr": 1.4019206283508169e-07}
{"train_loss": 0.03757980093359947, "global_step": 264990, "epoch": 2977, "lr": 1.401486833651322e-07}
{"train_loss": 0.04091892018914223, "global_step": 264991, "epoch": 2977, "lr": 1.4010531059823196e-07}
{"train_loss": 0.025589419528841972, "global_step": 264992, "epoch": 2977, "lr": 1.4006194453438648e-07}
{"train_loss": 0.07857588678598404, "global_step": 264993, "epoch": 2977, "lr": 1.4001858517360688e-07}
{"train_loss": 0.03379369154572487, "global_step": 264994, "epoch": 2977, "lr": 1.3997523251588762e-07}
{"train_loss": 0.035198383033275604, "global_step": 264995, "epoch": 2977, "lr": 1.3993188656124535e-07}
{"train_loss": 0.038165368139743805, "global_step": 264996, "epoch": 2977, "lr": 1.3988854730968005e-07}
{"train_loss": 0.03999810293316841, "global_step": 264997, "epoch": 2977, "lr": 1.3984521476119728e-07}
{"train_loss": 0.0541490875184536, "global_step": 264998, "epoch": 2977, "lr": 1.398018889158026e-07}
{"train_loss": 0.050181467086076736, "global_step": 264999, "epoch": 2977, "lr": 1.3975856977351266e-07}
{"train_loss": 0.12190182507038116, "global_step": 265000, "epoch": 2977, "lr": 1.3971525733431634e-07}
{"train_loss": 0.059032142162323, "global_step": 265001, "epoch": 2977, "lr": 1.3967195159823032e-07}
{"train_loss": 0.021007999777793884, "global_step": 265002, "epoch": 2977, "lr": 1.3962865256526015e-07}
{"train_loss": 0.02468792162835598, "global_step": 265003, "epoch": 2977, "lr": 1.395853602354058e-07}
{"train_loss": 0.05706403777003288, "global_step": 265004, "epoch": 2977, "lr": 1.395420746086784e-07}
{"train_loss": 0.028076816350221634, "global_step": 265005, "epoch": 2977, "lr": 1.3949879568507797e-07}
{"train_loss": 0.06544962525367737, "global_step": 265006, "epoch": 2977, "lr": 1.3945552346461554e-07}
{"train_loss": 0.0718579888343811, "global_step": 265007, "epoch": 2977, "lr": 1.3941225794730228e-07}
{"train_loss": 0.024822140112519264, "global_step": 265008, "epoch": 2977, "lr": 1.3936899913312706e-07}
{"train_loss": 0.029448291286826134, "global_step": 265009, "epoch": 2977, "lr": 1.393257470221121e-07}
{"train_loss": 0.05123082175850868, "global_step": 265010, "epoch": 2977, "lr": 1.3928250161425737e-07}
{"train_loss": 0.025466399267315865, "global_step": 265011, "epoch": 2977, "lr": 1.3923926290956845e-07}
{"train_loss": 0.010091079398989677, "global_step": 265012, "epoch": 2977, "lr": 1.391960309080509e-07}
{"train_loss": 0.02347966842353344, "global_step": 265013, "epoch": 2977, "lr": 1.3915280560970467e-07}
{"train_loss": 0.030506683513522148, "global_step": 265014, "epoch": 2977, "lr": 1.3910958701455202e-07}
{"train_loss": 0.018118195235729218, "global_step": 265015, "epoch": 2977, "lr": 1.390663751225818e-07}
{"train_loss": 0.013721298426389694, "global_step": 265016, "epoch": 2977, "lr": 1.3902316993380516e-07}
{"train_loss": 0.02628360316157341, "global_step": 265017, "epoch": 2977, "lr": 1.3897997144823315e-07}
{"train_loss": 0.021474869921803474, "global_step": 265018, "epoch": 2977, "lr": 1.3893677966586582e-07}
{"train_loss": 0.05581239238381386, "global_step": 265019, "epoch": 2977, "lr": 1.388935945867087e-07}
{"train_loss": 0.03256380558013916, "global_step": 265020, "epoch": 2977, "lr": 1.3885041621077288e-07}
{"train_loss": 0.033195845782756805, "global_step": 265021, "epoch": 2977, "lr": 1.3880724453805838e-07}
{"train_loss": 0.030483629554510117, "global_step": 265022, "epoch": 2977, "lr": 1.387640795685763e-07}
{"train_loss": 0.06058821454644203, "global_step": 265023, "epoch": 2977, "lr": 1.3872092130232662e-07}
{"train_loss": 0.02193889021873474, "global_step": 265024, "epoch": 2977, "lr": 1.3867776973932045e-07}
{"train_loss": 0.043141014873981476, "global_step": 265025, "epoch": 2977, "lr": 1.3863462487956336e-07}
{"train_loss": 0.03422238305211067, "global_step": 265026, "epoch": 2977, "lr": 1.3859148672305532e-07}
{"train_loss": 0.007617939729243517, "global_step": 265027, "epoch": 2977, "lr": 1.3854835526980746e-07}
{"train_loss": 0.013872629962861538, "global_step": 265028, "epoch": 2977, "lr": 1.3850523051982533e-07}
{"train_loss": 0.0247963834553957, "global_step": 265029, "epoch": 2977, "lr": 1.3846211247310892e-07}
{"train_loss": 0.023118413984775543, "global_step": 265030, "epoch": 2977, "lr": 1.3841900112966933e-07}
{"train_loss": 0.08319852501153946, "global_step": 265031, "epoch": 2977, "lr": 1.3837589648951767e-07}
{"train_loss": 0.03395622968673706, "global_step": 265032, "epoch": 2977, "lr": 1.3833279855264835e-07}
{"train_loss": 0.03690839186310768, "global_step": 265033, "epoch": 2977, "lr": 1.3828970731907255e-07}
{"train_loss": 0.034231483936309814, "global_step": 265034, "epoch": 2977, "lr": 1.3824662278879574e-07}
{"train_loss": 0.018870247527956963, "global_step": 265035, "epoch": 2977, "lr": 1.3820354496182907e-07}
{"train_loss": 0.0368649959564209, "global_step": 265036, "epoch": 2977, "lr": 1.3816047383816698e-07}
{"train_loss": 0.01961607113480568, "global_step": 265037, "epoch": 2977, "lr": 1.3811740941782613e-07}
{"train_loss": 0.043240323662757874, "global_step": 265038, "epoch": 2977, "lr": 1.3807435170080652e-07}
{"train_loss": 0.020466266199946404, "global_step": 265039, "epoch": 2977, "lr": 1.3803130068711369e-07}
{"train_loss": 0.0419229120016098, "global_step": 265040, "epoch": 2977, "lr": 1.3798825637675317e-07}
{"train_loss": 0.04136111339805334, "global_step": 265041, "epoch": 2977, "lr": 1.379452187697361e-07, "val_loss": 9.655328750610352}
{"train_loss": 0.028253845870494843, "global_step": 265042, "epoch": 2978, "lr": 1.3790218786606247e-07}
{"train_loss": 0.03825657442212105, "global_step": 265043, "epoch": 2978, "lr": 1.378591636657378e-07}
{"train_loss": 0.11205562204122543, "global_step": 265044, "epoch": 2978, "lr": 1.3781614616877324e-07}
{"train_loss": 0.05909564718604088, "global_step": 265045, "epoch": 2978, "lr": 1.3777313537517433e-07}
{"train_loss": 0.03675675392150879, "global_step": 265046, "epoch": 2978, "lr": 1.3773013128494106e-07}
{"train_loss": 0.03965550661087036, "global_step": 265047, "epoch": 2978, "lr": 1.3768713389807898e-07}
{"train_loss": 0.03592635691165924, "global_step": 265048, "epoch": 2978, "lr": 1.3764414321460472e-07}
{"train_loss": 0.06607413291931152, "global_step": 265049, "epoch": 2978, "lr": 1.3760115923450722e-07}
{"train_loss": 0.04613092914223671, "global_step": 265050, "epoch": 2978, "lr": 1.3755818195780867e-07}
{"train_loss": 0.0422801673412323, "global_step": 265051, "epoch": 2978, "lr": 1.375152113845035e-07}
{"train_loss": 0.08888614922761917, "global_step": 265052, "epoch": 2978, "lr": 1.374722475146084e-07}
{"train_loss": 0.052753694355487823, "global_step": 265053, "epoch": 2978, "lr": 1.3742929034811225e-07}
{"train_loss": 0.05205491557717323, "global_step": 265054, "epoch": 2978, "lr": 1.3738633988503725e-07}
{"train_loss": 0.04175471514463425, "global_step": 265055, "epoch": 2978, "lr": 1.3734339612538337e-07}
{"train_loss": 0.039598945528268814, "global_step": 265056, "epoch": 2978, "lr": 1.3730045906915623e-07}
{"train_loss": 0.0610320083796978, "global_step": 265057, "epoch": 2978, "lr": 1.3725752871635578e-07}
{"train_loss": 0.03075646050274372, "global_step": 265058, "epoch": 2978, "lr": 1.372146050669987e-07}
{"train_loss": 0.05753433704376221, "global_step": 265059, "epoch": 2978, "lr": 1.3717168812108493e-07}
{"train_loss": 0.024825319647789, "global_step": 265060, "epoch": 2978, "lr": 1.371287778786201e-07}
{"train_loss": 0.045194998383522034, "global_step": 265061, "epoch": 2978, "lr": 1.3708587433960974e-07}
{"train_loss": 0.02037167362868786, "global_step": 265062, "epoch": 2978, "lr": 1.3704297750405937e-07}
{"train_loss": 0.08071862161159515, "global_step": 265063, "epoch": 2978, "lr": 1.3700008737198012e-07}
{"train_loss": 0.04863761365413666, "global_step": 265064, "epoch": 2978, "lr": 1.3695720394337198e-07}
{"train_loss": 0.03585125878453255, "global_step": 265065, "epoch": 2978, "lr": 1.369143272182405e-07}
{"train_loss": 0.06124464049935341, "global_step": 265066, "epoch": 2978, "lr": 1.3687145719659124e-07}
{"train_loss": 0.03522590920329094, "global_step": 265067, "epoch": 2978, "lr": 1.3682859387843528e-07}
{"train_loss": 0.03368654102087021, "global_step": 265068, "epoch": 2978, "lr": 1.3678573726377264e-07}
{"train_loss": 0.093513123691082, "global_step": 265069, "epoch": 2978, "lr": 1.3674288735261443e-07}
{"train_loss": 0.03973275423049927, "global_step": 265070, "epoch": 2978, "lr": 1.3670004414496064e-07}
{"train_loss": 0.044175032526254654, "global_step": 265071, "epoch": 2978, "lr": 1.3665720764082234e-07}
{"train_loss": 0.033676233142614365, "global_step": 265072, "epoch": 2978, "lr": 1.366143778401996e-07}
{"train_loss": 0.02450476959347725, "global_step": 265073, "epoch": 2978, "lr": 1.3657155474310347e-07}
{"train_loss": 0.03854425996541977, "global_step": 265074, "epoch": 2978, "lr": 1.3652873834953394e-07}
{"train_loss": 0.044742830097675323, "global_step": 265075, "epoch": 2978, "lr": 1.364859286595077e-07}
{"train_loss": 0.057936254888772964, "global_step": 265076, "epoch": 2978, "lr": 1.3644312567301366e-07}
{"train_loss": 0.030945563688874245, "global_step": 265077, "epoch": 2978, "lr": 1.3640032939007398e-07}
{"train_loss": 0.039061229676008224, "global_step": 265078, "epoch": 2978, "lr": 1.3635753981068312e-07}
{"train_loss": 0.03233940154314041, "global_step": 265079, "epoch": 2978, "lr": 1.3631475693485775e-07}
{"train_loss": 0.036760181188583374, "global_step": 265080, "epoch": 2978, "lr": 1.3627198076259228e-07}
{"train_loss": 0.048840250819921494, "global_step": 265081, "epoch": 2978, "lr": 1.3622921129389787e-07}
{"train_loss": 0.060629576444625854, "global_step": 265082, "epoch": 2978, "lr": 1.3618644852878005e-07}
{"train_loss": 0.043674569576978683, "global_step": 265083, "epoch": 2978, "lr": 1.3614369246724434e-07}
{"train_loss": 0.03413398563861847, "global_step": 265084, "epoch": 2978, "lr": 1.3610094310929633e-07}
{"train_loss": 0.04201062396168709, "global_step": 265085, "epoch": 2978, "lr": 1.3605820045494155e-07}
{"train_loss": 0.040016304701566696, "global_step": 265086, "epoch": 2978, "lr": 1.3601546450418557e-07}
{"train_loss": 0.0744517520070076, "global_step": 265087, "epoch": 2978, "lr": 1.359727352570339e-07}
{"train_loss": 0.031650200486183167, "global_step": 265088, "epoch": 2978, "lr": 1.359300127134977e-07}
{"train_loss": 0.03370126336812973, "global_step": 265089, "epoch": 2978, "lr": 1.3588729687357138e-07}
{"train_loss": 0.052860744297504425, "global_step": 265090, "epoch": 2978, "lr": 1.358445877372716e-07}
{"train_loss": 0.03448091074824333, "global_step": 265091, "epoch": 2978, "lr": 1.3580188530459836e-07}
{"train_loss": 0.08375944942235947, "global_step": 265092, "epoch": 2978, "lr": 1.357591895755572e-07}
{"train_loss": 0.044014349579811096, "global_step": 265093, "epoch": 2978, "lr": 1.3571650055015926e-07}
{"train_loss": 0.053947754204273224, "global_step": 265094, "epoch": 2978, "lr": 1.3567381822840453e-07}
{"train_loss": 0.02765166386961937, "global_step": 265095, "epoch": 2978, "lr": 1.3563114261029853e-07}
{"train_loss": 0.07443351298570633, "global_step": 265096, "epoch": 2978, "lr": 1.3558847369585237e-07}
{"train_loss": 0.029539361596107483, "global_step": 265097, "epoch": 2978, "lr": 1.3554581148506606e-07}
{"train_loss": 0.029316462576389313, "global_step": 265098, "epoch": 2978, "lr": 1.355031559779507e-07}
{"train_loss": 0.054570045322179794, "global_step": 265099, "epoch": 2978, "lr": 1.3546050717450632e-07}
{"train_loss": 0.02786918170750141, "global_step": 265100, "epoch": 2978, "lr": 1.3541786507474398e-07}
{"train_loss": 0.04319353401660919, "global_step": 265101, "epoch": 2978, "lr": 1.3537522967866367e-07}
{"train_loss": 0.07274550199508667, "global_step": 265102, "epoch": 2978, "lr": 1.3533260098627652e-07}
{"train_loss": 0.02647612988948822, "global_step": 265103, "epoch": 2978, "lr": 1.3528997899758812e-07}
{"train_loss": 0.019860731437802315, "global_step": 265104, "epoch": 2978, "lr": 1.3524736371259838e-07}
{"train_loss": 0.08422327786684036, "global_step": 265105, "epoch": 2978, "lr": 1.3520475513131847e-07}
{"train_loss": 0.018107939511537552, "global_step": 265106, "epoch": 2978, "lr": 1.3516215325375392e-07}
{"train_loss": 0.03960900753736496, "global_step": 265107, "epoch": 2978, "lr": 1.3511955807991027e-07}
{"train_loss": 0.03913484141230583, "global_step": 265108, "epoch": 2978, "lr": 1.3507696960978755e-07}
{"train_loss": 0.02492101863026619, "global_step": 265109, "epoch": 2978, "lr": 1.3503438784339684e-07}
{"train_loss": 0.04405071586370468, "global_step": 265110, "epoch": 2978, "lr": 1.3499181278074367e-07}
{"train_loss": 0.01707044057548046, "global_step": 265111, "epoch": 2978, "lr": 1.3494924442183364e-07}
{"train_loss": 0.04801817238330841, "global_step": 265112, "epoch": 2978, "lr": 1.349066827666723e-07}
{"train_loss": 0.04551137238740921, "global_step": 265113, "epoch": 2978, "lr": 1.3486412781526513e-07}
{"train_loss": 0.016024500131607056, "global_step": 265114, "epoch": 2978, "lr": 1.3482157956761776e-07}
{"train_loss": 0.025640694424510002, "global_step": 265115, "epoch": 2978, "lr": 1.3477903802373571e-07}
{"train_loss": 0.0334569551050663, "global_step": 265116, "epoch": 2978, "lr": 1.347365031836245e-07}
{"train_loss": 0.06710859388113022, "global_step": 265117, "epoch": 2978, "lr": 1.3469397504728976e-07}
{"train_loss": 0.0420621819794178, "global_step": 265118, "epoch": 2978, "lr": 1.3465145361473698e-07}
{"train_loss": 0.04333309829235077, "global_step": 265119, "epoch": 2978, "lr": 1.346089388859717e-07}
{"train_loss": 0.06257136166095734, "global_step": 265120, "epoch": 2978, "lr": 1.3456643086100506e-07}
{"train_loss": 0.04172346740961075, "global_step": 265121, "epoch": 2978, "lr": 1.3452392953983152e-07}
{"train_loss": 0.058579958975315094, "global_step": 265122, "epoch": 2978, "lr": 1.3448143492246767e-07}
{"train_loss": 0.05166219174861908, "global_step": 265123, "epoch": 2978, "lr": 1.3443894700891356e-07}
{"train_loss": 0.05239962413907051, "global_step": 265124, "epoch": 2978, "lr": 1.343964657991803e-07}
{"train_loss": 0.05446344614028931, "global_step": 265125, "epoch": 2978, "lr": 1.3435399129326785e-07}
{"train_loss": 0.08453325927257538, "global_step": 265126, "epoch": 2978, "lr": 1.3431152349118182e-07}
{"train_loss": 0.03684220090508461, "global_step": 265127, "epoch": 2978, "lr": 1.342690623929277e-07}
{"train_loss": 0.03545289486646652, "global_step": 265128, "epoch": 2978, "lr": 1.3422660799852215e-07}
{"train_loss": 0.07588887214660645, "global_step": 265129, "epoch": 2978, "lr": 1.341841603079541e-07}
{"train_loss": 0.04615402604673016, "global_step": 265130, "epoch": 2978, "lr": 1.341417193212402e-07, "val_loss": 9.669710159301758}
{"train_loss": 0.03366805985569954, "global_step": 265131, "epoch": 2979, "lr": 1.3409928503838043e-07}
{"train_loss": 0.05411316081881523, "global_step": 265132, "epoch": 2979, "lr": 1.340568574593859e-07}
{"train_loss": 0.012098017148673534, "global_step": 265133, "epoch": 2979, "lr": 1.340144365842566e-07}
{"train_loss": 0.02475583367049694, "global_step": 265134, "epoch": 2979, "lr": 1.3397202241300366e-07}
{"train_loss": 0.02541288733482361, "global_step": 265135, "epoch": 2979, "lr": 1.339296149456326e-07}
{"train_loss": 0.043342724442481995, "global_step": 265136, "epoch": 2979, "lr": 1.3388721418214345e-07}
{"train_loss": 0.04959049075841904, "global_step": 265137, "epoch": 2979, "lr": 1.338448201225473e-07}
{"train_loss": 0.014438718557357788, "global_step": 265138, "epoch": 2979, "lr": 1.3380243276684966e-07}
{"train_loss": 0.03286968916654587, "global_step": 265139, "epoch": 2979, "lr": 1.337600521150506e-07}
{"train_loss": 0.034105416387319565, "global_step": 265140, "epoch": 2979, "lr": 1.337176781671612e-07}
{"train_loss": 0.048033036291599274, "global_step": 265141, "epoch": 2979, "lr": 1.3367531092318698e-07}
{"train_loss": 0.02052832581102848, "global_step": 265142, "epoch": 2979, "lr": 1.3363295038312795e-07}
{"train_loss": 0.027386192232370377, "global_step": 265143, "epoch": 2979, "lr": 1.3359059654700079e-07}
{"train_loss": 0.030169380828738213, "global_step": 265144, "epoch": 2979, "lr": 1.3354824941479992e-07}
{"train_loss": 0.036328453570604324, "global_step": 265145, "epoch": 2979, "lr": 1.3350590898653648e-07}
{"train_loss": 0.0889144316315651, "global_step": 265146, "epoch": 2979, "lr": 1.3346357526221597e-07}
{"train_loss": 0.033151861280202866, "global_step": 265147, "epoch": 2979, "lr": 1.33421248241844e-07}
{"train_loss": 0.027455180883407593, "global_step": 265148, "epoch": 2979, "lr": 1.333789279254205e-07}
{"train_loss": 0.06261462718248367, "global_step": 265149, "epoch": 2979, "lr": 1.3333661431296218e-07}
{"train_loss": 0.05221739783883095, "global_step": 265150, "epoch": 2979, "lr": 1.33294307404469e-07}
{"train_loss": 0.04899049550294876, "global_step": 265151, "epoch": 2979, "lr": 1.3325200719994658e-07}
{"train_loss": 0.07001665979623795, "global_step": 265152, "epoch": 2979, "lr": 1.3320971369939484e-07}
{"train_loss": 0.012396409176290035, "global_step": 265153, "epoch": 2979, "lr": 1.3316742690283045e-07}
{"train_loss": 0.0281669944524765, "global_step": 265154, "epoch": 2979, "lr": 1.3312514681025345e-07}
{"train_loss": 0.02158544398844242, "global_step": 265155, "epoch": 2979, "lr": 1.3308287342166935e-07}
{"train_loss": 0.03906950354576111, "global_step": 265156, "epoch": 2979, "lr": 1.3304060673708374e-07}
{"train_loss": 0.05675642192363739, "global_step": 265157, "epoch": 2979, "lr": 1.3299834675650767e-07}
{"train_loss": 0.05731009319424629, "global_step": 265158, "epoch": 2979, "lr": 1.3295609347993565e-07}
{"train_loss": 0.06719693541526794, "global_step": 265159, "epoch": 2979, "lr": 1.3291384690738428e-07}
{"train_loss": 0.04641185328364372, "global_step": 265160, "epoch": 2979, "lr": 1.328716070388536e-07}
{"train_loss": 0.06866712123155594, "global_step": 265161, "epoch": 2979, "lr": 1.3282937387434913e-07}
{"train_loss": 0.04573708400130272, "global_step": 265162, "epoch": 2979, "lr": 1.32787147413882e-07}
{"train_loss": 0.016618438065052032, "global_step": 265163, "epoch": 2979, "lr": 1.327449276574466e-07}
{"train_loss": 0.03362071141600609, "global_step": 265164, "epoch": 2979, "lr": 1.3270271460506522e-07}
{"train_loss": 0.027569163590669632, "global_step": 265165, "epoch": 2979, "lr": 1.326605082567267e-07}
{"train_loss": 0.026207054033875465, "global_step": 265166, "epoch": 2979, "lr": 1.3261830861244774e-07}
{"train_loss": 0.01995999924838543, "global_step": 265167, "epoch": 2979, "lr": 1.3257611567222827e-07}
{"train_loss": 0.03347529098391533, "global_step": 265168, "epoch": 2979, "lr": 1.3253392943607945e-07}
{"train_loss": 0.01708257384598255, "global_step": 265169, "epoch": 2979, "lr": 1.3249174990400127e-07}
{"train_loss": 0.04730341583490372, "global_step": 265170, "epoch": 2979, "lr": 1.3244957707599926e-07}
{"train_loss": 0.11940988898277283, "global_step": 265171, "epoch": 2979, "lr": 1.3240741095208452e-07}
{"train_loss": 0.029997486621141434, "global_step": 265172, "epoch": 2979, "lr": 1.3236525153226266e-07}
{"train_loss": 0.042228274047374725, "global_step": 265173, "epoch": 2979, "lr": 1.3232309881652803e-07}
{"train_loss": 0.015955230221152306, "global_step": 265174, "epoch": 2979, "lr": 1.3228095280490293e-07}
{"train_loss": 0.012089604511857033, "global_step": 265175, "epoch": 2979, "lr": 1.3223881349738178e-07}
{"train_loss": 0.05370211973786354, "global_step": 265176, "epoch": 2979, "lr": 1.3219668089397563e-07}
{"train_loss": 0.08820229023694992, "global_step": 265177, "epoch": 2979, "lr": 1.3215455499468453e-07}
{"train_loss": 0.035245560109615326, "global_step": 265178, "epoch": 2979, "lr": 1.321124357995196e-07}
{"train_loss": 0.023224903270602226, "global_step": 265179, "epoch": 2979, "lr": 1.3207032330848635e-07}
{"train_loss": 0.039844125509262085, "global_step": 265180, "epoch": 2979, "lr": 1.3202821752158477e-07}
{"train_loss": 0.024770792573690414, "global_step": 265181, "epoch": 2979, "lr": 1.3198611843882602e-07}
{"train_loss": 0.02930215373635292, "global_step": 265182, "epoch": 2979, "lr": 1.3194402606021006e-07}
{"train_loss": 0.046004727482795715, "global_step": 265183, "epoch": 2979, "lr": 1.31901940385748e-07}
{"train_loss": 0.014775973744690418, "global_step": 265184, "epoch": 2979, "lr": 1.3185986141544538e-07}
{"train_loss": 0.01936199516057968, "global_step": 265185, "epoch": 2979, "lr": 1.3181778914930777e-07}
{"train_loss": 0.030564503744244576, "global_step": 265186, "epoch": 2979, "lr": 1.3177572358733514e-07}
{"train_loss": 0.039927732199430466, "global_step": 265187, "epoch": 2979, "lr": 1.317336647295442e-07}
{"train_loss": 0.05869392305612564, "global_step": 265188, "epoch": 2979, "lr": 1.3169161257592931e-07}
{"train_loss": 0.013768491335213184, "global_step": 265189, "epoch": 2979, "lr": 1.3164956712650168e-07}
{"train_loss": 0.06891122460365295, "global_step": 265190, "epoch": 2979, "lr": 1.3160752838126123e-07}
{"train_loss": 0.0599522665143013, "global_step": 265191, "epoch": 2979, "lr": 1.3156549634022463e-07}
{"train_loss": 0.05326740816235542, "global_step": 265192, "epoch": 2979, "lr": 1.315234710033919e-07}
{"train_loss": 0.06458825618028641, "global_step": 265193, "epoch": 2979, "lr": 1.3148145237076303e-07}
{"train_loss": 0.0938432514667511, "global_step": 265194, "epoch": 2979, "lr": 1.3143944044234914e-07}
{"train_loss": 0.028256988152861595, "global_step": 265195, "epoch": 2979, "lr": 1.3139743521815573e-07}
{"train_loss": 0.08547412604093552, "global_step": 265196, "epoch": 2979, "lr": 1.3135543669819395e-07}
{"train_loss": 0.052162326872348785, "global_step": 265197, "epoch": 2979, "lr": 1.3131344488245822e-07}
{"train_loss": 0.022304194048047066, "global_step": 265198, "epoch": 2979, "lr": 1.3127145977095968e-07}
{"train_loss": 0.05492855980992317, "global_step": 265199, "epoch": 2979, "lr": 1.3122948136370384e-07}
{"train_loss": 0.01679455116391182, "global_step": 265200, "epoch": 2979, "lr": 1.311875096607018e-07}
{"train_loss": 0.028223764151334763, "global_step": 265201, "epoch": 2979, "lr": 1.3114554466194806e-07}
{"train_loss": 0.04711981490254402, "global_step": 265202, "epoch": 2979, "lr": 1.3110358636745368e-07}
{"train_loss": 0.03630648925900459, "global_step": 265203, "epoch": 2979, "lr": 1.310616347772242e-07}
{"train_loss": 0.06823298335075378, "global_step": 265204, "epoch": 2979, "lr": 1.3101968989127078e-07}
{"train_loss": 0.03593927249312401, "global_step": 265205, "epoch": 2979, "lr": 1.3097775170958782e-07}
{"train_loss": 0.056899502873420715, "global_step": 265206, "epoch": 2979, "lr": 1.3093582023219196e-07}
{"train_loss": 0.04321989044547081, "global_step": 265207, "epoch": 2979, "lr": 1.3089389545908325e-07}
{"train_loss": 0.060779470950365067, "global_step": 265208, "epoch": 2979, "lr": 1.308519773902672e-07}
{"train_loss": 0.023964449763298035, "global_step": 265209, "epoch": 2979, "lr": 1.3081006602574942e-07}
{"train_loss": 0.032485660165548325, "global_step": 265210, "epoch": 2979, "lr": 1.3076816136553537e-07}
{"train_loss": 0.020614072680473328, "global_step": 265211, "epoch": 2979, "lr": 1.3072626340963623e-07}
{"train_loss": 0.04926289990544319, "global_step": 265212, "epoch": 2979, "lr": 1.306843721580464e-07}
{"train_loss": 0.029424097388982773, "global_step": 265213, "epoch": 2979, "lr": 1.3064248761078257e-07}
{"train_loss": 0.020483514294028282, "global_step": 265214, "epoch": 2979, "lr": 1.3060060976784471e-07}
{"train_loss": 0.024039514362812042, "global_step": 265215, "epoch": 2979, "lr": 1.3055873862924394e-07}
{"train_loss": 0.02500748261809349, "global_step": 265216, "epoch": 2979, "lr": 1.3051687419498026e-07}
{"train_loss": 0.027572402730584145, "global_step": 265217, "epoch": 2979, "lr": 1.3047501646505922e-07}
{"train_loss": 0.03453199937939644, "global_step": 265218, "epoch": 2979, "lr": 1.3043316543948636e-07}
{"train_loss": 0.040167948984530535, "global_step": 265219, "epoch": 2979, "lr": 1.3039132111826723e-07, "val_loss": 9.676607131958008}
{"train_loss": 0.03859896585345268, "global_step": 265220, "epoch": 2980, "lr": 1.3034948350141296e-07}
{"train_loss": 0.024242376908659935, "global_step": 265221, "epoch": 2980, "lr": 1.3030765258892352e-07}
{"train_loss": 0.02759302407503128, "global_step": 265222, "epoch": 2980, "lr": 1.3026582838081002e-07}
{"train_loss": 0.02865464612841606, "global_step": 265223, "epoch": 2980, "lr": 1.3022401087707247e-07}
{"train_loss": 0.018820839002728462, "global_step": 265224, "epoch": 2980, "lr": 1.3018220007771642e-07}
{"train_loss": 0.05902406945824623, "global_step": 265225, "epoch": 2980, "lr": 1.3014039598275296e-07}
{"train_loss": 0.0663202777504921, "global_step": 265226, "epoch": 2980, "lr": 1.300985985921821e-07}
{"train_loss": 0.04186078906059265, "global_step": 265227, "epoch": 2980, "lr": 1.300568079060094e-07}
{"train_loss": 0.04667247459292412, "global_step": 265228, "epoch": 2980, "lr": 1.3001502392424592e-07}
{"train_loss": 0.03547239676117897, "global_step": 265229, "epoch": 2980, "lr": 1.299732466468917e-07}
{"train_loss": 0.036431409418582916, "global_step": 265230, "epoch": 2980, "lr": 1.2993147607395784e-07}
{"train_loss": 0.08508564531803131, "global_step": 265231, "epoch": 2980, "lr": 1.2988971220544432e-07}
{"train_loss": 0.06860890984535217, "global_step": 265232, "epoch": 2980, "lr": 1.2984795504136227e-07}
{"train_loss": 0.018638186156749725, "global_step": 265233, "epoch": 2980, "lr": 1.2980620458171167e-07}
{"train_loss": 0.011818831786513329, "global_step": 265234, "epoch": 2980, "lr": 1.2976446082650362e-07}
{"train_loss": 0.03585272282361984, "global_step": 265235, "epoch": 2980, "lr": 1.2972272377573814e-07}
{"train_loss": 0.04311143979430199, "global_step": 265236, "epoch": 2980, "lr": 1.296809934294263e-07}
{"train_loss": 0.022872626781463623, "global_step": 265237, "epoch": 2980, "lr": 1.2963926978756812e-07}
{"train_loss": 0.08889943361282349, "global_step": 265238, "epoch": 2980, "lr": 1.2959755285016918e-07}
{"train_loss": 0.044211361557245255, "global_step": 265239, "epoch": 2980, "lr": 1.2955584261724607e-07}
{"train_loss": 0.022656908258795738, "global_step": 265240, "epoch": 2980, "lr": 1.2951413908878775e-07}
{"train_loss": 0.05834595859050751, "global_step": 265241, "epoch": 2980, "lr": 1.2947244226481635e-07}
{"train_loss": 0.04627073556184769, "global_step": 265242, "epoch": 2980, "lr": 1.294307521453264e-07}
{"train_loss": 0.015550943091511726, "global_step": 265243, "epoch": 2980, "lr": 1.293890687303234e-07}
{"train_loss": 0.052298080176115036, "global_step": 265244, "epoch": 2980, "lr": 1.293473920198185e-07}
{"train_loss": 0.07178616523742676, "global_step": 265245, "epoch": 2980, "lr": 1.2930572201381718e-07}
{"train_loss": 0.06491594016551971, "global_step": 265246, "epoch": 2980, "lr": 1.2926405871231951e-07}
{"train_loss": 0.04030345007777214, "global_step": 265247, "epoch": 2980, "lr": 1.2922240211533653e-07}
{"train_loss": 0.05408184975385666, "global_step": 265248, "epoch": 2980, "lr": 1.2918075222287385e-07}
{"train_loss": 0.02570226602256298, "global_step": 265249, "epoch": 2980, "lr": 1.2913910903493143e-07}
{"train_loss": 0.07154437899589539, "global_step": 265250, "epoch": 2980, "lr": 1.2909747255152038e-07}
{"train_loss": 0.038126833736896515, "global_step": 265251, "epoch": 2980, "lr": 1.2905584277264627e-07}
{"train_loss": 0.06763952225446701, "global_step": 265252, "epoch": 2980, "lr": 1.2901421969830907e-07}
{"train_loss": 0.04198823869228363, "global_step": 265253, "epoch": 2980, "lr": 1.289726033285199e-07}
{"train_loss": 0.038023173809051514, "global_step": 265254, "epoch": 2980, "lr": 1.2893099366327876e-07}
{"train_loss": 0.030209919437766075, "global_step": 265255, "epoch": 2980, "lr": 1.2888939070259675e-07}
{"train_loss": 0.022660784423351288, "global_step": 265256, "epoch": 2980, "lr": 1.2884779444647944e-07}
{"train_loss": 0.05280793085694313, "global_step": 265257, "epoch": 2980, "lr": 1.2880620489493234e-07}
{"train_loss": 0.08601956069469452, "global_step": 265258, "epoch": 2980, "lr": 1.2876462204795547e-07}
{"train_loss": 0.07651575654745102, "global_step": 265259, "epoch": 2980, "lr": 1.2872304590555994e-07}
{"train_loss": 0.02827860787510872, "global_step": 265260, "epoch": 2980, "lr": 1.286814764677513e-07}
{"train_loss": 0.05162970349192619, "global_step": 265261, "epoch": 2980, "lr": 1.2863991373452955e-07}
{"train_loss": 0.05197950825095177, "global_step": 265262, "epoch": 2980, "lr": 1.285983577059058e-07}
{"train_loss": 0.03431347385048866, "global_step": 265263, "epoch": 2980, "lr": 1.2855680838188554e-07}
{"train_loss": 0.0495125949382782, "global_step": 265264, "epoch": 2980, "lr": 1.285152657624744e-07}
{"train_loss": 0.06362323462963104, "global_step": 265265, "epoch": 2980, "lr": 1.2847372984767236e-07}
{"train_loss": 0.03551190719008446, "global_step": 265266, "epoch": 2980, "lr": 1.2843220063749606e-07}
{"train_loss": 0.027178512886166573, "global_step": 265267, "epoch": 2980, "lr": 1.2839067813193994e-07}
{"train_loss": 0.014353906735777855, "global_step": 265268, "epoch": 2980, "lr": 1.2834916233100958e-07}
{"train_loss": 0.019322194159030914, "global_step": 265269, "epoch": 2980, "lr": 1.2830765323472162e-07}
{"train_loss": 0.04711242765188217, "global_step": 265270, "epoch": 2980, "lr": 1.2826615084307048e-07}
{"train_loss": 0.01297254953533411, "global_step": 265271, "epoch": 2980, "lr": 1.282246551560673e-07}
{"train_loss": 0.059857822954654694, "global_step": 265272, "epoch": 2980, "lr": 1.2818316617371761e-07}
{"train_loss": 0.03361693397164345, "global_step": 265273, "epoch": 2980, "lr": 1.28141683896027e-07}
{"train_loss": 0.03137129172682762, "global_step": 265274, "epoch": 2980, "lr": 1.281002083229954e-07}
{"train_loss": 0.024763504043221474, "global_step": 265275, "epoch": 2980, "lr": 1.2805873945463954e-07}
{"train_loss": 0.02839609980583191, "global_step": 265276, "epoch": 2980, "lr": 1.280172772909538e-07}
{"train_loss": 0.0725223496556282, "global_step": 265277, "epoch": 2980, "lr": 1.2797582183194935e-07}
{"train_loss": 0.032723475247621536, "global_step": 265278, "epoch": 2980, "lr": 1.279343730776261e-07}
{"train_loss": 0.03409077972173691, "global_step": 265279, "epoch": 2980, "lr": 1.2789293102800083e-07}
{"train_loss": 0.029680823907256126, "global_step": 265280, "epoch": 2980, "lr": 1.2785149568306788e-07}
{"train_loss": 0.06844376027584076, "global_step": 265281, "epoch": 2980, "lr": 1.2781006704283838e-07}
{"train_loss": 0.0345534048974514, "global_step": 265282, "epoch": 2980, "lr": 1.277686451073179e-07}
{"train_loss": 0.03323172777891159, "global_step": 265283, "epoch": 2980, "lr": 1.2772722987651197e-07}
{"train_loss": 0.014953482896089554, "global_step": 265284, "epoch": 2980, "lr": 1.2768582135042617e-07}
{"train_loss": 0.05124105140566826, "global_step": 265285, "epoch": 2980, "lr": 1.2764441952906603e-07}
{"train_loss": 0.04571281373500824, "global_step": 265286, "epoch": 2980, "lr": 1.2760302441243155e-07}
{"train_loss": 0.03902721405029297, "global_step": 265287, "epoch": 2980, "lr": 1.2756163600053383e-07}
{"train_loss": 0.021126436069607735, "global_step": 265288, "epoch": 2980, "lr": 1.2752025429337843e-07}
{"train_loss": 0.020222194492816925, "global_step": 265289, "epoch": 2980, "lr": 1.2747887929097092e-07}
{"train_loss": 0.032272931188344955, "global_step": 265290, "epoch": 2980, "lr": 1.274375109933168e-07}
{"train_loss": 0.027405891567468643, "global_step": 265291, "epoch": 2980, "lr": 1.2739614940042167e-07}
{"train_loss": 0.06739792227745056, "global_step": 265292, "epoch": 2980, "lr": 1.2735479451228549e-07}
{"train_loss": 0.05123048648238182, "global_step": 265293, "epoch": 2980, "lr": 1.2731344632892494e-07}
{"train_loss": 0.025460733100771904, "global_step": 265294, "epoch": 2980, "lr": 1.2727210485033448e-07}
{"train_loss": 0.02745230123400688, "global_step": 265295, "epoch": 2980, "lr": 1.272307700765252e-07}
{"train_loss": 0.05815141648054123, "global_step": 265296, "epoch": 2980, "lr": 1.271894420075026e-07}
{"train_loss": 0.03168507292866707, "global_step": 265297, "epoch": 2980, "lr": 1.271481206432723e-07}
{"train_loss": 0.10086670517921448, "global_step": 265298, "epoch": 2980, "lr": 1.2710680598383428e-07}
{"train_loss": 0.030059916898608208, "global_step": 265299, "epoch": 2980, "lr": 1.270654980292052e-07}
{"train_loss": 0.05543581768870354, "global_step": 265300, "epoch": 2980, "lr": 1.2702419677937948e-07}
{"train_loss": 0.045102864503860474, "global_step": 265301, "epoch": 2980, "lr": 1.2698290223436825e-07}
{"train_loss": 0.03646238520741463, "global_step": 265302, "epoch": 2980, "lr": 1.269416143941826e-07}
{"train_loss": 0.09639996290206909, "global_step": 265303, "epoch": 2980, "lr": 1.26900333258817e-07}
{"train_loss": 0.031176390126347542, "global_step": 265304, "epoch": 2980, "lr": 1.2685905882828253e-07}
{"train_loss": 0.04152390733361244, "global_step": 265305, "epoch": 2980, "lr": 1.2681779110257918e-07}
{"train_loss": 0.06440038979053497, "global_step": 265306, "epoch": 2980, "lr": 1.2677653008172363e-07}
{"train_loss": 0.024074124172329903, "global_step": 265307, "epoch": 2980, "lr": 1.2673527576571031e-07}
{"train_loss": 0.043492921795486736, "global_step": 265308, "epoch": 2980, "lr": 1.2669402815455588e-07, "val_loss": 9.65638542175293, "train_action_mse_error": 15.968643188476562}
{"train_loss": 0.0284259095788002, "global_step": 265309, "epoch": 2981, "lr": 1.266527872482548e-07}
{"train_loss": 0.03142501041293144, "global_step": 265310, "epoch": 2981, "lr": 1.2661155304681815e-07}
{"train_loss": 0.026547996327280998, "global_step": 265311, "epoch": 2981, "lr": 1.265703255502515e-07}
{"train_loss": 0.0555172823369503, "global_step": 265312, "epoch": 2981, "lr": 1.265291047585604e-07}
{"train_loss": 0.01778969168663025, "global_step": 265313, "epoch": 2981, "lr": 1.264878906717448e-07}
{"train_loss": 0.04106249287724495, "global_step": 265314, "epoch": 2981, "lr": 1.2644668328982146e-07}
{"train_loss": 0.08594135195016861, "global_step": 265315, "epoch": 2981, "lr": 1.2640548261278474e-07}
{"train_loss": 0.03780962899327278, "global_step": 265316, "epoch": 2981, "lr": 1.2636428864064576e-07}
{"train_loss": 0.02016463316977024, "global_step": 265317, "epoch": 2981, "lr": 1.263231013734101e-07}
{"train_loss": 0.05716782808303833, "global_step": 265318, "epoch": 2981, "lr": 1.2628192081108325e-07}
{"train_loss": 0.06336616724729538, "global_step": 265319, "epoch": 2981, "lr": 1.2624074695366527e-07}
{"train_loss": 0.032806772738695145, "global_step": 265320, "epoch": 2981, "lr": 1.261995798011728e-07}
{"train_loss": 0.06407762318849564, "global_step": 265321, "epoch": 2981, "lr": 1.2615841935360028e-07}
{"train_loss": 0.020369328558444977, "global_step": 265322, "epoch": 2981, "lr": 1.2611726561095883e-07}
{"train_loss": 0.03397876024246216, "global_step": 265323, "epoch": 2981, "lr": 1.260761185732595e-07}
{"train_loss": 0.05693860352039337, "global_step": 265324, "epoch": 2981, "lr": 1.2603497824049682e-07}
{"train_loss": 0.06125783175230026, "global_step": 265325, "epoch": 2981, "lr": 1.2599384461267627e-07}
{"train_loss": 0.03343874216079712, "global_step": 265326, "epoch": 2981, "lr": 1.2595271768981455e-07}
{"train_loss": 0.02831781655550003, "global_step": 265327, "epoch": 2981, "lr": 1.2591159747190605e-07}
{"train_loss": 0.03028496727347374, "global_step": 265328, "epoch": 2981, "lr": 1.2587048395896195e-07}
{"train_loss": 0.00901907216757536, "global_step": 265329, "epoch": 2981, "lr": 1.2582937715098775e-07}
{"train_loss": 0.058068763464689255, "global_step": 265330, "epoch": 2981, "lr": 1.25788277047989e-07}
{"train_loss": 0.046002138406038284, "global_step": 265331, "epoch": 2981, "lr": 1.2574718364996573e-07}
{"train_loss": 0.04746278002858162, "global_step": 265332, "epoch": 2981, "lr": 1.2570609695693458e-07}
{"train_loss": 0.03948473930358887, "global_step": 265333, "epoch": 2981, "lr": 1.2566501696889e-07}
{"train_loss": 0.02926608920097351, "global_step": 265334, "epoch": 2981, "lr": 1.2562394368583751e-07}
{"train_loss": 0.04063175991177559, "global_step": 265335, "epoch": 2981, "lr": 1.255828771077938e-07}
{"train_loss": 0.07508662343025208, "global_step": 265336, "epoch": 2981, "lr": 1.2554181723475889e-07}
{"train_loss": 0.05051443725824356, "global_step": 265337, "epoch": 2981, "lr": 1.2550076406673272e-07}
{"train_loss": 0.042440611869096756, "global_step": 265338, "epoch": 2981, "lr": 1.2545971760372643e-07}
{"train_loss": 0.015093907713890076, "global_step": 265339, "epoch": 2981, "lr": 1.2541867784574002e-07}
{"train_loss": 0.006912085693329573, "global_step": 265340, "epoch": 2981, "lr": 1.2537764479279013e-07}
{"train_loss": 0.054884810000658035, "global_step": 265341, "epoch": 2981, "lr": 1.253366184448712e-07}
{"train_loss": 0.015781991183757782, "global_step": 265342, "epoch": 2981, "lr": 1.2529559880199436e-07}
{"train_loss": 0.04476957395672798, "global_step": 265343, "epoch": 2981, "lr": 1.252545858641596e-07}
{"train_loss": 0.027679260820150375, "global_step": 265344, "epoch": 2981, "lr": 1.2521357963138358e-07}
{"train_loss": 0.05664317309856415, "global_step": 265345, "epoch": 2981, "lr": 1.2517258010366073e-07}
{"train_loss": 0.0259373988956213, "global_step": 265346, "epoch": 2981, "lr": 1.2513158728100217e-07}
{"train_loss": 0.09523668140172958, "global_step": 265347, "epoch": 2981, "lr": 1.2509060116340788e-07}
{"train_loss": 0.02661825157701969, "global_step": 265348, "epoch": 2981, "lr": 1.2504962175089452e-07}
{"train_loss": 0.04659571126103401, "global_step": 265349, "epoch": 2981, "lr": 1.25008649043451e-07}
{"train_loss": 0.026364728808403015, "global_step": 265350, "epoch": 2981, "lr": 1.2496768304109952e-07}
{"train_loss": 0.05031057447195053, "global_step": 265351, "epoch": 2981, "lr": 1.2492672374383452e-07}
{"train_loss": 0.03438594564795494, "global_step": 265352, "epoch": 2981, "lr": 1.248857711516671e-07}
{"train_loss": 0.05285096913576126, "global_step": 265353, "epoch": 2981, "lr": 1.248448252645973e-07}
{"train_loss": 0.03704249486327171, "global_step": 265354, "epoch": 2981, "lr": 1.2480388608263616e-07}
{"train_loss": 0.05852867290377617, "global_step": 265355, "epoch": 2981, "lr": 1.2476295360578926e-07}
{"train_loss": 0.01775078848004341, "global_step": 265356, "epoch": 2981, "lr": 1.2472202783406216e-07}
{"train_loss": 0.04168960452079773, "global_step": 265357, "epoch": 2981, "lr": 1.2468110876745486e-07}
{"train_loss": 0.05540333315730095, "global_step": 265358, "epoch": 2981, "lr": 1.2464019640597292e-07}
{"train_loss": 0.02602669782936573, "global_step": 265359, "epoch": 2981, "lr": 1.2459929074963296e-07}
{"train_loss": 0.07723512500524521, "global_step": 265360, "epoch": 2981, "lr": 1.2455839179842944e-07}
{"train_loss": 0.05357668176293373, "global_step": 265361, "epoch": 2981, "lr": 1.245174995523679e-07}
{"train_loss": 0.055365800857543945, "global_step": 265362, "epoch": 2981, "lr": 1.244766140114595e-07}
{"train_loss": 0.03650102764368057, "global_step": 265363, "epoch": 2981, "lr": 1.2443573517570971e-07}
{"train_loss": 0.027021897956728935, "global_step": 265364, "epoch": 2981, "lr": 1.2439486304511862e-07}
{"train_loss": 0.0311370100826025, "global_step": 265365, "epoch": 2981, "lr": 1.2435399761969724e-07}
{"train_loss": 0.041504353284835815, "global_step": 265366, "epoch": 2981, "lr": 1.2431313889944563e-07}
{"train_loss": 0.02520921640098095, "global_step": 265367, "epoch": 2981, "lr": 1.2427228688437487e-07}
{"train_loss": 0.04910535365343094, "global_step": 265368, "epoch": 2981, "lr": 1.2423144157448496e-07}
{"train_loss": 0.03296128287911415, "global_step": 265369, "epoch": 2981, "lr": 1.2419060296978702e-07}
{"train_loss": 0.038250841200351715, "global_step": 265370, "epoch": 2981, "lr": 1.2414977107028102e-07}
{"train_loss": 0.02631603553891182, "global_step": 265371, "epoch": 2981, "lr": 1.241089458759781e-07}
{"train_loss": 0.02610853873193264, "global_step": 265372, "epoch": 2981, "lr": 1.2406812738687823e-07}
{"train_loss": 0.058984044939279556, "global_step": 265373, "epoch": 2981, "lr": 1.2402731560298696e-07}
{"train_loss": 0.016810694709420204, "global_step": 265374, "epoch": 2981, "lr": 1.239865105243154e-07}
{"train_loss": 0.038339827209711075, "global_step": 265375, "epoch": 2981, "lr": 1.2394571215086915e-07}
{"train_loss": 0.053407568484544754, "global_step": 265376, "epoch": 2981, "lr": 1.2390492048264813e-07}
{"train_loss": 0.05311061069369316, "global_step": 265377, "epoch": 2981, "lr": 1.2386413551965793e-07}
{"train_loss": 0.04360826313495636, "global_step": 265378, "epoch": 2981, "lr": 1.2382335726190964e-07}
{"train_loss": 0.07195825129747391, "global_step": 265379, "epoch": 2981, "lr": 1.2378258570940327e-07}
{"train_loss": 0.01283788587898016, "global_step": 265380, "epoch": 2981, "lr": 1.2374182086214992e-07}
{"train_loss": 0.04168824478983879, "global_step": 265381, "epoch": 2981, "lr": 1.2370106272014404e-07}
{"train_loss": 0.056087471544742584, "global_step": 265382, "epoch": 2981, "lr": 1.2366031128340783e-07}
{"train_loss": 0.03449932858347893, "global_step": 265383, "epoch": 2981, "lr": 1.236195665519302e-07}
{"train_loss": 0.06894689798355103, "global_step": 265384, "epoch": 2981, "lr": 1.2357882852572776e-07}
{"train_loss": 0.05217014253139496, "global_step": 265385, "epoch": 2981, "lr": 1.235380972048006e-07}
{"train_loss": 0.06731762737035751, "global_step": 265386, "epoch": 2981, "lr": 1.2349737258915417e-07}
{"train_loss": 0.07568760216236115, "global_step": 265387, "epoch": 2981, "lr": 1.2345665467879964e-07}
{"train_loss": 0.023307908326387405, "global_step": 265388, "epoch": 2981, "lr": 1.2341594347373698e-07}
{"train_loss": 0.04007434844970703, "global_step": 265389, "epoch": 2981, "lr": 1.2337523897397173e-07}
{"train_loss": 0.040740594267845154, "global_step": 265390, "epoch": 2981, "lr": 1.2333454117951504e-07}
{"train_loss": 0.041487861424684525, "global_step": 265391, "epoch": 2981, "lr": 1.2329385009036132e-07}
{"train_loss": 0.09140150994062424, "global_step": 265392, "epoch": 2981, "lr": 1.2325316570652722e-07}
{"train_loss": 0.042717766016721725, "global_step": 265393, "epoch": 2981, "lr": 1.2321248802801278e-07}
{"train_loss": 0.06815997511148453, "global_step": 265394, "epoch": 2981, "lr": 1.231718170548235e-07}
{"train_loss": 0.04174071177840233, "global_step": 265395, "epoch": 2981, "lr": 1.2313115278697052e-07}
{"train_loss": 0.06301099807024002, "global_step": 265396, "epoch": 2981, "lr": 1.2309049522444827e-07}
{"train_loss": 0.042950991347557706, "global_step": 265397, "epoch": 2981, "lr": 1.230498443672734e-07, "val_loss": 9.558222770690918}
{"train_loss": 0.037497203797101974, "global_step": 265398, "epoch": 2982, "lr": 1.2300920021544592e-07}
{"train_loss": 0.04736408591270447, "global_step": 265399, "epoch": 2982, "lr": 1.229685627689714e-07}
{"train_loss": 0.03683202713727951, "global_step": 265400, "epoch": 2982, "lr": 1.2292793202785536e-07}
{"train_loss": 0.019657155498862267, "global_step": 265401, "epoch": 2982, "lr": 1.2288730799210336e-07}
{"train_loss": 0.027608495205640793, "global_step": 265402, "epoch": 2982, "lr": 1.2284669066172093e-07}
{"train_loss": 0.025457527488470078, "global_step": 265403, "epoch": 2982, "lr": 1.2280608003671368e-07}
{"train_loss": 0.03429625555872917, "global_step": 265404, "epoch": 2982, "lr": 1.2276547611709265e-07}
{"train_loss": 0.026813944801688194, "global_step": 265405, "epoch": 2982, "lr": 1.2272487890285233e-07}
{"train_loss": 0.05541253462433815, "global_step": 265406, "epoch": 2982, "lr": 1.226842883940038e-07}
{"train_loss": 0.037097468972206116, "global_step": 265407, "epoch": 2982, "lr": 1.2264370459055262e-07}
{"train_loss": 0.0921189934015274, "global_step": 265408, "epoch": 2982, "lr": 1.2260312749250436e-07}
{"train_loss": 0.035081155598163605, "global_step": 265409, "epoch": 2982, "lr": 1.2256255709986454e-07}
{"train_loss": 0.04440166801214218, "global_step": 265410, "epoch": 2982, "lr": 1.225219934126387e-07}
{"train_loss": 0.04456508904695511, "global_step": 265411, "epoch": 2982, "lr": 1.2248143643083242e-07}
{"train_loss": 0.031122393906116486, "global_step": 265412, "epoch": 2982, "lr": 1.2244088615445126e-07}
{"train_loss": 0.08595003187656403, "global_step": 265413, "epoch": 2982, "lr": 1.224003425834952e-07}
{"train_loss": 0.021958284080028534, "global_step": 265414, "epoch": 2982, "lr": 1.2235980571798089e-07}
{"train_loss": 0.047609634697437286, "global_step": 265415, "epoch": 2982, "lr": 1.223192755579028e-07}
{"train_loss": 0.05129239335656166, "global_step": 265416, "epoch": 2982, "lr": 1.22278752103272e-07}
{"train_loss": 0.02153637260198593, "global_step": 265417, "epoch": 2982, "lr": 1.2223823535409407e-07}
{"train_loss": 0.019352784380316734, "global_step": 265418, "epoch": 2982, "lr": 1.2219772531037455e-07}
{"train_loss": 0.03832325339317322, "global_step": 265419, "epoch": 2982, "lr": 1.2215722197211344e-07}
{"train_loss": 0.0060713323764503, "global_step": 265420, "epoch": 2982, "lr": 1.2211672533932184e-07}
{"train_loss": 0.04048057645559311, "global_step": 265421, "epoch": 2982, "lr": 1.220762354120053e-07}
{"train_loss": 0.022835206240415573, "global_step": 265422, "epoch": 2982, "lr": 1.2203575219016384e-07}
{"train_loss": 0.03318076580762863, "global_step": 265423, "epoch": 2982, "lr": 1.2199527567381407e-07}
{"train_loss": 0.05538516491651535, "global_step": 265424, "epoch": 2982, "lr": 1.2195480586295049e-07}
{"train_loss": 0.09938151389360428, "global_step": 265425, "epoch": 2982, "lr": 1.2191434275757863e-07}
{"train_loss": 0.04567323997616768, "global_step": 265426, "epoch": 2982, "lr": 1.2187388635770956e-07}
{"train_loss": 0.019073279574513435, "global_step": 265427, "epoch": 2982, "lr": 1.218334366633489e-07}
{"train_loss": 0.0339219830930233, "global_step": 265428, "epoch": 2982, "lr": 1.2179299367449658e-07}
{"train_loss": 0.046680957078933716, "global_step": 265429, "epoch": 2982, "lr": 1.2175255739116376e-07}
{"train_loss": 0.019818417727947235, "global_step": 265430, "epoch": 2982, "lr": 1.217121278133504e-07}
{"train_loss": 0.023945976048707962, "global_step": 265431, "epoch": 2982, "lr": 1.216717049410676e-07}
{"train_loss": 0.03609664365649223, "global_step": 265432, "epoch": 2982, "lr": 1.216312887743154e-07}
{"train_loss": 0.05145309492945671, "global_step": 265433, "epoch": 2982, "lr": 1.2159087931310486e-07}
{"train_loss": 0.04217499494552612, "global_step": 265434, "epoch": 2982, "lr": 1.2155047655743601e-07}
{"train_loss": 0.05649028345942497, "global_step": 265435, "epoch": 2982, "lr": 1.2151008050731993e-07}
{"train_loss": 0.03484414145350456, "global_step": 265436, "epoch": 2982, "lr": 1.2146969116275665e-07}
{"train_loss": 0.02627839893102646, "global_step": 265437, "epoch": 2982, "lr": 1.214293085237572e-07}
{"train_loss": 0.023264911025762558, "global_step": 265438, "epoch": 2982, "lr": 1.2138893259032169e-07}
{"train_loss": 0.02906242571771145, "global_step": 265439, "epoch": 2982, "lr": 1.213485633624556e-07}
{"train_loss": 0.02652806229889393, "global_step": 265440, "epoch": 2982, "lr": 1.2130820084017002e-07}
{"train_loss": 0.02860010787844658, "global_step": 265441, "epoch": 2982, "lr": 1.2126784502345944e-07}
{"train_loss": 0.01276916079223156, "global_step": 265442, "epoch": 2982, "lr": 1.2122749591234607e-07}
{"train_loss": 0.04001036286354065, "global_step": 265443, "epoch": 2982, "lr": 1.2118715350681876e-07}
{"train_loss": 0.04563548043370247, "global_step": 265444, "epoch": 2982, "lr": 1.211468178068942e-07}
{"train_loss": 0.05067510902881622, "global_step": 265445, "epoch": 2982, "lr": 1.211064888125668e-07}
{"train_loss": 0.050556715577840805, "global_step": 265446, "epoch": 2982, "lr": 1.2106616652385327e-07}
{"train_loss": 0.055141665041446686, "global_step": 265447, "epoch": 2982, "lr": 1.2102585094075357e-07}
{"train_loss": 0.027584008872509003, "global_step": 265448, "epoch": 2982, "lr": 1.2098554206327884e-07}
{"train_loss": 0.043153513222932816, "global_step": 265449, "epoch": 2982, "lr": 1.2094523989142348e-07}
{"train_loss": 0.026035772636532784, "global_step": 265450, "epoch": 2982, "lr": 1.2090494442520418e-07}
{"train_loss": 0.02705404907464981, "global_step": 265451, "epoch": 2982, "lr": 1.2086465566461536e-07}
{"train_loss": 0.04651818424463272, "global_step": 265452, "epoch": 2982, "lr": 1.208243736096737e-07}
{"train_loss": 0.03411094844341278, "global_step": 265453, "epoch": 2982, "lr": 1.2078409826037363e-07}
{"train_loss": 0.030173301696777344, "global_step": 265454, "epoch": 2982, "lr": 1.207438296167318e-07}
{"train_loss": 0.052345942705869675, "global_step": 265455, "epoch": 2982, "lr": 1.2070356767874268e-07}
{"train_loss": 0.052406199276447296, "global_step": 265456, "epoch": 2982, "lr": 1.2066331244642294e-07}
{"train_loss": 0.01904165931046009, "global_step": 265457, "epoch": 2982, "lr": 1.2062306391976698e-07}
{"train_loss": 0.07665510475635529, "global_step": 265458, "epoch": 2982, "lr": 1.205828220987859e-07}
{"train_loss": 0.03456255421042442, "global_step": 265459, "epoch": 2982, "lr": 1.205425869834853e-07}
{"train_loss": 0.03437051177024841, "global_step": 265460, "epoch": 2982, "lr": 1.205023585738707e-07}
{"train_loss": 0.06722259521484375, "global_step": 265461, "epoch": 2982, "lr": 1.2046213686994768e-07}
{"train_loss": 0.02701498754322529, "global_step": 265462, "epoch": 2982, "lr": 1.2042192187172176e-07}
{"train_loss": 0.08076222240924835, "global_step": 265463, "epoch": 2982, "lr": 1.2038171357919291e-07}
{"train_loss": 0.06081722304224968, "global_step": 265464, "epoch": 2982, "lr": 1.2034151199237232e-07}
{"train_loss": 0.04576161131262779, "global_step": 265465, "epoch": 2982, "lr": 1.2030131711126548e-07}
{"train_loss": 0.01753091625869274, "global_step": 265466, "epoch": 2982, "lr": 1.202611289358724e-07}
{"train_loss": 0.048565980046987534, "global_step": 265467, "epoch": 2982, "lr": 1.2022094746620415e-07}
{"train_loss": 0.007756849285215139, "global_step": 265468, "epoch": 2982, "lr": 1.2018077270226636e-07}
{"train_loss": 0.05453743413090706, "global_step": 265469, "epoch": 2982, "lr": 1.2014060464405895e-07}
{"train_loss": 0.031006990000605583, "global_step": 265470, "epoch": 2982, "lr": 1.201004432915931e-07}
{"train_loss": 0.06112747639417648, "global_step": 265471, "epoch": 2982, "lr": 1.2006028864487428e-07}
{"train_loss": 0.050765328109264374, "global_step": 265472, "epoch": 2982, "lr": 1.2002014070390255e-07}
{"train_loss": 0.019512075930833817, "global_step": 265473, "epoch": 2982, "lr": 1.1997999946868344e-07}
{"train_loss": 0.04933341220021248, "global_step": 265474, "epoch": 2982, "lr": 1.1993986493922805e-07}
{"train_loss": 0.04958581551909447, "global_step": 265475, "epoch": 2982, "lr": 1.1989973711553636e-07}
{"train_loss": 0.032792381942272186, "global_step": 265476, "epoch": 2982, "lr": 1.1985961599761952e-07}
{"train_loss": 0.061234619468450546, "global_step": 265477, "epoch": 2982, "lr": 1.198195015854775e-07}
{"train_loss": 0.03909415751695633, "global_step": 265478, "epoch": 2982, "lr": 1.1977939387912141e-07}
{"train_loss": 0.05652580410242081, "global_step": 265479, "epoch": 2982, "lr": 1.1973929287854568e-07}
{"train_loss": 0.03390726447105408, "global_step": 265480, "epoch": 2982, "lr": 1.1969919858377254e-07}
{"train_loss": 0.08266317844390869, "global_step": 265481, "epoch": 2982, "lr": 1.1965911099479088e-07}
{"train_loss": 0.017423372715711594, "global_step": 265482, "epoch": 2982, "lr": 1.1961903011161735e-07}
{"train_loss": 0.051864542067050934, "global_step": 265483, "epoch": 2982, "lr": 1.195789559342464e-07}
{"train_loss": 0.03067135438323021, "global_step": 265484, "epoch": 2982, "lr": 1.1953888846269467e-07}
{"train_loss": 0.026966460049152374, "global_step": 265485, "epoch": 2982, "lr": 1.194988276969622e-07}
{"train_loss": 0.04061864401224289, "global_step": 265486, "epoch": 2982, "lr": 1.1945877363706005e-07, "val_loss": 9.665169715881348}
{"train_loss": 0.053563062101602554, "global_step": 265487, "epoch": 2983, "lr": 1.1941872628298267e-07}
{"train_loss": 0.049026090651750565, "global_step": 265488, "epoch": 2983, "lr": 1.193786856347412e-07}
{"train_loss": 0.049470461905002594, "global_step": 265489, "epoch": 2983, "lr": 1.193386516923467e-07}
{"train_loss": 0.03651467710733414, "global_step": 265490, "epoch": 2983, "lr": 1.1929862445579364e-07}
{"train_loss": 0.06273076683282852, "global_step": 265491, "epoch": 2983, "lr": 1.1925860392509313e-07}
{"train_loss": 0.021327724680304527, "global_step": 265492, "epoch": 2983, "lr": 1.1921859010025627e-07}
{"train_loss": 0.025753429159522057, "global_step": 265493, "epoch": 2983, "lr": 1.1917858298127748e-07}
{"train_loss": 0.05858505517244339, "global_step": 265494, "epoch": 2983, "lr": 1.191385825681679e-07}
{"train_loss": 0.01935824379324913, "global_step": 265495, "epoch": 2983, "lr": 1.1909858886093306e-07}
{"train_loss": 0.05474647507071495, "global_step": 265496, "epoch": 2983, "lr": 1.1905860185957851e-07}
{"train_loss": 0.06924503296613693, "global_step": 265497, "epoch": 2983, "lr": 1.1901862156410427e-07}
{"train_loss": 0.029196567833423615, "global_step": 265498, "epoch": 2983, "lr": 1.1897864797452141e-07}
{"train_loss": 0.027937790378928185, "global_step": 265499, "epoch": 2983, "lr": 1.1893868109083551e-07}
{"train_loss": 0.02378099411725998, "global_step": 265500, "epoch": 2983, "lr": 1.188987209130521e-07}
{"train_loss": 0.03099185973405838, "global_step": 265501, "epoch": 2983, "lr": 1.1885876744117119e-07}
{"train_loss": 0.047595661133527756, "global_step": 265502, "epoch": 2983, "lr": 1.1881882067520389e-07}
{"train_loss": 0.035882215946912766, "global_step": 265503, "epoch": 2983, "lr": 1.187788806151502e-07}
{"train_loss": 0.018857410177588463, "global_step": 265504, "epoch": 2983, "lr": 1.1873894726102119e-07}
{"train_loss": 0.026993082836270332, "global_step": 265505, "epoch": 2983, "lr": 1.1869902061282245e-07}
{"train_loss": 0.03481435030698776, "global_step": 265506, "epoch": 2983, "lr": 1.1865910067055397e-07}
{"train_loss": 0.04981818422675133, "global_step": 265507, "epoch": 2983, "lr": 1.1861918743422129e-07}
{"train_loss": 0.042680900543928146, "global_step": 265508, "epoch": 2983, "lr": 1.1857928090383552e-07}
{"train_loss": 0.01909601129591465, "global_step": 265509, "epoch": 2983, "lr": 1.1853938107940221e-07}
{"train_loss": 0.03779730200767517, "global_step": 265510, "epoch": 2983, "lr": 1.1849948796091581e-07}
{"train_loss": 0.04034631699323654, "global_step": 265511, "epoch": 2983, "lr": 1.1845960154839298e-07}
{"train_loss": 0.015978945419192314, "global_step": 265512, "epoch": 2983, "lr": 1.1841972184183369e-07}
{"train_loss": 0.03876115009188652, "global_step": 265513, "epoch": 2983, "lr": 1.1837984884124909e-07}
{"train_loss": 0.06579387933015823, "global_step": 265514, "epoch": 2983, "lr": 1.1833998254663359e-07}
{"train_loss": 0.02871585264801979, "global_step": 265515, "epoch": 2983, "lr": 1.1830012295800386e-07}
{"train_loss": 0.04542410001158714, "global_step": 265516, "epoch": 2983, "lr": 1.1826027007535989e-07}
{"train_loss": 0.012672930024564266, "global_step": 265517, "epoch": 2983, "lr": 1.1822042389871279e-07}
{"train_loss": 0.03874966874718666, "global_step": 265518, "epoch": 2983, "lr": 1.1818058442805701e-07}
{"train_loss": 0.04294611141085625, "global_step": 265519, "epoch": 2983, "lr": 1.1814075166340366e-07}
{"train_loss": 0.03310490772128105, "global_step": 265520, "epoch": 2983, "lr": 1.1810092560475827e-07}
{"train_loss": 0.02866242825984955, "global_step": 265521, "epoch": 2983, "lr": 1.1806110625212641e-07}
{"train_loss": 0.012790163047611713, "global_step": 265522, "epoch": 2983, "lr": 1.1802129360551917e-07}
{"train_loss": 0.0726085975766182, "global_step": 265523, "epoch": 2983, "lr": 1.17981487664931e-07}
{"train_loss": 0.038430534303188324, "global_step": 265524, "epoch": 2983, "lr": 1.1794168843037301e-07}
{"train_loss": 0.0426035150885582, "global_step": 265525, "epoch": 2983, "lr": 1.179018959018452e-07}
{"train_loss": 0.05307915806770325, "global_step": 265526, "epoch": 2983, "lr": 1.1786211007936421e-07}
{"train_loss": 0.05186495929956436, "global_step": 265527, "epoch": 2983, "lr": 1.178223309629245e-07}
{"train_loss": 0.02315359190106392, "global_step": 265528, "epoch": 2983, "lr": 1.1778255855253717e-07}
{"train_loss": 0.04607892408967018, "global_step": 265529, "epoch": 2983, "lr": 1.1774279284820777e-07}
{"train_loss": 0.06145437806844711, "global_step": 265530, "epoch": 2983, "lr": 1.177030338499363e-07}
{"train_loss": 0.038315288722515106, "global_step": 265531, "epoch": 2983, "lr": 1.1766328155773387e-07}
{"train_loss": 0.04455988109111786, "global_step": 265532, "epoch": 2983, "lr": 1.1762353597160602e-07}
{"train_loss": 0.05846073105931282, "global_step": 265533, "epoch": 2983, "lr": 1.1758379709155275e-07}
{"train_loss": 0.04753265902400017, "global_step": 265534, "epoch": 2983, "lr": 1.1754406491757964e-07}
{"train_loss": 0.03992728516459465, "global_step": 265535, "epoch": 2983, "lr": 1.1750433944969774e-07}
{"train_loss": 0.029330208897590637, "global_step": 265536, "epoch": 2983, "lr": 1.1746462068790709e-07}
{"train_loss": 0.05980744957923889, "global_step": 265537, "epoch": 2983, "lr": 1.1742490863221878e-07}
{"train_loss": 0.0427250936627388, "global_step": 265538, "epoch": 2983, "lr": 1.173852032826328e-07}
{"train_loss": 0.049359194934368134, "global_step": 265539, "epoch": 2983, "lr": 1.1734550463916027e-07}
{"train_loss": 0.04600318521261215, "global_step": 265540, "epoch": 2983, "lr": 1.1730581270179563e-07}
{"train_loss": 0.04028725251555443, "global_step": 265541, "epoch": 2983, "lr": 1.1726612747055555e-07}
{"train_loss": 0.05116434022784233, "global_step": 265542, "epoch": 2983, "lr": 1.1722644894544e-07}
{"train_loss": 0.11345105618238449, "global_step": 265543, "epoch": 2983, "lr": 1.1718677712646009e-07}
{"train_loss": 0.019547272473573685, "global_step": 265544, "epoch": 2983, "lr": 1.1714711201361028e-07}
{"train_loss": 0.054280851036310196, "global_step": 265545, "epoch": 2983, "lr": 1.1710745360690723e-07}
{"train_loss": 0.03662737458944321, "global_step": 265546, "epoch": 2983, "lr": 1.1706780190634536e-07}
{"train_loss": 0.02725868858397007, "global_step": 265547, "epoch": 2983, "lr": 1.1702815691194135e-07}
{"train_loss": 0.05731113627552986, "global_step": 265548, "epoch": 2983, "lr": 1.1698851862368964e-07}
{"train_loss": 0.06965873390436172, "global_step": 265549, "epoch": 2983, "lr": 1.1694888704160135e-07}
{"train_loss": 0.08828974515199661, "global_step": 265550, "epoch": 2983, "lr": 1.1690926216568199e-07}
{"train_loss": 0.04481002688407898, "global_step": 265551, "epoch": 2983, "lr": 1.1686964399593713e-07}
{"train_loss": 0.0357653833925724, "global_step": 265552, "epoch": 2983, "lr": 1.1683003253237235e-07}
{"train_loss": 0.017894869670271873, "global_step": 265553, "epoch": 2983, "lr": 1.1679042777499316e-07}
{"train_loss": 0.028813019394874573, "global_step": 265554, "epoch": 2983, "lr": 1.1675082972379958e-07}
{"train_loss": 0.08559481799602509, "global_step": 265555, "epoch": 2983, "lr": 1.1671123837880271e-07}
{"train_loss": 0.04228268563747406, "global_step": 265556, "epoch": 2983, "lr": 1.166716537400081e-07}
{"train_loss": 0.052162084728479385, "global_step": 265557, "epoch": 2983, "lr": 1.1663207580741575e-07}
{"train_loss": 0.05230317637324333, "global_step": 265558, "epoch": 2983, "lr": 1.1659250458103121e-07}
{"train_loss": 0.03632044047117233, "global_step": 265559, "epoch": 2983, "lr": 1.1655294006087114e-07}
{"train_loss": 0.06641994416713715, "global_step": 265560, "epoch": 2983, "lr": 1.1651338224692443e-07}
{"train_loss": 0.036926187574863434, "global_step": 265561, "epoch": 2983, "lr": 1.1647383113920773e-07}
{"train_loss": 0.052336424589157104, "global_step": 265562, "epoch": 2983, "lr": 1.164342867377266e-07}
{"train_loss": 0.03772752359509468, "global_step": 265563, "epoch": 2983, "lr": 1.1639474904247549e-07}
{"train_loss": 0.09562867879867554, "global_step": 265564, "epoch": 2983, "lr": 1.163552180534766e-07}
{"train_loss": 0.028643475845456123, "global_step": 265565, "epoch": 2983, "lr": 1.1631569377071883e-07}
{"train_loss": 0.03550175204873085, "global_step": 265566, "epoch": 2983, "lr": 1.1627617619421326e-07}
{"train_loss": 0.042498502880334854, "global_step": 265567, "epoch": 2983, "lr": 1.1623666532397104e-07}
{"train_loss": 0.0462283119559288, "global_step": 265568, "epoch": 2983, "lr": 1.1619716115999213e-07}
{"train_loss": 0.05415816605091095, "global_step": 265569, "epoch": 2983, "lr": 1.1615766370228209e-07}
{"train_loss": 0.06256970763206482, "global_step": 265570, "epoch": 2983, "lr": 1.1611817295084648e-07}
{"train_loss": 0.038613468408584595, "global_step": 265571, "epoch": 2983, "lr": 1.1607868890568529e-07}
{"train_loss": 0.05583353340625763, "global_step": 265572, "epoch": 2983, "lr": 1.1603921156681519e-07}
{"train_loss": 0.05988050624728203, "global_step": 265573, "epoch": 2983, "lr": 1.1599974093423616e-07}
{"train_loss": 0.049992918968200684, "global_step": 265574, "epoch": 2983, "lr": 1.159602770079482e-07}
{"train_loss": 0.04494266048743484, "global_step": 265575, "epoch": 2983, "lr": 1.1592081978796798e-07, "val_loss": 9.602409362792969}
{"train_loss": 0.0422709621489048, "global_step": 265576, "epoch": 2984, "lr": 1.1588136927428995e-07}
{"train_loss": 0.038638610392808914, "global_step": 265577, "epoch": 2984, "lr": 1.1584192546692518e-07}
{"train_loss": 0.02407660335302353, "global_step": 265578, "epoch": 2984, "lr": 1.158024883658737e-07}
{"train_loss": 0.051024507731199265, "global_step": 265579, "epoch": 2984, "lr": 1.1576305797115216e-07}
{"train_loss": 0.04490092769265175, "global_step": 265580, "epoch": 2984, "lr": 1.1572363428274946e-07}
{"train_loss": 0.038996372371912, "global_step": 265581, "epoch": 2984, "lr": 1.1568421730068779e-07}
{"train_loss": 0.0813281461596489, "global_step": 265582, "epoch": 2984, "lr": 1.1564480702496161e-07}
{"train_loss": 0.05480644106864929, "global_step": 265583, "epoch": 2984, "lr": 1.1560540345557646e-07}
{"train_loss": 0.04110515117645264, "global_step": 265584, "epoch": 2984, "lr": 1.1556600659254346e-07}
{"train_loss": 0.033694833517074585, "global_step": 265585, "epoch": 2984, "lr": 1.155266164358626e-07}
{"train_loss": 0.035347580909729004, "global_step": 265586, "epoch": 2984, "lr": 1.1548723298553943e-07}
{"train_loss": 0.08536157757043839, "global_step": 265587, "epoch": 2984, "lr": 1.1544785624158506e-07}
{"train_loss": 0.07139982283115387, "global_step": 265588, "epoch": 2984, "lr": 1.1540848620399391e-07}
{"train_loss": 0.029416121542453766, "global_step": 265589, "epoch": 2984, "lr": 1.1536912287278823e-07}
{"train_loss": 0.07338733226060867, "global_step": 265590, "epoch": 2984, "lr": 1.1532976624795688e-07}
{"train_loss": 0.037484634667634964, "global_step": 265591, "epoch": 2984, "lr": 1.1529041632951099e-07}
{"train_loss": 0.03316784277558327, "global_step": 265592, "epoch": 2984, "lr": 1.1525107311745609e-07}
{"train_loss": 0.1300640106201172, "global_step": 265593, "epoch": 2984, "lr": 1.1521173661180329e-07}
{"train_loss": 0.04622018709778786, "global_step": 265594, "epoch": 2984, "lr": 1.1517240681254704e-07}
{"train_loss": 0.05492609739303589, "global_step": 265595, "epoch": 2984, "lr": 1.1513308371969844e-07}
{"train_loss": 0.05378999188542366, "global_step": 265596, "epoch": 2984, "lr": 1.1509376733326305e-07}
{"train_loss": 0.10727217048406601, "global_step": 265597, "epoch": 2984, "lr": 1.150544576532464e-07}
{"train_loss": 0.03768467158079147, "global_step": 265598, "epoch": 2984, "lr": 1.1501515467964852e-07}
{"train_loss": 0.08475228399038315, "global_step": 265599, "epoch": 2984, "lr": 1.1497585841248049e-07}
{"train_loss": 0.03271399810910225, "global_step": 265600, "epoch": 2984, "lr": 1.1493656885174786e-07}
{"train_loss": 0.0399351604282856, "global_step": 265601, "epoch": 2984, "lr": 1.1489728599745619e-07}
{"train_loss": 0.06150682643055916, "global_step": 265602, "epoch": 2984, "lr": 1.1485800984960548e-07}
{"train_loss": 0.03172340616583824, "global_step": 265603, "epoch": 2984, "lr": 1.1481874040820128e-07}
{"train_loss": 0.028587792068719864, "global_step": 265604, "epoch": 2984, "lr": 1.147794776732547e-07}
{"train_loss": 0.02948061190545559, "global_step": 265605, "epoch": 2984, "lr": 1.1474022164477127e-07}
{"train_loss": 0.014230087399482727, "global_step": 265606, "epoch": 2984, "lr": 1.1470097232274546e-07}
{"train_loss": 0.04145757108926773, "global_step": 265607, "epoch": 2984, "lr": 1.1466172970719392e-07}
{"train_loss": 0.05605883523821831, "global_step": 265608, "epoch": 2984, "lr": 1.1462249379812218e-07}
{"train_loss": 0.038007333874702454, "global_step": 265609, "epoch": 2984, "lr": 1.145832645955247e-07}
{"train_loss": 0.03357377275824547, "global_step": 265610, "epoch": 2984, "lr": 1.1454404209941815e-07}
{"train_loss": 0.04058166593313217, "global_step": 265611, "epoch": 2984, "lr": 1.1450482630980252e-07}
{"train_loss": 0.034745439887046814, "global_step": 265612, "epoch": 2984, "lr": 1.1446561722667781e-07}
{"train_loss": 0.04981236904859543, "global_step": 265613, "epoch": 2984, "lr": 1.1442641485006067e-07}
{"train_loss": 0.033311910927295685, "global_step": 265614, "epoch": 2984, "lr": 1.143872191799511e-07}
{"train_loss": 0.029904453083872795, "global_step": 265615, "epoch": 2984, "lr": 1.1434803021635465e-07}
{"train_loss": 0.05094224214553833, "global_step": 265616, "epoch": 2984, "lr": 1.1430884795927132e-07}
{"train_loss": 0.06558236479759216, "global_step": 265617, "epoch": 2984, "lr": 1.1426967240871778e-07}
{"train_loss": 0.060794033110141754, "global_step": 265618, "epoch": 2984, "lr": 1.1423050356468845e-07}
{"train_loss": 0.024901699274778366, "global_step": 265619, "epoch": 2984, "lr": 1.1419134142719445e-07}
{"train_loss": 0.039938993752002716, "global_step": 265620, "epoch": 2984, "lr": 1.1415218599623579e-07}
{"train_loss": 0.03650844469666481, "global_step": 265621, "epoch": 2984, "lr": 1.141130372718291e-07}
{"train_loss": 0.055536817759275436, "global_step": 265622, "epoch": 2984, "lr": 1.1407389525396328e-07}
{"train_loss": 0.053360722959041595, "global_step": 265623, "epoch": 2984, "lr": 1.1403475994266055e-07}
{"train_loss": 0.023732515051960945, "global_step": 265624, "epoch": 2984, "lr": 1.139956313379098e-07}
{"train_loss": 0.08253448456525803, "global_step": 265625, "epoch": 2984, "lr": 1.1395650943972769e-07}
{"train_loss": 0.03042704612016678, "global_step": 265626, "epoch": 2984, "lr": 1.139173942481142e-07}
{"train_loss": 0.03857072442770004, "global_step": 265627, "epoch": 2984, "lr": 1.1387828576308047e-07}
{"train_loss": 0.06451305001974106, "global_step": 265628, "epoch": 2984, "lr": 1.1383918398462645e-07}
{"train_loss": 0.0660383552312851, "global_step": 265629, "epoch": 2984, "lr": 1.1380008891275773e-07}
{"train_loss": 0.03798149526119232, "global_step": 265630, "epoch": 2984, "lr": 1.1376100054747985e-07}
{"train_loss": 0.03326490893959999, "global_step": 265631, "epoch": 2984, "lr": 1.1372191888879835e-07}
{"train_loss": 0.036598436534404755, "global_step": 265632, "epoch": 2984, "lr": 1.1368284393672435e-07}
{"train_loss": 0.03478991240262985, "global_step": 265633, "epoch": 2984, "lr": 1.1364377569125229e-07}
{"train_loss": 0.027771418914198875, "global_step": 265634, "epoch": 2984, "lr": 1.1360471415239327e-07}
{"train_loss": 0.01883360557258129, "global_step": 265635, "epoch": 2984, "lr": 1.1356565932015284e-07}
{"train_loss": 0.023845912888646126, "global_step": 265636, "epoch": 2984, "lr": 1.1352661119453656e-07}
{"train_loss": 0.031747665256261826, "global_step": 265637, "epoch": 2984, "lr": 1.1348756977554997e-07}
{"train_loss": 0.04494452849030495, "global_step": 265638, "epoch": 2984, "lr": 1.1344853506319308e-07}
{"train_loss": 0.04076285660266876, "global_step": 265639, "epoch": 2984, "lr": 1.1340950705747699e-07}
{"train_loss": 0.010531810112297535, "global_step": 265640, "epoch": 2984, "lr": 1.133704857584017e-07}
{"train_loss": 0.08690020442008972, "global_step": 265641, "epoch": 2984, "lr": 1.1333147116597832e-07}
{"train_loss": 0.04870511218905449, "global_step": 265642, "epoch": 2984, "lr": 1.1329246328021237e-07}
{"train_loss": 0.06081925705075264, "global_step": 265643, "epoch": 2984, "lr": 1.1325346210110387e-07}
{"train_loss": 0.021552015095949173, "global_step": 265644, "epoch": 2984, "lr": 1.1321446762865839e-07}
{"train_loss": 0.050053104758262634, "global_step": 265645, "epoch": 2984, "lr": 1.1317547986288702e-07}
{"train_loss": 0.03486088663339615, "global_step": 265646, "epoch": 2984, "lr": 1.1313649880378418e-07}
{"train_loss": 0.02486756443977356, "global_step": 265647, "epoch": 2984, "lr": 1.1309752445137211e-07}
{"train_loss": 0.05520760267972946, "global_step": 265648, "epoch": 2984, "lr": 1.130585568056397e-07}
{"train_loss": 0.02195596694946289, "global_step": 265649, "epoch": 2984, "lr": 1.1301959586659805e-07}
{"train_loss": 0.0598374679684639, "global_step": 265650, "epoch": 2984, "lr": 1.129806416342527e-07}
{"train_loss": 0.03663470596075058, "global_step": 265651, "epoch": 2984, "lr": 1.1294169410861477e-07}
{"train_loss": 0.05593927577137947, "global_step": 265652, "epoch": 2984, "lr": 1.129027532896787e-07}
{"train_loss": 0.04773861542344093, "global_step": 265653, "epoch": 2984, "lr": 1.1286381917745559e-07}
{"train_loss": 0.030575038865208626, "global_step": 265654, "epoch": 2984, "lr": 1.1282489177195099e-07}
{"train_loss": 0.038395702838897705, "global_step": 265655, "epoch": 2984, "lr": 1.1278597107317045e-07}
{"train_loss": 0.034365251660346985, "global_step": 265656, "epoch": 2984, "lr": 1.12747057081114e-07}
{"train_loss": 0.045401431620121, "global_step": 265657, "epoch": 2984, "lr": 1.127081497957927e-07}
{"train_loss": 0.03832591697573662, "global_step": 265658, "epoch": 2984, "lr": 1.1266924921720657e-07}
{"train_loss": 0.03417038917541504, "global_step": 265659, "epoch": 2984, "lr": 1.126303553453667e-07}
{"train_loss": 0.03180289268493652, "global_step": 265660, "epoch": 2984, "lr": 1.1259146818027866e-07}
{"train_loss": 0.03211130574345589, "global_step": 265661, "epoch": 2984, "lr": 1.1255258772194243e-07}
{"train_loss": 0.05843903496861458, "global_step": 265662, "epoch": 2984, "lr": 1.1251371397036359e-07}
{"train_loss": 0.06901601701974869, "global_step": 265663, "epoch": 2984, "lr": 1.1247484692554766e-07}
{"train_loss": 0.04505410029688913, "global_step": 265664, "epoch": 2984, "lr": 1.1243598658750576e-07, "val_loss": 9.631613731384277}
{"train_loss": 0.05330285057425499, "global_step": 265665, "epoch": 2985, "lr": 1.1239713295623789e-07}
{"train_loss": 0.02555582858622074, "global_step": 265666, "epoch": 2985, "lr": 1.1235828603174959e-07}
{"train_loss": 0.062285371124744415, "global_step": 265667, "epoch": 2985, "lr": 1.1231944581404641e-07}
{"train_loss": 0.09345279633998871, "global_step": 265668, "epoch": 2985, "lr": 1.1228061230313391e-07}
{"train_loss": 0.011727383360266685, "global_step": 265669, "epoch": 2985, "lr": 1.1224178549901765e-07}
{"train_loss": 0.031412359327077866, "global_step": 265670, "epoch": 2985, "lr": 1.1220296540170316e-07}
{"train_loss": 0.08419963717460632, "global_step": 265671, "epoch": 2985, "lr": 1.1216415201119601e-07}
{"train_loss": 0.012892443686723709, "global_step": 265672, "epoch": 2985, "lr": 1.1212534532749618e-07}
{"train_loss": 0.01798100955784321, "global_step": 265673, "epoch": 2985, "lr": 1.120865453506148e-07}
{"train_loss": 0.03091561608016491, "global_step": 265674, "epoch": 2985, "lr": 1.120477520805574e-07}
{"train_loss": 0.04777849465608597, "global_step": 265675, "epoch": 2985, "lr": 1.1200896551732398e-07}
{"train_loss": 0.03540775179862976, "global_step": 265676, "epoch": 2985, "lr": 1.1197018566092565e-07}
{"train_loss": 0.030785387381911278, "global_step": 265677, "epoch": 2985, "lr": 1.1193141251136242e-07}
{"train_loss": 0.02766266278922558, "global_step": 265678, "epoch": 2985, "lr": 1.1189264606864536e-07}
{"train_loss": 0.035028062760829926, "global_step": 265679, "epoch": 2985, "lr": 1.118538863327745e-07}
{"train_loss": 0.0451829619705677, "global_step": 265680, "epoch": 2985, "lr": 1.1181513330375538e-07}
{"train_loss": 0.05407118424773216, "global_step": 265681, "epoch": 2985, "lr": 1.1177638698159355e-07}
{"train_loss": 0.025241795927286148, "global_step": 265682, "epoch": 2985, "lr": 1.1173764736630011e-07}
{"train_loss": 0.06825323402881622, "global_step": 265683, "epoch": 2985, "lr": 1.116989144578695e-07}
{"train_loss": 0.12652993202209473, "global_step": 265684, "epoch": 2985, "lr": 1.116601882563184e-07}
{"train_loss": 0.04328696429729462, "global_step": 265685, "epoch": 2985, "lr": 1.1162146876164126e-07}
{"train_loss": 0.06323572248220444, "global_step": 265686, "epoch": 2985, "lr": 1.115827559738547e-07}
{"train_loss": 0.05765362083911896, "global_step": 265687, "epoch": 2985, "lr": 1.115440498929532e-07}
{"train_loss": 0.0545029453933239, "global_step": 265688, "epoch": 2985, "lr": 1.1150535051894784e-07}
{"train_loss": 0.04431683197617531, "global_step": 265689, "epoch": 2985, "lr": 1.1146665785184418e-07}
{"train_loss": 0.026292948052287102, "global_step": 265690, "epoch": 2985, "lr": 1.1142797189164223e-07}
{"train_loss": 0.015942636877298355, "global_step": 265691, "epoch": 2985, "lr": 1.1138929263835307e-07}
{"train_loss": 0.05499666929244995, "global_step": 265692, "epoch": 2985, "lr": 1.1135062009198227e-07}
{"train_loss": 0.07327254861593246, "global_step": 265693, "epoch": 2985, "lr": 1.1131195425252983e-07}
{"train_loss": 0.025678960606455803, "global_step": 265694, "epoch": 2985, "lr": 1.1127329512000128e-07}
{"train_loss": 0.0417356938123703, "global_step": 265695, "epoch": 2985, "lr": 1.1123464269440775e-07}
{"train_loss": 0.0320795401930809, "global_step": 265696, "epoch": 2985, "lr": 1.1119599697574922e-07}
{"train_loss": 0.033439818769693375, "global_step": 265697, "epoch": 2985, "lr": 1.1115735796403126e-07}
{"train_loss": 0.07917316257953644, "global_step": 265698, "epoch": 2985, "lr": 1.111187256592594e-07}
{"train_loss": 0.024419136345386505, "global_step": 265699, "epoch": 2985, "lr": 1.1108010006143921e-07}
{"train_loss": 0.06287591904401779, "global_step": 265700, "epoch": 2985, "lr": 1.1104148117058177e-07}
{"train_loss": 0.015486304648220539, "global_step": 265701, "epoch": 2985, "lr": 1.1100286898668155e-07}
{"train_loss": 0.09895216673612595, "global_step": 265702, "epoch": 2985, "lr": 1.1096426350974964e-07}
{"train_loss": 0.016269203275442123, "global_step": 265703, "epoch": 2985, "lr": 1.109256647397916e-07}
{"train_loss": 0.04001786932349205, "global_step": 265704, "epoch": 2985, "lr": 1.1088707267680742e-07}
{"train_loss": 0.07478339225053787, "global_step": 265705, "epoch": 2985, "lr": 1.1084848732081376e-07}
{"train_loss": 0.03720301017165184, "global_step": 265706, "epoch": 2985, "lr": 1.1080990867180508e-07}
{"train_loss": 0.05225260928273201, "global_step": 265707, "epoch": 2985, "lr": 1.107713367297869e-07}
{"train_loss": 0.023444801568984985, "global_step": 265708, "epoch": 2985, "lr": 1.1073277149477035e-07}
{"train_loss": 0.05593214929103851, "global_step": 265709, "epoch": 2985, "lr": 1.1069421296676097e-07}
{"train_loss": 0.03720114380121231, "global_step": 265710, "epoch": 2985, "lr": 1.106556611457532e-07}
{"train_loss": 0.024670442566275597, "global_step": 265711, "epoch": 2985, "lr": 1.1061711603176372e-07}
{"train_loss": 0.04113500565290451, "global_step": 265712, "epoch": 2985, "lr": 1.1057857762479807e-07}
{"train_loss": 0.03915880247950554, "global_step": 265713, "epoch": 2985, "lr": 1.1054004592485067e-07}
{"train_loss": 0.06740397214889526, "global_step": 265714, "epoch": 2985, "lr": 1.1050152093193822e-07}
{"train_loss": 0.06113549694418907, "global_step": 265715, "epoch": 2985, "lr": 1.1046300264605513e-07}
{"train_loss": 0.01875164732336998, "global_step": 265716, "epoch": 2985, "lr": 1.1042449106721253e-07}
{"train_loss": 0.048184461891651154, "global_step": 265717, "epoch": 2985, "lr": 1.1038598619542151e-07}
{"train_loss": 0.07652407884597778, "global_step": 265718, "epoch": 2985, "lr": 1.1034748803067651e-07}
{"train_loss": 0.0280123520642519, "global_step": 265719, "epoch": 2985, "lr": 1.1030899657298866e-07}
{"train_loss": 0.03747102618217468, "global_step": 265720, "epoch": 2985, "lr": 1.1027051182235792e-07}
{"train_loss": 0.07975825667381287, "global_step": 265721, "epoch": 2985, "lr": 1.1023203377879543e-07}
{"train_loss": 0.054060399532318115, "global_step": 265722, "epoch": 2985, "lr": 1.1019356244230672e-07}
{"train_loss": 0.021895280107855797, "global_step": 265723, "epoch": 2985, "lr": 1.1015509781289179e-07}
{"train_loss": 0.043658457696437836, "global_step": 265724, "epoch": 2985, "lr": 1.1011663989055621e-07}
{"train_loss": 0.03143502399325371, "global_step": 265725, "epoch": 2985, "lr": 1.1007818867531105e-07}
{"train_loss": 0.0383063368499279, "global_step": 265726, "epoch": 2985, "lr": 1.1003974416715635e-07}
{"train_loss": 0.025873709470033646, "global_step": 265727, "epoch": 2985, "lr": 1.1000130636610317e-07}
{"train_loss": 0.08589352667331696, "global_step": 265728, "epoch": 2985, "lr": 1.09962875272146e-07}
{"train_loss": 0.050692118704319, "global_step": 265729, "epoch": 2985, "lr": 1.0992445088530145e-07}
{"train_loss": 0.023422202095389366, "global_step": 265730, "epoch": 2985, "lr": 1.0988603320556401e-07}
{"train_loss": 0.014941579662263393, "global_step": 265731, "epoch": 2985, "lr": 1.0984762223295031e-07}
{"train_loss": 0.029506534337997437, "global_step": 265732, "epoch": 2985, "lr": 1.0980921796745481e-07}
{"train_loss": 0.08126511424779892, "global_step": 265733, "epoch": 2985, "lr": 1.097708204090886e-07}
{"train_loss": 0.042912740260362625, "global_step": 265734, "epoch": 2985, "lr": 1.0973242955785724e-07}
{"train_loss": 0.026487654075026512, "global_step": 265735, "epoch": 2985, "lr": 1.0969404541376071e-07}
{"train_loss": 0.05046394094824791, "global_step": 265736, "epoch": 2985, "lr": 1.0965566797681016e-07}
{"train_loss": 0.0722498968243599, "global_step": 265737, "epoch": 2985, "lr": 1.0961729724701109e-07}
{"train_loss": 0.029870782047510147, "global_step": 265738, "epoch": 2985, "lr": 1.0957893322436352e-07}
{"train_loss": 0.03131420165300369, "global_step": 265739, "epoch": 2985, "lr": 1.09540575908873e-07}
{"train_loss": 0.058385979384183884, "global_step": 265740, "epoch": 2985, "lr": 1.095022253005451e-07}
{"train_loss": 0.07591617107391357, "global_step": 265741, "epoch": 2985, "lr": 1.094638813993909e-07}
{"train_loss": 0.024105938151478767, "global_step": 265742, "epoch": 2985, "lr": 1.094255442054104e-07}
{"train_loss": 0.05368777737021446, "global_step": 265743, "epoch": 2985, "lr": 1.0938721371860916e-07}
{"train_loss": 0.013596451841294765, "global_step": 265744, "epoch": 2985, "lr": 1.0934888993899272e-07}
{"train_loss": 0.025690890848636627, "global_step": 265745, "epoch": 2985, "lr": 1.0931057286656665e-07}
{"train_loss": 0.0660315677523613, "global_step": 265746, "epoch": 2985, "lr": 1.0927226250133649e-07}
{"train_loss": 0.050030697137117386, "global_step": 265747, "epoch": 2985, "lr": 1.0923395884330223e-07}
{"train_loss": 0.040604203939437866, "global_step": 265748, "epoch": 2985, "lr": 1.0919566189248053e-07}
{"train_loss": 0.014241342432796955, "global_step": 265749, "epoch": 2985, "lr": 1.0915737164886586e-07}
{"train_loss": 0.025246988981962204, "global_step": 265750, "epoch": 2985, "lr": 1.0911908811246374e-07}
{"train_loss": 0.04905053600668907, "global_step": 265751, "epoch": 2985, "lr": 1.0908081128328528e-07}
{"train_loss": 0.011918768286705017, "global_step": 265752, "epoch": 2985, "lr": 1.0904254116133605e-07}
{"train_loss": 0.044031549119547514, "global_step": 265753, "epoch": 2985, "lr": 1.0900427774661048e-07, "val_loss": 9.672501564025879, "train_action_mse_error": 9.694866180419922}
{"train_loss": 0.03125789389014244, "global_step": 265754, "epoch": 2986, "lr": 1.0896602103912523e-07}
{"train_loss": 0.040122684091329575, "global_step": 265755, "epoch": 2986, "lr": 1.089277710388803e-07}
{"train_loss": 0.10385162383317947, "global_step": 265756, "epoch": 2986, "lr": 1.0888952774588679e-07}
{"train_loss": 0.01690858043730259, "global_step": 265757, "epoch": 2986, "lr": 1.0885129116013915e-07}
{"train_loss": 0.027246801182627678, "global_step": 265758, "epoch": 2986, "lr": 1.0881306128164848e-07}
{"train_loss": 0.027505047619342804, "global_step": 265759, "epoch": 2986, "lr": 1.0877483811042589e-07}
{"train_loss": 0.046586088836193085, "global_step": 265760, "epoch": 2986, "lr": 1.0873662164646581e-07}
{"train_loss": 0.06317880749702454, "global_step": 265761, "epoch": 2986, "lr": 1.0869841188977936e-07}
{"train_loss": 0.038963738828897476, "global_step": 265762, "epoch": 2986, "lr": 1.0866020884036654e-07}
{"train_loss": 0.044258683919906616, "global_step": 265763, "epoch": 2986, "lr": 1.0862201249823845e-07}
{"train_loss": 0.0766775980591774, "global_step": 265764, "epoch": 2986, "lr": 1.0858382286340063e-07}
{"train_loss": 0.022133661434054375, "global_step": 265765, "epoch": 2986, "lr": 1.085456399358531e-07}
{"train_loss": 0.04784683883190155, "global_step": 265766, "epoch": 2986, "lr": 1.0850746371560694e-07}
{"train_loss": 0.04942655563354492, "global_step": 265767, "epoch": 2986, "lr": 1.0846929420265661e-07}
{"train_loss": 0.03178025037050247, "global_step": 265768, "epoch": 2986, "lr": 1.0843113139701877e-07}
{"train_loss": 0.0532904677093029, "global_step": 265769, "epoch": 2986, "lr": 1.0839297529869897e-07}
{"train_loss": 0.08413171768188477, "global_step": 265770, "epoch": 2986, "lr": 1.0835482590769164e-07}
{"train_loss": 0.028360048308968544, "global_step": 265771, "epoch": 2986, "lr": 1.083166832240079e-07}
{"train_loss": 0.04643009975552559, "global_step": 265772, "epoch": 2986, "lr": 1.082785472476533e-07}
{"train_loss": 0.09316365420818329, "global_step": 265773, "epoch": 2986, "lr": 1.0824041797863338e-07}
{"train_loss": 0.06658954918384552, "global_step": 265774, "epoch": 2986, "lr": 1.082022954169537e-07}
{"train_loss": 0.026471208781003952, "global_step": 265775, "epoch": 2986, "lr": 1.0816417956261981e-07}
{"train_loss": 0.053762659430503845, "global_step": 265776, "epoch": 2986, "lr": 1.0812607041563172e-07}
{"train_loss": 0.04992000758647919, "global_step": 265777, "epoch": 2986, "lr": 1.0808796797599497e-07}
{"train_loss": 0.012124057859182358, "global_step": 265778, "epoch": 2986, "lr": 1.080498722437262e-07}
{"train_loss": 0.0631835088133812, "global_step": 265779, "epoch": 2986, "lr": 1.0801178321881433e-07}
{"train_loss": 0.02455875836312771, "global_step": 265780, "epoch": 2986, "lr": 1.07973700901276e-07}
{"train_loss": 0.07730241119861603, "global_step": 265781, "epoch": 2986, "lr": 1.0793562529111123e-07}
{"train_loss": 0.05163119360804558, "global_step": 265782, "epoch": 2986, "lr": 1.078975563883311e-07}
{"train_loss": 0.01088760793209076, "global_step": 265783, "epoch": 2986, "lr": 1.0785949419293006e-07}
{"train_loss": 0.02869287133216858, "global_step": 265784, "epoch": 2986, "lr": 1.0782143870491923e-07}
{"train_loss": 0.054493267089128494, "global_step": 265785, "epoch": 2986, "lr": 1.077833899243097e-07}
{"train_loss": 0.046461038291454315, "global_step": 265786, "epoch": 2986, "lr": 1.0774534785109036e-07}
{"train_loss": 0.05715125799179077, "global_step": 265787, "epoch": 2986, "lr": 1.0770731248528343e-07}
{"train_loss": 0.03373595327138901, "global_step": 265788, "epoch": 2986, "lr": 1.076692838268889e-07}
{"train_loss": 0.0514540821313858, "global_step": 265789, "epoch": 2986, "lr": 1.0763126187590678e-07}
{"train_loss": 0.04214484244585037, "global_step": 265790, "epoch": 2986, "lr": 1.0759324663234815e-07}
{"train_loss": 0.07385373115539551, "global_step": 265791, "epoch": 2986, "lr": 1.0755523809621305e-07}
{"train_loss": 0.06462310999631882, "global_step": 265792, "epoch": 2986, "lr": 1.0751723626750699e-07}
{"train_loss": 0.05002233386039734, "global_step": 265793, "epoch": 2986, "lr": 1.0747924114624108e-07}
{"train_loss": 0.03916516155004501, "global_step": 265794, "epoch": 2986, "lr": 1.0744125273241534e-07}
{"train_loss": 0.06344344466924667, "global_step": 265795, "epoch": 2986, "lr": 1.0740327102603531e-07}
{"train_loss": 0.032883983105421066, "global_step": 265796, "epoch": 2986, "lr": 1.0736529602710655e-07}
{"train_loss": 0.04295416548848152, "global_step": 265797, "epoch": 2986, "lr": 1.0732732773563459e-07}
{"train_loss": 0.035354577004909515, "global_step": 265798, "epoch": 2986, "lr": 1.0728936615162499e-07}
{"train_loss": 0.03904323652386665, "global_step": 265799, "epoch": 2986, "lr": 1.0725141127508331e-07}
{"train_loss": 0.014898889698088169, "global_step": 265800, "epoch": 2986, "lr": 1.072134631060151e-07}
{"train_loss": 0.042660463601350784, "global_step": 265801, "epoch": 2986, "lr": 1.0717552164442035e-07}
{"train_loss": 0.035141199827194214, "global_step": 265802, "epoch": 2986, "lr": 1.0713758689031017e-07}
{"train_loss": 0.04113220050930977, "global_step": 265803, "epoch": 2986, "lr": 1.0709965884368456e-07}
{"train_loss": 0.026471532881259918, "global_step": 265804, "epoch": 2986, "lr": 1.0706173750454907e-07}
{"train_loss": 0.03580141440033913, "global_step": 265805, "epoch": 2986, "lr": 1.0702382287291479e-07}
{"train_loss": 0.044823057949543, "global_step": 265806, "epoch": 2986, "lr": 1.0698591494878729e-07}
{"train_loss": 0.07659772038459778, "global_step": 265807, "epoch": 2986, "lr": 1.06948013732161e-07}
{"train_loss": 0.05130985751748085, "global_step": 265808, "epoch": 2986, "lr": 1.0691011922304706e-07}
{"train_loss": 0.06461726129055023, "global_step": 265809, "epoch": 2986, "lr": 1.0687223142145653e-07}
{"train_loss": 0.0329732745885849, "global_step": 265810, "epoch": 2986, "lr": 1.0683435032738387e-07}
{"train_loss": 0.03176668658852577, "global_step": 265811, "epoch": 2986, "lr": 1.0679647594084019e-07}
{"train_loss": 0.016788655892014503, "global_step": 265812, "epoch": 2986, "lr": 1.0675860826183104e-07}
{"train_loss": 0.038777150213718414, "global_step": 265813, "epoch": 2986, "lr": 1.0672074729035641e-07}
{"train_loss": 0.04217907786369324, "global_step": 265814, "epoch": 2986, "lr": 1.0668289302643297e-07}
{"train_loss": 0.052848733961582184, "global_step": 265815, "epoch": 2986, "lr": 1.0664504547004961e-07}
{"train_loss": 0.07833682000637054, "global_step": 265816, "epoch": 2986, "lr": 1.0660720462122297e-07}
{"train_loss": 0.026916468515992165, "global_step": 265817, "epoch": 2986, "lr": 1.0656937047995307e-07}
{"train_loss": 0.06886643171310425, "global_step": 265818, "epoch": 2986, "lr": 1.0653154304625102e-07}
{"train_loss": 0.06867343932390213, "global_step": 265819, "epoch": 2986, "lr": 1.0649372232011679e-07}
{"train_loss": 0.02125837653875351, "global_step": 265820, "epoch": 2986, "lr": 1.0645590830155594e-07}
{"train_loss": 0.023519856855273247, "global_step": 265821, "epoch": 2986, "lr": 1.0641810099057403e-07}
{"train_loss": 0.055175598710775375, "global_step": 265822, "epoch": 2986, "lr": 1.0638030038717106e-07}
{"train_loss": 0.08128020912408829, "global_step": 265823, "epoch": 2986, "lr": 1.0634250649136368e-07}
{"train_loss": 0.025442566722631454, "global_step": 265824, "epoch": 2986, "lr": 1.0630471930314634e-07}
{"train_loss": 0.03312254324555397, "global_step": 265825, "epoch": 2986, "lr": 1.0626693882253013e-07}
{"train_loss": 0.038566574454307556, "global_step": 265826, "epoch": 2986, "lr": 1.0622916504951508e-07}
{"train_loss": 0.0763867124915123, "global_step": 265827, "epoch": 2986, "lr": 1.0619139798411226e-07}
{"train_loss": 0.024083932861685753, "global_step": 265828, "epoch": 2986, "lr": 1.0615363762632724e-07}
{"train_loss": 0.029956351965665817, "global_step": 265829, "epoch": 2986, "lr": 1.0611588397616002e-07}
{"train_loss": 0.11396355926990509, "global_step": 265830, "epoch": 2986, "lr": 1.0607813703361058e-07}
{"train_loss": 0.0907287448644638, "global_step": 265831, "epoch": 2986, "lr": 1.0604039679870115e-07}
{"train_loss": 0.046541787683963776, "global_step": 265832, "epoch": 2986, "lr": 1.0600266327141505e-07}
{"train_loss": 0.006445364095270634, "global_step": 265833, "epoch": 2986, "lr": 1.0596493645178007e-07}
{"train_loss": 0.089901864528656, "global_step": 265834, "epoch": 2986, "lr": 1.0592721633977953e-07}
{"train_loss": 0.05609777569770813, "global_step": 265835, "epoch": 2986, "lr": 1.0588950293543564e-07}
{"train_loss": 0.07253650575876236, "global_step": 265836, "epoch": 2986, "lr": 1.0585179623874841e-07}
{"train_loss": 0.05338023230433464, "global_step": 265837, "epoch": 2986, "lr": 1.0581409624971783e-07}
{"train_loss": 0.02020285464823246, "global_step": 265838, "epoch": 2986, "lr": 1.0577640296834945e-07}
{"train_loss": 0.0336446687579155, "global_step": 265839, "epoch": 2986, "lr": 1.0573871639465993e-07}
{"train_loss": 0.021239185705780983, "global_step": 265840, "epoch": 2986, "lr": 1.0570103652863816e-07}
{"train_loss": 0.05996976047754288, "global_step": 265841, "epoch": 2986, "lr": 1.056633633703008e-07}
{"train_loss": 0.04709829804435205, "global_step": 265842, "epoch": 2986, "lr": 1.056256969196423e-07, "val_loss": 9.660061836242676}
{"train_loss": 0.021301371976733208, "global_step": 265843, "epoch": 2987, "lr": 1.0558803717667931e-07}
{"train_loss": 0.059311170130968094, "global_step": 265844, "epoch": 2987, "lr": 1.0555038414141183e-07}
{"train_loss": 0.06584976613521576, "global_step": 265845, "epoch": 2987, "lr": 1.0551273781384541e-07}
{"train_loss": 0.04525630176067352, "global_step": 265846, "epoch": 2987, "lr": 1.054750981939856e-07}
{"train_loss": 0.03372727334499359, "global_step": 265847, "epoch": 2987, "lr": 1.054374652818324e-07}
{"train_loss": 0.03971529379487038, "global_step": 265848, "epoch": 2987, "lr": 1.0539983907740247e-07}
{"train_loss": 0.02257811836898327, "global_step": 265849, "epoch": 2987, "lr": 1.0536221958068471e-07}
{"train_loss": 0.03758808970451355, "global_step": 265850, "epoch": 2987, "lr": 1.0532460679169575e-07}
{"train_loss": 0.06849023699760437, "global_step": 265851, "epoch": 2987, "lr": 1.0528700071044118e-07}
{"train_loss": 0.03999634459614754, "global_step": 265852, "epoch": 2987, "lr": 1.0524940133692096e-07}
{"train_loss": 0.025106480345129967, "global_step": 265853, "epoch": 2987, "lr": 1.0521180867114066e-07}
{"train_loss": 0.03321933373808861, "global_step": 265854, "epoch": 2987, "lr": 1.0517422271310584e-07}
{"train_loss": 0.07301438599824905, "global_step": 265855, "epoch": 2987, "lr": 1.0513664346282204e-07}
{"train_loss": 0.03887610882520676, "global_step": 265856, "epoch": 2987, "lr": 1.0509907092030035e-07}
{"train_loss": 0.032387979328632355, "global_step": 265857, "epoch": 2987, "lr": 1.050615050855297e-07}
{"train_loss": 0.05309517681598663, "global_step": 265858, "epoch": 2987, "lr": 1.0502394595853226e-07}
{"train_loss": 0.04827060177922249, "global_step": 265859, "epoch": 2987, "lr": 1.0498639353930806e-07}
{"train_loss": 0.04312821477651596, "global_step": 265860, "epoch": 2987, "lr": 1.0494884782785708e-07}
{"train_loss": 0.034511055797338486, "global_step": 265861, "epoch": 2987, "lr": 1.0491130882419043e-07}
{"train_loss": 0.04238149896264076, "global_step": 265862, "epoch": 2987, "lr": 1.0487377652830255e-07}
{"train_loss": 0.04903430491685867, "global_step": 265863, "epoch": 2987, "lr": 1.0483625094021566e-07}
{"train_loss": 0.034872785210609436, "global_step": 265864, "epoch": 2987, "lr": 1.0479873205991864e-07}
{"train_loss": 0.07152784615755081, "global_step": 265865, "epoch": 2987, "lr": 1.0476121988742815e-07}
{"train_loss": 0.0726587176322937, "global_step": 265866, "epoch": 2987, "lr": 1.0472371442273865e-07}
{"train_loss": 0.070529043674469, "global_step": 265867, "epoch": 2987, "lr": 1.0468621566586679e-07}
{"train_loss": 0.042939841747283936, "global_step": 265868, "epoch": 2987, "lr": 1.04648723616807e-07}
{"train_loss": 0.08893806487321854, "global_step": 265869, "epoch": 2987, "lr": 1.046112382755704e-07}
{"train_loss": 0.051322899758815765, "global_step": 265870, "epoch": 2987, "lr": 1.0457375964216254e-07}
{"train_loss": 0.058226242661476135, "global_step": 265871, "epoch": 2987, "lr": 1.0453628771658896e-07}
{"train_loss": 0.04030134901404381, "global_step": 265872, "epoch": 2987, "lr": 1.0449882249884412e-07}
{"train_loss": 0.07608038187026978, "global_step": 265873, "epoch": 2987, "lr": 1.0446136398895024e-07}
{"train_loss": 0.06138394773006439, "global_step": 265874, "epoch": 2987, "lr": 1.0442391218689617e-07}
{"train_loss": 0.026323895901441574, "global_step": 265875, "epoch": 2987, "lr": 1.0438646709269862e-07}
{"train_loss": 0.08652269095182419, "global_step": 265876, "epoch": 2987, "lr": 1.0434902870635754e-07}
{"train_loss": 0.038604337722063065, "global_step": 265877, "epoch": 2987, "lr": 1.0431159702787851e-07}
{"train_loss": 0.02290135808289051, "global_step": 265878, "epoch": 2987, "lr": 1.0427417205726153e-07}
{"train_loss": 0.06792833656072617, "global_step": 265879, "epoch": 2987, "lr": 1.0423675379452324e-07}
{"train_loss": 0.03367577865719795, "global_step": 265880, "epoch": 2987, "lr": 1.041993422396581e-07}
{"train_loss": 0.014384270645678043, "global_step": 265881, "epoch": 2987, "lr": 1.0416193739267721e-07}
{"train_loss": 0.04502459615468979, "global_step": 265882, "epoch": 2987, "lr": 1.0412453925358612e-07}
{"train_loss": 0.04144476354122162, "global_step": 265883, "epoch": 2987, "lr": 1.0408714782238482e-07}
{"train_loss": 0.04144834727048874, "global_step": 265884, "epoch": 2987, "lr": 1.0404976309907888e-07}
{"train_loss": 0.012499715201556683, "global_step": 265885, "epoch": 2987, "lr": 1.0401238508367939e-07}
{"train_loss": 0.018824832513928413, "global_step": 265886, "epoch": 2987, "lr": 1.0397501377618634e-07}
{"train_loss": 0.05274175852537155, "global_step": 265887, "epoch": 2987, "lr": 1.0393764917659976e-07}
{"train_loss": 0.030772916972637177, "global_step": 265888, "epoch": 2987, "lr": 1.0390029128493628e-07}
{"train_loss": 0.023027600720524788, "global_step": 265889, "epoch": 2987, "lr": 1.038629401011959e-07}
{"train_loss": 0.09471560269594193, "global_step": 265890, "epoch": 2987, "lr": 1.0382559562538419e-07}
{"train_loss": 0.01086333952844143, "global_step": 265891, "epoch": 2987, "lr": 1.0378825785750113e-07}
{"train_loss": 0.035661403089761734, "global_step": 265892, "epoch": 2987, "lr": 1.0375092679755783e-07}
{"train_loss": 0.017846977338194847, "global_step": 265893, "epoch": 2987, "lr": 1.0371360244555428e-07}
{"train_loss": 0.062450703233480453, "global_step": 265894, "epoch": 2987, "lr": 1.036762848015016e-07}
{"train_loss": 0.060004957020282745, "global_step": 265895, "epoch": 2987, "lr": 1.0363897386539979e-07}
{"train_loss": 0.08071320503950119, "global_step": 265896, "epoch": 2987, "lr": 1.0360166963725438e-07}
{"train_loss": 0.061857856810092926, "global_step": 265897, "epoch": 2987, "lr": 1.0356437211707092e-07}
{"train_loss": 0.05119994282722473, "global_step": 265898, "epoch": 2987, "lr": 1.0352708130486055e-07}
{"train_loss": 0.03522101789712906, "global_step": 265899, "epoch": 2987, "lr": 1.0348979720061769e-07}
{"train_loss": 0.0695500448346138, "global_step": 265900, "epoch": 2987, "lr": 1.0345251980435344e-07}
{"train_loss": 0.019476529210805893, "global_step": 265901, "epoch": 2987, "lr": 1.0341524911607336e-07}
{"train_loss": 0.02898849919438362, "global_step": 265902, "epoch": 2987, "lr": 1.0337798513577746e-07}
{"train_loss": 0.026397492736577988, "global_step": 265903, "epoch": 2987, "lr": 1.0334072786347682e-07}
{"train_loss": 0.033733684569597244, "global_step": 265904, "epoch": 2987, "lr": 1.0330347729917699e-07}
{"train_loss": 0.06126639246940613, "global_step": 265905, "epoch": 2987, "lr": 1.03266233442878e-07}
{"train_loss": 0.02754521742463112, "global_step": 265906, "epoch": 2987, "lr": 1.0322899629458538e-07}
{"train_loss": 0.0658964216709137, "global_step": 265907, "epoch": 2987, "lr": 1.0319176585430468e-07}
{"train_loss": 0.04294880852103233, "global_step": 265908, "epoch": 2987, "lr": 1.0315454212204145e-07}
{"train_loss": 0.028023025020956993, "global_step": 265909, "epoch": 2987, "lr": 1.0311732509780125e-07}
{"train_loss": 0.023752203211188316, "global_step": 265910, "epoch": 2987, "lr": 1.0308011478158963e-07}
{"train_loss": 0.08186464011669159, "global_step": 265911, "epoch": 2987, "lr": 1.0304291117341214e-07}
{"train_loss": 0.04693945497274399, "global_step": 265912, "epoch": 2987, "lr": 1.0300571427326877e-07}
{"train_loss": 0.03632136061787605, "global_step": 265913, "epoch": 2987, "lr": 1.0296852408117063e-07}
{"train_loss": 0.026662752032279968, "global_step": 265914, "epoch": 2987, "lr": 1.0293134059711773e-07}
{"train_loss": 0.0990387499332428, "global_step": 265915, "epoch": 2987, "lr": 1.0289416382112115e-07}
{"train_loss": 0.06354234367609024, "global_step": 265916, "epoch": 2987, "lr": 1.0285699375318092e-07}
{"train_loss": 0.06710764765739441, "global_step": 265917, "epoch": 2987, "lr": 1.0281983039330257e-07}
{"train_loss": 0.028561774641275406, "global_step": 265918, "epoch": 2987, "lr": 1.0278267374149164e-07}
{"train_loss": 0.021500134840607643, "global_step": 265919, "epoch": 2987, "lr": 1.0274552379775371e-07}
{"train_loss": 0.01386383268982172, "global_step": 265920, "epoch": 2987, "lr": 1.027083805620943e-07}
{"train_loss": 0.05891205742955208, "global_step": 265921, "epoch": 2987, "lr": 1.0267124403451899e-07}
{"train_loss": 0.04039997607469559, "global_step": 265922, "epoch": 2987, "lr": 1.0263411421503332e-07}
{"train_loss": 0.07532130926847458, "global_step": 265923, "epoch": 2987, "lr": 1.0259699110363175e-07}
{"train_loss": 0.05259641632437706, "global_step": 265924, "epoch": 2987, "lr": 1.0255987470033646e-07}
{"train_loss": 0.03182245418429375, "global_step": 265925, "epoch": 2987, "lr": 1.0252276500513636e-07}
{"train_loss": 0.04267125576734543, "global_step": 265926, "epoch": 2987, "lr": 1.024856620180481e-07}
{"train_loss": 0.047342412173748016, "global_step": 265927, "epoch": 2987, "lr": 1.0244856573907169e-07}
{"train_loss": 0.05529182404279709, "global_step": 265928, "epoch": 2987, "lr": 1.0241147616821822e-07}
{"train_loss": 0.030703360214829445, "global_step": 265929, "epoch": 2987, "lr": 1.0237439330548215e-07}
{"train_loss": 0.018922392278909683, "global_step": 265930, "epoch": 2987, "lr": 1.0233731715087458e-07}
{"train_loss": 0.045394988196870587, "global_step": 265931, "epoch": 2987, "lr": 1.023002477043955e-07, "val_loss": 9.640463829040527}
{"train_loss": 0.028003470972180367, "global_step": 265932, "epoch": 2988, "lr": 1.0226318496606158e-07}
{"train_loss": 0.041289933025836945, "global_step": 265933, "epoch": 2988, "lr": 1.022261289358617e-07}
{"train_loss": 0.016843022778630257, "global_step": 265934, "epoch": 2988, "lr": 1.0218907961381807e-07}
{"train_loss": 0.026717379689216614, "global_step": 265935, "epoch": 2988, "lr": 1.021520369999196e-07}
{"train_loss": 0.06489069759845734, "global_step": 265936, "epoch": 2988, "lr": 1.0211500109418293e-07}
{"train_loss": 0.012070843018591404, "global_step": 265937, "epoch": 2988, "lr": 1.0207797189660806e-07}
{"train_loss": 0.029131056740880013, "global_step": 265938, "epoch": 2988, "lr": 1.0204094940720055e-07}
{"train_loss": 0.03962576761841774, "global_step": 265939, "epoch": 2988, "lr": 1.0200393362596594e-07}
{"train_loss": 0.05214538797736168, "global_step": 265940, "epoch": 2988, "lr": 1.0196692455290424e-07}
{"train_loss": 0.031262435019016266, "global_step": 265941, "epoch": 2988, "lr": 1.019299221880321e-07}
{"train_loss": 0.021803095936775208, "global_step": 265942, "epoch": 2988, "lr": 1.0189292653134397e-07}
{"train_loss": 0.04575056582689285, "global_step": 265943, "epoch": 2988, "lr": 1.018559375828454e-07}
{"train_loss": 0.021319005638360977, "global_step": 265944, "epoch": 2988, "lr": 1.0181895534254748e-07}
{"train_loss": 0.046003300696611404, "global_step": 265945, "epoch": 2988, "lr": 1.0178197981045024e-07}
{"train_loss": 0.04125208407640457, "global_step": 265946, "epoch": 2988, "lr": 1.0174501098655919e-07}
{"train_loss": 0.022519709542393684, "global_step": 265947, "epoch": 2988, "lr": 1.0170804887088548e-07}
{"train_loss": 0.05129549652338028, "global_step": 265948, "epoch": 2988, "lr": 1.0167109346342352e-07}
{"train_loss": 0.018509317189455032, "global_step": 265949, "epoch": 2988, "lr": 1.0163414476418998e-07}
{"train_loss": 0.045256536453962326, "global_step": 265950, "epoch": 2988, "lr": 1.0159720277317375e-07}
{"train_loss": 0.0744752511382103, "global_step": 265951, "epoch": 2988, "lr": 1.0156026749039704e-07}
{"train_loss": 0.005298974458128214, "global_step": 265952, "epoch": 2988, "lr": 1.015233389158543e-07}
{"train_loss": 0.06280245631933212, "global_step": 265953, "epoch": 2988, "lr": 1.0148641704955663e-07}
{"train_loss": 0.07005918771028519, "global_step": 265954, "epoch": 2988, "lr": 1.0144950189150404e-07}
{"train_loss": 0.05913367122411728, "global_step": 265955, "epoch": 2988, "lr": 1.0141259344170206e-07}
{"train_loss": 0.022874532267451286, "global_step": 265956, "epoch": 2988, "lr": 1.0137569170016181e-07}
{"train_loss": 0.053402889519929886, "global_step": 265957, "epoch": 2988, "lr": 1.0133879666687773e-07}
{"train_loss": 0.043397508561611176, "global_step": 265958, "epoch": 2988, "lr": 1.0130190834186094e-07}
{"train_loss": 0.05682838708162308, "global_step": 265959, "epoch": 2988, "lr": 1.0126502672511696e-07}
{"train_loss": 0.0406254306435585, "global_step": 265960, "epoch": 2988, "lr": 1.0122815181665135e-07}
{"train_loss": 0.041016776114702225, "global_step": 265961, "epoch": 2988, "lr": 1.0119128361646968e-07}
{"train_loss": 0.08149819076061249, "global_step": 265962, "epoch": 2988, "lr": 1.0115442212457193e-07}
{"train_loss": 0.07280487567186356, "global_step": 265963, "epoch": 2988, "lr": 1.0111756734096367e-07}
{"train_loss": 0.029328016564249992, "global_step": 265964, "epoch": 2988, "lr": 1.0108071926565599e-07}
{"train_loss": 0.040190257132053375, "global_step": 265965, "epoch": 2988, "lr": 1.0104387789864333e-07}
{"train_loss": 0.05979890003800392, "global_step": 265966, "epoch": 2988, "lr": 1.0100704323994237e-07}
{"train_loss": 0.030754201114177704, "global_step": 265967, "epoch": 2988, "lr": 1.0097021528955309e-07}
{"train_loss": 0.12291894853115082, "global_step": 265968, "epoch": 2988, "lr": 1.0093339404748104e-07}
{"train_loss": 0.06240357458591461, "global_step": 265969, "epoch": 2988, "lr": 1.0089657951372623e-07}
{"train_loss": 0.05837680771946907, "global_step": 265970, "epoch": 2988, "lr": 1.0085977168829975e-07}
{"train_loss": 0.023151936009526253, "global_step": 265971, "epoch": 2988, "lr": 1.008229705712016e-07}
{"train_loss": 0.05099894851446152, "global_step": 265972, "epoch": 2988, "lr": 1.007861761624429e-07}
{"train_loss": 0.07235924899578094, "global_step": 265973, "epoch": 2988, "lr": 1.007493884620292e-07}
{"train_loss": 0.04544461891055107, "global_step": 265974, "epoch": 2988, "lr": 1.0071260746995492e-07}
{"train_loss": 0.020012248307466507, "global_step": 265975, "epoch": 2988, "lr": 1.0067583318623119e-07}
{"train_loss": 0.021129218861460686, "global_step": 265976, "epoch": 2988, "lr": 1.0063906561086911e-07}
{"train_loss": 0.021674450486898422, "global_step": 265977, "epoch": 2988, "lr": 1.0060230474386312e-07}
{"train_loss": 0.037042733281850815, "global_step": 265978, "epoch": 2988, "lr": 1.0056555058522433e-07}
{"train_loss": 0.045865803956985474, "global_step": 265979, "epoch": 2988, "lr": 1.0052880313495828e-07}
{"train_loss": 0.02189936488866806, "global_step": 265980, "epoch": 2988, "lr": 1.0049206239306496e-07}
{"train_loss": 0.03485392406582832, "global_step": 265981, "epoch": 2988, "lr": 1.0045532835955551e-07}
{"train_loss": 0.07205719500780106, "global_step": 265982, "epoch": 2988, "lr": 1.0041860103442991e-07}
{"train_loss": 0.035186417400836945, "global_step": 265983, "epoch": 2988, "lr": 1.003818804176937e-07}
{"train_loss": 0.04994722828269005, "global_step": 265984, "epoch": 2988, "lr": 1.0034516650935799e-07}
{"train_loss": 0.016650931909680367, "global_step": 265985, "epoch": 2988, "lr": 1.0030845930941723e-07}
{"train_loss": 0.024947350844740868, "global_step": 265986, "epoch": 2988, "lr": 1.0027175881788253e-07}
{"train_loss": 0.03431478515267372, "global_step": 265987, "epoch": 2988, "lr": 1.0023506503475943e-07}
{"train_loss": 0.012502630241215229, "global_step": 265988, "epoch": 2988, "lr": 1.0019837796004793e-07}
{"train_loss": 0.03926832601428032, "global_step": 265989, "epoch": 2988, "lr": 1.0016169759375916e-07}
{"train_loss": 0.06564533710479736, "global_step": 265990, "epoch": 2988, "lr": 1.0012502393589862e-07}
{"train_loss": 0.04540077969431877, "global_step": 265991, "epoch": 2988, "lr": 1.0008835698646635e-07}
{"train_loss": 0.036954786628484726, "global_step": 265992, "epoch": 2988, "lr": 1.0005169674546788e-07}
{"train_loss": 0.04538826644420624, "global_step": 265993, "epoch": 2988, "lr": 1.0001504321290878e-07}
{"train_loss": 0.033802930265665054, "global_step": 265994, "epoch": 2988, "lr": 9.997839638879459e-08}
{"train_loss": 0.030013708397746086, "global_step": 265995, "epoch": 2988, "lr": 9.994175627313085e-08}
{"train_loss": 0.04665989428758621, "global_step": 265996, "epoch": 2988, "lr": 9.990512286592313e-08}
{"train_loss": 0.039727091789245605, "global_step": 265997, "epoch": 2988, "lr": 9.986849616717142e-08}
{"train_loss": 0.006165849510580301, "global_step": 265998, "epoch": 2988, "lr": 9.983187617688683e-08}
{"train_loss": 0.029334813356399536, "global_step": 265999, "epoch": 2988, "lr": 9.979526289506935e-08}
{"train_loss": 0.024745609611272812, "global_step": 266000, "epoch": 2988, "lr": 9.97586563217301e-08}
{"train_loss": 0.08022420853376389, "global_step": 266001, "epoch": 2988, "lr": 9.97220564568635e-08}
{"train_loss": 0.017390644177794456, "global_step": 266002, "epoch": 2988, "lr": 9.968546330048623e-08}
{"train_loss": 0.014454723335802555, "global_step": 266003, "epoch": 2988, "lr": 9.964887685259827e-08}
{"train_loss": 0.0515984483063221, "global_step": 266004, "epoch": 2988, "lr": 9.961229711320518e-08}
{"train_loss": 0.04658122360706329, "global_step": 266005, "epoch": 2988, "lr": 9.957572408230698e-08}
{"train_loss": 0.04479539766907692, "global_step": 266006, "epoch": 2988, "lr": 9.953915775991474e-08}
{"train_loss": 0.01733144000172615, "global_step": 266007, "epoch": 2988, "lr": 9.950259814603402e-08}
{"train_loss": 0.033987581729888916, "global_step": 266008, "epoch": 2988, "lr": 9.946604524065928e-08}
{"train_loss": 0.03152943029999733, "global_step": 266009, "epoch": 2988, "lr": 9.94294990438127e-08}
{"train_loss": 0.03334920480847359, "global_step": 266010, "epoch": 2988, "lr": 9.939295955548322e-08}
{"train_loss": 0.06758561730384827, "global_step": 266011, "epoch": 2988, "lr": 9.935642677568191e-08}
{"train_loss": 0.03322555124759674, "global_step": 266012, "epoch": 2988, "lr": 9.931990070441988e-08}
{"train_loss": 0.006850455421954393, "global_step": 266013, "epoch": 2988, "lr": 9.928338134169158e-08}
{"train_loss": 0.009757989086210728, "global_step": 266014, "epoch": 2988, "lr": 9.92468686875081e-08}
{"train_loss": 0.0571121945977211, "global_step": 266015, "epoch": 2988, "lr": 9.921036274186946e-08}
{"train_loss": 0.054544392973184586, "global_step": 266016, "epoch": 2988, "lr": 9.917386350478675e-08}
{"train_loss": 0.037266653031110764, "global_step": 266017, "epoch": 2988, "lr": 9.913737097625997e-08}
{"train_loss": 0.0488949790596962, "global_step": 266018, "epoch": 2988, "lr": 9.910088515629468e-08}
{"train_loss": 0.03985731303691864, "global_step": 266019, "epoch": 2988, "lr": 9.906440604490197e-08}
{"train_loss": 0.040220584694212407, "global_step": 266020, "epoch": 2988, "lr": 9.902793364208185e-08, "val_loss": 9.6404390335083}
{"train_loss": 0.025728493928909302, "global_step": 266021, "epoch": 2989, "lr": 9.899146794783432e-08}
{"train_loss": 0.05889429524540901, "global_step": 266022, "epoch": 2989, "lr": 9.895500896217047e-08}
{"train_loss": 0.11195812374353409, "global_step": 266023, "epoch": 2989, "lr": 9.891855668509587e-08}
{"train_loss": 0.03241437301039696, "global_step": 266024, "epoch": 2989, "lr": 9.88821111166105e-08}
{"train_loss": 0.025834474712610245, "global_step": 266025, "epoch": 2989, "lr": 9.884567225672548e-08}
{"train_loss": 0.03690676391124725, "global_step": 266026, "epoch": 2989, "lr": 9.88092401054408e-08}
{"train_loss": 0.007360502146184444, "global_step": 266027, "epoch": 2989, "lr": 9.877281466276201e-08}
{"train_loss": 0.06090027838945389, "global_step": 266028, "epoch": 2989, "lr": 9.87363959287002e-08}
{"train_loss": 0.01578051783144474, "global_step": 266029, "epoch": 2989, "lr": 9.869998390324986e-08}
{"train_loss": 0.0451614111661911, "global_step": 266030, "epoch": 2989, "lr": 9.866357858642205e-08}
{"train_loss": 0.07160146534442902, "global_step": 266031, "epoch": 2989, "lr": 9.862717997822235e-08}
{"train_loss": 0.027976250275969505, "global_step": 266032, "epoch": 2989, "lr": 9.859078807865074e-08}
{"train_loss": 0.01822843961417675, "global_step": 266033, "epoch": 2989, "lr": 9.855440288771833e-08}
{"train_loss": 0.04595758393406868, "global_step": 266034, "epoch": 2989, "lr": 9.851802440542512e-08}
{"train_loss": 0.04617020860314369, "global_step": 266035, "epoch": 2989, "lr": 9.848165263178221e-08}
{"train_loss": 0.059390872716903687, "global_step": 266036, "epoch": 2989, "lr": 9.844528756678406e-08}
{"train_loss": 0.04680073261260986, "global_step": 266037, "epoch": 2989, "lr": 9.84089292104473e-08}
{"train_loss": 0.03317675366997719, "global_step": 266038, "epoch": 2989, "lr": 9.83725775627664e-08}
{"train_loss": 0.03425626829266548, "global_step": 266039, "epoch": 2989, "lr": 9.833623262375246e-08}
{"train_loss": 0.05905047804117203, "global_step": 266040, "epoch": 2989, "lr": 9.829989439341103e-08}
{"train_loss": 0.06700007617473602, "global_step": 266041, "epoch": 2989, "lr": 9.82635628717421e-08}
{"train_loss": 0.06053099036216736, "global_step": 266042, "epoch": 2989, "lr": 9.822723805875678e-08}
{"train_loss": 0.04381674528121948, "global_step": 266043, "epoch": 2989, "lr": 9.819091995445506e-08}
{"train_loss": 0.02460855059325695, "global_step": 266044, "epoch": 2989, "lr": 9.815460855884251e-08}
{"train_loss": 0.042518120259046555, "global_step": 266045, "epoch": 2989, "lr": 9.811830387192467e-08}
{"train_loss": 0.03513709828257561, "global_step": 266046, "epoch": 2989, "lr": 9.808200589370708e-08}
{"train_loss": 0.03465116024017334, "global_step": 266047, "epoch": 2989, "lr": 9.804571462419532e-08}
{"train_loss": 0.07458707690238953, "global_step": 266048, "epoch": 2989, "lr": 9.800943006339491e-08}
{"train_loss": 0.07077217847108841, "global_step": 266049, "epoch": 2989, "lr": 9.797315221130588e-08}
{"train_loss": 0.05367995798587799, "global_step": 266050, "epoch": 2989, "lr": 9.79368810679393e-08}
{"train_loss": 0.033474668860435486, "global_step": 266051, "epoch": 2989, "lr": 9.790061663328964e-08}
{"train_loss": 0.026990314945578575, "global_step": 266052, "epoch": 2989, "lr": 9.786435890737356e-08}
{"train_loss": 0.03852423280477524, "global_step": 266053, "epoch": 2989, "lr": 9.782810789019103e-08}
{"train_loss": 0.03441661223769188, "global_step": 266054, "epoch": 2989, "lr": 9.779186358175319e-08}
{"train_loss": 0.02690538950264454, "global_step": 266055, "epoch": 2989, "lr": 9.775562598204891e-08}
{"train_loss": 0.030473798513412476, "global_step": 266056, "epoch": 2989, "lr": 9.77193950911004e-08}
{"train_loss": 0.04576929658651352, "global_step": 266057, "epoch": 2989, "lr": 9.768317090890211e-08}
{"train_loss": 0.030982688069343567, "global_step": 266058, "epoch": 2989, "lr": 9.764695343546515e-08}
{"train_loss": 0.03151673078536987, "global_step": 266059, "epoch": 2989, "lr": 9.761074267078951e-08}
{"train_loss": 0.05767073854804039, "global_step": 266060, "epoch": 2989, "lr": 9.75745386148752e-08}
{"train_loss": 0.05687225982546806, "global_step": 266061, "epoch": 2989, "lr": 9.753834126774442e-08}
{"train_loss": 0.01366053894162178, "global_step": 266062, "epoch": 2989, "lr": 9.750215062938605e-08}
{"train_loss": 0.06302423775196075, "global_step": 266063, "epoch": 2989, "lr": 9.746596669981123e-08}
{"train_loss": 0.04837984964251518, "global_step": 266064, "epoch": 2989, "lr": 9.742978947901993e-08}
{"train_loss": 0.03390032425522804, "global_step": 266065, "epoch": 2989, "lr": 9.73936189670288e-08}
{"train_loss": 0.016619248315691948, "global_step": 266066, "epoch": 2989, "lr": 9.735745516382677e-08}
{"train_loss": 0.023121671751141548, "global_step": 266067, "epoch": 2989, "lr": 9.732129806943047e-08}
{"train_loss": 0.026972942054271698, "global_step": 266068, "epoch": 2989, "lr": 9.728514768384545e-08}
{"train_loss": 0.05231967940926552, "global_step": 266069, "epoch": 2989, "lr": 9.724900400706616e-08}
{"train_loss": 0.01389777660369873, "global_step": 266070, "epoch": 2989, "lr": 9.72128670391037e-08}
{"train_loss": 0.06936710327863693, "global_step": 266071, "epoch": 2989, "lr": 9.717673677996364e-08}
{"train_loss": 0.057455919682979584, "global_step": 266072, "epoch": 2989, "lr": 9.714061322965151e-08}
{"train_loss": 0.05002427101135254, "global_step": 266073, "epoch": 2989, "lr": 9.710449638817287e-08}
{"train_loss": 0.043699707835912704, "global_step": 266074, "epoch": 2989, "lr": 9.706838625552771e-08}
{"train_loss": 0.02490432932972908, "global_step": 266075, "epoch": 2989, "lr": 9.70322828317216e-08}
{"train_loss": 0.05348319187760353, "global_step": 266076, "epoch": 2989, "lr": 9.699618611676564e-08}
{"train_loss": 0.04280325025320053, "global_step": 266077, "epoch": 2989, "lr": 9.696009611065426e-08}
{"train_loss": 0.05411326512694359, "global_step": 266078, "epoch": 2989, "lr": 9.692401281340413e-08}
{"train_loss": 0.03208891674876213, "global_step": 266079, "epoch": 2989, "lr": 9.688793622500969e-08}
{"train_loss": 0.0498277023434639, "global_step": 266080, "epoch": 2989, "lr": 9.685186634548204e-08}
{"train_loss": 0.07088030129671097, "global_step": 266081, "epoch": 2989, "lr": 9.681580317482675e-08}
{"train_loss": 0.0341857448220253, "global_step": 266082, "epoch": 2989, "lr": 9.67797467130438e-08}
{"train_loss": 0.1355571150779724, "global_step": 266083, "epoch": 2989, "lr": 9.674369696013874e-08}
{"train_loss": 0.04452294111251831, "global_step": 266084, "epoch": 2989, "lr": 9.670765391612269e-08}
{"train_loss": 0.025662757456302643, "global_step": 266085, "epoch": 2989, "lr": 9.667161758099563e-08}
{"train_loss": 0.06352630257606506, "global_step": 266086, "epoch": 2989, "lr": 9.663558795475758e-08}
{"train_loss": 0.046658750623464584, "global_step": 266087, "epoch": 2989, "lr": 9.659956503742517e-08}
{"train_loss": 0.05496164783835411, "global_step": 266088, "epoch": 2989, "lr": 9.656354882899288e-08}
{"train_loss": 0.04063279926776886, "global_step": 266089, "epoch": 2989, "lr": 9.652753932947178e-08}
{"train_loss": 0.039546284824609756, "global_step": 266090, "epoch": 2989, "lr": 9.64915365388619e-08}
{"train_loss": 0.04031522944569588, "global_step": 266091, "epoch": 2989, "lr": 9.64555404571743e-08}
{"train_loss": 0.019447533413767815, "global_step": 266092, "epoch": 2989, "lr": 9.641955108440348e-08}
{"train_loss": 0.03555666282773018, "global_step": 266093, "epoch": 2989, "lr": 9.638356842056607e-08}
{"train_loss": 0.05704597011208534, "global_step": 266094, "epoch": 2989, "lr": 9.634759246566205e-08}
{"train_loss": 0.07157066464424133, "global_step": 266095, "epoch": 2989, "lr": 9.631162321969145e-08}
{"train_loss": 0.05290212109684944, "global_step": 266096, "epoch": 2989, "lr": 9.627566068266536e-08}
{"train_loss": 0.014489366672933102, "global_step": 266097, "epoch": 2989, "lr": 9.623970485458933e-08}
{"train_loss": 0.05036689341068268, "global_step": 266098, "epoch": 2989, "lr": 9.620375573546336e-08}
{"train_loss": 0.039185866713523865, "global_step": 266099, "epoch": 2989, "lr": 9.616781332529301e-08}
{"train_loss": 0.02569068782031536, "global_step": 266100, "epoch": 2989, "lr": 9.613187762408937e-08}
{"train_loss": 0.022326786071062088, "global_step": 266101, "epoch": 2989, "lr": 9.60959486318469e-08}
{"train_loss": 0.023848503828048706, "global_step": 266102, "epoch": 2989, "lr": 9.606002634858224e-08}
{"train_loss": 0.01870005391538143, "global_step": 266103, "epoch": 2989, "lr": 9.602411077428985e-08}
{"train_loss": 0.04008576273918152, "global_step": 266104, "epoch": 2989, "lr": 9.598820190898083e-08}
{"train_loss": 0.054314542561769485, "global_step": 266105, "epoch": 2989, "lr": 9.595229975265518e-08}
{"train_loss": 0.03511837497353554, "global_step": 266106, "epoch": 2989, "lr": 9.5916404305324e-08}
{"train_loss": 0.02945462055504322, "global_step": 266107, "epoch": 2989, "lr": 9.58805155669873e-08}
{"train_loss": 0.04511840268969536, "global_step": 266108, "epoch": 2989, "lr": 9.58446335376506e-08}
{"train_loss": 0.04301297503492136, "global_step": 266109, "epoch": 2989, "lr": 9.580875821731949e-08, "val_loss": 9.550779342651367}
{"train_loss": 0.028911687433719635, "global_step": 266110, "epoch": 2990, "lr": 9.57728896059995e-08}
{"train_loss": 0.01900276355445385, "global_step": 266111, "epoch": 2990, "lr": 9.573702770369064e-08}
{"train_loss": 0.056688155978918076, "global_step": 266112, "epoch": 2990, "lr": 9.570117251040401e-08}
{"train_loss": 0.03352656215429306, "global_step": 266113, "epoch": 2990, "lr": 9.566532402614514e-08}
{"train_loss": 0.04166962951421738, "global_step": 266114, "epoch": 2990, "lr": 9.562948225091406e-08}
{"train_loss": 0.06474663317203522, "global_step": 266115, "epoch": 2990, "lr": 9.55936471847163e-08}
{"train_loss": 0.05264805257320404, "global_step": 266116, "epoch": 2990, "lr": 9.555781882755744e-08}
{"train_loss": 0.02313661202788353, "global_step": 266117, "epoch": 2990, "lr": 9.552199717944299e-08}
{"train_loss": 0.019960958510637283, "global_step": 266118, "epoch": 2990, "lr": 9.548618224037852e-08}
{"train_loss": 0.06365856528282166, "global_step": 266119, "epoch": 2990, "lr": 9.545037401036405e-08}
{"train_loss": 0.0579821951687336, "global_step": 266120, "epoch": 2990, "lr": 9.541457248941065e-08}
{"train_loss": 0.07285130023956299, "global_step": 266121, "epoch": 2990, "lr": 9.53787776775239e-08}
{"train_loss": 0.029949543997645378, "global_step": 266122, "epoch": 2990, "lr": 9.534298957469823e-08}
{"train_loss": 0.03281471133232117, "global_step": 266123, "epoch": 2990, "lr": 9.530720818095029e-08}
{"train_loss": 0.04388923943042755, "global_step": 266124, "epoch": 2990, "lr": 9.52714334962801e-08}
{"train_loss": 0.04945309832692146, "global_step": 266125, "epoch": 2990, "lr": 9.523566552069318e-08}
{"train_loss": 0.05521530285477638, "global_step": 266126, "epoch": 2990, "lr": 9.519990425418956e-08}
{"train_loss": 0.028566649183630943, "global_step": 266127, "epoch": 2990, "lr": 9.516414969678033e-08}
{"train_loss": 0.017116939648985863, "global_step": 266128, "epoch": 2990, "lr": 9.512840184847105e-08}
{"train_loss": 0.0823795422911644, "global_step": 266129, "epoch": 2990, "lr": 9.509266070925616e-08}
{"train_loss": 0.05594368278980255, "global_step": 266130, "epoch": 2990, "lr": 9.50569262791523e-08}
{"train_loss": 0.039087653160095215, "global_step": 266131, "epoch": 2990, "lr": 9.502119855816504e-08}
{"train_loss": 0.055218491703271866, "global_step": 266132, "epoch": 2990, "lr": 9.498547754628328e-08}
{"train_loss": 0.0371796190738678, "global_step": 266133, "epoch": 2990, "lr": 9.494976324352923e-08}
{"train_loss": 0.04715321958065033, "global_step": 266134, "epoch": 2990, "lr": 9.49140556498973e-08}
{"train_loss": 0.030424801632761955, "global_step": 266135, "epoch": 2990, "lr": 9.487835476539864e-08}
{"train_loss": 0.012593808583915234, "global_step": 266136, "epoch": 2990, "lr": 9.484266059003877e-08}
{"train_loss": 0.05803077295422554, "global_step": 266137, "epoch": 2990, "lr": 9.480697312381215e-08}
{"train_loss": 0.03116745501756668, "global_step": 266138, "epoch": 2990, "lr": 9.477129236673543e-08}
{"train_loss": 0.04230721667408943, "global_step": 266139, "epoch": 2990, "lr": 9.473561831880307e-08}
{"train_loss": 0.035278186202049255, "global_step": 266140, "epoch": 2990, "lr": 9.46999509800317e-08}
{"train_loss": 0.05897863581776619, "global_step": 266141, "epoch": 2990, "lr": 9.46642903504158e-08}
{"train_loss": 0.048733487725257874, "global_step": 266142, "epoch": 2990, "lr": 9.46286364299609e-08}
{"train_loss": 0.02901664935052395, "global_step": 266143, "epoch": 2990, "lr": 9.459298921867809e-08}
{"train_loss": 0.06758376955986023, "global_step": 266144, "epoch": 2990, "lr": 9.455734871657296e-08}
{"train_loss": 0.033541351556777954, "global_step": 266145, "epoch": 2990, "lr": 9.452171492363993e-08}
{"train_loss": 0.045233044773340225, "global_step": 266146, "epoch": 2990, "lr": 9.448608783989565e-08}
{"train_loss": 0.03181910514831543, "global_step": 266147, "epoch": 2990, "lr": 9.445046746533458e-08}
{"train_loss": 0.07682375609874725, "global_step": 266148, "epoch": 2990, "lr": 9.441485379996784e-08}
{"train_loss": 0.017285620793700218, "global_step": 266149, "epoch": 2990, "lr": 9.437924684380095e-08}
{"train_loss": 0.0470304898917675, "global_step": 266150, "epoch": 2990, "lr": 9.434364659683392e-08}
{"train_loss": 0.041897766292095184, "global_step": 266151, "epoch": 2990, "lr": 9.430805305907785e-08}
{"train_loss": 0.0525733083486557, "global_step": 266152, "epoch": 2990, "lr": 9.427246623053276e-08}
{"train_loss": 0.02432461827993393, "global_step": 266153, "epoch": 2990, "lr": 9.423688611119863e-08}
{"train_loss": 0.01943756453692913, "global_step": 266154, "epoch": 2990, "lr": 9.420131270109211e-08}
{"train_loss": 0.060271065682172775, "global_step": 266155, "epoch": 2990, "lr": 9.416574600021322e-08}
{"train_loss": 0.014841774478554726, "global_step": 266156, "epoch": 2990, "lr": 9.413018600856194e-08}
{"train_loss": 0.037146855145692825, "global_step": 266157, "epoch": 2990, "lr": 9.409463272614382e-08}
{"train_loss": 0.05392605438828468, "global_step": 266158, "epoch": 2990, "lr": 9.405908615296998e-08}
{"train_loss": 0.03847675397992134, "global_step": 266159, "epoch": 2990, "lr": 9.402354628904597e-08}
{"train_loss": 0.054058849811553955, "global_step": 266160, "epoch": 2990, "lr": 9.398801313436623e-08}
{"train_loss": 0.040621448308229446, "global_step": 266161, "epoch": 2990, "lr": 9.395248668894185e-08}
{"train_loss": 0.04551936313509941, "global_step": 266162, "epoch": 2990, "lr": 9.391696695277841e-08}
{"train_loss": 0.046369586139917374, "global_step": 266163, "epoch": 2990, "lr": 9.388145392588144e-08}
{"train_loss": 0.0936347246170044, "global_step": 266164, "epoch": 2990, "lr": 9.384594760825094e-08}
{"train_loss": 0.044540226459503174, "global_step": 266165, "epoch": 2990, "lr": 9.381044799989803e-08}
{"train_loss": 0.03694693744182587, "global_step": 266166, "epoch": 2990, "lr": 9.377495510081713e-08}
{"train_loss": 0.029662828892469406, "global_step": 266167, "epoch": 2990, "lr": 9.373946891102492e-08}
{"train_loss": 0.021601548418402672, "global_step": 266168, "epoch": 2990, "lr": 9.370398943052139e-08}
{"train_loss": 0.03339817374944687, "global_step": 266169, "epoch": 2990, "lr": 9.366851665931208e-08}
{"train_loss": 0.04797443002462387, "global_step": 266170, "epoch": 2990, "lr": 9.363305059739702e-08}
{"train_loss": 0.03855013847351074, "global_step": 266171, "epoch": 2990, "lr": 9.359759124478728e-08}
{"train_loss": 0.05419953539967537, "global_step": 266172, "epoch": 2990, "lr": 9.356213860148843e-08}
{"train_loss": 0.07323549687862396, "global_step": 266173, "epoch": 2990, "lr": 9.352669266749493e-08}
{"train_loss": 0.04116981104016304, "global_step": 266174, "epoch": 2990, "lr": 9.349125344282339e-08}
{"train_loss": 0.06407836824655533, "global_step": 266175, "epoch": 2990, "lr": 9.345582092746829e-08}
{"train_loss": 0.030904976651072502, "global_step": 266176, "epoch": 2990, "lr": 9.342039512144629e-08}
{"train_loss": 0.05872447043657303, "global_step": 266177, "epoch": 2990, "lr": 9.338497602475183e-08}
{"train_loss": 0.00981631875038147, "global_step": 266178, "epoch": 2990, "lr": 9.3349563637396e-08}
{"train_loss": 0.07137271016836166, "global_step": 266179, "epoch": 2990, "lr": 9.331415795937882e-08}
{"train_loss": 0.03929847851395607, "global_step": 266180, "epoch": 2990, "lr": 9.327875899070582e-08}
{"train_loss": 0.037225089967250824, "global_step": 266181, "epoch": 2990, "lr": 9.324336673138811e-08}
{"train_loss": 0.05575317144393921, "global_step": 266182, "epoch": 2990, "lr": 9.320798118142016e-08}
{"train_loss": 0.04611064866185188, "global_step": 266183, "epoch": 2990, "lr": 9.317260234081305e-08}
{"train_loss": 0.04289618134498596, "global_step": 266184, "epoch": 2990, "lr": 9.313723020957788e-08}
{"train_loss": 0.03850964084267616, "global_step": 266185, "epoch": 2990, "lr": 9.310186478770355e-08}
{"train_loss": 0.029879624024033546, "global_step": 266186, "epoch": 2990, "lr": 9.306650607520672e-08}
{"train_loss": 0.05918600782752037, "global_step": 266187, "epoch": 2990, "lr": 9.30311540720874e-08}
{"train_loss": 0.008118724450469017, "global_step": 266188, "epoch": 2990, "lr": 9.299580877835112e-08}
{"train_loss": 0.0270698219537735, "global_step": 266189, "epoch": 2990, "lr": 9.2960470194009e-08}
{"train_loss": 0.0600760281085968, "global_step": 266190, "epoch": 2990, "lr": 9.292513831905548e-08}
{"train_loss": 0.041735291481018066, "global_step": 266191, "epoch": 2990, "lr": 9.288981315350164e-08}
{"train_loss": 0.06826981157064438, "global_step": 266192, "epoch": 2990, "lr": 9.285449469734753e-08}
{"train_loss": 0.027519412338733673, "global_step": 266193, "epoch": 2990, "lr": 9.281918295060421e-08}
{"train_loss": 0.03470558300614357, "global_step": 266194, "epoch": 2990, "lr": 9.278387791326614e-08}
{"train_loss": 0.04392257332801819, "global_step": 266195, "epoch": 2990, "lr": 9.274857958535554e-08}
{"train_loss": 0.04712286591529846, "global_step": 266196, "epoch": 2990, "lr": 9.271328796685575e-08}
{"train_loss": 0.017288994044065475, "global_step": 266197, "epoch": 2990, "lr": 9.267800305778896e-08}
{"train_loss": 0.042731541104279874, "global_step": 266198, "epoch": 2990, "lr": 9.264272485815517e-08, "val_loss": 9.700324058532715, "train_action_mse_error": 10.070089340209961}
{"train_loss": 0.06501813977956772, "global_step": 266199, "epoch": 2991, "lr": 9.26074533679544e-08}
{"train_loss": 0.04866994544863701, "global_step": 266200, "epoch": 2991, "lr": 9.25721885871922e-08}
{"train_loss": 0.06439752876758575, "global_step": 266201, "epoch": 2991, "lr": 9.253693051587963e-08}
{"train_loss": 0.03216831386089325, "global_step": 266202, "epoch": 2991, "lr": 9.25016791540112e-08}
{"train_loss": 0.058744948357343674, "global_step": 266203, "epoch": 2991, "lr": 9.246643450160353e-08}
{"train_loss": 0.03638102114200592, "global_step": 266204, "epoch": 2991, "lr": 9.243119655865662e-08}
{"train_loss": 0.02225741557776928, "global_step": 266205, "epoch": 2991, "lr": 9.239596532517048e-08}
{"train_loss": 0.05129266902804375, "global_step": 266206, "epoch": 2991, "lr": 9.236074080115064e-08}
{"train_loss": 0.04639444500207901, "global_step": 266207, "epoch": 2991, "lr": 9.232552298660824e-08}
{"train_loss": 0.02572532370686531, "global_step": 266208, "epoch": 2991, "lr": 9.229031188154324e-08}
{"train_loss": 0.027898572385311127, "global_step": 266209, "epoch": 2991, "lr": 9.225510748596122e-08}
{"train_loss": 0.05363974720239639, "global_step": 266210, "epoch": 2991, "lr": 9.221990979987328e-08}
{"train_loss": 0.04255843535065651, "global_step": 266211, "epoch": 2991, "lr": 9.218471882326828e-08}
{"train_loss": 0.04807940870523453, "global_step": 266212, "epoch": 2991, "lr": 9.214953455616848e-08}
{"train_loss": 0.07226329296827316, "global_step": 266213, "epoch": 2991, "lr": 9.21143569985683e-08}
{"train_loss": 0.06913714855909348, "global_step": 266214, "epoch": 2991, "lr": 9.207918615047329e-08}
{"train_loss": 0.061670105904340744, "global_step": 266215, "epoch": 2991, "lr": 9.2044022011889e-08}
{"train_loss": 0.011568675749003887, "global_step": 266216, "epoch": 2991, "lr": 9.200886458282653e-08}
{"train_loss": 0.03293910250067711, "global_step": 266217, "epoch": 2991, "lr": 9.197371386328035e-08}
{"train_loss": 0.04593603312969208, "global_step": 266218, "epoch": 2991, "lr": 9.193856985326155e-08}
{"train_loss": 0.08515223115682602, "global_step": 266219, "epoch": 2991, "lr": 9.190343255277566e-08}
{"train_loss": 0.013202860951423645, "global_step": 266220, "epoch": 2991, "lr": 9.186830196182273e-08}
{"train_loss": 0.04855057969689369, "global_step": 266221, "epoch": 2991, "lr": 9.183317808040825e-08}
{"train_loss": 0.029515434056520462, "global_step": 266222, "epoch": 2991, "lr": 9.179806090854337e-08}
{"train_loss": 0.06603807955980301, "global_step": 266223, "epoch": 2991, "lr": 9.176295044622252e-08}
{"train_loss": 0.015588982962071896, "global_step": 266224, "epoch": 2991, "lr": 9.17278466934568e-08}
{"train_loss": 0.036630865186452866, "global_step": 266225, "epoch": 2991, "lr": 9.169274965025176e-08}
{"train_loss": 0.06449282914400101, "global_step": 266226, "epoch": 2991, "lr": 9.165765931660741e-08}
{"train_loss": 0.04334524646401405, "global_step": 266227, "epoch": 2991, "lr": 9.162257569253485e-08}
{"train_loss": 0.04567181318998337, "global_step": 266228, "epoch": 2991, "lr": 9.158749877803407e-08}
{"train_loss": 0.027871429920196533, "global_step": 266229, "epoch": 2991, "lr": 9.155242857311063e-08}
{"train_loss": 0.0735497772693634, "global_step": 266230, "epoch": 2991, "lr": 9.151736507777009e-08}
{"train_loss": 0.045363761484622955, "global_step": 266231, "epoch": 2991, "lr": 9.148230829201799e-08}
{"train_loss": 0.07825778424739838, "global_step": 266232, "epoch": 2991, "lr": 9.144725821585987e-08}
{"train_loss": 0.01590261608362198, "global_step": 266233, "epoch": 2991, "lr": 9.141221484929574e-08}
{"train_loss": 0.02876204252243042, "global_step": 266234, "epoch": 2991, "lr": 9.137717819233116e-08}
{"train_loss": 0.011883950792253017, "global_step": 266235, "epoch": 2991, "lr": 9.134214824497167e-08}
{"train_loss": 0.028323765844106674, "global_step": 266236, "epoch": 2991, "lr": 9.130712500722838e-08}
{"train_loss": 0.0259955246001482, "global_step": 266237, "epoch": 2991, "lr": 9.127210847909572e-08}
{"train_loss": 0.02720823511481285, "global_step": 266238, "epoch": 2991, "lr": 9.123709866058483e-08}
{"train_loss": 0.014538244344294071, "global_step": 266239, "epoch": 2991, "lr": 9.120209555170122e-08}
{"train_loss": 0.07816871255636215, "global_step": 266240, "epoch": 2991, "lr": 9.116709915243937e-08}
{"train_loss": 0.030248699709773064, "global_step": 266241, "epoch": 2991, "lr": 9.113210946282147e-08}
{"train_loss": 0.06439772993326187, "global_step": 266242, "epoch": 2991, "lr": 9.109712648283642e-08}
{"train_loss": 0.04216562956571579, "global_step": 266243, "epoch": 2991, "lr": 9.106215021249531e-08}
{"train_loss": 0.02899141237139702, "global_step": 266244, "epoch": 2991, "lr": 9.102718065180926e-08}
{"train_loss": 0.06217797100543976, "global_step": 266245, "epoch": 2991, "lr": 9.099221780076717e-08}
{"train_loss": 0.0684027448296547, "global_step": 266246, "epoch": 2991, "lr": 9.095726165939123e-08}
{"train_loss": 0.03966713324189186, "global_step": 266247, "epoch": 2991, "lr": 9.092231222767034e-08}
{"train_loss": 0.06073639914393425, "global_step": 266248, "epoch": 2991, "lr": 9.088736950562115e-08}
{"train_loss": 0.05658285319805145, "global_step": 266249, "epoch": 2991, "lr": 9.085243349324368e-08}
{"train_loss": 0.05604731664061546, "global_step": 266250, "epoch": 2991, "lr": 9.081750419053792e-08}
{"train_loss": 0.02305612713098526, "global_step": 266251, "epoch": 2991, "lr": 9.078258159752051e-08}
{"train_loss": 0.031056784093379974, "global_step": 266252, "epoch": 2991, "lr": 9.074766571418592e-08}
{"train_loss": 0.039440788328647614, "global_step": 266253, "epoch": 2991, "lr": 9.071275654053968e-08}
{"train_loss": 0.059517741203308105, "global_step": 266254, "epoch": 2991, "lr": 9.06778540765929e-08}
{"train_loss": 0.017601758241653442, "global_step": 266255, "epoch": 2991, "lr": 9.064295832234005e-08}
{"train_loss": 0.05720672011375427, "global_step": 266256, "epoch": 2991, "lr": 9.060806927779774e-08}
{"train_loss": 0.02575470507144928, "global_step": 266257, "epoch": 2991, "lr": 9.057318694296602e-08}
{"train_loss": 0.06745952367782593, "global_step": 266258, "epoch": 2991, "lr": 9.053831131784485e-08}
{"train_loss": 0.04520468786358833, "global_step": 266259, "epoch": 2991, "lr": 9.05034424024398e-08}
{"train_loss": 0.05457974225282669, "global_step": 266260, "epoch": 2991, "lr": 9.046858019676197e-08}
{"train_loss": 0.0426466204226017, "global_step": 266261, "epoch": 2991, "lr": 9.043372470081135e-08}
{"train_loss": 0.0517125129699707, "global_step": 266262, "epoch": 2991, "lr": 9.039887591459351e-08}
{"train_loss": 0.04157261922955513, "global_step": 266263, "epoch": 2991, "lr": 9.036403383811398e-08}
{"train_loss": 0.026514314115047455, "global_step": 266264, "epoch": 2991, "lr": 9.032919847137833e-08}
{"train_loss": 0.019558699801564217, "global_step": 266265, "epoch": 2991, "lr": 9.029436981438655e-08}
{"train_loss": 0.04446382075548172, "global_step": 266266, "epoch": 2991, "lr": 9.025954786714974e-08}
{"train_loss": 0.04502894729375839, "global_step": 266267, "epoch": 2991, "lr": 9.02247326296679e-08}
{"train_loss": 0.05220934748649597, "global_step": 266268, "epoch": 2991, "lr": 9.01899241019466e-08}
{"train_loss": 0.03927907347679138, "global_step": 266269, "epoch": 2991, "lr": 9.015512228399136e-08}
{"train_loss": 0.05901776999235153, "global_step": 266270, "epoch": 2991, "lr": 9.012032717580221e-08}
{"train_loss": 0.059850238263607025, "global_step": 266271, "epoch": 2991, "lr": 9.008553877739578e-08}
{"train_loss": 0.03642972558736801, "global_step": 266272, "epoch": 2991, "lr": 9.005075708876099e-08}
{"train_loss": 0.03217830881476402, "global_step": 266273, "epoch": 2991, "lr": 9.001598210992001e-08}
{"train_loss": 0.08247154951095581, "global_step": 266274, "epoch": 2991, "lr": 8.998121384086178e-08}
{"train_loss": 0.05853414162993431, "global_step": 266275, "epoch": 2991, "lr": 8.994645228159738e-08}
{"train_loss": 0.0444253608584404, "global_step": 266276, "epoch": 2991, "lr": 8.991169743213235e-08}
{"train_loss": 0.039391882717609406, "global_step": 266277, "epoch": 2991, "lr": 8.98769492924667e-08}
{"train_loss": 0.06907835602760315, "global_step": 266278, "epoch": 2991, "lr": 8.984220786261155e-08}
{"train_loss": 0.03571900352835655, "global_step": 266279, "epoch": 2991, "lr": 8.980747314257243e-08}
{"train_loss": 0.09403160959482193, "global_step": 266280, "epoch": 2991, "lr": 8.977274513234934e-08}
{"train_loss": 0.03969525918364525, "global_step": 266281, "epoch": 2991, "lr": 8.97380238319423e-08}
{"train_loss": 0.0685286670923233, "global_step": 266282, "epoch": 2991, "lr": 8.970330924136794e-08}
{"train_loss": 0.011295443400740623, "global_step": 266283, "epoch": 2991, "lr": 8.966860136062072e-08}
{"train_loss": 0.05065825581550598, "global_step": 266284, "epoch": 2991, "lr": 8.963390018971174e-08}
{"train_loss": 0.0314197763800621, "global_step": 266285, "epoch": 2991, "lr": 8.9599205728641e-08}
{"train_loss": 0.06679433584213257, "global_step": 266286, "epoch": 2991, "lr": 8.956451797741406e-08}
{"train_loss": 0.04537697190816483, "global_step": 266287, "epoch": 2991, "lr": 8.9529836936042e-08, "val_loss": 9.662599563598633}
{"train_loss": 0.05188655108213425, "global_step": 266288, "epoch": 2992, "lr": 8.949516260451929e-08}
{"train_loss": 0.040508661419153214, "global_step": 266289, "epoch": 2992, "lr": 8.946049498285702e-08}
{"train_loss": 0.05953469127416611, "global_step": 266290, "epoch": 2992, "lr": 8.942583407106076e-08}
{"train_loss": 0.08638717234134674, "global_step": 266291, "epoch": 2992, "lr": 8.939117986913049e-08}
{"train_loss": 0.07015834748744965, "global_step": 266292, "epoch": 2992, "lr": 8.93565323770773e-08}
{"train_loss": 0.04080289974808693, "global_step": 266293, "epoch": 2992, "lr": 8.932189159489568e-08}
{"train_loss": 0.04126375913619995, "global_step": 266294, "epoch": 2992, "lr": 8.928725752260226e-08}
{"train_loss": 0.026896189898252487, "global_step": 266295, "epoch": 2992, "lr": 8.925263016019147e-08}
{"train_loss": 0.016907166689634323, "global_step": 266296, "epoch": 2992, "lr": 8.921800950767444e-08}
{"train_loss": 0.05408279597759247, "global_step": 266297, "epoch": 2992, "lr": 8.918339556505118e-08}
{"train_loss": 0.050319913774728775, "global_step": 266298, "epoch": 2992, "lr": 8.914878833233275e-08}
{"train_loss": 0.02677791751921177, "global_step": 266299, "epoch": 2992, "lr": 8.911418780951918e-08}
{"train_loss": 0.042645908892154694, "global_step": 266300, "epoch": 2992, "lr": 8.907959399661048e-08}
{"train_loss": 0.046790145337581635, "global_step": 266301, "epoch": 2992, "lr": 8.904500689362328e-08}
{"train_loss": 0.048743970692157745, "global_step": 266302, "epoch": 2992, "lr": 8.901042650055203e-08}
{"train_loss": 0.017265992239117622, "global_step": 266303, "epoch": 2992, "lr": 8.897585281740783e-08}
{"train_loss": 0.044202886521816254, "global_step": 266304, "epoch": 2992, "lr": 8.89412858441907e-08}
{"train_loss": 0.030054211616516113, "global_step": 266305, "epoch": 2992, "lr": 8.890672558090618e-08}
{"train_loss": 0.05644214153289795, "global_step": 266306, "epoch": 2992, "lr": 8.887217202755982e-08}
{"train_loss": 0.04334898293018341, "global_step": 266307, "epoch": 2992, "lr": 8.883762518415717e-08}
{"train_loss": 0.03482155501842499, "global_step": 266308, "epoch": 2992, "lr": 8.880308505070378e-08}
{"train_loss": 0.022813377901911736, "global_step": 266309, "epoch": 2992, "lr": 8.876855162719966e-08}
{"train_loss": 0.03257171809673309, "global_step": 266310, "epoch": 2992, "lr": 8.87340249136559e-08}
{"train_loss": 0.009431851096451283, "global_step": 266311, "epoch": 2992, "lr": 8.869950491006695e-08}
{"train_loss": 0.05096554011106491, "global_step": 266312, "epoch": 2992, "lr": 8.866499161644948e-08}
{"train_loss": 0.05815519392490387, "global_step": 266313, "epoch": 2992, "lr": 8.863048503280346e-08}
{"train_loss": 0.03916165232658386, "global_step": 266314, "epoch": 2992, "lr": 8.859598515912892e-08}
{"train_loss": 0.059736721217632294, "global_step": 266315, "epoch": 2992, "lr": 8.856149199543695e-08}
{"train_loss": 0.02407153509557247, "global_step": 266316, "epoch": 2992, "lr": 8.852700554172755e-08}
{"train_loss": 0.06074531376361847, "global_step": 266317, "epoch": 2992, "lr": 8.849252579801181e-08}
{"train_loss": 0.01718420907855034, "global_step": 266318, "epoch": 2992, "lr": 8.845805276428421e-08}
{"train_loss": 0.052923865616321564, "global_step": 266319, "epoch": 2992, "lr": 8.842358644055582e-08}
{"train_loss": 0.05351388081908226, "global_step": 266320, "epoch": 2992, "lr": 8.83891268268322e-08}
{"train_loss": 0.04315069317817688, "global_step": 266321, "epoch": 2992, "lr": 8.835467392311892e-08}
{"train_loss": 0.04624639451503754, "global_step": 266322, "epoch": 2992, "lr": 8.832022772941595e-08}
{"train_loss": 0.07077408581972122, "global_step": 266323, "epoch": 2992, "lr": 8.828578824572886e-08}
{"train_loss": 0.04023948684334755, "global_step": 266324, "epoch": 2992, "lr": 8.82513554720632e-08}
{"train_loss": 0.03286752477288246, "global_step": 266325, "epoch": 2992, "lr": 8.821692940842453e-08}
{"train_loss": 0.03727901726961136, "global_step": 266326, "epoch": 2992, "lr": 8.818251005481837e-08}
{"train_loss": 0.039799127727746964, "global_step": 266327, "epoch": 2992, "lr": 8.814809741124475e-08}
{"train_loss": 0.05194249376654625, "global_step": 266328, "epoch": 2992, "lr": 8.81136914777092e-08}
{"train_loss": 0.09459668397903442, "global_step": 266329, "epoch": 2992, "lr": 8.807929225422285e-08}
{"train_loss": 0.034173931926488876, "global_step": 266330, "epoch": 2992, "lr": 8.804489974078567e-08}
{"train_loss": 0.04969793185591698, "global_step": 266331, "epoch": 2992, "lr": 8.801051393740322e-08}
{"train_loss": 0.024243419989943504, "global_step": 266332, "epoch": 2992, "lr": 8.797613484407552e-08}
{"train_loss": 0.025438863784074783, "global_step": 266333, "epoch": 2992, "lr": 8.794176246081364e-08}
{"train_loss": 0.04683182016015053, "global_step": 266334, "epoch": 2992, "lr": 8.790739678761761e-08}
{"train_loss": 0.03125009313225746, "global_step": 266335, "epoch": 2992, "lr": 8.787303782449852e-08}
{"train_loss": 0.04402076080441475, "global_step": 266336, "epoch": 2992, "lr": 8.783868557145636e-08}
{"train_loss": 0.031211668625473976, "global_step": 266337, "epoch": 2992, "lr": 8.780434002849114e-08}
{"train_loss": 0.036119841039180756, "global_step": 266338, "epoch": 2992, "lr": 8.777000119561396e-08}
{"train_loss": 0.09852626919746399, "global_step": 266339, "epoch": 2992, "lr": 8.773566907283037e-08}
{"train_loss": 0.0900009423494339, "global_step": 266340, "epoch": 2992, "lr": 8.770134366014038e-08}
{"train_loss": 0.0690181702375412, "global_step": 266341, "epoch": 2992, "lr": 8.766702495754953e-08}
{"train_loss": 0.06246349588036537, "global_step": 266342, "epoch": 2992, "lr": 8.763271296506336e-08}
{"train_loss": 0.049872491508722305, "global_step": 266343, "epoch": 2992, "lr": 8.759840768269301e-08}
{"train_loss": 0.0787300243973732, "global_step": 266344, "epoch": 2992, "lr": 8.756410911042734e-08}
{"train_loss": 0.06975551694631577, "global_step": 266345, "epoch": 2992, "lr": 8.752981724828857e-08}
{"train_loss": 0.02243148721754551, "global_step": 266346, "epoch": 2992, "lr": 8.74955320962656e-08}
{"train_loss": 0.022723069414496422, "global_step": 266347, "epoch": 2992, "lr": 8.746125365437507e-08}
{"train_loss": 0.05655293166637421, "global_step": 266348, "epoch": 2992, "lr": 8.7426981922617e-08}
{"train_loss": 0.03447781503200531, "global_step": 266349, "epoch": 2992, "lr": 8.739271690099694e-08}
{"train_loss": 0.027673855423927307, "global_step": 266350, "epoch": 2992, "lr": 8.735845858951485e-08}
{"train_loss": 0.0569831058382988, "global_step": 266351, "epoch": 2992, "lr": 8.732420698818188e-08}
{"train_loss": 0.03473940119147301, "global_step": 266352, "epoch": 2992, "lr": 8.728996209699802e-08}
{"train_loss": 0.018262626603245735, "global_step": 266353, "epoch": 2992, "lr": 8.725572391596881e-08}
{"train_loss": 0.06517133861780167, "global_step": 266354, "epoch": 2992, "lr": 8.722149244509981e-08}
{"train_loss": 0.02098136395215988, "global_step": 266355, "epoch": 2992, "lr": 8.71872676844021e-08}
{"train_loss": 0.035045064985752106, "global_step": 266356, "epoch": 2992, "lr": 8.715304963386462e-08}
{"train_loss": 0.09830577671527863, "global_step": 266357, "epoch": 2992, "lr": 8.711883829350952e-08}
{"train_loss": 0.014202514663338661, "global_step": 266358, "epoch": 2992, "lr": 8.708463366332575e-08}
{"train_loss": 0.03741030767560005, "global_step": 266359, "epoch": 2992, "lr": 8.705043574332994e-08}
{"train_loss": 0.04709957540035248, "global_step": 266360, "epoch": 2992, "lr": 8.701624453352209e-08}
{"train_loss": 0.0469912625849247, "global_step": 266361, "epoch": 2992, "lr": 8.698206003390219e-08}
{"train_loss": 0.022085046395659447, "global_step": 266362, "epoch": 2992, "lr": 8.694788224448136e-08}
{"train_loss": 0.03686661645770073, "global_step": 266363, "epoch": 2992, "lr": 8.691371116525959e-08}
{"train_loss": 0.04294516518712044, "global_step": 266364, "epoch": 2992, "lr": 8.687954679624799e-08}
{"train_loss": 0.03302663937211037, "global_step": 266365, "epoch": 2992, "lr": 8.684538913744656e-08}
{"train_loss": 0.07836336642503738, "global_step": 266366, "epoch": 2992, "lr": 8.681123818886083e-08}
{"train_loss": 0.0838080421090126, "global_step": 266367, "epoch": 2992, "lr": 8.677709395049083e-08}
{"train_loss": 0.05195238068699837, "global_step": 266368, "epoch": 2992, "lr": 8.674295642234764e-08}
{"train_loss": 0.02146921679377556, "global_step": 266369, "epoch": 2992, "lr": 8.670882560443128e-08}
{"train_loss": 0.013660842552781105, "global_step": 266370, "epoch": 2992, "lr": 8.667470149675284e-08}
{"train_loss": 0.04654654860496521, "global_step": 266371, "epoch": 2992, "lr": 8.664058409931231e-08}
{"train_loss": 0.0538303516805172, "global_step": 266372, "epoch": 2992, "lr": 8.66064734121097e-08}
{"train_loss": 0.027845287695527077, "global_step": 266373, "epoch": 2992, "lr": 8.657236943516167e-08}
{"train_loss": 0.026640085503458977, "global_step": 266374, "epoch": 2992, "lr": 8.653827216845712e-08}
{"train_loss": 0.03375312313437462, "global_step": 266375, "epoch": 2992, "lr": 8.650418161201823e-08}
{"train_loss": 0.0444253667707691, "global_step": 266376, "epoch": 2992, "lr": 8.647009776583392e-08, "val_loss": 9.653145790100098}
{"train_loss": 0.03150704875588417, "global_step": 266377, "epoch": 2993, "lr": 8.643602062992084e-08}
{"train_loss": 0.056759826838970184, "global_step": 266378, "epoch": 2993, "lr": 8.640195020427344e-08}
{"train_loss": 0.0477762296795845, "global_step": 266379, "epoch": 2993, "lr": 8.636788648890282e-08}
{"train_loss": 0.024693233892321587, "global_step": 266380, "epoch": 2993, "lr": 8.633382948380897e-08}
{"train_loss": 0.03971724584698677, "global_step": 266381, "epoch": 2993, "lr": 8.6299779189003e-08}
{"train_loss": 0.037520632147789, "global_step": 266382, "epoch": 2993, "lr": 8.626573560448492e-08}
{"train_loss": 0.05494629591703415, "global_step": 266383, "epoch": 2993, "lr": 8.623169873025472e-08}
{"train_loss": 0.034749675542116165, "global_step": 266384, "epoch": 2993, "lr": 8.619766856632905e-08}
{"train_loss": 0.01133696734905243, "global_step": 266385, "epoch": 2993, "lr": 8.616364511270237e-08}
{"train_loss": 0.0521119199693203, "global_step": 266386, "epoch": 2993, "lr": 8.612962836938577e-08}
{"train_loss": 0.04118339717388153, "global_step": 266387, "epoch": 2993, "lr": 8.609561833637925e-08}
{"train_loss": 0.0628608763217926, "global_step": 266388, "epoch": 2993, "lr": 8.606161501368837e-08}
{"train_loss": 0.08571887761354446, "global_step": 266389, "epoch": 2993, "lr": 8.602761840131313e-08}
{"train_loss": 0.08303776383399963, "global_step": 266390, "epoch": 2993, "lr": 8.599362849927017e-08}
{"train_loss": 0.0685630515217781, "global_step": 266391, "epoch": 2993, "lr": 8.595964530755396e-08}
{"train_loss": 0.034536782652139664, "global_step": 266392, "epoch": 2993, "lr": 8.592566882617559e-08}
{"train_loss": 0.02088802307844162, "global_step": 266393, "epoch": 2993, "lr": 8.589169905513506e-08}
{"train_loss": 0.04289254918694496, "global_step": 266394, "epoch": 2993, "lr": 8.585773599443237e-08}
{"train_loss": 0.04544387757778168, "global_step": 266395, "epoch": 2993, "lr": 8.582377964408416e-08}
{"train_loss": 0.039282508194446564, "global_step": 266396, "epoch": 2993, "lr": 8.578983000408492e-08}
{"train_loss": 0.02985917031764984, "global_step": 266397, "epoch": 2993, "lr": 8.57558870744457e-08}
{"train_loss": 0.06774831563234329, "global_step": 266398, "epoch": 2993, "lr": 8.572195085516655e-08}
{"train_loss": 0.04809725657105446, "global_step": 266399, "epoch": 2993, "lr": 8.568802134625298e-08}
{"train_loss": 0.034532852470874786, "global_step": 266400, "epoch": 2993, "lr": 8.565409854771055e-08}
{"train_loss": 0.017026714980602264, "global_step": 266401, "epoch": 2993, "lr": 8.562018245954484e-08}
{"train_loss": 0.02140475995838642, "global_step": 266402, "epoch": 2993, "lr": 8.558627308175581e-08}
{"train_loss": 0.026593362912535667, "global_step": 266403, "epoch": 2993, "lr": 8.55523704143546e-08}
{"train_loss": 0.02131265215575695, "global_step": 266404, "epoch": 2993, "lr": 8.551847445734118e-08}
{"train_loss": 0.0356837660074234, "global_step": 266405, "epoch": 2993, "lr": 8.548458521072666e-08}
{"train_loss": 0.026245195418596268, "global_step": 266406, "epoch": 2993, "lr": 8.545070267449995e-08}
{"train_loss": 0.06365618854761124, "global_step": 266407, "epoch": 2993, "lr": 8.541682684868324e-08}
{"train_loss": 0.03269645944237709, "global_step": 266408, "epoch": 2993, "lr": 8.538295773327654e-08}
{"train_loss": 0.06083143875002861, "global_step": 266409, "epoch": 2993, "lr": 8.534909532827428e-08}
{"train_loss": 0.05172570049762726, "global_step": 266410, "epoch": 2993, "lr": 8.531523963369314e-08}
{"train_loss": 0.03537122905254364, "global_step": 266411, "epoch": 2993, "lr": 8.528139064953311e-08}
{"train_loss": 0.04186417907476425, "global_step": 266412, "epoch": 2993, "lr": 8.524754837579419e-08}
{"train_loss": 0.11199073493480682, "global_step": 266413, "epoch": 2993, "lr": 8.521371281248746e-08}
{"train_loss": 0.05036718025803566, "global_step": 266414, "epoch": 2993, "lr": 8.517988395961851e-08}
{"train_loss": 0.05344256013631821, "global_step": 266415, "epoch": 2993, "lr": 8.514606181718731e-08}
{"train_loss": 0.03845130652189255, "global_step": 266416, "epoch": 2993, "lr": 8.511224638519944e-08}
{"train_loss": 0.013263721019029617, "global_step": 266417, "epoch": 2993, "lr": 8.507843766366041e-08}
{"train_loss": 0.04629592224955559, "global_step": 266418, "epoch": 2993, "lr": 8.504463565257025e-08}
{"train_loss": 0.020554954186081886, "global_step": 266419, "epoch": 2993, "lr": 8.501084035194007e-08}
{"train_loss": 0.02798488549888134, "global_step": 266420, "epoch": 2993, "lr": 8.49770517617754e-08}
{"train_loss": 0.08481629937887192, "global_step": 266421, "epoch": 2993, "lr": 8.494326988207068e-08}
{"train_loss": 0.03595232591032982, "global_step": 266422, "epoch": 2993, "lr": 8.49094947128426e-08}
{"train_loss": 0.05732899531722069, "global_step": 266423, "epoch": 2993, "lr": 8.487572625408557e-08}
{"train_loss": 0.03981184959411621, "global_step": 266424, "epoch": 2993, "lr": 8.484196450581073e-08}
{"train_loss": 0.05136728286743164, "global_step": 266425, "epoch": 2993, "lr": 8.480820946802359e-08}
{"train_loss": 0.06617983430624008, "global_step": 266426, "epoch": 2993, "lr": 8.477446114071864e-08}
{"train_loss": 0.0461515411734581, "global_step": 266427, "epoch": 2993, "lr": 8.474071952391249e-08}
{"train_loss": 0.023402784019708633, "global_step": 266428, "epoch": 2993, "lr": 8.470698461760518e-08}
{"train_loss": 0.03649793192744255, "global_step": 266429, "epoch": 2993, "lr": 8.467325642179669e-08}
{"train_loss": 0.09591421484947205, "global_step": 266430, "epoch": 2993, "lr": 8.463953493649813e-08}
{"train_loss": 0.06998499482870102, "global_step": 266431, "epoch": 2993, "lr": 8.460582016171503e-08}
{"train_loss": 0.030840644612908363, "global_step": 266432, "epoch": 2993, "lr": 8.457211209744187e-08}
{"train_loss": 0.030645160004496574, "global_step": 266433, "epoch": 2993, "lr": 8.453841074368973e-08}
{"train_loss": 0.062420397996902466, "global_step": 266434, "epoch": 2993, "lr": 8.450471610046417e-08}
{"train_loss": 0.046393707394599915, "global_step": 266435, "epoch": 2993, "lr": 8.447102816777076e-08}
{"train_loss": 0.086695097386837, "global_step": 266436, "epoch": 2993, "lr": 8.4437346945615e-08}
{"train_loss": 0.031990669667720795, "global_step": 266437, "epoch": 2993, "lr": 8.440367243399139e-08}
{"train_loss": 0.07560723274946213, "global_step": 266438, "epoch": 2993, "lr": 8.437000463291655e-08}
{"train_loss": 0.022044679149985313, "global_step": 266439, "epoch": 2993, "lr": 8.433634354238496e-08}
{"train_loss": 0.03231354430317879, "global_step": 266440, "epoch": 2993, "lr": 8.430268916240768e-08}
{"train_loss": 0.08684100955724716, "global_step": 266441, "epoch": 2993, "lr": 8.42690414929903e-08}
{"train_loss": 0.03990926221013069, "global_step": 266442, "epoch": 2993, "lr": 8.423540053412726e-08}
{"train_loss": 0.030080176889896393, "global_step": 266443, "epoch": 2993, "lr": 8.42017662858352e-08}
{"train_loss": 0.011356960982084274, "global_step": 266444, "epoch": 2993, "lr": 8.416813874811414e-08}
{"train_loss": 0.09290304034948349, "global_step": 266445, "epoch": 2993, "lr": 8.41345179209696e-08}
{"train_loss": 0.04782654345035553, "global_step": 266446, "epoch": 2993, "lr": 8.41009038044016e-08}
{"train_loss": 0.041604846715927124, "global_step": 266447, "epoch": 2993, "lr": 8.406729639842125e-08}
{"train_loss": 0.047118235379457474, "global_step": 266448, "epoch": 2993, "lr": 8.403369570302855e-08}
{"train_loss": 0.05144667997956276, "global_step": 266449, "epoch": 2993, "lr": 8.400010171822903e-08}
{"train_loss": 0.04924749583005905, "global_step": 266450, "epoch": 2993, "lr": 8.396651444402826e-08}
{"train_loss": 0.03371261805295944, "global_step": 266451, "epoch": 2993, "lr": 8.393293388042622e-08}
{"train_loss": 0.07523150742053986, "global_step": 266452, "epoch": 2993, "lr": 8.389936002743403e-08}
{"train_loss": 0.033017247915267944, "global_step": 266453, "epoch": 2993, "lr": 8.38657928850517e-08}
{"train_loss": 0.06131621077656746, "global_step": 266454, "epoch": 2993, "lr": 8.383223245328476e-08}
{"train_loss": 0.040781453251838684, "global_step": 266455, "epoch": 2993, "lr": 8.379867873213876e-08}
{"train_loss": 0.040243491530418396, "global_step": 266456, "epoch": 2993, "lr": 8.376513172161926e-08}
{"train_loss": 0.023882608860731125, "global_step": 266457, "epoch": 2993, "lr": 8.373159142172627e-08}
{"train_loss": 0.04136770963668823, "global_step": 266458, "epoch": 2993, "lr": 8.369805783247087e-08}
{"train_loss": 0.10834848880767822, "global_step": 266459, "epoch": 2993, "lr": 8.366453095385307e-08}
{"train_loss": 0.04600323736667633, "global_step": 266460, "epoch": 2993, "lr": 8.363101078587288e-08}
{"train_loss": 0.03689073026180267, "global_step": 266461, "epoch": 2993, "lr": 8.359749732854693e-08}
{"train_loss": 0.0551447756588459, "global_step": 266462, "epoch": 2993, "lr": 8.35639905818697e-08}
{"train_loss": 0.018665598705410957, "global_step": 266463, "epoch": 2993, "lr": 8.353049054584672e-08}
{"train_loss": 0.08931850641965866, "global_step": 266464, "epoch": 2993, "lr": 8.34969972204891e-08}
{"train_loss": 0.04687458381391643, "global_step": 266465, "epoch": 2993, "lr": 8.346351060579682e-08, "val_loss": 9.59692668914795}
{"train_loss": 0.0098689254373312, "global_step": 266466, "epoch": 2994, "lr": 8.343003070176991e-08}
{"train_loss": 0.01188475638628006, "global_step": 266467, "epoch": 2994, "lr": 8.339655750841945e-08}
{"train_loss": 0.06482403725385666, "global_step": 266468, "epoch": 2994, "lr": 8.336309102575102e-08}
{"train_loss": 0.04135555773973465, "global_step": 266469, "epoch": 2994, "lr": 8.332963125376459e-08}
{"train_loss": 0.019802924245595932, "global_step": 266470, "epoch": 2994, "lr": 8.329617819246572e-08}
{"train_loss": 0.021073509007692337, "global_step": 266471, "epoch": 2994, "lr": 8.326273184185996e-08}
{"train_loss": 0.03329421207308769, "global_step": 266472, "epoch": 2994, "lr": 8.32292922019473e-08}
{"train_loss": 0.042563457041978836, "global_step": 266473, "epoch": 2994, "lr": 8.319585927273887e-08}
{"train_loss": 0.028623225167393684, "global_step": 266474, "epoch": 2994, "lr": 8.31624330542402e-08}
{"train_loss": 0.049026861786842346, "global_step": 266475, "epoch": 2994, "lr": 8.312901354644575e-08}
{"train_loss": 0.06591461598873138, "global_step": 266476, "epoch": 2994, "lr": 8.309560074936662e-08}
{"train_loss": 0.015859737992286682, "global_step": 266477, "epoch": 2994, "lr": 8.306219466300835e-08}
{"train_loss": 0.03896283730864525, "global_step": 266478, "epoch": 2994, "lr": 8.30287952873765e-08}
{"train_loss": 0.04891832172870636, "global_step": 266479, "epoch": 2994, "lr": 8.299540262247108e-08}
{"train_loss": 0.02933008037507534, "global_step": 266480, "epoch": 2994, "lr": 8.296201666829762e-08}
{"train_loss": 0.03986256942152977, "global_step": 266481, "epoch": 2994, "lr": 8.292863742486168e-08}
{"train_loss": 0.038261380046606064, "global_step": 266482, "epoch": 2994, "lr": 8.289526489216881e-08}
{"train_loss": 0.10641182214021683, "global_step": 266483, "epoch": 2994, "lr": 8.286189907022457e-08}
{"train_loss": 0.04304661974310875, "global_step": 266484, "epoch": 2994, "lr": 8.28285399590234e-08}
{"train_loss": 0.03433889523148537, "global_step": 266485, "epoch": 2994, "lr": 8.279518755858195e-08}
{"train_loss": 0.07595513015985489, "global_step": 266486, "epoch": 2994, "lr": 8.276184186890023e-08}
{"train_loss": 0.044692717492580414, "global_step": 266487, "epoch": 2994, "lr": 8.27285028899838e-08}
{"train_loss": 0.04978238046169281, "global_step": 266488, "epoch": 2994, "lr": 8.269517062183818e-08}
{"train_loss": 0.018353652209043503, "global_step": 266489, "epoch": 2994, "lr": 8.26618450644634e-08}
{"train_loss": 0.048358090221881866, "global_step": 266490, "epoch": 2994, "lr": 8.262852621786499e-08}
{"train_loss": 0.05019855871796608, "global_step": 266491, "epoch": 2994, "lr": 8.259521408204851e-08}
{"train_loss": 0.05183367431163788, "global_step": 266492, "epoch": 2994, "lr": 8.256190865701952e-08}
{"train_loss": 0.0714101642370224, "global_step": 266493, "epoch": 2994, "lr": 8.252860994278356e-08}
{"train_loss": 0.051447637379169464, "global_step": 266494, "epoch": 2994, "lr": 8.249531793934062e-08}
{"train_loss": 0.0477743037045002, "global_step": 266495, "epoch": 2994, "lr": 8.246203264670183e-08}
{"train_loss": 0.09585605561733246, "global_step": 266496, "epoch": 2994, "lr": 8.242875406486161e-08}
{"train_loss": 0.03995374217629433, "global_step": 266497, "epoch": 2994, "lr": 8.239548219383664e-08}
{"train_loss": 0.038547828793525696, "global_step": 266498, "epoch": 2994, "lr": 8.236221703362134e-08}
{"train_loss": 0.04048362746834755, "global_step": 266499, "epoch": 2994, "lr": 8.23289585842213e-08}
{"train_loss": 0.05029883235692978, "global_step": 266500, "epoch": 2994, "lr": 8.229570684564758e-08}
{"train_loss": 0.0628027394413948, "global_step": 266501, "epoch": 2994, "lr": 8.22624618179002e-08}
{"train_loss": 0.04261505603790283, "global_step": 266502, "epoch": 2994, "lr": 8.222922350098472e-08}
{"train_loss": 0.029504375532269478, "global_step": 266503, "epoch": 2994, "lr": 8.219599189490668e-08}
{"train_loss": 0.03806111216545105, "global_step": 266504, "epoch": 2994, "lr": 8.216276699966608e-08}
{"train_loss": 0.041249603033065796, "global_step": 266505, "epoch": 2994, "lr": 8.212954881527401e-08}
{"train_loss": 0.02857508882880211, "global_step": 266506, "epoch": 2994, "lr": 8.209633734172495e-08}
{"train_loss": 0.03793525695800781, "global_step": 266507, "epoch": 2994, "lr": 8.206313257903554e-08}
{"train_loss": 0.04460619390010834, "global_step": 266508, "epoch": 2994, "lr": 8.20299345272002e-08}
{"train_loss": 0.060534827411174774, "global_step": 266509, "epoch": 2994, "lr": 8.199674318623008e-08}
{"train_loss": 0.027806658297777176, "global_step": 266510, "epoch": 2994, "lr": 8.196355855612514e-08}
{"train_loss": 0.053131502121686935, "global_step": 266511, "epoch": 2994, "lr": 8.193038063689651e-08}
{"train_loss": 0.03357240930199623, "global_step": 266512, "epoch": 2994, "lr": 8.189720942853862e-08}
{"train_loss": 0.06966113299131393, "global_step": 266513, "epoch": 2994, "lr": 8.186404493106258e-08}
{"train_loss": 0.04887145385146141, "global_step": 266514, "epoch": 2994, "lr": 8.183088714447395e-08}
{"train_loss": 0.03656192496418953, "global_step": 266515, "epoch": 2994, "lr": 8.179773606877272e-08}
{"train_loss": 0.02713155373930931, "global_step": 266516, "epoch": 2994, "lr": 8.176459170396444e-08}
{"train_loss": 0.02351006120443344, "global_step": 266517, "epoch": 2994, "lr": 8.173145405005467e-08}
{"train_loss": 0.0530434288084507, "global_step": 266518, "epoch": 2994, "lr": 8.169832310704895e-08}
{"train_loss": 0.026933884248137474, "global_step": 266519, "epoch": 2994, "lr": 8.166519887494728e-08}
{"train_loss": 0.013797068037092686, "global_step": 266520, "epoch": 2994, "lr": 8.163208135376077e-08}
{"train_loss": 0.01914806105196476, "global_step": 266521, "epoch": 2994, "lr": 8.159897054348941e-08}
{"train_loss": 0.043666109442710876, "global_step": 266522, "epoch": 2994, "lr": 8.156586644413877e-08}
{"train_loss": 0.03148745745420456, "global_step": 266523, "epoch": 2994, "lr": 8.153276905571439e-08}
{"train_loss": 0.02910649962723255, "global_step": 266524, "epoch": 2994, "lr": 8.149967837821626e-08}
{"train_loss": 0.044852808117866516, "global_step": 266525, "epoch": 2994, "lr": 8.146659441165549e-08}
{"train_loss": 0.05229746550321579, "global_step": 266526, "epoch": 2994, "lr": 8.143351715603209e-08}
{"train_loss": 0.04715665057301521, "global_step": 266527, "epoch": 2994, "lr": 8.14004466113516e-08}
{"train_loss": 0.032083190977573395, "global_step": 266528, "epoch": 2994, "lr": 8.136738277761958e-08}
{"train_loss": 0.04911554977297783, "global_step": 266529, "epoch": 2994, "lr": 8.1334325654836e-08}
{"train_loss": 0.03115690127015114, "global_step": 266530, "epoch": 2994, "lr": 8.1301275243012e-08}
{"train_loss": 0.08027953654527664, "global_step": 266531, "epoch": 2994, "lr": 8.126823154214758e-08}
{"train_loss": 0.019655080512166023, "global_step": 266532, "epoch": 2994, "lr": 8.12351945522427e-08}
{"train_loss": 0.03905589506030083, "global_step": 266533, "epoch": 2994, "lr": 8.120216427331407e-08}
{"train_loss": 0.03496624901890755, "global_step": 266534, "epoch": 2994, "lr": 8.116914070535608e-08}
{"train_loss": 0.028708940371870995, "global_step": 266535, "epoch": 2994, "lr": 8.113612384837988e-08}
{"train_loss": 0.02788211964070797, "global_step": 266536, "epoch": 2994, "lr": 8.110311370238543e-08}
{"train_loss": 0.051301322877407074, "global_step": 266537, "epoch": 2994, "lr": 8.107011026737832e-08}
{"train_loss": 0.013165966607630253, "global_step": 266538, "epoch": 2994, "lr": 8.103711354335852e-08}
{"train_loss": 0.03520432114601135, "global_step": 266539, "epoch": 2994, "lr": 8.100412353033715e-08}
{"train_loss": 0.053696323186159134, "global_step": 266540, "epoch": 2994, "lr": 8.097114022831975e-08}
{"train_loss": 0.013040299527347088, "global_step": 266541, "epoch": 2994, "lr": 8.093816363730633e-08}
{"train_loss": 0.042016420513391495, "global_step": 266542, "epoch": 2994, "lr": 8.090519375730243e-08}
{"train_loss": 0.022896558046340942, "global_step": 266543, "epoch": 2994, "lr": 8.087223058830807e-08}
{"train_loss": 0.03129133582115173, "global_step": 266544, "epoch": 2994, "lr": 8.083927413033432e-08}
{"train_loss": 0.028624599799513817, "global_step": 266545, "epoch": 2994, "lr": 8.080632438338675e-08}
{"train_loss": 0.017922965809702873, "global_step": 266546, "epoch": 2994, "lr": 8.077338134745982e-08}
{"train_loss": 0.03543311730027199, "global_step": 266547, "epoch": 2994, "lr": 8.074044502257017e-08}
{"train_loss": 0.049623165279626846, "global_step": 266548, "epoch": 2994, "lr": 8.070751540871779e-08}
{"train_loss": 0.08946024626493454, "global_step": 266549, "epoch": 2994, "lr": 8.06745925059027e-08}
{"train_loss": 0.062393564730882645, "global_step": 266550, "epoch": 2994, "lr": 8.064167631413044e-08}
{"train_loss": 0.04400055482983589, "global_step": 266551, "epoch": 2994, "lr": 8.060876683341212e-08}
{"train_loss": 0.06164829805493355, "global_step": 266552, "epoch": 2994, "lr": 8.057586406374218e-08}
{"train_loss": 0.06628484278917313, "global_step": 266553, "epoch": 2994, "lr": 8.054296800513172e-08}
{"train_loss": 0.042148292828560543, "global_step": 266554, "epoch": 2994, "lr": 8.05100786575863e-08, "val_loss": 9.670641899108887}
{"train_loss": 0.05420735105872154, "global_step": 266555, "epoch": 2995, "lr": 8.047719602110593e-08}
{"train_loss": 0.037621766328811646, "global_step": 266556, "epoch": 2995, "lr": 8.044432009569614e-08}
{"train_loss": 0.0353381372988224, "global_step": 266557, "epoch": 2995, "lr": 8.041145088136803e-08}
{"train_loss": 0.04384525120258331, "global_step": 266558, "epoch": 2995, "lr": 8.037858837811607e-08}
{"train_loss": 0.06604153662919998, "global_step": 266559, "epoch": 2995, "lr": 8.034573258594579e-08}
{"train_loss": 0.042430490255355835, "global_step": 266560, "epoch": 2995, "lr": 8.031288350486832e-08}
{"train_loss": 0.016664044931530952, "global_step": 266561, "epoch": 2995, "lr": 8.028004113488364e-08}
{"train_loss": 0.06960064172744751, "global_step": 266562, "epoch": 2995, "lr": 8.024720547599729e-08}
{"train_loss": 0.031641099601984024, "global_step": 266563, "epoch": 2995, "lr": 8.021437652821483e-08}
{"train_loss": 0.043502189218997955, "global_step": 266564, "epoch": 2995, "lr": 8.018155429153629e-08}
{"train_loss": 0.04800908640027046, "global_step": 266565, "epoch": 2995, "lr": 8.014873876596718e-08}
{"train_loss": 0.044417351484298706, "global_step": 266566, "epoch": 2995, "lr": 8.011592995151863e-08}
{"train_loss": 0.055454596877098083, "global_step": 266567, "epoch": 2995, "lr": 8.008312784818506e-08}
{"train_loss": 0.026661258190870285, "global_step": 266568, "epoch": 2995, "lr": 8.00503324559776e-08}
{"train_loss": 0.04037974774837494, "global_step": 266569, "epoch": 2995, "lr": 8.001754377490178e-08}
{"train_loss": 0.035155072808265686, "global_step": 266570, "epoch": 2995, "lr": 7.998476180495762e-08}
{"train_loss": 0.03348285332322121, "global_step": 266571, "epoch": 2995, "lr": 7.99519865461451e-08}
{"train_loss": 0.04083459824323654, "global_step": 266572, "epoch": 2995, "lr": 7.99192179984809e-08}
{"train_loss": 0.024592336267232895, "global_step": 266573, "epoch": 2995, "lr": 7.9886456161965e-08}
{"train_loss": 0.05207858607172966, "global_step": 266574, "epoch": 2995, "lr": 7.985370103659739e-08}
{"train_loss": 0.10592588782310486, "global_step": 266575, "epoch": 2995, "lr": 7.982095262238365e-08}
{"train_loss": 0.01904534175992012, "global_step": 266576, "epoch": 2995, "lr": 7.978821091932931e-08}
{"train_loss": 0.036063358187675476, "global_step": 266577, "epoch": 2995, "lr": 7.975547592743992e-08}
{"train_loss": 0.04578089714050293, "global_step": 266578, "epoch": 2995, "lr": 7.972274764672105e-08}
{"train_loss": 0.029252858832478523, "global_step": 266579, "epoch": 2995, "lr": 7.969002607717269e-08}
{"train_loss": 0.03135475516319275, "global_step": 266580, "epoch": 2995, "lr": 7.965731121880038e-08}
{"train_loss": 0.025771548971533775, "global_step": 266581, "epoch": 2995, "lr": 7.962460307161524e-08}
{"train_loss": 0.05849651247262955, "global_step": 266582, "epoch": 2995, "lr": 7.95919016356117e-08}
{"train_loss": 0.10682601481676102, "global_step": 266583, "epoch": 2995, "lr": 7.955920691079532e-08}
{"train_loss": 0.05239773541688919, "global_step": 266584, "epoch": 2995, "lr": 7.952651889718277e-08}
{"train_loss": 0.017909323796629906, "global_step": 266585, "epoch": 2995, "lr": 7.949383759476293e-08}
{"train_loss": 0.030782561749219894, "global_step": 266586, "epoch": 2995, "lr": 7.946116300354689e-08}
{"train_loss": 0.04978083819150925, "global_step": 266587, "epoch": 2995, "lr": 7.942849512354023e-08}
{"train_loss": 0.04592670127749443, "global_step": 266588, "epoch": 2995, "lr": 7.939583395474848e-08}
{"train_loss": 0.02238999679684639, "global_step": 266589, "epoch": 2995, "lr": 7.936317949717165e-08}
{"train_loss": 0.02311970293521881, "global_step": 266590, "epoch": 2995, "lr": 7.933053175081529e-08}
{"train_loss": 0.02076798863708973, "global_step": 266591, "epoch": 2995, "lr": 7.929789071568494e-08}
{"train_loss": 0.035658758133649826, "global_step": 266592, "epoch": 2995, "lr": 7.926525639178062e-08}
{"train_loss": 0.0273145642131567, "global_step": 266593, "epoch": 2995, "lr": 7.923262877911342e-08}
{"train_loss": 0.1126171201467514, "global_step": 266594, "epoch": 2995, "lr": 7.920000787768889e-08}
{"train_loss": 0.02929283306002617, "global_step": 266595, "epoch": 2995, "lr": 7.916739368750148e-08}
{"train_loss": 0.07718931883573532, "global_step": 266596, "epoch": 2995, "lr": 7.91347862085623e-08}
{"train_loss": 0.03083670698106289, "global_step": 266597, "epoch": 2995, "lr": 7.910218544087689e-08}
{"train_loss": 0.04534400999546051, "global_step": 266598, "epoch": 2995, "lr": 7.90695913844508e-08}
{"train_loss": 0.02491372637450695, "global_step": 266599, "epoch": 2995, "lr": 7.903700403928405e-08}
{"train_loss": 0.03606846556067467, "global_step": 266600, "epoch": 2995, "lr": 7.900442340537661e-08}
{"train_loss": 0.08326222747564316, "global_step": 266601, "epoch": 2995, "lr": 7.897184948274517e-08}
{"train_loss": 0.06368188560009003, "global_step": 266602, "epoch": 2995, "lr": 7.893928227138414e-08}
{"train_loss": 0.045025311410427094, "global_step": 266603, "epoch": 2995, "lr": 7.89067217712991e-08}
{"train_loss": 0.044708698987960815, "global_step": 266604, "epoch": 2995, "lr": 7.887416798250113e-08}
{"train_loss": 0.012754233554005623, "global_step": 266605, "epoch": 2995, "lr": 7.884162090499026e-08}
{"train_loss": 0.023164192214608192, "global_step": 266606, "epoch": 2995, "lr": 7.880908053876645e-08}
{"train_loss": 0.02776353806257248, "global_step": 266607, "epoch": 2995, "lr": 7.877654688384084e-08}
{"train_loss": 0.0719282329082489, "global_step": 266608, "epoch": 2995, "lr": 7.874401994021896e-08}
{"train_loss": 0.06661393493413925, "global_step": 266609, "epoch": 2995, "lr": 7.871149970789526e-08}
{"train_loss": 0.029040127992630005, "global_step": 266610, "epoch": 2995, "lr": 7.867898618688085e-08}
{"train_loss": 0.0394805446267128, "global_step": 266611, "epoch": 2995, "lr": 7.864647937718128e-08}
{"train_loss": 0.038991428911685944, "global_step": 266612, "epoch": 2995, "lr": 7.86139792788021e-08}
{"train_loss": 0.030141351744532585, "global_step": 266613, "epoch": 2995, "lr": 7.85814858917433e-08}
{"train_loss": 0.020046086981892586, "global_step": 266614, "epoch": 2995, "lr": 7.854899921601044e-08}
{"train_loss": 0.09417387843132019, "global_step": 266615, "epoch": 2995, "lr": 7.851651925160353e-08}
{"train_loss": 0.042359210550785065, "global_step": 266616, "epoch": 2995, "lr": 7.84840459985392e-08}
{"train_loss": 0.05138731747865677, "global_step": 266617, "epoch": 2995, "lr": 7.845157945680637e-08}
{"train_loss": 0.0054663410410285, "global_step": 266618, "epoch": 2995, "lr": 7.841911962642167e-08}
{"train_loss": 0.03994106873869896, "global_step": 266619, "epoch": 2995, "lr": 7.838666650738513e-08}
{"train_loss": 0.023491356521844864, "global_step": 266620, "epoch": 2995, "lr": 7.835422009970228e-08}
{"train_loss": 0.05353269726037979, "global_step": 266621, "epoch": 2995, "lr": 7.832178040337312e-08}
{"train_loss": 0.080345019698143, "global_step": 266622, "epoch": 2995, "lr": 7.828934741840877e-08}
{"train_loss": 0.05544406920671463, "global_step": 266623, "epoch": 2995, "lr": 7.82569211448092e-08}
{"train_loss": 0.06286745518445969, "global_step": 266624, "epoch": 2995, "lr": 7.822450158257444e-08}
{"train_loss": 0.06362853199243546, "global_step": 266625, "epoch": 2995, "lr": 7.819208873172113e-08}
{"train_loss": 0.038173843175172806, "global_step": 266626, "epoch": 2995, "lr": 7.815968259224371e-08}
{"train_loss": 0.025914747267961502, "global_step": 266627, "epoch": 2995, "lr": 7.812728316414775e-08}
{"train_loss": 0.06580612063407898, "global_step": 266628, "epoch": 2995, "lr": 7.809489044743879e-08}
{"train_loss": 0.030056288465857506, "global_step": 266629, "epoch": 2995, "lr": 7.806250444212238e-08}
{"train_loss": 0.027712378650903702, "global_step": 266630, "epoch": 2995, "lr": 7.803012514820407e-08}
{"train_loss": 0.08007387816905975, "global_step": 266631, "epoch": 2995, "lr": 7.799775256568387e-08}
{"train_loss": 0.044956859201192856, "global_step": 266632, "epoch": 2995, "lr": 7.796538669456732e-08}
{"train_loss": 0.052481845021247864, "global_step": 266633, "epoch": 2995, "lr": 7.793302753485998e-08}
{"train_loss": 0.017995890229940414, "global_step": 266634, "epoch": 2995, "lr": 7.790067508656739e-08}
{"train_loss": 0.023254986852407455, "global_step": 266635, "epoch": 2995, "lr": 7.786832934969513e-08}
{"train_loss": 0.059644997119903564, "global_step": 266636, "epoch": 2995, "lr": 7.783599032424315e-08}
{"train_loss": 0.04069919139146805, "global_step": 266637, "epoch": 2995, "lr": 7.780365801021705e-08}
{"train_loss": 0.04773644730448723, "global_step": 266638, "epoch": 2995, "lr": 7.777133240762235e-08}
{"train_loss": 0.051281802356243134, "global_step": 266639, "epoch": 2995, "lr": 7.773901351645907e-08}
{"train_loss": 0.03435501083731651, "global_step": 266640, "epoch": 2995, "lr": 7.77067013367383e-08}
{"train_loss": 0.037396389991045, "global_step": 266641, "epoch": 2995, "lr": 7.767439586846558e-08}
{"train_loss": 0.06268797069787979, "global_step": 266642, "epoch": 2995, "lr": 7.764209711163539e-08}
{"train_loss": 0.04418939194024614, "global_step": 266643, "epoch": 2995, "lr": 7.760980506625881e-08, "val_loss": 9.621644020080566, "train_action_mse_error": 19.57741355895996}
{"train_loss": 0.046373460441827774, "global_step": 266644, "epoch": 2996, "lr": 7.75775197323414e-08}
{"train_loss": 0.039698511362075806, "global_step": 266645, "epoch": 2996, "lr": 7.754524110988315e-08}
{"train_loss": 0.02770429290831089, "global_step": 266646, "epoch": 2996, "lr": 7.751296919888962e-08}
{"train_loss": 0.07318031787872314, "global_step": 266647, "epoch": 2996, "lr": 7.748070399936636e-08}
{"train_loss": 0.030492380261421204, "global_step": 266648, "epoch": 2996, "lr": 7.744844551131891e-08}
{"train_loss": 0.03764650970697403, "global_step": 266649, "epoch": 2996, "lr": 7.741619373475284e-08}
{"train_loss": 0.019823649898171425, "global_step": 266650, "epoch": 2996, "lr": 7.738394866966259e-08}
{"train_loss": 0.01404188759624958, "global_step": 266651, "epoch": 2996, "lr": 7.735171031606481e-08}
{"train_loss": 0.05107221007347107, "global_step": 266652, "epoch": 2996, "lr": 7.73194786739595e-08}
{"train_loss": 0.020061643794178963, "global_step": 266653, "epoch": 2996, "lr": 7.728725374334667e-08}
{"train_loss": 0.05487636849284172, "global_step": 266654, "epoch": 2996, "lr": 7.725503552423186e-08}
{"train_loss": 0.057249732315540314, "global_step": 266655, "epoch": 2996, "lr": 7.722282401662617e-08}
{"train_loss": 0.039478741586208344, "global_step": 266656, "epoch": 2996, "lr": 7.719061922052961e-08}
{"train_loss": 0.0761849656701088, "global_step": 266657, "epoch": 2996, "lr": 7.715842113594218e-08}
{"train_loss": 0.050409939140081406, "global_step": 266658, "epoch": 2996, "lr": 7.712622976288054e-08}
{"train_loss": 0.03213715925812721, "global_step": 266659, "epoch": 2996, "lr": 7.709404510133356e-08}
{"train_loss": 0.03262808173894882, "global_step": 266660, "epoch": 2996, "lr": 7.706186715131236e-08}
{"train_loss": 0.05444091930985451, "global_step": 266661, "epoch": 2996, "lr": 7.70296959128225e-08}
{"train_loss": 0.05426233634352684, "global_step": 266662, "epoch": 2996, "lr": 7.699753138586952e-08}
{"train_loss": 0.028915541246533394, "global_step": 266663, "epoch": 2996, "lr": 7.696537357045896e-08}
{"train_loss": 0.046746183186769485, "global_step": 266664, "epoch": 2996, "lr": 7.69332224665853e-08}
{"train_loss": 0.06530002504587173, "global_step": 266665, "epoch": 2996, "lr": 7.690107807426516e-08}
{"train_loss": 0.07924098521471024, "global_step": 266666, "epoch": 2996, "lr": 7.686894039349301e-08}
{"train_loss": 0.06932632625102997, "global_step": 266667, "epoch": 2996, "lr": 7.683680942427995e-08}
{"train_loss": 0.03810776397585869, "global_step": 266668, "epoch": 2996, "lr": 7.680468516662598e-08}
{"train_loss": 0.05512212589383125, "global_step": 266669, "epoch": 2996, "lr": 7.677256762053664e-08}
{"train_loss": 0.03659198060631752, "global_step": 266670, "epoch": 2996, "lr": 7.67404567860175e-08}
{"train_loss": 0.03238030895590782, "global_step": 266671, "epoch": 2996, "lr": 7.670835266307407e-08}
{"train_loss": 0.03294290974736214, "global_step": 266672, "epoch": 2996, "lr": 7.66762552517064e-08}
{"train_loss": 0.05501218140125275, "global_step": 266673, "epoch": 2996, "lr": 7.664416455192003e-08}
{"train_loss": 0.03404620662331581, "global_step": 266674, "epoch": 2996, "lr": 7.661208056372049e-08}
{"train_loss": 0.03302908316254616, "global_step": 266675, "epoch": 2996, "lr": 7.658000328711889e-08}
{"train_loss": 0.03579607605934143, "global_step": 266676, "epoch": 2996, "lr": 7.654793272210415e-08}
{"train_loss": 0.04229248687624931, "global_step": 266677, "epoch": 2996, "lr": 7.651586886869288e-08}
{"train_loss": 0.043876174837350845, "global_step": 266678, "epoch": 2996, "lr": 7.648381172688512e-08}
{"train_loss": 0.01979185827076435, "global_step": 266679, "epoch": 2996, "lr": 7.645176129668641e-08}
{"train_loss": 0.0306592658162117, "global_step": 266680, "epoch": 2996, "lr": 7.641971757809674e-08}
{"train_loss": 0.0632796436548233, "global_step": 266681, "epoch": 2996, "lr": 7.638768057112722e-08}
{"train_loss": 0.04774646833539009, "global_step": 266682, "epoch": 2996, "lr": 7.635565027577785e-08}
{"train_loss": 0.06195712089538574, "global_step": 266683, "epoch": 2996, "lr": 7.632362669205417e-08}
{"train_loss": 0.06966877728700638, "global_step": 266684, "epoch": 2996, "lr": 7.629160981996175e-08}
{"train_loss": 0.053101807832717896, "global_step": 266685, "epoch": 2996, "lr": 7.625959965950058e-08}
{"train_loss": 0.019638270139694214, "global_step": 266686, "epoch": 2996, "lr": 7.622759621068177e-08}
{"train_loss": 0.045282796025276184, "global_step": 266687, "epoch": 2996, "lr": 7.619559947349975e-08}
{"train_loss": 0.06799285858869553, "global_step": 266688, "epoch": 2996, "lr": 7.616360944797118e-08}
{"train_loss": 0.04526054114103317, "global_step": 266689, "epoch": 2996, "lr": 7.613162613409053e-08}
{"train_loss": 0.05945521593093872, "global_step": 266690, "epoch": 2996, "lr": 7.609964953186333e-08}
{"train_loss": 0.01826884038746357, "global_step": 266691, "epoch": 2996, "lr": 7.606767964130068e-08}
{"train_loss": 0.03894694894552231, "global_step": 266692, "epoch": 2996, "lr": 7.60357164624026e-08}
{"train_loss": 0.03555090352892876, "global_step": 266693, "epoch": 2996, "lr": 7.600375999516907e-08}
{"train_loss": 0.0683755949139595, "global_step": 266694, "epoch": 2996, "lr": 7.59718102396112e-08}
{"train_loss": 0.056989118456840515, "global_step": 266695, "epoch": 2996, "lr": 7.593986719572344e-08}
{"train_loss": 0.03012477234005928, "global_step": 266696, "epoch": 2996, "lr": 7.590793086352799e-08}
{"train_loss": 0.03830903768539429, "global_step": 266697, "epoch": 2996, "lr": 7.58760012430082e-08}
{"train_loss": 0.04216626659035683, "global_step": 266698, "epoch": 2996, "lr": 7.584407833418628e-08}
{"train_loss": 0.008533748798072338, "global_step": 266699, "epoch": 2996, "lr": 7.581216213705666e-08}
{"train_loss": 0.041109826415777206, "global_step": 266700, "epoch": 2996, "lr": 7.578025265162492e-08}
{"train_loss": 0.033517319709062576, "global_step": 266701, "epoch": 2996, "lr": 7.574834987789658e-08}
{"train_loss": 0.02315329760313034, "global_step": 266702, "epoch": 2996, "lr": 7.571645381587167e-08}
{"train_loss": 0.045790839940309525, "global_step": 266703, "epoch": 2996, "lr": 7.568456446556127e-08}
{"train_loss": 0.05548154562711716, "global_step": 266704, "epoch": 2996, "lr": 7.565268182696539e-08}
{"train_loss": 0.04368296638131142, "global_step": 266705, "epoch": 2996, "lr": 7.562080590008958e-08}
{"train_loss": 0.017521584406495094, "global_step": 266706, "epoch": 2996, "lr": 7.558893668493938e-08}
{"train_loss": 0.05440649390220642, "global_step": 266707, "epoch": 2996, "lr": 7.555707418152036e-08}
{"train_loss": 0.04744623973965645, "global_step": 266708, "epoch": 2996, "lr": 7.552521838982696e-08}
{"train_loss": 0.021237745881080627, "global_step": 266709, "epoch": 2996, "lr": 7.549336930987582e-08}
{"train_loss": 0.03151644393801689, "global_step": 266710, "epoch": 2996, "lr": 7.546152694166697e-08}
{"train_loss": 0.0673767477273941, "global_step": 266711, "epoch": 2996, "lr": 7.542969128520039e-08}
{"train_loss": 0.07475343346595764, "global_step": 266712, "epoch": 2996, "lr": 7.539786234048718e-08}
{"train_loss": 0.021166006103157997, "global_step": 266713, "epoch": 2996, "lr": 7.536604010752735e-08}
{"train_loss": 0.06785617023706436, "global_step": 266714, "epoch": 2996, "lr": 7.533422458632089e-08}
{"train_loss": 0.03663061559200287, "global_step": 266715, "epoch": 2996, "lr": 7.530241577688446e-08}
{"train_loss": 0.07493266463279724, "global_step": 266716, "epoch": 2996, "lr": 7.52706136792125e-08}
{"train_loss": 0.03523829206824303, "global_step": 266717, "epoch": 2996, "lr": 7.523881829331059e-08}
{"train_loss": 0.07656613737344742, "global_step": 266718, "epoch": 2996, "lr": 7.520702961918424e-08}
{"train_loss": 0.0553370863199234, "global_step": 266719, "epoch": 2996, "lr": 7.517524765683904e-08}
{"train_loss": 0.029640784487128258, "global_step": 266720, "epoch": 2996, "lr": 7.51434724062805e-08}
{"train_loss": 0.02152409218251705, "global_step": 266721, "epoch": 2996, "lr": 7.51117038675031e-08}
{"train_loss": 0.04325580596923828, "global_step": 266722, "epoch": 2996, "lr": 7.507994204052904e-08}
{"train_loss": 0.06025824695825577, "global_step": 266723, "epoch": 2996, "lr": 7.504818692534166e-08}
{"train_loss": 0.04004710912704468, "global_step": 266724, "epoch": 2996, "lr": 7.501643852196316e-08}
{"train_loss": 0.029977528378367424, "global_step": 266725, "epoch": 2996, "lr": 7.498469683038245e-08}
{"train_loss": 0.03588739037513733, "global_step": 266726, "epoch": 2996, "lr": 7.495296185062173e-08}
{"train_loss": 0.030271422117948532, "global_step": 266727, "epoch": 2996, "lr": 7.492123358266434e-08}
{"train_loss": 0.06613019108772278, "global_step": 266728, "epoch": 2996, "lr": 7.48895120265325e-08}
{"train_loss": 0.029186028987169266, "global_step": 266729, "epoch": 2996, "lr": 7.485779718222064e-08}
{"train_loss": 0.03785925731062889, "global_step": 266730, "epoch": 2996, "lr": 7.482608904973431e-08}
{"train_loss": 0.03748450428247452, "global_step": 266731, "epoch": 2996, "lr": 7.479438762908464e-08}
{"train_loss": 0.044147530204375805, "global_step": 266732, "epoch": 2996, "lr": 7.476269292026605e-08, "val_loss": 9.667985916137695}
{"train_loss": 0.03853525221347809, "global_step": 266733, "epoch": 2997, "lr": 7.473100492328411e-08}
{"train_loss": 0.04959874972701073, "global_step": 266734, "epoch": 2997, "lr": 7.469932363815546e-08}
{"train_loss": 0.04434404522180557, "global_step": 266735, "epoch": 2997, "lr": 7.466764906486346e-08}
{"train_loss": 0.017009476199746132, "global_step": 266736, "epoch": 2997, "lr": 7.463598120343029e-08}
{"train_loss": 0.03207853436470032, "global_step": 266737, "epoch": 2997, "lr": 7.460432005385598e-08}
{"train_loss": 0.044806092977523804, "global_step": 266738, "epoch": 2997, "lr": 7.457266561614052e-08}
{"train_loss": 0.05388632044196129, "global_step": 266739, "epoch": 2997, "lr": 7.454101789028944e-08}
{"train_loss": 0.02913837321102619, "global_step": 266740, "epoch": 2997, "lr": 7.450937687630832e-08}
{"train_loss": 0.04343106597661972, "global_step": 266741, "epoch": 2997, "lr": 7.44777425742027e-08}
{"train_loss": 0.013449955731630325, "global_step": 266742, "epoch": 2997, "lr": 7.444611498396703e-08}
{"train_loss": 0.026404205709695816, "global_step": 266743, "epoch": 2997, "lr": 7.441449410562351e-08}
{"train_loss": 0.044069934636354446, "global_step": 266744, "epoch": 2997, "lr": 7.438287993916104e-08}
{"train_loss": 0.019553549587726593, "global_step": 266745, "epoch": 2997, "lr": 7.435127248459073e-08}
{"train_loss": 0.03846049681305885, "global_step": 266746, "epoch": 2997, "lr": 7.431967174191256e-08}
{"train_loss": 0.05229487270116806, "global_step": 266747, "epoch": 2997, "lr": 7.428807771113767e-08}
{"train_loss": 0.06756914407014847, "global_step": 266748, "epoch": 2997, "lr": 7.4256490392266e-08}
{"train_loss": 0.032138071954250336, "global_step": 266749, "epoch": 2997, "lr": 7.422490978530316e-08}
{"train_loss": 0.06938740611076355, "global_step": 266750, "epoch": 2997, "lr": 7.419333589024913e-08}
{"train_loss": 0.056598637253046036, "global_step": 266751, "epoch": 2997, "lr": 7.416176870710945e-08}
{"train_loss": 0.03820331022143364, "global_step": 266752, "epoch": 2997, "lr": 7.413020823589522e-08}
{"train_loss": 0.10940264910459518, "global_step": 266753, "epoch": 2997, "lr": 7.409865447660091e-08}
{"train_loss": 0.03766991198062897, "global_step": 266754, "epoch": 2997, "lr": 7.406710742923761e-08}
{"train_loss": 0.022080108523368835, "global_step": 266755, "epoch": 2997, "lr": 7.403556709381088e-08}
{"train_loss": 0.05578775703907013, "global_step": 266756, "epoch": 2997, "lr": 7.400403347031514e-08}
{"train_loss": 0.027325136587023735, "global_step": 266757, "epoch": 2997, "lr": 7.397250655876709e-08}
{"train_loss": 0.0549759604036808, "global_step": 266758, "epoch": 2997, "lr": 7.394098635916114e-08}
{"train_loss": 0.03124392405152321, "global_step": 266759, "epoch": 2997, "lr": 7.39094728715084e-08}
{"train_loss": 0.0211581252515316, "global_step": 266760, "epoch": 2997, "lr": 7.387796609580888e-08}
{"train_loss": 0.05155752971768379, "global_step": 266761, "epoch": 2997, "lr": 7.384646603206257e-08}
{"train_loss": 0.03377848491072655, "global_step": 266762, "epoch": 2997, "lr": 7.381497268028614e-08}
{"train_loss": 0.056645095348358154, "global_step": 266763, "epoch": 2997, "lr": 7.378348604047402e-08}
{"train_loss": 0.055921852588653564, "global_step": 266764, "epoch": 2997, "lr": 7.375200611263178e-08}
{"train_loss": 0.048218242824077606, "global_step": 266765, "epoch": 2997, "lr": 7.37205328967705e-08}
{"train_loss": 0.07802319526672363, "global_step": 266766, "epoch": 2997, "lr": 7.368906639288465e-08}
{"train_loss": 0.016136562451720238, "global_step": 266767, "epoch": 2997, "lr": 7.365760660097976e-08}
{"train_loss": 0.028328821063041687, "global_step": 266768, "epoch": 2997, "lr": 7.362615352106694e-08}
{"train_loss": 0.034712277352809906, "global_step": 266769, "epoch": 2997, "lr": 7.35947071531462e-08}
{"train_loss": 0.029201064258813858, "global_step": 266770, "epoch": 2997, "lr": 7.356326749722308e-08}
{"train_loss": 0.031123049557209015, "global_step": 266771, "epoch": 2997, "lr": 7.353183455330314e-08}
{"train_loss": 0.05655212327837944, "global_step": 266772, "epoch": 2997, "lr": 7.350040832138083e-08}
{"train_loss": 0.02845815382897854, "global_step": 266773, "epoch": 2997, "lr": 7.34689888014728e-08}
{"train_loss": 0.0864795595407486, "global_step": 266774, "epoch": 2997, "lr": 7.343757599357904e-08}
{"train_loss": 0.04827730730175972, "global_step": 266775, "epoch": 2997, "lr": 7.340616989769955e-08}
{"train_loss": 0.04152550548315048, "global_step": 266776, "epoch": 2997, "lr": 7.337477051384545e-08}
{"train_loss": 0.025611352175474167, "global_step": 266777, "epoch": 2997, "lr": 7.334337784201673e-08}
{"train_loss": 0.028646759688854218, "global_step": 266778, "epoch": 2997, "lr": 7.331199188221893e-08}
{"train_loss": 0.03032596781849861, "global_step": 266779, "epoch": 2997, "lr": 7.328061263445763e-08}
{"train_loss": 0.03827086463570595, "global_step": 266780, "epoch": 2997, "lr": 7.32492400987328e-08}
{"train_loss": 0.04578050598502159, "global_step": 266781, "epoch": 2997, "lr": 7.321787427505e-08}
{"train_loss": 0.05761861428618431, "global_step": 266782, "epoch": 2997, "lr": 7.318651516341479e-08}
{"train_loss": 0.054684318602085114, "global_step": 266783, "epoch": 2997, "lr": 7.315516276383271e-08}
{"train_loss": 0.02584242634475231, "global_step": 266784, "epoch": 2997, "lr": 7.312381707630378e-08}
{"train_loss": 0.05288441851735115, "global_step": 266785, "epoch": 2997, "lr": 7.309247810083908e-08}
{"train_loss": 0.01470235362648964, "global_step": 266786, "epoch": 2997, "lr": 7.306114583743306e-08}
{"train_loss": 0.03619641438126564, "global_step": 266787, "epoch": 2997, "lr": 7.302982028610239e-08}
{"train_loss": 0.06366700679063797, "global_step": 266788, "epoch": 2997, "lr": 7.299850144683596e-08}
{"train_loss": 0.03966150060296059, "global_step": 266789, "epoch": 2997, "lr": 7.296718931965041e-08}
{"train_loss": 0.02824118547141552, "global_step": 266790, "epoch": 2997, "lr": 7.29358839045513e-08}
{"train_loss": 0.08611645549535751, "global_step": 266791, "epoch": 2997, "lr": 7.290458520152754e-08}
{"train_loss": 0.06090551242232323, "global_step": 266792, "epoch": 2997, "lr": 7.287329321060132e-08}
{"train_loss": 0.07100193947553635, "global_step": 266793, "epoch": 2997, "lr": 7.28420079317671e-08}
{"train_loss": 0.02970796450972557, "global_step": 266794, "epoch": 2997, "lr": 7.281072936502486e-08}
{"train_loss": 0.0032347708474844694, "global_step": 266795, "epoch": 2997, "lr": 7.277945751039128e-08}
{"train_loss": 0.02312486246228218, "global_step": 266796, "epoch": 2997, "lr": 7.274819236786079e-08}
{"train_loss": 0.04432244971394539, "global_step": 266797, "epoch": 2997, "lr": 7.27169339374445e-08}
{"train_loss": 0.05038677528500557, "global_step": 266798, "epoch": 2997, "lr": 7.268568221913685e-08}
{"train_loss": 0.03723255172371864, "global_step": 266799, "epoch": 2997, "lr": 7.26544372129545e-08}
{"train_loss": 0.03701665624976158, "global_step": 266800, "epoch": 2997, "lr": 7.26231989188919e-08}
{"train_loss": 0.020398110151290894, "global_step": 266801, "epoch": 2997, "lr": 7.259196733695461e-08}
{"train_loss": 0.09415841102600098, "global_step": 266802, "epoch": 2997, "lr": 7.256074246715372e-08}
{"train_loss": 0.03432402014732361, "global_step": 266803, "epoch": 2997, "lr": 7.252952430948368e-08}
{"train_loss": 0.015350213274359703, "global_step": 266804, "epoch": 2997, "lr": 7.249831286395559e-08}
{"train_loss": 0.01990807056427002, "global_step": 266805, "epoch": 2997, "lr": 7.246710813056944e-08}
{"train_loss": 0.010039957240223885, "global_step": 266806, "epoch": 2997, "lr": 7.243591010933637e-08}
{"train_loss": 0.08804294466972351, "global_step": 266807, "epoch": 2997, "lr": 7.240471880025079e-08}
{"train_loss": 0.06553547084331512, "global_step": 266808, "epoch": 2997, "lr": 7.237353420332383e-08}
{"train_loss": 0.07389400899410248, "global_step": 266809, "epoch": 2997, "lr": 7.234235631856101e-08}
{"train_loss": 0.040079329162836075, "global_step": 266810, "epoch": 2997, "lr": 7.23111851459568e-08}
{"train_loss": 0.0368112176656723, "global_step": 266811, "epoch": 2997, "lr": 7.228002068552785e-08}
{"train_loss": 0.03379334136843681, "global_step": 266812, "epoch": 2997, "lr": 7.224886293726863e-08}
{"train_loss": 0.024533282965421677, "global_step": 266813, "epoch": 2997, "lr": 7.22177119011902e-08}
{"train_loss": 0.05863835662603378, "global_step": 266814, "epoch": 2997, "lr": 7.218656757729258e-08}
{"train_loss": 0.066196508705616, "global_step": 266815, "epoch": 2997, "lr": 7.215542996557578e-08}
{"train_loss": 0.023091759532690048, "global_step": 266816, "epoch": 2997, "lr": 7.212429906605645e-08}
{"train_loss": 0.05302467197179794, "global_step": 266817, "epoch": 2997, "lr": 7.209317487872902e-08}
{"train_loss": 0.028874486684799194, "global_step": 266818, "epoch": 2997, "lr": 7.206205740359906e-08}
{"train_loss": 0.029529007151722908, "global_step": 266819, "epoch": 2997, "lr": 7.203094664067212e-08}
{"train_loss": 0.022636085748672485, "global_step": 266820, "epoch": 2997, "lr": 7.199984258995374e-08}
{"train_loss": 0.042439943710551334, "global_step": 266821, "epoch": 2997, "lr": 7.196874525144393e-08, "val_loss": 9.593435287475586}
{"train_loss": 0.0457332544028759, "global_step": 266822, "epoch": 2998, "lr": 7.193765462514823e-08}
{"train_loss": 0.04422849044203758, "global_step": 266823, "epoch": 2998, "lr": 7.190657071107221e-08}
{"train_loss": 0.022562861442565918, "global_step": 266824, "epoch": 2998, "lr": 7.18754935092214e-08}
{"train_loss": 0.06946495920419693, "global_step": 266825, "epoch": 2998, "lr": 7.184442301960137e-08}
{"train_loss": 0.01871473714709282, "global_step": 266826, "epoch": 2998, "lr": 7.18133592422121e-08}
{"train_loss": 0.01645807735621929, "global_step": 266827, "epoch": 2998, "lr": 7.178230217705361e-08}
{"train_loss": 0.05342661589384079, "global_step": 266828, "epoch": 2998, "lr": 7.175125182414255e-08}
{"train_loss": 0.07309390604496002, "global_step": 266829, "epoch": 2998, "lr": 7.172020818346781e-08}
{"train_loss": 0.05876566842198372, "global_step": 266830, "epoch": 2998, "lr": 7.168917125505159e-08}
{"train_loss": 0.028896693140268326, "global_step": 266831, "epoch": 2998, "lr": 7.16581410388828e-08}
{"train_loss": 0.032630957663059235, "global_step": 266832, "epoch": 2998, "lr": 7.162711753497253e-08}
{"train_loss": 0.026957638561725616, "global_step": 266833, "epoch": 2998, "lr": 7.15961007433208e-08}
{"train_loss": 0.03221499174833298, "global_step": 266834, "epoch": 2998, "lr": 7.156509066393313e-08}
{"train_loss": 0.06343938410282135, "global_step": 266835, "epoch": 2998, "lr": 7.153408729682066e-08}
{"train_loss": 0.04462835565209389, "global_step": 266836, "epoch": 2998, "lr": 7.15030906419778e-08}
{"train_loss": 0.04957618564367294, "global_step": 266837, "epoch": 2998, "lr": 7.147210069941568e-08}
{"train_loss": 0.023880857974290848, "global_step": 266838, "epoch": 2998, "lr": 7.144111746913429e-08}
{"train_loss": 0.031181836500763893, "global_step": 266839, "epoch": 2998, "lr": 7.141014095113918e-08}
{"train_loss": 0.02262818068265915, "global_step": 266840, "epoch": 2998, "lr": 7.137917114543036e-08}
{"train_loss": 0.07477302849292755, "global_step": 266841, "epoch": 2998, "lr": 7.134820805202447e-08}
{"train_loss": 0.01717495732009411, "global_step": 266842, "epoch": 2998, "lr": 7.131725167091042e-08}
{"train_loss": 0.05935991555452347, "global_step": 266843, "epoch": 2998, "lr": 7.12863020020993e-08}
{"train_loss": 0.00741244200617075, "global_step": 266844, "epoch": 2998, "lr": 7.125535904559666e-08}
{"train_loss": 0.02314773201942444, "global_step": 266845, "epoch": 2998, "lr": 7.122442280140806e-08}
{"train_loss": 0.046521034091711044, "global_step": 266846, "epoch": 2998, "lr": 7.11934932695335e-08}
{"train_loss": 0.012953746132552624, "global_step": 266847, "epoch": 2998, "lr": 7.116257044997854e-08}
{"train_loss": 0.05314913019537926, "global_step": 266848, "epoch": 2998, "lr": 7.113165434274315e-08}
{"train_loss": 0.027357446029782295, "global_step": 266849, "epoch": 2998, "lr": 7.110074494783848e-08}
{"train_loss": 0.05172228440642357, "global_step": 266850, "epoch": 2998, "lr": 7.106984226527003e-08}
{"train_loss": 0.051320336759090424, "global_step": 266851, "epoch": 2998, "lr": 7.103894629503228e-08}
{"train_loss": 0.023381002247333527, "global_step": 266852, "epoch": 2998, "lr": 7.100805703713631e-08}
{"train_loss": 0.04465655982494354, "global_step": 266853, "epoch": 2998, "lr": 7.097717449158215e-08}
{"train_loss": 0.051501985639333725, "global_step": 266854, "epoch": 2998, "lr": 7.094629865838087e-08}
{"train_loss": 0.08205491304397583, "global_step": 266855, "epoch": 2998, "lr": 7.091542953752694e-08}
{"train_loss": 0.046083275228738785, "global_step": 266856, "epoch": 2998, "lr": 7.088456712903702e-08}
{"train_loss": 0.056223634630441666, "global_step": 266857, "epoch": 2998, "lr": 7.085371143290553e-08}
{"train_loss": 0.019794786348938942, "global_step": 266858, "epoch": 2998, "lr": 7.082286244913806e-08}
{"train_loss": 0.02274424582719803, "global_step": 266859, "epoch": 2998, "lr": 7.079202017773456e-08}
{"train_loss": 0.04606493189930916, "global_step": 266860, "epoch": 2998, "lr": 7.076118461871174e-08}
{"train_loss": 0.04141111299395561, "global_step": 266861, "epoch": 2998, "lr": 7.0730355772064e-08}
{"train_loss": 0.0613926462829113, "global_step": 266862, "epoch": 2998, "lr": 7.069953363780246e-08}
{"train_loss": 0.04487822577357292, "global_step": 266863, "epoch": 2998, "lr": 7.066871821592158e-08}
{"train_loss": 0.027285851538181305, "global_step": 266864, "epoch": 2998, "lr": 7.063790950643245e-08}
{"train_loss": 0.04221025109291077, "global_step": 266865, "epoch": 2998, "lr": 7.060710750933508e-08}
{"train_loss": 0.028103139251470566, "global_step": 266866, "epoch": 2998, "lr": 7.057631222464057e-08}
{"train_loss": 0.03012029081583023, "global_step": 266867, "epoch": 2998, "lr": 7.054552365234334e-08}
{"train_loss": 0.08478710055351257, "global_step": 266868, "epoch": 2998, "lr": 7.051474179246009e-08}
{"train_loss": 0.052232734858989716, "global_step": 266869, "epoch": 2998, "lr": 7.048396664497969e-08}
{"train_loss": 0.060952700674533844, "global_step": 266870, "epoch": 2998, "lr": 7.04531982099188e-08}
{"train_loss": 0.04851089417934418, "global_step": 266871, "epoch": 2998, "lr": 7.042243648727742e-08}
{"train_loss": 0.03040141612291336, "global_step": 266872, "epoch": 2998, "lr": 7.039168147705555e-08}
{"train_loss": 0.05230076611042023, "global_step": 266873, "epoch": 2998, "lr": 7.036093317926429e-08}
{"train_loss": 0.05534396320581436, "global_step": 266874, "epoch": 2998, "lr": 7.033019159390364e-08}
{"train_loss": 0.02969488687813282, "global_step": 266875, "epoch": 2998, "lr": 7.02994567209736e-08}
{"train_loss": 0.06071963533759117, "global_step": 266876, "epoch": 2998, "lr": 7.026872856049083e-08}
{"train_loss": 0.037161391228437424, "global_step": 266877, "epoch": 2998, "lr": 7.023800711244421e-08}
{"train_loss": 0.07948502153158188, "global_step": 266878, "epoch": 2998, "lr": 7.020729237685043e-08}
{"train_loss": 0.03375060856342316, "global_step": 266879, "epoch": 2998, "lr": 7.017658435370944e-08}
{"train_loss": 0.03924323990941048, "global_step": 266880, "epoch": 2998, "lr": 7.014588304302128e-08}
{"train_loss": 0.028176013380289078, "global_step": 266881, "epoch": 2998, "lr": 7.011518844479703e-08}
{"train_loss": 0.04483615979552269, "global_step": 266882, "epoch": 2998, "lr": 7.00845005590367e-08}
{"train_loss": 0.07403682172298431, "global_step": 266883, "epoch": 2998, "lr": 7.005381938574585e-08}
{"train_loss": 0.013654534704983234, "global_step": 266884, "epoch": 2998, "lr": 7.002314492492445e-08}
{"train_loss": 0.05508977919816971, "global_step": 266885, "epoch": 2998, "lr": 6.999247717657809e-08}
{"train_loss": 0.1262516975402832, "global_step": 266886, "epoch": 2998, "lr": 6.996181614071784e-08}
{"train_loss": 0.02916557900607586, "global_step": 266887, "epoch": 2998, "lr": 6.993116181734371e-08}
{"train_loss": 0.06395146995782852, "global_step": 266888, "epoch": 2998, "lr": 6.990051420645017e-08}
{"train_loss": 0.04571853578090668, "global_step": 266889, "epoch": 2998, "lr": 6.986987330805939e-08}
{"train_loss": 0.06396136432886124, "global_step": 266890, "epoch": 2998, "lr": 6.983923912216028e-08}
{"train_loss": 0.07487206161022186, "global_step": 266891, "epoch": 2998, "lr": 6.980861164876395e-08}
{"train_loss": 0.052211835980415344, "global_step": 266892, "epoch": 2998, "lr": 6.977799088787596e-08}
{"train_loss": 0.06063007935881615, "global_step": 266893, "epoch": 2998, "lr": 6.974737683949628e-08}
{"train_loss": 0.027922656387090683, "global_step": 266894, "epoch": 2998, "lr": 6.971676950363049e-08}
{"train_loss": 0.04218030720949173, "global_step": 266895, "epoch": 2998, "lr": 6.968616888028412e-08}
{"train_loss": 0.006058147177100182, "global_step": 266896, "epoch": 2998, "lr": 6.965557496945719e-08}
{"train_loss": 0.013947314582765102, "global_step": 266897, "epoch": 2998, "lr": 6.962498777115523e-08}
{"train_loss": 0.09361634403467178, "global_step": 266898, "epoch": 2998, "lr": 6.95944072853838e-08}
{"train_loss": 0.04593652859330177, "global_step": 266899, "epoch": 2998, "lr": 6.956383351214846e-08}
{"train_loss": 0.032849591225385666, "global_step": 266900, "epoch": 2998, "lr": 6.953326645145475e-08}
{"train_loss": 0.035906992852687836, "global_step": 266901, "epoch": 2998, "lr": 6.950270610329713e-08}
{"train_loss": 0.02727958746254444, "global_step": 266902, "epoch": 2998, "lr": 6.947215246769223e-08}
{"train_loss": 0.055798549205064774, "global_step": 266903, "epoch": 2998, "lr": 6.944160554463453e-08}
{"train_loss": 0.046094462275505066, "global_step": 266904, "epoch": 2998, "lr": 6.94110653341351e-08}
{"train_loss": 0.039065901190042496, "global_step": 266905, "epoch": 2998, "lr": 6.938053183619398e-08}
{"train_loss": 0.04732753708958626, "global_step": 266906, "epoch": 2998, "lr": 6.935000505081668e-08}
{"train_loss": 0.03610198199748993, "global_step": 266907, "epoch": 2998, "lr": 6.931948497800323e-08}
{"train_loss": 0.08161506056785583, "global_step": 266908, "epoch": 2998, "lr": 6.92889716177647e-08}
{"train_loss": 0.035757455974817276, "global_step": 266909, "epoch": 2998, "lr": 6.925846497009558e-08}
{"train_loss": 0.04450071127980613, "global_step": 266910, "epoch": 2998, "lr": 6.92279650350125e-08, "val_loss": 9.589442253112793}
{"train_loss": 0.019535185769200325, "global_step": 266911, "epoch": 2999, "lr": 6.91974718125099e-08}
{"train_loss": 0.010002883151173592, "global_step": 266912, "epoch": 2999, "lr": 6.91669853025989e-08}
{"train_loss": 0.05022917315363884, "global_step": 266913, "epoch": 2999, "lr": 6.913650550527396e-08}
{"train_loss": 0.04433154687285423, "global_step": 266914, "epoch": 2999, "lr": 6.910603242055169e-08}
{"train_loss": 0.031201593577861786, "global_step": 266915, "epoch": 2999, "lr": 6.907556604842657e-08}
{"train_loss": 0.020452212542295456, "global_step": 266916, "epoch": 2999, "lr": 6.904510638890416e-08}
{"train_loss": 0.024713704362511635, "global_step": 266917, "epoch": 2999, "lr": 6.901465344198998e-08}
{"train_loss": 0.060167182236909866, "global_step": 266918, "epoch": 2999, "lr": 6.898420720768961e-08}
{"train_loss": 0.04531662538647652, "global_step": 266919, "epoch": 2999, "lr": 6.895376768600858e-08}
{"train_loss": 0.013415183871984482, "global_step": 266920, "epoch": 2999, "lr": 6.892333487694136e-08}
{"train_loss": 0.04407588019967079, "global_step": 266921, "epoch": 2999, "lr": 6.889290878050459e-08}
{"train_loss": 0.01995118148624897, "global_step": 266922, "epoch": 2999, "lr": 6.886248939669271e-08}
{"train_loss": 0.03280957043170929, "global_step": 266923, "epoch": 2999, "lr": 6.883207672551683e-08}
{"train_loss": 0.018266817554831505, "global_step": 266924, "epoch": 2999, "lr": 6.880167076697697e-08}
{"train_loss": 0.023321067914366722, "global_step": 266925, "epoch": 2999, "lr": 6.877127152107865e-08}
{"train_loss": 0.053815264254808426, "global_step": 266926, "epoch": 2999, "lr": 6.874087898782744e-08}
{"train_loss": 0.027834393084049225, "global_step": 266927, "epoch": 2999, "lr": 6.871049316722334e-08}
{"train_loss": 0.05796029791235924, "global_step": 266928, "epoch": 2999, "lr": 6.868011405927188e-08}
{"train_loss": 0.07667282968759537, "global_step": 266929, "epoch": 2999, "lr": 6.864974166397864e-08}
{"train_loss": 0.018253473564982414, "global_step": 266930, "epoch": 2999, "lr": 6.86193759813436e-08}
{"train_loss": 0.03819951415061951, "global_step": 266931, "epoch": 2999, "lr": 6.858901701137788e-08}
{"train_loss": 0.028903141617774963, "global_step": 266932, "epoch": 2999, "lr": 6.855866475408145e-08}
{"train_loss": 0.06638793647289276, "global_step": 266933, "epoch": 2999, "lr": 6.852831920945436e-08}
{"train_loss": 0.02878795564174652, "global_step": 266934, "epoch": 2999, "lr": 6.84979803775132e-08}
{"train_loss": 0.046933118253946304, "global_step": 266935, "epoch": 2999, "lr": 6.846764825824692e-08}
{"train_loss": 0.033670637756586075, "global_step": 266936, "epoch": 2999, "lr": 6.84373228516666e-08}
{"train_loss": 0.053945962339639664, "global_step": 266937, "epoch": 2999, "lr": 6.840700415777779e-08}
{"train_loss": 0.050504133105278015, "global_step": 266938, "epoch": 2999, "lr": 6.837669217658604e-08}
{"train_loss": 0.07816480845212936, "global_step": 266939, "epoch": 2999, "lr": 6.834638690809136e-08}
{"train_loss": 0.08654534816741943, "global_step": 266940, "epoch": 2999, "lr": 6.83160883522993e-08}
{"train_loss": 0.07280288636684418, "global_step": 266941, "epoch": 2999, "lr": 6.828579650920985e-08}
{"train_loss": 0.04985439032316208, "global_step": 266942, "epoch": 2999, "lr": 6.825551137883412e-08}
{"train_loss": 0.05378173664212227, "global_step": 266943, "epoch": 2999, "lr": 6.822523296117212e-08}
{"train_loss": 0.08108428120613098, "global_step": 266944, "epoch": 2999, "lr": 6.819496125622937e-08}
{"train_loss": 0.03935671225190163, "global_step": 266945, "epoch": 2999, "lr": 6.81646962640059e-08}
{"train_loss": 0.060539111495018005, "global_step": 266946, "epoch": 2999, "lr": 6.81344379845128e-08}
{"train_loss": 0.05145082250237465, "global_step": 266947, "epoch": 2999, "lr": 6.810418641775007e-08}
{"train_loss": 0.019938252866268158, "global_step": 266948, "epoch": 2999, "lr": 6.807394156371771e-08}
{"train_loss": 0.020877446979284286, "global_step": 266949, "epoch": 2999, "lr": 6.804370342242682e-08}
{"train_loss": 0.04513738676905632, "global_step": 266950, "epoch": 2999, "lr": 6.801347199388297e-08}
{"train_loss": 0.02236955054104328, "global_step": 266951, "epoch": 2999, "lr": 6.798324727808059e-08}
{"train_loss": 0.03524322435259819, "global_step": 266952, "epoch": 2999, "lr": 6.795302927503078e-08}
{"train_loss": 0.031753867864608765, "global_step": 266953, "epoch": 2999, "lr": 6.79228179847391e-08}
{"train_loss": 0.01985745131969452, "global_step": 266954, "epoch": 2999, "lr": 6.789261340720554e-08}
{"train_loss": 0.02014138549566269, "global_step": 266955, "epoch": 2999, "lr": 6.786241554243012e-08}
{"train_loss": 0.04066159948706627, "global_step": 266956, "epoch": 2999, "lr": 6.783222439042946e-08}
{"train_loss": 0.02088240720331669, "global_step": 266957, "epoch": 2999, "lr": 6.780203995119249e-08}
{"train_loss": 0.026966050267219543, "global_step": 266958, "epoch": 2999, "lr": 6.777186222473586e-08}
{"train_loss": 0.07183285802602768, "global_step": 266959, "epoch": 2999, "lr": 6.774169121105956e-08}
{"train_loss": 0.011931903660297394, "global_step": 266960, "epoch": 2999, "lr": 6.771152691016358e-08}
{"train_loss": 0.044984571635723114, "global_step": 266961, "epoch": 2999, "lr": 6.768136932205904e-08}
{"train_loss": 0.05093471705913544, "global_step": 266962, "epoch": 2999, "lr": 6.765121844674038e-08}
{"train_loss": 0.07355479896068573, "global_step": 266963, "epoch": 2999, "lr": 6.762107428422426e-08}
{"train_loss": 0.020261254161596298, "global_step": 266964, "epoch": 2999, "lr": 6.759093683450512e-08}
{"train_loss": 0.046868834644556046, "global_step": 266965, "epoch": 2999, "lr": 6.756080609758852e-08}
{"train_loss": 0.05410923436284065, "global_step": 266966, "epoch": 2999, "lr": 6.753068207348001e-08}
{"train_loss": 0.09564031660556793, "global_step": 266967, "epoch": 2999, "lr": 6.750056476218514e-08}
{"train_loss": 0.0313197560608387, "global_step": 266968, "epoch": 2999, "lr": 6.74704541637039e-08}
{"train_loss": 0.0285789854824543, "global_step": 266969, "epoch": 2999, "lr": 6.744035027804185e-08}
{"train_loss": 0.08746425807476044, "global_step": 266970, "epoch": 2999, "lr": 6.741025310521009e-08}
{"train_loss": 0.04262252897024155, "global_step": 266971, "epoch": 2999, "lr": 6.738016264520308e-08}
{"train_loss": 0.024264121428132057, "global_step": 266972, "epoch": 2999, "lr": 6.735007889802636e-08}
{"train_loss": 0.07015850394964218, "global_step": 266973, "epoch": 2999, "lr": 6.732000186368548e-08}
{"train_loss": 0.09319663792848587, "global_step": 266974, "epoch": 2999, "lr": 6.728993154218599e-08}
{"train_loss": 0.047275397926568985, "global_step": 266975, "epoch": 2999, "lr": 6.72598679335279e-08}
{"train_loss": 0.04668242856860161, "global_step": 266976, "epoch": 2999, "lr": 6.72298110377223e-08}
{"train_loss": 0.07728773355484009, "global_step": 266977, "epoch": 2999, "lr": 6.71997608547692e-08}
{"train_loss": 0.04402932524681091, "global_step": 266978, "epoch": 2999, "lr": 6.716971738467414e-08}
{"train_loss": 0.023437853902578354, "global_step": 266979, "epoch": 2999, "lr": 6.713968062743159e-08}
{"train_loss": 0.0420314222574234, "global_step": 266980, "epoch": 2999, "lr": 6.710965058306373e-08}
{"train_loss": 0.043532367795705795, "global_step": 266981, "epoch": 2999, "lr": 6.707962725155392e-08}
{"train_loss": 0.06366857886314392, "global_step": 266982, "epoch": 2999, "lr": 6.704961063292436e-08}
{"train_loss": 0.04805171117186546, "global_step": 266983, "epoch": 2999, "lr": 6.70196007271695e-08}
{"train_loss": 0.06501765549182892, "global_step": 266984, "epoch": 2999, "lr": 6.69895975342949e-08}
{"train_loss": 0.04826754704117775, "global_step": 266985, "epoch": 2999, "lr": 6.695960105430055e-08}
{"train_loss": 0.05055837333202362, "global_step": 266986, "epoch": 2999, "lr": 6.69296112872031e-08}
{"train_loss": 0.02778027020394802, "global_step": 266987, "epoch": 2999, "lr": 6.689962823299146e-08}
{"train_loss": 0.03115035593509674, "global_step": 266988, "epoch": 2999, "lr": 6.686965189168226e-08}
{"train_loss": 0.06174950301647186, "global_step": 266989, "epoch": 2999, "lr": 6.683968226326998e-08}
{"train_loss": 0.04294666275382042, "global_step": 266990, "epoch": 2999, "lr": 6.680971934776015e-08}
{"train_loss": 0.016698209568858147, "global_step": 266991, "epoch": 2999, "lr": 6.677976314516388e-08}
{"train_loss": 0.06230214983224869, "global_step": 266992, "epoch": 2999, "lr": 6.674981365548117e-08}
{"train_loss": 0.02796129696071148, "global_step": 266993, "epoch": 2999, "lr": 6.671987087871202e-08}
{"train_loss": 0.05130883306264877, "global_step": 266994, "epoch": 2999, "lr": 6.668993481486752e-08}
{"train_loss": 0.057178933173418045, "global_step": 266995, "epoch": 2999, "lr": 6.666000546394213e-08}
{"train_loss": 0.059397514909505844, "global_step": 266996, "epoch": 2999, "lr": 6.663008282594696e-08}
{"train_loss": 0.05518711358308792, "global_step": 266997, "epoch": 2999, "lr": 6.660016690088755e-08}
{"train_loss": 0.034778814762830734, "global_step": 266998, "epoch": 2999, "lr": 6.65702576887639e-08}
{"train_loss": 0.04403230673476551, "global_step": 266999, "epoch": 2999, "lr": 6.654035518958713e-08, "val_loss": 9.6715669631958}
{"train_loss": 0.01972830668091774, "global_step": 267000, "epoch": 3000, "lr": 6.65104594033461e-08}
{"train_loss": 0.04374231770634651, "global_step": 267001, "epoch": 3000, "lr": 6.64805703300575e-08}
{"train_loss": 0.05777287110686302, "global_step": 267002, "epoch": 3000, "lr": 6.645068796972687e-08}
{"train_loss": 0.14183549582958221, "global_step": 267003, "epoch": 3000, "lr": 6.642081232234865e-08}
{"train_loss": 0.042695142328739166, "global_step": 267004, "epoch": 3000, "lr": 6.639094338792839e-08}
{"train_loss": 0.018467027693986893, "global_step": 267005, "epoch": 3000, "lr": 6.636108116647722e-08}
{"train_loss": 0.043047383427619934, "global_step": 267006, "epoch": 3000, "lr": 6.63312256579951e-08}
{"train_loss": 0.04338068515062332, "global_step": 267007, "epoch": 3000, "lr": 6.630137686248761e-08}
{"train_loss": 0.04309748113155365, "global_step": 267008, "epoch": 3000, "lr": 6.627153477995473e-08}
{"train_loss": 0.031361132860183716, "global_step": 267009, "epoch": 3000, "lr": 6.624169941040203e-08}
{"train_loss": 0.012823831290006638, "global_step": 267010, "epoch": 3000, "lr": 6.621187075384061e-08}
{"train_loss": 0.06387846916913986, "global_step": 267011, "epoch": 3000, "lr": 6.618204881025936e-08}
{"train_loss": 0.020427638664841652, "global_step": 267012, "epoch": 3000, "lr": 6.615223357968048e-08}
{"train_loss": 0.06944955885410309, "global_step": 267013, "epoch": 3000, "lr": 6.612242506209287e-08}
{"train_loss": 0.05824669077992439, "global_step": 267014, "epoch": 3000, "lr": 6.609262325750764e-08}
{"train_loss": 0.03786986321210861, "global_step": 267015, "epoch": 3000, "lr": 6.606282816593035e-08}
{"train_loss": 0.048799678683280945, "global_step": 267016, "epoch": 3000, "lr": 6.603303978736097e-08}
{"train_loss": 0.022216804325580597, "global_step": 267017, "epoch": 3000, "lr": 6.600325812179952e-08}
{"train_loss": 0.059526897966861725, "global_step": 267018, "epoch": 3000, "lr": 6.597348316926266e-08}
{"train_loss": 0.029000479727983475, "global_step": 267019, "epoch": 3000, "lr": 6.594371492974482e-08}
{"train_loss": 0.035779621452093124, "global_step": 267020, "epoch": 3000, "lr": 6.591395340325157e-08}
{"train_loss": 0.04700228571891785, "global_step": 267021, "epoch": 3000, "lr": 6.58841985897829e-08}
{"train_loss": 0.05405215173959732, "global_step": 267022, "epoch": 3000, "lr": 6.585445048935545e-08}
{"train_loss": 0.024703186005353928, "global_step": 267023, "epoch": 3000, "lr": 6.582470910195815e-08}
{"train_loss": 0.05432388186454773, "global_step": 267024, "epoch": 3000, "lr": 6.579497442760763e-08}
{"train_loss": 0.07706961035728455, "global_step": 267025, "epoch": 3000, "lr": 6.576524646630389e-08}
{"train_loss": 0.043525390326976776, "global_step": 267026, "epoch": 3000, "lr": 6.573552521804138e-08}
{"train_loss": 0.048145636916160583, "global_step": 267027, "epoch": 3000, "lr": 6.570581068283677e-08}
{"train_loss": 0.02652544155716896, "global_step": 267028, "epoch": 3000, "lr": 6.567610286069004e-08}
{"train_loss": 0.032708894461393356, "global_step": 267029, "epoch": 3000, "lr": 6.564640175160675e-08}
{"train_loss": 0.06247205659747124, "global_step": 267030, "epoch": 3000, "lr": 6.561670735558134e-08}
{"train_loss": 0.03232385218143463, "global_step": 267031, "epoch": 3000, "lr": 6.558701967263048e-08}
{"train_loss": 0.04038160294294357, "global_step": 267032, "epoch": 3000, "lr": 6.555733870275415e-08}
{"train_loss": 0.02503151074051857, "global_step": 267033, "epoch": 3000, "lr": 6.552766444595238e-08}
{"train_loss": 0.003932362422347069, "global_step": 267034, "epoch": 3000, "lr": 6.549799690223069e-08}
{"train_loss": 0.030615489929914474, "global_step": 267035, "epoch": 3000, "lr": 6.546833607160019e-08}
{"train_loss": 0.06014464795589447, "global_step": 267036, "epoch": 3000, "lr": 6.543868195404979e-08}
{"train_loss": 0.04060463234782219, "global_step": 267037, "epoch": 3000, "lr": 6.540903454960167e-08}
{"train_loss": 0.06467843800783157, "global_step": 267038, "epoch": 3000, "lr": 6.537939385824477e-08}
{"train_loss": 0.04247601330280304, "global_step": 267039, "epoch": 3000, "lr": 6.534975987999015e-08}
{"train_loss": 0.056106626987457275, "global_step": 267040, "epoch": 3000, "lr": 6.532013261484338e-08}
{"train_loss": 0.029538242146372795, "global_step": 267041, "epoch": 3000, "lr": 6.52905120627989e-08}
{"train_loss": 0.07179813086986542, "global_step": 267042, "epoch": 3000, "lr": 6.526089822387338e-08}
{"train_loss": 0.08055724203586578, "global_step": 267043, "epoch": 3000, "lr": 6.523129109806681e-08}
{"train_loss": 0.03200995922088623, "global_step": 267044, "epoch": 3000, "lr": 6.520169068537363e-08}
{"train_loss": 0.02774244360625744, "global_step": 267045, "epoch": 3000, "lr": 6.517209698581051e-08}
{"train_loss": 0.01429419033229351, "global_step": 267046, "epoch": 3000, "lr": 6.514250999937743e-08}
{"train_loss": 0.05354626476764679, "global_step": 267047, "epoch": 3000, "lr": 6.511292972607441e-08}
{"train_loss": 0.0590522438287735, "global_step": 267048, "epoch": 3000, "lr": 6.508335616591255e-08}
{"train_loss": 0.03359794244170189, "global_step": 267049, "epoch": 3000, "lr": 6.505378931888628e-08}
{"train_loss": 0.037411708384752274, "global_step": 267050, "epoch": 3000, "lr": 6.502422918500672e-08}
{"train_loss": 0.050334636121988297, "global_step": 267051, "epoch": 3000, "lr": 6.499467576427388e-08}
{"train_loss": 0.05014309287071228, "global_step": 267052, "epoch": 3000, "lr": 6.496512905669883e-08}
{"train_loss": 0.10985706001520157, "global_step": 267053, "epoch": 3000, "lr": 6.493558906227603e-08}
{"train_loss": 0.06311561912298203, "global_step": 267054, "epoch": 3000, "lr": 6.49060557810166e-08}
{"train_loss": 0.02736786939203739, "global_step": 267055, "epoch": 3000, "lr": 6.487652921292053e-08}
{"train_loss": 0.05669734999537468, "global_step": 267056, "epoch": 3000, "lr": 6.484700935799892e-08}
{"train_loss": 0.04208746552467346, "global_step": 267057, "epoch": 3000, "lr": 6.481749621624067e-08}
{"train_loss": 0.03164353966712952, "global_step": 267058, "epoch": 3000, "lr": 6.478798978766798e-08}
{"train_loss": 0.03368717432022095, "global_step": 267059, "epoch": 3000, "lr": 6.475849007226975e-08}
{"train_loss": 0.060433465987443924, "global_step": 267060, "epoch": 3000, "lr": 6.472899707005709e-08}
{"train_loss": 0.028178619220852852, "global_step": 267061, "epoch": 3000, "lr": 6.469951078103553e-08}
{"train_loss": 0.044027697294950485, "global_step": 267062, "epoch": 3000, "lr": 6.467003120520509e-08}
{"train_loss": 0.04884649068117142, "global_step": 267063, "epoch": 3000, "lr": 6.464055834257132e-08}
{"train_loss": 0.06782792508602142, "global_step": 267064, "epoch": 3000, "lr": 6.461109219313422e-08}
{"train_loss": 0.03212057054042816, "global_step": 267065, "epoch": 3000, "lr": 6.458163275690487e-08}
{"train_loss": 0.054894670844078064, "global_step": 267066, "epoch": 3000, "lr": 6.45521800338833e-08}
{"train_loss": 0.021350761875510216, "global_step": 267067, "epoch": 3000, "lr": 6.452273402407505e-08}
{"train_loss": 0.0295743215829134, "global_step": 267068, "epoch": 3000, "lr": 6.449329472748567e-08}
{"train_loss": 0.021679922938346863, "global_step": 267069, "epoch": 3000, "lr": 6.44638621441096e-08}
{"train_loss": 0.06688711047172546, "global_step": 267070, "epoch": 3000, "lr": 6.44344362739635e-08}
{"train_loss": 0.028910038992762566, "global_step": 267071, "epoch": 3000, "lr": 6.440501711704183e-08}
{"train_loss": 0.03853076323866844, "global_step": 267072, "epoch": 3000, "lr": 6.437560467335568e-08}
{"train_loss": 0.0757470354437828, "global_step": 267073, "epoch": 3000, "lr": 6.434619894290506e-08}
{"train_loss": 0.03919285535812378, "global_step": 267074, "epoch": 3000, "lr": 6.431679992568995e-08}
{"train_loss": 0.07075568288564682, "global_step": 267075, "epoch": 3000, "lr": 6.428740762172702e-08}
{"train_loss": 0.05487629771232605, "global_step": 267076, "epoch": 3000, "lr": 6.425802203100517e-08}
{"train_loss": 0.05675740912556648, "global_step": 267077, "epoch": 3000, "lr": 6.42286431535355e-08}
{"train_loss": 0.07570162415504456, "global_step": 267078, "epoch": 3000, "lr": 6.419927098932354e-08}
{"train_loss": 0.06013152003288269, "global_step": 267079, "epoch": 3000, "lr": 6.416990553837488e-08}
{"train_loss": 0.011550940573215485, "global_step": 267080, "epoch": 3000, "lr": 6.414054680068393e-08}
{"train_loss": 0.06944118440151215, "global_step": 267081, "epoch": 3000, "lr": 6.411119477626737e-08}
{"train_loss": 0.011263689957559109, "global_step": 267082, "epoch": 3000, "lr": 6.408184946511409e-08}
{"train_loss": 0.05829988047480583, "global_step": 267083, "epoch": 3000, "lr": 6.405251086724074e-08}
{"train_loss": 0.06342756748199463, "global_step": 267084, "epoch": 3000, "lr": 6.402317898264732e-08}
{"train_loss": 0.0990564301609993, "global_step": 267085, "epoch": 3000, "lr": 6.399385381133938e-08}
{"train_loss": 0.04470003396272659, "global_step": 267086, "epoch": 3000, "lr": 6.396453535331692e-08}
{"train_loss": 0.0449841246008873, "global_step": 267087, "epoch": 3000, "lr": 6.393522360857995e-08}
{"train_loss": 0.04656740343930681, "global_step": 267088, "epoch": 3000, "lr": 6.390591857714511e-08, "train/sim_max_reward_0": 0.8219336196558827, "train/sim_max_reward_1": 0.9015042288479473, "train/sim_max_reward_2": 0.6591859772689598, "train/sim_max_reward_3": 0.502549205700502, "train/sim_max_reward_4": 0.9684293924657739, "train/sim_max_reward_5": 0.9793599439733991, "test/sim_max_reward_4500000": 1.0, "test/sim_max_reward_4500001": 0.0, "test/sim_max_reward_4500002": 0.8854783826160929, "test/sim_max_reward_4500003": 0.18009820070376237, "test/sim_max_reward_4500004": 0.7541432560134765, "test/sim_max_reward_4500005": 0.4015810959542543, "test/sim_max_reward_4500006": 0.9574931871125119, "test/sim_max_reward_4500007": 0.8968749874444512, "test/sim_max_reward_4500008": 0.9450480988789313, "test/sim_max_reward_4500009": 0.04932542963446854, "test/sim_max_reward_4500010": 0.0, "test/sim_max_reward_4500011": 1.0, "test/sim_max_reward_4500012": 1.0, "test/sim_max_reward_4500013": 0.45097602023170347, "test/sim_max_reward_4500014": 1.0, "test/sim_max_reward_4500015": 0.9828132312828246, "test/sim_max_reward_4500016": 0.943011713836128, "test/sim_max_reward_4500017": 0.5002650788024614, "test/sim_max_reward_4500018": 1.0, "test/sim_max_reward_4500019": 0.4124388999665175, "test/sim_max_reward_4500020": 0.0, "test/sim_max_reward_4500021": 0.9700710422828613, "test/sim_max_reward_4500022": 0.9849262793302462, "test/sim_max_reward_4500023": 0.6460361838710816, "test/sim_max_reward_4500024": 0.0, "test/sim_max_reward_4500025": 0.43061010689556595, "test/sim_max_reward_4500026": 0.926368271089663, "test/sim_max_reward_4500027": 0.9793229333820821, "test/sim_max_reward_4500028": 0.9730646581281993, "test/sim_max_reward_4500029": 1.0, "test/sim_max_reward_4500030": 1.0, "test/sim_max_reward_4500031": 0.4471208875264433, "test/sim_max_reward_4500032": 0.4516018625386982, "test/sim_max_reward_4500033": 0.9738510273639986, "test/sim_max_reward_4500034": 0.5674177153480416, "test/sim_max_reward_4500035": 0.980950203518697, "test/sim_max_reward_4500036": 1.0, "test/sim_max_reward_4500037": 0.9124365124430468, "test/sim_max_reward_4500038": 0.16240908880695706, "test/sim_max_reward_4500039": 0.9832580879744276, "test/sim_max_reward_4500040": 1.0, "test/sim_max_reward_4500041": 0.467989159503594, "test/sim_max_reward_4500042": 0.0, "test/sim_max_reward_4500043": 0.5679524146808121, "test/sim_max_reward_4500044": 0.0, "test/sim_max_reward_4500045": 0.9581637444037724, "test/sim_max_reward_4500046": 0.6178094751621459, "test/sim_max_reward_4500047": 0.0, "test/sim_max_reward_4500048": 0.9699667973299796, "test/sim_max_reward_4500049": 0.9510120965983185, "train/mean_score": 0.8054937279854109, "test/mean_score": 0.6656377226131244, "val_loss": 9.76695442199707, "train_action_mse_error": 6.315728187561035}
{"train_loss": 0.05855712294578552, "global_step": 267089, "epoch": 3001, "lr": 6.387662025900687e-08}
{"train_loss": 0.05606979876756668, "global_step": 267090, "epoch": 3001, "lr": 6.38473286541763e-08}
{"train_loss": 0.05830858647823334, "global_step": 267091, "epoch": 3001, "lr": 6.381804376264788e-08}
{"train_loss": 0.0420946329832077, "global_step": 267092, "epoch": 3001, "lr": 6.37887655844327e-08}
{"train_loss": 0.02100687101483345, "global_step": 267093, "epoch": 3001, "lr": 6.375949411953075e-08}
{"train_loss": 0.04961349070072174, "global_step": 267094, "epoch": 3001, "lr": 6.373022936794759e-08}
{"train_loss": 0.04167252033948898, "global_step": 267095, "epoch": 3001, "lr": 6.370097132968878e-08}
{"train_loss": 0.09275977313518524, "global_step": 267096, "epoch": 3001, "lr": 6.36717200047543e-08}
{"train_loss": 0.03160887584090233, "global_step": 267097, "epoch": 3001, "lr": 6.364247539315526e-08}
{"train_loss": 0.055589742958545685, "global_step": 267098, "epoch": 3001, "lr": 6.361323749488612e-08}
{"train_loss": 0.041137438267469406, "global_step": 267099, "epoch": 3001, "lr": 6.358400630995797e-08}
{"train_loss": 0.030969848856329918, "global_step": 267100, "epoch": 3001, "lr": 6.355478183837082e-08}
{"train_loss": 0.03633645549416542, "global_step": 267101, "epoch": 3001, "lr": 6.352556408013022e-08}
{"train_loss": 0.019137250259518623, "global_step": 267102, "epoch": 3001, "lr": 6.34963530352417e-08}
{"train_loss": 0.02534930594265461, "global_step": 267103, "epoch": 3001, "lr": 6.346714870370529e-08}
{"train_loss": 0.021615106612443924, "global_step": 267104, "epoch": 3001, "lr": 6.343795108552653e-08}
{"train_loss": 0.04587935656309128, "global_step": 267105, "epoch": 3001, "lr": 6.340876018071096e-08}
{"train_loss": 0.06065698713064194, "global_step": 267106, "epoch": 3001, "lr": 6.337957598926413e-08}
{"train_loss": 0.04625518620014191, "global_step": 267107, "epoch": 3001, "lr": 6.335039851118607e-08}
{"train_loss": 0.07622237503528595, "global_step": 267108, "epoch": 3001, "lr": 6.33212277464823e-08}
{"train_loss": 0.018962187692523003, "global_step": 267109, "epoch": 3001, "lr": 6.329206369515283e-08}
{"train_loss": 0.03071165271103382, "global_step": 267110, "epoch": 3001, "lr": 6.326290635720877e-08}
{"train_loss": 0.029433006420731544, "global_step": 267111, "epoch": 3001, "lr": 6.323375573265011e-08}
{"train_loss": 0.05361497029662132, "global_step": 267112, "epoch": 3001, "lr": 6.320461182147686e-08}
{"train_loss": 0.028506072238087654, "global_step": 267113, "epoch": 3001, "lr": 6.31754746237001e-08}
{"train_loss": 0.0729115903377533, "global_step": 267114, "epoch": 3001, "lr": 6.314634413931985e-08}
{"train_loss": 0.01941244676709175, "global_step": 267115, "epoch": 3001, "lr": 6.311722036834167e-08}
{"train_loss": 0.02614227496087551, "global_step": 267116, "epoch": 3001, "lr": 6.308810331077109e-08}
{"train_loss": 0.033909983932971954, "global_step": 267117, "epoch": 3001, "lr": 6.30589929666081e-08}
{"train_loss": 0.03567219153046608, "global_step": 267118, "epoch": 3001, "lr": 6.302988933585829e-08}
{"train_loss": 0.039466340094804764, "global_step": 267119, "epoch": 3001, "lr": 6.300079241852164e-08}
{"train_loss": 0.027904421091079712, "global_step": 267120, "epoch": 3001, "lr": 6.297170221460924e-08}
{"train_loss": 0.07475673407316208, "global_step": 267121, "epoch": 3001, "lr": 6.294261872412111e-08}
{"train_loss": 0.05459728091955185, "global_step": 267122, "epoch": 3001, "lr": 6.291354194706278e-08}
{"train_loss": 0.012328565120697021, "global_step": 267123, "epoch": 3001, "lr": 6.288447188343982e-08}
{"train_loss": 0.038637883961200714, "global_step": 267124, "epoch": 3001, "lr": 6.285540853324667e-08}
{"train_loss": 0.03859559819102287, "global_step": 267125, "epoch": 3001, "lr": 6.282635189649999e-08}
{"train_loss": 0.04834473505616188, "global_step": 267126, "epoch": 3001, "lr": 6.279730197319423e-08}
{"train_loss": 0.036821458488702774, "global_step": 267127, "epoch": 3001, "lr": 6.276825876333491e-08}
{"train_loss": 0.06259404867887497, "global_step": 267128, "epoch": 3001, "lr": 6.273922226693318e-08}
{"train_loss": 0.02538437396287918, "global_step": 267129, "epoch": 3001, "lr": 6.271019248398346e-08}
{"train_loss": 0.03189074248075485, "global_step": 267130, "epoch": 3001, "lr": 6.268116941449132e-08}
{"train_loss": 0.04211607202887535, "global_step": 267131, "epoch": 3001, "lr": 6.265215305846784e-08}
{"train_loss": 0.05999056622385979, "global_step": 267132, "epoch": 3001, "lr": 6.262314341591302e-08}
{"train_loss": 0.0661744549870491, "global_step": 267133, "epoch": 3001, "lr": 6.259414048682688e-08}
{"train_loss": 0.05461067706346512, "global_step": 267134, "epoch": 3001, "lr": 6.256514427121496e-08}
{"train_loss": 0.0342341810464859, "global_step": 267135, "epoch": 3001, "lr": 6.253615476908837e-08}
{"train_loss": 0.018738320097327232, "global_step": 267136, "epoch": 3001, "lr": 6.250717198043598e-08}
{"train_loss": 0.025623975321650505, "global_step": 267137, "epoch": 3001, "lr": 6.247819590528003e-08}
{"train_loss": 0.05774978920817375, "global_step": 267138, "epoch": 3001, "lr": 6.24492265436094e-08}
{"train_loss": 0.08398429304361343, "global_step": 267139, "epoch": 3001, "lr": 6.24202638954352e-08}
{"train_loss": 0.02145126461982727, "global_step": 267140, "epoch": 3001, "lr": 6.239130796076298e-08}
{"train_loss": 0.08094608038663864, "global_step": 267141, "epoch": 3001, "lr": 6.236235873959274e-08}
{"train_loss": 0.05152220278978348, "global_step": 267142, "epoch": 3001, "lr": 6.233341623192446e-08}
{"train_loss": 0.03539276495575905, "global_step": 267143, "epoch": 3001, "lr": 6.230448043776927e-08}
{"train_loss": 0.005353443790227175, "global_step": 267144, "epoch": 3001, "lr": 6.22755513571327e-08}
{"train_loss": 0.021282115951180458, "global_step": 267145, "epoch": 3001, "lr": 6.224662899000921e-08}
{"train_loss": 0.11175807565450668, "global_step": 267146, "epoch": 3001, "lr": 6.221771333640991e-08}
{"train_loss": 0.04729542136192322, "global_step": 267147, "epoch": 3001, "lr": 6.218880439633478e-08}
{"train_loss": 0.0378863625228405, "global_step": 267148, "epoch": 3001, "lr": 6.215990216979495e-08}
{"train_loss": 0.0524902306497097, "global_step": 267149, "epoch": 3001, "lr": 6.213100665678485e-08}
{"train_loss": 0.06182849779725075, "global_step": 267150, "epoch": 3001, "lr": 6.210211785731557e-08}
{"train_loss": 0.07266056537628174, "global_step": 267151, "epoch": 3001, "lr": 6.207323577138157e-08}
{"train_loss": 0.05904172360897064, "global_step": 267152, "epoch": 3001, "lr": 6.204436039899952e-08}
{"train_loss": 0.011996965855360031, "global_step": 267153, "epoch": 3001, "lr": 6.201549174016386e-08}
{"train_loss": 0.03806598111987114, "global_step": 267154, "epoch": 3001, "lr": 6.198662979488567e-08}
{"train_loss": 0.04202372953295708, "global_step": 267155, "epoch": 3001, "lr": 6.195777456315943e-08}
{"train_loss": 0.03707500547170639, "global_step": 267156, "epoch": 3001, "lr": 6.192892604499622e-08}
{"train_loss": 0.07551608234643936, "global_step": 267157, "epoch": 3001, "lr": 6.190008424040161e-08}
{"train_loss": 0.0346815288066864, "global_step": 267158, "epoch": 3001, "lr": 6.187124914937004e-08}
{"train_loss": 0.039424918591976166, "global_step": 267159, "epoch": 3001, "lr": 6.18424207719126e-08}
{"train_loss": 0.052332326769828796, "global_step": 267160, "epoch": 3001, "lr": 6.181359910803486e-08}
{"train_loss": 0.06777216494083405, "global_step": 267161, "epoch": 3001, "lr": 6.178478415773681e-08}
{"train_loss": 0.08810967952013016, "global_step": 267162, "epoch": 3001, "lr": 6.1755975921024e-08}
{"train_loss": 0.062392234802246094, "global_step": 267163, "epoch": 3001, "lr": 6.172717439789643e-08}
{"train_loss": 0.028156766667962074, "global_step": 267164, "epoch": 3001, "lr": 6.169837958835967e-08}
{"train_loss": 0.07772770524024963, "global_step": 267165, "epoch": 3001, "lr": 6.16695914924248e-08}
{"train_loss": 0.04421013593673706, "global_step": 267166, "epoch": 3001, "lr": 6.164081011008627e-08}
{"train_loss": 0.07194837927818298, "global_step": 267167, "epoch": 3001, "lr": 6.161203544134963e-08}
{"train_loss": 0.025718726217746735, "global_step": 267168, "epoch": 3001, "lr": 6.158326748622045e-08}
{"train_loss": 0.013698555529117584, "global_step": 267169, "epoch": 3001, "lr": 6.155450624470426e-08}
{"train_loss": 0.049738410860300064, "global_step": 267170, "epoch": 3001, "lr": 6.152575171680664e-08}
{"train_loss": 0.034694138914346695, "global_step": 267171, "epoch": 3001, "lr": 6.149700390252755e-08}
{"train_loss": 0.05917244777083397, "global_step": 267172, "epoch": 3001, "lr": 6.146826280186702e-08}
{"train_loss": 0.05809566378593445, "global_step": 267173, "epoch": 3001, "lr": 6.143952841483613e-08}
{"train_loss": 0.031710073351860046, "global_step": 267174, "epoch": 3001, "lr": 6.141080074143491e-08}
{"train_loss": 0.057552043348550797, "global_step": 267175, "epoch": 3001, "lr": 6.138207978166888e-08}
{"train_loss": 0.08126340061426163, "global_step": 267176, "epoch": 3001, "lr": 6.13533655355436e-08}
{"train_loss": 0.045743338266659654, "global_step": 267177, "epoch": 3001, "lr": 6.132465800305909e-08, "val_loss": 9.626115798950195}
{"train_loss": 0.01202432718127966, "global_step": 267178, "epoch": 3002, "lr": 6.129595718422088e-08}
{"train_loss": 0.06698501110076904, "global_step": 267179, "epoch": 3002, "lr": 6.126726307902897e-08}
{"train_loss": 0.019241217523813248, "global_step": 267180, "epoch": 3002, "lr": 6.123857568749447e-08}
{"train_loss": 0.08665432035923004, "global_step": 267181, "epoch": 3002, "lr": 6.120989500961738e-08}
{"train_loss": 0.02636229433119297, "global_step": 267182, "epoch": 3002, "lr": 6.118122104540325e-08}
{"train_loss": 0.04229123517870903, "global_step": 267183, "epoch": 3002, "lr": 6.115255379485762e-08}
{"train_loss": 0.023532692342996597, "global_step": 267184, "epoch": 3002, "lr": 6.112389325797496e-08}
{"train_loss": 0.05647958070039749, "global_step": 267185, "epoch": 3002, "lr": 6.109523943476635e-08}
{"train_loss": 0.03468216955661774, "global_step": 267186, "epoch": 3002, "lr": 6.106659232523737e-08}
{"train_loss": 0.05661550536751747, "global_step": 267187, "epoch": 3002, "lr": 6.103795192938799e-08}
{"train_loss": 0.04418950527906418, "global_step": 267188, "epoch": 3002, "lr": 6.100931824722377e-08}
{"train_loss": 0.026723679155111313, "global_step": 267189, "epoch": 3002, "lr": 6.098069127875028e-08}
{"train_loss": 0.05381575971841812, "global_step": 267190, "epoch": 3002, "lr": 6.09520710239675e-08}
{"train_loss": 0.06125100702047348, "global_step": 267191, "epoch": 3002, "lr": 6.092345748288098e-08}
{"train_loss": 0.03637440502643585, "global_step": 267192, "epoch": 3002, "lr": 6.089485065549072e-08}
{"train_loss": 0.041368041187524796, "global_step": 267193, "epoch": 3002, "lr": 6.086625054180783e-08}
{"train_loss": 0.07670722901821136, "global_step": 267194, "epoch": 3002, "lr": 6.083765714183787e-08}
{"train_loss": 0.04652900621294975, "global_step": 267195, "epoch": 3002, "lr": 6.080907045557527e-08}
{"train_loss": 0.06981830298900604, "global_step": 267196, "epoch": 3002, "lr": 6.078049048302559e-08}
{"train_loss": 0.013163472525775433, "global_step": 267197, "epoch": 3002, "lr": 6.075191722419992e-08}
{"train_loss": 0.024735447019338608, "global_step": 267198, "epoch": 3002, "lr": 6.072335067909274e-08}
{"train_loss": 0.02500755526125431, "global_step": 267199, "epoch": 3002, "lr": 6.069479084771512e-08}
{"train_loss": 0.0323767326772213, "global_step": 267200, "epoch": 3002, "lr": 6.066623773006708e-08}
{"train_loss": 0.10420887172222137, "global_step": 267201, "epoch": 3002, "lr": 6.063769132615416e-08}
{"train_loss": 0.05213160812854767, "global_step": 267202, "epoch": 3002, "lr": 6.06091516359819e-08}
{"train_loss": 0.043257247656583786, "global_step": 267203, "epoch": 3002, "lr": 6.058061865955034e-08}
{"train_loss": 0.04708890616893768, "global_step": 267204, "epoch": 3002, "lr": 6.0552092396865e-08}
{"train_loss": 0.0393916517496109, "global_step": 267205, "epoch": 3002, "lr": 6.052357284793143e-08}
{"train_loss": 0.019319918006658554, "global_step": 267206, "epoch": 3002, "lr": 6.049506001274963e-08}
{"train_loss": 0.03499068692326546, "global_step": 267207, "epoch": 3002, "lr": 6.046655389132516e-08}
{"train_loss": 0.030481982976198196, "global_step": 267208, "epoch": 3002, "lr": 6.043805448366358e-08}
{"train_loss": 0.053036198019981384, "global_step": 267209, "epoch": 3002, "lr": 6.040956178977042e-08}
{"train_loss": 0.03186364099383354, "global_step": 267210, "epoch": 3002, "lr": 6.038107580964569e-08}
{"train_loss": 0.022374117746949196, "global_step": 267211, "epoch": 3002, "lr": 6.03525965432894e-08}
{"train_loss": 0.03141967952251434, "global_step": 267212, "epoch": 3002, "lr": 6.032412399071263e-08}
{"train_loss": 0.06191102787852287, "global_step": 267213, "epoch": 3002, "lr": 6.02956581519154e-08}
{"train_loss": 0.012013815343379974, "global_step": 267214, "epoch": 3002, "lr": 6.026719902690325e-08}
{"train_loss": 0.05626175180077553, "global_step": 267215, "epoch": 3002, "lr": 6.023874661568175e-08}
{"train_loss": 0.04072336107492447, "global_step": 267216, "epoch": 3002, "lr": 6.021030091825087e-08}
{"train_loss": 0.04297838732600212, "global_step": 267217, "epoch": 3002, "lr": 6.018186193461617e-08}
{"train_loss": 0.009179549291729927, "global_step": 267218, "epoch": 3002, "lr": 6.015342966478321e-08}
{"train_loss": 0.04986865818500519, "global_step": 267219, "epoch": 3002, "lr": 6.0125004108752e-08}
{"train_loss": 0.056974027305841446, "global_step": 267220, "epoch": 3002, "lr": 6.009658526652806e-08}
{"train_loss": 0.06982338428497314, "global_step": 267221, "epoch": 3002, "lr": 6.006817313811697e-08}
{"train_loss": 0.05898940563201904, "global_step": 267222, "epoch": 3002, "lr": 6.003976772351871e-08}
{"train_loss": 0.08017735928297043, "global_step": 267223, "epoch": 3002, "lr": 6.001136902273885e-08}
{"train_loss": 0.05364660546183586, "global_step": 267224, "epoch": 3002, "lr": 5.998297703578293e-08}
{"train_loss": 0.053897272795438766, "global_step": 267225, "epoch": 3002, "lr": 5.995459176265095e-08}
{"train_loss": 0.042125873267650604, "global_step": 267226, "epoch": 3002, "lr": 5.992621320335401e-08}
{"train_loss": 0.09964317083358765, "global_step": 267227, "epoch": 3002, "lr": 5.989784135789211e-08}
{"train_loss": 0.05444344878196716, "global_step": 267228, "epoch": 3002, "lr": 5.986947622626526e-08}
{"train_loss": 0.04257723316550255, "global_step": 267229, "epoch": 3002, "lr": 5.984111780847901e-08}
{"train_loss": 0.08726278692483902, "global_step": 267230, "epoch": 3002, "lr": 5.981276610454445e-08}
{"train_loss": 0.04238319769501686, "global_step": 267231, "epoch": 3002, "lr": 5.978442111445049e-08}
{"train_loss": 0.03725256025791168, "global_step": 267232, "epoch": 3002, "lr": 5.975608283821932e-08}
{"train_loss": 0.06045074388384819, "global_step": 267233, "epoch": 3002, "lr": 5.972775127583985e-08}
{"train_loss": 0.08782581984996796, "global_step": 267234, "epoch": 3002, "lr": 5.969942642732318e-08}
{"train_loss": 0.05847662687301636, "global_step": 267235, "epoch": 3002, "lr": 5.967110829266931e-08}
{"train_loss": 0.04361250624060631, "global_step": 267236, "epoch": 3002, "lr": 5.964279687188378e-08}
{"train_loss": 0.04631228372454643, "global_step": 267237, "epoch": 3002, "lr": 5.961449216497217e-08}
{"train_loss": 0.03223547339439392, "global_step": 267238, "epoch": 3002, "lr": 5.958619417193445e-08}
{"train_loss": 0.022161230444908142, "global_step": 267239, "epoch": 3002, "lr": 5.955790289278174e-08}
{"train_loss": 0.06674744933843613, "global_step": 267240, "epoch": 3002, "lr": 5.9529618327508474e-08}
{"train_loss": 0.03937317803502083, "global_step": 267241, "epoch": 3002, "lr": 5.950134047612022e-08}
{"train_loss": 0.0985223650932312, "global_step": 267242, "epoch": 3002, "lr": 5.947306933862806e-08}
{"train_loss": 0.0383567251265049, "global_step": 267243, "epoch": 3002, "lr": 5.9444804915032015e-08}
{"train_loss": 0.028408197686076164, "global_step": 267244, "epoch": 3002, "lr": 5.9416547205332076e-08}
{"train_loss": 0.03352919965982437, "global_step": 267245, "epoch": 3002, "lr": 5.9388296209533787e-08}
{"train_loss": 0.04069546237587929, "global_step": 267246, "epoch": 3002, "lr": 5.9360051927642714e-08}
{"train_loss": 0.019802240654826164, "global_step": 267247, "epoch": 3002, "lr": 5.9331814359664396e-08}
{"train_loss": 0.04875348508358002, "global_step": 267248, "epoch": 3002, "lr": 5.930358350559884e-08}
{"train_loss": 0.02773667313158512, "global_step": 267249, "epoch": 3002, "lr": 5.927535936545159e-08}
{"train_loss": 0.035610366612672806, "global_step": 267250, "epoch": 3002, "lr": 5.924714193922265e-08}
{"train_loss": 0.010402040556073189, "global_step": 267251, "epoch": 3002, "lr": 5.921893122692312e-08}
{"train_loss": 0.06674374639987946, "global_step": 267252, "epoch": 3002, "lr": 5.9190727228553014e-08}
{"train_loss": 0.03399239480495453, "global_step": 267253, "epoch": 3002, "lr": 5.916252994411786e-08}
{"train_loss": 0.023032071068882942, "global_step": 267254, "epoch": 3002, "lr": 5.913433937361212e-08}
{"train_loss": 0.053094279021024704, "global_step": 267255, "epoch": 3002, "lr": 5.9106155517052456e-08}
{"train_loss": 0.03251764923334122, "global_step": 267256, "epoch": 3002, "lr": 5.907797837443885e-08}
{"train_loss": 0.04847043380141258, "global_step": 267257, "epoch": 3002, "lr": 5.9049807945771304e-08}
{"train_loss": 0.0640718936920166, "global_step": 267258, "epoch": 3002, "lr": 5.9021644231055384e-08}
{"train_loss": 0.019001508131623268, "global_step": 267259, "epoch": 3002, "lr": 5.899348723029663e-08}
{"train_loss": 0.03363736346364021, "global_step": 267260, "epoch": 3002, "lr": 5.8965336943495045e-08}
{"train_loss": 0.026235660538077354, "global_step": 267261, "epoch": 3002, "lr": 5.893719337065618e-08}
{"train_loss": 0.02753671444952488, "global_step": 267262, "epoch": 3002, "lr": 5.890905651179113e-08}
{"train_loss": 0.043536264449357986, "global_step": 267263, "epoch": 3002, "lr": 5.88809263668888e-08}
{"train_loss": 0.05629204958677292, "global_step": 267264, "epoch": 3002, "lr": 5.885280293596584e-08}
{"train_loss": 0.02352103963494301, "global_step": 267265, "epoch": 3002, "lr": 5.8824686219022264e-08}
{"train_loss": 0.044668761343601045, "global_step": 267266, "epoch": 3002, "lr": 5.879657621605805e-08, "val_loss": 9.647095680236816}
{"train_loss": 0.08544889837503433, "global_step": 267267, "epoch": 3003, "lr": 5.8768472927084316e-08}
{"train_loss": 0.04230276495218277, "global_step": 267268, "epoch": 3003, "lr": 5.8740376352095505e-08}
{"train_loss": 0.06344790756702423, "global_step": 267269, "epoch": 3003, "lr": 5.8712286491102717e-08}
{"train_loss": 0.04460642859339714, "global_step": 267270, "epoch": 3003, "lr": 5.868420334410596e-08}
{"train_loss": 0.027301188558340073, "global_step": 267271, "epoch": 3003, "lr": 5.865612691111078e-08}
{"train_loss": 0.03120737336575985, "global_step": 267272, "epoch": 3003, "lr": 5.862805719212272e-08}
{"train_loss": 0.02370550110936165, "global_step": 267273, "epoch": 3003, "lr": 5.859999418714179e-08}
{"train_loss": 0.042687732726335526, "global_step": 267274, "epoch": 3003, "lr": 5.8571937896173544e-08}
{"train_loss": 0.036218516528606415, "global_step": 267275, "epoch": 3003, "lr": 5.8543888319223526e-08}
{"train_loss": 0.04680119827389717, "global_step": 267276, "epoch": 3003, "lr": 5.851584545629174e-08}
{"train_loss": 0.03532705456018448, "global_step": 267277, "epoch": 3003, "lr": 5.848780930738373e-08}
{"train_loss": 0.024887511506676674, "global_step": 267278, "epoch": 3003, "lr": 5.845977987249951e-08}
{"train_loss": 0.04581790789961815, "global_step": 267279, "epoch": 3003, "lr": 5.843175715165572e-08}
{"train_loss": 0.027914060279726982, "global_step": 267280, "epoch": 3003, "lr": 5.840374114484126e-08}
{"train_loss": 0.07248970121145248, "global_step": 267281, "epoch": 3003, "lr": 5.8375731852061686e-08}
{"train_loss": 0.0949641689658165, "global_step": 267282, "epoch": 3003, "lr": 5.834772927333365e-08}
{"train_loss": 0.02862773835659027, "global_step": 267283, "epoch": 3003, "lr": 5.831973340864605e-08}
{"train_loss": 0.03088168427348137, "global_step": 267284, "epoch": 3003, "lr": 5.829174425800998e-08}
{"train_loss": 0.03474447876214981, "global_step": 267285, "epoch": 3003, "lr": 5.8263761821431004e-08}
{"train_loss": 0.027584630995988846, "global_step": 267286, "epoch": 3003, "lr": 5.823578609890357e-08}
{"train_loss": 0.10363797843456268, "global_step": 267287, "epoch": 3003, "lr": 5.820781709044432e-08}
{"train_loss": 0.05358244478702545, "global_step": 267288, "epoch": 3003, "lr": 5.8179854796042155e-08}
{"train_loss": 0.043737929314374924, "global_step": 267289, "epoch": 3003, "lr": 5.815189921571929e-08}
{"train_loss": 0.05531519651412964, "global_step": 267290, "epoch": 3003, "lr": 5.8123950349459055e-08}
{"train_loss": 0.057346493005752563, "global_step": 267291, "epoch": 3003, "lr": 5.809600819728367e-08}
{"train_loss": 0.05946466699242592, "global_step": 267292, "epoch": 3003, "lr": 5.806807275918202e-08}
{"train_loss": 0.12605144083499908, "global_step": 267293, "epoch": 3003, "lr": 5.8040144035170775e-08}
{"train_loss": 0.026220276951789856, "global_step": 267294, "epoch": 3003, "lr": 5.801222202524437e-08}
{"train_loss": 0.05464017763733864, "global_step": 267295, "epoch": 3003, "lr": 5.798430672940836e-08}
{"train_loss": 0.02762162685394287, "global_step": 267296, "epoch": 3003, "lr": 5.795639814766829e-08}
{"train_loss": 0.0539541020989418, "global_step": 267297, "epoch": 3003, "lr": 5.792849628002417e-08}
{"train_loss": 0.047816529870033264, "global_step": 267298, "epoch": 3003, "lr": 5.79006011264871e-08}
{"train_loss": 0.049854010343551636, "global_step": 267299, "epoch": 3003, "lr": 5.7872712687057076e-08}
{"train_loss": 0.07644366472959518, "global_step": 267300, "epoch": 3003, "lr": 5.78448309617341e-08}
{"train_loss": 0.013638605363667011, "global_step": 267301, "epoch": 3003, "lr": 5.781695595052927e-08}
{"train_loss": 0.056295644491910934, "global_step": 267302, "epoch": 3003, "lr": 5.778908765343705e-08}
{"train_loss": 0.04155568405985832, "global_step": 267303, "epoch": 3003, "lr": 5.776122607046852e-08}
{"train_loss": 0.04117605835199356, "global_step": 267304, "epoch": 3003, "lr": 5.773337120162925e-08}
{"train_loss": 0.07194087654352188, "global_step": 267305, "epoch": 3003, "lr": 5.770552304691368e-08}
{"train_loss": 0.06658721715211868, "global_step": 267306, "epoch": 3003, "lr": 5.7677681606332914e-08}
{"train_loss": 0.012002096511423588, "global_step": 267307, "epoch": 3003, "lr": 5.76498468798925e-08}
{"train_loss": 0.052410442382097244, "global_step": 267308, "epoch": 3003, "lr": 5.7622018867586894e-08}
{"train_loss": 0.019913960248231888, "global_step": 267309, "epoch": 3003, "lr": 5.759419756942719e-08}
{"train_loss": 0.02548624388873577, "global_step": 267310, "epoch": 3003, "lr": 5.756638298541339e-08}
{"train_loss": 0.054108671844005585, "global_step": 267311, "epoch": 3003, "lr": 5.7538575115556606e-08}
{"train_loss": 0.039828501641750336, "global_step": 267312, "epoch": 3003, "lr": 5.751077395984572e-08}
{"train_loss": 0.05282416194677353, "global_step": 267313, "epoch": 3003, "lr": 5.748297951830295e-08}
{"train_loss": 0.025474973022937775, "global_step": 267314, "epoch": 3003, "lr": 5.7455191790917185e-08}
{"train_loss": 0.07651552557945251, "global_step": 267315, "epoch": 3003, "lr": 5.742741077769953e-08}
{"train_loss": 0.0458887442946434, "global_step": 267316, "epoch": 3003, "lr": 5.739963647865554e-08}
{"train_loss": 0.02794305793941021, "global_step": 267317, "epoch": 3003, "lr": 5.7371868893779653e-08}
{"train_loss": 0.05102863907814026, "global_step": 267318, "epoch": 3003, "lr": 5.7344108023082986e-08}
{"train_loss": 0.013707714155316353, "global_step": 267319, "epoch": 3003, "lr": 5.7316353866571084e-08}
{"train_loss": 0.05350145697593689, "global_step": 267320, "epoch": 3003, "lr": 5.728860642424394e-08}
{"train_loss": 0.05919962003827095, "global_step": 267321, "epoch": 3003, "lr": 5.726086569610156e-08}
{"train_loss": 0.08203369379043579, "global_step": 267322, "epoch": 3003, "lr": 5.723313168215505e-08}
{"train_loss": 0.057734254747629166, "global_step": 267323, "epoch": 3003, "lr": 5.7205404382404404e-08}
{"train_loss": 0.0319635309278965, "global_step": 267324, "epoch": 3003, "lr": 5.7177683796849626e-08}
{"train_loss": 0.015273276716470718, "global_step": 267325, "epoch": 3003, "lr": 5.7149969925501814e-08}
{"train_loss": 0.06214703619480133, "global_step": 267326, "epoch": 3003, "lr": 5.712226276836097e-08}
{"train_loss": 0.049211058765649796, "global_step": 267327, "epoch": 3003, "lr": 5.709456232543264e-08}
{"train_loss": 0.08282685279846191, "global_step": 267328, "epoch": 3003, "lr": 5.706686859671684e-08}
{"train_loss": 0.06390024721622467, "global_step": 267329, "epoch": 3003, "lr": 5.7039181582224653e-08}
{"train_loss": 0.03157460317015648, "global_step": 267330, "epoch": 3003, "lr": 5.701150128195054e-08}
{"train_loss": 0.008212573826313019, "global_step": 267331, "epoch": 3003, "lr": 5.6983827695905604e-08}
{"train_loss": 0.05363483354449272, "global_step": 267332, "epoch": 3003, "lr": 5.695616082408428e-08}
{"train_loss": 0.0683169960975647, "global_step": 267333, "epoch": 3003, "lr": 5.692850066650324e-08}
{"train_loss": 0.030126692727208138, "global_step": 267334, "epoch": 3003, "lr": 5.690084722315692e-08}
{"train_loss": 0.024565938860177994, "global_step": 267335, "epoch": 3003, "lr": 5.687320049405087e-08}
{"train_loss": 0.016613738611340523, "global_step": 267336, "epoch": 3003, "lr": 5.684556047919065e-08}
{"train_loss": 0.050222016870975494, "global_step": 267337, "epoch": 3003, "lr": 5.6817927178576257e-08}
{"train_loss": 0.03843884915113449, "global_step": 267338, "epoch": 3003, "lr": 5.679030059221879e-08}
{"train_loss": 0.025162648409605026, "global_step": 267339, "epoch": 3003, "lr": 5.6762680720112704e-08}
{"train_loss": 0.026346886530518532, "global_step": 267340, "epoch": 3003, "lr": 5.6735067562269096e-08}
{"train_loss": 0.03548083081841469, "global_step": 267341, "epoch": 3003, "lr": 5.670746111868797e-08}
{"train_loss": 0.0463770255446434, "global_step": 267342, "epoch": 3003, "lr": 5.6679861389369314e-08}
{"train_loss": 0.03716563060879707, "global_step": 267343, "epoch": 3003, "lr": 5.6652268374324244e-08}
{"train_loss": 0.023378020152449608, "global_step": 267344, "epoch": 3003, "lr": 5.662468207355831e-08}
{"train_loss": 0.033684141933918, "global_step": 267345, "epoch": 3003, "lr": 5.6597102487065954e-08}
{"train_loss": 0.011306852102279663, "global_step": 267346, "epoch": 3003, "lr": 5.6569529614858287e-08}
{"train_loss": 0.04462837055325508, "global_step": 267347, "epoch": 3003, "lr": 5.654196345692975e-08}
{"train_loss": 0.031366173177957535, "global_step": 267348, "epoch": 3003, "lr": 5.6514404013296997e-08}
{"train_loss": 0.03863580524921417, "global_step": 267349, "epoch": 3003, "lr": 5.648685128395448e-08}
{"train_loss": 0.022248534485697746, "global_step": 267350, "epoch": 3003, "lr": 5.6459305268907746e-08}
{"train_loss": 0.03077239915728569, "global_step": 267351, "epoch": 3003, "lr": 5.643176596816235e-08}
{"train_loss": 0.041994608938694, "global_step": 267352, "epoch": 3003, "lr": 5.6404233381723846e-08}
{"train_loss": 0.03462297469377518, "global_step": 267353, "epoch": 3003, "lr": 5.637670750958668e-08}
{"train_loss": 0.10110101848840714, "global_step": 267354, "epoch": 3003, "lr": 5.63491883517675e-08}
{"train_loss": 0.04505479465542215, "global_step": 267355, "epoch": 3003, "lr": 5.632167590826076e-08, "val_loss": 9.58018684387207}
{"train_loss": 0.015485026873648167, "global_step": 267356, "epoch": 3004, "lr": 5.629417017907201e-08}
{"train_loss": 0.038599446415901184, "global_step": 267357, "epoch": 3004, "lr": 5.626667116420681e-08}
{"train_loss": 0.06513553112745285, "global_step": 267358, "epoch": 3004, "lr": 5.623917886366514e-08}
{"train_loss": 0.016924411058425903, "global_step": 267359, "epoch": 3004, "lr": 5.621169327745812e-08}
{"train_loss": 0.022898662835359573, "global_step": 267360, "epoch": 3004, "lr": 5.618421440558019e-08}
{"train_loss": 0.0676640197634697, "global_step": 267361, "epoch": 3004, "lr": 5.615674224804246e-08}
{"train_loss": 0.016596144065260887, "global_step": 267362, "epoch": 3004, "lr": 5.6129276804844924e-08}
{"train_loss": 0.07990951836109161, "global_step": 267363, "epoch": 3004, "lr": 5.6101818075993126e-08}
{"train_loss": 0.0496232770383358, "global_step": 267364, "epoch": 3004, "lr": 5.607436606148708e-08}
{"train_loss": 0.04225059971213341, "global_step": 267365, "epoch": 3004, "lr": 5.604692076133233e-08}
{"train_loss": 0.03709726780653, "global_step": 267366, "epoch": 3004, "lr": 5.601948217553998e-08}
{"train_loss": 0.032897695899009705, "global_step": 267367, "epoch": 3004, "lr": 5.599205030409893e-08}
{"train_loss": 0.059569209814071655, "global_step": 267368, "epoch": 3004, "lr": 5.596462514702583e-08}
{"train_loss": 0.0397728830575943, "global_step": 267369, "epoch": 3004, "lr": 5.593720670432068e-08}
{"train_loss": 0.04072989523410797, "global_step": 267370, "epoch": 3004, "lr": 5.590979497598348e-08}
{"train_loss": 0.048135850578546524, "global_step": 267371, "epoch": 3004, "lr": 5.588238996201978e-08}
{"train_loss": 0.021064622327685356, "global_step": 267372, "epoch": 3004, "lr": 5.585499166243513e-08}
{"train_loss": 0.06050844490528107, "global_step": 267373, "epoch": 3004, "lr": 5.582760007722953e-08}
{"train_loss": 0.048991959542036057, "global_step": 267374, "epoch": 3004, "lr": 5.5800215206414097e-08}
{"train_loss": 0.03784824162721634, "global_step": 267375, "epoch": 3004, "lr": 5.577283704998326e-08}
{"train_loss": 0.04901335388422012, "global_step": 267376, "epoch": 3004, "lr": 5.574546560794813e-08}
{"train_loss": 0.07156763970851898, "global_step": 267377, "epoch": 3004, "lr": 5.571810088030871e-08}
{"train_loss": 0.017128542065620422, "global_step": 267378, "epoch": 3004, "lr": 5.5690742867070547e-08}
{"train_loss": 0.0592997707426548, "global_step": 267379, "epoch": 3004, "lr": 5.5663391568233637e-08}
{"train_loss": 0.012990044429898262, "global_step": 267380, "epoch": 3004, "lr": 5.5636046983803533e-08}
{"train_loss": 0.02217385545372963, "global_step": 267381, "epoch": 3004, "lr": 5.5608709113785795e-08}
{"train_loss": 0.01894671656191349, "global_step": 267382, "epoch": 3004, "lr": 5.5581377958180416e-08}
{"train_loss": 0.014545731246471405, "global_step": 267383, "epoch": 3004, "lr": 5.555405351699294e-08}
{"train_loss": 0.04487548768520355, "global_step": 267384, "epoch": 3004, "lr": 5.552673579023449e-08}
{"train_loss": 0.08298012614250183, "global_step": 267385, "epoch": 3004, "lr": 5.549942477789394e-08}
{"train_loss": 0.05004812404513359, "global_step": 267386, "epoch": 3004, "lr": 5.5472120479987956e-08}
{"train_loss": 0.022638577967882156, "global_step": 267387, "epoch": 3004, "lr": 5.544482289651098e-08}
{"train_loss": 0.021602001041173935, "global_step": 267388, "epoch": 3004, "lr": 5.5417532027474126e-08}
{"train_loss": 0.02555343322455883, "global_step": 267389, "epoch": 3004, "lr": 5.539024787287739e-08}
{"train_loss": 0.026490872725844383, "global_step": 267390, "epoch": 3004, "lr": 5.5362970432720765e-08}
{"train_loss": 0.05486157163977623, "global_step": 267391, "epoch": 3004, "lr": 5.533569970701536e-08}
{"train_loss": 0.05383291840553284, "global_step": 267392, "epoch": 3004, "lr": 5.5308435695761165e-08}
{"train_loss": 0.024819064885377884, "global_step": 267393, "epoch": 3004, "lr": 5.528117839896374e-08}
{"train_loss": 0.10171715915203094, "global_step": 267394, "epoch": 3004, "lr": 5.5253927816617536e-08}
{"train_loss": 0.06283396482467651, "global_step": 267395, "epoch": 3004, "lr": 5.52266839487392e-08}
{"train_loss": 0.028822822496294975, "global_step": 267396, "epoch": 3004, "lr": 5.519944679532873e-08}
{"train_loss": 0.0180969201028347, "global_step": 267397, "epoch": 3004, "lr": 5.5172216356386143e-08}
{"train_loss": 0.03135942295193672, "global_step": 267398, "epoch": 3004, "lr": 5.514499263191697e-08}
{"train_loss": 0.058585651218891144, "global_step": 267399, "epoch": 3004, "lr": 5.511777562192122e-08}
{"train_loss": 0.08986213058233261, "global_step": 267400, "epoch": 3004, "lr": 5.509056532640999e-08}
{"train_loss": 0.042854271829128265, "global_step": 267401, "epoch": 3004, "lr": 5.506336174538329e-08}
{"train_loss": 0.046431124210357666, "global_step": 267402, "epoch": 3004, "lr": 5.503616487884111e-08}
{"train_loss": 0.03562881425023079, "global_step": 267403, "epoch": 3004, "lr": 5.500897472679456e-08}
{"train_loss": 0.032170943915843964, "global_step": 267404, "epoch": 3004, "lr": 5.498179128924363e-08}
{"train_loss": 0.02685832418501377, "global_step": 267405, "epoch": 3004, "lr": 5.4954614566188335e-08}
{"train_loss": 0.04102500155568123, "global_step": 267406, "epoch": 3004, "lr": 5.492744455763976e-08}
{"train_loss": 0.023576559498906136, "global_step": 267407, "epoch": 3004, "lr": 5.490028126359237e-08}
{"train_loss": 0.02406121790409088, "global_step": 267408, "epoch": 3004, "lr": 5.4873124684057255e-08}
{"train_loss": 0.021131744608283043, "global_step": 267409, "epoch": 3004, "lr": 5.484597481903442e-08}
{"train_loss": 0.033608853816986084, "global_step": 267410, "epoch": 3004, "lr": 5.4818831668534966e-08}
{"train_loss": 0.038847532123327255, "global_step": 267411, "epoch": 3004, "lr": 5.479169523254779e-08}
{"train_loss": 0.05320960283279419, "global_step": 267412, "epoch": 3004, "lr": 5.476456551108955e-08}
{"train_loss": 0.013899765908718109, "global_step": 267413, "epoch": 3004, "lr": 5.473744250416024e-08}
{"train_loss": 0.04359407722949982, "global_step": 267414, "epoch": 3004, "lr": 5.471032621175987e-08}
{"train_loss": 0.06726418435573578, "global_step": 267415, "epoch": 3004, "lr": 5.468321663389952e-08}
{"train_loss": 0.03224674239754677, "global_step": 267416, "epoch": 3004, "lr": 5.465611377057367e-08}
{"train_loss": 0.04195544496178627, "global_step": 267417, "epoch": 3004, "lr": 5.46290176217934e-08}
{"train_loss": 0.0516316257417202, "global_step": 267418, "epoch": 3004, "lr": 5.4601928187558716e-08}
{"train_loss": 0.052423249930143356, "global_step": 267419, "epoch": 3004, "lr": 5.457484546786962e-08}
{"train_loss": 0.04554132744669914, "global_step": 267420, "epoch": 3004, "lr": 5.4547769462737205e-08}
{"train_loss": 0.10311327874660492, "global_step": 267421, "epoch": 3004, "lr": 5.4520700172167036e-08}
{"train_loss": 0.04208967834711075, "global_step": 267422, "epoch": 3004, "lr": 5.449363759615356e-08}
{"train_loss": 0.030645325779914856, "global_step": 267423, "epoch": 3004, "lr": 5.446658173470231e-08}
{"train_loss": 0.04677971452474594, "global_step": 267424, "epoch": 3004, "lr": 5.4439532587824414e-08}
{"train_loss": 0.029145022854208946, "global_step": 267425, "epoch": 3004, "lr": 5.4412490155514305e-08}
{"train_loss": 0.028315067291259766, "global_step": 267426, "epoch": 3004, "lr": 5.438545443778309e-08}
{"train_loss": 0.028223617002367973, "global_step": 267427, "epoch": 3004, "lr": 5.4358425434625214e-08}
{"train_loss": 0.021930143237113953, "global_step": 267428, "epoch": 3004, "lr": 5.433140314605734e-08}
{"train_loss": 0.049940694123506546, "global_step": 267429, "epoch": 3004, "lr": 5.43043875720739e-08}
{"train_loss": 0.05028684064745903, "global_step": 267430, "epoch": 3004, "lr": 5.427737871267491e-08}
{"train_loss": 0.019419997930526733, "global_step": 267431, "epoch": 3004, "lr": 5.4250376567877014e-08}
{"train_loss": 0.0597601979970932, "global_step": 267432, "epoch": 3004, "lr": 5.422338113766911e-08}
{"train_loss": 0.07005266100168228, "global_step": 267433, "epoch": 3004, "lr": 5.4196392422067866e-08}
{"train_loss": 0.05133118852972984, "global_step": 267434, "epoch": 3004, "lr": 5.416941042106771e-08}
{"train_loss": 0.045174095779657364, "global_step": 267435, "epoch": 3004, "lr": 5.414243513467976e-08}
{"train_loss": 0.018066970631480217, "global_step": 267436, "epoch": 3004, "lr": 5.411546656289845e-08}
{"train_loss": 0.019428705796599388, "global_step": 267437, "epoch": 3004, "lr": 5.408850470574045e-08}
{"train_loss": 0.038341790437698364, "global_step": 267438, "epoch": 3004, "lr": 5.4061549563194645e-08}
{"train_loss": 0.07011500000953674, "global_step": 267439, "epoch": 3004, "lr": 5.4034601135272143e-08}
{"train_loss": 0.03972616791725159, "global_step": 267440, "epoch": 3004, "lr": 5.40076594219785e-08}
{"train_loss": 0.02217148430645466, "global_step": 267441, "epoch": 3004, "lr": 5.39807244233137e-08}
{"train_loss": 0.023499874398112297, "global_step": 267442, "epoch": 3004, "lr": 5.395379613928331e-08}
{"train_loss": 0.044196199625730515, "global_step": 267443, "epoch": 3004, "lr": 5.392687456988732e-08}
{"train_loss": 0.04150690072414915, "global_step": 267444, "epoch": 3004, "lr": 5.3899959715131286e-08, "val_loss": 9.622322082519531}
{"train_loss": 0.031323689967393875, "global_step": 267445, "epoch": 3005, "lr": 5.387305157502076e-08}
{"train_loss": 0.03812617063522339, "global_step": 267446, "epoch": 3005, "lr": 5.3846150149561295e-08}
{"train_loss": 0.02718820422887802, "global_step": 267447, "epoch": 3005, "lr": 5.381925543875288e-08}
{"train_loss": 0.02219889871776104, "global_step": 267448, "epoch": 3005, "lr": 5.3792367442595524e-08}
{"train_loss": 0.026258863508701324, "global_step": 267449, "epoch": 3005, "lr": 5.3765486161100334e-08}
{"train_loss": 0.04522398114204407, "global_step": 267450, "epoch": 3005, "lr": 5.37386115942673e-08}
{"train_loss": 0.02538742870092392, "global_step": 267451, "epoch": 3005, "lr": 5.3711743742101975e-08}
{"train_loss": 0.078040212392807, "global_step": 267452, "epoch": 3005, "lr": 5.368488260460436e-08}
{"train_loss": 0.05158264935016632, "global_step": 267453, "epoch": 3005, "lr": 5.3658028181780006e-08}
{"train_loss": 0.04321783408522606, "global_step": 267454, "epoch": 3005, "lr": 5.363118047363447e-08}
{"train_loss": 0.06704976409673691, "global_step": 267455, "epoch": 3005, "lr": 5.3604339480167745e-08}
{"train_loss": 0.036163780838251114, "global_step": 267456, "epoch": 3005, "lr": 5.357750520138538e-08}
{"train_loss": 0.044381771236658096, "global_step": 267457, "epoch": 3005, "lr": 5.355067763728738e-08}
{"train_loss": 0.022973304614424706, "global_step": 267458, "epoch": 3005, "lr": 5.352385678788485e-08}
{"train_loss": 0.048143643885850906, "global_step": 267459, "epoch": 3005, "lr": 5.349704265317779e-08}
{"train_loss": 0.04631417244672775, "global_step": 267460, "epoch": 3005, "lr": 5.347023523317174e-08}
{"train_loss": 0.03315254673361778, "global_step": 267461, "epoch": 3005, "lr": 5.3443434527861156e-08}
{"train_loss": 0.047179002314805984, "global_step": 267462, "epoch": 3005, "lr": 5.341664053725715e-08}
{"train_loss": 0.08370967209339142, "global_step": 267463, "epoch": 3005, "lr": 5.338985326136525e-08}
{"train_loss": 0.05210593342781067, "global_step": 267464, "epoch": 3005, "lr": 5.336307270018548e-08}
{"train_loss": 0.028496377170085907, "global_step": 267465, "epoch": 3005, "lr": 5.3336298853723374e-08}
{"train_loss": 0.0678095668554306, "global_step": 267466, "epoch": 3005, "lr": 5.330953172197895e-08}
{"train_loss": 0.03676975145936012, "global_step": 267467, "epoch": 3005, "lr": 5.3282771304963285e-08}
{"train_loss": 0.020488472655415535, "global_step": 267468, "epoch": 3005, "lr": 5.325601760267085e-08}
{"train_loss": 0.04859839007258415, "global_step": 267469, "epoch": 3005, "lr": 5.322927061511274e-08}
{"train_loss": 0.06279746443033218, "global_step": 267470, "epoch": 3005, "lr": 5.32025303422834e-08}
{"train_loss": 0.06647627055644989, "global_step": 267471, "epoch": 3005, "lr": 5.317579678419393e-08}
{"train_loss": 0.03339463844895363, "global_step": 267472, "epoch": 3005, "lr": 5.3149069940849893e-08}
{"train_loss": 0.0771857425570488, "global_step": 267473, "epoch": 3005, "lr": 5.312234981225128e-08}
{"train_loss": 0.056837957352399826, "global_step": 267474, "epoch": 3005, "lr": 5.30956363983981e-08}
{"train_loss": 0.05548134818673134, "global_step": 267475, "epoch": 3005, "lr": 5.3068929699295887e-08}
{"train_loss": 0.05257004126906395, "global_step": 267476, "epoch": 3005, "lr": 5.304222971495576e-08}
{"train_loss": 0.03643372654914856, "global_step": 267477, "epoch": 3005, "lr": 5.301553644537216e-08}
{"train_loss": 0.041642285883426666, "global_step": 267478, "epoch": 3005, "lr": 5.298884989055064e-08}
{"train_loss": 0.044807180762290955, "global_step": 267479, "epoch": 3005, "lr": 5.296217005049675e-08}
{"train_loss": 0.04064132273197174, "global_step": 267480, "epoch": 3005, "lr": 5.2935496925216044e-08}
{"train_loss": 0.017467789351940155, "global_step": 267481, "epoch": 3005, "lr": 5.2908830514702965e-08}
{"train_loss": 0.014908942393958569, "global_step": 267482, "epoch": 3005, "lr": 5.288217081897418e-08}
{"train_loss": 0.02740458957850933, "global_step": 267483, "epoch": 3005, "lr": 5.285551783802412e-08}
{"train_loss": 0.028538953512907028, "global_step": 267484, "epoch": 3005, "lr": 5.282887157185834e-08}
{"train_loss": 0.03672075644135475, "global_step": 267485, "epoch": 3005, "lr": 5.28022320204824e-08}
{"train_loss": 0.08018336445093155, "global_step": 267486, "epoch": 3005, "lr": 5.2775599183896296e-08}
{"train_loss": 0.05087461695075035, "global_step": 267487, "epoch": 3005, "lr": 5.274897306210558e-08}
{"train_loss": 0.07276734709739685, "global_step": 267488, "epoch": 3005, "lr": 5.27223536551158e-08}
{"train_loss": 0.03269296512007713, "global_step": 267489, "epoch": 3005, "lr": 5.2695740962926954e-08}
{"train_loss": 0.024323564022779465, "global_step": 267490, "epoch": 3005, "lr": 5.266913498555015e-08}
{"train_loss": 0.03823354095220566, "global_step": 267491, "epoch": 3005, "lr": 5.264253572297428e-08}
{"train_loss": 0.029491446912288666, "global_step": 267492, "epoch": 3005, "lr": 5.261594317522156e-08}
{"train_loss": 0.010541264899075031, "global_step": 267493, "epoch": 3005, "lr": 5.258935734227533e-08}
{"train_loss": 0.025945108383893967, "global_step": 267494, "epoch": 3005, "lr": 5.256277822415778e-08}
{"train_loss": 0.06006322801113129, "global_step": 267495, "epoch": 3005, "lr": 5.2536205820863385e-08}
{"train_loss": 0.05907459557056427, "global_step": 267496, "epoch": 3005, "lr": 5.2509640132392126e-08}
{"train_loss": 0.052910078316926956, "global_step": 267497, "epoch": 3005, "lr": 5.248308115875511e-08}
{"train_loss": 0.026929402723908424, "global_step": 267498, "epoch": 3005, "lr": 5.245652889995234e-08}
{"train_loss": 0.04294511675834656, "global_step": 267499, "epoch": 3005, "lr": 5.2429983355989364e-08}
{"train_loss": 0.04618477448821068, "global_step": 267500, "epoch": 3005, "lr": 5.2403444526866185e-08}
{"train_loss": 0.03849572315812111, "global_step": 267501, "epoch": 3005, "lr": 5.23769124125939e-08}
{"train_loss": 0.054945509880781174, "global_step": 267502, "epoch": 3005, "lr": 5.235038701316697e-08}
{"train_loss": 0.0306007768958807, "global_step": 267503, "epoch": 3005, "lr": 5.232386832859093e-08}
{"train_loss": 0.10163770616054535, "global_step": 267504, "epoch": 3005, "lr": 5.229735635887134e-08}
{"train_loss": 0.02803586795926094, "global_step": 267505, "epoch": 3005, "lr": 5.2270851104013754e-08}
{"train_loss": 0.041242815554142, "global_step": 267506, "epoch": 3005, "lr": 5.2244352564018164e-08}
{"train_loss": 0.018534639850258827, "global_step": 267507, "epoch": 3005, "lr": 5.221786073889012e-08}
{"train_loss": 0.028602031990885735, "global_step": 267508, "epoch": 3005, "lr": 5.219137562863519e-08}
{"train_loss": 0.021104929968714714, "global_step": 267509, "epoch": 3005, "lr": 5.2164897233253353e-08}
{"train_loss": 0.052621711045503616, "global_step": 267510, "epoch": 3005, "lr": 5.213842555274462e-08}
{"train_loss": 0.028397900983691216, "global_step": 267511, "epoch": 3005, "lr": 5.211196058712009e-08}
{"train_loss": 0.09179656207561493, "global_step": 267512, "epoch": 3005, "lr": 5.2085502336379766e-08}
{"train_loss": 0.05113844946026802, "global_step": 267513, "epoch": 3005, "lr": 5.2059050800529197e-08}
{"train_loss": 0.044849563390016556, "global_step": 267514, "epoch": 3005, "lr": 5.203260597956838e-08}
{"train_loss": 0.05559256672859192, "global_step": 267515, "epoch": 3005, "lr": 5.2006167873508424e-08}
{"train_loss": 0.039395470172166824, "global_step": 267516, "epoch": 3005, "lr": 5.197973648234378e-08}
{"train_loss": 0.06349803507328033, "global_step": 267517, "epoch": 3005, "lr": 5.195331180607999e-08}
{"train_loss": 0.042514629662036896, "global_step": 267518, "epoch": 3005, "lr": 5.19268938447226e-08}
{"train_loss": 0.04556717723608017, "global_step": 267519, "epoch": 3005, "lr": 5.190048259827718e-08}
{"train_loss": 0.09365111589431763, "global_step": 267520, "epoch": 3005, "lr": 5.187407806674371e-08}
{"train_loss": 0.0466291606426239, "global_step": 267521, "epoch": 3005, "lr": 5.184768025012776e-08}
{"train_loss": 0.05099497362971306, "global_step": 267522, "epoch": 3005, "lr": 5.1821289148429316e-08}
{"train_loss": 0.04171611741185188, "global_step": 267523, "epoch": 3005, "lr": 5.179490476165394e-08}
{"train_loss": 0.08148262649774551, "global_step": 267524, "epoch": 3005, "lr": 5.1768527089812726e-08}
{"train_loss": 0.060434740036726, "global_step": 267525, "epoch": 3005, "lr": 5.174215613289457e-08}
{"train_loss": 0.03193478658795357, "global_step": 267526, "epoch": 3005, "lr": 5.171579189091613e-08}
{"train_loss": 0.03618262708187103, "global_step": 267527, "epoch": 3005, "lr": 5.1689434363877406e-08}
{"train_loss": 0.074557825922966, "global_step": 267528, "epoch": 3005, "lr": 5.1663083551772853e-08}
{"train_loss": 0.0484805665910244, "global_step": 267529, "epoch": 3005, "lr": 5.163673945461911e-08}
{"train_loss": 0.012366024777293205, "global_step": 267530, "epoch": 3005, "lr": 5.161040207241619e-08}
{"train_loss": 0.05802152305841446, "global_step": 267531, "epoch": 3005, "lr": 5.158407140515853e-08}
{"train_loss": 0.01953575760126114, "global_step": 267532, "epoch": 3005, "lr": 5.155774745286279e-08}
{"train_loss": 0.0451334851553266, "global_step": 267533, "epoch": 3005, "lr": 5.153143021552342e-08, "val_loss": 9.63074779510498, "train_action_mse_error": 14.599868774414062}
{"train_loss": 0.06431818753480911, "global_step": 267534, "epoch": 3006, "lr": 5.150511969314598e-08}
{"train_loss": 0.06558755785226822, "global_step": 267535, "epoch": 3006, "lr": 5.1478815885736e-08}
{"train_loss": 0.06866468489170074, "global_step": 267536, "epoch": 3006, "lr": 5.1452518793293494e-08}
{"train_loss": 0.030840201303362846, "global_step": 267537, "epoch": 3006, "lr": 5.142622841582956e-08}
{"train_loss": 0.0325935035943985, "global_step": 267538, "epoch": 3006, "lr": 5.139994475333865e-08}
{"train_loss": 0.02451702579855919, "global_step": 267539, "epoch": 3006, "lr": 5.137366780582631e-08}
{"train_loss": 0.04154244810342789, "global_step": 267540, "epoch": 3006, "lr": 5.134739757330364e-08}
{"train_loss": 0.03334396332502365, "global_step": 267541, "epoch": 3006, "lr": 5.1321134055765105e-08}
{"train_loss": 0.08776512742042542, "global_step": 267542, "epoch": 3006, "lr": 5.12948772532218e-08}
{"train_loss": 0.03688183054327965, "global_step": 267543, "epoch": 3006, "lr": 5.126862716566816e-08}
{"train_loss": 0.02918560802936554, "global_step": 267544, "epoch": 3006, "lr": 5.12423837931153e-08}
{"train_loss": 0.006527369376271963, "global_step": 267545, "epoch": 3006, "lr": 5.121614713556322e-08}
{"train_loss": 0.07104483246803284, "global_step": 267546, "epoch": 3006, "lr": 5.1189917193011916e-08}
{"train_loss": 0.05490799620747566, "global_step": 267547, "epoch": 3006, "lr": 5.116369396547804e-08}
{"train_loss": 0.032870057970285416, "global_step": 267548, "epoch": 3006, "lr": 5.11374774529505e-08}
{"train_loss": 0.04456215351819992, "global_step": 267549, "epoch": 3006, "lr": 5.1111267655440384e-08}
{"train_loss": 0.06875908374786377, "global_step": 267550, "epoch": 3006, "lr": 5.108506457294771e-08}
{"train_loss": 0.050419729202985764, "global_step": 267551, "epoch": 3006, "lr": 5.105886820547801e-08}
{"train_loss": 0.03680950775742531, "global_step": 267552, "epoch": 3006, "lr": 5.103267855303684e-08}
{"train_loss": 0.019500261172652245, "global_step": 267553, "epoch": 3006, "lr": 5.1006495615624204e-08}
{"train_loss": 0.03700895980000496, "global_step": 267554, "epoch": 3006, "lr": 5.098031939325121e-08}
{"train_loss": 0.05273301899433136, "global_step": 267555, "epoch": 3006, "lr": 5.095414988590674e-08}
{"train_loss": 0.04530950263142586, "global_step": 267556, "epoch": 3006, "lr": 5.092798709360747e-08}
{"train_loss": 0.0516238659620285, "global_step": 267557, "epoch": 3006, "lr": 5.0901831016347826e-08}
{"train_loss": 0.07114749401807785, "global_step": 267558, "epoch": 3006, "lr": 5.087568165413892e-08}
{"train_loss": 0.02846320904791355, "global_step": 267559, "epoch": 3006, "lr": 5.084953900698075e-08}
{"train_loss": 0.0437496043741703, "global_step": 267560, "epoch": 3006, "lr": 5.082340307487332e-08}
{"train_loss": 0.04131046310067177, "global_step": 267561, "epoch": 3006, "lr": 5.079727385782773e-08}
{"train_loss": 0.05006147548556328, "global_step": 267562, "epoch": 3006, "lr": 5.0771151355843984e-08}
{"train_loss": 0.057189006358385086, "global_step": 267563, "epoch": 3006, "lr": 5.074503556892207e-08}
{"train_loss": 0.04488534480333328, "global_step": 267564, "epoch": 3006, "lr": 5.071892649707311e-08}
{"train_loss": 0.044629283249378204, "global_step": 267565, "epoch": 3006, "lr": 5.0692824140291525e-08}
{"train_loss": 0.04747937247157097, "global_step": 267566, "epoch": 3006, "lr": 5.066672849858845e-08}
{"train_loss": 0.04401577264070511, "global_step": 267567, "epoch": 3006, "lr": 5.064063957196385e-08}
{"train_loss": 0.014885387383401394, "global_step": 267568, "epoch": 3006, "lr": 5.061455736042331e-08}
{"train_loss": 0.054467979818582535, "global_step": 267569, "epoch": 3006, "lr": 5.058848186396681e-08}
{"train_loss": 0.028402728959918022, "global_step": 267570, "epoch": 3006, "lr": 5.05624130825999e-08}
{"train_loss": 0.030856717377901077, "global_step": 267571, "epoch": 3006, "lr": 5.053635101632814e-08}
{"train_loss": 0.018623866140842438, "global_step": 267572, "epoch": 3006, "lr": 5.051029566515153e-08}
{"train_loss": 0.022819338366389275, "global_step": 267573, "epoch": 3006, "lr": 5.048424702907562e-08}
{"train_loss": 0.03851400688290596, "global_step": 267574, "epoch": 3006, "lr": 5.04582051081004e-08}
{"train_loss": 0.01879425346851349, "global_step": 267575, "epoch": 3006, "lr": 5.043216990223143e-08}
{"train_loss": 0.022916510701179504, "global_step": 267576, "epoch": 3006, "lr": 5.0406141411479815e-08}
{"train_loss": 0.025922873988747597, "global_step": 267577, "epoch": 3006, "lr": 5.038011963583999e-08}
{"train_loss": 0.05885623022913933, "global_step": 267578, "epoch": 3006, "lr": 5.035410457531198e-08}
{"train_loss": 0.014999561943113804, "global_step": 267579, "epoch": 3006, "lr": 5.032809622991241e-08}
{"train_loss": 0.057510290294885635, "global_step": 267580, "epoch": 3006, "lr": 5.030209459963575e-08}
{"train_loss": 0.059737011790275574, "global_step": 267581, "epoch": 3006, "lr": 5.027609968448199e-08}
{"train_loss": 0.01628274843096733, "global_step": 267582, "epoch": 3006, "lr": 5.025011148446779e-08}
{"train_loss": 0.02698325552046299, "global_step": 267583, "epoch": 3006, "lr": 5.022412999958204e-08}
{"train_loss": 0.02333211712539196, "global_step": 267584, "epoch": 3006, "lr": 5.019815522983584e-08}
{"train_loss": 0.03361224755644798, "global_step": 267585, "epoch": 3006, "lr": 5.017218717523475e-08}
{"train_loss": 0.025040093809366226, "global_step": 267586, "epoch": 3006, "lr": 5.014622583577877e-08}
{"train_loss": 0.06547273695468903, "global_step": 267587, "epoch": 3006, "lr": 5.0120271211467895e-08}
{"train_loss": 0.02545611932873726, "global_step": 267588, "epoch": 3006, "lr": 5.009432330231323e-08}
{"train_loss": 0.04814015328884125, "global_step": 267589, "epoch": 3006, "lr": 5.006838210831477e-08}
{"train_loss": 0.08343105763196945, "global_step": 267590, "epoch": 3006, "lr": 5.004244762947252e-08}
{"train_loss": 0.013995268382132053, "global_step": 267591, "epoch": 3006, "lr": 5.0016519865797585e-08}
{"train_loss": 0.06776341795921326, "global_step": 267592, "epoch": 3006, "lr": 4.999059881728441e-08}
{"train_loss": 0.040449049323797226, "global_step": 267593, "epoch": 3006, "lr": 4.996468448394409e-08}
{"train_loss": 0.03911025449633598, "global_step": 267594, "epoch": 3006, "lr": 4.993877686577664e-08}
{"train_loss": 0.016821205615997314, "global_step": 267595, "epoch": 3006, "lr": 4.99128759627876e-08}
{"train_loss": 0.020494991913437843, "global_step": 267596, "epoch": 3006, "lr": 4.988698177497697e-08}
{"train_loss": 0.027777323499321938, "global_step": 267597, "epoch": 3006, "lr": 4.9861094302350306e-08}
{"train_loss": 0.06193757802248001, "global_step": 267598, "epoch": 3006, "lr": 4.9835213544913164e-08}
{"train_loss": 0.04630529135465622, "global_step": 267599, "epoch": 3006, "lr": 4.9809339502665533e-08}
{"train_loss": 0.033676180988550186, "global_step": 267600, "epoch": 3006, "lr": 4.9783472175612967e-08}
{"train_loss": 0.038799650967121124, "global_step": 267601, "epoch": 3006, "lr": 4.975761156375547e-08}
{"train_loss": 0.06036073714494705, "global_step": 267602, "epoch": 3006, "lr": 4.973175766710414e-08}
{"train_loss": 0.02948508784174919, "global_step": 267603, "epoch": 3006, "lr": 4.970591048565343e-08}
{"train_loss": 0.042245786637067795, "global_step": 267604, "epoch": 3006, "lr": 4.968007001940889e-08}
{"train_loss": 0.03886805847287178, "global_step": 267605, "epoch": 3006, "lr": 4.9654236268381613e-08}
{"train_loss": 0.01188054122030735, "global_step": 267606, "epoch": 3006, "lr": 4.9628409232566065e-08}
{"train_loss": 0.01337268203496933, "global_step": 267607, "epoch": 3006, "lr": 4.9602588911973334e-08}
{"train_loss": 0.04110543429851532, "global_step": 267608, "epoch": 3006, "lr": 4.957677530659788e-08}
{"train_loss": 0.019937029108405113, "global_step": 267609, "epoch": 3006, "lr": 4.955096841645079e-08}
{"train_loss": 0.04683493822813034, "global_step": 267610, "epoch": 3006, "lr": 4.952516824153208e-08}
{"train_loss": 0.04039423540234566, "global_step": 267611, "epoch": 3006, "lr": 4.9499374781841744e-08}
{"train_loss": 0.05306465923786163, "global_step": 267612, "epoch": 3006, "lr": 4.9473588037390884e-08}
{"train_loss": 0.07780724763870239, "global_step": 267613, "epoch": 3006, "lr": 4.94478080081795e-08}
{"train_loss": 0.04044535383582115, "global_step": 267614, "epoch": 3006, "lr": 4.942203469421314e-08}
{"train_loss": 0.041591040790081024, "global_step": 267615, "epoch": 3006, "lr": 4.93962680954918e-08}
{"train_loss": 0.04039420560002327, "global_step": 267616, "epoch": 3006, "lr": 4.937050821202105e-08}
{"train_loss": 0.043980441987514496, "global_step": 267617, "epoch": 3006, "lr": 4.9344755043800875e-08}
{"train_loss": 0.051702141761779785, "global_step": 267618, "epoch": 3006, "lr": 4.9319008590842375e-08}
{"train_loss": 0.04384797811508179, "global_step": 267619, "epoch": 3006, "lr": 4.9293268853140005e-08}
{"train_loss": 0.06605462729930878, "global_step": 267620, "epoch": 3006, "lr": 4.9267535830699315e-08}
{"train_loss": 0.013590747490525246, "global_step": 267621, "epoch": 3006, "lr": 4.924180952353141e-08}
{"train_loss": 0.04064603548057461, "global_step": 267622, "epoch": 3006, "lr": 4.921608993163074e-08, "val_loss": 9.619994163513184}
{"train_loss": 0.02870520018041134, "global_step": 267623, "epoch": 3007, "lr": 4.919037705500284e-08}
{"train_loss": 0.04171794280409813, "global_step": 267624, "epoch": 3007, "lr": 4.916467089365884e-08}
{"train_loss": 0.04840104281902313, "global_step": 267625, "epoch": 3007, "lr": 4.913897144758761e-08}
{"train_loss": 0.031739406287670135, "global_step": 267626, "epoch": 3007, "lr": 4.911327871680582e-08}
{"train_loss": 0.03329192474484444, "global_step": 267627, "epoch": 3007, "lr": 4.9087592701307915e-08}
{"train_loss": 0.061796288937330246, "global_step": 267628, "epoch": 3007, "lr": 4.9061913401104996e-08}
{"train_loss": 0.04546958953142166, "global_step": 267629, "epoch": 3007, "lr": 4.9036240816197064e-08}
{"train_loss": 0.042346782982349396, "global_step": 267630, "epoch": 3007, "lr": 4.901057494658967e-08}
{"train_loss": 0.07168345898389816, "global_step": 267631, "epoch": 3007, "lr": 4.8984915792277264e-08}
{"train_loss": 0.025576507672667503, "global_step": 267632, "epoch": 3007, "lr": 4.8959263353276494e-08}
{"train_loss": 0.011834266595542431, "global_step": 267633, "epoch": 3007, "lr": 4.8933617629581816e-08}
{"train_loss": 0.03896572068333626, "global_step": 267634, "epoch": 3007, "lr": 4.890797862119878e-08}
{"train_loss": 0.08905380964279175, "global_step": 267635, "epoch": 3007, "lr": 4.888234632812738e-08}
{"train_loss": 0.026277055963873863, "global_step": 267636, "epoch": 3007, "lr": 4.885672075037873e-08}
{"train_loss": 0.02903050184249878, "global_step": 267637, "epoch": 3007, "lr": 4.8831101887952816e-08}
{"train_loss": 0.012000035494565964, "global_step": 267638, "epoch": 3007, "lr": 4.880548974085519e-08}
{"train_loss": 0.05996328219771385, "global_step": 267639, "epoch": 3007, "lr": 4.877988430908032e-08}
{"train_loss": 0.09137468039989471, "global_step": 267640, "epoch": 3007, "lr": 4.8754285592644835e-08}
{"train_loss": 0.01400847639888525, "global_step": 267641, "epoch": 3007, "lr": 4.87286935915432e-08}
{"train_loss": 0.0526091568171978, "global_step": 267642, "epoch": 3007, "lr": 4.8703108305780956e-08}
{"train_loss": 0.01626824028789997, "global_step": 267643, "epoch": 3007, "lr": 4.8677529735358105e-08}
{"train_loss": 0.04269237816333771, "global_step": 267644, "epoch": 3007, "lr": 4.865195788028576e-08}
{"train_loss": 0.023196518421173096, "global_step": 267645, "epoch": 3007, "lr": 4.862639274056391e-08}
{"train_loss": 0.067886583507061, "global_step": 267646, "epoch": 3007, "lr": 4.8600834316192555e-08}
{"train_loss": 0.03684063255786896, "global_step": 267647, "epoch": 3007, "lr": 4.8575282607182804e-08}
{"train_loss": 0.03869197890162468, "global_step": 267648, "epoch": 3007, "lr": 4.85497376135291e-08}
{"train_loss": 0.07545915991067886, "global_step": 267649, "epoch": 3007, "lr": 4.8524199335237e-08}
{"train_loss": 0.024228213354945183, "global_step": 267650, "epoch": 3007, "lr": 4.849866777231759e-08}
{"train_loss": 0.03912919759750366, "global_step": 267651, "epoch": 3007, "lr": 4.8473142924765345e-08}
{"train_loss": 0.038788653910160065, "global_step": 267652, "epoch": 3007, "lr": 4.84476247925858e-08}
{"train_loss": 0.04900258779525757, "global_step": 267653, "epoch": 3007, "lr": 4.842211337578451e-08}
{"train_loss": 0.046013038605451584, "global_step": 267654, "epoch": 3007, "lr": 4.8396608674367015e-08}
{"train_loss": 0.027122385799884796, "global_step": 267655, "epoch": 3007, "lr": 4.837111068833333e-08}
{"train_loss": 0.04597756266593933, "global_step": 267656, "epoch": 3007, "lr": 4.834561941768345e-08}
{"train_loss": 0.05952651426196098, "global_step": 267657, "epoch": 3007, "lr": 4.832013486242848e-08}
{"train_loss": 0.02474486082792282, "global_step": 267658, "epoch": 3007, "lr": 4.8294657022568415e-08}
{"train_loss": 0.04954966530203819, "global_step": 267659, "epoch": 3007, "lr": 4.8269185898103253e-08}
{"train_loss": 0.048095423728227615, "global_step": 267660, "epoch": 3007, "lr": 4.82437214890441e-08}
{"train_loss": 0.051919057965278625, "global_step": 267661, "epoch": 3007, "lr": 4.821826379538541e-08}
{"train_loss": 0.03203507140278816, "global_step": 267662, "epoch": 3007, "lr": 4.8192812817132724e-08}
{"train_loss": 0.038217537105083466, "global_step": 267663, "epoch": 3007, "lr": 4.816736855429715e-08}
{"train_loss": 0.03853963315486908, "global_step": 267664, "epoch": 3007, "lr": 4.814193100687314e-08}
{"train_loss": 0.0225595161318779, "global_step": 267665, "epoch": 3007, "lr": 4.811650017487179e-08}
{"train_loss": 0.0347459577023983, "global_step": 267666, "epoch": 3007, "lr": 4.8091076058287555e-08}
{"train_loss": 0.03370053693652153, "global_step": 267667, "epoch": 3007, "lr": 4.806565865713153e-08}
{"train_loss": 0.04077385738492012, "global_step": 267668, "epoch": 3007, "lr": 4.8040247971403716e-08}
{"train_loss": 0.07234863191843033, "global_step": 267669, "epoch": 3007, "lr": 4.801484400110412e-08}
{"train_loss": 0.01713070087134838, "global_step": 267670, "epoch": 3007, "lr": 4.798944674624384e-08}
{"train_loss": 0.06210450828075409, "global_step": 267671, "epoch": 3007, "lr": 4.796405620682287e-08}
{"train_loss": 0.01775369979441166, "global_step": 267672, "epoch": 3007, "lr": 4.7938672382846774e-08}
{"train_loss": 0.03854401409626007, "global_step": 267673, "epoch": 3007, "lr": 4.791329527430999e-08}
{"train_loss": 0.06137210130691528, "global_step": 267674, "epoch": 3007, "lr": 4.788792488122918e-08}
{"train_loss": 0.03920358419418335, "global_step": 267675, "epoch": 3007, "lr": 4.7862561203598774e-08}
{"train_loss": 0.03464437276124954, "global_step": 267676, "epoch": 3007, "lr": 4.7837204241418795e-08}
{"train_loss": 0.045293085277080536, "global_step": 267677, "epoch": 3007, "lr": 4.781185399470589e-08}
{"train_loss": 0.028727801516652107, "global_step": 267678, "epoch": 3007, "lr": 4.7786510463454506e-08}
{"train_loss": 0.016797466203570366, "global_step": 267679, "epoch": 3007, "lr": 4.776117364766464e-08}
{"train_loss": 0.036513932049274445, "global_step": 267680, "epoch": 3007, "lr": 4.7735843547347394e-08}
{"train_loss": 0.08278576284646988, "global_step": 267681, "epoch": 3007, "lr": 4.771052016250277e-08}
{"train_loss": 0.07084739208221436, "global_step": 267682, "epoch": 3007, "lr": 4.7685203493136323e-08}
{"train_loss": 0.03604516014456749, "global_step": 267683, "epoch": 3007, "lr": 4.765989353924805e-08}
{"train_loss": 0.03365709260106087, "global_step": 267684, "epoch": 3007, "lr": 4.7634590300843495e-08}
{"train_loss": 0.06671985983848572, "global_step": 267685, "epoch": 3007, "lr": 4.760929377792822e-08}
{"train_loss": 0.029485588893294334, "global_step": 267686, "epoch": 3007, "lr": 4.758400397050222e-08}
{"train_loss": 0.0395696759223938, "global_step": 267687, "epoch": 3007, "lr": 4.75587208785655e-08}
{"train_loss": 0.006821149494498968, "global_step": 267688, "epoch": 3007, "lr": 4.7533444502129154e-08}
{"train_loss": 0.0680960938334465, "global_step": 267689, "epoch": 3007, "lr": 4.7508174841193185e-08}
{"train_loss": 0.053872834891080856, "global_step": 267690, "epoch": 3007, "lr": 4.7482911895757595e-08}
{"train_loss": 0.050161246210336685, "global_step": 267691, "epoch": 3007, "lr": 4.745765566583349e-08}
{"train_loss": 0.031859077513217926, "global_step": 267692, "epoch": 3007, "lr": 4.7432406151415306e-08}
{"train_loss": 0.02693886309862137, "global_step": 267693, "epoch": 3007, "lr": 4.740716335251416e-08}
{"train_loss": 0.025897232815623283, "global_step": 267694, "epoch": 3007, "lr": 4.738192726913004e-08}
{"train_loss": 0.03293963521718979, "global_step": 267695, "epoch": 3007, "lr": 4.735669790126851e-08}
{"train_loss": 0.03791055083274841, "global_step": 267696, "epoch": 3007, "lr": 4.7331475248929554e-08}
{"train_loss": 0.05328982323408127, "global_step": 267697, "epoch": 3007, "lr": 4.7306259312113185e-08}
{"train_loss": 0.021394934505224228, "global_step": 267698, "epoch": 3007, "lr": 4.728105009083605e-08}
{"train_loss": 0.02404008060693741, "global_step": 267699, "epoch": 3007, "lr": 4.725584758508705e-08}
{"train_loss": 0.031913939863443375, "global_step": 267700, "epoch": 3007, "lr": 4.723065179487729e-08}
{"train_loss": 0.05792493745684624, "global_step": 267701, "epoch": 3007, "lr": 4.7205462720206763e-08}
{"train_loss": 0.07092998921871185, "global_step": 267702, "epoch": 3007, "lr": 4.718028036108657e-08}
{"train_loss": 0.02641657181084156, "global_step": 267703, "epoch": 3007, "lr": 4.7155104717505616e-08}
{"train_loss": 0.08132736384868622, "global_step": 267704, "epoch": 3007, "lr": 4.712993578948055e-08}
{"train_loss": 0.022164825350046158, "global_step": 267705, "epoch": 3007, "lr": 4.710477357701137e-08}
{"train_loss": 0.021541191264986992, "global_step": 267706, "epoch": 3007, "lr": 4.707961808009809e-08}
{"train_loss": 0.0498540997505188, "global_step": 267707, "epoch": 3007, "lr": 4.705446929874624e-08}
{"train_loss": 0.021863479167222977, "global_step": 267708, "epoch": 3007, "lr": 4.702932723296138e-08}
{"train_loss": 0.03905235230922699, "global_step": 267709, "epoch": 3007, "lr": 4.700419188274352e-08}
{"train_loss": 0.03843320906162262, "global_step": 267710, "epoch": 3007, "lr": 4.6979063248092645e-08}
{"train_loss": 0.040991903706506065, "global_step": 267711, "epoch": 3007, "lr": 4.6953941329019866e-08, "val_loss": 9.651395797729492}
{"train_loss": 0.045094430446624756, "global_step": 267712, "epoch": 3008, "lr": 4.692882612552518e-08}
{"train_loss": 0.037676289677619934, "global_step": 267713, "epoch": 3008, "lr": 4.690371763761414e-08}
{"train_loss": 0.017796870321035385, "global_step": 267714, "epoch": 3008, "lr": 4.6878615865286744e-08}
{"train_loss": 0.03717256337404251, "global_step": 267715, "epoch": 3008, "lr": 4.6853520808548544e-08}
{"train_loss": 0.04514424502849579, "global_step": 267716, "epoch": 3008, "lr": 4.682843246739954e-08}
{"train_loss": 0.08683433383703232, "global_step": 267717, "epoch": 3008, "lr": 4.6803350841845285e-08}
{"train_loss": 0.06271813064813614, "global_step": 267718, "epoch": 3008, "lr": 4.6778275931891324e-08}
{"train_loss": 0.047123368829488754, "global_step": 267719, "epoch": 3008, "lr": 4.675320773754322e-08}
{"train_loss": 0.06450875103473663, "global_step": 267720, "epoch": 3008, "lr": 4.67281462587954e-08}
{"train_loss": 0.02499522641301155, "global_step": 267721, "epoch": 3008, "lr": 4.6703091495653436e-08}
{"train_loss": 0.047498904168605804, "global_step": 267722, "epoch": 3008, "lr": 4.6678043448128425e-08}
{"train_loss": 0.04424582049250603, "global_step": 267723, "epoch": 3008, "lr": 4.6653002116214815e-08}
{"train_loss": 0.037262920290231705, "global_step": 267724, "epoch": 3008, "lr": 4.662796749992371e-08}
{"train_loss": 0.019105590879917145, "global_step": 267725, "epoch": 3008, "lr": 4.6602939599255105e-08}
{"train_loss": 0.030680492520332336, "global_step": 267726, "epoch": 3008, "lr": 4.6577918414209e-08}
{"train_loss": 0.045165594667196274, "global_step": 267727, "epoch": 3008, "lr": 4.655290394479095e-08}
{"train_loss": 0.055100180208683014, "global_step": 267728, "epoch": 3008, "lr": 4.6527896191006503e-08}
{"train_loss": 0.03556947410106659, "global_step": 267729, "epoch": 3008, "lr": 4.6502895152855664e-08}
{"train_loss": 0.06102412939071655, "global_step": 267730, "epoch": 3008, "lr": 4.647790083034953e-08}
{"train_loss": 0.06614954769611359, "global_step": 267731, "epoch": 3008, "lr": 4.6452913223477e-08}
{"train_loss": 0.03620009869337082, "global_step": 267732, "epoch": 3008, "lr": 4.642793233225473e-08}
{"train_loss": 0.03272610157728195, "global_step": 267733, "epoch": 3008, "lr": 4.640295815668272e-08}
{"train_loss": 0.10023651272058487, "global_step": 267734, "epoch": 3008, "lr": 4.637799069676096e-08}
{"train_loss": 0.024819692596793175, "global_step": 267735, "epoch": 3008, "lr": 4.635302995249502e-08}
{"train_loss": 0.03838224709033966, "global_step": 267736, "epoch": 3008, "lr": 4.632807592389044e-08}
{"train_loss": 0.02745858207345009, "global_step": 267737, "epoch": 3008, "lr": 4.630312861094721e-08}
{"train_loss": 0.0582151785492897, "global_step": 267738, "epoch": 3008, "lr": 4.627818801366535e-08}
{"train_loss": 0.08159150928258896, "global_step": 267739, "epoch": 3008, "lr": 4.625325413205594e-08}
{"train_loss": 0.07424059510231018, "global_step": 267740, "epoch": 3008, "lr": 4.6228326966119003e-08}
{"train_loss": 0.036192476749420166, "global_step": 267741, "epoch": 3008, "lr": 4.620340651586008e-08}
{"train_loss": 0.05268551781773567, "global_step": 267742, "epoch": 3008, "lr": 4.617849278127917e-08}
{"train_loss": 0.023118769749999046, "global_step": 267743, "epoch": 3008, "lr": 4.6153585762376273e-08}
{"train_loss": 0.05248415470123291, "global_step": 267744, "epoch": 3008, "lr": 4.612868545916249e-08}
{"train_loss": 0.06415614485740662, "global_step": 267745, "epoch": 3008, "lr": 4.610379187163782e-08}
{"train_loss": 0.028309836983680725, "global_step": 267746, "epoch": 3008, "lr": 4.607890499980783e-08}
{"train_loss": 0.0713421180844307, "global_step": 267747, "epoch": 3008, "lr": 4.605402484366694e-08}
{"train_loss": 0.08694878965616226, "global_step": 267748, "epoch": 3008, "lr": 4.602915140323183e-08}
{"train_loss": 0.028744583949446678, "global_step": 267749, "epoch": 3008, "lr": 4.600428467849693e-08}
{"train_loss": 0.023302042856812477, "global_step": 267750, "epoch": 3008, "lr": 4.597942466946781e-08}
{"train_loss": 0.030309580266475677, "global_step": 267751, "epoch": 3008, "lr": 4.595457137615e-08}
{"train_loss": 0.01832403428852558, "global_step": 267752, "epoch": 3008, "lr": 4.592972479853797e-08}
{"train_loss": 0.03465835005044937, "global_step": 267753, "epoch": 3008, "lr": 4.5904884936648354e-08}
{"train_loss": 0.018614884465932846, "global_step": 267754, "epoch": 3008, "lr": 4.588005179047561e-08}
{"train_loss": 0.058642975986003876, "global_step": 267755, "epoch": 3008, "lr": 4.585522536002529e-08}
{"train_loss": 0.04681295156478882, "global_step": 267756, "epoch": 3008, "lr": 4.5830405645302946e-08}
{"train_loss": 0.0707642063498497, "global_step": 267757, "epoch": 3008, "lr": 4.580559264630857e-08}
{"train_loss": 0.0817364752292633, "global_step": 267758, "epoch": 3008, "lr": 4.578078636304772e-08}
{"train_loss": 0.059066593647003174, "global_step": 267759, "epoch": 3008, "lr": 4.57559867955204e-08}
{"train_loss": 0.03264316916465759, "global_step": 267760, "epoch": 3008, "lr": 4.5731193943732154e-08}
{"train_loss": 0.04290541633963585, "global_step": 267761, "epoch": 3008, "lr": 4.570640780768853e-08}
{"train_loss": 0.018430151045322418, "global_step": 267762, "epoch": 3008, "lr": 4.568162838739509e-08}
{"train_loss": 0.02744261734187603, "global_step": 267763, "epoch": 3008, "lr": 4.565685568284628e-08}
{"train_loss": 0.06107069551944733, "global_step": 267764, "epoch": 3008, "lr": 4.563208969404764e-08}
{"train_loss": 0.05240150913596153, "global_step": 267765, "epoch": 3008, "lr": 4.5607330421004735e-08}
{"train_loss": 0.04418296366930008, "global_step": 267766, "epoch": 3008, "lr": 4.558257786372311e-08}
{"train_loss": 0.008174166083335876, "global_step": 267767, "epoch": 3008, "lr": 4.555783202220831e-08}
{"train_loss": 0.025167040526866913, "global_step": 267768, "epoch": 3008, "lr": 4.5533092896454796e-08}
{"train_loss": 0.022311247885227203, "global_step": 267769, "epoch": 3008, "lr": 4.550836048647367e-08}
{"train_loss": 0.0293864905834198, "global_step": 267770, "epoch": 3008, "lr": 4.5483634792259364e-08}
{"train_loss": 0.020678944885730743, "global_step": 267771, "epoch": 3008, "lr": 4.5458915813823e-08}
{"train_loss": 0.042362045496702194, "global_step": 267772, "epoch": 3008, "lr": 4.543420355117012e-08}
{"train_loss": 0.07893601059913635, "global_step": 267773, "epoch": 3008, "lr": 4.540949800429517e-08}
{"train_loss": 0.06259641796350479, "global_step": 267774, "epoch": 3008, "lr": 4.5384799173209256e-08}
{"train_loss": 0.044625937938690186, "global_step": 267775, "epoch": 3008, "lr": 4.5360107057912385e-08}
{"train_loss": 0.051611628383398056, "global_step": 267776, "epoch": 3008, "lr": 4.533542165840454e-08}
{"train_loss": 0.016400879248976707, "global_step": 267777, "epoch": 3008, "lr": 4.5310742974691286e-08}
{"train_loss": 0.07967416197061539, "global_step": 267778, "epoch": 3008, "lr": 4.528607100678373e-08}
{"train_loss": 0.05556949973106384, "global_step": 267779, "epoch": 3008, "lr": 4.5261405754676304e-08}
{"train_loss": 0.08504270762205124, "global_step": 267780, "epoch": 3008, "lr": 4.523674721837456e-08}
{"train_loss": 0.017279164865612984, "global_step": 267781, "epoch": 3008, "lr": 4.5212095397878515e-08}
{"train_loss": 0.03267303854227066, "global_step": 267782, "epoch": 3008, "lr": 4.518745029319926e-08}
{"train_loss": 0.0355803519487381, "global_step": 267783, "epoch": 3008, "lr": 4.516281190433125e-08}
{"train_loss": 0.020763136446475983, "global_step": 267784, "epoch": 3008, "lr": 4.513818023128558e-08}
{"train_loss": 0.01975274458527565, "global_step": 267785, "epoch": 3008, "lr": 4.511355527405669e-08}
{"train_loss": 0.017823703587055206, "global_step": 267786, "epoch": 3008, "lr": 4.5088937032661264e-08}
{"train_loss": 0.017012931406497955, "global_step": 267787, "epoch": 3008, "lr": 4.5064325507088166e-08}
{"train_loss": 0.044284846633672714, "global_step": 267788, "epoch": 3008, "lr": 4.503972069735407e-08}
{"train_loss": 0.04701387509703636, "global_step": 267789, "epoch": 3008, "lr": 4.501512260345342e-08}
{"train_loss": 0.05670400708913803, "global_step": 267790, "epoch": 3008, "lr": 4.499053122538621e-08}
{"train_loss": 0.026409750804305077, "global_step": 267791, "epoch": 3008, "lr": 4.496594656316355e-08}
{"train_loss": 0.04063882306218147, "global_step": 267792, "epoch": 3008, "lr": 4.494136861679099e-08}
{"train_loss": 0.02274871990084648, "global_step": 267793, "epoch": 3008, "lr": 4.4916797386262975e-08}
{"train_loss": 0.022685013711452484, "global_step": 267794, "epoch": 3008, "lr": 4.489223287159061e-08}
{"train_loss": 0.052868157625198364, "global_step": 267795, "epoch": 3008, "lr": 4.486767507276834e-08}
{"train_loss": 0.02018173784017563, "global_step": 267796, "epoch": 3008, "lr": 4.484312398980728e-08}
{"train_loss": 0.02637643925845623, "global_step": 267797, "epoch": 3008, "lr": 4.4818579622707416e-08}
{"train_loss": 0.014262563548982143, "global_step": 267798, "epoch": 3008, "lr": 4.47940419714743e-08}
{"train_loss": 0.03391367942094803, "global_step": 267799, "epoch": 3008, "lr": 4.4769511036113484e-08}
{"train_loss": 0.04237313965188988, "global_step": 267800, "epoch": 3008, "lr": 4.4744986816619425e-08, "val_loss": 9.593656539916992}
{"train_loss": 0.025540215894579887, "global_step": 267801, "epoch": 3009, "lr": 4.472046931299767e-08}
{"train_loss": 0.04365886002779007, "global_step": 267802, "epoch": 3009, "lr": 4.469595852525932e-08}
{"train_loss": 0.03825289011001587, "global_step": 267803, "epoch": 3009, "lr": 4.467145445340437e-08}
{"train_loss": 0.031950581818819046, "global_step": 267804, "epoch": 3009, "lr": 4.464695709742728e-08}
{"train_loss": 0.07602942734956741, "global_step": 267805, "epoch": 3009, "lr": 4.462246645734469e-08}
{"train_loss": 0.05764404684305191, "global_step": 267806, "epoch": 3009, "lr": 4.459798253315106e-08}
{"train_loss": 0.027615545317530632, "global_step": 267807, "epoch": 3009, "lr": 4.457350532485749e-08}
{"train_loss": 0.04442615807056427, "global_step": 267808, "epoch": 3009, "lr": 4.4549034832458424e-08}
{"train_loss": 0.06678842753171921, "global_step": 267809, "epoch": 3009, "lr": 4.452457105595942e-08}
{"train_loss": 0.04229920729994774, "global_step": 267810, "epoch": 3009, "lr": 4.450011399536602e-08}
{"train_loss": 0.09020211547613144, "global_step": 267811, "epoch": 3009, "lr": 4.447566365067823e-08}
{"train_loss": 0.037307258695364, "global_step": 267812, "epoch": 3009, "lr": 4.445122002190716e-08}
{"train_loss": 0.02300785854458809, "global_step": 267813, "epoch": 3009, "lr": 4.4426783109047245e-08}
{"train_loss": 0.06877479702234268, "global_step": 267814, "epoch": 3009, "lr": 4.440235291210404e-08}
{"train_loss": 0.046827662736177444, "global_step": 267815, "epoch": 3009, "lr": 4.4377929431088646e-08}
{"train_loss": 0.04220864549279213, "global_step": 267816, "epoch": 3009, "lr": 4.435351266598997e-08}
{"train_loss": 0.028031475841999054, "global_step": 267817, "epoch": 3009, "lr": 4.432910261682466e-08}
{"train_loss": 0.04927093908190727, "global_step": 267818, "epoch": 3009, "lr": 4.430469928358716e-08}
{"train_loss": 0.04758346453309059, "global_step": 267819, "epoch": 3009, "lr": 4.4280302666283025e-08}
{"train_loss": 0.053138308227062225, "global_step": 267820, "epoch": 3009, "lr": 4.425591276491781e-08}
{"train_loss": 0.04692633077502251, "global_step": 267821, "epoch": 3009, "lr": 4.4231529579497056e-08}
{"train_loss": 0.03523538261651993, "global_step": 267822, "epoch": 3009, "lr": 4.420715311001522e-08}
{"train_loss": 0.08445313572883606, "global_step": 267823, "epoch": 3009, "lr": 4.4182783356483404e-08}
{"train_loss": 0.05925668403506279, "global_step": 267824, "epoch": 3009, "lr": 4.4158420318901604e-08}
{"train_loss": 0.02941988781094551, "global_step": 267825, "epoch": 3009, "lr": 4.413406399727538e-08}
{"train_loss": 0.050955381244421005, "global_step": 267826, "epoch": 3009, "lr": 4.4109714391604717e-08}
{"train_loss": 0.07628218829631805, "global_step": 267827, "epoch": 3009, "lr": 4.408537150189518e-08}
{"train_loss": 0.0762999951839447, "global_step": 267828, "epoch": 3009, "lr": 4.406103532815231e-08}
{"train_loss": 0.029069267213344574, "global_step": 267829, "epoch": 3009, "lr": 4.403670587037612e-08}
{"train_loss": 0.029292985796928406, "global_step": 267830, "epoch": 3009, "lr": 4.4012383128566595e-08}
{"train_loss": 0.04422205686569214, "global_step": 267831, "epoch": 3009, "lr": 4.398806710272929e-08}
{"train_loss": 0.027574511244893074, "global_step": 267832, "epoch": 3009, "lr": 4.3963757792875316e-08}
{"train_loss": 0.042167194187641144, "global_step": 267833, "epoch": 3009, "lr": 4.393945519899911e-08}
{"train_loss": 0.031849682331085205, "global_step": 267834, "epoch": 3009, "lr": 4.391515932110068e-08}
{"train_loss": 0.05161765590310097, "global_step": 267835, "epoch": 3009, "lr": 4.389087015919668e-08}
{"train_loss": 0.052138131111860275, "global_step": 267836, "epoch": 3009, "lr": 4.386658771327601e-08}
{"train_loss": 0.0290512852370739, "global_step": 267837, "epoch": 3009, "lr": 4.384231198335531e-08}
{"train_loss": 0.021130818873643875, "global_step": 267838, "epoch": 3009, "lr": 4.381804296942904e-08}
{"train_loss": 0.025653043761849403, "global_step": 267839, "epoch": 3009, "lr": 4.37937806714972e-08}
{"train_loss": 0.05966925621032715, "global_step": 267840, "epoch": 3009, "lr": 4.376952508957643e-08}
{"train_loss": 0.02876989170908928, "global_step": 267841, "epoch": 3009, "lr": 4.374527622365565e-08}
{"train_loss": 0.03892084211111069, "global_step": 267842, "epoch": 3009, "lr": 4.37210340737515e-08}
{"train_loss": 0.029171064496040344, "global_step": 267843, "epoch": 3009, "lr": 4.3696798639852876e-08}
{"train_loss": 0.04395449906587601, "global_step": 267844, "epoch": 3009, "lr": 4.367256992197644e-08}
{"train_loss": 0.07134091109037399, "global_step": 267845, "epoch": 3009, "lr": 4.364834792011663e-08}
{"train_loss": 0.030792100355029106, "global_step": 267846, "epoch": 3009, "lr": 4.3624132634279e-08}
{"train_loss": 0.025903915986418724, "global_step": 267847, "epoch": 3009, "lr": 4.359992406446356e-08}
{"train_loss": 0.037905480712652206, "global_step": 267848, "epoch": 3009, "lr": 4.35757222106814e-08}
{"train_loss": 0.06391724944114685, "global_step": 267849, "epoch": 3009, "lr": 4.355152707293253e-08}
{"train_loss": 0.04026948660612106, "global_step": 267850, "epoch": 3009, "lr": 4.352733865121694e-08}
{"train_loss": 0.008989290334284306, "global_step": 267851, "epoch": 3009, "lr": 4.350315694554019e-08}
{"train_loss": 0.04704812169075012, "global_step": 267852, "epoch": 3009, "lr": 4.347898195590783e-08}
{"train_loss": 0.03080814704298973, "global_step": 267853, "epoch": 3009, "lr": 4.345481368231985e-08}
{"train_loss": 0.03807203471660614, "global_step": 267854, "epoch": 3009, "lr": 4.343065212478181e-08}
{"train_loss": 0.041318655014038086, "global_step": 267855, "epoch": 3009, "lr": 4.340649728329371e-08}
{"train_loss": 0.04085327684879303, "global_step": 267856, "epoch": 3009, "lr": 4.33823491578611e-08}
{"train_loss": 0.037658900022506714, "global_step": 267857, "epoch": 3009, "lr": 4.335820774848398e-08}
{"train_loss": 0.05102978274226189, "global_step": 267858, "epoch": 3009, "lr": 4.333407305517345e-08}
{"train_loss": 0.037006620317697525, "global_step": 267859, "epoch": 3009, "lr": 4.330994507792396e-08}
{"train_loss": 0.0340723916888237, "global_step": 267860, "epoch": 3009, "lr": 4.3285823816741065e-08}
{"train_loss": 0.03190149366855621, "global_step": 267861, "epoch": 3009, "lr": 4.3261709271635866e-08}
{"train_loss": 0.03338766470551491, "global_step": 267862, "epoch": 3009, "lr": 4.32376014426028e-08}
{"train_loss": 0.03595888987183571, "global_step": 267863, "epoch": 3009, "lr": 4.321350032964189e-08}
{"train_loss": 0.03968213498592377, "global_step": 267864, "epoch": 3009, "lr": 4.3189405932769765e-08}
{"train_loss": 0.06442036479711533, "global_step": 267865, "epoch": 3009, "lr": 4.316531825198089e-08}
{"train_loss": 0.021125124767422676, "global_step": 267866, "epoch": 3009, "lr": 4.314123728727526e-08}
{"train_loss": 0.045981138944625854, "global_step": 267867, "epoch": 3009, "lr": 4.3117163038663974e-08}
{"train_loss": 0.04689912870526314, "global_step": 267868, "epoch": 3009, "lr": 4.309309550614149e-08}
{"train_loss": 0.02698642946779728, "global_step": 267869, "epoch": 3009, "lr": 4.3069034689724454e-08}
{"train_loss": 0.05415076017379761, "global_step": 267870, "epoch": 3009, "lr": 4.304498058940176e-08}
{"train_loss": 0.052136171609163284, "global_step": 267871, "epoch": 3009, "lr": 4.302093320518452e-08}
{"train_loss": 0.03137683495879173, "global_step": 267872, "epoch": 3009, "lr": 4.299689253707273e-08}
{"train_loss": 0.07607556879520416, "global_step": 267873, "epoch": 3009, "lr": 4.297285858507749e-08}
{"train_loss": 0.03920700401067734, "global_step": 267874, "epoch": 3009, "lr": 4.2948831349187703e-08}
{"train_loss": 0.07089033722877502, "global_step": 267875, "epoch": 3009, "lr": 4.2924810829420017e-08}
{"train_loss": 0.03023940697312355, "global_step": 267876, "epoch": 3009, "lr": 4.290079702576888e-08}
{"train_loss": 0.018514610826969147, "global_step": 267877, "epoch": 3009, "lr": 4.287678993823985e-08}
{"train_loss": 0.045908939093351364, "global_step": 267878, "epoch": 3009, "lr": 4.285278956683847e-08}
{"train_loss": 0.03914881497621536, "global_step": 267879, "epoch": 3009, "lr": 4.28287959115703e-08}
{"train_loss": 0.03137059509754181, "global_step": 267880, "epoch": 3009, "lr": 4.280480897242978e-08}
{"train_loss": 0.06424273550510406, "global_step": 267881, "epoch": 3009, "lr": 4.278082874942801e-08}
{"train_loss": 0.01620689406991005, "global_step": 267882, "epoch": 3009, "lr": 4.275685524256501e-08}
{"train_loss": 0.049348875880241394, "global_step": 267883, "epoch": 3009, "lr": 4.273288845184631e-08}
{"train_loss": 0.0400688573718071, "global_step": 267884, "epoch": 3009, "lr": 4.2708928377271915e-08}
{"train_loss": 0.036548662930727005, "global_step": 267885, "epoch": 3009, "lr": 4.2684975018847386e-08}
{"train_loss": 0.013442836701869965, "global_step": 267886, "epoch": 3009, "lr": 4.2661028376572706e-08}
{"train_loss": 0.04640452563762665, "global_step": 267887, "epoch": 3009, "lr": 4.2637088450453446e-08}
{"train_loss": 0.0405023954808712, "global_step": 267888, "epoch": 3009, "lr": 4.261315524049514e-08}
{"train_loss": 0.042406475642340236, "global_step": 267889, "epoch": 3009, "lr": 4.2589228746692245e-08, "val_loss": 9.569748878479004}
{"train_loss": 0.030414216220378876, "global_step": 267890, "epoch": 3010, "lr": 4.2565308969061414e-08}
{"train_loss": 0.03006601706147194, "global_step": 267891, "epoch": 3010, "lr": 4.2541395907597094e-08}
{"train_loss": 0.03658702224493027, "global_step": 267892, "epoch": 3010, "lr": 4.251748956230484e-08}
{"train_loss": 0.03358397260308266, "global_step": 267893, "epoch": 3010, "lr": 4.249358993318464e-08}
{"train_loss": 0.03416908532381058, "global_step": 267894, "epoch": 3010, "lr": 4.246969702024206e-08}
{"train_loss": 0.06863750517368317, "global_step": 267895, "epoch": 3010, "lr": 4.2445810823482646e-08}
{"train_loss": 0.05848286673426628, "global_step": 267896, "epoch": 3010, "lr": 4.2421931342906395e-08}
{"train_loss": 0.010917962528765202, "global_step": 267897, "epoch": 3010, "lr": 4.239805857851886e-08}
{"train_loss": 0.028681214898824692, "global_step": 267898, "epoch": 3010, "lr": 4.2374192530320044e-08}
{"train_loss": 0.03778214007616043, "global_step": 267899, "epoch": 3010, "lr": 4.2350333198315495e-08}
{"train_loss": 0.03426861763000488, "global_step": 267900, "epoch": 3010, "lr": 4.2326480582510764e-08}
{"train_loss": 0.03772151470184326, "global_step": 267901, "epoch": 3010, "lr": 4.23026346829003e-08}
{"train_loss": 0.026120217517018318, "global_step": 267902, "epoch": 3010, "lr": 4.227879549950076e-08}
{"train_loss": 0.049949418753385544, "global_step": 267903, "epoch": 3010, "lr": 4.2254963032301034e-08}
{"train_loss": 0.014165666885674, "global_step": 267904, "epoch": 3010, "lr": 4.223113728131778e-08}
{"train_loss": 0.046672120690345764, "global_step": 267905, "epoch": 3010, "lr": 4.220731824654545e-08}
{"train_loss": 0.035052698105573654, "global_step": 267906, "epoch": 3010, "lr": 4.218350592798403e-08}
{"train_loss": 0.06254318356513977, "global_step": 267907, "epoch": 3010, "lr": 4.21597003256502e-08}
{"train_loss": 0.04178513586521149, "global_step": 267908, "epoch": 3010, "lr": 4.2135901439532833e-08}
{"train_loss": 0.014045261777937412, "global_step": 267909, "epoch": 3010, "lr": 4.211210926964859e-08}
{"train_loss": 0.02854854427278042, "global_step": 267910, "epoch": 3010, "lr": 4.208832381598638e-08}
{"train_loss": 0.03463830053806305, "global_step": 267911, "epoch": 3010, "lr": 4.206454507855728e-08}
{"train_loss": 0.06039192155003548, "global_step": 267912, "epoch": 3010, "lr": 4.2040773057366864e-08}
{"train_loss": 0.01755567081272602, "global_step": 267913, "epoch": 3010, "lr": 4.201700775241513e-08}
{"train_loss": 0.03731856122612953, "global_step": 267914, "epoch": 3010, "lr": 4.199324916370206e-08}
{"train_loss": 0.04436242952942848, "global_step": 267915, "epoch": 3010, "lr": 4.196949729123878e-08}
{"train_loss": 0.03013019636273384, "global_step": 267916, "epoch": 3010, "lr": 4.194575213501972e-08}
{"train_loss": 0.041839100420475006, "global_step": 267917, "epoch": 3010, "lr": 4.1922013695050445e-08}
{"train_loss": 0.0525796040892601, "global_step": 267918, "epoch": 3010, "lr": 4.189828197134204e-08}
{"train_loss": 0.04498198255896568, "global_step": 267919, "epoch": 3010, "lr": 4.187455696388898e-08}
{"train_loss": 0.01256322767585516, "global_step": 267920, "epoch": 3010, "lr": 4.185083867269679e-08}
{"train_loss": 0.017738506197929382, "global_step": 267921, "epoch": 3010, "lr": 4.182712709776548e-08}
{"train_loss": 0.05542968586087227, "global_step": 267922, "epoch": 3010, "lr": 4.180342223910616e-08}
{"train_loss": 0.02916828915476799, "global_step": 267923, "epoch": 3010, "lr": 4.177972409671882e-08}
{"train_loss": 0.05675187334418297, "global_step": 267924, "epoch": 3010, "lr": 4.1756032670597914e-08}
{"train_loss": 0.018698858097195625, "global_step": 267925, "epoch": 3010, "lr": 4.1732347960760085e-08}
{"train_loss": 0.03677172586321831, "global_step": 267926, "epoch": 3010, "lr": 4.17086699671998e-08}
{"train_loss": 0.03771697357296944, "global_step": 267927, "epoch": 3010, "lr": 4.1684998689922595e-08}
{"train_loss": 0.008114788681268692, "global_step": 267928, "epoch": 3010, "lr": 4.1661334128934024e-08}
{"train_loss": 0.022042538970708847, "global_step": 267929, "epoch": 3010, "lr": 4.1637676284234096e-08}
{"train_loss": 0.01947074383497238, "global_step": 267930, "epoch": 3010, "lr": 4.161402515582835e-08}
{"train_loss": 0.032582174986600876, "global_step": 267931, "epoch": 3010, "lr": 4.15903807437168e-08}
{"train_loss": 0.04728284105658531, "global_step": 267932, "epoch": 3010, "lr": 4.156674304790498e-08}
{"train_loss": 0.04642379283905029, "global_step": 267933, "epoch": 3010, "lr": 4.1543112068398446e-08}
{"train_loss": 0.0613381527364254, "global_step": 267934, "epoch": 3010, "lr": 4.151948780519721e-08}
{"train_loss": 0.019486000761389732, "global_step": 267935, "epoch": 3010, "lr": 4.149587025830126e-08}
{"train_loss": 0.04514233395457268, "global_step": 267936, "epoch": 3010, "lr": 4.147225942771615e-08}
{"train_loss": 0.017980169504880905, "global_step": 267937, "epoch": 3010, "lr": 4.144865531345299e-08}
{"train_loss": 0.06117163971066475, "global_step": 267938, "epoch": 3010, "lr": 4.142505791550066e-08}
{"train_loss": 0.06645762175321579, "global_step": 267939, "epoch": 3010, "lr": 4.1401467233875837e-08}
{"train_loss": 0.023441143333911896, "global_step": 267940, "epoch": 3010, "lr": 4.1377883268572946e-08}
{"train_loss": 0.03241140767931938, "global_step": 267941, "epoch": 3010, "lr": 4.1354306019597556e-08}
{"train_loss": 0.044793255627155304, "global_step": 267942, "epoch": 3010, "lr": 4.133073548694966e-08}
{"train_loss": 0.040888652205467224, "global_step": 267943, "epoch": 3010, "lr": 4.130717167064035e-08}
{"train_loss": 0.014376034960150719, "global_step": 267944, "epoch": 3010, "lr": 4.128361457066965e-08}
{"train_loss": 0.04491223022341728, "global_step": 267945, "epoch": 3010, "lr": 4.126006418703754e-08}
{"train_loss": 0.053740642964839935, "global_step": 267946, "epoch": 3010, "lr": 4.123652051974958e-08}
{"train_loss": 0.022195443511009216, "global_step": 267947, "epoch": 3010, "lr": 4.121298356880576e-08}
{"train_loss": 0.07096236944198608, "global_step": 267948, "epoch": 3010, "lr": 4.11894533342172e-08}
{"train_loss": 0.06922180950641632, "global_step": 267949, "epoch": 3010, "lr": 4.116592981597833e-08}
{"train_loss": 0.031600210815668106, "global_step": 267950, "epoch": 3010, "lr": 4.1142413014094716e-08}
{"train_loss": 0.05437645688652992, "global_step": 267951, "epoch": 3010, "lr": 4.11189029285719e-08}
{"train_loss": 0.08495612442493439, "global_step": 267952, "epoch": 3010, "lr": 4.1095399559409883e-08}
{"train_loss": 0.04294953867793083, "global_step": 267953, "epoch": 3010, "lr": 4.107190290661978e-08}
{"train_loss": 0.014906994998455048, "global_step": 267954, "epoch": 3010, "lr": 4.104841297019046e-08}
{"train_loss": 0.015573814511299133, "global_step": 267955, "epoch": 3010, "lr": 4.102492975013861e-08}
{"train_loss": 0.043993037194013596, "global_step": 267956, "epoch": 3010, "lr": 4.100145324646421e-08}
{"train_loss": 0.034248195588588715, "global_step": 267957, "epoch": 3010, "lr": 4.097798345916171e-08}
{"train_loss": 0.04574243351817131, "global_step": 267958, "epoch": 3010, "lr": 4.0954520388242214e-08}
{"train_loss": 0.04124445095658302, "global_step": 267959, "epoch": 3010, "lr": 4.0931064033711274e-08}
{"train_loss": 0.047621145844459534, "global_step": 267960, "epoch": 3010, "lr": 4.0907614395568894e-08}
{"train_loss": 0.07253574579954147, "global_step": 267961, "epoch": 3010, "lr": 4.088417147381507e-08}
{"train_loss": 0.031034914776682854, "global_step": 267962, "epoch": 3010, "lr": 4.086073526845535e-08}
{"train_loss": 0.018076732754707336, "global_step": 267963, "epoch": 3010, "lr": 4.0837305779489744e-08}
{"train_loss": 0.029565727338194847, "global_step": 267964, "epoch": 3010, "lr": 4.081388300692934e-08}
{"train_loss": 0.03766869753599167, "global_step": 267965, "epoch": 3010, "lr": 4.07904669507686e-08}
{"train_loss": 0.030620573088526726, "global_step": 267966, "epoch": 3010, "lr": 4.0767057611018624e-08}
{"train_loss": 0.04750290513038635, "global_step": 267967, "epoch": 3010, "lr": 4.07436549876794e-08}
{"train_loss": 0.07950693368911743, "global_step": 267968, "epoch": 3010, "lr": 4.072025908075094e-08}
{"train_loss": 0.06631261855363846, "global_step": 267969, "epoch": 3010, "lr": 4.06968698902388e-08}
{"train_loss": 0.029090316966176033, "global_step": 267970, "epoch": 3010, "lr": 4.067348741614296e-08}
{"train_loss": 0.04900197684764862, "global_step": 267971, "epoch": 3010, "lr": 4.065011165847454e-08}
{"train_loss": 0.049931131303310394, "global_step": 267972, "epoch": 3010, "lr": 4.0626742617227986e-08}
{"train_loss": 0.10716678202152252, "global_step": 267973, "epoch": 3010, "lr": 4.060338029241439e-08}
{"train_loss": 0.03614353761076927, "global_step": 267974, "epoch": 3010, "lr": 4.058002468402822e-08}
{"train_loss": 0.04155031591653824, "global_step": 267975, "epoch": 3010, "lr": 4.0556675792080556e-08}
{"train_loss": 0.05786163732409477, "global_step": 267976, "epoch": 3010, "lr": 4.053333361657141e-08}
{"train_loss": 0.04656751826405525, "global_step": 267977, "epoch": 3010, "lr": 4.050999815750078e-08}
{"train_loss": 0.040045881036962015, "global_step": 267978, "epoch": 3010, "lr": 4.0486669414874224e-08, "val_loss": 9.647109985351562, "train_action_mse_error": 20.05146598815918}
{"train_loss": 0.06462549418210983, "global_step": 267979, "epoch": 3011, "lr": 4.046334738869728e-08}
{"train_loss": 0.02402576617896557, "global_step": 267980, "epoch": 3011, "lr": 4.0440032078975506e-08}
{"train_loss": 0.03579612821340561, "global_step": 267981, "epoch": 3011, "lr": 4.0416723485703354e-08}
{"train_loss": 0.03599292039871216, "global_step": 267982, "epoch": 3011, "lr": 4.0393421608886373e-08}
{"train_loss": 0.10237850993871689, "global_step": 267983, "epoch": 3011, "lr": 4.037012644853011e-08}
{"train_loss": 0.05281820893287659, "global_step": 267984, "epoch": 3011, "lr": 4.034683800464012e-08}
{"train_loss": 0.04279269278049469, "global_step": 267985, "epoch": 3011, "lr": 4.032355627721085e-08}
{"train_loss": 0.06623540818691254, "global_step": 267986, "epoch": 3011, "lr": 4.0300281266258954e-08}
{"train_loss": 0.048457201570272446, "global_step": 267987, "epoch": 3011, "lr": 4.027701297177333e-08}
{"train_loss": 0.07089436799287796, "global_step": 267988, "epoch": 3011, "lr": 4.0253751393770636e-08}
{"train_loss": 0.03205429017543793, "global_step": 267989, "epoch": 3011, "lr": 4.023049653223976e-08}
{"train_loss": 0.05391183868050575, "global_step": 267990, "epoch": 3011, "lr": 4.020724838719181e-08}
{"train_loss": 0.023591935634613037, "global_step": 267991, "epoch": 3011, "lr": 4.018400695863234e-08}
{"train_loss": 0.024309629574418068, "global_step": 267992, "epoch": 3011, "lr": 4.016077224656134e-08}
{"train_loss": 0.008054525591433048, "global_step": 267993, "epoch": 3011, "lr": 4.013754425097882e-08}
{"train_loss": 0.07761402428150177, "global_step": 267994, "epoch": 3011, "lr": 4.011432297189033e-08}
{"train_loss": 0.08831299096345901, "global_step": 267995, "epoch": 3011, "lr": 4.0091108409301414e-08}
{"train_loss": 0.04175584763288498, "global_step": 267996, "epoch": 3011, "lr": 4.006790056321763e-08}
{"train_loss": 0.02156789042055607, "global_step": 267997, "epoch": 3011, "lr": 4.004469943363343e-08}
{"train_loss": 0.04766011983156204, "global_step": 267998, "epoch": 3011, "lr": 4.002150502055435e-08}
{"train_loss": 0.043643444776535034, "global_step": 267999, "epoch": 3011, "lr": 3.999831732398596e-08}
{"train_loss": 0.04865189641714096, "global_step": 268000, "epoch": 3011, "lr": 3.9975136343933793e-08}
{"train_loss": 0.03439955785870552, "global_step": 268001, "epoch": 3011, "lr": 3.9951962080397866e-08}
{"train_loss": 0.05472680926322937, "global_step": 268002, "epoch": 3011, "lr": 3.992879453337817e-08}
{"train_loss": 0.04860924184322357, "global_step": 268003, "epoch": 3011, "lr": 3.9905633702885806e-08}
{"train_loss": 0.019823474809527397, "global_step": 268004, "epoch": 3011, "lr": 3.9882479588915224e-08}
{"train_loss": 0.03204663470387459, "global_step": 268005, "epoch": 3011, "lr": 3.985933219147753e-08}
{"train_loss": 0.06095340475440025, "global_step": 268006, "epoch": 3011, "lr": 3.9836191510567166e-08}
{"train_loss": 0.03052077442407608, "global_step": 268007, "epoch": 3011, "lr": 3.981305754619524e-08}
{"train_loss": 0.03342324122786522, "global_step": 268008, "epoch": 3011, "lr": 3.9789930298356204e-08}
{"train_loss": 0.02081623487174511, "global_step": 268009, "epoch": 3011, "lr": 3.97668097670667e-08}
{"train_loss": 0.0358850359916687, "global_step": 268010, "epoch": 3011, "lr": 3.974369595231564e-08}
{"train_loss": 0.023633213713765144, "global_step": 268011, "epoch": 3011, "lr": 3.972058885411411e-08}
{"train_loss": 0.08080445975065231, "global_step": 268012, "epoch": 3011, "lr": 3.969748847246213e-08}
{"train_loss": 0.047181278467178345, "global_step": 268013, "epoch": 3011, "lr": 3.967439480736523e-08}
{"train_loss": 0.09425973892211914, "global_step": 268014, "epoch": 3011, "lr": 3.9651307858823425e-08}
{"train_loss": 0.043243296444416046, "global_step": 268015, "epoch": 3011, "lr": 3.9628227626842265e-08}
{"train_loss": 0.06597276031970978, "global_step": 268016, "epoch": 3011, "lr": 3.960515411142729e-08}
{"train_loss": 0.06987939774990082, "global_step": 268017, "epoch": 3011, "lr": 3.958208731257851e-08}
{"train_loss": 0.06779927760362625, "global_step": 268018, "epoch": 3011, "lr": 3.955902723029592e-08}
{"train_loss": 0.035301871597766876, "global_step": 268019, "epoch": 3011, "lr": 3.953597386458508e-08}
{"train_loss": 0.06590571254491806, "global_step": 268020, "epoch": 3011, "lr": 3.951292721545152e-08}
{"train_loss": 0.057041700929403305, "global_step": 268021, "epoch": 3011, "lr": 3.9489887282895265e-08}
{"train_loss": 0.038492072373628616, "global_step": 268022, "epoch": 3011, "lr": 3.9466854066921855e-08}
{"train_loss": 0.07593966275453568, "global_step": 268023, "epoch": 3011, "lr": 3.9443827567531285e-08}
{"train_loss": 0.011949689127504826, "global_step": 268024, "epoch": 3011, "lr": 3.942080778472912e-08}
{"train_loss": 0.1079004779458046, "global_step": 268025, "epoch": 3011, "lr": 3.9397794718515345e-08}
{"train_loss": 0.06804206967353821, "global_step": 268026, "epoch": 3011, "lr": 3.937478836890107e-08}
{"train_loss": 0.004661478102207184, "global_step": 268027, "epoch": 3011, "lr": 3.9351788735880744e-08}
{"train_loss": 0.06635736674070358, "global_step": 268028, "epoch": 3011, "lr": 3.932879581946547e-08}
{"train_loss": 0.06487490236759186, "global_step": 268029, "epoch": 3011, "lr": 3.9305809619644144e-08}
{"train_loss": 0.04566333442926407, "global_step": 268030, "epoch": 3011, "lr": 3.928283013643896e-08}
{"train_loss": 0.021732354536652565, "global_step": 268031, "epoch": 3011, "lr": 3.925985736983884e-08}
{"train_loss": 0.047265876084566116, "global_step": 268032, "epoch": 3011, "lr": 3.923689131984931e-08}
{"train_loss": 0.09078842401504517, "global_step": 268033, "epoch": 3011, "lr": 3.9213931986475936e-08}
{"train_loss": 0.07080314308404922, "global_step": 268034, "epoch": 3011, "lr": 3.919097936972427e-08}
{"train_loss": 0.036587439477443695, "global_step": 268035, "epoch": 3011, "lr": 3.91680334695943e-08}
{"train_loss": 0.057940516620874405, "global_step": 268036, "epoch": 3011, "lr": 3.9145094286086034e-08}
{"train_loss": 0.05617523565888405, "global_step": 268037, "epoch": 3011, "lr": 3.912216181921058e-08}
{"train_loss": 0.045759283006191254, "global_step": 268038, "epoch": 3011, "lr": 3.909923606896237e-08}
{"train_loss": 0.05347302556037903, "global_step": 268039, "epoch": 3011, "lr": 3.9076317035352525e-08}
{"train_loss": 0.022481214255094528, "global_step": 268040, "epoch": 3011, "lr": 3.905340471837549e-08}
{"train_loss": 0.06967654079198837, "global_step": 268041, "epoch": 3011, "lr": 3.9030499118036804e-08}
{"train_loss": 0.04574977234005928, "global_step": 268042, "epoch": 3011, "lr": 3.900760023434757e-08}
{"train_loss": 0.025474529713392258, "global_step": 268043, "epoch": 3011, "lr": 3.8984708067302254e-08}
{"train_loss": 0.024000128731131554, "global_step": 268044, "epoch": 3011, "lr": 3.8961822616906395e-08}
{"train_loss": 0.05570670962333679, "global_step": 268045, "epoch": 3011, "lr": 3.8938943883159995e-08}
{"train_loss": 0.03871481493115425, "global_step": 268046, "epoch": 3011, "lr": 3.891607186607416e-08}
{"train_loss": 0.06629384309053421, "global_step": 268047, "epoch": 3011, "lr": 3.8893206565643324e-08}
{"train_loss": 0.039510007947683334, "global_step": 268048, "epoch": 3011, "lr": 3.8870347981878605e-08}
{"train_loss": 0.02255776710808277, "global_step": 268049, "epoch": 3011, "lr": 3.884749611477445e-08}
{"train_loss": 0.05076667293906212, "global_step": 268050, "epoch": 3011, "lr": 3.882465096434196e-08}
{"train_loss": 0.04728031903505325, "global_step": 268051, "epoch": 3011, "lr": 3.8801812530581126e-08}
{"train_loss": 0.09665825963020325, "global_step": 268052, "epoch": 3011, "lr": 3.877898081349196e-08}
{"train_loss": 0.02651878073811531, "global_step": 268053, "epoch": 3011, "lr": 3.875615581308001e-08}
{"train_loss": 0.04896803945302963, "global_step": 268054, "epoch": 3011, "lr": 3.8733337529350823e-08}
{"train_loss": 0.04572644829750061, "global_step": 268055, "epoch": 3011, "lr": 3.871052596229885e-08}
{"train_loss": 0.06511875987052917, "global_step": 268056, "epoch": 3011, "lr": 3.868772111194075e-08}
{"train_loss": 0.062250543385744095, "global_step": 268057, "epoch": 3011, "lr": 3.866492297826541e-08}
{"train_loss": 0.027118882164359093, "global_step": 268058, "epoch": 3011, "lr": 3.864213156128949e-08}
{"train_loss": 0.09431066364049911, "global_step": 268059, "epoch": 3011, "lr": 3.8619346861001885e-08}
{"train_loss": 0.03510715812444687, "global_step": 268060, "epoch": 3011, "lr": 3.859656887741925e-08}
{"train_loss": 0.019105643033981323, "global_step": 268061, "epoch": 3011, "lr": 3.857379761053603e-08}
{"train_loss": 0.06796329468488693, "global_step": 268062, "epoch": 3011, "lr": 3.855103306035779e-08}
{"train_loss": 0.07408727705478668, "global_step": 268063, "epoch": 3011, "lr": 3.8528275226884516e-08}
{"train_loss": 0.0679180920124054, "global_step": 268064, "epoch": 3011, "lr": 3.8505524110127314e-08}
{"train_loss": 0.013968445360660553, "global_step": 268065, "epoch": 3011, "lr": 3.848277971008063e-08}
{"train_loss": 0.045519229024648666, "global_step": 268066, "epoch": 3011, "lr": 3.8460042026750024e-08}
{"train_loss": 0.049698856041839955, "global_step": 268067, "epoch": 3011, "lr": 3.843731106014103e-08, "val_loss": 9.59588623046875}
{"train_loss": 0.031034434214234352, "global_step": 268068, "epoch": 3012, "lr": 3.841458681025922e-08}
{"train_loss": 0.00769337173551321, "global_step": 268069, "epoch": 3012, "lr": 3.839186927709903e-08}
{"train_loss": 0.04276824742555618, "global_step": 268070, "epoch": 3012, "lr": 3.8369158460671576e-08}
{"train_loss": 0.04747087508440018, "global_step": 268071, "epoch": 3012, "lr": 3.834645436097129e-08}
{"train_loss": 0.06116524711251259, "global_step": 268072, "epoch": 3012, "lr": 3.8323756978009276e-08}
{"train_loss": 0.05896684527397156, "global_step": 268073, "epoch": 3012, "lr": 3.8301066311785537e-08}
{"train_loss": 0.04593811556696892, "global_step": 268074, "epoch": 3012, "lr": 3.8278382362305634e-08}
{"train_loss": 0.036395277827978134, "global_step": 268075, "epoch": 3012, "lr": 3.8255705129564e-08}
{"train_loss": 0.027690142393112183, "global_step": 268076, "epoch": 3012, "lr": 3.82330346135773e-08}
{"train_loss": 0.02058037742972374, "global_step": 268077, "epoch": 3012, "lr": 3.8210370814334426e-08}
{"train_loss": 0.027277546003460884, "global_step": 268078, "epoch": 3012, "lr": 3.818771373185204e-08}
{"train_loss": 0.06310579925775528, "global_step": 268079, "epoch": 3012, "lr": 3.816506336611903e-08}
{"train_loss": 0.047275006771087646, "global_step": 268080, "epoch": 3012, "lr": 3.814241971715205e-08}
{"train_loss": 0.05365406721830368, "global_step": 268081, "epoch": 3012, "lr": 3.811978278494555e-08}
{"train_loss": 0.02344011515378952, "global_step": 268082, "epoch": 3012, "lr": 3.809715256950508e-08}
{"train_loss": 0.019686978310346603, "global_step": 268083, "epoch": 3012, "lr": 3.807452907083064e-08}
{"train_loss": 0.04002103582024574, "global_step": 268084, "epoch": 3012, "lr": 3.805191228893334e-08}
{"train_loss": 0.04398650676012039, "global_step": 268085, "epoch": 3012, "lr": 3.802930222380763e-08}
{"train_loss": 0.010340851731598377, "global_step": 268086, "epoch": 3012, "lr": 3.800669887545904e-08}
{"train_loss": 0.006259516812860966, "global_step": 268087, "epoch": 3012, "lr": 3.798410224389315e-08}
{"train_loss": 0.038338348269462585, "global_step": 268088, "epoch": 3012, "lr": 3.7961512329109936e-08}
{"train_loss": 0.048097386956214905, "global_step": 268089, "epoch": 3012, "lr": 3.793892913111496e-08}
{"train_loss": 0.035181619226932526, "global_step": 268090, "epoch": 3012, "lr": 3.7916352649913776e-08}
{"train_loss": 0.06181630864739418, "global_step": 268091, "epoch": 3012, "lr": 3.789378288550083e-08}
{"train_loss": 0.03198791295289993, "global_step": 268092, "epoch": 3012, "lr": 3.787121983788722e-08}
{"train_loss": 0.07901498675346375, "global_step": 268093, "epoch": 3012, "lr": 3.784866350706739e-08}
{"train_loss": 0.04469429701566696, "global_step": 268094, "epoch": 3012, "lr": 3.7826113893058015e-08}
{"train_loss": 0.004839234985411167, "global_step": 268095, "epoch": 3012, "lr": 3.780357099584797e-08}
{"train_loss": 0.04344742000102997, "global_step": 268096, "epoch": 3012, "lr": 3.778103481544837e-08}
{"train_loss": 0.08110275864601135, "global_step": 268097, "epoch": 3012, "lr": 3.7758505351859205e-08}
{"train_loss": 0.055426426231861115, "global_step": 268098, "epoch": 3012, "lr": 3.773598260508604e-08}
{"train_loss": 0.016713015735149384, "global_step": 268099, "epoch": 3012, "lr": 3.7713466575128864e-08}
{"train_loss": 0.05798780545592308, "global_step": 268100, "epoch": 3012, "lr": 3.769095726199323e-08}
{"train_loss": 0.044666968286037445, "global_step": 268101, "epoch": 3012, "lr": 3.766845466567914e-08}
{"train_loss": 0.02273380011320114, "global_step": 268102, "epoch": 3012, "lr": 3.764595878619215e-08}
{"train_loss": 0.04812399297952652, "global_step": 268103, "epoch": 3012, "lr": 3.76234696235378e-08}
{"train_loss": 0.05842133238911629, "global_step": 268104, "epoch": 3012, "lr": 3.760098717771054e-08}
{"train_loss": 0.022332660853862762, "global_step": 268105, "epoch": 3012, "lr": 3.757851144872149e-08}
{"train_loss": 0.0726165845990181, "global_step": 268106, "epoch": 3012, "lr": 3.7556042436570624e-08}
{"train_loss": 0.03406553342938423, "global_step": 268107, "epoch": 3012, "lr": 3.753358014126351e-08}
{"train_loss": 0.011321756057441235, "global_step": 268108, "epoch": 3012, "lr": 3.751112456280015e-08}
{"train_loss": 0.06503389030694962, "global_step": 268109, "epoch": 3012, "lr": 3.748867570118608e-08}
{"train_loss": 0.055305179208517075, "global_step": 268110, "epoch": 3012, "lr": 3.7466233556421314e-08}
{"train_loss": 0.08327752351760864, "global_step": 268111, "epoch": 3012, "lr": 3.74437981285114e-08}
{"train_loss": 0.056367289274930954, "global_step": 268112, "epoch": 3012, "lr": 3.742136941745633e-08}
{"train_loss": 0.05420270934700966, "global_step": 268113, "epoch": 3012, "lr": 3.7398947423261665e-08}
{"train_loss": 0.028496529906988144, "global_step": 268114, "epoch": 3012, "lr": 3.73765321459274e-08}
{"train_loss": 0.03625594452023506, "global_step": 268115, "epoch": 3012, "lr": 3.735412358545909e-08}
{"train_loss": 0.030738748610019684, "global_step": 268116, "epoch": 3012, "lr": 3.733172174186228e-08}
{"train_loss": 0.04628417268395424, "global_step": 268117, "epoch": 3012, "lr": 3.730932661513698e-08}
{"train_loss": 0.056340817362070084, "global_step": 268118, "epoch": 3012, "lr": 3.7286938205283175e-08}
{"train_loss": 0.04285489767789841, "global_step": 268119, "epoch": 3012, "lr": 3.726455651231198e-08}
{"train_loss": 0.033359166234731674, "global_step": 268120, "epoch": 3012, "lr": 3.724218153621783e-08}
{"train_loss": 0.044841449707746506, "global_step": 268121, "epoch": 3012, "lr": 3.721981327701185e-08}
{"train_loss": 0.03894795849919319, "global_step": 268122, "epoch": 3012, "lr": 3.719745173468847e-08}
{"train_loss": 0.03988635167479515, "global_step": 268123, "epoch": 3012, "lr": 3.71750969092588e-08}
{"train_loss": 0.05245804414153099, "global_step": 268124, "epoch": 3012, "lr": 3.7152748800717286e-08}
{"train_loss": 0.02128831297159195, "global_step": 268125, "epoch": 3012, "lr": 3.713040740907503e-08}
{"train_loss": 0.0417850986123085, "global_step": 268126, "epoch": 3012, "lr": 3.7108072734326485e-08}
{"train_loss": 0.024972321465611458, "global_step": 268127, "epoch": 3012, "lr": 3.7085744776488294e-08}
{"train_loss": 0.07060980796813965, "global_step": 268128, "epoch": 3012, "lr": 3.706342353554937e-08}
{"train_loss": 0.07255984842777252, "global_step": 268129, "epoch": 3012, "lr": 3.704110901151525e-08}
{"train_loss": 0.0479709655046463, "global_step": 268130, "epoch": 3012, "lr": 3.7018801204397046e-08}
{"train_loss": 0.045013755559921265, "global_step": 268131, "epoch": 3012, "lr": 3.6996500114194755e-08}
{"train_loss": 0.0072385454550385475, "global_step": 268132, "epoch": 3012, "lr": 3.697420574090282e-08}
{"train_loss": 0.05389697849750519, "global_step": 268133, "epoch": 3012, "lr": 3.6951918084537904e-08}
{"train_loss": 0.03947457671165466, "global_step": 268134, "epoch": 3012, "lr": 3.69296371450889e-08}
{"train_loss": 0.049469638615846634, "global_step": 268135, "epoch": 3012, "lr": 3.690736292257246e-08}
{"train_loss": 0.020536893978714943, "global_step": 268136, "epoch": 3012, "lr": 3.688509541698304e-08}
{"train_loss": 0.06579649448394775, "global_step": 268137, "epoch": 3012, "lr": 3.6862834628320633e-08}
{"train_loss": 0.0529291070997715, "global_step": 268138, "epoch": 3012, "lr": 3.684058055659634e-08}
{"train_loss": 0.04607386142015457, "global_step": 268139, "epoch": 3012, "lr": 3.681833320181016e-08}
{"train_loss": 0.021164407953619957, "global_step": 268140, "epoch": 3012, "lr": 3.6796092563962104e-08}
{"train_loss": 0.03898189216852188, "global_step": 268141, "epoch": 3012, "lr": 3.677385864306326e-08}
{"train_loss": 0.05306791514158249, "global_step": 268142, "epoch": 3012, "lr": 3.675163143910809e-08}
{"train_loss": 0.08827152103185654, "global_step": 268143, "epoch": 3012, "lr": 3.672941095210214e-08}
{"train_loss": 0.06753946095705032, "global_step": 268144, "epoch": 3012, "lr": 3.670719718205096e-08}
{"train_loss": 0.031190287321805954, "global_step": 268145, "epoch": 3012, "lr": 3.668499012895454e-08}
{"train_loss": 0.046670034527778625, "global_step": 268146, "epoch": 3012, "lr": 3.6662789792812904e-08}
{"train_loss": 0.054958321154117584, "global_step": 268147, "epoch": 3012, "lr": 3.664059617363713e-08}
{"train_loss": 0.0358780175447464, "global_step": 268148, "epoch": 3012, "lr": 3.6618409271427233e-08}
{"train_loss": 0.020643245428800583, "global_step": 268149, "epoch": 3012, "lr": 3.659622908618321e-08}
{"train_loss": 0.12239514291286469, "global_step": 268150, "epoch": 3012, "lr": 3.657405561791061e-08}
{"train_loss": 0.09376517683267593, "global_step": 268151, "epoch": 3012, "lr": 3.6551888866609427e-08}
{"train_loss": 0.07753627747297287, "global_step": 268152, "epoch": 3012, "lr": 3.6529728832285226e-08}
{"train_loss": 0.017849383875727654, "global_step": 268153, "epoch": 3012, "lr": 3.650757551494355e-08}
{"train_loss": 0.061585281044244766, "global_step": 268154, "epoch": 3012, "lr": 3.64854289145844e-08}
{"train_loss": 0.06342989951372147, "global_step": 268155, "epoch": 3012, "lr": 3.646328903120777e-08}
{"train_loss": 0.044572485610842705, "global_step": 268156, "epoch": 3012, "lr": 3.644115586481922e-08, "val_loss": 9.6355619430542}
{"train_loss": 0.03596947714686394, "global_step": 268157, "epoch": 3013, "lr": 3.64190294154243e-08}
{"train_loss": 0.06499403715133667, "global_step": 268158, "epoch": 3013, "lr": 3.639690968302301e-08}
{"train_loss": 0.014104939065873623, "global_step": 268159, "epoch": 3013, "lr": 3.6374796667620895e-08}
{"train_loss": 0.052025482058525085, "global_step": 268160, "epoch": 3013, "lr": 3.635269036921796e-08}
{"train_loss": 0.04499700665473938, "global_step": 268161, "epoch": 3013, "lr": 3.633059078781975e-08}
{"train_loss": 0.04319414123892784, "global_step": 268162, "epoch": 3013, "lr": 3.630849792342628e-08}
{"train_loss": 0.05687137693166733, "global_step": 268163, "epoch": 3013, "lr": 3.628641177604308e-08}
{"train_loss": 0.04210975766181946, "global_step": 268164, "epoch": 3013, "lr": 3.626433234567017e-08}
{"train_loss": 0.009197914972901344, "global_step": 268165, "epoch": 3013, "lr": 3.6242259632313094e-08}
{"train_loss": 0.04065362364053726, "global_step": 268166, "epoch": 3013, "lr": 3.62201936359774e-08}
{"train_loss": 0.020362243056297302, "global_step": 268167, "epoch": 3013, "lr": 3.619813435665753e-08}
{"train_loss": 0.03530525788664818, "global_step": 268168, "epoch": 3013, "lr": 3.617608179437015e-08}
{"train_loss": 0.033773668110370636, "global_step": 268169, "epoch": 3013, "lr": 3.615403594910416e-08}
{"train_loss": 0.04515934735536575, "global_step": 268170, "epoch": 3013, "lr": 3.6131996820865096e-08}
{"train_loss": 0.022000601515173912, "global_step": 268171, "epoch": 3013, "lr": 3.6109964409664076e-08}
{"train_loss": 0.012254558503627777, "global_step": 268172, "epoch": 3013, "lr": 3.608793871550109e-08}
{"train_loss": 0.06902986019849777, "global_step": 268173, "epoch": 3013, "lr": 3.606591973837059e-08}
{"train_loss": 0.05046197399497032, "global_step": 268174, "epoch": 3013, "lr": 3.604390747828923e-08}
{"train_loss": 0.04385673627257347, "global_step": 268175, "epoch": 3013, "lr": 3.6021901935245905e-08}
{"train_loss": 0.06351371854543686, "global_step": 268176, "epoch": 3013, "lr": 3.5999903109257275e-08}
{"train_loss": 0.056110017001628876, "global_step": 268177, "epoch": 3013, "lr": 3.5977911000312225e-08}
{"train_loss": 0.019909659400582314, "global_step": 268178, "epoch": 3013, "lr": 3.595592560842742e-08}
{"train_loss": 0.0313519649207592, "global_step": 268179, "epoch": 3013, "lr": 3.593394693359175e-08}
{"train_loss": 0.07597649842500687, "global_step": 268180, "epoch": 3013, "lr": 3.591197497582188e-08}
{"train_loss": 0.04811488091945648, "global_step": 268181, "epoch": 3013, "lr": 3.589000973511225e-08}
{"train_loss": 0.04672127962112427, "global_step": 268182, "epoch": 3013, "lr": 3.5868051211473965e-08}
{"train_loss": 0.027623886242508888, "global_step": 268183, "epoch": 3013, "lr": 3.5846099404895914e-08}
{"train_loss": 0.07323987036943436, "global_step": 268184, "epoch": 3013, "lr": 3.582415431539476e-08}
{"train_loss": 0.07166355103254318, "global_step": 268185, "epoch": 3013, "lr": 3.580221594296496e-08}
{"train_loss": 0.01972501166164875, "global_step": 268186, "epoch": 3013, "lr": 3.578028428761759e-08}
{"train_loss": 0.026866786181926727, "global_step": 268187, "epoch": 3013, "lr": 3.5758359349347126e-08}
{"train_loss": 0.05734725669026375, "global_step": 268188, "epoch": 3013, "lr": 3.5736441128159104e-08}
{"train_loss": 0.07919437438249588, "global_step": 268189, "epoch": 3013, "lr": 3.571452962405908e-08}
{"train_loss": 0.052560608834028244, "global_step": 268190, "epoch": 3013, "lr": 3.569262483704705e-08}
{"train_loss": 0.03511928394436836, "global_step": 268191, "epoch": 3013, "lr": 3.567072676712302e-08}
{"train_loss": 0.04019290953874588, "global_step": 268192, "epoch": 3013, "lr": 3.564883541429809e-08}
{"train_loss": 0.027804654091596603, "global_step": 268193, "epoch": 3013, "lr": 3.562695077857225e-08}
{"train_loss": 0.05293959379196167, "global_step": 268194, "epoch": 3013, "lr": 3.560507285995107e-08}
{"train_loss": 0.03550991415977478, "global_step": 268195, "epoch": 3013, "lr": 3.558320165842899e-08}
{"train_loss": 0.06645122915506363, "global_step": 268196, "epoch": 3013, "lr": 3.556133717401156e-08}
{"train_loss": 0.02730718068778515, "global_step": 268197, "epoch": 3013, "lr": 3.553947940670987e-08}
{"train_loss": 0.03713993728160858, "global_step": 268198, "epoch": 3013, "lr": 3.551762835651285e-08}
{"train_loss": 0.061583083122968674, "global_step": 268199, "epoch": 3013, "lr": 3.549578402343712e-08}
{"train_loss": 0.051429569721221924, "global_step": 268200, "epoch": 3013, "lr": 3.5473946407482696e-08}
{"train_loss": 0.04738261550664902, "global_step": 268201, "epoch": 3013, "lr": 3.5452115508644025e-08}
{"train_loss": 0.04984639212489128, "global_step": 268202, "epoch": 3013, "lr": 3.543029132693221e-08}
{"train_loss": 0.10101095587015152, "global_step": 268203, "epoch": 3013, "lr": 3.540847386234725e-08}
{"train_loss": 0.036612335592508316, "global_step": 268204, "epoch": 3013, "lr": 3.538666311489469e-08}
{"train_loss": 0.018762944266200066, "global_step": 268205, "epoch": 3013, "lr": 3.536485908456899e-08}
{"train_loss": 0.021321892738342285, "global_step": 268206, "epoch": 3013, "lr": 3.53430617713868e-08}
{"train_loss": 0.036685511469841, "global_step": 268207, "epoch": 3013, "lr": 3.532127117534256e-08}
{"train_loss": 0.05188077688217163, "global_step": 268208, "epoch": 3013, "lr": 3.529948729643628e-08}
{"train_loss": 0.04134165495634079, "global_step": 268209, "epoch": 3013, "lr": 3.5277710134679066e-08}
{"train_loss": 0.045127466320991516, "global_step": 268210, "epoch": 3013, "lr": 3.5255939690070905e-08}
{"train_loss": 0.07438594847917557, "global_step": 268211, "epoch": 3013, "lr": 3.5234175962611806e-08}
{"train_loss": 0.06194411590695381, "global_step": 268212, "epoch": 3013, "lr": 3.5212418952301764e-08}
{"train_loss": 0.05661996081471443, "global_step": 268213, "epoch": 3013, "lr": 3.519066865915188e-08}
{"train_loss": 0.051766108721494675, "global_step": 268214, "epoch": 3013, "lr": 3.5168925083162166e-08}
{"train_loss": 0.06554433703422546, "global_step": 268215, "epoch": 3013, "lr": 3.5147188224338156e-08}
{"train_loss": 0.041519422084093094, "global_step": 268216, "epoch": 3013, "lr": 3.512545808267431e-08}
{"train_loss": 0.061346642673015594, "global_step": 268217, "epoch": 3013, "lr": 3.510373465818173e-08}
{"train_loss": 0.025459887459874153, "global_step": 268218, "epoch": 3013, "lr": 3.508201795086041e-08}
{"train_loss": 0.027577361091971397, "global_step": 268219, "epoch": 3013, "lr": 3.506030796071036e-08}
{"train_loss": 0.06956573575735092, "global_step": 268220, "epoch": 3013, "lr": 3.503860468774267e-08}
{"train_loss": 0.03396026790142059, "global_step": 268221, "epoch": 3013, "lr": 3.50169081319518e-08}
{"train_loss": 0.046388447284698486, "global_step": 268222, "epoch": 3013, "lr": 3.499521829334329e-08}
{"train_loss": 0.04310345649719238, "global_step": 268223, "epoch": 3013, "lr": 3.49735351719227e-08}
{"train_loss": 0.036247819662094116, "global_step": 268224, "epoch": 3013, "lr": 3.495185876769003e-08}
{"train_loss": 0.009613274596631527, "global_step": 268225, "epoch": 3013, "lr": 3.4930189080650824e-08}
{"train_loss": 0.04916702210903168, "global_step": 268226, "epoch": 3013, "lr": 3.4908526110805084e-08}
{"train_loss": 0.06741827726364136, "global_step": 268227, "epoch": 3013, "lr": 3.488686985815282e-08}
{"train_loss": 0.014091302640736103, "global_step": 268228, "epoch": 3013, "lr": 3.486522032270512e-08}
{"train_loss": 0.05922964587807655, "global_step": 268229, "epoch": 3013, "lr": 3.484357750446199e-08}
{"train_loss": 0.0469784140586853, "global_step": 268230, "epoch": 3013, "lr": 3.482194140342343e-08}
{"train_loss": 0.0409897118806839, "global_step": 268231, "epoch": 3013, "lr": 3.4800312019594996e-08}
{"train_loss": 0.022375954315066338, "global_step": 268232, "epoch": 3013, "lr": 3.477868935297668e-08}
{"train_loss": 0.04945656284689903, "global_step": 268233, "epoch": 3013, "lr": 3.4757073403574034e-08}
{"train_loss": 0.032689373940229416, "global_step": 268234, "epoch": 3013, "lr": 3.473546417138707e-08}
{"train_loss": 0.03600693494081497, "global_step": 268235, "epoch": 3013, "lr": 3.471386165642687e-08}
{"train_loss": 0.03769694268703461, "global_step": 268236, "epoch": 3013, "lr": 3.46922658586879e-08}
{"train_loss": 0.020871464163064957, "global_step": 268237, "epoch": 3013, "lr": 3.46706767781757e-08}
{"train_loss": 0.09087219834327698, "global_step": 268238, "epoch": 3013, "lr": 3.464909441489028e-08}
{"train_loss": 0.04699701443314552, "global_step": 268239, "epoch": 3013, "lr": 3.4627518768837184e-08}
{"train_loss": 0.03820274397730827, "global_step": 268240, "epoch": 3013, "lr": 3.460594984002197e-08}
{"train_loss": 0.06910338997840881, "global_step": 268241, "epoch": 3013, "lr": 3.458438762844463e-08}
{"train_loss": 0.020572859793901443, "global_step": 268242, "epoch": 3013, "lr": 3.456283213411071e-08}
{"train_loss": 0.04049922153353691, "global_step": 268243, "epoch": 3013, "lr": 3.454128335701468e-08}
{"train_loss": 0.030183840543031693, "global_step": 268244, "epoch": 3013, "lr": 3.451974129717317e-08}
{"train_loss": 0.04441810477776139, "global_step": 268245, "epoch": 3013, "lr": 3.4498205954575094e-08, "val_loss": 9.684999465942383}
{"train_loss": 0.04922203719615936, "global_step": 268246, "epoch": 3014, "lr": 3.447667732923155e-08}
{"train_loss": 0.0560130700469017, "global_step": 268247, "epoch": 3014, "lr": 3.4455155421142524e-08}
{"train_loss": 0.032807204872369766, "global_step": 268248, "epoch": 3014, "lr": 3.443364023031359e-08}
{"train_loss": 0.06009243428707123, "global_step": 268249, "epoch": 3014, "lr": 3.441213175675029e-08}
{"train_loss": 0.08024843037128448, "global_step": 268250, "epoch": 3014, "lr": 3.439063000044707e-08}
{"train_loss": 0.033608172088861465, "global_step": 268251, "epoch": 3014, "lr": 3.436913496140948e-08}
{"train_loss": 0.06063460558652878, "global_step": 268252, "epoch": 3014, "lr": 3.4347646639643075e-08}
{"train_loss": 0.016146602109074593, "global_step": 268253, "epoch": 3014, "lr": 3.432616503514785e-08}
{"train_loss": 0.017150094732642174, "global_step": 268254, "epoch": 3014, "lr": 3.4304690147934915e-08}
{"train_loss": 0.09215293079614639, "global_step": 268255, "epoch": 3014, "lr": 3.4283221977993165e-08}
{"train_loss": 0.02855107933282852, "global_step": 268256, "epoch": 3014, "lr": 3.4261760525339246e-08}
{"train_loss": 0.027217162773013115, "global_step": 268257, "epoch": 3014, "lr": 3.424030578996762e-08}
{"train_loss": 0.020961850881576538, "global_step": 268258, "epoch": 3014, "lr": 3.4218857771878276e-08}
{"train_loss": 0.041782818734645844, "global_step": 268259, "epoch": 3014, "lr": 3.4197416471087874e-08}
{"train_loss": 0.05798864737153053, "global_step": 268260, "epoch": 3014, "lr": 3.417598188758531e-08}
{"train_loss": 0.07431315630674362, "global_step": 268261, "epoch": 3014, "lr": 3.415455402138168e-08}
{"train_loss": 0.03145718201994896, "global_step": 268262, "epoch": 3014, "lr": 3.413313287247144e-08}
{"train_loss": 0.023276692256331444, "global_step": 268263, "epoch": 3014, "lr": 3.411171844086569e-08}
{"train_loss": 0.04844345524907112, "global_step": 268264, "epoch": 3014, "lr": 3.409031072656998e-08}
{"train_loss": 0.08409009873867035, "global_step": 268265, "epoch": 3014, "lr": 3.406890972957322e-08}
{"train_loss": 0.043759897351264954, "global_step": 268266, "epoch": 3014, "lr": 3.404751544989204e-08}
{"train_loss": 0.02360847406089306, "global_step": 268267, "epoch": 3014, "lr": 3.402612788752646e-08}
{"train_loss": 0.06836556643247604, "global_step": 268268, "epoch": 3014, "lr": 3.400474704247092e-08}
{"train_loss": 0.04130927100777626, "global_step": 268269, "epoch": 3014, "lr": 3.3983372914742074e-08}
{"train_loss": 0.049732256680727005, "global_step": 268270, "epoch": 3014, "lr": 3.396200550432882e-08}
{"train_loss": 0.020599838346242905, "global_step": 268271, "epoch": 3014, "lr": 3.394064481124226e-08}
{"train_loss": 0.06222324073314667, "global_step": 268272, "epoch": 3014, "lr": 3.3919290835482396e-08}
{"train_loss": 0.03582271561026573, "global_step": 268273, "epoch": 3014, "lr": 3.389794357705478e-08}
{"train_loss": 0.047155555337667465, "global_step": 268274, "epoch": 3014, "lr": 3.387660303595941e-08}
{"train_loss": 0.05758840590715408, "global_step": 268275, "epoch": 3014, "lr": 3.3855269212201835e-08}
{"train_loss": 0.0606311559677124, "global_step": 268276, "epoch": 3014, "lr": 3.383394210578206e-08}
{"train_loss": 0.03373132646083832, "global_step": 268277, "epoch": 3014, "lr": 3.381262171670563e-08}
{"train_loss": 0.036084432154893875, "global_step": 268278, "epoch": 3014, "lr": 3.3791308044972546e-08}
{"train_loss": 0.021895159035921097, "global_step": 268279, "epoch": 3014, "lr": 3.377000109058837e-08}
{"train_loss": 0.06200716644525528, "global_step": 268280, "epoch": 3014, "lr": 3.3748700853553086e-08}
{"train_loss": 0.02312408946454525, "global_step": 268281, "epoch": 3014, "lr": 3.372740733387225e-08}
{"train_loss": 0.029173292219638824, "global_step": 268282, "epoch": 3014, "lr": 3.370612053154587e-08}
{"train_loss": 0.033224474638700485, "global_step": 268283, "epoch": 3014, "lr": 3.368484044657949e-08}
{"train_loss": 0.04108424112200737, "global_step": 268284, "epoch": 3014, "lr": 3.366356707897866e-08}
{"train_loss": 0.03470901772379875, "global_step": 268285, "epoch": 3014, "lr": 3.364230042873784e-08}
{"train_loss": 0.001344730262644589, "global_step": 268286, "epoch": 3014, "lr": 3.3621040495868115e-08}
{"train_loss": 0.06833496689796448, "global_step": 268287, "epoch": 3014, "lr": 3.35997872803695e-08}
{"train_loss": 0.023327041417360306, "global_step": 268288, "epoch": 3014, "lr": 3.3578540782241986e-08}
{"train_loss": 0.042184554040431976, "global_step": 268289, "epoch": 3014, "lr": 3.355730100149112e-08}
{"train_loss": 0.043487321585416794, "global_step": 268290, "epoch": 3014, "lr": 3.3536067938122475e-08}
{"train_loss": 0.029400574043393135, "global_step": 268291, "epoch": 3014, "lr": 3.351484159213603e-08}
{"train_loss": 0.023347843438386917, "global_step": 268292, "epoch": 3014, "lr": 3.3493621963531785e-08}
{"train_loss": 0.07021091878414154, "global_step": 268293, "epoch": 3014, "lr": 3.34724090523153e-08}
{"train_loss": 0.09060785174369812, "global_step": 268294, "epoch": 3014, "lr": 3.345120285848658e-08}
{"train_loss": 0.023179857060313225, "global_step": 268295, "epoch": 3014, "lr": 3.3430003382056706e-08}
{"train_loss": 0.06098604202270508, "global_step": 268296, "epoch": 3014, "lr": 3.3408810623020146e-08}
{"train_loss": 0.025634661316871643, "global_step": 268297, "epoch": 3014, "lr": 3.3387624581387997e-08}
{"train_loss": 0.060153935104608536, "global_step": 268298, "epoch": 3014, "lr": 3.336644525714916e-08}
{"train_loss": 0.09162311255931854, "global_step": 268299, "epoch": 3014, "lr": 3.334527265032028e-08}
{"train_loss": 0.045064184814691544, "global_step": 268300, "epoch": 3014, "lr": 3.3324106760901365e-08}
{"train_loss": 0.02712026610970497, "global_step": 268301, "epoch": 3014, "lr": 3.3302947588886855e-08}
{"train_loss": 0.026250360533595085, "global_step": 268302, "epoch": 3014, "lr": 3.328179513429341e-08}
{"train_loss": 0.025436580181121826, "global_step": 268303, "epoch": 3014, "lr": 3.326064939710993e-08}
{"train_loss": 0.0410902202129364, "global_step": 268304, "epoch": 3014, "lr": 3.3239510377347515e-08}
{"train_loss": 0.028643440455198288, "global_step": 268305, "epoch": 3014, "lr": 3.3218378075006165e-08}
{"train_loss": 0.06400347501039505, "global_step": 268306, "epoch": 3014, "lr": 3.319725249009142e-08}
{"train_loss": 0.024187680333852768, "global_step": 268307, "epoch": 3014, "lr": 3.31761336226033e-08}
{"train_loss": 0.06535063683986664, "global_step": 268308, "epoch": 3014, "lr": 3.3155021472547345e-08}
{"train_loss": 0.017235469073057175, "global_step": 268309, "epoch": 3014, "lr": 3.313391603992355e-08}
{"train_loss": 0.023862609639763832, "global_step": 268310, "epoch": 3014, "lr": 3.311281732473193e-08}
{"train_loss": 0.01563960127532482, "global_step": 268311, "epoch": 3014, "lr": 3.309172532698357e-08}
{"train_loss": 0.06610836833715439, "global_step": 268312, "epoch": 3014, "lr": 3.307064004667848e-08}
{"train_loss": 0.06703685224056244, "global_step": 268313, "epoch": 3014, "lr": 3.304956148381111e-08}
{"train_loss": 0.030107973143458366, "global_step": 268314, "epoch": 3014, "lr": 3.302848963839811e-08}
{"train_loss": 0.045653872191905975, "global_step": 268315, "epoch": 3014, "lr": 3.300742451043393e-08}
{"train_loss": 0.03691290691494942, "global_step": 268316, "epoch": 3014, "lr": 3.298636609992411e-08}
{"train_loss": 0.026382438838481903, "global_step": 268317, "epoch": 3014, "lr": 3.2965314406868675e-08}
{"train_loss": 0.04637778177857399, "global_step": 268318, "epoch": 3014, "lr": 3.29442694312676e-08}
{"train_loss": 0.04990921542048454, "global_step": 268319, "epoch": 3014, "lr": 3.2923231173132006e-08}
{"train_loss": 0.02546306885778904, "global_step": 268320, "epoch": 3014, "lr": 3.2902199632461886e-08}
{"train_loss": 0.016491733491420746, "global_step": 268321, "epoch": 3014, "lr": 3.2881174809257234e-08}
{"train_loss": 0.0401429645717144, "global_step": 268322, "epoch": 3014, "lr": 3.2860156703523604e-08}
{"train_loss": 0.04758691042661667, "global_step": 268323, "epoch": 3014, "lr": 3.2839145315266553e-08}
{"train_loss": 0.07110009342432022, "global_step": 268324, "epoch": 3014, "lr": 3.2818140644480524e-08}
{"train_loss": 0.04143986850976944, "global_step": 268325, "epoch": 3014, "lr": 3.279714269117107e-08}
{"train_loss": 0.051475152373313904, "global_step": 268326, "epoch": 3014, "lr": 3.277615145534929e-08}
{"train_loss": 0.04995454475283623, "global_step": 268327, "epoch": 3014, "lr": 3.275516693700964e-08}
{"train_loss": 0.06951729953289032, "global_step": 268328, "epoch": 3014, "lr": 3.2734189136157666e-08}
{"train_loss": 0.036478351801633835, "global_step": 268329, "epoch": 3014, "lr": 3.271321805279337e-08}
{"train_loss": 0.06537488102912903, "global_step": 268330, "epoch": 3014, "lr": 3.269225368691675e-08}
{"train_loss": 0.0374995656311512, "global_step": 268331, "epoch": 3014, "lr": 3.267129603854446e-08}
{"train_loss": 0.055488184094429016, "global_step": 268332, "epoch": 3014, "lr": 3.2650345107665404e-08}
{"train_loss": 0.026985308155417442, "global_step": 268333, "epoch": 3014, "lr": 3.262940089429067e-08}
{"train_loss": 0.043719668539104924, "global_step": 268334, "epoch": 3014, "lr": 3.260846339842027e-08, "val_loss": 9.63961124420166}
{"train_loss": 0.029371213167905807, "global_step": 268335, "epoch": 3015, "lr": 3.258753262005421e-08}
{"train_loss": 0.0062582120299339294, "global_step": 268336, "epoch": 3015, "lr": 3.256660855919802e-08}
{"train_loss": 0.06323075294494629, "global_step": 268337, "epoch": 3015, "lr": 3.2545691215857264e-08}
{"train_loss": 0.04953734204173088, "global_step": 268338, "epoch": 3015, "lr": 3.2524780590026393e-08}
{"train_loss": 0.04023383930325508, "global_step": 268339, "epoch": 3015, "lr": 3.2503876681716506e-08}
{"train_loss": 0.036464061588048935, "global_step": 268340, "epoch": 3015, "lr": 3.24829794909276e-08}
{"train_loss": 0.07976153492927551, "global_step": 268341, "epoch": 3015, "lr": 3.246208901765968e-08}
{"train_loss": 0.030346985906362534, "global_step": 268342, "epoch": 3015, "lr": 3.2441205261923844e-08}
{"train_loss": 0.06114218011498451, "global_step": 268343, "epoch": 3015, "lr": 3.2420328223714544e-08}
{"train_loss": 0.03940185531973839, "global_step": 268344, "epoch": 3015, "lr": 3.239945790303733e-08}
{"train_loss": 0.05843200534582138, "global_step": 268345, "epoch": 3015, "lr": 3.2378594299892205e-08}
{"train_loss": 0.040042899549007416, "global_step": 268346, "epoch": 3015, "lr": 3.235773741428472e-08}
{"train_loss": 0.040541402995586395, "global_step": 268347, "epoch": 3015, "lr": 3.2336887246220415e-08}
{"train_loss": 0.030190031975507736, "global_step": 268348, "epoch": 3015, "lr": 3.23160437956993e-08}
{"train_loss": 0.0565345399081707, "global_step": 268349, "epoch": 3015, "lr": 3.229520706272693e-08}
{"train_loss": 0.09201215207576752, "global_step": 268350, "epoch": 3015, "lr": 3.2274377047297743e-08}
{"train_loss": 0.01999915950000286, "global_step": 268351, "epoch": 3015, "lr": 3.225355374942285e-08}
{"train_loss": 0.01857919991016388, "global_step": 268352, "epoch": 3015, "lr": 3.223273716910225e-08}
{"train_loss": 0.06999414414167404, "global_step": 268353, "epoch": 3015, "lr": 3.221192730634148e-08}
{"train_loss": 0.042457569390535355, "global_step": 268354, "epoch": 3015, "lr": 3.219112416114056e-08}
{"train_loss": 0.02055482752621174, "global_step": 268355, "epoch": 3015, "lr": 3.217032773349948e-08}
{"train_loss": 0.06219159811735153, "global_step": 268356, "epoch": 3015, "lr": 3.2149538023423795e-08}
{"train_loss": 0.047448039054870605, "global_step": 268357, "epoch": 3015, "lr": 3.212875503091905e-08}
{"train_loss": 0.04717426002025604, "global_step": 268358, "epoch": 3015, "lr": 3.2107978755985255e-08}
{"train_loss": 0.015674646943807602, "global_step": 268359, "epoch": 3015, "lr": 3.20872091986224e-08}
{"train_loss": 0.035805974155664444, "global_step": 268360, "epoch": 3015, "lr": 3.2066446358836046e-08}
{"train_loss": 0.0833987146615982, "global_step": 268361, "epoch": 3015, "lr": 3.204569023663173e-08}
{"train_loss": 0.04735932499170303, "global_step": 268362, "epoch": 3015, "lr": 3.202494083201502e-08}
{"train_loss": 0.04133720323443413, "global_step": 268363, "epoch": 3015, "lr": 3.2004198144974794e-08}
{"train_loss": 0.018768582493066788, "global_step": 268364, "epoch": 3015, "lr": 3.1983462175527724e-08}
{"train_loss": 0.05239272490143776, "global_step": 268365, "epoch": 3015, "lr": 3.196273292366825e-08}
{"train_loss": 0.022320404648780823, "global_step": 268366, "epoch": 3015, "lr": 3.1942010389407475e-08}
{"train_loss": 0.03700282797217369, "global_step": 268367, "epoch": 3015, "lr": 3.192129457273985e-08}
{"train_loss": 0.018784144893288612, "global_step": 268368, "epoch": 3015, "lr": 3.190058547367092e-08}
{"train_loss": 0.04707503318786621, "global_step": 268369, "epoch": 3015, "lr": 3.187988309220069e-08}
{"train_loss": 0.048786818981170654, "global_step": 268370, "epoch": 3015, "lr": 3.1859187428340264e-08}
{"train_loss": 0.016151979565620422, "global_step": 268371, "epoch": 3015, "lr": 3.183849848208409e-08}
{"train_loss": 0.05411437898874283, "global_step": 268372, "epoch": 3015, "lr": 3.1817816253443264e-08}
{"train_loss": 0.034594543278217316, "global_step": 268373, "epoch": 3015, "lr": 3.1797140742412244e-08}
{"train_loss": 0.06456893682479858, "global_step": 268374, "epoch": 3015, "lr": 3.177647194899658e-08}
{"train_loss": 0.035913869738578796, "global_step": 268375, "epoch": 3015, "lr": 3.175580987320182e-08}
{"train_loss": 0.05786601081490517, "global_step": 268376, "epoch": 3015, "lr": 3.1735154515022405e-08}
{"train_loss": 0.03940769284963608, "global_step": 268377, "epoch": 3015, "lr": 3.1714505874474996e-08}
{"train_loss": 0.02985958941280842, "global_step": 268378, "epoch": 3015, "lr": 3.16938639515485e-08}
{"train_loss": 0.07272718101739883, "global_step": 268379, "epoch": 3015, "lr": 3.1673228746254e-08}
{"train_loss": 0.023617571219801903, "global_step": 268380, "epoch": 3015, "lr": 3.165260025859151e-08}
{"train_loss": 0.024820223450660706, "global_step": 268381, "epoch": 3015, "lr": 3.163197848856658e-08}
{"train_loss": 0.038334283977746964, "global_step": 268382, "epoch": 3015, "lr": 3.161136343617921e-08}
{"train_loss": 0.019491558894515038, "global_step": 268383, "epoch": 3015, "lr": 3.159075510142939e-08}
{"train_loss": 0.01702992245554924, "global_step": 268384, "epoch": 3015, "lr": 3.1570153484328235e-08}
{"train_loss": 0.052736274898052216, "global_step": 268385, "epoch": 3015, "lr": 3.154955858487019e-08}
{"train_loss": 0.04058640077710152, "global_step": 268386, "epoch": 3015, "lr": 3.15289704030608e-08}
{"train_loss": 0.04946625977754593, "global_step": 268387, "epoch": 3015, "lr": 3.150838893890562e-08}
{"train_loss": 0.0402328185737133, "global_step": 268388, "epoch": 3015, "lr": 3.148781419240465e-08}
{"train_loss": 0.028727907687425613, "global_step": 268389, "epoch": 3015, "lr": 3.146724616355789e-08}
{"train_loss": 0.02996521070599556, "global_step": 268390, "epoch": 3015, "lr": 3.144668485237645e-08}
{"train_loss": 0.016462601721286774, "global_step": 268391, "epoch": 3015, "lr": 3.142613025885477e-08}
{"train_loss": 0.03288945183157921, "global_step": 268392, "epoch": 3015, "lr": 3.14055823829984e-08}
{"train_loss": 0.02899170108139515, "global_step": 268393, "epoch": 3015, "lr": 3.1385041224807344e-08}
{"train_loss": 0.09147464483976364, "global_step": 268394, "epoch": 3015, "lr": 3.1364506784292705e-08}
{"train_loss": 0.05419911816716194, "global_step": 268395, "epoch": 3015, "lr": 3.134397906145447e-08}
{"train_loss": 0.030216064304113388, "global_step": 268396, "epoch": 3015, "lr": 3.1323458056287115e-08}
{"train_loss": 0.03963856026530266, "global_step": 268397, "epoch": 3015, "lr": 3.1302943768801716e-08}
{"train_loss": 0.03321387991309166, "global_step": 268398, "epoch": 3015, "lr": 3.128243619899829e-08}
{"train_loss": 0.014880799688398838, "global_step": 268399, "epoch": 3015, "lr": 3.126193534687683e-08}
{"train_loss": 0.042180538177490234, "global_step": 268400, "epoch": 3015, "lr": 3.124144121244843e-08}
{"train_loss": 0.04594017565250397, "global_step": 268401, "epoch": 3015, "lr": 3.1220953795707555e-08}
{"train_loss": 0.055039796978235245, "global_step": 268402, "epoch": 3015, "lr": 3.1200473096659744e-08}
{"train_loss": 0.03739839792251587, "global_step": 268403, "epoch": 3015, "lr": 3.1179999115310554e-08}
{"train_loss": 0.048978663980960846, "global_step": 268404, "epoch": 3015, "lr": 3.1159531851659986e-08}
{"train_loss": 0.04527116194367409, "global_step": 268405, "epoch": 3015, "lr": 3.113907130570803e-08}
{"train_loss": 0.0404638908803463, "global_step": 268406, "epoch": 3015, "lr": 3.1118617477465805e-08}
{"train_loss": 0.037910882383584976, "global_step": 268407, "epoch": 3015, "lr": 3.109817036692775e-08}
{"train_loss": 0.02329639531672001, "global_step": 268408, "epoch": 3015, "lr": 3.107772997409386e-08}
{"train_loss": 0.0561462938785553, "global_step": 268409, "epoch": 3015, "lr": 3.10572962989808e-08}
{"train_loss": 0.05762201175093651, "global_step": 268410, "epoch": 3015, "lr": 3.1036869341577454e-08}
{"train_loss": 0.05881023406982422, "global_step": 268411, "epoch": 3015, "lr": 3.101644910189494e-08}
{"train_loss": 0.03527216613292694, "global_step": 268412, "epoch": 3015, "lr": 3.099603557993325e-08}
{"train_loss": 0.019533684477210045, "global_step": 268413, "epoch": 3015, "lr": 3.097562877569793e-08}
{"train_loss": 0.033387765288352966, "global_step": 268414, "epoch": 3015, "lr": 3.0955228689183436e-08}
{"train_loss": 0.06219155341386795, "global_step": 268415, "epoch": 3015, "lr": 3.093483532040087e-08}
{"train_loss": 0.04365804046392441, "global_step": 268416, "epoch": 3015, "lr": 3.091444866935023e-08}
{"train_loss": 0.02269928529858589, "global_step": 268417, "epoch": 3015, "lr": 3.089406873603151e-08}
{"train_loss": 0.03315337747335434, "global_step": 268418, "epoch": 3015, "lr": 3.087369552045028e-08}
{"train_loss": 0.00399198429659009, "global_step": 268419, "epoch": 3015, "lr": 3.085332902261207e-08}
{"train_loss": 0.018376784399151802, "global_step": 268420, "epoch": 3015, "lr": 3.0832969242516884e-08}
{"train_loss": 0.03552056849002838, "global_step": 268421, "epoch": 3015, "lr": 3.081261618016473e-08}
{"train_loss": 0.01950610801577568, "global_step": 268422, "epoch": 3015, "lr": 3.079226983556116e-08}
{"train_loss": 0.04034794663508119, "global_step": 268423, "epoch": 3015, "lr": 3.077193020870617e-08, "val_loss": 9.622221946716309, "train_action_mse_error": 15.95744514465332}
{"train_loss": 0.03136645257472992, "global_step": 268424, "epoch": 3016, "lr": 3.0751597299610857e-08}
{"train_loss": 0.021349169313907623, "global_step": 268425, "epoch": 3016, "lr": 3.073127110826413e-08}
{"train_loss": 0.015872426331043243, "global_step": 268426, "epoch": 3016, "lr": 3.0710951634682625e-08}
{"train_loss": 0.07572762668132782, "global_step": 268427, "epoch": 3016, "lr": 3.069063887886081e-08}
{"train_loss": 0.02463831938803196, "global_step": 268428, "epoch": 3016, "lr": 3.067033284080423e-08}
{"train_loss": 0.0466683991253376, "global_step": 268429, "epoch": 3016, "lr": 3.065003352051288e-08}
{"train_loss": 0.027499115094542503, "global_step": 268430, "epoch": 3016, "lr": 3.0629740917997864e-08}
{"train_loss": 0.031004222109913826, "global_step": 268431, "epoch": 3016, "lr": 3.060945503324808e-08}
{"train_loss": 0.025541620329022408, "global_step": 268432, "epoch": 3016, "lr": 3.0589175866280186e-08}
{"train_loss": 0.0783718004822731, "global_step": 268433, "epoch": 3016, "lr": 3.056890341708307e-08}
{"train_loss": 0.05589722841978073, "global_step": 268434, "epoch": 3016, "lr": 3.05486376856734e-08}
{"train_loss": 0.054882436990737915, "global_step": 268435, "epoch": 3016, "lr": 3.052837867204561e-08}
{"train_loss": 0.049381256103515625, "global_step": 268436, "epoch": 3016, "lr": 3.050812637620526e-08}
{"train_loss": 0.037564221769571304, "global_step": 268437, "epoch": 3016, "lr": 3.0487880798152346e-08}
{"train_loss": 0.02502804435789585, "global_step": 268438, "epoch": 3016, "lr": 3.046764193788687e-08}
{"train_loss": 0.043515242636203766, "global_step": 268439, "epoch": 3016, "lr": 3.044740979542548e-08}
{"train_loss": 0.042105577886104584, "global_step": 268440, "epoch": 3016, "lr": 3.042718437075154e-08}
{"train_loss": 0.061499182134866714, "global_step": 268441, "epoch": 3016, "lr": 3.040696566388723e-08}
{"train_loss": 0.04619339853525162, "global_step": 268442, "epoch": 3016, "lr": 3.0386753674821465e-08}
{"train_loss": 0.06701646000146866, "global_step": 268443, "epoch": 3016, "lr": 3.036654840355979e-08}
{"train_loss": 0.05113895982503891, "global_step": 268444, "epoch": 3016, "lr": 3.034634985010776e-08}
{"train_loss": 0.027191199362277985, "global_step": 268445, "epoch": 3016, "lr": 3.0326158014465364e-08}
{"train_loss": 0.08466708660125732, "global_step": 268446, "epoch": 3016, "lr": 3.0305972896638166e-08}
{"train_loss": 0.04843704029917717, "global_step": 268447, "epoch": 3016, "lr": 3.028579449662616e-08}
{"train_loss": 0.08053608238697052, "global_step": 268448, "epoch": 3016, "lr": 3.026562281442935e-08}
{"train_loss": 0.04439887776970863, "global_step": 268449, "epoch": 3016, "lr": 3.0245457850058834e-08}
{"train_loss": 0.024244695901870728, "global_step": 268450, "epoch": 3016, "lr": 3.0225299603514615e-08}
{"train_loss": 0.04455358907580376, "global_step": 268451, "epoch": 3016, "lr": 3.020514807479669e-08}
{"train_loss": 0.05166420340538025, "global_step": 268452, "epoch": 3016, "lr": 3.018500326390505e-08}
{"train_loss": 0.06757627427577972, "global_step": 268453, "epoch": 3016, "lr": 3.016486517085082e-08}
{"train_loss": 0.04273242503404617, "global_step": 268454, "epoch": 3016, "lr": 3.014473379562843e-08}
{"train_loss": 0.04172975569963455, "global_step": 268455, "epoch": 3016, "lr": 3.0124609138243444e-08}
{"train_loss": 0.021621178835630417, "global_step": 268456, "epoch": 3016, "lr": 3.01044911987014e-08}
{"train_loss": 0.0440329909324646, "global_step": 268457, "epoch": 3016, "lr": 3.008437997700231e-08}
{"train_loss": 0.03593841940164566, "global_step": 268458, "epoch": 3016, "lr": 3.006427547315171e-08}
{"train_loss": 0.020359832793474197, "global_step": 268459, "epoch": 3016, "lr": 3.004417768714407e-08}
{"train_loss": 0.045721374452114105, "global_step": 268460, "epoch": 3016, "lr": 3.0024086618990475e-08}
{"train_loss": 0.03933166339993477, "global_step": 268461, "epoch": 3016, "lr": 3.000400226869094e-08}
{"train_loss": 0.030278628692030907, "global_step": 268462, "epoch": 3016, "lr": 2.998392463625099e-08}
{"train_loss": 0.046864427626132965, "global_step": 268463, "epoch": 3016, "lr": 2.9963853721670654e-08}
{"train_loss": 0.034435611218214035, "global_step": 268464, "epoch": 3016, "lr": 2.9943789524949915e-08}
{"train_loss": 0.01404067687690258, "global_step": 268465, "epoch": 3016, "lr": 2.992373204609433e-08}
{"train_loss": 0.049104101955890656, "global_step": 268466, "epoch": 3016, "lr": 2.99036812851039e-08}
{"train_loss": 0.0432058721780777, "global_step": 268467, "epoch": 3016, "lr": 2.988363724198973e-08}
{"train_loss": 0.026287822052836418, "global_step": 268468, "epoch": 3016, "lr": 2.986359991674626e-08}
{"train_loss": 0.049444880336523056, "global_step": 268469, "epoch": 3016, "lr": 2.984356930937904e-08}
{"train_loss": 0.06022895127534866, "global_step": 268470, "epoch": 3016, "lr": 2.9823545419888076e-08}
{"train_loss": 0.03414039686322212, "global_step": 268471, "epoch": 3016, "lr": 2.9803528248278924e-08}
{"train_loss": 0.059434156864881516, "global_step": 268472, "epoch": 3016, "lr": 2.9783517794557124e-08}
{"train_loss": 0.12925834953784943, "global_step": 268473, "epoch": 3016, "lr": 2.9763514058717136e-08}
{"train_loss": 0.028827210888266563, "global_step": 268474, "epoch": 3016, "lr": 2.974351704077005e-08}
{"train_loss": 0.021483954042196274, "global_step": 268475, "epoch": 3016, "lr": 2.9723526740710327e-08}
{"train_loss": 0.042594894766807556, "global_step": 268476, "epoch": 3016, "lr": 2.9703543158549062e-08}
{"train_loss": 0.07074692845344543, "global_step": 268477, "epoch": 3016, "lr": 2.968356629428626e-08}
{"train_loss": 0.06936276704072952, "global_step": 268478, "epoch": 3016, "lr": 2.9663596147921913e-08}
{"train_loss": 0.06005954369902611, "global_step": 268479, "epoch": 3016, "lr": 2.9643632719456027e-08}
{"train_loss": 0.05836936831474304, "global_step": 268480, "epoch": 3016, "lr": 2.9623676008899703e-08}
{"train_loss": 0.02888670563697815, "global_step": 268481, "epoch": 3016, "lr": 2.960372601625294e-08}
{"train_loss": 0.06503160297870636, "global_step": 268482, "epoch": 3016, "lr": 2.9583782741515742e-08}
{"train_loss": 0.021634722128510475, "global_step": 268483, "epoch": 3016, "lr": 2.9563846184688106e-08}
{"train_loss": 0.027873309329152107, "global_step": 268484, "epoch": 3016, "lr": 2.954391634578113e-08}
{"train_loss": 0.026913993060588837, "global_step": 268485, "epoch": 3016, "lr": 2.952399322478927e-08}
{"train_loss": 0.03521270304918289, "global_step": 268486, "epoch": 3016, "lr": 2.9504076821723624e-08}
{"train_loss": 0.039272136986255646, "global_step": 268487, "epoch": 3016, "lr": 2.9484167136578645e-08}
{"train_loss": 0.03022909164428711, "global_step": 268488, "epoch": 3016, "lr": 2.946426416935988e-08}
{"train_loss": 0.03824881836771965, "global_step": 268489, "epoch": 3016, "lr": 2.944436792007288e-08}
{"train_loss": 0.008931180462241173, "global_step": 268490, "epoch": 3016, "lr": 2.9424478388717646e-08}
{"train_loss": 0.034982599318027496, "global_step": 268491, "epoch": 3016, "lr": 2.9404595575299733e-08}
{"train_loss": 0.017551207914948463, "global_step": 268492, "epoch": 3016, "lr": 2.9384719479813583e-08}
{"train_loss": 0.03609491139650345, "global_step": 268493, "epoch": 3016, "lr": 2.9364850102270302e-08}
{"train_loss": 0.04704797640442848, "global_step": 268494, "epoch": 3016, "lr": 2.9344987442669892e-08}
{"train_loss": 0.06746421754360199, "global_step": 268495, "epoch": 3016, "lr": 2.93251315010179e-08}
{"train_loss": 0.10292521864175797, "global_step": 268496, "epoch": 3016, "lr": 2.9305282277308777e-08}
{"train_loss": 0.04572756588459015, "global_step": 268497, "epoch": 3016, "lr": 2.9285439771553624e-08}
{"train_loss": 0.02075907029211521, "global_step": 268498, "epoch": 3016, "lr": 2.9265603983752444e-08}
{"train_loss": 0.02695957012474537, "global_step": 268499, "epoch": 3016, "lr": 2.9245774913905232e-08}
{"train_loss": 0.09027454257011414, "global_step": 268500, "epoch": 3016, "lr": 2.9225952562017545e-08}
{"train_loss": 0.06402481347322464, "global_step": 268501, "epoch": 3016, "lr": 2.920613692808938e-08}
{"train_loss": 0.04755041375756264, "global_step": 268502, "epoch": 3016, "lr": 2.9186328012131837e-08}
{"train_loss": 0.0638822540640831, "global_step": 268503, "epoch": 3016, "lr": 2.9166525814133817e-08}
{"train_loss": 0.06246211752295494, "global_step": 268504, "epoch": 3016, "lr": 2.914673033411197e-08}
{"train_loss": 0.02341635152697563, "global_step": 268505, "epoch": 3016, "lr": 2.91269415720552e-08}
{"train_loss": 0.0329381488263607, "global_step": 268506, "epoch": 3016, "lr": 2.9107159527980155e-08}
{"train_loss": 0.00733377318829298, "global_step": 268507, "epoch": 3016, "lr": 2.9087384201875732e-08}
{"train_loss": 0.02837243303656578, "global_step": 268508, "epoch": 3016, "lr": 2.906761559375859e-08}
{"train_loss": 0.041019268333911896, "global_step": 268509, "epoch": 3016, "lr": 2.9047853703617623e-08}
{"train_loss": 0.03238162398338318, "global_step": 268510, "epoch": 3016, "lr": 2.9028098531463932e-08}
{"train_loss": 0.0425238236784935, "global_step": 268511, "epoch": 3016, "lr": 2.900835007730307e-08}
{"train_loss": 0.04372438506924369, "global_step": 268512, "epoch": 3016, "lr": 2.8988608341129486e-08, "val_loss": 9.528752326965332}
{"train_loss": 0.03458114713430405, "global_step": 268513, "epoch": 3017, "lr": 2.896887332294873e-08}
{"train_loss": 0.029311593621969223, "global_step": 268514, "epoch": 3017, "lr": 2.8949145022760804e-08}
{"train_loss": 0.02982381545007229, "global_step": 268515, "epoch": 3017, "lr": 2.892942344057681e-08}
{"train_loss": 0.04670456051826477, "global_step": 268516, "epoch": 3017, "lr": 2.8909708576391192e-08}
{"train_loss": 0.03672529384493828, "global_step": 268517, "epoch": 3017, "lr": 2.8890000430215058e-08}
{"train_loss": 0.03602339327335358, "global_step": 268518, "epoch": 3017, "lr": 2.8870299002042854e-08}
{"train_loss": 0.016457006335258484, "global_step": 268519, "epoch": 3017, "lr": 2.8850604291874584e-08}
{"train_loss": 0.05212298408150673, "global_step": 268520, "epoch": 3017, "lr": 2.8830916299726895e-08}
{"train_loss": 0.02503613755106926, "global_step": 268521, "epoch": 3017, "lr": 2.881123502558869e-08}
{"train_loss": 0.047989148646593094, "global_step": 268522, "epoch": 3017, "lr": 2.8791560469465516e-08}
{"train_loss": 0.05972066521644592, "global_step": 268523, "epoch": 3017, "lr": 2.877189263136848e-08}
{"train_loss": 0.0471554733812809, "global_step": 268524, "epoch": 3017, "lr": 2.8752231511286477e-08}
{"train_loss": 0.035766009241342545, "global_step": 268525, "epoch": 3017, "lr": 2.873257710923616e-08}
{"train_loss": 0.06665033102035522, "global_step": 268526, "epoch": 3017, "lr": 2.8712929425211976e-08}
{"train_loss": 0.05728977546095848, "global_step": 268527, "epoch": 3017, "lr": 2.869328845921948e-08}
{"train_loss": 0.04320777207612991, "global_step": 268528, "epoch": 3017, "lr": 2.867365421125867e-08}
{"train_loss": 0.049858592450618744, "global_step": 268529, "epoch": 3017, "lr": 2.8654026681329548e-08}
{"train_loss": 0.08271922916173935, "global_step": 268530, "epoch": 3017, "lr": 2.8634405869443216e-08}
{"train_loss": 0.10227520018815994, "global_step": 268531, "epoch": 3017, "lr": 2.861479177559967e-08}
{"train_loss": 0.060976266860961914, "global_step": 268532, "epoch": 3017, "lr": 2.8595184399798914e-08}
{"train_loss": 0.04629220440983772, "global_step": 268533, "epoch": 3017, "lr": 2.8575583742040947e-08}
{"train_loss": 0.05102277919650078, "global_step": 268534, "epoch": 3017, "lr": 2.855598980233687e-08}
{"train_loss": 0.04210563376545906, "global_step": 268535, "epoch": 3017, "lr": 2.853640258068113e-08}
{"train_loss": 0.0408109687268734, "global_step": 268536, "epoch": 3017, "lr": 2.8516822077079286e-08}
{"train_loss": 0.012013946659862995, "global_step": 268537, "epoch": 3017, "lr": 2.8497248291536882e-08}
{"train_loss": 0.06092492863535881, "global_step": 268538, "epoch": 3017, "lr": 2.847768122404837e-08}
{"train_loss": 0.028778232634067535, "global_step": 268539, "epoch": 3017, "lr": 2.8458120874624848e-08}
{"train_loss": 0.07247747480869293, "global_step": 268540, "epoch": 3017, "lr": 2.8438567243271874e-08}
{"train_loss": 0.031817302107810974, "global_step": 268541, "epoch": 3017, "lr": 2.8419020329978342e-08}
{"train_loss": 0.032233934849500656, "global_step": 268542, "epoch": 3017, "lr": 2.8399480134760903e-08}
{"train_loss": 0.03792060911655426, "global_step": 268543, "epoch": 3017, "lr": 2.8379946657608457e-08}
{"train_loss": 0.02403469756245613, "global_step": 268544, "epoch": 3017, "lr": 2.836041989853766e-08}
{"train_loss": 0.047556858509778976, "global_step": 268545, "epoch": 3017, "lr": 2.8340899857542958e-08}
{"train_loss": 0.06503455340862274, "global_step": 268546, "epoch": 3017, "lr": 2.83213865346299e-08}
{"train_loss": 0.03388267010450363, "global_step": 268547, "epoch": 3017, "lr": 2.830187992979849e-08}
{"train_loss": 0.08137796819210052, "global_step": 268548, "epoch": 3017, "lr": 2.828238004304873e-08}
{"train_loss": 0.015923669561743736, "global_step": 268549, "epoch": 3017, "lr": 2.8262886874397266e-08}
{"train_loss": 0.03708614036440849, "global_step": 268550, "epoch": 3017, "lr": 2.824340042382745e-08}
{"train_loss": 0.0411188006401062, "global_step": 268551, "epoch": 3017, "lr": 2.822392069135593e-08}
{"train_loss": 0.05564629286527634, "global_step": 268552, "epoch": 3017, "lr": 2.8204447676977162e-08}
{"train_loss": 0.044679105281829834, "global_step": 268553, "epoch": 3017, "lr": 2.8184981380702248e-08}
{"train_loss": 0.04370657727122307, "global_step": 268554, "epoch": 3017, "lr": 2.816552180252563e-08}
{"train_loss": 0.030092909932136536, "global_step": 268555, "epoch": 3017, "lr": 2.8146068942452864e-08}
{"train_loss": 0.06582114100456238, "global_step": 268556, "epoch": 3017, "lr": 2.81266228004895e-08}
{"train_loss": 0.05477248504757881, "global_step": 268557, "epoch": 3017, "lr": 2.8107183376635536e-08}
{"train_loss": 0.02153938263654709, "global_step": 268558, "epoch": 3017, "lr": 2.8087750670890976e-08}
{"train_loss": 0.033138979226350784, "global_step": 268559, "epoch": 3017, "lr": 2.806832468326137e-08}
{"train_loss": 0.027684520930051804, "global_step": 268560, "epoch": 3017, "lr": 2.8048905413746718e-08}
{"train_loss": 0.021986985579133034, "global_step": 268561, "epoch": 3017, "lr": 2.8029492862352568e-08}
{"train_loss": 0.01109097059816122, "global_step": 268562, "epoch": 3017, "lr": 2.8010087029084476e-08}
{"train_loss": 0.05247046425938606, "global_step": 268563, "epoch": 3017, "lr": 2.7990687913942437e-08}
{"train_loss": 0.04804093390703201, "global_step": 268564, "epoch": 3017, "lr": 2.7971295516920902e-08}
{"train_loss": 0.017939036712050438, "global_step": 268565, "epoch": 3017, "lr": 2.7951909838036526e-08}
{"train_loss": 0.05339658260345459, "global_step": 268566, "epoch": 3017, "lr": 2.7932530877278206e-08}
{"train_loss": 0.017931289970874786, "global_step": 268567, "epoch": 3017, "lr": 2.7913158634662594e-08}
{"train_loss": 0.035046786069869995, "global_step": 268568, "epoch": 3017, "lr": 2.789379311017859e-08}
{"train_loss": 0.08983911573886871, "global_step": 268569, "epoch": 3017, "lr": 2.7874434303837294e-08}
{"train_loss": 0.018706925213336945, "global_step": 268570, "epoch": 3017, "lr": 2.785508221564426e-08}
{"train_loss": 0.05354752391576767, "global_step": 268571, "epoch": 3017, "lr": 2.783573684558838e-08}
{"train_loss": 0.03609386086463928, "global_step": 268572, "epoch": 3017, "lr": 2.781639819368631e-08}
{"train_loss": 0.017574286088347435, "global_step": 268573, "epoch": 3017, "lr": 2.7797066259932503e-08}
{"train_loss": 0.05876610800623894, "global_step": 268574, "epoch": 3017, "lr": 2.7777741044338058e-08}
{"train_loss": 0.015471497550606728, "global_step": 268575, "epoch": 3017, "lr": 2.775842254689187e-08}
{"train_loss": 0.03892536088824272, "global_step": 268576, "epoch": 3017, "lr": 2.7739110767610598e-08}
{"train_loss": 0.04891399294137955, "global_step": 268577, "epoch": 3017, "lr": 2.7719805706488688e-08}
{"train_loss": 0.0296010859310627, "global_step": 268578, "epoch": 3017, "lr": 2.770050736353169e-08}
{"train_loss": 0.02536933682858944, "global_step": 268579, "epoch": 3017, "lr": 2.7681215738739607e-08}
{"train_loss": 0.013706985861063004, "global_step": 268580, "epoch": 3017, "lr": 2.7661930832117988e-08}
{"train_loss": 0.025424007326364517, "global_step": 268581, "epoch": 3017, "lr": 2.7642652643666835e-08}
{"train_loss": 0.030582230538129807, "global_step": 268582, "epoch": 3017, "lr": 2.7623381173386143e-08}
{"train_loss": 0.03611062467098236, "global_step": 268583, "epoch": 3017, "lr": 2.7604116421287018e-08}
{"train_loss": 0.05640376731753349, "global_step": 268584, "epoch": 3017, "lr": 2.758485838736946e-08}
{"train_loss": 0.03141322359442711, "global_step": 268585, "epoch": 3017, "lr": 2.756560707163347e-08}
{"train_loss": 0.023685980588197708, "global_step": 268586, "epoch": 3017, "lr": 2.7546362474079047e-08}
{"train_loss": 0.030175453051924706, "global_step": 268587, "epoch": 3017, "lr": 2.752712459471174e-08}
{"train_loss": 0.04575193673372269, "global_step": 268588, "epoch": 3017, "lr": 2.75078934335371e-08}
{"train_loss": 0.011201245710253716, "global_step": 268589, "epoch": 3017, "lr": 2.7488668990555132e-08}
{"train_loss": 0.038006484508514404, "global_step": 268590, "epoch": 3017, "lr": 2.7469451265765833e-08}
{"train_loss": 0.04430139437317848, "global_step": 268591, "epoch": 3017, "lr": 2.7450240259174752e-08}
{"train_loss": 0.05630219355225563, "global_step": 268592, "epoch": 3017, "lr": 2.7431035970787446e-08}
{"train_loss": 0.04779297113418579, "global_step": 268593, "epoch": 3017, "lr": 2.7411838400598355e-08}
{"train_loss": 0.03301273286342621, "global_step": 268594, "epoch": 3017, "lr": 2.7392647548613038e-08}
{"train_loss": 0.059602636843919754, "global_step": 268595, "epoch": 3017, "lr": 2.7373463414842593e-08}
{"train_loss": 0.057173024863004684, "global_step": 268596, "epoch": 3017, "lr": 2.735428599927592e-08}
{"train_loss": 0.06580127775669098, "global_step": 268597, "epoch": 3017, "lr": 2.7335115301929672e-08}
{"train_loss": 0.03835970535874367, "global_step": 268598, "epoch": 3017, "lr": 2.7315951322792744e-08}
{"train_loss": 0.022237515076994896, "global_step": 268599, "epoch": 3017, "lr": 2.729679406187624e-08}
{"train_loss": 0.036585450172424316, "global_step": 268600, "epoch": 3017, "lr": 2.7277643519180163e-08}
{"train_loss": 0.041504838912004835, "global_step": 268601, "epoch": 3017, "lr": 2.725849969471006e-08, "val_loss": 9.626203536987305}
{"train_loss": 0.014923976734280586, "global_step": 268602, "epoch": 3018, "lr": 2.7239362588460383e-08}
{"train_loss": 0.022188853472471237, "global_step": 268603, "epoch": 3018, "lr": 2.722023220044223e-08}
{"train_loss": 0.064249686896801, "global_step": 268604, "epoch": 3018, "lr": 2.7201108530661156e-08}
{"train_loss": 0.025095678865909576, "global_step": 268605, "epoch": 3018, "lr": 2.718199157910606e-08}
{"train_loss": 0.05685869976878166, "global_step": 268606, "epoch": 3018, "lr": 2.7162881345788037e-08}
{"train_loss": 0.06278432905673981, "global_step": 268607, "epoch": 3018, "lr": 2.7143777830712647e-08}
{"train_loss": 0.08131517469882965, "global_step": 268608, "epoch": 3018, "lr": 2.7124681033874334e-08}
{"train_loss": 0.035719141364097595, "global_step": 268609, "epoch": 3018, "lr": 2.7105590955284198e-08}
{"train_loss": 0.03627840057015419, "global_step": 268610, "epoch": 3018, "lr": 2.7086507594942245e-08}
{"train_loss": 0.048363592475652695, "global_step": 268611, "epoch": 3018, "lr": 2.706743095284292e-08}
{"train_loss": 0.03251195698976517, "global_step": 268612, "epoch": 3018, "lr": 2.7048361028997327e-08}
{"train_loss": 0.03803800046443939, "global_step": 268613, "epoch": 3018, "lr": 2.7029297823405465e-08}
{"train_loss": 0.05423295497894287, "global_step": 268614, "epoch": 3018, "lr": 2.7010241336072884e-08}
{"train_loss": 0.01264920737594366, "global_step": 268615, "epoch": 3018, "lr": 2.6991191566999585e-08}
{"train_loss": 0.04333772510290146, "global_step": 268616, "epoch": 3018, "lr": 2.6972148516191122e-08}
{"train_loss": 0.05183917656540871, "global_step": 268617, "epoch": 3018, "lr": 2.6953112183641937e-08}
{"train_loss": 0.035490166395902634, "global_step": 268618, "epoch": 3018, "lr": 2.693408256936314e-08}
{"train_loss": 0.07330913841724396, "global_step": 268619, "epoch": 3018, "lr": 2.6915059673349174e-08}
{"train_loss": 0.05093047767877579, "global_step": 268620, "epoch": 3018, "lr": 2.6896043495611146e-08}
{"train_loss": 0.02339092642068863, "global_step": 268621, "epoch": 3018, "lr": 2.687703403614905e-08}
{"train_loss": 0.029096823185682297, "global_step": 268622, "epoch": 3018, "lr": 2.685803129495734e-08}
{"train_loss": 0.048566896468400955, "global_step": 268623, "epoch": 3018, "lr": 2.683903527205267e-08}
{"train_loss": 0.03434893116354942, "global_step": 268624, "epoch": 3018, "lr": 2.6820045967429485e-08}
{"train_loss": 0.03450930491089821, "global_step": 268625, "epoch": 3018, "lr": 2.680106338108779e-08}
{"train_loss": 0.07949382811784744, "global_step": 268626, "epoch": 3018, "lr": 2.6782087513038677e-08}
{"train_loss": 0.03095983900129795, "global_step": 268627, "epoch": 3018, "lr": 2.6763118363276606e-08}
{"train_loss": 0.0523032546043396, "global_step": 268628, "epoch": 3018, "lr": 2.6744155931807126e-08}
{"train_loss": 0.02695237658917904, "global_step": 268629, "epoch": 3018, "lr": 2.6725200218630232e-08}
{"train_loss": 0.04054024815559387, "global_step": 268630, "epoch": 3018, "lr": 2.670625122375703e-08}
{"train_loss": 0.031177381053566933, "global_step": 268631, "epoch": 3018, "lr": 2.6687308947176413e-08}
{"train_loss": 0.06547269225120544, "global_step": 268632, "epoch": 3018, "lr": 2.6668373388905044e-08}
{"train_loss": 0.03788459300994873, "global_step": 268633, "epoch": 3018, "lr": 2.6649444548937362e-08}
{"train_loss": 0.021804189309477806, "global_step": 268634, "epoch": 3018, "lr": 2.6630522427278926e-08}
{"train_loss": 0.036376163363456726, "global_step": 268635, "epoch": 3018, "lr": 2.661160702392973e-08}
{"train_loss": 0.04237297177314758, "global_step": 268636, "epoch": 3018, "lr": 2.6592698338889777e-08}
{"train_loss": 0.02185673639178276, "global_step": 268637, "epoch": 3018, "lr": 2.6573796372170168e-08}
{"train_loss": 0.04533087834715843, "global_step": 268638, "epoch": 3018, "lr": 2.655490112376535e-08}
{"train_loss": 0.042519014328718185, "global_step": 268639, "epoch": 3018, "lr": 2.653601259368088e-08}
{"train_loss": 0.06616777181625366, "global_step": 268640, "epoch": 3018, "lr": 2.6517130781922306e-08}
{"train_loss": 0.07058664411306381, "global_step": 268641, "epoch": 3018, "lr": 2.6498255688489627e-08}
{"train_loss": 0.02961481735110283, "global_step": 268642, "epoch": 3018, "lr": 2.647938731338284e-08}
{"train_loss": 0.09048259258270264, "global_step": 268643, "epoch": 3018, "lr": 2.6460525656607505e-08}
{"train_loss": 0.03605463355779648, "global_step": 268644, "epoch": 3018, "lr": 2.6441670718169163e-08}
{"train_loss": 0.05770621448755264, "global_step": 268645, "epoch": 3018, "lr": 2.6422822498062272e-08}
{"train_loss": 0.03793826699256897, "global_step": 268646, "epoch": 3018, "lr": 2.6403980996292378e-08}
{"train_loss": 0.03042871318757534, "global_step": 268647, "epoch": 3018, "lr": 2.638514621287058e-08}
{"train_loss": 0.09586050361394882, "global_step": 268648, "epoch": 3018, "lr": 2.6366318147785784e-08}
{"train_loss": 0.028213385492563248, "global_step": 268649, "epoch": 3018, "lr": 2.6347496801049087e-08}
{"train_loss": 0.03837582841515541, "global_step": 268650, "epoch": 3018, "lr": 2.632868217266049e-08}
{"train_loss": 0.048933371901512146, "global_step": 268651, "epoch": 3018, "lr": 2.6309874262619993e-08}
{"train_loss": 0.07163263112306595, "global_step": 268652, "epoch": 3018, "lr": 2.62910730709387e-08}
{"train_loss": 0.047256141901016235, "global_step": 268653, "epoch": 3018, "lr": 2.6272278597611054e-08}
{"train_loss": 0.04834932088851929, "global_step": 268654, "epoch": 3018, "lr": 2.6253490842642614e-08}
{"train_loss": 0.03223397582769394, "global_step": 268655, "epoch": 3018, "lr": 2.6234709806033376e-08}
{"train_loss": 0.021053563803434372, "global_step": 268656, "epoch": 3018, "lr": 2.621593548778889e-08}
{"train_loss": 0.03878903388977051, "global_step": 268657, "epoch": 3018, "lr": 2.619716788790916e-08}
{"train_loss": 0.024676065891981125, "global_step": 268658, "epoch": 3018, "lr": 2.617840700639973e-08}
{"train_loss": 0.032720889896154404, "global_step": 268659, "epoch": 3018, "lr": 2.6159652843260606e-08}
{"train_loss": 0.008710560388863087, "global_step": 268660, "epoch": 3018, "lr": 2.614090539849734e-08}
{"train_loss": 0.031413353979587555, "global_step": 268661, "epoch": 3018, "lr": 2.6122164672109927e-08}
{"train_loss": 0.03067779168486595, "global_step": 268662, "epoch": 3018, "lr": 2.6103430664098372e-08}
{"train_loss": 0.05736551806330681, "global_step": 268663, "epoch": 3018, "lr": 2.6084703374468223e-08}
{"train_loss": 0.0432039238512516, "global_step": 268664, "epoch": 3018, "lr": 2.606598280322503e-08}
{"train_loss": 0.04005984216928482, "global_step": 268665, "epoch": 3018, "lr": 2.6047268950368797e-08}
{"train_loss": 0.023713309317827225, "global_step": 268666, "epoch": 3018, "lr": 2.602856181589952e-08}
{"train_loss": 0.023429660126566887, "global_step": 268667, "epoch": 3018, "lr": 2.6009861399822753e-08}
{"train_loss": 0.05421477556228638, "global_step": 268668, "epoch": 3018, "lr": 2.5991167702138497e-08}
{"train_loss": 0.024993956089019775, "global_step": 268669, "epoch": 3018, "lr": 2.5972480722846747e-08}
{"train_loss": 0.0294126458466053, "global_step": 268670, "epoch": 3018, "lr": 2.5953800461958612e-08}
{"train_loss": 0.06806021928787231, "global_step": 268671, "epoch": 3018, "lr": 2.5935126919474084e-08}
{"train_loss": 0.036088455468416214, "global_step": 268672, "epoch": 3018, "lr": 2.591646009538762e-08}
{"train_loss": 0.049244437366724014, "global_step": 268673, "epoch": 3018, "lr": 2.5897799989710314e-08}
{"train_loss": 0.02175106666982174, "global_step": 268674, "epoch": 3018, "lr": 2.587914660244217e-08}
{"train_loss": 0.04027479141950607, "global_step": 268675, "epoch": 3018, "lr": 2.5860499933583192e-08}
{"train_loss": 0.01964220032095909, "global_step": 268676, "epoch": 3018, "lr": 2.5841859983144477e-08}
{"train_loss": 0.05491304397583008, "global_step": 268677, "epoch": 3018, "lr": 2.5823226751114925e-08}
{"train_loss": 0.034293584525585175, "global_step": 268678, "epoch": 3018, "lr": 2.5804600237511188e-08}
{"train_loss": 0.031544726341962814, "global_step": 268679, "epoch": 3018, "lr": 2.5785980442322167e-08}
{"train_loss": 0.025778774172067642, "global_step": 268680, "epoch": 3018, "lr": 2.576736736556451e-08}
{"train_loss": 0.05998508632183075, "global_step": 268681, "epoch": 3018, "lr": 2.5748761007227118e-08}
{"train_loss": 0.05717211216688156, "global_step": 268682, "epoch": 3018, "lr": 2.5730161367321093e-08}
{"train_loss": 0.05425795167684555, "global_step": 268683, "epoch": 3018, "lr": 2.5711568445846436e-08}
{"train_loss": 0.034767597913742065, "global_step": 268684, "epoch": 3018, "lr": 2.5692982242803144e-08}
{"train_loss": 0.10073370486497879, "global_step": 268685, "epoch": 3018, "lr": 2.567440275820232e-08}
{"train_loss": 0.023281587287783623, "global_step": 268686, "epoch": 3018, "lr": 2.5655829992032865e-08}
{"train_loss": 0.06141495704650879, "global_step": 268687, "epoch": 3018, "lr": 2.563726394431143e-08}
{"train_loss": 0.020110031589865685, "global_step": 268688, "epoch": 3018, "lr": 2.5618704615032464e-08}
{"train_loss": 0.019230704754590988, "global_step": 268689, "epoch": 3018, "lr": 2.5600152004195966e-08}
{"train_loss": 0.04239718101165268, "global_step": 268690, "epoch": 3018, "lr": 2.5581606111813038e-08, "val_loss": 9.695754051208496}
{"train_loss": 0.07859942317008972, "global_step": 268691, "epoch": 3019, "lr": 2.5563066937878134e-08}
{"train_loss": 0.008170930668711662, "global_step": 268692, "epoch": 3019, "lr": 2.55445344823968e-08}
{"train_loss": 0.02356754243373871, "global_step": 268693, "epoch": 3019, "lr": 2.5526008745374586e-08}
{"train_loss": 0.04903154447674751, "global_step": 268694, "epoch": 3019, "lr": 2.5507489726811495e-08}
{"train_loss": 0.051161106675863266, "global_step": 268695, "epoch": 3019, "lr": 2.548897742671308e-08}
{"train_loss": 0.044389382004737854, "global_step": 268696, "epoch": 3019, "lr": 2.5470471845073784e-08}
{"train_loss": 0.03245066478848457, "global_step": 268697, "epoch": 3019, "lr": 2.5451972981899164e-08}
{"train_loss": 0.03466904163360596, "global_step": 268698, "epoch": 3019, "lr": 2.5433480837194766e-08}
{"train_loss": 0.037810709327459335, "global_step": 268699, "epoch": 3019, "lr": 2.5414995410966147e-08}
{"train_loss": 0.059502068907022476, "global_step": 268700, "epoch": 3019, "lr": 2.5396516703207752e-08}
{"train_loss": 0.030534712597727776, "global_step": 268701, "epoch": 3019, "lr": 2.5378044713925132e-08}
{"train_loss": 0.04826731979846954, "global_step": 268702, "epoch": 3019, "lr": 2.535957944312384e-08}
{"train_loss": 0.033399488776922226, "global_step": 268703, "epoch": 3019, "lr": 2.534112089079832e-08}
{"train_loss": 0.020718077197670937, "global_step": 268704, "epoch": 3019, "lr": 2.5322669056959682e-08}
{"train_loss": 0.05403386801481247, "global_step": 268705, "epoch": 3019, "lr": 2.530422394160792e-08}
{"train_loss": 0.024793727323412895, "global_step": 268706, "epoch": 3019, "lr": 2.5285785544748587e-08}
{"train_loss": 0.01992521807551384, "global_step": 268707, "epoch": 3019, "lr": 2.526735386637613e-08}
{"train_loss": 0.06282860040664673, "global_step": 268708, "epoch": 3019, "lr": 2.5248928906496106e-08}
{"train_loss": 0.05558260157704353, "global_step": 268709, "epoch": 3019, "lr": 2.523051066511406e-08}
{"train_loss": 0.0488959439098835, "global_step": 268710, "epoch": 3019, "lr": 2.5212099142229993e-08}
{"train_loss": 0.03764275088906288, "global_step": 268711, "epoch": 3019, "lr": 2.519369433784946e-08}
{"train_loss": 0.09211643785238266, "global_step": 268712, "epoch": 3019, "lr": 2.5175296251972457e-08}
{"train_loss": 0.04213789105415344, "global_step": 268713, "epoch": 3019, "lr": 2.5156904884598985e-08}
{"train_loss": 0.025266576558351517, "global_step": 268714, "epoch": 3019, "lr": 2.5138520235734596e-08}
{"train_loss": 0.07213611900806427, "global_step": 268715, "epoch": 3019, "lr": 2.5120142305384843e-08}
{"train_loss": 0.03671577200293541, "global_step": 268716, "epoch": 3019, "lr": 2.510177109354417e-08}
{"train_loss": 0.03712373599410057, "global_step": 268717, "epoch": 3019, "lr": 2.508340660021813e-08}
{"train_loss": 0.04205596446990967, "global_step": 268718, "epoch": 3019, "lr": 2.5065048825417826e-08}
{"train_loss": 0.047670476138591766, "global_step": 268719, "epoch": 3019, "lr": 2.5046697769132156e-08}
{"train_loss": 0.05900293216109276, "global_step": 268720, "epoch": 3019, "lr": 2.502835343137222e-08}
{"train_loss": 0.05109628289937973, "global_step": 268721, "epoch": 3019, "lr": 2.501001581213802e-08}
{"train_loss": 0.03569496050477028, "global_step": 268722, "epoch": 3019, "lr": 2.499168491142956e-08}
{"train_loss": 0.07214505225419998, "global_step": 268723, "epoch": 3019, "lr": 2.497336072925238e-08}
{"train_loss": 0.04641349986195564, "global_step": 268724, "epoch": 3019, "lr": 2.4955043265612044e-08}
{"train_loss": 0.013227354735136032, "global_step": 268725, "epoch": 3019, "lr": 2.493673252050299e-08}
{"train_loss": 0.024757428094744682, "global_step": 268726, "epoch": 3019, "lr": 2.491842849393633e-08}
{"train_loss": 0.02342980168759823, "global_step": 268727, "epoch": 3019, "lr": 2.4900131185906507e-08}
{"train_loss": 0.024607855826616287, "global_step": 268728, "epoch": 3019, "lr": 2.488184059642462e-08}
{"train_loss": 0.03421517834067345, "global_step": 268729, "epoch": 3019, "lr": 2.4863556725485126e-08}
{"train_loss": 0.03251475840806961, "global_step": 268730, "epoch": 3019, "lr": 2.4845279573093573e-08}
{"train_loss": 0.0332360677421093, "global_step": 268731, "epoch": 3019, "lr": 2.4827009139249957e-08}
{"train_loss": 0.06969822943210602, "global_step": 268732, "epoch": 3019, "lr": 2.4808745423959835e-08}
{"train_loss": 0.029863491654396057, "global_step": 268733, "epoch": 3019, "lr": 2.4790488427228753e-08}
{"train_loss": 0.03562922775745392, "global_step": 268734, "epoch": 3019, "lr": 2.4772238149056714e-08}
{"train_loss": 0.08246280997991562, "global_step": 268735, "epoch": 3019, "lr": 2.4753994589438167e-08}
{"train_loss": 0.02854854241013527, "global_step": 268736, "epoch": 3019, "lr": 2.4735757748389766e-08}
{"train_loss": 0.03332606330513954, "global_step": 268737, "epoch": 3019, "lr": 2.4717527625900406e-08}
{"train_loss": 0.030039630830287933, "global_step": 268738, "epoch": 3019, "lr": 2.469930422198119e-08}
{"train_loss": 0.05612149089574814, "global_step": 268739, "epoch": 3019, "lr": 2.4681087536632118e-08}
{"train_loss": 0.05577186867594719, "global_step": 268740, "epoch": 3019, "lr": 2.4662877569853192e-08}
{"train_loss": 0.06677877902984619, "global_step": 268741, "epoch": 3019, "lr": 2.4644674321655513e-08}
{"train_loss": 0.0491192601621151, "global_step": 268742, "epoch": 3019, "lr": 2.462647779202798e-08}
{"train_loss": 0.0641779750585556, "global_step": 268743, "epoch": 3019, "lr": 2.4608287980987244e-08}
{"train_loss": 0.013618828728795052, "global_step": 268744, "epoch": 3019, "lr": 2.4590104888522203e-08}
{"train_loss": 0.02273610420525074, "global_step": 268745, "epoch": 3019, "lr": 2.457192851464396e-08}
{"train_loss": 0.06006421893835068, "global_step": 268746, "epoch": 3019, "lr": 2.4553758859352517e-08}
{"train_loss": 0.034426555037498474, "global_step": 268747, "epoch": 3019, "lr": 2.453559592265342e-08}
{"train_loss": 0.066934734582901, "global_step": 268748, "epoch": 3019, "lr": 2.4517439704546674e-08}
{"train_loss": 0.047257401049137115, "global_step": 268749, "epoch": 3019, "lr": 2.4499290205032275e-08}
{"train_loss": 0.03939659893512726, "global_step": 268750, "epoch": 3019, "lr": 2.448114742411578e-08}
{"train_loss": 0.0456416942179203, "global_step": 268751, "epoch": 3019, "lr": 2.4463011361797182e-08}
{"train_loss": 0.024150878190994263, "global_step": 268752, "epoch": 3019, "lr": 2.444488201808204e-08}
{"train_loss": 0.07291794568300247, "global_step": 268753, "epoch": 3019, "lr": 2.4426759392970345e-08}
{"train_loss": 0.04201661795377731, "global_step": 268754, "epoch": 3019, "lr": 2.4408643486467654e-08}
{"train_loss": 0.01953674852848053, "global_step": 268755, "epoch": 3019, "lr": 2.4390534298568414e-08}
{"train_loss": 0.04795237258076668, "global_step": 268756, "epoch": 3019, "lr": 2.4372431829283726e-08}
{"train_loss": 0.04953831434249878, "global_step": 268757, "epoch": 3019, "lr": 2.4354336078613593e-08}
{"train_loss": 0.08328080177307129, "global_step": 268758, "epoch": 3019, "lr": 2.4336247046563565e-08}
{"train_loss": 0.034820377826690674, "global_step": 268759, "epoch": 3019, "lr": 2.431816473312809e-08}
{"train_loss": 0.033323075622320175, "global_step": 268760, "epoch": 3019, "lr": 2.430008913831272e-08}
{"train_loss": 0.04723989963531494, "global_step": 268761, "epoch": 3019, "lr": 2.4282020262123006e-08}
{"train_loss": 0.030436217784881592, "global_step": 268762, "epoch": 3019, "lr": 2.426395810455895e-08}
{"train_loss": 0.0537339486181736, "global_step": 268763, "epoch": 3019, "lr": 2.4245902665626097e-08}
{"train_loss": 0.060492467135190964, "global_step": 268764, "epoch": 3019, "lr": 2.42278539453189e-08}
{"train_loss": 0.034551817923784256, "global_step": 268765, "epoch": 3019, "lr": 2.4209811943648462e-08}
{"train_loss": 0.029137831181287766, "global_step": 268766, "epoch": 3019, "lr": 2.4191776660614784e-08}
{"train_loss": 0.029036201536655426, "global_step": 268767, "epoch": 3019, "lr": 2.417374809621786e-08}
{"train_loss": 0.05889444798231125, "global_step": 268768, "epoch": 3019, "lr": 2.4155726250463252e-08}
{"train_loss": 0.07446140795946121, "global_step": 268769, "epoch": 3019, "lr": 2.4137711123350948e-08}
{"train_loss": 0.03325321525335312, "global_step": 268770, "epoch": 3019, "lr": 2.4119702714880953e-08}
{"train_loss": 0.026556089520454407, "global_step": 268771, "epoch": 3019, "lr": 2.410170102506437e-08}
{"train_loss": 0.03409941866993904, "global_step": 268772, "epoch": 3019, "lr": 2.408370605389565e-08}
{"train_loss": 0.04727248102426529, "global_step": 268773, "epoch": 3019, "lr": 2.4065717801380337e-08}
{"train_loss": 0.024411272257566452, "global_step": 268774, "epoch": 3019, "lr": 2.404773626751844e-08}
{"train_loss": 0.0724417120218277, "global_step": 268775, "epoch": 3019, "lr": 2.4029761452315502e-08}
{"train_loss": 0.036265160888433456, "global_step": 268776, "epoch": 3019, "lr": 2.4011793355771527e-08}
{"train_loss": 0.07685863226652145, "global_step": 268777, "epoch": 3019, "lr": 2.399383197789207e-08}
{"train_loss": 0.04803203418850899, "global_step": 268778, "epoch": 3019, "lr": 2.397587731867712e-08}
{"train_loss": 0.04432714161243331, "global_step": 268779, "epoch": 3019, "lr": 2.395792937813224e-08, "val_loss": 9.654273986816406}
{"train_loss": 0.09968680143356323, "global_step": 268780, "epoch": 3020, "lr": 2.3939988156251867e-08}
{"train_loss": 0.05457918718457222, "global_step": 268781, "epoch": 3020, "lr": 2.3922053653047117e-08}
{"train_loss": 0.044283896684646606, "global_step": 268782, "epoch": 3020, "lr": 2.390412586851243e-08}
{"train_loss": 0.09327899664640427, "global_step": 268783, "epoch": 3020, "lr": 2.3886204802658908e-08}
{"train_loss": 0.020514285191893578, "global_step": 268784, "epoch": 3020, "lr": 2.3868290455486553e-08}
{"train_loss": 0.02858394756913185, "global_step": 268785, "epoch": 3020, "lr": 2.3850382826989813e-08}
{"train_loss": 0.025911232456564903, "global_step": 268786, "epoch": 3020, "lr": 2.3832481917179795e-08}
{"train_loss": 0.03161340206861496, "global_step": 268787, "epoch": 3020, "lr": 2.381458772605649e-08}
{"train_loss": 0.026322318241000175, "global_step": 268788, "epoch": 3020, "lr": 2.379670025362546e-08}
{"train_loss": 0.029229551553726196, "global_step": 268789, "epoch": 3020, "lr": 2.3778819499881145e-08}
{"train_loss": 0.024517318233847618, "global_step": 268790, "epoch": 3020, "lr": 2.376094546483465e-08}
{"train_loss": 0.023137126117944717, "global_step": 268791, "epoch": 3020, "lr": 2.3743078148480423e-08}
{"train_loss": 0.04097701236605644, "global_step": 268792, "epoch": 3020, "lr": 2.372521755082957e-08}
{"train_loss": 0.08269940316677094, "global_step": 268793, "epoch": 3020, "lr": 2.3707363671876538e-08}
{"train_loss": 0.05912807211279869, "global_step": 268794, "epoch": 3020, "lr": 2.3689516511626876e-08}
{"train_loss": 0.014932733029127121, "global_step": 268795, "epoch": 3020, "lr": 2.3671676070080586e-08}
{"train_loss": 0.026502827182412148, "global_step": 268796, "epoch": 3020, "lr": 2.3653842347248767e-08}
{"train_loss": 0.04652809351682663, "global_step": 268797, "epoch": 3020, "lr": 2.3636015343125872e-08}
{"train_loss": 0.038597457110881805, "global_step": 268798, "epoch": 3020, "lr": 2.36181950577119e-08}
{"train_loss": 0.012018272653222084, "global_step": 268799, "epoch": 3020, "lr": 2.3600381491017954e-08}
{"train_loss": 0.03306137025356293, "global_step": 268800, "epoch": 3020, "lr": 2.358257464303848e-08}
{"train_loss": 0.02301044575870037, "global_step": 268801, "epoch": 3020, "lr": 2.356477451377903e-08}
{"train_loss": 0.09199601411819458, "global_step": 268802, "epoch": 3020, "lr": 2.3546981103245157e-08}
{"train_loss": 0.05764199048280716, "global_step": 268803, "epoch": 3020, "lr": 2.352919441143131e-08}
{"train_loss": 0.0407729335129261, "global_step": 268804, "epoch": 3020, "lr": 2.351141443834859e-08}
{"train_loss": 0.0473516471683979, "global_step": 268805, "epoch": 3020, "lr": 2.3493641183996997e-08}
{"train_loss": 0.027440963312983513, "global_step": 268806, "epoch": 3020, "lr": 2.3475874648376528e-08}
{"train_loss": 0.055700305849313736, "global_step": 268807, "epoch": 3020, "lr": 2.3458114831487187e-08}
{"train_loss": 0.03459453210234642, "global_step": 268808, "epoch": 3020, "lr": 2.3440361733340078e-08}
{"train_loss": 0.06919824331998825, "global_step": 268809, "epoch": 3020, "lr": 2.3422615353929644e-08}
{"train_loss": 0.031893759965896606, "global_step": 268810, "epoch": 3020, "lr": 2.340487569326144e-08}
{"train_loss": 0.051520831882953644, "global_step": 268811, "epoch": 3020, "lr": 2.3387142751335468e-08}
{"train_loss": 0.02707745134830475, "global_step": 268812, "epoch": 3020, "lr": 2.3369416528157272e-08}
{"train_loss": 0.03105703555047512, "global_step": 268813, "epoch": 3020, "lr": 2.335169702372686e-08}
{"train_loss": 0.06282070279121399, "global_step": 268814, "epoch": 3020, "lr": 2.3333984238049776e-08}
{"train_loss": 0.07598565518856049, "global_step": 268815, "epoch": 3020, "lr": 2.3316278171126028e-08}
{"train_loss": 0.032054416835308075, "global_step": 268816, "epoch": 3020, "lr": 2.329857882295561e-08}
{"train_loss": 0.03520626202225685, "global_step": 268817, "epoch": 3020, "lr": 2.3280886193544072e-08}
{"train_loss": 0.039267636835575104, "global_step": 268818, "epoch": 3020, "lr": 2.326320028289697e-08}
{"train_loss": 0.01566467434167862, "global_step": 268819, "epoch": 3020, "lr": 2.3245521091008748e-08}
{"train_loss": 0.03589873015880585, "global_step": 268820, "epoch": 3020, "lr": 2.3227848617890513e-08}
{"train_loss": 0.03418877720832825, "global_step": 268821, "epoch": 3020, "lr": 2.3210182863536712e-08}
{"train_loss": 0.04511713981628418, "global_step": 268822, "epoch": 3020, "lr": 2.3192523827952893e-08}
{"train_loss": 0.06356614828109741, "global_step": 268823, "epoch": 3020, "lr": 2.3174871511144612e-08}
{"train_loss": 0.07298509776592255, "global_step": 268824, "epoch": 3020, "lr": 2.3157225913106317e-08}
{"train_loss": 0.0396389439702034, "global_step": 268825, "epoch": 3020, "lr": 2.313958703384911e-08}
{"train_loss": 0.04477889835834503, "global_step": 268826, "epoch": 3020, "lr": 2.3121954873372987e-08}
{"train_loss": 0.054185837507247925, "global_step": 268827, "epoch": 3020, "lr": 2.3104329431677952e-08}
{"train_loss": 0.05240655690431595, "global_step": 268828, "epoch": 3020, "lr": 2.3086710708764004e-08}
{"train_loss": 0.029219044372439384, "global_step": 268829, "epoch": 3020, "lr": 2.3069098704642243e-08}
{"train_loss": 0.0358310267329216, "global_step": 268830, "epoch": 3020, "lr": 2.3051493419307124e-08}
{"train_loss": 0.026810064911842346, "global_step": 268831, "epoch": 3020, "lr": 2.3033894852764193e-08}
{"train_loss": 0.03821491450071335, "global_step": 268832, "epoch": 3020, "lr": 2.3016303005013452e-08}
{"train_loss": 0.10507825762033463, "global_step": 268833, "epoch": 3020, "lr": 2.299871787606045e-08}
{"train_loss": 0.06577356159687042, "global_step": 268834, "epoch": 3020, "lr": 2.298113946590519e-08}
{"train_loss": 0.020181767642498016, "global_step": 268835, "epoch": 3020, "lr": 2.296356777455322e-08}
{"train_loss": 0.04347146302461624, "global_step": 268836, "epoch": 3020, "lr": 2.2946002801998988e-08}
{"train_loss": 0.04138973355293274, "global_step": 268837, "epoch": 3020, "lr": 2.2928444548259153e-08}
{"train_loss": 0.05601239576935768, "global_step": 268838, "epoch": 3020, "lr": 2.291089301332261e-08}
{"train_loss": 0.06281644850969315, "global_step": 268839, "epoch": 3020, "lr": 2.289334819719491e-08}
{"train_loss": 0.028791651129722595, "global_step": 268840, "epoch": 3020, "lr": 2.28758100998816e-08}
{"train_loss": 0.03613211214542389, "global_step": 268841, "epoch": 3020, "lr": 2.2858278721382688e-08}
{"train_loss": 0.05769459530711174, "global_step": 268842, "epoch": 3020, "lr": 2.2840754061703717e-08}
{"train_loss": 0.031150367110967636, "global_step": 268843, "epoch": 3020, "lr": 2.2823236120844693e-08}
{"train_loss": 0.05508000776171684, "global_step": 268844, "epoch": 3020, "lr": 2.2805724898805613e-08}
{"train_loss": 0.012086275033652782, "global_step": 268845, "epoch": 3020, "lr": 2.278822039559203e-08}
{"train_loss": 0.029417352750897408, "global_step": 268846, "epoch": 3020, "lr": 2.2770722611209495e-08}
{"train_loss": 0.045929402112960815, "global_step": 268847, "epoch": 3020, "lr": 2.2753231545652453e-08}
{"train_loss": 0.08018795400857925, "global_step": 268848, "epoch": 3020, "lr": 2.273574719892646e-08}
{"train_loss": 0.043641265481710434, "global_step": 268849, "epoch": 3020, "lr": 2.2718269571031515e-08}
{"train_loss": 0.06999241560697556, "global_step": 268850, "epoch": 3020, "lr": 2.2700798661978716e-08}
{"train_loss": 0.024079134687781334, "global_step": 268851, "epoch": 3020, "lr": 2.268333447176252e-08}
{"train_loss": 0.01791507378220558, "global_step": 268852, "epoch": 3020, "lr": 2.266587700038847e-08}
{"train_loss": 0.01823488436639309, "global_step": 268853, "epoch": 3020, "lr": 2.264842624785657e-08}
{"train_loss": 0.02391129359602928, "global_step": 268854, "epoch": 3020, "lr": 2.263098221417237e-08}
{"train_loss": 0.0500078946352005, "global_step": 268855, "epoch": 3020, "lr": 2.261354489933032e-08}
{"train_loss": 0.03764371573925018, "global_step": 268856, "epoch": 3020, "lr": 2.259611430334707e-08}
{"train_loss": 0.044913388788700104, "global_step": 268857, "epoch": 3020, "lr": 2.2578690426211524e-08}
{"train_loss": 0.0297696553170681, "global_step": 268858, "epoch": 3020, "lr": 2.2561273267929227e-08}
{"train_loss": 0.03959139063954353, "global_step": 268859, "epoch": 3020, "lr": 2.2543862828511287e-08}
{"train_loss": 0.05777287855744362, "global_step": 268860, "epoch": 3020, "lr": 2.2526459107946597e-08}
{"train_loss": 0.02861529029905796, "global_step": 268861, "epoch": 3020, "lr": 2.250906210624626e-08}
{"train_loss": 0.0400359109044075, "global_step": 268862, "epoch": 3020, "lr": 2.2491671823415827e-08}
{"train_loss": 0.04080519080162048, "global_step": 268863, "epoch": 3020, "lr": 2.24742882594442e-08}
{"train_loss": 0.03749372065067291, "global_step": 268864, "epoch": 3020, "lr": 2.2456911414348025e-08}
{"train_loss": 0.028822090476751328, "global_step": 268865, "epoch": 3020, "lr": 2.2439541288116206e-08}
{"train_loss": 0.019361747428774834, "global_step": 268866, "epoch": 3020, "lr": 2.2422177880765393e-08}
{"train_loss": 0.03710178658366203, "global_step": 268867, "epoch": 3020, "lr": 2.2404821192284486e-08}
{"train_loss": 0.04215881313207779, "global_step": 268868, "epoch": 3020, "lr": 2.238747122269014e-08, "val_loss": 9.65082836151123, "train_action_mse_error": 14.342124938964844}
{"train_loss": 0.0574435219168663, "global_step": 268869, "epoch": 3021, "lr": 2.2370127971971245e-08}
{"train_loss": 0.028851250186562538, "global_step": 268870, "epoch": 3021, "lr": 2.235279144013336e-08}
{"train_loss": 0.056146666407585144, "global_step": 268871, "epoch": 3021, "lr": 2.2335461627187583e-08}
{"train_loss": 0.05494486913084984, "global_step": 268872, "epoch": 3021, "lr": 2.231813853312281e-08}
{"train_loss": 0.06518509238958359, "global_step": 268873, "epoch": 3021, "lr": 2.2300822157950153e-08}
{"train_loss": 0.05437320098280907, "global_step": 268874, "epoch": 3021, "lr": 2.228351250167515e-08}
{"train_loss": 0.11838509142398834, "global_step": 268875, "epoch": 3021, "lr": 2.226620956429226e-08}
{"train_loss": 0.05943613499403, "global_step": 268876, "epoch": 3021, "lr": 2.224891334580148e-08}
{"train_loss": 0.046492308378219604, "global_step": 268877, "epoch": 3021, "lr": 2.2231623846213912e-08}
{"train_loss": 0.07571763545274734, "global_step": 268878, "epoch": 3021, "lr": 2.2214341065529555e-08}
{"train_loss": 0.04087917134165764, "global_step": 268879, "epoch": 3021, "lr": 2.2197065003748406e-08}
{"train_loss": 0.0541980117559433, "global_step": 268880, "epoch": 3021, "lr": 2.2179795660870473e-08}
{"train_loss": 0.08730100095272064, "global_step": 268881, "epoch": 3021, "lr": 2.216253303690685e-08}
{"train_loss": 0.04266129434108734, "global_step": 268882, "epoch": 3021, "lr": 2.2145277131851994e-08}
{"train_loss": 0.020462162792682648, "global_step": 268883, "epoch": 3021, "lr": 2.212802794571145e-08}
{"train_loss": 0.02345438301563263, "global_step": 268884, "epoch": 3021, "lr": 2.211078547848522e-08}
{"train_loss": 0.04813297092914581, "global_step": 268885, "epoch": 3021, "lr": 2.2093549730178854e-08}
{"train_loss": 0.029675547033548355, "global_step": 268886, "epoch": 3021, "lr": 2.20763207007868e-08}
{"train_loss": 0.020922118797898293, "global_step": 268887, "epoch": 3021, "lr": 2.2059098390325717e-08}
{"train_loss": 0.048602357506752014, "global_step": 268888, "epoch": 3021, "lr": 2.2041882798784497e-08}
{"train_loss": 0.05464410409331322, "global_step": 268889, "epoch": 3021, "lr": 2.2024673926174245e-08}
{"train_loss": 0.04151124134659767, "global_step": 268890, "epoch": 3021, "lr": 2.2007471772489407e-08}
{"train_loss": 0.028638450428843498, "global_step": 268891, "epoch": 3021, "lr": 2.1990276337735537e-08}
{"train_loss": 0.04181630164384842, "global_step": 268892, "epoch": 3021, "lr": 2.197308762192374e-08}
{"train_loss": 0.03799503296613693, "global_step": 268893, "epoch": 3021, "lr": 2.1955905625042904e-08}
{"train_loss": 0.0857635959982872, "global_step": 268894, "epoch": 3021, "lr": 2.193873034709859e-08}
{"train_loss": 0.04009192809462547, "global_step": 268895, "epoch": 3021, "lr": 2.1921561788101896e-08}
{"train_loss": 0.04061546549201012, "global_step": 268896, "epoch": 3021, "lr": 2.190439994804172e-08}
{"train_loss": 0.040438346564769745, "global_step": 268897, "epoch": 3021, "lr": 2.1887244826934716e-08}
{"train_loss": 0.0406462699174881, "global_step": 268898, "epoch": 3021, "lr": 2.1870096424769782e-08}
{"train_loss": 0.044901322573423386, "global_step": 268899, "epoch": 3021, "lr": 2.1852954741558018e-08}
{"train_loss": 0.019425367936491966, "global_step": 268900, "epoch": 3021, "lr": 2.1835819777299428e-08}
{"train_loss": 0.026363415643572807, "global_step": 268901, "epoch": 3021, "lr": 2.1818691531994007e-08}
{"train_loss": 0.02216632291674614, "global_step": 268902, "epoch": 3021, "lr": 2.180157000564731e-08}
{"train_loss": 0.03331286832690239, "global_step": 268903, "epoch": 3021, "lr": 2.1784455198259335e-08}
{"train_loss": 0.036482810974121094, "global_step": 268904, "epoch": 3021, "lr": 2.1767347109835633e-08}
{"train_loss": 0.031402282416820526, "global_step": 268905, "epoch": 3021, "lr": 2.1750245740370656e-08}
{"train_loss": 0.02997191622853279, "global_step": 268906, "epoch": 3021, "lr": 2.17331510898755e-08}
{"train_loss": 0.03816712647676468, "global_step": 268907, "epoch": 3021, "lr": 2.1716063158350177e-08}
{"train_loss": 0.040786582976579666, "global_step": 268908, "epoch": 3021, "lr": 2.1698981945800223e-08}
{"train_loss": 0.07578232139348984, "global_step": 268909, "epoch": 3021, "lr": 2.1681907452214544e-08}
{"train_loss": 0.06985021382570267, "global_step": 268910, "epoch": 3021, "lr": 2.1664839677609795e-08}
{"train_loss": 0.042993441224098206, "global_step": 268911, "epoch": 3021, "lr": 2.1647778621980418e-08}
{"train_loss": 0.05976187065243721, "global_step": 268912, "epoch": 3021, "lr": 2.1630724285337522e-08}
{"train_loss": 0.05922270566225052, "global_step": 268913, "epoch": 3021, "lr": 2.1613676667670002e-08}
{"train_loss": 0.034140679985284805, "global_step": 268914, "epoch": 3021, "lr": 2.159663576898896e-08}
{"train_loss": 0.04058251529932022, "global_step": 268915, "epoch": 3021, "lr": 2.1579601589299946e-08}
{"train_loss": 0.0340365506708622, "global_step": 268916, "epoch": 3021, "lr": 2.156257412859741e-08}
{"train_loss": 0.04211295396089554, "global_step": 268917, "epoch": 3021, "lr": 2.1545553386881355e-08}
{"train_loss": 0.03115100972354412, "global_step": 268918, "epoch": 3021, "lr": 2.152853936416843e-08}
{"train_loss": 0.04609905183315277, "global_step": 268919, "epoch": 3021, "lr": 2.1511532060447538e-08}
{"train_loss": 0.07402398437261581, "global_step": 268920, "epoch": 3021, "lr": 2.1494531475724224e-08}
{"train_loss": 0.016970796510577202, "global_step": 268921, "epoch": 3021, "lr": 2.147753761000404e-08}
{"train_loss": 0.01806352287530899, "global_step": 268922, "epoch": 3021, "lr": 2.1460550463281436e-08}
{"train_loss": 0.04061540588736534, "global_step": 268923, "epoch": 3021, "lr": 2.1443570035567518e-08}
{"train_loss": 0.03677797690033913, "global_step": 268924, "epoch": 3021, "lr": 2.142659632686228e-08}
{"train_loss": 0.025477580726146698, "global_step": 268925, "epoch": 3021, "lr": 2.1409629337171277e-08}
{"train_loss": 0.03248479962348938, "global_step": 268926, "epoch": 3021, "lr": 2.139266906648896e-08}
{"train_loss": 0.06325577199459076, "global_step": 268927, "epoch": 3021, "lr": 2.1375715514820872e-08}
{"train_loss": 0.041103608906269073, "global_step": 268928, "epoch": 3021, "lr": 2.135876868216702e-08}
{"train_loss": 0.022600790485739708, "global_step": 268929, "epoch": 3021, "lr": 2.1341828568532953e-08}
{"train_loss": 0.033321820199489594, "global_step": 268930, "epoch": 3021, "lr": 2.132489517392422e-08}
{"train_loss": 0.02035430260002613, "global_step": 268931, "epoch": 3021, "lr": 2.1307968498335272e-08}
{"train_loss": 0.051680658012628555, "global_step": 268932, "epoch": 3021, "lr": 2.1291048541777216e-08}
{"train_loss": 0.04420792683959007, "global_step": 268933, "epoch": 3021, "lr": 2.1274135304244493e-08}
{"train_loss": 0.017074864357709885, "global_step": 268934, "epoch": 3021, "lr": 2.1257228785742655e-08}
{"train_loss": 0.0483720563352108, "global_step": 268935, "epoch": 3021, "lr": 2.1240328986271706e-08}
{"train_loss": 0.05849417671561241, "global_step": 268936, "epoch": 3021, "lr": 2.1223435905837198e-08}
{"train_loss": 0.026766009628772736, "global_step": 268937, "epoch": 3021, "lr": 2.1206549544439127e-08}
{"train_loss": 0.023373669013381004, "global_step": 268938, "epoch": 3021, "lr": 2.118966990208304e-08}
{"train_loss": 0.033543411642313004, "global_step": 268939, "epoch": 3021, "lr": 2.117279697876895e-08}
{"train_loss": 0.021490411832928658, "global_step": 268940, "epoch": 3021, "lr": 2.1155930774496846e-08}
{"train_loss": 0.03469020500779152, "global_step": 268941, "epoch": 3021, "lr": 2.1139071289272284e-08}
{"train_loss": 0.034173931926488876, "global_step": 268942, "epoch": 3021, "lr": 2.112221852309526e-08}
{"train_loss": 0.04168262332677841, "global_step": 268943, "epoch": 3021, "lr": 2.110537247597133e-08}
{"train_loss": 0.05056547746062279, "global_step": 268944, "epoch": 3021, "lr": 2.108853314790049e-08}
{"train_loss": 0.04767395555973053, "global_step": 268945, "epoch": 3021, "lr": 2.1071700538882744e-08}
{"train_loss": 0.02463005669414997, "global_step": 268946, "epoch": 3021, "lr": 2.105487464892364e-08}
{"train_loss": 0.0515265092253685, "global_step": 268947, "epoch": 3021, "lr": 2.1038055478023177e-08}
{"train_loss": 0.02433180809020996, "global_step": 268948, "epoch": 3021, "lr": 2.102124302618691e-08}
{"train_loss": 0.03201070800423622, "global_step": 268949, "epoch": 3021, "lr": 2.1004437293414834e-08}
{"train_loss": 0.03146113082766533, "global_step": 268950, "epoch": 3021, "lr": 2.0987638279712507e-08}
{"train_loss": 0.06307818740606308, "global_step": 268951, "epoch": 3021, "lr": 2.0970845985074373e-08}
{"train_loss": 0.033516302704811096, "global_step": 268952, "epoch": 3021, "lr": 2.0954060409511534e-08}
{"train_loss": 0.01740426756441593, "global_step": 268953, "epoch": 3021, "lr": 2.093728155301844e-08}
{"train_loss": 0.022007562220096588, "global_step": 268954, "epoch": 3021, "lr": 2.0920509415606193e-08}
{"train_loss": 0.03671363741159439, "global_step": 268955, "epoch": 3021, "lr": 2.0903743997269243e-08}
{"train_loss": 0.02904171124100685, "global_step": 268956, "epoch": 3021, "lr": 2.088698529801314e-08}
{"train_loss": 0.04187165110717329, "global_step": 268957, "epoch": 3021, "lr": 2.0870233317837883e-08, "val_loss": 9.70409107208252}
{"train_loss": 0.03803500905632973, "global_step": 268958, "epoch": 3022, "lr": 2.0853488056749025e-08}
{"train_loss": 0.041120342910289764, "global_step": 268959, "epoch": 3022, "lr": 2.0836749514746567e-08}
{"train_loss": 0.060707271099090576, "global_step": 268960, "epoch": 3022, "lr": 2.0820017691830507e-08}
{"train_loss": 0.028940191492438316, "global_step": 268961, "epoch": 3022, "lr": 2.0803292588011948e-08}
{"train_loss": 0.04309166595339775, "global_step": 268962, "epoch": 3022, "lr": 2.0786574203285336e-08}
{"train_loss": 0.08452590554952621, "global_step": 268963, "epoch": 3022, "lr": 2.0769862537650675e-08}
{"train_loss": 0.03727976977825165, "global_step": 268964, "epoch": 3022, "lr": 2.0753157591119066e-08}
{"train_loss": 0.02808760665357113, "global_step": 268965, "epoch": 3022, "lr": 2.0736459363684957e-08}
{"train_loss": 0.05926733464002609, "global_step": 268966, "epoch": 3022, "lr": 2.07197678553539e-08}
{"train_loss": 0.028804052621126175, "global_step": 268967, "epoch": 3022, "lr": 2.0703083066131446e-08}
{"train_loss": 0.07287626713514328, "global_step": 268968, "epoch": 3022, "lr": 2.0686404996012044e-08}
{"train_loss": 0.03479105234146118, "global_step": 268969, "epoch": 3022, "lr": 2.0669733645006795e-08}
{"train_loss": 0.04362073913216591, "global_step": 268970, "epoch": 3022, "lr": 2.065306901311015e-08}
{"train_loss": 0.10321700572967529, "global_step": 268971, "epoch": 3022, "lr": 2.0636411100327657e-08}
{"train_loss": 0.04498886317014694, "global_step": 268972, "epoch": 3022, "lr": 2.061975990665932e-08}
{"train_loss": 0.042307160794734955, "global_step": 268973, "epoch": 3022, "lr": 2.0603115432116237e-08}
{"train_loss": 0.03744620829820633, "global_step": 268974, "epoch": 3022, "lr": 2.058647767668731e-08}
{"train_loss": 0.07299315929412842, "global_step": 268975, "epoch": 3022, "lr": 2.0569846640383638e-08}
{"train_loss": 0.071803517639637, "global_step": 268976, "epoch": 3022, "lr": 2.0553222323205224e-08}
{"train_loss": 0.016531579196453094, "global_step": 268977, "epoch": 3022, "lr": 2.0536604725152063e-08}
{"train_loss": 0.040061768144369125, "global_step": 268978, "epoch": 3022, "lr": 2.0519993846235263e-08}
{"train_loss": 0.028815321624279022, "global_step": 268979, "epoch": 3022, "lr": 2.050338968644372e-08}
{"train_loss": 0.05173882097005844, "global_step": 268980, "epoch": 3022, "lr": 2.0486792245788533e-08}
{"train_loss": 0.02271134965121746, "global_step": 268981, "epoch": 3022, "lr": 2.0470201524269705e-08}
{"train_loss": 0.04629635065793991, "global_step": 268982, "epoch": 3022, "lr": 2.0453617521887237e-08}
{"train_loss": 0.0379762165248394, "global_step": 268983, "epoch": 3022, "lr": 2.0437040238652227e-08}
{"train_loss": 0.04922480508685112, "global_step": 268984, "epoch": 3022, "lr": 2.0420469674553578e-08}
{"train_loss": 0.025426045060157776, "global_step": 268985, "epoch": 3022, "lr": 2.0403905829602387e-08}
{"train_loss": 0.03802480548620224, "global_step": 268986, "epoch": 3022, "lr": 2.0387348703798658e-08}
{"train_loss": 0.0276788379997015, "global_step": 268987, "epoch": 3022, "lr": 2.0370798297147942e-08}
{"train_loss": 0.01095561869442463, "global_step": 268988, "epoch": 3022, "lr": 2.0354254609644684e-08}
{"train_loss": 0.04229283332824707, "global_step": 268989, "epoch": 3022, "lr": 2.0337717641299992e-08}
{"train_loss": 0.022073011845350266, "global_step": 268990, "epoch": 3022, "lr": 2.032118739210831e-08}
{"train_loss": 0.04440614581108093, "global_step": 268991, "epoch": 3022, "lr": 2.030466386207519e-08}
{"train_loss": 0.09589958935976028, "global_step": 268992, "epoch": 3022, "lr": 2.0288147051200634e-08}
{"train_loss": 0.030195681378245354, "global_step": 268993, "epoch": 3022, "lr": 2.0271636959495743e-08}
{"train_loss": 0.0360853374004364, "global_step": 268994, "epoch": 3022, "lr": 2.0255133586949414e-08}
{"train_loss": 0.052063338458538055, "global_step": 268995, "epoch": 3022, "lr": 2.0238636933578304e-08}
{"train_loss": 0.06989090889692307, "global_step": 268996, "epoch": 3022, "lr": 2.0222146999371304e-08}
{"train_loss": 0.05279573053121567, "global_step": 268997, "epoch": 3022, "lr": 2.020566378433397e-08}
{"train_loss": 0.06813826411962509, "global_step": 268998, "epoch": 3022, "lr": 2.0189187288477405e-08}
{"train_loss": 0.04649541899561882, "global_step": 268999, "epoch": 3022, "lr": 2.0172717511796058e-08}
{"train_loss": 0.06035088747739792, "global_step": 269000, "epoch": 3022, "lr": 2.0156254454289925e-08}
{"train_loss": 0.07473310828208923, "global_step": 269001, "epoch": 3022, "lr": 2.013979811596456e-08}
{"train_loss": 0.054083745926618576, "global_step": 269002, "epoch": 3022, "lr": 2.012334849682551e-08}
{"train_loss": 0.020350931212306023, "global_step": 269003, "epoch": 3022, "lr": 2.010690559686723e-08}
{"train_loss": 0.042065393179655075, "global_step": 269004, "epoch": 3022, "lr": 2.009046941610082e-08}
{"train_loss": 0.020291097462177277, "global_step": 269005, "epoch": 3022, "lr": 2.0074039954520728e-08}
{"train_loss": 0.04066765308380127, "global_step": 269006, "epoch": 3022, "lr": 2.005761721213806e-08}
{"train_loss": 0.021793900057673454, "global_step": 269007, "epoch": 3022, "lr": 2.0041201188941705e-08}
{"train_loss": 0.05568819120526314, "global_step": 269008, "epoch": 3022, "lr": 2.0024791884948324e-08}
{"train_loss": 0.01525947917252779, "global_step": 269009, "epoch": 3022, "lr": 2.0008389300146814e-08}
{"train_loss": 0.029497617855668068, "global_step": 269010, "epoch": 3022, "lr": 1.9991993434553825e-08}
{"train_loss": 0.034328266978263855, "global_step": 269011, "epoch": 3022, "lr": 1.9975604288158255e-08}
{"train_loss": 0.03233402594923973, "global_step": 269012, "epoch": 3022, "lr": 1.995922186097121e-08}
{"train_loss": 0.06938838958740234, "global_step": 269013, "epoch": 3022, "lr": 1.994284615298714e-08}
{"train_loss": 0.056607022881507874, "global_step": 269014, "epoch": 3022, "lr": 1.992647716421714e-08}
{"train_loss": 0.03405213728547096, "global_step": 269015, "epoch": 3022, "lr": 1.9910114894655662e-08}
{"train_loss": 0.04820716381072998, "global_step": 269016, "epoch": 3022, "lr": 1.989375934430826e-08}
{"train_loss": 0.07698909193277359, "global_step": 269017, "epoch": 3022, "lr": 1.9877410513174933e-08}
{"train_loss": 0.12339179217815399, "global_step": 269018, "epoch": 3022, "lr": 1.986106840126678e-08}
{"train_loss": 0.04736586660146713, "global_step": 269019, "epoch": 3022, "lr": 1.9844733008572702e-08}
{"train_loss": 0.039946019649505615, "global_step": 269020, "epoch": 3022, "lr": 1.98284043351038e-08}
{"train_loss": 0.06129766255617142, "global_step": 269021, "epoch": 3022, "lr": 1.9812082380860075e-08}
{"train_loss": 0.052908264100551605, "global_step": 269022, "epoch": 3022, "lr": 1.9795767145841527e-08}
{"train_loss": 0.0360361784696579, "global_step": 269023, "epoch": 3022, "lr": 1.9779458630059256e-08}
{"train_loss": 0.03839758411049843, "global_step": 269024, "epoch": 3022, "lr": 1.976315683350216e-08}
{"train_loss": 0.05283186584711075, "global_step": 269025, "epoch": 3022, "lr": 1.9746861756175793e-08}
{"train_loss": 0.033412519842386246, "global_step": 269026, "epoch": 3022, "lr": 1.9730573398091257e-08}
{"train_loss": 0.025366462767124176, "global_step": 269027, "epoch": 3022, "lr": 1.9714291759243e-08}
{"train_loss": 0.013423740863800049, "global_step": 269028, "epoch": 3022, "lr": 1.9698016839636567e-08}
{"train_loss": 0.02378225326538086, "global_step": 269029, "epoch": 3022, "lr": 1.9681748639271968e-08}
{"train_loss": 0.05118003860116005, "global_step": 269030, "epoch": 3022, "lr": 1.9665487158149198e-08}
{"train_loss": 0.05691239982843399, "global_step": 269031, "epoch": 3022, "lr": 1.964923239627381e-08}
{"train_loss": 0.03453600034117699, "global_step": 269032, "epoch": 3022, "lr": 1.96329843536458e-08}
{"train_loss": 0.07161019742488861, "global_step": 269033, "epoch": 3022, "lr": 1.9616743030276273e-08}
{"train_loss": 0.04415284842252731, "global_step": 269034, "epoch": 3022, "lr": 1.960050842615413e-08}
{"train_loss": 0.033528465777635574, "global_step": 269035, "epoch": 3022, "lr": 1.9584280541284916e-08}
{"train_loss": 0.02679119072854519, "global_step": 269036, "epoch": 3022, "lr": 1.9568059375679738e-08}
{"train_loss": 0.05288470536470413, "global_step": 269037, "epoch": 3022, "lr": 1.955184492933304e-08}
{"train_loss": 0.03314674645662308, "global_step": 269038, "epoch": 3022, "lr": 1.9535637202244826e-08}
{"train_loss": 0.07681272178888321, "global_step": 269039, "epoch": 3022, "lr": 1.95194361944262e-08}
{"train_loss": 0.07505317777395248, "global_step": 269040, "epoch": 3022, "lr": 1.9503241905871605e-08}
{"train_loss": 0.06603153795003891, "global_step": 269041, "epoch": 3022, "lr": 1.9487054336581047e-08}
{"train_loss": 0.031498175114393234, "global_step": 269042, "epoch": 3022, "lr": 1.947087348656562e-08}
{"train_loss": 0.03514780104160309, "global_step": 269043, "epoch": 3022, "lr": 1.9454699355825334e-08}
{"train_loss": 0.0479460284113884, "global_step": 269044, "epoch": 3022, "lr": 1.9438531944354633e-08}
{"train_loss": 0.04176943749189377, "global_step": 269045, "epoch": 3022, "lr": 1.9422371252164616e-08}
{"train_loss": 0.046179679599119706, "global_step": 269046, "epoch": 3022, "lr": 1.940621727925529e-08, "val_loss": 9.681600570678711}
{"train_loss": 0.03706379607319832, "global_step": 269047, "epoch": 3023, "lr": 1.939007002562665e-08}
{"train_loss": 0.0737120732665062, "global_step": 269048, "epoch": 3023, "lr": 1.9373929491284247e-08}
{"train_loss": 0.06769877672195435, "global_step": 269049, "epoch": 3023, "lr": 1.9357795676222534e-08}
{"train_loss": 0.05164629966020584, "global_step": 269050, "epoch": 3023, "lr": 1.934166858045261e-08}
{"train_loss": 0.0549551360309124, "global_step": 269051, "epoch": 3023, "lr": 1.9325548203968924e-08}
{"train_loss": 0.09185360372066498, "global_step": 269052, "epoch": 3023, "lr": 1.9309434546782578e-08}
{"train_loss": 0.05068107321858406, "global_step": 269053, "epoch": 3023, "lr": 1.9293327608888024e-08}
{"train_loss": 0.037770312279462814, "global_step": 269054, "epoch": 3023, "lr": 1.927722739029081e-08}
{"train_loss": 0.049529753625392914, "global_step": 269055, "epoch": 3023, "lr": 1.9261133890996484e-08}
{"train_loss": 0.0937776118516922, "global_step": 269056, "epoch": 3023, "lr": 1.9245047110999502e-08}
{"train_loss": 0.03581991046667099, "global_step": 269057, "epoch": 3023, "lr": 1.922896705030541e-08}
{"train_loss": 0.03886345028877258, "global_step": 269058, "epoch": 3023, "lr": 1.9212893708914213e-08}
{"train_loss": 0.018046384677290916, "global_step": 269059, "epoch": 3023, "lr": 1.919682708683701e-08}
{"train_loss": 0.06661826372146606, "global_step": 269060, "epoch": 3023, "lr": 1.9180767184062698e-08}
{"train_loss": 0.03950326517224312, "global_step": 269061, "epoch": 3023, "lr": 1.916471400060793e-08}
{"train_loss": 0.02372397482395172, "global_step": 269062, "epoch": 3023, "lr": 1.9148667536461605e-08}
{"train_loss": 0.04906178265810013, "global_step": 269063, "epoch": 3023, "lr": 1.913262779163483e-08}
{"train_loss": 0.04151427373290062, "global_step": 269064, "epoch": 3023, "lr": 1.9116594766127594e-08}
{"train_loss": 0.024544304236769676, "global_step": 269065, "epoch": 3023, "lr": 1.9100568459934352e-08}
{"train_loss": 0.05836345627903938, "global_step": 269066, "epoch": 3023, "lr": 1.9084548873071762e-08}
{"train_loss": 0.031038055196404457, "global_step": 269067, "epoch": 3023, "lr": 1.9068536005528714e-08}
{"train_loss": 0.030342530459165573, "global_step": 269068, "epoch": 3023, "lr": 1.9052529857316315e-08}
{"train_loss": 0.04451500251889229, "global_step": 269069, "epoch": 3023, "lr": 1.9036530428429013e-08}
{"train_loss": 0.03217324614524841, "global_step": 269070, "epoch": 3023, "lr": 1.9020537718877908e-08}
{"train_loss": 0.042727600783109665, "global_step": 269071, "epoch": 3023, "lr": 1.900455172865745e-08}
{"train_loss": 0.011674197390675545, "global_step": 269072, "epoch": 3023, "lr": 1.898857245777319e-08}
{"train_loss": 0.041558656841516495, "global_step": 269073, "epoch": 3023, "lr": 1.897259990622513e-08}
{"train_loss": 0.0692124217748642, "global_step": 269074, "epoch": 3023, "lr": 1.895663407401882e-08}
{"train_loss": 0.014073225669562817, "global_step": 269075, "epoch": 3023, "lr": 1.8940674961154258e-08}
{"train_loss": 0.05959448963403702, "global_step": 269076, "epoch": 3023, "lr": 1.8924722567631447e-08}
{"train_loss": 0.036453261971473694, "global_step": 269077, "epoch": 3023, "lr": 1.890877689346149e-08}
{"train_loss": 0.012128254398703575, "global_step": 269078, "epoch": 3023, "lr": 1.8892837938633277e-08}
{"train_loss": 0.016377225518226624, "global_step": 269079, "epoch": 3023, "lr": 1.887690570315792e-08}
{"train_loss": 0.03524726256728172, "global_step": 269080, "epoch": 3023, "lr": 1.8860980187035416e-08}
{"train_loss": 0.05856488272547722, "global_step": 269081, "epoch": 3023, "lr": 1.884506139027131e-08}
{"train_loss": 0.07168383151292801, "global_step": 269082, "epoch": 3023, "lr": 1.882914931286006e-08}
{"train_loss": 0.06030390411615372, "global_step": 269083, "epoch": 3023, "lr": 1.8813243954807215e-08}
{"train_loss": 0.02936064824461937, "global_step": 269084, "epoch": 3023, "lr": 1.879734531611832e-08}
{"train_loss": 0.026348205283284187, "global_step": 269085, "epoch": 3023, "lr": 1.878145339679338e-08}
{"train_loss": 0.08665114641189575, "global_step": 269086, "epoch": 3023, "lr": 1.8765568196832393e-08}
{"train_loss": 0.04240541160106659, "global_step": 269087, "epoch": 3023, "lr": 1.8749689716240915e-08}
{"train_loss": 0.028898118063807487, "global_step": 269088, "epoch": 3023, "lr": 1.873381795501894e-08}
{"train_loss": 0.08863981068134308, "global_step": 269089, "epoch": 3023, "lr": 1.871795291316647e-08}
{"train_loss": 0.008028371259570122, "global_step": 269090, "epoch": 3023, "lr": 1.8702094590694607e-08}
{"train_loss": 0.05487813800573349, "global_step": 269091, "epoch": 3023, "lr": 1.8686242987592252e-08}
{"train_loss": 0.018844980746507645, "global_step": 269092, "epoch": 3023, "lr": 1.86703981038705e-08}
{"train_loss": 0.05237504839897156, "global_step": 269093, "epoch": 3023, "lr": 1.8654559939529358e-08}
{"train_loss": 0.04535472393035889, "global_step": 269094, "epoch": 3023, "lr": 1.8638728494568825e-08}
{"train_loss": 0.03975139930844307, "global_step": 269095, "epoch": 3023, "lr": 1.8622903769e-08}
{"train_loss": 0.03835447132587433, "global_step": 269096, "epoch": 3023, "lr": 1.8607085762811783e-08}
{"train_loss": 0.05195213109254837, "global_step": 269097, "epoch": 3023, "lr": 1.8591274476015273e-08}
{"train_loss": 0.053806766867637634, "global_step": 269098, "epoch": 3023, "lr": 1.8575469908604927e-08}
{"train_loss": 0.013441763818264008, "global_step": 269099, "epoch": 3023, "lr": 1.855967206059739e-08}
{"train_loss": 0.014304086565971375, "global_step": 269100, "epoch": 3023, "lr": 1.854388093197601e-08}
{"train_loss": 0.059263166040182114, "global_step": 269101, "epoch": 3023, "lr": 1.8528096522757444e-08}
{"train_loss": 0.02284703031182289, "global_step": 269102, "epoch": 3023, "lr": 1.8512318832936138e-08}
{"train_loss": 0.037604495882987976, "global_step": 269103, "epoch": 3023, "lr": 1.8496547862517644e-08}
{"train_loss": 0.03787267208099365, "global_step": 269104, "epoch": 3023, "lr": 1.8480783611507514e-08}
{"train_loss": 0.034826185554265976, "global_step": 269105, "epoch": 3023, "lr": 1.8465026079900193e-08}
{"train_loss": 0.03799774870276451, "global_step": 269106, "epoch": 3023, "lr": 1.8449275267695687e-08}
{"train_loss": 0.08861170709133148, "global_step": 269107, "epoch": 3023, "lr": 1.8433531174910647e-08}
{"train_loss": 0.06558703631162643, "global_step": 269108, "epoch": 3023, "lr": 1.8417793801528415e-08}
{"train_loss": 0.04250704497098923, "global_step": 269109, "epoch": 3023, "lr": 1.840206314756565e-08}
{"train_loss": 0.052004534751176834, "global_step": 269110, "epoch": 3023, "lr": 1.8386339213022353e-08}
{"train_loss": 0.06868652254343033, "global_step": 269111, "epoch": 3023, "lr": 1.8370621997892967e-08}
{"train_loss": 0.03083236888051033, "global_step": 269112, "epoch": 3023, "lr": 1.83549115021886e-08}
{"train_loss": 0.06928793340921402, "global_step": 269113, "epoch": 3023, "lr": 1.8339207725903695e-08}
{"train_loss": 0.010288111865520477, "global_step": 269114, "epoch": 3023, "lr": 1.832351066904936e-08}
{"train_loss": 0.017840776592493057, "global_step": 269115, "epoch": 3023, "lr": 1.830782033161449e-08}
{"train_loss": 0.06591566652059555, "global_step": 269116, "epoch": 3023, "lr": 1.829213671361574e-08}
{"train_loss": 0.030038287863135338, "global_step": 269117, "epoch": 3023, "lr": 1.8276459815042003e-08}
{"train_loss": 0.03296886757016182, "global_step": 269118, "epoch": 3023, "lr": 1.8260789635904386e-08}
{"train_loss": 0.05701782926917076, "global_step": 269119, "epoch": 3023, "lr": 1.8245126176202887e-08}
{"train_loss": 0.070830799639225, "global_step": 269120, "epoch": 3023, "lr": 1.8229469435943057e-08}
{"train_loss": 0.024638675153255463, "global_step": 269121, "epoch": 3023, "lr": 1.821381941511935e-08}
{"train_loss": 0.08716439455747604, "global_step": 269122, "epoch": 3023, "lr": 1.819817611373731e-08}
{"train_loss": 0.04163532331585884, "global_step": 269123, "epoch": 3023, "lr": 1.8182539531796937e-08}
{"train_loss": 0.038671184331178665, "global_step": 269124, "epoch": 3023, "lr": 1.816690966930934e-08}
{"train_loss": 0.01828305423259735, "global_step": 269125, "epoch": 3023, "lr": 1.815128652626341e-08}
{"train_loss": 0.03231217712163925, "global_step": 269126, "epoch": 3023, "lr": 1.8135670102670255e-08}
{"train_loss": 0.06177792325615883, "global_step": 269127, "epoch": 3023, "lr": 1.812006039852987e-08}
{"train_loss": 0.029750198125839233, "global_step": 269128, "epoch": 3023, "lr": 1.810445741384781e-08}
{"train_loss": 0.03499047830700874, "global_step": 269129, "epoch": 3023, "lr": 1.8088861148618518e-08}
{"train_loss": 0.06876169145107269, "global_step": 269130, "epoch": 3023, "lr": 1.8073271602847552e-08}
{"train_loss": 0.04151090979576111, "global_step": 269131, "epoch": 3023, "lr": 1.805768877653491e-08}
{"train_loss": 0.015531167387962341, "global_step": 269132, "epoch": 3023, "lr": 1.804211266969169e-08}
{"train_loss": 0.02395482361316681, "global_step": 269133, "epoch": 3023, "lr": 1.8026543282306797e-08}
{"train_loss": 0.049443319439888, "global_step": 269134, "epoch": 3023, "lr": 1.801098061439688e-08}
{"train_loss": 0.043857124257372336, "global_step": 269135, "epoch": 3023, "lr": 1.7995424665950832e-08, "val_loss": 9.640426635742188}
{"train_loss": 0.024453816935420036, "global_step": 269136, "epoch": 3024, "lr": 1.7979875436974215e-08}
{"train_loss": 0.07204689830541611, "global_step": 269137, "epoch": 3024, "lr": 1.7964332927478123e-08}
{"train_loss": 0.03966904059052467, "global_step": 269138, "epoch": 3024, "lr": 1.7948797137451457e-08}
{"train_loss": 0.03968173265457153, "global_step": 269139, "epoch": 3024, "lr": 1.7933268066905318e-08}
{"train_loss": 0.060688566416502, "global_step": 269140, "epoch": 3024, "lr": 1.7917745715834155e-08}
{"train_loss": 0.03650045767426491, "global_step": 269141, "epoch": 3024, "lr": 1.790223008424907e-08}
{"train_loss": 0.03467855602502823, "global_step": 269142, "epoch": 3024, "lr": 1.7886721172150067e-08}
{"train_loss": 0.059709250926971436, "global_step": 269143, "epoch": 3024, "lr": 1.787121897953159e-08}
{"train_loss": 0.047994840890169144, "global_step": 269144, "epoch": 3024, "lr": 1.785572350640474e-08}
{"train_loss": 0.026915093883872032, "global_step": 269145, "epoch": 3024, "lr": 1.7840234752763972e-08}
{"train_loss": 0.05228348821401596, "global_step": 269146, "epoch": 3024, "lr": 1.7824752718620386e-08}
{"train_loss": 0.032616134732961655, "global_step": 269147, "epoch": 3024, "lr": 1.780927740396843e-08}
{"train_loss": 0.04109877720475197, "global_step": 269148, "epoch": 3024, "lr": 1.779380880881365e-08}
{"train_loss": 0.042646944522857666, "global_step": 269149, "epoch": 3024, "lr": 1.7778346933156054e-08}
{"train_loss": 0.03700549528002739, "global_step": 269150, "epoch": 3024, "lr": 1.776289177700119e-08}
{"train_loss": 0.08111906796693802, "global_step": 269151, "epoch": 3024, "lr": 1.7747443340349057e-08}
{"train_loss": 0.04324116185307503, "global_step": 269152, "epoch": 3024, "lr": 1.7732001623199657e-08}
{"train_loss": 0.033652257174253464, "global_step": 269153, "epoch": 3024, "lr": 1.771656662555854e-08}
{"train_loss": 0.07337328046560287, "global_step": 269154, "epoch": 3024, "lr": 1.7701138347420154e-08}
{"train_loss": 0.03867549076676369, "global_step": 269155, "epoch": 3024, "lr": 1.7685716788801155e-08}
{"train_loss": 0.018487298861145973, "global_step": 269156, "epoch": 3024, "lr": 1.7670301949690437e-08}
{"train_loss": 0.03612827509641647, "global_step": 269157, "epoch": 3024, "lr": 1.7654893830093556e-08}
{"train_loss": 0.0684308111667633, "global_step": 269158, "epoch": 3024, "lr": 1.7639492430016057e-08}
{"train_loss": 0.06562154740095139, "global_step": 269159, "epoch": 3024, "lr": 1.7624097749457945e-08}
{"train_loss": 0.02030741237103939, "global_step": 269160, "epoch": 3024, "lr": 1.760870978842477e-08}
{"train_loss": 0.06426279991865158, "global_step": 269161, "epoch": 3024, "lr": 1.7593328546910982e-08}
{"train_loss": 0.07522939890623093, "global_step": 269162, "epoch": 3024, "lr": 1.7577954024922127e-08}
{"train_loss": 0.023419272154569626, "global_step": 269163, "epoch": 3024, "lr": 1.7562586222458212e-08}
{"train_loss": 0.016586024314165115, "global_step": 269164, "epoch": 3024, "lr": 1.7547225139530332e-08}
{"train_loss": 0.07844775170087814, "global_step": 269165, "epoch": 3024, "lr": 1.7531870776127392e-08}
{"train_loss": 0.04537301883101463, "global_step": 269166, "epoch": 3024, "lr": 1.7516523132260486e-08}
{"train_loss": 0.03688163682818413, "global_step": 269167, "epoch": 3024, "lr": 1.7501182207929622e-08}
{"train_loss": 0.023226890712976456, "global_step": 269168, "epoch": 3024, "lr": 1.7485848003140348e-08}
{"train_loss": 0.03907567635178566, "global_step": 269169, "epoch": 3024, "lr": 1.747052051788711e-08}
{"train_loss": 0.021864432841539383, "global_step": 269170, "epoch": 3024, "lr": 1.7455199752175465e-08}
{"train_loss": 0.04440508410334587, "global_step": 269171, "epoch": 3024, "lr": 1.743988570601096e-08}
{"train_loss": 0.03113521821796894, "global_step": 269172, "epoch": 3024, "lr": 1.7424578379388046e-08}
{"train_loss": 0.04859670251607895, "global_step": 269173, "epoch": 3024, "lr": 1.7409277772312273e-08}
{"train_loss": 0.053130924701690674, "global_step": 269174, "epoch": 3024, "lr": 1.7393983884789187e-08}
{"train_loss": 0.04288960248231888, "global_step": 269175, "epoch": 3024, "lr": 1.73786967168188e-08}
{"train_loss": 0.07139196246862411, "global_step": 269176, "epoch": 3024, "lr": 1.7363416268401102e-08}
{"train_loss": 0.0392974391579628, "global_step": 269177, "epoch": 3024, "lr": 1.7348142539541644e-08}
{"train_loss": 0.03344111889600754, "global_step": 269178, "epoch": 3024, "lr": 1.7332875530240435e-08}
{"train_loss": 0.03552723303437233, "global_step": 269179, "epoch": 3024, "lr": 1.7317615240503014e-08}
{"train_loss": 0.04118150845170021, "global_step": 269180, "epoch": 3024, "lr": 1.730236167032384e-08}
{"train_loss": 0.052136849611997604, "global_step": 269181, "epoch": 3024, "lr": 1.728711481970846e-08}
{"train_loss": 0.0469229482114315, "global_step": 269182, "epoch": 3024, "lr": 1.7271874688656874e-08}
{"train_loss": 0.03397101163864136, "global_step": 269183, "epoch": 3024, "lr": 1.7256641277180185e-08}
{"train_loss": 0.05298449099063873, "global_step": 269184, "epoch": 3024, "lr": 1.7241414585272842e-08}
{"train_loss": 0.033705927431583405, "global_step": 269185, "epoch": 3024, "lr": 1.7226194612934842e-08}
{"train_loss": 0.0703234076499939, "global_step": 269186, "epoch": 3024, "lr": 1.721098136017174e-08}
{"train_loss": 0.0496111735701561, "global_step": 269187, "epoch": 3024, "lr": 1.7195774826989085e-08}
{"train_loss": 0.06608305871486664, "global_step": 269188, "epoch": 3024, "lr": 1.7180575013381328e-08}
{"train_loss": 0.0725293904542923, "global_step": 269189, "epoch": 3024, "lr": 1.716538191935957e-08}
{"train_loss": 0.01620216853916645, "global_step": 269190, "epoch": 3024, "lr": 1.7150195544918256e-08}
{"train_loss": 0.042844876646995544, "global_step": 269191, "epoch": 3024, "lr": 1.7135015890057393e-08}
{"train_loss": 0.0558607392013073, "global_step": 269192, "epoch": 3024, "lr": 1.7119842954788078e-08}
{"train_loss": 0.0647856667637825, "global_step": 269193, "epoch": 3024, "lr": 1.7104676739110315e-08}
{"train_loss": 0.08150387555360794, "global_step": 269194, "epoch": 3024, "lr": 1.7089517243018548e-08}
{"train_loss": 0.028178874403238297, "global_step": 269195, "epoch": 3024, "lr": 1.7074364466523885e-08}
{"train_loss": 0.039171669632196426, "global_step": 269196, "epoch": 3024, "lr": 1.705921840962077e-08}
{"train_loss": 0.06859786808490753, "global_step": 269197, "epoch": 3024, "lr": 1.7044079072314755e-08}
{"train_loss": 0.0188737940043211, "global_step": 269198, "epoch": 3024, "lr": 1.7028946454611393e-08}
{"train_loss": 0.03595174849033356, "global_step": 269199, "epoch": 3024, "lr": 1.7013820556505135e-08}
{"train_loss": 0.03165388107299805, "global_step": 269200, "epoch": 3024, "lr": 1.6998701378007077e-08}
{"train_loss": 0.05450301244854927, "global_step": 269201, "epoch": 3024, "lr": 1.6983588919111672e-08}
{"train_loss": 0.037640951573848724, "global_step": 269202, "epoch": 3024, "lr": 1.6968483179824468e-08}
{"train_loss": 0.028267627581954002, "global_step": 269203, "epoch": 3024, "lr": 1.695338416014547e-08}
{"train_loss": 0.03349025547504425, "global_step": 269204, "epoch": 3024, "lr": 1.6938291860074673e-08}
{"train_loss": 0.022109435871243477, "global_step": 269205, "epoch": 3024, "lr": 1.6923206279623183e-08}
{"train_loss": 0.02358434721827507, "global_step": 269206, "epoch": 3024, "lr": 1.6908127418785446e-08}
{"train_loss": 0.04774407669901848, "global_step": 269207, "epoch": 3024, "lr": 1.6893055277561466e-08}
{"train_loss": 0.01264894288033247, "global_step": 269208, "epoch": 3024, "lr": 1.687798985596234e-08}
{"train_loss": 0.06621064990758896, "global_step": 269209, "epoch": 3024, "lr": 1.686293115398252e-08}
{"train_loss": 0.07589253783226013, "global_step": 269210, "epoch": 3024, "lr": 1.6847879171627557e-08}
{"train_loss": 0.05361693724989891, "global_step": 269211, "epoch": 3024, "lr": 1.683283390889745e-08}
{"train_loss": 0.04047301039099693, "global_step": 269212, "epoch": 3024, "lr": 1.68177953657922e-08}
{"train_loss": 0.021927252411842346, "global_step": 269213, "epoch": 3024, "lr": 1.6802763542317356e-08}
{"train_loss": 0.06772523373365402, "global_step": 269214, "epoch": 3024, "lr": 1.6787738438478474e-08}
{"train_loss": 0.034207284450531006, "global_step": 269215, "epoch": 3024, "lr": 1.6772720054269997e-08}
{"train_loss": 0.042383842170238495, "global_step": 269216, "epoch": 3024, "lr": 1.675770838969748e-08}
{"train_loss": 0.031610533595085144, "global_step": 269217, "epoch": 3024, "lr": 1.674270344476092e-08}
{"train_loss": 0.06605184078216553, "global_step": 269218, "epoch": 3024, "lr": 1.6727705219465874e-08}
{"train_loss": 0.03480195254087448, "global_step": 269219, "epoch": 3024, "lr": 1.6712713713812333e-08}
{"train_loss": 0.044977687299251556, "global_step": 269220, "epoch": 3024, "lr": 1.6697728927805855e-08}
{"train_loss": 0.03607232868671417, "global_step": 269221, "epoch": 3024, "lr": 1.668275086144089e-08}
{"train_loss": 0.0495082251727581, "global_step": 269222, "epoch": 3024, "lr": 1.666777951472298e-08}
{"train_loss": 0.05553135275840759, "global_step": 269223, "epoch": 3024, "lr": 1.6652814887657687e-08}
{"train_loss": 0.04463260801864809, "global_step": 269224, "epoch": 3024, "lr": 1.6637856980245002e-08, "val_loss": 9.629706382751465}
{"train_loss": 0.06106162816286087, "global_step": 269225, "epoch": 3025, "lr": 1.6622905792484934e-08}
{"train_loss": 0.049428585916757584, "global_step": 269226, "epoch": 3025, "lr": 1.6607961324383026e-08}
{"train_loss": 0.05947016924619675, "global_step": 269227, "epoch": 3025, "lr": 1.6593023575933732e-08}
{"train_loss": 0.027381235733628273, "global_step": 269228, "epoch": 3025, "lr": 1.657809254714815e-08}
{"train_loss": 0.04005521908402443, "global_step": 269229, "epoch": 3025, "lr": 1.6563168238020733e-08}
{"train_loss": 0.030888238921761513, "global_step": 269230, "epoch": 3025, "lr": 1.654825064856258e-08}
{"train_loss": 0.0283499862998724, "global_step": 269231, "epoch": 3025, "lr": 1.6533339778768142e-08}
{"train_loss": 0.03050255961716175, "global_step": 269232, "epoch": 3025, "lr": 1.651843562863742e-08}
{"train_loss": 0.030713923275470734, "global_step": 269233, "epoch": 3025, "lr": 1.650353819818151e-08}
{"train_loss": 0.0892108604311943, "global_step": 269234, "epoch": 3025, "lr": 1.648864748739487e-08}
{"train_loss": 0.06221363693475723, "global_step": 269235, "epoch": 3025, "lr": 1.6473763496283046e-08}
{"train_loss": 0.028413476422429085, "global_step": 269236, "epoch": 3025, "lr": 1.6458886224846036e-08}
{"train_loss": 0.03743560612201691, "global_step": 269237, "epoch": 3025, "lr": 1.6444015673089398e-08}
{"train_loss": 0.024407878518104553, "global_step": 269238, "epoch": 3025, "lr": 1.6429151841013123e-08}
{"train_loss": 0.077851802110672, "global_step": 269239, "epoch": 3025, "lr": 1.641429472861722e-08}
{"train_loss": 0.04209071025252342, "global_step": 269240, "epoch": 3025, "lr": 1.639944433590168e-08}
{"train_loss": 0.05727691948413849, "global_step": 269241, "epoch": 3025, "lr": 1.6384600662877614e-08}
{"train_loss": 0.026916958391666412, "global_step": 269242, "epoch": 3025, "lr": 1.6369763709539466e-08}
{"train_loss": 0.03261120244860649, "global_step": 269243, "epoch": 3025, "lr": 1.6354933475892787e-08}
{"train_loss": 0.041632529348134995, "global_step": 269244, "epoch": 3025, "lr": 1.6340109961932028e-08}
{"train_loss": 0.04102810099720955, "global_step": 269245, "epoch": 3025, "lr": 1.632529316766829e-08}
{"train_loss": 0.06950138509273529, "global_step": 269246, "epoch": 3025, "lr": 1.631048309310157e-08}
{"train_loss": 0.03850944712758064, "global_step": 269247, "epoch": 3025, "lr": 1.6295679738231872e-08}
{"train_loss": 0.03925780579447746, "global_step": 269248, "epoch": 3025, "lr": 1.6280883103064747e-08}
{"train_loss": 0.06283238530158997, "global_step": 269249, "epoch": 3025, "lr": 1.6266093187594645e-08}
{"train_loss": 0.044190436601638794, "global_step": 269250, "epoch": 3025, "lr": 1.6251309991832664e-08}
{"train_loss": 0.05910023674368858, "global_step": 269251, "epoch": 3025, "lr": 1.6236533515773256e-08}
{"train_loss": 0.06774098426103592, "global_step": 269252, "epoch": 3025, "lr": 1.622176375942752e-08}
{"train_loss": 0.046117495745420456, "global_step": 269253, "epoch": 3025, "lr": 1.620700072278436e-08}
{"train_loss": 0.023838864639401436, "global_step": 269254, "epoch": 3025, "lr": 1.6192244405854873e-08}
{"train_loss": 0.03904974088072777, "global_step": 269255, "epoch": 3025, "lr": 1.617749480863906e-08}
{"train_loss": 0.046640872955322266, "global_step": 269256, "epoch": 3025, "lr": 1.6162751931142472e-08}
{"train_loss": 0.03875105828046799, "global_step": 269257, "epoch": 3025, "lr": 1.6148015773359558e-08}
{"train_loss": 0.0767010748386383, "global_step": 269258, "epoch": 3025, "lr": 1.613328633530142e-08}
{"train_loss": 0.04319952800869942, "global_step": 269259, "epoch": 3025, "lr": 1.611856361695696e-08}
{"train_loss": 0.03435683250427246, "global_step": 269260, "epoch": 3025, "lr": 1.6103847618342827e-08}
{"train_loss": 0.0530628077685833, "global_step": 269261, "epoch": 3025, "lr": 1.608913833945347e-08}
{"train_loss": 0.05758604779839516, "global_step": 269262, "epoch": 3025, "lr": 1.6074435780288887e-08}
{"train_loss": 0.01951863430440426, "global_step": 269263, "epoch": 3025, "lr": 1.6059739940854633e-08}
{"train_loss": 0.06489737331867218, "global_step": 269264, "epoch": 3025, "lr": 1.604505082115071e-08}
{"train_loss": 0.022262729704380035, "global_step": 269265, "epoch": 3025, "lr": 1.6030368421182663e-08}
{"train_loss": 0.026259884238243103, "global_step": 269266, "epoch": 3025, "lr": 1.6015692740950495e-08}
{"train_loss": 0.018656650558114052, "global_step": 269267, "epoch": 3025, "lr": 1.6001023780454205e-08}
{"train_loss": 0.09616570919752121, "global_step": 269268, "epoch": 3025, "lr": 1.5986361539699345e-08}
{"train_loss": 0.056549280881881714, "global_step": 269269, "epoch": 3025, "lr": 1.5971706018685915e-08}
{"train_loss": 0.028866194188594818, "global_step": 269270, "epoch": 3025, "lr": 1.595705721741392e-08}
{"train_loss": 0.0636720061302185, "global_step": 269271, "epoch": 3025, "lr": 1.594241513588335e-08}
{"train_loss": 0.057538170367479324, "global_step": 269272, "epoch": 3025, "lr": 1.592777977410531e-08}
{"train_loss": 0.14717437326908112, "global_step": 269273, "epoch": 3025, "lr": 1.5913151132079806e-08}
{"train_loss": 0.05149712413549423, "global_step": 269274, "epoch": 3025, "lr": 1.5898529209801284e-08}
{"train_loss": 0.051779404282569885, "global_step": 269275, "epoch": 3025, "lr": 1.5883914007275293e-08}
{"train_loss": 0.07822743058204651, "global_step": 269276, "epoch": 3025, "lr": 1.5869305524501832e-08}
{"train_loss": 0.05020981281995773, "global_step": 269277, "epoch": 3025, "lr": 1.5854703761492006e-08}
{"train_loss": 0.015225893817842007, "global_step": 269278, "epoch": 3025, "lr": 1.5840108718240264e-08}
{"train_loss": 0.07154513150453568, "global_step": 269279, "epoch": 3025, "lr": 1.5825520394746607e-08}
{"train_loss": 0.04262072592973709, "global_step": 269280, "epoch": 3025, "lr": 1.5810938791016582e-08}
{"train_loss": 0.005603868514299393, "global_step": 269281, "epoch": 3025, "lr": 1.5796363907050193e-08}
{"train_loss": 0.03915383666753769, "global_step": 269282, "epoch": 3025, "lr": 1.578179574285299e-08}
{"train_loss": 0.04980861768126488, "global_step": 269283, "epoch": 3025, "lr": 1.576723429842497e-08}
{"train_loss": 0.03603600710630417, "global_step": 269284, "epoch": 3025, "lr": 1.5752679573766138e-08}
{"train_loss": 0.023558055981993675, "global_step": 269285, "epoch": 3025, "lr": 1.5738131568882042e-08}
{"train_loss": 0.041136059910058975, "global_step": 269286, "epoch": 3025, "lr": 1.572359028377268e-08}
{"train_loss": 0.060149841010570526, "global_step": 269287, "epoch": 3025, "lr": 1.570905571843806e-08}
{"train_loss": 0.05399506911635399, "global_step": 269288, "epoch": 3025, "lr": 1.5694527872883725e-08}
{"train_loss": 0.04482763633131981, "global_step": 269289, "epoch": 3025, "lr": 1.5680006747109676e-08}
{"train_loss": 0.07637841254472733, "global_step": 269290, "epoch": 3025, "lr": 1.566549234112147e-08}
{"train_loss": 0.05829213932156563, "global_step": 269291, "epoch": 3025, "lr": 1.5650984654913547e-08}
{"train_loss": 0.057268187403678894, "global_step": 269292, "epoch": 3025, "lr": 1.5636483688491465e-08}
{"train_loss": 0.07106634974479675, "global_step": 269293, "epoch": 3025, "lr": 1.5621989441860774e-08}
{"train_loss": 0.020365756005048752, "global_step": 269294, "epoch": 3025, "lr": 1.5607501915021472e-08}
{"train_loss": 0.016186922788619995, "global_step": 269295, "epoch": 3025, "lr": 1.559302110796801e-08}
{"train_loss": 0.036423541605472565, "global_step": 269296, "epoch": 3025, "lr": 1.5578547020717037e-08}
{"train_loss": 0.029012659564614296, "global_step": 269297, "epoch": 3025, "lr": 1.5564079653257456e-08}
{"train_loss": 0.08580930531024933, "global_step": 269298, "epoch": 3025, "lr": 1.554961900559482e-08}
{"train_loss": 0.06402215361595154, "global_step": 269299, "epoch": 3025, "lr": 1.553516507773467e-08}
{"train_loss": 0.051703788340091705, "global_step": 269300, "epoch": 3025, "lr": 1.5520717869677016e-08}
{"train_loss": 0.04435698315501213, "global_step": 269301, "epoch": 3025, "lr": 1.5506277381421853e-08}
{"train_loss": 0.04776477813720703, "global_step": 269302, "epoch": 3025, "lr": 1.5491843612974732e-08}
{"train_loss": 0.062074676156044006, "global_step": 269303, "epoch": 3025, "lr": 1.5477416564335655e-08}
{"train_loss": 0.05583740398287773, "global_step": 269304, "epoch": 3025, "lr": 1.5462996235504623e-08}
{"train_loss": 0.051703695207834244, "global_step": 269305, "epoch": 3025, "lr": 1.5448582626487184e-08}
{"train_loss": 0.025268318131566048, "global_step": 269306, "epoch": 3025, "lr": 1.5434175737283342e-08}
{"train_loss": 0.021678689867258072, "global_step": 269307, "epoch": 3025, "lr": 1.541977556789309e-08}
{"train_loss": 0.027941877022385597, "global_step": 269308, "epoch": 3025, "lr": 1.5405382118321987e-08}
{"train_loss": 0.041908107697963715, "global_step": 269309, "epoch": 3025, "lr": 1.5390995388564477e-08}
{"train_loss": 0.04528488591313362, "global_step": 269310, "epoch": 3025, "lr": 1.5376615378637216e-08}
{"train_loss": 0.060085948556661606, "global_step": 269311, "epoch": 3025, "lr": 1.53622420885291e-08}
{"train_loss": 0.005785314831882715, "global_step": 269312, "epoch": 3025, "lr": 1.534787551824568e-08}
{"train_loss": 0.04713579051175647, "global_step": 269313, "epoch": 3025, "lr": 1.5333515667792508e-08, "val_loss": 9.70526123046875, "train_action_mse_error": 16.19965934753418}
{"train_loss": 0.0511745885014534, "global_step": 269314, "epoch": 3026, "lr": 1.5319162537164034e-08}
{"train_loss": 0.010308506898581982, "global_step": 269315, "epoch": 3026, "lr": 1.5304816126371358e-08}
{"train_loss": 0.014548234641551971, "global_step": 269316, "epoch": 3026, "lr": 1.5290476435408928e-08}
{"train_loss": 0.029984118416905403, "global_step": 269317, "epoch": 3026, "lr": 1.52761434642823e-08}
{"train_loss": 0.05068480968475342, "global_step": 269318, "epoch": 3026, "lr": 1.526181721299147e-08}
{"train_loss": 0.055623091757297516, "global_step": 269319, "epoch": 3026, "lr": 1.524749768154199e-08}
{"train_loss": 0.046719029545784, "global_step": 269320, "epoch": 3026, "lr": 1.523318486993386e-08}
{"train_loss": 0.06105206534266472, "global_step": 269321, "epoch": 3026, "lr": 1.5218878778167077e-08}
{"train_loss": 0.04645464941859245, "global_step": 269322, "epoch": 3026, "lr": 1.5204579406241648e-08}
{"train_loss": 0.05148616060614586, "global_step": 269323, "epoch": 3026, "lr": 1.519028675416867e-08}
{"train_loss": 0.014951616525650024, "global_step": 269324, "epoch": 3026, "lr": 1.517600082193704e-08}
{"train_loss": 0.04680735617876053, "global_step": 269325, "epoch": 3026, "lr": 1.5161721609563417e-08}
{"train_loss": 0.07661362737417221, "global_step": 269326, "epoch": 3026, "lr": 1.5147449117036694e-08}
{"train_loss": 0.029308326542377472, "global_step": 269327, "epoch": 3026, "lr": 1.5133183344362422e-08}
{"train_loss": 0.06471030414104462, "global_step": 269328, "epoch": 3026, "lr": 1.5118924291551705e-08}
{"train_loss": 0.042694441974163055, "global_step": 269329, "epoch": 3026, "lr": 1.5104671958593443e-08}
{"train_loss": 0.03183874860405922, "global_step": 269330, "epoch": 3026, "lr": 1.509042634549318e-08}
{"train_loss": 0.016494836658239365, "global_step": 269331, "epoch": 3026, "lr": 1.5076187452262024e-08}
{"train_loss": 0.017397506162524223, "global_step": 269332, "epoch": 3026, "lr": 1.5061955278888872e-08}
{"train_loss": 0.07217761874198914, "global_step": 269333, "epoch": 3026, "lr": 1.5047729825384826e-08}
{"train_loss": 0.04403083026409149, "global_step": 269334, "epoch": 3026, "lr": 1.5033511091744336e-08}
{"train_loss": 0.020017175003886223, "global_step": 269335, "epoch": 3026, "lr": 1.5019299077978498e-08}
{"train_loss": 0.06730623543262482, "global_step": 269336, "epoch": 3026, "lr": 1.500509378408177e-08}
{"train_loss": 0.023124802857637405, "global_step": 269337, "epoch": 3026, "lr": 1.4990895210054147e-08}
{"train_loss": 0.027691148221492767, "global_step": 269338, "epoch": 3026, "lr": 1.497670335590673e-08}
{"train_loss": 0.028304830193519592, "global_step": 269339, "epoch": 3026, "lr": 1.496251822163397e-08}
{"train_loss": 0.013223251327872276, "global_step": 269340, "epoch": 3026, "lr": 1.494833980724142e-08}
{"train_loss": 0.05197543650865555, "global_step": 269341, "epoch": 3026, "lr": 1.4934168112734626e-08}
{"train_loss": 0.05037611350417137, "global_step": 269342, "epoch": 3026, "lr": 1.492000313810249e-08}
{"train_loss": 0.030182678252458572, "global_step": 269343, "epoch": 3026, "lr": 1.4905844883361662e-08}
{"train_loss": 0.04466494545340538, "global_step": 269344, "epoch": 3026, "lr": 1.4891693348501045e-08}
{"train_loss": 0.04220423474907875, "global_step": 269345, "epoch": 3026, "lr": 1.4877548533531737e-08}
{"train_loss": 0.04597419127821922, "global_step": 269346, "epoch": 3026, "lr": 1.486341043845374e-08}
{"train_loss": 0.034052591770887375, "global_step": 269347, "epoch": 3026, "lr": 1.4849279063272603e-08}
{"train_loss": 0.04483265057206154, "global_step": 269348, "epoch": 3026, "lr": 1.4835154407977225e-08}
{"train_loss": 0.04283124953508377, "global_step": 269349, "epoch": 3026, "lr": 1.482103647258426e-08}
{"train_loss": 0.008935629390180111, "global_step": 269350, "epoch": 3026, "lr": 1.4806925257088155e-08}
{"train_loss": 0.03316270560026169, "global_step": 269351, "epoch": 3026, "lr": 1.4792820761488912e-08}
{"train_loss": 0.06677115708589554, "global_step": 269352, "epoch": 3026, "lr": 1.4778722985792082e-08}
{"train_loss": 0.04367301985621452, "global_step": 269353, "epoch": 3026, "lr": 1.4764631930003214e-08}
{"train_loss": 0.029596513137221336, "global_step": 269354, "epoch": 3026, "lr": 1.475054759411676e-08}
{"train_loss": 0.026817647740244865, "global_step": 269355, "epoch": 3026, "lr": 1.4736469978138268e-08}
{"train_loss": 0.03316495940089226, "global_step": 269356, "epoch": 3026, "lr": 1.472239908206774e-08}
{"train_loss": 0.026727281510829926, "global_step": 269357, "epoch": 3026, "lr": 1.4708334905910726e-08}
{"train_loss": 0.029620161280035973, "global_step": 269358, "epoch": 3026, "lr": 1.4694277449667226e-08}
{"train_loss": 0.03152146935462952, "global_step": 269359, "epoch": 3026, "lr": 1.4680226713337241e-08}
{"train_loss": 0.04729844629764557, "global_step": 269360, "epoch": 3026, "lr": 1.4666182696926323e-08}
{"train_loss": 0.05971193686127663, "global_step": 269361, "epoch": 3026, "lr": 1.465214540043447e-08}
{"train_loss": 0.023817678913474083, "global_step": 269362, "epoch": 3026, "lr": 1.4638114823861681e-08}
{"train_loss": 0.026728132739663124, "global_step": 269363, "epoch": 3026, "lr": 1.462409096721351e-08}
{"train_loss": 0.04874785244464874, "global_step": 269364, "epoch": 3026, "lr": 1.4610073830484405e-08}
{"train_loss": 0.026398098096251488, "global_step": 269365, "epoch": 3026, "lr": 1.4596063413685467e-08}
{"train_loss": 0.06318788230419159, "global_step": 269366, "epoch": 3026, "lr": 1.4582059716816698e-08}
{"train_loss": 0.032540544867515564, "global_step": 269367, "epoch": 3026, "lr": 1.4568062739878096e-08}
{"train_loss": 0.04440554603934288, "global_step": 269368, "epoch": 3026, "lr": 1.4554072482869663e-08}
{"train_loss": 0.021553728729486465, "global_step": 269369, "epoch": 3026, "lr": 1.4540088945796947e-08}
{"train_loss": 0.03977705538272858, "global_step": 269370, "epoch": 3026, "lr": 1.4526112128654401e-08}
{"train_loss": 0.08005574345588684, "global_step": 269371, "epoch": 3026, "lr": 1.4512142031453124e-08}
{"train_loss": 0.019152404740452766, "global_step": 269372, "epoch": 3026, "lr": 1.4498178654193118e-08}
{"train_loss": 0.012395190075039864, "global_step": 269373, "epoch": 3026, "lr": 1.4484221996874381e-08}
{"train_loss": 0.0951344221830368, "global_step": 269374, "epoch": 3026, "lr": 1.4470272059496915e-08}
{"train_loss": 0.05463669076561928, "global_step": 269375, "epoch": 3026, "lr": 1.445632884206627e-08}
{"train_loss": 0.04741925373673439, "global_step": 269376, "epoch": 3026, "lr": 1.4442392344582445e-08}
{"train_loss": 0.05793513357639313, "global_step": 269377, "epoch": 3026, "lr": 1.4428462567050993e-08}
{"train_loss": 0.05316641181707382, "global_step": 269378, "epoch": 3026, "lr": 1.4414539509466363e-08}
{"train_loss": 0.024341344833374023, "global_step": 269379, "epoch": 3026, "lr": 1.4400623171834105e-08}
{"train_loss": 0.08515046536922455, "global_step": 269380, "epoch": 3026, "lr": 1.438671355415977e-08}
{"train_loss": 0.06997577100992203, "global_step": 269381, "epoch": 3026, "lr": 1.4372810656437807e-08}
{"train_loss": 0.04447294771671295, "global_step": 269382, "epoch": 3026, "lr": 1.4358914478673769e-08}
{"train_loss": 0.014580000191926956, "global_step": 269383, "epoch": 3026, "lr": 1.4345025020873205e-08}
{"train_loss": 0.046534422785043716, "global_step": 269384, "epoch": 3026, "lr": 1.4331142283036114e-08}
{"train_loss": 0.06899335235357285, "global_step": 269385, "epoch": 3026, "lr": 1.4317266265156948e-08}
{"train_loss": 0.04811232164502144, "global_step": 269386, "epoch": 3026, "lr": 1.4303396967246807e-08}
{"train_loss": 0.031875330954790115, "global_step": 269387, "epoch": 3026, "lr": 1.4289534389305693e-08}
{"train_loss": 0.03378671407699585, "global_step": 269388, "epoch": 3026, "lr": 1.4275678531333603e-08}
{"train_loss": 0.06435590982437134, "global_step": 269389, "epoch": 3026, "lr": 1.426182939333054e-08}
{"train_loss": 0.025866471230983734, "global_step": 269390, "epoch": 3026, "lr": 1.4247986975302052e-08}
{"train_loss": 0.0657983347773552, "global_step": 269391, "epoch": 3026, "lr": 1.4234151277253693e-08}
{"train_loss": 0.013628566637635231, "global_step": 269392, "epoch": 3026, "lr": 1.422032229917436e-08}
{"train_loss": 0.051010869443416595, "global_step": 269393, "epoch": 3026, "lr": 1.4206500041080706e-08}
{"train_loss": 0.00831360463052988, "global_step": 269394, "epoch": 3026, "lr": 1.4192684502961628e-08}
{"train_loss": 0.02023889310657978, "global_step": 269395, "epoch": 3026, "lr": 1.417887568482823e-08}
{"train_loss": 0.03402376174926758, "global_step": 269396, "epoch": 3026, "lr": 1.4165073586680511e-08}
{"train_loss": 0.028688184916973114, "global_step": 269397, "epoch": 3026, "lr": 1.415127820851847e-08}
{"train_loss": 0.044995322823524475, "global_step": 269398, "epoch": 3026, "lr": 1.413748955034211e-08}
{"train_loss": 0.06682279706001282, "global_step": 269399, "epoch": 3026, "lr": 1.4123707612156978e-08}
{"train_loss": 0.04776942357420921, "global_step": 269400, "epoch": 3026, "lr": 1.4109932393963077e-08}
{"train_loss": 0.035693779587745667, "global_step": 269401, "epoch": 3026, "lr": 1.4096163895760405e-08}
{"train_loss": 0.04095173956763543, "global_step": 269402, "epoch": 3026, "lr": 1.4082402117560067e-08, "val_loss": 9.616164207458496}
{"train_loss": 0.03802086412906647, "global_step": 269403, "epoch": 3027, "lr": 1.4068647059350958e-08}
{"train_loss": 0.04448666796088219, "global_step": 269404, "epoch": 3027, "lr": 1.4054898721144182e-08}
{"train_loss": 0.020699135959148407, "global_step": 269405, "epoch": 3027, "lr": 1.4041157102939739e-08}
{"train_loss": 0.03879004716873169, "global_step": 269406, "epoch": 3027, "lr": 1.4027422204737628e-08}
{"train_loss": 0.04141422361135483, "global_step": 269407, "epoch": 3027, "lr": 1.4013694026537849e-08}
{"train_loss": 0.08244621753692627, "global_step": 269408, "epoch": 3027, "lr": 1.3999972568345954e-08}
{"train_loss": 0.018258314579725266, "global_step": 269409, "epoch": 3027, "lr": 1.3986257830161941e-08}
{"train_loss": 0.05592586100101471, "global_step": 269410, "epoch": 3027, "lr": 1.3972549811985813e-08}
{"train_loss": 0.026732830330729485, "global_step": 269411, "epoch": 3027, "lr": 1.395884851382867e-08}
{"train_loss": 0.057985514402389526, "global_step": 269412, "epoch": 3027, "lr": 1.3945153935679411e-08}
{"train_loss": 0.06908568739891052, "global_step": 269413, "epoch": 3027, "lr": 1.3931466077549137e-08}
{"train_loss": 0.054773252457380295, "global_step": 269414, "epoch": 3027, "lr": 1.3917784939432299e-08}
{"train_loss": 0.06462360918521881, "global_step": 269415, "epoch": 3027, "lr": 1.3904110521339996e-08}
{"train_loss": 0.007623175159096718, "global_step": 269416, "epoch": 3027, "lr": 1.3890442823266681e-08}
{"train_loss": 0.050028566271066666, "global_step": 269417, "epoch": 3027, "lr": 1.3876781845217902e-08}
{"train_loss": 0.030198831111192703, "global_step": 269418, "epoch": 3027, "lr": 1.3863127587193658e-08}
{"train_loss": 0.05411113053560257, "global_step": 269419, "epoch": 3027, "lr": 1.3849480049193953e-08}
{"train_loss": 0.04529178515076637, "global_step": 269420, "epoch": 3027, "lr": 1.3835839231224335e-08}
{"train_loss": 0.06622146815061569, "global_step": 269421, "epoch": 3027, "lr": 1.3822205133290356e-08}
{"train_loss": 0.03373894840478897, "global_step": 269422, "epoch": 3027, "lr": 1.3808577755380914e-08}
{"train_loss": 0.040134746581315994, "global_step": 269423, "epoch": 3027, "lr": 1.3794957097507111e-08}
{"train_loss": 0.0609687976539135, "global_step": 269424, "epoch": 3027, "lr": 1.3781343159674498e-08}
{"train_loss": 0.020736705511808395, "global_step": 269425, "epoch": 3027, "lr": 1.3767735941877524e-08}
{"train_loss": 0.05151228606700897, "global_step": 269426, "epoch": 3027, "lr": 1.375413544411619e-08}
{"train_loss": 0.07801792025566101, "global_step": 269427, "epoch": 3027, "lr": 1.3740541666401596e-08}
{"train_loss": 0.018035273998975754, "global_step": 269428, "epoch": 3027, "lr": 1.3726954608728193e-08}
{"train_loss": 0.054602473974227905, "global_step": 269429, "epoch": 3027, "lr": 1.371337427109598e-08}
{"train_loss": 0.03267590329051018, "global_step": 269430, "epoch": 3027, "lr": 1.3699800653516059e-08}
{"train_loss": 0.02302674949169159, "global_step": 269431, "epoch": 3027, "lr": 1.368623375598288e-08}
{"train_loss": 0.07911664247512817, "global_step": 269432, "epoch": 3027, "lr": 1.3672673578501994e-08}
{"train_loss": 0.028592916205525398, "global_step": 269433, "epoch": 3027, "lr": 1.36591201210734e-08}
{"train_loss": 0.03539067506790161, "global_step": 269434, "epoch": 3027, "lr": 1.3645573383697097e-08}
{"train_loss": 0.03818075731396675, "global_step": 269435, "epoch": 3027, "lr": 1.3632033366378638e-08}
{"train_loss": 0.01794428378343582, "global_step": 269436, "epoch": 3027, "lr": 1.3618500069118023e-08}
{"train_loss": 0.025810889899730682, "global_step": 269437, "epoch": 3027, "lr": 1.3604973491915252e-08}
{"train_loss": 0.05426863580942154, "global_step": 269438, "epoch": 3027, "lr": 1.3591453634775875e-08}
{"train_loss": 0.034281037747859955, "global_step": 269439, "epoch": 3027, "lr": 1.3577940497699892e-08}
{"train_loss": 0.01858973316848278, "global_step": 269440, "epoch": 3027, "lr": 1.3564434080687306e-08}
{"train_loss": 0.058290109038352966, "global_step": 269441, "epoch": 3027, "lr": 1.3550934383743663e-08}
{"train_loss": 0.0701526552438736, "global_step": 269442, "epoch": 3027, "lr": 1.3537441406868967e-08}
{"train_loss": 0.039863426238298416, "global_step": 269443, "epoch": 3027, "lr": 1.3523955150063217e-08}
{"train_loss": 0.07963528484106064, "global_step": 269444, "epoch": 3027, "lr": 1.3510475613331964e-08}
{"train_loss": 0.07676158100366592, "global_step": 269445, "epoch": 3027, "lr": 1.3497002796675206e-08}
{"train_loss": 0.06415041536092758, "global_step": 269446, "epoch": 3027, "lr": 1.3483536700092947e-08}
{"train_loss": 0.043728508055210114, "global_step": 269447, "epoch": 3027, "lr": 1.3470077323585185e-08}
{"train_loss": 0.05833176523447037, "global_step": 269448, "epoch": 3027, "lr": 1.345662466716302e-08}
{"train_loss": 0.03223686292767525, "global_step": 269449, "epoch": 3027, "lr": 1.3443178730820904e-08}
{"train_loss": 0.04231393709778786, "global_step": 269450, "epoch": 3027, "lr": 1.3429739514558837e-08}
{"train_loss": 0.07788269966840744, "global_step": 269451, "epoch": 3027, "lr": 1.3416307018387918e-08}
{"train_loss": 0.0581209734082222, "global_step": 269452, "epoch": 3027, "lr": 1.340288124229705e-08}
{"train_loss": 0.03903630003333092, "global_step": 269453, "epoch": 3027, "lr": 1.338946218629733e-08}
{"train_loss": 0.05007040873169899, "global_step": 269454, "epoch": 3027, "lr": 1.337604985038876e-08}
{"train_loss": 0.054598111659288406, "global_step": 269455, "epoch": 3027, "lr": 1.3362644234576893e-08}
{"train_loss": 0.06692153960466385, "global_step": 269456, "epoch": 3027, "lr": 1.3349245338850625e-08}
{"train_loss": 0.05583632364869118, "global_step": 269457, "epoch": 3027, "lr": 1.3335853163226608e-08}
{"train_loss": 0.039459697902202606, "global_step": 269458, "epoch": 3027, "lr": 1.3322467707699293e-08}
{"train_loss": 0.04803341254591942, "global_step": 269459, "epoch": 3027, "lr": 1.330908897226868e-08}
{"train_loss": 0.021075677126646042, "global_step": 269460, "epoch": 3027, "lr": 1.3295716956940319e-08}
{"train_loss": 0.0810965746641159, "global_step": 269461, "epoch": 3027, "lr": 1.3282351661719761e-08}
{"train_loss": 0.02489999681711197, "global_step": 269462, "epoch": 3027, "lr": 1.3268993086601455e-08}
{"train_loss": 0.058510225266218185, "global_step": 269463, "epoch": 3027, "lr": 1.3255641231585403e-08}
{"train_loss": 0.021361825987696648, "global_step": 269464, "epoch": 3027, "lr": 1.3242296096682705e-08}
{"train_loss": 0.018257085233926773, "global_step": 269465, "epoch": 3027, "lr": 1.322895768188781e-08}
{"train_loss": 0.07158267498016357, "global_step": 269466, "epoch": 3027, "lr": 1.3215625987206271e-08}
{"train_loss": 0.06522582471370697, "global_step": 269467, "epoch": 3027, "lr": 1.3202301012638085e-08}
{"train_loss": 0.060687173157930374, "global_step": 269468, "epoch": 3027, "lr": 1.3188982758188805e-08}
{"train_loss": 0.06409450620412827, "global_step": 269469, "epoch": 3027, "lr": 1.317567122385288e-08}
{"train_loss": 0.014348248019814491, "global_step": 269470, "epoch": 3027, "lr": 1.3162366409641414e-08}
{"train_loss": 0.022919798269867897, "global_step": 269471, "epoch": 3027, "lr": 1.31490683155433e-08}
{"train_loss": 0.050852276384830475, "global_step": 269472, "epoch": 3027, "lr": 1.3135776941575195e-08}
{"train_loss": 0.029650220647454262, "global_step": 269473, "epoch": 3027, "lr": 1.3122492287725996e-08}
{"train_loss": 0.02700984477996826, "global_step": 269474, "epoch": 3027, "lr": 1.3109214354006805e-08}
{"train_loss": 0.05626023933291435, "global_step": 269475, "epoch": 3027, "lr": 1.3095943140417622e-08}
{"train_loss": 0.020511822775006294, "global_step": 269476, "epoch": 3027, "lr": 1.3082678646952895e-08}
{"train_loss": 0.03992322087287903, "global_step": 269477, "epoch": 3027, "lr": 1.3069420873623728e-08}
{"train_loss": 0.019114816561341286, "global_step": 269478, "epoch": 3027, "lr": 1.3056169820424569e-08}
{"train_loss": 0.04226003214716911, "global_step": 269479, "epoch": 3027, "lr": 1.304292548736652e-08}
{"train_loss": 0.058172859251499176, "global_step": 269480, "epoch": 3027, "lr": 1.302968787443848e-08}
{"train_loss": 0.04058326408267021, "global_step": 269481, "epoch": 3027, "lr": 1.3016456981657099e-08}
{"train_loss": 0.030217867344617844, "global_step": 269482, "epoch": 3027, "lr": 1.3003232809011278e-08}
{"train_loss": 0.054435212165117264, "global_step": 269483, "epoch": 3027, "lr": 1.299001535651212e-08}
{"train_loss": 0.019321374595165253, "global_step": 269484, "epoch": 3027, "lr": 1.2976804624154071e-08}
{"train_loss": 0.026562651619315147, "global_step": 269485, "epoch": 3027, "lr": 1.2963600611937132e-08}
{"train_loss": 0.03378506749868393, "global_step": 269486, "epoch": 3027, "lr": 1.2950403319877957e-08}
{"train_loss": 0.01887902431190014, "global_step": 269487, "epoch": 3027, "lr": 1.2937212747959892e-08}
{"train_loss": 0.03910764679312706, "global_step": 269488, "epoch": 3027, "lr": 1.2924028896199591e-08}
{"train_loss": 0.08882701396942139, "global_step": 269489, "epoch": 3027, "lr": 1.2910851764585953e-08}
{"train_loss": 0.05907014012336731, "global_step": 269490, "epoch": 3027, "lr": 1.2897681353130076e-08}
{"train_loss": 0.04507962693826536, "global_step": 269491, "epoch": 3027, "lr": 1.2884517661831964e-08, "val_loss": 9.619722366333008}
{"train_loss": 0.026349198073148727, "global_step": 269492, "epoch": 3028, "lr": 1.2871360690691613e-08}
{"train_loss": 0.06335795670747757, "global_step": 269493, "epoch": 3028, "lr": 1.285821043971458e-08}
{"train_loss": 0.015087681822478771, "global_step": 269494, "epoch": 3028, "lr": 1.2845066908895308e-08}
{"train_loss": 0.016019796952605247, "global_step": 269495, "epoch": 3028, "lr": 1.2831930098244904e-08}
{"train_loss": 0.05413869768381119, "global_step": 269496, "epoch": 3028, "lr": 1.2818800007752264e-08}
{"train_loss": 0.05621771514415741, "global_step": 269497, "epoch": 3028, "lr": 1.2805676637434039e-08}
{"train_loss": 0.029499182477593422, "global_step": 269498, "epoch": 3028, "lr": 1.279255998727913e-08}
{"train_loss": 0.062244564294815063, "global_step": 269499, "epoch": 3028, "lr": 1.2779450057298637e-08}
{"train_loss": 0.06472432613372803, "global_step": 269500, "epoch": 3028, "lr": 1.276634684748701e-08}
{"train_loss": 0.02503998391330242, "global_step": 269501, "epoch": 3028, "lr": 1.275325035785535e-08}
{"train_loss": 0.03841660916805267, "global_step": 269502, "epoch": 3028, "lr": 1.2740160588392558e-08}
{"train_loss": 0.04375210776925087, "global_step": 269503, "epoch": 3028, "lr": 1.2727077539115284e-08}
{"train_loss": 0.040473390370607376, "global_step": 269504, "epoch": 3028, "lr": 1.2714001210012427e-08}
{"train_loss": 0.0666351243853569, "global_step": 269505, "epoch": 3028, "lr": 1.270093160109509e-08}
{"train_loss": 0.017370032146573067, "global_step": 269506, "epoch": 3028, "lr": 1.268786871235772e-08}
{"train_loss": 0.07261555641889572, "global_step": 269507, "epoch": 3028, "lr": 1.267481254380587e-08}
{"train_loss": 0.04441571980714798, "global_step": 269508, "epoch": 3028, "lr": 1.2661763095439538e-08}
{"train_loss": 0.024932075291872025, "global_step": 269509, "epoch": 3028, "lr": 1.2648720367258725e-08}
{"train_loss": 0.02455013245344162, "global_step": 269510, "epoch": 3028, "lr": 1.2635684359274535e-08}
{"train_loss": 0.060133110731840134, "global_step": 269511, "epoch": 3028, "lr": 1.2622655071475864e-08}
{"train_loss": 0.05597969889640808, "global_step": 269512, "epoch": 3028, "lr": 1.2609632503879365e-08}
{"train_loss": 0.02457149513065815, "global_step": 269513, "epoch": 3028, "lr": 1.2596616656468385e-08}
{"train_loss": 0.07461697608232498, "global_step": 269514, "epoch": 3028, "lr": 1.2583607529259578e-08}
{"train_loss": 0.03211306408047676, "global_step": 269515, "epoch": 3028, "lr": 1.2570605122252942e-08}
{"train_loss": 0.12622039020061493, "global_step": 269516, "epoch": 3028, "lr": 1.255760943544293e-08}
{"train_loss": 0.0297309011220932, "global_step": 269517, "epoch": 3028, "lr": 1.254462046883509e-08}
{"train_loss": 0.061387110501527786, "global_step": 269518, "epoch": 3028, "lr": 1.2531638222434971e-08}
{"train_loss": 0.07100239396095276, "global_step": 269519, "epoch": 3028, "lr": 1.2518662696237027e-08}
{"train_loss": 0.03494514524936676, "global_step": 269520, "epoch": 3028, "lr": 1.2505693890246806e-08}
{"train_loss": 0.05770416557788849, "global_step": 269521, "epoch": 3028, "lr": 1.2492731804464309e-08}
{"train_loss": 0.07780688256025314, "global_step": 269522, "epoch": 3028, "lr": 1.2479776438895086e-08}
{"train_loss": 0.05682745948433876, "global_step": 269523, "epoch": 3028, "lr": 1.2466827793539137e-08}
{"train_loss": 0.02533465065062046, "global_step": 269524, "epoch": 3028, "lr": 1.2453885868396465e-08}
{"train_loss": 0.04213903844356537, "global_step": 269525, "epoch": 3028, "lr": 1.2440950663472617e-08}
{"train_loss": 0.07323990762233734, "global_step": 269526, "epoch": 3028, "lr": 1.2428022178767595e-08}
{"train_loss": 0.056932833045721054, "global_step": 269527, "epoch": 3028, "lr": 1.24151004142814e-08}
{"train_loss": 0.03560841083526611, "global_step": 269528, "epoch": 3028, "lr": 1.240218537001403e-08}
{"train_loss": 0.0802888423204422, "global_step": 269529, "epoch": 3028, "lr": 1.2389277045971037e-08}
{"train_loss": 0.03749706968665123, "global_step": 269530, "epoch": 3028, "lr": 1.2376375442152421e-08}
{"train_loss": 0.03522077575325966, "global_step": 269531, "epoch": 3028, "lr": 1.2363480558563734e-08}
{"train_loss": 0.028752122074365616, "global_step": 269532, "epoch": 3028, "lr": 1.2350592395199423e-08}
{"train_loss": 0.03119424730539322, "global_step": 269533, "epoch": 3028, "lr": 1.233771095206504e-08}
{"train_loss": 0.05563125014305115, "global_step": 269534, "epoch": 3028, "lr": 1.2324836229166137e-08}
{"train_loss": 0.04414976015686989, "global_step": 269535, "epoch": 3028, "lr": 1.2311968226497162e-08}
{"train_loss": 0.09705964475870132, "global_step": 269536, "epoch": 3028, "lr": 1.2299106944069216e-08}
{"train_loss": 0.01282557938247919, "global_step": 269537, "epoch": 3028, "lr": 1.22862523818712e-08}
{"train_loss": 0.05340252444148064, "global_step": 269538, "epoch": 3028, "lr": 1.2273404539919765e-08}
{"train_loss": 0.06903088092803955, "global_step": 269539, "epoch": 3028, "lr": 1.2260563418203808e-08}
{"train_loss": 0.013045735657215118, "global_step": 269540, "epoch": 3028, "lr": 1.2247729016728882e-08}
{"train_loss": 0.06282410025596619, "global_step": 269541, "epoch": 3028, "lr": 1.2234901335500537e-08}
{"train_loss": 0.0829794704914093, "global_step": 269542, "epoch": 3028, "lr": 1.2222080374518773e-08}
{"train_loss": 0.06585142761468887, "global_step": 269543, "epoch": 3028, "lr": 1.2209266133783592e-08}
{"train_loss": 0.04828575253486633, "global_step": 269544, "epoch": 3028, "lr": 1.2196458613294992e-08}
{"train_loss": 0.03230537101626396, "global_step": 269545, "epoch": 3028, "lr": 1.2183657813064076e-08}
{"train_loss": 0.07779090851545334, "global_step": 269546, "epoch": 3028, "lr": 1.217086373307974e-08}
{"train_loss": 0.03571311756968498, "global_step": 269547, "epoch": 3028, "lr": 1.215807637335309e-08}
{"train_loss": 0.040519025176763535, "global_step": 269548, "epoch": 3028, "lr": 1.214529573387857e-08}
{"train_loss": 0.06721348315477371, "global_step": 269549, "epoch": 3028, "lr": 1.2132521814667285e-08}
{"train_loss": 0.02488321252167225, "global_step": 269550, "epoch": 3028, "lr": 1.2119754615708135e-08}
{"train_loss": 0.05775332450866699, "global_step": 269551, "epoch": 3028, "lr": 1.2106994137017769e-08}
{"train_loss": 0.04651777446269989, "global_step": 269552, "epoch": 3028, "lr": 1.2094240378585086e-08}
{"train_loss": 0.05058910325169563, "global_step": 269553, "epoch": 3028, "lr": 1.208149334042119e-08}
{"train_loss": 0.07466552406549454, "global_step": 269554, "epoch": 3028, "lr": 1.206875302252053e-08}
{"train_loss": 0.03948955982923508, "global_step": 269555, "epoch": 3028, "lr": 1.2056019424888654e-08}
{"train_loss": 0.11335215717554092, "global_step": 269556, "epoch": 3028, "lr": 1.2043292547525564e-08}
{"train_loss": 0.04440416768193245, "global_step": 269557, "epoch": 3028, "lr": 1.2030572390436812e-08}
{"train_loss": 0.030064962804317474, "global_step": 269558, "epoch": 3028, "lr": 1.2017858953622397e-08}
{"train_loss": 0.02490665391087532, "global_step": 269559, "epoch": 3028, "lr": 1.2005152237076767e-08}
{"train_loss": 0.0784851685166359, "global_step": 269560, "epoch": 3028, "lr": 1.1992452240811024e-08}
{"train_loss": 0.04118203744292259, "global_step": 269561, "epoch": 3028, "lr": 1.1979758964825171e-08}
{"train_loss": 0.03987452760338783, "global_step": 269562, "epoch": 3028, "lr": 1.1967072409119207e-08}
{"train_loss": 0.00807487964630127, "global_step": 269563, "epoch": 3028, "lr": 1.1954392573693129e-08}
{"train_loss": 0.025343069806694984, "global_step": 269564, "epoch": 3028, "lr": 1.1941719458552491e-08}
{"train_loss": 0.020459812134504318, "global_step": 269565, "epoch": 3028, "lr": 1.1929053063697294e-08}
{"train_loss": 0.022990725934505463, "global_step": 269566, "epoch": 3028, "lr": 1.1916393389127533e-08}
{"train_loss": 0.02242141216993332, "global_step": 269567, "epoch": 3028, "lr": 1.1903740434843213e-08}
{"train_loss": 0.07203863561153412, "global_step": 269568, "epoch": 3028, "lr": 1.1891094200855434e-08}
{"train_loss": 0.01872766949236393, "global_step": 269569, "epoch": 3028, "lr": 1.1878454687153096e-08}
{"train_loss": 0.006361131090670824, "global_step": 269570, "epoch": 3028, "lr": 1.1865821893752849e-08}
{"train_loss": 0.0588158443570137, "global_step": 269571, "epoch": 3028, "lr": 1.185319582063804e-08}
{"train_loss": 0.024653077125549316, "global_step": 269572, "epoch": 3028, "lr": 1.1840576467825327e-08}
{"train_loss": 0.043124835938215256, "global_step": 269573, "epoch": 3028, "lr": 1.1827963835314705e-08}
{"train_loss": 0.016622863709926605, "global_step": 269574, "epoch": 3028, "lr": 1.1815357923100624e-08}
{"train_loss": 0.04500740393996239, "global_step": 269575, "epoch": 3028, "lr": 1.1802758731188634e-08}
{"train_loss": 0.03267168253660202, "global_step": 269576, "epoch": 3028, "lr": 1.1790166259578739e-08}
{"train_loss": 0.061838530004024506, "global_step": 269577, "epoch": 3028, "lr": 1.1777580508282038e-08}
{"train_loss": 0.08421897143125534, "global_step": 269578, "epoch": 3028, "lr": 1.1765001477281878e-08}
{"train_loss": 0.021285902708768845, "global_step": 269579, "epoch": 3028, "lr": 1.1752429166600465e-08}
{"train_loss": 0.04645933159051484, "global_step": 269580, "epoch": 3028, "lr": 1.1739863576221144e-08, "val_loss": 9.599520683288574}
{"train_loss": 0.010824688710272312, "global_step": 269581, "epoch": 3029, "lr": 1.1727304706160569e-08}
{"train_loss": 0.04481842741370201, "global_step": 269582, "epoch": 3029, "lr": 1.1714752556413188e-08}
{"train_loss": 0.013678880408406258, "global_step": 269583, "epoch": 3029, "lr": 1.1702207126979003e-08}
{"train_loss": 0.08854062110185623, "global_step": 269584, "epoch": 3029, "lr": 1.1689668417863564e-08}
{"train_loss": 0.042433399707078934, "global_step": 269585, "epoch": 3029, "lr": 1.167713642906687e-08}
{"train_loss": 0.03376594930887222, "global_step": 269586, "epoch": 3029, "lr": 1.1664611160588923e-08}
{"train_loss": 0.025643976405262947, "global_step": 269587, "epoch": 3029, "lr": 1.1652092612435273e-08}
{"train_loss": 0.046681150794029236, "global_step": 269588, "epoch": 3029, "lr": 1.163958078460592e-08}
{"train_loss": 0.04765848070383072, "global_step": 269589, "epoch": 3029, "lr": 1.1627075677100863e-08}
{"train_loss": 0.03020324930548668, "global_step": 269590, "epoch": 3029, "lr": 1.1614577289925654e-08}
{"train_loss": 0.07594497501850128, "global_step": 269591, "epoch": 3029, "lr": 1.1602085623080294e-08}
{"train_loss": 0.062089432030916214, "global_step": 269592, "epoch": 3029, "lr": 1.1589600676564782e-08}
{"train_loss": 0.06265329569578171, "global_step": 269593, "epoch": 3029, "lr": 1.1577122450379119e-08}
{"train_loss": 0.04442138224840164, "global_step": 269594, "epoch": 3029, "lr": 1.1564650944528854e-08}
{"train_loss": 0.06500700116157532, "global_step": 269595, "epoch": 3029, "lr": 1.155218615901399e-08}
{"train_loss": 0.08010007441043854, "global_step": 269596, "epoch": 3029, "lr": 1.1539728093840074e-08}
{"train_loss": 0.045235566794872284, "global_step": 269597, "epoch": 3029, "lr": 1.152727674900156e-08}
{"train_loss": 0.020604576915502548, "global_step": 269598, "epoch": 3029, "lr": 1.1514832124503993e-08}
{"train_loss": 0.01878589577972889, "global_step": 269599, "epoch": 3029, "lr": 1.150239422035293e-08}
{"train_loss": 0.042445335537195206, "global_step": 269600, "epoch": 3029, "lr": 1.1489963036542817e-08}
{"train_loss": 0.0332622192800045, "global_step": 269601, "epoch": 3029, "lr": 1.1477538573079206e-08}
{"train_loss": 0.033599771559238434, "global_step": 269602, "epoch": 3029, "lr": 1.1465120829962096e-08}
{"train_loss": 0.03621695935726166, "global_step": 269603, "epoch": 3029, "lr": 1.1452709807197037e-08}
{"train_loss": 0.04045725613832474, "global_step": 269604, "epoch": 3029, "lr": 1.1440305504778482e-08}
{"train_loss": 0.03005000576376915, "global_step": 269605, "epoch": 3029, "lr": 1.142790792271753e-08}
{"train_loss": 0.044684991240501404, "global_step": 269606, "epoch": 3029, "lr": 1.141551706100863e-08}
{"train_loss": 0.03863067552447319, "global_step": 269607, "epoch": 3029, "lr": 1.1403132919657334e-08}
{"train_loss": 0.010723263956606388, "global_step": 269608, "epoch": 3029, "lr": 1.139075549865809e-08}
{"train_loss": 0.017329970374703407, "global_step": 269609, "epoch": 3029, "lr": 1.1378384798022002e-08}
{"train_loss": 0.022284355014562607, "global_step": 269610, "epoch": 3029, "lr": 1.1366020817749068e-08}
{"train_loss": 0.01650288701057434, "global_step": 269611, "epoch": 3029, "lr": 1.1353663557833738e-08}
{"train_loss": 0.06874038279056549, "global_step": 269612, "epoch": 3029, "lr": 1.1341313018287115e-08}
{"train_loss": 0.0410824790596962, "global_step": 269613, "epoch": 3029, "lr": 1.1328969199103646e-08}
{"train_loss": 0.062089305371046066, "global_step": 269614, "epoch": 3029, "lr": 1.1316632100288883e-08}
{"train_loss": 0.03351971507072449, "global_step": 269615, "epoch": 3029, "lr": 1.1304301721842824e-08}
{"train_loss": 0.03460663929581642, "global_step": 269616, "epoch": 3029, "lr": 1.1291978063771025e-08}
{"train_loss": 0.030369877815246582, "global_step": 269617, "epoch": 3029, "lr": 1.127966112606793e-08}
{"train_loss": 0.03233284875750542, "global_step": 269618, "epoch": 3029, "lr": 1.1267350908739094e-08}
{"train_loss": 0.07461606711149216, "global_step": 269619, "epoch": 3029, "lr": 1.1255047411784514e-08}
{"train_loss": 0.02465740032494068, "global_step": 269620, "epoch": 3029, "lr": 1.1242750635209742e-08}
{"train_loss": 0.03407532721757889, "global_step": 269621, "epoch": 3029, "lr": 1.1230460579014778e-08}
{"train_loss": 0.0631476566195488, "global_step": 269622, "epoch": 3029, "lr": 1.1218177243199623e-08}
{"train_loss": 0.050786081701517105, "global_step": 269623, "epoch": 3029, "lr": 1.1205900627764277e-08}
{"train_loss": 0.06641331315040588, "global_step": 269624, "epoch": 3029, "lr": 1.119363073271984e-08}
{"train_loss": 0.04223434999585152, "global_step": 269625, "epoch": 3029, "lr": 1.1181367558055212e-08}
{"train_loss": 0.02816808968782425, "global_step": 269626, "epoch": 3029, "lr": 1.1169111103781494e-08}
{"train_loss": 0.03250175714492798, "global_step": 269627, "epoch": 3029, "lr": 1.1156861369893134e-08}
{"train_loss": 0.07415807247161865, "global_step": 269628, "epoch": 3029, "lr": 1.1144618356395686e-08}
{"train_loss": 0.03146718442440033, "global_step": 269629, "epoch": 3029, "lr": 1.11323820632947e-08}
{"train_loss": 0.06787067651748657, "global_step": 269630, "epoch": 3029, "lr": 1.1120152490584624e-08}
{"train_loss": 0.037749744951725006, "global_step": 269631, "epoch": 3029, "lr": 1.1107929638271008e-08}
{"train_loss": 0.027310583740472794, "global_step": 269632, "epoch": 3029, "lr": 1.1095713506353856e-08}
{"train_loss": 0.0201340913772583, "global_step": 269633, "epoch": 3029, "lr": 1.1083504094833164e-08}
{"train_loss": 0.04692781716585159, "global_step": 269634, "epoch": 3029, "lr": 1.1071301403720035e-08}
{"train_loss": 0.045372139662504196, "global_step": 269635, "epoch": 3029, "lr": 1.1059105433003369e-08}
{"train_loss": 0.02822369895875454, "global_step": 269636, "epoch": 3029, "lr": 1.1046916182694268e-08}
{"train_loss": 0.018328305333852768, "global_step": 269637, "epoch": 3029, "lr": 1.1034733652787176e-08}
{"train_loss": 0.035633936524391174, "global_step": 269638, "epoch": 3029, "lr": 1.102255784328765e-08}
{"train_loss": 0.020442906767129898, "global_step": 269639, "epoch": 3029, "lr": 1.101038875420124e-08}
{"train_loss": 0.0520441010594368, "global_step": 269640, "epoch": 3029, "lr": 1.0998226385516841e-08}
{"train_loss": 0.03218066319823265, "global_step": 269641, "epoch": 3029, "lr": 1.0986070737251109e-08}
{"train_loss": 0.04569265618920326, "global_step": 269642, "epoch": 3029, "lr": 1.097392180939849e-08}
{"train_loss": 0.02123303897678852, "global_step": 269643, "epoch": 3029, "lr": 1.0961779601958989e-08}
{"train_loss": 0.022866373881697655, "global_step": 269644, "epoch": 3029, "lr": 1.0949644114938152e-08}
{"train_loss": 0.031004466116428375, "global_step": 269645, "epoch": 3029, "lr": 1.0937515348335981e-08}
{"train_loss": 0.04367133975028992, "global_step": 269646, "epoch": 3029, "lr": 1.0925393302152476e-08}
{"train_loss": 0.0600283145904541, "global_step": 269647, "epoch": 3029, "lr": 1.0913277976393188e-08}
{"train_loss": 0.07066456973552704, "global_step": 269648, "epoch": 3029, "lr": 1.0901169371052567e-08}
{"train_loss": 0.024896739050745964, "global_step": 269649, "epoch": 3029, "lr": 1.0889067486141713e-08}
{"train_loss": 0.0392041951417923, "global_step": 269650, "epoch": 3029, "lr": 1.0876972321655077e-08}
{"train_loss": 0.03016371838748455, "global_step": 269651, "epoch": 3029, "lr": 1.0864883877598208e-08}
{"train_loss": 0.06141357868909836, "global_step": 269652, "epoch": 3029, "lr": 1.0852802153971109e-08}
{"train_loss": 0.0316632054746151, "global_step": 269653, "epoch": 3029, "lr": 1.0840727150773777e-08}
{"train_loss": 0.04495733603835106, "global_step": 269654, "epoch": 3029, "lr": 1.0828658868011765e-08}
{"train_loss": 0.026886258274316788, "global_step": 269655, "epoch": 3029, "lr": 1.081659730567952e-08}
{"train_loss": 0.03191626816987991, "global_step": 269656, "epoch": 3029, "lr": 1.0804542463788148e-08}
{"train_loss": 0.019669372588396072, "global_step": 269657, "epoch": 3029, "lr": 1.0792494342337645e-08}
{"train_loss": 0.026264794170856476, "global_step": 269658, "epoch": 3029, "lr": 1.0780452941322461e-08}
{"train_loss": 0.04065890237689018, "global_step": 269659, "epoch": 3029, "lr": 1.0768418260748148e-08}
{"train_loss": 0.021229540929198265, "global_step": 269660, "epoch": 3029, "lr": 1.0756390300620256e-08}
{"train_loss": 0.04542228579521179, "global_step": 269661, "epoch": 3029, "lr": 1.0744369060933234e-08}
{"train_loss": 0.04907098785042763, "global_step": 269662, "epoch": 3029, "lr": 1.0732354541692634e-08}
{"train_loss": 0.02978314273059368, "global_step": 269663, "epoch": 3029, "lr": 1.0720346742904008e-08}
{"train_loss": 0.04617638885974884, "global_step": 269664, "epoch": 3029, "lr": 1.0708345664561802e-08}
{"train_loss": 0.07580477744340897, "global_step": 269665, "epoch": 3029, "lr": 1.069635130667157e-08}
{"train_loss": 0.08150919526815414, "global_step": 269666, "epoch": 3029, "lr": 1.068436366923331e-08}
{"train_loss": 0.06223275884985924, "global_step": 269667, "epoch": 3029, "lr": 1.0672382752247023e-08}
{"train_loss": 0.08346552401781082, "global_step": 269668, "epoch": 3029, "lr": 1.066040855571826e-08}
{"train_loss": 0.04166695146999332, "global_step": 269669, "epoch": 3029, "lr": 1.064844107964702e-08, "val_loss": 9.652119636535645}
{"train_loss": 0.06485410034656525, "global_step": 269670, "epoch": 3030, "lr": 1.0636480324033305e-08}
{"train_loss": 0.05457790568470955, "global_step": 269671, "epoch": 3030, "lr": 1.0624526288882663e-08}
{"train_loss": 0.05606885999441147, "global_step": 269672, "epoch": 3030, "lr": 1.0612578974195097e-08}
{"train_loss": 0.056098733097314835, "global_step": 269673, "epoch": 3030, "lr": 1.0600638379970606e-08}
{"train_loss": 0.06962884217500687, "global_step": 269674, "epoch": 3030, "lr": 1.058870450620919e-08}
{"train_loss": 0.03544267266988754, "global_step": 269675, "epoch": 3030, "lr": 1.0576777352916401e-08}
{"train_loss": 0.0670715868473053, "global_step": 269676, "epoch": 3030, "lr": 1.0564856920092237e-08}
{"train_loss": 0.03549161180853844, "global_step": 269677, "epoch": 3030, "lr": 1.055294320774225e-08}
{"train_loss": 0.03647245466709137, "global_step": 269678, "epoch": 3030, "lr": 1.0541036215855338e-08}
{"train_loss": 0.039967019110918045, "global_step": 269679, "epoch": 3030, "lr": 1.0529135944448155e-08}
{"train_loss": 0.030960535630583763, "global_step": 269680, "epoch": 3030, "lr": 1.0517242393515148e-08}
{"train_loss": 0.06131378188729286, "global_step": 269681, "epoch": 3030, "lr": 1.0505355563056318e-08}
{"train_loss": 0.033514466136693954, "global_step": 269682, "epoch": 3030, "lr": 1.0493475453082768e-08}
{"train_loss": 0.05773962289094925, "global_step": 269683, "epoch": 3030, "lr": 1.0481602063583396e-08}
{"train_loss": 0.0376887321472168, "global_step": 269684, "epoch": 3030, "lr": 1.0469735394563751e-08}
{"train_loss": 0.061395592987537384, "global_step": 269685, "epoch": 3030, "lr": 1.0457875446029387e-08}
{"train_loss": 0.05813111364841461, "global_step": 269686, "epoch": 3030, "lr": 1.04460222179803e-08}
{"train_loss": 0.04388835281133652, "global_step": 269687, "epoch": 3030, "lr": 1.0434175710416494e-08}
{"train_loss": 0.023908540606498718, "global_step": 269688, "epoch": 3030, "lr": 1.0422335923343518e-08}
{"train_loss": 0.02973932959139347, "global_step": 269689, "epoch": 3030, "lr": 1.0410502856755822e-08}
{"train_loss": 0.04208452254533768, "global_step": 269690, "epoch": 3030, "lr": 1.0398676510658956e-08}
{"train_loss": 0.04128771275281906, "global_step": 269691, "epoch": 3030, "lr": 1.0386856885058471e-08}
{"train_loss": 0.06431907415390015, "global_step": 269692, "epoch": 3030, "lr": 1.0375043979948817e-08}
{"train_loss": 0.034748323261737823, "global_step": 269693, "epoch": 3030, "lr": 1.0363237795341097e-08}
{"train_loss": 0.07159353792667389, "global_step": 269694, "epoch": 3030, "lr": 1.0351438331224206e-08}
{"train_loss": 0.07627514004707336, "global_step": 269695, "epoch": 3030, "lr": 1.0339645587609248e-08}
{"train_loss": 0.041980765759944916, "global_step": 269696, "epoch": 3030, "lr": 1.0327859564490671e-08}
{"train_loss": 0.014869049191474915, "global_step": 269697, "epoch": 3030, "lr": 1.0316080261879579e-08}
{"train_loss": 0.050954580307006836, "global_step": 269698, "epoch": 3030, "lr": 1.030430767977042e-08}
{"train_loss": 0.03925340250134468, "global_step": 269699, "epoch": 3030, "lr": 1.0292541818163193e-08}
{"train_loss": 0.038649529218673706, "global_step": 269700, "epoch": 3030, "lr": 1.0280782677069001e-08}
{"train_loss": 0.04255257546901703, "global_step": 269701, "epoch": 3030, "lr": 1.0269030256476741e-08}
{"train_loss": 0.0619010329246521, "global_step": 269702, "epoch": 3030, "lr": 1.0257284556397517e-08}
{"train_loss": 0.038961660116910934, "global_step": 269703, "epoch": 3030, "lr": 1.0245545576831328e-08}
{"train_loss": 0.05921453237533569, "global_step": 269704, "epoch": 3030, "lr": 1.0233813317778174e-08}
{"train_loss": 0.05788663402199745, "global_step": 269705, "epoch": 3030, "lr": 1.0222087779238055e-08}
{"train_loss": 0.05159362778067589, "global_step": 269706, "epoch": 3030, "lr": 1.021036896121652e-08}
{"train_loss": 0.04707714170217514, "global_step": 269707, "epoch": 3030, "lr": 1.0198656863708022e-08}
{"train_loss": 0.002399310702458024, "global_step": 269708, "epoch": 3030, "lr": 1.0186951486723662e-08}
{"train_loss": 0.05688167363405228, "global_step": 269709, "epoch": 3030, "lr": 1.0175252830263438e-08}
{"train_loss": 0.036298807710409164, "global_step": 269710, "epoch": 3030, "lr": 1.01635608943218e-08}
{"train_loss": 0.0914028137922287, "global_step": 269711, "epoch": 3030, "lr": 1.0151875678898747e-08}
{"train_loss": 0.07728466391563416, "global_step": 269712, "epoch": 3030, "lr": 1.0140197184010935e-08}
{"train_loss": 0.028215032070875168, "global_step": 269713, "epoch": 3030, "lr": 1.012852540964726e-08}
{"train_loss": 0.05805421620607376, "global_step": 269714, "epoch": 3030, "lr": 1.0116860355807723e-08}
{"train_loss": 0.031823012977838516, "global_step": 269715, "epoch": 3030, "lr": 1.0105202022503424e-08}
{"train_loss": 0.03882969915866852, "global_step": 269716, "epoch": 3030, "lr": 1.0093550409734366e-08}
{"train_loss": 0.0737457349896431, "global_step": 269717, "epoch": 3030, "lr": 1.0081905517494994e-08}
{"train_loss": 0.04251273721456528, "global_step": 269718, "epoch": 3030, "lr": 1.0070267345790862e-08}
{"train_loss": 0.02514839358627796, "global_step": 269719, "epoch": 3030, "lr": 1.005863589462197e-08}
{"train_loss": 0.05687371641397476, "global_step": 269720, "epoch": 3030, "lr": 1.004701116399387e-08}
{"train_loss": 0.06624740362167358, "global_step": 269721, "epoch": 3030, "lr": 1.0035393153901008e-08}
{"train_loss": 0.039815306663513184, "global_step": 269722, "epoch": 3030, "lr": 1.0023781864354487e-08}
{"train_loss": 0.04596526548266411, "global_step": 269723, "epoch": 3030, "lr": 1.0012177295348758e-08}
{"train_loss": 0.0427335686981678, "global_step": 269724, "epoch": 3030, "lr": 1.000057944688937e-08}
{"train_loss": 0.04451579973101616, "global_step": 269725, "epoch": 3030, "lr": 9.988988318976323e-09}
{"train_loss": 0.05627349764108658, "global_step": 269726, "epoch": 3030, "lr": 9.977403911609618e-09}
{"train_loss": 0.029170576483011246, "global_step": 269727, "epoch": 3030, "lr": 9.965826224794806e-09}
{"train_loss": 0.06057349592447281, "global_step": 269728, "epoch": 3030, "lr": 9.954255258526335e-09}
{"train_loss": 0.03774646297097206, "global_step": 269729, "epoch": 3030, "lr": 9.942691012809758e-09}
{"train_loss": 0.04185716062784195, "global_step": 269730, "epoch": 3030, "lr": 9.931133487650623e-09}
{"train_loss": 0.0284368097782135, "global_step": 269731, "epoch": 3030, "lr": 9.919582683048933e-09}
{"train_loss": 0.054788243025541306, "global_step": 269732, "epoch": 3030, "lr": 9.908038598999137e-09}
{"train_loss": 0.020987380295991898, "global_step": 269733, "epoch": 3030, "lr": 9.896501235512335e-09}
{"train_loss": 0.049374811351299286, "global_step": 269734, "epoch": 3030, "lr": 9.884970592582977e-09}
{"train_loss": 0.0804157704114914, "global_step": 269735, "epoch": 3030, "lr": 9.873446670216613e-09}
{"train_loss": 0.024792863056063652, "global_step": 269736, "epoch": 3030, "lr": 9.861929468413245e-09}
{"train_loss": 0.03674232214689255, "global_step": 269737, "epoch": 3030, "lr": 9.85041898717287e-09}
{"train_loss": 0.07286933064460754, "global_step": 269738, "epoch": 3030, "lr": 9.838915226501044e-09}
{"train_loss": 0.03955743834376335, "global_step": 269739, "epoch": 3030, "lr": 9.827418186392212e-09}
{"train_loss": 0.02785331755876541, "global_step": 269740, "epoch": 3030, "lr": 9.815927866857477e-09}
{"train_loss": 0.040890324860811234, "global_step": 269741, "epoch": 3030, "lr": 9.804444267891288e-09}
{"train_loss": 0.04839538782835007, "global_step": 269742, "epoch": 3030, "lr": 9.792967389499197e-09}
{"train_loss": 0.06488409638404846, "global_step": 269743, "epoch": 3030, "lr": 9.781497231681202e-09}
{"train_loss": 0.04193639010190964, "global_step": 269744, "epoch": 3030, "lr": 9.770033794442857e-09}
{"train_loss": 0.06897728145122528, "global_step": 269745, "epoch": 3030, "lr": 9.758577077773057e-09}
{"train_loss": 0.05018585920333862, "global_step": 269746, "epoch": 3030, "lr": 9.747127081688457e-09}
{"train_loss": 0.052120886743068695, "global_step": 269747, "epoch": 3030, "lr": 9.735683806183505e-09}
{"train_loss": 0.03095470555126667, "global_step": 269748, "epoch": 3030, "lr": 9.7242472512582e-09}
{"train_loss": 0.07192912697792053, "global_step": 269749, "epoch": 3030, "lr": 9.712817416918097e-09}
{"train_loss": 0.019652416929602623, "global_step": 269750, "epoch": 3030, "lr": 9.70139430315764e-09}
{"train_loss": 0.02070913277566433, "global_step": 269751, "epoch": 3030, "lr": 9.689977909987935e-09}
{"train_loss": 0.10613179951906204, "global_step": 269752, "epoch": 3030, "lr": 9.67856823740898e-09}
{"train_loss": 0.03327794373035431, "global_step": 269753, "epoch": 3030, "lr": 9.667165285415225e-09}
{"train_loss": 0.060963500291109085, "global_step": 269754, "epoch": 3030, "lr": 9.65576905401222e-09}
{"train_loss": 0.017723005264997482, "global_step": 269755, "epoch": 3030, "lr": 9.644379543199966e-09}
{"train_loss": 0.07184910774230957, "global_step": 269756, "epoch": 3030, "lr": 9.632996752984013e-09}
{"train_loss": 0.05357791855931282, "global_step": 269757, "epoch": 3030, "lr": 9.621620683364362e-09}
{"train_loss": 0.04831738404019244, "global_step": 269758, "epoch": 3030, "lr": 9.610251334341014e-09, "val_loss": 9.613892555236816, "train_action_mse_error": 13.22599983215332}
{"train_loss": 0.03368408605456352, "global_step": 269759, "epoch": 3031, "lr": 9.598888705919518e-09}
{"train_loss": 0.05650961771607399, "global_step": 269760, "epoch": 3031, "lr": 9.587532798094323e-09}
{"train_loss": 0.026654640212655067, "global_step": 269761, "epoch": 3031, "lr": 9.576183610870981e-09}
{"train_loss": 0.04367867484688759, "global_step": 269762, "epoch": 3031, "lr": 9.564841144255043e-09}
{"train_loss": 0.052518993616104126, "global_step": 269763, "epoch": 3031, "lr": 9.553505398235407e-09}
{"train_loss": 0.039400555193424225, "global_step": 269764, "epoch": 3031, "lr": 9.542176372828727e-09}
{"train_loss": 0.01816515624523163, "global_step": 269765, "epoch": 3031, "lr": 9.530854068029449e-09}
{"train_loss": 0.04349318519234657, "global_step": 269766, "epoch": 3031, "lr": 9.519538483837576e-09}
{"train_loss": 0.06011589616537094, "global_step": 269767, "epoch": 3031, "lr": 9.508229620258658e-09}
{"train_loss": 0.0518677718937397, "global_step": 269768, "epoch": 3031, "lr": 9.496927477287143e-09}
{"train_loss": 0.03496905416250229, "global_step": 269769, "epoch": 3031, "lr": 9.485632054934135e-09}
{"train_loss": 0.05183994770050049, "global_step": 269770, "epoch": 3031, "lr": 9.474343353194082e-09}
{"train_loss": 0.04861694946885109, "global_step": 269771, "epoch": 3031, "lr": 9.463061372072534e-09}
{"train_loss": 0.04849854111671448, "global_step": 269772, "epoch": 3031, "lr": 9.451786111569494e-09}
{"train_loss": 0.06864968687295914, "global_step": 269773, "epoch": 3031, "lr": 9.440517571684959e-09}
{"train_loss": 0.02449013851583004, "global_step": 269774, "epoch": 3031, "lr": 9.429255752424482e-09}
{"train_loss": 0.021425796672701836, "global_step": 269775, "epoch": 3031, "lr": 9.41800065378251e-09}
{"train_loss": 0.05393227934837341, "global_step": 269776, "epoch": 3031, "lr": 9.406752275770148e-09}
{"train_loss": 0.019045483320951462, "global_step": 269777, "epoch": 3031, "lr": 9.39551061837629e-09}
{"train_loss": 0.027161192148923874, "global_step": 269778, "epoch": 3031, "lr": 9.384275681617593e-09}
{"train_loss": 0.07886803150177002, "global_step": 269779, "epoch": 3031, "lr": 9.373047465482953e-09}
{"train_loss": 0.060213737189769745, "global_step": 269780, "epoch": 3031, "lr": 9.361825969983474e-09}
{"train_loss": 0.0506221167743206, "global_step": 269781, "epoch": 3031, "lr": 9.350611195113602e-09}
{"train_loss": 0.06620371341705322, "global_step": 269782, "epoch": 3031, "lr": 9.339403140873338e-09}
{"train_loss": 0.050102028995752335, "global_step": 269783, "epoch": 3031, "lr": 9.328201807273784e-09}
{"train_loss": 0.04758790507912636, "global_step": 269784, "epoch": 3031, "lr": 9.31700719430939e-09}
{"train_loss": 0.056095581501722336, "global_step": 269785, "epoch": 3031, "lr": 9.305819301980157e-09}
{"train_loss": 0.05393614247441292, "global_step": 269786, "epoch": 3031, "lr": 9.294638130291632e-09}
{"train_loss": 0.006761814001947641, "global_step": 269787, "epoch": 3031, "lr": 9.283463679249372e-09}
{"train_loss": 0.06532879918813705, "global_step": 269788, "epoch": 3031, "lr": 9.27229594884227e-09}
{"train_loss": 0.014343117363750935, "global_step": 269789, "epoch": 3031, "lr": 9.261134939081428e-09}
{"train_loss": 0.04681025445461273, "global_step": 269790, "epoch": 3031, "lr": 9.24998064996685e-09}
{"train_loss": 0.049316685646772385, "global_step": 269791, "epoch": 3031, "lr": 9.238833081498533e-09}
{"train_loss": 0.019719058647751808, "global_step": 269792, "epoch": 3031, "lr": 9.227692233682028e-09}
{"train_loss": 0.04352254047989845, "global_step": 269793, "epoch": 3031, "lr": 9.216558106511786e-09}
{"train_loss": 0.06099740043282509, "global_step": 269794, "epoch": 3031, "lr": 9.205430699998907e-09}
{"train_loss": 0.05565086379647255, "global_step": 269795, "epoch": 3031, "lr": 9.19431001413229e-09}
{"train_loss": 0.05581791326403618, "global_step": 269796, "epoch": 3031, "lr": 9.183196048923038e-09}
{"train_loss": 0.021793916821479797, "global_step": 269797, "epoch": 3031, "lr": 9.172088804371148e-09}
{"train_loss": 0.026628462597727776, "global_step": 269798, "epoch": 3031, "lr": 9.160988280471073e-09}
{"train_loss": 0.048338331282138824, "global_step": 269799, "epoch": 3031, "lr": 9.149894477233912e-09}
{"train_loss": 0.09828732162714005, "global_step": 269800, "epoch": 3031, "lr": 9.138807394659666e-09}
{"train_loss": 0.023929215967655182, "global_step": 269801, "epoch": 3031, "lr": 9.127727032742783e-09}
{"train_loss": 0.06402867287397385, "global_step": 269802, "epoch": 3031, "lr": 9.116653391494367e-09}
{"train_loss": 0.052742715924978256, "global_step": 269803, "epoch": 3031, "lr": 9.105586470903316e-09}
{"train_loss": 0.060713883489370346, "global_step": 269804, "epoch": 3031, "lr": 9.09452627098628e-09}
{"train_loss": 0.06072487309575081, "global_step": 269805, "epoch": 3031, "lr": 9.083472791732162e-09}
{"train_loss": 0.02703743614256382, "global_step": 269806, "epoch": 3031, "lr": 9.072426033152059e-09}
{"train_loss": 0.058799199759960175, "global_step": 269807, "epoch": 3031, "lr": 9.061385995240424e-09}
{"train_loss": 0.03075045347213745, "global_step": 269808, "epoch": 3031, "lr": 9.050352678002804e-09}
{"train_loss": 0.0305673535913229, "global_step": 269809, "epoch": 3031, "lr": 9.039326081433652e-09}
{"train_loss": 0.03610147908329964, "global_step": 269810, "epoch": 3031, "lr": 9.028306205544069e-09}
{"train_loss": 0.05306950584053993, "global_step": 269811, "epoch": 3031, "lr": 9.017293050334052e-09}
{"train_loss": 0.023773757740855217, "global_step": 269812, "epoch": 3031, "lr": 9.006286615798055e-09}
{"train_loss": 0.024913577362895012, "global_step": 269813, "epoch": 3031, "lr": 8.995286901941624e-09}
{"train_loss": 0.08056919276714325, "global_step": 269814, "epoch": 3031, "lr": 8.984293908770314e-09}
{"train_loss": 0.020329657942056656, "global_step": 269815, "epoch": 3031, "lr": 8.973307636278572e-09}
{"train_loss": 0.028281064704060555, "global_step": 269816, "epoch": 3031, "lr": 8.962328084471949e-09}
{"train_loss": 0.05701804906129837, "global_step": 269817, "epoch": 3031, "lr": 8.951355253350446e-09}
{"train_loss": 0.07632257044315338, "global_step": 269818, "epoch": 3031, "lr": 8.940389142914063e-09}
{"train_loss": 0.01684013567864895, "global_step": 269819, "epoch": 3031, "lr": 8.929429753173901e-09}
{"train_loss": 0.023672472685575485, "global_step": 269820, "epoch": 3031, "lr": 8.918477084113309e-09}
{"train_loss": 0.026219680905342102, "global_step": 269821, "epoch": 3031, "lr": 8.907531135754487e-09}
{"train_loss": 0.0555206723511219, "global_step": 269822, "epoch": 3031, "lr": 8.896591908080786e-09}
{"train_loss": 0.03505602106451988, "global_step": 269823, "epoch": 3031, "lr": 8.885659401108859e-09}
{"train_loss": 0.04640771448612213, "global_step": 269824, "epoch": 3031, "lr": 8.874733614827602e-09}
{"train_loss": 0.030822530388832092, "global_step": 269825, "epoch": 3031, "lr": 8.863814549242566e-09}
{"train_loss": 0.04126561060547829, "global_step": 269826, "epoch": 3031, "lr": 8.852902204359303e-09}
{"train_loss": 0.051424719393253326, "global_step": 269827, "epoch": 3031, "lr": 8.841996580177814e-09}
{"train_loss": 0.030700678005814552, "global_step": 269828, "epoch": 3031, "lr": 8.831097676698097e-09}
{"train_loss": 0.018265116959810257, "global_step": 269829, "epoch": 3031, "lr": 8.820205493920153e-09}
{"train_loss": 0.023678790777921677, "global_step": 269830, "epoch": 3031, "lr": 8.809320031849533e-09}
{"train_loss": 0.045373234897851944, "global_step": 269831, "epoch": 3031, "lr": 8.798441290480685e-09}
{"train_loss": 0.014528116211295128, "global_step": 269832, "epoch": 3031, "lr": 8.787569269819163e-09}
{"train_loss": 0.03541043773293495, "global_step": 269833, "epoch": 3031, "lr": 8.776703969870515e-09}
{"train_loss": 0.03561567887663841, "global_step": 269834, "epoch": 3031, "lr": 8.765845390634741e-09}
{"train_loss": 0.0357663631439209, "global_step": 269835, "epoch": 3031, "lr": 8.754993532106293e-09}
{"train_loss": 0.024466589093208313, "global_step": 269836, "epoch": 3031, "lr": 8.744148394290718e-09}
{"train_loss": 0.043495792895555496, "global_step": 269837, "epoch": 3031, "lr": 8.73330997719357e-09}
{"train_loss": 0.08664847165346146, "global_step": 269838, "epoch": 3031, "lr": 8.72247828081485e-09}
{"train_loss": 0.04515934735536575, "global_step": 269839, "epoch": 3031, "lr": 8.711653305149003e-09}
{"train_loss": 0.06282753497362137, "global_step": 269840, "epoch": 3031, "lr": 8.700835050201583e-09}
{"train_loss": 0.033600617200136185, "global_step": 269841, "epoch": 3031, "lr": 8.690023515978141e-09}
{"train_loss": 0.055147286504507065, "global_step": 269842, "epoch": 3031, "lr": 8.679218702478675e-09}
{"train_loss": 0.061612799763679504, "global_step": 269843, "epoch": 3031, "lr": 8.668420609703188e-09}
{"train_loss": 0.06031738594174385, "global_step": 269844, "epoch": 3031, "lr": 8.657629237651676e-09}
{"train_loss": 0.05800875648856163, "global_step": 269845, "epoch": 3031, "lr": 8.646844586324143e-09}
{"train_loss": 0.0445873960852623, "global_step": 269846, "epoch": 3031, "lr": 8.636066655726139e-09}
{"train_loss": 0.04345101310107648, "global_step": 269847, "epoch": 3031, "lr": 8.625295445857662e-09, "val_loss": 9.629548072814941}
{"train_loss": 0.0587802529335022, "global_step": 269848, "epoch": 3032, "lr": 8.614530956724265e-09}
{"train_loss": 0.029530560597777367, "global_step": 269849, "epoch": 3032, "lr": 8.603773188320396e-09}
{"train_loss": 0.03797160089015961, "global_step": 269850, "epoch": 3032, "lr": 8.593022140646057e-09}
{"train_loss": 0.06511736661195755, "global_step": 269851, "epoch": 3032, "lr": 8.582277813712348e-09}
{"train_loss": 0.03569452837109566, "global_step": 269852, "epoch": 3032, "lr": 8.571540207513718e-09}
{"train_loss": 0.010639016516506672, "global_step": 269853, "epoch": 3032, "lr": 8.560809322055719e-09}
{"train_loss": 0.04293452203273773, "global_step": 269854, "epoch": 3032, "lr": 8.5500851573328e-09}
{"train_loss": 0.0374685674905777, "global_step": 269855, "epoch": 3032, "lr": 8.539367713356061e-09}
{"train_loss": 0.04365390166640282, "global_step": 269856, "epoch": 3032, "lr": 8.528656990119954e-09}
{"train_loss": 0.057849232107400894, "global_step": 269857, "epoch": 3032, "lr": 8.517952987624479e-09}
{"train_loss": 0.03049444407224655, "global_step": 269858, "epoch": 3032, "lr": 8.507255705875183e-09}
{"train_loss": 0.049487460404634476, "global_step": 269859, "epoch": 3032, "lr": 8.496565144877622e-09}
{"train_loss": 0.0361674465239048, "global_step": 269860, "epoch": 3032, "lr": 8.485881304626242e-09}
{"train_loss": 0.06889832764863968, "global_step": 269861, "epoch": 3032, "lr": 8.475204185121044e-09}
{"train_loss": 0.06137251481413841, "global_step": 269862, "epoch": 3032, "lr": 8.46453378637313e-09}
{"train_loss": 0.032175179570913315, "global_step": 269863, "epoch": 3032, "lr": 8.453870108371398e-09}
{"train_loss": 0.04899930953979492, "global_step": 269864, "epoch": 3032, "lr": 8.44321315112695e-09}
{"train_loss": 0.024327881634235382, "global_step": 269865, "epoch": 3032, "lr": 8.432562914639786e-09}
{"train_loss": 0.03088013269007206, "global_step": 269866, "epoch": 3032, "lr": 8.421919398909905e-09}
{"train_loss": 0.05232999473810196, "global_step": 269867, "epoch": 3032, "lr": 8.41128260393731e-09}
{"train_loss": 0.020633505657315254, "global_step": 269868, "epoch": 3032, "lr": 8.400652529721998e-09}
{"train_loss": 0.05219583213329315, "global_step": 269869, "epoch": 3032, "lr": 8.39002917626952e-09}
{"train_loss": 0.0953981876373291, "global_step": 269870, "epoch": 3032, "lr": 8.379412543579879e-09}
{"train_loss": 0.012700824998319149, "global_step": 269871, "epoch": 3032, "lr": 8.368802631653072e-09}
{"train_loss": 0.10882805287837982, "global_step": 269872, "epoch": 3032, "lr": 8.358199440494652e-09}
{"train_loss": 0.03252626582980156, "global_step": 269873, "epoch": 3032, "lr": 8.347602970099066e-09}
{"train_loss": 0.04929478093981743, "global_step": 269874, "epoch": 3032, "lr": 8.337013220477418e-09}
{"train_loss": 0.07972817122936249, "global_step": 269875, "epoch": 3032, "lr": 8.326430191618607e-09}
{"train_loss": 0.031041935086250305, "global_step": 269876, "epoch": 3032, "lr": 8.31585388353373e-09}
{"train_loss": 0.03887464106082916, "global_step": 269877, "epoch": 3032, "lr": 8.305284296222792e-09}
{"train_loss": 0.06134990602731705, "global_step": 269878, "epoch": 3032, "lr": 8.294721429685793e-09}
{"train_loss": 0.035545967519283295, "global_step": 269879, "epoch": 3032, "lr": 8.284165283922729e-09}
{"train_loss": 0.04086615517735481, "global_step": 269880, "epoch": 3032, "lr": 8.273615858939155e-09}
{"train_loss": 0.031157253310084343, "global_step": 269881, "epoch": 3032, "lr": 8.263073154735068e-09}
{"train_loss": 0.020695986226201057, "global_step": 269882, "epoch": 3032, "lr": 8.25253717130492e-09}
{"train_loss": 0.0171651691198349, "global_step": 269883, "epoch": 3032, "lr": 8.24200790865981e-09}
{"train_loss": 0.02035459317266941, "global_step": 269884, "epoch": 3032, "lr": 8.23148536679974e-09}
{"train_loss": 0.02647313103079796, "global_step": 269885, "epoch": 3032, "lr": 8.220969545719159e-09}
{"train_loss": 0.037825245410203934, "global_step": 269886, "epoch": 3032, "lr": 8.210460445423617e-09}
{"train_loss": 0.013862153515219688, "global_step": 269887, "epoch": 3032, "lr": 8.199958065918667e-09}
{"train_loss": 0.04189015552401543, "global_step": 269888, "epoch": 3032, "lr": 8.189462407198756e-09}
{"train_loss": 0.027274737134575844, "global_step": 269889, "epoch": 3032, "lr": 8.178973469269436e-09}
{"train_loss": 0.04553590342402458, "global_step": 269890, "epoch": 3032, "lr": 8.168491252130706e-09}
{"train_loss": 0.062181100249290466, "global_step": 269891, "epoch": 3032, "lr": 8.158015755782567e-09}
{"train_loss": 0.04524564743041992, "global_step": 269892, "epoch": 3032, "lr": 8.14754698023057e-09}
{"train_loss": 0.040680211037397385, "global_step": 269893, "epoch": 3032, "lr": 8.137084925474714e-09}
{"train_loss": 0.031756043434143066, "global_step": 269894, "epoch": 3032, "lr": 8.126629591515e-09}
{"train_loss": 0.022195491939783096, "global_step": 269895, "epoch": 3032, "lr": 8.116180978351429e-09}
{"train_loss": 0.03828633204102516, "global_step": 269896, "epoch": 3032, "lr": 8.10573908598955e-09}
{"train_loss": 0.05383006110787392, "global_step": 269897, "epoch": 3032, "lr": 8.095303914429363e-09}
{"train_loss": 0.03143306076526642, "global_step": 269898, "epoch": 3032, "lr": 8.08487546366532e-09}
{"train_loss": 0.028062062337994576, "global_step": 269899, "epoch": 3032, "lr": 8.074453733714071e-09}
{"train_loss": 0.015160842798650265, "global_step": 269900, "epoch": 3032, "lr": 8.064038724558964e-09}
{"train_loss": 0.048670895397663116, "global_step": 269901, "epoch": 3032, "lr": 8.053630436216652e-09}
{"train_loss": 0.06051316112279892, "global_step": 269902, "epoch": 3032, "lr": 8.043228868681585e-09}
{"train_loss": 0.07115665823221207, "global_step": 269903, "epoch": 3032, "lr": 8.03283402195376e-09}
{"train_loss": 0.07564879953861237, "global_step": 269904, "epoch": 3032, "lr": 8.022445896038733e-09}
{"train_loss": 0.03858286142349243, "global_step": 269905, "epoch": 3032, "lr": 8.012064490930949e-09}
{"train_loss": 0.03457002714276314, "global_step": 269906, "epoch": 3032, "lr": 8.00168980664151e-09}
{"train_loss": 0.049114372581243515, "global_step": 269907, "epoch": 3032, "lr": 7.991321843164867e-09}
{"train_loss": 0.04457340016961098, "global_step": 269908, "epoch": 3032, "lr": 7.980960600506572e-09}
{"train_loss": 0.017085615545511246, "global_step": 269909, "epoch": 3032, "lr": 7.970606078661069e-09}
{"train_loss": 0.039473436772823334, "global_step": 269910, "epoch": 3032, "lr": 7.960258277639465e-09}
{"train_loss": 0.06969400495290756, "global_step": 269911, "epoch": 3032, "lr": 7.949917197436207e-09}
{"train_loss": 0.04453440383076668, "global_step": 269912, "epoch": 3032, "lr": 7.939582838056848e-09}
{"train_loss": 0.06623842567205429, "global_step": 269913, "epoch": 3032, "lr": 7.929255199501383e-09}
{"train_loss": 0.046243175864219666, "global_step": 269914, "epoch": 3032, "lr": 7.918934281764268e-09}
{"train_loss": 0.060853876173496246, "global_step": 269915, "epoch": 3032, "lr": 7.908620084856599e-09}
{"train_loss": 0.04590393230319023, "global_step": 269916, "epoch": 3032, "lr": 7.89831260877838e-09}
{"train_loss": 0.029469603672623634, "global_step": 269917, "epoch": 3032, "lr": 7.888011853529608e-09}
{"train_loss": 0.02789805643260479, "global_step": 269918, "epoch": 3032, "lr": 7.877717819104736e-09}
{"train_loss": 0.06582764536142349, "global_step": 269919, "epoch": 3032, "lr": 7.867430505514861e-09}
{"train_loss": 0.019246965646743774, "global_step": 269920, "epoch": 3032, "lr": 7.857149912759987e-09}
{"train_loss": 0.034869637340307236, "global_step": 269921, "epoch": 3032, "lr": 7.846876040840111e-09}
{"train_loss": 0.020247630774974823, "global_step": 269922, "epoch": 3032, "lr": 7.836608889749686e-09}
{"train_loss": 0.05224641412496567, "global_step": 269923, "epoch": 3032, "lr": 7.82634845949981e-09}
{"train_loss": 0.024206077679991722, "global_step": 269924, "epoch": 3032, "lr": 7.816094750090485e-09}
{"train_loss": 0.058877021074295044, "global_step": 269925, "epoch": 3032, "lr": 7.805847761521712e-09}
{"train_loss": 0.020598534494638443, "global_step": 269926, "epoch": 3032, "lr": 7.795607493787938e-09}
{"train_loss": 0.05304199829697609, "global_step": 269927, "epoch": 3032, "lr": 7.785373946900266e-09}
{"train_loss": 0.029717203229665756, "global_step": 269928, "epoch": 3032, "lr": 7.775147120858695e-09}
{"train_loss": 0.04018038883805275, "global_step": 269929, "epoch": 3032, "lr": 7.764927015663226e-09}
{"train_loss": 0.06024368479847908, "global_step": 269930, "epoch": 3032, "lr": 7.754713631308308e-09}
{"train_loss": 0.025525256991386414, "global_step": 269931, "epoch": 3032, "lr": 7.744506967805043e-09}
{"train_loss": 0.038734082132577896, "global_step": 269932, "epoch": 3032, "lr": 7.734307025153431e-09}
{"train_loss": 0.022422773763537407, "global_step": 269933, "epoch": 3032, "lr": 7.724113803347921e-09}
{"train_loss": 0.018358342349529266, "global_step": 269934, "epoch": 3032, "lr": 7.713927302399615e-09}
{"train_loss": 0.03608429431915283, "global_step": 269935, "epoch": 3032, "lr": 7.70374752229741e-09}
{"train_loss": 0.04155051621367757, "global_step": 269936, "epoch": 3032, "lr": 7.693574463057963e-09, "val_loss": 9.608773231506348}
{"train_loss": 0.04565638303756714, "global_step": 269937, "epoch": 3033, "lr": 7.683408124670167e-09}
{"train_loss": 0.04315287247300148, "global_step": 269938, "epoch": 3033, "lr": 7.673248507139575e-09}
{"train_loss": 0.039742182940244675, "global_step": 269939, "epoch": 3033, "lr": 7.663095610471737e-09}
{"train_loss": 0.021015590056777, "global_step": 269940, "epoch": 3033, "lr": 7.652949434661106e-09}
{"train_loss": 0.0643276572227478, "global_step": 269941, "epoch": 3033, "lr": 7.642809979713228e-09}
{"train_loss": 0.03350408375263214, "global_step": 269942, "epoch": 3033, "lr": 7.632677245628106e-09}
{"train_loss": 0.03298349305987358, "global_step": 269943, "epoch": 3033, "lr": 7.62255123241129e-09}
{"train_loss": 0.0343509167432785, "global_step": 269944, "epoch": 3033, "lr": 7.61243194005723e-09}
{"train_loss": 0.06992984563112259, "global_step": 269945, "epoch": 3033, "lr": 7.602319368571476e-09}
{"train_loss": 0.030775511637330055, "global_step": 269946, "epoch": 3033, "lr": 7.592213517948477e-09}
{"train_loss": 0.0660196840763092, "global_step": 269947, "epoch": 3033, "lr": 7.582114388204886e-09}
{"train_loss": 0.048494938760995865, "global_step": 269948, "epoch": 3033, "lr": 7.572021979324052e-09}
{"train_loss": 0.033987369388341904, "global_step": 269949, "epoch": 3033, "lr": 7.561936291322624e-09}
{"train_loss": 0.035984788089990616, "global_step": 269950, "epoch": 3033, "lr": 7.551857324189504e-09}
{"train_loss": 0.025056451559066772, "global_step": 269951, "epoch": 3033, "lr": 7.541785077935793e-09}
{"train_loss": 0.04925239831209183, "global_step": 269952, "epoch": 3033, "lr": 7.531719552555938e-09}
{"train_loss": 0.0482870377600193, "global_step": 269953, "epoch": 3033, "lr": 7.521660748055492e-09}
{"train_loss": 0.08864544332027435, "global_step": 269954, "epoch": 3033, "lr": 7.511608664434455e-09}
{"train_loss": 0.05663086846470833, "global_step": 269955, "epoch": 3033, "lr": 7.501563301692826e-09}
{"train_loss": 0.039484862238168716, "global_step": 269956, "epoch": 3033, "lr": 7.491524659836157e-09}
{"train_loss": 0.050099484622478485, "global_step": 269957, "epoch": 3033, "lr": 7.481492738858898e-09}
{"train_loss": 0.05657554790377617, "global_step": 269958, "epoch": 3033, "lr": 7.471467538772147e-09}
{"train_loss": 0.06485117971897125, "global_step": 269959, "epoch": 3033, "lr": 7.461449059564807e-09}
{"train_loss": 0.030712071806192398, "global_step": 269960, "epoch": 3033, "lr": 7.451437301247976e-09}
{"train_loss": 0.02881111577153206, "global_step": 269961, "epoch": 3033, "lr": 7.441432263821657e-09}
{"train_loss": 0.03718055412173271, "global_step": 269962, "epoch": 3033, "lr": 7.431433947280297e-09}
{"train_loss": 0.030884606763720512, "global_step": 269963, "epoch": 3033, "lr": 7.421442351635e-09}
{"train_loss": 0.03391160070896149, "global_step": 269964, "epoch": 3033, "lr": 7.411457476885763e-09}
{"train_loss": 0.06079617142677307, "global_step": 269965, "epoch": 3033, "lr": 7.401479323027038e-09}
{"train_loss": 0.038441773504018784, "global_step": 269966, "epoch": 3033, "lr": 7.391507890058824e-09}
{"train_loss": 0.03459978848695755, "global_step": 269967, "epoch": 3033, "lr": 7.381543177992223e-09}
{"train_loss": 0.051924124360084534, "global_step": 269968, "epoch": 3033, "lr": 7.371585186827235e-09}
{"train_loss": 0.04586939513683319, "global_step": 269969, "epoch": 3033, "lr": 7.361633916558308e-09}
{"train_loss": 0.08774203807115555, "global_step": 269970, "epoch": 3033, "lr": 7.351689367190995e-09}
{"train_loss": 0.03314438834786415, "global_step": 269971, "epoch": 3033, "lr": 7.341751538725294e-09}
{"train_loss": 0.0613168329000473, "global_step": 269972, "epoch": 3033, "lr": 7.331820431161207e-09}
{"train_loss": 0.023391839116811752, "global_step": 269973, "epoch": 3033, "lr": 7.321896044504284e-09}
{"train_loss": 0.0390704981982708, "global_step": 269974, "epoch": 3033, "lr": 7.3119783787545245e-09}
{"train_loss": 0.06167779117822647, "global_step": 269975, "epoch": 3033, "lr": 7.30206743391193e-09}
{"train_loss": 0.05520452931523323, "global_step": 269976, "epoch": 3033, "lr": 7.292163209976499e-09}
{"train_loss": 0.06103838235139847, "global_step": 269977, "epoch": 3033, "lr": 7.282265706953784e-09}
{"train_loss": 0.06221852824091911, "global_step": 269978, "epoch": 3033, "lr": 7.272374924838232e-09}
{"train_loss": 0.026247404515743256, "global_step": 269979, "epoch": 3033, "lr": 7.262490863640947e-09}
{"train_loss": 0.09489784389734268, "global_step": 269980, "epoch": 3033, "lr": 7.252613523356377e-09}
{"train_loss": 0.053513094782829285, "global_step": 269981, "epoch": 3033, "lr": 7.242742903984523e-09}
{"train_loss": 0.07684314996004105, "global_step": 269982, "epoch": 3033, "lr": 7.2328790055309345e-09}
{"train_loss": 0.023188862949609756, "global_step": 269983, "epoch": 3033, "lr": 7.223021827995613e-09}
{"train_loss": 0.012487033382058144, "global_step": 269984, "epoch": 3033, "lr": 7.2131713713841084e-09}
{"train_loss": 0.03364552929997444, "global_step": 269985, "epoch": 3033, "lr": 7.20332763568532e-09}
{"train_loss": 0.05382855609059334, "global_step": 269986, "epoch": 3033, "lr": 7.193490620915899e-09}
{"train_loss": 0.027885332703590393, "global_step": 269987, "epoch": 3033, "lr": 7.183660327064745e-09}
{"train_loss": 0.01110669132322073, "global_step": 269988, "epoch": 3033, "lr": 7.173836754142959e-09}
{"train_loss": 0.05736835300922394, "global_step": 269989, "epoch": 3033, "lr": 7.164019902144992e-09}
{"train_loss": 0.019098598510026932, "global_step": 269990, "epoch": 3033, "lr": 7.154209771076392e-09}
{"train_loss": 0.04656057432293892, "global_step": 269991, "epoch": 3033, "lr": 7.14440636093161e-09}
{"train_loss": 0.038094662129879, "global_step": 269992, "epoch": 3033, "lr": 7.134609671721748e-09}
{"train_loss": 0.051699038594961166, "global_step": 269993, "epoch": 3033, "lr": 7.124819703441255e-09}
{"train_loss": 0.04576648771762848, "global_step": 269994, "epoch": 3033, "lr": 7.1150364560956804e-09}
{"train_loss": 0.07500022649765015, "global_step": 269995, "epoch": 3033, "lr": 7.105259929679475e-09}
{"train_loss": 0.02442355267703533, "global_step": 269996, "epoch": 3033, "lr": 7.0954901242037405e-09}
{"train_loss": 0.021044332534074783, "global_step": 269997, "epoch": 3033, "lr": 7.085727039662926e-09}
{"train_loss": 0.030451523140072823, "global_step": 269998, "epoch": 3033, "lr": 7.075970676057031e-09}
{"train_loss": 0.07541828602552414, "global_step": 269999, "epoch": 3033, "lr": 7.066221033397158e-09}
{"train_loss": 0.07785425335168839, "global_step": 270000, "epoch": 3033, "lr": 7.056478111672205e-09}
{"train_loss": 0.02035311423242092, "global_step": 270001, "epoch": 3033, "lr": 7.046741910893273e-09}
{"train_loss": 0.05554917827248573, "global_step": 270002, "epoch": 3033, "lr": 7.037012431054812e-09}
{"train_loss": 0.04867204651236534, "global_step": 270003, "epoch": 3033, "lr": 7.027289672162373e-09}
{"train_loss": 0.04795818403363228, "global_step": 270004, "epoch": 3033, "lr": 7.017573634210406e-09}
{"train_loss": 0.021648487076163292, "global_step": 270005, "epoch": 3033, "lr": 7.007864317210011e-09}
{"train_loss": 0.03872813284397125, "global_step": 270006, "epoch": 3033, "lr": 6.9981617211611895e-09}
{"train_loss": 0.0168171189725399, "global_step": 270007, "epoch": 3033, "lr": 6.98846584605839e-09}
{"train_loss": 0.0905022993683815, "global_step": 270008, "epoch": 3033, "lr": 6.978776691907163e-09}
{"train_loss": 0.05994963273406029, "global_step": 270009, "epoch": 3033, "lr": 6.969094258707509e-09}
{"train_loss": 0.049343954771757126, "global_step": 270010, "epoch": 3033, "lr": 6.95941854646498e-09}
{"train_loss": 0.09059175848960876, "global_step": 270011, "epoch": 3033, "lr": 6.949749555174024e-09}
{"train_loss": 0.06434907019138336, "global_step": 270012, "epoch": 3033, "lr": 6.940087284840191e-09}
{"train_loss": 0.025755546987056732, "global_step": 270013, "epoch": 3033, "lr": 6.930431735463483e-09}
{"train_loss": 0.017372597008943558, "global_step": 270014, "epoch": 3033, "lr": 6.920782907043899e-09}
{"train_loss": 0.05062916502356529, "global_step": 270015, "epoch": 3033, "lr": 6.91114079958699e-09}
{"train_loss": 0.0626024529337883, "global_step": 270016, "epoch": 3033, "lr": 6.901505413092757e-09}
{"train_loss": 0.025370564311742783, "global_step": 270017, "epoch": 3033, "lr": 6.891876747561199e-09}
{"train_loss": 0.026712607592344284, "global_step": 270018, "epoch": 3033, "lr": 6.882254802992316e-09}
{"train_loss": 0.054207395762205124, "global_step": 270019, "epoch": 3033, "lr": 6.872639579386109e-09}
{"train_loss": 0.06290854513645172, "global_step": 270020, "epoch": 3033, "lr": 6.8630310767481275e-09}
{"train_loss": 0.061950236558914185, "global_step": 270021, "epoch": 3033, "lr": 6.853429295078373e-09}
{"train_loss": 0.0708429291844368, "global_step": 270022, "epoch": 3033, "lr": 6.843834234382396e-09}
{"train_loss": 0.04803324490785599, "global_step": 270023, "epoch": 3033, "lr": 6.834245894649094e-09}
{"train_loss": 0.05511370673775673, "global_step": 270024, "epoch": 3033, "lr": 6.824664275895121e-09}
{"train_loss": 0.04643085334198863, "global_step": 270025, "epoch": 3033, "lr": 6.815089378109374e-09, "val_loss": 9.665003776550293}
{"train_loss": 0.03360461816191673, "global_step": 270026, "epoch": 3034, "lr": 6.805521201297405e-09}
{"train_loss": 0.05202064290642738, "global_step": 270027, "epoch": 3034, "lr": 6.795959745464764e-09}
{"train_loss": 0.04827064648270607, "global_step": 270028, "epoch": 3034, "lr": 6.786405010605901e-09}
{"train_loss": 0.03301915526390076, "global_step": 270029, "epoch": 3034, "lr": 6.7768569967263665e-09}
{"train_loss": 0.04437331482768059, "global_step": 270030, "epoch": 3034, "lr": 6.7673157038317116e-09}
{"train_loss": 0.017701322212815285, "global_step": 270031, "epoch": 3034, "lr": 6.7577811319108344e-09}
{"train_loss": 0.019308974966406822, "global_step": 270032, "epoch": 3034, "lr": 6.748253280974837e-09}
{"train_loss": 0.05915816128253937, "global_step": 270033, "epoch": 3034, "lr": 6.738732151018168e-09}
{"train_loss": 0.027526838704943657, "global_step": 270034, "epoch": 3034, "lr": 6.72921774205193e-09}
{"train_loss": 0.028632672503590584, "global_step": 270035, "epoch": 3034, "lr": 6.719710054070572e-09}
{"train_loss": 0.04059295728802681, "global_step": 270036, "epoch": 3034, "lr": 6.710209087074093e-09}
{"train_loss": 0.022908171638846397, "global_step": 270037, "epoch": 3034, "lr": 6.700714841068045e-09}
{"train_loss": 0.06393187493085861, "global_step": 270038, "epoch": 3034, "lr": 6.691227316046878e-09}
{"train_loss": 0.05634835362434387, "global_step": 270039, "epoch": 3034, "lr": 6.6817465120216914e-09}
{"train_loss": 0.04371194913983345, "global_step": 270040, "epoch": 3034, "lr": 6.672272428986937e-09}
{"train_loss": 0.039697855710983276, "global_step": 270041, "epoch": 3034, "lr": 6.662805066948164e-09}
{"train_loss": 0.08389914035797119, "global_step": 270042, "epoch": 3034, "lr": 6.653344425899821e-09}
{"train_loss": 0.01999296434223652, "global_step": 270043, "epoch": 3034, "lr": 6.643890505853012e-09}
{"train_loss": 0.011209524236619473, "global_step": 270044, "epoch": 3034, "lr": 6.634443306802185e-09}
{"train_loss": 0.04682218283414841, "global_step": 270045, "epoch": 3034, "lr": 6.62500282874734e-09}
{"train_loss": 0.008577731437981129, "global_step": 270046, "epoch": 3034, "lr": 6.615569071694028e-09}
{"train_loss": 0.06610839813947678, "global_step": 270047, "epoch": 3034, "lr": 6.6061420356422484e-09}
{"train_loss": 0.022083880379796028, "global_step": 270048, "epoch": 3034, "lr": 6.596721720592003e-09}
{"train_loss": 0.050377171486616135, "global_step": 270049, "epoch": 3034, "lr": 6.58730812654329e-09}
{"train_loss": 0.048226431012153625, "global_step": 270050, "epoch": 3034, "lr": 6.577901253507213e-09}
{"train_loss": 0.04786539822816849, "global_step": 270051, "epoch": 3034, "lr": 6.568501101472668e-09}
{"train_loss": 0.017634503543376923, "global_step": 270052, "epoch": 3034, "lr": 6.559107670445208e-09}
{"train_loss": 0.054654430598020554, "global_step": 270053, "epoch": 3034, "lr": 6.549720960424832e-09}
{"train_loss": 0.04805651307106018, "global_step": 270054, "epoch": 3034, "lr": 6.540340971417091e-09}
{"train_loss": 0.03062584064900875, "global_step": 270055, "epoch": 3034, "lr": 6.530967703416435e-09}
{"train_loss": 0.04606471210718155, "global_step": 270056, "epoch": 3034, "lr": 6.521601156433965e-09}
{"train_loss": 0.027559766545891762, "global_step": 270057, "epoch": 3034, "lr": 6.51224133046413e-09}
{"train_loss": 0.042581938207149506, "global_step": 270058, "epoch": 3034, "lr": 6.502888225506931e-09}
{"train_loss": 0.009494693018496037, "global_step": 270059, "epoch": 3034, "lr": 6.493541841567918e-09}
{"train_loss": 0.05232716351747513, "global_step": 270060, "epoch": 3034, "lr": 6.484202178647092e-09}
{"train_loss": 0.02916371263563633, "global_step": 270061, "epoch": 3034, "lr": 6.474869236744452e-09}
{"train_loss": 0.06370298564434052, "global_step": 270062, "epoch": 3034, "lr": 6.4655430158599985e-09}
{"train_loss": 0.07518136501312256, "global_step": 270063, "epoch": 3034, "lr": 6.456223515999282e-09}
{"train_loss": 0.035663314163684845, "global_step": 270064, "epoch": 3034, "lr": 6.4469107371623036e-09}
{"train_loss": 0.049651652574539185, "global_step": 270065, "epoch": 3034, "lr": 6.437604679349063e-09}
{"train_loss": 0.01737932488322258, "global_step": 270066, "epoch": 3034, "lr": 6.4283053425595595e-09}
{"train_loss": 0.08452753722667694, "global_step": 270067, "epoch": 3034, "lr": 6.419012726793794e-09}
{"train_loss": 0.06658222526311874, "global_step": 270068, "epoch": 3034, "lr": 6.4097268320573166e-09}
{"train_loss": 0.038995254784822464, "global_step": 270069, "epoch": 3034, "lr": 6.400447658350128e-09}
{"train_loss": 0.052322711795568466, "global_step": 270070, "epoch": 3034, "lr": 6.39117520567778e-09}
{"train_loss": 0.02376176230609417, "global_step": 270071, "epoch": 3034, "lr": 6.381909474029169e-09}
{"train_loss": 0.07296015322208405, "global_step": 270072, "epoch": 3034, "lr": 6.372650463415397e-09}
{"train_loss": 0.022921517491340637, "global_step": 270073, "epoch": 3034, "lr": 6.363398173836466e-09}
{"train_loss": 0.07288063317537308, "global_step": 270074, "epoch": 3034, "lr": 6.3541526052923735e-09}
{"train_loss": 0.06520941853523254, "global_step": 270075, "epoch": 3034, "lr": 6.344913757788673e-09}
{"train_loss": 0.05265926942229271, "global_step": 270076, "epoch": 3034, "lr": 6.3356816313198115e-09}
{"train_loss": 0.025154316797852516, "global_step": 270077, "epoch": 3034, "lr": 6.326456225885791e-09}
{"train_loss": 0.07212410867214203, "global_step": 270078, "epoch": 3034, "lr": 6.317237541497712e-09}
{"train_loss": 0.029650619253516197, "global_step": 270079, "epoch": 3034, "lr": 6.308025578144472e-09}
{"train_loss": 0.05427079647779465, "global_step": 270080, "epoch": 3034, "lr": 6.2988203358371745e-09}
{"train_loss": 0.02996259555220604, "global_step": 270081, "epoch": 3034, "lr": 6.289621814575819e-09}
{"train_loss": 0.050432465970516205, "global_step": 270082, "epoch": 3034, "lr": 6.280430014354855e-09}
{"train_loss": 0.08992327749729156, "global_step": 270083, "epoch": 3034, "lr": 6.271244935185383e-09}
{"train_loss": 0.04460190609097481, "global_step": 270084, "epoch": 3034, "lr": 6.262066577061854e-09}
{"train_loss": 0.0584380067884922, "global_step": 270085, "epoch": 3034, "lr": 6.252894939984267e-09}
{"train_loss": 0.028725335374474525, "global_step": 270086, "epoch": 3034, "lr": 6.243730023958172e-09}
{"train_loss": 0.04886777698993683, "global_step": 270087, "epoch": 3034, "lr": 6.234571828983571e-09}
{"train_loss": 0.05696706473827362, "global_step": 270088, "epoch": 3034, "lr": 6.225420355060463e-09}
{"train_loss": 0.04306541383266449, "global_step": 270089, "epoch": 3034, "lr": 6.216275602194399e-09}
{"train_loss": 0.04202619194984436, "global_step": 270090, "epoch": 3034, "lr": 6.207137570379828e-09}
{"train_loss": 0.03776965290307999, "global_step": 270091, "epoch": 3034, "lr": 6.198006259622302e-09}
{"train_loss": 0.06058748438954353, "global_step": 270092, "epoch": 3034, "lr": 6.18888166992182e-09}
{"train_loss": 0.04429148882627487, "global_step": 270093, "epoch": 3034, "lr": 6.179763801283933e-09}
{"train_loss": 0.04437275230884552, "global_step": 270094, "epoch": 3034, "lr": 6.17065265369754e-09}
{"train_loss": 0.01914185844361782, "global_step": 270095, "epoch": 3034, "lr": 6.161548227179292e-09}
{"train_loss": 0.025504732504487038, "global_step": 270096, "epoch": 3034, "lr": 6.152450521718089e-09}
{"train_loss": 0.02298288233578205, "global_step": 270097, "epoch": 3034, "lr": 6.143359537325033e-09}
{"train_loss": 0.06189124286174774, "global_step": 270098, "epoch": 3034, "lr": 6.134275273994572e-09}
{"train_loss": 0.027180342003703117, "global_step": 270099, "epoch": 3034, "lr": 6.1251977317322576e-09}
{"train_loss": 0.020408596843481064, "global_step": 270100, "epoch": 3034, "lr": 6.116126910532538e-09}
{"train_loss": 0.03198843076825142, "global_step": 270101, "epoch": 3034, "lr": 6.107062810406517e-09}
{"train_loss": 0.06093420088291168, "global_step": 270102, "epoch": 3034, "lr": 6.0980054313430905e-09}
{"train_loss": 0.05566442385315895, "global_step": 270103, "epoch": 3034, "lr": 6.0889547733589126e-09}
{"train_loss": 0.024620404466986656, "global_step": 270104, "epoch": 3034, "lr": 6.07991083643733e-09}
{"train_loss": 0.05233576148748398, "global_step": 270105, "epoch": 3034, "lr": 6.070873620594997e-09}
{"train_loss": 0.0408284030854702, "global_step": 270106, "epoch": 3034, "lr": 6.061843125826361e-09}
{"train_loss": 0.042524781078100204, "global_step": 270107, "epoch": 3034, "lr": 6.052819352131423e-09}
{"train_loss": 0.03177754208445549, "global_step": 270108, "epoch": 3034, "lr": 6.043802299515733e-09}
{"train_loss": 0.04537598788738251, "global_step": 270109, "epoch": 3034, "lr": 6.0347919679792915e-09}
{"train_loss": 0.06590396910905838, "global_step": 270110, "epoch": 3034, "lr": 6.0257883575221e-09}
{"train_loss": 0.0650186836719513, "global_step": 270111, "epoch": 3034, "lr": 6.016791468138605e-09}
{"train_loss": 0.04221687093377113, "global_step": 270112, "epoch": 3034, "lr": 6.0078012998454614e-09}
{"train_loss": 0.030180150642991066, "global_step": 270113, "epoch": 3034, "lr": 5.998817852631566e-09}
{"train_loss": 0.04363390155597014, "global_step": 270114, "epoch": 3034, "lr": 5.9898411265024714e-09, "val_loss": 9.63388729095459}
{"train_loss": 0.016383590176701546, "global_step": 270115, "epoch": 3035, "lr": 5.980871121452625e-09}
{"train_loss": 0.05287247896194458, "global_step": 270116, "epoch": 3035, "lr": 5.9719078374986805e-09}
{"train_loss": 0.01870473101735115, "global_step": 270117, "epoch": 3035, "lr": 5.9629512746295355e-09}
{"train_loss": 0.10531216114759445, "global_step": 270118, "epoch": 3035, "lr": 5.954001432845191e-09}
{"train_loss": 0.06026873365044594, "global_step": 270119, "epoch": 3035, "lr": 5.945058312156748e-09}
{"train_loss": 0.056907858699560165, "global_step": 270120, "epoch": 3035, "lr": 5.936121912553105e-09}
{"train_loss": 0.05707404389977455, "global_step": 270121, "epoch": 3035, "lr": 5.927192234045364e-09}
{"train_loss": 0.05156250670552254, "global_step": 270122, "epoch": 3035, "lr": 5.918269276633526e-09}
{"train_loss": 0.014118140563368797, "global_step": 270123, "epoch": 3035, "lr": 5.909353040312038e-09}
{"train_loss": 0.033841442316770554, "global_step": 270124, "epoch": 3035, "lr": 5.900443525092003e-09}
{"train_loss": 0.05736074224114418, "global_step": 270125, "epoch": 3035, "lr": 5.89154073096787e-09}
{"train_loss": 0.036413319408893585, "global_step": 270126, "epoch": 3035, "lr": 5.88264465793964e-09}
{"train_loss": 0.03770025819540024, "global_step": 270127, "epoch": 3035, "lr": 5.873755306012863e-09}
{"train_loss": 0.06851984560489655, "global_step": 270128, "epoch": 3035, "lr": 5.864872675187539e-09}
{"train_loss": 0.07843359559774399, "global_step": 270129, "epoch": 3035, "lr": 5.855996765463667e-09}
{"train_loss": 0.11774658411741257, "global_step": 270130, "epoch": 3035, "lr": 5.847127576841249e-09}
{"train_loss": 0.026753563433885574, "global_step": 270131, "epoch": 3035, "lr": 5.838265109325835e-09}
{"train_loss": 0.03185737878084183, "global_step": 270132, "epoch": 3035, "lr": 5.829409362917426e-09}
{"train_loss": 0.07692374289035797, "global_step": 270133, "epoch": 3035, "lr": 5.82056033761047e-09}
{"train_loss": 0.07332293689250946, "global_step": 270134, "epoch": 3035, "lr": 5.811718033416069e-09}
{"train_loss": 0.04374516010284424, "global_step": 270135, "epoch": 3035, "lr": 5.802882450334224e-09}
{"train_loss": 0.04958491772413254, "global_step": 270136, "epoch": 3035, "lr": 5.7940535883538315e-09}
{"train_loss": 0.04845463111996651, "global_step": 270137, "epoch": 3035, "lr": 5.785231447491546e-09}
{"train_loss": 0.052864931523799896, "global_step": 270138, "epoch": 3035, "lr": 5.776416027741816e-09}
{"train_loss": 0.017039094120264053, "global_step": 270139, "epoch": 3035, "lr": 5.767607329104641e-09}
{"train_loss": 0.06464867293834686, "global_step": 270140, "epoch": 3035, "lr": 5.758805351580021e-09}
{"train_loss": 0.04369549825787544, "global_step": 270141, "epoch": 3035, "lr": 5.750010095173508e-09}
{"train_loss": 0.07571207731962204, "global_step": 270142, "epoch": 3035, "lr": 5.741221559890653e-09}
{"train_loss": 0.012850855477154255, "global_step": 270143, "epoch": 3035, "lr": 5.732439745720353e-09}
{"train_loss": 0.03146786615252495, "global_step": 270144, "epoch": 3035, "lr": 5.723664652668159e-09}
{"train_loss": 0.027115602046251297, "global_step": 270145, "epoch": 3035, "lr": 5.714896280739623e-09}
{"train_loss": 0.014888955280184746, "global_step": 270146, "epoch": 3035, "lr": 5.706134629934745e-09}
{"train_loss": 0.05051084980368614, "global_step": 270147, "epoch": 3035, "lr": 5.697379700253524e-09}
{"train_loss": 0.04509693384170532, "global_step": 270148, "epoch": 3035, "lr": 5.688631491695962e-09}
{"train_loss": 0.055648840963840485, "global_step": 270149, "epoch": 3035, "lr": 5.679890004262056e-09}
{"train_loss": 0.017315899953246117, "global_step": 270150, "epoch": 3035, "lr": 5.671155237957359e-09}
{"train_loss": 0.06583090126514435, "global_step": 270151, "epoch": 3035, "lr": 5.662427192781872e-09}
{"train_loss": 0.017942028120160103, "global_step": 270152, "epoch": 3035, "lr": 5.653705868735592e-09}
{"train_loss": 0.01902107335627079, "global_step": 270153, "epoch": 3035, "lr": 5.644991265818522e-09}
{"train_loss": 0.0808987021446228, "global_step": 270154, "epoch": 3035, "lr": 5.6362833840306605e-09}
{"train_loss": 0.06464903801679611, "global_step": 270155, "epoch": 3035, "lr": 5.6275822233775584e-09}
{"train_loss": 0.040095917880535126, "global_step": 270156, "epoch": 3035, "lr": 5.618887783859217e-09}
{"train_loss": 0.058723755180835724, "global_step": 270157, "epoch": 3035, "lr": 5.610200065475635e-09}
{"train_loss": 0.043725330382585526, "global_step": 270158, "epoch": 3035, "lr": 5.6015190682323635e-09}
{"train_loss": 0.07510224729776382, "global_step": 270159, "epoch": 3035, "lr": 5.592844792123853e-09}
{"train_loss": 0.05346893146634102, "global_step": 270160, "epoch": 3035, "lr": 5.584177237150101e-09}
{"train_loss": 0.07071062177419662, "global_step": 270161, "epoch": 3035, "lr": 5.5755164033222116e-09}
{"train_loss": 0.044440917670726776, "global_step": 270162, "epoch": 3035, "lr": 5.566862290629082e-09}
{"train_loss": 0.030643772333860397, "global_step": 270163, "epoch": 3035, "lr": 5.558214899081815e-09}
{"train_loss": 0.06503310054540634, "global_step": 270164, "epoch": 3035, "lr": 5.54957422868041e-09}
{"train_loss": 0.05073627084493637, "global_step": 270165, "epoch": 3035, "lr": 5.540940279419315e-09}
{"train_loss": 0.049378570169210434, "global_step": 270166, "epoch": 3035, "lr": 5.532313051304083e-09}
{"train_loss": 0.035338204354047775, "global_step": 270167, "epoch": 3035, "lr": 5.5236925443347136e-09}
{"train_loss": 0.038192979991436005, "global_step": 270168, "epoch": 3035, "lr": 5.515078758516756e-09}
{"train_loss": 0.040065038949251175, "global_step": 270169, "epoch": 3035, "lr": 5.506471693850213e-09}
{"train_loss": 0.0605192631483078, "global_step": 270170, "epoch": 3035, "lr": 5.497871350329531e-09}
{"train_loss": 0.037758033722639084, "global_step": 270171, "epoch": 3035, "lr": 5.489277727960263e-09}
{"train_loss": 0.07101237028837204, "global_step": 270172, "epoch": 3035, "lr": 5.480690826742407e-09}
{"train_loss": 0.022550079971551895, "global_step": 270173, "epoch": 3035, "lr": 5.472110646681517e-09}
{"train_loss": 0.042684152722358704, "global_step": 270174, "epoch": 3035, "lr": 5.46353718777759e-09}
{"train_loss": 0.04517266899347305, "global_step": 270175, "epoch": 3035, "lr": 5.454970450025076e-09}
{"train_loss": 0.05113985016942024, "global_step": 270176, "epoch": 3035, "lr": 5.446410433429527e-09}
{"train_loss": 0.05029664933681488, "global_step": 270177, "epoch": 3035, "lr": 5.437857137996494e-09}
{"train_loss": 0.04554042965173721, "global_step": 270178, "epoch": 3035, "lr": 5.429310563720424e-09}
{"train_loss": 0.0701102763414383, "global_step": 270179, "epoch": 3035, "lr": 5.420770710606871e-09}
{"train_loss": 0.010079418309032917, "global_step": 270180, "epoch": 3035, "lr": 5.412237578650281e-09}
{"train_loss": 0.018774917349219322, "global_step": 270181, "epoch": 3035, "lr": 5.403711167861758e-09}
{"train_loss": 0.037547480314970016, "global_step": 270182, "epoch": 3035, "lr": 5.39519147823575e-09}
{"train_loss": 0.05337950214743614, "global_step": 270183, "epoch": 3035, "lr": 5.386678509772259e-09}
{"train_loss": 0.03531437739729881, "global_step": 270184, "epoch": 3035, "lr": 5.3781722624823835e-09}
{"train_loss": 0.02316911332309246, "global_step": 270185, "epoch": 3035, "lr": 5.369672736355025e-09}
{"train_loss": 0.006919011939316988, "global_step": 270186, "epoch": 3035, "lr": 5.361179931395732e-09}
{"train_loss": 0.056484613567590714, "global_step": 270187, "epoch": 3035, "lr": 5.3526938476045065e-09}
{"train_loss": 0.04414733499288559, "global_step": 270188, "epoch": 3035, "lr": 5.344214484986898e-09}
{"train_loss": 0.04289254918694496, "global_step": 270189, "epoch": 3035, "lr": 5.3357418435429076e-09}
{"train_loss": 0.059718091040849686, "global_step": 270190, "epoch": 3035, "lr": 5.327275923272534e-09}
{"train_loss": 0.03945765271782875, "global_step": 270191, "epoch": 3035, "lr": 5.318816724175779e-09}
{"train_loss": 0.03981295973062515, "global_step": 270192, "epoch": 3035, "lr": 5.3103642462526414e-09}
{"train_loss": 0.03913263976573944, "global_step": 270193, "epoch": 3035, "lr": 5.301918489508673e-09}
{"train_loss": 0.02695534937083721, "global_step": 270194, "epoch": 3035, "lr": 5.2934794539383215e-09}
{"train_loss": 0.032785169780254364, "global_step": 270195, "epoch": 3035, "lr": 5.28504713955269e-09}
{"train_loss": 0.04508394002914429, "global_step": 270196, "epoch": 3035, "lr": 5.276621546340676e-09}
{"train_loss": 0.04635414481163025, "global_step": 270197, "epoch": 3035, "lr": 5.268202674313383e-09}
{"train_loss": 0.05039786547422409, "global_step": 270198, "epoch": 3035, "lr": 5.259790523470809e-09}
{"train_loss": 0.0624937042593956, "global_step": 270199, "epoch": 3035, "lr": 5.251385093807404e-09}
{"train_loss": 0.03393113613128662, "global_step": 270200, "epoch": 3035, "lr": 5.2429863853287185e-09}
{"train_loss": 0.03927759826183319, "global_step": 270201, "epoch": 3035, "lr": 5.234594398040305e-09}
{"train_loss": 0.08736871182918549, "global_step": 270202, "epoch": 3035, "lr": 5.22620913193661e-09}
{"train_loss": 0.0466185445750781, "global_step": 270203, "epoch": 3035, "lr": 5.217830587017636e-09, "val_loss": 9.59857177734375, "train_action_mse_error": 20.08146095275879}
{"train_loss": 0.037575460970401764, "global_step": 270204, "epoch": 3036, "lr": 5.2094587632944835e-09}
{"train_loss": 0.05640896037220955, "global_step": 270205, "epoch": 3036, "lr": 5.201093660756051e-09}
{"train_loss": 0.08675752580165863, "global_step": 270206, "epoch": 3036, "lr": 5.19273527940789e-09}
{"train_loss": 0.07621829211711884, "global_step": 270207, "epoch": 3036, "lr": 5.1843836192555505e-09}
{"train_loss": 0.009148339740931988, "global_step": 270208, "epoch": 3036, "lr": 5.176038680299034e-09}
{"train_loss": 0.03054976649582386, "global_step": 270209, "epoch": 3036, "lr": 5.167700462532787e-09}
{"train_loss": 0.037656981498003006, "global_step": 270210, "epoch": 3036, "lr": 5.159368965962363e-09}
{"train_loss": 0.02129262499511242, "global_step": 270211, "epoch": 3036, "lr": 5.151044190593313e-09}
{"train_loss": 0.023965485394001007, "global_step": 270212, "epoch": 3036, "lr": 5.1427261364145325e-09}
{"train_loss": 0.013862567022442818, "global_step": 270213, "epoch": 3036, "lr": 5.1344148034426775e-09}
{"train_loss": 0.07902128249406815, "global_step": 270214, "epoch": 3036, "lr": 5.1261101916666445e-09}
{"train_loss": 0.028971532359719276, "global_step": 270215, "epoch": 3036, "lr": 5.117812301097535e-09}
{"train_loss": 0.01623268984258175, "global_step": 270216, "epoch": 3036, "lr": 5.109521131724249e-09}
{"train_loss": 0.0602247528731823, "global_step": 270217, "epoch": 3036, "lr": 5.101236683557886e-09}
{"train_loss": 0.022937919944524765, "global_step": 270218, "epoch": 3036, "lr": 5.0929589565984485e-09}
{"train_loss": 0.05429692938923836, "global_step": 270219, "epoch": 3036, "lr": 5.0846879508403836e-09}
{"train_loss": 0.04743720218539238, "global_step": 270220, "epoch": 3036, "lr": 5.076423666289243e-09}
{"train_loss": 0.048360466957092285, "global_step": 270221, "epoch": 3036, "lr": 5.0681661029505775e-09}
{"train_loss": 0.004281681030988693, "global_step": 270222, "epoch": 3036, "lr": 5.059915260818837e-09}
{"train_loss": 0.06843414902687073, "global_step": 270223, "epoch": 3036, "lr": 5.051671139899572e-09}
{"train_loss": 0.053044166415929794, "global_step": 270224, "epoch": 3036, "lr": 5.043433740187231e-09}
{"train_loss": 0.021923048421740532, "global_step": 270225, "epoch": 3036, "lr": 5.035203061687366e-09}
{"train_loss": 0.03353176265954971, "global_step": 270226, "epoch": 3036, "lr": 5.026979104405527e-09}
{"train_loss": 0.014268049038946629, "global_step": 270227, "epoch": 3036, "lr": 5.0187618683361634e-09}
{"train_loss": 0.05072759464383125, "global_step": 270228, "epoch": 3036, "lr": 5.010551353484827e-09}
{"train_loss": 0.06555187702178955, "global_step": 270229, "epoch": 3036, "lr": 5.002347559845965e-09}
{"train_loss": 0.012417535297572613, "global_step": 270230, "epoch": 3036, "lr": 4.994150487430682e-09}
{"train_loss": 0.0331619493663311, "global_step": 270231, "epoch": 3036, "lr": 4.9859601362334254e-09}
{"train_loss": 0.032213322818279266, "global_step": 270232, "epoch": 3036, "lr": 4.977776506254195e-09}
{"train_loss": 0.06572949141263962, "global_step": 270233, "epoch": 3036, "lr": 4.969599597498542e-09}
{"train_loss": 0.06379325687885284, "global_step": 270234, "epoch": 3036, "lr": 4.961429409960916e-09}
{"train_loss": 0.018292732536792755, "global_step": 270235, "epoch": 3036, "lr": 4.953265943652419e-09}
{"train_loss": 0.037777408957481384, "global_step": 270236, "epoch": 3036, "lr": 4.945109198561948e-09}
{"train_loss": 0.05865376815199852, "global_step": 270237, "epoch": 3036, "lr": 4.936959174700606e-09}
{"train_loss": 0.05943187326192856, "global_step": 270238, "epoch": 3036, "lr": 4.928815872068393e-09}
{"train_loss": 0.0456034317612648, "global_step": 270239, "epoch": 3036, "lr": 4.920679290665308e-09}
{"train_loss": 0.04761330783367157, "global_step": 270240, "epoch": 3036, "lr": 4.912549430485802e-09}
{"train_loss": 0.04109274223446846, "global_step": 270241, "epoch": 3036, "lr": 4.9044262915409754e-09}
{"train_loss": 0.022853970527648926, "global_step": 270242, "epoch": 3036, "lr": 4.896309873825277e-09}
{"train_loss": 0.06584185361862183, "global_step": 270243, "epoch": 3036, "lr": 4.88820017734426e-09}
{"train_loss": 0.028117159381508827, "global_step": 270244, "epoch": 3036, "lr": 4.880097202092371e-09}
{"train_loss": 0.04061676189303398, "global_step": 270245, "epoch": 3036, "lr": 4.8720009480751615e-09}
{"train_loss": 0.03760592266917229, "global_step": 270246, "epoch": 3036, "lr": 4.863911415298183e-09}
{"train_loss": 0.10673549771308899, "global_step": 270247, "epoch": 3036, "lr": 4.855828603755885e-09}
{"train_loss": 0.03214188665151596, "global_step": 270248, "epoch": 3036, "lr": 4.847752513448267e-09}
{"train_loss": 0.013578291982412338, "global_step": 270249, "epoch": 3036, "lr": 4.839683144386431e-09}
{"train_loss": 0.023231443017721176, "global_step": 270250, "epoch": 3036, "lr": 4.831620496559275e-09}
{"train_loss": 0.08610739558935165, "global_step": 270251, "epoch": 3036, "lr": 4.8235645699723496e-09}
{"train_loss": 0.06326659768819809, "global_step": 270252, "epoch": 3036, "lr": 4.815515364631207e-09}
{"train_loss": 0.022603696212172508, "global_step": 270253, "epoch": 3036, "lr": 4.807472880530295e-09}
{"train_loss": 0.05986785516142845, "global_step": 270254, "epoch": 3036, "lr": 4.799437117675165e-09}
{"train_loss": 0.05498463287949562, "global_step": 270255, "epoch": 3036, "lr": 4.7914080760658176e-09}
{"train_loss": 0.03885234519839287, "global_step": 270256, "epoch": 3036, "lr": 4.783385755702252e-09}
{"train_loss": 0.00872253067791462, "global_step": 270257, "epoch": 3036, "lr": 4.775370156584469e-09}
{"train_loss": 0.03641560301184654, "global_step": 270258, "epoch": 3036, "lr": 4.767361278718019e-09}
{"train_loss": 0.04225980490446091, "global_step": 270259, "epoch": 3036, "lr": 4.759359122102902e-09}
{"train_loss": 0.04864252731204033, "global_step": 270260, "epoch": 3036, "lr": 4.751363686733567e-09}
{"train_loss": 0.016568105667829514, "global_step": 270261, "epoch": 3036, "lr": 4.7433749726155665e-09}
{"train_loss": 0.017566535621881485, "global_step": 270262, "epoch": 3036, "lr": 4.7353929797544495e-09}
{"train_loss": 0.03090243972837925, "global_step": 270263, "epoch": 3036, "lr": 4.727417708144666e-09}
{"train_loss": 0.031851429492235184, "global_step": 270264, "epoch": 3036, "lr": 4.719449157791767e-09}
{"train_loss": 0.023650620132684708, "global_step": 270265, "epoch": 3036, "lr": 4.711487328695751e-09}
{"train_loss": 0.03610595688223839, "global_step": 270266, "epoch": 3036, "lr": 4.703532220856621e-09}
{"train_loss": 0.016631655395030975, "global_step": 270267, "epoch": 3036, "lr": 4.695583834274375e-09}
{"train_loss": 0.04386454448103905, "global_step": 270268, "epoch": 3036, "lr": 4.687642168949013e-09}
{"train_loss": 0.03362041711807251, "global_step": 270269, "epoch": 3036, "lr": 4.679707224891639e-09}
{"train_loss": 0.09795499593019485, "global_step": 270270, "epoch": 3036, "lr": 4.671779002091148e-09}
{"train_loss": 0.028084829449653625, "global_step": 270271, "epoch": 3036, "lr": 4.663857500547542e-09}
{"train_loss": 0.039996903389692307, "global_step": 270272, "epoch": 3036, "lr": 4.655942720277473e-09}
{"train_loss": 0.04001207277178764, "global_step": 270273, "epoch": 3036, "lr": 4.648034661264289e-09}
{"train_loss": 0.10527322441339493, "global_step": 270274, "epoch": 3036, "lr": 4.6401333235190915e-09}
{"train_loss": 0.037585582584142685, "global_step": 270275, "epoch": 3036, "lr": 4.63223870704188e-09}
{"train_loss": 0.0660775676369667, "global_step": 270276, "epoch": 3036, "lr": 4.6243508118326564e-09}
{"train_loss": 0.02672048844397068, "global_step": 270277, "epoch": 3036, "lr": 4.616469637891418e-09}
{"train_loss": 0.08434906601905823, "global_step": 270278, "epoch": 3036, "lr": 4.608595185223719e-09}
{"train_loss": 0.03273046016693115, "global_step": 270279, "epoch": 3036, "lr": 4.600727453824006e-09}
{"train_loss": 0.03427481651306152, "global_step": 270280, "epoch": 3036, "lr": 4.59286644369783e-09}
{"train_loss": 0.08119982481002808, "global_step": 270281, "epoch": 3036, "lr": 4.585012154839641e-09}
{"train_loss": 0.038492050021886826, "global_step": 270282, "epoch": 3036, "lr": 4.577164587260541e-09}
{"train_loss": 0.012102334760129452, "global_step": 270283, "epoch": 3036, "lr": 4.5693237409549786e-09}
{"train_loss": 0.04935875162482262, "global_step": 270284, "epoch": 3036, "lr": 4.561489615928505e-09}
{"train_loss": 0.0342913419008255, "global_step": 270285, "epoch": 3036, "lr": 4.553662212175569e-09}
{"train_loss": 0.07616142928600311, "global_step": 270286, "epoch": 3036, "lr": 4.545841529701722e-09}
{"train_loss": 0.06949801743030548, "global_step": 270287, "epoch": 3036, "lr": 4.538027568506964e-09}
{"train_loss": 0.034145817160606384, "global_step": 270288, "epoch": 3036, "lr": 4.530220328596846e-09}
{"train_loss": 0.05654891952872276, "global_step": 270289, "epoch": 3036, "lr": 4.522419809965816e-09}
{"train_loss": 0.03981032222509384, "global_step": 270290, "epoch": 3036, "lr": 4.514626012613876e-09}
{"train_loss": 0.0442398265004158, "global_step": 270291, "epoch": 3036, "lr": 4.506838936552127e-09}
{"train_loss": 0.042905619660957475, "global_step": 270292, "epoch": 3036, "lr": 4.499058581769466e-09, "val_loss": 9.736735343933105}
{"train_loss": 0.025327511131763458, "global_step": 270293, "epoch": 3037, "lr": 4.491284948271446e-09}
{"train_loss": 0.0663992166519165, "global_step": 270294, "epoch": 3037, "lr": 4.4835180360636164e-09}
{"train_loss": 0.047056894749403, "global_step": 270295, "epoch": 3037, "lr": 4.475757845145978e-09}
{"train_loss": 0.07980386167764664, "global_step": 270296, "epoch": 3037, "lr": 4.468004375512979e-09}
{"train_loss": 0.02655785344541073, "global_step": 270297, "epoch": 3037, "lr": 4.4602576271701726e-09}
{"train_loss": 0.032893527299165726, "global_step": 270298, "epoch": 3037, "lr": 4.452517600123107e-09}
{"train_loss": 0.07259570062160492, "global_step": 270299, "epoch": 3037, "lr": 4.4447842943662335e-09}
{"train_loss": 0.040583137422800064, "global_step": 270300, "epoch": 3037, "lr": 4.43705770989955e-09}
{"train_loss": 0.03971336781978607, "global_step": 270301, "epoch": 3037, "lr": 4.4293378467286095e-09}
{"train_loss": 0.07105869054794312, "global_step": 270302, "epoch": 3037, "lr": 4.42162470484786e-09}
{"train_loss": 0.038528893142938614, "global_step": 270303, "epoch": 3037, "lr": 4.413918284268403e-09}
{"train_loss": 0.02888287790119648, "global_step": 270304, "epoch": 3037, "lr": 4.406218584984689e-09}
{"train_loss": 0.05347168818116188, "global_step": 270305, "epoch": 3037, "lr": 4.398525607002269e-09}
{"train_loss": 0.010571612976491451, "global_step": 270306, "epoch": 3037, "lr": 4.3908393503155896e-09}
{"train_loss": 0.06216227635741234, "global_step": 270307, "epoch": 3037, "lr": 4.383159814930205e-09}
{"train_loss": 0.076596699655056, "global_step": 270308, "epoch": 3037, "lr": 4.375487000846112e-09}
{"train_loss": 0.03719763830304146, "global_step": 270309, "epoch": 3037, "lr": 4.367820908068865e-09}
{"train_loss": 0.04261422157287598, "global_step": 270310, "epoch": 3037, "lr": 4.360161536587359e-09}
{"train_loss": 0.03899601846933365, "global_step": 270311, "epoch": 3037, "lr": 4.352508886412698e-09}
{"train_loss": 0.07750169932842255, "global_step": 270312, "epoch": 3037, "lr": 4.344862957544882e-09}
{"train_loss": 0.07034245133399963, "global_step": 270313, "epoch": 3037, "lr": 4.337223749983909e-09}
{"train_loss": 0.01608697511255741, "global_step": 270314, "epoch": 3037, "lr": 4.329591263729782e-09}
{"train_loss": 0.04690492898225784, "global_step": 270315, "epoch": 3037, "lr": 4.321965498782499e-09}
{"train_loss": 0.02142886258661747, "global_step": 270316, "epoch": 3037, "lr": 4.314346455147611e-09}
{"train_loss": 0.020300842821598053, "global_step": 270317, "epoch": 3037, "lr": 4.306734132825119e-09}
{"train_loss": 0.02404496632516384, "global_step": 270318, "epoch": 3037, "lr": 4.299128531809471e-09}
{"train_loss": 0.008425436913967133, "global_step": 270319, "epoch": 3037, "lr": 4.291529652106219e-09}
{"train_loss": 0.04581746831536293, "global_step": 270320, "epoch": 3037, "lr": 4.283937493720913e-09}
{"train_loss": 0.012036246247589588, "global_step": 270321, "epoch": 3037, "lr": 4.276352056648003e-09}
{"train_loss": 0.05746631324291229, "global_step": 270322, "epoch": 3037, "lr": 4.268773340887489e-09}
{"train_loss": 0.029940757900476456, "global_step": 270323, "epoch": 3037, "lr": 4.261201346450472e-09}
{"train_loss": 0.09541816264390945, "global_step": 270324, "epoch": 3037, "lr": 4.253636073325851e-09}
{"train_loss": 0.03594373166561127, "global_step": 270325, "epoch": 3037, "lr": 4.246077521519176e-09}
{"train_loss": 0.07754284888505936, "global_step": 270326, "epoch": 3037, "lr": 4.238525691035999e-09}
{"train_loss": 0.050753552466630936, "global_step": 270327, "epoch": 3037, "lr": 4.230980581870769e-09}
{"train_loss": 0.05273667722940445, "global_step": 270328, "epoch": 3037, "lr": 4.223442194029037e-09}
{"train_loss": 0.062475766986608505, "global_step": 270329, "epoch": 3037, "lr": 4.215910527510802e-09}
{"train_loss": 0.05805050581693649, "global_step": 270330, "epoch": 3037, "lr": 4.2083855823160655e-09}
{"train_loss": 0.056167151778936386, "global_step": 270331, "epoch": 3037, "lr": 4.2008673584448264e-09}
{"train_loss": 0.02437855675816536, "global_step": 270332, "epoch": 3037, "lr": 4.193355855902637e-09}
{"train_loss": 0.053732920438051224, "global_step": 270333, "epoch": 3037, "lr": 4.185851074683944e-09}
{"train_loss": 0.09532981365919113, "global_step": 270334, "epoch": 3037, "lr": 4.178353014794301e-09}
{"train_loss": 0.04016125947237015, "global_step": 270335, "epoch": 3037, "lr": 4.170861676233706e-09}
{"train_loss": 0.025859419256448746, "global_step": 270336, "epoch": 3037, "lr": 4.16337705900216e-09}
{"train_loss": 0.08344051241874695, "global_step": 270337, "epoch": 3037, "lr": 4.155899163099663e-09}
{"train_loss": 0.04870956763625145, "global_step": 270338, "epoch": 3037, "lr": 4.148427988531767e-09}
{"train_loss": 0.03717079386115074, "global_step": 270339, "epoch": 3037, "lr": 4.1409635352929186e-09}
{"train_loss": 0.04212206229567528, "global_step": 270340, "epoch": 3037, "lr": 4.133505803394222e-09}
{"train_loss": 0.03678514063358307, "global_step": 270341, "epoch": 3037, "lr": 4.126054792824574e-09}
{"train_loss": 0.03541701287031174, "global_step": 270342, "epoch": 3037, "lr": 4.118610503595077e-09}
{"train_loss": 0.06651486456394196, "global_step": 270343, "epoch": 3037, "lr": 4.111172935700181e-09}
{"train_loss": 0.0331275574862957, "global_step": 270344, "epoch": 3037, "lr": 4.1037420891454345e-09}
{"train_loss": 0.07811494171619415, "global_step": 270345, "epoch": 3037, "lr": 4.096317963925289e-09}
{"train_loss": 0.06480099260807037, "global_step": 270346, "epoch": 3037, "lr": 4.088900560050846e-09}
{"train_loss": 0.03275707736611366, "global_step": 270347, "epoch": 3037, "lr": 4.081489877511002e-09}
{"train_loss": 0.03825538977980614, "global_step": 270348, "epoch": 3037, "lr": 4.074085916316861e-09}
{"train_loss": 0.023088596761226654, "global_step": 270349, "epoch": 3037, "lr": 4.066688676468422e-09}
{"train_loss": 0.04459795355796814, "global_step": 270350, "epoch": 3037, "lr": 4.0592981579601345e-09}
{"train_loss": 0.040143441408872604, "global_step": 270351, "epoch": 3037, "lr": 4.051914360797549e-09}
{"train_loss": 0.07407090067863464, "global_step": 270352, "epoch": 3037, "lr": 4.0445372849806655e-09}
{"train_loss": 0.07640494406223297, "global_step": 270353, "epoch": 3037, "lr": 4.037166930509484e-09}
{"train_loss": 0.026704128831624985, "global_step": 270354, "epoch": 3037, "lr": 4.029803297384005e-09}
{"train_loss": 0.05448140949010849, "global_step": 270355, "epoch": 3037, "lr": 4.022446385609779e-09}
{"train_loss": 0.007605640217661858, "global_step": 270356, "epoch": 3037, "lr": 4.015096195186807e-09}
{"train_loss": 0.02382754720747471, "global_step": 270357, "epoch": 3037, "lr": 4.007752726115088e-09}
{"train_loss": 0.03652575984597206, "global_step": 270358, "epoch": 3037, "lr": 4.000415978394623e-09}
{"train_loss": 0.03291889652609825, "global_step": 270359, "epoch": 3037, "lr": 3.99308595202541e-09}
{"train_loss": 0.02431240864098072, "global_step": 270360, "epoch": 3037, "lr": 3.9857626470130025e-09}
{"train_loss": 0.06456685066223145, "global_step": 270361, "epoch": 3037, "lr": 3.978446063351848e-09}
{"train_loss": 0.0356869213283062, "global_step": 270362, "epoch": 3037, "lr": 3.9711362010474985e-09}
{"train_loss": 0.04464607313275337, "global_step": 270363, "epoch": 3037, "lr": 3.963833060099953e-09}
{"train_loss": 0.016638783738017082, "global_step": 270364, "epoch": 3037, "lr": 3.9565366405092115e-09}
{"train_loss": 0.04835120588541031, "global_step": 270365, "epoch": 3037, "lr": 3.949246942275275e-09}
{"train_loss": 0.02543153241276741, "global_step": 270366, "epoch": 3037, "lr": 3.9419639654036944e-09}
{"train_loss": 0.08206102252006531, "global_step": 270367, "epoch": 3037, "lr": 3.934687709894469e-09}
{"train_loss": 0.020742813125252724, "global_step": 270368, "epoch": 3037, "lr": 3.927418175747599e-09}
{"train_loss": 0.034361667931079865, "global_step": 270369, "epoch": 3037, "lr": 3.920155362957534e-09}
{"train_loss": 0.08310946077108383, "global_step": 270370, "epoch": 3037, "lr": 3.912899271535375e-09}
{"train_loss": 0.041511647403240204, "global_step": 270371, "epoch": 3037, "lr": 3.905649901475572e-09}
{"train_loss": 0.054115235805511475, "global_step": 270372, "epoch": 3037, "lr": 3.898407252783676e-09}
{"train_loss": 0.06165969744324684, "global_step": 270373, "epoch": 3037, "lr": 3.891171325459686e-09}
{"train_loss": 0.06204751133918762, "global_step": 270374, "epoch": 3037, "lr": 3.8839421194980516e-09}
{"train_loss": 0.06624536216259003, "global_step": 270375, "epoch": 3037, "lr": 3.876719634909875e-09}
{"train_loss": 0.009634377434849739, "global_step": 270376, "epoch": 3037, "lr": 3.869503871684055e-09}
{"train_loss": 0.04184996336698532, "global_step": 270377, "epoch": 3037, "lr": 3.862294829837243e-09}
{"train_loss": 0.12180895358324051, "global_step": 270378, "epoch": 3037, "lr": 3.8550925093527866e-09}
{"train_loss": 0.023613259196281433, "global_step": 270379, "epoch": 3037, "lr": 3.8478969102473394e-09}
{"train_loss": 0.020361747592687607, "global_step": 270380, "epoch": 3037, "lr": 3.84070803251535e-09}
{"train_loss": 0.046317432096667506, "global_step": 270381, "epoch": 3037, "lr": 3.833525876156818e-09, "val_loss": 9.650734901428223}
{"train_loss": 0.021065931767225266, "global_step": 270382, "epoch": 3038, "lr": 3.826350441171744e-09}
{"train_loss": 0.05379806086421013, "global_step": 270383, "epoch": 3038, "lr": 3.819181727560128e-09}
{"train_loss": 0.036129362881183624, "global_step": 270384, "epoch": 3038, "lr": 3.812019735333072e-09}
{"train_loss": 0.04134282469749451, "global_step": 270385, "epoch": 3038, "lr": 3.804864464479474e-09}
{"train_loss": 0.031797923147678375, "global_step": 270386, "epoch": 3038, "lr": 3.797715915004885e-09}
{"train_loss": 0.04029765725135803, "global_step": 270387, "epoch": 3038, "lr": 3.790574086909304e-09}
{"train_loss": 0.023551547899842262, "global_step": 270388, "epoch": 3038, "lr": 3.783438980198284e-09}
{"train_loss": 0.025140631943941116, "global_step": 270389, "epoch": 3038, "lr": 3.776310594866272e-09}
{"train_loss": 0.025368837639689445, "global_step": 270390, "epoch": 3038, "lr": 3.769188930918821e-09}
{"train_loss": 0.045918602496385574, "global_step": 270391, "epoch": 3038, "lr": 3.762073988350378e-09}
{"train_loss": 0.05168098956346512, "global_step": 270392, "epoch": 3038, "lr": 3.754965767172047e-09}
{"train_loss": 0.03892152011394501, "global_step": 270393, "epoch": 3038, "lr": 3.747864267378276e-09}
{"train_loss": 0.06000998616218567, "global_step": 270394, "epoch": 3038, "lr": 3.740769488974616e-09}
{"train_loss": 0.03882943093776703, "global_step": 270395, "epoch": 3038, "lr": 3.733681431955516e-09}
{"train_loss": 0.10125218331813812, "global_step": 270396, "epoch": 3038, "lr": 3.726600096320976e-09}
{"train_loss": 0.03470192849636078, "global_step": 270397, "epoch": 3038, "lr": 3.719525482082098e-09}
{"train_loss": 0.013061916455626488, "global_step": 270398, "epoch": 3038, "lr": 3.7124575892333314e-09}
{"train_loss": 0.037899747490882874, "global_step": 270399, "epoch": 3038, "lr": 3.705396417774676e-09}
{"train_loss": 0.014200934208929539, "global_step": 270400, "epoch": 3038, "lr": 3.6983419677061315e-09}
{"train_loss": 0.04232248663902283, "global_step": 270401, "epoch": 3038, "lr": 3.6912942390332495e-09}
{"train_loss": 0.02573430724442005, "global_step": 270402, "epoch": 3038, "lr": 3.6842532317560297e-09}
{"train_loss": 0.04567818343639374, "global_step": 270403, "epoch": 3038, "lr": 3.677218945868921e-09}
{"train_loss": 0.054296933114528656, "global_step": 270404, "epoch": 3038, "lr": 3.6701913813830258e-09}
{"train_loss": 0.039401330053806305, "global_step": 270405, "epoch": 3038, "lr": 3.663170538292793e-09}
{"train_loss": 0.059198543429374695, "global_step": 270406, "epoch": 3038, "lr": 3.656156416598222e-09}
{"train_loss": 0.11479820311069489, "global_step": 270407, "epoch": 3038, "lr": 3.649149016304865e-09}
{"train_loss": 0.029519986361265182, "global_step": 270408, "epoch": 3038, "lr": 3.642148337412721e-09}
{"train_loss": 0.04495133459568024, "global_step": 270409, "epoch": 3038, "lr": 3.6351543799217904e-09}
{"train_loss": 0.05858490243554115, "global_step": 270410, "epoch": 3038, "lr": 3.628167143832073e-09}
{"train_loss": 0.035604167729616165, "global_step": 270411, "epoch": 3038, "lr": 3.621186629143569e-09}
{"train_loss": 0.028170306235551834, "global_step": 270412, "epoch": 3038, "lr": 3.61421283586183e-09}
{"train_loss": 0.062138546258211136, "global_step": 270413, "epoch": 3038, "lr": 3.607245763981304e-09}
{"train_loss": 0.06401462852954865, "global_step": 270414, "epoch": 3038, "lr": 3.6002854135075426e-09}
{"train_loss": 0.043173160403966904, "global_step": 270415, "epoch": 3038, "lr": 3.5933317844405455e-09}
{"train_loss": 0.032832272350788116, "global_step": 270416, "epoch": 3038, "lr": 3.5863848767858644e-09}
{"train_loss": 0.031054629012942314, "global_step": 270417, "epoch": 3038, "lr": 3.5794446905323964e-09}
{"train_loss": 0.06666213274002075, "global_step": 270418, "epoch": 3038, "lr": 3.572511225691244e-09}
{"train_loss": 0.019293194636702538, "global_step": 270419, "epoch": 3038, "lr": 3.5655844822624073e-09}
{"train_loss": 0.06705562025308609, "global_step": 270420, "epoch": 3038, "lr": 3.558664460240335e-09}
{"train_loss": 0.0069504971615970135, "global_step": 270421, "epoch": 3038, "lr": 3.5517511596305787e-09}
{"train_loss": 0.0407007671892643, "global_step": 270422, "epoch": 3038, "lr": 3.5448445804386887e-09}
{"train_loss": 0.0060471720062196255, "global_step": 270423, "epoch": 3038, "lr": 3.5379447226591146e-09}
{"train_loss": 0.034005746245384216, "global_step": 270424, "epoch": 3038, "lr": 3.531051586291856e-09}
{"train_loss": 0.03514937683939934, "global_step": 270425, "epoch": 3038, "lr": 3.524165171342464e-09}
{"train_loss": 0.020703334361314774, "global_step": 270426, "epoch": 3038, "lr": 3.5172854778109386e-09}
{"train_loss": 0.041136231273412704, "global_step": 270427, "epoch": 3038, "lr": 3.5104125056917293e-09}
{"train_loss": 0.04189962148666382, "global_step": 270428, "epoch": 3038, "lr": 3.5035462549959376e-09}
{"train_loss": 0.035494282841682434, "global_step": 270429, "epoch": 3038, "lr": 3.4966867257180127e-09}
{"train_loss": 0.04706117510795593, "global_step": 270430, "epoch": 3038, "lr": 3.489833917863505e-09}
{"train_loss": 0.041211023926734924, "global_step": 270431, "epoch": 3038, "lr": 3.482987831426865e-09}
{"train_loss": 0.03978995606303215, "global_step": 270432, "epoch": 3038, "lr": 3.4761484664191933e-09}
{"train_loss": 0.04145733639597893, "global_step": 270433, "epoch": 3038, "lr": 3.469315822829389e-09}
{"train_loss": 0.08342933654785156, "global_step": 270434, "epoch": 3038, "lr": 3.462489900663002e-09}
{"train_loss": 0.061653755605220795, "global_step": 270435, "epoch": 3038, "lr": 3.455670699920033e-09}
{"train_loss": 0.056616660207509995, "global_step": 270436, "epoch": 3038, "lr": 3.4488582206060326e-09}
{"train_loss": 0.05955582857131958, "global_step": 270437, "epoch": 3038, "lr": 3.4420524627210014e-09}
{"train_loss": 0.037942301481962204, "global_step": 270438, "epoch": 3038, "lr": 3.4352534262593882e-09}
{"train_loss": 0.07842693477869034, "global_step": 270439, "epoch": 3038, "lr": 3.428461111232295e-09}
{"train_loss": 0.05325362831354141, "global_step": 270440, "epoch": 3038, "lr": 3.4216755176286196e-09}
{"train_loss": 0.02898244932293892, "global_step": 270441, "epoch": 3038, "lr": 3.4148966454594646e-09}
{"train_loss": 0.02081632800400257, "global_step": 270442, "epoch": 3038, "lr": 3.408124494719278e-09}
{"train_loss": 0.04950147122144699, "global_step": 270443, "epoch": 3038, "lr": 3.401359065413612e-09}
{"train_loss": 0.025746183469891548, "global_step": 270444, "epoch": 3038, "lr": 3.3946003575424656e-09}
{"train_loss": 0.033085279166698456, "global_step": 270445, "epoch": 3038, "lr": 3.3878483711002884e-09}
{"train_loss": 0.05776827409863472, "global_step": 270446, "epoch": 3038, "lr": 3.3811031060981823e-09}
{"train_loss": 0.02227906696498394, "global_step": 270447, "epoch": 3038, "lr": 3.3743645625305965e-09}
{"train_loss": 0.09979911148548126, "global_step": 270448, "epoch": 3038, "lr": 3.3676327403975307e-09}
{"train_loss": 0.04576367512345314, "global_step": 270449, "epoch": 3038, "lr": 3.360907639704536e-09}
{"train_loss": 0.06742540001869202, "global_step": 270450, "epoch": 3038, "lr": 3.3541892604516125e-09}
{"train_loss": 0.032379306852817535, "global_step": 270451, "epoch": 3038, "lr": 3.34747760263876e-09}
{"train_loss": 0.04148384928703308, "global_step": 270452, "epoch": 3038, "lr": 3.340772666265979e-09}
{"train_loss": 0.03342355415225029, "global_step": 270453, "epoch": 3038, "lr": 3.3340744513332688e-09}
{"train_loss": 0.05179787799715996, "global_step": 270454, "epoch": 3038, "lr": 3.3273829578406302e-09}
{"train_loss": 0.013836474157869816, "global_step": 270455, "epoch": 3038, "lr": 3.3206981857936136e-09}
{"train_loss": 0.054609514772892, "global_step": 270456, "epoch": 3038, "lr": 3.3140201351922194e-09}
{"train_loss": 0.024238435551524162, "global_step": 270457, "epoch": 3038, "lr": 3.3073488060308967e-09}
{"train_loss": 0.02778492495417595, "global_step": 270458, "epoch": 3038, "lr": 3.300684198320747e-09}
{"train_loss": 0.013568151742219925, "global_step": 270459, "epoch": 3038, "lr": 3.29402631205622e-09}
{"train_loss": 0.02486417442560196, "global_step": 270460, "epoch": 3038, "lr": 3.2873751472373148e-09}
{"train_loss": 0.07316659390926361, "global_step": 270461, "epoch": 3038, "lr": 3.2807307038695835e-09}
{"train_loss": 0.046496495604515076, "global_step": 270462, "epoch": 3038, "lr": 3.2740929819530254e-09}
{"train_loss": 0.016207491979002953, "global_step": 270463, "epoch": 3038, "lr": 3.2674619814820896e-09}
{"train_loss": 0.0784827321767807, "global_step": 270464, "epoch": 3038, "lr": 3.2608377024678782e-09}
{"train_loss": 0.03726719692349434, "global_step": 270465, "epoch": 3038, "lr": 3.254220144899289e-09}
{"train_loss": 0.06381579488515854, "global_step": 270466, "epoch": 3038, "lr": 3.247609308787425e-09}
{"train_loss": 0.029645368456840515, "global_step": 270467, "epoch": 3038, "lr": 3.241005194132285e-09}
{"train_loss": 0.03926759958267212, "global_step": 270468, "epoch": 3038, "lr": 3.2344078009283187e-09}
{"train_loss": 0.03329762443900108, "global_step": 270469, "epoch": 3038, "lr": 3.2278171291810767e-09}
{"train_loss": 0.042707171971292306, "global_step": 270470, "epoch": 3038, "lr": 3.221233178890559e-09, "val_loss": 9.644655227661133}
{"train_loss": 0.0287515576928854, "global_step": 270471, "epoch": 3039, "lr": 3.214655950056766e-09}
{"train_loss": 0.03493451699614525, "global_step": 270472, "epoch": 3039, "lr": 3.208085442679698e-09}
{"train_loss": 0.044046550989151, "global_step": 270473, "epoch": 3039, "lr": 3.201521656764905e-09}
{"train_loss": 0.01640438102185726, "global_step": 270474, "epoch": 3039, "lr": 3.1949645923123882e-09}
{"train_loss": 0.05658217519521713, "global_step": 270475, "epoch": 3039, "lr": 3.1884142493165956e-09}
{"train_loss": 0.05181169882416725, "global_step": 270476, "epoch": 3039, "lr": 3.181870627783079e-09}
{"train_loss": 0.05428747460246086, "global_step": 270477, "epoch": 3039, "lr": 3.175333727717389e-09}
{"train_loss": 0.03936711326241493, "global_step": 270478, "epoch": 3039, "lr": 3.1688035491084233e-09}
{"train_loss": 0.060937076807022095, "global_step": 270479, "epoch": 3039, "lr": 3.1622800919672846e-09}
{"train_loss": 0.05050782486796379, "global_step": 270480, "epoch": 3039, "lr": 3.1557633562939727e-09}
{"train_loss": 0.062280163168907166, "global_step": 270481, "epoch": 3039, "lr": 3.1492533420829364e-09}
{"train_loss": 0.029101040214300156, "global_step": 270482, "epoch": 3039, "lr": 3.142750049339727e-09}
{"train_loss": 0.021904356777668, "global_step": 270483, "epoch": 3039, "lr": 3.136253478069895e-09}
{"train_loss": 0.030572256073355675, "global_step": 270484, "epoch": 3039, "lr": 3.1297636282623388e-09}
{"train_loss": 0.07064864784479141, "global_step": 270485, "epoch": 3039, "lr": 3.1232804999281608e-09}
{"train_loss": 0.04393637180328369, "global_step": 270486, "epoch": 3039, "lr": 3.116804093061809e-09}
{"train_loss": 0.023280102759599686, "global_step": 270487, "epoch": 3039, "lr": 3.110334407668836e-09}
{"train_loss": 0.04293307662010193, "global_step": 270488, "epoch": 3039, "lr": 3.10387144374924e-09}
{"train_loss": 0.019441265612840652, "global_step": 270489, "epoch": 3039, "lr": 3.097415201303022e-09}
{"train_loss": 0.056491319090127945, "global_step": 270490, "epoch": 3039, "lr": 3.0909656803301823e-09}
{"train_loss": 0.08646474778652191, "global_step": 270491, "epoch": 3039, "lr": 3.08452288083072e-09}
{"train_loss": 0.006781730335205793, "global_step": 270492, "epoch": 3039, "lr": 3.078086802810187e-09}
{"train_loss": 0.033719997853040695, "global_step": 270493, "epoch": 3039, "lr": 3.0716574462630322e-09}
{"train_loss": 0.027653396129608154, "global_step": 270494, "epoch": 3039, "lr": 3.065234811194806e-09}
{"train_loss": 0.033516816794872284, "global_step": 270495, "epoch": 3039, "lr": 3.0588188976055086e-09}
{"train_loss": 0.019065992906689644, "global_step": 270496, "epoch": 3039, "lr": 3.0524097054951404e-09}
{"train_loss": 0.04370092228055, "global_step": 270497, "epoch": 3039, "lr": 3.0460072348637014e-09}
{"train_loss": 0.03450818359851837, "global_step": 270498, "epoch": 3039, "lr": 3.0396114857167424e-09}
{"train_loss": 0.041049934923648834, "global_step": 270499, "epoch": 3039, "lr": 3.033222458048712e-09}
{"train_loss": 0.07157265394926071, "global_step": 270500, "epoch": 3039, "lr": 3.0268401518651623e-09}
{"train_loss": 0.030633412301540375, "global_step": 270501, "epoch": 3039, "lr": 3.0204645671660925e-09}
{"train_loss": 0.05766039714217186, "global_step": 270502, "epoch": 3039, "lr": 3.0140957039515026e-09}
{"train_loss": 0.017803801223635674, "global_step": 270503, "epoch": 3039, "lr": 3.007733562221393e-09}
{"train_loss": 0.03318650647997856, "global_step": 270504, "epoch": 3039, "lr": 3.0013781419757637e-09}
{"train_loss": 0.05547773465514183, "global_step": 270505, "epoch": 3039, "lr": 2.9950294432201655e-09}
{"train_loss": 0.07681037485599518, "global_step": 270506, "epoch": 3039, "lr": 2.988687465949047e-09}
{"train_loss": 0.08031568676233292, "global_step": 270507, "epoch": 3039, "lr": 2.9823522101735113e-09}
{"train_loss": 0.05003806948661804, "global_step": 270508, "epoch": 3039, "lr": 2.9760236758824555e-09}
{"train_loss": 0.08037729561328888, "global_step": 270509, "epoch": 3039, "lr": 2.9697018630814312e-09}
{"train_loss": 0.06998039036989212, "global_step": 270510, "epoch": 3039, "lr": 2.9633867717704377e-09}
{"train_loss": 0.013807019218802452, "global_step": 270511, "epoch": 3039, "lr": 2.9570784019550267e-09}
{"train_loss": 0.05056733265519142, "global_step": 270512, "epoch": 3039, "lr": 2.9507767536296472e-09}
{"train_loss": 0.03363591805100441, "global_step": 270513, "epoch": 3039, "lr": 2.94448182679985e-09}
{"train_loss": 0.07859723269939423, "global_step": 270514, "epoch": 3039, "lr": 2.9381936214656348e-09}
{"train_loss": 0.04054615646600723, "global_step": 270515, "epoch": 3039, "lr": 2.931912137627002e-09}
{"train_loss": 0.030213257297873497, "global_step": 270516, "epoch": 3039, "lr": 2.9256373752839515e-09}
{"train_loss": 0.03858533874154091, "global_step": 270517, "epoch": 3039, "lr": 2.919369334442035e-09}
{"train_loss": 0.05415309593081474, "global_step": 270518, "epoch": 3039, "lr": 2.913108015090149e-09}
{"train_loss": 0.01834961399435997, "global_step": 270519, "epoch": 3039, "lr": 2.906853417244948e-09}
{"train_loss": 0.0216047503054142, "global_step": 270520, "epoch": 3039, "lr": 2.900605540895329e-09}
{"train_loss": 0.025124073028564453, "global_step": 270521, "epoch": 3039, "lr": 2.8943643860468437e-09}
{"train_loss": 0.02707449160516262, "global_step": 270522, "epoch": 3039, "lr": 2.888129952699492e-09}
{"train_loss": 0.03240508958697319, "global_step": 270523, "epoch": 3039, "lr": 2.881902240853274e-09}
{"train_loss": 0.05799536034464836, "global_step": 270524, "epoch": 3039, "lr": 2.8756812505137397e-09}
{"train_loss": 0.030733294785022736, "global_step": 270525, "epoch": 3039, "lr": 2.8694669816753396e-09}
{"train_loss": 0.0459604486823082, "global_step": 270526, "epoch": 3039, "lr": 2.8632594343436235e-09}
{"train_loss": 0.03990890085697174, "global_step": 270527, "epoch": 3039, "lr": 2.8570586085185926e-09}
{"train_loss": 0.02943672612309456, "global_step": 270528, "epoch": 3039, "lr": 2.8508645042002457e-09}
{"train_loss": 0.0851631835103035, "global_step": 270529, "epoch": 3039, "lr": 2.8446771213830325e-09}
{"train_loss": 0.03457766771316528, "global_step": 270530, "epoch": 3039, "lr": 2.838496460078055e-09}
{"train_loss": 0.07011844962835312, "global_step": 270531, "epoch": 3039, "lr": 2.8323225202853133e-09}
{"train_loss": 0.023333216086030006, "global_step": 270532, "epoch": 3039, "lr": 2.826155301999256e-09}
{"train_loss": 0.12460324168205261, "global_step": 270533, "epoch": 3039, "lr": 2.8199948052254345e-09}
{"train_loss": 0.04305974394083023, "global_step": 270534, "epoch": 3039, "lr": 2.8138410299582974e-09}
{"train_loss": 0.01700763963162899, "global_step": 270535, "epoch": 3039, "lr": 2.807693976208947e-09}
{"train_loss": 0.07785988599061966, "global_step": 270536, "epoch": 3039, "lr": 2.801553643971833e-09}
{"train_loss": 0.09993565827608109, "global_step": 270537, "epoch": 3039, "lr": 2.7954200332469538e-09}
{"train_loss": 0.05461821332573891, "global_step": 270538, "epoch": 3039, "lr": 2.7892931440343107e-09}
{"train_loss": 0.09272877126932144, "global_step": 270539, "epoch": 3039, "lr": 2.783172976339454e-09}
{"train_loss": 0.03157086670398712, "global_step": 270540, "epoch": 3039, "lr": 2.7770595301623848e-09}
{"train_loss": 0.03993351384997368, "global_step": 270541, "epoch": 3039, "lr": 2.770952805503102e-09}
{"train_loss": 0.0159318670630455, "global_step": 270542, "epoch": 3039, "lr": 2.764852802361606e-09}
{"train_loss": 0.03637157753109932, "global_step": 270543, "epoch": 3039, "lr": 2.758759520737897e-09}
{"train_loss": 0.0721239298582077, "global_step": 270544, "epoch": 3039, "lr": 2.7526729606319744e-09}
{"train_loss": 0.01920287497341633, "global_step": 270545, "epoch": 3039, "lr": 2.74659312204939e-09}
{"train_loss": 0.031079303473234177, "global_step": 270546, "epoch": 3039, "lr": 2.7405200049901436e-09}
{"train_loss": 0.04689382016658783, "global_step": 270547, "epoch": 3039, "lr": 2.734453609448684e-09}
{"train_loss": 0.04754266515374184, "global_step": 270548, "epoch": 3039, "lr": 2.728393935430562e-09}
{"train_loss": 0.07103593647480011, "global_step": 270549, "epoch": 3039, "lr": 2.7223409829413295e-09}
{"train_loss": 0.07227402180433273, "global_step": 270550, "epoch": 3039, "lr": 2.7162947519698833e-09}
{"train_loss": 0.005788602400571108, "global_step": 270551, "epoch": 3039, "lr": 2.7102552425273263e-09}
{"train_loss": 0.024316148832440376, "global_step": 270552, "epoch": 3039, "lr": 2.7042224546136585e-09}
{"train_loss": 0.02916383184492588, "global_step": 270553, "epoch": 3039, "lr": 2.6981963882233284e-09}
{"train_loss": 0.03644714131951332, "global_step": 270554, "epoch": 3039, "lr": 2.6921770433618876e-09}
{"train_loss": 0.017072293907403946, "global_step": 270555, "epoch": 3039, "lr": 2.6861644200237848e-09}
{"train_loss": 0.04741562530398369, "global_step": 270556, "epoch": 3039, "lr": 2.6801585182201216e-09}
{"train_loss": 0.05951004847884178, "global_step": 270557, "epoch": 3039, "lr": 2.674159337950899e-09}
{"train_loss": 0.07478060573339462, "global_step": 270558, "epoch": 3039, "lr": 2.668166879205014e-09}
{"train_loss": 0.04491570447519254, "global_step": 270559, "epoch": 3039, "lr": 2.6621811419991203e-09, "val_loss": 9.61519718170166}
{"train_loss": 0.038589686155319214, "global_step": 270560, "epoch": 3040, "lr": 2.656202126316565e-09}
{"train_loss": 0.06334001570940018, "global_step": 270561, "epoch": 3040, "lr": 2.6502298321740004e-09}
{"train_loss": 0.08493297547101974, "global_step": 270562, "epoch": 3040, "lr": 2.6442642595658763e-09}
{"train_loss": 0.020548654720187187, "global_step": 270563, "epoch": 3040, "lr": 2.638305408486641e-09}
{"train_loss": 0.041188180446624756, "global_step": 270564, "epoch": 3040, "lr": 2.632353278947397e-09}
{"train_loss": 0.08166038990020752, "global_step": 270565, "epoch": 3040, "lr": 2.6264078709481443e-09}
{"train_loss": 0.0341770276427269, "global_step": 270566, "epoch": 3040, "lr": 2.620469184483332e-09}
{"train_loss": 0.05680057406425476, "global_step": 270567, "epoch": 3040, "lr": 2.6145372195585105e-09}
{"train_loss": 0.007001462392508984, "global_step": 270568, "epoch": 3040, "lr": 2.6086119761681295e-09}
{"train_loss": 0.0401630736887455, "global_step": 270569, "epoch": 3040, "lr": 2.6026934543232906e-09}
{"train_loss": 0.026495449244976044, "global_step": 270570, "epoch": 3040, "lr": 2.596781654012892e-09}
{"train_loss": 0.039990879595279694, "global_step": 270571, "epoch": 3040, "lr": 2.5908765752480357e-09}
{"train_loss": 0.0895242840051651, "global_step": 270572, "epoch": 3040, "lr": 2.584978218023171e-09}
{"train_loss": 0.06531227380037308, "global_step": 270573, "epoch": 3040, "lr": 2.5790865823438482e-09}
{"train_loss": 0.10945156961679459, "global_step": 270574, "epoch": 3040, "lr": 2.573201668210068e-09}
{"train_loss": 0.02616584300994873, "global_step": 270575, "epoch": 3040, "lr": 2.567323475616279e-09}
{"train_loss": 0.03720064461231232, "global_step": 270576, "epoch": 3040, "lr": 2.5614520045735837e-09}
{"train_loss": 0.024506717920303345, "global_step": 270577, "epoch": 3040, "lr": 2.555587255070879e-09}
{"train_loss": 0.024527983739972115, "global_step": 270578, "epoch": 3040, "lr": 2.5497292271192684e-09}
{"train_loss": 0.03934759274125099, "global_step": 270579, "epoch": 3040, "lr": 2.5438779207132e-09}
{"train_loss": 0.06331496685743332, "global_step": 270580, "epoch": 3040, "lr": 2.5380333358582254e-09}
{"train_loss": 0.029555153101682663, "global_step": 270581, "epoch": 3040, "lr": 2.5321954725487927e-09}
{"train_loss": 0.03745675086975098, "global_step": 270582, "epoch": 3040, "lr": 2.526364330796005e-09}
{"train_loss": 0.01721956394612789, "global_step": 270583, "epoch": 3040, "lr": 2.5205399105887593e-09}
{"train_loss": 0.06799823045730591, "global_step": 270584, "epoch": 3040, "lr": 2.5147222119326073e-09}
{"train_loss": 0.02672579325735569, "global_step": 270585, "epoch": 3040, "lr": 2.5089112348331e-09}
{"train_loss": 0.010763753205537796, "global_step": 270586, "epoch": 3040, "lr": 2.5031069792846863e-09}
{"train_loss": 0.04003370180726051, "global_step": 270587, "epoch": 3040, "lr": 2.4973094452873657e-09}
{"train_loss": 0.09914404153823853, "global_step": 270588, "epoch": 3040, "lr": 2.49151863284669e-09}
{"train_loss": 0.05670730024576187, "global_step": 270589, "epoch": 3040, "lr": 2.4857345419626587e-09}
{"train_loss": 0.0575077161192894, "global_step": 270590, "epoch": 3040, "lr": 2.4799571726352723e-09}
{"train_loss": 0.03303666040301323, "global_step": 270591, "epoch": 3040, "lr": 2.474186524870081e-09}
{"train_loss": 0.06177143752574921, "global_step": 270592, "epoch": 3040, "lr": 2.4684225986559838e-09}
{"train_loss": 0.03675515577197075, "global_step": 270593, "epoch": 3040, "lr": 2.4626653940040824e-09}
{"train_loss": 0.025554684922099113, "global_step": 270594, "epoch": 3040, "lr": 2.456914910908825e-09}
{"train_loss": 0.054847925901412964, "global_step": 270595, "epoch": 3040, "lr": 2.451171149375764e-09}
{"train_loss": 0.020598800852894783, "global_step": 270596, "epoch": 3040, "lr": 2.445434109399347e-09}
{"train_loss": 0.009598803706467152, "global_step": 270597, "epoch": 3040, "lr": 2.4397037909906773e-09}
{"train_loss": 0.08091289550065994, "global_step": 270598, "epoch": 3040, "lr": 2.433980194144203e-09}
{"train_loss": 0.023791605606675148, "global_step": 270599, "epoch": 3040, "lr": 2.4282633188543734e-09}
{"train_loss": 0.09303157031536102, "global_step": 270600, "epoch": 3040, "lr": 2.422553165137842e-09}
{"train_loss": 0.016854487359523773, "global_step": 270601, "epoch": 3040, "lr": 2.416849732977955e-09}
{"train_loss": 0.07556486874818802, "global_step": 270602, "epoch": 3040, "lr": 2.411153022391366e-09}
{"train_loss": 0.04082319885492325, "global_step": 270603, "epoch": 3040, "lr": 2.4054630333669727e-09}
{"train_loss": 0.07259852439165115, "global_step": 270604, "epoch": 3040, "lr": 2.399779765910326e-09}
{"train_loss": 0.05887257680296898, "global_step": 270605, "epoch": 3040, "lr": 2.3941032200214266e-09}
{"train_loss": 0.02614668197929859, "global_step": 270606, "epoch": 3040, "lr": 2.3884333957002734e-09}
{"train_loss": 0.05650464817881584, "global_step": 270607, "epoch": 3040, "lr": 2.382770292952419e-09}
{"train_loss": 0.016099072992801666, "global_step": 270608, "epoch": 3040, "lr": 2.3771139117723107e-09}
{"train_loss": 0.0255962572991848, "global_step": 270609, "epoch": 3040, "lr": 2.3714642521599494e-09}
{"train_loss": 0.025378018617630005, "global_step": 270610, "epoch": 3040, "lr": 2.3658213141208863e-09}
{"train_loss": 0.029752153903245926, "global_step": 270611, "epoch": 3040, "lr": 2.360185097660672e-09}
{"train_loss": 0.012730947695672512, "global_step": 270612, "epoch": 3040, "lr": 2.3545556027682046e-09}
{"train_loss": 0.11197519302368164, "global_step": 270613, "epoch": 3040, "lr": 2.3489328294490353e-09}
{"train_loss": 0.009454306215047836, "global_step": 270614, "epoch": 3040, "lr": 2.343316777708715e-09}
{"train_loss": 0.07527606934309006, "global_step": 270615, "epoch": 3040, "lr": 2.3377074475416926e-09}
{"train_loss": 0.03339884802699089, "global_step": 270616, "epoch": 3040, "lr": 2.3321048389479684e-09}
{"train_loss": 0.022750576958060265, "global_step": 270617, "epoch": 3040, "lr": 2.326508951933093e-09}
{"train_loss": 0.040885258466005325, "global_step": 270618, "epoch": 3040, "lr": 2.3209197864970667e-09}
{"train_loss": 0.04404429346323013, "global_step": 270619, "epoch": 3040, "lr": 2.3153373426398896e-09}
{"train_loss": 0.07229471951723099, "global_step": 270620, "epoch": 3040, "lr": 2.3097616203615613e-09}
{"train_loss": 0.04746229574084282, "global_step": 270621, "epoch": 3040, "lr": 2.3041926196620824e-09}
{"train_loss": 0.04627237468957901, "global_step": 270622, "epoch": 3040, "lr": 2.298630340547003e-09}
{"train_loss": 0.059129901230335236, "global_step": 270623, "epoch": 3040, "lr": 2.2930747830107737e-09}
{"train_loss": 0.055949319154024124, "global_step": 270624, "epoch": 3040, "lr": 2.287525947058944e-09}
{"train_loss": 0.023987185209989548, "global_step": 270625, "epoch": 3040, "lr": 2.2819838326859633e-09}
{"train_loss": 0.12111344188451767, "global_step": 270626, "epoch": 3040, "lr": 2.276448439902934e-09}
{"train_loss": 0.05324617773294449, "global_step": 270627, "epoch": 3040, "lr": 2.2709197686987538e-09}
{"train_loss": 0.02258864790201187, "global_step": 270628, "epoch": 3040, "lr": 2.2653978190845247e-09}
{"train_loss": 0.01965256594121456, "global_step": 270629, "epoch": 3040, "lr": 2.2598825910546962e-09}
{"train_loss": 0.03622684255242348, "global_step": 270630, "epoch": 3040, "lr": 2.2543740846092674e-09}
{"train_loss": 0.013993820175528526, "global_step": 270631, "epoch": 3040, "lr": 2.2488722997537903e-09}
{"train_loss": 0.06288569420576096, "global_step": 270632, "epoch": 3040, "lr": 2.2433772364827133e-09}
{"train_loss": 0.054457779973745346, "global_step": 270633, "epoch": 3040, "lr": 2.2378888948071384e-09}
{"train_loss": 0.046345826238393784, "global_step": 270634, "epoch": 3040, "lr": 2.232407274715964e-09}
{"train_loss": 0.04181984066963196, "global_step": 270635, "epoch": 3040, "lr": 2.226932376214741e-09}
{"train_loss": 0.020604651421308517, "global_step": 270636, "epoch": 3040, "lr": 2.22146419930902e-09}
{"train_loss": 0.021368416026234627, "global_step": 270637, "epoch": 3040, "lr": 2.216002743993251e-09}
{"train_loss": 0.08648607134819031, "global_step": 270638, "epoch": 3040, "lr": 2.2105480102674325e-09}
{"train_loss": 0.012150061316788197, "global_step": 270639, "epoch": 3040, "lr": 2.205099998142668e-09}
{"train_loss": 0.021948209032416344, "global_step": 270640, "epoch": 3040, "lr": 2.1996587076023033e-09}
{"train_loss": 0.017116352915763855, "global_step": 270641, "epoch": 3040, "lr": 2.1942241386629925e-09}
{"train_loss": 0.0435514859855175, "global_step": 270642, "epoch": 3040, "lr": 2.188796291319184e-09}
{"train_loss": 0.0643250122666359, "global_step": 270643, "epoch": 3040, "lr": 2.1833751655708777e-09}
{"train_loss": 0.08408845216035843, "global_step": 270644, "epoch": 3040, "lr": 2.177960761418074e-09}
{"train_loss": 0.04064653068780899, "global_step": 270645, "epoch": 3040, "lr": 2.172553078866324e-09}
{"train_loss": 0.05216812714934349, "global_step": 270646, "epoch": 3040, "lr": 2.1671521179100763e-09}
{"train_loss": 0.046255163848400116, "global_step": 270647, "epoch": 3040, "lr": 2.161757878554882e-09}
{"train_loss": 0.045413899685392216, "global_step": 270648, "epoch": 3040, "lr": 2.1563703607951903e-09, "val_loss": 9.614675521850586, "train_action_mse_error": 29.055252075195312}
{"train_loss": 0.03380128741264343, "global_step": 270649, "epoch": 3041, "lr": 2.1509895646421028e-09}
{"train_loss": 0.04412219673395157, "global_step": 270650, "epoch": 3041, "lr": 2.1456154900845183e-09}
{"train_loss": 0.09602221846580505, "global_step": 270651, "epoch": 3041, "lr": 2.1402481371335383e-09}
{"train_loss": 0.09384606033563614, "global_step": 270652, "epoch": 3041, "lr": 2.1348875057836116e-09}
{"train_loss": 0.049360744655132294, "global_step": 270653, "epoch": 3041, "lr": 2.1295335960402894e-09}
{"train_loss": 0.028656819835305214, "global_step": 270654, "epoch": 3041, "lr": 2.124186407898021e-09}
{"train_loss": 0.03556981682777405, "global_step": 270655, "epoch": 3041, "lr": 2.1188459413623575e-09}
{"train_loss": 0.03193395584821701, "global_step": 270656, "epoch": 3041, "lr": 2.113512196433298e-09}
{"train_loss": 0.025633588433265686, "global_step": 270657, "epoch": 3041, "lr": 2.108185173105293e-09}
{"train_loss": 0.03312964364886284, "global_step": 270658, "epoch": 3041, "lr": 2.102864871389443e-09}
{"train_loss": 0.04399928078055382, "global_step": 270659, "epoch": 3041, "lr": 2.0975512912801975e-09}
{"train_loss": 0.024309419095516205, "global_step": 270660, "epoch": 3041, "lr": 2.092244432783108e-09}
{"train_loss": 0.05072590336203575, "global_step": 270661, "epoch": 3041, "lr": 2.0869442958926235e-09}
{"train_loss": 0.05716968700289726, "global_step": 270662, "epoch": 3041, "lr": 2.0816508806087432e-09}
{"train_loss": 0.035507239401340485, "global_step": 270663, "epoch": 3041, "lr": 2.0763641869370187e-09}
{"train_loss": 0.0342601053416729, "global_step": 270664, "epoch": 3041, "lr": 2.07108421487745e-09}
{"train_loss": 0.07331804931163788, "global_step": 270665, "epoch": 3041, "lr": 2.0658109644300374e-09}
{"train_loss": 0.06755181401968002, "global_step": 270666, "epoch": 3041, "lr": 2.0605444356003313e-09}
{"train_loss": 0.0435975044965744, "global_step": 270667, "epoch": 3041, "lr": 2.0552846283772297e-09}
{"train_loss": 0.06395546346902847, "global_step": 270668, "epoch": 3041, "lr": 2.050031542771835e-09}
{"train_loss": 0.017581909894943237, "global_step": 270669, "epoch": 3041, "lr": 2.0447851787785966e-09}
{"train_loss": 0.06007008254528046, "global_step": 270670, "epoch": 3041, "lr": 2.039545536408616e-09}
{"train_loss": 0.02386314608156681, "global_step": 270671, "epoch": 3041, "lr": 2.0343126156452397e-09}
{"train_loss": 0.017658965662121773, "global_step": 270672, "epoch": 3041, "lr": 2.0290864165051216e-09}
{"train_loss": 0.043867163360118866, "global_step": 270673, "epoch": 3041, "lr": 2.02386693898271e-09}
{"train_loss": 0.05200216919183731, "global_step": 270674, "epoch": 3041, "lr": 2.018654183078006e-09}
{"train_loss": 0.04263908416032791, "global_step": 270675, "epoch": 3041, "lr": 2.0134481487910083e-09}
{"train_loss": 0.02217332273721695, "global_step": 270676, "epoch": 3041, "lr": 2.008248836121718e-09}
{"train_loss": 0.07438946515321732, "global_step": 270677, "epoch": 3041, "lr": 2.003056245075685e-09}
{"train_loss": 0.027838116511702538, "global_step": 270678, "epoch": 3041, "lr": 1.9978703756529106e-09}
{"train_loss": 0.0408635213971138, "global_step": 270679, "epoch": 3041, "lr": 1.9926912278478427e-09}
{"train_loss": 0.04973699152469635, "global_step": 270680, "epoch": 3041, "lr": 1.987518801671584e-09}
{"train_loss": 0.024526186287403107, "global_step": 270681, "epoch": 3041, "lr": 1.982353097113032e-09}
{"train_loss": 0.07649081945419312, "global_step": 270682, "epoch": 3041, "lr": 1.9771941141777383e-09}
{"train_loss": 0.03744003176689148, "global_step": 270683, "epoch": 3041, "lr": 1.972041852871254e-09}
{"train_loss": 0.019838063046336174, "global_step": 270684, "epoch": 3041, "lr": 1.966896313188027e-09}
{"train_loss": 0.05767018347978592, "global_step": 270685, "epoch": 3041, "lr": 1.9617574951336093e-09}
{"train_loss": 0.019141197204589844, "global_step": 270686, "epoch": 3041, "lr": 1.956625398708001e-09}
{"train_loss": 0.06418188661336899, "global_step": 270687, "epoch": 3041, "lr": 1.95150002390565e-09}
{"train_loss": 0.05525951460003853, "global_step": 270688, "epoch": 3041, "lr": 1.9463813707321088e-09}
{"train_loss": 0.04602137207984924, "global_step": 270689, "epoch": 3041, "lr": 1.9412694391873764e-09}
{"train_loss": 0.02426989935338497, "global_step": 270690, "epoch": 3041, "lr": 1.9361642292714533e-09}
{"train_loss": 0.031977325677871704, "global_step": 270691, "epoch": 3041, "lr": 1.9310657409843392e-09}
{"train_loss": 0.02341541089117527, "global_step": 270692, "epoch": 3041, "lr": 1.9259739743315853e-09}
{"train_loss": 0.04526631161570549, "global_step": 270693, "epoch": 3041, "lr": 1.9208889293131915e-09}
{"train_loss": 0.03034335933625698, "global_step": 270694, "epoch": 3041, "lr": 1.915810605918056e-09}
{"train_loss": 0.042657434940338135, "global_step": 270695, "epoch": 3041, "lr": 1.9107390041628316e-09}
{"train_loss": 0.04725699871778488, "global_step": 270696, "epoch": 3041, "lr": 1.9056741240419674e-09}
{"train_loss": 0.02815769426524639, "global_step": 270697, "epoch": 3041, "lr": 1.9006159655554635e-09}
{"train_loss": 0.028968077152967453, "global_step": 270698, "epoch": 3041, "lr": 1.8955645286977685e-09}
{"train_loss": 0.05633971095085144, "global_step": 270699, "epoch": 3041, "lr": 1.8905198134799852e-09}
{"train_loss": 0.03714793547987938, "global_step": 270700, "epoch": 3041, "lr": 1.885481819902113e-09}
{"train_loss": 0.011112553998827934, "global_step": 270701, "epoch": 3041, "lr": 1.8804505479586008e-09}
{"train_loss": 0.05044224113225937, "global_step": 270702, "epoch": 3041, "lr": 1.8754259976494492e-09}
{"train_loss": 0.0230439193546772, "global_step": 270703, "epoch": 3041, "lr": 1.870408168980209e-09}
{"train_loss": 0.0579308457672596, "global_step": 270704, "epoch": 3041, "lr": 1.8653970619508798e-09}
{"train_loss": 0.06016337871551514, "global_step": 270705, "epoch": 3041, "lr": 1.8603926765670133e-09}
{"train_loss": 0.02165287174284458, "global_step": 270706, "epoch": 3041, "lr": 1.8553950128175067e-09}
{"train_loss": 0.04395546391606331, "global_step": 270707, "epoch": 3041, "lr": 1.8504040707079118e-09}
{"train_loss": 0.07979732751846313, "global_step": 270708, "epoch": 3041, "lr": 1.8454198502437792e-09}
{"train_loss": 0.04451098293066025, "global_step": 270709, "epoch": 3041, "lr": 1.840442351419558e-09}
{"train_loss": 0.06813310086727142, "global_step": 270710, "epoch": 3041, "lr": 1.8354715742407991e-09}
{"train_loss": 0.08291290700435638, "global_step": 270711, "epoch": 3041, "lr": 1.8305075187019516e-09}
{"train_loss": 0.0539059191942215, "global_step": 270712, "epoch": 3041, "lr": 1.8255501848141177e-09}
{"train_loss": 0.04256228730082512, "global_step": 270713, "epoch": 3041, "lr": 1.8205995725661951e-09}
{"train_loss": 0.0433393269777298, "global_step": 270714, "epoch": 3041, "lr": 1.815655681963735e-09}
{"train_loss": 0.025439199060201645, "global_step": 270715, "epoch": 3041, "lr": 1.8107185130122883e-09}
{"train_loss": 0.04239128902554512, "global_step": 270716, "epoch": 3041, "lr": 1.8057880657063043e-09}
{"train_loss": 0.029867997393012047, "global_step": 270717, "epoch": 3041, "lr": 1.8008643400457826e-09}
{"train_loss": 0.041100118309259415, "global_step": 270718, "epoch": 3041, "lr": 1.7959473360362745e-09}
{"train_loss": 0.021153084933757782, "global_step": 270719, "epoch": 3041, "lr": 1.7910370536722288e-09}
{"train_loss": 0.08325573056936264, "global_step": 270720, "epoch": 3041, "lr": 1.7861334929591966e-09}
{"train_loss": 0.029298122972249985, "global_step": 270721, "epoch": 3041, "lr": 1.7812366539027292e-09}
{"train_loss": 0.04758116975426674, "global_step": 270722, "epoch": 3041, "lr": 1.7763465364917242e-09}
{"train_loss": 0.04660356417298317, "global_step": 270723, "epoch": 3041, "lr": 1.7714631407317328e-09}
{"train_loss": 0.0373251736164093, "global_step": 270724, "epoch": 3041, "lr": 1.7665864666283061e-09}
{"train_loss": 0.029770202934741974, "global_step": 270725, "epoch": 3041, "lr": 1.7617165141758929e-09}
{"train_loss": 0.03155127167701721, "global_step": 270726, "epoch": 3041, "lr": 1.7568532833744934e-09}
{"train_loss": 0.06024667248129845, "global_step": 270727, "epoch": 3041, "lr": 1.7519967742296584e-09}
{"train_loss": 0.031809091567993164, "global_step": 270728, "epoch": 3041, "lr": 1.747146986741388e-09}
{"train_loss": 0.06179371476173401, "global_step": 270729, "epoch": 3041, "lr": 1.7423039209041313e-09}
{"train_loss": 0.06300343573093414, "global_step": 270730, "epoch": 3041, "lr": 1.7374675767289905e-09}
{"train_loss": 0.07554043084383011, "global_step": 270731, "epoch": 3041, "lr": 1.7326379542104143e-09}
{"train_loss": 0.037551119923591614, "global_step": 270732, "epoch": 3041, "lr": 1.7278150533428517e-09}
{"train_loss": 0.040422473102808, "global_step": 270733, "epoch": 3041, "lr": 1.7229988741429559e-09}
{"train_loss": 0.05962776020169258, "global_step": 270734, "epoch": 3041, "lr": 1.7181894165940736e-09}
{"train_loss": 0.0317881740629673, "global_step": 270735, "epoch": 3041, "lr": 1.7133866807073074e-09}
{"train_loss": 0.07881027460098267, "global_step": 270736, "epoch": 3041, "lr": 1.7085906664826567e-09}
{"train_loss": 0.044341401019123164, "global_step": 270737, "epoch": 3041, "lr": 1.703801373920122e-09, "val_loss": 9.685140609741211}
{"train_loss": 0.04535141959786415, "global_step": 270738, "epoch": 3042, "lr": 1.6990188030141519e-09}
{"train_loss": 0.08904167264699936, "global_step": 270739, "epoch": 3042, "lr": 1.6942429537758486e-09}
{"train_loss": 0.02900708094239235, "global_step": 270740, "epoch": 3042, "lr": 1.68947382619411e-09}
{"train_loss": 0.01577455922961235, "global_step": 270741, "epoch": 3042, "lr": 1.6847114202800385e-09}
{"train_loss": 0.04793691262602806, "global_step": 270742, "epoch": 3042, "lr": 1.6799557360280827e-09}
{"train_loss": 0.02904173918068409, "global_step": 270743, "epoch": 3042, "lr": 1.6752067734437937e-09}
{"train_loss": 0.027605468407273293, "global_step": 270744, "epoch": 3042, "lr": 1.6704645325216206e-09}
{"train_loss": 0.04330418258905411, "global_step": 270745, "epoch": 3042, "lr": 1.6657290132671144e-09}
{"train_loss": 0.028709622099995613, "global_step": 270746, "epoch": 3042, "lr": 1.6610002156802752e-09}
{"train_loss": 0.07957955449819565, "global_step": 270747, "epoch": 3042, "lr": 1.6562781397611028e-09}
{"train_loss": 0.05297337844967842, "global_step": 270748, "epoch": 3042, "lr": 1.6515627855095972e-09}
{"train_loss": 0.03668827563524246, "global_step": 270749, "epoch": 3042, "lr": 1.6468541529202075e-09}
{"train_loss": 0.021707477048039436, "global_step": 270750, "epoch": 3042, "lr": 1.642152242009587e-09}
{"train_loss": 0.07239485532045364, "global_step": 270751, "epoch": 3042, "lr": 1.637457052761082e-09}
{"train_loss": 0.04851085692644119, "global_step": 270752, "epoch": 3042, "lr": 1.6327685851857955e-09}
{"train_loss": 0.0415806844830513, "global_step": 270753, "epoch": 3042, "lr": 1.6280868392837268e-09}
{"train_loss": 0.06949988752603531, "global_step": 270754, "epoch": 3042, "lr": 1.623411815049325e-09}
{"train_loss": 0.0278824120759964, "global_step": 270755, "epoch": 3042, "lr": 1.6187435124936924e-09}
{"train_loss": 0.05350438132882118, "global_step": 270756, "epoch": 3042, "lr": 1.6140819316057265e-09}
{"train_loss": 0.030612587928771973, "global_step": 270757, "epoch": 3042, "lr": 1.6094270723909789e-09}
{"train_loss": 0.03643212094902992, "global_step": 270758, "epoch": 3042, "lr": 1.6047789348550002e-09}
{"train_loss": 0.06184716895222664, "global_step": 270759, "epoch": 3042, "lr": 1.6001375189866886e-09}
{"train_loss": 0.024632081389427185, "global_step": 270760, "epoch": 3042, "lr": 1.5955028248026972e-09}
{"train_loss": 0.041654668748378754, "global_step": 270761, "epoch": 3042, "lr": 1.5908748522863725e-09}
{"train_loss": 0.02716163732111454, "global_step": 270762, "epoch": 3042, "lr": 1.5862536014543683e-09}
{"train_loss": 0.04142715781927109, "global_step": 270763, "epoch": 3042, "lr": 1.581639072295582e-09}
{"train_loss": 0.0453714020550251, "global_step": 270764, "epoch": 3042, "lr": 1.5770312648100138e-09}
{"train_loss": 0.021679528057575226, "global_step": 270765, "epoch": 3042, "lr": 1.5724301790087658e-09}
{"train_loss": 0.044024378061294556, "global_step": 270766, "epoch": 3042, "lr": 1.567835814886287e-09}
{"train_loss": 0.04660874605178833, "global_step": 270767, "epoch": 3042, "lr": 1.5632481724425774e-09}
{"train_loss": 0.04604654386639595, "global_step": 270768, "epoch": 3042, "lr": 1.558667251683188e-09}
{"train_loss": 0.08446632325649261, "global_step": 270769, "epoch": 3042, "lr": 1.5540930526025676e-09}
{"train_loss": 0.03255084529519081, "global_step": 270770, "epoch": 3042, "lr": 1.5495255752007166e-09}
{"train_loss": 0.025455649942159653, "global_step": 270771, "epoch": 3042, "lr": 1.5449648194831856e-09}
{"train_loss": 0.04173397272825241, "global_step": 270772, "epoch": 3042, "lr": 1.540410785449975e-09}
{"train_loss": 0.040746476501226425, "global_step": 270773, "epoch": 3042, "lr": 1.5358634730955335e-09}
{"train_loss": 0.06859778612852097, "global_step": 270774, "epoch": 3042, "lr": 1.5313228824309633e-09}
{"train_loss": 0.01104147918522358, "global_step": 270775, "epoch": 3042, "lr": 1.5267890134507134e-09}
{"train_loss": 0.024247443303465843, "global_step": 270776, "epoch": 3042, "lr": 1.5222618661492326e-09}
{"train_loss": 0.031072406098246574, "global_step": 270777, "epoch": 3042, "lr": 1.5177414405376233e-09}
{"train_loss": 0.013459917157888412, "global_step": 270778, "epoch": 3042, "lr": 1.5132277366158854e-09}
{"train_loss": 0.022517971694469452, "global_step": 270779, "epoch": 3042, "lr": 1.5087207543784675e-09}
{"train_loss": 0.020452793687582016, "global_step": 270780, "epoch": 3042, "lr": 1.50422049382537e-09}
{"train_loss": 0.06607171148061752, "global_step": 270781, "epoch": 3042, "lr": 1.499726954967695e-09}
{"train_loss": 0.08188840001821518, "global_step": 270782, "epoch": 3042, "lr": 1.4952401377943404e-09}
{"train_loss": 0.024848347529768944, "global_step": 270783, "epoch": 3042, "lr": 1.490760042310857e-09}
{"train_loss": 0.058084651827812195, "global_step": 270784, "epoch": 3042, "lr": 1.486286668517245e-09}
{"train_loss": 0.03556416556239128, "global_step": 270785, "epoch": 3042, "lr": 1.4818200164190554e-09}
{"train_loss": 0.013997034169733524, "global_step": 270786, "epoch": 3042, "lr": 1.4773600860051862e-09}
{"train_loss": 0.03264614939689636, "global_step": 270787, "epoch": 3042, "lr": 1.4729068772867394e-09}
{"train_loss": 0.06584344059228897, "global_step": 270788, "epoch": 3042, "lr": 1.468460390263715e-09}
{"train_loss": 0.04323649778962135, "global_step": 270789, "epoch": 3042, "lr": 1.4640206249305622e-09}
{"train_loss": 0.04548724368214607, "global_step": 270790, "epoch": 3042, "lr": 1.4595875812928317e-09}
{"train_loss": 0.047236550599336624, "global_step": 270791, "epoch": 3042, "lr": 1.4551612593449726e-09}
{"train_loss": 0.05165763944387436, "global_step": 270792, "epoch": 3042, "lr": 1.4507416590980872e-09}
{"train_loss": 0.03329114243388176, "global_step": 270793, "epoch": 3042, "lr": 1.446328780541073e-09}
{"train_loss": 0.038142409175634384, "global_step": 270794, "epoch": 3042, "lr": 1.4419226236850326e-09}
{"train_loss": 0.051101550459861755, "global_step": 270795, "epoch": 3042, "lr": 1.4375231885244144e-09}
{"train_loss": 0.028220221400260925, "global_step": 270796, "epoch": 3042, "lr": 1.433130475059219e-09}
{"train_loss": 0.07599867135286331, "global_step": 270797, "epoch": 3042, "lr": 1.428744483294997e-09}
{"train_loss": 0.11030346155166626, "global_step": 270798, "epoch": 3042, "lr": 1.4243652132261976e-09}
{"train_loss": 0.05354227125644684, "global_step": 270799, "epoch": 3042, "lr": 1.4199926648583717e-09}
{"train_loss": 0.053743626922369, "global_step": 270800, "epoch": 3042, "lr": 1.4156268381859683e-09}
{"train_loss": 0.033290375024080276, "global_step": 270801, "epoch": 3042, "lr": 1.4112677332200897e-09}
{"train_loss": 0.0363362692296505, "global_step": 270802, "epoch": 3042, "lr": 1.4069153499551845e-09}
{"train_loss": 0.022611405700445175, "global_step": 270803, "epoch": 3042, "lr": 1.402569688385702e-09}
{"train_loss": 0.04510825127363205, "global_step": 270804, "epoch": 3042, "lr": 1.3982307485227442e-09}
{"train_loss": 0.024775544181466103, "global_step": 270805, "epoch": 3042, "lr": 1.39389853036076e-09}
{"train_loss": 0.09021684527397156, "global_step": 270806, "epoch": 3042, "lr": 1.3895730339053005e-09}
{"train_loss": 0.02111702226102352, "global_step": 270807, "epoch": 3042, "lr": 1.3852542591508144e-09}
{"train_loss": 0.017445536330342293, "global_step": 270808, "epoch": 3042, "lr": 1.380942206097302e-09}
{"train_loss": 0.06544119864702225, "global_step": 270809, "epoch": 3042, "lr": 1.3766368747503144e-09}
{"train_loss": 0.027817295864224434, "global_step": 270810, "epoch": 3042, "lr": 1.3723382651154027e-09}
{"train_loss": 0.058550119400024414, "global_step": 270811, "epoch": 3042, "lr": 1.3680463771814644e-09}
{"train_loss": 0.05057962238788605, "global_step": 270812, "epoch": 3042, "lr": 1.363761210954051e-09}
{"train_loss": 0.019617358222603798, "global_step": 270813, "epoch": 3042, "lr": 1.3594827664331621e-09}
{"train_loss": 0.027008363977074623, "global_step": 270814, "epoch": 3042, "lr": 1.3552110436187981e-09}
{"train_loss": 0.04761692136526108, "global_step": 270815, "epoch": 3042, "lr": 1.35094604251651e-09}
{"train_loss": 0.04244829714298248, "global_step": 270816, "epoch": 3042, "lr": 1.3466877631207465e-09}
{"train_loss": 0.04030874744057655, "global_step": 270817, "epoch": 3042, "lr": 1.3424362054370588e-09}
{"train_loss": 0.043280523270368576, "global_step": 270818, "epoch": 3042, "lr": 1.338191369465447e-09}
{"train_loss": 0.019387202337384224, "global_step": 270819, "epoch": 3042, "lr": 1.3339532552003597e-09}
{"train_loss": 0.05175742506980896, "global_step": 270820, "epoch": 3042, "lr": 1.3297218626473484e-09}
{"train_loss": 0.05268386751413345, "global_step": 270821, "epoch": 3042, "lr": 1.325497191806413e-09}
{"train_loss": 0.06116338074207306, "global_step": 270822, "epoch": 3042, "lr": 1.3212792426775534e-09}
{"train_loss": 0.018049612641334534, "global_step": 270823, "epoch": 3042, "lr": 1.3170680152607695e-09}
{"train_loss": 0.03275975212454796, "global_step": 270824, "epoch": 3042, "lr": 1.3128635095560615e-09}
{"train_loss": 0.01151295192539692, "global_step": 270825, "epoch": 3042, "lr": 1.3086657255689804e-09}
{"train_loss": 0.04191534921233909, "global_step": 270826, "epoch": 3042, "lr": 1.3044746632939753e-09, "val_loss": 9.546091079711914}
{"train_loss": 0.056973639875650406, "global_step": 270827, "epoch": 3043, "lr": 1.300290322736597e-09}
{"train_loss": 0.03960712254047394, "global_step": 270828, "epoch": 3043, "lr": 1.2961127038968456e-09}
{"train_loss": 0.04401451721787453, "global_step": 270829, "epoch": 3043, "lr": 1.2919418067691702e-09}
{"train_loss": 0.04110701382160187, "global_step": 270830, "epoch": 3043, "lr": 1.2877776313591217e-09}
{"train_loss": 0.044555164873600006, "global_step": 270831, "epoch": 3043, "lr": 1.2836201776667e-09}
{"train_loss": 0.048926934599876404, "global_step": 270832, "epoch": 3043, "lr": 1.2794694456919054e-09}
{"train_loss": 0.043449945747852325, "global_step": 270833, "epoch": 3043, "lr": 1.2753254354402888e-09}
{"train_loss": 0.0542653426527977, "global_step": 270834, "epoch": 3043, "lr": 1.271188146900748e-09}
{"train_loss": 0.05198213830590248, "global_step": 270835, "epoch": 3043, "lr": 1.2670575800843854e-09}
{"train_loss": 0.011074463836848736, "global_step": 270836, "epoch": 3043, "lr": 1.2629337349912008e-09}
{"train_loss": 0.023305842652916908, "global_step": 270837, "epoch": 3043, "lr": 1.258816611615643e-09}
{"train_loss": 0.02347630076110363, "global_step": 270838, "epoch": 3043, "lr": 1.2547062099632634e-09}
{"train_loss": 0.05585571378469467, "global_step": 270839, "epoch": 3043, "lr": 1.2506025300285107e-09}
{"train_loss": 0.029195616021752357, "global_step": 270840, "epoch": 3043, "lr": 1.2465055718169361e-09}
{"train_loss": 0.027519231662154198, "global_step": 270841, "epoch": 3043, "lr": 1.2424153353340905e-09}
{"train_loss": 0.06918247044086456, "global_step": 270842, "epoch": 3043, "lr": 1.238331820568872e-09}
{"train_loss": 0.019670480862259865, "global_step": 270843, "epoch": 3043, "lr": 1.2342550275323827e-09}
{"train_loss": 0.043834276497364044, "global_step": 270844, "epoch": 3043, "lr": 1.2301849562135203e-09}
{"train_loss": 0.03478091210126877, "global_step": 270845, "epoch": 3043, "lr": 1.2261216066289384e-09}
{"train_loss": 0.026852084323763847, "global_step": 270846, "epoch": 3043, "lr": 1.222064978761983e-09}
{"train_loss": 0.009147657081484795, "global_step": 270847, "epoch": 3043, "lr": 1.2180150726293082e-09}
{"train_loss": 0.03833239525556564, "global_step": 270848, "epoch": 3043, "lr": 1.2139718882142604e-09}
{"train_loss": 0.05254432559013367, "global_step": 270849, "epoch": 3043, "lr": 1.2099354255334927e-09}
{"train_loss": 0.02171292155981064, "global_step": 270850, "epoch": 3043, "lr": 1.2059056845759032e-09}
{"train_loss": 0.03743967413902283, "global_step": 270851, "epoch": 3043, "lr": 1.2018826653525938e-09}
{"train_loss": 0.04023782163858414, "global_step": 270852, "epoch": 3043, "lr": 1.1978663678524626e-09}
{"train_loss": 0.02506210468709469, "global_step": 270853, "epoch": 3043, "lr": 1.1938567920866117e-09}
{"train_loss": 0.05029090866446495, "global_step": 270854, "epoch": 3043, "lr": 1.1898539380494899e-09}
{"train_loss": 0.029388051480054855, "global_step": 270855, "epoch": 3043, "lr": 1.1858578057410973e-09}
{"train_loss": 0.05017763376235962, "global_step": 270856, "epoch": 3043, "lr": 1.1818683951669852e-09}
{"train_loss": 0.06118056923151016, "global_step": 270857, "epoch": 3043, "lr": 1.177885706321602e-09}
{"train_loss": 0.02424120157957077, "global_step": 270858, "epoch": 3043, "lr": 1.1739097392104991e-09}
{"train_loss": 0.014231501147150993, "global_step": 270859, "epoch": 3043, "lr": 1.1699404938281256e-09}
{"train_loss": 0.06845858693122864, "global_step": 270860, "epoch": 3043, "lr": 1.1659779701855833e-09}
{"train_loss": 0.09694590419530869, "global_step": 270861, "epoch": 3043, "lr": 1.1620221682717703e-09}
{"train_loss": 0.0612172894179821, "global_step": 270862, "epoch": 3043, "lr": 1.1580730880922374e-09}
{"train_loss": 0.05899457633495331, "global_step": 270863, "epoch": 3043, "lr": 1.1541307296525361e-09}
{"train_loss": 0.02934742346405983, "global_step": 270864, "epoch": 3043, "lr": 1.150195092941564e-09}
{"train_loss": 0.07025256752967834, "global_step": 270865, "epoch": 3043, "lr": 1.1462661779704232e-09}
{"train_loss": 0.0699247494339943, "global_step": 270866, "epoch": 3043, "lr": 1.1423439847391137e-09}
{"train_loss": 0.09687666594982147, "global_step": 270867, "epoch": 3043, "lr": 1.1384285132365335e-09}
{"train_loss": 0.050701163709163666, "global_step": 270868, "epoch": 3043, "lr": 1.1345197634793358e-09}
{"train_loss": 0.0435686856508255, "global_step": 270869, "epoch": 3043, "lr": 1.1306177354564185e-09}
{"train_loss": 0.061339542269706726, "global_step": 270870, "epoch": 3043, "lr": 1.1267224291733325e-09}
{"train_loss": 0.023953642696142197, "global_step": 270871, "epoch": 3043, "lr": 1.1228338446245268e-09}
{"train_loss": 0.04313956946134567, "global_step": 270872, "epoch": 3043, "lr": 1.1189519818211035e-09}
{"train_loss": 0.018282907083630562, "global_step": 270873, "epoch": 3043, "lr": 1.1150768407575118e-09}
{"train_loss": 0.024320684373378754, "global_step": 270874, "epoch": 3043, "lr": 1.1112084214337515e-09}
{"train_loss": 0.048693422228097916, "global_step": 270875, "epoch": 3043, "lr": 1.1073467238498226e-09}
{"train_loss": 0.07814698666334152, "global_step": 270876, "epoch": 3043, "lr": 1.103491748005725e-09}
{"train_loss": 0.07396957278251648, "global_step": 270877, "epoch": 3043, "lr": 1.09964349390701e-09}
{"train_loss": 0.038706667721271515, "global_step": 270878, "epoch": 3043, "lr": 1.0958019615536773e-09}
{"train_loss": 0.01283432450145483, "global_step": 270879, "epoch": 3043, "lr": 1.0919671509401764e-09}
{"train_loss": 0.02267083153128624, "global_step": 270880, "epoch": 3043, "lr": 1.0881390620720577e-09}
{"train_loss": 0.06102699041366577, "global_step": 270881, "epoch": 3043, "lr": 1.0843176949437705e-09}
{"train_loss": 0.04928341507911682, "global_step": 270882, "epoch": 3043, "lr": 1.080503049566417e-09}
{"train_loss": 0.060679614543914795, "global_step": 270883, "epoch": 3043, "lr": 1.0766951259288948e-09}
{"train_loss": 0.05811253935098648, "global_step": 270884, "epoch": 3043, "lr": 1.0728939240423064e-09}
{"train_loss": 0.06546498835086823, "global_step": 270885, "epoch": 3043, "lr": 1.0690994438955493e-09}
{"train_loss": 0.021488131955266, "global_step": 270886, "epoch": 3043, "lr": 1.0653116854997257e-09}
{"train_loss": 0.04790835082530975, "global_step": 270887, "epoch": 3043, "lr": 1.061530648854836e-09}
{"train_loss": 0.09519326686859131, "global_step": 270888, "epoch": 3043, "lr": 1.0577563339553288e-09}
{"train_loss": 0.05825832858681679, "global_step": 270889, "epoch": 3043, "lr": 1.053988740801204e-09}
{"train_loss": 0.041358936578035355, "global_step": 270890, "epoch": 3043, "lr": 1.0502278693980127e-09}
{"train_loss": 0.054885607212781906, "global_step": 270891, "epoch": 3043, "lr": 1.0464737197457553e-09}
{"train_loss": 0.0372367687523365, "global_step": 270892, "epoch": 3043, "lr": 1.0427262918388802e-09}
{"train_loss": 0.05148956924676895, "global_step": 270893, "epoch": 3043, "lr": 1.038985585682939e-09}
{"train_loss": 0.034435443580150604, "global_step": 270894, "epoch": 3043, "lr": 1.0352516012834824e-09}
{"train_loss": 0.018763013184070587, "global_step": 270895, "epoch": 3043, "lr": 1.0315243386294083e-09}
{"train_loss": 0.04473387822508812, "global_step": 270896, "epoch": 3043, "lr": 1.027803797731819e-09}
{"train_loss": 0.06901605427265167, "global_step": 270897, "epoch": 3043, "lr": 1.0240899785851632e-09}
{"train_loss": 0.03468703106045723, "global_step": 270898, "epoch": 3043, "lr": 1.0203828811894412e-09}
{"train_loss": 0.05744686722755432, "global_step": 270899, "epoch": 3043, "lr": 1.0166825055502039e-09}
{"train_loss": 0.01723206788301468, "global_step": 270900, "epoch": 3043, "lr": 1.0129888516619002e-09}
{"train_loss": 0.05132148787379265, "global_step": 270901, "epoch": 3043, "lr": 1.0093019195300813e-09}
{"train_loss": 0.03917510807514191, "global_step": 270902, "epoch": 3043, "lr": 1.005621709154747e-09}
{"train_loss": 0.07692122459411621, "global_step": 270903, "epoch": 3043, "lr": 1.0019482205303466e-09}
{"train_loss": 0.06434096395969391, "global_step": 270904, "epoch": 3043, "lr": 9.982814536624307e-10}
{"train_loss": 0.027292145416140556, "global_step": 270905, "epoch": 3043, "lr": 9.94621408556551e-10}
{"train_loss": 0.02536371909081936, "global_step": 270906, "epoch": 3043, "lr": 9.909680852016046e-10}
{"train_loss": 0.021779803559184074, "global_step": 270907, "epoch": 3043, "lr": 9.873214836086942e-10}
{"train_loss": 0.010144330561161041, "global_step": 270908, "epoch": 3043, "lr": 9.836816037667174e-10}
{"train_loss": 0.023627083748579025, "global_step": 270909, "epoch": 3043, "lr": 9.800484456867764e-10}
{"train_loss": 0.05483223870396614, "global_step": 270910, "epoch": 3043, "lr": 9.764220093688714e-10}
{"train_loss": 0.046649787575006485, "global_step": 270911, "epoch": 3043, "lr": 9.728022948074511e-10}
{"train_loss": 0.01790633611381054, "global_step": 270912, "epoch": 3043, "lr": 9.691893020080668e-10}
{"train_loss": 0.01792885735630989, "global_step": 270913, "epoch": 3043, "lr": 9.65583030965167e-10}
{"train_loss": 0.05143061280250549, "global_step": 270914, "epoch": 3043, "lr": 9.619834816843031e-10}
{"train_loss": 0.04372879896271095, "global_step": 270915, "epoch": 3043, "lr": 9.583906541654752e-10, "val_loss": 9.621437072753906}
{"train_loss": 0.02188713476061821, "global_step": 270916, "epoch": 3044, "lr": 9.548045484086832e-10}
{"train_loss": 0.03574949502944946, "global_step": 270917, "epoch": 3044, "lr": 9.51225164413927e-10}
{"train_loss": 0.06346195936203003, "global_step": 270918, "epoch": 3044, "lr": 9.476525021812065e-10}
{"train_loss": 0.04613141715526581, "global_step": 270919, "epoch": 3044, "lr": 9.440865617105222e-10}
{"train_loss": 0.04742875695228577, "global_step": 270920, "epoch": 3044, "lr": 9.405273430074246e-10}
{"train_loss": 0.015300233848392963, "global_step": 270921, "epoch": 3044, "lr": 9.369748460608119e-10}
{"train_loss": 0.0508432500064373, "global_step": 270922, "epoch": 3044, "lr": 9.334290708873373e-10}
{"train_loss": 0.053519442677497864, "global_step": 270923, "epoch": 3044, "lr": 9.298900174758984e-10}
{"train_loss": 0.041290685534477234, "global_step": 270924, "epoch": 3044, "lr": 9.263576858264955e-10}
{"train_loss": 0.021952396258711815, "global_step": 270925, "epoch": 3044, "lr": 9.228320759502307e-10}
{"train_loss": 0.03218749538064003, "global_step": 270926, "epoch": 3044, "lr": 9.193131878360018e-10}
{"train_loss": 0.027374060824513435, "global_step": 270927, "epoch": 3044, "lr": 9.158010214893598e-10}
{"train_loss": 0.07776132971048355, "global_step": 270928, "epoch": 3044, "lr": 9.122955769103047e-10}
{"train_loss": 0.03902576491236687, "global_step": 270929, "epoch": 3044, "lr": 9.087968540988367e-10}
{"train_loss": 0.03392775356769562, "global_step": 270930, "epoch": 3044, "lr": 9.053048530549557e-10}
{"train_loss": 0.06480619311332703, "global_step": 270931, "epoch": 3044, "lr": 9.018195737786617e-10}
{"train_loss": 0.04357224330306053, "global_step": 270932, "epoch": 3044, "lr": 8.983410162755057e-10}
{"train_loss": 0.04940560832619667, "global_step": 270933, "epoch": 3044, "lr": 8.948691805399368e-10}
{"train_loss": 0.01847163960337639, "global_step": 270934, "epoch": 3044, "lr": 8.914040665775059e-10}
{"train_loss": 0.029343049973249435, "global_step": 270935, "epoch": 3044, "lr": 8.87945674382662e-10}
{"train_loss": 0.05374394729733467, "global_step": 270936, "epoch": 3044, "lr": 8.844940039609562e-10}
{"train_loss": 0.03206317871809006, "global_step": 270937, "epoch": 3044, "lr": 8.810490553123885e-10}
{"train_loss": 0.047830983996391296, "global_step": 270938, "epoch": 3044, "lr": 8.776108284314077e-10}
{"train_loss": 0.05611756071448326, "global_step": 270939, "epoch": 3044, "lr": 8.741793233235651e-10}
{"train_loss": 0.014653973281383514, "global_step": 270940, "epoch": 3044, "lr": 8.707545399944118e-10}
{"train_loss": 0.02495148777961731, "global_step": 270941, "epoch": 3044, "lr": 8.673364784328453e-10}
{"train_loss": 0.09147866070270538, "global_step": 270942, "epoch": 3044, "lr": 8.639251386444169e-10}
{"train_loss": 0.032463982701301575, "global_step": 270943, "epoch": 3044, "lr": 8.605205206346778e-10}
{"train_loss": 0.08517295867204666, "global_step": 270944, "epoch": 3044, "lr": 8.571226243980767e-10}
{"train_loss": 0.032158028334379196, "global_step": 270945, "epoch": 3044, "lr": 8.537314499401649e-10}
{"train_loss": 0.05247274786233902, "global_step": 270946, "epoch": 3044, "lr": 8.503469972553912e-10}
{"train_loss": 0.01462398748844862, "global_step": 270947, "epoch": 3044, "lr": 8.469692663437556e-10}
{"train_loss": 0.04839478060603142, "global_step": 270948, "epoch": 3044, "lr": 8.435982572108092e-10}
{"train_loss": 0.015698058530688286, "global_step": 270949, "epoch": 3044, "lr": 8.40233969856552e-10}
{"train_loss": 0.05856921151280403, "global_step": 270950, "epoch": 3044, "lr": 8.36876404280984e-10}
{"train_loss": 0.011649995110929012, "global_step": 270951, "epoch": 3044, "lr": 8.335255604841053e-10}
{"train_loss": 0.07687502354383469, "global_step": 270952, "epoch": 3044, "lr": 8.301814384603645e-10}
{"train_loss": 0.0500037856400013, "global_step": 270953, "epoch": 3044, "lr": 8.268440382208642e-10}
{"train_loss": 0.04815052077174187, "global_step": 270954, "epoch": 3044, "lr": 8.23513359754502e-10}
{"train_loss": 0.03409381955862045, "global_step": 270955, "epoch": 3044, "lr": 8.2018940307238e-10}
{"train_loss": 0.011568634770810604, "global_step": 270956, "epoch": 3044, "lr": 8.168721681744985e-10}
{"train_loss": 0.051339298486709595, "global_step": 270957, "epoch": 3044, "lr": 8.135616550497549e-10}
{"train_loss": 0.04209670051932335, "global_step": 270958, "epoch": 3044, "lr": 8.102578637092517e-10}
{"train_loss": 0.038997501134872437, "global_step": 270959, "epoch": 3044, "lr": 8.069607941529889e-10}
{"train_loss": 0.03141913190484047, "global_step": 270960, "epoch": 3044, "lr": 8.036704463754152e-10}
{"train_loss": 0.053439054638147354, "global_step": 270961, "epoch": 3044, "lr": 8.003868203820819e-10}
{"train_loss": 0.038420747965574265, "global_step": 270962, "epoch": 3044, "lr": 7.971099161729889e-10}
{"train_loss": 0.017491938546299934, "global_step": 270963, "epoch": 3044, "lr": 7.938397337425852e-10}
{"train_loss": 0.0825885534286499, "global_step": 270964, "epoch": 3044, "lr": 7.905762731019728e-10}
{"train_loss": 0.026759108528494835, "global_step": 270965, "epoch": 3044, "lr": 7.873195342400497e-10}
{"train_loss": 0.0497460775077343, "global_step": 270966, "epoch": 3044, "lr": 7.84069517167918e-10}
{"train_loss": 0.025630995631217957, "global_step": 270967, "epoch": 3044, "lr": 7.808262218800266e-10}
{"train_loss": 0.044781606644392014, "global_step": 270968, "epoch": 3044, "lr": 7.775896483763756e-10}
{"train_loss": 0.042206697165966034, "global_step": 270969, "epoch": 3044, "lr": 7.743597966569649e-10}
{"train_loss": 0.032068800181150436, "global_step": 270970, "epoch": 3044, "lr": 7.711366667273457e-10}
{"train_loss": 0.07837743312120438, "global_step": 270971, "epoch": 3044, "lr": 7.679202585819667e-10}
{"train_loss": 0.06581352651119232, "global_step": 270972, "epoch": 3044, "lr": 7.647105722263793e-10}
{"train_loss": 0.0494760163128376, "global_step": 270973, "epoch": 3044, "lr": 7.615076076550321e-10}
{"train_loss": 0.04101374372839928, "global_step": 270974, "epoch": 3044, "lr": 7.583113648734763e-10}
{"train_loss": 0.015327662229537964, "global_step": 270975, "epoch": 3044, "lr": 7.55121843881712e-10}
{"train_loss": 0.04203964024782181, "global_step": 270976, "epoch": 3044, "lr": 7.519390446797392e-10}
{"train_loss": 0.052730802446603775, "global_step": 270977, "epoch": 3044, "lr": 7.487629672675577e-10}
{"train_loss": 0.04518873617053032, "global_step": 270978, "epoch": 3044, "lr": 7.455936116396167e-10}
{"train_loss": 0.03995799273252487, "global_step": 270979, "epoch": 3044, "lr": 7.424309778070182e-10}
{"train_loss": 0.008609690703451633, "global_step": 270980, "epoch": 3044, "lr": 7.392750657642112e-10}
{"train_loss": 0.01778888702392578, "global_step": 270981, "epoch": 3044, "lr": 7.361258755111955e-10}
{"train_loss": 0.046585384756326675, "global_step": 270982, "epoch": 3044, "lr": 7.329834070535225e-10}
{"train_loss": 0.055677738040685654, "global_step": 270983, "epoch": 3044, "lr": 7.298476603856407e-10}
{"train_loss": 0.02114500291645527, "global_step": 270984, "epoch": 3044, "lr": 7.267186355131018e-10}
{"train_loss": 0.06761554628610611, "global_step": 270985, "epoch": 3044, "lr": 7.235963324303541e-10}
{"train_loss": 0.02192062698304653, "global_step": 270986, "epoch": 3044, "lr": 7.204807511373979e-10}
{"train_loss": 0.015808600932359695, "global_step": 270987, "epoch": 3044, "lr": 7.173718916453354e-10}
{"train_loss": 0.0608493909239769, "global_step": 270988, "epoch": 3044, "lr": 7.142697539430643e-10}
{"train_loss": 0.048030029982328415, "global_step": 270989, "epoch": 3044, "lr": 7.111743380361357e-10}
{"train_loss": 0.07374297082424164, "global_step": 270990, "epoch": 3044, "lr": 7.080856439245498e-10}
{"train_loss": 0.03663887456059456, "global_step": 270991, "epoch": 3044, "lr": 7.050036716083063e-10}
{"train_loss": 0.04522891715168953, "global_step": 270992, "epoch": 3044, "lr": 7.019284210874055e-10}
{"train_loss": 0.032473646104335785, "global_step": 270993, "epoch": 3044, "lr": 6.988598923618473e-10}
{"train_loss": 0.03822614252567291, "global_step": 270994, "epoch": 3044, "lr": 6.957980854371826e-10}
{"train_loss": 0.07267679274082184, "global_step": 270995, "epoch": 3044, "lr": 6.927430003023094e-10}
{"train_loss": 0.048928480595350266, "global_step": 270996, "epoch": 3044, "lr": 6.89694636973881e-10}
{"train_loss": 0.08094670623540878, "global_step": 270997, "epoch": 3044, "lr": 6.86652995435244e-10}
{"train_loss": 0.032780371606349945, "global_step": 270998, "epoch": 3044, "lr": 6.836180756975008e-10}
{"train_loss": 0.04254806041717529, "global_step": 270999, "epoch": 3044, "lr": 6.805898777606512e-10}
{"train_loss": 0.07108862698078156, "global_step": 271000, "epoch": 3044, "lr": 6.775684016191442e-10}
{"train_loss": 0.04929328337311745, "global_step": 271001, "epoch": 3044, "lr": 6.745536472840818e-10}
{"train_loss": 0.06797216832637787, "global_step": 271002, "epoch": 3044, "lr": 6.715456147443621e-10}
{"train_loss": 0.023586003109812737, "global_step": 271003, "epoch": 3044, "lr": 6.685443039999851e-10}
{"train_loss": 0.04339135566914684, "global_step": 271004, "epoch": 3044, "lr": 6.655497150620526e-10, "val_loss": 9.667793273925781}
{"train_loss": 0.011712531559169292, "global_step": 271005, "epoch": 3045, "lr": 6.625618479250139e-10}
{"train_loss": 0.06292934715747833, "global_step": 271006, "epoch": 3045, "lr": 6.59580702588869e-10}
{"train_loss": 0.02173824980854988, "global_step": 271007, "epoch": 3045, "lr": 6.566062790536176e-10}
{"train_loss": 0.024749016389250755, "global_step": 271008, "epoch": 3045, "lr": 6.536385773248111e-10}
{"train_loss": 0.04078363627195358, "global_step": 271009, "epoch": 3045, "lr": 6.50677597391347e-10}
{"train_loss": 0.038335174322128296, "global_step": 271010, "epoch": 3045, "lr": 6.477233392698789e-10}
{"train_loss": 0.04416367784142494, "global_step": 271011, "epoch": 3045, "lr": 6.447758029437534e-10}
{"train_loss": 0.06247527897357941, "global_step": 271012, "epoch": 3045, "lr": 6.418349884240726e-10}
{"train_loss": 0.038073938339948654, "global_step": 271013, "epoch": 3045, "lr": 6.389008957108366e-10}
{"train_loss": 0.03861669823527336, "global_step": 271014, "epoch": 3045, "lr": 6.359735248040455e-10}
{"train_loss": 0.03620375320315361, "global_step": 271015, "epoch": 3045, "lr": 6.33052875698148e-10}
{"train_loss": 0.059533603489398956, "global_step": 271016, "epoch": 3045, "lr": 6.301389483986952e-10}
{"train_loss": 0.0561741478741169, "global_step": 271017, "epoch": 3045, "lr": 6.272317429056873e-10}
{"train_loss": 0.01947135291993618, "global_step": 271018, "epoch": 3045, "lr": 6.243312592191242e-10}
{"train_loss": 0.014417092315852642, "global_step": 271019, "epoch": 3045, "lr": 6.214374973390058e-10}
{"train_loss": 0.047653231769800186, "global_step": 271020, "epoch": 3045, "lr": 6.185504572653322e-10}
{"train_loss": 0.019682709127664566, "global_step": 271021, "epoch": 3045, "lr": 6.156701389981034e-10}
{"train_loss": 0.041602298617362976, "global_step": 271022, "epoch": 3045, "lr": 6.127965425373194e-10}
{"train_loss": 0.07948428392410278, "global_step": 271023, "epoch": 3045, "lr": 6.099296678885314e-10}
{"train_loss": 0.04700544849038124, "global_step": 271024, "epoch": 3045, "lr": 6.070695150461881e-10}
{"train_loss": 0.05162360519170761, "global_step": 271025, "epoch": 3045, "lr": 6.042160840158406e-10}
{"train_loss": 0.02337571792304516, "global_step": 271026, "epoch": 3045, "lr": 6.01369374791938e-10}
{"train_loss": 0.03090466372668743, "global_step": 271027, "epoch": 3045, "lr": 5.985293873800313e-10}
{"train_loss": 0.07464417815208435, "global_step": 271028, "epoch": 3045, "lr": 5.956961217745694e-10}
{"train_loss": 0.02343210019171238, "global_step": 271029, "epoch": 3045, "lr": 5.928695779811033e-10}
{"train_loss": 0.06196339428424835, "global_step": 271030, "epoch": 3045, "lr": 5.900497559996332e-10}
{"train_loss": 0.03405579924583435, "global_step": 271031, "epoch": 3045, "lr": 5.87236655830159e-10}
{"train_loss": 0.021171942353248596, "global_step": 271032, "epoch": 3045, "lr": 5.844302774726807e-10}
{"train_loss": 0.018194930627942085, "global_step": 271033, "epoch": 3045, "lr": 5.816306209271982e-10}
{"train_loss": 0.04511948674917221, "global_step": 271034, "epoch": 3045, "lr": 5.788376861881605e-10}
{"train_loss": 0.07224900275468826, "global_step": 271035, "epoch": 3045, "lr": 5.7605147326667e-10}
{"train_loss": 0.030108347535133362, "global_step": 271036, "epoch": 3045, "lr": 5.732719821627263e-10}
{"train_loss": 0.041811440140008926, "global_step": 271037, "epoch": 3045, "lr": 5.704992128652275e-10}
{"train_loss": 0.06297574192285538, "global_step": 271038, "epoch": 3045, "lr": 5.677331653852758e-10}
{"train_loss": 0.05387035012245178, "global_step": 271039, "epoch": 3045, "lr": 5.649738397173199e-10}
{"train_loss": 0.02830285020172596, "global_step": 271040, "epoch": 3045, "lr": 5.62221235866911e-10}
{"train_loss": 0.03928612545132637, "global_step": 271041, "epoch": 3045, "lr": 5.59475353828498e-10}
{"train_loss": 0.036873459815979004, "global_step": 271042, "epoch": 3045, "lr": 5.567361936076321e-10}
{"train_loss": 0.03917582705616951, "global_step": 271043, "epoch": 3045, "lr": 5.54003755198762e-10}
{"train_loss": 0.036701250821352005, "global_step": 271044, "epoch": 3045, "lr": 5.51278038607439e-10}
{"train_loss": 0.045181721448898315, "global_step": 271045, "epoch": 3045, "lr": 5.485590438336629e-10}
{"train_loss": 0.015094910748302937, "global_step": 271046, "epoch": 3045, "lr": 5.458467708774339e-10}
{"train_loss": 0.03457533195614815, "global_step": 271047, "epoch": 3045, "lr": 5.43141219738752e-10}
{"train_loss": 0.07293877005577087, "global_step": 271048, "epoch": 3045, "lr": 5.404423904176171e-10}
{"train_loss": 0.020676886662840843, "global_step": 271049, "epoch": 3045, "lr": 5.37750282914029e-10}
{"train_loss": 0.032917916774749756, "global_step": 271050, "epoch": 3045, "lr": 5.350648972279881e-10}
{"train_loss": 0.04225578159093857, "global_step": 271051, "epoch": 3045, "lr": 5.323862333594942e-10}
{"train_loss": 0.03146137669682503, "global_step": 271052, "epoch": 3045, "lr": 5.297142913140984e-10}
{"train_loss": 0.060708705335855484, "global_step": 271053, "epoch": 3045, "lr": 5.270490710862497e-10}
{"train_loss": 0.029750706627964973, "global_step": 271054, "epoch": 3045, "lr": 5.243905726759479e-10}
{"train_loss": 0.01838851347565651, "global_step": 271055, "epoch": 3045, "lr": 5.217387960887443e-10}
{"train_loss": 0.04402884095907211, "global_step": 271056, "epoch": 3045, "lr": 5.190937413190877e-10}
{"train_loss": 0.012872779741883278, "global_step": 271057, "epoch": 3045, "lr": 5.164554083725293e-10}
{"train_loss": 0.0325641930103302, "global_step": 271058, "epoch": 3045, "lr": 5.138237972435178e-10}
{"train_loss": 0.034608833491802216, "global_step": 271059, "epoch": 3045, "lr": 5.111989079376045e-10}
{"train_loss": 0.05767814815044403, "global_step": 271060, "epoch": 3045, "lr": 5.085807404547893e-10}
{"train_loss": 0.06768222153186798, "global_step": 271061, "epoch": 3045, "lr": 5.059692947950722e-10}
{"train_loss": 0.08703461289405823, "global_step": 271062, "epoch": 3045, "lr": 5.033645709584533e-10}
{"train_loss": 0.0375983901321888, "global_step": 271063, "epoch": 3045, "lr": 5.007665689393814e-10}
{"train_loss": 0.049730218946933746, "global_step": 271064, "epoch": 3045, "lr": 4.981752887489588e-10}
{"train_loss": 0.08021800220012665, "global_step": 271065, "epoch": 3045, "lr": 4.955907303816342e-10}
{"train_loss": 0.052830517292022705, "global_step": 271066, "epoch": 3045, "lr": 4.930128938374079e-10}
{"train_loss": 0.026493407785892487, "global_step": 271067, "epoch": 3045, "lr": 4.904417791162796e-10}
{"train_loss": 0.034564681351184845, "global_step": 271068, "epoch": 3045, "lr": 4.878773862182495e-10}
{"train_loss": 0.06357010453939438, "global_step": 271069, "epoch": 3045, "lr": 4.853197151488686e-10}
{"train_loss": 0.07272409647703171, "global_step": 271070, "epoch": 3045, "lr": 4.827687659025859e-10}
{"train_loss": 0.042090147733688354, "global_step": 271071, "epoch": 3045, "lr": 4.802245384849524e-10}
{"train_loss": 0.035903140902519226, "global_step": 271072, "epoch": 3045, "lr": 4.77687032890417e-10}
{"train_loss": 0.04653225839138031, "global_step": 271073, "epoch": 3045, "lr": 4.751562491245309e-10}
{"train_loss": 0.028649095445871353, "global_step": 271074, "epoch": 3045, "lr": 4.72632187181743e-10}
{"train_loss": 0.06553921103477478, "global_step": 271075, "epoch": 3045, "lr": 4.701148470676042e-10}
{"train_loss": 0.031231926754117012, "global_step": 271076, "epoch": 3045, "lr": 4.676042287821147e-10}
{"train_loss": 0.06238359957933426, "global_step": 271077, "epoch": 3045, "lr": 4.651003323252745e-10}
{"train_loss": 0.02900034934282303, "global_step": 271078, "epoch": 3045, "lr": 4.626031576915324e-10}
{"train_loss": 0.053798433393239975, "global_step": 271079, "epoch": 3045, "lr": 4.6011270489199066e-10}
{"train_loss": 0.016160383820533752, "global_step": 271080, "epoch": 3045, "lr": 4.576289739210982e-10}
{"train_loss": 0.020223475992679596, "global_step": 271081, "epoch": 3045, "lr": 4.551519647733038e-10}
{"train_loss": 0.03409413993358612, "global_step": 271082, "epoch": 3045, "lr": 4.526816774597098e-10}
{"train_loss": 0.029702724888920784, "global_step": 271083, "epoch": 3045, "lr": 4.5021811197476506e-10}
{"train_loss": 0.05818580836057663, "global_step": 271084, "epoch": 3045, "lr": 4.4776126831846955e-10}
{"train_loss": 0.04535298049449921, "global_step": 271085, "epoch": 3045, "lr": 4.453111464963744e-10}
{"train_loss": 0.04228636994957924, "global_step": 271086, "epoch": 3045, "lr": 4.428677465029285e-10}
{"train_loss": 0.046357639133930206, "global_step": 271087, "epoch": 3045, "lr": 4.4043106833813186e-10}
{"train_loss": 0.06726909428834915, "global_step": 271088, "epoch": 3045, "lr": 4.3800111200753557e-10}
{"train_loss": 0.06652803719043732, "global_step": 271089, "epoch": 3045, "lr": 4.3557787750558855e-10}
{"train_loss": 0.05415859818458557, "global_step": 271090, "epoch": 3045, "lr": 4.3316136483784185e-10}
{"train_loss": 0.01855296827852726, "global_step": 271091, "epoch": 3045, "lr": 4.307515740042956e-10}
{"train_loss": 0.02522524818778038, "global_step": 271092, "epoch": 3045, "lr": 4.2834850499939847e-10}
{"train_loss": 0.04190030414611101, "global_step": 271093, "epoch": 3045, "lr": 4.259521578342529e-10, "val_loss": 9.693907737731934, "train_action_mse_error": 26.338478088378906}
{"train_loss": 0.03087826631963253, "global_step": 271094, "epoch": 3046, "lr": 4.235625324977566e-10}
{"train_loss": 0.08571299910545349, "global_step": 271095, "epoch": 3046, "lr": 4.211796289954606e-10}
{"train_loss": 0.04828503355383873, "global_step": 271096, "epoch": 3046, "lr": 4.1880344732181386e-10}
{"train_loss": 0.07680630683898926, "global_step": 271097, "epoch": 3046, "lr": 4.164339874879186e-10}
{"train_loss": 0.05808026343584061, "global_step": 271098, "epoch": 3046, "lr": 4.1407124948822375e-10}
{"train_loss": 0.030628660693764687, "global_step": 271099, "epoch": 3046, "lr": 4.117152333227292e-10}
{"train_loss": 0.048376865684986115, "global_step": 271100, "epoch": 3046, "lr": 4.093659389969862e-10}
{"train_loss": 0.01436076033860445, "global_step": 271101, "epoch": 3046, "lr": 4.070233664998924e-10}
{"train_loss": 0.035812366753816605, "global_step": 271102, "epoch": 3046, "lr": 4.0468751584255006e-10}
{"train_loss": 0.023442333564162254, "global_step": 271103, "epoch": 3046, "lr": 4.023583870194081e-10}
{"train_loss": 0.04935245215892792, "global_step": 271104, "epoch": 3046, "lr": 4.000359800360176e-10}
{"train_loss": 0.035203967243433, "global_step": 271105, "epoch": 3046, "lr": 3.977202948868275e-10}
{"train_loss": 0.020726613700389862, "global_step": 271106, "epoch": 3046, "lr": 3.954113315718377e-10}
{"train_loss": 0.08544805645942688, "global_step": 271107, "epoch": 3046, "lr": 3.931090901021506e-10}
{"train_loss": 0.03445839509367943, "global_step": 271108, "epoch": 3046, "lr": 3.9081357046111266e-10}
{"train_loss": 0.02369445003569126, "global_step": 271109, "epoch": 3046, "lr": 3.885247726653773e-10}
{"train_loss": 0.03846723958849907, "global_step": 271110, "epoch": 3046, "lr": 3.862426967038424e-10}
{"train_loss": 0.03211928904056549, "global_step": 271111, "epoch": 3046, "lr": 3.839673425820589e-10}
{"train_loss": 0.052144963294267654, "global_step": 271112, "epoch": 3046, "lr": 3.816987103000269e-10}
{"train_loss": 0.07280483841896057, "global_step": 271113, "epoch": 3046, "lr": 3.794367998577464e-10}
{"train_loss": 0.09986996650695801, "global_step": 271114, "epoch": 3046, "lr": 3.7718161125521734e-10}
{"train_loss": 0.042273156344890594, "global_step": 271115, "epoch": 3046, "lr": 3.7493314449243977e-10}
{"train_loss": 0.05197823792695999, "global_step": 271116, "epoch": 3046, "lr": 3.726913995694137e-10}
{"train_loss": 0.07195258140563965, "global_step": 271117, "epoch": 3046, "lr": 3.7045637648613907e-10}
{"train_loss": 0.041018035262823105, "global_step": 271118, "epoch": 3046, "lr": 3.6822807524261594e-10}
{"train_loss": 0.03195428103208542, "global_step": 271119, "epoch": 3046, "lr": 3.660064958388443e-10}
{"train_loss": 0.05375194922089577, "global_step": 271120, "epoch": 3046, "lr": 3.637916382803752e-10}
{"train_loss": 0.03275679424405098, "global_step": 271121, "epoch": 3046, "lr": 3.615835025616576e-10}
{"train_loss": 0.01693630963563919, "global_step": 271122, "epoch": 3046, "lr": 3.593820886882426e-10}
{"train_loss": 0.0797802209854126, "global_step": 271123, "epoch": 3046, "lr": 3.57187396649028e-10}
{"train_loss": 0.03439982980489731, "global_step": 271124, "epoch": 3046, "lr": 3.549994264606671e-10}
{"train_loss": 0.016659632325172424, "global_step": 271125, "epoch": 3046, "lr": 3.5281817811205764e-10}
{"train_loss": 0.043712615966796875, "global_step": 271126, "epoch": 3046, "lr": 3.506436516087508e-10}
{"train_loss": 0.0884057804942131, "global_step": 271127, "epoch": 3046, "lr": 3.484758469451954e-10}
{"train_loss": 0.05783558636903763, "global_step": 271128, "epoch": 3046, "lr": 3.4631476412694265e-10}
{"train_loss": 0.021747268736362457, "global_step": 271129, "epoch": 3046, "lr": 3.441604031539924e-10}
{"train_loss": 0.04452696070075035, "global_step": 271130, "epoch": 3046, "lr": 3.4201276402079373e-10}
{"train_loss": 0.04292283579707146, "global_step": 271131, "epoch": 3046, "lr": 3.398718467384487e-10}
{"train_loss": 0.03048158437013626, "global_step": 271132, "epoch": 3046, "lr": 3.3773765129585524e-10}
{"train_loss": 0.09207092970609665, "global_step": 271133, "epoch": 3046, "lr": 3.356101777041154e-10}
{"train_loss": 0.04694196581840515, "global_step": 271134, "epoch": 3046, "lr": 3.3348942595212705e-10}
{"train_loss": 0.04092305898666382, "global_step": 271135, "epoch": 3046, "lr": 3.313753960454413e-10}
{"train_loss": 0.013881632126867771, "global_step": 271136, "epoch": 3046, "lr": 3.292680879896093e-10}
{"train_loss": 0.03437887504696846, "global_step": 271137, "epoch": 3046, "lr": 3.2716750177907984e-10}
{"train_loss": 0.0534689798951149, "global_step": 271138, "epoch": 3046, "lr": 3.25073637413853e-10}
{"train_loss": 0.035242047160863876, "global_step": 271139, "epoch": 3046, "lr": 3.2298649489392876e-10}
{"train_loss": 0.06847316771745682, "global_step": 271140, "epoch": 3046, "lr": 3.2090607422485817e-10}
{"train_loss": 0.03379509598016739, "global_step": 271141, "epoch": 3046, "lr": 3.188323754010902e-10}
{"train_loss": 0.018099071457982063, "global_step": 271142, "epoch": 3046, "lr": 3.1676539842262485e-10}
{"train_loss": 0.08720388263463974, "global_step": 271143, "epoch": 3046, "lr": 3.1470514329501323e-10}
{"train_loss": 0.05795977637171745, "global_step": 271144, "epoch": 3046, "lr": 3.126516100182553e-10}
{"train_loss": 0.07726181298494339, "global_step": 271145, "epoch": 3046, "lr": 3.106047985867999e-10}
{"train_loss": 0.04882877692580223, "global_step": 271146, "epoch": 3046, "lr": 3.085647090061983e-10}
{"train_loss": 0.036157190799713135, "global_step": 271147, "epoch": 3046, "lr": 3.065313412708992e-10}
{"train_loss": 0.04781954735517502, "global_step": 271148, "epoch": 3046, "lr": 3.045046953864539e-10}
{"train_loss": 0.033793434500694275, "global_step": 271149, "epoch": 3046, "lr": 3.0248477135286224e-10}
{"train_loss": 0.04520174115896225, "global_step": 271150, "epoch": 3046, "lr": 3.004715691701243e-10}
{"train_loss": 0.021775083616375923, "global_step": 271151, "epoch": 3046, "lr": 2.984650888382401e-10}
{"train_loss": 0.026908578351140022, "global_step": 271152, "epoch": 3046, "lr": 2.9646533035165846e-10}
{"train_loss": 0.014848188497126102, "global_step": 271153, "epoch": 3046, "lr": 2.944722937214817e-10}
{"train_loss": 0.04236074537038803, "global_step": 271154, "epoch": 3046, "lr": 2.9248597893660744e-10}
{"train_loss": 0.028875408694148064, "global_step": 271155, "epoch": 3046, "lr": 2.9050638600813804e-10}
{"train_loss": 0.037297554314136505, "global_step": 271156, "epoch": 3046, "lr": 2.885335149305224e-10}
{"train_loss": 0.053624771535396576, "global_step": 271157, "epoch": 3046, "lr": 2.8656736570376043e-10}
{"train_loss": 0.047403886914253235, "global_step": 271158, "epoch": 3046, "lr": 2.8460793832785216e-10}
{"train_loss": 0.058111533522605896, "global_step": 271159, "epoch": 3046, "lr": 2.8265523280279757e-10}
{"train_loss": 0.014282013289630413, "global_step": 271160, "epoch": 3046, "lr": 2.807092491341479e-10}
{"train_loss": 0.0604877695441246, "global_step": 271161, "epoch": 3046, "lr": 2.7876998731635185e-10}
{"train_loss": 0.04969248175621033, "global_step": 271162, "epoch": 3046, "lr": 2.7683744734940955e-10}
{"train_loss": 0.040917277336120605, "global_step": 271163, "epoch": 3046, "lr": 2.7491162923887204e-10}
{"train_loss": 0.010259416885674, "global_step": 271164, "epoch": 3046, "lr": 2.729925329847394e-10}
{"train_loss": 0.03761725127696991, "global_step": 271165, "epoch": 3046, "lr": 2.710801585814604e-10}
{"train_loss": 0.04072627052664757, "global_step": 271166, "epoch": 3046, "lr": 2.6917450602903514e-10}
{"train_loss": 0.08448993414640427, "global_step": 271167, "epoch": 3046, "lr": 2.672755753385658e-10}
{"train_loss": 0.02361445315182209, "global_step": 271168, "epoch": 3046, "lr": 2.6538336649895024e-10}
{"train_loss": 0.0527963861823082, "global_step": 271169, "epoch": 3046, "lr": 2.6349787951018835e-10}
{"train_loss": 0.11849173903465271, "global_step": 271170, "epoch": 3046, "lr": 2.6161911438338236e-10}
{"train_loss": 0.07586880773305893, "global_step": 271171, "epoch": 3046, "lr": 2.597470711074301e-10}
{"train_loss": 0.044501010328531265, "global_step": 271172, "epoch": 3046, "lr": 2.578817496878827e-10}
{"train_loss": 0.02577829547226429, "global_step": 271173, "epoch": 3046, "lr": 2.560231501247401e-10}
{"train_loss": 0.05170745030045509, "global_step": 271174, "epoch": 3046, "lr": 2.541712724180023e-10}
{"train_loss": 0.03997469320893288, "global_step": 271175, "epoch": 3046, "lr": 2.5232611656766935e-10}
{"train_loss": 0.03162063658237457, "global_step": 271176, "epoch": 3046, "lr": 2.5048768257374123e-10}
{"train_loss": 0.06726862490177155, "global_step": 271177, "epoch": 3046, "lr": 2.486559704362179e-10}
{"train_loss": 0.05436677485704422, "global_step": 271178, "epoch": 3046, "lr": 2.468309801550994e-10}
{"train_loss": 0.06423881649971008, "global_step": 271179, "epoch": 3046, "lr": 2.450127117359369e-10}
{"train_loss": 0.027945447713136673, "global_step": 271180, "epoch": 3046, "lr": 2.43201165167628e-10}
{"train_loss": 0.05231243371963501, "global_step": 271181, "epoch": 3046, "lr": 2.4139634046127514e-10}
{"train_loss": 0.04671242279576117, "global_step": 271182, "epoch": 3046, "lr": 2.395982376168782e-10, "val_loss": 9.578714370727539}
{"train_loss": 0.06413781642913818, "global_step": 271183, "epoch": 3047, "lr": 2.378068566233349e-10}
{"train_loss": 0.08290725201368332, "global_step": 271184, "epoch": 3047, "lr": 2.360221974917476e-10}
{"train_loss": 0.06443504989147186, "global_step": 271185, "epoch": 3047, "lr": 2.342442602221162e-10}
{"train_loss": 0.020979423075914383, "global_step": 271186, "epoch": 3047, "lr": 2.3247304480333853e-10}
{"train_loss": 0.05407385155558586, "global_step": 271187, "epoch": 3047, "lr": 2.307085512520679e-10}
{"train_loss": 0.04483355954289436, "global_step": 271188, "epoch": 3047, "lr": 2.2895077955720213e-10}
{"train_loss": 0.03249947354197502, "global_step": 271189, "epoch": 3047, "lr": 2.2719972972429226e-10}
{"train_loss": 0.04020669311285019, "global_step": 271190, "epoch": 3047, "lr": 2.2545540174778724e-10}
{"train_loss": 0.04473298788070679, "global_step": 271191, "epoch": 3047, "lr": 2.2371779563323814e-10}
{"train_loss": 0.04137849435210228, "global_step": 271192, "epoch": 3047, "lr": 2.2198691137509387e-10}
{"train_loss": 0.04280293360352516, "global_step": 271193, "epoch": 3047, "lr": 2.2026274898445666e-10}
{"train_loss": 0.03937675803899765, "global_step": 271194, "epoch": 3047, "lr": 2.1854530845022424e-10}
{"train_loss": 0.07717839628458023, "global_step": 271195, "epoch": 3047, "lr": 2.168345897779478e-10}
{"train_loss": 0.01652969792485237, "global_step": 271196, "epoch": 3047, "lr": 2.151305929731784e-10}
{"train_loss": 0.0481611005961895, "global_step": 271197, "epoch": 3047, "lr": 2.134333180248138e-10}
{"train_loss": 0.02332398295402527, "global_step": 271198, "epoch": 3047, "lr": 2.1174276493840518e-10}
{"train_loss": 0.01721031963825226, "global_step": 271199, "epoch": 3047, "lr": 2.1005893371395245e-10}
{"train_loss": 0.056357622146606445, "global_step": 271200, "epoch": 3047, "lr": 2.083818243514557e-10}
{"train_loss": 0.09617721289396286, "global_step": 271201, "epoch": 3047, "lr": 2.0671143685091488e-10}
{"train_loss": 0.04645306617021561, "global_step": 271202, "epoch": 3047, "lr": 2.0504777121233e-10}
{"train_loss": 0.020631929859519005, "global_step": 271203, "epoch": 3047, "lr": 2.0339082744125216e-10}
{"train_loss": 0.03403192758560181, "global_step": 271204, "epoch": 3047, "lr": 2.0174060553213025e-10}
{"train_loss": 0.042050182819366455, "global_step": 271205, "epoch": 3047, "lr": 2.000971054849643e-10}
{"train_loss": 0.07879467308521271, "global_step": 271206, "epoch": 3047, "lr": 1.9846032729975427e-10}
{"train_loss": 0.01750463806092739, "global_step": 271207, "epoch": 3047, "lr": 1.968302709820513e-10}
{"train_loss": 0.015403777360916138, "global_step": 271208, "epoch": 3047, "lr": 1.952069365263043e-10}
{"train_loss": 0.038242924958467484, "global_step": 271209, "epoch": 3047, "lr": 1.9359032393806432e-10}
{"train_loss": 0.034904297441244125, "global_step": 271210, "epoch": 3047, "lr": 1.9198043321178026e-10}
{"train_loss": 0.04210127145051956, "global_step": 271211, "epoch": 3047, "lr": 1.9037726434745217e-10}
{"train_loss": 0.04575541242957115, "global_step": 271212, "epoch": 3047, "lr": 1.8878081735063113e-10}
{"train_loss": 0.09466585516929626, "global_step": 271213, "epoch": 3047, "lr": 1.8719109222131715e-10}
{"train_loss": 0.02748100645840168, "global_step": 271214, "epoch": 3047, "lr": 1.856080889539591e-10}
{"train_loss": 0.01671016961336136, "global_step": 271215, "epoch": 3047, "lr": 1.840318075541081e-10}
{"train_loss": 0.021900789812207222, "global_step": 271216, "epoch": 3047, "lr": 1.8246224802176414e-10}
{"train_loss": 0.02713540941476822, "global_step": 271217, "epoch": 3047, "lr": 1.8089941035137614e-10}
{"train_loss": 0.027555400505661964, "global_step": 271218, "epoch": 3047, "lr": 1.793432945484952e-10}
{"train_loss": 0.023754429072141647, "global_step": 271219, "epoch": 3047, "lr": 1.7779390061312128e-10}
{"train_loss": 0.09618952870368958, "global_step": 271220, "epoch": 3047, "lr": 1.7625122854525445e-10}
{"train_loss": 0.030924296006560326, "global_step": 271221, "epoch": 3047, "lr": 1.7471527834489466e-10}
{"train_loss": 0.06628265976905823, "global_step": 271222, "epoch": 3047, "lr": 1.7318605001204192e-10}
{"train_loss": 0.058258626610040665, "global_step": 271223, "epoch": 3047, "lr": 1.716635435411451e-10}
{"train_loss": 0.049096088856458664, "global_step": 271224, "epoch": 3047, "lr": 1.7014775894330647e-10}
{"train_loss": 0.03215339407324791, "global_step": 271225, "epoch": 3047, "lr": 1.686386962129749e-10}
{"train_loss": 0.024994129315018654, "global_step": 271226, "epoch": 3047, "lr": 1.6713635534459926e-10}
{"train_loss": 0.0461437851190567, "global_step": 271227, "epoch": 3047, "lr": 1.656407363492818e-10}
{"train_loss": 0.043100256472826004, "global_step": 271228, "epoch": 3047, "lr": 1.6415183922147136e-10}
{"train_loss": 0.0891399085521698, "global_step": 271229, "epoch": 3047, "lr": 1.62669663961168e-10}
{"train_loss": 0.0533166378736496, "global_step": 271230, "epoch": 3047, "lr": 1.611942105739228e-10}
{"train_loss": 0.03271514177322388, "global_step": 271231, "epoch": 3047, "lr": 1.5972547904863355e-10}
{"train_loss": 0.04312008246779442, "global_step": 271232, "epoch": 3047, "lr": 1.5826346939640246e-10}
{"train_loss": 0.02572987973690033, "global_step": 271233, "epoch": 3047, "lr": 1.5680818161167843e-10}
{"train_loss": 0.0682755634188652, "global_step": 271234, "epoch": 3047, "lr": 1.5535961570001256e-10}
{"train_loss": 0.027637464925646782, "global_step": 271235, "epoch": 3047, "lr": 1.5391777165585376e-10}
{"train_loss": 0.037165567278862, "global_step": 271236, "epoch": 3047, "lr": 1.5248264947920199e-10}
{"train_loss": 0.004624093417078257, "global_step": 271237, "epoch": 3047, "lr": 1.510542491756084e-10}
{"train_loss": 0.040546439588069916, "global_step": 271238, "epoch": 3047, "lr": 1.4963257074507299e-10}
{"train_loss": 0.09481510519981384, "global_step": 271239, "epoch": 3047, "lr": 1.482176141820446e-10}
{"train_loss": 0.04583865776658058, "global_step": 271240, "epoch": 3047, "lr": 1.4680937948652328e-10}
{"train_loss": 0.05610369145870209, "global_step": 271241, "epoch": 3047, "lr": 1.4540786666406014e-10}
{"train_loss": 0.040431298315525055, "global_step": 271242, "epoch": 3047, "lr": 1.4401307571465518e-10}
{"train_loss": 0.051772259175777435, "global_step": 271243, "epoch": 3047, "lr": 1.4262500663275724e-10}
{"train_loss": 0.024490943178534508, "global_step": 271244, "epoch": 3047, "lr": 1.412436594294686e-10}
{"train_loss": 0.018778882920742035, "global_step": 271245, "epoch": 3047, "lr": 1.398690340881359e-10}
{"train_loss": 0.01978161558508873, "global_step": 271246, "epoch": 3047, "lr": 1.3850113062541247e-10}
{"train_loss": 0.009969583712518215, "global_step": 271247, "epoch": 3047, "lr": 1.3713994903019612e-10}
{"train_loss": 0.03764324262738228, "global_step": 271248, "epoch": 3047, "lr": 1.3578548931358903e-10}
{"train_loss": 0.03707500547170639, "global_step": 271249, "epoch": 3047, "lr": 1.34437751464489e-10}
{"train_loss": 0.01860031485557556, "global_step": 271250, "epoch": 3047, "lr": 1.3309673548844715e-10}
{"train_loss": 0.03540270775556564, "global_step": 271251, "epoch": 3047, "lr": 1.3176244138546345e-10}
{"train_loss": 0.049276843667030334, "global_step": 271252, "epoch": 3047, "lr": 1.3043486915553792e-10}
{"train_loss": 0.01887967437505722, "global_step": 271253, "epoch": 3047, "lr": 1.2911401879867058e-10}
{"train_loss": 0.04164697229862213, "global_step": 271254, "epoch": 3047, "lr": 1.277998903148614e-10}
{"train_loss": 0.029420463368296623, "global_step": 271255, "epoch": 3047, "lr": 1.2649248370411037e-10}
{"train_loss": 0.03976571559906006, "global_step": 271256, "epoch": 3047, "lr": 1.2519179896641752e-10}
{"train_loss": 0.036632515490055084, "global_step": 271257, "epoch": 3047, "lr": 1.2389783610178284e-10}
{"train_loss": 0.08372747153043747, "global_step": 271258, "epoch": 3047, "lr": 1.2261059511020635e-10}
{"train_loss": 0.05311695486307144, "global_step": 271259, "epoch": 3047, "lr": 1.213300759972391e-10}
{"train_loss": 0.038190387189388275, "global_step": 271260, "epoch": 3047, "lr": 1.2005627875733005e-10}
{"train_loss": 0.06569948047399521, "global_step": 271261, "epoch": 3047, "lr": 1.1878920339047916e-10}
{"train_loss": 0.05216849595308304, "global_step": 271262, "epoch": 3047, "lr": 1.1752884989668644e-10}
{"train_loss": 0.06785765290260315, "global_step": 271263, "epoch": 3047, "lr": 1.1627521828150301e-10}
{"train_loss": 0.08081667125225067, "global_step": 271264, "epoch": 3047, "lr": 1.1502830853937774e-10}
{"train_loss": 0.02125602401793003, "global_step": 271265, "epoch": 3047, "lr": 1.1378812067031063e-10}
{"train_loss": 0.05223408341407776, "global_step": 271266, "epoch": 3047, "lr": 1.1255465467985282e-10}
{"train_loss": 0.0247703418135643, "global_step": 271267, "epoch": 3047, "lr": 1.1132791056245318e-10}
{"train_loss": 0.06983712315559387, "global_step": 271268, "epoch": 3047, "lr": 1.1010788832366281e-10}
{"train_loss": 0.02494567260146141, "global_step": 271269, "epoch": 3047, "lr": 1.0889458795793061e-10}
{"train_loss": 0.04907253012061119, "global_step": 271270, "epoch": 3047, "lr": 1.0768800947080771e-10}
{"train_loss": 0.04407812311219868, "global_step": 271271, "epoch": 3047, "lr": 1.0648815285674297e-10, "val_loss": 9.74958324432373}
{"train_loss": 0.015429447405040264, "global_step": 271272, "epoch": 3048, "lr": 1.052950181212875e-10}
{"train_loss": 0.01749584451317787, "global_step": 271273, "epoch": 3048, "lr": 1.0410860525889022e-10}
{"train_loss": 0.015846548601984978, "global_step": 271274, "epoch": 3048, "lr": 1.0292891427510221e-10}
{"train_loss": 0.026460761204361916, "global_step": 271275, "epoch": 3048, "lr": 1.0175594516992348e-10}
{"train_loss": 0.04442794248461723, "global_step": 271276, "epoch": 3048, "lr": 1.0058969793780292e-10}
{"train_loss": 0.06605293601751328, "global_step": 271277, "epoch": 3048, "lr": 9.943017258429166e-11}
{"train_loss": 0.06643328815698624, "global_step": 271278, "epoch": 3048, "lr": 9.827736910938967e-11}
{"train_loss": 0.06087511405348778, "global_step": 271279, "epoch": 3048, "lr": 9.713128751309697e-11}
{"train_loss": 0.021766075864434242, "global_step": 271280, "epoch": 3048, "lr": 9.599192778986244e-11}
{"train_loss": 0.036102116107940674, "global_step": 271281, "epoch": 3048, "lr": 9.485928994523718e-11}
{"train_loss": 0.06254459917545319, "global_step": 271282, "epoch": 3048, "lr": 9.373337398477233e-11}
{"train_loss": 0.00960064958781004, "global_step": 271283, "epoch": 3048, "lr": 9.261417989736565e-11}
{"train_loss": 0.06276320666074753, "global_step": 271284, "epoch": 3048, "lr": 9.150170768301714e-11}
{"train_loss": 0.029226167127490044, "global_step": 271285, "epoch": 3048, "lr": 9.039595735282902e-11}
{"train_loss": 0.02550988458096981, "global_step": 271286, "epoch": 3048, "lr": 8.929692890125019e-11}
{"train_loss": 0.036986518651247025, "global_step": 271287, "epoch": 3048, "lr": 8.820462232828065e-11}
{"train_loss": 0.06284556537866592, "global_step": 271288, "epoch": 3048, "lr": 8.711903763392037e-11}
{"train_loss": 0.062307510524988174, "global_step": 271289, "epoch": 3048, "lr": 8.604017481816939e-11}
{"train_loss": 0.05674479156732559, "global_step": 271290, "epoch": 3048, "lr": 8.496803388102769e-11}
{"train_loss": 0.02815709076821804, "global_step": 271291, "epoch": 3048, "lr": 8.390261482249528e-11}
{"train_loss": 0.048805370926856995, "global_step": 271292, "epoch": 3048, "lr": 8.284391764257215e-11}
{"train_loss": 0.054307952523231506, "global_step": 271293, "epoch": 3048, "lr": 8.17919423412583e-11}
{"train_loss": 0.03736817091703415, "global_step": 271294, "epoch": 3048, "lr": 8.074668892410486e-11}
{"train_loss": 0.032604072242975235, "global_step": 271295, "epoch": 3048, "lr": 7.97081573855607e-11}
{"train_loss": 0.06064799055457115, "global_step": 271296, "epoch": 3048, "lr": 7.867634772562582e-11}
{"train_loss": 0.05362780764698982, "global_step": 271297, "epoch": 3048, "lr": 7.765125994430023e-11}
{"train_loss": 0.08310997486114502, "global_step": 271298, "epoch": 3048, "lr": 7.663289404158391e-11}
{"train_loss": 0.03771120682358742, "global_step": 271299, "epoch": 3048, "lr": 7.5621250023028e-11}
{"train_loss": 0.016248296946287155, "global_step": 271300, "epoch": 3048, "lr": 7.461632788308138e-11}
{"train_loss": 0.05033751577138901, "global_step": 271301, "epoch": 3048, "lr": 7.361812762729514e-11}
{"train_loss": 0.03168831765651703, "global_step": 271302, "epoch": 3048, "lr": 7.262664925011819e-11}
{"train_loss": 0.056228116154670715, "global_step": 271303, "epoch": 3048, "lr": 7.164189275155054e-11}
{"train_loss": 0.03209160640835762, "global_step": 271304, "epoch": 3048, "lr": 7.066385813714327e-11}
{"train_loss": 0.04524710401892662, "global_step": 271305, "epoch": 3048, "lr": 6.969254540134529e-11}
{"train_loss": 0.019630515947937965, "global_step": 271306, "epoch": 3048, "lr": 6.872795454415659e-11}
{"train_loss": 0.06855196505784988, "global_step": 271307, "epoch": 3048, "lr": 6.77700855711283e-11}
{"train_loss": 0.043466828763484955, "global_step": 271308, "epoch": 3048, "lr": 6.68189384822604e-11}
{"train_loss": 0.03225976228713989, "global_step": 271309, "epoch": 3048, "lr": 6.587451327200178e-11}
{"train_loss": 0.09898407012224197, "global_step": 271310, "epoch": 3048, "lr": 6.493680994590357e-11}
{"train_loss": 0.02784881554543972, "global_step": 271311, "epoch": 3048, "lr": 6.400582849841463e-11}
{"train_loss": 0.054420486092567444, "global_step": 271312, "epoch": 3048, "lr": 6.30815689350861e-11}
{"train_loss": 0.03282369673252106, "global_step": 271313, "epoch": 3048, "lr": 6.216403125036686e-11}
{"train_loss": 0.03892335668206215, "global_step": 271314, "epoch": 3048, "lr": 6.125321545535911e-11}
{"train_loss": 0.03970615565776825, "global_step": 271315, "epoch": 3048, "lr": 6.034912153340954e-11}
{"train_loss": 0.050812989473342896, "global_step": 271316, "epoch": 3048, "lr": 5.94517495011715e-11}
{"train_loss": 0.06076033040881157, "global_step": 271317, "epoch": 3048, "lr": 5.856109934754272e-11}
{"train_loss": 0.037527360022068024, "global_step": 271318, "epoch": 3048, "lr": 5.7677171078074357e-11}
{"train_loss": 0.019752608612179756, "global_step": 271319, "epoch": 3048, "lr": 5.6799964692766386e-11}
{"train_loss": 0.03761160001158714, "global_step": 271320, "epoch": 3048, "lr": 5.592948019161881e-11}
{"train_loss": 0.04083823412656784, "global_step": 271321, "epoch": 3048, "lr": 5.506571756908052e-11}
{"train_loss": 0.04489805921912193, "global_step": 271322, "epoch": 3048, "lr": 5.420867683070263e-11}
{"train_loss": 0.04486484080553055, "global_step": 271323, "epoch": 3048, "lr": 5.335835797648514e-11}
{"train_loss": 0.031562671065330505, "global_step": 271324, "epoch": 3048, "lr": 5.251476100642805e-11}
{"train_loss": 0.062063951045274734, "global_step": 271325, "epoch": 3048, "lr": 5.1677885920531354e-11}
{"train_loss": 0.03148844838142395, "global_step": 271326, "epoch": 3048, "lr": 5.084773271324395e-11}
{"train_loss": 0.02812742255628109, "global_step": 271327, "epoch": 3048, "lr": 5.002430139566805e-11}
{"train_loss": 0.03924747183918953, "global_step": 271328, "epoch": 3048, "lr": 4.920759195670144e-11}
{"train_loss": 0.01642140932381153, "global_step": 271329, "epoch": 3048, "lr": 4.839760440744634e-11}
{"train_loss": 0.028060007840394974, "global_step": 271330, "epoch": 3048, "lr": 4.7594338736800523e-11}
{"train_loss": 0.05669340491294861, "global_step": 271331, "epoch": 3048, "lr": 4.679779495586623e-11}
{"train_loss": 0.04312722757458687, "global_step": 271332, "epoch": 3048, "lr": 4.600797305354121e-11}
{"train_loss": 0.029069561511278152, "global_step": 271333, "epoch": 3048, "lr": 4.522487304092771e-11}
{"train_loss": 0.05879339203238487, "global_step": 271334, "epoch": 3048, "lr": 4.44484949124746e-11}
{"train_loss": 0.07608532905578613, "global_step": 271335, "epoch": 3048, "lr": 4.367883866263078e-11}
{"train_loss": 0.062460169196128845, "global_step": 271336, "epoch": 3048, "lr": 4.2915904302498476e-11}
{"train_loss": 0.06049126014113426, "global_step": 271337, "epoch": 3048, "lr": 4.2159691826526573e-11}
{"train_loss": 0.043690405786037445, "global_step": 271338, "epoch": 3048, "lr": 4.141020123471506e-11}
{"train_loss": 0.02385169081389904, "global_step": 271339, "epoch": 3048, "lr": 4.0667432527063955e-11}
{"train_loss": 0.05418550968170166, "global_step": 271340, "epoch": 3048, "lr": 3.993138570912436e-11}
{"train_loss": 0.015212995000183582, "global_step": 271341, "epoch": 3048, "lr": 3.9202060769794044e-11}
{"train_loss": 0.049597930163145065, "global_step": 271342, "epoch": 3048, "lr": 3.847945772017525e-11}
{"train_loss": 0.02886781468987465, "global_step": 271343, "epoch": 3048, "lr": 3.776357655471685e-11}
{"train_loss": 0.016425449401140213, "global_step": 271344, "epoch": 3048, "lr": 3.705441727341885e-11}
{"train_loss": 0.015603184700012207, "global_step": 271345, "epoch": 3048, "lr": 3.6351979881832366e-11}
{"train_loss": 0.03464630991220474, "global_step": 271346, "epoch": 3048, "lr": 3.565626437440628e-11}
{"train_loss": 0.0743396207690239, "global_step": 271347, "epoch": 3048, "lr": 3.496727075114059e-11}
{"train_loss": 0.04771358147263527, "global_step": 271348, "epoch": 3048, "lr": 3.4284999012035304e-11}
{"train_loss": 0.06200231611728668, "global_step": 271349, "epoch": 3048, "lr": 3.3609449162641524e-11}
{"train_loss": 0.03135320171713829, "global_step": 271350, "epoch": 3048, "lr": 3.294062119740815e-11}
{"train_loss": 0.05992212891578674, "global_step": 271351, "epoch": 3048, "lr": 3.227851511633517e-11}
{"train_loss": 0.05419261008501053, "global_step": 271352, "epoch": 3048, "lr": 3.162313092497371e-11}
{"train_loss": 0.055948540568351746, "global_step": 271353, "epoch": 3048, "lr": 3.0974468617772646e-11}
{"train_loss": 0.01585189253091812, "global_step": 271354, "epoch": 3048, "lr": 3.0332528200283095e-11}
{"train_loss": 0.03718063607811928, "global_step": 271355, "epoch": 3048, "lr": 2.969730966695394e-11}
{"train_loss": 0.038385938853025436, "global_step": 271356, "epoch": 3048, "lr": 2.906881301778519e-11}
{"train_loss": 0.08511364459991455, "global_step": 271357, "epoch": 3048, "lr": 2.844703825832795e-11}
{"train_loss": 0.01906457729637623, "global_step": 271358, "epoch": 3048, "lr": 2.7831985388582227e-11}
{"train_loss": 0.03222870081663132, "global_step": 271359, "epoch": 3048, "lr": 2.72236544029969e-11}
{"train_loss": 0.04266402118033572, "global_step": 271360, "epoch": 3048, "lr": 2.6622045301571975e-11, "val_loss": 9.589632987976074}
{"train_loss": 0.08342646062374115, "global_step": 271361, "epoch": 3049, "lr": 2.602715808985856e-11}
{"train_loss": 0.057387709617614746, "global_step": 271362, "epoch": 3049, "lr": 2.5438992762305546e-11}
{"train_loss": 0.06151581555604935, "global_step": 271363, "epoch": 3049, "lr": 2.4857549324464047e-11}
{"train_loss": 0.04428136348724365, "global_step": 271364, "epoch": 3049, "lr": 2.4282827776334062e-11}
{"train_loss": 0.07586285471916199, "global_step": 271365, "epoch": 3049, "lr": 2.3714828112364473e-11}
{"train_loss": 0.03677904233336449, "global_step": 271366, "epoch": 3049, "lr": 2.3153550332555286e-11}
{"train_loss": 0.032542888075113297, "global_step": 271367, "epoch": 3049, "lr": 2.2598994448008725e-11}
{"train_loss": 0.02914413996040821, "global_step": 271368, "epoch": 3049, "lr": 2.205116044207145e-11}
{"train_loss": 0.047996774315834045, "global_step": 271369, "epoch": 3049, "lr": 2.1510048331396804e-11}
{"train_loss": 0.06285418570041656, "global_step": 271370, "epoch": 3049, "lr": 2.0975658104882556e-11}
{"train_loss": 0.030885295942425728, "global_step": 271371, "epoch": 3049, "lr": 2.044798976807982e-11}
{"train_loss": 0.02646416611969471, "global_step": 271372, "epoch": 3049, "lr": 1.992704331543749e-11}
{"train_loss": 0.058660075068473816, "global_step": 271373, "epoch": 3049, "lr": 1.941281875250667e-11}
{"train_loss": 0.06866587698459625, "global_step": 271374, "epoch": 3049, "lr": 1.8905316079287362e-11}
{"train_loss": 0.05169622600078583, "global_step": 271375, "epoch": 3049, "lr": 1.840453529577957e-11}
{"train_loss": 0.050389159470796585, "global_step": 271376, "epoch": 3049, "lr": 1.7910476396432173e-11}
{"train_loss": 0.030197059735655785, "global_step": 271377, "epoch": 3049, "lr": 1.7423139386796294e-11}
{"train_loss": 0.051020435988903046, "global_step": 271378, "epoch": 3049, "lr": 1.6942524261320813e-11}
{"train_loss": 0.0664915069937706, "global_step": 271379, "epoch": 3049, "lr": 1.6468631031107963e-11}
{"train_loss": 0.04640601575374603, "global_step": 271380, "epoch": 3049, "lr": 1.6001459685055508e-11}
{"train_loss": 0.031911589205265045, "global_step": 271381, "epoch": 3049, "lr": 1.554101022871457e-11}
{"train_loss": 0.01777241937816143, "global_step": 271382, "epoch": 3049, "lr": 1.5087282662085144e-11}
{"train_loss": 0.032462067902088165, "global_step": 271383, "epoch": 3049, "lr": 1.464027697961612e-11}
{"train_loss": 0.04061353951692581, "global_step": 271384, "epoch": 3049, "lr": 1.4199993186858607e-11}
{"train_loss": 0.05046265199780464, "global_step": 271385, "epoch": 3049, "lr": 1.3766431289363723e-11}
{"train_loss": 0.013284561224281788, "global_step": 271386, "epoch": 3049, "lr": 1.3339591276029239e-11}
{"train_loss": 0.023195501416921616, "global_step": 271387, "epoch": 3049, "lr": 1.2919473146855155e-11}
{"train_loss": 0.042579080909490585, "global_step": 271388, "epoch": 3049, "lr": 1.2506076912943699e-11}
{"train_loss": 0.06405884772539139, "global_step": 271389, "epoch": 3049, "lr": 1.209940256319264e-11}
{"train_loss": 0.07679376751184464, "global_step": 271390, "epoch": 3049, "lr": 1.1699450108704213e-11}
{"train_loss": 0.04843109846115112, "global_step": 271391, "epoch": 3049, "lr": 1.1306219538376183e-11}
{"train_loss": 0.014848470687866211, "global_step": 271392, "epoch": 3049, "lr": 1.0919710857759669e-11}
{"train_loss": 0.10145122557878494, "global_step": 271393, "epoch": 3049, "lr": 1.0539924066854667e-11}
{"train_loss": 0.019337667152285576, "global_step": 271394, "epoch": 3049, "lr": 1.016685916566118e-11}
{"train_loss": 0.05142652988433838, "global_step": 271395, "epoch": 3049, "lr": 9.800516154179207e-12}
{"train_loss": 0.07987741380929947, "global_step": 271396, "epoch": 3049, "lr": 9.440895032408747e-12}
{"train_loss": 0.04185570776462555, "global_step": 271397, "epoch": 3049, "lr": 9.087995794798688e-12}
{"train_loss": 0.041114602237939835, "global_step": 271398, "epoch": 3049, "lr": 8.741818452451256e-12}
{"train_loss": 0.06622575968503952, "global_step": 271399, "epoch": 3049, "lr": 8.402362994264223e-12}
{"train_loss": 0.07635417580604553, "global_step": 271400, "epoch": 3049, "lr": 8.06962943133982e-12}
{"train_loss": 0.0744343027472496, "global_step": 271401, "epoch": 3049, "lr": 7.743617752575816e-12}
{"train_loss": 0.03216007724404335, "global_step": 271402, "epoch": 3049, "lr": 7.42432796907444e-12}
{"train_loss": 0.05972066894173622, "global_step": 271403, "epoch": 3049, "lr": 7.111760069733464e-12}
{"train_loss": 0.06587442010641098, "global_step": 271404, "epoch": 3049, "lr": 6.8059140656551164e-12}
{"train_loss": 0.047380007803440094, "global_step": 271405, "epoch": 3049, "lr": 6.506789945737169e-12}
{"train_loss": 0.03923090174794197, "global_step": 271406, "epoch": 3049, "lr": 6.214387715530734e-12}
{"train_loss": 0.05210055783390999, "global_step": 271407, "epoch": 3049, "lr": 5.928707380586928e-12}
{"train_loss": 0.03515344113111496, "global_step": 271408, "epoch": 3049, "lr": 5.649748929803522e-12}
{"train_loss": 0.018445730209350586, "global_step": 271409, "epoch": 3049, "lr": 5.377512374282745e-12}
{"train_loss": 0.03772379830479622, "global_step": 271410, "epoch": 3049, "lr": 5.1119977029223664e-12}
{"train_loss": 0.03745553642511368, "global_step": 271411, "epoch": 3049, "lr": 4.8532049268246174e-12}
{"train_loss": 0.053023844957351685, "global_step": 271412, "epoch": 3049, "lr": 4.6011340348872665e-12}
{"train_loss": 0.06555154174566269, "global_step": 271413, "epoch": 3049, "lr": 4.355785038212545e-12}
{"train_loss": 0.03744323551654816, "global_step": 271414, "epoch": 3049, "lr": 4.117157925698223e-12}
{"train_loss": 0.03447393700480461, "global_step": 271415, "epoch": 3049, "lr": 3.88525270844653e-12}
{"train_loss": 0.0386788472533226, "global_step": 271416, "epoch": 3049, "lr": 3.6600693809063505e-12}
{"train_loss": 0.043329354375600815, "global_step": 271417, "epoch": 3049, "lr": 3.44160793752657e-12}
{"train_loss": 0.014389563351869583, "global_step": 271418, "epoch": 3049, "lr": 3.2298683894094185e-12}
{"train_loss": 0.07404130697250366, "global_step": 271419, "epoch": 3049, "lr": 3.024850731003781e-12}
{"train_loss": 0.017431050539016724, "global_step": 271420, "epoch": 3049, "lr": 2.8265549623096575e-12}
{"train_loss": 0.03503527492284775, "global_step": 271421, "epoch": 3049, "lr": 2.6349810888781632e-12}
{"train_loss": 0.01681540161371231, "global_step": 271422, "epoch": 3049, "lr": 2.450129099607068e-12}
{"train_loss": 0.0627095103263855, "global_step": 271423, "epoch": 3049, "lr": 2.2719990000474868e-12}
{"train_loss": 0.041814710944890976, "global_step": 271424, "epoch": 3049, "lr": 2.1005907957505346e-12}
{"train_loss": 0.05427488684654236, "global_step": 271425, "epoch": 3049, "lr": 1.935904475613981e-12}
{"train_loss": 0.02239362709224224, "global_step": 271426, "epoch": 3049, "lr": 1.7779400507400567e-12}
{"train_loss": 0.01438729465007782, "global_step": 271427, "epoch": 3049, "lr": 1.6266975155776465e-12}
{"train_loss": 0.030870208516716957, "global_step": 271428, "epoch": 3049, "lr": 1.4821768645756351e-12}
{"train_loss": 0.08228832483291626, "global_step": 271429, "epoch": 3049, "lr": 1.3443781088362528e-12}
{"train_loss": 0.004878438543528318, "global_step": 271430, "epoch": 3049, "lr": 1.2133012483594997e-12}
{"train_loss": 0.04529508203268051, "global_step": 271431, "epoch": 3049, "lr": 1.0889462720431453e-12}
{"train_loss": 0.01927824504673481, "global_step": 271432, "epoch": 3049, "lr": 9.71313185438305e-13}
{"train_loss": 0.027506032958626747, "global_step": 271433, "epoch": 3049, "lr": 8.604019940960939e-13}
{"train_loss": 0.05498441308736801, "global_step": 271434, "epoch": 3049, "lr": 7.562126869142816e-13}
{"train_loss": 0.038320545107126236, "global_step": 271435, "epoch": 3049, "lr": 6.587452749950984e-13}
{"train_loss": 0.04011091962456703, "global_step": 271436, "epoch": 3049, "lr": 5.679997527874292e-13}
{"train_loss": 0.03824840113520622, "global_step": 271437, "epoch": 3049, "lr": 4.839761202912741e-13}
{"train_loss": 0.038458727300167084, "global_step": 271438, "epoch": 3049, "lr": 4.0667437750663284e-13}
{"train_loss": 0.041275639086961746, "global_step": 271439, "epoch": 3049, "lr": 3.360945299846208e-13}
{"train_loss": 0.05375512316823006, "global_step": 271440, "epoch": 3049, "lr": 2.7223656662300754e-13}
{"train_loss": 0.09127713739871979, "global_step": 271441, "epoch": 3049, "lr": 2.1510049852402347e-13}
{"train_loss": 0.03610950708389282, "global_step": 271442, "epoch": 3049, "lr": 1.646863201365534e-13}
{"train_loss": 0.01731666922569275, "global_step": 271443, "epoch": 3049, "lr": 1.209940314605973e-13}
{"train_loss": 0.061185359954833984, "global_step": 271444, "epoch": 3049, "lr": 8.40236324961552e-14}
{"train_loss": 0.04226218909025192, "global_step": 271445, "epoch": 3049, "lr": 5.377512324322709e-14}
{"train_loss": 0.03464478254318237, "global_step": 271446, "epoch": 3049, "lr": 3.0248509252928106e-14}
{"train_loss": 0.06864964216947556, "global_step": 271447, "epoch": 3049, "lr": 1.344377942302799e-14}
{"train_loss": 0.05418616533279419, "global_step": 271448, "epoch": 3049, "lr": 3.3609448557569977e-15}
{"train_loss": 0.0454273012969015, "global_step": 271449, "epoch": 3049, "lr": 0.0, "val_loss": 9.692619323730469}